diff --git "a/checkpoints/trainer_state.json" "b/checkpoints/trainer_state.json" new file mode 100644--- /dev/null +++ "b/checkpoints/trainer_state.json" @@ -0,0 +1,201564 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 1.0, + "eval_steps": 500, + "global_step": 33589, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 1.9841269841269841e-07, + "loss": 0.898, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 3.9682539682539683e-07, + "loss": 0.9437, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 5.952380952380952e-07, + "loss": 1.0606, + "step": 3 + }, + { + "epoch": 0.0, + "learning_rate": 7.936507936507937e-07, + "loss": 1.0237, + "step": 4 + }, + { + "epoch": 0.0, + "learning_rate": 9.92063492063492e-07, + "loss": 0.8949, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 1.1904761904761904e-06, + "loss": 0.9257, + "step": 6 + }, + { + "epoch": 0.0, + "learning_rate": 1.388888888888889e-06, + "loss": 0.9678, + "step": 7 + }, + { + "epoch": 0.0, + "learning_rate": 1.5873015873015873e-06, + "loss": 0.8161, + "step": 8 + }, + { + "epoch": 0.0, + "learning_rate": 1.7857142857142857e-06, + "loss": 0.797, + "step": 9 + }, + { + "epoch": 0.0, + "learning_rate": 1.984126984126984e-06, + "loss": 0.7925, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 2.1825396825396824e-06, + "loss": 0.9081, + "step": 11 + }, + { + "epoch": 0.0, + "learning_rate": 2.3809523809523808e-06, + "loss": 1.0106, + "step": 12 + }, + { + "epoch": 0.0, + "learning_rate": 2.5793650793650795e-06, + "loss": 0.9151, + "step": 13 + }, + { + "epoch": 0.0, + "learning_rate": 2.777777777777778e-06, + "loss": 0.9795, + "step": 14 + }, + { + "epoch": 0.0, + "learning_rate": 2.9761904761904763e-06, + "loss": 0.9787, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 3.1746031746031746e-06, + "loss": 1.0094, + "step": 16 + }, + { + "epoch": 0.0, + "learning_rate": 3.373015873015873e-06, + "loss": 0.9701, + "step": 17 + }, + { + "epoch": 0.0, + "learning_rate": 3.5714285714285714e-06, + "loss": 0.9785, + "step": 18 + }, + { + "epoch": 0.0, + "learning_rate": 3.7698412698412697e-06, + "loss": 0.8526, + "step": 19 + }, + { + "epoch": 0.0, + "learning_rate": 3.968253968253968e-06, + "loss": 1.0763, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 4.166666666666667e-06, + "loss": 0.9321, + "step": 21 + }, + { + "epoch": 0.0, + "learning_rate": 4.365079365079365e-06, + "loss": 0.9708, + "step": 22 + }, + { + "epoch": 0.0, + "learning_rate": 4.563492063492064e-06, + "loss": 0.9754, + "step": 23 + }, + { + "epoch": 0.0, + "learning_rate": 4.7619047619047615e-06, + "loss": 0.9257, + "step": 24 + }, + { + "epoch": 0.0, + "learning_rate": 4.96031746031746e-06, + "loss": 0.9066, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 5.158730158730159e-06, + "loss": 0.9949, + "step": 26 + }, + { + "epoch": 0.0, + "learning_rate": 5.357142857142857e-06, + "loss": 0.8573, + "step": 27 + }, + { + "epoch": 0.0, + "learning_rate": 5.555555555555556e-06, + "loss": 1.0619, + "step": 28 + }, + { + "epoch": 0.0, + "learning_rate": 5.753968253968254e-06, + "loss": 1.0049, + "step": 29 + }, + { + "epoch": 0.0, + "learning_rate": 5.9523809523809525e-06, + "loss": 0.9702, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 6.1507936507936505e-06, + "loss": 0.9662, + "step": 31 + }, + { + "epoch": 0.0, + "learning_rate": 6.349206349206349e-06, + "loss": 1.0182, + "step": 32 + }, + { + "epoch": 0.0, + "learning_rate": 6.547619047619048e-06, + "loss": 0.9921, + "step": 33 + }, + { + "epoch": 0.0, + "learning_rate": 6.746031746031746e-06, + "loss": 0.9373, + "step": 34 + }, + { + "epoch": 0.0, + "learning_rate": 6.944444444444445e-06, + "loss": 0.9532, + "step": 35 + }, + { + "epoch": 0.0, + "learning_rate": 7.142857142857143e-06, + "loss": 0.9544, + "step": 36 + }, + { + "epoch": 0.0, + "learning_rate": 7.3412698412698415e-06, + "loss": 0.8428, + "step": 37 + }, + { + "epoch": 0.0, + "learning_rate": 7.5396825396825394e-06, + "loss": 0.8256, + "step": 38 + }, + { + "epoch": 0.0, + "learning_rate": 7.738095238095238e-06, + "loss": 1.0346, + "step": 39 + }, + { + "epoch": 0.0, + "learning_rate": 7.936507936507936e-06, + "loss": 0.9539, + "step": 40 + }, + { + "epoch": 0.0, + "learning_rate": 8.134920634920636e-06, + "loss": 0.8626, + "step": 41 + }, + { + "epoch": 0.0, + "learning_rate": 8.333333333333334e-06, + "loss": 0.9052, + "step": 42 + }, + { + "epoch": 0.0, + "learning_rate": 8.531746031746032e-06, + "loss": 0.897, + "step": 43 + }, + { + "epoch": 0.0, + "learning_rate": 8.73015873015873e-06, + "loss": 0.9306, + "step": 44 + }, + { + "epoch": 0.0, + "learning_rate": 8.92857142857143e-06, + "loss": 0.8981, + "step": 45 + }, + { + "epoch": 0.0, + "learning_rate": 9.126984126984127e-06, + "loss": 0.9112, + "step": 46 + }, + { + "epoch": 0.0, + "learning_rate": 9.325396825396827e-06, + "loss": 0.9711, + "step": 47 + }, + { + "epoch": 0.0, + "learning_rate": 9.523809523809523e-06, + "loss": 0.9242, + "step": 48 + }, + { + "epoch": 0.0, + "learning_rate": 9.722222222222223e-06, + "loss": 0.8989, + "step": 49 + }, + { + "epoch": 0.0, + "learning_rate": 9.92063492063492e-06, + "loss": 0.977, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 1.011904761904762e-05, + "loss": 1.0136, + "step": 51 + }, + { + "epoch": 0.0, + "learning_rate": 1.0317460317460318e-05, + "loss": 0.9141, + "step": 52 + }, + { + "epoch": 0.0, + "learning_rate": 1.0515873015873016e-05, + "loss": 0.8457, + "step": 53 + }, + { + "epoch": 0.0, + "learning_rate": 1.0714285714285714e-05, + "loss": 0.869, + "step": 54 + }, + { + "epoch": 0.0, + "learning_rate": 1.0912698412698414e-05, + "loss": 0.8702, + "step": 55 + }, + { + "epoch": 0.0, + "learning_rate": 1.1111111111111112e-05, + "loss": 0.7668, + "step": 56 + }, + { + "epoch": 0.0, + "learning_rate": 1.130952380952381e-05, + "loss": 0.882, + "step": 57 + }, + { + "epoch": 0.0, + "learning_rate": 1.1507936507936508e-05, + "loss": 0.8072, + "step": 58 + }, + { + "epoch": 0.0, + "learning_rate": 1.1706349206349207e-05, + "loss": 0.8771, + "step": 59 + }, + { + "epoch": 0.0, + "learning_rate": 1.1904761904761905e-05, + "loss": 0.7272, + "step": 60 + }, + { + "epoch": 0.0, + "learning_rate": 1.2103174603174603e-05, + "loss": 0.9419, + "step": 61 + }, + { + "epoch": 0.0, + "learning_rate": 1.2301587301587301e-05, + "loss": 0.8828, + "step": 62 + }, + { + "epoch": 0.0, + "learning_rate": 1.25e-05, + "loss": 0.8769, + "step": 63 + }, + { + "epoch": 0.0, + "learning_rate": 1.2698412698412699e-05, + "loss": 0.8345, + "step": 64 + }, + { + "epoch": 0.0, + "learning_rate": 1.2896825396825398e-05, + "loss": 0.7218, + "step": 65 + }, + { + "epoch": 0.0, + "learning_rate": 1.3095238095238096e-05, + "loss": 0.854, + "step": 66 + }, + { + "epoch": 0.0, + "learning_rate": 1.3293650793650794e-05, + "loss": 0.9002, + "step": 67 + }, + { + "epoch": 0.0, + "learning_rate": 1.3492063492063492e-05, + "loss": 0.9019, + "step": 68 + }, + { + "epoch": 0.0, + "learning_rate": 1.3690476190476192e-05, + "loss": 0.835, + "step": 69 + }, + { + "epoch": 0.0, + "learning_rate": 1.388888888888889e-05, + "loss": 0.865, + "step": 70 + }, + { + "epoch": 0.0, + "learning_rate": 1.4087301587301587e-05, + "loss": 0.7525, + "step": 71 + }, + { + "epoch": 0.0, + "learning_rate": 1.4285714285714285e-05, + "loss": 0.8532, + "step": 72 + }, + { + "epoch": 0.0, + "learning_rate": 1.4484126984126987e-05, + "loss": 0.7937, + "step": 73 + }, + { + "epoch": 0.0, + "learning_rate": 1.4682539682539683e-05, + "loss": 0.8466, + "step": 74 + }, + { + "epoch": 0.0, + "learning_rate": 1.4880952380952381e-05, + "loss": 0.7435, + "step": 75 + }, + { + "epoch": 0.0, + "learning_rate": 1.5079365079365079e-05, + "loss": 0.9223, + "step": 76 + }, + { + "epoch": 0.0, + "learning_rate": 1.527777777777778e-05, + "loss": 0.811, + "step": 77 + }, + { + "epoch": 0.0, + "learning_rate": 1.5476190476190476e-05, + "loss": 0.8137, + "step": 78 + }, + { + "epoch": 0.0, + "learning_rate": 1.5674603174603176e-05, + "loss": 0.7937, + "step": 79 + }, + { + "epoch": 0.0, + "learning_rate": 1.5873015873015872e-05, + "loss": 0.786, + "step": 80 + }, + { + "epoch": 0.0, + "learning_rate": 1.6071428571428572e-05, + "loss": 0.7678, + "step": 81 + }, + { + "epoch": 0.0, + "learning_rate": 1.626984126984127e-05, + "loss": 0.7884, + "step": 82 + }, + { + "epoch": 0.0, + "learning_rate": 1.6468253968253968e-05, + "loss": 0.8645, + "step": 83 + }, + { + "epoch": 0.0, + "learning_rate": 1.6666666666666667e-05, + "loss": 0.8268, + "step": 84 + }, + { + "epoch": 0.0, + "learning_rate": 1.6865079365079367e-05, + "loss": 0.759, + "step": 85 + }, + { + "epoch": 0.0, + "learning_rate": 1.7063492063492063e-05, + "loss": 0.7425, + "step": 86 + }, + { + "epoch": 0.0, + "learning_rate": 1.7261904761904763e-05, + "loss": 0.8029, + "step": 87 + }, + { + "epoch": 0.0, + "learning_rate": 1.746031746031746e-05, + "loss": 0.7619, + "step": 88 + }, + { + "epoch": 0.0, + "learning_rate": 1.7658730158730162e-05, + "loss": 0.8521, + "step": 89 + }, + { + "epoch": 0.0, + "learning_rate": 1.785714285714286e-05, + "loss": 0.8716, + "step": 90 + }, + { + "epoch": 0.0, + "learning_rate": 1.8055555555555555e-05, + "loss": 0.8211, + "step": 91 + }, + { + "epoch": 0.0, + "learning_rate": 1.8253968253968254e-05, + "loss": 0.7918, + "step": 92 + }, + { + "epoch": 0.0, + "learning_rate": 1.8452380952380954e-05, + "loss": 0.7602, + "step": 93 + }, + { + "epoch": 0.0, + "learning_rate": 1.8650793650793654e-05, + "loss": 0.7613, + "step": 94 + }, + { + "epoch": 0.0, + "learning_rate": 1.884920634920635e-05, + "loss": 0.7957, + "step": 95 + }, + { + "epoch": 0.0, + "learning_rate": 1.9047619047619046e-05, + "loss": 0.7839, + "step": 96 + }, + { + "epoch": 0.0, + "learning_rate": 1.924603174603175e-05, + "loss": 0.7839, + "step": 97 + }, + { + "epoch": 0.0, + "learning_rate": 1.9444444444444445e-05, + "loss": 0.7617, + "step": 98 + }, + { + "epoch": 0.0, + "learning_rate": 1.9642857142857145e-05, + "loss": 0.8121, + "step": 99 + }, + { + "epoch": 0.0, + "learning_rate": 1.984126984126984e-05, + "loss": 0.6893, + "step": 100 + }, + { + "epoch": 0.0, + "learning_rate": 2.003968253968254e-05, + "loss": 0.7574, + "step": 101 + }, + { + "epoch": 0.0, + "learning_rate": 2.023809523809524e-05, + "loss": 0.7108, + "step": 102 + }, + { + "epoch": 0.0, + "learning_rate": 2.0436507936507937e-05, + "loss": 0.7104, + "step": 103 + }, + { + "epoch": 0.0, + "learning_rate": 2.0634920634920636e-05, + "loss": 0.8691, + "step": 104 + }, + { + "epoch": 0.0, + "learning_rate": 2.0833333333333336e-05, + "loss": 0.7018, + "step": 105 + }, + { + "epoch": 0.0, + "learning_rate": 2.1031746031746032e-05, + "loss": 0.7495, + "step": 106 + }, + { + "epoch": 0.0, + "learning_rate": 2.1230158730158732e-05, + "loss": 0.7199, + "step": 107 + }, + { + "epoch": 0.0, + "learning_rate": 2.1428571428571428e-05, + "loss": 0.6716, + "step": 108 + }, + { + "epoch": 0.0, + "learning_rate": 2.1626984126984128e-05, + "loss": 0.7214, + "step": 109 + }, + { + "epoch": 0.0, + "learning_rate": 2.1825396825396827e-05, + "loss": 0.7915, + "step": 110 + }, + { + "epoch": 0.0, + "learning_rate": 2.2023809523809524e-05, + "loss": 0.6599, + "step": 111 + }, + { + "epoch": 0.0, + "learning_rate": 2.2222222222222223e-05, + "loss": 0.8162, + "step": 112 + }, + { + "epoch": 0.0, + "learning_rate": 2.2420634920634923e-05, + "loss": 0.7007, + "step": 113 + }, + { + "epoch": 0.0, + "learning_rate": 2.261904761904762e-05, + "loss": 0.762, + "step": 114 + }, + { + "epoch": 0.0, + "learning_rate": 2.281746031746032e-05, + "loss": 0.816, + "step": 115 + }, + { + "epoch": 0.0, + "learning_rate": 2.3015873015873015e-05, + "loss": 0.745, + "step": 116 + }, + { + "epoch": 0.0, + "learning_rate": 2.3214285714285715e-05, + "loss": 0.6821, + "step": 117 + }, + { + "epoch": 0.0, + "learning_rate": 2.3412698412698414e-05, + "loss": 0.8315, + "step": 118 + }, + { + "epoch": 0.0, + "learning_rate": 2.361111111111111e-05, + "loss": 0.659, + "step": 119 + }, + { + "epoch": 0.0, + "learning_rate": 2.380952380952381e-05, + "loss": 0.7215, + "step": 120 + }, + { + "epoch": 0.0, + "learning_rate": 2.400793650793651e-05, + "loss": 0.7351, + "step": 121 + }, + { + "epoch": 0.0, + "learning_rate": 2.4206349206349206e-05, + "loss": 0.5598, + "step": 122 + }, + { + "epoch": 0.0, + "learning_rate": 2.4404761904761906e-05, + "loss": 0.6263, + "step": 123 + }, + { + "epoch": 0.0, + "learning_rate": 2.4603174603174602e-05, + "loss": 0.7462, + "step": 124 + }, + { + "epoch": 0.0, + "learning_rate": 2.4801587301587305e-05, + "loss": 0.7999, + "step": 125 + }, + { + "epoch": 0.0, + "learning_rate": 2.5e-05, + "loss": 0.7186, + "step": 126 + }, + { + "epoch": 0.0, + "learning_rate": 2.5198412698412697e-05, + "loss": 0.7827, + "step": 127 + }, + { + "epoch": 0.0, + "learning_rate": 2.5396825396825397e-05, + "loss": 0.6996, + "step": 128 + }, + { + "epoch": 0.0, + "learning_rate": 2.5595238095238093e-05, + "loss": 0.6132, + "step": 129 + }, + { + "epoch": 0.0, + "learning_rate": 2.5793650793650796e-05, + "loss": 0.6324, + "step": 130 + }, + { + "epoch": 0.0, + "learning_rate": 2.5992063492063496e-05, + "loss": 0.6403, + "step": 131 + }, + { + "epoch": 0.0, + "learning_rate": 2.6190476190476192e-05, + "loss": 0.6824, + "step": 132 + }, + { + "epoch": 0.0, + "learning_rate": 2.6388888888888892e-05, + "loss": 0.7067, + "step": 133 + }, + { + "epoch": 0.0, + "learning_rate": 2.6587301587301588e-05, + "loss": 0.644, + "step": 134 + }, + { + "epoch": 0.0, + "learning_rate": 2.6785714285714288e-05, + "loss": 0.691, + "step": 135 + }, + { + "epoch": 0.0, + "learning_rate": 2.6984126984126984e-05, + "loss": 0.7106, + "step": 136 + }, + { + "epoch": 0.0, + "learning_rate": 2.718253968253968e-05, + "loss": 0.735, + "step": 137 + }, + { + "epoch": 0.0, + "learning_rate": 2.7380952380952383e-05, + "loss": 0.6893, + "step": 138 + }, + { + "epoch": 0.0, + "learning_rate": 2.7579365079365083e-05, + "loss": 0.6308, + "step": 139 + }, + { + "epoch": 0.0, + "learning_rate": 2.777777777777778e-05, + "loss": 0.6297, + "step": 140 + }, + { + "epoch": 0.0, + "learning_rate": 2.797619047619048e-05, + "loss": 0.6301, + "step": 141 + }, + { + "epoch": 0.0, + "learning_rate": 2.8174603174603175e-05, + "loss": 0.7773, + "step": 142 + }, + { + "epoch": 0.0, + "learning_rate": 2.8373015873015875e-05, + "loss": 0.5519, + "step": 143 + }, + { + "epoch": 0.0, + "learning_rate": 2.857142857142857e-05, + "loss": 0.6365, + "step": 144 + }, + { + "epoch": 0.0, + "learning_rate": 2.876984126984127e-05, + "loss": 0.695, + "step": 145 + }, + { + "epoch": 0.0, + "learning_rate": 2.8968253968253974e-05, + "loss": 0.7744, + "step": 146 + }, + { + "epoch": 0.0, + "learning_rate": 2.916666666666667e-05, + "loss": 0.6444, + "step": 147 + }, + { + "epoch": 0.0, + "learning_rate": 2.9365079365079366e-05, + "loss": 0.7552, + "step": 148 + }, + { + "epoch": 0.0, + "learning_rate": 2.9563492063492066e-05, + "loss": 0.6701, + "step": 149 + }, + { + "epoch": 0.0, + "learning_rate": 2.9761904761904762e-05, + "loss": 0.7936, + "step": 150 + }, + { + "epoch": 0.0, + "learning_rate": 2.996031746031746e-05, + "loss": 0.6361, + "step": 151 + }, + { + "epoch": 0.0, + "learning_rate": 3.0158730158730158e-05, + "loss": 0.663, + "step": 152 + }, + { + "epoch": 0.0, + "learning_rate": 3.0357142857142857e-05, + "loss": 0.6349, + "step": 153 + }, + { + "epoch": 0.0, + "learning_rate": 3.055555555555556e-05, + "loss": 0.6423, + "step": 154 + }, + { + "epoch": 0.0, + "learning_rate": 3.075396825396826e-05, + "loss": 0.6308, + "step": 155 + }, + { + "epoch": 0.0, + "learning_rate": 3.095238095238095e-05, + "loss": 0.6855, + "step": 156 + }, + { + "epoch": 0.0, + "learning_rate": 3.1150793650793656e-05, + "loss": 0.5921, + "step": 157 + }, + { + "epoch": 0.0, + "learning_rate": 3.134920634920635e-05, + "loss": 0.807, + "step": 158 + }, + { + "epoch": 0.0, + "learning_rate": 3.154761904761905e-05, + "loss": 0.6349, + "step": 159 + }, + { + "epoch": 0.0, + "learning_rate": 3.1746031746031745e-05, + "loss": 0.6952, + "step": 160 + }, + { + "epoch": 0.0, + "learning_rate": 3.194444444444444e-05, + "loss": 0.7238, + "step": 161 + }, + { + "epoch": 0.0, + "learning_rate": 3.2142857142857144e-05, + "loss": 0.7016, + "step": 162 + }, + { + "epoch": 0.0, + "learning_rate": 3.234126984126985e-05, + "loss": 0.6988, + "step": 163 + }, + { + "epoch": 0.0, + "learning_rate": 3.253968253968254e-05, + "loss": 0.6731, + "step": 164 + }, + { + "epoch": 0.0, + "learning_rate": 3.273809523809524e-05, + "loss": 0.719, + "step": 165 + }, + { + "epoch": 0.0, + "learning_rate": 3.2936507936507936e-05, + "loss": 0.7759, + "step": 166 + }, + { + "epoch": 0.0, + "learning_rate": 3.313492063492064e-05, + "loss": 0.6453, + "step": 167 + }, + { + "epoch": 0.01, + "learning_rate": 3.3333333333333335e-05, + "loss": 0.5486, + "step": 168 + }, + { + "epoch": 0.01, + "learning_rate": 3.353174603174603e-05, + "loss": 0.6977, + "step": 169 + }, + { + "epoch": 0.01, + "learning_rate": 3.3730158730158734e-05, + "loss": 0.7584, + "step": 170 + }, + { + "epoch": 0.01, + "learning_rate": 3.392857142857143e-05, + "loss": 0.7276, + "step": 171 + }, + { + "epoch": 0.01, + "learning_rate": 3.412698412698413e-05, + "loss": 0.6578, + "step": 172 + }, + { + "epoch": 0.01, + "learning_rate": 3.432539682539683e-05, + "loss": 0.6432, + "step": 173 + }, + { + "epoch": 0.01, + "learning_rate": 3.4523809523809526e-05, + "loss": 0.6699, + "step": 174 + }, + { + "epoch": 0.01, + "learning_rate": 3.472222222222222e-05, + "loss": 0.8727, + "step": 175 + }, + { + "epoch": 0.01, + "learning_rate": 3.492063492063492e-05, + "loss": 0.6616, + "step": 176 + }, + { + "epoch": 0.01, + "learning_rate": 3.511904761904762e-05, + "loss": 0.7029, + "step": 177 + }, + { + "epoch": 0.01, + "learning_rate": 3.5317460317460324e-05, + "loss": 0.7055, + "step": 178 + }, + { + "epoch": 0.01, + "learning_rate": 3.551587301587302e-05, + "loss": 0.6402, + "step": 179 + }, + { + "epoch": 0.01, + "learning_rate": 3.571428571428572e-05, + "loss": 0.6244, + "step": 180 + }, + { + "epoch": 0.01, + "learning_rate": 3.591269841269841e-05, + "loss": 0.679, + "step": 181 + }, + { + "epoch": 0.01, + "learning_rate": 3.611111111111111e-05, + "loss": 0.7162, + "step": 182 + }, + { + "epoch": 0.01, + "learning_rate": 3.630952380952381e-05, + "loss": 0.6929, + "step": 183 + }, + { + "epoch": 0.01, + "learning_rate": 3.650793650793651e-05, + "loss": 0.7311, + "step": 184 + }, + { + "epoch": 0.01, + "learning_rate": 3.6706349206349205e-05, + "loss": 0.6766, + "step": 185 + }, + { + "epoch": 0.01, + "learning_rate": 3.690476190476191e-05, + "loss": 0.7436, + "step": 186 + }, + { + "epoch": 0.01, + "learning_rate": 3.7103174603174604e-05, + "loss": 0.7054, + "step": 187 + }, + { + "epoch": 0.01, + "learning_rate": 3.730158730158731e-05, + "loss": 0.7039, + "step": 188 + }, + { + "epoch": 0.01, + "learning_rate": 3.7500000000000003e-05, + "loss": 0.5641, + "step": 189 + }, + { + "epoch": 0.01, + "learning_rate": 3.76984126984127e-05, + "loss": 0.691, + "step": 190 + }, + { + "epoch": 0.01, + "learning_rate": 3.7896825396825396e-05, + "loss": 0.6515, + "step": 191 + }, + { + "epoch": 0.01, + "learning_rate": 3.809523809523809e-05, + "loss": 0.6133, + "step": 192 + }, + { + "epoch": 0.01, + "learning_rate": 3.8293650793650795e-05, + "loss": 0.6521, + "step": 193 + }, + { + "epoch": 0.01, + "learning_rate": 3.84920634920635e-05, + "loss": 0.6345, + "step": 194 + }, + { + "epoch": 0.01, + "learning_rate": 3.8690476190476195e-05, + "loss": 0.6908, + "step": 195 + }, + { + "epoch": 0.01, + "learning_rate": 3.888888888888889e-05, + "loss": 0.6731, + "step": 196 + }, + { + "epoch": 0.01, + "learning_rate": 3.908730158730159e-05, + "loss": 0.5897, + "step": 197 + }, + { + "epoch": 0.01, + "learning_rate": 3.928571428571429e-05, + "loss": 0.7487, + "step": 198 + }, + { + "epoch": 0.01, + "learning_rate": 3.9484126984126986e-05, + "loss": 0.602, + "step": 199 + }, + { + "epoch": 0.01, + "learning_rate": 3.968253968253968e-05, + "loss": 0.6424, + "step": 200 + }, + { + "epoch": 0.01, + "learning_rate": 3.9880952380952386e-05, + "loss": 0.5364, + "step": 201 + }, + { + "epoch": 0.01, + "learning_rate": 4.007936507936508e-05, + "loss": 0.6414, + "step": 202 + }, + { + "epoch": 0.01, + "learning_rate": 4.027777777777778e-05, + "loss": 0.7005, + "step": 203 + }, + { + "epoch": 0.01, + "learning_rate": 4.047619047619048e-05, + "loss": 0.654, + "step": 204 + }, + { + "epoch": 0.01, + "learning_rate": 4.067460317460318e-05, + "loss": 0.6414, + "step": 205 + }, + { + "epoch": 0.01, + "learning_rate": 4.0873015873015874e-05, + "loss": 0.7245, + "step": 206 + }, + { + "epoch": 0.01, + "learning_rate": 4.107142857142857e-05, + "loss": 0.6804, + "step": 207 + }, + { + "epoch": 0.01, + "learning_rate": 4.126984126984127e-05, + "loss": 0.6141, + "step": 208 + }, + { + "epoch": 0.01, + "learning_rate": 4.1468253968253976e-05, + "loss": 0.5919, + "step": 209 + }, + { + "epoch": 0.01, + "learning_rate": 4.166666666666667e-05, + "loss": 0.7872, + "step": 210 + }, + { + "epoch": 0.01, + "learning_rate": 4.186507936507937e-05, + "loss": 0.7313, + "step": 211 + }, + { + "epoch": 0.01, + "learning_rate": 4.2063492063492065e-05, + "loss": 0.6479, + "step": 212 + }, + { + "epoch": 0.01, + "learning_rate": 4.226190476190476e-05, + "loss": 0.5105, + "step": 213 + }, + { + "epoch": 0.01, + "learning_rate": 4.2460317460317464e-05, + "loss": 0.5469, + "step": 214 + }, + { + "epoch": 0.01, + "learning_rate": 4.265873015873016e-05, + "loss": 0.6445, + "step": 215 + }, + { + "epoch": 0.01, + "learning_rate": 4.2857142857142856e-05, + "loss": 0.7022, + "step": 216 + }, + { + "epoch": 0.01, + "learning_rate": 4.305555555555556e-05, + "loss": 0.7241, + "step": 217 + }, + { + "epoch": 0.01, + "learning_rate": 4.3253968253968256e-05, + "loss": 0.6216, + "step": 218 + }, + { + "epoch": 0.01, + "learning_rate": 4.345238095238096e-05, + "loss": 0.6699, + "step": 219 + }, + { + "epoch": 0.01, + "learning_rate": 4.3650793650793655e-05, + "loss": 0.5787, + "step": 220 + }, + { + "epoch": 0.01, + "learning_rate": 4.384920634920635e-05, + "loss": 0.6313, + "step": 221 + }, + { + "epoch": 0.01, + "learning_rate": 4.404761904761905e-05, + "loss": 0.7016, + "step": 222 + }, + { + "epoch": 0.01, + "learning_rate": 4.4246031746031744e-05, + "loss": 0.5827, + "step": 223 + }, + { + "epoch": 0.01, + "learning_rate": 4.4444444444444447e-05, + "loss": 0.5799, + "step": 224 + }, + { + "epoch": 0.01, + "learning_rate": 4.464285714285715e-05, + "loss": 0.751, + "step": 225 + }, + { + "epoch": 0.01, + "learning_rate": 4.4841269841269846e-05, + "loss": 0.6446, + "step": 226 + }, + { + "epoch": 0.01, + "learning_rate": 4.503968253968254e-05, + "loss": 0.7477, + "step": 227 + }, + { + "epoch": 0.01, + "learning_rate": 4.523809523809524e-05, + "loss": 0.7189, + "step": 228 + }, + { + "epoch": 0.01, + "learning_rate": 4.543650793650794e-05, + "loss": 0.7043, + "step": 229 + }, + { + "epoch": 0.01, + "learning_rate": 4.563492063492064e-05, + "loss": 0.7033, + "step": 230 + }, + { + "epoch": 0.01, + "learning_rate": 4.5833333333333334e-05, + "loss": 0.6254, + "step": 231 + }, + { + "epoch": 0.01, + "learning_rate": 4.603174603174603e-05, + "loss": 0.7004, + "step": 232 + }, + { + "epoch": 0.01, + "learning_rate": 4.623015873015873e-05, + "loss": 0.6479, + "step": 233 + }, + { + "epoch": 0.01, + "learning_rate": 4.642857142857143e-05, + "loss": 0.6501, + "step": 234 + }, + { + "epoch": 0.01, + "learning_rate": 4.662698412698413e-05, + "loss": 0.6979, + "step": 235 + }, + { + "epoch": 0.01, + "learning_rate": 4.682539682539683e-05, + "loss": 0.7345, + "step": 236 + }, + { + "epoch": 0.01, + "learning_rate": 4.7023809523809525e-05, + "loss": 0.632, + "step": 237 + }, + { + "epoch": 0.01, + "learning_rate": 4.722222222222222e-05, + "loss": 0.722, + "step": 238 + }, + { + "epoch": 0.01, + "learning_rate": 4.7420634920634924e-05, + "loss": 0.7307, + "step": 239 + }, + { + "epoch": 0.01, + "learning_rate": 4.761904761904762e-05, + "loss": 0.6131, + "step": 240 + }, + { + "epoch": 0.01, + "learning_rate": 4.781746031746032e-05, + "loss": 0.5927, + "step": 241 + }, + { + "epoch": 0.01, + "learning_rate": 4.801587301587302e-05, + "loss": 0.6395, + "step": 242 + }, + { + "epoch": 0.01, + "learning_rate": 4.8214285714285716e-05, + "loss": 0.7283, + "step": 243 + }, + { + "epoch": 0.01, + "learning_rate": 4.841269841269841e-05, + "loss": 0.5445, + "step": 244 + }, + { + "epoch": 0.01, + "learning_rate": 4.8611111111111115e-05, + "loss": 0.5858, + "step": 245 + }, + { + "epoch": 0.01, + "learning_rate": 4.880952380952381e-05, + "loss": 0.6013, + "step": 246 + }, + { + "epoch": 0.01, + "learning_rate": 4.900793650793651e-05, + "loss": 0.5912, + "step": 247 + }, + { + "epoch": 0.01, + "learning_rate": 4.9206349206349204e-05, + "loss": 0.636, + "step": 248 + }, + { + "epoch": 0.01, + "learning_rate": 4.940476190476191e-05, + "loss": 0.7017, + "step": 249 + }, + { + "epoch": 0.01, + "learning_rate": 4.960317460317461e-05, + "loss": 0.6491, + "step": 250 + }, + { + "epoch": 0.01, + "learning_rate": 4.9801587301587306e-05, + "loss": 0.7415, + "step": 251 + }, + { + "epoch": 0.01, + "learning_rate": 5e-05, + "loss": 0.5958, + "step": 252 + }, + { + "epoch": 0.01, + "learning_rate": 5.01984126984127e-05, + "loss": 0.6192, + "step": 253 + }, + { + "epoch": 0.01, + "learning_rate": 5.0396825396825395e-05, + "loss": 0.6936, + "step": 254 + }, + { + "epoch": 0.01, + "learning_rate": 5.05952380952381e-05, + "loss": 0.5967, + "step": 255 + }, + { + "epoch": 0.01, + "learning_rate": 5.0793650793650794e-05, + "loss": 0.5144, + "step": 256 + }, + { + "epoch": 0.01, + "learning_rate": 5.099206349206349e-05, + "loss": 0.6287, + "step": 257 + }, + { + "epoch": 0.01, + "learning_rate": 5.119047619047619e-05, + "loss": 0.5911, + "step": 258 + }, + { + "epoch": 0.01, + "learning_rate": 5.138888888888889e-05, + "loss": 0.4781, + "step": 259 + }, + { + "epoch": 0.01, + "learning_rate": 5.158730158730159e-05, + "loss": 0.607, + "step": 260 + }, + { + "epoch": 0.01, + "learning_rate": 5.1785714285714296e-05, + "loss": 0.6945, + "step": 261 + }, + { + "epoch": 0.01, + "learning_rate": 5.198412698412699e-05, + "loss": 0.7408, + "step": 262 + }, + { + "epoch": 0.01, + "learning_rate": 5.218253968253969e-05, + "loss": 0.5664, + "step": 263 + }, + { + "epoch": 0.01, + "learning_rate": 5.2380952380952384e-05, + "loss": 0.5773, + "step": 264 + }, + { + "epoch": 0.01, + "learning_rate": 5.257936507936508e-05, + "loss": 0.5704, + "step": 265 + }, + { + "epoch": 0.01, + "learning_rate": 5.2777777777777784e-05, + "loss": 0.5538, + "step": 266 + }, + { + "epoch": 0.01, + "learning_rate": 5.297619047619048e-05, + "loss": 0.6289, + "step": 267 + }, + { + "epoch": 0.01, + "learning_rate": 5.3174603174603176e-05, + "loss": 0.6082, + "step": 268 + }, + { + "epoch": 0.01, + "learning_rate": 5.337301587301587e-05, + "loss": 0.6555, + "step": 269 + }, + { + "epoch": 0.01, + "learning_rate": 5.3571428571428575e-05, + "loss": 0.5427, + "step": 270 + }, + { + "epoch": 0.01, + "learning_rate": 5.376984126984127e-05, + "loss": 0.5631, + "step": 271 + }, + { + "epoch": 0.01, + "learning_rate": 5.396825396825397e-05, + "loss": 0.5594, + "step": 272 + }, + { + "epoch": 0.01, + "learning_rate": 5.4166666666666664e-05, + "loss": 0.6782, + "step": 273 + }, + { + "epoch": 0.01, + "learning_rate": 5.436507936507936e-05, + "loss": 0.6543, + "step": 274 + }, + { + "epoch": 0.01, + "learning_rate": 5.4563492063492063e-05, + "loss": 0.6396, + "step": 275 + }, + { + "epoch": 0.01, + "learning_rate": 5.4761904761904766e-05, + "loss": 0.6258, + "step": 276 + }, + { + "epoch": 0.01, + "learning_rate": 5.496031746031747e-05, + "loss": 0.5311, + "step": 277 + }, + { + "epoch": 0.01, + "learning_rate": 5.5158730158730166e-05, + "loss": 0.5128, + "step": 278 + }, + { + "epoch": 0.01, + "learning_rate": 5.535714285714286e-05, + "loss": 0.5872, + "step": 279 + }, + { + "epoch": 0.01, + "learning_rate": 5.555555555555556e-05, + "loss": 0.6763, + "step": 280 + }, + { + "epoch": 0.01, + "learning_rate": 5.575396825396826e-05, + "loss": 0.6229, + "step": 281 + }, + { + "epoch": 0.01, + "learning_rate": 5.595238095238096e-05, + "loss": 0.6068, + "step": 282 + }, + { + "epoch": 0.01, + "learning_rate": 5.6150793650793654e-05, + "loss": 0.683, + "step": 283 + }, + { + "epoch": 0.01, + "learning_rate": 5.634920634920635e-05, + "loss": 0.7285, + "step": 284 + }, + { + "epoch": 0.01, + "learning_rate": 5.6547619047619046e-05, + "loss": 0.6153, + "step": 285 + }, + { + "epoch": 0.01, + "learning_rate": 5.674603174603175e-05, + "loss": 0.6377, + "step": 286 + }, + { + "epoch": 0.01, + "learning_rate": 5.6944444444444445e-05, + "loss": 0.6532, + "step": 287 + }, + { + "epoch": 0.01, + "learning_rate": 5.714285714285714e-05, + "loss": 0.6446, + "step": 288 + }, + { + "epoch": 0.01, + "learning_rate": 5.734126984126984e-05, + "loss": 0.7136, + "step": 289 + }, + { + "epoch": 0.01, + "learning_rate": 5.753968253968254e-05, + "loss": 0.5714, + "step": 290 + }, + { + "epoch": 0.01, + "learning_rate": 5.773809523809524e-05, + "loss": 0.7389, + "step": 291 + }, + { + "epoch": 0.01, + "learning_rate": 5.793650793650795e-05, + "loss": 0.5901, + "step": 292 + }, + { + "epoch": 0.01, + "learning_rate": 5.813492063492064e-05, + "loss": 0.7087, + "step": 293 + }, + { + "epoch": 0.01, + "learning_rate": 5.833333333333334e-05, + "loss": 0.6658, + "step": 294 + }, + { + "epoch": 0.01, + "learning_rate": 5.8531746031746036e-05, + "loss": 0.5931, + "step": 295 + }, + { + "epoch": 0.01, + "learning_rate": 5.873015873015873e-05, + "loss": 0.6193, + "step": 296 + }, + { + "epoch": 0.01, + "learning_rate": 5.8928571428571435e-05, + "loss": 0.7073, + "step": 297 + }, + { + "epoch": 0.01, + "learning_rate": 5.912698412698413e-05, + "loss": 0.5922, + "step": 298 + }, + { + "epoch": 0.01, + "learning_rate": 5.932539682539683e-05, + "loss": 0.6709, + "step": 299 + }, + { + "epoch": 0.01, + "learning_rate": 5.9523809523809524e-05, + "loss": 0.5687, + "step": 300 + }, + { + "epoch": 0.01, + "learning_rate": 5.972222222222223e-05, + "loss": 0.5526, + "step": 301 + }, + { + "epoch": 0.01, + "learning_rate": 5.992063492063492e-05, + "loss": 0.6134, + "step": 302 + }, + { + "epoch": 0.01, + "learning_rate": 6.011904761904762e-05, + "loss": 0.734, + "step": 303 + }, + { + "epoch": 0.01, + "learning_rate": 6.0317460317460316e-05, + "loss": 0.6823, + "step": 304 + }, + { + "epoch": 0.01, + "learning_rate": 6.051587301587301e-05, + "loss": 0.717, + "step": 305 + }, + { + "epoch": 0.01, + "learning_rate": 6.0714285714285715e-05, + "loss": 0.6155, + "step": 306 + }, + { + "epoch": 0.01, + "learning_rate": 6.091269841269841e-05, + "loss": 0.605, + "step": 307 + }, + { + "epoch": 0.01, + "learning_rate": 6.111111111111112e-05, + "loss": 0.6486, + "step": 308 + }, + { + "epoch": 0.01, + "learning_rate": 6.130952380952381e-05, + "loss": 0.6647, + "step": 309 + }, + { + "epoch": 0.01, + "learning_rate": 6.150793650793651e-05, + "loss": 0.5956, + "step": 310 + }, + { + "epoch": 0.01, + "learning_rate": 6.170634920634922e-05, + "loss": 0.63, + "step": 311 + }, + { + "epoch": 0.01, + "learning_rate": 6.19047619047619e-05, + "loss": 0.6477, + "step": 312 + }, + { + "epoch": 0.01, + "learning_rate": 6.210317460317461e-05, + "loss": 0.5366, + "step": 313 + }, + { + "epoch": 0.01, + "learning_rate": 6.230158730158731e-05, + "loss": 0.6072, + "step": 314 + }, + { + "epoch": 0.01, + "learning_rate": 6.25e-05, + "loss": 0.5827, + "step": 315 + }, + { + "epoch": 0.01, + "learning_rate": 6.26984126984127e-05, + "loss": 0.6751, + "step": 316 + }, + { + "epoch": 0.01, + "learning_rate": 6.28968253968254e-05, + "loss": 0.5821, + "step": 317 + }, + { + "epoch": 0.01, + "learning_rate": 6.30952380952381e-05, + "loss": 0.6176, + "step": 318 + }, + { + "epoch": 0.01, + "learning_rate": 6.32936507936508e-05, + "loss": 0.5533, + "step": 319 + }, + { + "epoch": 0.01, + "learning_rate": 6.349206349206349e-05, + "loss": 0.6208, + "step": 320 + }, + { + "epoch": 0.01, + "learning_rate": 6.369047619047619e-05, + "loss": 0.7685, + "step": 321 + }, + { + "epoch": 0.01, + "learning_rate": 6.388888888888888e-05, + "loss": 0.5346, + "step": 322 + }, + { + "epoch": 0.01, + "learning_rate": 6.40873015873016e-05, + "loss": 0.6414, + "step": 323 + }, + { + "epoch": 0.01, + "learning_rate": 6.428571428571429e-05, + "loss": 0.6823, + "step": 324 + }, + { + "epoch": 0.01, + "learning_rate": 6.448412698412699e-05, + "loss": 0.6111, + "step": 325 + }, + { + "epoch": 0.01, + "learning_rate": 6.46825396825397e-05, + "loss": 0.6957, + "step": 326 + }, + { + "epoch": 0.01, + "learning_rate": 6.488095238095238e-05, + "loss": 0.5557, + "step": 327 + }, + { + "epoch": 0.01, + "learning_rate": 6.507936507936509e-05, + "loss": 0.6467, + "step": 328 + }, + { + "epoch": 0.01, + "learning_rate": 6.527777777777778e-05, + "loss": 0.6431, + "step": 329 + }, + { + "epoch": 0.01, + "learning_rate": 6.547619047619048e-05, + "loss": 0.6142, + "step": 330 + }, + { + "epoch": 0.01, + "learning_rate": 6.567460317460318e-05, + "loss": 0.6268, + "step": 331 + }, + { + "epoch": 0.01, + "learning_rate": 6.587301587301587e-05, + "loss": 0.6144, + "step": 332 + }, + { + "epoch": 0.01, + "learning_rate": 6.607142857142857e-05, + "loss": 0.6336, + "step": 333 + }, + { + "epoch": 0.01, + "learning_rate": 6.626984126984128e-05, + "loss": 0.5891, + "step": 334 + }, + { + "epoch": 0.01, + "learning_rate": 6.646825396825397e-05, + "loss": 0.5852, + "step": 335 + }, + { + "epoch": 0.01, + "learning_rate": 6.666666666666667e-05, + "loss": 0.6482, + "step": 336 + }, + { + "epoch": 0.01, + "learning_rate": 6.686507936507936e-05, + "loss": 0.5843, + "step": 337 + }, + { + "epoch": 0.01, + "learning_rate": 6.706349206349206e-05, + "loss": 0.6715, + "step": 338 + }, + { + "epoch": 0.01, + "learning_rate": 6.726190476190477e-05, + "loss": 0.6662, + "step": 339 + }, + { + "epoch": 0.01, + "learning_rate": 6.746031746031747e-05, + "loss": 0.7197, + "step": 340 + }, + { + "epoch": 0.01, + "learning_rate": 6.765873015873017e-05, + "loss": 0.704, + "step": 341 + }, + { + "epoch": 0.01, + "learning_rate": 6.785714285714286e-05, + "loss": 0.6357, + "step": 342 + }, + { + "epoch": 0.01, + "learning_rate": 6.805555555555556e-05, + "loss": 0.6048, + "step": 343 + }, + { + "epoch": 0.01, + "learning_rate": 6.825396825396825e-05, + "loss": 0.5315, + "step": 344 + }, + { + "epoch": 0.01, + "learning_rate": 6.845238095238096e-05, + "loss": 0.6498, + "step": 345 + }, + { + "epoch": 0.01, + "learning_rate": 6.865079365079366e-05, + "loss": 0.7215, + "step": 346 + }, + { + "epoch": 0.01, + "learning_rate": 6.884920634920635e-05, + "loss": 0.701, + "step": 347 + }, + { + "epoch": 0.01, + "learning_rate": 6.904761904761905e-05, + "loss": 0.6267, + "step": 348 + }, + { + "epoch": 0.01, + "learning_rate": 6.924603174603174e-05, + "loss": 0.5949, + "step": 349 + }, + { + "epoch": 0.01, + "learning_rate": 6.944444444444444e-05, + "loss": 0.6427, + "step": 350 + }, + { + "epoch": 0.01, + "learning_rate": 6.964285714285715e-05, + "loss": 0.5603, + "step": 351 + }, + { + "epoch": 0.01, + "learning_rate": 6.984126984126984e-05, + "loss": 0.5888, + "step": 352 + }, + { + "epoch": 0.01, + "learning_rate": 7.003968253968254e-05, + "loss": 0.6361, + "step": 353 + }, + { + "epoch": 0.01, + "learning_rate": 7.023809523809524e-05, + "loss": 0.5912, + "step": 354 + }, + { + "epoch": 0.01, + "learning_rate": 7.043650793650795e-05, + "loss": 0.8426, + "step": 355 + }, + { + "epoch": 0.01, + "learning_rate": 7.063492063492065e-05, + "loss": 0.5162, + "step": 356 + }, + { + "epoch": 0.01, + "learning_rate": 7.083333333333334e-05, + "loss": 0.6435, + "step": 357 + }, + { + "epoch": 0.01, + "learning_rate": 7.103174603174604e-05, + "loss": 0.6104, + "step": 358 + }, + { + "epoch": 0.01, + "learning_rate": 7.123015873015873e-05, + "loss": 0.6292, + "step": 359 + }, + { + "epoch": 0.01, + "learning_rate": 7.142857142857143e-05, + "loss": 0.5804, + "step": 360 + }, + { + "epoch": 0.01, + "learning_rate": 7.162698412698414e-05, + "loss": 0.6802, + "step": 361 + }, + { + "epoch": 0.01, + "learning_rate": 7.182539682539683e-05, + "loss": 0.5847, + "step": 362 + }, + { + "epoch": 0.01, + "learning_rate": 7.202380952380953e-05, + "loss": 0.5685, + "step": 363 + }, + { + "epoch": 0.01, + "learning_rate": 7.222222222222222e-05, + "loss": 0.6111, + "step": 364 + }, + { + "epoch": 0.01, + "learning_rate": 7.242063492063492e-05, + "loss": 0.4893, + "step": 365 + }, + { + "epoch": 0.01, + "learning_rate": 7.261904761904762e-05, + "loss": 0.5743, + "step": 366 + }, + { + "epoch": 0.01, + "learning_rate": 7.281746031746031e-05, + "loss": 0.6204, + "step": 367 + }, + { + "epoch": 0.01, + "learning_rate": 7.301587301587302e-05, + "loss": 0.5589, + "step": 368 + }, + { + "epoch": 0.01, + "learning_rate": 7.321428571428571e-05, + "loss": 0.5907, + "step": 369 + }, + { + "epoch": 0.01, + "learning_rate": 7.341269841269841e-05, + "loss": 0.5845, + "step": 370 + }, + { + "epoch": 0.01, + "learning_rate": 7.361111111111111e-05, + "loss": 0.5835, + "step": 371 + }, + { + "epoch": 0.01, + "learning_rate": 7.380952380952382e-05, + "loss": 0.5294, + "step": 372 + }, + { + "epoch": 0.01, + "learning_rate": 7.400793650793652e-05, + "loss": 0.7188, + "step": 373 + }, + { + "epoch": 0.01, + "learning_rate": 7.420634920634921e-05, + "loss": 0.5954, + "step": 374 + }, + { + "epoch": 0.01, + "learning_rate": 7.440476190476191e-05, + "loss": 0.6494, + "step": 375 + }, + { + "epoch": 0.01, + "learning_rate": 7.460317460317461e-05, + "loss": 0.6593, + "step": 376 + }, + { + "epoch": 0.01, + "learning_rate": 7.48015873015873e-05, + "loss": 0.653, + "step": 377 + }, + { + "epoch": 0.01, + "learning_rate": 7.500000000000001e-05, + "loss": 0.6301, + "step": 378 + }, + { + "epoch": 0.01, + "learning_rate": 7.51984126984127e-05, + "loss": 0.5263, + "step": 379 + }, + { + "epoch": 0.01, + "learning_rate": 7.53968253968254e-05, + "loss": 0.5814, + "step": 380 + }, + { + "epoch": 0.01, + "learning_rate": 7.55952380952381e-05, + "loss": 0.6907, + "step": 381 + }, + { + "epoch": 0.01, + "learning_rate": 7.579365079365079e-05, + "loss": 0.5889, + "step": 382 + }, + { + "epoch": 0.01, + "learning_rate": 7.59920634920635e-05, + "loss": 0.5535, + "step": 383 + }, + { + "epoch": 0.01, + "learning_rate": 7.619047619047618e-05, + "loss": 0.6338, + "step": 384 + }, + { + "epoch": 0.01, + "learning_rate": 7.638888888888889e-05, + "loss": 0.5875, + "step": 385 + }, + { + "epoch": 0.01, + "learning_rate": 7.658730158730159e-05, + "loss": 0.6428, + "step": 386 + }, + { + "epoch": 0.01, + "learning_rate": 7.67857142857143e-05, + "loss": 0.4404, + "step": 387 + }, + { + "epoch": 0.01, + "learning_rate": 7.6984126984127e-05, + "loss": 0.5535, + "step": 388 + }, + { + "epoch": 0.01, + "learning_rate": 7.718253968253969e-05, + "loss": 0.657, + "step": 389 + }, + { + "epoch": 0.01, + "learning_rate": 7.738095238095239e-05, + "loss": 0.6174, + "step": 390 + }, + { + "epoch": 0.01, + "learning_rate": 7.757936507936508e-05, + "loss": 0.5151, + "step": 391 + }, + { + "epoch": 0.01, + "learning_rate": 7.777777777777778e-05, + "loss": 0.6695, + "step": 392 + }, + { + "epoch": 0.01, + "learning_rate": 7.797619047619048e-05, + "loss": 0.6018, + "step": 393 + }, + { + "epoch": 0.01, + "learning_rate": 7.817460317460317e-05, + "loss": 0.5893, + "step": 394 + }, + { + "epoch": 0.01, + "learning_rate": 7.837301587301588e-05, + "loss": 0.5774, + "step": 395 + }, + { + "epoch": 0.01, + "learning_rate": 7.857142857142858e-05, + "loss": 0.5471, + "step": 396 + }, + { + "epoch": 0.01, + "learning_rate": 7.876984126984127e-05, + "loss": 0.6622, + "step": 397 + }, + { + "epoch": 0.01, + "learning_rate": 7.896825396825397e-05, + "loss": 0.6808, + "step": 398 + }, + { + "epoch": 0.01, + "learning_rate": 7.916666666666666e-05, + "loss": 0.5237, + "step": 399 + }, + { + "epoch": 0.01, + "learning_rate": 7.936507936507937e-05, + "loss": 0.6218, + "step": 400 + }, + { + "epoch": 0.01, + "learning_rate": 7.956349206349207e-05, + "loss": 0.6347, + "step": 401 + }, + { + "epoch": 0.01, + "learning_rate": 7.976190476190477e-05, + "loss": 0.5972, + "step": 402 + }, + { + "epoch": 0.01, + "learning_rate": 7.996031746031747e-05, + "loss": 0.6526, + "step": 403 + }, + { + "epoch": 0.01, + "learning_rate": 8.015873015873016e-05, + "loss": 0.5943, + "step": 404 + }, + { + "epoch": 0.01, + "learning_rate": 8.035714285714287e-05, + "loss": 0.606, + "step": 405 + }, + { + "epoch": 0.01, + "learning_rate": 8.055555555555556e-05, + "loss": 0.6506, + "step": 406 + }, + { + "epoch": 0.01, + "learning_rate": 8.075396825396826e-05, + "loss": 0.5291, + "step": 407 + }, + { + "epoch": 0.01, + "learning_rate": 8.095238095238096e-05, + "loss": 0.6399, + "step": 408 + }, + { + "epoch": 0.01, + "learning_rate": 8.115079365079365e-05, + "loss": 0.5421, + "step": 409 + }, + { + "epoch": 0.01, + "learning_rate": 8.134920634920635e-05, + "loss": 0.626, + "step": 410 + }, + { + "epoch": 0.01, + "learning_rate": 8.154761904761904e-05, + "loss": 0.5027, + "step": 411 + }, + { + "epoch": 0.01, + "learning_rate": 8.174603174603175e-05, + "loss": 0.616, + "step": 412 + }, + { + "epoch": 0.01, + "learning_rate": 8.194444444444445e-05, + "loss": 0.7486, + "step": 413 + }, + { + "epoch": 0.01, + "learning_rate": 8.214285714285714e-05, + "loss": 0.6606, + "step": 414 + }, + { + "epoch": 0.01, + "learning_rate": 8.234126984126984e-05, + "loss": 0.4418, + "step": 415 + }, + { + "epoch": 0.01, + "learning_rate": 8.253968253968255e-05, + "loss": 0.5274, + "step": 416 + }, + { + "epoch": 0.01, + "learning_rate": 8.273809523809524e-05, + "loss": 0.5746, + "step": 417 + }, + { + "epoch": 0.01, + "learning_rate": 8.293650793650795e-05, + "loss": 0.5849, + "step": 418 + }, + { + "epoch": 0.01, + "learning_rate": 8.313492063492064e-05, + "loss": 0.6497, + "step": 419 + }, + { + "epoch": 0.01, + "learning_rate": 8.333333333333334e-05, + "loss": 0.7048, + "step": 420 + }, + { + "epoch": 0.01, + "learning_rate": 8.353174603174603e-05, + "loss": 0.6669, + "step": 421 + }, + { + "epoch": 0.01, + "learning_rate": 8.373015873015874e-05, + "loss": 0.5875, + "step": 422 + }, + { + "epoch": 0.01, + "learning_rate": 8.392857142857144e-05, + "loss": 0.5043, + "step": 423 + }, + { + "epoch": 0.01, + "learning_rate": 8.412698412698413e-05, + "loss": 0.6041, + "step": 424 + }, + { + "epoch": 0.01, + "learning_rate": 8.432539682539683e-05, + "loss": 0.6077, + "step": 425 + }, + { + "epoch": 0.01, + "learning_rate": 8.452380952380952e-05, + "loss": 0.6177, + "step": 426 + }, + { + "epoch": 0.01, + "learning_rate": 8.472222222222222e-05, + "loss": 0.6084, + "step": 427 + }, + { + "epoch": 0.01, + "learning_rate": 8.492063492063493e-05, + "loss": 0.6472, + "step": 428 + }, + { + "epoch": 0.01, + "learning_rate": 8.511904761904762e-05, + "loss": 0.6064, + "step": 429 + }, + { + "epoch": 0.01, + "learning_rate": 8.531746031746032e-05, + "loss": 0.6315, + "step": 430 + }, + { + "epoch": 0.01, + "learning_rate": 8.551587301587301e-05, + "loss": 0.6066, + "step": 431 + }, + { + "epoch": 0.01, + "learning_rate": 8.571428571428571e-05, + "loss": 0.5614, + "step": 432 + }, + { + "epoch": 0.01, + "learning_rate": 8.591269841269842e-05, + "loss": 0.6501, + "step": 433 + }, + { + "epoch": 0.01, + "learning_rate": 8.611111111111112e-05, + "loss": 0.7144, + "step": 434 + }, + { + "epoch": 0.01, + "learning_rate": 8.630952380952382e-05, + "loss": 0.5757, + "step": 435 + }, + { + "epoch": 0.01, + "learning_rate": 8.650793650793651e-05, + "loss": 0.577, + "step": 436 + }, + { + "epoch": 0.01, + "learning_rate": 8.670634920634921e-05, + "loss": 0.5314, + "step": 437 + }, + { + "epoch": 0.01, + "learning_rate": 8.690476190476192e-05, + "loss": 0.591, + "step": 438 + }, + { + "epoch": 0.01, + "learning_rate": 8.71031746031746e-05, + "loss": 0.719, + "step": 439 + }, + { + "epoch": 0.01, + "learning_rate": 8.730158730158731e-05, + "loss": 0.7256, + "step": 440 + }, + { + "epoch": 0.01, + "learning_rate": 8.75e-05, + "loss": 0.6289, + "step": 441 + }, + { + "epoch": 0.01, + "learning_rate": 8.76984126984127e-05, + "loss": 0.6068, + "step": 442 + }, + { + "epoch": 0.01, + "learning_rate": 8.78968253968254e-05, + "loss": 0.4823, + "step": 443 + }, + { + "epoch": 0.01, + "learning_rate": 8.80952380952381e-05, + "loss": 0.6923, + "step": 444 + }, + { + "epoch": 0.01, + "learning_rate": 8.82936507936508e-05, + "loss": 0.7015, + "step": 445 + }, + { + "epoch": 0.01, + "learning_rate": 8.849206349206349e-05, + "loss": 0.6315, + "step": 446 + }, + { + "epoch": 0.01, + "learning_rate": 8.869047619047619e-05, + "loss": 0.67, + "step": 447 + }, + { + "epoch": 0.01, + "learning_rate": 8.888888888888889e-05, + "loss": 0.5746, + "step": 448 + }, + { + "epoch": 0.01, + "learning_rate": 8.90873015873016e-05, + "loss": 0.5857, + "step": 449 + }, + { + "epoch": 0.01, + "learning_rate": 8.92857142857143e-05, + "loss": 0.5708, + "step": 450 + }, + { + "epoch": 0.01, + "learning_rate": 8.948412698412699e-05, + "loss": 0.5721, + "step": 451 + }, + { + "epoch": 0.01, + "learning_rate": 8.968253968253969e-05, + "loss": 0.527, + "step": 452 + }, + { + "epoch": 0.01, + "learning_rate": 8.988095238095238e-05, + "loss": 0.5687, + "step": 453 + }, + { + "epoch": 0.01, + "learning_rate": 9.007936507936508e-05, + "loss": 0.6109, + "step": 454 + }, + { + "epoch": 0.01, + "learning_rate": 9.027777777777779e-05, + "loss": 0.5666, + "step": 455 + }, + { + "epoch": 0.01, + "learning_rate": 9.047619047619048e-05, + "loss": 0.6079, + "step": 456 + }, + { + "epoch": 0.01, + "learning_rate": 9.067460317460318e-05, + "loss": 0.6231, + "step": 457 + }, + { + "epoch": 0.01, + "learning_rate": 9.087301587301588e-05, + "loss": 0.6556, + "step": 458 + }, + { + "epoch": 0.01, + "learning_rate": 9.107142857142857e-05, + "loss": 0.7161, + "step": 459 + }, + { + "epoch": 0.01, + "learning_rate": 9.126984126984128e-05, + "loss": 0.6076, + "step": 460 + }, + { + "epoch": 0.01, + "learning_rate": 9.146825396825396e-05, + "loss": 0.7785, + "step": 461 + }, + { + "epoch": 0.01, + "learning_rate": 9.166666666666667e-05, + "loss": 0.5724, + "step": 462 + }, + { + "epoch": 0.01, + "learning_rate": 9.186507936507937e-05, + "loss": 0.669, + "step": 463 + }, + { + "epoch": 0.01, + "learning_rate": 9.206349206349206e-05, + "loss": 0.5919, + "step": 464 + }, + { + "epoch": 0.01, + "learning_rate": 9.226190476190478e-05, + "loss": 0.5645, + "step": 465 + }, + { + "epoch": 0.01, + "learning_rate": 9.246031746031747e-05, + "loss": 0.6475, + "step": 466 + }, + { + "epoch": 0.01, + "learning_rate": 9.265873015873017e-05, + "loss": 0.6372, + "step": 467 + }, + { + "epoch": 0.01, + "learning_rate": 9.285714285714286e-05, + "loss": 0.4834, + "step": 468 + }, + { + "epoch": 0.01, + "learning_rate": 9.305555555555556e-05, + "loss": 0.7282, + "step": 469 + }, + { + "epoch": 0.01, + "learning_rate": 9.325396825396826e-05, + "loss": 0.6216, + "step": 470 + }, + { + "epoch": 0.01, + "learning_rate": 9.345238095238095e-05, + "loss": 0.6187, + "step": 471 + }, + { + "epoch": 0.01, + "learning_rate": 9.365079365079366e-05, + "loss": 0.5911, + "step": 472 + }, + { + "epoch": 0.01, + "learning_rate": 9.384920634920635e-05, + "loss": 0.5754, + "step": 473 + }, + { + "epoch": 0.01, + "learning_rate": 9.404761904761905e-05, + "loss": 0.5957, + "step": 474 + }, + { + "epoch": 0.01, + "learning_rate": 9.424603174603175e-05, + "loss": 0.5444, + "step": 475 + }, + { + "epoch": 0.01, + "learning_rate": 9.444444444444444e-05, + "loss": 0.5114, + "step": 476 + }, + { + "epoch": 0.01, + "learning_rate": 9.464285714285715e-05, + "loss": 0.6617, + "step": 477 + }, + { + "epoch": 0.01, + "learning_rate": 9.484126984126985e-05, + "loss": 0.6057, + "step": 478 + }, + { + "epoch": 0.01, + "learning_rate": 9.503968253968254e-05, + "loss": 0.6657, + "step": 479 + }, + { + "epoch": 0.01, + "learning_rate": 9.523809523809524e-05, + "loss": 0.5363, + "step": 480 + }, + { + "epoch": 0.01, + "learning_rate": 9.543650793650794e-05, + "loss": 0.5357, + "step": 481 + }, + { + "epoch": 0.01, + "learning_rate": 9.563492063492065e-05, + "loss": 0.6387, + "step": 482 + }, + { + "epoch": 0.01, + "learning_rate": 9.583333333333334e-05, + "loss": 0.5977, + "step": 483 + }, + { + "epoch": 0.01, + "learning_rate": 9.603174603174604e-05, + "loss": 0.554, + "step": 484 + }, + { + "epoch": 0.01, + "learning_rate": 9.623015873015874e-05, + "loss": 0.6068, + "step": 485 + }, + { + "epoch": 0.01, + "learning_rate": 9.642857142857143e-05, + "loss": 0.6756, + "step": 486 + }, + { + "epoch": 0.01, + "learning_rate": 9.662698412698413e-05, + "loss": 0.7167, + "step": 487 + }, + { + "epoch": 0.01, + "learning_rate": 9.682539682539682e-05, + "loss": 0.6089, + "step": 488 + }, + { + "epoch": 0.01, + "learning_rate": 9.702380952380953e-05, + "loss": 0.6318, + "step": 489 + }, + { + "epoch": 0.01, + "learning_rate": 9.722222222222223e-05, + "loss": 0.5645, + "step": 490 + }, + { + "epoch": 0.01, + "learning_rate": 9.742063492063492e-05, + "loss": 0.5802, + "step": 491 + }, + { + "epoch": 0.01, + "learning_rate": 9.761904761904762e-05, + "loss": 0.6911, + "step": 492 + }, + { + "epoch": 0.01, + "learning_rate": 9.781746031746031e-05, + "loss": 0.5249, + "step": 493 + }, + { + "epoch": 0.01, + "learning_rate": 9.801587301587302e-05, + "loss": 0.6086, + "step": 494 + }, + { + "epoch": 0.01, + "learning_rate": 9.821428571428572e-05, + "loss": 0.6338, + "step": 495 + }, + { + "epoch": 0.01, + "learning_rate": 9.841269841269841e-05, + "loss": 0.6251, + "step": 496 + }, + { + "epoch": 0.01, + "learning_rate": 9.861111111111112e-05, + "loss": 0.5742, + "step": 497 + }, + { + "epoch": 0.01, + "learning_rate": 9.880952380952381e-05, + "loss": 0.6692, + "step": 498 + }, + { + "epoch": 0.01, + "learning_rate": 9.900793650793652e-05, + "loss": 0.6653, + "step": 499 + }, + { + "epoch": 0.01, + "learning_rate": 9.920634920634922e-05, + "loss": 0.5403, + "step": 500 + }, + { + "epoch": 0.01, + "learning_rate": 9.940476190476191e-05, + "loss": 0.6439, + "step": 501 + }, + { + "epoch": 0.01, + "learning_rate": 9.960317460317461e-05, + "loss": 0.6391, + "step": 502 + }, + { + "epoch": 0.01, + "learning_rate": 9.98015873015873e-05, + "loss": 0.5594, + "step": 503 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001, + "loss": 0.5864, + "step": 504 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010019841269841271, + "loss": 0.5467, + "step": 505 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001003968253968254, + "loss": 0.6634, + "step": 506 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001005952380952381, + "loss": 0.5858, + "step": 507 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010079365079365079, + "loss": 0.64, + "step": 508 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010099206349206349, + "loss": 0.5473, + "step": 509 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001011904761904762, + "loss": 0.5443, + "step": 510 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010138888888888889, + "loss": 0.6851, + "step": 511 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010158730158730159, + "loss": 0.7659, + "step": 512 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010178571428571428, + "loss": 0.6519, + "step": 513 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010198412698412698, + "loss": 0.6311, + "step": 514 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010218253968253968, + "loss": 0.6508, + "step": 515 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010238095238095237, + "loss": 0.7558, + "step": 516 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010257936507936508, + "loss": 0.6666, + "step": 517 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010277777777777778, + "loss": 0.5105, + "step": 518 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010297619047619047, + "loss": 0.5636, + "step": 519 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010317460317460319, + "loss": 0.5856, + "step": 520 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010337301587301589, + "loss": 0.5347, + "step": 521 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010357142857142859, + "loss": 0.5918, + "step": 522 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010376984126984128, + "loss": 0.6906, + "step": 523 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010396825396825398, + "loss": 0.61, + "step": 524 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010416666666666667, + "loss": 0.5889, + "step": 525 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010436507936507938, + "loss": 0.5799, + "step": 526 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010456349206349208, + "loss": 0.6315, + "step": 527 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010476190476190477, + "loss": 0.6846, + "step": 528 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010496031746031747, + "loss": 0.6354, + "step": 529 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010515873015873016, + "loss": 0.5604, + "step": 530 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010535714285714286, + "loss": 0.6581, + "step": 531 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010555555555555557, + "loss": 0.6581, + "step": 532 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010575396825396826, + "loss": 0.5659, + "step": 533 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010595238095238096, + "loss": 0.7038, + "step": 534 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010615079365079366, + "loss": 0.6216, + "step": 535 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010634920634920635, + "loss": 0.6273, + "step": 536 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010654761904761906, + "loss": 0.7273, + "step": 537 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010674603174603174, + "loss": 0.6685, + "step": 538 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010694444444444445, + "loss": 0.6729, + "step": 539 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010714285714285715, + "loss": 0.5569, + "step": 540 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010734126984126984, + "loss": 0.6327, + "step": 541 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010753968253968254, + "loss": 0.5352, + "step": 542 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010773809523809523, + "loss": 0.6745, + "step": 543 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010793650793650794, + "loss": 0.6013, + "step": 544 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010813492063492064, + "loss": 0.6075, + "step": 545 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010833333333333333, + "loss": 0.6478, + "step": 546 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010853174603174603, + "loss": 0.6244, + "step": 547 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010873015873015872, + "loss": 0.6139, + "step": 548 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010892857142857142, + "loss": 0.5347, + "step": 549 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010912698412698413, + "loss": 0.602, + "step": 550 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010932539682539682, + "loss": 0.608, + "step": 551 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010952380952380953, + "loss": 0.5166, + "step": 552 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010972222222222224, + "loss": 0.6605, + "step": 553 + }, + { + "epoch": 0.02, + "learning_rate": 0.00010992063492063494, + "loss": 0.6199, + "step": 554 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011011904761904763, + "loss": 0.6404, + "step": 555 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011031746031746033, + "loss": 0.5891, + "step": 556 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011051587301587303, + "loss": 0.5231, + "step": 557 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011071428571428572, + "loss": 0.6431, + "step": 558 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011091269841269843, + "loss": 0.581, + "step": 559 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011111111111111112, + "loss": 0.5868, + "step": 560 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011130952380952382, + "loss": 0.6038, + "step": 561 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011150793650793652, + "loss": 0.621, + "step": 562 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011170634920634921, + "loss": 0.5246, + "step": 563 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011190476190476191, + "loss": 0.5548, + "step": 564 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001121031746031746, + "loss": 0.6626, + "step": 565 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011230158730158731, + "loss": 0.6297, + "step": 566 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011250000000000001, + "loss": 0.5686, + "step": 567 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001126984126984127, + "loss": 0.6431, + "step": 568 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001128968253968254, + "loss": 0.551, + "step": 569 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011309523809523809, + "loss": 0.6512, + "step": 570 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001132936507936508, + "loss": 0.5273, + "step": 571 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001134920634920635, + "loss": 0.6097, + "step": 572 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011369047619047619, + "loss": 0.5724, + "step": 573 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011388888888888889, + "loss": 0.6209, + "step": 574 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011408730158730158, + "loss": 0.6587, + "step": 575 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011428571428571428, + "loss": 0.6074, + "step": 576 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011448412698412699, + "loss": 0.593, + "step": 577 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011468253968253968, + "loss": 0.7166, + "step": 578 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011488095238095238, + "loss": 0.6875, + "step": 579 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011507936507936508, + "loss": 0.6061, + "step": 580 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011527777777777777, + "loss": 0.5682, + "step": 581 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011547619047619047, + "loss": 0.6284, + "step": 582 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011567460317460319, + "loss": 0.497, + "step": 583 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001158730158730159, + "loss": 0.5523, + "step": 584 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011607142857142858, + "loss": 0.6942, + "step": 585 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011626984126984129, + "loss": 0.5189, + "step": 586 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011646825396825398, + "loss": 0.5919, + "step": 587 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011666666666666668, + "loss": 0.4944, + "step": 588 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011686507936507938, + "loss": 0.5977, + "step": 589 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011706349206349207, + "loss": 0.5673, + "step": 590 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011726190476190477, + "loss": 0.5261, + "step": 591 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011746031746031746, + "loss": 0.6583, + "step": 592 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011765873015873017, + "loss": 0.5498, + "step": 593 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011785714285714287, + "loss": 0.6906, + "step": 594 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011805555555555556, + "loss": 0.4861, + "step": 595 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011825396825396826, + "loss": 0.7152, + "step": 596 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011845238095238097, + "loss": 0.6274, + "step": 597 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011865079365079366, + "loss": 0.5662, + "step": 598 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011884920634920636, + "loss": 0.619, + "step": 599 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011904761904761905, + "loss": 0.5628, + "step": 600 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011924603174603175, + "loss": 0.6372, + "step": 601 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011944444444444445, + "loss": 0.5879, + "step": 602 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011964285714285714, + "loss": 0.5425, + "step": 603 + }, + { + "epoch": 0.02, + "learning_rate": 0.00011984126984126985, + "loss": 0.5724, + "step": 604 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012003968253968254, + "loss": 0.561, + "step": 605 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012023809523809524, + "loss": 0.571, + "step": 606 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012043650793650794, + "loss": 0.6514, + "step": 607 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012063492063492063, + "loss": 0.6023, + "step": 608 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012083333333333333, + "loss": 0.6271, + "step": 609 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012103174603174602, + "loss": 0.593, + "step": 610 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012123015873015873, + "loss": 0.6554, + "step": 611 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012142857142857143, + "loss": 0.6571, + "step": 612 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012162698412698412, + "loss": 0.6189, + "step": 613 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012182539682539682, + "loss": 0.6016, + "step": 614 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012202380952380954, + "loss": 0.5631, + "step": 615 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012222222222222224, + "loss": 0.6032, + "step": 616 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012242063492063493, + "loss": 0.6868, + "step": 617 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012261904761904762, + "loss": 0.5534, + "step": 618 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012281746031746034, + "loss": 0.6579, + "step": 619 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012301587301587303, + "loss": 0.5344, + "step": 620 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012321428571428572, + "loss": 0.6887, + "step": 621 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012341269841269843, + "loss": 0.5957, + "step": 622 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012361111111111112, + "loss": 0.6408, + "step": 623 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001238095238095238, + "loss": 0.7625, + "step": 624 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012400793650793653, + "loss": 0.5622, + "step": 625 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012420634920634922, + "loss": 0.6117, + "step": 626 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001244047619047619, + "loss": 0.6017, + "step": 627 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012460317460317462, + "loss": 0.61, + "step": 628 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001248015873015873, + "loss": 0.5982, + "step": 629 + }, + { + "epoch": 0.02, + "learning_rate": 0.000125, + "loss": 0.5345, + "step": 630 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001251984126984127, + "loss": 0.5405, + "step": 631 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001253968253968254, + "loss": 0.5127, + "step": 632 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001255952380952381, + "loss": 0.5678, + "step": 633 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001257936507936508, + "loss": 0.6278, + "step": 634 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001259920634920635, + "loss": 0.6634, + "step": 635 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001261904761904762, + "loss": 0.6239, + "step": 636 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012638888888888888, + "loss": 0.6486, + "step": 637 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001265873015873016, + "loss": 0.5421, + "step": 638 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001267857142857143, + "loss": 0.5461, + "step": 639 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012698412698412698, + "loss": 0.5671, + "step": 640 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012718253968253967, + "loss": 0.6231, + "step": 641 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012738095238095238, + "loss": 0.5623, + "step": 642 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012757936507936507, + "loss": 0.6552, + "step": 643 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012777777777777776, + "loss": 0.6291, + "step": 644 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012797619047619048, + "loss": 0.5694, + "step": 645 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001281746031746032, + "loss": 0.6296, + "step": 646 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012837301587301589, + "loss": 0.6084, + "step": 647 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012857142857142858, + "loss": 0.5382, + "step": 648 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001287698412698413, + "loss": 0.5267, + "step": 649 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012896825396825398, + "loss": 0.5231, + "step": 650 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012916666666666667, + "loss": 0.5477, + "step": 651 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001293650793650794, + "loss": 0.6554, + "step": 652 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012956349206349208, + "loss": 0.6971, + "step": 653 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012976190476190477, + "loss": 0.5703, + "step": 654 + }, + { + "epoch": 0.02, + "learning_rate": 0.00012996031746031748, + "loss": 0.6902, + "step": 655 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013015873015873017, + "loss": 0.5423, + "step": 656 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013035714285714286, + "loss": 0.5164, + "step": 657 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013055555555555555, + "loss": 0.5582, + "step": 658 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013075396825396827, + "loss": 0.5494, + "step": 659 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013095238095238096, + "loss": 0.5447, + "step": 660 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013115079365079365, + "loss": 0.566, + "step": 661 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013134920634920636, + "loss": 0.6133, + "step": 662 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013154761904761905, + "loss": 0.5594, + "step": 663 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013174603174603174, + "loss": 0.5443, + "step": 664 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013194444444444446, + "loss": 0.5856, + "step": 665 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013214285714285715, + "loss": 0.5478, + "step": 666 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013234126984126984, + "loss": 0.4644, + "step": 667 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013253968253968255, + "loss": 0.513, + "step": 668 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013273809523809524, + "loss": 0.5235, + "step": 669 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013293650793650793, + "loss": 0.5832, + "step": 670 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013313492063492062, + "loss": 0.5123, + "step": 671 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013333333333333334, + "loss": 0.5969, + "step": 672 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013353174603174603, + "loss": 0.5245, + "step": 673 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013373015873015872, + "loss": 0.5568, + "step": 674 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013392857142857144, + "loss": 0.5253, + "step": 675 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013412698412698412, + "loss": 0.5688, + "step": 676 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013432539682539681, + "loss": 0.499, + "step": 677 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013452380952380953, + "loss": 0.5732, + "step": 678 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013472222222222225, + "loss": 0.5814, + "step": 679 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013492063492063494, + "loss": 0.5, + "step": 680 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013511904761904763, + "loss": 0.5432, + "step": 681 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013531746031746034, + "loss": 0.5753, + "step": 682 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013551587301587303, + "loss": 0.6946, + "step": 683 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013571428571428572, + "loss": 0.5983, + "step": 684 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013591269841269844, + "loss": 0.5787, + "step": 685 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013611111111111113, + "loss": 0.5494, + "step": 686 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013630952380952382, + "loss": 0.6744, + "step": 687 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001365079365079365, + "loss": 0.6524, + "step": 688 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013670634920634922, + "loss": 0.6237, + "step": 689 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001369047619047619, + "loss": 0.4791, + "step": 690 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001371031746031746, + "loss": 0.5042, + "step": 691 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013730158730158732, + "loss": 0.5614, + "step": 692 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001375, + "loss": 0.5458, + "step": 693 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001376984126984127, + "loss": 0.5328, + "step": 694 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013789682539682541, + "loss": 0.6203, + "step": 695 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001380952380952381, + "loss": 0.5658, + "step": 696 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001382936507936508, + "loss": 0.5797, + "step": 697 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013849206349206348, + "loss": 0.605, + "step": 698 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001386904761904762, + "loss": 0.5915, + "step": 699 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001388888888888889, + "loss": 0.6583, + "step": 700 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013908730158730158, + "loss": 0.6283, + "step": 701 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001392857142857143, + "loss": 0.6748, + "step": 702 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013948412698412698, + "loss": 0.4811, + "step": 703 + }, + { + "epoch": 0.02, + "learning_rate": 0.00013968253968253967, + "loss": 0.5369, + "step": 704 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001398809523809524, + "loss": 0.7503, + "step": 705 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014007936507936508, + "loss": 0.4761, + "step": 706 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014027777777777777, + "loss": 0.5651, + "step": 707 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014047619047619049, + "loss": 0.5474, + "step": 708 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001406746031746032, + "loss": 0.6253, + "step": 709 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001408730158730159, + "loss": 0.661, + "step": 710 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014107142857142858, + "loss": 0.6596, + "step": 711 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001412698412698413, + "loss": 0.5516, + "step": 712 + }, + { + "epoch": 0.02, + "learning_rate": 0.000141468253968254, + "loss": 0.599, + "step": 713 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014166666666666668, + "loss": 0.6841, + "step": 714 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014186507936507937, + "loss": 0.5925, + "step": 715 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014206349206349208, + "loss": 0.6928, + "step": 716 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014226190476190477, + "loss": 0.6012, + "step": 717 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014246031746031746, + "loss": 0.6322, + "step": 718 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014265873015873018, + "loss": 0.5591, + "step": 719 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014285714285714287, + "loss": 0.5583, + "step": 720 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014305555555555556, + "loss": 0.5132, + "step": 721 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014325396825396827, + "loss": 0.664, + "step": 722 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014345238095238096, + "loss": 0.646, + "step": 723 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014365079365079365, + "loss": 0.5798, + "step": 724 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014384920634920634, + "loss": 0.5603, + "step": 725 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014404761904761906, + "loss": 0.5157, + "step": 726 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014424603174603175, + "loss": 0.8202, + "step": 727 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014444444444444444, + "loss": 0.598, + "step": 728 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014464285714285715, + "loss": 0.5346, + "step": 729 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014484126984126984, + "loss": 0.6771, + "step": 730 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014503968253968253, + "loss": 0.5321, + "step": 731 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014523809523809525, + "loss": 0.5905, + "step": 732 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014543650793650794, + "loss": 0.6566, + "step": 733 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014563492063492063, + "loss": 0.5961, + "step": 734 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014583333333333335, + "loss": 0.5996, + "step": 735 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014603174603174603, + "loss": 0.5282, + "step": 736 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014623015873015872, + "loss": 0.7372, + "step": 737 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014642857142857141, + "loss": 0.6625, + "step": 738 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014662698412698413, + "loss": 0.6668, + "step": 739 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014682539682539682, + "loss": 0.6968, + "step": 740 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014702380952380954, + "loss": 0.6523, + "step": 741 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014722222222222223, + "loss": 0.5504, + "step": 742 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014742063492063494, + "loss": 0.6566, + "step": 743 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014761904761904763, + "loss": 0.6224, + "step": 744 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014781746031746032, + "loss": 0.5992, + "step": 745 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014801587301587304, + "loss": 0.4742, + "step": 746 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014821428571428573, + "loss": 0.5905, + "step": 747 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014841269841269842, + "loss": 0.5515, + "step": 748 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014861111111111113, + "loss": 0.6159, + "step": 749 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014880952380952382, + "loss": 0.6582, + "step": 750 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001490079365079365, + "loss": 0.6098, + "step": 751 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014920634920634923, + "loss": 0.5662, + "step": 752 + }, + { + "epoch": 0.02, + "learning_rate": 0.00014940476190476192, + "loss": 0.5681, + "step": 753 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001496031746031746, + "loss": 0.542, + "step": 754 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001498015873015873, + "loss": 0.5864, + "step": 755 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015000000000000001, + "loss": 0.5982, + "step": 756 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001501984126984127, + "loss": 0.6506, + "step": 757 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001503968253968254, + "loss": 0.6617, + "step": 758 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001505952380952381, + "loss": 0.6071, + "step": 759 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001507936507936508, + "loss": 0.5273, + "step": 760 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001509920634920635, + "loss": 0.5889, + "step": 761 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001511904761904762, + "loss": 0.5849, + "step": 762 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001513888888888889, + "loss": 0.6203, + "step": 763 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015158730158730158, + "loss": 0.489, + "step": 764 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015178571428571427, + "loss": 0.6182, + "step": 765 + }, + { + "epoch": 0.02, + "learning_rate": 0.000151984126984127, + "loss": 0.6064, + "step": 766 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015218253968253968, + "loss": 0.605, + "step": 767 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015238095238095237, + "loss": 0.5718, + "step": 768 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015257936507936509, + "loss": 0.6284, + "step": 769 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015277777777777777, + "loss": 0.5796, + "step": 770 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015297619047619046, + "loss": 0.585, + "step": 771 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015317460317460318, + "loss": 0.6889, + "step": 772 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001533730158730159, + "loss": 0.551, + "step": 773 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001535714285714286, + "loss": 0.6337, + "step": 774 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015376984126984128, + "loss": 0.5539, + "step": 775 + }, + { + "epoch": 0.02, + "learning_rate": 0.000153968253968254, + "loss": 0.7314, + "step": 776 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015416666666666668, + "loss": 0.7143, + "step": 777 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015436507936507937, + "loss": 0.5675, + "step": 778 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001545634920634921, + "loss": 0.5554, + "step": 779 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015476190476190478, + "loss": 0.7169, + "step": 780 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015496031746031747, + "loss": 0.619, + "step": 781 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015515873015873016, + "loss": 0.5866, + "step": 782 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015535714285714287, + "loss": 0.6841, + "step": 783 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015555555555555556, + "loss": 0.6245, + "step": 784 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015575396825396825, + "loss": 0.6089, + "step": 785 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015595238095238097, + "loss": 0.7067, + "step": 786 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015615079365079366, + "loss": 0.5267, + "step": 787 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015634920634920635, + "loss": 0.6536, + "step": 788 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015654761904761906, + "loss": 0.5952, + "step": 789 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015674603174603175, + "loss": 0.6781, + "step": 790 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015694444444444444, + "loss": 0.6562, + "step": 791 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015714285714285716, + "loss": 0.5796, + "step": 792 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015734126984126985, + "loss": 0.556, + "step": 793 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015753968253968254, + "loss": 0.6289, + "step": 794 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015773809523809523, + "loss": 0.5872, + "step": 795 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015793650793650795, + "loss": 0.5921, + "step": 796 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015813492063492063, + "loss": 0.5871, + "step": 797 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015833333333333332, + "loss": 0.5834, + "step": 798 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015853174603174604, + "loss": 0.6879, + "step": 799 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015873015873015873, + "loss": 0.6374, + "step": 800 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015892857142857142, + "loss": 0.6849, + "step": 801 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015912698412698414, + "loss": 0.6065, + "step": 802 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015932539682539683, + "loss": 0.4463, + "step": 803 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015952380952380954, + "loss": 0.5655, + "step": 804 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015972222222222223, + "loss": 0.5536, + "step": 805 + }, + { + "epoch": 0.02, + "learning_rate": 0.00015992063492063495, + "loss": 0.7011, + "step": 806 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016011904761904764, + "loss": 0.567, + "step": 807 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016031746031746033, + "loss": 0.6564, + "step": 808 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016051587301587304, + "loss": 0.6228, + "step": 809 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016071428571428573, + "loss": 0.6643, + "step": 810 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016091269841269842, + "loss": 0.6247, + "step": 811 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001611111111111111, + "loss": 0.494, + "step": 812 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016130952380952383, + "loss": 0.6536, + "step": 813 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016150793650793652, + "loss": 0.7174, + "step": 814 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001617063492063492, + "loss": 0.5312, + "step": 815 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016190476190476192, + "loss": 0.5508, + "step": 816 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016210317460317461, + "loss": 0.5957, + "step": 817 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001623015873015873, + "loss": 0.6072, + "step": 818 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016250000000000002, + "loss": 0.5881, + "step": 819 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001626984126984127, + "loss": 0.6251, + "step": 820 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001628968253968254, + "loss": 0.6686, + "step": 821 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001630952380952381, + "loss": 0.543, + "step": 822 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001632936507936508, + "loss": 0.5853, + "step": 823 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001634920634920635, + "loss": 0.5797, + "step": 824 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016369047619047618, + "loss": 0.5661, + "step": 825 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001638888888888889, + "loss": 0.6463, + "step": 826 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001640873015873016, + "loss": 0.5461, + "step": 827 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016428571428571428, + "loss": 0.5895, + "step": 828 + }, + { + "epoch": 0.02, + "learning_rate": 0.000164484126984127, + "loss": 0.488, + "step": 829 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016468253968253969, + "loss": 0.5349, + "step": 830 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016488095238095237, + "loss": 0.5895, + "step": 831 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001650793650793651, + "loss": 0.664, + "step": 832 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016527777777777778, + "loss": 0.5883, + "step": 833 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016547619047619047, + "loss": 0.7084, + "step": 834 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001656746031746032, + "loss": 0.5541, + "step": 835 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001658730158730159, + "loss": 0.685, + "step": 836 + }, + { + "epoch": 0.02, + "learning_rate": 0.0001660714285714286, + "loss": 0.4906, + "step": 837 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016626984126984128, + "loss": 0.6592, + "step": 838 + }, + { + "epoch": 0.02, + "learning_rate": 0.00016646825396825397, + "loss": 0.6274, + "step": 839 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001666666666666667, + "loss": 0.6269, + "step": 840 + }, + { + "epoch": 0.03, + "learning_rate": 0.00016686507936507938, + "loss": 0.6763, + "step": 841 + }, + { + "epoch": 0.03, + "learning_rate": 0.00016706349206349207, + "loss": 0.6322, + "step": 842 + }, + { + "epoch": 0.03, + "learning_rate": 0.00016726190476190478, + "loss": 0.613, + "step": 843 + }, + { + "epoch": 0.03, + "learning_rate": 0.00016746031746031747, + "loss": 0.6134, + "step": 844 + }, + { + "epoch": 0.03, + "learning_rate": 0.00016765873015873016, + "loss": 0.5857, + "step": 845 + }, + { + "epoch": 0.03, + "learning_rate": 0.00016785714285714288, + "loss": 0.6561, + "step": 846 + }, + { + "epoch": 0.03, + "learning_rate": 0.00016805555555555557, + "loss": 0.6166, + "step": 847 + }, + { + "epoch": 0.03, + "learning_rate": 0.00016825396825396826, + "loss": 0.4867, + "step": 848 + }, + { + "epoch": 0.03, + "learning_rate": 0.00016845238095238097, + "loss": 0.6235, + "step": 849 + }, + { + "epoch": 0.03, + "learning_rate": 0.00016865079365079366, + "loss": 0.6102, + "step": 850 + }, + { + "epoch": 0.03, + "learning_rate": 0.00016884920634920635, + "loss": 0.5555, + "step": 851 + }, + { + "epoch": 0.03, + "learning_rate": 0.00016904761904761904, + "loss": 0.5592, + "step": 852 + }, + { + "epoch": 0.03, + "learning_rate": 0.00016924603174603176, + "loss": 0.6546, + "step": 853 + }, + { + "epoch": 0.03, + "learning_rate": 0.00016944444444444445, + "loss": 0.6227, + "step": 854 + }, + { + "epoch": 0.03, + "learning_rate": 0.00016964285714285714, + "loss": 0.6311, + "step": 855 + }, + { + "epoch": 0.03, + "learning_rate": 0.00016984126984126986, + "loss": 0.4425, + "step": 856 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017003968253968254, + "loss": 0.5649, + "step": 857 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017023809523809523, + "loss": 0.6347, + "step": 858 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017043650793650795, + "loss": 0.4978, + "step": 859 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017063492063492064, + "loss": 0.6033, + "step": 860 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017083333333333333, + "loss": 0.6866, + "step": 861 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017103174603174602, + "loss": 0.5378, + "step": 862 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017123015873015874, + "loss": 0.5155, + "step": 863 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017142857142857143, + "loss": 0.677, + "step": 864 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017162698412698411, + "loss": 0.6907, + "step": 865 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017182539682539683, + "loss": 0.5827, + "step": 866 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017202380952380955, + "loss": 0.7159, + "step": 867 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017222222222222224, + "loss": 0.7951, + "step": 868 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017242063492063493, + "loss": 0.5441, + "step": 869 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017261904761904764, + "loss": 0.5972, + "step": 870 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017281746031746033, + "loss": 0.5144, + "step": 871 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017301587301587302, + "loss": 0.5735, + "step": 872 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017321428571428574, + "loss": 0.5025, + "step": 873 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017341269841269843, + "loss": 0.6683, + "step": 874 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017361111111111112, + "loss": 0.5903, + "step": 875 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017380952380952383, + "loss": 0.5012, + "step": 876 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017400793650793652, + "loss": 0.6292, + "step": 877 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001742063492063492, + "loss": 0.4923, + "step": 878 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001744047619047619, + "loss": 0.5454, + "step": 879 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017460317460317462, + "loss": 0.5788, + "step": 880 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001748015873015873, + "loss": 0.4471, + "step": 881 + }, + { + "epoch": 0.03, + "learning_rate": 0.000175, + "loss": 0.5714, + "step": 882 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017519841269841271, + "loss": 0.408, + "step": 883 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001753968253968254, + "loss": 0.5842, + "step": 884 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001755952380952381, + "loss": 0.6419, + "step": 885 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001757936507936508, + "loss": 0.5885, + "step": 886 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001759920634920635, + "loss": 0.5812, + "step": 887 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001761904761904762, + "loss": 0.6771, + "step": 888 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001763888888888889, + "loss": 0.4521, + "step": 889 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001765873015873016, + "loss": 0.674, + "step": 890 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017678571428571428, + "loss": 0.6208, + "step": 891 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017698412698412697, + "loss": 0.66, + "step": 892 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001771825396825397, + "loss": 0.6522, + "step": 893 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017738095238095238, + "loss": 0.5743, + "step": 894 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017757936507936507, + "loss": 0.602, + "step": 895 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017777777777777779, + "loss": 0.611, + "step": 896 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017797619047619048, + "loss": 0.6818, + "step": 897 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001781746031746032, + "loss": 0.6162, + "step": 898 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017837301587301588, + "loss": 0.6502, + "step": 899 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001785714285714286, + "loss": 0.5833, + "step": 900 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001787698412698413, + "loss": 0.5555, + "step": 901 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017896825396825398, + "loss": 0.5627, + "step": 902 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001791666666666667, + "loss": 0.4828, + "step": 903 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017936507936507938, + "loss": 0.58, + "step": 904 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017956349206349207, + "loss": 0.6444, + "step": 905 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017976190476190476, + "loss": 0.5278, + "step": 906 + }, + { + "epoch": 0.03, + "learning_rate": 0.00017996031746031748, + "loss": 0.7012, + "step": 907 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018015873015873017, + "loss": 0.5312, + "step": 908 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018035714285714286, + "loss": 0.5832, + "step": 909 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018055555555555557, + "loss": 0.6169, + "step": 910 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018075396825396826, + "loss": 0.6055, + "step": 911 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018095238095238095, + "loss": 0.7029, + "step": 912 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018115079365079367, + "loss": 0.5129, + "step": 913 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018134920634920636, + "loss": 0.5869, + "step": 914 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018154761904761905, + "loss": 0.5025, + "step": 915 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018174603174603177, + "loss": 0.6189, + "step": 916 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018194444444444445, + "loss": 0.6318, + "step": 917 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018214285714285714, + "loss": 0.6288, + "step": 918 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018234126984126983, + "loss": 0.6842, + "step": 919 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018253968253968255, + "loss": 0.5691, + "step": 920 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018273809523809524, + "loss": 0.6079, + "step": 921 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018293650793650793, + "loss": 0.6349, + "step": 922 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018313492063492065, + "loss": 0.5974, + "step": 923 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018333333333333334, + "loss": 0.7069, + "step": 924 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018353174603174602, + "loss": 0.6202, + "step": 925 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018373015873015874, + "loss": 0.6968, + "step": 926 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018392857142857143, + "loss": 0.5835, + "step": 927 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018412698412698412, + "loss": 0.5379, + "step": 928 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001843253968253968, + "loss": 0.5651, + "step": 929 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018452380952380955, + "loss": 0.5885, + "step": 930 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018472222222222224, + "loss": 0.5267, + "step": 931 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018492063492063493, + "loss": 0.6, + "step": 932 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018511904761904765, + "loss": 0.6612, + "step": 933 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018531746031746034, + "loss": 0.6257, + "step": 934 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018551587301587303, + "loss": 0.6391, + "step": 935 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018571428571428572, + "loss": 0.6365, + "step": 936 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018591269841269843, + "loss": 0.693, + "step": 937 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018611111111111112, + "loss": 0.5895, + "step": 938 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001863095238095238, + "loss": 0.73, + "step": 939 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018650793650793653, + "loss": 0.6081, + "step": 940 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018670634920634922, + "loss": 0.6191, + "step": 941 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001869047619047619, + "loss": 0.6074, + "step": 942 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018710317460317462, + "loss": 0.6819, + "step": 943 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018730158730158731, + "loss": 0.7019, + "step": 944 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001875, + "loss": 0.6181, + "step": 945 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001876984126984127, + "loss": 0.5908, + "step": 946 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001878968253968254, + "loss": 0.5829, + "step": 947 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001880952380952381, + "loss": 0.4679, + "step": 948 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001882936507936508, + "loss": 0.6063, + "step": 949 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001884920634920635, + "loss": 0.6484, + "step": 950 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001886904761904762, + "loss": 0.5964, + "step": 951 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018888888888888888, + "loss": 0.6028, + "step": 952 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001890873015873016, + "loss": 0.6103, + "step": 953 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001892857142857143, + "loss": 0.6986, + "step": 954 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018948412698412698, + "loss": 0.5723, + "step": 955 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001896825396825397, + "loss": 0.5318, + "step": 956 + }, + { + "epoch": 0.03, + "learning_rate": 0.00018988095238095239, + "loss": 0.5895, + "step": 957 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019007936507936508, + "loss": 0.5805, + "step": 958 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019027777777777776, + "loss": 0.6134, + "step": 959 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019047619047619048, + "loss": 0.6158, + "step": 960 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001906746031746032, + "loss": 0.5896, + "step": 961 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001908730158730159, + "loss": 0.5235, + "step": 962 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019107142857142858, + "loss": 0.5546, + "step": 963 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001912698412698413, + "loss": 0.573, + "step": 964 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019146825396825398, + "loss": 0.6564, + "step": 965 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019166666666666667, + "loss": 0.6111, + "step": 966 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001918650793650794, + "loss": 0.7892, + "step": 967 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019206349206349208, + "loss": 0.5685, + "step": 968 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019226190476190477, + "loss": 0.5836, + "step": 969 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019246031746031748, + "loss": 0.5234, + "step": 970 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019265873015873017, + "loss": 0.6071, + "step": 971 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019285714285714286, + "loss": 0.6011, + "step": 972 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019305555555555558, + "loss": 0.627, + "step": 973 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019325396825396827, + "loss": 0.5898, + "step": 974 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019345238095238096, + "loss": 0.5293, + "step": 975 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019365079365079365, + "loss": 0.5605, + "step": 976 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019384920634920637, + "loss": 0.5933, + "step": 977 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019404761904761905, + "loss": 0.6449, + "step": 978 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019424603174603174, + "loss": 0.6046, + "step": 979 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019444444444444446, + "loss": 0.5486, + "step": 980 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019464285714285715, + "loss": 0.5572, + "step": 981 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019484126984126984, + "loss": 0.5741, + "step": 982 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019503968253968256, + "loss": 0.629, + "step": 983 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019523809523809525, + "loss": 0.6463, + "step": 984 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019543650793650793, + "loss": 0.585, + "step": 985 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019563492063492062, + "loss": 0.5442, + "step": 986 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019583333333333334, + "loss": 0.5955, + "step": 987 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019603174603174603, + "loss": 0.5208, + "step": 988 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019623015873015872, + "loss": 0.5363, + "step": 989 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019642857142857144, + "loss": 0.5562, + "step": 990 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019662698412698413, + "loss": 0.6383, + "step": 991 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019682539682539682, + "loss": 0.6123, + "step": 992 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019702380952380953, + "loss": 0.5277, + "step": 993 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019722222222222225, + "loss": 0.5493, + "step": 994 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019742063492063494, + "loss": 0.6565, + "step": 995 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019761904761904763, + "loss": 0.6672, + "step": 996 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019781746031746034, + "loss": 0.4664, + "step": 997 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019801587301587303, + "loss": 0.5387, + "step": 998 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019821428571428572, + "loss": 0.6326, + "step": 999 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019841269841269844, + "loss": 0.6634, + "step": 1000 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019861111111111113, + "loss": 0.724, + "step": 1001 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019880952380952382, + "loss": 0.5684, + "step": 1002 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001990079365079365, + "loss": 0.6699, + "step": 1003 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019920634920634922, + "loss": 0.6414, + "step": 1004 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019940476190476191, + "loss": 0.5924, + "step": 1005 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001996031746031746, + "loss": 0.5732, + "step": 1006 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019980158730158732, + "loss": 0.5262, + "step": 1007 + }, + { + "epoch": 0.03, + "learning_rate": 0.0002, + "loss": 0.5515, + "step": 1008 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999999953511995, + "loss": 0.5045, + "step": 1009 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999999814047974, + "loss": 0.5457, + "step": 1010 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999999581607938, + "loss": 0.4904, + "step": 1011 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999999256191893, + "loss": 0.6328, + "step": 1012 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999998837799842, + "loss": 0.6102, + "step": 1013 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999999832643179, + "loss": 0.5151, + "step": 1014 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999997722087735, + "loss": 0.5901, + "step": 1015 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999997024767687, + "loss": 0.5247, + "step": 1016 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999996234471652, + "loss": 0.6967, + "step": 1017 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999995351199638, + "loss": 0.5498, + "step": 1018 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999994374951654, + "loss": 0.5841, + "step": 1019 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999993305727708, + "loss": 0.6728, + "step": 1020 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999999214352781, + "loss": 0.5868, + "step": 1021 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999990888351968, + "loss": 0.572, + "step": 1022 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999989540200197, + "loss": 0.6459, + "step": 1023 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999988099072508, + "loss": 0.6655, + "step": 1024 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999986564968919, + "loss": 0.5725, + "step": 1025 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999998493788944, + "loss": 0.5854, + "step": 1026 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999983217834085, + "loss": 0.607, + "step": 1027 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999998140480287, + "loss": 0.6451, + "step": 1028 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999997949879582, + "loss": 0.5616, + "step": 1029 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999977499812936, + "loss": 0.5841, + "step": 1030 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999975407854255, + "loss": 0.6826, + "step": 1031 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999973222919787, + "loss": 0.5931, + "step": 1032 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999997094500955, + "loss": 0.6044, + "step": 1033 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999996857412357, + "loss": 0.6371, + "step": 1034 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999996611026187, + "loss": 0.6728, + "step": 1035 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999996355342447, + "loss": 0.5446, + "step": 1036 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999960903611393, + "loss": 0.6269, + "step": 1037 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999958160822666, + "loss": 0.5906, + "step": 1038 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999955325058314, + "loss": 0.5561, + "step": 1039 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999952396318365, + "loss": 0.6141, + "step": 1040 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999994937460284, + "loss": 0.5774, + "step": 1041 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999946259911772, + "loss": 0.4953, + "step": 1042 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999994305224519, + "loss": 0.6045, + "step": 1043 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999939751603126, + "loss": 0.6856, + "step": 1044 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999936357985605, + "loss": 0.5648, + "step": 1045 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999932871392662, + "loss": 0.5774, + "step": 1046 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999929291824326, + "loss": 0.6383, + "step": 1047 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999925619280638, + "loss": 0.532, + "step": 1048 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999921853761622, + "loss": 0.7097, + "step": 1049 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999991799526732, + "loss": 0.5988, + "step": 1050 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999991404379777, + "loss": 0.5862, + "step": 1051 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999909999353, + "loss": 0.5181, + "step": 1052 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999905861933055, + "loss": 0.6629, + "step": 1053 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999990163153797, + "loss": 0.4974, + "step": 1054 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999897308167785, + "loss": 0.5314, + "step": 1055 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999989289182254, + "loss": 0.5985, + "step": 1056 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999888382502278, + "loss": 0.686, + "step": 1057 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999883780207037, + "loss": 0.6168, + "step": 1058 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999879084936864, + "loss": 0.5448, + "step": 1059 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999874296691797, + "loss": 0.5747, + "step": 1060 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999986941547189, + "loss": 0.7829, + "step": 1061 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999864441277182, + "loss": 0.69, + "step": 1062 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999985937410772, + "loss": 0.6053, + "step": 1063 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999854213963546, + "loss": 0.6596, + "step": 1064 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999848960844716, + "loss": 0.5911, + "step": 1065 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999843614751278, + "loss": 0.5897, + "step": 1066 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999838175683278, + "loss": 0.5096, + "step": 1067 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999832643640767, + "loss": 0.6648, + "step": 1068 + }, + { + "epoch": 0.03, + "learning_rate": 0.000199998270186238, + "loss": 0.5637, + "step": 1069 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999821300632425, + "loss": 0.5569, + "step": 1070 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999815489666698, + "loss": 0.4769, + "step": 1071 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999980958572667, + "loss": 0.6772, + "step": 1072 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999803588812402, + "loss": 0.5805, + "step": 1073 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999797498923943, + "loss": 0.6437, + "step": 1074 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999791316061355, + "loss": 0.5955, + "step": 1075 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999978504022469, + "loss": 0.4881, + "step": 1076 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999778671414008, + "loss": 0.536, + "step": 1077 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999772209629372, + "loss": 0.5491, + "step": 1078 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999976565487084, + "loss": 0.5371, + "step": 1079 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999759007138473, + "loss": 0.5053, + "step": 1080 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999975226643233, + "loss": 0.515, + "step": 1081 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999745432752478, + "loss": 0.6162, + "step": 1082 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999738506098976, + "loss": 0.6324, + "step": 1083 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999731486471892, + "loss": 0.5884, + "step": 1084 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999972437387129, + "loss": 0.5732, + "step": 1085 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999717168297235, + "loss": 0.5982, + "step": 1086 + }, + { + "epoch": 0.03, + "learning_rate": 0.000199997098697498, + "loss": 0.539, + "step": 1087 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999702478229044, + "loss": 0.6207, + "step": 1088 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999969499373504, + "loss": 0.6723, + "step": 1089 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999687416267858, + "loss": 0.6425, + "step": 1090 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999679745827568, + "loss": 0.7411, + "step": 1091 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999967198241424, + "loss": 0.6878, + "step": 1092 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999966412602795, + "loss": 0.5291, + "step": 1093 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999965617666877, + "loss": 0.6786, + "step": 1094 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999648134336767, + "loss": 0.5919, + "step": 1095 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999639999032022, + "loss": 0.6681, + "step": 1096 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999631770754612, + "loss": 0.5399, + "step": 1097 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999623449504613, + "loss": 0.5276, + "step": 1098 + }, + { + "epoch": 0.03, + "learning_rate": 0.000199996150352821, + "loss": 0.6136, + "step": 1099 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999606528087145, + "loss": 0.5849, + "step": 1100 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999959792791984, + "loss": 0.6269, + "step": 1101 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999958923478026, + "loss": 0.6232, + "step": 1102 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999958044866848, + "loss": 0.5595, + "step": 1103 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999957156958459, + "loss": 0.688, + "step": 1104 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999562597528667, + "loss": 0.6288, + "step": 1105 + }, + { + "epoch": 0.03, + "learning_rate": 0.000199995535325008, + "loss": 0.5841, + "step": 1106 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999544374501064, + "loss": 0.4862, + "step": 1107 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999535123529553, + "loss": 0.6298, + "step": 1108 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999525779586353, + "loss": 0.6231, + "step": 1109 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999516342671542, + "loss": 0.4857, + "step": 1110 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999506812785217, + "loss": 0.6009, + "step": 1111 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999949718992746, + "loss": 0.6217, + "step": 1112 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999487474098366, + "loss": 0.5896, + "step": 1113 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999477665298023, + "loss": 0.6084, + "step": 1114 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999946776352652, + "loss": 0.5634, + "step": 1115 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999457768783954, + "loss": 0.5738, + "step": 1116 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999447681070414, + "loss": 0.5742, + "step": 1117 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999437500385994, + "loss": 0.6251, + "step": 1118 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999427226730793, + "loss": 0.6586, + "step": 1119 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999416860104897, + "loss": 0.5637, + "step": 1120 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999406400508413, + "loss": 0.675, + "step": 1121 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999939584794143, + "loss": 0.5559, + "step": 1122 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999385202404052, + "loss": 0.5663, + "step": 1123 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999374463896375, + "loss": 0.6252, + "step": 1124 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999363632418503, + "loss": 0.6239, + "step": 1125 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999935270797053, + "loss": 0.5586, + "step": 1126 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999341690552563, + "loss": 0.604, + "step": 1127 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999330580164698, + "loss": 0.6043, + "step": 1128 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999319376807047, + "loss": 0.5674, + "step": 1129 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999308080479708, + "loss": 0.5542, + "step": 1130 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999296691182789, + "loss": 0.5676, + "step": 1131 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999285208916392, + "loss": 0.5686, + "step": 1132 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999927363368063, + "loss": 0.515, + "step": 1133 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999261965475602, + "loss": 0.6167, + "step": 1134 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999250204301425, + "loss": 0.5897, + "step": 1135 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999238350158205, + "loss": 0.5813, + "step": 1136 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999226403046052, + "loss": 0.5581, + "step": 1137 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999214362965075, + "loss": 0.5724, + "step": 1138 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999202229915392, + "loss": 0.5602, + "step": 1139 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999919000389711, + "loss": 0.5291, + "step": 1140 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999177684910344, + "loss": 0.6849, + "step": 1141 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999165272955208, + "loss": 0.5547, + "step": 1142 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999152768031823, + "loss": 0.5685, + "step": 1143 + }, + { + "epoch": 0.03, + "learning_rate": 0.000199991401701403, + "loss": 0.6087, + "step": 1144 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999127479280752, + "loss": 0.6408, + "step": 1145 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999114695453305, + "loss": 0.5035, + "step": 1146 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999101818658074, + "loss": 0.5746, + "step": 1147 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999088848895183, + "loss": 0.5323, + "step": 1148 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999075786164745, + "loss": 0.6158, + "step": 1149 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999062630466887, + "loss": 0.5559, + "step": 1150 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999904938180173, + "loss": 0.7457, + "step": 1151 + }, + { + "epoch": 0.03, + "learning_rate": 0.000199990360401694, + "loss": 0.5598, + "step": 1152 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019999022605570013, + "loss": 0.6346, + "step": 1153 + }, + { + "epoch": 0.03, + "learning_rate": 0.000199990090780037, + "loss": 0.6526, + "step": 1154 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019998995457470587, + "loss": 0.7168, + "step": 1155 + }, + { + "epoch": 0.03, + "learning_rate": 0.000199989817439708, + "loss": 0.6392, + "step": 1156 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019998967937504464, + "loss": 0.558, + "step": 1157 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019998954038071714, + "loss": 0.5919, + "step": 1158 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999894004567267, + "loss": 0.619, + "step": 1159 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019998925960307469, + "loss": 0.5798, + "step": 1160 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019998911781976238, + "loss": 0.486, + "step": 1161 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999889751067911, + "loss": 0.6392, + "step": 1162 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999888314641622, + "loss": 0.558, + "step": 1163 + }, + { + "epoch": 0.03, + "learning_rate": 0.000199988686891877, + "loss": 0.8221, + "step": 1164 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019998854138993682, + "loss": 0.5741, + "step": 1165 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019998839495834303, + "loss": 0.5494, + "step": 1166 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019998824759709702, + "loss": 0.5541, + "step": 1167 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019998809930620013, + "loss": 0.6498, + "step": 1168 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019998795008565373, + "loss": 0.6463, + "step": 1169 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019998779993545923, + "loss": 0.6018, + "step": 1170 + }, + { + "epoch": 0.03, + "learning_rate": 0.000199987648855618, + "loss": 0.5551, + "step": 1171 + }, + { + "epoch": 0.03, + "learning_rate": 0.0001999874968461315, + "loss": 0.5462, + "step": 1172 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019998734390700107, + "loss": 0.5956, + "step": 1173 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019998719003822818, + "loss": 0.6861, + "step": 1174 + }, + { + "epoch": 0.03, + "learning_rate": 0.00019998703523981424, + "loss": 0.6185, + "step": 1175 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999868795117607, + "loss": 0.6514, + "step": 1176 + }, + { + "epoch": 0.04, + "learning_rate": 0.000199986722854069, + "loss": 0.5903, + "step": 1177 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998656526674062, + "loss": 0.6463, + "step": 1178 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998640674977701, + "loss": 0.5655, + "step": 1179 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998624730317962, + "loss": 0.4932, + "step": 1180 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998608692694996, + "loss": 0.4594, + "step": 1181 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998592562108953, + "loss": 0.5525, + "step": 1182 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998576338559981, + "loss": 0.7135, + "step": 1183 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999856002204823, + "loss": 0.579, + "step": 1184 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998543612573851, + "loss": 0.5929, + "step": 1185 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998527110137002, + "loss": 0.6591, + "step": 1186 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998510514737832, + "loss": 0.5977, + "step": 1187 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998493826376497, + "loss": 0.6184, + "step": 1188 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999847704505315, + "loss": 0.4669, + "step": 1189 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999846017076795, + "loss": 0.5019, + "step": 1190 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998443203521048, + "loss": 0.5031, + "step": 1191 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998426143312613, + "loss": 0.563, + "step": 1192 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998408990142788, + "loss": 0.6605, + "step": 1193 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999839174401175, + "loss": 0.5507, + "step": 1194 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998374404919645, + "loss": 0.6116, + "step": 1195 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999835697286664, + "loss": 0.4716, + "step": 1196 + }, + { + "epoch": 0.04, + "learning_rate": 0.000199983394478529, + "loss": 0.6183, + "step": 1197 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998321829878583, + "loss": 0.6269, + "step": 1198 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998304118943856, + "loss": 0.618, + "step": 1199 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999828631504888, + "loss": 0.5412, + "step": 1200 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998268418193823, + "loss": 0.5322, + "step": 1201 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998250428378853, + "loss": 0.553, + "step": 1202 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998232345604133, + "loss": 0.5958, + "step": 1203 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998214169869835, + "loss": 0.5913, + "step": 1204 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998195901176126, + "loss": 0.5723, + "step": 1205 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998177539523178, + "loss": 0.5919, + "step": 1206 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998159084911156, + "loss": 0.6463, + "step": 1207 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999814053734024, + "loss": 0.5064, + "step": 1208 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998121896810597, + "loss": 0.5445, + "step": 1209 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998103163322402, + "loss": 0.6226, + "step": 1210 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998084336875826, + "loss": 0.5773, + "step": 1211 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998065417471052, + "loss": 0.5766, + "step": 1212 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998046405108244, + "loss": 0.6426, + "step": 1213 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019998027299787587, + "loss": 0.7223, + "step": 1214 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999800810150926, + "loss": 0.6018, + "step": 1215 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997988810273437, + "loss": 0.4709, + "step": 1216 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997969426080299, + "loss": 0.5783, + "step": 1217 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997949948930024, + "loss": 0.5161, + "step": 1218 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997930378822795, + "loss": 0.7237, + "step": 1219 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997910715758798, + "loss": 0.4558, + "step": 1220 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999789095973821, + "loss": 0.6384, + "step": 1221 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997871110761218, + "loss": 0.5598, + "step": 1222 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997851168828, + "loss": 0.6803, + "step": 1223 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997831133938753, + "loss": 0.6274, + "step": 1224 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997811006093654, + "loss": 0.6571, + "step": 1225 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997790785292893, + "loss": 0.4962, + "step": 1226 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997770471536657, + "loss": 0.5287, + "step": 1227 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997750064825137, + "loss": 0.5585, + "step": 1228 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997729565158522, + "loss": 0.5531, + "step": 1229 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997708972537, + "loss": 0.6471, + "step": 1230 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999768828696077, + "loss": 0.5641, + "step": 1231 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997667508430013, + "loss": 0.7049, + "step": 1232 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999764663694493, + "loss": 0.6877, + "step": 1233 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997625672505716, + "loss": 0.5167, + "step": 1234 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997604615112563, + "loss": 0.6515, + "step": 1235 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997583464765662, + "loss": 0.5242, + "step": 1236 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999756222146522, + "loss": 0.7035, + "step": 1237 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997540885211427, + "loss": 0.506, + "step": 1238 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999751945600448, + "loss": 0.5874, + "step": 1239 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997497933844588, + "loss": 0.5592, + "step": 1240 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999747631873194, + "loss": 0.6121, + "step": 1241 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997454610666746, + "loss": 0.5845, + "step": 1242 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997432809649198, + "loss": 0.5718, + "step": 1243 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997410915679507, + "loss": 0.6041, + "step": 1244 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997388928757876, + "loss": 0.5654, + "step": 1245 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997366848884505, + "loss": 0.58, + "step": 1246 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997344676059604, + "loss": 0.5552, + "step": 1247 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997322410283375, + "loss": 0.4936, + "step": 1248 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997300051556027, + "loss": 0.5468, + "step": 1249 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997277599877766, + "loss": 0.5798, + "step": 1250 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997255055248804, + "loss": 0.6448, + "step": 1251 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999723241766935, + "loss": 0.6326, + "step": 1252 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999720968713961, + "loss": 0.5859, + "step": 1253 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997186863659802, + "loss": 0.5162, + "step": 1254 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997163947230135, + "loss": 0.5658, + "step": 1255 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999714093785082, + "loss": 0.6211, + "step": 1256 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997117835522077, + "loss": 0.5797, + "step": 1257 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997094640244111, + "loss": 0.5959, + "step": 1258 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997071352017147, + "loss": 0.6242, + "step": 1259 + }, + { + "epoch": 0.04, + "learning_rate": 0.000199970479708414, + "loss": 0.5642, + "step": 1260 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997024496717081, + "loss": 0.707, + "step": 1261 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019997000929644415, + "loss": 0.6269, + "step": 1262 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999697726962362, + "loss": 0.5817, + "step": 1263 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999695351665491, + "loss": 0.6598, + "step": 1264 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996929670738515, + "loss": 0.6285, + "step": 1265 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996905731874655, + "loss": 0.5498, + "step": 1266 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996881700063543, + "loss": 0.6222, + "step": 1267 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999685757530541, + "loss": 0.6523, + "step": 1268 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996833357600484, + "loss": 0.6494, + "step": 1269 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996809046948983, + "loss": 0.612, + "step": 1270 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999678464335114, + "loss": 0.515, + "step": 1271 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996760146807172, + "loss": 0.5516, + "step": 1272 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996735557317315, + "loss": 0.5781, + "step": 1273 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996710874881793, + "loss": 0.6455, + "step": 1274 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999668609950084, + "loss": 0.5295, + "step": 1275 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996661231174681, + "loss": 0.5797, + "step": 1276 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996636269903554, + "loss": 0.6137, + "step": 1277 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996611215687687, + "loss": 0.6307, + "step": 1278 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999658606852731, + "loss": 0.4995, + "step": 1279 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996560828422665, + "loss": 0.6466, + "step": 1280 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996535495373977, + "loss": 0.5215, + "step": 1281 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996510069381486, + "loss": 0.5882, + "step": 1282 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999648455044543, + "loss": 0.5642, + "step": 1283 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996458938566046, + "loss": 0.6785, + "step": 1284 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999643323374357, + "loss": 0.6071, + "step": 1285 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996407435978242, + "loss": 0.6577, + "step": 1286 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996381545270303, + "loss": 0.6633, + "step": 1287 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999635556161999, + "loss": 0.5746, + "step": 1288 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996329485027548, + "loss": 0.6054, + "step": 1289 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996303315493222, + "loss": 0.6999, + "step": 1290 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996277053017247, + "loss": 0.6401, + "step": 1291 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996250697599873, + "loss": 0.5985, + "step": 1292 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996224249241347, + "loss": 0.5909, + "step": 1293 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999619770794191, + "loss": 0.6682, + "step": 1294 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999617107370181, + "loss": 0.645, + "step": 1295 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996144346521298, + "loss": 0.631, + "step": 1296 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996117526400618, + "loss": 0.6604, + "step": 1297 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996090613340025, + "loss": 0.5792, + "step": 1298 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996063607339762, + "loss": 0.5665, + "step": 1299 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019996036508400082, + "loss": 0.5442, + "step": 1300 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999600931652124, + "loss": 0.5258, + "step": 1301 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995982031703488, + "loss": 0.6102, + "step": 1302 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995954653947079, + "loss": 0.6532, + "step": 1303 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995927183252268, + "loss": 0.5349, + "step": 1304 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995899619619308, + "loss": 0.5727, + "step": 1305 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999587196304846, + "loss": 0.6453, + "step": 1306 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995844213539977, + "loss": 0.6424, + "step": 1307 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999581637109412, + "loss": 0.5657, + "step": 1308 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995788435711145, + "loss": 0.5143, + "step": 1309 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995760407391313, + "loss": 0.5346, + "step": 1310 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995732286134886, + "loss": 0.5565, + "step": 1311 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999570407194212, + "loss": 0.5735, + "step": 1312 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995675764813287, + "loss": 0.5779, + "step": 1313 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999564736474864, + "loss": 0.6248, + "step": 1314 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999561887174845, + "loss": 0.6568, + "step": 1315 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999559028581298, + "loss": 0.5324, + "step": 1316 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995561606942495, + "loss": 0.51, + "step": 1317 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999553283513726, + "loss": 0.5517, + "step": 1318 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995503970397546, + "loss": 0.6756, + "step": 1319 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995475012723622, + "loss": 0.4792, + "step": 1320 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995445962115753, + "loss": 0.5206, + "step": 1321 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999541681857421, + "loss": 0.5483, + "step": 1322 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995387582099266, + "loss": 0.5358, + "step": 1323 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995358252691196, + "loss": 0.6748, + "step": 1324 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995328830350263, + "loss": 0.6304, + "step": 1325 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999529931507675, + "loss": 0.6369, + "step": 1326 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995269706870925, + "loss": 0.6025, + "step": 1327 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999524000573307, + "loss": 0.5978, + "step": 1328 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995210211663452, + "loss": 0.6438, + "step": 1329 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995180324662356, + "loss": 0.6372, + "step": 1330 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995150344730058, + "loss": 0.4669, + "step": 1331 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995120271866835, + "loss": 0.5718, + "step": 1332 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995090106072967, + "loss": 0.6571, + "step": 1333 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995059847348733, + "loss": 0.6807, + "step": 1334 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019995029495694419, + "loss": 0.6282, + "step": 1335 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994999051110304, + "loss": 0.554, + "step": 1336 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999496851359667, + "loss": 0.6597, + "step": 1337 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994937883153804, + "loss": 0.6308, + "step": 1338 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994907159781987, + "loss": 0.5824, + "step": 1339 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994876343481507, + "loss": 0.5995, + "step": 1340 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994845434252652, + "loss": 0.6179, + "step": 1341 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994814432095707, + "loss": 0.4733, + "step": 1342 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994783337010962, + "loss": 0.5902, + "step": 1343 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994752148998703, + "loss": 0.6278, + "step": 1344 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994720868059224, + "loss": 0.6791, + "step": 1345 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994689494192813, + "loss": 0.5432, + "step": 1346 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999465802739976, + "loss": 0.5676, + "step": 1347 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994626467680364, + "loss": 0.6448, + "step": 1348 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994594815034912, + "loss": 0.4989, + "step": 1349 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994563069463704, + "loss": 0.5898, + "step": 1350 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999453123096703, + "loss": 0.606, + "step": 1351 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999449929954519, + "loss": 0.7287, + "step": 1352 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994467275198473, + "loss": 0.6514, + "step": 1353 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999443515792719, + "loss": 0.5411, + "step": 1354 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994402947731628, + "loss": 0.5043, + "step": 1355 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994370644612093, + "loss": 0.6207, + "step": 1356 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999433824856888, + "loss": 0.6469, + "step": 1357 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994305759602296, + "loss": 0.5157, + "step": 1358 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994273177712638, + "loss": 0.5696, + "step": 1359 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994240502900214, + "loss": 0.618, + "step": 1360 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994207735165326, + "loss": 0.6865, + "step": 1361 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994174874508276, + "loss": 0.722, + "step": 1362 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999414192092937, + "loss": 0.5479, + "step": 1363 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994108874428915, + "loss": 0.654, + "step": 1364 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994075735007224, + "loss": 0.6394, + "step": 1365 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994042502664594, + "loss": 0.6225, + "step": 1366 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019994009177401342, + "loss": 0.674, + "step": 1367 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999397575921778, + "loss": 0.557, + "step": 1368 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999394224811421, + "loss": 0.7401, + "step": 1369 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999390864409095, + "loss": 0.6742, + "step": 1370 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999387494714831, + "loss": 0.5504, + "step": 1371 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993841157286605, + "loss": 0.6375, + "step": 1372 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993807274506145, + "loss": 0.6427, + "step": 1373 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999377329880725, + "loss": 0.5865, + "step": 1374 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993739230190233, + "loss": 0.5825, + "step": 1375 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993705068655415, + "loss": 0.6928, + "step": 1376 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999367081420311, + "loss": 0.6407, + "step": 1377 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993636466833633, + "loss": 0.6313, + "step": 1378 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993602026547312, + "loss": 0.4961, + "step": 1379 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993567493344457, + "loss": 0.6562, + "step": 1380 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993532867225398, + "loss": 0.747, + "step": 1381 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993498148190454, + "loss": 0.5555, + "step": 1382 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993463336239945, + "loss": 0.6189, + "step": 1383 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993428431374198, + "loss": 0.614, + "step": 1384 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993393433593535, + "loss": 0.5858, + "step": 1385 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993358342898284, + "loss": 0.6125, + "step": 1386 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993323159288773, + "loss": 0.5726, + "step": 1387 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993287882765322, + "loss": 0.5929, + "step": 1388 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993252513328264, + "loss": 0.5079, + "step": 1389 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993217050977928, + "loss": 0.6168, + "step": 1390 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993181495714642, + "loss": 0.537, + "step": 1391 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993145847538738, + "loss": 0.5658, + "step": 1392 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993110106450549, + "loss": 0.6164, + "step": 1393 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993074272450402, + "loss": 0.5147, + "step": 1394 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993038345538633, + "loss": 0.6507, + "step": 1395 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019993002325715574, + "loss": 0.5823, + "step": 1396 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999296621298157, + "loss": 0.5601, + "step": 1397 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992930007336942, + "loss": 0.6477, + "step": 1398 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992893708782033, + "loss": 0.6218, + "step": 1399 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999285731731719, + "loss": 0.5202, + "step": 1400 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992820832942735, + "loss": 0.6217, + "step": 1401 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992784255659014, + "loss": 0.6527, + "step": 1402 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999274758546637, + "loss": 0.6386, + "step": 1403 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992710822365142, + "loss": 0.6416, + "step": 1404 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999267396635567, + "loss": 0.6286, + "step": 1405 + }, + { + "epoch": 0.04, + "learning_rate": 0.000199926370174383, + "loss": 0.5478, + "step": 1406 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999259997561337, + "loss": 0.5328, + "step": 1407 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992562840881234, + "loss": 0.6828, + "step": 1408 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992525613242229, + "loss": 0.5267, + "step": 1409 + }, + { + "epoch": 0.04, + "learning_rate": 0.000199924882926967, + "loss": 0.5324, + "step": 1410 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992450879245002, + "loss": 0.6607, + "step": 1411 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992413372887474, + "loss": 0.5727, + "step": 1412 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992375773624472, + "loss": 0.5513, + "step": 1413 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992338081456344, + "loss": 0.6382, + "step": 1414 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992300296383438, + "loss": 0.4833, + "step": 1415 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992262418406105, + "loss": 0.5509, + "step": 1416 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992224447524696, + "loss": 0.548, + "step": 1417 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992186383739572, + "loss": 0.6084, + "step": 1418 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992148227051078, + "loss": 0.6556, + "step": 1419 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992109977459574, + "loss": 0.5648, + "step": 1420 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992071634965415, + "loss": 0.6856, + "step": 1421 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019992033199568954, + "loss": 0.5619, + "step": 1422 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999199467127055, + "loss": 0.66, + "step": 1423 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991956050070564, + "loss": 0.5907, + "step": 1424 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991917335969355, + "loss": 0.5533, + "step": 1425 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991878528967277, + "loss": 0.5304, + "step": 1426 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991839629064696, + "loss": 0.6456, + "step": 1427 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991800636261973, + "loss": 0.7233, + "step": 1428 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991761550559468, + "loss": 0.5736, + "step": 1429 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991722371957548, + "loss": 0.6957, + "step": 1430 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991683100456577, + "loss": 0.5643, + "step": 1431 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991643736056916, + "loss": 0.6556, + "step": 1432 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991604278758938, + "loss": 0.6916, + "step": 1433 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991564728563002, + "loss": 0.5869, + "step": 1434 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991525085469482, + "loss": 0.4792, + "step": 1435 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991485349478742, + "loss": 0.5998, + "step": 1436 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991445520591155, + "loss": 0.6035, + "step": 1437 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991405598807088, + "loss": 0.533, + "step": 1438 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991365584126915, + "loss": 0.6104, + "step": 1439 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991325476551006, + "loss": 0.5543, + "step": 1440 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991285276079737, + "loss": 0.569, + "step": 1441 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991244982713478, + "loss": 0.5537, + "step": 1442 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991204596452606, + "loss": 0.5791, + "step": 1443 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991164117297495, + "loss": 0.5747, + "step": 1444 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999112354524852, + "loss": 0.6068, + "step": 1445 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019991082880306063, + "loss": 0.5996, + "step": 1446 + }, + { + "epoch": 0.04, + "learning_rate": 0.000199910421224705, + "loss": 0.6229, + "step": 1447 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999100127174221, + "loss": 0.5121, + "step": 1448 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999096032812157, + "loss": 0.5279, + "step": 1449 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019990919291608964, + "loss": 0.555, + "step": 1450 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999087816220477, + "loss": 0.6329, + "step": 1451 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019990836939909375, + "loss": 0.6976, + "step": 1452 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999079562472316, + "loss": 0.5401, + "step": 1453 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999075421664651, + "loss": 0.4261, + "step": 1454 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999071271567981, + "loss": 0.5984, + "step": 1455 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019990671121823442, + "loss": 0.5615, + "step": 1456 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019990629435077796, + "loss": 0.547, + "step": 1457 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999058765544326, + "loss": 0.5617, + "step": 1458 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019990545782920223, + "loss": 0.6577, + "step": 1459 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019990503817509075, + "loss": 0.5689, + "step": 1460 + }, + { + "epoch": 0.04, + "learning_rate": 0.000199904617592102, + "loss": 0.7218, + "step": 1461 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019990419608023995, + "loss": 0.5222, + "step": 1462 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999037736395085, + "loss": 0.6122, + "step": 1463 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999033502699116, + "loss": 0.6446, + "step": 1464 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019990292597145314, + "loss": 0.6262, + "step": 1465 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999025007441371, + "loss": 0.5413, + "step": 1466 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019990207458796743, + "loss": 0.6193, + "step": 1467 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999016475029481, + "loss": 0.5707, + "step": 1468 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019990121948908307, + "loss": 0.5924, + "step": 1469 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019990079054637628, + "loss": 0.5735, + "step": 1470 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001999003606748318, + "loss": 0.5879, + "step": 1471 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019989992987445355, + "loss": 0.6826, + "step": 1472 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019989949814524558, + "loss": 0.5381, + "step": 1473 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001998990654872119, + "loss": 0.5326, + "step": 1474 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001998986319003565, + "loss": 0.6549, + "step": 1475 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019989819738468344, + "loss": 0.6103, + "step": 1476 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001998977619401968, + "loss": 0.6572, + "step": 1477 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019989732556690055, + "loss": 0.5619, + "step": 1478 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019989688826479876, + "loss": 0.6602, + "step": 1479 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019989645003389556, + "loss": 0.5756, + "step": 1480 + }, + { + "epoch": 0.04, + "learning_rate": 0.000199896010874195, + "loss": 0.5468, + "step": 1481 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001998955707857011, + "loss": 0.5816, + "step": 1482 + }, + { + "epoch": 0.04, + "learning_rate": 0.000199895129768418, + "loss": 0.5348, + "step": 1483 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019989468782234982, + "loss": 0.5547, + "step": 1484 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019989424494750062, + "loss": 0.5692, + "step": 1485 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019989380114387455, + "loss": 0.5693, + "step": 1486 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019989335641147575, + "loss": 0.5489, + "step": 1487 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019989291075030833, + "loss": 0.5596, + "step": 1488 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019989246416037644, + "loss": 0.6997, + "step": 1489 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019989201664168422, + "loss": 0.5573, + "step": 1490 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019989156819423586, + "loss": 0.7969, + "step": 1491 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001998911188180355, + "loss": 0.5582, + "step": 1492 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019989066851308732, + "loss": 0.6123, + "step": 1493 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019989021727939552, + "loss": 0.6022, + "step": 1494 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001998897651169643, + "loss": 0.6203, + "step": 1495 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019988931202579786, + "loss": 0.5024, + "step": 1496 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001998888580059004, + "loss": 0.5379, + "step": 1497 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019988840305727617, + "loss": 0.642, + "step": 1498 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019988794717992937, + "loss": 0.6107, + "step": 1499 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001998874903738642, + "loss": 0.4509, + "step": 1500 + }, + { + "epoch": 0.04, + "learning_rate": 0.000199887032639085, + "loss": 0.5519, + "step": 1501 + }, + { + "epoch": 0.04, + "learning_rate": 0.000199886573975596, + "loss": 0.6126, + "step": 1502 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001998861143834014, + "loss": 0.5728, + "step": 1503 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019988565386250558, + "loss": 0.6099, + "step": 1504 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001998851924129127, + "loss": 0.4713, + "step": 1505 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019988473003462715, + "loss": 0.7062, + "step": 1506 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019988426672765317, + "loss": 0.5164, + "step": 1507 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001998838024919951, + "loss": 0.5318, + "step": 1508 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019988333732765722, + "loss": 0.62, + "step": 1509 + }, + { + "epoch": 0.04, + "learning_rate": 0.0001998828712346439, + "loss": 0.6251, + "step": 1510 + }, + { + "epoch": 0.04, + "learning_rate": 0.00019988240421295946, + "loss": 0.628, + "step": 1511 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019988193626260822, + "loss": 0.5658, + "step": 1512 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019988146738359456, + "loss": 0.595, + "step": 1513 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019988099757592282, + "loss": 0.6072, + "step": 1514 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019988052683959736, + "loss": 0.567, + "step": 1515 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998800551746226, + "loss": 0.6016, + "step": 1516 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019987958258100286, + "loss": 0.5776, + "step": 1517 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998791090587426, + "loss": 0.6595, + "step": 1518 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019987863460784616, + "loss": 0.5473, + "step": 1519 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019987815922831802, + "loss": 0.5631, + "step": 1520 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998776829201625, + "loss": 0.5145, + "step": 1521 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019987720568338414, + "loss": 0.5935, + "step": 1522 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019987672751798732, + "loss": 0.6001, + "step": 1523 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998762484239765, + "loss": 0.5005, + "step": 1524 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019987576840135613, + "loss": 0.6005, + "step": 1525 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019987528745013067, + "loss": 0.6034, + "step": 1526 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019987480557030459, + "loss": 0.5043, + "step": 1527 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019987432276188236, + "loss": 0.5102, + "step": 1528 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998738390248685, + "loss": 0.6204, + "step": 1529 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998733543592675, + "loss": 0.5449, + "step": 1530 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019987286876508385, + "loss": 0.6001, + "step": 1531 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019987238224232205, + "loss": 0.5295, + "step": 1532 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019987189479098668, + "loss": 0.6066, + "step": 1533 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998714064110822, + "loss": 0.6494, + "step": 1534 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019987091710261322, + "loss": 0.5642, + "step": 1535 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019987042686558424, + "loss": 0.5141, + "step": 1536 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019986993569999983, + "loss": 0.463, + "step": 1537 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019986944360586456, + "loss": 0.634, + "step": 1538 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019986895058318304, + "loss": 0.6091, + "step": 1539 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998684566319598, + "loss": 0.4564, + "step": 1540 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019986796175219948, + "loss": 0.695, + "step": 1541 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998674659439066, + "loss": 0.6529, + "step": 1542 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019986696920708586, + "loss": 0.7167, + "step": 1543 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019986647154174184, + "loss": 0.6709, + "step": 1544 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019986597294787916, + "loss": 0.5405, + "step": 1545 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998654734255025, + "loss": 0.6748, + "step": 1546 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019986497297461644, + "loss": 0.6577, + "step": 1547 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019986447159522566, + "loss": 0.5656, + "step": 1548 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019986396928733485, + "loss": 0.491, + "step": 1549 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019986346605094861, + "loss": 0.6824, + "step": 1550 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998629618860717, + "loss": 0.5844, + "step": 1551 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019986245679270875, + "loss": 0.6988, + "step": 1552 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998619507708645, + "loss": 0.5006, + "step": 1553 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019986144382054365, + "loss": 0.6376, + "step": 1554 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019986093594175086, + "loss": 0.6193, + "step": 1555 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998604271344909, + "loss": 0.7169, + "step": 1556 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019985991739876848, + "loss": 0.5622, + "step": 1557 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019985940673458835, + "loss": 0.6155, + "step": 1558 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998588951419553, + "loss": 0.4988, + "step": 1559 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019985838262087397, + "loss": 0.5518, + "step": 1560 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019985786917134927, + "loss": 0.5783, + "step": 1561 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019985735479338584, + "loss": 0.6504, + "step": 1562 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019985683948698854, + "loss": 0.5917, + "step": 1563 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019985632325216217, + "loss": 0.5266, + "step": 1564 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019985580608891147, + "loss": 0.5114, + "step": 1565 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019985528799724135, + "loss": 0.4645, + "step": 1566 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998547689771565, + "loss": 0.5969, + "step": 1567 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998542490286618, + "loss": 0.5125, + "step": 1568 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998537281517621, + "loss": 0.5635, + "step": 1569 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019985320634646227, + "loss": 0.6321, + "step": 1570 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998526836127671, + "loss": 0.535, + "step": 1571 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019985215995068148, + "loss": 0.6338, + "step": 1572 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019985163536021027, + "loss": 0.5268, + "step": 1573 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019985110984135836, + "loss": 0.5673, + "step": 1574 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019985058339413062, + "loss": 0.6334, + "step": 1575 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019985005601853194, + "loss": 0.6453, + "step": 1576 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019984952771456724, + "loss": 0.6679, + "step": 1577 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019984899848224144, + "loss": 0.5795, + "step": 1578 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019984846832155943, + "loss": 0.6101, + "step": 1579 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019984793723252617, + "loss": 0.5667, + "step": 1580 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019984740521514658, + "loss": 0.5645, + "step": 1581 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019984687226942563, + "loss": 0.6618, + "step": 1582 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019984633839536823, + "loss": 0.5643, + "step": 1583 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019984580359297939, + "loss": 0.5142, + "step": 1584 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019984526786226406, + "loss": 0.5775, + "step": 1585 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019984473120322722, + "loss": 0.4924, + "step": 1586 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019984419361587384, + "loss": 0.7076, + "step": 1587 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019984365510020897, + "loss": 0.5998, + "step": 1588 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998431156562376, + "loss": 0.6029, + "step": 1589 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019984257528396474, + "loss": 0.6857, + "step": 1590 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019984203398339535, + "loss": 0.6259, + "step": 1591 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019984149175453455, + "loss": 0.6361, + "step": 1592 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998409485973874, + "loss": 0.6015, + "step": 1593 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019984040451195885, + "loss": 0.5314, + "step": 1594 + }, + { + "epoch": 0.05, + "learning_rate": 0.000199839859498254, + "loss": 0.5792, + "step": 1595 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019983931355627797, + "loss": 0.5409, + "step": 1596 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019983876668603576, + "loss": 0.6142, + "step": 1597 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998382188875325, + "loss": 0.6959, + "step": 1598 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019983767016077327, + "loss": 0.514, + "step": 1599 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998371205057632, + "loss": 0.5975, + "step": 1600 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019983656992250736, + "loss": 0.6602, + "step": 1601 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019983601841101086, + "loss": 0.588, + "step": 1602 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019983546597127888, + "loss": 0.4904, + "step": 1603 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019983491260331654, + "loss": 0.6091, + "step": 1604 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019983435830712892, + "loss": 0.5866, + "step": 1605 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998338030827213, + "loss": 0.5324, + "step": 1606 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998332469300987, + "loss": 0.5033, + "step": 1607 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998326898492664, + "loss": 0.6192, + "step": 1608 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998321318402295, + "loss": 0.62, + "step": 1609 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019983157290299328, + "loss": 0.5408, + "step": 1610 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019983101303756287, + "loss": 0.6408, + "step": 1611 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019983045224394346, + "loss": 0.5206, + "step": 1612 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019982989052214032, + "loss": 0.5738, + "step": 1613 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019982932787215862, + "loss": 0.6248, + "step": 1614 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019982876429400364, + "loss": 0.6046, + "step": 1615 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019982819978768058, + "loss": 0.504, + "step": 1616 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019982763435319472, + "loss": 0.5549, + "step": 1617 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998270679905513, + "loss": 0.6409, + "step": 1618 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998265006997556, + "loss": 0.5327, + "step": 1619 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998259324808129, + "loss": 0.5878, + "step": 1620 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998253633337284, + "loss": 0.633, + "step": 1621 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019982479325850752, + "loss": 0.5452, + "step": 1622 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998242222551555, + "loss": 0.5847, + "step": 1623 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998236503236776, + "loss": 0.5837, + "step": 1624 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998230774640792, + "loss": 0.6221, + "step": 1625 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019982250367636564, + "loss": 0.5767, + "step": 1626 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019982192896054218, + "loss": 0.5896, + "step": 1627 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019982135331661426, + "loss": 0.5838, + "step": 1628 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019982077674458716, + "loss": 0.6533, + "step": 1629 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019982019924446624, + "loss": 0.6048, + "step": 1630 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019981962081625691, + "loss": 0.5234, + "step": 1631 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019981904145996456, + "loss": 0.4747, + "step": 1632 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998184611755945, + "loss": 0.5374, + "step": 1633 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998178799631522, + "loss": 0.5129, + "step": 1634 + }, + { + "epoch": 0.05, + "learning_rate": 0.000199817297822643, + "loss": 0.5565, + "step": 1635 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998167147540724, + "loss": 0.408, + "step": 1636 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019981613075744574, + "loss": 0.5661, + "step": 1637 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998155458327685, + "loss": 0.5275, + "step": 1638 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019981495998004607, + "loss": 0.5526, + "step": 1639 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019981437319928393, + "loss": 0.5223, + "step": 1640 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019981378549048753, + "loss": 0.5047, + "step": 1641 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019981319685366232, + "loss": 0.5671, + "step": 1642 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019981260728881383, + "loss": 0.6015, + "step": 1643 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019981201679594748, + "loss": 0.5897, + "step": 1644 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019981142537506876, + "loss": 0.5061, + "step": 1645 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998108330261832, + "loss": 0.5071, + "step": 1646 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998102397492963, + "loss": 0.572, + "step": 1647 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019980964554441356, + "loss": 0.546, + "step": 1648 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019980905041154054, + "loss": 0.6038, + "step": 1649 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019980845435068272, + "loss": 0.5543, + "step": 1650 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998078573618457, + "loss": 0.6216, + "step": 1651 + }, + { + "epoch": 0.05, + "learning_rate": 0.000199807259445035, + "loss": 0.6434, + "step": 1652 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019980666060025616, + "loss": 0.5498, + "step": 1653 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019980606082751476, + "loss": 0.4759, + "step": 1654 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998054601268164, + "loss": 0.5543, + "step": 1655 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019980485849816667, + "loss": 0.6546, + "step": 1656 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019980425594157112, + "loss": 0.603, + "step": 1657 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001998036524570354, + "loss": 0.6373, + "step": 1658 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019980304804456504, + "loss": 0.6228, + "step": 1659 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019980244270416576, + "loss": 0.6437, + "step": 1660 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019980183643584316, + "loss": 0.4571, + "step": 1661 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019980122923960283, + "loss": 0.5194, + "step": 1662 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019980062111545047, + "loss": 0.6543, + "step": 1663 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019980001206339169, + "loss": 0.6986, + "step": 1664 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019979940208343219, + "loss": 0.5451, + "step": 1665 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997987911755776, + "loss": 0.6562, + "step": 1666 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019979817933983367, + "loss": 0.5443, + "step": 1667 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019979756657620602, + "loss": 0.5592, + "step": 1668 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019979695288470034, + "loss": 0.5408, + "step": 1669 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019979633826532243, + "loss": 0.4971, + "step": 1670 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997957227180779, + "loss": 0.5477, + "step": 1671 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019979510624297252, + "loss": 0.6198, + "step": 1672 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019979448884001203, + "loss": 0.6579, + "step": 1673 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019979387050920216, + "loss": 0.5694, + "step": 1674 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019979325125054864, + "loss": 0.4985, + "step": 1675 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019979263106405725, + "loss": 0.599, + "step": 1676 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019979200994973376, + "loss": 0.6108, + "step": 1677 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997913879075839, + "loss": 0.6882, + "step": 1678 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019979076493761353, + "loss": 0.5981, + "step": 1679 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019979014103982837, + "loss": 0.585, + "step": 1680 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019978951621423425, + "loss": 0.6775, + "step": 1681 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019978889046083702, + "loss": 0.5308, + "step": 1682 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019978826377964242, + "loss": 0.6168, + "step": 1683 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019978763617065632, + "loss": 0.4973, + "step": 1684 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019978700763388456, + "loss": 0.521, + "step": 1685 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019978637816933298, + "loss": 0.5243, + "step": 1686 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997857477770074, + "loss": 0.6235, + "step": 1687 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019978511645691373, + "loss": 0.6632, + "step": 1688 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019978448420905783, + "loss": 0.4981, + "step": 1689 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019978385103344555, + "loss": 0.5851, + "step": 1690 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997832169300828, + "loss": 0.564, + "step": 1691 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019978258189897545, + "loss": 0.5504, + "step": 1692 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019978194594012943, + "loss": 0.5839, + "step": 1693 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019978130905355066, + "loss": 0.482, + "step": 1694 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019978067123924503, + "loss": 0.5677, + "step": 1695 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019978003249721847, + "loss": 0.6657, + "step": 1696 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019977939282747698, + "loss": 0.5733, + "step": 1697 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019977875223002643, + "loss": 0.6225, + "step": 1698 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019977811070487285, + "loss": 0.673, + "step": 1699 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019977746825202214, + "loss": 0.6187, + "step": 1700 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997768248714803, + "loss": 0.6623, + "step": 1701 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997761805632533, + "loss": 0.563, + "step": 1702 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997755353273471, + "loss": 0.5387, + "step": 1703 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019977488916376782, + "loss": 0.6564, + "step": 1704 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019977424207252135, + "loss": 0.4925, + "step": 1705 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019977359405361377, + "loss": 0.4563, + "step": 1706 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019977294510705102, + "loss": 0.5571, + "step": 1707 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019977229523283923, + "loss": 0.6034, + "step": 1708 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997716444309844, + "loss": 0.5365, + "step": 1709 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997709927014926, + "loss": 0.4752, + "step": 1710 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019977034004436983, + "loss": 0.6697, + "step": 1711 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019976968645962225, + "loss": 0.6202, + "step": 1712 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019976903194725588, + "loss": 0.6088, + "step": 1713 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997683765072768, + "loss": 0.4937, + "step": 1714 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019976772013969112, + "loss": 0.6401, + "step": 1715 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019976706284450494, + "loss": 0.5585, + "step": 1716 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019976640462172438, + "loss": 0.6373, + "step": 1717 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019976574547135557, + "loss": 0.6437, + "step": 1718 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019976508539340458, + "loss": 0.6812, + "step": 1719 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997644243878776, + "loss": 0.4721, + "step": 1720 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019976376245478078, + "loss": 0.5931, + "step": 1721 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019976309959412027, + "loss": 0.5779, + "step": 1722 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019976243580590222, + "loss": 0.6539, + "step": 1723 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019976177109013276, + "loss": 0.6021, + "step": 1724 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019976110544681817, + "loss": 0.6401, + "step": 1725 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019976043887596454, + "loss": 0.6592, + "step": 1726 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019975977137757814, + "loss": 0.6916, + "step": 1727 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019975910295166513, + "loss": 0.6984, + "step": 1728 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019975843359823175, + "loss": 0.645, + "step": 1729 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997577633172842, + "loss": 0.5802, + "step": 1730 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997570921088287, + "loss": 0.702, + "step": 1731 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019975641997287157, + "loss": 0.6521, + "step": 1732 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019975574690941897, + "loss": 0.6844, + "step": 1733 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019975507291847725, + "loss": 0.8543, + "step": 1734 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019975439800005255, + "loss": 0.687, + "step": 1735 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019975372215415128, + "loss": 0.6361, + "step": 1736 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019975304538077962, + "loss": 0.5649, + "step": 1737 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997523676799439, + "loss": 0.5817, + "step": 1738 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019975168905165042, + "loss": 0.7124, + "step": 1739 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019975100949590552, + "loss": 0.5568, + "step": 1740 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019975032901271543, + "loss": 0.6946, + "step": 1741 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997496476020866, + "loss": 0.6551, + "step": 1742 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019974896526402528, + "loss": 0.6545, + "step": 1743 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019974828199853786, + "loss": 0.7778, + "step": 1744 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019974759780563064, + "loss": 0.6349, + "step": 1745 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019974691268531003, + "loss": 0.7772, + "step": 1746 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019974622663758237, + "loss": 0.6861, + "step": 1747 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019974553966245405, + "loss": 0.8056, + "step": 1748 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019974485175993146, + "loss": 0.7117, + "step": 1749 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019974416293002097, + "loss": 0.5633, + "step": 1750 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019974347317272903, + "loss": 0.5972, + "step": 1751 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019974278248806204, + "loss": 0.6226, + "step": 1752 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997420908760264, + "loss": 0.6157, + "step": 1753 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019974139833662854, + "loss": 0.638, + "step": 1754 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019974070486987493, + "loss": 0.6375, + "step": 1755 + }, + { + "epoch": 0.05, + "learning_rate": 0.000199740010475772, + "loss": 0.6222, + "step": 1756 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019973931515432619, + "loss": 0.7313, + "step": 1757 + }, + { + "epoch": 0.05, + "learning_rate": 0.000199738618905544, + "loss": 0.5074, + "step": 1758 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019973792172943186, + "loss": 0.7096, + "step": 1759 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019973722362599626, + "loss": 0.7178, + "step": 1760 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019973652459524374, + "loss": 0.5868, + "step": 1761 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019973582463718076, + "loss": 0.4609, + "step": 1762 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019973512375181385, + "loss": 0.7266, + "step": 1763 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997344219391495, + "loss": 0.7453, + "step": 1764 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019973371919919424, + "loss": 0.6646, + "step": 1765 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997330155319546, + "loss": 0.6291, + "step": 1766 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997323109374371, + "loss": 0.6117, + "step": 1767 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997316054156484, + "loss": 0.7267, + "step": 1768 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019973089896659492, + "loss": 0.7254, + "step": 1769 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997301915902833, + "loss": 0.652, + "step": 1770 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019972948328672013, + "loss": 0.6173, + "step": 1771 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019972877405591196, + "loss": 0.4972, + "step": 1772 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019972806389786542, + "loss": 0.5704, + "step": 1773 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019972735281258705, + "loss": 0.5621, + "step": 1774 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019972664080008352, + "loss": 0.5807, + "step": 1775 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019972592786036142, + "loss": 0.6863, + "step": 1776 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997252139934274, + "loss": 0.6167, + "step": 1777 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019972449919928808, + "loss": 0.7289, + "step": 1778 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019972378347795015, + "loss": 0.7017, + "step": 1779 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019972306682942016, + "loss": 0.6069, + "step": 1780 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019972234925370486, + "loss": 0.6738, + "step": 1781 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019972163075081094, + "loss": 0.6047, + "step": 1782 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019972091132074501, + "loss": 0.6282, + "step": 1783 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997201909635138, + "loss": 0.6499, + "step": 1784 + }, + { + "epoch": 0.05, + "learning_rate": 0.000199719469679124, + "loss": 0.4935, + "step": 1785 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997187474675823, + "loss": 0.7256, + "step": 1786 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019971802432889547, + "loss": 0.6365, + "step": 1787 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019971730026307016, + "loss": 0.5835, + "step": 1788 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019971657527011316, + "loss": 0.6156, + "step": 1789 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019971584935003117, + "loss": 0.6379, + "step": 1790 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019971512250283095, + "loss": 0.5845, + "step": 1791 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019971439472851925, + "loss": 0.6164, + "step": 1792 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997136660271029, + "loss": 0.5351, + "step": 1793 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019971293639858858, + "loss": 0.7021, + "step": 1794 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019971220584298312, + "loss": 0.5739, + "step": 1795 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019971147436029333, + "loss": 0.4836, + "step": 1796 + }, + { + "epoch": 0.05, + "learning_rate": 0.000199710741950526, + "loss": 0.792, + "step": 1797 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997100086136879, + "loss": 0.6613, + "step": 1798 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019970927434978593, + "loss": 0.7254, + "step": 1799 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019970853915882684, + "loss": 0.6581, + "step": 1800 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997078030408175, + "loss": 0.5686, + "step": 1801 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997070659957647, + "loss": 0.5996, + "step": 1802 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019970632802367538, + "loss": 0.6176, + "step": 1803 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019970558912455638, + "loss": 0.4943, + "step": 1804 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997048492984145, + "loss": 0.5232, + "step": 1805 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019970410854525672, + "loss": 0.6017, + "step": 1806 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019970336686508984, + "loss": 0.67, + "step": 1807 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997026242579208, + "loss": 0.6384, + "step": 1808 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997018807237565, + "loss": 0.805, + "step": 1809 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019970113626260387, + "loss": 0.6193, + "step": 1810 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001997003908744698, + "loss": 0.5662, + "step": 1811 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019969964455936122, + "loss": 0.6323, + "step": 1812 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019969889731728512, + "loss": 0.5971, + "step": 1813 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019969814914824836, + "loss": 0.6589, + "step": 1814 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019969740005225796, + "loss": 0.5887, + "step": 1815 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019969665002932086, + "loss": 0.5865, + "step": 1816 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001996958990794441, + "loss": 0.5492, + "step": 1817 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019969514720263453, + "loss": 0.6311, + "step": 1818 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019969439439889927, + "loss": 0.6136, + "step": 1819 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019969364066824526, + "loss": 0.5815, + "step": 1820 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019969288601067947, + "loss": 0.6199, + "step": 1821 + }, + { + "epoch": 0.05, + "learning_rate": 0.000199692130426209, + "loss": 0.6164, + "step": 1822 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019969137391484082, + "loss": 0.5969, + "step": 1823 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019969061647658198, + "loss": 0.5674, + "step": 1824 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001996898581114395, + "loss": 0.6498, + "step": 1825 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001996890988194205, + "loss": 0.5502, + "step": 1826 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019968833860053192, + "loss": 0.5652, + "step": 1827 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019968757745478097, + "loss": 0.6222, + "step": 1828 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019968681538217462, + "loss": 0.6254, + "step": 1829 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019968605238272001, + "loss": 0.558, + "step": 1830 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001996852884564242, + "loss": 0.5418, + "step": 1831 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019968452360329432, + "loss": 0.7673, + "step": 1832 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019968375782333748, + "loss": 0.6518, + "step": 1833 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019968299111656077, + "loss": 0.5039, + "step": 1834 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019968222348297135, + "loss": 0.7107, + "step": 1835 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019968145492257634, + "loss": 0.5962, + "step": 1836 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019968068543538286, + "loss": 0.6926, + "step": 1837 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019967991502139817, + "loss": 0.5828, + "step": 1838 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001996791436806293, + "loss": 0.576, + "step": 1839 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019967837141308348, + "loss": 0.5834, + "step": 1840 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019967759821876792, + "loss": 0.5533, + "step": 1841 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019967682409768975, + "loss": 0.7027, + "step": 1842 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019967604904985622, + "loss": 0.6502, + "step": 1843 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019967527307527448, + "loss": 0.5114, + "step": 1844 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019967449617395183, + "loss": 0.4802, + "step": 1845 + }, + { + "epoch": 0.05, + "learning_rate": 0.0001996737183458954, + "loss": 0.5781, + "step": 1846 + }, + { + "epoch": 0.05, + "learning_rate": 0.00019967293959111248, + "loss": 0.6529, + "step": 1847 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996721599096103, + "loss": 0.7049, + "step": 1848 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019967137930139607, + "loss": 0.7186, + "step": 1849 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996705977664771, + "loss": 0.5274, + "step": 1850 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019966981530486065, + "loss": 0.5558, + "step": 1851 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019966903191655398, + "loss": 0.599, + "step": 1852 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019966824760156437, + "loss": 0.6984, + "step": 1853 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019966746235989912, + "loss": 0.49, + "step": 1854 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996666761915655, + "loss": 0.5748, + "step": 1855 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019966588909657086, + "loss": 0.5836, + "step": 1856 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019966510107492254, + "loss": 0.4839, + "step": 1857 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996643121266278, + "loss": 0.6294, + "step": 1858 + }, + { + "epoch": 0.06, + "learning_rate": 0.000199663522251694, + "loss": 0.5513, + "step": 1859 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019966273145012852, + "loss": 0.5553, + "step": 1860 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019966193972193868, + "loss": 0.539, + "step": 1861 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019966114706713184, + "loss": 0.5544, + "step": 1862 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019966035348571536, + "loss": 0.6093, + "step": 1863 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019965955897769666, + "loss": 0.6165, + "step": 1864 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996587635430831, + "loss": 0.6522, + "step": 1865 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019965796718188206, + "loss": 0.5369, + "step": 1866 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019965716989410096, + "loss": 0.5441, + "step": 1867 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019965637167974722, + "loss": 0.6339, + "step": 1868 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019965557253882828, + "loss": 0.5124, + "step": 1869 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019965477247135152, + "loss": 0.5046, + "step": 1870 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019965397147732441, + "loss": 0.5186, + "step": 1871 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019965316955675437, + "loss": 0.5945, + "step": 1872 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996523667096489, + "loss": 0.5734, + "step": 1873 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019965156293601545, + "loss": 0.6158, + "step": 1874 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019965075823586147, + "loss": 0.5896, + "step": 1875 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019964995260919445, + "loss": 0.6327, + "step": 1876 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019964914605602186, + "loss": 0.7217, + "step": 1877 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019964833857635128, + "loss": 0.5275, + "step": 1878 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019964753017019015, + "loss": 0.5753, + "step": 1879 + }, + { + "epoch": 0.06, + "learning_rate": 0.000199646720837546, + "loss": 0.6051, + "step": 1880 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019964591057842633, + "loss": 0.5784, + "step": 1881 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019964509939283874, + "loss": 0.6886, + "step": 1882 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996442872807907, + "loss": 0.625, + "step": 1883 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996434742422898, + "loss": 0.6864, + "step": 1884 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996426602773436, + "loss": 0.7044, + "step": 1885 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019964184538595964, + "loss": 0.6445, + "step": 1886 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019964102956814553, + "loss": 0.5379, + "step": 1887 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019964021282390884, + "loss": 0.5392, + "step": 1888 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019963939515325717, + "loss": 0.6907, + "step": 1889 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019963857655619808, + "loss": 0.6246, + "step": 1890 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019963775703273924, + "loss": 0.5744, + "step": 1891 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019963693658288825, + "loss": 0.542, + "step": 1892 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019963611520665273, + "loss": 0.5785, + "step": 1893 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019963529290404033, + "loss": 0.6106, + "step": 1894 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019963446967505869, + "loss": 0.6805, + "step": 1895 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019963364551971546, + "loss": 0.5014, + "step": 1896 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019963282043801827, + "loss": 0.585, + "step": 1897 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996319944299749, + "loss": 0.5521, + "step": 1898 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996311674955929, + "loss": 0.6257, + "step": 1899 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019963033963488, + "loss": 0.5022, + "step": 1900 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019962951084784395, + "loss": 0.5886, + "step": 1901 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019962868113449238, + "loss": 0.5287, + "step": 1902 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019962785049483305, + "loss": 0.6952, + "step": 1903 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996270189288737, + "loss": 0.6808, + "step": 1904 + }, + { + "epoch": 0.06, + "learning_rate": 0.000199626186436622, + "loss": 0.5421, + "step": 1905 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019962535301808578, + "loss": 0.6547, + "step": 1906 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019962451867327268, + "loss": 0.5929, + "step": 1907 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019962368340219056, + "loss": 0.4897, + "step": 1908 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996228472048471, + "loss": 0.6208, + "step": 1909 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019962201008125012, + "loss": 0.5628, + "step": 1910 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996211720314074, + "loss": 0.5554, + "step": 1911 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019962033305532672, + "loss": 0.5948, + "step": 1912 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996194931530159, + "loss": 0.5494, + "step": 1913 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019961865232448272, + "loss": 0.5243, + "step": 1914 + }, + { + "epoch": 0.06, + "learning_rate": 0.000199617810569735, + "loss": 0.7264, + "step": 1915 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996169678887806, + "loss": 0.5368, + "step": 1916 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019961612428162737, + "loss": 0.5061, + "step": 1917 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996152797482831, + "loss": 0.6396, + "step": 1918 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019961443428875562, + "loss": 0.5668, + "step": 1919 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019961358790305286, + "loss": 0.6603, + "step": 1920 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996127405911827, + "loss": 0.5583, + "step": 1921 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019961189235315294, + "loss": 0.6642, + "step": 1922 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996110431889715, + "loss": 0.6494, + "step": 1923 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996101930986463, + "loss": 0.6101, + "step": 1924 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019960934208218523, + "loss": 0.5451, + "step": 1925 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019960849013959616, + "loss": 0.5784, + "step": 1926 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996076372708871, + "loss": 0.5833, + "step": 1927 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019960678347606588, + "loss": 0.5945, + "step": 1928 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019960592875514053, + "loss": 0.6132, + "step": 1929 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019960507310811892, + "loss": 0.5294, + "step": 1930 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019960421653500906, + "loss": 0.6498, + "step": 1931 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996033590358189, + "loss": 0.5259, + "step": 1932 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019960250061055638, + "loss": 0.6047, + "step": 1933 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019960164125922952, + "loss": 0.5885, + "step": 1934 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001996007809818463, + "loss": 0.5833, + "step": 1935 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995999197784147, + "loss": 0.6514, + "step": 1936 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995990576489428, + "loss": 0.6191, + "step": 1937 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019959819459343847, + "loss": 0.526, + "step": 1938 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019959733061190987, + "loss": 0.6411, + "step": 1939 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019959646570436496, + "loss": 0.5039, + "step": 1940 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019959559987081185, + "loss": 0.623, + "step": 1941 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995947331112585, + "loss": 0.6567, + "step": 1942 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019959386542571305, + "loss": 0.6026, + "step": 1943 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019959299681418356, + "loss": 0.6799, + "step": 1944 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019959212727667803, + "loss": 0.485, + "step": 1945 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019959125681320462, + "loss": 0.5733, + "step": 1946 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019959038542377138, + "loss": 0.5523, + "step": 1947 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019958951310838642, + "loss": 0.6804, + "step": 1948 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019958863986705786, + "loss": 0.5694, + "step": 1949 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019958776569979384, + "loss": 0.612, + "step": 1950 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019958689060660246, + "loss": 0.5464, + "step": 1951 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019958601458749186, + "loss": 0.6034, + "step": 1952 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019958513764247018, + "loss": 0.5896, + "step": 1953 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995842597715456, + "loss": 0.4939, + "step": 1954 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019958338097472624, + "loss": 0.5923, + "step": 1955 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019958250125202032, + "loss": 0.5641, + "step": 1956 + }, + { + "epoch": 0.06, + "learning_rate": 0.000199581620603436, + "loss": 0.6753, + "step": 1957 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995807390289814, + "loss": 0.5403, + "step": 1958 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019957985652866486, + "loss": 0.5588, + "step": 1959 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019957897310249445, + "loss": 0.6481, + "step": 1960 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019957808875047843, + "loss": 0.6188, + "step": 1961 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019957720347262505, + "loss": 0.5978, + "step": 1962 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019957631726894254, + "loss": 0.4948, + "step": 1963 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019957543013943908, + "loss": 0.6486, + "step": 1964 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019957454208412298, + "loss": 0.5654, + "step": 1965 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019957365310300243, + "loss": 0.5845, + "step": 1966 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995727631960858, + "loss": 0.6471, + "step": 1967 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019957187236338127, + "loss": 0.6006, + "step": 1968 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019957098060489716, + "loss": 0.5674, + "step": 1969 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019957008792064178, + "loss": 0.5529, + "step": 1970 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019956919431062336, + "loss": 0.5787, + "step": 1971 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995682997748503, + "loss": 0.5416, + "step": 1972 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019956740431333088, + "loss": 0.6617, + "step": 1973 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019956650792607339, + "loss": 0.5633, + "step": 1974 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019956561061308618, + "loss": 0.6073, + "step": 1975 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019956471237437762, + "loss": 0.6157, + "step": 1976 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019956381320995607, + "loss": 0.4262, + "step": 1977 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019956291311982985, + "loss": 0.5913, + "step": 1978 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019956201210400733, + "loss": 0.6145, + "step": 1979 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019956111016249692, + "loss": 0.7232, + "step": 1980 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019956020729530697, + "loss": 0.5627, + "step": 1981 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019955930350244594, + "loss": 0.5317, + "step": 1982 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019955839878392216, + "loss": 0.7249, + "step": 1983 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019955749313974405, + "loss": 0.6128, + "step": 1984 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019955658656992003, + "loss": 0.6204, + "step": 1985 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995556790744586, + "loss": 0.5895, + "step": 1986 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019955477065336811, + "loss": 0.6095, + "step": 1987 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019955386130665706, + "loss": 0.614, + "step": 1988 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019955295103433392, + "loss": 0.6301, + "step": 1989 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019955203983640707, + "loss": 0.6112, + "step": 1990 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019955112771288501, + "loss": 0.5735, + "step": 1991 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995502146637763, + "loss": 0.6228, + "step": 1992 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019954930068908936, + "loss": 0.6127, + "step": 1993 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019954838578883265, + "loss": 0.5792, + "step": 1994 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019954746996301477, + "loss": 0.549, + "step": 1995 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019954655321164418, + "loss": 0.4713, + "step": 1996 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019954563553472942, + "loss": 0.579, + "step": 1997 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019954471693227902, + "loss": 0.6037, + "step": 1998 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019954379740430147, + "loss": 0.504, + "step": 1999 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995428769508054, + "loss": 0.6556, + "step": 2000 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019954195557179936, + "loss": 0.6364, + "step": 2001 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019954103326729185, + "loss": 0.5154, + "step": 2002 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019954011003729148, + "loss": 0.5439, + "step": 2003 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995391858818069, + "loss": 0.5261, + "step": 2004 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019953826080084658, + "loss": 0.6003, + "step": 2005 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995373347944192, + "loss": 0.5424, + "step": 2006 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019953640786253337, + "loss": 0.6059, + "step": 2007 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019953548000519765, + "loss": 0.4755, + "step": 2008 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019953455122242075, + "loss": 0.5364, + "step": 2009 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019953362151421127, + "loss": 0.5305, + "step": 2010 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019953269088057783, + "loss": 0.5085, + "step": 2011 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019953175932152907, + "loss": 0.5514, + "step": 2012 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019953082683707373, + "loss": 0.6151, + "step": 2013 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995298934272204, + "loss": 0.5549, + "step": 2014 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019952895909197782, + "loss": 0.5161, + "step": 2015 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995280238313546, + "loss": 0.655, + "step": 2016 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019952708764535952, + "loss": 0.5587, + "step": 2017 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019952615053400123, + "loss": 0.6435, + "step": 2018 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995252124972885, + "loss": 0.5507, + "step": 2019 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019952427353522994, + "loss": 0.5341, + "step": 2020 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019952333364783438, + "loss": 0.596, + "step": 2021 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019952239283511057, + "loss": 0.6413, + "step": 2022 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019952145109706715, + "loss": 0.5413, + "step": 2023 + }, + { + "epoch": 0.06, + "learning_rate": 0.000199520508433713, + "loss": 0.4437, + "step": 2024 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995195648450568, + "loss": 0.6136, + "step": 2025 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019951862033110735, + "loss": 0.5508, + "step": 2026 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019951767489187345, + "loss": 0.566, + "step": 2027 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019951672852736385, + "loss": 0.6263, + "step": 2028 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019951578123758736, + "loss": 0.6638, + "step": 2029 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019951483302255282, + "loss": 0.6059, + "step": 2030 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019951388388226904, + "loss": 0.5155, + "step": 2031 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019951293381674482, + "loss": 0.6134, + "step": 2032 + }, + { + "epoch": 0.06, + "learning_rate": 0.000199511982825989, + "loss": 0.6256, + "step": 2033 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019951103091001043, + "loss": 0.5477, + "step": 2034 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019951007806881795, + "loss": 0.5515, + "step": 2035 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019950912430242044, + "loss": 0.64, + "step": 2036 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019950816961082673, + "loss": 0.6379, + "step": 2037 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995072139940457, + "loss": 0.5464, + "step": 2038 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995062574520863, + "loss": 0.6498, + "step": 2039 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019950529998495736, + "loss": 0.5951, + "step": 2040 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995043415926678, + "loss": 0.7258, + "step": 2041 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019950338227522655, + "loss": 0.572, + "step": 2042 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019950242203264248, + "loss": 0.5314, + "step": 2043 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019950146086492455, + "loss": 0.5922, + "step": 2044 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001995004987720817, + "loss": 0.5325, + "step": 2045 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019949953575412285, + "loss": 0.5851, + "step": 2046 + }, + { + "epoch": 0.06, + "learning_rate": 0.000199498571811057, + "loss": 0.5999, + "step": 2047 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019949760694289308, + "loss": 0.6389, + "step": 2048 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019949664114964007, + "loss": 0.7072, + "step": 2049 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019949567443130692, + "loss": 0.6188, + "step": 2050 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019949470678790268, + "loss": 0.6095, + "step": 2051 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001994937382194363, + "loss": 0.6096, + "step": 2052 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019949276872591678, + "loss": 0.5959, + "step": 2053 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019949179830735316, + "loss": 0.6644, + "step": 2054 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019949082696375445, + "loss": 0.6409, + "step": 2055 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019948985469512968, + "loss": 0.5784, + "step": 2056 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019948888150148793, + "loss": 0.6284, + "step": 2057 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019948790738283817, + "loss": 0.6314, + "step": 2058 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019948693233918952, + "loss": 0.7273, + "step": 2059 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019948595637055104, + "loss": 0.5788, + "step": 2060 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019948497947693173, + "loss": 0.6979, + "step": 2061 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001994840016583408, + "loss": 0.6467, + "step": 2062 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019948302291478723, + "loss": 0.5175, + "step": 2063 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019948204324628019, + "loss": 0.7018, + "step": 2064 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019948106265282872, + "loss": 0.6202, + "step": 2065 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019948008113444203, + "loss": 0.5602, + "step": 2066 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019947909869112915, + "loss": 0.6274, + "step": 2067 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001994781153228993, + "loss": 0.5121, + "step": 2068 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019947713102976155, + "loss": 0.5299, + "step": 2069 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019947614581172508, + "loss": 0.4967, + "step": 2070 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019947515966879908, + "loss": 0.5457, + "step": 2071 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019947417260099268, + "loss": 0.5161, + "step": 2072 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019947318460831508, + "loss": 0.4824, + "step": 2073 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019947219569077545, + "loss": 0.4887, + "step": 2074 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019947120584838296, + "loss": 0.7096, + "step": 2075 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001994702150811469, + "loss": 0.5787, + "step": 2076 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019946922338907639, + "loss": 0.5363, + "step": 2077 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001994682307721807, + "loss": 0.624, + "step": 2078 + }, + { + "epoch": 0.06, + "learning_rate": 0.000199467237230469, + "loss": 0.5507, + "step": 2079 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001994662427639506, + "loss": 0.5996, + "step": 2080 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019946524737263473, + "loss": 0.6354, + "step": 2081 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019946425105653063, + "loss": 0.481, + "step": 2082 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019946325381564757, + "loss": 0.4885, + "step": 2083 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019946225564999482, + "loss": 0.4887, + "step": 2084 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019946125655958166, + "loss": 0.5115, + "step": 2085 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019946025654441735, + "loss": 0.5016, + "step": 2086 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019945925560451122, + "loss": 0.6644, + "step": 2087 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019945825373987258, + "loss": 0.5974, + "step": 2088 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019945725095051075, + "loss": 0.6127, + "step": 2089 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019945624723643506, + "loss": 0.5624, + "step": 2090 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019945524259765477, + "loss": 0.509, + "step": 2091 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001994542370341793, + "loss": 0.6115, + "step": 2092 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019945323054601798, + "loss": 0.5981, + "step": 2093 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019945222313318015, + "loss": 0.7036, + "step": 2094 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019945121479567515, + "loss": 0.6112, + "step": 2095 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001994502055335125, + "loss": 0.6194, + "step": 2096 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019944919534670136, + "loss": 0.559, + "step": 2097 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019944818423525131, + "loss": 0.5354, + "step": 2098 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019944717219917165, + "loss": 0.6318, + "step": 2099 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019944615923847182, + "loss": 0.6165, + "step": 2100 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019944514535316123, + "loss": 0.6537, + "step": 2101 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019944413054324934, + "loss": 0.5784, + "step": 2102 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019944311480874554, + "loss": 0.6172, + "step": 2103 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019944209814965928, + "loss": 0.5272, + "step": 2104 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019944108056600004, + "loss": 0.5952, + "step": 2105 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019944006205777727, + "loss": 0.6198, + "step": 2106 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019943904262500044, + "loss": 0.5221, + "step": 2107 + }, + { + "epoch": 0.06, + "learning_rate": 0.000199438022267679, + "loss": 0.4837, + "step": 2108 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019943700098582246, + "loss": 0.6062, + "step": 2109 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019943597877944034, + "loss": 0.4241, + "step": 2110 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019943495564854211, + "loss": 0.53, + "step": 2111 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019943393159313732, + "loss": 0.4592, + "step": 2112 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019943290661323542, + "loss": 0.5605, + "step": 2113 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019943188070884602, + "loss": 0.6413, + "step": 2114 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019943085387997858, + "loss": 0.6326, + "step": 2115 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019942982612664273, + "loss": 0.5037, + "step": 2116 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019942879744884798, + "loss": 0.5873, + "step": 2117 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019942776784660389, + "loss": 0.5598, + "step": 2118 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019942673731992004, + "loss": 0.6669, + "step": 2119 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019942570586880604, + "loss": 0.4377, + "step": 2120 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019942467349327143, + "loss": 0.4891, + "step": 2121 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019942364019332582, + "loss": 0.6013, + "step": 2122 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019942260596897887, + "loss": 0.4896, + "step": 2123 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019942157082024014, + "loss": 0.5699, + "step": 2124 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019942053474711924, + "loss": 0.5624, + "step": 2125 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019941949774962585, + "loss": 0.6044, + "step": 2126 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019941845982776962, + "loss": 0.5448, + "step": 2127 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019941742098156015, + "loss": 0.6337, + "step": 2128 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019941638121100713, + "loss": 0.5441, + "step": 2129 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001994153405161202, + "loss": 0.5798, + "step": 2130 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019941429889690907, + "loss": 0.6506, + "step": 2131 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019941325635338344, + "loss": 0.5529, + "step": 2132 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019941221288555293, + "loss": 0.5178, + "step": 2133 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001994111684934273, + "loss": 0.6303, + "step": 2134 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001994101231770163, + "loss": 0.494, + "step": 2135 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019940907693632954, + "loss": 0.6794, + "step": 2136 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001994080297713768, + "loss": 0.5939, + "step": 2137 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019940698168216786, + "loss": 0.682, + "step": 2138 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019940593266871238, + "loss": 0.6267, + "step": 2139 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019940488273102015, + "loss": 0.5903, + "step": 2140 + }, + { + "epoch": 0.06, + "learning_rate": 0.000199403831869101, + "loss": 0.568, + "step": 2141 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001994027800829646, + "loss": 0.6291, + "step": 2142 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019940172737262075, + "loss": 0.6597, + "step": 2143 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001994006737380793, + "loss": 0.6698, + "step": 2144 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019939961917934997, + "loss": 0.5697, + "step": 2145 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001993985636964426, + "loss": 0.6088, + "step": 2146 + }, + { + "epoch": 0.06, + "learning_rate": 0.000199397507289367, + "loss": 0.5766, + "step": 2147 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019939644995813302, + "loss": 0.5498, + "step": 2148 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019939539170275045, + "loss": 0.5197, + "step": 2149 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019939433252322915, + "loss": 0.5686, + "step": 2150 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019939327241957894, + "loss": 0.5433, + "step": 2151 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019939221139180971, + "loss": 0.5036, + "step": 2152 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001993911494399313, + "loss": 0.5697, + "step": 2153 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019939008656395362, + "loss": 0.4631, + "step": 2154 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019938902276388652, + "loss": 0.5764, + "step": 2155 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001993879580397399, + "loss": 0.6174, + "step": 2156 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019938689239152365, + "loss": 0.5486, + "step": 2157 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019938582581924768, + "loss": 0.5446, + "step": 2158 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019938475832292194, + "loss": 0.5529, + "step": 2159 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019938368990255628, + "loss": 0.4656, + "step": 2160 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019938262055816072, + "loss": 0.6747, + "step": 2161 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019938155028974516, + "loss": 0.6088, + "step": 2162 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019938047909731955, + "loss": 0.5069, + "step": 2163 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019937940698089385, + "loss": 0.6348, + "step": 2164 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019937833394047805, + "loss": 0.6088, + "step": 2165 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001993772599760821, + "loss": 0.4809, + "step": 2166 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019937618508771598, + "loss": 0.4756, + "step": 2167 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019937510927538972, + "loss": 0.5512, + "step": 2168 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001993740325391133, + "loss": 0.6031, + "step": 2169 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001993729548788967, + "loss": 0.6206, + "step": 2170 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019937187629474998, + "loss": 0.5511, + "step": 2171 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019937079678668317, + "loss": 0.6214, + "step": 2172 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019936971635470632, + "loss": 0.4837, + "step": 2173 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019936863499882943, + "loss": 0.6879, + "step": 2174 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001993675527190626, + "loss": 0.5899, + "step": 2175 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019936646951541583, + "loss": 0.6486, + "step": 2176 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019936538538789926, + "loss": 0.5466, + "step": 2177 + }, + { + "epoch": 0.06, + "learning_rate": 0.0001993643003365229, + "loss": 0.6344, + "step": 2178 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019936321436129692, + "loss": 0.4883, + "step": 2179 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019936212746223138, + "loss": 0.5952, + "step": 2180 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019936103963933633, + "loss": 0.5778, + "step": 2181 + }, + { + "epoch": 0.06, + "learning_rate": 0.000199359950892622, + "loss": 0.6549, + "step": 2182 + }, + { + "epoch": 0.06, + "learning_rate": 0.00019935886122209842, + "loss": 0.628, + "step": 2183 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019935777062777573, + "loss": 0.6273, + "step": 2184 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001993566791096641, + "loss": 0.6877, + "step": 2185 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019935558666777365, + "loss": 0.6708, + "step": 2186 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019935449330211457, + "loss": 0.6137, + "step": 2187 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019935339901269706, + "loss": 0.5511, + "step": 2188 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019935230379953116, + "loss": 0.5625, + "step": 2189 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001993512076626272, + "loss": 0.4995, + "step": 2190 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019935011060199528, + "loss": 0.5931, + "step": 2191 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019934901261764562, + "loss": 0.6205, + "step": 2192 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019934791370958846, + "loss": 0.511, + "step": 2193 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019934681387783398, + "loss": 0.6193, + "step": 2194 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001993457131223924, + "loss": 0.5392, + "step": 2195 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019934461144327403, + "loss": 0.6233, + "step": 2196 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019934350884048902, + "loss": 0.5611, + "step": 2197 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019934240531404765, + "loss": 0.6435, + "step": 2198 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019934130086396022, + "loss": 0.6221, + "step": 2199 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019934019549023698, + "loss": 0.5634, + "step": 2200 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019933908919288814, + "loss": 0.6002, + "step": 2201 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019933798197192408, + "loss": 0.5308, + "step": 2202 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019933687382735503, + "loss": 0.6043, + "step": 2203 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019933576475919136, + "loss": 0.4071, + "step": 2204 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001993346547674433, + "loss": 0.62, + "step": 2205 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019933354385212124, + "loss": 0.6318, + "step": 2206 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019933243201323547, + "loss": 0.7017, + "step": 2207 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001993313192507963, + "loss": 0.5684, + "step": 2208 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019933020556481417, + "loss": 0.4529, + "step": 2209 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019932909095529935, + "loss": 0.4507, + "step": 2210 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019932797542226226, + "loss": 0.532, + "step": 2211 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001993268589657132, + "loss": 0.6643, + "step": 2212 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001993257415856626, + "loss": 0.6081, + "step": 2213 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019932462328212087, + "loss": 0.5073, + "step": 2214 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019932350405509834, + "loss": 0.59, + "step": 2215 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001993223839046055, + "loss": 0.6085, + "step": 2216 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001993212628306527, + "loss": 0.5683, + "step": 2217 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001993201408332504, + "loss": 0.6635, + "step": 2218 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019931901791240896, + "loss": 0.5003, + "step": 2219 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019931789406813893, + "loss": 0.4927, + "step": 2220 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019931676930045068, + "loss": 0.6069, + "step": 2221 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001993156436093547, + "loss": 0.5245, + "step": 2222 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019931451699486147, + "loss": 0.5877, + "step": 2223 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001993133894569814, + "loss": 0.6764, + "step": 2224 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019931226099572505, + "loss": 0.6312, + "step": 2225 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019931113161110288, + "loss": 0.581, + "step": 2226 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019931000130312537, + "loss": 0.5764, + "step": 2227 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019930887007180308, + "loss": 0.5341, + "step": 2228 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019930773791714646, + "loss": 0.5679, + "step": 2229 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001993066048391661, + "loss": 0.5353, + "step": 2230 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001993054708378725, + "loss": 0.6306, + "step": 2231 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001993043359132762, + "loss": 0.5753, + "step": 2232 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019930320006538777, + "loss": 0.6035, + "step": 2233 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001993020632942178, + "loss": 0.5359, + "step": 2234 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019930092559977678, + "loss": 0.6329, + "step": 2235 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019929978698207532, + "loss": 0.6063, + "step": 2236 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019929864744112406, + "loss": 0.4563, + "step": 2237 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019929750697693352, + "loss": 0.6147, + "step": 2238 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019929636558951437, + "loss": 0.6433, + "step": 2239 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019929522327887715, + "loss": 0.5419, + "step": 2240 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019929408004503254, + "loss": 0.468, + "step": 2241 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019929293588799113, + "loss": 0.5975, + "step": 2242 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992917908077636, + "loss": 0.5891, + "step": 2243 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019929064480436057, + "loss": 0.5826, + "step": 2244 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019928949787779269, + "loss": 0.6589, + "step": 2245 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019928835002807062, + "loss": 0.5057, + "step": 2246 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019928720125520508, + "loss": 0.6662, + "step": 2247 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019928605155920668, + "loss": 0.6992, + "step": 2248 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019928490094008616, + "loss": 0.5438, + "step": 2249 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019928374939785418, + "loss": 0.6016, + "step": 2250 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019928259693252151, + "loss": 0.6012, + "step": 2251 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992814435440988, + "loss": 0.5792, + "step": 2252 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992802892325968, + "loss": 0.6381, + "step": 2253 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019927913399802623, + "loss": 0.5373, + "step": 2254 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019927797784039783, + "loss": 0.4898, + "step": 2255 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992768207597224, + "loss": 0.4916, + "step": 2256 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019927566275601064, + "loss": 0.5191, + "step": 2257 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019927450382927332, + "loss": 0.5872, + "step": 2258 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019927334397952127, + "loss": 0.605, + "step": 2259 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992721832067652, + "loss": 0.5167, + "step": 2260 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019927102151101592, + "loss": 0.6147, + "step": 2261 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992698588922843, + "loss": 0.5449, + "step": 2262 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019926869535058105, + "loss": 0.5273, + "step": 2263 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019926753088591706, + "loss": 0.6578, + "step": 2264 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019926636549830314, + "loss": 0.5708, + "step": 2265 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019926519918775007, + "loss": 0.5441, + "step": 2266 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992640319542688, + "loss": 0.7059, + "step": 2267 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019926286379787008, + "loss": 0.4886, + "step": 2268 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019926169471856485, + "loss": 0.5304, + "step": 2269 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992605247163639, + "loss": 0.6569, + "step": 2270 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019925935379127817, + "loss": 0.5953, + "step": 2271 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019925818194331857, + "loss": 0.5377, + "step": 2272 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992570091724959, + "loss": 0.6273, + "step": 2273 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019925583547882115, + "loss": 0.5561, + "step": 2274 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992546608623052, + "loss": 0.5967, + "step": 2275 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019925348532295897, + "loss": 0.6786, + "step": 2276 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019925230886079338, + "loss": 0.5786, + "step": 2277 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992511314758194, + "loss": 0.7009, + "step": 2278 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019924995316804794, + "loss": 0.5441, + "step": 2279 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019924877393748998, + "loss": 0.5222, + "step": 2280 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019924759378415647, + "loss": 0.5191, + "step": 2281 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019924641270805838, + "loss": 0.5753, + "step": 2282 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992452307092067, + "loss": 0.5194, + "step": 2283 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019924404778761244, + "loss": 0.5602, + "step": 2284 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019924286394328657, + "loss": 0.7327, + "step": 2285 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019924167917624014, + "loss": 0.5983, + "step": 2286 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019924049348648407, + "loss": 0.5515, + "step": 2287 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019923930687402948, + "loss": 0.5696, + "step": 2288 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019923811933888737, + "loss": 0.5992, + "step": 2289 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019923693088106878, + "loss": 0.7252, + "step": 2290 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019923574150058473, + "loss": 0.6499, + "step": 2291 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992345511974463, + "loss": 0.5809, + "step": 2292 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992333599716646, + "loss": 0.6155, + "step": 2293 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019923216782325066, + "loss": 0.478, + "step": 2294 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992309747522156, + "loss": 0.5656, + "step": 2295 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019922978075857044, + "loss": 0.6176, + "step": 2296 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019922858584232636, + "loss": 0.529, + "step": 2297 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992273900034944, + "loss": 0.5517, + "step": 2298 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019922619324208576, + "loss": 0.5463, + "step": 2299 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992249955581115, + "loss": 0.6152, + "step": 2300 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992237969515828, + "loss": 0.69, + "step": 2301 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019922259742251075, + "loss": 0.5814, + "step": 2302 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019922139697090653, + "loss": 0.5825, + "step": 2303 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019922019559678135, + "loss": 0.5725, + "step": 2304 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992189933001463, + "loss": 0.6662, + "step": 2305 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992177900810126, + "loss": 0.4956, + "step": 2306 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019921658593939143, + "loss": 0.5331, + "step": 2307 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019921538087529401, + "loss": 0.689, + "step": 2308 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992141748887315, + "loss": 0.6699, + "step": 2309 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019921296797971517, + "loss": 0.6441, + "step": 2310 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019921176014825616, + "loss": 0.6001, + "step": 2311 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019921055139436578, + "loss": 0.5678, + "step": 2312 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019920934171805522, + "loss": 0.562, + "step": 2313 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019920813111933575, + "loss": 0.6083, + "step": 2314 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992069195982186, + "loss": 0.6467, + "step": 2315 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019920570715471506, + "loss": 0.631, + "step": 2316 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992044937888364, + "loss": 0.5955, + "step": 2317 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992032795005939, + "loss": 0.5437, + "step": 2318 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992020642899988, + "loss": 0.5161, + "step": 2319 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001992008481570625, + "loss": 0.614, + "step": 2320 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019919963110179626, + "loss": 0.5437, + "step": 2321 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991984131242113, + "loss": 0.5602, + "step": 2322 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019919719422431913, + "loss": 0.5947, + "step": 2323 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019919597440213094, + "loss": 0.5671, + "step": 2324 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019919475365765812, + "loss": 0.6108, + "step": 2325 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019919353199091203, + "loss": 0.6546, + "step": 2326 + }, + { + "epoch": 0.07, + "learning_rate": 0.000199192309401904, + "loss": 0.6278, + "step": 2327 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019919108589064538, + "loss": 0.6471, + "step": 2328 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019918986145714765, + "loss": 0.5926, + "step": 2329 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019918863610142204, + "loss": 0.625, + "step": 2330 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019918740982348009, + "loss": 0.6461, + "step": 2331 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991861826233331, + "loss": 0.592, + "step": 2332 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019918495450099252, + "loss": 0.5635, + "step": 2333 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019918372545646977, + "loss": 0.6403, + "step": 2334 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019918249548977628, + "loss": 0.6583, + "step": 2335 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019918126460092346, + "loss": 0.6125, + "step": 2336 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991800327899228, + "loss": 0.5569, + "step": 2337 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991788000567857, + "loss": 0.5793, + "step": 2338 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019917756640152362, + "loss": 0.7725, + "step": 2339 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991763318241481, + "loss": 0.6215, + "step": 2340 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019917509632467057, + "loss": 0.5616, + "step": 2341 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991738599031025, + "loss": 0.6052, + "step": 2342 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019917262255945544, + "loss": 0.617, + "step": 2343 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991713842937408, + "loss": 0.601, + "step": 2344 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019917014510597024, + "loss": 0.5257, + "step": 2345 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019916890499615514, + "loss": 0.6212, + "step": 2346 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991676639643071, + "loss": 0.5454, + "step": 2347 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019916642201043766, + "loss": 0.5357, + "step": 2348 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019916517913455832, + "loss": 0.6876, + "step": 2349 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019916393533668067, + "loss": 0.527, + "step": 2350 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991626906168163, + "loss": 0.599, + "step": 2351 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019916144497497676, + "loss": 0.5617, + "step": 2352 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991601984111736, + "loss": 0.6855, + "step": 2353 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019915895092541844, + "loss": 0.6384, + "step": 2354 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019915770251772289, + "loss": 0.7285, + "step": 2355 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019915645318809851, + "loss": 0.6479, + "step": 2356 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019915520293655695, + "loss": 0.5768, + "step": 2357 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019915395176310988, + "loss": 0.4946, + "step": 2358 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019915269966776884, + "loss": 0.4935, + "step": 2359 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019915144665054552, + "loss": 0.6656, + "step": 2360 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019915019271145158, + "loss": 0.5439, + "step": 2361 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991489378504987, + "loss": 0.6375, + "step": 2362 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019914768206769846, + "loss": 0.6489, + "step": 2363 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991464253630626, + "loss": 0.5559, + "step": 2364 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991451677366028, + "loss": 0.5205, + "step": 2365 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019914390918833076, + "loss": 0.5275, + "step": 2366 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019914264971825817, + "loss": 0.5553, + "step": 2367 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991413893263967, + "loss": 0.5679, + "step": 2368 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019914012801275813, + "loss": 0.4984, + "step": 2369 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019913886577735418, + "loss": 0.5939, + "step": 2370 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019913760262019653, + "loss": 0.5711, + "step": 2371 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019913633854129698, + "loss": 0.6858, + "step": 2372 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019913507354066726, + "loss": 0.5736, + "step": 2373 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019913380761831914, + "loss": 0.5334, + "step": 2374 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991325407742644, + "loss": 0.5902, + "step": 2375 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991312730085148, + "loss": 0.5428, + "step": 2376 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019913000432108212, + "loss": 0.5641, + "step": 2377 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019912873471197818, + "loss": 0.6312, + "step": 2378 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019912746418121477, + "loss": 0.574, + "step": 2379 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019912619272880372, + "loss": 0.6603, + "step": 2380 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019912492035475686, + "loss": 0.6131, + "step": 2381 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019912364705908594, + "loss": 0.5768, + "step": 2382 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019912237284180288, + "loss": 0.6339, + "step": 2383 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991210977029195, + "loss": 0.5365, + "step": 2384 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991198216424477, + "loss": 0.5823, + "step": 2385 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019911854466039925, + "loss": 0.6679, + "step": 2386 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991172667567861, + "loss": 0.542, + "step": 2387 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001991159879316201, + "loss": 0.4806, + "step": 2388 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019911470818491316, + "loss": 0.5794, + "step": 2389 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019911342751667715, + "loss": 0.6207, + "step": 2390 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019911214592692401, + "loss": 0.5811, + "step": 2391 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019911086341566563, + "loss": 0.6077, + "step": 2392 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019910957998291396, + "loss": 0.6713, + "step": 2393 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019910829562868088, + "loss": 0.5709, + "step": 2394 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019910701035297842, + "loss": 0.4753, + "step": 2395 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019910572415581844, + "loss": 0.6574, + "step": 2396 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019910443703721292, + "loss": 0.5624, + "step": 2397 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019910314899717387, + "loss": 0.6042, + "step": 2398 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019910186003571325, + "loss": 0.6344, + "step": 2399 + }, + { + "epoch": 0.07, + "learning_rate": 0.000199100570152843, + "loss": 0.4799, + "step": 2400 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019909927934857517, + "loss": 0.5221, + "step": 2401 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019909798762292175, + "loss": 0.6781, + "step": 2402 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990966949758947, + "loss": 0.5176, + "step": 2403 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990954014075061, + "loss": 0.492, + "step": 2404 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019909410691776796, + "loss": 0.5685, + "step": 2405 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990928115066923, + "loss": 0.5987, + "step": 2406 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019909151517429117, + "loss": 0.3969, + "step": 2407 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019909021792057665, + "loss": 0.454, + "step": 2408 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019908891974556073, + "loss": 0.6053, + "step": 2409 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019908762064925557, + "loss": 0.5772, + "step": 2410 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990863206316732, + "loss": 0.5232, + "step": 2411 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019908501969282568, + "loss": 0.5567, + "step": 2412 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019908371783272516, + "loss": 0.4834, + "step": 2413 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019908241505138375, + "loss": 0.5452, + "step": 2414 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019908111134881348, + "loss": 0.5421, + "step": 2415 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019907980672502658, + "loss": 0.6573, + "step": 2416 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019907850118003508, + "loss": 0.6335, + "step": 2417 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019907719471385117, + "loss": 0.5104, + "step": 2418 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019907588732648703, + "loss": 0.5402, + "step": 2419 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019907457901795477, + "loss": 0.6379, + "step": 2420 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019907326978826654, + "loss": 0.5197, + "step": 2421 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019907195963743453, + "loss": 0.5657, + "step": 2422 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019907064856547092, + "loss": 0.5454, + "step": 2423 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019906933657238792, + "loss": 0.5887, + "step": 2424 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019906802365819774, + "loss": 0.6211, + "step": 2425 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019906670982291252, + "loss": 0.5817, + "step": 2426 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019906539506654453, + "loss": 0.5636, + "step": 2427 + }, + { + "epoch": 0.07, + "learning_rate": 0.000199064079389106, + "loss": 0.6346, + "step": 2428 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019906276279060912, + "loss": 0.5161, + "step": 2429 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019906144527106616, + "loss": 0.6137, + "step": 2430 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019906012683048937, + "loss": 0.4409, + "step": 2431 + }, + { + "epoch": 0.07, + "learning_rate": 0.000199058807468891, + "loss": 0.6156, + "step": 2432 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990574871862833, + "loss": 0.5891, + "step": 2433 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019905616598267858, + "loss": 0.5813, + "step": 2434 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019905484385808911, + "loss": 0.594, + "step": 2435 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019905352081252717, + "loss": 0.6305, + "step": 2436 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019905219684600506, + "loss": 0.5389, + "step": 2437 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019905087195853515, + "loss": 0.6754, + "step": 2438 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990495461501297, + "loss": 0.5719, + "step": 2439 + }, + { + "epoch": 0.07, + "learning_rate": 0.000199048219420801, + "loss": 0.6946, + "step": 2440 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019904689177056145, + "loss": 0.6027, + "step": 2441 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019904556319942342, + "loss": 0.5776, + "step": 2442 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019904423370739916, + "loss": 0.6851, + "step": 2443 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019904290329450115, + "loss": 0.4978, + "step": 2444 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019904157196074163, + "loss": 0.5959, + "step": 2445 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990402397061331, + "loss": 0.6349, + "step": 2446 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990389065306879, + "loss": 0.6229, + "step": 2447 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019903757243441841, + "loss": 0.5927, + "step": 2448 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019903623741733705, + "loss": 0.5676, + "step": 2449 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990349014794562, + "loss": 0.5954, + "step": 2450 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990335646207883, + "loss": 0.629, + "step": 2451 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990322268413458, + "loss": 0.5619, + "step": 2452 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990308881411411, + "loss": 0.5159, + "step": 2453 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019902954852018675, + "loss": 0.6254, + "step": 2454 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990282079784951, + "loss": 0.53, + "step": 2455 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990268665160786, + "loss": 0.5534, + "step": 2456 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019902552413294977, + "loss": 0.6331, + "step": 2457 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990241808291211, + "loss": 0.6378, + "step": 2458 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019902283660460507, + "loss": 0.6227, + "step": 2459 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019902149145941417, + "loss": 0.5487, + "step": 2460 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990201453935609, + "loss": 0.7171, + "step": 2461 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019901879840705778, + "loss": 0.5302, + "step": 2462 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019901745049991733, + "loss": 0.5194, + "step": 2463 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990161016721521, + "loss": 0.7042, + "step": 2464 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019901475192377464, + "loss": 0.6583, + "step": 2465 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019901340125479746, + "loss": 0.5697, + "step": 2466 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019901204966523315, + "loss": 0.5659, + "step": 2467 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019901069715509427, + "loss": 0.5965, + "step": 2468 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019900934372439337, + "loss": 0.5953, + "step": 2469 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019900798937314305, + "loss": 0.5985, + "step": 2470 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019900663410135595, + "loss": 0.5011, + "step": 2471 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019900527790904456, + "loss": 0.5291, + "step": 2472 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019900392079622158, + "loss": 0.5871, + "step": 2473 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990025627628996, + "loss": 0.5377, + "step": 2474 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001990012038090913, + "loss": 0.6655, + "step": 2475 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001989998439348092, + "loss": 0.5139, + "step": 2476 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019899848314006605, + "loss": 0.5176, + "step": 2477 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019899712142487443, + "loss": 0.6352, + "step": 2478 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019899575878924705, + "loss": 0.5743, + "step": 2479 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019899439523319657, + "loss": 0.5458, + "step": 2480 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019899303075673563, + "loss": 0.5818, + "step": 2481 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019899166535987696, + "loss": 0.5755, + "step": 2482 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019899029904263323, + "loss": 0.5604, + "step": 2483 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001989889318050172, + "loss": 0.5241, + "step": 2484 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019898756364704147, + "loss": 0.7378, + "step": 2485 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019898619456871887, + "loss": 0.5574, + "step": 2486 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019898482457006206, + "loss": 0.6133, + "step": 2487 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001989834536510838, + "loss": 0.6701, + "step": 2488 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019898208181179684, + "loss": 0.6285, + "step": 2489 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019898070905221397, + "loss": 0.5033, + "step": 2490 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019897933537234787, + "loss": 0.6396, + "step": 2491 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001989779607722114, + "loss": 0.5539, + "step": 2492 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001989765852518173, + "loss": 0.5182, + "step": 2493 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001989752088111783, + "loss": 0.5721, + "step": 2494 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001989738314503073, + "loss": 0.6115, + "step": 2495 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019897245316921705, + "loss": 0.6925, + "step": 2496 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019897107396792037, + "loss": 0.6266, + "step": 2497 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019896969384643014, + "loss": 0.522, + "step": 2498 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001989683128047591, + "loss": 0.5293, + "step": 2499 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019896693084292012, + "loss": 0.6697, + "step": 2500 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019896554796092607, + "loss": 0.7282, + "step": 2501 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019896416415878982, + "loss": 0.593, + "step": 2502 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001989627794365242, + "loss": 0.7241, + "step": 2503 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019896139379414212, + "loss": 0.6713, + "step": 2504 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019896000723165641, + "loss": 0.6796, + "step": 2505 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019895861974908005, + "loss": 0.5757, + "step": 2506 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001989572313464258, + "loss": 0.597, + "step": 2507 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001989558420237067, + "loss": 0.5565, + "step": 2508 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019895445178093563, + "loss": 0.6108, + "step": 2509 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001989530606181255, + "loss": 0.4837, + "step": 2510 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019895166853528925, + "loss": 0.6471, + "step": 2511 + }, + { + "epoch": 0.07, + "learning_rate": 0.0001989502755324398, + "loss": 0.5363, + "step": 2512 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019894888160959016, + "loss": 0.6459, + "step": 2513 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019894748676675323, + "loss": 0.5564, + "step": 2514 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019894609100394201, + "loss": 0.5064, + "step": 2515 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019894469432116949, + "loss": 0.5396, + "step": 2516 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019894329671844862, + "loss": 0.5597, + "step": 2517 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019894189819579243, + "loss": 0.4991, + "step": 2518 + }, + { + "epoch": 0.07, + "learning_rate": 0.00019894049875321388, + "loss": 0.6179, + "step": 2519 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019893909839072604, + "loss": 0.5833, + "step": 2520 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019893769710834186, + "loss": 0.4787, + "step": 2521 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019893629490607442, + "loss": 0.5824, + "step": 2522 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019893489178393674, + "loss": 0.5935, + "step": 2523 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001989334877419419, + "loss": 0.6934, + "step": 2524 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019893208278010287, + "loss": 0.4845, + "step": 2525 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019893067689843278, + "loss": 0.6197, + "step": 2526 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019892927009694468, + "loss": 0.5604, + "step": 2527 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019892786237565168, + "loss": 0.5474, + "step": 2528 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019892645373456683, + "loss": 0.5726, + "step": 2529 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019892504417370327, + "loss": 0.6672, + "step": 2530 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019892363369307403, + "loss": 0.468, + "step": 2531 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019892222229269231, + "loss": 0.6583, + "step": 2532 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019892080997257116, + "loss": 0.5836, + "step": 2533 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019891939673272378, + "loss": 0.5354, + "step": 2534 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019891798257316325, + "loss": 0.6756, + "step": 2535 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019891656749390272, + "loss": 0.5481, + "step": 2536 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019891515149495544, + "loss": 0.4565, + "step": 2537 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019891373457633444, + "loss": 0.509, + "step": 2538 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019891231673805298, + "loss": 0.5882, + "step": 2539 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019891089798012424, + "loss": 0.5201, + "step": 2540 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019890947830256136, + "loss": 0.6313, + "step": 2541 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019890805770537759, + "loss": 0.6484, + "step": 2542 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019890663618858613, + "loss": 0.4766, + "step": 2543 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019890521375220016, + "loss": 0.6747, + "step": 2544 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019890379039623298, + "loss": 0.5338, + "step": 2545 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019890236612069774, + "loss": 0.5366, + "step": 2546 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001989009409256077, + "loss": 0.521, + "step": 2547 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019889951481097615, + "loss": 0.6417, + "step": 2548 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019889808777681634, + "loss": 0.5895, + "step": 2549 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019889665982314152, + "loss": 0.5021, + "step": 2550 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019889523094996498, + "loss": 0.5479, + "step": 2551 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019889380115729996, + "loss": 0.6876, + "step": 2552 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019889237044515984, + "loss": 0.5037, + "step": 2553 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019889093881355786, + "loss": 0.6661, + "step": 2554 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019888950626250737, + "loss": 0.5897, + "step": 2555 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988880727920216, + "loss": 0.4833, + "step": 2556 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019888663840211398, + "loss": 0.6693, + "step": 2557 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988852030927978, + "loss": 0.523, + "step": 2558 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019888376686408643, + "loss": 0.6326, + "step": 2559 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988823297159932, + "loss": 0.523, + "step": 2560 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019888089164853145, + "loss": 0.5281, + "step": 2561 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019887945266171462, + "loss": 0.4863, + "step": 2562 + }, + { + "epoch": 0.08, + "learning_rate": 0.000198878012755556, + "loss": 0.5291, + "step": 2563 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988765719300691, + "loss": 0.5865, + "step": 2564 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988751301852672, + "loss": 0.5812, + "step": 2565 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019887368752116374, + "loss": 0.6617, + "step": 2566 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019887224393777214, + "loss": 0.5817, + "step": 2567 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988707994351058, + "loss": 0.5945, + "step": 2568 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019886935401317827, + "loss": 0.4612, + "step": 2569 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988679076720028, + "loss": 0.6116, + "step": 2570 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019886646041159295, + "loss": 0.6123, + "step": 2571 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019886501223196216, + "loss": 0.6425, + "step": 2572 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988635631331239, + "loss": 0.5218, + "step": 2573 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019886211311509163, + "loss": 0.6272, + "step": 2574 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019886066217787883, + "loss": 0.5587, + "step": 2575 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019885921032149897, + "loss": 0.5795, + "step": 2576 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988577575459656, + "loss": 0.6224, + "step": 2577 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988563038512922, + "loss": 0.5261, + "step": 2578 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988548492374923, + "loss": 0.6667, + "step": 2579 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019885339370457938, + "loss": 0.4917, + "step": 2580 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019885193725256703, + "loss": 0.5264, + "step": 2581 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019885047988146877, + "loss": 0.5737, + "step": 2582 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019884902159129813, + "loss": 0.4706, + "step": 2583 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988475623820687, + "loss": 0.5477, + "step": 2584 + }, + { + "epoch": 0.08, + "learning_rate": 0.000198846102253794, + "loss": 0.6218, + "step": 2585 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019884464120648767, + "loss": 0.6744, + "step": 2586 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019884317924016322, + "loss": 0.5558, + "step": 2587 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019884171635483434, + "loss": 0.5723, + "step": 2588 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019884025255051456, + "loss": 0.6683, + "step": 2589 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019883878782721748, + "loss": 0.6874, + "step": 2590 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019883732218495674, + "loss": 0.5498, + "step": 2591 + }, + { + "epoch": 0.08, + "learning_rate": 0.000198835855623746, + "loss": 0.6114, + "step": 2592 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019883438814359886, + "loss": 0.4606, + "step": 2593 + }, + { + "epoch": 0.08, + "learning_rate": 0.000198832919744529, + "loss": 0.5206, + "step": 2594 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019883145042655, + "loss": 0.6136, + "step": 2595 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019882998018967555, + "loss": 0.5638, + "step": 2596 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019882850903391936, + "loss": 0.5987, + "step": 2597 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019882703695929508, + "loss": 0.4876, + "step": 2598 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988255639658164, + "loss": 0.6137, + "step": 2599 + }, + { + "epoch": 0.08, + "learning_rate": 0.000198824090053497, + "loss": 0.6377, + "step": 2600 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019882261522235063, + "loss": 0.5913, + "step": 2601 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019882113947239092, + "loss": 0.568, + "step": 2602 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988196628036317, + "loss": 0.5062, + "step": 2603 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019881818521608663, + "loss": 0.4622, + "step": 2604 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019881670670976944, + "loss": 0.5416, + "step": 2605 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988152272846939, + "loss": 0.4326, + "step": 2606 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019881374694087373, + "loss": 0.548, + "step": 2607 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019881226567832277, + "loss": 0.5636, + "step": 2608 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019881078349705473, + "loss": 0.6483, + "step": 2609 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019880930039708342, + "loss": 0.7129, + "step": 2610 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988078163784226, + "loss": 0.6362, + "step": 2611 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988063314410861, + "loss": 0.5243, + "step": 2612 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019880484558508767, + "loss": 0.5938, + "step": 2613 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001988033588104412, + "loss": 0.6089, + "step": 2614 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019880187111716047, + "loss": 0.5535, + "step": 2615 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019880038250525935, + "loss": 0.6966, + "step": 2616 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001987988929747516, + "loss": 0.6342, + "step": 2617 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019879740252565114, + "loss": 0.7131, + "step": 2618 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001987959111579718, + "loss": 0.6247, + "step": 2619 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001987944188717275, + "loss": 0.4871, + "step": 2620 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019879292566693204, + "loss": 0.6405, + "step": 2621 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001987914315435993, + "loss": 0.5902, + "step": 2622 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019878993650174326, + "loss": 0.5504, + "step": 2623 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019878844054137772, + "loss": 0.6202, + "step": 2624 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019878694366251668, + "loss": 0.6643, + "step": 2625 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019878544586517395, + "loss": 0.6127, + "step": 2626 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019878394714936353, + "loss": 0.5481, + "step": 2627 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019878244751509938, + "loss": 0.5999, + "step": 2628 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019878094696239537, + "loss": 0.5289, + "step": 2629 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019877944549126552, + "loss": 0.4943, + "step": 2630 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019877794310172372, + "loss": 0.7803, + "step": 2631 + }, + { + "epoch": 0.08, + "learning_rate": 0.000198776439793784, + "loss": 0.6477, + "step": 2632 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019877493556746029, + "loss": 0.64, + "step": 2633 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001987734304227666, + "loss": 0.6612, + "step": 2634 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019877192435971693, + "loss": 0.6005, + "step": 2635 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019877041737832526, + "loss": 0.5518, + "step": 2636 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019876890947860562, + "loss": 0.4803, + "step": 2637 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019876740066057202, + "loss": 0.4978, + "step": 2638 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019876589092423852, + "loss": 0.5358, + "step": 2639 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019876438026961913, + "loss": 0.6669, + "step": 2640 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019876286869672787, + "loss": 0.5606, + "step": 2641 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019876135620557883, + "loss": 0.5785, + "step": 2642 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019875984279618604, + "loss": 0.5322, + "step": 2643 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019875832846856362, + "loss": 0.5014, + "step": 2644 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019875681322272564, + "loss": 0.6255, + "step": 2645 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001987552970586861, + "loss": 0.5319, + "step": 2646 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019875377997645926, + "loss": 0.6045, + "step": 2647 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019875226197605908, + "loss": 0.5115, + "step": 2648 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019875074305749973, + "loss": 0.6316, + "step": 2649 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019874922322079535, + "loss": 0.6225, + "step": 2650 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019874770246596003, + "loss": 0.4997, + "step": 2651 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019874618079300792, + "loss": 0.5837, + "step": 2652 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001987446582019532, + "loss": 0.602, + "step": 2653 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019874313469281, + "loss": 0.6963, + "step": 2654 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019874161026559248, + "loss": 0.5055, + "step": 2655 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019874008492031485, + "loss": 0.5291, + "step": 2656 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019873855865699123, + "loss": 0.5939, + "step": 2657 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019873703147563586, + "loss": 0.5965, + "step": 2658 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019873550337626291, + "loss": 0.5645, + "step": 2659 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001987339743588866, + "loss": 0.5388, + "step": 2660 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001987324444235212, + "loss": 0.5066, + "step": 2661 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019873091357018084, + "loss": 0.6011, + "step": 2662 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019872938179887978, + "loss": 0.5636, + "step": 2663 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019872784910963232, + "loss": 0.5353, + "step": 2664 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019872631550245262, + "loss": 0.5454, + "step": 2665 + }, + { + "epoch": 0.08, + "learning_rate": 0.000198724780977355, + "loss": 0.7244, + "step": 2666 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019872324553435374, + "loss": 0.6162, + "step": 2667 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001987217091734631, + "loss": 0.6021, + "step": 2668 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019872017189469728, + "loss": 0.5138, + "step": 2669 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019871863369807073, + "loss": 0.5129, + "step": 2670 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019871709458359763, + "loss": 0.6776, + "step": 2671 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019871555455129234, + "loss": 0.6105, + "step": 2672 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019871401360116916, + "loss": 0.5776, + "step": 2673 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019871247173324243, + "loss": 0.6925, + "step": 2674 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019871092894752649, + "loss": 0.7074, + "step": 2675 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019870938524403566, + "loss": 0.6546, + "step": 2676 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019870784062278433, + "loss": 0.6007, + "step": 2677 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001987062950837868, + "loss": 0.6583, + "step": 2678 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001987047486270575, + "loss": 0.5886, + "step": 2679 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001987032012526108, + "loss": 0.5429, + "step": 2680 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019870165296046106, + "loss": 0.5949, + "step": 2681 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019870010375062268, + "loss": 0.6683, + "step": 2682 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001986985536231101, + "loss": 0.4702, + "step": 2683 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019869700257793769, + "loss": 0.6733, + "step": 2684 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001986954506151199, + "loss": 0.5813, + "step": 2685 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001986938977346711, + "loss": 0.6051, + "step": 2686 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019869234393660583, + "loss": 0.5216, + "step": 2687 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019869078922093848, + "loss": 0.5886, + "step": 2688 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019868923358768345, + "loss": 0.6207, + "step": 2689 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001986876770368553, + "loss": 0.5347, + "step": 2690 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019868611956846846, + "loss": 0.577, + "step": 2691 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019868456118253742, + "loss": 0.5009, + "step": 2692 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019868300187907666, + "loss": 0.6811, + "step": 2693 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019868144165810067, + "loss": 0.5668, + "step": 2694 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019867988051962396, + "loss": 0.5775, + "step": 2695 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019867831846366105, + "loss": 0.5864, + "step": 2696 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019867675549022648, + "loss": 0.6008, + "step": 2697 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019867519159933478, + "loss": 0.5281, + "step": 2698 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019867362679100044, + "loss": 0.513, + "step": 2699 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019867206106523807, + "loss": 0.6297, + "step": 2700 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019867049442206218, + "loss": 0.6155, + "step": 2701 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019866892686148737, + "loss": 0.563, + "step": 2702 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001986673583835282, + "loss": 0.541, + "step": 2703 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001986657889881993, + "loss": 0.4257, + "step": 2704 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019866421867551515, + "loss": 0.6151, + "step": 2705 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019866264744549045, + "loss": 0.6128, + "step": 2706 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019866107529813978, + "loss": 0.6224, + "step": 2707 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019865950223347774, + "loss": 0.6007, + "step": 2708 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019865792825151898, + "loss": 0.5582, + "step": 2709 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019865635335227815, + "loss": 0.55, + "step": 2710 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019865477753576983, + "loss": 0.6262, + "step": 2711 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019865320080200873, + "loss": 0.5921, + "step": 2712 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019865162315100946, + "loss": 0.556, + "step": 2713 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019865004458278677, + "loss": 0.5462, + "step": 2714 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019864846509735521, + "loss": 0.5856, + "step": 2715 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019864688469472958, + "loss": 0.5551, + "step": 2716 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019864530337492453, + "loss": 0.5298, + "step": 2717 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019864372113795477, + "loss": 0.4195, + "step": 2718 + }, + { + "epoch": 0.08, + "learning_rate": 0.000198642137983835, + "loss": 0.5806, + "step": 2719 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019864055391257996, + "loss": 0.672, + "step": 2720 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019863896892420436, + "loss": 0.4915, + "step": 2721 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019863738301872293, + "loss": 0.6474, + "step": 2722 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019863579619615042, + "loss": 0.6486, + "step": 2723 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001986342084565016, + "loss": 0.5406, + "step": 2724 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001986326197997912, + "loss": 0.522, + "step": 2725 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019863103022603405, + "loss": 0.5093, + "step": 2726 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019862943973524488, + "loss": 0.711, + "step": 2727 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019862784832743848, + "loss": 0.5469, + "step": 2728 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019862625600262966, + "loss": 0.597, + "step": 2729 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001986246627608332, + "loss": 0.4541, + "step": 2730 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019862306860206396, + "loss": 0.6006, + "step": 2731 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019862147352633672, + "loss": 0.5782, + "step": 2732 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019861987753366634, + "loss": 0.5893, + "step": 2733 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019861828062406765, + "loss": 0.5815, + "step": 2734 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019861668279755549, + "loss": 0.5556, + "step": 2735 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001986150840541447, + "loss": 0.5555, + "step": 2736 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019861348439385016, + "loss": 0.5642, + "step": 2737 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019861188381668673, + "loss": 0.6109, + "step": 2738 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019861028232266936, + "loss": 0.638, + "step": 2739 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019860867991181285, + "loss": 0.5896, + "step": 2740 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019860707658413216, + "loss": 0.5775, + "step": 2741 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019860547233964214, + "loss": 0.5411, + "step": 2742 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019860386717835776, + "loss": 0.5436, + "step": 2743 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019860226110029393, + "loss": 0.5141, + "step": 2744 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019860065410546557, + "loss": 0.6587, + "step": 2745 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019859904619388762, + "loss": 0.4846, + "step": 2746 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019859743736557503, + "loss": 0.4962, + "step": 2747 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019859582762054277, + "loss": 0.6553, + "step": 2748 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001985942169588058, + "loss": 0.547, + "step": 2749 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019859260538037908, + "loss": 0.7094, + "step": 2750 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019859099288527764, + "loss": 0.6665, + "step": 2751 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019858937947351643, + "loss": 0.6647, + "step": 2752 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019858776514511048, + "loss": 0.5633, + "step": 2753 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019858614990007477, + "loss": 0.4822, + "step": 2754 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019858453373842433, + "loss": 0.5323, + "step": 2755 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001985829166601742, + "loss": 0.5902, + "step": 2756 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001985812986653394, + "loss": 0.5721, + "step": 2757 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019857967975393494, + "loss": 0.484, + "step": 2758 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019857805992597595, + "loss": 0.5804, + "step": 2759 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001985764391814774, + "loss": 0.6259, + "step": 2760 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019857481752045446, + "loss": 0.5345, + "step": 2761 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001985731949429221, + "loss": 0.6286, + "step": 2762 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001985715714488955, + "loss": 0.5671, + "step": 2763 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019856994703838974, + "loss": 0.5601, + "step": 2764 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019856832171141986, + "loss": 0.5052, + "step": 2765 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019856669546800101, + "loss": 0.5779, + "step": 2766 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019856506830814833, + "loss": 0.603, + "step": 2767 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019856344023187693, + "loss": 0.5278, + "step": 2768 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001985618112392019, + "loss": 0.6445, + "step": 2769 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019856018133013848, + "loss": 0.5245, + "step": 2770 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019855855050470177, + "loss": 0.6285, + "step": 2771 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019855691876290691, + "loss": 0.5369, + "step": 2772 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019855528610476913, + "loss": 0.5957, + "step": 2773 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019855365253030355, + "loss": 0.5865, + "step": 2774 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019855201803952545, + "loss": 0.5956, + "step": 2775 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019855038263244992, + "loss": 0.495, + "step": 2776 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001985487463090922, + "loss": 0.5474, + "step": 2777 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019854710906946755, + "loss": 0.5216, + "step": 2778 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001985454709135911, + "loss": 0.6032, + "step": 2779 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001985438318414782, + "loss": 0.7243, + "step": 2780 + }, + { + "epoch": 0.08, + "learning_rate": 0.000198542191853144, + "loss": 0.5003, + "step": 2781 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019854055094860375, + "loss": 0.4954, + "step": 2782 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019853890912787279, + "loss": 0.684, + "step": 2783 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001985372663909663, + "loss": 0.5701, + "step": 2784 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019853562273789956, + "loss": 0.5422, + "step": 2785 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019853397816868787, + "loss": 0.7356, + "step": 2786 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019853233268334654, + "loss": 0.4535, + "step": 2787 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019853068628189087, + "loss": 0.8021, + "step": 2788 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019852903896433612, + "loss": 0.5191, + "step": 2789 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019852739073069765, + "loss": 0.5602, + "step": 2790 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019852574158099078, + "loss": 0.5345, + "step": 2791 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001985240915152308, + "loss": 0.5336, + "step": 2792 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019852244053343313, + "loss": 0.5411, + "step": 2793 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019852078863561302, + "loss": 0.5119, + "step": 2794 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019851913582178595, + "loss": 0.5559, + "step": 2795 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019851748209196717, + "loss": 0.5161, + "step": 2796 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019851582744617212, + "loss": 0.5559, + "step": 2797 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019851417188441616, + "loss": 0.5341, + "step": 2798 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019851251540671473, + "loss": 0.5291, + "step": 2799 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019851085801308313, + "loss": 0.5394, + "step": 2800 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001985091997035369, + "loss": 0.4747, + "step": 2801 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019850754047809135, + "loss": 0.6258, + "step": 2802 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019850588033676198, + "loss": 0.59, + "step": 2803 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019850421927956416, + "loss": 0.6802, + "step": 2804 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019850255730651342, + "loss": 0.5658, + "step": 2805 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019850089441762514, + "loss": 0.5224, + "step": 2806 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001984992306129148, + "loss": 0.5573, + "step": 2807 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019849756589239786, + "loss": 0.5654, + "step": 2808 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001984959002560898, + "loss": 0.6196, + "step": 2809 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019849423370400614, + "loss": 0.5708, + "step": 2810 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019849256623616235, + "loss": 0.5139, + "step": 2811 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019849089785257392, + "loss": 0.5589, + "step": 2812 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001984892285532564, + "loss": 0.5407, + "step": 2813 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019848755833822525, + "loss": 0.5479, + "step": 2814 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001984858872074961, + "loss": 0.5693, + "step": 2815 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019848421516108436, + "loss": 0.4904, + "step": 2816 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019848254219900565, + "loss": 0.5903, + "step": 2817 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019848086832127554, + "loss": 0.6189, + "step": 2818 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019847919352790957, + "loss": 0.6048, + "step": 2819 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019847751781892328, + "loss": 0.5414, + "step": 2820 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001984758411943323, + "loss": 0.5217, + "step": 2821 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019847416365415216, + "loss": 0.5049, + "step": 2822 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019847248519839851, + "loss": 0.562, + "step": 2823 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019847080582708696, + "loss": 0.6678, + "step": 2824 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001984691255402331, + "loss": 0.6177, + "step": 2825 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019846744433785256, + "loss": 0.5134, + "step": 2826 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019846576221996097, + "loss": 0.5679, + "step": 2827 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019846407918657394, + "loss": 0.6308, + "step": 2828 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019846239523770714, + "loss": 0.6294, + "step": 2829 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019846071037337625, + "loss": 0.7646, + "step": 2830 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019845902459359694, + "loss": 0.5243, + "step": 2831 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019845733789838486, + "loss": 0.6187, + "step": 2832 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019845565028775566, + "loss": 0.6774, + "step": 2833 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019845396176172508, + "loss": 0.6836, + "step": 2834 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001984522723203088, + "loss": 0.6036, + "step": 2835 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019845058196352256, + "loss": 0.5456, + "step": 2836 + }, + { + "epoch": 0.08, + "learning_rate": 0.000198448890691382, + "loss": 0.5128, + "step": 2837 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001984471985039029, + "loss": 0.5082, + "step": 2838 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019844550540110102, + "loss": 0.6337, + "step": 2839 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019844381138299204, + "loss": 0.6133, + "step": 2840 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019844211644959176, + "loss": 0.577, + "step": 2841 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001984404206009159, + "loss": 0.6479, + "step": 2842 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019843872383698024, + "loss": 0.5241, + "step": 2843 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019843702615780057, + "loss": 0.5624, + "step": 2844 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019843532756339265, + "loss": 0.5465, + "step": 2845 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019843362805377227, + "loss": 0.5323, + "step": 2846 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019843192762895528, + "loss": 0.5694, + "step": 2847 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019843022628895747, + "loss": 0.6256, + "step": 2848 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019842852403379463, + "loss": 0.5862, + "step": 2849 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001984268208634826, + "loss": 0.5782, + "step": 2850 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001984251167780372, + "loss": 0.4655, + "step": 2851 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019842341177747432, + "loss": 0.6224, + "step": 2852 + }, + { + "epoch": 0.08, + "learning_rate": 0.0001984217058618098, + "loss": 0.5915, + "step": 2853 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019841999903105946, + "loss": 0.5238, + "step": 2854 + }, + { + "epoch": 0.08, + "learning_rate": 0.00019841829128523918, + "loss": 0.5077, + "step": 2855 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019841658262436486, + "loss": 0.5922, + "step": 2856 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001984148730484524, + "loss": 0.5947, + "step": 2857 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019841316255751767, + "loss": 0.5074, + "step": 2858 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019841145115157658, + "loss": 0.56, + "step": 2859 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019840973883064505, + "loss": 0.6049, + "step": 2860 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019840802559473896, + "loss": 0.669, + "step": 2861 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019840631144387428, + "loss": 0.5359, + "step": 2862 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019840459637806697, + "loss": 0.5265, + "step": 2863 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001984028803973329, + "loss": 0.4657, + "step": 2864 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019840116350168806, + "loss": 0.4324, + "step": 2865 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019839944569114847, + "loss": 0.5795, + "step": 2866 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019839772696573, + "loss": 0.6724, + "step": 2867 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019839600732544872, + "loss": 0.5715, + "step": 2868 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019839428677032058, + "loss": 0.5531, + "step": 2869 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019839256530036154, + "loss": 0.5926, + "step": 2870 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019839084291558767, + "loss": 0.63, + "step": 2871 + }, + { + "epoch": 0.09, + "learning_rate": 0.000198389119616015, + "loss": 0.6671, + "step": 2872 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019838739540165942, + "loss": 0.5166, + "step": 2873 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001983856702725371, + "loss": 0.7077, + "step": 2874 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019838394422866404, + "loss": 0.611, + "step": 2875 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001983822172700563, + "loss": 0.6655, + "step": 2876 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019838048939672986, + "loss": 0.4697, + "step": 2877 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019837876060870086, + "loss": 0.5808, + "step": 2878 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019837703090598536, + "loss": 0.6325, + "step": 2879 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019837530028859942, + "loss": 0.4332, + "step": 2880 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019837356875655914, + "loss": 0.5524, + "step": 2881 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019837183630988064, + "loss": 0.5226, + "step": 2882 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019837010294858004, + "loss": 0.5717, + "step": 2883 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001983683686726734, + "loss": 0.6449, + "step": 2884 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019836663348217688, + "loss": 0.5259, + "step": 2885 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019836489737710662, + "loss": 0.5782, + "step": 2886 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019836316035747872, + "loss": 0.5473, + "step": 2887 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019836142242330937, + "loss": 0.6436, + "step": 2888 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019835968357461472, + "loss": 0.6324, + "step": 2889 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019835794381141094, + "loss": 0.6358, + "step": 2890 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019835620313371418, + "loss": 0.5801, + "step": 2891 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019835446154154067, + "loss": 0.5756, + "step": 2892 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019835271903490654, + "loss": 0.4562, + "step": 2893 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019835097561382808, + "loss": 0.5932, + "step": 2894 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001983492312783214, + "loss": 0.596, + "step": 2895 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019834748602840283, + "loss": 0.6072, + "step": 2896 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019834573986408848, + "loss": 0.5607, + "step": 2897 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019834399278539463, + "loss": 0.5463, + "step": 2898 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019834224479233756, + "loss": 0.449, + "step": 2899 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019834049588493349, + "loss": 0.6671, + "step": 2900 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019833874606319866, + "loss": 0.5926, + "step": 2901 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001983369953271494, + "loss": 0.4738, + "step": 2902 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001983352436768019, + "loss": 0.5877, + "step": 2903 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019833349111217255, + "loss": 0.5473, + "step": 2904 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019833173763327756, + "loss": 0.5127, + "step": 2905 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019832998324013326, + "loss": 0.5616, + "step": 2906 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019832822793275598, + "loss": 0.5798, + "step": 2907 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019832647171116202, + "loss": 0.6732, + "step": 2908 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019832471457536773, + "loss": 0.606, + "step": 2909 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001983229565253894, + "loss": 0.4327, + "step": 2910 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019832119756124346, + "loss": 0.59, + "step": 2911 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019831943768294617, + "loss": 0.6131, + "step": 2912 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019831767689051396, + "loss": 0.5717, + "step": 2913 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019831591518396316, + "loss": 0.5731, + "step": 2914 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019831415256331016, + "loss": 0.5939, + "step": 2915 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019831238902857137, + "loss": 0.5152, + "step": 2916 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019831062457976314, + "loss": 0.5393, + "step": 2917 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019830885921690193, + "loss": 0.6332, + "step": 2918 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019830709294000416, + "loss": 0.6452, + "step": 2919 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019830532574908615, + "loss": 0.5496, + "step": 2920 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019830355764416445, + "loss": 0.5887, + "step": 2921 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019830178862525544, + "loss": 0.5795, + "step": 2922 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019830001869237558, + "loss": 0.6913, + "step": 2923 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019829824784554134, + "loss": 0.5135, + "step": 2924 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019829647608476917, + "loss": 0.6104, + "step": 2925 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001982947034100755, + "loss": 0.5591, + "step": 2926 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001982929298214769, + "loss": 0.5616, + "step": 2927 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019829115531898976, + "loss": 0.5541, + "step": 2928 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001982893799026307, + "loss": 0.5807, + "step": 2929 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019828760357241612, + "loss": 0.5033, + "step": 2930 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019828582632836256, + "loss": 0.5462, + "step": 2931 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019828404817048658, + "loss": 0.5222, + "step": 2932 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001982822690988047, + "loss": 0.6241, + "step": 2933 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019828048911333345, + "loss": 0.6137, + "step": 2934 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019827870821408937, + "loss": 0.5698, + "step": 2935 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019827692640108906, + "loss": 0.5218, + "step": 2936 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019827514367434904, + "loss": 0.5999, + "step": 2937 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001982733600338859, + "loss": 0.6323, + "step": 2938 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019827157547971623, + "loss": 0.5595, + "step": 2939 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019826979001185663, + "loss": 0.5533, + "step": 2940 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019826800363032367, + "loss": 0.6491, + "step": 2941 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019826621633513395, + "loss": 0.5329, + "step": 2942 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019826442812630418, + "loss": 0.6641, + "step": 2943 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019826263900385088, + "loss": 0.5346, + "step": 2944 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019826084896779072, + "loss": 0.6114, + "step": 2945 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001982590580181404, + "loss": 0.5443, + "step": 2946 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019825726615491646, + "loss": 0.5689, + "step": 2947 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019825547337813565, + "loss": 0.6328, + "step": 2948 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019825367968781462, + "loss": 0.5538, + "step": 2949 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019825188508397005, + "loss": 0.585, + "step": 2950 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019825008956661857, + "loss": 0.5897, + "step": 2951 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019824829313577693, + "loss": 0.6107, + "step": 2952 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019824649579146183, + "loss": 0.704, + "step": 2953 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019824469753368991, + "loss": 0.4803, + "step": 2954 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019824289836247803, + "loss": 0.468, + "step": 2955 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019824109827784277, + "loss": 0.4928, + "step": 2956 + }, + { + "epoch": 0.09, + "learning_rate": 0.000198239297279801, + "loss": 0.4943, + "step": 2957 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019823749536836935, + "loss": 0.5529, + "step": 2958 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019823569254356465, + "loss": 0.4368, + "step": 2959 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019823388880540362, + "loss": 0.5099, + "step": 2960 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019823208415390304, + "loss": 0.5217, + "step": 2961 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001982302785890797, + "loss": 0.5674, + "step": 2962 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019822847211095038, + "loss": 0.6155, + "step": 2963 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019822666471953188, + "loss": 0.624, + "step": 2964 + }, + { + "epoch": 0.09, + "learning_rate": 0.000198224856414841, + "loss": 0.5698, + "step": 2965 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019822304719689453, + "loss": 0.6026, + "step": 2966 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019822123706570935, + "loss": 0.5638, + "step": 2967 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019821942602130224, + "loss": 0.5354, + "step": 2968 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019821761406369005, + "loss": 0.6804, + "step": 2969 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019821580119288963, + "loss": 0.5386, + "step": 2970 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019821398740891782, + "loss": 0.5272, + "step": 2971 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019821217271179153, + "loss": 0.56, + "step": 2972 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001982103571015276, + "loss": 0.5825, + "step": 2973 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001982085405781429, + "loss": 0.4834, + "step": 2974 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001982067231416543, + "loss": 0.6985, + "step": 2975 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019820490479207877, + "loss": 0.575, + "step": 2976 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019820308552943319, + "loss": 0.5424, + "step": 2977 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019820126535373444, + "loss": 0.5976, + "step": 2978 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019819944426499947, + "loss": 0.6109, + "step": 2979 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001981976222632452, + "loss": 0.6854, + "step": 2980 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019819579934848855, + "loss": 0.5453, + "step": 2981 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019819397552074653, + "loss": 0.5729, + "step": 2982 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019819215078003603, + "loss": 0.5664, + "step": 2983 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001981903251263741, + "loss": 0.5427, + "step": 2984 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001981884985597776, + "loss": 0.4878, + "step": 2985 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019818667108026361, + "loss": 0.5728, + "step": 2986 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001981848426878491, + "loss": 0.5835, + "step": 2987 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019818301338255103, + "loss": 0.6568, + "step": 2988 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019818118316438645, + "loss": 0.6521, + "step": 2989 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019817935203337236, + "loss": 0.5164, + "step": 2990 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019817751998952576, + "loss": 0.5767, + "step": 2991 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001981756870328637, + "loss": 0.6022, + "step": 2992 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019817385316340326, + "loss": 0.6177, + "step": 2993 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019817201838116146, + "loss": 0.5243, + "step": 2994 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019817018268615538, + "loss": 0.5248, + "step": 2995 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019816834607840202, + "loss": 0.5136, + "step": 2996 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019816650855791851, + "loss": 0.5207, + "step": 2997 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019816467012472198, + "loss": 0.5487, + "step": 2998 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019816283077882943, + "loss": 0.5689, + "step": 2999 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019816099052025803, + "loss": 0.5944, + "step": 3000 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019815914934902482, + "loss": 0.466, + "step": 3001 + }, + { + "epoch": 0.09, + "learning_rate": 0.000198157307265147, + "loss": 0.5036, + "step": 3002 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019815546426864164, + "loss": 0.7083, + "step": 3003 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019815362035952591, + "loss": 0.4932, + "step": 3004 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019815177553781692, + "loss": 0.552, + "step": 3005 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019814992980353186, + "loss": 0.6439, + "step": 3006 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019814808315668785, + "loss": 0.5914, + "step": 3007 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001981462355973021, + "loss": 0.5679, + "step": 3008 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019814438712539175, + "loss": 0.5493, + "step": 3009 + }, + { + "epoch": 0.09, + "learning_rate": 0.000198142537740974, + "loss": 0.5358, + "step": 3010 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019814068744406605, + "loss": 0.6244, + "step": 3011 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019813883623468514, + "loss": 0.6332, + "step": 3012 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001981369841128484, + "loss": 0.6418, + "step": 3013 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019813513107857315, + "loss": 0.7051, + "step": 3014 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019813327713187652, + "loss": 0.5319, + "step": 3015 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019813142227277582, + "loss": 0.5668, + "step": 3016 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019812956650128825, + "loss": 0.7144, + "step": 3017 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019812770981743108, + "loss": 0.5972, + "step": 3018 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019812585222122158, + "loss": 0.5593, + "step": 3019 + }, + { + "epoch": 0.09, + "learning_rate": 0.000198123993712677, + "loss": 0.4967, + "step": 3020 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019812213429181466, + "loss": 0.6348, + "step": 3021 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001981202739586518, + "loss": 0.632, + "step": 3022 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019811841271320574, + "loss": 0.5451, + "step": 3023 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001981165505554938, + "loss": 0.611, + "step": 3024 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019811468748553326, + "loss": 0.6983, + "step": 3025 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001981128235033415, + "loss": 0.6586, + "step": 3026 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019811095860893576, + "loss": 0.6463, + "step": 3027 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019810909280233344, + "loss": 0.5299, + "step": 3028 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001981072260835519, + "loss": 0.5849, + "step": 3029 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019810535845260846, + "loss": 0.5519, + "step": 3030 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019810348990952051, + "loss": 0.5567, + "step": 3031 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001981016204543054, + "loss": 0.6003, + "step": 3032 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001980997500869805, + "loss": 0.5854, + "step": 3033 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019809787880756328, + "loss": 0.6241, + "step": 3034 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019809600661607104, + "loss": 0.4762, + "step": 3035 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019809413351252125, + "loss": 0.5447, + "step": 3036 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019809225949693127, + "loss": 0.5187, + "step": 3037 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019809038456931858, + "loss": 0.536, + "step": 3038 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019808850872970058, + "loss": 0.5631, + "step": 3039 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019808663197809473, + "loss": 0.5399, + "step": 3040 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019808475431451846, + "loss": 0.605, + "step": 3041 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019808287573898925, + "loss": 0.5115, + "step": 3042 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019808099625152456, + "loss": 0.6598, + "step": 3043 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019807911585214183, + "loss": 0.6347, + "step": 3044 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019807723454085858, + "loss": 0.5407, + "step": 3045 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001980753523176923, + "loss": 0.5343, + "step": 3046 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019807346918266047, + "loss": 0.6175, + "step": 3047 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019807158513578064, + "loss": 0.5625, + "step": 3048 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019806970017707028, + "loss": 0.6304, + "step": 3049 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019806781430654693, + "loss": 0.4836, + "step": 3050 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001980659275242281, + "loss": 0.5448, + "step": 3051 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019806403983013145, + "loss": 0.642, + "step": 3052 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019806215122427436, + "loss": 0.5795, + "step": 3053 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019806026170667452, + "loss": 0.5424, + "step": 3054 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019805837127734942, + "loss": 0.5376, + "step": 3055 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019805647993631666, + "loss": 0.61, + "step": 3056 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019805458768359384, + "loss": 0.6864, + "step": 3057 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019805269451919857, + "loss": 0.4946, + "step": 3058 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001980508004431484, + "loss": 0.5143, + "step": 3059 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019804890545546097, + "loss": 0.6286, + "step": 3060 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001980470095561539, + "loss": 0.5368, + "step": 3061 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019804511274524483, + "loss": 0.6732, + "step": 3062 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019804321502275133, + "loss": 0.5307, + "step": 3063 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019804131638869114, + "loss": 0.5478, + "step": 3064 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019803941684308184, + "loss": 0.5451, + "step": 3065 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019803751638594112, + "loss": 0.6096, + "step": 3066 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019803561501728666, + "loss": 0.5269, + "step": 3067 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019803371273713612, + "loss": 0.5352, + "step": 3068 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001980318095455072, + "loss": 0.5022, + "step": 3069 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019802990544241756, + "loss": 0.4561, + "step": 3070 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019802800042788494, + "loss": 0.5849, + "step": 3071 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019802609450192703, + "loss": 0.5199, + "step": 3072 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001980241876645616, + "loss": 0.5652, + "step": 3073 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019802227991580634, + "loss": 0.5625, + "step": 3074 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019802037125567895, + "loss": 0.5772, + "step": 3075 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019801846168419725, + "loss": 0.5414, + "step": 3076 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019801655120137892, + "loss": 0.6209, + "step": 3077 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001980146398072418, + "loss": 0.6184, + "step": 3078 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001980127275018036, + "loss": 0.6209, + "step": 3079 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019801081428508214, + "loss": 0.5694, + "step": 3080 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001980089001570952, + "loss": 0.5412, + "step": 3081 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019800698511786052, + "loss": 0.5104, + "step": 3082 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019800506916739603, + "loss": 0.5812, + "step": 3083 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019800315230571942, + "loss": 0.4573, + "step": 3084 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019800123453284858, + "loss": 0.4989, + "step": 3085 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979993158488013, + "loss": 0.5211, + "step": 3086 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019799739625359544, + "loss": 0.6955, + "step": 3087 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019799547574724885, + "loss": 0.531, + "step": 3088 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019799355432977938, + "loss": 0.5055, + "step": 3089 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979916320012049, + "loss": 0.5648, + "step": 3090 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979897087615433, + "loss": 0.6171, + "step": 3091 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019798778461081244, + "loss": 0.5669, + "step": 3092 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019798585954903023, + "loss": 0.5081, + "step": 3093 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979839335762145, + "loss": 0.5333, + "step": 3094 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019798200669238325, + "loss": 0.5726, + "step": 3095 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019798007889755431, + "loss": 0.7047, + "step": 3096 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979781501917457, + "loss": 0.4924, + "step": 3097 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019797622057497527, + "loss": 0.5337, + "step": 3098 + }, + { + "epoch": 0.09, + "learning_rate": 0.000197974290047261, + "loss": 0.6234, + "step": 3099 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019797235860862083, + "loss": 0.4583, + "step": 3100 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979704262590727, + "loss": 0.6354, + "step": 3101 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979684929986346, + "loss": 0.5131, + "step": 3102 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979665588273245, + "loss": 0.5431, + "step": 3103 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979646237451604, + "loss": 0.6878, + "step": 3104 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019796268775216028, + "loss": 0.5447, + "step": 3105 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979607508483421, + "loss": 0.6029, + "step": 3106 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979588130337239, + "loss": 0.5798, + "step": 3107 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019795687430832374, + "loss": 0.6334, + "step": 3108 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019795493467215954, + "loss": 0.627, + "step": 3109 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019795299412524945, + "loss": 0.5721, + "step": 3110 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019795105266761146, + "loss": 0.6502, + "step": 3111 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019794911029926362, + "loss": 0.5539, + "step": 3112 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019794716702022398, + "loss": 0.6169, + "step": 3113 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979452228305106, + "loss": 0.7411, + "step": 3114 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019794327773014158, + "loss": 0.6961, + "step": 3115 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019794133171913502, + "loss": 0.6045, + "step": 3116 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019793938479750902, + "loss": 0.615, + "step": 3117 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019793743696528158, + "loss": 0.6834, + "step": 3118 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019793548822247096, + "loss": 0.5851, + "step": 3119 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019793353856909515, + "loss": 0.6305, + "step": 3120 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019793158800517237, + "loss": 0.6903, + "step": 3121 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979296365307207, + "loss": 0.6338, + "step": 3122 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019792768414575828, + "loss": 0.6191, + "step": 3123 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019792573085030332, + "loss": 0.5181, + "step": 3124 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019792377664437392, + "loss": 0.5228, + "step": 3125 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979218215279883, + "loss": 0.6483, + "step": 3126 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019791986550116457, + "loss": 0.63, + "step": 3127 + }, + { + "epoch": 0.09, + "learning_rate": 0.000197917908563921, + "loss": 0.5957, + "step": 3128 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979159507162757, + "loss": 0.5778, + "step": 3129 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019791399195824693, + "loss": 0.5815, + "step": 3130 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979120322898529, + "loss": 0.4363, + "step": 3131 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019791007171111183, + "loss": 0.6075, + "step": 3132 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019790811022204192, + "loss": 0.6431, + "step": 3133 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019790614782266142, + "loss": 0.6798, + "step": 3134 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019790418451298857, + "loss": 0.4998, + "step": 3135 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019790222029304164, + "loss": 0.5916, + "step": 3136 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001979002551628389, + "loss": 0.6518, + "step": 3137 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978982891223986, + "loss": 0.5089, + "step": 3138 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019789632217173902, + "loss": 0.5658, + "step": 3139 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019789435431087843, + "loss": 0.5544, + "step": 3140 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978923855398352, + "loss": 0.6117, + "step": 3141 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019789041585862755, + "loss": 0.5286, + "step": 3142 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019788844526727383, + "loss": 0.6004, + "step": 3143 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978864737657924, + "loss": 0.5754, + "step": 3144 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978845013542015, + "loss": 0.5691, + "step": 3145 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019788252803251956, + "loss": 0.5475, + "step": 3146 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019788055380076487, + "loss": 0.5291, + "step": 3147 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019787857865895579, + "loss": 0.6044, + "step": 3148 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019787660260711068, + "loss": 0.5279, + "step": 3149 + }, + { + "epoch": 0.09, + "learning_rate": 0.000197874625645248, + "loss": 0.6198, + "step": 3150 + }, + { + "epoch": 0.09, + "learning_rate": 0.000197872647773386, + "loss": 0.5537, + "step": 3151 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019787066899154315, + "loss": 0.5669, + "step": 3152 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978686892997378, + "loss": 0.5352, + "step": 3153 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978667086979884, + "loss": 0.6681, + "step": 3154 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019786472718631335, + "loss": 0.5604, + "step": 3155 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019786274476473107, + "loss": 0.572, + "step": 3156 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019786076143326, + "loss": 0.6023, + "step": 3157 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978587771919186, + "loss": 0.5987, + "step": 3158 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019785679204072524, + "loss": 0.555, + "step": 3159 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019785480597969848, + "loss": 0.5613, + "step": 3160 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978528190088567, + "loss": 0.5237, + "step": 3161 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978508311282184, + "loss": 0.5538, + "step": 3162 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978488423378021, + "loss": 0.5626, + "step": 3163 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019784685263762625, + "loss": 0.5922, + "step": 3164 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019784486202770936, + "loss": 0.6294, + "step": 3165 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019784287050806995, + "loss": 0.5316, + "step": 3166 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978408780787265, + "loss": 0.7551, + "step": 3167 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978388847396976, + "loss": 0.5335, + "step": 3168 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019783689049100175, + "loss": 0.4354, + "step": 3169 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978348953326574, + "loss": 0.5404, + "step": 3170 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019783289926468328, + "loss": 0.6201, + "step": 3171 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019783090228709782, + "loss": 0.6543, + "step": 3172 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978289043999196, + "loss": 0.6344, + "step": 3173 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019782690560316728, + "loss": 0.6199, + "step": 3174 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019782490589685933, + "loss": 0.527, + "step": 3175 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978229052810144, + "loss": 0.5976, + "step": 3176 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978209037556511, + "loss": 0.4538, + "step": 3177 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019781890132078802, + "loss": 0.5201, + "step": 3178 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978168979764438, + "loss": 0.5998, + "step": 3179 + }, + { + "epoch": 0.09, + "learning_rate": 0.000197814893722637, + "loss": 0.5515, + "step": 3180 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019781288855938634, + "loss": 0.408, + "step": 3181 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019781088248671042, + "loss": 0.469, + "step": 3182 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978088755046279, + "loss": 0.5812, + "step": 3183 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019780686761315742, + "loss": 0.5577, + "step": 3184 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978048588123177, + "loss": 0.6236, + "step": 3185 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019780284910212736, + "loss": 0.4826, + "step": 3186 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001978008384826051, + "loss": 0.5806, + "step": 3187 + }, + { + "epoch": 0.09, + "learning_rate": 0.00019779882695376965, + "loss": 0.6964, + "step": 3188 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001977968145156397, + "loss": 0.5486, + "step": 3189 + }, + { + "epoch": 0.09, + "learning_rate": 0.0001977948011682339, + "loss": 0.5277, + "step": 3190 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019779278691157103, + "loss": 0.5175, + "step": 3191 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019779077174566983, + "loss": 0.5638, + "step": 3192 + }, + { + "epoch": 0.1, + "learning_rate": 0.000197788755670549, + "loss": 0.5195, + "step": 3193 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001977867386862273, + "loss": 0.6144, + "step": 3194 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019778472079272344, + "loss": 0.5271, + "step": 3195 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019778270199005622, + "loss": 0.5406, + "step": 3196 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019778068227824445, + "loss": 0.6174, + "step": 3197 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019777866165730684, + "loss": 0.5389, + "step": 3198 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019777664012726222, + "loss": 0.4724, + "step": 3199 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019777461768812936, + "loss": 0.6405, + "step": 3200 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019777259433992708, + "loss": 0.668, + "step": 3201 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019777057008267418, + "loss": 0.6457, + "step": 3202 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019776854491638948, + "loss": 0.5244, + "step": 3203 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019776651884109185, + "loss": 0.5951, + "step": 3204 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019776449185680007, + "loss": 0.5459, + "step": 3205 + }, + { + "epoch": 0.1, + "learning_rate": 0.000197762463963533, + "loss": 0.5071, + "step": 3206 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001977604351613095, + "loss": 0.5747, + "step": 3207 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019775840545014844, + "loss": 0.6199, + "step": 3208 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019775637483006873, + "loss": 0.5317, + "step": 3209 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019775434330108918, + "loss": 0.6238, + "step": 3210 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001977523108632287, + "loss": 0.5692, + "step": 3211 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019775027751650618, + "loss": 0.5241, + "step": 3212 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019774824326094058, + "loss": 0.5471, + "step": 3213 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019774620809655073, + "loss": 0.5529, + "step": 3214 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019774417202335563, + "loss": 0.6043, + "step": 3215 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019774213504137418, + "loss": 0.5673, + "step": 3216 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019774009715062526, + "loss": 0.5172, + "step": 3217 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019773805835112792, + "loss": 0.5166, + "step": 3218 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019773601864290106, + "loss": 0.7016, + "step": 3219 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019773397802596367, + "loss": 0.5847, + "step": 3220 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019773193650033467, + "loss": 0.5566, + "step": 3221 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019772989406603307, + "loss": 0.6281, + "step": 3222 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019772785072307792, + "loss": 0.5093, + "step": 3223 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019772580647148814, + "loss": 0.5992, + "step": 3224 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019772376131128275, + "loss": 0.6374, + "step": 3225 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019772171524248076, + "loss": 0.4823, + "step": 3226 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019771966826510125, + "loss": 0.5371, + "step": 3227 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019771762037916317, + "loss": 0.5285, + "step": 3228 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019771557158468564, + "loss": 0.5147, + "step": 3229 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019771352188168763, + "loss": 0.5695, + "step": 3230 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019771147127018828, + "loss": 0.7543, + "step": 3231 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001977094197502066, + "loss": 0.673, + "step": 3232 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019770736732176167, + "loss": 0.5831, + "step": 3233 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019770531398487259, + "loss": 0.634, + "step": 3234 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019770325973955842, + "loss": 0.4967, + "step": 3235 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001977012045858383, + "loss": 0.522, + "step": 3236 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019769914852373134, + "loss": 0.5622, + "step": 3237 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001976970915532566, + "loss": 0.5777, + "step": 3238 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019769503367443326, + "loss": 0.575, + "step": 3239 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019769297488728043, + "loss": 0.4671, + "step": 3240 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019769091519181723, + "loss": 0.559, + "step": 3241 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019768885458806287, + "loss": 0.6284, + "step": 3242 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001976867930760365, + "loss": 0.612, + "step": 3243 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019768473065575722, + "loss": 0.5825, + "step": 3244 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019768266732724424, + "loss": 0.5383, + "step": 3245 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019768060309051678, + "loss": 0.579, + "step": 3246 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019767853794559395, + "loss": 0.5953, + "step": 3247 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001976764718924951, + "loss": 0.6709, + "step": 3248 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001976744049312393, + "loss": 0.7081, + "step": 3249 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019767233706184577, + "loss": 0.6429, + "step": 3250 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019767026828433382, + "loss": 0.6717, + "step": 3251 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019766819859872264, + "loss": 0.5874, + "step": 3252 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019766612800503145, + "loss": 0.603, + "step": 3253 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019766405650327955, + "loss": 0.4992, + "step": 3254 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019766198409348616, + "loss": 0.5189, + "step": 3255 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001976599107756706, + "loss": 0.6052, + "step": 3256 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001976578365498521, + "loss": 0.6756, + "step": 3257 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019765576141604995, + "loss": 0.5863, + "step": 3258 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019765368537428345, + "loss": 0.6025, + "step": 3259 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019765160842457192, + "loss": 0.6064, + "step": 3260 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019764953056693467, + "loss": 0.5984, + "step": 3261 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019764745180139096, + "loss": 0.7312, + "step": 3262 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019764537212796018, + "loss": 0.7511, + "step": 3263 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019764329154666165, + "loss": 0.4513, + "step": 3264 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019764121005751473, + "loss": 0.6267, + "step": 3265 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001976391276605387, + "loss": 0.4866, + "step": 3266 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019763704435575303, + "loss": 0.5399, + "step": 3267 + }, + { + "epoch": 0.1, + "learning_rate": 0.000197634960143177, + "loss": 0.4943, + "step": 3268 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019763287502283007, + "loss": 0.5775, + "step": 3269 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019763078899473152, + "loss": 0.7, + "step": 3270 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019762870205890085, + "loss": 0.5755, + "step": 3271 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019762661421535743, + "loss": 0.5061, + "step": 3272 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019762452546412065, + "loss": 0.526, + "step": 3273 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019762243580520993, + "loss": 0.6233, + "step": 3274 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001976203452386447, + "loss": 0.5592, + "step": 3275 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001976182537644444, + "loss": 0.548, + "step": 3276 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019761616138262852, + "loss": 0.5399, + "step": 3277 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019761406809321646, + "loss": 0.5834, + "step": 3278 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001976119738962277, + "loss": 0.5917, + "step": 3279 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019760987879168172, + "loss": 0.5747, + "step": 3280 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019760778277959796, + "loss": 0.5451, + "step": 3281 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019760568585999594, + "loss": 0.5011, + "step": 3282 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019760358803289522, + "loss": 0.4899, + "step": 3283 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019760148929831516, + "loss": 0.6701, + "step": 3284 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001975993896562754, + "loss": 0.5483, + "step": 3285 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019759728910679537, + "loss": 0.6166, + "step": 3286 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019759518764989468, + "loss": 0.5092, + "step": 3287 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019759308528559283, + "loss": 0.4707, + "step": 3288 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019759098201390938, + "loss": 0.5521, + "step": 3289 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019758887783486383, + "loss": 0.5341, + "step": 3290 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019758677274847583, + "loss": 0.6095, + "step": 3291 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001975846667547649, + "loss": 0.5644, + "step": 3292 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019758255985375065, + "loss": 0.5758, + "step": 3293 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019758045204545263, + "loss": 0.5095, + "step": 3294 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019757834332989046, + "loss": 0.4966, + "step": 3295 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019757623370708375, + "loss": 0.4683, + "step": 3296 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019757412317705214, + "loss": 0.5493, + "step": 3297 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019757201173981517, + "loss": 0.4323, + "step": 3298 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019756989939539253, + "loss": 0.56, + "step": 3299 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019756778614380386, + "loss": 0.674, + "step": 3300 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019756567198506882, + "loss": 0.5726, + "step": 3301 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019756355691920703, + "loss": 0.6171, + "step": 3302 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019756144094623817, + "loss": 0.5468, + "step": 3303 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019755932406618194, + "loss": 0.607, + "step": 3304 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019755720627905797, + "loss": 0.4563, + "step": 3305 + }, + { + "epoch": 0.1, + "learning_rate": 0.000197555087584886, + "loss": 0.522, + "step": 3306 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001975529679836857, + "loss": 0.5937, + "step": 3307 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019755084747547676, + "loss": 0.6061, + "step": 3308 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019754872606027896, + "loss": 0.6029, + "step": 3309 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019754660373811196, + "loss": 0.5428, + "step": 3310 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001975444805089955, + "loss": 0.6271, + "step": 3311 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019754235637294937, + "loss": 0.5804, + "step": 3312 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019754023132999328, + "loss": 0.481, + "step": 3313 + }, + { + "epoch": 0.1, + "learning_rate": 0.000197538105380147, + "loss": 0.5757, + "step": 3314 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019753597852343026, + "loss": 0.639, + "step": 3315 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001975338507598629, + "loss": 0.5445, + "step": 3316 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019753172208946467, + "loss": 0.6215, + "step": 3317 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019752959251225532, + "loss": 0.6263, + "step": 3318 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019752746202825473, + "loss": 0.6431, + "step": 3319 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019752533063748266, + "loss": 0.4841, + "step": 3320 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019752319833995894, + "loss": 0.5999, + "step": 3321 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001975210651357034, + "loss": 0.3841, + "step": 3322 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001975189310247358, + "loss": 0.592, + "step": 3323 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001975167960070761, + "loss": 0.6357, + "step": 3324 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001975146600827441, + "loss": 0.6049, + "step": 3325 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019751252325175966, + "loss": 0.5402, + "step": 3326 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019751038551414263, + "loss": 0.6393, + "step": 3327 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001975082468699129, + "loss": 0.5779, + "step": 3328 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019750610731909034, + "loss": 0.5735, + "step": 3329 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019750396686169487, + "loss": 0.55, + "step": 3330 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019750182549774638, + "loss": 0.5695, + "step": 3331 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019749968322726476, + "loss": 0.573, + "step": 3332 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019749754005026998, + "loss": 0.5921, + "step": 3333 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001974953959667819, + "loss": 0.4419, + "step": 3334 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001974932509768205, + "loss": 0.5632, + "step": 3335 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019749110508040572, + "loss": 0.5911, + "step": 3336 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019748895827755746, + "loss": 0.6843, + "step": 3337 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019748681056829578, + "loss": 0.5464, + "step": 3338 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019748466195264054, + "loss": 0.5744, + "step": 3339 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019748251243061177, + "loss": 0.649, + "step": 3340 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019748036200222948, + "loss": 0.5589, + "step": 3341 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001974782106675136, + "loss": 0.5595, + "step": 3342 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019747605842648422, + "loss": 0.5934, + "step": 3343 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019747390527916124, + "loss": 0.7093, + "step": 3344 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019747175122556478, + "loss": 0.424, + "step": 3345 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001974695962657148, + "loss": 0.5377, + "step": 3346 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001974674403996314, + "loss": 0.5152, + "step": 3347 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019746528362733456, + "loss": 0.5647, + "step": 3348 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019746312594884436, + "loss": 0.611, + "step": 3349 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019746096736418086, + "loss": 0.6254, + "step": 3350 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019745880787336412, + "loss": 0.57, + "step": 3351 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019745664747641424, + "loss": 0.5088, + "step": 3352 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001974544861733513, + "loss": 0.5668, + "step": 3353 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019745232396419535, + "loss": 0.713, + "step": 3354 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019745016084896658, + "loss": 0.6042, + "step": 3355 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019744799682768505, + "loss": 0.5667, + "step": 3356 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019744583190037086, + "loss": 0.6749, + "step": 3357 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019744366606704416, + "loss": 0.591, + "step": 3358 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001974414993277251, + "loss": 0.7285, + "step": 3359 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019743933168243382, + "loss": 0.5321, + "step": 3360 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019743716313119048, + "loss": 0.5401, + "step": 3361 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019743499367401522, + "loss": 0.6953, + "step": 3362 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001974328233109282, + "loss": 0.5652, + "step": 3363 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019743065204194966, + "loss": 0.62, + "step": 3364 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019742847986709973, + "loss": 0.5569, + "step": 3365 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019742630678639862, + "loss": 0.6305, + "step": 3366 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019742413279986656, + "loss": 0.5498, + "step": 3367 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019742195790752373, + "loss": 0.5786, + "step": 3368 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019741978210939038, + "loss": 0.6695, + "step": 3369 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019741760540548667, + "loss": 0.575, + "step": 3370 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019741542779583298, + "loss": 0.6114, + "step": 3371 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001974132492804494, + "loss": 0.5662, + "step": 3372 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019741106985935627, + "loss": 0.6305, + "step": 3373 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019740888953257384, + "loss": 0.6888, + "step": 3374 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019740670830012237, + "loss": 0.5546, + "step": 3375 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019740452616202216, + "loss": 0.5786, + "step": 3376 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001974023431182935, + "loss": 0.6498, + "step": 3377 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019740015916895663, + "loss": 0.4735, + "step": 3378 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019739797431403194, + "loss": 0.5484, + "step": 3379 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019739578855353972, + "loss": 0.5575, + "step": 3380 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019739360188750023, + "loss": 0.6008, + "step": 3381 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019739141431593385, + "loss": 0.6264, + "step": 3382 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019738922583886095, + "loss": 0.6327, + "step": 3383 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001973870364563018, + "loss": 0.5717, + "step": 3384 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019738484616827685, + "loss": 0.5431, + "step": 3385 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001973826549748064, + "loss": 0.524, + "step": 3386 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001973804628759108, + "loss": 0.6221, + "step": 3387 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001973782698716105, + "loss": 0.6296, + "step": 3388 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019737607596192585, + "loss": 0.5983, + "step": 3389 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019737388114687726, + "loss": 0.502, + "step": 3390 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019737168542648513, + "loss": 0.5352, + "step": 3391 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019736948880076985, + "loss": 0.6775, + "step": 3392 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001973672912697519, + "loss": 0.5694, + "step": 3393 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001973650928334517, + "loss": 0.6744, + "step": 3394 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019736289349188966, + "loss": 0.5367, + "step": 3395 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019736069324508622, + "loss": 0.5609, + "step": 3396 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019735849209306187, + "loss": 0.6323, + "step": 3397 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019735629003583708, + "loss": 0.5716, + "step": 3398 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001973540870734323, + "loss": 0.5923, + "step": 3399 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019735188320586802, + "loss": 0.6449, + "step": 3400 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019734967843316472, + "loss": 0.6077, + "step": 3401 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001973474727553429, + "loss": 0.6522, + "step": 3402 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001973452661724231, + "loss": 0.4808, + "step": 3403 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001973430586844258, + "loss": 0.5169, + "step": 3404 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019734085029137155, + "loss": 0.6327, + "step": 3405 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001973386409932809, + "loss": 0.5576, + "step": 3406 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001973364307901743, + "loss": 0.6394, + "step": 3407 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019733421968207238, + "loss": 0.5481, + "step": 3408 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001973320076689957, + "loss": 0.5822, + "step": 3409 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019732979475096477, + "loss": 0.5198, + "step": 3410 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019732758092800022, + "loss": 0.498, + "step": 3411 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019732536620012262, + "loss": 0.642, + "step": 3412 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019732315056735254, + "loss": 0.5878, + "step": 3413 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019732093402971058, + "loss": 0.6933, + "step": 3414 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019731871658721742, + "loss": 0.5718, + "step": 3415 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019731649823989358, + "loss": 0.6102, + "step": 3416 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019731427898775974, + "loss": 0.5191, + "step": 3417 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001973120588308365, + "loss": 0.6158, + "step": 3418 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019730983776914453, + "loss": 0.6075, + "step": 3419 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019730761580270448, + "loss": 0.552, + "step": 3420 + }, + { + "epoch": 0.1, + "learning_rate": 0.000197305392931537, + "loss": 0.5632, + "step": 3421 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019730316915566275, + "loss": 0.6424, + "step": 3422 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019730094447510243, + "loss": 0.7095, + "step": 3423 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001972987188898767, + "loss": 0.59, + "step": 3424 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019729649240000626, + "loss": 0.5729, + "step": 3425 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001972942650055118, + "loss": 0.528, + "step": 3426 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019729203670641407, + "loss": 0.6031, + "step": 3427 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019728980750273375, + "loss": 0.6151, + "step": 3428 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019728757739449155, + "loss": 0.5633, + "step": 3429 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001972853463817083, + "loss": 0.6274, + "step": 3430 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001972831144644046, + "loss": 0.563, + "step": 3431 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019728088164260127, + "loss": 0.5479, + "step": 3432 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019727864791631912, + "loss": 0.5584, + "step": 3433 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019727641328557886, + "loss": 0.5959, + "step": 3434 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019727417775040128, + "loss": 0.6331, + "step": 3435 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019727194131080716, + "loss": 0.5326, + "step": 3436 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001972697039668173, + "loss": 0.5961, + "step": 3437 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001972674657184525, + "loss": 0.6409, + "step": 3438 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019726522656573358, + "loss": 0.5366, + "step": 3439 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019726298650868133, + "loss": 0.6699, + "step": 3440 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001972607455473166, + "loss": 0.5758, + "step": 3441 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019725850368166022, + "loss": 0.5604, + "step": 3442 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001972562609117331, + "loss": 0.6646, + "step": 3443 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019725401723755595, + "loss": 0.6183, + "step": 3444 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001972517726591497, + "loss": 0.6157, + "step": 3445 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001972495271765353, + "loss": 0.5289, + "step": 3446 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001972472807897335, + "loss": 0.5621, + "step": 3447 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019724503349876526, + "loss": 0.7092, + "step": 3448 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019724278530365146, + "loss": 0.5858, + "step": 3449 + }, + { + "epoch": 0.1, + "learning_rate": 0.000197240536204413, + "loss": 0.5641, + "step": 3450 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001972382862010708, + "loss": 0.635, + "step": 3451 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019723603529364577, + "loss": 0.583, + "step": 3452 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001972337834821588, + "loss": 0.5002, + "step": 3453 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001972315307666309, + "loss": 0.5294, + "step": 3454 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019722927714708297, + "loss": 0.5536, + "step": 3455 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019722702262353598, + "loss": 0.5428, + "step": 3456 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019722476719601087, + "loss": 0.6664, + "step": 3457 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019722251086452862, + "loss": 0.58, + "step": 3458 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019722025362911024, + "loss": 0.578, + "step": 3459 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019721799548977668, + "loss": 0.5968, + "step": 3460 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019721573644654891, + "loss": 0.554, + "step": 3461 + }, + { + "epoch": 0.1, + "learning_rate": 0.000197213476499448, + "loss": 0.6758, + "step": 3462 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019721121564849497, + "loss": 0.5415, + "step": 3463 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019720895389371075, + "loss": 0.5194, + "step": 3464 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019720669123511644, + "loss": 0.4819, + "step": 3465 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019720442767273303, + "loss": 0.6003, + "step": 3466 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001972021632065816, + "loss": 0.6102, + "step": 3467 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019719989783668324, + "loss": 0.4844, + "step": 3468 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019719763156305893, + "loss": 0.7534, + "step": 3469 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019719536438572979, + "loss": 0.657, + "step": 3470 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001971930963047169, + "loss": 0.6254, + "step": 3471 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001971908273200413, + "loss": 0.5325, + "step": 3472 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019718855743172418, + "loss": 0.6441, + "step": 3473 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019718628663978655, + "loss": 0.6122, + "step": 3474 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019718401494424957, + "loss": 0.6789, + "step": 3475 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019718174234513436, + "loss": 0.5762, + "step": 3476 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019717946884246205, + "loss": 0.6052, + "step": 3477 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019717719443625376, + "loss": 0.5893, + "step": 3478 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019717491912653063, + "loss": 0.4999, + "step": 3479 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019717264291331388, + "loss": 0.5824, + "step": 3480 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019717036579662458, + "loss": 0.5614, + "step": 3481 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019716808777648398, + "loss": 0.6466, + "step": 3482 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019716580885291318, + "loss": 0.5558, + "step": 3483 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019716352902593345, + "loss": 0.488, + "step": 3484 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019716124829556594, + "loss": 0.5111, + "step": 3485 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019715896666183186, + "loss": 0.6214, + "step": 3486 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001971566841247525, + "loss": 0.604, + "step": 3487 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001971544006843489, + "loss": 0.6678, + "step": 3488 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001971521163406425, + "loss": 0.6153, + "step": 3489 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001971498310936544, + "loss": 0.5373, + "step": 3490 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019714754494340589, + "loss": 0.576, + "step": 3491 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019714525788991824, + "loss": 0.6152, + "step": 3492 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019714296993321272, + "loss": 0.6492, + "step": 3493 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019714068107331054, + "loss": 0.6308, + "step": 3494 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019713839131023305, + "loss": 0.5181, + "step": 3495 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001971361006440015, + "loss": 0.581, + "step": 3496 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001971338090746372, + "loss": 0.6581, + "step": 3497 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019713151660216146, + "loss": 0.6263, + "step": 3498 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001971292232265956, + "loss": 0.5729, + "step": 3499 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019712692894796092, + "loss": 0.5976, + "step": 3500 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001971246337662788, + "loss": 0.5052, + "step": 3501 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019712233768157052, + "loss": 0.5091, + "step": 3502 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019712004069385745, + "loss": 0.7034, + "step": 3503 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019711774280316093, + "loss": 0.4848, + "step": 3504 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019711544400950238, + "loss": 0.5281, + "step": 3505 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001971131443129031, + "loss": 0.6421, + "step": 3506 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019711084371338455, + "loss": 0.6263, + "step": 3507 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019710854221096806, + "loss": 0.583, + "step": 3508 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019710623980567505, + "loss": 0.479, + "step": 3509 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019710393649752695, + "loss": 0.752, + "step": 3510 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019710163228654512, + "loss": 0.5424, + "step": 3511 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019709932717275103, + "loss": 0.6423, + "step": 3512 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001970970211561661, + "loss": 0.5305, + "step": 3513 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019709471423681174, + "loss": 0.5159, + "step": 3514 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019709240641470945, + "loss": 0.5922, + "step": 3515 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019709009768988062, + "loss": 0.4923, + "step": 3516 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019708778806234683, + "loss": 0.4821, + "step": 3517 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019708547753212943, + "loss": 0.5672, + "step": 3518 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019708316609924994, + "loss": 0.6629, + "step": 3519 + }, + { + "epoch": 0.1, + "learning_rate": 0.0001970808537637299, + "loss": 0.5374, + "step": 3520 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019707854052559078, + "loss": 0.5534, + "step": 3521 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019707622638485405, + "loss": 0.5518, + "step": 3522 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019707391134154127, + "loss": 0.5746, + "step": 3523 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019707159539567398, + "loss": 0.4791, + "step": 3524 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019706927854727367, + "loss": 0.5923, + "step": 3525 + }, + { + "epoch": 0.1, + "learning_rate": 0.00019706696079636188, + "loss": 0.5736, + "step": 3526 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001970646421429602, + "loss": 0.5255, + "step": 3527 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019706232258709014, + "loss": 0.6936, + "step": 3528 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001970600021287733, + "loss": 0.5728, + "step": 3529 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019705768076803124, + "loss": 0.6664, + "step": 3530 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019705535850488554, + "loss": 0.5756, + "step": 3531 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001970530353393578, + "loss": 0.7055, + "step": 3532 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019705071127146964, + "loss": 0.754, + "step": 3533 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019704838630124264, + "loss": 0.5816, + "step": 3534 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019704606042869843, + "loss": 0.5277, + "step": 3535 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001970437336538586, + "loss": 0.6347, + "step": 3536 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019704140597674483, + "loss": 0.5934, + "step": 3537 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019703907739737874, + "loss": 0.5333, + "step": 3538 + }, + { + "epoch": 0.11, + "learning_rate": 0.000197036747915782, + "loss": 0.6027, + "step": 3539 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019703441753197624, + "loss": 0.5414, + "step": 3540 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019703208624598313, + "loss": 0.7016, + "step": 3541 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019702975405782436, + "loss": 0.6544, + "step": 3542 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019702742096752162, + "loss": 0.6639, + "step": 3543 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019702508697509662, + "loss": 0.5847, + "step": 3544 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019702275208057098, + "loss": 0.6008, + "step": 3545 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019702041628396648, + "loss": 0.6848, + "step": 3546 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001970180795853048, + "loss": 0.5293, + "step": 3547 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019701574198460772, + "loss": 0.5493, + "step": 3548 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001970134034818969, + "loss": 0.6394, + "step": 3549 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019701106407719413, + "loss": 0.5803, + "step": 3550 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019700872377052117, + "loss": 0.5548, + "step": 3551 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019700638256189975, + "loss": 0.4906, + "step": 3552 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019700404045135166, + "loss": 0.5948, + "step": 3553 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019700169743889863, + "loss": 0.5324, + "step": 3554 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019699935352456252, + "loss": 0.6038, + "step": 3555 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019699700870836505, + "loss": 0.5016, + "step": 3556 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019699466299032806, + "loss": 0.6388, + "step": 3557 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019699231637047336, + "loss": 0.5143, + "step": 3558 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019698996884882274, + "loss": 0.6607, + "step": 3559 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019698762042539807, + "loss": 0.6174, + "step": 3560 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019698527110022116, + "loss": 0.5633, + "step": 3561 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019698292087331383, + "loss": 0.5812, + "step": 3562 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019698056974469798, + "loss": 0.4958, + "step": 3563 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019697821771439542, + "loss": 0.6323, + "step": 3564 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019697586478242807, + "loss": 0.5683, + "step": 3565 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001969735109488178, + "loss": 0.5936, + "step": 3566 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019697115621358641, + "loss": 0.6031, + "step": 3567 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001969688005767559, + "loss": 0.4507, + "step": 3568 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001969664440383481, + "loss": 0.5612, + "step": 3569 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019696408659838498, + "loss": 0.5407, + "step": 3570 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019696172825688844, + "loss": 0.6322, + "step": 3571 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019695936901388037, + "loss": 0.481, + "step": 3572 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019695700886938276, + "loss": 0.609, + "step": 3573 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001969546478234175, + "loss": 0.4871, + "step": 3574 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019695228587600656, + "loss": 0.5806, + "step": 3575 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001969499230271719, + "loss": 0.5981, + "step": 3576 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019694755927693552, + "loss": 0.4551, + "step": 3577 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019694519462531938, + "loss": 0.6354, + "step": 3578 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019694282907234544, + "loss": 0.5879, + "step": 3579 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019694046261803573, + "loss": 0.7299, + "step": 3580 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001969380952624122, + "loss": 0.5638, + "step": 3581 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019693572700549692, + "loss": 0.5784, + "step": 3582 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019693335784731192, + "loss": 0.6097, + "step": 3583 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019693098778787913, + "loss": 0.5617, + "step": 3584 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019692861682722067, + "loss": 0.578, + "step": 3585 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019692624496535856, + "loss": 0.5736, + "step": 3586 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019692387220231487, + "loss": 0.4892, + "step": 3587 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019692149853811164, + "loss": 0.6238, + "step": 3588 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019691912397277098, + "loss": 0.5466, + "step": 3589 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019691674850631487, + "loss": 0.5117, + "step": 3590 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001969143721387655, + "loss": 0.4978, + "step": 3591 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019691199487014492, + "loss": 0.6686, + "step": 3592 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019690961670047524, + "loss": 0.5603, + "step": 3593 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019690723762977853, + "loss": 0.5218, + "step": 3594 + }, + { + "epoch": 0.11, + "learning_rate": 0.000196904857658077, + "loss": 0.5975, + "step": 3595 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019690247678539268, + "loss": 0.5447, + "step": 3596 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019690009501174778, + "loss": 0.5667, + "step": 3597 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001968977123371644, + "loss": 0.5212, + "step": 3598 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019689532876166475, + "loss": 0.5351, + "step": 3599 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001968929442852709, + "loss": 0.6121, + "step": 3600 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001968905589080051, + "loss": 0.5877, + "step": 3601 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001968881726298895, + "loss": 0.6114, + "step": 3602 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019688578545094628, + "loss": 0.6358, + "step": 3603 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019688339737119765, + "loss": 0.6442, + "step": 3604 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001968810083906658, + "loss": 0.5443, + "step": 3605 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019687861850937294, + "loss": 0.5493, + "step": 3606 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019687622772734131, + "loss": 0.595, + "step": 3607 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019687383604459312, + "loss": 0.5386, + "step": 3608 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001968714434611506, + "loss": 0.5804, + "step": 3609 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019686904997703603, + "loss": 0.5181, + "step": 3610 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019686665559227163, + "loss": 0.5028, + "step": 3611 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019686426030687967, + "loss": 0.5369, + "step": 3612 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019686186412088243, + "loss": 0.5986, + "step": 3613 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019685946703430218, + "loss": 0.6374, + "step": 3614 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001968570690471612, + "loss": 0.6483, + "step": 3615 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019685467015948183, + "loss": 0.6117, + "step": 3616 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001968522703712863, + "loss": 0.6023, + "step": 3617 + }, + { + "epoch": 0.11, + "learning_rate": 0.000196849869682597, + "loss": 0.5534, + "step": 3618 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019684746809343616, + "loss": 0.5384, + "step": 3619 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001968450656038262, + "loss": 0.5919, + "step": 3620 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019684266221378942, + "loss": 0.453, + "step": 3621 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019684025792334814, + "loss": 0.6182, + "step": 3622 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019683785273252476, + "loss": 0.5319, + "step": 3623 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001968354466413416, + "loss": 0.5432, + "step": 3624 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019683303964982107, + "loss": 0.6083, + "step": 3625 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019683063175798552, + "loss": 0.5752, + "step": 3626 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019682822296585736, + "loss": 0.6003, + "step": 3627 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019682581327345897, + "loss": 0.6229, + "step": 3628 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019682340268081276, + "loss": 0.5676, + "step": 3629 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019682099118794116, + "loss": 0.555, + "step": 3630 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019681857879486655, + "loss": 0.562, + "step": 3631 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019681616550161138, + "loss": 0.5729, + "step": 3632 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001968137513081981, + "loss": 0.5458, + "step": 3633 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019681133621464914, + "loss": 0.6172, + "step": 3634 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019680892022098698, + "loss": 0.5533, + "step": 3635 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019680650332723403, + "loss": 0.5833, + "step": 3636 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019680408553341284, + "loss": 0.6381, + "step": 3637 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001968016668395458, + "loss": 0.6318, + "step": 3638 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019679924724565545, + "loss": 0.5413, + "step": 3639 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001967968267517643, + "loss": 0.537, + "step": 3640 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019679440535789482, + "loss": 0.648, + "step": 3641 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019679198306406955, + "loss": 0.5462, + "step": 3642 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019678955987031098, + "loss": 0.489, + "step": 3643 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019678713577664166, + "loss": 0.5382, + "step": 3644 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019678471078308417, + "loss": 0.5448, + "step": 3645 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019678228488966096, + "loss": 0.5738, + "step": 3646 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019677985809639467, + "loss": 0.6512, + "step": 3647 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019677743040330783, + "loss": 0.6787, + "step": 3648 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019677500181042299, + "loss": 0.4915, + "step": 3649 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001967725723177628, + "loss": 0.4961, + "step": 3650 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019677014192534976, + "loss": 0.6628, + "step": 3651 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001967677106332065, + "loss": 0.5401, + "step": 3652 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019676527844135567, + "loss": 0.6828, + "step": 3653 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019676284534981983, + "loss": 0.6416, + "step": 3654 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019676041135862162, + "loss": 0.452, + "step": 3655 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019675797646778368, + "loss": 0.5756, + "step": 3656 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019675554067732863, + "loss": 0.5855, + "step": 3657 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019675310398727913, + "loss": 0.5567, + "step": 3658 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019675066639765785, + "loss": 0.6508, + "step": 3659 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001967482279084874, + "loss": 0.6144, + "step": 3660 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001967457885197905, + "loss": 0.6499, + "step": 3661 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019674334823158981, + "loss": 0.5269, + "step": 3662 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019674090704390804, + "loss": 0.4969, + "step": 3663 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019673846495676787, + "loss": 0.7023, + "step": 3664 + }, + { + "epoch": 0.11, + "learning_rate": 0.000196736021970192, + "loss": 0.6704, + "step": 3665 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019673357808420316, + "loss": 0.5891, + "step": 3666 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019673113329882406, + "loss": 0.6867, + "step": 3667 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019672868761407746, + "loss": 0.6834, + "step": 3668 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019672624102998605, + "loss": 0.7643, + "step": 3669 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001967237935465726, + "loss": 0.5849, + "step": 3670 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019672134516385986, + "loss": 0.5455, + "step": 3671 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019671889588187064, + "loss": 0.6204, + "step": 3672 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019671644570062763, + "loss": 0.5353, + "step": 3673 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001967139946201537, + "loss": 0.4616, + "step": 3674 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019671154264047159, + "loss": 0.558, + "step": 3675 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019670908976160404, + "loss": 0.5725, + "step": 3676 + }, + { + "epoch": 0.11, + "learning_rate": 0.000196706635983574, + "loss": 0.5167, + "step": 3677 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019670418130640415, + "loss": 0.5894, + "step": 3678 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019670172573011739, + "loss": 0.5051, + "step": 3679 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019669926925473653, + "loss": 0.5492, + "step": 3680 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019669681188028436, + "loss": 0.6087, + "step": 3681 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001966943536067838, + "loss": 0.5022, + "step": 3682 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019669189443425767, + "loss": 0.6539, + "step": 3683 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019668943436272886, + "loss": 0.603, + "step": 3684 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019668697339222021, + "loss": 0.6822, + "step": 3685 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019668451152275466, + "loss": 0.5254, + "step": 3686 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019668204875435502, + "loss": 0.6193, + "step": 3687 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001966795850870442, + "loss": 0.634, + "step": 3688 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019667712052084518, + "loss": 0.5428, + "step": 3689 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001966746550557808, + "loss": 0.4925, + "step": 3690 + }, + { + "epoch": 0.11, + "learning_rate": 0.000196672188691874, + "loss": 0.5837, + "step": 3691 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019666972142914774, + "loss": 0.6939, + "step": 3692 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019666725326762495, + "loss": 0.6163, + "step": 3693 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001966647842073285, + "loss": 0.5804, + "step": 3694 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019666231424828149, + "loss": 0.6185, + "step": 3695 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019665984339050678, + "loss": 0.5711, + "step": 3696 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019665737163402736, + "loss": 0.6258, + "step": 3697 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019665489897886622, + "loss": 0.5535, + "step": 3698 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019665242542504636, + "loss": 0.5762, + "step": 3699 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001966499509725908, + "loss": 0.5632, + "step": 3700 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019664747562152246, + "loss": 0.5079, + "step": 3701 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019664499937186447, + "loss": 0.5343, + "step": 3702 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019664252222363976, + "loss": 0.539, + "step": 3703 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019664004417687138, + "loss": 0.5497, + "step": 3704 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019663756523158243, + "loss": 0.5574, + "step": 3705 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001966350853877959, + "loss": 0.5998, + "step": 3706 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019663260464553484, + "loss": 0.5959, + "step": 3707 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019663012300482233, + "loss": 0.6526, + "step": 3708 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001966276404656815, + "loss": 0.418, + "step": 3709 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019662515702813534, + "loss": 0.4956, + "step": 3710 + }, + { + "epoch": 0.11, + "learning_rate": 0.000196622672692207, + "loss": 0.6129, + "step": 3711 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019662018745791952, + "loss": 0.5881, + "step": 3712 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001966177013252961, + "loss": 0.5457, + "step": 3713 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019661521429435976, + "loss": 0.531, + "step": 3714 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001966127263651337, + "loss": 0.5427, + "step": 3715 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019661023753764098, + "loss": 0.4944, + "step": 3716 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001966077478119048, + "loss": 0.649, + "step": 3717 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001966052571879483, + "loss": 0.639, + "step": 3718 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001966027656657946, + "loss": 0.5442, + "step": 3719 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019660027324546687, + "loss": 0.524, + "step": 3720 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019659777992698832, + "loss": 0.5887, + "step": 3721 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001965952857103821, + "loss": 0.5214, + "step": 3722 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019659279059567144, + "loss": 0.4948, + "step": 3723 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001965902945828795, + "loss": 0.5767, + "step": 3724 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001965877976720295, + "loss": 0.5447, + "step": 3725 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019658529986314464, + "loss": 0.6041, + "step": 3726 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019658280115624814, + "loss": 0.5248, + "step": 3727 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001965803015513633, + "loss": 0.5536, + "step": 3728 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019657780104851327, + "loss": 0.5983, + "step": 3729 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019657529964772134, + "loss": 0.5551, + "step": 3730 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001965727973490108, + "loss": 0.5338, + "step": 3731 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019657029415240485, + "loss": 0.7143, + "step": 3732 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019656779005792677, + "loss": 0.5843, + "step": 3733 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001965652850655999, + "loss": 0.6785, + "step": 3734 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001965627791754475, + "loss": 0.5739, + "step": 3735 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019656027238749285, + "loss": 0.6091, + "step": 3736 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019655776470175928, + "loss": 0.6167, + "step": 3737 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001965552561182701, + "loss": 0.6288, + "step": 3738 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019655274663704864, + "loss": 0.7043, + "step": 3739 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001965502362581182, + "loss": 0.6261, + "step": 3740 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019654772498150215, + "loss": 0.5713, + "step": 3741 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019654521280722387, + "loss": 0.6261, + "step": 3742 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019654269973530663, + "loss": 0.5226, + "step": 3743 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019654018576577387, + "loss": 0.6776, + "step": 3744 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019653767089864894, + "loss": 0.5011, + "step": 3745 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019653515513395523, + "loss": 0.5644, + "step": 3746 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001965326384717161, + "loss": 0.6525, + "step": 3747 + }, + { + "epoch": 0.11, + "learning_rate": 0.000196530120911955, + "loss": 0.5579, + "step": 3748 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019652760245469529, + "loss": 0.556, + "step": 3749 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001965250830999604, + "loss": 0.515, + "step": 3750 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019652256284777376, + "loss": 0.6273, + "step": 3751 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019652004169815878, + "loss": 0.5751, + "step": 3752 + }, + { + "epoch": 0.11, + "learning_rate": 0.000196517519651139, + "loss": 0.5768, + "step": 3753 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001965149967067377, + "loss": 0.6494, + "step": 3754 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019651247286497846, + "loss": 0.4559, + "step": 3755 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019650994812588473, + "loss": 0.5304, + "step": 3756 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019650742248947996, + "loss": 0.5599, + "step": 3757 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019650489595578763, + "loss": 0.5928, + "step": 3758 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019650236852483124, + "loss": 0.5991, + "step": 3759 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001964998401966343, + "loss": 0.5275, + "step": 3760 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001964973109712203, + "loss": 0.5833, + "step": 3761 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019649478084861276, + "loss": 0.6219, + "step": 3762 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001964922498288352, + "loss": 0.5497, + "step": 3763 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019648971791191117, + "loss": 0.5789, + "step": 3764 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019648718509786425, + "loss": 0.6261, + "step": 3765 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019648465138671786, + "loss": 0.4697, + "step": 3766 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019648211677849567, + "loss": 0.6453, + "step": 3767 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019647958127322118, + "loss": 0.6204, + "step": 3768 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019647704487091804, + "loss": 0.5037, + "step": 3769 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019647450757160977, + "loss": 0.61, + "step": 3770 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019647196937532, + "loss": 0.6834, + "step": 3771 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001964694302820723, + "loss": 0.5438, + "step": 3772 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019646689029189024, + "loss": 0.5045, + "step": 3773 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019646434940479756, + "loss": 0.5942, + "step": 3774 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019646180762081773, + "loss": 0.573, + "step": 3775 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001964592649399745, + "loss": 0.6184, + "step": 3776 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019645672136229148, + "loss": 0.5514, + "step": 3777 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001964541768877923, + "loss": 0.573, + "step": 3778 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001964516315165006, + "loss": 0.5512, + "step": 3779 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001964490852484401, + "loss": 0.5057, + "step": 3780 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019644653808363445, + "loss": 0.5167, + "step": 3781 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001964439900221073, + "loss": 0.5746, + "step": 3782 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019644144106388238, + "loss": 0.5102, + "step": 3783 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001964388912089834, + "loss": 0.5696, + "step": 3784 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019643634045743403, + "loss": 0.6081, + "step": 3785 + }, + { + "epoch": 0.11, + "learning_rate": 0.000196433788809258, + "loss": 0.5914, + "step": 3786 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019643123626447904, + "loss": 0.553, + "step": 3787 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019642868282312086, + "loss": 0.596, + "step": 3788 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019642612848520726, + "loss": 0.5595, + "step": 3789 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001964235732507619, + "loss": 0.5472, + "step": 3790 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019642101711980862, + "loss": 0.5046, + "step": 3791 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019641846009237115, + "loss": 0.6419, + "step": 3792 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019641590216847328, + "loss": 0.6312, + "step": 3793 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019641334334813876, + "loss": 0.516, + "step": 3794 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019641078363139138, + "loss": 0.5078, + "step": 3795 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019640822301825498, + "loss": 0.6224, + "step": 3796 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019640566150875335, + "loss": 0.6573, + "step": 3797 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001964030991029103, + "loss": 0.4634, + "step": 3798 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019640053580074966, + "loss": 0.5142, + "step": 3799 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019639797160229527, + "loss": 0.5286, + "step": 3800 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019639540650757094, + "loss": 0.5124, + "step": 3801 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019639284051660053, + "loss": 0.4271, + "step": 3802 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019639027362940793, + "loss": 0.6196, + "step": 3803 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019638770584601699, + "loss": 0.5857, + "step": 3804 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019638513716645158, + "loss": 0.5815, + "step": 3805 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019638256759073555, + "loss": 0.4898, + "step": 3806 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019637999711889282, + "loss": 0.6762, + "step": 3807 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019637742575094732, + "loss": 0.5294, + "step": 3808 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001963748534869229, + "loss": 0.5173, + "step": 3809 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019637228032684352, + "loss": 0.5356, + "step": 3810 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001963697062707331, + "loss": 0.5029, + "step": 3811 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019636713131861554, + "loss": 0.6193, + "step": 3812 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001963645554705148, + "loss": 0.5015, + "step": 3813 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019636197872645482, + "loss": 0.4854, + "step": 3814 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001963594010864596, + "loss": 0.5713, + "step": 3815 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019635682255055305, + "loss": 0.6182, + "step": 3816 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019635424311875918, + "loss": 0.5244, + "step": 3817 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019635166279110194, + "loss": 0.5305, + "step": 3818 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019634908156760536, + "loss": 0.5581, + "step": 3819 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019634649944829345, + "loss": 0.6886, + "step": 3820 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019634391643319013, + "loss": 0.5275, + "step": 3821 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019634133252231952, + "loss": 0.6348, + "step": 3822 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019633874771570557, + "loss": 0.6143, + "step": 3823 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019633616201337236, + "loss": 0.5923, + "step": 3824 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001963335754153439, + "loss": 0.585, + "step": 3825 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019633098792164424, + "loss": 0.5488, + "step": 3826 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019632839953229745, + "loss": 0.5418, + "step": 3827 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001963258102473276, + "loss": 0.5424, + "step": 3828 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019632322006675877, + "loss": 0.6726, + "step": 3829 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019632062899061504, + "loss": 0.5606, + "step": 3830 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019631803701892047, + "loss": 0.4937, + "step": 3831 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001963154441516992, + "loss": 0.641, + "step": 3832 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001963128503889753, + "loss": 0.4888, + "step": 3833 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019631025573077287, + "loss": 0.6287, + "step": 3834 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001963076601771161, + "loss": 0.5382, + "step": 3835 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001963050637280291, + "loss": 0.5309, + "step": 3836 + }, + { + "epoch": 0.11, + "learning_rate": 0.000196302466383536, + "loss": 0.4735, + "step": 3837 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019629986814366092, + "loss": 0.467, + "step": 3838 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001962972690084281, + "loss": 0.6606, + "step": 3839 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019629466897786162, + "loss": 0.6214, + "step": 3840 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001962920680519857, + "loss": 0.5021, + "step": 3841 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019628946623082447, + "loss": 0.5992, + "step": 3842 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001962868635144022, + "loss": 0.6651, + "step": 3843 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019628425990274304, + "loss": 0.5525, + "step": 3844 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001962816553958712, + "loss": 0.6087, + "step": 3845 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019627904999381093, + "loss": 0.6086, + "step": 3846 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001962764436965864, + "loss": 0.6112, + "step": 3847 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019627383650422186, + "loss": 0.3889, + "step": 3848 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019627122841674158, + "loss": 0.5828, + "step": 3849 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019626861943416977, + "loss": 0.6229, + "step": 3850 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001962660095565307, + "loss": 0.5018, + "step": 3851 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019626339878384868, + "loss": 0.5796, + "step": 3852 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019626078711614792, + "loss": 0.5126, + "step": 3853 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001962581745534527, + "loss": 0.5236, + "step": 3854 + }, + { + "epoch": 0.11, + "learning_rate": 0.0001962555610957874, + "loss": 0.5397, + "step": 3855 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019625294674317624, + "loss": 0.6608, + "step": 3856 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019625033149564352, + "loss": 0.5438, + "step": 3857 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019624771535321358, + "loss": 0.5652, + "step": 3858 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019624509831591073, + "loss": 0.5306, + "step": 3859 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019624248038375934, + "loss": 0.6672, + "step": 3860 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019623986155678372, + "loss": 0.5833, + "step": 3861 + }, + { + "epoch": 0.11, + "learning_rate": 0.00019623724183500825, + "loss": 0.6416, + "step": 3862 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019623462121845723, + "loss": 0.6334, + "step": 3863 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019623199970715506, + "loss": 0.588, + "step": 3864 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019622937730112613, + "loss": 0.5239, + "step": 3865 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019622675400039476, + "loss": 0.5198, + "step": 3866 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019622412980498542, + "loss": 0.6146, + "step": 3867 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019622150471492247, + "loss": 0.5971, + "step": 3868 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001962188787302303, + "loss": 0.5362, + "step": 3869 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019621625185093334, + "loss": 0.5744, + "step": 3870 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019621362407705605, + "loss": 0.5704, + "step": 3871 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019621099540862282, + "loss": 0.564, + "step": 3872 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019620836584565812, + "loss": 0.5725, + "step": 3873 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019620573538818635, + "loss": 0.6275, + "step": 3874 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019620310403623199, + "loss": 0.5985, + "step": 3875 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001962004717898195, + "loss": 0.462, + "step": 3876 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001961978386489734, + "loss": 0.5868, + "step": 3877 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019619520461371814, + "loss": 0.591, + "step": 3878 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019619256968407819, + "loss": 0.5722, + "step": 3879 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019618993386007807, + "loss": 0.5529, + "step": 3880 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019618729714174227, + "loss": 0.6195, + "step": 3881 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019618465952909534, + "loss": 0.5501, + "step": 3882 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019618202102216175, + "loss": 0.6003, + "step": 3883 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001961793816209661, + "loss": 0.6602, + "step": 3884 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001961767413255329, + "loss": 0.5289, + "step": 3885 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019617410013588664, + "loss": 0.4464, + "step": 3886 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019617145805205194, + "loss": 0.7061, + "step": 3887 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001961688150740534, + "loss": 0.5159, + "step": 3888 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019616617120191552, + "loss": 0.6052, + "step": 3889 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001961635264356629, + "loss": 0.6965, + "step": 3890 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019616088077532013, + "loss": 0.6103, + "step": 3891 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019615823422091183, + "loss": 0.5116, + "step": 3892 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019615558677246256, + "loss": 0.5424, + "step": 3893 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019615293842999702, + "loss": 0.488, + "step": 3894 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019615028919353977, + "loss": 0.6191, + "step": 3895 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019614763906311545, + "loss": 0.6711, + "step": 3896 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001961449880387487, + "loss": 0.6951, + "step": 3897 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019614233612046415, + "loss": 0.5132, + "step": 3898 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001961396833082865, + "loss": 0.4632, + "step": 3899 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019613702960224037, + "loss": 0.6254, + "step": 3900 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001961343750023505, + "loss": 0.5178, + "step": 3901 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001961317195086415, + "loss": 0.5788, + "step": 3902 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019612906312113806, + "loss": 0.596, + "step": 3903 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019612640583986494, + "loss": 0.5433, + "step": 3904 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019612374766484682, + "loss": 0.4993, + "step": 3905 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019612108859610837, + "loss": 0.5094, + "step": 3906 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019611842863367435, + "loss": 0.5524, + "step": 3907 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001961157677775695, + "loss": 0.6105, + "step": 3908 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019611310602781856, + "loss": 0.5913, + "step": 3909 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019611044338444626, + "loss": 0.5104, + "step": 3910 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019610777984747734, + "loss": 0.6695, + "step": 3911 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019610511541693664, + "loss": 0.6134, + "step": 3912 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019610245009284884, + "loss": 0.6235, + "step": 3913 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019609978387523876, + "loss": 0.5336, + "step": 3914 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019609711676413122, + "loss": 0.6106, + "step": 3915 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019609444875955096, + "loss": 0.4624, + "step": 3916 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019609177986152284, + "loss": 0.5611, + "step": 3917 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019608911007007164, + "loss": 0.5395, + "step": 3918 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019608643938522218, + "loss": 0.5578, + "step": 3919 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001960837678069993, + "loss": 0.52, + "step": 3920 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019608109533542786, + "loss": 0.7495, + "step": 3921 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019607842197053268, + "loss": 0.5688, + "step": 3922 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019607574771233864, + "loss": 0.4618, + "step": 3923 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019607307256087057, + "loss": 0.5038, + "step": 3924 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019607039651615336, + "loss": 0.5115, + "step": 3925 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001960677195782119, + "loss": 0.4485, + "step": 3926 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001960650417470711, + "loss": 0.6096, + "step": 3927 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001960623630227558, + "loss": 0.4531, + "step": 3928 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019605968340529094, + "loss": 0.6053, + "step": 3929 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019605700289470143, + "loss": 0.5388, + "step": 3930 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001960543214910122, + "loss": 0.5933, + "step": 3931 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019605163919424816, + "loss": 0.5183, + "step": 3932 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019604895600443424, + "loss": 0.4967, + "step": 3933 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019604627192159547, + "loss": 0.6566, + "step": 3934 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001960435869457567, + "loss": 0.6396, + "step": 3935 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019604090107694293, + "loss": 0.5786, + "step": 3936 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019603821431517915, + "loss": 0.5977, + "step": 3937 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019603552666049034, + "loss": 0.5249, + "step": 3938 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019603283811290146, + "loss": 0.692, + "step": 3939 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019603014867243757, + "loss": 0.6142, + "step": 3940 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019602745833912358, + "loss": 0.5535, + "step": 3941 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001960247671129846, + "loss": 0.4973, + "step": 3942 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019602207499404557, + "loss": 0.6088, + "step": 3943 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019601938198233158, + "loss": 0.6918, + "step": 3944 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019601668807786763, + "loss": 0.6105, + "step": 3945 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001960139932806788, + "loss": 0.6138, + "step": 3946 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019601129759079012, + "loss": 0.5549, + "step": 3947 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019600860100822668, + "loss": 0.6237, + "step": 3948 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019600590353301348, + "loss": 0.6455, + "step": 3949 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001960032051651757, + "loss": 0.5315, + "step": 3950 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019600050590473837, + "loss": 0.5493, + "step": 3951 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019599780575172657, + "loss": 0.6079, + "step": 3952 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001959951047061655, + "loss": 0.6747, + "step": 3953 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019599240276808012, + "loss": 0.5938, + "step": 3954 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001959896999374957, + "loss": 0.6123, + "step": 3955 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001959869962144373, + "loss": 0.5703, + "step": 3956 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019598429159893005, + "loss": 0.5366, + "step": 3957 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019598158609099913, + "loss": 0.5132, + "step": 3958 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019597887969066965, + "loss": 0.5375, + "step": 3959 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019597617239796683, + "loss": 0.5414, + "step": 3960 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019597346421291578, + "loss": 0.4851, + "step": 3961 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019597075513554173, + "loss": 0.6048, + "step": 3962 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019596804516586986, + "loss": 0.4745, + "step": 3963 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019596533430392535, + "loss": 0.6425, + "step": 3964 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019596262254973338, + "loss": 0.513, + "step": 3965 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019595990990331922, + "loss": 0.5493, + "step": 3966 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019595719636470804, + "loss": 0.5375, + "step": 3967 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019595448193392514, + "loss": 0.6104, + "step": 3968 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019595176661099567, + "loss": 0.6222, + "step": 3969 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001959490503959449, + "loss": 0.49, + "step": 3970 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019594633328879815, + "loss": 0.497, + "step": 3971 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019594361528958062, + "loss": 0.4533, + "step": 3972 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019594089639831758, + "loss": 0.4845, + "step": 3973 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019593817661503433, + "loss": 0.5587, + "step": 3974 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019593545593975616, + "loss": 0.4808, + "step": 3975 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019593273437250834, + "loss": 0.4997, + "step": 3976 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001959300119133162, + "loss": 0.563, + "step": 3977 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019592728856220504, + "loss": 0.5569, + "step": 3978 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001959245643192002, + "loss": 0.5525, + "step": 3979 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019592183918432695, + "loss": 0.5506, + "step": 3980 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019591911315761072, + "loss": 0.5461, + "step": 3981 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019591638623907677, + "loss": 0.5667, + "step": 3982 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019591365842875052, + "loss": 0.5308, + "step": 3983 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019591092972665727, + "loss": 0.6816, + "step": 3984 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019590820013282243, + "loss": 0.4812, + "step": 3985 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001959054696472714, + "loss": 0.609, + "step": 3986 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001959027382700295, + "loss": 0.6572, + "step": 3987 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019590000600112217, + "loss": 0.6019, + "step": 3988 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001958972728405748, + "loss": 0.5887, + "step": 3989 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019589453878841284, + "loss": 0.5647, + "step": 3990 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019589180384466166, + "loss": 0.5509, + "step": 3991 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001958890680093467, + "loss": 0.6036, + "step": 3992 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001958863312824934, + "loss": 0.5652, + "step": 3993 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019588359366412724, + "loss": 0.5606, + "step": 3994 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001958808551542736, + "loss": 0.6594, + "step": 3995 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019587811575295802, + "loss": 0.5056, + "step": 3996 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001958753754602059, + "loss": 0.6127, + "step": 3997 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019587263427604278, + "loss": 0.5609, + "step": 3998 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019586989220049409, + "loss": 0.5436, + "step": 3999 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019586714923358538, + "loss": 0.5174, + "step": 4000 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001958644053753421, + "loss": 0.5315, + "step": 4001 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019586166062578978, + "loss": 0.5183, + "step": 4002 + }, + { + "epoch": 0.12, + "learning_rate": 0.000195858914984954, + "loss": 0.493, + "step": 4003 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001958561684528602, + "loss": 0.6487, + "step": 4004 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019585342102953392, + "loss": 0.4812, + "step": 4005 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019585067271500077, + "loss": 0.6016, + "step": 4006 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019584792350928627, + "loss": 0.577, + "step": 4007 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019584517341241596, + "loss": 0.5934, + "step": 4008 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019584242242441543, + "loss": 0.5408, + "step": 4009 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019583967054531028, + "loss": 0.5837, + "step": 4010 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019583691777512604, + "loss": 0.6078, + "step": 4011 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019583416411388834, + "loss": 0.5445, + "step": 4012 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001958314095616228, + "loss": 0.6053, + "step": 4013 + }, + { + "epoch": 0.12, + "learning_rate": 0.000195828654118355, + "loss": 0.4966, + "step": 4014 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019582589778411054, + "loss": 0.5943, + "step": 4015 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001958231405589151, + "loss": 0.5419, + "step": 4016 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001958203824427943, + "loss": 0.6141, + "step": 4017 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019581762343577378, + "loss": 0.6724, + "step": 4018 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019581486353787914, + "loss": 0.6107, + "step": 4019 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019581210274913612, + "loss": 0.5114, + "step": 4020 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019580934106957034, + "loss": 0.6882, + "step": 4021 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001958065784992075, + "loss": 0.6133, + "step": 4022 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001958038150380733, + "loss": 0.478, + "step": 4023 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019580105068619337, + "loss": 0.6383, + "step": 4024 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001957982854435935, + "loss": 0.6559, + "step": 4025 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019579551931029932, + "loss": 0.5462, + "step": 4026 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019579275228633658, + "loss": 0.6639, + "step": 4027 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019578998437173102, + "loss": 0.8526, + "step": 4028 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019578721556650836, + "loss": 0.5766, + "step": 4029 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019578444587069435, + "loss": 0.6037, + "step": 4030 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019578167528431473, + "loss": 0.5797, + "step": 4031 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001957789038073953, + "loss": 0.6479, + "step": 4032 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019577613143996177, + "loss": 0.6205, + "step": 4033 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019577335818203993, + "loss": 0.4617, + "step": 4034 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001957705840336556, + "loss": 0.5386, + "step": 4035 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019576780899483454, + "loss": 0.5221, + "step": 4036 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019576503306560254, + "loss": 0.6132, + "step": 4037 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019576225624598546, + "loss": 0.5819, + "step": 4038 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001957594785360091, + "loss": 0.6272, + "step": 4039 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019575669993569924, + "loss": 0.6039, + "step": 4040 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001957539204450818, + "loss": 0.6042, + "step": 4041 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001957511400641825, + "loss": 0.5939, + "step": 4042 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001957483587930273, + "loss": 0.5493, + "step": 4043 + }, + { + "epoch": 0.12, + "learning_rate": 0.000195745576631642, + "loss": 0.5495, + "step": 4044 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019574279358005253, + "loss": 0.5338, + "step": 4045 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019574000963828468, + "loss": 0.471, + "step": 4046 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019573722480636441, + "loss": 0.6804, + "step": 4047 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019573443908431758, + "loss": 0.6259, + "step": 4048 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019573165247217008, + "loss": 0.5582, + "step": 4049 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019572886496994785, + "loss": 0.4996, + "step": 4050 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019572607657767675, + "loss": 0.5881, + "step": 4051 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001957232872953828, + "loss": 0.5035, + "step": 4052 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019572049712309183, + "loss": 0.5957, + "step": 4053 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019571770606082985, + "loss": 0.5633, + "step": 4054 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019571491410862274, + "loss": 0.7296, + "step": 4055 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019571212126649656, + "loss": 0.6489, + "step": 4056 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001957093275344772, + "loss": 0.4961, + "step": 4057 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019570653291259068, + "loss": 0.4493, + "step": 4058 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019570373740086295, + "loss": 0.5876, + "step": 4059 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019570094099932, + "loss": 0.6046, + "step": 4060 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019569814370798787, + "loss": 0.5459, + "step": 4061 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019569534552689251, + "loss": 0.6733, + "step": 4062 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019569254645605998, + "loss": 0.5257, + "step": 4063 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001956897464955163, + "loss": 0.4872, + "step": 4064 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019568694564528748, + "loss": 0.4773, + "step": 4065 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019568414390539962, + "loss": 0.62, + "step": 4066 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019568134127587866, + "loss": 0.6778, + "step": 4067 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019567853775675077, + "loss": 0.557, + "step": 4068 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019567573334804196, + "loss": 0.5757, + "step": 4069 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001956729280497783, + "loss": 0.5691, + "step": 4070 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019567012186198592, + "loss": 0.5746, + "step": 4071 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019566731478469084, + "loss": 0.5273, + "step": 4072 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019566450681791922, + "loss": 0.7, + "step": 4073 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019566169796169714, + "loss": 0.5646, + "step": 4074 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001956588882160507, + "loss": 0.6148, + "step": 4075 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019565607758100607, + "loss": 0.6005, + "step": 4076 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019565326605658934, + "loss": 0.649, + "step": 4077 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019565045364282666, + "loss": 0.5015, + "step": 4078 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019564764033974422, + "loss": 0.637, + "step": 4079 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001956448261473681, + "loss": 0.5301, + "step": 4080 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019564201106572453, + "loss": 0.5286, + "step": 4081 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019563919509483964, + "loss": 0.5361, + "step": 4082 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019563637823473963, + "loss": 0.572, + "step": 4083 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001956335604854507, + "loss": 0.5246, + "step": 4084 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019563074184699906, + "loss": 0.5374, + "step": 4085 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001956279223194109, + "loss": 0.5055, + "step": 4086 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001956251019027124, + "loss": 0.6072, + "step": 4087 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019562228059692984, + "loss": 0.5646, + "step": 4088 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019561945840208938, + "loss": 0.5652, + "step": 4089 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019561663531821738, + "loss": 0.5112, + "step": 4090 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019561381134533998, + "loss": 0.5064, + "step": 4091 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019561098648348345, + "loss": 0.5641, + "step": 4092 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001956081607326741, + "loss": 0.5375, + "step": 4093 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019560533409293816, + "loss": 0.5994, + "step": 4094 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019560250656430194, + "loss": 0.628, + "step": 4095 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001955996781467917, + "loss": 0.5165, + "step": 4096 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001955968488404338, + "loss": 0.6188, + "step": 4097 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019559401864525445, + "loss": 0.5973, + "step": 4098 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019559118756128002, + "loss": 0.5782, + "step": 4099 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019558835558853684, + "loss": 0.6553, + "step": 4100 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019558552272705125, + "loss": 0.4858, + "step": 4101 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019558268897684953, + "loss": 0.5902, + "step": 4102 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019557985433795808, + "loss": 0.4949, + "step": 4103 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019557701881040321, + "loss": 0.4949, + "step": 4104 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019557418239421134, + "loss": 0.5255, + "step": 4105 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019557134508940883, + "loss": 0.5935, + "step": 4106 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019556850689602203, + "loss": 0.5628, + "step": 4107 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019556566781407731, + "loss": 0.525, + "step": 4108 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019556282784360116, + "loss": 0.6482, + "step": 4109 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001955599869846199, + "loss": 0.568, + "step": 4110 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019555714523715993, + "loss": 0.4895, + "step": 4111 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019555430260124776, + "loss": 0.6542, + "step": 4112 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019555145907690976, + "loss": 0.5473, + "step": 4113 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019554861466417235, + "loss": 0.6414, + "step": 4114 + }, + { + "epoch": 0.12, + "learning_rate": 0.000195545769363062, + "loss": 0.5108, + "step": 4115 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019554292317360518, + "loss": 0.5423, + "step": 4116 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019554007609582836, + "loss": 0.5834, + "step": 4117 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019553722812975797, + "loss": 0.6078, + "step": 4118 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019553437927542053, + "loss": 0.4726, + "step": 4119 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001955315295328425, + "loss": 0.5075, + "step": 4120 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019552867890205037, + "loss": 0.5375, + "step": 4121 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019552582738307066, + "loss": 0.5047, + "step": 4122 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019552297497592988, + "loss": 0.5061, + "step": 4123 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019552012168065456, + "loss": 0.6132, + "step": 4124 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019551726749727122, + "loss": 0.5641, + "step": 4125 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019551441242580638, + "loss": 0.5662, + "step": 4126 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019551155646628662, + "loss": 0.5845, + "step": 4127 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019550869961873848, + "loss": 0.5176, + "step": 4128 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001955058418831885, + "loss": 0.6344, + "step": 4129 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019550298325966326, + "loss": 0.6608, + "step": 4130 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019550012374818936, + "loss": 0.5979, + "step": 4131 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019549726334879336, + "loss": 0.6023, + "step": 4132 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001954944020615019, + "loss": 0.6975, + "step": 4133 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001954915398863415, + "loss": 0.7031, + "step": 4134 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019548867682333887, + "loss": 0.56, + "step": 4135 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019548581287252056, + "loss": 0.5772, + "step": 4136 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019548294803391317, + "loss": 0.5453, + "step": 4137 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019548008230754348, + "loss": 0.5555, + "step": 4138 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019547721569343796, + "loss": 0.5076, + "step": 4139 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019547434819162338, + "loss": 0.4895, + "step": 4140 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019547147980212635, + "loss": 0.6458, + "step": 4141 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019546861052497356, + "loss": 0.6566, + "step": 4142 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019546574036019166, + "loss": 0.5403, + "step": 4143 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019546286930780739, + "loss": 0.5277, + "step": 4144 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001954599973678474, + "loss": 0.5738, + "step": 4145 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019545712454033836, + "loss": 0.5987, + "step": 4146 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019545425082530707, + "loss": 0.547, + "step": 4147 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001954513762227802, + "loss": 0.7156, + "step": 4148 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019544850073278445, + "loss": 0.6526, + "step": 4149 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001954456243553466, + "loss": 0.5954, + "step": 4150 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019544274709049338, + "loss": 0.4585, + "step": 4151 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001954398689382515, + "loss": 0.6103, + "step": 4152 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019543698989864778, + "loss": 0.6396, + "step": 4153 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019543410997170896, + "loss": 0.6736, + "step": 4154 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019543122915746184, + "loss": 0.5648, + "step": 4155 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019542834745593318, + "loss": 0.5567, + "step": 4156 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019542546486714976, + "loss": 0.6397, + "step": 4157 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019542258139113844, + "loss": 0.7311, + "step": 4158 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019541969702792594, + "loss": 0.5833, + "step": 4159 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019541681177753916, + "loss": 0.6069, + "step": 4160 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019541392564000488, + "loss": 0.67, + "step": 4161 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019541103861534998, + "loss": 0.5642, + "step": 4162 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019540815070360123, + "loss": 0.6047, + "step": 4163 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019540526190478554, + "loss": 0.5971, + "step": 4164 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019540237221892973, + "loss": 0.5765, + "step": 4165 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001953994816460607, + "loss": 0.4806, + "step": 4166 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019539659018620532, + "loss": 0.6757, + "step": 4167 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019539369783939047, + "loss": 0.6164, + "step": 4168 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019539080460564303, + "loss": 0.5291, + "step": 4169 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001953879104849899, + "loss": 0.6708, + "step": 4170 + }, + { + "epoch": 0.12, + "learning_rate": 0.000195385015477458, + "loss": 0.5298, + "step": 4171 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019538211958307423, + "loss": 0.5784, + "step": 4172 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019537922280186555, + "loss": 0.6057, + "step": 4173 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019537632513385886, + "loss": 0.5804, + "step": 4174 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001953734265790811, + "loss": 0.648, + "step": 4175 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019537052713755926, + "loss": 0.4959, + "step": 4176 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019536762680932027, + "loss": 0.5559, + "step": 4177 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019536472559439108, + "loss": 0.4472, + "step": 4178 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019536182349279868, + "loss": 0.4459, + "step": 4179 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019535892050457005, + "loss": 0.4867, + "step": 4180 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019535601662973217, + "loss": 0.5595, + "step": 4181 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019535311186831208, + "loss": 0.6448, + "step": 4182 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019535020622033676, + "loss": 0.52, + "step": 4183 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001953472996858332, + "loss": 0.5803, + "step": 4184 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019534439226482847, + "loss": 0.6344, + "step": 4185 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019534148395734956, + "loss": 0.5253, + "step": 4186 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019533857476342357, + "loss": 0.5724, + "step": 4187 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001953356646830775, + "loss": 0.5721, + "step": 4188 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019533275371633839, + "loss": 0.4675, + "step": 4189 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019532984186323337, + "loss": 0.5908, + "step": 4190 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019532692912378944, + "loss": 0.6377, + "step": 4191 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001953240154980337, + "loss": 0.5625, + "step": 4192 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001953211009859933, + "loss": 0.522, + "step": 4193 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019531818558769527, + "loss": 0.6222, + "step": 4194 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019531526930316676, + "loss": 0.495, + "step": 4195 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019531235213243487, + "loss": 0.4496, + "step": 4196 + }, + { + "epoch": 0.12, + "learning_rate": 0.0001953094340755267, + "loss": 0.5649, + "step": 4197 + }, + { + "epoch": 0.12, + "learning_rate": 0.00019530651513246939, + "loss": 0.4892, + "step": 4198 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001953035953032901, + "loss": 0.4732, + "step": 4199 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019530067458801599, + "loss": 0.5081, + "step": 4200 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019529775298667414, + "loss": 0.6261, + "step": 4201 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019529483049929178, + "loss": 0.6124, + "step": 4202 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019529190712589607, + "loss": 0.4959, + "step": 4203 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001952889828665142, + "loss": 0.5095, + "step": 4204 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019528605772117335, + "loss": 0.5381, + "step": 4205 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001952831316899007, + "loss": 0.5572, + "step": 4206 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019528020477272345, + "loss": 0.4694, + "step": 4207 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019527727696966887, + "loss": 0.595, + "step": 4208 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001952743482807641, + "loss": 0.5752, + "step": 4209 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019527141870603646, + "loss": 0.6257, + "step": 4210 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001952684882455131, + "loss": 0.4938, + "step": 4211 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001952655568992213, + "loss": 0.5744, + "step": 4212 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019526262466718834, + "loss": 0.5209, + "step": 4213 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019525969154944147, + "loss": 0.627, + "step": 4214 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019525675754600795, + "loss": 0.6473, + "step": 4215 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019525382265691508, + "loss": 0.5496, + "step": 4216 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001952508868821901, + "loss": 0.6082, + "step": 4217 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019524795022186035, + "loss": 0.5158, + "step": 4218 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019524501267595315, + "loss": 0.5553, + "step": 4219 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019524207424449574, + "loss": 0.531, + "step": 4220 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019523913492751552, + "loss": 0.5134, + "step": 4221 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019523619472503975, + "loss": 0.6108, + "step": 4222 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019523325363709582, + "loss": 0.5757, + "step": 4223 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019523031166371105, + "loss": 0.6628, + "step": 4224 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001952273688049128, + "loss": 0.4632, + "step": 4225 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001952244250607284, + "loss": 0.6499, + "step": 4226 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019522148043118527, + "loss": 0.4863, + "step": 4227 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019521853491631075, + "loss": 0.4446, + "step": 4228 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019521558851613228, + "loss": 0.5772, + "step": 4229 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019521264123067717, + "loss": 0.4996, + "step": 4230 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001952096930599729, + "loss": 0.477, + "step": 4231 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019520674400404686, + "loss": 0.5769, + "step": 4232 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019520379406292646, + "loss": 0.6543, + "step": 4233 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001952008432366391, + "loss": 0.5976, + "step": 4234 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019519789152521225, + "loss": 0.6429, + "step": 4235 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019519493892867335, + "loss": 0.5606, + "step": 4236 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019519198544704986, + "loss": 0.6336, + "step": 4237 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019518903108036923, + "loss": 0.5266, + "step": 4238 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019518607582865892, + "loss": 0.6629, + "step": 4239 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001951831196919464, + "loss": 0.5524, + "step": 4240 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001951801626702592, + "loss": 0.6549, + "step": 4241 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019517720476362475, + "loss": 0.6608, + "step": 4242 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001951742459720706, + "loss": 0.5949, + "step": 4243 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019517128629562426, + "loss": 0.537, + "step": 4244 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019516832573431318, + "loss": 0.6379, + "step": 4245 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019516536428816497, + "loss": 0.638, + "step": 4246 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019516240195720713, + "loss": 0.5858, + "step": 4247 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019515943874146724, + "loss": 0.5283, + "step": 4248 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019515647464097278, + "loss": 0.5095, + "step": 4249 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019515350965575135, + "loss": 0.5251, + "step": 4250 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001951505437858305, + "loss": 0.5978, + "step": 4251 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019514757703123782, + "loss": 0.4549, + "step": 4252 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001951446093920009, + "loss": 0.5165, + "step": 4253 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019514164086814734, + "loss": 0.5465, + "step": 4254 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019513867145970468, + "loss": 0.5255, + "step": 4255 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001951357011667006, + "loss": 0.687, + "step": 4256 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019513272998916267, + "loss": 0.6084, + "step": 4257 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019512975792711855, + "loss": 0.513, + "step": 4258 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019512678498059588, + "loss": 0.6214, + "step": 4259 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019512381114962222, + "loss": 0.6424, + "step": 4260 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001951208364342253, + "loss": 0.4884, + "step": 4261 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019511786083443278, + "loss": 0.5983, + "step": 4262 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019511488435027227, + "loss": 0.5306, + "step": 4263 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019511190698177152, + "loss": 0.4688, + "step": 4264 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019510892872895812, + "loss": 0.5369, + "step": 4265 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001951059495918598, + "loss": 0.5972, + "step": 4266 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019510296957050426, + "loss": 0.6109, + "step": 4267 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019509998866491928, + "loss": 0.5966, + "step": 4268 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019509700687513247, + "loss": 0.6078, + "step": 4269 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019509402420117157, + "loss": 0.4654, + "step": 4270 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019509104064306435, + "loss": 0.5826, + "step": 4271 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019508805620083855, + "loss": 0.6075, + "step": 4272 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019508507087452186, + "loss": 0.5774, + "step": 4273 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019508208466414214, + "loss": 0.4424, + "step": 4274 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019507909756972703, + "loss": 0.5771, + "step": 4275 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001950761095913044, + "loss": 0.6812, + "step": 4276 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019507312072890198, + "loss": 0.5801, + "step": 4277 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001950701309825476, + "loss": 0.64, + "step": 4278 + }, + { + "epoch": 0.13, + "learning_rate": 0.000195067140352269, + "loss": 0.565, + "step": 4279 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019506414883809406, + "loss": 0.6484, + "step": 4280 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001950611564400505, + "loss": 0.5847, + "step": 4281 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019505816315816625, + "loss": 0.7675, + "step": 4282 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019505516899246907, + "loss": 0.532, + "step": 4283 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019505217394298678, + "loss": 0.6511, + "step": 4284 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019504917800974728, + "loss": 0.6146, + "step": 4285 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019504618119277844, + "loss": 0.5573, + "step": 4286 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019504318349210803, + "loss": 0.626, + "step": 4287 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019504018490776404, + "loss": 0.6275, + "step": 4288 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019503718543977425, + "loss": 0.7059, + "step": 4289 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001950341850881666, + "loss": 0.5121, + "step": 4290 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019503118385296898, + "loss": 0.6212, + "step": 4291 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019502818173420928, + "loss": 0.6361, + "step": 4292 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019502517873191543, + "loss": 0.5514, + "step": 4293 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001950221748461153, + "loss": 0.5136, + "step": 4294 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019501917007683692, + "loss": 0.6179, + "step": 4295 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019501616442410815, + "loss": 0.5244, + "step": 4296 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019501315788795692, + "loss": 0.5222, + "step": 4297 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019501015046841124, + "loss": 0.5384, + "step": 4298 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019500714216549907, + "loss": 0.5157, + "step": 4299 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019500413297924833, + "loss": 0.7408, + "step": 4300 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019500112290968705, + "loss": 0.5972, + "step": 4301 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019499811195684318, + "loss": 0.4788, + "step": 4302 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001949951001207447, + "loss": 0.6274, + "step": 4303 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019499208740141969, + "loss": 0.6831, + "step": 4304 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019498907379889607, + "loss": 0.5088, + "step": 4305 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019498605931320194, + "loss": 0.6155, + "step": 4306 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019498304394436524, + "loss": 0.4975, + "step": 4307 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019498002769241407, + "loss": 0.6382, + "step": 4308 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001949770105573765, + "loss": 0.4721, + "step": 4309 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019497399253928048, + "loss": 0.5804, + "step": 4310 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019497097363815416, + "loss": 0.4779, + "step": 4311 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019496795385402557, + "loss": 0.6927, + "step": 4312 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001949649331869228, + "loss": 0.6088, + "step": 4313 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019496191163687393, + "loss": 0.6611, + "step": 4314 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019495888920390705, + "loss": 0.4969, + "step": 4315 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019495586588805027, + "loss": 0.5664, + "step": 4316 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019495284168933173, + "loss": 0.6435, + "step": 4317 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019494981660777945, + "loss": 0.5479, + "step": 4318 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019494679064342165, + "loss": 0.6159, + "step": 4319 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019494376379628644, + "loss": 0.5713, + "step": 4320 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019494073606640195, + "loss": 0.5602, + "step": 4321 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019493770745379632, + "loss": 0.4616, + "step": 4322 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019493467795849773, + "loss": 0.5062, + "step": 4323 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019493164758053436, + "loss": 0.4811, + "step": 4324 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019492861631993433, + "loss": 0.5843, + "step": 4325 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001949255841767259, + "loss": 0.596, + "step": 4326 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019492255115093721, + "loss": 0.6198, + "step": 4327 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019491951724259646, + "loss": 0.7074, + "step": 4328 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019491648245173186, + "loss": 0.5607, + "step": 4329 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019491344677837166, + "loss": 0.5891, + "step": 4330 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019491041022254405, + "loss": 0.5806, + "step": 4331 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001949073727842773, + "loss": 0.543, + "step": 4332 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001949043344635996, + "loss": 0.6031, + "step": 4333 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019490129526053923, + "loss": 0.465, + "step": 4334 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001948982551751244, + "loss": 0.5103, + "step": 4335 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019489521420738346, + "loss": 0.5142, + "step": 4336 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019489217235734465, + "loss": 0.5646, + "step": 4337 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001948891296250362, + "loss": 0.6449, + "step": 4338 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019488608601048646, + "loss": 0.6129, + "step": 4339 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019488304151372372, + "loss": 0.5806, + "step": 4340 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019487999613477627, + "loss": 0.5385, + "step": 4341 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019487694987367244, + "loss": 0.5385, + "step": 4342 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019487390273044053, + "loss": 0.5778, + "step": 4343 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001948708547051089, + "loss": 0.62, + "step": 4344 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019486780579770584, + "loss": 0.5467, + "step": 4345 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001948647560082598, + "loss": 0.613, + "step": 4346 + }, + { + "epoch": 0.13, + "learning_rate": 0.000194861705336799, + "loss": 0.4932, + "step": 4347 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001948586537833519, + "loss": 0.4415, + "step": 4348 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019485560134794685, + "loss": 0.5247, + "step": 4349 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001948525480306122, + "loss": 0.4296, + "step": 4350 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019484949383137638, + "loss": 0.5273, + "step": 4351 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001948464387502678, + "loss": 0.6297, + "step": 4352 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001948433827873148, + "loss": 0.5641, + "step": 4353 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019484032594254582, + "loss": 0.5922, + "step": 4354 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001948372682159893, + "loss": 0.499, + "step": 4355 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019483420960767368, + "loss": 0.6679, + "step": 4356 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019483115011762733, + "loss": 0.6353, + "step": 4357 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019482808974587878, + "loss": 0.5955, + "step": 4358 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019482502849245642, + "loss": 0.6421, + "step": 4359 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019482196635738877, + "loss": 0.6101, + "step": 4360 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019481890334070425, + "loss": 0.5977, + "step": 4361 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019481583944243135, + "loss": 0.6787, + "step": 4362 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001948127746625986, + "loss": 0.6266, + "step": 4363 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001948097090012344, + "loss": 0.6796, + "step": 4364 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019480664245836735, + "loss": 0.5166, + "step": 4365 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019480357503402594, + "loss": 0.4893, + "step": 4366 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019480050672823865, + "loss": 0.6428, + "step": 4367 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019479743754103403, + "loss": 0.5446, + "step": 4368 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001947943674724406, + "loss": 0.5729, + "step": 4369 + }, + { + "epoch": 0.13, + "learning_rate": 0.000194791296522487, + "loss": 0.5245, + "step": 4370 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019478822469120165, + "loss": 0.5481, + "step": 4371 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019478515197861315, + "loss": 0.5446, + "step": 4372 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001947820783847501, + "loss": 0.5389, + "step": 4373 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019477900390964107, + "loss": 0.5653, + "step": 4374 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019477592855331465, + "loss": 0.5605, + "step": 4375 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001947728523157994, + "loss": 0.5052, + "step": 4376 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019476977519712393, + "loss": 0.5777, + "step": 4377 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019476669719731689, + "loss": 0.5744, + "step": 4378 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019476361831640687, + "loss": 0.4862, + "step": 4379 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019476053855442248, + "loss": 0.7038, + "step": 4380 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019475745791139234, + "loss": 0.599, + "step": 4381 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019475437638734517, + "loss": 0.596, + "step": 4382 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019475129398230956, + "loss": 0.555, + "step": 4383 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001947482106963142, + "loss": 0.5719, + "step": 4384 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001947451265293877, + "loss": 0.5709, + "step": 4385 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001947420414815588, + "loss": 0.6711, + "step": 4386 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019473895555285618, + "loss": 0.5092, + "step": 4387 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019473586874330846, + "loss": 0.5992, + "step": 4388 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001947327810529444, + "loss": 0.682, + "step": 4389 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019472969248179273, + "loss": 0.608, + "step": 4390 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019472660302988212, + "loss": 0.5568, + "step": 4391 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019472351269724132, + "loss": 0.5803, + "step": 4392 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019472042148389903, + "loss": 0.5012, + "step": 4393 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019471732938988403, + "loss": 0.7382, + "step": 4394 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019471423641522506, + "loss": 0.4468, + "step": 4395 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019471114255995088, + "loss": 0.4731, + "step": 4396 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019470804782409018, + "loss": 0.6226, + "step": 4397 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019470495220767187, + "loss": 0.5926, + "step": 4398 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019470185571072461, + "loss": 0.5113, + "step": 4399 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019469875833327723, + "loss": 0.465, + "step": 4400 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001946956600753586, + "loss": 0.6505, + "step": 4401 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019469256093699743, + "loss": 0.5493, + "step": 4402 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019468946091822257, + "loss": 0.5282, + "step": 4403 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019468636001906283, + "loss": 0.5601, + "step": 4404 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019468325823954706, + "loss": 0.5647, + "step": 4405 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019468015557970412, + "loss": 0.6387, + "step": 4406 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001946770520395628, + "loss": 0.4886, + "step": 4407 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019467394761915198, + "loss": 0.6322, + "step": 4408 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019467084231850054, + "loss": 0.5616, + "step": 4409 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019466773613763735, + "loss": 0.4316, + "step": 4410 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019466462907659127, + "loss": 0.5475, + "step": 4411 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019466152113539121, + "loss": 0.6005, + "step": 4412 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019465841231406603, + "loss": 0.5779, + "step": 4413 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019465530261264467, + "loss": 0.4933, + "step": 4414 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019465219203115602, + "loss": 0.5011, + "step": 4415 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019464908056962904, + "loss": 0.6941, + "step": 4416 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019464596822809264, + "loss": 0.5442, + "step": 4417 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001946428550065757, + "loss": 0.558, + "step": 4418 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019463974090510726, + "loss": 0.4945, + "step": 4419 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019463662592371622, + "loss": 0.5721, + "step": 4420 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019463351006243155, + "loss": 0.4998, + "step": 4421 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001946303933212822, + "loss": 0.4973, + "step": 4422 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001946272757002972, + "loss": 0.642, + "step": 4423 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019462415719950546, + "loss": 0.5022, + "step": 4424 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019462103781893607, + "loss": 0.5487, + "step": 4425 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019461791755861796, + "loss": 0.6643, + "step": 4426 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019461479641858016, + "loss": 0.5418, + "step": 4427 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001946116743988517, + "loss": 0.5549, + "step": 4428 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001946085514994616, + "loss": 0.5246, + "step": 4429 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019460542772043889, + "loss": 0.5865, + "step": 4430 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019460230306181262, + "loss": 0.672, + "step": 4431 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019459917752361183, + "loss": 0.5834, + "step": 4432 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019459605110586564, + "loss": 0.6169, + "step": 4433 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019459292380860302, + "loss": 0.5786, + "step": 4434 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001945897956318531, + "loss": 0.595, + "step": 4435 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019458666657564498, + "loss": 0.5823, + "step": 4436 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001945835366400077, + "loss": 0.6167, + "step": 4437 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019458040582497041, + "loss": 0.5775, + "step": 4438 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019457727413056222, + "loss": 0.574, + "step": 4439 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001945741415568122, + "loss": 0.5851, + "step": 4440 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019457100810374954, + "loss": 0.5922, + "step": 4441 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001945678737714033, + "loss": 0.5517, + "step": 4442 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019456473855980271, + "loss": 0.6085, + "step": 4443 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019456160246897683, + "loss": 0.6584, + "step": 4444 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019455846549895487, + "loss": 0.5724, + "step": 4445 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019455532764976597, + "loss": 0.533, + "step": 4446 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019455218892143935, + "loss": 0.552, + "step": 4447 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019454904931400416, + "loss": 0.511, + "step": 4448 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019454590882748957, + "loss": 0.6081, + "step": 4449 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019454276746192482, + "loss": 0.574, + "step": 4450 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019453962521733908, + "loss": 0.586, + "step": 4451 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001945364820937616, + "loss": 0.4491, + "step": 4452 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001945333380912216, + "loss": 0.4876, + "step": 4453 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019453019320974827, + "loss": 0.6135, + "step": 4454 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001945270474493709, + "loss": 0.5922, + "step": 4455 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001945239008101187, + "loss": 0.5788, + "step": 4456 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019452075329202097, + "loss": 0.6573, + "step": 4457 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019451760489510694, + "loss": 0.5575, + "step": 4458 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019451445561940588, + "loss": 0.4586, + "step": 4459 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001945113054649471, + "loss": 0.5874, + "step": 4460 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019450815443175984, + "loss": 0.5411, + "step": 4461 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019450500251987345, + "loss": 0.6051, + "step": 4462 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001945018497293172, + "loss": 0.5135, + "step": 4463 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001944986960601204, + "loss": 0.5444, + "step": 4464 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001944955415123124, + "loss": 0.4772, + "step": 4465 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019449238608592256, + "loss": 0.5709, + "step": 4466 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019448922978098013, + "loss": 0.5568, + "step": 4467 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019448607259751451, + "loss": 0.5654, + "step": 4468 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019448291453555506, + "loss": 0.5247, + "step": 4469 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019447975559513112, + "loss": 0.538, + "step": 4470 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019447659577627205, + "loss": 0.509, + "step": 4471 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019447343507900727, + "loss": 0.5621, + "step": 4472 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019447027350336614, + "loss": 0.6038, + "step": 4473 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019446711104937806, + "loss": 0.6512, + "step": 4474 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001944639477170724, + "loss": 0.6529, + "step": 4475 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019446078350647864, + "loss": 0.6154, + "step": 4476 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019445761841762615, + "loss": 0.6627, + "step": 4477 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001944544524505444, + "loss": 0.6209, + "step": 4478 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019445128560526275, + "loss": 0.6495, + "step": 4479 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001944481178818107, + "loss": 0.5825, + "step": 4480 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019444494928021772, + "loss": 0.6352, + "step": 4481 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019444177980051323, + "loss": 0.4571, + "step": 4482 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001944386094427267, + "loss": 0.6578, + "step": 4483 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019443543820688765, + "loss": 0.5233, + "step": 4484 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019443226609302551, + "loss": 0.4835, + "step": 4485 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019442909310116982, + "loss": 0.5399, + "step": 4486 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019442591923135002, + "loss": 0.5367, + "step": 4487 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019442274448359568, + "loss": 0.4729, + "step": 4488 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001944195688579363, + "loss": 0.5776, + "step": 4489 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001944163923544014, + "loss": 0.5974, + "step": 4490 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001944132149730205, + "loss": 0.6291, + "step": 4491 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019441003671382317, + "loss": 0.5484, + "step": 4492 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001944068575768389, + "loss": 0.5971, + "step": 4493 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019440367756209736, + "loss": 0.4864, + "step": 4494 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019440049666962804, + "loss": 0.6724, + "step": 4495 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001943973148994605, + "loss": 0.4767, + "step": 4496 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019439413225162436, + "loss": 0.572, + "step": 4497 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019439094872614922, + "loss": 0.5599, + "step": 4498 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019438776432306463, + "loss": 0.5095, + "step": 4499 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019438457904240024, + "loss": 0.4368, + "step": 4500 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019438139288418564, + "loss": 0.5283, + "step": 4501 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001943782058484505, + "loss": 0.4939, + "step": 4502 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001943750179352244, + "loss": 0.5981, + "step": 4503 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019437182914453698, + "loss": 0.6528, + "step": 4504 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001943686394764179, + "loss": 0.5271, + "step": 4505 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019436544893089685, + "loss": 0.6077, + "step": 4506 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019436225750800348, + "loss": 0.6044, + "step": 4507 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019435906520776742, + "loss": 0.5388, + "step": 4508 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019435587203021836, + "loss": 0.4806, + "step": 4509 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019435267797538602, + "loss": 0.5441, + "step": 4510 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019434948304330012, + "loss": 0.4905, + "step": 4511 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019434628723399028, + "loss": 0.6516, + "step": 4512 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019434309054748632, + "loss": 0.6493, + "step": 4513 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019433989298381788, + "loss": 0.5493, + "step": 4514 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001943366945430147, + "loss": 0.4555, + "step": 4515 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019433349522510657, + "loss": 0.6137, + "step": 4516 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019433029503012315, + "loss": 0.6343, + "step": 4517 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001943270939580943, + "loss": 0.6284, + "step": 4518 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019432389200904968, + "loss": 0.5371, + "step": 4519 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001943206891830191, + "loss": 0.6353, + "step": 4520 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019431748548003237, + "loss": 0.4647, + "step": 4521 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019431428090011926, + "loss": 0.6413, + "step": 4522 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019431107544330952, + "loss": 0.6318, + "step": 4523 + }, + { + "epoch": 0.13, + "learning_rate": 0.000194307869109633, + "loss": 0.6261, + "step": 4524 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019430466189911951, + "loss": 0.6218, + "step": 4525 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019430145381179887, + "loss": 0.5882, + "step": 4526 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001942982448477009, + "loss": 0.6073, + "step": 4527 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019429503500685542, + "loss": 0.4862, + "step": 4528 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019429182428929229, + "loss": 0.4275, + "step": 4529 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019428861269504136, + "loss": 0.5238, + "step": 4530 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019428540022413247, + "loss": 0.6397, + "step": 4531 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019428218687659552, + "loss": 0.6451, + "step": 4532 + }, + { + "epoch": 0.13, + "learning_rate": 0.00019427897265246038, + "loss": 0.6321, + "step": 4533 + }, + { + "epoch": 0.13, + "learning_rate": 0.0001942757575517569, + "loss": 0.5337, + "step": 4534 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019427254157451508, + "loss": 0.5234, + "step": 4535 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019426932472076466, + "loss": 0.6079, + "step": 4536 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019426610699053564, + "loss": 0.5498, + "step": 4537 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019426288838385796, + "loss": 0.4854, + "step": 4538 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001942596689007615, + "loss": 0.672, + "step": 4539 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019425644854127622, + "loss": 0.6967, + "step": 4540 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019425322730543202, + "loss": 0.5628, + "step": 4541 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001942500051932589, + "loss": 0.5538, + "step": 4542 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019424678220478678, + "loss": 0.5492, + "step": 4543 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019424355834004567, + "loss": 0.6858, + "step": 4544 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019424033359906546, + "loss": 0.6147, + "step": 4545 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019423710798187625, + "loss": 0.5307, + "step": 4546 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019423388148850793, + "loss": 0.5242, + "step": 4547 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019423065411899054, + "loss": 0.5309, + "step": 4548 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019422742587335413, + "loss": 0.6002, + "step": 4549 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001942241967516286, + "loss": 0.5984, + "step": 4550 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019422096675384413, + "loss": 0.6079, + "step": 4551 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019421773588003058, + "loss": 0.5729, + "step": 4552 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001942145041302181, + "loss": 0.5678, + "step": 4553 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019421127150443673, + "loss": 0.5618, + "step": 4554 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019420803800271647, + "loss": 0.5489, + "step": 4555 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019420480362508747, + "loss": 0.547, + "step": 4556 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019420156837157972, + "loss": 0.5542, + "step": 4557 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019419833224222334, + "loss": 0.5381, + "step": 4558 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019419509523704842, + "loss": 0.4717, + "step": 4559 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019419185735608502, + "loss": 0.5747, + "step": 4560 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019418861859936328, + "loss": 0.5846, + "step": 4561 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019418537896691334, + "loss": 0.6123, + "step": 4562 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019418213845876525, + "loss": 0.5672, + "step": 4563 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019417889707494917, + "loss": 0.6076, + "step": 4564 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019417565481549525, + "loss": 0.5932, + "step": 4565 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019417241168043364, + "loss": 0.6055, + "step": 4566 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019416916766979445, + "loss": 0.5563, + "step": 4567 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001941659227836079, + "loss": 0.542, + "step": 4568 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019416267702190412, + "loss": 0.619, + "step": 4569 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019415943038471328, + "loss": 0.5153, + "step": 4570 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019415618287206562, + "loss": 0.565, + "step": 4571 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019415293448399128, + "loss": 0.5961, + "step": 4572 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019414968522052045, + "loss": 0.6446, + "step": 4573 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019414643508168343, + "loss": 0.5626, + "step": 4574 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019414318406751034, + "loss": 0.591, + "step": 4575 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019413993217803145, + "loss": 0.6477, + "step": 4576 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019413667941327702, + "loss": 0.5382, + "step": 4577 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019413342577327724, + "loss": 0.5431, + "step": 4578 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019413017125806238, + "loss": 0.5025, + "step": 4579 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019412691586766272, + "loss": 0.5651, + "step": 4580 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019412365960210849, + "loss": 0.5806, + "step": 4581 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019412040246143, + "loss": 0.5936, + "step": 4582 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019411714444565753, + "loss": 0.5102, + "step": 4583 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019411388555482138, + "loss": 0.6888, + "step": 4584 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001941106257889518, + "loss": 0.6356, + "step": 4585 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019410736514807915, + "loss": 0.6014, + "step": 4586 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001941041036322337, + "loss": 0.5884, + "step": 4587 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001941008412414458, + "loss": 0.5848, + "step": 4588 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001940975779757458, + "loss": 0.4532, + "step": 4589 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019409431383516404, + "loss": 0.574, + "step": 4590 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001940910488197308, + "loss": 0.5645, + "step": 4591 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019408778292947653, + "loss": 0.5731, + "step": 4592 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019408451616443158, + "loss": 0.4726, + "step": 4593 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019408124852462623, + "loss": 0.4957, + "step": 4594 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019407798001009097, + "loss": 0.6653, + "step": 4595 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019407471062085613, + "loss": 0.4806, + "step": 4596 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019407144035695213, + "loss": 0.5226, + "step": 4597 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019406816921840937, + "loss": 0.5985, + "step": 4598 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019406489720525828, + "loss": 0.4986, + "step": 4599 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019406162431752927, + "loss": 0.6065, + "step": 4600 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019405835055525273, + "loss": 0.6579, + "step": 4601 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019405507591845918, + "loss": 0.492, + "step": 4602 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019405180040717897, + "loss": 0.4654, + "step": 4603 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019404852402144264, + "loss": 0.5812, + "step": 4604 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019404524676128063, + "loss": 0.6567, + "step": 4605 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019404196862672337, + "loss": 0.5771, + "step": 4606 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001940386896178014, + "loss": 0.5649, + "step": 4607 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019403540973454513, + "loss": 0.5237, + "step": 4608 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019403212897698514, + "loss": 0.5152, + "step": 4609 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019402884734515184, + "loss": 0.552, + "step": 4610 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019402556483907583, + "loss": 0.5727, + "step": 4611 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019402228145878758, + "loss": 0.6033, + "step": 4612 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019401899720431764, + "loss": 0.4902, + "step": 4613 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019401571207569655, + "loss": 0.6023, + "step": 4614 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019401242607295483, + "loss": 0.578, + "step": 4615 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019400913919612303, + "loss": 0.6189, + "step": 4616 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019400585144523173, + "loss": 0.5373, + "step": 4617 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001940025628203115, + "loss": 0.623, + "step": 4618 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019399927332139287, + "loss": 0.6024, + "step": 4619 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001939959829485065, + "loss": 0.5662, + "step": 4620 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019399269170168292, + "loss": 0.5273, + "step": 4621 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019398939958095276, + "loss": 0.588, + "step": 4622 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019398610658634665, + "loss": 0.4359, + "step": 4623 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019398281271789514, + "loss": 0.4525, + "step": 4624 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001939795179756289, + "loss": 0.6508, + "step": 4625 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019397622235957859, + "loss": 0.6157, + "step": 4626 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001939729258697748, + "loss": 0.5026, + "step": 4627 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001939696285062482, + "loss": 0.564, + "step": 4628 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019396633026902947, + "loss": 0.6071, + "step": 4629 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001939630311581492, + "loss": 0.6548, + "step": 4630 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019395973117363816, + "loss": 0.5107, + "step": 4631 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019395643031552698, + "loss": 0.6545, + "step": 4632 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019395312858384633, + "loss": 0.6095, + "step": 4633 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019394982597862698, + "loss": 0.6161, + "step": 4634 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019394652249989956, + "loss": 0.5947, + "step": 4635 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019394321814769484, + "loss": 0.5246, + "step": 4636 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001939399129220435, + "loss": 0.5638, + "step": 4637 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001939366068229763, + "loss": 0.6256, + "step": 4638 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019393329985052397, + "loss": 0.6264, + "step": 4639 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019392999200471724, + "loss": 0.6858, + "step": 4640 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001939266832855869, + "loss": 0.563, + "step": 4641 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001939233736931637, + "loss": 0.5764, + "step": 4642 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001939200632274784, + "loss": 0.5944, + "step": 4643 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019391675188856175, + "loss": 0.6811, + "step": 4644 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019391343967644462, + "loss": 0.5959, + "step": 4645 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019391012659115774, + "loss": 0.6521, + "step": 4646 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019390681263273195, + "loss": 0.5485, + "step": 4647 + }, + { + "epoch": 0.14, + "learning_rate": 0.000193903497801198, + "loss": 0.4935, + "step": 4648 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001939001820965868, + "loss": 0.5014, + "step": 4649 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019389686551892908, + "loss": 0.5509, + "step": 4650 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019389354806825576, + "loss": 0.5241, + "step": 4651 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019389022974459767, + "loss": 0.4688, + "step": 4652 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019388691054798564, + "loss": 0.5627, + "step": 4653 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019388359047845053, + "loss": 0.5638, + "step": 4654 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001938802695360232, + "loss": 0.7082, + "step": 4655 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019387694772073454, + "loss": 0.5707, + "step": 4656 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019387362503261546, + "loss": 0.7105, + "step": 4657 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001938703014716968, + "loss": 0.7596, + "step": 4658 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001938669770380095, + "loss": 0.5394, + "step": 4659 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019386365173158448, + "loss": 0.6534, + "step": 4660 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019386032555245258, + "loss": 0.6497, + "step": 4661 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019385699850064483, + "loss": 0.5424, + "step": 4662 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001938536705761921, + "loss": 0.5588, + "step": 4663 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019385034177912534, + "loss": 0.6116, + "step": 4664 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019384701210947553, + "loss": 0.5126, + "step": 4665 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001938436815672736, + "loss": 0.5987, + "step": 4666 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019384035015255054, + "loss": 0.5177, + "step": 4667 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019383701786533726, + "loss": 0.5486, + "step": 4668 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001938336847056648, + "loss": 0.5157, + "step": 4669 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019383035067356414, + "loss": 0.5319, + "step": 4670 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001938270157690663, + "loss": 0.5626, + "step": 4671 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019382367999220223, + "loss": 0.5244, + "step": 4672 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019382034334300302, + "loss": 0.6183, + "step": 4673 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019381700582149963, + "loss": 0.4799, + "step": 4674 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001938136674277231, + "loss": 0.5888, + "step": 4675 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019381032816170448, + "loss": 0.5161, + "step": 4676 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019380698802347484, + "loss": 0.5174, + "step": 4677 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019380364701306518, + "loss": 0.6001, + "step": 4678 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019380030513050663, + "loss": 0.5682, + "step": 4679 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019379696237583022, + "loss": 0.5813, + "step": 4680 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019379361874906703, + "loss": 0.6035, + "step": 4681 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019379027425024813, + "loss": 0.5333, + "step": 4682 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019378692887940467, + "loss": 0.4177, + "step": 4683 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019378358263656773, + "loss": 0.4688, + "step": 4684 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001937802355217684, + "loss": 0.5666, + "step": 4685 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001937768875350378, + "loss": 0.6179, + "step": 4686 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019377353867640714, + "loss": 0.572, + "step": 4687 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019377018894590746, + "loss": 0.5027, + "step": 4688 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001937668383435699, + "loss": 0.5642, + "step": 4689 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019376348686942572, + "loss": 0.4513, + "step": 4690 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019376013452350593, + "loss": 0.5615, + "step": 4691 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019375678130584186, + "loss": 0.5231, + "step": 4692 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019375342721646456, + "loss": 0.7215, + "step": 4693 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001937500722554052, + "loss": 0.5482, + "step": 4694 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019374671642269513, + "loss": 0.6652, + "step": 4695 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001937433597183654, + "loss": 0.6078, + "step": 4696 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019374000214244732, + "loss": 0.6062, + "step": 4697 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019373664369497204, + "loss": 0.5609, + "step": 4698 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001937332843759708, + "loss": 0.5535, + "step": 4699 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019372992418547483, + "loss": 0.5671, + "step": 4700 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001937265631235154, + "loss": 0.5486, + "step": 4701 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019372320119012373, + "loss": 0.5287, + "step": 4702 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001937198383853311, + "loss": 0.5755, + "step": 4703 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019371647470916876, + "loss": 0.4894, + "step": 4704 + }, + { + "epoch": 0.14, + "learning_rate": 0.000193713110161668, + "loss": 0.5962, + "step": 4705 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019370974474286007, + "loss": 0.7014, + "step": 4706 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001937063784527763, + "loss": 0.572, + "step": 4707 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019370301129144796, + "loss": 0.5701, + "step": 4708 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001936996432589064, + "loss": 0.4773, + "step": 4709 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019369627435518284, + "loss": 0.6175, + "step": 4710 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019369290458030872, + "loss": 0.4477, + "step": 4711 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001936895339343153, + "loss": 0.5966, + "step": 4712 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019368616241723392, + "loss": 0.6248, + "step": 4713 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019368279002909594, + "loss": 0.6006, + "step": 4714 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019367941676993273, + "loss": 0.4622, + "step": 4715 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019367604263977563, + "loss": 0.5576, + "step": 4716 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019367266763865604, + "loss": 0.4969, + "step": 4717 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001936692917666053, + "loss": 0.5476, + "step": 4718 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019366591502365484, + "loss": 0.4869, + "step": 4719 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019366253740983602, + "loss": 0.6218, + "step": 4720 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019365915892518025, + "loss": 0.5818, + "step": 4721 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019365577956971896, + "loss": 0.5604, + "step": 4722 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019365239934348357, + "loss": 0.6218, + "step": 4723 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019364901824650546, + "loss": 0.6413, + "step": 4724 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019364563627881617, + "loss": 0.5052, + "step": 4725 + }, + { + "epoch": 0.14, + "learning_rate": 0.000193642253440447, + "loss": 0.6923, + "step": 4726 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019363886973142953, + "loss": 0.5955, + "step": 4727 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019363548515179515, + "loss": 0.5511, + "step": 4728 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019363209970157534, + "loss": 0.5603, + "step": 4729 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019362871338080159, + "loss": 0.54, + "step": 4730 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001936253261895054, + "loss": 0.5008, + "step": 4731 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019362193812771822, + "loss": 0.5951, + "step": 4732 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001936185491954716, + "loss": 0.6113, + "step": 4733 + }, + { + "epoch": 0.14, + "learning_rate": 0.000193615159392797, + "loss": 0.7294, + "step": 4734 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019361176871972598, + "loss": 0.5267, + "step": 4735 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019360837717629004, + "loss": 0.5124, + "step": 4736 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019360498476252066, + "loss": 0.6689, + "step": 4737 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019360159147844953, + "loss": 0.5975, + "step": 4738 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019359819732410805, + "loss": 0.5696, + "step": 4739 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019359480229952784, + "loss": 0.6084, + "step": 4740 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019359140640474046, + "loss": 0.5299, + "step": 4741 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019358800963977752, + "loss": 0.5596, + "step": 4742 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019358461200467053, + "loss": 0.6132, + "step": 4743 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019358121349945113, + "loss": 0.6217, + "step": 4744 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001935778141241509, + "loss": 0.6488, + "step": 4745 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019357441387880147, + "loss": 0.5323, + "step": 4746 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001935710127634344, + "loss": 0.5403, + "step": 4747 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019356761077808136, + "loss": 0.5034, + "step": 4748 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019356420792277402, + "loss": 0.5586, + "step": 4749 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001935608041975439, + "loss": 0.5079, + "step": 4750 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019355739960242273, + "loss": 0.4738, + "step": 4751 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019355399413744216, + "loss": 0.6764, + "step": 4752 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001935505878026338, + "loss": 0.6041, + "step": 4753 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019354718059802942, + "loss": 0.5105, + "step": 4754 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001935437725236606, + "loss": 0.5865, + "step": 4755 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019354036357955906, + "loss": 0.6165, + "step": 4756 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019353695376575654, + "loss": 0.5641, + "step": 4757 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019353354308228467, + "loss": 0.5917, + "step": 4758 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001935301315291752, + "loss": 0.5825, + "step": 4759 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019352671910645986, + "loss": 0.5444, + "step": 4760 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019352330581417034, + "loss": 0.5689, + "step": 4761 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001935198916523384, + "loss": 0.6559, + "step": 4762 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019351647662099579, + "loss": 0.5961, + "step": 4763 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019351306072017422, + "loss": 0.545, + "step": 4764 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019350964394990554, + "loss": 0.5878, + "step": 4765 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019350622631022143, + "loss": 0.6372, + "step": 4766 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001935028078011537, + "loss": 0.583, + "step": 4767 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001934993884227341, + "loss": 0.601, + "step": 4768 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001934959681749945, + "loss": 0.5462, + "step": 4769 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019349254705796663, + "loss": 0.7197, + "step": 4770 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019348912507168234, + "loss": 0.6165, + "step": 4771 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001934857022161734, + "loss": 0.5474, + "step": 4772 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019348227849147167, + "loss": 0.4641, + "step": 4773 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019347885389760897, + "loss": 0.5548, + "step": 4774 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019347542843461716, + "loss": 0.5003, + "step": 4775 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019347200210252804, + "loss": 0.4825, + "step": 4776 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019346857490137352, + "loss": 0.5598, + "step": 4777 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019346514683118545, + "loss": 0.6691, + "step": 4778 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019346171789199566, + "loss": 0.5075, + "step": 4779 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019345828808383611, + "loss": 0.4575, + "step": 4780 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019345485740673862, + "loss": 0.5392, + "step": 4781 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019345142586073513, + "loss": 0.539, + "step": 4782 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001934479934458575, + "loss": 0.5358, + "step": 4783 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001934445601621377, + "loss": 0.4459, + "step": 4784 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019344112600960762, + "loss": 0.5668, + "step": 4785 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019343769098829918, + "loss": 0.6649, + "step": 4786 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019343425509824435, + "loss": 0.6282, + "step": 4787 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019343081833947504, + "loss": 0.5667, + "step": 4788 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019342738071202322, + "loss": 0.5545, + "step": 4789 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019342394221592084, + "loss": 0.5865, + "step": 4790 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001934205028511999, + "loss": 0.5495, + "step": 4791 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019341706261789235, + "loss": 0.5873, + "step": 4792 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019341362151603018, + "loss": 0.4532, + "step": 4793 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019341017954564537, + "loss": 0.5605, + "step": 4794 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019340673670677, + "loss": 0.5957, + "step": 4795 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019340329299943597, + "loss": 0.526, + "step": 4796 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019339984842367536, + "loss": 0.6162, + "step": 4797 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001933964029795202, + "loss": 0.5379, + "step": 4798 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019339295666700247, + "loss": 0.7814, + "step": 4799 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019338950948615428, + "loss": 0.5926, + "step": 4800 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019338606143700767, + "loss": 0.6276, + "step": 4801 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019338261251959464, + "loss": 0.5347, + "step": 4802 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001933791627339473, + "loss": 0.5468, + "step": 4803 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019337571208009777, + "loss": 0.5593, + "step": 4804 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019337226055807805, + "loss": 0.522, + "step": 4805 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019336880816792028, + "loss": 0.6532, + "step": 4806 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019336535490965652, + "loss": 0.5124, + "step": 4807 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019336190078331891, + "loss": 0.6019, + "step": 4808 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019335844578893956, + "loss": 0.5928, + "step": 4809 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001933549899265506, + "loss": 0.6284, + "step": 4810 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019335153319618414, + "loss": 0.5305, + "step": 4811 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019334807559787237, + "loss": 0.647, + "step": 4812 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019334461713164735, + "loss": 0.5845, + "step": 4813 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001933411577975413, + "loss": 0.5732, + "step": 4814 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019333769759558635, + "loss": 0.5555, + "step": 4815 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001933342365258147, + "loss": 0.5475, + "step": 4816 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019333077458825852, + "loss": 0.5816, + "step": 4817 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019332731178295002, + "loss": 0.6399, + "step": 4818 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019332384810992131, + "loss": 0.6531, + "step": 4819 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019332038356920472, + "loss": 0.5858, + "step": 4820 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019331691816083237, + "loss": 0.4064, + "step": 4821 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001933134518848365, + "loss": 0.6555, + "step": 4822 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019330998474124936, + "loss": 0.5414, + "step": 4823 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019330651673010317, + "loss": 0.5577, + "step": 4824 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019330304785143015, + "loss": 0.5565, + "step": 4825 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019329957810526263, + "loss": 0.5221, + "step": 4826 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019329610749163276, + "loss": 0.5727, + "step": 4827 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001932926360105729, + "loss": 0.6983, + "step": 4828 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001932891636621153, + "loss": 0.5905, + "step": 4829 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001932856904462922, + "loss": 0.5673, + "step": 4830 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019328221636313598, + "loss": 0.5803, + "step": 4831 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019327874141267888, + "loss": 0.5045, + "step": 4832 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001932752655949532, + "loss": 0.6108, + "step": 4833 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019327178890999128, + "loss": 0.5227, + "step": 4834 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019326831135782545, + "loss": 0.468, + "step": 4835 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019326483293848802, + "loss": 0.4792, + "step": 4836 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019326135365201134, + "loss": 0.5518, + "step": 4837 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001932578734984278, + "loss": 0.4904, + "step": 4838 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019325439247776967, + "loss": 0.5492, + "step": 4839 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019325091059006942, + "loss": 0.6077, + "step": 4840 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001932474278353593, + "loss": 0.5781, + "step": 4841 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019324394421367186, + "loss": 0.5599, + "step": 4842 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001932404597250393, + "loss": 0.65, + "step": 4843 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019323697436949414, + "loss": 0.4704, + "step": 4844 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019323348814706876, + "loss": 0.5967, + "step": 4845 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019323000105779555, + "loss": 0.5549, + "step": 4846 + }, + { + "epoch": 0.14, + "learning_rate": 0.000193226513101707, + "loss": 0.656, + "step": 4847 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019322302427883541, + "loss": 0.5027, + "step": 4848 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019321953458921335, + "loss": 0.559, + "step": 4849 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001932160440328732, + "loss": 0.4755, + "step": 4850 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019321255260984737, + "loss": 0.5257, + "step": 4851 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019320906032016844, + "loss": 0.5702, + "step": 4852 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019320556716386883, + "loss": 0.5216, + "step": 4853 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019320207314098097, + "loss": 0.6012, + "step": 4854 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019319857825153737, + "loss": 0.5783, + "step": 4855 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019319508249557058, + "loss": 0.6812, + "step": 4856 + }, + { + "epoch": 0.14, + "learning_rate": 0.000193191585873113, + "loss": 0.5645, + "step": 4857 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001931880883841972, + "loss": 0.5184, + "step": 4858 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019318459002885577, + "loss": 0.4177, + "step": 4859 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019318109080712108, + "loss": 0.5872, + "step": 4860 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001931775907190258, + "loss": 0.627, + "step": 4861 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019317408976460239, + "loss": 0.5793, + "step": 4862 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019317058794388343, + "loss": 0.5088, + "step": 4863 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019316708525690148, + "loss": 0.5475, + "step": 4864 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019316358170368912, + "loss": 0.5596, + "step": 4865 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019316007728427888, + "loss": 0.5673, + "step": 4866 + }, + { + "epoch": 0.14, + "learning_rate": 0.0001931565719987034, + "loss": 0.6056, + "step": 4867 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019315306584699523, + "loss": 0.5518, + "step": 4868 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019314955882918696, + "loss": 0.628, + "step": 4869 + }, + { + "epoch": 0.14, + "learning_rate": 0.00019314605094531125, + "loss": 0.6152, + "step": 4870 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019314254219540068, + "loss": 0.5119, + "step": 4871 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019313903257948788, + "loss": 0.6015, + "step": 4872 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019313552209760542, + "loss": 0.6338, + "step": 4873 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019313201074978604, + "loss": 0.4931, + "step": 4874 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019312849853606235, + "loss": 0.5571, + "step": 4875 + }, + { + "epoch": 0.15, + "learning_rate": 0.000193124985456467, + "loss": 0.5693, + "step": 4876 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019312147151103265, + "loss": 0.5654, + "step": 4877 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019311795669979195, + "loss": 0.542, + "step": 4878 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001931144410227776, + "loss": 0.5816, + "step": 4879 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001931109244800223, + "loss": 0.632, + "step": 4880 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019310740707155873, + "loss": 0.4723, + "step": 4881 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019310388879741964, + "loss": 0.5832, + "step": 4882 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019310036965763765, + "loss": 0.5314, + "step": 4883 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019309684965224556, + "loss": 0.5181, + "step": 4884 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019309332878127605, + "loss": 0.5547, + "step": 4885 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019308980704476185, + "loss": 0.6374, + "step": 4886 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019308628444273575, + "loss": 0.6199, + "step": 4887 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001930827609752305, + "loss": 0.5735, + "step": 4888 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001930792366422788, + "loss": 0.6114, + "step": 4889 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019307571144391346, + "loss": 0.6407, + "step": 4890 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019307218538016727, + "loss": 0.5828, + "step": 4891 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019306865845107298, + "loss": 0.6735, + "step": 4892 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001930651306566634, + "loss": 0.4643, + "step": 4893 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019306160199697133, + "loss": 0.6568, + "step": 4894 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001930580724720296, + "loss": 0.5192, + "step": 4895 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019305454208187097, + "loss": 0.6543, + "step": 4896 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001930510108265283, + "loss": 0.579, + "step": 4897 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019304747870603443, + "loss": 0.5397, + "step": 4898 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019304394572042217, + "loss": 0.6306, + "step": 4899 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001930404118697244, + "loss": 0.4782, + "step": 4900 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019303687715397397, + "loss": 0.5401, + "step": 4901 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001930333415732037, + "loss": 0.601, + "step": 4902 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019302980512744654, + "loss": 0.5215, + "step": 4903 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001930262678167353, + "loss": 0.5713, + "step": 4904 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019302272964110293, + "loss": 0.4977, + "step": 4905 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019301919060058228, + "loss": 0.5329, + "step": 4906 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019301565069520626, + "loss": 0.548, + "step": 4907 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001930121099250078, + "loss": 0.6185, + "step": 4908 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001930085682900198, + "loss": 0.6605, + "step": 4909 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019300502579027522, + "loss": 0.5997, + "step": 4910 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019300148242580698, + "loss": 0.5673, + "step": 4911 + }, + { + "epoch": 0.15, + "learning_rate": 0.000192997938196648, + "loss": 0.5734, + "step": 4912 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019299439310283128, + "loss": 0.6418, + "step": 4913 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019299084714438975, + "loss": 0.5771, + "step": 4914 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001929873003213564, + "loss": 0.5936, + "step": 4915 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019298375263376416, + "loss": 0.6121, + "step": 4916 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019298020408164607, + "loss": 0.622, + "step": 4917 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001929766546650351, + "loss": 0.5839, + "step": 4918 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019297310438396427, + "loss": 0.5735, + "step": 4919 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019296955323846652, + "loss": 0.4676, + "step": 4920 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019296600122857495, + "loss": 0.4292, + "step": 4921 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019296244835432255, + "loss": 0.5431, + "step": 4922 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019295889461574234, + "loss": 0.5314, + "step": 4923 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019295534001286741, + "loss": 0.5464, + "step": 4924 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019295178454573075, + "loss": 0.6328, + "step": 4925 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019294822821436548, + "loss": 0.5341, + "step": 4926 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019294467101880458, + "loss": 0.5208, + "step": 4927 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019294111295908118, + "loss": 0.564, + "step": 4928 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001929375540352284, + "loss": 0.704, + "step": 4929 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019293399424727924, + "loss": 0.5392, + "step": 4930 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019293043359526685, + "loss": 0.564, + "step": 4931 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019292687207922438, + "loss": 0.5728, + "step": 4932 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019292330969918482, + "loss": 0.6198, + "step": 4933 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019291974645518139, + "loss": 0.5948, + "step": 4934 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019291618234724717, + "loss": 0.5352, + "step": 4935 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019291261737541536, + "loss": 0.6154, + "step": 4936 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019290905153971903, + "loss": 0.7003, + "step": 4937 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001929054848401914, + "loss": 0.5384, + "step": 4938 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019290191727686556, + "loss": 0.6431, + "step": 4939 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019289834884977475, + "loss": 0.5719, + "step": 4940 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001928947795589521, + "loss": 0.5044, + "step": 4941 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019289120940443083, + "loss": 0.5119, + "step": 4942 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001928876383862441, + "loss": 0.5815, + "step": 4943 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019288406650442514, + "loss": 0.554, + "step": 4944 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019288049375900716, + "loss": 0.4885, + "step": 4945 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019287692015002334, + "loss": 0.5454, + "step": 4946 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019287334567750697, + "loss": 0.6325, + "step": 4947 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019286977034149123, + "loss": 0.621, + "step": 4948 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019286619414200936, + "loss": 0.6574, + "step": 4949 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001928626170790947, + "loss": 0.5382, + "step": 4950 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019285903915278033, + "loss": 0.5552, + "step": 4951 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001928554603630997, + "loss": 0.5301, + "step": 4952 + }, + { + "epoch": 0.15, + "learning_rate": 0.000192851880710086, + "loss": 0.6469, + "step": 4953 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001928483001937725, + "loss": 0.5668, + "step": 4954 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001928447188141925, + "loss": 0.5637, + "step": 4955 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019284113657137933, + "loss": 0.6401, + "step": 4956 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019283755346536627, + "loss": 0.6543, + "step": 4957 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019283396949618668, + "loss": 0.5298, + "step": 4958 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001928303846638738, + "loss": 0.6018, + "step": 4959 + }, + { + "epoch": 0.15, + "learning_rate": 0.000192826798968461, + "loss": 0.6176, + "step": 4960 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001928232124099816, + "loss": 0.6364, + "step": 4961 + }, + { + "epoch": 0.15, + "learning_rate": 0.000192819624988469, + "loss": 0.5342, + "step": 4962 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001928160367039565, + "loss": 0.5863, + "step": 4963 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019281244755647752, + "loss": 0.514, + "step": 4964 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019280885754606537, + "loss": 0.6498, + "step": 4965 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019280526667275343, + "loss": 0.5786, + "step": 4966 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019280167493657512, + "loss": 0.6514, + "step": 4967 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019279808233756387, + "loss": 0.6474, + "step": 4968 + }, + { + "epoch": 0.15, + "learning_rate": 0.000192794488875753, + "loss": 0.4906, + "step": 4969 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019279089455117595, + "loss": 0.5639, + "step": 4970 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019278729936386614, + "loss": 0.5073, + "step": 4971 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019278370331385702, + "loss": 0.5262, + "step": 4972 + }, + { + "epoch": 0.15, + "learning_rate": 0.000192780106401182, + "loss": 0.5085, + "step": 4973 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019277650862587454, + "loss": 0.4588, + "step": 4974 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019277290998796808, + "loss": 0.7017, + "step": 4975 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019276931048749603, + "loss": 0.6542, + "step": 4976 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019276571012449197, + "loss": 0.4986, + "step": 4977 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001927621088989893, + "loss": 0.4459, + "step": 4978 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001927585068110215, + "loss": 0.5752, + "step": 4979 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019275490386062204, + "loss": 0.6047, + "step": 4980 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001927513000478245, + "loss": 0.4995, + "step": 4981 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019274769537266234, + "loss": 0.627, + "step": 4982 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019274408983516902, + "loss": 0.5593, + "step": 4983 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019274048343537818, + "loss": 0.5589, + "step": 4984 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019273687617332326, + "loss": 0.5444, + "step": 4985 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019273326804903783, + "loss": 0.5898, + "step": 4986 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019272965906255542, + "loss": 0.6049, + "step": 4987 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001927260492139096, + "loss": 0.6004, + "step": 4988 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019272243850313397, + "loss": 0.5524, + "step": 4989 + }, + { + "epoch": 0.15, + "learning_rate": 0.000192718826930262, + "loss": 0.6039, + "step": 4990 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019271521449532738, + "loss": 0.5304, + "step": 4991 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019271160119836364, + "loss": 0.6275, + "step": 4992 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001927079870394044, + "loss": 0.5044, + "step": 4993 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019270437201848322, + "loss": 0.6306, + "step": 4994 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019270075613563373, + "loss": 0.6363, + "step": 4995 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019269713939088957, + "loss": 0.5533, + "step": 4996 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019269352178428434, + "loss": 0.5502, + "step": 4997 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001926899033158517, + "loss": 0.5694, + "step": 4998 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019268628398562526, + "loss": 0.6017, + "step": 4999 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019268266379363872, + "loss": 0.6022, + "step": 5000 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001926790427399257, + "loss": 0.528, + "step": 5001 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019267542082451988, + "loss": 0.5223, + "step": 5002 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019267179804745497, + "loss": 0.5511, + "step": 5003 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019266817440876456, + "loss": 0.5837, + "step": 5004 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019266454990848244, + "loss": 0.5713, + "step": 5005 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019266092454664225, + "loss": 0.6295, + "step": 5006 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001926572983232777, + "loss": 0.5918, + "step": 5007 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019265367123842252, + "loss": 0.6049, + "step": 5008 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019265004329211048, + "loss": 0.6312, + "step": 5009 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001926464144843752, + "loss": 0.5433, + "step": 5010 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001926427848152505, + "loss": 0.6461, + "step": 5011 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019263915428477016, + "loss": 0.5455, + "step": 5012 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019263552289296786, + "loss": 0.5979, + "step": 5013 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019263189063987735, + "loss": 0.6273, + "step": 5014 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019262825752553247, + "loss": 0.5645, + "step": 5015 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019262462354996697, + "loss": 0.64, + "step": 5016 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019262098871321466, + "loss": 0.6011, + "step": 5017 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019261735301530927, + "loss": 0.4858, + "step": 5018 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001926137164562847, + "loss": 0.5449, + "step": 5019 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019261007903617462, + "loss": 0.527, + "step": 5020 + }, + { + "epoch": 0.15, + "learning_rate": 0.000192606440755013, + "loss": 0.5981, + "step": 5021 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001926028016128336, + "loss": 0.5425, + "step": 5022 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019259916160967023, + "loss": 0.5601, + "step": 5023 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001925955207455568, + "loss": 0.588, + "step": 5024 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001925918790205271, + "loss": 0.616, + "step": 5025 + }, + { + "epoch": 0.15, + "learning_rate": 0.000192588236434615, + "loss": 0.6436, + "step": 5026 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001925845929878544, + "loss": 0.7188, + "step": 5027 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019258094868027917, + "loss": 0.5264, + "step": 5028 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019257730351192312, + "loss": 0.6196, + "step": 5029 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019257365748282025, + "loss": 0.494, + "step": 5030 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001925700105930044, + "loss": 0.6445, + "step": 5031 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019256636284250948, + "loss": 0.5451, + "step": 5032 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019256271423136943, + "loss": 0.5782, + "step": 5033 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019255906475961813, + "loss": 0.5677, + "step": 5034 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019255541442728954, + "loss": 0.5076, + "step": 5035 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001925517632344176, + "loss": 0.534, + "step": 5036 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019254811118103623, + "loss": 0.4782, + "step": 5037 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019254445826717947, + "loss": 0.5381, + "step": 5038 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019254080449288116, + "loss": 0.5579, + "step": 5039 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019253714985817535, + "loss": 0.6303, + "step": 5040 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019253349436309603, + "loss": 0.694, + "step": 5041 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019252983800767713, + "loss": 0.6393, + "step": 5042 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001925261807919527, + "loss": 0.4471, + "step": 5043 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001925225227159567, + "loss": 0.5631, + "step": 5044 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019251886377972315, + "loss": 0.5603, + "step": 5045 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019251520398328608, + "loss": 0.653, + "step": 5046 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019251154332667955, + "loss": 0.6693, + "step": 5047 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001925078818099375, + "loss": 0.5576, + "step": 5048 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001925042194330941, + "loss": 0.4805, + "step": 5049 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001925005561961833, + "loss": 0.5319, + "step": 5050 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001924968920992392, + "loss": 0.5714, + "step": 5051 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019249322714229587, + "loss": 0.5513, + "step": 5052 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019248956132538738, + "loss": 0.563, + "step": 5053 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019248589464854781, + "loss": 0.5501, + "step": 5054 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019248222711181124, + "loss": 0.6022, + "step": 5055 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019247855871521181, + "loss": 0.5917, + "step": 5056 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019247488945878355, + "loss": 0.5196, + "step": 5057 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019247121934256068, + "loss": 0.6102, + "step": 5058 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019246754836657726, + "loss": 0.5914, + "step": 5059 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001924638765308674, + "loss": 0.59, + "step": 5060 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019246020383546528, + "loss": 0.5271, + "step": 5061 + }, + { + "epoch": 0.15, + "learning_rate": 0.000192456530280405, + "loss": 0.5599, + "step": 5062 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019245285586572082, + "loss": 0.4917, + "step": 5063 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001924491805914468, + "loss": 0.7159, + "step": 5064 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019244550445761716, + "loss": 0.4718, + "step": 5065 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019244182746426603, + "loss": 0.6469, + "step": 5066 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019243814961142767, + "loss": 0.5303, + "step": 5067 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019243447089913623, + "loss": 0.6907, + "step": 5068 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019243079132742592, + "loss": 0.7265, + "step": 5069 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019242711089633092, + "loss": 0.7013, + "step": 5070 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001924234296058855, + "loss": 0.6327, + "step": 5071 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019241974745612387, + "loss": 0.6976, + "step": 5072 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001924160644470803, + "loss": 0.7235, + "step": 5073 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019241238057878896, + "loss": 0.5171, + "step": 5074 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019240869585128412, + "loss": 0.4608, + "step": 5075 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019240501026460008, + "loss": 0.635, + "step": 5076 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001924013238187711, + "loss": 0.5243, + "step": 5077 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001923976365138314, + "loss": 0.5845, + "step": 5078 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019239394834981533, + "loss": 0.5743, + "step": 5079 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019239025932675714, + "loss": 0.5586, + "step": 5080 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019238656944469115, + "loss": 0.5631, + "step": 5081 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019238287870365165, + "loss": 0.5336, + "step": 5082 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019237918710367295, + "loss": 0.5363, + "step": 5083 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001923754946447894, + "loss": 0.6324, + "step": 5084 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019237180132703534, + "loss": 0.5981, + "step": 5085 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019236810715044506, + "loss": 0.5738, + "step": 5086 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019236441211505296, + "loss": 0.527, + "step": 5087 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019236071622089334, + "loss": 0.5205, + "step": 5088 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001923570194680006, + "loss": 0.6503, + "step": 5089 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001923533218564091, + "loss": 0.5204, + "step": 5090 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019234962338615324, + "loss": 0.611, + "step": 5091 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019234592405726738, + "loss": 0.6019, + "step": 5092 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019234222386978593, + "loss": 0.4889, + "step": 5093 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019233852282374327, + "loss": 0.5927, + "step": 5094 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019233482091917384, + "loss": 0.5167, + "step": 5095 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019233111815611205, + "loss": 0.5909, + "step": 5096 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001923274145345923, + "loss": 0.6066, + "step": 5097 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019232371005464905, + "loss": 0.4869, + "step": 5098 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019232000471631678, + "loss": 0.6257, + "step": 5099 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019231629851962985, + "loss": 0.477, + "step": 5100 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001923125914646228, + "loss": 0.6595, + "step": 5101 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019230888355133006, + "loss": 0.5471, + "step": 5102 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001923051747797861, + "loss": 0.6155, + "step": 5103 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019230146515002546, + "loss": 0.5989, + "step": 5104 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019229775466208252, + "loss": 0.5302, + "step": 5105 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019229404331599188, + "loss": 0.4928, + "step": 5106 + }, + { + "epoch": 0.15, + "learning_rate": 0.000192290331111788, + "loss": 0.5386, + "step": 5107 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001922866180495054, + "loss": 0.4993, + "step": 5108 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019228290412917862, + "loss": 0.448, + "step": 5109 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019227918935084217, + "loss": 0.5413, + "step": 5110 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001922754737145306, + "loss": 0.5519, + "step": 5111 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019227175722027847, + "loss": 0.6126, + "step": 5112 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001922680398681203, + "loss": 0.5385, + "step": 5113 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019226432165809064, + "loss": 0.7098, + "step": 5114 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019226060259022414, + "loss": 0.6184, + "step": 5115 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001922568826645553, + "loss": 0.6216, + "step": 5116 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019225316188111876, + "loss": 0.5911, + "step": 5117 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019224944023994906, + "loss": 0.5065, + "step": 5118 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019224571774108086, + "loss": 0.5655, + "step": 5119 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019224199438454872, + "loss": 0.6102, + "step": 5120 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019223827017038727, + "loss": 0.5637, + "step": 5121 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019223454509863117, + "loss": 0.456, + "step": 5122 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019223081916931503, + "loss": 0.7053, + "step": 5123 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019222709238247347, + "loss": 0.5783, + "step": 5124 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019222336473814118, + "loss": 0.5467, + "step": 5125 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001922196362363528, + "loss": 0.529, + "step": 5126 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019221590687714302, + "loss": 0.562, + "step": 5127 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019221217666054647, + "loss": 0.559, + "step": 5128 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019220844558659784, + "loss": 0.5359, + "step": 5129 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019220471365533187, + "loss": 0.5916, + "step": 5130 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019220098086678318, + "loss": 0.6571, + "step": 5131 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019219724722098654, + "loss": 0.4867, + "step": 5132 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019219351271797664, + "loss": 0.435, + "step": 5133 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001921897773577882, + "loss": 0.6054, + "step": 5134 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019218604114045596, + "loss": 0.5424, + "step": 5135 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019218230406601463, + "loss": 0.4855, + "step": 5136 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019217856613449897, + "loss": 0.6187, + "step": 5137 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001921748273459438, + "loss": 0.5233, + "step": 5138 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019217108770038378, + "loss": 0.5637, + "step": 5139 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019216734719785372, + "loss": 0.6088, + "step": 5140 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001921636058383884, + "loss": 0.637, + "step": 5141 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019215986362202262, + "loss": 0.595, + "step": 5142 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019215612054879119, + "loss": 0.5445, + "step": 5143 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001921523766187288, + "loss": 0.6106, + "step": 5144 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019214863183187038, + "loss": 0.5419, + "step": 5145 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019214488618825076, + "loss": 0.5101, + "step": 5146 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019214113968790464, + "loss": 0.6321, + "step": 5147 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019213739233086695, + "loss": 0.5943, + "step": 5148 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001921336441171725, + "loss": 0.5367, + "step": 5149 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001921298950468562, + "loss": 0.57, + "step": 5150 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001921261451199528, + "loss": 0.5204, + "step": 5151 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019212239433649724, + "loss": 0.6288, + "step": 5152 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019211864269652436, + "loss": 0.5518, + "step": 5153 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019211489020006905, + "loss": 0.5623, + "step": 5154 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019211113684716624, + "loss": 0.5552, + "step": 5155 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019210738263785075, + "loss": 0.5198, + "step": 5156 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019210362757215754, + "loss": 0.5951, + "step": 5157 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019209987165012154, + "loss": 0.5121, + "step": 5158 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019209611487177762, + "loss": 0.6001, + "step": 5159 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019209235723716071, + "loss": 0.5114, + "step": 5160 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019208859874630577, + "loss": 0.4978, + "step": 5161 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019208483939924778, + "loss": 0.5698, + "step": 5162 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001920810791960216, + "loss": 0.5292, + "step": 5163 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001920773181366623, + "loss": 0.5393, + "step": 5164 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019207355622120477, + "loss": 0.4631, + "step": 5165 + }, + { + "epoch": 0.15, + "learning_rate": 0.000192069793449684, + "loss": 0.6132, + "step": 5166 + }, + { + "epoch": 0.15, + "learning_rate": 0.000192066029822135, + "loss": 0.57, + "step": 5167 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019206226533859276, + "loss": 0.5757, + "step": 5168 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019205849999909226, + "loss": 0.5134, + "step": 5169 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001920547338036685, + "loss": 0.6471, + "step": 5170 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019205096675235653, + "loss": 0.5246, + "step": 5171 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019204719884519136, + "loss": 0.566, + "step": 5172 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019204343008220802, + "loss": 0.629, + "step": 5173 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019203966046344156, + "loss": 0.6454, + "step": 5174 + }, + { + "epoch": 0.15, + "learning_rate": 0.000192035889988927, + "loss": 0.4902, + "step": 5175 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019203211865869942, + "loss": 0.71, + "step": 5176 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019202834647279388, + "loss": 0.5965, + "step": 5177 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019202457343124547, + "loss": 0.5514, + "step": 5178 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019202079953408924, + "loss": 0.7373, + "step": 5179 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019201702478136031, + "loss": 0.5573, + "step": 5180 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019201324917309374, + "loss": 0.6096, + "step": 5181 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019200947270932466, + "loss": 0.5823, + "step": 5182 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019200569539008818, + "loss": 0.5383, + "step": 5183 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001920019172154194, + "loss": 0.5615, + "step": 5184 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019199813818535346, + "loss": 0.6547, + "step": 5185 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001919943582999255, + "loss": 0.5457, + "step": 5186 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019199057755917067, + "loss": 0.581, + "step": 5187 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019198679596312413, + "loss": 0.5241, + "step": 5188 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019198301351182098, + "loss": 0.6336, + "step": 5189 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019197923020529648, + "loss": 0.5873, + "step": 5190 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019197544604358574, + "loss": 0.6464, + "step": 5191 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019197166102672393, + "loss": 0.4974, + "step": 5192 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001919678751547463, + "loss": 0.5499, + "step": 5193 + }, + { + "epoch": 0.15, + "learning_rate": 0.000191964088427688, + "loss": 0.5273, + "step": 5194 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019196030084558428, + "loss": 0.5175, + "step": 5195 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019195651240847035, + "loss": 0.6561, + "step": 5196 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001919527231163814, + "loss": 0.5354, + "step": 5197 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019194893296935268, + "loss": 0.5933, + "step": 5198 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001919451419674194, + "loss": 0.5211, + "step": 5199 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019194135011061687, + "loss": 0.5462, + "step": 5200 + }, + { + "epoch": 0.15, + "learning_rate": 0.0001919375573989803, + "loss": 0.5889, + "step": 5201 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019193376383254496, + "loss": 0.6258, + "step": 5202 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019192996941134614, + "loss": 0.5821, + "step": 5203 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019192617413541912, + "loss": 0.54, + "step": 5204 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019192237800479914, + "loss": 0.5489, + "step": 5205 + }, + { + "epoch": 0.15, + "learning_rate": 0.00019191858101952152, + "loss": 0.5581, + "step": 5206 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019191478317962158, + "loss": 0.6001, + "step": 5207 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001919109844851346, + "loss": 0.5309, + "step": 5208 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019190718493609596, + "loss": 0.4777, + "step": 5209 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001919033845325409, + "loss": 0.5781, + "step": 5210 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019189958327450483, + "loss": 0.591, + "step": 5211 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019189578116202307, + "loss": 0.5891, + "step": 5212 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019189197819513093, + "loss": 0.6193, + "step": 5213 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019188817437386385, + "loss": 0.5119, + "step": 5214 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001918843696982571, + "loss": 0.5097, + "step": 5215 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019188056416834614, + "loss": 0.6932, + "step": 5216 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019187675778416629, + "loss": 0.4681, + "step": 5217 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019187295054575297, + "loss": 0.3664, + "step": 5218 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019186914245314157, + "loss": 0.5487, + "step": 5219 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001918653335063675, + "loss": 0.7132, + "step": 5220 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019186152370546618, + "loss": 0.6013, + "step": 5221 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019185771305047303, + "loss": 0.5088, + "step": 5222 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019185390154142343, + "loss": 0.5973, + "step": 5223 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001918500891783529, + "loss": 0.6283, + "step": 5224 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019184627596129686, + "loss": 0.5725, + "step": 5225 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001918424618902907, + "loss": 0.6663, + "step": 5226 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019183864696537, + "loss": 0.5313, + "step": 5227 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019183483118657014, + "loss": 0.5564, + "step": 5228 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001918310145539266, + "loss": 0.5485, + "step": 5229 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001918271970674749, + "loss": 0.5109, + "step": 5230 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019182337872725053, + "loss": 0.5661, + "step": 5231 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019181955953328898, + "loss": 0.5726, + "step": 5232 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019181573948562577, + "loss": 0.5455, + "step": 5233 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001918119185842964, + "loss": 0.5366, + "step": 5234 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001918080968293364, + "loss": 0.5509, + "step": 5235 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001918042742207813, + "loss": 0.6704, + "step": 5236 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019180045075866664, + "loss": 0.7458, + "step": 5237 + }, + { + "epoch": 0.16, + "learning_rate": 0.000191796626443028, + "loss": 0.6371, + "step": 5238 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001917928012739009, + "loss": 0.6968, + "step": 5239 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019178897525132094, + "loss": 0.5717, + "step": 5240 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019178514837532365, + "loss": 0.4855, + "step": 5241 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019178132064594464, + "loss": 0.5965, + "step": 5242 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019177749206321948, + "loss": 0.5049, + "step": 5243 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019177366262718376, + "loss": 0.592, + "step": 5244 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019176983233787316, + "loss": 0.4741, + "step": 5245 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001917660011953232, + "loss": 0.6184, + "step": 5246 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019176216919956953, + "loss": 0.6478, + "step": 5247 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001917583363506478, + "loss": 0.625, + "step": 5248 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019175450264859363, + "loss": 0.6354, + "step": 5249 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019175066809344265, + "loss": 0.5491, + "step": 5250 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019174683268523052, + "loss": 0.4732, + "step": 5251 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001917429964239929, + "loss": 0.5997, + "step": 5252 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019173915930976548, + "loss": 0.5035, + "step": 5253 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019173532134258395, + "loss": 0.6247, + "step": 5254 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019173148252248393, + "loss": 0.4911, + "step": 5255 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019172764284950116, + "loss": 0.6058, + "step": 5256 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019172380232367132, + "loss": 0.5685, + "step": 5257 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019171996094503012, + "loss": 0.6276, + "step": 5258 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019171611871361327, + "loss": 0.5968, + "step": 5259 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019171227562945654, + "loss": 0.5228, + "step": 5260 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019170843169259559, + "loss": 0.6129, + "step": 5261 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001917045869030662, + "loss": 0.5623, + "step": 5262 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001917007412609041, + "loss": 0.548, + "step": 5263 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019169689476614506, + "loss": 0.5982, + "step": 5264 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019169304741882486, + "loss": 0.5745, + "step": 5265 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019168919921897925, + "loss": 0.5278, + "step": 5266 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019168535016664398, + "loss": 0.6005, + "step": 5267 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001916815002618549, + "loss": 0.6231, + "step": 5268 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019167764950464775, + "loss": 0.6002, + "step": 5269 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019167379789505838, + "loss": 0.5178, + "step": 5270 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019166994543312252, + "loss": 0.5379, + "step": 5271 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001916660921188761, + "loss": 0.6561, + "step": 5272 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019166223795235484, + "loss": 0.4862, + "step": 5273 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019165838293359468, + "loss": 0.5545, + "step": 5274 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019165452706263136, + "loss": 0.543, + "step": 5275 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001916506703395008, + "loss": 0.5769, + "step": 5276 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019164681276423885, + "loss": 0.5493, + "step": 5277 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019164295433688134, + "loss": 0.5657, + "step": 5278 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019163909505746416, + "loss": 0.6348, + "step": 5279 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001916352349260232, + "loss": 0.5836, + "step": 5280 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019163137394259436, + "loss": 0.6488, + "step": 5281 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019162751210721353, + "loss": 0.5802, + "step": 5282 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019162364941991658, + "loss": 0.494, + "step": 5283 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001916197858807395, + "loss": 0.5675, + "step": 5284 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019161592148971815, + "loss": 0.5635, + "step": 5285 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019161205624688846, + "loss": 0.6417, + "step": 5286 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001916081901522864, + "loss": 0.6006, + "step": 5287 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001916043232059479, + "loss": 0.5613, + "step": 5288 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001916004554079089, + "loss": 0.5597, + "step": 5289 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019159658675820538, + "loss": 0.547, + "step": 5290 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001915927172568733, + "loss": 0.564, + "step": 5291 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019158884690394863, + "loss": 0.5513, + "step": 5292 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001915849756994674, + "loss": 0.6053, + "step": 5293 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019158110364346552, + "loss": 0.5158, + "step": 5294 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019157723073597908, + "loss": 0.6199, + "step": 5295 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019157335697704404, + "loss": 0.4869, + "step": 5296 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019156948236669644, + "loss": 0.5732, + "step": 5297 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019156560690497223, + "loss": 0.5322, + "step": 5298 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019156173059190756, + "loss": 0.5266, + "step": 5299 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019155785342753838, + "loss": 0.5512, + "step": 5300 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001915539754119008, + "loss": 0.4965, + "step": 5301 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001915500965450308, + "loss": 0.5489, + "step": 5302 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019154621682696452, + "loss": 0.5932, + "step": 5303 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019154233625773802, + "loss": 0.5846, + "step": 5304 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019153845483738734, + "loss": 0.4663, + "step": 5305 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019153457256594858, + "loss": 0.67, + "step": 5306 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019153068944345787, + "loss": 0.5378, + "step": 5307 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019152680546995128, + "loss": 0.5877, + "step": 5308 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019152292064546493, + "loss": 0.5668, + "step": 5309 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019151903497003495, + "loss": 0.5919, + "step": 5310 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019151514844369745, + "loss": 0.6933, + "step": 5311 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019151126106648858, + "loss": 0.5106, + "step": 5312 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019150737283844447, + "loss": 0.5502, + "step": 5313 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019150348375960127, + "loss": 0.5374, + "step": 5314 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001914995938299952, + "loss": 0.4434, + "step": 5315 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001914957030496623, + "loss": 0.5181, + "step": 5316 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019149181141863886, + "loss": 0.61, + "step": 5317 + }, + { + "epoch": 0.16, + "learning_rate": 0.000191487918936961, + "loss": 0.4896, + "step": 5318 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019148402560466497, + "loss": 0.6359, + "step": 5319 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001914801314217869, + "loss": 0.6087, + "step": 5320 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019147623638836306, + "loss": 0.5875, + "step": 5321 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001914723405044296, + "loss": 0.5879, + "step": 5322 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001914684437700228, + "loss": 0.6187, + "step": 5323 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019146454618517884, + "loss": 0.5135, + "step": 5324 + }, + { + "epoch": 0.16, + "learning_rate": 0.000191460647749934, + "loss": 0.5504, + "step": 5325 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001914567484643245, + "loss": 0.6094, + "step": 5326 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019145284832838662, + "loss": 0.6255, + "step": 5327 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019144894734215659, + "loss": 0.5762, + "step": 5328 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001914450455056707, + "loss": 0.5844, + "step": 5329 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019144114281896524, + "loss": 0.5888, + "step": 5330 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019143723928207647, + "loss": 0.5718, + "step": 5331 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019143333489504068, + "loss": 0.6326, + "step": 5332 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019142942965789418, + "loss": 0.5007, + "step": 5333 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001914255235706733, + "loss": 0.674, + "step": 5334 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019142161663341433, + "loss": 0.5982, + "step": 5335 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001914177088461536, + "loss": 0.5116, + "step": 5336 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019141380020892747, + "loss": 0.6025, + "step": 5337 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019140989072177224, + "loss": 0.5449, + "step": 5338 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001914059803847243, + "loss": 0.5608, + "step": 5339 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019140206919781997, + "loss": 0.5564, + "step": 5340 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019139815716109565, + "loss": 0.4989, + "step": 5341 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019139424427458765, + "loss": 0.5799, + "step": 5342 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019139033053833243, + "loss": 0.5156, + "step": 5343 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019138641595236631, + "loss": 0.5808, + "step": 5344 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019138250051672575, + "loss": 0.5125, + "step": 5345 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001913785842314471, + "loss": 0.4718, + "step": 5346 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019137466709656683, + "loss": 0.4319, + "step": 5347 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019137074911212127, + "loss": 0.5327, + "step": 5348 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019136683027814693, + "loss": 0.5851, + "step": 5349 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019136291059468024, + "loss": 0.5489, + "step": 5350 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019135899006175757, + "loss": 0.546, + "step": 5351 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019135506867941544, + "loss": 0.5101, + "step": 5352 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001913511464476903, + "loss": 0.6279, + "step": 5353 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001913472233666186, + "loss": 0.4953, + "step": 5354 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019134329943623686, + "loss": 0.5497, + "step": 5355 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019133937465658148, + "loss": 0.6008, + "step": 5356 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019133544902768902, + "loss": 0.5692, + "step": 5357 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019133152254959598, + "loss": 0.5774, + "step": 5358 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019132759522233882, + "loss": 0.6021, + "step": 5359 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001913236670459541, + "loss": 0.6302, + "step": 5360 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019131973802047827, + "loss": 0.5748, + "step": 5361 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019131580814594793, + "loss": 0.5143, + "step": 5362 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019131187742239965, + "loss": 0.5517, + "step": 5363 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019130794584986988, + "loss": 0.5074, + "step": 5364 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019130401342839525, + "loss": 0.4268, + "step": 5365 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019130008015801227, + "loss": 0.6583, + "step": 5366 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019129614603875754, + "loss": 0.5967, + "step": 5367 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019129221107066764, + "loss": 0.5547, + "step": 5368 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019128827525377912, + "loss": 0.6559, + "step": 5369 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019128433858812862, + "loss": 0.5941, + "step": 5370 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019128040107375274, + "loss": 0.6777, + "step": 5371 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019127646271068803, + "loss": 0.5898, + "step": 5372 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019127252349897118, + "loss": 0.5893, + "step": 5373 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001912685834386388, + "loss": 0.5605, + "step": 5374 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019126464252972747, + "loss": 0.6268, + "step": 5375 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001912607007722739, + "loss": 0.5467, + "step": 5376 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019125675816631467, + "loss": 0.5936, + "step": 5377 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001912528147118865, + "loss": 0.612, + "step": 5378 + }, + { + "epoch": 0.16, + "learning_rate": 0.000191248870409026, + "loss": 0.5079, + "step": 5379 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001912449252577699, + "loss": 0.6186, + "step": 5380 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019124097925815486, + "loss": 0.4849, + "step": 5381 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019123703241021752, + "loss": 0.4842, + "step": 5382 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019123308471399465, + "loss": 0.5732, + "step": 5383 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001912291361695229, + "loss": 0.5574, + "step": 5384 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019122518677683903, + "loss": 0.6389, + "step": 5385 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001912212365359797, + "loss": 0.4915, + "step": 5386 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001912172854469817, + "loss": 0.4539, + "step": 5387 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001912133335098817, + "loss": 0.586, + "step": 5388 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019120938072471653, + "loss": 0.5862, + "step": 5389 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019120542709152286, + "loss": 0.5693, + "step": 5390 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019120147261033745, + "loss": 0.6138, + "step": 5391 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019119751728119714, + "loss": 0.5394, + "step": 5392 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019119356110413867, + "loss": 0.5868, + "step": 5393 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019118960407919877, + "loss": 0.5028, + "step": 5394 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001911856462064143, + "loss": 0.4485, + "step": 5395 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019118168748582203, + "loss": 0.6167, + "step": 5396 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001911777279174588, + "loss": 0.5688, + "step": 5397 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019117376750136137, + "loss": 0.6581, + "step": 5398 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001911698062375666, + "loss": 0.6039, + "step": 5399 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019116584412611127, + "loss": 0.5609, + "step": 5400 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001911618811670323, + "loss": 0.6029, + "step": 5401 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019115791736036645, + "loss": 0.5946, + "step": 5402 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019115395270615065, + "loss": 0.6419, + "step": 5403 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019114998720442175, + "loss": 0.5936, + "step": 5404 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019114602085521656, + "loss": 0.4828, + "step": 5405 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019114205365857202, + "loss": 0.5178, + "step": 5406 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019113808561452495, + "loss": 0.5596, + "step": 5407 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001911341167231123, + "loss": 0.5221, + "step": 5408 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019113014698437099, + "loss": 0.5891, + "step": 5409 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019112617639833784, + "loss": 0.582, + "step": 5410 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019112220496504985, + "loss": 0.4551, + "step": 5411 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019111823268454394, + "loss": 0.5555, + "step": 5412 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019111425955685698, + "loss": 0.493, + "step": 5413 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019111028558202597, + "loss": 0.5505, + "step": 5414 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019110631076008782, + "loss": 0.6206, + "step": 5415 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019110233509107952, + "loss": 0.7091, + "step": 5416 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019109835857503802, + "loss": 0.6677, + "step": 5417 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019109438121200028, + "loss": 0.6701, + "step": 5418 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019109040300200326, + "loss": 0.6237, + "step": 5419 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019108642394508401, + "loss": 0.6427, + "step": 5420 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019108244404127952, + "loss": 0.6536, + "step": 5421 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019107846329062674, + "loss": 0.5339, + "step": 5422 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019107448169316271, + "loss": 0.6017, + "step": 5423 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019107049924892445, + "loss": 0.6236, + "step": 5424 + }, + { + "epoch": 0.16, + "learning_rate": 0.000191066515957949, + "loss": 0.5569, + "step": 5425 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019106253182027335, + "loss": 0.5909, + "step": 5426 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019105854683593458, + "loss": 0.4834, + "step": 5427 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019105456100496974, + "loss": 0.4931, + "step": 5428 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019105057432741588, + "loss": 0.5176, + "step": 5429 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001910465868033101, + "loss": 0.6463, + "step": 5430 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019104259843268938, + "loss": 0.6345, + "step": 5431 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001910386092155909, + "loss": 0.5869, + "step": 5432 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019103461915205173, + "loss": 0.502, + "step": 5433 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019103062824210896, + "loss": 0.5998, + "step": 5434 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019102663648579965, + "loss": 0.6334, + "step": 5435 + }, + { + "epoch": 0.16, + "learning_rate": 0.000191022643883161, + "loss": 0.5687, + "step": 5436 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019101865043423004, + "loss": 0.6648, + "step": 5437 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019101465613904396, + "loss": 0.5805, + "step": 5438 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001910106609976399, + "loss": 0.6762, + "step": 5439 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019100666501005497, + "loss": 0.6455, + "step": 5440 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019100266817632635, + "loss": 0.6022, + "step": 5441 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019099867049649118, + "loss": 0.5347, + "step": 5442 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019099467197058664, + "loss": 0.6466, + "step": 5443 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019099067259864994, + "loss": 0.513, + "step": 5444 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001909866723807182, + "loss": 0.5618, + "step": 5445 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019098267131682862, + "loss": 0.4733, + "step": 5446 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019097866940701846, + "loss": 0.6075, + "step": 5447 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001909746666513249, + "loss": 0.5778, + "step": 5448 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019097066304978513, + "loss": 0.524, + "step": 5449 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019096665860243638, + "loss": 0.5844, + "step": 5450 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019096265330931594, + "loss": 0.6858, + "step": 5451 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019095864717046096, + "loss": 0.6279, + "step": 5452 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019095464018590875, + "loss": 0.6355, + "step": 5453 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001909506323556965, + "loss": 0.543, + "step": 5454 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001909466236798616, + "loss": 0.4337, + "step": 5455 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001909426141584412, + "loss": 0.547, + "step": 5456 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001909386037914726, + "loss": 0.5986, + "step": 5457 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019093459257899314, + "loss": 0.5092, + "step": 5458 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019093058052104008, + "loss": 0.5219, + "step": 5459 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001909265676176507, + "loss": 0.5929, + "step": 5460 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019092255386886234, + "loss": 0.5828, + "step": 5461 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019091853927471233, + "loss": 0.5137, + "step": 5462 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019091452383523797, + "loss": 0.5985, + "step": 5463 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019091050755047658, + "loss": 0.6572, + "step": 5464 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019090649042046557, + "loss": 0.5527, + "step": 5465 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019090247244524223, + "loss": 0.4674, + "step": 5466 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001908984536248439, + "loss": 0.6424, + "step": 5467 + }, + { + "epoch": 0.16, + "learning_rate": 0.000190894433959308, + "loss": 0.6324, + "step": 5468 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001908904134486719, + "loss": 0.469, + "step": 5469 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019088639209297292, + "loss": 0.6773, + "step": 5470 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019088236989224847, + "loss": 0.6229, + "step": 5471 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019087834684653601, + "loss": 0.5512, + "step": 5472 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019087432295587293, + "loss": 0.6377, + "step": 5473 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019087029822029657, + "loss": 0.5537, + "step": 5474 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001908662726398444, + "loss": 0.6383, + "step": 5475 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019086224621455385, + "loss": 0.6046, + "step": 5476 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019085821894446235, + "loss": 0.6146, + "step": 5477 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001908541908296073, + "loss": 0.5772, + "step": 5478 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019085016187002626, + "loss": 0.5445, + "step": 5479 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001908461320657566, + "loss": 0.6691, + "step": 5480 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001908421014168358, + "loss": 0.5697, + "step": 5481 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019083806992330137, + "loss": 0.7055, + "step": 5482 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019083403758519073, + "loss": 0.5661, + "step": 5483 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019083000440254147, + "loss": 0.5733, + "step": 5484 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019082597037539098, + "loss": 0.4644, + "step": 5485 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019082193550377686, + "loss": 0.5698, + "step": 5486 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019081789978773653, + "loss": 0.5102, + "step": 5487 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019081386322730758, + "loss": 0.5773, + "step": 5488 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019080982582252756, + "loss": 0.5893, + "step": 5489 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019080578757343393, + "loss": 0.5486, + "step": 5490 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019080174848006432, + "loss": 0.7169, + "step": 5491 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001907977085424562, + "loss": 0.5367, + "step": 5492 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001907936677606472, + "loss": 0.5104, + "step": 5493 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019078962613467486, + "loss": 0.5803, + "step": 5494 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019078558366457675, + "loss": 0.5187, + "step": 5495 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001907815403503905, + "loss": 0.6483, + "step": 5496 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019077749619215363, + "loss": 0.5667, + "step": 5497 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001907734511899038, + "loss": 0.4697, + "step": 5498 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001907694053436786, + "loss": 0.5799, + "step": 5499 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019076535865351563, + "loss": 0.5216, + "step": 5500 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019076131111945256, + "loss": 0.568, + "step": 5501 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019075726274152697, + "loss": 0.5627, + "step": 5502 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019075321351977653, + "loss": 0.5192, + "step": 5503 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019074916345423887, + "loss": 0.6265, + "step": 5504 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019074511254495166, + "loss": 0.574, + "step": 5505 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001907410607919526, + "loss": 0.6488, + "step": 5506 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001907370081952793, + "loss": 0.4954, + "step": 5507 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019073295475496944, + "loss": 0.5588, + "step": 5508 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019072890047106073, + "loss": 0.5725, + "step": 5509 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001907248453435909, + "loss": 0.5064, + "step": 5510 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019072078937259755, + "loss": 0.5478, + "step": 5511 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019071673255811853, + "loss": 0.6349, + "step": 5512 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019071267490019148, + "loss": 0.5995, + "step": 5513 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001907086163988541, + "loss": 0.5442, + "step": 5514 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019070455705414417, + "loss": 0.6582, + "step": 5515 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001907004968660994, + "loss": 0.5275, + "step": 5516 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001906964358347576, + "loss": 0.5591, + "step": 5517 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019069237396015644, + "loss": 0.5996, + "step": 5518 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001906883112423338, + "loss": 0.4582, + "step": 5519 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019068424768132734, + "loss": 0.522, + "step": 5520 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019068018327717486, + "loss": 0.6119, + "step": 5521 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019067611802991424, + "loss": 0.6693, + "step": 5522 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019067205193958316, + "loss": 0.5047, + "step": 5523 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019066798500621954, + "loss": 0.5068, + "step": 5524 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001906639172298611, + "loss": 0.6455, + "step": 5525 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001906598486105457, + "loss": 0.5148, + "step": 5526 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019065577914831113, + "loss": 0.5998, + "step": 5527 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001906517088431953, + "loss": 0.5303, + "step": 5528 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019064763769523598, + "loss": 0.5541, + "step": 5529 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019064356570447108, + "loss": 0.5096, + "step": 5530 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019063949287093841, + "loss": 0.4796, + "step": 5531 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019063541919467586, + "loss": 0.7103, + "step": 5532 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001906313446757213, + "loss": 0.5919, + "step": 5533 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019062726931411267, + "loss": 0.6226, + "step": 5534 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019062319310988776, + "loss": 0.4697, + "step": 5535 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019061911606308454, + "loss": 0.554, + "step": 5536 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019061503817374088, + "loss": 0.6564, + "step": 5537 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019061095944189474, + "loss": 0.6153, + "step": 5538 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019060687986758397, + "loss": 0.5787, + "step": 5539 + }, + { + "epoch": 0.16, + "learning_rate": 0.0001906027994508466, + "loss": 0.6193, + "step": 5540 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019059871819172047, + "loss": 0.5182, + "step": 5541 + }, + { + "epoch": 0.16, + "learning_rate": 0.00019059463609024356, + "loss": 0.4717, + "step": 5542 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019059055314645386, + "loss": 0.6917, + "step": 5543 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019058646936038928, + "loss": 0.462, + "step": 5544 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019058238473208783, + "loss": 0.5748, + "step": 5545 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019057829926158744, + "loss": 0.5754, + "step": 5546 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019057421294892617, + "loss": 0.7042, + "step": 5547 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019057012579414194, + "loss": 0.6605, + "step": 5548 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019056603779727275, + "loss": 0.5931, + "step": 5549 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019056194895835667, + "loss": 0.4528, + "step": 5550 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001905578592774317, + "loss": 0.632, + "step": 5551 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019055376875453582, + "loss": 0.5356, + "step": 5552 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019054967738970708, + "loss": 0.5042, + "step": 5553 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019054558518298356, + "loss": 0.5059, + "step": 5554 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019054149213440326, + "loss": 0.5699, + "step": 5555 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019053739824400427, + "loss": 0.4937, + "step": 5556 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019053330351182461, + "loss": 0.6354, + "step": 5557 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019052920793790238, + "loss": 0.5823, + "step": 5558 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001905251115222757, + "loss": 0.5697, + "step": 5559 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019052101426498256, + "loss": 0.5659, + "step": 5560 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019051691616606112, + "loss": 0.6921, + "step": 5561 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001905128172255495, + "loss": 0.5784, + "step": 5562 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019050871744348574, + "loss": 0.5323, + "step": 5563 + }, + { + "epoch": 0.17, + "learning_rate": 0.000190504616819908, + "loss": 0.5213, + "step": 5564 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019050051535485444, + "loss": 0.672, + "step": 5565 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019049641304836313, + "loss": 0.632, + "step": 5566 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019049230990047223, + "loss": 0.5461, + "step": 5567 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019048820591121993, + "loss": 0.5682, + "step": 5568 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019048410108064433, + "loss": 0.5341, + "step": 5569 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019047999540878363, + "loss": 0.6366, + "step": 5570 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019047588889567601, + "loss": 0.5277, + "step": 5571 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001904717815413596, + "loss": 0.5875, + "step": 5572 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019046767334587265, + "loss": 0.5411, + "step": 5573 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019046356430925333, + "loss": 0.5162, + "step": 5574 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019045945443153984, + "loss": 0.4746, + "step": 5575 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001904553437127704, + "loss": 0.6114, + "step": 5576 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001904512321529832, + "loss": 0.6354, + "step": 5577 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019044711975221655, + "loss": 0.6368, + "step": 5578 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001904430065105086, + "loss": 0.5671, + "step": 5579 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001904388924278976, + "loss": 0.6085, + "step": 5580 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019043477750442183, + "loss": 0.5822, + "step": 5581 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019043066174011957, + "loss": 0.5605, + "step": 5582 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019042654513502902, + "loss": 0.5827, + "step": 5583 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019042242768918852, + "loss": 0.6109, + "step": 5584 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019041830940263633, + "loss": 0.55, + "step": 5585 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019041419027541074, + "loss": 0.5305, + "step": 5586 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019041007030755003, + "loss": 0.5663, + "step": 5587 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019040594949909254, + "loss": 0.5586, + "step": 5588 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019040182785007653, + "loss": 0.6248, + "step": 5589 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019039770536054038, + "loss": 0.5225, + "step": 5590 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001903935820305224, + "loss": 0.5948, + "step": 5591 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001903894578600609, + "loss": 0.4979, + "step": 5592 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019038533284919425, + "loss": 0.5329, + "step": 5593 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001903812069979608, + "loss": 0.4844, + "step": 5594 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019037708030639894, + "loss": 0.535, + "step": 5595 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019037295277454696, + "loss": 0.5498, + "step": 5596 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019036882440244332, + "loss": 0.5637, + "step": 5597 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019036469519012635, + "loss": 0.6205, + "step": 5598 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019036056513763448, + "loss": 0.5504, + "step": 5599 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019035643424500606, + "loss": 0.6667, + "step": 5600 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019035230251227955, + "loss": 0.5097, + "step": 5601 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019034816993949335, + "loss": 0.6201, + "step": 5602 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019034403652668584, + "loss": 0.5705, + "step": 5603 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001903399022738955, + "loss": 0.6262, + "step": 5604 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019033576718116075, + "loss": 0.4933, + "step": 5605 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019033163124852008, + "loss": 0.4724, + "step": 5606 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019032749447601185, + "loss": 0.5192, + "step": 5607 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001903233568636746, + "loss": 0.5162, + "step": 5608 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019031921841154682, + "loss": 0.4993, + "step": 5609 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019031507911966688, + "loss": 0.4738, + "step": 5610 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019031093898807334, + "loss": 0.6084, + "step": 5611 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001903067980168047, + "loss": 0.5917, + "step": 5612 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019030265620589946, + "loss": 0.5055, + "step": 5613 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019029851355539608, + "loss": 0.5314, + "step": 5614 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019029437006533315, + "loss": 0.5956, + "step": 5615 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019029022573574913, + "loss": 0.6359, + "step": 5616 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019028608056668257, + "loss": 0.6318, + "step": 5617 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019028193455817202, + "loss": 0.5004, + "step": 5618 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019027778771025603, + "loss": 0.5509, + "step": 5619 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019027364002297318, + "loss": 0.6345, + "step": 5620 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019026949149636196, + "loss": 0.4427, + "step": 5621 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019026534213046101, + "loss": 0.6122, + "step": 5622 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019026119192530887, + "loss": 0.5738, + "step": 5623 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019025704088094416, + "loss": 0.664, + "step": 5624 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019025288899740544, + "loss": 0.5408, + "step": 5625 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019024873627473135, + "loss": 0.5614, + "step": 5626 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019024458271296046, + "loss": 0.5427, + "step": 5627 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019024042831213142, + "loss": 0.5747, + "step": 5628 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019023627307228284, + "loss": 0.5888, + "step": 5629 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019023211699345337, + "loss": 0.5599, + "step": 5630 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019022796007568161, + "loss": 0.4008, + "step": 5631 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019022380231900626, + "loss": 0.4662, + "step": 5632 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019021964372346596, + "loss": 0.5319, + "step": 5633 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019021548428909934, + "loss": 0.591, + "step": 5634 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019021132401594515, + "loss": 0.6088, + "step": 5635 + }, + { + "epoch": 0.17, + "learning_rate": 0.000190207162904042, + "loss": 0.5972, + "step": 5636 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001902030009534286, + "loss": 0.6089, + "step": 5637 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019019883816414362, + "loss": 0.5309, + "step": 5638 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019019467453622584, + "loss": 0.5965, + "step": 5639 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019019051006971392, + "loss": 0.5235, + "step": 5640 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019018634476464653, + "loss": 0.3863, + "step": 5641 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019018217862106252, + "loss": 0.6062, + "step": 5642 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019017801163900052, + "loss": 0.5302, + "step": 5643 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001901738438184993, + "loss": 0.6121, + "step": 5644 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019016967515959765, + "loss": 0.5119, + "step": 5645 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019016550566233428, + "loss": 0.5606, + "step": 5646 + }, + { + "epoch": 0.17, + "learning_rate": 0.000190161335326748, + "loss": 0.5742, + "step": 5647 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019015716415287753, + "loss": 0.4636, + "step": 5648 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001901529921407617, + "loss": 0.5596, + "step": 5649 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019014881929043927, + "loss": 0.5943, + "step": 5650 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019014464560194902, + "loss": 0.5256, + "step": 5651 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019014047107532983, + "loss": 0.5125, + "step": 5652 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019013629571062044, + "loss": 0.6027, + "step": 5653 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019013211950785972, + "loss": 0.5399, + "step": 5654 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019012794246708645, + "loss": 0.5555, + "step": 5655 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001901237645883395, + "loss": 0.5623, + "step": 5656 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001901195858716577, + "loss": 0.5326, + "step": 5657 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019011540631707995, + "loss": 0.4466, + "step": 5658 + }, + { + "epoch": 0.17, + "learning_rate": 0.000190111225924645, + "loss": 0.6678, + "step": 5659 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001901070446943918, + "loss": 0.5618, + "step": 5660 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019010286262635928, + "loss": 0.7095, + "step": 5661 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001900986797205862, + "loss": 0.5952, + "step": 5662 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001900944959771115, + "loss": 0.4662, + "step": 5663 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019009031139597413, + "loss": 0.5314, + "step": 5664 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001900861259772129, + "loss": 0.4972, + "step": 5665 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001900819397208668, + "loss": 0.6259, + "step": 5666 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019007775262697474, + "loss": 0.5112, + "step": 5667 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001900735646955756, + "loss": 0.5883, + "step": 5668 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001900693759267084, + "loss": 0.4438, + "step": 5669 + }, + { + "epoch": 0.17, + "learning_rate": 0.000190065186320412, + "loss": 0.563, + "step": 5670 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019006099587672544, + "loss": 0.571, + "step": 5671 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001900568045956876, + "loss": 0.7452, + "step": 5672 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019005261247733747, + "loss": 0.6503, + "step": 5673 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019004841952171406, + "loss": 0.5844, + "step": 5674 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019004422572885634, + "loss": 0.4916, + "step": 5675 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019004003109880332, + "loss": 0.6381, + "step": 5676 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019003583563159392, + "loss": 0.6434, + "step": 5677 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019003163932726724, + "loss": 0.5902, + "step": 5678 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019002744218586226, + "loss": 0.5239, + "step": 5679 + }, + { + "epoch": 0.17, + "learning_rate": 0.000190023244207418, + "loss": 0.5384, + "step": 5680 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001900190453919735, + "loss": 0.6137, + "step": 5681 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019001484573956778, + "loss": 0.6162, + "step": 5682 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001900106452502399, + "loss": 0.5153, + "step": 5683 + }, + { + "epoch": 0.17, + "learning_rate": 0.00019000644392402894, + "loss": 0.577, + "step": 5684 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001900022417609739, + "loss": 0.5715, + "step": 5685 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018999803876111393, + "loss": 0.4993, + "step": 5686 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018999383492448804, + "loss": 0.5392, + "step": 5687 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018998963025113535, + "loss": 0.5624, + "step": 5688 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018998542474109496, + "loss": 0.5454, + "step": 5689 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001899812183944059, + "loss": 0.5236, + "step": 5690 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018997701121110737, + "loss": 0.5701, + "step": 5691 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018997280319123846, + "loss": 0.5463, + "step": 5692 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018996859433483828, + "loss": 0.5566, + "step": 5693 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018996438464194597, + "loss": 0.4968, + "step": 5694 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018996017411260064, + "loss": 0.5651, + "step": 5695 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001899559627468415, + "loss": 0.6824, + "step": 5696 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018995175054470765, + "loss": 0.4563, + "step": 5697 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001899475375062383, + "loss": 0.6218, + "step": 5698 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001899433236314726, + "loss": 0.5836, + "step": 5699 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018993910892044967, + "loss": 0.5412, + "step": 5700 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001899348933732088, + "loss": 0.5753, + "step": 5701 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018993067698978912, + "loss": 0.5469, + "step": 5702 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018992645977022985, + "loss": 0.6605, + "step": 5703 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001899222417145702, + "loss": 0.5762, + "step": 5704 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018991802282284937, + "loss": 0.4735, + "step": 5705 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018991380309510663, + "loss": 0.6065, + "step": 5706 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018990958253138117, + "loss": 0.6371, + "step": 5707 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018990536113171224, + "loss": 0.5603, + "step": 5708 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001899011388961391, + "loss": 0.5646, + "step": 5709 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018989691582470101, + "loss": 0.621, + "step": 5710 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001898926919174372, + "loss": 0.5225, + "step": 5711 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018988846717438702, + "loss": 0.4842, + "step": 5712 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018988424159558967, + "loss": 0.5726, + "step": 5713 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018988001518108444, + "loss": 0.5529, + "step": 5714 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018987578793091067, + "loss": 0.5565, + "step": 5715 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018987155984510765, + "loss": 0.6137, + "step": 5716 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001898673309237147, + "loss": 0.5181, + "step": 5717 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001898631011667711, + "loss": 0.5493, + "step": 5718 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018985887057431622, + "loss": 0.558, + "step": 5719 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018985463914638934, + "loss": 0.6355, + "step": 5720 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018985040688302988, + "loss": 0.6351, + "step": 5721 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018984617378427711, + "loss": 0.5036, + "step": 5722 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018984193985017048, + "loss": 0.5855, + "step": 5723 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018983770508074926, + "loss": 0.6084, + "step": 5724 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018983346947605285, + "loss": 0.6854, + "step": 5725 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001898292330361207, + "loss": 0.5951, + "step": 5726 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001898249957609921, + "loss": 0.586, + "step": 5727 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001898207576507065, + "loss": 0.5436, + "step": 5728 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018981651870530328, + "loss": 0.5199, + "step": 5729 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018981227892482188, + "loss": 0.5488, + "step": 5730 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018980803830930172, + "loss": 0.6653, + "step": 5731 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001898037968587822, + "loss": 0.5291, + "step": 5732 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018979955457330277, + "loss": 0.6239, + "step": 5733 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001897953114529029, + "loss": 0.6288, + "step": 5734 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018979106749762194, + "loss": 0.4782, + "step": 5735 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018978682270749949, + "loss": 0.5489, + "step": 5736 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018978257708257492, + "loss": 0.5655, + "step": 5737 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018977833062288772, + "loss": 0.5955, + "step": 5738 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001897740833284774, + "loss": 0.5111, + "step": 5739 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018976983519938343, + "loss": 0.5663, + "step": 5740 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018976558623564534, + "loss": 0.5503, + "step": 5741 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018976133643730257, + "loss": 0.4995, + "step": 5742 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018975708580439466, + "loss": 0.561, + "step": 5743 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018975283433696118, + "loss": 0.612, + "step": 5744 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018974858203504158, + "loss": 0.5754, + "step": 5745 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018974432889867546, + "loss": 0.5403, + "step": 5746 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018974007492790233, + "loss": 0.6365, + "step": 5747 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018973582012276173, + "loss": 0.5431, + "step": 5748 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018973156448329328, + "loss": 0.5811, + "step": 5749 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001897273080095365, + "loss": 0.5072, + "step": 5750 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018972305070153096, + "loss": 0.5213, + "step": 5751 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018971879255931623, + "loss": 0.5155, + "step": 5752 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018971453358293197, + "loss": 0.6883, + "step": 5753 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001897102737724177, + "loss": 0.5739, + "step": 5754 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001897060131278131, + "loss": 0.703, + "step": 5755 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001897017516491577, + "loss": 0.5613, + "step": 5756 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018969748933649115, + "loss": 0.6214, + "step": 5757 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018969322618985314, + "loss": 0.5399, + "step": 5758 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018968896220928328, + "loss": 0.4992, + "step": 5759 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018968469739482114, + "loss": 0.6045, + "step": 5760 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018968043174650645, + "loss": 0.6579, + "step": 5761 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018967616526437885, + "loss": 0.551, + "step": 5762 + }, + { + "epoch": 0.17, + "learning_rate": 0.000189671897948478, + "loss": 0.6479, + "step": 5763 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001896676297988436, + "loss": 0.5546, + "step": 5764 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001896633608155153, + "loss": 0.5541, + "step": 5765 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001896590909985328, + "loss": 0.5679, + "step": 5766 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018965482034793583, + "loss": 0.6522, + "step": 5767 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018965054886376406, + "loss": 0.477, + "step": 5768 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018964627654605723, + "loss": 0.5563, + "step": 5769 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018964200339485502, + "loss": 0.5707, + "step": 5770 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018963772941019721, + "loss": 0.5061, + "step": 5771 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018963345459212353, + "loss": 0.5762, + "step": 5772 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018962917894067372, + "loss": 0.4595, + "step": 5773 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001896249024558875, + "loss": 0.5414, + "step": 5774 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018962062513780466, + "loss": 0.5995, + "step": 5775 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018961634698646496, + "loss": 0.559, + "step": 5776 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001896120680019082, + "loss": 0.5981, + "step": 5777 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018960778818417415, + "loss": 0.5549, + "step": 5778 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018960350753330257, + "loss": 0.4765, + "step": 5779 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018959922604933334, + "loss": 0.7169, + "step": 5780 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018959494373230618, + "loss": 0.5839, + "step": 5781 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018959066058226096, + "loss": 0.4908, + "step": 5782 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018958637659923747, + "loss": 0.5751, + "step": 5783 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018958209178327558, + "loss": 0.6004, + "step": 5784 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018957780613441507, + "loss": 0.5085, + "step": 5785 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018957351965269583, + "loss": 0.5597, + "step": 5786 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018956923233815772, + "loss": 0.5452, + "step": 5787 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018956494419084056, + "loss": 0.5222, + "step": 5788 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018956065521078428, + "loss": 0.5524, + "step": 5789 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018955636539802871, + "loss": 0.5051, + "step": 5790 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018955207475261374, + "loss": 0.4255, + "step": 5791 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001895477832745793, + "loss": 0.6189, + "step": 5792 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018954349096396523, + "loss": 0.5739, + "step": 5793 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018953919782081148, + "loss": 0.6601, + "step": 5794 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018953490384515792, + "loss": 0.6134, + "step": 5795 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018953060903704455, + "loss": 0.5885, + "step": 5796 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018952631339651127, + "loss": 0.5068, + "step": 5797 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018952201692359799, + "loss": 0.6074, + "step": 5798 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018951771961834467, + "loss": 0.5224, + "step": 5799 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001895134214807913, + "loss": 0.6568, + "step": 5800 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001895091225109778, + "loss": 0.5096, + "step": 5801 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018950482270894414, + "loss": 0.6296, + "step": 5802 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018950052207473031, + "loss": 0.5719, + "step": 5803 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001894962206083763, + "loss": 0.4744, + "step": 5804 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018949191830992213, + "loss": 0.563, + "step": 5805 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018948761517940775, + "loss": 0.426, + "step": 5806 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018948331121687318, + "loss": 0.5369, + "step": 5807 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018947900642235846, + "loss": 0.5752, + "step": 5808 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001894747007959036, + "loss": 0.5901, + "step": 5809 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018947039433754863, + "loss": 0.5865, + "step": 5810 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018946608704733357, + "loss": 0.4889, + "step": 5811 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018946177892529854, + "loss": 0.5967, + "step": 5812 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001894574699714835, + "loss": 0.5741, + "step": 5813 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018945316018592859, + "loss": 0.4602, + "step": 5814 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018944884956867384, + "loss": 0.6134, + "step": 5815 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018944453811975932, + "loss": 0.5783, + "step": 5816 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018944022583922517, + "loss": 0.4877, + "step": 5817 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001894359127271114, + "loss": 0.5806, + "step": 5818 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001894315987834582, + "loss": 0.4792, + "step": 5819 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001894272840083056, + "loss": 0.5522, + "step": 5820 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001894229684016938, + "loss": 0.5665, + "step": 5821 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018941865196366283, + "loss": 0.4786, + "step": 5822 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018941433469425292, + "loss": 0.6428, + "step": 5823 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018941001659350415, + "loss": 0.5634, + "step": 5824 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018940569766145665, + "loss": 0.4859, + "step": 5825 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018940137789815066, + "loss": 0.6087, + "step": 5826 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018939705730362626, + "loss": 0.5221, + "step": 5827 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018939273587792365, + "loss": 0.5034, + "step": 5828 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018938841362108302, + "loss": 0.6544, + "step": 5829 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018938409053314455, + "loss": 0.5797, + "step": 5830 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001893797666141484, + "loss": 0.6334, + "step": 5831 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018937544186413484, + "loss": 0.476, + "step": 5832 + }, + { + "epoch": 0.17, + "learning_rate": 0.000189371116283144, + "loss": 0.633, + "step": 5833 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018936678987121618, + "loss": 0.668, + "step": 5834 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018936246262839156, + "loss": 0.6083, + "step": 5835 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018935813455471038, + "loss": 0.6106, + "step": 5836 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018935380565021287, + "loss": 0.56, + "step": 5837 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018934947591493927, + "loss": 0.4741, + "step": 5838 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018934514534892988, + "loss": 0.5524, + "step": 5839 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018934081395222493, + "loss": 0.5504, + "step": 5840 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018933648172486471, + "loss": 0.5227, + "step": 5841 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001893321486668895, + "loss": 0.5328, + "step": 5842 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018932781477833954, + "loss": 0.6154, + "step": 5843 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001893234800592552, + "loss": 0.5077, + "step": 5844 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001893191445096767, + "loss": 0.5975, + "step": 5845 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018931480812964444, + "loss": 0.4874, + "step": 5846 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018931047091919867, + "loss": 0.5435, + "step": 5847 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001893061328783797, + "loss": 0.6465, + "step": 5848 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018930179400722795, + "loss": 0.4977, + "step": 5849 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001892974543057837, + "loss": 0.6444, + "step": 5850 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018929311377408728, + "loss": 0.4583, + "step": 5851 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018928877241217913, + "loss": 0.6115, + "step": 5852 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018928443022009952, + "loss": 0.6498, + "step": 5853 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001892800871978889, + "loss": 0.5774, + "step": 5854 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018927574334558756, + "loss": 0.5281, + "step": 5855 + }, + { + "epoch": 0.17, + "learning_rate": 0.000189271398663236, + "loss": 0.6437, + "step": 5856 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001892670531508745, + "loss": 0.5589, + "step": 5857 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018926270680854357, + "loss": 0.664, + "step": 5858 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018925835963628355, + "loss": 0.5337, + "step": 5859 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018925401163413485, + "loss": 0.5421, + "step": 5860 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018924966280213794, + "loss": 0.5309, + "step": 5861 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018924531314033324, + "loss": 0.6753, + "step": 5862 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018924096264876118, + "loss": 0.5637, + "step": 5863 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018923661132746222, + "loss": 0.6813, + "step": 5864 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001892322591764768, + "loss": 0.5878, + "step": 5865 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018922790619584542, + "loss": 0.5181, + "step": 5866 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018922355238560856, + "loss": 0.5368, + "step": 5867 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018921919774580662, + "loss": 0.5374, + "step": 5868 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018921484227648016, + "loss": 0.6629, + "step": 5869 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018921048597766963, + "loss": 0.5454, + "step": 5870 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001892061288494156, + "loss": 0.5062, + "step": 5871 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018920177089175853, + "loss": 0.6696, + "step": 5872 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018919741210473892, + "loss": 0.4941, + "step": 5873 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018919305248839733, + "loss": 0.5768, + "step": 5874 + }, + { + "epoch": 0.17, + "learning_rate": 0.0001891886920427743, + "loss": 0.4768, + "step": 5875 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018918433076791033, + "loss": 0.4817, + "step": 5876 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018917996866384603, + "loss": 0.601, + "step": 5877 + }, + { + "epoch": 0.17, + "learning_rate": 0.00018917560573062191, + "loss": 0.6329, + "step": 5878 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018917124196827855, + "loss": 0.49, + "step": 5879 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018916687737685655, + "loss": 0.5978, + "step": 5880 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018916251195639643, + "loss": 0.585, + "step": 5881 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018915814570693882, + "loss": 0.6217, + "step": 5882 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001891537786285243, + "loss": 0.4818, + "step": 5883 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018914941072119347, + "loss": 0.5368, + "step": 5884 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018914504198498699, + "loss": 0.4534, + "step": 5885 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018914067241994543, + "loss": 0.5607, + "step": 5886 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001891363020261094, + "loss": 0.6144, + "step": 5887 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018913193080351956, + "loss": 0.5473, + "step": 5888 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018912755875221654, + "loss": 0.642, + "step": 5889 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018912318587224105, + "loss": 0.6311, + "step": 5890 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018911881216363368, + "loss": 0.5749, + "step": 5891 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001891144376264351, + "loss": 0.5081, + "step": 5892 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018911006226068598, + "loss": 0.5813, + "step": 5893 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018910568606642704, + "loss": 0.5449, + "step": 5894 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018910130904369895, + "loss": 0.6183, + "step": 5895 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018909693119254238, + "loss": 0.5166, + "step": 5896 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001890925525129981, + "loss": 0.6962, + "step": 5897 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018908817300510676, + "loss": 0.4487, + "step": 5898 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018908379266890907, + "loss": 0.4931, + "step": 5899 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018907941150444576, + "loss": 0.4828, + "step": 5900 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018907502951175765, + "loss": 0.5666, + "step": 5901 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018907064669088536, + "loss": 0.5226, + "step": 5902 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018906626304186972, + "loss": 0.5642, + "step": 5903 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018906187856475147, + "loss": 0.5973, + "step": 5904 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018905749325957136, + "loss": 0.5822, + "step": 5905 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001890531071263702, + "loss": 0.6326, + "step": 5906 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001890487201651887, + "loss": 0.5741, + "step": 5907 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018904433237606772, + "loss": 0.4978, + "step": 5908 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018903994375904802, + "loss": 0.641, + "step": 5909 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001890355543141704, + "loss": 0.6046, + "step": 5910 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001890311640414757, + "loss": 0.776, + "step": 5911 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001890267729410047, + "loss": 0.5505, + "step": 5912 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018902238101279828, + "loss": 0.609, + "step": 5913 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001890179882568972, + "loss": 0.5326, + "step": 5914 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018901359467334236, + "loss": 0.5342, + "step": 5915 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018900920026217462, + "loss": 0.5757, + "step": 5916 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018900480502343482, + "loss": 0.5791, + "step": 5917 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018900040895716375, + "loss": 0.5738, + "step": 5918 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018899601206340242, + "loss": 0.494, + "step": 5919 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018899161434219157, + "loss": 0.6582, + "step": 5920 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018898721579357222, + "loss": 0.5309, + "step": 5921 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018898281641758518, + "loss": 0.4948, + "step": 5922 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018897841621427137, + "loss": 0.5096, + "step": 5923 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001889740151836717, + "loss": 0.5365, + "step": 5924 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018896961332582715, + "loss": 0.4891, + "step": 5925 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018896521064077856, + "loss": 0.5817, + "step": 5926 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001889608071285669, + "loss": 0.4677, + "step": 5927 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018895640278923312, + "loss": 0.6313, + "step": 5928 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018895199762281815, + "loss": 0.4513, + "step": 5929 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018894759162936295, + "loss": 0.5823, + "step": 5930 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001889431848089085, + "loss": 0.4706, + "step": 5931 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001889387771614958, + "loss": 0.5492, + "step": 5932 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018893436868716576, + "loss": 0.5917, + "step": 5933 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018892995938595944, + "loss": 0.5576, + "step": 5934 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018892554925791774, + "loss": 0.4145, + "step": 5935 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018892113830308179, + "loss": 0.5173, + "step": 5936 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001889167265214925, + "loss": 0.5052, + "step": 5937 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018891231391319094, + "loss": 0.4564, + "step": 5938 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018890790047821813, + "loss": 0.6466, + "step": 5939 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018890348621661507, + "loss": 0.5781, + "step": 5940 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018889907112842284, + "loss": 0.4632, + "step": 5941 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001888946552136825, + "loss": 0.5543, + "step": 5942 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018889023847243507, + "loss": 0.6597, + "step": 5943 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001888858209047216, + "loss": 0.5004, + "step": 5944 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018888140251058325, + "loss": 0.6614, + "step": 5945 + }, + { + "epoch": 0.18, + "learning_rate": 0.000188876983290061, + "loss": 0.5963, + "step": 5946 + }, + { + "epoch": 0.18, + "learning_rate": 0.000188872563243196, + "loss": 0.543, + "step": 5947 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018886814237002933, + "loss": 0.4906, + "step": 5948 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001888637206706021, + "loss": 0.5023, + "step": 5949 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018885929814495538, + "loss": 0.5981, + "step": 5950 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018885487479313034, + "loss": 0.5111, + "step": 5951 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018885045061516807, + "loss": 0.5099, + "step": 5952 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018884602561110975, + "loss": 0.5651, + "step": 5953 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018884159978099652, + "loss": 0.5155, + "step": 5954 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018883717312486947, + "loss": 0.6102, + "step": 5955 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018883274564276977, + "loss": 0.4762, + "step": 5956 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018882831733473863, + "loss": 0.5047, + "step": 5957 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018882388820081722, + "loss": 0.5585, + "step": 5958 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018881945824104668, + "loss": 0.4807, + "step": 5959 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001888150274554682, + "loss": 0.4371, + "step": 5960 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018881059584412301, + "loss": 0.587, + "step": 5961 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001888061634070523, + "loss": 0.606, + "step": 5962 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018880173014429726, + "loss": 0.6753, + "step": 5963 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018879729605589916, + "loss": 0.4907, + "step": 5964 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018879286114189915, + "loss": 0.4832, + "step": 5965 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018878842540233853, + "loss": 0.5029, + "step": 5966 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001887839888372585, + "loss": 0.5199, + "step": 5967 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018877955144670034, + "loss": 0.5079, + "step": 5968 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018877511323070532, + "loss": 0.5853, + "step": 5969 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018877067418931464, + "loss": 0.5442, + "step": 5970 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018876623432256964, + "loss": 0.5376, + "step": 5971 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018876179363051153, + "loss": 0.5248, + "step": 5972 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018875735211318172, + "loss": 0.5454, + "step": 5973 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018875290977062137, + "loss": 0.5408, + "step": 5974 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018874846660287182, + "loss": 0.4783, + "step": 5975 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001887440226099744, + "loss": 0.4927, + "step": 5976 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001887395777919705, + "loss": 0.4923, + "step": 5977 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001887351321489013, + "loss": 0.7186, + "step": 5978 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018873068568080824, + "loss": 0.5522, + "step": 5979 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018872623838773263, + "loss": 0.5067, + "step": 5980 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018872179026971578, + "loss": 0.6157, + "step": 5981 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018871734132679913, + "loss": 0.6675, + "step": 5982 + }, + { + "epoch": 0.18, + "learning_rate": 0.000188712891559024, + "loss": 0.5306, + "step": 5983 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018870844096643171, + "loss": 0.5949, + "step": 5984 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018870398954906373, + "loss": 0.4656, + "step": 5985 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018869953730696142, + "loss": 0.5525, + "step": 5986 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018869508424016615, + "loss": 0.5982, + "step": 5987 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018869063034871933, + "loss": 0.533, + "step": 5988 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001886861756326624, + "loss": 0.5613, + "step": 5989 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018868172009203673, + "loss": 0.493, + "step": 5990 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018867726372688378, + "loss": 0.5943, + "step": 5991 + }, + { + "epoch": 0.18, + "learning_rate": 0.000188672806537245, + "loss": 0.5635, + "step": 5992 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018866834852316176, + "loss": 0.5404, + "step": 5993 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018866388968467556, + "loss": 0.4512, + "step": 5994 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001886594300218279, + "loss": 0.5944, + "step": 5995 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018865496953466014, + "loss": 0.6983, + "step": 5996 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018865050822321387, + "loss": 0.6213, + "step": 5997 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018864604608753044, + "loss": 0.5438, + "step": 5998 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018864158312765144, + "loss": 0.5575, + "step": 5999 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001886371193436183, + "loss": 0.6188, + "step": 6000 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001886326547354726, + "loss": 0.5496, + "step": 6001 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018862818930325578, + "loss": 0.5661, + "step": 6002 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018862372304700938, + "loss": 0.6481, + "step": 6003 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018861925596677493, + "loss": 0.5584, + "step": 6004 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018861478806259395, + "loss": 0.545, + "step": 6005 + }, + { + "epoch": 0.18, + "learning_rate": 0.000188610319334508, + "loss": 0.5833, + "step": 6006 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018860584978255858, + "loss": 0.589, + "step": 6007 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018860137940678734, + "loss": 0.6438, + "step": 6008 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018859690820723573, + "loss": 0.5906, + "step": 6009 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018859243618394542, + "loss": 0.5725, + "step": 6010 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001885879633369579, + "loss": 0.656, + "step": 6011 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018858348966631488, + "loss": 0.5221, + "step": 6012 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018857901517205784, + "loss": 0.6168, + "step": 6013 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001885745398542284, + "loss": 0.6558, + "step": 6014 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018857006371286823, + "loss": 0.6011, + "step": 6015 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018856558674801887, + "loss": 0.5213, + "step": 6016 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018856110895972198, + "loss": 0.6188, + "step": 6017 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018855663034801924, + "loss": 0.4999, + "step": 6018 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018855215091295223, + "loss": 0.5746, + "step": 6019 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001885476706545626, + "loss": 0.6087, + "step": 6020 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018854318957289203, + "loss": 0.6196, + "step": 6021 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001885387076679822, + "loss": 0.5198, + "step": 6022 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001885342249398747, + "loss": 0.5662, + "step": 6023 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018852974138861128, + "loss": 0.5471, + "step": 6024 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018852525701423361, + "loss": 0.4577, + "step": 6025 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001885207718167834, + "loss": 0.634, + "step": 6026 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018851628579630232, + "loss": 0.6227, + "step": 6027 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018851179895283207, + "loss": 0.5209, + "step": 6028 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018850731128641445, + "loss": 0.5674, + "step": 6029 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018850282279709107, + "loss": 0.5862, + "step": 6030 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018849833348490375, + "loss": 0.5759, + "step": 6031 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001884938433498942, + "loss": 0.5652, + "step": 6032 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018848935239210418, + "loss": 0.5817, + "step": 6033 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018848486061157538, + "loss": 0.56, + "step": 6034 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018848036800834964, + "loss": 0.6297, + "step": 6035 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001884758745824687, + "loss": 0.4662, + "step": 6036 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018847138033397434, + "loss": 0.5107, + "step": 6037 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018846688526290834, + "loss": 0.5935, + "step": 6038 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018846238936931248, + "loss": 0.5531, + "step": 6039 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018845789265322865, + "loss": 0.5017, + "step": 6040 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018845339511469852, + "loss": 0.5777, + "step": 6041 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018844889675376401, + "loss": 0.509, + "step": 6042 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001884443975704669, + "loss": 0.6526, + "step": 6043 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018843989756484907, + "loss": 0.5527, + "step": 6044 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018843539673695227, + "loss": 0.5338, + "step": 6045 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018843089508681842, + "loss": 0.6665, + "step": 6046 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018842639261448934, + "loss": 0.6229, + "step": 6047 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001884218893200069, + "loss": 0.5656, + "step": 6048 + }, + { + "epoch": 0.18, + "learning_rate": 0.000188417385203413, + "loss": 0.5228, + "step": 6049 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001884128802647495, + "loss": 0.5993, + "step": 6050 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001884083745040582, + "loss": 0.5453, + "step": 6051 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018840386792138113, + "loss": 0.5983, + "step": 6052 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018839936051676015, + "loss": 0.6141, + "step": 6053 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018839485229023712, + "loss": 0.4575, + "step": 6054 + }, + { + "epoch": 0.18, + "learning_rate": 0.000188390343241854, + "loss": 0.6121, + "step": 6055 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001883858333716527, + "loss": 0.6109, + "step": 6056 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018838132267967513, + "loss": 0.5812, + "step": 6057 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018837681116596323, + "loss": 0.5138, + "step": 6058 + }, + { + "epoch": 0.18, + "learning_rate": 0.000188372298830559, + "loss": 0.6359, + "step": 6059 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018836778567350436, + "loss": 0.5802, + "step": 6060 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018836327169484123, + "loss": 0.5417, + "step": 6061 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018835875689461164, + "loss": 0.5416, + "step": 6062 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018835424127285755, + "loss": 0.6232, + "step": 6063 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018834972482962092, + "loss": 0.6054, + "step": 6064 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018834520756494378, + "loss": 0.6317, + "step": 6065 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001883406894788681, + "loss": 0.5821, + "step": 6066 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001883361705714359, + "loss": 0.5903, + "step": 6067 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001883316508426892, + "loss": 0.5502, + "step": 6068 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018832713029267, + "loss": 0.5988, + "step": 6069 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018832260892142038, + "loss": 0.5108, + "step": 6070 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018831808672898232, + "loss": 0.5751, + "step": 6071 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001883135637153979, + "loss": 0.5348, + "step": 6072 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018830903988070917, + "loss": 0.5837, + "step": 6073 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018830451522495816, + "loss": 0.6357, + "step": 6074 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018829998974818698, + "loss": 0.5239, + "step": 6075 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018829546345043764, + "loss": 0.5605, + "step": 6076 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018829093633175234, + "loss": 0.533, + "step": 6077 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018828640839217307, + "loss": 0.6389, + "step": 6078 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018828187963174195, + "loss": 0.587, + "step": 6079 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018827735005050113, + "loss": 0.5436, + "step": 6080 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018827281964849268, + "loss": 0.5214, + "step": 6081 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018826828842575873, + "loss": 0.5639, + "step": 6082 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018826375638234143, + "loss": 0.6547, + "step": 6083 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018825922351828288, + "loss": 0.5697, + "step": 6084 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018825468983362526, + "loss": 0.5571, + "step": 6085 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001882501553284107, + "loss": 0.5646, + "step": 6086 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018824562000268138, + "loss": 0.6405, + "step": 6087 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018824108385647947, + "loss": 0.608, + "step": 6088 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001882365468898471, + "loss": 0.5934, + "step": 6089 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001882320091028265, + "loss": 0.6282, + "step": 6090 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018822747049545985, + "loss": 0.5415, + "step": 6091 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018822293106778934, + "loss": 0.556, + "step": 6092 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018821839081985714, + "loss": 0.5386, + "step": 6093 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018821384975170556, + "loss": 0.5133, + "step": 6094 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018820930786337675, + "loss": 0.5897, + "step": 6095 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018820476515491296, + "loss": 0.5652, + "step": 6096 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018820022162635637, + "loss": 0.6072, + "step": 6097 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018819567727774932, + "loss": 0.5471, + "step": 6098 + }, + { + "epoch": 0.18, + "learning_rate": 0.000188191132109134, + "loss": 0.6063, + "step": 6099 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001881865861205527, + "loss": 0.6187, + "step": 6100 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001881820393120476, + "loss": 0.5719, + "step": 6101 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001881774916836611, + "loss": 0.4831, + "step": 6102 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018817294323543544, + "loss": 0.6038, + "step": 6103 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018816839396741287, + "loss": 0.5789, + "step": 6104 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018816384387963572, + "loss": 0.4502, + "step": 6105 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001881592929721463, + "loss": 0.6239, + "step": 6106 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001881547412449869, + "loss": 0.6924, + "step": 6107 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018815018869819988, + "loss": 0.5489, + "step": 6108 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018814563533182748, + "loss": 0.6504, + "step": 6109 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018814108114591213, + "loss": 0.5014, + "step": 6110 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018813652614049612, + "loss": 0.5841, + "step": 6111 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018813197031562184, + "loss": 0.606, + "step": 6112 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018812741367133163, + "loss": 0.5889, + "step": 6113 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018812285620766782, + "loss": 0.5983, + "step": 6114 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018811829792467284, + "loss": 0.5976, + "step": 6115 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018811373882238902, + "loss": 0.5422, + "step": 6116 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001881091789008588, + "loss": 0.6436, + "step": 6117 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018810461816012454, + "loss": 0.6095, + "step": 6118 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018810005660022867, + "loss": 0.5255, + "step": 6119 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018809549422121356, + "loss": 0.5333, + "step": 6120 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018809093102312169, + "loss": 0.5203, + "step": 6121 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018808636700599543, + "loss": 0.5985, + "step": 6122 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018808180216987725, + "loss": 0.5986, + "step": 6123 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018807723651480956, + "loss": 0.4869, + "step": 6124 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018807267004083487, + "loss": 0.5421, + "step": 6125 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018806810274799555, + "loss": 0.6238, + "step": 6126 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018806353463633415, + "loss": 0.5229, + "step": 6127 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018805896570589308, + "loss": 0.4903, + "step": 6128 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018805439595671484, + "loss": 0.5537, + "step": 6129 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018804982538884194, + "loss": 0.6086, + "step": 6130 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018804525400231684, + "loss": 0.6233, + "step": 6131 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018804068179718205, + "loss": 0.6166, + "step": 6132 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001880361087734801, + "loss": 0.6186, + "step": 6133 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018803153493125352, + "loss": 0.545, + "step": 6134 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001880269602705448, + "loss": 0.6254, + "step": 6135 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018802238479139648, + "loss": 0.5587, + "step": 6136 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018801780849385112, + "loss": 0.6124, + "step": 6137 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018801323137795123, + "loss": 0.5594, + "step": 6138 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018800865344373945, + "loss": 0.5924, + "step": 6139 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018800407469125822, + "loss": 0.5918, + "step": 6140 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018799949512055022, + "loss": 0.5353, + "step": 6141 + }, + { + "epoch": 0.18, + "learning_rate": 0.000187994914731658, + "loss": 0.6055, + "step": 6142 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018799033352462406, + "loss": 0.5379, + "step": 6143 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018798575149949116, + "loss": 0.5932, + "step": 6144 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018798116865630175, + "loss": 0.5906, + "step": 6145 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018797658499509852, + "loss": 0.4683, + "step": 6146 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018797200051592406, + "loss": 0.6396, + "step": 6147 + }, + { + "epoch": 0.18, + "learning_rate": 0.000187967415218821, + "loss": 0.4682, + "step": 6148 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018796282910383202, + "loss": 0.5913, + "step": 6149 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018795824217099965, + "loss": 0.5557, + "step": 6150 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018795365442036662, + "loss": 0.5099, + "step": 6151 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001879490658519756, + "loss": 0.4384, + "step": 6152 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018794447646586918, + "loss": 0.5602, + "step": 6153 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018793988626209006, + "loss": 0.4814, + "step": 6154 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018793529524068094, + "loss": 0.6255, + "step": 6155 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018793070340168453, + "loss": 0.5691, + "step": 6156 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018792611074514343, + "loss": 0.5996, + "step": 6157 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018792151727110043, + "loss": 0.5164, + "step": 6158 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001879169229795982, + "loss": 0.5191, + "step": 6159 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018791232787067946, + "loss": 0.6169, + "step": 6160 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018790773194438692, + "loss": 0.5509, + "step": 6161 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018790313520076332, + "loss": 0.6163, + "step": 6162 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018789853763985143, + "loss": 0.5587, + "step": 6163 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018789393926169393, + "loss": 0.4767, + "step": 6164 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018788934006633366, + "loss": 0.5665, + "step": 6165 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018788474005381335, + "loss": 0.663, + "step": 6166 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018788013922417572, + "loss": 0.5278, + "step": 6167 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018787553757746355, + "loss": 0.5367, + "step": 6168 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001878709351137197, + "loss": 0.6005, + "step": 6169 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001878663318329869, + "loss": 0.5168, + "step": 6170 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018786172773530796, + "loss": 0.5795, + "step": 6171 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018785712282072572, + "loss": 0.4526, + "step": 6172 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018785251708928295, + "loss": 0.5186, + "step": 6173 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018784791054102248, + "loss": 0.5089, + "step": 6174 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018784330317598717, + "loss": 0.6263, + "step": 6175 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018783869499421982, + "loss": 0.4218, + "step": 6176 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018783408599576333, + "loss": 0.5616, + "step": 6177 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001878294761806605, + "loss": 0.5449, + "step": 6178 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001878248655489542, + "loss": 0.4657, + "step": 6179 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001878202541006873, + "loss": 0.5164, + "step": 6180 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018781564183590268, + "loss": 0.5508, + "step": 6181 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001878110287546432, + "loss": 0.5794, + "step": 6182 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001878064148569518, + "loss": 0.5202, + "step": 6183 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018780180014287134, + "loss": 0.5658, + "step": 6184 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018779718461244476, + "loss": 0.4916, + "step": 6185 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018779256826571491, + "loss": 0.5897, + "step": 6186 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018778795110272479, + "loss": 0.6561, + "step": 6187 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018778333312351726, + "loss": 0.5569, + "step": 6188 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018777871432813532, + "loss": 0.6148, + "step": 6189 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018777409471662184, + "loss": 0.5628, + "step": 6190 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018776947428901983, + "loss": 0.5344, + "step": 6191 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001877648530453722, + "loss": 0.57, + "step": 6192 + }, + { + "epoch": 0.18, + "learning_rate": 0.000187760230985722, + "loss": 0.5297, + "step": 6193 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018775560811011213, + "loss": 0.624, + "step": 6194 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018775098441858558, + "loss": 0.6161, + "step": 6195 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018774635991118536, + "loss": 0.5529, + "step": 6196 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018774173458795446, + "loss": 0.5876, + "step": 6197 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018773710844893586, + "loss": 0.6072, + "step": 6198 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018773248149417263, + "loss": 0.551, + "step": 6199 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018772785372370772, + "loss": 0.5732, + "step": 6200 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001877232251375842, + "loss": 0.5792, + "step": 6201 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018771859573584512, + "loss": 0.5479, + "step": 6202 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018771396551853348, + "loss": 0.6167, + "step": 6203 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018770933448569233, + "loss": 0.4738, + "step": 6204 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018770470263736476, + "loss": 0.6417, + "step": 6205 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018770006997359385, + "loss": 0.5712, + "step": 6206 + }, + { + "epoch": 0.18, + "learning_rate": 0.0001876954364944226, + "loss": 0.5999, + "step": 6207 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018769080219989412, + "loss": 0.5509, + "step": 6208 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018768616709005154, + "loss": 0.6866, + "step": 6209 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018768153116493794, + "loss": 0.5664, + "step": 6210 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018767689442459638, + "loss": 0.5918, + "step": 6211 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018767225686907004, + "loss": 0.7387, + "step": 6212 + }, + { + "epoch": 0.18, + "learning_rate": 0.00018766761849840197, + "loss": 0.6189, + "step": 6213 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018766297931263533, + "loss": 0.5603, + "step": 6214 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018765833931181324, + "loss": 0.5415, + "step": 6215 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018765369849597886, + "loss": 0.5965, + "step": 6216 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018764905686517532, + "loss": 0.6597, + "step": 6217 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001876444144194458, + "loss": 0.5716, + "step": 6218 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018763977115883346, + "loss": 0.5683, + "step": 6219 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018763512708338138, + "loss": 0.5156, + "step": 6220 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018763048219313285, + "loss": 0.5658, + "step": 6221 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018762583648813106, + "loss": 0.5669, + "step": 6222 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018762118996841915, + "loss": 0.6433, + "step": 6223 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018761654263404034, + "loss": 0.6167, + "step": 6224 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018761189448503781, + "loss": 0.4811, + "step": 6225 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018760724552145485, + "loss": 0.4917, + "step": 6226 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018760259574333458, + "loss": 0.6987, + "step": 6227 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018759794515072034, + "loss": 0.5434, + "step": 6228 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001875932937436553, + "loss": 0.6228, + "step": 6229 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018758864152218271, + "loss": 0.5829, + "step": 6230 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018758398848634587, + "loss": 0.6255, + "step": 6231 + }, + { + "epoch": 0.19, + "learning_rate": 0.000187579334636188, + "loss": 0.5737, + "step": 6232 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018757467997175238, + "loss": 0.6341, + "step": 6233 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001875700244930823, + "loss": 0.4896, + "step": 6234 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018756536820022102, + "loss": 0.6184, + "step": 6235 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018756071109321182, + "loss": 0.4588, + "step": 6236 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001875560531720981, + "loss": 0.5934, + "step": 6237 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018755139443692303, + "loss": 0.4567, + "step": 6238 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018754673488773, + "loss": 0.6334, + "step": 6239 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018754207452456234, + "loss": 0.5255, + "step": 6240 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018753741334746334, + "loss": 0.5962, + "step": 6241 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018753275135647638, + "loss": 0.6191, + "step": 6242 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001875280885516448, + "loss": 0.6852, + "step": 6243 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018752342493301191, + "loss": 0.5868, + "step": 6244 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001875187605006211, + "loss": 0.7117, + "step": 6245 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018751409525451576, + "loss": 0.6339, + "step": 6246 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001875094291947392, + "loss": 0.5653, + "step": 6247 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018750476232133487, + "loss": 0.5101, + "step": 6248 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018750009463434615, + "loss": 0.6162, + "step": 6249 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001874954261338164, + "loss": 0.5942, + "step": 6250 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018749075681978906, + "loss": 0.7091, + "step": 6251 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018748608669230752, + "loss": 0.5714, + "step": 6252 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018748141575141523, + "loss": 0.5804, + "step": 6253 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001874767439971556, + "loss": 0.5135, + "step": 6254 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018747207142957208, + "loss": 0.5764, + "step": 6255 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001874673980487081, + "loss": 0.6444, + "step": 6256 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001874627238546071, + "loss": 0.4968, + "step": 6257 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018745804884731255, + "loss": 0.6738, + "step": 6258 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018745337302686794, + "loss": 0.595, + "step": 6259 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001874486963933167, + "loss": 0.662, + "step": 6260 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018744401894670235, + "loss": 0.5533, + "step": 6261 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018743934068706834, + "loss": 0.5183, + "step": 6262 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018743466161445823, + "loss": 0.6388, + "step": 6263 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018742998172891544, + "loss": 0.5992, + "step": 6264 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018742530103048357, + "loss": 0.4929, + "step": 6265 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018742061951920605, + "loss": 0.5792, + "step": 6266 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001874159371951265, + "loss": 0.5774, + "step": 6267 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018741125405828836, + "loss": 0.6752, + "step": 6268 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018740657010873525, + "loss": 0.6038, + "step": 6269 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001874018853465107, + "loss": 0.5072, + "step": 6270 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018739719977165824, + "loss": 0.5984, + "step": 6271 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001873925133842214, + "loss": 0.5953, + "step": 6272 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001873878261842439, + "loss": 0.4334, + "step": 6273 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018738313817176914, + "loss": 0.4865, + "step": 6274 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001873784493468408, + "loss": 0.6136, + "step": 6275 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018737375970950248, + "loss": 0.5174, + "step": 6276 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018736906925979777, + "loss": 0.5646, + "step": 6277 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001873643779977703, + "loss": 0.6531, + "step": 6278 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018735968592346362, + "loss": 0.6452, + "step": 6279 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018735499303692142, + "loss": 0.5832, + "step": 6280 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001873502993381873, + "loss": 0.4752, + "step": 6281 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018734560482730491, + "loss": 0.5881, + "step": 6282 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018734090950431792, + "loss": 0.4985, + "step": 6283 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018733621336926995, + "loss": 0.5273, + "step": 6284 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018733151642220467, + "loss": 0.5663, + "step": 6285 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018732681866316577, + "loss": 0.5366, + "step": 6286 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001873221200921969, + "loss": 0.6585, + "step": 6287 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018731742070934176, + "loss": 0.4931, + "step": 6288 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018731272051464407, + "loss": 0.4951, + "step": 6289 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018730801950814747, + "loss": 0.5615, + "step": 6290 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018730331768989573, + "loss": 0.5076, + "step": 6291 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001872986150599325, + "loss": 0.6251, + "step": 6292 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001872939116183016, + "loss": 0.5587, + "step": 6293 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018728920736504664, + "loss": 0.4482, + "step": 6294 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018728450230021145, + "loss": 0.641, + "step": 6295 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018727979642383974, + "loss": 0.5365, + "step": 6296 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018727508973597528, + "loss": 0.4556, + "step": 6297 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018727038223666182, + "loss": 0.602, + "step": 6298 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018726567392594312, + "loss": 0.6088, + "step": 6299 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018726096480386296, + "loss": 0.5144, + "step": 6300 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018725625487046513, + "loss": 0.5273, + "step": 6301 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018725154412579343, + "loss": 0.5985, + "step": 6302 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018724683256989163, + "loss": 0.5642, + "step": 6303 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018724212020280358, + "loss": 0.6956, + "step": 6304 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018723740702457307, + "loss": 0.624, + "step": 6305 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001872326930352439, + "loss": 0.6249, + "step": 6306 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001872279782348599, + "loss": 0.6105, + "step": 6307 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018722326262346495, + "loss": 0.5926, + "step": 6308 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018721854620110286, + "loss": 0.6682, + "step": 6309 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001872138289678175, + "loss": 0.6053, + "step": 6310 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018720911092365273, + "loss": 0.6069, + "step": 6311 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018720439206865237, + "loss": 0.5682, + "step": 6312 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018719967240286031, + "loss": 0.5118, + "step": 6313 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001871949519263205, + "loss": 0.5837, + "step": 6314 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018719023063907674, + "loss": 0.6676, + "step": 6315 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018718550854117302, + "loss": 0.6439, + "step": 6316 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018718078563265315, + "loss": 0.648, + "step": 6317 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001871760619135611, + "loss": 0.5973, + "step": 6318 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018717133738394077, + "loss": 0.532, + "step": 6319 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018716661204383605, + "loss": 0.6292, + "step": 6320 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018716188589329097, + "loss": 0.5781, + "step": 6321 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018715715893234937, + "loss": 0.6126, + "step": 6322 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018715243116105526, + "loss": 0.5751, + "step": 6323 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018714770257945257, + "loss": 0.6022, + "step": 6324 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001871429731875853, + "loss": 0.5007, + "step": 6325 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018713824298549741, + "loss": 0.5893, + "step": 6326 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018713351197323282, + "loss": 0.4898, + "step": 6327 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001871287801508356, + "loss": 0.584, + "step": 6328 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018712404751834968, + "loss": 0.622, + "step": 6329 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018711931407581909, + "loss": 0.6215, + "step": 6330 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018711457982328785, + "loss": 0.4111, + "step": 6331 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018710984476079997, + "loss": 0.5515, + "step": 6332 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018710510888839948, + "loss": 0.5278, + "step": 6333 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018710037220613039, + "loss": 0.6123, + "step": 6334 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018709563471403678, + "loss": 0.522, + "step": 6335 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018709089641216262, + "loss": 0.5763, + "step": 6336 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018708615730055205, + "loss": 0.6642, + "step": 6337 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001870814173792491, + "loss": 0.618, + "step": 6338 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018707667664829783, + "loss": 0.6736, + "step": 6339 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018707193510774235, + "loss": 0.5443, + "step": 6340 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018706719275762666, + "loss": 0.5216, + "step": 6341 + }, + { + "epoch": 0.19, + "learning_rate": 0.000187062449597995, + "loss": 0.6489, + "step": 6342 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001870577056288913, + "loss": 0.5732, + "step": 6343 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001870529608503598, + "loss": 0.5601, + "step": 6344 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018704821526244455, + "loss": 0.6172, + "step": 6345 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018704346886518972, + "loss": 0.6084, + "step": 6346 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001870387216586394, + "loss": 0.5056, + "step": 6347 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001870339736428377, + "loss": 0.5319, + "step": 6348 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001870292248178288, + "loss": 0.5813, + "step": 6349 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018702447518365687, + "loss": 0.6331, + "step": 6350 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018701972474036607, + "loss": 0.5524, + "step": 6351 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018701497348800055, + "loss": 0.569, + "step": 6352 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018701022142660447, + "loss": 0.5774, + "step": 6353 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018700546855622202, + "loss": 0.5631, + "step": 6354 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018700071487689744, + "loss": 0.6863, + "step": 6355 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018699596038867483, + "loss": 0.619, + "step": 6356 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001869912050915985, + "loss": 0.6453, + "step": 6357 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018698644898571264, + "loss": 0.6044, + "step": 6358 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018698169207106137, + "loss": 0.5696, + "step": 6359 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018697693434768903, + "loss": 0.5659, + "step": 6360 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018697217581563984, + "loss": 0.5145, + "step": 6361 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018696741647495802, + "loss": 0.5144, + "step": 6362 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001869626563256878, + "loss": 0.5206, + "step": 6363 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001869578953678735, + "loss": 0.54, + "step": 6364 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001869531336015593, + "loss": 0.5768, + "step": 6365 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018694837102678955, + "loss": 0.4733, + "step": 6366 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001869436076436085, + "loss": 0.5634, + "step": 6367 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001869388434520604, + "loss": 0.5775, + "step": 6368 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018693407845218963, + "loss": 0.6605, + "step": 6369 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018692931264404044, + "loss": 0.572, + "step": 6370 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018692454602765713, + "loss": 0.5963, + "step": 6371 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018691977860308401, + "loss": 0.5345, + "step": 6372 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001869150103703655, + "loss": 0.5484, + "step": 6373 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018691024132954582, + "loss": 0.5729, + "step": 6374 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018690547148066938, + "loss": 0.5513, + "step": 6375 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018690070082378045, + "loss": 0.4868, + "step": 6376 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001868959293589235, + "loss": 0.6343, + "step": 6377 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018689115708614282, + "loss": 0.4867, + "step": 6378 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018688638400548278, + "loss": 0.5774, + "step": 6379 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018688161011698776, + "loss": 0.4881, + "step": 6380 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018687683542070217, + "loss": 0.5503, + "step": 6381 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001868720599166704, + "loss": 0.612, + "step": 6382 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018686728360493683, + "loss": 0.5415, + "step": 6383 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018686250648554587, + "loss": 0.5594, + "step": 6384 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018685772855854196, + "loss": 0.6365, + "step": 6385 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001868529498239695, + "loss": 0.5446, + "step": 6386 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018684817028187292, + "loss": 0.6357, + "step": 6387 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001868433899322967, + "loss": 0.5148, + "step": 6388 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018683860877528522, + "loss": 0.6855, + "step": 6389 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018683382681088296, + "loss": 0.527, + "step": 6390 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001868290440391344, + "loss": 0.5847, + "step": 6391 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018682426046008401, + "loss": 0.4616, + "step": 6392 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018681947607377625, + "loss": 0.5855, + "step": 6393 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018681469088025558, + "loss": 0.4967, + "step": 6394 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018680990487956653, + "loss": 0.5955, + "step": 6395 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001868051180717536, + "loss": 0.6716, + "step": 6396 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018680033045686124, + "loss": 0.6092, + "step": 6397 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018679554203493404, + "loss": 0.5207, + "step": 6398 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001867907528060165, + "loss": 0.5306, + "step": 6399 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018678596277015307, + "loss": 0.5188, + "step": 6400 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018678117192738838, + "loss": 0.5937, + "step": 6401 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018677638027776695, + "loss": 0.4814, + "step": 6402 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001867715878213333, + "loss": 0.6226, + "step": 6403 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018676679455813198, + "loss": 0.6319, + "step": 6404 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018676200048820766, + "loss": 0.5359, + "step": 6405 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001867572056116048, + "loss": 0.5062, + "step": 6406 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018675240992836801, + "loss": 0.4762, + "step": 6407 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018674761343854193, + "loss": 0.5592, + "step": 6408 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018674281614217107, + "loss": 0.5799, + "step": 6409 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001867380180393001, + "loss": 0.5357, + "step": 6410 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001867332191299736, + "loss": 0.5788, + "step": 6411 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001867284194142362, + "loss": 0.4499, + "step": 6412 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018672361889213252, + "loss": 0.6552, + "step": 6413 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018671881756370721, + "loss": 0.5738, + "step": 6414 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018671401542900488, + "loss": 0.5171, + "step": 6415 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018670921248807023, + "loss": 0.4931, + "step": 6416 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018670440874094785, + "loss": 0.6495, + "step": 6417 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001866996041876824, + "loss": 0.4957, + "step": 6418 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018669479882831865, + "loss": 0.5609, + "step": 6419 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018668999266290118, + "loss": 0.7496, + "step": 6420 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018668518569147472, + "loss": 0.6479, + "step": 6421 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018668037791408395, + "loss": 0.5261, + "step": 6422 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018667556933077358, + "loss": 0.5196, + "step": 6423 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001866707599415883, + "loss": 0.6322, + "step": 6424 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018666594974657287, + "loss": 0.5408, + "step": 6425 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018666113874577195, + "loss": 0.5597, + "step": 6426 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001866563269392303, + "loss": 0.6109, + "step": 6427 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018665151432699267, + "loss": 0.5187, + "step": 6428 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001866467009091038, + "loss": 0.5938, + "step": 6429 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001866418866856084, + "loss": 0.6576, + "step": 6430 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018663707165655132, + "loss": 0.4683, + "step": 6431 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018663225582197726, + "loss": 0.527, + "step": 6432 + }, + { + "epoch": 0.19, + "learning_rate": 0.000186627439181931, + "loss": 0.5883, + "step": 6433 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018662262173645732, + "loss": 0.5734, + "step": 6434 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018661780348560108, + "loss": 0.6018, + "step": 6435 + }, + { + "epoch": 0.19, + "learning_rate": 0.000186612984429407, + "loss": 0.5228, + "step": 6436 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001866081645679199, + "loss": 0.5868, + "step": 6437 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018660334390118463, + "loss": 0.6517, + "step": 6438 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018659852242924595, + "loss": 0.6518, + "step": 6439 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018659370015214878, + "loss": 0.4789, + "step": 6440 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018658887706993787, + "loss": 0.5045, + "step": 6441 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018658405318265807, + "loss": 0.6565, + "step": 6442 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018657922849035425, + "loss": 0.6042, + "step": 6443 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001865744029930713, + "loss": 0.6752, + "step": 6444 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018656957669085404, + "loss": 0.558, + "step": 6445 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018656474958374738, + "loss": 0.6858, + "step": 6446 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018655992167179614, + "loss": 0.679, + "step": 6447 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001865550929550453, + "loss": 0.7524, + "step": 6448 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018655026343353964, + "loss": 0.5269, + "step": 6449 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018654543310732419, + "loss": 0.6148, + "step": 6450 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018654060197644378, + "loss": 0.6322, + "step": 6451 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018653577004094332, + "loss": 0.6511, + "step": 6452 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018653093730086777, + "loss": 0.4931, + "step": 6453 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018652610375626206, + "loss": 0.4798, + "step": 6454 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001865212694071711, + "loss": 0.5228, + "step": 6455 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001865164342536399, + "loss": 0.4613, + "step": 6456 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018651159829571334, + "loss": 0.5048, + "step": 6457 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018650676153343643, + "loss": 0.591, + "step": 6458 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018650192396685414, + "loss": 0.5417, + "step": 6459 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018649708559601144, + "loss": 0.4873, + "step": 6460 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001864922464209533, + "loss": 0.5503, + "step": 6461 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018648740644172472, + "loss": 0.6292, + "step": 6462 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018648256565837072, + "loss": 0.5876, + "step": 6463 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001864777240709363, + "loss": 0.5043, + "step": 6464 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018647288167946644, + "loss": 0.5478, + "step": 6465 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001864680384840062, + "loss": 0.4987, + "step": 6466 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018646319448460062, + "loss": 0.5355, + "step": 6467 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018645834968129469, + "loss": 0.555, + "step": 6468 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018645350407413348, + "loss": 0.5068, + "step": 6469 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018644865766316208, + "loss": 0.5714, + "step": 6470 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018644381044842552, + "loss": 0.5907, + "step": 6471 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018643896242996884, + "loss": 0.4128, + "step": 6472 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018643411360783714, + "loss": 0.5021, + "step": 6473 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018642926398207548, + "loss": 0.5752, + "step": 6474 + }, + { + "epoch": 0.19, + "learning_rate": 0.000186424413552729, + "loss": 0.4596, + "step": 6475 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018641956231984277, + "loss": 0.6518, + "step": 6476 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001864147102834619, + "loss": 0.6038, + "step": 6477 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001864098574436315, + "loss": 0.5679, + "step": 6478 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018640500380039666, + "loss": 0.5926, + "step": 6479 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018640014935380252, + "loss": 0.6509, + "step": 6480 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018639529410389424, + "loss": 0.5889, + "step": 6481 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018639043805071697, + "loss": 0.5704, + "step": 6482 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001863855811943158, + "loss": 0.6387, + "step": 6483 + }, + { + "epoch": 0.19, + "learning_rate": 0.000186380723534736, + "loss": 0.5069, + "step": 6484 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018637586507202258, + "loss": 0.5783, + "step": 6485 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018637100580622083, + "loss": 0.4365, + "step": 6486 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001863661457373759, + "loss": 0.5933, + "step": 6487 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018636128486553295, + "loss": 0.6767, + "step": 6488 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001863564231907372, + "loss": 0.5939, + "step": 6489 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001863515607130338, + "loss": 0.4672, + "step": 6490 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018634669743246808, + "loss": 0.5355, + "step": 6491 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018634183334908513, + "loss": 0.5965, + "step": 6492 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018633696846293027, + "loss": 0.5948, + "step": 6493 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018633210277404865, + "loss": 0.5485, + "step": 6494 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018632723628248557, + "loss": 0.6265, + "step": 6495 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018632236898828624, + "loss": 0.6635, + "step": 6496 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018631750089149592, + "loss": 0.4873, + "step": 6497 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001863126319921599, + "loss": 0.5493, + "step": 6498 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001863077622903234, + "loss": 0.5706, + "step": 6499 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018630289178603175, + "loss": 0.5494, + "step": 6500 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018629802047933022, + "loss": 0.5495, + "step": 6501 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018629314837026407, + "loss": 0.5595, + "step": 6502 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001862882754588786, + "loss": 0.51, + "step": 6503 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018628340174521918, + "loss": 0.6642, + "step": 6504 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018627852722933108, + "loss": 0.575, + "step": 6505 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018627365191125958, + "loss": 0.6347, + "step": 6506 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001862687757910501, + "loss": 0.6181, + "step": 6507 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018626389886874787, + "loss": 0.672, + "step": 6508 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018625902114439833, + "loss": 0.5691, + "step": 6509 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018625414261804678, + "loss": 0.6513, + "step": 6510 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018624926328973857, + "loss": 0.5661, + "step": 6511 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001862443831595191, + "loss": 0.616, + "step": 6512 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018623950222743373, + "loss": 0.5945, + "step": 6513 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018623462049352786, + "loss": 0.5108, + "step": 6514 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018622973795784684, + "loss": 0.5981, + "step": 6515 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018622485462043605, + "loss": 0.6147, + "step": 6516 + }, + { + "epoch": 0.19, + "learning_rate": 0.000186219970481341, + "loss": 0.5939, + "step": 6517 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018621508554060698, + "loss": 0.6077, + "step": 6518 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018621019979827948, + "loss": 0.6411, + "step": 6519 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018620531325440389, + "loss": 0.5173, + "step": 6520 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018620042590902566, + "loss": 0.4823, + "step": 6521 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001861955377621902, + "loss": 0.6711, + "step": 6522 + }, + { + "epoch": 0.19, + "learning_rate": 0.000186190648813943, + "loss": 0.5319, + "step": 6523 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001861857590643295, + "loss": 0.6357, + "step": 6524 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001861808685133952, + "loss": 0.5299, + "step": 6525 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018617597716118548, + "loss": 0.6429, + "step": 6526 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018617108500774588, + "loss": 0.5152, + "step": 6527 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001861661920531219, + "loss": 0.5261, + "step": 6528 + }, + { + "epoch": 0.19, + "learning_rate": 0.000186161298297359, + "loss": 0.5389, + "step": 6529 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018615640374050268, + "loss": 0.5861, + "step": 6530 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018615150838259843, + "loss": 0.5767, + "step": 6531 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018614661222369182, + "loss": 0.5431, + "step": 6532 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018614171526382838, + "loss": 0.4746, + "step": 6533 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018613681750305355, + "loss": 0.5668, + "step": 6534 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018613191894141292, + "loss": 0.6111, + "step": 6535 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018612701957895208, + "loss": 0.5495, + "step": 6536 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001861221194157165, + "loss": 0.6582, + "step": 6537 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001861172184517518, + "loss": 0.5856, + "step": 6538 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018611231668710354, + "loss": 0.5187, + "step": 6539 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018610741412181727, + "loss": 0.6215, + "step": 6540 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018610251075593858, + "loss": 0.5334, + "step": 6541 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018609760658951303, + "loss": 0.4749, + "step": 6542 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018609270162258626, + "loss": 0.5009, + "step": 6543 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001860877958552039, + "loss": 0.4944, + "step": 6544 + }, + { + "epoch": 0.19, + "learning_rate": 0.0001860828892874115, + "loss": 0.6047, + "step": 6545 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018607798191925471, + "loss": 0.5661, + "step": 6546 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018607307375077917, + "loss": 0.5209, + "step": 6547 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018606816478203045, + "loss": 0.561, + "step": 6548 + }, + { + "epoch": 0.19, + "learning_rate": 0.00018606325501305428, + "loss": 0.5599, + "step": 6549 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018605834444389622, + "loss": 0.5236, + "step": 6550 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018605343307460205, + "loss": 0.6168, + "step": 6551 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018604852090521728, + "loss": 0.4413, + "step": 6552 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018604360793578767, + "loss": 0.509, + "step": 6553 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001860386941663589, + "loss": 0.6191, + "step": 6554 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018603377959697665, + "loss": 0.617, + "step": 6555 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018602886422768662, + "loss": 0.5057, + "step": 6556 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001860239480585345, + "loss": 0.6803, + "step": 6557 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018601903108956598, + "loss": 0.5032, + "step": 6558 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018601411332082676, + "loss": 0.562, + "step": 6559 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018600919475236265, + "loss": 0.5325, + "step": 6560 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018600427538421928, + "loss": 0.5904, + "step": 6561 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018599935521644247, + "loss": 0.6284, + "step": 6562 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018599443424907795, + "loss": 0.5089, + "step": 6563 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018598951248217143, + "loss": 0.556, + "step": 6564 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018598458991576865, + "loss": 0.5226, + "step": 6565 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018597966654991548, + "loss": 0.5926, + "step": 6566 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018597474238465762, + "loss": 0.5968, + "step": 6567 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018596981742004088, + "loss": 0.7224, + "step": 6568 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018596489165611103, + "loss": 0.687, + "step": 6569 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001859599650929139, + "loss": 0.4944, + "step": 6570 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018595503773049526, + "loss": 0.5825, + "step": 6571 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018595010956890093, + "loss": 0.5483, + "step": 6572 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018594518060817674, + "loss": 0.7008, + "step": 6573 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018594025084836852, + "loss": 0.5476, + "step": 6574 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018593532028952212, + "loss": 0.5161, + "step": 6575 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018593038893168333, + "loss": 0.6653, + "step": 6576 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018592545677489802, + "loss": 0.5472, + "step": 6577 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001859205238192121, + "loss": 0.5925, + "step": 6578 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001859155900646714, + "loss": 0.7006, + "step": 6579 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018591065551132175, + "loss": 0.5015, + "step": 6580 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018590572015920908, + "loss": 0.5217, + "step": 6581 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018590078400837926, + "loss": 0.4675, + "step": 6582 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018589584705887822, + "loss": 0.5003, + "step": 6583 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018589090931075177, + "loss": 0.5967, + "step": 6584 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001858859707640459, + "loss": 0.5547, + "step": 6585 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018588103141880654, + "loss": 0.5735, + "step": 6586 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018587609127507953, + "loss": 0.6565, + "step": 6587 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018587115033291088, + "loss": 0.6223, + "step": 6588 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018586620859234647, + "loss": 0.5955, + "step": 6589 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018586126605343232, + "loss": 0.6025, + "step": 6590 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018585632271621432, + "loss": 0.5971, + "step": 6591 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018585137858073843, + "loss": 0.5565, + "step": 6592 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018584643364705062, + "loss": 0.6509, + "step": 6593 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018584148791519695, + "loss": 0.5291, + "step": 6594 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018583654138522331, + "loss": 0.5671, + "step": 6595 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018583159405717573, + "loss": 0.6649, + "step": 6596 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018582664593110018, + "loss": 0.5519, + "step": 6597 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018582169700704265, + "loss": 0.5652, + "step": 6598 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018581674728504924, + "loss": 0.5853, + "step": 6599 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001858117967651659, + "loss": 0.6484, + "step": 6600 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018580684544743868, + "loss": 0.5497, + "step": 6601 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001858018933319136, + "loss": 0.5329, + "step": 6602 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018579694041863674, + "loss": 0.5112, + "step": 6603 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018579198670765407, + "loss": 0.6184, + "step": 6604 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018578703219901174, + "loss": 0.6017, + "step": 6605 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018578207689275576, + "loss": 0.5569, + "step": 6606 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018577712078893222, + "loss": 0.5073, + "step": 6607 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001857721638875872, + "loss": 0.4411, + "step": 6608 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018576720618876678, + "loss": 0.5644, + "step": 6609 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018576224769251712, + "loss": 0.546, + "step": 6610 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001857572883988842, + "loss": 0.4754, + "step": 6611 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001857523283079142, + "loss": 0.5427, + "step": 6612 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001857473674196532, + "loss": 0.5335, + "step": 6613 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018574240573414742, + "loss": 0.5602, + "step": 6614 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001857374432514429, + "loss": 0.5329, + "step": 6615 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018573247997158578, + "loss": 0.5279, + "step": 6616 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018572751589462225, + "loss": 0.5773, + "step": 6617 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018572255102059843, + "loss": 0.584, + "step": 6618 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018571758534956053, + "loss": 0.659, + "step": 6619 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018571261888155466, + "loss": 0.5329, + "step": 6620 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018570765161662705, + "loss": 0.4312, + "step": 6621 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018570268355482382, + "loss": 0.5953, + "step": 6622 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001856977146961912, + "loss": 0.5327, + "step": 6623 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018569274504077538, + "loss": 0.4695, + "step": 6624 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001856877745886226, + "loss": 0.5718, + "step": 6625 + }, + { + "epoch": 0.2, + "learning_rate": 0.000185682803339779, + "loss": 0.5815, + "step": 6626 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001856778312942909, + "loss": 0.5572, + "step": 6627 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001856728584522044, + "loss": 0.5274, + "step": 6628 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018566788481356588, + "loss": 0.5788, + "step": 6629 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018566291037842145, + "loss": 0.556, + "step": 6630 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018565793514681747, + "loss": 0.5796, + "step": 6631 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018565295911880013, + "loss": 0.592, + "step": 6632 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018564798229441572, + "loss": 0.5425, + "step": 6633 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018564300467371049, + "loss": 0.6238, + "step": 6634 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018563802625673074, + "loss": 0.5252, + "step": 6635 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018563304704352277, + "loss": 0.5301, + "step": 6636 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018562806703413285, + "loss": 0.5262, + "step": 6637 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018562308622860728, + "loss": 0.5847, + "step": 6638 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001856181046269924, + "loss": 0.5337, + "step": 6639 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018561312222933448, + "loss": 0.4581, + "step": 6640 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018560813903567992, + "loss": 0.546, + "step": 6641 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018560315504607494, + "loss": 0.5358, + "step": 6642 + }, + { + "epoch": 0.2, + "learning_rate": 0.000185598170260566, + "loss": 0.4923, + "step": 6643 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018559318467919935, + "loss": 0.6653, + "step": 6644 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018558819830202138, + "loss": 0.5779, + "step": 6645 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001855832111290785, + "loss": 0.5118, + "step": 6646 + }, + { + "epoch": 0.2, + "learning_rate": 0.000185578223160417, + "loss": 0.6586, + "step": 6647 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001855732343960833, + "loss": 0.6111, + "step": 6648 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018556824483612373, + "loss": 0.692, + "step": 6649 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018556325448058477, + "loss": 0.5938, + "step": 6650 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018555826332951276, + "loss": 0.4796, + "step": 6651 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018555327138295412, + "loss": 0.5638, + "step": 6652 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018554827864095529, + "loss": 0.5857, + "step": 6653 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001855432851035626, + "loss": 0.5822, + "step": 6654 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018553829077082258, + "loss": 0.6347, + "step": 6655 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018553329564278165, + "loss": 0.5726, + "step": 6656 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018552829971948617, + "loss": 0.5943, + "step": 6657 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018552330300098267, + "loss": 0.5764, + "step": 6658 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018551830548731757, + "loss": 0.5432, + "step": 6659 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018551330717853738, + "loss": 0.5765, + "step": 6660 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018550830807468854, + "loss": 0.5744, + "step": 6661 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001855033081758175, + "loss": 0.6036, + "step": 6662 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001854983074819708, + "loss": 0.4705, + "step": 6663 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018549330599319493, + "loss": 0.6511, + "step": 6664 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018548830370953634, + "loss": 0.5308, + "step": 6665 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018548330063104159, + "loss": 0.492, + "step": 6666 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018547829675775716, + "loss": 0.5433, + "step": 6667 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018547329208972962, + "loss": 0.5949, + "step": 6668 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001854682866270055, + "loss": 0.6156, + "step": 6669 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001854632803696313, + "loss": 0.4906, + "step": 6670 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018545827331765353, + "loss": 0.5311, + "step": 6671 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018545326547111883, + "loss": 0.5997, + "step": 6672 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018544825683007373, + "loss": 0.5894, + "step": 6673 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001854432473945648, + "loss": 0.5182, + "step": 6674 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001854382371646386, + "loss": 0.7371, + "step": 6675 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018543322614034176, + "loss": 0.5193, + "step": 6676 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018542821432172079, + "loss": 0.5941, + "step": 6677 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018542320170882232, + "loss": 0.5295, + "step": 6678 + }, + { + "epoch": 0.2, + "learning_rate": 0.000185418188301693, + "loss": 0.5365, + "step": 6679 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018541317410037943, + "loss": 0.4425, + "step": 6680 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018540815910492815, + "loss": 0.5633, + "step": 6681 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001854031433153859, + "loss": 0.5138, + "step": 6682 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001853981267317992, + "loss": 0.6023, + "step": 6683 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018539310935421482, + "loss": 0.7139, + "step": 6684 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001853880911826793, + "loss": 0.6062, + "step": 6685 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018538307221723936, + "loss": 0.6534, + "step": 6686 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018537805245794165, + "loss": 0.5626, + "step": 6687 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018537303190483282, + "loss": 0.6188, + "step": 6688 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018536801055795955, + "loss": 0.5958, + "step": 6689 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001853629884173686, + "loss": 0.546, + "step": 6690 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018535796548310656, + "loss": 0.4963, + "step": 6691 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018535294175522018, + "loss": 0.6656, + "step": 6692 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018534791723375616, + "loss": 0.4691, + "step": 6693 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018534289191876125, + "loss": 0.6162, + "step": 6694 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018533786581028213, + "loss": 0.5929, + "step": 6695 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018533283890836554, + "loss": 0.6372, + "step": 6696 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018532781121305826, + "loss": 0.5914, + "step": 6697 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018532278272440696, + "loss": 0.5627, + "step": 6698 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018531775344245844, + "loss": 0.5747, + "step": 6699 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018531272336725943, + "loss": 0.6253, + "step": 6700 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018530769249885677, + "loss": 0.4675, + "step": 6701 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018530266083729717, + "loss": 0.5882, + "step": 6702 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001852976283826274, + "loss": 0.5427, + "step": 6703 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018529259513489431, + "loss": 0.6172, + "step": 6704 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018528756109414463, + "loss": 0.5541, + "step": 6705 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018528252626042523, + "loss": 0.556, + "step": 6706 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001852774906337829, + "loss": 0.578, + "step": 6707 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018527245421426445, + "loss": 0.5931, + "step": 6708 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018526741700191667, + "loss": 0.6323, + "step": 6709 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018526237899678646, + "loss": 0.5706, + "step": 6710 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001852573401989206, + "loss": 0.598, + "step": 6711 + }, + { + "epoch": 0.2, + "learning_rate": 0.000185252300608366, + "loss": 0.6802, + "step": 6712 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001852472602251695, + "loss": 0.5839, + "step": 6713 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018524221904937793, + "loss": 0.4916, + "step": 6714 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001852371770810382, + "loss": 0.52, + "step": 6715 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018523213432019713, + "loss": 0.6354, + "step": 6716 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018522709076690167, + "loss": 0.5382, + "step": 6717 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001852220464211987, + "loss": 0.6034, + "step": 6718 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018521700128313508, + "loss": 0.5794, + "step": 6719 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018521195535275777, + "loss": 0.5269, + "step": 6720 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018520690863011366, + "loss": 0.5803, + "step": 6721 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018520186111524965, + "loss": 0.5138, + "step": 6722 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001851968128082127, + "loss": 0.5188, + "step": 6723 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018519176370904977, + "loss": 0.5196, + "step": 6724 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018518671381780773, + "loss": 0.6097, + "step": 6725 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018518166313453361, + "loss": 0.669, + "step": 6726 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018517661165927434, + "loss": 0.6398, + "step": 6727 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018517155939207686, + "loss": 0.6366, + "step": 6728 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001851665063329882, + "loss": 0.5831, + "step": 6729 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018516145248205528, + "loss": 0.6118, + "step": 6730 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018515639783932514, + "loss": 0.5938, + "step": 6731 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018515134240484473, + "loss": 0.5494, + "step": 6732 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018514628617866108, + "loss": 0.5892, + "step": 6733 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018514122916082117, + "loss": 0.4976, + "step": 6734 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018513617135137208, + "loss": 0.6208, + "step": 6735 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001851311127503608, + "loss": 0.7108, + "step": 6736 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018512605335783438, + "loss": 0.5823, + "step": 6737 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018512099317383982, + "loss": 0.4356, + "step": 6738 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001851159321984242, + "loss": 0.5702, + "step": 6739 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018511087043163453, + "loss": 0.5956, + "step": 6740 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018510580787351794, + "loss": 0.5943, + "step": 6741 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018510074452412147, + "loss": 0.5887, + "step": 6742 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001850956803834922, + "loss": 0.4918, + "step": 6743 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001850906154516772, + "loss": 0.521, + "step": 6744 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001850855497287236, + "loss": 0.6833, + "step": 6745 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001850804832146784, + "loss": 0.6133, + "step": 6746 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018507541590958885, + "loss": 0.5658, + "step": 6747 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018507034781350197, + "loss": 0.6579, + "step": 6748 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001850652789264649, + "loss": 0.67, + "step": 6749 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018506020924852475, + "loss": 0.592, + "step": 6750 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001850551387797287, + "loss": 0.7322, + "step": 6751 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018505006752012385, + "loss": 0.546, + "step": 6752 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018504499546975737, + "loss": 0.6375, + "step": 6753 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018503992262867644, + "loss": 0.5162, + "step": 6754 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018503484899692817, + "loss": 0.5564, + "step": 6755 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001850297745745598, + "loss": 0.5318, + "step": 6756 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018502469936161847, + "loss": 0.4817, + "step": 6757 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018501962335815134, + "loss": 0.5461, + "step": 6758 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018501454656420565, + "loss": 0.5794, + "step": 6759 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001850094689798286, + "loss": 0.5935, + "step": 6760 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001850043906050674, + "loss": 0.6561, + "step": 6761 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018499931143996925, + "loss": 0.5488, + "step": 6762 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018499423148458138, + "loss": 0.7143, + "step": 6763 + }, + { + "epoch": 0.2, + "learning_rate": 0.000184989150738951, + "loss": 0.5942, + "step": 6764 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001849840692031254, + "loss": 0.5757, + "step": 6765 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018497898687715178, + "loss": 0.454, + "step": 6766 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018497390376107743, + "loss": 0.5295, + "step": 6767 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018496881985494956, + "loss": 0.5096, + "step": 6768 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018496373515881545, + "loss": 0.5946, + "step": 6769 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018495864967272243, + "loss": 0.5757, + "step": 6770 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018495356339671775, + "loss": 0.5941, + "step": 6771 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018494847633084867, + "loss": 0.4819, + "step": 6772 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018494338847516251, + "loss": 0.65, + "step": 6773 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018493829982970663, + "loss": 0.5845, + "step": 6774 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018493321039452827, + "loss": 0.5963, + "step": 6775 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018492812016967472, + "loss": 0.503, + "step": 6776 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018492302915519343, + "loss": 0.5269, + "step": 6777 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001849179373511316, + "loss": 0.5512, + "step": 6778 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001849128447575367, + "loss": 0.7161, + "step": 6779 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018490775137445598, + "loss": 0.5239, + "step": 6780 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018490265720193683, + "loss": 0.5668, + "step": 6781 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001848975622400266, + "loss": 0.6176, + "step": 6782 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018489246648877266, + "loss": 0.6127, + "step": 6783 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018488736994822243, + "loss": 0.6915, + "step": 6784 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018488227261842326, + "loss": 0.6856, + "step": 6785 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018487717449942255, + "loss": 0.6012, + "step": 6786 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001848720755912677, + "loss": 0.5935, + "step": 6787 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018486697589400612, + "loss": 0.5771, + "step": 6788 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001848618754076852, + "loss": 0.6118, + "step": 6789 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001848567741323524, + "loss": 0.6501, + "step": 6790 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018485167206805513, + "loss": 0.7041, + "step": 6791 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018484656921484086, + "loss": 0.5528, + "step": 6792 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018484146557275698, + "loss": 0.5976, + "step": 6793 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018483636114185098, + "loss": 0.5722, + "step": 6794 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001848312559221703, + "loss": 0.5591, + "step": 6795 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018482614991376242, + "loss": 0.4589, + "step": 6796 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018482104311667478, + "loss": 0.6638, + "step": 6797 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018481593553095493, + "loss": 0.5271, + "step": 6798 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001848108271566503, + "loss": 0.5471, + "step": 6799 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018480571799380836, + "loss": 0.654, + "step": 6800 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018480060804247669, + "loss": 0.5822, + "step": 6801 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018479549730270276, + "loss": 0.5824, + "step": 6802 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018479038577453411, + "loss": 0.5083, + "step": 6803 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018478527345801822, + "loss": 0.6474, + "step": 6804 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018478016035320267, + "loss": 0.5954, + "step": 6805 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018477504646013498, + "loss": 0.5816, + "step": 6806 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001847699317788627, + "loss": 0.7287, + "step": 6807 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018476481630943335, + "loss": 0.5802, + "step": 6808 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018475970005189454, + "loss": 0.479, + "step": 6809 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018475458300629382, + "loss": 0.6079, + "step": 6810 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018474946517267878, + "loss": 0.6, + "step": 6811 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018474434655109696, + "loss": 0.6274, + "step": 6812 + }, + { + "epoch": 0.2, + "learning_rate": 0.000184739227141596, + "loss": 0.5713, + "step": 6813 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018473410694422348, + "loss": 0.6508, + "step": 6814 + }, + { + "epoch": 0.2, + "learning_rate": 0.000184728985959027, + "loss": 0.6139, + "step": 6815 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018472386418605418, + "loss": 0.6686, + "step": 6816 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018471874162535266, + "loss": 0.5078, + "step": 6817 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018471361827697002, + "loss": 0.5683, + "step": 6818 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018470849414095394, + "loss": 0.4925, + "step": 6819 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018470336921735201, + "loss": 0.6023, + "step": 6820 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018469824350621195, + "loss": 0.6731, + "step": 6821 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018469311700758137, + "loss": 0.5956, + "step": 6822 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001846879897215079, + "loss": 0.5443, + "step": 6823 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001846828616480393, + "loss": 0.6272, + "step": 6824 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001846777327872232, + "loss": 0.4979, + "step": 6825 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018467260313910727, + "loss": 0.5255, + "step": 6826 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018466747270373926, + "loss": 0.6057, + "step": 6827 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001846623414811668, + "loss": 0.5623, + "step": 6828 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018465720947143766, + "loss": 0.4983, + "step": 6829 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001846520766745995, + "loss": 0.6187, + "step": 6830 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018464694309070008, + "loss": 0.6603, + "step": 6831 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001846418087197871, + "loss": 0.5015, + "step": 6832 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018463667356190836, + "loss": 0.5659, + "step": 6833 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018463153761711154, + "loss": 0.5162, + "step": 6834 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001846264008854444, + "loss": 0.5016, + "step": 6835 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018462126336695473, + "loss": 0.6847, + "step": 6836 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018461612506169027, + "loss": 0.6668, + "step": 6837 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018461098596969883, + "loss": 0.5538, + "step": 6838 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018460584609102814, + "loss": 0.507, + "step": 6839 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018460070542572604, + "loss": 0.5566, + "step": 6840 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018459556397384025, + "loss": 0.4537, + "step": 6841 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018459042173541866, + "loss": 0.5719, + "step": 6842 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018458527871050904, + "loss": 0.5522, + "step": 6843 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001845801348991592, + "loss": 0.5132, + "step": 6844 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018457499030141697, + "loss": 0.5843, + "step": 6845 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018456984491733023, + "loss": 0.5603, + "step": 6846 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018456469874694674, + "loss": 0.5693, + "step": 6847 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001845595517903144, + "loss": 0.6112, + "step": 6848 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018455440404748104, + "loss": 0.5408, + "step": 6849 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018454925551849454, + "loss": 0.4572, + "step": 6850 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018454410620340273, + "loss": 0.5597, + "step": 6851 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018453895610225356, + "loss": 0.6044, + "step": 6852 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018453380521509484, + "loss": 0.5484, + "step": 6853 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018452865354197452, + "loss": 0.5172, + "step": 6854 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018452350108294044, + "loss": 0.5616, + "step": 6855 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018451834783804054, + "loss": 0.6427, + "step": 6856 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018451319380732276, + "loss": 0.6063, + "step": 6857 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018450803899083492, + "loss": 0.6072, + "step": 6858 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018450288338862507, + "loss": 0.5301, + "step": 6859 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001844977270007411, + "loss": 0.6223, + "step": 6860 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001844925698272309, + "loss": 0.5049, + "step": 6861 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018448741186814248, + "loss": 0.6305, + "step": 6862 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001844822531235238, + "loss": 0.5173, + "step": 6863 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018447709359342276, + "loss": 0.4744, + "step": 6864 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001844719332778874, + "loss": 0.5166, + "step": 6865 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018446677217696564, + "loss": 0.5558, + "step": 6866 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018446161029070553, + "loss": 0.6056, + "step": 6867 + }, + { + "epoch": 0.2, + "learning_rate": 0.000184456447619155, + "loss": 0.4726, + "step": 6868 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001844512841623621, + "loss": 0.6616, + "step": 6869 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018444611992037485, + "loss": 0.5841, + "step": 6870 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018444095489324118, + "loss": 0.5787, + "step": 6871 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001844357890810092, + "loss": 0.6998, + "step": 6872 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018443062248372688, + "loss": 0.6259, + "step": 6873 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018442545510144234, + "loss": 0.5148, + "step": 6874 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001844202869342035, + "loss": 0.6129, + "step": 6875 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018441511798205853, + "loss": 0.5207, + "step": 6876 + }, + { + "epoch": 0.2, + "learning_rate": 0.0001844099482450554, + "loss": 0.4618, + "step": 6877 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018440477772324222, + "loss": 0.5279, + "step": 6878 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018439960641666707, + "loss": 0.4925, + "step": 6879 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018439443432537801, + "loss": 0.5744, + "step": 6880 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018438926144942315, + "loss": 0.6097, + "step": 6881 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018438408778885055, + "loss": 0.5695, + "step": 6882 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018437891334370832, + "loss": 0.5088, + "step": 6883 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018437373811404465, + "loss": 0.5463, + "step": 6884 + }, + { + "epoch": 0.2, + "learning_rate": 0.00018436856209990752, + "loss": 0.5826, + "step": 6885 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018436338530134516, + "loss": 0.5098, + "step": 6886 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018435820771840565, + "loss": 0.5649, + "step": 6887 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018435302935113716, + "loss": 0.4571, + "step": 6888 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018434785019958786, + "loss": 0.5218, + "step": 6889 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001843426702638058, + "loss": 0.5579, + "step": 6890 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018433748954383922, + "loss": 0.5868, + "step": 6891 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001843323080397363, + "loss": 0.5846, + "step": 6892 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001843271257515452, + "loss": 0.6457, + "step": 6893 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018432194267931407, + "loss": 0.5362, + "step": 6894 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018431675882309113, + "loss": 0.5976, + "step": 6895 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018431157418292459, + "loss": 0.6235, + "step": 6896 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018430638875886258, + "loss": 0.5468, + "step": 6897 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001843012025509534, + "loss": 0.5572, + "step": 6898 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018429601555924527, + "loss": 0.5413, + "step": 6899 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018429082778378635, + "loss": 0.4987, + "step": 6900 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001842856392246249, + "loss": 0.6235, + "step": 6901 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001842804498818092, + "loss": 0.5683, + "step": 6902 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018427525975538747, + "loss": 0.6124, + "step": 6903 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018427006884540794, + "loss": 0.5889, + "step": 6904 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018426487715191893, + "loss": 0.5146, + "step": 6905 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018425968467496864, + "loss": 0.4693, + "step": 6906 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018425449141460537, + "loss": 0.4442, + "step": 6907 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018424929737087745, + "loss": 0.499, + "step": 6908 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018424410254383315, + "loss": 0.5663, + "step": 6909 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018423890693352078, + "loss": 0.5587, + "step": 6910 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018423371053998856, + "loss": 0.6502, + "step": 6911 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018422851336328492, + "loss": 0.5295, + "step": 6912 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001842233154034581, + "loss": 0.5681, + "step": 6913 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001842181166605565, + "loss": 0.5976, + "step": 6914 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001842129171346284, + "loss": 0.5613, + "step": 6915 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001842077168257222, + "loss": 0.59, + "step": 6916 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001842025157338861, + "loss": 0.634, + "step": 6917 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018419731385916865, + "loss": 0.5622, + "step": 6918 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018419211120161813, + "loss": 0.5404, + "step": 6919 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018418690776128292, + "loss": 0.571, + "step": 6920 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001841817035382114, + "loss": 0.497, + "step": 6921 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018417649853245193, + "loss": 0.6131, + "step": 6922 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001841712927440529, + "loss": 0.5395, + "step": 6923 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018416608617306278, + "loss": 0.5265, + "step": 6924 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018416087881952992, + "loss": 0.5111, + "step": 6925 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018415567068350277, + "loss": 0.5059, + "step": 6926 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018415046176502973, + "loss": 0.6335, + "step": 6927 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001841452520641592, + "loss": 0.6976, + "step": 6928 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018414004158093968, + "loss": 0.5252, + "step": 6929 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001841348303154196, + "loss": 0.6017, + "step": 6930 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001841296182676474, + "loss": 0.5089, + "step": 6931 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018412440543767153, + "loss": 0.613, + "step": 6932 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018411919182554048, + "loss": 0.5734, + "step": 6933 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001841139774313027, + "loss": 0.55, + "step": 6934 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018410876225500667, + "loss": 0.574, + "step": 6935 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001841035462967009, + "loss": 0.5718, + "step": 6936 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001840983295564339, + "loss": 0.5794, + "step": 6937 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001840931120342541, + "loss": 0.5052, + "step": 6938 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018408789373021008, + "loss": 0.4829, + "step": 6939 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018408267464435036, + "loss": 0.5646, + "step": 6940 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018407745477672343, + "loss": 0.6425, + "step": 6941 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018407223412737785, + "loss": 0.586, + "step": 6942 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018406701269636213, + "loss": 0.572, + "step": 6943 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018406179048372482, + "loss": 0.6549, + "step": 6944 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001840565674895145, + "loss": 0.6611, + "step": 6945 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018405134371377973, + "loss": 0.5511, + "step": 6946 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018404611915656904, + "loss": 0.5763, + "step": 6947 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018404089381793103, + "loss": 0.5696, + "step": 6948 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001840356676979143, + "loss": 0.4409, + "step": 6949 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018403044079656744, + "loss": 0.6485, + "step": 6950 + }, + { + "epoch": 0.21, + "learning_rate": 0.000184025213113939, + "loss": 0.6268, + "step": 6951 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018401998465007765, + "loss": 0.4646, + "step": 6952 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018401475540503193, + "loss": 0.6264, + "step": 6953 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001840095253788505, + "loss": 0.6102, + "step": 6954 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018400429457158203, + "loss": 0.5005, + "step": 6955 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001839990629832751, + "loss": 0.6444, + "step": 6956 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018399383061397834, + "loss": 0.4716, + "step": 6957 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018398859746374043, + "loss": 0.5404, + "step": 6958 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018398336353261002, + "loss": 0.5869, + "step": 6959 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018397812882063575, + "loss": 0.5867, + "step": 6960 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018397289332786634, + "loss": 0.5173, + "step": 6961 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001839676570543504, + "loss": 0.4995, + "step": 6962 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001839624200001367, + "loss": 0.4852, + "step": 6963 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018395718216527383, + "loss": 0.5031, + "step": 6964 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018395194354981056, + "loss": 0.4896, + "step": 6965 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001839467041537956, + "loss": 0.6054, + "step": 6966 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001839414639772776, + "loss": 0.6005, + "step": 6967 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018393622302030534, + "loss": 0.5753, + "step": 6968 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018393098128292756, + "loss": 0.5227, + "step": 6969 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018392573876519294, + "loss": 0.6142, + "step": 6970 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018392049546715023, + "loss": 0.6742, + "step": 6971 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018391525138884826, + "loss": 0.5982, + "step": 6972 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001839100065303357, + "loss": 0.6334, + "step": 6973 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018390476089166134, + "loss": 0.5867, + "step": 6974 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018389951447287395, + "loss": 0.6107, + "step": 6975 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018389426727402232, + "loss": 0.5906, + "step": 6976 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018388901929515522, + "loss": 0.5565, + "step": 6977 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018388377053632146, + "loss": 0.5336, + "step": 6978 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018387852099756982, + "loss": 0.4658, + "step": 6979 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018387327067894915, + "loss": 0.6093, + "step": 6980 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018386801958050823, + "loss": 0.6106, + "step": 6981 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018386276770229588, + "loss": 0.6172, + "step": 6982 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018385751504436095, + "loss": 0.4828, + "step": 6983 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018385226160675227, + "loss": 0.5733, + "step": 6984 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018384700738951866, + "loss": 0.516, + "step": 6985 + }, + { + "epoch": 0.21, + "learning_rate": 0.000183841752392709, + "loss": 0.5611, + "step": 6986 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018383649661637218, + "loss": 0.5607, + "step": 6987 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018383124006055699, + "loss": 0.6517, + "step": 6988 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018382598272531235, + "loss": 0.6652, + "step": 6989 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018382072461068714, + "loss": 0.5289, + "step": 6990 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018381546571673022, + "loss": 0.677, + "step": 6991 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018381020604349053, + "loss": 0.5753, + "step": 6992 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001838049455910169, + "loss": 0.584, + "step": 6993 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018379968435935836, + "loss": 0.5744, + "step": 6994 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001837944223485637, + "loss": 0.5595, + "step": 6995 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018378915955868193, + "loss": 0.5056, + "step": 6996 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018378389598976193, + "loss": 0.6247, + "step": 6997 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018377863164185265, + "loss": 0.6732, + "step": 6998 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018377336651500308, + "loss": 0.5767, + "step": 6999 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001837681006092621, + "loss": 0.5878, + "step": 7000 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018376283392467874, + "loss": 0.5746, + "step": 7001 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001837575664613019, + "loss": 0.4985, + "step": 7002 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001837522982191806, + "loss": 0.5489, + "step": 7003 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001837470291983638, + "loss": 0.5128, + "step": 7004 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018374175939890055, + "loss": 0.6179, + "step": 7005 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018373648882083975, + "loss": 0.5757, + "step": 7006 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018373121746423045, + "loss": 0.5422, + "step": 7007 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001837259453291217, + "loss": 0.6522, + "step": 7008 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018372067241556245, + "loss": 0.5715, + "step": 7009 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018371539872360178, + "loss": 0.5737, + "step": 7010 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001837101242532887, + "loss": 0.5194, + "step": 7011 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018370484900467223, + "loss": 0.5815, + "step": 7012 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001836995729778014, + "loss": 0.5955, + "step": 7013 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018369429617272535, + "loss": 0.5549, + "step": 7014 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018368901858949307, + "loss": 0.6601, + "step": 7015 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018368374022815367, + "loss": 0.5527, + "step": 7016 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018367846108875618, + "loss": 0.6079, + "step": 7017 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018367318117134973, + "loss": 0.5508, + "step": 7018 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018366790047598337, + "loss": 0.5799, + "step": 7019 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018366261900270626, + "loss": 0.5934, + "step": 7020 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018365733675156743, + "loss": 0.6502, + "step": 7021 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018365205372261603, + "loss": 0.4388, + "step": 7022 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018364676991590114, + "loss": 0.5602, + "step": 7023 + }, + { + "epoch": 0.21, + "learning_rate": 0.000183641485331472, + "loss": 0.4772, + "step": 7024 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001836361999693776, + "loss": 0.5929, + "step": 7025 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018363091382966719, + "loss": 0.6066, + "step": 7026 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018362562691238984, + "loss": 0.4543, + "step": 7027 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018362033921759475, + "loss": 0.6772, + "step": 7028 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001836150507453311, + "loss": 0.627, + "step": 7029 + }, + { + "epoch": 0.21, + "learning_rate": 0.000183609761495648, + "loss": 0.5634, + "step": 7030 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001836044714685947, + "loss": 0.493, + "step": 7031 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001835991806642203, + "loss": 0.6145, + "step": 7032 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018359388908257404, + "loss": 0.5128, + "step": 7033 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018358859672370513, + "loss": 0.5512, + "step": 7034 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001835833035876628, + "loss": 0.5879, + "step": 7035 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018357800967449617, + "loss": 0.5607, + "step": 7036 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018357271498425454, + "loss": 0.6846, + "step": 7037 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001835674195169871, + "loss": 0.4654, + "step": 7038 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018356212327274313, + "loss": 0.63, + "step": 7039 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018355682625157183, + "loss": 0.434, + "step": 7040 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018355152845352246, + "loss": 0.6028, + "step": 7041 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018354622987864426, + "loss": 0.5717, + "step": 7042 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018354093052698654, + "loss": 0.517, + "step": 7043 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018353563039859853, + "loss": 0.6131, + "step": 7044 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018353032949352955, + "loss": 0.6435, + "step": 7045 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018352502781182882, + "loss": 0.694, + "step": 7046 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018351972535354572, + "loss": 0.5791, + "step": 7047 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018351442211872945, + "loss": 0.5694, + "step": 7048 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018350911810742938, + "loss": 0.5304, + "step": 7049 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018350381331969483, + "loss": 0.5536, + "step": 7050 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001834985077555751, + "loss": 0.5667, + "step": 7051 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018349320141511951, + "loss": 0.508, + "step": 7052 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001834878942983774, + "loss": 0.6179, + "step": 7053 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018348258640539817, + "loss": 0.6159, + "step": 7054 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018347727773623107, + "loss": 0.6143, + "step": 7055 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018347196829092557, + "loss": 0.627, + "step": 7056 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018346665806953094, + "loss": 0.6237, + "step": 7057 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001834613470720966, + "loss": 0.5436, + "step": 7058 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018345603529867192, + "loss": 0.5097, + "step": 7059 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018345072274930627, + "loss": 0.5743, + "step": 7060 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018344540942404907, + "loss": 0.5111, + "step": 7061 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018344009532294972, + "loss": 0.5252, + "step": 7062 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001834347804460576, + "loss": 0.5454, + "step": 7063 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018342946479342218, + "loss": 0.5035, + "step": 7064 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018342414836509282, + "loss": 0.6018, + "step": 7065 + }, + { + "epoch": 0.21, + "learning_rate": 0.000183418831161119, + "loss": 0.5031, + "step": 7066 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018341351318155013, + "loss": 0.5715, + "step": 7067 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018340819442643565, + "loss": 0.531, + "step": 7068 + }, + { + "epoch": 0.21, + "learning_rate": 0.000183402874895825, + "loss": 0.6641, + "step": 7069 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018339755458976768, + "loss": 0.5096, + "step": 7070 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018339223350831313, + "loss": 0.5385, + "step": 7071 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018338691165151086, + "loss": 0.5316, + "step": 7072 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001833815890194103, + "loss": 0.582, + "step": 7073 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018337626561206095, + "loss": 0.5673, + "step": 7074 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001833709414295123, + "loss": 0.603, + "step": 7075 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018336561647181392, + "loss": 0.5713, + "step": 7076 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018336029073901523, + "loss": 0.5845, + "step": 7077 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018335496423116577, + "loss": 0.5216, + "step": 7078 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018334963694831505, + "loss": 0.6573, + "step": 7079 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001833443088905127, + "loss": 0.6534, + "step": 7080 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018333898005780811, + "loss": 0.5148, + "step": 7081 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018333365045025094, + "loss": 0.4973, + "step": 7082 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018332832006789072, + "loss": 0.57, + "step": 7083 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018332298891077695, + "loss": 0.5903, + "step": 7084 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018331765697895925, + "loss": 0.5589, + "step": 7085 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018331232427248718, + "loss": 0.6682, + "step": 7086 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018330699079141034, + "loss": 0.5714, + "step": 7087 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001833016565357783, + "loss": 0.6251, + "step": 7088 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018329632150564063, + "loss": 0.5196, + "step": 7089 + }, + { + "epoch": 0.21, + "learning_rate": 0.000183290985701047, + "loss": 0.5765, + "step": 7090 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018328564912204698, + "loss": 0.6902, + "step": 7091 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018328031176869015, + "loss": 0.4673, + "step": 7092 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018327497364102618, + "loss": 0.4633, + "step": 7093 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018326963473910474, + "loss": 0.663, + "step": 7094 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001832642950629754, + "loss": 0.4974, + "step": 7095 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001832589546126878, + "loss": 0.529, + "step": 7096 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018325361338829166, + "loss": 0.5419, + "step": 7097 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001832482713898366, + "loss": 0.6133, + "step": 7098 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018324292861737228, + "loss": 0.4087, + "step": 7099 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001832375850709484, + "loss": 0.5055, + "step": 7100 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018323224075061462, + "loss": 0.4835, + "step": 7101 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018322689565642065, + "loss": 0.5271, + "step": 7102 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018322154978841617, + "loss": 0.4539, + "step": 7103 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018321620314665086, + "loss": 0.5899, + "step": 7104 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018321085573117448, + "loss": 0.5176, + "step": 7105 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018320550754203672, + "loss": 0.5244, + "step": 7106 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018320015857928735, + "loss": 0.6097, + "step": 7107 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018319480884297607, + "loss": 0.5159, + "step": 7108 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018318945833315257, + "loss": 0.5355, + "step": 7109 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018318410704986667, + "loss": 0.5905, + "step": 7110 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018317875499316807, + "loss": 0.5669, + "step": 7111 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001831734021631066, + "loss": 0.5342, + "step": 7112 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018316804855973194, + "loss": 0.6093, + "step": 7113 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018316269418309393, + "loss": 0.4823, + "step": 7114 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018315733903324235, + "loss": 0.5241, + "step": 7115 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018315198311022696, + "loss": 0.5459, + "step": 7116 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018314662641409757, + "loss": 0.6462, + "step": 7117 + }, + { + "epoch": 0.21, + "learning_rate": 0.000183141268944904, + "loss": 0.6001, + "step": 7118 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018313591070269602, + "loss": 0.5328, + "step": 7119 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001831305516875235, + "loss": 0.5398, + "step": 7120 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018312519189943625, + "loss": 0.6063, + "step": 7121 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018311983133848406, + "loss": 0.5586, + "step": 7122 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018311447000471685, + "loss": 0.627, + "step": 7123 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001831091078981844, + "loss": 0.5606, + "step": 7124 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018310374501893662, + "loss": 0.553, + "step": 7125 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018309838136702328, + "loss": 0.5505, + "step": 7126 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018309301694249437, + "loss": 0.519, + "step": 7127 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001830876517453997, + "loss": 0.6717, + "step": 7128 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001830822857757891, + "loss": 0.604, + "step": 7129 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018307691903371258, + "loss": 0.483, + "step": 7130 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018307155151921995, + "loss": 0.6079, + "step": 7131 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018306618323236115, + "loss": 0.4322, + "step": 7132 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001830608141731861, + "loss": 0.534, + "step": 7133 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001830554443417447, + "loss": 0.628, + "step": 7134 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018305007373808688, + "loss": 0.6492, + "step": 7135 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018304470236226255, + "loss": 0.5396, + "step": 7136 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001830393302143217, + "loss": 0.5502, + "step": 7137 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018303395729431422, + "loss": 0.5972, + "step": 7138 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018302858360229015, + "loss": 0.53, + "step": 7139 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018302320913829938, + "loss": 0.5343, + "step": 7140 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018301783390239188, + "loss": 0.6009, + "step": 7141 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001830124578946177, + "loss": 0.565, + "step": 7142 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001830070811150267, + "loss": 0.481, + "step": 7143 + }, + { + "epoch": 0.21, + "learning_rate": 0.000183001703563669, + "loss": 0.6066, + "step": 7144 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018299632524059454, + "loss": 0.6371, + "step": 7145 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001829909461458533, + "loss": 0.511, + "step": 7146 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018298556627949534, + "loss": 0.5388, + "step": 7147 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018298018564157065, + "loss": 0.622, + "step": 7148 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018297480423212927, + "loss": 0.5696, + "step": 7149 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018296942205122123, + "loss": 0.5802, + "step": 7150 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018296403909889657, + "loss": 0.62, + "step": 7151 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018295865537520532, + "loss": 0.4932, + "step": 7152 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018295327088019758, + "loss": 0.6478, + "step": 7153 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001829478856139234, + "loss": 0.5417, + "step": 7154 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018294249957643283, + "loss": 0.5959, + "step": 7155 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018293711276777595, + "loss": 0.5593, + "step": 7156 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018293172518800286, + "loss": 0.5027, + "step": 7157 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018292633683716363, + "loss": 0.6304, + "step": 7158 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001829209477153084, + "loss": 0.5042, + "step": 7159 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001829155578224872, + "loss": 0.6959, + "step": 7160 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018291016715875022, + "loss": 0.5901, + "step": 7161 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018290477572414755, + "loss": 0.5225, + "step": 7162 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001828993835187293, + "loss": 0.6763, + "step": 7163 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018289399054254564, + "loss": 0.5267, + "step": 7164 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018288859679564668, + "loss": 0.5893, + "step": 7165 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001828832022780826, + "loss": 0.5206, + "step": 7166 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018287780698990352, + "loss": 0.6315, + "step": 7167 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018287241093115964, + "loss": 0.5901, + "step": 7168 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001828670141019011, + "loss": 0.5825, + "step": 7169 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001828616165021781, + "loss": 0.5753, + "step": 7170 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018285621813204082, + "loss": 0.6711, + "step": 7171 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018285081899153943, + "loss": 0.5321, + "step": 7172 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018284541908072417, + "loss": 0.574, + "step": 7173 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018284001839964523, + "loss": 0.5218, + "step": 7174 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018283461694835275, + "loss": 0.3838, + "step": 7175 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018282921472689706, + "loss": 0.6443, + "step": 7176 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018282381173532838, + "loss": 0.5231, + "step": 7177 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018281840797369686, + "loss": 0.6445, + "step": 7178 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001828130034420528, + "loss": 0.6149, + "step": 7179 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018280759814044646, + "loss": 0.5297, + "step": 7180 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018280219206892807, + "loss": 0.6132, + "step": 7181 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018279678522754785, + "loss": 0.6646, + "step": 7182 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018279137761635617, + "loss": 0.6015, + "step": 7183 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018278596923540326, + "loss": 0.6191, + "step": 7184 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001827805600847394, + "loss": 0.4478, + "step": 7185 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018277515016441488, + "loss": 0.5145, + "step": 7186 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018276973947447997, + "loss": 0.5725, + "step": 7187 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018276432801498507, + "loss": 0.5647, + "step": 7188 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001827589157859804, + "loss": 0.6022, + "step": 7189 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001827535027875163, + "loss": 0.6397, + "step": 7190 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018274808901964313, + "loss": 0.6745, + "step": 7191 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018274267448241123, + "loss": 0.6204, + "step": 7192 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018273725917587093, + "loss": 0.4932, + "step": 7193 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018273184310007255, + "loss": 0.6243, + "step": 7194 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018272642625506645, + "loss": 0.5982, + "step": 7195 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018272100864090302, + "loss": 0.7411, + "step": 7196 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001827155902576326, + "loss": 0.5922, + "step": 7197 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018271017110530564, + "loss": 0.5726, + "step": 7198 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018270475118397242, + "loss": 0.5241, + "step": 7199 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018269933049368345, + "loss": 0.5225, + "step": 7200 + }, + { + "epoch": 0.21, + "learning_rate": 0.000182693909034489, + "loss": 0.5586, + "step": 7201 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018268848680643958, + "loss": 0.4724, + "step": 7202 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018268306380958554, + "loss": 0.588, + "step": 7203 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018267764004397736, + "loss": 0.6738, + "step": 7204 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001826722155096654, + "loss": 0.5274, + "step": 7205 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018266679020670018, + "loss": 0.5046, + "step": 7206 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018266136413513207, + "loss": 0.6222, + "step": 7207 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018265593729501148, + "loss": 0.4998, + "step": 7208 + }, + { + "epoch": 0.21, + "learning_rate": 0.000182650509686389, + "loss": 0.5592, + "step": 7209 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018264508130931502, + "loss": 0.6372, + "step": 7210 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018263965216384, + "loss": 0.5463, + "step": 7211 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001826342222500144, + "loss": 0.5799, + "step": 7212 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001826287915678888, + "loss": 0.6807, + "step": 7213 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018262336011751358, + "loss": 0.585, + "step": 7214 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001826179278989393, + "loss": 0.5129, + "step": 7215 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018261249491221645, + "loss": 0.5962, + "step": 7216 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018260706115739558, + "loss": 0.5987, + "step": 7217 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018260162663452713, + "loss": 0.5087, + "step": 7218 + }, + { + "epoch": 0.21, + "learning_rate": 0.0001825961913436617, + "loss": 0.5894, + "step": 7219 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018259075528484982, + "loss": 0.5387, + "step": 7220 + }, + { + "epoch": 0.21, + "learning_rate": 0.00018258531845814197, + "loss": 0.6346, + "step": 7221 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001825798808635888, + "loss": 0.5079, + "step": 7222 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018257444250124077, + "loss": 0.6131, + "step": 7223 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018256900337114851, + "loss": 0.5895, + "step": 7224 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018256356347336254, + "loss": 0.5491, + "step": 7225 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018255812280793348, + "loss": 0.505, + "step": 7226 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001825526813749119, + "loss": 0.5815, + "step": 7227 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018254723917434838, + "loss": 0.6313, + "step": 7228 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018254179620629354, + "loss": 0.5204, + "step": 7229 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018253635247079798, + "loss": 0.6205, + "step": 7230 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018253090796791228, + "loss": 0.4854, + "step": 7231 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001825254626976871, + "loss": 0.6038, + "step": 7232 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018252001666017309, + "loss": 0.5219, + "step": 7233 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018251456985542085, + "loss": 0.5686, + "step": 7234 + }, + { + "epoch": 0.22, + "learning_rate": 0.000182509122283481, + "loss": 0.5905, + "step": 7235 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018250367394440423, + "loss": 0.5987, + "step": 7236 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001824982248382412, + "loss": 0.6038, + "step": 7237 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018249277496504252, + "loss": 0.665, + "step": 7238 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001824873243248589, + "loss": 0.5246, + "step": 7239 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018248187291774106, + "loss": 0.7015, + "step": 7240 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001824764207437396, + "loss": 0.5291, + "step": 7241 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018247096780290526, + "loss": 0.4852, + "step": 7242 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018246551409528873, + "loss": 0.6159, + "step": 7243 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018246005962094074, + "loss": 0.535, + "step": 7244 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018245460437991195, + "loss": 0.5442, + "step": 7245 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001824491483722531, + "loss": 0.6899, + "step": 7246 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018244369159801496, + "loss": 0.5774, + "step": 7247 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018243823405724824, + "loss": 0.674, + "step": 7248 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018243277575000365, + "loss": 0.5695, + "step": 7249 + }, + { + "epoch": 0.22, + "learning_rate": 0.000182427316676332, + "loss": 0.5179, + "step": 7250 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018242185683628397, + "loss": 0.5375, + "step": 7251 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001824163962299104, + "loss": 0.6083, + "step": 7252 + }, + { + "epoch": 0.22, + "learning_rate": 0.000182410934857262, + "loss": 0.4969, + "step": 7253 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001824054727183896, + "loss": 0.5431, + "step": 7254 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018240000981334392, + "loss": 0.5933, + "step": 7255 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018239454614217584, + "loss": 0.5588, + "step": 7256 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018238908170493606, + "loss": 0.5114, + "step": 7257 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018238361650167547, + "loss": 0.6853, + "step": 7258 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018237815053244482, + "loss": 0.5311, + "step": 7259 + }, + { + "epoch": 0.22, + "learning_rate": 0.000182372683797295, + "loss": 0.5822, + "step": 7260 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018236721629627674, + "loss": 0.5085, + "step": 7261 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018236174802944096, + "loss": 0.5334, + "step": 7262 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001823562789968385, + "loss": 0.6337, + "step": 7263 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018235080919852016, + "loss": 0.523, + "step": 7264 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001823453386345368, + "loss": 0.4707, + "step": 7265 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018233986730493932, + "loss": 0.522, + "step": 7266 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018233439520977856, + "loss": 0.4981, + "step": 7267 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018232892234910542, + "loss": 0.5526, + "step": 7268 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018232344872297078, + "loss": 0.5349, + "step": 7269 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018231797433142553, + "loss": 0.7539, + "step": 7270 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018231249917452052, + "loss": 0.5651, + "step": 7271 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018230702325230675, + "loss": 0.5806, + "step": 7272 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018230154656483506, + "loss": 0.586, + "step": 7273 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018229606911215638, + "loss": 0.6148, + "step": 7274 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001822905908943217, + "loss": 0.5925, + "step": 7275 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018228511191138182, + "loss": 0.6218, + "step": 7276 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001822796321633878, + "loss": 0.5719, + "step": 7277 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018227415165039058, + "loss": 0.4642, + "step": 7278 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018226867037244107, + "loss": 0.5587, + "step": 7279 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018226318832959026, + "loss": 0.6552, + "step": 7280 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001822577055218891, + "loss": 0.496, + "step": 7281 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001822522219493886, + "loss": 0.5932, + "step": 7282 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001822467376121397, + "loss": 0.5037, + "step": 7283 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001822412525101934, + "loss": 0.5999, + "step": 7284 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018223576664360075, + "loss": 0.6123, + "step": 7285 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018223028001241272, + "loss": 0.5534, + "step": 7286 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018222479261668028, + "loss": 0.6739, + "step": 7287 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001822193044564545, + "loss": 0.6113, + "step": 7288 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018221381553178641, + "loss": 0.4926, + "step": 7289 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018220832584272704, + "loss": 0.4942, + "step": 7290 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001822028353893274, + "loss": 0.5636, + "step": 7291 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018219734417163855, + "loss": 0.5767, + "step": 7292 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001821918521897116, + "loss": 0.6514, + "step": 7293 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018218635944359753, + "loss": 0.5586, + "step": 7294 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018218086593334745, + "loss": 0.6244, + "step": 7295 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018217537165901243, + "loss": 0.5915, + "step": 7296 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018216987662064357, + "loss": 0.6339, + "step": 7297 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018216438081829198, + "loss": 0.5247, + "step": 7298 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018215888425200866, + "loss": 0.4674, + "step": 7299 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001821533869218448, + "loss": 0.5849, + "step": 7300 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018214788882785153, + "loss": 0.5769, + "step": 7301 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018214238997007987, + "loss": 0.5318, + "step": 7302 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018213689034858103, + "loss": 0.5238, + "step": 7303 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018213138996340614, + "loss": 0.6653, + "step": 7304 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001821258888146063, + "loss": 0.4694, + "step": 7305 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001821203869022327, + "loss": 0.5197, + "step": 7306 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018211488422633645, + "loss": 0.5285, + "step": 7307 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018210938078696873, + "loss": 0.6677, + "step": 7308 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018210387658418073, + "loss": 0.5913, + "step": 7309 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001820983716180236, + "loss": 0.5658, + "step": 7310 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018209286588854852, + "loss": 0.5256, + "step": 7311 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018208735939580673, + "loss": 0.5354, + "step": 7312 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018208185213984936, + "loss": 0.4973, + "step": 7313 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018207634412072764, + "loss": 0.5807, + "step": 7314 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001820708353384928, + "loss": 0.503, + "step": 7315 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018206532579319605, + "loss": 0.551, + "step": 7316 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001820598154848886, + "loss": 0.7053, + "step": 7317 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018205430441362169, + "loss": 0.5914, + "step": 7318 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018204879257944655, + "loss": 0.6006, + "step": 7319 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018204327998241443, + "loss": 0.5286, + "step": 7320 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018203776662257658, + "loss": 0.5329, + "step": 7321 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018203225249998432, + "loss": 0.5561, + "step": 7322 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018202673761468887, + "loss": 0.559, + "step": 7323 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018202122196674147, + "loss": 0.6079, + "step": 7324 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018201570555619346, + "loss": 0.5382, + "step": 7325 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018201018838309612, + "loss": 0.472, + "step": 7326 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001820046704475007, + "loss": 0.5059, + "step": 7327 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018199915174945856, + "loss": 0.6404, + "step": 7328 + }, + { + "epoch": 0.22, + "learning_rate": 0.000181993632289021, + "loss": 0.5143, + "step": 7329 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018198811206623932, + "loss": 0.748, + "step": 7330 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018198259108116483, + "loss": 0.6795, + "step": 7331 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001819770693338489, + "loss": 0.5864, + "step": 7332 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018197154682434284, + "loss": 0.6114, + "step": 7333 + }, + { + "epoch": 0.22, + "learning_rate": 0.000181966023552698, + "loss": 0.5553, + "step": 7334 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001819604995189658, + "loss": 0.4821, + "step": 7335 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001819549747231975, + "loss": 0.6278, + "step": 7336 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001819494491654445, + "loss": 0.6514, + "step": 7337 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018194392284575817, + "loss": 0.5485, + "step": 7338 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018193839576418995, + "loss": 0.5008, + "step": 7339 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018193286792079114, + "loss": 0.5496, + "step": 7340 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001819273393156132, + "loss": 0.5725, + "step": 7341 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018192180994870753, + "loss": 0.6747, + "step": 7342 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001819162798201255, + "loss": 0.6123, + "step": 7343 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018191074892991854, + "loss": 0.5106, + "step": 7344 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001819052172781381, + "loss": 0.5293, + "step": 7345 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001818996848648356, + "loss": 0.5281, + "step": 7346 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018189415169006247, + "loss": 0.5321, + "step": 7347 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018188861775387017, + "loss": 0.6358, + "step": 7348 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001818830830563101, + "loss": 0.5568, + "step": 7349 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018187754759743378, + "loss": 0.6366, + "step": 7350 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018187201137729263, + "loss": 0.5766, + "step": 7351 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018186647439593818, + "loss": 0.6165, + "step": 7352 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018186093665342187, + "loss": 0.6142, + "step": 7353 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001818553981497952, + "loss": 0.5529, + "step": 7354 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001818498588851097, + "loss": 0.5688, + "step": 7355 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018184431885941678, + "loss": 0.5427, + "step": 7356 + }, + { + "epoch": 0.22, + "learning_rate": 0.000181838778072768, + "loss": 0.581, + "step": 7357 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018183323652521492, + "loss": 0.509, + "step": 7358 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018182769421680898, + "loss": 0.6235, + "step": 7359 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018182215114760178, + "loss": 0.7009, + "step": 7360 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018181660731764483, + "loss": 0.4928, + "step": 7361 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018181106272698968, + "loss": 0.5221, + "step": 7362 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018180551737568786, + "loss": 0.5694, + "step": 7363 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018179997126379092, + "loss": 0.5461, + "step": 7364 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001817944243913505, + "loss": 0.6163, + "step": 7365 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018178887675841812, + "loss": 0.54, + "step": 7366 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018178332836504534, + "loss": 0.5878, + "step": 7367 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018177777921128376, + "loss": 0.6357, + "step": 7368 + }, + { + "epoch": 0.22, + "learning_rate": 0.000181772229297185, + "loss": 0.6299, + "step": 7369 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018176667862280065, + "loss": 0.6111, + "step": 7370 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018176112718818228, + "loss": 0.5229, + "step": 7371 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018175557499338155, + "loss": 0.6499, + "step": 7372 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018175002203845006, + "loss": 0.572, + "step": 7373 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018174446832343945, + "loss": 0.5759, + "step": 7374 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018173891384840136, + "loss": 0.5345, + "step": 7375 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018173335861338744, + "loss": 0.5991, + "step": 7376 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001817278026184493, + "loss": 0.5997, + "step": 7377 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018172224586363863, + "loss": 0.4844, + "step": 7378 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018171668834900711, + "loss": 0.5475, + "step": 7379 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001817111300746064, + "loss": 0.606, + "step": 7380 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018170557104048817, + "loss": 0.5794, + "step": 7381 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018170001124670403, + "loss": 0.5263, + "step": 7382 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018169445069330582, + "loss": 0.6154, + "step": 7383 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018168888938034517, + "loss": 0.6372, + "step": 7384 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018168332730787373, + "loss": 0.5952, + "step": 7385 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018167776447594334, + "loss": 0.61, + "step": 7386 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001816722008846056, + "loss": 0.5107, + "step": 7387 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018166663653391228, + "loss": 0.5986, + "step": 7388 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018166107142391518, + "loss": 0.7035, + "step": 7389 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018165550555466591, + "loss": 0.6564, + "step": 7390 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018164993892621635, + "loss": 0.5636, + "step": 7391 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001816443715386182, + "loss": 0.5323, + "step": 7392 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001816388033919232, + "loss": 0.5866, + "step": 7393 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018163323448618316, + "loss": 0.5641, + "step": 7394 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001816276648214498, + "loss": 0.742, + "step": 7395 + }, + { + "epoch": 0.22, + "learning_rate": 0.000181622094397775, + "loss": 0.6404, + "step": 7396 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018161652321521045, + "loss": 0.5315, + "step": 7397 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018161095127380806, + "loss": 0.5039, + "step": 7398 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001816053785736195, + "loss": 0.4779, + "step": 7399 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001815998051146967, + "loss": 0.5342, + "step": 7400 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018159423089709141, + "loss": 0.5048, + "step": 7401 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001815886559208555, + "loss": 0.5129, + "step": 7402 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001815830801860408, + "loss": 0.5122, + "step": 7403 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001815775036926991, + "loss": 0.6114, + "step": 7404 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001815719264408823, + "loss": 0.5902, + "step": 7405 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018156634843064225, + "loss": 0.5502, + "step": 7406 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018156076966203077, + "loss": 0.5217, + "step": 7407 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001815551901350998, + "loss": 0.5041, + "step": 7408 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018154960984990113, + "loss": 0.4688, + "step": 7409 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018154402880648675, + "loss": 0.5255, + "step": 7410 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018153844700490845, + "loss": 0.5625, + "step": 7411 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001815328644452182, + "loss": 0.5173, + "step": 7412 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001815272811274678, + "loss": 0.5506, + "step": 7413 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001815216970517093, + "loss": 0.6026, + "step": 7414 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018151611221799453, + "loss": 0.6019, + "step": 7415 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018151052662637547, + "loss": 0.5528, + "step": 7416 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018150494027690397, + "loss": 0.5181, + "step": 7417 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018149935316963205, + "loss": 0.5808, + "step": 7418 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018149376530461159, + "loss": 0.5066, + "step": 7419 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018148817668189464, + "loss": 0.6189, + "step": 7420 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018148258730153307, + "loss": 0.5834, + "step": 7421 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001814769971635789, + "loss": 0.4858, + "step": 7422 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018147140626808403, + "loss": 0.5141, + "step": 7423 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018146581461510054, + "loss": 0.4664, + "step": 7424 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018146022220468034, + "loss": 0.57, + "step": 7425 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018145462903687549, + "loss": 0.5915, + "step": 7426 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018144903511173798, + "loss": 0.5759, + "step": 7427 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018144344042931977, + "loss": 0.549, + "step": 7428 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018143784498967294, + "loss": 0.6362, + "step": 7429 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018143224879284945, + "loss": 0.626, + "step": 7430 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001814266518389014, + "loss": 0.5425, + "step": 7431 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018142105412788078, + "loss": 0.4961, + "step": 7432 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018141545565983965, + "loss": 0.4599, + "step": 7433 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018140985643483006, + "loss": 0.6264, + "step": 7434 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018140425645290405, + "loss": 0.6245, + "step": 7435 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018139865571411377, + "loss": 0.6414, + "step": 7436 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018139305421851115, + "loss": 0.5786, + "step": 7437 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001813874519661484, + "loss": 0.5465, + "step": 7438 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018138184895707756, + "loss": 0.4774, + "step": 7439 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018137624519135072, + "loss": 0.5392, + "step": 7440 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018137064066902, + "loss": 0.6424, + "step": 7441 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018136503539013747, + "loss": 0.6403, + "step": 7442 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018135942935475528, + "loss": 0.5871, + "step": 7443 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018135382256292555, + "loss": 0.6023, + "step": 7444 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001813482150147004, + "loss": 0.613, + "step": 7445 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018134260671013195, + "loss": 0.4517, + "step": 7446 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001813369976492724, + "loss": 0.5701, + "step": 7447 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001813313878321738, + "loss": 0.506, + "step": 7448 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001813257772588884, + "loss": 0.418, + "step": 7449 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018132016592946836, + "loss": 0.6103, + "step": 7450 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018131455384396583, + "loss": 0.6048, + "step": 7451 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018130894100243296, + "loss": 0.4372, + "step": 7452 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018130332740492198, + "loss": 0.5519, + "step": 7453 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018129771305148507, + "loss": 0.5369, + "step": 7454 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018129209794217442, + "loss": 0.586, + "step": 7455 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018128648207704222, + "loss": 0.5376, + "step": 7456 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018128086545614073, + "loss": 0.4318, + "step": 7457 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018127524807952215, + "loss": 0.5148, + "step": 7458 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018126962994723868, + "loss": 0.5063, + "step": 7459 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018126401105934261, + "loss": 0.5276, + "step": 7460 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018125839141588617, + "loss": 0.6062, + "step": 7461 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018125277101692158, + "loss": 0.5561, + "step": 7462 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018124714986250107, + "loss": 0.6603, + "step": 7463 + }, + { + "epoch": 0.22, + "learning_rate": 0.000181241527952677, + "loss": 0.524, + "step": 7464 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018123590528750153, + "loss": 0.4872, + "step": 7465 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018123028186702703, + "loss": 0.6753, + "step": 7466 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018122465769130573, + "loss": 0.5714, + "step": 7467 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018121903276038993, + "loss": 0.5839, + "step": 7468 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018121340707433193, + "loss": 0.5068, + "step": 7469 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018120778063318403, + "loss": 0.6143, + "step": 7470 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001812021534369986, + "loss": 0.5864, + "step": 7471 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018119652548582787, + "loss": 0.5803, + "step": 7472 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018119089677972423, + "loss": 0.6309, + "step": 7473 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018118526731873997, + "loss": 0.6791, + "step": 7474 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018117963710292747, + "loss": 0.6447, + "step": 7475 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018117400613233906, + "loss": 0.5313, + "step": 7476 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018116837440702708, + "loss": 0.6227, + "step": 7477 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001811627419270439, + "loss": 0.593, + "step": 7478 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018115710869244192, + "loss": 0.5501, + "step": 7479 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018115147470327346, + "loss": 0.5374, + "step": 7480 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018114583995959096, + "loss": 0.5254, + "step": 7481 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018114020446144676, + "loss": 0.5454, + "step": 7482 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018113456820889327, + "loss": 0.4764, + "step": 7483 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018112893120198292, + "loss": 0.5208, + "step": 7484 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001811232934407681, + "loss": 0.7181, + "step": 7485 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018111765492530125, + "loss": 0.4804, + "step": 7486 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018111201565563473, + "loss": 0.6803, + "step": 7487 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018110637563182106, + "loss": 0.67, + "step": 7488 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018110073485391262, + "loss": 0.4797, + "step": 7489 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018109509332196185, + "loss": 0.5936, + "step": 7490 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018108945103602125, + "loss": 0.6097, + "step": 7491 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018108380799614326, + "loss": 0.5789, + "step": 7492 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001810781642023803, + "loss": 0.5523, + "step": 7493 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018107251965478495, + "loss": 0.5262, + "step": 7494 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018106687435340954, + "loss": 0.5283, + "step": 7495 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001810612282983067, + "loss": 0.405, + "step": 7496 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018105558148952887, + "loss": 0.5402, + "step": 7497 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018104993392712852, + "loss": 0.6335, + "step": 7498 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001810442856111582, + "loss": 0.5464, + "step": 7499 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001810386365416704, + "loss": 0.6122, + "step": 7500 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018103298671871766, + "loss": 0.5562, + "step": 7501 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018102733614235253, + "loss": 0.6172, + "step": 7502 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018102168481262752, + "loss": 0.541, + "step": 7503 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018101603272959515, + "loss": 0.4544, + "step": 7504 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018101037989330803, + "loss": 0.5417, + "step": 7505 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018100472630381867, + "loss": 0.5611, + "step": 7506 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018099907196117966, + "loss": 0.5804, + "step": 7507 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001809934168654435, + "loss": 0.6012, + "step": 7508 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001809877610166629, + "loss": 0.4927, + "step": 7509 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018098210441489036, + "loss": 0.6023, + "step": 7510 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001809764470601785, + "loss": 0.5188, + "step": 7511 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018097078895257989, + "loss": 0.5672, + "step": 7512 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018096513009214715, + "loss": 0.3838, + "step": 7513 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018095947047893293, + "loss": 0.4724, + "step": 7514 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018095381011298977, + "loss": 0.5564, + "step": 7515 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001809481489943704, + "loss": 0.5635, + "step": 7516 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001809424871231274, + "loss": 0.6637, + "step": 7517 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001809368244993134, + "loss": 0.5268, + "step": 7518 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018093116112298104, + "loss": 0.5908, + "step": 7519 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018092549699418302, + "loss": 0.5149, + "step": 7520 + }, + { + "epoch": 0.22, + "learning_rate": 0.000180919832112972, + "loss": 0.6168, + "step": 7521 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018091416647940061, + "loss": 0.6283, + "step": 7522 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018090850009352155, + "loss": 0.5986, + "step": 7523 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001809028329553875, + "loss": 0.5793, + "step": 7524 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018089716506505119, + "loss": 0.5636, + "step": 7525 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018089149642256525, + "loss": 0.5332, + "step": 7526 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018088582702798243, + "loss": 0.544, + "step": 7527 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001808801568813554, + "loss": 0.4345, + "step": 7528 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018087448598273694, + "loss": 0.5332, + "step": 7529 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018086881433217972, + "loss": 0.6092, + "step": 7530 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001808631419297365, + "loss": 0.5935, + "step": 7531 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018085746877546003, + "loss": 0.6343, + "step": 7532 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018085179486940304, + "loss": 0.5693, + "step": 7533 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001808461202116183, + "loss": 0.5507, + "step": 7534 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018084044480215852, + "loss": 0.5593, + "step": 7535 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001808347686410765, + "loss": 0.4659, + "step": 7536 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018082909172842507, + "loss": 0.5943, + "step": 7537 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001808234140642569, + "loss": 0.693, + "step": 7538 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018081773564862486, + "loss": 0.5289, + "step": 7539 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018081205648158176, + "loss": 0.6168, + "step": 7540 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018080637656318034, + "loss": 0.5628, + "step": 7541 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001808006958934734, + "loss": 0.4903, + "step": 7542 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018079501447251385, + "loss": 0.5894, + "step": 7543 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018078933230035444, + "loss": 0.585, + "step": 7544 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018078364937704802, + "loss": 0.5319, + "step": 7545 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018077796570264743, + "loss": 0.529, + "step": 7546 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001807722812772055, + "loss": 0.5094, + "step": 7547 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001807665961007751, + "loss": 0.5181, + "step": 7548 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018076091017340905, + "loss": 0.5188, + "step": 7549 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018075522349516029, + "loss": 0.6473, + "step": 7550 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001807495360660816, + "loss": 0.5711, + "step": 7551 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018074384788622594, + "loss": 0.4963, + "step": 7552 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018073815895564615, + "loss": 0.5464, + "step": 7553 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018073246927439516, + "loss": 0.5905, + "step": 7554 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001807267788425258, + "loss": 0.6349, + "step": 7555 + }, + { + "epoch": 0.22, + "learning_rate": 0.00018072108766009105, + "loss": 0.5776, + "step": 7556 + }, + { + "epoch": 0.22, + "learning_rate": 0.0001807153957271438, + "loss": 0.5056, + "step": 7557 + }, + { + "epoch": 0.23, + "learning_rate": 0.000180709703043737, + "loss": 0.5127, + "step": 7558 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018070400960992352, + "loss": 0.6501, + "step": 7559 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018069831542575634, + "loss": 0.5216, + "step": 7560 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018069262049128838, + "loss": 0.7309, + "step": 7561 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018068692480657257, + "loss": 0.4881, + "step": 7562 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001806812283716619, + "loss": 0.518, + "step": 7563 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018067553118660933, + "loss": 0.4519, + "step": 7564 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018066983325146785, + "loss": 0.5553, + "step": 7565 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018066413456629038, + "loss": 0.587, + "step": 7566 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018065843513112996, + "loss": 0.4571, + "step": 7567 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018065273494603952, + "loss": 0.6181, + "step": 7568 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018064703401107215, + "loss": 0.5429, + "step": 7569 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018064133232628077, + "loss": 0.6505, + "step": 7570 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018063562989171844, + "loss": 0.5994, + "step": 7571 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018062992670743816, + "loss": 0.5001, + "step": 7572 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018062422277349293, + "loss": 0.5642, + "step": 7573 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018061851808993582, + "loss": 0.6242, + "step": 7574 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018061281265681987, + "loss": 0.5401, + "step": 7575 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018060710647419808, + "loss": 0.6577, + "step": 7576 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018060139954212358, + "loss": 0.7513, + "step": 7577 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018059569186064942, + "loss": 0.536, + "step": 7578 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018058998342982858, + "loss": 0.5486, + "step": 7579 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018058427424971424, + "loss": 0.4539, + "step": 7580 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018057856432035938, + "loss": 0.6328, + "step": 7581 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001805728536418172, + "loss": 0.5679, + "step": 7582 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001805671422141407, + "loss": 0.5438, + "step": 7583 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018056143003738307, + "loss": 0.577, + "step": 7584 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018055571711159734, + "loss": 0.5639, + "step": 7585 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018055000343683664, + "loss": 0.5024, + "step": 7586 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018054428901315413, + "loss": 0.5397, + "step": 7587 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018053857384060294, + "loss": 0.6485, + "step": 7588 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018053285791923615, + "loss": 0.5079, + "step": 7589 + }, + { + "epoch": 0.23, + "learning_rate": 0.000180527141249107, + "loss": 0.5997, + "step": 7590 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018052142383026854, + "loss": 0.6513, + "step": 7591 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018051570566277397, + "loss": 0.6267, + "step": 7592 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001805099867466765, + "loss": 0.5531, + "step": 7593 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018050426708202927, + "loss": 0.5307, + "step": 7594 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001804985466688854, + "loss": 0.6095, + "step": 7595 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018049282550729813, + "loss": 0.5846, + "step": 7596 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018048710359732067, + "loss": 0.5724, + "step": 7597 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001804813809390062, + "loss": 0.4902, + "step": 7598 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018047565753240793, + "loss": 0.5168, + "step": 7599 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018046993337757905, + "loss": 0.536, + "step": 7600 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001804642084745728, + "loss": 0.5359, + "step": 7601 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018045848282344245, + "loss": 0.5289, + "step": 7602 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018045275642424117, + "loss": 0.5012, + "step": 7603 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018044702927702221, + "loss": 0.6225, + "step": 7604 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018044130138183885, + "loss": 0.6325, + "step": 7605 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018043557273874435, + "loss": 0.6183, + "step": 7606 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018042984334779193, + "loss": 0.6695, + "step": 7607 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018042411320903489, + "loss": 0.5978, + "step": 7608 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001804183823225265, + "loss": 0.6311, + "step": 7609 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018041265068832003, + "loss": 0.5256, + "step": 7610 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001804069183064688, + "loss": 0.5856, + "step": 7611 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018040118517702607, + "loss": 0.6504, + "step": 7612 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018039545130004518, + "loss": 0.5653, + "step": 7613 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018038971667557943, + "loss": 0.6386, + "step": 7614 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018038398130368214, + "loss": 0.5622, + "step": 7615 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018037824518440663, + "loss": 0.547, + "step": 7616 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018037250831780618, + "loss": 0.6533, + "step": 7617 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018036677070393424, + "loss": 0.5766, + "step": 7618 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018036103234284406, + "loss": 0.4234, + "step": 7619 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018035529323458905, + "loss": 0.5852, + "step": 7620 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018034955337922255, + "loss": 0.558, + "step": 7621 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018034381277679792, + "loss": 0.4763, + "step": 7622 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018033807142736857, + "loss": 0.5657, + "step": 7623 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001803323293309878, + "loss": 0.5698, + "step": 7624 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018032658648770911, + "loss": 0.5124, + "step": 7625 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018032084289758578, + "loss": 0.5003, + "step": 7626 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018031509856067127, + "loss": 0.6395, + "step": 7627 + }, + { + "epoch": 0.23, + "learning_rate": 0.000180309353477019, + "loss": 0.5664, + "step": 7628 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018030360764668239, + "loss": 0.6335, + "step": 7629 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001802978610697148, + "loss": 0.4981, + "step": 7630 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001802921137461697, + "loss": 0.5333, + "step": 7631 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001802863656761006, + "loss": 0.5651, + "step": 7632 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001802806168595608, + "loss": 0.6853, + "step": 7633 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018027486729660385, + "loss": 0.5428, + "step": 7634 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018026911698728316, + "loss": 0.5186, + "step": 7635 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018026336593165225, + "loss": 0.6019, + "step": 7636 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018025761412976453, + "loss": 0.6224, + "step": 7637 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001802518615816735, + "loss": 0.5875, + "step": 7638 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018024610828743263, + "loss": 0.6124, + "step": 7639 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018024035424709544, + "loss": 0.4499, + "step": 7640 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018023459946071542, + "loss": 0.5062, + "step": 7641 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018022884392834607, + "loss": 0.5518, + "step": 7642 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001802230876500409, + "loss": 0.4441, + "step": 7643 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018021733062585345, + "loss": 0.5767, + "step": 7644 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001802115728558372, + "loss": 0.6413, + "step": 7645 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018020581434004574, + "loss": 0.5987, + "step": 7646 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001802000550785326, + "loss": 0.5875, + "step": 7647 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018019429507135128, + "loss": 0.5838, + "step": 7648 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018018853431855537, + "loss": 0.7439, + "step": 7649 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018018277282019846, + "loss": 0.6289, + "step": 7650 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018017701057633406, + "loss": 0.669, + "step": 7651 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018017124758701577, + "loss": 0.4801, + "step": 7652 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018016548385229718, + "loss": 0.531, + "step": 7653 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001801597193722319, + "loss": 0.5805, + "step": 7654 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018015395414687344, + "loss": 0.5167, + "step": 7655 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001801481881762755, + "loss": 0.5901, + "step": 7656 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018014242146049164, + "loss": 0.6273, + "step": 7657 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018013665399957548, + "loss": 0.4313, + "step": 7658 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018013088579358065, + "loss": 0.6515, + "step": 7659 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001801251168425608, + "loss": 0.5129, + "step": 7660 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018011934714656956, + "loss": 0.5503, + "step": 7661 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018011357670566052, + "loss": 0.5612, + "step": 7662 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001801078055198874, + "loss": 0.5945, + "step": 7663 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018010203358930383, + "loss": 0.6381, + "step": 7664 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018009626091396345, + "loss": 0.5593, + "step": 7665 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018009048749391998, + "loss": 0.6063, + "step": 7666 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001800847133292271, + "loss": 0.5834, + "step": 7667 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018007893841993843, + "loss": 0.5832, + "step": 7668 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018007316276610774, + "loss": 0.5977, + "step": 7669 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001800673863677887, + "loss": 0.5743, + "step": 7670 + }, + { + "epoch": 0.23, + "learning_rate": 0.000180061609225035, + "loss": 0.5465, + "step": 7671 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018005583133790036, + "loss": 0.5462, + "step": 7672 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001800500527064385, + "loss": 0.5668, + "step": 7673 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001800442733307032, + "loss": 0.6384, + "step": 7674 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018003849321074813, + "loss": 0.527, + "step": 7675 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018003271234662703, + "loss": 0.6332, + "step": 7676 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018002693073839368, + "loss": 0.6745, + "step": 7677 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001800211483861018, + "loss": 0.6473, + "step": 7678 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018001536528980522, + "loss": 0.6384, + "step": 7679 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018000958144955763, + "loss": 0.5384, + "step": 7680 + }, + { + "epoch": 0.23, + "learning_rate": 0.00018000379686541285, + "loss": 0.6143, + "step": 7681 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017999801153742464, + "loss": 0.5581, + "step": 7682 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017999222546564685, + "loss": 0.407, + "step": 7683 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017998643865013318, + "loss": 0.5631, + "step": 7684 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017998065109093748, + "loss": 0.6815, + "step": 7685 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017997486278811359, + "loss": 0.5647, + "step": 7686 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001799690737417153, + "loss": 0.5249, + "step": 7687 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017996328395179645, + "loss": 0.5645, + "step": 7688 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017995749341841082, + "loss": 0.5447, + "step": 7689 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001799517021416123, + "loss": 0.7087, + "step": 7690 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017994591012145472, + "loss": 0.5826, + "step": 7691 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001799401173579919, + "loss": 0.5208, + "step": 7692 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017993432385127777, + "loss": 0.6399, + "step": 7693 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017992852960136614, + "loss": 0.5874, + "step": 7694 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001799227346083109, + "loss": 0.5667, + "step": 7695 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017991693887216593, + "loss": 0.5688, + "step": 7696 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017991114239298512, + "loss": 0.5989, + "step": 7697 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017990534517082235, + "loss": 0.6074, + "step": 7698 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017989954720573154, + "loss": 0.6518, + "step": 7699 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017989374849776657, + "loss": 0.5482, + "step": 7700 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017988794904698138, + "loss": 0.5352, + "step": 7701 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017988214885342988, + "loss": 0.5078, + "step": 7702 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017987634791716598, + "loss": 0.6122, + "step": 7703 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017987054623824365, + "loss": 0.4835, + "step": 7704 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001798647438167168, + "loss": 0.53, + "step": 7705 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001798589406526394, + "loss": 0.6285, + "step": 7706 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001798531367460654, + "loss": 0.522, + "step": 7707 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017984733209704877, + "loss": 0.544, + "step": 7708 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017984152670564346, + "loss": 0.5503, + "step": 7709 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017983572057190346, + "loss": 0.634, + "step": 7710 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017982991369588274, + "loss": 0.5951, + "step": 7711 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001798241060776353, + "loss": 0.5528, + "step": 7712 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017981829771721516, + "loss": 0.5589, + "step": 7713 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017981248861467629, + "loss": 0.5519, + "step": 7714 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001798066787700727, + "loss": 0.578, + "step": 7715 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001798008681834584, + "loss": 0.632, + "step": 7716 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017979505685488745, + "loss": 0.6403, + "step": 7717 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017978924478441387, + "loss": 0.5674, + "step": 7718 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017978343197209168, + "loss": 0.5557, + "step": 7719 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017977761841797495, + "loss": 0.4687, + "step": 7720 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017977180412211773, + "loss": 0.5716, + "step": 7721 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017976598908457406, + "loss": 0.5981, + "step": 7722 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017976017330539803, + "loss": 0.5572, + "step": 7723 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017975435678464367, + "loss": 0.6117, + "step": 7724 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017974853952236513, + "loss": 0.4986, + "step": 7725 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001797427215186164, + "loss": 0.5604, + "step": 7726 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017973690277345164, + "loss": 0.5332, + "step": 7727 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017973108328692497, + "loss": 0.5046, + "step": 7728 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017972526305909043, + "loss": 0.5271, + "step": 7729 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001797194420900022, + "loss": 0.5598, + "step": 7730 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017971362037971436, + "loss": 0.5088, + "step": 7731 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017970779792828102, + "loss": 0.4323, + "step": 7732 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017970197473575638, + "loss": 0.5524, + "step": 7733 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017969615080219454, + "loss": 0.4775, + "step": 7734 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017969032612764963, + "loss": 0.65, + "step": 7735 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017968450071217587, + "loss": 0.5656, + "step": 7736 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017967867455582733, + "loss": 0.5699, + "step": 7737 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017967284765865827, + "loss": 0.5983, + "step": 7738 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017966702002072282, + "loss": 0.5657, + "step": 7739 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017966119164207516, + "loss": 0.4874, + "step": 7740 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017965536252276948, + "loss": 0.5833, + "step": 7741 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017964953266286, + "loss": 0.5796, + "step": 7742 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017964370206240092, + "loss": 0.5348, + "step": 7743 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017963787072144643, + "loss": 0.5375, + "step": 7744 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017963203864005078, + "loss": 0.5795, + "step": 7745 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017962620581826814, + "loss": 0.538, + "step": 7746 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001796203722561528, + "loss": 0.5918, + "step": 7747 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017961453795375896, + "loss": 0.5072, + "step": 7748 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017960870291114085, + "loss": 0.5642, + "step": 7749 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001796028671283528, + "loss": 0.5052, + "step": 7750 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017959703060544898, + "loss": 0.5597, + "step": 7751 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001795911933424837, + "loss": 0.5171, + "step": 7752 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017958535533951124, + "loss": 0.5315, + "step": 7753 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017957951659658584, + "loss": 0.6067, + "step": 7754 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017957367711376185, + "loss": 0.557, + "step": 7755 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017956783689109353, + "loss": 0.6161, + "step": 7756 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017956199592863515, + "loss": 0.4638, + "step": 7757 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017955615422644104, + "loss": 0.5711, + "step": 7758 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001795503117845655, + "loss": 0.6965, + "step": 7759 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017954446860306292, + "loss": 0.5889, + "step": 7760 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001795386246819875, + "loss": 0.659, + "step": 7761 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001795327800213937, + "loss": 0.4532, + "step": 7762 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001795269346213358, + "loss": 0.5284, + "step": 7763 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017952108848186816, + "loss": 0.5787, + "step": 7764 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017951524160304512, + "loss": 0.5499, + "step": 7765 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001795093939849211, + "loss": 0.4922, + "step": 7766 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017950354562755038, + "loss": 0.4604, + "step": 7767 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001794976965309874, + "loss": 0.5256, + "step": 7768 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017949184669528652, + "loss": 0.5693, + "step": 7769 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017948599612050212, + "loss": 0.5615, + "step": 7770 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001794801448066886, + "loss": 0.6113, + "step": 7771 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017947429275390038, + "loss": 0.5056, + "step": 7772 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017946843996219188, + "loss": 0.5828, + "step": 7773 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017946258643161746, + "loss": 0.6079, + "step": 7774 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017945673216223162, + "loss": 0.589, + "step": 7775 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017945087715408873, + "loss": 0.5427, + "step": 7776 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017944502140724327, + "loss": 0.5613, + "step": 7777 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017943916492174964, + "loss": 0.5754, + "step": 7778 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017943330769766233, + "loss": 0.5491, + "step": 7779 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001794274497350358, + "loss": 0.5484, + "step": 7780 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001794215910339245, + "loss": 0.4832, + "step": 7781 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017941573159438288, + "loss": 0.5916, + "step": 7782 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001794098714164654, + "loss": 0.5203, + "step": 7783 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017940401050022664, + "loss": 0.6151, + "step": 7784 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017939814884572103, + "loss": 0.6121, + "step": 7785 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001793922864530031, + "loss": 0.4939, + "step": 7786 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017938642332212726, + "loss": 0.6087, + "step": 7787 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017938055945314818, + "loss": 0.558, + "step": 7788 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017937469484612025, + "loss": 0.4961, + "step": 7789 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017936882950109803, + "loss": 0.6469, + "step": 7790 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001793629634181361, + "loss": 0.5684, + "step": 7791 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017935709659728898, + "loss": 0.5482, + "step": 7792 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017935122903861114, + "loss": 0.6355, + "step": 7793 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017934536074215725, + "loss": 0.491, + "step": 7794 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001793394917079818, + "loss": 0.5096, + "step": 7795 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017933362193613942, + "loss": 0.5042, + "step": 7796 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001793277514266846, + "loss": 0.4923, + "step": 7797 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017932188017967196, + "loss": 0.5855, + "step": 7798 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017931600819515612, + "loss": 0.5026, + "step": 7799 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001793101354731916, + "loss": 0.5792, + "step": 7800 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001793042620138331, + "loss": 0.4929, + "step": 7801 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017929838781713514, + "loss": 0.518, + "step": 7802 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001792925128831524, + "loss": 0.6123, + "step": 7803 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001792866372119395, + "loss": 0.6334, + "step": 7804 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017928076080355102, + "loss": 0.5831, + "step": 7805 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017927488365804161, + "loss": 0.5447, + "step": 7806 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017926900577546598, + "loss": 0.5885, + "step": 7807 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017926312715587866, + "loss": 0.5511, + "step": 7808 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001792572477993344, + "loss": 0.4828, + "step": 7809 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017925136770588786, + "loss": 0.6113, + "step": 7810 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017924548687559368, + "loss": 0.6337, + "step": 7811 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017923960530850653, + "loss": 0.4747, + "step": 7812 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017923372300468113, + "loss": 0.5364, + "step": 7813 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017922783996417217, + "loss": 0.556, + "step": 7814 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001792219561870343, + "loss": 0.6262, + "step": 7815 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017921607167332225, + "loss": 0.5637, + "step": 7816 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017921018642309076, + "loss": 0.5199, + "step": 7817 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017920430043639454, + "loss": 0.5274, + "step": 7818 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017919841371328828, + "loss": 0.5619, + "step": 7819 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017919252625382673, + "loss": 0.5621, + "step": 7820 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017918663805806465, + "loss": 0.5143, + "step": 7821 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017918074912605674, + "loss": 0.541, + "step": 7822 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017917485945785782, + "loss": 0.4856, + "step": 7823 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001791689690535226, + "loss": 0.4653, + "step": 7824 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017916307791310585, + "loss": 0.5323, + "step": 7825 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017915718603666236, + "loss": 0.6479, + "step": 7826 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017915129342424687, + "loss": 0.6399, + "step": 7827 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017914540007591426, + "loss": 0.499, + "step": 7828 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017913950599171923, + "loss": 0.3553, + "step": 7829 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001791336111717166, + "loss": 0.5345, + "step": 7830 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017912771561596124, + "loss": 0.5858, + "step": 7831 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001791218193245079, + "loss": 0.4467, + "step": 7832 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017911592229741138, + "loss": 0.6223, + "step": 7833 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017911002453472658, + "loss": 0.5025, + "step": 7834 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017910412603650832, + "loss": 0.5877, + "step": 7835 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001790982268028114, + "loss": 0.6223, + "step": 7836 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017909232683369067, + "loss": 0.5476, + "step": 7837 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017908642612920106, + "loss": 0.4946, + "step": 7838 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017908052468939734, + "loss": 0.4535, + "step": 7839 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017907462251433445, + "loss": 0.6235, + "step": 7840 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017906871960406723, + "loss": 0.4879, + "step": 7841 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017906281595865054, + "loss": 0.6868, + "step": 7842 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017905691157813934, + "loss": 0.6504, + "step": 7843 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017905100646258844, + "loss": 0.6165, + "step": 7844 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001790451006120528, + "loss": 0.4882, + "step": 7845 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017903919402658734, + "loss": 0.5938, + "step": 7846 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017903328670624695, + "loss": 0.4519, + "step": 7847 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017902737865108657, + "loss": 0.6495, + "step": 7848 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001790214698611611, + "loss": 0.4492, + "step": 7849 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001790155603365255, + "loss": 0.5829, + "step": 7850 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017900965007723474, + "loss": 0.5774, + "step": 7851 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017900373908334371, + "loss": 0.5853, + "step": 7852 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001789978273549074, + "loss": 0.53, + "step": 7853 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017899191489198079, + "loss": 0.5631, + "step": 7854 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017898600169461885, + "loss": 0.514, + "step": 7855 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017898008776287653, + "loss": 0.4287, + "step": 7856 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017897417309680882, + "loss": 0.4561, + "step": 7857 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017896825769647077, + "loss": 0.6755, + "step": 7858 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001789623415619173, + "loss": 0.4514, + "step": 7859 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017895642469320343, + "loss": 0.5671, + "step": 7860 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017895050709038424, + "loss": 0.6722, + "step": 7861 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017894458875351466, + "loss": 0.6054, + "step": 7862 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001789386696826498, + "loss": 0.7077, + "step": 7863 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017893274987784463, + "loss": 0.5115, + "step": 7864 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001789268293391542, + "loss": 0.5468, + "step": 7865 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017892090806663358, + "loss": 0.6862, + "step": 7866 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001789149860603378, + "loss": 0.5477, + "step": 7867 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017890906332032196, + "loss": 0.6233, + "step": 7868 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017890313984664105, + "loss": 0.5335, + "step": 7869 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017889721563935023, + "loss": 0.5295, + "step": 7870 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017889129069850453, + "loss": 0.5568, + "step": 7871 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017888536502415906, + "loss": 0.485, + "step": 7872 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001788794386163689, + "loss": 0.4788, + "step": 7873 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017887351147518918, + "loss": 0.5984, + "step": 7874 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017886758360067498, + "loss": 0.5392, + "step": 7875 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017886165499288138, + "loss": 0.5335, + "step": 7876 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017885572565186358, + "loss": 0.5158, + "step": 7877 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017884979557767668, + "loss": 0.5428, + "step": 7878 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017884386477037581, + "loss": 0.4979, + "step": 7879 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017883793323001607, + "loss": 0.5194, + "step": 7880 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001788320009566527, + "loss": 0.5287, + "step": 7881 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017882606795034076, + "loss": 0.5389, + "step": 7882 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017882013421113545, + "loss": 0.5821, + "step": 7883 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017881419973909198, + "loss": 0.553, + "step": 7884 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017880826453426548, + "loss": 0.6096, + "step": 7885 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017880232859671116, + "loss": 0.623, + "step": 7886 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017879639192648416, + "loss": 0.5864, + "step": 7887 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017879045452363977, + "loss": 0.6132, + "step": 7888 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001787845163882331, + "loss": 0.4832, + "step": 7889 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001787785775203194, + "loss": 0.5494, + "step": 7890 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017877263791995387, + "loss": 0.4905, + "step": 7891 + }, + { + "epoch": 0.23, + "learning_rate": 0.00017876669758719178, + "loss": 0.567, + "step": 7892 + }, + { + "epoch": 0.23, + "learning_rate": 0.0001787607565220883, + "loss": 0.5928, + "step": 7893 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017875481472469872, + "loss": 0.6507, + "step": 7894 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017874887219507826, + "loss": 0.6948, + "step": 7895 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017874292893328215, + "loss": 0.545, + "step": 7896 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001787369849393657, + "loss": 0.5485, + "step": 7897 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017873104021338412, + "loss": 0.5109, + "step": 7898 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001787250947553927, + "loss": 0.5035, + "step": 7899 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017871914856544676, + "loss": 0.5804, + "step": 7900 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017871320164360152, + "loss": 0.6229, + "step": 7901 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017870725398991234, + "loss": 0.6171, + "step": 7902 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017870130560443444, + "loss": 0.5961, + "step": 7903 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001786953564872232, + "loss": 0.5259, + "step": 7904 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017868940663833387, + "loss": 0.648, + "step": 7905 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017868345605782183, + "loss": 0.5548, + "step": 7906 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017867750474574235, + "loss": 0.5227, + "step": 7907 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001786715527021508, + "loss": 0.597, + "step": 7908 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017866559992710251, + "loss": 0.5708, + "step": 7909 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017865964642065283, + "loss": 0.5947, + "step": 7910 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001786536921828571, + "loss": 0.5512, + "step": 7911 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001786477372137707, + "loss": 0.616, + "step": 7912 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017864178151344895, + "loss": 0.5495, + "step": 7913 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001786358250819473, + "loss": 0.5376, + "step": 7914 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017862986791932105, + "loss": 0.6408, + "step": 7915 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001786239100256257, + "loss": 0.5436, + "step": 7916 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001786179514009165, + "loss": 0.6077, + "step": 7917 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017861199204524894, + "loss": 0.565, + "step": 7918 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017860603195867842, + "loss": 0.5323, + "step": 7919 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001786000711412603, + "loss": 0.5828, + "step": 7920 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017859410959305012, + "loss": 0.5152, + "step": 7921 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001785881473141032, + "loss": 0.5807, + "step": 7922 + }, + { + "epoch": 0.24, + "learning_rate": 0.000178582184304475, + "loss": 0.541, + "step": 7923 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017857622056422097, + "loss": 0.6507, + "step": 7924 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017857025609339656, + "loss": 0.6328, + "step": 7925 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017856429089205725, + "loss": 0.7023, + "step": 7926 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017855832496025843, + "loss": 0.5156, + "step": 7927 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017855235829805562, + "loss": 0.5531, + "step": 7928 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017854639090550433, + "loss": 0.6575, + "step": 7929 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017854042278266002, + "loss": 0.5934, + "step": 7930 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017853445392957808, + "loss": 0.5809, + "step": 7931 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017852848434631415, + "loss": 0.6091, + "step": 7932 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017852251403292364, + "loss": 0.5462, + "step": 7933 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001785165429894621, + "loss": 0.5178, + "step": 7934 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017851057121598505, + "loss": 0.4805, + "step": 7935 + }, + { + "epoch": 0.24, + "learning_rate": 0.000178504598712548, + "loss": 0.5613, + "step": 7936 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017849862547920648, + "loss": 0.6687, + "step": 7937 + }, + { + "epoch": 0.24, + "learning_rate": 0.000178492651516016, + "loss": 0.556, + "step": 7938 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017848667682303216, + "loss": 0.5581, + "step": 7939 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017848070140031046, + "loss": 0.5327, + "step": 7940 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017847472524790646, + "loss": 0.6096, + "step": 7941 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001784687483658758, + "loss": 0.526, + "step": 7942 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017846277075427396, + "loss": 0.5873, + "step": 7943 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017845679241315654, + "loss": 0.532, + "step": 7944 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017845081334257913, + "loss": 0.5714, + "step": 7945 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017844483354259735, + "loss": 0.6239, + "step": 7946 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017843885301326674, + "loss": 0.6207, + "step": 7947 + }, + { + "epoch": 0.24, + "learning_rate": 0.000178432871754643, + "loss": 0.5184, + "step": 7948 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001784268897667816, + "loss": 0.62, + "step": 7949 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001784209070497383, + "loss": 0.5811, + "step": 7950 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017841492360356863, + "loss": 0.6035, + "step": 7951 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017840893942832825, + "loss": 0.5616, + "step": 7952 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001784029545240728, + "loss": 0.493, + "step": 7953 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017839696889085794, + "loss": 0.6188, + "step": 7954 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001783909825287393, + "loss": 0.6191, + "step": 7955 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001783849954377726, + "loss": 0.4913, + "step": 7956 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001783790076180134, + "loss": 0.5679, + "step": 7957 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017837301906951743, + "loss": 0.5297, + "step": 7958 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017836702979234036, + "loss": 0.4953, + "step": 7959 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001783610397865379, + "loss": 0.7162, + "step": 7960 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001783550490521657, + "loss": 0.4994, + "step": 7961 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001783490575892795, + "loss": 0.5696, + "step": 7962 + }, + { + "epoch": 0.24, + "learning_rate": 0.000178343065397935, + "loss": 0.508, + "step": 7963 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017833707247818788, + "loss": 0.5688, + "step": 7964 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001783310788300939, + "loss": 0.4102, + "step": 7965 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017832508445370877, + "loss": 0.4937, + "step": 7966 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001783190893490882, + "loss": 0.4485, + "step": 7967 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017831309351628796, + "loss": 0.5794, + "step": 7968 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001783070969553638, + "loss": 0.5334, + "step": 7969 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017830109966637144, + "loss": 0.5684, + "step": 7970 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017829510164936669, + "loss": 0.5068, + "step": 7971 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017828910290440528, + "loss": 0.6735, + "step": 7972 + }, + { + "epoch": 0.24, + "learning_rate": 0.000178283103431543, + "loss": 0.6273, + "step": 7973 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001782771032308356, + "loss": 0.534, + "step": 7974 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001782711023023389, + "loss": 0.5412, + "step": 7975 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017826510064610873, + "loss": 0.4999, + "step": 7976 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001782590982622008, + "loss": 0.5416, + "step": 7977 + }, + { + "epoch": 0.24, + "learning_rate": 0.000178253095150671, + "loss": 0.5138, + "step": 7978 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017824709131157506, + "loss": 0.6024, + "step": 7979 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017824108674496892, + "loss": 0.6293, + "step": 7980 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017823508145090825, + "loss": 0.6003, + "step": 7981 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017822907542944905, + "loss": 0.5886, + "step": 7982 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017822306868064706, + "loss": 0.681, + "step": 7983 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017821706120455818, + "loss": 0.6048, + "step": 7984 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001782110530012382, + "loss": 0.5645, + "step": 7985 + }, + { + "epoch": 0.24, + "learning_rate": 0.000178205044070743, + "loss": 0.6142, + "step": 7986 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017819903441312852, + "loss": 0.5346, + "step": 7987 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017819302402845058, + "loss": 0.5549, + "step": 7988 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017818701291676505, + "loss": 0.5083, + "step": 7989 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017818100107812784, + "loss": 0.558, + "step": 7990 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017817498851259485, + "loss": 0.611, + "step": 7991 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017816897522022197, + "loss": 0.5552, + "step": 7992 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017816296120106513, + "loss": 0.583, + "step": 7993 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017815694645518019, + "loss": 0.6345, + "step": 7994 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017815093098262315, + "loss": 0.5657, + "step": 7995 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017814491478344987, + "loss": 0.589, + "step": 7996 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017813889785771634, + "loss": 0.5224, + "step": 7997 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017813288020547847, + "loss": 0.6686, + "step": 7998 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017812686182679222, + "loss": 0.5448, + "step": 7999 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017812084272171355, + "loss": 0.5631, + "step": 8000 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001781148228902984, + "loss": 0.5519, + "step": 8001 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017810880233260282, + "loss": 0.7335, + "step": 8002 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001781027810486827, + "loss": 0.5566, + "step": 8003 + }, + { + "epoch": 0.24, + "learning_rate": 0.000178096759038594, + "loss": 0.5876, + "step": 8004 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017809073630239283, + "loss": 0.6109, + "step": 8005 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017808471284013508, + "loss": 0.5603, + "step": 8006 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017807868865187681, + "loss": 0.5322, + "step": 8007 + }, + { + "epoch": 0.24, + "learning_rate": 0.000178072663737674, + "loss": 0.5373, + "step": 8008 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017806663809758267, + "loss": 0.5481, + "step": 8009 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017806061173165886, + "loss": 0.502, + "step": 8010 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017805458463995862, + "loss": 0.5555, + "step": 8011 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017804855682253793, + "loss": 0.5197, + "step": 8012 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017804252827945287, + "loss": 0.5053, + "step": 8013 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017803649901075949, + "loss": 0.4706, + "step": 8014 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017803046901651386, + "loss": 0.5579, + "step": 8015 + }, + { + "epoch": 0.24, + "learning_rate": 0.000178024438296772, + "loss": 0.5749, + "step": 8016 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017801840685159004, + "loss": 0.5504, + "step": 8017 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017801237468102402, + "loss": 0.5643, + "step": 8018 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017800634178513, + "loss": 0.5825, + "step": 8019 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017800030816396414, + "loss": 0.5411, + "step": 8020 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017799427381758252, + "loss": 0.5322, + "step": 8021 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017798823874604118, + "loss": 0.5525, + "step": 8022 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001779822029493963, + "loss": 0.6517, + "step": 8023 + }, + { + "epoch": 0.24, + "learning_rate": 0.000177976166427704, + "loss": 0.6001, + "step": 8024 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017797012918102035, + "loss": 0.5462, + "step": 8025 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017796409120940151, + "loss": 0.5782, + "step": 8026 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017795805251290365, + "loss": 0.6595, + "step": 8027 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017795201309158288, + "loss": 0.6531, + "step": 8028 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017794597294549535, + "loss": 0.4755, + "step": 8029 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017793993207469724, + "loss": 0.5827, + "step": 8030 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001779338904792447, + "loss": 0.5576, + "step": 8031 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017792784815919393, + "loss": 0.5335, + "step": 8032 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017792180511460107, + "loss": 0.569, + "step": 8033 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017791576134552233, + "loss": 0.5197, + "step": 8034 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017790971685201386, + "loss": 0.4739, + "step": 8035 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017790367163413192, + "loss": 0.5726, + "step": 8036 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001778976256919327, + "loss": 0.5348, + "step": 8037 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001778915790254724, + "loss": 0.5132, + "step": 8038 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017788553163480724, + "loss": 0.6343, + "step": 8039 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017787948351999345, + "loss": 0.5696, + "step": 8040 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017787343468108726, + "loss": 0.4947, + "step": 8041 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001778673851181449, + "loss": 0.5376, + "step": 8042 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017786133483122261, + "loss": 0.5332, + "step": 8043 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017785528382037672, + "loss": 0.5491, + "step": 8044 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017784923208566337, + "loss": 0.4662, + "step": 8045 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001778431796271389, + "loss": 0.7596, + "step": 8046 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017783712644485963, + "loss": 0.5368, + "step": 8047 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017783107253888173, + "loss": 0.6017, + "step": 8048 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017782501790926153, + "loss": 0.5135, + "step": 8049 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017781896255605537, + "loss": 0.5317, + "step": 8050 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017781290647931946, + "loss": 0.5679, + "step": 8051 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017780684967911024, + "loss": 0.5784, + "step": 8052 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017780079215548388, + "loss": 0.5616, + "step": 8053 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017779473390849677, + "loss": 0.5439, + "step": 8054 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017778867493820525, + "loss": 0.5258, + "step": 8055 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001777826152446656, + "loss": 0.5348, + "step": 8056 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001777765548279342, + "loss": 0.5544, + "step": 8057 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017777049368806742, + "loss": 0.6048, + "step": 8058 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017776443182512157, + "loss": 0.5943, + "step": 8059 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017775836923915303, + "loss": 0.6019, + "step": 8060 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017775230593021816, + "loss": 0.4702, + "step": 8061 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017774624189837335, + "loss": 0.5029, + "step": 8062 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017774017714367493, + "loss": 0.5658, + "step": 8063 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001777341116661794, + "loss": 0.528, + "step": 8064 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017772804546594304, + "loss": 0.5149, + "step": 8065 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001777219785430223, + "loss": 0.5533, + "step": 8066 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017771591089747355, + "loss": 0.629, + "step": 8067 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017770984252935327, + "loss": 0.5592, + "step": 8068 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017770377343871783, + "loss": 0.5763, + "step": 8069 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017769770362562366, + "loss": 0.6301, + "step": 8070 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017769163309012723, + "loss": 0.5612, + "step": 8071 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017768556183228495, + "loss": 0.7009, + "step": 8072 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001776794898521533, + "loss": 0.5854, + "step": 8073 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001776734171497887, + "loss": 0.4451, + "step": 8074 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017766734372524762, + "loss": 0.6002, + "step": 8075 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017766126957858652, + "loss": 0.5749, + "step": 8076 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017765519470986193, + "loss": 0.4915, + "step": 8077 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017764911911913024, + "loss": 0.5567, + "step": 8078 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017764304280644804, + "loss": 0.5712, + "step": 8079 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017763696577187171, + "loss": 0.5855, + "step": 8080 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017763088801545787, + "loss": 0.5441, + "step": 8081 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017762480953726294, + "loss": 0.5082, + "step": 8082 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017761873033734352, + "loss": 0.6714, + "step": 8083 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017761265041575604, + "loss": 0.5079, + "step": 8084 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017760656977255706, + "loss": 0.5256, + "step": 8085 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017760048840780314, + "loss": 0.5392, + "step": 8086 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001775944063215508, + "loss": 0.5452, + "step": 8087 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001775883235138566, + "loss": 0.6519, + "step": 8088 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001775822399847771, + "loss": 0.5693, + "step": 8089 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017757615573436885, + "loss": 0.6155, + "step": 8090 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017757007076268843, + "loss": 0.5404, + "step": 8091 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001775639850697924, + "loss": 0.4899, + "step": 8092 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017755789865573734, + "loss": 0.5586, + "step": 8093 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017755181152057987, + "loss": 0.5624, + "step": 8094 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017754572366437656, + "loss": 0.4121, + "step": 8095 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017753963508718404, + "loss": 0.5036, + "step": 8096 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017753354578905884, + "loss": 0.5351, + "step": 8097 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017752745577005768, + "loss": 0.6235, + "step": 8098 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001775213650302371, + "loss": 0.6636, + "step": 8099 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001775152735696538, + "loss": 0.7344, + "step": 8100 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017750918138836435, + "loss": 0.5194, + "step": 8101 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017750308848642546, + "loss": 0.5832, + "step": 8102 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017749699486389374, + "loss": 0.5014, + "step": 8103 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017749090052082583, + "loss": 0.5008, + "step": 8104 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001774848054572784, + "loss": 0.6332, + "step": 8105 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017747870967330814, + "loss": 0.5116, + "step": 8106 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017747261316897173, + "loss": 0.5681, + "step": 8107 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017746651594432585, + "loss": 0.5828, + "step": 8108 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017746041799942714, + "loss": 0.5478, + "step": 8109 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017745431933433237, + "loss": 0.5709, + "step": 8110 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017744821994909819, + "loss": 0.6372, + "step": 8111 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017744211984378131, + "loss": 0.5277, + "step": 8112 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017743601901843852, + "loss": 0.5247, + "step": 8113 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017742991747312647, + "loss": 0.5128, + "step": 8114 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017742381520790187, + "loss": 0.671, + "step": 8115 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017741771222282154, + "loss": 0.5452, + "step": 8116 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017741160851794216, + "loss": 0.5764, + "step": 8117 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001774055040933205, + "loss": 0.5523, + "step": 8118 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001773993989490133, + "loss": 0.5208, + "step": 8119 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017739329308507734, + "loss": 0.5274, + "step": 8120 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001773871865015694, + "loss": 0.5361, + "step": 8121 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017738107919854624, + "loss": 0.6046, + "step": 8122 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017737497117606464, + "loss": 0.5646, + "step": 8123 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001773688624341814, + "loss": 0.5311, + "step": 8124 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001773627529729533, + "loss": 0.5696, + "step": 8125 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001773566427924372, + "loss": 0.4082, + "step": 8126 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017735053189268983, + "loss": 0.5344, + "step": 8127 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017734442027376806, + "loss": 0.5186, + "step": 8128 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017733830793572868, + "loss": 0.6001, + "step": 8129 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017733219487862852, + "loss": 0.5221, + "step": 8130 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001773260811025245, + "loss": 0.6402, + "step": 8131 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017731996660747334, + "loss": 0.552, + "step": 8132 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017731385139353197, + "loss": 0.5433, + "step": 8133 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001773077354607572, + "loss": 0.6006, + "step": 8134 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017730161880920596, + "loss": 0.6099, + "step": 8135 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017729550143893503, + "loss": 0.5226, + "step": 8136 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017728938335000136, + "loss": 0.5449, + "step": 8137 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001772832645424618, + "loss": 0.5936, + "step": 8138 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017727714501637325, + "loss": 0.5099, + "step": 8139 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017727102477179264, + "loss": 0.6127, + "step": 8140 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001772649038087768, + "loss": 0.6088, + "step": 8141 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017725878212738265, + "loss": 0.6462, + "step": 8142 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001772526597276672, + "loss": 0.5857, + "step": 8143 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017724653660968726, + "loss": 0.5255, + "step": 8144 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017724041277349982, + "loss": 0.563, + "step": 8145 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017723428821916183, + "loss": 0.6229, + "step": 8146 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001772281629467302, + "loss": 0.6236, + "step": 8147 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017722203695626188, + "loss": 0.5867, + "step": 8148 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017721591024781385, + "loss": 0.5569, + "step": 8149 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017720978282144306, + "loss": 0.4964, + "step": 8150 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001772036546772065, + "loss": 0.6197, + "step": 8151 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001771975258151611, + "loss": 0.5487, + "step": 8152 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001771913962353639, + "loss": 0.5404, + "step": 8153 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017718526593787183, + "loss": 0.565, + "step": 8154 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017717913492274194, + "loss": 0.521, + "step": 8155 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001771730031900312, + "loss": 0.5513, + "step": 8156 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017716687073979665, + "loss": 0.4556, + "step": 8157 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001771607375720953, + "loss": 0.6608, + "step": 8158 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017715460368698414, + "loss": 0.6111, + "step": 8159 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017714846908452024, + "loss": 0.5852, + "step": 8160 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017714233376476063, + "loss": 0.6204, + "step": 8161 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017713619772776235, + "loss": 0.5817, + "step": 8162 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017713006097358242, + "loss": 0.5849, + "step": 8163 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017712392350227796, + "loss": 0.5597, + "step": 8164 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017711778531390596, + "loss": 0.5241, + "step": 8165 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017711164640852353, + "loss": 0.5714, + "step": 8166 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017710550678618777, + "loss": 0.4442, + "step": 8167 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017709936644695574, + "loss": 0.5463, + "step": 8168 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017709322539088452, + "loss": 0.5449, + "step": 8169 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001770870836180312, + "loss": 0.5747, + "step": 8170 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017708094112845295, + "loss": 0.509, + "step": 8171 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001770747979222068, + "loss": 0.4771, + "step": 8172 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017706865399934985, + "loss": 0.59, + "step": 8173 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017706250935993934, + "loss": 0.5453, + "step": 8174 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017705636400403233, + "loss": 0.4494, + "step": 8175 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017705021793168593, + "loss": 0.6128, + "step": 8176 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001770440711429573, + "loss": 0.5507, + "step": 8177 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017703792363790363, + "loss": 0.6429, + "step": 8178 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017703177541658204, + "loss": 0.5896, + "step": 8179 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001770256264790497, + "loss": 0.6471, + "step": 8180 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017701947682536383, + "loss": 0.5554, + "step": 8181 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017701332645558151, + "loss": 0.5457, + "step": 8182 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017700717536975997, + "loss": 0.4273, + "step": 8183 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017700102356795646, + "loss": 0.4696, + "step": 8184 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017699487105022805, + "loss": 0.5603, + "step": 8185 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017698871781663206, + "loss": 0.5627, + "step": 8186 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017698256386722565, + "loss": 0.5645, + "step": 8187 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017697640920206605, + "loss": 0.6155, + "step": 8188 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017697025382121045, + "loss": 0.6937, + "step": 8189 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017696409772471615, + "loss": 0.5139, + "step": 8190 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017695794091264033, + "loss": 0.5421, + "step": 8191 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017695178338504022, + "loss": 0.6025, + "step": 8192 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017694562514197312, + "loss": 0.5695, + "step": 8193 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017693946618349627, + "loss": 0.5056, + "step": 8194 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001769333065096669, + "loss": 0.6671, + "step": 8195 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017692714612054235, + "loss": 0.6201, + "step": 8196 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017692098501617985, + "loss": 0.621, + "step": 8197 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017691482319663666, + "loss": 0.5726, + "step": 8198 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001769086606619701, + "loss": 0.5657, + "step": 8199 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017690249741223748, + "loss": 0.6044, + "step": 8200 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001768963334474961, + "loss": 0.6555, + "step": 8201 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017689016876780325, + "loss": 0.5179, + "step": 8202 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017688400337321624, + "loss": 0.5407, + "step": 8203 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017687783726379243, + "loss": 0.737, + "step": 8204 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017687167043958912, + "loss": 0.5472, + "step": 8205 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017686550290066367, + "loss": 0.6045, + "step": 8206 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001768593346470734, + "loss": 0.6218, + "step": 8207 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017685316567887566, + "loss": 0.6671, + "step": 8208 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001768469959961278, + "loss": 0.5354, + "step": 8209 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017684082559888724, + "loss": 0.5798, + "step": 8210 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017683465448721132, + "loss": 0.6195, + "step": 8211 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017682848266115735, + "loss": 0.5274, + "step": 8212 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001768223101207828, + "loss": 0.6485, + "step": 8213 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017681613686614504, + "loss": 0.4538, + "step": 8214 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017680996289730144, + "loss": 0.5993, + "step": 8215 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017680378821430942, + "loss": 0.453, + "step": 8216 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001767976128172264, + "loss": 0.4614, + "step": 8217 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017679143670610976, + "loss": 0.5576, + "step": 8218 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017678525988101696, + "loss": 0.6264, + "step": 8219 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017677908234200543, + "loss": 0.5569, + "step": 8220 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017677290408913256, + "loss": 0.5589, + "step": 8221 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017676672512245586, + "loss": 0.5011, + "step": 8222 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017676054544203272, + "loss": 0.5344, + "step": 8223 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017675436504792065, + "loss": 0.5132, + "step": 8224 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001767481839401771, + "loss": 0.5315, + "step": 8225 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017674200211885947, + "loss": 0.5393, + "step": 8226 + }, + { + "epoch": 0.24, + "learning_rate": 0.00017673581958402532, + "loss": 0.687, + "step": 8227 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001767296363357321, + "loss": 0.4993, + "step": 8228 + }, + { + "epoch": 0.24, + "learning_rate": 0.0001767234523740373, + "loss": 0.5594, + "step": 8229 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017671726769899842, + "loss": 0.5972, + "step": 8230 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017671108231067298, + "loss": 0.52, + "step": 8231 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017670489620911844, + "loss": 0.5498, + "step": 8232 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017669870939439236, + "loss": 0.5687, + "step": 8233 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017669252186655225, + "loss": 0.5451, + "step": 8234 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017668633362565562, + "loss": 0.56, + "step": 8235 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017668014467176005, + "loss": 0.5867, + "step": 8236 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017667395500492306, + "loss": 0.6254, + "step": 8237 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001766677646252022, + "loss": 0.6461, + "step": 8238 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017666157353265498, + "loss": 0.5902, + "step": 8239 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017665538172733904, + "loss": 0.5586, + "step": 8240 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017664918920931187, + "loss": 0.5924, + "step": 8241 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017664299597863113, + "loss": 0.5913, + "step": 8242 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017663680203535437, + "loss": 0.5127, + "step": 8243 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017663060737953913, + "loss": 0.5894, + "step": 8244 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017662441201124306, + "loss": 0.5046, + "step": 8245 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017661821593052376, + "loss": 0.6085, + "step": 8246 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017661201913743882, + "loss": 0.6527, + "step": 8247 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017660582163204585, + "loss": 0.578, + "step": 8248 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001765996234144025, + "loss": 0.4535, + "step": 8249 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017659342448456637, + "loss": 0.5071, + "step": 8250 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001765872248425951, + "loss": 0.5793, + "step": 8251 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017658102448854637, + "loss": 0.6693, + "step": 8252 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017657482342247775, + "loss": 0.5832, + "step": 8253 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017656862164444696, + "loss": 0.5293, + "step": 8254 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017656241915451167, + "loss": 0.6908, + "step": 8255 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017655621595272946, + "loss": 0.4943, + "step": 8256 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017655001203915812, + "loss": 0.6143, + "step": 8257 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017654380741385526, + "loss": 0.6138, + "step": 8258 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017653760207687858, + "loss": 0.5995, + "step": 8259 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001765313960282858, + "loss": 0.6126, + "step": 8260 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001765251892681346, + "loss": 0.6159, + "step": 8261 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017651898179648265, + "loss": 0.4803, + "step": 8262 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017651277361338777, + "loss": 0.4147, + "step": 8263 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017650656471890756, + "loss": 0.5286, + "step": 8264 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017650035511309982, + "loss": 0.5896, + "step": 8265 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017649414479602225, + "loss": 0.573, + "step": 8266 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017648793376773267, + "loss": 0.5707, + "step": 8267 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017648172202828873, + "loss": 0.5951, + "step": 8268 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001764755095777482, + "loss": 0.5524, + "step": 8269 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001764692964161689, + "loss": 0.5508, + "step": 8270 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017646308254360853, + "loss": 0.6902, + "step": 8271 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001764568679601249, + "loss": 0.4048, + "step": 8272 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001764506526657758, + "loss": 0.6898, + "step": 8273 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017644443666061898, + "loss": 0.5222, + "step": 8274 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017643821994471227, + "loss": 0.6379, + "step": 8275 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017643200251811348, + "loss": 0.5501, + "step": 8276 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017642578438088036, + "loss": 0.4846, + "step": 8277 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017641956553307076, + "loss": 0.5986, + "step": 8278 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017641334597474247, + "loss": 0.5282, + "step": 8279 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017640712570595338, + "loss": 0.5983, + "step": 8280 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001764009047267613, + "loss": 0.5204, + "step": 8281 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017639468303722405, + "loss": 0.6074, + "step": 8282 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017638846063739946, + "loss": 0.6203, + "step": 8283 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017638223752734544, + "loss": 0.5715, + "step": 8284 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001763760137071198, + "loss": 0.6181, + "step": 8285 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001763697891767804, + "loss": 0.6089, + "step": 8286 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001763635639363852, + "loss": 0.6374, + "step": 8287 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017635733798599196, + "loss": 0.5253, + "step": 8288 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017635111132565867, + "loss": 0.6224, + "step": 8289 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017634488395544315, + "loss": 0.6393, + "step": 8290 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017633865587540334, + "loss": 0.5863, + "step": 8291 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017633242708559713, + "loss": 0.5663, + "step": 8292 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017632619758608243, + "loss": 0.5027, + "step": 8293 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017631996737691718, + "loss": 0.5657, + "step": 8294 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017631373645815929, + "loss": 0.604, + "step": 8295 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017630750482986666, + "loss": 0.6411, + "step": 8296 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001763012724920973, + "loss": 0.4213, + "step": 8297 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001762950394449091, + "loss": 0.5588, + "step": 8298 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017628880568836003, + "loss": 0.5912, + "step": 8299 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017628257122250806, + "loss": 0.586, + "step": 8300 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017627633604741115, + "loss": 0.6604, + "step": 8301 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017627010016312727, + "loss": 0.6218, + "step": 8302 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017626386356971437, + "loss": 0.5899, + "step": 8303 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017625762626723051, + "loss": 0.6382, + "step": 8304 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001762513882557336, + "loss": 0.5621, + "step": 8305 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017624514953528165, + "loss": 0.6722, + "step": 8306 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017623891010593273, + "loss": 0.6049, + "step": 8307 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001762326699677448, + "loss": 0.6078, + "step": 8308 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001762264291207759, + "loss": 0.5672, + "step": 8309 + }, + { + "epoch": 0.25, + "learning_rate": 0.000176220187565084, + "loss": 0.644, + "step": 8310 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017621394530072722, + "loss": 0.5749, + "step": 8311 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017620770232776352, + "loss": 0.4746, + "step": 8312 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017620145864625099, + "loss": 0.5182, + "step": 8313 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017619521425624767, + "loss": 0.5559, + "step": 8314 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017618896915781162, + "loss": 0.5115, + "step": 8315 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001761827233510009, + "loss": 0.4437, + "step": 8316 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017617647683587358, + "loss": 0.48, + "step": 8317 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017617022961248773, + "loss": 0.4292, + "step": 8318 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017616398168090146, + "loss": 0.5196, + "step": 8319 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017615773304117282, + "loss": 0.6501, + "step": 8320 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017615148369335996, + "loss": 0.5739, + "step": 8321 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017614523363752093, + "loss": 0.5895, + "step": 8322 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001761389828737139, + "loss": 0.5683, + "step": 8323 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017613273140199692, + "loss": 0.5285, + "step": 8324 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017612647922242818, + "loss": 0.6442, + "step": 8325 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017612022633506575, + "loss": 0.6344, + "step": 8326 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001761139727399678, + "loss": 0.6023, + "step": 8327 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017610771843719247, + "loss": 0.5437, + "step": 8328 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001761014634267979, + "loss": 0.5936, + "step": 8329 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017609520770884226, + "loss": 0.5039, + "step": 8330 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001760889512833837, + "loss": 0.5232, + "step": 8331 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001760826941504804, + "loss": 0.5581, + "step": 8332 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017607643631019056, + "loss": 0.4759, + "step": 8333 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001760701777625723, + "loss": 0.5551, + "step": 8334 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017606391850768388, + "loss": 0.5657, + "step": 8335 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017605765854558343, + "loss": 0.5408, + "step": 8336 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017605139787632917, + "loss": 0.6159, + "step": 8337 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017604513649997937, + "loss": 0.6276, + "step": 8338 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017603887441659216, + "loss": 0.5577, + "step": 8339 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001760326116262258, + "loss": 0.5728, + "step": 8340 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001760263481289386, + "loss": 0.5944, + "step": 8341 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017602008392478863, + "loss": 0.623, + "step": 8342 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017601381901383425, + "loss": 0.5752, + "step": 8343 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017600755339613366, + "loss": 0.511, + "step": 8344 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017600128707174516, + "loss": 0.6174, + "step": 8345 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017599502004072697, + "loss": 0.7062, + "step": 8346 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017598875230313737, + "loss": 0.6329, + "step": 8347 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001759824838590346, + "loss": 0.585, + "step": 8348 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017597621470847704, + "loss": 0.5557, + "step": 8349 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001759699448515229, + "loss": 0.5058, + "step": 8350 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001759636742882305, + "loss": 0.5712, + "step": 8351 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001759574030186581, + "loss": 0.5238, + "step": 8352 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017595113104286404, + "loss": 0.5552, + "step": 8353 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001759448583609067, + "loss": 0.5257, + "step": 8354 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017593858497284427, + "loss": 0.6545, + "step": 8355 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017593231087873517, + "loss": 0.5525, + "step": 8356 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017592603607863769, + "loss": 0.6141, + "step": 8357 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001759197605726102, + "loss": 0.4513, + "step": 8358 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017591348436071103, + "loss": 0.6183, + "step": 8359 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017590720744299854, + "loss": 0.6483, + "step": 8360 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001759009298195311, + "loss": 0.66, + "step": 8361 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017589465149036703, + "loss": 0.626, + "step": 8362 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017588837245556478, + "loss": 0.6375, + "step": 8363 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017588209271518265, + "loss": 0.5783, + "step": 8364 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001758758122692791, + "loss": 0.5015, + "step": 8365 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017586953111791247, + "loss": 0.6404, + "step": 8366 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001758632492611412, + "loss": 0.4566, + "step": 8367 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017585696669902365, + "loss": 0.683, + "step": 8368 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017585068343161826, + "loss": 0.5834, + "step": 8369 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017584439945898345, + "loss": 0.5118, + "step": 8370 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017583811478117766, + "loss": 0.4992, + "step": 8371 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001758318293982593, + "loss": 0.6573, + "step": 8372 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001758255433102868, + "loss": 0.6376, + "step": 8373 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001758192565173186, + "loss": 0.5586, + "step": 8374 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001758129690194132, + "loss": 0.5503, + "step": 8375 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017580668081662904, + "loss": 0.524, + "step": 8376 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017580039190902456, + "loss": 0.5681, + "step": 8377 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017579410229665824, + "loss": 0.5876, + "step": 8378 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017578781197958855, + "loss": 0.4979, + "step": 8379 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017578152095787402, + "loss": 0.5154, + "step": 8380 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017577522923157313, + "loss": 0.4577, + "step": 8381 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017576893680074432, + "loss": 0.6222, + "step": 8382 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017576264366544615, + "loss": 0.5122, + "step": 8383 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017575634982573712, + "loss": 0.5646, + "step": 8384 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001757500552816757, + "loss": 0.5886, + "step": 8385 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017574376003332053, + "loss": 0.6231, + "step": 8386 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017573746408073, + "loss": 0.487, + "step": 8387 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017573116742396278, + "loss": 0.5031, + "step": 8388 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017572487006307732, + "loss": 0.6274, + "step": 8389 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001757185719981322, + "loss": 0.574, + "step": 8390 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017571227322918595, + "loss": 0.5657, + "step": 8391 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017570597375629718, + "loss": 0.6313, + "step": 8392 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017569967357952443, + "loss": 0.4919, + "step": 8393 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001756933726989263, + "loss": 0.7115, + "step": 8394 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017568707111456134, + "loss": 0.4458, + "step": 8395 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017568076882648818, + "loss": 0.5943, + "step": 8396 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017567446583476541, + "loss": 0.6326, + "step": 8397 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017566816213945158, + "loss": 0.5322, + "step": 8398 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017566185774060537, + "loss": 0.5195, + "step": 8399 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017565555263828535, + "loss": 0.5905, + "step": 8400 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017564924683255014, + "loss": 0.6085, + "step": 8401 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001756429403234584, + "loss": 0.6542, + "step": 8402 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017563663311106874, + "loss": 0.5636, + "step": 8403 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017563032519543977, + "loss": 0.5747, + "step": 8404 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017562401657663024, + "loss": 0.5793, + "step": 8405 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017561770725469872, + "loss": 0.6336, + "step": 8406 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017561139722970393, + "loss": 0.6091, + "step": 8407 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017560508650170446, + "loss": 0.5714, + "step": 8408 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017559877507075905, + "loss": 0.4826, + "step": 8409 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017559246293692636, + "loss": 0.6632, + "step": 8410 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001755861501002651, + "loss": 0.6148, + "step": 8411 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001755798365608339, + "loss": 0.5665, + "step": 8412 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017557352231869152, + "loss": 0.4902, + "step": 8413 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017556720737389663, + "loss": 0.622, + "step": 8414 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017556089172650802, + "loss": 0.669, + "step": 8415 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001755545753765843, + "loss": 0.5634, + "step": 8416 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017554825832418426, + "loss": 0.6055, + "step": 8417 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017554194056936665, + "loss": 0.5792, + "step": 8418 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017553562211219018, + "loss": 0.62, + "step": 8419 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017552930295271357, + "loss": 0.7456, + "step": 8420 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017552298309099563, + "loss": 0.5896, + "step": 8421 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017551666252709505, + "loss": 0.6319, + "step": 8422 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001755103412610707, + "loss": 0.5276, + "step": 8423 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001755040192929813, + "loss": 0.4642, + "step": 8424 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001754976966228856, + "loss": 0.6922, + "step": 8425 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017549137325084235, + "loss": 0.485, + "step": 8426 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017548504917691047, + "loss": 0.635, + "step": 8427 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017547872440114865, + "loss": 0.599, + "step": 8428 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017547239892361574, + "loss": 0.6151, + "step": 8429 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017546607274437058, + "loss": 0.6146, + "step": 8430 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001754597458634719, + "loss": 0.5714, + "step": 8431 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017545341828097862, + "loss": 0.6483, + "step": 8432 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017544708999694952, + "loss": 0.5061, + "step": 8433 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017544076101144344, + "loss": 0.5702, + "step": 8434 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017543443132451923, + "loss": 0.6257, + "step": 8435 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017542810093623576, + "loss": 0.4894, + "step": 8436 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017542176984665185, + "loss": 0.7456, + "step": 8437 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001754154380558264, + "loss": 0.4999, + "step": 8438 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017540910556381825, + "loss": 0.6369, + "step": 8439 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001754027723706863, + "loss": 0.5895, + "step": 8440 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001753964384764894, + "loss": 0.5987, + "step": 8441 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001753901038812865, + "loss": 0.4823, + "step": 8442 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017538376858513646, + "loss": 0.5604, + "step": 8443 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017537743258809818, + "loss": 0.6038, + "step": 8444 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017537109589023056, + "loss": 0.4032, + "step": 8445 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017536475849159254, + "loss": 0.5668, + "step": 8446 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017535842039224304, + "loss": 0.5541, + "step": 8447 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017535208159224098, + "loss": 0.6365, + "step": 8448 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001753457420916453, + "loss": 0.5609, + "step": 8449 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017533940189051492, + "loss": 0.5381, + "step": 8450 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017533306098890884, + "loss": 0.5662, + "step": 8451 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017532671938688595, + "loss": 0.575, + "step": 8452 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017532037708450524, + "loss": 0.5548, + "step": 8453 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001753140340818257, + "loss": 0.4732, + "step": 8454 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001753076903789063, + "loss": 0.5227, + "step": 8455 + }, + { + "epoch": 0.25, + "learning_rate": 0.000175301345975806, + "loss": 0.5836, + "step": 8456 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001752950008725838, + "loss": 0.5568, + "step": 8457 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001752886550692987, + "loss": 0.6292, + "step": 8458 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017528230856600968, + "loss": 0.6575, + "step": 8459 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017527596136277578, + "loss": 0.5482, + "step": 8460 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017526961345965596, + "loss": 0.6206, + "step": 8461 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017526326485670928, + "loss": 0.5822, + "step": 8462 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017525691555399478, + "loss": 0.5124, + "step": 8463 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017525056555157145, + "loss": 0.6631, + "step": 8464 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017524421484949837, + "loss": 0.568, + "step": 8465 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017523786344783457, + "loss": 0.5413, + "step": 8466 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017523151134663908, + "loss": 0.61, + "step": 8467 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017522515854597102, + "loss": 0.6109, + "step": 8468 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001752188050458894, + "loss": 0.6728, + "step": 8469 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001752124508464533, + "loss": 0.5399, + "step": 8470 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017520609594772182, + "loss": 0.5512, + "step": 8471 + }, + { + "epoch": 0.25, + "learning_rate": 0.000175199740349754, + "loss": 0.6321, + "step": 8472 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017519338405260902, + "loss": 0.5707, + "step": 8473 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017518702705634592, + "loss": 0.6565, + "step": 8474 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017518066936102376, + "loss": 0.543, + "step": 8475 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017517431096670174, + "loss": 0.4969, + "step": 8476 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017516795187343893, + "loss": 0.5476, + "step": 8477 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017516159208129446, + "loss": 0.5595, + "step": 8478 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017515523159032745, + "loss": 0.5296, + "step": 8479 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017514887040059705, + "loss": 0.5254, + "step": 8480 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017514250851216246, + "loss": 0.5241, + "step": 8481 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017513614592508273, + "loss": 0.4074, + "step": 8482 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017512978263941706, + "loss": 0.5027, + "step": 8483 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017512341865522462, + "loss": 0.6548, + "step": 8484 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017511705397256456, + "loss": 0.5474, + "step": 8485 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001751106885914961, + "loss": 0.507, + "step": 8486 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017510432251207839, + "loss": 0.4757, + "step": 8487 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001750979557343706, + "loss": 0.5765, + "step": 8488 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017509158825843195, + "loss": 0.5405, + "step": 8489 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017508522008432165, + "loss": 0.5509, + "step": 8490 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017507885121209891, + "loss": 0.5365, + "step": 8491 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017507248164182294, + "loss": 0.5535, + "step": 8492 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017506611137355294, + "loss": 0.6371, + "step": 8493 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017505974040734815, + "loss": 0.4973, + "step": 8494 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017505336874326782, + "loss": 0.6049, + "step": 8495 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001750469963813712, + "loss": 0.5045, + "step": 8496 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017504062332171746, + "loss": 0.7106, + "step": 8497 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017503424956436596, + "loss": 0.5636, + "step": 8498 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001750278751093759, + "loss": 0.5465, + "step": 8499 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017502149995680654, + "loss": 0.6185, + "step": 8500 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017501512410671722, + "loss": 0.6018, + "step": 8501 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017500874755916715, + "loss": 0.6144, + "step": 8502 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017500237031421563, + "loss": 0.5736, + "step": 8503 + }, + { + "epoch": 0.25, + "learning_rate": 0.000174995992371922, + "loss": 0.5292, + "step": 8504 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017498961373234548, + "loss": 0.5124, + "step": 8505 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017498323439554546, + "loss": 0.5824, + "step": 8506 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017497685436158116, + "loss": 0.7242, + "step": 8507 + }, + { + "epoch": 0.25, + "learning_rate": 0.000174970473630512, + "loss": 0.6172, + "step": 8508 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017496409220239722, + "loss": 0.6564, + "step": 8509 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001749577100772962, + "loss": 0.5364, + "step": 8510 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017495132725526828, + "loss": 0.5628, + "step": 8511 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017494494373637278, + "loss": 0.4465, + "step": 8512 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017493855952066908, + "loss": 0.6352, + "step": 8513 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017493217460821651, + "loss": 0.5076, + "step": 8514 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017492578899907445, + "loss": 0.5238, + "step": 8515 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017491940269330226, + "loss": 0.574, + "step": 8516 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001749130156909593, + "loss": 0.5232, + "step": 8517 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017490662799210502, + "loss": 0.5715, + "step": 8518 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017490023959679872, + "loss": 0.566, + "step": 8519 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001748938505050999, + "loss": 0.595, + "step": 8520 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017488746071706788, + "loss": 0.5325, + "step": 8521 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001748810702327621, + "loss": 0.6409, + "step": 8522 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017487467905224197, + "loss": 0.6641, + "step": 8523 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017486828717556692, + "loss": 0.5657, + "step": 8524 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017486189460279636, + "loss": 0.6303, + "step": 8525 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017485550133398978, + "loss": 0.5288, + "step": 8526 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017484910736920655, + "loss": 0.5204, + "step": 8527 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017484271270850618, + "loss": 0.496, + "step": 8528 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017483631735194804, + "loss": 0.5953, + "step": 8529 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001748299212995917, + "loss": 0.4898, + "step": 8530 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017482352455149656, + "loss": 0.5992, + "step": 8531 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017481712710772212, + "loss": 0.5959, + "step": 8532 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017481072896832786, + "loss": 0.6718, + "step": 8533 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017480433013337322, + "loss": 0.5615, + "step": 8534 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017479793060291775, + "loss": 0.4345, + "step": 8535 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017479153037702093, + "loss": 0.4396, + "step": 8536 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017478512945574227, + "loss": 0.4735, + "step": 8537 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017477872783914128, + "loss": 0.6433, + "step": 8538 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001747723255272775, + "loss": 0.5763, + "step": 8539 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017476592252021043, + "loss": 0.4985, + "step": 8540 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001747595188179996, + "loss": 0.4562, + "step": 8541 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017475311442070459, + "loss": 0.6317, + "step": 8542 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001747467093283849, + "loss": 0.6605, + "step": 8543 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001747403035411001, + "loss": 0.5182, + "step": 8544 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017473389705890973, + "loss": 0.5656, + "step": 8545 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017472748988187338, + "loss": 0.634, + "step": 8546 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017472108201005063, + "loss": 0.5058, + "step": 8547 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017471467344350102, + "loss": 0.4609, + "step": 8548 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017470826418228416, + "loss": 0.493, + "step": 8549 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017470185422645962, + "loss": 0.6055, + "step": 8550 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017469544357608701, + "loss": 0.6125, + "step": 8551 + }, + { + "epoch": 0.25, + "learning_rate": 0.000174689032231226, + "loss": 0.5652, + "step": 8552 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001746826201919361, + "loss": 0.5878, + "step": 8553 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017467620745827692, + "loss": 0.6606, + "step": 8554 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017466979403030818, + "loss": 0.4419, + "step": 8555 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017466337990808942, + "loss": 0.5746, + "step": 8556 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017465696509168032, + "loss": 0.5807, + "step": 8557 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017465054958114054, + "loss": 0.5599, + "step": 8558 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017464413337652966, + "loss": 0.6331, + "step": 8559 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017463771647790744, + "loss": 0.5775, + "step": 8560 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017463129888533344, + "loss": 0.6254, + "step": 8561 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017462488059886736, + "loss": 0.4928, + "step": 8562 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017461846161856886, + "loss": 0.5157, + "step": 8563 + }, + { + "epoch": 0.25, + "learning_rate": 0.0001746120419444977, + "loss": 0.419, + "step": 8564 + }, + { + "epoch": 0.25, + "learning_rate": 0.00017460562157671348, + "loss": 0.6583, + "step": 8565 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017459920051527593, + "loss": 0.5138, + "step": 8566 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017459277876024477, + "loss": 0.5341, + "step": 8567 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017458635631167964, + "loss": 0.545, + "step": 8568 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001745799331696403, + "loss": 0.5733, + "step": 8569 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001745735093341865, + "loss": 0.6091, + "step": 8570 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017456708480537793, + "loss": 0.4257, + "step": 8571 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017456065958327432, + "loss": 0.5397, + "step": 8572 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001745542336679354, + "loss": 0.6495, + "step": 8573 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017454780705942094, + "loss": 0.4832, + "step": 8574 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017454137975779065, + "loss": 0.5424, + "step": 8575 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017453495176310434, + "loss": 0.5473, + "step": 8576 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017452852307542178, + "loss": 0.4627, + "step": 8577 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001745220936948027, + "loss": 0.573, + "step": 8578 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001745156636213069, + "loss": 0.661, + "step": 8579 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017450923285499415, + "loss": 0.574, + "step": 8580 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017450280139592422, + "loss": 0.5104, + "step": 8581 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017449636924415698, + "loss": 0.6058, + "step": 8582 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017448993639975217, + "loss": 0.5624, + "step": 8583 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017448350286276964, + "loss": 0.6419, + "step": 8584 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017447706863326918, + "loss": 0.5712, + "step": 8585 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017447063371131061, + "loss": 0.5395, + "step": 8586 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001744641980969538, + "loss": 0.4752, + "step": 8587 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001744577617902585, + "loss": 0.5945, + "step": 8588 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017445132479128463, + "loss": 0.5606, + "step": 8589 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017444488710009203, + "loss": 0.5451, + "step": 8590 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017443844871674053, + "loss": 0.6349, + "step": 8591 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017443200964129002, + "loss": 0.5606, + "step": 8592 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001744255698738003, + "loss": 0.5337, + "step": 8593 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017441912941433134, + "loss": 0.582, + "step": 8594 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017441268826294296, + "loss": 0.4831, + "step": 8595 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017440624641969508, + "loss": 0.5011, + "step": 8596 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017439980388464754, + "loss": 0.5708, + "step": 8597 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017439336065786033, + "loss": 0.611, + "step": 8598 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017438691673939323, + "loss": 0.5097, + "step": 8599 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017438047212930627, + "loss": 0.4989, + "step": 8600 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017437402682765932, + "loss": 0.6787, + "step": 8601 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017436758083451233, + "loss": 0.4393, + "step": 8602 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017436113414992518, + "loss": 0.6687, + "step": 8603 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017435468677395787, + "loss": 0.562, + "step": 8604 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001743482387066703, + "loss": 0.591, + "step": 8605 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017434178994812243, + "loss": 0.4986, + "step": 8606 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017433534049837425, + "loss": 0.6344, + "step": 8607 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001743288903574857, + "loss": 0.5641, + "step": 8608 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017432243952551673, + "loss": 0.5715, + "step": 8609 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017431598800252734, + "loss": 0.4636, + "step": 8610 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017430953578857755, + "loss": 0.5621, + "step": 8611 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017430308288372728, + "loss": 0.6176, + "step": 8612 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017429662928803656, + "loss": 0.4897, + "step": 8613 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001742901750015654, + "loss": 0.457, + "step": 8614 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017428372002437382, + "loss": 0.5358, + "step": 8615 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001742772643565218, + "loss": 0.4641, + "step": 8616 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017427080799806935, + "loss": 0.6725, + "step": 8617 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001742643509490766, + "loss": 0.6016, + "step": 8618 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017425789320960346, + "loss": 0.5167, + "step": 8619 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017425143477971003, + "loss": 0.6368, + "step": 8620 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017424497565945638, + "loss": 0.5172, + "step": 8621 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017423851584890253, + "loss": 0.5522, + "step": 8622 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017423205534810852, + "loss": 0.6283, + "step": 8623 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017422559415713446, + "loss": 0.5805, + "step": 8624 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017421913227604042, + "loss": 0.5518, + "step": 8625 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017421266970488644, + "loss": 0.563, + "step": 8626 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017420620644373268, + "loss": 0.5306, + "step": 8627 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017419974249263914, + "loss": 0.6117, + "step": 8628 + }, + { + "epoch": 0.26, + "learning_rate": 0.000174193277851666, + "loss": 0.5449, + "step": 8629 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001741868125208733, + "loss": 0.5789, + "step": 8630 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017418034650032124, + "loss": 0.6312, + "step": 8631 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017417387979006986, + "loss": 0.625, + "step": 8632 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001741674123901793, + "loss": 0.6298, + "step": 8633 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017416094430070967, + "loss": 0.6218, + "step": 8634 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001741544755217212, + "loss": 0.526, + "step": 8635 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017414800605327393, + "loss": 0.5842, + "step": 8636 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017414153589542805, + "loss": 0.558, + "step": 8637 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017413506504824374, + "loss": 0.6625, + "step": 8638 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017412859351178113, + "loss": 0.5588, + "step": 8639 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001741221212861004, + "loss": 0.52, + "step": 8640 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017411564837126176, + "loss": 0.5108, + "step": 8641 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017410917476732532, + "loss": 0.5533, + "step": 8642 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001741027004743513, + "loss": 0.6104, + "step": 8643 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017409622549239995, + "loss": 0.5008, + "step": 8644 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001740897498215314, + "loss": 0.5459, + "step": 8645 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017408327346180591, + "loss": 0.4302, + "step": 8646 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017407679641328366, + "loss": 0.4392, + "step": 8647 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017407031867602485, + "loss": 0.592, + "step": 8648 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017406384025008977, + "loss": 0.5447, + "step": 8649 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017405736113553863, + "loss": 0.4899, + "step": 8650 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017405088133243165, + "loss": 0.5389, + "step": 8651 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017404440084082908, + "loss": 0.61, + "step": 8652 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017403791966079122, + "loss": 0.5636, + "step": 8653 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017403143779237822, + "loss": 0.4817, + "step": 8654 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017402495523565048, + "loss": 0.6046, + "step": 8655 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001740184719906682, + "loss": 0.6551, + "step": 8656 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017401198805749166, + "loss": 0.6706, + "step": 8657 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017400550343618114, + "loss": 0.5666, + "step": 8658 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017399901812679695, + "loss": 0.572, + "step": 8659 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001739925321293994, + "loss": 0.5724, + "step": 8660 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017398604544404875, + "loss": 0.5142, + "step": 8661 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017397955807080536, + "loss": 0.5472, + "step": 8662 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017397307000972952, + "loss": 0.529, + "step": 8663 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017396658126088156, + "loss": 0.5244, + "step": 8664 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017396009182432177, + "loss": 0.4955, + "step": 8665 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017395360170011057, + "loss": 0.4734, + "step": 8666 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017394711088830823, + "loss": 0.4925, + "step": 8667 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017394061938897515, + "loss": 0.5582, + "step": 8668 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017393412720217164, + "loss": 0.4593, + "step": 8669 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001739276343279581, + "loss": 0.6021, + "step": 8670 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017392114076639486, + "loss": 0.5425, + "step": 8671 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017391464651754236, + "loss": 0.5004, + "step": 8672 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001739081515814609, + "loss": 0.5127, + "step": 8673 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017390165595821086, + "loss": 0.4928, + "step": 8674 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017389515964785272, + "loss": 0.4846, + "step": 8675 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017388866265044686, + "loss": 0.5736, + "step": 8676 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017388216496605363, + "loss": 0.635, + "step": 8677 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001738756665947335, + "loss": 0.5125, + "step": 8678 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017386916753654686, + "loss": 0.4894, + "step": 8679 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017386266779155415, + "loss": 0.5499, + "step": 8680 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017385616735981577, + "loss": 0.5814, + "step": 8681 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001738496662413922, + "loss": 0.5577, + "step": 8682 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017384316443634385, + "loss": 0.5228, + "step": 8683 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001738366619447312, + "loss": 0.5993, + "step": 8684 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001738301587666147, + "loss": 0.463, + "step": 8685 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001738236549020548, + "loss": 0.514, + "step": 8686 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017381715035111198, + "loss": 0.5151, + "step": 8687 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017381064511384672, + "loss": 0.7114, + "step": 8688 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001738041391903195, + "loss": 0.6504, + "step": 8689 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017379763258059083, + "loss": 0.4638, + "step": 8690 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017379112528472115, + "loss": 0.6988, + "step": 8691 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017378461730277104, + "loss": 0.612, + "step": 8692 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017377810863480092, + "loss": 0.5551, + "step": 8693 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017377159928087137, + "loss": 0.6198, + "step": 8694 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017376508924104287, + "loss": 0.515, + "step": 8695 + }, + { + "epoch": 0.26, + "learning_rate": 0.000173758578515376, + "loss": 0.5943, + "step": 8696 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017375206710393127, + "loss": 0.5848, + "step": 8697 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001737455550067692, + "loss": 0.6121, + "step": 8698 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017373904222395034, + "loss": 0.6036, + "step": 8699 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017373252875553527, + "loss": 0.498, + "step": 8700 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017372601460158453, + "loss": 0.5234, + "step": 8701 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017371949976215868, + "loss": 0.5331, + "step": 8702 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017371298423731832, + "loss": 0.4895, + "step": 8703 + }, + { + "epoch": 0.26, + "learning_rate": 0.000173706468027124, + "loss": 0.5905, + "step": 8704 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017369995113163632, + "loss": 0.575, + "step": 8705 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001736934335509159, + "loss": 0.6113, + "step": 8706 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017368691528502325, + "loss": 0.6427, + "step": 8707 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017368039633401907, + "loss": 0.4784, + "step": 8708 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017367387669796393, + "loss": 0.4981, + "step": 8709 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017366735637691842, + "loss": 0.5067, + "step": 8710 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001736608353709432, + "loss": 0.5464, + "step": 8711 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001736543136800989, + "loss": 0.579, + "step": 8712 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017364779130444618, + "loss": 0.5809, + "step": 8713 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001736412682440456, + "loss": 0.6021, + "step": 8714 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017363474449895786, + "loss": 0.5334, + "step": 8715 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017362822006924364, + "loss": 0.5588, + "step": 8716 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017362169495496355, + "loss": 0.5907, + "step": 8717 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001736151691561783, + "loss": 0.5211, + "step": 8718 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017360864267294853, + "loss": 0.6235, + "step": 8719 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017360211550533497, + "loss": 0.6545, + "step": 8720 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017359558765339824, + "loss": 0.5764, + "step": 8721 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001735890591171991, + "loss": 0.5898, + "step": 8722 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001735825298967982, + "loss": 0.6221, + "step": 8723 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017357599999225627, + "loss": 0.5608, + "step": 8724 + }, + { + "epoch": 0.26, + "learning_rate": 0.000173569469403634, + "loss": 0.518, + "step": 8725 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017356293813099215, + "loss": 0.5375, + "step": 8726 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001735564061743914, + "loss": 0.5345, + "step": 8727 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017354987353389252, + "loss": 0.5084, + "step": 8728 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017354334020955623, + "loss": 0.6261, + "step": 8729 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017353680620144326, + "loss": 0.5982, + "step": 8730 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017353027150961438, + "loss": 0.557, + "step": 8731 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017352373613413032, + "loss": 0.5407, + "step": 8732 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017351720007505192, + "loss": 0.5342, + "step": 8733 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017351066333243985, + "loss": 0.7046, + "step": 8734 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017350412590635495, + "loss": 0.5849, + "step": 8735 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017349758779685798, + "loss": 0.62, + "step": 8736 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001734910490040097, + "loss": 0.5742, + "step": 8737 + }, + { + "epoch": 0.26, + "learning_rate": 0.000173484509527871, + "loss": 0.5642, + "step": 8738 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017347796936850255, + "loss": 0.505, + "step": 8739 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017347142852596528, + "loss": 0.6426, + "step": 8740 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017346488700031992, + "loss": 0.5712, + "step": 8741 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017345834479162733, + "loss": 0.6201, + "step": 8742 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017345180189994833, + "loss": 0.5648, + "step": 8743 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017344525832534375, + "loss": 0.6327, + "step": 8744 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017343871406787443, + "loss": 0.5026, + "step": 8745 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001734321691276012, + "loss": 0.5356, + "step": 8746 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017342562350458499, + "loss": 0.5756, + "step": 8747 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017341907719888654, + "loss": 0.52, + "step": 8748 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017341253021056676, + "loss": 0.537, + "step": 8749 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017340598253968656, + "loss": 0.6616, + "step": 8750 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001733994341863068, + "loss": 0.5827, + "step": 8751 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017339288515048836, + "loss": 0.5809, + "step": 8752 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001733863354322921, + "loss": 0.5928, + "step": 8753 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017337978503177893, + "loss": 0.5169, + "step": 8754 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001733732339490098, + "loss": 0.4354, + "step": 8755 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017336668218404557, + "loss": 0.4967, + "step": 8756 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017336012973694719, + "loss": 0.5683, + "step": 8757 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017335357660777553, + "loss": 0.5408, + "step": 8758 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017334702279659157, + "loss": 0.7308, + "step": 8759 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001733404683034562, + "loss": 0.5275, + "step": 8760 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001733339131284304, + "loss": 0.5056, + "step": 8761 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017332735727157512, + "loss": 0.5279, + "step": 8762 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001733208007329513, + "loss": 0.5548, + "step": 8763 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017331424351261986, + "loss": 0.5594, + "step": 8764 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017330768561064184, + "loss": 0.548, + "step": 8765 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017330112702707817, + "loss": 0.4902, + "step": 8766 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017329456776198982, + "loss": 0.5892, + "step": 8767 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001732880078154378, + "loss": 0.4197, + "step": 8768 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001732814471874831, + "loss": 0.5956, + "step": 8769 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001732748858781867, + "loss": 0.5523, + "step": 8770 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017326832388760963, + "loss": 0.5345, + "step": 8771 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017326176121581287, + "loss": 0.639, + "step": 8772 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017325519786285745, + "loss": 0.5604, + "step": 8773 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017324863382880442, + "loss": 0.5828, + "step": 8774 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001732420691137148, + "loss": 0.4504, + "step": 8775 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017323550371764956, + "loss": 0.6178, + "step": 8776 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017322893764066982, + "loss": 0.6535, + "step": 8777 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017322237088283662, + "loss": 0.5387, + "step": 8778 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017321580344421098, + "loss": 0.4828, + "step": 8779 + }, + { + "epoch": 0.26, + "learning_rate": 0.000173209235324854, + "loss": 0.5571, + "step": 8780 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017320266652482672, + "loss": 0.538, + "step": 8781 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001731960970441902, + "loss": 0.579, + "step": 8782 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017318952688300557, + "loss": 0.5579, + "step": 8783 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017318295604133388, + "loss": 0.6544, + "step": 8784 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017317638451923623, + "loss": 0.5141, + "step": 8785 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017316981231677372, + "loss": 0.5643, + "step": 8786 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017316323943400747, + "loss": 0.6848, + "step": 8787 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017315666587099858, + "loss": 0.5827, + "step": 8788 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017315009162780816, + "loss": 0.5167, + "step": 8789 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017314351670449734, + "loss": 0.5654, + "step": 8790 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017313694110112723, + "loss": 0.4887, + "step": 8791 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017313036481775903, + "loss": 0.5144, + "step": 8792 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001731237878544538, + "loss": 0.6254, + "step": 8793 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017311721021127282, + "loss": 0.6336, + "step": 8794 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001731106318882771, + "loss": 0.6226, + "step": 8795 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017310405288552784, + "loss": 0.5382, + "step": 8796 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001730974732030863, + "loss": 0.6459, + "step": 8797 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017309089284101354, + "loss": 0.6153, + "step": 8798 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001730843117993708, + "loss": 0.5615, + "step": 8799 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017307773007821926, + "loss": 0.5841, + "step": 8800 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017307114767762012, + "loss": 0.5839, + "step": 8801 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017306456459763455, + "loss": 0.5394, + "step": 8802 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001730579808383238, + "loss": 0.5684, + "step": 8803 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017305139639974906, + "loss": 0.5628, + "step": 8804 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017304481128197155, + "loss": 0.5129, + "step": 8805 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017303822548505248, + "loss": 0.5494, + "step": 8806 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017303163900905311, + "loss": 0.5125, + "step": 8807 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017302505185403465, + "loss": 0.5173, + "step": 8808 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017301846402005842, + "loss": 0.585, + "step": 8809 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017301187550718556, + "loss": 0.6351, + "step": 8810 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001730052863154774, + "loss": 0.4762, + "step": 8811 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017299869644499518, + "loss": 0.6058, + "step": 8812 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017299210589580017, + "loss": 0.556, + "step": 8813 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017298551466795367, + "loss": 0.5946, + "step": 8814 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017297892276151692, + "loss": 0.5391, + "step": 8815 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017297233017655124, + "loss": 0.6186, + "step": 8816 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017296573691311795, + "loss": 0.5886, + "step": 8817 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001729591429712783, + "loss": 0.5484, + "step": 8818 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001729525483510936, + "loss": 0.4976, + "step": 8819 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001729459530526252, + "loss": 0.5707, + "step": 8820 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017293935707593438, + "loss": 0.6283, + "step": 8821 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001729327604210825, + "loss": 0.5432, + "step": 8822 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001729261630881309, + "loss": 0.6314, + "step": 8823 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001729195650771409, + "loss": 0.6828, + "step": 8824 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001729129663881738, + "loss": 0.4776, + "step": 8825 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017290636702129106, + "loss": 0.5647, + "step": 8826 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017289976697655395, + "loss": 0.5102, + "step": 8827 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017289316625402384, + "loss": 0.5536, + "step": 8828 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017288656485376212, + "loss": 0.5742, + "step": 8829 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017287996277583022, + "loss": 0.5804, + "step": 8830 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017287336002028944, + "loss": 0.4852, + "step": 8831 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001728667565872012, + "loss": 0.4845, + "step": 8832 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017286015247662691, + "loss": 0.5535, + "step": 8833 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017285354768862798, + "loss": 0.5019, + "step": 8834 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001728469422232658, + "loss": 0.4955, + "step": 8835 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017284033608060172, + "loss": 0.4748, + "step": 8836 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017283372926069727, + "loss": 0.5234, + "step": 8837 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017282712176361383, + "loss": 0.5623, + "step": 8838 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017282051358941288, + "loss": 0.6217, + "step": 8839 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017281390473815575, + "loss": 0.564, + "step": 8840 + }, + { + "epoch": 0.26, + "learning_rate": 0.000172807295209904, + "loss": 0.6337, + "step": 8841 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017280068500471902, + "loss": 0.4872, + "step": 8842 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017279407412266229, + "loss": 0.576, + "step": 8843 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017278746256379525, + "loss": 0.6019, + "step": 8844 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017278085032817943, + "loss": 0.6193, + "step": 8845 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017277423741587626, + "loss": 0.5109, + "step": 8846 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017276762382694723, + "loss": 0.4264, + "step": 8847 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017276100956145385, + "loss": 0.6326, + "step": 8848 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017275439461945756, + "loss": 0.5183, + "step": 8849 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017274777900101995, + "loss": 0.5438, + "step": 8850 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017274116270620245, + "loss": 0.5079, + "step": 8851 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017273454573506664, + "loss": 0.6905, + "step": 8852 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017272792808767403, + "loss": 0.5522, + "step": 8853 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001727213097640861, + "loss": 0.5817, + "step": 8854 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001727146907643644, + "loss": 0.5993, + "step": 8855 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001727080710885705, + "loss": 0.5569, + "step": 8856 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017270145073676596, + "loss": 0.6612, + "step": 8857 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001726948297090123, + "loss": 0.5438, + "step": 8858 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017268820800537106, + "loss": 0.5978, + "step": 8859 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017268158562590386, + "loss": 0.6171, + "step": 8860 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017267496257067225, + "loss": 0.572, + "step": 8861 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017266833883973777, + "loss": 0.5147, + "step": 8862 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017266171443316205, + "loss": 0.5071, + "step": 8863 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017265508935100668, + "loss": 0.5273, + "step": 8864 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017264846359333323, + "loss": 0.5109, + "step": 8865 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017264183716020332, + "loss": 0.5046, + "step": 8866 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017263521005167858, + "loss": 0.6025, + "step": 8867 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001726285822678206, + "loss": 0.4614, + "step": 8868 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017262195380869102, + "loss": 0.6577, + "step": 8869 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001726153246743514, + "loss": 0.5609, + "step": 8870 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001726086948648635, + "loss": 0.6555, + "step": 8871 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017260206438028886, + "loss": 0.5743, + "step": 8872 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001725954332206892, + "loss": 0.4561, + "step": 8873 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001725888013861261, + "loss": 0.5995, + "step": 8874 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017258216887666125, + "loss": 0.5232, + "step": 8875 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017257553569235634, + "loss": 0.6532, + "step": 8876 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017256890183327302, + "loss": 0.5675, + "step": 8877 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017256226729947297, + "loss": 0.5266, + "step": 8878 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001725556320910179, + "loss": 0.5184, + "step": 8879 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017254899620796946, + "loss": 0.485, + "step": 8880 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017254235965038936, + "loss": 0.4406, + "step": 8881 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017253572241833934, + "loss": 0.5775, + "step": 8882 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017252908451188107, + "loss": 0.5906, + "step": 8883 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017252244593107626, + "loss": 0.6016, + "step": 8884 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017251580667598668, + "loss": 0.6712, + "step": 8885 + }, + { + "epoch": 0.26, + "learning_rate": 0.000172509166746674, + "loss": 0.6331, + "step": 8886 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017250252614320002, + "loss": 0.5754, + "step": 8887 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017249588486562646, + "loss": 0.5285, + "step": 8888 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017248924291401503, + "loss": 0.4629, + "step": 8889 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017248260028842748, + "loss": 0.5141, + "step": 8890 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017247595698892564, + "loss": 0.5441, + "step": 8891 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017246931301557124, + "loss": 0.4634, + "step": 8892 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017246266836842605, + "loss": 0.5504, + "step": 8893 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017245602304755185, + "loss": 0.4723, + "step": 8894 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001724493770530104, + "loss": 0.6448, + "step": 8895 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017244273038486357, + "loss": 0.6283, + "step": 8896 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017243608304317305, + "loss": 0.6032, + "step": 8897 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017242943502800072, + "loss": 0.5165, + "step": 8898 + }, + { + "epoch": 0.26, + "learning_rate": 0.0001724227863394084, + "loss": 0.4861, + "step": 8899 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017241613697745783, + "loss": 0.498, + "step": 8900 + }, + { + "epoch": 0.26, + "learning_rate": 0.00017240948694221091, + "loss": 0.5752, + "step": 8901 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017240283623372942, + "loss": 0.6685, + "step": 8902 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017239618485207525, + "loss": 0.5886, + "step": 8903 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017238953279731018, + "loss": 0.5416, + "step": 8904 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017238288006949607, + "loss": 0.6522, + "step": 8905 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017237622666869485, + "loss": 0.4954, + "step": 8906 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017236957259496826, + "loss": 0.6427, + "step": 8907 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017236291784837823, + "loss": 0.5129, + "step": 8908 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017235626242898666, + "loss": 0.5872, + "step": 8909 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001723496063368554, + "loss": 0.4783, + "step": 8910 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017234294957204634, + "loss": 0.5346, + "step": 8911 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017233629213462136, + "loss": 0.5316, + "step": 8912 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017232963402464234, + "loss": 0.5338, + "step": 8913 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017232297524217123, + "loss": 0.4954, + "step": 8914 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017231631578726994, + "loss": 0.4656, + "step": 8915 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017230965566000035, + "loss": 0.4369, + "step": 8916 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001723029948604244, + "loss": 0.4914, + "step": 8917 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017229633338860403, + "loss": 0.5318, + "step": 8918 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017228967124460117, + "loss": 0.602, + "step": 8919 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017228300842847771, + "loss": 0.4863, + "step": 8920 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001722763449402957, + "loss": 0.5561, + "step": 8921 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017226968078011703, + "loss": 0.5825, + "step": 8922 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017226301594800366, + "loss": 0.5046, + "step": 8923 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017225635044401758, + "loss": 0.5559, + "step": 8924 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017224968426822078, + "loss": 0.5809, + "step": 8925 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017224301742067516, + "loss": 0.5613, + "step": 8926 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017223634990144276, + "loss": 0.6849, + "step": 8927 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001722296817105856, + "loss": 0.6691, + "step": 8928 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017222301284816561, + "loss": 0.5226, + "step": 8929 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017221634331424486, + "loss": 0.5145, + "step": 8930 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017220967310888535, + "loss": 0.6007, + "step": 8931 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017220300223214907, + "loss": 0.6026, + "step": 8932 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017219633068409803, + "loss": 0.5568, + "step": 8933 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001721896584647943, + "loss": 0.613, + "step": 8934 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017218298557429988, + "loss": 0.5475, + "step": 8935 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017217631201267683, + "loss": 0.5318, + "step": 8936 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001721696377799872, + "loss": 0.6594, + "step": 8937 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017216296287629308, + "loss": 0.749, + "step": 8938 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017215628730165645, + "loss": 0.5163, + "step": 8939 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017214961105613944, + "loss": 0.5642, + "step": 8940 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017214293413980408, + "loss": 0.6731, + "step": 8941 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017213625655271248, + "loss": 0.5434, + "step": 8942 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017212957829492673, + "loss": 0.5753, + "step": 8943 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001721228993665089, + "loss": 0.5229, + "step": 8944 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001721162197675211, + "loss": 0.6174, + "step": 8945 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017210953949802545, + "loss": 0.6554, + "step": 8946 + }, + { + "epoch": 0.27, + "learning_rate": 0.000172102858558084, + "loss": 0.7495, + "step": 8947 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017209617694775894, + "loss": 0.5398, + "step": 8948 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017208949466711234, + "loss": 0.4404, + "step": 8949 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017208281171620637, + "loss": 0.464, + "step": 8950 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017207612809510314, + "loss": 0.5764, + "step": 8951 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017206944380386481, + "loss": 0.5911, + "step": 8952 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001720627588425535, + "loss": 0.5221, + "step": 8953 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001720560732112314, + "loss": 0.4586, + "step": 8954 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001720493869099606, + "loss": 0.6481, + "step": 8955 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017204269993880336, + "loss": 0.4911, + "step": 8956 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017203601229782178, + "loss": 0.5189, + "step": 8957 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001720293239870781, + "loss": 0.6579, + "step": 8958 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017202263500663443, + "loss": 0.5111, + "step": 8959 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017201594535655305, + "loss": 0.5589, + "step": 8960 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001720092550368961, + "loss": 0.5033, + "step": 8961 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001720025640477258, + "loss": 0.4546, + "step": 8962 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017199587238910435, + "loss": 0.5036, + "step": 8963 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017198918006109396, + "loss": 0.525, + "step": 8964 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001719824870637569, + "loss": 0.5126, + "step": 8965 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017197579339715534, + "loss": 0.6089, + "step": 8966 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017196909906135152, + "loss": 0.6886, + "step": 8967 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017196240405640775, + "loss": 0.5324, + "step": 8968 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001719557083823862, + "loss": 0.4888, + "step": 8969 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017194901203934914, + "loss": 0.4882, + "step": 8970 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017194231502735887, + "loss": 0.609, + "step": 8971 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001719356173464776, + "loss": 0.4656, + "step": 8972 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017192891899676767, + "loss": 0.5164, + "step": 8973 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017192221997829128, + "loss": 0.5826, + "step": 8974 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017191552029111078, + "loss": 0.4875, + "step": 8975 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017190881993528842, + "loss": 0.5393, + "step": 8976 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017190211891088652, + "loss": 0.4854, + "step": 8977 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017189541721796738, + "loss": 0.5603, + "step": 8978 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001718887148565933, + "loss": 0.4687, + "step": 8979 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001718820118268266, + "loss": 0.4534, + "step": 8980 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001718753081287296, + "loss": 0.5308, + "step": 8981 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001718686037623646, + "loss": 0.5576, + "step": 8982 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017186189872779402, + "loss": 0.5436, + "step": 8983 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017185519302508016, + "loss": 0.4975, + "step": 8984 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001718484866542853, + "loss": 0.6027, + "step": 8985 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017184177961547187, + "loss": 0.6096, + "step": 8986 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001718350719087022, + "loss": 0.6351, + "step": 8987 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017182836353403866, + "loss": 0.6977, + "step": 8988 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017182165449154365, + "loss": 0.5576, + "step": 8989 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017181494478127947, + "loss": 0.5288, + "step": 8990 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001718082344033086, + "loss": 0.3489, + "step": 8991 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017180152335769338, + "loss": 0.6172, + "step": 8992 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017179481164449623, + "loss": 0.511, + "step": 8993 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001717880992637795, + "loss": 0.6035, + "step": 8994 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017178138621560564, + "loss": 0.758, + "step": 8995 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001717746725000371, + "loss": 0.6559, + "step": 8996 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017176795811713622, + "loss": 0.5538, + "step": 8997 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017176124306696546, + "loss": 0.5402, + "step": 8998 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001717545273495873, + "loss": 0.5865, + "step": 8999 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017174781096506413, + "loss": 0.5332, + "step": 9000 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001717410939134584, + "loss": 0.5292, + "step": 9001 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017173437619483257, + "loss": 0.5699, + "step": 9002 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017172765780924912, + "loss": 0.4754, + "step": 9003 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017172093875677047, + "loss": 0.5458, + "step": 9004 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017171421903745915, + "loss": 0.5467, + "step": 9005 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001717074986513776, + "loss": 0.4332, + "step": 9006 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017170077759858825, + "loss": 0.5266, + "step": 9007 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001716940558791537, + "loss": 0.57, + "step": 9008 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017168733349313638, + "loss": 0.5739, + "step": 9009 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001716806104405988, + "loss": 0.5811, + "step": 9010 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001716738867216035, + "loss": 0.5038, + "step": 9011 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017166716233621293, + "loss": 0.6862, + "step": 9012 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017166043728448967, + "loss": 0.4911, + "step": 9013 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017165371156649623, + "loss": 0.5495, + "step": 9014 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017164698518229513, + "loss": 0.5602, + "step": 9015 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017164025813194892, + "loss": 0.5236, + "step": 9016 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017163353041552016, + "loss": 0.6303, + "step": 9017 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017162680203307136, + "loss": 0.5349, + "step": 9018 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001716200729846651, + "loss": 0.5224, + "step": 9019 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017161334327036394, + "loss": 0.4907, + "step": 9020 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017160661289023047, + "loss": 0.6651, + "step": 9021 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017159988184432728, + "loss": 0.4556, + "step": 9022 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017159315013271688, + "loss": 0.5472, + "step": 9023 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017158641775546194, + "loss": 0.5287, + "step": 9024 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017157968471262497, + "loss": 0.6848, + "step": 9025 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017157295100426865, + "loss": 0.6463, + "step": 9026 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017156621663045555, + "loss": 0.7151, + "step": 9027 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017155948159124833, + "loss": 0.5656, + "step": 9028 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017155274588670952, + "loss": 0.5173, + "step": 9029 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017154600951690182, + "loss": 0.545, + "step": 9030 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017153927248188787, + "loss": 0.5234, + "step": 9031 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017153253478173022, + "loss": 0.547, + "step": 9032 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001715257964164916, + "loss": 0.5177, + "step": 9033 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017151905738623465, + "loss": 0.6472, + "step": 9034 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017151231769102197, + "loss": 0.5724, + "step": 9035 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017150557733091627, + "loss": 0.5205, + "step": 9036 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017149883630598021, + "loss": 0.4433, + "step": 9037 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001714920946162765, + "loss": 0.6528, + "step": 9038 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017148535226186774, + "loss": 0.5569, + "step": 9039 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001714786092428167, + "loss": 0.5331, + "step": 9040 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017147186555918604, + "loss": 0.6506, + "step": 9041 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017146512121103845, + "loss": 0.531, + "step": 9042 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017145837619843664, + "loss": 0.6041, + "step": 9043 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017145163052144336, + "loss": 0.5566, + "step": 9044 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017144488418012127, + "loss": 0.6302, + "step": 9045 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017143813717453312, + "loss": 0.4944, + "step": 9046 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017143138950474166, + "loss": 0.5992, + "step": 9047 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017142464117080958, + "loss": 0.6055, + "step": 9048 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017141789217279968, + "loss": 0.4908, + "step": 9049 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017141114251077467, + "loss": 0.539, + "step": 9050 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017140439218479733, + "loss": 0.6637, + "step": 9051 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017139764119493043, + "loss": 0.6095, + "step": 9052 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017139088954123667, + "loss": 0.493, + "step": 9053 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001713841372237789, + "loss": 0.6277, + "step": 9054 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017137738424261988, + "loss": 0.5704, + "step": 9055 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017137063059782237, + "loss": 0.5421, + "step": 9056 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017136387628944918, + "loss": 0.4438, + "step": 9057 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017135712131756314, + "loss": 0.6364, + "step": 9058 + }, + { + "epoch": 0.27, + "learning_rate": 0.000171350365682227, + "loss": 0.6547, + "step": 9059 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001713436093835036, + "loss": 0.5582, + "step": 9060 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017133685242145576, + "loss": 0.6367, + "step": 9061 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001713300947961463, + "loss": 0.5938, + "step": 9062 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017132333650763805, + "loss": 0.4247, + "step": 9063 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001713165775559938, + "loss": 0.5742, + "step": 9064 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017130981794127653, + "loss": 0.5385, + "step": 9065 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001713030576635489, + "loss": 0.5087, + "step": 9066 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001712962967228739, + "loss": 0.5546, + "step": 9067 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017128953511931437, + "loss": 0.5183, + "step": 9068 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017128277285293313, + "loss": 0.5578, + "step": 9069 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017127600992379306, + "loss": 0.5869, + "step": 9070 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001712692463319571, + "loss": 0.5924, + "step": 9071 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017126248207748805, + "loss": 0.5256, + "step": 9072 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017125571716044888, + "loss": 0.4972, + "step": 9073 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017124895158090244, + "loss": 0.5442, + "step": 9074 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017124218533891164, + "loss": 0.6256, + "step": 9075 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001712354184345394, + "loss": 0.5451, + "step": 9076 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017122865086784863, + "loss": 0.5163, + "step": 9077 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017122188263890225, + "loss": 0.5516, + "step": 9078 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001712151137477632, + "loss": 0.5159, + "step": 9079 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017120834419449442, + "loss": 0.6282, + "step": 9080 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001712015739791588, + "loss": 0.7048, + "step": 9081 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017119480310181931, + "loss": 0.5001, + "step": 9082 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017118803156253897, + "loss": 0.5583, + "step": 9083 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017118125936138067, + "loss": 0.5248, + "step": 9084 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017117448649840738, + "loss": 0.5395, + "step": 9085 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017116771297368208, + "loss": 0.6072, + "step": 9086 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017116093878726774, + "loss": 0.5904, + "step": 9087 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017115416393922737, + "loss": 0.5594, + "step": 9088 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017114738842962393, + "loss": 0.5843, + "step": 9089 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017114061225852046, + "loss": 0.5733, + "step": 9090 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001711338354259799, + "loss": 0.4723, + "step": 9091 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017112705793206526, + "loss": 0.558, + "step": 9092 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017112027977683964, + "loss": 0.5767, + "step": 9093 + }, + { + "epoch": 0.27, + "learning_rate": 0.000171113500960366, + "loss": 0.5653, + "step": 9094 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017110672148270734, + "loss": 0.5672, + "step": 9095 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017109994134392674, + "loss": 0.5504, + "step": 9096 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001710931605440872, + "loss": 0.6085, + "step": 9097 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017108637908325186, + "loss": 0.6004, + "step": 9098 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017107959696148364, + "loss": 0.4983, + "step": 9099 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017107281417884566, + "loss": 0.5215, + "step": 9100 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017106603073540098, + "loss": 0.4824, + "step": 9101 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017105924663121268, + "loss": 0.6143, + "step": 9102 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017105246186634382, + "loss": 0.4594, + "step": 9103 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001710456764408575, + "loss": 0.6059, + "step": 9104 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001710388903548168, + "loss": 0.5683, + "step": 9105 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001710321036082848, + "loss": 0.4887, + "step": 9106 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017102531620132461, + "loss": 0.6182, + "step": 9107 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017101852813399934, + "loss": 0.4997, + "step": 9108 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017101173940637211, + "loss": 0.5327, + "step": 9109 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017100495001850602, + "loss": 0.5444, + "step": 9110 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001709981599704642, + "loss": 0.5062, + "step": 9111 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017099136926230978, + "loss": 0.5662, + "step": 9112 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017098457789410594, + "loss": 0.6479, + "step": 9113 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017097778586591574, + "loss": 0.5384, + "step": 9114 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001709709931778024, + "loss": 0.525, + "step": 9115 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017096419982982905, + "loss": 0.6271, + "step": 9116 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017095740582205886, + "loss": 0.6143, + "step": 9117 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017095061115455498, + "loss": 0.5047, + "step": 9118 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001709438158273806, + "loss": 0.5316, + "step": 9119 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017093701984059888, + "loss": 0.5154, + "step": 9120 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017093022319427306, + "loss": 0.5767, + "step": 9121 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017092342588846625, + "loss": 0.5947, + "step": 9122 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001709166279232417, + "loss": 0.6217, + "step": 9123 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017090982929866263, + "loss": 0.6713, + "step": 9124 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017090303001479225, + "loss": 0.5939, + "step": 9125 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001708962300716937, + "loss": 0.4966, + "step": 9126 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017088942946943028, + "loss": 0.598, + "step": 9127 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017088262820806523, + "loss": 0.6152, + "step": 9128 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001708758262876617, + "loss": 0.5456, + "step": 9129 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017086902370828304, + "loss": 0.5427, + "step": 9130 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017086222046999242, + "loss": 0.4549, + "step": 9131 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001708554165728531, + "loss": 0.4957, + "step": 9132 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017084861201692838, + "loss": 0.6306, + "step": 9133 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001708418068022815, + "loss": 0.5697, + "step": 9134 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017083500092897574, + "loss": 0.5493, + "step": 9135 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017082819439707435, + "loss": 0.6358, + "step": 9136 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017082138720664065, + "loss": 0.6196, + "step": 9137 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001708145793577379, + "loss": 0.5353, + "step": 9138 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017080777085042945, + "loss": 0.5337, + "step": 9139 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017080096168477855, + "loss": 0.5654, + "step": 9140 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017079415186084854, + "loss": 0.5273, + "step": 9141 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017078734137870272, + "loss": 0.5533, + "step": 9142 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001707805302384044, + "loss": 0.6586, + "step": 9143 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001707737184400169, + "loss": 0.5582, + "step": 9144 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001707669059836036, + "loss": 0.5871, + "step": 9145 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017076009286922782, + "loss": 0.5452, + "step": 9146 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001707532790969529, + "loss": 0.6156, + "step": 9147 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001707464646668422, + "loss": 0.5604, + "step": 9148 + }, + { + "epoch": 0.27, + "learning_rate": 0.000170739649578959, + "loss": 0.5702, + "step": 9149 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001707328338333668, + "loss": 0.5787, + "step": 9150 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017072601743012888, + "loss": 0.512, + "step": 9151 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017071920036930863, + "loss": 0.524, + "step": 9152 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017071238265096948, + "loss": 0.5744, + "step": 9153 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017070556427517474, + "loss": 0.4841, + "step": 9154 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017069874524198783, + "loss": 0.6124, + "step": 9155 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001706919255514722, + "loss": 0.6284, + "step": 9156 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001706851052036912, + "loss": 0.5258, + "step": 9157 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017067828419870828, + "loss": 0.5451, + "step": 9158 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017067146253658684, + "loss": 0.5891, + "step": 9159 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017066464021739034, + "loss": 0.5479, + "step": 9160 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017065781724118212, + "loss": 0.5097, + "step": 9161 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001706509936080257, + "loss": 0.487, + "step": 9162 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001706441693179845, + "loss": 0.5752, + "step": 9163 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017063734437112196, + "loss": 0.6545, + "step": 9164 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001706305187675016, + "loss": 0.5706, + "step": 9165 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017062369250718678, + "loss": 0.5437, + "step": 9166 + }, + { + "epoch": 0.27, + "learning_rate": 0.000170616865590241, + "loss": 0.5367, + "step": 9167 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017061003801672778, + "loss": 0.5071, + "step": 9168 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017060320978671057, + "loss": 0.5816, + "step": 9169 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017059638090025282, + "loss": 0.4918, + "step": 9170 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001705895513574181, + "loss": 0.5909, + "step": 9171 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017058272115826983, + "loss": 0.6203, + "step": 9172 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017057589030287159, + "loss": 0.5577, + "step": 9173 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001705690587912868, + "loss": 0.6586, + "step": 9174 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017056222662357906, + "loss": 0.6167, + "step": 9175 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017055539379981185, + "loss": 0.624, + "step": 9176 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001705485603200487, + "loss": 0.586, + "step": 9177 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017054172618435318, + "loss": 0.5938, + "step": 9178 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001705348913927888, + "loss": 0.5549, + "step": 9179 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001705280559454191, + "loss": 0.5638, + "step": 9180 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017052121984230764, + "loss": 0.6045, + "step": 9181 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017051438308351798, + "loss": 0.6194, + "step": 9182 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017050754566911373, + "loss": 0.5035, + "step": 9183 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017050070759915838, + "loss": 0.6048, + "step": 9184 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017049386887371557, + "loss": 0.4911, + "step": 9185 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017048702949284885, + "loss": 0.5746, + "step": 9186 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017048018945662184, + "loss": 0.5629, + "step": 9187 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001704733487650981, + "loss": 0.7228, + "step": 9188 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017046650741834126, + "loss": 0.6324, + "step": 9189 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017045966541641495, + "loss": 0.5616, + "step": 9190 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017045282275938273, + "loss": 0.4803, + "step": 9191 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017044597944730822, + "loss": 0.6279, + "step": 9192 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001704391354802551, + "loss": 0.5403, + "step": 9193 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017043229085828693, + "loss": 0.5421, + "step": 9194 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017042544558146746, + "loss": 0.5504, + "step": 9195 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001704185996498602, + "loss": 0.5243, + "step": 9196 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001704117530635289, + "loss": 0.6448, + "step": 9197 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001704049058225372, + "loss": 0.526, + "step": 9198 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001703980579269487, + "loss": 0.6048, + "step": 9199 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017039120937682717, + "loss": 0.5653, + "step": 9200 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017038436017223618, + "loss": 0.5507, + "step": 9201 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017037751031323949, + "loss": 0.558, + "step": 9202 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017037065979990076, + "loss": 0.5765, + "step": 9203 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017036380863228367, + "loss": 0.6524, + "step": 9204 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017035695681045195, + "loss": 0.6345, + "step": 9205 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017035010433446925, + "loss": 0.4883, + "step": 9206 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017034325120439934, + "loss": 0.5091, + "step": 9207 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001703363974203059, + "loss": 0.4498, + "step": 9208 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001703295429822527, + "loss": 0.5926, + "step": 9209 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017032268789030343, + "loss": 0.6207, + "step": 9210 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017031583214452184, + "loss": 0.5889, + "step": 9211 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017030897574497167, + "loss": 0.5104, + "step": 9212 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017030211869171664, + "loss": 0.5892, + "step": 9213 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017029526098482056, + "loss": 0.5612, + "step": 9214 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017028840262434716, + "loss": 0.5648, + "step": 9215 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001702815436103602, + "loss": 0.5909, + "step": 9216 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017027468394292342, + "loss": 0.5946, + "step": 9217 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017026782362210068, + "loss": 0.6392, + "step": 9218 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001702609626479557, + "loss": 0.5313, + "step": 9219 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017025410102055233, + "loss": 0.5987, + "step": 9220 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001702472387399543, + "loss": 0.5258, + "step": 9221 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017024037580622544, + "loss": 0.5738, + "step": 9222 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017023351221942955, + "loss": 0.5238, + "step": 9223 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017022664797963048, + "loss": 0.6144, + "step": 9224 + }, + { + "epoch": 0.27, + "learning_rate": 0.000170219783086892, + "loss": 0.5881, + "step": 9225 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017021291754127803, + "loss": 0.6817, + "step": 9226 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017020605134285225, + "loss": 0.4873, + "step": 9227 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017019918449167864, + "loss": 0.6453, + "step": 9228 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017019231698782097, + "loss": 0.5689, + "step": 9229 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001701854488313431, + "loss": 0.5645, + "step": 9230 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017017858002230896, + "loss": 0.4841, + "step": 9231 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001701717105607823, + "loss": 0.5533, + "step": 9232 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017016484044682706, + "loss": 0.6496, + "step": 9233 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017015796968050708, + "loss": 0.5937, + "step": 9234 + }, + { + "epoch": 0.27, + "learning_rate": 0.0001701510982618863, + "loss": 0.4604, + "step": 9235 + }, + { + "epoch": 0.27, + "learning_rate": 0.00017014422619102853, + "loss": 0.5742, + "step": 9236 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001701373534679977, + "loss": 0.5768, + "step": 9237 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017013048009285772, + "loss": 0.6813, + "step": 9238 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017012360606567253, + "loss": 0.6501, + "step": 9239 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017011673138650593, + "loss": 0.4963, + "step": 9240 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017010985605542196, + "loss": 0.5215, + "step": 9241 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017010298007248445, + "loss": 0.5899, + "step": 9242 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017009610343775742, + "loss": 0.4998, + "step": 9243 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001700892261513047, + "loss": 0.6088, + "step": 9244 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017008234821319033, + "loss": 0.5036, + "step": 9245 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001700754696234782, + "loss": 0.6068, + "step": 9246 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017006859038223228, + "loss": 0.6398, + "step": 9247 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017006171048951655, + "loss": 0.5009, + "step": 9248 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017005482994539496, + "loss": 0.4959, + "step": 9249 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017004794874993148, + "loss": 0.5457, + "step": 9250 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017004106690319008, + "loss": 0.6034, + "step": 9251 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017003418440523474, + "loss": 0.5306, + "step": 9252 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017002730125612953, + "loss": 0.6072, + "step": 9253 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017002041745593834, + "loss": 0.6697, + "step": 9254 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017001353300472524, + "loss": 0.6279, + "step": 9255 + }, + { + "epoch": 0.28, + "learning_rate": 0.00017000664790255417, + "loss": 0.6235, + "step": 9256 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016999976214948921, + "loss": 0.6242, + "step": 9257 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016999287574559434, + "loss": 0.4722, + "step": 9258 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016998598869093364, + "loss": 0.6073, + "step": 9259 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001699791009855711, + "loss": 0.5977, + "step": 9260 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016997221262957078, + "loss": 0.6527, + "step": 9261 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001699653236229967, + "loss": 0.6482, + "step": 9262 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016995843396591293, + "loss": 0.5116, + "step": 9263 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016995154365838353, + "loss": 0.5602, + "step": 9264 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016994465270047258, + "loss": 0.5638, + "step": 9265 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016993776109224407, + "loss": 0.5618, + "step": 9266 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016993086883376217, + "loss": 0.5536, + "step": 9267 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016992397592509092, + "loss": 0.493, + "step": 9268 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001699170823662944, + "loss": 0.6144, + "step": 9269 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001699101881574367, + "loss": 0.5995, + "step": 9270 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016990329329858196, + "loss": 0.5334, + "step": 9271 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016989639778979427, + "loss": 0.4989, + "step": 9272 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001698895016311377, + "loss": 0.5682, + "step": 9273 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016988260482267645, + "loss": 0.5778, + "step": 9274 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016987570736447453, + "loss": 0.5809, + "step": 9275 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001698688092565962, + "loss": 0.6455, + "step": 9276 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016986191049910544, + "loss": 0.5388, + "step": 9277 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016985501109206653, + "loss": 0.5822, + "step": 9278 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016984811103554356, + "loss": 0.5293, + "step": 9279 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016984121032960072, + "loss": 0.6111, + "step": 9280 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001698343089743021, + "loss": 0.609, + "step": 9281 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016982740696971194, + "loss": 0.5095, + "step": 9282 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016982050431589438, + "loss": 0.5467, + "step": 9283 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016981360101291359, + "loss": 0.604, + "step": 9284 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016980669706083377, + "loss": 0.6097, + "step": 9285 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001697997924597191, + "loss": 0.4998, + "step": 9286 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016979288720963377, + "loss": 0.5291, + "step": 9287 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016978598131064197, + "loss": 0.4842, + "step": 9288 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016977907476280795, + "loss": 0.4799, + "step": 9289 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016977216756619595, + "loss": 0.4957, + "step": 9290 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016976525972087008, + "loss": 0.5464, + "step": 9291 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016975835122689465, + "loss": 0.5563, + "step": 9292 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001697514420843339, + "loss": 0.5737, + "step": 9293 + }, + { + "epoch": 0.28, + "learning_rate": 0.000169744532293252, + "loss": 0.6405, + "step": 9294 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016973762185371326, + "loss": 0.527, + "step": 9295 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001697307107657819, + "loss": 0.5183, + "step": 9296 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016972379902952217, + "loss": 0.6703, + "step": 9297 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016971688664499837, + "loss": 0.532, + "step": 9298 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016970997361227474, + "loss": 0.5781, + "step": 9299 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016970305993141554, + "loss": 0.5984, + "step": 9300 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001696961456024851, + "loss": 0.5049, + "step": 9301 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016968923062554765, + "loss": 0.4454, + "step": 9302 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016968231500066752, + "loss": 0.5316, + "step": 9303 + }, + { + "epoch": 0.28, + "learning_rate": 0.000169675398727909, + "loss": 0.5495, + "step": 9304 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001696684818073364, + "loss": 0.5425, + "step": 9305 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016966156423901398, + "loss": 0.5766, + "step": 9306 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016965464602300615, + "loss": 0.5055, + "step": 9307 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016964772715937718, + "loss": 0.6091, + "step": 9308 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016964080764819136, + "loss": 0.5715, + "step": 9309 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001696338874895131, + "loss": 0.5603, + "step": 9310 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016962696668340672, + "loss": 0.5407, + "step": 9311 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016962004522993654, + "loss": 0.5221, + "step": 9312 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001696131231291669, + "loss": 0.6382, + "step": 9313 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016960620038116223, + "loss": 0.5256, + "step": 9314 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016959927698598684, + "loss": 0.5037, + "step": 9315 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001695923529437051, + "loss": 0.5936, + "step": 9316 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001695854282543814, + "loss": 0.588, + "step": 9317 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016957850291808007, + "loss": 0.6967, + "step": 9318 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016957157693486564, + "loss": 0.5679, + "step": 9319 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016956465030480236, + "loss": 0.591, + "step": 9320 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016955772302795473, + "loss": 0.4898, + "step": 9321 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001695507951043871, + "loss": 0.5382, + "step": 9322 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016954386653416387, + "loss": 0.5583, + "step": 9323 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016953693731734952, + "loss": 0.4943, + "step": 9324 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016953000745400842, + "loss": 0.5103, + "step": 9325 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016952307694420503, + "loss": 0.5436, + "step": 9326 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001695161457880038, + "loss": 0.5421, + "step": 9327 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016950921398546913, + "loss": 0.5466, + "step": 9328 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016950228153666549, + "loss": 0.5215, + "step": 9329 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016949534844165737, + "loss": 0.5594, + "step": 9330 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016948841470050912, + "loss": 0.587, + "step": 9331 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016948148031328537, + "loss": 0.5926, + "step": 9332 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016947454528005044, + "loss": 0.5689, + "step": 9333 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016946760960086894, + "loss": 0.5953, + "step": 9334 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016946067327580525, + "loss": 0.5456, + "step": 9335 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016945373630492392, + "loss": 0.5284, + "step": 9336 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016944679868828943, + "loss": 0.5978, + "step": 9337 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016943986042596628, + "loss": 0.5521, + "step": 9338 + }, + { + "epoch": 0.28, + "learning_rate": 0.000169432921518019, + "loss": 0.6332, + "step": 9339 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016942598196451208, + "loss": 0.6587, + "step": 9340 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016941904176551002, + "loss": 0.5722, + "step": 9341 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016941210092107745, + "loss": 0.6254, + "step": 9342 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016940515943127875, + "loss": 0.5947, + "step": 9343 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001693982172961786, + "loss": 0.5827, + "step": 9344 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016939127451584145, + "loss": 0.4158, + "step": 9345 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001693843310903319, + "loss": 0.5877, + "step": 9346 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001693773870197145, + "loss": 0.5423, + "step": 9347 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001693704423040538, + "loss": 0.5448, + "step": 9348 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016936349694341435, + "loss": 0.6275, + "step": 9349 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016935655093786077, + "loss": 0.6967, + "step": 9350 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016934960428745762, + "loss": 0.5548, + "step": 9351 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001693426569922695, + "loss": 0.4623, + "step": 9352 + }, + { + "epoch": 0.28, + "learning_rate": 0.000169335709052361, + "loss": 0.59, + "step": 9353 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001693287604677967, + "loss": 0.5914, + "step": 9354 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001693218112386412, + "loss": 0.507, + "step": 9355 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016931486136495915, + "loss": 0.6589, + "step": 9356 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016930791084681515, + "loss": 0.5874, + "step": 9357 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001693009596842738, + "loss": 0.5707, + "step": 9358 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016929400787739978, + "loss": 0.5647, + "step": 9359 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016928705542625764, + "loss": 0.6191, + "step": 9360 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016928010233091211, + "loss": 0.5426, + "step": 9361 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001692731485914278, + "loss": 0.5106, + "step": 9362 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016926619420786935, + "loss": 0.5242, + "step": 9363 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016925923918030143, + "loss": 0.6464, + "step": 9364 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001692522835087887, + "loss": 0.5294, + "step": 9365 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001692453271933959, + "loss": 0.5727, + "step": 9366 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001692383702341876, + "loss": 0.6389, + "step": 9367 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016923141263122852, + "loss": 0.5728, + "step": 9368 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001692244543845834, + "loss": 0.6303, + "step": 9369 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016921749549431682, + "loss": 0.6102, + "step": 9370 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001692105359604936, + "loss": 0.5446, + "step": 9371 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001692035757831784, + "loss": 0.5938, + "step": 9372 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016919661496243594, + "loss": 0.5913, + "step": 9373 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016918965349833093, + "loss": 0.5347, + "step": 9374 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001691826913909281, + "loss": 0.5745, + "step": 9375 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016917572864029218, + "loss": 0.6124, + "step": 9376 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001691687652464879, + "loss": 0.595, + "step": 9377 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016916180120958, + "loss": 0.5598, + "step": 9378 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016915483652963324, + "loss": 0.6157, + "step": 9379 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016914787120671237, + "loss": 0.4887, + "step": 9380 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016914090524088217, + "loss": 0.6397, + "step": 9381 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016913393863220735, + "loss": 0.5332, + "step": 9382 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001691269713807528, + "loss": 0.5583, + "step": 9383 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016912000348658315, + "loss": 0.5762, + "step": 9384 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001691130349497633, + "loss": 0.5953, + "step": 9385 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016910606577035797, + "loss": 0.5082, + "step": 9386 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016909909594843198, + "loss": 0.466, + "step": 9387 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016909212548405015, + "loss": 0.528, + "step": 9388 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001690851543772773, + "loss": 0.5836, + "step": 9389 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016907818262817817, + "loss": 0.5651, + "step": 9390 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016907121023681767, + "loss": 0.6006, + "step": 9391 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016906423720326058, + "loss": 0.5728, + "step": 9392 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001690572635275717, + "loss": 0.6717, + "step": 9393 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016905028920981594, + "loss": 0.4968, + "step": 9394 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001690433142500581, + "loss": 0.6529, + "step": 9395 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016903633864836305, + "loss": 0.5153, + "step": 9396 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016902936240479564, + "loss": 0.7082, + "step": 9397 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001690223855194207, + "loss": 0.4916, + "step": 9398 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016901540799230317, + "loss": 0.4986, + "step": 9399 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016900842982350787, + "loss": 0.5158, + "step": 9400 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001690014510130997, + "loss": 0.5239, + "step": 9401 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016899447156114352, + "loss": 0.5944, + "step": 9402 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016898749146770426, + "loss": 0.5362, + "step": 9403 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001689805107328468, + "loss": 0.459, + "step": 9404 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016897352935663605, + "loss": 0.6162, + "step": 9405 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001689665473391369, + "loss": 0.4832, + "step": 9406 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016895956468041426, + "loss": 0.577, + "step": 9407 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001689525813805331, + "loss": 0.6252, + "step": 9408 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016894559743955832, + "loss": 0.5913, + "step": 9409 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016893861285755487, + "loss": 0.5546, + "step": 9410 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016893162763458763, + "loss": 0.6055, + "step": 9411 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016892464177072162, + "loss": 0.6622, + "step": 9412 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016891765526602174, + "loss": 0.6617, + "step": 9413 + }, + { + "epoch": 0.28, + "learning_rate": 0.000168910668120553, + "loss": 0.5266, + "step": 9414 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001689036803343803, + "loss": 0.5384, + "step": 9415 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016889669190756868, + "loss": 0.5041, + "step": 9416 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016888970284018303, + "loss": 0.5717, + "step": 9417 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001688827131322884, + "loss": 0.612, + "step": 9418 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001688757227839498, + "loss": 0.599, + "step": 9419 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016886873179523213, + "loss": 0.4855, + "step": 9420 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016886174016620043, + "loss": 0.6412, + "step": 9421 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016885474789691976, + "loss": 0.5568, + "step": 9422 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016884775498745506, + "loss": 0.536, + "step": 9423 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001688407614378714, + "loss": 0.6092, + "step": 9424 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016883376724823377, + "loss": 0.5565, + "step": 9425 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016882677241860717, + "loss": 0.6184, + "step": 9426 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001688197769490567, + "loss": 0.5505, + "step": 9427 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016881278083964734, + "loss": 0.6174, + "step": 9428 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016880578409044422, + "loss": 0.6562, + "step": 9429 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016879878670151234, + "loss": 0.595, + "step": 9430 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016879178867291675, + "loss": 0.5614, + "step": 9431 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001687847900047225, + "loss": 0.5485, + "step": 9432 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016877779069699472, + "loss": 0.6216, + "step": 9433 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016877079074979843, + "loss": 0.4625, + "step": 9434 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016876379016319873, + "loss": 0.6243, + "step": 9435 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016875678893726076, + "loss": 0.5924, + "step": 9436 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016874978707204954, + "loss": 0.5746, + "step": 9437 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001687427845676302, + "loss": 0.5446, + "step": 9438 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016873578142406786, + "loss": 0.6352, + "step": 9439 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001687287776414276, + "loss": 0.6143, + "step": 9440 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016872177321977457, + "loss": 0.5874, + "step": 9441 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001687147681591739, + "loss": 0.5301, + "step": 9442 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001687077624596907, + "loss": 0.5623, + "step": 9443 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016870075612139006, + "loss": 0.4729, + "step": 9444 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016869374914433723, + "loss": 0.5966, + "step": 9445 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016868674152859726, + "loss": 0.5226, + "step": 9446 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016867973327423536, + "loss": 0.517, + "step": 9447 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016867272438131668, + "loss": 0.4716, + "step": 9448 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016866571484990637, + "loss": 0.5366, + "step": 9449 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001686587046800696, + "loss": 0.6049, + "step": 9450 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016865169387187158, + "loss": 0.5057, + "step": 9451 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016864468242537746, + "loss": 0.575, + "step": 9452 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016863767034065244, + "loss": 0.5271, + "step": 9453 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016863065761776172, + "loss": 0.5171, + "step": 9454 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016862364425677048, + "loss": 0.6057, + "step": 9455 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016861663025774396, + "loss": 0.5342, + "step": 9456 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001686096156207474, + "loss": 0.4827, + "step": 9457 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016860260034584594, + "loss": 0.6271, + "step": 9458 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016859558443310486, + "loss": 0.6058, + "step": 9459 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016858856788258937, + "loss": 0.6388, + "step": 9460 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001685815506943647, + "loss": 0.4727, + "step": 9461 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001685745328684961, + "loss": 0.5035, + "step": 9462 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016856751440504883, + "loss": 0.5019, + "step": 9463 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016856049530408816, + "loss": 0.5297, + "step": 9464 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001685534755656793, + "loss": 0.563, + "step": 9465 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016854645518988757, + "loss": 0.6422, + "step": 9466 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001685394341767782, + "loss": 0.5232, + "step": 9467 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016853241252641653, + "loss": 0.5143, + "step": 9468 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016852539023886774, + "loss": 0.6173, + "step": 9469 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016851836731419722, + "loss": 0.6645, + "step": 9470 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016851134375247024, + "loss": 0.4725, + "step": 9471 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016850431955375207, + "loss": 0.5321, + "step": 9472 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016849729471810806, + "loss": 0.5544, + "step": 9473 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001684902692456035, + "loss": 0.6671, + "step": 9474 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001684832431363037, + "loss": 0.5761, + "step": 9475 + }, + { + "epoch": 0.28, + "learning_rate": 0.000168476216390274, + "loss": 0.5206, + "step": 9476 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016846918900757972, + "loss": 0.5221, + "step": 9477 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016846216098828623, + "loss": 0.5315, + "step": 9478 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016845513233245887, + "loss": 0.6116, + "step": 9479 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016844810304016296, + "loss": 0.6183, + "step": 9480 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016844107311146385, + "loss": 0.589, + "step": 9481 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016843404254642693, + "loss": 0.5703, + "step": 9482 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016842701134511757, + "loss": 0.6236, + "step": 9483 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016841997950760114, + "loss": 0.5231, + "step": 9484 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016841294703394297, + "loss": 0.4671, + "step": 9485 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001684059139242085, + "loss": 0.5144, + "step": 9486 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016839888017846312, + "loss": 0.6531, + "step": 9487 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016839184579677224, + "loss": 0.5756, + "step": 9488 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016838481077920117, + "loss": 0.4128, + "step": 9489 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001683777751258154, + "loss": 0.4556, + "step": 9490 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016837073883668038, + "loss": 0.5795, + "step": 9491 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016836370191186144, + "loss": 0.5608, + "step": 9492 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016835666435142406, + "loss": 0.5219, + "step": 9493 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016834962615543364, + "loss": 0.5365, + "step": 9494 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016834258732395565, + "loss": 0.5272, + "step": 9495 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016833554785705552, + "loss": 0.516, + "step": 9496 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016832850775479872, + "loss": 0.6082, + "step": 9497 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016832146701725064, + "loss": 0.5047, + "step": 9498 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016831442564447686, + "loss": 0.5492, + "step": 9499 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016830738363654273, + "loss": 0.5432, + "step": 9500 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016830034099351378, + "loss": 0.6333, + "step": 9501 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016829329771545547, + "loss": 0.6017, + "step": 9502 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016828625380243333, + "loss": 0.5793, + "step": 9503 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016827920925451277, + "loss": 0.5875, + "step": 9504 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016827216407175937, + "loss": 0.6297, + "step": 9505 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001682651182542386, + "loss": 0.4965, + "step": 9506 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016825807180201595, + "loss": 0.6063, + "step": 9507 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016825102471515696, + "loss": 0.4449, + "step": 9508 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016824397699372715, + "loss": 0.4061, + "step": 9509 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016823692863779206, + "loss": 0.6579, + "step": 9510 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016822987964741718, + "loss": 0.6395, + "step": 9511 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016822283002266807, + "loss": 0.6428, + "step": 9512 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016821577976361028, + "loss": 0.5436, + "step": 9513 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016820872887030936, + "loss": 0.5838, + "step": 9514 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016820167734283087, + "loss": 0.7217, + "step": 9515 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016819462518124036, + "loss": 0.5918, + "step": 9516 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001681875723856034, + "loss": 0.5123, + "step": 9517 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016818051895598556, + "loss": 0.6275, + "step": 9518 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001681734648924525, + "loss": 0.4948, + "step": 9519 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016816641019506965, + "loss": 0.564, + "step": 9520 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001681593548639027, + "loss": 0.6182, + "step": 9521 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016815229889901727, + "loss": 0.6146, + "step": 9522 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001681452423004789, + "loss": 0.5017, + "step": 9523 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016813818506835327, + "loss": 0.5784, + "step": 9524 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001681311272027059, + "loss": 0.5697, + "step": 9525 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016812406870360248, + "loss": 0.6596, + "step": 9526 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001681170095711086, + "loss": 0.4987, + "step": 9527 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016810994980528995, + "loss": 0.5686, + "step": 9528 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016810288940621212, + "loss": 0.523, + "step": 9529 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016809582837394075, + "loss": 0.5625, + "step": 9530 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001680887667085415, + "loss": 0.6451, + "step": 9531 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016808170441008012, + "loss": 0.5786, + "step": 9532 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001680746414786221, + "loss": 0.5211, + "step": 9533 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016806757791423325, + "loss": 0.6307, + "step": 9534 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016806051371697913, + "loss": 0.513, + "step": 9535 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016805344888692553, + "loss": 0.6303, + "step": 9536 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016804638342413804, + "loss": 0.5763, + "step": 9537 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016803931732868244, + "loss": 0.4772, + "step": 9538 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016803225060062435, + "loss": 0.5847, + "step": 9539 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016802518324002955, + "loss": 0.6496, + "step": 9540 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016801811524696367, + "loss": 0.4977, + "step": 9541 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001680110466214925, + "loss": 0.6155, + "step": 9542 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016800397736368167, + "loss": 0.4642, + "step": 9543 + }, + { + "epoch": 0.28, + "learning_rate": 0.000167996907473597, + "loss": 0.578, + "step": 9544 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001679898369513042, + "loss": 0.595, + "step": 9545 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016798276579686898, + "loss": 0.5841, + "step": 9546 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016797569401035708, + "loss": 0.5294, + "step": 9547 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001679686215918343, + "loss": 0.5163, + "step": 9548 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016796154854136636, + "loss": 0.5288, + "step": 9549 + }, + { + "epoch": 0.28, + "learning_rate": 0.000167954474859019, + "loss": 0.6203, + "step": 9550 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016794740054485805, + "loss": 0.5335, + "step": 9551 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016794032559894925, + "loss": 0.5389, + "step": 9552 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016793325002135835, + "loss": 0.7254, + "step": 9553 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016792617381215119, + "loss": 0.4822, + "step": 9554 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016791909697139352, + "loss": 0.6599, + "step": 9555 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016791201949915117, + "loss": 0.5047, + "step": 9556 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016790494139548993, + "loss": 0.5235, + "step": 9557 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001678978626604756, + "loss": 0.6257, + "step": 9558 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016789078329417404, + "loss": 0.5254, + "step": 9559 + }, + { + "epoch": 0.28, + "learning_rate": 0.000167883703296651, + "loss": 0.5995, + "step": 9560 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016787662266797235, + "loss": 0.5207, + "step": 9561 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016786954140820393, + "loss": 0.6135, + "step": 9562 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016786245951741155, + "loss": 0.5086, + "step": 9563 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016785537699566105, + "loss": 0.5092, + "step": 9564 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016784829384301831, + "loss": 0.5613, + "step": 9565 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001678412100595492, + "loss": 0.5526, + "step": 9566 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016783412564531954, + "loss": 0.5343, + "step": 9567 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016782704060039523, + "loss": 0.5586, + "step": 9568 + }, + { + "epoch": 0.28, + "learning_rate": 0.0001678199549248421, + "loss": 0.7657, + "step": 9569 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016781286861872606, + "loss": 0.5428, + "step": 9570 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016780578168211304, + "loss": 0.6055, + "step": 9571 + }, + { + "epoch": 0.28, + "learning_rate": 0.00016779869411506884, + "loss": 0.5485, + "step": 9572 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016779160591765942, + "loss": 0.4679, + "step": 9573 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016778451708995067, + "loss": 0.5046, + "step": 9574 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001677774276320085, + "loss": 0.4987, + "step": 9575 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001677703375438988, + "loss": 0.5165, + "step": 9576 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016776324682568755, + "loss": 0.5949, + "step": 9577 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016775615547744063, + "loss": 0.5768, + "step": 9578 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016774906349922394, + "loss": 0.6362, + "step": 9579 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001677419708911035, + "loss": 0.5323, + "step": 9580 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001677348776531452, + "loss": 0.5879, + "step": 9581 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016772778378541502, + "loss": 0.6185, + "step": 9582 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016772068928797888, + "loss": 0.6268, + "step": 9583 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016771359416090274, + "loss": 0.6355, + "step": 9584 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016770649840425265, + "loss": 0.5197, + "step": 9585 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016769940201809448, + "loss": 0.6287, + "step": 9586 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016769230500249428, + "loss": 0.5557, + "step": 9587 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016768520735751797, + "loss": 0.5966, + "step": 9588 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001676781090832316, + "loss": 0.6008, + "step": 9589 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016767101017970117, + "loss": 0.5543, + "step": 9590 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016766391064699263, + "loss": 0.6164, + "step": 9591 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016765681048517203, + "loss": 0.4797, + "step": 9592 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016764970969430535, + "loss": 0.5554, + "step": 9593 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016764260827445865, + "loss": 0.6348, + "step": 9594 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016763550622569793, + "loss": 0.5856, + "step": 9595 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016762840354808925, + "loss": 0.6393, + "step": 9596 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001676213002416986, + "loss": 0.4993, + "step": 9597 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016761419630659206, + "loss": 0.5896, + "step": 9598 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016760709174283567, + "loss": 0.5315, + "step": 9599 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001675999865504955, + "loss": 0.6037, + "step": 9600 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016759288072963754, + "loss": 0.5275, + "step": 9601 + }, + { + "epoch": 0.29, + "learning_rate": 0.000167585774280328, + "loss": 0.6217, + "step": 9602 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001675786672026328, + "loss": 0.5159, + "step": 9603 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016757155949661813, + "loss": 0.5869, + "step": 9604 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016756445116235003, + "loss": 0.4869, + "step": 9605 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016755734219989457, + "loss": 0.5818, + "step": 9606 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016755023260931785, + "loss": 0.5376, + "step": 9607 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016754312239068602, + "loss": 0.5413, + "step": 9608 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016753601154406519, + "loss": 0.668, + "step": 9609 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001675289000695214, + "loss": 0.5518, + "step": 9610 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001675217879671208, + "loss": 0.5384, + "step": 9611 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016751467523692956, + "loss": 0.5222, + "step": 9612 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016750756187901376, + "loss": 0.5761, + "step": 9613 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016750044789343957, + "loss": 0.6547, + "step": 9614 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001674933332802731, + "loss": 0.5005, + "step": 9615 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001674862180395805, + "loss": 0.6046, + "step": 9616 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016747910217142798, + "loss": 0.5452, + "step": 9617 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016747198567588164, + "loss": 0.5858, + "step": 9618 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016746486855300768, + "loss": 0.5009, + "step": 9619 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016745775080287227, + "loss": 0.5639, + "step": 9620 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016745063242554155, + "loss": 0.5463, + "step": 9621 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016744351342108177, + "loss": 0.5301, + "step": 9622 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016743639378955907, + "loss": 0.5329, + "step": 9623 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016742927353103964, + "loss": 0.4577, + "step": 9624 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016742215264558974, + "loss": 0.6517, + "step": 9625 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016741503113327548, + "loss": 0.5062, + "step": 9626 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016740790899416318, + "loss": 0.578, + "step": 9627 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016740078622831897, + "loss": 0.4631, + "step": 9628 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016739366283580914, + "loss": 0.6118, + "step": 9629 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001673865388166999, + "loss": 0.6477, + "step": 9630 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016737941417105745, + "loss": 0.5252, + "step": 9631 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016737228889894805, + "loss": 0.508, + "step": 9632 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016736516300043798, + "loss": 0.5113, + "step": 9633 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016735803647559348, + "loss": 0.5285, + "step": 9634 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016735090932448081, + "loss": 0.5792, + "step": 9635 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001673437815471662, + "loss": 0.5839, + "step": 9636 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016733665314371598, + "loss": 0.6286, + "step": 9637 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016732952411419638, + "loss": 0.577, + "step": 9638 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001673223944586737, + "loss": 0.6007, + "step": 9639 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016731526417721425, + "loss": 0.5559, + "step": 9640 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016730813326988427, + "loss": 0.4589, + "step": 9641 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016730100173675008, + "loss": 0.4944, + "step": 9642 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016729386957787807, + "loss": 0.4965, + "step": 9643 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016728673679333443, + "loss": 0.5727, + "step": 9644 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016727960338318556, + "loss": 0.4812, + "step": 9645 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001672724693474977, + "loss": 0.5993, + "step": 9646 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016726533468633728, + "loss": 0.4588, + "step": 9647 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016725819939977056, + "loss": 0.5128, + "step": 9648 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016725106348786393, + "loss": 0.4942, + "step": 9649 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016724392695068372, + "loss": 0.5139, + "step": 9650 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016723678978829625, + "loss": 0.4771, + "step": 9651 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001672296520007679, + "loss": 0.5498, + "step": 9652 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016722251358816507, + "loss": 0.5261, + "step": 9653 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016721537455055408, + "loss": 0.5337, + "step": 9654 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016720823488800134, + "loss": 0.5331, + "step": 9655 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016720109460057324, + "loss": 0.7094, + "step": 9656 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001671939536883361, + "loss": 0.5319, + "step": 9657 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016718681215135641, + "loss": 0.5353, + "step": 9658 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016717966998970046, + "loss": 0.5619, + "step": 9659 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016717252720343476, + "loss": 0.5978, + "step": 9660 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016716538379262565, + "loss": 0.5815, + "step": 9661 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001671582397573396, + "loss": 0.5352, + "step": 9662 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016715109509764296, + "loss": 0.6727, + "step": 9663 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016714394981360225, + "loss": 0.6542, + "step": 9664 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001671368039052838, + "loss": 0.59, + "step": 9665 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016712965737275415, + "loss": 0.4694, + "step": 9666 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001671225102160797, + "loss": 0.5786, + "step": 9667 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016711536243532686, + "loss": 0.5996, + "step": 9668 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016710821403056217, + "loss": 0.5029, + "step": 9669 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016710106500185206, + "loss": 0.6447, + "step": 9670 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016709391534926296, + "loss": 0.4758, + "step": 9671 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001670867650728614, + "loss": 0.5795, + "step": 9672 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016707961417271382, + "loss": 0.7074, + "step": 9673 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016707246264888672, + "loss": 0.5383, + "step": 9674 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001670653105014466, + "loss": 0.5386, + "step": 9675 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016705815773045997, + "loss": 0.5909, + "step": 9676 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016705100433599327, + "loss": 0.4885, + "step": 9677 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016704385031811311, + "loss": 0.5502, + "step": 9678 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016703669567688593, + "loss": 0.6606, + "step": 9679 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016702954041237825, + "loss": 0.5116, + "step": 9680 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016702238452465665, + "loss": 0.5203, + "step": 9681 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001670152280137876, + "loss": 0.6995, + "step": 9682 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001670080708798377, + "loss": 0.5891, + "step": 9683 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016700091312287348, + "loss": 0.5455, + "step": 9684 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016699375474296143, + "loss": 0.5753, + "step": 9685 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016698659574016818, + "loss": 0.6118, + "step": 9686 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016697943611456025, + "loss": 0.4911, + "step": 9687 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001669722758662042, + "loss": 0.5948, + "step": 9688 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016696511499516664, + "loss": 0.5514, + "step": 9689 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016695795350151412, + "loss": 0.5896, + "step": 9690 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016695079138531326, + "loss": 0.5896, + "step": 9691 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001669436286466306, + "loss": 0.5533, + "step": 9692 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016693646528553275, + "loss": 0.6588, + "step": 9693 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016692930130208635, + "loss": 0.5552, + "step": 9694 + }, + { + "epoch": 0.29, + "learning_rate": 0.000166922136696358, + "loss": 0.628, + "step": 9695 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016691497146841425, + "loss": 0.5596, + "step": 9696 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001669078056183218, + "loss": 0.4811, + "step": 9697 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001669006391461472, + "loss": 0.4838, + "step": 9698 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001668934720519572, + "loss": 0.5398, + "step": 9699 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016688630433581828, + "loss": 0.6423, + "step": 9700 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016687913599779716, + "loss": 0.5925, + "step": 9701 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016687196703796053, + "loss": 0.4749, + "step": 9702 + }, + { + "epoch": 0.29, + "learning_rate": 0.000166864797456375, + "loss": 0.5612, + "step": 9703 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001668576272531072, + "loss": 0.6556, + "step": 9704 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016685045642822385, + "loss": 0.684, + "step": 9705 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001668432849817916, + "loss": 0.5659, + "step": 9706 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016683611291387713, + "loss": 0.5971, + "step": 9707 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001668289402245471, + "loss": 0.4838, + "step": 9708 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016682176691386826, + "loss": 0.5245, + "step": 9709 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016681459298190725, + "loss": 0.6841, + "step": 9710 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016680741842873075, + "loss": 0.6614, + "step": 9711 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016680024325440554, + "loss": 0.4864, + "step": 9712 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016679306745899828, + "loss": 0.6119, + "step": 9713 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016678589104257568, + "loss": 0.5591, + "step": 9714 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016677871400520454, + "loss": 0.5721, + "step": 9715 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016677153634695146, + "loss": 0.551, + "step": 9716 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016676435806788328, + "loss": 0.5707, + "step": 9717 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001667571791680667, + "loss": 0.5279, + "step": 9718 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001667499996475685, + "loss": 0.5037, + "step": 9719 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001667428195064554, + "loss": 0.5288, + "step": 9720 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016673563874479414, + "loss": 0.6319, + "step": 9721 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016672845736265152, + "loss": 0.6025, + "step": 9722 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001667212753600943, + "loss": 0.6204, + "step": 9723 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016671409273718928, + "loss": 0.6376, + "step": 9724 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001667069094940032, + "loss": 0.5166, + "step": 9725 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016669972563060283, + "loss": 0.4864, + "step": 9726 + }, + { + "epoch": 0.29, + "learning_rate": 0.000166692541147055, + "loss": 0.5905, + "step": 9727 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016668535604342656, + "loss": 0.6712, + "step": 9728 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001666781703197842, + "loss": 0.5712, + "step": 9729 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016667098397619482, + "loss": 0.5353, + "step": 9730 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016666379701272519, + "loss": 0.5416, + "step": 9731 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016665660942944216, + "loss": 0.5142, + "step": 9732 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016664942122641252, + "loss": 0.5908, + "step": 9733 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016664223240370313, + "loss": 0.4488, + "step": 9734 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016663504296138084, + "loss": 0.593, + "step": 9735 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001666278528995125, + "loss": 0.5289, + "step": 9736 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016662066221816488, + "loss": 0.5599, + "step": 9737 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016661347091740492, + "loss": 0.4655, + "step": 9738 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016660627899729944, + "loss": 0.7034, + "step": 9739 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016659908645791537, + "loss": 0.5042, + "step": 9740 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001665918932993195, + "loss": 0.5832, + "step": 9741 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016658469952157877, + "loss": 0.5949, + "step": 9742 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016657750512476, + "loss": 0.4546, + "step": 9743 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016657031010893015, + "loss": 0.5066, + "step": 9744 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001665631144741561, + "loss": 0.5464, + "step": 9745 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016655591822050476, + "loss": 0.5955, + "step": 9746 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016654872134804296, + "loss": 0.5428, + "step": 9747 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001665415238568377, + "loss": 0.6967, + "step": 9748 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016653432574695591, + "loss": 0.5436, + "step": 9749 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016652712701846443, + "loss": 0.8166, + "step": 9750 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016651992767143024, + "loss": 0.4991, + "step": 9751 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001665127277059203, + "loss": 0.5831, + "step": 9752 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016650552712200152, + "loss": 0.5204, + "step": 9753 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016649832591974083, + "loss": 0.5537, + "step": 9754 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016649112409920525, + "loss": 0.5249, + "step": 9755 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016648392166046167, + "loss": 0.4465, + "step": 9756 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001664767186035771, + "loss": 0.5751, + "step": 9757 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001664695149286185, + "loss": 0.6207, + "step": 9758 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001664623106356528, + "loss": 0.665, + "step": 9759 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016645510572474708, + "loss": 0.5748, + "step": 9760 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016644790019596824, + "loss": 0.5172, + "step": 9761 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001664406940493833, + "loss": 0.6, + "step": 9762 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001664334872850593, + "loss": 0.5449, + "step": 9763 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001664262799030632, + "loss": 0.6208, + "step": 9764 + }, + { + "epoch": 0.29, + "learning_rate": 0.000166419071903462, + "loss": 0.7002, + "step": 9765 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016641186328632273, + "loss": 0.5972, + "step": 9766 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001664046540517125, + "loss": 0.5979, + "step": 9767 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001663974441996982, + "loss": 0.5886, + "step": 9768 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016639023373034694, + "loss": 0.5446, + "step": 9769 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016638302264372574, + "loss": 0.5421, + "step": 9770 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001663758109399017, + "loss": 0.5015, + "step": 9771 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001663685986189418, + "loss": 0.5494, + "step": 9772 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016636138568091308, + "loss": 0.5955, + "step": 9773 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016635417212588268, + "loss": 0.5675, + "step": 9774 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016634695795391763, + "loss": 0.4791, + "step": 9775 + }, + { + "epoch": 0.29, + "learning_rate": 0.000166339743165085, + "loss": 0.5389, + "step": 9776 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016633252775945188, + "loss": 0.4995, + "step": 9777 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016632531173708535, + "loss": 0.4558, + "step": 9778 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016631809509805255, + "loss": 0.5384, + "step": 9779 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001663108778424205, + "loss": 0.5553, + "step": 9780 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016630365997025634, + "loss": 0.6308, + "step": 9781 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016629644148162718, + "loss": 0.6614, + "step": 9782 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016628922237660012, + "loss": 0.6333, + "step": 9783 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001662820026552423, + "loss": 0.5593, + "step": 9784 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016627478231762081, + "loss": 0.6073, + "step": 9785 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016626756136380285, + "loss": 0.4686, + "step": 9786 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016626033979385548, + "loss": 0.6294, + "step": 9787 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016625311760784593, + "loss": 0.5343, + "step": 9788 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016624589480584122, + "loss": 0.5495, + "step": 9789 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016623867138790868, + "loss": 0.5397, + "step": 9790 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001662314473541153, + "loss": 0.5186, + "step": 9791 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016622422270452835, + "loss": 0.4967, + "step": 9792 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016621699743921492, + "loss": 0.5325, + "step": 9793 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001662097715582423, + "loss": 0.6119, + "step": 9794 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001662025450616776, + "loss": 0.5651, + "step": 9795 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016619531794958802, + "loss": 0.6426, + "step": 9796 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016618809022204075, + "loss": 0.6891, + "step": 9797 + }, + { + "epoch": 0.29, + "learning_rate": 0.000166180861879103, + "loss": 0.5977, + "step": 9798 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016617363292084197, + "loss": 0.5893, + "step": 9799 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016616640334732487, + "loss": 0.5911, + "step": 9800 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016615917315861896, + "loss": 0.5108, + "step": 9801 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016615194235479138, + "loss": 0.5417, + "step": 9802 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001661447109359094, + "loss": 0.4828, + "step": 9803 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016613747890204032, + "loss": 0.5403, + "step": 9804 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016613024625325125, + "loss": 0.6785, + "step": 9805 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016612301298960956, + "loss": 0.5185, + "step": 9806 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016611577911118243, + "loss": 0.5539, + "step": 9807 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016610854461803713, + "loss": 0.5617, + "step": 9808 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016610130951024096, + "loss": 0.6267, + "step": 9809 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016609407378786111, + "loss": 0.6222, + "step": 9810 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016608683745096493, + "loss": 0.6196, + "step": 9811 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001660796004996197, + "loss": 0.6585, + "step": 9812 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016607236293389265, + "loss": 0.6609, + "step": 9813 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016606512475385112, + "loss": 0.6121, + "step": 9814 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016605788595956239, + "loss": 0.6032, + "step": 9815 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016605064655109375, + "loss": 0.525, + "step": 9816 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016604340652851254, + "loss": 0.522, + "step": 9817 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016603616589188605, + "loss": 0.6567, + "step": 9818 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001660289246412816, + "loss": 0.591, + "step": 9819 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016602168277676652, + "loss": 0.6341, + "step": 9820 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001660144402984082, + "loss": 0.6592, + "step": 9821 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016600719720627388, + "loss": 0.5044, + "step": 9822 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016599995350043094, + "loss": 0.6286, + "step": 9823 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016599270918094675, + "loss": 0.5679, + "step": 9824 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016598546424788864, + "loss": 0.5712, + "step": 9825 + }, + { + "epoch": 0.29, + "learning_rate": 0.000165978218701324, + "loss": 0.5241, + "step": 9826 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001659709725413202, + "loss": 0.4493, + "step": 9827 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016596372576794453, + "loss": 0.5673, + "step": 9828 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016595647838126446, + "loss": 0.6438, + "step": 9829 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001659492303813474, + "loss": 0.5267, + "step": 9830 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001659419817682606, + "loss": 0.481, + "step": 9831 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016593473254207156, + "loss": 0.5527, + "step": 9832 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016592748270284765, + "loss": 0.5634, + "step": 9833 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001659202322506563, + "loss": 0.6049, + "step": 9834 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001659129811855649, + "loss": 0.5731, + "step": 9835 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016590572950764084, + "loss": 0.6335, + "step": 9836 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016589847721695163, + "loss": 0.6374, + "step": 9837 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001658912243135646, + "loss": 0.5386, + "step": 9838 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016588397079754725, + "loss": 0.633, + "step": 9839 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016587671666896698, + "loss": 0.5675, + "step": 9840 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016586946192789122, + "loss": 0.5054, + "step": 9841 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016586220657438752, + "loss": 0.5779, + "step": 9842 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016585495060852327, + "loss": 0.5894, + "step": 9843 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016584769403036592, + "loss": 0.5768, + "step": 9844 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016584043683998294, + "loss": 0.5452, + "step": 9845 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016583317903744185, + "loss": 0.4712, + "step": 9846 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016582592062281006, + "loss": 0.5309, + "step": 9847 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016581866159615514, + "loss": 0.5938, + "step": 9848 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001658114019575445, + "loss": 0.6671, + "step": 9849 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016580414170704567, + "loss": 0.5548, + "step": 9850 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001657968808447262, + "loss": 0.548, + "step": 9851 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016578961937065354, + "loss": 0.5868, + "step": 9852 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016578235728489522, + "loss": 0.6163, + "step": 9853 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016577509458751876, + "loss": 0.4737, + "step": 9854 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016576783127859166, + "loss": 0.4901, + "step": 9855 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001657605673581815, + "loss": 0.4678, + "step": 9856 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001657533028263558, + "loss": 0.5949, + "step": 9857 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016574603768318208, + "loss": 0.5787, + "step": 9858 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001657387719287279, + "loss": 0.5359, + "step": 9859 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016573150556306086, + "loss": 0.5696, + "step": 9860 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016572423858624843, + "loss": 0.6453, + "step": 9861 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016571697099835827, + "loss": 0.589, + "step": 9862 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016570970279945787, + "loss": 0.5923, + "step": 9863 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016570243398961488, + "loss": 0.5451, + "step": 9864 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001656951645688968, + "loss": 0.5027, + "step": 9865 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016568789453737132, + "loss": 0.5455, + "step": 9866 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001656806238951059, + "loss": 0.5993, + "step": 9867 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016567335264216829, + "loss": 0.5757, + "step": 9868 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016566608077862597, + "loss": 0.518, + "step": 9869 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016565880830454662, + "loss": 0.5466, + "step": 9870 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016565153521999783, + "loss": 0.5394, + "step": 9871 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016564426152504725, + "loss": 0.5379, + "step": 9872 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001656369872197625, + "loss": 0.5181, + "step": 9873 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016562971230421118, + "loss": 0.5521, + "step": 9874 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016562243677846095, + "loss": 0.6481, + "step": 9875 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016561516064257949, + "loss": 0.5467, + "step": 9876 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001656078838966344, + "loss": 0.6038, + "step": 9877 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016560060654069334, + "loss": 0.5528, + "step": 9878 + }, + { + "epoch": 0.29, + "learning_rate": 0.000165593328574824, + "loss": 0.4822, + "step": 9879 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016558604999909401, + "loss": 0.5845, + "step": 9880 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016557877081357109, + "loss": 0.693, + "step": 9881 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001655714910183229, + "loss": 0.5734, + "step": 9882 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001655642106134171, + "loss": 0.5526, + "step": 9883 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001655569295989214, + "loss": 0.5293, + "step": 9884 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016554964797490355, + "loss": 0.5633, + "step": 9885 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016554236574143114, + "loss": 0.5675, + "step": 9886 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016553508289857195, + "loss": 0.5486, + "step": 9887 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001655277994463937, + "loss": 0.555, + "step": 9888 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016552051538496406, + "loss": 0.5631, + "step": 9889 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001655132307143508, + "loss": 0.5171, + "step": 9890 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016550594543462163, + "loss": 0.6638, + "step": 9891 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016549865954584427, + "loss": 0.5088, + "step": 9892 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001654913730480865, + "loss": 0.6234, + "step": 9893 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016548408594141604, + "loss": 0.5628, + "step": 9894 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016547679822590065, + "loss": 0.6163, + "step": 9895 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001654695099016081, + "loss": 0.542, + "step": 9896 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001654622209686061, + "loss": 0.5573, + "step": 9897 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001654549314269625, + "loss": 0.5785, + "step": 9898 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016544764127674502, + "loss": 0.5862, + "step": 9899 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016544035051802144, + "loss": 0.5697, + "step": 9900 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016543305915085957, + "loss": 0.4906, + "step": 9901 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001654257671753272, + "loss": 0.6337, + "step": 9902 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016541847459149213, + "loss": 0.5018, + "step": 9903 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016541118139942217, + "loss": 0.6941, + "step": 9904 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001654038875991851, + "loss": 0.606, + "step": 9905 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016539659319084874, + "loss": 0.5504, + "step": 9906 + }, + { + "epoch": 0.29, + "learning_rate": 0.00016538929817448095, + "loss": 0.5683, + "step": 9907 + }, + { + "epoch": 0.29, + "learning_rate": 0.0001653820025501495, + "loss": 0.551, + "step": 9908 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016537470631792223, + "loss": 0.5406, + "step": 9909 + }, + { + "epoch": 0.3, + "learning_rate": 0.000165367409477867, + "loss": 0.5721, + "step": 9910 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016536011203005171, + "loss": 0.5954, + "step": 9911 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016535281397454408, + "loss": 0.611, + "step": 9912 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001653455153114121, + "loss": 0.6122, + "step": 9913 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016533821604072351, + "loss": 0.5322, + "step": 9914 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016533091616254626, + "loss": 0.5368, + "step": 9915 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016532361567694816, + "loss": 0.5778, + "step": 9916 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016531631458399713, + "loss": 0.5828, + "step": 9917 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016530901288376107, + "loss": 0.5235, + "step": 9918 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016530171057630778, + "loss": 0.4549, + "step": 9919 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016529440766170525, + "loss": 0.4932, + "step": 9920 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016528710414002133, + "loss": 0.483, + "step": 9921 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016527980001132393, + "loss": 0.6083, + "step": 9922 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016527249527568095, + "loss": 0.5033, + "step": 9923 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016526518993316037, + "loss": 0.5613, + "step": 9924 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016525788398383004, + "loss": 0.487, + "step": 9925 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016525057742775793, + "loss": 0.5295, + "step": 9926 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016524327026501195, + "loss": 0.5594, + "step": 9927 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016523596249566004, + "loss": 0.57, + "step": 9928 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016522865411977016, + "loss": 0.5484, + "step": 9929 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016522134513741022, + "loss": 0.5361, + "step": 9930 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016521403554864823, + "loss": 0.5107, + "step": 9931 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016520672535355214, + "loss": 0.5266, + "step": 9932 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001651994145521899, + "loss": 0.5261, + "step": 9933 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001651921031446295, + "loss": 0.437, + "step": 9934 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016518479113093889, + "loss": 0.5105, + "step": 9935 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016517747851118607, + "loss": 0.4828, + "step": 9936 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016517016528543906, + "loss": 0.482, + "step": 9937 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001651628514537658, + "loss": 0.5587, + "step": 9938 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016515553701623432, + "loss": 0.4895, + "step": 9939 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016514822197291262, + "loss": 0.5746, + "step": 9940 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016514090632386875, + "loss": 0.579, + "step": 9941 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016513359006917067, + "loss": 0.5583, + "step": 9942 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016512627320888643, + "loss": 0.5031, + "step": 9943 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016511895574308408, + "loss": 0.5442, + "step": 9944 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001651116376718316, + "loss": 0.5643, + "step": 9945 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016510431899519708, + "loss": 0.5118, + "step": 9946 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016509699971324854, + "loss": 0.5815, + "step": 9947 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016508967982605408, + "loss": 0.5664, + "step": 9948 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016508235933368167, + "loss": 0.537, + "step": 9949 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016507503823619946, + "loss": 0.5014, + "step": 9950 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016506771653367546, + "loss": 0.4553, + "step": 9951 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016506039422617777, + "loss": 0.5884, + "step": 9952 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016505307131377449, + "loss": 0.479, + "step": 9953 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001650457477965336, + "loss": 0.5326, + "step": 9954 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016503842367452336, + "loss": 0.5753, + "step": 9955 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016503109894781175, + "loss": 0.5828, + "step": 9956 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016502377361646688, + "loss": 0.4536, + "step": 9957 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001650164476805569, + "loss": 0.5632, + "step": 9958 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001650091211401499, + "loss": 0.6747, + "step": 9959 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016500179399531398, + "loss": 0.595, + "step": 9960 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016499446624611733, + "loss": 0.6152, + "step": 9961 + }, + { + "epoch": 0.3, + "learning_rate": 0.000164987137892628, + "loss": 0.5644, + "step": 9962 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001649798089349142, + "loss": 0.5483, + "step": 9963 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016497247937304396, + "loss": 0.5807, + "step": 9964 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016496514920708556, + "loss": 0.5641, + "step": 9965 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001649578184371071, + "loss": 0.5264, + "step": 9966 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016495048706317674, + "loss": 0.5951, + "step": 9967 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001649431550853626, + "loss": 0.4854, + "step": 9968 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001649358225037329, + "loss": 0.4904, + "step": 9969 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001649284893183558, + "loss": 0.5279, + "step": 9970 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001649211555292995, + "loss": 0.4442, + "step": 9971 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016491382113663217, + "loss": 0.5347, + "step": 9972 + }, + { + "epoch": 0.3, + "learning_rate": 0.000164906486140422, + "loss": 0.6043, + "step": 9973 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001648991505407372, + "loss": 0.5365, + "step": 9974 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016489181433764594, + "loss": 0.5728, + "step": 9975 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016488447753121646, + "loss": 0.5885, + "step": 9976 + }, + { + "epoch": 0.3, + "learning_rate": 0.000164877140121517, + "loss": 0.5728, + "step": 9977 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016486980210861572, + "loss": 0.5422, + "step": 9978 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016486246349258087, + "loss": 0.5255, + "step": 9979 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016485512427348069, + "loss": 0.5724, + "step": 9980 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016484778445138344, + "loss": 0.4637, + "step": 9981 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001648404440263573, + "loss": 0.5913, + "step": 9982 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016483310299847056, + "loss": 0.4492, + "step": 9983 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001648257613677915, + "loss": 0.5512, + "step": 9984 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001648184191343883, + "loss": 0.6083, + "step": 9985 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001648110762983293, + "loss": 0.5825, + "step": 9986 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016480373285968276, + "loss": 0.5315, + "step": 9987 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016479638881851695, + "loss": 0.5517, + "step": 9988 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016478904417490008, + "loss": 0.5329, + "step": 9989 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016478169892890053, + "loss": 0.6452, + "step": 9990 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016477435308058657, + "loss": 0.5716, + "step": 9991 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016476700663002651, + "loss": 0.5494, + "step": 9992 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001647596595772886, + "loss": 0.6521, + "step": 9993 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016475231192244122, + "loss": 0.4755, + "step": 9994 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001647449636655526, + "loss": 0.5821, + "step": 9995 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016473761480669116, + "loss": 0.5313, + "step": 9996 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016473026534592514, + "loss": 0.6307, + "step": 9997 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016472291528332294, + "loss": 0.467, + "step": 9998 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016471556461895287, + "loss": 0.5784, + "step": 9999 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016470821335288328, + "loss": 0.4428, + "step": 10000 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001647008614851825, + "loss": 0.5374, + "step": 10001 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001646935090159189, + "loss": 0.5233, + "step": 10002 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016468615594516083, + "loss": 0.666, + "step": 10003 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016467880227297667, + "loss": 0.5669, + "step": 10004 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016467144799943477, + "loss": 0.5301, + "step": 10005 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016466409312460356, + "loss": 0.5552, + "step": 10006 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016465673764855135, + "loss": 0.5991, + "step": 10007 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016464938157134657, + "loss": 0.4928, + "step": 10008 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001646420248930576, + "loss": 0.5479, + "step": 10009 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016463466761375286, + "loss": 0.6852, + "step": 10010 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016462730973350074, + "loss": 0.5927, + "step": 10011 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016461995125236967, + "loss": 0.518, + "step": 10012 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016461259217042802, + "loss": 0.6184, + "step": 10013 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016460523248774424, + "loss": 0.5748, + "step": 10014 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016459787220438673, + "loss": 0.5996, + "step": 10015 + }, + { + "epoch": 0.3, + "learning_rate": 0.000164590511320424, + "loss": 0.6424, + "step": 10016 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016458314983592445, + "loss": 0.5201, + "step": 10017 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016457578775095648, + "loss": 0.5902, + "step": 10018 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016456842506558858, + "loss": 0.5909, + "step": 10019 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001645610617798892, + "loss": 0.5461, + "step": 10020 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001645536978939268, + "loss": 0.5137, + "step": 10021 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016454633340776985, + "loss": 0.5634, + "step": 10022 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001645389683214868, + "loss": 0.5043, + "step": 10023 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016453160263514615, + "loss": 0.4907, + "step": 10024 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001645242363488164, + "loss": 0.6849, + "step": 10025 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016451686946256599, + "loss": 0.6394, + "step": 10026 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016450950197646345, + "loss": 0.6088, + "step": 10027 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016450213389057727, + "loss": 0.4846, + "step": 10028 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016449476520497596, + "loss": 0.5241, + "step": 10029 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016448739591972805, + "loss": 0.6183, + "step": 10030 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016448002603490201, + "loss": 0.5349, + "step": 10031 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001644726555505664, + "loss": 0.5743, + "step": 10032 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016446528446678973, + "loss": 0.542, + "step": 10033 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016445791278364054, + "loss": 0.5359, + "step": 10034 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016445054050118736, + "loss": 0.5515, + "step": 10035 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016444316761949872, + "loss": 0.4919, + "step": 10036 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016443579413864324, + "loss": 0.5759, + "step": 10037 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001644284200586894, + "loss": 0.5417, + "step": 10038 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001644210453797058, + "loss": 0.5879, + "step": 10039 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016441367010176097, + "loss": 0.5109, + "step": 10040 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016440629422492353, + "loss": 0.5677, + "step": 10041 + }, + { + "epoch": 0.3, + "learning_rate": 0.000164398917749262, + "loss": 0.5458, + "step": 10042 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016439154067484507, + "loss": 0.545, + "step": 10043 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016438416300174123, + "loss": 0.5775, + "step": 10044 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001643767847300191, + "loss": 0.5438, + "step": 10045 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016436940585974724, + "loss": 0.5138, + "step": 10046 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016436202639099435, + "loss": 0.5751, + "step": 10047 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016435464632382898, + "loss": 0.5364, + "step": 10048 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016434726565831977, + "loss": 0.5745, + "step": 10049 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001643398843945353, + "loss": 0.6082, + "step": 10050 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016433250253254428, + "loss": 0.4264, + "step": 10051 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016432512007241522, + "loss": 0.6108, + "step": 10052 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001643177370142169, + "loss": 0.4893, + "step": 10053 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016431035335801784, + "loss": 0.4853, + "step": 10054 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016430296910388678, + "loss": 0.6085, + "step": 10055 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016429558425189232, + "loss": 0.5611, + "step": 10056 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016428819880210318, + "loss": 0.5253, + "step": 10057 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016428081275458796, + "loss": 0.4856, + "step": 10058 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016427342610941538, + "loss": 0.7048, + "step": 10059 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016426603886665405, + "loss": 0.5782, + "step": 10060 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016425865102637276, + "loss": 0.6287, + "step": 10061 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016425126258864015, + "loss": 0.598, + "step": 10062 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016424387355352485, + "loss": 0.6034, + "step": 10063 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016423648392109566, + "loss": 0.6441, + "step": 10064 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016422909369142124, + "loss": 0.5508, + "step": 10065 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016422170286457032, + "loss": 0.573, + "step": 10066 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016421431144061156, + "loss": 0.6115, + "step": 10067 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016420691941961376, + "loss": 0.5244, + "step": 10068 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016419952680164562, + "loss": 0.5762, + "step": 10069 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016419213358677585, + "loss": 0.642, + "step": 10070 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016418473977507323, + "loss": 0.5856, + "step": 10071 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016417734536660646, + "loss": 0.6397, + "step": 10072 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001641699503614443, + "loss": 0.4875, + "step": 10073 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016416255475965554, + "loss": 0.4584, + "step": 10074 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001641551585613089, + "loss": 0.5523, + "step": 10075 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016414776176647318, + "loss": 0.5302, + "step": 10076 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001641403643752171, + "loss": 0.5845, + "step": 10077 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016413296638760952, + "loss": 0.6398, + "step": 10078 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016412556780371918, + "loss": 0.6194, + "step": 10079 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016411816862361487, + "loss": 0.5526, + "step": 10080 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016411076884736536, + "loss": 0.5921, + "step": 10081 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016410336847503949, + "loss": 0.5789, + "step": 10082 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016409596750670605, + "loss": 0.5306, + "step": 10083 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016408856594243382, + "loss": 0.6297, + "step": 10084 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016408116378229171, + "loss": 0.6234, + "step": 10085 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001640737610263484, + "loss": 0.613, + "step": 10086 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016406635767467285, + "loss": 0.6188, + "step": 10087 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001640589537273338, + "loss": 0.5176, + "step": 10088 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016405154918440018, + "loss": 0.549, + "step": 10089 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016404414404594075, + "loss": 0.5352, + "step": 10090 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016403673831202442, + "loss": 0.6149, + "step": 10091 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016402933198271999, + "loss": 0.5905, + "step": 10092 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016402192505809637, + "loss": 0.5934, + "step": 10093 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016401451753822238, + "loss": 0.5742, + "step": 10094 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016400710942316696, + "loss": 0.6455, + "step": 10095 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016399970071299892, + "loss": 0.583, + "step": 10096 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001639922914077872, + "loss": 0.646, + "step": 10097 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016398488150760063, + "loss": 0.5706, + "step": 10098 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016397747101250815, + "loss": 0.5508, + "step": 10099 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016397005992257864, + "loss": 0.6655, + "step": 10100 + }, + { + "epoch": 0.3, + "learning_rate": 0.000163962648237881, + "loss": 0.5694, + "step": 10101 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016395523595848414, + "loss": 0.6546, + "step": 10102 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016394782308445702, + "loss": 0.6444, + "step": 10103 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001639404096158685, + "loss": 0.579, + "step": 10104 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016393299555278755, + "loss": 0.5017, + "step": 10105 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016392558089528306, + "loss": 0.5224, + "step": 10106 + }, + { + "epoch": 0.3, + "learning_rate": 0.000163918165643424, + "loss": 0.4784, + "step": 10107 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016391074979727933, + "loss": 0.5868, + "step": 10108 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016390333335691802, + "loss": 0.5391, + "step": 10109 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016389591632240893, + "loss": 0.6018, + "step": 10110 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001638884986938211, + "loss": 0.7051, + "step": 10111 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016388108047122346, + "loss": 0.5466, + "step": 10112 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016387366165468502, + "loss": 0.4453, + "step": 10113 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001638662422442747, + "loss": 0.5283, + "step": 10114 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016385882224006154, + "loss": 0.5209, + "step": 10115 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016385140164211446, + "loss": 0.5713, + "step": 10116 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016384398045050252, + "loss": 0.746, + "step": 10117 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016383655866529474, + "loss": 0.5902, + "step": 10118 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016382913628656, + "loss": 0.453, + "step": 10119 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016382171331436746, + "loss": 0.589, + "step": 10120 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016381428974878603, + "loss": 0.4963, + "step": 10121 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016380686558988477, + "loss": 0.4842, + "step": 10122 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001637994408377327, + "loss": 0.5274, + "step": 10123 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016379201549239887, + "loss": 0.4323, + "step": 10124 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001637845895539523, + "loss": 0.6069, + "step": 10125 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016377716302246205, + "loss": 0.5008, + "step": 10126 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016376973589799713, + "loss": 0.4496, + "step": 10127 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016376230818062668, + "loss": 0.5521, + "step": 10128 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016375487987041964, + "loss": 0.6051, + "step": 10129 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016374745096744517, + "loss": 0.4776, + "step": 10130 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016374002147177228, + "loss": 0.5101, + "step": 10131 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001637325913834701, + "loss": 0.5615, + "step": 10132 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016372516070260767, + "loss": 0.5609, + "step": 10133 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001637177294292541, + "loss": 0.4723, + "step": 10134 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016371029756347846, + "loss": 0.4864, + "step": 10135 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001637028651053499, + "loss": 0.4973, + "step": 10136 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016369543205493748, + "loss": 0.5279, + "step": 10137 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001636879984123103, + "loss": 0.4687, + "step": 10138 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001636805641775375, + "loss": 0.6157, + "step": 10139 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001636731293506882, + "loss": 0.595, + "step": 10140 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016366569393183155, + "loss": 0.5873, + "step": 10141 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016365825792103662, + "loss": 0.5569, + "step": 10142 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001636508213183726, + "loss": 0.6898, + "step": 10143 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001636433841239086, + "loss": 0.7103, + "step": 10144 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016363594633771376, + "loss": 0.5703, + "step": 10145 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016362850795985728, + "loss": 0.5238, + "step": 10146 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016362106899040827, + "loss": 0.6586, + "step": 10147 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016361362942943593, + "loss": 0.4725, + "step": 10148 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016360618927700944, + "loss": 0.6694, + "step": 10149 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016359874853319792, + "loss": 0.5503, + "step": 10150 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001635913071980706, + "loss": 0.6169, + "step": 10151 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016358386527169665, + "loss": 0.4414, + "step": 10152 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016357642275414527, + "loss": 0.6154, + "step": 10153 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001635689796454856, + "loss": 0.4991, + "step": 10154 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016356153594578697, + "loss": 0.5328, + "step": 10155 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016355409165511844, + "loss": 0.5998, + "step": 10156 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016354664677354936, + "loss": 0.6116, + "step": 10157 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016353920130114883, + "loss": 0.4445, + "step": 10158 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016353175523798618, + "loss": 0.5591, + "step": 10159 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016352430858413057, + "loss": 0.526, + "step": 10160 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016351686133965124, + "loss": 0.5269, + "step": 10161 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016350941350461748, + "loss": 0.5153, + "step": 10162 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016350196507909848, + "loss": 0.4978, + "step": 10163 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016349451606316353, + "loss": 0.4623, + "step": 10164 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016348706645688188, + "loss": 0.4619, + "step": 10165 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016347961626032279, + "loss": 0.5951, + "step": 10166 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001634721654735555, + "loss": 0.5813, + "step": 10167 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016346471409664934, + "loss": 0.5725, + "step": 10168 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016345726212967356, + "loss": 0.5556, + "step": 10169 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016344980957269745, + "loss": 0.489, + "step": 10170 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001634423564257903, + "loss": 0.6053, + "step": 10171 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016343490268902139, + "loss": 0.5819, + "step": 10172 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016342744836246004, + "loss": 0.5679, + "step": 10173 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016341999344617557, + "loss": 0.5491, + "step": 10174 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016341253794023726, + "loss": 0.5047, + "step": 10175 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016340508184471445, + "loss": 0.613, + "step": 10176 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001633976251596765, + "loss": 0.5952, + "step": 10177 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016339016788519267, + "loss": 0.5598, + "step": 10178 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016338271002133229, + "loss": 0.5837, + "step": 10179 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016337525156816476, + "loss": 0.6161, + "step": 10180 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001633677925257594, + "loss": 0.5243, + "step": 10181 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001633603328941856, + "loss": 0.5206, + "step": 10182 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001633528726735126, + "loss": 0.5256, + "step": 10183 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001633454118638099, + "loss": 0.5602, + "step": 10184 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001633379504651468, + "loss": 0.5325, + "step": 10185 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016333048847759268, + "loss": 0.4764, + "step": 10186 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016332302590121688, + "loss": 0.4734, + "step": 10187 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016331556273608885, + "loss": 0.5588, + "step": 10188 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016330809898227794, + "loss": 0.5486, + "step": 10189 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016330063463985362, + "loss": 0.5564, + "step": 10190 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016329316970888514, + "loss": 0.4661, + "step": 10191 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016328570418944205, + "loss": 0.5766, + "step": 10192 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016327823808159368, + "loss": 0.5424, + "step": 10193 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001632707713854095, + "loss": 0.6204, + "step": 10194 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001632633041009589, + "loss": 0.6356, + "step": 10195 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001632558362283113, + "loss": 0.4723, + "step": 10196 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016324836776753613, + "loss": 0.5363, + "step": 10197 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016324089871870288, + "loss": 0.4642, + "step": 10198 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016323342908188093, + "loss": 0.5653, + "step": 10199 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016322595885713976, + "loss": 0.549, + "step": 10200 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016321848804454883, + "loss": 0.5969, + "step": 10201 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001632110166441776, + "loss": 0.6177, + "step": 10202 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016320354465609554, + "loss": 0.5344, + "step": 10203 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001631960720803721, + "loss": 0.5163, + "step": 10204 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016318859891707675, + "loss": 0.4948, + "step": 10205 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016318112516627902, + "loss": 0.4858, + "step": 10206 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016317365082804837, + "loss": 0.4696, + "step": 10207 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016316617590245427, + "loss": 0.5552, + "step": 10208 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016315870038956628, + "loss": 0.6065, + "step": 10209 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016315122428945381, + "loss": 0.5922, + "step": 10210 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016314374760218646, + "loss": 0.5623, + "step": 10211 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016313627032783372, + "loss": 0.5975, + "step": 10212 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016312879246646508, + "loss": 0.6606, + "step": 10213 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016312131401815012, + "loss": 0.6058, + "step": 10214 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016311383498295834, + "loss": 0.607, + "step": 10215 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016310635536095924, + "loss": 0.4352, + "step": 10216 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016309887515222242, + "loss": 0.6576, + "step": 10217 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016309139435681738, + "loss": 0.6069, + "step": 10218 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016308391297481376, + "loss": 0.4636, + "step": 10219 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016307643100628102, + "loss": 0.4872, + "step": 10220 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016306894845128877, + "loss": 0.4869, + "step": 10221 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016306146530990656, + "loss": 0.6778, + "step": 10222 + }, + { + "epoch": 0.3, + "learning_rate": 0.000163053981582204, + "loss": 0.5637, + "step": 10223 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016304649726825066, + "loss": 0.5633, + "step": 10224 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001630390123681161, + "loss": 0.5176, + "step": 10225 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001630315268818699, + "loss": 0.4798, + "step": 10226 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016302404080958174, + "loss": 0.6242, + "step": 10227 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016301655415132112, + "loss": 0.5748, + "step": 10228 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016300906690715775, + "loss": 0.5819, + "step": 10229 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001630015790771611, + "loss": 0.561, + "step": 10230 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016299409066140096, + "loss": 0.5468, + "step": 10231 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016298660165994683, + "loss": 0.5554, + "step": 10232 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001629791120728684, + "loss": 0.5332, + "step": 10233 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016297162190023528, + "loss": 0.4447, + "step": 10234 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016296413114211714, + "loss": 0.4924, + "step": 10235 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016295663979858357, + "loss": 0.6281, + "step": 10236 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001629491478697043, + "loss": 0.5652, + "step": 10237 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016294165535554888, + "loss": 0.4658, + "step": 10238 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016293416225618708, + "loss": 0.6288, + "step": 10239 + }, + { + "epoch": 0.3, + "learning_rate": 0.0001629266685716885, + "loss": 0.6097, + "step": 10240 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016291917430212287, + "loss": 0.6007, + "step": 10241 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016291167944755978, + "loss": 0.6045, + "step": 10242 + }, + { + "epoch": 0.3, + "learning_rate": 0.000162904184008069, + "loss": 0.5955, + "step": 10243 + }, + { + "epoch": 0.3, + "learning_rate": 0.00016289668798372017, + "loss": 0.6196, + "step": 10244 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016288919137458306, + "loss": 0.5439, + "step": 10245 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016288169418072723, + "loss": 0.4969, + "step": 10246 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016287419640222256, + "loss": 0.5945, + "step": 10247 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001628666980391386, + "loss": 0.5757, + "step": 10248 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016285919909154518, + "loss": 0.5525, + "step": 10249 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016285169955951195, + "loss": 0.6279, + "step": 10250 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016284419944310866, + "loss": 0.5716, + "step": 10251 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016283669874240508, + "loss": 0.6553, + "step": 10252 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001628291974574709, + "loss": 0.5305, + "step": 10253 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001628216955883759, + "loss": 0.5763, + "step": 10254 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016281419313518978, + "loss": 0.5289, + "step": 10255 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016280669009798238, + "loss": 0.5604, + "step": 10256 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016279918647682336, + "loss": 0.6009, + "step": 10257 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016279168227178258, + "loss": 0.548, + "step": 10258 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016278417748292974, + "loss": 0.5304, + "step": 10259 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016277667211033462, + "loss": 0.6159, + "step": 10260 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016276916615406707, + "loss": 0.5706, + "step": 10261 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001627616596141968, + "loss": 0.6437, + "step": 10262 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016275415249079363, + "loss": 0.5709, + "step": 10263 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016274664478392736, + "loss": 0.4755, + "step": 10264 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001627391364936678, + "loss": 0.6527, + "step": 10265 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016273162762008475, + "loss": 0.637, + "step": 10266 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016272411816324804, + "loss": 0.6182, + "step": 10267 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016271660812322745, + "loss": 0.6169, + "step": 10268 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016270909750009288, + "loss": 0.5507, + "step": 10269 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001627015862939141, + "loss": 0.5577, + "step": 10270 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016269407450476093, + "loss": 0.6508, + "step": 10271 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001626865621327033, + "loss": 0.7077, + "step": 10272 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016267904917781096, + "loss": 0.532, + "step": 10273 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016267153564015382, + "loss": 0.4103, + "step": 10274 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001626640215198017, + "loss": 0.546, + "step": 10275 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016265650681682447, + "loss": 0.582, + "step": 10276 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016264899153129204, + "loss": 0.5963, + "step": 10277 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016264147566327425, + "loss": 0.4786, + "step": 10278 + }, + { + "epoch": 0.31, + "learning_rate": 0.000162633959212841, + "loss": 0.5723, + "step": 10279 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016262644218006212, + "loss": 0.5485, + "step": 10280 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001626189245650076, + "loss": 0.6248, + "step": 10281 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016261140636774724, + "loss": 0.5504, + "step": 10282 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016260388758835098, + "loss": 0.5707, + "step": 10283 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016259636822688872, + "loss": 0.6463, + "step": 10284 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001625888482834304, + "loss": 0.7231, + "step": 10285 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016258132775804588, + "loss": 0.584, + "step": 10286 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016257380665080513, + "loss": 0.6116, + "step": 10287 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016256628496177808, + "loss": 0.6115, + "step": 10288 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016255876269103463, + "loss": 0.5588, + "step": 10289 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016255123983864475, + "loss": 0.5627, + "step": 10290 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016254371640467836, + "loss": 0.5821, + "step": 10291 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016253619238920544, + "loss": 0.4852, + "step": 10292 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001625286677922959, + "loss": 0.547, + "step": 10293 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016252114261401977, + "loss": 0.6256, + "step": 10294 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016251361685444695, + "loss": 0.7437, + "step": 10295 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016250609051364742, + "loss": 0.523, + "step": 10296 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016249856359169124, + "loss": 0.5162, + "step": 10297 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016249103608864825, + "loss": 0.5523, + "step": 10298 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016248350800458853, + "loss": 0.6663, + "step": 10299 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016247597933958206, + "loss": 0.5926, + "step": 10300 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016246845009369886, + "loss": 0.5145, + "step": 10301 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016246092026700888, + "loss": 0.6809, + "step": 10302 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016245338985958218, + "loss": 0.5451, + "step": 10303 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016244585887148873, + "loss": 0.4869, + "step": 10304 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016243832730279857, + "loss": 0.422, + "step": 10305 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016243079515358173, + "loss": 0.648, + "step": 10306 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016242326242390826, + "loss": 0.585, + "step": 10307 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016241572911384816, + "loss": 0.5851, + "step": 10308 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001624081952234715, + "loss": 0.5241, + "step": 10309 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001624006607528483, + "loss": 0.5619, + "step": 10310 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016239312570204864, + "loss": 0.5434, + "step": 10311 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016238559007114253, + "loss": 0.5922, + "step": 10312 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016237805386020012, + "loss": 0.5361, + "step": 10313 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016237051706929141, + "loss": 0.5062, + "step": 10314 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016236297969848647, + "loss": 0.5357, + "step": 10315 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016235544174785543, + "loss": 0.5027, + "step": 10316 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016234790321746832, + "loss": 0.579, + "step": 10317 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016234036410739527, + "loss": 0.6067, + "step": 10318 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016233282441770633, + "loss": 0.5502, + "step": 10319 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016232528414847165, + "loss": 0.4791, + "step": 10320 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016231774329976135, + "loss": 0.4698, + "step": 10321 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016231020187164547, + "loss": 0.5294, + "step": 10322 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016230265986419418, + "loss": 0.6493, + "step": 10323 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001622951172774776, + "loss": 0.5069, + "step": 10324 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016228757411156582, + "loss": 0.6458, + "step": 10325 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016228003036652903, + "loss": 0.5911, + "step": 10326 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016227248604243734, + "loss": 0.6019, + "step": 10327 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016226494113936088, + "loss": 0.6399, + "step": 10328 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001622573956573698, + "loss": 0.579, + "step": 10329 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016224984959653432, + "loss": 0.5593, + "step": 10330 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001622423029569245, + "loss": 0.5609, + "step": 10331 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016223475573861059, + "loss": 0.522, + "step": 10332 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016222720794166267, + "loss": 0.5943, + "step": 10333 + }, + { + "epoch": 0.31, + "learning_rate": 0.000162219659566151, + "loss": 0.5595, + "step": 10334 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016221211061214576, + "loss": 0.5998, + "step": 10335 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001622045610797171, + "loss": 0.5907, + "step": 10336 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001621970109689352, + "loss": 0.5682, + "step": 10337 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016218946027987032, + "loss": 0.436, + "step": 10338 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016218190901259259, + "loss": 0.5243, + "step": 10339 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016217435716717226, + "loss": 0.6385, + "step": 10340 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016216680474367956, + "loss": 0.465, + "step": 10341 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016215925174218465, + "loss": 0.5284, + "step": 10342 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001621516981627578, + "loss": 0.5107, + "step": 10343 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016214414400546928, + "loss": 0.5189, + "step": 10344 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016213658927038922, + "loss": 0.476, + "step": 10345 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016212903395758794, + "loss": 0.6024, + "step": 10346 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016212147806713567, + "loss": 0.5416, + "step": 10347 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016211392159910264, + "loss": 0.5345, + "step": 10348 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016210636455355913, + "loss": 0.4676, + "step": 10349 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016209880693057537, + "loss": 0.4577, + "step": 10350 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001620912487302217, + "loss": 0.5656, + "step": 10351 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016208368995256832, + "loss": 0.5708, + "step": 10352 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016207613059768554, + "loss": 0.5617, + "step": 10353 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016206857066564364, + "loss": 0.5195, + "step": 10354 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001620610101565129, + "loss": 0.6311, + "step": 10355 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016205344907036366, + "loss": 0.6299, + "step": 10356 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016204588740726617, + "loss": 0.5881, + "step": 10357 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001620383251672907, + "loss": 0.533, + "step": 10358 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016203076235050764, + "loss": 0.6134, + "step": 10359 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016202319895698727, + "loss": 0.6615, + "step": 10360 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016201563498679993, + "loss": 0.569, + "step": 10361 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001620080704400159, + "loss": 0.5561, + "step": 10362 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016200050531670561, + "loss": 0.6427, + "step": 10363 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016199293961693926, + "loss": 0.6222, + "step": 10364 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001619853733407873, + "loss": 0.6228, + "step": 10365 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016197780648832003, + "loss": 0.4826, + "step": 10366 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016197023905960785, + "loss": 0.5304, + "step": 10367 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016196267105472105, + "loss": 0.5467, + "step": 10368 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016195510247373008, + "loss": 0.5587, + "step": 10369 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016194753331670518, + "loss": 0.5928, + "step": 10370 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016193996358371688, + "loss": 0.5863, + "step": 10371 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016193239327483543, + "loss": 0.4885, + "step": 10372 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016192482239013129, + "loss": 0.5287, + "step": 10373 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016191725092967485, + "loss": 0.59, + "step": 10374 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016190967889353648, + "loss": 0.5325, + "step": 10375 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016190210628178658, + "loss": 0.6607, + "step": 10376 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001618945330944956, + "loss": 0.6203, + "step": 10377 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016188695933173387, + "loss": 0.5316, + "step": 10378 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016187938499357188, + "loss": 0.6559, + "step": 10379 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016187181008008007, + "loss": 0.5646, + "step": 10380 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016186423459132876, + "loss": 0.5805, + "step": 10381 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001618566585273885, + "loss": 0.574, + "step": 10382 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016184908188832966, + "loss": 0.5568, + "step": 10383 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001618415046742227, + "loss": 0.5982, + "step": 10384 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001618339268851381, + "loss": 0.606, + "step": 10385 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016182634852114627, + "loss": 0.6655, + "step": 10386 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016181876958231773, + "loss": 0.5137, + "step": 10387 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001618111900687229, + "loss": 0.5575, + "step": 10388 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016180360998043224, + "loss": 0.6123, + "step": 10389 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016179602931751626, + "loss": 0.5466, + "step": 10390 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016178844808004543, + "loss": 0.532, + "step": 10391 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001617808662680902, + "loss": 0.6898, + "step": 10392 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016177328388172115, + "loss": 0.4685, + "step": 10393 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001617657009210087, + "loss": 0.5618, + "step": 10394 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016175811738602337, + "loss": 0.6042, + "step": 10395 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016175053327683572, + "loss": 0.5247, + "step": 10396 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001617429485935162, + "loss": 0.5971, + "step": 10397 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016173536333613533, + "loss": 0.6235, + "step": 10398 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001617277775047637, + "loss": 0.6311, + "step": 10399 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016172019109947177, + "loss": 0.5956, + "step": 10400 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001617126041203301, + "loss": 0.6285, + "step": 10401 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016170501656740925, + "loss": 0.4953, + "step": 10402 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016169742844077974, + "loss": 0.57, + "step": 10403 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016168983974051211, + "loss": 0.4669, + "step": 10404 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016168225046667695, + "loss": 0.6109, + "step": 10405 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016167466061934484, + "loss": 0.5725, + "step": 10406 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016166707019858628, + "loss": 0.4588, + "step": 10407 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001616594792044719, + "loss": 0.6091, + "step": 10408 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016165188763707223, + "loss": 0.5107, + "step": 10409 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001616442954964579, + "loss": 0.5256, + "step": 10410 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001616367027826995, + "loss": 0.4307, + "step": 10411 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016162910949586755, + "loss": 0.5462, + "step": 10412 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016162151563603276, + "loss": 0.6507, + "step": 10413 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016161392120326565, + "loss": 0.5388, + "step": 10414 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016160632619763687, + "loss": 0.63, + "step": 10415 + }, + { + "epoch": 0.31, + "learning_rate": 0.000161598730619217, + "loss": 0.5097, + "step": 10416 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001615911344680767, + "loss": 0.5237, + "step": 10417 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016158353774428657, + "loss": 0.5893, + "step": 10418 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016157594044791725, + "loss": 0.4998, + "step": 10419 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001615683425790394, + "loss": 0.4698, + "step": 10420 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016156074413772363, + "loss": 0.4809, + "step": 10421 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001615531451240406, + "loss": 0.6231, + "step": 10422 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016154554553806093, + "loss": 0.5096, + "step": 10423 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016153794537985533, + "loss": 0.573, + "step": 10424 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016153034464949447, + "loss": 0.5432, + "step": 10425 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016152274334704894, + "loss": 0.5821, + "step": 10426 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016151514147258946, + "loss": 0.549, + "step": 10427 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016150753902618675, + "loss": 0.525, + "step": 10428 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016149993600791143, + "loss": 0.5061, + "step": 10429 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001614923324178342, + "loss": 0.5841, + "step": 10430 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001614847282560258, + "loss": 0.5267, + "step": 10431 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016147712352255691, + "loss": 0.5905, + "step": 10432 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016146951821749823, + "loss": 0.6357, + "step": 10433 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016146191234092045, + "loss": 0.6332, + "step": 10434 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001614543058928943, + "loss": 0.5794, + "step": 10435 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016144669887349053, + "loss": 0.598, + "step": 10436 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016143909128277982, + "loss": 0.4749, + "step": 10437 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016143148312083294, + "loss": 0.5407, + "step": 10438 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016142387438772062, + "loss": 0.4766, + "step": 10439 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001614162650835136, + "loss": 0.5713, + "step": 10440 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001614086552082826, + "loss": 0.5331, + "step": 10441 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016140104476209841, + "loss": 0.6553, + "step": 10442 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001613934337450318, + "loss": 0.6336, + "step": 10443 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001613858221571535, + "loss": 0.5456, + "step": 10444 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016137820999853427, + "loss": 0.5933, + "step": 10445 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016137059726924492, + "loss": 0.5671, + "step": 10446 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016136298396935623, + "loss": 0.553, + "step": 10447 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016135537009893895, + "loss": 0.5317, + "step": 10448 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001613477556580639, + "loss": 0.6222, + "step": 10449 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001613401406468019, + "loss": 0.5459, + "step": 10450 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016133252506522367, + "loss": 0.5068, + "step": 10451 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001613249089134001, + "loss": 0.56, + "step": 10452 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016131729219140195, + "loss": 0.6081, + "step": 10453 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016130967489930007, + "loss": 0.4735, + "step": 10454 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016130205703716524, + "loss": 0.5444, + "step": 10455 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016129443860506837, + "loss": 0.5299, + "step": 10456 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016128681960308016, + "loss": 0.5596, + "step": 10457 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001612792000312716, + "loss": 0.5121, + "step": 10458 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016127157988971345, + "loss": 0.5516, + "step": 10459 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016126395917847653, + "loss": 0.8985, + "step": 10460 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016125633789763178, + "loss": 0.5035, + "step": 10461 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016124871604725002, + "loss": 0.5404, + "step": 10462 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016124109362740204, + "loss": 0.6488, + "step": 10463 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016123347063815886, + "loss": 0.5607, + "step": 10464 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016122584707959123, + "loss": 0.5595, + "step": 10465 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016121822295177008, + "loss": 0.6533, + "step": 10466 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001612105982547663, + "loss": 0.5837, + "step": 10467 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016120297298865077, + "loss": 0.513, + "step": 10468 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001611953471534944, + "loss": 0.4324, + "step": 10469 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016118772074936807, + "loss": 0.531, + "step": 10470 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001611800937763427, + "loss": 0.5094, + "step": 10471 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001611724662344892, + "loss": 0.4517, + "step": 10472 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016116483812387848, + "loss": 0.4699, + "step": 10473 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001611572094445815, + "loss": 0.6812, + "step": 10474 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016114958019666914, + "loss": 0.5254, + "step": 10475 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001611419503802124, + "loss": 0.578, + "step": 10476 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016113431999528212, + "loss": 0.5709, + "step": 10477 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016112668904194935, + "loss": 0.5923, + "step": 10478 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016111905752028496, + "loss": 0.5992, + "step": 10479 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001611114254303599, + "loss": 0.5444, + "step": 10480 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001611037927722452, + "loss": 0.5114, + "step": 10481 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001610961595460118, + "loss": 0.7115, + "step": 10482 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016108852575173063, + "loss": 0.4661, + "step": 10483 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016108089138947267, + "loss": 0.7, + "step": 10484 + }, + { + "epoch": 0.31, + "learning_rate": 0.000161073256459309, + "loss": 0.5725, + "step": 10485 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016106562096131048, + "loss": 0.5442, + "step": 10486 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001610579848955482, + "loss": 0.566, + "step": 10487 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016105034826209303, + "loss": 0.6143, + "step": 10488 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001610427110610161, + "loss": 0.7416, + "step": 10489 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016103507329238837, + "loss": 0.5528, + "step": 10490 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001610274349562809, + "loss": 0.4282, + "step": 10491 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016101979605276463, + "loss": 0.5767, + "step": 10492 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001610121565819106, + "loss": 0.5518, + "step": 10493 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016100451654378986, + "loss": 0.7614, + "step": 10494 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016099687593847345, + "loss": 0.4824, + "step": 10495 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001609892347660324, + "loss": 0.4991, + "step": 10496 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016098159302653774, + "loss": 0.4662, + "step": 10497 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016097395072006053, + "loss": 0.5499, + "step": 10498 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016096630784667184, + "loss": 0.4971, + "step": 10499 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001609586644064427, + "loss": 0.5884, + "step": 10500 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016095102039944426, + "loss": 0.4052, + "step": 10501 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016094337582574743, + "loss": 0.6664, + "step": 10502 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016093573068542345, + "loss": 0.5273, + "step": 10503 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001609280849785433, + "loss": 0.5465, + "step": 10504 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001609204387051781, + "loss": 0.7049, + "step": 10505 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016091279186539894, + "loss": 0.5144, + "step": 10506 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016090514445927693, + "loss": 0.5988, + "step": 10507 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016089749648688317, + "loss": 0.484, + "step": 10508 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016088984794828876, + "loss": 0.5835, + "step": 10509 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001608821988435648, + "loss": 0.492, + "step": 10510 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016087454917278242, + "loss": 0.6032, + "step": 10511 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016086689893601275, + "loss": 0.6907, + "step": 10512 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016085924813332691, + "loss": 0.59, + "step": 10513 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016085159676479602, + "loss": 0.4856, + "step": 10514 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016084394483049125, + "loss": 0.6392, + "step": 10515 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016083629233048374, + "loss": 0.4684, + "step": 10516 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001608286392648446, + "loss": 0.578, + "step": 10517 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016082098563364506, + "loss": 0.5231, + "step": 10518 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016081333143695622, + "loss": 0.4649, + "step": 10519 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016080567667484924, + "loss": 0.6663, + "step": 10520 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016079802134739532, + "loss": 0.4369, + "step": 10521 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016079036545466567, + "loss": 0.4765, + "step": 10522 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001607827089967314, + "loss": 0.4544, + "step": 10523 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016077505197366373, + "loss": 0.5263, + "step": 10524 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016076739438553383, + "loss": 0.551, + "step": 10525 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016075973623241294, + "loss": 0.5713, + "step": 10526 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016075207751437224, + "loss": 0.5511, + "step": 10527 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001607444182314829, + "loss": 0.4434, + "step": 10528 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016073675838381622, + "loss": 0.606, + "step": 10529 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016072909797144335, + "loss": 0.65, + "step": 10530 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016072143699443552, + "loss": 0.5024, + "step": 10531 + }, + { + "epoch": 0.31, + "learning_rate": 0.000160713775452864, + "loss": 0.622, + "step": 10532 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016070611334679998, + "loss": 0.4707, + "step": 10533 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016069845067631468, + "loss": 0.5359, + "step": 10534 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001606907874414794, + "loss": 0.6097, + "step": 10535 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016068312364236537, + "loss": 0.494, + "step": 10536 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016067545927904385, + "loss": 0.563, + "step": 10537 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001606677943515861, + "loss": 0.4711, + "step": 10538 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016066012886006335, + "loss": 0.6492, + "step": 10539 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001606524628045469, + "loss": 0.456, + "step": 10540 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016064479618510804, + "loss": 0.6313, + "step": 10541 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016063712900181803, + "loss": 0.4611, + "step": 10542 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016062946125474818, + "loss": 0.5911, + "step": 10543 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016062179294396974, + "loss": 0.5489, + "step": 10544 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016061412406955403, + "loss": 0.5715, + "step": 10545 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016060645463157237, + "loss": 0.4694, + "step": 10546 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016059878463009603, + "loss": 0.592, + "step": 10547 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016059111406519636, + "loss": 0.5018, + "step": 10548 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016058344293694464, + "loss": 0.4854, + "step": 10549 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016057577124541222, + "loss": 0.5436, + "step": 10550 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016056809899067044, + "loss": 0.4721, + "step": 10551 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001605604261727906, + "loss": 0.6051, + "step": 10552 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016055275279184408, + "loss": 0.5425, + "step": 10553 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016054507884790214, + "loss": 0.6201, + "step": 10554 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016053740434103628, + "loss": 0.4847, + "step": 10555 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001605297292713177, + "loss": 0.5363, + "step": 10556 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001605220536388178, + "loss": 0.5736, + "step": 10557 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016051437744360803, + "loss": 0.5512, + "step": 10558 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016050670068575967, + "loss": 0.4799, + "step": 10559 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001604990233653441, + "loss": 0.4786, + "step": 10560 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016049134548243275, + "loss": 0.7402, + "step": 10561 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016048366703709697, + "loss": 0.5106, + "step": 10562 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016047598802940816, + "loss": 0.5825, + "step": 10563 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001604683084594377, + "loss": 0.595, + "step": 10564 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016046062832725702, + "loss": 0.5774, + "step": 10565 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016045294763293753, + "loss": 0.6171, + "step": 10566 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016044526637655062, + "loss": 0.6027, + "step": 10567 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001604375845581677, + "loss": 0.5384, + "step": 10568 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016042990217786021, + "loss": 0.5832, + "step": 10569 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001604222192356996, + "loss": 0.5166, + "step": 10570 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016041453573175727, + "loss": 0.6089, + "step": 10571 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016040685166610462, + "loss": 0.4931, + "step": 10572 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016039916703881318, + "loss": 0.5715, + "step": 10573 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016039148184995438, + "loss": 0.4891, + "step": 10574 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016038379609959963, + "loss": 0.623, + "step": 10575 + }, + { + "epoch": 0.31, + "learning_rate": 0.0001603761097878204, + "loss": 0.5691, + "step": 10576 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016036842291468817, + "loss": 0.5035, + "step": 10577 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016036073548027438, + "loss": 0.5361, + "step": 10578 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016035304748465058, + "loss": 0.4727, + "step": 10579 + }, + { + "epoch": 0.31, + "learning_rate": 0.00016034535892788817, + "loss": 0.7255, + "step": 10580 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016033766981005866, + "loss": 0.5648, + "step": 10581 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016032998013123354, + "loss": 0.4899, + "step": 10582 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001603222898914843, + "loss": 0.5561, + "step": 10583 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016031459909088248, + "loss": 0.5265, + "step": 10584 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016030690772949956, + "loss": 0.5904, + "step": 10585 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016029921580740702, + "loss": 0.5153, + "step": 10586 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016029152332467644, + "loss": 0.573, + "step": 10587 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001602838302813793, + "loss": 0.5477, + "step": 10588 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001602761366775871, + "loss": 0.6003, + "step": 10589 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016026844251337144, + "loss": 0.631, + "step": 10590 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001602607477888038, + "loss": 0.5944, + "step": 10591 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016025305250395576, + "loss": 0.6841, + "step": 10592 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001602453566588988, + "loss": 0.5535, + "step": 10593 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016023766025370458, + "loss": 0.6916, + "step": 10594 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001602299632884446, + "loss": 0.5692, + "step": 10595 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001602222657631904, + "loss": 0.5867, + "step": 10596 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001602145676780136, + "loss": 0.5653, + "step": 10597 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016020686903298574, + "loss": 0.4429, + "step": 10598 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016019916982817843, + "loss": 0.5664, + "step": 10599 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016019147006366322, + "loss": 0.5821, + "step": 10600 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016018376973951168, + "loss": 0.5662, + "step": 10601 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016017606885579546, + "loss": 0.5407, + "step": 10602 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016016836741258614, + "loss": 0.5415, + "step": 10603 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001601606654099553, + "loss": 0.4518, + "step": 10604 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016015296284797458, + "loss": 0.6232, + "step": 10605 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001601452597267156, + "loss": 0.438, + "step": 10606 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016013755604624995, + "loss": 0.5286, + "step": 10607 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001601298518066493, + "loss": 0.6127, + "step": 10608 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016012214700798522, + "loss": 0.5249, + "step": 10609 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001601144416503294, + "loss": 0.5438, + "step": 10610 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016010673573375346, + "loss": 0.6172, + "step": 10611 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016009902925832905, + "loss": 0.5377, + "step": 10612 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001600913222241278, + "loss": 0.5107, + "step": 10613 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016008361463122137, + "loss": 0.5192, + "step": 10614 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016007590647968147, + "loss": 0.5219, + "step": 10615 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016006819776957977, + "loss": 0.6413, + "step": 10616 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016006048850098785, + "loss": 0.5541, + "step": 10617 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016005277867397742, + "loss": 0.6238, + "step": 10618 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016004506828862027, + "loss": 0.5251, + "step": 10619 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016003735734498794, + "loss": 0.4855, + "step": 10620 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016002964584315224, + "loss": 0.5276, + "step": 10621 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001600219337831848, + "loss": 0.554, + "step": 10622 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016001422116515733, + "loss": 0.4692, + "step": 10623 + }, + { + "epoch": 0.32, + "learning_rate": 0.00016000650798914155, + "loss": 0.5226, + "step": 10624 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015999879425520922, + "loss": 0.6375, + "step": 10625 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015999107996343196, + "loss": 0.5294, + "step": 10626 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015998336511388156, + "loss": 0.6149, + "step": 10627 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015997564970662972, + "loss": 0.6447, + "step": 10628 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015996793374174826, + "loss": 0.5163, + "step": 10629 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001599602172193088, + "loss": 0.5718, + "step": 10630 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015995250013938316, + "loss": 0.502, + "step": 10631 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015994478250204303, + "loss": 0.4193, + "step": 10632 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015993706430736024, + "loss": 0.5832, + "step": 10633 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001599293455554065, + "loss": 0.6115, + "step": 10634 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001599216262462536, + "loss": 0.5729, + "step": 10635 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001599139063799733, + "loss": 0.566, + "step": 10636 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001599061859566374, + "loss": 0.629, + "step": 10637 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015989846497631763, + "loss": 0.6283, + "step": 10638 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001598907434390858, + "loss": 0.5231, + "step": 10639 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015988302134501372, + "loss": 0.5467, + "step": 10640 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015987529869417318, + "loss": 0.5475, + "step": 10641 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015986757548663597, + "loss": 0.5635, + "step": 10642 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001598598517224739, + "loss": 0.4469, + "step": 10643 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015985212740175878, + "loss": 0.6001, + "step": 10644 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015984440252456245, + "loss": 0.4715, + "step": 10645 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015983667709095674, + "loss": 0.6061, + "step": 10646 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015982895110101342, + "loss": 0.5644, + "step": 10647 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015982122455480434, + "loss": 0.4898, + "step": 10648 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001598134974524014, + "loss": 0.5859, + "step": 10649 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015980576979387635, + "loss": 0.4594, + "step": 10650 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015979804157930115, + "loss": 0.5566, + "step": 10651 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015979031280874753, + "loss": 0.5758, + "step": 10652 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015978258348228747, + "loss": 0.4459, + "step": 10653 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015977485359999272, + "loss": 0.5732, + "step": 10654 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015976712316193524, + "loss": 0.5225, + "step": 10655 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015975939216818688, + "loss": 0.5087, + "step": 10656 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001597516606188195, + "loss": 0.6653, + "step": 10657 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015974392851390498, + "loss": 0.5675, + "step": 10658 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001597361958535152, + "loss": 0.569, + "step": 10659 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015972846263772214, + "loss": 0.5085, + "step": 10660 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001597207288665976, + "loss": 0.6319, + "step": 10661 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015971299454021356, + "loss": 0.6212, + "step": 10662 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015970525965864187, + "loss": 0.5323, + "step": 10663 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001596975242219545, + "loss": 0.6151, + "step": 10664 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001596897882302233, + "loss": 0.6177, + "step": 10665 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015968205168352025, + "loss": 0.6119, + "step": 10666 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001596743145819173, + "loss": 0.5644, + "step": 10667 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015966657692548632, + "loss": 0.5877, + "step": 10668 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001596588387142993, + "loss": 0.5272, + "step": 10669 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015965109994842816, + "loss": 0.4916, + "step": 10670 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015964336062794488, + "loss": 0.5968, + "step": 10671 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015963562075292144, + "loss": 0.6347, + "step": 10672 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015962788032342973, + "loss": 0.4987, + "step": 10673 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015962013933954174, + "loss": 0.5494, + "step": 10674 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015961239780132948, + "loss": 0.5509, + "step": 10675 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001596046557088649, + "loss": 0.6047, + "step": 10676 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015959691306222, + "loss": 0.59, + "step": 10677 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015958916986146674, + "loss": 0.4381, + "step": 10678 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015958142610667715, + "loss": 0.5445, + "step": 10679 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015957368179792317, + "loss": 0.5582, + "step": 10680 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015956593693527686, + "loss": 0.5433, + "step": 10681 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001595581915188102, + "loss": 0.4401, + "step": 10682 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015955044554859525, + "loss": 0.4919, + "step": 10683 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015954269902470394, + "loss": 0.6183, + "step": 10684 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015953495194720837, + "loss": 0.5655, + "step": 10685 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015952720431618053, + "loss": 0.5821, + "step": 10686 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001595194561316925, + "loss": 0.5176, + "step": 10687 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015951170739381628, + "loss": 0.4751, + "step": 10688 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001595039581026239, + "loss": 0.451, + "step": 10689 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015949620825818746, + "loss": 0.528, + "step": 10690 + }, + { + "epoch": 0.32, + "learning_rate": 0.000159488457860579, + "loss": 0.5905, + "step": 10691 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015948070690987055, + "loss": 0.5495, + "step": 10692 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015947295540613416, + "loss": 0.5785, + "step": 10693 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015946520334944196, + "loss": 0.5322, + "step": 10694 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015945745073986603, + "loss": 0.5376, + "step": 10695 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015944969757747835, + "loss": 0.5189, + "step": 10696 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015944194386235117, + "loss": 0.6179, + "step": 10697 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001594341895945564, + "loss": 0.5963, + "step": 10698 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001594264347741663, + "loss": 0.4923, + "step": 10699 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015941867940125282, + "loss": 0.5559, + "step": 10700 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001594109234758882, + "loss": 0.4875, + "step": 10701 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015940316699814444, + "loss": 0.5456, + "step": 10702 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015939540996809376, + "loss": 0.5721, + "step": 10703 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015938765238580823, + "loss": 0.52, + "step": 10704 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015937989425135994, + "loss": 0.5881, + "step": 10705 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015937213556482108, + "loss": 0.5467, + "step": 10706 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015936437632626374, + "loss": 0.4927, + "step": 10707 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015935661653576015, + "loss": 0.4768, + "step": 10708 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015934885619338233, + "loss": 0.5235, + "step": 10709 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015934109529920255, + "loss": 0.5765, + "step": 10710 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015933333385329288, + "loss": 0.4871, + "step": 10711 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015932557185572554, + "loss": 0.4916, + "step": 10712 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015931780930657268, + "loss": 0.5727, + "step": 10713 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015931004620590647, + "loss": 0.5667, + "step": 10714 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015930228255379907, + "loss": 0.634, + "step": 10715 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001592945183503227, + "loss": 0.654, + "step": 10716 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001592867535955495, + "loss": 0.5537, + "step": 10717 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015927898828955173, + "loss": 0.5261, + "step": 10718 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015927122243240155, + "loss": 0.5399, + "step": 10719 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015926345602417115, + "loss": 0.6097, + "step": 10720 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001592556890649328, + "loss": 0.5488, + "step": 10721 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015924792155475864, + "loss": 0.4329, + "step": 10722 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001592401534937209, + "loss": 0.6091, + "step": 10723 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015923238488189183, + "loss": 0.5794, + "step": 10724 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015922461571934367, + "loss": 0.5572, + "step": 10725 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015921684600614863, + "loss": 0.5073, + "step": 10726 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015920907574237895, + "loss": 0.6021, + "step": 10727 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001592013049281069, + "loss": 0.603, + "step": 10728 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001591935335634047, + "loss": 0.5399, + "step": 10729 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015918576164834463, + "loss": 0.5872, + "step": 10730 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015917798918299892, + "loss": 0.47, + "step": 10731 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015917021616743986, + "loss": 0.5407, + "step": 10732 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001591624426017397, + "loss": 0.5206, + "step": 10733 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015915466848597074, + "loss": 0.555, + "step": 10734 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015914689382020525, + "loss": 0.5869, + "step": 10735 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015913911860451551, + "loss": 0.5545, + "step": 10736 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015913134283897384, + "loss": 0.5712, + "step": 10737 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015912356652365248, + "loss": 0.5419, + "step": 10738 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015911578965862377, + "loss": 0.504, + "step": 10739 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015910801224396, + "loss": 0.4881, + "step": 10740 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015910023427973354, + "loss": 0.6009, + "step": 10741 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015909245576601662, + "loss": 0.5477, + "step": 10742 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001590846767028816, + "loss": 0.543, + "step": 10743 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015907689709040076, + "loss": 0.5947, + "step": 10744 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015906911692864655, + "loss": 0.5023, + "step": 10745 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001590613362176912, + "loss": 0.6842, + "step": 10746 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001590535549576071, + "loss": 0.6734, + "step": 10747 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015904577314846655, + "loss": 0.5654, + "step": 10748 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015903799079034197, + "loss": 0.5507, + "step": 10749 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015903020788330565, + "loss": 0.5425, + "step": 10750 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015902242442743003, + "loss": 0.4898, + "step": 10751 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015901464042278738, + "loss": 0.6518, + "step": 10752 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015900685586945016, + "loss": 0.6006, + "step": 10753 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001589990707674907, + "loss": 0.5926, + "step": 10754 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015899128511698139, + "loss": 0.5411, + "step": 10755 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001589834989179946, + "loss": 0.5136, + "step": 10756 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015897571217060282, + "loss": 0.4711, + "step": 10757 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001589679248748783, + "loss": 0.5895, + "step": 10758 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015896013703089356, + "loss": 0.5276, + "step": 10759 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015895234863872094, + "loss": 0.5278, + "step": 10760 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001589445596984329, + "loss": 0.6248, + "step": 10761 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015893677021010182, + "loss": 0.5641, + "step": 10762 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015892898017380014, + "loss": 0.5453, + "step": 10763 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001589211895896003, + "loss": 0.5742, + "step": 10764 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015891339845757472, + "loss": 0.5558, + "step": 10765 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015890560677779583, + "loss": 0.5227, + "step": 10766 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001588978145503361, + "loss": 0.6001, + "step": 10767 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015889002177526794, + "loss": 0.5624, + "step": 10768 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015888222845266388, + "loss": 0.4829, + "step": 10769 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015887443458259627, + "loss": 0.5936, + "step": 10770 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015886664016513766, + "loss": 0.6822, + "step": 10771 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001588588452003605, + "loss": 0.6001, + "step": 10772 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015885104968833724, + "loss": 0.5642, + "step": 10773 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015884325362914038, + "loss": 0.6333, + "step": 10774 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001588354570228424, + "loss": 0.5792, + "step": 10775 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001588276598695158, + "loss": 0.6732, + "step": 10776 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015881986216923305, + "loss": 0.5705, + "step": 10777 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015881206392206668, + "loss": 0.5886, + "step": 10778 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015880426512808918, + "loss": 0.5904, + "step": 10779 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015879646578737303, + "loss": 0.6516, + "step": 10780 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015878866589999079, + "loss": 0.6339, + "step": 10781 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015878086546601498, + "loss": 0.5246, + "step": 10782 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001587730644855181, + "loss": 0.5346, + "step": 10783 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001587652629585727, + "loss": 0.5015, + "step": 10784 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001587574608852513, + "loss": 0.6002, + "step": 10785 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015874965826562644, + "loss": 0.5441, + "step": 10786 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015874185509977067, + "loss": 0.6265, + "step": 10787 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015873405138775657, + "loss": 0.5863, + "step": 10788 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015872624712965662, + "loss": 0.5399, + "step": 10789 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015871844232554348, + "loss": 0.5736, + "step": 10790 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015871063697548964, + "loss": 0.5011, + "step": 10791 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015870283107956773, + "loss": 0.6855, + "step": 10792 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015869502463785023, + "loss": 0.5556, + "step": 10793 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015868721765040987, + "loss": 0.5432, + "step": 10794 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015867941011731907, + "loss": 0.6008, + "step": 10795 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015867160203865055, + "loss": 0.5713, + "step": 10796 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015866379341447683, + "loss": 0.6373, + "step": 10797 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001586559842448706, + "loss": 0.5758, + "step": 10798 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001586481745299043, + "loss": 0.5898, + "step": 10799 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015864036426965075, + "loss": 0.4644, + "step": 10800 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001586325534641824, + "loss": 0.4706, + "step": 10801 + }, + { + "epoch": 0.32, + "learning_rate": 0.000158624742113572, + "loss": 0.5671, + "step": 10802 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015861693021789204, + "loss": 0.528, + "step": 10803 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015860911777721527, + "loss": 0.559, + "step": 10804 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015860130479161425, + "loss": 0.7193, + "step": 10805 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001585934912611617, + "loss": 0.5924, + "step": 10806 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001585856771859302, + "loss": 0.5575, + "step": 10807 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001585778625659924, + "loss": 0.5353, + "step": 10808 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015857004740142103, + "loss": 0.5888, + "step": 10809 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015856223169228865, + "loss": 0.4872, + "step": 10810 + }, + { + "epoch": 0.32, + "learning_rate": 0.000158554415438668, + "loss": 0.6057, + "step": 10811 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015854659864063172, + "loss": 0.5709, + "step": 10812 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015853878129825255, + "loss": 0.6048, + "step": 10813 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015853096341160306, + "loss": 0.6425, + "step": 10814 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015852314498075604, + "loss": 0.5493, + "step": 10815 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015851532600578413, + "loss": 0.4944, + "step": 10816 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015850750648676004, + "loss": 0.6396, + "step": 10817 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001584996864237565, + "loss": 0.5605, + "step": 10818 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015849186581684617, + "loss": 0.4588, + "step": 10819 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015848404466610179, + "loss": 0.5496, + "step": 10820 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015847622297159605, + "loss": 0.6632, + "step": 10821 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001584684007334017, + "loss": 0.5981, + "step": 10822 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015846057795159148, + "loss": 0.563, + "step": 10823 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015845275462623808, + "loss": 0.5261, + "step": 10824 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001584449307574143, + "loss": 0.5877, + "step": 10825 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001584371063451928, + "loss": 0.5075, + "step": 10826 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015842928138964644, + "loss": 0.5992, + "step": 10827 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001584214558908479, + "loss": 0.477, + "step": 10828 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001584136298488699, + "loss": 0.495, + "step": 10829 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015840580326378527, + "loss": 0.6505, + "step": 10830 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015839797613566677, + "loss": 0.6151, + "step": 10831 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015839014846458715, + "loss": 0.4894, + "step": 10832 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001583823202506192, + "loss": 0.6196, + "step": 10833 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015837449149383575, + "loss": 0.5626, + "step": 10834 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015836666219430949, + "loss": 0.4673, + "step": 10835 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015835883235211328, + "loss": 0.5325, + "step": 10836 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001583510019673199, + "loss": 0.5826, + "step": 10837 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015834317104000217, + "loss": 0.6037, + "step": 10838 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015833533957023292, + "loss": 0.5944, + "step": 10839 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001583275075580849, + "loss": 0.5531, + "step": 10840 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015831967500363095, + "loss": 0.5117, + "step": 10841 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015831184190694394, + "loss": 0.6927, + "step": 10842 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015830400826809664, + "loss": 0.5129, + "step": 10843 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015829617408716194, + "loss": 0.5099, + "step": 10844 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001582883393642126, + "loss": 0.5884, + "step": 10845 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015828050409932157, + "loss": 0.5172, + "step": 10846 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015827266829256158, + "loss": 0.5732, + "step": 10847 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015826483194400558, + "loss": 0.5496, + "step": 10848 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015825699505372636, + "loss": 0.5319, + "step": 10849 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015824915762179686, + "loss": 0.4866, + "step": 10850 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015824131964828987, + "loss": 0.5625, + "step": 10851 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015823348113327835, + "loss": 0.5625, + "step": 10852 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001582256420768351, + "loss": 0.5415, + "step": 10853 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015821780247903306, + "loss": 0.712, + "step": 10854 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015820996233994507, + "loss": 0.472, + "step": 10855 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015820212165964406, + "loss": 0.5743, + "step": 10856 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015819428043820292, + "loss": 0.6385, + "step": 10857 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015818643867569456, + "loss": 0.5751, + "step": 10858 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015817859637219187, + "loss": 0.6482, + "step": 10859 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001581707535277678, + "loss": 0.5822, + "step": 10860 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015816291014249525, + "loss": 0.6164, + "step": 10861 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001581550662164471, + "loss": 0.5393, + "step": 10862 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015814722174969635, + "loss": 0.6224, + "step": 10863 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001581393767423159, + "loss": 0.6209, + "step": 10864 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015813153119437873, + "loss": 0.4473, + "step": 10865 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015812368510595775, + "loss": 0.5563, + "step": 10866 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015811583847712588, + "loss": 0.4379, + "step": 10867 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001581079913079561, + "loss": 0.5519, + "step": 10868 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001581001435985214, + "loss": 0.6192, + "step": 10869 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001580922953488947, + "loss": 0.5604, + "step": 10870 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015808444655914902, + "loss": 0.6076, + "step": 10871 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015807659722935727, + "loss": 0.6099, + "step": 10872 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001580687473595925, + "loss": 0.4329, + "step": 10873 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015806089694992765, + "loss": 0.5419, + "step": 10874 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015805304600043573, + "loss": 0.5379, + "step": 10875 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001580451945111897, + "loss": 0.529, + "step": 10876 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015803734248226263, + "loss": 0.5933, + "step": 10877 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015802948991372744, + "loss": 0.4511, + "step": 10878 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001580216368056572, + "loss": 0.5041, + "step": 10879 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001580137831581249, + "loss": 0.5352, + "step": 10880 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015800592897120355, + "loss": 0.4923, + "step": 10881 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015799807424496623, + "loss": 0.6481, + "step": 10882 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015799021897948588, + "loss": 0.5749, + "step": 10883 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015798236317483562, + "loss": 0.5142, + "step": 10884 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015797450683108845, + "loss": 0.5393, + "step": 10885 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015796664994831741, + "loss": 0.4756, + "step": 10886 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001579587925265956, + "loss": 0.5798, + "step": 10887 + }, + { + "epoch": 0.32, + "learning_rate": 0.000157950934565996, + "loss": 0.6377, + "step": 10888 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015794307606659173, + "loss": 0.673, + "step": 10889 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001579352170284558, + "loss": 0.523, + "step": 10890 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015792735745166135, + "loss": 0.6135, + "step": 10891 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001579194973362814, + "loss": 0.5921, + "step": 10892 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015791163668238907, + "loss": 0.6845, + "step": 10893 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001579037754900574, + "loss": 0.5509, + "step": 10894 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001578959137593595, + "loss": 0.602, + "step": 10895 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001578880514903685, + "loss": 0.6162, + "step": 10896 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015788018868315745, + "loss": 0.5183, + "step": 10897 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015787232533779943, + "loss": 0.6846, + "step": 10898 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015786446145436766, + "loss": 0.5157, + "step": 10899 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015785659703293514, + "loss": 0.4935, + "step": 10900 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015784873207357508, + "loss": 0.5235, + "step": 10901 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015784086657636055, + "loss": 0.5114, + "step": 10902 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015783300054136471, + "loss": 0.4539, + "step": 10903 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015782513396866068, + "loss": 0.5155, + "step": 10904 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015781726685832156, + "loss": 0.5265, + "step": 10905 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001578093992104206, + "loss": 0.5999, + "step": 10906 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015780153102503083, + "loss": 0.5655, + "step": 10907 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001577936623022255, + "loss": 0.6311, + "step": 10908 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001577857930420777, + "loss": 0.6569, + "step": 10909 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015777792324466066, + "loss": 0.6262, + "step": 10910 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015777005291004747, + "loss": 0.5798, + "step": 10911 + }, + { + "epoch": 0.32, + "learning_rate": 0.0001577621820383114, + "loss": 0.5916, + "step": 10912 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015775431062952557, + "loss": 0.5955, + "step": 10913 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015774643868376317, + "loss": 0.4979, + "step": 10914 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015773856620109741, + "loss": 0.5684, + "step": 10915 + }, + { + "epoch": 0.32, + "learning_rate": 0.00015773069318160147, + "loss": 0.494, + "step": 10916 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015772281962534853, + "loss": 0.6527, + "step": 10917 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015771494553241185, + "loss": 0.4391, + "step": 10918 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001577070709028646, + "loss": 0.5779, + "step": 10919 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015769919573678002, + "loss": 0.6177, + "step": 10920 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015769132003423127, + "loss": 0.5448, + "step": 10921 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015768344379529168, + "loss": 0.4642, + "step": 10922 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015767556702003436, + "loss": 0.6066, + "step": 10923 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015766768970853264, + "loss": 0.5393, + "step": 10924 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015765981186085975, + "loss": 0.5315, + "step": 10925 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015765193347708887, + "loss": 0.5198, + "step": 10926 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001576440545572933, + "loss": 0.722, + "step": 10927 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015763617510154628, + "loss": 0.4719, + "step": 10928 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015762829510992108, + "loss": 0.5644, + "step": 10929 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015762041458249098, + "loss": 0.6477, + "step": 10930 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015761253351932917, + "loss": 0.6148, + "step": 10931 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015760465192050905, + "loss": 0.6367, + "step": 10932 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015759676978610384, + "loss": 0.6099, + "step": 10933 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015758888711618678, + "loss": 0.5954, + "step": 10934 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001575810039108312, + "loss": 0.655, + "step": 10935 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001575731201701104, + "loss": 0.5973, + "step": 10936 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001575652358940977, + "loss": 0.4775, + "step": 10937 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015755735108286635, + "loss": 0.5338, + "step": 10938 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015754946573648973, + "loss": 0.5263, + "step": 10939 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015754157985504107, + "loss": 0.54, + "step": 10940 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015753369343859378, + "loss": 0.509, + "step": 10941 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015752580648722107, + "loss": 0.6323, + "step": 10942 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001575179190009964, + "loss": 0.4884, + "step": 10943 + }, + { + "epoch": 0.33, + "learning_rate": 0.000157510030979993, + "loss": 0.5333, + "step": 10944 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015750214242428432, + "loss": 0.611, + "step": 10945 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015749425333394355, + "loss": 0.4983, + "step": 10946 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001574863637090442, + "loss": 0.4854, + "step": 10947 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001574784735496595, + "loss": 0.5555, + "step": 10948 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001574705828558629, + "loss": 0.6315, + "step": 10949 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015746269162772772, + "loss": 0.5817, + "step": 10950 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001574547998653273, + "loss": 0.5957, + "step": 10951 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015744690756873509, + "loss": 0.5523, + "step": 10952 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015743901473802443, + "loss": 0.6243, + "step": 10953 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015743112137326867, + "loss": 0.593, + "step": 10954 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015742322747454125, + "loss": 0.4585, + "step": 10955 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015741533304191555, + "loss": 0.5624, + "step": 10956 + }, + { + "epoch": 0.33, + "learning_rate": 0.000157407438075465, + "loss": 0.6039, + "step": 10957 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015739954257526292, + "loss": 0.5542, + "step": 10958 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015739164654138282, + "loss": 0.6208, + "step": 10959 + }, + { + "epoch": 0.33, + "learning_rate": 0.000157383749973898, + "loss": 0.6112, + "step": 10960 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015737585287288198, + "loss": 0.4465, + "step": 10961 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015736795523840815, + "loss": 0.6138, + "step": 10962 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015736005707054992, + "loss": 0.5654, + "step": 10963 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015735215836938074, + "loss": 0.6312, + "step": 10964 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015734425913497408, + "loss": 0.576, + "step": 10965 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015733635936740337, + "loss": 0.6524, + "step": 10966 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015732845906674199, + "loss": 0.5514, + "step": 10967 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015732055823306347, + "loss": 0.54, + "step": 10968 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015731265686644123, + "loss": 0.5881, + "step": 10969 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015730475496694878, + "loss": 0.584, + "step": 10970 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015729685253465953, + "loss": 0.5064, + "step": 10971 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015728894956964698, + "loss": 0.5336, + "step": 10972 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001572810460719846, + "loss": 0.5986, + "step": 10973 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015727314204174592, + "loss": 0.5971, + "step": 10974 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015726523747900436, + "loss": 0.6692, + "step": 10975 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015725733238383349, + "loss": 0.4999, + "step": 10976 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001572494267563067, + "loss": 0.5061, + "step": 10977 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015724152059649763, + "loss": 0.5958, + "step": 10978 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015723361390447966, + "loss": 0.4472, + "step": 10979 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015722570668032638, + "loss": 0.5553, + "step": 10980 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015721779892411127, + "loss": 0.6573, + "step": 10981 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015720989063590788, + "loss": 0.5915, + "step": 10982 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015720198181578976, + "loss": 0.6209, + "step": 10983 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001571940724638304, + "loss": 0.6287, + "step": 10984 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001571861625801033, + "loss": 0.4939, + "step": 10985 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015717825216468213, + "loss": 0.629, + "step": 10986 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015717034121764026, + "loss": 0.5624, + "step": 10987 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015716242973905144, + "loss": 0.6418, + "step": 10988 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015715451772898908, + "loss": 0.6162, + "step": 10989 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015714660518752682, + "loss": 0.5534, + "step": 10990 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015713869211473816, + "loss": 0.544, + "step": 10991 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015713077851069675, + "loss": 0.439, + "step": 10992 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001571228643754761, + "loss": 0.5912, + "step": 10993 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001571149497091499, + "loss": 0.6109, + "step": 10994 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001571070345117916, + "loss": 0.5297, + "step": 10995 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015709911878347482, + "loss": 0.5381, + "step": 10996 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015709120252427322, + "loss": 0.4518, + "step": 10997 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015708328573426042, + "loss": 0.5913, + "step": 10998 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015707536841350991, + "loss": 0.5395, + "step": 10999 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015706745056209544, + "loss": 0.6156, + "step": 11000 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001570595321800905, + "loss": 0.5398, + "step": 11001 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015705161326756882, + "loss": 0.6277, + "step": 11002 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015704369382460397, + "loss": 0.7262, + "step": 11003 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015703577385126955, + "loss": 0.5123, + "step": 11004 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001570278533476393, + "loss": 0.5497, + "step": 11005 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015701993231378674, + "loss": 0.5677, + "step": 11006 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015701201074978563, + "loss": 0.6017, + "step": 11007 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001570040886557095, + "loss": 0.5203, + "step": 11008 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001569961660316321, + "loss": 0.6575, + "step": 11009 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015698824287762708, + "loss": 0.5592, + "step": 11010 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001569803191937681, + "loss": 0.5405, + "step": 11011 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015697239498012878, + "loss": 0.4779, + "step": 11012 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015696447023678289, + "loss": 0.6267, + "step": 11013 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015695654496380404, + "loss": 0.5993, + "step": 11014 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015694861916126592, + "loss": 0.5723, + "step": 11015 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015694069282924224, + "loss": 0.6372, + "step": 11016 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015693276596780671, + "loss": 0.4459, + "step": 11017 + }, + { + "epoch": 0.33, + "learning_rate": 0.000156924838577033, + "loss": 0.7205, + "step": 11018 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015691691065699483, + "loss": 0.6126, + "step": 11019 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001569089822077659, + "loss": 0.5919, + "step": 11020 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015690105322941996, + "loss": 0.5523, + "step": 11021 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015689312372203066, + "loss": 0.5911, + "step": 11022 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015688519368567183, + "loss": 0.5184, + "step": 11023 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001568772631204171, + "loss": 0.4918, + "step": 11024 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015686933202634028, + "loss": 0.5522, + "step": 11025 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015686140040351503, + "loss": 0.4392, + "step": 11026 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015685346825201516, + "loss": 0.537, + "step": 11027 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015684553557191442, + "loss": 0.4897, + "step": 11028 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015683760236328653, + "loss": 0.6047, + "step": 11029 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015682966862620528, + "loss": 0.5593, + "step": 11030 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001568217343607444, + "loss": 0.5994, + "step": 11031 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001568137995669777, + "loss": 0.6419, + "step": 11032 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015680586424497895, + "loss": 0.5695, + "step": 11033 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001567979283948219, + "loss": 0.5497, + "step": 11034 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015678999201658032, + "loss": 0.5543, + "step": 11035 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015678205511032806, + "loss": 0.5251, + "step": 11036 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001567741176761389, + "loss": 0.4827, + "step": 11037 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001567661797140866, + "loss": 0.5079, + "step": 11038 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015675824122424498, + "loss": 0.5536, + "step": 11039 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015675030220668786, + "loss": 0.4197, + "step": 11040 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015674236266148903, + "loss": 0.5332, + "step": 11041 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015673442258872235, + "loss": 0.5598, + "step": 11042 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001567264819884616, + "loss": 0.5692, + "step": 11043 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015671854086078067, + "loss": 0.7241, + "step": 11044 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015671059920575333, + "loss": 0.489, + "step": 11045 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001567026570234534, + "loss": 0.6129, + "step": 11046 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001566947143139548, + "loss": 0.6131, + "step": 11047 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001566867710773314, + "loss": 0.5682, + "step": 11048 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015667882731365692, + "loss": 0.5861, + "step": 11049 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001566708830230053, + "loss": 0.5445, + "step": 11050 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001566629382054504, + "loss": 0.5824, + "step": 11051 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015665499286106612, + "loss": 0.6854, + "step": 11052 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015664704698992624, + "loss": 0.5795, + "step": 11053 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015663910059210477, + "loss": 0.5046, + "step": 11054 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015663115366767544, + "loss": 0.5065, + "step": 11055 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015662320621671224, + "loss": 0.5463, + "step": 11056 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015661525823928906, + "loss": 0.5591, + "step": 11057 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015660730973547974, + "loss": 0.4647, + "step": 11058 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015659936070535824, + "loss": 0.6553, + "step": 11059 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015659141114899845, + "loss": 0.6242, + "step": 11060 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015658346106647427, + "loss": 0.4707, + "step": 11061 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001565755104578596, + "loss": 0.5796, + "step": 11062 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001565675593232284, + "loss": 0.5642, + "step": 11063 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015655960766265453, + "loss": 0.5255, + "step": 11064 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015655165547621204, + "loss": 0.4796, + "step": 11065 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015654370276397476, + "loss": 0.4899, + "step": 11066 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015653574952601668, + "loss": 0.5899, + "step": 11067 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015652779576241175, + "loss": 0.6241, + "step": 11068 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015651984147323388, + "loss": 0.5765, + "step": 11069 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015651188665855706, + "loss": 0.6885, + "step": 11070 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015650393131845522, + "loss": 0.5647, + "step": 11071 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015649597545300237, + "loss": 0.5693, + "step": 11072 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015648801906227244, + "loss": 0.5624, + "step": 11073 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015648006214633945, + "loss": 0.4828, + "step": 11074 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015647210470527733, + "loss": 0.5547, + "step": 11075 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001564641467391601, + "loss": 0.5125, + "step": 11076 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015645618824806175, + "loss": 0.6152, + "step": 11077 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015644822923205622, + "loss": 0.4758, + "step": 11078 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001564402696912176, + "loss": 0.6083, + "step": 11079 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015643230962561982, + "loss": 0.4958, + "step": 11080 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001564243490353369, + "loss": 0.704, + "step": 11081 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001564163879204429, + "loss": 0.5906, + "step": 11082 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001564084262810118, + "loss": 0.6678, + "step": 11083 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015640046411711763, + "loss": 0.6769, + "step": 11084 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015639250142883445, + "loss": 0.4724, + "step": 11085 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015638453821623625, + "loss": 0.6646, + "step": 11086 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015637657447939709, + "loss": 0.5024, + "step": 11087 + }, + { + "epoch": 0.33, + "learning_rate": 0.000156368610218391, + "loss": 0.5596, + "step": 11088 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015636064543329202, + "loss": 0.4321, + "step": 11089 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015635268012417425, + "loss": 0.5499, + "step": 11090 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015634471429111174, + "loss": 0.3994, + "step": 11091 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001563367479341785, + "loss": 0.554, + "step": 11092 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015632878105344865, + "loss": 0.7301, + "step": 11093 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015632081364899624, + "loss": 0.559, + "step": 11094 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015631284572089534, + "loss": 0.496, + "step": 11095 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015630487726922006, + "loss": 0.6589, + "step": 11096 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015629690829404447, + "loss": 0.5151, + "step": 11097 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015628893879544262, + "loss": 0.5518, + "step": 11098 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001562809687734887, + "loss": 0.4884, + "step": 11099 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015627299822825675, + "loss": 0.6355, + "step": 11100 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001562650271598209, + "loss": 0.534, + "step": 11101 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015625705556825524, + "loss": 0.4787, + "step": 11102 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001562490834536339, + "loss": 0.4281, + "step": 11103 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015624111081603104, + "loss": 0.5137, + "step": 11104 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001562331376555207, + "loss": 0.555, + "step": 11105 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001562251639721771, + "loss": 0.5128, + "step": 11106 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001562171897660743, + "loss": 0.5301, + "step": 11107 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015620921503728646, + "loss": 0.5748, + "step": 11108 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015620123978588778, + "loss": 0.7015, + "step": 11109 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015619326401195236, + "loss": 0.5408, + "step": 11110 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015618528771555435, + "loss": 0.6421, + "step": 11111 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015617731089676796, + "loss": 0.4946, + "step": 11112 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015616933355566727, + "loss": 0.5687, + "step": 11113 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015616135569232657, + "loss": 0.5466, + "step": 11114 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015615337730681992, + "loss": 0.5389, + "step": 11115 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015614539839922158, + "loss": 0.5195, + "step": 11116 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015613741896960569, + "loss": 0.6538, + "step": 11117 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015612943901804644, + "loss": 0.5498, + "step": 11118 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015612145854461805, + "loss": 0.4791, + "step": 11119 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015611347754939468, + "loss": 0.4856, + "step": 11120 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001561054960324506, + "loss": 0.6109, + "step": 11121 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015609751399385996, + "loss": 0.5794, + "step": 11122 + }, + { + "epoch": 0.33, + "learning_rate": 0.000156089531433697, + "loss": 0.5714, + "step": 11123 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015608154835203592, + "loss": 0.4556, + "step": 11124 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015607356474895098, + "loss": 0.4888, + "step": 11125 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015606558062451637, + "loss": 0.4712, + "step": 11126 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015605759597880632, + "loss": 0.5498, + "step": 11127 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001560496108118951, + "loss": 0.5026, + "step": 11128 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015604162512385694, + "loss": 0.498, + "step": 11129 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015603363891476607, + "loss": 0.5867, + "step": 11130 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015602565218469678, + "loss": 0.5389, + "step": 11131 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001560176649337233, + "loss": 0.6368, + "step": 11132 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001560096771619199, + "loss": 0.5712, + "step": 11133 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015600168886936085, + "loss": 0.6202, + "step": 11134 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001559937000561204, + "loss": 0.5954, + "step": 11135 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015598571072227287, + "loss": 0.5691, + "step": 11136 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001559777208678925, + "loss": 0.5867, + "step": 11137 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015596973049305358, + "loss": 0.6, + "step": 11138 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015596173959783041, + "loss": 0.6178, + "step": 11139 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001559537481822973, + "loss": 0.5503, + "step": 11140 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015594575624652853, + "loss": 0.5561, + "step": 11141 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015593776379059843, + "loss": 0.6431, + "step": 11142 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015592977081458132, + "loss": 0.4894, + "step": 11143 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015592177731855143, + "loss": 0.4607, + "step": 11144 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001559137833025832, + "loss": 0.5846, + "step": 11145 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015590578876675085, + "loss": 0.7042, + "step": 11146 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001558977937111288, + "loss": 0.5945, + "step": 11147 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015588979813579133, + "loss": 0.5499, + "step": 11148 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001558818020408128, + "loss": 0.5967, + "step": 11149 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015587380542626754, + "loss": 0.6426, + "step": 11150 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015586580829222987, + "loss": 0.7168, + "step": 11151 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015585781063877423, + "loss": 0.5082, + "step": 11152 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015584981246597493, + "loss": 0.5916, + "step": 11153 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001558418137739063, + "loss": 0.4911, + "step": 11154 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015583381456264276, + "loss": 0.5085, + "step": 11155 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015582581483225867, + "loss": 0.5737, + "step": 11156 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001558178145828284, + "loss": 0.5612, + "step": 11157 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001558098138144263, + "loss": 0.5158, + "step": 11158 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015580181252712686, + "loss": 0.6085, + "step": 11159 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015579381072100437, + "loss": 0.5262, + "step": 11160 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015578580839613326, + "loss": 0.5757, + "step": 11161 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015577780555258794, + "loss": 0.491, + "step": 11162 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015576980219044284, + "loss": 0.6443, + "step": 11163 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015576179830977232, + "loss": 0.636, + "step": 11164 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015575379391065085, + "loss": 0.5504, + "step": 11165 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001557457889931528, + "loss": 0.525, + "step": 11166 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015573778355735263, + "loss": 0.5388, + "step": 11167 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015572977760332476, + "loss": 0.4893, + "step": 11168 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015572177113114366, + "loss": 0.5841, + "step": 11169 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001557137641408837, + "loss": 0.6749, + "step": 11170 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001557057566326194, + "loss": 0.5767, + "step": 11171 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015569774860642514, + "loss": 0.5263, + "step": 11172 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015568974006237546, + "loss": 0.4839, + "step": 11173 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015568173100054473, + "loss": 0.6179, + "step": 11174 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015567372142100746, + "loss": 0.5464, + "step": 11175 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015566571132383814, + "loss": 0.6051, + "step": 11176 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001556577007091112, + "loss": 0.456, + "step": 11177 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015564968957690117, + "loss": 0.6291, + "step": 11178 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015564167792728246, + "loss": 0.4891, + "step": 11179 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015563366576032964, + "loss": 0.4946, + "step": 11180 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015562565307611717, + "loss": 0.5674, + "step": 11181 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015561763987471955, + "loss": 0.5576, + "step": 11182 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015560962615621127, + "loss": 0.5942, + "step": 11183 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015560161192066684, + "loss": 0.6175, + "step": 11184 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001555935971681608, + "loss": 0.6342, + "step": 11185 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015558558189876764, + "loss": 0.5154, + "step": 11186 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001555775661125619, + "loss": 0.4967, + "step": 11187 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001555695498096181, + "loss": 0.5699, + "step": 11188 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015556153299001077, + "loss": 0.5615, + "step": 11189 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015555351565381447, + "loss": 0.3993, + "step": 11190 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015554549780110369, + "loss": 0.5513, + "step": 11191 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015553747943195304, + "loss": 0.5778, + "step": 11192 + }, + { + "epoch": 0.33, + "learning_rate": 0.000155529460546437, + "loss": 0.4501, + "step": 11193 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015552144114463023, + "loss": 0.5712, + "step": 11194 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015551342122660717, + "loss": 0.5686, + "step": 11195 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015550540079244248, + "loss": 0.6323, + "step": 11196 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015549737984221068, + "loss": 0.5775, + "step": 11197 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015548935837598639, + "loss": 0.5565, + "step": 11198 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015548133639384413, + "loss": 0.5369, + "step": 11199 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015547331389585852, + "loss": 0.4938, + "step": 11200 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001554652908821042, + "loss": 0.5399, + "step": 11201 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015545726735265562, + "loss": 0.4241, + "step": 11202 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015544924330758753, + "loss": 0.5797, + "step": 11203 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015544121874697445, + "loss": 0.4437, + "step": 11204 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015543319367089105, + "loss": 0.5907, + "step": 11205 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015542516807941188, + "loss": 0.5435, + "step": 11206 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001554171419726116, + "loss": 0.5377, + "step": 11207 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015540911535056478, + "loss": 0.4866, + "step": 11208 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015540108821334616, + "loss": 0.5785, + "step": 11209 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015539306056103027, + "loss": 0.5253, + "step": 11210 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015538503239369175, + "loss": 0.5741, + "step": 11211 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015537700371140531, + "loss": 0.5859, + "step": 11212 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015536897451424558, + "loss": 0.6273, + "step": 11213 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001553609448022871, + "loss": 0.5519, + "step": 11214 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015535291457560472, + "loss": 0.615, + "step": 11215 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015534488383427294, + "loss": 0.5337, + "step": 11216 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015533685257836655, + "loss": 0.598, + "step": 11217 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001553288208079601, + "loss": 0.5298, + "step": 11218 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015532078852312835, + "loss": 0.5983, + "step": 11219 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015531275572394594, + "loss": 0.6835, + "step": 11220 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015530472241048762, + "loss": 0.4947, + "step": 11221 + }, + { + "epoch": 0.33, + "learning_rate": 0.000155296688582828, + "loss": 0.5897, + "step": 11222 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001552886542410418, + "loss": 0.6058, + "step": 11223 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015528061938520376, + "loss": 0.5216, + "step": 11224 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001552725840153885, + "loss": 0.618, + "step": 11225 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015526454813167084, + "loss": 0.6193, + "step": 11226 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015525651173412542, + "loss": 0.5596, + "step": 11227 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015524847482282695, + "loss": 0.6048, + "step": 11228 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001552404373978502, + "loss": 0.6397, + "step": 11229 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015523239945926988, + "loss": 0.55, + "step": 11230 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015522436100716075, + "loss": 0.5978, + "step": 11231 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015521632204159748, + "loss": 0.4136, + "step": 11232 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015520828256265488, + "loss": 0.5501, + "step": 11233 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001552002425704077, + "loss": 0.4878, + "step": 11234 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015519220206493058, + "loss": 0.5672, + "step": 11235 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015518416104629842, + "loss": 0.5412, + "step": 11236 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015517611951458592, + "loss": 0.5896, + "step": 11237 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015516807746986786, + "loss": 0.5614, + "step": 11238 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015516003491221897, + "loss": 0.6577, + "step": 11239 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015515199184171409, + "loss": 0.5937, + "step": 11240 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015514394825842794, + "loss": 0.4632, + "step": 11241 + }, + { + "epoch": 0.33, + "learning_rate": 0.0001551359041624354, + "loss": 0.6342, + "step": 11242 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015512785955381115, + "loss": 0.5813, + "step": 11243 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015511981443263004, + "loss": 0.6287, + "step": 11244 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015511176879896686, + "loss": 0.5108, + "step": 11245 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015510372265289642, + "loss": 0.4899, + "step": 11246 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015509567599449354, + "loss": 0.6109, + "step": 11247 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015508762882383303, + "loss": 0.5777, + "step": 11248 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015507958114098968, + "loss": 0.5732, + "step": 11249 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015507153294603833, + "loss": 0.4924, + "step": 11250 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015506348423905385, + "loss": 0.6439, + "step": 11251 + }, + { + "epoch": 0.33, + "learning_rate": 0.00015505543502011105, + "loss": 0.5408, + "step": 11252 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001550473852892847, + "loss": 0.5178, + "step": 11253 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015503933504664973, + "loss": 0.5775, + "step": 11254 + }, + { + "epoch": 0.34, + "learning_rate": 0.000155031284292281, + "loss": 0.5216, + "step": 11255 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015502323302625325, + "loss": 0.5418, + "step": 11256 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015501518124864144, + "loss": 0.5643, + "step": 11257 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015500712895952042, + "loss": 0.5706, + "step": 11258 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015499907615896502, + "loss": 0.5546, + "step": 11259 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015499102284705015, + "loss": 0.6638, + "step": 11260 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015498296902385062, + "loss": 0.4842, + "step": 11261 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001549749146894414, + "loss": 0.5898, + "step": 11262 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015496685984389732, + "loss": 0.4531, + "step": 11263 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015495880448729329, + "loss": 0.5445, + "step": 11264 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001549507486197042, + "loss": 0.6218, + "step": 11265 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015494269224120493, + "loss": 0.6405, + "step": 11266 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015493463535187044, + "loss": 0.7043, + "step": 11267 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001549265779517756, + "loss": 0.5376, + "step": 11268 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015491852004099533, + "loss": 0.5913, + "step": 11269 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001549104616196045, + "loss": 0.6132, + "step": 11270 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015490240268767815, + "loss": 0.4912, + "step": 11271 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015489434324529114, + "loss": 0.5877, + "step": 11272 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015488628329251837, + "loss": 0.626, + "step": 11273 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001548782228294348, + "loss": 0.4758, + "step": 11274 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015487016185611544, + "loss": 0.5267, + "step": 11275 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001548621003726351, + "loss": 0.5905, + "step": 11276 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015485403837906892, + "loss": 0.6044, + "step": 11277 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001548459758754917, + "loss": 0.5781, + "step": 11278 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015483791286197845, + "loss": 0.5504, + "step": 11279 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015482984933860413, + "loss": 0.6204, + "step": 11280 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015482178530544375, + "loss": 0.5248, + "step": 11281 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015481372076257225, + "loss": 0.5735, + "step": 11282 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015480565571006463, + "loss": 0.567, + "step": 11283 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015479759014799588, + "loss": 0.496, + "step": 11284 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015478952407644095, + "loss": 0.538, + "step": 11285 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015478145749547485, + "loss": 0.536, + "step": 11286 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015477339040517266, + "loss": 0.5782, + "step": 11287 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015476532280560921, + "loss": 0.6504, + "step": 11288 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001547572546968597, + "loss": 0.5922, + "step": 11289 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015474918607899903, + "loss": 0.5025, + "step": 11290 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015474111695210226, + "loss": 0.5818, + "step": 11291 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001547330473162444, + "loss": 0.6471, + "step": 11292 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015472497717150047, + "loss": 0.4757, + "step": 11293 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015471690651794554, + "loss": 0.6218, + "step": 11294 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001547088353556546, + "loss": 0.5984, + "step": 11295 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001547007636847027, + "loss": 0.6592, + "step": 11296 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015469269150516493, + "loss": 0.5198, + "step": 11297 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001546846188171163, + "loss": 0.5195, + "step": 11298 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015467654562063187, + "loss": 0.5219, + "step": 11299 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015466847191578673, + "loss": 0.6306, + "step": 11300 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001546603977026559, + "loss": 0.6047, + "step": 11301 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015465232298131451, + "loss": 0.4413, + "step": 11302 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015464424775183758, + "loss": 0.4795, + "step": 11303 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015463617201430023, + "loss": 0.5229, + "step": 11304 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015462809576877754, + "loss": 0.5509, + "step": 11305 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015462001901534454, + "loss": 0.4485, + "step": 11306 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001546119417540764, + "loss": 0.5658, + "step": 11307 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015460386398504817, + "loss": 0.5518, + "step": 11308 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015459578570833503, + "loss": 0.5405, + "step": 11309 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015458770692401198, + "loss": 0.5755, + "step": 11310 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015457962763215424, + "loss": 0.6476, + "step": 11311 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015457154783283681, + "loss": 0.5309, + "step": 11312 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015456346752613494, + "loss": 0.6138, + "step": 11313 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015455538671212365, + "loss": 0.6157, + "step": 11314 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015454730539087814, + "loss": 0.5694, + "step": 11315 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015453922356247353, + "loss": 0.534, + "step": 11316 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015453114122698493, + "loss": 0.5512, + "step": 11317 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015452305838448754, + "loss": 0.4523, + "step": 11318 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015451497503505648, + "loss": 0.5721, + "step": 11319 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001545068911787669, + "loss": 0.5671, + "step": 11320 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015449880681569396, + "loss": 0.4264, + "step": 11321 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015449072194591285, + "loss": 0.6439, + "step": 11322 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015448263656949872, + "loss": 0.6147, + "step": 11323 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015447455068652677, + "loss": 0.5447, + "step": 11324 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001544664642970721, + "loss": 0.5224, + "step": 11325 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015445837740121006, + "loss": 0.5673, + "step": 11326 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015445028999901562, + "loss": 0.603, + "step": 11327 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015444220209056417, + "loss": 0.5738, + "step": 11328 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015443411367593076, + "loss": 0.6155, + "step": 11329 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001544260247551907, + "loss": 0.5881, + "step": 11330 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015441793532841916, + "loss": 0.5256, + "step": 11331 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001544098453956913, + "loss": 0.6417, + "step": 11332 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015440175495708242, + "loss": 0.5154, + "step": 11333 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001543936640126677, + "loss": 0.5721, + "step": 11334 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015438557256252234, + "loss": 0.6421, + "step": 11335 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015437748060672163, + "loss": 0.6059, + "step": 11336 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015436938814534078, + "loss": 0.5677, + "step": 11337 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015436129517845502, + "loss": 0.5911, + "step": 11338 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015435320170613962, + "loss": 0.5311, + "step": 11339 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001543451077284698, + "loss": 0.5551, + "step": 11340 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001543370132455208, + "loss": 0.577, + "step": 11341 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015432891825736795, + "loss": 0.6646, + "step": 11342 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015432082276408646, + "loss": 0.5111, + "step": 11343 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001543127267657516, + "loss": 0.5559, + "step": 11344 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015430463026243867, + "loss": 0.4179, + "step": 11345 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001542965332542229, + "loss": 0.5599, + "step": 11346 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015428843574117967, + "loss": 0.7288, + "step": 11347 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015428033772338414, + "loss": 0.6026, + "step": 11348 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015427223920091169, + "loss": 0.6003, + "step": 11349 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015426414017383756, + "loss": 0.6379, + "step": 11350 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001542560406422371, + "loss": 0.5076, + "step": 11351 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001542479406061856, + "loss": 0.3956, + "step": 11352 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001542398400657584, + "loss": 0.4801, + "step": 11353 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015423173902103074, + "loss": 0.5359, + "step": 11354 + }, + { + "epoch": 0.34, + "learning_rate": 0.000154223637472078, + "loss": 0.5819, + "step": 11355 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015421553541897549, + "loss": 0.5809, + "step": 11356 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015420743286179852, + "loss": 0.4208, + "step": 11357 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015419932980062249, + "loss": 0.5694, + "step": 11358 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015419122623552267, + "loss": 0.6669, + "step": 11359 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015418312216657442, + "loss": 0.5909, + "step": 11360 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001541750175938531, + "loss": 0.5468, + "step": 11361 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015416691251743405, + "loss": 0.5149, + "step": 11362 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015415880693739266, + "loss": 0.5285, + "step": 11363 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015415070085380425, + "loss": 0.606, + "step": 11364 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015414259426674423, + "loss": 0.538, + "step": 11365 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015413448717628794, + "loss": 0.5446, + "step": 11366 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015412637958251072, + "loss": 0.4692, + "step": 11367 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015411827148548803, + "loss": 0.5114, + "step": 11368 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015411016288529523, + "loss": 0.5227, + "step": 11369 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001541020537820077, + "loss": 0.4587, + "step": 11370 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015409394417570082, + "loss": 0.5263, + "step": 11371 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015408583406645003, + "loss": 0.5934, + "step": 11372 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001540777234543307, + "loss": 0.6485, + "step": 11373 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015406961233941822, + "loss": 0.5699, + "step": 11374 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015406150072178808, + "loss": 0.5092, + "step": 11375 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015405338860151563, + "loss": 0.5085, + "step": 11376 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001540452759786763, + "loss": 0.491, + "step": 11377 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015403716285334557, + "loss": 0.4911, + "step": 11378 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015402904922559883, + "loss": 0.4529, + "step": 11379 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001540209350955115, + "loss": 0.5915, + "step": 11380 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015401282046315906, + "loss": 0.5901, + "step": 11381 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015400470532861692, + "loss": 0.5378, + "step": 11382 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001539965896919606, + "loss": 0.5074, + "step": 11383 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015398847355326548, + "loss": 0.5276, + "step": 11384 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001539803569126071, + "loss": 0.6025, + "step": 11385 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015397223977006082, + "loss": 0.6687, + "step": 11386 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015396412212570216, + "loss": 0.5884, + "step": 11387 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015395600397960664, + "loss": 0.6435, + "step": 11388 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015394788533184967, + "loss": 0.5355, + "step": 11389 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015393976618250678, + "loss": 0.6211, + "step": 11390 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001539316465316534, + "loss": 0.6053, + "step": 11391 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015392352637936515, + "loss": 0.47, + "step": 11392 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015391540572571737, + "loss": 0.4525, + "step": 11393 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015390728457078567, + "loss": 0.5312, + "step": 11394 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015389916291464552, + "loss": 0.4732, + "step": 11395 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001538910407573724, + "loss": 0.6924, + "step": 11396 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015388291809904193, + "loss": 0.6342, + "step": 11397 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015387479493972952, + "loss": 0.6074, + "step": 11398 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015386667127951074, + "loss": 0.51, + "step": 11399 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015385854711846112, + "loss": 0.5396, + "step": 11400 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001538504224566562, + "loss": 0.5308, + "step": 11401 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015384229729417152, + "loss": 0.6195, + "step": 11402 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015383417163108263, + "loss": 0.5711, + "step": 11403 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015382604546746505, + "loss": 0.5428, + "step": 11404 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015381791880339436, + "loss": 0.5796, + "step": 11405 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015380979163894608, + "loss": 0.5417, + "step": 11406 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015380166397419584, + "loss": 0.563, + "step": 11407 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015379353580921915, + "loss": 0.4879, + "step": 11408 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001537854071440916, + "loss": 0.4775, + "step": 11409 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015377727797888878, + "loss": 0.5848, + "step": 11410 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001537691483136863, + "loss": 0.4897, + "step": 11411 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015376101814855965, + "loss": 0.5647, + "step": 11412 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001537528874835845, + "loss": 0.6277, + "step": 11413 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015374475631883645, + "loss": 0.4981, + "step": 11414 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015373662465439106, + "loss": 0.472, + "step": 11415 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015372849249032394, + "loss": 0.5273, + "step": 11416 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015372035982671073, + "loss": 0.4985, + "step": 11417 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015371222666362702, + "loss": 0.5682, + "step": 11418 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015370409300114843, + "loss": 0.5759, + "step": 11419 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001536959588393506, + "loss": 0.5354, + "step": 11420 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001536878241783091, + "loss": 0.5556, + "step": 11421 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015367968901809967, + "loss": 0.5925, + "step": 11422 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015367155335879786, + "loss": 0.5395, + "step": 11423 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015366341720047933, + "loss": 0.6238, + "step": 11424 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015365528054321975, + "loss": 0.5014, + "step": 11425 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015364714338709473, + "loss": 0.502, + "step": 11426 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015363900573218, + "loss": 0.5206, + "step": 11427 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001536308675785511, + "loss": 0.4973, + "step": 11428 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015362272892628385, + "loss": 0.4502, + "step": 11429 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001536145897754538, + "loss": 0.5654, + "step": 11430 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015360645012613665, + "loss": 0.5351, + "step": 11431 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001535983099784081, + "loss": 0.479, + "step": 11432 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015359016933234385, + "loss": 0.5516, + "step": 11433 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015358202818801955, + "loss": 0.6043, + "step": 11434 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015357388654551091, + "loss": 0.5461, + "step": 11435 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015356574440489359, + "loss": 0.7354, + "step": 11436 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015355760176624336, + "loss": 0.653, + "step": 11437 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001535494586296359, + "loss": 0.5481, + "step": 11438 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001535413149951469, + "loss": 0.6356, + "step": 11439 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015353317086285213, + "loss": 0.6615, + "step": 11440 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001535250262328272, + "loss": 0.4237, + "step": 11441 + }, + { + "epoch": 0.34, + "learning_rate": 0.000153516881105148, + "loss": 0.642, + "step": 11442 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015350873547989008, + "loss": 0.6517, + "step": 11443 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015350058935712933, + "loss": 0.5994, + "step": 11444 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001534924427369414, + "loss": 0.5101, + "step": 11445 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015348429561940204, + "loss": 0.4418, + "step": 11446 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015347614800458704, + "loss": 0.6066, + "step": 11447 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015346799989257215, + "loss": 0.5476, + "step": 11448 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015345985128343306, + "loss": 0.5843, + "step": 11449 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001534517021772456, + "loss": 0.5068, + "step": 11450 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015344355257408552, + "loss": 0.6064, + "step": 11451 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015343540247402858, + "loss": 0.6789, + "step": 11452 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015342725187715057, + "loss": 0.4879, + "step": 11453 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001534191007835273, + "loss": 0.6789, + "step": 11454 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015341094919323447, + "loss": 0.4991, + "step": 11455 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015340279710634796, + "loss": 0.5611, + "step": 11456 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015339464452294348, + "loss": 0.4248, + "step": 11457 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001533864914430969, + "loss": 0.7057, + "step": 11458 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015337833786688401, + "loss": 0.512, + "step": 11459 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001533701837943806, + "loss": 0.5915, + "step": 11460 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015336202922566248, + "loss": 0.5564, + "step": 11461 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015335387416080552, + "loss": 0.4983, + "step": 11462 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015334571859988548, + "loss": 0.5843, + "step": 11463 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001533375625429782, + "loss": 0.6163, + "step": 11464 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015332940599015952, + "loss": 0.51, + "step": 11465 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015332124894150529, + "loss": 0.5387, + "step": 11466 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015331309139709137, + "loss": 0.5407, + "step": 11467 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015330493335699352, + "loss": 0.5085, + "step": 11468 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015329677482128765, + "loss": 0.6054, + "step": 11469 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015328861579004963, + "loss": 0.5501, + "step": 11470 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015328045626335533, + "loss": 0.5267, + "step": 11471 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015327229624128054, + "loss": 0.7077, + "step": 11472 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015326413572390118, + "loss": 0.5191, + "step": 11473 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015325597471129308, + "loss": 0.5631, + "step": 11474 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001532478132035322, + "loss": 0.5446, + "step": 11475 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015323965120069434, + "loss": 0.6175, + "step": 11476 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015323148870285546, + "loss": 0.5523, + "step": 11477 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001532233257100914, + "loss": 0.6761, + "step": 11478 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001532151622224781, + "loss": 0.5323, + "step": 11479 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001532069982400914, + "loss": 0.6129, + "step": 11480 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001531988337630072, + "loss": 0.5999, + "step": 11481 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001531906687913015, + "loss": 0.5645, + "step": 11482 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015318250332505015, + "loss": 0.5074, + "step": 11483 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015317433736432908, + "loss": 0.5921, + "step": 11484 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001531661709092142, + "loss": 0.5338, + "step": 11485 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015315800395978145, + "loss": 0.5759, + "step": 11486 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015314983651610678, + "loss": 0.5229, + "step": 11487 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001531416685782661, + "loss": 0.5676, + "step": 11488 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015313350014633538, + "loss": 0.5266, + "step": 11489 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001531253312203906, + "loss": 0.5023, + "step": 11490 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001531171618005076, + "loss": 0.521, + "step": 11491 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001531089918867624, + "loss": 0.5797, + "step": 11492 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015310082147923097, + "loss": 0.5438, + "step": 11493 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015309265057798928, + "loss": 0.5534, + "step": 11494 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015308447918311327, + "loss": 0.6168, + "step": 11495 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015307630729467894, + "loss": 0.4855, + "step": 11496 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001530681349127623, + "loss": 0.554, + "step": 11497 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015305996203743923, + "loss": 0.5645, + "step": 11498 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001530517886687858, + "loss": 0.5531, + "step": 11499 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015304361480687797, + "loss": 0.6323, + "step": 11500 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015303544045179175, + "loss": 0.6319, + "step": 11501 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015302726560360318, + "loss": 0.5801, + "step": 11502 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015301909026238817, + "loss": 0.5061, + "step": 11503 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015301091442822284, + "loss": 0.7138, + "step": 11504 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015300273810118313, + "loss": 0.6032, + "step": 11505 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001529945612813451, + "loss": 0.5997, + "step": 11506 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015298638396878475, + "loss": 0.659, + "step": 11507 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015297820616357814, + "loss": 0.5379, + "step": 11508 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015297002786580127, + "loss": 0.524, + "step": 11509 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001529618490755302, + "loss": 0.5059, + "step": 11510 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015295366979284095, + "loss": 0.4651, + "step": 11511 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001529454900178096, + "loss": 0.5239, + "step": 11512 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015293730975051215, + "loss": 0.5422, + "step": 11513 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015292912899102476, + "loss": 0.6996, + "step": 11514 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015292094773942336, + "loss": 0.5211, + "step": 11515 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001529127659957841, + "loss": 0.5876, + "step": 11516 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015290458376018305, + "loss": 0.4842, + "step": 11517 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015289640103269625, + "loss": 0.5326, + "step": 11518 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001528882178133998, + "loss": 0.5726, + "step": 11519 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015288003410236979, + "loss": 0.5402, + "step": 11520 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015287184989968227, + "loss": 0.4806, + "step": 11521 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001528636652054134, + "loss": 0.4568, + "step": 11522 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015285548001963922, + "loss": 0.5272, + "step": 11523 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015284729434243585, + "loss": 0.5986, + "step": 11524 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001528391081738794, + "loss": 0.5626, + "step": 11525 + }, + { + "epoch": 0.34, + "learning_rate": 0.000152830921514046, + "loss": 0.5674, + "step": 11526 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001528227343630117, + "loss": 0.4811, + "step": 11527 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001528145467208527, + "loss": 0.5491, + "step": 11528 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001528063585876451, + "loss": 0.6088, + "step": 11529 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015279816996346504, + "loss": 0.525, + "step": 11530 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015278998084838864, + "loss": 0.5943, + "step": 11531 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015278179124249196, + "loss": 0.7224, + "step": 11532 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015277360114585131, + "loss": 0.5706, + "step": 11533 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001527654105585427, + "loss": 0.6366, + "step": 11534 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015275721948064237, + "loss": 0.622, + "step": 11535 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015274902791222641, + "loss": 0.5547, + "step": 11536 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015274083585337103, + "loss": 0.5529, + "step": 11537 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015273264330415237, + "loss": 0.5166, + "step": 11538 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001527244502646466, + "loss": 0.6043, + "step": 11539 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015271625673492993, + "loss": 0.6593, + "step": 11540 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015270806271507848, + "loss": 0.4619, + "step": 11541 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015269986820516852, + "loss": 0.5595, + "step": 11542 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015269167320527616, + "loss": 0.4465, + "step": 11543 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001526834777154776, + "loss": 0.5429, + "step": 11544 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015267528173584913, + "loss": 0.4909, + "step": 11545 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015266708526646681, + "loss": 0.5487, + "step": 11546 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015265888830740698, + "loss": 0.5491, + "step": 11547 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015265069085874575, + "loss": 0.5261, + "step": 11548 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015264249292055938, + "loss": 0.4898, + "step": 11549 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015263429449292412, + "loss": 0.538, + "step": 11550 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015262609557591619, + "loss": 0.5813, + "step": 11551 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015261789616961176, + "loss": 0.3797, + "step": 11552 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001526096962740871, + "loss": 0.5656, + "step": 11553 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015260149588941846, + "loss": 0.6857, + "step": 11554 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001525932950156821, + "loss": 0.6227, + "step": 11555 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001525850936529542, + "loss": 0.5445, + "step": 11556 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001525768918013111, + "loss": 0.5133, + "step": 11557 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015256868946082896, + "loss": 0.5083, + "step": 11558 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015256048663158417, + "loss": 0.5153, + "step": 11559 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015255228331365286, + "loss": 0.5144, + "step": 11560 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015254407950711143, + "loss": 0.4618, + "step": 11561 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015253587521203604, + "loss": 0.6096, + "step": 11562 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015252767042850304, + "loss": 0.5823, + "step": 11563 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001525194651565887, + "loss": 0.5805, + "step": 11564 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015251125939636927, + "loss": 0.6064, + "step": 11565 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015250305314792114, + "loss": 0.5869, + "step": 11566 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001524948464113205, + "loss": 0.6297, + "step": 11567 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001524866391866437, + "loss": 0.6217, + "step": 11568 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015247843147396704, + "loss": 0.5268, + "step": 11569 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015247022327336687, + "loss": 0.6041, + "step": 11570 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015246201458491947, + "loss": 0.6024, + "step": 11571 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015245380540870114, + "loss": 0.5797, + "step": 11572 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001524455957447883, + "loss": 0.4797, + "step": 11573 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015243738559325717, + "loss": 0.6526, + "step": 11574 + }, + { + "epoch": 0.34, + "learning_rate": 0.0001524291749541841, + "loss": 0.5058, + "step": 11575 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015242096382764552, + "loss": 0.6355, + "step": 11576 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015241275221371766, + "loss": 0.5418, + "step": 11577 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015240454011247696, + "loss": 0.5429, + "step": 11578 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015239632752399968, + "loss": 0.4949, + "step": 11579 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015238811444836226, + "loss": 0.5003, + "step": 11580 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015237990088564103, + "loss": 0.5951, + "step": 11581 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015237168683591236, + "loss": 0.5075, + "step": 11582 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015236347229925264, + "loss": 0.6, + "step": 11583 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015235525727573822, + "loss": 0.546, + "step": 11584 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015234704176544543, + "loss": 0.5195, + "step": 11585 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015233882576845076, + "loss": 0.6484, + "step": 11586 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015233060928483054, + "loss": 0.5008, + "step": 11587 + }, + { + "epoch": 0.34, + "learning_rate": 0.00015232239231466115, + "loss": 0.5057, + "step": 11588 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015231417485801903, + "loss": 0.6032, + "step": 11589 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015230595691498056, + "loss": 0.5833, + "step": 11590 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015229773848562215, + "loss": 0.6193, + "step": 11591 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015228951957002023, + "loss": 0.6061, + "step": 11592 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015228130016825116, + "loss": 0.5604, + "step": 11593 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015227308028039144, + "loss": 0.6419, + "step": 11594 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015226485990651743, + "loss": 0.6163, + "step": 11595 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015225663904670555, + "loss": 0.5699, + "step": 11596 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015224841770103233, + "loss": 0.5023, + "step": 11597 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001522401958695741, + "loss": 0.5314, + "step": 11598 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015223197355240738, + "loss": 0.4781, + "step": 11599 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015222375074960858, + "loss": 0.5299, + "step": 11600 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015221552746125415, + "loss": 0.6049, + "step": 11601 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015220730368742055, + "loss": 0.5725, + "step": 11602 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015219907942818427, + "loss": 0.6524, + "step": 11603 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015219085468362175, + "loss": 0.5347, + "step": 11604 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015218262945380944, + "loss": 0.5176, + "step": 11605 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015217440373882382, + "loss": 0.484, + "step": 11606 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015216617753874146, + "loss": 0.6922, + "step": 11607 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015215795085363873, + "loss": 0.5794, + "step": 11608 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001521497236835922, + "loss": 0.5989, + "step": 11609 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015214149602867826, + "loss": 0.4724, + "step": 11610 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015213326788897352, + "loss": 0.6081, + "step": 11611 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015212503926455441, + "loss": 0.5555, + "step": 11612 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015211681015549745, + "loss": 0.5248, + "step": 11613 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001521085805618792, + "loss": 0.5421, + "step": 11614 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015210035048377613, + "loss": 0.5876, + "step": 11615 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015209211992126472, + "loss": 0.5632, + "step": 11616 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001520838888744216, + "loss": 0.5071, + "step": 11617 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015207565734332322, + "loss": 0.542, + "step": 11618 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015206742532804612, + "loss": 0.5324, + "step": 11619 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015205919282866684, + "loss": 0.6581, + "step": 11620 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015205095984526196, + "loss": 0.4859, + "step": 11621 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015204272637790798, + "loss": 0.5067, + "step": 11622 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015203449242668152, + "loss": 0.6384, + "step": 11623 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015202625799165903, + "loss": 0.543, + "step": 11624 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015201802307291716, + "loss": 0.6666, + "step": 11625 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001520097876705324, + "loss": 0.6293, + "step": 11626 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015200155178458143, + "loss": 0.5421, + "step": 11627 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015199331541514074, + "loss": 0.6032, + "step": 11628 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001519850785622869, + "loss": 0.528, + "step": 11629 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015197684122609652, + "loss": 0.5966, + "step": 11630 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015196860340664617, + "loss": 0.6181, + "step": 11631 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001519603651040125, + "loss": 0.5986, + "step": 11632 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015195212631827207, + "loss": 0.5292, + "step": 11633 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015194388704950143, + "loss": 0.5955, + "step": 11634 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015193564729777727, + "loss": 0.5218, + "step": 11635 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015192740706317616, + "loss": 0.5544, + "step": 11636 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015191916634577467, + "loss": 0.5106, + "step": 11637 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015191092514564954, + "loss": 0.5623, + "step": 11638 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015190268346287723, + "loss": 0.5826, + "step": 11639 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015189444129753453, + "loss": 0.5607, + "step": 11640 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015188619864969796, + "loss": 0.5557, + "step": 11641 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001518779555194442, + "loss": 0.5935, + "step": 11642 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015186971190684988, + "loss": 0.5157, + "step": 11643 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015186146781199166, + "loss": 0.5759, + "step": 11644 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001518532232349462, + "loss": 0.5725, + "step": 11645 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015184497817579011, + "loss": 0.53, + "step": 11646 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001518367326346001, + "loss": 0.54, + "step": 11647 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015182848661145278, + "loss": 0.6148, + "step": 11648 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015182024010642486, + "loss": 0.6489, + "step": 11649 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015181199311959303, + "loss": 0.5538, + "step": 11650 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001518037456510339, + "loss": 0.5179, + "step": 11651 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001517954977008242, + "loss": 0.6014, + "step": 11652 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001517872492690406, + "loss": 0.5206, + "step": 11653 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001517790003557598, + "loss": 0.5551, + "step": 11654 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001517707509610585, + "loss": 0.5808, + "step": 11655 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001517625010850134, + "loss": 0.5602, + "step": 11656 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015175425072770117, + "loss": 0.5731, + "step": 11657 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015174599988919857, + "loss": 0.5843, + "step": 11658 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015173774856958228, + "loss": 0.5085, + "step": 11659 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015172949676892902, + "loss": 0.4865, + "step": 11660 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015172124448731552, + "loss": 0.5278, + "step": 11661 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015171299172481848, + "loss": 0.5968, + "step": 11662 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015170473848151466, + "loss": 0.5059, + "step": 11663 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015169648475748084, + "loss": 0.5327, + "step": 11664 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015168823055279366, + "loss": 0.5737, + "step": 11665 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015167997586752993, + "loss": 0.5009, + "step": 11666 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015167172070176636, + "loss": 0.6521, + "step": 11667 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015166346505557973, + "loss": 0.5062, + "step": 11668 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015165520892904683, + "loss": 0.5407, + "step": 11669 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015164695232224436, + "loss": 0.5099, + "step": 11670 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001516386952352491, + "loss": 0.5577, + "step": 11671 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015163043766813785, + "loss": 0.5841, + "step": 11672 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015162217962098735, + "loss": 0.559, + "step": 11673 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015161392109387442, + "loss": 0.5188, + "step": 11674 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001516056620868758, + "loss": 0.4512, + "step": 11675 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001515974026000683, + "loss": 0.4791, + "step": 11676 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015158914263352876, + "loss": 0.5862, + "step": 11677 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001515808821873339, + "loss": 0.5807, + "step": 11678 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015157262126156055, + "loss": 0.5025, + "step": 11679 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015156435985628553, + "loss": 0.5505, + "step": 11680 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015155609797158566, + "loss": 0.5025, + "step": 11681 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001515478356075377, + "loss": 0.4592, + "step": 11682 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015153957276421855, + "loss": 0.569, + "step": 11683 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015153130944170498, + "loss": 0.6007, + "step": 11684 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001515230456400738, + "loss": 0.5533, + "step": 11685 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001515147813594019, + "loss": 0.4971, + "step": 11686 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015150651659976615, + "loss": 0.4866, + "step": 11687 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015149825136124328, + "loss": 0.5239, + "step": 11688 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001514899856439102, + "loss": 0.5933, + "step": 11689 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015148171944784377, + "loss": 0.5266, + "step": 11690 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015147345277312079, + "loss": 0.5911, + "step": 11691 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001514651856198182, + "loss": 0.5363, + "step": 11692 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015145691798801282, + "loss": 0.5943, + "step": 11693 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015144864987778154, + "loss": 0.5901, + "step": 11694 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001514403812892012, + "loss": 0.6195, + "step": 11695 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001514321122223487, + "loss": 0.4644, + "step": 11696 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001514238426773009, + "loss": 0.5921, + "step": 11697 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015141557265413474, + "loss": 0.5301, + "step": 11698 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015140730215292705, + "loss": 0.5983, + "step": 11699 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015139903117375477, + "loss": 0.481, + "step": 11700 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015139075971669477, + "loss": 0.5735, + "step": 11701 + }, + { + "epoch": 0.35, + "learning_rate": 0.000151382487781824, + "loss": 0.4742, + "step": 11702 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001513742153692193, + "loss": 0.5918, + "step": 11703 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015136594247895763, + "loss": 0.6055, + "step": 11704 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001513576691111159, + "loss": 0.5451, + "step": 11705 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015134939526577105, + "loss": 0.6272, + "step": 11706 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015134112094299996, + "loss": 0.4967, + "step": 11707 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015133284614287962, + "loss": 0.4931, + "step": 11708 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015132457086548695, + "loss": 0.5219, + "step": 11709 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015131629511089885, + "loss": 0.4601, + "step": 11710 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015130801887919227, + "loss": 0.5354, + "step": 11711 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001512997421704442, + "loss": 0.5564, + "step": 11712 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015129146498473156, + "loss": 0.4934, + "step": 11713 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015128318732213133, + "loss": 0.5612, + "step": 11714 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015127490918272047, + "loss": 0.6305, + "step": 11715 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015126663056657592, + "loss": 0.5442, + "step": 11716 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001512583514737747, + "loss": 0.4684, + "step": 11717 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015125007190439375, + "loss": 0.5686, + "step": 11718 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015124179185851006, + "loss": 0.5284, + "step": 11719 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015123351133620062, + "loss": 0.5195, + "step": 11720 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001512252303375424, + "loss": 0.6247, + "step": 11721 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001512169488626124, + "loss": 0.5688, + "step": 11722 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015120866691148763, + "loss": 0.5329, + "step": 11723 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015120038448424508, + "loss": 0.5818, + "step": 11724 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015119210158096174, + "loss": 0.5503, + "step": 11725 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015118381820171468, + "loss": 0.5315, + "step": 11726 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015117553434658087, + "loss": 0.5453, + "step": 11727 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015116725001563733, + "loss": 0.5499, + "step": 11728 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001511589652089611, + "loss": 0.4827, + "step": 11729 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015115067992662917, + "loss": 0.5313, + "step": 11730 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015114239416871866, + "loss": 0.5176, + "step": 11731 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015113410793530648, + "loss": 0.5894, + "step": 11732 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015112582122646983, + "loss": 0.542, + "step": 11733 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015111753404228558, + "loss": 0.4652, + "step": 11734 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015110924638283096, + "loss": 0.4538, + "step": 11735 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015110095824818288, + "loss": 0.5627, + "step": 11736 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015109266963841847, + "loss": 0.5731, + "step": 11737 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001510843805536148, + "loss": 0.6101, + "step": 11738 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001510760909938489, + "loss": 0.5671, + "step": 11739 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015106780095919787, + "loss": 0.5176, + "step": 11740 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015105951044973878, + "loss": 0.504, + "step": 11741 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001510512194655487, + "loss": 0.6169, + "step": 11742 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015104292800670476, + "loss": 0.5728, + "step": 11743 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015103463607328397, + "loss": 0.6859, + "step": 11744 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015102634366536351, + "loss": 0.5462, + "step": 11745 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015101805078302042, + "loss": 0.4976, + "step": 11746 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015100975742633185, + "loss": 0.6277, + "step": 11747 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015100146359537487, + "loss": 0.5426, + "step": 11748 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015099316929022662, + "loss": 0.6175, + "step": 11749 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015098487451096424, + "loss": 0.3888, + "step": 11750 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015097657925766477, + "loss": 0.5775, + "step": 11751 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015096828353040537, + "loss": 0.6206, + "step": 11752 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015095998732926323, + "loss": 0.5627, + "step": 11753 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001509516906543154, + "loss": 0.6409, + "step": 11754 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001509433935056391, + "loss": 0.5763, + "step": 11755 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015093509588331137, + "loss": 0.5556, + "step": 11756 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001509267977874095, + "loss": 0.5512, + "step": 11757 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015091849921801046, + "loss": 0.5135, + "step": 11758 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015091020017519158, + "loss": 0.5034, + "step": 11759 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015090190065902997, + "loss": 0.5936, + "step": 11760 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015089360066960271, + "loss": 0.5131, + "step": 11761 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015088530020698708, + "loss": 0.6779, + "step": 11762 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001508769992712602, + "loss": 0.5159, + "step": 11763 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015086869786249925, + "loss": 0.5722, + "step": 11764 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001508603959807814, + "loss": 0.4561, + "step": 11765 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001508520936261839, + "loss": 0.5632, + "step": 11766 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015084379079878388, + "loss": 0.6274, + "step": 11767 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015083548749865856, + "loss": 0.6177, + "step": 11768 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015082718372588514, + "loss": 0.5673, + "step": 11769 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015081887948054082, + "loss": 0.5885, + "step": 11770 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015081057476270285, + "loss": 0.5202, + "step": 11771 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015080226957244833, + "loss": 0.4538, + "step": 11772 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001507939639098546, + "loss": 0.5681, + "step": 11773 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015078565777499887, + "loss": 0.5544, + "step": 11774 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001507773511679583, + "loss": 0.5551, + "step": 11775 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015076904408881014, + "loss": 0.6037, + "step": 11776 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015076073653763167, + "loss": 0.5632, + "step": 11777 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001507524285145001, + "loss": 0.4628, + "step": 11778 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015074412001949265, + "loss": 0.6089, + "step": 11779 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015073581105268664, + "loss": 0.5113, + "step": 11780 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015072750161415925, + "loss": 0.59, + "step": 11781 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015071919170398776, + "loss": 0.5093, + "step": 11782 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001507108813222495, + "loss": 0.4781, + "step": 11783 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001507025704690216, + "loss": 0.5707, + "step": 11784 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015069425914438145, + "loss": 0.584, + "step": 11785 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015068594734840628, + "loss": 0.5741, + "step": 11786 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015067763508117335, + "loss": 0.4719, + "step": 11787 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015066932234276, + "loss": 0.6136, + "step": 11788 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015066100913324344, + "loss": 0.563, + "step": 11789 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015065269545270108, + "loss": 0.6465, + "step": 11790 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015064438130121008, + "loss": 0.5274, + "step": 11791 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015063606667884783, + "loss": 0.5755, + "step": 11792 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015062775158569163, + "loss": 0.3819, + "step": 11793 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015061943602181875, + "loss": 0.5839, + "step": 11794 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015061111998730656, + "loss": 0.4952, + "step": 11795 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001506028034822323, + "loss": 0.4761, + "step": 11796 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001505944865066734, + "loss": 0.5234, + "step": 11797 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015058616906070705, + "loss": 0.4345, + "step": 11798 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015057785114441075, + "loss": 0.5215, + "step": 11799 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001505695327578617, + "loss": 0.6076, + "step": 11800 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001505612139011373, + "loss": 0.5838, + "step": 11801 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015055289457431488, + "loss": 0.5331, + "step": 11802 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015054457477747178, + "loss": 0.6203, + "step": 11803 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001505362545106854, + "loss": 0.5465, + "step": 11804 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015052793377403305, + "loss": 0.5227, + "step": 11805 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001505196125675921, + "loss": 0.524, + "step": 11806 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015051129089143996, + "loss": 0.6051, + "step": 11807 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015050296874565393, + "loss": 0.5423, + "step": 11808 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015049464613031143, + "loss": 0.5683, + "step": 11809 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015048632304548988, + "loss": 0.5299, + "step": 11810 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015047799949126657, + "loss": 0.5148, + "step": 11811 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015046967546771897, + "loss": 0.6291, + "step": 11812 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015046135097492442, + "loss": 0.6204, + "step": 11813 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015045302601296036, + "loss": 0.5272, + "step": 11814 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015044470058190416, + "loss": 0.5267, + "step": 11815 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015043637468183322, + "loss": 0.54, + "step": 11816 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015042804831282504, + "loss": 0.3756, + "step": 11817 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001504197214749569, + "loss": 0.6586, + "step": 11818 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015041139416830632, + "loss": 0.4946, + "step": 11819 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015040306639295064, + "loss": 0.4852, + "step": 11820 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001503947381489674, + "loss": 0.5331, + "step": 11821 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015038640943643391, + "loss": 0.499, + "step": 11822 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001503780802554277, + "loss": 0.4652, + "step": 11823 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001503697506060262, + "loss": 0.5614, + "step": 11824 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001503614204883068, + "loss": 0.5845, + "step": 11825 + }, + { + "epoch": 0.35, + "learning_rate": 0.000150353089902347, + "loss": 0.6114, + "step": 11826 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015034475884822426, + "loss": 0.5865, + "step": 11827 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015033642732601598, + "loss": 0.5168, + "step": 11828 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015032809533579967, + "loss": 0.4616, + "step": 11829 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015031976287765278, + "loss": 0.548, + "step": 11830 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015031142995165282, + "loss": 0.4563, + "step": 11831 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015030309655787721, + "loss": 0.5545, + "step": 11832 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001502947626964035, + "loss": 0.4278, + "step": 11833 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015028642836730906, + "loss": 0.5194, + "step": 11834 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015027809357067147, + "loss": 0.4961, + "step": 11835 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015026975830656828, + "loss": 0.64, + "step": 11836 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015026142257507682, + "loss": 0.5084, + "step": 11837 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015025308637627474, + "loss": 0.5245, + "step": 11838 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015024474971023945, + "loss": 0.5544, + "step": 11839 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015023641257704852, + "loss": 0.4584, + "step": 11840 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015022807497677948, + "loss": 0.6568, + "step": 11841 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015021973690950978, + "loss": 0.7072, + "step": 11842 + }, + { + "epoch": 0.35, + "learning_rate": 0.000150211398375317, + "loss": 0.537, + "step": 11843 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015020305937427864, + "loss": 0.4418, + "step": 11844 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015019471990647225, + "loss": 0.6171, + "step": 11845 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015018637997197537, + "loss": 0.6405, + "step": 11846 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001501780395708655, + "loss": 0.5424, + "step": 11847 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015016969870322025, + "loss": 0.5939, + "step": 11848 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001501613573691171, + "loss": 0.6211, + "step": 11849 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001501530155686337, + "loss": 0.4942, + "step": 11850 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001501446733018475, + "loss": 0.4777, + "step": 11851 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001501363305688361, + "loss": 0.539, + "step": 11852 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015012798736967713, + "loss": 0.581, + "step": 11853 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015011964370444808, + "loss": 0.5401, + "step": 11854 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015011129957322655, + "loss": 0.525, + "step": 11855 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015010295497609017, + "loss": 0.6349, + "step": 11856 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015009460991311646, + "loss": 0.5573, + "step": 11857 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015008626438438305, + "loss": 0.6619, + "step": 11858 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001500779183899675, + "loss": 0.4982, + "step": 11859 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015006957192994743, + "loss": 0.5537, + "step": 11860 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015006122500440042, + "loss": 0.5113, + "step": 11861 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001500528776134041, + "loss": 0.5382, + "step": 11862 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001500445297570361, + "loss": 0.4786, + "step": 11863 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015003618143537398, + "loss": 0.5339, + "step": 11864 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015002783264849538, + "loss": 0.5244, + "step": 11865 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015001948339647792, + "loss": 0.5917, + "step": 11866 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015001113367939928, + "loss": 0.5943, + "step": 11867 + }, + { + "epoch": 0.35, + "learning_rate": 0.00015000278349733704, + "loss": 0.5749, + "step": 11868 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014999443285036886, + "loss": 0.4688, + "step": 11869 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014998608173857233, + "loss": 0.5246, + "step": 11870 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014997773016202514, + "loss": 0.5121, + "step": 11871 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014996937812080497, + "loss": 0.607, + "step": 11872 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001499610256149894, + "loss": 0.5801, + "step": 11873 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014995267264465616, + "loss": 0.5043, + "step": 11874 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014994431920988286, + "loss": 0.4434, + "step": 11875 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014993596531074718, + "loss": 0.6132, + "step": 11876 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001499276109473268, + "loss": 0.5672, + "step": 11877 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014991925611969938, + "loss": 0.6118, + "step": 11878 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014991090082794266, + "loss": 0.637, + "step": 11879 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014990254507213423, + "loss": 0.5303, + "step": 11880 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014989418885235186, + "loss": 0.5214, + "step": 11881 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014988583216867318, + "loss": 0.5527, + "step": 11882 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001498774750211759, + "loss": 0.4527, + "step": 11883 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001498691174099378, + "loss": 0.5411, + "step": 11884 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014986075933503648, + "loss": 0.5926, + "step": 11885 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001498524007965497, + "loss": 0.5809, + "step": 11886 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014984404179455515, + "loss": 0.5865, + "step": 11887 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014983568232913058, + "loss": 0.5935, + "step": 11888 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014982732240035367, + "loss": 0.5362, + "step": 11889 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014981896200830222, + "loss": 0.5926, + "step": 11890 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001498106011530539, + "loss": 0.5171, + "step": 11891 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014980223983468647, + "loss": 0.5283, + "step": 11892 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014979387805327766, + "loss": 0.5225, + "step": 11893 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001497855158089052, + "loss": 0.5314, + "step": 11894 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001497771531016469, + "loss": 0.7016, + "step": 11895 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001497687899315804, + "loss": 0.5127, + "step": 11896 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014976042629878358, + "loss": 0.504, + "step": 11897 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014975206220333416, + "loss": 0.5338, + "step": 11898 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014974369764530985, + "loss": 0.6273, + "step": 11899 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001497353326247885, + "loss": 0.5361, + "step": 11900 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014972696714184782, + "loss": 0.5416, + "step": 11901 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001497186011965656, + "loss": 0.5844, + "step": 11902 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014971023478901972, + "loss": 0.5128, + "step": 11903 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001497018679192878, + "loss": 0.5829, + "step": 11904 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001496935005874478, + "loss": 0.5438, + "step": 11905 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014968513279357742, + "loss": 0.4646, + "step": 11906 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014967676453775446, + "loss": 0.7286, + "step": 11907 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014966839582005673, + "loss": 0.5135, + "step": 11908 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014966002664056204, + "loss": 0.6455, + "step": 11909 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014965165699934826, + "loss": 0.6137, + "step": 11910 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014964328689649314, + "loss": 0.5499, + "step": 11911 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014963491633207453, + "loss": 0.6588, + "step": 11912 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014962654530617023, + "loss": 0.6171, + "step": 11913 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014961817381885811, + "loss": 0.5168, + "step": 11914 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014960980187021602, + "loss": 0.4909, + "step": 11915 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001496014294603217, + "loss": 0.5553, + "step": 11916 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014959305658925312, + "loss": 0.5844, + "step": 11917 + }, + { + "epoch": 0.35, + "learning_rate": 0.000149584683257088, + "loss": 0.6376, + "step": 11918 + }, + { + "epoch": 0.35, + "learning_rate": 0.0001495763094639043, + "loss": 0.5142, + "step": 11919 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014956793520977984, + "loss": 0.5601, + "step": 11920 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014955956049479246, + "loss": 0.5825, + "step": 11921 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014955118531902005, + "loss": 0.6167, + "step": 11922 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014954280968254043, + "loss": 0.5359, + "step": 11923 + }, + { + "epoch": 0.35, + "learning_rate": 0.00014953443358543152, + "loss": 0.5602, + "step": 11924 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014952605702777124, + "loss": 0.4971, + "step": 11925 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014951768000963738, + "loss": 0.5348, + "step": 11926 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014950930253110787, + "loss": 0.6008, + "step": 11927 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014950092459226058, + "loss": 0.6024, + "step": 11928 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014949254619317346, + "loss": 0.448, + "step": 11929 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014948416733392435, + "loss": 0.5696, + "step": 11930 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014947578801459118, + "loss": 0.5665, + "step": 11931 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014946740823525186, + "loss": 0.5675, + "step": 11932 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014945902799598428, + "loss": 0.487, + "step": 11933 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014945064729686636, + "loss": 0.6136, + "step": 11934 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014944226613797606, + "loss": 0.583, + "step": 11935 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014943388451939127, + "loss": 0.6604, + "step": 11936 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001494255024411899, + "loss": 0.6273, + "step": 11937 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014941711990344993, + "loss": 0.5766, + "step": 11938 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014940873690624928, + "loss": 0.4879, + "step": 11939 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014940035344966586, + "loss": 0.6215, + "step": 11940 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014939196953377766, + "loss": 0.501, + "step": 11941 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001493835851586626, + "loss": 0.5431, + "step": 11942 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014937520032439866, + "loss": 0.6018, + "step": 11943 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014936681503106378, + "loss": 0.6476, + "step": 11944 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014935842927873595, + "loss": 0.7064, + "step": 11945 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001493500430674931, + "loss": 0.5937, + "step": 11946 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014934165639741322, + "loss": 0.6361, + "step": 11947 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014933326926857426, + "loss": 0.5637, + "step": 11948 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014932488168105428, + "loss": 0.4993, + "step": 11949 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014931649363493117, + "loss": 0.4721, + "step": 11950 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014930810513028297, + "loss": 0.5644, + "step": 11951 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014929971616718763, + "loss": 0.5188, + "step": 11952 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014929132674572322, + "loss": 0.5703, + "step": 11953 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014928293686596766, + "loss": 0.5421, + "step": 11954 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014927454652799904, + "loss": 0.5242, + "step": 11955 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001492661557318953, + "loss": 0.615, + "step": 11956 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014925776447773448, + "loss": 0.6909, + "step": 11957 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001492493727655946, + "loss": 0.4666, + "step": 11958 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001492409805955537, + "loss": 0.438, + "step": 11959 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014923258796768976, + "loss": 0.6995, + "step": 11960 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014922419488208084, + "loss": 0.5486, + "step": 11961 + }, + { + "epoch": 0.36, + "learning_rate": 0.000149215801338805, + "loss": 0.497, + "step": 11962 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014920740733794023, + "loss": 0.5178, + "step": 11963 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001491990128795646, + "loss": 0.4652, + "step": 11964 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014919061796375618, + "loss": 0.6028, + "step": 11965 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014918222259059297, + "loss": 0.53, + "step": 11966 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014917382676015307, + "loss": 0.6519, + "step": 11967 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014916543047251455, + "loss": 0.5712, + "step": 11968 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014915703372775542, + "loss": 0.5163, + "step": 11969 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001491486365259538, + "loss": 0.5439, + "step": 11970 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014914023886718772, + "loss": 0.5943, + "step": 11971 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014913184075153533, + "loss": 0.5319, + "step": 11972 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001491234421790746, + "loss": 0.5353, + "step": 11973 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014911504314988377, + "loss": 0.7208, + "step": 11974 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014910664366404077, + "loss": 0.6323, + "step": 11975 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001490982437216238, + "loss": 0.5786, + "step": 11976 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001490898433227109, + "loss": 0.6267, + "step": 11977 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014908144246738024, + "loss": 0.5093, + "step": 11978 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001490730411557099, + "loss": 0.6932, + "step": 11979 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014906463938777796, + "loss": 0.5219, + "step": 11980 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014905623716366257, + "loss": 0.5665, + "step": 11981 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014904783448344182, + "loss": 0.6317, + "step": 11982 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014903943134719384, + "loss": 0.5072, + "step": 11983 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001490310277549968, + "loss": 0.5634, + "step": 11984 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014902262370692878, + "loss": 0.5955, + "step": 11985 + }, + { + "epoch": 0.36, + "learning_rate": 0.000149014219203068, + "loss": 0.599, + "step": 11986 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014900581424349247, + "loss": 0.5104, + "step": 11987 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014899740882828047, + "loss": 0.6732, + "step": 11988 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014898900295751008, + "loss": 0.5425, + "step": 11989 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014898059663125945, + "loss": 0.5483, + "step": 11990 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014897218984960678, + "loss": 0.5239, + "step": 11991 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014896378261263014, + "loss": 0.4733, + "step": 11992 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014895537492040783, + "loss": 0.5489, + "step": 11993 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014894696677301793, + "loss": 0.5901, + "step": 11994 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014893855817053867, + "loss": 0.487, + "step": 11995 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014893014911304816, + "loss": 0.6398, + "step": 11996 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014892173960062466, + "loss": 0.6018, + "step": 11997 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001489133296333463, + "loss": 0.5471, + "step": 11998 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001489049192112913, + "loss": 0.4881, + "step": 11999 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014889650833453785, + "loss": 0.5418, + "step": 12000 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014888809700316417, + "loss": 0.6223, + "step": 12001 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014887968521724843, + "loss": 0.6122, + "step": 12002 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014887127297686886, + "loss": 0.5991, + "step": 12003 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014886286028210368, + "loss": 0.5642, + "step": 12004 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014885444713303112, + "loss": 0.5825, + "step": 12005 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014884603352972934, + "loss": 0.508, + "step": 12006 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014883761947227664, + "loss": 0.555, + "step": 12007 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014882920496075118, + "loss": 0.5726, + "step": 12008 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001488207899952313, + "loss": 0.5079, + "step": 12009 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014881237457579512, + "loss": 0.6081, + "step": 12010 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014880395870252094, + "loss": 0.5278, + "step": 12011 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014879554237548704, + "loss": 0.6722, + "step": 12012 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014878712559477159, + "loss": 0.5381, + "step": 12013 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014877870836045289, + "loss": 0.5196, + "step": 12014 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014877029067260925, + "loss": 0.518, + "step": 12015 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014876187253131884, + "loss": 0.5898, + "step": 12016 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014875345393666, + "loss": 0.5094, + "step": 12017 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014874503488871094, + "loss": 0.535, + "step": 12018 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014873661538755, + "loss": 0.6137, + "step": 12019 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014872819543325545, + "loss": 0.5408, + "step": 12020 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014871977502590555, + "loss": 0.4839, + "step": 12021 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014871135416557858, + "loss": 0.6487, + "step": 12022 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001487029328523529, + "loss": 0.5807, + "step": 12023 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001486945110863067, + "loss": 0.5676, + "step": 12024 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014868608886751838, + "loss": 0.6514, + "step": 12025 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001486776661960662, + "loss": 0.5299, + "step": 12026 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001486692430720285, + "loss": 0.6176, + "step": 12027 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014866081949548355, + "loss": 0.5269, + "step": 12028 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014865239546650968, + "loss": 0.6393, + "step": 12029 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014864397098518527, + "loss": 0.5496, + "step": 12030 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001486355460515886, + "loss": 0.5466, + "step": 12031 + }, + { + "epoch": 0.36, + "learning_rate": 0.000148627120665798, + "loss": 0.5749, + "step": 12032 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014861869482789178, + "loss": 0.6198, + "step": 12033 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014861026853794834, + "loss": 0.6045, + "step": 12034 + }, + { + "epoch": 0.36, + "learning_rate": 0.000148601841796046, + "loss": 0.5679, + "step": 12035 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014859341460226312, + "loss": 0.5522, + "step": 12036 + }, + { + "epoch": 0.36, + "learning_rate": 0.000148584986956678, + "loss": 0.4921, + "step": 12037 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014857655885936906, + "loss": 0.4768, + "step": 12038 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014856813031041463, + "loss": 0.617, + "step": 12039 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014855970130989306, + "loss": 0.556, + "step": 12040 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001485512718578828, + "loss": 0.5432, + "step": 12041 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001485428419544621, + "loss": 0.4738, + "step": 12042 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014853441159970947, + "loss": 0.5432, + "step": 12043 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014852598079370316, + "loss": 0.5663, + "step": 12044 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014851754953652165, + "loss": 0.6879, + "step": 12045 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001485091178282433, + "loss": 0.607, + "step": 12046 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001485006856689465, + "loss": 0.4564, + "step": 12047 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001484922530587097, + "loss": 0.56, + "step": 12048 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001484838199976112, + "loss": 0.5682, + "step": 12049 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014847538648572948, + "loss": 0.4597, + "step": 12050 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014846695252314295, + "loss": 0.6572, + "step": 12051 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014845851810992998, + "loss": 0.5154, + "step": 12052 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014845008324616908, + "loss": 0.563, + "step": 12053 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014844164793193857, + "loss": 0.5866, + "step": 12054 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014843321216731693, + "loss": 0.4709, + "step": 12055 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001484247759523826, + "loss": 0.539, + "step": 12056 + }, + { + "epoch": 0.36, + "learning_rate": 0.000148416339287214, + "loss": 0.4599, + "step": 12057 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014840790217188958, + "loss": 0.5782, + "step": 12058 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014839946460648776, + "loss": 0.5119, + "step": 12059 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014839102659108702, + "loss": 0.5417, + "step": 12060 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014838258812576582, + "loss": 0.6269, + "step": 12061 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014837414921060259, + "loss": 0.4313, + "step": 12062 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014836570984567578, + "loss": 0.7268, + "step": 12063 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014835727003106386, + "loss": 0.5791, + "step": 12064 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014834882976684537, + "loss": 0.6249, + "step": 12065 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001483403890530987, + "loss": 0.5625, + "step": 12066 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014833194788990237, + "loss": 0.5415, + "step": 12067 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014832350627733482, + "loss": 0.4893, + "step": 12068 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014831506421547462, + "loss": 0.5854, + "step": 12069 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014830662170440016, + "loss": 0.6325, + "step": 12070 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014829817874419002, + "loss": 0.617, + "step": 12071 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014828973533492266, + "loss": 0.6581, + "step": 12072 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001482812914766766, + "loss": 0.5713, + "step": 12073 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014827284716953027, + "loss": 0.4552, + "step": 12074 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001482644024135623, + "loss": 0.6226, + "step": 12075 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014825595720885113, + "loss": 0.5461, + "step": 12076 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014824751155547534, + "loss": 0.6126, + "step": 12077 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001482390654535134, + "loss": 0.5218, + "step": 12078 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014823061890304384, + "loss": 0.7166, + "step": 12079 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001482221719041452, + "loss": 0.63, + "step": 12080 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014821372445689604, + "loss": 0.5501, + "step": 12081 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001482052765613749, + "loss": 0.4377, + "step": 12082 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001481968282176603, + "loss": 0.49, + "step": 12083 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014818837942583079, + "loss": 0.5775, + "step": 12084 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014817993018596495, + "loss": 0.4796, + "step": 12085 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001481714804981413, + "loss": 0.4963, + "step": 12086 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014816303036243846, + "loss": 0.4448, + "step": 12087 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014815457977893494, + "loss": 0.5381, + "step": 12088 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001481461287477093, + "loss": 0.6603, + "step": 12089 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001481376772688402, + "loss": 0.6921, + "step": 12090 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014812922534240613, + "loss": 0.6127, + "step": 12091 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014812077296848573, + "loss": 0.5571, + "step": 12092 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014811232014715753, + "loss": 0.6393, + "step": 12093 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014810386687850018, + "loss": 0.6181, + "step": 12094 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014809541316259222, + "loss": 0.5436, + "step": 12095 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001480869589995123, + "loss": 0.5305, + "step": 12096 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014807850438933897, + "loss": 0.6347, + "step": 12097 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014807004933215092, + "loss": 0.5086, + "step": 12098 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014806159382802666, + "loss": 0.5506, + "step": 12099 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014805313787704486, + "loss": 0.5412, + "step": 12100 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014804468147928417, + "loss": 0.5028, + "step": 12101 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014803622463482316, + "loss": 0.4793, + "step": 12102 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014802776734374048, + "loss": 0.6378, + "step": 12103 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014801930960611477, + "loss": 0.5741, + "step": 12104 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014801085142202464, + "loss": 0.6539, + "step": 12105 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014800239279154877, + "loss": 0.5885, + "step": 12106 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014799393371476576, + "loss": 0.5733, + "step": 12107 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001479854741917543, + "loss": 0.5178, + "step": 12108 + }, + { + "epoch": 0.36, + "learning_rate": 0.000147977014222593, + "loss": 0.501, + "step": 12109 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001479685538073606, + "loss": 0.5092, + "step": 12110 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014796009294613564, + "loss": 0.6193, + "step": 12111 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001479516316389969, + "loss": 0.5212, + "step": 12112 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014794316988602296, + "loss": 0.6995, + "step": 12113 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014793470768729257, + "loss": 0.4536, + "step": 12114 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014792624504288436, + "loss": 0.4919, + "step": 12115 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014791778195287703, + "loss": 0.5408, + "step": 12116 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014790931841734926, + "loss": 0.5346, + "step": 12117 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014790085443637974, + "loss": 0.6203, + "step": 12118 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014789239001004715, + "loss": 0.5518, + "step": 12119 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014788392513843024, + "loss": 0.4355, + "step": 12120 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014787545982160766, + "loss": 0.6456, + "step": 12121 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014786699405965813, + "loss": 0.5052, + "step": 12122 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014785852785266038, + "loss": 0.62, + "step": 12123 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001478500612006931, + "loss": 0.6073, + "step": 12124 + }, + { + "epoch": 0.36, + "learning_rate": 0.000147841594103835, + "loss": 0.5506, + "step": 12125 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014783312656216485, + "loss": 0.5726, + "step": 12126 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001478246585757614, + "loss": 0.563, + "step": 12127 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014781619014470328, + "loss": 0.6618, + "step": 12128 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014780772126906928, + "loss": 0.581, + "step": 12129 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014779925194893817, + "loss": 0.5566, + "step": 12130 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014779078218438865, + "loss": 0.5623, + "step": 12131 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014778231197549947, + "loss": 0.6241, + "step": 12132 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001477738413223494, + "loss": 0.6773, + "step": 12133 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014776537022501723, + "loss": 0.641, + "step": 12134 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014775689868358168, + "loss": 0.5412, + "step": 12135 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001477484266981215, + "loss": 0.4535, + "step": 12136 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014773995426871548, + "loss": 0.6408, + "step": 12137 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014773148139544237, + "loss": 0.596, + "step": 12138 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014772300807838103, + "loss": 0.52, + "step": 12139 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001477145343176101, + "loss": 0.5666, + "step": 12140 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014770606011320854, + "loss": 0.6266, + "step": 12141 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014769758546525496, + "loss": 0.5676, + "step": 12142 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001476891103738283, + "loss": 0.5711, + "step": 12143 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014768063483900726, + "loss": 0.4773, + "step": 12144 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014767215886087073, + "loss": 0.5127, + "step": 12145 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001476636824394974, + "loss": 0.4595, + "step": 12146 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001476552055749662, + "loss": 0.4267, + "step": 12147 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014764672826735587, + "loss": 0.5385, + "step": 12148 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001476382505167453, + "loss": 0.4636, + "step": 12149 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014762977232321317, + "loss": 0.6426, + "step": 12150 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014762129368683848, + "loss": 0.5265, + "step": 12151 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014761281460769993, + "loss": 0.6333, + "step": 12152 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014760433508587646, + "loss": 0.4706, + "step": 12153 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014759585512144684, + "loss": 0.51, + "step": 12154 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001475873747144899, + "loss": 0.4995, + "step": 12155 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014757889386508453, + "loss": 0.5248, + "step": 12156 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014757041257330955, + "loss": 0.5749, + "step": 12157 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014756193083924387, + "loss": 0.6259, + "step": 12158 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014755344866296626, + "loss": 0.4075, + "step": 12159 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001475449660445557, + "loss": 0.5087, + "step": 12160 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014753648298409099, + "loss": 0.6025, + "step": 12161 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014752799948165098, + "loss": 0.5838, + "step": 12162 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014751951553731456, + "loss": 0.4425, + "step": 12163 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014751103115116066, + "loss": 0.5758, + "step": 12164 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001475025463232681, + "loss": 0.5215, + "step": 12165 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001474940610537158, + "loss": 0.5983, + "step": 12166 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014748557534258267, + "loss": 0.5037, + "step": 12167 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014747708918994755, + "loss": 0.5238, + "step": 12168 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014746860259588938, + "loss": 0.5498, + "step": 12169 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014746011556048712, + "loss": 0.5792, + "step": 12170 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001474516280838196, + "loss": 0.498, + "step": 12171 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014744314016596573, + "loss": 0.5696, + "step": 12172 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014743465180700445, + "loss": 0.6132, + "step": 12173 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001474261630070147, + "loss": 0.6008, + "step": 12174 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014741767376607538, + "loss": 0.5386, + "step": 12175 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014740918408426542, + "loss": 0.5703, + "step": 12176 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014740069396166377, + "loss": 0.5886, + "step": 12177 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001473922033983494, + "loss": 0.6606, + "step": 12178 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014738371239440113, + "loss": 0.538, + "step": 12179 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014737522094989804, + "loss": 0.6209, + "step": 12180 + }, + { + "epoch": 0.36, + "learning_rate": 0.000147366729064919, + "loss": 0.5133, + "step": 12181 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014735823673954303, + "loss": 0.5876, + "step": 12182 + }, + { + "epoch": 0.36, + "learning_rate": 0.000147349743973849, + "loss": 0.5453, + "step": 12183 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014734125076791598, + "loss": 0.5703, + "step": 12184 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014733275712182283, + "loss": 0.6601, + "step": 12185 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001473242630356486, + "loss": 0.5394, + "step": 12186 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014731576850947223, + "loss": 0.5065, + "step": 12187 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001473072735433727, + "loss": 0.4702, + "step": 12188 + }, + { + "epoch": 0.36, + "learning_rate": 0.000147298778137429, + "loss": 0.6085, + "step": 12189 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014729028229172012, + "loss": 0.5838, + "step": 12190 + }, + { + "epoch": 0.36, + "learning_rate": 0.000147281786006325, + "loss": 0.4367, + "step": 12191 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001472732892813227, + "loss": 0.5263, + "step": 12192 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014726479211679222, + "loss": 0.6299, + "step": 12193 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001472562945128125, + "loss": 0.5377, + "step": 12194 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014724779646946263, + "loss": 0.6089, + "step": 12195 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014723929798682157, + "loss": 0.5255, + "step": 12196 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014723079906496834, + "loss": 0.5808, + "step": 12197 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014722229970398195, + "loss": 0.5049, + "step": 12198 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014721379990394147, + "loss": 0.6522, + "step": 12199 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014720529966492585, + "loss": 0.4456, + "step": 12200 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001471967989870142, + "loss": 0.6891, + "step": 12201 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001471882978702855, + "loss": 0.593, + "step": 12202 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014717979631481888, + "loss": 0.6037, + "step": 12203 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014717129432069323, + "loss": 0.634, + "step": 12204 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014716279188798774, + "loss": 0.5462, + "step": 12205 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014715428901678138, + "loss": 0.5494, + "step": 12206 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014714578570715326, + "loss": 0.6218, + "step": 12207 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014713728195918239, + "loss": 0.6053, + "step": 12208 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014712877777294789, + "loss": 0.462, + "step": 12209 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001471202731485288, + "loss": 0.5379, + "step": 12210 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014711176808600415, + "loss": 0.473, + "step": 12211 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014710326258545307, + "loss": 0.5431, + "step": 12212 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014709475664695466, + "loss": 0.5291, + "step": 12213 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001470862502705879, + "loss": 0.507, + "step": 12214 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014707774345643204, + "loss": 0.5985, + "step": 12215 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014706923620456602, + "loss": 0.563, + "step": 12216 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014706072851506904, + "loss": 0.5883, + "step": 12217 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014705222038802014, + "loss": 0.5878, + "step": 12218 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014704371182349844, + "loss": 0.5796, + "step": 12219 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014703520282158313, + "loss": 0.7033, + "step": 12220 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001470266933823532, + "loss": 0.5753, + "step": 12221 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001470181835058878, + "loss": 0.5258, + "step": 12222 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014700967319226607, + "loss": 0.453, + "step": 12223 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014700116244156715, + "loss": 0.5663, + "step": 12224 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014699265125387015, + "loss": 0.5577, + "step": 12225 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014698413962925422, + "loss": 0.4583, + "step": 12226 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001469756275677985, + "loss": 0.4681, + "step": 12227 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014696711506958208, + "loss": 0.6116, + "step": 12228 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014695860213468414, + "loss": 0.6214, + "step": 12229 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014695008876318386, + "loss": 0.6114, + "step": 12230 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014694157495516032, + "loss": 0.5274, + "step": 12231 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001469330607106928, + "loss": 0.5568, + "step": 12232 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014692454602986032, + "loss": 0.5668, + "step": 12233 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014691603091274217, + "loss": 0.659, + "step": 12234 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014690751535941744, + "loss": 0.4866, + "step": 12235 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001468989993699653, + "loss": 0.5757, + "step": 12236 + }, + { + "epoch": 0.36, + "learning_rate": 0.000146890482944465, + "loss": 0.5705, + "step": 12237 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014688196608299566, + "loss": 0.6117, + "step": 12238 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014687344878563648, + "loss": 0.5356, + "step": 12239 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014686493105246665, + "loss": 0.6031, + "step": 12240 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014685641288356537, + "loss": 0.5548, + "step": 12241 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014684789427901187, + "loss": 0.6141, + "step": 12242 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014683937523888527, + "loss": 0.4956, + "step": 12243 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014683085576326489, + "loss": 0.4357, + "step": 12244 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001468223358522298, + "loss": 0.4878, + "step": 12245 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014681381550585935, + "loss": 0.5493, + "step": 12246 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001468052947242327, + "loss": 0.6611, + "step": 12247 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014679677350742904, + "loss": 0.5051, + "step": 12248 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014678825185552764, + "loss": 0.6222, + "step": 12249 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014677972976860775, + "loss": 0.5904, + "step": 12250 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014677120724674857, + "loss": 0.4831, + "step": 12251 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014676268429002933, + "loss": 0.6003, + "step": 12252 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014675416089852928, + "loss": 0.6119, + "step": 12253 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014674563707232772, + "loss": 0.516, + "step": 12254 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014673711281150382, + "loss": 0.4503, + "step": 12255 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014672858811613687, + "loss": 0.4795, + "step": 12256 + }, + { + "epoch": 0.36, + "learning_rate": 0.00014672006298630617, + "loss": 0.5806, + "step": 12257 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001467115374220909, + "loss": 0.6435, + "step": 12258 + }, + { + "epoch": 0.36, + "learning_rate": 0.0001467030114235704, + "loss": 0.6041, + "step": 12259 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001466944849908239, + "loss": 0.5073, + "step": 12260 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014668595812393072, + "loss": 0.61, + "step": 12261 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014667743082297005, + "loss": 0.5363, + "step": 12262 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001466689030880213, + "loss": 0.5728, + "step": 12263 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014666037491916365, + "loss": 0.5154, + "step": 12264 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014665184631647643, + "loss": 0.5106, + "step": 12265 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014664331728003892, + "loss": 0.5598, + "step": 12266 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014663478780993047, + "loss": 0.5834, + "step": 12267 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014662625790623035, + "loss": 0.4992, + "step": 12268 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014661772756901785, + "loss": 0.6098, + "step": 12269 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001466091967983723, + "loss": 0.5891, + "step": 12270 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014660066559437303, + "loss": 0.5279, + "step": 12271 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014659213395709934, + "loss": 0.6048, + "step": 12272 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014658360188663053, + "loss": 0.5395, + "step": 12273 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014657506938304597, + "loss": 0.5759, + "step": 12274 + }, + { + "epoch": 0.37, + "learning_rate": 0.000146566536446425, + "loss": 0.5778, + "step": 12275 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014655800307684694, + "loss": 0.5326, + "step": 12276 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001465494692743911, + "loss": 0.5604, + "step": 12277 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014654093503913685, + "loss": 0.5771, + "step": 12278 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014653240037116355, + "loss": 0.4869, + "step": 12279 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014652386527055052, + "loss": 0.5741, + "step": 12280 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014651532973737713, + "loss": 0.5854, + "step": 12281 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014650679377172275, + "loss": 0.5771, + "step": 12282 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014649825737366674, + "loss": 0.5389, + "step": 12283 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014648972054328847, + "loss": 0.6357, + "step": 12284 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001464811832806673, + "loss": 0.5162, + "step": 12285 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001464726455858826, + "loss": 0.5227, + "step": 12286 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001464641074590138, + "loss": 0.525, + "step": 12287 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001464555689001402, + "loss": 0.4806, + "step": 12288 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014644702990934127, + "loss": 0.5431, + "step": 12289 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014643849048669635, + "loss": 0.5568, + "step": 12290 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014642995063228486, + "loss": 0.5552, + "step": 12291 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014642141034618622, + "loss": 0.5718, + "step": 12292 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014641286962847975, + "loss": 0.4951, + "step": 12293 + }, + { + "epoch": 0.37, + "learning_rate": 0.000146404328479245, + "loss": 0.529, + "step": 12294 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014639578689856121, + "loss": 0.5486, + "step": 12295 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014638724488650796, + "loss": 0.5452, + "step": 12296 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014637870244316455, + "loss": 0.4404, + "step": 12297 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014637015956861044, + "loss": 0.5529, + "step": 12298 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001463616162629251, + "loss": 0.486, + "step": 12299 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014635307252618794, + "loss": 0.5239, + "step": 12300 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014634452835847833, + "loss": 0.481, + "step": 12301 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014633598375987582, + "loss": 0.6316, + "step": 12302 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014632743873045978, + "loss": 0.6846, + "step": 12303 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014631889327030968, + "loss": 0.5871, + "step": 12304 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014631034737950496, + "loss": 0.5687, + "step": 12305 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014630180105812513, + "loss": 0.509, + "step": 12306 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014629325430624954, + "loss": 0.4933, + "step": 12307 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014628470712395777, + "loss": 0.5138, + "step": 12308 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014627615951132923, + "loss": 0.4895, + "step": 12309 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001462676114684434, + "loss": 0.7628, + "step": 12310 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014625906299537976, + "loss": 0.5823, + "step": 12311 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001462505140922178, + "loss": 0.4985, + "step": 12312 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014624196475903698, + "loss": 0.6388, + "step": 12313 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014623341499591682, + "loss": 0.5688, + "step": 12314 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014622486480293678, + "loss": 0.5791, + "step": 12315 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014621631418017635, + "loss": 0.5267, + "step": 12316 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014620776312771508, + "loss": 0.5881, + "step": 12317 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014619921164563243, + "loss": 0.669, + "step": 12318 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014619065973400794, + "loss": 0.602, + "step": 12319 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001461821073929211, + "loss": 0.5704, + "step": 12320 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001461735546224514, + "loss": 0.6268, + "step": 12321 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014616500142267843, + "loss": 0.5443, + "step": 12322 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001461564477936817, + "loss": 0.7198, + "step": 12323 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014614789373554066, + "loss": 0.5404, + "step": 12324 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014613933924833493, + "loss": 0.5808, + "step": 12325 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014613078433214398, + "loss": 0.5661, + "step": 12326 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001461222289870474, + "loss": 0.5856, + "step": 12327 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014611367321312472, + "loss": 0.6651, + "step": 12328 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001461051170104555, + "loss": 0.5889, + "step": 12329 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014609656037911924, + "loss": 0.5174, + "step": 12330 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014608800331919556, + "loss": 0.5905, + "step": 12331 + }, + { + "epoch": 0.37, + "learning_rate": 0.000146079445830764, + "loss": 0.5526, + "step": 12332 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014607088791390412, + "loss": 0.5705, + "step": 12333 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014606232956869547, + "loss": 0.4812, + "step": 12334 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014605377079521766, + "loss": 0.4619, + "step": 12335 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001460452115935502, + "loss": 0.5858, + "step": 12336 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014603665196377276, + "loss": 0.5658, + "step": 12337 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014602809190596486, + "loss": 0.4756, + "step": 12338 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001460195314202061, + "loss": 0.4935, + "step": 12339 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014601097050657608, + "loss": 0.5302, + "step": 12340 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001460024091651544, + "loss": 0.5674, + "step": 12341 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014599384739602067, + "loss": 0.5148, + "step": 12342 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014598528519925446, + "loss": 0.5407, + "step": 12343 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001459767225749354, + "loss": 0.632, + "step": 12344 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014596815952314308, + "loss": 0.5219, + "step": 12345 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014595959604395717, + "loss": 0.5949, + "step": 12346 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014595103213745723, + "loss": 0.575, + "step": 12347 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001459424678037229, + "loss": 0.5938, + "step": 12348 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014593390304283384, + "loss": 0.6719, + "step": 12349 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014592533785486963, + "loss": 0.5131, + "step": 12350 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014591677223990996, + "loss": 0.5852, + "step": 12351 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014590820619803444, + "loss": 0.5149, + "step": 12352 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001458996397293227, + "loss": 0.5597, + "step": 12353 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014589107283385442, + "loss": 0.6808, + "step": 12354 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014588250551170922, + "loss": 0.5081, + "step": 12355 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014587393776296678, + "loss": 0.5955, + "step": 12356 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014586536958770673, + "loss": 0.5587, + "step": 12357 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014585680098600876, + "loss": 0.5501, + "step": 12358 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014584823195795256, + "loss": 0.5231, + "step": 12359 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014583966250361776, + "loss": 0.5622, + "step": 12360 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014583109262308403, + "loss": 0.6098, + "step": 12361 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001458225223164311, + "loss": 0.6231, + "step": 12362 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014581395158373857, + "loss": 0.6287, + "step": 12363 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001458053804250862, + "loss": 0.5755, + "step": 12364 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014579680884055368, + "loss": 0.5741, + "step": 12365 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014578823683022066, + "loss": 0.5006, + "step": 12366 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014577966439416687, + "loss": 0.594, + "step": 12367 + }, + { + "epoch": 0.37, + "learning_rate": 0.000145771091532472, + "loss": 0.5737, + "step": 12368 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014576251824521578, + "loss": 0.5897, + "step": 12369 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014575394453247793, + "loss": 0.5844, + "step": 12370 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014574537039433805, + "loss": 0.4549, + "step": 12371 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014573679583087601, + "loss": 0.4648, + "step": 12372 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014572822084217148, + "loss": 0.4676, + "step": 12373 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014571964542830416, + "loss": 0.5744, + "step": 12374 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001457110695893538, + "loss": 0.5533, + "step": 12375 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014570249332540017, + "loss": 0.5662, + "step": 12376 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014569391663652294, + "loss": 0.633, + "step": 12377 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014568533952280188, + "loss": 0.6511, + "step": 12378 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014567676198431673, + "loss": 0.5389, + "step": 12379 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014566818402114728, + "loss": 0.6892, + "step": 12380 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014565960563337323, + "loss": 0.4741, + "step": 12381 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014565102682107442, + "loss": 0.6548, + "step": 12382 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014564244758433046, + "loss": 0.5762, + "step": 12383 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001456338679232213, + "loss": 0.7214, + "step": 12384 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001456252878378266, + "loss": 0.6286, + "step": 12385 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014561670732822615, + "loss": 0.6469, + "step": 12386 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014560812639449974, + "loss": 0.558, + "step": 12387 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014559954503672712, + "loss": 0.4859, + "step": 12388 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014559096325498816, + "loss": 0.524, + "step": 12389 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014558238104936256, + "loss": 0.4906, + "step": 12390 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014557379841993016, + "loss": 0.5156, + "step": 12391 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014556521536677073, + "loss": 0.4351, + "step": 12392 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001455566318899641, + "loss": 0.577, + "step": 12393 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014554804798959004, + "loss": 0.5235, + "step": 12394 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014553946366572843, + "loss": 0.5394, + "step": 12395 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014553087891845903, + "loss": 0.6359, + "step": 12396 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014552229374786163, + "loss": 0.4506, + "step": 12397 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014551370815401609, + "loss": 0.5282, + "step": 12398 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014550512213700224, + "loss": 0.7092, + "step": 12399 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001454965356968999, + "loss": 0.6097, + "step": 12400 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014548794883378893, + "loss": 0.586, + "step": 12401 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001454793615477491, + "loss": 0.5319, + "step": 12402 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014547077383886032, + "loss": 0.5471, + "step": 12403 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001454621857072024, + "loss": 0.5156, + "step": 12404 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001454535971528552, + "loss": 0.5844, + "step": 12405 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014544500817589858, + "loss": 0.5714, + "step": 12406 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014543641877641237, + "loss": 0.4839, + "step": 12407 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014542782895447646, + "loss": 0.6732, + "step": 12408 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001454192387101707, + "loss": 0.4728, + "step": 12409 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014541064804357494, + "loss": 0.551, + "step": 12410 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001454020569547691, + "loss": 0.6013, + "step": 12411 + }, + { + "epoch": 0.37, + "learning_rate": 0.000145393465443833, + "loss": 0.6342, + "step": 12412 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014538487351084658, + "loss": 0.5428, + "step": 12413 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014537628115588968, + "loss": 0.5688, + "step": 12414 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001453676883790422, + "loss": 0.5066, + "step": 12415 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014535909518038404, + "loss": 0.5701, + "step": 12416 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001453505015599951, + "loss": 0.7146, + "step": 12417 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001453419075179553, + "loss": 0.4879, + "step": 12418 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014533331305434442, + "loss": 0.4602, + "step": 12419 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014532471816924256, + "loss": 0.5641, + "step": 12420 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014531612286272947, + "loss": 0.5132, + "step": 12421 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014530752713488518, + "loss": 0.6283, + "step": 12422 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014529893098578955, + "loss": 0.5308, + "step": 12423 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001452903344155225, + "loss": 0.6671, + "step": 12424 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014528173742416398, + "loss": 0.5862, + "step": 12425 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001452731400117939, + "loss": 0.5437, + "step": 12426 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014526454217849222, + "loss": 0.5101, + "step": 12427 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014525594392433887, + "loss": 0.5974, + "step": 12428 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014524734524941376, + "loss": 0.5256, + "step": 12429 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001452387461537969, + "loss": 0.5276, + "step": 12430 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001452301466375682, + "loss": 0.5879, + "step": 12431 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001452215467008076, + "loss": 0.498, + "step": 12432 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014521294634359513, + "loss": 0.5212, + "step": 12433 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001452043455660107, + "loss": 0.6191, + "step": 12434 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014519574436813425, + "loss": 0.5674, + "step": 12435 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001451871427500458, + "loss": 0.6139, + "step": 12436 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014517854071182532, + "loss": 0.5143, + "step": 12437 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014516993825355274, + "loss": 0.5365, + "step": 12438 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001451613353753081, + "loss": 0.6071, + "step": 12439 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014515273207717135, + "loss": 0.6762, + "step": 12440 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001451441283592225, + "loss": 0.6014, + "step": 12441 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014513552422154154, + "loss": 0.5931, + "step": 12442 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014512691966420843, + "loss": 0.5064, + "step": 12443 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014511831468730328, + "loss": 0.5721, + "step": 12444 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014510970929090594, + "loss": 0.454, + "step": 12445 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014510110347509657, + "loss": 0.6074, + "step": 12446 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001450924972399551, + "loss": 0.5673, + "step": 12447 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014508389058556152, + "loss": 0.5267, + "step": 12448 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014507528351199596, + "loss": 0.6542, + "step": 12449 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014506667601933834, + "loss": 0.6228, + "step": 12450 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014505806810766872, + "loss": 0.4524, + "step": 12451 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014504945977706715, + "loss": 0.5257, + "step": 12452 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014504085102761368, + "loss": 0.5107, + "step": 12453 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001450322418593883, + "loss": 0.5383, + "step": 12454 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001450236322724711, + "loss": 0.6728, + "step": 12455 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001450150222669421, + "loss": 0.5466, + "step": 12456 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014500641184288137, + "loss": 0.6475, + "step": 12457 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014499780100036896, + "loss": 0.5481, + "step": 12458 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014498918973948494, + "loss": 0.6136, + "step": 12459 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014498057806030935, + "loss": 0.5827, + "step": 12460 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014497196596292229, + "loss": 0.6348, + "step": 12461 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001449633534474038, + "loss": 0.599, + "step": 12462 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014495474051383396, + "loss": 0.4047, + "step": 12463 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014494612716229287, + "loss": 0.5686, + "step": 12464 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014493751339286062, + "loss": 0.6843, + "step": 12465 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014492889920561725, + "loss": 0.5866, + "step": 12466 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001449202846006429, + "loss": 0.5484, + "step": 12467 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014491166957801766, + "loss": 0.5329, + "step": 12468 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001449030541378216, + "loss": 0.5055, + "step": 12469 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014489443828013485, + "loss": 0.4621, + "step": 12470 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014488582200503748, + "loss": 0.5903, + "step": 12471 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014487720531260965, + "loss": 0.5874, + "step": 12472 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014486858820293147, + "loss": 0.539, + "step": 12473 + }, + { + "epoch": 0.37, + "learning_rate": 0.000144859970676083, + "loss": 0.6839, + "step": 12474 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014485135273214442, + "loss": 0.6102, + "step": 12475 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014484273437119583, + "loss": 0.6145, + "step": 12476 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014483411559331738, + "loss": 0.5684, + "step": 12477 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014482549639858918, + "loss": 0.47, + "step": 12478 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001448168767870914, + "loss": 0.4657, + "step": 12479 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014480825675890414, + "loss": 0.4774, + "step": 12480 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014479963631410759, + "loss": 0.5341, + "step": 12481 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014479101545278184, + "loss": 0.5078, + "step": 12482 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014478239417500713, + "loss": 0.5686, + "step": 12483 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014477377248086354, + "loss": 0.6546, + "step": 12484 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014476515037043126, + "loss": 0.5519, + "step": 12485 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014475652784379044, + "loss": 0.5646, + "step": 12486 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014474790490102128, + "loss": 0.5852, + "step": 12487 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014473928154220392, + "loss": 0.5788, + "step": 12488 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014473065776741855, + "loss": 0.6391, + "step": 12489 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001447220335767454, + "loss": 0.5494, + "step": 12490 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014471340897026456, + "loss": 0.4646, + "step": 12491 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001447047839480563, + "loss": 0.5526, + "step": 12492 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014469615851020072, + "loss": 0.559, + "step": 12493 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001446875326567781, + "loss": 0.4803, + "step": 12494 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014467890638786862, + "loss": 0.5242, + "step": 12495 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014467027970355246, + "loss": 0.4839, + "step": 12496 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014466165260390988, + "loss": 0.6726, + "step": 12497 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014465302508902102, + "loss": 0.5313, + "step": 12498 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014464439715896615, + "loss": 0.6612, + "step": 12499 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014463576881382543, + "loss": 0.6422, + "step": 12500 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014462714005367917, + "loss": 0.5154, + "step": 12501 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014461851087860752, + "loss": 0.7325, + "step": 12502 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014460988128869073, + "loss": 0.5613, + "step": 12503 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014460125128400907, + "loss": 0.6154, + "step": 12504 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014459262086464274, + "loss": 0.5665, + "step": 12505 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014458399003067202, + "loss": 0.4834, + "step": 12506 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014457535878217707, + "loss": 0.6293, + "step": 12507 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014456672711923824, + "loss": 0.4644, + "step": 12508 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014455809504193575, + "loss": 0.5274, + "step": 12509 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014454946255034986, + "loss": 0.5338, + "step": 12510 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014454082964456078, + "loss": 0.5505, + "step": 12511 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014453219632464884, + "loss": 0.5443, + "step": 12512 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001445235625906943, + "loss": 0.5126, + "step": 12513 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001445149284427774, + "loss": 0.5432, + "step": 12514 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014450629388097847, + "loss": 0.5054, + "step": 12515 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001444976589053777, + "loss": 0.5801, + "step": 12516 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014448902351605548, + "loss": 0.563, + "step": 12517 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014448038771309205, + "loss": 0.6421, + "step": 12518 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001444717514965677, + "loss": 0.5362, + "step": 12519 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001444631148665627, + "loss": 0.5518, + "step": 12520 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014445447782315742, + "loss": 0.4346, + "step": 12521 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014444584036643209, + "loss": 0.6155, + "step": 12522 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014443720249646708, + "loss": 0.4902, + "step": 12523 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014442856421334263, + "loss": 0.4654, + "step": 12524 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014441992551713915, + "loss": 0.6001, + "step": 12525 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014441128640793686, + "loss": 0.455, + "step": 12526 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014440264688581614, + "loss": 0.6316, + "step": 12527 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014439400695085732, + "loss": 0.5388, + "step": 12528 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014438536660314072, + "loss": 0.4434, + "step": 12529 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014437672584274668, + "loss": 0.5472, + "step": 12530 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001443680846697555, + "loss": 0.5236, + "step": 12531 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001443594430842476, + "loss": 0.7461, + "step": 12532 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014435080108630323, + "loss": 0.6115, + "step": 12533 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014434215867600278, + "loss": 0.5697, + "step": 12534 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014433351585342664, + "loss": 0.5888, + "step": 12535 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014432487261865513, + "loss": 0.5576, + "step": 12536 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014431622897176865, + "loss": 0.6773, + "step": 12537 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014430758491284747, + "loss": 0.4122, + "step": 12538 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014429894044197208, + "loss": 0.4893, + "step": 12539 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014429029555922278, + "loss": 0.6105, + "step": 12540 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014428165026467998, + "loss": 0.5756, + "step": 12541 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014427300455842402, + "loss": 0.5926, + "step": 12542 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014426435844053532, + "loss": 0.4667, + "step": 12543 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014425571191109426, + "loss": 0.4926, + "step": 12544 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014424706497018124, + "loss": 0.5773, + "step": 12545 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001442384176178766, + "loss": 0.5372, + "step": 12546 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014422976985426086, + "loss": 0.6581, + "step": 12547 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014422112167941428, + "loss": 0.5424, + "step": 12548 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014421247309341737, + "loss": 0.6142, + "step": 12549 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001442038240963505, + "loss": 0.5322, + "step": 12550 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014419517468829406, + "loss": 0.5592, + "step": 12551 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014418652486932856, + "loss": 0.5357, + "step": 12552 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014417787463953433, + "loss": 0.4862, + "step": 12553 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014416922399899184, + "loss": 0.4902, + "step": 12554 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014416057294778148, + "loss": 0.4725, + "step": 12555 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014415192148598373, + "loss": 0.4602, + "step": 12556 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014414326961367902, + "loss": 0.5274, + "step": 12557 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014413461733094778, + "loss": 0.5333, + "step": 12558 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014412596463787044, + "loss": 0.6077, + "step": 12559 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014411731153452748, + "loss": 0.5367, + "step": 12560 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014410865802099937, + "loss": 0.529, + "step": 12561 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001441000040973665, + "loss": 0.4244, + "step": 12562 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014409134976370937, + "loss": 0.4415, + "step": 12563 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014408269502010846, + "loss": 0.5607, + "step": 12564 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001440740398666442, + "loss": 0.5116, + "step": 12565 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001440653843033971, + "loss": 0.503, + "step": 12566 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014405672833044762, + "loss": 0.4214, + "step": 12567 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014404807194787626, + "loss": 0.5875, + "step": 12568 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014403941515576344, + "loss": 0.6439, + "step": 12569 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001440307579541897, + "loss": 0.6195, + "step": 12570 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014402210034323554, + "loss": 0.5406, + "step": 12571 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014401344232298143, + "loss": 0.5822, + "step": 12572 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014400478389350787, + "loss": 0.4136, + "step": 12573 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014399612505489536, + "loss": 0.5982, + "step": 12574 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014398746580722443, + "loss": 0.4987, + "step": 12575 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014397880615057558, + "loss": 0.5153, + "step": 12576 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001439701460850293, + "loss": 0.5164, + "step": 12577 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014396148561066614, + "loss": 0.5185, + "step": 12578 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001439528247275666, + "loss": 0.4864, + "step": 12579 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014394416343581123, + "loss": 0.4985, + "step": 12580 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014393550173548054, + "loss": 0.5831, + "step": 12581 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014392683962665502, + "loss": 0.5015, + "step": 12582 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001439181771094153, + "loss": 0.5158, + "step": 12583 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014390951418384186, + "loss": 0.5088, + "step": 12584 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014390085085001528, + "loss": 0.435, + "step": 12585 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014389218710801605, + "loss": 0.4492, + "step": 12586 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014388352295792477, + "loss": 0.5051, + "step": 12587 + }, + { + "epoch": 0.37, + "learning_rate": 0.000143874858399822, + "loss": 0.537, + "step": 12588 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014386619343378825, + "loss": 0.591, + "step": 12589 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014385752805990415, + "loss": 0.4588, + "step": 12590 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001438488622782502, + "loss": 0.5669, + "step": 12591 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014384019608890705, + "loss": 0.6688, + "step": 12592 + }, + { + "epoch": 0.37, + "learning_rate": 0.0001438315294919552, + "loss": 0.4908, + "step": 12593 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014382286248747526, + "loss": 0.5612, + "step": 12594 + }, + { + "epoch": 0.37, + "learning_rate": 0.00014381419507554784, + "loss": 0.6507, + "step": 12595 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014380552725625344, + "loss": 0.4934, + "step": 12596 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014379685902967277, + "loss": 0.5446, + "step": 12597 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014378819039588629, + "loss": 0.4741, + "step": 12598 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014377952135497473, + "loss": 0.5267, + "step": 12599 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001437708519070186, + "loss": 0.604, + "step": 12600 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014376218205209852, + "loss": 0.4926, + "step": 12601 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014375351179029514, + "loss": 0.6818, + "step": 12602 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014374484112168905, + "loss": 0.5883, + "step": 12603 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014373617004636086, + "loss": 0.5835, + "step": 12604 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014372749856439118, + "loss": 0.6441, + "step": 12605 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014371882667586064, + "loss": 0.59, + "step": 12606 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001437101543808499, + "loss": 0.5937, + "step": 12607 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014370148167943954, + "loss": 0.4807, + "step": 12608 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014369280857171024, + "loss": 0.4589, + "step": 12609 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001436841350577426, + "loss": 0.5698, + "step": 12610 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001436754611376173, + "loss": 0.5352, + "step": 12611 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014366678681141493, + "loss": 0.4788, + "step": 12612 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014365811207921625, + "loss": 0.4933, + "step": 12613 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001436494369411018, + "loss": 0.5456, + "step": 12614 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014364076139715228, + "loss": 0.622, + "step": 12615 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001436320854474484, + "loss": 0.501, + "step": 12616 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014362340909207072, + "loss": 0.6469, + "step": 12617 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014361473233109997, + "loss": 0.5612, + "step": 12618 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014360605516461686, + "loss": 0.5305, + "step": 12619 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014359737759270203, + "loss": 0.5663, + "step": 12620 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014358869961543614, + "loss": 0.5876, + "step": 12621 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014358002123289988, + "loss": 0.4615, + "step": 12622 + }, + { + "epoch": 0.38, + "learning_rate": 0.000143571342445174, + "loss": 0.5708, + "step": 12623 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014356266325233908, + "loss": 0.6487, + "step": 12624 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014355398365447592, + "loss": 0.6555, + "step": 12625 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001435453036516652, + "loss": 0.5005, + "step": 12626 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014353662324398752, + "loss": 0.7602, + "step": 12627 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014352794243152375, + "loss": 0.5506, + "step": 12628 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014351926121435447, + "loss": 0.7974, + "step": 12629 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014351057959256046, + "loss": 0.4004, + "step": 12630 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014350189756622244, + "loss": 0.5425, + "step": 12631 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001434932151354211, + "loss": 0.5773, + "step": 12632 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014348453230023718, + "loss": 0.5347, + "step": 12633 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014347584906075143, + "loss": 0.4899, + "step": 12634 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014346716541704455, + "loss": 0.5323, + "step": 12635 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014345848136919728, + "loss": 0.4908, + "step": 12636 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014344979691729038, + "loss": 0.5308, + "step": 12637 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014344111206140457, + "loss": 0.5429, + "step": 12638 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014343242680162063, + "loss": 0.4661, + "step": 12639 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001434237411380193, + "loss": 0.6153, + "step": 12640 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001434150550706813, + "loss": 0.5491, + "step": 12641 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014340636859968748, + "loss": 0.4387, + "step": 12642 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001433976817251185, + "loss": 0.59, + "step": 12643 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014338899444705518, + "loss": 0.6012, + "step": 12644 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001433803067655783, + "loss": 0.4251, + "step": 12645 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014337161868076858, + "loss": 0.4967, + "step": 12646 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001433629301927069, + "loss": 0.6244, + "step": 12647 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001433542413014739, + "loss": 0.4603, + "step": 12648 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014334555200715048, + "loss": 0.6032, + "step": 12649 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014333686230981742, + "loss": 0.5289, + "step": 12650 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014332817220955545, + "loss": 0.639, + "step": 12651 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014331948170644542, + "loss": 0.6038, + "step": 12652 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001433107908005681, + "loss": 0.548, + "step": 12653 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014330209949200433, + "loss": 0.5113, + "step": 12654 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014329340778083488, + "loss": 0.5614, + "step": 12655 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014328471566714057, + "loss": 0.4978, + "step": 12656 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014327602315100226, + "loss": 0.7184, + "step": 12657 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001432673302325007, + "loss": 0.5618, + "step": 12658 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014325863691171678, + "loss": 0.5333, + "step": 12659 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014324994318873127, + "loss": 0.4791, + "step": 12660 + }, + { + "epoch": 0.38, + "learning_rate": 0.000143241249063625, + "loss": 0.473, + "step": 12661 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001432325545364789, + "loss": 0.5523, + "step": 12662 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001432238596073737, + "loss": 0.4897, + "step": 12663 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001432151642763903, + "loss": 0.5715, + "step": 12664 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001432064685436095, + "loss": 0.5741, + "step": 12665 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001431977724091122, + "loss": 0.5616, + "step": 12666 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001431890758729792, + "loss": 0.5847, + "step": 12667 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014318037893529142, + "loss": 0.5379, + "step": 12668 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014317168159612965, + "loss": 0.5959, + "step": 12669 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014316298385557482, + "loss": 0.4947, + "step": 12670 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014315428571370776, + "loss": 0.4923, + "step": 12671 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014314558717060932, + "loss": 0.623, + "step": 12672 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014313688822636047, + "loss": 0.4863, + "step": 12673 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014312818888104198, + "loss": 0.6059, + "step": 12674 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001431194891347348, + "loss": 0.5907, + "step": 12675 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014311078898751977, + "loss": 0.593, + "step": 12676 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014310208843947781, + "loss": 0.633, + "step": 12677 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014309338749068982, + "loss": 0.629, + "step": 12678 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001430846861412367, + "loss": 0.6048, + "step": 12679 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014307598439119933, + "loss": 0.4756, + "step": 12680 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014306728224065864, + "loss": 0.5039, + "step": 12681 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014305857968969552, + "loss": 0.547, + "step": 12682 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014304987673839086, + "loss": 0.5909, + "step": 12683 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014304117338682562, + "loss": 0.5064, + "step": 12684 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014303246963508074, + "loss": 0.6402, + "step": 12685 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014302376548323706, + "loss": 0.6424, + "step": 12686 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014301506093137559, + "loss": 0.482, + "step": 12687 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014300635597957722, + "loss": 0.588, + "step": 12688 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014299765062792286, + "loss": 0.5605, + "step": 12689 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014298894487649351, + "loss": 0.571, + "step": 12690 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014298023872537008, + "loss": 0.5266, + "step": 12691 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001429715321746335, + "loss": 0.414, + "step": 12692 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014296282522436476, + "loss": 0.4737, + "step": 12693 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014295411787464482, + "loss": 0.6085, + "step": 12694 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001429454101255546, + "loss": 0.6253, + "step": 12695 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014293670197717504, + "loss": 0.558, + "step": 12696 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014292799342958715, + "loss": 0.56, + "step": 12697 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014291928448287192, + "loss": 0.5746, + "step": 12698 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014291057513711024, + "loss": 0.6827, + "step": 12699 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014290186539238318, + "loss": 0.4857, + "step": 12700 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001428931552487716, + "loss": 0.5515, + "step": 12701 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014288444470635664, + "loss": 0.6188, + "step": 12702 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014287573376521918, + "loss": 0.4868, + "step": 12703 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014286702242544022, + "loss": 0.543, + "step": 12704 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001428583106871008, + "loss": 0.4951, + "step": 12705 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014284959855028184, + "loss": 0.535, + "step": 12706 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001428408860150644, + "loss": 0.6313, + "step": 12707 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014283217308152954, + "loss": 0.6742, + "step": 12708 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014282345974975816, + "loss": 0.5261, + "step": 12709 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001428147460198313, + "loss": 0.5493, + "step": 12710 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014280603189183004, + "loss": 0.4927, + "step": 12711 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014279731736583533, + "loss": 0.4976, + "step": 12712 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014278860244192821, + "loss": 0.5985, + "step": 12713 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014277988712018975, + "loss": 0.6736, + "step": 12714 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014277117140070096, + "loss": 0.5836, + "step": 12715 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014276245528354283, + "loss": 0.5969, + "step": 12716 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014275373876879646, + "loss": 0.5879, + "step": 12717 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001427450218565428, + "loss": 0.61, + "step": 12718 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014273630454686304, + "loss": 0.5517, + "step": 12719 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001427275868398381, + "loss": 0.5236, + "step": 12720 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014271886873554914, + "loss": 0.5582, + "step": 12721 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014271015023407713, + "loss": 0.5137, + "step": 12722 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014270143133550316, + "loss": 0.56, + "step": 12723 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014269271203990833, + "loss": 0.4751, + "step": 12724 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014268399234737366, + "loss": 0.5158, + "step": 12725 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014267527225798025, + "loss": 0.5363, + "step": 12726 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014266655177180916, + "loss": 0.5945, + "step": 12727 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001426578308889415, + "loss": 0.4416, + "step": 12728 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001426491096094583, + "loss": 0.5907, + "step": 12729 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014264038793344066, + "loss": 0.5767, + "step": 12730 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014263166586096974, + "loss": 0.6779, + "step": 12731 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014262294339212653, + "loss": 0.5486, + "step": 12732 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014261422052699223, + "loss": 0.6698, + "step": 12733 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014260549726564784, + "loss": 0.5903, + "step": 12734 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014259677360817458, + "loss": 0.4981, + "step": 12735 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014258804955465343, + "loss": 0.6417, + "step": 12736 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001425793251051656, + "loss": 0.5493, + "step": 12737 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001425706002597922, + "loss": 0.6115, + "step": 12738 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001425618750186143, + "loss": 0.5114, + "step": 12739 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014255314938171304, + "loss": 0.4724, + "step": 12740 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014254442334916959, + "loss": 0.5885, + "step": 12741 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014253569692106502, + "loss": 0.5548, + "step": 12742 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001425269700974805, + "loss": 0.5392, + "step": 12743 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014251824287849716, + "loss": 0.61, + "step": 12744 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014250951526419617, + "loss": 0.5064, + "step": 12745 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014250078725465862, + "loss": 0.645, + "step": 12746 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014249205884996568, + "loss": 0.5102, + "step": 12747 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014248333005019854, + "loss": 0.4981, + "step": 12748 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014247460085543829, + "loss": 0.4777, + "step": 12749 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014246587126576617, + "loss": 0.5838, + "step": 12750 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001424571412812633, + "loss": 0.5693, + "step": 12751 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014244841090201084, + "loss": 0.5864, + "step": 12752 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014243968012808998, + "loss": 0.6042, + "step": 12753 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014243094895958187, + "loss": 0.5147, + "step": 12754 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014242221739656775, + "loss": 0.644, + "step": 12755 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001424134854391287, + "loss": 0.4934, + "step": 12756 + }, + { + "epoch": 0.38, + "learning_rate": 0.000142404753087346, + "loss": 0.6131, + "step": 12757 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001423960203413008, + "loss": 0.4794, + "step": 12758 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014238728720107431, + "loss": 0.5906, + "step": 12759 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014237855366674773, + "loss": 0.4818, + "step": 12760 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014236981973840223, + "loss": 0.5188, + "step": 12761 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014236108541611906, + "loss": 0.56, + "step": 12762 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014235235069997933, + "loss": 0.6616, + "step": 12763 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014234361559006443, + "loss": 0.5443, + "step": 12764 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014233488008645536, + "loss": 0.5782, + "step": 12765 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014232614418923352, + "loss": 0.5388, + "step": 12766 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014231740789848003, + "loss": 0.577, + "step": 12767 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014230867121427617, + "loss": 0.5357, + "step": 12768 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014229993413670313, + "loss": 0.5267, + "step": 12769 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014229119666584215, + "loss": 0.584, + "step": 12770 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014228245880177452, + "loss": 0.465, + "step": 12771 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014227372054458138, + "loss": 0.4725, + "step": 12772 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001422649818943441, + "loss": 0.5246, + "step": 12773 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014225624285114382, + "loss": 0.6564, + "step": 12774 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014224750341506183, + "loss": 0.5258, + "step": 12775 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014223876358617944, + "loss": 0.5751, + "step": 12776 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001422300233645778, + "loss": 0.5947, + "step": 12777 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014222128275033827, + "loss": 0.4864, + "step": 12778 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001422125417435421, + "loss": 0.5723, + "step": 12779 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001422038003442705, + "loss": 0.524, + "step": 12780 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001421950585526048, + "loss": 0.409, + "step": 12781 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014218631636862627, + "loss": 0.5454, + "step": 12782 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001421775737924162, + "loss": 0.6276, + "step": 12783 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014216883082405582, + "loss": 0.5013, + "step": 12784 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001421600874636265, + "loss": 0.5671, + "step": 12785 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014215134371120945, + "loss": 0.492, + "step": 12786 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014214259956688602, + "loss": 0.5696, + "step": 12787 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014213385503073747, + "loss": 0.5285, + "step": 12788 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014212511010284516, + "loss": 0.5611, + "step": 12789 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014211636478329037, + "loss": 0.4794, + "step": 12790 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014210761907215442, + "loss": 0.4873, + "step": 12791 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014209887296951857, + "loss": 0.5238, + "step": 12792 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001420901264754642, + "loss": 0.578, + "step": 12793 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001420813795900726, + "loss": 0.6757, + "step": 12794 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014207263231342515, + "loss": 0.5067, + "step": 12795 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014206388464560305, + "loss": 0.5223, + "step": 12796 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014205513658668777, + "loss": 0.5635, + "step": 12797 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014204638813676058, + "loss": 0.544, + "step": 12798 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014203763929590283, + "loss": 0.578, + "step": 12799 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014202889006419587, + "loss": 0.5531, + "step": 12800 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014202014044172105, + "loss": 0.5798, + "step": 12801 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001420113904285597, + "loss": 0.6966, + "step": 12802 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001420026400247932, + "loss": 0.5102, + "step": 12803 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014199388923050285, + "loss": 0.5013, + "step": 12804 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001419851380457701, + "loss": 0.5907, + "step": 12805 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014197638647067625, + "loss": 0.5636, + "step": 12806 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014196763450530271, + "loss": 0.4933, + "step": 12807 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014195888214973076, + "loss": 0.4988, + "step": 12808 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014195012940404193, + "loss": 0.5951, + "step": 12809 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014194137626831748, + "loss": 0.5549, + "step": 12810 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014193262274263884, + "loss": 0.6625, + "step": 12811 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001419238688270874, + "loss": 0.4722, + "step": 12812 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014191511452174452, + "loss": 0.478, + "step": 12813 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001419063598266916, + "loss": 0.5639, + "step": 12814 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014189760474201006, + "loss": 0.6112, + "step": 12815 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014188884926778127, + "loss": 0.5308, + "step": 12816 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001418800934040867, + "loss": 0.5822, + "step": 12817 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001418713371510077, + "loss": 0.5119, + "step": 12818 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014186258050862566, + "loss": 0.7035, + "step": 12819 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014185382347702206, + "loss": 0.6619, + "step": 12820 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014184506605627827, + "loss": 0.517, + "step": 12821 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014183630824647576, + "loss": 0.4557, + "step": 12822 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001418275500476959, + "loss": 0.6062, + "step": 12823 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001418187914600202, + "loss": 0.582, + "step": 12824 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014181003248353, + "loss": 0.4988, + "step": 12825 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001418012731183068, + "loss": 0.4961, + "step": 12826 + }, + { + "epoch": 0.38, + "learning_rate": 0.000141792513364432, + "loss": 0.4382, + "step": 12827 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001417837532219871, + "loss": 0.4783, + "step": 12828 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001417749926910535, + "loss": 0.4379, + "step": 12829 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014176623177171267, + "loss": 0.4932, + "step": 12830 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014175747046404606, + "loss": 0.5178, + "step": 12831 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014174870876813516, + "loss": 0.5358, + "step": 12832 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014173994668406138, + "loss": 0.6544, + "step": 12833 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014173118421190617, + "loss": 0.671, + "step": 12834 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014172242135175112, + "loss": 0.5935, + "step": 12835 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014171365810367758, + "loss": 0.5711, + "step": 12836 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014170489446776706, + "loss": 0.5099, + "step": 12837 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001416961304441011, + "loss": 0.5388, + "step": 12838 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014168736603276108, + "loss": 0.5049, + "step": 12839 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001416786012338286, + "loss": 0.4831, + "step": 12840 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014166983604738505, + "loss": 0.4545, + "step": 12841 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014166107047351202, + "loss": 0.507, + "step": 12842 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014165230451229096, + "loss": 0.5286, + "step": 12843 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014164353816380334, + "loss": 0.6742, + "step": 12844 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014163477142813073, + "loss": 0.5815, + "step": 12845 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014162600430535458, + "loss": 0.5267, + "step": 12846 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014161723679555644, + "loss": 0.4605, + "step": 12847 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014160846889881782, + "loss": 0.5742, + "step": 12848 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014159970061522023, + "loss": 0.5591, + "step": 12849 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014159093194484522, + "loss": 0.5322, + "step": 12850 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001415821628877743, + "loss": 0.6102, + "step": 12851 + }, + { + "epoch": 0.38, + "learning_rate": 0.000141573393444089, + "loss": 0.513, + "step": 12852 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014156462361387086, + "loss": 0.5447, + "step": 12853 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001415558533972014, + "loss": 0.5443, + "step": 12854 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001415470827941622, + "loss": 0.5723, + "step": 12855 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014153831180483477, + "loss": 0.586, + "step": 12856 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001415295404293007, + "loss": 0.4067, + "step": 12857 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014152076866764146, + "loss": 0.6066, + "step": 12858 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001415119965199387, + "loss": 0.4642, + "step": 12859 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014150322398627394, + "loss": 0.5011, + "step": 12860 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014149445106672874, + "loss": 0.5294, + "step": 12861 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014148567776138467, + "loss": 0.5605, + "step": 12862 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001414769040703233, + "loss": 0.6151, + "step": 12863 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014146812999362622, + "loss": 0.5915, + "step": 12864 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014145935553137496, + "loss": 0.5284, + "step": 12865 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001414505806836512, + "loss": 0.6403, + "step": 12866 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014144180545053642, + "loss": 0.4473, + "step": 12867 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014143302983211223, + "loss": 0.5562, + "step": 12868 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001414242538284603, + "loss": 0.5862, + "step": 12869 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001414154774396621, + "loss": 0.4936, + "step": 12870 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014140670066579934, + "loss": 0.5399, + "step": 12871 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014139792350695356, + "loss": 0.6003, + "step": 12872 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001413891459632064, + "loss": 0.4567, + "step": 12873 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014138036803463944, + "loss": 0.5915, + "step": 12874 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014137158972133435, + "loss": 0.4258, + "step": 12875 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014136281102337267, + "loss": 0.5445, + "step": 12876 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001413540319408361, + "loss": 0.4895, + "step": 12877 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014134525247380618, + "loss": 0.5184, + "step": 12878 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001413364726223646, + "loss": 0.5972, + "step": 12879 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014132769238659298, + "loss": 0.4933, + "step": 12880 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014131891176657292, + "loss": 0.5491, + "step": 12881 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001413101307623861, + "loss": 0.5858, + "step": 12882 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014130134937411417, + "loss": 0.466, + "step": 12883 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014129256760183874, + "loss": 0.4552, + "step": 12884 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014128378544564144, + "loss": 0.5384, + "step": 12885 + }, + { + "epoch": 0.38, + "learning_rate": 0.000141275002905604, + "loss": 0.6193, + "step": 12886 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014126621998180802, + "loss": 0.5903, + "step": 12887 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001412574366743352, + "loss": 0.5288, + "step": 12888 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014124865298326713, + "loss": 0.6124, + "step": 12889 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014123986890868554, + "loss": 0.5713, + "step": 12890 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001412310844506721, + "loss": 0.5208, + "step": 12891 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014122229960930846, + "loss": 0.5756, + "step": 12892 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014121351438467632, + "loss": 0.5644, + "step": 12893 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014120472877685733, + "loss": 0.4833, + "step": 12894 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001411959427859332, + "loss": 0.4927, + "step": 12895 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014118715641198563, + "loss": 0.6501, + "step": 12896 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014117836965509624, + "loss": 0.6097, + "step": 12897 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014116958251534687, + "loss": 0.5844, + "step": 12898 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014116079499281906, + "loss": 0.5164, + "step": 12899 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014115200708759464, + "loss": 0.5286, + "step": 12900 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014114321879975518, + "loss": 0.5338, + "step": 12901 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014113443012938253, + "loss": 0.4939, + "step": 12902 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014112564107655832, + "loss": 0.5436, + "step": 12903 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014111685164136433, + "loss": 0.5897, + "step": 12904 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001411080618238822, + "loss": 0.6035, + "step": 12905 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014109927162419373, + "loss": 0.6286, + "step": 12906 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014109048104238056, + "loss": 0.4783, + "step": 12907 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014108169007852453, + "loss": 0.5259, + "step": 12908 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001410728987327073, + "loss": 0.5533, + "step": 12909 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014106410700501058, + "loss": 0.5475, + "step": 12910 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014105531489551622, + "loss": 0.6066, + "step": 12911 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014104652240430588, + "loss": 0.5119, + "step": 12912 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014103772953146136, + "loss": 0.5765, + "step": 12913 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014102893627706436, + "loss": 0.5448, + "step": 12914 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014102014264119665, + "loss": 0.5593, + "step": 12915 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014101134862394001, + "loss": 0.6161, + "step": 12916 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014100255422537622, + "loss": 0.5909, + "step": 12917 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014099375944558702, + "loss": 0.6082, + "step": 12918 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014098496428465416, + "loss": 0.6076, + "step": 12919 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014097616874265945, + "loss": 0.5345, + "step": 12920 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014096737281968466, + "loss": 0.6884, + "step": 12921 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014095857651581155, + "loss": 0.4652, + "step": 12922 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014094977983112195, + "loss": 0.5179, + "step": 12923 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001409409827656976, + "loss": 0.5196, + "step": 12924 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001409321853196203, + "loss": 0.5265, + "step": 12925 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014092338749297187, + "loss": 0.5787, + "step": 12926 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014091458928583406, + "loss": 0.3801, + "step": 12927 + }, + { + "epoch": 0.38, + "learning_rate": 0.00014090579069828872, + "loss": 0.5201, + "step": 12928 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001408969917304177, + "loss": 0.5521, + "step": 12929 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001408881923823027, + "loss": 0.5495, + "step": 12930 + }, + { + "epoch": 0.38, + "learning_rate": 0.0001408793926540256, + "loss": 0.6006, + "step": 12931 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014087059254566816, + "loss": 0.5298, + "step": 12932 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014086179205731228, + "loss": 0.5815, + "step": 12933 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001408529911890397, + "loss": 0.552, + "step": 12934 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014084418994093234, + "loss": 0.5106, + "step": 12935 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014083538831307198, + "loss": 0.5413, + "step": 12936 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014082658630554042, + "loss": 0.5927, + "step": 12937 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014081778391841955, + "loss": 0.5618, + "step": 12938 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014080898115179119, + "loss": 0.5481, + "step": 12939 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014080017800573722, + "loss": 0.5025, + "step": 12940 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014079137448033943, + "loss": 0.6289, + "step": 12941 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014078257057567966, + "loss": 0.4294, + "step": 12942 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014077376629183987, + "loss": 0.5062, + "step": 12943 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014076496162890182, + "loss": 0.4331, + "step": 12944 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014075615658694742, + "loss": 0.5592, + "step": 12945 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001407473511660585, + "loss": 0.6252, + "step": 12946 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014073854536631692, + "loss": 0.5304, + "step": 12947 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014072973918780466, + "loss": 0.5956, + "step": 12948 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014072093263060344, + "loss": 0.446, + "step": 12949 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001407121256947953, + "loss": 0.5552, + "step": 12950 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014070331838046197, + "loss": 0.5845, + "step": 12951 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001406945106876854, + "loss": 0.4399, + "step": 12952 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014068570261654756, + "loss": 0.4958, + "step": 12953 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001406768941671302, + "loss": 0.4892, + "step": 12954 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014066808533951534, + "loss": 0.5941, + "step": 12955 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014065927613378482, + "loss": 0.5631, + "step": 12956 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014065046655002054, + "loss": 0.6192, + "step": 12957 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014064165658830445, + "loss": 0.5608, + "step": 12958 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001406328462487184, + "loss": 0.4903, + "step": 12959 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001406240355313444, + "loss": 0.6095, + "step": 12960 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014061522443626422, + "loss": 0.4987, + "step": 12961 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014060641296355993, + "loss": 0.6882, + "step": 12962 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001405976011133134, + "loss": 0.5967, + "step": 12963 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014058878888560653, + "loss": 0.5915, + "step": 12964 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001405799762805213, + "loss": 0.5892, + "step": 12965 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014057116329813958, + "loss": 0.6285, + "step": 12966 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014056234993854342, + "loss": 0.548, + "step": 12967 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014055353620181466, + "loss": 0.5853, + "step": 12968 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001405447220880353, + "loss": 0.4678, + "step": 12969 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014053590759728726, + "loss": 0.5276, + "step": 12970 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014052709272965253, + "loss": 0.5985, + "step": 12971 + }, + { + "epoch": 0.39, + "learning_rate": 0.000140518277485213, + "loss": 0.5874, + "step": 12972 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014050946186405067, + "loss": 0.6176, + "step": 12973 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014050064586624758, + "loss": 0.5731, + "step": 12974 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014049182949188555, + "loss": 0.5318, + "step": 12975 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001404830127410467, + "loss": 0.5333, + "step": 12976 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014047419561381288, + "loss": 0.6287, + "step": 12977 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014046537811026612, + "loss": 0.5136, + "step": 12978 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014045656023048842, + "loss": 0.5275, + "step": 12979 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014044774197456174, + "loss": 0.5426, + "step": 12980 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014043892334256812, + "loss": 0.4754, + "step": 12981 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014043010433458943, + "loss": 0.5432, + "step": 12982 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014042128495070778, + "loss": 0.5743, + "step": 12983 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014041246519100515, + "loss": 0.5497, + "step": 12984 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001404036450555635, + "loss": 0.5091, + "step": 12985 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001403948245444649, + "loss": 0.4696, + "step": 12986 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014038600365779127, + "loss": 0.5947, + "step": 12987 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014037718239562471, + "loss": 0.532, + "step": 12988 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014036836075804717, + "loss": 0.5858, + "step": 12989 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001403595387451407, + "loss": 0.5471, + "step": 12990 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014035071635698737, + "loss": 0.5686, + "step": 12991 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014034189359366913, + "loss": 0.5488, + "step": 12992 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014033307045526806, + "loss": 0.6143, + "step": 12993 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014032424694186615, + "loss": 0.5372, + "step": 12994 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014031542305354547, + "loss": 0.5145, + "step": 12995 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014030659879038808, + "loss": 0.5224, + "step": 12996 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014029777415247599, + "loss": 0.7467, + "step": 12997 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014028894913989124, + "loss": 0.5201, + "step": 12998 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001402801237527159, + "loss": 0.4419, + "step": 12999 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014027129799103202, + "loss": 0.5859, + "step": 13000 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001402624718549217, + "loss": 0.5546, + "step": 13001 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014025364534446693, + "loss": 0.5813, + "step": 13002 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001402448184597498, + "loss": 0.4986, + "step": 13003 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001402359912008524, + "loss": 0.6038, + "step": 13004 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001402271635678568, + "loss": 0.5168, + "step": 13005 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014021833556084504, + "loss": 0.5299, + "step": 13006 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014020950717989926, + "loss": 0.594, + "step": 13007 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001402006784251015, + "loss": 0.4873, + "step": 13008 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014019184929653383, + "loss": 0.5602, + "step": 13009 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014018301979427838, + "loss": 0.5172, + "step": 13010 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014017418991841725, + "loss": 0.5328, + "step": 13011 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001401653596690325, + "loss": 0.5163, + "step": 13012 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001401565290462062, + "loss": 0.6567, + "step": 13013 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014014769805002054, + "loss": 0.6991, + "step": 13014 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001401388666805576, + "loss": 0.5876, + "step": 13015 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014013003493789944, + "loss": 0.563, + "step": 13016 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014012120282212823, + "loss": 0.5716, + "step": 13017 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014011237033332606, + "loss": 0.5876, + "step": 13018 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014010353747157507, + "loss": 0.5158, + "step": 13019 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014009470423695736, + "loss": 0.5895, + "step": 13020 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014008587062955505, + "loss": 0.6915, + "step": 13021 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014007703664945033, + "loss": 0.6028, + "step": 13022 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014006820229672527, + "loss": 0.5023, + "step": 13023 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014005936757146206, + "loss": 0.474, + "step": 13024 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014005053247374277, + "loss": 0.5983, + "step": 13025 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014004169700364963, + "loss": 0.5954, + "step": 13026 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014003286116126472, + "loss": 0.647, + "step": 13027 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014002402494667026, + "loss": 0.5049, + "step": 13028 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014001518835994836, + "loss": 0.4851, + "step": 13029 + }, + { + "epoch": 0.39, + "learning_rate": 0.00014000635140118116, + "loss": 0.4535, + "step": 13030 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001399975140704509, + "loss": 0.5627, + "step": 13031 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013998867636783965, + "loss": 0.5823, + "step": 13032 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013997983829342964, + "loss": 0.488, + "step": 13033 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013997099984730305, + "loss": 0.5053, + "step": 13034 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013996216102954198, + "loss": 0.517, + "step": 13035 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001399533218402287, + "loss": 0.5389, + "step": 13036 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013994448227944536, + "loss": 0.5838, + "step": 13037 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013993564234727413, + "loss": 0.5701, + "step": 13038 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013992680204379722, + "loss": 0.4919, + "step": 13039 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013991796136909683, + "loss": 0.6719, + "step": 13040 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013990912032325515, + "loss": 0.6735, + "step": 13041 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013990027890635432, + "loss": 0.6162, + "step": 13042 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013989143711847668, + "loss": 0.5638, + "step": 13043 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001398825949597043, + "loss": 0.5818, + "step": 13044 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013987375243011945, + "loss": 0.568, + "step": 13045 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013986490952980438, + "loss": 0.6323, + "step": 13046 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013985606625884124, + "loss": 0.4825, + "step": 13047 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001398472226173123, + "loss": 0.5736, + "step": 13048 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013983837860529974, + "loss": 0.6093, + "step": 13049 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013982953422288582, + "loss": 0.5465, + "step": 13050 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013982068947015276, + "loss": 0.5866, + "step": 13051 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013981184434718282, + "loss": 0.482, + "step": 13052 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001398029988540582, + "loss": 0.4675, + "step": 13053 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001397941529908612, + "loss": 0.5155, + "step": 13054 + }, + { + "epoch": 0.39, + "learning_rate": 0.000139785306757674, + "loss": 0.6001, + "step": 13055 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013977646015457884, + "loss": 0.6242, + "step": 13056 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013976761318165804, + "loss": 0.5879, + "step": 13057 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013975876583899384, + "loss": 0.5383, + "step": 13058 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013974991812666844, + "loss": 0.4415, + "step": 13059 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013974107004476418, + "loss": 0.5074, + "step": 13060 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013973222159336328, + "loss": 0.5956, + "step": 13061 + }, + { + "epoch": 0.39, + "learning_rate": 0.000139723372772548, + "loss": 0.5681, + "step": 13062 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013971452358240067, + "loss": 0.629, + "step": 13063 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013970567402300349, + "loss": 0.5598, + "step": 13064 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001396968240944388, + "loss": 0.5237, + "step": 13065 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013968797379678883, + "loss": 0.5326, + "step": 13066 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013967912313013595, + "loss": 0.6717, + "step": 13067 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013967027209456237, + "loss": 0.5381, + "step": 13068 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001396614206901504, + "loss": 0.716, + "step": 13069 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013965256891698236, + "loss": 0.5539, + "step": 13070 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013964371677514055, + "loss": 0.4821, + "step": 13071 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013963486426470724, + "loss": 0.4767, + "step": 13072 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013962601138576476, + "loss": 0.5361, + "step": 13073 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013961715813839545, + "loss": 0.4909, + "step": 13074 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013960830452268156, + "loss": 0.5772, + "step": 13075 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013959945053870545, + "loss": 0.6112, + "step": 13076 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013959059618654944, + "loss": 0.6412, + "step": 13077 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013958174146629584, + "loss": 0.5213, + "step": 13078 + }, + { + "epoch": 0.39, + "learning_rate": 0.000139572886378027, + "loss": 0.5329, + "step": 13079 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001395640309218252, + "loss": 0.6247, + "step": 13080 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013955517509777282, + "loss": 0.6923, + "step": 13081 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001395463189059522, + "loss": 0.5521, + "step": 13082 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013953746234644565, + "loss": 0.5709, + "step": 13083 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013952860541933558, + "loss": 0.5703, + "step": 13084 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013951974812470423, + "loss": 0.4969, + "step": 13085 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013951089046263405, + "loss": 0.4785, + "step": 13086 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013950203243320733, + "loss": 0.49, + "step": 13087 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013949317403650648, + "loss": 0.4634, + "step": 13088 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013948431527261384, + "loss": 0.4402, + "step": 13089 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013947545614161178, + "loss": 0.6841, + "step": 13090 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013946659664358264, + "loss": 0.4715, + "step": 13091 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013945773677860883, + "loss": 0.6004, + "step": 13092 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001394488765467727, + "loss": 0.511, + "step": 13093 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013944001594815662, + "loss": 0.5595, + "step": 13094 + }, + { + "epoch": 0.39, + "learning_rate": 0.000139431154982843, + "loss": 0.5699, + "step": 13095 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013942229365091426, + "loss": 0.6792, + "step": 13096 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001394134319524527, + "loss": 0.5545, + "step": 13097 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013940456988754079, + "loss": 0.5459, + "step": 13098 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013939570745626084, + "loss": 0.6392, + "step": 13099 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013938684465869536, + "loss": 0.5055, + "step": 13100 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013937798149492666, + "loss": 0.5719, + "step": 13101 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013936911796503723, + "loss": 0.5928, + "step": 13102 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013936025406910939, + "loss": 0.5013, + "step": 13103 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013935138980722563, + "loss": 0.5232, + "step": 13104 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013934252517946828, + "loss": 0.5731, + "step": 13105 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013933366018591985, + "loss": 0.5465, + "step": 13106 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001393247948266627, + "loss": 0.5102, + "step": 13107 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001393159291017793, + "loss": 0.4942, + "step": 13108 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013930706301135208, + "loss": 0.5073, + "step": 13109 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001392981965554634, + "loss": 0.4951, + "step": 13110 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001392893297341958, + "loss": 0.5017, + "step": 13111 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013928046254763166, + "loss": 0.7003, + "step": 13112 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013927159499585342, + "loss": 0.546, + "step": 13113 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013926272707894354, + "loss": 0.5306, + "step": 13114 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001392538587969845, + "loss": 0.5701, + "step": 13115 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001392449901500587, + "loss": 0.5172, + "step": 13116 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013923612113824865, + "loss": 0.6395, + "step": 13117 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013922725176163673, + "loss": 0.6396, + "step": 13118 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001392183820203055, + "loss": 0.6574, + "step": 13119 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013920951191433736, + "loss": 0.585, + "step": 13120 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013920064144381483, + "loss": 0.5777, + "step": 13121 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013919177060882037, + "loss": 0.6291, + "step": 13122 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013918289940943642, + "loss": 0.5541, + "step": 13123 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001391740278457455, + "loss": 0.5782, + "step": 13124 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001391651559178301, + "loss": 0.5979, + "step": 13125 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013915628362577264, + "loss": 0.5334, + "step": 13126 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013914741096965571, + "loss": 0.5029, + "step": 13127 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001391385379495617, + "loss": 0.5754, + "step": 13128 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013912966456557324, + "loss": 0.505, + "step": 13129 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013912079081777269, + "loss": 0.4671, + "step": 13130 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013911191670624263, + "loss": 0.5463, + "step": 13131 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013910304223106558, + "loss": 0.5697, + "step": 13132 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013909416739232402, + "loss": 0.5948, + "step": 13133 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013908529219010046, + "loss": 0.4932, + "step": 13134 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013907641662447745, + "loss": 0.5969, + "step": 13135 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001390675406955375, + "loss": 0.5604, + "step": 13136 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013905866440336312, + "loss": 0.6085, + "step": 13137 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013904978774803683, + "loss": 0.5651, + "step": 13138 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001390409107296412, + "loss": 0.6369, + "step": 13139 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001390320333482587, + "loss": 0.4444, + "step": 13140 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013902315560397195, + "loss": 0.5175, + "step": 13141 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013901427749686345, + "loss": 0.5547, + "step": 13142 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013900539902701574, + "loss": 0.5638, + "step": 13143 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013899652019451136, + "loss": 0.5111, + "step": 13144 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001389876409994329, + "loss": 0.5705, + "step": 13145 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013897876144186287, + "loss": 0.4484, + "step": 13146 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013896988152188385, + "loss": 0.5206, + "step": 13147 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013896100123957845, + "loss": 0.5323, + "step": 13148 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013895212059502913, + "loss": 0.4999, + "step": 13149 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013894323958831857, + "loss": 0.555, + "step": 13150 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013893435821952924, + "loss": 0.5298, + "step": 13151 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001389254764887438, + "loss": 0.626, + "step": 13152 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013891659439604478, + "loss": 0.5883, + "step": 13153 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001389077119415148, + "loss": 0.565, + "step": 13154 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001388988291252364, + "loss": 0.4521, + "step": 13155 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013888994594729218, + "loss": 0.5453, + "step": 13156 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013888106240776473, + "loss": 0.5693, + "step": 13157 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013887217850673672, + "loss": 0.5087, + "step": 13158 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013886329424429064, + "loss": 0.4787, + "step": 13159 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013885440962050914, + "loss": 0.6098, + "step": 13160 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001388455246354748, + "loss": 0.6145, + "step": 13161 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013883663928927032, + "loss": 0.4811, + "step": 13162 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001388277535819782, + "loss": 0.6946, + "step": 13163 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013881886751368114, + "loss": 0.5873, + "step": 13164 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013880998108446168, + "loss": 0.6306, + "step": 13165 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001388010942944025, + "loss": 0.5354, + "step": 13166 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013879220714358621, + "loss": 0.5542, + "step": 13167 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013878331963209547, + "loss": 0.4491, + "step": 13168 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013877443176001287, + "loss": 0.5029, + "step": 13169 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013876554352742102, + "loss": 0.5555, + "step": 13170 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001387566549344026, + "loss": 0.6622, + "step": 13171 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013874776598104028, + "loss": 0.6111, + "step": 13172 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013873887666741668, + "loss": 0.4702, + "step": 13173 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013872998699361442, + "loss": 0.6937, + "step": 13174 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013872109695971618, + "loss": 0.5442, + "step": 13175 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001387122065658046, + "loss": 0.6108, + "step": 13176 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001387033158119624, + "loss": 0.6081, + "step": 13177 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013869442469827215, + "loss": 0.5625, + "step": 13178 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001386855332248166, + "loss": 0.5972, + "step": 13179 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013867664139167833, + "loss": 0.5509, + "step": 13180 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001386677491989401, + "loss": 0.6258, + "step": 13181 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013865885664668452, + "loss": 0.5426, + "step": 13182 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013864996373499434, + "loss": 0.4166, + "step": 13183 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013864107046395217, + "loss": 0.5598, + "step": 13184 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013863217683364074, + "loss": 0.5219, + "step": 13185 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001386232828441427, + "loss": 0.6236, + "step": 13186 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013861438849554082, + "loss": 0.5245, + "step": 13187 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001386054937879177, + "loss": 0.4509, + "step": 13188 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013859659872135613, + "loss": 0.5237, + "step": 13189 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001385877032959387, + "loss": 0.5397, + "step": 13190 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013857880751174826, + "loss": 0.5088, + "step": 13191 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001385699113688674, + "loss": 0.4646, + "step": 13192 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013856101486737888, + "loss": 0.4877, + "step": 13193 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013855211800736542, + "loss": 0.491, + "step": 13194 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013854322078890973, + "loss": 0.4185, + "step": 13195 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013853432321209454, + "loss": 0.6111, + "step": 13196 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001385254252770026, + "loss": 0.5545, + "step": 13197 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013851652698371655, + "loss": 0.4664, + "step": 13198 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013850762833231923, + "loss": 0.6489, + "step": 13199 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013849872932289328, + "loss": 0.5384, + "step": 13200 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013848982995552157, + "loss": 0.6578, + "step": 13201 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013848093023028667, + "loss": 0.6195, + "step": 13202 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013847203014727148, + "loss": 0.4985, + "step": 13203 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013846312970655866, + "loss": 0.5291, + "step": 13204 + }, + { + "epoch": 0.39, + "learning_rate": 0.000138454228908231, + "loss": 0.4391, + "step": 13205 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013844532775237126, + "loss": 0.5372, + "step": 13206 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013843642623906214, + "loss": 0.4944, + "step": 13207 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013842752436838652, + "loss": 0.5526, + "step": 13208 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013841862214042703, + "loss": 0.5415, + "step": 13209 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013840971955526654, + "loss": 0.4279, + "step": 13210 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013840081661298774, + "loss": 0.6076, + "step": 13211 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013839191331367347, + "loss": 0.5614, + "step": 13212 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013838300965740653, + "loss": 0.5295, + "step": 13213 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013837410564426962, + "loss": 0.5764, + "step": 13214 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001383652012743456, + "loss": 0.4805, + "step": 13215 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013835629654771722, + "loss": 0.6221, + "step": 13216 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013834739146446727, + "loss": 0.4283, + "step": 13217 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001383384860246786, + "loss": 0.5974, + "step": 13218 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001383295802284339, + "loss": 0.5471, + "step": 13219 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013832067407581607, + "loss": 0.5779, + "step": 13220 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013831176756690789, + "loss": 0.5557, + "step": 13221 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001383028607017922, + "loss": 0.6205, + "step": 13222 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013829395348055173, + "loss": 0.5392, + "step": 13223 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013828504590326934, + "loss": 0.5246, + "step": 13224 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013827613797002787, + "loss": 0.4826, + "step": 13225 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013826722968091014, + "loss": 0.5705, + "step": 13226 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001382583210359989, + "loss": 0.4789, + "step": 13227 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001382494120353771, + "loss": 0.508, + "step": 13228 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001382405026791275, + "loss": 0.5112, + "step": 13229 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013823159296733292, + "loss": 0.5455, + "step": 13230 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013822268290007624, + "loss": 0.4239, + "step": 13231 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013821377247744032, + "loss": 0.5786, + "step": 13232 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013820486169950792, + "loss": 0.4467, + "step": 13233 + }, + { + "epoch": 0.39, + "learning_rate": 0.000138195950566362, + "loss": 0.465, + "step": 13234 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001381870390780853, + "loss": 0.5966, + "step": 13235 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001381781272347608, + "loss": 0.5929, + "step": 13236 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001381692150364712, + "loss": 0.55, + "step": 13237 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013816030248329952, + "loss": 0.4225, + "step": 13238 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013815138957532853, + "loss": 0.5173, + "step": 13239 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001381424763126411, + "loss": 0.5101, + "step": 13240 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001381335626953202, + "loss": 0.5543, + "step": 13241 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013812464872344857, + "loss": 0.5466, + "step": 13242 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001381157343971092, + "loss": 0.6282, + "step": 13243 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001381068197163849, + "loss": 0.4667, + "step": 13244 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013809790468135856, + "loss": 0.563, + "step": 13245 + }, + { + "epoch": 0.39, + "learning_rate": 0.0001380889892921131, + "loss": 0.5979, + "step": 13246 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013808007354873137, + "loss": 0.5304, + "step": 13247 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013807115745129633, + "loss": 0.675, + "step": 13248 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013806224099989085, + "loss": 0.4856, + "step": 13249 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013805332419459778, + "loss": 0.4879, + "step": 13250 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013804440703550012, + "loss": 0.5484, + "step": 13251 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013803548952268067, + "loss": 0.6221, + "step": 13252 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013802657165622246, + "loss": 0.5845, + "step": 13253 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013801765343620826, + "loss": 0.6345, + "step": 13254 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013800873486272114, + "loss": 0.5167, + "step": 13255 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013799981593584393, + "loss": 0.7293, + "step": 13256 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013799089665565958, + "loss": 0.5337, + "step": 13257 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013798197702225102, + "loss": 0.5382, + "step": 13258 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013797305703570117, + "loss": 0.511, + "step": 13259 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013796413669609297, + "loss": 0.5465, + "step": 13260 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013795521600350937, + "loss": 0.6037, + "step": 13261 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013794629495803327, + "loss": 0.5192, + "step": 13262 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013793737355974767, + "loss": 0.5511, + "step": 13263 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013792845180873547, + "loss": 0.5415, + "step": 13264 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013791952970507968, + "loss": 0.68, + "step": 13265 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013791060724886319, + "loss": 0.5624, + "step": 13266 + }, + { + "epoch": 0.39, + "learning_rate": 0.00013790168444016896, + "loss": 0.5423, + "step": 13267 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013789276127908, + "loss": 0.5616, + "step": 13268 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013788383776567925, + "loss": 0.5049, + "step": 13269 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013787491390004967, + "loss": 0.5757, + "step": 13270 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013786598968227424, + "loss": 0.5581, + "step": 13271 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013785706511243592, + "loss": 0.5238, + "step": 13272 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013784814019061773, + "loss": 0.567, + "step": 13273 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013783921491690256, + "loss": 0.5306, + "step": 13274 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001378302892913735, + "loss": 0.5456, + "step": 13275 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013782136331411346, + "loss": 0.5533, + "step": 13276 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013781243698520547, + "loss": 0.5365, + "step": 13277 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001378035103047325, + "loss": 0.5122, + "step": 13278 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001377945832727776, + "loss": 0.6047, + "step": 13279 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013778565588942366, + "loss": 0.6714, + "step": 13280 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013777672815475377, + "loss": 0.4435, + "step": 13281 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013776780006885097, + "loss": 0.5232, + "step": 13282 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013775887163179818, + "loss": 0.5195, + "step": 13283 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013774994284367846, + "loss": 0.4978, + "step": 13284 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013774101370457478, + "loss": 0.532, + "step": 13285 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013773208421457024, + "loss": 0.6759, + "step": 13286 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001377231543737478, + "loss": 0.5024, + "step": 13287 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013771422418219054, + "loss": 0.5965, + "step": 13288 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013770529363998143, + "loss": 0.5908, + "step": 13289 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001376963627472035, + "loss": 0.646, + "step": 13290 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013768743150393986, + "loss": 0.6388, + "step": 13291 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013767849991027346, + "loss": 0.5415, + "step": 13292 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001376695679662874, + "loss": 0.4734, + "step": 13293 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013766063567206474, + "loss": 0.5281, + "step": 13294 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013765170302768845, + "loss": 0.6073, + "step": 13295 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013764277003324167, + "loss": 0.655, + "step": 13296 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013763383668880742, + "loss": 0.465, + "step": 13297 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013762490299446873, + "loss": 0.6886, + "step": 13298 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001376159689503087, + "loss": 0.4566, + "step": 13299 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013760703455641037, + "loss": 0.5844, + "step": 13300 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013759809981285683, + "loss": 0.5227, + "step": 13301 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013758916471973115, + "loss": 0.5588, + "step": 13302 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001375802292771164, + "loss": 0.5904, + "step": 13303 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013757129348509564, + "loss": 0.4452, + "step": 13304 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013756235734375198, + "loss": 0.5108, + "step": 13305 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013755342085316848, + "loss": 0.6383, + "step": 13306 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013754448401342826, + "loss": 0.5565, + "step": 13307 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013753554682461442, + "loss": 0.5019, + "step": 13308 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013752660928680995, + "loss": 0.5119, + "step": 13309 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001375176714000981, + "loss": 0.6137, + "step": 13310 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013750873316456183, + "loss": 0.6206, + "step": 13311 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013749979458028432, + "loss": 0.6361, + "step": 13312 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001374908556473487, + "loss": 0.5425, + "step": 13313 + }, + { + "epoch": 0.4, + "learning_rate": 0.000137481916365838, + "loss": 0.4641, + "step": 13314 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013747297673583544, + "loss": 0.6328, + "step": 13315 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013746403675742402, + "loss": 0.5589, + "step": 13316 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013745509643068695, + "loss": 0.5029, + "step": 13317 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013744615575570732, + "loss": 0.5929, + "step": 13318 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013743721473256825, + "loss": 0.496, + "step": 13319 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013742827336135285, + "loss": 0.6198, + "step": 13320 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013741933164214434, + "loss": 0.4743, + "step": 13321 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013741038957502578, + "loss": 0.5265, + "step": 13322 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013740144716008033, + "loss": 0.5887, + "step": 13323 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013739250439739108, + "loss": 0.712, + "step": 13324 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001373835612870413, + "loss": 0.559, + "step": 13325 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013737461782911402, + "loss": 0.4562, + "step": 13326 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013736567402369246, + "loss": 0.5534, + "step": 13327 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013735672987085973, + "loss": 0.6291, + "step": 13328 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013734778537069904, + "loss": 0.5505, + "step": 13329 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013733884052329352, + "loss": 0.5819, + "step": 13330 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013732989532872633, + "loss": 0.5295, + "step": 13331 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013732094978708068, + "loss": 0.5843, + "step": 13332 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013731200389843968, + "loss": 0.4914, + "step": 13333 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013730305766288656, + "loss": 0.5273, + "step": 13334 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013729411108050449, + "loss": 0.5031, + "step": 13335 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001372851641513766, + "loss": 0.4767, + "step": 13336 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013727621687558615, + "loss": 0.5743, + "step": 13337 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001372672692532163, + "loss": 0.5398, + "step": 13338 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013725832128435022, + "loss": 0.6503, + "step": 13339 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001372493729690711, + "loss": 0.5015, + "step": 13340 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013724042430746217, + "loss": 0.6113, + "step": 13341 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001372314752996066, + "loss": 0.6894, + "step": 13342 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013722252594558761, + "loss": 0.5293, + "step": 13343 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013721357624548846, + "loss": 0.5384, + "step": 13344 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013720462619939226, + "loss": 0.6171, + "step": 13345 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001371956758073823, + "loss": 0.5097, + "step": 13346 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013718672506954176, + "loss": 0.5527, + "step": 13347 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013717777398595385, + "loss": 0.4471, + "step": 13348 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013716882255670185, + "loss": 0.5685, + "step": 13349 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013715987078186893, + "loss": 0.5257, + "step": 13350 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001371509186615384, + "loss": 0.5975, + "step": 13351 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013714196619579337, + "loss": 0.6008, + "step": 13352 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013713301338471714, + "loss": 0.5299, + "step": 13353 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013712406022839296, + "loss": 0.5513, + "step": 13354 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013711510672690405, + "loss": 0.6119, + "step": 13355 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013710615288033373, + "loss": 0.5723, + "step": 13356 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013709719868876513, + "loss": 0.4771, + "step": 13357 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013708824415228162, + "loss": 0.4867, + "step": 13358 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013707928927096633, + "loss": 0.5534, + "step": 13359 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013707033404490265, + "loss": 0.6355, + "step": 13360 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013706137847417372, + "loss": 0.5786, + "step": 13361 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001370524225588629, + "loss": 0.471, + "step": 13362 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013704346629905343, + "loss": 0.5467, + "step": 13363 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013703450969482857, + "loss": 0.4553, + "step": 13364 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001370255527462716, + "loss": 0.5499, + "step": 13365 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013701659545346578, + "loss": 0.4805, + "step": 13366 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001370076378164944, + "loss": 0.4601, + "step": 13367 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013699867983544076, + "loss": 0.5633, + "step": 13368 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013698972151038812, + "loss": 0.6447, + "step": 13369 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013698076284141987, + "loss": 0.5326, + "step": 13370 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013697180382861917, + "loss": 0.5831, + "step": 13371 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013696284447206936, + "loss": 0.5789, + "step": 13372 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001369538847718538, + "loss": 0.4766, + "step": 13373 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013694492472805572, + "loss": 0.6604, + "step": 13374 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013693596434075848, + "loss": 0.5687, + "step": 13375 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001369270036100453, + "loss": 0.605, + "step": 13376 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013691804253599963, + "loss": 0.5915, + "step": 13377 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001369090811187047, + "loss": 0.4696, + "step": 13378 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001369001193582438, + "loss": 0.61, + "step": 13379 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013689115725470035, + "loss": 0.5086, + "step": 13380 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001368821948081576, + "loss": 0.5469, + "step": 13381 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001368732320186989, + "loss": 0.5094, + "step": 13382 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001368642688864076, + "loss": 0.5528, + "step": 13383 + }, + { + "epoch": 0.4, + "learning_rate": 0.000136855305411367, + "loss": 0.4466, + "step": 13384 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013684634159366046, + "loss": 0.5104, + "step": 13385 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013683737743337132, + "loss": 0.5575, + "step": 13386 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013682841293058295, + "loss": 0.5016, + "step": 13387 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013681944808537866, + "loss": 0.4193, + "step": 13388 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013681048289784182, + "loss": 0.6372, + "step": 13389 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013680151736805577, + "loss": 0.4995, + "step": 13390 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013679255149610389, + "loss": 0.5801, + "step": 13391 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013678358528206952, + "loss": 0.4599, + "step": 13392 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013677461872603605, + "loss": 0.4951, + "step": 13393 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013676565182808682, + "loss": 0.5664, + "step": 13394 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001367566845883052, + "loss": 0.5284, + "step": 13395 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013674771700677462, + "loss": 0.6667, + "step": 13396 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001367387490835784, + "loss": 0.5753, + "step": 13397 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013672978081879987, + "loss": 0.5743, + "step": 13398 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013672081221252259, + "loss": 0.6273, + "step": 13399 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013671184326482974, + "loss": 0.5107, + "step": 13400 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013670287397580483, + "loss": 0.527, + "step": 13401 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013669390434553123, + "loss": 0.576, + "step": 13402 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013668493437409232, + "loss": 0.5904, + "step": 13403 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013667596406157154, + "loss": 0.5692, + "step": 13404 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013666699340805224, + "loss": 0.4379, + "step": 13405 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001366580224136179, + "loss": 0.5785, + "step": 13406 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001366490510783518, + "loss": 0.4824, + "step": 13407 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013664007940233748, + "loss": 0.5373, + "step": 13408 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013663110738565827, + "loss": 0.6544, + "step": 13409 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013662213502839765, + "loss": 0.5706, + "step": 13410 + }, + { + "epoch": 0.4, + "learning_rate": 0.000136613162330639, + "loss": 0.5849, + "step": 13411 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013660418929246578, + "loss": 0.4652, + "step": 13412 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001365952159139614, + "loss": 0.6261, + "step": 13413 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013658624219520926, + "loss": 0.6192, + "step": 13414 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013657726813629282, + "loss": 0.6193, + "step": 13415 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013656829373729553, + "loss": 0.5958, + "step": 13416 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001365593189983008, + "loss": 0.4627, + "step": 13417 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013655034391939216, + "loss": 0.4914, + "step": 13418 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001365413685006529, + "loss": 0.5169, + "step": 13419 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013653239274216664, + "loss": 0.4817, + "step": 13420 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001365234166440167, + "loss": 0.5859, + "step": 13421 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001365144402062866, + "loss": 0.5797, + "step": 13422 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001365054634290598, + "loss": 0.5624, + "step": 13423 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013649648631241975, + "loss": 0.5602, + "step": 13424 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013648750885644992, + "loss": 0.5594, + "step": 13425 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013647853106123377, + "loss": 0.6114, + "step": 13426 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013646955292685476, + "loss": 0.7069, + "step": 13427 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013646057445339642, + "loss": 0.5557, + "step": 13428 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013645159564094214, + "loss": 0.6942, + "step": 13429 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001364426164895755, + "loss": 0.6304, + "step": 13430 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001364336369993799, + "loss": 0.6317, + "step": 13431 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001364246571704389, + "loss": 0.6864, + "step": 13432 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001364156770028359, + "loss": 0.4991, + "step": 13433 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001364066964966545, + "loss": 0.6647, + "step": 13434 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013639771565197815, + "loss": 0.5774, + "step": 13435 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001363887344688903, + "loss": 0.6644, + "step": 13436 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013637975294747455, + "loss": 0.474, + "step": 13437 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013637077108781434, + "loss": 0.4975, + "step": 13438 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013636178888999317, + "loss": 0.5504, + "step": 13439 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001363528063540946, + "loss": 0.5713, + "step": 13440 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001363438234802021, + "loss": 0.6037, + "step": 13441 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013633484026839927, + "loss": 0.5249, + "step": 13442 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013632585671876954, + "loss": 0.5502, + "step": 13443 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013631687283139647, + "loss": 0.5226, + "step": 13444 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001363078886063636, + "loss": 0.6039, + "step": 13445 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013629890404375446, + "loss": 0.6131, + "step": 13446 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013628991914365257, + "loss": 0.6413, + "step": 13447 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013628093390614146, + "loss": 0.626, + "step": 13448 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013627194833130472, + "loss": 0.4721, + "step": 13449 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013626296241922583, + "loss": 0.644, + "step": 13450 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013625397616998837, + "loss": 0.5703, + "step": 13451 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001362449895836759, + "loss": 0.5936, + "step": 13452 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013623600266037195, + "loss": 0.5942, + "step": 13453 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013622701540016012, + "loss": 0.5969, + "step": 13454 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013621802780312391, + "loss": 0.5715, + "step": 13455 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013620903986934693, + "loss": 0.5649, + "step": 13456 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013620005159891273, + "loss": 0.5705, + "step": 13457 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013619106299190488, + "loss": 0.6503, + "step": 13458 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013618207404840694, + "loss": 0.6249, + "step": 13459 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013617308476850245, + "loss": 0.5156, + "step": 13460 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013616409515227512, + "loss": 0.4777, + "step": 13461 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001361551051998084, + "loss": 0.4717, + "step": 13462 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013614611491118595, + "loss": 0.6358, + "step": 13463 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013613712428649127, + "loss": 0.574, + "step": 13464 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013612813332580808, + "loss": 0.5642, + "step": 13465 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013611914202921988, + "loss": 0.6372, + "step": 13466 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013611015039681026, + "loss": 0.5151, + "step": 13467 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013610115842866292, + "loss": 0.5571, + "step": 13468 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013609216612486136, + "loss": 0.5, + "step": 13469 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013608317348548922, + "loss": 0.4928, + "step": 13470 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013607418051063012, + "loss": 0.5207, + "step": 13471 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013606518720036767, + "loss": 0.5734, + "step": 13472 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001360561935547855, + "loss": 0.6159, + "step": 13473 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013604719957396717, + "loss": 0.4852, + "step": 13474 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013603820525799636, + "loss": 0.5028, + "step": 13475 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001360292106069567, + "loss": 0.5427, + "step": 13476 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013602021562093182, + "loss": 0.4748, + "step": 13477 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013601122030000527, + "loss": 0.6688, + "step": 13478 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013600222464426077, + "loss": 0.5367, + "step": 13479 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013599322865378197, + "loss": 0.6015, + "step": 13480 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013598423232865243, + "loss": 0.6167, + "step": 13481 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013597523566895588, + "loss": 0.5808, + "step": 13482 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013596623867477589, + "loss": 0.6435, + "step": 13483 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013595724134619617, + "loss": 0.5196, + "step": 13484 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013594824368330034, + "loss": 0.5154, + "step": 13485 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013593924568617208, + "loss": 0.5498, + "step": 13486 + }, + { + "epoch": 0.4, + "learning_rate": 0.000135930247354895, + "loss": 0.4921, + "step": 13487 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013592124868955284, + "loss": 0.4718, + "step": 13488 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001359122496902292, + "loss": 0.5504, + "step": 13489 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013590325035700782, + "loss": 0.5775, + "step": 13490 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013589425068997226, + "loss": 0.5772, + "step": 13491 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013588525068920632, + "loss": 0.4886, + "step": 13492 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001358762503547936, + "loss": 0.5619, + "step": 13493 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013586724968681782, + "loss": 0.6095, + "step": 13494 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001358582486853626, + "loss": 0.4988, + "step": 13495 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013584924735051172, + "loss": 0.5386, + "step": 13496 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013584024568234884, + "loss": 0.5893, + "step": 13497 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013583124368095762, + "loss": 0.5774, + "step": 13498 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013582224134642176, + "loss": 0.5892, + "step": 13499 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013581323867882498, + "loss": 0.5084, + "step": 13500 + }, + { + "epoch": 0.4, + "learning_rate": 0.000135804235678251, + "loss": 0.6364, + "step": 13501 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013579523234478352, + "loss": 0.4368, + "step": 13502 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001357862286785062, + "loss": 0.5477, + "step": 13503 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001357772246795028, + "loss": 0.6009, + "step": 13504 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013576822034785704, + "loss": 0.5247, + "step": 13505 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001357592156836526, + "loss": 0.5971, + "step": 13506 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013575021068697323, + "loss": 0.6264, + "step": 13507 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013574120535790268, + "loss": 0.4618, + "step": 13508 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001357321996965246, + "loss": 0.5285, + "step": 13509 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013572319370292276, + "loss": 0.4391, + "step": 13510 + }, + { + "epoch": 0.4, + "learning_rate": 0.000135714187377181, + "loss": 0.5296, + "step": 13511 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013570518071938286, + "loss": 0.7069, + "step": 13512 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013569617372961222, + "loss": 0.5373, + "step": 13513 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013568716640795277, + "loss": 0.5601, + "step": 13514 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013567815875448829, + "loss": 0.4963, + "step": 13515 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013566915076930247, + "loss": 0.4713, + "step": 13516 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013566014245247913, + "loss": 0.6675, + "step": 13517 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013565113380410203, + "loss": 0.6256, + "step": 13518 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013564212482425485, + "loss": 0.6305, + "step": 13519 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001356331155130214, + "loss": 0.5347, + "step": 13520 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013562410587048546, + "loss": 0.578, + "step": 13521 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013561509589673078, + "loss": 0.6105, + "step": 13522 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013560608559184115, + "loss": 0.5507, + "step": 13523 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013559707495590032, + "loss": 0.5033, + "step": 13524 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013558806398899207, + "loss": 0.4412, + "step": 13525 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013557905269120016, + "loss": 0.5906, + "step": 13526 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001355700410626084, + "loss": 0.4439, + "step": 13527 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001355610291033006, + "loss": 0.5096, + "step": 13528 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001355520168133605, + "loss": 0.634, + "step": 13529 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013554300419287195, + "loss": 0.5631, + "step": 13530 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013553399124191872, + "loss": 0.5636, + "step": 13531 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001355249779605846, + "loss": 0.5799, + "step": 13532 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013551596434895335, + "loss": 0.6361, + "step": 13533 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013550695040710883, + "loss": 0.5714, + "step": 13534 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013549793613513487, + "loss": 0.6247, + "step": 13535 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013548892153311524, + "loss": 0.5514, + "step": 13536 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013547990660113375, + "loss": 0.5273, + "step": 13537 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013547089133927423, + "loss": 0.4117, + "step": 13538 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001354618757476205, + "loss": 0.5837, + "step": 13539 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013545285982625638, + "loss": 0.4829, + "step": 13540 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013544384357526567, + "loss": 0.5257, + "step": 13541 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001354348269947323, + "loss": 0.6488, + "step": 13542 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013542581008473996, + "loss": 0.4899, + "step": 13543 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001354167928453726, + "loss": 0.6888, + "step": 13544 + }, + { + "epoch": 0.4, + "learning_rate": 0.000135407775276714, + "loss": 0.7155, + "step": 13545 + }, + { + "epoch": 0.4, + "learning_rate": 0.000135398757378848, + "loss": 0.5822, + "step": 13546 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001353897391518585, + "loss": 0.5882, + "step": 13547 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001353807205958293, + "loss": 0.5476, + "step": 13548 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013537170171084426, + "loss": 0.588, + "step": 13549 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013536268249698724, + "loss": 0.5496, + "step": 13550 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001353536629543421, + "loss": 0.4907, + "step": 13551 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013534464308299267, + "loss": 0.472, + "step": 13552 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013533562288302282, + "loss": 0.5041, + "step": 13553 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013532660235451646, + "loss": 0.4318, + "step": 13554 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013531758149755741, + "loss": 0.5531, + "step": 13555 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001353085603122296, + "loss": 0.6344, + "step": 13556 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013529953879861683, + "loss": 0.5897, + "step": 13557 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013529051695680304, + "loss": 0.512, + "step": 13558 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001352814947868721, + "loss": 0.4482, + "step": 13559 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013527247228890785, + "loss": 0.4389, + "step": 13560 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013526344946299422, + "loss": 0.5829, + "step": 13561 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013525442630921513, + "loss": 0.5691, + "step": 13562 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013524540282765438, + "loss": 0.5087, + "step": 13563 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013523637901839596, + "loss": 0.4546, + "step": 13564 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001352273548815237, + "loss": 0.6335, + "step": 13565 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013521833041712158, + "loss": 0.5989, + "step": 13566 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013520930562527342, + "loss": 0.5766, + "step": 13567 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013520028050606318, + "loss": 0.5516, + "step": 13568 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013519125505957478, + "loss": 0.5528, + "step": 13569 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013518222928589208, + "loss": 0.6009, + "step": 13570 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013517320318509908, + "loss": 0.4932, + "step": 13571 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013516417675727958, + "loss": 0.4881, + "step": 13572 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013515515000251765, + "loss": 0.5977, + "step": 13573 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013514612292089712, + "loss": 0.6195, + "step": 13574 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013513709551250198, + "loss": 0.4257, + "step": 13575 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001351280677774161, + "loss": 0.6478, + "step": 13576 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013511903971572345, + "loss": 0.5097, + "step": 13577 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013511001132750797, + "loss": 0.5167, + "step": 13578 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013510098261285358, + "loss": 0.5069, + "step": 13579 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013509195357184426, + "loss": 0.4487, + "step": 13580 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013508292420456395, + "loss": 0.5829, + "step": 13581 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013507389451109658, + "loss": 0.5079, + "step": 13582 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013506486449152611, + "loss": 0.5193, + "step": 13583 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001350558341459365, + "loss": 0.5978, + "step": 13584 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013504680347441175, + "loss": 0.5532, + "step": 13585 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013503777247703575, + "loss": 0.5842, + "step": 13586 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013502874115389255, + "loss": 0.5123, + "step": 13587 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013501970950506603, + "loss": 0.5609, + "step": 13588 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013501067753064026, + "loss": 0.5361, + "step": 13589 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001350016452306991, + "loss": 0.6188, + "step": 13590 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013499261260532664, + "loss": 0.557, + "step": 13591 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013498357965460682, + "loss": 0.5793, + "step": 13592 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001349745463786236, + "loss": 0.5569, + "step": 13593 + }, + { + "epoch": 0.4, + "learning_rate": 0.000134965512777461, + "loss": 0.6311, + "step": 13594 + }, + { + "epoch": 0.4, + "learning_rate": 0.000134956478851203, + "loss": 0.533, + "step": 13595 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013494744459993356, + "loss": 0.6247, + "step": 13596 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013493841002373675, + "loss": 0.583, + "step": 13597 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001349293751226965, + "loss": 0.5776, + "step": 13598 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013492033989689686, + "loss": 0.4989, + "step": 13599 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001349113043464218, + "loss": 0.5761, + "step": 13600 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001349022684713553, + "loss": 0.6044, + "step": 13601 + }, + { + "epoch": 0.4, + "learning_rate": 0.00013489323227178153, + "loss": 0.456, + "step": 13602 + }, + { + "epoch": 0.4, + "learning_rate": 0.0001348841957477843, + "loss": 0.6471, + "step": 13603 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013487515889944778, + "loss": 0.5536, + "step": 13604 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013486612172685592, + "loss": 0.5267, + "step": 13605 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013485708423009274, + "loss": 0.5408, + "step": 13606 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013484804640924227, + "loss": 0.5239, + "step": 13607 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013483900826438856, + "loss": 0.4939, + "step": 13608 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013482996979561569, + "loss": 0.5463, + "step": 13609 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013482093100300758, + "loss": 0.5211, + "step": 13610 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013481189188664836, + "loss": 0.5578, + "step": 13611 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013480285244662205, + "loss": 0.5795, + "step": 13612 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013479381268301265, + "loss": 0.5229, + "step": 13613 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001347847725959043, + "loss": 0.6346, + "step": 13614 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013477573218538097, + "loss": 0.5337, + "step": 13615 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013476669145152674, + "loss": 0.6341, + "step": 13616 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013475765039442565, + "loss": 0.5852, + "step": 13617 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013474860901416182, + "loss": 0.6309, + "step": 13618 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013473956731081924, + "loss": 0.5473, + "step": 13619 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013473052528448201, + "loss": 0.4768, + "step": 13620 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013472148293523423, + "loss": 0.502, + "step": 13621 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013471244026315993, + "loss": 0.6951, + "step": 13622 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001347033972683432, + "loss": 0.5129, + "step": 13623 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013469435395086808, + "loss": 0.6089, + "step": 13624 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013468531031081874, + "loss": 0.55, + "step": 13625 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013467626634827918, + "loss": 0.5915, + "step": 13626 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013466722206333348, + "loss": 0.5446, + "step": 13627 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001346581774560658, + "loss": 0.5054, + "step": 13628 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001346491325265602, + "loss": 0.5596, + "step": 13629 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001346400872749008, + "loss": 0.5703, + "step": 13630 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013463104170117162, + "loss": 0.4813, + "step": 13631 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013462199580545684, + "loss": 0.5697, + "step": 13632 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001346129495878406, + "loss": 0.5884, + "step": 13633 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013460390304840687, + "loss": 0.5506, + "step": 13634 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013459485618723987, + "loss": 0.481, + "step": 13635 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013458580900442367, + "loss": 0.4958, + "step": 13636 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001345767615000424, + "loss": 0.5717, + "step": 13637 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001345677136741802, + "loss": 0.6041, + "step": 13638 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001345586655269212, + "loss": 0.5586, + "step": 13639 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001345496170583495, + "loss": 0.6102, + "step": 13640 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013454056826854918, + "loss": 0.4619, + "step": 13641 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013453151915760446, + "loss": 0.5016, + "step": 13642 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013452246972559944, + "loss": 0.5172, + "step": 13643 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001345134199726182, + "loss": 0.5645, + "step": 13644 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013450436989874501, + "loss": 0.6353, + "step": 13645 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013449531950406388, + "loss": 0.5476, + "step": 13646 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001344862687886591, + "loss": 0.4775, + "step": 13647 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013447721775261467, + "loss": 0.5102, + "step": 13648 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013446816639601483, + "loss": 0.5164, + "step": 13649 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001344591147189437, + "loss": 0.4777, + "step": 13650 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013445006272148546, + "loss": 0.6155, + "step": 13651 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013444101040372428, + "loss": 0.488, + "step": 13652 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001344319577657443, + "loss": 0.5071, + "step": 13653 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001344229048076297, + "loss": 0.5917, + "step": 13654 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013441385152946467, + "loss": 0.5581, + "step": 13655 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013440479793133336, + "loss": 0.5309, + "step": 13656 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013439574401331994, + "loss": 0.6154, + "step": 13657 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013438668977550859, + "loss": 0.5042, + "step": 13658 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013437763521798355, + "loss": 0.4927, + "step": 13659 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001343685803408289, + "loss": 0.5329, + "step": 13660 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013435952514412894, + "loss": 0.6258, + "step": 13661 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013435046962796777, + "loss": 0.4028, + "step": 13662 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013434141379242964, + "loss": 0.4486, + "step": 13663 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013433235763759874, + "loss": 0.5744, + "step": 13664 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013432330116355925, + "loss": 0.6399, + "step": 13665 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013431424437039538, + "loss": 0.4626, + "step": 13666 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013430518725819136, + "loss": 0.5196, + "step": 13667 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013429612982703137, + "loss": 0.5829, + "step": 13668 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013428707207699964, + "loss": 0.62, + "step": 13669 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013427801400818038, + "loss": 0.5364, + "step": 13670 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001342689556206578, + "loss": 0.532, + "step": 13671 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013425989691451612, + "loss": 0.5549, + "step": 13672 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001342508378898396, + "loss": 0.4965, + "step": 13673 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013424177854671243, + "loss": 0.5306, + "step": 13674 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013423271888521882, + "loss": 0.6202, + "step": 13675 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013422365890544308, + "loss": 0.607, + "step": 13676 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013421459860746938, + "loss": 0.5489, + "step": 13677 + }, + { + "epoch": 0.41, + "learning_rate": 0.000134205537991382, + "loss": 0.6149, + "step": 13678 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001341964770572651, + "loss": 0.5822, + "step": 13679 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013418741580520306, + "loss": 0.5761, + "step": 13680 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013417835423528, + "loss": 0.6163, + "step": 13681 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013416929234758024, + "loss": 0.5687, + "step": 13682 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013416023014218802, + "loss": 0.5082, + "step": 13683 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001341511676191876, + "loss": 0.5213, + "step": 13684 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013414210477866325, + "loss": 0.5937, + "step": 13685 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001341330416206992, + "loss": 0.5335, + "step": 13686 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001341239781453797, + "loss": 0.4762, + "step": 13687 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001341149143527891, + "loss": 0.5545, + "step": 13688 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001341058502430116, + "loss": 0.5228, + "step": 13689 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001340967858161315, + "loss": 0.5991, + "step": 13690 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013408772107223308, + "loss": 0.5254, + "step": 13691 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013407865601140065, + "loss": 0.5333, + "step": 13692 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001340695906337184, + "loss": 0.4607, + "step": 13693 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001340605249392707, + "loss": 0.6282, + "step": 13694 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013405145892814182, + "loss": 0.5106, + "step": 13695 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013404239260041603, + "loss": 0.6477, + "step": 13696 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013403332595617767, + "loss": 0.5686, + "step": 13697 + }, + { + "epoch": 0.41, + "learning_rate": 0.000134024258995511, + "loss": 0.5279, + "step": 13698 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001340151917185003, + "loss": 0.451, + "step": 13699 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013400612412522991, + "loss": 0.5493, + "step": 13700 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013399705621578418, + "loss": 0.4644, + "step": 13701 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013398798799024736, + "loss": 0.3792, + "step": 13702 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013397891944870373, + "loss": 0.5313, + "step": 13703 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013396985059123767, + "loss": 0.5903, + "step": 13704 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013396078141793348, + "loss": 0.5169, + "step": 13705 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013395171192887546, + "loss": 0.5282, + "step": 13706 + }, + { + "epoch": 0.41, + "learning_rate": 0.000133942642124148, + "loss": 0.4439, + "step": 13707 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013393357200383534, + "loss": 0.5895, + "step": 13708 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001339245015680219, + "loss": 0.4563, + "step": 13709 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001339154308167919, + "loss": 0.5254, + "step": 13710 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013390635975022978, + "loss": 0.5427, + "step": 13711 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013389728836841984, + "loss": 0.5472, + "step": 13712 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001338882166714464, + "loss": 0.5466, + "step": 13713 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013387914465939386, + "loss": 0.5587, + "step": 13714 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013387007233234654, + "loss": 0.6004, + "step": 13715 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013386099969038876, + "loss": 0.4868, + "step": 13716 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001338519267336049, + "loss": 0.6191, + "step": 13717 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013384285346207934, + "loss": 0.4751, + "step": 13718 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001338337798758964, + "loss": 0.5983, + "step": 13719 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013382470597514046, + "loss": 0.6373, + "step": 13720 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001338156317598959, + "loss": 0.6351, + "step": 13721 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013380655723024704, + "loss": 0.4868, + "step": 13722 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013379748238627834, + "loss": 0.5335, + "step": 13723 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013378840722807407, + "loss": 0.4984, + "step": 13724 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013377933175571868, + "loss": 0.5706, + "step": 13725 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001337702559692965, + "loss": 0.6143, + "step": 13726 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013376117986889195, + "loss": 0.4923, + "step": 13727 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013375210345458943, + "loss": 0.5592, + "step": 13728 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001337430267264733, + "loss": 0.558, + "step": 13729 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013373394968462792, + "loss": 0.5463, + "step": 13730 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013372487232913777, + "loss": 0.4295, + "step": 13731 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013371579466008718, + "loss": 0.6102, + "step": 13732 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013370671667756057, + "loss": 0.4794, + "step": 13733 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013369763838164234, + "loss": 0.4385, + "step": 13734 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013368855977241687, + "loss": 0.484, + "step": 13735 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013367948084996864, + "loss": 0.6163, + "step": 13736 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013367040161438198, + "loss": 0.611, + "step": 13737 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001336613220657414, + "loss": 0.5246, + "step": 13738 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013365224220413118, + "loss": 0.6673, + "step": 13739 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001336431620296359, + "loss": 0.6341, + "step": 13740 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001336340815423399, + "loss": 0.5524, + "step": 13741 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001336250007423276, + "loss": 0.5787, + "step": 13742 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013361591962968343, + "loss": 0.5457, + "step": 13743 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013360683820449185, + "loss": 0.4741, + "step": 13744 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001335977564668373, + "loss": 0.6813, + "step": 13745 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001335886744168042, + "loss": 0.4702, + "step": 13746 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013357959205447697, + "loss": 0.521, + "step": 13747 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013357050937994007, + "loss": 0.6098, + "step": 13748 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013356142639327798, + "loss": 0.54, + "step": 13749 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001335523430945751, + "loss": 0.6652, + "step": 13750 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001335432594839159, + "loss": 0.5814, + "step": 13751 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001335341755613849, + "loss": 0.4795, + "step": 13752 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013352509132706644, + "loss": 0.6861, + "step": 13753 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013351600678104507, + "loss": 0.5296, + "step": 13754 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013350692192340522, + "loss": 0.4468, + "step": 13755 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013349783675423136, + "loss": 0.4944, + "step": 13756 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013348875127360796, + "loss": 0.5469, + "step": 13757 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013347966548161953, + "loss": 0.4614, + "step": 13758 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013347057937835048, + "loss": 0.7008, + "step": 13759 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013346149296388534, + "loss": 0.5136, + "step": 13760 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013345240623830856, + "loss": 0.5556, + "step": 13761 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013344331920170466, + "loss": 0.5951, + "step": 13762 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001334342318541581, + "loss": 0.5186, + "step": 13763 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013342514419575338, + "loss": 0.5398, + "step": 13764 + }, + { + "epoch": 0.41, + "learning_rate": 0.000133416056226575, + "loss": 0.5052, + "step": 13765 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001334069679467074, + "loss": 0.5298, + "step": 13766 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001333978793562352, + "loss": 0.6107, + "step": 13767 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013338879045524276, + "loss": 0.521, + "step": 13768 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001333797012438147, + "loss": 0.5342, + "step": 13769 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013337061172203544, + "loss": 0.5029, + "step": 13770 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013336152188998958, + "loss": 0.5595, + "step": 13771 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013335243174776156, + "loss": 0.618, + "step": 13772 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001333433412954359, + "loss": 0.6277, + "step": 13773 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013333425053309718, + "loss": 0.5475, + "step": 13774 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013332515946082988, + "loss": 0.4607, + "step": 13775 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013331606807871852, + "loss": 0.5341, + "step": 13776 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013330697638684765, + "loss": 0.6073, + "step": 13777 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013329788438530178, + "loss": 0.5987, + "step": 13778 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013328879207416544, + "loss": 0.5581, + "step": 13779 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001332796994535232, + "loss": 0.5725, + "step": 13780 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013327060652345958, + "loss": 0.513, + "step": 13781 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013326151328405908, + "loss": 0.5337, + "step": 13782 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001332524197354063, + "loss": 0.6917, + "step": 13783 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001332433258775858, + "loss": 0.5637, + "step": 13784 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013323423171068211, + "loss": 0.6189, + "step": 13785 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013322513723477975, + "loss": 0.57, + "step": 13786 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013321604244996332, + "loss": 0.5441, + "step": 13787 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001332069473563174, + "loss": 0.4895, + "step": 13788 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013319785195392645, + "loss": 0.5559, + "step": 13789 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013318875624287514, + "loss": 0.552, + "step": 13790 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013317966022324801, + "loss": 0.5543, + "step": 13791 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013317056389512962, + "loss": 0.6303, + "step": 13792 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013316146725860455, + "loss": 0.4758, + "step": 13793 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013315237031375735, + "loss": 0.5314, + "step": 13794 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013314327306067263, + "loss": 0.6067, + "step": 13795 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013313417549943497, + "loss": 0.5173, + "step": 13796 + }, + { + "epoch": 0.41, + "learning_rate": 0.000133125077630129, + "loss": 0.6335, + "step": 13797 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001331159794528392, + "loss": 0.4956, + "step": 13798 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013310688096765023, + "loss": 0.4678, + "step": 13799 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001330977821746467, + "loss": 0.5191, + "step": 13800 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013308868307391314, + "loss": 0.634, + "step": 13801 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013307958366553426, + "loss": 0.6254, + "step": 13802 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013307048394959454, + "loss": 0.579, + "step": 13803 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013306138392617863, + "loss": 0.5421, + "step": 13804 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001330522835953712, + "loss": 0.5272, + "step": 13805 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001330431829572568, + "loss": 0.599, + "step": 13806 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013303408201192002, + "loss": 0.5598, + "step": 13807 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013302498075944556, + "loss": 0.5949, + "step": 13808 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013301587919991795, + "loss": 0.5625, + "step": 13809 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013300677733342188, + "loss": 0.4686, + "step": 13810 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013299767516004193, + "loss": 0.4987, + "step": 13811 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013298857267986277, + "loss": 0.497, + "step": 13812 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013297946989296897, + "loss": 0.4413, + "step": 13813 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013297036679944525, + "loss": 0.7032, + "step": 13814 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013296126339937617, + "loss": 0.637, + "step": 13815 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013295215969284641, + "loss": 0.5417, + "step": 13816 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013294305567994062, + "loss": 0.5813, + "step": 13817 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001329339513607434, + "loss": 0.661, + "step": 13818 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013292484673533943, + "loss": 0.5786, + "step": 13819 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013291574180381336, + "loss": 0.5465, + "step": 13820 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013290663656624987, + "loss": 0.5442, + "step": 13821 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013289753102273356, + "loss": 0.618, + "step": 13822 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013288842517334912, + "loss": 0.5227, + "step": 13823 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001328793190181812, + "loss": 0.5554, + "step": 13824 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013287021255731446, + "loss": 0.5088, + "step": 13825 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013286110579083363, + "loss": 0.4396, + "step": 13826 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013285199871882332, + "loss": 0.6253, + "step": 13827 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013284289134136824, + "loss": 0.4899, + "step": 13828 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013283378365855301, + "loss": 0.4838, + "step": 13829 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013282467567046234, + "loss": 0.5373, + "step": 13830 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013281556737718096, + "loss": 0.5147, + "step": 13831 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001328064587787935, + "loss": 0.531, + "step": 13832 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013279734987538467, + "loss": 0.4881, + "step": 13833 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013278824066703912, + "loss": 0.5623, + "step": 13834 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001327791311538416, + "loss": 0.6564, + "step": 13835 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013277002133587677, + "loss": 0.5214, + "step": 13836 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013276091121322935, + "loss": 0.6298, + "step": 13837 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013275180078598404, + "loss": 0.4081, + "step": 13838 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013274269005422552, + "loss": 0.5052, + "step": 13839 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013273357901803854, + "loss": 0.571, + "step": 13840 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013272446767750778, + "loss": 0.6173, + "step": 13841 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013271535603271795, + "loss": 0.4214, + "step": 13842 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013270624408375382, + "loss": 0.5345, + "step": 13843 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001326971318307, + "loss": 0.6622, + "step": 13844 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013268801927364133, + "loss": 0.5023, + "step": 13845 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013267890641266246, + "loss": 0.5999, + "step": 13846 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013266979324784816, + "loss": 0.6244, + "step": 13847 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013266067977928315, + "loss": 0.5427, + "step": 13848 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013265156600705214, + "loss": 0.4877, + "step": 13849 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001326424519312399, + "loss": 0.4777, + "step": 13850 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013263333755193115, + "loss": 0.438, + "step": 13851 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001326242228692106, + "loss": 0.5364, + "step": 13852 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013261510788316306, + "loss": 0.5146, + "step": 13853 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013260599259387325, + "loss": 0.4816, + "step": 13854 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013259687700142592, + "loss": 0.5362, + "step": 13855 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001325877611059058, + "loss": 0.5917, + "step": 13856 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013257864490739767, + "loss": 0.4869, + "step": 13857 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013256952840598625, + "loss": 0.5881, + "step": 13858 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001325604116017564, + "loss": 0.506, + "step": 13859 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013255129449479277, + "loss": 0.457, + "step": 13860 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013254217708518018, + "loss": 0.5055, + "step": 13861 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013253305937300345, + "loss": 0.5216, + "step": 13862 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013252394135834726, + "loss": 0.4809, + "step": 13863 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013251482304129646, + "loss": 0.753, + "step": 13864 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013250570442193573, + "loss": 0.526, + "step": 13865 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013249658550034996, + "loss": 0.7054, + "step": 13866 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001324874662766239, + "loss": 0.5074, + "step": 13867 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001324783467508423, + "loss": 0.6219, + "step": 13868 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013246922692309, + "loss": 0.5524, + "step": 13869 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013246010679345176, + "loss": 0.4168, + "step": 13870 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013245098636201235, + "loss": 0.4679, + "step": 13871 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013244186562885664, + "loss": 0.4496, + "step": 13872 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013243274459406935, + "loss": 0.5342, + "step": 13873 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001324236232577354, + "loss": 0.6288, + "step": 13874 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013241450161993946, + "loss": 0.6478, + "step": 13875 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013240537968076644, + "loss": 0.5804, + "step": 13876 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001323962574403011, + "loss": 0.5451, + "step": 13877 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001323871348986283, + "loss": 0.4914, + "step": 13878 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013237801205583278, + "loss": 0.6243, + "step": 13879 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013236888891199943, + "loss": 0.582, + "step": 13880 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013235976546721308, + "loss": 0.6541, + "step": 13881 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013235064172155848, + "loss": 0.4333, + "step": 13882 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013234151767512053, + "loss": 0.4822, + "step": 13883 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013233239332798404, + "loss": 0.5089, + "step": 13884 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013232326868023382, + "loss": 0.6261, + "step": 13885 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013231414373195477, + "loss": 0.5213, + "step": 13886 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013230501848323164, + "loss": 0.5761, + "step": 13887 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013229589293414938, + "loss": 0.5861, + "step": 13888 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013228676708479275, + "loss": 0.7008, + "step": 13889 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013227764093524664, + "loss": 0.5703, + "step": 13890 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013226851448559586, + "loss": 0.557, + "step": 13891 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001322593877359253, + "loss": 0.4691, + "step": 13892 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013225026068631985, + "loss": 0.5147, + "step": 13893 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001322411333368643, + "loss": 0.4698, + "step": 13894 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013223200568764356, + "loss": 0.5765, + "step": 13895 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013222287773874246, + "loss": 0.6482, + "step": 13896 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001322137494902459, + "loss": 0.4859, + "step": 13897 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001322046209422387, + "loss": 0.6044, + "step": 13898 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001321954920948058, + "loss": 0.5183, + "step": 13899 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013218636294803207, + "loss": 0.5544, + "step": 13900 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013217723350200233, + "loss": 0.5688, + "step": 13901 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013216810375680152, + "loss": 0.5233, + "step": 13902 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013215897371251446, + "loss": 0.4586, + "step": 13903 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001321498433692261, + "loss": 0.5369, + "step": 13904 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013214071272702133, + "loss": 0.6074, + "step": 13905 + }, + { + "epoch": 0.41, + "learning_rate": 0.000132131581785985, + "loss": 0.5319, + "step": 13906 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013212245054620208, + "loss": 0.6412, + "step": 13907 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013211331900775733, + "loss": 0.4368, + "step": 13908 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001321041871707358, + "loss": 0.5919, + "step": 13909 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013209505503522233, + "loss": 0.5838, + "step": 13910 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013208592260130183, + "loss": 0.6055, + "step": 13911 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013207678986905917, + "loss": 0.3878, + "step": 13912 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013206765683857933, + "loss": 0.4822, + "step": 13913 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001320585235099472, + "loss": 0.5252, + "step": 13914 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001320493898832477, + "loss": 0.6242, + "step": 13915 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013204025595856575, + "loss": 0.498, + "step": 13916 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013203112173598626, + "loss": 0.5676, + "step": 13917 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013202198721559415, + "loss": 0.5202, + "step": 13918 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013201285239747442, + "loss": 0.4705, + "step": 13919 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001320037172817119, + "loss": 0.5556, + "step": 13920 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013199458186839157, + "loss": 0.7016, + "step": 13921 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013198544615759838, + "loss": 0.6133, + "step": 13922 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013197631014941726, + "loss": 0.6267, + "step": 13923 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013196717384393315, + "loss": 0.5327, + "step": 13924 + }, + { + "epoch": 0.41, + "learning_rate": 0.000131958037241231, + "loss": 0.4658, + "step": 13925 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001319489003413958, + "loss": 0.5495, + "step": 13926 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001319397631445124, + "loss": 0.5078, + "step": 13927 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013193062565066582, + "loss": 0.4105, + "step": 13928 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013192148785994102, + "loss": 0.6653, + "step": 13929 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013191234977242295, + "loss": 0.5978, + "step": 13930 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013190321138819658, + "loss": 0.5687, + "step": 13931 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013189407270734687, + "loss": 0.635, + "step": 13932 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013188493372995875, + "loss": 0.5808, + "step": 13933 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013187579445611723, + "loss": 0.5285, + "step": 13934 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001318666548859073, + "loss": 0.553, + "step": 13935 + }, + { + "epoch": 0.41, + "learning_rate": 0.0001318575150194139, + "loss": 0.5683, + "step": 13936 + }, + { + "epoch": 0.41, + "learning_rate": 0.000131848374856722, + "loss": 0.6, + "step": 13937 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013183923439791668, + "loss": 0.5705, + "step": 13938 + }, + { + "epoch": 0.41, + "learning_rate": 0.00013183009364308275, + "loss": 0.4866, + "step": 13939 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013182095259230535, + "loss": 0.5462, + "step": 13940 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013181181124566943, + "loss": 0.592, + "step": 13941 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013180266960325992, + "loss": 0.517, + "step": 13942 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013179352766516192, + "loss": 0.5448, + "step": 13943 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013178438543146035, + "loss": 0.5094, + "step": 13944 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013177524290224022, + "loss": 0.551, + "step": 13945 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013176610007758653, + "loss": 0.4492, + "step": 13946 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013175695695758435, + "loss": 0.5846, + "step": 13947 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001317478135423186, + "loss": 0.6764, + "step": 13948 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013173866983187433, + "loss": 0.5686, + "step": 13949 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001317295258263366, + "loss": 0.5436, + "step": 13950 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013172038152579034, + "loss": 0.4656, + "step": 13951 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013171123693032065, + "loss": 0.5172, + "step": 13952 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001317020920400125, + "loss": 0.6125, + "step": 13953 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013169294685495093, + "loss": 0.5532, + "step": 13954 + }, + { + "epoch": 0.42, + "learning_rate": 0.000131683801375221, + "loss": 0.4749, + "step": 13955 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013167465560090768, + "loss": 0.5238, + "step": 13956 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013166550953209603, + "loss": 0.5825, + "step": 13957 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013165636316887113, + "loss": 0.5517, + "step": 13958 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013164721651131797, + "loss": 0.5552, + "step": 13959 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013163806955952158, + "loss": 0.5951, + "step": 13960 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013162892231356703, + "loss": 0.5459, + "step": 13961 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001316197747735394, + "loss": 0.6068, + "step": 13962 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013161062693952367, + "loss": 0.6476, + "step": 13963 + }, + { + "epoch": 0.42, + "learning_rate": 0.000131601478811605, + "loss": 0.5959, + "step": 13964 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013159233038986828, + "loss": 0.581, + "step": 13965 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013158318167439873, + "loss": 0.5185, + "step": 13966 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001315740326652813, + "loss": 0.5454, + "step": 13967 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013156488336260114, + "loss": 0.6254, + "step": 13968 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013155573376644325, + "loss": 0.5644, + "step": 13969 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001315465838768927, + "loss": 0.4192, + "step": 13970 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013153743369403464, + "loss": 0.5182, + "step": 13971 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013152828321795409, + "loss": 0.5995, + "step": 13972 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001315191324487361, + "loss": 0.5521, + "step": 13973 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013150998138646575, + "loss": 0.6692, + "step": 13974 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013150083003122818, + "loss": 0.5478, + "step": 13975 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013149167838310847, + "loss": 0.6147, + "step": 13976 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001314825264421917, + "loss": 0.5303, + "step": 13977 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001314733742085629, + "loss": 0.5853, + "step": 13978 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013146422168230722, + "loss": 0.5084, + "step": 13979 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013145506886350972, + "loss": 0.505, + "step": 13980 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001314459157522556, + "loss": 0.692, + "step": 13981 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013143676234862983, + "loss": 0.6348, + "step": 13982 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013142760865271762, + "loss": 0.7346, + "step": 13983 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013141845466460396, + "loss": 0.5073, + "step": 13984 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001314093003843741, + "loss": 0.6639, + "step": 13985 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013140014581211303, + "loss": 0.5335, + "step": 13986 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013139099094790595, + "loss": 0.5471, + "step": 13987 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013138183579183797, + "loss": 0.535, + "step": 13988 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013137268034399416, + "loss": 0.6105, + "step": 13989 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013136352460445965, + "loss": 0.5525, + "step": 13990 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013135436857331963, + "loss": 0.6285, + "step": 13991 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013134521225065916, + "loss": 0.6382, + "step": 13992 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001313360556365634, + "loss": 0.5961, + "step": 13993 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001313268987311175, + "loss": 0.5068, + "step": 13994 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001313177415344066, + "loss": 0.653, + "step": 13995 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013130858404651578, + "loss": 0.5946, + "step": 13996 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013129942626753023, + "loss": 0.536, + "step": 13997 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013129026819753512, + "loss": 0.6728, + "step": 13998 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013128110983661552, + "loss": 0.6369, + "step": 13999 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013127195118485668, + "loss": 0.5733, + "step": 14000 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013126279224234366, + "loss": 0.6451, + "step": 14001 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013125363300916167, + "loss": 0.5265, + "step": 14002 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013124447348539586, + "loss": 0.518, + "step": 14003 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013123531367113139, + "loss": 0.5028, + "step": 14004 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013122615356645344, + "loss": 0.5872, + "step": 14005 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013121699317144713, + "loss": 0.7577, + "step": 14006 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013120783248619765, + "loss": 0.612, + "step": 14007 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001311986715107902, + "loss": 0.5129, + "step": 14008 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013118951024530993, + "loss": 0.489, + "step": 14009 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013118034868984206, + "loss": 0.5521, + "step": 14010 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001311711868444717, + "loss": 0.603, + "step": 14011 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001311620247092841, + "loss": 0.5548, + "step": 14012 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013115286228436437, + "loss": 0.5478, + "step": 14013 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013114369956979775, + "loss": 0.5951, + "step": 14014 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013113453656566945, + "loss": 0.5531, + "step": 14015 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013112537327206463, + "loss": 0.5607, + "step": 14016 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001311162096890685, + "loss": 0.5667, + "step": 14017 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013110704581676622, + "loss": 0.6152, + "step": 14018 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013109788165524306, + "loss": 0.5528, + "step": 14019 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013108871720458418, + "loss": 0.4916, + "step": 14020 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013107955246487483, + "loss": 0.6405, + "step": 14021 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013107038743620012, + "loss": 0.6275, + "step": 14022 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013106122211864536, + "loss": 0.5276, + "step": 14023 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013105205651229578, + "loss": 0.5111, + "step": 14024 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001310428906172365, + "loss": 0.5423, + "step": 14025 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013103372443355285, + "loss": 0.6162, + "step": 14026 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013102455796132993, + "loss": 0.5198, + "step": 14027 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013101539120065307, + "loss": 0.631, + "step": 14028 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013100622415160748, + "loss": 0.5681, + "step": 14029 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013099705681427835, + "loss": 0.5764, + "step": 14030 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013098788918875095, + "loss": 0.5537, + "step": 14031 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001309787212751105, + "loss": 0.5772, + "step": 14032 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013096955307344225, + "loss": 0.6061, + "step": 14033 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013096038458383148, + "loss": 0.6621, + "step": 14034 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013095121580636334, + "loss": 0.5222, + "step": 14035 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013094204674112315, + "loss": 0.4568, + "step": 14036 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013093287738819614, + "loss": 0.539, + "step": 14037 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013092370774766757, + "loss": 0.633, + "step": 14038 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013091453781962267, + "loss": 0.6528, + "step": 14039 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001309053676041467, + "loss": 0.5259, + "step": 14040 + }, + { + "epoch": 0.42, + "learning_rate": 0.000130896197101325, + "loss": 0.5281, + "step": 14041 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013088702631124273, + "loss": 0.5825, + "step": 14042 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013087785523398525, + "loss": 0.5984, + "step": 14043 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001308686838696377, + "loss": 0.5049, + "step": 14044 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001308595122182855, + "loss": 0.4863, + "step": 14045 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001308503402800138, + "loss": 0.4846, + "step": 14046 + }, + { + "epoch": 0.42, + "learning_rate": 0.000130841168054908, + "loss": 0.5069, + "step": 14047 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013083199554305325, + "loss": 0.5699, + "step": 14048 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001308228227445349, + "loss": 0.612, + "step": 14049 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013081364965943827, + "loss": 0.5908, + "step": 14050 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001308044762878486, + "loss": 0.5668, + "step": 14051 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013079530262985115, + "loss": 0.5761, + "step": 14052 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013078612868553128, + "loss": 0.5945, + "step": 14053 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001307769544549742, + "loss": 0.6286, + "step": 14054 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013076777993826534, + "loss": 0.5372, + "step": 14055 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013075860513548986, + "loss": 0.531, + "step": 14056 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013074943004673321, + "loss": 0.5388, + "step": 14057 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013074025467208053, + "loss": 0.5042, + "step": 14058 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013073107901161727, + "loss": 0.6106, + "step": 14059 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013072190306542865, + "loss": 0.5268, + "step": 14060 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013071272683360004, + "loss": 0.5025, + "step": 14061 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013070355031621678, + "loss": 0.438, + "step": 14062 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013069437351336406, + "loss": 0.6473, + "step": 14063 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013068519642512734, + "loss": 0.6127, + "step": 14064 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013067601905159188, + "loss": 0.5894, + "step": 14065 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013066684139284302, + "loss": 0.531, + "step": 14066 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001306576634489661, + "loss": 0.6092, + "step": 14067 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013064848522004643, + "loss": 0.5424, + "step": 14068 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013063930670616938, + "loss": 0.537, + "step": 14069 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013063012790742024, + "loss": 0.5485, + "step": 14070 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001306209488238844, + "loss": 0.4981, + "step": 14071 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001306117694556472, + "loss": 0.6135, + "step": 14072 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001306025898027939, + "loss": 0.4856, + "step": 14073 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013059340986541, + "loss": 0.5859, + "step": 14074 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001305842296435807, + "loss": 0.6115, + "step": 14075 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013057504913739144, + "loss": 0.4871, + "step": 14076 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001305658683469276, + "loss": 0.5621, + "step": 14077 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013055668727227445, + "loss": 0.5444, + "step": 14078 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013054750591351744, + "loss": 0.5419, + "step": 14079 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013053832427074185, + "loss": 0.597, + "step": 14080 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013052914234403312, + "loss": 0.5377, + "step": 14081 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001305199601334766, + "loss": 0.6513, + "step": 14082 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013051077763915765, + "loss": 0.5028, + "step": 14083 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013050159486116165, + "loss": 0.5735, + "step": 14084 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013049241179957397, + "loss": 0.624, + "step": 14085 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013048322845448, + "loss": 0.5798, + "step": 14086 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001304740448259651, + "loss": 0.5213, + "step": 14087 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001304648609141147, + "loss": 0.5831, + "step": 14088 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013045567671901418, + "loss": 0.5091, + "step": 14089 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013044649224074888, + "loss": 0.5268, + "step": 14090 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013043730747940425, + "loss": 0.4835, + "step": 14091 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013042812243506567, + "loss": 0.5164, + "step": 14092 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013041893710781857, + "loss": 0.6244, + "step": 14093 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013040975149774826, + "loss": 0.6239, + "step": 14094 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013040056560494021, + "loss": 0.6796, + "step": 14095 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013039137942947983, + "loss": 0.6949, + "step": 14096 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001303821929714525, + "loss": 0.4262, + "step": 14097 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001303730062309437, + "loss": 0.6126, + "step": 14098 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013036381920803874, + "loss": 0.4254, + "step": 14099 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013035463190282312, + "loss": 0.42, + "step": 14100 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001303454443153822, + "loss": 0.5237, + "step": 14101 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013033625644580146, + "loss": 0.6023, + "step": 14102 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001303270682941663, + "loss": 0.4848, + "step": 14103 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001303178798605621, + "loss": 0.6111, + "step": 14104 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001303086911450744, + "loss": 0.5565, + "step": 14105 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013029950214778853, + "loss": 0.7013, + "step": 14106 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013029031286878996, + "loss": 0.5945, + "step": 14107 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013028112330816416, + "loss": 0.5484, + "step": 14108 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013027193346599653, + "loss": 0.6211, + "step": 14109 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013026274334237254, + "loss": 0.6089, + "step": 14110 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013025355293737762, + "loss": 0.5315, + "step": 14111 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001302443622510972, + "loss": 0.5025, + "step": 14112 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001302351712836168, + "loss": 0.6692, + "step": 14113 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001302259800350218, + "loss": 0.4052, + "step": 14114 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013021678850539767, + "loss": 0.5368, + "step": 14115 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013020759669482988, + "loss": 0.6125, + "step": 14116 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013019840460340394, + "loss": 0.4494, + "step": 14117 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013018921223120522, + "loss": 0.5292, + "step": 14118 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001301800195783193, + "loss": 0.5758, + "step": 14119 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001301708266448315, + "loss": 0.6489, + "step": 14120 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013016163343082746, + "loss": 0.5304, + "step": 14121 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013015243993639252, + "loss": 0.5503, + "step": 14122 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013014324616161224, + "loss": 0.4848, + "step": 14123 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013013405210657205, + "loss": 0.5912, + "step": 14124 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013012485777135749, + "loss": 0.5017, + "step": 14125 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013011566315605398, + "loss": 0.5631, + "step": 14126 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013010646826074705, + "loss": 0.6354, + "step": 14127 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013009727308552216, + "loss": 0.5928, + "step": 14128 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013008807763046485, + "loss": 0.4476, + "step": 14129 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013007888189566055, + "loss": 0.581, + "step": 14130 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013006968588119482, + "loss": 0.5159, + "step": 14131 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001300604895871531, + "loss": 0.5236, + "step": 14132 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013005129301362097, + "loss": 0.6248, + "step": 14133 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013004209616068388, + "loss": 0.6339, + "step": 14134 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013003289902842734, + "loss": 0.4554, + "step": 14135 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001300237016169369, + "loss": 0.6264, + "step": 14136 + }, + { + "epoch": 0.42, + "learning_rate": 0.000130014503926298, + "loss": 0.5208, + "step": 14137 + }, + { + "epoch": 0.42, + "learning_rate": 0.00013000530595659625, + "loss": 0.5799, + "step": 14138 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012999610770791714, + "loss": 0.4871, + "step": 14139 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012998690918034615, + "loss": 0.5293, + "step": 14140 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012997771037396883, + "loss": 0.6599, + "step": 14141 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001299685112888707, + "loss": 0.4958, + "step": 14142 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001299593119251373, + "loss": 0.5417, + "step": 14143 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001299501122828542, + "loss": 0.5622, + "step": 14144 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001299409123621068, + "loss": 0.5575, + "step": 14145 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001299317121629808, + "loss": 0.5854, + "step": 14146 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012992251168556166, + "loss": 0.5894, + "step": 14147 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012991331092993496, + "loss": 0.5303, + "step": 14148 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012990410989618617, + "loss": 0.57, + "step": 14149 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012989490858440089, + "loss": 0.4868, + "step": 14150 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012988570699466466, + "loss": 0.5015, + "step": 14151 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012987650512706308, + "loss": 0.4904, + "step": 14152 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001298673029816816, + "loss": 0.5195, + "step": 14153 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001298581005586059, + "loss": 0.6162, + "step": 14154 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012984889785792146, + "loss": 0.479, + "step": 14155 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012983969487971386, + "loss": 0.694, + "step": 14156 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012983049162406865, + "loss": 0.619, + "step": 14157 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012982128809107147, + "loss": 0.5982, + "step": 14158 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001298120842808078, + "loss": 0.4559, + "step": 14159 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012980288019336325, + "loss": 0.4813, + "step": 14160 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012979367582882341, + "loss": 0.574, + "step": 14161 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012978447118727385, + "loss": 0.6296, + "step": 14162 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012977526626880014, + "loss": 0.5737, + "step": 14163 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012976606107348786, + "loss": 0.6018, + "step": 14164 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012975685560142266, + "loss": 0.5563, + "step": 14165 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012974764985269001, + "loss": 0.4379, + "step": 14166 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012973844382737563, + "loss": 0.6789, + "step": 14167 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012972923752556498, + "loss": 0.5357, + "step": 14168 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012972003094734378, + "loss": 0.5761, + "step": 14169 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012971082409279759, + "loss": 0.4543, + "step": 14170 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012970161696201194, + "loss": 0.5449, + "step": 14171 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012969240955507255, + "loss": 0.5167, + "step": 14172 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012968320187206492, + "loss": 0.6347, + "step": 14173 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012967399391307474, + "loss": 0.4431, + "step": 14174 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001296647856781876, + "loss": 0.5411, + "step": 14175 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012965557716748907, + "loss": 0.4743, + "step": 14176 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012964636838106488, + "loss": 0.6257, + "step": 14177 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001296371593190005, + "loss": 0.5493, + "step": 14178 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012962794998138167, + "loss": 0.4673, + "step": 14179 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001296187403682939, + "loss": 0.4754, + "step": 14180 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012960953047982295, + "loss": 0.5991, + "step": 14181 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012960032031605437, + "loss": 0.5484, + "step": 14182 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001295911098770738, + "loss": 0.5518, + "step": 14183 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001295818991629669, + "loss": 0.5569, + "step": 14184 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012957268817381928, + "loss": 0.5466, + "step": 14185 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001295634769097166, + "loss": 0.5718, + "step": 14186 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001295542653707445, + "loss": 0.6079, + "step": 14187 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001295450535569886, + "loss": 0.5197, + "step": 14188 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001295358414685346, + "loss": 0.6123, + "step": 14189 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012952662910546806, + "loss": 0.6505, + "step": 14190 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012951741646787475, + "loss": 0.5421, + "step": 14191 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012950820355584023, + "loss": 0.4593, + "step": 14192 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001294989903694502, + "loss": 0.5503, + "step": 14193 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012948977690879033, + "loss": 0.6545, + "step": 14194 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012948056317394627, + "loss": 0.5967, + "step": 14195 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012947134916500367, + "loss": 0.6186, + "step": 14196 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012946213488204819, + "loss": 0.5827, + "step": 14197 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012945292032516558, + "loss": 0.5455, + "step": 14198 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001294437054944414, + "loss": 0.5376, + "step": 14199 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001294344903899614, + "loss": 0.5754, + "step": 14200 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012942527501181124, + "loss": 0.5862, + "step": 14201 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012941605936007662, + "loss": 0.4151, + "step": 14202 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012940684343484317, + "loss": 0.584, + "step": 14203 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012939762723619664, + "loss": 0.5408, + "step": 14204 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001293884107642227, + "loss": 0.6148, + "step": 14205 + }, + { + "epoch": 0.42, + "learning_rate": 0.000129379194019007, + "loss": 0.5196, + "step": 14206 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012936997700063523, + "loss": 0.5544, + "step": 14207 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001293607597091932, + "loss": 0.5091, + "step": 14208 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012935154214476646, + "loss": 0.5008, + "step": 14209 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012934232430744083, + "loss": 0.5973, + "step": 14210 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012933310619730193, + "loss": 0.5858, + "step": 14211 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001293238878144355, + "loss": 0.5335, + "step": 14212 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012931466915892727, + "loss": 0.6181, + "step": 14213 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012930545023086292, + "loss": 0.6094, + "step": 14214 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012929623103032816, + "loss": 0.4757, + "step": 14215 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012928701155740874, + "loss": 0.5514, + "step": 14216 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012927779181219035, + "loss": 0.5675, + "step": 14217 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012926857179475874, + "loss": 0.5826, + "step": 14218 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001292593515051996, + "loss": 0.5658, + "step": 14219 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012925013094359865, + "loss": 0.6813, + "step": 14220 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012924091011004164, + "loss": 0.5071, + "step": 14221 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012923168900461435, + "loss": 0.5083, + "step": 14222 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001292224676274024, + "loss": 0.5104, + "step": 14223 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012921324597849168, + "loss": 0.466, + "step": 14224 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012920402405796775, + "loss": 0.6396, + "step": 14225 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001291948018659165, + "loss": 0.5552, + "step": 14226 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001291855794024236, + "loss": 0.5642, + "step": 14227 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001291763566675748, + "loss": 0.657, + "step": 14228 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012916713366145592, + "loss": 0.5075, + "step": 14229 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001291579103841526, + "loss": 0.5399, + "step": 14230 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012914868683575067, + "loss": 0.6612, + "step": 14231 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012913946301633587, + "loss": 0.5727, + "step": 14232 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012913023892599396, + "loss": 0.5544, + "step": 14233 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001291210145648107, + "loss": 0.5622, + "step": 14234 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012911178993287186, + "loss": 0.5988, + "step": 14235 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012910256503026316, + "loss": 0.4961, + "step": 14236 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012909333985707043, + "loss": 0.6012, + "step": 14237 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012908411441337942, + "loss": 0.5682, + "step": 14238 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012907488869927594, + "loss": 0.4799, + "step": 14239 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012906566271484569, + "loss": 0.4567, + "step": 14240 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001290564364601745, + "loss": 0.5769, + "step": 14241 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012904720993534816, + "loss": 0.5654, + "step": 14242 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012903798314045244, + "loss": 0.5644, + "step": 14243 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001290287560755731, + "loss": 0.5589, + "step": 14244 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012901952874079598, + "loss": 0.5064, + "step": 14245 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012901030113620686, + "loss": 0.5042, + "step": 14246 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012900107326189151, + "loss": 0.4689, + "step": 14247 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012899184511793571, + "loss": 0.5358, + "step": 14248 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012898261670442533, + "loss": 0.5095, + "step": 14249 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012897338802144608, + "loss": 0.6032, + "step": 14250 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012896415906908388, + "loss": 0.4293, + "step": 14251 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001289549298474244, + "loss": 0.5157, + "step": 14252 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001289457003565536, + "loss": 0.5926, + "step": 14253 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012893647059655715, + "loss": 0.6429, + "step": 14254 + }, + { + "epoch": 0.42, + "learning_rate": 0.000128927240567521, + "loss": 0.5528, + "step": 14255 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012891801026953083, + "loss": 0.5135, + "step": 14256 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012890877970267255, + "loss": 0.5977, + "step": 14257 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012889954886703198, + "loss": 0.5382, + "step": 14258 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001288903177626949, + "loss": 0.5467, + "step": 14259 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012888108638974715, + "loss": 0.5276, + "step": 14260 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001288718547482746, + "loss": 0.6052, + "step": 14261 + }, + { + "epoch": 0.42, + "learning_rate": 0.000128862622838363, + "loss": 0.4976, + "step": 14262 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001288533906600983, + "loss": 0.5144, + "step": 14263 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012884415821356625, + "loss": 0.5761, + "step": 14264 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001288349254988527, + "loss": 0.492, + "step": 14265 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001288256925160435, + "loss": 0.4596, + "step": 14266 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012881645926522453, + "loss": 0.6219, + "step": 14267 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012880722574648158, + "loss": 0.5933, + "step": 14268 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012879799195990053, + "loss": 0.5425, + "step": 14269 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012878875790556723, + "loss": 0.5354, + "step": 14270 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012877952358356754, + "loss": 0.5028, + "step": 14271 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012877028899398733, + "loss": 0.4982, + "step": 14272 + }, + { + "epoch": 0.42, + "learning_rate": 0.0001287610541369124, + "loss": 0.5889, + "step": 14273 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012875181901242865, + "loss": 0.6212, + "step": 14274 + }, + { + "epoch": 0.42, + "learning_rate": 0.00012874258362062196, + "loss": 0.471, + "step": 14275 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012873334796157821, + "loss": 0.6437, + "step": 14276 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001287241120353832, + "loss": 0.5648, + "step": 14277 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012871487584212288, + "loss": 0.4948, + "step": 14278 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012870563938188306, + "loss": 0.6178, + "step": 14279 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012869640265474966, + "loss": 0.5289, + "step": 14280 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012868716566080855, + "loss": 0.5519, + "step": 14281 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001286779284001456, + "loss": 0.6909, + "step": 14282 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001286686908728467, + "loss": 0.6023, + "step": 14283 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012865945307899776, + "loss": 0.6331, + "step": 14284 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012865021501868462, + "loss": 0.5137, + "step": 14285 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012864097669199325, + "loss": 0.5218, + "step": 14286 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001286317380990094, + "loss": 0.5121, + "step": 14287 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012862249923981914, + "loss": 0.4752, + "step": 14288 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012861326011450824, + "loss": 0.5803, + "step": 14289 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012860402072316268, + "loss": 0.5933, + "step": 14290 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012859478106586836, + "loss": 0.6092, + "step": 14291 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012858554114271114, + "loss": 0.6375, + "step": 14292 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012857630095377694, + "loss": 0.62, + "step": 14293 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012856706049915167, + "loss": 0.4724, + "step": 14294 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012855781977892127, + "loss": 0.4757, + "step": 14295 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012854857879317165, + "loss": 0.532, + "step": 14296 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012853933754198871, + "loss": 0.5166, + "step": 14297 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012853009602545844, + "loss": 0.5745, + "step": 14298 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012852085424366663, + "loss": 0.6656, + "step": 14299 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001285116121966993, + "loss": 0.4576, + "step": 14300 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001285023698846424, + "loss": 0.4273, + "step": 14301 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001284931273075818, + "loss": 0.539, + "step": 14302 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012848388446560352, + "loss": 0.5316, + "step": 14303 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012847464135879333, + "loss": 0.5508, + "step": 14304 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012846539798723735, + "loss": 0.6254, + "step": 14305 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001284561543510214, + "loss": 0.6045, + "step": 14306 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001284469104502315, + "loss": 0.5632, + "step": 14307 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012843766628495355, + "loss": 0.5056, + "step": 14308 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012842842185527352, + "loss": 0.5606, + "step": 14309 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012841917716127733, + "loss": 0.6184, + "step": 14310 + }, + { + "epoch": 0.43, + "learning_rate": 0.000128409932203051, + "loss": 0.4922, + "step": 14311 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001284006869806804, + "loss": 0.4932, + "step": 14312 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012839144149425158, + "loss": 0.4104, + "step": 14313 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001283821957438504, + "loss": 0.5747, + "step": 14314 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012837294972956292, + "loss": 0.4956, + "step": 14315 + }, + { + "epoch": 0.43, + "learning_rate": 0.000128363703451475, + "loss": 0.4558, + "step": 14316 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001283544569096727, + "loss": 0.5586, + "step": 14317 + }, + { + "epoch": 0.43, + "learning_rate": 0.000128345210104242, + "loss": 0.5818, + "step": 14318 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001283359630352688, + "loss": 0.4949, + "step": 14319 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012832671570283912, + "loss": 0.6168, + "step": 14320 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012831746810703894, + "loss": 0.5467, + "step": 14321 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001283082202479542, + "loss": 0.5584, + "step": 14322 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012829897212567092, + "loss": 0.5881, + "step": 14323 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012828972374027509, + "loss": 0.5448, + "step": 14324 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012828047509185267, + "loss": 0.4499, + "step": 14325 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012827122618048965, + "loss": 0.5369, + "step": 14326 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012826197700627207, + "loss": 0.5214, + "step": 14327 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012825272756928588, + "loss": 0.5698, + "step": 14328 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012824347786961709, + "loss": 0.483, + "step": 14329 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001282342279073517, + "loss": 0.5013, + "step": 14330 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012822497768257566, + "loss": 0.5772, + "step": 14331 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001282157271953751, + "loss": 0.5584, + "step": 14332 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001282064764458359, + "loss": 0.5888, + "step": 14333 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001281972254340442, + "loss": 0.69, + "step": 14334 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012818797416008586, + "loss": 0.6328, + "step": 14335 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012817872262404698, + "loss": 0.7574, + "step": 14336 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001281694708260136, + "loss": 0.6035, + "step": 14337 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012816021876607168, + "loss": 0.6386, + "step": 14338 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012815096644430728, + "loss": 0.5887, + "step": 14339 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012814171386080642, + "loss": 0.5598, + "step": 14340 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012813246101565513, + "loss": 0.5222, + "step": 14341 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012812320790893942, + "loss": 0.6012, + "step": 14342 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001281139545407453, + "loss": 0.5649, + "step": 14343 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001281047009111589, + "loss": 0.5528, + "step": 14344 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012809544702026613, + "loss": 0.549, + "step": 14345 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012808619286815314, + "loss": 0.412, + "step": 14346 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012807693845490588, + "loss": 0.6231, + "step": 14347 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012806768378061044, + "loss": 0.5602, + "step": 14348 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012805842884535288, + "loss": 0.6026, + "step": 14349 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012804917364921925, + "loss": 0.5595, + "step": 14350 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012803991819229554, + "loss": 0.5044, + "step": 14351 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012803066247466788, + "loss": 0.4478, + "step": 14352 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001280214064964223, + "loss": 0.5629, + "step": 14353 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012801215025764483, + "loss": 0.4949, + "step": 14354 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012800289375842157, + "loss": 0.5656, + "step": 14355 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012799363699883854, + "loss": 0.465, + "step": 14356 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012798437997898186, + "loss": 0.6225, + "step": 14357 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012797512269893752, + "loss": 0.6258, + "step": 14358 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012796586515879168, + "loss": 0.5402, + "step": 14359 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012795660735863035, + "loss": 0.4701, + "step": 14360 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012794734929853964, + "loss": 0.5386, + "step": 14361 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001279380909786056, + "loss": 0.5508, + "step": 14362 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012792883239891436, + "loss": 0.418, + "step": 14363 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012791957355955192, + "loss": 0.6609, + "step": 14364 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012791031446060446, + "loss": 0.5845, + "step": 14365 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012790105510215798, + "loss": 0.5652, + "step": 14366 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012789179548429864, + "loss": 0.551, + "step": 14367 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001278825356071125, + "loss": 0.545, + "step": 14368 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012787327547068562, + "loss": 0.4782, + "step": 14369 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012786401507510416, + "loss": 0.6823, + "step": 14370 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001278547544204542, + "loss": 0.642, + "step": 14371 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012784549350682183, + "loss": 0.5462, + "step": 14372 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012783623233429315, + "loss": 0.6514, + "step": 14373 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012782697090295424, + "loss": 0.5343, + "step": 14374 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012781770921289134, + "loss": 0.5883, + "step": 14375 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001278084472641904, + "loss": 0.4877, + "step": 14376 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012779918505693763, + "loss": 0.7439, + "step": 14377 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012778992259121908, + "loss": 0.5604, + "step": 14378 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012778065986712093, + "loss": 0.5895, + "step": 14379 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001277713968847293, + "loss": 0.5156, + "step": 14380 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012776213364413025, + "loss": 0.4711, + "step": 14381 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012775287014540998, + "loss": 0.5592, + "step": 14382 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012774360638865458, + "loss": 0.6032, + "step": 14383 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012773434237395013, + "loss": 0.6328, + "step": 14384 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012772507810138287, + "loss": 0.524, + "step": 14385 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012771581357103889, + "loss": 0.6044, + "step": 14386 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001277065487830043, + "loss": 0.6433, + "step": 14387 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012769728373736525, + "loss": 0.5195, + "step": 14388 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012768801843420795, + "loss": 0.5479, + "step": 14389 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012767875287361845, + "loss": 0.5347, + "step": 14390 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012766948705568294, + "loss": 0.5191, + "step": 14391 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012766022098048758, + "loss": 0.4661, + "step": 14392 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012765095464811846, + "loss": 0.5316, + "step": 14393 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012764168805866186, + "loss": 0.523, + "step": 14394 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001276324212122038, + "loss": 0.5945, + "step": 14395 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012762315410883054, + "loss": 0.4498, + "step": 14396 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012761388674862817, + "loss": 0.4431, + "step": 14397 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001276046191316829, + "loss": 0.5568, + "step": 14398 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012759535125808088, + "loss": 0.4932, + "step": 14399 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012758608312790832, + "loss": 0.5965, + "step": 14400 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001275768147412513, + "loss": 0.4815, + "step": 14401 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012756754609819608, + "loss": 0.5213, + "step": 14402 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001275582771988288, + "loss": 0.6026, + "step": 14403 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012754900804323564, + "loss": 0.5356, + "step": 14404 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012753973863150275, + "loss": 0.7279, + "step": 14405 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001275304689637164, + "loss": 0.561, + "step": 14406 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001275211990399627, + "loss": 0.497, + "step": 14407 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012751192886032787, + "loss": 0.6222, + "step": 14408 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012750265842489806, + "loss": 0.6259, + "step": 14409 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012749338773375952, + "loss": 0.5824, + "step": 14410 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012748411678699842, + "loss": 0.5375, + "step": 14411 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012747484558470093, + "loss": 0.7324, + "step": 14412 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001274655741269533, + "loss": 0.5932, + "step": 14413 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001274563024138417, + "loss": 0.6051, + "step": 14414 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012744703044545235, + "loss": 0.5358, + "step": 14415 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012743775822187146, + "loss": 0.5808, + "step": 14416 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001274284857431852, + "loss": 0.5274, + "step": 14417 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012741921300947983, + "loss": 0.6233, + "step": 14418 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012740994002084153, + "loss": 0.606, + "step": 14419 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012740066677735656, + "loss": 0.5198, + "step": 14420 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001273913932791111, + "loss": 0.5647, + "step": 14421 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001273821195261914, + "loss": 0.6024, + "step": 14422 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012737284551868363, + "loss": 0.6566, + "step": 14423 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012736357125667404, + "loss": 0.4993, + "step": 14424 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012735429674024892, + "loss": 0.6371, + "step": 14425 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001273450219694944, + "loss": 0.6274, + "step": 14426 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012733574694449683, + "loss": 0.6324, + "step": 14427 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012732647166534232, + "loss": 0.4575, + "step": 14428 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001273171961321172, + "loss": 0.6396, + "step": 14429 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012730792034490766, + "loss": 0.597, + "step": 14430 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012729864430379996, + "loss": 0.5505, + "step": 14431 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001272893680088804, + "loss": 0.6176, + "step": 14432 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001272800914602351, + "loss": 0.5798, + "step": 14433 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001272708146579504, + "loss": 0.5778, + "step": 14434 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012726153760211255, + "loss": 0.6049, + "step": 14435 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012725226029280777, + "loss": 0.7029, + "step": 14436 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012724298273012236, + "loss": 0.533, + "step": 14437 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012723370491414253, + "loss": 0.7645, + "step": 14438 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001272244268449546, + "loss": 0.5846, + "step": 14439 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012721514852264475, + "loss": 0.5853, + "step": 14440 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012720586994729934, + "loss": 0.5293, + "step": 14441 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012719659111900457, + "loss": 0.6821, + "step": 14442 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012718731203784675, + "loss": 0.5045, + "step": 14443 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012717803270391212, + "loss": 0.4803, + "step": 14444 + }, + { + "epoch": 0.43, + "learning_rate": 0.000127168753117287, + "loss": 0.559, + "step": 14445 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012715947327805765, + "loss": 0.5402, + "step": 14446 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001271501931863103, + "loss": 0.5317, + "step": 14447 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001271409128421313, + "loss": 0.6169, + "step": 14448 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012713163224560692, + "loss": 0.5543, + "step": 14449 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001271223513968234, + "loss": 0.6271, + "step": 14450 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001271130702958671, + "loss": 0.6438, + "step": 14451 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012710378894282423, + "loss": 0.4452, + "step": 14452 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001270945073377812, + "loss": 0.5914, + "step": 14453 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001270852254808242, + "loss": 0.5161, + "step": 14454 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001270759433720396, + "loss": 0.5251, + "step": 14455 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012706666101151365, + "loss": 0.5805, + "step": 14456 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012705737839933265, + "loss": 0.4865, + "step": 14457 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012704809553558296, + "loss": 0.583, + "step": 14458 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012703881242035085, + "loss": 0.5084, + "step": 14459 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012702952905372262, + "loss": 0.6258, + "step": 14460 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012702024543578464, + "loss": 0.4814, + "step": 14461 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012701096156662317, + "loss": 0.6572, + "step": 14462 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012700167744632453, + "loss": 0.5196, + "step": 14463 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012699239307497508, + "loss": 0.5475, + "step": 14464 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001269831084526611, + "loss": 0.4843, + "step": 14465 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012697382357946895, + "loss": 0.5025, + "step": 14466 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001269645384554849, + "loss": 0.4741, + "step": 14467 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012695525308079535, + "loss": 0.6242, + "step": 14468 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012694596745548657, + "loss": 0.5721, + "step": 14469 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012693668157964498, + "loss": 0.6307, + "step": 14470 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012692739545335681, + "loss": 0.5347, + "step": 14471 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012691810907670847, + "loss": 0.5865, + "step": 14472 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001269088224497863, + "loss": 0.511, + "step": 14473 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012689953557267657, + "loss": 0.4676, + "step": 14474 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012689024844546572, + "loss": 0.514, + "step": 14475 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012688096106824005, + "loss": 0.5881, + "step": 14476 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012687167344108594, + "loss": 0.6711, + "step": 14477 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012686238556408968, + "loss": 0.5786, + "step": 14478 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012685309743733767, + "loss": 0.5372, + "step": 14479 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001268438090609163, + "loss": 0.5545, + "step": 14480 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012683452043491184, + "loss": 0.6724, + "step": 14481 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012682523155941076, + "loss": 0.5443, + "step": 14482 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001268159424344993, + "loss": 0.5296, + "step": 14483 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012680665306026394, + "loss": 0.5817, + "step": 14484 + }, + { + "epoch": 0.43, + "learning_rate": 0.000126797363436791, + "loss": 0.4783, + "step": 14485 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012678807356416688, + "loss": 0.5017, + "step": 14486 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012677878344247788, + "loss": 0.4721, + "step": 14487 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012676949307181045, + "loss": 0.5998, + "step": 14488 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012676020245225094, + "loss": 0.5029, + "step": 14489 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012675091158388573, + "loss": 0.5208, + "step": 14490 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001267416204668012, + "loss": 0.6271, + "step": 14491 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012673232910108377, + "loss": 0.5447, + "step": 14492 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012672303748681977, + "loss": 0.6167, + "step": 14493 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012671374562409565, + "loss": 0.5227, + "step": 14494 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012670445351299774, + "loss": 0.6119, + "step": 14495 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012669516115361247, + "loss": 0.6574, + "step": 14496 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012668586854602622, + "loss": 0.5769, + "step": 14497 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001266765756903254, + "loss": 0.5375, + "step": 14498 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012666728258659644, + "loss": 0.5261, + "step": 14499 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012665798923492568, + "loss": 0.5824, + "step": 14500 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001266486956353996, + "loss": 0.5374, + "step": 14501 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012663940178810453, + "loss": 0.6622, + "step": 14502 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012663010769312693, + "loss": 0.5726, + "step": 14503 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012662081335055323, + "loss": 0.4623, + "step": 14504 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001266115187604698, + "loss": 0.6451, + "step": 14505 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001266022239229631, + "loss": 0.5361, + "step": 14506 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012659292883811945, + "loss": 0.6864, + "step": 14507 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012658363350602543, + "loss": 0.4279, + "step": 14508 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012657433792676734, + "loss": 0.569, + "step": 14509 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012656504210043163, + "loss": 0.5071, + "step": 14510 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001265557460271048, + "loss": 0.5154, + "step": 14511 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012654644970687317, + "loss": 0.5692, + "step": 14512 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012653715313982325, + "loss": 0.5692, + "step": 14513 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012652785632604147, + "loss": 0.6796, + "step": 14514 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012651855926561426, + "loss": 0.5738, + "step": 14515 + }, + { + "epoch": 0.43, + "learning_rate": 0.000126509261958628, + "loss": 0.6057, + "step": 14516 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012649996440516927, + "loss": 0.5608, + "step": 14517 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001264906666053244, + "loss": 0.5282, + "step": 14518 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012648136855917988, + "loss": 0.6158, + "step": 14519 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001264720702668221, + "loss": 0.5869, + "step": 14520 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001264627717283376, + "loss": 0.4925, + "step": 14521 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001264534729438128, + "loss": 0.4586, + "step": 14522 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012644417391333414, + "loss": 0.5644, + "step": 14523 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001264348746369881, + "loss": 0.5692, + "step": 14524 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012642557511486114, + "loss": 0.4672, + "step": 14525 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001264162753470397, + "loss": 0.4739, + "step": 14526 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012640697533361026, + "loss": 0.5022, + "step": 14527 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001263976750746593, + "loss": 0.5009, + "step": 14528 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012638837457027325, + "loss": 0.556, + "step": 14529 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012637907382053863, + "loss": 0.652, + "step": 14530 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012636977282554188, + "loss": 0.5874, + "step": 14531 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012636047158536953, + "loss": 0.6405, + "step": 14532 + }, + { + "epoch": 0.43, + "learning_rate": 0.000126351170100108, + "loss": 0.6201, + "step": 14533 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012634186836984375, + "loss": 0.5904, + "step": 14534 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012633256639466334, + "loss": 0.5201, + "step": 14535 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012632326417465323, + "loss": 0.5119, + "step": 14536 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012631396170989994, + "loss": 0.5949, + "step": 14537 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012630465900048985, + "loss": 0.4851, + "step": 14538 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012629535604650956, + "loss": 0.5603, + "step": 14539 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012628605284804552, + "loss": 0.6891, + "step": 14540 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012627674940518424, + "loss": 0.5444, + "step": 14541 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012626744571801222, + "loss": 0.4828, + "step": 14542 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012625814178661597, + "loss": 0.5311, + "step": 14543 + }, + { + "epoch": 0.43, + "learning_rate": 0.000126248837611082, + "loss": 0.5937, + "step": 14544 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012623953319149673, + "loss": 0.5205, + "step": 14545 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001262302285279468, + "loss": 0.5325, + "step": 14546 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012622092362051863, + "loss": 0.5155, + "step": 14547 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001262116184692988, + "loss": 0.5418, + "step": 14548 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012620231307437373, + "loss": 0.4882, + "step": 14549 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012619300743583006, + "loss": 0.5696, + "step": 14550 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001261837015537542, + "loss": 0.4948, + "step": 14551 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012617439542823272, + "loss": 0.5397, + "step": 14552 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012616508905935215, + "loss": 0.5927, + "step": 14553 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012615578244719902, + "loss": 0.4384, + "step": 14554 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012614647559185986, + "loss": 0.5335, + "step": 14555 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012613716849342116, + "loss": 0.6149, + "step": 14556 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012612786115196952, + "loss": 0.5526, + "step": 14557 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001261185535675914, + "loss": 0.5964, + "step": 14558 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012610924574037343, + "loss": 0.56, + "step": 14559 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012609993767040206, + "loss": 0.5004, + "step": 14560 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012609062935776388, + "loss": 0.6086, + "step": 14561 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001260813208025454, + "loss": 0.526, + "step": 14562 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012607201200483324, + "loss": 0.491, + "step": 14563 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001260627029647139, + "loss": 0.5023, + "step": 14564 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012605339368227392, + "loss": 0.5108, + "step": 14565 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012604408415759985, + "loss": 0.5102, + "step": 14566 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012603477439077827, + "loss": 0.5493, + "step": 14567 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012602546438189577, + "loss": 0.4936, + "step": 14568 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012601615413103886, + "loss": 0.5733, + "step": 14569 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001260068436382941, + "loss": 0.4534, + "step": 14570 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012599753290374813, + "loss": 0.6197, + "step": 14571 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012598822192748737, + "loss": 0.6152, + "step": 14572 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012597891070959856, + "loss": 0.5418, + "step": 14573 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012596959925016812, + "loss": 0.6204, + "step": 14574 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012596028754928277, + "loss": 0.5432, + "step": 14575 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012595097560702895, + "loss": 0.5196, + "step": 14576 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012594166342349335, + "loss": 0.5046, + "step": 14577 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012593235099876248, + "loss": 0.658, + "step": 14578 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012592303833292297, + "loss": 0.5051, + "step": 14579 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012591372542606136, + "loss": 0.556, + "step": 14580 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012590441227826427, + "loss": 0.5595, + "step": 14581 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012589509888961826, + "loss": 0.557, + "step": 14582 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012588578526020994, + "loss": 0.5476, + "step": 14583 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012587647139012593, + "loss": 0.5665, + "step": 14584 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001258671572794528, + "loss": 0.4627, + "step": 14585 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012585784292827712, + "loss": 0.572, + "step": 14586 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012584852833668556, + "loss": 0.5686, + "step": 14587 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012583921350476466, + "loss": 0.4058, + "step": 14588 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012582989843260107, + "loss": 0.5513, + "step": 14589 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012582058312028138, + "loss": 0.5656, + "step": 14590 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012581126756789216, + "loss": 0.7358, + "step": 14591 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001258019517755201, + "loss": 0.6673, + "step": 14592 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012579263574325175, + "loss": 0.4878, + "step": 14593 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001257833194711738, + "loss": 0.5326, + "step": 14594 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012577400295937276, + "loss": 0.5335, + "step": 14595 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012576468620793536, + "loss": 0.4687, + "step": 14596 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012575536921694816, + "loss": 0.5798, + "step": 14597 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001257460519864978, + "loss": 0.5715, + "step": 14598 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012573673451667092, + "loss": 0.4937, + "step": 14599 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012572741680755414, + "loss": 0.5767, + "step": 14600 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012571809885923406, + "loss": 0.6038, + "step": 14601 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001257087806717974, + "loss": 0.533, + "step": 14602 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001256994622453307, + "loss": 0.6216, + "step": 14603 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012569014357992066, + "loss": 0.5553, + "step": 14604 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012568082467565389, + "loss": 0.4761, + "step": 14605 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001256715055326171, + "loss": 0.6036, + "step": 14606 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001256621861508968, + "loss": 0.5739, + "step": 14607 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012565286653057976, + "loss": 0.6553, + "step": 14608 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012564354667175261, + "loss": 0.5446, + "step": 14609 + }, + { + "epoch": 0.43, + "learning_rate": 0.00012563422657450196, + "loss": 0.4805, + "step": 14610 + }, + { + "epoch": 0.43, + "learning_rate": 0.0001256249062389145, + "loss": 0.6201, + "step": 14611 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012561558566507688, + "loss": 0.5205, + "step": 14612 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012560626485307575, + "loss": 0.5011, + "step": 14613 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012559694380299779, + "loss": 0.4609, + "step": 14614 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012558762251492958, + "loss": 0.5718, + "step": 14615 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012557830098895796, + "loss": 0.6421, + "step": 14616 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001255689792251694, + "loss": 0.6091, + "step": 14617 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012555965722365074, + "loss": 0.5086, + "step": 14618 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001255503349844885, + "loss": 0.6293, + "step": 14619 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001255410125077695, + "loss": 0.6078, + "step": 14620 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001255316897935803, + "loss": 0.6124, + "step": 14621 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012552236684200757, + "loss": 0.516, + "step": 14622 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012551304365313813, + "loss": 0.5935, + "step": 14623 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012550372022705853, + "loss": 0.4951, + "step": 14624 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012549439656385557, + "loss": 0.5076, + "step": 14625 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012548507266361582, + "loss": 0.5941, + "step": 14626 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012547574852642604, + "loss": 0.6445, + "step": 14627 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001254664241523729, + "loss": 0.4352, + "step": 14628 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012545709954154307, + "loss": 0.6046, + "step": 14629 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001254477746940233, + "loss": 0.5799, + "step": 14630 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012543844960990027, + "loss": 0.5528, + "step": 14631 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012542912428926067, + "loss": 0.476, + "step": 14632 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001254197987321912, + "loss": 0.4879, + "step": 14633 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012541047293877856, + "loss": 0.6087, + "step": 14634 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012540114690910953, + "loss": 0.4511, + "step": 14635 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012539182064327066, + "loss": 0.4307, + "step": 14636 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012538249414134885, + "loss": 0.4932, + "step": 14637 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012537316740343068, + "loss": 0.5868, + "step": 14638 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012536384042960292, + "loss": 0.5739, + "step": 14639 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001253545132199523, + "loss": 0.6417, + "step": 14640 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001253451857745655, + "loss": 0.5324, + "step": 14641 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012533585809352928, + "loss": 0.6381, + "step": 14642 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012532653017693034, + "loss": 0.5802, + "step": 14643 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012531720202485543, + "loss": 0.4998, + "step": 14644 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012530787363739126, + "loss": 0.5146, + "step": 14645 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012529854501462453, + "loss": 0.5621, + "step": 14646 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012528921615664206, + "loss": 0.658, + "step": 14647 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012527988706353052, + "loss": 0.5556, + "step": 14648 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012527055773537668, + "loss": 0.6371, + "step": 14649 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012526122817226723, + "loss": 0.4845, + "step": 14650 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012525189837428897, + "loss": 0.5426, + "step": 14651 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012524256834152864, + "loss": 0.4444, + "step": 14652 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012523323807407295, + "loss": 0.6452, + "step": 14653 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001252239075720087, + "loss": 0.4378, + "step": 14654 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012521457683542256, + "loss": 0.3875, + "step": 14655 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012520524586440135, + "loss": 0.5912, + "step": 14656 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012519591465903182, + "loss": 0.5127, + "step": 14657 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012518658321940073, + "loss": 0.6163, + "step": 14658 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012517725154559482, + "loss": 0.5037, + "step": 14659 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012516791963770085, + "loss": 0.4856, + "step": 14660 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001251585874958056, + "loss": 0.6167, + "step": 14661 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001251492551199958, + "loss": 0.5716, + "step": 14662 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001251399225103583, + "loss": 0.4123, + "step": 14663 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001251305896669798, + "loss": 0.6568, + "step": 14664 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012512125658994705, + "loss": 0.541, + "step": 14665 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012511192327934695, + "loss": 0.5618, + "step": 14666 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012510258973526614, + "loss": 0.582, + "step": 14667 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012509325595779148, + "loss": 0.6144, + "step": 14668 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012508392194700974, + "loss": 0.5205, + "step": 14669 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012507458770300763, + "loss": 0.5797, + "step": 14670 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012506525322587207, + "loss": 0.5332, + "step": 14671 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012505591851568975, + "loss": 0.5468, + "step": 14672 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012504658357254746, + "loss": 0.5956, + "step": 14673 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012503724839653204, + "loss": 0.563, + "step": 14674 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012502791298773028, + "loss": 0.5838, + "step": 14675 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012501857734622892, + "loss": 0.5467, + "step": 14676 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001250092414721148, + "loss": 0.5382, + "step": 14677 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012499990536547477, + "loss": 0.5002, + "step": 14678 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012499056902639554, + "loss": 0.5243, + "step": 14679 + }, + { + "epoch": 0.44, + "learning_rate": 0.000124981232454964, + "loss": 0.4394, + "step": 14680 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001249718956512669, + "loss": 0.5812, + "step": 14681 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012496255861539103, + "loss": 0.5388, + "step": 14682 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012495322134742327, + "loss": 0.5388, + "step": 14683 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001249438838474504, + "loss": 0.5511, + "step": 14684 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012493454611555926, + "loss": 0.5924, + "step": 14685 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001249252081518366, + "loss": 0.5904, + "step": 14686 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012491586995636933, + "loss": 0.5851, + "step": 14687 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012490653152924424, + "loss": 0.5898, + "step": 14688 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001248971928705481, + "loss": 0.4428, + "step": 14689 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012488785398036782, + "loss": 0.5296, + "step": 14690 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012487851485879018, + "loss": 0.5008, + "step": 14691 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012486917550590202, + "loss": 0.7319, + "step": 14692 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001248598359217902, + "loss": 0.5534, + "step": 14693 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001248504961065415, + "loss": 0.5624, + "step": 14694 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012484115606024282, + "loss": 0.5014, + "step": 14695 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012483181578298096, + "loss": 0.5255, + "step": 14696 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012482247527484278, + "loss": 0.613, + "step": 14697 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001248131345359151, + "loss": 0.6111, + "step": 14698 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012480379356628483, + "loss": 0.5262, + "step": 14699 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001247944523660387, + "loss": 0.572, + "step": 14700 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001247851109352637, + "loss": 0.538, + "step": 14701 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001247757692740466, + "loss": 0.541, + "step": 14702 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012476642738247425, + "loss": 0.5729, + "step": 14703 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012475708526063354, + "loss": 0.5262, + "step": 14704 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012474774290861134, + "loss": 0.4741, + "step": 14705 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012473840032649447, + "loss": 0.503, + "step": 14706 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001247290575143698, + "loss": 0.587, + "step": 14707 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012471971447232424, + "loss": 0.6581, + "step": 14708 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001247103712004446, + "loss": 0.4697, + "step": 14709 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012470102769881778, + "loss": 0.5622, + "step": 14710 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012469168396753069, + "loss": 0.4298, + "step": 14711 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001246823400066701, + "loss": 0.5818, + "step": 14712 + }, + { + "epoch": 0.44, + "learning_rate": 0.000124672995816323, + "loss": 0.5259, + "step": 14713 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012466365139657617, + "loss": 0.5244, + "step": 14714 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001246543067475166, + "loss": 0.6002, + "step": 14715 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012464496186923108, + "loss": 0.4879, + "step": 14716 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012463561676180653, + "loss": 0.5436, + "step": 14717 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012462627142532983, + "loss": 0.6015, + "step": 14718 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012461692585988786, + "loss": 0.5543, + "step": 14719 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012460758006556757, + "loss": 0.4653, + "step": 14720 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012459823404245576, + "loss": 0.555, + "step": 14721 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001245888877906394, + "loss": 0.5357, + "step": 14722 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012457954131020536, + "loss": 0.555, + "step": 14723 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012457019460124054, + "loss": 0.5327, + "step": 14724 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012456084766383185, + "loss": 0.5568, + "step": 14725 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012455150049806618, + "loss": 0.4385, + "step": 14726 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001245421531040304, + "loss": 0.4979, + "step": 14727 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012453280548181152, + "loss": 0.5896, + "step": 14728 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012452345763149636, + "loss": 0.6226, + "step": 14729 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012451410955317192, + "loss": 0.6078, + "step": 14730 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012450476124692498, + "loss": 0.5808, + "step": 14731 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012449541271284257, + "loss": 0.6243, + "step": 14732 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001244860639510116, + "loss": 0.5453, + "step": 14733 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012447671496151894, + "loss": 0.4685, + "step": 14734 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012446736574445156, + "loss": 0.5732, + "step": 14735 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012445801629989636, + "loss": 0.4791, + "step": 14736 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012444866662794025, + "loss": 0.4758, + "step": 14737 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001244393167286702, + "loss": 0.5779, + "step": 14738 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001244299666021731, + "loss": 0.5326, + "step": 14739 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012442061624853596, + "loss": 0.5825, + "step": 14740 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001244112656678456, + "loss": 0.6186, + "step": 14741 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012440191486018907, + "loss": 0.5508, + "step": 14742 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012439256382565324, + "loss": 0.5647, + "step": 14743 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012438321256432507, + "loss": 0.4568, + "step": 14744 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001243738610762915, + "loss": 0.5253, + "step": 14745 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012436450936163953, + "loss": 0.5483, + "step": 14746 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012435515742045602, + "loss": 0.5882, + "step": 14747 + }, + { + "epoch": 0.44, + "learning_rate": 0.000124345805252828, + "loss": 0.6052, + "step": 14748 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012433645285884234, + "loss": 0.4442, + "step": 14749 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012432710023858607, + "loss": 0.5431, + "step": 14750 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001243177473921461, + "loss": 0.5933, + "step": 14751 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012430839431960944, + "loss": 0.665, + "step": 14752 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012429904102106298, + "loss": 0.4246, + "step": 14753 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012428968749659375, + "loss": 0.6291, + "step": 14754 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012428033374628866, + "loss": 0.587, + "step": 14755 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012427097977023473, + "loss": 0.5284, + "step": 14756 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012426162556851888, + "loss": 0.4958, + "step": 14757 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001242522711412281, + "loss": 0.5386, + "step": 14758 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012424291648844942, + "loss": 0.6968, + "step": 14759 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001242335616102697, + "loss": 0.4608, + "step": 14760 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012422420650677604, + "loss": 0.6313, + "step": 14761 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012421485117805533, + "loss": 0.4635, + "step": 14762 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012420549562419458, + "loss": 0.5919, + "step": 14763 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012419613984528078, + "loss": 0.6242, + "step": 14764 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012418678384140092, + "loss": 0.5104, + "step": 14765 + }, + { + "epoch": 0.44, + "learning_rate": 0.000124177427612642, + "loss": 0.5508, + "step": 14766 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012416807115909098, + "loss": 0.5155, + "step": 14767 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012415871448083487, + "loss": 0.5275, + "step": 14768 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012414935757796067, + "loss": 0.5321, + "step": 14769 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012414000045055535, + "loss": 0.5669, + "step": 14770 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012413064309870593, + "loss": 0.6189, + "step": 14771 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001241212855224994, + "loss": 0.5954, + "step": 14772 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001241119277220228, + "loss": 0.6051, + "step": 14773 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012410256969736308, + "loss": 0.5974, + "step": 14774 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012409321144860727, + "loss": 0.6211, + "step": 14775 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001240838529758424, + "loss": 0.5981, + "step": 14776 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012407449427915548, + "loss": 0.5091, + "step": 14777 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012406513535863348, + "loss": 0.5078, + "step": 14778 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012405577621436343, + "loss": 0.5416, + "step": 14779 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012404641684643242, + "loss": 0.5069, + "step": 14780 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012403705725492735, + "loss": 0.6073, + "step": 14781 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001240276974399353, + "loss": 0.5321, + "step": 14782 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012401833740154333, + "loss": 0.6297, + "step": 14783 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012400897713983844, + "loss": 0.3787, + "step": 14784 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012399961665490763, + "loss": 0.5603, + "step": 14785 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012399025594683795, + "loss": 0.6009, + "step": 14786 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012398089501571642, + "loss": 0.5495, + "step": 14787 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012397153386163012, + "loss": 0.6117, + "step": 14788 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012396217248466598, + "loss": 0.488, + "step": 14789 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001239528108849112, + "loss": 0.5691, + "step": 14790 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012394344906245267, + "loss": 0.5742, + "step": 14791 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012393408701737754, + "loss": 0.5019, + "step": 14792 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012392472474977276, + "loss": 0.5531, + "step": 14793 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012391536225972547, + "loss": 0.5546, + "step": 14794 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012390599954732262, + "loss": 0.5266, + "step": 14795 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012389663661265137, + "loss": 0.5698, + "step": 14796 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012388727345579868, + "loss": 0.5938, + "step": 14797 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012387791007685165, + "loss": 0.5393, + "step": 14798 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012386854647589734, + "loss": 0.5313, + "step": 14799 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012385918265302281, + "loss": 0.6407, + "step": 14800 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012384981860831507, + "loss": 0.6155, + "step": 14801 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012384045434186123, + "loss": 0.521, + "step": 14802 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012383108985374836, + "loss": 0.5919, + "step": 14803 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012382172514406354, + "loss": 0.4396, + "step": 14804 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012381236021289377, + "loss": 0.4766, + "step": 14805 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001238029950603262, + "loss": 0.6106, + "step": 14806 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012379362968644784, + "loss": 0.5334, + "step": 14807 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001237842640913458, + "loss": 0.6286, + "step": 14808 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012377489827510718, + "loss": 0.581, + "step": 14809 + }, + { + "epoch": 0.44, + "learning_rate": 0.000123765532237819, + "loss": 0.6343, + "step": 14810 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012375616597956837, + "loss": 0.5773, + "step": 14811 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001237467995004424, + "loss": 0.4931, + "step": 14812 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012373743280052812, + "loss": 0.5419, + "step": 14813 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001237280658799127, + "loss": 0.5893, + "step": 14814 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012371869873868314, + "loss": 0.5552, + "step": 14815 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001237093313769266, + "loss": 0.5648, + "step": 14816 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012369996379473015, + "loss": 0.5516, + "step": 14817 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012369059599218085, + "loss": 0.5705, + "step": 14818 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012368122796936585, + "loss": 0.7004, + "step": 14819 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001236718597263722, + "loss": 0.4595, + "step": 14820 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001236624912632871, + "loss": 0.5636, + "step": 14821 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012365312258019753, + "loss": 0.5196, + "step": 14822 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012364375367719067, + "loss": 0.5277, + "step": 14823 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001236343845543536, + "loss": 0.5095, + "step": 14824 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012362501521177347, + "loss": 0.5427, + "step": 14825 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012361564564953735, + "loss": 0.5839, + "step": 14826 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012360627586773237, + "loss": 0.4687, + "step": 14827 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012359690586644564, + "loss": 0.5307, + "step": 14828 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001235875356457643, + "loss": 0.4434, + "step": 14829 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012357816520577544, + "loss": 0.6067, + "step": 14830 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012356879454656622, + "loss": 0.4971, + "step": 14831 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001235594236682237, + "loss": 0.5216, + "step": 14832 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012355005257083508, + "loss": 0.5539, + "step": 14833 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012354068125448745, + "loss": 0.6563, + "step": 14834 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012353130971926794, + "loss": 0.5207, + "step": 14835 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012352193796526367, + "loss": 0.5972, + "step": 14836 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001235125659925618, + "loss": 0.4815, + "step": 14837 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001235031938012495, + "loss": 0.4471, + "step": 14838 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001234938213914139, + "loss": 0.5201, + "step": 14839 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012348444876314206, + "loss": 0.5622, + "step": 14840 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012347507591652118, + "loss": 0.5024, + "step": 14841 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001234657028516384, + "loss": 0.5237, + "step": 14842 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012345632956858087, + "loss": 0.4129, + "step": 14843 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012344695606743573, + "loss": 0.5408, + "step": 14844 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012343758234829015, + "loss": 0.5263, + "step": 14845 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012342820841123125, + "loss": 0.5345, + "step": 14846 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012341883425634627, + "loss": 0.4225, + "step": 14847 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012340945988372222, + "loss": 0.6119, + "step": 14848 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012340008529344637, + "loss": 0.4819, + "step": 14849 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001233907104856059, + "loss": 0.5343, + "step": 14850 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012338133546028785, + "loss": 0.5721, + "step": 14851 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012337196021757954, + "loss": 0.6274, + "step": 14852 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012336258475756799, + "loss": 0.5134, + "step": 14853 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012335320908034048, + "loss": 0.611, + "step": 14854 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012334383318598412, + "loss": 0.5979, + "step": 14855 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012333445707458612, + "loss": 0.6196, + "step": 14856 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012332508074623362, + "loss": 0.4777, + "step": 14857 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001233157042010138, + "loss": 0.6002, + "step": 14858 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001233063274390139, + "loss": 0.6431, + "step": 14859 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012329695046032102, + "loss": 0.5462, + "step": 14860 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001232875732650224, + "loss": 0.4586, + "step": 14861 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001232781958532052, + "loss": 0.5861, + "step": 14862 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001232688182249566, + "loss": 0.4747, + "step": 14863 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012325944038036384, + "loss": 0.6103, + "step": 14864 + }, + { + "epoch": 0.44, + "learning_rate": 0.000123250062319514, + "loss": 0.4579, + "step": 14865 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012324068404249446, + "loss": 0.6275, + "step": 14866 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001232313055493922, + "loss": 0.5371, + "step": 14867 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012322192684029455, + "loss": 0.6565, + "step": 14868 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001232125479152887, + "loss": 0.5927, + "step": 14869 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001232031687744618, + "loss": 0.5382, + "step": 14870 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012319378941790112, + "loss": 0.5251, + "step": 14871 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012318440984569382, + "loss": 0.4653, + "step": 14872 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001231750300579271, + "loss": 0.5505, + "step": 14873 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001231656500546882, + "loss": 0.5453, + "step": 14874 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001231562698360643, + "loss": 0.5593, + "step": 14875 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012314688940214268, + "loss": 0.5383, + "step": 14876 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012313750875301045, + "loss": 0.5344, + "step": 14877 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012312812788875494, + "loss": 0.5756, + "step": 14878 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012311874680946328, + "loss": 0.435, + "step": 14879 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001231093655152227, + "loss": 0.5733, + "step": 14880 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001230999840061205, + "loss": 0.678, + "step": 14881 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012309060228224384, + "loss": 0.4946, + "step": 14882 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012308122034368, + "loss": 0.5257, + "step": 14883 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012307183819051611, + "loss": 0.6194, + "step": 14884 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001230624558228395, + "loss": 0.609, + "step": 14885 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012305307324073734, + "loss": 0.622, + "step": 14886 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001230436904442969, + "loss": 0.5458, + "step": 14887 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012303430743360543, + "loss": 0.4849, + "step": 14888 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012302492420875015, + "loss": 0.6377, + "step": 14889 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012301554076981828, + "loss": 0.5868, + "step": 14890 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001230061571168971, + "loss": 0.6048, + "step": 14891 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012299677325007383, + "loss": 0.6319, + "step": 14892 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012298738916943574, + "loss": 0.5257, + "step": 14893 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012297800487507006, + "loss": 0.564, + "step": 14894 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012296862036706405, + "loss": 0.5923, + "step": 14895 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012295923564550497, + "loss": 0.5204, + "step": 14896 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012294985071048005, + "loss": 0.5765, + "step": 14897 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012294046556207656, + "loss": 0.6557, + "step": 14898 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001229310802003818, + "loss": 0.4997, + "step": 14899 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012292169462548299, + "loss": 0.4791, + "step": 14900 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012291230883746735, + "loss": 0.592, + "step": 14901 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012290292283642225, + "loss": 0.5966, + "step": 14902 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001228935366224349, + "loss": 0.5394, + "step": 14903 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012288415019559253, + "loss": 0.5496, + "step": 14904 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012287476355598247, + "loss": 0.568, + "step": 14905 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012286537670369195, + "loss": 0.6051, + "step": 14906 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012285598963880832, + "loss": 0.586, + "step": 14907 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001228466023614188, + "loss": 0.4948, + "step": 14908 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012283721487161067, + "loss": 0.4858, + "step": 14909 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012282782716947118, + "loss": 0.7355, + "step": 14910 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001228184392550877, + "loss": 0.4859, + "step": 14911 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012280905112854745, + "loss": 0.5841, + "step": 14912 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012279966278993772, + "loss": 0.5634, + "step": 14913 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012279027423934584, + "loss": 0.5837, + "step": 14914 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012278088547685906, + "loss": 0.515, + "step": 14915 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001227714965025647, + "loss": 0.6044, + "step": 14916 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012276210731655002, + "loss": 0.5586, + "step": 14917 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012275271791890234, + "loss": 0.5311, + "step": 14918 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012274332830970898, + "loss": 0.473, + "step": 14919 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001227339384890572, + "loss": 0.5032, + "step": 14920 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012272454845703432, + "loss": 0.4228, + "step": 14921 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012271515821372765, + "loss": 0.6572, + "step": 14922 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001227057677592245, + "loss": 0.6365, + "step": 14923 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012269637709361216, + "loss": 0.5782, + "step": 14924 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012268698621697793, + "loss": 0.6206, + "step": 14925 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001226775951294092, + "loss": 0.4991, + "step": 14926 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012266820383099317, + "loss": 0.4836, + "step": 14927 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012265881232181725, + "loss": 0.5242, + "step": 14928 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012264942060196868, + "loss": 0.5609, + "step": 14929 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012264002867153488, + "loss": 0.4853, + "step": 14930 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012263063653060308, + "loss": 0.5794, + "step": 14931 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012262124417926066, + "loss": 0.5624, + "step": 14932 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012261185161759495, + "loss": 0.6144, + "step": 14933 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012260245884569321, + "loss": 0.5199, + "step": 14934 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012259306586364285, + "loss": 0.5661, + "step": 14935 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012258367267153114, + "loss": 0.5391, + "step": 14936 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012257427926944546, + "loss": 0.4715, + "step": 14937 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012256488565747314, + "loss": 0.4919, + "step": 14938 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012255549183570147, + "loss": 0.5268, + "step": 14939 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001225460978042179, + "loss": 0.5463, + "step": 14940 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001225367035631096, + "loss": 0.6053, + "step": 14941 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012252730911246407, + "loss": 0.5605, + "step": 14942 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012251791445236858, + "loss": 0.5667, + "step": 14943 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001225085195829105, + "loss": 0.5558, + "step": 14944 + }, + { + "epoch": 0.44, + "learning_rate": 0.0001224991245041772, + "loss": 0.4769, + "step": 14945 + }, + { + "epoch": 0.44, + "learning_rate": 0.000122489729216256, + "loss": 0.4787, + "step": 14946 + }, + { + "epoch": 0.44, + "learning_rate": 0.00012248033371923425, + "loss": 0.4962, + "step": 14947 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001224709380131993, + "loss": 0.5322, + "step": 14948 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012246154209823854, + "loss": 0.506, + "step": 14949 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012245214597443932, + "loss": 0.5556, + "step": 14950 + }, + { + "epoch": 0.45, + "learning_rate": 0.000122442749641889, + "loss": 0.4797, + "step": 14951 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012243335310067492, + "loss": 0.5049, + "step": 14952 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012242395635088447, + "loss": 0.5402, + "step": 14953 + }, + { + "epoch": 0.45, + "learning_rate": 0.000122414559392605, + "loss": 0.5616, + "step": 14954 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012240516222592393, + "loss": 0.5156, + "step": 14955 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012239576485092855, + "loss": 0.6063, + "step": 14956 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001223863672677063, + "loss": 0.5955, + "step": 14957 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012237696947634452, + "loss": 0.5821, + "step": 14958 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001223675714769306, + "loss": 0.5719, + "step": 14959 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001223581732695519, + "loss": 0.4666, + "step": 14960 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012234877485429584, + "loss": 0.5454, + "step": 14961 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012233937623124975, + "loss": 0.5736, + "step": 14962 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012232997740050107, + "loss": 0.5859, + "step": 14963 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012232057836213716, + "loss": 0.589, + "step": 14964 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012231117911624538, + "loss": 0.5076, + "step": 14965 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012230177966291318, + "loss": 0.6488, + "step": 14966 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001222923800022279, + "loss": 0.534, + "step": 14967 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012228298013427694, + "loss": 0.6472, + "step": 14968 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012227358005914777, + "loss": 0.644, + "step": 14969 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012226417977692769, + "loss": 0.4532, + "step": 14970 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012225477928770414, + "loss": 0.5866, + "step": 14971 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001222453785915645, + "loss": 0.5476, + "step": 14972 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012223597768859626, + "loss": 0.5439, + "step": 14973 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001222265765788867, + "loss": 0.5614, + "step": 14974 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001222171752625233, + "loss": 0.4584, + "step": 14975 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001222077737395935, + "loss": 0.5633, + "step": 14976 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012219837201018466, + "loss": 0.492, + "step": 14977 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012218897007438417, + "loss": 0.6362, + "step": 14978 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001221795679322795, + "loss": 0.5279, + "step": 14979 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012217016558395803, + "loss": 0.592, + "step": 14980 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012216076302950724, + "loss": 0.5392, + "step": 14981 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001221513602690145, + "loss": 0.553, + "step": 14982 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012214195730256722, + "loss": 0.5606, + "step": 14983 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012213255413025283, + "loss": 0.4788, + "step": 14984 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001221231507521588, + "loss": 0.5131, + "step": 14985 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012211374716837253, + "loss": 0.5304, + "step": 14986 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001221043433789814, + "loss": 0.4993, + "step": 14987 + }, + { + "epoch": 0.45, + "learning_rate": 0.000122094939384073, + "loss": 0.5425, + "step": 14988 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012208553518373456, + "loss": 0.5575, + "step": 14989 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012207613077805366, + "loss": 0.5778, + "step": 14990 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001220667261671177, + "loss": 0.5236, + "step": 14991 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001220573213510141, + "loss": 0.5951, + "step": 14992 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012204791632983032, + "loss": 0.6817, + "step": 14993 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012203851110365381, + "loss": 0.5833, + "step": 14994 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012202910567257199, + "loss": 0.5776, + "step": 14995 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012201970003667236, + "loss": 0.4815, + "step": 14996 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001220102941960423, + "loss": 0.6069, + "step": 14997 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012200088815076932, + "loss": 0.5222, + "step": 14998 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001219914819009408, + "loss": 0.4853, + "step": 14999 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012198207544664433, + "loss": 0.5681, + "step": 15000 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012197266878796723, + "loss": 0.6072, + "step": 15001 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012196326192499703, + "loss": 0.6007, + "step": 15002 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012195385485782116, + "loss": 0.5856, + "step": 15003 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001219444475865271, + "loss": 0.5894, + "step": 15004 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012193504011120231, + "loss": 0.4637, + "step": 15005 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012192563243193426, + "loss": 0.6704, + "step": 15006 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012191622454881044, + "loss": 0.4958, + "step": 15007 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012190681646191828, + "loss": 0.5737, + "step": 15008 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012189740817134529, + "loss": 0.5855, + "step": 15009 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001218879996771789, + "loss": 0.5038, + "step": 15010 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012187859097950662, + "loss": 0.5447, + "step": 15011 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012186918207841595, + "loss": 0.5421, + "step": 15012 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012185977297399429, + "loss": 0.5943, + "step": 15013 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012185036366632922, + "loss": 0.6122, + "step": 15014 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012184095415550816, + "loss": 0.568, + "step": 15015 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001218315444416186, + "loss": 0.5886, + "step": 15016 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012182213452474807, + "loss": 0.566, + "step": 15017 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012181272440498399, + "loss": 0.5147, + "step": 15018 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012180331408241394, + "loss": 0.4674, + "step": 15019 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012179390355712532, + "loss": 0.48, + "step": 15020 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012178449282920569, + "loss": 0.5753, + "step": 15021 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012177508189874254, + "loss": 0.5697, + "step": 15022 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012176567076582335, + "loss": 0.6458, + "step": 15023 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012175625943053564, + "loss": 0.4508, + "step": 15024 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012174684789296687, + "loss": 0.4932, + "step": 15025 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012173743615320461, + "loss": 0.5336, + "step": 15026 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012172802421133632, + "loss": 0.4868, + "step": 15027 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001217186120674495, + "loss": 0.5986, + "step": 15028 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012170919972163171, + "loss": 0.6479, + "step": 15029 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012169978717397039, + "loss": 0.5378, + "step": 15030 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012169037442455315, + "loss": 0.6443, + "step": 15031 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012168096147346741, + "loss": 0.6312, + "step": 15032 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012167154832080076, + "loss": 0.6162, + "step": 15033 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012166213496664068, + "loss": 0.5807, + "step": 15034 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012165272141107471, + "loss": 0.5855, + "step": 15035 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012164330765419036, + "loss": 0.4476, + "step": 15036 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012163389369607516, + "loss": 0.5109, + "step": 15037 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012162447953681664, + "loss": 0.4962, + "step": 15038 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012161506517650235, + "loss": 0.5383, + "step": 15039 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012160565061521978, + "loss": 0.5767, + "step": 15040 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012159623585305646, + "loss": 0.5454, + "step": 15041 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012158682089009997, + "loss": 0.5981, + "step": 15042 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012157740572643783, + "loss": 0.4179, + "step": 15043 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012156799036215755, + "loss": 0.4387, + "step": 15044 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012155857479734671, + "loss": 0.5394, + "step": 15045 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012154915903209283, + "loss": 0.4827, + "step": 15046 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012153974306648346, + "loss": 0.4969, + "step": 15047 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012153032690060615, + "loss": 0.6365, + "step": 15048 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012152091053454841, + "loss": 0.5556, + "step": 15049 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012151149396839787, + "loss": 0.5327, + "step": 15050 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012150207720224198, + "loss": 0.5386, + "step": 15051 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012149266023616839, + "loss": 0.6898, + "step": 15052 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012148324307026457, + "loss": 0.6226, + "step": 15053 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012147382570461816, + "loss": 0.5043, + "step": 15054 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012146440813931666, + "loss": 0.4999, + "step": 15055 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012145499037444765, + "loss": 0.5887, + "step": 15056 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012144557241009869, + "loss": 0.4731, + "step": 15057 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012143615424635733, + "loss": 0.517, + "step": 15058 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012142673588331117, + "loss": 0.5733, + "step": 15059 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012141731732104775, + "loss": 0.5283, + "step": 15060 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012140789855965465, + "loss": 0.5033, + "step": 15061 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012139847959921943, + "loss": 0.598, + "step": 15062 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001213890604398297, + "loss": 0.526, + "step": 15063 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012137964108157303, + "loss": 0.5266, + "step": 15064 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012137022152453693, + "loss": 0.4959, + "step": 15065 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012136080176880904, + "loss": 0.5575, + "step": 15066 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012135138181447691, + "loss": 0.5361, + "step": 15067 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001213419616616282, + "loss": 0.6644, + "step": 15068 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001213325413103504, + "loss": 0.613, + "step": 15069 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012132312076073113, + "loss": 0.5149, + "step": 15070 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012131370001285798, + "loss": 0.6223, + "step": 15071 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012130427906681857, + "loss": 0.7133, + "step": 15072 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001212948579227004, + "loss": 0.4908, + "step": 15073 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012128543658059122, + "loss": 0.5484, + "step": 15074 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012127601504057847, + "loss": 0.4639, + "step": 15075 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012126659330274983, + "loss": 0.5178, + "step": 15076 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012125717136719288, + "loss": 0.5199, + "step": 15077 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012124774923399523, + "loss": 0.5346, + "step": 15078 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012123832690324446, + "loss": 0.6158, + "step": 15079 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012122890437502818, + "loss": 0.4125, + "step": 15080 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012121948164943406, + "loss": 0.5698, + "step": 15081 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012121005872654962, + "loss": 0.519, + "step": 15082 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012120063560646251, + "loss": 0.4856, + "step": 15083 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012119121228926033, + "loss": 0.5145, + "step": 15084 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012118178877503073, + "loss": 0.5508, + "step": 15085 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012117236506386129, + "loss": 0.7079, + "step": 15086 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012116294115583964, + "loss": 0.4858, + "step": 15087 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012115351705105338, + "loss": 0.6042, + "step": 15088 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012114409274959017, + "loss": 0.4673, + "step": 15089 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012113466825153758, + "loss": 0.5051, + "step": 15090 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012112524355698331, + "loss": 0.623, + "step": 15091 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012111581866601494, + "loss": 0.5402, + "step": 15092 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012110639357872009, + "loss": 0.5467, + "step": 15093 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001210969682951864, + "loss": 0.6443, + "step": 15094 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012108754281550155, + "loss": 0.4977, + "step": 15095 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012107811713975305, + "loss": 0.6001, + "step": 15096 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012106869126802869, + "loss": 0.5416, + "step": 15097 + }, + { + "epoch": 0.45, + "learning_rate": 0.000121059265200416, + "loss": 0.4997, + "step": 15098 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012104983893700266, + "loss": 0.5211, + "step": 15099 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012104041247787632, + "loss": 0.5189, + "step": 15100 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001210309858231246, + "loss": 0.5843, + "step": 15101 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012102155897283516, + "loss": 0.5375, + "step": 15102 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012101213192709564, + "loss": 0.5073, + "step": 15103 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012100270468599369, + "loss": 0.4611, + "step": 15104 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012099327724961698, + "loss": 0.495, + "step": 15105 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012098384961805311, + "loss": 0.4415, + "step": 15106 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012097442179138983, + "loss": 0.5382, + "step": 15107 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012096499376971466, + "loss": 0.4983, + "step": 15108 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012095556555311538, + "loss": 0.4847, + "step": 15109 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001209461371416796, + "loss": 0.5828, + "step": 15110 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012093670853549497, + "loss": 0.4774, + "step": 15111 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012092727973464918, + "loss": 0.5931, + "step": 15112 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012091785073922987, + "loss": 0.4457, + "step": 15113 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012090842154932476, + "loss": 0.5658, + "step": 15114 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012089899216502145, + "loss": 0.6086, + "step": 15115 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012088956258640765, + "loss": 0.6163, + "step": 15116 + }, + { + "epoch": 0.45, + "learning_rate": 0.000120880132813571, + "loss": 0.4317, + "step": 15117 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012087070284659923, + "loss": 0.6408, + "step": 15118 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012086127268557996, + "loss": 0.5363, + "step": 15119 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001208518423306009, + "loss": 0.628, + "step": 15120 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012084241178174968, + "loss": 0.5014, + "step": 15121 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012083298103911409, + "loss": 0.3875, + "step": 15122 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001208235501027817, + "loss": 0.6016, + "step": 15123 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012081411897284027, + "loss": 0.603, + "step": 15124 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012080468764937744, + "loss": 0.5004, + "step": 15125 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001207952561324809, + "loss": 0.4796, + "step": 15126 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001207858244222384, + "loss": 0.6034, + "step": 15127 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012077639251873756, + "loss": 0.5515, + "step": 15128 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012076696042206612, + "loss": 0.569, + "step": 15129 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012075752813231174, + "loss": 0.4892, + "step": 15130 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012074809564956215, + "loss": 0.4786, + "step": 15131 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012073866297390503, + "loss": 0.6064, + "step": 15132 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001207292301054281, + "loss": 0.613, + "step": 15133 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012071979704421903, + "loss": 0.5577, + "step": 15134 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012071036379036554, + "loss": 0.6175, + "step": 15135 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012070093034395538, + "loss": 0.654, + "step": 15136 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012069149670507616, + "loss": 0.6325, + "step": 15137 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012068206287381571, + "loss": 0.5401, + "step": 15138 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012067262885026163, + "loss": 0.5502, + "step": 15139 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012066319463450173, + "loss": 0.6014, + "step": 15140 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012065376022662365, + "loss": 0.5423, + "step": 15141 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012064432562671517, + "loss": 0.5232, + "step": 15142 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012063489083486396, + "loss": 0.6401, + "step": 15143 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012062545585115777, + "loss": 0.6137, + "step": 15144 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001206160206756843, + "loss": 0.6021, + "step": 15145 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012060658530853129, + "loss": 0.5925, + "step": 15146 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012059714974978646, + "loss": 0.5654, + "step": 15147 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012058771399953754, + "loss": 0.5193, + "step": 15148 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012057827805787227, + "loss": 0.5631, + "step": 15149 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012056884192487838, + "loss": 0.6439, + "step": 15150 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012055940560064359, + "loss": 0.5761, + "step": 15151 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012054996908525562, + "loss": 0.6791, + "step": 15152 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012054053237880225, + "loss": 0.5302, + "step": 15153 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012053109548137118, + "loss": 0.4652, + "step": 15154 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012052165839305018, + "loss": 0.5755, + "step": 15155 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012051222111392697, + "loss": 0.5077, + "step": 15156 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012050278364408931, + "loss": 0.4928, + "step": 15157 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012049334598362494, + "loss": 0.4113, + "step": 15158 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001204839081326216, + "loss": 0.5231, + "step": 15159 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012047447009116706, + "loss": 0.5061, + "step": 15160 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012046503185934905, + "loss": 0.7474, + "step": 15161 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012045559343725532, + "loss": 0.5382, + "step": 15162 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012044615482497366, + "loss": 0.6487, + "step": 15163 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012043671602259176, + "loss": 0.5095, + "step": 15164 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012042727703019747, + "loss": 0.5997, + "step": 15165 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012041783784787846, + "loss": 0.6309, + "step": 15166 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012040839847572256, + "loss": 0.509, + "step": 15167 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012039895891381747, + "loss": 0.4813, + "step": 15168 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012038951916225103, + "loss": 0.518, + "step": 15169 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012038007922111094, + "loss": 0.6425, + "step": 15170 + }, + { + "epoch": 0.45, + "learning_rate": 0.000120370639090485, + "loss": 0.465, + "step": 15171 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012036119877046097, + "loss": 0.4677, + "step": 15172 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012035175826112666, + "loss": 0.5858, + "step": 15173 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001203423175625698, + "loss": 0.4891, + "step": 15174 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012033287667487816, + "loss": 0.5269, + "step": 15175 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012032343559813954, + "loss": 0.5896, + "step": 15176 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012031399433244175, + "loss": 0.5093, + "step": 15177 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001203045528778725, + "loss": 0.5924, + "step": 15178 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012029511123451961, + "loss": 0.5513, + "step": 15179 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001202856694024709, + "loss": 0.4715, + "step": 15180 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001202762273818141, + "loss": 0.5376, + "step": 15181 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012026678517263701, + "loss": 0.5878, + "step": 15182 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012025734277502741, + "loss": 0.5464, + "step": 15183 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012024790018907317, + "loss": 0.5877, + "step": 15184 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012023845741486197, + "loss": 0.5489, + "step": 15185 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001202290144524817, + "loss": 0.6153, + "step": 15186 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012021957130202006, + "loss": 0.6232, + "step": 15187 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012021012796356495, + "loss": 0.5378, + "step": 15188 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012020068443720412, + "loss": 0.5095, + "step": 15189 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012019124072302536, + "loss": 0.5778, + "step": 15190 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012018179682111651, + "loss": 0.5258, + "step": 15191 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012017235273156538, + "loss": 0.5075, + "step": 15192 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001201629084544597, + "loss": 0.5552, + "step": 15193 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012015346398988736, + "loss": 0.5317, + "step": 15194 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012014401933793615, + "loss": 0.5052, + "step": 15195 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012013457449869387, + "loss": 0.7456, + "step": 15196 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012012512947224832, + "loss": 0.4225, + "step": 15197 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012011568425868737, + "loss": 0.5787, + "step": 15198 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012010623885809879, + "loss": 0.4432, + "step": 15199 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012009679327057044, + "loss": 0.5861, + "step": 15200 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012008734749619006, + "loss": 0.5343, + "step": 15201 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012007790153504557, + "loss": 0.631, + "step": 15202 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012006845538722476, + "loss": 0.4965, + "step": 15203 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012005900905281545, + "loss": 0.5419, + "step": 15204 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012004956253190544, + "loss": 0.5813, + "step": 15205 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012004011582458261, + "loss": 0.4787, + "step": 15206 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012003066893093475, + "loss": 0.6338, + "step": 15207 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012002122185104973, + "loss": 0.5376, + "step": 15208 + }, + { + "epoch": 0.45, + "learning_rate": 0.00012001177458501536, + "loss": 0.6434, + "step": 15209 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001200023271329195, + "loss": 0.5565, + "step": 15210 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011999287949484998, + "loss": 0.5834, + "step": 15211 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011998343167089463, + "loss": 0.6478, + "step": 15212 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011997398366114128, + "loss": 0.525, + "step": 15213 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011996453546567778, + "loss": 0.5136, + "step": 15214 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011995508708459202, + "loss": 0.5414, + "step": 15215 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011994563851797178, + "loss": 0.5856, + "step": 15216 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011993618976590499, + "loss": 0.5397, + "step": 15217 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011992674082847942, + "loss": 0.5679, + "step": 15218 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011991729170578294, + "loss": 0.5704, + "step": 15219 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011990784239790344, + "loss": 0.5098, + "step": 15220 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011989839290492876, + "loss": 0.4162, + "step": 15221 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011988894322694677, + "loss": 0.6138, + "step": 15222 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011987949336404529, + "loss": 0.5413, + "step": 15223 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001198700433163122, + "loss": 0.6202, + "step": 15224 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011986059308383537, + "loss": 0.5694, + "step": 15225 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011985114266670267, + "loss": 0.6212, + "step": 15226 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011984169206500195, + "loss": 0.4845, + "step": 15227 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011983224127882109, + "loss": 0.5146, + "step": 15228 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011982279030824797, + "loss": 0.5539, + "step": 15229 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011981333915337041, + "loss": 0.4892, + "step": 15230 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011980388781427636, + "loss": 0.5956, + "step": 15231 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001197944362910536, + "loss": 0.5244, + "step": 15232 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011978498458379011, + "loss": 0.4933, + "step": 15233 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011977553269257368, + "loss": 0.5616, + "step": 15234 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011976608061749226, + "loss": 0.6536, + "step": 15235 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001197566283586337, + "loss": 0.5788, + "step": 15236 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011974717591608589, + "loss": 0.6159, + "step": 15237 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011973772328993669, + "loss": 0.4918, + "step": 15238 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011972827048027399, + "loss": 0.4547, + "step": 15239 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001197188174871857, + "loss": 0.5068, + "step": 15240 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011970936431075974, + "loss": 0.551, + "step": 15241 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011969991095108393, + "loss": 0.4988, + "step": 15242 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011969045740824622, + "loss": 0.4401, + "step": 15243 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001196810036823345, + "loss": 0.5807, + "step": 15244 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001196715497734366, + "loss": 0.5756, + "step": 15245 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011966209568164053, + "loss": 0.5991, + "step": 15246 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011965264140703407, + "loss": 0.4897, + "step": 15247 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011964318694970525, + "loss": 0.6536, + "step": 15248 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011963373230974184, + "loss": 0.5651, + "step": 15249 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011962427748723183, + "loss": 0.5744, + "step": 15250 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011961482248226315, + "loss": 0.4671, + "step": 15251 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011960536729492365, + "loss": 0.4885, + "step": 15252 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011959591192530125, + "loss": 0.5713, + "step": 15253 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011958645637348385, + "loss": 0.6063, + "step": 15254 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001195770006395594, + "loss": 0.475, + "step": 15255 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011956754472361583, + "loss": 0.543, + "step": 15256 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011955808862574098, + "loss": 0.5162, + "step": 15257 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011954863234602285, + "loss": 0.5369, + "step": 15258 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001195391758845493, + "loss": 0.4361, + "step": 15259 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011952971924140832, + "loss": 0.4919, + "step": 15260 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011952026241668777, + "loss": 0.5615, + "step": 15261 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011951080541047559, + "loss": 0.4642, + "step": 15262 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011950134822285975, + "loss": 0.5069, + "step": 15263 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011949189085392811, + "loss": 0.5307, + "step": 15264 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011948243330376867, + "loss": 0.6614, + "step": 15265 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011947297557246931, + "loss": 0.553, + "step": 15266 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011946351766011798, + "loss": 0.5406, + "step": 15267 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011945405956680265, + "loss": 0.4954, + "step": 15268 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011944460129261119, + "loss": 0.6209, + "step": 15269 + }, + { + "epoch": 0.45, + "learning_rate": 0.0001194351428376316, + "loss": 0.6577, + "step": 15270 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011942568420195177, + "loss": 0.6124, + "step": 15271 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011941622538565973, + "loss": 0.5578, + "step": 15272 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011940676638884331, + "loss": 0.5161, + "step": 15273 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011939730721159055, + "loss": 0.4504, + "step": 15274 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011938784785398931, + "loss": 0.552, + "step": 15275 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011937838831612762, + "loss": 0.5027, + "step": 15276 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011936892859809341, + "loss": 0.6854, + "step": 15277 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011935946869997459, + "loss": 0.5133, + "step": 15278 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011935000862185918, + "loss": 0.4493, + "step": 15279 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011934054836383509, + "loss": 0.4792, + "step": 15280 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011933108792599026, + "loss": 0.675, + "step": 15281 + }, + { + "epoch": 0.45, + "learning_rate": 0.00011932162730841273, + "loss": 0.4622, + "step": 15282 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011931216651119037, + "loss": 0.5256, + "step": 15283 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011930270553441118, + "loss": 0.6484, + "step": 15284 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011929324437816316, + "loss": 0.6159, + "step": 15285 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011928378304253422, + "loss": 0.5752, + "step": 15286 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011927432152761235, + "loss": 0.5451, + "step": 15287 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011926485983348551, + "loss": 0.5324, + "step": 15288 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011925539796024173, + "loss": 0.4551, + "step": 15289 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011924593590796888, + "loss": 0.5548, + "step": 15290 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011923647367675503, + "loss": 0.4261, + "step": 15291 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011922701126668809, + "loss": 0.6139, + "step": 15292 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011921754867785606, + "loss": 0.5491, + "step": 15293 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011920808591034694, + "loss": 0.4913, + "step": 15294 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011919862296424869, + "loss": 0.5784, + "step": 15295 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011918915983964929, + "loss": 0.5047, + "step": 15296 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011917969653663672, + "loss": 0.5158, + "step": 15297 + }, + { + "epoch": 0.46, + "learning_rate": 0.000119170233055299, + "loss": 0.4875, + "step": 15298 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011916076939572407, + "loss": 0.4035, + "step": 15299 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011915130555799996, + "loss": 0.5848, + "step": 15300 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011914184154221462, + "loss": 0.4902, + "step": 15301 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011913237734845605, + "loss": 0.5813, + "step": 15302 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011912291297681233, + "loss": 0.6197, + "step": 15303 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011911344842737134, + "loss": 0.4887, + "step": 15304 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011910398370022116, + "loss": 0.6161, + "step": 15305 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011909451879544971, + "loss": 0.5385, + "step": 15306 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011908505371314505, + "loss": 0.5438, + "step": 15307 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011907558845339516, + "loss": 0.4415, + "step": 15308 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011906612301628806, + "loss": 0.5275, + "step": 15309 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011905665740191176, + "loss": 0.5644, + "step": 15310 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011904719161035422, + "loss": 0.5874, + "step": 15311 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011903772564170352, + "loss": 0.5613, + "step": 15312 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011902825949604763, + "loss": 0.442, + "step": 15313 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011901879317347455, + "loss": 0.617, + "step": 15314 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011900932667407231, + "loss": 0.5235, + "step": 15315 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011899985999792895, + "loss": 0.5414, + "step": 15316 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011899039314513245, + "loss": 0.4862, + "step": 15317 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011898092611577085, + "loss": 0.5838, + "step": 15318 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011897145890993213, + "loss": 0.505, + "step": 15319 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001189619915277044, + "loss": 0.499, + "step": 15320 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011895252396917558, + "loss": 0.4244, + "step": 15321 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001189430562344338, + "loss": 0.6147, + "step": 15322 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011893358832356698, + "loss": 0.5817, + "step": 15323 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011892412023666324, + "loss": 0.5287, + "step": 15324 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011891465197381056, + "loss": 0.51, + "step": 15325 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011890518353509697, + "loss": 0.5953, + "step": 15326 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011889571492061054, + "loss": 0.63, + "step": 15327 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011888624613043926, + "loss": 0.4904, + "step": 15328 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001188767771646712, + "loss": 0.5769, + "step": 15329 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011886730802339441, + "loss": 0.4891, + "step": 15330 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011885783870669688, + "loss": 0.5138, + "step": 15331 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001188483692146667, + "loss": 0.5417, + "step": 15332 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011883889954739189, + "loss": 0.4751, + "step": 15333 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011882942970496052, + "loss": 0.5153, + "step": 15334 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011881995968746058, + "loss": 0.5059, + "step": 15335 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011881048949498019, + "loss": 0.5369, + "step": 15336 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011880101912760732, + "loss": 0.5898, + "step": 15337 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011879154858543012, + "loss": 0.6141, + "step": 15338 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011878207786853657, + "loss": 0.5809, + "step": 15339 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011877260697701473, + "loss": 0.5109, + "step": 15340 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001187631359109527, + "loss": 0.5921, + "step": 15341 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011875366467043849, + "loss": 0.6056, + "step": 15342 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011874419325556018, + "loss": 0.4789, + "step": 15343 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011873472166640587, + "loss": 0.6064, + "step": 15344 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011872524990306355, + "loss": 0.6701, + "step": 15345 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011871577796562132, + "loss": 0.555, + "step": 15346 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011870630585416724, + "loss": 0.5324, + "step": 15347 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001186968335687894, + "loss": 0.6809, + "step": 15348 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011868736110957586, + "loss": 0.6663, + "step": 15349 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011867788847661466, + "loss": 0.6367, + "step": 15350 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011866841566999392, + "loss": 0.4963, + "step": 15351 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011865894268980167, + "loss": 0.5436, + "step": 15352 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011864946953612604, + "loss": 0.5985, + "step": 15353 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011863999620905501, + "loss": 0.5335, + "step": 15354 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011863052270867678, + "loss": 0.5109, + "step": 15355 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011862104903507936, + "loss": 0.5092, + "step": 15356 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011861157518835087, + "loss": 0.521, + "step": 15357 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011860210116857935, + "loss": 0.542, + "step": 15358 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001185926269758529, + "loss": 0.525, + "step": 15359 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011858315261025963, + "loss": 0.5014, + "step": 15360 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011857367807188761, + "loss": 0.4474, + "step": 15361 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011856420336082492, + "loss": 0.5287, + "step": 15362 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011855472847715969, + "loss": 0.55, + "step": 15363 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011854525342097994, + "loss": 0.6321, + "step": 15364 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011853577819237389, + "loss": 0.5274, + "step": 15365 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011852630279142951, + "loss": 0.5287, + "step": 15366 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011851682721823498, + "loss": 0.5856, + "step": 15367 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011850735147287834, + "loss": 0.4463, + "step": 15368 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011849787555544774, + "loss": 0.4875, + "step": 15369 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011848839946603128, + "loss": 0.5813, + "step": 15370 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011847892320471703, + "loss": 0.5876, + "step": 15371 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001184694467715931, + "loss": 0.6501, + "step": 15372 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011845997016674766, + "loss": 0.5709, + "step": 15373 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011845049339026875, + "loss": 0.5527, + "step": 15374 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001184410164422445, + "loss": 0.4961, + "step": 15375 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011843153932276304, + "loss": 0.5916, + "step": 15376 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011842206203191248, + "loss": 0.6462, + "step": 15377 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001184125845697809, + "loss": 0.5548, + "step": 15378 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011840310693645649, + "loss": 0.5089, + "step": 15379 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001183936291320273, + "loss": 0.5865, + "step": 15380 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011838415115658146, + "loss": 0.583, + "step": 15381 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011837467301020715, + "loss": 0.5588, + "step": 15382 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011836519469299241, + "loss": 0.5647, + "step": 15383 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011835571620502546, + "loss": 0.5248, + "step": 15384 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011834623754639433, + "loss": 0.5588, + "step": 15385 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011833675871718721, + "loss": 0.4339, + "step": 15386 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011832727971749222, + "loss": 0.5175, + "step": 15387 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001183178005473975, + "loss": 0.5171, + "step": 15388 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011830832120699116, + "loss": 0.5805, + "step": 15389 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011829884169636134, + "loss": 0.4315, + "step": 15390 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011828936201559619, + "loss": 0.6897, + "step": 15391 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011827988216478385, + "loss": 0.5214, + "step": 15392 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011827040214401244, + "loss": 0.6706, + "step": 15393 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011826092195337013, + "loss": 0.6981, + "step": 15394 + }, + { + "epoch": 0.46, + "learning_rate": 0.000118251441592945, + "loss": 0.4899, + "step": 15395 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001182419610628253, + "loss": 0.4762, + "step": 15396 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011823248036309906, + "loss": 0.4814, + "step": 15397 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011822299949385454, + "loss": 0.5957, + "step": 15398 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011821351845517979, + "loss": 0.4974, + "step": 15399 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011820403724716302, + "loss": 0.5109, + "step": 15400 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011819455586989237, + "loss": 0.5604, + "step": 15401 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011818507432345598, + "loss": 0.4245, + "step": 15402 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011817559260794202, + "loss": 0.4822, + "step": 15403 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011816611072343866, + "loss": 0.4858, + "step": 15404 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011815662867003402, + "loss": 0.6108, + "step": 15405 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011814714644781631, + "loss": 0.5685, + "step": 15406 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011813766405687363, + "loss": 0.5319, + "step": 15407 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011812818149729424, + "loss": 0.4648, + "step": 15408 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001181186987691662, + "loss": 0.64, + "step": 15409 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011810921587257771, + "loss": 0.4861, + "step": 15410 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011809973280761696, + "loss": 0.4935, + "step": 15411 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011809024957437209, + "loss": 0.5841, + "step": 15412 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011808076617293134, + "loss": 0.5959, + "step": 15413 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011807128260338278, + "loss": 0.516, + "step": 15414 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011806179886581468, + "loss": 0.5418, + "step": 15415 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011805231496031513, + "loss": 0.7737, + "step": 15416 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011804283088697236, + "loss": 0.5859, + "step": 15417 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011803334664587455, + "loss": 0.6512, + "step": 15418 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011802386223710988, + "loss": 0.4516, + "step": 15419 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011801437766076651, + "loss": 0.626, + "step": 15420 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011800489291693264, + "loss": 0.6197, + "step": 15421 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011799540800569646, + "loss": 0.6911, + "step": 15422 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011798592292714614, + "loss": 0.6021, + "step": 15423 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011797643768136985, + "loss": 0.5249, + "step": 15424 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011796695226845586, + "loss": 0.6007, + "step": 15425 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011795746668849228, + "loss": 0.5681, + "step": 15426 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011794798094156736, + "loss": 0.4538, + "step": 15427 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011793849502776922, + "loss": 0.5357, + "step": 15428 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011792900894718612, + "loss": 0.6304, + "step": 15429 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011791952269990625, + "loss": 0.4216, + "step": 15430 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001179100362860178, + "loss": 0.5658, + "step": 15431 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011790054970560896, + "loss": 0.5182, + "step": 15432 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011789106295876795, + "loss": 0.4858, + "step": 15433 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011788157604558297, + "loss": 0.5229, + "step": 15434 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011787208896614222, + "loss": 0.5076, + "step": 15435 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011786260172053392, + "loss": 0.5554, + "step": 15436 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011785311430884624, + "loss": 0.5219, + "step": 15437 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011784362673116743, + "loss": 0.5508, + "step": 15438 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001178341389875857, + "loss": 0.4295, + "step": 15439 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011782465107818925, + "loss": 0.5613, + "step": 15440 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011781516300306632, + "loss": 0.4962, + "step": 15441 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011780567476230509, + "loss": 0.5969, + "step": 15442 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011779618635599377, + "loss": 0.5388, + "step": 15443 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011778669778422062, + "loss": 0.5674, + "step": 15444 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011777720904707383, + "loss": 0.5397, + "step": 15445 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011776772014464169, + "loss": 0.5951, + "step": 15446 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001177582310770123, + "loss": 0.5265, + "step": 15447 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011774874184427398, + "loss": 0.5039, + "step": 15448 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011773925244651494, + "loss": 0.6494, + "step": 15449 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011772976288382338, + "loss": 0.5181, + "step": 15450 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011772027315628758, + "loss": 0.4731, + "step": 15451 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011771078326399572, + "loss": 0.5866, + "step": 15452 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011770129320703607, + "loss": 0.537, + "step": 15453 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011769180298549684, + "loss": 0.6092, + "step": 15454 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011768231259946627, + "loss": 0.4532, + "step": 15455 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011767282204903265, + "loss": 0.5887, + "step": 15456 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011766333133428411, + "loss": 0.4762, + "step": 15457 + }, + { + "epoch": 0.46, + "learning_rate": 0.000117653840455309, + "loss": 0.5842, + "step": 15458 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011764434941219551, + "loss": 0.4991, + "step": 15459 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011763485820503188, + "loss": 0.5658, + "step": 15460 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011762536683390636, + "loss": 0.5057, + "step": 15461 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011761587529890723, + "loss": 0.5688, + "step": 15462 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011760638360012271, + "loss": 0.6119, + "step": 15463 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011759689173764104, + "loss": 0.5458, + "step": 15464 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011758739971155048, + "loss": 0.504, + "step": 15465 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011757790752193931, + "loss": 0.5053, + "step": 15466 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011756841516889575, + "loss": 0.5484, + "step": 15467 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011755892265250805, + "loss": 0.5822, + "step": 15468 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011754942997286449, + "loss": 0.5613, + "step": 15469 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011753993713005335, + "loss": 0.6526, + "step": 15470 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011753044412416283, + "loss": 0.5594, + "step": 15471 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011752095095528126, + "loss": 0.4152, + "step": 15472 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011751145762349684, + "loss": 0.4003, + "step": 15473 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011750196412889789, + "loss": 0.5882, + "step": 15474 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011749247047157266, + "loss": 0.6538, + "step": 15475 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011748297665160938, + "loss": 0.5106, + "step": 15476 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011747348266909639, + "loss": 0.5102, + "step": 15477 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011746398852412188, + "loss": 0.5471, + "step": 15478 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011745449421677418, + "loss": 0.5976, + "step": 15479 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011744499974714158, + "loss": 0.667, + "step": 15480 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011743550511531229, + "loss": 0.5507, + "step": 15481 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011742601032137464, + "loss": 0.57, + "step": 15482 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011741651536541689, + "loss": 0.5988, + "step": 15483 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011740702024752732, + "loss": 0.5263, + "step": 15484 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011739752496779421, + "loss": 0.4987, + "step": 15485 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011738802952630583, + "loss": 0.483, + "step": 15486 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011737853392315052, + "loss": 0.5408, + "step": 15487 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011736903815841648, + "loss": 0.5509, + "step": 15488 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001173595422321921, + "loss": 0.5291, + "step": 15489 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011735004614456558, + "loss": 0.6357, + "step": 15490 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011734054989562524, + "loss": 0.3853, + "step": 15491 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011733105348545941, + "loss": 0.5229, + "step": 15492 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011732155691415634, + "loss": 0.6898, + "step": 15493 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011731206018180431, + "loss": 0.6243, + "step": 15494 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011730256328849168, + "loss": 0.4543, + "step": 15495 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001172930662343067, + "loss": 0.5252, + "step": 15496 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011728356901933768, + "loss": 0.6107, + "step": 15497 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011727407164367292, + "loss": 0.4879, + "step": 15498 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011726457410740075, + "loss": 0.4848, + "step": 15499 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001172550764106094, + "loss": 0.5255, + "step": 15500 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011724557855338729, + "loss": 0.5808, + "step": 15501 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011723608053582263, + "loss": 0.65, + "step": 15502 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011722658235800377, + "loss": 0.5901, + "step": 15503 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011721708402001898, + "loss": 0.4395, + "step": 15504 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011720758552195665, + "loss": 0.5448, + "step": 15505 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011719808686390504, + "loss": 0.6014, + "step": 15506 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011718858804595246, + "loss": 0.5794, + "step": 15507 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011717908906818726, + "loss": 0.4884, + "step": 15508 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001171695899306977, + "loss": 0.5526, + "step": 15509 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011716009063357216, + "loss": 0.6833, + "step": 15510 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011715059117689895, + "loss": 0.5062, + "step": 15511 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011714109156076636, + "loss": 0.5221, + "step": 15512 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011713159178526273, + "loss": 0.5472, + "step": 15513 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001171220918504764, + "loss": 0.4928, + "step": 15514 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011711259175649567, + "loss": 0.5756, + "step": 15515 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001171030915034089, + "loss": 0.5514, + "step": 15516 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011709359109130436, + "loss": 0.536, + "step": 15517 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011708409052027047, + "loss": 0.6255, + "step": 15518 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011707458979039549, + "loss": 0.575, + "step": 15519 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011706508890176782, + "loss": 0.5454, + "step": 15520 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001170555878544757, + "loss": 0.5172, + "step": 15521 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011704608664860755, + "loss": 0.5805, + "step": 15522 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011703658528425166, + "loss": 0.5187, + "step": 15523 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001170270837614964, + "loss": 0.4376, + "step": 15524 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011701758208043012, + "loss": 0.6181, + "step": 15525 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011700808024114114, + "loss": 0.4389, + "step": 15526 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011699857824371781, + "loss": 0.5682, + "step": 15527 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011698907608824845, + "loss": 0.5584, + "step": 15528 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011697957377482144, + "loss": 0.4857, + "step": 15529 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011697007130352514, + "loss": 0.5698, + "step": 15530 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011696056867444784, + "loss": 0.5754, + "step": 15531 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011695106588767799, + "loss": 0.553, + "step": 15532 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011694156294330383, + "loss": 0.5859, + "step": 15533 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011693205984141382, + "loss": 0.6115, + "step": 15534 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011692255658209621, + "loss": 0.5242, + "step": 15535 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011691305316543945, + "loss": 0.5043, + "step": 15536 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011690354959153187, + "loss": 0.489, + "step": 15537 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001168940458604618, + "loss": 0.5034, + "step": 15538 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011688454197231764, + "loss": 0.5176, + "step": 15539 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011687503792718773, + "loss": 0.5689, + "step": 15540 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011686553372516045, + "loss": 0.5991, + "step": 15541 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011685602936632416, + "loss": 0.5801, + "step": 15542 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011684652485076722, + "loss": 0.4774, + "step": 15543 + }, + { + "epoch": 0.46, + "learning_rate": 0.000116837020178578, + "loss": 0.5723, + "step": 15544 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011682751534984491, + "loss": 0.5449, + "step": 15545 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011681801036465626, + "loss": 0.5827, + "step": 15546 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011680850522310045, + "loss": 0.5415, + "step": 15547 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011679899992526587, + "loss": 0.6794, + "step": 15548 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011678949447124089, + "loss": 0.639, + "step": 15549 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011677998886111386, + "loss": 0.4801, + "step": 15550 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001167704830949732, + "loss": 0.6252, + "step": 15551 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011676097717290723, + "loss": 0.575, + "step": 15552 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011675147109500444, + "loss": 0.5848, + "step": 15553 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011674196486135311, + "loss": 0.5157, + "step": 15554 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011673245847204166, + "loss": 0.43, + "step": 15555 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001167229519271585, + "loss": 0.6298, + "step": 15556 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011671344522679199, + "loss": 0.5322, + "step": 15557 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011670393837103053, + "loss": 0.506, + "step": 15558 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001166944313599625, + "loss": 0.5868, + "step": 15559 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001166849241936763, + "loss": 0.631, + "step": 15560 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011667541687226036, + "loss": 0.5819, + "step": 15561 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011666590939580297, + "loss": 0.5902, + "step": 15562 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011665640176439267, + "loss": 0.6513, + "step": 15563 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011664689397811775, + "loss": 0.5933, + "step": 15564 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011663738603706667, + "loss": 0.6133, + "step": 15565 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011662787794132774, + "loss": 0.4644, + "step": 15566 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011661836969098949, + "loss": 0.5129, + "step": 15567 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011660886128614023, + "loss": 0.5353, + "step": 15568 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011659935272686841, + "loss": 0.4979, + "step": 15569 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011658984401326243, + "loss": 0.5499, + "step": 15570 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011658033514541069, + "loss": 0.5013, + "step": 15571 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001165708261234016, + "loss": 0.4971, + "step": 15572 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011656131694732357, + "loss": 0.5539, + "step": 15573 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011655180761726503, + "loss": 0.5913, + "step": 15574 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011654229813331436, + "loss": 0.5884, + "step": 15575 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011653278849556002, + "loss": 0.6024, + "step": 15576 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011652327870409039, + "loss": 0.4848, + "step": 15577 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001165137687589939, + "loss": 0.6352, + "step": 15578 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011650425866035893, + "loss": 0.4925, + "step": 15579 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011649474840827399, + "loss": 0.4373, + "step": 15580 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011648523800282743, + "loss": 0.5183, + "step": 15581 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011647572744410773, + "loss": 0.6998, + "step": 15582 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011646621673220324, + "loss": 0.5849, + "step": 15583 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011645670586720245, + "loss": 0.6118, + "step": 15584 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011644719484919378, + "loss": 0.5237, + "step": 15585 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011643768367826562, + "loss": 0.6204, + "step": 15586 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011642817235450644, + "loss": 0.5058, + "step": 15587 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011641866087800469, + "loss": 0.5746, + "step": 15588 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011640914924884874, + "loss": 0.6685, + "step": 15589 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011639963746712708, + "loss": 0.4713, + "step": 15590 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011639012553292808, + "loss": 0.6628, + "step": 15591 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011638061344634028, + "loss": 0.5526, + "step": 15592 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011637110120745202, + "loss": 0.3802, + "step": 15593 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011636158881635184, + "loss": 0.4422, + "step": 15594 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001163520762731281, + "loss": 0.6992, + "step": 15595 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011634256357786926, + "loss": 0.5202, + "step": 15596 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011633305073066381, + "loss": 0.5194, + "step": 15597 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011632353773160014, + "loss": 0.4644, + "step": 15598 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011631402458076673, + "loss": 0.5873, + "step": 15599 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011630451127825201, + "loss": 0.5098, + "step": 15600 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011629499782414446, + "loss": 0.4739, + "step": 15601 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011628548421853249, + "loss": 0.6051, + "step": 15602 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011627597046150461, + "loss": 0.5695, + "step": 15603 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011626645655314922, + "loss": 0.492, + "step": 15604 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011625694249355479, + "loss": 0.4986, + "step": 15605 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011624742828280984, + "loss": 0.5917, + "step": 15606 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011623791392100274, + "loss": 0.5827, + "step": 15607 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011622839940822199, + "loss": 0.577, + "step": 15608 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011621888474455605, + "loss": 0.6101, + "step": 15609 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011620936993009336, + "loss": 0.5067, + "step": 15610 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011619985496492244, + "loss": 0.586, + "step": 15611 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011619033984913173, + "loss": 0.6376, + "step": 15612 + }, + { + "epoch": 0.46, + "learning_rate": 0.0001161808245828097, + "loss": 0.5661, + "step": 15613 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011617130916604477, + "loss": 0.4566, + "step": 15614 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011616179359892547, + "loss": 0.511, + "step": 15615 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011615227788154026, + "loss": 0.6217, + "step": 15616 + }, + { + "epoch": 0.46, + "learning_rate": 0.00011614276201397762, + "loss": 0.5162, + "step": 15617 + }, + { + "epoch": 0.46, + "learning_rate": 0.000116133245996326, + "loss": 0.5093, + "step": 15618 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001161237298286739, + "loss": 0.5512, + "step": 15619 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011611421351110979, + "loss": 0.5584, + "step": 15620 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011610469704372211, + "loss": 0.415, + "step": 15621 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001160951804265994, + "loss": 0.5268, + "step": 15622 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011608566365983013, + "loss": 0.5623, + "step": 15623 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011607614674350273, + "loss": 0.4911, + "step": 15624 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011606662967770578, + "loss": 0.5162, + "step": 15625 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011605711246252766, + "loss": 0.4521, + "step": 15626 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011604759509805694, + "loss": 0.5253, + "step": 15627 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011603807758438205, + "loss": 0.6318, + "step": 15628 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011602855992159154, + "loss": 0.5651, + "step": 15629 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011601904210977384, + "loss": 0.3998, + "step": 15630 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011600952414901749, + "loss": 0.48, + "step": 15631 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011600000603941094, + "loss": 0.5858, + "step": 15632 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011599048778104275, + "loss": 0.7435, + "step": 15633 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011598096937400136, + "loss": 0.5939, + "step": 15634 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011597145081837528, + "loss": 0.5641, + "step": 15635 + }, + { + "epoch": 0.47, + "learning_rate": 0.000115961932114253, + "loss": 0.5635, + "step": 15636 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011595241326172308, + "loss": 0.5037, + "step": 15637 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011594289426087392, + "loss": 0.5573, + "step": 15638 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011593337511179414, + "loss": 0.5543, + "step": 15639 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011592385581457217, + "loss": 0.6026, + "step": 15640 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011591433636929651, + "loss": 0.5632, + "step": 15641 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011590481677605573, + "loss": 0.5193, + "step": 15642 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011589529703493829, + "loss": 0.5405, + "step": 15643 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011588577714603273, + "loss": 0.5426, + "step": 15644 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011587625710942751, + "loss": 0.4958, + "step": 15645 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011586673692521119, + "loss": 0.6133, + "step": 15646 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011585721659347228, + "loss": 0.5793, + "step": 15647 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011584769611429929, + "loss": 0.5058, + "step": 15648 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011583817548778073, + "loss": 0.4742, + "step": 15649 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011582865471400514, + "loss": 0.5147, + "step": 15650 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011581913379306105, + "loss": 0.5672, + "step": 15651 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011580961272503692, + "loss": 0.6131, + "step": 15652 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011580009151002129, + "loss": 0.5549, + "step": 15653 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011579057014810278, + "loss": 0.5897, + "step": 15654 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011578104863936976, + "loss": 0.5332, + "step": 15655 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011577152698391092, + "loss": 0.6695, + "step": 15656 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011576200518181463, + "loss": 0.6047, + "step": 15657 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011575248323316954, + "loss": 0.6146, + "step": 15658 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011574296113806416, + "loss": 0.6039, + "step": 15659 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011573343889658696, + "loss": 0.5641, + "step": 15660 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011572391650882654, + "loss": 0.5704, + "step": 15661 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001157143939748714, + "loss": 0.5283, + "step": 15662 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011570487129481009, + "loss": 0.452, + "step": 15663 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011569534846873116, + "loss": 0.5269, + "step": 15664 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011568582549672313, + "loss": 0.4801, + "step": 15665 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011567630237887455, + "loss": 0.6405, + "step": 15666 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011566677911527391, + "loss": 0.6139, + "step": 15667 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011565725570600987, + "loss": 0.5555, + "step": 15668 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011564773215117087, + "loss": 0.4958, + "step": 15669 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011563820845084553, + "loss": 0.5574, + "step": 15670 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011562868460512228, + "loss": 0.5893, + "step": 15671 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011561916061408981, + "loss": 0.4476, + "step": 15672 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011560963647783662, + "loss": 0.5282, + "step": 15673 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011560011219645122, + "loss": 0.6159, + "step": 15674 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011559058777002222, + "loss": 0.511, + "step": 15675 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011558106319863813, + "loss": 0.5885, + "step": 15676 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011557153848238753, + "loss": 0.6218, + "step": 15677 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011556201362135897, + "loss": 0.5523, + "step": 15678 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011555248861564101, + "loss": 0.6806, + "step": 15679 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011554296346532221, + "loss": 0.5515, + "step": 15680 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011553343817049114, + "loss": 0.5603, + "step": 15681 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011552391273123633, + "loss": 0.5638, + "step": 15682 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001155143871476464, + "loss": 0.4453, + "step": 15683 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011550486141980983, + "loss": 0.5875, + "step": 15684 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011549533554781531, + "loss": 0.6639, + "step": 15685 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011548580953175125, + "loss": 0.6236, + "step": 15686 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001154762833717064, + "loss": 0.4895, + "step": 15687 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011546675706776915, + "loss": 0.4819, + "step": 15688 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001154572306200282, + "loss": 0.5769, + "step": 15689 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011544770402857207, + "loss": 0.5555, + "step": 15690 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011543817729348935, + "loss": 0.5218, + "step": 15691 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011542865041486862, + "loss": 0.485, + "step": 15692 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011541912339279843, + "loss": 0.61, + "step": 15693 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011540959622736738, + "loss": 0.5348, + "step": 15694 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011540006891866406, + "loss": 0.562, + "step": 15695 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011539054146677703, + "loss": 0.4617, + "step": 15696 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011538101387179488, + "loss": 0.5283, + "step": 15697 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011537148613380617, + "loss": 0.4119, + "step": 15698 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011536195825289955, + "loss": 0.5347, + "step": 15699 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011535243022916351, + "loss": 0.6387, + "step": 15700 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011534290206268674, + "loss": 0.6395, + "step": 15701 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011533337375355775, + "loss": 0.6326, + "step": 15702 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001153238453018652, + "loss": 0.3813, + "step": 15703 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011531431670769764, + "loss": 0.5508, + "step": 15704 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011530478797114361, + "loss": 0.4918, + "step": 15705 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011529525909229184, + "loss": 0.5441, + "step": 15706 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011528573007123078, + "loss": 0.6009, + "step": 15707 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011527620090804913, + "loss": 0.5004, + "step": 15708 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011526667160283544, + "loss": 0.559, + "step": 15709 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011525714215567833, + "loss": 0.5163, + "step": 15710 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011524761256666638, + "loss": 0.5546, + "step": 15711 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011523808283588824, + "loss": 0.5543, + "step": 15712 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011522855296343245, + "loss": 0.4558, + "step": 15713 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011521902294938764, + "loss": 0.4278, + "step": 15714 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011520949279384241, + "loss": 0.5063, + "step": 15715 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011519996249688542, + "loss": 0.4637, + "step": 15716 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001151904320586052, + "loss": 0.7085, + "step": 15717 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011518090147909042, + "loss": 0.5964, + "step": 15718 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011517137075842965, + "loss": 0.5979, + "step": 15719 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011516183989671152, + "loss": 0.524, + "step": 15720 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011515230889402467, + "loss": 0.5361, + "step": 15721 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011514277775045768, + "loss": 0.5272, + "step": 15722 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011513324646609918, + "loss": 0.5722, + "step": 15723 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011512371504103779, + "loss": 0.4847, + "step": 15724 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001151141834753621, + "loss": 0.5696, + "step": 15725 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011510465176916077, + "loss": 0.6028, + "step": 15726 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001150951199225224, + "loss": 0.4603, + "step": 15727 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011508558793553565, + "loss": 0.5356, + "step": 15728 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011507605580828907, + "loss": 0.5309, + "step": 15729 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001150665235408714, + "loss": 0.4813, + "step": 15730 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011505699113337111, + "loss": 0.5908, + "step": 15731 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011504745858587698, + "loss": 0.6289, + "step": 15732 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011503792589847757, + "loss": 0.4771, + "step": 15733 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001150283930712615, + "loss": 0.5786, + "step": 15734 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011501886010431744, + "loss": 0.6666, + "step": 15735 + }, + { + "epoch": 0.47, + "learning_rate": 0.000115009326997734, + "loss": 0.5759, + "step": 15736 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011499979375159984, + "loss": 0.5598, + "step": 15737 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011499026036600353, + "loss": 0.5424, + "step": 15738 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011498072684103379, + "loss": 0.4972, + "step": 15739 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001149711931767792, + "loss": 0.4848, + "step": 15740 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011496165937332844, + "loss": 0.4719, + "step": 15741 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011495212543077012, + "loss": 0.5665, + "step": 15742 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011494259134919289, + "loss": 0.5703, + "step": 15743 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011493305712868543, + "loss": 0.6425, + "step": 15744 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011492352276933633, + "loss": 0.5964, + "step": 15745 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011491398827123427, + "loss": 0.5682, + "step": 15746 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011490445363446792, + "loss": 0.5711, + "step": 15747 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011489491885912585, + "loss": 0.5165, + "step": 15748 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001148853839452968, + "loss": 0.4829, + "step": 15749 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011487584889306935, + "loss": 0.4771, + "step": 15750 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011486631370253222, + "loss": 0.59, + "step": 15751 + }, + { + "epoch": 0.47, + "learning_rate": 0.000114856778373774, + "loss": 0.4569, + "step": 15752 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001148472429068834, + "loss": 0.6114, + "step": 15753 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011483770730194905, + "loss": 0.611, + "step": 15754 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011482817155905961, + "loss": 0.5718, + "step": 15755 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011481863567830373, + "loss": 0.6244, + "step": 15756 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011480909965977009, + "loss": 0.5837, + "step": 15757 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011479956350354733, + "loss": 0.64, + "step": 15758 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011479002720972416, + "loss": 0.7224, + "step": 15759 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011478049077838917, + "loss": 0.5387, + "step": 15760 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011477095420963111, + "loss": 0.4523, + "step": 15761 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011476141750353857, + "loss": 0.4256, + "step": 15762 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011475188066020027, + "loss": 0.5156, + "step": 15763 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011474234367970485, + "loss": 0.4584, + "step": 15764 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011473280656214101, + "loss": 0.5892, + "step": 15765 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001147232693075974, + "loss": 0.5168, + "step": 15766 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001147137319161627, + "loss": 0.4702, + "step": 15767 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011470419438792559, + "loss": 0.6711, + "step": 15768 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011469465672297473, + "loss": 0.5531, + "step": 15769 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011468511892139881, + "loss": 0.4909, + "step": 15770 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011467558098328653, + "loss": 0.576, + "step": 15771 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011466604290872652, + "loss": 0.505, + "step": 15772 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011465650469780751, + "loss": 0.5432, + "step": 15773 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011464696635061814, + "loss": 0.4927, + "step": 15774 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011463742786724712, + "loss": 0.5655, + "step": 15775 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011462788924778314, + "loss": 0.5576, + "step": 15776 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011461835049231484, + "loss": 0.5946, + "step": 15777 + }, + { + "epoch": 0.47, + "learning_rate": 0.000114608811600931, + "loss": 0.5865, + "step": 15778 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001145992725737202, + "loss": 0.588, + "step": 15779 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011458973341077124, + "loss": 0.4702, + "step": 15780 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011458019411217268, + "loss": 0.637, + "step": 15781 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011457065467801333, + "loss": 0.4714, + "step": 15782 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011456111510838185, + "loss": 0.4845, + "step": 15783 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011455157540336692, + "loss": 0.6558, + "step": 15784 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011454203556305722, + "loss": 0.6767, + "step": 15785 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011453249558754148, + "loss": 0.5312, + "step": 15786 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011452295547690841, + "loss": 0.5794, + "step": 15787 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011451341523124668, + "loss": 0.5419, + "step": 15788 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011450387485064496, + "loss": 0.5557, + "step": 15789 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011449433433519204, + "loss": 0.5765, + "step": 15790 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011448479368497653, + "loss": 0.6922, + "step": 15791 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011447525290008724, + "loss": 0.5487, + "step": 15792 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011446571198061275, + "loss": 0.5459, + "step": 15793 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011445617092664189, + "loss": 0.5018, + "step": 15794 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001144466297382633, + "loss": 0.5411, + "step": 15795 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011443708841556571, + "loss": 0.5967, + "step": 15796 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001144275469586378, + "loss": 0.5469, + "step": 15797 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011441800536756833, + "loss": 0.5065, + "step": 15798 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011440846364244599, + "loss": 0.6215, + "step": 15799 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011439892178335949, + "loss": 0.4798, + "step": 15800 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011438937979039757, + "loss": 0.6018, + "step": 15801 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011437983766364893, + "loss": 0.5959, + "step": 15802 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011437029540320227, + "loss": 0.5256, + "step": 15803 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011436075300914635, + "loss": 0.5198, + "step": 15804 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011435121048156987, + "loss": 0.5872, + "step": 15805 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011434166782056155, + "loss": 0.4826, + "step": 15806 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011433212502621012, + "loss": 0.5092, + "step": 15807 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011432258209860427, + "loss": 0.6668, + "step": 15808 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011431303903783281, + "loss": 0.4967, + "step": 15809 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011430349584398438, + "loss": 0.5356, + "step": 15810 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011429395251714781, + "loss": 0.6638, + "step": 15811 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011428440905741169, + "loss": 0.5584, + "step": 15812 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011427486546486485, + "loss": 0.5151, + "step": 15813 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011426532173959603, + "loss": 0.5297, + "step": 15814 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011425577788169391, + "loss": 0.5353, + "step": 15815 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011424623389124725, + "loss": 0.5907, + "step": 15816 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011423668976834481, + "loss": 0.5267, + "step": 15817 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011422714551307528, + "loss": 0.5829, + "step": 15818 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011421760112552741, + "loss": 0.4926, + "step": 15819 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011420805660578995, + "loss": 0.4576, + "step": 15820 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011419851195395168, + "loss": 0.5223, + "step": 15821 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011418896717010126, + "loss": 0.589, + "step": 15822 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011417942225432751, + "loss": 0.4727, + "step": 15823 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011416987720671913, + "loss": 0.5895, + "step": 15824 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011416033202736489, + "loss": 0.4902, + "step": 15825 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001141507867163535, + "loss": 0.5614, + "step": 15826 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011414124127377376, + "loss": 0.5003, + "step": 15827 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001141316956997144, + "loss": 0.4327, + "step": 15828 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011412214999426416, + "loss": 0.558, + "step": 15829 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011411260415751179, + "loss": 0.5455, + "step": 15830 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011410305818954605, + "loss": 0.5234, + "step": 15831 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011409351209045571, + "loss": 0.4608, + "step": 15832 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001140839658603295, + "loss": 0.6161, + "step": 15833 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011407441949925617, + "loss": 0.5308, + "step": 15834 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011406487300732454, + "loss": 0.5039, + "step": 15835 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011405532638462331, + "loss": 0.5705, + "step": 15836 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011404577963124126, + "loss": 0.5263, + "step": 15837 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011403623274726714, + "loss": 0.5164, + "step": 15838 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011402668573278969, + "loss": 0.5617, + "step": 15839 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011401713858789775, + "loss": 0.497, + "step": 15840 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011400759131268002, + "loss": 0.5705, + "step": 15841 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011399804390722532, + "loss": 0.5143, + "step": 15842 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011398849637162235, + "loss": 0.5673, + "step": 15843 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011397894870595993, + "loss": 0.6475, + "step": 15844 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011396940091032682, + "loss": 0.6188, + "step": 15845 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001139598529848118, + "loss": 0.5411, + "step": 15846 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011395030492950361, + "loss": 0.4306, + "step": 15847 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011394075674449106, + "loss": 0.4751, + "step": 15848 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011393120842986289, + "loss": 0.5741, + "step": 15849 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011392165998570791, + "loss": 0.521, + "step": 15850 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011391211141211489, + "loss": 0.605, + "step": 15851 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011390256270917261, + "loss": 0.6591, + "step": 15852 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001138930138769698, + "loss": 0.5624, + "step": 15853 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011388346491559534, + "loss": 0.5705, + "step": 15854 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001138739158251379, + "loss": 0.5052, + "step": 15855 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011386436660568635, + "loss": 0.6484, + "step": 15856 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011385481725732945, + "loss": 0.4859, + "step": 15857 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011384526778015598, + "loss": 0.4917, + "step": 15858 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011383571817425474, + "loss": 0.5671, + "step": 15859 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011382616843971448, + "loss": 0.5086, + "step": 15860 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011381661857662403, + "loss": 0.4649, + "step": 15861 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011380706858507217, + "loss": 0.5273, + "step": 15862 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011379751846514768, + "loss": 0.4856, + "step": 15863 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011378796821693938, + "loss": 0.5498, + "step": 15864 + }, + { + "epoch": 0.47, + "learning_rate": 0.000113778417840536, + "loss": 0.5121, + "step": 15865 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011376886733602646, + "loss": 0.4872, + "step": 15866 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001137593167034994, + "loss": 0.6122, + "step": 15867 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011374976594304376, + "loss": 0.5538, + "step": 15868 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011374021505474825, + "loss": 0.613, + "step": 15869 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011373066403870167, + "loss": 0.5116, + "step": 15870 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011372111289499289, + "loss": 0.528, + "step": 15871 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011371156162371064, + "loss": 0.5294, + "step": 15872 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001137020102249438, + "loss": 0.6073, + "step": 15873 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011369245869878107, + "loss": 0.5223, + "step": 15874 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011368290704531135, + "loss": 0.5428, + "step": 15875 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011367335526462342, + "loss": 0.5608, + "step": 15876 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011366380335680606, + "loss": 0.4812, + "step": 15877 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011365425132194811, + "loss": 0.5543, + "step": 15878 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001136446991601384, + "loss": 0.524, + "step": 15879 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011363514687146568, + "loss": 0.5856, + "step": 15880 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011362559445601882, + "loss": 0.4304, + "step": 15881 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011361604191388659, + "loss": 0.5917, + "step": 15882 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011360648924515786, + "loss": 0.5063, + "step": 15883 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011359693644992138, + "loss": 0.5609, + "step": 15884 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011358738352826605, + "loss": 0.6658, + "step": 15885 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011357783048028059, + "loss": 0.5906, + "step": 15886 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001135682773060539, + "loss": 0.6319, + "step": 15887 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001135587240056748, + "loss": 0.6497, + "step": 15888 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011354917057923206, + "loss": 0.6093, + "step": 15889 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011353961702681456, + "loss": 0.4975, + "step": 15890 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011353006334851107, + "loss": 0.6106, + "step": 15891 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011352050954441042, + "loss": 0.4651, + "step": 15892 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001135109556146015, + "loss": 0.4532, + "step": 15893 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011350140155917308, + "loss": 0.6019, + "step": 15894 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011349184737821404, + "loss": 0.5238, + "step": 15895 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011348229307181312, + "loss": 0.6289, + "step": 15896 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001134727386400593, + "loss": 0.5826, + "step": 15897 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011346318408304124, + "loss": 0.5607, + "step": 15898 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001134536294008479, + "loss": 0.6153, + "step": 15899 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011344407459356808, + "loss": 0.5771, + "step": 15900 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011343451966129061, + "loss": 0.5265, + "step": 15901 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011342496460410435, + "loss": 0.4909, + "step": 15902 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011341540942209807, + "loss": 0.5412, + "step": 15903 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011340585411536069, + "loss": 0.5608, + "step": 15904 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011339629868398103, + "loss": 0.5943, + "step": 15905 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001133867431280479, + "loss": 0.5922, + "step": 15906 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011337718744765018, + "loss": 0.6329, + "step": 15907 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011336763164287672, + "loss": 0.6676, + "step": 15908 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011335807571381631, + "loss": 0.5166, + "step": 15909 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011334851966055786, + "loss": 0.659, + "step": 15910 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001133389634831902, + "loss": 0.5089, + "step": 15911 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011332940718180216, + "loss": 0.5558, + "step": 15912 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011331985075648259, + "loss": 0.4669, + "step": 15913 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001133102942073204, + "loss": 0.6206, + "step": 15914 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011330073753440436, + "loss": 0.6473, + "step": 15915 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011329118073782341, + "loss": 0.4503, + "step": 15916 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001132816238176663, + "loss": 0.5785, + "step": 15917 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011327206677402198, + "loss": 0.5583, + "step": 15918 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011326250960697926, + "loss": 0.5316, + "step": 15919 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011325295231662702, + "loss": 0.5076, + "step": 15920 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011324339490305412, + "loss": 0.581, + "step": 15921 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001132338373663494, + "loss": 0.5259, + "step": 15922 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011322427970660174, + "loss": 0.5926, + "step": 15923 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011321472192389999, + "loss": 0.5021, + "step": 15924 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011320516401833301, + "loss": 0.5761, + "step": 15925 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011319560598998973, + "loss": 0.5455, + "step": 15926 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011318604783895889, + "loss": 0.4154, + "step": 15927 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011317648956532951, + "loss": 0.6599, + "step": 15928 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011316693116919033, + "loss": 0.5734, + "step": 15929 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011315737265063028, + "loss": 0.5669, + "step": 15930 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011314781400973824, + "loss": 0.4368, + "step": 15931 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011313825524660304, + "loss": 0.5553, + "step": 15932 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001131286963613136, + "loss": 0.5377, + "step": 15933 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011311913735395876, + "loss": 0.5671, + "step": 15934 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011310957822462742, + "loss": 0.4604, + "step": 15935 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011310001897340845, + "loss": 0.4744, + "step": 15936 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011309045960039073, + "loss": 0.5869, + "step": 15937 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011308090010566312, + "loss": 0.5411, + "step": 15938 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011307134048931452, + "loss": 0.5981, + "step": 15939 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011306178075143381, + "loss": 0.472, + "step": 15940 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011305222089210985, + "loss": 0.5394, + "step": 15941 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011304266091143156, + "loss": 0.4657, + "step": 15942 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011303310080948782, + "loss": 0.6801, + "step": 15943 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011302354058636744, + "loss": 0.5106, + "step": 15944 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011301398024215944, + "loss": 0.5289, + "step": 15945 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001130044197769526, + "loss": 0.5618, + "step": 15946 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011299485919083588, + "loss": 0.6164, + "step": 15947 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001129852984838981, + "loss": 0.5553, + "step": 15948 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001129757376562282, + "loss": 0.6469, + "step": 15949 + }, + { + "epoch": 0.47, + "learning_rate": 0.0001129661767079151, + "loss": 0.5316, + "step": 15950 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011295661563904763, + "loss": 0.5626, + "step": 15951 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011294705444971469, + "loss": 0.5812, + "step": 15952 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011293749314000524, + "loss": 0.5518, + "step": 15953 + }, + { + "epoch": 0.47, + "learning_rate": 0.00011292793171000811, + "loss": 0.5376, + "step": 15954 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011291837015981222, + "loss": 0.5326, + "step": 15955 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011290880848950647, + "loss": 0.502, + "step": 15956 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011289924669917979, + "loss": 0.5528, + "step": 15957 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011288968478892101, + "loss": 0.4839, + "step": 15958 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011288012275881915, + "loss": 0.4732, + "step": 15959 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011287056060896297, + "loss": 0.4773, + "step": 15960 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011286099833944148, + "loss": 0.5871, + "step": 15961 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011285143595034355, + "loss": 0.5428, + "step": 15962 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001128418734417581, + "loss": 0.4468, + "step": 15963 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011283231081377402, + "loss": 0.591, + "step": 15964 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001128227480664802, + "loss": 0.5845, + "step": 15965 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011281318519996563, + "loss": 0.5747, + "step": 15966 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011280362221431915, + "loss": 0.6247, + "step": 15967 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001127940591096297, + "loss": 0.542, + "step": 15968 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011278449588598616, + "loss": 0.6023, + "step": 15969 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011277493254347751, + "loss": 0.6708, + "step": 15970 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011276536908219262, + "loss": 0.4954, + "step": 15971 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001127558055022204, + "loss": 0.6122, + "step": 15972 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011274624180364978, + "loss": 0.5708, + "step": 15973 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001127366779865697, + "loss": 0.5665, + "step": 15974 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011272711405106904, + "loss": 0.5597, + "step": 15975 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011271754999723679, + "loss": 0.5347, + "step": 15976 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001127079858251618, + "loss": 0.4859, + "step": 15977 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011269842153493304, + "loss": 0.5536, + "step": 15978 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011268885712663939, + "loss": 0.7029, + "step": 15979 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011267929260036982, + "loss": 0.5903, + "step": 15980 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011266972795621323, + "loss": 0.514, + "step": 15981 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011266016319425858, + "loss": 0.5532, + "step": 15982 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011265059831459477, + "loss": 0.5801, + "step": 15983 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011264103331731072, + "loss": 0.5554, + "step": 15984 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011263146820249541, + "loss": 0.5197, + "step": 15985 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011262190297023773, + "loss": 0.5612, + "step": 15986 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011261233762062659, + "loss": 0.4602, + "step": 15987 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011260277215375103, + "loss": 0.5904, + "step": 15988 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011259320656969987, + "loss": 0.534, + "step": 15989 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011258364086856215, + "loss": 0.5699, + "step": 15990 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011257407505042667, + "loss": 0.497, + "step": 15991 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001125645091153825, + "loss": 0.5261, + "step": 15992 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011255494306351852, + "loss": 0.5731, + "step": 15993 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001125453768949237, + "loss": 0.533, + "step": 15994 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011253581060968696, + "loss": 0.548, + "step": 15995 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011252624420789723, + "loss": 0.6197, + "step": 15996 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011251667768964349, + "loss": 0.5947, + "step": 15997 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011250711105501468, + "loss": 0.6121, + "step": 15998 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011249754430409973, + "loss": 0.5782, + "step": 15999 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011248797743698759, + "loss": 0.5109, + "step": 16000 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011247841045376721, + "loss": 0.4665, + "step": 16001 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011246884335452754, + "loss": 0.5683, + "step": 16002 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011245927613935754, + "loss": 0.5726, + "step": 16003 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011244970880834616, + "loss": 0.4846, + "step": 16004 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011244014136158234, + "loss": 0.5599, + "step": 16005 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011243057379915502, + "loss": 0.5003, + "step": 16006 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011242100612115325, + "loss": 0.4985, + "step": 16007 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011241143832766585, + "loss": 0.5912, + "step": 16008 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011240187041878187, + "loss": 0.6606, + "step": 16009 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011239230239459022, + "loss": 0.5594, + "step": 16010 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011238273425517991, + "loss": 0.6026, + "step": 16011 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011237316600063985, + "loss": 0.4506, + "step": 16012 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011236359763105904, + "loss": 0.4455, + "step": 16013 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011235402914652642, + "loss": 0.6732, + "step": 16014 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011234446054713097, + "loss": 0.5702, + "step": 16015 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011233489183296165, + "loss": 0.492, + "step": 16016 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011232532300410741, + "loss": 0.5701, + "step": 16017 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001123157540606572, + "loss": 0.5566, + "step": 16018 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011230618500270007, + "loss": 0.6522, + "step": 16019 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001122966158303249, + "loss": 0.5148, + "step": 16020 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011228704654362073, + "loss": 0.5541, + "step": 16021 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011227747714267644, + "loss": 0.5674, + "step": 16022 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011226790762758113, + "loss": 0.5908, + "step": 16023 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011225833799842366, + "loss": 0.5736, + "step": 16024 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011224876825529307, + "loss": 0.5109, + "step": 16025 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001122391983982783, + "loss": 0.6028, + "step": 16026 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011222962842746836, + "loss": 0.556, + "step": 16027 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001122200583429522, + "loss": 0.5659, + "step": 16028 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011221048814481881, + "loss": 0.5479, + "step": 16029 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011220091783315714, + "loss": 0.6213, + "step": 16030 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011219134740805623, + "loss": 0.5894, + "step": 16031 + }, + { + "epoch": 0.48, + "learning_rate": 0.000112181776869605, + "loss": 0.5032, + "step": 16032 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011217220621789253, + "loss": 0.5548, + "step": 16033 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011216263545300769, + "loss": 0.4592, + "step": 16034 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011215306457503951, + "loss": 0.6159, + "step": 16035 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011214349358407701, + "loss": 0.526, + "step": 16036 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011213392248020909, + "loss": 0.5967, + "step": 16037 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011212435126352486, + "loss": 0.4927, + "step": 16038 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011211477993411319, + "loss": 0.6223, + "step": 16039 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011210520849206316, + "loss": 0.5103, + "step": 16040 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001120956369374637, + "loss": 0.4544, + "step": 16041 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011208606527040384, + "loss": 0.6433, + "step": 16042 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011207649349097258, + "loss": 0.475, + "step": 16043 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011206692159925887, + "loss": 0.4314, + "step": 16044 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011205734959535175, + "loss": 0.5495, + "step": 16045 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011204777747934019, + "loss": 0.45, + "step": 16046 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001120382052513132, + "loss": 0.4975, + "step": 16047 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011202863291135979, + "loss": 0.6366, + "step": 16048 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011201906045956891, + "loss": 0.6135, + "step": 16049 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011200948789602963, + "loss": 0.5541, + "step": 16050 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011199991522083088, + "loss": 0.5965, + "step": 16051 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011199034243406177, + "loss": 0.4855, + "step": 16052 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011198076953581115, + "loss": 0.4746, + "step": 16053 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011197119652616817, + "loss": 0.4669, + "step": 16054 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011196162340522174, + "loss": 0.562, + "step": 16055 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011195205017306092, + "loss": 0.5142, + "step": 16056 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001119424768297747, + "loss": 0.5523, + "step": 16057 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001119329033754521, + "loss": 0.6059, + "step": 16058 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011192332981018211, + "loss": 0.5363, + "step": 16059 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011191375613405375, + "loss": 0.614, + "step": 16060 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011190418234715603, + "loss": 0.5123, + "step": 16061 + }, + { + "epoch": 0.48, + "learning_rate": 0.000111894608449578, + "loss": 0.6141, + "step": 16062 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011188503444140859, + "loss": 0.5608, + "step": 16063 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001118754603227369, + "loss": 0.4664, + "step": 16064 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011186588609365189, + "loss": 0.5333, + "step": 16065 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011185631175424263, + "loss": 0.5142, + "step": 16066 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011184673730459811, + "loss": 0.6192, + "step": 16067 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011183716274480732, + "loss": 0.5898, + "step": 16068 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011182758807495935, + "loss": 0.5564, + "step": 16069 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011181801329514312, + "loss": 0.5749, + "step": 16070 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011180843840544775, + "loss": 0.5451, + "step": 16071 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011179886340596222, + "loss": 0.623, + "step": 16072 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011178928829677555, + "loss": 0.5078, + "step": 16073 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001117797130779768, + "loss": 0.5931, + "step": 16074 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011177013774965494, + "loss": 0.5904, + "step": 16075 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011176056231189907, + "loss": 0.5188, + "step": 16076 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011175098676479816, + "loss": 0.5162, + "step": 16077 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011174141110844125, + "loss": 0.6285, + "step": 16078 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011173183534291737, + "loss": 0.4119, + "step": 16079 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011172225946831558, + "loss": 0.5284, + "step": 16080 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001117126834847249, + "loss": 0.4845, + "step": 16081 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011170310739223433, + "loss": 0.5451, + "step": 16082 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011169353119093297, + "loss": 0.6222, + "step": 16083 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001116839548809098, + "loss": 0.5356, + "step": 16084 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011167437846225386, + "loss": 0.5466, + "step": 16085 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011166480193505422, + "loss": 0.5366, + "step": 16086 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011165522529939989, + "loss": 0.6068, + "step": 16087 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011164564855537991, + "loss": 0.5503, + "step": 16088 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011163607170308337, + "loss": 0.5508, + "step": 16089 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011162649474259923, + "loss": 0.3952, + "step": 16090 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001116169176740166, + "loss": 0.6968, + "step": 16091 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011160734049742448, + "loss": 0.5076, + "step": 16092 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011159776321291197, + "loss": 0.5318, + "step": 16093 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011158818582056803, + "loss": 0.6616, + "step": 16094 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011157860832048182, + "loss": 0.4797, + "step": 16095 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011156903071274227, + "loss": 0.5511, + "step": 16096 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001115594529974385, + "loss": 0.4397, + "step": 16097 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011154987517465953, + "loss": 0.4615, + "step": 16098 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011154029724449446, + "loss": 0.5831, + "step": 16099 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001115307192070323, + "loss": 0.5428, + "step": 16100 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011152114106236209, + "loss": 0.606, + "step": 16101 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001115115628105729, + "loss": 0.494, + "step": 16102 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011150198445175379, + "loss": 0.6085, + "step": 16103 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011149240598599382, + "loss": 0.5523, + "step": 16104 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011148282741338203, + "loss": 0.6786, + "step": 16105 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001114732487340075, + "loss": 0.5344, + "step": 16106 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011146366994795926, + "loss": 0.5636, + "step": 16107 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001114540910553264, + "loss": 0.5743, + "step": 16108 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011144451205619797, + "loss": 0.4045, + "step": 16109 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011143493295066303, + "loss": 0.5524, + "step": 16110 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011142535373881061, + "loss": 0.5731, + "step": 16111 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011141577442072986, + "loss": 0.4157, + "step": 16112 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011140619499650973, + "loss": 0.502, + "step": 16113 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011139661546623939, + "loss": 0.5846, + "step": 16114 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011138703583000782, + "loss": 0.4928, + "step": 16115 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011137745608790415, + "loss": 0.4897, + "step": 16116 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011136787624001743, + "loss": 0.4706, + "step": 16117 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011135829628643672, + "loss": 0.5108, + "step": 16118 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011134871622725111, + "loss": 0.4359, + "step": 16119 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011133913606254967, + "loss": 0.5174, + "step": 16120 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011132955579242144, + "loss": 0.5335, + "step": 16121 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011131997541695551, + "loss": 0.6303, + "step": 16122 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011131039493624095, + "loss": 0.5843, + "step": 16123 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001113008143503669, + "loss": 0.5291, + "step": 16124 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011129123365942234, + "loss": 0.6302, + "step": 16125 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011128165286349641, + "loss": 0.5342, + "step": 16126 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011127207196267815, + "loss": 0.5907, + "step": 16127 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011126249095705666, + "loss": 0.5037, + "step": 16128 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011125290984672103, + "loss": 0.455, + "step": 16129 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011124332863176032, + "loss": 0.5546, + "step": 16130 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011123374731226361, + "loss": 0.5566, + "step": 16131 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011122416588832003, + "loss": 0.6598, + "step": 16132 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001112145843600186, + "loss": 0.5869, + "step": 16133 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011120500272744845, + "loss": 0.5251, + "step": 16134 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011119542099069864, + "loss": 0.4385, + "step": 16135 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011118583914985827, + "loss": 0.5825, + "step": 16136 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011117625720501643, + "loss": 0.5687, + "step": 16137 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001111666751562622, + "loss": 0.5385, + "step": 16138 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011115709300368469, + "loss": 0.5426, + "step": 16139 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011114751074737295, + "loss": 0.5603, + "step": 16140 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011113792838741613, + "loss": 0.6887, + "step": 16141 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011112834592390325, + "loss": 0.6062, + "step": 16142 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011111876335692348, + "loss": 0.6259, + "step": 16143 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011110918068656584, + "loss": 0.5855, + "step": 16144 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011109959791291951, + "loss": 0.6607, + "step": 16145 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001110900150360735, + "loss": 0.461, + "step": 16146 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011108043205611699, + "loss": 0.6129, + "step": 16147 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011107084897313901, + "loss": 0.5057, + "step": 16148 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001110612657872287, + "loss": 0.5562, + "step": 16149 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011105168249847518, + "loss": 0.5755, + "step": 16150 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011104209910696748, + "loss": 0.5481, + "step": 16151 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011103251561279475, + "loss": 0.5368, + "step": 16152 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001110229320160461, + "loss": 0.5888, + "step": 16153 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011101334831681058, + "loss": 0.6348, + "step": 16154 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001110037645151774, + "loss": 0.5095, + "step": 16155 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011099418061123556, + "loss": 0.5321, + "step": 16156 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011098459660507426, + "loss": 0.4709, + "step": 16157 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011097501249678249, + "loss": 0.6106, + "step": 16158 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011096542828644947, + "loss": 0.5782, + "step": 16159 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011095584397416425, + "loss": 0.5495, + "step": 16160 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011094625956001598, + "loss": 0.4552, + "step": 16161 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011093667504409373, + "loss": 0.4873, + "step": 16162 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011092709042648666, + "loss": 0.5459, + "step": 16163 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011091750570728382, + "loss": 0.5464, + "step": 16164 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011090792088657441, + "loss": 0.6199, + "step": 16165 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011089833596444747, + "loss": 0.5717, + "step": 16166 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011088875094099215, + "loss": 0.5108, + "step": 16167 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011087916581629758, + "loss": 0.5935, + "step": 16168 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011086958059045284, + "loss": 0.595, + "step": 16169 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011085999526354709, + "loss": 0.5988, + "step": 16170 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011085040983566941, + "loss": 0.4785, + "step": 16171 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011084082430690896, + "loss": 0.5806, + "step": 16172 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011083123867735481, + "loss": 0.4897, + "step": 16173 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011082165294709619, + "loss": 0.6146, + "step": 16174 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001108120671162221, + "loss": 0.4998, + "step": 16175 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011080248118482175, + "loss": 0.5388, + "step": 16176 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011079289515298419, + "loss": 0.5635, + "step": 16177 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011078330902079863, + "loss": 0.4591, + "step": 16178 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011077372278835416, + "loss": 0.5651, + "step": 16179 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011076413645573988, + "loss": 0.5459, + "step": 16180 + }, + { + "epoch": 0.48, + "learning_rate": 0.000110754550023045, + "loss": 0.59, + "step": 16181 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011074496349035856, + "loss": 0.4198, + "step": 16182 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011073537685776975, + "loss": 0.5248, + "step": 16183 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011072579012536768, + "loss": 0.628, + "step": 16184 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011071620329324145, + "loss": 0.5399, + "step": 16185 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001107066163614803, + "loss": 0.5331, + "step": 16186 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011069702933017325, + "loss": 0.5407, + "step": 16187 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011068744219940951, + "loss": 0.4952, + "step": 16188 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011067785496927818, + "loss": 0.5464, + "step": 16189 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011066826763986842, + "loss": 0.6266, + "step": 16190 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011065868021126935, + "loss": 0.6282, + "step": 16191 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011064909268357013, + "loss": 0.4755, + "step": 16192 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001106395050568599, + "loss": 0.5886, + "step": 16193 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011062991733122776, + "loss": 0.633, + "step": 16194 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011062032950676291, + "loss": 0.5384, + "step": 16195 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011061074158355447, + "loss": 0.4978, + "step": 16196 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011060115356169154, + "loss": 0.5888, + "step": 16197 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011059156544126336, + "loss": 0.5244, + "step": 16198 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011058197722235904, + "loss": 0.6574, + "step": 16199 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011057238890506767, + "loss": 0.5854, + "step": 16200 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011056280048947844, + "loss": 0.522, + "step": 16201 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011055321197568051, + "loss": 0.5984, + "step": 16202 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011054362336376305, + "loss": 0.5021, + "step": 16203 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011053403465381516, + "loss": 0.5747, + "step": 16204 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011052444584592603, + "loss": 0.5295, + "step": 16205 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011051485694018475, + "loss": 0.7645, + "step": 16206 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011050526793668057, + "loss": 0.5667, + "step": 16207 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011049567883550256, + "loss": 0.5419, + "step": 16208 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011048608963673994, + "loss": 0.5164, + "step": 16209 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011047650034048183, + "loss": 0.6068, + "step": 16210 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011046691094681738, + "loss": 0.4778, + "step": 16211 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011045732145583579, + "loss": 0.493, + "step": 16212 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011044773186762618, + "loss": 0.4911, + "step": 16213 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011043814218227773, + "loss": 0.5558, + "step": 16214 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011042855239987958, + "loss": 0.4857, + "step": 16215 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011041896252052089, + "loss": 0.5798, + "step": 16216 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011040937254429089, + "loss": 0.4878, + "step": 16217 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011039978247127865, + "loss": 0.5223, + "step": 16218 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011039019230157342, + "loss": 0.5276, + "step": 16219 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011038060203526426, + "loss": 0.5187, + "step": 16220 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011037101167244044, + "loss": 0.6443, + "step": 16221 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011036142121319109, + "loss": 0.5272, + "step": 16222 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011035183065760535, + "loss": 0.5096, + "step": 16223 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011034224000577244, + "loss": 0.6264, + "step": 16224 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001103326492577815, + "loss": 0.5426, + "step": 16225 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001103230584137217, + "loss": 0.576, + "step": 16226 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011031346747368221, + "loss": 0.5551, + "step": 16227 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001103038764377522, + "loss": 0.4682, + "step": 16228 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001102942853060209, + "loss": 0.6771, + "step": 16229 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001102846940785774, + "loss": 0.6049, + "step": 16230 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011027510275551094, + "loss": 0.5535, + "step": 16231 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011026551133691065, + "loss": 0.4755, + "step": 16232 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011025591982286572, + "loss": 0.6765, + "step": 16233 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011024632821346537, + "loss": 0.6576, + "step": 16234 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011023673650879868, + "loss": 0.5933, + "step": 16235 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011022714470895497, + "loss": 0.5968, + "step": 16236 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011021755281402328, + "loss": 0.6647, + "step": 16237 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011020796082409288, + "loss": 0.5372, + "step": 16238 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011019836873925295, + "loss": 0.5524, + "step": 16239 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011018877655959263, + "loss": 0.4874, + "step": 16240 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011017918428520113, + "loss": 0.5417, + "step": 16241 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011016959191616765, + "loss": 0.5021, + "step": 16242 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011015999945258132, + "loss": 0.553, + "step": 16243 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001101504068945314, + "loss": 0.571, + "step": 16244 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011014081424210702, + "loss": 0.6343, + "step": 16245 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001101312214953974, + "loss": 0.5863, + "step": 16246 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011012162865449172, + "loss": 0.5839, + "step": 16247 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011011203571947919, + "loss": 0.6638, + "step": 16248 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011010244269044894, + "loss": 0.5667, + "step": 16249 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011009284956749024, + "loss": 0.6379, + "step": 16250 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011008325635069222, + "loss": 0.6379, + "step": 16251 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001100736630401441, + "loss": 0.472, + "step": 16252 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001100640696359351, + "loss": 0.5818, + "step": 16253 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011005447613815439, + "loss": 0.4766, + "step": 16254 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011004488254689116, + "loss": 0.5075, + "step": 16255 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011003528886223461, + "loss": 0.5062, + "step": 16256 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011002569508427396, + "loss": 0.6713, + "step": 16257 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011001610121309838, + "loss": 0.492, + "step": 16258 + }, + { + "epoch": 0.48, + "learning_rate": 0.00011000650724879706, + "loss": 0.6556, + "step": 16259 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010999691319145928, + "loss": 0.5982, + "step": 16260 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010998731904117414, + "loss": 0.6061, + "step": 16261 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010997772479803093, + "loss": 0.4821, + "step": 16262 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010996813046211878, + "loss": 0.6119, + "step": 16263 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001099585360335269, + "loss": 0.4446, + "step": 16264 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010994894151234458, + "loss": 0.6419, + "step": 16265 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010993934689866093, + "loss": 0.7792, + "step": 16266 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010992975219256523, + "loss": 0.4438, + "step": 16267 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010992015739414659, + "loss": 0.498, + "step": 16268 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010991056250349431, + "loss": 0.5229, + "step": 16269 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010990096752069758, + "loss": 0.5835, + "step": 16270 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001098913724458456, + "loss": 0.5234, + "step": 16271 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010988177727902757, + "loss": 0.6202, + "step": 16272 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010987218202033273, + "loss": 0.5006, + "step": 16273 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010986258666985026, + "loss": 0.5831, + "step": 16274 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010985299122766941, + "loss": 0.5515, + "step": 16275 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010984339569387935, + "loss": 0.6421, + "step": 16276 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010983380006856934, + "loss": 0.5648, + "step": 16277 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010982420435182855, + "loss": 0.512, + "step": 16278 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010981460854374625, + "loss": 0.5321, + "step": 16279 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010980501264441161, + "loss": 0.6036, + "step": 16280 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010979541665391391, + "loss": 0.5341, + "step": 16281 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010978582057234227, + "loss": 0.6976, + "step": 16282 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010977622439978601, + "loss": 0.4936, + "step": 16283 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001097666281363343, + "loss": 0.5679, + "step": 16284 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010975703178207639, + "loss": 0.6025, + "step": 16285 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010974743533710148, + "loss": 0.4507, + "step": 16286 + }, + { + "epoch": 0.48, + "learning_rate": 0.0001097378388014988, + "loss": 0.5529, + "step": 16287 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010972824217535756, + "loss": 0.6663, + "step": 16288 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010971864545876702, + "loss": 0.5579, + "step": 16289 + }, + { + "epoch": 0.48, + "learning_rate": 0.00010970904865181637, + "loss": 0.5899, + "step": 16290 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010969945175459488, + "loss": 0.6674, + "step": 16291 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010968985476719172, + "loss": 0.5931, + "step": 16292 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010968025768969621, + "loss": 0.5336, + "step": 16293 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010967066052219746, + "loss": 0.6597, + "step": 16294 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010966106326478477, + "loss": 0.5472, + "step": 16295 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010965146591754742, + "loss": 0.6074, + "step": 16296 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010964186848057454, + "loss": 0.548, + "step": 16297 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010963227095395544, + "loss": 0.5575, + "step": 16298 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010962267333777927, + "loss": 0.5683, + "step": 16299 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010961307563213537, + "loss": 0.6869, + "step": 16300 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010960347783711288, + "loss": 0.4815, + "step": 16301 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010959387995280113, + "loss": 0.6415, + "step": 16302 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010958428197928927, + "loss": 0.5277, + "step": 16303 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001095746839166666, + "loss": 0.5991, + "step": 16304 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010956508576502233, + "loss": 0.4425, + "step": 16305 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010955548752444569, + "loss": 0.5418, + "step": 16306 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010954588919502596, + "loss": 0.6154, + "step": 16307 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010953629077685235, + "loss": 0.4881, + "step": 16308 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010952669227001407, + "loss": 0.5842, + "step": 16309 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010951709367460045, + "loss": 0.5593, + "step": 16310 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010950749499070066, + "loss": 0.5926, + "step": 16311 + }, + { + "epoch": 0.49, + "learning_rate": 0.000109497896218404, + "loss": 0.5749, + "step": 16312 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010948829735779964, + "loss": 0.4905, + "step": 16313 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010947869840897692, + "loss": 0.5857, + "step": 16314 + }, + { + "epoch": 0.49, + "learning_rate": 0.000109469099372025, + "loss": 0.6391, + "step": 16315 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010945950024703319, + "loss": 0.495, + "step": 16316 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010944990103409072, + "loss": 0.5621, + "step": 16317 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010944030173328683, + "loss": 0.4478, + "step": 16318 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001094307023447108, + "loss": 0.4955, + "step": 16319 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010942110286845184, + "loss": 0.4946, + "step": 16320 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001094115033045992, + "loss": 0.5554, + "step": 16321 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010940190365324219, + "loss": 0.6729, + "step": 16322 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010939230391446998, + "loss": 0.4877, + "step": 16323 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010938270408837192, + "loss": 0.5796, + "step": 16324 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010937310417503717, + "loss": 0.5689, + "step": 16325 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010936350417455508, + "loss": 0.595, + "step": 16326 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010935390408701483, + "loss": 0.4806, + "step": 16327 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010934430391250572, + "loss": 0.512, + "step": 16328 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010933470365111702, + "loss": 0.5074, + "step": 16329 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010932510330293793, + "loss": 0.6087, + "step": 16330 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010931550286805773, + "loss": 0.5284, + "step": 16331 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010930590234656572, + "loss": 0.5112, + "step": 16332 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010929630173855117, + "loss": 0.4888, + "step": 16333 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010928670104410329, + "loss": 0.544, + "step": 16334 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010927710026331134, + "loss": 0.4675, + "step": 16335 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010926749939626464, + "loss": 0.5856, + "step": 16336 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001092578984430524, + "loss": 0.5183, + "step": 16337 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010924829740376392, + "loss": 0.5947, + "step": 16338 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010923869627848847, + "loss": 0.5536, + "step": 16339 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010922909506731527, + "loss": 0.7317, + "step": 16340 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010921949377033367, + "loss": 0.4846, + "step": 16341 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010920989238763284, + "loss": 0.5215, + "step": 16342 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010920029091930217, + "loss": 0.5012, + "step": 16343 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001091906893654308, + "loss": 0.5128, + "step": 16344 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010918108772610807, + "loss": 0.524, + "step": 16345 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010917148600142326, + "loss": 0.4333, + "step": 16346 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010916188419146563, + "loss": 0.6214, + "step": 16347 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010915228229632446, + "loss": 0.4867, + "step": 16348 + }, + { + "epoch": 0.49, + "learning_rate": 0.000109142680316089, + "loss": 0.5076, + "step": 16349 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010913307825084855, + "loss": 0.6319, + "step": 16350 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010912347610069238, + "loss": 0.4797, + "step": 16351 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010911387386570974, + "loss": 0.6731, + "step": 16352 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010910427154598999, + "loss": 0.545, + "step": 16353 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001090946691416223, + "loss": 0.5928, + "step": 16354 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010908506665269605, + "loss": 0.5464, + "step": 16355 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001090754640793004, + "loss": 0.5594, + "step": 16356 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010906586142152476, + "loss": 0.4841, + "step": 16357 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010905625867945833, + "loss": 0.6721, + "step": 16358 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010904665585319043, + "loss": 0.537, + "step": 16359 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010903705294281033, + "loss": 0.5785, + "step": 16360 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010902744994840731, + "loss": 0.5003, + "step": 16361 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010901784687007064, + "loss": 0.5022, + "step": 16362 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010900824370788964, + "loss": 0.5284, + "step": 16363 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010899864046195359, + "loss": 0.4531, + "step": 16364 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010898903713235175, + "loss": 0.4995, + "step": 16365 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010897943371917342, + "loss": 0.5839, + "step": 16366 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010896983022250794, + "loss": 0.5639, + "step": 16367 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010896022664244449, + "loss": 0.6099, + "step": 16368 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010895062297907244, + "loss": 0.6463, + "step": 16369 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001089410192324811, + "loss": 0.4925, + "step": 16370 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010893141540275968, + "loss": 0.5903, + "step": 16371 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010892181148999758, + "loss": 0.5381, + "step": 16372 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010891220749428397, + "loss": 0.6065, + "step": 16373 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010890260341570824, + "loss": 0.6483, + "step": 16374 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010889299925435965, + "loss": 0.5741, + "step": 16375 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010888339501032749, + "loss": 0.5651, + "step": 16376 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010887379068370108, + "loss": 0.4955, + "step": 16377 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010886418627456967, + "loss": 0.5148, + "step": 16378 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010885458178302261, + "loss": 0.5626, + "step": 16379 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010884497720914917, + "loss": 0.5996, + "step": 16380 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010883537255303865, + "loss": 0.5736, + "step": 16381 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010882576781478037, + "loss": 0.5424, + "step": 16382 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010881616299446363, + "loss": 0.5909, + "step": 16383 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010880655809217772, + "loss": 0.6525, + "step": 16384 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001087969531080119, + "loss": 0.5847, + "step": 16385 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010878734804205557, + "loss": 0.605, + "step": 16386 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010877774289439795, + "loss": 0.5101, + "step": 16387 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010876813766512836, + "loss": 0.44, + "step": 16388 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010875853235433616, + "loss": 0.5554, + "step": 16389 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010874892696211058, + "loss": 0.5724, + "step": 16390 + }, + { + "epoch": 0.49, + "learning_rate": 0.000108739321488541, + "loss": 0.5939, + "step": 16391 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010872971593371668, + "loss": 0.5646, + "step": 16392 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010872011029772693, + "loss": 0.4435, + "step": 16393 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010871050458066108, + "loss": 0.5967, + "step": 16394 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010870089878260842, + "loss": 0.6713, + "step": 16395 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001086912929036583, + "loss": 0.517, + "step": 16396 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010868168694389997, + "loss": 0.4814, + "step": 16397 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010867208090342277, + "loss": 0.5576, + "step": 16398 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010866247478231604, + "loss": 0.5616, + "step": 16399 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010865286858066905, + "loss": 0.5157, + "step": 16400 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010864326229857118, + "loss": 0.5446, + "step": 16401 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010863365593611163, + "loss": 0.5209, + "step": 16402 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010862404949337988, + "loss": 0.5715, + "step": 16403 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010861444297046508, + "loss": 0.4744, + "step": 16404 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010860483636745665, + "loss": 0.5304, + "step": 16405 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010859522968444387, + "loss": 0.4509, + "step": 16406 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010858562292151608, + "loss": 0.6312, + "step": 16407 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001085760160787626, + "loss": 0.6119, + "step": 16408 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010856640915627272, + "loss": 0.6027, + "step": 16409 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001085568021541358, + "loss": 0.5063, + "step": 16410 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010854719507244113, + "loss": 0.5477, + "step": 16411 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010853758791127806, + "loss": 0.506, + "step": 16412 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001085279806707359, + "loss": 0.5344, + "step": 16413 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010851837335090394, + "loss": 0.5073, + "step": 16414 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010850876595187159, + "loss": 0.5678, + "step": 16415 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001084991584737281, + "loss": 0.5847, + "step": 16416 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010848955091656283, + "loss": 0.5516, + "step": 16417 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010847994328046507, + "loss": 0.573, + "step": 16418 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010847033556552422, + "loss": 0.616, + "step": 16419 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010846072777182954, + "loss": 0.532, + "step": 16420 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001084511198994704, + "loss": 0.6478, + "step": 16421 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010844151194853611, + "loss": 0.4668, + "step": 16422 + }, + { + "epoch": 0.49, + "learning_rate": 0.000108431903919116, + "loss": 0.5172, + "step": 16423 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010842229581129942, + "loss": 0.5646, + "step": 16424 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010841268762517566, + "loss": 0.5259, + "step": 16425 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010840307936083409, + "loss": 0.5434, + "step": 16426 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010839347101836407, + "loss": 0.5927, + "step": 16427 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010838386259785492, + "loss": 0.5416, + "step": 16428 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001083742540993959, + "loss": 0.5723, + "step": 16429 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010836464552307643, + "loss": 0.5883, + "step": 16430 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001083550368689858, + "loss": 0.4578, + "step": 16431 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001083454281372134, + "loss": 0.5417, + "step": 16432 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010833581932784849, + "loss": 0.4751, + "step": 16433 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010832621044098052, + "loss": 0.4401, + "step": 16434 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010831660147669871, + "loss": 0.5847, + "step": 16435 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010830699243509246, + "loss": 0.603, + "step": 16436 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010829738331625112, + "loss": 0.4872, + "step": 16437 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010828777412026401, + "loss": 0.4781, + "step": 16438 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010827816484722049, + "loss": 0.632, + "step": 16439 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010826855549720986, + "loss": 0.4754, + "step": 16440 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010825894607032154, + "loss": 0.5912, + "step": 16441 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010824933656664479, + "loss": 0.5681, + "step": 16442 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010823972698626903, + "loss": 0.6131, + "step": 16443 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010823011732928356, + "loss": 0.5534, + "step": 16444 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010822050759577774, + "loss": 0.535, + "step": 16445 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010821089778584093, + "loss": 0.5588, + "step": 16446 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010820128789956243, + "loss": 0.582, + "step": 16447 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010819167793703166, + "loss": 0.5672, + "step": 16448 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010818206789833791, + "loss": 0.548, + "step": 16449 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010817245778357057, + "loss": 0.5319, + "step": 16450 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010816284759281896, + "loss": 0.5524, + "step": 16451 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010815323732617245, + "loss": 0.6001, + "step": 16452 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001081436269837204, + "loss": 0.5254, + "step": 16453 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010813401656555213, + "loss": 0.5602, + "step": 16454 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010812440607175704, + "loss": 0.4929, + "step": 16455 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010811479550242445, + "loss": 0.5786, + "step": 16456 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010810518485764371, + "loss": 0.4817, + "step": 16457 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010809557413750424, + "loss": 0.5631, + "step": 16458 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010808596334209528, + "loss": 0.4994, + "step": 16459 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010807635247150631, + "loss": 0.4368, + "step": 16460 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010806674152582663, + "loss": 0.4831, + "step": 16461 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010805713050514556, + "loss": 0.5597, + "step": 16462 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010804751940955255, + "loss": 0.6134, + "step": 16463 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010803790823913689, + "loss": 0.6617, + "step": 16464 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010802829699398797, + "loss": 0.7627, + "step": 16465 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010801868567419512, + "loss": 0.55, + "step": 16466 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010800907427984775, + "loss": 0.5827, + "step": 16467 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010799946281103519, + "loss": 0.5961, + "step": 16468 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010798985126784682, + "loss": 0.5709, + "step": 16469 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010798023965037201, + "loss": 0.506, + "step": 16470 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010797062795870009, + "loss": 0.5383, + "step": 16471 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010796101619292045, + "loss": 0.5211, + "step": 16472 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010795140435312247, + "loss": 0.5463, + "step": 16473 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010794179243939549, + "loss": 0.5498, + "step": 16474 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010793218045182889, + "loss": 0.5334, + "step": 16475 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010792256839051203, + "loss": 0.4611, + "step": 16476 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001079129562555343, + "loss": 0.583, + "step": 16477 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010790334404698503, + "loss": 0.6557, + "step": 16478 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010789373176495368, + "loss": 0.5539, + "step": 16479 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010788411940952949, + "loss": 0.5717, + "step": 16480 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010787450698080192, + "loss": 0.5725, + "step": 16481 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010786489447886032, + "loss": 0.6068, + "step": 16482 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010785528190379408, + "loss": 0.4844, + "step": 16483 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010784566925569257, + "loss": 0.5997, + "step": 16484 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010783605653464515, + "loss": 0.5646, + "step": 16485 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010782644374074118, + "loss": 0.6218, + "step": 16486 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010781683087407007, + "loss": 0.4449, + "step": 16487 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010780721793472115, + "loss": 0.4675, + "step": 16488 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010779760492278387, + "loss": 0.5802, + "step": 16489 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010778799183834754, + "loss": 0.6728, + "step": 16490 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001077783786815016, + "loss": 0.3581, + "step": 16491 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010776876545233533, + "loss": 0.5577, + "step": 16492 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010775915215093823, + "loss": 0.6529, + "step": 16493 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010774953877739964, + "loss": 0.5088, + "step": 16494 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010773992533180888, + "loss": 0.6711, + "step": 16495 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010773031181425543, + "loss": 0.5567, + "step": 16496 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010772069822482857, + "loss": 0.5676, + "step": 16497 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010771108456361776, + "loss": 0.5356, + "step": 16498 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010770147083071235, + "loss": 0.5237, + "step": 16499 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010769185702620175, + "loss": 0.6527, + "step": 16500 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010768224315017532, + "loss": 0.4935, + "step": 16501 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010767262920272246, + "loss": 0.5242, + "step": 16502 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010766301518393254, + "loss": 0.6121, + "step": 16503 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010765340109389496, + "loss": 0.5797, + "step": 16504 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010764378693269912, + "loss": 0.5892, + "step": 16505 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010763417270043439, + "loss": 0.5209, + "step": 16506 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010762455839719014, + "loss": 0.5068, + "step": 16507 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010761494402305583, + "loss": 0.547, + "step": 16508 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010760532957812077, + "loss": 0.5311, + "step": 16509 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010759571506247442, + "loss": 0.5639, + "step": 16510 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001075861004762061, + "loss": 0.4504, + "step": 16511 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010757648581940525, + "loss": 0.4494, + "step": 16512 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010756687109216126, + "loss": 0.5326, + "step": 16513 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010755725629456352, + "loss": 0.4918, + "step": 16514 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010754764142670145, + "loss": 0.5967, + "step": 16515 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010753802648866438, + "loss": 0.4765, + "step": 16516 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010752841148054176, + "loss": 0.5745, + "step": 16517 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010751879640242297, + "loss": 0.6059, + "step": 16518 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001075091812543974, + "loss": 0.5617, + "step": 16519 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010749956603655447, + "loss": 0.4935, + "step": 16520 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010748995074898354, + "loss": 0.5402, + "step": 16521 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010748033539177407, + "loss": 0.5824, + "step": 16522 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001074707199650154, + "loss": 0.6315, + "step": 16523 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010746110446879694, + "loss": 0.6598, + "step": 16524 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001074514889032081, + "loss": 0.595, + "step": 16525 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010744187326833833, + "loss": 0.5701, + "step": 16526 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010743225756427698, + "loss": 0.4938, + "step": 16527 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010742264179111343, + "loss": 0.5766, + "step": 16528 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010741302594893711, + "loss": 0.6168, + "step": 16529 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010740341003783747, + "loss": 0.5088, + "step": 16530 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010739379405790387, + "loss": 0.4507, + "step": 16531 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010738417800922569, + "loss": 0.5118, + "step": 16532 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010737456189189237, + "loss": 0.5662, + "step": 16533 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010736494570599333, + "loss": 0.506, + "step": 16534 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010735532945161794, + "loss": 0.6234, + "step": 16535 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010734571312885565, + "loss": 0.6043, + "step": 16536 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010733609673779584, + "loss": 0.4806, + "step": 16537 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001073264802785279, + "loss": 0.6212, + "step": 16538 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010731686375114132, + "loss": 0.5559, + "step": 16539 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001073072471557254, + "loss": 0.5396, + "step": 16540 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010729763049236964, + "loss": 0.5482, + "step": 16541 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010728801376116343, + "loss": 0.5183, + "step": 16542 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010727839696219614, + "loss": 0.5552, + "step": 16543 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010726878009555725, + "loss": 0.5283, + "step": 16544 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010725916316133612, + "loss": 0.6003, + "step": 16545 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001072495461596222, + "loss": 0.5909, + "step": 16546 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010723992909050487, + "loss": 0.52, + "step": 16547 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010723031195407358, + "loss": 0.6432, + "step": 16548 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010722069475041771, + "loss": 0.5488, + "step": 16549 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001072110774796267, + "loss": 0.5257, + "step": 16550 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010720146014179, + "loss": 0.7238, + "step": 16551 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010719184273699696, + "loss": 0.4987, + "step": 16552 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010718222526533706, + "loss": 0.5428, + "step": 16553 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010717260772689966, + "loss": 0.476, + "step": 16554 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010716299012177422, + "loss": 0.528, + "step": 16555 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010715337245005017, + "loss": 0.5993, + "step": 16556 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001071437547118169, + "loss": 0.5829, + "step": 16557 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010713413690716385, + "loss": 0.4618, + "step": 16558 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010712451903618044, + "loss": 0.6117, + "step": 16559 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001071149010989561, + "loss": 0.6821, + "step": 16560 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010710528309558023, + "loss": 0.5365, + "step": 16561 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010709566502614227, + "loss": 0.3837, + "step": 16562 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010708604689073165, + "loss": 0.5253, + "step": 16563 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010707642868943782, + "loss": 0.5683, + "step": 16564 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010706681042235014, + "loss": 0.4281, + "step": 16565 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010705719208955806, + "loss": 0.5258, + "step": 16566 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010704757369115104, + "loss": 0.6047, + "step": 16567 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010703795522721852, + "loss": 0.531, + "step": 16568 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010702833669784982, + "loss": 0.5113, + "step": 16569 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010701871810313453, + "loss": 0.5283, + "step": 16570 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010700909944316193, + "loss": 0.5475, + "step": 16571 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010699948071802155, + "loss": 0.6284, + "step": 16572 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010698986192780279, + "loss": 0.6351, + "step": 16573 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010698024307259506, + "loss": 0.5169, + "step": 16574 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010697062415248785, + "loss": 0.6433, + "step": 16575 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010696100516757052, + "loss": 0.5695, + "step": 16576 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010695138611793256, + "loss": 0.5079, + "step": 16577 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010694176700366336, + "loss": 0.5338, + "step": 16578 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010693214782485239, + "loss": 0.517, + "step": 16579 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010692252858158909, + "loss": 0.7076, + "step": 16580 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010691290927396283, + "loss": 0.572, + "step": 16581 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010690328990206315, + "loss": 0.5093, + "step": 16582 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010689367046597937, + "loss": 0.5919, + "step": 16583 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010688405096580108, + "loss": 0.4686, + "step": 16584 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010687443140161755, + "loss": 0.6022, + "step": 16585 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001068648117735183, + "loss": 0.4675, + "step": 16586 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001068551920815928, + "loss": 0.5752, + "step": 16587 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010684557232593045, + "loss": 0.5531, + "step": 16588 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010683595250662069, + "loss": 0.6132, + "step": 16589 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010682633262375297, + "loss": 0.5744, + "step": 16590 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010681671267741673, + "loss": 0.554, + "step": 16591 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010680709266770142, + "loss": 0.6041, + "step": 16592 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010679747259469645, + "loss": 0.5061, + "step": 16593 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001067878524584913, + "loss": 0.514, + "step": 16594 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010677823225917542, + "loss": 0.5538, + "step": 16595 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010676861199683824, + "loss": 0.5933, + "step": 16596 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010675899167156917, + "loss": 0.5007, + "step": 16597 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001067493712834577, + "loss": 0.5547, + "step": 16598 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010673975083259329, + "loss": 0.5316, + "step": 16599 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010673013031906533, + "loss": 0.5188, + "step": 16600 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010672050974296334, + "loss": 0.4609, + "step": 16601 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010671088910437667, + "loss": 0.5037, + "step": 16602 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010670126840339487, + "loss": 0.5477, + "step": 16603 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010669164764010735, + "loss": 0.647, + "step": 16604 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010668202681460355, + "loss": 0.4554, + "step": 16605 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010667240592697294, + "loss": 0.5052, + "step": 16606 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010666278497730493, + "loss": 0.5715, + "step": 16607 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010665316396568903, + "loss": 0.5166, + "step": 16608 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010664354289221465, + "loss": 0.6489, + "step": 16609 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010663392175697127, + "loss": 0.5427, + "step": 16610 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010662430056004831, + "loss": 0.615, + "step": 16611 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010661467930153524, + "loss": 0.5065, + "step": 16612 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010660505798152153, + "loss": 0.6145, + "step": 16613 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001065954366000966, + "loss": 0.4687, + "step": 16614 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010658581515734998, + "loss": 0.584, + "step": 16615 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010657619365337101, + "loss": 0.5335, + "step": 16616 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010656657208824927, + "loss": 0.494, + "step": 16617 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010655695046207412, + "loss": 0.5667, + "step": 16618 + }, + { + "epoch": 0.49, + "learning_rate": 0.0001065473287749351, + "loss": 0.5319, + "step": 16619 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010653770702692159, + "loss": 0.5686, + "step": 16620 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010652808521812309, + "loss": 0.6238, + "step": 16621 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010651846334862908, + "loss": 0.4954, + "step": 16622 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010650884141852897, + "loss": 0.692, + "step": 16623 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010649921942791224, + "loss": 0.5723, + "step": 16624 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010648959737686841, + "loss": 0.5533, + "step": 16625 + }, + { + "epoch": 0.49, + "learning_rate": 0.00010647997526548685, + "loss": 0.6407, + "step": 16626 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010647035309385707, + "loss": 0.5717, + "step": 16627 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010646073086206852, + "loss": 0.493, + "step": 16628 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010645110857021066, + "loss": 0.5748, + "step": 16629 + }, + { + "epoch": 0.5, + "learning_rate": 0.000106441486218373, + "loss": 0.5935, + "step": 16630 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010643186380664492, + "loss": 0.6181, + "step": 16631 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010642224133511599, + "loss": 0.4883, + "step": 16632 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010641261880387557, + "loss": 0.5718, + "step": 16633 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010640299621301317, + "loss": 0.6045, + "step": 16634 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001063933735626183, + "loss": 0.5011, + "step": 16635 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010638375085278038, + "loss": 0.5449, + "step": 16636 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010637412808358888, + "loss": 0.5779, + "step": 16637 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010636450525513328, + "loss": 0.5545, + "step": 16638 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010635488236750306, + "loss": 0.4859, + "step": 16639 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010634525942078768, + "loss": 0.7024, + "step": 16640 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010633563641507658, + "loss": 0.5731, + "step": 16641 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010632601335045929, + "loss": 0.6446, + "step": 16642 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001063163902270252, + "loss": 0.5229, + "step": 16643 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010630676704486389, + "loss": 0.4952, + "step": 16644 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010629714380406473, + "loss": 0.4585, + "step": 16645 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010628752050471727, + "loss": 0.4751, + "step": 16646 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010627789714691092, + "loss": 0.5448, + "step": 16647 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001062682737307352, + "loss": 0.5931, + "step": 16648 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010625865025627957, + "loss": 0.5772, + "step": 16649 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010624902672363351, + "loss": 0.6544, + "step": 16650 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010623940313288647, + "loss": 0.6127, + "step": 16651 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010622977948412796, + "loss": 0.542, + "step": 16652 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010622015577744744, + "loss": 0.5762, + "step": 16653 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010621053201293439, + "loss": 0.6105, + "step": 16654 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010620090819067827, + "loss": 0.5737, + "step": 16655 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010619128431076861, + "loss": 0.4566, + "step": 16656 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010618166037329484, + "loss": 0.5373, + "step": 16657 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010617203637834647, + "loss": 0.5247, + "step": 16658 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010616241232601296, + "loss": 0.5071, + "step": 16659 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010615278821638377, + "loss": 0.5441, + "step": 16660 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010614316404954844, + "loss": 0.4111, + "step": 16661 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001061335398255964, + "loss": 0.4895, + "step": 16662 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010612391554461718, + "loss": 0.5746, + "step": 16663 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001061142912067002, + "loss": 0.5606, + "step": 16664 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010610466681193499, + "loss": 0.4675, + "step": 16665 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010609504236041102, + "loss": 0.5111, + "step": 16666 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010608541785221778, + "loss": 0.5524, + "step": 16667 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010607579328744474, + "loss": 0.5419, + "step": 16668 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010606616866618141, + "loss": 0.5718, + "step": 16669 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010605654398851725, + "loss": 0.4807, + "step": 16670 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010604691925454177, + "loss": 0.5757, + "step": 16671 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010603729446434443, + "loss": 0.5813, + "step": 16672 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010602766961801476, + "loss": 0.4842, + "step": 16673 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010601804471564216, + "loss": 0.6391, + "step": 16674 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010600841975731626, + "loss": 0.5892, + "step": 16675 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001059987947431264, + "loss": 0.5108, + "step": 16676 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010598916967316218, + "loss": 0.5941, + "step": 16677 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010597954454751302, + "loss": 0.5118, + "step": 16678 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010596991936626843, + "loss": 0.5412, + "step": 16679 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010596029412951794, + "loss": 0.4087, + "step": 16680 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010595066883735101, + "loss": 0.5309, + "step": 16681 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010594104348985712, + "loss": 0.459, + "step": 16682 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010593141808712576, + "loss": 0.61, + "step": 16683 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010592179262924646, + "loss": 0.5103, + "step": 16684 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010591216711630869, + "loss": 0.5943, + "step": 16685 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010590254154840194, + "loss": 0.5122, + "step": 16686 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010589291592561574, + "loss": 0.5909, + "step": 16687 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010588329024803951, + "loss": 0.706, + "step": 16688 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010587366451576284, + "loss": 0.5537, + "step": 16689 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010586403872887512, + "loss": 0.5777, + "step": 16690 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010585441288746594, + "loss": 0.5329, + "step": 16691 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001058447869916248, + "loss": 0.524, + "step": 16692 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010583516104144109, + "loss": 0.4657, + "step": 16693 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010582553503700443, + "loss": 0.604, + "step": 16694 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010581590897840423, + "loss": 0.5741, + "step": 16695 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010580628286573005, + "loss": 0.5534, + "step": 16696 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010579665669907134, + "loss": 0.631, + "step": 16697 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010578703047851765, + "loss": 0.6302, + "step": 16698 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010577740420415845, + "loss": 0.6119, + "step": 16699 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010576777787608325, + "loss": 0.4677, + "step": 16700 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010575815149438155, + "loss": 0.5053, + "step": 16701 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010574852505914286, + "loss": 0.5313, + "step": 16702 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010573889857045665, + "loss": 0.5359, + "step": 16703 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010572927202841247, + "loss": 0.5392, + "step": 16704 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010571964543309978, + "loss": 0.5187, + "step": 16705 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010571001878460816, + "loss": 0.4535, + "step": 16706 + }, + { + "epoch": 0.5, + "learning_rate": 0.000105700392083027, + "loss": 0.5716, + "step": 16707 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010569076532844588, + "loss": 0.4831, + "step": 16708 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010568113852095432, + "loss": 0.4555, + "step": 16709 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010567151166064178, + "loss": 0.5064, + "step": 16710 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010566188474759779, + "loss": 0.5019, + "step": 16711 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010565225778191186, + "loss": 0.6817, + "step": 16712 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010564263076367347, + "loss": 0.4971, + "step": 16713 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010563300369297215, + "loss": 0.6031, + "step": 16714 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010562337656989743, + "loss": 0.4959, + "step": 16715 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010561374939453878, + "loss": 0.573, + "step": 16716 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001056041221669857, + "loss": 0.5444, + "step": 16717 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010559449488732778, + "loss": 0.5525, + "step": 16718 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010558486755565442, + "loss": 0.5605, + "step": 16719 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010557524017205525, + "loss": 0.5238, + "step": 16720 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010556561273661965, + "loss": 0.5061, + "step": 16721 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010555598524943724, + "loss": 0.5685, + "step": 16722 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001055463577105975, + "loss": 0.4734, + "step": 16723 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010553673012018992, + "loss": 0.4017, + "step": 16724 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010552710247830406, + "loss": 0.6399, + "step": 16725 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010551747478502935, + "loss": 0.4924, + "step": 16726 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001055078470404554, + "loss": 0.4548, + "step": 16727 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010549821924467166, + "loss": 0.5138, + "step": 16728 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010548859139776768, + "loss": 0.642, + "step": 16729 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010547896349983297, + "loss": 0.6381, + "step": 16730 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010546933555095703, + "loss": 0.6304, + "step": 16731 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001054597075512294, + "loss": 0.5191, + "step": 16732 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010545007950073957, + "loss": 0.5943, + "step": 16733 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010544045139957708, + "loss": 0.6301, + "step": 16734 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010543082324783146, + "loss": 0.5793, + "step": 16735 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010542119504559216, + "loss": 0.4994, + "step": 16736 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010541156679294881, + "loss": 0.7061, + "step": 16737 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010540193848999083, + "loss": 0.5839, + "step": 16738 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010539231013680779, + "loss": 0.5266, + "step": 16739 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001053826817334892, + "loss": 0.5906, + "step": 16740 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010537305328012457, + "loss": 0.5325, + "step": 16741 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010536342477680345, + "loss": 0.5739, + "step": 16742 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010535379622361534, + "loss": 0.4468, + "step": 16743 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010534416762064975, + "loss": 0.4306, + "step": 16744 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010533453896799626, + "loss": 0.6534, + "step": 16745 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010532491026574431, + "loss": 0.5766, + "step": 16746 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010531528151398349, + "loss": 0.6041, + "step": 16747 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010530565271280326, + "loss": 0.5328, + "step": 16748 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010529602386229325, + "loss": 0.6265, + "step": 16749 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010528639496254286, + "loss": 0.5941, + "step": 16750 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010527676601364173, + "loss": 0.6653, + "step": 16751 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001052671370156793, + "loss": 0.4933, + "step": 16752 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010525750796874514, + "loss": 0.5575, + "step": 16753 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010524787887292876, + "loss": 0.5591, + "step": 16754 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010523824972831969, + "loss": 0.5564, + "step": 16755 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010522862053500748, + "loss": 0.6326, + "step": 16756 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010521899129308162, + "loss": 0.5941, + "step": 16757 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010520936200263168, + "loss": 0.6835, + "step": 16758 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010519973266374717, + "loss": 0.5165, + "step": 16759 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010519010327651759, + "loss": 0.5346, + "step": 16760 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010518047384103251, + "loss": 0.5583, + "step": 16761 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010517084435738144, + "loss": 0.5863, + "step": 16762 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010516121482565395, + "loss": 0.6252, + "step": 16763 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010515158524593953, + "loss": 0.4166, + "step": 16764 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010514195561832769, + "loss": 0.5049, + "step": 16765 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010513232594290805, + "loss": 0.6614, + "step": 16766 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010512269621977004, + "loss": 0.5514, + "step": 16767 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010511306644900329, + "loss": 0.627, + "step": 16768 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010510343663069724, + "loss": 0.5135, + "step": 16769 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010509380676494149, + "loss": 0.5714, + "step": 16770 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010508417685182557, + "loss": 0.5244, + "step": 16771 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010507454689143899, + "loss": 0.5921, + "step": 16772 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010506491688387127, + "loss": 0.5672, + "step": 16773 + }, + { + "epoch": 0.5, + "learning_rate": 0.000105055286829212, + "loss": 0.5698, + "step": 16774 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010504565672755069, + "loss": 0.5484, + "step": 16775 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010503602657897687, + "loss": 0.5779, + "step": 16776 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010502639638358007, + "loss": 0.5943, + "step": 16777 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010501676614144987, + "loss": 0.5614, + "step": 16778 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010500713585267571, + "loss": 0.4875, + "step": 16779 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010499750551734729, + "loss": 0.5304, + "step": 16780 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010498787513555397, + "loss": 0.4746, + "step": 16781 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010497824470738544, + "loss": 0.5336, + "step": 16782 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010496861423293112, + "loss": 0.5968, + "step": 16783 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010495898371228064, + "loss": 0.5668, + "step": 16784 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001049493531455235, + "loss": 0.602, + "step": 16785 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010493972253274924, + "loss": 0.5749, + "step": 16786 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001049300918740474, + "loss": 0.5441, + "step": 16787 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010492046116950756, + "loss": 0.5316, + "step": 16788 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010491083041921923, + "loss": 0.4883, + "step": 16789 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010490119962327194, + "loss": 0.513, + "step": 16790 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010489156878175525, + "loss": 0.6099, + "step": 16791 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010488193789475871, + "loss": 0.5614, + "step": 16792 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010487230696237184, + "loss": 0.55, + "step": 16793 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010486267598468422, + "loss": 0.5385, + "step": 16794 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010485304496178539, + "loss": 0.4947, + "step": 16795 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010484341389376484, + "loss": 0.4614, + "step": 16796 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001048337827807122, + "loss": 0.5415, + "step": 16797 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010482415162271694, + "loss": 0.6136, + "step": 16798 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010481452041986867, + "loss": 0.6181, + "step": 16799 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010480488917225688, + "loss": 0.468, + "step": 16800 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010479525787997118, + "loss": 0.5147, + "step": 16801 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010478562654310105, + "loss": 0.5072, + "step": 16802 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010477599516173608, + "loss": 0.5537, + "step": 16803 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010476636373596582, + "loss": 0.5738, + "step": 16804 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010475673226587981, + "loss": 0.5959, + "step": 16805 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010474710075156759, + "loss": 0.4929, + "step": 16806 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010473746919311873, + "loss": 0.6244, + "step": 16807 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010472783759062277, + "loss": 0.5812, + "step": 16808 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010471820594416925, + "loss": 0.5541, + "step": 16809 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010470857425384772, + "loss": 0.3639, + "step": 16810 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010469894251974779, + "loss": 0.5425, + "step": 16811 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010468931074195891, + "loss": 0.4851, + "step": 16812 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010467967892057073, + "loss": 0.6835, + "step": 16813 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010467004705567274, + "loss": 0.5257, + "step": 16814 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010466041514735451, + "loss": 0.4289, + "step": 16815 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001046507831957056, + "loss": 0.5721, + "step": 16816 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010464115120081557, + "loss": 0.7151, + "step": 16817 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010463151916277396, + "loss": 0.6363, + "step": 16818 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010462188708167033, + "loss": 0.4255, + "step": 16819 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010461225495759426, + "loss": 0.4763, + "step": 16820 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010460262279063524, + "loss": 0.4684, + "step": 16821 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001045929905808829, + "loss": 0.5721, + "step": 16822 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010458335832842676, + "loss": 0.5438, + "step": 16823 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010457372603335635, + "loss": 0.5632, + "step": 16824 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001045640936957613, + "loss": 0.5462, + "step": 16825 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001045544613157311, + "loss": 0.4901, + "step": 16826 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010454482889335533, + "loss": 0.6279, + "step": 16827 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010453519642872358, + "loss": 0.6146, + "step": 16828 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010452556392192534, + "loss": 0.6006, + "step": 16829 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010451593137305027, + "loss": 0.5593, + "step": 16830 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010450629878218781, + "loss": 0.5436, + "step": 16831 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010449666614942761, + "loss": 0.4738, + "step": 16832 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001044870334748592, + "loss": 0.5443, + "step": 16833 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010447740075857213, + "loss": 0.5209, + "step": 16834 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010446776800065599, + "loss": 0.4907, + "step": 16835 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010445813520120031, + "loss": 0.5052, + "step": 16836 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010444850236029467, + "loss": 0.4835, + "step": 16837 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010443886947802862, + "loss": 0.5392, + "step": 16838 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001044292365544917, + "loss": 0.4617, + "step": 16839 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010441960358977358, + "loss": 0.5509, + "step": 16840 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010440997058396369, + "loss": 0.6198, + "step": 16841 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010440033753715169, + "loss": 0.5095, + "step": 16842 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010439070444942706, + "loss": 0.5094, + "step": 16843 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010438107132087945, + "loss": 0.5293, + "step": 16844 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010437143815159836, + "loss": 0.5498, + "step": 16845 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010436180494167337, + "loss": 0.5371, + "step": 16846 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010435217169119409, + "loss": 0.4436, + "step": 16847 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010434253840025005, + "loss": 0.5277, + "step": 16848 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010433290506893082, + "loss": 0.5324, + "step": 16849 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010432327169732594, + "loss": 0.7066, + "step": 16850 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010431363828552502, + "loss": 0.6291, + "step": 16851 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010430400483361759, + "loss": 0.5717, + "step": 16852 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010429437134169324, + "loss": 0.5843, + "step": 16853 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010428473780984158, + "loss": 0.5266, + "step": 16854 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001042751042381521, + "loss": 0.5849, + "step": 16855 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010426547062671442, + "loss": 0.5077, + "step": 16856 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001042558369756181, + "loss": 0.5762, + "step": 16857 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010424620328495265, + "loss": 0.6312, + "step": 16858 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010423656955480776, + "loss": 0.4705, + "step": 16859 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001042269357852729, + "loss": 0.7152, + "step": 16860 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001042173019764377, + "loss": 0.51, + "step": 16861 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010420766812839167, + "loss": 0.4446, + "step": 16862 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010419803424122445, + "loss": 0.515, + "step": 16863 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010418840031502557, + "loss": 0.447, + "step": 16864 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010417876634988461, + "loss": 0.4498, + "step": 16865 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010416913234589116, + "loss": 0.4814, + "step": 16866 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010415949830313476, + "loss": 0.6029, + "step": 16867 + }, + { + "epoch": 0.5, + "learning_rate": 0.000104149864221705, + "loss": 0.5353, + "step": 16868 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010414023010169149, + "loss": 0.6504, + "step": 16869 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010413059594318372, + "loss": 0.5526, + "step": 16870 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010412096174627137, + "loss": 0.4973, + "step": 16871 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010411132751104391, + "loss": 0.5032, + "step": 16872 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010410169323759103, + "loss": 0.484, + "step": 16873 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010409205892600217, + "loss": 0.6118, + "step": 16874 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010408242457636701, + "loss": 0.5517, + "step": 16875 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010407279018877509, + "loss": 0.5117, + "step": 16876 + }, + { + "epoch": 0.5, + "learning_rate": 0.000104063155763316, + "loss": 0.5881, + "step": 16877 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010405352130007928, + "loss": 0.5138, + "step": 16878 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010404388679915456, + "loss": 0.4449, + "step": 16879 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001040342522606314, + "loss": 0.5525, + "step": 16880 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010402461768459934, + "loss": 0.5097, + "step": 16881 + }, + { + "epoch": 0.5, + "learning_rate": 0.000104014983071148, + "loss": 0.5803, + "step": 16882 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010400534842036697, + "loss": 0.636, + "step": 16883 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010399571373234575, + "loss": 0.485, + "step": 16884 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010398607900717406, + "loss": 0.6229, + "step": 16885 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010397644424494132, + "loss": 0.5723, + "step": 16886 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010396680944573724, + "loss": 0.4705, + "step": 16887 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010395717460965133, + "loss": 0.492, + "step": 16888 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010394753973677316, + "loss": 0.5412, + "step": 16889 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010393790482719238, + "loss": 0.5892, + "step": 16890 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010392826988099849, + "loss": 0.5308, + "step": 16891 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010391863489828116, + "loss": 0.4829, + "step": 16892 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010390899987912987, + "loss": 0.5336, + "step": 16893 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010389936482363429, + "loss": 0.494, + "step": 16894 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010388972973188398, + "loss": 0.5205, + "step": 16895 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001038800946039685, + "loss": 0.5843, + "step": 16896 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010387045943997744, + "loss": 0.4857, + "step": 16897 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010386082424000039, + "loss": 0.6004, + "step": 16898 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010385118900412693, + "loss": 0.535, + "step": 16899 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010384155373244668, + "loss": 0.5512, + "step": 16900 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010383191842504916, + "loss": 0.5092, + "step": 16901 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010382228308202402, + "loss": 0.5484, + "step": 16902 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010381264770346077, + "loss": 0.5188, + "step": 16903 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001038030122894491, + "loss": 0.5548, + "step": 16904 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010379337684007848, + "loss": 0.5469, + "step": 16905 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001037837413554386, + "loss": 0.4739, + "step": 16906 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010377410583561897, + "loss": 0.5687, + "step": 16907 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010376447028070924, + "loss": 0.4794, + "step": 16908 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010375483469079894, + "loss": 0.5531, + "step": 16909 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010374519906597771, + "loss": 0.5452, + "step": 16910 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010373556340633508, + "loss": 0.5608, + "step": 16911 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010372592771196068, + "loss": 0.4889, + "step": 16912 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001037162919829441, + "loss": 0.5047, + "step": 16913 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010370665621937491, + "loss": 0.5744, + "step": 16914 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010369702042134268, + "loss": 0.6192, + "step": 16915 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001036873845889371, + "loss": 0.5434, + "step": 16916 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001036777487222476, + "loss": 0.5533, + "step": 16917 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010366811282136393, + "loss": 0.4399, + "step": 16918 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010365847688637555, + "loss": 0.5047, + "step": 16919 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010364884091737215, + "loss": 0.5266, + "step": 16920 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001036392049144433, + "loss": 0.4606, + "step": 16921 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010362956887767852, + "loss": 0.498, + "step": 16922 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010361993280716751, + "loss": 0.5709, + "step": 16923 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010361029670299975, + "loss": 0.6176, + "step": 16924 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010360066056526492, + "loss": 0.5672, + "step": 16925 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010359102439405259, + "loss": 0.4978, + "step": 16926 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010358138818945233, + "loss": 0.5082, + "step": 16927 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010357175195155377, + "loss": 0.4808, + "step": 16928 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010356211568044647, + "loss": 0.5056, + "step": 16929 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010355247937622004, + "loss": 0.5706, + "step": 16930 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010354284303896408, + "loss": 0.4867, + "step": 16931 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010353320666876816, + "loss": 0.6252, + "step": 16932 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001035235702657219, + "loss": 0.5133, + "step": 16933 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010351393382991488, + "loss": 0.5651, + "step": 16934 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010350429736143675, + "loss": 0.5564, + "step": 16935 + }, + { + "epoch": 0.5, + "learning_rate": 0.000103494660860377, + "loss": 0.5869, + "step": 16936 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010348502432682529, + "loss": 0.546, + "step": 16937 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010347538776087124, + "loss": 0.5064, + "step": 16938 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010346575116260442, + "loss": 0.5001, + "step": 16939 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010345611453211443, + "loss": 0.5947, + "step": 16940 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010344647786949084, + "loss": 0.559, + "step": 16941 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010343684117482331, + "loss": 0.5545, + "step": 16942 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010342720444820138, + "loss": 0.5074, + "step": 16943 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010341756768971464, + "loss": 0.653, + "step": 16944 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010340793089945276, + "loss": 0.5997, + "step": 16945 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010339829407750526, + "loss": 0.58, + "step": 16946 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010338865722396182, + "loss": 0.5123, + "step": 16947 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010337902033891195, + "loss": 0.5272, + "step": 16948 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010336938342244536, + "loss": 0.5804, + "step": 16949 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001033597464746515, + "loss": 0.5441, + "step": 16950 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010335010949562012, + "loss": 0.5908, + "step": 16951 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010334047248544072, + "loss": 0.6104, + "step": 16952 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010333083544420297, + "loss": 0.4359, + "step": 16953 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010332119837199645, + "loss": 0.5598, + "step": 16954 + }, + { + "epoch": 0.5, + "learning_rate": 0.0001033115612689107, + "loss": 0.4843, + "step": 16955 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010330192413503542, + "loss": 0.6161, + "step": 16956 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010329228697046014, + "loss": 0.4985, + "step": 16957 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010328264977527448, + "loss": 0.5641, + "step": 16958 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010327301254956807, + "loss": 0.5213, + "step": 16959 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010326337529343049, + "loss": 0.5427, + "step": 16960 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010325373800695135, + "loss": 0.5362, + "step": 16961 + }, + { + "epoch": 0.5, + "learning_rate": 0.00010324410069022023, + "loss": 0.5112, + "step": 16962 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010323446334332675, + "loss": 0.5973, + "step": 16963 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010322482596636059, + "loss": 0.5769, + "step": 16964 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010321518855941119, + "loss": 0.5495, + "step": 16965 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010320555112256832, + "loss": 0.5991, + "step": 16966 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010319591365592147, + "loss": 0.629, + "step": 16967 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010318627615956029, + "loss": 0.5113, + "step": 16968 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010317663863357441, + "loss": 0.5021, + "step": 16969 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010316700107805339, + "loss": 0.6415, + "step": 16970 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010315736349308687, + "loss": 0.4858, + "step": 16971 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010314772587876445, + "loss": 0.693, + "step": 16972 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001031380882351757, + "loss": 0.6031, + "step": 16973 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010312845056241028, + "loss": 0.5375, + "step": 16974 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010311881286055774, + "loss": 0.4932, + "step": 16975 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010310917512970776, + "loss": 0.5365, + "step": 16976 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010309953736994989, + "loss": 0.516, + "step": 16977 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010308989958137378, + "loss": 0.5786, + "step": 16978 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010308026176406898, + "loss": 0.3788, + "step": 16979 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010307062391812516, + "loss": 0.6079, + "step": 16980 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010306098604363188, + "loss": 0.5153, + "step": 16981 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010305134814067879, + "loss": 0.5784, + "step": 16982 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010304171020935547, + "loss": 0.5591, + "step": 16983 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010303207224975155, + "loss": 0.5836, + "step": 16984 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010302243426195664, + "loss": 0.6549, + "step": 16985 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010301279624606033, + "loss": 0.4761, + "step": 16986 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010300315820215223, + "loss": 0.6465, + "step": 16987 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010299352013032197, + "loss": 0.5161, + "step": 16988 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010298388203065916, + "loss": 0.5504, + "step": 16989 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010297424390325341, + "loss": 0.5404, + "step": 16990 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001029646057481943, + "loss": 0.5647, + "step": 16991 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001029549675655715, + "loss": 0.465, + "step": 16992 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010294532935547456, + "loss": 0.5217, + "step": 16993 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010293569111799313, + "loss": 0.6306, + "step": 16994 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010292605285321684, + "loss": 0.5037, + "step": 16995 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010291641456123522, + "loss": 0.4583, + "step": 16996 + }, + { + "epoch": 0.51, + "learning_rate": 0.000102906776242138, + "loss": 0.5543, + "step": 16997 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010289713789601469, + "loss": 0.7142, + "step": 16998 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010288749952295496, + "loss": 0.5984, + "step": 16999 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010287786112304842, + "loss": 0.6804, + "step": 17000 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010286822269638468, + "loss": 0.494, + "step": 17001 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010285858424305332, + "loss": 0.6153, + "step": 17002 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010284894576314402, + "loss": 0.5069, + "step": 17003 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010283930725674633, + "loss": 0.6294, + "step": 17004 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001028296687239499, + "loss": 0.6896, + "step": 17005 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010282003016484432, + "loss": 0.3866, + "step": 17006 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010281039157951926, + "loss": 0.5592, + "step": 17007 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010280075296806426, + "loss": 0.5457, + "step": 17008 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010279111433056903, + "loss": 0.5806, + "step": 17009 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010278147566712306, + "loss": 0.6269, + "step": 17010 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010277183697781607, + "loss": 0.7078, + "step": 17011 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010276219826273765, + "loss": 0.5614, + "step": 17012 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010275255952197742, + "loss": 0.4925, + "step": 17013 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010274292075562497, + "loss": 0.529, + "step": 17014 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010273328196376997, + "loss": 0.6071, + "step": 17015 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010272364314650198, + "loss": 0.5945, + "step": 17016 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010271400430391063, + "loss": 0.5291, + "step": 17017 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010270436543608556, + "loss": 0.6018, + "step": 17018 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010269472654311638, + "loss": 0.4285, + "step": 17019 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001026850876250927, + "loss": 0.5475, + "step": 17020 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010267544868210417, + "loss": 0.3992, + "step": 17021 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010266580971424035, + "loss": 0.5101, + "step": 17022 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010265617072159093, + "loss": 0.5539, + "step": 17023 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010264653170424546, + "loss": 0.5088, + "step": 17024 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001026368926622936, + "loss": 0.4508, + "step": 17025 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010262725359582497, + "loss": 0.5168, + "step": 17026 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010261761450492917, + "loss": 0.5856, + "step": 17027 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010260797538969587, + "loss": 0.5879, + "step": 17028 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001025983362502146, + "loss": 0.5567, + "step": 17029 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010258869708657508, + "loss": 0.4864, + "step": 17030 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010257905789886686, + "loss": 0.5322, + "step": 17031 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001025694186871796, + "loss": 0.4651, + "step": 17032 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010255977945160292, + "loss": 0.6345, + "step": 17033 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010255014019222642, + "loss": 0.6824, + "step": 17034 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010254050090913972, + "loss": 0.5364, + "step": 17035 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010253086160243246, + "loss": 0.5879, + "step": 17036 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010252122227219425, + "loss": 0.5026, + "step": 17037 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010251158291851475, + "loss": 0.4907, + "step": 17038 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010250194354148352, + "loss": 0.5961, + "step": 17039 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010249230414119027, + "loss": 0.5152, + "step": 17040 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010248266471772449, + "loss": 0.538, + "step": 17041 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010247302527117595, + "loss": 0.5548, + "step": 17042 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010246338580163417, + "loss": 0.5369, + "step": 17043 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010245374630918882, + "loss": 0.5281, + "step": 17044 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010244410679392953, + "loss": 0.5824, + "step": 17045 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010243446725594589, + "loss": 0.5667, + "step": 17046 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010242482769532755, + "loss": 0.4981, + "step": 17047 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010241518811216413, + "loss": 0.5617, + "step": 17048 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010240554850654527, + "loss": 0.5557, + "step": 17049 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010239590887856055, + "loss": 0.5971, + "step": 17050 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010238626922829961, + "loss": 0.4616, + "step": 17051 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010237662955585215, + "loss": 0.6056, + "step": 17052 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010236698986130771, + "loss": 0.5261, + "step": 17053 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010235735014475592, + "loss": 0.5879, + "step": 17054 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010234771040628644, + "loss": 0.4905, + "step": 17055 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010233807064598886, + "loss": 0.6432, + "step": 17056 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010232843086395287, + "loss": 0.6237, + "step": 17057 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010231879106026802, + "loss": 0.5856, + "step": 17058 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010230915123502404, + "loss": 0.5771, + "step": 17059 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010229951138831042, + "loss": 0.5745, + "step": 17060 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001022898715202169, + "loss": 0.4906, + "step": 17061 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010228023163083306, + "loss": 0.4976, + "step": 17062 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010227059172024853, + "loss": 0.571, + "step": 17063 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010226095178855295, + "loss": 0.5362, + "step": 17064 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010225131183583594, + "loss": 0.5587, + "step": 17065 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010224167186218711, + "loss": 0.4978, + "step": 17066 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010223203186769614, + "loss": 0.5583, + "step": 17067 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001022223918524526, + "loss": 0.5162, + "step": 17068 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010221275181654617, + "loss": 0.5364, + "step": 17069 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010220311176006642, + "loss": 0.4999, + "step": 17070 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010219347168310307, + "loss": 0.5778, + "step": 17071 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010218383158574565, + "loss": 0.5294, + "step": 17072 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010217419146808385, + "loss": 0.5094, + "step": 17073 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010216455133020728, + "loss": 0.5427, + "step": 17074 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010215491117220557, + "loss": 0.4397, + "step": 17075 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010214527099416837, + "loss": 0.527, + "step": 17076 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010213563079618529, + "loss": 0.5986, + "step": 17077 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010212599057834596, + "loss": 0.5653, + "step": 17078 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010211635034074003, + "loss": 0.5204, + "step": 17079 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001021067100834571, + "loss": 0.6139, + "step": 17080 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010209706980658682, + "loss": 0.4519, + "step": 17081 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010208742951021881, + "loss": 0.5675, + "step": 17082 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010207778919444275, + "loss": 0.6373, + "step": 17083 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010206814885934821, + "loss": 0.5256, + "step": 17084 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010205850850502488, + "loss": 0.5126, + "step": 17085 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010204886813156232, + "loss": 0.4631, + "step": 17086 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001020392277390502, + "loss": 0.5725, + "step": 17087 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010202958732757817, + "loss": 0.7214, + "step": 17088 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010201994689723582, + "loss": 0.4972, + "step": 17089 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010201030644811285, + "loss": 0.543, + "step": 17090 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010200066598029881, + "loss": 0.5099, + "step": 17091 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010199102549388339, + "loss": 0.4111, + "step": 17092 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010198138498895621, + "loss": 0.56, + "step": 17093 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010197174446560691, + "loss": 0.5629, + "step": 17094 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010196210392392514, + "loss": 0.5226, + "step": 17095 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010195246336400046, + "loss": 0.4992, + "step": 17096 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010194282278592257, + "loss": 0.569, + "step": 17097 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001019331821897811, + "loss": 0.4518, + "step": 17098 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010192354157566565, + "loss": 0.4589, + "step": 17099 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001019139009436659, + "loss": 0.5497, + "step": 17100 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010190426029387145, + "loss": 0.5643, + "step": 17101 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010189461962637196, + "loss": 0.5439, + "step": 17102 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010188497894125704, + "loss": 0.629, + "step": 17103 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010187533823861635, + "loss": 0.6186, + "step": 17104 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001018656975185395, + "loss": 0.5283, + "step": 17105 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010185605678111614, + "loss": 0.5493, + "step": 17106 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001018464160264359, + "loss": 0.5138, + "step": 17107 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010183677525458844, + "loss": 0.5805, + "step": 17108 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010182713446566335, + "loss": 0.5547, + "step": 17109 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001018174936597503, + "loss": 0.4657, + "step": 17110 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010180785283693892, + "loss": 0.5132, + "step": 17111 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010179821199731885, + "loss": 0.5263, + "step": 17112 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010178857114097968, + "loss": 0.6511, + "step": 17113 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010177893026801116, + "loss": 0.599, + "step": 17114 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001017692893785028, + "loss": 0.6603, + "step": 17115 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010175964847254434, + "loss": 0.5806, + "step": 17116 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010175000755022532, + "loss": 0.6484, + "step": 17117 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010174036661163544, + "loss": 0.5097, + "step": 17118 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010173072565686437, + "loss": 0.7142, + "step": 17119 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010172108468600163, + "loss": 0.6435, + "step": 17120 + }, + { + "epoch": 0.51, + "learning_rate": 0.000101711443699137, + "loss": 0.5007, + "step": 17121 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010170180269636, + "loss": 0.5593, + "step": 17122 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010169216167776035, + "loss": 0.5118, + "step": 17123 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010168252064342763, + "loss": 0.4817, + "step": 17124 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010167287959345151, + "loss": 0.577, + "step": 17125 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010166323852792164, + "loss": 0.5259, + "step": 17126 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010165359744692762, + "loss": 0.6045, + "step": 17127 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001016439563505591, + "loss": 0.5933, + "step": 17128 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010163431523890574, + "loss": 0.5562, + "step": 17129 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010162467411205717, + "loss": 0.522, + "step": 17130 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010161503297010304, + "loss": 0.6175, + "step": 17131 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010160539181313295, + "loss": 0.5038, + "step": 17132 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001015957506412366, + "loss": 0.6042, + "step": 17133 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010158610945450358, + "loss": 0.4573, + "step": 17134 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010157646825302353, + "loss": 0.4838, + "step": 17135 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010156682703688614, + "loss": 0.4867, + "step": 17136 + }, + { + "epoch": 0.51, + "learning_rate": 0.000101557185806181, + "loss": 0.4895, + "step": 17137 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010154754456099777, + "loss": 0.4635, + "step": 17138 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010153790330142608, + "loss": 0.6027, + "step": 17139 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010152826202755559, + "loss": 0.6136, + "step": 17140 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001015186207394759, + "loss": 0.5734, + "step": 17141 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001015089794372767, + "loss": 0.5475, + "step": 17142 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010149933812104764, + "loss": 0.6113, + "step": 17143 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010148969679087827, + "loss": 0.6822, + "step": 17144 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010148005544685837, + "loss": 0.5786, + "step": 17145 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010147041408907743, + "loss": 0.526, + "step": 17146 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010146077271762519, + "loss": 0.5171, + "step": 17147 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010145113133259127, + "loss": 0.6599, + "step": 17148 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010144148993406533, + "loss": 0.5004, + "step": 17149 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010143184852213698, + "loss": 0.5097, + "step": 17150 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010142220709689587, + "loss": 0.5993, + "step": 17151 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010141256565843167, + "loss": 0.4755, + "step": 17152 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010140292420683395, + "loss": 0.5324, + "step": 17153 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010139328274219242, + "loss": 0.6279, + "step": 17154 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001013836412645967, + "loss": 0.5687, + "step": 17155 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010137399977413644, + "loss": 0.5696, + "step": 17156 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010136435827090127, + "loss": 0.4782, + "step": 17157 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010135471675498085, + "loss": 0.5231, + "step": 17158 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010134507522646481, + "loss": 0.5355, + "step": 17159 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001013354336854428, + "loss": 0.4494, + "step": 17160 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010132579213200444, + "loss": 0.45, + "step": 17161 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010131615056623943, + "loss": 0.5877, + "step": 17162 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010130650898823732, + "loss": 0.6058, + "step": 17163 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010129686739808789, + "loss": 0.6775, + "step": 17164 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010128722579588063, + "loss": 0.5264, + "step": 17165 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001012775841817053, + "loss": 0.5319, + "step": 17166 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010126794255565148, + "loss": 0.6191, + "step": 17167 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010125830091780886, + "loss": 0.537, + "step": 17168 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010124865926826703, + "loss": 0.5321, + "step": 17169 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010123901760711567, + "loss": 0.4735, + "step": 17170 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010122937593444444, + "loss": 0.5834, + "step": 17171 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010121973425034295, + "loss": 0.4736, + "step": 17172 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010121009255490084, + "loss": 0.481, + "step": 17173 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001012004508482078, + "loss": 0.601, + "step": 17174 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010119080913035342, + "loss": 0.6889, + "step": 17175 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001011811674014274, + "loss": 0.5411, + "step": 17176 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010117152566151933, + "loss": 0.5104, + "step": 17177 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010116188391071889, + "loss": 0.5341, + "step": 17178 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010115224214911571, + "loss": 0.502, + "step": 17179 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010114260037679945, + "loss": 0.5671, + "step": 17180 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010113295859385976, + "loss": 0.5405, + "step": 17181 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010112331680038627, + "loss": 0.5857, + "step": 17182 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010111367499646862, + "loss": 0.5666, + "step": 17183 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010110403318219645, + "loss": 0.5751, + "step": 17184 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010109439135765943, + "loss": 0.5763, + "step": 17185 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010108474952294722, + "loss": 0.5346, + "step": 17186 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001010751076781494, + "loss": 0.4783, + "step": 17187 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010106546582335567, + "loss": 0.5696, + "step": 17188 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001010558239586557, + "loss": 0.5043, + "step": 17189 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010104618208413905, + "loss": 0.4127, + "step": 17190 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010103654019989544, + "loss": 0.5308, + "step": 17191 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010102689830601446, + "loss": 0.5703, + "step": 17192 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010101725640258586, + "loss": 0.5125, + "step": 17193 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010100761448969913, + "loss": 0.5147, + "step": 17194 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010099797256744408, + "loss": 0.5359, + "step": 17195 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010098833063591023, + "loss": 0.6418, + "step": 17196 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010097868869518727, + "loss": 0.4853, + "step": 17197 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010096904674536486, + "loss": 0.5392, + "step": 17198 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010095940478653267, + "loss": 0.529, + "step": 17199 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010094976281878031, + "loss": 0.5081, + "step": 17200 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010094012084219742, + "loss": 0.5757, + "step": 17201 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010093047885687365, + "loss": 0.5992, + "step": 17202 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010092083686289868, + "loss": 0.5604, + "step": 17203 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001009111948603621, + "loss": 0.5198, + "step": 17204 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010090155284935363, + "loss": 0.6283, + "step": 17205 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010089191082996285, + "loss": 0.5666, + "step": 17206 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001008822688022795, + "loss": 0.4627, + "step": 17207 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001008726267663931, + "loss": 0.6264, + "step": 17208 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001008629847223934, + "loss": 0.6138, + "step": 17209 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010085334267036999, + "loss": 0.4942, + "step": 17210 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010084370061041255, + "loss": 0.4967, + "step": 17211 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010083405854261072, + "loss": 0.5703, + "step": 17212 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010082441646705416, + "loss": 0.5725, + "step": 17213 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010081477438383249, + "loss": 0.5856, + "step": 17214 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010080513229303537, + "loss": 0.6205, + "step": 17215 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010079549019475245, + "loss": 0.4387, + "step": 17216 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010078584808907339, + "loss": 0.5968, + "step": 17217 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010077620597608781, + "loss": 0.5191, + "step": 17218 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001007665638558854, + "loss": 0.5836, + "step": 17219 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010075692172855575, + "loss": 0.5135, + "step": 17220 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010074727959418857, + "loss": 0.5154, + "step": 17221 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010073763745287348, + "loss": 0.5832, + "step": 17222 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010072799530470012, + "loss": 0.4747, + "step": 17223 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010071835314975817, + "loss": 0.6028, + "step": 17224 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010070871098813723, + "loss": 0.5056, + "step": 17225 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010069906881992702, + "loss": 0.5187, + "step": 17226 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001006894266452171, + "loss": 0.5913, + "step": 17227 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010067978446409719, + "loss": 0.4263, + "step": 17228 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001006701422766569, + "loss": 0.52, + "step": 17229 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010066050008298591, + "loss": 0.5864, + "step": 17230 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010065085788317385, + "loss": 0.5026, + "step": 17231 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010064121567731038, + "loss": 0.4327, + "step": 17232 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010063157346548513, + "loss": 0.6828, + "step": 17233 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010062193124778776, + "loss": 0.586, + "step": 17234 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010061228902430791, + "loss": 0.601, + "step": 17235 + }, + { + "epoch": 0.51, + "learning_rate": 0.0001006026467951353, + "loss": 0.4847, + "step": 17236 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010059300456035945, + "loss": 0.593, + "step": 17237 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010058336232007013, + "loss": 0.4782, + "step": 17238 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010057372007435689, + "loss": 0.5267, + "step": 17239 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010056407782330945, + "loss": 0.4971, + "step": 17240 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010055443556701744, + "loss": 0.5416, + "step": 17241 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010054479330557055, + "loss": 0.6462, + "step": 17242 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010053515103905833, + "loss": 0.5427, + "step": 17243 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010052550876757052, + "loss": 0.6681, + "step": 17244 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010051586649119673, + "loss": 0.61, + "step": 17245 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010050622421002662, + "loss": 0.5266, + "step": 17246 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010049658192414985, + "loss": 0.5116, + "step": 17247 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010048693963365605, + "loss": 0.6249, + "step": 17248 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010047729733863484, + "loss": 0.5927, + "step": 17249 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010046765503917599, + "loss": 0.6125, + "step": 17250 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010045801273536904, + "loss": 0.4866, + "step": 17251 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010044837042730364, + "loss": 0.5815, + "step": 17252 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010043872811506949, + "loss": 0.5491, + "step": 17253 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010042908579875621, + "loss": 0.6055, + "step": 17254 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010041944347845349, + "loss": 0.6177, + "step": 17255 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010040980115425091, + "loss": 0.5248, + "step": 17256 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010040015882623821, + "loss": 0.6792, + "step": 17257 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010039051649450495, + "loss": 0.6725, + "step": 17258 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010038087415914084, + "loss": 0.5653, + "step": 17259 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010037123182023551, + "loss": 0.6541, + "step": 17260 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010036158947787862, + "loss": 0.5252, + "step": 17261 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010035194713215981, + "loss": 0.5138, + "step": 17262 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010034230478316875, + "loss": 0.5244, + "step": 17263 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010033266243099505, + "loss": 0.6133, + "step": 17264 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010032302007572843, + "loss": 0.5559, + "step": 17265 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010031337771745844, + "loss": 0.5385, + "step": 17266 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010030373535627485, + "loss": 0.6583, + "step": 17267 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010029409299226719, + "loss": 0.5673, + "step": 17268 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010028445062552523, + "loss": 0.5683, + "step": 17269 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010027480825613851, + "loss": 0.6251, + "step": 17270 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010026516588419675, + "loss": 0.4854, + "step": 17271 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010025552350978959, + "loss": 0.5282, + "step": 17272 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010024588113300669, + "loss": 0.5884, + "step": 17273 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010023623875393766, + "loss": 0.6901, + "step": 17274 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010022659637267218, + "loss": 0.6067, + "step": 17275 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010021695398929991, + "loss": 0.5349, + "step": 17276 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010020731160391048, + "loss": 0.5532, + "step": 17277 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010019766921659355, + "loss": 0.5977, + "step": 17278 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010018802682743877, + "loss": 0.608, + "step": 17279 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010017838443653575, + "loss": 0.6439, + "step": 17280 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010016874204397426, + "loss": 0.516, + "step": 17281 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010015909964984382, + "loss": 0.5303, + "step": 17282 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010014945725423417, + "loss": 0.543, + "step": 17283 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010013981485723491, + "loss": 0.5668, + "step": 17284 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010013017245893567, + "loss": 0.6453, + "step": 17285 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010012053005942621, + "loss": 0.4879, + "step": 17286 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010011088765879606, + "loss": 0.5711, + "step": 17287 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010010124525713496, + "loss": 0.5759, + "step": 17288 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010009160285453248, + "loss": 0.4631, + "step": 17289 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010008196045107832, + "loss": 0.5155, + "step": 17290 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010007231804686212, + "loss": 0.455, + "step": 17291 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010006267564197355, + "loss": 0.4515, + "step": 17292 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010005303323650225, + "loss": 0.5165, + "step": 17293 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010004339083053786, + "loss": 0.606, + "step": 17294 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010003374842417006, + "loss": 0.5068, + "step": 17295 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010002410601748847, + "loss": 0.5728, + "step": 17296 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010001446361058271, + "loss": 0.5105, + "step": 17297 + }, + { + "epoch": 0.51, + "learning_rate": 0.00010000482120354253, + "loss": 0.6074, + "step": 17298 + }, + { + "epoch": 0.52, + "learning_rate": 9.99951787964575e-05, + "loss": 0.6891, + "step": 17299 + }, + { + "epoch": 0.52, + "learning_rate": 9.998553638941731e-05, + "loss": 0.5168, + "step": 17300 + }, + { + "epoch": 0.52, + "learning_rate": 9.997589398251156e-05, + "loss": 0.4993, + "step": 17301 + }, + { + "epoch": 0.52, + "learning_rate": 9.996625157582998e-05, + "loss": 0.4962, + "step": 17302 + }, + { + "epoch": 0.52, + "learning_rate": 9.995660916946218e-05, + "loss": 0.5573, + "step": 17303 + }, + { + "epoch": 0.52, + "learning_rate": 9.994696676349776e-05, + "loss": 0.5691, + "step": 17304 + }, + { + "epoch": 0.52, + "learning_rate": 9.993732435802646e-05, + "loss": 0.5401, + "step": 17305 + }, + { + "epoch": 0.52, + "learning_rate": 9.992768195313787e-05, + "loss": 0.5233, + "step": 17306 + }, + { + "epoch": 0.52, + "learning_rate": 9.99180395489217e-05, + "loss": 0.4423, + "step": 17307 + }, + { + "epoch": 0.52, + "learning_rate": 9.990839714546756e-05, + "loss": 0.6081, + "step": 17308 + }, + { + "epoch": 0.52, + "learning_rate": 9.989875474286507e-05, + "loss": 0.6297, + "step": 17309 + }, + { + "epoch": 0.52, + "learning_rate": 9.988911234120394e-05, + "loss": 0.6364, + "step": 17310 + }, + { + "epoch": 0.52, + "learning_rate": 9.987946994057384e-05, + "loss": 0.6388, + "step": 17311 + }, + { + "epoch": 0.52, + "learning_rate": 9.986982754106434e-05, + "loss": 0.5626, + "step": 17312 + }, + { + "epoch": 0.52, + "learning_rate": 9.986018514276512e-05, + "loss": 0.517, + "step": 17313 + }, + { + "epoch": 0.52, + "learning_rate": 9.985054274576585e-05, + "loss": 0.4994, + "step": 17314 + }, + { + "epoch": 0.52, + "learning_rate": 9.984090035015621e-05, + "loss": 0.5334, + "step": 17315 + }, + { + "epoch": 0.52, + "learning_rate": 9.983125795602576e-05, + "loss": 0.4866, + "step": 17316 + }, + { + "epoch": 0.52, + "learning_rate": 9.982161556346423e-05, + "loss": 0.4484, + "step": 17317 + }, + { + "epoch": 0.52, + "learning_rate": 9.981197317256128e-05, + "loss": 0.7202, + "step": 17318 + }, + { + "epoch": 0.52, + "learning_rate": 9.980233078340649e-05, + "loss": 0.4913, + "step": 17319 + }, + { + "epoch": 0.52, + "learning_rate": 9.979268839608954e-05, + "loss": 0.4504, + "step": 17320 + }, + { + "epoch": 0.52, + "learning_rate": 9.978304601070012e-05, + "loss": 0.503, + "step": 17321 + }, + { + "epoch": 0.52, + "learning_rate": 9.977340362732787e-05, + "loss": 0.6213, + "step": 17322 + }, + { + "epoch": 0.52, + "learning_rate": 9.976376124606238e-05, + "loss": 0.5971, + "step": 17323 + }, + { + "epoch": 0.52, + "learning_rate": 9.975411886699334e-05, + "loss": 0.6646, + "step": 17324 + }, + { + "epoch": 0.52, + "learning_rate": 9.97444764902104e-05, + "loss": 0.5456, + "step": 17325 + }, + { + "epoch": 0.52, + "learning_rate": 9.973483411580326e-05, + "loss": 0.5881, + "step": 17326 + }, + { + "epoch": 0.52, + "learning_rate": 9.972519174386151e-05, + "loss": 0.6223, + "step": 17327 + }, + { + "epoch": 0.52, + "learning_rate": 9.971554937447479e-05, + "loss": 0.6065, + "step": 17328 + }, + { + "epoch": 0.52, + "learning_rate": 9.97059070077328e-05, + "loss": 0.5716, + "step": 17329 + }, + { + "epoch": 0.52, + "learning_rate": 9.96962646437252e-05, + "loss": 0.4827, + "step": 17330 + }, + { + "epoch": 0.52, + "learning_rate": 9.968662228254158e-05, + "loss": 0.5261, + "step": 17331 + }, + { + "epoch": 0.52, + "learning_rate": 9.967697992427161e-05, + "loss": 0.5223, + "step": 17332 + }, + { + "epoch": 0.52, + "learning_rate": 9.966733756900495e-05, + "loss": 0.5468, + "step": 17333 + }, + { + "epoch": 0.52, + "learning_rate": 9.96576952168313e-05, + "loss": 0.5032, + "step": 17334 + }, + { + "epoch": 0.52, + "learning_rate": 9.96480528678402e-05, + "loss": 0.4382, + "step": 17335 + }, + { + "epoch": 0.52, + "learning_rate": 9.963841052212141e-05, + "loss": 0.5335, + "step": 17336 + }, + { + "epoch": 0.52, + "learning_rate": 9.962876817976448e-05, + "loss": 0.4923, + "step": 17337 + }, + { + "epoch": 0.52, + "learning_rate": 9.961912584085919e-05, + "loss": 0.5889, + "step": 17338 + }, + { + "epoch": 0.52, + "learning_rate": 9.960948350549508e-05, + "loss": 0.5823, + "step": 17339 + }, + { + "epoch": 0.52, + "learning_rate": 9.959984117376182e-05, + "loss": 0.4648, + "step": 17340 + }, + { + "epoch": 0.52, + "learning_rate": 9.95901988457491e-05, + "loss": 0.5059, + "step": 17341 + }, + { + "epoch": 0.52, + "learning_rate": 9.958055652154656e-05, + "loss": 0.566, + "step": 17342 + }, + { + "epoch": 0.52, + "learning_rate": 9.957091420124381e-05, + "loss": 0.6177, + "step": 17343 + }, + { + "epoch": 0.52, + "learning_rate": 9.956127188493054e-05, + "loss": 0.5002, + "step": 17344 + }, + { + "epoch": 0.52, + "learning_rate": 9.955162957269638e-05, + "loss": 0.6043, + "step": 17345 + }, + { + "epoch": 0.52, + "learning_rate": 9.954198726463102e-05, + "loss": 0.4762, + "step": 17346 + }, + { + "epoch": 0.52, + "learning_rate": 9.953234496082403e-05, + "loss": 0.5842, + "step": 17347 + }, + { + "epoch": 0.52, + "learning_rate": 9.952270266136514e-05, + "loss": 0.552, + "step": 17348 + }, + { + "epoch": 0.52, + "learning_rate": 9.9513060366344e-05, + "loss": 0.5075, + "step": 17349 + }, + { + "epoch": 0.52, + "learning_rate": 9.950341807585017e-05, + "loss": 0.5252, + "step": 17350 + }, + { + "epoch": 0.52, + "learning_rate": 9.94937757899734e-05, + "loss": 0.5926, + "step": 17351 + }, + { + "epoch": 0.52, + "learning_rate": 9.94841335088033e-05, + "loss": 0.5218, + "step": 17352 + }, + { + "epoch": 0.52, + "learning_rate": 9.947449123242952e-05, + "loss": 0.4801, + "step": 17353 + }, + { + "epoch": 0.52, + "learning_rate": 9.946484896094168e-05, + "loss": 0.5494, + "step": 17354 + }, + { + "epoch": 0.52, + "learning_rate": 9.945520669442949e-05, + "loss": 0.5041, + "step": 17355 + }, + { + "epoch": 0.52, + "learning_rate": 9.944556443298254e-05, + "loss": 0.4987, + "step": 17356 + }, + { + "epoch": 0.52, + "learning_rate": 9.943592217669056e-05, + "loss": 0.6247, + "step": 17357 + }, + { + "epoch": 0.52, + "learning_rate": 9.942627992564312e-05, + "loss": 0.5602, + "step": 17358 + }, + { + "epoch": 0.52, + "learning_rate": 9.941663767992991e-05, + "loss": 0.5618, + "step": 17359 + }, + { + "epoch": 0.52, + "learning_rate": 9.940699543964056e-05, + "loss": 0.5686, + "step": 17360 + }, + { + "epoch": 0.52, + "learning_rate": 9.939735320486477e-05, + "loss": 0.4662, + "step": 17361 + }, + { + "epoch": 0.52, + "learning_rate": 9.93877109756921e-05, + "loss": 0.4871, + "step": 17362 + }, + { + "epoch": 0.52, + "learning_rate": 9.937806875221225e-05, + "loss": 0.5595, + "step": 17363 + }, + { + "epoch": 0.52, + "learning_rate": 9.936842653451489e-05, + "loss": 0.5795, + "step": 17364 + }, + { + "epoch": 0.52, + "learning_rate": 9.935878432268967e-05, + "loss": 0.5455, + "step": 17365 + }, + { + "epoch": 0.52, + "learning_rate": 9.934914211682616e-05, + "loss": 0.5631, + "step": 17366 + }, + { + "epoch": 0.52, + "learning_rate": 9.933949991701411e-05, + "loss": 0.5357, + "step": 17367 + }, + { + "epoch": 0.52, + "learning_rate": 9.932985772334309e-05, + "loss": 0.5595, + "step": 17368 + }, + { + "epoch": 0.52, + "learning_rate": 9.932021553590284e-05, + "loss": 0.5741, + "step": 17369 + }, + { + "epoch": 0.52, + "learning_rate": 9.931057335478292e-05, + "loss": 0.4876, + "step": 17370 + }, + { + "epoch": 0.52, + "learning_rate": 9.9300931180073e-05, + "loss": 0.4962, + "step": 17371 + }, + { + "epoch": 0.52, + "learning_rate": 9.929128901186276e-05, + "loss": 0.5115, + "step": 17372 + }, + { + "epoch": 0.52, + "learning_rate": 9.928164685024187e-05, + "loss": 0.4973, + "step": 17373 + }, + { + "epoch": 0.52, + "learning_rate": 9.92720046952999e-05, + "loss": 0.4655, + "step": 17374 + }, + { + "epoch": 0.52, + "learning_rate": 9.926236254712653e-05, + "loss": 0.5282, + "step": 17375 + }, + { + "epoch": 0.52, + "learning_rate": 9.925272040581144e-05, + "loss": 0.5224, + "step": 17376 + }, + { + "epoch": 0.52, + "learning_rate": 9.924307827144428e-05, + "loss": 0.6952, + "step": 17377 + }, + { + "epoch": 0.52, + "learning_rate": 9.923343614411463e-05, + "loss": 0.5385, + "step": 17378 + }, + { + "epoch": 0.52, + "learning_rate": 9.922379402391221e-05, + "loss": 0.5369, + "step": 17379 + }, + { + "epoch": 0.52, + "learning_rate": 9.921415191092662e-05, + "loss": 0.5546, + "step": 17380 + }, + { + "epoch": 0.52, + "learning_rate": 9.920450980524757e-05, + "loss": 0.589, + "step": 17381 + }, + { + "epoch": 0.52, + "learning_rate": 9.919486770696466e-05, + "loss": 0.5096, + "step": 17382 + }, + { + "epoch": 0.52, + "learning_rate": 9.918522561616755e-05, + "loss": 0.5507, + "step": 17383 + }, + { + "epoch": 0.52, + "learning_rate": 9.917558353294589e-05, + "loss": 0.4769, + "step": 17384 + }, + { + "epoch": 0.52, + "learning_rate": 9.91659414573893e-05, + "loss": 0.5033, + "step": 17385 + }, + { + "epoch": 0.52, + "learning_rate": 9.915629938958747e-05, + "loss": 0.576, + "step": 17386 + }, + { + "epoch": 0.52, + "learning_rate": 9.914665732963e-05, + "loss": 0.5635, + "step": 17387 + }, + { + "epoch": 0.52, + "learning_rate": 9.913701527760664e-05, + "loss": 0.4401, + "step": 17388 + }, + { + "epoch": 0.52, + "learning_rate": 9.912737323360692e-05, + "loss": 0.5469, + "step": 17389 + }, + { + "epoch": 0.52, + "learning_rate": 9.911773119772053e-05, + "loss": 0.4413, + "step": 17390 + }, + { + "epoch": 0.52, + "learning_rate": 9.910808917003713e-05, + "loss": 0.5268, + "step": 17391 + }, + { + "epoch": 0.52, + "learning_rate": 9.90984471506464e-05, + "loss": 0.5379, + "step": 17392 + }, + { + "epoch": 0.52, + "learning_rate": 9.908880513963791e-05, + "loss": 0.4784, + "step": 17393 + }, + { + "epoch": 0.52, + "learning_rate": 9.907916313710135e-05, + "loss": 0.5869, + "step": 17394 + }, + { + "epoch": 0.52, + "learning_rate": 9.906952114312637e-05, + "loss": 0.5637, + "step": 17395 + }, + { + "epoch": 0.52, + "learning_rate": 9.905987915780264e-05, + "loss": 0.5272, + "step": 17396 + }, + { + "epoch": 0.52, + "learning_rate": 9.905023718121973e-05, + "loss": 0.558, + "step": 17397 + }, + { + "epoch": 0.52, + "learning_rate": 9.904059521346735e-05, + "loss": 0.6105, + "step": 17398 + }, + { + "epoch": 0.52, + "learning_rate": 9.903095325463512e-05, + "loss": 0.5369, + "step": 17399 + }, + { + "epoch": 0.52, + "learning_rate": 9.902131130481275e-05, + "loss": 0.5403, + "step": 17400 + }, + { + "epoch": 0.52, + "learning_rate": 9.901166936408981e-05, + "loss": 0.5166, + "step": 17401 + }, + { + "epoch": 0.52, + "learning_rate": 9.900202743255594e-05, + "loss": 0.5898, + "step": 17402 + }, + { + "epoch": 0.52, + "learning_rate": 9.899238551030085e-05, + "loss": 0.5524, + "step": 17403 + }, + { + "epoch": 0.52, + "learning_rate": 9.89827435974142e-05, + "loss": 0.595, + "step": 17404 + }, + { + "epoch": 0.52, + "learning_rate": 9.897310169398555e-05, + "loss": 0.5745, + "step": 17405 + }, + { + "epoch": 0.52, + "learning_rate": 9.896345980010457e-05, + "loss": 0.5217, + "step": 17406 + }, + { + "epoch": 0.52, + "learning_rate": 9.895381791586098e-05, + "loss": 0.583, + "step": 17407 + }, + { + "epoch": 0.52, + "learning_rate": 9.894417604134436e-05, + "loss": 0.5599, + "step": 17408 + }, + { + "epoch": 0.52, + "learning_rate": 9.893453417664434e-05, + "loss": 0.4872, + "step": 17409 + }, + { + "epoch": 0.52, + "learning_rate": 9.892489232185061e-05, + "loss": 0.5196, + "step": 17410 + }, + { + "epoch": 0.52, + "learning_rate": 9.891525047705279e-05, + "loss": 0.6521, + "step": 17411 + }, + { + "epoch": 0.52, + "learning_rate": 9.890560864234059e-05, + "loss": 0.4624, + "step": 17412 + }, + { + "epoch": 0.52, + "learning_rate": 9.889596681780356e-05, + "loss": 0.5702, + "step": 17413 + }, + { + "epoch": 0.52, + "learning_rate": 9.888632500353141e-05, + "loss": 0.5536, + "step": 17414 + }, + { + "epoch": 0.52, + "learning_rate": 9.887668319961378e-05, + "loss": 0.5189, + "step": 17415 + }, + { + "epoch": 0.52, + "learning_rate": 9.886704140614026e-05, + "loss": 0.5629, + "step": 17416 + }, + { + "epoch": 0.52, + "learning_rate": 9.885739962320056e-05, + "loss": 0.4707, + "step": 17417 + }, + { + "epoch": 0.52, + "learning_rate": 9.884775785088429e-05, + "loss": 0.4353, + "step": 17418 + }, + { + "epoch": 0.52, + "learning_rate": 9.883811608928112e-05, + "loss": 0.559, + "step": 17419 + }, + { + "epoch": 0.52, + "learning_rate": 9.88284743384807e-05, + "loss": 0.5905, + "step": 17420 + }, + { + "epoch": 0.52, + "learning_rate": 9.881883259857262e-05, + "loss": 0.4726, + "step": 17421 + }, + { + "epoch": 0.52, + "learning_rate": 9.880919086964658e-05, + "loss": 0.4829, + "step": 17422 + }, + { + "epoch": 0.52, + "learning_rate": 9.879954915179224e-05, + "loss": 0.6011, + "step": 17423 + }, + { + "epoch": 0.52, + "learning_rate": 9.878990744509918e-05, + "loss": 0.6024, + "step": 17424 + }, + { + "epoch": 0.52, + "learning_rate": 9.878026574965708e-05, + "loss": 0.5946, + "step": 17425 + }, + { + "epoch": 0.52, + "learning_rate": 9.877062406555558e-05, + "loss": 0.6057, + "step": 17426 + }, + { + "epoch": 0.52, + "learning_rate": 9.876098239288436e-05, + "loss": 0.5656, + "step": 17427 + }, + { + "epoch": 0.52, + "learning_rate": 9.8751340731733e-05, + "loss": 0.7204, + "step": 17428 + }, + { + "epoch": 0.52, + "learning_rate": 9.874169908219118e-05, + "loss": 0.6512, + "step": 17429 + }, + { + "epoch": 0.52, + "learning_rate": 9.873205744434852e-05, + "loss": 0.612, + "step": 17430 + }, + { + "epoch": 0.52, + "learning_rate": 9.872241581829474e-05, + "loss": 0.4889, + "step": 17431 + }, + { + "epoch": 0.52, + "learning_rate": 9.871277420411938e-05, + "loss": 0.4803, + "step": 17432 + }, + { + "epoch": 0.52, + "learning_rate": 9.870313260191214e-05, + "loss": 0.6589, + "step": 17433 + }, + { + "epoch": 0.52, + "learning_rate": 9.869349101176266e-05, + "loss": 0.4303, + "step": 17434 + }, + { + "epoch": 0.52, + "learning_rate": 9.868384943376061e-05, + "loss": 0.548, + "step": 17435 + }, + { + "epoch": 0.52, + "learning_rate": 9.867420786799559e-05, + "loss": 0.4251, + "step": 17436 + }, + { + "epoch": 0.52, + "learning_rate": 9.866456631455722e-05, + "loss": 0.5727, + "step": 17437 + }, + { + "epoch": 0.52, + "learning_rate": 9.865492477353521e-05, + "loss": 0.5042, + "step": 17438 + }, + { + "epoch": 0.52, + "learning_rate": 9.86452832450192e-05, + "loss": 0.5234, + "step": 17439 + }, + { + "epoch": 0.52, + "learning_rate": 9.863564172909875e-05, + "loss": 0.607, + "step": 17440 + }, + { + "epoch": 0.52, + "learning_rate": 9.862600022586359e-05, + "loss": 0.651, + "step": 17441 + }, + { + "epoch": 0.52, + "learning_rate": 9.86163587354033e-05, + "loss": 0.5083, + "step": 17442 + }, + { + "epoch": 0.52, + "learning_rate": 9.860671725780762e-05, + "loss": 0.6472, + "step": 17443 + }, + { + "epoch": 0.52, + "learning_rate": 9.859707579316609e-05, + "loss": 0.5733, + "step": 17444 + }, + { + "epoch": 0.52, + "learning_rate": 9.858743434156836e-05, + "loss": 0.4571, + "step": 17445 + }, + { + "epoch": 0.52, + "learning_rate": 9.857779290310414e-05, + "loss": 0.5863, + "step": 17446 + }, + { + "epoch": 0.52, + "learning_rate": 9.856815147786304e-05, + "loss": 0.4731, + "step": 17447 + }, + { + "epoch": 0.52, + "learning_rate": 9.85585100659347e-05, + "loss": 0.5821, + "step": 17448 + }, + { + "epoch": 0.52, + "learning_rate": 9.854886866740871e-05, + "loss": 0.5415, + "step": 17449 + }, + { + "epoch": 0.52, + "learning_rate": 9.853922728237482e-05, + "loss": 0.4007, + "step": 17450 + }, + { + "epoch": 0.52, + "learning_rate": 9.85295859109226e-05, + "loss": 0.4371, + "step": 17451 + }, + { + "epoch": 0.52, + "learning_rate": 9.851994455314166e-05, + "loss": 0.6082, + "step": 17452 + }, + { + "epoch": 0.52, + "learning_rate": 9.851030320912173e-05, + "loss": 0.533, + "step": 17453 + }, + { + "epoch": 0.52, + "learning_rate": 9.85006618789524e-05, + "loss": 0.6577, + "step": 17454 + }, + { + "epoch": 0.52, + "learning_rate": 9.849102056272332e-05, + "loss": 0.5214, + "step": 17455 + }, + { + "epoch": 0.52, + "learning_rate": 9.84813792605241e-05, + "loss": 0.5872, + "step": 17456 + }, + { + "epoch": 0.52, + "learning_rate": 9.847173797244444e-05, + "loss": 0.5528, + "step": 17457 + }, + { + "epoch": 0.52, + "learning_rate": 9.846209669857396e-05, + "loss": 0.6115, + "step": 17458 + }, + { + "epoch": 0.52, + "learning_rate": 9.845245543900225e-05, + "loss": 0.5602, + "step": 17459 + }, + { + "epoch": 0.52, + "learning_rate": 9.844281419381902e-05, + "loss": 0.5391, + "step": 17460 + }, + { + "epoch": 0.52, + "learning_rate": 9.843317296311387e-05, + "loss": 0.5866, + "step": 17461 + }, + { + "epoch": 0.52, + "learning_rate": 9.842353174697648e-05, + "loss": 0.6152, + "step": 17462 + }, + { + "epoch": 0.52, + "learning_rate": 9.841389054549644e-05, + "loss": 0.572, + "step": 17463 + }, + { + "epoch": 0.52, + "learning_rate": 9.840424935876342e-05, + "loss": 0.5082, + "step": 17464 + }, + { + "epoch": 0.52, + "learning_rate": 9.839460818686705e-05, + "loss": 0.6389, + "step": 17465 + }, + { + "epoch": 0.52, + "learning_rate": 9.8384967029897e-05, + "loss": 0.544, + "step": 17466 + }, + { + "epoch": 0.52, + "learning_rate": 9.837532588794286e-05, + "loss": 0.5195, + "step": 17467 + }, + { + "epoch": 0.52, + "learning_rate": 9.836568476109427e-05, + "loss": 0.5998, + "step": 17468 + }, + { + "epoch": 0.52, + "learning_rate": 9.83560436494409e-05, + "loss": 0.5338, + "step": 17469 + }, + { + "epoch": 0.52, + "learning_rate": 9.834640255307244e-05, + "loss": 0.4935, + "step": 17470 + }, + { + "epoch": 0.52, + "learning_rate": 9.83367614720784e-05, + "loss": 0.5148, + "step": 17471 + }, + { + "epoch": 0.52, + "learning_rate": 9.832712040654851e-05, + "loss": 0.5167, + "step": 17472 + }, + { + "epoch": 0.52, + "learning_rate": 9.831747935657237e-05, + "loss": 0.4245, + "step": 17473 + }, + { + "epoch": 0.52, + "learning_rate": 9.830783832223969e-05, + "loss": 0.5524, + "step": 17474 + }, + { + "epoch": 0.52, + "learning_rate": 9.829819730364003e-05, + "loss": 0.5141, + "step": 17475 + }, + { + "epoch": 0.52, + "learning_rate": 9.828855630086302e-05, + "loss": 0.5875, + "step": 17476 + }, + { + "epoch": 0.52, + "learning_rate": 9.827891531399835e-05, + "loss": 0.7096, + "step": 17477 + }, + { + "epoch": 0.52, + "learning_rate": 9.826927434313567e-05, + "loss": 0.5772, + "step": 17478 + }, + { + "epoch": 0.52, + "learning_rate": 9.825963338836457e-05, + "loss": 0.5115, + "step": 17479 + }, + { + "epoch": 0.52, + "learning_rate": 9.824999244977469e-05, + "loss": 0.4352, + "step": 17480 + }, + { + "epoch": 0.52, + "learning_rate": 9.824035152745571e-05, + "loss": 0.5628, + "step": 17481 + }, + { + "epoch": 0.52, + "learning_rate": 9.823071062149722e-05, + "loss": 0.6077, + "step": 17482 + }, + { + "epoch": 0.52, + "learning_rate": 9.822106973198885e-05, + "loss": 0.4782, + "step": 17483 + }, + { + "epoch": 0.52, + "learning_rate": 9.82114288590203e-05, + "loss": 0.4664, + "step": 17484 + }, + { + "epoch": 0.52, + "learning_rate": 9.820178800268119e-05, + "loss": 0.5801, + "step": 17485 + }, + { + "epoch": 0.52, + "learning_rate": 9.81921471630611e-05, + "loss": 0.5973, + "step": 17486 + }, + { + "epoch": 0.52, + "learning_rate": 9.818250634024972e-05, + "loss": 0.5793, + "step": 17487 + }, + { + "epoch": 0.52, + "learning_rate": 9.817286553433667e-05, + "loss": 0.4655, + "step": 17488 + }, + { + "epoch": 0.52, + "learning_rate": 9.816322474541161e-05, + "loss": 0.6369, + "step": 17489 + }, + { + "epoch": 0.52, + "learning_rate": 9.815358397356412e-05, + "loss": 0.4263, + "step": 17490 + }, + { + "epoch": 0.52, + "learning_rate": 9.81439432188839e-05, + "loss": 0.5034, + "step": 17491 + }, + { + "epoch": 0.52, + "learning_rate": 9.81343024814605e-05, + "loss": 0.5508, + "step": 17492 + }, + { + "epoch": 0.52, + "learning_rate": 9.812466176138369e-05, + "loss": 0.4754, + "step": 17493 + }, + { + "epoch": 0.52, + "learning_rate": 9.811502105874298e-05, + "loss": 0.531, + "step": 17494 + }, + { + "epoch": 0.52, + "learning_rate": 9.810538037362805e-05, + "loss": 0.5967, + "step": 17495 + }, + { + "epoch": 0.52, + "learning_rate": 9.809573970612856e-05, + "loss": 0.626, + "step": 17496 + }, + { + "epoch": 0.52, + "learning_rate": 9.808609905633412e-05, + "loss": 0.6267, + "step": 17497 + }, + { + "epoch": 0.52, + "learning_rate": 9.807645842433437e-05, + "loss": 0.5087, + "step": 17498 + }, + { + "epoch": 0.52, + "learning_rate": 9.806681781021891e-05, + "loss": 0.4822, + "step": 17499 + }, + { + "epoch": 0.52, + "learning_rate": 9.805717721407745e-05, + "loss": 0.5064, + "step": 17500 + }, + { + "epoch": 0.52, + "learning_rate": 9.804753663599959e-05, + "loss": 0.6296, + "step": 17501 + }, + { + "epoch": 0.52, + "learning_rate": 9.80378960760749e-05, + "loss": 0.5531, + "step": 17502 + }, + { + "epoch": 0.52, + "learning_rate": 9.80282555343931e-05, + "loss": 0.5178, + "step": 17503 + }, + { + "epoch": 0.52, + "learning_rate": 9.801861501104378e-05, + "loss": 0.4281, + "step": 17504 + }, + { + "epoch": 0.52, + "learning_rate": 9.800897450611662e-05, + "loss": 0.4966, + "step": 17505 + }, + { + "epoch": 0.52, + "learning_rate": 9.79993340197012e-05, + "loss": 0.5612, + "step": 17506 + }, + { + "epoch": 0.52, + "learning_rate": 9.798969355188716e-05, + "loss": 0.5793, + "step": 17507 + }, + { + "epoch": 0.52, + "learning_rate": 9.798005310276418e-05, + "loss": 0.5002, + "step": 17508 + }, + { + "epoch": 0.52, + "learning_rate": 9.797041267242187e-05, + "loss": 0.501, + "step": 17509 + }, + { + "epoch": 0.52, + "learning_rate": 9.796077226094983e-05, + "loss": 0.5513, + "step": 17510 + }, + { + "epoch": 0.52, + "learning_rate": 9.79511318684377e-05, + "loss": 0.5682, + "step": 17511 + }, + { + "epoch": 0.52, + "learning_rate": 9.794149149497516e-05, + "loss": 0.596, + "step": 17512 + }, + { + "epoch": 0.52, + "learning_rate": 9.793185114065182e-05, + "loss": 0.4921, + "step": 17513 + }, + { + "epoch": 0.52, + "learning_rate": 9.792221080555726e-05, + "loss": 0.627, + "step": 17514 + }, + { + "epoch": 0.52, + "learning_rate": 9.791257048978118e-05, + "loss": 0.4412, + "step": 17515 + }, + { + "epoch": 0.52, + "learning_rate": 9.790293019341321e-05, + "loss": 0.5507, + "step": 17516 + }, + { + "epoch": 0.52, + "learning_rate": 9.789328991654291e-05, + "loss": 0.6003, + "step": 17517 + }, + { + "epoch": 0.52, + "learning_rate": 9.788364965926e-05, + "loss": 0.4846, + "step": 17518 + }, + { + "epoch": 0.52, + "learning_rate": 9.787400942165405e-05, + "loss": 0.5275, + "step": 17519 + }, + { + "epoch": 0.52, + "learning_rate": 9.786436920381475e-05, + "loss": 0.5191, + "step": 17520 + }, + { + "epoch": 0.52, + "learning_rate": 9.785472900583166e-05, + "loss": 0.6337, + "step": 17521 + }, + { + "epoch": 0.52, + "learning_rate": 9.784508882779445e-05, + "loss": 0.5021, + "step": 17522 + }, + { + "epoch": 0.52, + "learning_rate": 9.783544866979271e-05, + "loss": 0.5346, + "step": 17523 + }, + { + "epoch": 0.52, + "learning_rate": 9.782580853191618e-05, + "loss": 0.5211, + "step": 17524 + }, + { + "epoch": 0.52, + "learning_rate": 9.781616841425437e-05, + "loss": 0.5263, + "step": 17525 + }, + { + "epoch": 0.52, + "learning_rate": 9.780652831689695e-05, + "loss": 0.4559, + "step": 17526 + }, + { + "epoch": 0.52, + "learning_rate": 9.779688823993357e-05, + "loss": 0.5148, + "step": 17527 + }, + { + "epoch": 0.52, + "learning_rate": 9.778724818345387e-05, + "loss": 0.6019, + "step": 17528 + }, + { + "epoch": 0.52, + "learning_rate": 9.777760814754743e-05, + "loss": 0.5497, + "step": 17529 + }, + { + "epoch": 0.52, + "learning_rate": 9.776796813230387e-05, + "loss": 0.5658, + "step": 17530 + }, + { + "epoch": 0.52, + "learning_rate": 9.77583281378129e-05, + "loss": 0.5195, + "step": 17531 + }, + { + "epoch": 0.52, + "learning_rate": 9.774868816416411e-05, + "loss": 0.5413, + "step": 17532 + }, + { + "epoch": 0.52, + "learning_rate": 9.773904821144708e-05, + "loss": 0.6052, + "step": 17533 + }, + { + "epoch": 0.52, + "learning_rate": 9.772940827975149e-05, + "loss": 0.4408, + "step": 17534 + }, + { + "epoch": 0.52, + "learning_rate": 9.771976836916695e-05, + "loss": 0.437, + "step": 17535 + }, + { + "epoch": 0.52, + "learning_rate": 9.771012847978313e-05, + "loss": 0.5171, + "step": 17536 + }, + { + "epoch": 0.52, + "learning_rate": 9.770048861168959e-05, + "loss": 0.5351, + "step": 17537 + }, + { + "epoch": 0.52, + "learning_rate": 9.769084876497598e-05, + "loss": 0.5333, + "step": 17538 + }, + { + "epoch": 0.52, + "learning_rate": 9.768120893973196e-05, + "loss": 0.5411, + "step": 17539 + }, + { + "epoch": 0.52, + "learning_rate": 9.767156913604716e-05, + "loss": 0.4118, + "step": 17540 + }, + { + "epoch": 0.52, + "learning_rate": 9.766192935401116e-05, + "loss": 0.6161, + "step": 17541 + }, + { + "epoch": 0.52, + "learning_rate": 9.765228959371359e-05, + "loss": 0.5287, + "step": 17542 + }, + { + "epoch": 0.52, + "learning_rate": 9.76426498552441e-05, + "loss": 0.5644, + "step": 17543 + }, + { + "epoch": 0.52, + "learning_rate": 9.763301013869235e-05, + "loss": 0.6166, + "step": 17544 + }, + { + "epoch": 0.52, + "learning_rate": 9.762337044414787e-05, + "loss": 0.5291, + "step": 17545 + }, + { + "epoch": 0.52, + "learning_rate": 9.761373077170038e-05, + "loss": 0.4811, + "step": 17546 + }, + { + "epoch": 0.52, + "learning_rate": 9.760409112143949e-05, + "loss": 0.4786, + "step": 17547 + }, + { + "epoch": 0.52, + "learning_rate": 9.759445149345477e-05, + "loss": 0.6135, + "step": 17548 + }, + { + "epoch": 0.52, + "learning_rate": 9.75848118878359e-05, + "loss": 0.4766, + "step": 17549 + }, + { + "epoch": 0.52, + "learning_rate": 9.757517230467247e-05, + "loss": 0.6027, + "step": 17550 + }, + { + "epoch": 0.52, + "learning_rate": 9.756553274405416e-05, + "loss": 0.4679, + "step": 17551 + }, + { + "epoch": 0.52, + "learning_rate": 9.755589320607049e-05, + "loss": 0.4732, + "step": 17552 + }, + { + "epoch": 0.52, + "learning_rate": 9.75462536908112e-05, + "loss": 0.5408, + "step": 17553 + }, + { + "epoch": 0.52, + "learning_rate": 9.753661419836583e-05, + "loss": 0.6857, + "step": 17554 + }, + { + "epoch": 0.52, + "learning_rate": 9.752697472882408e-05, + "loss": 0.6129, + "step": 17555 + }, + { + "epoch": 0.52, + "learning_rate": 9.751733528227553e-05, + "loss": 0.5264, + "step": 17556 + }, + { + "epoch": 0.52, + "learning_rate": 9.750769585880976e-05, + "loss": 0.5993, + "step": 17557 + }, + { + "epoch": 0.52, + "learning_rate": 9.749805645851647e-05, + "loss": 0.5405, + "step": 17558 + }, + { + "epoch": 0.52, + "learning_rate": 9.748841708148529e-05, + "loss": 0.5812, + "step": 17559 + }, + { + "epoch": 0.52, + "learning_rate": 9.747877772780577e-05, + "loss": 0.501, + "step": 17560 + }, + { + "epoch": 0.52, + "learning_rate": 9.746913839756755e-05, + "loss": 0.678, + "step": 17561 + }, + { + "epoch": 0.52, + "learning_rate": 9.745949909086029e-05, + "loss": 0.5034, + "step": 17562 + }, + { + "epoch": 0.52, + "learning_rate": 9.744985980777364e-05, + "loss": 0.5225, + "step": 17563 + }, + { + "epoch": 0.52, + "learning_rate": 9.74402205483971e-05, + "loss": 0.5634, + "step": 17564 + }, + { + "epoch": 0.52, + "learning_rate": 9.743058131282042e-05, + "loss": 0.5904, + "step": 17565 + }, + { + "epoch": 0.52, + "learning_rate": 9.742094210113313e-05, + "loss": 0.5011, + "step": 17566 + }, + { + "epoch": 0.52, + "learning_rate": 9.741130291342495e-05, + "loss": 0.6109, + "step": 17567 + }, + { + "epoch": 0.52, + "learning_rate": 9.740166374978542e-05, + "loss": 0.4924, + "step": 17568 + }, + { + "epoch": 0.52, + "learning_rate": 9.739202461030416e-05, + "loss": 0.5885, + "step": 17569 + }, + { + "epoch": 0.52, + "learning_rate": 9.738238549507083e-05, + "loss": 0.479, + "step": 17570 + }, + { + "epoch": 0.52, + "learning_rate": 9.737274640417507e-05, + "loss": 0.4749, + "step": 17571 + }, + { + "epoch": 0.52, + "learning_rate": 9.736310733770644e-05, + "loss": 0.4798, + "step": 17572 + }, + { + "epoch": 0.52, + "learning_rate": 9.735346829575455e-05, + "loss": 0.4637, + "step": 17573 + }, + { + "epoch": 0.52, + "learning_rate": 9.73438292784091e-05, + "loss": 0.4942, + "step": 17574 + }, + { + "epoch": 0.52, + "learning_rate": 9.733419028575968e-05, + "loss": 0.5533, + "step": 17575 + }, + { + "epoch": 0.52, + "learning_rate": 9.732455131789587e-05, + "loss": 0.6078, + "step": 17576 + }, + { + "epoch": 0.52, + "learning_rate": 9.731491237490731e-05, + "loss": 0.6071, + "step": 17577 + }, + { + "epoch": 0.52, + "learning_rate": 9.730527345688366e-05, + "loss": 0.504, + "step": 17578 + }, + { + "epoch": 0.52, + "learning_rate": 9.729563456391446e-05, + "loss": 0.4936, + "step": 17579 + }, + { + "epoch": 0.52, + "learning_rate": 9.72859956960894e-05, + "loss": 0.5638, + "step": 17580 + }, + { + "epoch": 0.52, + "learning_rate": 9.727635685349805e-05, + "loss": 0.4721, + "step": 17581 + }, + { + "epoch": 0.52, + "learning_rate": 9.726671803623008e-05, + "loss": 0.4811, + "step": 17582 + }, + { + "epoch": 0.52, + "learning_rate": 9.725707924437504e-05, + "loss": 0.562, + "step": 17583 + }, + { + "epoch": 0.52, + "learning_rate": 9.72474404780226e-05, + "loss": 0.4824, + "step": 17584 + }, + { + "epoch": 0.52, + "learning_rate": 9.723780173726233e-05, + "loss": 0.5025, + "step": 17585 + }, + { + "epoch": 0.52, + "learning_rate": 9.722816302218394e-05, + "loss": 0.5287, + "step": 17586 + }, + { + "epoch": 0.52, + "learning_rate": 9.721852433287697e-05, + "loss": 0.5265, + "step": 17587 + }, + { + "epoch": 0.52, + "learning_rate": 9.7208885669431e-05, + "loss": 0.503, + "step": 17588 + }, + { + "epoch": 0.52, + "learning_rate": 9.719924703193575e-05, + "loss": 0.4749, + "step": 17589 + }, + { + "epoch": 0.52, + "learning_rate": 9.718960842048078e-05, + "loss": 0.5195, + "step": 17590 + }, + { + "epoch": 0.52, + "learning_rate": 9.71799698351557e-05, + "loss": 0.5406, + "step": 17591 + }, + { + "epoch": 0.52, + "learning_rate": 9.717033127605014e-05, + "loss": 0.4833, + "step": 17592 + }, + { + "epoch": 0.52, + "learning_rate": 9.71606927432537e-05, + "loss": 0.5642, + "step": 17593 + }, + { + "epoch": 0.52, + "learning_rate": 9.715105423685605e-05, + "loss": 0.6576, + "step": 17594 + }, + { + "epoch": 0.52, + "learning_rate": 9.71414157569467e-05, + "loss": 0.6046, + "step": 17595 + }, + { + "epoch": 0.52, + "learning_rate": 9.713177730361536e-05, + "loss": 0.4859, + "step": 17596 + }, + { + "epoch": 0.52, + "learning_rate": 9.712213887695159e-05, + "loss": 0.6354, + "step": 17597 + }, + { + "epoch": 0.52, + "learning_rate": 9.711250047704505e-05, + "loss": 0.5506, + "step": 17598 + }, + { + "epoch": 0.52, + "learning_rate": 9.710286210398533e-05, + "loss": 0.4606, + "step": 17599 + }, + { + "epoch": 0.52, + "learning_rate": 9.709322375786201e-05, + "loss": 0.6291, + "step": 17600 + }, + { + "epoch": 0.52, + "learning_rate": 9.708358543876477e-05, + "loss": 0.5339, + "step": 17601 + }, + { + "epoch": 0.52, + "learning_rate": 9.70739471467832e-05, + "loss": 0.6273, + "step": 17602 + }, + { + "epoch": 0.52, + "learning_rate": 9.706430888200689e-05, + "loss": 0.5152, + "step": 17603 + }, + { + "epoch": 0.52, + "learning_rate": 9.705467064452545e-05, + "loss": 0.5629, + "step": 17604 + }, + { + "epoch": 0.52, + "learning_rate": 9.704503243442853e-05, + "loss": 0.5079, + "step": 17605 + }, + { + "epoch": 0.52, + "learning_rate": 9.703539425180573e-05, + "loss": 0.6206, + "step": 17606 + }, + { + "epoch": 0.52, + "learning_rate": 9.702575609674661e-05, + "loss": 0.5052, + "step": 17607 + }, + { + "epoch": 0.52, + "learning_rate": 9.701611796934086e-05, + "loss": 0.5564, + "step": 17608 + }, + { + "epoch": 0.52, + "learning_rate": 9.700647986967801e-05, + "loss": 0.5355, + "step": 17609 + }, + { + "epoch": 0.52, + "learning_rate": 9.699684179784778e-05, + "loss": 0.5748, + "step": 17610 + }, + { + "epoch": 0.52, + "learning_rate": 9.698720375393969e-05, + "loss": 0.468, + "step": 17611 + }, + { + "epoch": 0.52, + "learning_rate": 9.697756573804336e-05, + "loss": 0.536, + "step": 17612 + }, + { + "epoch": 0.52, + "learning_rate": 9.696792775024849e-05, + "loss": 0.5233, + "step": 17613 + }, + { + "epoch": 0.52, + "learning_rate": 9.695828979064454e-05, + "loss": 0.4903, + "step": 17614 + }, + { + "epoch": 0.52, + "learning_rate": 9.694865185932122e-05, + "loss": 0.6173, + "step": 17615 + }, + { + "epoch": 0.52, + "learning_rate": 9.693901395636811e-05, + "loss": 0.438, + "step": 17616 + }, + { + "epoch": 0.52, + "learning_rate": 9.692937608187486e-05, + "loss": 0.5943, + "step": 17617 + }, + { + "epoch": 0.52, + "learning_rate": 9.691973823593104e-05, + "loss": 0.4875, + "step": 17618 + }, + { + "epoch": 0.52, + "learning_rate": 9.691010041862623e-05, + "loss": 0.5804, + "step": 17619 + }, + { + "epoch": 0.52, + "learning_rate": 9.690046263005011e-05, + "loss": 0.5373, + "step": 17620 + }, + { + "epoch": 0.52, + "learning_rate": 9.689082487029227e-05, + "loss": 0.4545, + "step": 17621 + }, + { + "epoch": 0.52, + "learning_rate": 9.688118713944229e-05, + "loss": 0.6493, + "step": 17622 + }, + { + "epoch": 0.52, + "learning_rate": 9.687154943758975e-05, + "loss": 0.5452, + "step": 17623 + }, + { + "epoch": 0.52, + "learning_rate": 9.686191176482432e-05, + "loss": 0.5696, + "step": 17624 + }, + { + "epoch": 0.52, + "learning_rate": 9.685227412123561e-05, + "loss": 0.6095, + "step": 17625 + }, + { + "epoch": 0.52, + "learning_rate": 9.684263650691315e-05, + "loss": 0.5821, + "step": 17626 + }, + { + "epoch": 0.52, + "learning_rate": 9.683299892194662e-05, + "loss": 0.6732, + "step": 17627 + }, + { + "epoch": 0.52, + "learning_rate": 9.68233613664256e-05, + "loss": 0.4795, + "step": 17628 + }, + { + "epoch": 0.52, + "learning_rate": 9.681372384043973e-05, + "loss": 0.5129, + "step": 17629 + }, + { + "epoch": 0.52, + "learning_rate": 9.680408634407856e-05, + "loss": 0.5698, + "step": 17630 + }, + { + "epoch": 0.52, + "learning_rate": 9.67944488774317e-05, + "loss": 0.4875, + "step": 17631 + }, + { + "epoch": 0.52, + "learning_rate": 9.678481144058881e-05, + "loss": 0.4223, + "step": 17632 + }, + { + "epoch": 0.52, + "learning_rate": 9.677517403363948e-05, + "loss": 0.5025, + "step": 17633 + }, + { + "epoch": 0.52, + "learning_rate": 9.676553665667326e-05, + "loss": 0.6147, + "step": 17634 + }, + { + "epoch": 0.53, + "learning_rate": 9.675589930977978e-05, + "loss": 0.4982, + "step": 17635 + }, + { + "epoch": 0.53, + "learning_rate": 9.674626199304869e-05, + "loss": 0.55, + "step": 17636 + }, + { + "epoch": 0.53, + "learning_rate": 9.673662470656956e-05, + "loss": 0.5078, + "step": 17637 + }, + { + "epoch": 0.53, + "learning_rate": 9.672698745043195e-05, + "loss": 0.6865, + "step": 17638 + }, + { + "epoch": 0.53, + "learning_rate": 9.671735022472554e-05, + "loss": 0.5148, + "step": 17639 + }, + { + "epoch": 0.53, + "learning_rate": 9.670771302953987e-05, + "loss": 0.5333, + "step": 17640 + }, + { + "epoch": 0.53, + "learning_rate": 9.669807586496462e-05, + "loss": 0.5147, + "step": 17641 + }, + { + "epoch": 0.53, + "learning_rate": 9.668843873108932e-05, + "loss": 0.4885, + "step": 17642 + }, + { + "epoch": 0.53, + "learning_rate": 9.667880162800357e-05, + "loss": 0.4994, + "step": 17643 + }, + { + "epoch": 0.53, + "learning_rate": 9.666916455579708e-05, + "loss": 0.4456, + "step": 17644 + }, + { + "epoch": 0.53, + "learning_rate": 9.66595275145593e-05, + "loss": 0.5829, + "step": 17645 + }, + { + "epoch": 0.53, + "learning_rate": 9.664989050437991e-05, + "loss": 0.5131, + "step": 17646 + }, + { + "epoch": 0.53, + "learning_rate": 9.664025352534848e-05, + "loss": 0.6258, + "step": 17647 + }, + { + "epoch": 0.53, + "learning_rate": 9.66306165775547e-05, + "loss": 0.5385, + "step": 17648 + }, + { + "epoch": 0.53, + "learning_rate": 9.662097966108807e-05, + "loss": 0.5333, + "step": 17649 + }, + { + "epoch": 0.53, + "learning_rate": 9.66113427760382e-05, + "loss": 0.6337, + "step": 17650 + }, + { + "epoch": 0.53, + "learning_rate": 9.660170592249474e-05, + "loss": 0.6324, + "step": 17651 + }, + { + "epoch": 0.53, + "learning_rate": 9.659206910054729e-05, + "loss": 0.517, + "step": 17652 + }, + { + "epoch": 0.53, + "learning_rate": 9.658243231028537e-05, + "loss": 0.6256, + "step": 17653 + }, + { + "epoch": 0.53, + "learning_rate": 9.657279555179866e-05, + "loss": 0.5016, + "step": 17654 + }, + { + "epoch": 0.53, + "learning_rate": 9.656315882517673e-05, + "loss": 0.5647, + "step": 17655 + }, + { + "epoch": 0.53, + "learning_rate": 9.655352213050919e-05, + "loss": 0.648, + "step": 17656 + }, + { + "epoch": 0.53, + "learning_rate": 9.654388546788559e-05, + "loss": 0.4514, + "step": 17657 + }, + { + "epoch": 0.53, + "learning_rate": 9.65342488373956e-05, + "loss": 0.5315, + "step": 17658 + }, + { + "epoch": 0.53, + "learning_rate": 9.652461223912875e-05, + "loss": 0.5424, + "step": 17659 + }, + { + "epoch": 0.53, + "learning_rate": 9.651497567317472e-05, + "loss": 0.4372, + "step": 17660 + }, + { + "epoch": 0.53, + "learning_rate": 9.650533913962303e-05, + "loss": 0.5882, + "step": 17661 + }, + { + "epoch": 0.53, + "learning_rate": 9.649570263856328e-05, + "loss": 0.5084, + "step": 17662 + }, + { + "epoch": 0.53, + "learning_rate": 9.648606617008512e-05, + "loss": 0.5625, + "step": 17663 + }, + { + "epoch": 0.53, + "learning_rate": 9.647642973427813e-05, + "loss": 0.4618, + "step": 17664 + }, + { + "epoch": 0.53, + "learning_rate": 9.646679333123188e-05, + "loss": 0.5633, + "step": 17665 + }, + { + "epoch": 0.53, + "learning_rate": 9.645715696103594e-05, + "loss": 0.5264, + "step": 17666 + }, + { + "epoch": 0.53, + "learning_rate": 9.644752062377997e-05, + "loss": 0.4613, + "step": 17667 + }, + { + "epoch": 0.53, + "learning_rate": 9.643788431955358e-05, + "loss": 0.5628, + "step": 17668 + }, + { + "epoch": 0.53, + "learning_rate": 9.642824804844626e-05, + "loss": 0.5184, + "step": 17669 + }, + { + "epoch": 0.53, + "learning_rate": 9.641861181054768e-05, + "loss": 0.5416, + "step": 17670 + }, + { + "epoch": 0.53, + "learning_rate": 9.64089756059474e-05, + "loss": 0.4532, + "step": 17671 + }, + { + "epoch": 0.53, + "learning_rate": 9.639933943473509e-05, + "loss": 0.5456, + "step": 17672 + }, + { + "epoch": 0.53, + "learning_rate": 9.638970329700027e-05, + "loss": 0.5732, + "step": 17673 + }, + { + "epoch": 0.53, + "learning_rate": 9.638006719283251e-05, + "loss": 0.5364, + "step": 17674 + }, + { + "epoch": 0.53, + "learning_rate": 9.637043112232146e-05, + "loss": 0.4442, + "step": 17675 + }, + { + "epoch": 0.53, + "learning_rate": 9.636079508555674e-05, + "loss": 0.5133, + "step": 17676 + }, + { + "epoch": 0.53, + "learning_rate": 9.635115908262786e-05, + "loss": 0.5204, + "step": 17677 + }, + { + "epoch": 0.53, + "learning_rate": 9.634152311362443e-05, + "loss": 0.4853, + "step": 17678 + }, + { + "epoch": 0.53, + "learning_rate": 9.633188717863611e-05, + "loss": 0.5584, + "step": 17679 + }, + { + "epoch": 0.53, + "learning_rate": 9.632225127775242e-05, + "loss": 0.5721, + "step": 17680 + }, + { + "epoch": 0.53, + "learning_rate": 9.631261541106294e-05, + "loss": 0.5817, + "step": 17681 + }, + { + "epoch": 0.53, + "learning_rate": 9.630297957865731e-05, + "loss": 0.6533, + "step": 17682 + }, + { + "epoch": 0.53, + "learning_rate": 9.629334378062514e-05, + "loss": 0.5937, + "step": 17683 + }, + { + "epoch": 0.53, + "learning_rate": 9.628370801705593e-05, + "loss": 0.6238, + "step": 17684 + }, + { + "epoch": 0.53, + "learning_rate": 9.627407228803934e-05, + "loss": 0.5074, + "step": 17685 + }, + { + "epoch": 0.53, + "learning_rate": 9.626443659366495e-05, + "loss": 0.6226, + "step": 17686 + }, + { + "epoch": 0.53, + "learning_rate": 9.625480093402236e-05, + "loss": 0.4907, + "step": 17687 + }, + { + "epoch": 0.53, + "learning_rate": 9.624516530920109e-05, + "loss": 0.5694, + "step": 17688 + }, + { + "epoch": 0.53, + "learning_rate": 9.623552971929079e-05, + "loss": 0.5221, + "step": 17689 + }, + { + "epoch": 0.53, + "learning_rate": 9.622589416438101e-05, + "loss": 0.5803, + "step": 17690 + }, + { + "epoch": 0.53, + "learning_rate": 9.621625864456143e-05, + "loss": 0.4258, + "step": 17691 + }, + { + "epoch": 0.53, + "learning_rate": 9.620662315992153e-05, + "loss": 0.5423, + "step": 17692 + }, + { + "epoch": 0.53, + "learning_rate": 9.61969877105509e-05, + "loss": 0.5797, + "step": 17693 + }, + { + "epoch": 0.53, + "learning_rate": 9.618735229653922e-05, + "loss": 0.5071, + "step": 17694 + }, + { + "epoch": 0.53, + "learning_rate": 9.617771691797602e-05, + "loss": 0.6114, + "step": 17695 + }, + { + "epoch": 0.53, + "learning_rate": 9.616808157495086e-05, + "loss": 0.4849, + "step": 17696 + }, + { + "epoch": 0.53, + "learning_rate": 9.615844626755334e-05, + "loss": 0.6678, + "step": 17697 + }, + { + "epoch": 0.53, + "learning_rate": 9.614881099587308e-05, + "loss": 0.561, + "step": 17698 + }, + { + "epoch": 0.53, + "learning_rate": 9.613917575999965e-05, + "loss": 0.5439, + "step": 17699 + }, + { + "epoch": 0.53, + "learning_rate": 9.612954056002258e-05, + "loss": 0.5347, + "step": 17700 + }, + { + "epoch": 0.53, + "learning_rate": 9.611990539603152e-05, + "loss": 0.5389, + "step": 17701 + }, + { + "epoch": 0.53, + "learning_rate": 9.611027026811603e-05, + "loss": 0.4653, + "step": 17702 + }, + { + "epoch": 0.53, + "learning_rate": 9.610063517636573e-05, + "loss": 0.585, + "step": 17703 + }, + { + "epoch": 0.53, + "learning_rate": 9.609100012087015e-05, + "loss": 0.5594, + "step": 17704 + }, + { + "epoch": 0.53, + "learning_rate": 9.608136510171885e-05, + "loss": 0.5009, + "step": 17705 + }, + { + "epoch": 0.53, + "learning_rate": 9.607173011900151e-05, + "loss": 0.4655, + "step": 17706 + }, + { + "epoch": 0.53, + "learning_rate": 9.606209517280765e-05, + "loss": 0.5773, + "step": 17707 + }, + { + "epoch": 0.53, + "learning_rate": 9.605246026322686e-05, + "loss": 0.4863, + "step": 17708 + }, + { + "epoch": 0.53, + "learning_rate": 9.604282539034869e-05, + "loss": 0.5294, + "step": 17709 + }, + { + "epoch": 0.53, + "learning_rate": 9.603319055426279e-05, + "loss": 0.5526, + "step": 17710 + }, + { + "epoch": 0.53, + "learning_rate": 9.60235557550587e-05, + "loss": 0.5705, + "step": 17711 + }, + { + "epoch": 0.53, + "learning_rate": 9.601392099282597e-05, + "loss": 0.5297, + "step": 17712 + }, + { + "epoch": 0.53, + "learning_rate": 9.600428626765423e-05, + "loss": 0.5434, + "step": 17713 + }, + { + "epoch": 0.53, + "learning_rate": 9.599465157963308e-05, + "loss": 0.5043, + "step": 17714 + }, + { + "epoch": 0.53, + "learning_rate": 9.598501692885201e-05, + "loss": 0.5766, + "step": 17715 + }, + { + "epoch": 0.53, + "learning_rate": 9.597538231540068e-05, + "loss": 0.5844, + "step": 17716 + }, + { + "epoch": 0.53, + "learning_rate": 9.596574773936863e-05, + "loss": 0.6731, + "step": 17717 + }, + { + "epoch": 0.53, + "learning_rate": 9.595611320084548e-05, + "loss": 0.573, + "step": 17718 + }, + { + "epoch": 0.53, + "learning_rate": 9.594647869992073e-05, + "loss": 0.5459, + "step": 17719 + }, + { + "epoch": 0.53, + "learning_rate": 9.593684423668403e-05, + "loss": 0.5046, + "step": 17720 + }, + { + "epoch": 0.53, + "learning_rate": 9.59272098112249e-05, + "loss": 0.5266, + "step": 17721 + }, + { + "epoch": 0.53, + "learning_rate": 9.591757542363301e-05, + "loss": 0.4898, + "step": 17722 + }, + { + "epoch": 0.53, + "learning_rate": 9.590794107399785e-05, + "loss": 0.5156, + "step": 17723 + }, + { + "epoch": 0.53, + "learning_rate": 9.589830676240901e-05, + "loss": 0.4811, + "step": 17724 + }, + { + "epoch": 0.53, + "learning_rate": 9.588867248895608e-05, + "loss": 0.5256, + "step": 17725 + }, + { + "epoch": 0.53, + "learning_rate": 9.587903825372867e-05, + "loss": 0.5636, + "step": 17726 + }, + { + "epoch": 0.53, + "learning_rate": 9.58694040568163e-05, + "loss": 0.6099, + "step": 17727 + }, + { + "epoch": 0.53, + "learning_rate": 9.585976989830854e-05, + "loss": 0.5896, + "step": 17728 + }, + { + "epoch": 0.53, + "learning_rate": 9.5850135778295e-05, + "loss": 0.6061, + "step": 17729 + }, + { + "epoch": 0.53, + "learning_rate": 9.584050169686528e-05, + "loss": 0.5231, + "step": 17730 + }, + { + "epoch": 0.53, + "learning_rate": 9.583086765410887e-05, + "loss": 0.5433, + "step": 17731 + }, + { + "epoch": 0.53, + "learning_rate": 9.582123365011541e-05, + "loss": 0.5155, + "step": 17732 + }, + { + "epoch": 0.53, + "learning_rate": 9.581159968497443e-05, + "loss": 0.456, + "step": 17733 + }, + { + "epoch": 0.53, + "learning_rate": 9.580196575877558e-05, + "loss": 0.5729, + "step": 17734 + }, + { + "epoch": 0.53, + "learning_rate": 9.579233187160835e-05, + "loss": 0.502, + "step": 17735 + }, + { + "epoch": 0.53, + "learning_rate": 9.578269802356232e-05, + "loss": 0.4725, + "step": 17736 + }, + { + "epoch": 0.53, + "learning_rate": 9.57730642147271e-05, + "loss": 0.4473, + "step": 17737 + }, + { + "epoch": 0.53, + "learning_rate": 9.576343044519228e-05, + "loss": 0.5721, + "step": 17738 + }, + { + "epoch": 0.53, + "learning_rate": 9.575379671504736e-05, + "loss": 0.5329, + "step": 17739 + }, + { + "epoch": 0.53, + "learning_rate": 9.574416302438193e-05, + "loss": 0.6036, + "step": 17740 + }, + { + "epoch": 0.53, + "learning_rate": 9.573452937328559e-05, + "loss": 0.5055, + "step": 17741 + }, + { + "epoch": 0.53, + "learning_rate": 9.572489576184794e-05, + "loss": 0.4764, + "step": 17742 + }, + { + "epoch": 0.53, + "learning_rate": 9.571526219015843e-05, + "loss": 0.5972, + "step": 17743 + }, + { + "epoch": 0.53, + "learning_rate": 9.570562865830676e-05, + "loss": 0.4389, + "step": 17744 + }, + { + "epoch": 0.53, + "learning_rate": 9.569599516638243e-05, + "loss": 0.6055, + "step": 17745 + }, + { + "epoch": 0.53, + "learning_rate": 9.5686361714475e-05, + "loss": 0.5556, + "step": 17746 + }, + { + "epoch": 0.53, + "learning_rate": 9.567672830267408e-05, + "loss": 0.5313, + "step": 17747 + }, + { + "epoch": 0.53, + "learning_rate": 9.566709493106921e-05, + "loss": 0.4944, + "step": 17748 + }, + { + "epoch": 0.53, + "learning_rate": 9.565746159975e-05, + "loss": 0.4668, + "step": 17749 + }, + { + "epoch": 0.53, + "learning_rate": 9.564782830880592e-05, + "loss": 0.551, + "step": 17750 + }, + { + "epoch": 0.53, + "learning_rate": 9.563819505832664e-05, + "loss": 0.578, + "step": 17751 + }, + { + "epoch": 0.53, + "learning_rate": 9.562856184840164e-05, + "loss": 0.5594, + "step": 17752 + }, + { + "epoch": 0.53, + "learning_rate": 9.561892867912057e-05, + "loss": 0.552, + "step": 17753 + }, + { + "epoch": 0.53, + "learning_rate": 9.560929555057296e-05, + "loss": 0.6103, + "step": 17754 + }, + { + "epoch": 0.53, + "learning_rate": 9.559966246284832e-05, + "loss": 0.4663, + "step": 17755 + }, + { + "epoch": 0.53, + "learning_rate": 9.559002941603631e-05, + "loss": 0.5073, + "step": 17756 + }, + { + "epoch": 0.53, + "learning_rate": 9.558039641022647e-05, + "loss": 0.5676, + "step": 17757 + }, + { + "epoch": 0.53, + "learning_rate": 9.55707634455083e-05, + "loss": 0.5814, + "step": 17758 + }, + { + "epoch": 0.53, + "learning_rate": 9.55611305219714e-05, + "loss": 0.493, + "step": 17759 + }, + { + "epoch": 0.53, + "learning_rate": 9.555149763970536e-05, + "loss": 0.6267, + "step": 17760 + }, + { + "epoch": 0.53, + "learning_rate": 9.554186479879974e-05, + "loss": 0.5815, + "step": 17761 + }, + { + "epoch": 0.53, + "learning_rate": 9.553223199934404e-05, + "loss": 0.4703, + "step": 17762 + }, + { + "epoch": 0.53, + "learning_rate": 9.552259924142789e-05, + "loss": 0.4968, + "step": 17763 + }, + { + "epoch": 0.53, + "learning_rate": 9.551296652514081e-05, + "loss": 0.4856, + "step": 17764 + }, + { + "epoch": 0.53, + "learning_rate": 9.550333385057241e-05, + "loss": 0.573, + "step": 17765 + }, + { + "epoch": 0.53, + "learning_rate": 9.549370121781221e-05, + "loss": 0.5903, + "step": 17766 + }, + { + "epoch": 0.53, + "learning_rate": 9.548406862694975e-05, + "loss": 0.5034, + "step": 17767 + }, + { + "epoch": 0.53, + "learning_rate": 9.547443607807465e-05, + "loss": 0.505, + "step": 17768 + }, + { + "epoch": 0.53, + "learning_rate": 9.546480357127646e-05, + "loss": 0.6415, + "step": 17769 + }, + { + "epoch": 0.53, + "learning_rate": 9.54551711066447e-05, + "loss": 0.5338, + "step": 17770 + }, + { + "epoch": 0.53, + "learning_rate": 9.544553868426892e-05, + "loss": 0.6028, + "step": 17771 + }, + { + "epoch": 0.53, + "learning_rate": 9.543590630423873e-05, + "loss": 0.4696, + "step": 17772 + }, + { + "epoch": 0.53, + "learning_rate": 9.542627396664367e-05, + "loss": 0.4513, + "step": 17773 + }, + { + "epoch": 0.53, + "learning_rate": 9.541664167157327e-05, + "loss": 0.545, + "step": 17774 + }, + { + "epoch": 0.53, + "learning_rate": 9.540700941911713e-05, + "loss": 0.6084, + "step": 17775 + }, + { + "epoch": 0.53, + "learning_rate": 9.539737720936479e-05, + "loss": 0.5777, + "step": 17776 + }, + { + "epoch": 0.53, + "learning_rate": 9.538774504240577e-05, + "loss": 0.596, + "step": 17777 + }, + { + "epoch": 0.53, + "learning_rate": 9.537811291832968e-05, + "loss": 0.5398, + "step": 17778 + }, + { + "epoch": 0.53, + "learning_rate": 9.536848083722603e-05, + "loss": 0.5439, + "step": 17779 + }, + { + "epoch": 0.53, + "learning_rate": 9.535884879918448e-05, + "loss": 0.5441, + "step": 17780 + }, + { + "epoch": 0.53, + "learning_rate": 9.534921680429442e-05, + "loss": 0.5106, + "step": 17781 + }, + { + "epoch": 0.53, + "learning_rate": 9.533958485264551e-05, + "loss": 0.4633, + "step": 17782 + }, + { + "epoch": 0.53, + "learning_rate": 9.532995294432727e-05, + "loss": 0.5885, + "step": 17783 + }, + { + "epoch": 0.53, + "learning_rate": 9.532032107942931e-05, + "loss": 0.5855, + "step": 17784 + }, + { + "epoch": 0.53, + "learning_rate": 9.531068925804111e-05, + "loss": 0.5643, + "step": 17785 + }, + { + "epoch": 0.53, + "learning_rate": 9.530105748025224e-05, + "loss": 0.5359, + "step": 17786 + }, + { + "epoch": 0.53, + "learning_rate": 9.529142574615227e-05, + "loss": 0.5431, + "step": 17787 + }, + { + "epoch": 0.53, + "learning_rate": 9.528179405583078e-05, + "loss": 0.5155, + "step": 17788 + }, + { + "epoch": 0.53, + "learning_rate": 9.527216240937725e-05, + "loss": 0.5294, + "step": 17789 + }, + { + "epoch": 0.53, + "learning_rate": 9.526253080688129e-05, + "loss": 0.4426, + "step": 17790 + }, + { + "epoch": 0.53, + "learning_rate": 9.525289924843242e-05, + "loss": 0.6111, + "step": 17791 + }, + { + "epoch": 0.53, + "learning_rate": 9.524326773412023e-05, + "loss": 0.5511, + "step": 17792 + }, + { + "epoch": 0.53, + "learning_rate": 9.52336362640342e-05, + "loss": 0.5723, + "step": 17793 + }, + { + "epoch": 0.53, + "learning_rate": 9.522400483826394e-05, + "loss": 0.6037, + "step": 17794 + }, + { + "epoch": 0.53, + "learning_rate": 9.521437345689895e-05, + "loss": 0.6092, + "step": 17795 + }, + { + "epoch": 0.53, + "learning_rate": 9.520474212002886e-05, + "loss": 0.6257, + "step": 17796 + }, + { + "epoch": 0.53, + "learning_rate": 9.519511082774314e-05, + "loss": 0.6827, + "step": 17797 + }, + { + "epoch": 0.53, + "learning_rate": 9.518547958013135e-05, + "loss": 0.551, + "step": 17798 + }, + { + "epoch": 0.53, + "learning_rate": 9.517584837728307e-05, + "loss": 0.4739, + "step": 17799 + }, + { + "epoch": 0.53, + "learning_rate": 9.516621721928785e-05, + "loss": 0.5172, + "step": 17800 + }, + { + "epoch": 0.53, + "learning_rate": 9.51565861062352e-05, + "loss": 0.5168, + "step": 17801 + }, + { + "epoch": 0.53, + "learning_rate": 9.514695503821464e-05, + "loss": 0.6666, + "step": 17802 + }, + { + "epoch": 0.53, + "learning_rate": 9.513732401531579e-05, + "loss": 0.5581, + "step": 17803 + }, + { + "epoch": 0.53, + "learning_rate": 9.512769303762818e-05, + "loss": 0.5662, + "step": 17804 + }, + { + "epoch": 0.53, + "learning_rate": 9.511806210524131e-05, + "loss": 0.5466, + "step": 17805 + }, + { + "epoch": 0.53, + "learning_rate": 9.510843121824477e-05, + "loss": 0.4847, + "step": 17806 + }, + { + "epoch": 0.53, + "learning_rate": 9.509880037672806e-05, + "loss": 0.5721, + "step": 17807 + }, + { + "epoch": 0.53, + "learning_rate": 9.50891695807808e-05, + "loss": 0.4943, + "step": 17808 + }, + { + "epoch": 0.53, + "learning_rate": 9.507953883049245e-05, + "loss": 0.4976, + "step": 17809 + }, + { + "epoch": 0.53, + "learning_rate": 9.506990812595258e-05, + "loss": 0.5286, + "step": 17810 + }, + { + "epoch": 0.53, + "learning_rate": 9.50602774672508e-05, + "loss": 0.5852, + "step": 17811 + }, + { + "epoch": 0.53, + "learning_rate": 9.505064685447654e-05, + "loss": 0.5072, + "step": 17812 + }, + { + "epoch": 0.53, + "learning_rate": 9.504101628771938e-05, + "loss": 0.652, + "step": 17813 + }, + { + "epoch": 0.53, + "learning_rate": 9.503138576706887e-05, + "loss": 0.4736, + "step": 17814 + }, + { + "epoch": 0.53, + "learning_rate": 9.502175529261461e-05, + "loss": 0.574, + "step": 17815 + }, + { + "epoch": 0.53, + "learning_rate": 9.501212486444604e-05, + "loss": 0.5279, + "step": 17816 + }, + { + "epoch": 0.53, + "learning_rate": 9.500249448265275e-05, + "loss": 0.6248, + "step": 17817 + }, + { + "epoch": 0.53, + "learning_rate": 9.499286414732428e-05, + "loss": 0.6386, + "step": 17818 + }, + { + "epoch": 0.53, + "learning_rate": 9.49832338585502e-05, + "loss": 0.4819, + "step": 17819 + }, + { + "epoch": 0.53, + "learning_rate": 9.497360361641995e-05, + "loss": 0.5421, + "step": 17820 + }, + { + "epoch": 0.53, + "learning_rate": 9.496397342102315e-05, + "loss": 0.6322, + "step": 17821 + }, + { + "epoch": 0.53, + "learning_rate": 9.495434327244934e-05, + "loss": 0.6226, + "step": 17822 + }, + { + "epoch": 0.53, + "learning_rate": 9.494471317078804e-05, + "loss": 0.5652, + "step": 17823 + }, + { + "epoch": 0.53, + "learning_rate": 9.493508311612874e-05, + "loss": 0.477, + "step": 17824 + }, + { + "epoch": 0.53, + "learning_rate": 9.492545310856105e-05, + "loss": 0.5303, + "step": 17825 + }, + { + "epoch": 0.53, + "learning_rate": 9.491582314817444e-05, + "loss": 0.5376, + "step": 17826 + }, + { + "epoch": 0.53, + "learning_rate": 9.490619323505853e-05, + "loss": 0.3951, + "step": 17827 + }, + { + "epoch": 0.53, + "learning_rate": 9.489656336930278e-05, + "loss": 0.5472, + "step": 17828 + }, + { + "epoch": 0.53, + "learning_rate": 9.488693355099673e-05, + "loss": 0.5479, + "step": 17829 + }, + { + "epoch": 0.53, + "learning_rate": 9.487730378022995e-05, + "loss": 0.5336, + "step": 17830 + }, + { + "epoch": 0.53, + "learning_rate": 9.486767405709199e-05, + "loss": 0.5317, + "step": 17831 + }, + { + "epoch": 0.53, + "learning_rate": 9.485804438167233e-05, + "loss": 0.581, + "step": 17832 + }, + { + "epoch": 0.53, + "learning_rate": 9.484841475406049e-05, + "loss": 0.5026, + "step": 17833 + }, + { + "epoch": 0.53, + "learning_rate": 9.483878517434606e-05, + "loss": 0.5784, + "step": 17834 + }, + { + "epoch": 0.53, + "learning_rate": 9.482915564261858e-05, + "loss": 0.5732, + "step": 17835 + }, + { + "epoch": 0.53, + "learning_rate": 9.481952615896751e-05, + "loss": 0.458, + "step": 17836 + }, + { + "epoch": 0.53, + "learning_rate": 9.480989672348242e-05, + "loss": 0.4853, + "step": 17837 + }, + { + "epoch": 0.53, + "learning_rate": 9.480026733625284e-05, + "loss": 0.6032, + "step": 17838 + }, + { + "epoch": 0.53, + "learning_rate": 9.479063799736835e-05, + "loss": 0.5233, + "step": 17839 + }, + { + "epoch": 0.53, + "learning_rate": 9.478100870691839e-05, + "loss": 0.5468, + "step": 17840 + }, + { + "epoch": 0.53, + "learning_rate": 9.477137946499253e-05, + "loss": 0.5504, + "step": 17841 + }, + { + "epoch": 0.53, + "learning_rate": 9.476175027168035e-05, + "loss": 0.5238, + "step": 17842 + }, + { + "epoch": 0.53, + "learning_rate": 9.475212112707127e-05, + "loss": 0.5053, + "step": 17843 + }, + { + "epoch": 0.53, + "learning_rate": 9.474249203125489e-05, + "loss": 0.4846, + "step": 17844 + }, + { + "epoch": 0.53, + "learning_rate": 9.473286298432071e-05, + "loss": 0.5739, + "step": 17845 + }, + { + "epoch": 0.53, + "learning_rate": 9.47232339863583e-05, + "loss": 0.588, + "step": 17846 + }, + { + "epoch": 0.53, + "learning_rate": 9.471360503745716e-05, + "loss": 0.5861, + "step": 17847 + }, + { + "epoch": 0.53, + "learning_rate": 9.470397613770678e-05, + "loss": 0.5227, + "step": 17848 + }, + { + "epoch": 0.53, + "learning_rate": 9.469434728719673e-05, + "loss": 0.5087, + "step": 17849 + }, + { + "epoch": 0.53, + "learning_rate": 9.468471848601656e-05, + "loss": 0.5249, + "step": 17850 + }, + { + "epoch": 0.53, + "learning_rate": 9.467508973425572e-05, + "loss": 0.556, + "step": 17851 + }, + { + "epoch": 0.53, + "learning_rate": 9.466546103200378e-05, + "loss": 0.5096, + "step": 17852 + }, + { + "epoch": 0.53, + "learning_rate": 9.465583237935027e-05, + "loss": 0.5241, + "step": 17853 + }, + { + "epoch": 0.53, + "learning_rate": 9.464620377638471e-05, + "loss": 0.5988, + "step": 17854 + }, + { + "epoch": 0.53, + "learning_rate": 9.463657522319658e-05, + "loss": 0.5384, + "step": 17855 + }, + { + "epoch": 0.53, + "learning_rate": 9.462694671987544e-05, + "loss": 0.5844, + "step": 17856 + }, + { + "epoch": 0.53, + "learning_rate": 9.461731826651081e-05, + "loss": 0.5316, + "step": 17857 + }, + { + "epoch": 0.53, + "learning_rate": 9.460768986319224e-05, + "loss": 0.6491, + "step": 17858 + }, + { + "epoch": 0.53, + "learning_rate": 9.45980615100092e-05, + "loss": 0.4873, + "step": 17859 + }, + { + "epoch": 0.53, + "learning_rate": 9.458843320705121e-05, + "loss": 0.4785, + "step": 17860 + }, + { + "epoch": 0.53, + "learning_rate": 9.457880495440783e-05, + "loss": 0.5333, + "step": 17861 + }, + { + "epoch": 0.53, + "learning_rate": 9.456917675216859e-05, + "loss": 0.4958, + "step": 17862 + }, + { + "epoch": 0.53, + "learning_rate": 9.455954860042296e-05, + "loss": 0.6108, + "step": 17863 + }, + { + "epoch": 0.53, + "learning_rate": 9.454992049926045e-05, + "loss": 0.5261, + "step": 17864 + }, + { + "epoch": 0.53, + "learning_rate": 9.454029244877062e-05, + "loss": 0.6251, + "step": 17865 + }, + { + "epoch": 0.53, + "learning_rate": 9.4530664449043e-05, + "loss": 0.5313, + "step": 17866 + }, + { + "epoch": 0.53, + "learning_rate": 9.452103650016705e-05, + "loss": 0.5258, + "step": 17867 + }, + { + "epoch": 0.53, + "learning_rate": 9.451140860223235e-05, + "loss": 0.548, + "step": 17868 + }, + { + "epoch": 0.53, + "learning_rate": 9.450178075532834e-05, + "loss": 0.5178, + "step": 17869 + }, + { + "epoch": 0.53, + "learning_rate": 9.449215295954464e-05, + "loss": 0.6068, + "step": 17870 + }, + { + "epoch": 0.53, + "learning_rate": 9.448252521497067e-05, + "loss": 0.4588, + "step": 17871 + }, + { + "epoch": 0.53, + "learning_rate": 9.447289752169597e-05, + "loss": 0.6301, + "step": 17872 + }, + { + "epoch": 0.53, + "learning_rate": 9.446326987981009e-05, + "loss": 0.5523, + "step": 17873 + }, + { + "epoch": 0.53, + "learning_rate": 9.445364228940253e-05, + "loss": 0.5013, + "step": 17874 + }, + { + "epoch": 0.53, + "learning_rate": 9.444401475056278e-05, + "loss": 0.7336, + "step": 17875 + }, + { + "epoch": 0.53, + "learning_rate": 9.443438726338034e-05, + "loss": 0.5254, + "step": 17876 + }, + { + "epoch": 0.53, + "learning_rate": 9.442475982794481e-05, + "loss": 0.5488, + "step": 17877 + }, + { + "epoch": 0.53, + "learning_rate": 9.44151324443456e-05, + "loss": 0.5606, + "step": 17878 + }, + { + "epoch": 0.53, + "learning_rate": 9.440550511267224e-05, + "loss": 0.6923, + "step": 17879 + }, + { + "epoch": 0.53, + "learning_rate": 9.439587783301431e-05, + "loss": 0.6365, + "step": 17880 + }, + { + "epoch": 0.53, + "learning_rate": 9.438625060546127e-05, + "loss": 0.6024, + "step": 17881 + }, + { + "epoch": 0.53, + "learning_rate": 9.437662343010259e-05, + "loss": 0.4237, + "step": 17882 + }, + { + "epoch": 0.53, + "learning_rate": 9.436699630702787e-05, + "loss": 0.6045, + "step": 17883 + }, + { + "epoch": 0.53, + "learning_rate": 9.435736923632655e-05, + "loss": 0.5106, + "step": 17884 + }, + { + "epoch": 0.53, + "learning_rate": 9.43477422180882e-05, + "loss": 0.5525, + "step": 17885 + }, + { + "epoch": 0.53, + "learning_rate": 9.433811525240224e-05, + "loss": 0.4524, + "step": 17886 + }, + { + "epoch": 0.53, + "learning_rate": 9.432848833935825e-05, + "loss": 0.6209, + "step": 17887 + }, + { + "epoch": 0.53, + "learning_rate": 9.431886147904568e-05, + "loss": 0.5645, + "step": 17888 + }, + { + "epoch": 0.53, + "learning_rate": 9.430923467155413e-05, + "loss": 0.548, + "step": 17889 + }, + { + "epoch": 0.53, + "learning_rate": 9.429960791697303e-05, + "loss": 0.4186, + "step": 17890 + }, + { + "epoch": 0.53, + "learning_rate": 9.428998121539186e-05, + "loss": 0.5831, + "step": 17891 + }, + { + "epoch": 0.53, + "learning_rate": 9.428035456690021e-05, + "loss": 0.5106, + "step": 17892 + }, + { + "epoch": 0.53, + "learning_rate": 9.427072797158756e-05, + "loss": 0.5128, + "step": 17893 + }, + { + "epoch": 0.53, + "learning_rate": 9.426110142954338e-05, + "loss": 0.5077, + "step": 17894 + }, + { + "epoch": 0.53, + "learning_rate": 9.425147494085717e-05, + "loss": 0.5131, + "step": 17895 + }, + { + "epoch": 0.53, + "learning_rate": 9.424184850561846e-05, + "loss": 0.5191, + "step": 17896 + }, + { + "epoch": 0.53, + "learning_rate": 9.42322221239168e-05, + "loss": 0.6552, + "step": 17897 + }, + { + "epoch": 0.53, + "learning_rate": 9.422259579584157e-05, + "loss": 0.5485, + "step": 17898 + }, + { + "epoch": 0.53, + "learning_rate": 9.421296952148237e-05, + "loss": 0.5682, + "step": 17899 + }, + { + "epoch": 0.53, + "learning_rate": 9.420334330092867e-05, + "loss": 0.5521, + "step": 17900 + }, + { + "epoch": 0.53, + "learning_rate": 9.419371713426999e-05, + "loss": 0.5621, + "step": 17901 + }, + { + "epoch": 0.53, + "learning_rate": 9.41840910215958e-05, + "loss": 0.626, + "step": 17902 + }, + { + "epoch": 0.53, + "learning_rate": 9.41744649629956e-05, + "loss": 0.5746, + "step": 17903 + }, + { + "epoch": 0.53, + "learning_rate": 9.41648389585589e-05, + "loss": 0.5552, + "step": 17904 + }, + { + "epoch": 0.53, + "learning_rate": 9.415521300837526e-05, + "loss": 0.5216, + "step": 17905 + }, + { + "epoch": 0.53, + "learning_rate": 9.414558711253409e-05, + "loss": 0.5111, + "step": 17906 + }, + { + "epoch": 0.53, + "learning_rate": 9.413596127112487e-05, + "loss": 0.4615, + "step": 17907 + }, + { + "epoch": 0.53, + "learning_rate": 9.412633548423722e-05, + "loss": 0.5872, + "step": 17908 + }, + { + "epoch": 0.53, + "learning_rate": 9.411670975196052e-05, + "loss": 0.6361, + "step": 17909 + }, + { + "epoch": 0.53, + "learning_rate": 9.41070840743843e-05, + "loss": 0.6131, + "step": 17910 + }, + { + "epoch": 0.53, + "learning_rate": 9.409745845159806e-05, + "loss": 0.4123, + "step": 17911 + }, + { + "epoch": 0.53, + "learning_rate": 9.408783288369135e-05, + "loss": 0.5025, + "step": 17912 + }, + { + "epoch": 0.53, + "learning_rate": 9.407820737075355e-05, + "loss": 0.4879, + "step": 17913 + }, + { + "epoch": 0.53, + "learning_rate": 9.406858191287425e-05, + "loss": 0.4924, + "step": 17914 + }, + { + "epoch": 0.53, + "learning_rate": 9.405895651014289e-05, + "loss": 0.4391, + "step": 17915 + }, + { + "epoch": 0.53, + "learning_rate": 9.404933116264904e-05, + "loss": 0.6181, + "step": 17916 + }, + { + "epoch": 0.53, + "learning_rate": 9.403970587048207e-05, + "loss": 0.6951, + "step": 17917 + }, + { + "epoch": 0.53, + "learning_rate": 9.403008063373158e-05, + "loss": 0.5489, + "step": 17918 + }, + { + "epoch": 0.53, + "learning_rate": 9.402045545248699e-05, + "loss": 0.5927, + "step": 17919 + }, + { + "epoch": 0.53, + "learning_rate": 9.401083032683786e-05, + "loss": 0.568, + "step": 17920 + }, + { + "epoch": 0.53, + "learning_rate": 9.400120525687363e-05, + "loss": 0.6646, + "step": 17921 + }, + { + "epoch": 0.53, + "learning_rate": 9.399158024268378e-05, + "loss": 0.5628, + "step": 17922 + }, + { + "epoch": 0.53, + "learning_rate": 9.398195528435782e-05, + "loss": 0.5931, + "step": 17923 + }, + { + "epoch": 0.53, + "learning_rate": 9.397233038198529e-05, + "loss": 0.4945, + "step": 17924 + }, + { + "epoch": 0.53, + "learning_rate": 9.396270553565561e-05, + "loss": 0.4885, + "step": 17925 + }, + { + "epoch": 0.53, + "learning_rate": 9.395308074545825e-05, + "loss": 0.5092, + "step": 17926 + }, + { + "epoch": 0.53, + "learning_rate": 9.394345601148277e-05, + "loss": 0.5785, + "step": 17927 + }, + { + "epoch": 0.53, + "learning_rate": 9.393383133381864e-05, + "loss": 0.5371, + "step": 17928 + }, + { + "epoch": 0.53, + "learning_rate": 9.392420671255528e-05, + "loss": 0.5036, + "step": 17929 + }, + { + "epoch": 0.53, + "learning_rate": 9.391458214778223e-05, + "loss": 0.6523, + "step": 17930 + }, + { + "epoch": 0.53, + "learning_rate": 9.390495763958898e-05, + "loss": 0.5384, + "step": 17931 + }, + { + "epoch": 0.53, + "learning_rate": 9.389533318806502e-05, + "loss": 0.5181, + "step": 17932 + }, + { + "epoch": 0.53, + "learning_rate": 9.388570879329982e-05, + "loss": 0.6033, + "step": 17933 + }, + { + "epoch": 0.53, + "learning_rate": 9.387608445538284e-05, + "loss": 0.5688, + "step": 17934 + }, + { + "epoch": 0.53, + "learning_rate": 9.386646017440359e-05, + "loss": 0.6269, + "step": 17935 + }, + { + "epoch": 0.53, + "learning_rate": 9.385683595045158e-05, + "loss": 0.527, + "step": 17936 + }, + { + "epoch": 0.53, + "learning_rate": 9.384721178361625e-05, + "loss": 0.4178, + "step": 17937 + }, + { + "epoch": 0.53, + "learning_rate": 9.383758767398706e-05, + "loss": 0.5309, + "step": 17938 + }, + { + "epoch": 0.53, + "learning_rate": 9.382796362165355e-05, + "loss": 0.5552, + "step": 17939 + }, + { + "epoch": 0.53, + "learning_rate": 9.381833962670519e-05, + "loss": 0.5094, + "step": 17940 + }, + { + "epoch": 0.53, + "learning_rate": 9.38087156892314e-05, + "loss": 0.487, + "step": 17941 + }, + { + "epoch": 0.53, + "learning_rate": 9.379909180932173e-05, + "loss": 0.4958, + "step": 17942 + }, + { + "epoch": 0.53, + "learning_rate": 9.378946798706565e-05, + "loss": 0.49, + "step": 17943 + }, + { + "epoch": 0.53, + "learning_rate": 9.377984422255258e-05, + "loss": 0.473, + "step": 17944 + }, + { + "epoch": 0.53, + "learning_rate": 9.377022051587206e-05, + "loss": 0.4157, + "step": 17945 + }, + { + "epoch": 0.53, + "learning_rate": 9.376059686711354e-05, + "loss": 0.5149, + "step": 17946 + }, + { + "epoch": 0.53, + "learning_rate": 9.375097327636655e-05, + "loss": 0.4645, + "step": 17947 + }, + { + "epoch": 0.53, + "learning_rate": 9.374134974372046e-05, + "loss": 0.4978, + "step": 17948 + }, + { + "epoch": 0.53, + "learning_rate": 9.373172626926484e-05, + "loss": 0.5222, + "step": 17949 + }, + { + "epoch": 0.53, + "learning_rate": 9.372210285308909e-05, + "loss": 0.4892, + "step": 17950 + }, + { + "epoch": 0.53, + "learning_rate": 9.371247949528277e-05, + "loss": 0.5863, + "step": 17951 + }, + { + "epoch": 0.53, + "learning_rate": 9.37028561959353e-05, + "loss": 0.5028, + "step": 17952 + }, + { + "epoch": 0.53, + "learning_rate": 9.369323295513614e-05, + "loss": 0.3958, + "step": 17953 + }, + { + "epoch": 0.53, + "learning_rate": 9.368360977297481e-05, + "loss": 0.5737, + "step": 17954 + }, + { + "epoch": 0.53, + "learning_rate": 9.367398664954075e-05, + "loss": 0.5683, + "step": 17955 + }, + { + "epoch": 0.53, + "learning_rate": 9.366436358492343e-05, + "loss": 0.5015, + "step": 17956 + }, + { + "epoch": 0.53, + "learning_rate": 9.365474057921236e-05, + "loss": 0.5916, + "step": 17957 + }, + { + "epoch": 0.53, + "learning_rate": 9.364511763249695e-05, + "loss": 0.4589, + "step": 17958 + }, + { + "epoch": 0.53, + "learning_rate": 9.363549474486675e-05, + "loss": 0.5549, + "step": 17959 + }, + { + "epoch": 0.53, + "learning_rate": 9.362587191641114e-05, + "loss": 0.584, + "step": 17960 + }, + { + "epoch": 0.53, + "learning_rate": 9.361624914721965e-05, + "loss": 0.5569, + "step": 17961 + }, + { + "epoch": 0.53, + "learning_rate": 9.36066264373817e-05, + "loss": 0.569, + "step": 17962 + }, + { + "epoch": 0.53, + "learning_rate": 9.359700378698684e-05, + "loss": 0.6139, + "step": 17963 + }, + { + "epoch": 0.53, + "learning_rate": 9.358738119612447e-05, + "loss": 0.5548, + "step": 17964 + }, + { + "epoch": 0.53, + "learning_rate": 9.357775866488404e-05, + "loss": 0.5446, + "step": 17965 + }, + { + "epoch": 0.53, + "learning_rate": 9.356813619335508e-05, + "loss": 0.4644, + "step": 17966 + }, + { + "epoch": 0.53, + "learning_rate": 9.355851378162705e-05, + "loss": 0.5172, + "step": 17967 + }, + { + "epoch": 0.53, + "learning_rate": 9.354889142978937e-05, + "loss": 0.6283, + "step": 17968 + }, + { + "epoch": 0.53, + "learning_rate": 9.353926913793149e-05, + "loss": 0.5072, + "step": 17969 + }, + { + "epoch": 0.53, + "learning_rate": 9.352964690614296e-05, + "loss": 0.6174, + "step": 17970 + }, + { + "epoch": 0.54, + "learning_rate": 9.35200247345132e-05, + "loss": 0.5493, + "step": 17971 + }, + { + "epoch": 0.54, + "learning_rate": 9.351040262313162e-05, + "loss": 0.6239, + "step": 17972 + }, + { + "epoch": 0.54, + "learning_rate": 9.350078057208776e-05, + "loss": 0.5032, + "step": 17973 + }, + { + "epoch": 0.54, + "learning_rate": 9.349115858147106e-05, + "loss": 0.4858, + "step": 17974 + }, + { + "epoch": 0.54, + "learning_rate": 9.348153665137093e-05, + "loss": 0.4681, + "step": 17975 + }, + { + "epoch": 0.54, + "learning_rate": 9.347191478187692e-05, + "loss": 0.4366, + "step": 17976 + }, + { + "epoch": 0.54, + "learning_rate": 9.346229297307841e-05, + "loss": 0.4873, + "step": 17977 + }, + { + "epoch": 0.54, + "learning_rate": 9.345267122506496e-05, + "loss": 0.4988, + "step": 17978 + }, + { + "epoch": 0.54, + "learning_rate": 9.344304953792589e-05, + "loss": 0.4712, + "step": 17979 + }, + { + "epoch": 0.54, + "learning_rate": 9.343342791175076e-05, + "loss": 0.5872, + "step": 17980 + }, + { + "epoch": 0.54, + "learning_rate": 9.342380634662897e-05, + "loss": 0.4734, + "step": 17981 + }, + { + "epoch": 0.54, + "learning_rate": 9.341418484265007e-05, + "loss": 0.5625, + "step": 17982 + }, + { + "epoch": 0.54, + "learning_rate": 9.340456339990342e-05, + "loss": 0.6048, + "step": 17983 + }, + { + "epoch": 0.54, + "learning_rate": 9.339494201847848e-05, + "loss": 0.5513, + "step": 17984 + }, + { + "epoch": 0.54, + "learning_rate": 9.338532069846477e-05, + "loss": 0.545, + "step": 17985 + }, + { + "epoch": 0.54, + "learning_rate": 9.337569943995174e-05, + "loss": 0.536, + "step": 17986 + }, + { + "epoch": 0.54, + "learning_rate": 9.336607824302877e-05, + "loss": 0.5973, + "step": 17987 + }, + { + "epoch": 0.54, + "learning_rate": 9.335645710778537e-05, + "loss": 0.4334, + "step": 17988 + }, + { + "epoch": 0.54, + "learning_rate": 9.334683603431099e-05, + "loss": 0.5535, + "step": 17989 + }, + { + "epoch": 0.54, + "learning_rate": 9.33372150226951e-05, + "loss": 0.5779, + "step": 17990 + }, + { + "epoch": 0.54, + "learning_rate": 9.33275940730271e-05, + "loss": 0.4365, + "step": 17991 + }, + { + "epoch": 0.54, + "learning_rate": 9.331797318539647e-05, + "loss": 0.5155, + "step": 17992 + }, + { + "epoch": 0.54, + "learning_rate": 9.330835235989264e-05, + "loss": 0.44, + "step": 17993 + }, + { + "epoch": 0.54, + "learning_rate": 9.329873159660514e-05, + "loss": 0.545, + "step": 17994 + }, + { + "epoch": 0.54, + "learning_rate": 9.328911089562334e-05, + "loss": 0.4743, + "step": 17995 + }, + { + "epoch": 0.54, + "learning_rate": 9.327949025703668e-05, + "loss": 0.4782, + "step": 17996 + }, + { + "epoch": 0.54, + "learning_rate": 9.326986968093468e-05, + "loss": 0.3964, + "step": 17997 + }, + { + "epoch": 0.54, + "learning_rate": 9.326024916740676e-05, + "loss": 0.4633, + "step": 17998 + }, + { + "epoch": 0.54, + "learning_rate": 9.325062871654233e-05, + "loss": 0.5735, + "step": 17999 + }, + { + "epoch": 0.54, + "learning_rate": 9.324100832843085e-05, + "loss": 0.6436, + "step": 18000 + }, + { + "epoch": 0.54, + "learning_rate": 9.32313880031618e-05, + "loss": 0.5814, + "step": 18001 + }, + { + "epoch": 0.54, + "learning_rate": 9.322176774082463e-05, + "loss": 0.5267, + "step": 18002 + }, + { + "epoch": 0.54, + "learning_rate": 9.321214754150872e-05, + "loss": 0.5402, + "step": 18003 + }, + { + "epoch": 0.54, + "learning_rate": 9.320252740530357e-05, + "loss": 0.5538, + "step": 18004 + }, + { + "epoch": 0.54, + "learning_rate": 9.31929073322986e-05, + "loss": 0.5413, + "step": 18005 + }, + { + "epoch": 0.54, + "learning_rate": 9.31832873225833e-05, + "loss": 0.6661, + "step": 18006 + }, + { + "epoch": 0.54, + "learning_rate": 9.317366737624706e-05, + "loss": 0.4647, + "step": 18007 + }, + { + "epoch": 0.54, + "learning_rate": 9.316404749337931e-05, + "loss": 0.6118, + "step": 18008 + }, + { + "epoch": 0.54, + "learning_rate": 9.31544276740696e-05, + "loss": 0.492, + "step": 18009 + }, + { + "epoch": 0.54, + "learning_rate": 9.314480791840723e-05, + "loss": 0.6024, + "step": 18010 + }, + { + "epoch": 0.54, + "learning_rate": 9.31351882264817e-05, + "loss": 0.4758, + "step": 18011 + }, + { + "epoch": 0.54, + "learning_rate": 9.312556859838246e-05, + "loss": 0.538, + "step": 18012 + }, + { + "epoch": 0.54, + "learning_rate": 9.311594903419898e-05, + "loss": 0.642, + "step": 18013 + }, + { + "epoch": 0.54, + "learning_rate": 9.310632953402064e-05, + "loss": 0.4965, + "step": 18014 + }, + { + "epoch": 0.54, + "learning_rate": 9.309671009793686e-05, + "loss": 0.4644, + "step": 18015 + }, + { + "epoch": 0.54, + "learning_rate": 9.308709072603717e-05, + "loss": 0.5565, + "step": 18016 + }, + { + "epoch": 0.54, + "learning_rate": 9.307747141841097e-05, + "loss": 0.524, + "step": 18017 + }, + { + "epoch": 0.54, + "learning_rate": 9.306785217514762e-05, + "loss": 0.4794, + "step": 18018 + }, + { + "epoch": 0.54, + "learning_rate": 9.305823299633665e-05, + "loss": 0.4234, + "step": 18019 + }, + { + "epoch": 0.54, + "learning_rate": 9.304861388206747e-05, + "loss": 0.5851, + "step": 18020 + }, + { + "epoch": 0.54, + "learning_rate": 9.303899483242951e-05, + "loss": 0.5629, + "step": 18021 + }, + { + "epoch": 0.54, + "learning_rate": 9.302937584751218e-05, + "loss": 0.6069, + "step": 18022 + }, + { + "epoch": 0.54, + "learning_rate": 9.301975692740495e-05, + "loss": 0.4895, + "step": 18023 + }, + { + "epoch": 0.54, + "learning_rate": 9.301013807219721e-05, + "loss": 0.5612, + "step": 18024 + }, + { + "epoch": 0.54, + "learning_rate": 9.300051928197847e-05, + "loss": 0.5631, + "step": 18025 + }, + { + "epoch": 0.54, + "learning_rate": 9.299090055683809e-05, + "loss": 0.5584, + "step": 18026 + }, + { + "epoch": 0.54, + "learning_rate": 9.298128189686549e-05, + "loss": 0.4959, + "step": 18027 + }, + { + "epoch": 0.54, + "learning_rate": 9.297166330215017e-05, + "loss": 0.4363, + "step": 18028 + }, + { + "epoch": 0.54, + "learning_rate": 9.296204477278155e-05, + "loss": 0.5916, + "step": 18029 + }, + { + "epoch": 0.54, + "learning_rate": 9.2952426308849e-05, + "loss": 0.5318, + "step": 18030 + }, + { + "epoch": 0.54, + "learning_rate": 9.294280791044195e-05, + "loss": 0.5717, + "step": 18031 + }, + { + "epoch": 0.54, + "learning_rate": 9.29331895776499e-05, + "loss": 0.4865, + "step": 18032 + }, + { + "epoch": 0.54, + "learning_rate": 9.292357131056224e-05, + "loss": 0.5633, + "step": 18033 + }, + { + "epoch": 0.54, + "learning_rate": 9.291395310926836e-05, + "loss": 0.5328, + "step": 18034 + }, + { + "epoch": 0.54, + "learning_rate": 9.290433497385774e-05, + "loss": 0.5432, + "step": 18035 + }, + { + "epoch": 0.54, + "learning_rate": 9.289471690441976e-05, + "loss": 0.583, + "step": 18036 + }, + { + "epoch": 0.54, + "learning_rate": 9.288509890104393e-05, + "loss": 0.453, + "step": 18037 + }, + { + "epoch": 0.54, + "learning_rate": 9.287548096381959e-05, + "loss": 0.4251, + "step": 18038 + }, + { + "epoch": 0.54, + "learning_rate": 9.286586309283615e-05, + "loss": 0.5193, + "step": 18039 + }, + { + "epoch": 0.54, + "learning_rate": 9.285624528818314e-05, + "loss": 0.4738, + "step": 18040 + }, + { + "epoch": 0.54, + "learning_rate": 9.284662754994986e-05, + "loss": 0.6188, + "step": 18041 + }, + { + "epoch": 0.54, + "learning_rate": 9.28370098782258e-05, + "loss": 0.6388, + "step": 18042 + }, + { + "epoch": 0.54, + "learning_rate": 9.282739227310034e-05, + "loss": 0.537, + "step": 18043 + }, + { + "epoch": 0.54, + "learning_rate": 9.281777473466298e-05, + "loss": 0.4803, + "step": 18044 + }, + { + "epoch": 0.54, + "learning_rate": 9.280815726300307e-05, + "loss": 0.4596, + "step": 18045 + }, + { + "epoch": 0.54, + "learning_rate": 9.279853985821002e-05, + "loss": 0.542, + "step": 18046 + }, + { + "epoch": 0.54, + "learning_rate": 9.27889225203733e-05, + "loss": 0.6891, + "step": 18047 + }, + { + "epoch": 0.54, + "learning_rate": 9.277930524958233e-05, + "loss": 0.5838, + "step": 18048 + }, + { + "epoch": 0.54, + "learning_rate": 9.276968804592645e-05, + "loss": 0.541, + "step": 18049 + }, + { + "epoch": 0.54, + "learning_rate": 9.276007090949515e-05, + "loss": 0.5668, + "step": 18050 + }, + { + "epoch": 0.54, + "learning_rate": 9.275045384037783e-05, + "loss": 0.6064, + "step": 18051 + }, + { + "epoch": 0.54, + "learning_rate": 9.274083683866393e-05, + "loss": 0.5378, + "step": 18052 + }, + { + "epoch": 0.54, + "learning_rate": 9.273121990444278e-05, + "loss": 0.5362, + "step": 18053 + }, + { + "epoch": 0.54, + "learning_rate": 9.272160303780388e-05, + "loss": 0.5697, + "step": 18054 + }, + { + "epoch": 0.54, + "learning_rate": 9.271198623883658e-05, + "loss": 0.5494, + "step": 18055 + }, + { + "epoch": 0.54, + "learning_rate": 9.270236950763038e-05, + "loss": 0.5526, + "step": 18056 + }, + { + "epoch": 0.54, + "learning_rate": 9.26927528442746e-05, + "loss": 0.5356, + "step": 18057 + }, + { + "epoch": 0.54, + "learning_rate": 9.26831362488587e-05, + "loss": 0.4738, + "step": 18058 + }, + { + "epoch": 0.54, + "learning_rate": 9.267351972147209e-05, + "loss": 0.6272, + "step": 18059 + }, + { + "epoch": 0.54, + "learning_rate": 9.26639032622042e-05, + "loss": 0.5123, + "step": 18060 + }, + { + "epoch": 0.54, + "learning_rate": 9.265428687114439e-05, + "loss": 0.5846, + "step": 18061 + }, + { + "epoch": 0.54, + "learning_rate": 9.264467054838207e-05, + "loss": 0.504, + "step": 18062 + }, + { + "epoch": 0.54, + "learning_rate": 9.263505429400669e-05, + "loss": 0.4872, + "step": 18063 + }, + { + "epoch": 0.54, + "learning_rate": 9.262543810810767e-05, + "loss": 0.5474, + "step": 18064 + }, + { + "epoch": 0.54, + "learning_rate": 9.261582199077434e-05, + "loss": 0.4488, + "step": 18065 + }, + { + "epoch": 0.54, + "learning_rate": 9.260620594209617e-05, + "loss": 0.5571, + "step": 18066 + }, + { + "epoch": 0.54, + "learning_rate": 9.259658996216254e-05, + "loss": 0.554, + "step": 18067 + }, + { + "epoch": 0.54, + "learning_rate": 9.25869740510629e-05, + "loss": 0.4689, + "step": 18068 + }, + { + "epoch": 0.54, + "learning_rate": 9.25773582088866e-05, + "loss": 0.574, + "step": 18069 + }, + { + "epoch": 0.54, + "learning_rate": 9.256774243572303e-05, + "loss": 0.5959, + "step": 18070 + }, + { + "epoch": 0.54, + "learning_rate": 9.255812673166168e-05, + "loss": 0.6292, + "step": 18071 + }, + { + "epoch": 0.54, + "learning_rate": 9.25485110967919e-05, + "loss": 0.4734, + "step": 18072 + }, + { + "epoch": 0.54, + "learning_rate": 9.253889553120308e-05, + "loss": 0.5846, + "step": 18073 + }, + { + "epoch": 0.54, + "learning_rate": 9.252928003498462e-05, + "loss": 0.548, + "step": 18074 + }, + { + "epoch": 0.54, + "learning_rate": 9.251966460822597e-05, + "loss": 0.4104, + "step": 18075 + }, + { + "epoch": 0.54, + "learning_rate": 9.251004925101648e-05, + "loss": 0.5139, + "step": 18076 + }, + { + "epoch": 0.54, + "learning_rate": 9.250043396344554e-05, + "loss": 0.5168, + "step": 18077 + }, + { + "epoch": 0.54, + "learning_rate": 9.249081874560261e-05, + "loss": 0.5145, + "step": 18078 + }, + { + "epoch": 0.54, + "learning_rate": 9.248120359757707e-05, + "loss": 0.522, + "step": 18079 + }, + { + "epoch": 0.54, + "learning_rate": 9.247158851945825e-05, + "loss": 0.4842, + "step": 18080 + }, + { + "epoch": 0.54, + "learning_rate": 9.246197351133563e-05, + "loss": 0.5509, + "step": 18081 + }, + { + "epoch": 0.54, + "learning_rate": 9.245235857329856e-05, + "loss": 0.5524, + "step": 18082 + }, + { + "epoch": 0.54, + "learning_rate": 9.24427437054365e-05, + "loss": 0.5598, + "step": 18083 + }, + { + "epoch": 0.54, + "learning_rate": 9.243312890783875e-05, + "loss": 0.4884, + "step": 18084 + }, + { + "epoch": 0.54, + "learning_rate": 9.242351418059475e-05, + "loss": 0.546, + "step": 18085 + }, + { + "epoch": 0.54, + "learning_rate": 9.241389952379391e-05, + "loss": 0.5353, + "step": 18086 + }, + { + "epoch": 0.54, + "learning_rate": 9.240428493752563e-05, + "loss": 0.5861, + "step": 18087 + }, + { + "epoch": 0.54, + "learning_rate": 9.239467042187926e-05, + "loss": 0.6874, + "step": 18088 + }, + { + "epoch": 0.54, + "learning_rate": 9.23850559769442e-05, + "loss": 0.5327, + "step": 18089 + }, + { + "epoch": 0.54, + "learning_rate": 9.237544160280986e-05, + "loss": 0.6204, + "step": 18090 + }, + { + "epoch": 0.54, + "learning_rate": 9.236582729956566e-05, + "loss": 0.5143, + "step": 18091 + }, + { + "epoch": 0.54, + "learning_rate": 9.23562130673009e-05, + "loss": 0.573, + "step": 18092 + }, + { + "epoch": 0.54, + "learning_rate": 9.234659890610505e-05, + "loss": 0.5802, + "step": 18093 + }, + { + "epoch": 0.54, + "learning_rate": 9.233698481606747e-05, + "loss": 0.5368, + "step": 18094 + }, + { + "epoch": 0.54, + "learning_rate": 9.232737079727759e-05, + "loss": 0.5686, + "step": 18095 + }, + { + "epoch": 0.54, + "learning_rate": 9.23177568498247e-05, + "loss": 0.5547, + "step": 18096 + }, + { + "epoch": 0.54, + "learning_rate": 9.230814297379827e-05, + "loss": 0.5726, + "step": 18097 + }, + { + "epoch": 0.54, + "learning_rate": 9.229852916928764e-05, + "loss": 0.5538, + "step": 18098 + }, + { + "epoch": 0.54, + "learning_rate": 9.228891543638226e-05, + "loss": 0.5125, + "step": 18099 + }, + { + "epoch": 0.54, + "learning_rate": 9.227930177517145e-05, + "loss": 0.5269, + "step": 18100 + }, + { + "epoch": 0.54, + "learning_rate": 9.226968818574459e-05, + "loss": 0.4525, + "step": 18101 + }, + { + "epoch": 0.54, + "learning_rate": 9.22600746681911e-05, + "loss": 0.5954, + "step": 18102 + }, + { + "epoch": 0.54, + "learning_rate": 9.225046122260039e-05, + "loss": 0.5549, + "step": 18103 + }, + { + "epoch": 0.54, + "learning_rate": 9.224084784906179e-05, + "loss": 0.6248, + "step": 18104 + }, + { + "epoch": 0.54, + "learning_rate": 9.223123454766465e-05, + "loss": 0.409, + "step": 18105 + }, + { + "epoch": 0.54, + "learning_rate": 9.222162131849845e-05, + "loss": 0.487, + "step": 18106 + }, + { + "epoch": 0.54, + "learning_rate": 9.221200816165248e-05, + "loss": 0.5155, + "step": 18107 + }, + { + "epoch": 0.54, + "learning_rate": 9.220239507721615e-05, + "loss": 0.4409, + "step": 18108 + }, + { + "epoch": 0.54, + "learning_rate": 9.219278206527886e-05, + "loss": 0.4957, + "step": 18109 + }, + { + "epoch": 0.54, + "learning_rate": 9.218316912592998e-05, + "loss": 0.4923, + "step": 18110 + }, + { + "epoch": 0.54, + "learning_rate": 9.217355625925884e-05, + "loss": 0.4989, + "step": 18111 + }, + { + "epoch": 0.54, + "learning_rate": 9.216394346535489e-05, + "loss": 0.5359, + "step": 18112 + }, + { + "epoch": 0.54, + "learning_rate": 9.215433074430744e-05, + "loss": 0.5338, + "step": 18113 + }, + { + "epoch": 0.54, + "learning_rate": 9.214471809620596e-05, + "loss": 0.4853, + "step": 18114 + }, + { + "epoch": 0.54, + "learning_rate": 9.213510552113969e-05, + "loss": 0.5855, + "step": 18115 + }, + { + "epoch": 0.54, + "learning_rate": 9.212549301919809e-05, + "loss": 0.564, + "step": 18116 + }, + { + "epoch": 0.54, + "learning_rate": 9.211588059047052e-05, + "loss": 0.6978, + "step": 18117 + }, + { + "epoch": 0.54, + "learning_rate": 9.210626823504637e-05, + "loss": 0.5679, + "step": 18118 + }, + { + "epoch": 0.54, + "learning_rate": 9.209665595301499e-05, + "loss": 0.6669, + "step": 18119 + }, + { + "epoch": 0.54, + "learning_rate": 9.208704374446572e-05, + "loss": 0.5948, + "step": 18120 + }, + { + "epoch": 0.54, + "learning_rate": 9.207743160948798e-05, + "loss": 0.532, + "step": 18121 + }, + { + "epoch": 0.54, + "learning_rate": 9.206781954817115e-05, + "loss": 0.5384, + "step": 18122 + }, + { + "epoch": 0.54, + "learning_rate": 9.205820756060452e-05, + "loss": 0.5909, + "step": 18123 + }, + { + "epoch": 0.54, + "learning_rate": 9.204859564687756e-05, + "loss": 0.4318, + "step": 18124 + }, + { + "epoch": 0.54, + "learning_rate": 9.203898380707956e-05, + "loss": 0.5697, + "step": 18125 + }, + { + "epoch": 0.54, + "learning_rate": 9.202937204129996e-05, + "loss": 0.5982, + "step": 18126 + }, + { + "epoch": 0.54, + "learning_rate": 9.201976034962801e-05, + "loss": 0.5282, + "step": 18127 + }, + { + "epoch": 0.54, + "learning_rate": 9.201014873215319e-05, + "loss": 0.4703, + "step": 18128 + }, + { + "epoch": 0.54, + "learning_rate": 9.200053718896479e-05, + "loss": 0.4815, + "step": 18129 + }, + { + "epoch": 0.54, + "learning_rate": 9.199092572015226e-05, + "loss": 0.568, + "step": 18130 + }, + { + "epoch": 0.54, + "learning_rate": 9.19813143258049e-05, + "loss": 0.4577, + "step": 18131 + }, + { + "epoch": 0.54, + "learning_rate": 9.197170300601205e-05, + "loss": 0.4654, + "step": 18132 + }, + { + "epoch": 0.54, + "learning_rate": 9.196209176086312e-05, + "loss": 0.6012, + "step": 18133 + }, + { + "epoch": 0.54, + "learning_rate": 9.195248059044749e-05, + "loss": 0.5243, + "step": 18134 + }, + { + "epoch": 0.54, + "learning_rate": 9.194286949485446e-05, + "loss": 0.5343, + "step": 18135 + }, + { + "epoch": 0.54, + "learning_rate": 9.19332584741734e-05, + "loss": 0.566, + "step": 18136 + }, + { + "epoch": 0.54, + "learning_rate": 9.192364752849372e-05, + "loss": 0.5626, + "step": 18137 + }, + { + "epoch": 0.54, + "learning_rate": 9.191403665790474e-05, + "loss": 0.5065, + "step": 18138 + }, + { + "epoch": 0.54, + "learning_rate": 9.190442586249579e-05, + "loss": 0.5368, + "step": 18139 + }, + { + "epoch": 0.54, + "learning_rate": 9.18948151423563e-05, + "loss": 0.4445, + "step": 18140 + }, + { + "epoch": 0.54, + "learning_rate": 9.188520449757559e-05, + "loss": 0.5065, + "step": 18141 + }, + { + "epoch": 0.54, + "learning_rate": 9.187559392824297e-05, + "loss": 0.5631, + "step": 18142 + }, + { + "epoch": 0.54, + "learning_rate": 9.186598343444788e-05, + "loss": 0.4822, + "step": 18143 + }, + { + "epoch": 0.54, + "learning_rate": 9.18563730162796e-05, + "loss": 0.5588, + "step": 18144 + }, + { + "epoch": 0.54, + "learning_rate": 9.184676267382759e-05, + "loss": 0.5453, + "step": 18145 + }, + { + "epoch": 0.54, + "learning_rate": 9.183715240718107e-05, + "loss": 0.6094, + "step": 18146 + }, + { + "epoch": 0.54, + "learning_rate": 9.182754221642946e-05, + "loss": 0.5438, + "step": 18147 + }, + { + "epoch": 0.54, + "learning_rate": 9.18179321016621e-05, + "loss": 0.4932, + "step": 18148 + }, + { + "epoch": 0.54, + "learning_rate": 9.180832206296838e-05, + "loss": 0.6161, + "step": 18149 + }, + { + "epoch": 0.54, + "learning_rate": 9.17987121004376e-05, + "loss": 0.4966, + "step": 18150 + }, + { + "epoch": 0.54, + "learning_rate": 9.178910221415909e-05, + "loss": 0.562, + "step": 18151 + }, + { + "epoch": 0.54, + "learning_rate": 9.177949240422227e-05, + "loss": 0.6612, + "step": 18152 + }, + { + "epoch": 0.54, + "learning_rate": 9.176988267071647e-05, + "loss": 0.369, + "step": 18153 + }, + { + "epoch": 0.54, + "learning_rate": 9.176027301373098e-05, + "loss": 0.5983, + "step": 18154 + }, + { + "epoch": 0.54, + "learning_rate": 9.175066343335522e-05, + "loss": 0.3667, + "step": 18155 + }, + { + "epoch": 0.54, + "learning_rate": 9.174105392967849e-05, + "loss": 0.537, + "step": 18156 + }, + { + "epoch": 0.54, + "learning_rate": 9.173144450279016e-05, + "loss": 0.5261, + "step": 18157 + }, + { + "epoch": 0.54, + "learning_rate": 9.172183515277955e-05, + "loss": 0.5636, + "step": 18158 + }, + { + "epoch": 0.54, + "learning_rate": 9.171222587973601e-05, + "loss": 0.4666, + "step": 18159 + }, + { + "epoch": 0.54, + "learning_rate": 9.170261668374889e-05, + "loss": 0.622, + "step": 18160 + }, + { + "epoch": 0.54, + "learning_rate": 9.169300756490757e-05, + "loss": 0.5372, + "step": 18161 + }, + { + "epoch": 0.54, + "learning_rate": 9.168339852330132e-05, + "loss": 0.5621, + "step": 18162 + }, + { + "epoch": 0.54, + "learning_rate": 9.167378955901951e-05, + "loss": 0.6236, + "step": 18163 + }, + { + "epoch": 0.54, + "learning_rate": 9.16641806721515e-05, + "loss": 0.6645, + "step": 18164 + }, + { + "epoch": 0.54, + "learning_rate": 9.165457186278664e-05, + "loss": 0.6586, + "step": 18165 + }, + { + "epoch": 0.54, + "learning_rate": 9.164496313101422e-05, + "loss": 0.4143, + "step": 18166 + }, + { + "epoch": 0.54, + "learning_rate": 9.163535447692359e-05, + "loss": 0.5064, + "step": 18167 + }, + { + "epoch": 0.54, + "learning_rate": 9.16257459006041e-05, + "loss": 0.4903, + "step": 18168 + }, + { + "epoch": 0.54, + "learning_rate": 9.161613740214515e-05, + "loss": 0.4939, + "step": 18169 + }, + { + "epoch": 0.54, + "learning_rate": 9.160652898163594e-05, + "loss": 0.5279, + "step": 18170 + }, + { + "epoch": 0.54, + "learning_rate": 9.159692063916589e-05, + "loss": 0.6152, + "step": 18171 + }, + { + "epoch": 0.54, + "learning_rate": 9.158731237482437e-05, + "loss": 0.523, + "step": 18172 + }, + { + "epoch": 0.54, + "learning_rate": 9.15777041887006e-05, + "loss": 0.4895, + "step": 18173 + }, + { + "epoch": 0.54, + "learning_rate": 9.156809608088403e-05, + "loss": 0.414, + "step": 18174 + }, + { + "epoch": 0.54, + "learning_rate": 9.15584880514639e-05, + "loss": 0.6348, + "step": 18175 + }, + { + "epoch": 0.54, + "learning_rate": 9.154888010052966e-05, + "loss": 0.5745, + "step": 18176 + }, + { + "epoch": 0.54, + "learning_rate": 9.153927222817049e-05, + "loss": 0.4996, + "step": 18177 + }, + { + "epoch": 0.54, + "learning_rate": 9.152966443447582e-05, + "loss": 0.4829, + "step": 18178 + }, + { + "epoch": 0.54, + "learning_rate": 9.152005671953492e-05, + "loss": 0.6712, + "step": 18179 + }, + { + "epoch": 0.54, + "learning_rate": 9.151044908343721e-05, + "loss": 0.6481, + "step": 18180 + }, + { + "epoch": 0.54, + "learning_rate": 9.150084152627194e-05, + "loss": 0.5895, + "step": 18181 + }, + { + "epoch": 0.54, + "learning_rate": 9.149123404812842e-05, + "loss": 0.5703, + "step": 18182 + }, + { + "epoch": 0.54, + "learning_rate": 9.148162664909606e-05, + "loss": 0.4555, + "step": 18183 + }, + { + "epoch": 0.54, + "learning_rate": 9.147201932926415e-05, + "loss": 0.4531, + "step": 18184 + }, + { + "epoch": 0.54, + "learning_rate": 9.146241208872196e-05, + "loss": 0.4567, + "step": 18185 + }, + { + "epoch": 0.54, + "learning_rate": 9.145280492755888e-05, + "loss": 0.5881, + "step": 18186 + }, + { + "epoch": 0.54, + "learning_rate": 9.144319784586423e-05, + "loss": 0.4679, + "step": 18187 + }, + { + "epoch": 0.54, + "learning_rate": 9.143359084372731e-05, + "loss": 0.4541, + "step": 18188 + }, + { + "epoch": 0.54, + "learning_rate": 9.142398392123744e-05, + "loss": 0.5453, + "step": 18189 + }, + { + "epoch": 0.54, + "learning_rate": 9.141437707848393e-05, + "loss": 0.6002, + "step": 18190 + }, + { + "epoch": 0.54, + "learning_rate": 9.140477031555613e-05, + "loss": 0.6979, + "step": 18191 + }, + { + "epoch": 0.54, + "learning_rate": 9.139516363254338e-05, + "loss": 0.5042, + "step": 18192 + }, + { + "epoch": 0.54, + "learning_rate": 9.138555702953496e-05, + "loss": 0.606, + "step": 18193 + }, + { + "epoch": 0.54, + "learning_rate": 9.137595050662016e-05, + "loss": 0.3779, + "step": 18194 + }, + { + "epoch": 0.54, + "learning_rate": 9.136634406388835e-05, + "loss": 0.5024, + "step": 18195 + }, + { + "epoch": 0.54, + "learning_rate": 9.135673770142887e-05, + "loss": 0.6193, + "step": 18196 + }, + { + "epoch": 0.54, + "learning_rate": 9.134713141933098e-05, + "loss": 0.5789, + "step": 18197 + }, + { + "epoch": 0.54, + "learning_rate": 9.133752521768398e-05, + "loss": 0.4275, + "step": 18198 + }, + { + "epoch": 0.54, + "learning_rate": 9.132791909657724e-05, + "loss": 0.4246, + "step": 18199 + }, + { + "epoch": 0.54, + "learning_rate": 9.131831305610008e-05, + "loss": 0.6326, + "step": 18200 + }, + { + "epoch": 0.54, + "learning_rate": 9.130870709634174e-05, + "loss": 0.6014, + "step": 18201 + }, + { + "epoch": 0.54, + "learning_rate": 9.12991012173916e-05, + "loss": 0.5861, + "step": 18202 + }, + { + "epoch": 0.54, + "learning_rate": 9.128949541933896e-05, + "loss": 0.4715, + "step": 18203 + }, + { + "epoch": 0.54, + "learning_rate": 9.12798897022731e-05, + "loss": 0.5468, + "step": 18204 + }, + { + "epoch": 0.54, + "learning_rate": 9.127028406628335e-05, + "loss": 0.5022, + "step": 18205 + }, + { + "epoch": 0.54, + "learning_rate": 9.126067851145899e-05, + "loss": 0.4479, + "step": 18206 + }, + { + "epoch": 0.54, + "learning_rate": 9.125107303788944e-05, + "loss": 0.4238, + "step": 18207 + }, + { + "epoch": 0.54, + "learning_rate": 9.124146764566387e-05, + "loss": 0.5157, + "step": 18208 + }, + { + "epoch": 0.54, + "learning_rate": 9.123186233487165e-05, + "loss": 0.6553, + "step": 18209 + }, + { + "epoch": 0.54, + "learning_rate": 9.122225710560206e-05, + "loss": 0.561, + "step": 18210 + }, + { + "epoch": 0.54, + "learning_rate": 9.121265195794448e-05, + "loss": 0.5856, + "step": 18211 + }, + { + "epoch": 0.54, + "learning_rate": 9.120304689198812e-05, + "loss": 0.4567, + "step": 18212 + }, + { + "epoch": 0.54, + "learning_rate": 9.119344190782231e-05, + "loss": 0.6319, + "step": 18213 + }, + { + "epoch": 0.54, + "learning_rate": 9.118383700553638e-05, + "loss": 0.5518, + "step": 18214 + }, + { + "epoch": 0.54, + "learning_rate": 9.117423218521965e-05, + "loss": 0.6455, + "step": 18215 + }, + { + "epoch": 0.54, + "learning_rate": 9.116462744696135e-05, + "loss": 0.6261, + "step": 18216 + }, + { + "epoch": 0.54, + "learning_rate": 9.115502279085083e-05, + "loss": 0.5013, + "step": 18217 + }, + { + "epoch": 0.54, + "learning_rate": 9.11454182169774e-05, + "loss": 0.5318, + "step": 18218 + }, + { + "epoch": 0.54, + "learning_rate": 9.113581372543037e-05, + "loss": 0.5238, + "step": 18219 + }, + { + "epoch": 0.54, + "learning_rate": 9.112620931629896e-05, + "loss": 0.5722, + "step": 18220 + }, + { + "epoch": 0.54, + "learning_rate": 9.111660498967254e-05, + "loss": 0.5212, + "step": 18221 + }, + { + "epoch": 0.54, + "learning_rate": 9.110700074564036e-05, + "loss": 0.5229, + "step": 18222 + }, + { + "epoch": 0.54, + "learning_rate": 9.109739658429179e-05, + "loss": 0.5242, + "step": 18223 + }, + { + "epoch": 0.54, + "learning_rate": 9.108779250571606e-05, + "loss": 0.4898, + "step": 18224 + }, + { + "epoch": 0.54, + "learning_rate": 9.107818851000245e-05, + "loss": 0.6231, + "step": 18225 + }, + { + "epoch": 0.54, + "learning_rate": 9.106858459724032e-05, + "loss": 0.5746, + "step": 18226 + }, + { + "epoch": 0.54, + "learning_rate": 9.105898076751894e-05, + "loss": 0.464, + "step": 18227 + }, + { + "epoch": 0.54, + "learning_rate": 9.104937702092758e-05, + "loss": 0.4623, + "step": 18228 + }, + { + "epoch": 0.54, + "learning_rate": 9.103977335755552e-05, + "loss": 0.57, + "step": 18229 + }, + { + "epoch": 0.54, + "learning_rate": 9.10301697774921e-05, + "loss": 0.568, + "step": 18230 + }, + { + "epoch": 0.54, + "learning_rate": 9.102056628082661e-05, + "loss": 0.5285, + "step": 18231 + }, + { + "epoch": 0.54, + "learning_rate": 9.101096286764828e-05, + "loss": 0.693, + "step": 18232 + }, + { + "epoch": 0.54, + "learning_rate": 9.100135953804643e-05, + "loss": 0.592, + "step": 18233 + }, + { + "epoch": 0.54, + "learning_rate": 9.099175629211037e-05, + "loss": 0.6198, + "step": 18234 + }, + { + "epoch": 0.54, + "learning_rate": 9.098215312992938e-05, + "loss": 0.4633, + "step": 18235 + }, + { + "epoch": 0.54, + "learning_rate": 9.097255005159273e-05, + "loss": 0.4946, + "step": 18236 + }, + { + "epoch": 0.54, + "learning_rate": 9.096294705718968e-05, + "loss": 0.536, + "step": 18237 + }, + { + "epoch": 0.54, + "learning_rate": 9.095334414680961e-05, + "loss": 0.5862, + "step": 18238 + }, + { + "epoch": 0.54, + "learning_rate": 9.094374132054168e-05, + "loss": 0.4615, + "step": 18239 + }, + { + "epoch": 0.54, + "learning_rate": 9.093413857847527e-05, + "loss": 0.5907, + "step": 18240 + }, + { + "epoch": 0.54, + "learning_rate": 9.09245359206996e-05, + "loss": 0.5833, + "step": 18241 + }, + { + "epoch": 0.54, + "learning_rate": 9.091493334730402e-05, + "loss": 0.645, + "step": 18242 + }, + { + "epoch": 0.54, + "learning_rate": 9.090533085837772e-05, + "loss": 0.6028, + "step": 18243 + }, + { + "epoch": 0.54, + "learning_rate": 9.089572845401003e-05, + "loss": 0.4991, + "step": 18244 + }, + { + "epoch": 0.54, + "learning_rate": 9.088612613429025e-05, + "loss": 0.6147, + "step": 18245 + }, + { + "epoch": 0.54, + "learning_rate": 9.087652389930766e-05, + "loss": 0.4528, + "step": 18246 + }, + { + "epoch": 0.54, + "learning_rate": 9.086692174915147e-05, + "loss": 0.4847, + "step": 18247 + }, + { + "epoch": 0.54, + "learning_rate": 9.085731968391102e-05, + "loss": 0.6176, + "step": 18248 + }, + { + "epoch": 0.54, + "learning_rate": 9.084771770367554e-05, + "loss": 0.5236, + "step": 18249 + }, + { + "epoch": 0.54, + "learning_rate": 9.08381158085344e-05, + "loss": 0.4923, + "step": 18250 + }, + { + "epoch": 0.54, + "learning_rate": 9.082851399857675e-05, + "loss": 0.516, + "step": 18251 + }, + { + "epoch": 0.54, + "learning_rate": 9.081891227389194e-05, + "loss": 0.5767, + "step": 18252 + }, + { + "epoch": 0.54, + "learning_rate": 9.08093106345692e-05, + "loss": 0.5315, + "step": 18253 + }, + { + "epoch": 0.54, + "learning_rate": 9.079970908069788e-05, + "loss": 0.5051, + "step": 18254 + }, + { + "epoch": 0.54, + "learning_rate": 9.079010761236717e-05, + "loss": 0.5474, + "step": 18255 + }, + { + "epoch": 0.54, + "learning_rate": 9.078050622966635e-05, + "loss": 0.5659, + "step": 18256 + }, + { + "epoch": 0.54, + "learning_rate": 9.077090493268472e-05, + "loss": 0.4527, + "step": 18257 + }, + { + "epoch": 0.54, + "learning_rate": 9.076130372151157e-05, + "loss": 0.4955, + "step": 18258 + }, + { + "epoch": 0.54, + "learning_rate": 9.075170259623609e-05, + "loss": 0.5628, + "step": 18259 + }, + { + "epoch": 0.54, + "learning_rate": 9.074210155694762e-05, + "loss": 0.5013, + "step": 18260 + }, + { + "epoch": 0.54, + "learning_rate": 9.07325006037354e-05, + "loss": 0.4087, + "step": 18261 + }, + { + "epoch": 0.54, + "learning_rate": 9.072289973668871e-05, + "loss": 0.5803, + "step": 18262 + }, + { + "epoch": 0.54, + "learning_rate": 9.071329895589675e-05, + "loss": 0.5265, + "step": 18263 + }, + { + "epoch": 0.54, + "learning_rate": 9.070369826144887e-05, + "loss": 0.5751, + "step": 18264 + }, + { + "epoch": 0.54, + "learning_rate": 9.069409765343426e-05, + "loss": 0.6669, + "step": 18265 + }, + { + "epoch": 0.54, + "learning_rate": 9.068449713194228e-05, + "loss": 0.5282, + "step": 18266 + }, + { + "epoch": 0.54, + "learning_rate": 9.067489669706211e-05, + "loss": 0.4569, + "step": 18267 + }, + { + "epoch": 0.54, + "learning_rate": 9.066529634888301e-05, + "loss": 0.5827, + "step": 18268 + }, + { + "epoch": 0.54, + "learning_rate": 9.065569608749432e-05, + "loss": 0.5236, + "step": 18269 + }, + { + "epoch": 0.54, + "learning_rate": 9.06460959129852e-05, + "loss": 0.4566, + "step": 18270 + }, + { + "epoch": 0.54, + "learning_rate": 9.063649582544496e-05, + "loss": 0.4512, + "step": 18271 + }, + { + "epoch": 0.54, + "learning_rate": 9.062689582496282e-05, + "loss": 0.5281, + "step": 18272 + }, + { + "epoch": 0.54, + "learning_rate": 9.061729591162812e-05, + "loss": 0.5169, + "step": 18273 + }, + { + "epoch": 0.54, + "learning_rate": 9.060769608553005e-05, + "loss": 0.4982, + "step": 18274 + }, + { + "epoch": 0.54, + "learning_rate": 9.059809634675784e-05, + "loss": 0.6837, + "step": 18275 + }, + { + "epoch": 0.54, + "learning_rate": 9.058849669540081e-05, + "loss": 0.4849, + "step": 18276 + }, + { + "epoch": 0.54, + "learning_rate": 9.057889713154821e-05, + "loss": 0.551, + "step": 18277 + }, + { + "epoch": 0.54, + "learning_rate": 9.056929765528923e-05, + "loss": 0.5655, + "step": 18278 + }, + { + "epoch": 0.54, + "learning_rate": 9.055969826671318e-05, + "loss": 0.5429, + "step": 18279 + }, + { + "epoch": 0.54, + "learning_rate": 9.055009896590927e-05, + "loss": 0.5351, + "step": 18280 + }, + { + "epoch": 0.54, + "learning_rate": 9.054049975296684e-05, + "loss": 0.5971, + "step": 18281 + }, + { + "epoch": 0.54, + "learning_rate": 9.053090062797502e-05, + "loss": 0.6104, + "step": 18282 + }, + { + "epoch": 0.54, + "learning_rate": 9.05213015910231e-05, + "loss": 0.4847, + "step": 18283 + }, + { + "epoch": 0.54, + "learning_rate": 9.051170264220035e-05, + "loss": 0.527, + "step": 18284 + }, + { + "epoch": 0.54, + "learning_rate": 9.050210378159604e-05, + "loss": 0.5257, + "step": 18285 + }, + { + "epoch": 0.54, + "learning_rate": 9.049250500929937e-05, + "loss": 0.6139, + "step": 18286 + }, + { + "epoch": 0.54, + "learning_rate": 9.048290632539957e-05, + "loss": 0.7062, + "step": 18287 + }, + { + "epoch": 0.54, + "learning_rate": 9.047330772998593e-05, + "loss": 0.5399, + "step": 18288 + }, + { + "epoch": 0.54, + "learning_rate": 9.04637092231477e-05, + "loss": 0.4859, + "step": 18289 + }, + { + "epoch": 0.54, + "learning_rate": 9.045411080497407e-05, + "loss": 0.5959, + "step": 18290 + }, + { + "epoch": 0.54, + "learning_rate": 9.044451247555432e-05, + "loss": 0.5153, + "step": 18291 + }, + { + "epoch": 0.54, + "learning_rate": 9.04349142349777e-05, + "loss": 0.6598, + "step": 18292 + }, + { + "epoch": 0.54, + "learning_rate": 9.042531608333345e-05, + "loss": 0.6144, + "step": 18293 + }, + { + "epoch": 0.54, + "learning_rate": 9.041571802071074e-05, + "loss": 0.6207, + "step": 18294 + }, + { + "epoch": 0.54, + "learning_rate": 9.04061200471989e-05, + "loss": 0.576, + "step": 18295 + }, + { + "epoch": 0.54, + "learning_rate": 9.03965221628871e-05, + "loss": 0.5799, + "step": 18296 + }, + { + "epoch": 0.54, + "learning_rate": 9.038692436786467e-05, + "loss": 0.5232, + "step": 18297 + }, + { + "epoch": 0.54, + "learning_rate": 9.037732666222074e-05, + "loss": 0.5199, + "step": 18298 + }, + { + "epoch": 0.54, + "learning_rate": 9.036772904604459e-05, + "loss": 0.5177, + "step": 18299 + }, + { + "epoch": 0.54, + "learning_rate": 9.035813151942547e-05, + "loss": 0.5662, + "step": 18300 + }, + { + "epoch": 0.54, + "learning_rate": 9.034853408245262e-05, + "loss": 0.5847, + "step": 18301 + }, + { + "epoch": 0.54, + "learning_rate": 9.033893673521524e-05, + "loss": 0.5338, + "step": 18302 + }, + { + "epoch": 0.54, + "learning_rate": 9.032933947780254e-05, + "loss": 0.4798, + "step": 18303 + }, + { + "epoch": 0.54, + "learning_rate": 9.031974231030384e-05, + "loss": 0.6419, + "step": 18304 + }, + { + "epoch": 0.54, + "learning_rate": 9.031014523280829e-05, + "loss": 0.4816, + "step": 18305 + }, + { + "epoch": 0.54, + "learning_rate": 9.030054824540514e-05, + "loss": 0.5769, + "step": 18306 + }, + { + "epoch": 0.55, + "learning_rate": 9.029095134818364e-05, + "loss": 0.492, + "step": 18307 + }, + { + "epoch": 0.55, + "learning_rate": 9.028135454123303e-05, + "loss": 0.5665, + "step": 18308 + }, + { + "epoch": 0.55, + "learning_rate": 9.027175782464245e-05, + "loss": 0.5721, + "step": 18309 + }, + { + "epoch": 0.55, + "learning_rate": 9.026216119850124e-05, + "loss": 0.5002, + "step": 18310 + }, + { + "epoch": 0.55, + "learning_rate": 9.025256466289853e-05, + "loss": 0.5564, + "step": 18311 + }, + { + "epoch": 0.55, + "learning_rate": 9.024296821792366e-05, + "loss": 0.4727, + "step": 18312 + }, + { + "epoch": 0.55, + "learning_rate": 9.023337186366572e-05, + "loss": 0.6865, + "step": 18313 + }, + { + "epoch": 0.55, + "learning_rate": 9.022377560021401e-05, + "loss": 0.5872, + "step": 18314 + }, + { + "epoch": 0.55, + "learning_rate": 9.021417942765774e-05, + "loss": 0.5314, + "step": 18315 + }, + { + "epoch": 0.55, + "learning_rate": 9.020458334608615e-05, + "loss": 0.4161, + "step": 18316 + }, + { + "epoch": 0.55, + "learning_rate": 9.019498735558841e-05, + "loss": 0.5777, + "step": 18317 + }, + { + "epoch": 0.55, + "learning_rate": 9.018539145625377e-05, + "loss": 0.5926, + "step": 18318 + }, + { + "epoch": 0.55, + "learning_rate": 9.017579564817146e-05, + "loss": 0.5887, + "step": 18319 + }, + { + "epoch": 0.55, + "learning_rate": 9.016619993143071e-05, + "loss": 0.4948, + "step": 18320 + }, + { + "epoch": 0.55, + "learning_rate": 9.015660430612067e-05, + "loss": 0.6801, + "step": 18321 + }, + { + "epoch": 0.55, + "learning_rate": 9.014700877233062e-05, + "loss": 0.507, + "step": 18322 + }, + { + "epoch": 0.55, + "learning_rate": 9.013741333014975e-05, + "loss": 0.4988, + "step": 18323 + }, + { + "epoch": 0.55, + "learning_rate": 9.012781797966732e-05, + "loss": 0.4573, + "step": 18324 + }, + { + "epoch": 0.55, + "learning_rate": 9.011822272097245e-05, + "loss": 0.5254, + "step": 18325 + }, + { + "epoch": 0.55, + "learning_rate": 9.010862755415442e-05, + "loss": 0.596, + "step": 18326 + }, + { + "epoch": 0.55, + "learning_rate": 9.009903247930242e-05, + "loss": 0.522, + "step": 18327 + }, + { + "epoch": 0.55, + "learning_rate": 9.00894374965057e-05, + "loss": 0.5628, + "step": 18328 + }, + { + "epoch": 0.55, + "learning_rate": 9.007984260585342e-05, + "loss": 0.6021, + "step": 18329 + }, + { + "epoch": 0.55, + "learning_rate": 9.00702478074348e-05, + "loss": 0.5235, + "step": 18330 + }, + { + "epoch": 0.55, + "learning_rate": 9.006065310133907e-05, + "loss": 0.5454, + "step": 18331 + }, + { + "epoch": 0.55, + "learning_rate": 9.005105848765546e-05, + "loss": 0.4659, + "step": 18332 + }, + { + "epoch": 0.55, + "learning_rate": 9.004146396647311e-05, + "loss": 0.5928, + "step": 18333 + }, + { + "epoch": 0.55, + "learning_rate": 9.003186953788124e-05, + "loss": 0.6498, + "step": 18334 + }, + { + "epoch": 0.55, + "learning_rate": 9.002227520196912e-05, + "loss": 0.4967, + "step": 18335 + }, + { + "epoch": 0.55, + "learning_rate": 9.001268095882589e-05, + "loss": 0.5042, + "step": 18336 + }, + { + "epoch": 0.55, + "learning_rate": 9.000308680854074e-05, + "loss": 0.5544, + "step": 18337 + }, + { + "epoch": 0.55, + "learning_rate": 8.999349275120294e-05, + "loss": 0.6671, + "step": 18338 + }, + { + "epoch": 0.55, + "learning_rate": 8.998389878690167e-05, + "loss": 0.5137, + "step": 18339 + }, + { + "epoch": 0.55, + "learning_rate": 8.997430491572607e-05, + "loss": 0.5235, + "step": 18340 + }, + { + "epoch": 0.55, + "learning_rate": 8.996471113776541e-05, + "loss": 0.3935, + "step": 18341 + }, + { + "epoch": 0.55, + "learning_rate": 8.995511745310885e-05, + "loss": 0.5762, + "step": 18342 + }, + { + "epoch": 0.55, + "learning_rate": 8.994552386184566e-05, + "loss": 0.5782, + "step": 18343 + }, + { + "epoch": 0.55, + "learning_rate": 8.993593036406492e-05, + "loss": 0.5915, + "step": 18344 + }, + { + "epoch": 0.55, + "learning_rate": 8.992633695985591e-05, + "loss": 0.5005, + "step": 18345 + }, + { + "epoch": 0.55, + "learning_rate": 8.99167436493078e-05, + "loss": 0.478, + "step": 18346 + }, + { + "epoch": 0.55, + "learning_rate": 8.99071504325098e-05, + "loss": 0.5837, + "step": 18347 + }, + { + "epoch": 0.55, + "learning_rate": 8.98975573095511e-05, + "loss": 0.69, + "step": 18348 + }, + { + "epoch": 0.55, + "learning_rate": 8.988796428052084e-05, + "loss": 0.4818, + "step": 18349 + }, + { + "epoch": 0.55, + "learning_rate": 8.98783713455083e-05, + "loss": 0.5904, + "step": 18350 + }, + { + "epoch": 0.55, + "learning_rate": 8.986877850460264e-05, + "loss": 0.5706, + "step": 18351 + }, + { + "epoch": 0.55, + "learning_rate": 8.985918575789299e-05, + "loss": 0.5956, + "step": 18352 + }, + { + "epoch": 0.55, + "learning_rate": 8.984959310546863e-05, + "loss": 0.5172, + "step": 18353 + }, + { + "epoch": 0.55, + "learning_rate": 8.984000054741868e-05, + "loss": 0.4189, + "step": 18354 + }, + { + "epoch": 0.55, + "learning_rate": 8.98304080838324e-05, + "loss": 0.49, + "step": 18355 + }, + { + "epoch": 0.55, + "learning_rate": 8.98208157147989e-05, + "loss": 0.5315, + "step": 18356 + }, + { + "epoch": 0.55, + "learning_rate": 8.981122344040738e-05, + "loss": 0.5995, + "step": 18357 + }, + { + "epoch": 0.55, + "learning_rate": 8.980163126074706e-05, + "loss": 0.6025, + "step": 18358 + }, + { + "epoch": 0.55, + "learning_rate": 8.979203917590713e-05, + "loss": 0.526, + "step": 18359 + }, + { + "epoch": 0.55, + "learning_rate": 8.978244718597674e-05, + "loss": 0.6084, + "step": 18360 + }, + { + "epoch": 0.55, + "learning_rate": 8.977285529104505e-05, + "loss": 0.6602, + "step": 18361 + }, + { + "epoch": 0.55, + "learning_rate": 8.976326349120131e-05, + "loss": 0.5185, + "step": 18362 + }, + { + "epoch": 0.55, + "learning_rate": 8.975367178653469e-05, + "loss": 0.7081, + "step": 18363 + }, + { + "epoch": 0.55, + "learning_rate": 8.97440801771343e-05, + "loss": 0.5113, + "step": 18364 + }, + { + "epoch": 0.55, + "learning_rate": 8.973448866308937e-05, + "loss": 0.632, + "step": 18365 + }, + { + "epoch": 0.55, + "learning_rate": 8.972489724448908e-05, + "loss": 0.5696, + "step": 18366 + }, + { + "epoch": 0.55, + "learning_rate": 8.971530592142264e-05, + "loss": 0.5444, + "step": 18367 + }, + { + "epoch": 0.55, + "learning_rate": 8.970571469397912e-05, + "loss": 0.5791, + "step": 18368 + }, + { + "epoch": 0.55, + "learning_rate": 8.969612356224778e-05, + "loss": 0.5426, + "step": 18369 + }, + { + "epoch": 0.55, + "learning_rate": 8.968653252631782e-05, + "loss": 0.4537, + "step": 18370 + }, + { + "epoch": 0.55, + "learning_rate": 8.967694158627833e-05, + "loss": 0.5443, + "step": 18371 + }, + { + "epoch": 0.55, + "learning_rate": 8.966735074221852e-05, + "loss": 0.4183, + "step": 18372 + }, + { + "epoch": 0.55, + "learning_rate": 8.965775999422757e-05, + "loss": 0.5753, + "step": 18373 + }, + { + "epoch": 0.55, + "learning_rate": 8.964816934239469e-05, + "loss": 0.4598, + "step": 18374 + }, + { + "epoch": 0.55, + "learning_rate": 8.963857878680895e-05, + "loss": 0.5623, + "step": 18375 + }, + { + "epoch": 0.55, + "learning_rate": 8.962898832755959e-05, + "loss": 0.5116, + "step": 18376 + }, + { + "epoch": 0.55, + "learning_rate": 8.961939796473575e-05, + "loss": 0.6306, + "step": 18377 + }, + { + "epoch": 0.55, + "learning_rate": 8.960980769842664e-05, + "loss": 0.5403, + "step": 18378 + }, + { + "epoch": 0.55, + "learning_rate": 8.960021752872139e-05, + "loss": 0.6129, + "step": 18379 + }, + { + "epoch": 0.55, + "learning_rate": 8.959062745570914e-05, + "loss": 0.4824, + "step": 18380 + }, + { + "epoch": 0.55, + "learning_rate": 8.958103747947911e-05, + "loss": 0.5218, + "step": 18381 + }, + { + "epoch": 0.55, + "learning_rate": 8.957144760012046e-05, + "loss": 0.3968, + "step": 18382 + }, + { + "epoch": 0.55, + "learning_rate": 8.956185781772231e-05, + "loss": 0.594, + "step": 18383 + }, + { + "epoch": 0.55, + "learning_rate": 8.955226813237385e-05, + "loss": 0.5004, + "step": 18384 + }, + { + "epoch": 0.55, + "learning_rate": 8.954267854416422e-05, + "loss": 0.5965, + "step": 18385 + }, + { + "epoch": 0.55, + "learning_rate": 8.953308905318266e-05, + "loss": 0.5395, + "step": 18386 + }, + { + "epoch": 0.55, + "learning_rate": 8.952349965951819e-05, + "loss": 0.5011, + "step": 18387 + }, + { + "epoch": 0.55, + "learning_rate": 8.951391036326009e-05, + "loss": 0.6593, + "step": 18388 + }, + { + "epoch": 0.55, + "learning_rate": 8.950432116449744e-05, + "loss": 0.5228, + "step": 18389 + }, + { + "epoch": 0.55, + "learning_rate": 8.949473206331946e-05, + "loss": 0.5539, + "step": 18390 + }, + { + "epoch": 0.55, + "learning_rate": 8.948514305981527e-05, + "loss": 0.5516, + "step": 18391 + }, + { + "epoch": 0.55, + "learning_rate": 8.947555415407401e-05, + "loss": 0.4892, + "step": 18392 + }, + { + "epoch": 0.55, + "learning_rate": 8.946596534618485e-05, + "loss": 0.6652, + "step": 18393 + }, + { + "epoch": 0.55, + "learning_rate": 8.945637663623699e-05, + "loss": 0.4946, + "step": 18394 + }, + { + "epoch": 0.55, + "learning_rate": 8.94467880243195e-05, + "loss": 0.5643, + "step": 18395 + }, + { + "epoch": 0.55, + "learning_rate": 8.943719951052157e-05, + "loss": 0.5592, + "step": 18396 + }, + { + "epoch": 0.55, + "learning_rate": 8.942761109493236e-05, + "loss": 0.4685, + "step": 18397 + }, + { + "epoch": 0.55, + "learning_rate": 8.941802277764102e-05, + "loss": 0.5488, + "step": 18398 + }, + { + "epoch": 0.55, + "learning_rate": 8.940843455873665e-05, + "loss": 0.5997, + "step": 18399 + }, + { + "epoch": 0.55, + "learning_rate": 8.939884643830844e-05, + "loss": 0.6436, + "step": 18400 + }, + { + "epoch": 0.55, + "learning_rate": 8.938925841644558e-05, + "loss": 0.5553, + "step": 18401 + }, + { + "epoch": 0.55, + "learning_rate": 8.937967049323712e-05, + "loss": 0.5881, + "step": 18402 + }, + { + "epoch": 0.55, + "learning_rate": 8.937008266877227e-05, + "loss": 0.4384, + "step": 18403 + }, + { + "epoch": 0.55, + "learning_rate": 8.936049494314011e-05, + "loss": 0.5254, + "step": 18404 + }, + { + "epoch": 0.55, + "learning_rate": 8.93509073164299e-05, + "loss": 0.535, + "step": 18405 + }, + { + "epoch": 0.55, + "learning_rate": 8.934131978873066e-05, + "loss": 0.5915, + "step": 18406 + }, + { + "epoch": 0.55, + "learning_rate": 8.933173236013161e-05, + "loss": 0.5498, + "step": 18407 + }, + { + "epoch": 0.55, + "learning_rate": 8.932214503072182e-05, + "loss": 0.5515, + "step": 18408 + }, + { + "epoch": 0.55, + "learning_rate": 8.931255780059051e-05, + "loss": 0.6856, + "step": 18409 + }, + { + "epoch": 0.55, + "learning_rate": 8.930297066982676e-05, + "loss": 0.4993, + "step": 18410 + }, + { + "epoch": 0.55, + "learning_rate": 8.929338363851972e-05, + "loss": 0.5368, + "step": 18411 + }, + { + "epoch": 0.55, + "learning_rate": 8.928379670675854e-05, + "loss": 0.4574, + "step": 18412 + }, + { + "epoch": 0.55, + "learning_rate": 8.927420987463236e-05, + "loss": 0.6052, + "step": 18413 + }, + { + "epoch": 0.55, + "learning_rate": 8.926462314223027e-05, + "loss": 0.5786, + "step": 18414 + }, + { + "epoch": 0.55, + "learning_rate": 8.925503650964146e-05, + "loss": 0.5414, + "step": 18415 + }, + { + "epoch": 0.55, + "learning_rate": 8.9245449976955e-05, + "loss": 0.5069, + "step": 18416 + }, + { + "epoch": 0.55, + "learning_rate": 8.923586354426014e-05, + "loss": 0.5416, + "step": 18417 + }, + { + "epoch": 0.55, + "learning_rate": 8.922627721164587e-05, + "loss": 0.4888, + "step": 18418 + }, + { + "epoch": 0.55, + "learning_rate": 8.92166909792014e-05, + "loss": 0.511, + "step": 18419 + }, + { + "epoch": 0.55, + "learning_rate": 8.92071048470158e-05, + "loss": 0.6564, + "step": 18420 + }, + { + "epoch": 0.55, + "learning_rate": 8.91975188151783e-05, + "loss": 0.5802, + "step": 18421 + }, + { + "epoch": 0.55, + "learning_rate": 8.918793288377794e-05, + "loss": 0.5294, + "step": 18422 + }, + { + "epoch": 0.55, + "learning_rate": 8.917834705290385e-05, + "loss": 0.5021, + "step": 18423 + }, + { + "epoch": 0.55, + "learning_rate": 8.916876132264518e-05, + "loss": 0.5308, + "step": 18424 + }, + { + "epoch": 0.55, + "learning_rate": 8.915917569309109e-05, + "loss": 0.5564, + "step": 18425 + }, + { + "epoch": 0.55, + "learning_rate": 8.914959016433061e-05, + "loss": 0.5308, + "step": 18426 + }, + { + "epoch": 0.55, + "learning_rate": 8.914000473645295e-05, + "loss": 0.526, + "step": 18427 + }, + { + "epoch": 0.55, + "learning_rate": 8.913041940954718e-05, + "loss": 0.5074, + "step": 18428 + }, + { + "epoch": 0.55, + "learning_rate": 8.912083418370247e-05, + "loss": 0.4142, + "step": 18429 + }, + { + "epoch": 0.55, + "learning_rate": 8.911124905900787e-05, + "loss": 0.4774, + "step": 18430 + }, + { + "epoch": 0.55, + "learning_rate": 8.910166403555255e-05, + "loss": 0.4701, + "step": 18431 + }, + { + "epoch": 0.55, + "learning_rate": 8.90920791134256e-05, + "loss": 0.61, + "step": 18432 + }, + { + "epoch": 0.55, + "learning_rate": 8.908249429271619e-05, + "loss": 0.4926, + "step": 18433 + }, + { + "epoch": 0.55, + "learning_rate": 8.907290957351337e-05, + "loss": 0.479, + "step": 18434 + }, + { + "epoch": 0.55, + "learning_rate": 8.906332495590626e-05, + "loss": 0.5052, + "step": 18435 + }, + { + "epoch": 0.55, + "learning_rate": 8.905374043998404e-05, + "loss": 0.5139, + "step": 18436 + }, + { + "epoch": 0.55, + "learning_rate": 8.904415602583576e-05, + "loss": 0.5171, + "step": 18437 + }, + { + "epoch": 0.55, + "learning_rate": 8.903457171355057e-05, + "loss": 0.5124, + "step": 18438 + }, + { + "epoch": 0.55, + "learning_rate": 8.902498750321751e-05, + "loss": 0.5416, + "step": 18439 + }, + { + "epoch": 0.55, + "learning_rate": 8.90154033949258e-05, + "loss": 0.4937, + "step": 18440 + }, + { + "epoch": 0.55, + "learning_rate": 8.900581938876447e-05, + "loss": 0.6331, + "step": 18441 + }, + { + "epoch": 0.55, + "learning_rate": 8.899623548482261e-05, + "loss": 0.5279, + "step": 18442 + }, + { + "epoch": 0.55, + "learning_rate": 8.89866516831894e-05, + "loss": 0.4896, + "step": 18443 + }, + { + "epoch": 0.55, + "learning_rate": 8.897706798395394e-05, + "loss": 0.482, + "step": 18444 + }, + { + "epoch": 0.55, + "learning_rate": 8.896748438720526e-05, + "loss": 0.5343, + "step": 18445 + }, + { + "epoch": 0.55, + "learning_rate": 8.895790089303255e-05, + "loss": 0.6193, + "step": 18446 + }, + { + "epoch": 0.55, + "learning_rate": 8.894831750152483e-05, + "loss": 0.5297, + "step": 18447 + }, + { + "epoch": 0.55, + "learning_rate": 8.893873421277133e-05, + "loss": 0.6353, + "step": 18448 + }, + { + "epoch": 0.55, + "learning_rate": 8.8929151026861e-05, + "loss": 0.4557, + "step": 18449 + }, + { + "epoch": 0.55, + "learning_rate": 8.891956794388304e-05, + "loss": 0.5594, + "step": 18450 + }, + { + "epoch": 0.55, + "learning_rate": 8.89099849639265e-05, + "loss": 0.5273, + "step": 18451 + }, + { + "epoch": 0.55, + "learning_rate": 8.890040208708053e-05, + "loss": 0.5951, + "step": 18452 + }, + { + "epoch": 0.55, + "learning_rate": 8.889081931343418e-05, + "loss": 0.6411, + "step": 18453 + }, + { + "epoch": 0.55, + "learning_rate": 8.888123664307655e-05, + "loss": 0.5629, + "step": 18454 + }, + { + "epoch": 0.55, + "learning_rate": 8.887165407609676e-05, + "loss": 0.5628, + "step": 18455 + }, + { + "epoch": 0.55, + "learning_rate": 8.886207161258394e-05, + "loss": 0.5399, + "step": 18456 + }, + { + "epoch": 0.55, + "learning_rate": 8.885248925262707e-05, + "loss": 0.4951, + "step": 18457 + }, + { + "epoch": 0.55, + "learning_rate": 8.884290699631535e-05, + "loss": 0.5188, + "step": 18458 + }, + { + "epoch": 0.55, + "learning_rate": 8.883332484373781e-05, + "loss": 0.5997, + "step": 18459 + }, + { + "epoch": 0.55, + "learning_rate": 8.882374279498362e-05, + "loss": 0.568, + "step": 18460 + }, + { + "epoch": 0.55, + "learning_rate": 8.881416085014175e-05, + "loss": 0.6062, + "step": 18461 + }, + { + "epoch": 0.55, + "learning_rate": 8.880457900930139e-05, + "loss": 0.6322, + "step": 18462 + }, + { + "epoch": 0.55, + "learning_rate": 8.879499727255156e-05, + "loss": 0.4975, + "step": 18463 + }, + { + "epoch": 0.55, + "learning_rate": 8.878541563998141e-05, + "loss": 0.607, + "step": 18464 + }, + { + "epoch": 0.55, + "learning_rate": 8.877583411168e-05, + "loss": 0.5298, + "step": 18465 + }, + { + "epoch": 0.55, + "learning_rate": 8.876625268773637e-05, + "loss": 0.5476, + "step": 18466 + }, + { + "epoch": 0.55, + "learning_rate": 8.87566713682397e-05, + "loss": 0.5714, + "step": 18467 + }, + { + "epoch": 0.55, + "learning_rate": 8.8747090153279e-05, + "loss": 0.6235, + "step": 18468 + }, + { + "epoch": 0.55, + "learning_rate": 8.873750904294336e-05, + "loss": 0.5864, + "step": 18469 + }, + { + "epoch": 0.55, + "learning_rate": 8.872792803732186e-05, + "loss": 0.4933, + "step": 18470 + }, + { + "epoch": 0.55, + "learning_rate": 8.871834713650364e-05, + "loss": 0.5223, + "step": 18471 + }, + { + "epoch": 0.55, + "learning_rate": 8.87087663405777e-05, + "loss": 0.5577, + "step": 18472 + }, + { + "epoch": 0.55, + "learning_rate": 8.869918564963312e-05, + "loss": 0.5633, + "step": 18473 + }, + { + "epoch": 0.55, + "learning_rate": 8.868960506375904e-05, + "loss": 0.5003, + "step": 18474 + }, + { + "epoch": 0.55, + "learning_rate": 8.868002458304453e-05, + "loss": 0.591, + "step": 18475 + }, + { + "epoch": 0.55, + "learning_rate": 8.867044420757859e-05, + "loss": 0.5959, + "step": 18476 + }, + { + "epoch": 0.55, + "learning_rate": 8.866086393745037e-05, + "loss": 0.4991, + "step": 18477 + }, + { + "epoch": 0.55, + "learning_rate": 8.86512837727489e-05, + "loss": 0.5136, + "step": 18478 + }, + { + "epoch": 0.55, + "learning_rate": 8.864170371356331e-05, + "loss": 0.5502, + "step": 18479 + }, + { + "epoch": 0.55, + "learning_rate": 8.86321237599826e-05, + "loss": 0.5143, + "step": 18480 + }, + { + "epoch": 0.55, + "learning_rate": 8.862254391209588e-05, + "loss": 0.4304, + "step": 18481 + }, + { + "epoch": 0.55, + "learning_rate": 8.861296416999219e-05, + "loss": 0.4552, + "step": 18482 + }, + { + "epoch": 0.55, + "learning_rate": 8.860338453376066e-05, + "loss": 0.5574, + "step": 18483 + }, + { + "epoch": 0.55, + "learning_rate": 8.859380500349031e-05, + "loss": 0.5247, + "step": 18484 + }, + { + "epoch": 0.55, + "learning_rate": 8.858422557927018e-05, + "loss": 0.4851, + "step": 18485 + }, + { + "epoch": 0.55, + "learning_rate": 8.857464626118939e-05, + "loss": 0.55, + "step": 18486 + }, + { + "epoch": 0.55, + "learning_rate": 8.856506704933702e-05, + "loss": 0.5542, + "step": 18487 + }, + { + "epoch": 0.55, + "learning_rate": 8.855548794380206e-05, + "loss": 0.5546, + "step": 18488 + }, + { + "epoch": 0.55, + "learning_rate": 8.854590894467361e-05, + "loss": 0.5704, + "step": 18489 + }, + { + "epoch": 0.55, + "learning_rate": 8.853633005204075e-05, + "loss": 0.5484, + "step": 18490 + }, + { + "epoch": 0.55, + "learning_rate": 8.852675126599253e-05, + "loss": 0.5825, + "step": 18491 + }, + { + "epoch": 0.55, + "learning_rate": 8.851717258661798e-05, + "loss": 0.4895, + "step": 18492 + }, + { + "epoch": 0.55, + "learning_rate": 8.85075940140062e-05, + "loss": 0.5877, + "step": 18493 + }, + { + "epoch": 0.55, + "learning_rate": 8.849801554824621e-05, + "loss": 0.4723, + "step": 18494 + }, + { + "epoch": 0.55, + "learning_rate": 8.848843718942711e-05, + "loss": 0.5771, + "step": 18495 + }, + { + "epoch": 0.55, + "learning_rate": 8.847885893763794e-05, + "loss": 0.5815, + "step": 18496 + }, + { + "epoch": 0.55, + "learning_rate": 8.846928079296772e-05, + "loss": 0.5542, + "step": 18497 + }, + { + "epoch": 0.55, + "learning_rate": 8.845970275550555e-05, + "loss": 0.5548, + "step": 18498 + }, + { + "epoch": 0.55, + "learning_rate": 8.845012482534048e-05, + "loss": 0.6397, + "step": 18499 + }, + { + "epoch": 0.55, + "learning_rate": 8.844054700256152e-05, + "loss": 0.6086, + "step": 18500 + }, + { + "epoch": 0.55, + "learning_rate": 8.843096928725774e-05, + "loss": 0.5146, + "step": 18501 + }, + { + "epoch": 0.55, + "learning_rate": 8.842139167951823e-05, + "loss": 0.5145, + "step": 18502 + }, + { + "epoch": 0.55, + "learning_rate": 8.841181417943198e-05, + "loss": 0.4566, + "step": 18503 + }, + { + "epoch": 0.55, + "learning_rate": 8.840223678708805e-05, + "loss": 0.6208, + "step": 18504 + }, + { + "epoch": 0.55, + "learning_rate": 8.839265950257552e-05, + "loss": 0.5363, + "step": 18505 + }, + { + "epoch": 0.55, + "learning_rate": 8.838308232598343e-05, + "loss": 0.6447, + "step": 18506 + }, + { + "epoch": 0.55, + "learning_rate": 8.837350525740078e-05, + "loss": 0.5281, + "step": 18507 + }, + { + "epoch": 0.55, + "learning_rate": 8.836392829691667e-05, + "loss": 0.5725, + "step": 18508 + }, + { + "epoch": 0.55, + "learning_rate": 8.835435144462007e-05, + "loss": 0.4364, + "step": 18509 + }, + { + "epoch": 0.55, + "learning_rate": 8.834477470060016e-05, + "loss": 0.5225, + "step": 18510 + }, + { + "epoch": 0.55, + "learning_rate": 8.83351980649458e-05, + "loss": 0.5257, + "step": 18511 + }, + { + "epoch": 0.55, + "learning_rate": 8.832562153774618e-05, + "loss": 0.5395, + "step": 18512 + }, + { + "epoch": 0.55, + "learning_rate": 8.831604511909021e-05, + "loss": 0.54, + "step": 18513 + }, + { + "epoch": 0.55, + "learning_rate": 8.830646880906707e-05, + "loss": 0.408, + "step": 18514 + }, + { + "epoch": 0.55, + "learning_rate": 8.829689260776569e-05, + "loss": 0.5197, + "step": 18515 + }, + { + "epoch": 0.55, + "learning_rate": 8.828731651527511e-05, + "loss": 0.4834, + "step": 18516 + }, + { + "epoch": 0.55, + "learning_rate": 8.827774053168443e-05, + "loss": 0.541, + "step": 18517 + }, + { + "epoch": 0.55, + "learning_rate": 8.826816465708265e-05, + "loss": 0.5577, + "step": 18518 + }, + { + "epoch": 0.55, + "learning_rate": 8.825858889155878e-05, + "loss": 0.5249, + "step": 18519 + }, + { + "epoch": 0.55, + "learning_rate": 8.824901323520188e-05, + "loss": 0.4895, + "step": 18520 + }, + { + "epoch": 0.55, + "learning_rate": 8.823943768810095e-05, + "loss": 0.5407, + "step": 18521 + }, + { + "epoch": 0.55, + "learning_rate": 8.82298622503451e-05, + "loss": 0.5631, + "step": 18522 + }, + { + "epoch": 0.55, + "learning_rate": 8.822028692202323e-05, + "loss": 0.4578, + "step": 18523 + }, + { + "epoch": 0.55, + "learning_rate": 8.821071170322447e-05, + "loss": 0.495, + "step": 18524 + }, + { + "epoch": 0.55, + "learning_rate": 8.820113659403779e-05, + "loss": 0.5844, + "step": 18525 + }, + { + "epoch": 0.55, + "learning_rate": 8.819156159455228e-05, + "loss": 0.6527, + "step": 18526 + }, + { + "epoch": 0.55, + "learning_rate": 8.818198670485691e-05, + "loss": 0.4625, + "step": 18527 + }, + { + "epoch": 0.55, + "learning_rate": 8.817241192504069e-05, + "loss": 0.472, + "step": 18528 + }, + { + "epoch": 0.55, + "learning_rate": 8.81628372551927e-05, + "loss": 0.5687, + "step": 18529 + }, + { + "epoch": 0.55, + "learning_rate": 8.815326269540194e-05, + "loss": 0.4556, + "step": 18530 + }, + { + "epoch": 0.55, + "learning_rate": 8.814368824575741e-05, + "loss": 0.4762, + "step": 18531 + }, + { + "epoch": 0.55, + "learning_rate": 8.81341139063481e-05, + "loss": 0.5687, + "step": 18532 + }, + { + "epoch": 0.55, + "learning_rate": 8.812453967726312e-05, + "loss": 0.506, + "step": 18533 + }, + { + "epoch": 0.55, + "learning_rate": 8.811496555859143e-05, + "loss": 0.5008, + "step": 18534 + }, + { + "epoch": 0.55, + "learning_rate": 8.810539155042203e-05, + "loss": 0.6097, + "step": 18535 + }, + { + "epoch": 0.55, + "learning_rate": 8.809581765284397e-05, + "loss": 0.5898, + "step": 18536 + }, + { + "epoch": 0.55, + "learning_rate": 8.808624386594629e-05, + "loss": 0.6773, + "step": 18537 + }, + { + "epoch": 0.55, + "learning_rate": 8.807667018981791e-05, + "loss": 0.4555, + "step": 18538 + }, + { + "epoch": 0.55, + "learning_rate": 8.806709662454792e-05, + "loss": 0.5852, + "step": 18539 + }, + { + "epoch": 0.55, + "learning_rate": 8.805752317022528e-05, + "loss": 0.4673, + "step": 18540 + }, + { + "epoch": 0.55, + "learning_rate": 8.804794982693911e-05, + "loss": 0.5594, + "step": 18541 + }, + { + "epoch": 0.55, + "learning_rate": 8.803837659477827e-05, + "loss": 0.5747, + "step": 18542 + }, + { + "epoch": 0.55, + "learning_rate": 8.802880347383187e-05, + "loss": 0.5654, + "step": 18543 + }, + { + "epoch": 0.55, + "learning_rate": 8.801923046418884e-05, + "loss": 0.5303, + "step": 18544 + }, + { + "epoch": 0.55, + "learning_rate": 8.800965756593828e-05, + "loss": 0.5111, + "step": 18545 + }, + { + "epoch": 0.55, + "learning_rate": 8.800008477916913e-05, + "loss": 0.6214, + "step": 18546 + }, + { + "epoch": 0.55, + "learning_rate": 8.799051210397038e-05, + "loss": 0.5101, + "step": 18547 + }, + { + "epoch": 0.55, + "learning_rate": 8.79809395404311e-05, + "loss": 0.485, + "step": 18548 + }, + { + "epoch": 0.55, + "learning_rate": 8.797136708864026e-05, + "loss": 0.5553, + "step": 18549 + }, + { + "epoch": 0.55, + "learning_rate": 8.79617947486868e-05, + "loss": 0.5189, + "step": 18550 + }, + { + "epoch": 0.55, + "learning_rate": 8.795222252065982e-05, + "loss": 0.4511, + "step": 18551 + }, + { + "epoch": 0.55, + "learning_rate": 8.794265040464825e-05, + "loss": 0.54, + "step": 18552 + }, + { + "epoch": 0.55, + "learning_rate": 8.793307840074117e-05, + "loss": 0.6608, + "step": 18553 + }, + { + "epoch": 0.55, + "learning_rate": 8.792350650902745e-05, + "loss": 0.5304, + "step": 18554 + }, + { + "epoch": 0.55, + "learning_rate": 8.791393472959617e-05, + "loss": 0.5872, + "step": 18555 + }, + { + "epoch": 0.55, + "learning_rate": 8.79043630625363e-05, + "loss": 0.6478, + "step": 18556 + }, + { + "epoch": 0.55, + "learning_rate": 8.789479150793687e-05, + "loss": 0.6049, + "step": 18557 + }, + { + "epoch": 0.55, + "learning_rate": 8.788522006588684e-05, + "loss": 0.5049, + "step": 18558 + }, + { + "epoch": 0.55, + "learning_rate": 8.787564873647517e-05, + "loss": 0.574, + "step": 18559 + }, + { + "epoch": 0.55, + "learning_rate": 8.78660775197909e-05, + "loss": 0.4612, + "step": 18560 + }, + { + "epoch": 0.55, + "learning_rate": 8.785650641592305e-05, + "loss": 0.5037, + "step": 18561 + }, + { + "epoch": 0.55, + "learning_rate": 8.78469354249605e-05, + "loss": 0.6117, + "step": 18562 + }, + { + "epoch": 0.55, + "learning_rate": 8.783736454699233e-05, + "loss": 0.5929, + "step": 18563 + }, + { + "epoch": 0.55, + "learning_rate": 8.782779378210749e-05, + "loss": 0.6302, + "step": 18564 + }, + { + "epoch": 0.55, + "learning_rate": 8.781822313039501e-05, + "loss": 0.6089, + "step": 18565 + }, + { + "epoch": 0.55, + "learning_rate": 8.780865259194378e-05, + "loss": 0.4961, + "step": 18566 + }, + { + "epoch": 0.55, + "learning_rate": 8.779908216684286e-05, + "loss": 0.4524, + "step": 18567 + }, + { + "epoch": 0.55, + "learning_rate": 8.778951185518124e-05, + "loss": 0.5187, + "step": 18568 + }, + { + "epoch": 0.55, + "learning_rate": 8.777994165704783e-05, + "loss": 0.4914, + "step": 18569 + }, + { + "epoch": 0.55, + "learning_rate": 8.777037157253166e-05, + "loss": 0.5192, + "step": 18570 + }, + { + "epoch": 0.55, + "learning_rate": 8.77608016017217e-05, + "loss": 0.4615, + "step": 18571 + }, + { + "epoch": 0.55, + "learning_rate": 8.775123174470697e-05, + "loss": 0.5024, + "step": 18572 + }, + { + "epoch": 0.55, + "learning_rate": 8.774166200157636e-05, + "loss": 0.5581, + "step": 18573 + }, + { + "epoch": 0.55, + "learning_rate": 8.773209237241891e-05, + "loss": 0.71, + "step": 18574 + }, + { + "epoch": 0.55, + "learning_rate": 8.772252285732354e-05, + "loss": 0.6215, + "step": 18575 + }, + { + "epoch": 0.55, + "learning_rate": 8.771295345637931e-05, + "loss": 0.4529, + "step": 18576 + }, + { + "epoch": 0.55, + "learning_rate": 8.770338416967513e-05, + "loss": 0.5145, + "step": 18577 + }, + { + "epoch": 0.55, + "learning_rate": 8.769381499729996e-05, + "loss": 0.6073, + "step": 18578 + }, + { + "epoch": 0.55, + "learning_rate": 8.768424593934281e-05, + "loss": 0.4617, + "step": 18579 + }, + { + "epoch": 0.55, + "learning_rate": 8.767467699589263e-05, + "loss": 0.5936, + "step": 18580 + }, + { + "epoch": 0.55, + "learning_rate": 8.766510816703838e-05, + "loss": 0.4933, + "step": 18581 + }, + { + "epoch": 0.55, + "learning_rate": 8.765553945286905e-05, + "loss": 0.6386, + "step": 18582 + }, + { + "epoch": 0.55, + "learning_rate": 8.764597085347357e-05, + "loss": 0.5442, + "step": 18583 + }, + { + "epoch": 0.55, + "learning_rate": 8.7636402368941e-05, + "loss": 0.5962, + "step": 18584 + }, + { + "epoch": 0.55, + "learning_rate": 8.762683399936016e-05, + "loss": 0.5452, + "step": 18585 + }, + { + "epoch": 0.55, + "learning_rate": 8.761726574482012e-05, + "loss": 0.6314, + "step": 18586 + }, + { + "epoch": 0.55, + "learning_rate": 8.760769760540976e-05, + "loss": 0.5235, + "step": 18587 + }, + { + "epoch": 0.55, + "learning_rate": 8.759812958121816e-05, + "loss": 0.5516, + "step": 18588 + }, + { + "epoch": 0.55, + "learning_rate": 8.758856167233417e-05, + "loss": 0.5771, + "step": 18589 + }, + { + "epoch": 0.55, + "learning_rate": 8.757899387884678e-05, + "loss": 0.5894, + "step": 18590 + }, + { + "epoch": 0.55, + "learning_rate": 8.756942620084498e-05, + "loss": 0.4857, + "step": 18591 + }, + { + "epoch": 0.55, + "learning_rate": 8.75598586384177e-05, + "loss": 0.6412, + "step": 18592 + }, + { + "epoch": 0.55, + "learning_rate": 8.755029119165387e-05, + "loss": 0.593, + "step": 18593 + }, + { + "epoch": 0.55, + "learning_rate": 8.754072386064248e-05, + "loss": 0.6263, + "step": 18594 + }, + { + "epoch": 0.55, + "learning_rate": 8.753115664547249e-05, + "loss": 0.5718, + "step": 18595 + }, + { + "epoch": 0.55, + "learning_rate": 8.752158954623284e-05, + "loss": 0.5622, + "step": 18596 + }, + { + "epoch": 0.55, + "learning_rate": 8.751202256301244e-05, + "loss": 0.5405, + "step": 18597 + }, + { + "epoch": 0.55, + "learning_rate": 8.75024556959003e-05, + "loss": 0.569, + "step": 18598 + }, + { + "epoch": 0.55, + "learning_rate": 8.749288894498537e-05, + "loss": 0.6427, + "step": 18599 + }, + { + "epoch": 0.55, + "learning_rate": 8.748332231035652e-05, + "loss": 0.5638, + "step": 18600 + }, + { + "epoch": 0.55, + "learning_rate": 8.747375579210278e-05, + "loss": 0.4989, + "step": 18601 + }, + { + "epoch": 0.55, + "learning_rate": 8.746418939031304e-05, + "loss": 0.5471, + "step": 18602 + }, + { + "epoch": 0.55, + "learning_rate": 8.745462310507633e-05, + "loss": 0.5454, + "step": 18603 + }, + { + "epoch": 0.55, + "learning_rate": 8.744505693648149e-05, + "loss": 0.5484, + "step": 18604 + }, + { + "epoch": 0.55, + "learning_rate": 8.743549088461752e-05, + "loss": 0.5209, + "step": 18605 + }, + { + "epoch": 0.55, + "learning_rate": 8.742592494957332e-05, + "loss": 0.3848, + "step": 18606 + }, + { + "epoch": 0.55, + "learning_rate": 8.74163591314379e-05, + "loss": 0.5789, + "step": 18607 + }, + { + "epoch": 0.55, + "learning_rate": 8.740679343030015e-05, + "loss": 0.5996, + "step": 18608 + }, + { + "epoch": 0.55, + "learning_rate": 8.739722784624899e-05, + "loss": 0.5493, + "step": 18609 + }, + { + "epoch": 0.55, + "learning_rate": 8.738766237937339e-05, + "loss": 0.5411, + "step": 18610 + }, + { + "epoch": 0.55, + "learning_rate": 8.737809702976232e-05, + "loss": 0.5023, + "step": 18611 + }, + { + "epoch": 0.55, + "learning_rate": 8.736853179750461e-05, + "loss": 0.4841, + "step": 18612 + }, + { + "epoch": 0.55, + "learning_rate": 8.735896668268929e-05, + "loss": 0.5938, + "step": 18613 + }, + { + "epoch": 0.55, + "learning_rate": 8.734940168540524e-05, + "loss": 0.5192, + "step": 18614 + }, + { + "epoch": 0.55, + "learning_rate": 8.733983680574147e-05, + "loss": 0.5116, + "step": 18615 + }, + { + "epoch": 0.55, + "learning_rate": 8.733027204378678e-05, + "loss": 0.5311, + "step": 18616 + }, + { + "epoch": 0.55, + "learning_rate": 8.73207073996302e-05, + "loss": 0.4746, + "step": 18617 + }, + { + "epoch": 0.55, + "learning_rate": 8.731114287336062e-05, + "loss": 0.6127, + "step": 18618 + }, + { + "epoch": 0.55, + "learning_rate": 8.730157846506701e-05, + "loss": 0.4912, + "step": 18619 + }, + { + "epoch": 0.55, + "learning_rate": 8.729201417483823e-05, + "loss": 0.5108, + "step": 18620 + }, + { + "epoch": 0.55, + "learning_rate": 8.728245000276322e-05, + "loss": 0.5649, + "step": 18621 + }, + { + "epoch": 0.55, + "learning_rate": 8.727288594893094e-05, + "loss": 0.6115, + "step": 18622 + }, + { + "epoch": 0.55, + "learning_rate": 8.726332201343033e-05, + "loss": 0.5218, + "step": 18623 + }, + { + "epoch": 0.55, + "learning_rate": 8.725375819635023e-05, + "loss": 0.598, + "step": 18624 + }, + { + "epoch": 0.55, + "learning_rate": 8.724419449777962e-05, + "loss": 0.6036, + "step": 18625 + }, + { + "epoch": 0.55, + "learning_rate": 8.723463091780742e-05, + "loss": 0.5109, + "step": 18626 + }, + { + "epoch": 0.55, + "learning_rate": 8.722506745652254e-05, + "loss": 0.516, + "step": 18627 + }, + { + "epoch": 0.55, + "learning_rate": 8.721550411401385e-05, + "loss": 0.4348, + "step": 18628 + }, + { + "epoch": 0.55, + "learning_rate": 8.720594089037032e-05, + "loss": 0.5141, + "step": 18629 + }, + { + "epoch": 0.55, + "learning_rate": 8.719637778568085e-05, + "loss": 0.4975, + "step": 18630 + }, + { + "epoch": 0.55, + "learning_rate": 8.718681480003438e-05, + "loss": 0.5071, + "step": 18631 + }, + { + "epoch": 0.55, + "learning_rate": 8.71772519335198e-05, + "loss": 0.6788, + "step": 18632 + }, + { + "epoch": 0.55, + "learning_rate": 8.716768918622599e-05, + "loss": 0.5253, + "step": 18633 + }, + { + "epoch": 0.55, + "learning_rate": 8.715812655824194e-05, + "loss": 0.624, + "step": 18634 + }, + { + "epoch": 0.55, + "learning_rate": 8.714856404965648e-05, + "loss": 0.5879, + "step": 18635 + }, + { + "epoch": 0.55, + "learning_rate": 8.713900166055855e-05, + "loss": 0.506, + "step": 18636 + }, + { + "epoch": 0.55, + "learning_rate": 8.712943939103704e-05, + "loss": 0.53, + "step": 18637 + }, + { + "epoch": 0.55, + "learning_rate": 8.71198772411809e-05, + "loss": 0.5115, + "step": 18638 + }, + { + "epoch": 0.55, + "learning_rate": 8.711031521107901e-05, + "loss": 0.572, + "step": 18639 + }, + { + "epoch": 0.55, + "learning_rate": 8.710075330082023e-05, + "loss": 0.47, + "step": 18640 + }, + { + "epoch": 0.55, + "learning_rate": 8.709119151049354e-05, + "loss": 0.561, + "step": 18641 + }, + { + "epoch": 0.56, + "learning_rate": 8.708162984018782e-05, + "loss": 0.5839, + "step": 18642 + }, + { + "epoch": 0.56, + "learning_rate": 8.707206828999193e-05, + "loss": 0.4319, + "step": 18643 + }, + { + "epoch": 0.56, + "learning_rate": 8.70625068599948e-05, + "loss": 0.5585, + "step": 18644 + }, + { + "epoch": 0.56, + "learning_rate": 8.70529455502853e-05, + "loss": 0.587, + "step": 18645 + }, + { + "epoch": 0.56, + "learning_rate": 8.704338436095244e-05, + "loss": 0.4634, + "step": 18646 + }, + { + "epoch": 0.56, + "learning_rate": 8.703382329208494e-05, + "loss": 0.5093, + "step": 18647 + }, + { + "epoch": 0.56, + "learning_rate": 8.702426234377181e-05, + "loss": 0.506, + "step": 18648 + }, + { + "epoch": 0.56, + "learning_rate": 8.70147015161019e-05, + "loss": 0.496, + "step": 18649 + }, + { + "epoch": 0.56, + "learning_rate": 8.700514080916416e-05, + "loss": 0.6084, + "step": 18650 + }, + { + "epoch": 0.56, + "learning_rate": 8.699558022304743e-05, + "loss": 0.6151, + "step": 18651 + }, + { + "epoch": 0.56, + "learning_rate": 8.698601975784057e-05, + "loss": 0.4947, + "step": 18652 + }, + { + "epoch": 0.56, + "learning_rate": 8.697645941363255e-05, + "loss": 0.6523, + "step": 18653 + }, + { + "epoch": 0.56, + "learning_rate": 8.696689919051224e-05, + "loss": 0.5854, + "step": 18654 + }, + { + "epoch": 0.56, + "learning_rate": 8.695733908856845e-05, + "loss": 0.5945, + "step": 18655 + }, + { + "epoch": 0.56, + "learning_rate": 8.694777910789016e-05, + "loss": 0.6519, + "step": 18656 + }, + { + "epoch": 0.56, + "learning_rate": 8.693821924856622e-05, + "loss": 0.4919, + "step": 18657 + }, + { + "epoch": 0.56, + "learning_rate": 8.692865951068553e-05, + "loss": 0.5639, + "step": 18658 + }, + { + "epoch": 0.56, + "learning_rate": 8.69190998943369e-05, + "loss": 0.4937, + "step": 18659 + }, + { + "epoch": 0.56, + "learning_rate": 8.69095403996093e-05, + "loss": 0.493, + "step": 18660 + }, + { + "epoch": 0.56, + "learning_rate": 8.689998102659154e-05, + "loss": 0.5236, + "step": 18661 + }, + { + "epoch": 0.56, + "learning_rate": 8.689042177537259e-05, + "loss": 0.4877, + "step": 18662 + }, + { + "epoch": 0.56, + "learning_rate": 8.688086264604125e-05, + "loss": 0.5256, + "step": 18663 + }, + { + "epoch": 0.56, + "learning_rate": 8.68713036386864e-05, + "loss": 0.4674, + "step": 18664 + }, + { + "epoch": 0.56, + "learning_rate": 8.686174475339697e-05, + "loss": 0.5422, + "step": 18665 + }, + { + "epoch": 0.56, + "learning_rate": 8.68521859902618e-05, + "loss": 0.5885, + "step": 18666 + }, + { + "epoch": 0.56, + "learning_rate": 8.684262734936974e-05, + "loss": 0.5156, + "step": 18667 + }, + { + "epoch": 0.56, + "learning_rate": 8.683306883080968e-05, + "loss": 0.5051, + "step": 18668 + }, + { + "epoch": 0.56, + "learning_rate": 8.682351043467054e-05, + "loss": 0.6043, + "step": 18669 + }, + { + "epoch": 0.56, + "learning_rate": 8.681395216104112e-05, + "loss": 0.6852, + "step": 18670 + }, + { + "epoch": 0.56, + "learning_rate": 8.68043940100103e-05, + "loss": 0.5375, + "step": 18671 + }, + { + "epoch": 0.56, + "learning_rate": 8.679483598166698e-05, + "loss": 0.6654, + "step": 18672 + }, + { + "epoch": 0.56, + "learning_rate": 8.678527807610005e-05, + "loss": 0.6405, + "step": 18673 + }, + { + "epoch": 0.56, + "learning_rate": 8.677572029339827e-05, + "loss": 0.4644, + "step": 18674 + }, + { + "epoch": 0.56, + "learning_rate": 8.676616263365062e-05, + "loss": 0.6434, + "step": 18675 + }, + { + "epoch": 0.56, + "learning_rate": 8.675660509694588e-05, + "loss": 0.4463, + "step": 18676 + }, + { + "epoch": 0.56, + "learning_rate": 8.674704768337301e-05, + "loss": 0.5267, + "step": 18677 + }, + { + "epoch": 0.56, + "learning_rate": 8.673749039302075e-05, + "loss": 0.4454, + "step": 18678 + }, + { + "epoch": 0.56, + "learning_rate": 8.672793322597805e-05, + "loss": 0.5694, + "step": 18679 + }, + { + "epoch": 0.56, + "learning_rate": 8.671837618233369e-05, + "loss": 0.6057, + "step": 18680 + }, + { + "epoch": 0.56, + "learning_rate": 8.670881926217664e-05, + "loss": 0.4828, + "step": 18681 + }, + { + "epoch": 0.56, + "learning_rate": 8.669926246559565e-05, + "loss": 0.6085, + "step": 18682 + }, + { + "epoch": 0.56, + "learning_rate": 8.668970579267961e-05, + "loss": 0.5671, + "step": 18683 + }, + { + "epoch": 0.56, + "learning_rate": 8.66801492435174e-05, + "loss": 0.5884, + "step": 18684 + }, + { + "epoch": 0.56, + "learning_rate": 8.667059281819787e-05, + "loss": 0.5829, + "step": 18685 + }, + { + "epoch": 0.56, + "learning_rate": 8.666103651680983e-05, + "loss": 0.5716, + "step": 18686 + }, + { + "epoch": 0.56, + "learning_rate": 8.665148033944215e-05, + "loss": 0.521, + "step": 18687 + }, + { + "epoch": 0.56, + "learning_rate": 8.664192428618367e-05, + "loss": 0.4689, + "step": 18688 + }, + { + "epoch": 0.56, + "learning_rate": 8.663236835712333e-05, + "loss": 0.5462, + "step": 18689 + }, + { + "epoch": 0.56, + "learning_rate": 8.662281255234984e-05, + "loss": 0.6058, + "step": 18690 + }, + { + "epoch": 0.56, + "learning_rate": 8.661325687195211e-05, + "loss": 0.6395, + "step": 18691 + }, + { + "epoch": 0.56, + "learning_rate": 8.660370131601898e-05, + "loss": 0.5575, + "step": 18692 + }, + { + "epoch": 0.56, + "learning_rate": 8.659414588463933e-05, + "loss": 0.5409, + "step": 18693 + }, + { + "epoch": 0.56, + "learning_rate": 8.658459057790195e-05, + "loss": 0.5236, + "step": 18694 + }, + { + "epoch": 0.56, + "learning_rate": 8.657503539589568e-05, + "loss": 0.6264, + "step": 18695 + }, + { + "epoch": 0.56, + "learning_rate": 8.656548033870941e-05, + "loss": 0.5548, + "step": 18696 + }, + { + "epoch": 0.56, + "learning_rate": 8.655592540643195e-05, + "loss": 0.5762, + "step": 18697 + }, + { + "epoch": 0.56, + "learning_rate": 8.654637059915211e-05, + "loss": 0.637, + "step": 18698 + }, + { + "epoch": 0.56, + "learning_rate": 8.653681591695875e-05, + "loss": 0.5004, + "step": 18699 + }, + { + "epoch": 0.56, + "learning_rate": 8.652726135994076e-05, + "loss": 0.4933, + "step": 18700 + }, + { + "epoch": 0.56, + "learning_rate": 8.651770692818689e-05, + "loss": 0.6643, + "step": 18701 + }, + { + "epoch": 0.56, + "learning_rate": 8.650815262178598e-05, + "loss": 0.5568, + "step": 18702 + }, + { + "epoch": 0.56, + "learning_rate": 8.649859844082693e-05, + "loss": 0.5499, + "step": 18703 + }, + { + "epoch": 0.56, + "learning_rate": 8.648904438539853e-05, + "loss": 0.512, + "step": 18704 + }, + { + "epoch": 0.56, + "learning_rate": 8.647949045558959e-05, + "loss": 0.5318, + "step": 18705 + }, + { + "epoch": 0.56, + "learning_rate": 8.646993665148897e-05, + "loss": 0.6209, + "step": 18706 + }, + { + "epoch": 0.56, + "learning_rate": 8.646038297318545e-05, + "loss": 0.4832, + "step": 18707 + }, + { + "epoch": 0.56, + "learning_rate": 8.645082942076799e-05, + "loss": 0.4209, + "step": 18708 + }, + { + "epoch": 0.56, + "learning_rate": 8.644127599432523e-05, + "loss": 0.4876, + "step": 18709 + }, + { + "epoch": 0.56, + "learning_rate": 8.643172269394611e-05, + "loss": 0.4451, + "step": 18710 + }, + { + "epoch": 0.56, + "learning_rate": 8.64221695197194e-05, + "loss": 0.5688, + "step": 18711 + }, + { + "epoch": 0.56, + "learning_rate": 8.641261647173398e-05, + "loss": 0.5985, + "step": 18712 + }, + { + "epoch": 0.56, + "learning_rate": 8.640306355007863e-05, + "loss": 0.4981, + "step": 18713 + }, + { + "epoch": 0.56, + "learning_rate": 8.639351075484216e-05, + "loss": 0.5595, + "step": 18714 + }, + { + "epoch": 0.56, + "learning_rate": 8.638395808611342e-05, + "loss": 0.5111, + "step": 18715 + }, + { + "epoch": 0.56, + "learning_rate": 8.637440554398122e-05, + "loss": 0.5452, + "step": 18716 + }, + { + "epoch": 0.56, + "learning_rate": 8.636485312853433e-05, + "loss": 0.5134, + "step": 18717 + }, + { + "epoch": 0.56, + "learning_rate": 8.635530083986163e-05, + "loss": 0.5683, + "step": 18718 + }, + { + "epoch": 0.56, + "learning_rate": 8.634574867805187e-05, + "loss": 0.4721, + "step": 18719 + }, + { + "epoch": 0.56, + "learning_rate": 8.633619664319398e-05, + "loss": 0.5919, + "step": 18720 + }, + { + "epoch": 0.56, + "learning_rate": 8.63266447353766e-05, + "loss": 0.5636, + "step": 18721 + }, + { + "epoch": 0.56, + "learning_rate": 8.631709295468866e-05, + "loss": 0.4681, + "step": 18722 + }, + { + "epoch": 0.56, + "learning_rate": 8.630754130121892e-05, + "loss": 0.6673, + "step": 18723 + }, + { + "epoch": 0.56, + "learning_rate": 8.629798977505624e-05, + "loss": 0.5812, + "step": 18724 + }, + { + "epoch": 0.56, + "learning_rate": 8.628843837628937e-05, + "loss": 0.4904, + "step": 18725 + }, + { + "epoch": 0.56, + "learning_rate": 8.627888710500712e-05, + "loss": 0.487, + "step": 18726 + }, + { + "epoch": 0.56, + "learning_rate": 8.626933596129833e-05, + "loss": 0.58, + "step": 18727 + }, + { + "epoch": 0.56, + "learning_rate": 8.62597849452518e-05, + "loss": 0.4526, + "step": 18728 + }, + { + "epoch": 0.56, + "learning_rate": 8.625023405695627e-05, + "loss": 0.6276, + "step": 18729 + }, + { + "epoch": 0.56, + "learning_rate": 8.62406832965006e-05, + "loss": 0.6094, + "step": 18730 + }, + { + "epoch": 0.56, + "learning_rate": 8.62311326639736e-05, + "loss": 0.5504, + "step": 18731 + }, + { + "epoch": 0.56, + "learning_rate": 8.622158215946401e-05, + "loss": 0.5974, + "step": 18732 + }, + { + "epoch": 0.56, + "learning_rate": 8.621203178306066e-05, + "loss": 0.5912, + "step": 18733 + }, + { + "epoch": 0.56, + "learning_rate": 8.620248153485233e-05, + "loss": 0.4756, + "step": 18734 + }, + { + "epoch": 0.56, + "learning_rate": 8.619293141492788e-05, + "loss": 0.5174, + "step": 18735 + }, + { + "epoch": 0.56, + "learning_rate": 8.618338142337598e-05, + "loss": 0.5272, + "step": 18736 + }, + { + "epoch": 0.56, + "learning_rate": 8.617383156028553e-05, + "loss": 0.6997, + "step": 18737 + }, + { + "epoch": 0.56, + "learning_rate": 8.616428182574527e-05, + "loss": 0.5488, + "step": 18738 + }, + { + "epoch": 0.56, + "learning_rate": 8.615473221984406e-05, + "loss": 0.4824, + "step": 18739 + }, + { + "epoch": 0.56, + "learning_rate": 8.614518274267056e-05, + "loss": 0.4566, + "step": 18740 + }, + { + "epoch": 0.56, + "learning_rate": 8.613563339431367e-05, + "loss": 0.4927, + "step": 18741 + }, + { + "epoch": 0.56, + "learning_rate": 8.612608417486209e-05, + "loss": 0.5867, + "step": 18742 + }, + { + "epoch": 0.56, + "learning_rate": 8.611653508440471e-05, + "loss": 0.6062, + "step": 18743 + }, + { + "epoch": 0.56, + "learning_rate": 8.610698612303022e-05, + "loss": 0.5104, + "step": 18744 + }, + { + "epoch": 0.56, + "learning_rate": 8.609743729082742e-05, + "loss": 0.5727, + "step": 18745 + }, + { + "epoch": 0.56, + "learning_rate": 8.608788858788512e-05, + "loss": 0.5308, + "step": 18746 + }, + { + "epoch": 0.56, + "learning_rate": 8.607834001429212e-05, + "loss": 0.5141, + "step": 18747 + }, + { + "epoch": 0.56, + "learning_rate": 8.606879157013712e-05, + "loss": 0.4435, + "step": 18748 + }, + { + "epoch": 0.56, + "learning_rate": 8.605924325550895e-05, + "loss": 0.529, + "step": 18749 + }, + { + "epoch": 0.56, + "learning_rate": 8.604969507049639e-05, + "loss": 0.5599, + "step": 18750 + }, + { + "epoch": 0.56, + "learning_rate": 8.604014701518825e-05, + "loss": 0.4749, + "step": 18751 + }, + { + "epoch": 0.56, + "learning_rate": 8.603059908967319e-05, + "loss": 0.4272, + "step": 18752 + }, + { + "epoch": 0.56, + "learning_rate": 8.602105129404008e-05, + "loss": 0.5119, + "step": 18753 + }, + { + "epoch": 0.56, + "learning_rate": 8.601150362837765e-05, + "loss": 0.5931, + "step": 18754 + }, + { + "epoch": 0.56, + "learning_rate": 8.600195609277471e-05, + "loss": 0.5624, + "step": 18755 + }, + { + "epoch": 0.56, + "learning_rate": 8.599240868731999e-05, + "loss": 0.6825, + "step": 18756 + }, + { + "epoch": 0.56, + "learning_rate": 8.598286141210226e-05, + "loss": 0.5363, + "step": 18757 + }, + { + "epoch": 0.56, + "learning_rate": 8.597331426721032e-05, + "loss": 0.4935, + "step": 18758 + }, + { + "epoch": 0.56, + "learning_rate": 8.596376725273292e-05, + "loss": 0.5055, + "step": 18759 + }, + { + "epoch": 0.56, + "learning_rate": 8.595422036875877e-05, + "loss": 0.6111, + "step": 18760 + }, + { + "epoch": 0.56, + "learning_rate": 8.594467361537671e-05, + "loss": 0.5042, + "step": 18761 + }, + { + "epoch": 0.56, + "learning_rate": 8.593512699267551e-05, + "loss": 0.5477, + "step": 18762 + }, + { + "epoch": 0.56, + "learning_rate": 8.592558050074386e-05, + "loss": 0.5158, + "step": 18763 + }, + { + "epoch": 0.56, + "learning_rate": 8.591603413967052e-05, + "loss": 0.5039, + "step": 18764 + }, + { + "epoch": 0.56, + "learning_rate": 8.590648790954432e-05, + "loss": 0.4531, + "step": 18765 + }, + { + "epoch": 0.56, + "learning_rate": 8.589694181045398e-05, + "loss": 0.4937, + "step": 18766 + }, + { + "epoch": 0.56, + "learning_rate": 8.588739584248824e-05, + "loss": 0.5604, + "step": 18767 + }, + { + "epoch": 0.56, + "learning_rate": 8.587785000573587e-05, + "loss": 0.5336, + "step": 18768 + }, + { + "epoch": 0.56, + "learning_rate": 8.586830430028561e-05, + "loss": 0.6025, + "step": 18769 + }, + { + "epoch": 0.56, + "learning_rate": 8.585875872622625e-05, + "loss": 0.4791, + "step": 18770 + }, + { + "epoch": 0.56, + "learning_rate": 8.58492132836465e-05, + "loss": 0.5081, + "step": 18771 + }, + { + "epoch": 0.56, + "learning_rate": 8.583966797263515e-05, + "loss": 0.5673, + "step": 18772 + }, + { + "epoch": 0.56, + "learning_rate": 8.583012279328087e-05, + "loss": 0.6153, + "step": 18773 + }, + { + "epoch": 0.56, + "learning_rate": 8.582057774567251e-05, + "loss": 0.6233, + "step": 18774 + }, + { + "epoch": 0.56, + "learning_rate": 8.581103282989876e-05, + "loss": 0.5215, + "step": 18775 + }, + { + "epoch": 0.56, + "learning_rate": 8.580148804604835e-05, + "loss": 0.4817, + "step": 18776 + }, + { + "epoch": 0.56, + "learning_rate": 8.579194339421005e-05, + "loss": 0.5857, + "step": 18777 + }, + { + "epoch": 0.56, + "learning_rate": 8.578239887447262e-05, + "loss": 0.4541, + "step": 18778 + }, + { + "epoch": 0.56, + "learning_rate": 8.577285448692474e-05, + "loss": 0.5753, + "step": 18779 + }, + { + "epoch": 0.56, + "learning_rate": 8.576331023165522e-05, + "loss": 0.5677, + "step": 18780 + }, + { + "epoch": 0.56, + "learning_rate": 8.575376610875275e-05, + "loss": 0.5562, + "step": 18781 + }, + { + "epoch": 0.56, + "learning_rate": 8.574422211830613e-05, + "loss": 0.4409, + "step": 18782 + }, + { + "epoch": 0.56, + "learning_rate": 8.5734678260404e-05, + "loss": 0.5339, + "step": 18783 + }, + { + "epoch": 0.56, + "learning_rate": 8.572513453513516e-05, + "loss": 0.51, + "step": 18784 + }, + { + "epoch": 0.56, + "learning_rate": 8.571559094258831e-05, + "loss": 0.5923, + "step": 18785 + }, + { + "epoch": 0.56, + "learning_rate": 8.570604748285225e-05, + "loss": 0.5336, + "step": 18786 + }, + { + "epoch": 0.56, + "learning_rate": 8.569650415601563e-05, + "loss": 0.7531, + "step": 18787 + }, + { + "epoch": 0.56, + "learning_rate": 8.56869609621672e-05, + "loss": 0.5084, + "step": 18788 + }, + { + "epoch": 0.56, + "learning_rate": 8.567741790139571e-05, + "loss": 0.5178, + "step": 18789 + }, + { + "epoch": 0.56, + "learning_rate": 8.566787497378993e-05, + "loss": 0.4888, + "step": 18790 + }, + { + "epoch": 0.56, + "learning_rate": 8.565833217943848e-05, + "loss": 0.4722, + "step": 18791 + }, + { + "epoch": 0.56, + "learning_rate": 8.564878951843016e-05, + "loss": 0.4533, + "step": 18792 + }, + { + "epoch": 0.56, + "learning_rate": 8.563924699085367e-05, + "loss": 0.5288, + "step": 18793 + }, + { + "epoch": 0.56, + "learning_rate": 8.562970459679778e-05, + "loss": 0.5834, + "step": 18794 + }, + { + "epoch": 0.56, + "learning_rate": 8.56201623363511e-05, + "loss": 0.5902, + "step": 18795 + }, + { + "epoch": 0.56, + "learning_rate": 8.561062020960245e-05, + "loss": 0.4637, + "step": 18796 + }, + { + "epoch": 0.56, + "learning_rate": 8.560107821664056e-05, + "loss": 0.6693, + "step": 18797 + }, + { + "epoch": 0.56, + "learning_rate": 8.559153635755403e-05, + "loss": 0.6565, + "step": 18798 + }, + { + "epoch": 0.56, + "learning_rate": 8.55819946324317e-05, + "loss": 0.5457, + "step": 18799 + }, + { + "epoch": 0.56, + "learning_rate": 8.55724530413622e-05, + "loss": 0.4914, + "step": 18800 + }, + { + "epoch": 0.56, + "learning_rate": 8.556291158443433e-05, + "loss": 0.4954, + "step": 18801 + }, + { + "epoch": 0.56, + "learning_rate": 8.555337026173673e-05, + "loss": 0.4899, + "step": 18802 + }, + { + "epoch": 0.56, + "learning_rate": 8.554382907335813e-05, + "loss": 0.5547, + "step": 18803 + }, + { + "epoch": 0.56, + "learning_rate": 8.553428801938723e-05, + "loss": 0.5178, + "step": 18804 + }, + { + "epoch": 0.56, + "learning_rate": 8.552474709991281e-05, + "loss": 0.4465, + "step": 18805 + }, + { + "epoch": 0.56, + "learning_rate": 8.55152063150235e-05, + "loss": 0.6149, + "step": 18806 + }, + { + "epoch": 0.56, + "learning_rate": 8.550566566480797e-05, + "loss": 0.495, + "step": 18807 + }, + { + "epoch": 0.56, + "learning_rate": 8.549612514935504e-05, + "loss": 0.5128, + "step": 18808 + }, + { + "epoch": 0.56, + "learning_rate": 8.548658476875337e-05, + "loss": 0.6156, + "step": 18809 + }, + { + "epoch": 0.56, + "learning_rate": 8.547704452309162e-05, + "loss": 0.5483, + "step": 18810 + }, + { + "epoch": 0.56, + "learning_rate": 8.546750441245853e-05, + "loss": 0.5226, + "step": 18811 + }, + { + "epoch": 0.56, + "learning_rate": 8.545796443694277e-05, + "loss": 0.439, + "step": 18812 + }, + { + "epoch": 0.56, + "learning_rate": 8.544842459663312e-05, + "loss": 0.5155, + "step": 18813 + }, + { + "epoch": 0.56, + "learning_rate": 8.543888489161817e-05, + "loss": 0.5775, + "step": 18814 + }, + { + "epoch": 0.56, + "learning_rate": 8.542934532198667e-05, + "loss": 0.5311, + "step": 18815 + }, + { + "epoch": 0.56, + "learning_rate": 8.54198058878273e-05, + "loss": 0.6145, + "step": 18816 + }, + { + "epoch": 0.56, + "learning_rate": 8.54102665892288e-05, + "loss": 0.619, + "step": 18817 + }, + { + "epoch": 0.56, + "learning_rate": 8.540072742627982e-05, + "loss": 0.4645, + "step": 18818 + }, + { + "epoch": 0.56, + "learning_rate": 8.5391188399069e-05, + "loss": 0.5669, + "step": 18819 + }, + { + "epoch": 0.56, + "learning_rate": 8.538164950768514e-05, + "loss": 0.6588, + "step": 18820 + }, + { + "epoch": 0.56, + "learning_rate": 8.537211075221691e-05, + "loss": 0.6376, + "step": 18821 + }, + { + "epoch": 0.56, + "learning_rate": 8.536257213275289e-05, + "loss": 0.5879, + "step": 18822 + }, + { + "epoch": 0.56, + "learning_rate": 8.535303364938187e-05, + "loss": 0.5627, + "step": 18823 + }, + { + "epoch": 0.56, + "learning_rate": 8.534349530219252e-05, + "loss": 0.5737, + "step": 18824 + }, + { + "epoch": 0.56, + "learning_rate": 8.533395709127352e-05, + "loss": 0.536, + "step": 18825 + }, + { + "epoch": 0.56, + "learning_rate": 8.53244190167135e-05, + "loss": 0.5128, + "step": 18826 + }, + { + "epoch": 0.56, + "learning_rate": 8.531488107860121e-05, + "loss": 0.4671, + "step": 18827 + }, + { + "epoch": 0.56, + "learning_rate": 8.53053432770253e-05, + "loss": 0.5694, + "step": 18828 + }, + { + "epoch": 0.56, + "learning_rate": 8.529580561207445e-05, + "loss": 0.5103, + "step": 18829 + }, + { + "epoch": 0.56, + "learning_rate": 8.528626808383732e-05, + "loss": 0.5582, + "step": 18830 + }, + { + "epoch": 0.56, + "learning_rate": 8.527673069240261e-05, + "loss": 0.5451, + "step": 18831 + }, + { + "epoch": 0.56, + "learning_rate": 8.526719343785901e-05, + "loss": 0.5424, + "step": 18832 + }, + { + "epoch": 0.56, + "learning_rate": 8.525765632029516e-05, + "loss": 0.4914, + "step": 18833 + }, + { + "epoch": 0.56, + "learning_rate": 8.524811933979976e-05, + "loss": 0.5506, + "step": 18834 + }, + { + "epoch": 0.56, + "learning_rate": 8.523858249646144e-05, + "loss": 0.6122, + "step": 18835 + }, + { + "epoch": 0.56, + "learning_rate": 8.522904579036894e-05, + "loss": 0.5323, + "step": 18836 + }, + { + "epoch": 0.56, + "learning_rate": 8.521950922161085e-05, + "loss": 0.6256, + "step": 18837 + }, + { + "epoch": 0.56, + "learning_rate": 8.520997279027587e-05, + "loss": 0.5988, + "step": 18838 + }, + { + "epoch": 0.56, + "learning_rate": 8.520043649645267e-05, + "loss": 0.5737, + "step": 18839 + }, + { + "epoch": 0.56, + "learning_rate": 8.519090034022994e-05, + "loss": 0.6095, + "step": 18840 + }, + { + "epoch": 0.56, + "learning_rate": 8.518136432169629e-05, + "loss": 0.5336, + "step": 18841 + }, + { + "epoch": 0.56, + "learning_rate": 8.517182844094042e-05, + "loss": 0.5139, + "step": 18842 + }, + { + "epoch": 0.56, + "learning_rate": 8.516229269805095e-05, + "loss": 0.6171, + "step": 18843 + }, + { + "epoch": 0.56, + "learning_rate": 8.515275709311665e-05, + "loss": 0.545, + "step": 18844 + }, + { + "epoch": 0.56, + "learning_rate": 8.5143221626226e-05, + "loss": 0.4386, + "step": 18845 + }, + { + "epoch": 0.56, + "learning_rate": 8.51336862974678e-05, + "loss": 0.5429, + "step": 18846 + }, + { + "epoch": 0.56, + "learning_rate": 8.512415110693063e-05, + "loss": 0.481, + "step": 18847 + }, + { + "epoch": 0.56, + "learning_rate": 8.511461605470323e-05, + "loss": 0.6894, + "step": 18848 + }, + { + "epoch": 0.56, + "learning_rate": 8.510508114087418e-05, + "loss": 0.6328, + "step": 18849 + }, + { + "epoch": 0.56, + "learning_rate": 8.50955463655321e-05, + "loss": 0.4689, + "step": 18850 + }, + { + "epoch": 0.56, + "learning_rate": 8.508601172876574e-05, + "loss": 0.5451, + "step": 18851 + }, + { + "epoch": 0.56, + "learning_rate": 8.50764772306637e-05, + "loss": 0.5605, + "step": 18852 + }, + { + "epoch": 0.56, + "learning_rate": 8.50669428713146e-05, + "loss": 0.539, + "step": 18853 + }, + { + "epoch": 0.56, + "learning_rate": 8.505740865080712e-05, + "loss": 0.4968, + "step": 18854 + }, + { + "epoch": 0.56, + "learning_rate": 8.504787456922988e-05, + "loss": 0.5113, + "step": 18855 + }, + { + "epoch": 0.56, + "learning_rate": 8.503834062667161e-05, + "loss": 0.5446, + "step": 18856 + }, + { + "epoch": 0.56, + "learning_rate": 8.502880682322082e-05, + "loss": 0.463, + "step": 18857 + }, + { + "epoch": 0.56, + "learning_rate": 8.501927315896624e-05, + "loss": 0.6078, + "step": 18858 + }, + { + "epoch": 0.56, + "learning_rate": 8.500973963399646e-05, + "loss": 0.5594, + "step": 18859 + }, + { + "epoch": 0.56, + "learning_rate": 8.50002062484002e-05, + "loss": 0.5356, + "step": 18860 + }, + { + "epoch": 0.56, + "learning_rate": 8.499067300226602e-05, + "loss": 0.5509, + "step": 18861 + }, + { + "epoch": 0.56, + "learning_rate": 8.498113989568255e-05, + "loss": 0.5749, + "step": 18862 + }, + { + "epoch": 0.56, + "learning_rate": 8.49716069287385e-05, + "loss": 0.6465, + "step": 18863 + }, + { + "epoch": 0.56, + "learning_rate": 8.496207410152246e-05, + "loss": 0.5443, + "step": 18864 + }, + { + "epoch": 0.56, + "learning_rate": 8.495254141412304e-05, + "loss": 0.6167, + "step": 18865 + }, + { + "epoch": 0.56, + "learning_rate": 8.494300886662887e-05, + "loss": 0.4919, + "step": 18866 + }, + { + "epoch": 0.56, + "learning_rate": 8.493347645912866e-05, + "loss": 0.585, + "step": 18867 + }, + { + "epoch": 0.56, + "learning_rate": 8.492394419171094e-05, + "loss": 0.6279, + "step": 18868 + }, + { + "epoch": 0.56, + "learning_rate": 8.491441206446437e-05, + "loss": 0.3775, + "step": 18869 + }, + { + "epoch": 0.56, + "learning_rate": 8.490488007747761e-05, + "loss": 0.5651, + "step": 18870 + }, + { + "epoch": 0.56, + "learning_rate": 8.489534823083927e-05, + "loss": 0.6572, + "step": 18871 + }, + { + "epoch": 0.56, + "learning_rate": 8.488581652463791e-05, + "loss": 0.5252, + "step": 18872 + }, + { + "epoch": 0.56, + "learning_rate": 8.487628495896225e-05, + "loss": 0.6007, + "step": 18873 + }, + { + "epoch": 0.56, + "learning_rate": 8.486675353390083e-05, + "loss": 0.5427, + "step": 18874 + }, + { + "epoch": 0.56, + "learning_rate": 8.485722224954237e-05, + "loss": 0.5139, + "step": 18875 + }, + { + "epoch": 0.56, + "learning_rate": 8.484769110597535e-05, + "loss": 0.6151, + "step": 18876 + }, + { + "epoch": 0.56, + "learning_rate": 8.483816010328849e-05, + "loss": 0.513, + "step": 18877 + }, + { + "epoch": 0.56, + "learning_rate": 8.482862924157036e-05, + "loss": 0.5028, + "step": 18878 + }, + { + "epoch": 0.56, + "learning_rate": 8.481909852090961e-05, + "loss": 0.5362, + "step": 18879 + }, + { + "epoch": 0.56, + "learning_rate": 8.480956794139484e-05, + "loss": 0.4949, + "step": 18880 + }, + { + "epoch": 0.56, + "learning_rate": 8.48000375031146e-05, + "loss": 0.5443, + "step": 18881 + }, + { + "epoch": 0.56, + "learning_rate": 8.479050720615759e-05, + "loss": 0.4569, + "step": 18882 + }, + { + "epoch": 0.56, + "learning_rate": 8.47809770506124e-05, + "loss": 0.5868, + "step": 18883 + }, + { + "epoch": 0.56, + "learning_rate": 8.477144703656757e-05, + "loss": 0.5189, + "step": 18884 + }, + { + "epoch": 0.56, + "learning_rate": 8.47619171641118e-05, + "loss": 0.4993, + "step": 18885 + }, + { + "epoch": 0.56, + "learning_rate": 8.47523874333336e-05, + "loss": 0.5206, + "step": 18886 + }, + { + "epoch": 0.56, + "learning_rate": 8.474285784432171e-05, + "loss": 0.6081, + "step": 18887 + }, + { + "epoch": 0.56, + "learning_rate": 8.473332839716458e-05, + "loss": 0.5366, + "step": 18888 + }, + { + "epoch": 0.56, + "learning_rate": 8.472379909195089e-05, + "loss": 0.53, + "step": 18889 + }, + { + "epoch": 0.56, + "learning_rate": 8.471426992876921e-05, + "loss": 0.6087, + "step": 18890 + }, + { + "epoch": 0.56, + "learning_rate": 8.47047409077082e-05, + "loss": 0.4501, + "step": 18891 + }, + { + "epoch": 0.56, + "learning_rate": 8.46952120288564e-05, + "loss": 0.5348, + "step": 18892 + }, + { + "epoch": 0.56, + "learning_rate": 8.468568329230238e-05, + "loss": 0.541, + "step": 18893 + }, + { + "epoch": 0.56, + "learning_rate": 8.467615469813484e-05, + "loss": 0.4768, + "step": 18894 + }, + { + "epoch": 0.56, + "learning_rate": 8.466662624644226e-05, + "loss": 0.5491, + "step": 18895 + }, + { + "epoch": 0.56, + "learning_rate": 8.465709793731327e-05, + "loss": 0.5483, + "step": 18896 + }, + { + "epoch": 0.56, + "learning_rate": 8.464756977083649e-05, + "loss": 0.441, + "step": 18897 + }, + { + "epoch": 0.56, + "learning_rate": 8.46380417471005e-05, + "loss": 0.488, + "step": 18898 + }, + { + "epoch": 0.56, + "learning_rate": 8.462851386619386e-05, + "loss": 0.6755, + "step": 18899 + }, + { + "epoch": 0.56, + "learning_rate": 8.461898612820514e-05, + "loss": 0.5021, + "step": 18900 + }, + { + "epoch": 0.56, + "learning_rate": 8.460945853322299e-05, + "loss": 0.5767, + "step": 18901 + }, + { + "epoch": 0.56, + "learning_rate": 8.4599931081336e-05, + "loss": 0.523, + "step": 18902 + }, + { + "epoch": 0.56, + "learning_rate": 8.459040377263263e-05, + "loss": 0.5195, + "step": 18903 + }, + { + "epoch": 0.56, + "learning_rate": 8.458087660720159e-05, + "loss": 0.4787, + "step": 18904 + }, + { + "epoch": 0.56, + "learning_rate": 8.457134958513139e-05, + "loss": 0.5182, + "step": 18905 + }, + { + "epoch": 0.56, + "learning_rate": 8.456182270651067e-05, + "loss": 0.6029, + "step": 18906 + }, + { + "epoch": 0.56, + "learning_rate": 8.455229597142795e-05, + "loss": 0.5572, + "step": 18907 + }, + { + "epoch": 0.56, + "learning_rate": 8.454276937997182e-05, + "loss": 0.5387, + "step": 18908 + }, + { + "epoch": 0.56, + "learning_rate": 8.453324293223085e-05, + "loss": 0.5643, + "step": 18909 + }, + { + "epoch": 0.56, + "learning_rate": 8.452371662829366e-05, + "loss": 0.5068, + "step": 18910 + }, + { + "epoch": 0.56, + "learning_rate": 8.451419046824876e-05, + "loss": 0.5112, + "step": 18911 + }, + { + "epoch": 0.56, + "learning_rate": 8.450466445218472e-05, + "loss": 0.5121, + "step": 18912 + }, + { + "epoch": 0.56, + "learning_rate": 8.449513858019017e-05, + "loss": 0.6823, + "step": 18913 + }, + { + "epoch": 0.56, + "learning_rate": 8.448561285235365e-05, + "loss": 0.4922, + "step": 18914 + }, + { + "epoch": 0.56, + "learning_rate": 8.447608726876368e-05, + "loss": 0.5485, + "step": 18915 + }, + { + "epoch": 0.56, + "learning_rate": 8.446656182950889e-05, + "loss": 0.5065, + "step": 18916 + }, + { + "epoch": 0.56, + "learning_rate": 8.445703653467778e-05, + "loss": 0.5858, + "step": 18917 + }, + { + "epoch": 0.56, + "learning_rate": 8.444751138435904e-05, + "loss": 0.5496, + "step": 18918 + }, + { + "epoch": 0.56, + "learning_rate": 8.443798637864106e-05, + "loss": 0.589, + "step": 18919 + }, + { + "epoch": 0.56, + "learning_rate": 8.44284615176125e-05, + "loss": 0.5388, + "step": 18920 + }, + { + "epoch": 0.56, + "learning_rate": 8.441893680136187e-05, + "loss": 0.6213, + "step": 18921 + }, + { + "epoch": 0.56, + "learning_rate": 8.440941222997781e-05, + "loss": 0.5011, + "step": 18922 + }, + { + "epoch": 0.56, + "learning_rate": 8.439988780354879e-05, + "loss": 0.4745, + "step": 18923 + }, + { + "epoch": 0.56, + "learning_rate": 8.439036352216339e-05, + "loss": 0.4883, + "step": 18924 + }, + { + "epoch": 0.56, + "learning_rate": 8.438083938591019e-05, + "loss": 0.582, + "step": 18925 + }, + { + "epoch": 0.56, + "learning_rate": 8.437131539487773e-05, + "loss": 0.621, + "step": 18926 + }, + { + "epoch": 0.56, + "learning_rate": 8.436179154915451e-05, + "loss": 0.5845, + "step": 18927 + }, + { + "epoch": 0.56, + "learning_rate": 8.435226784882914e-05, + "loss": 0.5323, + "step": 18928 + }, + { + "epoch": 0.56, + "learning_rate": 8.434274429399017e-05, + "loss": 0.5421, + "step": 18929 + }, + { + "epoch": 0.56, + "learning_rate": 8.43332208847261e-05, + "loss": 0.5731, + "step": 18930 + }, + { + "epoch": 0.56, + "learning_rate": 8.432369762112548e-05, + "loss": 0.5655, + "step": 18931 + }, + { + "epoch": 0.56, + "learning_rate": 8.431417450327689e-05, + "loss": 0.5077, + "step": 18932 + }, + { + "epoch": 0.56, + "learning_rate": 8.430465153126889e-05, + "loss": 0.5057, + "step": 18933 + }, + { + "epoch": 0.56, + "learning_rate": 8.429512870518992e-05, + "loss": 0.613, + "step": 18934 + }, + { + "epoch": 0.56, + "learning_rate": 8.428560602512862e-05, + "loss": 0.5193, + "step": 18935 + }, + { + "epoch": 0.56, + "learning_rate": 8.427608349117347e-05, + "loss": 0.5398, + "step": 18936 + }, + { + "epoch": 0.56, + "learning_rate": 8.426656110341305e-05, + "loss": 0.4689, + "step": 18937 + }, + { + "epoch": 0.56, + "learning_rate": 8.425703886193588e-05, + "loss": 0.4767, + "step": 18938 + }, + { + "epoch": 0.56, + "learning_rate": 8.424751676683047e-05, + "loss": 0.4844, + "step": 18939 + }, + { + "epoch": 0.56, + "learning_rate": 8.423799481818536e-05, + "loss": 0.5286, + "step": 18940 + }, + { + "epoch": 0.56, + "learning_rate": 8.422847301608913e-05, + "loss": 0.5437, + "step": 18941 + }, + { + "epoch": 0.56, + "learning_rate": 8.421895136063025e-05, + "loss": 0.5173, + "step": 18942 + }, + { + "epoch": 0.56, + "learning_rate": 8.420942985189726e-05, + "loss": 0.6891, + "step": 18943 + }, + { + "epoch": 0.56, + "learning_rate": 8.41999084899787e-05, + "loss": 0.4968, + "step": 18944 + }, + { + "epoch": 0.56, + "learning_rate": 8.419038727496313e-05, + "loss": 0.6502, + "step": 18945 + }, + { + "epoch": 0.56, + "learning_rate": 8.418086620693899e-05, + "loss": 0.585, + "step": 18946 + }, + { + "epoch": 0.56, + "learning_rate": 8.417134528599487e-05, + "loss": 0.5497, + "step": 18947 + }, + { + "epoch": 0.56, + "learning_rate": 8.416182451221925e-05, + "loss": 0.5601, + "step": 18948 + }, + { + "epoch": 0.56, + "learning_rate": 8.415230388570075e-05, + "loss": 0.4702, + "step": 18949 + }, + { + "epoch": 0.56, + "learning_rate": 8.414278340652774e-05, + "loss": 0.5027, + "step": 18950 + }, + { + "epoch": 0.56, + "learning_rate": 8.413326307478883e-05, + "loss": 0.5441, + "step": 18951 + }, + { + "epoch": 0.56, + "learning_rate": 8.41237428905725e-05, + "loss": 0.4997, + "step": 18952 + }, + { + "epoch": 0.56, + "learning_rate": 8.411422285396732e-05, + "loss": 0.5783, + "step": 18953 + }, + { + "epoch": 0.56, + "learning_rate": 8.410470296506174e-05, + "loss": 0.4283, + "step": 18954 + }, + { + "epoch": 0.56, + "learning_rate": 8.409518322394428e-05, + "loss": 0.5267, + "step": 18955 + }, + { + "epoch": 0.56, + "learning_rate": 8.408566363070348e-05, + "loss": 0.5255, + "step": 18956 + }, + { + "epoch": 0.56, + "learning_rate": 8.407614418542787e-05, + "loss": 0.6623, + "step": 18957 + }, + { + "epoch": 0.56, + "learning_rate": 8.406662488820587e-05, + "loss": 0.4401, + "step": 18958 + }, + { + "epoch": 0.56, + "learning_rate": 8.405710573912607e-05, + "loss": 0.4558, + "step": 18959 + }, + { + "epoch": 0.56, + "learning_rate": 8.404758673827697e-05, + "loss": 0.525, + "step": 18960 + }, + { + "epoch": 0.56, + "learning_rate": 8.403806788574703e-05, + "loss": 0.5017, + "step": 18961 + }, + { + "epoch": 0.56, + "learning_rate": 8.402854918162475e-05, + "loss": 0.5198, + "step": 18962 + }, + { + "epoch": 0.56, + "learning_rate": 8.401903062599867e-05, + "loss": 0.5921, + "step": 18963 + }, + { + "epoch": 0.56, + "learning_rate": 8.40095122189573e-05, + "loss": 0.5144, + "step": 18964 + }, + { + "epoch": 0.56, + "learning_rate": 8.399999396058907e-05, + "loss": 0.4353, + "step": 18965 + }, + { + "epoch": 0.56, + "learning_rate": 8.399047585098253e-05, + "loss": 0.5152, + "step": 18966 + }, + { + "epoch": 0.56, + "learning_rate": 8.398095789022616e-05, + "loss": 0.6685, + "step": 18967 + }, + { + "epoch": 0.56, + "learning_rate": 8.397144007840849e-05, + "loss": 0.5309, + "step": 18968 + }, + { + "epoch": 0.56, + "learning_rate": 8.396192241561797e-05, + "loss": 0.6364, + "step": 18969 + }, + { + "epoch": 0.56, + "learning_rate": 8.39524049019431e-05, + "loss": 0.4824, + "step": 18970 + }, + { + "epoch": 0.56, + "learning_rate": 8.394288753747235e-05, + "loss": 0.5191, + "step": 18971 + }, + { + "epoch": 0.56, + "learning_rate": 8.393337032229427e-05, + "loss": 0.5352, + "step": 18972 + }, + { + "epoch": 0.56, + "learning_rate": 8.39238532564973e-05, + "loss": 0.5482, + "step": 18973 + }, + { + "epoch": 0.56, + "learning_rate": 8.391433634016989e-05, + "loss": 0.5663, + "step": 18974 + }, + { + "epoch": 0.56, + "learning_rate": 8.390481957340062e-05, + "loss": 0.6368, + "step": 18975 + }, + { + "epoch": 0.56, + "learning_rate": 8.389530295627793e-05, + "loss": 0.5946, + "step": 18976 + }, + { + "epoch": 0.56, + "learning_rate": 8.388578648889025e-05, + "loss": 0.5316, + "step": 18977 + }, + { + "epoch": 0.57, + "learning_rate": 8.387627017132612e-05, + "loss": 0.5646, + "step": 18978 + }, + { + "epoch": 0.57, + "learning_rate": 8.386675400367399e-05, + "loss": 0.5201, + "step": 18979 + }, + { + "epoch": 0.57, + "learning_rate": 8.385723798602242e-05, + "loss": 0.6054, + "step": 18980 + }, + { + "epoch": 0.57, + "learning_rate": 8.384772211845975e-05, + "loss": 0.6071, + "step": 18981 + }, + { + "epoch": 0.57, + "learning_rate": 8.383820640107454e-05, + "loss": 0.571, + "step": 18982 + }, + { + "epoch": 0.57, + "learning_rate": 8.382869083395522e-05, + "loss": 0.5541, + "step": 18983 + }, + { + "epoch": 0.57, + "learning_rate": 8.381917541719034e-05, + "loss": 0.532, + "step": 18984 + }, + { + "epoch": 0.57, + "learning_rate": 8.380966015086829e-05, + "loss": 0.6147, + "step": 18985 + }, + { + "epoch": 0.57, + "learning_rate": 8.380014503507757e-05, + "loss": 0.5628, + "step": 18986 + }, + { + "epoch": 0.57, + "learning_rate": 8.379063006990664e-05, + "loss": 0.6113, + "step": 18987 + }, + { + "epoch": 0.57, + "learning_rate": 8.378111525544399e-05, + "loss": 0.5649, + "step": 18988 + }, + { + "epoch": 0.57, + "learning_rate": 8.377160059177803e-05, + "loss": 0.5502, + "step": 18989 + }, + { + "epoch": 0.57, + "learning_rate": 8.37620860789973e-05, + "loss": 0.4949, + "step": 18990 + }, + { + "epoch": 0.57, + "learning_rate": 8.37525717171902e-05, + "loss": 0.6563, + "step": 18991 + }, + { + "epoch": 0.57, + "learning_rate": 8.374305750644523e-05, + "loss": 0.5434, + "step": 18992 + }, + { + "epoch": 0.57, + "learning_rate": 8.37335434468508e-05, + "loss": 0.4636, + "step": 18993 + }, + { + "epoch": 0.57, + "learning_rate": 8.372402953849542e-05, + "loss": 0.5418, + "step": 18994 + }, + { + "epoch": 0.57, + "learning_rate": 8.371451578146754e-05, + "loss": 0.5346, + "step": 18995 + }, + { + "epoch": 0.57, + "learning_rate": 8.370500217585558e-05, + "loss": 0.6058, + "step": 18996 + }, + { + "epoch": 0.57, + "learning_rate": 8.369548872174801e-05, + "loss": 0.519, + "step": 18997 + }, + { + "epoch": 0.57, + "learning_rate": 8.368597541923328e-05, + "loss": 0.5594, + "step": 18998 + }, + { + "epoch": 0.57, + "learning_rate": 8.367646226839989e-05, + "loss": 0.6026, + "step": 18999 + }, + { + "epoch": 0.57, + "learning_rate": 8.366694926933623e-05, + "loss": 0.609, + "step": 19000 + }, + { + "epoch": 0.57, + "learning_rate": 8.365743642213075e-05, + "loss": 0.6245, + "step": 19001 + }, + { + "epoch": 0.57, + "learning_rate": 8.364792372687191e-05, + "loss": 0.6861, + "step": 19002 + }, + { + "epoch": 0.57, + "learning_rate": 8.36384111836482e-05, + "loss": 0.6563, + "step": 19003 + }, + { + "epoch": 0.57, + "learning_rate": 8.362889879254799e-05, + "loss": 0.4545, + "step": 19004 + }, + { + "epoch": 0.57, + "learning_rate": 8.361938655365973e-05, + "loss": 0.5077, + "step": 19005 + }, + { + "epoch": 0.57, + "learning_rate": 8.360987446707192e-05, + "loss": 0.6132, + "step": 19006 + }, + { + "epoch": 0.57, + "learning_rate": 8.360036253287298e-05, + "loss": 0.5301, + "step": 19007 + }, + { + "epoch": 0.57, + "learning_rate": 8.359085075115128e-05, + "loss": 0.6039, + "step": 19008 + }, + { + "epoch": 0.57, + "learning_rate": 8.358133912199535e-05, + "loss": 0.4985, + "step": 19009 + }, + { + "epoch": 0.57, + "learning_rate": 8.357182764549355e-05, + "loss": 0.5087, + "step": 19010 + }, + { + "epoch": 0.57, + "learning_rate": 8.356231632173441e-05, + "loss": 0.4654, + "step": 19011 + }, + { + "epoch": 0.57, + "learning_rate": 8.355280515080626e-05, + "loss": 0.6449, + "step": 19012 + }, + { + "epoch": 0.57, + "learning_rate": 8.354329413279758e-05, + "loss": 0.7134, + "step": 19013 + }, + { + "epoch": 0.57, + "learning_rate": 8.353378326779676e-05, + "loss": 0.3932, + "step": 19014 + }, + { + "epoch": 0.57, + "learning_rate": 8.352427255589232e-05, + "loss": 0.5055, + "step": 19015 + }, + { + "epoch": 0.57, + "learning_rate": 8.351476199717258e-05, + "loss": 0.5743, + "step": 19016 + }, + { + "epoch": 0.57, + "learning_rate": 8.350525159172602e-05, + "loss": 0.4546, + "step": 19017 + }, + { + "epoch": 0.57, + "learning_rate": 8.349574133964107e-05, + "loss": 0.4246, + "step": 19018 + }, + { + "epoch": 0.57, + "learning_rate": 8.348623124100617e-05, + "loss": 0.4266, + "step": 19019 + }, + { + "epoch": 0.57, + "learning_rate": 8.347672129590965e-05, + "loss": 0.5195, + "step": 19020 + }, + { + "epoch": 0.57, + "learning_rate": 8.346721150444001e-05, + "loss": 0.5477, + "step": 19021 + }, + { + "epoch": 0.57, + "learning_rate": 8.345770186668563e-05, + "loss": 0.4893, + "step": 19022 + }, + { + "epoch": 0.57, + "learning_rate": 8.344819238273502e-05, + "loss": 0.56, + "step": 19023 + }, + { + "epoch": 0.57, + "learning_rate": 8.343868305267644e-05, + "loss": 0.5481, + "step": 19024 + }, + { + "epoch": 0.57, + "learning_rate": 8.342917387659841e-05, + "loss": 0.5265, + "step": 19025 + }, + { + "epoch": 0.57, + "learning_rate": 8.341966485458935e-05, + "loss": 0.5541, + "step": 19026 + }, + { + "epoch": 0.57, + "learning_rate": 8.341015598673758e-05, + "loss": 0.6361, + "step": 19027 + }, + { + "epoch": 0.57, + "learning_rate": 8.34006472731316e-05, + "loss": 0.5436, + "step": 19028 + }, + { + "epoch": 0.57, + "learning_rate": 8.339113871385976e-05, + "loss": 0.5732, + "step": 19029 + }, + { + "epoch": 0.57, + "learning_rate": 8.338163030901054e-05, + "loss": 0.519, + "step": 19030 + }, + { + "epoch": 0.57, + "learning_rate": 8.337212205867228e-05, + "loss": 0.7082, + "step": 19031 + }, + { + "epoch": 0.57, + "learning_rate": 8.336261396293337e-05, + "loss": 0.5992, + "step": 19032 + }, + { + "epoch": 0.57, + "learning_rate": 8.335310602188226e-05, + "loss": 0.5414, + "step": 19033 + }, + { + "epoch": 0.57, + "learning_rate": 8.334359823560738e-05, + "loss": 0.5668, + "step": 19034 + }, + { + "epoch": 0.57, + "learning_rate": 8.333409060419704e-05, + "loss": 0.4659, + "step": 19035 + }, + { + "epoch": 0.57, + "learning_rate": 8.332458312773968e-05, + "loss": 0.5508, + "step": 19036 + }, + { + "epoch": 0.57, + "learning_rate": 8.33150758063237e-05, + "loss": 0.5671, + "step": 19037 + }, + { + "epoch": 0.57, + "learning_rate": 8.330556864003754e-05, + "loss": 0.5478, + "step": 19038 + }, + { + "epoch": 0.57, + "learning_rate": 8.329606162896949e-05, + "loss": 0.5505, + "step": 19039 + }, + { + "epoch": 0.57, + "learning_rate": 8.328655477320802e-05, + "loss": 0.4829, + "step": 19040 + }, + { + "epoch": 0.57, + "learning_rate": 8.327704807284149e-05, + "loss": 0.5972, + "step": 19041 + }, + { + "epoch": 0.57, + "learning_rate": 8.326754152795837e-05, + "loss": 0.5648, + "step": 19042 + }, + { + "epoch": 0.57, + "learning_rate": 8.325803513864691e-05, + "loss": 0.558, + "step": 19043 + }, + { + "epoch": 0.57, + "learning_rate": 8.324852890499559e-05, + "loss": 0.5311, + "step": 19044 + }, + { + "epoch": 0.57, + "learning_rate": 8.323902282709275e-05, + "loss": 0.5324, + "step": 19045 + }, + { + "epoch": 0.57, + "learning_rate": 8.322951690502684e-05, + "loss": 0.539, + "step": 19046 + }, + { + "epoch": 0.57, + "learning_rate": 8.322001113888618e-05, + "loss": 0.5688, + "step": 19047 + }, + { + "epoch": 0.57, + "learning_rate": 8.321050552875912e-05, + "loss": 0.494, + "step": 19048 + }, + { + "epoch": 0.57, + "learning_rate": 8.320100007473414e-05, + "loss": 0.5404, + "step": 19049 + }, + { + "epoch": 0.57, + "learning_rate": 8.319149477689958e-05, + "loss": 0.4955, + "step": 19050 + }, + { + "epoch": 0.57, + "learning_rate": 8.318198963534375e-05, + "loss": 0.6189, + "step": 19051 + }, + { + "epoch": 0.57, + "learning_rate": 8.317248465015513e-05, + "loss": 0.5807, + "step": 19052 + }, + { + "epoch": 0.57, + "learning_rate": 8.316297982142198e-05, + "loss": 0.528, + "step": 19053 + }, + { + "epoch": 0.57, + "learning_rate": 8.315347514923282e-05, + "loss": 0.367, + "step": 19054 + }, + { + "epoch": 0.57, + "learning_rate": 8.314397063367586e-05, + "loss": 0.5477, + "step": 19055 + }, + { + "epoch": 0.57, + "learning_rate": 8.313446627483956e-05, + "loss": 0.5125, + "step": 19056 + }, + { + "epoch": 0.57, + "learning_rate": 8.312496207281226e-05, + "loss": 0.6728, + "step": 19057 + }, + { + "epoch": 0.57, + "learning_rate": 8.311545802768238e-05, + "loss": 0.478, + "step": 19058 + }, + { + "epoch": 0.57, + "learning_rate": 8.310595413953821e-05, + "loss": 0.4996, + "step": 19059 + }, + { + "epoch": 0.57, + "learning_rate": 8.309645040846814e-05, + "loss": 0.5234, + "step": 19060 + }, + { + "epoch": 0.57, + "learning_rate": 8.308694683456057e-05, + "loss": 0.7397, + "step": 19061 + }, + { + "epoch": 0.57, + "learning_rate": 8.30774434179038e-05, + "loss": 0.5715, + "step": 19062 + }, + { + "epoch": 0.57, + "learning_rate": 8.30679401585862e-05, + "loss": 0.6432, + "step": 19063 + }, + { + "epoch": 0.57, + "learning_rate": 8.305843705669618e-05, + "loss": 0.5302, + "step": 19064 + }, + { + "epoch": 0.57, + "learning_rate": 8.304893411232206e-05, + "loss": 0.5026, + "step": 19065 + }, + { + "epoch": 0.57, + "learning_rate": 8.303943132555218e-05, + "loss": 0.5343, + "step": 19066 + }, + { + "epoch": 0.57, + "learning_rate": 8.302992869647487e-05, + "loss": 0.4672, + "step": 19067 + }, + { + "epoch": 0.57, + "learning_rate": 8.302042622517857e-05, + "loss": 0.5691, + "step": 19068 + }, + { + "epoch": 0.57, + "learning_rate": 8.30109239117516e-05, + "loss": 0.5993, + "step": 19069 + }, + { + "epoch": 0.57, + "learning_rate": 8.300142175628223e-05, + "loss": 0.5361, + "step": 19070 + }, + { + "epoch": 0.57, + "learning_rate": 8.29919197588589e-05, + "loss": 0.5442, + "step": 19071 + }, + { + "epoch": 0.57, + "learning_rate": 8.298241791956988e-05, + "loss": 0.561, + "step": 19072 + }, + { + "epoch": 0.57, + "learning_rate": 8.297291623850361e-05, + "loss": 0.4794, + "step": 19073 + }, + { + "epoch": 0.57, + "learning_rate": 8.296341471574835e-05, + "loss": 0.5747, + "step": 19074 + }, + { + "epoch": 0.57, + "learning_rate": 8.295391335139249e-05, + "loss": 0.475, + "step": 19075 + }, + { + "epoch": 0.57, + "learning_rate": 8.29444121455243e-05, + "loss": 0.5081, + "step": 19076 + }, + { + "epoch": 0.57, + "learning_rate": 8.293491109823223e-05, + "loss": 0.5292, + "step": 19077 + }, + { + "epoch": 0.57, + "learning_rate": 8.292541020960453e-05, + "loss": 0.5409, + "step": 19078 + }, + { + "epoch": 0.57, + "learning_rate": 8.291590947972954e-05, + "loss": 0.5337, + "step": 19079 + }, + { + "epoch": 0.57, + "learning_rate": 8.290640890869564e-05, + "loss": 0.6316, + "step": 19080 + }, + { + "epoch": 0.57, + "learning_rate": 8.289690849659115e-05, + "loss": 0.5362, + "step": 19081 + }, + { + "epoch": 0.57, + "learning_rate": 8.288740824350434e-05, + "loss": 0.5377, + "step": 19082 + }, + { + "epoch": 0.57, + "learning_rate": 8.287790814952363e-05, + "loss": 0.541, + "step": 19083 + }, + { + "epoch": 0.57, + "learning_rate": 8.286840821473727e-05, + "loss": 0.5696, + "step": 19084 + }, + { + "epoch": 0.57, + "learning_rate": 8.28589084392337e-05, + "loss": 0.5057, + "step": 19085 + }, + { + "epoch": 0.57, + "learning_rate": 8.284940882310108e-05, + "loss": 0.3938, + "step": 19086 + }, + { + "epoch": 0.57, + "learning_rate": 8.283990936642786e-05, + "loss": 0.5211, + "step": 19087 + }, + { + "epoch": 0.57, + "learning_rate": 8.283041006930229e-05, + "loss": 0.4703, + "step": 19088 + }, + { + "epoch": 0.57, + "learning_rate": 8.282091093181278e-05, + "loss": 0.5011, + "step": 19089 + }, + { + "epoch": 0.57, + "learning_rate": 8.281141195404755e-05, + "loss": 0.499, + "step": 19090 + }, + { + "epoch": 0.57, + "learning_rate": 8.280191313609497e-05, + "loss": 0.5412, + "step": 19091 + }, + { + "epoch": 0.57, + "learning_rate": 8.279241447804337e-05, + "loss": 0.6287, + "step": 19092 + }, + { + "epoch": 0.57, + "learning_rate": 8.278291597998103e-05, + "loss": 0.5109, + "step": 19093 + }, + { + "epoch": 0.57, + "learning_rate": 8.277341764199624e-05, + "loss": 0.5827, + "step": 19094 + }, + { + "epoch": 0.57, + "learning_rate": 8.276391946417738e-05, + "loss": 0.5749, + "step": 19095 + }, + { + "epoch": 0.57, + "learning_rate": 8.275442144661276e-05, + "loss": 0.4702, + "step": 19096 + }, + { + "epoch": 0.57, + "learning_rate": 8.274492358939061e-05, + "loss": 0.5452, + "step": 19097 + }, + { + "epoch": 0.57, + "learning_rate": 8.273542589259928e-05, + "loss": 0.627, + "step": 19098 + }, + { + "epoch": 0.57, + "learning_rate": 8.272592835632709e-05, + "loss": 0.5805, + "step": 19099 + }, + { + "epoch": 0.57, + "learning_rate": 8.271643098066236e-05, + "loss": 0.6093, + "step": 19100 + }, + { + "epoch": 0.57, + "learning_rate": 8.270693376569331e-05, + "loss": 0.5346, + "step": 19101 + }, + { + "epoch": 0.57, + "learning_rate": 8.269743671150833e-05, + "loss": 0.5289, + "step": 19102 + }, + { + "epoch": 0.57, + "learning_rate": 8.268793981819568e-05, + "loss": 0.6345, + "step": 19103 + }, + { + "epoch": 0.57, + "learning_rate": 8.267844308584369e-05, + "loss": 0.6309, + "step": 19104 + }, + { + "epoch": 0.57, + "learning_rate": 8.266894651454063e-05, + "loss": 0.5098, + "step": 19105 + }, + { + "epoch": 0.57, + "learning_rate": 8.265945010437477e-05, + "loss": 0.492, + "step": 19106 + }, + { + "epoch": 0.57, + "learning_rate": 8.264995385543443e-05, + "loss": 0.5333, + "step": 19107 + }, + { + "epoch": 0.57, + "learning_rate": 8.264045776780793e-05, + "loss": 0.5809, + "step": 19108 + }, + { + "epoch": 0.57, + "learning_rate": 8.263096184158353e-05, + "loss": 0.5618, + "step": 19109 + }, + { + "epoch": 0.57, + "learning_rate": 8.26214660768495e-05, + "loss": 0.4824, + "step": 19110 + }, + { + "epoch": 0.57, + "learning_rate": 8.261197047369416e-05, + "loss": 0.575, + "step": 19111 + }, + { + "epoch": 0.57, + "learning_rate": 8.260247503220582e-05, + "loss": 0.4614, + "step": 19112 + }, + { + "epoch": 0.57, + "learning_rate": 8.25929797524727e-05, + "loss": 0.5078, + "step": 19113 + }, + { + "epoch": 0.57, + "learning_rate": 8.258348463458313e-05, + "loss": 0.4957, + "step": 19114 + }, + { + "epoch": 0.57, + "learning_rate": 8.257398967862536e-05, + "loss": 0.4847, + "step": 19115 + }, + { + "epoch": 0.57, + "learning_rate": 8.256449488468773e-05, + "loss": 0.5552, + "step": 19116 + }, + { + "epoch": 0.57, + "learning_rate": 8.255500025285846e-05, + "loss": 0.5303, + "step": 19117 + }, + { + "epoch": 0.57, + "learning_rate": 8.254550578322582e-05, + "loss": 0.6111, + "step": 19118 + }, + { + "epoch": 0.57, + "learning_rate": 8.25360114758781e-05, + "loss": 0.5358, + "step": 19119 + }, + { + "epoch": 0.57, + "learning_rate": 8.252651733090364e-05, + "loss": 0.5348, + "step": 19120 + }, + { + "epoch": 0.57, + "learning_rate": 8.251702334839063e-05, + "loss": 0.5132, + "step": 19121 + }, + { + "epoch": 0.57, + "learning_rate": 8.250752952842735e-05, + "loss": 0.6025, + "step": 19122 + }, + { + "epoch": 0.57, + "learning_rate": 8.24980358711021e-05, + "loss": 0.5402, + "step": 19123 + }, + { + "epoch": 0.57, + "learning_rate": 8.248854237650318e-05, + "loss": 0.6147, + "step": 19124 + }, + { + "epoch": 0.57, + "learning_rate": 8.247904904471875e-05, + "loss": 0.5542, + "step": 19125 + }, + { + "epoch": 0.57, + "learning_rate": 8.246955587583718e-05, + "loss": 0.4416, + "step": 19126 + }, + { + "epoch": 0.57, + "learning_rate": 8.24600628699467e-05, + "loss": 0.5611, + "step": 19127 + }, + { + "epoch": 0.57, + "learning_rate": 8.245057002713555e-05, + "loss": 0.5463, + "step": 19128 + }, + { + "epoch": 0.57, + "learning_rate": 8.244107734749197e-05, + "loss": 0.4636, + "step": 19129 + }, + { + "epoch": 0.57, + "learning_rate": 8.243158483110428e-05, + "loss": 0.5506, + "step": 19130 + }, + { + "epoch": 0.57, + "learning_rate": 8.242209247806074e-05, + "loss": 0.5458, + "step": 19131 + }, + { + "epoch": 0.57, + "learning_rate": 8.241260028844953e-05, + "loss": 0.4505, + "step": 19132 + }, + { + "epoch": 0.57, + "learning_rate": 8.240310826235899e-05, + "loss": 0.5616, + "step": 19133 + }, + { + "epoch": 0.57, + "learning_rate": 8.239361639987729e-05, + "loss": 0.6464, + "step": 19134 + }, + { + "epoch": 0.57, + "learning_rate": 8.238412470109278e-05, + "loss": 0.5412, + "step": 19135 + }, + { + "epoch": 0.57, + "learning_rate": 8.237463316609365e-05, + "loss": 0.6397, + "step": 19136 + }, + { + "epoch": 0.57, + "learning_rate": 8.236514179496814e-05, + "loss": 0.4754, + "step": 19137 + }, + { + "epoch": 0.57, + "learning_rate": 8.23556505878045e-05, + "loss": 0.5497, + "step": 19138 + }, + { + "epoch": 0.57, + "learning_rate": 8.234615954469103e-05, + "loss": 0.6324, + "step": 19139 + }, + { + "epoch": 0.57, + "learning_rate": 8.23366686657159e-05, + "loss": 0.7159, + "step": 19140 + }, + { + "epoch": 0.57, + "learning_rate": 8.232717795096738e-05, + "loss": 0.4737, + "step": 19141 + }, + { + "epoch": 0.57, + "learning_rate": 8.231768740053373e-05, + "loss": 0.5343, + "step": 19142 + }, + { + "epoch": 0.57, + "learning_rate": 8.23081970145032e-05, + "loss": 0.531, + "step": 19143 + }, + { + "epoch": 0.57, + "learning_rate": 8.229870679296395e-05, + "loss": 0.5936, + "step": 19144 + }, + { + "epoch": 0.57, + "learning_rate": 8.228921673600429e-05, + "loss": 0.5786, + "step": 19145 + }, + { + "epoch": 0.57, + "learning_rate": 8.227972684371242e-05, + "loss": 0.4878, + "step": 19146 + }, + { + "epoch": 0.57, + "learning_rate": 8.227023711617665e-05, + "loss": 0.4838, + "step": 19147 + }, + { + "epoch": 0.57, + "learning_rate": 8.22607475534851e-05, + "loss": 0.5206, + "step": 19148 + }, + { + "epoch": 0.57, + "learning_rate": 8.225125815572605e-05, + "loss": 0.5646, + "step": 19149 + }, + { + "epoch": 0.57, + "learning_rate": 8.224176892298771e-05, + "loss": 0.5274, + "step": 19150 + }, + { + "epoch": 0.57, + "learning_rate": 8.223227985535838e-05, + "loss": 0.512, + "step": 19151 + }, + { + "epoch": 0.57, + "learning_rate": 8.222279095292619e-05, + "loss": 0.5377, + "step": 19152 + }, + { + "epoch": 0.57, + "learning_rate": 8.221330221577939e-05, + "loss": 0.5139, + "step": 19153 + }, + { + "epoch": 0.57, + "learning_rate": 8.220381364400624e-05, + "loss": 0.5681, + "step": 19154 + }, + { + "epoch": 0.57, + "learning_rate": 8.219432523769496e-05, + "loss": 0.4557, + "step": 19155 + }, + { + "epoch": 0.57, + "learning_rate": 8.21848369969337e-05, + "loss": 0.5761, + "step": 19156 + }, + { + "epoch": 0.57, + "learning_rate": 8.217534892181076e-05, + "loss": 0.4371, + "step": 19157 + }, + { + "epoch": 0.57, + "learning_rate": 8.216586101241433e-05, + "loss": 0.5919, + "step": 19158 + }, + { + "epoch": 0.57, + "learning_rate": 8.21563732688326e-05, + "loss": 0.4673, + "step": 19159 + }, + { + "epoch": 0.57, + "learning_rate": 8.214688569115377e-05, + "loss": 0.4402, + "step": 19160 + }, + { + "epoch": 0.57, + "learning_rate": 8.213739827946612e-05, + "loss": 0.6516, + "step": 19161 + }, + { + "epoch": 0.57, + "learning_rate": 8.212791103385782e-05, + "loss": 0.3489, + "step": 19162 + }, + { + "epoch": 0.57, + "learning_rate": 8.211842395441705e-05, + "loss": 0.6733, + "step": 19163 + }, + { + "epoch": 0.57, + "learning_rate": 8.210893704123208e-05, + "loss": 0.5799, + "step": 19164 + }, + { + "epoch": 0.57, + "learning_rate": 8.209945029439105e-05, + "loss": 0.5421, + "step": 19165 + }, + { + "epoch": 0.57, + "learning_rate": 8.208996371398223e-05, + "loss": 0.417, + "step": 19166 + }, + { + "epoch": 0.57, + "learning_rate": 8.208047730009378e-05, + "loss": 0.5908, + "step": 19167 + }, + { + "epoch": 0.57, + "learning_rate": 8.20709910528139e-05, + "loss": 0.5386, + "step": 19168 + }, + { + "epoch": 0.57, + "learning_rate": 8.206150497223079e-05, + "loss": 0.5228, + "step": 19169 + }, + { + "epoch": 0.57, + "learning_rate": 8.20520190584327e-05, + "loss": 0.6106, + "step": 19170 + }, + { + "epoch": 0.57, + "learning_rate": 8.204253331150775e-05, + "loss": 0.4657, + "step": 19171 + }, + { + "epoch": 0.57, + "learning_rate": 8.203304773154416e-05, + "loss": 0.4934, + "step": 19172 + }, + { + "epoch": 0.57, + "learning_rate": 8.202356231863013e-05, + "loss": 0.5705, + "step": 19173 + }, + { + "epoch": 0.57, + "learning_rate": 8.20140770728539e-05, + "loss": 0.4809, + "step": 19174 + }, + { + "epoch": 0.57, + "learning_rate": 8.200459199430356e-05, + "loss": 0.5186, + "step": 19175 + }, + { + "epoch": 0.57, + "learning_rate": 8.199510708306737e-05, + "loss": 0.5574, + "step": 19176 + }, + { + "epoch": 0.57, + "learning_rate": 8.198562233923349e-05, + "loss": 0.4657, + "step": 19177 + }, + { + "epoch": 0.57, + "learning_rate": 8.197613776289016e-05, + "loss": 0.472, + "step": 19178 + }, + { + "epoch": 0.57, + "learning_rate": 8.196665335412546e-05, + "loss": 0.5099, + "step": 19179 + }, + { + "epoch": 0.57, + "learning_rate": 8.195716911302766e-05, + "loss": 0.6586, + "step": 19180 + }, + { + "epoch": 0.57, + "learning_rate": 8.194768503968487e-05, + "loss": 0.4688, + "step": 19181 + }, + { + "epoch": 0.57, + "learning_rate": 8.193820113418537e-05, + "loss": 0.6026, + "step": 19182 + }, + { + "epoch": 0.57, + "learning_rate": 8.192871739661726e-05, + "loss": 0.582, + "step": 19183 + }, + { + "epoch": 0.57, + "learning_rate": 8.191923382706868e-05, + "loss": 0.4608, + "step": 19184 + }, + { + "epoch": 0.57, + "learning_rate": 8.19097504256279e-05, + "loss": 0.6049, + "step": 19185 + }, + { + "epoch": 0.57, + "learning_rate": 8.190026719238309e-05, + "loss": 0.5287, + "step": 19186 + }, + { + "epoch": 0.57, + "learning_rate": 8.189078412742231e-05, + "loss": 0.5468, + "step": 19187 + }, + { + "epoch": 0.57, + "learning_rate": 8.188130123083384e-05, + "loss": 0.5285, + "step": 19188 + }, + { + "epoch": 0.57, + "learning_rate": 8.187181850270583e-05, + "loss": 0.5839, + "step": 19189 + }, + { + "epoch": 0.57, + "learning_rate": 8.18623359431264e-05, + "loss": 0.5364, + "step": 19190 + }, + { + "epoch": 0.57, + "learning_rate": 8.18528535521837e-05, + "loss": 0.5446, + "step": 19191 + }, + { + "epoch": 0.57, + "learning_rate": 8.184337132996598e-05, + "loss": 0.5807, + "step": 19192 + }, + { + "epoch": 0.57, + "learning_rate": 8.183388927656138e-05, + "loss": 0.4534, + "step": 19193 + }, + { + "epoch": 0.57, + "learning_rate": 8.182440739205799e-05, + "loss": 0.6051, + "step": 19194 + }, + { + "epoch": 0.57, + "learning_rate": 8.181492567654404e-05, + "loss": 0.5781, + "step": 19195 + }, + { + "epoch": 0.57, + "learning_rate": 8.180544413010763e-05, + "loss": 0.6117, + "step": 19196 + }, + { + "epoch": 0.57, + "learning_rate": 8.1795962752837e-05, + "loss": 0.5867, + "step": 19197 + }, + { + "epoch": 0.57, + "learning_rate": 8.178648154482023e-05, + "loss": 0.5118, + "step": 19198 + }, + { + "epoch": 0.57, + "learning_rate": 8.177700050614549e-05, + "loss": 0.5994, + "step": 19199 + }, + { + "epoch": 0.57, + "learning_rate": 8.176751963690094e-05, + "loss": 0.4349, + "step": 19200 + }, + { + "epoch": 0.57, + "learning_rate": 8.175803893717476e-05, + "loss": 0.4615, + "step": 19201 + }, + { + "epoch": 0.57, + "learning_rate": 8.174855840705502e-05, + "loss": 0.456, + "step": 19202 + }, + { + "epoch": 0.57, + "learning_rate": 8.173907804662991e-05, + "loss": 0.6829, + "step": 19203 + }, + { + "epoch": 0.57, + "learning_rate": 8.172959785598757e-05, + "loss": 0.4412, + "step": 19204 + }, + { + "epoch": 0.57, + "learning_rate": 8.172011783521619e-05, + "loss": 0.6627, + "step": 19205 + }, + { + "epoch": 0.57, + "learning_rate": 8.171063798440382e-05, + "loss": 0.5961, + "step": 19206 + }, + { + "epoch": 0.57, + "learning_rate": 8.170115830363867e-05, + "loss": 0.5482, + "step": 19207 + }, + { + "epoch": 0.57, + "learning_rate": 8.169167879300884e-05, + "loss": 0.5253, + "step": 19208 + }, + { + "epoch": 0.57, + "learning_rate": 8.168219945260255e-05, + "loss": 0.4877, + "step": 19209 + }, + { + "epoch": 0.57, + "learning_rate": 8.167272028250779e-05, + "loss": 0.5645, + "step": 19210 + }, + { + "epoch": 0.57, + "learning_rate": 8.16632412828128e-05, + "loss": 0.4915, + "step": 19211 + }, + { + "epoch": 0.57, + "learning_rate": 8.165376245360568e-05, + "loss": 0.5568, + "step": 19212 + }, + { + "epoch": 0.57, + "learning_rate": 8.164428379497459e-05, + "loss": 0.5541, + "step": 19213 + }, + { + "epoch": 0.57, + "learning_rate": 8.163480530700761e-05, + "loss": 0.5283, + "step": 19214 + }, + { + "epoch": 0.57, + "learning_rate": 8.162532698979287e-05, + "loss": 0.436, + "step": 19215 + }, + { + "epoch": 0.57, + "learning_rate": 8.161584884341854e-05, + "loss": 0.5987, + "step": 19216 + }, + { + "epoch": 0.57, + "learning_rate": 8.160637086797274e-05, + "loss": 0.5427, + "step": 19217 + }, + { + "epoch": 0.57, + "learning_rate": 8.159689306354355e-05, + "loss": 0.5007, + "step": 19218 + }, + { + "epoch": 0.57, + "learning_rate": 8.15874154302191e-05, + "loss": 0.4911, + "step": 19219 + }, + { + "epoch": 0.57, + "learning_rate": 8.157793796808753e-05, + "loss": 0.4902, + "step": 19220 + }, + { + "epoch": 0.57, + "learning_rate": 8.1568460677237e-05, + "loss": 0.5164, + "step": 19221 + }, + { + "epoch": 0.57, + "learning_rate": 8.155898355775553e-05, + "loss": 0.5701, + "step": 19222 + }, + { + "epoch": 0.57, + "learning_rate": 8.154950660973127e-05, + "loss": 0.5882, + "step": 19223 + }, + { + "epoch": 0.57, + "learning_rate": 8.15400298332524e-05, + "loss": 0.5316, + "step": 19224 + }, + { + "epoch": 0.57, + "learning_rate": 8.15305532284069e-05, + "loss": 0.6281, + "step": 19225 + }, + { + "epoch": 0.57, + "learning_rate": 8.1521076795283e-05, + "loss": 0.5125, + "step": 19226 + }, + { + "epoch": 0.57, + "learning_rate": 8.151160053396873e-05, + "loss": 0.4977, + "step": 19227 + }, + { + "epoch": 0.57, + "learning_rate": 8.150212444455227e-05, + "loss": 0.4492, + "step": 19228 + }, + { + "epoch": 0.57, + "learning_rate": 8.149264852712168e-05, + "loss": 0.6805, + "step": 19229 + }, + { + "epoch": 0.57, + "learning_rate": 8.148317278176504e-05, + "loss": 0.462, + "step": 19230 + }, + { + "epoch": 0.57, + "learning_rate": 8.14736972085705e-05, + "loss": 0.4848, + "step": 19231 + }, + { + "epoch": 0.57, + "learning_rate": 8.146422180762614e-05, + "loss": 0.5323, + "step": 19232 + }, + { + "epoch": 0.57, + "learning_rate": 8.145474657902006e-05, + "loss": 0.5073, + "step": 19233 + }, + { + "epoch": 0.57, + "learning_rate": 8.144527152284033e-05, + "loss": 0.5688, + "step": 19234 + }, + { + "epoch": 0.57, + "learning_rate": 8.143579663917509e-05, + "loss": 0.487, + "step": 19235 + }, + { + "epoch": 0.57, + "learning_rate": 8.142632192811244e-05, + "loss": 0.5253, + "step": 19236 + }, + { + "epoch": 0.57, + "learning_rate": 8.141684738974039e-05, + "loss": 0.5525, + "step": 19237 + }, + { + "epoch": 0.57, + "learning_rate": 8.140737302414711e-05, + "loss": 0.4723, + "step": 19238 + }, + { + "epoch": 0.57, + "learning_rate": 8.139789883142066e-05, + "loss": 0.4971, + "step": 19239 + }, + { + "epoch": 0.57, + "learning_rate": 8.138842481164917e-05, + "loss": 0.3969, + "step": 19240 + }, + { + "epoch": 0.57, + "learning_rate": 8.137895096492065e-05, + "loss": 0.5966, + "step": 19241 + }, + { + "epoch": 0.57, + "learning_rate": 8.136947729132323e-05, + "loss": 0.5619, + "step": 19242 + }, + { + "epoch": 0.57, + "learning_rate": 8.136000379094497e-05, + "loss": 0.5828, + "step": 19243 + }, + { + "epoch": 0.57, + "learning_rate": 8.135053046387402e-05, + "loss": 0.5132, + "step": 19244 + }, + { + "epoch": 0.57, + "learning_rate": 8.134105731019837e-05, + "loss": 0.4993, + "step": 19245 + }, + { + "epoch": 0.57, + "learning_rate": 8.13315843300061e-05, + "loss": 0.5249, + "step": 19246 + }, + { + "epoch": 0.57, + "learning_rate": 8.132211152338535e-05, + "loss": 0.5559, + "step": 19247 + }, + { + "epoch": 0.57, + "learning_rate": 8.131263889042419e-05, + "loss": 0.6186, + "step": 19248 + }, + { + "epoch": 0.57, + "learning_rate": 8.130316643121061e-05, + "loss": 0.5058, + "step": 19249 + }, + { + "epoch": 0.57, + "learning_rate": 8.129369414583277e-05, + "loss": 0.4465, + "step": 19250 + }, + { + "epoch": 0.57, + "learning_rate": 8.128422203437868e-05, + "loss": 0.5412, + "step": 19251 + }, + { + "epoch": 0.57, + "learning_rate": 8.12747500969365e-05, + "loss": 0.4954, + "step": 19252 + }, + { + "epoch": 0.57, + "learning_rate": 8.126527833359416e-05, + "loss": 0.5985, + "step": 19253 + }, + { + "epoch": 0.57, + "learning_rate": 8.125580674443983e-05, + "loss": 0.5842, + "step": 19254 + }, + { + "epoch": 0.57, + "learning_rate": 8.124633532956154e-05, + "loss": 0.5615, + "step": 19255 + }, + { + "epoch": 0.57, + "learning_rate": 8.123686408904731e-05, + "loss": 0.5271, + "step": 19256 + }, + { + "epoch": 0.57, + "learning_rate": 8.122739302298528e-05, + "loss": 0.6631, + "step": 19257 + }, + { + "epoch": 0.57, + "learning_rate": 8.121792213146344e-05, + "loss": 0.5392, + "step": 19258 + }, + { + "epoch": 0.57, + "learning_rate": 8.120845141456991e-05, + "loss": 0.6042, + "step": 19259 + }, + { + "epoch": 0.57, + "learning_rate": 8.119898087239269e-05, + "loss": 0.612, + "step": 19260 + }, + { + "epoch": 0.57, + "learning_rate": 8.118951050501983e-05, + "loss": 0.6402, + "step": 19261 + }, + { + "epoch": 0.57, + "learning_rate": 8.118004031253942e-05, + "loss": 0.5184, + "step": 19262 + }, + { + "epoch": 0.57, + "learning_rate": 8.117057029503953e-05, + "loss": 0.6087, + "step": 19263 + }, + { + "epoch": 0.57, + "learning_rate": 8.116110045260814e-05, + "loss": 0.5756, + "step": 19264 + }, + { + "epoch": 0.57, + "learning_rate": 8.11516307853333e-05, + "loss": 0.4747, + "step": 19265 + }, + { + "epoch": 0.57, + "learning_rate": 8.114216129330313e-05, + "loss": 0.6424, + "step": 19266 + }, + { + "epoch": 0.57, + "learning_rate": 8.113269197660564e-05, + "loss": 0.5053, + "step": 19267 + }, + { + "epoch": 0.57, + "learning_rate": 8.11232228353288e-05, + "loss": 0.4508, + "step": 19268 + }, + { + "epoch": 0.57, + "learning_rate": 8.111375386956075e-05, + "loss": 0.4956, + "step": 19269 + }, + { + "epoch": 0.57, + "learning_rate": 8.110428507938947e-05, + "loss": 0.4471, + "step": 19270 + }, + { + "epoch": 0.57, + "learning_rate": 8.109481646490304e-05, + "loss": 0.5658, + "step": 19271 + }, + { + "epoch": 0.57, + "learning_rate": 8.108534802618948e-05, + "loss": 0.529, + "step": 19272 + }, + { + "epoch": 0.57, + "learning_rate": 8.10758797633368e-05, + "loss": 0.6698, + "step": 19273 + }, + { + "epoch": 0.57, + "learning_rate": 8.106641167643301e-05, + "loss": 0.5914, + "step": 19274 + }, + { + "epoch": 0.57, + "learning_rate": 8.105694376556626e-05, + "loss": 0.4647, + "step": 19275 + }, + { + "epoch": 0.57, + "learning_rate": 8.104747603082443e-05, + "loss": 0.4512, + "step": 19276 + }, + { + "epoch": 0.57, + "learning_rate": 8.103800847229562e-05, + "loss": 0.438, + "step": 19277 + }, + { + "epoch": 0.57, + "learning_rate": 8.102854109006786e-05, + "loss": 0.608, + "step": 19278 + }, + { + "epoch": 0.57, + "learning_rate": 8.10190738842292e-05, + "loss": 0.5016, + "step": 19279 + }, + { + "epoch": 0.57, + "learning_rate": 8.100960685486757e-05, + "loss": 0.6003, + "step": 19280 + }, + { + "epoch": 0.57, + "learning_rate": 8.100014000207108e-05, + "loss": 0.5012, + "step": 19281 + }, + { + "epoch": 0.57, + "learning_rate": 8.099067332592768e-05, + "loss": 0.5243, + "step": 19282 + }, + { + "epoch": 0.57, + "learning_rate": 8.098120682652549e-05, + "loss": 0.5094, + "step": 19283 + }, + { + "epoch": 0.57, + "learning_rate": 8.09717405039524e-05, + "loss": 0.4982, + "step": 19284 + }, + { + "epoch": 0.57, + "learning_rate": 8.09622743582965e-05, + "loss": 0.5331, + "step": 19285 + }, + { + "epoch": 0.57, + "learning_rate": 8.095280838964576e-05, + "loss": 0.5645, + "step": 19286 + }, + { + "epoch": 0.57, + "learning_rate": 8.094334259808827e-05, + "loss": 0.4771, + "step": 19287 + }, + { + "epoch": 0.57, + "learning_rate": 8.093387698371195e-05, + "loss": 0.5799, + "step": 19288 + }, + { + "epoch": 0.57, + "learning_rate": 8.092441154660485e-05, + "loss": 0.6187, + "step": 19289 + }, + { + "epoch": 0.57, + "learning_rate": 8.091494628685499e-05, + "loss": 0.4853, + "step": 19290 + }, + { + "epoch": 0.57, + "learning_rate": 8.090548120455033e-05, + "loss": 0.447, + "step": 19291 + }, + { + "epoch": 0.57, + "learning_rate": 8.089601629977887e-05, + "loss": 0.5611, + "step": 19292 + }, + { + "epoch": 0.57, + "learning_rate": 8.088655157262866e-05, + "loss": 0.5793, + "step": 19293 + }, + { + "epoch": 0.57, + "learning_rate": 8.08770870231877e-05, + "loss": 0.4245, + "step": 19294 + }, + { + "epoch": 0.57, + "learning_rate": 8.086762265154396e-05, + "loss": 0.5883, + "step": 19295 + }, + { + "epoch": 0.57, + "learning_rate": 8.085815845778539e-05, + "loss": 0.4783, + "step": 19296 + }, + { + "epoch": 0.57, + "learning_rate": 8.084869444200008e-05, + "loss": 0.5757, + "step": 19297 + }, + { + "epoch": 0.57, + "learning_rate": 8.083923060427597e-05, + "loss": 0.5067, + "step": 19298 + }, + { + "epoch": 0.57, + "learning_rate": 8.082976694470103e-05, + "loss": 0.4786, + "step": 19299 + }, + { + "epoch": 0.57, + "learning_rate": 8.08203034633633e-05, + "loss": 0.5286, + "step": 19300 + }, + { + "epoch": 0.57, + "learning_rate": 8.081084016035072e-05, + "loss": 0.6082, + "step": 19301 + }, + { + "epoch": 0.57, + "learning_rate": 8.080137703575134e-05, + "loss": 0.637, + "step": 19302 + }, + { + "epoch": 0.57, + "learning_rate": 8.079191408965308e-05, + "loss": 0.5386, + "step": 19303 + }, + { + "epoch": 0.57, + "learning_rate": 8.078245132214396e-05, + "loss": 0.5137, + "step": 19304 + }, + { + "epoch": 0.57, + "learning_rate": 8.077298873331192e-05, + "loss": 0.5602, + "step": 19305 + }, + { + "epoch": 0.57, + "learning_rate": 8.0763526323245e-05, + "loss": 0.541, + "step": 19306 + }, + { + "epoch": 0.57, + "learning_rate": 8.075406409203114e-05, + "loss": 0.4455, + "step": 19307 + }, + { + "epoch": 0.57, + "learning_rate": 8.074460203975828e-05, + "loss": 0.4997, + "step": 19308 + }, + { + "epoch": 0.57, + "learning_rate": 8.073514016651448e-05, + "loss": 0.5622, + "step": 19309 + }, + { + "epoch": 0.57, + "learning_rate": 8.072567847238768e-05, + "loss": 0.5282, + "step": 19310 + }, + { + "epoch": 0.57, + "learning_rate": 8.071621695746579e-05, + "loss": 0.5651, + "step": 19311 + }, + { + "epoch": 0.57, + "learning_rate": 8.070675562183687e-05, + "loss": 0.5772, + "step": 19312 + }, + { + "epoch": 0.57, + "learning_rate": 8.06972944655888e-05, + "loss": 0.5236, + "step": 19313 + }, + { + "epoch": 0.58, + "learning_rate": 8.068783348880968e-05, + "loss": 0.5219, + "step": 19314 + }, + { + "epoch": 0.58, + "learning_rate": 8.067837269158731e-05, + "loss": 0.5817, + "step": 19315 + }, + { + "epoch": 0.58, + "learning_rate": 8.066891207400975e-05, + "loss": 0.6551, + "step": 19316 + }, + { + "epoch": 0.58, + "learning_rate": 8.065945163616492e-05, + "loss": 0.61, + "step": 19317 + }, + { + "epoch": 0.58, + "learning_rate": 8.064999137814085e-05, + "loss": 0.6233, + "step": 19318 + }, + { + "epoch": 0.58, + "learning_rate": 8.064053130002542e-05, + "loss": 0.4789, + "step": 19319 + }, + { + "epoch": 0.58, + "learning_rate": 8.06310714019066e-05, + "loss": 0.4621, + "step": 19320 + }, + { + "epoch": 0.58, + "learning_rate": 8.06216116838724e-05, + "loss": 0.6549, + "step": 19321 + }, + { + "epoch": 0.58, + "learning_rate": 8.061215214601071e-05, + "loss": 0.5598, + "step": 19322 + }, + { + "epoch": 0.58, + "learning_rate": 8.060269278840947e-05, + "loss": 0.5424, + "step": 19323 + }, + { + "epoch": 0.58, + "learning_rate": 8.05932336111567e-05, + "loss": 0.652, + "step": 19324 + }, + { + "epoch": 0.58, + "learning_rate": 8.058377461434032e-05, + "loss": 0.5263, + "step": 19325 + }, + { + "epoch": 0.58, + "learning_rate": 8.057431579804824e-05, + "loss": 0.5846, + "step": 19326 + }, + { + "epoch": 0.58, + "learning_rate": 8.056485716236841e-05, + "loss": 0.5955, + "step": 19327 + }, + { + "epoch": 0.58, + "learning_rate": 8.055539870738882e-05, + "loss": 0.5167, + "step": 19328 + }, + { + "epoch": 0.58, + "learning_rate": 8.05459404331974e-05, + "loss": 0.5777, + "step": 19329 + }, + { + "epoch": 0.58, + "learning_rate": 8.053648233988203e-05, + "loss": 0.5407, + "step": 19330 + }, + { + "epoch": 0.58, + "learning_rate": 8.05270244275307e-05, + "loss": 0.4856, + "step": 19331 + }, + { + "epoch": 0.58, + "learning_rate": 8.051756669623134e-05, + "loss": 0.5596, + "step": 19332 + }, + { + "epoch": 0.58, + "learning_rate": 8.05081091460719e-05, + "loss": 0.5817, + "step": 19333 + }, + { + "epoch": 0.58, + "learning_rate": 8.049865177714029e-05, + "loss": 0.6069, + "step": 19334 + }, + { + "epoch": 0.58, + "learning_rate": 8.048919458952442e-05, + "loss": 0.6007, + "step": 19335 + }, + { + "epoch": 0.58, + "learning_rate": 8.047973758331224e-05, + "loss": 0.4576, + "step": 19336 + }, + { + "epoch": 0.58, + "learning_rate": 8.047028075859172e-05, + "loss": 0.4316, + "step": 19337 + }, + { + "epoch": 0.58, + "learning_rate": 8.046082411545071e-05, + "loss": 0.4953, + "step": 19338 + }, + { + "epoch": 0.58, + "learning_rate": 8.045136765397716e-05, + "loss": 0.3846, + "step": 19339 + }, + { + "epoch": 0.58, + "learning_rate": 8.044191137425903e-05, + "loss": 0.4866, + "step": 19340 + }, + { + "epoch": 0.58, + "learning_rate": 8.043245527638422e-05, + "loss": 0.5294, + "step": 19341 + }, + { + "epoch": 0.58, + "learning_rate": 8.04229993604406e-05, + "loss": 0.5204, + "step": 19342 + }, + { + "epoch": 0.58, + "learning_rate": 8.041354362651616e-05, + "loss": 0.6472, + "step": 19343 + }, + { + "epoch": 0.58, + "learning_rate": 8.040408807469876e-05, + "loss": 0.6107, + "step": 19344 + }, + { + "epoch": 0.58, + "learning_rate": 8.039463270507642e-05, + "loss": 0.5706, + "step": 19345 + }, + { + "epoch": 0.58, + "learning_rate": 8.038517751773689e-05, + "loss": 0.5697, + "step": 19346 + }, + { + "epoch": 0.58, + "learning_rate": 8.037572251276818e-05, + "loss": 0.7143, + "step": 19347 + }, + { + "epoch": 0.58, + "learning_rate": 8.036626769025815e-05, + "loss": 0.6139, + "step": 19348 + }, + { + "epoch": 0.58, + "learning_rate": 8.03568130502948e-05, + "loss": 0.5043, + "step": 19349 + }, + { + "epoch": 0.58, + "learning_rate": 8.034735859296594e-05, + "loss": 0.5658, + "step": 19350 + }, + { + "epoch": 0.58, + "learning_rate": 8.033790431835951e-05, + "loss": 0.4691, + "step": 19351 + }, + { + "epoch": 0.58, + "learning_rate": 8.03284502265634e-05, + "loss": 0.4957, + "step": 19352 + }, + { + "epoch": 0.58, + "learning_rate": 8.031899631766555e-05, + "loss": 0.5107, + "step": 19353 + }, + { + "epoch": 0.58, + "learning_rate": 8.030954259175379e-05, + "loss": 0.5145, + "step": 19354 + }, + { + "epoch": 0.58, + "learning_rate": 8.030008904891607e-05, + "loss": 0.5475, + "step": 19355 + }, + { + "epoch": 0.58, + "learning_rate": 8.029063568924031e-05, + "loss": 0.5993, + "step": 19356 + }, + { + "epoch": 0.58, + "learning_rate": 8.028118251281431e-05, + "loss": 0.492, + "step": 19357 + }, + { + "epoch": 0.58, + "learning_rate": 8.027172951972602e-05, + "loss": 0.5663, + "step": 19358 + }, + { + "epoch": 0.58, + "learning_rate": 8.026227671006335e-05, + "loss": 0.6578, + "step": 19359 + }, + { + "epoch": 0.58, + "learning_rate": 8.025282408391416e-05, + "loss": 0.6588, + "step": 19360 + }, + { + "epoch": 0.58, + "learning_rate": 8.024337164136633e-05, + "loss": 0.5121, + "step": 19361 + }, + { + "epoch": 0.58, + "learning_rate": 8.023391938250775e-05, + "loss": 0.5792, + "step": 19362 + }, + { + "epoch": 0.58, + "learning_rate": 8.02244673074263e-05, + "loss": 0.5753, + "step": 19363 + }, + { + "epoch": 0.58, + "learning_rate": 8.021501541620992e-05, + "loss": 0.5966, + "step": 19364 + }, + { + "epoch": 0.58, + "learning_rate": 8.020556370894642e-05, + "loss": 0.6051, + "step": 19365 + }, + { + "epoch": 0.58, + "learning_rate": 8.019611218572367e-05, + "loss": 0.6096, + "step": 19366 + }, + { + "epoch": 0.58, + "learning_rate": 8.01866608466296e-05, + "loss": 0.5921, + "step": 19367 + }, + { + "epoch": 0.58, + "learning_rate": 8.017720969175207e-05, + "loss": 0.5712, + "step": 19368 + }, + { + "epoch": 0.58, + "learning_rate": 8.016775872117895e-05, + "loss": 0.5021, + "step": 19369 + }, + { + "epoch": 0.58, + "learning_rate": 8.015830793499806e-05, + "loss": 0.4904, + "step": 19370 + }, + { + "epoch": 0.58, + "learning_rate": 8.014885733329735e-05, + "loss": 0.602, + "step": 19371 + }, + { + "epoch": 0.58, + "learning_rate": 8.013940691616466e-05, + "loss": 0.6065, + "step": 19372 + }, + { + "epoch": 0.58, + "learning_rate": 8.012995668368781e-05, + "loss": 0.5231, + "step": 19373 + }, + { + "epoch": 0.58, + "learning_rate": 8.012050663595475e-05, + "loss": 0.5187, + "step": 19374 + }, + { + "epoch": 0.58, + "learning_rate": 8.011105677305324e-05, + "loss": 0.4061, + "step": 19375 + }, + { + "epoch": 0.58, + "learning_rate": 8.010160709507125e-05, + "loss": 0.5053, + "step": 19376 + }, + { + "epoch": 0.58, + "learning_rate": 8.009215760209657e-05, + "loss": 0.5833, + "step": 19377 + }, + { + "epoch": 0.58, + "learning_rate": 8.008270829421707e-05, + "loss": 0.6342, + "step": 19378 + }, + { + "epoch": 0.58, + "learning_rate": 8.007325917152059e-05, + "loss": 0.5786, + "step": 19379 + }, + { + "epoch": 0.58, + "learning_rate": 8.006381023409505e-05, + "loss": 0.6409, + "step": 19380 + }, + { + "epoch": 0.58, + "learning_rate": 8.005436148202823e-05, + "loss": 0.5908, + "step": 19381 + }, + { + "epoch": 0.58, + "learning_rate": 8.004491291540799e-05, + "loss": 0.5014, + "step": 19382 + }, + { + "epoch": 0.58, + "learning_rate": 8.003546453432222e-05, + "loss": 0.5903, + "step": 19383 + }, + { + "epoch": 0.58, + "learning_rate": 8.002601633885876e-05, + "loss": 0.5804, + "step": 19384 + }, + { + "epoch": 0.58, + "learning_rate": 8.00165683291054e-05, + "loss": 0.6612, + "step": 19385 + }, + { + "epoch": 0.58, + "learning_rate": 8.000712050515004e-05, + "loss": 0.5259, + "step": 19386 + }, + { + "epoch": 0.58, + "learning_rate": 7.999767286708053e-05, + "loss": 0.6064, + "step": 19387 + }, + { + "epoch": 0.58, + "learning_rate": 7.998822541498467e-05, + "loss": 0.4523, + "step": 19388 + }, + { + "epoch": 0.58, + "learning_rate": 7.997877814895029e-05, + "loss": 0.6384, + "step": 19389 + }, + { + "epoch": 0.58, + "learning_rate": 7.996933106906526e-05, + "loss": 0.4261, + "step": 19390 + }, + { + "epoch": 0.58, + "learning_rate": 7.995988417541745e-05, + "loss": 0.472, + "step": 19391 + }, + { + "epoch": 0.58, + "learning_rate": 7.995043746809459e-05, + "loss": 0.6192, + "step": 19392 + }, + { + "epoch": 0.58, + "learning_rate": 7.994099094718459e-05, + "loss": 0.5128, + "step": 19393 + }, + { + "epoch": 0.58, + "learning_rate": 7.993154461277524e-05, + "loss": 0.4955, + "step": 19394 + }, + { + "epoch": 0.58, + "learning_rate": 7.992209846495444e-05, + "loss": 0.5341, + "step": 19395 + }, + { + "epoch": 0.58, + "learning_rate": 7.991265250380995e-05, + "loss": 0.5561, + "step": 19396 + }, + { + "epoch": 0.58, + "learning_rate": 7.990320672942958e-05, + "loss": 0.6005, + "step": 19397 + }, + { + "epoch": 0.58, + "learning_rate": 7.989376114190122e-05, + "loss": 0.5421, + "step": 19398 + }, + { + "epoch": 0.58, + "learning_rate": 7.988431574131265e-05, + "loss": 0.5727, + "step": 19399 + }, + { + "epoch": 0.58, + "learning_rate": 7.98748705277517e-05, + "loss": 0.6157, + "step": 19400 + }, + { + "epoch": 0.58, + "learning_rate": 7.986542550130614e-05, + "loss": 0.4684, + "step": 19401 + }, + { + "epoch": 0.58, + "learning_rate": 7.985598066206387e-05, + "loss": 0.563, + "step": 19402 + }, + { + "epoch": 0.58, + "learning_rate": 7.984653601011267e-05, + "loss": 0.5917, + "step": 19403 + }, + { + "epoch": 0.58, + "learning_rate": 7.983709154554031e-05, + "loss": 0.5594, + "step": 19404 + }, + { + "epoch": 0.58, + "learning_rate": 7.982764726843466e-05, + "loss": 0.5907, + "step": 19405 + }, + { + "epoch": 0.58, + "learning_rate": 7.981820317888348e-05, + "loss": 0.5121, + "step": 19406 + }, + { + "epoch": 0.58, + "learning_rate": 7.980875927697465e-05, + "loss": 0.4585, + "step": 19407 + }, + { + "epoch": 0.58, + "learning_rate": 7.979931556279589e-05, + "loss": 0.5617, + "step": 19408 + }, + { + "epoch": 0.58, + "learning_rate": 7.978987203643506e-05, + "loss": 0.5074, + "step": 19409 + }, + { + "epoch": 0.58, + "learning_rate": 7.978042869797993e-05, + "loss": 0.5436, + "step": 19410 + }, + { + "epoch": 0.58, + "learning_rate": 7.977098554751835e-05, + "loss": 0.5948, + "step": 19411 + }, + { + "epoch": 0.58, + "learning_rate": 7.976154258513806e-05, + "loss": 0.5766, + "step": 19412 + }, + { + "epoch": 0.58, + "learning_rate": 7.975209981092686e-05, + "loss": 0.6358, + "step": 19413 + }, + { + "epoch": 0.58, + "learning_rate": 7.974265722497259e-05, + "loss": 0.5045, + "step": 19414 + }, + { + "epoch": 0.58, + "learning_rate": 7.973321482736303e-05, + "loss": 0.6041, + "step": 19415 + }, + { + "epoch": 0.58, + "learning_rate": 7.972377261818594e-05, + "loss": 0.572, + "step": 19416 + }, + { + "epoch": 0.58, + "learning_rate": 7.971433059752913e-05, + "loss": 0.4933, + "step": 19417 + }, + { + "epoch": 0.58, + "learning_rate": 7.970488876548038e-05, + "loss": 0.6488, + "step": 19418 + }, + { + "epoch": 0.58, + "learning_rate": 7.969544712212755e-05, + "loss": 0.6084, + "step": 19419 + }, + { + "epoch": 0.58, + "learning_rate": 7.968600566755828e-05, + "loss": 0.5369, + "step": 19420 + }, + { + "epoch": 0.58, + "learning_rate": 7.967656440186047e-05, + "loss": 0.5663, + "step": 19421 + }, + { + "epoch": 0.58, + "learning_rate": 7.966712332512188e-05, + "loss": 0.5303, + "step": 19422 + }, + { + "epoch": 0.58, + "learning_rate": 7.965768243743024e-05, + "loss": 0.6093, + "step": 19423 + }, + { + "epoch": 0.58, + "learning_rate": 7.964824173887338e-05, + "loss": 0.4659, + "step": 19424 + }, + { + "epoch": 0.58, + "learning_rate": 7.963880122953901e-05, + "loss": 0.5262, + "step": 19425 + }, + { + "epoch": 0.58, + "learning_rate": 7.962936090951502e-05, + "loss": 0.6204, + "step": 19426 + }, + { + "epoch": 0.58, + "learning_rate": 7.961992077888909e-05, + "loss": 0.4484, + "step": 19427 + }, + { + "epoch": 0.58, + "learning_rate": 7.961048083774899e-05, + "loss": 0.6991, + "step": 19428 + }, + { + "epoch": 0.58, + "learning_rate": 7.960104108618254e-05, + "loss": 0.4138, + "step": 19429 + }, + { + "epoch": 0.58, + "learning_rate": 7.95916015242775e-05, + "loss": 0.4965, + "step": 19430 + }, + { + "epoch": 0.58, + "learning_rate": 7.958216215212158e-05, + "loss": 0.4755, + "step": 19431 + }, + { + "epoch": 0.58, + "learning_rate": 7.957272296980256e-05, + "loss": 0.5396, + "step": 19432 + }, + { + "epoch": 0.58, + "learning_rate": 7.956328397740825e-05, + "loss": 0.5342, + "step": 19433 + }, + { + "epoch": 0.58, + "learning_rate": 7.95538451750264e-05, + "loss": 0.6, + "step": 19434 + }, + { + "epoch": 0.58, + "learning_rate": 7.95444065627447e-05, + "loss": 0.5476, + "step": 19435 + }, + { + "epoch": 0.58, + "learning_rate": 7.953496814065098e-05, + "loss": 0.5247, + "step": 19436 + }, + { + "epoch": 0.58, + "learning_rate": 7.952552990883295e-05, + "loss": 0.5689, + "step": 19437 + }, + { + "epoch": 0.58, + "learning_rate": 7.951609186737841e-05, + "loss": 0.5615, + "step": 19438 + }, + { + "epoch": 0.58, + "learning_rate": 7.95066540163751e-05, + "loss": 0.5892, + "step": 19439 + }, + { + "epoch": 0.58, + "learning_rate": 7.949721635591071e-05, + "loss": 0.603, + "step": 19440 + }, + { + "epoch": 0.58, + "learning_rate": 7.948777888607304e-05, + "loss": 0.6328, + "step": 19441 + }, + { + "epoch": 0.58, + "learning_rate": 7.947834160694984e-05, + "loss": 0.4313, + "step": 19442 + }, + { + "epoch": 0.58, + "learning_rate": 7.946890451862885e-05, + "loss": 0.5645, + "step": 19443 + }, + { + "epoch": 0.58, + "learning_rate": 7.945946762119777e-05, + "loss": 0.4818, + "step": 19444 + }, + { + "epoch": 0.58, + "learning_rate": 7.945003091474439e-05, + "loss": 0.5115, + "step": 19445 + }, + { + "epoch": 0.58, + "learning_rate": 7.944059439935646e-05, + "loss": 0.4994, + "step": 19446 + }, + { + "epoch": 0.58, + "learning_rate": 7.943115807512164e-05, + "loss": 0.6395, + "step": 19447 + }, + { + "epoch": 0.58, + "learning_rate": 7.942172194212774e-05, + "loss": 0.5903, + "step": 19448 + }, + { + "epoch": 0.58, + "learning_rate": 7.941228600046244e-05, + "loss": 0.4854, + "step": 19449 + }, + { + "epoch": 0.58, + "learning_rate": 7.940285025021357e-05, + "loss": 0.6029, + "step": 19450 + }, + { + "epoch": 0.58, + "learning_rate": 7.939341469146872e-05, + "loss": 0.5082, + "step": 19451 + }, + { + "epoch": 0.58, + "learning_rate": 7.938397932431572e-05, + "loss": 0.5317, + "step": 19452 + }, + { + "epoch": 0.58, + "learning_rate": 7.937454414884228e-05, + "loss": 0.545, + "step": 19453 + }, + { + "epoch": 0.58, + "learning_rate": 7.936510916513606e-05, + "loss": 0.5312, + "step": 19454 + }, + { + "epoch": 0.58, + "learning_rate": 7.935567437328486e-05, + "loss": 0.5101, + "step": 19455 + }, + { + "epoch": 0.58, + "learning_rate": 7.934623977337635e-05, + "loss": 0.6044, + "step": 19456 + }, + { + "epoch": 0.58, + "learning_rate": 7.93368053654983e-05, + "loss": 0.5711, + "step": 19457 + }, + { + "epoch": 0.58, + "learning_rate": 7.932737114973838e-05, + "loss": 0.4889, + "step": 19458 + }, + { + "epoch": 0.58, + "learning_rate": 7.931793712618431e-05, + "loss": 0.5847, + "step": 19459 + }, + { + "epoch": 0.58, + "learning_rate": 7.930850329492384e-05, + "loss": 0.4877, + "step": 19460 + }, + { + "epoch": 0.58, + "learning_rate": 7.929906965604467e-05, + "loss": 0.5081, + "step": 19461 + }, + { + "epoch": 0.58, + "learning_rate": 7.92896362096345e-05, + "loss": 0.5557, + "step": 19462 + }, + { + "epoch": 0.58, + "learning_rate": 7.9280202955781e-05, + "loss": 0.5209, + "step": 19463 + }, + { + "epoch": 0.58, + "learning_rate": 7.927076989457194e-05, + "loss": 0.4695, + "step": 19464 + }, + { + "epoch": 0.58, + "learning_rate": 7.926133702609502e-05, + "loss": 0.6263, + "step": 19465 + }, + { + "epoch": 0.58, + "learning_rate": 7.925190435043788e-05, + "loss": 0.5994, + "step": 19466 + }, + { + "epoch": 0.58, + "learning_rate": 7.924247186768829e-05, + "loss": 0.424, + "step": 19467 + }, + { + "epoch": 0.58, + "learning_rate": 7.923303957793389e-05, + "loss": 0.4786, + "step": 19468 + }, + { + "epoch": 0.58, + "learning_rate": 7.922360748126246e-05, + "loss": 0.4631, + "step": 19469 + }, + { + "epoch": 0.58, + "learning_rate": 7.921417557776163e-05, + "loss": 0.5142, + "step": 19470 + }, + { + "epoch": 0.58, + "learning_rate": 7.92047438675191e-05, + "loss": 0.5333, + "step": 19471 + }, + { + "epoch": 0.58, + "learning_rate": 7.919531235062257e-05, + "loss": 0.6306, + "step": 19472 + }, + { + "epoch": 0.58, + "learning_rate": 7.918588102715976e-05, + "loss": 0.4153, + "step": 19473 + }, + { + "epoch": 0.58, + "learning_rate": 7.917644989721832e-05, + "loss": 0.5242, + "step": 19474 + }, + { + "epoch": 0.58, + "learning_rate": 7.916701896088592e-05, + "loss": 0.4908, + "step": 19475 + }, + { + "epoch": 0.58, + "learning_rate": 7.91575882182503e-05, + "loss": 0.4423, + "step": 19476 + }, + { + "epoch": 0.58, + "learning_rate": 7.914815766939915e-05, + "loss": 0.6107, + "step": 19477 + }, + { + "epoch": 0.58, + "learning_rate": 7.913872731442008e-05, + "loss": 0.5512, + "step": 19478 + }, + { + "epoch": 0.58, + "learning_rate": 7.912929715340081e-05, + "loss": 0.6061, + "step": 19479 + }, + { + "epoch": 0.58, + "learning_rate": 7.9119867186429e-05, + "loss": 0.6012, + "step": 19480 + }, + { + "epoch": 0.58, + "learning_rate": 7.911043741359241e-05, + "loss": 0.503, + "step": 19481 + }, + { + "epoch": 0.58, + "learning_rate": 7.910100783497858e-05, + "loss": 0.6069, + "step": 19482 + }, + { + "epoch": 0.58, + "learning_rate": 7.909157845067528e-05, + "loss": 0.437, + "step": 19483 + }, + { + "epoch": 0.58, + "learning_rate": 7.908214926077011e-05, + "loss": 0.5296, + "step": 19484 + }, + { + "epoch": 0.58, + "learning_rate": 7.907272026535084e-05, + "loss": 0.4326, + "step": 19485 + }, + { + "epoch": 0.58, + "learning_rate": 7.906329146450505e-05, + "loss": 0.5663, + "step": 19486 + }, + { + "epoch": 0.58, + "learning_rate": 7.90538628583204e-05, + "loss": 0.6207, + "step": 19487 + }, + { + "epoch": 0.58, + "learning_rate": 7.904443444688465e-05, + "loss": 0.6036, + "step": 19488 + }, + { + "epoch": 0.58, + "learning_rate": 7.903500623028536e-05, + "loss": 0.5469, + "step": 19489 + }, + { + "epoch": 0.58, + "learning_rate": 7.90255782086102e-05, + "loss": 0.4568, + "step": 19490 + }, + { + "epoch": 0.58, + "learning_rate": 7.901615038194689e-05, + "loss": 0.4713, + "step": 19491 + }, + { + "epoch": 0.58, + "learning_rate": 7.900672275038307e-05, + "loss": 0.5556, + "step": 19492 + }, + { + "epoch": 0.58, + "learning_rate": 7.899729531400635e-05, + "loss": 0.5234, + "step": 19493 + }, + { + "epoch": 0.58, + "learning_rate": 7.898786807290437e-05, + "loss": 0.5996, + "step": 19494 + }, + { + "epoch": 0.58, + "learning_rate": 7.897844102716486e-05, + "loss": 0.4477, + "step": 19495 + }, + { + "epoch": 0.58, + "learning_rate": 7.896901417687545e-05, + "loss": 0.5978, + "step": 19496 + }, + { + "epoch": 0.58, + "learning_rate": 7.89595875221237e-05, + "loss": 0.5742, + "step": 19497 + }, + { + "epoch": 0.58, + "learning_rate": 7.895016106299735e-05, + "loss": 0.6107, + "step": 19498 + }, + { + "epoch": 0.58, + "learning_rate": 7.8940734799584e-05, + "loss": 0.5688, + "step": 19499 + }, + { + "epoch": 0.58, + "learning_rate": 7.893130873197134e-05, + "loss": 0.5409, + "step": 19500 + }, + { + "epoch": 0.58, + "learning_rate": 7.892188286024696e-05, + "loss": 0.5512, + "step": 19501 + }, + { + "epoch": 0.58, + "learning_rate": 7.891245718449849e-05, + "loss": 0.488, + "step": 19502 + }, + { + "epoch": 0.58, + "learning_rate": 7.89030317048136e-05, + "loss": 0.5577, + "step": 19503 + }, + { + "epoch": 0.58, + "learning_rate": 7.889360642127994e-05, + "loss": 0.5789, + "step": 19504 + }, + { + "epoch": 0.58, + "learning_rate": 7.88841813339851e-05, + "loss": 0.6042, + "step": 19505 + }, + { + "epoch": 0.58, + "learning_rate": 7.88747564430167e-05, + "loss": 0.5577, + "step": 19506 + }, + { + "epoch": 0.58, + "learning_rate": 7.886533174846241e-05, + "loss": 0.4295, + "step": 19507 + }, + { + "epoch": 0.58, + "learning_rate": 7.885590725040987e-05, + "loss": 0.6031, + "step": 19508 + }, + { + "epoch": 0.58, + "learning_rate": 7.884648294894663e-05, + "loss": 0.5516, + "step": 19509 + }, + { + "epoch": 0.58, + "learning_rate": 7.883705884416038e-05, + "loss": 0.6122, + "step": 19510 + }, + { + "epoch": 0.58, + "learning_rate": 7.882763493613872e-05, + "loss": 0.5351, + "step": 19511 + }, + { + "epoch": 0.58, + "learning_rate": 7.881821122496931e-05, + "loss": 0.5882, + "step": 19512 + }, + { + "epoch": 0.58, + "learning_rate": 7.880878771073968e-05, + "loss": 0.4989, + "step": 19513 + }, + { + "epoch": 0.58, + "learning_rate": 7.879936439353751e-05, + "loss": 0.5499, + "step": 19514 + }, + { + "epoch": 0.58, + "learning_rate": 7.878994127345039e-05, + "loss": 0.5838, + "step": 19515 + }, + { + "epoch": 0.58, + "learning_rate": 7.878051835056598e-05, + "loss": 0.602, + "step": 19516 + }, + { + "epoch": 0.58, + "learning_rate": 7.877109562497183e-05, + "loss": 0.4861, + "step": 19517 + }, + { + "epoch": 0.58, + "learning_rate": 7.876167309675555e-05, + "loss": 0.4551, + "step": 19518 + }, + { + "epoch": 0.58, + "learning_rate": 7.875225076600481e-05, + "loss": 0.5839, + "step": 19519 + }, + { + "epoch": 0.58, + "learning_rate": 7.874282863280716e-05, + "loss": 0.5547, + "step": 19520 + }, + { + "epoch": 0.58, + "learning_rate": 7.873340669725018e-05, + "loss": 0.6048, + "step": 19521 + }, + { + "epoch": 0.58, + "learning_rate": 7.872398495942154e-05, + "loss": 0.6376, + "step": 19522 + }, + { + "epoch": 0.58, + "learning_rate": 7.871456341940884e-05, + "loss": 0.7011, + "step": 19523 + }, + { + "epoch": 0.58, + "learning_rate": 7.87051420772996e-05, + "loss": 0.5606, + "step": 19524 + }, + { + "epoch": 0.58, + "learning_rate": 7.869572093318146e-05, + "loss": 0.4497, + "step": 19525 + }, + { + "epoch": 0.58, + "learning_rate": 7.868629998714203e-05, + "loss": 0.4331, + "step": 19526 + }, + { + "epoch": 0.58, + "learning_rate": 7.867687923926892e-05, + "loss": 0.5234, + "step": 19527 + }, + { + "epoch": 0.58, + "learning_rate": 7.866745868964963e-05, + "loss": 0.5574, + "step": 19528 + }, + { + "epoch": 0.58, + "learning_rate": 7.865803833837184e-05, + "loss": 0.5486, + "step": 19529 + }, + { + "epoch": 0.58, + "learning_rate": 7.864861818552307e-05, + "loss": 0.4969, + "step": 19530 + }, + { + "epoch": 0.58, + "learning_rate": 7.863919823119099e-05, + "loss": 0.439, + "step": 19531 + }, + { + "epoch": 0.58, + "learning_rate": 7.862977847546311e-05, + "loss": 0.5225, + "step": 19532 + }, + { + "epoch": 0.58, + "learning_rate": 7.8620358918427e-05, + "loss": 0.5124, + "step": 19533 + }, + { + "epoch": 0.58, + "learning_rate": 7.861093956017031e-05, + "loss": 0.4808, + "step": 19534 + }, + { + "epoch": 0.58, + "learning_rate": 7.860152040078059e-05, + "loss": 0.533, + "step": 19535 + }, + { + "epoch": 0.58, + "learning_rate": 7.859210144034539e-05, + "loss": 0.5347, + "step": 19536 + }, + { + "epoch": 0.58, + "learning_rate": 7.858268267895227e-05, + "loss": 0.5586, + "step": 19537 + }, + { + "epoch": 0.58, + "learning_rate": 7.857326411668885e-05, + "loss": 0.577, + "step": 19538 + }, + { + "epoch": 0.58, + "learning_rate": 7.856384575364271e-05, + "loss": 0.5384, + "step": 19539 + }, + { + "epoch": 0.58, + "learning_rate": 7.855442758990135e-05, + "loss": 0.5197, + "step": 19540 + }, + { + "epoch": 0.58, + "learning_rate": 7.854500962555238e-05, + "loss": 0.6201, + "step": 19541 + }, + { + "epoch": 0.58, + "learning_rate": 7.853559186068334e-05, + "loss": 0.5884, + "step": 19542 + }, + { + "epoch": 0.58, + "learning_rate": 7.852617429538187e-05, + "loss": 0.5835, + "step": 19543 + }, + { + "epoch": 0.58, + "learning_rate": 7.851675692973543e-05, + "loss": 0.6375, + "step": 19544 + }, + { + "epoch": 0.58, + "learning_rate": 7.850733976383164e-05, + "loss": 0.5309, + "step": 19545 + }, + { + "epoch": 0.58, + "learning_rate": 7.849792279775801e-05, + "loss": 0.5338, + "step": 19546 + }, + { + "epoch": 0.58, + "learning_rate": 7.848850603160217e-05, + "loss": 0.6121, + "step": 19547 + }, + { + "epoch": 0.58, + "learning_rate": 7.84790894654516e-05, + "loss": 0.4789, + "step": 19548 + }, + { + "epoch": 0.58, + "learning_rate": 7.846967309939387e-05, + "loss": 0.5525, + "step": 19549 + }, + { + "epoch": 0.58, + "learning_rate": 7.846025693351655e-05, + "loss": 0.6914, + "step": 19550 + }, + { + "epoch": 0.58, + "learning_rate": 7.84508409679072e-05, + "loss": 0.5586, + "step": 19551 + }, + { + "epoch": 0.58, + "learning_rate": 7.84414252026533e-05, + "loss": 0.4959, + "step": 19552 + }, + { + "epoch": 0.58, + "learning_rate": 7.843200963784246e-05, + "loss": 0.6289, + "step": 19553 + }, + { + "epoch": 0.58, + "learning_rate": 7.84225942735622e-05, + "loss": 0.5984, + "step": 19554 + }, + { + "epoch": 0.58, + "learning_rate": 7.841317910990006e-05, + "loss": 0.4601, + "step": 19555 + }, + { + "epoch": 0.58, + "learning_rate": 7.840376414694354e-05, + "loss": 0.6088, + "step": 19556 + }, + { + "epoch": 0.58, + "learning_rate": 7.839434938478026e-05, + "loss": 0.6078, + "step": 19557 + }, + { + "epoch": 0.58, + "learning_rate": 7.83849348234977e-05, + "loss": 0.5505, + "step": 19558 + }, + { + "epoch": 0.58, + "learning_rate": 7.837552046318337e-05, + "loss": 0.5929, + "step": 19559 + }, + { + "epoch": 0.58, + "learning_rate": 7.836610630392485e-05, + "loss": 0.5322, + "step": 19560 + }, + { + "epoch": 0.58, + "learning_rate": 7.835669234580965e-05, + "loss": 0.5835, + "step": 19561 + }, + { + "epoch": 0.58, + "learning_rate": 7.834727858892533e-05, + "loss": 0.5564, + "step": 19562 + }, + { + "epoch": 0.58, + "learning_rate": 7.833786503335935e-05, + "loss": 0.5453, + "step": 19563 + }, + { + "epoch": 0.58, + "learning_rate": 7.832845167919925e-05, + "loss": 0.465, + "step": 19564 + }, + { + "epoch": 0.58, + "learning_rate": 7.83190385265326e-05, + "loss": 0.5949, + "step": 19565 + }, + { + "epoch": 0.58, + "learning_rate": 7.83096255754469e-05, + "loss": 0.5848, + "step": 19566 + }, + { + "epoch": 0.58, + "learning_rate": 7.830021282602965e-05, + "loss": 0.4609, + "step": 19567 + }, + { + "epoch": 0.58, + "learning_rate": 7.829080027836832e-05, + "loss": 0.4774, + "step": 19568 + }, + { + "epoch": 0.58, + "learning_rate": 7.828138793255052e-05, + "loss": 0.5049, + "step": 19569 + }, + { + "epoch": 0.58, + "learning_rate": 7.827197578866375e-05, + "loss": 0.4424, + "step": 19570 + }, + { + "epoch": 0.58, + "learning_rate": 7.826256384679542e-05, + "loss": 0.5828, + "step": 19571 + }, + { + "epoch": 0.58, + "learning_rate": 7.825315210703314e-05, + "loss": 0.5375, + "step": 19572 + }, + { + "epoch": 0.58, + "learning_rate": 7.824374056946437e-05, + "loss": 0.5922, + "step": 19573 + }, + { + "epoch": 0.58, + "learning_rate": 7.823432923417667e-05, + "loss": 0.5172, + "step": 19574 + }, + { + "epoch": 0.58, + "learning_rate": 7.822491810125748e-05, + "loss": 0.4955, + "step": 19575 + }, + { + "epoch": 0.58, + "learning_rate": 7.821550717079432e-05, + "loss": 0.5767, + "step": 19576 + }, + { + "epoch": 0.58, + "learning_rate": 7.820609644287467e-05, + "loss": 0.6794, + "step": 19577 + }, + { + "epoch": 0.58, + "learning_rate": 7.81966859175861e-05, + "loss": 0.5573, + "step": 19578 + }, + { + "epoch": 0.58, + "learning_rate": 7.818727559501602e-05, + "loss": 0.6447, + "step": 19579 + }, + { + "epoch": 0.58, + "learning_rate": 7.817786547525194e-05, + "loss": 0.5849, + "step": 19580 + }, + { + "epoch": 0.58, + "learning_rate": 7.81684555583814e-05, + "loss": 0.5684, + "step": 19581 + }, + { + "epoch": 0.58, + "learning_rate": 7.815904584449188e-05, + "loss": 0.4385, + "step": 19582 + }, + { + "epoch": 0.58, + "learning_rate": 7.81496363336708e-05, + "loss": 0.5616, + "step": 19583 + }, + { + "epoch": 0.58, + "learning_rate": 7.814022702600571e-05, + "loss": 0.5615, + "step": 19584 + }, + { + "epoch": 0.58, + "learning_rate": 7.81308179215841e-05, + "loss": 0.4957, + "step": 19585 + }, + { + "epoch": 0.58, + "learning_rate": 7.812140902049341e-05, + "loss": 0.5632, + "step": 19586 + }, + { + "epoch": 0.58, + "learning_rate": 7.811200032282112e-05, + "loss": 0.4445, + "step": 19587 + }, + { + "epoch": 0.58, + "learning_rate": 7.810259182865473e-05, + "loss": 0.5421, + "step": 19588 + }, + { + "epoch": 0.58, + "learning_rate": 7.809318353808175e-05, + "loss": 0.515, + "step": 19589 + }, + { + "epoch": 0.58, + "learning_rate": 7.808377545118958e-05, + "loss": 0.6059, + "step": 19590 + }, + { + "epoch": 0.58, + "learning_rate": 7.807436756806575e-05, + "loss": 0.5563, + "step": 19591 + }, + { + "epoch": 0.58, + "learning_rate": 7.806495988879768e-05, + "loss": 0.5325, + "step": 19592 + }, + { + "epoch": 0.58, + "learning_rate": 7.805555241347293e-05, + "loss": 0.573, + "step": 19593 + }, + { + "epoch": 0.58, + "learning_rate": 7.804614514217888e-05, + "loss": 0.4854, + "step": 19594 + }, + { + "epoch": 0.58, + "learning_rate": 7.8036738075003e-05, + "loss": 0.5558, + "step": 19595 + }, + { + "epoch": 0.58, + "learning_rate": 7.802733121203278e-05, + "loss": 0.482, + "step": 19596 + }, + { + "epoch": 0.58, + "learning_rate": 7.801792455335572e-05, + "loss": 0.4334, + "step": 19597 + }, + { + "epoch": 0.58, + "learning_rate": 7.800851809905921e-05, + "loss": 0.5608, + "step": 19598 + }, + { + "epoch": 0.58, + "learning_rate": 7.799911184923069e-05, + "loss": 0.5314, + "step": 19599 + }, + { + "epoch": 0.58, + "learning_rate": 7.798970580395771e-05, + "loss": 0.5155, + "step": 19600 + }, + { + "epoch": 0.58, + "learning_rate": 7.798029996332769e-05, + "loss": 0.4489, + "step": 19601 + }, + { + "epoch": 0.58, + "learning_rate": 7.797089432742802e-05, + "loss": 0.5219, + "step": 19602 + }, + { + "epoch": 0.58, + "learning_rate": 7.796148889634621e-05, + "loss": 0.5099, + "step": 19603 + }, + { + "epoch": 0.58, + "learning_rate": 7.795208367016967e-05, + "loss": 0.6096, + "step": 19604 + }, + { + "epoch": 0.58, + "learning_rate": 7.794267864898592e-05, + "loss": 0.5807, + "step": 19605 + }, + { + "epoch": 0.58, + "learning_rate": 7.793327383288234e-05, + "loss": 0.5334, + "step": 19606 + }, + { + "epoch": 0.58, + "learning_rate": 7.792386922194637e-05, + "loss": 0.4884, + "step": 19607 + }, + { + "epoch": 0.58, + "learning_rate": 7.791446481626545e-05, + "loss": 0.539, + "step": 19608 + }, + { + "epoch": 0.58, + "learning_rate": 7.790506061592707e-05, + "loss": 0.5693, + "step": 19609 + }, + { + "epoch": 0.58, + "learning_rate": 7.789565662101861e-05, + "loss": 0.5511, + "step": 19610 + }, + { + "epoch": 0.58, + "learning_rate": 7.78862528316275e-05, + "loss": 0.4322, + "step": 19611 + }, + { + "epoch": 0.58, + "learning_rate": 7.787684924784122e-05, + "loss": 0.5455, + "step": 19612 + }, + { + "epoch": 0.58, + "learning_rate": 7.786744586974722e-05, + "loss": 0.5431, + "step": 19613 + }, + { + "epoch": 0.58, + "learning_rate": 7.785804269743281e-05, + "loss": 0.5389, + "step": 19614 + }, + { + "epoch": 0.58, + "learning_rate": 7.784863973098554e-05, + "loss": 0.4968, + "step": 19615 + }, + { + "epoch": 0.58, + "learning_rate": 7.783923697049276e-05, + "loss": 0.5459, + "step": 19616 + }, + { + "epoch": 0.58, + "learning_rate": 7.782983441604199e-05, + "loss": 0.5253, + "step": 19617 + }, + { + "epoch": 0.58, + "learning_rate": 7.782043206772052e-05, + "loss": 0.448, + "step": 19618 + }, + { + "epoch": 0.58, + "learning_rate": 7.781102992561584e-05, + "loss": 0.5357, + "step": 19619 + }, + { + "epoch": 0.58, + "learning_rate": 7.780162798981539e-05, + "loss": 0.6819, + "step": 19620 + }, + { + "epoch": 0.58, + "learning_rate": 7.779222626040652e-05, + "loss": 0.5891, + "step": 19621 + }, + { + "epoch": 0.58, + "learning_rate": 7.77828247374767e-05, + "loss": 0.5951, + "step": 19622 + }, + { + "epoch": 0.58, + "learning_rate": 7.777342342111329e-05, + "loss": 0.5387, + "step": 19623 + }, + { + "epoch": 0.58, + "learning_rate": 7.776402231140378e-05, + "loss": 0.5705, + "step": 19624 + }, + { + "epoch": 0.58, + "learning_rate": 7.77546214084355e-05, + "loss": 0.5479, + "step": 19625 + }, + { + "epoch": 0.58, + "learning_rate": 7.774522071229587e-05, + "loss": 0.5211, + "step": 19626 + }, + { + "epoch": 0.58, + "learning_rate": 7.773582022307234e-05, + "loss": 0.6047, + "step": 19627 + }, + { + "epoch": 0.58, + "learning_rate": 7.772641994085228e-05, + "loss": 0.5905, + "step": 19628 + }, + { + "epoch": 0.58, + "learning_rate": 7.771701986572307e-05, + "loss": 0.4754, + "step": 19629 + }, + { + "epoch": 0.58, + "learning_rate": 7.770761999777212e-05, + "loss": 0.5478, + "step": 19630 + }, + { + "epoch": 0.58, + "learning_rate": 7.769822033708685e-05, + "loss": 0.5937, + "step": 19631 + }, + { + "epoch": 0.58, + "learning_rate": 7.768882088375467e-05, + "loss": 0.4475, + "step": 19632 + }, + { + "epoch": 0.58, + "learning_rate": 7.767942163786287e-05, + "loss": 0.5831, + "step": 19633 + }, + { + "epoch": 0.58, + "learning_rate": 7.767002259949895e-05, + "loss": 0.6302, + "step": 19634 + }, + { + "epoch": 0.58, + "learning_rate": 7.766062376875025e-05, + "loss": 0.5463, + "step": 19635 + }, + { + "epoch": 0.58, + "learning_rate": 7.76512251457042e-05, + "loss": 0.4553, + "step": 19636 + }, + { + "epoch": 0.58, + "learning_rate": 7.764182673044814e-05, + "loss": 0.5585, + "step": 19637 + }, + { + "epoch": 0.58, + "learning_rate": 7.763242852306944e-05, + "loss": 0.5283, + "step": 19638 + }, + { + "epoch": 0.58, + "learning_rate": 7.762303052365549e-05, + "loss": 0.5212, + "step": 19639 + }, + { + "epoch": 0.58, + "learning_rate": 7.761363273229374e-05, + "loss": 0.4578, + "step": 19640 + }, + { + "epoch": 0.58, + "learning_rate": 7.760423514907148e-05, + "loss": 0.5582, + "step": 19641 + }, + { + "epoch": 0.58, + "learning_rate": 7.75948377740761e-05, + "loss": 0.5348, + "step": 19642 + }, + { + "epoch": 0.58, + "learning_rate": 7.7585440607395e-05, + "loss": 0.5224, + "step": 19643 + }, + { + "epoch": 0.58, + "learning_rate": 7.757604364911558e-05, + "loss": 0.6414, + "step": 19644 + }, + { + "epoch": 0.58, + "learning_rate": 7.756664689932509e-05, + "loss": 0.5194, + "step": 19645 + }, + { + "epoch": 0.58, + "learning_rate": 7.755725035811104e-05, + "loss": 0.4801, + "step": 19646 + }, + { + "epoch": 0.58, + "learning_rate": 7.754785402556067e-05, + "loss": 0.5356, + "step": 19647 + }, + { + "epoch": 0.58, + "learning_rate": 7.75384579017615e-05, + "loss": 0.5437, + "step": 19648 + }, + { + "epoch": 0.58, + "learning_rate": 7.752906198680073e-05, + "loss": 0.5884, + "step": 19649 + }, + { + "epoch": 0.59, + "learning_rate": 7.751966628076579e-05, + "loss": 0.504, + "step": 19650 + }, + { + "epoch": 0.59, + "learning_rate": 7.751027078374407e-05, + "loss": 0.5258, + "step": 19651 + }, + { + "epoch": 0.59, + "learning_rate": 7.750087549582284e-05, + "loss": 0.5844, + "step": 19652 + }, + { + "epoch": 0.59, + "learning_rate": 7.749148041708951e-05, + "loss": 0.5485, + "step": 19653 + }, + { + "epoch": 0.59, + "learning_rate": 7.748208554763141e-05, + "loss": 0.6562, + "step": 19654 + }, + { + "epoch": 0.59, + "learning_rate": 7.747269088753596e-05, + "loss": 0.4709, + "step": 19655 + }, + { + "epoch": 0.59, + "learning_rate": 7.746329643689042e-05, + "loss": 0.4944, + "step": 19656 + }, + { + "epoch": 0.59, + "learning_rate": 7.745390219578214e-05, + "loss": 0.6319, + "step": 19657 + }, + { + "epoch": 0.59, + "learning_rate": 7.744450816429852e-05, + "loss": 0.4705, + "step": 19658 + }, + { + "epoch": 0.59, + "learning_rate": 7.743511434252691e-05, + "loss": 0.531, + "step": 19659 + }, + { + "epoch": 0.59, + "learning_rate": 7.742572073055457e-05, + "loss": 0.5146, + "step": 19660 + }, + { + "epoch": 0.59, + "learning_rate": 7.741632732846886e-05, + "loss": 0.612, + "step": 19661 + }, + { + "epoch": 0.59, + "learning_rate": 7.740693413635719e-05, + "loss": 0.5536, + "step": 19662 + }, + { + "epoch": 0.59, + "learning_rate": 7.739754115430682e-05, + "loss": 0.5862, + "step": 19663 + }, + { + "epoch": 0.59, + "learning_rate": 7.738814838240508e-05, + "loss": 0.4825, + "step": 19664 + }, + { + "epoch": 0.59, + "learning_rate": 7.737875582073935e-05, + "loss": 0.5449, + "step": 19665 + }, + { + "epoch": 0.59, + "learning_rate": 7.736936346939691e-05, + "loss": 0.543, + "step": 19666 + }, + { + "epoch": 0.59, + "learning_rate": 7.735997132846515e-05, + "loss": 0.6101, + "step": 19667 + }, + { + "epoch": 0.59, + "learning_rate": 7.735057939803133e-05, + "loss": 0.5863, + "step": 19668 + }, + { + "epoch": 0.59, + "learning_rate": 7.734118767818277e-05, + "loss": 0.4141, + "step": 19669 + }, + { + "epoch": 0.59, + "learning_rate": 7.733179616900684e-05, + "loss": 0.4735, + "step": 19670 + }, + { + "epoch": 0.59, + "learning_rate": 7.732240487059086e-05, + "loss": 0.5332, + "step": 19671 + }, + { + "epoch": 0.59, + "learning_rate": 7.73130137830221e-05, + "loss": 0.5104, + "step": 19672 + }, + { + "epoch": 0.59, + "learning_rate": 7.730362290638788e-05, + "loss": 0.5261, + "step": 19673 + }, + { + "epoch": 0.59, + "learning_rate": 7.729423224077553e-05, + "loss": 0.5282, + "step": 19674 + }, + { + "epoch": 0.59, + "learning_rate": 7.72848417862724e-05, + "loss": 0.553, + "step": 19675 + }, + { + "epoch": 0.59, + "learning_rate": 7.72754515429657e-05, + "loss": 0.4879, + "step": 19676 + }, + { + "epoch": 0.59, + "learning_rate": 7.726606151094283e-05, + "loss": 0.6433, + "step": 19677 + }, + { + "epoch": 0.59, + "learning_rate": 7.725667169029103e-05, + "loss": 0.5153, + "step": 19678 + }, + { + "epoch": 0.59, + "learning_rate": 7.724728208109767e-05, + "loss": 0.5436, + "step": 19679 + }, + { + "epoch": 0.59, + "learning_rate": 7.723789268345e-05, + "loss": 0.5672, + "step": 19680 + }, + { + "epoch": 0.59, + "learning_rate": 7.722850349743533e-05, + "loss": 0.6011, + "step": 19681 + }, + { + "epoch": 0.59, + "learning_rate": 7.721911452314094e-05, + "loss": 0.5276, + "step": 19682 + }, + { + "epoch": 0.59, + "learning_rate": 7.720972576065417e-05, + "loss": 0.5158, + "step": 19683 + }, + { + "epoch": 0.59, + "learning_rate": 7.720033721006227e-05, + "loss": 0.5733, + "step": 19684 + }, + { + "epoch": 0.59, + "learning_rate": 7.719094887145255e-05, + "loss": 0.4831, + "step": 19685 + }, + { + "epoch": 0.59, + "learning_rate": 7.718156074491231e-05, + "loss": 0.5179, + "step": 19686 + }, + { + "epoch": 0.59, + "learning_rate": 7.717217283052882e-05, + "loss": 0.4865, + "step": 19687 + }, + { + "epoch": 0.59, + "learning_rate": 7.716278512838935e-05, + "loss": 0.5395, + "step": 19688 + }, + { + "epoch": 0.59, + "learning_rate": 7.715339763858121e-05, + "loss": 0.5215, + "step": 19689 + }, + { + "epoch": 0.59, + "learning_rate": 7.71440103611917e-05, + "loss": 0.5272, + "step": 19690 + }, + { + "epoch": 0.59, + "learning_rate": 7.713462329630805e-05, + "loss": 0.6399, + "step": 19691 + }, + { + "epoch": 0.59, + "learning_rate": 7.712523644401754e-05, + "loss": 0.5891, + "step": 19692 + }, + { + "epoch": 0.59, + "learning_rate": 7.711584980440749e-05, + "loss": 0.4984, + "step": 19693 + }, + { + "epoch": 0.59, + "learning_rate": 7.710646337756516e-05, + "loss": 0.545, + "step": 19694 + }, + { + "epoch": 0.59, + "learning_rate": 7.709707716357778e-05, + "loss": 0.4879, + "step": 19695 + }, + { + "epoch": 0.59, + "learning_rate": 7.708769116253266e-05, + "loss": 0.4562, + "step": 19696 + }, + { + "epoch": 0.59, + "learning_rate": 7.707830537451704e-05, + "loss": 0.5116, + "step": 19697 + }, + { + "epoch": 0.59, + "learning_rate": 7.706891979961823e-05, + "loss": 0.5582, + "step": 19698 + }, + { + "epoch": 0.59, + "learning_rate": 7.705953443792345e-05, + "loss": 0.6, + "step": 19699 + }, + { + "epoch": 0.59, + "learning_rate": 7.705014928951996e-05, + "loss": 0.5022, + "step": 19700 + }, + { + "epoch": 0.59, + "learning_rate": 7.704076435449506e-05, + "loss": 0.4491, + "step": 19701 + }, + { + "epoch": 0.59, + "learning_rate": 7.703137963293598e-05, + "loss": 0.5348, + "step": 19702 + }, + { + "epoch": 0.59, + "learning_rate": 7.702199512492998e-05, + "loss": 0.5273, + "step": 19703 + }, + { + "epoch": 0.59, + "learning_rate": 7.701261083056428e-05, + "loss": 0.5143, + "step": 19704 + }, + { + "epoch": 0.59, + "learning_rate": 7.700322674992619e-05, + "loss": 0.5138, + "step": 19705 + }, + { + "epoch": 0.59, + "learning_rate": 7.699384288310294e-05, + "loss": 0.543, + "step": 19706 + }, + { + "epoch": 0.59, + "learning_rate": 7.698445923018173e-05, + "loss": 0.5908, + "step": 19707 + }, + { + "epoch": 0.59, + "learning_rate": 7.697507579124989e-05, + "loss": 0.5249, + "step": 19708 + }, + { + "epoch": 0.59, + "learning_rate": 7.696569256639457e-05, + "loss": 0.5671, + "step": 19709 + }, + { + "epoch": 0.59, + "learning_rate": 7.69563095557031e-05, + "loss": 0.467, + "step": 19710 + }, + { + "epoch": 0.59, + "learning_rate": 7.694692675926268e-05, + "loss": 0.5869, + "step": 19711 + }, + { + "epoch": 0.59, + "learning_rate": 7.693754417716054e-05, + "loss": 0.5671, + "step": 19712 + }, + { + "epoch": 0.59, + "learning_rate": 7.69281618094839e-05, + "loss": 0.5368, + "step": 19713 + }, + { + "epoch": 0.59, + "learning_rate": 7.691877965632007e-05, + "loss": 0.4444, + "step": 19714 + }, + { + "epoch": 0.59, + "learning_rate": 7.690939771775618e-05, + "loss": 0.6304, + "step": 19715 + }, + { + "epoch": 0.59, + "learning_rate": 7.69000159938795e-05, + "loss": 0.5726, + "step": 19716 + }, + { + "epoch": 0.59, + "learning_rate": 7.68906344847773e-05, + "loss": 0.6064, + "step": 19717 + }, + { + "epoch": 0.59, + "learning_rate": 7.688125319053676e-05, + "loss": 0.5514, + "step": 19718 + }, + { + "epoch": 0.59, + "learning_rate": 7.68718721112451e-05, + "loss": 0.5492, + "step": 19719 + }, + { + "epoch": 0.59, + "learning_rate": 7.686249124698955e-05, + "loss": 0.5484, + "step": 19720 + }, + { + "epoch": 0.59, + "learning_rate": 7.685311059785737e-05, + "loss": 0.4845, + "step": 19721 + }, + { + "epoch": 0.59, + "learning_rate": 7.684373016393573e-05, + "loss": 0.5304, + "step": 19722 + }, + { + "epoch": 0.59, + "learning_rate": 7.683434994531181e-05, + "loss": 0.569, + "step": 19723 + }, + { + "epoch": 0.59, + "learning_rate": 7.682496994207292e-05, + "loss": 0.4744, + "step": 19724 + }, + { + "epoch": 0.59, + "learning_rate": 7.681559015430623e-05, + "loss": 0.6151, + "step": 19725 + }, + { + "epoch": 0.59, + "learning_rate": 7.68062105820989e-05, + "loss": 0.5309, + "step": 19726 + }, + { + "epoch": 0.59, + "learning_rate": 7.67968312255382e-05, + "loss": 0.6992, + "step": 19727 + }, + { + "epoch": 0.59, + "learning_rate": 7.678745208471131e-05, + "loss": 0.6195, + "step": 19728 + }, + { + "epoch": 0.59, + "learning_rate": 7.677807315970547e-05, + "loss": 0.503, + "step": 19729 + }, + { + "epoch": 0.59, + "learning_rate": 7.676869445060783e-05, + "loss": 0.581, + "step": 19730 + }, + { + "epoch": 0.59, + "learning_rate": 7.675931595750559e-05, + "loss": 0.5234, + "step": 19731 + }, + { + "epoch": 0.59, + "learning_rate": 7.674993768048598e-05, + "loss": 0.5226, + "step": 19732 + }, + { + "epoch": 0.59, + "learning_rate": 7.674055961963621e-05, + "loss": 0.5793, + "step": 19733 + }, + { + "epoch": 0.59, + "learning_rate": 7.673118177504343e-05, + "loss": 0.4289, + "step": 19734 + }, + { + "epoch": 0.59, + "learning_rate": 7.672180414679482e-05, + "loss": 0.4632, + "step": 19735 + }, + { + "epoch": 0.59, + "learning_rate": 7.671242673497762e-05, + "loss": 0.6218, + "step": 19736 + }, + { + "epoch": 0.59, + "learning_rate": 7.670304953967901e-05, + "loss": 0.5079, + "step": 19737 + }, + { + "epoch": 0.59, + "learning_rate": 7.669367256098613e-05, + "loss": 0.5152, + "step": 19738 + }, + { + "epoch": 0.59, + "learning_rate": 7.668429579898622e-05, + "loss": 0.5121, + "step": 19739 + }, + { + "epoch": 0.59, + "learning_rate": 7.66749192537664e-05, + "loss": 0.4935, + "step": 19740 + }, + { + "epoch": 0.59, + "learning_rate": 7.666554292541392e-05, + "loss": 0.5622, + "step": 19741 + }, + { + "epoch": 0.59, + "learning_rate": 7.665616681401592e-05, + "loss": 0.5818, + "step": 19742 + }, + { + "epoch": 0.59, + "learning_rate": 7.664679091965956e-05, + "loss": 0.5462, + "step": 19743 + }, + { + "epoch": 0.59, + "learning_rate": 7.663741524243202e-05, + "loss": 0.4602, + "step": 19744 + }, + { + "epoch": 0.59, + "learning_rate": 7.662803978242052e-05, + "loss": 0.4762, + "step": 19745 + }, + { + "epoch": 0.59, + "learning_rate": 7.661866453971218e-05, + "loss": 0.5634, + "step": 19746 + }, + { + "epoch": 0.59, + "learning_rate": 7.660928951439413e-05, + "loss": 0.5509, + "step": 19747 + }, + { + "epoch": 0.59, + "learning_rate": 7.659991470655365e-05, + "loss": 0.4798, + "step": 19748 + }, + { + "epoch": 0.59, + "learning_rate": 7.65905401162778e-05, + "loss": 0.4262, + "step": 19749 + }, + { + "epoch": 0.59, + "learning_rate": 7.658116574365378e-05, + "loss": 0.5449, + "step": 19750 + }, + { + "epoch": 0.59, + "learning_rate": 7.657179158876875e-05, + "loss": 0.5962, + "step": 19751 + }, + { + "epoch": 0.59, + "learning_rate": 7.656241765170988e-05, + "loss": 0.544, + "step": 19752 + }, + { + "epoch": 0.59, + "learning_rate": 7.65530439325643e-05, + "loss": 0.5097, + "step": 19753 + }, + { + "epoch": 0.59, + "learning_rate": 7.654367043141915e-05, + "loss": 0.5587, + "step": 19754 + }, + { + "epoch": 0.59, + "learning_rate": 7.653429714836162e-05, + "loss": 0.6662, + "step": 19755 + }, + { + "epoch": 0.59, + "learning_rate": 7.652492408347886e-05, + "loss": 0.5869, + "step": 19756 + }, + { + "epoch": 0.59, + "learning_rate": 7.651555123685798e-05, + "loss": 0.6344, + "step": 19757 + }, + { + "epoch": 0.59, + "learning_rate": 7.650617860858615e-05, + "loss": 0.5953, + "step": 19758 + }, + { + "epoch": 0.59, + "learning_rate": 7.649680619875048e-05, + "loss": 0.583, + "step": 19759 + }, + { + "epoch": 0.59, + "learning_rate": 7.648743400743818e-05, + "loss": 0.3933, + "step": 19760 + }, + { + "epoch": 0.59, + "learning_rate": 7.647806203473634e-05, + "loss": 0.5127, + "step": 19761 + }, + { + "epoch": 0.59, + "learning_rate": 7.646869028073208e-05, + "loss": 0.5962, + "step": 19762 + }, + { + "epoch": 0.59, + "learning_rate": 7.645931874551258e-05, + "loss": 0.4831, + "step": 19763 + }, + { + "epoch": 0.59, + "learning_rate": 7.644994742916497e-05, + "loss": 0.5928, + "step": 19764 + }, + { + "epoch": 0.59, + "learning_rate": 7.644057633177634e-05, + "loss": 0.5672, + "step": 19765 + }, + { + "epoch": 0.59, + "learning_rate": 7.643120545343382e-05, + "loss": 0.531, + "step": 19766 + }, + { + "epoch": 0.59, + "learning_rate": 7.642183479422458e-05, + "loss": 0.5683, + "step": 19767 + }, + { + "epoch": 0.59, + "learning_rate": 7.641246435423574e-05, + "loss": 0.5207, + "step": 19768 + }, + { + "epoch": 0.59, + "learning_rate": 7.640309413355438e-05, + "loss": 0.5842, + "step": 19769 + }, + { + "epoch": 0.59, + "learning_rate": 7.639372413226766e-05, + "loss": 0.5275, + "step": 19770 + }, + { + "epoch": 0.59, + "learning_rate": 7.638435435046265e-05, + "loss": 0.502, + "step": 19771 + }, + { + "epoch": 0.59, + "learning_rate": 7.637498478822655e-05, + "loss": 0.5387, + "step": 19772 + }, + { + "epoch": 0.59, + "learning_rate": 7.636561544564641e-05, + "loss": 0.5782, + "step": 19773 + }, + { + "epoch": 0.59, + "learning_rate": 7.635624632280936e-05, + "loss": 0.7303, + "step": 19774 + }, + { + "epoch": 0.59, + "learning_rate": 7.634687741980248e-05, + "loss": 0.5617, + "step": 19775 + }, + { + "epoch": 0.59, + "learning_rate": 7.633750873671295e-05, + "loss": 0.4617, + "step": 19776 + }, + { + "epoch": 0.59, + "learning_rate": 7.63281402736278e-05, + "loss": 0.4983, + "step": 19777 + }, + { + "epoch": 0.59, + "learning_rate": 7.631877203063416e-05, + "loss": 0.5946, + "step": 19778 + }, + { + "epoch": 0.59, + "learning_rate": 7.630940400781916e-05, + "loss": 0.6534, + "step": 19779 + }, + { + "epoch": 0.59, + "learning_rate": 7.63000362052699e-05, + "loss": 0.4357, + "step": 19780 + }, + { + "epoch": 0.59, + "learning_rate": 7.629066862307341e-05, + "loss": 0.5192, + "step": 19781 + }, + { + "epoch": 0.59, + "learning_rate": 7.628130126131686e-05, + "loss": 0.5675, + "step": 19782 + }, + { + "epoch": 0.59, + "learning_rate": 7.627193412008734e-05, + "loss": 0.5377, + "step": 19783 + }, + { + "epoch": 0.59, + "learning_rate": 7.626256719947189e-05, + "loss": 0.6523, + "step": 19784 + }, + { + "epoch": 0.59, + "learning_rate": 7.625320049955762e-05, + "loss": 0.5855, + "step": 19785 + }, + { + "epoch": 0.59, + "learning_rate": 7.624383402043164e-05, + "loss": 0.5501, + "step": 19786 + }, + { + "epoch": 0.59, + "learning_rate": 7.623446776218104e-05, + "loss": 0.5804, + "step": 19787 + }, + { + "epoch": 0.59, + "learning_rate": 7.622510172489286e-05, + "loss": 0.5818, + "step": 19788 + }, + { + "epoch": 0.59, + "learning_rate": 7.621573590865422e-05, + "loss": 0.5, + "step": 19789 + }, + { + "epoch": 0.59, + "learning_rate": 7.620637031355216e-05, + "loss": 0.5739, + "step": 19790 + }, + { + "epoch": 0.59, + "learning_rate": 7.619700493967384e-05, + "loss": 0.556, + "step": 19791 + }, + { + "epoch": 0.59, + "learning_rate": 7.618763978710626e-05, + "loss": 0.5618, + "step": 19792 + }, + { + "epoch": 0.59, + "learning_rate": 7.617827485593649e-05, + "loss": 0.513, + "step": 19793 + }, + { + "epoch": 0.59, + "learning_rate": 7.616891014625164e-05, + "loss": 0.4329, + "step": 19794 + }, + { + "epoch": 0.59, + "learning_rate": 7.615954565813879e-05, + "loss": 0.5712, + "step": 19795 + }, + { + "epoch": 0.59, + "learning_rate": 7.615018139168496e-05, + "loss": 0.5557, + "step": 19796 + }, + { + "epoch": 0.59, + "learning_rate": 7.614081734697722e-05, + "loss": 0.4296, + "step": 19797 + }, + { + "epoch": 0.59, + "learning_rate": 7.613145352410267e-05, + "loss": 0.5122, + "step": 19798 + }, + { + "epoch": 0.59, + "learning_rate": 7.612208992314838e-05, + "loss": 0.6226, + "step": 19799 + }, + { + "epoch": 0.59, + "learning_rate": 7.611272654420133e-05, + "loss": 0.4643, + "step": 19800 + }, + { + "epoch": 0.59, + "learning_rate": 7.610336338734865e-05, + "loss": 0.6037, + "step": 19801 + }, + { + "epoch": 0.59, + "learning_rate": 7.609400045267736e-05, + "loss": 0.7254, + "step": 19802 + }, + { + "epoch": 0.59, + "learning_rate": 7.608463774027457e-05, + "loss": 0.5025, + "step": 19803 + }, + { + "epoch": 0.59, + "learning_rate": 7.607527525022725e-05, + "loss": 0.5108, + "step": 19804 + }, + { + "epoch": 0.59, + "learning_rate": 7.606591298262248e-05, + "loss": 0.4872, + "step": 19805 + }, + { + "epoch": 0.59, + "learning_rate": 7.605655093754732e-05, + "loss": 0.5332, + "step": 19806 + }, + { + "epoch": 0.59, + "learning_rate": 7.604718911508884e-05, + "loss": 0.372, + "step": 19807 + }, + { + "epoch": 0.59, + "learning_rate": 7.603782751533401e-05, + "loss": 0.4818, + "step": 19808 + }, + { + "epoch": 0.59, + "learning_rate": 7.602846613836991e-05, + "loss": 0.5455, + "step": 19809 + }, + { + "epoch": 0.59, + "learning_rate": 7.601910498428359e-05, + "loss": 0.5954, + "step": 19810 + }, + { + "epoch": 0.59, + "learning_rate": 7.60097440531621e-05, + "loss": 0.5996, + "step": 19811 + }, + { + "epoch": 0.59, + "learning_rate": 7.600038334509238e-05, + "loss": 0.6497, + "step": 19812 + }, + { + "epoch": 0.59, + "learning_rate": 7.599102286016157e-05, + "loss": 0.5184, + "step": 19813 + }, + { + "epoch": 0.59, + "learning_rate": 7.59816625984567e-05, + "loss": 0.5816, + "step": 19814 + }, + { + "epoch": 0.59, + "learning_rate": 7.597230256006471e-05, + "loss": 0.5824, + "step": 19815 + }, + { + "epoch": 0.59, + "learning_rate": 7.596294274507266e-05, + "loss": 0.505, + "step": 19816 + }, + { + "epoch": 0.59, + "learning_rate": 7.595358315356763e-05, + "loss": 0.5245, + "step": 19817 + }, + { + "epoch": 0.59, + "learning_rate": 7.594422378563659e-05, + "loss": 0.5869, + "step": 19818 + }, + { + "epoch": 0.59, + "learning_rate": 7.593486464136654e-05, + "loss": 0.4738, + "step": 19819 + }, + { + "epoch": 0.59, + "learning_rate": 7.592550572084454e-05, + "loss": 0.5422, + "step": 19820 + }, + { + "epoch": 0.59, + "learning_rate": 7.591614702415759e-05, + "loss": 0.5462, + "step": 19821 + }, + { + "epoch": 0.59, + "learning_rate": 7.590678855139274e-05, + "loss": 0.5356, + "step": 19822 + }, + { + "epoch": 0.59, + "learning_rate": 7.589743030263694e-05, + "loss": 0.4528, + "step": 19823 + }, + { + "epoch": 0.59, + "learning_rate": 7.588807227797721e-05, + "loss": 0.6185, + "step": 19824 + }, + { + "epoch": 0.59, + "learning_rate": 7.58787144775006e-05, + "loss": 0.6583, + "step": 19825 + }, + { + "epoch": 0.59, + "learning_rate": 7.586935690129411e-05, + "loss": 0.5179, + "step": 19826 + }, + { + "epoch": 0.59, + "learning_rate": 7.58599995494447e-05, + "loss": 0.569, + "step": 19827 + }, + { + "epoch": 0.59, + "learning_rate": 7.585064242203937e-05, + "loss": 0.4689, + "step": 19828 + }, + { + "epoch": 0.59, + "learning_rate": 7.584128551916515e-05, + "loss": 0.4459, + "step": 19829 + }, + { + "epoch": 0.59, + "learning_rate": 7.583192884090907e-05, + "loss": 0.4675, + "step": 19830 + }, + { + "epoch": 0.59, + "learning_rate": 7.582257238735803e-05, + "loss": 0.4668, + "step": 19831 + }, + { + "epoch": 0.59, + "learning_rate": 7.58132161585991e-05, + "loss": 0.4214, + "step": 19832 + }, + { + "epoch": 0.59, + "learning_rate": 7.580386015471923e-05, + "loss": 0.545, + "step": 19833 + }, + { + "epoch": 0.59, + "learning_rate": 7.579450437580546e-05, + "loss": 0.6173, + "step": 19834 + }, + { + "epoch": 0.59, + "learning_rate": 7.578514882194472e-05, + "loss": 0.473, + "step": 19835 + }, + { + "epoch": 0.59, + "learning_rate": 7.5775793493224e-05, + "loss": 0.4476, + "step": 19836 + }, + { + "epoch": 0.59, + "learning_rate": 7.57664383897303e-05, + "loss": 0.5194, + "step": 19837 + }, + { + "epoch": 0.59, + "learning_rate": 7.575708351155063e-05, + "loss": 0.6583, + "step": 19838 + }, + { + "epoch": 0.59, + "learning_rate": 7.574772885877191e-05, + "loss": 0.5126, + "step": 19839 + }, + { + "epoch": 0.59, + "learning_rate": 7.573837443148113e-05, + "loss": 0.5351, + "step": 19840 + }, + { + "epoch": 0.59, + "learning_rate": 7.572902022976528e-05, + "loss": 0.6012, + "step": 19841 + }, + { + "epoch": 0.59, + "learning_rate": 7.571966625371138e-05, + "loss": 0.5919, + "step": 19842 + }, + { + "epoch": 0.59, + "learning_rate": 7.571031250340628e-05, + "loss": 0.5545, + "step": 19843 + }, + { + "epoch": 0.59, + "learning_rate": 7.570095897893703e-05, + "loss": 0.6096, + "step": 19844 + }, + { + "epoch": 0.59, + "learning_rate": 7.569160568039057e-05, + "loss": 0.5313, + "step": 19845 + }, + { + "epoch": 0.59, + "learning_rate": 7.56822526078539e-05, + "loss": 0.5267, + "step": 19846 + }, + { + "epoch": 0.59, + "learning_rate": 7.567289976141395e-05, + "loss": 0.5036, + "step": 19847 + }, + { + "epoch": 0.59, + "learning_rate": 7.566354714115768e-05, + "loss": 0.4595, + "step": 19848 + }, + { + "epoch": 0.59, + "learning_rate": 7.565419474717206e-05, + "loss": 0.5017, + "step": 19849 + }, + { + "epoch": 0.59, + "learning_rate": 7.5644842579544e-05, + "loss": 0.4672, + "step": 19850 + }, + { + "epoch": 0.59, + "learning_rate": 7.56354906383605e-05, + "loss": 0.6234, + "step": 19851 + }, + { + "epoch": 0.59, + "learning_rate": 7.562613892370847e-05, + "loss": 0.5342, + "step": 19852 + }, + { + "epoch": 0.59, + "learning_rate": 7.561678743567495e-05, + "loss": 0.5208, + "step": 19853 + }, + { + "epoch": 0.59, + "learning_rate": 7.560743617434679e-05, + "loss": 0.5945, + "step": 19854 + }, + { + "epoch": 0.59, + "learning_rate": 7.559808513981094e-05, + "loss": 0.4689, + "step": 19855 + }, + { + "epoch": 0.59, + "learning_rate": 7.558873433215439e-05, + "loss": 0.5802, + "step": 19856 + }, + { + "epoch": 0.59, + "learning_rate": 7.557938375146408e-05, + "loss": 0.6371, + "step": 19857 + }, + { + "epoch": 0.59, + "learning_rate": 7.557003339782691e-05, + "loss": 0.5272, + "step": 19858 + }, + { + "epoch": 0.59, + "learning_rate": 7.556068327132982e-05, + "loss": 0.6047, + "step": 19859 + }, + { + "epoch": 0.59, + "learning_rate": 7.555133337205976e-05, + "loss": 0.529, + "step": 19860 + }, + { + "epoch": 0.59, + "learning_rate": 7.554198370010369e-05, + "loss": 0.4517, + "step": 19861 + }, + { + "epoch": 0.59, + "learning_rate": 7.553263425554847e-05, + "loss": 0.4972, + "step": 19862 + }, + { + "epoch": 0.59, + "learning_rate": 7.552328503848107e-05, + "loss": 0.5849, + "step": 19863 + }, + { + "epoch": 0.59, + "learning_rate": 7.55139360489884e-05, + "loss": 0.4959, + "step": 19864 + }, + { + "epoch": 0.59, + "learning_rate": 7.550458728715744e-05, + "loss": 0.5935, + "step": 19865 + }, + { + "epoch": 0.59, + "learning_rate": 7.549523875307503e-05, + "loss": 0.5735, + "step": 19866 + }, + { + "epoch": 0.59, + "learning_rate": 7.548589044682812e-05, + "loss": 0.4832, + "step": 19867 + }, + { + "epoch": 0.59, + "learning_rate": 7.547654236850363e-05, + "loss": 0.5157, + "step": 19868 + }, + { + "epoch": 0.59, + "learning_rate": 7.54671945181885e-05, + "loss": 0.5177, + "step": 19869 + }, + { + "epoch": 0.59, + "learning_rate": 7.545784689596962e-05, + "loss": 0.5084, + "step": 19870 + }, + { + "epoch": 0.59, + "learning_rate": 7.544849950193386e-05, + "loss": 0.5849, + "step": 19871 + }, + { + "epoch": 0.59, + "learning_rate": 7.543915233616818e-05, + "loss": 0.5398, + "step": 19872 + }, + { + "epoch": 0.59, + "learning_rate": 7.542980539875951e-05, + "loss": 0.4879, + "step": 19873 + }, + { + "epoch": 0.59, + "learning_rate": 7.542045868979466e-05, + "loss": 0.5791, + "step": 19874 + }, + { + "epoch": 0.59, + "learning_rate": 7.541111220936062e-05, + "loss": 0.5415, + "step": 19875 + }, + { + "epoch": 0.59, + "learning_rate": 7.540176595754424e-05, + "loss": 0.5255, + "step": 19876 + }, + { + "epoch": 0.59, + "learning_rate": 7.539241993443247e-05, + "loss": 0.5354, + "step": 19877 + }, + { + "epoch": 0.59, + "learning_rate": 7.538307414011215e-05, + "loss": 0.5783, + "step": 19878 + }, + { + "epoch": 0.59, + "learning_rate": 7.53737285746702e-05, + "loss": 0.6789, + "step": 19879 + }, + { + "epoch": 0.59, + "learning_rate": 7.536438323819353e-05, + "loss": 0.4994, + "step": 19880 + }, + { + "epoch": 0.59, + "learning_rate": 7.535503813076895e-05, + "loss": 0.484, + "step": 19881 + }, + { + "epoch": 0.59, + "learning_rate": 7.534569325248345e-05, + "loss": 0.5195, + "step": 19882 + }, + { + "epoch": 0.59, + "learning_rate": 7.533634860342381e-05, + "loss": 0.5159, + "step": 19883 + }, + { + "epoch": 0.59, + "learning_rate": 7.532700418367704e-05, + "loss": 0.5479, + "step": 19884 + }, + { + "epoch": 0.59, + "learning_rate": 7.531765999332991e-05, + "loss": 0.5232, + "step": 19885 + }, + { + "epoch": 0.59, + "learning_rate": 7.530831603246934e-05, + "loss": 0.4791, + "step": 19886 + }, + { + "epoch": 0.59, + "learning_rate": 7.529897230118223e-05, + "loss": 0.5311, + "step": 19887 + }, + { + "epoch": 0.59, + "learning_rate": 7.528962879955544e-05, + "loss": 0.5562, + "step": 19888 + }, + { + "epoch": 0.59, + "learning_rate": 7.528028552767581e-05, + "loss": 0.5993, + "step": 19889 + }, + { + "epoch": 0.59, + "learning_rate": 7.527094248563022e-05, + "loss": 0.4678, + "step": 19890 + }, + { + "epoch": 0.59, + "learning_rate": 7.526159967350555e-05, + "loss": 0.6346, + "step": 19891 + }, + { + "epoch": 0.59, + "learning_rate": 7.525225709138871e-05, + "loss": 0.6727, + "step": 19892 + }, + { + "epoch": 0.59, + "learning_rate": 7.524291473936648e-05, + "loss": 0.4567, + "step": 19893 + }, + { + "epoch": 0.59, + "learning_rate": 7.523357261752577e-05, + "loss": 0.4843, + "step": 19894 + }, + { + "epoch": 0.59, + "learning_rate": 7.522423072595341e-05, + "loss": 0.6035, + "step": 19895 + }, + { + "epoch": 0.59, + "learning_rate": 7.521488906473633e-05, + "loss": 0.4887, + "step": 19896 + }, + { + "epoch": 0.59, + "learning_rate": 7.52055476339613e-05, + "loss": 0.48, + "step": 19897 + }, + { + "epoch": 0.59, + "learning_rate": 7.51962064337152e-05, + "loss": 0.4522, + "step": 19898 + }, + { + "epoch": 0.59, + "learning_rate": 7.51868654640849e-05, + "loss": 0.4578, + "step": 19899 + }, + { + "epoch": 0.59, + "learning_rate": 7.517752472515726e-05, + "loss": 0.4974, + "step": 19900 + }, + { + "epoch": 0.59, + "learning_rate": 7.516818421701908e-05, + "loss": 0.6266, + "step": 19901 + }, + { + "epoch": 0.59, + "learning_rate": 7.515884393975719e-05, + "loss": 0.5332, + "step": 19902 + }, + { + "epoch": 0.59, + "learning_rate": 7.51495038934585e-05, + "loss": 0.4894, + "step": 19903 + }, + { + "epoch": 0.59, + "learning_rate": 7.514016407820986e-05, + "loss": 0.4978, + "step": 19904 + }, + { + "epoch": 0.59, + "learning_rate": 7.513082449409799e-05, + "loss": 0.6131, + "step": 19905 + }, + { + "epoch": 0.59, + "learning_rate": 7.512148514120984e-05, + "loss": 0.5469, + "step": 19906 + }, + { + "epoch": 0.59, + "learning_rate": 7.511214601963219e-05, + "loss": 0.4872, + "step": 19907 + }, + { + "epoch": 0.59, + "learning_rate": 7.510280712945191e-05, + "loss": 0.6419, + "step": 19908 + }, + { + "epoch": 0.59, + "learning_rate": 7.509346847075581e-05, + "loss": 0.6042, + "step": 19909 + }, + { + "epoch": 0.59, + "learning_rate": 7.508413004363069e-05, + "loss": 0.4577, + "step": 19910 + }, + { + "epoch": 0.59, + "learning_rate": 7.50747918481634e-05, + "loss": 0.4953, + "step": 19911 + }, + { + "epoch": 0.59, + "learning_rate": 7.506545388444079e-05, + "loss": 0.6181, + "step": 19912 + }, + { + "epoch": 0.59, + "learning_rate": 7.505611615254962e-05, + "loss": 0.5603, + "step": 19913 + }, + { + "epoch": 0.59, + "learning_rate": 7.504677865257673e-05, + "loss": 0.5867, + "step": 19914 + }, + { + "epoch": 0.59, + "learning_rate": 7.503744138460899e-05, + "loss": 0.5387, + "step": 19915 + }, + { + "epoch": 0.59, + "learning_rate": 7.502810434873315e-05, + "loss": 0.5434, + "step": 19916 + }, + { + "epoch": 0.59, + "learning_rate": 7.501876754503603e-05, + "loss": 0.5919, + "step": 19917 + }, + { + "epoch": 0.59, + "learning_rate": 7.500943097360446e-05, + "loss": 0.3815, + "step": 19918 + }, + { + "epoch": 0.59, + "learning_rate": 7.500009463452527e-05, + "loss": 0.634, + "step": 19919 + }, + { + "epoch": 0.59, + "learning_rate": 7.499075852788522e-05, + "loss": 0.5017, + "step": 19920 + }, + { + "epoch": 0.59, + "learning_rate": 7.498142265377108e-05, + "loss": 0.6172, + "step": 19921 + }, + { + "epoch": 0.59, + "learning_rate": 7.497208701226976e-05, + "loss": 0.5564, + "step": 19922 + }, + { + "epoch": 0.59, + "learning_rate": 7.4962751603468e-05, + "loss": 0.4884, + "step": 19923 + }, + { + "epoch": 0.59, + "learning_rate": 7.495341642745255e-05, + "loss": 0.4402, + "step": 19924 + }, + { + "epoch": 0.59, + "learning_rate": 7.494408148431029e-05, + "loss": 0.5808, + "step": 19925 + }, + { + "epoch": 0.59, + "learning_rate": 7.493474677412794e-05, + "loss": 0.7244, + "step": 19926 + }, + { + "epoch": 0.59, + "learning_rate": 7.492541229699237e-05, + "loss": 0.5304, + "step": 19927 + }, + { + "epoch": 0.59, + "learning_rate": 7.49160780529903e-05, + "loss": 0.4453, + "step": 19928 + }, + { + "epoch": 0.59, + "learning_rate": 7.490674404220852e-05, + "loss": 0.4754, + "step": 19929 + }, + { + "epoch": 0.59, + "learning_rate": 7.489741026473385e-05, + "loss": 0.4783, + "step": 19930 + }, + { + "epoch": 0.59, + "learning_rate": 7.488807672065308e-05, + "loss": 0.5037, + "step": 19931 + }, + { + "epoch": 0.59, + "learning_rate": 7.487874341005294e-05, + "loss": 0.6044, + "step": 19932 + }, + { + "epoch": 0.59, + "learning_rate": 7.486941033302021e-05, + "loss": 0.5114, + "step": 19933 + }, + { + "epoch": 0.59, + "learning_rate": 7.486007748964172e-05, + "loss": 0.6271, + "step": 19934 + }, + { + "epoch": 0.59, + "learning_rate": 7.485074488000421e-05, + "loss": 0.6666, + "step": 19935 + }, + { + "epoch": 0.59, + "learning_rate": 7.484141250419443e-05, + "loss": 0.4875, + "step": 19936 + }, + { + "epoch": 0.59, + "learning_rate": 7.483208036229918e-05, + "loss": 0.6129, + "step": 19937 + }, + { + "epoch": 0.59, + "learning_rate": 7.482274845440519e-05, + "loss": 0.5327, + "step": 19938 + }, + { + "epoch": 0.59, + "learning_rate": 7.48134167805993e-05, + "loss": 0.5497, + "step": 19939 + }, + { + "epoch": 0.59, + "learning_rate": 7.480408534096819e-05, + "loss": 0.5929, + "step": 19940 + }, + { + "epoch": 0.59, + "learning_rate": 7.479475413559867e-05, + "loss": 0.5478, + "step": 19941 + }, + { + "epoch": 0.59, + "learning_rate": 7.478542316457744e-05, + "loss": 0.5325, + "step": 19942 + }, + { + "epoch": 0.59, + "learning_rate": 7.477609242799136e-05, + "loss": 0.57, + "step": 19943 + }, + { + "epoch": 0.59, + "learning_rate": 7.476676192592708e-05, + "loss": 0.5958, + "step": 19944 + }, + { + "epoch": 0.59, + "learning_rate": 7.475743165847138e-05, + "loss": 0.5111, + "step": 19945 + }, + { + "epoch": 0.59, + "learning_rate": 7.474810162571104e-05, + "loss": 0.5123, + "step": 19946 + }, + { + "epoch": 0.59, + "learning_rate": 7.473877182773279e-05, + "loss": 0.5079, + "step": 19947 + }, + { + "epoch": 0.59, + "learning_rate": 7.472944226462336e-05, + "loss": 0.4913, + "step": 19948 + }, + { + "epoch": 0.59, + "learning_rate": 7.47201129364695e-05, + "loss": 0.4934, + "step": 19949 + }, + { + "epoch": 0.59, + "learning_rate": 7.471078384335797e-05, + "loss": 0.5622, + "step": 19950 + }, + { + "epoch": 0.59, + "learning_rate": 7.470145498537548e-05, + "loss": 0.5295, + "step": 19951 + }, + { + "epoch": 0.59, + "learning_rate": 7.469212636260877e-05, + "loss": 0.5347, + "step": 19952 + }, + { + "epoch": 0.59, + "learning_rate": 7.46827979751446e-05, + "loss": 0.5025, + "step": 19953 + }, + { + "epoch": 0.59, + "learning_rate": 7.467346982306971e-05, + "loss": 0.6056, + "step": 19954 + }, + { + "epoch": 0.59, + "learning_rate": 7.466414190647075e-05, + "loss": 0.4667, + "step": 19955 + }, + { + "epoch": 0.59, + "learning_rate": 7.465481422543452e-05, + "loss": 0.52, + "step": 19956 + }, + { + "epoch": 0.59, + "learning_rate": 7.464548678004771e-05, + "loss": 0.6081, + "step": 19957 + }, + { + "epoch": 0.59, + "learning_rate": 7.463615957039709e-05, + "loss": 0.6385, + "step": 19958 + }, + { + "epoch": 0.59, + "learning_rate": 7.462683259656934e-05, + "loss": 0.5811, + "step": 19959 + }, + { + "epoch": 0.59, + "learning_rate": 7.461750585865117e-05, + "loss": 0.5211, + "step": 19960 + }, + { + "epoch": 0.59, + "learning_rate": 7.460817935672933e-05, + "loss": 0.5755, + "step": 19961 + }, + { + "epoch": 0.59, + "learning_rate": 7.459885309089054e-05, + "loss": 0.5243, + "step": 19962 + }, + { + "epoch": 0.59, + "learning_rate": 7.458952706122146e-05, + "loss": 0.6519, + "step": 19963 + }, + { + "epoch": 0.59, + "learning_rate": 7.458020126780882e-05, + "loss": 0.5425, + "step": 19964 + }, + { + "epoch": 0.59, + "learning_rate": 7.457087571073936e-05, + "loss": 0.5541, + "step": 19965 + }, + { + "epoch": 0.59, + "learning_rate": 7.456155039009978e-05, + "loss": 0.5496, + "step": 19966 + }, + { + "epoch": 0.59, + "learning_rate": 7.455222530597672e-05, + "loss": 0.4583, + "step": 19967 + }, + { + "epoch": 0.59, + "learning_rate": 7.454290045845695e-05, + "loss": 0.5214, + "step": 19968 + }, + { + "epoch": 0.59, + "learning_rate": 7.453357584762712e-05, + "loss": 0.5521, + "step": 19969 + }, + { + "epoch": 0.59, + "learning_rate": 7.4524251473574e-05, + "loss": 0.529, + "step": 19970 + }, + { + "epoch": 0.59, + "learning_rate": 7.451492733638421e-05, + "loss": 0.5933, + "step": 19971 + }, + { + "epoch": 0.59, + "learning_rate": 7.450560343614446e-05, + "loss": 0.5005, + "step": 19972 + }, + { + "epoch": 0.59, + "learning_rate": 7.449627977294144e-05, + "loss": 0.6177, + "step": 19973 + }, + { + "epoch": 0.59, + "learning_rate": 7.44869563468619e-05, + "loss": 0.574, + "step": 19974 + }, + { + "epoch": 0.59, + "learning_rate": 7.447763315799244e-05, + "loss": 0.5236, + "step": 19975 + }, + { + "epoch": 0.59, + "learning_rate": 7.446831020641975e-05, + "loss": 0.5615, + "step": 19976 + }, + { + "epoch": 0.59, + "learning_rate": 7.445898749223055e-05, + "loss": 0.6428, + "step": 19977 + }, + { + "epoch": 0.59, + "learning_rate": 7.444966501551154e-05, + "loss": 0.4536, + "step": 19978 + }, + { + "epoch": 0.59, + "learning_rate": 7.444034277634931e-05, + "loss": 0.6061, + "step": 19979 + }, + { + "epoch": 0.59, + "learning_rate": 7.44310207748306e-05, + "loss": 0.4842, + "step": 19980 + }, + { + "epoch": 0.59, + "learning_rate": 7.44216990110421e-05, + "loss": 0.6558, + "step": 19981 + }, + { + "epoch": 0.59, + "learning_rate": 7.441237748507043e-05, + "loss": 0.5115, + "step": 19982 + }, + { + "epoch": 0.59, + "learning_rate": 7.440305619700225e-05, + "loss": 0.5875, + "step": 19983 + }, + { + "epoch": 0.59, + "learning_rate": 7.439373514692427e-05, + "loss": 0.529, + "step": 19984 + }, + { + "epoch": 0.59, + "learning_rate": 7.438441433492315e-05, + "loss": 0.4509, + "step": 19985 + }, + { + "epoch": 0.6, + "learning_rate": 7.437509376108552e-05, + "loss": 0.6321, + "step": 19986 + }, + { + "epoch": 0.6, + "learning_rate": 7.436577342549805e-05, + "loss": 0.5763, + "step": 19987 + }, + { + "epoch": 0.6, + "learning_rate": 7.43564533282474e-05, + "loss": 0.5738, + "step": 19988 + }, + { + "epoch": 0.6, + "learning_rate": 7.434713346942025e-05, + "loss": 0.4376, + "step": 19989 + }, + { + "epoch": 0.6, + "learning_rate": 7.433781384910321e-05, + "loss": 0.5396, + "step": 19990 + }, + { + "epoch": 0.6, + "learning_rate": 7.432849446738294e-05, + "loss": 0.4749, + "step": 19991 + }, + { + "epoch": 0.6, + "learning_rate": 7.431917532434611e-05, + "loss": 0.5887, + "step": 19992 + }, + { + "epoch": 0.6, + "learning_rate": 7.430985642007937e-05, + "loss": 0.5617, + "step": 19993 + }, + { + "epoch": 0.6, + "learning_rate": 7.430053775466934e-05, + "loss": 0.6266, + "step": 19994 + }, + { + "epoch": 0.6, + "learning_rate": 7.429121932820264e-05, + "loss": 0.4888, + "step": 19995 + }, + { + "epoch": 0.6, + "learning_rate": 7.428190114076595e-05, + "loss": 0.5424, + "step": 19996 + }, + { + "epoch": 0.6, + "learning_rate": 7.427258319244591e-05, + "loss": 0.4433, + "step": 19997 + }, + { + "epoch": 0.6, + "learning_rate": 7.42632654833291e-05, + "loss": 0.5987, + "step": 19998 + }, + { + "epoch": 0.6, + "learning_rate": 7.425394801350222e-05, + "loss": 0.6385, + "step": 19999 + }, + { + "epoch": 0.6, + "learning_rate": 7.424463078305184e-05, + "loss": 0.5217, + "step": 20000 + }, + { + "epoch": 0.6, + "learning_rate": 7.423531379206466e-05, + "loss": 0.6068, + "step": 20001 + }, + { + "epoch": 0.6, + "learning_rate": 7.422599704062725e-05, + "loss": 0.594, + "step": 20002 + }, + { + "epoch": 0.6, + "learning_rate": 7.421668052882623e-05, + "loss": 0.5359, + "step": 20003 + }, + { + "epoch": 0.6, + "learning_rate": 7.420736425674823e-05, + "loss": 0.503, + "step": 20004 + }, + { + "epoch": 0.6, + "learning_rate": 7.419804822447993e-05, + "loss": 0.4691, + "step": 20005 + }, + { + "epoch": 0.6, + "learning_rate": 7.418873243210787e-05, + "loss": 0.4261, + "step": 20006 + }, + { + "epoch": 0.6, + "learning_rate": 7.417941687971866e-05, + "loss": 0.4413, + "step": 20007 + }, + { + "epoch": 0.6, + "learning_rate": 7.417010156739896e-05, + "loss": 0.4936, + "step": 20008 + }, + { + "epoch": 0.6, + "learning_rate": 7.416078649523537e-05, + "loss": 0.6673, + "step": 20009 + }, + { + "epoch": 0.6, + "learning_rate": 7.415147166331446e-05, + "loss": 0.51, + "step": 20010 + }, + { + "epoch": 0.6, + "learning_rate": 7.414215707172289e-05, + "loss": 0.5257, + "step": 20011 + }, + { + "epoch": 0.6, + "learning_rate": 7.413284272054726e-05, + "loss": 0.6009, + "step": 20012 + }, + { + "epoch": 0.6, + "learning_rate": 7.412352860987409e-05, + "loss": 0.4893, + "step": 20013 + }, + { + "epoch": 0.6, + "learning_rate": 7.411421473979008e-05, + "loss": 0.5708, + "step": 20014 + }, + { + "epoch": 0.6, + "learning_rate": 7.410490111038177e-05, + "loss": 0.5373, + "step": 20015 + }, + { + "epoch": 0.6, + "learning_rate": 7.40955877217358e-05, + "loss": 0.5854, + "step": 20016 + }, + { + "epoch": 0.6, + "learning_rate": 7.408627457393867e-05, + "loss": 0.5643, + "step": 20017 + }, + { + "epoch": 0.6, + "learning_rate": 7.407696166707707e-05, + "loss": 0.6185, + "step": 20018 + }, + { + "epoch": 0.6, + "learning_rate": 7.406764900123751e-05, + "loss": 0.6174, + "step": 20019 + }, + { + "epoch": 0.6, + "learning_rate": 7.405833657650667e-05, + "loss": 0.6141, + "step": 20020 + }, + { + "epoch": 0.6, + "learning_rate": 7.404902439297105e-05, + "loss": 0.5747, + "step": 20021 + }, + { + "epoch": 0.6, + "learning_rate": 7.403971245071725e-05, + "loss": 0.5743, + "step": 20022 + }, + { + "epoch": 0.6, + "learning_rate": 7.403040074983186e-05, + "loss": 0.5107, + "step": 20023 + }, + { + "epoch": 0.6, + "learning_rate": 7.402108929040148e-05, + "loss": 0.4624, + "step": 20024 + }, + { + "epoch": 0.6, + "learning_rate": 7.401177807251265e-05, + "loss": 0.5239, + "step": 20025 + }, + { + "epoch": 0.6, + "learning_rate": 7.400246709625191e-05, + "loss": 0.509, + "step": 20026 + }, + { + "epoch": 0.6, + "learning_rate": 7.39931563617059e-05, + "loss": 0.5137, + "step": 20027 + }, + { + "epoch": 0.6, + "learning_rate": 7.398384586896118e-05, + "loss": 0.5775, + "step": 20028 + }, + { + "epoch": 0.6, + "learning_rate": 7.397453561810424e-05, + "loss": 0.4999, + "step": 20029 + }, + { + "epoch": 0.6, + "learning_rate": 7.396522560922172e-05, + "loss": 0.4643, + "step": 20030 + }, + { + "epoch": 0.6, + "learning_rate": 7.395591584240013e-05, + "loss": 0.4557, + "step": 20031 + }, + { + "epoch": 0.6, + "learning_rate": 7.394660631772611e-05, + "loss": 0.5632, + "step": 20032 + }, + { + "epoch": 0.6, + "learning_rate": 7.393729703528613e-05, + "loss": 0.5122, + "step": 20033 + }, + { + "epoch": 0.6, + "learning_rate": 7.392798799516676e-05, + "loss": 0.657, + "step": 20034 + }, + { + "epoch": 0.6, + "learning_rate": 7.391867919745459e-05, + "loss": 0.5084, + "step": 20035 + }, + { + "epoch": 0.6, + "learning_rate": 7.390937064223615e-05, + "loss": 0.598, + "step": 20036 + }, + { + "epoch": 0.6, + "learning_rate": 7.390006232959798e-05, + "loss": 0.5072, + "step": 20037 + }, + { + "epoch": 0.6, + "learning_rate": 7.38907542596266e-05, + "loss": 0.5001, + "step": 20038 + }, + { + "epoch": 0.6, + "learning_rate": 7.38814464324086e-05, + "loss": 0.5335, + "step": 20039 + }, + { + "epoch": 0.6, + "learning_rate": 7.387213884803053e-05, + "loss": 0.5492, + "step": 20040 + }, + { + "epoch": 0.6, + "learning_rate": 7.386283150657885e-05, + "loss": 0.5718, + "step": 20041 + }, + { + "epoch": 0.6, + "learning_rate": 7.385352440814018e-05, + "loss": 0.54, + "step": 20042 + }, + { + "epoch": 0.6, + "learning_rate": 7.384421755280098e-05, + "loss": 0.6828, + "step": 20043 + }, + { + "epoch": 0.6, + "learning_rate": 7.383491094064786e-05, + "loss": 0.5395, + "step": 20044 + }, + { + "epoch": 0.6, + "learning_rate": 7.38256045717673e-05, + "loss": 0.4944, + "step": 20045 + }, + { + "epoch": 0.6, + "learning_rate": 7.381629844624582e-05, + "loss": 0.5391, + "step": 20046 + }, + { + "epoch": 0.6, + "learning_rate": 7.380699256417e-05, + "loss": 0.5581, + "step": 20047 + }, + { + "epoch": 0.6, + "learning_rate": 7.379768692562628e-05, + "loss": 0.5576, + "step": 20048 + }, + { + "epoch": 0.6, + "learning_rate": 7.378838153070124e-05, + "loss": 0.5437, + "step": 20049 + }, + { + "epoch": 0.6, + "learning_rate": 7.377907637948137e-05, + "loss": 0.5133, + "step": 20050 + }, + { + "epoch": 0.6, + "learning_rate": 7.376977147205324e-05, + "loss": 0.5807, + "step": 20051 + }, + { + "epoch": 0.6, + "learning_rate": 7.376046680850328e-05, + "loss": 0.5249, + "step": 20052 + }, + { + "epoch": 0.6, + "learning_rate": 7.375116238891803e-05, + "loss": 0.5099, + "step": 20053 + }, + { + "epoch": 0.6, + "learning_rate": 7.374185821338404e-05, + "loss": 0.5545, + "step": 20054 + }, + { + "epoch": 0.6, + "learning_rate": 7.37325542819878e-05, + "loss": 0.5733, + "step": 20055 + }, + { + "epoch": 0.6, + "learning_rate": 7.372325059481578e-05, + "loss": 0.6442, + "step": 20056 + }, + { + "epoch": 0.6, + "learning_rate": 7.371394715195449e-05, + "loss": 0.6095, + "step": 20057 + }, + { + "epoch": 0.6, + "learning_rate": 7.370464395349046e-05, + "loss": 0.5036, + "step": 20058 + }, + { + "epoch": 0.6, + "learning_rate": 7.369534099951018e-05, + "loss": 0.5898, + "step": 20059 + }, + { + "epoch": 0.6, + "learning_rate": 7.368603829010011e-05, + "loss": 0.5285, + "step": 20060 + }, + { + "epoch": 0.6, + "learning_rate": 7.367673582534677e-05, + "loss": 0.4301, + "step": 20061 + }, + { + "epoch": 0.6, + "learning_rate": 7.366743360533664e-05, + "loss": 0.6522, + "step": 20062 + }, + { + "epoch": 0.6, + "learning_rate": 7.365813163015627e-05, + "loss": 0.5502, + "step": 20063 + }, + { + "epoch": 0.6, + "learning_rate": 7.364882989989206e-05, + "loss": 0.6023, + "step": 20064 + }, + { + "epoch": 0.6, + "learning_rate": 7.36395284146305e-05, + "loss": 0.4736, + "step": 20065 + }, + { + "epoch": 0.6, + "learning_rate": 7.363022717445813e-05, + "loss": 0.5241, + "step": 20066 + }, + { + "epoch": 0.6, + "learning_rate": 7.362092617946142e-05, + "loss": 0.6187, + "step": 20067 + }, + { + "epoch": 0.6, + "learning_rate": 7.361162542972679e-05, + "loss": 0.5867, + "step": 20068 + }, + { + "epoch": 0.6, + "learning_rate": 7.360232492534073e-05, + "loss": 0.5292, + "step": 20069 + }, + { + "epoch": 0.6, + "learning_rate": 7.359302466638977e-05, + "loss": 0.388, + "step": 20070 + }, + { + "epoch": 0.6, + "learning_rate": 7.358372465296035e-05, + "loss": 0.5257, + "step": 20071 + }, + { + "epoch": 0.6, + "learning_rate": 7.35744248851389e-05, + "loss": 0.5674, + "step": 20072 + }, + { + "epoch": 0.6, + "learning_rate": 7.356512536301193e-05, + "loss": 0.5239, + "step": 20073 + }, + { + "epoch": 0.6, + "learning_rate": 7.355582608666585e-05, + "loss": 0.5626, + "step": 20074 + }, + { + "epoch": 0.6, + "learning_rate": 7.35465270561872e-05, + "loss": 0.5174, + "step": 20075 + }, + { + "epoch": 0.6, + "learning_rate": 7.353722827166241e-05, + "loss": 0.4982, + "step": 20076 + }, + { + "epoch": 0.6, + "learning_rate": 7.35279297331779e-05, + "loss": 0.4652, + "step": 20077 + }, + { + "epoch": 0.6, + "learning_rate": 7.351863144082019e-05, + "loss": 0.6695, + "step": 20078 + }, + { + "epoch": 0.6, + "learning_rate": 7.350933339467563e-05, + "loss": 0.4938, + "step": 20079 + }, + { + "epoch": 0.6, + "learning_rate": 7.350003559483076e-05, + "loss": 0.531, + "step": 20080 + }, + { + "epoch": 0.6, + "learning_rate": 7.349073804137197e-05, + "loss": 0.498, + "step": 20081 + }, + { + "epoch": 0.6, + "learning_rate": 7.348144073438578e-05, + "loss": 0.5234, + "step": 20082 + }, + { + "epoch": 0.6, + "learning_rate": 7.347214367395856e-05, + "loss": 0.5371, + "step": 20083 + }, + { + "epoch": 0.6, + "learning_rate": 7.346284686017675e-05, + "loss": 0.5607, + "step": 20084 + }, + { + "epoch": 0.6, + "learning_rate": 7.345355029312684e-05, + "loss": 0.5464, + "step": 20085 + }, + { + "epoch": 0.6, + "learning_rate": 7.344425397289526e-05, + "loss": 0.5591, + "step": 20086 + }, + { + "epoch": 0.6, + "learning_rate": 7.34349578995684e-05, + "loss": 0.5248, + "step": 20087 + }, + { + "epoch": 0.6, + "learning_rate": 7.342566207323269e-05, + "loss": 0.5687, + "step": 20088 + }, + { + "epoch": 0.6, + "learning_rate": 7.34163664939746e-05, + "loss": 0.5001, + "step": 20089 + }, + { + "epoch": 0.6, + "learning_rate": 7.340707116188057e-05, + "loss": 0.5181, + "step": 20090 + }, + { + "epoch": 0.6, + "learning_rate": 7.339777607703695e-05, + "loss": 0.486, + "step": 20091 + }, + { + "epoch": 0.6, + "learning_rate": 7.338848123953022e-05, + "loss": 0.519, + "step": 20092 + }, + { + "epoch": 0.6, + "learning_rate": 7.337918664944677e-05, + "loss": 0.541, + "step": 20093 + }, + { + "epoch": 0.6, + "learning_rate": 7.336989230687308e-05, + "loss": 0.5483, + "step": 20094 + }, + { + "epoch": 0.6, + "learning_rate": 7.33605982118955e-05, + "loss": 0.6634, + "step": 20095 + }, + { + "epoch": 0.6, + "learning_rate": 7.335130436460041e-05, + "loss": 0.5005, + "step": 20096 + }, + { + "epoch": 0.6, + "learning_rate": 7.334201076507431e-05, + "loss": 0.5495, + "step": 20097 + }, + { + "epoch": 0.6, + "learning_rate": 7.33327174134036e-05, + "loss": 0.5856, + "step": 20098 + }, + { + "epoch": 0.6, + "learning_rate": 7.332342430967462e-05, + "loss": 0.5666, + "step": 20099 + }, + { + "epoch": 0.6, + "learning_rate": 7.331413145397379e-05, + "loss": 0.5496, + "step": 20100 + }, + { + "epoch": 0.6, + "learning_rate": 7.330483884638756e-05, + "loss": 0.5159, + "step": 20101 + }, + { + "epoch": 0.6, + "learning_rate": 7.329554648700232e-05, + "loss": 0.6577, + "step": 20102 + }, + { + "epoch": 0.6, + "learning_rate": 7.328625437590439e-05, + "loss": 0.5134, + "step": 20103 + }, + { + "epoch": 0.6, + "learning_rate": 7.327696251318025e-05, + "loss": 0.4996, + "step": 20104 + }, + { + "epoch": 0.6, + "learning_rate": 7.326767089891624e-05, + "loss": 0.5176, + "step": 20105 + }, + { + "epoch": 0.6, + "learning_rate": 7.325837953319881e-05, + "loss": 0.5236, + "step": 20106 + }, + { + "epoch": 0.6, + "learning_rate": 7.324908841611429e-05, + "loss": 0.5718, + "step": 20107 + }, + { + "epoch": 0.6, + "learning_rate": 7.323979754774908e-05, + "loss": 0.504, + "step": 20108 + }, + { + "epoch": 0.6, + "learning_rate": 7.323050692818956e-05, + "loss": 0.5386, + "step": 20109 + }, + { + "epoch": 0.6, + "learning_rate": 7.322121655752214e-05, + "loss": 0.5098, + "step": 20110 + }, + { + "epoch": 0.6, + "learning_rate": 7.321192643583317e-05, + "loss": 0.5664, + "step": 20111 + }, + { + "epoch": 0.6, + "learning_rate": 7.3202636563209e-05, + "loss": 0.4326, + "step": 20112 + }, + { + "epoch": 0.6, + "learning_rate": 7.319334693973607e-05, + "loss": 0.5029, + "step": 20113 + }, + { + "epoch": 0.6, + "learning_rate": 7.31840575655007e-05, + "loss": 0.6715, + "step": 20114 + }, + { + "epoch": 0.6, + "learning_rate": 7.317476844058926e-05, + "loss": 0.543, + "step": 20115 + }, + { + "epoch": 0.6, + "learning_rate": 7.316547956508817e-05, + "loss": 0.5542, + "step": 20116 + }, + { + "epoch": 0.6, + "learning_rate": 7.315619093908375e-05, + "loss": 0.6587, + "step": 20117 + }, + { + "epoch": 0.6, + "learning_rate": 7.314690256266237e-05, + "loss": 0.5147, + "step": 20118 + }, + { + "epoch": 0.6, + "learning_rate": 7.313761443591034e-05, + "loss": 0.5397, + "step": 20119 + }, + { + "epoch": 0.6, + "learning_rate": 7.31283265589141e-05, + "loss": 0.5953, + "step": 20120 + }, + { + "epoch": 0.6, + "learning_rate": 7.311903893175999e-05, + "loss": 0.573, + "step": 20121 + }, + { + "epoch": 0.6, + "learning_rate": 7.31097515545343e-05, + "loss": 0.5354, + "step": 20122 + }, + { + "epoch": 0.6, + "learning_rate": 7.310046442732344e-05, + "loss": 0.5499, + "step": 20123 + }, + { + "epoch": 0.6, + "learning_rate": 7.309117755021372e-05, + "loss": 0.6169, + "step": 20124 + }, + { + "epoch": 0.6, + "learning_rate": 7.308189092329155e-05, + "loss": 0.4829, + "step": 20125 + }, + { + "epoch": 0.6, + "learning_rate": 7.307260454664321e-05, + "loss": 0.5778, + "step": 20126 + }, + { + "epoch": 0.6, + "learning_rate": 7.306331842035503e-05, + "loss": 0.5689, + "step": 20127 + }, + { + "epoch": 0.6, + "learning_rate": 7.305403254451341e-05, + "loss": 0.5787, + "step": 20128 + }, + { + "epoch": 0.6, + "learning_rate": 7.304474691920468e-05, + "loss": 0.5104, + "step": 20129 + }, + { + "epoch": 0.6, + "learning_rate": 7.303546154451513e-05, + "loss": 0.5682, + "step": 20130 + }, + { + "epoch": 0.6, + "learning_rate": 7.302617642053109e-05, + "loss": 0.516, + "step": 20131 + }, + { + "epoch": 0.6, + "learning_rate": 7.301689154733893e-05, + "loss": 0.5413, + "step": 20132 + }, + { + "epoch": 0.6, + "learning_rate": 7.300760692502497e-05, + "loss": 0.4041, + "step": 20133 + }, + { + "epoch": 0.6, + "learning_rate": 7.29983225536755e-05, + "loss": 0.571, + "step": 20134 + }, + { + "epoch": 0.6, + "learning_rate": 7.298903843337686e-05, + "loss": 0.5366, + "step": 20135 + }, + { + "epoch": 0.6, + "learning_rate": 7.297975456421536e-05, + "loss": 0.5732, + "step": 20136 + }, + { + "epoch": 0.6, + "learning_rate": 7.297047094627738e-05, + "loss": 0.631, + "step": 20137 + }, + { + "epoch": 0.6, + "learning_rate": 7.296118757964918e-05, + "loss": 0.5195, + "step": 20138 + }, + { + "epoch": 0.6, + "learning_rate": 7.295190446441704e-05, + "loss": 0.5758, + "step": 20139 + }, + { + "epoch": 0.6, + "learning_rate": 7.294262160066735e-05, + "loss": 0.434, + "step": 20140 + }, + { + "epoch": 0.6, + "learning_rate": 7.29333389884864e-05, + "loss": 0.39, + "step": 20141 + }, + { + "epoch": 0.6, + "learning_rate": 7.292405662796044e-05, + "loss": 0.5108, + "step": 20142 + }, + { + "epoch": 0.6, + "learning_rate": 7.29147745191758e-05, + "loss": 0.5544, + "step": 20143 + }, + { + "epoch": 0.6, + "learning_rate": 7.290549266221884e-05, + "loss": 0.5516, + "step": 20144 + }, + { + "epoch": 0.6, + "learning_rate": 7.289621105717578e-05, + "loss": 0.5201, + "step": 20145 + }, + { + "epoch": 0.6, + "learning_rate": 7.288692970413293e-05, + "loss": 0.456, + "step": 20146 + }, + { + "epoch": 0.6, + "learning_rate": 7.28776486031766e-05, + "loss": 0.5516, + "step": 20147 + }, + { + "epoch": 0.6, + "learning_rate": 7.286836775439313e-05, + "loss": 0.523, + "step": 20148 + }, + { + "epoch": 0.6, + "learning_rate": 7.285908715786872e-05, + "loss": 0.6038, + "step": 20149 + }, + { + "epoch": 0.6, + "learning_rate": 7.284980681368972e-05, + "loss": 0.5938, + "step": 20150 + }, + { + "epoch": 0.6, + "learning_rate": 7.284052672194239e-05, + "loss": 0.6018, + "step": 20151 + }, + { + "epoch": 0.6, + "learning_rate": 7.283124688271304e-05, + "loss": 0.5755, + "step": 20152 + }, + { + "epoch": 0.6, + "learning_rate": 7.282196729608789e-05, + "loss": 0.5778, + "step": 20153 + }, + { + "epoch": 0.6, + "learning_rate": 7.281268796215328e-05, + "loss": 0.5404, + "step": 20154 + }, + { + "epoch": 0.6, + "learning_rate": 7.280340888099543e-05, + "loss": 0.5966, + "step": 20155 + }, + { + "epoch": 0.6, + "learning_rate": 7.279413005270069e-05, + "loss": 0.5456, + "step": 20156 + }, + { + "epoch": 0.6, + "learning_rate": 7.278485147735527e-05, + "loss": 0.5363, + "step": 20157 + }, + { + "epoch": 0.6, + "learning_rate": 7.277557315504542e-05, + "loss": 0.5429, + "step": 20158 + }, + { + "epoch": 0.6, + "learning_rate": 7.276629508585747e-05, + "loss": 0.609, + "step": 20159 + }, + { + "epoch": 0.6, + "learning_rate": 7.275701726987767e-05, + "loss": 0.524, + "step": 20160 + }, + { + "epoch": 0.6, + "learning_rate": 7.274773970719225e-05, + "loss": 0.5827, + "step": 20161 + }, + { + "epoch": 0.6, + "learning_rate": 7.273846239788747e-05, + "loss": 0.5214, + "step": 20162 + }, + { + "epoch": 0.6, + "learning_rate": 7.272918534204961e-05, + "loss": 0.4694, + "step": 20163 + }, + { + "epoch": 0.6, + "learning_rate": 7.271990853976495e-05, + "loss": 0.5107, + "step": 20164 + }, + { + "epoch": 0.6, + "learning_rate": 7.271063199111965e-05, + "loss": 0.4674, + "step": 20165 + }, + { + "epoch": 0.6, + "learning_rate": 7.270135569620004e-05, + "loss": 0.5801, + "step": 20166 + }, + { + "epoch": 0.6, + "learning_rate": 7.269207965509234e-05, + "loss": 0.4899, + "step": 20167 + }, + { + "epoch": 0.6, + "learning_rate": 7.268280386788282e-05, + "loss": 0.5298, + "step": 20168 + }, + { + "epoch": 0.6, + "learning_rate": 7.267352833465771e-05, + "loss": 0.5489, + "step": 20169 + }, + { + "epoch": 0.6, + "learning_rate": 7.26642530555032e-05, + "loss": 0.589, + "step": 20170 + }, + { + "epoch": 0.6, + "learning_rate": 7.265497803050558e-05, + "loss": 0.4576, + "step": 20171 + }, + { + "epoch": 0.6, + "learning_rate": 7.264570325975111e-05, + "loss": 0.5753, + "step": 20172 + }, + { + "epoch": 0.6, + "learning_rate": 7.263642874332597e-05, + "loss": 0.4837, + "step": 20173 + }, + { + "epoch": 0.6, + "learning_rate": 7.262715448131638e-05, + "loss": 0.4929, + "step": 20174 + }, + { + "epoch": 0.6, + "learning_rate": 7.261788047380863e-05, + "loss": 0.5079, + "step": 20175 + }, + { + "epoch": 0.6, + "learning_rate": 7.260860672088894e-05, + "loss": 0.677, + "step": 20176 + }, + { + "epoch": 0.6, + "learning_rate": 7.259933322264345e-05, + "loss": 0.5948, + "step": 20177 + }, + { + "epoch": 0.6, + "learning_rate": 7.259005997915846e-05, + "loss": 0.5924, + "step": 20178 + }, + { + "epoch": 0.6, + "learning_rate": 7.25807869905202e-05, + "loss": 0.5722, + "step": 20179 + }, + { + "epoch": 0.6, + "learning_rate": 7.25715142568148e-05, + "loss": 0.5721, + "step": 20180 + }, + { + "epoch": 0.6, + "learning_rate": 7.256224177812857e-05, + "loss": 0.5124, + "step": 20181 + }, + { + "epoch": 0.6, + "learning_rate": 7.255296955454767e-05, + "loss": 0.5786, + "step": 20182 + }, + { + "epoch": 0.6, + "learning_rate": 7.254369758615833e-05, + "loss": 0.4148, + "step": 20183 + }, + { + "epoch": 0.6, + "learning_rate": 7.253442587304671e-05, + "loss": 0.522, + "step": 20184 + }, + { + "epoch": 0.6, + "learning_rate": 7.252515441529908e-05, + "loss": 0.5641, + "step": 20185 + }, + { + "epoch": 0.6, + "learning_rate": 7.25158832130016e-05, + "loss": 0.5389, + "step": 20186 + }, + { + "epoch": 0.6, + "learning_rate": 7.250661226624051e-05, + "loss": 0.5554, + "step": 20187 + }, + { + "epoch": 0.6, + "learning_rate": 7.249734157510196e-05, + "loss": 0.4546, + "step": 20188 + }, + { + "epoch": 0.6, + "learning_rate": 7.248807113967216e-05, + "loss": 0.5132, + "step": 20189 + }, + { + "epoch": 0.6, + "learning_rate": 7.247880096003731e-05, + "loss": 0.4847, + "step": 20190 + }, + { + "epoch": 0.6, + "learning_rate": 7.246953103628363e-05, + "loss": 0.4914, + "step": 20191 + }, + { + "epoch": 0.6, + "learning_rate": 7.246026136849726e-05, + "loss": 0.5222, + "step": 20192 + }, + { + "epoch": 0.6, + "learning_rate": 7.245099195676438e-05, + "loss": 0.5905, + "step": 20193 + }, + { + "epoch": 0.6, + "learning_rate": 7.244172280117122e-05, + "loss": 0.4385, + "step": 20194 + }, + { + "epoch": 0.6, + "learning_rate": 7.243245390180395e-05, + "loss": 0.4878, + "step": 20195 + }, + { + "epoch": 0.6, + "learning_rate": 7.242318525874871e-05, + "loss": 0.4824, + "step": 20196 + }, + { + "epoch": 0.6, + "learning_rate": 7.241391687209172e-05, + "loss": 0.4932, + "step": 20197 + }, + { + "epoch": 0.6, + "learning_rate": 7.24046487419191e-05, + "loss": 0.6252, + "step": 20198 + }, + { + "epoch": 0.6, + "learning_rate": 7.239538086831712e-05, + "loss": 0.5068, + "step": 20199 + }, + { + "epoch": 0.6, + "learning_rate": 7.238611325137186e-05, + "loss": 0.5337, + "step": 20200 + }, + { + "epoch": 0.6, + "learning_rate": 7.237684589116947e-05, + "loss": 0.5311, + "step": 20201 + }, + { + "epoch": 0.6, + "learning_rate": 7.236757878779621e-05, + "loss": 0.5441, + "step": 20202 + }, + { + "epoch": 0.6, + "learning_rate": 7.235831194133819e-05, + "loss": 0.5238, + "step": 20203 + }, + { + "epoch": 0.6, + "learning_rate": 7.234904535188155e-05, + "loss": 0.3892, + "step": 20204 + }, + { + "epoch": 0.6, + "learning_rate": 7.233977901951246e-05, + "loss": 0.378, + "step": 20205 + }, + { + "epoch": 0.6, + "learning_rate": 7.233051294431707e-05, + "loss": 0.506, + "step": 20206 + }, + { + "epoch": 0.6, + "learning_rate": 7.23212471263816e-05, + "loss": 0.6583, + "step": 20207 + }, + { + "epoch": 0.6, + "learning_rate": 7.231198156579207e-05, + "loss": 0.6379, + "step": 20208 + }, + { + "epoch": 0.6, + "learning_rate": 7.230271626263473e-05, + "loss": 0.5469, + "step": 20209 + }, + { + "epoch": 0.6, + "learning_rate": 7.229345121699573e-05, + "loss": 0.4577, + "step": 20210 + }, + { + "epoch": 0.6, + "learning_rate": 7.228418642896114e-05, + "loss": 0.469, + "step": 20211 + }, + { + "epoch": 0.6, + "learning_rate": 7.227492189861713e-05, + "loss": 0.4569, + "step": 20212 + }, + { + "epoch": 0.6, + "learning_rate": 7.226565762604986e-05, + "loss": 0.4332, + "step": 20213 + }, + { + "epoch": 0.6, + "learning_rate": 7.225639361134548e-05, + "loss": 0.5421, + "step": 20214 + }, + { + "epoch": 0.6, + "learning_rate": 7.224712985459005e-05, + "loss": 0.5719, + "step": 20215 + }, + { + "epoch": 0.6, + "learning_rate": 7.223786635586977e-05, + "loss": 0.5608, + "step": 20216 + }, + { + "epoch": 0.6, + "learning_rate": 7.222860311527071e-05, + "loss": 0.5566, + "step": 20217 + }, + { + "epoch": 0.6, + "learning_rate": 7.221934013287908e-05, + "loss": 0.4355, + "step": 20218 + }, + { + "epoch": 0.6, + "learning_rate": 7.221007740878092e-05, + "loss": 0.6291, + "step": 20219 + }, + { + "epoch": 0.6, + "learning_rate": 7.22008149430624e-05, + "loss": 0.5574, + "step": 20220 + }, + { + "epoch": 0.6, + "learning_rate": 7.21915527358096e-05, + "loss": 0.462, + "step": 20221 + }, + { + "epoch": 0.6, + "learning_rate": 7.21822907871087e-05, + "loss": 0.6335, + "step": 20222 + }, + { + "epoch": 0.6, + "learning_rate": 7.217302909704575e-05, + "loss": 0.4911, + "step": 20223 + }, + { + "epoch": 0.6, + "learning_rate": 7.216376766570687e-05, + "loss": 0.5425, + "step": 20224 + }, + { + "epoch": 0.6, + "learning_rate": 7.215450649317818e-05, + "loss": 0.5753, + "step": 20225 + }, + { + "epoch": 0.6, + "learning_rate": 7.214524557954586e-05, + "loss": 0.5737, + "step": 20226 + }, + { + "epoch": 0.6, + "learning_rate": 7.213598492489586e-05, + "loss": 0.4532, + "step": 20227 + }, + { + "epoch": 0.6, + "learning_rate": 7.21267245293144e-05, + "loss": 0.579, + "step": 20228 + }, + { + "epoch": 0.6, + "learning_rate": 7.211746439288752e-05, + "loss": 0.4585, + "step": 20229 + }, + { + "epoch": 0.6, + "learning_rate": 7.210820451570139e-05, + "loss": 0.5748, + "step": 20230 + }, + { + "epoch": 0.6, + "learning_rate": 7.209894489784204e-05, + "loss": 0.4916, + "step": 20231 + }, + { + "epoch": 0.6, + "learning_rate": 7.208968553939556e-05, + "loss": 0.5763, + "step": 20232 + }, + { + "epoch": 0.6, + "learning_rate": 7.208042644044807e-05, + "loss": 0.6427, + "step": 20233 + }, + { + "epoch": 0.6, + "learning_rate": 7.207116760108569e-05, + "loss": 0.5423, + "step": 20234 + }, + { + "epoch": 0.6, + "learning_rate": 7.206190902139443e-05, + "loss": 0.4998, + "step": 20235 + }, + { + "epoch": 0.6, + "learning_rate": 7.205265070146037e-05, + "loss": 0.4118, + "step": 20236 + }, + { + "epoch": 0.6, + "learning_rate": 7.204339264136965e-05, + "loss": 0.4647, + "step": 20237 + }, + { + "epoch": 0.6, + "learning_rate": 7.203413484120836e-05, + "loss": 0.481, + "step": 20238 + }, + { + "epoch": 0.6, + "learning_rate": 7.202487730106249e-05, + "loss": 0.5217, + "step": 20239 + }, + { + "epoch": 0.6, + "learning_rate": 7.201562002101818e-05, + "loss": 0.584, + "step": 20240 + }, + { + "epoch": 0.6, + "learning_rate": 7.200636300116151e-05, + "loss": 0.5129, + "step": 20241 + }, + { + "epoch": 0.6, + "learning_rate": 7.199710624157846e-05, + "loss": 0.6563, + "step": 20242 + }, + { + "epoch": 0.6, + "learning_rate": 7.19878497423552e-05, + "loss": 0.4891, + "step": 20243 + }, + { + "epoch": 0.6, + "learning_rate": 7.197859350357771e-05, + "loss": 0.6799, + "step": 20244 + }, + { + "epoch": 0.6, + "learning_rate": 7.196933752533215e-05, + "loss": 0.4024, + "step": 20245 + }, + { + "epoch": 0.6, + "learning_rate": 7.196008180770446e-05, + "loss": 0.6023, + "step": 20246 + }, + { + "epoch": 0.6, + "learning_rate": 7.195082635078077e-05, + "loss": 0.489, + "step": 20247 + }, + { + "epoch": 0.6, + "learning_rate": 7.19415711546471e-05, + "loss": 0.5191, + "step": 20248 + }, + { + "epoch": 0.6, + "learning_rate": 7.193231621938956e-05, + "loss": 0.5532, + "step": 20249 + }, + { + "epoch": 0.6, + "learning_rate": 7.192306154509414e-05, + "loss": 0.6612, + "step": 20250 + }, + { + "epoch": 0.6, + "learning_rate": 7.191380713184689e-05, + "loss": 0.5197, + "step": 20251 + }, + { + "epoch": 0.6, + "learning_rate": 7.190455297973388e-05, + "loss": 0.489, + "step": 20252 + }, + { + "epoch": 0.6, + "learning_rate": 7.189529908884117e-05, + "loss": 0.5002, + "step": 20253 + }, + { + "epoch": 0.6, + "learning_rate": 7.188604545925472e-05, + "loss": 0.5381, + "step": 20254 + }, + { + "epoch": 0.6, + "learning_rate": 7.187679209106062e-05, + "loss": 0.592, + "step": 20255 + }, + { + "epoch": 0.6, + "learning_rate": 7.18675389843449e-05, + "loss": 0.5974, + "step": 20256 + }, + { + "epoch": 0.6, + "learning_rate": 7.185828613919363e-05, + "loss": 0.5432, + "step": 20257 + }, + { + "epoch": 0.6, + "learning_rate": 7.184903355569274e-05, + "loss": 0.4272, + "step": 20258 + }, + { + "epoch": 0.6, + "learning_rate": 7.183978123392834e-05, + "loss": 0.6748, + "step": 20259 + }, + { + "epoch": 0.6, + "learning_rate": 7.18305291739864e-05, + "loss": 0.4755, + "step": 20260 + }, + { + "epoch": 0.6, + "learning_rate": 7.182127737595303e-05, + "loss": 0.4907, + "step": 20261 + }, + { + "epoch": 0.6, + "learning_rate": 7.181202583991418e-05, + "loss": 0.5356, + "step": 20262 + }, + { + "epoch": 0.6, + "learning_rate": 7.180277456595585e-05, + "loss": 0.6947, + "step": 20263 + }, + { + "epoch": 0.6, + "learning_rate": 7.179352355416409e-05, + "loss": 0.4037, + "step": 20264 + }, + { + "epoch": 0.6, + "learning_rate": 7.178427280462493e-05, + "loss": 0.5506, + "step": 20265 + }, + { + "epoch": 0.6, + "learning_rate": 7.177502231742435e-05, + "loss": 0.5367, + "step": 20266 + }, + { + "epoch": 0.6, + "learning_rate": 7.176577209264835e-05, + "loss": 0.5224, + "step": 20267 + }, + { + "epoch": 0.6, + "learning_rate": 7.175652213038295e-05, + "loss": 0.3853, + "step": 20268 + }, + { + "epoch": 0.6, + "learning_rate": 7.174727243071418e-05, + "loss": 0.5436, + "step": 20269 + }, + { + "epoch": 0.6, + "learning_rate": 7.173802299372796e-05, + "loss": 0.5999, + "step": 20270 + }, + { + "epoch": 0.6, + "learning_rate": 7.172877381951036e-05, + "loss": 0.489, + "step": 20271 + }, + { + "epoch": 0.6, + "learning_rate": 7.171952490814733e-05, + "loss": 0.5391, + "step": 20272 + }, + { + "epoch": 0.6, + "learning_rate": 7.171027625972492e-05, + "loss": 0.5494, + "step": 20273 + }, + { + "epoch": 0.6, + "learning_rate": 7.17010278743291e-05, + "loss": 0.5532, + "step": 20274 + }, + { + "epoch": 0.6, + "learning_rate": 7.16917797520458e-05, + "loss": 0.5404, + "step": 20275 + }, + { + "epoch": 0.6, + "learning_rate": 7.168253189296111e-05, + "loss": 0.5906, + "step": 20276 + }, + { + "epoch": 0.6, + "learning_rate": 7.167328429716089e-05, + "loss": 0.6117, + "step": 20277 + }, + { + "epoch": 0.6, + "learning_rate": 7.166403696473121e-05, + "loss": 0.5061, + "step": 20278 + }, + { + "epoch": 0.6, + "learning_rate": 7.1654789895758e-05, + "loss": 0.4405, + "step": 20279 + }, + { + "epoch": 0.6, + "learning_rate": 7.164554309032729e-05, + "loss": 0.6276, + "step": 20280 + }, + { + "epoch": 0.6, + "learning_rate": 7.163629654852501e-05, + "loss": 0.5386, + "step": 20281 + }, + { + "epoch": 0.6, + "learning_rate": 7.16270502704371e-05, + "loss": 0.628, + "step": 20282 + }, + { + "epoch": 0.6, + "learning_rate": 7.161780425614961e-05, + "loss": 0.5791, + "step": 20283 + }, + { + "epoch": 0.6, + "learning_rate": 7.160855850574848e-05, + "loss": 0.371, + "step": 20284 + }, + { + "epoch": 0.6, + "learning_rate": 7.159931301931964e-05, + "loss": 0.6107, + "step": 20285 + }, + { + "epoch": 0.6, + "learning_rate": 7.159006779694903e-05, + "loss": 0.5237, + "step": 20286 + }, + { + "epoch": 0.6, + "learning_rate": 7.158082283872267e-05, + "loss": 0.5529, + "step": 20287 + }, + { + "epoch": 0.6, + "learning_rate": 7.157157814472652e-05, + "loss": 0.587, + "step": 20288 + }, + { + "epoch": 0.6, + "learning_rate": 7.156233371504647e-05, + "loss": 0.472, + "step": 20289 + }, + { + "epoch": 0.6, + "learning_rate": 7.155308954976853e-05, + "loss": 0.4875, + "step": 20290 + }, + { + "epoch": 0.6, + "learning_rate": 7.154384564897859e-05, + "loss": 0.5101, + "step": 20291 + }, + { + "epoch": 0.6, + "learning_rate": 7.153460201276269e-05, + "loss": 0.6317, + "step": 20292 + }, + { + "epoch": 0.6, + "learning_rate": 7.152535864120669e-05, + "loss": 0.5171, + "step": 20293 + }, + { + "epoch": 0.6, + "learning_rate": 7.151611553439653e-05, + "loss": 0.6077, + "step": 20294 + }, + { + "epoch": 0.6, + "learning_rate": 7.150687269241818e-05, + "loss": 0.4791, + "step": 20295 + }, + { + "epoch": 0.6, + "learning_rate": 7.149763011535762e-05, + "loss": 0.5882, + "step": 20296 + }, + { + "epoch": 0.6, + "learning_rate": 7.148838780330069e-05, + "loss": 0.5216, + "step": 20297 + }, + { + "epoch": 0.6, + "learning_rate": 7.147914575633337e-05, + "loss": 0.5476, + "step": 20298 + }, + { + "epoch": 0.6, + "learning_rate": 7.14699039745416e-05, + "loss": 0.5, + "step": 20299 + }, + { + "epoch": 0.6, + "learning_rate": 7.146066245801131e-05, + "loss": 0.4474, + "step": 20300 + }, + { + "epoch": 0.6, + "learning_rate": 7.145142120682834e-05, + "loss": 0.6243, + "step": 20301 + }, + { + "epoch": 0.6, + "learning_rate": 7.144218022107874e-05, + "loss": 0.5341, + "step": 20302 + }, + { + "epoch": 0.6, + "learning_rate": 7.143293950084832e-05, + "loss": 0.5222, + "step": 20303 + }, + { + "epoch": 0.6, + "learning_rate": 7.14236990462231e-05, + "loss": 0.5958, + "step": 20304 + }, + { + "epoch": 0.6, + "learning_rate": 7.141445885728891e-05, + "loss": 0.6316, + "step": 20305 + }, + { + "epoch": 0.6, + "learning_rate": 7.140521893413166e-05, + "loss": 0.4802, + "step": 20306 + }, + { + "epoch": 0.6, + "learning_rate": 7.139597927683734e-05, + "loss": 0.5386, + "step": 20307 + }, + { + "epoch": 0.6, + "learning_rate": 7.138673988549177e-05, + "loss": 0.596, + "step": 20308 + }, + { + "epoch": 0.6, + "learning_rate": 7.13775007601809e-05, + "loss": 0.5947, + "step": 20309 + }, + { + "epoch": 0.6, + "learning_rate": 7.136826190099058e-05, + "loss": 0.5323, + "step": 20310 + }, + { + "epoch": 0.6, + "learning_rate": 7.135902330800682e-05, + "loss": 0.5136, + "step": 20311 + }, + { + "epoch": 0.6, + "learning_rate": 7.13497849813154e-05, + "loss": 0.4338, + "step": 20312 + }, + { + "epoch": 0.6, + "learning_rate": 7.134054692100225e-05, + "loss": 0.5767, + "step": 20313 + }, + { + "epoch": 0.6, + "learning_rate": 7.13313091271533e-05, + "loss": 0.5436, + "step": 20314 + }, + { + "epoch": 0.6, + "learning_rate": 7.132207159985443e-05, + "loss": 0.4932, + "step": 20315 + }, + { + "epoch": 0.6, + "learning_rate": 7.131283433919146e-05, + "loss": 0.6157, + "step": 20316 + }, + { + "epoch": 0.6, + "learning_rate": 7.130359734525034e-05, + "loss": 0.6693, + "step": 20317 + }, + { + "epoch": 0.6, + "learning_rate": 7.129436061811694e-05, + "loss": 0.5678, + "step": 20318 + }, + { + "epoch": 0.6, + "learning_rate": 7.128512415787715e-05, + "loss": 0.4804, + "step": 20319 + }, + { + "epoch": 0.6, + "learning_rate": 7.12758879646168e-05, + "loss": 0.6832, + "step": 20320 + }, + { + "epoch": 0.6, + "learning_rate": 7.126665203842182e-05, + "loss": 0.4831, + "step": 20321 + }, + { + "epoch": 0.61, + "learning_rate": 7.125741637937802e-05, + "loss": 0.5528, + "step": 20322 + }, + { + "epoch": 0.61, + "learning_rate": 7.124818098757136e-05, + "loss": 0.5666, + "step": 20323 + }, + { + "epoch": 0.61, + "learning_rate": 7.123894586308764e-05, + "loss": 0.4626, + "step": 20324 + }, + { + "epoch": 0.61, + "learning_rate": 7.12297110060127e-05, + "loss": 0.5629, + "step": 20325 + }, + { + "epoch": 0.61, + "learning_rate": 7.122047641643247e-05, + "loss": 0.5913, + "step": 20326 + }, + { + "epoch": 0.61, + "learning_rate": 7.121124209443279e-05, + "loss": 0.5745, + "step": 20327 + }, + { + "epoch": 0.61, + "learning_rate": 7.12020080400995e-05, + "loss": 0.5629, + "step": 20328 + }, + { + "epoch": 0.61, + "learning_rate": 7.119277425351844e-05, + "loss": 0.5237, + "step": 20329 + }, + { + "epoch": 0.61, + "learning_rate": 7.118354073477549e-05, + "loss": 0.5488, + "step": 20330 + }, + { + "epoch": 0.61, + "learning_rate": 7.117430748395653e-05, + "loss": 0.6322, + "step": 20331 + }, + { + "epoch": 0.61, + "learning_rate": 7.116507450114732e-05, + "loss": 0.429, + "step": 20332 + }, + { + "epoch": 0.61, + "learning_rate": 7.115584178643378e-05, + "loss": 0.4398, + "step": 20333 + }, + { + "epoch": 0.61, + "learning_rate": 7.114660933990171e-05, + "loss": 0.651, + "step": 20334 + }, + { + "epoch": 0.61, + "learning_rate": 7.1137377161637e-05, + "loss": 0.5368, + "step": 20335 + }, + { + "epoch": 0.61, + "learning_rate": 7.112814525172543e-05, + "loss": 0.4769, + "step": 20336 + }, + { + "epoch": 0.61, + "learning_rate": 7.111891361025285e-05, + "loss": 0.5126, + "step": 20337 + }, + { + "epoch": 0.61, + "learning_rate": 7.110968223730512e-05, + "loss": 0.6102, + "step": 20338 + }, + { + "epoch": 0.61, + "learning_rate": 7.110045113296807e-05, + "loss": 0.5925, + "step": 20339 + }, + { + "epoch": 0.61, + "learning_rate": 7.109122029732746e-05, + "loss": 0.5171, + "step": 20340 + }, + { + "epoch": 0.61, + "learning_rate": 7.108198973046916e-05, + "loss": 0.5195, + "step": 20341 + }, + { + "epoch": 0.61, + "learning_rate": 7.107275943247905e-05, + "loss": 0.5711, + "step": 20342 + }, + { + "epoch": 0.61, + "learning_rate": 7.106352940344286e-05, + "loss": 0.551, + "step": 20343 + }, + { + "epoch": 0.61, + "learning_rate": 7.105429964344641e-05, + "loss": 0.4544, + "step": 20344 + }, + { + "epoch": 0.61, + "learning_rate": 7.104507015257557e-05, + "loss": 0.5655, + "step": 20345 + }, + { + "epoch": 0.61, + "learning_rate": 7.103584093091617e-05, + "loss": 0.5197, + "step": 20346 + }, + { + "epoch": 0.61, + "learning_rate": 7.102661197855391e-05, + "loss": 0.5099, + "step": 20347 + }, + { + "epoch": 0.61, + "learning_rate": 7.10173832955747e-05, + "loss": 0.6182, + "step": 20348 + }, + { + "epoch": 0.61, + "learning_rate": 7.100815488206431e-05, + "loss": 0.7184, + "step": 20349 + }, + { + "epoch": 0.61, + "learning_rate": 7.099892673810855e-05, + "loss": 0.486, + "step": 20350 + }, + { + "epoch": 0.61, + "learning_rate": 7.098969886379318e-05, + "loss": 0.5505, + "step": 20351 + }, + { + "epoch": 0.61, + "learning_rate": 7.098047125920404e-05, + "loss": 0.5842, + "step": 20352 + }, + { + "epoch": 0.61, + "learning_rate": 7.097124392442689e-05, + "loss": 0.4312, + "step": 20353 + }, + { + "epoch": 0.61, + "learning_rate": 7.09620168595476e-05, + "loss": 0.5986, + "step": 20354 + }, + { + "epoch": 0.61, + "learning_rate": 7.095279006465187e-05, + "loss": 0.5095, + "step": 20355 + }, + { + "epoch": 0.61, + "learning_rate": 7.09435635398255e-05, + "loss": 0.5852, + "step": 20356 + }, + { + "epoch": 0.61, + "learning_rate": 7.093433728515432e-05, + "loss": 0.6288, + "step": 20357 + }, + { + "epoch": 0.61, + "learning_rate": 7.09251113007241e-05, + "loss": 0.6602, + "step": 20358 + }, + { + "epoch": 0.61, + "learning_rate": 7.09158855866206e-05, + "loss": 0.6114, + "step": 20359 + }, + { + "epoch": 0.61, + "learning_rate": 7.090666014292957e-05, + "loss": 0.5162, + "step": 20360 + }, + { + "epoch": 0.61, + "learning_rate": 7.089743496973685e-05, + "loss": 0.4929, + "step": 20361 + }, + { + "epoch": 0.61, + "learning_rate": 7.08882100671282e-05, + "loss": 0.5884, + "step": 20362 + }, + { + "epoch": 0.61, + "learning_rate": 7.087898543518932e-05, + "loss": 0.5205, + "step": 20363 + }, + { + "epoch": 0.61, + "learning_rate": 7.086976107400607e-05, + "loss": 0.4987, + "step": 20364 + }, + { + "epoch": 0.61, + "learning_rate": 7.086053698366413e-05, + "loss": 0.5702, + "step": 20365 + }, + { + "epoch": 0.61, + "learning_rate": 7.085131316424935e-05, + "loss": 0.5374, + "step": 20366 + }, + { + "epoch": 0.61, + "learning_rate": 7.084208961584741e-05, + "loss": 0.6115, + "step": 20367 + }, + { + "epoch": 0.61, + "learning_rate": 7.083286633854409e-05, + "loss": 0.421, + "step": 20368 + }, + { + "epoch": 0.61, + "learning_rate": 7.082364333242518e-05, + "loss": 0.5614, + "step": 20369 + }, + { + "epoch": 0.61, + "learning_rate": 7.081442059757642e-05, + "loss": 0.5837, + "step": 20370 + }, + { + "epoch": 0.61, + "learning_rate": 7.080519813408353e-05, + "loss": 0.5043, + "step": 20371 + }, + { + "epoch": 0.61, + "learning_rate": 7.079597594203224e-05, + "loss": 0.4851, + "step": 20372 + }, + { + "epoch": 0.61, + "learning_rate": 7.078675402150838e-05, + "loss": 0.5056, + "step": 20373 + }, + { + "epoch": 0.61, + "learning_rate": 7.077753237259762e-05, + "loss": 0.5525, + "step": 20374 + }, + { + "epoch": 0.61, + "learning_rate": 7.076831099538568e-05, + "loss": 0.5338, + "step": 20375 + }, + { + "epoch": 0.61, + "learning_rate": 7.075908988995837e-05, + "loss": 0.5054, + "step": 20376 + }, + { + "epoch": 0.61, + "learning_rate": 7.07498690564014e-05, + "loss": 0.4272, + "step": 20377 + }, + { + "epoch": 0.61, + "learning_rate": 7.074064849480044e-05, + "loss": 0.4656, + "step": 20378 + }, + { + "epoch": 0.61, + "learning_rate": 7.07314282052413e-05, + "loss": 0.6044, + "step": 20379 + }, + { + "epoch": 0.61, + "learning_rate": 7.072220818780967e-05, + "loss": 0.5896, + "step": 20380 + }, + { + "epoch": 0.61, + "learning_rate": 7.071298844259131e-05, + "loss": 0.5343, + "step": 20381 + }, + { + "epoch": 0.61, + "learning_rate": 7.070376896967186e-05, + "loss": 0.4982, + "step": 20382 + }, + { + "epoch": 0.61, + "learning_rate": 7.069454976913711e-05, + "loss": 0.5876, + "step": 20383 + }, + { + "epoch": 0.61, + "learning_rate": 7.068533084107274e-05, + "loss": 0.5907, + "step": 20384 + }, + { + "epoch": 0.61, + "learning_rate": 7.067611218556451e-05, + "loss": 0.5835, + "step": 20385 + }, + { + "epoch": 0.61, + "learning_rate": 7.06668938026981e-05, + "loss": 0.4981, + "step": 20386 + }, + { + "epoch": 0.61, + "learning_rate": 7.06576756925592e-05, + "loss": 0.513, + "step": 20387 + }, + { + "epoch": 0.61, + "learning_rate": 7.064845785523354e-05, + "loss": 0.481, + "step": 20388 + }, + { + "epoch": 0.61, + "learning_rate": 7.063924029080685e-05, + "loss": 0.55, + "step": 20389 + }, + { + "epoch": 0.61, + "learning_rate": 7.063002299936479e-05, + "loss": 0.4717, + "step": 20390 + }, + { + "epoch": 0.61, + "learning_rate": 7.062080598099303e-05, + "loss": 0.5319, + "step": 20391 + }, + { + "epoch": 0.61, + "learning_rate": 7.061158923577734e-05, + "loss": 0.5046, + "step": 20392 + }, + { + "epoch": 0.61, + "learning_rate": 7.06023727638034e-05, + "loss": 0.5651, + "step": 20393 + }, + { + "epoch": 0.61, + "learning_rate": 7.059315656515684e-05, + "loss": 0.4412, + "step": 20394 + }, + { + "epoch": 0.61, + "learning_rate": 7.058394063992341e-05, + "loss": 0.576, + "step": 20395 + }, + { + "epoch": 0.61, + "learning_rate": 7.057472498818875e-05, + "loss": 0.6119, + "step": 20396 + }, + { + "epoch": 0.61, + "learning_rate": 7.056550961003862e-05, + "loss": 0.5459, + "step": 20397 + }, + { + "epoch": 0.61, + "learning_rate": 7.055629450555863e-05, + "loss": 0.5571, + "step": 20398 + }, + { + "epoch": 0.61, + "learning_rate": 7.054707967483445e-05, + "loss": 0.4827, + "step": 20399 + }, + { + "epoch": 0.61, + "learning_rate": 7.05378651179518e-05, + "loss": 0.5719, + "step": 20400 + }, + { + "epoch": 0.61, + "learning_rate": 7.052865083499637e-05, + "loss": 0.5068, + "step": 20401 + }, + { + "epoch": 0.61, + "learning_rate": 7.051943682605377e-05, + "loss": 0.5481, + "step": 20402 + }, + { + "epoch": 0.61, + "learning_rate": 7.051022309120968e-05, + "loss": 0.5539, + "step": 20403 + }, + { + "epoch": 0.61, + "learning_rate": 7.05010096305498e-05, + "loss": 0.5992, + "step": 20404 + }, + { + "epoch": 0.61, + "learning_rate": 7.04917964441598e-05, + "loss": 0.5935, + "step": 20405 + }, + { + "epoch": 0.61, + "learning_rate": 7.048258353212527e-05, + "loss": 0.4382, + "step": 20406 + }, + { + "epoch": 0.61, + "learning_rate": 7.047337089453194e-05, + "loss": 0.5671, + "step": 20407 + }, + { + "epoch": 0.61, + "learning_rate": 7.046415853146546e-05, + "loss": 0.4736, + "step": 20408 + }, + { + "epoch": 0.61, + "learning_rate": 7.045494644301142e-05, + "loss": 0.6073, + "step": 20409 + }, + { + "epoch": 0.61, + "learning_rate": 7.044573462925552e-05, + "loss": 0.4443, + "step": 20410 + }, + { + "epoch": 0.61, + "learning_rate": 7.043652309028341e-05, + "loss": 0.588, + "step": 20411 + }, + { + "epoch": 0.61, + "learning_rate": 7.042731182618077e-05, + "loss": 0.4381, + "step": 20412 + }, + { + "epoch": 0.61, + "learning_rate": 7.041810083703313e-05, + "loss": 0.5157, + "step": 20413 + }, + { + "epoch": 0.61, + "learning_rate": 7.040889012292622e-05, + "loss": 0.5243, + "step": 20414 + }, + { + "epoch": 0.61, + "learning_rate": 7.039967968394563e-05, + "loss": 0.4476, + "step": 20415 + }, + { + "epoch": 0.61, + "learning_rate": 7.039046952017707e-05, + "loss": 0.5029, + "step": 20416 + }, + { + "epoch": 0.61, + "learning_rate": 7.038125963170611e-05, + "loss": 0.5241, + "step": 20417 + }, + { + "epoch": 0.61, + "learning_rate": 7.037205001861837e-05, + "loss": 0.7158, + "step": 20418 + }, + { + "epoch": 0.61, + "learning_rate": 7.03628406809995e-05, + "loss": 0.4049, + "step": 20419 + }, + { + "epoch": 0.61, + "learning_rate": 7.035363161893518e-05, + "loss": 0.5385, + "step": 20420 + }, + { + "epoch": 0.61, + "learning_rate": 7.034442283251092e-05, + "loss": 0.4806, + "step": 20421 + }, + { + "epoch": 0.61, + "learning_rate": 7.033521432181241e-05, + "loss": 0.5542, + "step": 20422 + }, + { + "epoch": 0.61, + "learning_rate": 7.032600608692525e-05, + "loss": 0.5419, + "step": 20423 + }, + { + "epoch": 0.61, + "learning_rate": 7.03167981279351e-05, + "loss": 0.5645, + "step": 20424 + }, + { + "epoch": 0.61, + "learning_rate": 7.030759044492748e-05, + "loss": 0.5615, + "step": 20425 + }, + { + "epoch": 0.61, + "learning_rate": 7.029838303798806e-05, + "loss": 0.5652, + "step": 20426 + }, + { + "epoch": 0.61, + "learning_rate": 7.028917590720244e-05, + "loss": 0.5491, + "step": 20427 + }, + { + "epoch": 0.61, + "learning_rate": 7.027996905265623e-05, + "loss": 0.5282, + "step": 20428 + }, + { + "epoch": 0.61, + "learning_rate": 7.027076247443502e-05, + "loss": 0.5527, + "step": 20429 + }, + { + "epoch": 0.61, + "learning_rate": 7.02615561726244e-05, + "loss": 0.6331, + "step": 20430 + }, + { + "epoch": 0.61, + "learning_rate": 7.025235014730998e-05, + "loss": 0.5212, + "step": 20431 + }, + { + "epoch": 0.61, + "learning_rate": 7.024314439857739e-05, + "loss": 0.4926, + "step": 20432 + }, + { + "epoch": 0.61, + "learning_rate": 7.023393892651216e-05, + "loss": 0.4274, + "step": 20433 + }, + { + "epoch": 0.61, + "learning_rate": 7.022473373119986e-05, + "loss": 0.5639, + "step": 20434 + }, + { + "epoch": 0.61, + "learning_rate": 7.021552881272616e-05, + "loss": 0.5021, + "step": 20435 + }, + { + "epoch": 0.61, + "learning_rate": 7.020632417117662e-05, + "loss": 0.4947, + "step": 20436 + }, + { + "epoch": 0.61, + "learning_rate": 7.019711980663676e-05, + "loss": 0.3794, + "step": 20437 + }, + { + "epoch": 0.61, + "learning_rate": 7.018791571919222e-05, + "loss": 0.5829, + "step": 20438 + }, + { + "epoch": 0.61, + "learning_rate": 7.017871190892858e-05, + "loss": 0.6456, + "step": 20439 + }, + { + "epoch": 0.61, + "learning_rate": 7.016950837593134e-05, + "loss": 0.491, + "step": 20440 + }, + { + "epoch": 0.61, + "learning_rate": 7.016030512028617e-05, + "loss": 0.7173, + "step": 20441 + }, + { + "epoch": 0.61, + "learning_rate": 7.015110214207854e-05, + "loss": 0.4947, + "step": 20442 + }, + { + "epoch": 0.61, + "learning_rate": 7.014189944139416e-05, + "loss": 0.5168, + "step": 20443 + }, + { + "epoch": 0.61, + "learning_rate": 7.01326970183184e-05, + "loss": 0.5297, + "step": 20444 + }, + { + "epoch": 0.61, + "learning_rate": 7.012349487293697e-05, + "loss": 0.603, + "step": 20445 + }, + { + "epoch": 0.61, + "learning_rate": 7.011429300533534e-05, + "loss": 0.7032, + "step": 20446 + }, + { + "epoch": 0.61, + "learning_rate": 7.010509141559914e-05, + "loss": 0.5821, + "step": 20447 + }, + { + "epoch": 0.61, + "learning_rate": 7.009589010381388e-05, + "loss": 0.3768, + "step": 20448 + }, + { + "epoch": 0.61, + "learning_rate": 7.008668907006508e-05, + "loss": 0.5485, + "step": 20449 + }, + { + "epoch": 0.61, + "learning_rate": 7.007748831443835e-05, + "loss": 0.4667, + "step": 20450 + }, + { + "epoch": 0.61, + "learning_rate": 7.006828783701921e-05, + "loss": 0.4605, + "step": 20451 + }, + { + "epoch": 0.61, + "learning_rate": 7.005908763789321e-05, + "loss": 0.6399, + "step": 20452 + }, + { + "epoch": 0.61, + "learning_rate": 7.004988771714585e-05, + "loss": 0.4418, + "step": 20453 + }, + { + "epoch": 0.61, + "learning_rate": 7.00406880748627e-05, + "loss": 0.5456, + "step": 20454 + }, + { + "epoch": 0.61, + "learning_rate": 7.003148871112934e-05, + "loss": 0.6603, + "step": 20455 + }, + { + "epoch": 0.61, + "learning_rate": 7.00222896260312e-05, + "loss": 0.5242, + "step": 20456 + }, + { + "epoch": 0.61, + "learning_rate": 7.001309081965388e-05, + "loss": 0.5403, + "step": 20457 + }, + { + "epoch": 0.61, + "learning_rate": 7.000389229208286e-05, + "loss": 0.564, + "step": 20458 + }, + { + "epoch": 0.61, + "learning_rate": 6.999469404340376e-05, + "loss": 0.524, + "step": 20459 + }, + { + "epoch": 0.61, + "learning_rate": 6.998549607370199e-05, + "loss": 0.4686, + "step": 20460 + }, + { + "epoch": 0.61, + "learning_rate": 6.997629838306313e-05, + "loss": 0.6138, + "step": 20461 + }, + { + "epoch": 0.61, + "learning_rate": 6.996710097157266e-05, + "loss": 0.4984, + "step": 20462 + }, + { + "epoch": 0.61, + "learning_rate": 6.995790383931616e-05, + "loss": 0.5329, + "step": 20463 + }, + { + "epoch": 0.61, + "learning_rate": 6.994870698637907e-05, + "loss": 0.3947, + "step": 20464 + }, + { + "epoch": 0.61, + "learning_rate": 6.99395104128469e-05, + "loss": 0.5458, + "step": 20465 + }, + { + "epoch": 0.61, + "learning_rate": 6.99303141188052e-05, + "loss": 0.4879, + "step": 20466 + }, + { + "epoch": 0.61, + "learning_rate": 6.99211181043395e-05, + "loss": 0.6046, + "step": 20467 + }, + { + "epoch": 0.61, + "learning_rate": 6.991192236953519e-05, + "loss": 0.433, + "step": 20468 + }, + { + "epoch": 0.61, + "learning_rate": 6.990272691447785e-05, + "loss": 0.6051, + "step": 20469 + }, + { + "epoch": 0.61, + "learning_rate": 6.989353173925296e-05, + "loss": 0.5588, + "step": 20470 + }, + { + "epoch": 0.61, + "learning_rate": 6.988433684394604e-05, + "loss": 0.4626, + "step": 20471 + }, + { + "epoch": 0.61, + "learning_rate": 6.987514222864254e-05, + "loss": 0.6229, + "step": 20472 + }, + { + "epoch": 0.61, + "learning_rate": 6.986594789342793e-05, + "loss": 0.6283, + "step": 20473 + }, + { + "epoch": 0.61, + "learning_rate": 6.985675383838781e-05, + "loss": 0.5365, + "step": 20474 + }, + { + "epoch": 0.61, + "learning_rate": 6.984756006360748e-05, + "loss": 0.4335, + "step": 20475 + }, + { + "epoch": 0.61, + "learning_rate": 6.983836656917258e-05, + "loss": 0.5288, + "step": 20476 + }, + { + "epoch": 0.61, + "learning_rate": 6.982917335516848e-05, + "loss": 0.4706, + "step": 20477 + }, + { + "epoch": 0.61, + "learning_rate": 6.981998042168075e-05, + "loss": 0.5151, + "step": 20478 + }, + { + "epoch": 0.61, + "learning_rate": 6.981078776879479e-05, + "loss": 0.6121, + "step": 20479 + }, + { + "epoch": 0.61, + "learning_rate": 6.980159539659607e-05, + "loss": 0.5608, + "step": 20480 + }, + { + "epoch": 0.61, + "learning_rate": 6.979240330517012e-05, + "loss": 0.628, + "step": 20481 + }, + { + "epoch": 0.61, + "learning_rate": 6.978321149460237e-05, + "loss": 0.5021, + "step": 20482 + }, + { + "epoch": 0.61, + "learning_rate": 6.977401996497824e-05, + "loss": 0.5435, + "step": 20483 + }, + { + "epoch": 0.61, + "learning_rate": 6.976482871638324e-05, + "loss": 0.459, + "step": 20484 + }, + { + "epoch": 0.61, + "learning_rate": 6.975563774890281e-05, + "loss": 0.5853, + "step": 20485 + }, + { + "epoch": 0.61, + "learning_rate": 6.974644706262243e-05, + "loss": 0.5287, + "step": 20486 + }, + { + "epoch": 0.61, + "learning_rate": 6.973725665762748e-05, + "loss": 0.5601, + "step": 20487 + }, + { + "epoch": 0.61, + "learning_rate": 6.972806653400349e-05, + "loss": 0.4795, + "step": 20488 + }, + { + "epoch": 0.61, + "learning_rate": 6.971887669183584e-05, + "loss": 0.4999, + "step": 20489 + }, + { + "epoch": 0.61, + "learning_rate": 6.970968713121005e-05, + "loss": 0.5177, + "step": 20490 + }, + { + "epoch": 0.61, + "learning_rate": 6.97004978522115e-05, + "loss": 0.5133, + "step": 20491 + }, + { + "epoch": 0.61, + "learning_rate": 6.969130885492562e-05, + "loss": 0.5319, + "step": 20492 + }, + { + "epoch": 0.61, + "learning_rate": 6.96821201394379e-05, + "loss": 0.4944, + "step": 20493 + }, + { + "epoch": 0.61, + "learning_rate": 6.967293170583376e-05, + "loss": 0.5185, + "step": 20494 + }, + { + "epoch": 0.61, + "learning_rate": 6.966374355419857e-05, + "loss": 0.5411, + "step": 20495 + }, + { + "epoch": 0.61, + "learning_rate": 6.965455568461781e-05, + "loss": 0.6315, + "step": 20496 + }, + { + "epoch": 0.61, + "learning_rate": 6.964536809717691e-05, + "loss": 0.6124, + "step": 20497 + }, + { + "epoch": 0.61, + "learning_rate": 6.96361807919613e-05, + "loss": 0.5462, + "step": 20498 + }, + { + "epoch": 0.61, + "learning_rate": 6.962699376905633e-05, + "loss": 0.5328, + "step": 20499 + }, + { + "epoch": 0.61, + "learning_rate": 6.96178070285475e-05, + "loss": 0.577, + "step": 20500 + }, + { + "epoch": 0.61, + "learning_rate": 6.960862057052016e-05, + "loss": 0.4642, + "step": 20501 + }, + { + "epoch": 0.61, + "learning_rate": 6.959943439505981e-05, + "loss": 0.6157, + "step": 20502 + }, + { + "epoch": 0.61, + "learning_rate": 6.959024850225177e-05, + "loss": 0.5201, + "step": 20503 + }, + { + "epoch": 0.61, + "learning_rate": 6.958106289218145e-05, + "loss": 0.5343, + "step": 20504 + }, + { + "epoch": 0.61, + "learning_rate": 6.957187756493435e-05, + "loss": 0.6209, + "step": 20505 + }, + { + "epoch": 0.61, + "learning_rate": 6.956269252059576e-05, + "loss": 0.5239, + "step": 20506 + }, + { + "epoch": 0.61, + "learning_rate": 6.955350775925112e-05, + "loss": 0.5334, + "step": 20507 + }, + { + "epoch": 0.61, + "learning_rate": 6.954432328098583e-05, + "loss": 0.5824, + "step": 20508 + }, + { + "epoch": 0.61, + "learning_rate": 6.953513908588531e-05, + "loss": 0.6639, + "step": 20509 + }, + { + "epoch": 0.61, + "learning_rate": 6.95259551740349e-05, + "loss": 0.5041, + "step": 20510 + }, + { + "epoch": 0.61, + "learning_rate": 6.951677154552001e-05, + "loss": 0.6576, + "step": 20511 + }, + { + "epoch": 0.61, + "learning_rate": 6.950758820042604e-05, + "loss": 0.4326, + "step": 20512 + }, + { + "epoch": 0.61, + "learning_rate": 6.949840513883839e-05, + "loss": 0.6646, + "step": 20513 + }, + { + "epoch": 0.61, + "learning_rate": 6.948922236084237e-05, + "loss": 0.5124, + "step": 20514 + }, + { + "epoch": 0.61, + "learning_rate": 6.94800398665234e-05, + "loss": 0.5377, + "step": 20515 + }, + { + "epoch": 0.61, + "learning_rate": 6.947085765596689e-05, + "loss": 0.47, + "step": 20516 + }, + { + "epoch": 0.61, + "learning_rate": 6.946167572925817e-05, + "loss": 0.5634, + "step": 20517 + }, + { + "epoch": 0.61, + "learning_rate": 6.94524940864826e-05, + "loss": 0.5223, + "step": 20518 + }, + { + "epoch": 0.61, + "learning_rate": 6.944331272772556e-05, + "loss": 0.5527, + "step": 20519 + }, + { + "epoch": 0.61, + "learning_rate": 6.943413165307241e-05, + "loss": 0.6057, + "step": 20520 + }, + { + "epoch": 0.61, + "learning_rate": 6.942495086260856e-05, + "loss": 0.5311, + "step": 20521 + }, + { + "epoch": 0.61, + "learning_rate": 6.941577035641932e-05, + "loss": 0.5602, + "step": 20522 + }, + { + "epoch": 0.61, + "learning_rate": 6.940659013459003e-05, + "loss": 0.3867, + "step": 20523 + }, + { + "epoch": 0.61, + "learning_rate": 6.939741019720609e-05, + "loss": 0.59, + "step": 20524 + }, + { + "epoch": 0.61, + "learning_rate": 6.938823054435286e-05, + "loss": 0.5701, + "step": 20525 + }, + { + "epoch": 0.61, + "learning_rate": 6.937905117611563e-05, + "loss": 0.5516, + "step": 20526 + }, + { + "epoch": 0.61, + "learning_rate": 6.936987209257977e-05, + "loss": 0.5481, + "step": 20527 + }, + { + "epoch": 0.61, + "learning_rate": 6.936069329383064e-05, + "loss": 0.4751, + "step": 20528 + }, + { + "epoch": 0.61, + "learning_rate": 6.93515147799536e-05, + "loss": 0.4438, + "step": 20529 + }, + { + "epoch": 0.61, + "learning_rate": 6.934233655103392e-05, + "loss": 0.5465, + "step": 20530 + }, + { + "epoch": 0.61, + "learning_rate": 6.9333158607157e-05, + "loss": 0.5719, + "step": 20531 + }, + { + "epoch": 0.61, + "learning_rate": 6.932398094840813e-05, + "loss": 0.547, + "step": 20532 + }, + { + "epoch": 0.61, + "learning_rate": 6.93148035748727e-05, + "loss": 0.5666, + "step": 20533 + }, + { + "epoch": 0.61, + "learning_rate": 6.930562648663597e-05, + "loss": 0.5385, + "step": 20534 + }, + { + "epoch": 0.61, + "learning_rate": 6.929644968378326e-05, + "loss": 0.5879, + "step": 20535 + }, + { + "epoch": 0.61, + "learning_rate": 6.928727316639996e-05, + "loss": 0.5046, + "step": 20536 + }, + { + "epoch": 0.61, + "learning_rate": 6.927809693457137e-05, + "loss": 0.5833, + "step": 20537 + }, + { + "epoch": 0.61, + "learning_rate": 6.926892098838275e-05, + "loss": 0.4581, + "step": 20538 + }, + { + "epoch": 0.61, + "learning_rate": 6.925974532791945e-05, + "loss": 0.4962, + "step": 20539 + }, + { + "epoch": 0.61, + "learning_rate": 6.925056995326684e-05, + "loss": 0.5143, + "step": 20540 + }, + { + "epoch": 0.61, + "learning_rate": 6.924139486451013e-05, + "loss": 0.5362, + "step": 20541 + }, + { + "epoch": 0.61, + "learning_rate": 6.923222006173468e-05, + "loss": 0.5398, + "step": 20542 + }, + { + "epoch": 0.61, + "learning_rate": 6.922304554502578e-05, + "loss": 0.4169, + "step": 20543 + }, + { + "epoch": 0.61, + "learning_rate": 6.921387131446877e-05, + "loss": 0.5204, + "step": 20544 + }, + { + "epoch": 0.61, + "learning_rate": 6.920469737014887e-05, + "loss": 0.5456, + "step": 20545 + }, + { + "epoch": 0.61, + "learning_rate": 6.919552371215144e-05, + "loss": 0.5245, + "step": 20546 + }, + { + "epoch": 0.61, + "learning_rate": 6.918635034056176e-05, + "loss": 0.4798, + "step": 20547 + }, + { + "epoch": 0.61, + "learning_rate": 6.917717725546513e-05, + "loss": 0.485, + "step": 20548 + }, + { + "epoch": 0.61, + "learning_rate": 6.916800445694676e-05, + "loss": 0.5422, + "step": 20549 + }, + { + "epoch": 0.61, + "learning_rate": 6.915883194509204e-05, + "loss": 0.5192, + "step": 20550 + }, + { + "epoch": 0.61, + "learning_rate": 6.914965971998618e-05, + "loss": 0.5308, + "step": 20551 + }, + { + "epoch": 0.61, + "learning_rate": 6.914048778171453e-05, + "loss": 0.6005, + "step": 20552 + }, + { + "epoch": 0.61, + "learning_rate": 6.91313161303623e-05, + "loss": 0.5151, + "step": 20553 + }, + { + "epoch": 0.61, + "learning_rate": 6.912214476601479e-05, + "loss": 0.6454, + "step": 20554 + }, + { + "epoch": 0.61, + "learning_rate": 6.911297368875727e-05, + "loss": 0.5813, + "step": 20555 + }, + { + "epoch": 0.61, + "learning_rate": 6.910380289867504e-05, + "loss": 0.4823, + "step": 20556 + }, + { + "epoch": 0.61, + "learning_rate": 6.909463239585331e-05, + "loss": 0.527, + "step": 20557 + }, + { + "epoch": 0.61, + "learning_rate": 6.908546218037735e-05, + "loss": 0.4801, + "step": 20558 + }, + { + "epoch": 0.61, + "learning_rate": 6.907629225233247e-05, + "loss": 0.5814, + "step": 20559 + }, + { + "epoch": 0.61, + "learning_rate": 6.906712261180391e-05, + "loss": 0.5089, + "step": 20560 + }, + { + "epoch": 0.61, + "learning_rate": 6.905795325887689e-05, + "loss": 0.5893, + "step": 20561 + }, + { + "epoch": 0.61, + "learning_rate": 6.90487841936367e-05, + "loss": 0.5461, + "step": 20562 + }, + { + "epoch": 0.61, + "learning_rate": 6.903961541616855e-05, + "loss": 0.5507, + "step": 20563 + }, + { + "epoch": 0.61, + "learning_rate": 6.903044692655776e-05, + "loss": 0.6543, + "step": 20564 + }, + { + "epoch": 0.61, + "learning_rate": 6.902127872488951e-05, + "loss": 0.4703, + "step": 20565 + }, + { + "epoch": 0.61, + "learning_rate": 6.901211081124905e-05, + "loss": 0.39, + "step": 20566 + }, + { + "epoch": 0.61, + "learning_rate": 6.900294318572165e-05, + "loss": 0.6183, + "step": 20567 + }, + { + "epoch": 0.61, + "learning_rate": 6.899377584839255e-05, + "loss": 0.5689, + "step": 20568 + }, + { + "epoch": 0.61, + "learning_rate": 6.898460879934696e-05, + "loss": 0.498, + "step": 20569 + }, + { + "epoch": 0.61, + "learning_rate": 6.897544203867008e-05, + "loss": 0.6431, + "step": 20570 + }, + { + "epoch": 0.61, + "learning_rate": 6.896627556644721e-05, + "loss": 0.4994, + "step": 20571 + }, + { + "epoch": 0.61, + "learning_rate": 6.895710938276352e-05, + "loss": 0.4836, + "step": 20572 + }, + { + "epoch": 0.61, + "learning_rate": 6.894794348770425e-05, + "loss": 0.5619, + "step": 20573 + }, + { + "epoch": 0.61, + "learning_rate": 6.893877788135464e-05, + "loss": 0.4991, + "step": 20574 + }, + { + "epoch": 0.61, + "learning_rate": 6.89296125637999e-05, + "loss": 0.5295, + "step": 20575 + }, + { + "epoch": 0.61, + "learning_rate": 6.892044753512521e-05, + "loss": 0.5707, + "step": 20576 + }, + { + "epoch": 0.61, + "learning_rate": 6.891128279541583e-05, + "loss": 0.4763, + "step": 20577 + }, + { + "epoch": 0.61, + "learning_rate": 6.890211834475695e-05, + "loss": 0.5963, + "step": 20578 + }, + { + "epoch": 0.61, + "learning_rate": 6.88929541832338e-05, + "loss": 0.4817, + "step": 20579 + }, + { + "epoch": 0.61, + "learning_rate": 6.888379031093152e-05, + "loss": 0.5172, + "step": 20580 + }, + { + "epoch": 0.61, + "learning_rate": 6.88746267279354e-05, + "loss": 0.556, + "step": 20581 + }, + { + "epoch": 0.61, + "learning_rate": 6.886546343433056e-05, + "loss": 0.5155, + "step": 20582 + }, + { + "epoch": 0.61, + "learning_rate": 6.885630043020226e-05, + "loss": 0.6231, + "step": 20583 + }, + { + "epoch": 0.61, + "learning_rate": 6.884713771563566e-05, + "loss": 0.5081, + "step": 20584 + }, + { + "epoch": 0.61, + "learning_rate": 6.883797529071594e-05, + "loss": 0.5765, + "step": 20585 + }, + { + "epoch": 0.61, + "learning_rate": 6.882881315552832e-05, + "loss": 0.5943, + "step": 20586 + }, + { + "epoch": 0.61, + "learning_rate": 6.881965131015797e-05, + "loss": 0.5773, + "step": 20587 + }, + { + "epoch": 0.61, + "learning_rate": 6.881048975469009e-05, + "loss": 0.5711, + "step": 20588 + }, + { + "epoch": 0.61, + "learning_rate": 6.88013284892098e-05, + "loss": 0.6291, + "step": 20589 + }, + { + "epoch": 0.61, + "learning_rate": 6.879216751380235e-05, + "loss": 0.6317, + "step": 20590 + }, + { + "epoch": 0.61, + "learning_rate": 6.878300682855292e-05, + "loss": 0.5556, + "step": 20591 + }, + { + "epoch": 0.61, + "learning_rate": 6.877384643354661e-05, + "loss": 0.4911, + "step": 20592 + }, + { + "epoch": 0.61, + "learning_rate": 6.876468632886864e-05, + "loss": 0.5118, + "step": 20593 + }, + { + "epoch": 0.61, + "learning_rate": 6.875552651460413e-05, + "loss": 0.507, + "step": 20594 + }, + { + "epoch": 0.61, + "learning_rate": 6.874636699083836e-05, + "loss": 0.559, + "step": 20595 + }, + { + "epoch": 0.61, + "learning_rate": 6.873720775765638e-05, + "loss": 0.5488, + "step": 20596 + }, + { + "epoch": 0.61, + "learning_rate": 6.872804881514334e-05, + "loss": 0.5162, + "step": 20597 + }, + { + "epoch": 0.61, + "learning_rate": 6.871889016338448e-05, + "loss": 0.5861, + "step": 20598 + }, + { + "epoch": 0.61, + "learning_rate": 6.870973180246492e-05, + "loss": 0.509, + "step": 20599 + }, + { + "epoch": 0.61, + "learning_rate": 6.87005737324698e-05, + "loss": 0.6945, + "step": 20600 + }, + { + "epoch": 0.61, + "learning_rate": 6.869141595348425e-05, + "loss": 0.565, + "step": 20601 + }, + { + "epoch": 0.61, + "learning_rate": 6.868225846559344e-05, + "loss": 0.4826, + "step": 20602 + }, + { + "epoch": 0.61, + "learning_rate": 6.867310126888253e-05, + "loss": 0.5029, + "step": 20603 + }, + { + "epoch": 0.61, + "learning_rate": 6.86639443634366e-05, + "loss": 0.633, + "step": 20604 + }, + { + "epoch": 0.61, + "learning_rate": 6.865478774934085e-05, + "loss": 0.4891, + "step": 20605 + }, + { + "epoch": 0.61, + "learning_rate": 6.864563142668041e-05, + "loss": 0.5339, + "step": 20606 + }, + { + "epoch": 0.61, + "learning_rate": 6.863647539554037e-05, + "loss": 0.535, + "step": 20607 + }, + { + "epoch": 0.61, + "learning_rate": 6.862731965600587e-05, + "loss": 0.5927, + "step": 20608 + }, + { + "epoch": 0.61, + "learning_rate": 6.861816420816205e-05, + "loss": 0.5551, + "step": 20609 + }, + { + "epoch": 0.61, + "learning_rate": 6.860900905209408e-05, + "loss": 0.6124, + "step": 20610 + }, + { + "epoch": 0.61, + "learning_rate": 6.859985418788697e-05, + "loss": 0.5372, + "step": 20611 + }, + { + "epoch": 0.61, + "learning_rate": 6.859069961562594e-05, + "loss": 0.5788, + "step": 20612 + }, + { + "epoch": 0.61, + "learning_rate": 6.858154533539602e-05, + "loss": 0.5236, + "step": 20613 + }, + { + "epoch": 0.61, + "learning_rate": 6.857239134728244e-05, + "loss": 0.5235, + "step": 20614 + }, + { + "epoch": 0.61, + "learning_rate": 6.85632376513702e-05, + "loss": 0.4835, + "step": 20615 + }, + { + "epoch": 0.61, + "learning_rate": 6.855408424774442e-05, + "loss": 0.528, + "step": 20616 + }, + { + "epoch": 0.61, + "learning_rate": 6.854493113649026e-05, + "loss": 0.4816, + "step": 20617 + }, + { + "epoch": 0.61, + "learning_rate": 6.853577831769283e-05, + "loss": 0.5604, + "step": 20618 + }, + { + "epoch": 0.61, + "learning_rate": 6.852662579143713e-05, + "loss": 0.5169, + "step": 20619 + }, + { + "epoch": 0.61, + "learning_rate": 6.851747355780835e-05, + "loss": 0.5183, + "step": 20620 + }, + { + "epoch": 0.61, + "learning_rate": 6.850832161689154e-05, + "loss": 0.531, + "step": 20621 + }, + { + "epoch": 0.61, + "learning_rate": 6.849916996877184e-05, + "loss": 0.6194, + "step": 20622 + }, + { + "epoch": 0.61, + "learning_rate": 6.849001861353426e-05, + "loss": 0.4933, + "step": 20623 + }, + { + "epoch": 0.61, + "learning_rate": 6.848086755126394e-05, + "loss": 0.4905, + "step": 20624 + }, + { + "epoch": 0.61, + "learning_rate": 6.847171678204594e-05, + "loss": 0.4787, + "step": 20625 + }, + { + "epoch": 0.61, + "learning_rate": 6.846256630596539e-05, + "loss": 0.5444, + "step": 20626 + }, + { + "epoch": 0.61, + "learning_rate": 6.845341612310729e-05, + "loss": 0.5906, + "step": 20627 + }, + { + "epoch": 0.61, + "learning_rate": 6.844426623355675e-05, + "loss": 0.497, + "step": 20628 + }, + { + "epoch": 0.61, + "learning_rate": 6.843511663739888e-05, + "loss": 0.5443, + "step": 20629 + }, + { + "epoch": 0.61, + "learning_rate": 6.842596733471872e-05, + "loss": 0.5411, + "step": 20630 + }, + { + "epoch": 0.61, + "learning_rate": 6.841681832560132e-05, + "loss": 0.5573, + "step": 20631 + }, + { + "epoch": 0.61, + "learning_rate": 6.840766961013172e-05, + "loss": 0.7015, + "step": 20632 + }, + { + "epoch": 0.61, + "learning_rate": 6.839852118839505e-05, + "loss": 0.5601, + "step": 20633 + }, + { + "epoch": 0.61, + "learning_rate": 6.838937306047635e-05, + "loss": 0.5807, + "step": 20634 + }, + { + "epoch": 0.61, + "learning_rate": 6.838022522646062e-05, + "loss": 0.6502, + "step": 20635 + }, + { + "epoch": 0.61, + "learning_rate": 6.837107768643296e-05, + "loss": 0.5986, + "step": 20636 + }, + { + "epoch": 0.61, + "learning_rate": 6.836193044047845e-05, + "loss": 0.4986, + "step": 20637 + }, + { + "epoch": 0.61, + "learning_rate": 6.835278348868205e-05, + "loss": 0.5433, + "step": 20638 + }, + { + "epoch": 0.61, + "learning_rate": 6.834363683112889e-05, + "loss": 0.4739, + "step": 20639 + }, + { + "epoch": 0.61, + "learning_rate": 6.833449046790396e-05, + "loss": 0.5633, + "step": 20640 + }, + { + "epoch": 0.61, + "learning_rate": 6.832534439909237e-05, + "loss": 0.5853, + "step": 20641 + }, + { + "epoch": 0.61, + "learning_rate": 6.831619862477903e-05, + "loss": 0.5322, + "step": 20642 + }, + { + "epoch": 0.61, + "learning_rate": 6.830705314504908e-05, + "loss": 0.5576, + "step": 20643 + }, + { + "epoch": 0.61, + "learning_rate": 6.829790795998751e-05, + "loss": 0.5801, + "step": 20644 + }, + { + "epoch": 0.61, + "learning_rate": 6.828876306967938e-05, + "loss": 0.5572, + "step": 20645 + }, + { + "epoch": 0.61, + "learning_rate": 6.827961847420967e-05, + "loss": 0.6313, + "step": 20646 + }, + { + "epoch": 0.61, + "learning_rate": 6.827047417366341e-05, + "loss": 0.523, + "step": 20647 + }, + { + "epoch": 0.61, + "learning_rate": 6.826133016812566e-05, + "loss": 0.4587, + "step": 20648 + }, + { + "epoch": 0.61, + "learning_rate": 6.825218645768143e-05, + "loss": 0.482, + "step": 20649 + }, + { + "epoch": 0.61, + "learning_rate": 6.824304304241568e-05, + "loss": 0.5868, + "step": 20650 + }, + { + "epoch": 0.61, + "learning_rate": 6.823389992241347e-05, + "loss": 0.4885, + "step": 20651 + }, + { + "epoch": 0.61, + "learning_rate": 6.82247570977598e-05, + "loss": 0.5558, + "step": 20652 + }, + { + "epoch": 0.61, + "learning_rate": 6.82156145685397e-05, + "loss": 0.5344, + "step": 20653 + }, + { + "epoch": 0.61, + "learning_rate": 6.82064723348381e-05, + "loss": 0.3654, + "step": 20654 + }, + { + "epoch": 0.61, + "learning_rate": 6.819733039674008e-05, + "loss": 0.5696, + "step": 20655 + }, + { + "epoch": 0.61, + "learning_rate": 6.818818875433057e-05, + "loss": 0.4823, + "step": 20656 + }, + { + "epoch": 0.61, + "learning_rate": 6.817904740769466e-05, + "loss": 0.532, + "step": 20657 + }, + { + "epoch": 0.62, + "learning_rate": 6.816990635691724e-05, + "loss": 0.5239, + "step": 20658 + }, + { + "epoch": 0.62, + "learning_rate": 6.816076560208336e-05, + "loss": 0.5368, + "step": 20659 + }, + { + "epoch": 0.62, + "learning_rate": 6.815162514327798e-05, + "loss": 0.5447, + "step": 20660 + }, + { + "epoch": 0.62, + "learning_rate": 6.814248498058613e-05, + "loss": 0.6479, + "step": 20661 + }, + { + "epoch": 0.62, + "learning_rate": 6.813334511409273e-05, + "loss": 0.4741, + "step": 20662 + }, + { + "epoch": 0.62, + "learning_rate": 6.812420554388278e-05, + "loss": 0.5816, + "step": 20663 + }, + { + "epoch": 0.62, + "learning_rate": 6.811506627004127e-05, + "loss": 0.5917, + "step": 20664 + }, + { + "epoch": 0.62, + "learning_rate": 6.81059272926532e-05, + "loss": 0.6068, + "step": 20665 + }, + { + "epoch": 0.62, + "learning_rate": 6.809678861180346e-05, + "loss": 0.4886, + "step": 20666 + }, + { + "epoch": 0.62, + "learning_rate": 6.808765022757708e-05, + "loss": 0.6215, + "step": 20667 + }, + { + "epoch": 0.62, + "learning_rate": 6.807851214005898e-05, + "loss": 0.5176, + "step": 20668 + }, + { + "epoch": 0.62, + "learning_rate": 6.806937434933421e-05, + "loss": 0.6057, + "step": 20669 + }, + { + "epoch": 0.62, + "learning_rate": 6.806023685548764e-05, + "loss": 0.6124, + "step": 20670 + }, + { + "epoch": 0.62, + "learning_rate": 6.805109965860423e-05, + "loss": 0.5353, + "step": 20671 + }, + { + "epoch": 0.62, + "learning_rate": 6.804196275876904e-05, + "loss": 0.5163, + "step": 20672 + }, + { + "epoch": 0.62, + "learning_rate": 6.803282615606687e-05, + "loss": 0.6805, + "step": 20673 + }, + { + "epoch": 0.62, + "learning_rate": 6.802368985058277e-05, + "loss": 0.5695, + "step": 20674 + }, + { + "epoch": 0.62, + "learning_rate": 6.801455384240163e-05, + "loss": 0.5738, + "step": 20675 + }, + { + "epoch": 0.62, + "learning_rate": 6.800541813160845e-05, + "loss": 0.5176, + "step": 20676 + }, + { + "epoch": 0.62, + "learning_rate": 6.799628271828816e-05, + "loss": 0.601, + "step": 20677 + }, + { + "epoch": 0.62, + "learning_rate": 6.798714760252562e-05, + "loss": 0.594, + "step": 20678 + }, + { + "epoch": 0.62, + "learning_rate": 6.797801278440585e-05, + "loss": 0.5565, + "step": 20679 + }, + { + "epoch": 0.62, + "learning_rate": 6.796887826401379e-05, + "loss": 0.6285, + "step": 20680 + }, + { + "epoch": 0.62, + "learning_rate": 6.795974404143428e-05, + "loss": 0.5876, + "step": 20681 + }, + { + "epoch": 0.62, + "learning_rate": 6.795061011675232e-05, + "loss": 0.6135, + "step": 20682 + }, + { + "epoch": 0.62, + "learning_rate": 6.794147649005281e-05, + "loss": 0.6005, + "step": 20683 + }, + { + "epoch": 0.62, + "learning_rate": 6.793234316142069e-05, + "loss": 0.5504, + "step": 20684 + }, + { + "epoch": 0.62, + "learning_rate": 6.792321013094085e-05, + "loss": 0.4045, + "step": 20685 + }, + { + "epoch": 0.62, + "learning_rate": 6.791407739869822e-05, + "loss": 0.5822, + "step": 20686 + }, + { + "epoch": 0.62, + "learning_rate": 6.790494496477768e-05, + "loss": 0.4865, + "step": 20687 + }, + { + "epoch": 0.62, + "learning_rate": 6.789581282926422e-05, + "loss": 0.4711, + "step": 20688 + }, + { + "epoch": 0.62, + "learning_rate": 6.788668099224268e-05, + "loss": 0.4953, + "step": 20689 + }, + { + "epoch": 0.62, + "learning_rate": 6.787754945379796e-05, + "loss": 0.6196, + "step": 20690 + }, + { + "epoch": 0.62, + "learning_rate": 6.7868418214015e-05, + "loss": 0.6217, + "step": 20691 + }, + { + "epoch": 0.62, + "learning_rate": 6.78592872729787e-05, + "loss": 0.5305, + "step": 20692 + }, + { + "epoch": 0.62, + "learning_rate": 6.785015663077393e-05, + "loss": 0.4162, + "step": 20693 + }, + { + "epoch": 0.62, + "learning_rate": 6.784102628748554e-05, + "loss": 0.4385, + "step": 20694 + }, + { + "epoch": 0.62, + "learning_rate": 6.783189624319852e-05, + "loss": 0.6651, + "step": 20695 + }, + { + "epoch": 0.62, + "learning_rate": 6.782276649799771e-05, + "loss": 0.5622, + "step": 20696 + }, + { + "epoch": 0.62, + "learning_rate": 6.781363705196798e-05, + "loss": 0.5961, + "step": 20697 + }, + { + "epoch": 0.62, + "learning_rate": 6.780450790519421e-05, + "loss": 0.4756, + "step": 20698 + }, + { + "epoch": 0.62, + "learning_rate": 6.779537905776128e-05, + "loss": 0.554, + "step": 20699 + }, + { + "epoch": 0.62, + "learning_rate": 6.778625050975414e-05, + "loss": 0.5945, + "step": 20700 + }, + { + "epoch": 0.62, + "learning_rate": 6.777712226125757e-05, + "loss": 0.4735, + "step": 20701 + }, + { + "epoch": 0.62, + "learning_rate": 6.776799431235645e-05, + "loss": 0.6238, + "step": 20702 + }, + { + "epoch": 0.62, + "learning_rate": 6.775886666313573e-05, + "loss": 0.5495, + "step": 20703 + }, + { + "epoch": 0.62, + "learning_rate": 6.774973931368017e-05, + "loss": 0.5072, + "step": 20704 + }, + { + "epoch": 0.62, + "learning_rate": 6.77406122640747e-05, + "loss": 0.5442, + "step": 20705 + }, + { + "epoch": 0.62, + "learning_rate": 6.773148551440413e-05, + "loss": 0.5129, + "step": 20706 + }, + { + "epoch": 0.62, + "learning_rate": 6.77223590647534e-05, + "loss": 0.5743, + "step": 20707 + }, + { + "epoch": 0.62, + "learning_rate": 6.771323291520728e-05, + "loss": 0.6054, + "step": 20708 + }, + { + "epoch": 0.62, + "learning_rate": 6.770410706585063e-05, + "loss": 0.4496, + "step": 20709 + }, + { + "epoch": 0.62, + "learning_rate": 6.769498151676834e-05, + "loss": 0.6611, + "step": 20710 + }, + { + "epoch": 0.62, + "learning_rate": 6.768585626804526e-05, + "loss": 0.6017, + "step": 20711 + }, + { + "epoch": 0.62, + "learning_rate": 6.767673131976617e-05, + "loss": 0.6077, + "step": 20712 + }, + { + "epoch": 0.62, + "learning_rate": 6.766760667201598e-05, + "loss": 0.5614, + "step": 20713 + }, + { + "epoch": 0.62, + "learning_rate": 6.76584823248795e-05, + "loss": 0.5291, + "step": 20714 + }, + { + "epoch": 0.62, + "learning_rate": 6.764935827844157e-05, + "loss": 0.4666, + "step": 20715 + }, + { + "epoch": 0.62, + "learning_rate": 6.764023453278696e-05, + "loss": 0.5145, + "step": 20716 + }, + { + "epoch": 0.62, + "learning_rate": 6.763111108800058e-05, + "loss": 0.5834, + "step": 20717 + }, + { + "epoch": 0.62, + "learning_rate": 6.762198794416721e-05, + "loss": 0.569, + "step": 20718 + }, + { + "epoch": 0.62, + "learning_rate": 6.761286510137175e-05, + "loss": 0.5197, + "step": 20719 + }, + { + "epoch": 0.62, + "learning_rate": 6.76037425596989e-05, + "loss": 0.4652, + "step": 20720 + }, + { + "epoch": 0.62, + "learning_rate": 6.759462031923356e-05, + "loss": 0.6, + "step": 20721 + }, + { + "epoch": 0.62, + "learning_rate": 6.758549838006053e-05, + "loss": 0.6144, + "step": 20722 + }, + { + "epoch": 0.62, + "learning_rate": 6.757637674226464e-05, + "loss": 0.4425, + "step": 20723 + }, + { + "epoch": 0.62, + "learning_rate": 6.756725540593066e-05, + "loss": 0.509, + "step": 20724 + }, + { + "epoch": 0.62, + "learning_rate": 6.755813437114338e-05, + "loss": 0.5746, + "step": 20725 + }, + { + "epoch": 0.62, + "learning_rate": 6.754901363798766e-05, + "loss": 0.5102, + "step": 20726 + }, + { + "epoch": 0.62, + "learning_rate": 6.75398932065483e-05, + "loss": 0.629, + "step": 20727 + }, + { + "epoch": 0.62, + "learning_rate": 6.753077307691004e-05, + "loss": 0.5412, + "step": 20728 + }, + { + "epoch": 0.62, + "learning_rate": 6.752165324915772e-05, + "loss": 0.5949, + "step": 20729 + }, + { + "epoch": 0.62, + "learning_rate": 6.751253372337611e-05, + "loss": 0.4697, + "step": 20730 + }, + { + "epoch": 0.62, + "learning_rate": 6.750341449965005e-05, + "loss": 0.6638, + "step": 20731 + }, + { + "epoch": 0.62, + "learning_rate": 6.749429557806428e-05, + "loss": 0.5694, + "step": 20732 + }, + { + "epoch": 0.62, + "learning_rate": 6.748517695870357e-05, + "loss": 0.4565, + "step": 20733 + }, + { + "epoch": 0.62, + "learning_rate": 6.747605864165273e-05, + "loss": 0.6039, + "step": 20734 + }, + { + "epoch": 0.62, + "learning_rate": 6.746694062699657e-05, + "loss": 0.5544, + "step": 20735 + }, + { + "epoch": 0.62, + "learning_rate": 6.745782291481981e-05, + "loss": 0.7344, + "step": 20736 + }, + { + "epoch": 0.62, + "learning_rate": 6.744870550520723e-05, + "loss": 0.5127, + "step": 20737 + }, + { + "epoch": 0.62, + "learning_rate": 6.743958839824364e-05, + "loss": 0.5234, + "step": 20738 + }, + { + "epoch": 0.62, + "learning_rate": 6.743047159401377e-05, + "loss": 0.555, + "step": 20739 + }, + { + "epoch": 0.62, + "learning_rate": 6.742135509260235e-05, + "loss": 0.5773, + "step": 20740 + }, + { + "epoch": 0.62, + "learning_rate": 6.741223889409423e-05, + "loss": 0.6838, + "step": 20741 + }, + { + "epoch": 0.62, + "learning_rate": 6.740312299857414e-05, + "loss": 0.4588, + "step": 20742 + }, + { + "epoch": 0.62, + "learning_rate": 6.739400740612676e-05, + "loss": 0.578, + "step": 20743 + }, + { + "epoch": 0.62, + "learning_rate": 6.738489211683695e-05, + "loss": 0.523, + "step": 20744 + }, + { + "epoch": 0.62, + "learning_rate": 6.737577713078939e-05, + "loss": 0.5175, + "step": 20745 + }, + { + "epoch": 0.62, + "learning_rate": 6.736666244806891e-05, + "loss": 0.4929, + "step": 20746 + }, + { + "epoch": 0.62, + "learning_rate": 6.735754806876012e-05, + "loss": 0.4778, + "step": 20747 + }, + { + "epoch": 0.62, + "learning_rate": 6.734843399294788e-05, + "loss": 0.5483, + "step": 20748 + }, + { + "epoch": 0.62, + "learning_rate": 6.733932022071685e-05, + "loss": 0.5124, + "step": 20749 + }, + { + "epoch": 0.62, + "learning_rate": 6.733020675215185e-05, + "loss": 0.5449, + "step": 20750 + }, + { + "epoch": 0.62, + "learning_rate": 6.732109358733756e-05, + "loss": 0.555, + "step": 20751 + }, + { + "epoch": 0.62, + "learning_rate": 6.731198072635866e-05, + "loss": 0.5289, + "step": 20752 + }, + { + "epoch": 0.62, + "learning_rate": 6.730286816929999e-05, + "loss": 0.5264, + "step": 20753 + }, + { + "epoch": 0.62, + "learning_rate": 6.729375591624623e-05, + "loss": 0.4962, + "step": 20754 + }, + { + "epoch": 0.62, + "learning_rate": 6.728464396728207e-05, + "loss": 0.5453, + "step": 20755 + }, + { + "epoch": 0.62, + "learning_rate": 6.727553232249225e-05, + "loss": 0.4921, + "step": 20756 + }, + { + "epoch": 0.62, + "learning_rate": 6.726642098196148e-05, + "loss": 0.4692, + "step": 20757 + }, + { + "epoch": 0.62, + "learning_rate": 6.725730994577451e-05, + "loss": 0.496, + "step": 20758 + }, + { + "epoch": 0.62, + "learning_rate": 6.724819921401599e-05, + "loss": 0.5103, + "step": 20759 + }, + { + "epoch": 0.62, + "learning_rate": 6.723908878677067e-05, + "loss": 0.4799, + "step": 20760 + }, + { + "epoch": 0.62, + "learning_rate": 6.722997866412324e-05, + "loss": 0.6085, + "step": 20761 + }, + { + "epoch": 0.62, + "learning_rate": 6.722086884615844e-05, + "loss": 0.579, + "step": 20762 + }, + { + "epoch": 0.62, + "learning_rate": 6.72117593329609e-05, + "loss": 0.4917, + "step": 20763 + }, + { + "epoch": 0.62, + "learning_rate": 6.720265012461535e-05, + "loss": 0.5698, + "step": 20764 + }, + { + "epoch": 0.62, + "learning_rate": 6.719354122120652e-05, + "loss": 0.5288, + "step": 20765 + }, + { + "epoch": 0.62, + "learning_rate": 6.718443262281906e-05, + "loss": 0.5499, + "step": 20766 + }, + { + "epoch": 0.62, + "learning_rate": 6.717532432953765e-05, + "loss": 0.5209, + "step": 20767 + }, + { + "epoch": 0.62, + "learning_rate": 6.7166216341447e-05, + "loss": 0.5086, + "step": 20768 + }, + { + "epoch": 0.62, + "learning_rate": 6.715710865863181e-05, + "loss": 0.5864, + "step": 20769 + }, + { + "epoch": 0.62, + "learning_rate": 6.71480012811767e-05, + "loss": 0.6226, + "step": 20770 + }, + { + "epoch": 0.62, + "learning_rate": 6.713889420916637e-05, + "loss": 0.5895, + "step": 20771 + }, + { + "epoch": 0.62, + "learning_rate": 6.712978744268552e-05, + "loss": 0.4583, + "step": 20772 + }, + { + "epoch": 0.62, + "learning_rate": 6.712068098181883e-05, + "loss": 0.5697, + "step": 20773 + }, + { + "epoch": 0.62, + "learning_rate": 6.711157482665091e-05, + "loss": 0.5862, + "step": 20774 + }, + { + "epoch": 0.62, + "learning_rate": 6.710246897726648e-05, + "loss": 0.5746, + "step": 20775 + }, + { + "epoch": 0.62, + "learning_rate": 6.709336343375014e-05, + "loss": 0.5804, + "step": 20776 + }, + { + "epoch": 0.62, + "learning_rate": 6.708425819618668e-05, + "loss": 0.5388, + "step": 20777 + }, + { + "epoch": 0.62, + "learning_rate": 6.70751532646606e-05, + "loss": 0.5174, + "step": 20778 + }, + { + "epoch": 0.62, + "learning_rate": 6.706604863925663e-05, + "loss": 0.5762, + "step": 20779 + }, + { + "epoch": 0.62, + "learning_rate": 6.70569443200594e-05, + "loss": 0.645, + "step": 20780 + }, + { + "epoch": 0.62, + "learning_rate": 6.704784030715361e-05, + "loss": 0.5524, + "step": 20781 + }, + { + "epoch": 0.62, + "learning_rate": 6.703873660062386e-05, + "loss": 0.5206, + "step": 20782 + }, + { + "epoch": 0.62, + "learning_rate": 6.702963320055476e-05, + "loss": 0.5655, + "step": 20783 + }, + { + "epoch": 0.62, + "learning_rate": 6.702053010703102e-05, + "loss": 0.5291, + "step": 20784 + }, + { + "epoch": 0.62, + "learning_rate": 6.701142732013728e-05, + "loss": 0.4598, + "step": 20785 + }, + { + "epoch": 0.62, + "learning_rate": 6.700232483995808e-05, + "loss": 0.5797, + "step": 20786 + }, + { + "epoch": 0.62, + "learning_rate": 6.699322266657814e-05, + "loss": 0.4763, + "step": 20787 + }, + { + "epoch": 0.62, + "learning_rate": 6.698412080008206e-05, + "loss": 0.4987, + "step": 20788 + }, + { + "epoch": 0.62, + "learning_rate": 6.697501924055449e-05, + "loss": 0.5533, + "step": 20789 + }, + { + "epoch": 0.62, + "learning_rate": 6.696591798807998e-05, + "loss": 0.5846, + "step": 20790 + }, + { + "epoch": 0.62, + "learning_rate": 6.695681704274324e-05, + "loss": 0.4747, + "step": 20791 + }, + { + "epoch": 0.62, + "learning_rate": 6.69477164046288e-05, + "loss": 0.552, + "step": 20792 + }, + { + "epoch": 0.62, + "learning_rate": 6.693861607382137e-05, + "loss": 0.5393, + "step": 20793 + }, + { + "epoch": 0.62, + "learning_rate": 6.692951605040549e-05, + "loss": 0.4801, + "step": 20794 + }, + { + "epoch": 0.62, + "learning_rate": 6.692041633446577e-05, + "loss": 0.5373, + "step": 20795 + }, + { + "epoch": 0.62, + "learning_rate": 6.691131692608685e-05, + "loss": 0.5007, + "step": 20796 + }, + { + "epoch": 0.62, + "learning_rate": 6.690221782535334e-05, + "loss": 0.6012, + "step": 20797 + }, + { + "epoch": 0.62, + "learning_rate": 6.689311903234981e-05, + "loss": 0.5708, + "step": 20798 + }, + { + "epoch": 0.62, + "learning_rate": 6.688402054716082e-05, + "loss": 0.5541, + "step": 20799 + }, + { + "epoch": 0.62, + "learning_rate": 6.687492236987106e-05, + "loss": 0.4912, + "step": 20800 + }, + { + "epoch": 0.62, + "learning_rate": 6.686582450056505e-05, + "loss": 0.5281, + "step": 20801 + }, + { + "epoch": 0.62, + "learning_rate": 6.685672693932738e-05, + "loss": 0.5345, + "step": 20802 + }, + { + "epoch": 0.62, + "learning_rate": 6.684762968624267e-05, + "loss": 0.596, + "step": 20803 + }, + { + "epoch": 0.62, + "learning_rate": 6.683853274139551e-05, + "loss": 0.4728, + "step": 20804 + }, + { + "epoch": 0.62, + "learning_rate": 6.68294361048704e-05, + "loss": 0.4434, + "step": 20805 + }, + { + "epoch": 0.62, + "learning_rate": 6.682033977675202e-05, + "loss": 0.5659, + "step": 20806 + }, + { + "epoch": 0.62, + "learning_rate": 6.681124375712486e-05, + "loss": 0.5818, + "step": 20807 + }, + { + "epoch": 0.62, + "learning_rate": 6.680214804607358e-05, + "loss": 0.4648, + "step": 20808 + }, + { + "epoch": 0.62, + "learning_rate": 6.679305264368265e-05, + "loss": 0.6184, + "step": 20809 + }, + { + "epoch": 0.62, + "learning_rate": 6.67839575500367e-05, + "loss": 0.4587, + "step": 20810 + }, + { + "epoch": 0.62, + "learning_rate": 6.677486276522025e-05, + "loss": 0.4864, + "step": 20811 + }, + { + "epoch": 0.62, + "learning_rate": 6.676576828931794e-05, + "loss": 0.562, + "step": 20812 + }, + { + "epoch": 0.62, + "learning_rate": 6.675667412241422e-05, + "loss": 0.4801, + "step": 20813 + }, + { + "epoch": 0.62, + "learning_rate": 6.674758026459369e-05, + "loss": 0.5791, + "step": 20814 + }, + { + "epoch": 0.62, + "learning_rate": 6.673848671594092e-05, + "loss": 0.4779, + "step": 20815 + }, + { + "epoch": 0.62, + "learning_rate": 6.672939347654048e-05, + "loss": 0.509, + "step": 20816 + }, + { + "epoch": 0.62, + "learning_rate": 6.672030054647682e-05, + "loss": 0.5675, + "step": 20817 + }, + { + "epoch": 0.62, + "learning_rate": 6.671120792583457e-05, + "loss": 0.5439, + "step": 20818 + }, + { + "epoch": 0.62, + "learning_rate": 6.670211561469825e-05, + "loss": 0.5623, + "step": 20819 + }, + { + "epoch": 0.62, + "learning_rate": 6.669302361315239e-05, + "loss": 0.6279, + "step": 20820 + }, + { + "epoch": 0.62, + "learning_rate": 6.668393192128149e-05, + "loss": 0.5878, + "step": 20821 + }, + { + "epoch": 0.62, + "learning_rate": 6.667484053917013e-05, + "loss": 0.4071, + "step": 20822 + }, + { + "epoch": 0.62, + "learning_rate": 6.66657494669028e-05, + "loss": 0.5766, + "step": 20823 + }, + { + "epoch": 0.62, + "learning_rate": 6.66566587045641e-05, + "loss": 0.507, + "step": 20824 + }, + { + "epoch": 0.62, + "learning_rate": 6.664756825223846e-05, + "loss": 0.5701, + "step": 20825 + }, + { + "epoch": 0.62, + "learning_rate": 6.663847811001043e-05, + "loss": 0.5491, + "step": 20826 + }, + { + "epoch": 0.62, + "learning_rate": 6.662938827796454e-05, + "loss": 0.4839, + "step": 20827 + }, + { + "epoch": 0.62, + "learning_rate": 6.662029875618534e-05, + "loss": 0.6516, + "step": 20828 + }, + { + "epoch": 0.62, + "learning_rate": 6.661120954475725e-05, + "loss": 0.4577, + "step": 20829 + }, + { + "epoch": 0.62, + "learning_rate": 6.660212064376483e-05, + "loss": 0.5496, + "step": 20830 + }, + { + "epoch": 0.62, + "learning_rate": 6.659303205329259e-05, + "loss": 0.651, + "step": 20831 + }, + { + "epoch": 0.62, + "learning_rate": 6.658394377342505e-05, + "loss": 0.5948, + "step": 20832 + }, + { + "epoch": 0.62, + "learning_rate": 6.657485580424665e-05, + "loss": 0.5313, + "step": 20833 + }, + { + "epoch": 0.62, + "learning_rate": 6.656576814584192e-05, + "loss": 0.6322, + "step": 20834 + }, + { + "epoch": 0.62, + "learning_rate": 6.655668079829537e-05, + "loss": 0.5437, + "step": 20835 + }, + { + "epoch": 0.62, + "learning_rate": 6.654759376169144e-05, + "loss": 0.6004, + "step": 20836 + }, + { + "epoch": 0.62, + "learning_rate": 6.653850703611468e-05, + "loss": 0.5431, + "step": 20837 + }, + { + "epoch": 0.62, + "learning_rate": 6.652942062164952e-05, + "loss": 0.5799, + "step": 20838 + }, + { + "epoch": 0.62, + "learning_rate": 6.652033451838052e-05, + "loss": 0.6996, + "step": 20839 + }, + { + "epoch": 0.62, + "learning_rate": 6.651124872639205e-05, + "loss": 0.5345, + "step": 20840 + }, + { + "epoch": 0.62, + "learning_rate": 6.650216324576865e-05, + "loss": 0.4457, + "step": 20841 + }, + { + "epoch": 0.62, + "learning_rate": 6.649307807659479e-05, + "loss": 0.6721, + "step": 20842 + }, + { + "epoch": 0.62, + "learning_rate": 6.648399321895496e-05, + "loss": 0.5518, + "step": 20843 + }, + { + "epoch": 0.62, + "learning_rate": 6.647490867293358e-05, + "loss": 0.5756, + "step": 20844 + }, + { + "epoch": 0.62, + "learning_rate": 6.646582443861514e-05, + "loss": 0.5055, + "step": 20845 + }, + { + "epoch": 0.62, + "learning_rate": 6.64567405160841e-05, + "loss": 0.5471, + "step": 20846 + }, + { + "epoch": 0.62, + "learning_rate": 6.644765690542494e-05, + "loss": 0.6449, + "step": 20847 + }, + { + "epoch": 0.62, + "learning_rate": 6.643857360672206e-05, + "loss": 0.52, + "step": 20848 + }, + { + "epoch": 0.62, + "learning_rate": 6.642949062005994e-05, + "loss": 0.4601, + "step": 20849 + }, + { + "epoch": 0.62, + "learning_rate": 6.642040794552307e-05, + "loss": 0.5506, + "step": 20850 + }, + { + "epoch": 0.62, + "learning_rate": 6.641132558319587e-05, + "loss": 0.6001, + "step": 20851 + }, + { + "epoch": 0.62, + "learning_rate": 6.640224353316273e-05, + "loss": 0.456, + "step": 20852 + }, + { + "epoch": 0.62, + "learning_rate": 6.639316179550816e-05, + "loss": 0.5462, + "step": 20853 + }, + { + "epoch": 0.62, + "learning_rate": 6.638408037031657e-05, + "loss": 0.5122, + "step": 20854 + }, + { + "epoch": 0.62, + "learning_rate": 6.637499925767243e-05, + "loss": 0.5692, + "step": 20855 + }, + { + "epoch": 0.62, + "learning_rate": 6.636591845766014e-05, + "loss": 0.5064, + "step": 20856 + }, + { + "epoch": 0.62, + "learning_rate": 6.635683797036409e-05, + "loss": 0.5583, + "step": 20857 + }, + { + "epoch": 0.62, + "learning_rate": 6.634775779586879e-05, + "loss": 0.5074, + "step": 20858 + }, + { + "epoch": 0.62, + "learning_rate": 6.633867793425866e-05, + "loss": 0.5613, + "step": 20859 + }, + { + "epoch": 0.62, + "learning_rate": 6.632959838561804e-05, + "loss": 0.6536, + "step": 20860 + }, + { + "epoch": 0.62, + "learning_rate": 6.63205191500314e-05, + "loss": 0.5042, + "step": 20861 + }, + { + "epoch": 0.62, + "learning_rate": 6.631144022758314e-05, + "loss": 0.6114, + "step": 20862 + }, + { + "epoch": 0.62, + "learning_rate": 6.630236161835773e-05, + "loss": 0.4405, + "step": 20863 + }, + { + "epoch": 0.62, + "learning_rate": 6.629328332243947e-05, + "loss": 0.4501, + "step": 20864 + }, + { + "epoch": 0.62, + "learning_rate": 6.628420533991284e-05, + "loss": 0.4989, + "step": 20865 + }, + { + "epoch": 0.62, + "learning_rate": 6.627512767086228e-05, + "loss": 0.5083, + "step": 20866 + }, + { + "epoch": 0.62, + "learning_rate": 6.626605031537207e-05, + "loss": 0.6015, + "step": 20867 + }, + { + "epoch": 0.62, + "learning_rate": 6.625697327352674e-05, + "loss": 0.4576, + "step": 20868 + }, + { + "epoch": 0.62, + "learning_rate": 6.624789654541057e-05, + "loss": 0.5668, + "step": 20869 + }, + { + "epoch": 0.62, + "learning_rate": 6.623882013110807e-05, + "loss": 0.5021, + "step": 20870 + }, + { + "epoch": 0.62, + "learning_rate": 6.622974403070351e-05, + "loss": 0.5936, + "step": 20871 + }, + { + "epoch": 0.62, + "learning_rate": 6.622066824428135e-05, + "loss": 0.6451, + "step": 20872 + }, + { + "epoch": 0.62, + "learning_rate": 6.621159277192594e-05, + "loss": 0.5221, + "step": 20873 + }, + { + "epoch": 0.62, + "learning_rate": 6.62025176137217e-05, + "loss": 0.5727, + "step": 20874 + }, + { + "epoch": 0.62, + "learning_rate": 6.619344276975297e-05, + "loss": 0.426, + "step": 20875 + }, + { + "epoch": 0.62, + "learning_rate": 6.618436824010413e-05, + "loss": 0.4423, + "step": 20876 + }, + { + "epoch": 0.62, + "learning_rate": 6.617529402485955e-05, + "loss": 0.4707, + "step": 20877 + }, + { + "epoch": 0.62, + "learning_rate": 6.616622012410363e-05, + "loss": 0.5312, + "step": 20878 + }, + { + "epoch": 0.62, + "learning_rate": 6.615714653792068e-05, + "loss": 0.4798, + "step": 20879 + }, + { + "epoch": 0.62, + "learning_rate": 6.614807326639511e-05, + "loss": 0.4638, + "step": 20880 + }, + { + "epoch": 0.62, + "learning_rate": 6.613900030961126e-05, + "loss": 0.4945, + "step": 20881 + }, + { + "epoch": 0.62, + "learning_rate": 6.612992766765351e-05, + "loss": 0.4806, + "step": 20882 + }, + { + "epoch": 0.62, + "learning_rate": 6.612085534060616e-05, + "loss": 0.5409, + "step": 20883 + }, + { + "epoch": 0.62, + "learning_rate": 6.611178332855361e-05, + "loss": 0.536, + "step": 20884 + }, + { + "epoch": 0.62, + "learning_rate": 6.610271163158017e-05, + "loss": 0.5578, + "step": 20885 + }, + { + "epoch": 0.62, + "learning_rate": 6.609364024977025e-05, + "loss": 0.5814, + "step": 20886 + }, + { + "epoch": 0.62, + "learning_rate": 6.608456918320812e-05, + "loss": 0.4911, + "step": 20887 + }, + { + "epoch": 0.62, + "learning_rate": 6.607549843197814e-05, + "loss": 0.5552, + "step": 20888 + }, + { + "epoch": 0.62, + "learning_rate": 6.606642799616467e-05, + "loss": 0.5674, + "step": 20889 + }, + { + "epoch": 0.62, + "learning_rate": 6.605735787585205e-05, + "loss": 0.6466, + "step": 20890 + }, + { + "epoch": 0.62, + "learning_rate": 6.604828807112456e-05, + "loss": 0.6632, + "step": 20891 + }, + { + "epoch": 0.62, + "learning_rate": 6.603921858206655e-05, + "loss": 0.5429, + "step": 20892 + }, + { + "epoch": 0.62, + "learning_rate": 6.603014940876236e-05, + "loss": 0.6089, + "step": 20893 + }, + { + "epoch": 0.62, + "learning_rate": 6.602108055129632e-05, + "loss": 0.4147, + "step": 20894 + }, + { + "epoch": 0.62, + "learning_rate": 6.60120120097527e-05, + "loss": 0.5667, + "step": 20895 + }, + { + "epoch": 0.62, + "learning_rate": 6.600294378421585e-05, + "loss": 0.5954, + "step": 20896 + }, + { + "epoch": 0.62, + "learning_rate": 6.599387587477007e-05, + "loss": 0.5536, + "step": 20897 + }, + { + "epoch": 0.62, + "learning_rate": 6.598480828149971e-05, + "loss": 0.5533, + "step": 20898 + }, + { + "epoch": 0.62, + "learning_rate": 6.597574100448904e-05, + "loss": 0.4482, + "step": 20899 + }, + { + "epoch": 0.62, + "learning_rate": 6.596667404382234e-05, + "loss": 0.5294, + "step": 20900 + }, + { + "epoch": 0.62, + "learning_rate": 6.5957607399584e-05, + "loss": 0.6241, + "step": 20901 + }, + { + "epoch": 0.62, + "learning_rate": 6.594854107185822e-05, + "loss": 0.515, + "step": 20902 + }, + { + "epoch": 0.62, + "learning_rate": 6.593947506072933e-05, + "loss": 0.5806, + "step": 20903 + }, + { + "epoch": 0.62, + "learning_rate": 6.593040936628161e-05, + "loss": 0.5296, + "step": 20904 + }, + { + "epoch": 0.62, + "learning_rate": 6.59213439885994e-05, + "loss": 0.5257, + "step": 20905 + }, + { + "epoch": 0.62, + "learning_rate": 6.591227892776694e-05, + "loss": 0.4549, + "step": 20906 + }, + { + "epoch": 0.62, + "learning_rate": 6.590321418386849e-05, + "loss": 0.5069, + "step": 20907 + }, + { + "epoch": 0.62, + "learning_rate": 6.58941497569884e-05, + "loss": 0.6069, + "step": 20908 + }, + { + "epoch": 0.62, + "learning_rate": 6.588508564721093e-05, + "loss": 0.652, + "step": 20909 + }, + { + "epoch": 0.62, + "learning_rate": 6.587602185462029e-05, + "loss": 0.6249, + "step": 20910 + }, + { + "epoch": 0.62, + "learning_rate": 6.586695837930084e-05, + "loss": 0.4931, + "step": 20911 + }, + { + "epoch": 0.62, + "learning_rate": 6.585789522133676e-05, + "loss": 0.4969, + "step": 20912 + }, + { + "epoch": 0.62, + "learning_rate": 6.584883238081243e-05, + "loss": 0.4765, + "step": 20913 + }, + { + "epoch": 0.62, + "learning_rate": 6.5839769857812e-05, + "loss": 0.5717, + "step": 20914 + }, + { + "epoch": 0.62, + "learning_rate": 6.583070765241977e-05, + "loss": 0.5092, + "step": 20915 + }, + { + "epoch": 0.62, + "learning_rate": 6.582164576472e-05, + "loss": 0.4904, + "step": 20916 + }, + { + "epoch": 0.62, + "learning_rate": 6.581258419479697e-05, + "loss": 0.5261, + "step": 20917 + }, + { + "epoch": 0.62, + "learning_rate": 6.580352294273491e-05, + "loss": 0.5973, + "step": 20918 + }, + { + "epoch": 0.62, + "learning_rate": 6.579446200861804e-05, + "loss": 0.4864, + "step": 20919 + }, + { + "epoch": 0.62, + "learning_rate": 6.578540139253062e-05, + "loss": 0.5246, + "step": 20920 + }, + { + "epoch": 0.62, + "learning_rate": 6.577634109455695e-05, + "loss": 0.5727, + "step": 20921 + }, + { + "epoch": 0.62, + "learning_rate": 6.576728111478119e-05, + "loss": 0.5921, + "step": 20922 + }, + { + "epoch": 0.62, + "learning_rate": 6.575822145328758e-05, + "loss": 0.5939, + "step": 20923 + }, + { + "epoch": 0.62, + "learning_rate": 6.574916211016042e-05, + "loss": 0.5256, + "step": 20924 + }, + { + "epoch": 0.62, + "learning_rate": 6.57401030854839e-05, + "loss": 0.5078, + "step": 20925 + }, + { + "epoch": 0.62, + "learning_rate": 6.573104437934221e-05, + "loss": 0.4748, + "step": 20926 + }, + { + "epoch": 0.62, + "learning_rate": 6.572198599181965e-05, + "loss": 0.6351, + "step": 20927 + }, + { + "epoch": 0.62, + "learning_rate": 6.571292792300037e-05, + "loss": 0.4773, + "step": 20928 + }, + { + "epoch": 0.62, + "learning_rate": 6.570387017296865e-05, + "loss": 0.627, + "step": 20929 + }, + { + "epoch": 0.62, + "learning_rate": 6.569481274180867e-05, + "loss": 0.5929, + "step": 20930 + }, + { + "epoch": 0.62, + "learning_rate": 6.568575562960462e-05, + "loss": 0.4988, + "step": 20931 + }, + { + "epoch": 0.62, + "learning_rate": 6.56766988364408e-05, + "loss": 0.5734, + "step": 20932 + }, + { + "epoch": 0.62, + "learning_rate": 6.566764236240129e-05, + "loss": 0.484, + "step": 20933 + }, + { + "epoch": 0.62, + "learning_rate": 6.565858620757039e-05, + "loss": 0.5212, + "step": 20934 + }, + { + "epoch": 0.62, + "learning_rate": 6.564953037203223e-05, + "loss": 0.4624, + "step": 20935 + }, + { + "epoch": 0.62, + "learning_rate": 6.56404748558711e-05, + "loss": 0.4928, + "step": 20936 + }, + { + "epoch": 0.62, + "learning_rate": 6.563141965917112e-05, + "loss": 0.5079, + "step": 20937 + }, + { + "epoch": 0.62, + "learning_rate": 6.562236478201649e-05, + "loss": 0.4724, + "step": 20938 + }, + { + "epoch": 0.62, + "learning_rate": 6.561331022449141e-05, + "loss": 0.4974, + "step": 20939 + }, + { + "epoch": 0.62, + "learning_rate": 6.56042559866801e-05, + "loss": 0.5876, + "step": 20940 + }, + { + "epoch": 0.62, + "learning_rate": 6.559520206866666e-05, + "loss": 0.5203, + "step": 20941 + }, + { + "epoch": 0.62, + "learning_rate": 6.558614847053535e-05, + "loss": 0.478, + "step": 20942 + }, + { + "epoch": 0.62, + "learning_rate": 6.557709519237029e-05, + "loss": 0.6809, + "step": 20943 + }, + { + "epoch": 0.62, + "learning_rate": 6.556804223425574e-05, + "loss": 0.5432, + "step": 20944 + }, + { + "epoch": 0.62, + "learning_rate": 6.555898959627575e-05, + "loss": 0.4484, + "step": 20945 + }, + { + "epoch": 0.62, + "learning_rate": 6.554993727851456e-05, + "loss": 0.6088, + "step": 20946 + }, + { + "epoch": 0.62, + "learning_rate": 6.55408852810563e-05, + "loss": 0.5767, + "step": 20947 + }, + { + "epoch": 0.62, + "learning_rate": 6.553183360398522e-05, + "loss": 0.5433, + "step": 20948 + }, + { + "epoch": 0.62, + "learning_rate": 6.552278224738537e-05, + "loss": 0.4592, + "step": 20949 + }, + { + "epoch": 0.62, + "learning_rate": 6.551373121134094e-05, + "loss": 0.6054, + "step": 20950 + }, + { + "epoch": 0.62, + "learning_rate": 6.55046804959361e-05, + "loss": 0.608, + "step": 20951 + }, + { + "epoch": 0.62, + "learning_rate": 6.549563010125504e-05, + "loss": 0.6035, + "step": 20952 + }, + { + "epoch": 0.62, + "learning_rate": 6.54865800273818e-05, + "loss": 0.4676, + "step": 20953 + }, + { + "epoch": 0.62, + "learning_rate": 6.54775302744006e-05, + "loss": 0.5271, + "step": 20954 + }, + { + "epoch": 0.62, + "learning_rate": 6.546848084239557e-05, + "loss": 0.4939, + "step": 20955 + }, + { + "epoch": 0.62, + "learning_rate": 6.545943173145085e-05, + "loss": 0.4562, + "step": 20956 + }, + { + "epoch": 0.62, + "learning_rate": 6.545038294165055e-05, + "loss": 0.635, + "step": 20957 + }, + { + "epoch": 0.62, + "learning_rate": 6.544133447307883e-05, + "loss": 0.5403, + "step": 20958 + }, + { + "epoch": 0.62, + "learning_rate": 6.543228632581978e-05, + "loss": 0.6212, + "step": 20959 + }, + { + "epoch": 0.62, + "learning_rate": 6.542323849995759e-05, + "loss": 0.5087, + "step": 20960 + }, + { + "epoch": 0.62, + "learning_rate": 6.541419099557635e-05, + "loss": 0.3954, + "step": 20961 + }, + { + "epoch": 0.62, + "learning_rate": 6.540514381276015e-05, + "loss": 0.5657, + "step": 20962 + }, + { + "epoch": 0.62, + "learning_rate": 6.539609695159314e-05, + "loss": 0.5236, + "step": 20963 + }, + { + "epoch": 0.62, + "learning_rate": 6.538705041215946e-05, + "loss": 0.6011, + "step": 20964 + }, + { + "epoch": 0.62, + "learning_rate": 6.537800419454316e-05, + "loss": 0.534, + "step": 20965 + }, + { + "epoch": 0.62, + "learning_rate": 6.536895829882838e-05, + "loss": 0.5188, + "step": 20966 + }, + { + "epoch": 0.62, + "learning_rate": 6.535991272509924e-05, + "loss": 0.5882, + "step": 20967 + }, + { + "epoch": 0.62, + "learning_rate": 6.535086747343982e-05, + "loss": 0.4732, + "step": 20968 + }, + { + "epoch": 0.62, + "learning_rate": 6.534182254393419e-05, + "loss": 0.5756, + "step": 20969 + }, + { + "epoch": 0.62, + "learning_rate": 6.533277793666652e-05, + "loss": 0.4707, + "step": 20970 + }, + { + "epoch": 0.62, + "learning_rate": 6.532373365172087e-05, + "loss": 0.4981, + "step": 20971 + }, + { + "epoch": 0.62, + "learning_rate": 6.531468968918129e-05, + "loss": 0.524, + "step": 20972 + }, + { + "epoch": 0.62, + "learning_rate": 6.530564604913192e-05, + "loss": 0.6009, + "step": 20973 + }, + { + "epoch": 0.62, + "learning_rate": 6.529660273165681e-05, + "loss": 0.5306, + "step": 20974 + }, + { + "epoch": 0.62, + "learning_rate": 6.52875597368401e-05, + "loss": 0.5268, + "step": 20975 + }, + { + "epoch": 0.62, + "learning_rate": 6.527851706476578e-05, + "loss": 0.5931, + "step": 20976 + }, + { + "epoch": 0.62, + "learning_rate": 6.526947471551798e-05, + "loss": 0.4993, + "step": 20977 + }, + { + "epoch": 0.62, + "learning_rate": 6.526043268918074e-05, + "loss": 0.5057, + "step": 20978 + }, + { + "epoch": 0.62, + "learning_rate": 6.525139098583822e-05, + "loss": 0.4442, + "step": 20979 + }, + { + "epoch": 0.62, + "learning_rate": 6.524234960557436e-05, + "loss": 0.5333, + "step": 20980 + }, + { + "epoch": 0.62, + "learning_rate": 6.523330854847329e-05, + "loss": 0.494, + "step": 20981 + }, + { + "epoch": 0.62, + "learning_rate": 6.522426781461906e-05, + "loss": 0.546, + "step": 20982 + }, + { + "epoch": 0.62, + "learning_rate": 6.521522740409576e-05, + "loss": 0.6127, + "step": 20983 + }, + { + "epoch": 0.62, + "learning_rate": 6.520618731698736e-05, + "loss": 0.4851, + "step": 20984 + }, + { + "epoch": 0.62, + "learning_rate": 6.519714755337798e-05, + "loss": 0.6789, + "step": 20985 + }, + { + "epoch": 0.62, + "learning_rate": 6.518810811335166e-05, + "loss": 0.5286, + "step": 20986 + }, + { + "epoch": 0.62, + "learning_rate": 6.517906899699247e-05, + "loss": 0.5585, + "step": 20987 + }, + { + "epoch": 0.62, + "learning_rate": 6.517003020438435e-05, + "loss": 0.6038, + "step": 20988 + }, + { + "epoch": 0.62, + "learning_rate": 6.516099173561143e-05, + "loss": 0.5768, + "step": 20989 + }, + { + "epoch": 0.62, + "learning_rate": 6.515195359075772e-05, + "loss": 0.5248, + "step": 20990 + }, + { + "epoch": 0.62, + "learning_rate": 6.51429157699073e-05, + "loss": 0.5342, + "step": 20991 + }, + { + "epoch": 0.62, + "learning_rate": 6.513387827314413e-05, + "loss": 0.6091, + "step": 20992 + }, + { + "epoch": 0.62, + "learning_rate": 6.512484110055224e-05, + "loss": 0.5949, + "step": 20993 + }, + { + "epoch": 0.63, + "learning_rate": 6.511580425221569e-05, + "loss": 0.6263, + "step": 20994 + }, + { + "epoch": 0.63, + "learning_rate": 6.51067677282185e-05, + "loss": 0.5577, + "step": 20995 + }, + { + "epoch": 0.63, + "learning_rate": 6.509773152864468e-05, + "loss": 0.4299, + "step": 20996 + }, + { + "epoch": 0.63, + "learning_rate": 6.508869565357822e-05, + "loss": 0.4649, + "step": 20997 + }, + { + "epoch": 0.63, + "learning_rate": 6.507966010310319e-05, + "loss": 0.472, + "step": 20998 + }, + { + "epoch": 0.63, + "learning_rate": 6.507062487730355e-05, + "loss": 0.5794, + "step": 20999 + }, + { + "epoch": 0.63, + "learning_rate": 6.506158997626328e-05, + "loss": 0.5963, + "step": 21000 + }, + { + "epoch": 0.63, + "learning_rate": 6.505255540006645e-05, + "loss": 0.5338, + "step": 21001 + }, + { + "epoch": 0.63, + "learning_rate": 6.504352114879706e-05, + "loss": 0.5569, + "step": 21002 + }, + { + "epoch": 0.63, + "learning_rate": 6.503448722253902e-05, + "loss": 0.5077, + "step": 21003 + }, + { + "epoch": 0.63, + "learning_rate": 6.502545362137642e-05, + "loss": 0.5178, + "step": 21004 + }, + { + "epoch": 0.63, + "learning_rate": 6.501642034539318e-05, + "loss": 0.5793, + "step": 21005 + }, + { + "epoch": 0.63, + "learning_rate": 6.500738739467338e-05, + "loss": 0.552, + "step": 21006 + }, + { + "epoch": 0.63, + "learning_rate": 6.49983547693009e-05, + "loss": 0.5634, + "step": 21007 + }, + { + "epoch": 0.63, + "learning_rate": 6.498932246935978e-05, + "loss": 0.5083, + "step": 21008 + }, + { + "epoch": 0.63, + "learning_rate": 6.498029049493397e-05, + "loss": 0.5661, + "step": 21009 + }, + { + "epoch": 0.63, + "learning_rate": 6.497125884610748e-05, + "loss": 0.4305, + "step": 21010 + }, + { + "epoch": 0.63, + "learning_rate": 6.496222752296426e-05, + "loss": 0.566, + "step": 21011 + }, + { + "epoch": 0.63, + "learning_rate": 6.495319652558827e-05, + "loss": 0.5329, + "step": 21012 + }, + { + "epoch": 0.63, + "learning_rate": 6.49441658540635e-05, + "loss": 0.4965, + "step": 21013 + }, + { + "epoch": 0.63, + "learning_rate": 6.493513550847392e-05, + "loss": 0.531, + "step": 21014 + }, + { + "epoch": 0.63, + "learning_rate": 6.492610548890344e-05, + "loss": 0.4692, + "step": 21015 + }, + { + "epoch": 0.63, + "learning_rate": 6.491707579543608e-05, + "loss": 0.5769, + "step": 21016 + }, + { + "epoch": 0.63, + "learning_rate": 6.490804642815575e-05, + "loss": 0.4829, + "step": 21017 + }, + { + "epoch": 0.63, + "learning_rate": 6.489901738714645e-05, + "loss": 0.5207, + "step": 21018 + }, + { + "epoch": 0.63, + "learning_rate": 6.488998867249207e-05, + "loss": 0.5272, + "step": 21019 + }, + { + "epoch": 0.63, + "learning_rate": 6.488096028427658e-05, + "loss": 0.4839, + "step": 21020 + }, + { + "epoch": 0.63, + "learning_rate": 6.487193222258391e-05, + "loss": 0.5405, + "step": 21021 + }, + { + "epoch": 0.63, + "learning_rate": 6.486290448749806e-05, + "loss": 0.5279, + "step": 21022 + }, + { + "epoch": 0.63, + "learning_rate": 6.485387707910289e-05, + "loss": 0.668, + "step": 21023 + }, + { + "epoch": 0.63, + "learning_rate": 6.484484999748234e-05, + "loss": 0.4426, + "step": 21024 + }, + { + "epoch": 0.63, + "learning_rate": 6.48358232427204e-05, + "loss": 0.5281, + "step": 21025 + }, + { + "epoch": 0.63, + "learning_rate": 6.482679681490097e-05, + "loss": 0.4984, + "step": 21026 + }, + { + "epoch": 0.63, + "learning_rate": 6.481777071410794e-05, + "loss": 0.5344, + "step": 21027 + }, + { + "epoch": 0.63, + "learning_rate": 6.480874494042524e-05, + "loss": 0.5656, + "step": 21028 + }, + { + "epoch": 0.63, + "learning_rate": 6.479971949393684e-05, + "loss": 0.564, + "step": 21029 + }, + { + "epoch": 0.63, + "learning_rate": 6.479069437472663e-05, + "loss": 0.4983, + "step": 21030 + }, + { + "epoch": 0.63, + "learning_rate": 6.478166958287846e-05, + "loss": 0.5047, + "step": 21031 + }, + { + "epoch": 0.63, + "learning_rate": 6.477264511847631e-05, + "loss": 0.5448, + "step": 21032 + }, + { + "epoch": 0.63, + "learning_rate": 6.476362098160409e-05, + "loss": 0.4836, + "step": 21033 + }, + { + "epoch": 0.63, + "learning_rate": 6.475459717234563e-05, + "loss": 0.4875, + "step": 21034 + }, + { + "epoch": 0.63, + "learning_rate": 6.47455736907849e-05, + "loss": 0.5189, + "step": 21035 + }, + { + "epoch": 0.63, + "learning_rate": 6.473655053700577e-05, + "loss": 0.5167, + "step": 21036 + }, + { + "epoch": 0.63, + "learning_rate": 6.472752771109218e-05, + "loss": 0.5135, + "step": 21037 + }, + { + "epoch": 0.63, + "learning_rate": 6.471850521312794e-05, + "loss": 0.5541, + "step": 21038 + }, + { + "epoch": 0.63, + "learning_rate": 6.470948304319697e-05, + "loss": 0.5442, + "step": 21039 + }, + { + "epoch": 0.63, + "learning_rate": 6.470046120138317e-05, + "loss": 0.5647, + "step": 21040 + }, + { + "epoch": 0.63, + "learning_rate": 6.469143968777043e-05, + "loss": 0.4635, + "step": 21041 + }, + { + "epoch": 0.63, + "learning_rate": 6.468241850244261e-05, + "loss": 0.5293, + "step": 21042 + }, + { + "epoch": 0.63, + "learning_rate": 6.467339764548355e-05, + "loss": 0.5346, + "step": 21043 + }, + { + "epoch": 0.63, + "learning_rate": 6.466437711697719e-05, + "loss": 0.6235, + "step": 21044 + }, + { + "epoch": 0.63, + "learning_rate": 6.465535691700738e-05, + "loss": 0.4881, + "step": 21045 + }, + { + "epoch": 0.63, + "learning_rate": 6.464633704565795e-05, + "loss": 0.5316, + "step": 21046 + }, + { + "epoch": 0.63, + "learning_rate": 6.463731750301278e-05, + "loss": 0.4796, + "step": 21047 + }, + { + "epoch": 0.63, + "learning_rate": 6.462829828915575e-05, + "loss": 0.4857, + "step": 21048 + }, + { + "epoch": 0.63, + "learning_rate": 6.461927940417074e-05, + "loss": 0.5111, + "step": 21049 + }, + { + "epoch": 0.63, + "learning_rate": 6.461026084814151e-05, + "loss": 0.4978, + "step": 21050 + }, + { + "epoch": 0.63, + "learning_rate": 6.4601242621152e-05, + "loss": 0.6019, + "step": 21051 + }, + { + "epoch": 0.63, + "learning_rate": 6.4592224723286e-05, + "loss": 0.4691, + "step": 21052 + }, + { + "epoch": 0.63, + "learning_rate": 6.458320715462742e-05, + "loss": 0.6568, + "step": 21053 + }, + { + "epoch": 0.63, + "learning_rate": 6.457418991526005e-05, + "loss": 0.3909, + "step": 21054 + }, + { + "epoch": 0.63, + "learning_rate": 6.456517300526773e-05, + "loss": 0.5502, + "step": 21055 + }, + { + "epoch": 0.63, + "learning_rate": 6.455615642473433e-05, + "loss": 0.6, + "step": 21056 + }, + { + "epoch": 0.63, + "learning_rate": 6.454714017374367e-05, + "loss": 0.513, + "step": 21057 + }, + { + "epoch": 0.63, + "learning_rate": 6.453812425237956e-05, + "loss": 0.5061, + "step": 21058 + }, + { + "epoch": 0.63, + "learning_rate": 6.45291086607258e-05, + "loss": 0.6328, + "step": 21059 + }, + { + "epoch": 0.63, + "learning_rate": 6.452009339886629e-05, + "loss": 0.6163, + "step": 21060 + }, + { + "epoch": 0.63, + "learning_rate": 6.451107846688481e-05, + "loss": 0.5745, + "step": 21061 + }, + { + "epoch": 0.63, + "learning_rate": 6.450206386486515e-05, + "loss": 0.5234, + "step": 21062 + }, + { + "epoch": 0.63, + "learning_rate": 6.449304959289118e-05, + "loss": 0.4269, + "step": 21063 + }, + { + "epoch": 0.63, + "learning_rate": 6.448403565104668e-05, + "loss": 0.4972, + "step": 21064 + }, + { + "epoch": 0.63, + "learning_rate": 6.447502203941544e-05, + "loss": 0.4829, + "step": 21065 + }, + { + "epoch": 0.63, + "learning_rate": 6.446600875808131e-05, + "loss": 0.5075, + "step": 21066 + }, + { + "epoch": 0.63, + "learning_rate": 6.445699580712804e-05, + "loss": 0.5891, + "step": 21067 + }, + { + "epoch": 0.63, + "learning_rate": 6.444798318663951e-05, + "loss": 0.5729, + "step": 21068 + }, + { + "epoch": 0.63, + "learning_rate": 6.443897089669942e-05, + "loss": 0.6221, + "step": 21069 + }, + { + "epoch": 0.63, + "learning_rate": 6.442995893739161e-05, + "loss": 0.473, + "step": 21070 + }, + { + "epoch": 0.63, + "learning_rate": 6.442094730879985e-05, + "loss": 0.4189, + "step": 21071 + }, + { + "epoch": 0.63, + "learning_rate": 6.441193601100799e-05, + "loss": 0.6076, + "step": 21072 + }, + { + "epoch": 0.63, + "learning_rate": 6.440292504409972e-05, + "loss": 0.4468, + "step": 21073 + }, + { + "epoch": 0.63, + "learning_rate": 6.439391440815887e-05, + "loss": 0.5912, + "step": 21074 + }, + { + "epoch": 0.63, + "learning_rate": 6.438490410326923e-05, + "loss": 0.5267, + "step": 21075 + }, + { + "epoch": 0.63, + "learning_rate": 6.437589412951457e-05, + "loss": 0.4606, + "step": 21076 + }, + { + "epoch": 0.63, + "learning_rate": 6.436688448697861e-05, + "loss": 0.5676, + "step": 21077 + }, + { + "epoch": 0.63, + "learning_rate": 6.435787517574517e-05, + "loss": 0.4887, + "step": 21078 + }, + { + "epoch": 0.63, + "learning_rate": 6.4348866195898e-05, + "loss": 0.4601, + "step": 21079 + }, + { + "epoch": 0.63, + "learning_rate": 6.43398575475209e-05, + "loss": 0.462, + "step": 21080 + }, + { + "epoch": 0.63, + "learning_rate": 6.433084923069754e-05, + "loss": 0.5741, + "step": 21081 + }, + { + "epoch": 0.63, + "learning_rate": 6.432184124551175e-05, + "loss": 0.5713, + "step": 21082 + }, + { + "epoch": 0.63, + "learning_rate": 6.431283359204724e-05, + "loss": 0.4839, + "step": 21083 + }, + { + "epoch": 0.63, + "learning_rate": 6.430382627038783e-05, + "loss": 0.5369, + "step": 21084 + }, + { + "epoch": 0.63, + "learning_rate": 6.429481928061718e-05, + "loss": 0.5848, + "step": 21085 + }, + { + "epoch": 0.63, + "learning_rate": 6.428581262281905e-05, + "loss": 0.4687, + "step": 21086 + }, + { + "epoch": 0.63, + "learning_rate": 6.427680629707722e-05, + "loss": 0.4615, + "step": 21087 + }, + { + "epoch": 0.63, + "learning_rate": 6.426780030347544e-05, + "loss": 0.5104, + "step": 21088 + }, + { + "epoch": 0.63, + "learning_rate": 6.425879464209737e-05, + "loss": 0.6482, + "step": 21089 + }, + { + "epoch": 0.63, + "learning_rate": 6.424978931302679e-05, + "loss": 0.5473, + "step": 21090 + }, + { + "epoch": 0.63, + "learning_rate": 6.424078431634742e-05, + "loss": 0.4692, + "step": 21091 + }, + { + "epoch": 0.63, + "learning_rate": 6.423177965214302e-05, + "loss": 0.6424, + "step": 21092 + }, + { + "epoch": 0.63, + "learning_rate": 6.422277532049723e-05, + "loss": 0.6617, + "step": 21093 + }, + { + "epoch": 0.63, + "learning_rate": 6.421377132149382e-05, + "loss": 0.5289, + "step": 21094 + }, + { + "epoch": 0.63, + "learning_rate": 6.42047676552165e-05, + "loss": 0.5619, + "step": 21095 + }, + { + "epoch": 0.63, + "learning_rate": 6.419576432174902e-05, + "loss": 0.4557, + "step": 21096 + }, + { + "epoch": 0.63, + "learning_rate": 6.418676132117503e-05, + "loss": 0.5614, + "step": 21097 + }, + { + "epoch": 0.63, + "learning_rate": 6.417775865357825e-05, + "loss": 0.6015, + "step": 21098 + }, + { + "epoch": 0.63, + "learning_rate": 6.416875631904244e-05, + "loss": 0.5277, + "step": 21099 + }, + { + "epoch": 0.63, + "learning_rate": 6.41597543176512e-05, + "loss": 0.4302, + "step": 21100 + }, + { + "epoch": 0.63, + "learning_rate": 6.41507526494883e-05, + "loss": 0.556, + "step": 21101 + }, + { + "epoch": 0.63, + "learning_rate": 6.414175131463738e-05, + "loss": 0.552, + "step": 21102 + }, + { + "epoch": 0.63, + "learning_rate": 6.413275031318222e-05, + "loss": 0.5609, + "step": 21103 + }, + { + "epoch": 0.63, + "learning_rate": 6.412374964520642e-05, + "loss": 0.4647, + "step": 21104 + }, + { + "epoch": 0.63, + "learning_rate": 6.411474931079369e-05, + "loss": 0.5395, + "step": 21105 + }, + { + "epoch": 0.63, + "learning_rate": 6.410574931002773e-05, + "loss": 0.5832, + "step": 21106 + }, + { + "epoch": 0.63, + "learning_rate": 6.409674964299224e-05, + "loss": 0.5674, + "step": 21107 + }, + { + "epoch": 0.63, + "learning_rate": 6.408775030977081e-05, + "loss": 0.4315, + "step": 21108 + }, + { + "epoch": 0.63, + "learning_rate": 6.407875131044718e-05, + "loss": 0.5278, + "step": 21109 + }, + { + "epoch": 0.63, + "learning_rate": 6.4069752645105e-05, + "loss": 0.5415, + "step": 21110 + }, + { + "epoch": 0.63, + "learning_rate": 6.406075431382799e-05, + "loss": 0.6446, + "step": 21111 + }, + { + "epoch": 0.63, + "learning_rate": 6.40517563166997e-05, + "loss": 0.5153, + "step": 21112 + }, + { + "epoch": 0.63, + "learning_rate": 6.404275865380388e-05, + "loss": 0.5907, + "step": 21113 + }, + { + "epoch": 0.63, + "learning_rate": 6.403376132522411e-05, + "loss": 0.517, + "step": 21114 + }, + { + "epoch": 0.63, + "learning_rate": 6.402476433104416e-05, + "loss": 0.6025, + "step": 21115 + }, + { + "epoch": 0.63, + "learning_rate": 6.401576767134761e-05, + "loss": 0.5592, + "step": 21116 + }, + { + "epoch": 0.63, + "learning_rate": 6.400677134621806e-05, + "loss": 0.6778, + "step": 21117 + }, + { + "epoch": 0.63, + "learning_rate": 6.399777535573923e-05, + "loss": 0.5515, + "step": 21118 + }, + { + "epoch": 0.63, + "learning_rate": 6.398877969999475e-05, + "loss": 0.4862, + "step": 21119 + }, + { + "epoch": 0.63, + "learning_rate": 6.397978437906821e-05, + "loss": 0.534, + "step": 21120 + }, + { + "epoch": 0.63, + "learning_rate": 6.397078939304331e-05, + "loss": 0.5074, + "step": 21121 + }, + { + "epoch": 0.63, + "learning_rate": 6.396179474200365e-05, + "loss": 0.5627, + "step": 21122 + }, + { + "epoch": 0.63, + "learning_rate": 6.395280042603287e-05, + "loss": 0.4901, + "step": 21123 + }, + { + "epoch": 0.63, + "learning_rate": 6.394380644521454e-05, + "loss": 0.5631, + "step": 21124 + }, + { + "epoch": 0.63, + "learning_rate": 6.393481279963236e-05, + "loss": 0.5216, + "step": 21125 + }, + { + "epoch": 0.63, + "learning_rate": 6.392581948936989e-05, + "loss": 0.529, + "step": 21126 + }, + { + "epoch": 0.63, + "learning_rate": 6.39168265145108e-05, + "loss": 0.5554, + "step": 21127 + }, + { + "epoch": 0.63, + "learning_rate": 6.390783387513868e-05, + "loss": 0.5493, + "step": 21128 + }, + { + "epoch": 0.63, + "learning_rate": 6.38988415713371e-05, + "loss": 0.589, + "step": 21129 + }, + { + "epoch": 0.63, + "learning_rate": 6.388984960318973e-05, + "loss": 0.5965, + "step": 21130 + }, + { + "epoch": 0.63, + "learning_rate": 6.388085797078015e-05, + "loss": 0.6146, + "step": 21131 + }, + { + "epoch": 0.63, + "learning_rate": 6.387186667419195e-05, + "loss": 0.631, + "step": 21132 + }, + { + "epoch": 0.63, + "learning_rate": 6.386287571350871e-05, + "loss": 0.5687, + "step": 21133 + }, + { + "epoch": 0.63, + "learning_rate": 6.385388508881409e-05, + "loss": 0.5221, + "step": 21134 + }, + { + "epoch": 0.63, + "learning_rate": 6.384489480019162e-05, + "loss": 0.5404, + "step": 21135 + }, + { + "epoch": 0.63, + "learning_rate": 6.38359048477249e-05, + "loss": 0.521, + "step": 21136 + }, + { + "epoch": 0.63, + "learning_rate": 6.382691523149753e-05, + "loss": 0.6567, + "step": 21137 + }, + { + "epoch": 0.63, + "learning_rate": 6.38179259515931e-05, + "loss": 0.4543, + "step": 21138 + }, + { + "epoch": 0.63, + "learning_rate": 6.380893700809516e-05, + "loss": 0.6613, + "step": 21139 + }, + { + "epoch": 0.63, + "learning_rate": 6.37999484010873e-05, + "loss": 0.5463, + "step": 21140 + }, + { + "epoch": 0.63, + "learning_rate": 6.379096013065307e-05, + "loss": 0.582, + "step": 21141 + }, + { + "epoch": 0.63, + "learning_rate": 6.378197219687612e-05, + "loss": 0.5592, + "step": 21142 + }, + { + "epoch": 0.63, + "learning_rate": 6.37729845998399e-05, + "loss": 0.6636, + "step": 21143 + }, + { + "epoch": 0.63, + "learning_rate": 6.376399733962806e-05, + "loss": 0.525, + "step": 21144 + }, + { + "epoch": 0.63, + "learning_rate": 6.37550104163241e-05, + "loss": 0.5089, + "step": 21145 + }, + { + "epoch": 0.63, + "learning_rate": 6.374602383001165e-05, + "loss": 0.5527, + "step": 21146 + }, + { + "epoch": 0.63, + "learning_rate": 6.37370375807742e-05, + "loss": 0.6115, + "step": 21147 + }, + { + "epoch": 0.63, + "learning_rate": 6.37280516686953e-05, + "loss": 0.5815, + "step": 21148 + }, + { + "epoch": 0.63, + "learning_rate": 6.371906609385855e-05, + "loss": 0.5158, + "step": 21149 + }, + { + "epoch": 0.63, + "learning_rate": 6.371008085634747e-05, + "loss": 0.5112, + "step": 21150 + }, + { + "epoch": 0.63, + "learning_rate": 6.370109595624556e-05, + "loss": 0.5175, + "step": 21151 + }, + { + "epoch": 0.63, + "learning_rate": 6.369211139363641e-05, + "loss": 0.5025, + "step": 21152 + }, + { + "epoch": 0.63, + "learning_rate": 6.368312716860354e-05, + "loss": 0.5437, + "step": 21153 + }, + { + "epoch": 0.63, + "learning_rate": 6.367414328123051e-05, + "loss": 0.5378, + "step": 21154 + }, + { + "epoch": 0.63, + "learning_rate": 6.366515973160077e-05, + "loss": 0.5623, + "step": 21155 + }, + { + "epoch": 0.63, + "learning_rate": 6.36561765197979e-05, + "loss": 0.6363, + "step": 21156 + }, + { + "epoch": 0.63, + "learning_rate": 6.36471936459054e-05, + "loss": 0.605, + "step": 21157 + }, + { + "epoch": 0.63, + "learning_rate": 6.363821111000685e-05, + "loss": 0.5961, + "step": 21158 + }, + { + "epoch": 0.63, + "learning_rate": 6.36292289121857e-05, + "loss": 0.6003, + "step": 21159 + }, + { + "epoch": 0.63, + "learning_rate": 6.362024705252547e-05, + "loss": 0.4353, + "step": 21160 + }, + { + "epoch": 0.63, + "learning_rate": 6.36112655311097e-05, + "loss": 0.521, + "step": 21161 + }, + { + "epoch": 0.63, + "learning_rate": 6.360228434802189e-05, + "loss": 0.7283, + "step": 21162 + }, + { + "epoch": 0.63, + "learning_rate": 6.359330350334553e-05, + "loss": 0.4618, + "step": 21163 + }, + { + "epoch": 0.63, + "learning_rate": 6.358432299716408e-05, + "loss": 0.5391, + "step": 21164 + }, + { + "epoch": 0.63, + "learning_rate": 6.357534282956114e-05, + "loss": 0.5754, + "step": 21165 + }, + { + "epoch": 0.63, + "learning_rate": 6.356636300062013e-05, + "loss": 0.5028, + "step": 21166 + }, + { + "epoch": 0.63, + "learning_rate": 6.355738351042452e-05, + "loss": 0.6351, + "step": 21167 + }, + { + "epoch": 0.63, + "learning_rate": 6.354840435905787e-05, + "loss": 0.6311, + "step": 21168 + }, + { + "epoch": 0.63, + "learning_rate": 6.353942554660364e-05, + "loss": 0.604, + "step": 21169 + }, + { + "epoch": 0.63, + "learning_rate": 6.353044707314524e-05, + "loss": 0.4714, + "step": 21170 + }, + { + "epoch": 0.63, + "learning_rate": 6.352146893876625e-05, + "loss": 0.4463, + "step": 21171 + }, + { + "epoch": 0.63, + "learning_rate": 6.351249114355008e-05, + "loss": 0.5905, + "step": 21172 + }, + { + "epoch": 0.63, + "learning_rate": 6.350351368758028e-05, + "loss": 0.5687, + "step": 21173 + }, + { + "epoch": 0.63, + "learning_rate": 6.349453657094021e-05, + "loss": 0.5725, + "step": 21174 + }, + { + "epoch": 0.63, + "learning_rate": 6.34855597937134e-05, + "loss": 0.565, + "step": 21175 + }, + { + "epoch": 0.63, + "learning_rate": 6.34765833559833e-05, + "loss": 0.557, + "step": 21176 + }, + { + "epoch": 0.63, + "learning_rate": 6.346760725783338e-05, + "loss": 0.5389, + "step": 21177 + }, + { + "epoch": 0.63, + "learning_rate": 6.345863149934709e-05, + "loss": 0.5379, + "step": 21178 + }, + { + "epoch": 0.63, + "learning_rate": 6.344965608060787e-05, + "loss": 0.5492, + "step": 21179 + }, + { + "epoch": 0.63, + "learning_rate": 6.344068100169918e-05, + "loss": 0.6208, + "step": 21180 + }, + { + "epoch": 0.63, + "learning_rate": 6.34317062627045e-05, + "loss": 0.6149, + "step": 21181 + }, + { + "epoch": 0.63, + "learning_rate": 6.34227318637072e-05, + "loss": 0.6413, + "step": 21182 + }, + { + "epoch": 0.63, + "learning_rate": 6.341375780479076e-05, + "loss": 0.5503, + "step": 21183 + }, + { + "epoch": 0.63, + "learning_rate": 6.340478408603864e-05, + "loss": 0.4752, + "step": 21184 + }, + { + "epoch": 0.63, + "learning_rate": 6.339581070753427e-05, + "loss": 0.4272, + "step": 21185 + }, + { + "epoch": 0.63, + "learning_rate": 6.3386837669361e-05, + "loss": 0.5729, + "step": 21186 + }, + { + "epoch": 0.63, + "learning_rate": 6.337786497160236e-05, + "loss": 0.6437, + "step": 21187 + }, + { + "epoch": 0.63, + "learning_rate": 6.336889261434173e-05, + "loss": 0.5019, + "step": 21188 + }, + { + "epoch": 0.63, + "learning_rate": 6.335992059766256e-05, + "loss": 0.5192, + "step": 21189 + }, + { + "epoch": 0.63, + "learning_rate": 6.335094892164822e-05, + "loss": 0.5339, + "step": 21190 + }, + { + "epoch": 0.63, + "learning_rate": 6.334197758638214e-05, + "loss": 0.4873, + "step": 21191 + }, + { + "epoch": 0.63, + "learning_rate": 6.333300659194776e-05, + "loss": 0.6172, + "step": 21192 + }, + { + "epoch": 0.63, + "learning_rate": 6.332403593842848e-05, + "loss": 0.5086, + "step": 21193 + }, + { + "epoch": 0.63, + "learning_rate": 6.33150656259077e-05, + "loss": 0.4967, + "step": 21194 + }, + { + "epoch": 0.63, + "learning_rate": 6.330609565446878e-05, + "loss": 0.488, + "step": 21195 + }, + { + "epoch": 0.63, + "learning_rate": 6.32971260241952e-05, + "loss": 0.5519, + "step": 21196 + }, + { + "epoch": 0.63, + "learning_rate": 6.32881567351703e-05, + "loss": 0.5851, + "step": 21197 + }, + { + "epoch": 0.63, + "learning_rate": 6.327918778747745e-05, + "loss": 0.5176, + "step": 21198 + }, + { + "epoch": 0.63, + "learning_rate": 6.327021918120011e-05, + "loss": 0.5212, + "step": 21199 + }, + { + "epoch": 0.63, + "learning_rate": 6.326125091642165e-05, + "loss": 0.5792, + "step": 21200 + }, + { + "epoch": 0.63, + "learning_rate": 6.325228299322541e-05, + "loss": 0.4811, + "step": 21201 + }, + { + "epoch": 0.63, + "learning_rate": 6.32433154116948e-05, + "loss": 0.4472, + "step": 21202 + }, + { + "epoch": 0.63, + "learning_rate": 6.323434817191319e-05, + "loss": 0.5427, + "step": 21203 + }, + { + "epoch": 0.63, + "learning_rate": 6.3225381273964e-05, + "loss": 0.5774, + "step": 21204 + }, + { + "epoch": 0.63, + "learning_rate": 6.321641471793049e-05, + "loss": 0.6247, + "step": 21205 + }, + { + "epoch": 0.63, + "learning_rate": 6.320744850389614e-05, + "loss": 0.5702, + "step": 21206 + }, + { + "epoch": 0.63, + "learning_rate": 6.319848263194423e-05, + "loss": 0.5279, + "step": 21207 + }, + { + "epoch": 0.63, + "learning_rate": 6.318951710215822e-05, + "loss": 0.4647, + "step": 21208 + }, + { + "epoch": 0.63, + "learning_rate": 6.318055191462137e-05, + "loss": 0.5808, + "step": 21209 + }, + { + "epoch": 0.63, + "learning_rate": 6.317158706941705e-05, + "loss": 0.645, + "step": 21210 + }, + { + "epoch": 0.63, + "learning_rate": 6.316262256662868e-05, + "loss": 0.5112, + "step": 21211 + }, + { + "epoch": 0.63, + "learning_rate": 6.315365840633958e-05, + "loss": 0.5896, + "step": 21212 + }, + { + "epoch": 0.63, + "learning_rate": 6.314469458863302e-05, + "loss": 0.4802, + "step": 21213 + }, + { + "epoch": 0.63, + "learning_rate": 6.313573111359243e-05, + "loss": 0.5679, + "step": 21214 + }, + { + "epoch": 0.63, + "learning_rate": 6.31267679813011e-05, + "loss": 0.6116, + "step": 21215 + }, + { + "epoch": 0.63, + "learning_rate": 6.311780519184245e-05, + "loss": 0.5713, + "step": 21216 + }, + { + "epoch": 0.63, + "learning_rate": 6.310884274529969e-05, + "loss": 0.5928, + "step": 21217 + }, + { + "epoch": 0.63, + "learning_rate": 6.30998806417562e-05, + "loss": 0.5931, + "step": 21218 + }, + { + "epoch": 0.63, + "learning_rate": 6.309091888129532e-05, + "loss": 0.5458, + "step": 21219 + }, + { + "epoch": 0.63, + "learning_rate": 6.308195746400039e-05, + "loss": 0.5252, + "step": 21220 + }, + { + "epoch": 0.63, + "learning_rate": 6.30729963899547e-05, + "loss": 0.5117, + "step": 21221 + }, + { + "epoch": 0.63, + "learning_rate": 6.306403565924156e-05, + "loss": 0.4545, + "step": 21222 + }, + { + "epoch": 0.63, + "learning_rate": 6.305507527194429e-05, + "loss": 0.7135, + "step": 21223 + }, + { + "epoch": 0.63, + "learning_rate": 6.304611522814623e-05, + "loss": 0.5367, + "step": 21224 + }, + { + "epoch": 0.63, + "learning_rate": 6.303715552793065e-05, + "loss": 0.5375, + "step": 21225 + }, + { + "epoch": 0.63, + "learning_rate": 6.302819617138085e-05, + "loss": 0.6305, + "step": 21226 + }, + { + "epoch": 0.63, + "learning_rate": 6.301923715858016e-05, + "loss": 0.4288, + "step": 21227 + }, + { + "epoch": 0.63, + "learning_rate": 6.301027848961187e-05, + "loss": 0.4856, + "step": 21228 + }, + { + "epoch": 0.63, + "learning_rate": 6.300132016455924e-05, + "loss": 0.4491, + "step": 21229 + }, + { + "epoch": 0.63, + "learning_rate": 6.299236218350562e-05, + "loss": 0.5999, + "step": 21230 + }, + { + "epoch": 0.63, + "learning_rate": 6.298340454653427e-05, + "loss": 0.6176, + "step": 21231 + }, + { + "epoch": 0.63, + "learning_rate": 6.297444725372845e-05, + "loss": 0.6716, + "step": 21232 + }, + { + "epoch": 0.63, + "learning_rate": 6.296549030517147e-05, + "loss": 0.5243, + "step": 21233 + }, + { + "epoch": 0.63, + "learning_rate": 6.295653370094658e-05, + "loss": 0.6443, + "step": 21234 + }, + { + "epoch": 0.63, + "learning_rate": 6.294757744113713e-05, + "loss": 0.5125, + "step": 21235 + }, + { + "epoch": 0.63, + "learning_rate": 6.293862152582629e-05, + "loss": 0.4802, + "step": 21236 + }, + { + "epoch": 0.63, + "learning_rate": 6.292966595509739e-05, + "loss": 0.5449, + "step": 21237 + }, + { + "epoch": 0.63, + "learning_rate": 6.292071072903366e-05, + "loss": 0.4042, + "step": 21238 + }, + { + "epoch": 0.63, + "learning_rate": 6.291175584771842e-05, + "loss": 0.4734, + "step": 21239 + }, + { + "epoch": 0.63, + "learning_rate": 6.290280131123488e-05, + "loss": 0.396, + "step": 21240 + }, + { + "epoch": 0.63, + "learning_rate": 6.289384711966628e-05, + "loss": 0.4777, + "step": 21241 + }, + { + "epoch": 0.63, + "learning_rate": 6.288489327309593e-05, + "loss": 0.4956, + "step": 21242 + }, + { + "epoch": 0.63, + "learning_rate": 6.287593977160706e-05, + "loss": 0.6558, + "step": 21243 + }, + { + "epoch": 0.63, + "learning_rate": 6.286698661528287e-05, + "loss": 0.6336, + "step": 21244 + }, + { + "epoch": 0.63, + "learning_rate": 6.285803380420667e-05, + "loss": 0.4379, + "step": 21245 + }, + { + "epoch": 0.63, + "learning_rate": 6.284908133846164e-05, + "loss": 0.5847, + "step": 21246 + }, + { + "epoch": 0.63, + "learning_rate": 6.28401292181311e-05, + "loss": 0.6886, + "step": 21247 + }, + { + "epoch": 0.63, + "learning_rate": 6.283117744329817e-05, + "loss": 0.4699, + "step": 21248 + }, + { + "epoch": 0.63, + "learning_rate": 6.282222601404616e-05, + "loss": 0.6109, + "step": 21249 + }, + { + "epoch": 0.63, + "learning_rate": 6.281327493045825e-05, + "loss": 0.538, + "step": 21250 + }, + { + "epoch": 0.63, + "learning_rate": 6.280432419261774e-05, + "loss": 0.5951, + "step": 21251 + }, + { + "epoch": 0.63, + "learning_rate": 6.279537380060776e-05, + "loss": 0.519, + "step": 21252 + }, + { + "epoch": 0.63, + "learning_rate": 6.278642375451156e-05, + "loss": 0.4761, + "step": 21253 + }, + { + "epoch": 0.63, + "learning_rate": 6.277747405441238e-05, + "loss": 0.5609, + "step": 21254 + }, + { + "epoch": 0.63, + "learning_rate": 6.276852470039342e-05, + "loss": 0.4798, + "step": 21255 + }, + { + "epoch": 0.63, + "learning_rate": 6.275957569253786e-05, + "loss": 0.6133, + "step": 21256 + }, + { + "epoch": 0.63, + "learning_rate": 6.275062703092894e-05, + "loss": 0.6254, + "step": 21257 + }, + { + "epoch": 0.63, + "learning_rate": 6.274167871564982e-05, + "loss": 0.4836, + "step": 21258 + }, + { + "epoch": 0.63, + "learning_rate": 6.273273074678376e-05, + "loss": 0.4568, + "step": 21259 + }, + { + "epoch": 0.63, + "learning_rate": 6.272378312441387e-05, + "loss": 0.3929, + "step": 21260 + }, + { + "epoch": 0.63, + "learning_rate": 6.271483584862341e-05, + "loss": 0.6019, + "step": 21261 + }, + { + "epoch": 0.63, + "learning_rate": 6.270588891949556e-05, + "loss": 0.5346, + "step": 21262 + }, + { + "epoch": 0.63, + "learning_rate": 6.269694233711345e-05, + "loss": 0.5551, + "step": 21263 + }, + { + "epoch": 0.63, + "learning_rate": 6.268799610156033e-05, + "loss": 0.5047, + "step": 21264 + }, + { + "epoch": 0.63, + "learning_rate": 6.267905021291933e-05, + "loss": 0.6147, + "step": 21265 + }, + { + "epoch": 0.63, + "learning_rate": 6.267010467127368e-05, + "loss": 0.556, + "step": 21266 + }, + { + "epoch": 0.63, + "learning_rate": 6.266115947670651e-05, + "loss": 0.5935, + "step": 21267 + }, + { + "epoch": 0.63, + "learning_rate": 6.2652214629301e-05, + "loss": 0.6106, + "step": 21268 + }, + { + "epoch": 0.63, + "learning_rate": 6.264327012914028e-05, + "loss": 0.4768, + "step": 21269 + }, + { + "epoch": 0.63, + "learning_rate": 6.263432597630759e-05, + "loss": 0.5828, + "step": 21270 + }, + { + "epoch": 0.63, + "learning_rate": 6.262538217088603e-05, + "loss": 0.552, + "step": 21271 + }, + { + "epoch": 0.63, + "learning_rate": 6.261643871295874e-05, + "loss": 0.5435, + "step": 21272 + }, + { + "epoch": 0.63, + "learning_rate": 6.260749560260892e-05, + "loss": 0.421, + "step": 21273 + }, + { + "epoch": 0.63, + "learning_rate": 6.259855283991973e-05, + "loss": 0.6536, + "step": 21274 + }, + { + "epoch": 0.63, + "learning_rate": 6.258961042497426e-05, + "loss": 0.4928, + "step": 21275 + }, + { + "epoch": 0.63, + "learning_rate": 6.258066835785567e-05, + "loss": 0.4577, + "step": 21276 + }, + { + "epoch": 0.63, + "learning_rate": 6.257172663864711e-05, + "loss": 0.5745, + "step": 21277 + }, + { + "epoch": 0.63, + "learning_rate": 6.25627852674318e-05, + "loss": 0.5303, + "step": 21278 + }, + { + "epoch": 0.63, + "learning_rate": 6.255384424429272e-05, + "loss": 0.5343, + "step": 21279 + }, + { + "epoch": 0.63, + "learning_rate": 6.254490356931308e-05, + "loss": 0.662, + "step": 21280 + }, + { + "epoch": 0.63, + "learning_rate": 6.253596324257598e-05, + "loss": 0.5172, + "step": 21281 + }, + { + "epoch": 0.63, + "learning_rate": 6.252702326416461e-05, + "loss": 0.5908, + "step": 21282 + }, + { + "epoch": 0.63, + "learning_rate": 6.251808363416201e-05, + "loss": 0.657, + "step": 21283 + }, + { + "epoch": 0.63, + "learning_rate": 6.250914435265131e-05, + "loss": 0.5317, + "step": 21284 + }, + { + "epoch": 0.63, + "learning_rate": 6.250020541971569e-05, + "loss": 0.5354, + "step": 21285 + }, + { + "epoch": 0.63, + "learning_rate": 6.249126683543821e-05, + "loss": 0.5172, + "step": 21286 + }, + { + "epoch": 0.63, + "learning_rate": 6.248232859990196e-05, + "loss": 0.5489, + "step": 21287 + }, + { + "epoch": 0.63, + "learning_rate": 6.247339071319006e-05, + "loss": 0.5899, + "step": 21288 + }, + { + "epoch": 0.63, + "learning_rate": 6.246445317538564e-05, + "loss": 0.5574, + "step": 21289 + }, + { + "epoch": 0.63, + "learning_rate": 6.245551598657177e-05, + "loss": 0.4931, + "step": 21290 + }, + { + "epoch": 0.63, + "learning_rate": 6.244657914683152e-05, + "loss": 0.461, + "step": 21291 + }, + { + "epoch": 0.63, + "learning_rate": 6.243764265624803e-05, + "loss": 0.5212, + "step": 21292 + }, + { + "epoch": 0.63, + "learning_rate": 6.242870651490435e-05, + "loss": 0.563, + "step": 21293 + }, + { + "epoch": 0.63, + "learning_rate": 6.241977072288363e-05, + "loss": 0.5007, + "step": 21294 + }, + { + "epoch": 0.63, + "learning_rate": 6.241083528026888e-05, + "loss": 0.4766, + "step": 21295 + }, + { + "epoch": 0.63, + "learning_rate": 6.240190018714316e-05, + "loss": 0.4687, + "step": 21296 + }, + { + "epoch": 0.63, + "learning_rate": 6.239296544358965e-05, + "loss": 0.498, + "step": 21297 + }, + { + "epoch": 0.63, + "learning_rate": 6.238403104969133e-05, + "loss": 0.5672, + "step": 21298 + }, + { + "epoch": 0.63, + "learning_rate": 6.23750970055313e-05, + "loss": 0.5345, + "step": 21299 + }, + { + "epoch": 0.63, + "learning_rate": 6.23661633111926e-05, + "loss": 0.6128, + "step": 21300 + }, + { + "epoch": 0.63, + "learning_rate": 6.235722996675836e-05, + "loss": 0.5183, + "step": 21301 + }, + { + "epoch": 0.63, + "learning_rate": 6.234829697231158e-05, + "loss": 0.5374, + "step": 21302 + }, + { + "epoch": 0.63, + "learning_rate": 6.233936432793528e-05, + "loss": 0.5743, + "step": 21303 + }, + { + "epoch": 0.63, + "learning_rate": 6.23304320337126e-05, + "loss": 0.5855, + "step": 21304 + }, + { + "epoch": 0.63, + "learning_rate": 6.232150008972658e-05, + "loss": 0.5229, + "step": 21305 + }, + { + "epoch": 0.63, + "learning_rate": 6.231256849606017e-05, + "loss": 0.5464, + "step": 21306 + }, + { + "epoch": 0.63, + "learning_rate": 6.230363725279651e-05, + "loss": 0.5738, + "step": 21307 + }, + { + "epoch": 0.63, + "learning_rate": 6.22947063600186e-05, + "loss": 0.609, + "step": 21308 + }, + { + "epoch": 0.63, + "learning_rate": 6.228577581780952e-05, + "loss": 0.3359, + "step": 21309 + }, + { + "epoch": 0.63, + "learning_rate": 6.22768456262522e-05, + "loss": 0.529, + "step": 21310 + }, + { + "epoch": 0.63, + "learning_rate": 6.226791578542978e-05, + "loss": 0.5716, + "step": 21311 + }, + { + "epoch": 0.63, + "learning_rate": 6.22589862954252e-05, + "loss": 0.5094, + "step": 21312 + }, + { + "epoch": 0.63, + "learning_rate": 6.225005715632157e-05, + "loss": 0.5833, + "step": 21313 + }, + { + "epoch": 0.63, + "learning_rate": 6.224112836820185e-05, + "loss": 0.5448, + "step": 21314 + }, + { + "epoch": 0.63, + "learning_rate": 6.223219993114904e-05, + "loss": 0.6251, + "step": 21315 + }, + { + "epoch": 0.63, + "learning_rate": 6.222327184524621e-05, + "loss": 0.5693, + "step": 21316 + }, + { + "epoch": 0.63, + "learning_rate": 6.221434411057636e-05, + "loss": 0.4554, + "step": 21317 + }, + { + "epoch": 0.63, + "learning_rate": 6.220541672722244e-05, + "loss": 0.5044, + "step": 21318 + }, + { + "epoch": 0.63, + "learning_rate": 6.21964896952675e-05, + "loss": 0.6129, + "step": 21319 + }, + { + "epoch": 0.63, + "learning_rate": 6.218756301479454e-05, + "loss": 0.5243, + "step": 21320 + }, + { + "epoch": 0.63, + "learning_rate": 6.217863668588658e-05, + "loss": 0.4487, + "step": 21321 + }, + { + "epoch": 0.63, + "learning_rate": 6.216971070862653e-05, + "loss": 0.5008, + "step": 21322 + }, + { + "epoch": 0.63, + "learning_rate": 6.216078508309745e-05, + "loss": 0.4968, + "step": 21323 + }, + { + "epoch": 0.63, + "learning_rate": 6.21518598093823e-05, + "loss": 0.5393, + "step": 21324 + }, + { + "epoch": 0.63, + "learning_rate": 6.21429348875641e-05, + "loss": 0.6501, + "step": 21325 + }, + { + "epoch": 0.63, + "learning_rate": 6.213401031772578e-05, + "loss": 0.5507, + "step": 21326 + }, + { + "epoch": 0.63, + "learning_rate": 6.212508609995034e-05, + "loss": 0.5418, + "step": 21327 + }, + { + "epoch": 0.63, + "learning_rate": 6.211616223432077e-05, + "loss": 0.4518, + "step": 21328 + }, + { + "epoch": 0.63, + "learning_rate": 6.210723872092002e-05, + "loss": 0.6128, + "step": 21329 + }, + { + "epoch": 0.64, + "learning_rate": 6.209831555983106e-05, + "loss": 0.5744, + "step": 21330 + }, + { + "epoch": 0.64, + "learning_rate": 6.208939275113684e-05, + "loss": 0.4565, + "step": 21331 + }, + { + "epoch": 0.64, + "learning_rate": 6.208047029492037e-05, + "loss": 0.5208, + "step": 21332 + }, + { + "epoch": 0.64, + "learning_rate": 6.207154819126455e-05, + "loss": 0.5364, + "step": 21333 + }, + { + "epoch": 0.64, + "learning_rate": 6.206262644025234e-05, + "loss": 0.5546, + "step": 21334 + }, + { + "epoch": 0.64, + "learning_rate": 6.205370504196674e-05, + "loss": 0.4789, + "step": 21335 + }, + { + "epoch": 0.64, + "learning_rate": 6.204478399649067e-05, + "loss": 0.6728, + "step": 21336 + }, + { + "epoch": 0.64, + "learning_rate": 6.203586330390705e-05, + "loss": 0.5262, + "step": 21337 + }, + { + "epoch": 0.64, + "learning_rate": 6.202694296429885e-05, + "loss": 0.5118, + "step": 21338 + }, + { + "epoch": 0.64, + "learning_rate": 6.201802297774896e-05, + "loss": 0.5025, + "step": 21339 + }, + { + "epoch": 0.64, + "learning_rate": 6.200910334434045e-05, + "loss": 0.5087, + "step": 21340 + }, + { + "epoch": 0.64, + "learning_rate": 6.200018406415609e-05, + "loss": 0.5869, + "step": 21341 + }, + { + "epoch": 0.64, + "learning_rate": 6.199126513727888e-05, + "loss": 0.4598, + "step": 21342 + }, + { + "epoch": 0.64, + "learning_rate": 6.198234656379172e-05, + "loss": 0.5325, + "step": 21343 + }, + { + "epoch": 0.64, + "learning_rate": 6.19734283437776e-05, + "loss": 0.5129, + "step": 21344 + }, + { + "epoch": 0.64, + "learning_rate": 6.196451047731934e-05, + "loss": 0.5282, + "step": 21345 + }, + { + "epoch": 0.64, + "learning_rate": 6.195559296449992e-05, + "loss": 0.4018, + "step": 21346 + }, + { + "epoch": 0.64, + "learning_rate": 6.194667580540223e-05, + "loss": 0.4526, + "step": 21347 + }, + { + "epoch": 0.64, + "learning_rate": 6.193775900010921e-05, + "loss": 0.5309, + "step": 21348 + }, + { + "epoch": 0.64, + "learning_rate": 6.192884254870368e-05, + "loss": 0.4514, + "step": 21349 + }, + { + "epoch": 0.64, + "learning_rate": 6.191992645126864e-05, + "loss": 0.4367, + "step": 21350 + }, + { + "epoch": 0.64, + "learning_rate": 6.191101070788694e-05, + "loss": 0.5288, + "step": 21351 + }, + { + "epoch": 0.64, + "learning_rate": 6.19020953186415e-05, + "loss": 0.3897, + "step": 21352 + }, + { + "epoch": 0.64, + "learning_rate": 6.189318028361514e-05, + "loss": 0.5538, + "step": 21353 + }, + { + "epoch": 0.64, + "learning_rate": 6.188426560289084e-05, + "loss": 0.6276, + "step": 21354 + }, + { + "epoch": 0.64, + "learning_rate": 6.187535127655141e-05, + "loss": 0.4994, + "step": 21355 + }, + { + "epoch": 0.64, + "learning_rate": 6.186643730467983e-05, + "loss": 0.5429, + "step": 21356 + }, + { + "epoch": 0.64, + "learning_rate": 6.185752368735889e-05, + "loss": 0.4985, + "step": 21357 + }, + { + "epoch": 0.64, + "learning_rate": 6.184861042467147e-05, + "loss": 0.498, + "step": 21358 + }, + { + "epoch": 0.64, + "learning_rate": 6.183969751670051e-05, + "loss": 0.5639, + "step": 21359 + }, + { + "epoch": 0.64, + "learning_rate": 6.183078496352881e-05, + "loss": 0.4609, + "step": 21360 + }, + { + "epoch": 0.64, + "learning_rate": 6.182187276523925e-05, + "loss": 0.4684, + "step": 21361 + }, + { + "epoch": 0.64, + "learning_rate": 6.181296092191468e-05, + "loss": 0.5895, + "step": 21362 + }, + { + "epoch": 0.64, + "learning_rate": 6.180404943363804e-05, + "loss": 0.5384, + "step": 21363 + }, + { + "epoch": 0.64, + "learning_rate": 6.17951383004921e-05, + "loss": 0.4669, + "step": 21364 + }, + { + "epoch": 0.64, + "learning_rate": 6.17862275225597e-05, + "loss": 0.4283, + "step": 21365 + }, + { + "epoch": 0.64, + "learning_rate": 6.177731709992375e-05, + "loss": 0.6485, + "step": 21366 + }, + { + "epoch": 0.64, + "learning_rate": 6.17684070326671e-05, + "loss": 0.5312, + "step": 21367 + }, + { + "epoch": 0.64, + "learning_rate": 6.175949732087253e-05, + "loss": 0.5061, + "step": 21368 + }, + { + "epoch": 0.64, + "learning_rate": 6.175058796462292e-05, + "loss": 0.5287, + "step": 21369 + }, + { + "epoch": 0.64, + "learning_rate": 6.174167896400109e-05, + "loss": 0.4624, + "step": 21370 + }, + { + "epoch": 0.64, + "learning_rate": 6.173277031908993e-05, + "loss": 0.5049, + "step": 21371 + }, + { + "epoch": 0.64, + "learning_rate": 6.172386202997216e-05, + "loss": 0.6393, + "step": 21372 + }, + { + "epoch": 0.64, + "learning_rate": 6.17149540967307e-05, + "loss": 0.521, + "step": 21373 + }, + { + "epoch": 0.64, + "learning_rate": 6.17060465194483e-05, + "loss": 0.5831, + "step": 21374 + }, + { + "epoch": 0.64, + "learning_rate": 6.169713929820786e-05, + "loss": 0.4164, + "step": 21375 + }, + { + "epoch": 0.64, + "learning_rate": 6.168823243309214e-05, + "loss": 0.5037, + "step": 21376 + }, + { + "epoch": 0.64, + "learning_rate": 6.167932592418393e-05, + "loss": 0.4912, + "step": 21377 + }, + { + "epoch": 0.64, + "learning_rate": 6.167041977156611e-05, + "loss": 0.5622, + "step": 21378 + }, + { + "epoch": 0.64, + "learning_rate": 6.166151397532147e-05, + "loss": 0.4405, + "step": 21379 + }, + { + "epoch": 0.64, + "learning_rate": 6.165260853553274e-05, + "loss": 0.6731, + "step": 21380 + }, + { + "epoch": 0.64, + "learning_rate": 6.164370345228281e-05, + "loss": 0.5734, + "step": 21381 + }, + { + "epoch": 0.64, + "learning_rate": 6.163479872565441e-05, + "loss": 0.5105, + "step": 21382 + }, + { + "epoch": 0.64, + "learning_rate": 6.16258943557304e-05, + "loss": 0.4946, + "step": 21383 + }, + { + "epoch": 0.64, + "learning_rate": 6.16169903425935e-05, + "loss": 0.5865, + "step": 21384 + }, + { + "epoch": 0.64, + "learning_rate": 6.160808668632654e-05, + "loss": 0.6121, + "step": 21385 + }, + { + "epoch": 0.64, + "learning_rate": 6.159918338701226e-05, + "loss": 0.5891, + "step": 21386 + }, + { + "epoch": 0.64, + "learning_rate": 6.159028044473351e-05, + "loss": 0.5424, + "step": 21387 + }, + { + "epoch": 0.64, + "learning_rate": 6.158137785957301e-05, + "loss": 0.4384, + "step": 21388 + }, + { + "epoch": 0.64, + "learning_rate": 6.157247563161352e-05, + "loss": 0.5864, + "step": 21389 + }, + { + "epoch": 0.64, + "learning_rate": 6.156357376093787e-05, + "loss": 0.5619, + "step": 21390 + }, + { + "epoch": 0.64, + "learning_rate": 6.155467224762878e-05, + "loss": 0.5644, + "step": 21391 + }, + { + "epoch": 0.64, + "learning_rate": 6.154577109176902e-05, + "loss": 0.514, + "step": 21392 + }, + { + "epoch": 0.64, + "learning_rate": 6.153687029344135e-05, + "loss": 0.4905, + "step": 21393 + }, + { + "epoch": 0.64, + "learning_rate": 6.152796985272855e-05, + "loss": 0.5455, + "step": 21394 + }, + { + "epoch": 0.64, + "learning_rate": 6.151906976971335e-05, + "loss": 0.5517, + "step": 21395 + }, + { + "epoch": 0.64, + "learning_rate": 6.151017004447848e-05, + "loss": 0.6005, + "step": 21396 + }, + { + "epoch": 0.64, + "learning_rate": 6.150127067710671e-05, + "loss": 0.5228, + "step": 21397 + }, + { + "epoch": 0.64, + "learning_rate": 6.149237166768083e-05, + "loss": 0.5784, + "step": 21398 + }, + { + "epoch": 0.64, + "learning_rate": 6.148347301628346e-05, + "loss": 0.5744, + "step": 21399 + }, + { + "epoch": 0.64, + "learning_rate": 6.147457472299745e-05, + "loss": 0.5272, + "step": 21400 + }, + { + "epoch": 0.64, + "learning_rate": 6.146567678790544e-05, + "loss": 0.5476, + "step": 21401 + }, + { + "epoch": 0.64, + "learning_rate": 6.14567792110903e-05, + "loss": 0.7453, + "step": 21402 + }, + { + "epoch": 0.64, + "learning_rate": 6.144788199263459e-05, + "loss": 0.517, + "step": 21403 + }, + { + "epoch": 0.64, + "learning_rate": 6.143898513262114e-05, + "loss": 0.5044, + "step": 21404 + }, + { + "epoch": 0.64, + "learning_rate": 6.14300886311326e-05, + "loss": 0.5608, + "step": 21405 + }, + { + "epoch": 0.64, + "learning_rate": 6.142119248825177e-05, + "loss": 0.519, + "step": 21406 + }, + { + "epoch": 0.64, + "learning_rate": 6.14122967040613e-05, + "loss": 0.4608, + "step": 21407 + }, + { + "epoch": 0.64, + "learning_rate": 6.140340127864391e-05, + "loss": 0.6351, + "step": 21408 + }, + { + "epoch": 0.64, + "learning_rate": 6.139450621208232e-05, + "loss": 0.5194, + "step": 21409 + }, + { + "epoch": 0.64, + "learning_rate": 6.138561150445923e-05, + "loss": 0.5233, + "step": 21410 + }, + { + "epoch": 0.64, + "learning_rate": 6.137671715585731e-05, + "loss": 0.5582, + "step": 21411 + }, + { + "epoch": 0.64, + "learning_rate": 6.136782316635929e-05, + "loss": 0.5924, + "step": 21412 + }, + { + "epoch": 0.64, + "learning_rate": 6.135892953604784e-05, + "loss": 0.4514, + "step": 21413 + }, + { + "epoch": 0.64, + "learning_rate": 6.135003626500573e-05, + "loss": 0.4808, + "step": 21414 + }, + { + "epoch": 0.64, + "learning_rate": 6.13411433533155e-05, + "loss": 0.5346, + "step": 21415 + }, + { + "epoch": 0.64, + "learning_rate": 6.133225080105993e-05, + "loss": 0.4862, + "step": 21416 + }, + { + "epoch": 0.64, + "learning_rate": 6.132335860832168e-05, + "loss": 0.4702, + "step": 21417 + }, + { + "epoch": 0.64, + "learning_rate": 6.131446677518345e-05, + "loss": 0.5166, + "step": 21418 + }, + { + "epoch": 0.64, + "learning_rate": 6.13055753017279e-05, + "loss": 0.5433, + "step": 21419 + }, + { + "epoch": 0.64, + "learning_rate": 6.129668418803762e-05, + "loss": 0.7136, + "step": 21420 + }, + { + "epoch": 0.64, + "learning_rate": 6.128779343419539e-05, + "loss": 0.4, + "step": 21421 + }, + { + "epoch": 0.64, + "learning_rate": 6.127890304028386e-05, + "loss": 0.5625, + "step": 21422 + }, + { + "epoch": 0.64, + "learning_rate": 6.12700130063856e-05, + "loss": 0.5691, + "step": 21423 + }, + { + "epoch": 0.64, + "learning_rate": 6.126112333258334e-05, + "loss": 0.481, + "step": 21424 + }, + { + "epoch": 0.64, + "learning_rate": 6.125223401895976e-05, + "loss": 0.5895, + "step": 21425 + }, + { + "epoch": 0.64, + "learning_rate": 6.124334506559743e-05, + "loss": 0.4677, + "step": 21426 + }, + { + "epoch": 0.64, + "learning_rate": 6.123445647257899e-05, + "loss": 0.5356, + "step": 21427 + }, + { + "epoch": 0.64, + "learning_rate": 6.122556823998718e-05, + "loss": 0.6074, + "step": 21428 + }, + { + "epoch": 0.64, + "learning_rate": 6.121668036790457e-05, + "loss": 0.4989, + "step": 21429 + }, + { + "epoch": 0.64, + "learning_rate": 6.120779285641378e-05, + "loss": 0.5265, + "step": 21430 + }, + { + "epoch": 0.64, + "learning_rate": 6.11989057055975e-05, + "loss": 0.4974, + "step": 21431 + }, + { + "epoch": 0.64, + "learning_rate": 6.11900189155383e-05, + "loss": 0.543, + "step": 21432 + }, + { + "epoch": 0.64, + "learning_rate": 6.11811324863189e-05, + "loss": 0.5168, + "step": 21433 + }, + { + "epoch": 0.64, + "learning_rate": 6.11722464180218e-05, + "loss": 0.5805, + "step": 21434 + }, + { + "epoch": 0.64, + "learning_rate": 6.11633607107297e-05, + "loss": 0.5839, + "step": 21435 + }, + { + "epoch": 0.64, + "learning_rate": 6.115447536452516e-05, + "loss": 0.4963, + "step": 21436 + }, + { + "epoch": 0.64, + "learning_rate": 6.114559037949088e-05, + "loss": 0.5283, + "step": 21437 + }, + { + "epoch": 0.64, + "learning_rate": 6.11367057557094e-05, + "loss": 0.4587, + "step": 21438 + }, + { + "epoch": 0.64, + "learning_rate": 6.11278214932633e-05, + "loss": 0.4757, + "step": 21439 + }, + { + "epoch": 0.64, + "learning_rate": 6.111893759223526e-05, + "loss": 0.5888, + "step": 21440 + }, + { + "epoch": 0.64, + "learning_rate": 6.111005405270786e-05, + "loss": 0.469, + "step": 21441 + }, + { + "epoch": 0.64, + "learning_rate": 6.110117087476362e-05, + "loss": 0.5627, + "step": 21442 + }, + { + "epoch": 0.64, + "learning_rate": 6.109228805848523e-05, + "loss": 0.406, + "step": 21443 + }, + { + "epoch": 0.64, + "learning_rate": 6.10834056039552e-05, + "loss": 0.5004, + "step": 21444 + }, + { + "epoch": 0.64, + "learning_rate": 6.107452351125623e-05, + "loss": 0.4853, + "step": 21445 + }, + { + "epoch": 0.64, + "learning_rate": 6.106564178047077e-05, + "loss": 0.5924, + "step": 21446 + }, + { + "epoch": 0.64, + "learning_rate": 6.105676041168146e-05, + "loss": 0.4782, + "step": 21447 + }, + { + "epoch": 0.64, + "learning_rate": 6.104787940497085e-05, + "loss": 0.5516, + "step": 21448 + }, + { + "epoch": 0.64, + "learning_rate": 6.103899876042158e-05, + "loss": 0.5138, + "step": 21449 + }, + { + "epoch": 0.64, + "learning_rate": 6.103011847811615e-05, + "loss": 0.5379, + "step": 21450 + }, + { + "epoch": 0.64, + "learning_rate": 6.102123855813713e-05, + "loss": 0.5728, + "step": 21451 + }, + { + "epoch": 0.64, + "learning_rate": 6.101235900056712e-05, + "loss": 0.4545, + "step": 21452 + }, + { + "epoch": 0.64, + "learning_rate": 6.1003479805488674e-05, + "loss": 0.5208, + "step": 21453 + }, + { + "epoch": 0.64, + "learning_rate": 6.099460097298429e-05, + "loss": 0.537, + "step": 21454 + }, + { + "epoch": 0.64, + "learning_rate": 6.098572250313657e-05, + "loss": 0.4504, + "step": 21455 + }, + { + "epoch": 0.64, + "learning_rate": 6.097684439602807e-05, + "loss": 0.608, + "step": 21456 + }, + { + "epoch": 0.64, + "learning_rate": 6.0967966651741326e-05, + "loss": 0.425, + "step": 21457 + }, + { + "epoch": 0.64, + "learning_rate": 6.095908927035883e-05, + "loss": 0.6306, + "step": 21458 + }, + { + "epoch": 0.64, + "learning_rate": 6.095021225196319e-05, + "loss": 0.3648, + "step": 21459 + }, + { + "epoch": 0.64, + "learning_rate": 6.094133559663693e-05, + "loss": 0.5776, + "step": 21460 + }, + { + "epoch": 0.64, + "learning_rate": 6.093245930446252e-05, + "loss": 0.5285, + "step": 21461 + }, + { + "epoch": 0.64, + "learning_rate": 6.092358337552255e-05, + "loss": 0.5863, + "step": 21462 + }, + { + "epoch": 0.64, + "learning_rate": 6.091470780989952e-05, + "loss": 0.6227, + "step": 21463 + }, + { + "epoch": 0.64, + "learning_rate": 6.0905832607675995e-05, + "loss": 0.5211, + "step": 21464 + }, + { + "epoch": 0.64, + "learning_rate": 6.089695776893443e-05, + "loss": 0.5122, + "step": 21465 + }, + { + "epoch": 0.64, + "learning_rate": 6.088808329375737e-05, + "loss": 0.4766, + "step": 21466 + }, + { + "epoch": 0.64, + "learning_rate": 6.087920918222732e-05, + "loss": 0.5236, + "step": 21467 + }, + { + "epoch": 0.64, + "learning_rate": 6.0870335434426795e-05, + "loss": 0.5324, + "step": 21468 + }, + { + "epoch": 0.64, + "learning_rate": 6.08614620504383e-05, + "loss": 0.5354, + "step": 21469 + }, + { + "epoch": 0.64, + "learning_rate": 6.085258903034431e-05, + "loss": 0.5433, + "step": 21470 + }, + { + "epoch": 0.64, + "learning_rate": 6.084371637422737e-05, + "loss": 0.6052, + "step": 21471 + }, + { + "epoch": 0.64, + "learning_rate": 6.083484408216995e-05, + "loss": 0.5128, + "step": 21472 + }, + { + "epoch": 0.64, + "learning_rate": 6.082597215425452e-05, + "loss": 0.4401, + "step": 21473 + }, + { + "epoch": 0.64, + "learning_rate": 6.081710059056359e-05, + "loss": 0.5423, + "step": 21474 + }, + { + "epoch": 0.64, + "learning_rate": 6.0808229391179626e-05, + "loss": 0.5049, + "step": 21475 + }, + { + "epoch": 0.64, + "learning_rate": 6.0799358556185195e-05, + "loss": 0.4605, + "step": 21476 + }, + { + "epoch": 0.64, + "learning_rate": 6.0790488085662645e-05, + "loss": 0.5659, + "step": 21477 + }, + { + "epoch": 0.64, + "learning_rate": 6.078161797969453e-05, + "loss": 0.6345, + "step": 21478 + }, + { + "epoch": 0.64, + "learning_rate": 6.0772748238363264e-05, + "loss": 0.5184, + "step": 21479 + }, + { + "epoch": 0.64, + "learning_rate": 6.07638788617514e-05, + "loss": 0.474, + "step": 21480 + }, + { + "epoch": 0.64, + "learning_rate": 6.0755009849941335e-05, + "loss": 0.578, + "step": 21481 + }, + { + "epoch": 0.64, + "learning_rate": 6.074614120301552e-05, + "loss": 0.5475, + "step": 21482 + }, + { + "epoch": 0.64, + "learning_rate": 6.0737272921056466e-05, + "loss": 0.6575, + "step": 21483 + }, + { + "epoch": 0.64, + "learning_rate": 6.072840500414661e-05, + "loss": 0.5587, + "step": 21484 + }, + { + "epoch": 0.64, + "learning_rate": 6.0719537452368355e-05, + "loss": 0.5773, + "step": 21485 + }, + { + "epoch": 0.64, + "learning_rate": 6.071067026580422e-05, + "loss": 0.533, + "step": 21486 + }, + { + "epoch": 0.64, + "learning_rate": 6.0701803444536596e-05, + "loss": 0.4921, + "step": 21487 + }, + { + "epoch": 0.64, + "learning_rate": 6.0692936988647974e-05, + "loss": 0.5278, + "step": 21488 + }, + { + "epoch": 0.64, + "learning_rate": 6.0684070898220715e-05, + "loss": 0.4805, + "step": 21489 + }, + { + "epoch": 0.64, + "learning_rate": 6.0675205173337315e-05, + "loss": 0.6011, + "step": 21490 + }, + { + "epoch": 0.64, + "learning_rate": 6.0666339814080186e-05, + "loss": 0.4783, + "step": 21491 + }, + { + "epoch": 0.64, + "learning_rate": 6.0657474820531735e-05, + "loss": 0.4886, + "step": 21492 + }, + { + "epoch": 0.64, + "learning_rate": 6.0648610192774414e-05, + "loss": 0.5831, + "step": 21493 + }, + { + "epoch": 0.64, + "learning_rate": 6.063974593089061e-05, + "loss": 0.4483, + "step": 21494 + }, + { + "epoch": 0.64, + "learning_rate": 6.063088203496281e-05, + "loss": 0.5358, + "step": 21495 + }, + { + "epoch": 0.64, + "learning_rate": 6.062201850507334e-05, + "loss": 0.4958, + "step": 21496 + }, + { + "epoch": 0.64, + "learning_rate": 6.061315534130467e-05, + "loss": 0.5409, + "step": 21497 + }, + { + "epoch": 0.64, + "learning_rate": 6.060429254373915e-05, + "loss": 0.5447, + "step": 21498 + }, + { + "epoch": 0.64, + "learning_rate": 6.059543011245925e-05, + "loss": 0.481, + "step": 21499 + }, + { + "epoch": 0.64, + "learning_rate": 6.058656804754733e-05, + "loss": 0.5864, + "step": 21500 + }, + { + "epoch": 0.64, + "learning_rate": 6.0577706349085775e-05, + "loss": 0.491, + "step": 21501 + }, + { + "epoch": 0.64, + "learning_rate": 6.0568845017156996e-05, + "loss": 0.5453, + "step": 21502 + }, + { + "epoch": 0.64, + "learning_rate": 6.055998405184341e-05, + "loss": 0.4492, + "step": 21503 + }, + { + "epoch": 0.64, + "learning_rate": 6.055112345322733e-05, + "loss": 0.5086, + "step": 21504 + }, + { + "epoch": 0.64, + "learning_rate": 6.054226322139121e-05, + "loss": 0.4802, + "step": 21505 + }, + { + "epoch": 0.64, + "learning_rate": 6.053340335641737e-05, + "loss": 0.5653, + "step": 21506 + }, + { + "epoch": 0.64, + "learning_rate": 6.052454385838828e-05, + "loss": 0.5431, + "step": 21507 + }, + { + "epoch": 0.64, + "learning_rate": 6.051568472738618e-05, + "loss": 0.5747, + "step": 21508 + }, + { + "epoch": 0.64, + "learning_rate": 6.050682596349354e-05, + "loss": 0.5652, + "step": 21509 + }, + { + "epoch": 0.64, + "learning_rate": 6.049796756679267e-05, + "loss": 0.5979, + "step": 21510 + }, + { + "epoch": 0.64, + "learning_rate": 6.0489109537365984e-05, + "loss": 0.477, + "step": 21511 + }, + { + "epoch": 0.64, + "learning_rate": 6.0480251875295804e-05, + "loss": 0.5653, + "step": 21512 + }, + { + "epoch": 0.64, + "learning_rate": 6.047139458066445e-05, + "loss": 0.5212, + "step": 21513 + }, + { + "epoch": 0.64, + "learning_rate": 6.046253765355435e-05, + "loss": 0.5051, + "step": 21514 + }, + { + "epoch": 0.64, + "learning_rate": 6.045368109404783e-05, + "loss": 0.56, + "step": 21515 + }, + { + "epoch": 0.64, + "learning_rate": 6.044482490222719e-05, + "loss": 0.5596, + "step": 21516 + }, + { + "epoch": 0.64, + "learning_rate": 6.0435969078174816e-05, + "loss": 0.5295, + "step": 21517 + }, + { + "epoch": 0.64, + "learning_rate": 6.042711362197302e-05, + "loss": 0.5439, + "step": 21518 + }, + { + "epoch": 0.64, + "learning_rate": 6.04182585337042e-05, + "loss": 0.508, + "step": 21519 + }, + { + "epoch": 0.64, + "learning_rate": 6.0409403813450584e-05, + "loss": 0.5283, + "step": 21520 + }, + { + "epoch": 0.64, + "learning_rate": 6.040054946129456e-05, + "loss": 0.6059, + "step": 21521 + }, + { + "epoch": 0.64, + "learning_rate": 6.039169547731845e-05, + "loss": 0.5178, + "step": 21522 + }, + { + "epoch": 0.64, + "learning_rate": 6.038284186160459e-05, + "loss": 0.6615, + "step": 21523 + }, + { + "epoch": 0.64, + "learning_rate": 6.0373988614235254e-05, + "loss": 0.5737, + "step": 21524 + }, + { + "epoch": 0.64, + "learning_rate": 6.0365135735292766e-05, + "loss": 0.585, + "step": 21525 + }, + { + "epoch": 0.64, + "learning_rate": 6.035628322485949e-05, + "loss": 0.5421, + "step": 21526 + }, + { + "epoch": 0.64, + "learning_rate": 6.034743108301766e-05, + "loss": 0.5703, + "step": 21527 + }, + { + "epoch": 0.64, + "learning_rate": 6.033857930984963e-05, + "loss": 0.5418, + "step": 21528 + }, + { + "epoch": 0.64, + "learning_rate": 6.032972790543765e-05, + "loss": 0.5002, + "step": 21529 + }, + { + "epoch": 0.64, + "learning_rate": 6.03208768698641e-05, + "loss": 0.5422, + "step": 21530 + }, + { + "epoch": 0.64, + "learning_rate": 6.031202620321118e-05, + "loss": 0.4865, + "step": 21531 + }, + { + "epoch": 0.64, + "learning_rate": 6.0303175905561224e-05, + "loss": 0.6185, + "step": 21532 + }, + { + "epoch": 0.64, + "learning_rate": 6.0294325976996524e-05, + "loss": 0.5781, + "step": 21533 + }, + { + "epoch": 0.64, + "learning_rate": 6.0285476417599384e-05, + "loss": 0.5171, + "step": 21534 + }, + { + "epoch": 0.64, + "learning_rate": 6.027662722745201e-05, + "loss": 0.5749, + "step": 21535 + }, + { + "epoch": 0.64, + "learning_rate": 6.026777840663675e-05, + "loss": 0.5133, + "step": 21536 + }, + { + "epoch": 0.64, + "learning_rate": 6.025892995523582e-05, + "loss": 0.5292, + "step": 21537 + }, + { + "epoch": 0.64, + "learning_rate": 6.025008187333159e-05, + "loss": 0.5121, + "step": 21538 + }, + { + "epoch": 0.64, + "learning_rate": 6.02412341610062e-05, + "loss": 0.5379, + "step": 21539 + }, + { + "epoch": 0.64, + "learning_rate": 6.023238681834198e-05, + "loss": 0.4523, + "step": 21540 + }, + { + "epoch": 0.64, + "learning_rate": 6.022353984542115e-05, + "loss": 0.5017, + "step": 21541 + }, + { + "epoch": 0.64, + "learning_rate": 6.021469324232605e-05, + "loss": 0.548, + "step": 21542 + }, + { + "epoch": 0.64, + "learning_rate": 6.020584700913885e-05, + "loss": 0.4619, + "step": 21543 + }, + { + "epoch": 0.64, + "learning_rate": 6.01970011459418e-05, + "loss": 0.4475, + "step": 21544 + }, + { + "epoch": 0.64, + "learning_rate": 6.0188155652817194e-05, + "loss": 0.6584, + "step": 21545 + }, + { + "epoch": 0.64, + "learning_rate": 6.0179310529847266e-05, + "loss": 0.6776, + "step": 21546 + }, + { + "epoch": 0.64, + "learning_rate": 6.017046577711419e-05, + "loss": 0.4769, + "step": 21547 + }, + { + "epoch": 0.64, + "learning_rate": 6.016162139470027e-05, + "loss": 0.5525, + "step": 21548 + }, + { + "epoch": 0.64, + "learning_rate": 6.015277738268772e-05, + "loss": 0.5002, + "step": 21549 + }, + { + "epoch": 0.64, + "learning_rate": 6.014393374115881e-05, + "loss": 0.5949, + "step": 21550 + }, + { + "epoch": 0.64, + "learning_rate": 6.013509047019565e-05, + "loss": 0.468, + "step": 21551 + }, + { + "epoch": 0.64, + "learning_rate": 6.012624756988056e-05, + "loss": 0.4828, + "step": 21552 + }, + { + "epoch": 0.64, + "learning_rate": 6.0117405040295704e-05, + "loss": 0.5161, + "step": 21553 + }, + { + "epoch": 0.64, + "learning_rate": 6.0108562881523355e-05, + "loss": 0.4882, + "step": 21554 + }, + { + "epoch": 0.64, + "learning_rate": 6.009972109364568e-05, + "loss": 0.4707, + "step": 21555 + }, + { + "epoch": 0.64, + "learning_rate": 6.0090879676744874e-05, + "loss": 0.4772, + "step": 21556 + }, + { + "epoch": 0.64, + "learning_rate": 6.00820386309032e-05, + "loss": 0.5451, + "step": 21557 + }, + { + "epoch": 0.64, + "learning_rate": 6.00731979562028e-05, + "loss": 0.5162, + "step": 21558 + }, + { + "epoch": 0.64, + "learning_rate": 6.006435765272587e-05, + "loss": 0.5055, + "step": 21559 + }, + { + "epoch": 0.64, + "learning_rate": 6.0055517720554646e-05, + "loss": 0.6946, + "step": 21560 + }, + { + "epoch": 0.64, + "learning_rate": 6.004667815977132e-05, + "loss": 0.4795, + "step": 21561 + }, + { + "epoch": 0.64, + "learning_rate": 6.003783897045805e-05, + "loss": 0.6061, + "step": 21562 + }, + { + "epoch": 0.64, + "learning_rate": 6.002900015269699e-05, + "loss": 0.5066, + "step": 21563 + }, + { + "epoch": 0.64, + "learning_rate": 6.0020161706570366e-05, + "loss": 0.4677, + "step": 21564 + }, + { + "epoch": 0.64, + "learning_rate": 6.0011323632160376e-05, + "loss": 0.5725, + "step": 21565 + }, + { + "epoch": 0.64, + "learning_rate": 6.0002485929549134e-05, + "loss": 0.4895, + "step": 21566 + }, + { + "epoch": 0.64, + "learning_rate": 5.999364859881884e-05, + "loss": 0.5866, + "step": 21567 + }, + { + "epoch": 0.64, + "learning_rate": 5.998481164005164e-05, + "loss": 0.5956, + "step": 21568 + }, + { + "epoch": 0.64, + "learning_rate": 5.997597505332977e-05, + "loss": 0.5006, + "step": 21569 + }, + { + "epoch": 0.64, + "learning_rate": 5.9967138838735285e-05, + "loss": 0.5867, + "step": 21570 + }, + { + "epoch": 0.64, + "learning_rate": 5.995830299635039e-05, + "loss": 0.4942, + "step": 21571 + }, + { + "epoch": 0.64, + "learning_rate": 5.994946752625723e-05, + "loss": 0.4904, + "step": 21572 + }, + { + "epoch": 0.64, + "learning_rate": 5.994063242853798e-05, + "loss": 0.5121, + "step": 21573 + }, + { + "epoch": 0.64, + "learning_rate": 5.993179770327476e-05, + "loss": 0.5653, + "step": 21574 + }, + { + "epoch": 0.64, + "learning_rate": 5.99229633505497e-05, + "loss": 0.6353, + "step": 21575 + }, + { + "epoch": 0.64, + "learning_rate": 5.991412937044496e-05, + "loss": 0.6776, + "step": 21576 + }, + { + "epoch": 0.64, + "learning_rate": 5.990529576304269e-05, + "loss": 0.517, + "step": 21577 + }, + { + "epoch": 0.64, + "learning_rate": 5.9896462528424954e-05, + "loss": 0.4685, + "step": 21578 + }, + { + "epoch": 0.64, + "learning_rate": 5.988762966667396e-05, + "loss": 0.5658, + "step": 21579 + }, + { + "epoch": 0.64, + "learning_rate": 5.987879717787177e-05, + "loss": 0.54, + "step": 21580 + }, + { + "epoch": 0.64, + "learning_rate": 5.98699650621006e-05, + "loss": 0.5213, + "step": 21581 + }, + { + "epoch": 0.64, + "learning_rate": 5.986113331944243e-05, + "loss": 0.5452, + "step": 21582 + }, + { + "epoch": 0.64, + "learning_rate": 5.9852301949979474e-05, + "loss": 0.5636, + "step": 21583 + }, + { + "epoch": 0.64, + "learning_rate": 5.984347095379379e-05, + "loss": 0.4862, + "step": 21584 + }, + { + "epoch": 0.64, + "learning_rate": 5.9834640330967554e-05, + "loss": 0.5306, + "step": 21585 + }, + { + "epoch": 0.64, + "learning_rate": 5.982581008158279e-05, + "loss": 0.5695, + "step": 21586 + }, + { + "epoch": 0.64, + "learning_rate": 5.9816980205721615e-05, + "loss": 0.4519, + "step": 21587 + }, + { + "epoch": 0.64, + "learning_rate": 5.980815070346617e-05, + "loss": 0.5232, + "step": 21588 + }, + { + "epoch": 0.64, + "learning_rate": 5.979932157489854e-05, + "loss": 0.5068, + "step": 21589 + }, + { + "epoch": 0.64, + "learning_rate": 5.979049282010075e-05, + "loss": 0.6396, + "step": 21590 + }, + { + "epoch": 0.64, + "learning_rate": 5.9781664439154964e-05, + "loss": 0.4886, + "step": 21591 + }, + { + "epoch": 0.64, + "learning_rate": 5.977283643214324e-05, + "loss": 0.5669, + "step": 21592 + }, + { + "epoch": 0.64, + "learning_rate": 5.976400879914763e-05, + "loss": 0.5838, + "step": 21593 + }, + { + "epoch": 0.64, + "learning_rate": 5.975518154025022e-05, + "loss": 0.5446, + "step": 21594 + }, + { + "epoch": 0.64, + "learning_rate": 5.97463546555331e-05, + "loss": 0.5318, + "step": 21595 + }, + { + "epoch": 0.64, + "learning_rate": 5.973752814507836e-05, + "loss": 0.4779, + "step": 21596 + }, + { + "epoch": 0.64, + "learning_rate": 5.972870200896799e-05, + "loss": 0.5061, + "step": 21597 + }, + { + "epoch": 0.64, + "learning_rate": 5.9719876247284125e-05, + "loss": 0.5756, + "step": 21598 + }, + { + "epoch": 0.64, + "learning_rate": 5.971105086010877e-05, + "loss": 0.5165, + "step": 21599 + }, + { + "epoch": 0.64, + "learning_rate": 5.970222584752405e-05, + "loss": 0.5996, + "step": 21600 + }, + { + "epoch": 0.64, + "learning_rate": 5.969340120961195e-05, + "loss": 0.6206, + "step": 21601 + }, + { + "epoch": 0.64, + "learning_rate": 5.968457694645454e-05, + "loss": 0.6092, + "step": 21602 + }, + { + "epoch": 0.64, + "learning_rate": 5.967575305813384e-05, + "loss": 0.5062, + "step": 21603 + }, + { + "epoch": 0.64, + "learning_rate": 5.966692954473197e-05, + "loss": 0.4821, + "step": 21604 + }, + { + "epoch": 0.64, + "learning_rate": 5.9658106406330894e-05, + "loss": 0.4614, + "step": 21605 + }, + { + "epoch": 0.64, + "learning_rate": 5.9649283643012646e-05, + "loss": 0.3938, + "step": 21606 + }, + { + "epoch": 0.64, + "learning_rate": 5.964046125485929e-05, + "loss": 0.3765, + "step": 21607 + }, + { + "epoch": 0.64, + "learning_rate": 5.9631639241952854e-05, + "loss": 0.5696, + "step": 21608 + }, + { + "epoch": 0.64, + "learning_rate": 5.9622817604375315e-05, + "loss": 0.4992, + "step": 21609 + }, + { + "epoch": 0.64, + "learning_rate": 5.961399634220874e-05, + "loss": 0.5757, + "step": 21610 + }, + { + "epoch": 0.64, + "learning_rate": 5.960517545553512e-05, + "loss": 0.4663, + "step": 21611 + }, + { + "epoch": 0.64, + "learning_rate": 5.959635494443654e-05, + "loss": 0.6533, + "step": 21612 + }, + { + "epoch": 0.64, + "learning_rate": 5.9587534808994884e-05, + "loss": 0.5117, + "step": 21613 + }, + { + "epoch": 0.64, + "learning_rate": 5.9578715049292236e-05, + "loss": 0.4592, + "step": 21614 + }, + { + "epoch": 0.64, + "learning_rate": 5.956989566541056e-05, + "loss": 0.4222, + "step": 21615 + }, + { + "epoch": 0.64, + "learning_rate": 5.9561076657431936e-05, + "loss": 0.5967, + "step": 21616 + }, + { + "epoch": 0.64, + "learning_rate": 5.955225802543828e-05, + "loss": 0.4662, + "step": 21617 + }, + { + "epoch": 0.64, + "learning_rate": 5.954343976951158e-05, + "loss": 0.5873, + "step": 21618 + }, + { + "epoch": 0.64, + "learning_rate": 5.9534621889733886e-05, + "loss": 0.5887, + "step": 21619 + }, + { + "epoch": 0.64, + "learning_rate": 5.952580438618716e-05, + "loss": 0.4727, + "step": 21620 + }, + { + "epoch": 0.64, + "learning_rate": 5.951698725895334e-05, + "loss": 0.5936, + "step": 21621 + }, + { + "epoch": 0.64, + "learning_rate": 5.950817050811445e-05, + "loss": 0.551, + "step": 21622 + }, + { + "epoch": 0.64, + "learning_rate": 5.9499354133752475e-05, + "loss": 0.4564, + "step": 21623 + }, + { + "epoch": 0.64, + "learning_rate": 5.9490538135949337e-05, + "loss": 0.5546, + "step": 21624 + }, + { + "epoch": 0.64, + "learning_rate": 5.948172251478702e-05, + "loss": 0.5652, + "step": 21625 + }, + { + "epoch": 0.64, + "learning_rate": 5.947290727034751e-05, + "loss": 0.5193, + "step": 21626 + }, + { + "epoch": 0.64, + "learning_rate": 5.946409240271278e-05, + "loss": 0.4707, + "step": 21627 + }, + { + "epoch": 0.64, + "learning_rate": 5.945527791196472e-05, + "loss": 0.5958, + "step": 21628 + }, + { + "epoch": 0.64, + "learning_rate": 5.9446463798185356e-05, + "loss": 0.5538, + "step": 21629 + }, + { + "epoch": 0.64, + "learning_rate": 5.943765006145658e-05, + "loss": 0.4909, + "step": 21630 + }, + { + "epoch": 0.64, + "learning_rate": 5.942883670186041e-05, + "loss": 0.5468, + "step": 21631 + }, + { + "epoch": 0.64, + "learning_rate": 5.942002371947872e-05, + "loss": 0.5086, + "step": 21632 + }, + { + "epoch": 0.64, + "learning_rate": 5.9411211114393495e-05, + "loss": 0.6084, + "step": 21633 + }, + { + "epoch": 0.64, + "learning_rate": 5.940239888668662e-05, + "loss": 0.5066, + "step": 21634 + }, + { + "epoch": 0.64, + "learning_rate": 5.9393587036440094e-05, + "loss": 0.6015, + "step": 21635 + }, + { + "epoch": 0.64, + "learning_rate": 5.9384775563735784e-05, + "loss": 0.4559, + "step": 21636 + }, + { + "epoch": 0.64, + "learning_rate": 5.937596446865564e-05, + "loss": 0.5371, + "step": 21637 + }, + { + "epoch": 0.64, + "learning_rate": 5.93671537512816e-05, + "loss": 0.4978, + "step": 21638 + }, + { + "epoch": 0.64, + "learning_rate": 5.9358343411695595e-05, + "loss": 0.5709, + "step": 21639 + }, + { + "epoch": 0.64, + "learning_rate": 5.9349533449979465e-05, + "loss": 0.4306, + "step": 21640 + }, + { + "epoch": 0.64, + "learning_rate": 5.934072386621521e-05, + "loss": 0.487, + "step": 21641 + }, + { + "epoch": 0.64, + "learning_rate": 5.9331914660484666e-05, + "loss": 0.5528, + "step": 21642 + }, + { + "epoch": 0.64, + "learning_rate": 5.932310583286983e-05, + "loss": 0.5236, + "step": 21643 + }, + { + "epoch": 0.64, + "learning_rate": 5.931429738345249e-05, + "loss": 0.6165, + "step": 21644 + }, + { + "epoch": 0.64, + "learning_rate": 5.9305489312314614e-05, + "loss": 0.5033, + "step": 21645 + }, + { + "epoch": 0.64, + "learning_rate": 5.929668161953805e-05, + "loss": 0.6002, + "step": 21646 + }, + { + "epoch": 0.64, + "learning_rate": 5.928787430520477e-05, + "loss": 0.5456, + "step": 21647 + }, + { + "epoch": 0.64, + "learning_rate": 5.927906736939658e-05, + "loss": 0.6177, + "step": 21648 + }, + { + "epoch": 0.64, + "learning_rate": 5.9270260812195374e-05, + "loss": 0.6913, + "step": 21649 + }, + { + "epoch": 0.64, + "learning_rate": 5.9261454633683064e-05, + "loss": 0.5345, + "step": 21650 + }, + { + "epoch": 0.64, + "learning_rate": 5.9252648833941545e-05, + "loss": 0.5705, + "step": 21651 + }, + { + "epoch": 0.64, + "learning_rate": 5.9243843413052616e-05, + "loss": 0.5122, + "step": 21652 + }, + { + "epoch": 0.64, + "learning_rate": 5.923503837109821e-05, + "loss": 0.4597, + "step": 21653 + }, + { + "epoch": 0.64, + "learning_rate": 5.922623370816015e-05, + "loss": 0.5997, + "step": 21654 + }, + { + "epoch": 0.64, + "learning_rate": 5.921742942432036e-05, + "loss": 0.5955, + "step": 21655 + }, + { + "epoch": 0.64, + "learning_rate": 5.92086255196606e-05, + "loss": 0.4951, + "step": 21656 + }, + { + "epoch": 0.64, + "learning_rate": 5.919982199426282e-05, + "loss": 0.5763, + "step": 21657 + }, + { + "epoch": 0.64, + "learning_rate": 5.919101884820884e-05, + "loss": 0.4822, + "step": 21658 + }, + { + "epoch": 0.64, + "learning_rate": 5.9182216081580455e-05, + "loss": 0.4076, + "step": 21659 + }, + { + "epoch": 0.64, + "learning_rate": 5.9173413694459586e-05, + "loss": 0.5697, + "step": 21660 + }, + { + "epoch": 0.64, + "learning_rate": 5.916461168692803e-05, + "loss": 0.5202, + "step": 21661 + }, + { + "epoch": 0.64, + "learning_rate": 5.9155810059067674e-05, + "loss": 0.5226, + "step": 21662 + }, + { + "epoch": 0.64, + "learning_rate": 5.914700881096029e-05, + "loss": 0.5719, + "step": 21663 + }, + { + "epoch": 0.64, + "learning_rate": 5.913820794268775e-05, + "loss": 0.5558, + "step": 21664 + }, + { + "epoch": 0.65, + "learning_rate": 5.912940745433184e-05, + "loss": 0.4633, + "step": 21665 + }, + { + "epoch": 0.65, + "learning_rate": 5.9120607345974446e-05, + "loss": 0.4755, + "step": 21666 + }, + { + "epoch": 0.65, + "learning_rate": 5.9111807617697345e-05, + "loss": 0.405, + "step": 21667 + }, + { + "epoch": 0.65, + "learning_rate": 5.9103008269582325e-05, + "loss": 0.6322, + "step": 21668 + }, + { + "epoch": 0.65, + "learning_rate": 5.909420930171126e-05, + "loss": 0.5238, + "step": 21669 + }, + { + "epoch": 0.65, + "learning_rate": 5.908541071416596e-05, + "loss": 0.5468, + "step": 21670 + }, + { + "epoch": 0.65, + "learning_rate": 5.9076612507028165e-05, + "loss": 0.4076, + "step": 21671 + }, + { + "epoch": 0.65, + "learning_rate": 5.906781468037972e-05, + "loss": 0.5383, + "step": 21672 + }, + { + "epoch": 0.65, + "learning_rate": 5.90590172343024e-05, + "loss": 0.5679, + "step": 21673 + }, + { + "epoch": 0.65, + "learning_rate": 5.90502201688781e-05, + "loss": 0.464, + "step": 21674 + }, + { + "epoch": 0.65, + "learning_rate": 5.9041423484188464e-05, + "loss": 0.4908, + "step": 21675 + }, + { + "epoch": 0.65, + "learning_rate": 5.903262718031537e-05, + "loss": 0.4763, + "step": 21676 + }, + { + "epoch": 0.65, + "learning_rate": 5.902383125734054e-05, + "loss": 0.6016, + "step": 21677 + }, + { + "epoch": 0.65, + "learning_rate": 5.901503571534586e-05, + "loss": 0.4638, + "step": 21678 + }, + { + "epoch": 0.65, + "learning_rate": 5.9006240554413015e-05, + "loss": 0.4735, + "step": 21679 + }, + { + "epoch": 0.65, + "learning_rate": 5.899744577462378e-05, + "loss": 0.466, + "step": 21680 + }, + { + "epoch": 0.65, + "learning_rate": 5.898865137605999e-05, + "loss": 0.5384, + "step": 21681 + }, + { + "epoch": 0.65, + "learning_rate": 5.8979857358803383e-05, + "loss": 0.5194, + "step": 21682 + }, + { + "epoch": 0.65, + "learning_rate": 5.897106372293567e-05, + "loss": 0.4701, + "step": 21683 + }, + { + "epoch": 0.65, + "learning_rate": 5.896227046853868e-05, + "loss": 0.5147, + "step": 21684 + }, + { + "epoch": 0.65, + "learning_rate": 5.8953477595694115e-05, + "loss": 0.5022, + "step": 21685 + }, + { + "epoch": 0.65, + "learning_rate": 5.8944685104483836e-05, + "loss": 0.5725, + "step": 21686 + }, + { + "epoch": 0.65, + "learning_rate": 5.893589299498943e-05, + "loss": 0.5011, + "step": 21687 + }, + { + "epoch": 0.65, + "learning_rate": 5.892710126729275e-05, + "loss": 0.5105, + "step": 21688 + }, + { + "epoch": 0.65, + "learning_rate": 5.8918309921475535e-05, + "loss": 0.5361, + "step": 21689 + }, + { + "epoch": 0.65, + "learning_rate": 5.890951895761946e-05, + "loss": 0.5335, + "step": 21690 + }, + { + "epoch": 0.65, + "learning_rate": 5.890072837580631e-05, + "loss": 0.5102, + "step": 21691 + }, + { + "epoch": 0.65, + "learning_rate": 5.88919381761178e-05, + "loss": 0.6183, + "step": 21692 + }, + { + "epoch": 0.65, + "learning_rate": 5.888314835863571e-05, + "loss": 0.5253, + "step": 21693 + }, + { + "epoch": 0.65, + "learning_rate": 5.887435892344169e-05, + "loss": 0.6125, + "step": 21694 + }, + { + "epoch": 0.65, + "learning_rate": 5.8865569870617485e-05, + "loss": 0.506, + "step": 21695 + }, + { + "epoch": 0.65, + "learning_rate": 5.8856781200244806e-05, + "loss": 0.4568, + "step": 21696 + }, + { + "epoch": 0.65, + "learning_rate": 5.884799291240541e-05, + "loss": 0.4902, + "step": 21697 + }, + { + "epoch": 0.65, + "learning_rate": 5.883920500718097e-05, + "loss": 0.5339, + "step": 21698 + }, + { + "epoch": 0.65, + "learning_rate": 5.8830417484653155e-05, + "loss": 0.5747, + "step": 21699 + }, + { + "epoch": 0.65, + "learning_rate": 5.8821630344903734e-05, + "loss": 0.5253, + "step": 21700 + }, + { + "epoch": 0.65, + "learning_rate": 5.8812843588014424e-05, + "loss": 0.5789, + "step": 21701 + }, + { + "epoch": 0.65, + "learning_rate": 5.880405721406681e-05, + "loss": 0.4874, + "step": 21702 + }, + { + "epoch": 0.65, + "learning_rate": 5.879527122314268e-05, + "loss": 0.5004, + "step": 21703 + }, + { + "epoch": 0.65, + "learning_rate": 5.878648561532369e-05, + "loss": 0.6002, + "step": 21704 + }, + { + "epoch": 0.65, + "learning_rate": 5.877770039069158e-05, + "loss": 0.5137, + "step": 21705 + }, + { + "epoch": 0.65, + "learning_rate": 5.876891554932793e-05, + "loss": 0.5893, + "step": 21706 + }, + { + "epoch": 0.65, + "learning_rate": 5.876013109131447e-05, + "loss": 0.4575, + "step": 21707 + }, + { + "epoch": 0.65, + "learning_rate": 5.8751347016732875e-05, + "loss": 0.5003, + "step": 21708 + }, + { + "epoch": 0.65, + "learning_rate": 5.8742563325664843e-05, + "loss": 0.5346, + "step": 21709 + }, + { + "epoch": 0.65, + "learning_rate": 5.8733780018192006e-05, + "loss": 0.5307, + "step": 21710 + }, + { + "epoch": 0.65, + "learning_rate": 5.8724997094396004e-05, + "loss": 0.4494, + "step": 21711 + }, + { + "epoch": 0.65, + "learning_rate": 5.8716214554358565e-05, + "loss": 0.5792, + "step": 21712 + }, + { + "epoch": 0.65, + "learning_rate": 5.870743239816131e-05, + "loss": 0.7093, + "step": 21713 + }, + { + "epoch": 0.65, + "learning_rate": 5.8698650625885856e-05, + "loss": 0.5413, + "step": 21714 + }, + { + "epoch": 0.65, + "learning_rate": 5.86898692376139e-05, + "loss": 0.4595, + "step": 21715 + }, + { + "epoch": 0.65, + "learning_rate": 5.868108823342707e-05, + "loss": 0.516, + "step": 21716 + }, + { + "epoch": 0.65, + "learning_rate": 5.8672307613407074e-05, + "loss": 0.5133, + "step": 21717 + }, + { + "epoch": 0.65, + "learning_rate": 5.866352737763543e-05, + "loss": 0.5369, + "step": 21718 + }, + { + "epoch": 0.65, + "learning_rate": 5.865474752619384e-05, + "loss": 0.5784, + "step": 21719 + }, + { + "epoch": 0.65, + "learning_rate": 5.8645968059163916e-05, + "loss": 0.4873, + "step": 21720 + }, + { + "epoch": 0.65, + "learning_rate": 5.863718897662733e-05, + "loss": 0.5142, + "step": 21721 + }, + { + "epoch": 0.65, + "learning_rate": 5.862841027866568e-05, + "loss": 0.6207, + "step": 21722 + }, + { + "epoch": 0.65, + "learning_rate": 5.861963196536053e-05, + "loss": 0.6279, + "step": 21723 + }, + { + "epoch": 0.65, + "learning_rate": 5.861085403679362e-05, + "loss": 0.5028, + "step": 21724 + }, + { + "epoch": 0.65, + "learning_rate": 5.8602076493046454e-05, + "loss": 0.5579, + "step": 21725 + }, + { + "epoch": 0.65, + "learning_rate": 5.859329933420067e-05, + "loss": 0.6218, + "step": 21726 + }, + { + "epoch": 0.65, + "learning_rate": 5.8584522560337904e-05, + "loss": 0.4615, + "step": 21727 + }, + { + "epoch": 0.65, + "learning_rate": 5.857574617153976e-05, + "loss": 0.5823, + "step": 21728 + }, + { + "epoch": 0.65, + "learning_rate": 5.856697016788779e-05, + "loss": 0.5224, + "step": 21729 + }, + { + "epoch": 0.65, + "learning_rate": 5.8558194549463606e-05, + "loss": 0.5435, + "step": 21730 + }, + { + "epoch": 0.65, + "learning_rate": 5.854941931634883e-05, + "loss": 0.6073, + "step": 21731 + }, + { + "epoch": 0.65, + "learning_rate": 5.8540644468625064e-05, + "loss": 0.5387, + "step": 21732 + }, + { + "epoch": 0.65, + "learning_rate": 5.85318700063738e-05, + "loss": 0.5103, + "step": 21733 + }, + { + "epoch": 0.65, + "learning_rate": 5.852309592967671e-05, + "loss": 0.5219, + "step": 21734 + }, + { + "epoch": 0.65, + "learning_rate": 5.8514322238615324e-05, + "loss": 0.555, + "step": 21735 + }, + { + "epoch": 0.65, + "learning_rate": 5.8505548933271285e-05, + "loss": 0.4971, + "step": 21736 + }, + { + "epoch": 0.65, + "learning_rate": 5.849677601372609e-05, + "loss": 0.4951, + "step": 21737 + }, + { + "epoch": 0.65, + "learning_rate": 5.848800348006132e-05, + "loss": 0.5292, + "step": 21738 + }, + { + "epoch": 0.65, + "learning_rate": 5.847923133235853e-05, + "loss": 0.5381, + "step": 21739 + }, + { + "epoch": 0.65, + "learning_rate": 5.847045957069934e-05, + "loss": 0.4651, + "step": 21740 + }, + { + "epoch": 0.65, + "learning_rate": 5.846168819516525e-05, + "loss": 0.5062, + "step": 21741 + }, + { + "epoch": 0.65, + "learning_rate": 5.845291720583781e-05, + "loss": 0.5604, + "step": 21742 + }, + { + "epoch": 0.65, + "learning_rate": 5.84441466027986e-05, + "loss": 0.4836, + "step": 21743 + }, + { + "epoch": 0.65, + "learning_rate": 5.843537638612917e-05, + "loss": 0.3741, + "step": 21744 + }, + { + "epoch": 0.65, + "learning_rate": 5.8426606555911014e-05, + "loss": 0.617, + "step": 21745 + }, + { + "epoch": 0.65, + "learning_rate": 5.8417837112225716e-05, + "loss": 0.4799, + "step": 21746 + }, + { + "epoch": 0.65, + "learning_rate": 5.840906805515477e-05, + "loss": 0.6941, + "step": 21747 + }, + { + "epoch": 0.65, + "learning_rate": 5.84002993847798e-05, + "loss": 0.5459, + "step": 21748 + }, + { + "epoch": 0.65, + "learning_rate": 5.839153110118221e-05, + "loss": 0.5516, + "step": 21749 + }, + { + "epoch": 0.65, + "learning_rate": 5.838276320444357e-05, + "loss": 0.5215, + "step": 21750 + }, + { + "epoch": 0.65, + "learning_rate": 5.837399569464546e-05, + "loss": 0.5828, + "step": 21751 + }, + { + "epoch": 0.65, + "learning_rate": 5.836522857186933e-05, + "loss": 0.4631, + "step": 21752 + }, + { + "epoch": 0.65, + "learning_rate": 5.835646183619669e-05, + "loss": 0.5041, + "step": 21753 + }, + { + "epoch": 0.65, + "learning_rate": 5.8347695487709066e-05, + "loss": 0.5714, + "step": 21754 + }, + { + "epoch": 0.65, + "learning_rate": 5.8338929526488e-05, + "loss": 0.4115, + "step": 21755 + }, + { + "epoch": 0.65, + "learning_rate": 5.8330163952614925e-05, + "loss": 0.4596, + "step": 21756 + }, + { + "epoch": 0.65, + "learning_rate": 5.832139876617142e-05, + "loss": 0.618, + "step": 21757 + }, + { + "epoch": 0.65, + "learning_rate": 5.831263396723891e-05, + "loss": 0.5464, + "step": 21758 + }, + { + "epoch": 0.65, + "learning_rate": 5.830386955589894e-05, + "loss": 0.4803, + "step": 21759 + }, + { + "epoch": 0.65, + "learning_rate": 5.829510553223294e-05, + "loss": 0.5021, + "step": 21760 + }, + { + "epoch": 0.65, + "learning_rate": 5.828634189632245e-05, + "loss": 0.5626, + "step": 21761 + }, + { + "epoch": 0.65, + "learning_rate": 5.827757864824891e-05, + "loss": 0.522, + "step": 21762 + }, + { + "epoch": 0.65, + "learning_rate": 5.826881578809384e-05, + "loss": 0.5531, + "step": 21763 + }, + { + "epoch": 0.65, + "learning_rate": 5.8260053315938654e-05, + "loss": 0.5491, + "step": 21764 + }, + { + "epoch": 0.65, + "learning_rate": 5.825129123186486e-05, + "loss": 0.5436, + "step": 21765 + }, + { + "epoch": 0.65, + "learning_rate": 5.824252953595395e-05, + "loss": 0.5343, + "step": 21766 + }, + { + "epoch": 0.65, + "learning_rate": 5.8233768228287364e-05, + "loss": 0.5443, + "step": 21767 + }, + { + "epoch": 0.65, + "learning_rate": 5.8225007308946524e-05, + "loss": 0.5686, + "step": 21768 + }, + { + "epoch": 0.65, + "learning_rate": 5.821624677801291e-05, + "loss": 0.7011, + "step": 21769 + }, + { + "epoch": 0.65, + "learning_rate": 5.820748663556801e-05, + "loss": 0.6227, + "step": 21770 + }, + { + "epoch": 0.65, + "learning_rate": 5.819872688169326e-05, + "loss": 0.4449, + "step": 21771 + }, + { + "epoch": 0.65, + "learning_rate": 5.818996751647003e-05, + "loss": 0.6379, + "step": 21772 + }, + { + "epoch": 0.65, + "learning_rate": 5.818120853997982e-05, + "loss": 0.6003, + "step": 21773 + }, + { + "epoch": 0.65, + "learning_rate": 5.8172449952304065e-05, + "loss": 0.4964, + "step": 21774 + }, + { + "epoch": 0.65, + "learning_rate": 5.816369175352425e-05, + "loss": 0.4391, + "step": 21775 + }, + { + "epoch": 0.65, + "learning_rate": 5.8154933943721756e-05, + "loss": 0.5752, + "step": 21776 + }, + { + "epoch": 0.65, + "learning_rate": 5.814617652297796e-05, + "loss": 0.6164, + "step": 21777 + }, + { + "epoch": 0.65, + "learning_rate": 5.813741949137433e-05, + "loss": 0.6139, + "step": 21778 + }, + { + "epoch": 0.65, + "learning_rate": 5.8128662848992345e-05, + "loss": 0.6418, + "step": 21779 + }, + { + "epoch": 0.65, + "learning_rate": 5.8119906595913356e-05, + "loss": 0.5169, + "step": 21780 + }, + { + "epoch": 0.65, + "learning_rate": 5.811115073221873e-05, + "loss": 0.4246, + "step": 21781 + }, + { + "epoch": 0.65, + "learning_rate": 5.810239525798995e-05, + "loss": 0.6351, + "step": 21782 + }, + { + "epoch": 0.65, + "learning_rate": 5.809364017330843e-05, + "loss": 0.4978, + "step": 21783 + }, + { + "epoch": 0.65, + "learning_rate": 5.80848854782555e-05, + "loss": 0.6407, + "step": 21784 + }, + { + "epoch": 0.65, + "learning_rate": 5.807613117291264e-05, + "loss": 0.592, + "step": 21785 + }, + { + "epoch": 0.65, + "learning_rate": 5.8067377257361165e-05, + "loss": 0.54, + "step": 21786 + }, + { + "epoch": 0.65, + "learning_rate": 5.805862373168254e-05, + "loss": 0.5486, + "step": 21787 + }, + { + "epoch": 0.65, + "learning_rate": 5.8049870595958076e-05, + "loss": 0.4976, + "step": 21788 + }, + { + "epoch": 0.65, + "learning_rate": 5.80411178502692e-05, + "loss": 0.4716, + "step": 21789 + }, + { + "epoch": 0.65, + "learning_rate": 5.803236549469737e-05, + "loss": 0.5716, + "step": 21790 + }, + { + "epoch": 0.65, + "learning_rate": 5.80236135293238e-05, + "loss": 0.4601, + "step": 21791 + }, + { + "epoch": 0.65, + "learning_rate": 5.8014861954229926e-05, + "loss": 0.4532, + "step": 21792 + }, + { + "epoch": 0.65, + "learning_rate": 5.8006110769497145e-05, + "loss": 0.4951, + "step": 21793 + }, + { + "epoch": 0.65, + "learning_rate": 5.799735997520684e-05, + "loss": 0.5393, + "step": 21794 + }, + { + "epoch": 0.65, + "learning_rate": 5.798860957144034e-05, + "loss": 0.521, + "step": 21795 + }, + { + "epoch": 0.65, + "learning_rate": 5.797985955827898e-05, + "loss": 0.5065, + "step": 21796 + }, + { + "epoch": 0.65, + "learning_rate": 5.7971109935804124e-05, + "loss": 0.4986, + "step": 21797 + }, + { + "epoch": 0.65, + "learning_rate": 5.79623607040972e-05, + "loss": 0.4318, + "step": 21798 + }, + { + "epoch": 0.65, + "learning_rate": 5.795361186323942e-05, + "loss": 0.5096, + "step": 21799 + }, + { + "epoch": 0.65, + "learning_rate": 5.7944863413312254e-05, + "loss": 0.4778, + "step": 21800 + }, + { + "epoch": 0.65, + "learning_rate": 5.793611535439695e-05, + "loss": 0.5376, + "step": 21801 + }, + { + "epoch": 0.65, + "learning_rate": 5.7927367686574915e-05, + "loss": 0.5701, + "step": 21802 + }, + { + "epoch": 0.65, + "learning_rate": 5.79186204099274e-05, + "loss": 0.5477, + "step": 21803 + }, + { + "epoch": 0.65, + "learning_rate": 5.790987352453583e-05, + "loss": 0.5123, + "step": 21804 + }, + { + "epoch": 0.65, + "learning_rate": 5.790112703048144e-05, + "loss": 0.586, + "step": 21805 + }, + { + "epoch": 0.65, + "learning_rate": 5.789238092784564e-05, + "loss": 0.4605, + "step": 21806 + }, + { + "epoch": 0.65, + "learning_rate": 5.788363521670963e-05, + "loss": 0.5524, + "step": 21807 + }, + { + "epoch": 0.65, + "learning_rate": 5.7874889897154814e-05, + "loss": 0.5588, + "step": 21808 + }, + { + "epoch": 0.65, + "learning_rate": 5.7866144969262524e-05, + "loss": 0.505, + "step": 21809 + }, + { + "epoch": 0.65, + "learning_rate": 5.785740043311403e-05, + "loss": 0.5484, + "step": 21810 + }, + { + "epoch": 0.65, + "learning_rate": 5.7848656288790585e-05, + "loss": 0.5228, + "step": 21811 + }, + { + "epoch": 0.65, + "learning_rate": 5.783991253637353e-05, + "loss": 0.5616, + "step": 21812 + }, + { + "epoch": 0.65, + "learning_rate": 5.78311691759442e-05, + "loss": 0.512, + "step": 21813 + }, + { + "epoch": 0.65, + "learning_rate": 5.782242620758387e-05, + "loss": 0.6836, + "step": 21814 + }, + { + "epoch": 0.65, + "learning_rate": 5.781368363137376e-05, + "loss": 0.508, + "step": 21815 + }, + { + "epoch": 0.65, + "learning_rate": 5.78049414473952e-05, + "loss": 0.5682, + "step": 21816 + }, + { + "epoch": 0.65, + "learning_rate": 5.779619965572949e-05, + "loss": 0.5491, + "step": 21817 + }, + { + "epoch": 0.65, + "learning_rate": 5.778745825645793e-05, + "loss": 0.5107, + "step": 21818 + }, + { + "epoch": 0.65, + "learning_rate": 5.777871724966175e-05, + "loss": 0.6146, + "step": 21819 + }, + { + "epoch": 0.65, + "learning_rate": 5.7769976635422194e-05, + "loss": 0.5787, + "step": 21820 + }, + { + "epoch": 0.65, + "learning_rate": 5.7761236413820606e-05, + "loss": 0.4559, + "step": 21821 + }, + { + "epoch": 0.65, + "learning_rate": 5.7752496584938174e-05, + "loss": 0.4662, + "step": 21822 + }, + { + "epoch": 0.65, + "learning_rate": 5.7743757148856225e-05, + "loss": 0.4922, + "step": 21823 + }, + { + "epoch": 0.65, + "learning_rate": 5.773501810565593e-05, + "loss": 0.6005, + "step": 21824 + }, + { + "epoch": 0.65, + "learning_rate": 5.772627945541864e-05, + "loss": 0.5138, + "step": 21825 + }, + { + "epoch": 0.65, + "learning_rate": 5.771754119822552e-05, + "loss": 0.5106, + "step": 21826 + }, + { + "epoch": 0.65, + "learning_rate": 5.7708803334157846e-05, + "loss": 0.5456, + "step": 21827 + }, + { + "epoch": 0.65, + "learning_rate": 5.7700065863296906e-05, + "loss": 0.4788, + "step": 21828 + }, + { + "epoch": 0.65, + "learning_rate": 5.769132878572389e-05, + "loss": 0.5348, + "step": 21829 + }, + { + "epoch": 0.65, + "learning_rate": 5.7682592101519996e-05, + "loss": 0.4747, + "step": 21830 + }, + { + "epoch": 0.65, + "learning_rate": 5.76738558107665e-05, + "loss": 0.4652, + "step": 21831 + }, + { + "epoch": 0.65, + "learning_rate": 5.7665119913544654e-05, + "loss": 0.4752, + "step": 21832 + }, + { + "epoch": 0.65, + "learning_rate": 5.765638440993566e-05, + "loss": 0.5191, + "step": 21833 + }, + { + "epoch": 0.65, + "learning_rate": 5.764764930002067e-05, + "loss": 0.6523, + "step": 21834 + }, + { + "epoch": 0.65, + "learning_rate": 5.763891458388098e-05, + "loss": 0.5694, + "step": 21835 + }, + { + "epoch": 0.65, + "learning_rate": 5.7630180261597765e-05, + "loss": 0.5224, + "step": 21836 + }, + { + "epoch": 0.65, + "learning_rate": 5.762144633325228e-05, + "loss": 0.5241, + "step": 21837 + }, + { + "epoch": 0.65, + "learning_rate": 5.761271279892571e-05, + "loss": 0.6202, + "step": 21838 + }, + { + "epoch": 0.65, + "learning_rate": 5.7603979658699205e-05, + "loss": 0.5608, + "step": 21839 + }, + { + "epoch": 0.65, + "learning_rate": 5.7595246912653985e-05, + "loss": 0.5164, + "step": 21840 + }, + { + "epoch": 0.65, + "learning_rate": 5.7586514560871316e-05, + "loss": 0.5563, + "step": 21841 + }, + { + "epoch": 0.65, + "learning_rate": 5.7577782603432316e-05, + "loss": 0.4132, + "step": 21842 + }, + { + "epoch": 0.65, + "learning_rate": 5.756905104041815e-05, + "loss": 0.4746, + "step": 21843 + }, + { + "epoch": 0.65, + "learning_rate": 5.756031987191004e-05, + "loss": 0.4826, + "step": 21844 + }, + { + "epoch": 0.65, + "learning_rate": 5.755158909798919e-05, + "loss": 0.505, + "step": 21845 + }, + { + "epoch": 0.65, + "learning_rate": 5.754285871873673e-05, + "loss": 0.5469, + "step": 21846 + }, + { + "epoch": 0.65, + "learning_rate": 5.753412873423386e-05, + "loss": 0.5898, + "step": 21847 + }, + { + "epoch": 0.65, + "learning_rate": 5.752539914456171e-05, + "loss": 0.5782, + "step": 21848 + }, + { + "epoch": 0.65, + "learning_rate": 5.7516669949801516e-05, + "loss": 0.6387, + "step": 21849 + }, + { + "epoch": 0.65, + "learning_rate": 5.750794115003434e-05, + "loss": 0.5659, + "step": 21850 + }, + { + "epoch": 0.65, + "learning_rate": 5.749921274534139e-05, + "loss": 0.5474, + "step": 21851 + }, + { + "epoch": 0.65, + "learning_rate": 5.749048473580392e-05, + "loss": 0.4631, + "step": 21852 + }, + { + "epoch": 0.65, + "learning_rate": 5.7481757121502876e-05, + "loss": 0.5985, + "step": 21853 + }, + { + "epoch": 0.65, + "learning_rate": 5.7473029902519524e-05, + "loss": 0.5904, + "step": 21854 + }, + { + "epoch": 0.65, + "learning_rate": 5.746430307893498e-05, + "loss": 0.5674, + "step": 21855 + }, + { + "epoch": 0.65, + "learning_rate": 5.745557665083044e-05, + "loss": 0.5301, + "step": 21856 + }, + { + "epoch": 0.65, + "learning_rate": 5.744685061828699e-05, + "loss": 0.4337, + "step": 21857 + }, + { + "epoch": 0.65, + "learning_rate": 5.743812498138572e-05, + "loss": 0.5427, + "step": 21858 + }, + { + "epoch": 0.65, + "learning_rate": 5.74293997402078e-05, + "loss": 0.5796, + "step": 21859 + }, + { + "epoch": 0.65, + "learning_rate": 5.7420674894834404e-05, + "loss": 0.5215, + "step": 21860 + }, + { + "epoch": 0.65, + "learning_rate": 5.7411950445346554e-05, + "loss": 0.5205, + "step": 21861 + }, + { + "epoch": 0.65, + "learning_rate": 5.740322639182546e-05, + "loss": 0.4696, + "step": 21862 + }, + { + "epoch": 0.65, + "learning_rate": 5.7394502734352154e-05, + "loss": 0.583, + "step": 21863 + }, + { + "epoch": 0.65, + "learning_rate": 5.738577947300781e-05, + "loss": 0.5358, + "step": 21864 + }, + { + "epoch": 0.65, + "learning_rate": 5.737705660787347e-05, + "loss": 0.493, + "step": 21865 + }, + { + "epoch": 0.65, + "learning_rate": 5.736833413903031e-05, + "loss": 0.5428, + "step": 21866 + }, + { + "epoch": 0.65, + "learning_rate": 5.735961206655933e-05, + "loss": 0.5515, + "step": 21867 + }, + { + "epoch": 0.65, + "learning_rate": 5.7350890390541745e-05, + "loss": 0.5769, + "step": 21868 + }, + { + "epoch": 0.65, + "learning_rate": 5.734216911105854e-05, + "loss": 0.5547, + "step": 21869 + }, + { + "epoch": 0.65, + "learning_rate": 5.7333448228190846e-05, + "loss": 0.5082, + "step": 21870 + }, + { + "epoch": 0.65, + "learning_rate": 5.7324727742019765e-05, + "loss": 0.5175, + "step": 21871 + }, + { + "epoch": 0.65, + "learning_rate": 5.731600765262637e-05, + "loss": 0.4479, + "step": 21872 + }, + { + "epoch": 0.65, + "learning_rate": 5.730728796009169e-05, + "loss": 0.6253, + "step": 21873 + }, + { + "epoch": 0.65, + "learning_rate": 5.729856866449683e-05, + "loss": 0.475, + "step": 21874 + }, + { + "epoch": 0.65, + "learning_rate": 5.728984976592289e-05, + "loss": 0.543, + "step": 21875 + }, + { + "epoch": 0.65, + "learning_rate": 5.728113126445092e-05, + "loss": 0.6245, + "step": 21876 + }, + { + "epoch": 0.65, + "learning_rate": 5.727241316016191e-05, + "loss": 0.515, + "step": 21877 + }, + { + "epoch": 0.65, + "learning_rate": 5.726369545313698e-05, + "loss": 0.3403, + "step": 21878 + }, + { + "epoch": 0.65, + "learning_rate": 5.725497814345717e-05, + "loss": 0.5624, + "step": 21879 + }, + { + "epoch": 0.65, + "learning_rate": 5.724626123120358e-05, + "loss": 0.472, + "step": 21880 + }, + { + "epoch": 0.65, + "learning_rate": 5.723754471645722e-05, + "loss": 0.5456, + "step": 21881 + }, + { + "epoch": 0.65, + "learning_rate": 5.722882859929908e-05, + "loss": 0.5116, + "step": 21882 + }, + { + "epoch": 0.65, + "learning_rate": 5.722011287981024e-05, + "loss": 0.4931, + "step": 21883 + }, + { + "epoch": 0.65, + "learning_rate": 5.721139755807179e-05, + "loss": 0.4904, + "step": 21884 + }, + { + "epoch": 0.65, + "learning_rate": 5.720268263416471e-05, + "loss": 0.4677, + "step": 21885 + }, + { + "epoch": 0.65, + "learning_rate": 5.7193968108169984e-05, + "loss": 0.5386, + "step": 21886 + }, + { + "epoch": 0.65, + "learning_rate": 5.718525398016872e-05, + "loss": 0.5164, + "step": 21887 + }, + { + "epoch": 0.65, + "learning_rate": 5.7176540250241864e-05, + "loss": 0.6248, + "step": 21888 + }, + { + "epoch": 0.65, + "learning_rate": 5.716782691847047e-05, + "loss": 0.4714, + "step": 21889 + }, + { + "epoch": 0.65, + "learning_rate": 5.7159113984935585e-05, + "loss": 0.5395, + "step": 21890 + }, + { + "epoch": 0.65, + "learning_rate": 5.7150401449718195e-05, + "loss": 0.457, + "step": 21891 + }, + { + "epoch": 0.65, + "learning_rate": 5.714168931289925e-05, + "loss": 0.5703, + "step": 21892 + }, + { + "epoch": 0.65, + "learning_rate": 5.713297757455979e-05, + "loss": 0.4358, + "step": 21893 + }, + { + "epoch": 0.65, + "learning_rate": 5.712426623478087e-05, + "loss": 0.5183, + "step": 21894 + }, + { + "epoch": 0.65, + "learning_rate": 5.711555529364341e-05, + "loss": 0.5646, + "step": 21895 + }, + { + "epoch": 0.65, + "learning_rate": 5.7106844751228405e-05, + "loss": 0.6264, + "step": 21896 + }, + { + "epoch": 0.65, + "learning_rate": 5.709813460761686e-05, + "loss": 0.4918, + "step": 21897 + }, + { + "epoch": 0.65, + "learning_rate": 5.708942486288975e-05, + "loss": 0.5308, + "step": 21898 + }, + { + "epoch": 0.65, + "learning_rate": 5.708071551712813e-05, + "loss": 0.5223, + "step": 21899 + }, + { + "epoch": 0.65, + "learning_rate": 5.707200657041289e-05, + "loss": 0.6092, + "step": 21900 + }, + { + "epoch": 0.65, + "learning_rate": 5.7063298022824984e-05, + "loss": 0.5718, + "step": 21901 + }, + { + "epoch": 0.65, + "learning_rate": 5.705458987444543e-05, + "loss": 0.5613, + "step": 21902 + }, + { + "epoch": 0.65, + "learning_rate": 5.7045882125355224e-05, + "loss": 0.5757, + "step": 21903 + }, + { + "epoch": 0.65, + "learning_rate": 5.703717477563527e-05, + "loss": 0.5878, + "step": 21904 + }, + { + "epoch": 0.65, + "learning_rate": 5.70284678253665e-05, + "loss": 0.7255, + "step": 21905 + }, + { + "epoch": 0.65, + "learning_rate": 5.7019761274629934e-05, + "loss": 0.459, + "step": 21906 + }, + { + "epoch": 0.65, + "learning_rate": 5.701105512350653e-05, + "loss": 0.534, + "step": 21907 + }, + { + "epoch": 0.65, + "learning_rate": 5.7002349372077154e-05, + "loss": 0.4684, + "step": 21908 + }, + { + "epoch": 0.65, + "learning_rate": 5.6993644020422835e-05, + "loss": 0.496, + "step": 21909 + }, + { + "epoch": 0.65, + "learning_rate": 5.698493906862443e-05, + "loss": 0.5704, + "step": 21910 + }, + { + "epoch": 0.65, + "learning_rate": 5.6976234516762975e-05, + "loss": 0.5128, + "step": 21911 + }, + { + "epoch": 0.65, + "learning_rate": 5.6967530364919285e-05, + "loss": 0.4984, + "step": 21912 + }, + { + "epoch": 0.65, + "learning_rate": 5.695882661317435e-05, + "loss": 0.507, + "step": 21913 + }, + { + "epoch": 0.65, + "learning_rate": 5.6950123261609146e-05, + "loss": 0.4533, + "step": 21914 + }, + { + "epoch": 0.65, + "learning_rate": 5.6941420310304514e-05, + "loss": 0.672, + "step": 21915 + }, + { + "epoch": 0.65, + "learning_rate": 5.693271775934137e-05, + "loss": 0.5571, + "step": 21916 + }, + { + "epoch": 0.65, + "learning_rate": 5.692401560880065e-05, + "loss": 0.5193, + "step": 21917 + }, + { + "epoch": 0.65, + "learning_rate": 5.6915313858763306e-05, + "loss": 0.4924, + "step": 21918 + }, + { + "epoch": 0.65, + "learning_rate": 5.69066125093102e-05, + "loss": 0.5278, + "step": 21919 + }, + { + "epoch": 0.65, + "learning_rate": 5.6897911560522196e-05, + "loss": 0.5026, + "step": 21920 + }, + { + "epoch": 0.65, + "learning_rate": 5.688921101248024e-05, + "loss": 0.6939, + "step": 21921 + }, + { + "epoch": 0.65, + "learning_rate": 5.688051086526525e-05, + "loss": 0.4984, + "step": 21922 + }, + { + "epoch": 0.65, + "learning_rate": 5.687181111895803e-05, + "loss": 0.5135, + "step": 21923 + }, + { + "epoch": 0.65, + "learning_rate": 5.686311177363958e-05, + "loss": 0.5187, + "step": 21924 + }, + { + "epoch": 0.65, + "learning_rate": 5.685441282939068e-05, + "loss": 0.5594, + "step": 21925 + }, + { + "epoch": 0.65, + "learning_rate": 5.6845714286292286e-05, + "loss": 0.5363, + "step": 21926 + }, + { + "epoch": 0.65, + "learning_rate": 5.6837016144425205e-05, + "loss": 0.6212, + "step": 21927 + }, + { + "epoch": 0.65, + "learning_rate": 5.682831840387039e-05, + "loss": 0.575, + "step": 21928 + }, + { + "epoch": 0.65, + "learning_rate": 5.681962106470862e-05, + "loss": 0.5383, + "step": 21929 + }, + { + "epoch": 0.65, + "learning_rate": 5.681092412702084e-05, + "loss": 0.5933, + "step": 21930 + }, + { + "epoch": 0.65, + "learning_rate": 5.6802227590887836e-05, + "loss": 0.4241, + "step": 21931 + }, + { + "epoch": 0.65, + "learning_rate": 5.67935314563905e-05, + "loss": 0.5282, + "step": 21932 + }, + { + "epoch": 0.65, + "learning_rate": 5.678483572360973e-05, + "loss": 0.5221, + "step": 21933 + }, + { + "epoch": 0.65, + "learning_rate": 5.6776140392626344e-05, + "loss": 0.5557, + "step": 21934 + }, + { + "epoch": 0.65, + "learning_rate": 5.676744546352113e-05, + "loss": 0.4691, + "step": 21935 + }, + { + "epoch": 0.65, + "learning_rate": 5.675875093637497e-05, + "loss": 0.5081, + "step": 21936 + }, + { + "epoch": 0.65, + "learning_rate": 5.675005681126876e-05, + "loss": 0.5549, + "step": 21937 + }, + { + "epoch": 0.65, + "learning_rate": 5.674136308828329e-05, + "loss": 0.434, + "step": 21938 + }, + { + "epoch": 0.65, + "learning_rate": 5.673266976749933e-05, + "loss": 0.5385, + "step": 21939 + }, + { + "epoch": 0.65, + "learning_rate": 5.6723976848997754e-05, + "loss": 0.4308, + "step": 21940 + }, + { + "epoch": 0.65, + "learning_rate": 5.671528433285941e-05, + "loss": 0.6045, + "step": 21941 + }, + { + "epoch": 0.65, + "learning_rate": 5.670659221916514e-05, + "loss": 0.493, + "step": 21942 + }, + { + "epoch": 0.65, + "learning_rate": 5.66979005079957e-05, + "loss": 0.4594, + "step": 21943 + }, + { + "epoch": 0.65, + "learning_rate": 5.668920919943191e-05, + "loss": 0.5678, + "step": 21944 + }, + { + "epoch": 0.65, + "learning_rate": 5.6680518293554585e-05, + "loss": 0.3878, + "step": 21945 + }, + { + "epoch": 0.65, + "learning_rate": 5.6671827790444576e-05, + "loss": 0.4583, + "step": 21946 + }, + { + "epoch": 0.65, + "learning_rate": 5.666313769018263e-05, + "loss": 0.4671, + "step": 21947 + }, + { + "epoch": 0.65, + "learning_rate": 5.6654447992849526e-05, + "loss": 0.4937, + "step": 21948 + }, + { + "epoch": 0.65, + "learning_rate": 5.664575869852609e-05, + "loss": 0.4917, + "step": 21949 + }, + { + "epoch": 0.65, + "learning_rate": 5.663706980729315e-05, + "loss": 0.5736, + "step": 21950 + }, + { + "epoch": 0.65, + "learning_rate": 5.6628381319231406e-05, + "loss": 0.5677, + "step": 21951 + }, + { + "epoch": 0.65, + "learning_rate": 5.661969323442173e-05, + "loss": 0.5612, + "step": 21952 + }, + { + "epoch": 0.65, + "learning_rate": 5.6611005552944854e-05, + "loss": 0.5008, + "step": 21953 + }, + { + "epoch": 0.65, + "learning_rate": 5.660231827488152e-05, + "loss": 0.535, + "step": 21954 + }, + { + "epoch": 0.65, + "learning_rate": 5.659363140031253e-05, + "loss": 0.4705, + "step": 21955 + }, + { + "epoch": 0.65, + "learning_rate": 5.658494492931866e-05, + "loss": 0.5951, + "step": 21956 + }, + { + "epoch": 0.65, + "learning_rate": 5.657625886198075e-05, + "loss": 0.6147, + "step": 21957 + }, + { + "epoch": 0.65, + "learning_rate": 5.6567573198379395e-05, + "loss": 0.6104, + "step": 21958 + }, + { + "epoch": 0.65, + "learning_rate": 5.655888793859543e-05, + "loss": 0.5052, + "step": 21959 + }, + { + "epoch": 0.65, + "learning_rate": 5.655020308270962e-05, + "loss": 0.5744, + "step": 21960 + }, + { + "epoch": 0.65, + "learning_rate": 5.654151863080274e-05, + "loss": 0.5452, + "step": 21961 + }, + { + "epoch": 0.65, + "learning_rate": 5.65328345829555e-05, + "loss": 0.6474, + "step": 21962 + }, + { + "epoch": 0.65, + "learning_rate": 5.652415093924859e-05, + "loss": 0.5132, + "step": 21963 + }, + { + "epoch": 0.65, + "learning_rate": 5.651546769976281e-05, + "loss": 0.405, + "step": 21964 + }, + { + "epoch": 0.65, + "learning_rate": 5.650678486457891e-05, + "loss": 0.646, + "step": 21965 + }, + { + "epoch": 0.65, + "learning_rate": 5.649810243377756e-05, + "loss": 0.4967, + "step": 21966 + }, + { + "epoch": 0.65, + "learning_rate": 5.648942040743954e-05, + "loss": 0.4862, + "step": 21967 + }, + { + "epoch": 0.65, + "learning_rate": 5.648073878564553e-05, + "loss": 0.4289, + "step": 21968 + }, + { + "epoch": 0.65, + "learning_rate": 5.647205756847629e-05, + "loss": 0.5199, + "step": 21969 + }, + { + "epoch": 0.65, + "learning_rate": 5.646337675601247e-05, + "loss": 0.5104, + "step": 21970 + }, + { + "epoch": 0.65, + "learning_rate": 5.645469634833487e-05, + "loss": 0.5579, + "step": 21971 + }, + { + "epoch": 0.65, + "learning_rate": 5.644601634552409e-05, + "loss": 0.4943, + "step": 21972 + }, + { + "epoch": 0.65, + "learning_rate": 5.643733674766094e-05, + "loss": 0.5048, + "step": 21973 + }, + { + "epoch": 0.65, + "learning_rate": 5.6428657554826035e-05, + "loss": 0.6262, + "step": 21974 + }, + { + "epoch": 0.65, + "learning_rate": 5.64199787671001e-05, + "loss": 0.6128, + "step": 21975 + }, + { + "epoch": 0.65, + "learning_rate": 5.641130038456388e-05, + "loss": 0.4848, + "step": 21976 + }, + { + "epoch": 0.65, + "learning_rate": 5.640262240729802e-05, + "loss": 0.4565, + "step": 21977 + }, + { + "epoch": 0.65, + "learning_rate": 5.6393944835383164e-05, + "loss": 0.4723, + "step": 21978 + }, + { + "epoch": 0.65, + "learning_rate": 5.638526766890001e-05, + "loss": 0.5852, + "step": 21979 + }, + { + "epoch": 0.65, + "learning_rate": 5.637659090792931e-05, + "loss": 0.5336, + "step": 21980 + }, + { + "epoch": 0.65, + "learning_rate": 5.636791455255167e-05, + "loss": 0.5153, + "step": 21981 + }, + { + "epoch": 0.65, + "learning_rate": 5.635923860284773e-05, + "loss": 0.6304, + "step": 21982 + }, + { + "epoch": 0.65, + "learning_rate": 5.6350563058898207e-05, + "loss": 0.515, + "step": 21983 + }, + { + "epoch": 0.65, + "learning_rate": 5.634188792078379e-05, + "loss": 0.532, + "step": 21984 + }, + { + "epoch": 0.65, + "learning_rate": 5.6333213188585055e-05, + "loss": 0.4939, + "step": 21985 + }, + { + "epoch": 0.65, + "learning_rate": 5.632453886238274e-05, + "loss": 0.5399, + "step": 21986 + }, + { + "epoch": 0.65, + "learning_rate": 5.631586494225741e-05, + "loss": 0.5227, + "step": 21987 + }, + { + "epoch": 0.65, + "learning_rate": 5.6307191428289806e-05, + "loss": 0.5164, + "step": 21988 + }, + { + "epoch": 0.65, + "learning_rate": 5.629851832056048e-05, + "loss": 0.5466, + "step": 21989 + }, + { + "epoch": 0.65, + "learning_rate": 5.628984561915015e-05, + "loss": 0.5704, + "step": 21990 + }, + { + "epoch": 0.65, + "learning_rate": 5.628117332413938e-05, + "loss": 0.6147, + "step": 21991 + }, + { + "epoch": 0.65, + "learning_rate": 5.627250143560887e-05, + "loss": 0.5397, + "step": 21992 + }, + { + "epoch": 0.65, + "learning_rate": 5.6263829953639166e-05, + "loss": 0.5781, + "step": 21993 + }, + { + "epoch": 0.65, + "learning_rate": 5.6255158878310955e-05, + "loss": 0.5089, + "step": 21994 + }, + { + "epoch": 0.65, + "learning_rate": 5.624648820970487e-05, + "loss": 0.5308, + "step": 21995 + }, + { + "epoch": 0.65, + "learning_rate": 5.6237817947901505e-05, + "loss": 0.5862, + "step": 21996 + }, + { + "epoch": 0.65, + "learning_rate": 5.6229148092981434e-05, + "loss": 0.4836, + "step": 21997 + }, + { + "epoch": 0.65, + "learning_rate": 5.622047864502529e-05, + "loss": 0.4686, + "step": 21998 + }, + { + "epoch": 0.65, + "learning_rate": 5.621180960411372e-05, + "loss": 0.4555, + "step": 21999 + }, + { + "epoch": 0.65, + "learning_rate": 5.620314097032731e-05, + "loss": 0.5138, + "step": 22000 + }, + { + "epoch": 0.66, + "learning_rate": 5.6194472743746585e-05, + "loss": 0.5956, + "step": 22001 + }, + { + "epoch": 0.66, + "learning_rate": 5.6185804924452203e-05, + "loss": 0.6728, + "step": 22002 + }, + { + "epoch": 0.66, + "learning_rate": 5.617713751252474e-05, + "loss": 0.6091, + "step": 22003 + }, + { + "epoch": 0.66, + "learning_rate": 5.616847050804481e-05, + "loss": 0.523, + "step": 22004 + }, + { + "epoch": 0.66, + "learning_rate": 5.615980391109299e-05, + "loss": 0.5381, + "step": 22005 + }, + { + "epoch": 0.66, + "learning_rate": 5.61511377217498e-05, + "loss": 0.5375, + "step": 22006 + }, + { + "epoch": 0.66, + "learning_rate": 5.6142471940095855e-05, + "loss": 0.6017, + "step": 22007 + }, + { + "epoch": 0.66, + "learning_rate": 5.613380656621177e-05, + "loss": 0.4846, + "step": 22008 + }, + { + "epoch": 0.66, + "learning_rate": 5.612514160017806e-05, + "loss": 0.4981, + "step": 22009 + }, + { + "epoch": 0.66, + "learning_rate": 5.611647704207526e-05, + "loss": 0.5489, + "step": 22010 + }, + { + "epoch": 0.66, + "learning_rate": 5.610781289198395e-05, + "loss": 0.5188, + "step": 22011 + }, + { + "epoch": 0.66, + "learning_rate": 5.6099149149984754e-05, + "loss": 0.5489, + "step": 22012 + }, + { + "epoch": 0.66, + "learning_rate": 5.609048581615814e-05, + "loss": 0.4377, + "step": 22013 + }, + { + "epoch": 0.66, + "learning_rate": 5.608182289058472e-05, + "loss": 0.4881, + "step": 22014 + }, + { + "epoch": 0.66, + "learning_rate": 5.607316037334498e-05, + "loss": 0.5665, + "step": 22015 + }, + { + "epoch": 0.66, + "learning_rate": 5.606449826451952e-05, + "loss": 0.5783, + "step": 22016 + }, + { + "epoch": 0.66, + "learning_rate": 5.605583656418879e-05, + "loss": 0.5539, + "step": 22017 + }, + { + "epoch": 0.66, + "learning_rate": 5.6047175272433395e-05, + "loss": 0.5857, + "step": 22018 + }, + { + "epoch": 0.66, + "learning_rate": 5.603851438933392e-05, + "loss": 0.5894, + "step": 22019 + }, + { + "epoch": 0.66, + "learning_rate": 5.602985391497073e-05, + "loss": 0.4946, + "step": 22020 + }, + { + "epoch": 0.66, + "learning_rate": 5.602119384942444e-05, + "loss": 0.5229, + "step": 22021 + }, + { + "epoch": 0.66, + "learning_rate": 5.601253419277556e-05, + "loss": 0.5626, + "step": 22022 + }, + { + "epoch": 0.66, + "learning_rate": 5.600387494510464e-05, + "loss": 0.4764, + "step": 22023 + }, + { + "epoch": 0.66, + "learning_rate": 5.5995216106492166e-05, + "loss": 0.5722, + "step": 22024 + }, + { + "epoch": 0.66, + "learning_rate": 5.5986557677018594e-05, + "loss": 0.6106, + "step": 22025 + }, + { + "epoch": 0.66, + "learning_rate": 5.5977899656764453e-05, + "loss": 0.5115, + "step": 22026 + }, + { + "epoch": 0.66, + "learning_rate": 5.596924204581031e-05, + "loss": 0.4854, + "step": 22027 + }, + { + "epoch": 0.66, + "learning_rate": 5.596058484423656e-05, + "loss": 0.525, + "step": 22028 + }, + { + "epoch": 0.66, + "learning_rate": 5.595192805212379e-05, + "loss": 0.4307, + "step": 22029 + }, + { + "epoch": 0.66, + "learning_rate": 5.5943271669552376e-05, + "loss": 0.5586, + "step": 22030 + }, + { + "epoch": 0.66, + "learning_rate": 5.5934615696602924e-05, + "loss": 0.5119, + "step": 22031 + }, + { + "epoch": 0.66, + "learning_rate": 5.592596013335581e-05, + "loss": 0.5174, + "step": 22032 + }, + { + "epoch": 0.66, + "learning_rate": 5.591730497989158e-05, + "loss": 0.5517, + "step": 22033 + }, + { + "epoch": 0.66, + "learning_rate": 5.590865023629064e-05, + "loss": 0.5314, + "step": 22034 + }, + { + "epoch": 0.66, + "learning_rate": 5.589999590263355e-05, + "loss": 0.4552, + "step": 22035 + }, + { + "epoch": 0.66, + "learning_rate": 5.5891341979000664e-05, + "loss": 0.44, + "step": 22036 + }, + { + "epoch": 0.66, + "learning_rate": 5.588268846547251e-05, + "loss": 0.5853, + "step": 22037 + }, + { + "epoch": 0.66, + "learning_rate": 5.587403536212957e-05, + "loss": 0.6081, + "step": 22038 + }, + { + "epoch": 0.66, + "learning_rate": 5.5865382669052277e-05, + "loss": 0.6279, + "step": 22039 + }, + { + "epoch": 0.66, + "learning_rate": 5.585673038632101e-05, + "loss": 0.5545, + "step": 22040 + }, + { + "epoch": 0.66, + "learning_rate": 5.584807851401628e-05, + "loss": 0.5468, + "step": 22041 + }, + { + "epoch": 0.66, + "learning_rate": 5.583942705221855e-05, + "loss": 0.5411, + "step": 22042 + }, + { + "epoch": 0.66, + "learning_rate": 5.5830776001008235e-05, + "loss": 0.5188, + "step": 22043 + }, + { + "epoch": 0.66, + "learning_rate": 5.5822125360465717e-05, + "loss": 0.5422, + "step": 22044 + }, + { + "epoch": 0.66, + "learning_rate": 5.581347513067147e-05, + "loss": 0.5867, + "step": 22045 + }, + { + "epoch": 0.66, + "learning_rate": 5.580482531170592e-05, + "loss": 0.5264, + "step": 22046 + }, + { + "epoch": 0.66, + "learning_rate": 5.5796175903649515e-05, + "loss": 0.4966, + "step": 22047 + }, + { + "epoch": 0.66, + "learning_rate": 5.578752690658267e-05, + "loss": 0.5384, + "step": 22048 + }, + { + "epoch": 0.66, + "learning_rate": 5.577887832058572e-05, + "loss": 0.5587, + "step": 22049 + }, + { + "epoch": 0.66, + "learning_rate": 5.57702301457392e-05, + "loss": 0.6234, + "step": 22050 + }, + { + "epoch": 0.66, + "learning_rate": 5.576158238212339e-05, + "loss": 0.4816, + "step": 22051 + }, + { + "epoch": 0.66, + "learning_rate": 5.575293502981881e-05, + "loss": 0.487, + "step": 22052 + }, + { + "epoch": 0.66, + "learning_rate": 5.574428808890576e-05, + "loss": 0.6131, + "step": 22053 + }, + { + "epoch": 0.66, + "learning_rate": 5.5735641559464714e-05, + "loss": 0.6523, + "step": 22054 + }, + { + "epoch": 0.66, + "learning_rate": 5.5726995441576005e-05, + "loss": 0.7258, + "step": 22055 + }, + { + "epoch": 0.66, + "learning_rate": 5.571834973532003e-05, + "loss": 0.5006, + "step": 22056 + }, + { + "epoch": 0.66, + "learning_rate": 5.5709704440777234e-05, + "loss": 0.4961, + "step": 22057 + }, + { + "epoch": 0.66, + "learning_rate": 5.570105955802796e-05, + "loss": 0.4372, + "step": 22058 + }, + { + "epoch": 0.66, + "learning_rate": 5.569241508715254e-05, + "loss": 0.5721, + "step": 22059 + }, + { + "epoch": 0.66, + "learning_rate": 5.568377102823138e-05, + "loss": 0.544, + "step": 22060 + }, + { + "epoch": 0.66, + "learning_rate": 5.567512738134489e-05, + "loss": 0.4736, + "step": 22061 + }, + { + "epoch": 0.66, + "learning_rate": 5.56664841465734e-05, + "loss": 0.5646, + "step": 22062 + }, + { + "epoch": 0.66, + "learning_rate": 5.565784132399725e-05, + "loss": 0.6827, + "step": 22063 + }, + { + "epoch": 0.66, + "learning_rate": 5.5649198913696795e-05, + "loss": 0.5948, + "step": 22064 + }, + { + "epoch": 0.66, + "learning_rate": 5.564055691575243e-05, + "loss": 0.4369, + "step": 22065 + }, + { + "epoch": 0.66, + "learning_rate": 5.563191533024451e-05, + "loss": 0.4833, + "step": 22066 + }, + { + "epoch": 0.66, + "learning_rate": 5.562327415725337e-05, + "loss": 0.5735, + "step": 22067 + }, + { + "epoch": 0.66, + "learning_rate": 5.5614633396859294e-05, + "loss": 0.5661, + "step": 22068 + }, + { + "epoch": 0.66, + "learning_rate": 5.5605993049142666e-05, + "loss": 0.5508, + "step": 22069 + }, + { + "epoch": 0.66, + "learning_rate": 5.559735311418386e-05, + "loss": 0.4213, + "step": 22070 + }, + { + "epoch": 0.66, + "learning_rate": 5.558871359206317e-05, + "loss": 0.5363, + "step": 22071 + }, + { + "epoch": 0.66, + "learning_rate": 5.558007448286089e-05, + "loss": 0.5906, + "step": 22072 + }, + { + "epoch": 0.66, + "learning_rate": 5.557143578665736e-05, + "loss": 0.5584, + "step": 22073 + }, + { + "epoch": 0.66, + "learning_rate": 5.556279750353296e-05, + "loss": 0.524, + "step": 22074 + }, + { + "epoch": 0.66, + "learning_rate": 5.555415963356792e-05, + "loss": 0.5169, + "step": 22075 + }, + { + "epoch": 0.66, + "learning_rate": 5.554552217684261e-05, + "loss": 0.5136, + "step": 22076 + }, + { + "epoch": 0.66, + "learning_rate": 5.55368851334373e-05, + "loss": 0.5376, + "step": 22077 + }, + { + "epoch": 0.66, + "learning_rate": 5.5528248503432344e-05, + "loss": 0.514, + "step": 22078 + }, + { + "epoch": 0.66, + "learning_rate": 5.5519612286907964e-05, + "loss": 0.6405, + "step": 22079 + }, + { + "epoch": 0.66, + "learning_rate": 5.55109764839445e-05, + "loss": 0.4942, + "step": 22080 + }, + { + "epoch": 0.66, + "learning_rate": 5.550234109462229e-05, + "loss": 0.4358, + "step": 22081 + }, + { + "epoch": 0.66, + "learning_rate": 5.5493706119021583e-05, + "loss": 0.6193, + "step": 22082 + }, + { + "epoch": 0.66, + "learning_rate": 5.548507155722261e-05, + "loss": 0.5709, + "step": 22083 + }, + { + "epoch": 0.66, + "learning_rate": 5.5476437409305704e-05, + "loss": 0.5339, + "step": 22084 + }, + { + "epoch": 0.66, + "learning_rate": 5.546780367535117e-05, + "loss": 0.4875, + "step": 22085 + }, + { + "epoch": 0.66, + "learning_rate": 5.545917035543926e-05, + "loss": 0.6263, + "step": 22086 + }, + { + "epoch": 0.66, + "learning_rate": 5.5450537449650186e-05, + "loss": 0.5866, + "step": 22087 + }, + { + "epoch": 0.66, + "learning_rate": 5.544190495806425e-05, + "loss": 0.5997, + "step": 22088 + }, + { + "epoch": 0.66, + "learning_rate": 5.5433272880761765e-05, + "loss": 0.4954, + "step": 22089 + }, + { + "epoch": 0.66, + "learning_rate": 5.5424641217822915e-05, + "loss": 0.437, + "step": 22090 + }, + { + "epoch": 0.66, + "learning_rate": 5.541600996932803e-05, + "loss": 0.6178, + "step": 22091 + }, + { + "epoch": 0.66, + "learning_rate": 5.5407379135357264e-05, + "loss": 0.5049, + "step": 22092 + }, + { + "epoch": 0.66, + "learning_rate": 5.539874871599095e-05, + "loss": 0.5005, + "step": 22093 + }, + { + "epoch": 0.66, + "learning_rate": 5.5390118711309266e-05, + "loss": 0.5374, + "step": 22094 + }, + { + "epoch": 0.66, + "learning_rate": 5.5381489121392515e-05, + "loss": 0.4364, + "step": 22095 + }, + { + "epoch": 0.66, + "learning_rate": 5.5372859946320856e-05, + "loss": 0.6353, + "step": 22096 + }, + { + "epoch": 0.66, + "learning_rate": 5.5364231186174595e-05, + "loss": 0.3807, + "step": 22097 + }, + { + "epoch": 0.66, + "learning_rate": 5.535560284103388e-05, + "loss": 0.5295, + "step": 22098 + }, + { + "epoch": 0.66, + "learning_rate": 5.5346974910978974e-05, + "loss": 0.5955, + "step": 22099 + }, + { + "epoch": 0.66, + "learning_rate": 5.533834739609015e-05, + "loss": 0.5009, + "step": 22100 + }, + { + "epoch": 0.66, + "learning_rate": 5.532972029644756e-05, + "loss": 0.5106, + "step": 22101 + }, + { + "epoch": 0.66, + "learning_rate": 5.53210936121314e-05, + "loss": 0.5541, + "step": 22102 + }, + { + "epoch": 0.66, + "learning_rate": 5.5312467343221896e-05, + "loss": 0.4894, + "step": 22103 + }, + { + "epoch": 0.66, + "learning_rate": 5.5303841489799305e-05, + "loss": 0.4319, + "step": 22104 + }, + { + "epoch": 0.66, + "learning_rate": 5.5295216051943776e-05, + "loss": 0.4654, + "step": 22105 + }, + { + "epoch": 0.66, + "learning_rate": 5.528659102973548e-05, + "loss": 0.541, + "step": 22106 + }, + { + "epoch": 0.66, + "learning_rate": 5.527796642325462e-05, + "loss": 0.5941, + "step": 22107 + }, + { + "epoch": 0.66, + "learning_rate": 5.526934223258142e-05, + "loss": 0.5408, + "step": 22108 + }, + { + "epoch": 0.66, + "learning_rate": 5.526071845779608e-05, + "loss": 0.5728, + "step": 22109 + }, + { + "epoch": 0.66, + "learning_rate": 5.525209509897875e-05, + "loss": 0.6252, + "step": 22110 + }, + { + "epoch": 0.66, + "learning_rate": 5.5243472156209576e-05, + "loss": 0.5688, + "step": 22111 + }, + { + "epoch": 0.66, + "learning_rate": 5.5234849629568753e-05, + "loss": 0.5579, + "step": 22112 + }, + { + "epoch": 0.66, + "learning_rate": 5.522622751913649e-05, + "loss": 0.5606, + "step": 22113 + }, + { + "epoch": 0.66, + "learning_rate": 5.521760582499291e-05, + "loss": 0.5216, + "step": 22114 + }, + { + "epoch": 0.66, + "learning_rate": 5.520898454721817e-05, + "loss": 0.5629, + "step": 22115 + }, + { + "epoch": 0.66, + "learning_rate": 5.520036368589245e-05, + "loss": 0.5101, + "step": 22116 + }, + { + "epoch": 0.66, + "learning_rate": 5.519174324109589e-05, + "loss": 0.4984, + "step": 22117 + }, + { + "epoch": 0.66, + "learning_rate": 5.5183123212908616e-05, + "loss": 0.589, + "step": 22118 + }, + { + "epoch": 0.66, + "learning_rate": 5.5174503601410835e-05, + "loss": 0.5685, + "step": 22119 + }, + { + "epoch": 0.66, + "learning_rate": 5.5165884406682666e-05, + "loss": 0.6017, + "step": 22120 + }, + { + "epoch": 0.66, + "learning_rate": 5.515726562880419e-05, + "loss": 0.4485, + "step": 22121 + }, + { + "epoch": 0.66, + "learning_rate": 5.5148647267855594e-05, + "loss": 0.5397, + "step": 22122 + }, + { + "epoch": 0.66, + "learning_rate": 5.5140029323916997e-05, + "loss": 0.4405, + "step": 22123 + }, + { + "epoch": 0.66, + "learning_rate": 5.51314117970686e-05, + "loss": 0.5215, + "step": 22124 + }, + { + "epoch": 0.66, + "learning_rate": 5.5122794687390375e-05, + "loss": 0.5815, + "step": 22125 + }, + { + "epoch": 0.66, + "learning_rate": 5.511417799496252e-05, + "loss": 0.5171, + "step": 22126 + }, + { + "epoch": 0.66, + "learning_rate": 5.510556171986515e-05, + "loss": 0.5733, + "step": 22127 + }, + { + "epoch": 0.66, + "learning_rate": 5.509694586217842e-05, + "loss": 0.485, + "step": 22128 + }, + { + "epoch": 0.66, + "learning_rate": 5.508833042198238e-05, + "loss": 0.5509, + "step": 22129 + }, + { + "epoch": 0.66, + "learning_rate": 5.50797153993571e-05, + "loss": 0.5577, + "step": 22130 + }, + { + "epoch": 0.66, + "learning_rate": 5.5071100794382737e-05, + "loss": 0.5668, + "step": 22131 + }, + { + "epoch": 0.66, + "learning_rate": 5.5062486607139405e-05, + "loss": 0.4286, + "step": 22132 + }, + { + "epoch": 0.66, + "learning_rate": 5.505387283770713e-05, + "loss": 0.5131, + "step": 22133 + }, + { + "epoch": 0.66, + "learning_rate": 5.5045259486166053e-05, + "loss": 0.6014, + "step": 22134 + }, + { + "epoch": 0.66, + "learning_rate": 5.503664655259622e-05, + "loss": 0.5616, + "step": 22135 + }, + { + "epoch": 0.66, + "learning_rate": 5.5028034037077744e-05, + "loss": 0.6283, + "step": 22136 + }, + { + "epoch": 0.66, + "learning_rate": 5.5019421939690655e-05, + "loss": 0.5164, + "step": 22137 + }, + { + "epoch": 0.66, + "learning_rate": 5.5010810260515086e-05, + "loss": 0.5808, + "step": 22138 + }, + { + "epoch": 0.66, + "learning_rate": 5.500219899963105e-05, + "loss": 0.4688, + "step": 22139 + }, + { + "epoch": 0.66, + "learning_rate": 5.499358815711866e-05, + "loss": 0.4619, + "step": 22140 + }, + { + "epoch": 0.66, + "learning_rate": 5.498497773305791e-05, + "loss": 0.4076, + "step": 22141 + }, + { + "epoch": 0.66, + "learning_rate": 5.49763677275289e-05, + "loss": 0.5195, + "step": 22142 + }, + { + "epoch": 0.66, + "learning_rate": 5.4967758140611715e-05, + "loss": 0.5506, + "step": 22143 + }, + { + "epoch": 0.66, + "learning_rate": 5.495914897238637e-05, + "loss": 0.4931, + "step": 22144 + }, + { + "epoch": 0.66, + "learning_rate": 5.495054022293287e-05, + "loss": 0.5872, + "step": 22145 + }, + { + "epoch": 0.66, + "learning_rate": 5.4941931892331275e-05, + "loss": 0.5537, + "step": 22146 + }, + { + "epoch": 0.66, + "learning_rate": 5.4933323980661686e-05, + "loss": 0.453, + "step": 22147 + }, + { + "epoch": 0.66, + "learning_rate": 5.4924716488004104e-05, + "loss": 0.5215, + "step": 22148 + }, + { + "epoch": 0.66, + "learning_rate": 5.4916109414438475e-05, + "loss": 0.7049, + "step": 22149 + }, + { + "epoch": 0.66, + "learning_rate": 5.4907502760044914e-05, + "loss": 0.5714, + "step": 22150 + }, + { + "epoch": 0.66, + "learning_rate": 5.489889652490345e-05, + "loss": 0.4644, + "step": 22151 + }, + { + "epoch": 0.66, + "learning_rate": 5.4890290709094036e-05, + "loss": 0.5605, + "step": 22152 + }, + { + "epoch": 0.66, + "learning_rate": 5.4881685312696774e-05, + "loss": 0.5785, + "step": 22153 + }, + { + "epoch": 0.66, + "learning_rate": 5.487308033579155e-05, + "loss": 0.5515, + "step": 22154 + }, + { + "epoch": 0.66, + "learning_rate": 5.48644757784585e-05, + "loss": 0.5141, + "step": 22155 + }, + { + "epoch": 0.66, + "learning_rate": 5.485587164077752e-05, + "loss": 0.4795, + "step": 22156 + }, + { + "epoch": 0.66, + "learning_rate": 5.48472679228287e-05, + "loss": 0.5287, + "step": 22157 + }, + { + "epoch": 0.66, + "learning_rate": 5.4838664624691924e-05, + "loss": 0.5379, + "step": 22158 + }, + { + "epoch": 0.66, + "learning_rate": 5.4830061746447306e-05, + "loss": 0.5266, + "step": 22159 + }, + { + "epoch": 0.66, + "learning_rate": 5.482145928817473e-05, + "loss": 0.5775, + "step": 22160 + }, + { + "epoch": 0.66, + "learning_rate": 5.4812857249954206e-05, + "loss": 0.4924, + "step": 22161 + }, + { + "epoch": 0.66, + "learning_rate": 5.480425563186577e-05, + "loss": 0.4623, + "step": 22162 + }, + { + "epoch": 0.66, + "learning_rate": 5.4795654433989354e-05, + "loss": 0.6871, + "step": 22163 + }, + { + "epoch": 0.66, + "learning_rate": 5.478705365640488e-05, + "loss": 0.4387, + "step": 22164 + }, + { + "epoch": 0.66, + "learning_rate": 5.477845329919238e-05, + "loss": 0.5201, + "step": 22165 + }, + { + "epoch": 0.66, + "learning_rate": 5.476985336243182e-05, + "loss": 0.467, + "step": 22166 + }, + { + "epoch": 0.66, + "learning_rate": 5.4761253846203145e-05, + "loss": 0.5102, + "step": 22167 + }, + { + "epoch": 0.66, + "learning_rate": 5.475265475058626e-05, + "loss": 0.5374, + "step": 22168 + }, + { + "epoch": 0.66, + "learning_rate": 5.474405607566115e-05, + "loss": 0.5368, + "step": 22169 + }, + { + "epoch": 0.66, + "learning_rate": 5.473545782150777e-05, + "loss": 0.6211, + "step": 22170 + }, + { + "epoch": 0.66, + "learning_rate": 5.472685998820613e-05, + "loss": 0.5705, + "step": 22171 + }, + { + "epoch": 0.66, + "learning_rate": 5.471826257583607e-05, + "loss": 0.5793, + "step": 22172 + }, + { + "epoch": 0.66, + "learning_rate": 5.470966558447752e-05, + "loss": 0.5478, + "step": 22173 + }, + { + "epoch": 0.66, + "learning_rate": 5.470106901421046e-05, + "loss": 0.5623, + "step": 22174 + }, + { + "epoch": 0.66, + "learning_rate": 5.469247286511484e-05, + "loss": 0.4781, + "step": 22175 + }, + { + "epoch": 0.66, + "learning_rate": 5.468387713727054e-05, + "loss": 0.5624, + "step": 22176 + }, + { + "epoch": 0.66, + "learning_rate": 5.4675281830757476e-05, + "loss": 0.4426, + "step": 22177 + }, + { + "epoch": 0.66, + "learning_rate": 5.466668694565555e-05, + "loss": 0.5064, + "step": 22178 + }, + { + "epoch": 0.66, + "learning_rate": 5.465809248204475e-05, + "loss": 0.6105, + "step": 22179 + }, + { + "epoch": 0.66, + "learning_rate": 5.464949844000491e-05, + "loss": 0.6349, + "step": 22180 + }, + { + "epoch": 0.66, + "learning_rate": 5.4640904819615966e-05, + "loss": 0.4759, + "step": 22181 + }, + { + "epoch": 0.66, + "learning_rate": 5.463231162095783e-05, + "loss": 0.5588, + "step": 22182 + }, + { + "epoch": 0.66, + "learning_rate": 5.462371884411035e-05, + "loss": 0.6075, + "step": 22183 + }, + { + "epoch": 0.66, + "learning_rate": 5.461512648915343e-05, + "loss": 0.5136, + "step": 22184 + }, + { + "epoch": 0.66, + "learning_rate": 5.4606534556166976e-05, + "loss": 0.6443, + "step": 22185 + }, + { + "epoch": 0.66, + "learning_rate": 5.459794304523096e-05, + "loss": 0.4339, + "step": 22186 + }, + { + "epoch": 0.66, + "learning_rate": 5.458935195642509e-05, + "loss": 0.5088, + "step": 22187 + }, + { + "epoch": 0.66, + "learning_rate": 5.4580761289829315e-05, + "loss": 0.4344, + "step": 22188 + }, + { + "epoch": 0.66, + "learning_rate": 5.457217104552355e-05, + "loss": 0.494, + "step": 22189 + }, + { + "epoch": 0.66, + "learning_rate": 5.456358122358764e-05, + "loss": 0.5273, + "step": 22190 + }, + { + "epoch": 0.66, + "learning_rate": 5.4554991824101466e-05, + "loss": 0.5899, + "step": 22191 + }, + { + "epoch": 0.66, + "learning_rate": 5.4546402847144805e-05, + "loss": 0.6675, + "step": 22192 + }, + { + "epoch": 0.66, + "learning_rate": 5.453781429279758e-05, + "loss": 0.6166, + "step": 22193 + }, + { + "epoch": 0.66, + "learning_rate": 5.452922616113969e-05, + "loss": 0.6341, + "step": 22194 + }, + { + "epoch": 0.66, + "learning_rate": 5.4520638452250886e-05, + "loss": 0.5521, + "step": 22195 + }, + { + "epoch": 0.66, + "learning_rate": 5.4512051166211095e-05, + "loss": 0.551, + "step": 22196 + }, + { + "epoch": 0.66, + "learning_rate": 5.4503464303100094e-05, + "loss": 0.501, + "step": 22197 + }, + { + "epoch": 0.66, + "learning_rate": 5.4494877862997786e-05, + "loss": 0.6225, + "step": 22198 + }, + { + "epoch": 0.66, + "learning_rate": 5.4486291845983916e-05, + "loss": 0.5723, + "step": 22199 + }, + { + "epoch": 0.66, + "learning_rate": 5.447770625213842e-05, + "loss": 0.4699, + "step": 22200 + }, + { + "epoch": 0.66, + "learning_rate": 5.446912108154101e-05, + "loss": 0.4839, + "step": 22201 + }, + { + "epoch": 0.66, + "learning_rate": 5.4460536334271615e-05, + "loss": 0.6004, + "step": 22202 + }, + { + "epoch": 0.66, + "learning_rate": 5.4451952010409955e-05, + "loss": 0.5644, + "step": 22203 + }, + { + "epoch": 0.66, + "learning_rate": 5.4443368110035896e-05, + "loss": 0.67, + "step": 22204 + }, + { + "epoch": 0.66, + "learning_rate": 5.443478463322929e-05, + "loss": 0.5648, + "step": 22205 + }, + { + "epoch": 0.66, + "learning_rate": 5.4426201580069894e-05, + "loss": 0.4398, + "step": 22206 + }, + { + "epoch": 0.66, + "learning_rate": 5.441761895063746e-05, + "loss": 0.6252, + "step": 22207 + }, + { + "epoch": 0.66, + "learning_rate": 5.440903674501185e-05, + "loss": 0.5891, + "step": 22208 + }, + { + "epoch": 0.66, + "learning_rate": 5.440045496327289e-05, + "loss": 0.5882, + "step": 22209 + }, + { + "epoch": 0.66, + "learning_rate": 5.439187360550031e-05, + "loss": 0.5086, + "step": 22210 + }, + { + "epoch": 0.66, + "learning_rate": 5.438329267177389e-05, + "loss": 0.6222, + "step": 22211 + }, + { + "epoch": 0.66, + "learning_rate": 5.437471216217342e-05, + "loss": 0.6404, + "step": 22212 + }, + { + "epoch": 0.66, + "learning_rate": 5.43661320767787e-05, + "loss": 0.5447, + "step": 22213 + }, + { + "epoch": 0.66, + "learning_rate": 5.435755241566953e-05, + "loss": 0.5732, + "step": 22214 + }, + { + "epoch": 0.66, + "learning_rate": 5.4348973178925644e-05, + "loss": 0.5761, + "step": 22215 + }, + { + "epoch": 0.66, + "learning_rate": 5.434039436662677e-05, + "loss": 0.5352, + "step": 22216 + }, + { + "epoch": 0.66, + "learning_rate": 5.433181597885275e-05, + "loss": 0.4902, + "step": 22217 + }, + { + "epoch": 0.66, + "learning_rate": 5.432323801568329e-05, + "loss": 0.5567, + "step": 22218 + }, + { + "epoch": 0.66, + "learning_rate": 5.431466047719816e-05, + "loss": 0.5443, + "step": 22219 + }, + { + "epoch": 0.66, + "learning_rate": 5.430608336347709e-05, + "loss": 0.5251, + "step": 22220 + }, + { + "epoch": 0.66, + "learning_rate": 5.429750667459987e-05, + "loss": 0.5448, + "step": 22221 + }, + { + "epoch": 0.66, + "learning_rate": 5.42889304106462e-05, + "loss": 0.5093, + "step": 22222 + }, + { + "epoch": 0.66, + "learning_rate": 5.4280354571695825e-05, + "loss": 0.5214, + "step": 22223 + }, + { + "epoch": 0.66, + "learning_rate": 5.427177915782853e-05, + "loss": 0.5693, + "step": 22224 + }, + { + "epoch": 0.66, + "learning_rate": 5.4263204169124e-05, + "loss": 0.5237, + "step": 22225 + }, + { + "epoch": 0.66, + "learning_rate": 5.4254629605661944e-05, + "loss": 0.653, + "step": 22226 + }, + { + "epoch": 0.66, + "learning_rate": 5.4246055467522104e-05, + "loss": 0.5429, + "step": 22227 + }, + { + "epoch": 0.66, + "learning_rate": 5.423748175478425e-05, + "loss": 0.4562, + "step": 22228 + }, + { + "epoch": 0.66, + "learning_rate": 5.422890846752804e-05, + "loss": 0.5938, + "step": 22229 + }, + { + "epoch": 0.66, + "learning_rate": 5.422033560583315e-05, + "loss": 0.523, + "step": 22230 + }, + { + "epoch": 0.66, + "learning_rate": 5.421176316977935e-05, + "loss": 0.3872, + "step": 22231 + }, + { + "epoch": 0.66, + "learning_rate": 5.420319115944632e-05, + "loss": 0.4287, + "step": 22232 + }, + { + "epoch": 0.66, + "learning_rate": 5.419461957491381e-05, + "loss": 0.5088, + "step": 22233 + }, + { + "epoch": 0.66, + "learning_rate": 5.418604841626147e-05, + "loss": 0.4337, + "step": 22234 + }, + { + "epoch": 0.66, + "learning_rate": 5.417747768356894e-05, + "loss": 0.5236, + "step": 22235 + }, + { + "epoch": 0.66, + "learning_rate": 5.4168907376915975e-05, + "loss": 0.5845, + "step": 22236 + }, + { + "epoch": 0.66, + "learning_rate": 5.416033749638227e-05, + "loss": 0.5069, + "step": 22237 + }, + { + "epoch": 0.66, + "learning_rate": 5.415176804204749e-05, + "loss": 0.4066, + "step": 22238 + }, + { + "epoch": 0.66, + "learning_rate": 5.4143199013991254e-05, + "loss": 0.5238, + "step": 22239 + }, + { + "epoch": 0.66, + "learning_rate": 5.413463041229328e-05, + "loss": 0.6915, + "step": 22240 + }, + { + "epoch": 0.66, + "learning_rate": 5.412606223703326e-05, + "loss": 0.6173, + "step": 22241 + }, + { + "epoch": 0.66, + "learning_rate": 5.41174944882908e-05, + "loss": 0.5656, + "step": 22242 + }, + { + "epoch": 0.66, + "learning_rate": 5.4108927166145614e-05, + "loss": 0.4837, + "step": 22243 + }, + { + "epoch": 0.66, + "learning_rate": 5.4100360270677305e-05, + "loss": 0.6005, + "step": 22244 + }, + { + "epoch": 0.66, + "learning_rate": 5.4091793801965594e-05, + "loss": 0.5448, + "step": 22245 + }, + { + "epoch": 0.66, + "learning_rate": 5.408322776009005e-05, + "loss": 0.6038, + "step": 22246 + }, + { + "epoch": 0.66, + "learning_rate": 5.407466214513035e-05, + "loss": 0.5436, + "step": 22247 + }, + { + "epoch": 0.66, + "learning_rate": 5.4066096957166214e-05, + "loss": 0.4956, + "step": 22248 + }, + { + "epoch": 0.66, + "learning_rate": 5.4057532196277113e-05, + "loss": 0.4884, + "step": 22249 + }, + { + "epoch": 0.66, + "learning_rate": 5.4048967862542785e-05, + "loss": 0.6254, + "step": 22250 + }, + { + "epoch": 0.66, + "learning_rate": 5.404040395604284e-05, + "loss": 0.5423, + "step": 22251 + }, + { + "epoch": 0.66, + "learning_rate": 5.403184047685692e-05, + "loss": 0.5375, + "step": 22252 + }, + { + "epoch": 0.66, + "learning_rate": 5.4023277425064645e-05, + "loss": 0.5125, + "step": 22253 + }, + { + "epoch": 0.66, + "learning_rate": 5.401471480074556e-05, + "loss": 0.5682, + "step": 22254 + }, + { + "epoch": 0.66, + "learning_rate": 5.4006152603979344e-05, + "loss": 0.6035, + "step": 22255 + }, + { + "epoch": 0.66, + "learning_rate": 5.399759083484561e-05, + "loss": 0.5809, + "step": 22256 + }, + { + "epoch": 0.66, + "learning_rate": 5.398902949342391e-05, + "loss": 0.5328, + "step": 22257 + }, + { + "epoch": 0.66, + "learning_rate": 5.398046857979393e-05, + "loss": 0.6716, + "step": 22258 + }, + { + "epoch": 0.66, + "learning_rate": 5.3971908094035164e-05, + "loss": 0.6013, + "step": 22259 + }, + { + "epoch": 0.66, + "learning_rate": 5.396334803622728e-05, + "loss": 0.4865, + "step": 22260 + }, + { + "epoch": 0.66, + "learning_rate": 5.3954788406449804e-05, + "loss": 0.5832, + "step": 22261 + }, + { + "epoch": 0.66, + "learning_rate": 5.3946229204782386e-05, + "loss": 0.438, + "step": 22262 + }, + { + "epoch": 0.66, + "learning_rate": 5.393767043130456e-05, + "loss": 0.5237, + "step": 22263 + }, + { + "epoch": 0.66, + "learning_rate": 5.392911208609592e-05, + "loss": 0.6387, + "step": 22264 + }, + { + "epoch": 0.66, + "learning_rate": 5.392055416923602e-05, + "loss": 0.5538, + "step": 22265 + }, + { + "epoch": 0.66, + "learning_rate": 5.391199668080442e-05, + "loss": 0.3463, + "step": 22266 + }, + { + "epoch": 0.66, + "learning_rate": 5.390343962088077e-05, + "loss": 0.5175, + "step": 22267 + }, + { + "epoch": 0.66, + "learning_rate": 5.389488298954455e-05, + "loss": 0.5622, + "step": 22268 + }, + { + "epoch": 0.66, + "learning_rate": 5.388632678687531e-05, + "loss": 0.6615, + "step": 22269 + }, + { + "epoch": 0.66, + "learning_rate": 5.387777101295259e-05, + "loss": 0.5036, + "step": 22270 + }, + { + "epoch": 0.66, + "learning_rate": 5.3869215667856046e-05, + "loss": 0.5525, + "step": 22271 + }, + { + "epoch": 0.66, + "learning_rate": 5.3860660751665136e-05, + "loss": 0.5381, + "step": 22272 + }, + { + "epoch": 0.66, + "learning_rate": 5.385210626445938e-05, + "loss": 0.6304, + "step": 22273 + }, + { + "epoch": 0.66, + "learning_rate": 5.3843552206318336e-05, + "loss": 0.5571, + "step": 22274 + }, + { + "epoch": 0.66, + "learning_rate": 5.3834998577321546e-05, + "loss": 0.5219, + "step": 22275 + }, + { + "epoch": 0.66, + "learning_rate": 5.38264453775486e-05, + "loss": 0.5566, + "step": 22276 + }, + { + "epoch": 0.66, + "learning_rate": 5.3817892607078945e-05, + "loss": 0.5119, + "step": 22277 + }, + { + "epoch": 0.66, + "learning_rate": 5.380934026599208e-05, + "loss": 0.6302, + "step": 22278 + }, + { + "epoch": 0.66, + "learning_rate": 5.380078835436756e-05, + "loss": 0.5742, + "step": 22279 + }, + { + "epoch": 0.66, + "learning_rate": 5.379223687228494e-05, + "loss": 0.4495, + "step": 22280 + }, + { + "epoch": 0.66, + "learning_rate": 5.3783685819823685e-05, + "loss": 0.4275, + "step": 22281 + }, + { + "epoch": 0.66, + "learning_rate": 5.3775135197063254e-05, + "loss": 0.6301, + "step": 22282 + }, + { + "epoch": 0.66, + "learning_rate": 5.376658500408323e-05, + "loss": 0.6137, + "step": 22283 + }, + { + "epoch": 0.66, + "learning_rate": 5.375803524096304e-05, + "loss": 0.5341, + "step": 22284 + }, + { + "epoch": 0.66, + "learning_rate": 5.374948590778221e-05, + "loss": 0.6059, + "step": 22285 + }, + { + "epoch": 0.66, + "learning_rate": 5.374093700462026e-05, + "loss": 0.6065, + "step": 22286 + }, + { + "epoch": 0.66, + "learning_rate": 5.3732388531556645e-05, + "loss": 0.5648, + "step": 22287 + }, + { + "epoch": 0.66, + "learning_rate": 5.3723840488670795e-05, + "loss": 0.5853, + "step": 22288 + }, + { + "epoch": 0.66, + "learning_rate": 5.3715292876042235e-05, + "loss": 0.5418, + "step": 22289 + }, + { + "epoch": 0.66, + "learning_rate": 5.370674569375044e-05, + "loss": 0.5202, + "step": 22290 + }, + { + "epoch": 0.66, + "learning_rate": 5.369819894187494e-05, + "loss": 0.5431, + "step": 22291 + }, + { + "epoch": 0.66, + "learning_rate": 5.3689652620495053e-05, + "loss": 0.5649, + "step": 22292 + }, + { + "epoch": 0.66, + "learning_rate": 5.3681106729690335e-05, + "loss": 0.5758, + "step": 22293 + }, + { + "epoch": 0.66, + "learning_rate": 5.3672561269540214e-05, + "loss": 0.628, + "step": 22294 + }, + { + "epoch": 0.66, + "learning_rate": 5.36640162401242e-05, + "loss": 0.6195, + "step": 22295 + }, + { + "epoch": 0.66, + "learning_rate": 5.3655471641521696e-05, + "loss": 0.6071, + "step": 22296 + }, + { + "epoch": 0.66, + "learning_rate": 5.36469274738121e-05, + "loss": 0.4753, + "step": 22297 + }, + { + "epoch": 0.66, + "learning_rate": 5.36383837370749e-05, + "loss": 0.5614, + "step": 22298 + }, + { + "epoch": 0.66, + "learning_rate": 5.3629840431389566e-05, + "loss": 0.5657, + "step": 22299 + }, + { + "epoch": 0.66, + "learning_rate": 5.362129755683546e-05, + "loss": 0.5541, + "step": 22300 + }, + { + "epoch": 0.66, + "learning_rate": 5.361275511349209e-05, + "loss": 0.6207, + "step": 22301 + }, + { + "epoch": 0.66, + "learning_rate": 5.3604213101438775e-05, + "loss": 0.5975, + "step": 22302 + }, + { + "epoch": 0.66, + "learning_rate": 5.359567152075504e-05, + "loss": 0.6444, + "step": 22303 + }, + { + "epoch": 0.66, + "learning_rate": 5.3587130371520234e-05, + "loss": 0.6996, + "step": 22304 + }, + { + "epoch": 0.66, + "learning_rate": 5.3578589653813814e-05, + "loss": 0.5062, + "step": 22305 + }, + { + "epoch": 0.66, + "learning_rate": 5.3570049367715133e-05, + "loss": 0.4447, + "step": 22306 + }, + { + "epoch": 0.66, + "learning_rate": 5.356150951330368e-05, + "loss": 0.5078, + "step": 22307 + }, + { + "epoch": 0.66, + "learning_rate": 5.355297009065874e-05, + "loss": 0.5373, + "step": 22308 + }, + { + "epoch": 0.66, + "learning_rate": 5.3544431099859784e-05, + "loss": 0.59, + "step": 22309 + }, + { + "epoch": 0.66, + "learning_rate": 5.353589254098623e-05, + "loss": 0.5767, + "step": 22310 + }, + { + "epoch": 0.66, + "learning_rate": 5.352735441411742e-05, + "loss": 0.5857, + "step": 22311 + }, + { + "epoch": 0.66, + "learning_rate": 5.351881671933272e-05, + "loss": 0.5506, + "step": 22312 + }, + { + "epoch": 0.66, + "learning_rate": 5.351027945671154e-05, + "loss": 0.4579, + "step": 22313 + }, + { + "epoch": 0.66, + "learning_rate": 5.3501742626333286e-05, + "loss": 0.4767, + "step": 22314 + }, + { + "epoch": 0.66, + "learning_rate": 5.349320622827729e-05, + "loss": 0.5242, + "step": 22315 + }, + { + "epoch": 0.66, + "learning_rate": 5.3484670262622895e-05, + "loss": 0.5237, + "step": 22316 + }, + { + "epoch": 0.66, + "learning_rate": 5.34761347294495e-05, + "loss": 0.4926, + "step": 22317 + }, + { + "epoch": 0.66, + "learning_rate": 5.3467599628836494e-05, + "loss": 0.6174, + "step": 22318 + }, + { + "epoch": 0.66, + "learning_rate": 5.3459064960863156e-05, + "loss": 0.4762, + "step": 22319 + }, + { + "epoch": 0.66, + "learning_rate": 5.3450530725608926e-05, + "loss": 0.5129, + "step": 22320 + }, + { + "epoch": 0.66, + "learning_rate": 5.3441996923153084e-05, + "loss": 0.5949, + "step": 22321 + }, + { + "epoch": 0.66, + "learning_rate": 5.3433463553575016e-05, + "loss": 0.5043, + "step": 22322 + }, + { + "epoch": 0.66, + "learning_rate": 5.342493061695402e-05, + "loss": 0.5192, + "step": 22323 + }, + { + "epoch": 0.66, + "learning_rate": 5.341639811336949e-05, + "loss": 0.5776, + "step": 22324 + }, + { + "epoch": 0.66, + "learning_rate": 5.340786604290069e-05, + "loss": 0.6508, + "step": 22325 + }, + { + "epoch": 0.66, + "learning_rate": 5.339933440562701e-05, + "loss": 0.5438, + "step": 22326 + }, + { + "epoch": 0.66, + "learning_rate": 5.33908032016277e-05, + "loss": 0.561, + "step": 22327 + }, + { + "epoch": 0.66, + "learning_rate": 5.338227243098215e-05, + "loss": 0.4869, + "step": 22328 + }, + { + "epoch": 0.66, + "learning_rate": 5.337374209376967e-05, + "loss": 0.5388, + "step": 22329 + }, + { + "epoch": 0.66, + "learning_rate": 5.336521219006956e-05, + "loss": 0.4614, + "step": 22330 + }, + { + "epoch": 0.66, + "learning_rate": 5.335668271996108e-05, + "loss": 0.555, + "step": 22331 + }, + { + "epoch": 0.66, + "learning_rate": 5.3348153683523574e-05, + "loss": 0.5754, + "step": 22332 + }, + { + "epoch": 0.66, + "learning_rate": 5.333962508083639e-05, + "loss": 0.6242, + "step": 22333 + }, + { + "epoch": 0.66, + "learning_rate": 5.333109691197876e-05, + "loss": 0.6355, + "step": 22334 + }, + { + "epoch": 0.66, + "learning_rate": 5.332256917702996e-05, + "loss": 0.592, + "step": 22335 + }, + { + "epoch": 0.66, + "learning_rate": 5.331404187606931e-05, + "loss": 0.4713, + "step": 22336 + }, + { + "epoch": 0.67, + "learning_rate": 5.3305515009176085e-05, + "loss": 0.4559, + "step": 22337 + }, + { + "epoch": 0.67, + "learning_rate": 5.329698857642962e-05, + "loss": 0.5329, + "step": 22338 + }, + { + "epoch": 0.67, + "learning_rate": 5.3288462577909125e-05, + "loss": 0.5424, + "step": 22339 + }, + { + "epoch": 0.67, + "learning_rate": 5.327993701369386e-05, + "loss": 0.4669, + "step": 22340 + }, + { + "epoch": 0.67, + "learning_rate": 5.327141188386311e-05, + "loss": 0.4945, + "step": 22341 + }, + { + "epoch": 0.67, + "learning_rate": 5.326288718849619e-05, + "loss": 0.4988, + "step": 22342 + }, + { + "epoch": 0.67, + "learning_rate": 5.325436292767233e-05, + "loss": 0.494, + "step": 22343 + }, + { + "epoch": 0.67, + "learning_rate": 5.3245839101470716e-05, + "loss": 0.5673, + "step": 22344 + }, + { + "epoch": 0.67, + "learning_rate": 5.323731570997067e-05, + "loss": 0.5626, + "step": 22345 + }, + { + "epoch": 0.67, + "learning_rate": 5.322879275325146e-05, + "loss": 0.5157, + "step": 22346 + }, + { + "epoch": 0.67, + "learning_rate": 5.322027023139226e-05, + "loss": 0.5465, + "step": 22347 + }, + { + "epoch": 0.67, + "learning_rate": 5.321174814447236e-05, + "loss": 0.6132, + "step": 22348 + }, + { + "epoch": 0.67, + "learning_rate": 5.3203226492571e-05, + "loss": 0.5083, + "step": 22349 + }, + { + "epoch": 0.67, + "learning_rate": 5.3194705275767345e-05, + "loss": 0.5818, + "step": 22350 + }, + { + "epoch": 0.67, + "learning_rate": 5.3186184494140656e-05, + "loss": 0.4696, + "step": 22351 + }, + { + "epoch": 0.67, + "learning_rate": 5.3177664147770175e-05, + "loss": 0.5764, + "step": 22352 + }, + { + "epoch": 0.67, + "learning_rate": 5.316914423673519e-05, + "loss": 0.509, + "step": 22353 + }, + { + "epoch": 0.67, + "learning_rate": 5.316062476111476e-05, + "loss": 0.5621, + "step": 22354 + }, + { + "epoch": 0.67, + "learning_rate": 5.3152105720988166e-05, + "loss": 0.5379, + "step": 22355 + }, + { + "epoch": 0.67, + "learning_rate": 5.314358711643462e-05, + "loss": 0.5541, + "step": 22356 + }, + { + "epoch": 0.67, + "learning_rate": 5.313506894753337e-05, + "loss": 0.5405, + "step": 22357 + }, + { + "epoch": 0.67, + "learning_rate": 5.312655121436356e-05, + "loss": 0.4376, + "step": 22358 + }, + { + "epoch": 0.67, + "learning_rate": 5.311803391700437e-05, + "loss": 0.6253, + "step": 22359 + }, + { + "epoch": 0.67, + "learning_rate": 5.310951705553502e-05, + "loss": 0.4286, + "step": 22360 + }, + { + "epoch": 0.67, + "learning_rate": 5.310100063003471e-05, + "loss": 0.5643, + "step": 22361 + }, + { + "epoch": 0.67, + "learning_rate": 5.309248464058259e-05, + "loss": 0.487, + "step": 22362 + }, + { + "epoch": 0.67, + "learning_rate": 5.308396908725787e-05, + "loss": 0.5728, + "step": 22363 + }, + { + "epoch": 0.67, + "learning_rate": 5.307545397013967e-05, + "loss": 0.5672, + "step": 22364 + }, + { + "epoch": 0.67, + "learning_rate": 5.306693928930724e-05, + "loss": 0.6212, + "step": 22365 + }, + { + "epoch": 0.67, + "learning_rate": 5.305842504483966e-05, + "loss": 0.51, + "step": 22366 + }, + { + "epoch": 0.67, + "learning_rate": 5.304991123681619e-05, + "loss": 0.5304, + "step": 22367 + }, + { + "epoch": 0.67, + "learning_rate": 5.304139786531588e-05, + "loss": 0.5242, + "step": 22368 + }, + { + "epoch": 0.67, + "learning_rate": 5.303288493041798e-05, + "loss": 0.4869, + "step": 22369 + }, + { + "epoch": 0.67, + "learning_rate": 5.302437243220155e-05, + "loss": 0.5416, + "step": 22370 + }, + { + "epoch": 0.67, + "learning_rate": 5.30158603707458e-05, + "loss": 0.4894, + "step": 22371 + }, + { + "epoch": 0.67, + "learning_rate": 5.300734874612987e-05, + "loss": 0.5073, + "step": 22372 + }, + { + "epoch": 0.67, + "learning_rate": 5.299883755843289e-05, + "loss": 0.6498, + "step": 22373 + }, + { + "epoch": 0.67, + "learning_rate": 5.299032680773395e-05, + "loss": 0.5254, + "step": 22374 + }, + { + "epoch": 0.67, + "learning_rate": 5.298181649411222e-05, + "loss": 0.5327, + "step": 22375 + }, + { + "epoch": 0.67, + "learning_rate": 5.297330661764686e-05, + "loss": 0.6902, + "step": 22376 + }, + { + "epoch": 0.67, + "learning_rate": 5.2964797178416936e-05, + "loss": 0.4979, + "step": 22377 + }, + { + "epoch": 0.67, + "learning_rate": 5.2956288176501555e-05, + "loss": 0.5074, + "step": 22378 + }, + { + "epoch": 0.67, + "learning_rate": 5.294777961197985e-05, + "loss": 0.5782, + "step": 22379 + }, + { + "epoch": 0.67, + "learning_rate": 5.293927148493099e-05, + "loss": 0.43, + "step": 22380 + }, + { + "epoch": 0.67, + "learning_rate": 5.293076379543398e-05, + "loss": 0.4975, + "step": 22381 + }, + { + "epoch": 0.67, + "learning_rate": 5.292225654356799e-05, + "loss": 0.4756, + "step": 22382 + }, + { + "epoch": 0.67, + "learning_rate": 5.291374972941208e-05, + "loss": 0.6188, + "step": 22383 + }, + { + "epoch": 0.67, + "learning_rate": 5.2905243353045386e-05, + "loss": 0.5174, + "step": 22384 + }, + { + "epoch": 0.67, + "learning_rate": 5.289673741454695e-05, + "loss": 0.539, + "step": 22385 + }, + { + "epoch": 0.67, + "learning_rate": 5.288823191399589e-05, + "loss": 0.4604, + "step": 22386 + }, + { + "epoch": 0.67, + "learning_rate": 5.287972685147124e-05, + "loss": 0.5351, + "step": 22387 + }, + { + "epoch": 0.67, + "learning_rate": 5.287122222705215e-05, + "loss": 0.4597, + "step": 22388 + }, + { + "epoch": 0.67, + "learning_rate": 5.286271804081762e-05, + "loss": 0.4845, + "step": 22389 + }, + { + "epoch": 0.67, + "learning_rate": 5.285421429284674e-05, + "loss": 0.396, + "step": 22390 + }, + { + "epoch": 0.67, + "learning_rate": 5.2845710983218635e-05, + "loss": 0.5576, + "step": 22391 + }, + { + "epoch": 0.67, + "learning_rate": 5.2837208112012306e-05, + "loss": 0.5419, + "step": 22392 + }, + { + "epoch": 0.67, + "learning_rate": 5.282870567930678e-05, + "loss": 0.4585, + "step": 22393 + }, + { + "epoch": 0.67, + "learning_rate": 5.282020368518116e-05, + "loss": 0.5966, + "step": 22394 + }, + { + "epoch": 0.67, + "learning_rate": 5.281170212971447e-05, + "loss": 0.5688, + "step": 22395 + }, + { + "epoch": 0.67, + "learning_rate": 5.280320101298585e-05, + "loss": 0.5088, + "step": 22396 + }, + { + "epoch": 0.67, + "learning_rate": 5.279470033507416e-05, + "loss": 0.4487, + "step": 22397 + }, + { + "epoch": 0.67, + "learning_rate": 5.278620009605856e-05, + "loss": 0.4942, + "step": 22398 + }, + { + "epoch": 0.67, + "learning_rate": 5.277770029601804e-05, + "loss": 0.5226, + "step": 22399 + }, + { + "epoch": 0.67, + "learning_rate": 5.2769200935031684e-05, + "loss": 0.4698, + "step": 22400 + }, + { + "epoch": 0.67, + "learning_rate": 5.276070201317847e-05, + "loss": 0.5244, + "step": 22401 + }, + { + "epoch": 0.67, + "learning_rate": 5.275220353053738e-05, + "loss": 0.5661, + "step": 22402 + }, + { + "epoch": 0.67, + "learning_rate": 5.274370548718748e-05, + "loss": 0.5236, + "step": 22403 + }, + { + "epoch": 0.67, + "learning_rate": 5.27352078832078e-05, + "loss": 0.3759, + "step": 22404 + }, + { + "epoch": 0.67, + "learning_rate": 5.272671071867733e-05, + "loss": 0.5668, + "step": 22405 + }, + { + "epoch": 0.67, + "learning_rate": 5.271821399367501e-05, + "loss": 0.4896, + "step": 22406 + }, + { + "epoch": 0.67, + "learning_rate": 5.2709717708279905e-05, + "loss": 0.7349, + "step": 22407 + }, + { + "epoch": 0.67, + "learning_rate": 5.270122186257103e-05, + "loss": 0.5271, + "step": 22408 + }, + { + "epoch": 0.67, + "learning_rate": 5.2692726456627306e-05, + "loss": 0.4997, + "step": 22409 + }, + { + "epoch": 0.67, + "learning_rate": 5.268423149052779e-05, + "loss": 0.4464, + "step": 22410 + }, + { + "epoch": 0.67, + "learning_rate": 5.267573696435144e-05, + "loss": 0.4686, + "step": 22411 + }, + { + "epoch": 0.67, + "learning_rate": 5.266724287817717e-05, + "loss": 0.4914, + "step": 22412 + }, + { + "epoch": 0.67, + "learning_rate": 5.2658749232084024e-05, + "loss": 0.4869, + "step": 22413 + }, + { + "epoch": 0.67, + "learning_rate": 5.265025602615097e-05, + "loss": 0.6292, + "step": 22414 + }, + { + "epoch": 0.67, + "learning_rate": 5.264176326045702e-05, + "loss": 0.5132, + "step": 22415 + }, + { + "epoch": 0.67, + "learning_rate": 5.2633270935081013e-05, + "loss": 0.4315, + "step": 22416 + }, + { + "epoch": 0.67, + "learning_rate": 5.2624779050101966e-05, + "loss": 0.5332, + "step": 22417 + }, + { + "epoch": 0.67, + "learning_rate": 5.2616287605598854e-05, + "loss": 0.5184, + "step": 22418 + }, + { + "epoch": 0.67, + "learning_rate": 5.260779660165065e-05, + "loss": 0.4988, + "step": 22419 + }, + { + "epoch": 0.67, + "learning_rate": 5.259930603833627e-05, + "loss": 0.5549, + "step": 22420 + }, + { + "epoch": 0.67, + "learning_rate": 5.25908159157346e-05, + "loss": 0.5869, + "step": 22421 + }, + { + "epoch": 0.67, + "learning_rate": 5.2582326233924626e-05, + "loss": 0.5352, + "step": 22422 + }, + { + "epoch": 0.67, + "learning_rate": 5.257383699298532e-05, + "loss": 0.5969, + "step": 22423 + }, + { + "epoch": 0.67, + "learning_rate": 5.256534819299556e-05, + "loss": 0.4587, + "step": 22424 + }, + { + "epoch": 0.67, + "learning_rate": 5.255685983403431e-05, + "loss": 0.4953, + "step": 22425 + }, + { + "epoch": 0.67, + "learning_rate": 5.254837191618043e-05, + "loss": 0.5131, + "step": 22426 + }, + { + "epoch": 0.67, + "learning_rate": 5.253988443951292e-05, + "loss": 0.5025, + "step": 22427 + }, + { + "epoch": 0.67, + "learning_rate": 5.253139740411059e-05, + "loss": 0.4935, + "step": 22428 + }, + { + "epoch": 0.67, + "learning_rate": 5.252291081005247e-05, + "loss": 0.5026, + "step": 22429 + }, + { + "epoch": 0.67, + "learning_rate": 5.2514424657417363e-05, + "loss": 0.4577, + "step": 22430 + }, + { + "epoch": 0.67, + "learning_rate": 5.250593894628424e-05, + "loss": 0.609, + "step": 22431 + }, + { + "epoch": 0.67, + "learning_rate": 5.249745367673192e-05, + "loss": 0.6756, + "step": 22432 + }, + { + "epoch": 0.67, + "learning_rate": 5.248896884883936e-05, + "loss": 0.5203, + "step": 22433 + }, + { + "epoch": 0.67, + "learning_rate": 5.2480484462685455e-05, + "loss": 0.5951, + "step": 22434 + }, + { + "epoch": 0.67, + "learning_rate": 5.2472000518349085e-05, + "loss": 0.677, + "step": 22435 + }, + { + "epoch": 0.67, + "learning_rate": 5.246351701590906e-05, + "loss": 0.6014, + "step": 22436 + }, + { + "epoch": 0.67, + "learning_rate": 5.245503395544431e-05, + "loss": 0.4912, + "step": 22437 + }, + { + "epoch": 0.67, + "learning_rate": 5.244655133703373e-05, + "loss": 0.5283, + "step": 22438 + }, + { + "epoch": 0.67, + "learning_rate": 5.2438069160756185e-05, + "loss": 0.5056, + "step": 22439 + }, + { + "epoch": 0.67, + "learning_rate": 5.242958742669046e-05, + "loss": 0.5089, + "step": 22440 + }, + { + "epoch": 0.67, + "learning_rate": 5.242110613491549e-05, + "loss": 0.5323, + "step": 22441 + }, + { + "epoch": 0.67, + "learning_rate": 5.2412625285510096e-05, + "loss": 0.5517, + "step": 22442 + }, + { + "epoch": 0.67, + "learning_rate": 5.2404144878553194e-05, + "loss": 0.4906, + "step": 22443 + }, + { + "epoch": 0.67, + "learning_rate": 5.239566491412358e-05, + "loss": 0.4868, + "step": 22444 + }, + { + "epoch": 0.67, + "learning_rate": 5.238718539230006e-05, + "loss": 0.5114, + "step": 22445 + }, + { + "epoch": 0.67, + "learning_rate": 5.237870631316155e-05, + "loss": 0.5989, + "step": 22446 + }, + { + "epoch": 0.67, + "learning_rate": 5.237022767678682e-05, + "loss": 0.5835, + "step": 22447 + }, + { + "epoch": 0.67, + "learning_rate": 5.2361749483254764e-05, + "loss": 0.4719, + "step": 22448 + }, + { + "epoch": 0.67, + "learning_rate": 5.235327173264413e-05, + "loss": 0.5606, + "step": 22449 + }, + { + "epoch": 0.67, + "learning_rate": 5.2344794425033826e-05, + "loss": 0.4764, + "step": 22450 + }, + { + "epoch": 0.67, + "learning_rate": 5.233631756050259e-05, + "loss": 0.4699, + "step": 22451 + }, + { + "epoch": 0.67, + "learning_rate": 5.2327841139129284e-05, + "loss": 0.5683, + "step": 22452 + }, + { + "epoch": 0.67, + "learning_rate": 5.231936516099274e-05, + "loss": 0.4277, + "step": 22453 + }, + { + "epoch": 0.67, + "learning_rate": 5.2310889626171746e-05, + "loss": 0.5587, + "step": 22454 + }, + { + "epoch": 0.67, + "learning_rate": 5.2302414534745046e-05, + "loss": 0.6298, + "step": 22455 + }, + { + "epoch": 0.67, + "learning_rate": 5.22939398867915e-05, + "loss": 0.458, + "step": 22456 + }, + { + "epoch": 0.67, + "learning_rate": 5.2285465682389865e-05, + "loss": 0.4454, + "step": 22457 + }, + { + "epoch": 0.67, + "learning_rate": 5.2276991921619046e-05, + "loss": 0.6074, + "step": 22458 + }, + { + "epoch": 0.67, + "learning_rate": 5.226851860455765e-05, + "loss": 0.5108, + "step": 22459 + }, + { + "epoch": 0.67, + "learning_rate": 5.226004573128455e-05, + "loss": 0.4938, + "step": 22460 + }, + { + "epoch": 0.67, + "learning_rate": 5.225157330187851e-05, + "loss": 0.5646, + "step": 22461 + }, + { + "epoch": 0.67, + "learning_rate": 5.224310131641835e-05, + "loss": 0.4582, + "step": 22462 + }, + { + "epoch": 0.67, + "learning_rate": 5.2234629774982816e-05, + "loss": 0.5045, + "step": 22463 + }, + { + "epoch": 0.67, + "learning_rate": 5.222615867765059e-05, + "loss": 0.5916, + "step": 22464 + }, + { + "epoch": 0.67, + "learning_rate": 5.221768802450052e-05, + "loss": 0.5349, + "step": 22465 + }, + { + "epoch": 0.67, + "learning_rate": 5.220921781561139e-05, + "loss": 0.5182, + "step": 22466 + }, + { + "epoch": 0.67, + "learning_rate": 5.2200748051061844e-05, + "loss": 0.525, + "step": 22467 + }, + { + "epoch": 0.67, + "learning_rate": 5.219227873093074e-05, + "loss": 0.5284, + "step": 22468 + }, + { + "epoch": 0.67, + "learning_rate": 5.218380985529674e-05, + "loss": 0.5765, + "step": 22469 + }, + { + "epoch": 0.67, + "learning_rate": 5.2175341424238654e-05, + "loss": 0.4701, + "step": 22470 + }, + { + "epoch": 0.67, + "learning_rate": 5.216687343783514e-05, + "loss": 0.6811, + "step": 22471 + }, + { + "epoch": 0.67, + "learning_rate": 5.215840589616501e-05, + "loss": 0.5179, + "step": 22472 + }, + { + "epoch": 0.67, + "learning_rate": 5.214993879930692e-05, + "loss": 0.4111, + "step": 22473 + }, + { + "epoch": 0.67, + "learning_rate": 5.2141472147339666e-05, + "loss": 0.6342, + "step": 22474 + }, + { + "epoch": 0.67, + "learning_rate": 5.21330059403419e-05, + "loss": 0.5182, + "step": 22475 + }, + { + "epoch": 0.67, + "learning_rate": 5.2124540178392343e-05, + "loss": 0.5668, + "step": 22476 + }, + { + "epoch": 0.67, + "learning_rate": 5.2116074861569784e-05, + "loss": 0.4795, + "step": 22477 + }, + { + "epoch": 0.67, + "learning_rate": 5.210760998995288e-05, + "loss": 0.5794, + "step": 22478 + }, + { + "epoch": 0.67, + "learning_rate": 5.209914556362028e-05, + "loss": 0.5228, + "step": 22479 + }, + { + "epoch": 0.67, + "learning_rate": 5.209068158265075e-05, + "loss": 0.5834, + "step": 22480 + }, + { + "epoch": 0.67, + "learning_rate": 5.208221804712299e-05, + "loss": 0.446, + "step": 22481 + }, + { + "epoch": 0.67, + "learning_rate": 5.2073754957115684e-05, + "loss": 0.5074, + "step": 22482 + }, + { + "epoch": 0.67, + "learning_rate": 5.206529231270745e-05, + "loss": 0.5882, + "step": 22483 + }, + { + "epoch": 0.67, + "learning_rate": 5.205683011397702e-05, + "loss": 0.5883, + "step": 22484 + }, + { + "epoch": 0.67, + "learning_rate": 5.204836836100313e-05, + "loss": 0.5845, + "step": 22485 + }, + { + "epoch": 0.67, + "learning_rate": 5.2039907053864345e-05, + "loss": 0.6267, + "step": 22486 + }, + { + "epoch": 0.67, + "learning_rate": 5.203144619263945e-05, + "loss": 0.5807, + "step": 22487 + }, + { + "epoch": 0.67, + "learning_rate": 5.202298577740699e-05, + "loss": 0.554, + "step": 22488 + }, + { + "epoch": 0.67, + "learning_rate": 5.2014525808245726e-05, + "loss": 0.4958, + "step": 22489 + }, + { + "epoch": 0.67, + "learning_rate": 5.2006066285234256e-05, + "loss": 0.467, + "step": 22490 + }, + { + "epoch": 0.67, + "learning_rate": 5.199760720845127e-05, + "loss": 0.4546, + "step": 22491 + }, + { + "epoch": 0.67, + "learning_rate": 5.198914857797537e-05, + "loss": 0.5244, + "step": 22492 + }, + { + "epoch": 0.67, + "learning_rate": 5.198069039388527e-05, + "loss": 0.5096, + "step": 22493 + }, + { + "epoch": 0.67, + "learning_rate": 5.197223265625954e-05, + "loss": 0.593, + "step": 22494 + }, + { + "epoch": 0.67, + "learning_rate": 5.1963775365176846e-05, + "loss": 0.5733, + "step": 22495 + }, + { + "epoch": 0.67, + "learning_rate": 5.195531852071586e-05, + "loss": 0.4772, + "step": 22496 + }, + { + "epoch": 0.67, + "learning_rate": 5.1946862122955165e-05, + "loss": 0.485, + "step": 22497 + }, + { + "epoch": 0.67, + "learning_rate": 5.1938406171973364e-05, + "loss": 0.5195, + "step": 22498 + }, + { + "epoch": 0.67, + "learning_rate": 5.192995066784911e-05, + "loss": 0.4557, + "step": 22499 + }, + { + "epoch": 0.67, + "learning_rate": 5.1921495610661045e-05, + "loss": 0.7043, + "step": 22500 + }, + { + "epoch": 0.67, + "learning_rate": 5.1913041000487764e-05, + "loss": 0.4614, + "step": 22501 + }, + { + "epoch": 0.67, + "learning_rate": 5.190458683740782e-05, + "loss": 0.5388, + "step": 22502 + }, + { + "epoch": 0.67, + "learning_rate": 5.189613312149986e-05, + "loss": 0.5033, + "step": 22503 + }, + { + "epoch": 0.67, + "learning_rate": 5.1887679852842466e-05, + "loss": 0.5099, + "step": 22504 + }, + { + "epoch": 0.67, + "learning_rate": 5.18792270315143e-05, + "loss": 0.4686, + "step": 22505 + }, + { + "epoch": 0.67, + "learning_rate": 5.18707746575939e-05, + "loss": 0.4902, + "step": 22506 + }, + { + "epoch": 0.67, + "learning_rate": 5.1862322731159816e-05, + "loss": 0.5608, + "step": 22507 + }, + { + "epoch": 0.67, + "learning_rate": 5.185387125229066e-05, + "loss": 0.4898, + "step": 22508 + }, + { + "epoch": 0.67, + "learning_rate": 5.184542022106509e-05, + "loss": 0.51, + "step": 22509 + }, + { + "epoch": 0.67, + "learning_rate": 5.1836969637561585e-05, + "loss": 0.4071, + "step": 22510 + }, + { + "epoch": 0.67, + "learning_rate": 5.18285195018587e-05, + "loss": 0.6109, + "step": 22511 + }, + { + "epoch": 0.67, + "learning_rate": 5.182006981403509e-05, + "loss": 0.5409, + "step": 22512 + }, + { + "epoch": 0.67, + "learning_rate": 5.181162057416922e-05, + "loss": 0.5623, + "step": 22513 + }, + { + "epoch": 0.67, + "learning_rate": 5.180317178233971e-05, + "loss": 0.5261, + "step": 22514 + }, + { + "epoch": 0.67, + "learning_rate": 5.1794723438625126e-05, + "loss": 0.5139, + "step": 22515 + }, + { + "epoch": 0.67, + "learning_rate": 5.1786275543104e-05, + "loss": 0.6566, + "step": 22516 + }, + { + "epoch": 0.67, + "learning_rate": 5.1777828095854816e-05, + "loss": 0.5254, + "step": 22517 + }, + { + "epoch": 0.67, + "learning_rate": 5.176938109695617e-05, + "loss": 0.5234, + "step": 22518 + }, + { + "epoch": 0.67, + "learning_rate": 5.1760934546486603e-05, + "loss": 0.5922, + "step": 22519 + }, + { + "epoch": 0.67, + "learning_rate": 5.175248844452472e-05, + "loss": 0.5708, + "step": 22520 + }, + { + "epoch": 0.67, + "learning_rate": 5.1744042791148884e-05, + "loss": 0.4779, + "step": 22521 + }, + { + "epoch": 0.67, + "learning_rate": 5.173559758643771e-05, + "loss": 0.5047, + "step": 22522 + }, + { + "epoch": 0.67, + "learning_rate": 5.1727152830469716e-05, + "loss": 0.6657, + "step": 22523 + }, + { + "epoch": 0.67, + "learning_rate": 5.1718708523323455e-05, + "loss": 0.5472, + "step": 22524 + }, + { + "epoch": 0.67, + "learning_rate": 5.171026466507739e-05, + "loss": 0.508, + "step": 22525 + }, + { + "epoch": 0.67, + "learning_rate": 5.170182125581e-05, + "loss": 0.4937, + "step": 22526 + }, + { + "epoch": 0.67, + "learning_rate": 5.169337829559984e-05, + "loss": 0.517, + "step": 22527 + }, + { + "epoch": 0.67, + "learning_rate": 5.1684935784525425e-05, + "loss": 0.6123, + "step": 22528 + }, + { + "epoch": 0.67, + "learning_rate": 5.167649372266517e-05, + "loss": 0.6046, + "step": 22529 + }, + { + "epoch": 0.67, + "learning_rate": 5.1668052110097664e-05, + "loss": 0.5542, + "step": 22530 + }, + { + "epoch": 0.67, + "learning_rate": 5.165961094690131e-05, + "loss": 0.5345, + "step": 22531 + }, + { + "epoch": 0.67, + "learning_rate": 5.1651170233154664e-05, + "loss": 0.4895, + "step": 22532 + }, + { + "epoch": 0.67, + "learning_rate": 5.164272996893613e-05, + "loss": 0.4953, + "step": 22533 + }, + { + "epoch": 0.67, + "learning_rate": 5.163429015432426e-05, + "loss": 0.5512, + "step": 22534 + }, + { + "epoch": 0.67, + "learning_rate": 5.1625850789397444e-05, + "loss": 0.6118, + "step": 22535 + }, + { + "epoch": 0.67, + "learning_rate": 5.161741187423423e-05, + "loss": 0.5336, + "step": 22536 + }, + { + "epoch": 0.67, + "learning_rate": 5.1608973408912995e-05, + "loss": 0.4761, + "step": 22537 + }, + { + "epoch": 0.67, + "learning_rate": 5.160053539351224e-05, + "loss": 0.5594, + "step": 22538 + }, + { + "epoch": 0.67, + "learning_rate": 5.159209782811044e-05, + "loss": 0.3977, + "step": 22539 + }, + { + "epoch": 0.67, + "learning_rate": 5.158366071278604e-05, + "loss": 0.503, + "step": 22540 + }, + { + "epoch": 0.67, + "learning_rate": 5.157522404761741e-05, + "loss": 0.5223, + "step": 22541 + }, + { + "epoch": 0.67, + "learning_rate": 5.1566787832683064e-05, + "loss": 0.5335, + "step": 22542 + }, + { + "epoch": 0.67, + "learning_rate": 5.1558352068061455e-05, + "loss": 0.5576, + "step": 22543 + }, + { + "epoch": 0.67, + "learning_rate": 5.1549916753830974e-05, + "loss": 0.5536, + "step": 22544 + }, + { + "epoch": 0.67, + "learning_rate": 5.154148189007002e-05, + "loss": 0.6885, + "step": 22545 + }, + { + "epoch": 0.67, + "learning_rate": 5.153304747685706e-05, + "loss": 0.5342, + "step": 22546 + }, + { + "epoch": 0.67, + "learning_rate": 5.152461351427054e-05, + "loss": 0.3668, + "step": 22547 + }, + { + "epoch": 0.67, + "learning_rate": 5.1516180002388814e-05, + "loss": 0.5136, + "step": 22548 + }, + { + "epoch": 0.67, + "learning_rate": 5.150774694129035e-05, + "loss": 0.542, + "step": 22549 + }, + { + "epoch": 0.67, + "learning_rate": 5.149931433105349e-05, + "loss": 0.5411, + "step": 22550 + }, + { + "epoch": 0.67, + "learning_rate": 5.149088217175673e-05, + "loss": 0.6214, + "step": 22551 + }, + { + "epoch": 0.67, + "learning_rate": 5.148245046347835e-05, + "loss": 0.5929, + "step": 22552 + }, + { + "epoch": 0.67, + "learning_rate": 5.147401920629687e-05, + "loss": 0.4992, + "step": 22553 + }, + { + "epoch": 0.67, + "learning_rate": 5.1465588400290564e-05, + "loss": 0.4781, + "step": 22554 + }, + { + "epoch": 0.67, + "learning_rate": 5.1457158045537925e-05, + "loss": 0.5767, + "step": 22555 + }, + { + "epoch": 0.67, + "learning_rate": 5.144872814211723e-05, + "loss": 0.6565, + "step": 22556 + }, + { + "epoch": 0.67, + "learning_rate": 5.144029869010691e-05, + "loss": 0.5879, + "step": 22557 + }, + { + "epoch": 0.67, + "learning_rate": 5.14318696895854e-05, + "loss": 0.5973, + "step": 22558 + }, + { + "epoch": 0.67, + "learning_rate": 5.142344114063098e-05, + "loss": 0.6022, + "step": 22559 + }, + { + "epoch": 0.67, + "learning_rate": 5.1415013043322026e-05, + "loss": 0.4409, + "step": 22560 + }, + { + "epoch": 0.67, + "learning_rate": 5.1406585397736906e-05, + "loss": 0.6126, + "step": 22561 + }, + { + "epoch": 0.67, + "learning_rate": 5.1398158203954e-05, + "loss": 0.5468, + "step": 22562 + }, + { + "epoch": 0.67, + "learning_rate": 5.138973146205171e-05, + "loss": 0.4578, + "step": 22563 + }, + { + "epoch": 0.67, + "learning_rate": 5.138130517210824e-05, + "loss": 0.5688, + "step": 22564 + }, + { + "epoch": 0.67, + "learning_rate": 5.1372879334202026e-05, + "loss": 0.4449, + "step": 22565 + }, + { + "epoch": 0.67, + "learning_rate": 5.1364453948411404e-05, + "loss": 0.5072, + "step": 22566 + }, + { + "epoch": 0.67, + "learning_rate": 5.1356029014814756e-05, + "loss": 0.4949, + "step": 22567 + }, + { + "epoch": 0.67, + "learning_rate": 5.1347604533490346e-05, + "loss": 0.6191, + "step": 22568 + }, + { + "epoch": 0.67, + "learning_rate": 5.133918050451647e-05, + "loss": 0.5191, + "step": 22569 + }, + { + "epoch": 0.67, + "learning_rate": 5.133075692797151e-05, + "loss": 0.636, + "step": 22570 + }, + { + "epoch": 0.67, + "learning_rate": 5.132233380393382e-05, + "loss": 0.5771, + "step": 22571 + }, + { + "epoch": 0.67, + "learning_rate": 5.1313911132481666e-05, + "loss": 0.5199, + "step": 22572 + }, + { + "epoch": 0.67, + "learning_rate": 5.1305488913693314e-05, + "loss": 0.5967, + "step": 22573 + }, + { + "epoch": 0.67, + "learning_rate": 5.129706714764713e-05, + "loss": 0.498, + "step": 22574 + }, + { + "epoch": 0.67, + "learning_rate": 5.128864583442143e-05, + "loss": 0.5817, + "step": 22575 + }, + { + "epoch": 0.67, + "learning_rate": 5.128022497409446e-05, + "loss": 0.5317, + "step": 22576 + }, + { + "epoch": 0.67, + "learning_rate": 5.127180456674457e-05, + "loss": 0.5709, + "step": 22577 + }, + { + "epoch": 0.67, + "learning_rate": 5.126338461245003e-05, + "loss": 0.5156, + "step": 22578 + }, + { + "epoch": 0.67, + "learning_rate": 5.1254965111289064e-05, + "loss": 0.464, + "step": 22579 + }, + { + "epoch": 0.67, + "learning_rate": 5.1246546063340016e-05, + "loss": 0.4712, + "step": 22580 + }, + { + "epoch": 0.67, + "learning_rate": 5.123812746868115e-05, + "loss": 0.5869, + "step": 22581 + }, + { + "epoch": 0.67, + "learning_rate": 5.122970932739082e-05, + "loss": 0.5727, + "step": 22582 + }, + { + "epoch": 0.67, + "learning_rate": 5.1221291639547117e-05, + "loss": 0.5637, + "step": 22583 + }, + { + "epoch": 0.67, + "learning_rate": 5.121287440522843e-05, + "loss": 0.4934, + "step": 22584 + }, + { + "epoch": 0.67, + "learning_rate": 5.120445762451298e-05, + "loss": 0.62, + "step": 22585 + }, + { + "epoch": 0.67, + "learning_rate": 5.119604129747907e-05, + "loss": 0.5831, + "step": 22586 + }, + { + "epoch": 0.67, + "learning_rate": 5.118762542420492e-05, + "loss": 0.5019, + "step": 22587 + }, + { + "epoch": 0.67, + "learning_rate": 5.117921000476874e-05, + "loss": 0.5749, + "step": 22588 + }, + { + "epoch": 0.67, + "learning_rate": 5.11707950392488e-05, + "loss": 0.5875, + "step": 22589 + }, + { + "epoch": 0.67, + "learning_rate": 5.1162380527723395e-05, + "loss": 0.5086, + "step": 22590 + }, + { + "epoch": 0.67, + "learning_rate": 5.1153966470270665e-05, + "loss": 0.5897, + "step": 22591 + }, + { + "epoch": 0.67, + "learning_rate": 5.1145552866968914e-05, + "loss": 0.4646, + "step": 22592 + }, + { + "epoch": 0.67, + "learning_rate": 5.113713971789633e-05, + "loss": 0.4792, + "step": 22593 + }, + { + "epoch": 0.67, + "learning_rate": 5.112872702313116e-05, + "loss": 0.5105, + "step": 22594 + }, + { + "epoch": 0.67, + "learning_rate": 5.112031478275159e-05, + "loss": 0.4887, + "step": 22595 + }, + { + "epoch": 0.67, + "learning_rate": 5.111190299683587e-05, + "loss": 0.5656, + "step": 22596 + }, + { + "epoch": 0.67, + "learning_rate": 5.1103491665462157e-05, + "loss": 0.6356, + "step": 22597 + }, + { + "epoch": 0.67, + "learning_rate": 5.1095080788708735e-05, + "loss": 0.6014, + "step": 22598 + }, + { + "epoch": 0.67, + "learning_rate": 5.108667036665372e-05, + "loss": 0.5224, + "step": 22599 + }, + { + "epoch": 0.67, + "learning_rate": 5.1078260399375356e-05, + "loss": 0.6708, + "step": 22600 + }, + { + "epoch": 0.67, + "learning_rate": 5.106985088695186e-05, + "loss": 0.4981, + "step": 22601 + }, + { + "epoch": 0.67, + "learning_rate": 5.106144182946139e-05, + "loss": 0.5495, + "step": 22602 + }, + { + "epoch": 0.67, + "learning_rate": 5.1053033226982086e-05, + "loss": 0.4879, + "step": 22603 + }, + { + "epoch": 0.67, + "learning_rate": 5.104462507959218e-05, + "loss": 0.6165, + "step": 22604 + }, + { + "epoch": 0.67, + "learning_rate": 5.103621738736987e-05, + "loss": 0.5332, + "step": 22605 + }, + { + "epoch": 0.67, + "learning_rate": 5.102781015039328e-05, + "loss": 0.5108, + "step": 22606 + }, + { + "epoch": 0.67, + "learning_rate": 5.1019403368740584e-05, + "loss": 0.4225, + "step": 22607 + }, + { + "epoch": 0.67, + "learning_rate": 5.101099704248994e-05, + "loss": 0.4729, + "step": 22608 + }, + { + "epoch": 0.67, + "learning_rate": 5.100259117171956e-05, + "loss": 0.5356, + "step": 22609 + }, + { + "epoch": 0.67, + "learning_rate": 5.099418575650752e-05, + "loss": 0.5273, + "step": 22610 + }, + { + "epoch": 0.67, + "learning_rate": 5.098578079693205e-05, + "loss": 0.5471, + "step": 22611 + }, + { + "epoch": 0.67, + "learning_rate": 5.097737629307121e-05, + "loss": 0.6915, + "step": 22612 + }, + { + "epoch": 0.67, + "learning_rate": 5.096897224500323e-05, + "loss": 0.5455, + "step": 22613 + }, + { + "epoch": 0.67, + "learning_rate": 5.0960568652806165e-05, + "loss": 0.5525, + "step": 22614 + }, + { + "epoch": 0.67, + "learning_rate": 5.0952165516558226e-05, + "loss": 0.553, + "step": 22615 + }, + { + "epoch": 0.67, + "learning_rate": 5.094376283633746e-05, + "loss": 0.4667, + "step": 22616 + }, + { + "epoch": 0.67, + "learning_rate": 5.093536061222208e-05, + "loss": 0.6637, + "step": 22617 + }, + { + "epoch": 0.67, + "learning_rate": 5.0926958844290126e-05, + "loss": 0.4902, + "step": 22618 + }, + { + "epoch": 0.67, + "learning_rate": 5.091855753261974e-05, + "loss": 0.492, + "step": 22619 + }, + { + "epoch": 0.67, + "learning_rate": 5.091015667728909e-05, + "loss": 0.4775, + "step": 22620 + }, + { + "epoch": 0.67, + "learning_rate": 5.0901756278376234e-05, + "loss": 0.5507, + "step": 22621 + }, + { + "epoch": 0.67, + "learning_rate": 5.089335633595924e-05, + "loss": 0.6087, + "step": 22622 + }, + { + "epoch": 0.67, + "learning_rate": 5.088495685011626e-05, + "loss": 0.5092, + "step": 22623 + }, + { + "epoch": 0.67, + "learning_rate": 5.087655782092536e-05, + "loss": 0.5467, + "step": 22624 + }, + { + "epoch": 0.67, + "learning_rate": 5.086815924846473e-05, + "loss": 0.5143, + "step": 22625 + }, + { + "epoch": 0.67, + "learning_rate": 5.0859761132812303e-05, + "loss": 0.5808, + "step": 22626 + }, + { + "epoch": 0.67, + "learning_rate": 5.085136347404622e-05, + "loss": 0.5776, + "step": 22627 + }, + { + "epoch": 0.67, + "learning_rate": 5.084296627224458e-05, + "loss": 0.6142, + "step": 22628 + }, + { + "epoch": 0.67, + "learning_rate": 5.083456952748549e-05, + "loss": 0.5858, + "step": 22629 + }, + { + "epoch": 0.67, + "learning_rate": 5.082617323984695e-05, + "loss": 0.5601, + "step": 22630 + }, + { + "epoch": 0.67, + "learning_rate": 5.081777740940704e-05, + "loss": 0.4529, + "step": 22631 + }, + { + "epoch": 0.67, + "learning_rate": 5.0809382036243834e-05, + "loss": 0.3977, + "step": 22632 + }, + { + "epoch": 0.67, + "learning_rate": 5.0800987120435415e-05, + "loss": 0.6065, + "step": 22633 + }, + { + "epoch": 0.67, + "learning_rate": 5.079259266205978e-05, + "loss": 0.4834, + "step": 22634 + }, + { + "epoch": 0.67, + "learning_rate": 5.078419866119504e-05, + "loss": 0.482, + "step": 22635 + }, + { + "epoch": 0.67, + "learning_rate": 5.077580511791915e-05, + "loss": 0.4859, + "step": 22636 + }, + { + "epoch": 0.67, + "learning_rate": 5.0767412032310266e-05, + "loss": 0.4764, + "step": 22637 + }, + { + "epoch": 0.67, + "learning_rate": 5.075901940444632e-05, + "loss": 0.4813, + "step": 22638 + }, + { + "epoch": 0.67, + "learning_rate": 5.075062723440543e-05, + "loss": 0.6001, + "step": 22639 + }, + { + "epoch": 0.67, + "learning_rate": 5.0742235522265525e-05, + "loss": 0.5161, + "step": 22640 + }, + { + "epoch": 0.67, + "learning_rate": 5.0733844268104724e-05, + "loss": 0.4924, + "step": 22641 + }, + { + "epoch": 0.67, + "learning_rate": 5.072545347200097e-05, + "loss": 0.6019, + "step": 22642 + }, + { + "epoch": 0.67, + "learning_rate": 5.071706313403232e-05, + "loss": 0.5043, + "step": 22643 + }, + { + "epoch": 0.67, + "learning_rate": 5.0708673254276794e-05, + "loss": 0.5573, + "step": 22644 + }, + { + "epoch": 0.67, + "learning_rate": 5.07002838328124e-05, + "loss": 0.5381, + "step": 22645 + }, + { + "epoch": 0.67, + "learning_rate": 5.069189486971706e-05, + "loss": 0.5126, + "step": 22646 + }, + { + "epoch": 0.67, + "learning_rate": 5.068350636506884e-05, + "loss": 0.5388, + "step": 22647 + }, + { + "epoch": 0.67, + "learning_rate": 5.067511831894576e-05, + "loss": 0.4843, + "step": 22648 + }, + { + "epoch": 0.67, + "learning_rate": 5.066673073142576e-05, + "loss": 0.571, + "step": 22649 + }, + { + "epoch": 0.67, + "learning_rate": 5.065834360258681e-05, + "loss": 0.5623, + "step": 22650 + }, + { + "epoch": 0.67, + "learning_rate": 5.064995693250692e-05, + "loss": 0.5408, + "step": 22651 + }, + { + "epoch": 0.67, + "learning_rate": 5.064157072126408e-05, + "loss": 0.5384, + "step": 22652 + }, + { + "epoch": 0.67, + "learning_rate": 5.063318496893622e-05, + "loss": 0.5639, + "step": 22653 + }, + { + "epoch": 0.67, + "learning_rate": 5.062479967560137e-05, + "loss": 0.4586, + "step": 22654 + }, + { + "epoch": 0.67, + "learning_rate": 5.06164148413374e-05, + "loss": 0.6161, + "step": 22655 + }, + { + "epoch": 0.67, + "learning_rate": 5.060803046622238e-05, + "loss": 0.6186, + "step": 22656 + }, + { + "epoch": 0.67, + "learning_rate": 5.059964655033415e-05, + "loss": 0.4821, + "step": 22657 + }, + { + "epoch": 0.67, + "learning_rate": 5.0591263093750776e-05, + "loss": 0.4877, + "step": 22658 + }, + { + "epoch": 0.67, + "learning_rate": 5.058288009655008e-05, + "loss": 0.5994, + "step": 22659 + }, + { + "epoch": 0.67, + "learning_rate": 5.057449755881013e-05, + "loss": 0.4735, + "step": 22660 + }, + { + "epoch": 0.67, + "learning_rate": 5.0566115480608755e-05, + "loss": 0.6364, + "step": 22661 + }, + { + "epoch": 0.67, + "learning_rate": 5.0557733862023935e-05, + "loss": 0.5588, + "step": 22662 + }, + { + "epoch": 0.67, + "learning_rate": 5.054935270313365e-05, + "loss": 0.4859, + "step": 22663 + }, + { + "epoch": 0.67, + "learning_rate": 5.054097200401576e-05, + "loss": 0.596, + "step": 22664 + }, + { + "epoch": 0.67, + "learning_rate": 5.053259176474817e-05, + "loss": 0.4218, + "step": 22665 + }, + { + "epoch": 0.67, + "learning_rate": 5.052421198540882e-05, + "loss": 0.5714, + "step": 22666 + }, + { + "epoch": 0.67, + "learning_rate": 5.051583266607567e-05, + "loss": 0.4282, + "step": 22667 + }, + { + "epoch": 0.67, + "learning_rate": 5.0507453806826586e-05, + "loss": 0.5202, + "step": 22668 + }, + { + "epoch": 0.67, + "learning_rate": 5.049907540773944e-05, + "loss": 0.5909, + "step": 22669 + }, + { + "epoch": 0.67, + "learning_rate": 5.049069746889214e-05, + "loss": 0.5342, + "step": 22670 + }, + { + "epoch": 0.67, + "learning_rate": 5.0482319990362626e-05, + "loss": 0.52, + "step": 22671 + }, + { + "epoch": 0.67, + "learning_rate": 5.0473942972228794e-05, + "loss": 0.5755, + "step": 22672 + }, + { + "epoch": 0.68, + "learning_rate": 5.04655664145685e-05, + "loss": 0.5569, + "step": 22673 + }, + { + "epoch": 0.68, + "learning_rate": 5.045719031745959e-05, + "loss": 0.482, + "step": 22674 + }, + { + "epoch": 0.68, + "learning_rate": 5.0448814680980014e-05, + "loss": 0.5483, + "step": 22675 + }, + { + "epoch": 0.68, + "learning_rate": 5.044043950520757e-05, + "loss": 0.623, + "step": 22676 + }, + { + "epoch": 0.68, + "learning_rate": 5.0432064790220214e-05, + "loss": 0.5306, + "step": 22677 + }, + { + "epoch": 0.68, + "learning_rate": 5.042369053609573e-05, + "loss": 0.5533, + "step": 22678 + }, + { + "epoch": 0.68, + "learning_rate": 5.041531674291202e-05, + "loss": 0.515, + "step": 22679 + }, + { + "epoch": 0.68, + "learning_rate": 5.040694341074692e-05, + "loss": 0.477, + "step": 22680 + }, + { + "epoch": 0.68, + "learning_rate": 5.0398570539678294e-05, + "loss": 0.4625, + "step": 22681 + }, + { + "epoch": 0.68, + "learning_rate": 5.039019812978403e-05, + "loss": 0.4768, + "step": 22682 + }, + { + "epoch": 0.68, + "learning_rate": 5.038182618114191e-05, + "loss": 0.4325, + "step": 22683 + }, + { + "epoch": 0.68, + "learning_rate": 5.037345469382977e-05, + "loss": 0.5813, + "step": 22684 + }, + { + "epoch": 0.68, + "learning_rate": 5.036508366792547e-05, + "loss": 0.5402, + "step": 22685 + }, + { + "epoch": 0.68, + "learning_rate": 5.0356713103506844e-05, + "loss": 0.4802, + "step": 22686 + }, + { + "epoch": 0.68, + "learning_rate": 5.0348343000651786e-05, + "loss": 0.4493, + "step": 22687 + }, + { + "epoch": 0.68, + "learning_rate": 5.0339973359437966e-05, + "loss": 0.4551, + "step": 22688 + }, + { + "epoch": 0.68, + "learning_rate": 5.0331604179943295e-05, + "loss": 0.6564, + "step": 22689 + }, + { + "epoch": 0.68, + "learning_rate": 5.032323546224555e-05, + "loss": 0.4575, + "step": 22690 + }, + { + "epoch": 0.68, + "learning_rate": 5.0314867206422625e-05, + "loss": 0.4744, + "step": 22691 + }, + { + "epoch": 0.68, + "learning_rate": 5.030649941255223e-05, + "loss": 0.4332, + "step": 22692 + }, + { + "epoch": 0.68, + "learning_rate": 5.029813208071219e-05, + "loss": 0.5565, + "step": 22693 + }, + { + "epoch": 0.68, + "learning_rate": 5.02897652109803e-05, + "loss": 0.4825, + "step": 22694 + }, + { + "epoch": 0.68, + "learning_rate": 5.028139880343439e-05, + "loss": 0.5367, + "step": 22695 + }, + { + "epoch": 0.68, + "learning_rate": 5.027303285815219e-05, + "loss": 0.6484, + "step": 22696 + }, + { + "epoch": 0.68, + "learning_rate": 5.0264667375211536e-05, + "loss": 0.4808, + "step": 22697 + }, + { + "epoch": 0.68, + "learning_rate": 5.025630235469017e-05, + "loss": 0.5474, + "step": 22698 + }, + { + "epoch": 0.68, + "learning_rate": 5.0247937796665886e-05, + "loss": 0.629, + "step": 22699 + }, + { + "epoch": 0.68, + "learning_rate": 5.0239573701216434e-05, + "loss": 0.5285, + "step": 22700 + }, + { + "epoch": 0.68, + "learning_rate": 5.0231210068419623e-05, + "loss": 0.5267, + "step": 22701 + }, + { + "epoch": 0.68, + "learning_rate": 5.022284689835314e-05, + "loss": 0.5765, + "step": 22702 + }, + { + "epoch": 0.68, + "learning_rate": 5.021448419109482e-05, + "loss": 0.4662, + "step": 22703 + }, + { + "epoch": 0.68, + "learning_rate": 5.020612194672236e-05, + "loss": 0.6008, + "step": 22704 + }, + { + "epoch": 0.68, + "learning_rate": 5.019776016531352e-05, + "loss": 0.566, + "step": 22705 + }, + { + "epoch": 0.68, + "learning_rate": 5.0189398846946115e-05, + "loss": 0.4553, + "step": 22706 + }, + { + "epoch": 0.68, + "learning_rate": 5.0181037991697813e-05, + "loss": 0.4885, + "step": 22707 + }, + { + "epoch": 0.68, + "learning_rate": 5.017267759964632e-05, + "loss": 0.4077, + "step": 22708 + }, + { + "epoch": 0.68, + "learning_rate": 5.0164317670869423e-05, + "loss": 0.5107, + "step": 22709 + }, + { + "epoch": 0.68, + "learning_rate": 5.015595820544488e-05, + "loss": 0.5153, + "step": 22710 + }, + { + "epoch": 0.68, + "learning_rate": 5.014759920345036e-05, + "loss": 0.6321, + "step": 22711 + }, + { + "epoch": 0.68, + "learning_rate": 5.013924066496356e-05, + "loss": 0.5238, + "step": 22712 + }, + { + "epoch": 0.68, + "learning_rate": 5.013088259006222e-05, + "loss": 0.5497, + "step": 22713 + }, + { + "epoch": 0.68, + "learning_rate": 5.0122524978824103e-05, + "loss": 0.5355, + "step": 22714 + }, + { + "epoch": 0.68, + "learning_rate": 5.011416783132683e-05, + "loss": 0.4912, + "step": 22715 + }, + { + "epoch": 0.68, + "learning_rate": 5.0105811147648184e-05, + "loss": 0.5219, + "step": 22716 + }, + { + "epoch": 0.68, + "learning_rate": 5.009745492786577e-05, + "loss": 0.5178, + "step": 22717 + }, + { + "epoch": 0.68, + "learning_rate": 5.0089099172057374e-05, + "loss": 0.596, + "step": 22718 + }, + { + "epoch": 0.68, + "learning_rate": 5.008074388030062e-05, + "loss": 0.4917, + "step": 22719 + }, + { + "epoch": 0.68, + "learning_rate": 5.007238905267323e-05, + "loss": 0.6323, + "step": 22720 + }, + { + "epoch": 0.68, + "learning_rate": 5.006403468925285e-05, + "loss": 0.5924, + "step": 22721 + }, + { + "epoch": 0.68, + "learning_rate": 5.0055680790117186e-05, + "loss": 0.6101, + "step": 22722 + }, + { + "epoch": 0.68, + "learning_rate": 5.004732735534388e-05, + "loss": 0.5636, + "step": 22723 + }, + { + "epoch": 0.68, + "learning_rate": 5.0038974385010595e-05, + "loss": 0.451, + "step": 22724 + }, + { + "epoch": 0.68, + "learning_rate": 5.003062187919507e-05, + "loss": 0.5904, + "step": 22725 + }, + { + "epoch": 0.68, + "learning_rate": 5.00222698379749e-05, + "loss": 0.5392, + "step": 22726 + }, + { + "epoch": 0.68, + "learning_rate": 5.001391826142771e-05, + "loss": 0.4188, + "step": 22727 + }, + { + "epoch": 0.68, + "learning_rate": 5.000556714963118e-05, + "loss": 0.4951, + "step": 22728 + }, + { + "epoch": 0.68, + "learning_rate": 4.999721650266296e-05, + "loss": 0.4676, + "step": 22729 + }, + { + "epoch": 0.68, + "learning_rate": 4.998886632060077e-05, + "loss": 0.6156, + "step": 22730 + }, + { + "epoch": 0.68, + "learning_rate": 4.99805166035221e-05, + "loss": 0.4698, + "step": 22731 + }, + { + "epoch": 0.68, + "learning_rate": 4.997216735150464e-05, + "loss": 0.5405, + "step": 22732 + }, + { + "epoch": 0.68, + "learning_rate": 4.9963818564626044e-05, + "loss": 0.5611, + "step": 22733 + }, + { + "epoch": 0.68, + "learning_rate": 4.9955470242963945e-05, + "loss": 0.55, + "step": 22734 + }, + { + "epoch": 0.68, + "learning_rate": 4.9947122386595936e-05, + "loss": 0.5292, + "step": 22735 + }, + { + "epoch": 0.68, + "learning_rate": 4.9938774995599605e-05, + "loss": 0.5198, + "step": 22736 + }, + { + "epoch": 0.68, + "learning_rate": 4.993042807005258e-05, + "loss": 0.4899, + "step": 22737 + }, + { + "epoch": 0.68, + "learning_rate": 4.9922081610032535e-05, + "loss": 0.4905, + "step": 22738 + }, + { + "epoch": 0.68, + "learning_rate": 4.9913735615616965e-05, + "loss": 0.5536, + "step": 22739 + }, + { + "epoch": 0.68, + "learning_rate": 4.9905390086883555e-05, + "loss": 0.4586, + "step": 22740 + }, + { + "epoch": 0.68, + "learning_rate": 4.989704502390987e-05, + "loss": 0.525, + "step": 22741 + }, + { + "epoch": 0.68, + "learning_rate": 4.9888700426773446e-05, + "loss": 0.5811, + "step": 22742 + }, + { + "epoch": 0.68, + "learning_rate": 4.9880356295551924e-05, + "loss": 0.4995, + "step": 22743 + }, + { + "epoch": 0.68, + "learning_rate": 4.987201263032291e-05, + "loss": 0.6015, + "step": 22744 + }, + { + "epoch": 0.68, + "learning_rate": 4.986366943116393e-05, + "loss": 0.5999, + "step": 22745 + }, + { + "epoch": 0.68, + "learning_rate": 4.9855326698152536e-05, + "loss": 0.49, + "step": 22746 + }, + { + "epoch": 0.68, + "learning_rate": 4.9846984431366336e-05, + "loss": 0.5335, + "step": 22747 + }, + { + "epoch": 0.68, + "learning_rate": 4.983864263088288e-05, + "loss": 0.5744, + "step": 22748 + }, + { + "epoch": 0.68, + "learning_rate": 4.983030129677981e-05, + "loss": 0.4534, + "step": 22749 + }, + { + "epoch": 0.68, + "learning_rate": 4.982196042913453e-05, + "loss": 0.7071, + "step": 22750 + }, + { + "epoch": 0.68, + "learning_rate": 4.981362002802466e-05, + "loss": 0.5729, + "step": 22751 + }, + { + "epoch": 0.68, + "learning_rate": 4.980528009352775e-05, + "loss": 0.4118, + "step": 22752 + }, + { + "epoch": 0.68, + "learning_rate": 4.979694062572138e-05, + "loss": 0.5367, + "step": 22753 + }, + { + "epoch": 0.68, + "learning_rate": 4.978860162468304e-05, + "loss": 0.4252, + "step": 22754 + }, + { + "epoch": 0.68, + "learning_rate": 4.9780263090490245e-05, + "loss": 0.5137, + "step": 22755 + }, + { + "epoch": 0.68, + "learning_rate": 4.977192502322053e-05, + "loss": 0.5051, + "step": 22756 + }, + { + "epoch": 0.68, + "learning_rate": 4.9763587422951486e-05, + "loss": 0.4418, + "step": 22757 + }, + { + "epoch": 0.68, + "learning_rate": 4.975525028976055e-05, + "loss": 0.5748, + "step": 22758 + }, + { + "epoch": 0.68, + "learning_rate": 4.97469136237253e-05, + "loss": 0.5518, + "step": 22759 + }, + { + "epoch": 0.68, + "learning_rate": 4.973857742492318e-05, + "loss": 0.5548, + "step": 22760 + }, + { + "epoch": 0.68, + "learning_rate": 4.973024169343178e-05, + "loss": 0.4792, + "step": 22761 + }, + { + "epoch": 0.68, + "learning_rate": 4.9721906429328514e-05, + "loss": 0.4786, + "step": 22762 + }, + { + "epoch": 0.68, + "learning_rate": 4.971357163269096e-05, + "loss": 0.5598, + "step": 22763 + }, + { + "epoch": 0.68, + "learning_rate": 4.970523730359654e-05, + "loss": 0.5128, + "step": 22764 + }, + { + "epoch": 0.68, + "learning_rate": 4.969690344212282e-05, + "loss": 0.7171, + "step": 22765 + }, + { + "epoch": 0.68, + "learning_rate": 4.968857004834719e-05, + "loss": 0.6354, + "step": 22766 + }, + { + "epoch": 0.68, + "learning_rate": 4.968023712234721e-05, + "loss": 0.4785, + "step": 22767 + }, + { + "epoch": 0.68, + "learning_rate": 4.967190466420034e-05, + "loss": 0.6153, + "step": 22768 + }, + { + "epoch": 0.68, + "learning_rate": 4.966357267398406e-05, + "loss": 0.6296, + "step": 22769 + }, + { + "epoch": 0.68, + "learning_rate": 4.965524115177578e-05, + "loss": 0.5324, + "step": 22770 + }, + { + "epoch": 0.68, + "learning_rate": 4.9646910097652986e-05, + "loss": 0.4748, + "step": 22771 + }, + { + "epoch": 0.68, + "learning_rate": 4.9638579511693205e-05, + "loss": 0.585, + "step": 22772 + }, + { + "epoch": 0.68, + "learning_rate": 4.963024939397384e-05, + "loss": 0.5703, + "step": 22773 + }, + { + "epoch": 0.68, + "learning_rate": 4.96219197445723e-05, + "loss": 0.5471, + "step": 22774 + }, + { + "epoch": 0.68, + "learning_rate": 4.961359056356607e-05, + "loss": 0.498, + "step": 22775 + }, + { + "epoch": 0.68, + "learning_rate": 4.960526185103263e-05, + "loss": 0.6022, + "step": 22776 + }, + { + "epoch": 0.68, + "learning_rate": 4.9596933607049346e-05, + "loss": 0.4665, + "step": 22777 + }, + { + "epoch": 0.68, + "learning_rate": 4.958860583169371e-05, + "loss": 0.5014, + "step": 22778 + }, + { + "epoch": 0.68, + "learning_rate": 4.95802785250431e-05, + "loss": 0.4946, + "step": 22779 + }, + { + "epoch": 0.68, + "learning_rate": 4.957195168717501e-05, + "loss": 0.5727, + "step": 22780 + }, + { + "epoch": 0.68, + "learning_rate": 4.956362531816676e-05, + "loss": 0.5999, + "step": 22781 + }, + { + "epoch": 0.68, + "learning_rate": 4.955529941809587e-05, + "loss": 0.3935, + "step": 22782 + }, + { + "epoch": 0.68, + "learning_rate": 4.954697398703966e-05, + "loss": 0.4618, + "step": 22783 + }, + { + "epoch": 0.68, + "learning_rate": 4.953864902507561e-05, + "loss": 0.594, + "step": 22784 + }, + { + "epoch": 0.68, + "learning_rate": 4.9530324532281055e-05, + "loss": 0.5141, + "step": 22785 + }, + { + "epoch": 0.68, + "learning_rate": 4.952200050873342e-05, + "loss": 0.5309, + "step": 22786 + }, + { + "epoch": 0.68, + "learning_rate": 4.951367695451016e-05, + "loss": 0.5639, + "step": 22787 + }, + { + "epoch": 0.68, + "learning_rate": 4.95053538696886e-05, + "loss": 0.4737, + "step": 22788 + }, + { + "epoch": 0.68, + "learning_rate": 4.949703125434609e-05, + "loss": 0.4913, + "step": 22789 + }, + { + "epoch": 0.68, + "learning_rate": 4.948870910856006e-05, + "loss": 0.5214, + "step": 22790 + }, + { + "epoch": 0.68, + "learning_rate": 4.948038743240788e-05, + "loss": 0.5936, + "step": 22791 + }, + { + "epoch": 0.68, + "learning_rate": 4.947206622596701e-05, + "loss": 0.4444, + "step": 22792 + }, + { + "epoch": 0.68, + "learning_rate": 4.946374548931464e-05, + "loss": 0.5726, + "step": 22793 + }, + { + "epoch": 0.68, + "learning_rate": 4.945542522252823e-05, + "loss": 0.4872, + "step": 22794 + }, + { + "epoch": 0.68, + "learning_rate": 4.9447105425685125e-05, + "loss": 0.5177, + "step": 22795 + }, + { + "epoch": 0.68, + "learning_rate": 4.943878609886273e-05, + "loss": 0.5044, + "step": 22796 + }, + { + "epoch": 0.68, + "learning_rate": 4.9430467242138344e-05, + "loss": 0.4977, + "step": 22797 + }, + { + "epoch": 0.68, + "learning_rate": 4.942214885558929e-05, + "loss": 0.5565, + "step": 22798 + }, + { + "epoch": 0.68, + "learning_rate": 4.941383093929293e-05, + "loss": 0.4965, + "step": 22799 + }, + { + "epoch": 0.68, + "learning_rate": 4.940551349332665e-05, + "loss": 0.5202, + "step": 22800 + }, + { + "epoch": 0.68, + "learning_rate": 4.93971965177677e-05, + "loss": 0.4476, + "step": 22801 + }, + { + "epoch": 0.68, + "learning_rate": 4.938888001269348e-05, + "loss": 0.5931, + "step": 22802 + }, + { + "epoch": 0.68, + "learning_rate": 4.938056397818125e-05, + "loss": 0.4525, + "step": 22803 + }, + { + "epoch": 0.68, + "learning_rate": 4.937224841430841e-05, + "loss": 0.612, + "step": 22804 + }, + { + "epoch": 0.68, + "learning_rate": 4.936393332115217e-05, + "loss": 0.5374, + "step": 22805 + }, + { + "epoch": 0.68, + "learning_rate": 4.935561869878994e-05, + "loss": 0.5531, + "step": 22806 + }, + { + "epoch": 0.68, + "learning_rate": 4.934730454729899e-05, + "loss": 0.5714, + "step": 22807 + }, + { + "epoch": 0.68, + "learning_rate": 4.933899086675657e-05, + "loss": 0.4719, + "step": 22808 + }, + { + "epoch": 0.68, + "learning_rate": 4.933067765724002e-05, + "loss": 0.6118, + "step": 22809 + }, + { + "epoch": 0.68, + "learning_rate": 4.9322364918826635e-05, + "loss": 0.5728, + "step": 22810 + }, + { + "epoch": 0.68, + "learning_rate": 4.931405265159375e-05, + "loss": 0.7112, + "step": 22811 + }, + { + "epoch": 0.68, + "learning_rate": 4.930574085561859e-05, + "loss": 0.6098, + "step": 22812 + }, + { + "epoch": 0.68, + "learning_rate": 4.929742953097841e-05, + "loss": 0.5794, + "step": 22813 + }, + { + "epoch": 0.68, + "learning_rate": 4.928911867775054e-05, + "loss": 0.51, + "step": 22814 + }, + { + "epoch": 0.68, + "learning_rate": 4.928080829601225e-05, + "loss": 0.6025, + "step": 22815 + }, + { + "epoch": 0.68, + "learning_rate": 4.9272498385840795e-05, + "loss": 0.4738, + "step": 22816 + }, + { + "epoch": 0.68, + "learning_rate": 4.92641889473134e-05, + "loss": 0.4321, + "step": 22817 + }, + { + "epoch": 0.68, + "learning_rate": 4.925587998050735e-05, + "loss": 0.5185, + "step": 22818 + }, + { + "epoch": 0.68, + "learning_rate": 4.924757148549993e-05, + "loss": 0.408, + "step": 22819 + }, + { + "epoch": 0.68, + "learning_rate": 4.9239263462368344e-05, + "loss": 0.5225, + "step": 22820 + }, + { + "epoch": 0.68, + "learning_rate": 4.923095591118988e-05, + "loss": 0.6059, + "step": 22821 + }, + { + "epoch": 0.68, + "learning_rate": 4.922264883204173e-05, + "loss": 0.5317, + "step": 22822 + }, + { + "epoch": 0.68, + "learning_rate": 4.921434222500117e-05, + "loss": 0.5145, + "step": 22823 + }, + { + "epoch": 0.68, + "learning_rate": 4.920603609014539e-05, + "loss": 0.5168, + "step": 22824 + }, + { + "epoch": 0.68, + "learning_rate": 4.919773042755168e-05, + "loss": 0.446, + "step": 22825 + }, + { + "epoch": 0.68, + "learning_rate": 4.918942523729719e-05, + "loss": 0.5013, + "step": 22826 + }, + { + "epoch": 0.68, + "learning_rate": 4.91811205194592e-05, + "loss": 0.5251, + "step": 22827 + }, + { + "epoch": 0.68, + "learning_rate": 4.917281627411487e-05, + "loss": 0.4759, + "step": 22828 + }, + { + "epoch": 0.68, + "learning_rate": 4.916451250134143e-05, + "loss": 0.5809, + "step": 22829 + }, + { + "epoch": 0.68, + "learning_rate": 4.915620920121613e-05, + "loss": 0.6061, + "step": 22830 + }, + { + "epoch": 0.68, + "learning_rate": 4.9147906373816145e-05, + "loss": 0.6149, + "step": 22831 + }, + { + "epoch": 0.68, + "learning_rate": 4.913960401921861e-05, + "loss": 0.5768, + "step": 22832 + }, + { + "epoch": 0.68, + "learning_rate": 4.9131302137500756e-05, + "loss": 0.5305, + "step": 22833 + }, + { + "epoch": 0.68, + "learning_rate": 4.912300072873983e-05, + "loss": 0.5589, + "step": 22834 + }, + { + "epoch": 0.68, + "learning_rate": 4.9114699793012963e-05, + "loss": 0.4768, + "step": 22835 + }, + { + "epoch": 0.68, + "learning_rate": 4.91063993303973e-05, + "loss": 0.4814, + "step": 22836 + }, + { + "epoch": 0.68, + "learning_rate": 4.909809934097006e-05, + "loss": 0.4116, + "step": 22837 + }, + { + "epoch": 0.68, + "learning_rate": 4.908979982480839e-05, + "loss": 0.5199, + "step": 22838 + }, + { + "epoch": 0.68, + "learning_rate": 4.908150078198952e-05, + "loss": 0.475, + "step": 22839 + }, + { + "epoch": 0.68, + "learning_rate": 4.9073202212590554e-05, + "loss": 0.5254, + "step": 22840 + }, + { + "epoch": 0.68, + "learning_rate": 4.9064904116688624e-05, + "loss": 0.4485, + "step": 22841 + }, + { + "epoch": 0.68, + "learning_rate": 4.9056606494360944e-05, + "loss": 0.4401, + "step": 22842 + }, + { + "epoch": 0.68, + "learning_rate": 4.904830934568461e-05, + "loss": 0.4915, + "step": 22843 + }, + { + "epoch": 0.68, + "learning_rate": 4.904001267073681e-05, + "loss": 0.5712, + "step": 22844 + }, + { + "epoch": 0.68, + "learning_rate": 4.903171646959462e-05, + "loss": 0.3998, + "step": 22845 + }, + { + "epoch": 0.68, + "learning_rate": 4.902342074233528e-05, + "loss": 0.5102, + "step": 22846 + }, + { + "epoch": 0.68, + "learning_rate": 4.901512548903579e-05, + "loss": 0.5405, + "step": 22847 + }, + { + "epoch": 0.68, + "learning_rate": 4.900683070977337e-05, + "loss": 0.5102, + "step": 22848 + }, + { + "epoch": 0.68, + "learning_rate": 4.8998536404625125e-05, + "loss": 0.5904, + "step": 22849 + }, + { + "epoch": 0.68, + "learning_rate": 4.899024257366818e-05, + "loss": 0.6082, + "step": 22850 + }, + { + "epoch": 0.68, + "learning_rate": 4.8981949216979585e-05, + "loss": 0.6329, + "step": 22851 + }, + { + "epoch": 0.68, + "learning_rate": 4.8973656334636495e-05, + "loss": 0.6435, + "step": 22852 + }, + { + "epoch": 0.68, + "learning_rate": 4.896536392671601e-05, + "loss": 0.5542, + "step": 22853 + }, + { + "epoch": 0.68, + "learning_rate": 4.89570719932953e-05, + "loss": 0.5877, + "step": 22854 + }, + { + "epoch": 0.68, + "learning_rate": 4.894878053445132e-05, + "loss": 0.5661, + "step": 22855 + }, + { + "epoch": 0.68, + "learning_rate": 4.894048955026124e-05, + "loss": 0.5108, + "step": 22856 + }, + { + "epoch": 0.68, + "learning_rate": 4.893219904080213e-05, + "loss": 0.5515, + "step": 22857 + }, + { + "epoch": 0.68, + "learning_rate": 4.892390900615111e-05, + "loss": 0.4936, + "step": 22858 + }, + { + "epoch": 0.68, + "learning_rate": 4.891561944638524e-05, + "loss": 0.5245, + "step": 22859 + }, + { + "epoch": 0.68, + "learning_rate": 4.890733036158154e-05, + "loss": 0.5237, + "step": 22860 + }, + { + "epoch": 0.68, + "learning_rate": 4.889904175181712e-05, + "loss": 0.6388, + "step": 22861 + }, + { + "epoch": 0.68, + "learning_rate": 4.889075361716907e-05, + "loss": 0.4825, + "step": 22862 + }, + { + "epoch": 0.68, + "learning_rate": 4.88824659577144e-05, + "loss": 0.5295, + "step": 22863 + }, + { + "epoch": 0.68, + "learning_rate": 4.887417877353021e-05, + "loss": 0.4465, + "step": 22864 + }, + { + "epoch": 0.68, + "learning_rate": 4.88658920646935e-05, + "loss": 0.564, + "step": 22865 + }, + { + "epoch": 0.68, + "learning_rate": 4.8857605831281385e-05, + "loss": 0.3965, + "step": 22866 + }, + { + "epoch": 0.68, + "learning_rate": 4.884932007337083e-05, + "loss": 0.6329, + "step": 22867 + }, + { + "epoch": 0.68, + "learning_rate": 4.884103479103894e-05, + "loss": 0.5581, + "step": 22868 + }, + { + "epoch": 0.68, + "learning_rate": 4.883274998436269e-05, + "loss": 0.467, + "step": 22869 + }, + { + "epoch": 0.68, + "learning_rate": 4.882446565341917e-05, + "loss": 0.4735, + "step": 22870 + }, + { + "epoch": 0.68, + "learning_rate": 4.881618179828533e-05, + "loss": 0.4858, + "step": 22871 + }, + { + "epoch": 0.68, + "learning_rate": 4.880789841903824e-05, + "loss": 0.5565, + "step": 22872 + }, + { + "epoch": 0.68, + "learning_rate": 4.879961551575494e-05, + "loss": 0.5498, + "step": 22873 + }, + { + "epoch": 0.68, + "learning_rate": 4.8791333088512414e-05, + "loss": 0.5211, + "step": 22874 + }, + { + "epoch": 0.68, + "learning_rate": 4.878305113738762e-05, + "loss": 0.5021, + "step": 22875 + }, + { + "epoch": 0.68, + "learning_rate": 4.877476966245761e-05, + "loss": 0.5835, + "step": 22876 + }, + { + "epoch": 0.68, + "learning_rate": 4.87664886637994e-05, + "loss": 0.4852, + "step": 22877 + }, + { + "epoch": 0.68, + "learning_rate": 4.875820814148997e-05, + "loss": 0.5863, + "step": 22878 + }, + { + "epoch": 0.68, + "learning_rate": 4.874992809560626e-05, + "loss": 0.5521, + "step": 22879 + }, + { + "epoch": 0.68, + "learning_rate": 4.874164852622529e-05, + "loss": 0.5264, + "step": 22880 + }, + { + "epoch": 0.68, + "learning_rate": 4.873336943342408e-05, + "loss": 0.5618, + "step": 22881 + }, + { + "epoch": 0.68, + "learning_rate": 4.872509081727953e-05, + "loss": 0.6384, + "step": 22882 + }, + { + "epoch": 0.68, + "learning_rate": 4.8716812677868684e-05, + "loss": 0.6537, + "step": 22883 + }, + { + "epoch": 0.68, + "learning_rate": 4.870853501526844e-05, + "loss": 0.6148, + "step": 22884 + }, + { + "epoch": 0.68, + "learning_rate": 4.870025782955584e-05, + "loss": 0.6081, + "step": 22885 + }, + { + "epoch": 0.68, + "learning_rate": 4.869198112080775e-05, + "loss": 0.4429, + "step": 22886 + }, + { + "epoch": 0.68, + "learning_rate": 4.868370488910121e-05, + "loss": 0.5572, + "step": 22887 + }, + { + "epoch": 0.68, + "learning_rate": 4.8675429134513085e-05, + "loss": 0.57, + "step": 22888 + }, + { + "epoch": 0.68, + "learning_rate": 4.8667153857120404e-05, + "loss": 0.6253, + "step": 22889 + }, + { + "epoch": 0.68, + "learning_rate": 4.865887905700004e-05, + "loss": 0.5258, + "step": 22890 + }, + { + "epoch": 0.68, + "learning_rate": 4.8650604734228944e-05, + "loss": 0.5538, + "step": 22891 + }, + { + "epoch": 0.68, + "learning_rate": 4.86423308888841e-05, + "loss": 0.6033, + "step": 22892 + }, + { + "epoch": 0.68, + "learning_rate": 4.8634057521042395e-05, + "loss": 0.5003, + "step": 22893 + }, + { + "epoch": 0.68, + "learning_rate": 4.862578463078073e-05, + "loss": 0.4605, + "step": 22894 + }, + { + "epoch": 0.68, + "learning_rate": 4.861751221817602e-05, + "loss": 0.4903, + "step": 22895 + }, + { + "epoch": 0.68, + "learning_rate": 4.860924028330521e-05, + "loss": 0.5312, + "step": 22896 + }, + { + "epoch": 0.68, + "learning_rate": 4.860096882624527e-05, + "loss": 0.6053, + "step": 22897 + }, + { + "epoch": 0.68, + "learning_rate": 4.859269784707298e-05, + "loss": 0.4409, + "step": 22898 + }, + { + "epoch": 0.68, + "learning_rate": 4.858442734586528e-05, + "loss": 0.5974, + "step": 22899 + }, + { + "epoch": 0.68, + "learning_rate": 4.857615732269909e-05, + "loss": 0.6058, + "step": 22900 + }, + { + "epoch": 0.68, + "learning_rate": 4.8567887777651324e-05, + "loss": 0.5134, + "step": 22901 + }, + { + "epoch": 0.68, + "learning_rate": 4.8559618710798835e-05, + "loss": 0.4657, + "step": 22902 + }, + { + "epoch": 0.68, + "learning_rate": 4.8551350122218486e-05, + "loss": 0.5998, + "step": 22903 + }, + { + "epoch": 0.68, + "learning_rate": 4.854308201198717e-05, + "loss": 0.5901, + "step": 22904 + }, + { + "epoch": 0.68, + "learning_rate": 4.85348143801818e-05, + "loss": 0.5376, + "step": 22905 + }, + { + "epoch": 0.68, + "learning_rate": 4.852654722687919e-05, + "loss": 0.5564, + "step": 22906 + }, + { + "epoch": 0.68, + "learning_rate": 4.851828055215626e-05, + "loss": 0.5333, + "step": 22907 + }, + { + "epoch": 0.68, + "learning_rate": 4.851001435608984e-05, + "loss": 0.52, + "step": 22908 + }, + { + "epoch": 0.68, + "learning_rate": 4.850174863875675e-05, + "loss": 0.5226, + "step": 22909 + }, + { + "epoch": 0.68, + "learning_rate": 4.8493483400233875e-05, + "loss": 0.4646, + "step": 22910 + }, + { + "epoch": 0.68, + "learning_rate": 4.848521864059809e-05, + "loss": 0.4938, + "step": 22911 + }, + { + "epoch": 0.68, + "learning_rate": 4.847695435992622e-05, + "loss": 0.5104, + "step": 22912 + }, + { + "epoch": 0.68, + "learning_rate": 4.846869055829505e-05, + "loss": 0.5132, + "step": 22913 + }, + { + "epoch": 0.68, + "learning_rate": 4.846042723578147e-05, + "loss": 0.5702, + "step": 22914 + }, + { + "epoch": 0.68, + "learning_rate": 4.8452164392462286e-05, + "loss": 0.5271, + "step": 22915 + }, + { + "epoch": 0.68, + "learning_rate": 4.844390202841437e-05, + "loss": 0.5149, + "step": 22916 + }, + { + "epoch": 0.68, + "learning_rate": 4.84356401437145e-05, + "loss": 0.5007, + "step": 22917 + }, + { + "epoch": 0.68, + "learning_rate": 4.8427378738439464e-05, + "loss": 0.5668, + "step": 22918 + }, + { + "epoch": 0.68, + "learning_rate": 4.84191178126661e-05, + "loss": 0.4968, + "step": 22919 + }, + { + "epoch": 0.68, + "learning_rate": 4.8410857366471275e-05, + "loss": 0.4121, + "step": 22920 + }, + { + "epoch": 0.68, + "learning_rate": 4.8402597399931724e-05, + "loss": 0.4751, + "step": 22921 + }, + { + "epoch": 0.68, + "learning_rate": 4.8394337913124224e-05, + "loss": 0.562, + "step": 22922 + }, + { + "epoch": 0.68, + "learning_rate": 4.83860789061256e-05, + "loss": 0.5281, + "step": 22923 + }, + { + "epoch": 0.68, + "learning_rate": 4.8377820379012676e-05, + "loss": 0.5423, + "step": 22924 + }, + { + "epoch": 0.68, + "learning_rate": 4.836956233186217e-05, + "loss": 0.5723, + "step": 22925 + }, + { + "epoch": 0.68, + "learning_rate": 4.836130476475094e-05, + "loss": 0.5836, + "step": 22926 + }, + { + "epoch": 0.68, + "learning_rate": 4.835304767775567e-05, + "loss": 0.6179, + "step": 22927 + }, + { + "epoch": 0.68, + "learning_rate": 4.834479107095321e-05, + "loss": 0.5817, + "step": 22928 + }, + { + "epoch": 0.68, + "learning_rate": 4.8336534944420276e-05, + "loss": 0.6457, + "step": 22929 + }, + { + "epoch": 0.68, + "learning_rate": 4.832827929823367e-05, + "loss": 0.4168, + "step": 22930 + }, + { + "epoch": 0.68, + "learning_rate": 4.8320024132470095e-05, + "loss": 0.665, + "step": 22931 + }, + { + "epoch": 0.68, + "learning_rate": 4.831176944720638e-05, + "loss": 0.5435, + "step": 22932 + }, + { + "epoch": 0.68, + "learning_rate": 4.830351524251919e-05, + "loss": 0.578, + "step": 22933 + }, + { + "epoch": 0.68, + "learning_rate": 4.8295261518485325e-05, + "loss": 0.5755, + "step": 22934 + }, + { + "epoch": 0.68, + "learning_rate": 4.828700827518153e-05, + "loss": 0.4934, + "step": 22935 + }, + { + "epoch": 0.68, + "learning_rate": 4.8278755512684526e-05, + "loss": 0.5003, + "step": 22936 + }, + { + "epoch": 0.68, + "learning_rate": 4.827050323107101e-05, + "loss": 0.5857, + "step": 22937 + }, + { + "epoch": 0.68, + "learning_rate": 4.8262251430417725e-05, + "loss": 0.5096, + "step": 22938 + }, + { + "epoch": 0.68, + "learning_rate": 4.825400011080145e-05, + "loss": 0.6086, + "step": 22939 + }, + { + "epoch": 0.68, + "learning_rate": 4.8245749272298865e-05, + "loss": 0.5699, + "step": 22940 + }, + { + "epoch": 0.68, + "learning_rate": 4.8237498914986634e-05, + "loss": 0.5225, + "step": 22941 + }, + { + "epoch": 0.68, + "learning_rate": 4.82292490389415e-05, + "loss": 0.5313, + "step": 22942 + }, + { + "epoch": 0.68, + "learning_rate": 4.822099964424022e-05, + "loss": 0.5286, + "step": 22943 + }, + { + "epoch": 0.68, + "learning_rate": 4.821275073095941e-05, + "loss": 0.543, + "step": 22944 + }, + { + "epoch": 0.68, + "learning_rate": 4.820450229917583e-05, + "loss": 0.5904, + "step": 22945 + }, + { + "epoch": 0.68, + "learning_rate": 4.819625434896612e-05, + "loss": 0.6297, + "step": 22946 + }, + { + "epoch": 0.68, + "learning_rate": 4.8188006880407024e-05, + "loss": 0.5107, + "step": 22947 + }, + { + "epoch": 0.68, + "learning_rate": 4.817975989357515e-05, + "loss": 0.5778, + "step": 22948 + }, + { + "epoch": 0.68, + "learning_rate": 4.8171513388547254e-05, + "loss": 0.488, + "step": 22949 + }, + { + "epoch": 0.68, + "learning_rate": 4.8163267365399934e-05, + "loss": 0.5006, + "step": 22950 + }, + { + "epoch": 0.68, + "learning_rate": 4.8155021824209936e-05, + "loss": 0.5823, + "step": 22951 + }, + { + "epoch": 0.68, + "learning_rate": 4.814677676505384e-05, + "loss": 0.6049, + "step": 22952 + }, + { + "epoch": 0.68, + "learning_rate": 4.8138532188008335e-05, + "loss": 0.4056, + "step": 22953 + }, + { + "epoch": 0.68, + "learning_rate": 4.8130288093150134e-05, + "loss": 0.404, + "step": 22954 + }, + { + "epoch": 0.68, + "learning_rate": 4.812204448055585e-05, + "loss": 0.5438, + "step": 22955 + }, + { + "epoch": 0.68, + "learning_rate": 4.8113801350302066e-05, + "loss": 0.6448, + "step": 22956 + }, + { + "epoch": 0.68, + "learning_rate": 4.810555870246549e-05, + "loss": 0.5515, + "step": 22957 + }, + { + "epoch": 0.68, + "learning_rate": 4.8097316537122736e-05, + "loss": 0.4454, + "step": 22958 + }, + { + "epoch": 0.68, + "learning_rate": 4.8089074854350535e-05, + "loss": 0.531, + "step": 22959 + }, + { + "epoch": 0.68, + "learning_rate": 4.8080833654225344e-05, + "loss": 0.5351, + "step": 22960 + }, + { + "epoch": 0.68, + "learning_rate": 4.807259293682387e-05, + "loss": 0.4796, + "step": 22961 + }, + { + "epoch": 0.68, + "learning_rate": 4.8064352702222726e-05, + "loss": 0.533, + "step": 22962 + }, + { + "epoch": 0.68, + "learning_rate": 4.8056112950498566e-05, + "loss": 0.5306, + "step": 22963 + }, + { + "epoch": 0.68, + "learning_rate": 4.804787368172797e-05, + "loss": 0.5026, + "step": 22964 + }, + { + "epoch": 0.68, + "learning_rate": 4.803963489598751e-05, + "loss": 0.6059, + "step": 22965 + }, + { + "epoch": 0.68, + "learning_rate": 4.803139659335381e-05, + "loss": 0.4844, + "step": 22966 + }, + { + "epoch": 0.68, + "learning_rate": 4.80231587739035e-05, + "loss": 0.5681, + "step": 22967 + }, + { + "epoch": 0.68, + "learning_rate": 4.801492143771311e-05, + "loss": 0.4882, + "step": 22968 + }, + { + "epoch": 0.68, + "learning_rate": 4.800668458485931e-05, + "loss": 0.5148, + "step": 22969 + }, + { + "epoch": 0.68, + "learning_rate": 4.799844821541862e-05, + "loss": 0.5489, + "step": 22970 + }, + { + "epoch": 0.68, + "learning_rate": 4.79902123294676e-05, + "loss": 0.6306, + "step": 22971 + }, + { + "epoch": 0.68, + "learning_rate": 4.798197692708286e-05, + "loss": 0.51, + "step": 22972 + }, + { + "epoch": 0.68, + "learning_rate": 4.7973742008341e-05, + "loss": 0.491, + "step": 22973 + }, + { + "epoch": 0.68, + "learning_rate": 4.7965507573318546e-05, + "loss": 0.5912, + "step": 22974 + }, + { + "epoch": 0.68, + "learning_rate": 4.7957273622092034e-05, + "loss": 0.4383, + "step": 22975 + }, + { + "epoch": 0.68, + "learning_rate": 4.794904015473805e-05, + "loss": 0.6167, + "step": 22976 + }, + { + "epoch": 0.68, + "learning_rate": 4.794080717133315e-05, + "loss": 0.5117, + "step": 22977 + }, + { + "epoch": 0.68, + "learning_rate": 4.79325746719539e-05, + "loss": 0.4728, + "step": 22978 + }, + { + "epoch": 0.68, + "learning_rate": 4.792434265667683e-05, + "loss": 0.5853, + "step": 22979 + }, + { + "epoch": 0.68, + "learning_rate": 4.7916111125578425e-05, + "loss": 0.6046, + "step": 22980 + }, + { + "epoch": 0.68, + "learning_rate": 4.790788007873526e-05, + "loss": 0.4436, + "step": 22981 + }, + { + "epoch": 0.68, + "learning_rate": 4.7899649516223896e-05, + "loss": 0.5558, + "step": 22982 + }, + { + "epoch": 0.68, + "learning_rate": 4.789141943812083e-05, + "loss": 0.5404, + "step": 22983 + }, + { + "epoch": 0.68, + "learning_rate": 4.7883189844502554e-05, + "loss": 0.425, + "step": 22984 + }, + { + "epoch": 0.68, + "learning_rate": 4.787496073544558e-05, + "loss": 0.4413, + "step": 22985 + }, + { + "epoch": 0.68, + "learning_rate": 4.7866732111026505e-05, + "loss": 0.5646, + "step": 22986 + }, + { + "epoch": 0.68, + "learning_rate": 4.7858503971321735e-05, + "loss": 0.4981, + "step": 22987 + }, + { + "epoch": 0.68, + "learning_rate": 4.785027631640785e-05, + "loss": 0.5311, + "step": 22988 + }, + { + "epoch": 0.68, + "learning_rate": 4.784204914636127e-05, + "loss": 0.5414, + "step": 22989 + }, + { + "epoch": 0.68, + "learning_rate": 4.783382246125857e-05, + "loss": 0.5736, + "step": 22990 + }, + { + "epoch": 0.68, + "learning_rate": 4.782559626117615e-05, + "loss": 0.5149, + "step": 22991 + }, + { + "epoch": 0.68, + "learning_rate": 4.781737054619059e-05, + "loss": 0.482, + "step": 22992 + }, + { + "epoch": 0.68, + "learning_rate": 4.780914531637828e-05, + "loss": 0.5013, + "step": 22993 + }, + { + "epoch": 0.68, + "learning_rate": 4.7800920571815774e-05, + "loss": 0.5171, + "step": 22994 + }, + { + "epoch": 0.68, + "learning_rate": 4.779269631257947e-05, + "loss": 0.6026, + "step": 22995 + }, + { + "epoch": 0.68, + "learning_rate": 4.778447253874586e-05, + "loss": 0.6167, + "step": 22996 + }, + { + "epoch": 0.68, + "learning_rate": 4.7776249250391456e-05, + "loss": 0.5069, + "step": 22997 + }, + { + "epoch": 0.68, + "learning_rate": 4.776802644759267e-05, + "loss": 0.3988, + "step": 22998 + }, + { + "epoch": 0.68, + "learning_rate": 4.7759804130425924e-05, + "loss": 0.5408, + "step": 22999 + }, + { + "epoch": 0.68, + "learning_rate": 4.775158229896769e-05, + "loss": 0.4848, + "step": 23000 + }, + { + "epoch": 0.68, + "learning_rate": 4.7743360953294455e-05, + "loss": 0.6122, + "step": 23001 + }, + { + "epoch": 0.68, + "learning_rate": 4.7735140093482635e-05, + "loss": 0.491, + "step": 23002 + }, + { + "epoch": 0.68, + "learning_rate": 4.772691971960861e-05, + "loss": 0.559, + "step": 23003 + }, + { + "epoch": 0.68, + "learning_rate": 4.7718699831748846e-05, + "loss": 0.542, + "step": 23004 + }, + { + "epoch": 0.68, + "learning_rate": 4.771048042997981e-05, + "loss": 0.6136, + "step": 23005 + }, + { + "epoch": 0.68, + "learning_rate": 4.770226151437785e-05, + "loss": 0.6686, + "step": 23006 + }, + { + "epoch": 0.68, + "learning_rate": 4.769404308501947e-05, + "loss": 0.6019, + "step": 23007 + }, + { + "epoch": 0.68, + "learning_rate": 4.7685825141980976e-05, + "loss": 0.4973, + "step": 23008 + }, + { + "epoch": 0.69, + "learning_rate": 4.7677607685338876e-05, + "loss": 0.5013, + "step": 23009 + }, + { + "epoch": 0.69, + "learning_rate": 4.766939071516948e-05, + "loss": 0.5851, + "step": 23010 + }, + { + "epoch": 0.69, + "learning_rate": 4.7661174231549276e-05, + "loss": 0.5078, + "step": 23011 + }, + { + "epoch": 0.69, + "learning_rate": 4.765295823455458e-05, + "loss": 0.6558, + "step": 23012 + }, + { + "epoch": 0.69, + "learning_rate": 4.764474272426185e-05, + "loss": 0.5724, + "step": 23013 + }, + { + "epoch": 0.69, + "learning_rate": 4.7636527700747394e-05, + "loss": 0.5176, + "step": 23014 + }, + { + "epoch": 0.69, + "learning_rate": 4.762831316408764e-05, + "loss": 0.6052, + "step": 23015 + }, + { + "epoch": 0.69, + "learning_rate": 4.762009911435899e-05, + "loss": 0.4287, + "step": 23016 + }, + { + "epoch": 0.69, + "learning_rate": 4.761188555163778e-05, + "loss": 0.4888, + "step": 23017 + }, + { + "epoch": 0.69, + "learning_rate": 4.760367247600034e-05, + "loss": 0.5894, + "step": 23018 + }, + { + "epoch": 0.69, + "learning_rate": 4.759545988752308e-05, + "loss": 0.5926, + "step": 23019 + }, + { + "epoch": 0.69, + "learning_rate": 4.758724778628233e-05, + "loss": 0.4671, + "step": 23020 + }, + { + "epoch": 0.69, + "learning_rate": 4.7579036172354554e-05, + "loss": 0.5687, + "step": 23021 + }, + { + "epoch": 0.69, + "learning_rate": 4.7570825045815914e-05, + "loss": 0.5725, + "step": 23022 + }, + { + "epoch": 0.69, + "learning_rate": 4.756261440674286e-05, + "loss": 0.491, + "step": 23023 + }, + { + "epoch": 0.69, + "learning_rate": 4.7554404255211716e-05, + "loss": 0.5751, + "step": 23024 + }, + { + "epoch": 0.69, + "learning_rate": 4.754619459129885e-05, + "loss": 0.4668, + "step": 23025 + }, + { + "epoch": 0.69, + "learning_rate": 4.7537985415080564e-05, + "loss": 0.6143, + "step": 23026 + }, + { + "epoch": 0.69, + "learning_rate": 4.7529776726633144e-05, + "loss": 0.4648, + "step": 23027 + }, + { + "epoch": 0.69, + "learning_rate": 4.752156852603295e-05, + "loss": 0.5256, + "step": 23028 + }, + { + "epoch": 0.69, + "learning_rate": 4.751336081335632e-05, + "loss": 0.6225, + "step": 23029 + }, + { + "epoch": 0.69, + "learning_rate": 4.750515358867952e-05, + "loss": 0.5259, + "step": 23030 + }, + { + "epoch": 0.69, + "learning_rate": 4.7496946852078914e-05, + "loss": 0.4742, + "step": 23031 + }, + { + "epoch": 0.69, + "learning_rate": 4.748874060363073e-05, + "loss": 0.522, + "step": 23032 + }, + { + "epoch": 0.69, + "learning_rate": 4.748053484341134e-05, + "loss": 0.4804, + "step": 23033 + }, + { + "epoch": 0.69, + "learning_rate": 4.7472329571496974e-05, + "loss": 0.5978, + "step": 23034 + }, + { + "epoch": 0.69, + "learning_rate": 4.746412478796399e-05, + "loss": 0.5306, + "step": 23035 + }, + { + "epoch": 0.69, + "learning_rate": 4.745592049288864e-05, + "loss": 0.5356, + "step": 23036 + }, + { + "epoch": 0.69, + "learning_rate": 4.744771668634715e-05, + "loss": 0.5685, + "step": 23037 + }, + { + "epoch": 0.69, + "learning_rate": 4.743951336841584e-05, + "loss": 0.5613, + "step": 23038 + }, + { + "epoch": 0.69, + "learning_rate": 4.7431310539171005e-05, + "loss": 0.507, + "step": 23039 + }, + { + "epoch": 0.69, + "learning_rate": 4.7423108198688924e-05, + "loss": 0.605, + "step": 23040 + }, + { + "epoch": 0.69, + "learning_rate": 4.741490634704582e-05, + "loss": 0.4364, + "step": 23041 + }, + { + "epoch": 0.69, + "learning_rate": 4.740670498431793e-05, + "loss": 0.549, + "step": 23042 + }, + { + "epoch": 0.69, + "learning_rate": 4.7398504110581535e-05, + "loss": 0.5344, + "step": 23043 + }, + { + "epoch": 0.69, + "learning_rate": 4.739030372591291e-05, + "loss": 0.5399, + "step": 23044 + }, + { + "epoch": 0.69, + "learning_rate": 4.738210383038828e-05, + "loss": 0.6332, + "step": 23045 + }, + { + "epoch": 0.69, + "learning_rate": 4.7373904424083845e-05, + "loss": 0.5397, + "step": 23046 + }, + { + "epoch": 0.69, + "learning_rate": 4.736570550707586e-05, + "loss": 0.5832, + "step": 23047 + }, + { + "epoch": 0.69, + "learning_rate": 4.735750707944062e-05, + "loss": 0.4638, + "step": 23048 + }, + { + "epoch": 0.69, + "learning_rate": 4.734930914125425e-05, + "loss": 0.4478, + "step": 23049 + }, + { + "epoch": 0.69, + "learning_rate": 4.734111169259307e-05, + "loss": 0.479, + "step": 23050 + }, + { + "epoch": 0.69, + "learning_rate": 4.733291473353318e-05, + "loss": 0.4546, + "step": 23051 + }, + { + "epoch": 0.69, + "learning_rate": 4.7324718264150925e-05, + "loss": 0.5003, + "step": 23052 + }, + { + "epoch": 0.69, + "learning_rate": 4.731652228452239e-05, + "loss": 0.494, + "step": 23053 + }, + { + "epoch": 0.69, + "learning_rate": 4.730832679472388e-05, + "loss": 0.4295, + "step": 23054 + }, + { + "epoch": 0.69, + "learning_rate": 4.73001317948315e-05, + "loss": 0.4809, + "step": 23055 + }, + { + "epoch": 0.69, + "learning_rate": 4.7291937284921536e-05, + "loss": 0.4616, + "step": 23056 + }, + { + "epoch": 0.69, + "learning_rate": 4.728374326507009e-05, + "loss": 0.5101, + "step": 23057 + }, + { + "epoch": 0.69, + "learning_rate": 4.727554973535339e-05, + "loss": 0.4958, + "step": 23058 + }, + { + "epoch": 0.69, + "learning_rate": 4.7267356695847656e-05, + "loss": 0.5837, + "step": 23059 + }, + { + "epoch": 0.69, + "learning_rate": 4.725916414662902e-05, + "loss": 0.5752, + "step": 23060 + }, + { + "epoch": 0.69, + "learning_rate": 4.7250972087773615e-05, + "loss": 0.6757, + "step": 23061 + }, + { + "epoch": 0.69, + "learning_rate": 4.724278051935764e-05, + "loss": 0.5537, + "step": 23062 + }, + { + "epoch": 0.69, + "learning_rate": 4.7234589441457275e-05, + "loss": 0.5205, + "step": 23063 + }, + { + "epoch": 0.69, + "learning_rate": 4.722639885414874e-05, + "loss": 0.5272, + "step": 23064 + }, + { + "epoch": 0.69, + "learning_rate": 4.721820875750805e-05, + "loss": 0.4972, + "step": 23065 + }, + { + "epoch": 0.69, + "learning_rate": 4.7210019151611405e-05, + "loss": 0.4497, + "step": 23066 + }, + { + "epoch": 0.69, + "learning_rate": 4.7201830036534965e-05, + "loss": 0.4836, + "step": 23067 + }, + { + "epoch": 0.69, + "learning_rate": 4.7193641412354904e-05, + "loss": 0.6032, + "step": 23068 + }, + { + "epoch": 0.69, + "learning_rate": 4.7185453279147327e-05, + "loss": 0.488, + "step": 23069 + }, + { + "epoch": 0.69, + "learning_rate": 4.717726563698831e-05, + "loss": 0.6026, + "step": 23070 + }, + { + "epoch": 0.69, + "learning_rate": 4.716907848595405e-05, + "loss": 0.5134, + "step": 23071 + }, + { + "epoch": 0.69, + "learning_rate": 4.7160891826120625e-05, + "loss": 0.5226, + "step": 23072 + }, + { + "epoch": 0.69, + "learning_rate": 4.715270565756415e-05, + "loss": 0.5572, + "step": 23073 + }, + { + "epoch": 0.69, + "learning_rate": 4.714451998036081e-05, + "loss": 0.5635, + "step": 23074 + }, + { + "epoch": 0.69, + "learning_rate": 4.713633479458666e-05, + "loss": 0.4805, + "step": 23075 + }, + { + "epoch": 0.69, + "learning_rate": 4.712815010031775e-05, + "loss": 0.4929, + "step": 23076 + }, + { + "epoch": 0.69, + "learning_rate": 4.7119965897630224e-05, + "loss": 0.4898, + "step": 23077 + }, + { + "epoch": 0.69, + "learning_rate": 4.7111782186600216e-05, + "loss": 0.5215, + "step": 23078 + }, + { + "epoch": 0.69, + "learning_rate": 4.710359896730379e-05, + "loss": 0.4697, + "step": 23079 + }, + { + "epoch": 0.69, + "learning_rate": 4.709541623981697e-05, + "loss": 0.5187, + "step": 23080 + }, + { + "epoch": 0.69, + "learning_rate": 4.70872340042159e-05, + "loss": 0.6165, + "step": 23081 + }, + { + "epoch": 0.69, + "learning_rate": 4.707905226057662e-05, + "loss": 0.4882, + "step": 23082 + }, + { + "epoch": 0.69, + "learning_rate": 4.707087100897527e-05, + "loss": 0.4052, + "step": 23083 + }, + { + "epoch": 0.69, + "learning_rate": 4.706269024948786e-05, + "loss": 0.6204, + "step": 23084 + }, + { + "epoch": 0.69, + "learning_rate": 4.705450998219042e-05, + "loss": 0.4659, + "step": 23085 + }, + { + "epoch": 0.69, + "learning_rate": 4.704633020715904e-05, + "loss": 0.5242, + "step": 23086 + }, + { + "epoch": 0.69, + "learning_rate": 4.703815092446983e-05, + "loss": 0.5515, + "step": 23087 + }, + { + "epoch": 0.69, + "learning_rate": 4.702997213419876e-05, + "loss": 0.5754, + "step": 23088 + }, + { + "epoch": 0.69, + "learning_rate": 4.702179383642188e-05, + "loss": 0.4623, + "step": 23089 + }, + { + "epoch": 0.69, + "learning_rate": 4.701361603121525e-05, + "loss": 0.4958, + "step": 23090 + }, + { + "epoch": 0.69, + "learning_rate": 4.700543871865491e-05, + "loss": 0.4558, + "step": 23091 + }, + { + "epoch": 0.69, + "learning_rate": 4.699726189881687e-05, + "loss": 0.4982, + "step": 23092 + }, + { + "epoch": 0.69, + "learning_rate": 4.698908557177719e-05, + "loss": 0.5774, + "step": 23093 + }, + { + "epoch": 0.69, + "learning_rate": 4.698090973761182e-05, + "loss": 0.5687, + "step": 23094 + }, + { + "epoch": 0.69, + "learning_rate": 4.697273439639687e-05, + "loss": 0.4331, + "step": 23095 + }, + { + "epoch": 0.69, + "learning_rate": 4.696455954820825e-05, + "loss": 0.5727, + "step": 23096 + }, + { + "epoch": 0.69, + "learning_rate": 4.6956385193122064e-05, + "loss": 0.6517, + "step": 23097 + }, + { + "epoch": 0.69, + "learning_rate": 4.694821133121423e-05, + "loss": 0.4494, + "step": 23098 + }, + { + "epoch": 0.69, + "learning_rate": 4.6940037962560824e-05, + "loss": 0.6235, + "step": 23099 + }, + { + "epoch": 0.69, + "learning_rate": 4.693186508723775e-05, + "loss": 0.5901, + "step": 23100 + }, + { + "epoch": 0.69, + "learning_rate": 4.692369270532105e-05, + "loss": 0.5185, + "step": 23101 + }, + { + "epoch": 0.69, + "learning_rate": 4.6915520816886736e-05, + "loss": 0.6642, + "step": 23102 + }, + { + "epoch": 0.69, + "learning_rate": 4.690734942201075e-05, + "loss": 0.5188, + "step": 23103 + }, + { + "epoch": 0.69, + "learning_rate": 4.6899178520769035e-05, + "loss": 0.5134, + "step": 23104 + }, + { + "epoch": 0.69, + "learning_rate": 4.68910081132376e-05, + "loss": 0.5137, + "step": 23105 + }, + { + "epoch": 0.69, + "learning_rate": 4.688283819949243e-05, + "loss": 0.6352, + "step": 23106 + }, + { + "epoch": 0.69, + "learning_rate": 4.687466877960946e-05, + "loss": 0.6141, + "step": 23107 + }, + { + "epoch": 0.69, + "learning_rate": 4.6866499853664626e-05, + "loss": 0.5867, + "step": 23108 + }, + { + "epoch": 0.69, + "learning_rate": 4.685833142173388e-05, + "loss": 0.6008, + "step": 23109 + }, + { + "epoch": 0.69, + "learning_rate": 4.6850163483893226e-05, + "loss": 0.4615, + "step": 23110 + }, + { + "epoch": 0.69, + "learning_rate": 4.684199604021855e-05, + "loss": 0.6639, + "step": 23111 + }, + { + "epoch": 0.69, + "learning_rate": 4.683382909078583e-05, + "loss": 0.5054, + "step": 23112 + }, + { + "epoch": 0.69, + "learning_rate": 4.682566263567094e-05, + "loss": 0.5833, + "step": 23113 + }, + { + "epoch": 0.69, + "learning_rate": 4.681749667494989e-05, + "loss": 0.5341, + "step": 23114 + }, + { + "epoch": 0.69, + "learning_rate": 4.6809331208698514e-05, + "loss": 0.5303, + "step": 23115 + }, + { + "epoch": 0.69, + "learning_rate": 4.680116623699282e-05, + "loss": 0.4524, + "step": 23116 + }, + { + "epoch": 0.69, + "learning_rate": 4.6793001759908636e-05, + "loss": 0.5401, + "step": 23117 + }, + { + "epoch": 0.69, + "learning_rate": 4.678483777752196e-05, + "loss": 0.4963, + "step": 23118 + }, + { + "epoch": 0.69, + "learning_rate": 4.677667428990862e-05, + "loss": 0.5694, + "step": 23119 + }, + { + "epoch": 0.69, + "learning_rate": 4.676851129714453e-05, + "loss": 0.5063, + "step": 23120 + }, + { + "epoch": 0.69, + "learning_rate": 4.676034879930566e-05, + "loss": 0.5837, + "step": 23121 + }, + { + "epoch": 0.69, + "learning_rate": 4.675218679646784e-05, + "loss": 0.5391, + "step": 23122 + }, + { + "epoch": 0.69, + "learning_rate": 4.674402528870694e-05, + "loss": 0.5206, + "step": 23123 + }, + { + "epoch": 0.69, + "learning_rate": 4.673586427609885e-05, + "loss": 0.4854, + "step": 23124 + }, + { + "epoch": 0.69, + "learning_rate": 4.6727703758719474e-05, + "loss": 0.5045, + "step": 23125 + }, + { + "epoch": 0.69, + "learning_rate": 4.671954373664476e-05, + "loss": 0.5152, + "step": 23126 + }, + { + "epoch": 0.69, + "learning_rate": 4.67113842099504e-05, + "loss": 0.5479, + "step": 23127 + }, + { + "epoch": 0.69, + "learning_rate": 4.670322517871236e-05, + "loss": 0.5643, + "step": 23128 + }, + { + "epoch": 0.69, + "learning_rate": 4.669506664300649e-05, + "loss": 0.5172, + "step": 23129 + }, + { + "epoch": 0.69, + "learning_rate": 4.668690860290868e-05, + "loss": 0.5354, + "step": 23130 + }, + { + "epoch": 0.69, + "learning_rate": 4.6678751058494744e-05, + "loss": 0.5848, + "step": 23131 + }, + { + "epoch": 0.69, + "learning_rate": 4.6670594009840494e-05, + "loss": 0.3576, + "step": 23132 + }, + { + "epoch": 0.69, + "learning_rate": 4.66624374570218e-05, + "loss": 0.589, + "step": 23133 + }, + { + "epoch": 0.69, + "learning_rate": 4.665428140011455e-05, + "loss": 0.4669, + "step": 23134 + }, + { + "epoch": 0.69, + "learning_rate": 4.66461258391945e-05, + "loss": 0.4348, + "step": 23135 + }, + { + "epoch": 0.69, + "learning_rate": 4.6637970774337525e-05, + "loss": 0.4503, + "step": 23136 + }, + { + "epoch": 0.69, + "learning_rate": 4.6629816205619446e-05, + "loss": 0.5062, + "step": 23137 + }, + { + "epoch": 0.69, + "learning_rate": 4.6621662133116016e-05, + "loss": 0.5192, + "step": 23138 + }, + { + "epoch": 0.69, + "learning_rate": 4.661350855690311e-05, + "loss": 0.5886, + "step": 23139 + }, + { + "epoch": 0.69, + "learning_rate": 4.660535547705654e-05, + "loss": 0.5625, + "step": 23140 + }, + { + "epoch": 0.69, + "learning_rate": 4.6597202893652117e-05, + "loss": 0.4675, + "step": 23141 + }, + { + "epoch": 0.69, + "learning_rate": 4.6589050806765564e-05, + "loss": 0.4463, + "step": 23142 + }, + { + "epoch": 0.69, + "learning_rate": 4.658089921647274e-05, + "loss": 0.4994, + "step": 23143 + }, + { + "epoch": 0.69, + "learning_rate": 4.657274812284942e-05, + "loss": 0.5824, + "step": 23144 + }, + { + "epoch": 0.69, + "learning_rate": 4.6564597525971434e-05, + "loss": 0.5426, + "step": 23145 + }, + { + "epoch": 0.69, + "learning_rate": 4.655644742591452e-05, + "loss": 0.5381, + "step": 23146 + }, + { + "epoch": 0.69, + "learning_rate": 4.6548297822754416e-05, + "loss": 0.5364, + "step": 23147 + }, + { + "epoch": 0.69, + "learning_rate": 4.654014871656693e-05, + "loss": 0.5248, + "step": 23148 + }, + { + "epoch": 0.69, + "learning_rate": 4.653200010742789e-05, + "loss": 0.693, + "step": 23149 + }, + { + "epoch": 0.69, + "learning_rate": 4.652385199541299e-05, + "loss": 0.6847, + "step": 23150 + }, + { + "epoch": 0.69, + "learning_rate": 4.651570438059797e-05, + "loss": 0.5607, + "step": 23151 + }, + { + "epoch": 0.69, + "learning_rate": 4.65075572630586e-05, + "loss": 0.5444, + "step": 23152 + }, + { + "epoch": 0.69, + "learning_rate": 4.649941064287069e-05, + "loss": 0.5147, + "step": 23153 + }, + { + "epoch": 0.69, + "learning_rate": 4.64912645201099e-05, + "loss": 0.4717, + "step": 23154 + }, + { + "epoch": 0.69, + "learning_rate": 4.648311889485204e-05, + "loss": 0.5985, + "step": 23155 + }, + { + "epoch": 0.69, + "learning_rate": 4.647497376717278e-05, + "loss": 0.5508, + "step": 23156 + }, + { + "epoch": 0.69, + "learning_rate": 4.6466829137147915e-05, + "loss": 0.5778, + "step": 23157 + }, + { + "epoch": 0.69, + "learning_rate": 4.6458685004853094e-05, + "loss": 0.5425, + "step": 23158 + }, + { + "epoch": 0.69, + "learning_rate": 4.645054137036413e-05, + "loss": 0.4201, + "step": 23159 + }, + { + "epoch": 0.69, + "learning_rate": 4.644239823375665e-05, + "loss": 0.4919, + "step": 23160 + }, + { + "epoch": 0.69, + "learning_rate": 4.643425559510642e-05, + "loss": 0.5778, + "step": 23161 + }, + { + "epoch": 0.69, + "learning_rate": 4.6426113454489116e-05, + "loss": 0.4955, + "step": 23162 + }, + { + "epoch": 0.69, + "learning_rate": 4.6417971811980456e-05, + "loss": 0.5548, + "step": 23163 + }, + { + "epoch": 0.69, + "learning_rate": 4.6409830667656176e-05, + "loss": 0.6023, + "step": 23164 + }, + { + "epoch": 0.69, + "learning_rate": 4.6401690021591926e-05, + "loss": 0.4851, + "step": 23165 + }, + { + "epoch": 0.69, + "learning_rate": 4.639354987386336e-05, + "loss": 0.4991, + "step": 23166 + }, + { + "epoch": 0.69, + "learning_rate": 4.638541022454621e-05, + "loss": 0.5095, + "step": 23167 + }, + { + "epoch": 0.69, + "learning_rate": 4.6377271073716174e-05, + "loss": 0.4187, + "step": 23168 + }, + { + "epoch": 0.69, + "learning_rate": 4.636913242144891e-05, + "loss": 0.4323, + "step": 23169 + }, + { + "epoch": 0.69, + "learning_rate": 4.636099426782004e-05, + "loss": 0.5438, + "step": 23170 + }, + { + "epoch": 0.69, + "learning_rate": 4.635285661290526e-05, + "loss": 0.5372, + "step": 23171 + }, + { + "epoch": 0.69, + "learning_rate": 4.6344719456780286e-05, + "loss": 0.4636, + "step": 23172 + }, + { + "epoch": 0.69, + "learning_rate": 4.633658279952068e-05, + "loss": 0.4701, + "step": 23173 + }, + { + "epoch": 0.69, + "learning_rate": 4.6328446641202173e-05, + "loss": 0.4451, + "step": 23174 + }, + { + "epoch": 0.69, + "learning_rate": 4.632031098190035e-05, + "loss": 0.5087, + "step": 23175 + }, + { + "epoch": 0.69, + "learning_rate": 4.6312175821690906e-05, + "loss": 0.461, + "step": 23176 + }, + { + "epoch": 0.69, + "learning_rate": 4.630404116064942e-05, + "loss": 0.5958, + "step": 23177 + }, + { + "epoch": 0.69, + "learning_rate": 4.6295906998851604e-05, + "loss": 0.5501, + "step": 23178 + }, + { + "epoch": 0.69, + "learning_rate": 4.6287773336373e-05, + "loss": 0.5089, + "step": 23179 + }, + { + "epoch": 0.69, + "learning_rate": 4.6279640173289306e-05, + "loss": 0.6029, + "step": 23180 + }, + { + "epoch": 0.69, + "learning_rate": 4.6271507509676074e-05, + "loss": 0.4791, + "step": 23181 + }, + { + "epoch": 0.69, + "learning_rate": 4.626337534560894e-05, + "loss": 0.5565, + "step": 23182 + }, + { + "epoch": 0.69, + "learning_rate": 4.6255243681163576e-05, + "loss": 0.5705, + "step": 23183 + }, + { + "epoch": 0.69, + "learning_rate": 4.624711251641552e-05, + "loss": 0.5585, + "step": 23184 + }, + { + "epoch": 0.69, + "learning_rate": 4.623898185144037e-05, + "loss": 0.4738, + "step": 23185 + }, + { + "epoch": 0.69, + "learning_rate": 4.623085168631373e-05, + "loss": 0.5947, + "step": 23186 + }, + { + "epoch": 0.69, + "learning_rate": 4.62227220211112e-05, + "loss": 0.5444, + "step": 23187 + }, + { + "epoch": 0.69, + "learning_rate": 4.621459285590843e-05, + "loss": 0.5862, + "step": 23188 + }, + { + "epoch": 0.69, + "learning_rate": 4.6206464190780883e-05, + "loss": 0.563, + "step": 23189 + }, + { + "epoch": 0.69, + "learning_rate": 4.619833602580418e-05, + "loss": 0.5137, + "step": 23190 + }, + { + "epoch": 0.69, + "learning_rate": 4.619020836105391e-05, + "loss": 0.467, + "step": 23191 + }, + { + "epoch": 0.69, + "learning_rate": 4.618208119660567e-05, + "loss": 0.4958, + "step": 23192 + }, + { + "epoch": 0.69, + "learning_rate": 4.6173954532535004e-05, + "loss": 0.5233, + "step": 23193 + }, + { + "epoch": 0.69, + "learning_rate": 4.616582836891741e-05, + "loss": 0.6283, + "step": 23194 + }, + { + "epoch": 0.69, + "learning_rate": 4.6157702705828475e-05, + "loss": 0.45, + "step": 23195 + }, + { + "epoch": 0.69, + "learning_rate": 4.614957754334381e-05, + "loss": 0.4776, + "step": 23196 + }, + { + "epoch": 0.69, + "learning_rate": 4.6141452881538875e-05, + "loss": 0.4455, + "step": 23197 + }, + { + "epoch": 0.69, + "learning_rate": 4.613332872048928e-05, + "loss": 0.5257, + "step": 23198 + }, + { + "epoch": 0.69, + "learning_rate": 4.612520506027049e-05, + "loss": 0.6152, + "step": 23199 + }, + { + "epoch": 0.69, + "learning_rate": 4.61170819009581e-05, + "loss": 0.4618, + "step": 23200 + }, + { + "epoch": 0.69, + "learning_rate": 4.610895924262758e-05, + "loss": 0.5081, + "step": 23201 + }, + { + "epoch": 0.69, + "learning_rate": 4.6100837085354506e-05, + "loss": 0.478, + "step": 23202 + }, + { + "epoch": 0.69, + "learning_rate": 4.609271542921437e-05, + "loss": 0.4942, + "step": 23203 + }, + { + "epoch": 0.69, + "learning_rate": 4.608459427428266e-05, + "loss": 0.5777, + "step": 23204 + }, + { + "epoch": 0.69, + "learning_rate": 4.6076473620634875e-05, + "loss": 0.5695, + "step": 23205 + }, + { + "epoch": 0.69, + "learning_rate": 4.606835346834656e-05, + "loss": 0.4703, + "step": 23206 + }, + { + "epoch": 0.69, + "learning_rate": 4.6060233817493236e-05, + "loss": 0.5062, + "step": 23207 + }, + { + "epoch": 0.69, + "learning_rate": 4.605211466815037e-05, + "loss": 0.5438, + "step": 23208 + }, + { + "epoch": 0.69, + "learning_rate": 4.6043996020393386e-05, + "loss": 0.4951, + "step": 23209 + }, + { + "epoch": 0.69, + "learning_rate": 4.603587787429783e-05, + "loss": 0.5228, + "step": 23210 + }, + { + "epoch": 0.69, + "learning_rate": 4.602776022993921e-05, + "loss": 0.5042, + "step": 23211 + }, + { + "epoch": 0.69, + "learning_rate": 4.601964308739297e-05, + "loss": 0.5622, + "step": 23212 + }, + { + "epoch": 0.69, + "learning_rate": 4.601152644673453e-05, + "loss": 0.6333, + "step": 23213 + }, + { + "epoch": 0.69, + "learning_rate": 4.6003410308039396e-05, + "loss": 0.5588, + "step": 23214 + }, + { + "epoch": 0.69, + "learning_rate": 4.599529467138308e-05, + "loss": 0.5303, + "step": 23215 + }, + { + "epoch": 0.69, + "learning_rate": 4.598717953684094e-05, + "loss": 0.4577, + "step": 23216 + }, + { + "epoch": 0.69, + "learning_rate": 4.5979064904488523e-05, + "loss": 0.5124, + "step": 23217 + }, + { + "epoch": 0.69, + "learning_rate": 4.597095077440119e-05, + "loss": 0.5546, + "step": 23218 + }, + { + "epoch": 0.69, + "learning_rate": 4.596283714665447e-05, + "loss": 0.4679, + "step": 23219 + }, + { + "epoch": 0.69, + "learning_rate": 4.5954724021323703e-05, + "loss": 0.5077, + "step": 23220 + }, + { + "epoch": 0.69, + "learning_rate": 4.5946611398484415e-05, + "loss": 0.4305, + "step": 23221 + }, + { + "epoch": 0.69, + "learning_rate": 4.5938499278211954e-05, + "loss": 0.5086, + "step": 23222 + }, + { + "epoch": 0.69, + "learning_rate": 4.5930387660581806e-05, + "loss": 0.4639, + "step": 23223 + }, + { + "epoch": 0.69, + "learning_rate": 4.592227654566934e-05, + "loss": 0.5991, + "step": 23224 + }, + { + "epoch": 0.69, + "learning_rate": 4.591416593354999e-05, + "loss": 0.5893, + "step": 23225 + }, + { + "epoch": 0.69, + "learning_rate": 4.59060558242992e-05, + "loss": 0.5581, + "step": 23226 + }, + { + "epoch": 0.69, + "learning_rate": 4.589794621799235e-05, + "loss": 0.5215, + "step": 23227 + }, + { + "epoch": 0.69, + "learning_rate": 4.588983711470479e-05, + "loss": 0.5594, + "step": 23228 + }, + { + "epoch": 0.69, + "learning_rate": 4.588172851451197e-05, + "loss": 0.6, + "step": 23229 + }, + { + "epoch": 0.69, + "learning_rate": 4.587362041748926e-05, + "loss": 0.4106, + "step": 23230 + }, + { + "epoch": 0.69, + "learning_rate": 4.586551282371212e-05, + "loss": 0.5903, + "step": 23231 + }, + { + "epoch": 0.69, + "learning_rate": 4.585740573325581e-05, + "loss": 0.6446, + "step": 23232 + }, + { + "epoch": 0.69, + "learning_rate": 4.5849299146195744e-05, + "loss": 0.5106, + "step": 23233 + }, + { + "epoch": 0.69, + "learning_rate": 4.584119306260737e-05, + "loss": 0.5032, + "step": 23234 + }, + { + "epoch": 0.69, + "learning_rate": 4.583308748256594e-05, + "loss": 0.5861, + "step": 23235 + }, + { + "epoch": 0.69, + "learning_rate": 4.5824982406146925e-05, + "loss": 0.4853, + "step": 23236 + }, + { + "epoch": 0.69, + "learning_rate": 4.581687783342559e-05, + "loss": 0.5391, + "step": 23237 + }, + { + "epoch": 0.69, + "learning_rate": 4.580877376447736e-05, + "loss": 0.6014, + "step": 23238 + }, + { + "epoch": 0.69, + "learning_rate": 4.5800670199377524e-05, + "loss": 0.5696, + "step": 23239 + }, + { + "epoch": 0.69, + "learning_rate": 4.579256713820146e-05, + "loss": 0.577, + "step": 23240 + }, + { + "epoch": 0.69, + "learning_rate": 4.578446458102453e-05, + "loss": 0.5879, + "step": 23241 + }, + { + "epoch": 0.69, + "learning_rate": 4.5776362527922046e-05, + "loss": 0.553, + "step": 23242 + }, + { + "epoch": 0.69, + "learning_rate": 4.576826097896929e-05, + "loss": 0.5308, + "step": 23243 + }, + { + "epoch": 0.69, + "learning_rate": 4.576015993424162e-05, + "loss": 0.5483, + "step": 23244 + }, + { + "epoch": 0.69, + "learning_rate": 4.5752059393814416e-05, + "loss": 0.5209, + "step": 23245 + }, + { + "epoch": 0.69, + "learning_rate": 4.5743959357762935e-05, + "loss": 0.6655, + "step": 23246 + }, + { + "epoch": 0.69, + "learning_rate": 4.573585982616247e-05, + "loss": 0.5068, + "step": 23247 + }, + { + "epoch": 0.69, + "learning_rate": 4.572776079908834e-05, + "loss": 0.5026, + "step": 23248 + }, + { + "epoch": 0.69, + "learning_rate": 4.571966227661586e-05, + "loss": 0.5461, + "step": 23249 + }, + { + "epoch": 0.69, + "learning_rate": 4.571156425882037e-05, + "loss": 0.5221, + "step": 23250 + }, + { + "epoch": 0.69, + "learning_rate": 4.570346674577711e-05, + "loss": 0.5939, + "step": 23251 + }, + { + "epoch": 0.69, + "learning_rate": 4.569536973756136e-05, + "loss": 0.5023, + "step": 23252 + }, + { + "epoch": 0.69, + "learning_rate": 4.5687273234248396e-05, + "loss": 0.5057, + "step": 23253 + }, + { + "epoch": 0.69, + "learning_rate": 4.567917723591356e-05, + "loss": 0.5031, + "step": 23254 + }, + { + "epoch": 0.69, + "learning_rate": 4.567108174263208e-05, + "loss": 0.5032, + "step": 23255 + }, + { + "epoch": 0.69, + "learning_rate": 4.56629867544792e-05, + "loss": 0.5834, + "step": 23256 + }, + { + "epoch": 0.69, + "learning_rate": 4.5654892271530215e-05, + "loss": 0.5361, + "step": 23257 + }, + { + "epoch": 0.69, + "learning_rate": 4.5646798293860415e-05, + "loss": 0.5008, + "step": 23258 + }, + { + "epoch": 0.69, + "learning_rate": 4.5638704821544985e-05, + "loss": 0.5396, + "step": 23259 + }, + { + "epoch": 0.69, + "learning_rate": 4.5630611854659246e-05, + "loss": 0.4691, + "step": 23260 + }, + { + "epoch": 0.69, + "learning_rate": 4.562251939327837e-05, + "loss": 0.4572, + "step": 23261 + }, + { + "epoch": 0.69, + "learning_rate": 4.561442743747767e-05, + "loss": 0.5128, + "step": 23262 + }, + { + "epoch": 0.69, + "learning_rate": 4.560633598733231e-05, + "loss": 0.6394, + "step": 23263 + }, + { + "epoch": 0.69, + "learning_rate": 4.559824504291761e-05, + "loss": 0.6275, + "step": 23264 + }, + { + "epoch": 0.69, + "learning_rate": 4.55901546043087e-05, + "loss": 0.4738, + "step": 23265 + }, + { + "epoch": 0.69, + "learning_rate": 4.558206467158088e-05, + "loss": 0.7, + "step": 23266 + }, + { + "epoch": 0.69, + "learning_rate": 4.5573975244809306e-05, + "loss": 0.5847, + "step": 23267 + }, + { + "epoch": 0.69, + "learning_rate": 4.556588632406922e-05, + "loss": 0.4138, + "step": 23268 + }, + { + "epoch": 0.69, + "learning_rate": 4.555779790943585e-05, + "loss": 0.4168, + "step": 23269 + }, + { + "epoch": 0.69, + "learning_rate": 4.554971000098438e-05, + "loss": 0.551, + "step": 23270 + }, + { + "epoch": 0.69, + "learning_rate": 4.554162259878998e-05, + "loss": 0.5289, + "step": 23271 + }, + { + "epoch": 0.69, + "learning_rate": 4.553353570292785e-05, + "loss": 0.4015, + "step": 23272 + }, + { + "epoch": 0.69, + "learning_rate": 4.552544931347324e-05, + "loss": 0.6277, + "step": 23273 + }, + { + "epoch": 0.69, + "learning_rate": 4.55173634305013e-05, + "loss": 0.4964, + "step": 23274 + }, + { + "epoch": 0.69, + "learning_rate": 4.5509278054087155e-05, + "loss": 0.6091, + "step": 23275 + }, + { + "epoch": 0.69, + "learning_rate": 4.550119318430603e-05, + "loss": 0.5523, + "step": 23276 + }, + { + "epoch": 0.69, + "learning_rate": 4.549310882123313e-05, + "loss": 0.6265, + "step": 23277 + }, + { + "epoch": 0.69, + "learning_rate": 4.548502496494353e-05, + "loss": 0.5547, + "step": 23278 + }, + { + "epoch": 0.69, + "learning_rate": 4.5476941615512494e-05, + "loss": 0.4973, + "step": 23279 + }, + { + "epoch": 0.69, + "learning_rate": 4.5468858773015074e-05, + "loss": 0.5483, + "step": 23280 + }, + { + "epoch": 0.69, + "learning_rate": 4.546077643752651e-05, + "loss": 0.4913, + "step": 23281 + }, + { + "epoch": 0.69, + "learning_rate": 4.545269460912187e-05, + "loss": 0.4916, + "step": 23282 + }, + { + "epoch": 0.69, + "learning_rate": 4.544461328787638e-05, + "loss": 0.4551, + "step": 23283 + }, + { + "epoch": 0.69, + "learning_rate": 4.543653247386509e-05, + "loss": 0.6737, + "step": 23284 + }, + { + "epoch": 0.69, + "learning_rate": 4.5428452167163215e-05, + "loss": 0.5117, + "step": 23285 + }, + { + "epoch": 0.69, + "learning_rate": 4.5420372367845795e-05, + "loss": 0.5378, + "step": 23286 + }, + { + "epoch": 0.69, + "learning_rate": 4.541229307598801e-05, + "loss": 0.4451, + "step": 23287 + }, + { + "epoch": 0.69, + "learning_rate": 4.5404214291665e-05, + "loss": 0.3983, + "step": 23288 + }, + { + "epoch": 0.69, + "learning_rate": 4.539613601495185e-05, + "loss": 0.5241, + "step": 23289 + }, + { + "epoch": 0.69, + "learning_rate": 4.538805824592363e-05, + "loss": 0.5458, + "step": 23290 + }, + { + "epoch": 0.69, + "learning_rate": 4.537998098465547e-05, + "loss": 0.489, + "step": 23291 + }, + { + "epoch": 0.69, + "learning_rate": 4.5371904231222474e-05, + "loss": 0.5255, + "step": 23292 + }, + { + "epoch": 0.69, + "learning_rate": 4.5363827985699825e-05, + "loss": 0.5697, + "step": 23293 + }, + { + "epoch": 0.69, + "learning_rate": 4.5355752248162445e-05, + "loss": 0.3852, + "step": 23294 + }, + { + "epoch": 0.69, + "learning_rate": 4.534767701868551e-05, + "loss": 0.5179, + "step": 23295 + }, + { + "epoch": 0.69, + "learning_rate": 4.5339602297344087e-05, + "loss": 0.4584, + "step": 23296 + }, + { + "epoch": 0.69, + "learning_rate": 4.533152808421329e-05, + "loss": 0.5548, + "step": 23297 + }, + { + "epoch": 0.69, + "learning_rate": 4.5323454379368156e-05, + "loss": 0.5148, + "step": 23298 + }, + { + "epoch": 0.69, + "learning_rate": 4.531538118288372e-05, + "loss": 0.517, + "step": 23299 + }, + { + "epoch": 0.69, + "learning_rate": 4.530730849483511e-05, + "loss": 0.6167, + "step": 23300 + }, + { + "epoch": 0.69, + "learning_rate": 4.529923631529731e-05, + "loss": 0.5667, + "step": 23301 + }, + { + "epoch": 0.69, + "learning_rate": 4.529116464434541e-05, + "loss": 0.6664, + "step": 23302 + }, + { + "epoch": 0.69, + "learning_rate": 4.528309348205448e-05, + "loss": 0.4913, + "step": 23303 + }, + { + "epoch": 0.69, + "learning_rate": 4.527502282849956e-05, + "loss": 0.5026, + "step": 23304 + }, + { + "epoch": 0.69, + "learning_rate": 4.526695268375562e-05, + "loss": 0.5206, + "step": 23305 + }, + { + "epoch": 0.69, + "learning_rate": 4.5258883047897736e-05, + "loss": 0.5643, + "step": 23306 + }, + { + "epoch": 0.69, + "learning_rate": 4.525081392100098e-05, + "loss": 0.4647, + "step": 23307 + }, + { + "epoch": 0.69, + "learning_rate": 4.524274530314034e-05, + "loss": 0.6222, + "step": 23308 + }, + { + "epoch": 0.69, + "learning_rate": 4.523467719439078e-05, + "loss": 0.6034, + "step": 23309 + }, + { + "epoch": 0.69, + "learning_rate": 4.5226609594827386e-05, + "loss": 0.5704, + "step": 23310 + }, + { + "epoch": 0.69, + "learning_rate": 4.521854250452513e-05, + "loss": 0.5863, + "step": 23311 + }, + { + "epoch": 0.69, + "learning_rate": 4.521047592355907e-05, + "loss": 0.55, + "step": 23312 + }, + { + "epoch": 0.69, + "learning_rate": 4.520240985200417e-05, + "loss": 0.5688, + "step": 23313 + }, + { + "epoch": 0.69, + "learning_rate": 4.519434428993539e-05, + "loss": 0.5879, + "step": 23314 + }, + { + "epoch": 0.69, + "learning_rate": 4.518627923742774e-05, + "loss": 0.5144, + "step": 23315 + }, + { + "epoch": 0.69, + "learning_rate": 4.517821469455626e-05, + "loss": 0.5162, + "step": 23316 + }, + { + "epoch": 0.69, + "learning_rate": 4.5170150661395905e-05, + "loss": 0.5852, + "step": 23317 + }, + { + "epoch": 0.69, + "learning_rate": 4.516208713802158e-05, + "loss": 0.4836, + "step": 23318 + }, + { + "epoch": 0.69, + "learning_rate": 4.5154024124508335e-05, + "loss": 0.4696, + "step": 23319 + }, + { + "epoch": 0.69, + "learning_rate": 4.5145961620931134e-05, + "loss": 0.6885, + "step": 23320 + }, + { + "epoch": 0.69, + "learning_rate": 4.513789962736488e-05, + "loss": 0.4383, + "step": 23321 + }, + { + "epoch": 0.69, + "learning_rate": 4.512983814388462e-05, + "loss": 0.5453, + "step": 23322 + }, + { + "epoch": 0.69, + "learning_rate": 4.5121777170565204e-05, + "loss": 0.5248, + "step": 23323 + }, + { + "epoch": 0.69, + "learning_rate": 4.511371670748168e-05, + "loss": 0.5043, + "step": 23324 + }, + { + "epoch": 0.69, + "learning_rate": 4.51056567547089e-05, + "loss": 0.5761, + "step": 23325 + }, + { + "epoch": 0.69, + "learning_rate": 4.509759731232188e-05, + "loss": 0.5424, + "step": 23326 + }, + { + "epoch": 0.69, + "learning_rate": 4.508953838039549e-05, + "loss": 0.5723, + "step": 23327 + }, + { + "epoch": 0.69, + "learning_rate": 4.5081479959004714e-05, + "loss": 0.5222, + "step": 23328 + }, + { + "epoch": 0.69, + "learning_rate": 4.507342204822441e-05, + "loss": 0.5209, + "step": 23329 + }, + { + "epoch": 0.69, + "learning_rate": 4.506536464812955e-05, + "loss": 0.4357, + "step": 23330 + }, + { + "epoch": 0.69, + "learning_rate": 4.5057307758795064e-05, + "loss": 0.574, + "step": 23331 + }, + { + "epoch": 0.69, + "learning_rate": 4.504925138029583e-05, + "loss": 0.5435, + "step": 23332 + }, + { + "epoch": 0.69, + "learning_rate": 4.5041195512706723e-05, + "loss": 0.5377, + "step": 23333 + }, + { + "epoch": 0.69, + "learning_rate": 4.5033140156102674e-05, + "loss": 0.5186, + "step": 23334 + }, + { + "epoch": 0.69, + "learning_rate": 4.5025085310558615e-05, + "loss": 0.5911, + "step": 23335 + }, + { + "epoch": 0.69, + "learning_rate": 4.501703097614941e-05, + "loss": 0.5634, + "step": 23336 + }, + { + "epoch": 0.69, + "learning_rate": 4.500897715294989e-05, + "loss": 0.51, + "step": 23337 + }, + { + "epoch": 0.69, + "learning_rate": 4.500092384103498e-05, + "loss": 0.6059, + "step": 23338 + }, + { + "epoch": 0.69, + "learning_rate": 4.499287104047961e-05, + "loss": 0.5459, + "step": 23339 + }, + { + "epoch": 0.69, + "learning_rate": 4.498481875135856e-05, + "loss": 0.5105, + "step": 23340 + }, + { + "epoch": 0.69, + "learning_rate": 4.497676697374678e-05, + "loss": 0.4714, + "step": 23341 + }, + { + "epoch": 0.69, + "learning_rate": 4.4968715707719036e-05, + "loss": 0.3765, + "step": 23342 + }, + { + "epoch": 0.69, + "learning_rate": 4.496066495335029e-05, + "loss": 0.4301, + "step": 23343 + }, + { + "epoch": 0.69, + "learning_rate": 4.49526147107153e-05, + "loss": 0.4927, + "step": 23344 + }, + { + "epoch": 0.7, + "learning_rate": 4.4944564979889006e-05, + "loss": 0.5384, + "step": 23345 + }, + { + "epoch": 0.7, + "learning_rate": 4.493651576094616e-05, + "loss": 0.4497, + "step": 23346 + }, + { + "epoch": 0.7, + "learning_rate": 4.492846705396169e-05, + "loss": 0.659, + "step": 23347 + }, + { + "epoch": 0.7, + "learning_rate": 4.492041885901034e-05, + "loss": 0.6571, + "step": 23348 + }, + { + "epoch": 0.7, + "learning_rate": 4.491237117616698e-05, + "loss": 0.5905, + "step": 23349 + }, + { + "epoch": 0.7, + "learning_rate": 4.490432400550648e-05, + "loss": 0.4936, + "step": 23350 + }, + { + "epoch": 0.7, + "learning_rate": 4.489627734710361e-05, + "loss": 0.495, + "step": 23351 + }, + { + "epoch": 0.7, + "learning_rate": 4.488823120103316e-05, + "loss": 0.5243, + "step": 23352 + }, + { + "epoch": 0.7, + "learning_rate": 4.488018556736997e-05, + "loss": 0.4892, + "step": 23353 + }, + { + "epoch": 0.7, + "learning_rate": 4.487214044618885e-05, + "loss": 0.4116, + "step": 23354 + }, + { + "epoch": 0.7, + "learning_rate": 4.4864095837564667e-05, + "loss": 0.5313, + "step": 23355 + }, + { + "epoch": 0.7, + "learning_rate": 4.485605174157207e-05, + "loss": 0.4822, + "step": 23356 + }, + { + "epoch": 0.7, + "learning_rate": 4.4848008158285924e-05, + "loss": 0.5541, + "step": 23357 + }, + { + "epoch": 0.7, + "learning_rate": 4.483996508778102e-05, + "loss": 0.5973, + "step": 23358 + }, + { + "epoch": 0.7, + "learning_rate": 4.4831922530132165e-05, + "loss": 0.5281, + "step": 23359 + }, + { + "epoch": 0.7, + "learning_rate": 4.482388048541411e-05, + "loss": 0.5528, + "step": 23360 + }, + { + "epoch": 0.7, + "learning_rate": 4.481583895370159e-05, + "loss": 0.5123, + "step": 23361 + }, + { + "epoch": 0.7, + "learning_rate": 4.4807797935069405e-05, + "loss": 0.5261, + "step": 23362 + }, + { + "epoch": 0.7, + "learning_rate": 4.4799757429592356e-05, + "loss": 0.529, + "step": 23363 + }, + { + "epoch": 0.7, + "learning_rate": 4.479171743734513e-05, + "loss": 0.4668, + "step": 23364 + }, + { + "epoch": 0.7, + "learning_rate": 4.478367795840254e-05, + "loss": 0.572, + "step": 23365 + }, + { + "epoch": 0.7, + "learning_rate": 4.477563899283931e-05, + "loss": 0.3796, + "step": 23366 + }, + { + "epoch": 0.7, + "learning_rate": 4.4767600540730134e-05, + "loss": 0.6045, + "step": 23367 + }, + { + "epoch": 0.7, + "learning_rate": 4.475956260214979e-05, + "loss": 0.6666, + "step": 23368 + }, + { + "epoch": 0.7, + "learning_rate": 4.475152517717307e-05, + "loss": 0.548, + "step": 23369 + }, + { + "epoch": 0.7, + "learning_rate": 4.474348826587465e-05, + "loss": 0.3734, + "step": 23370 + }, + { + "epoch": 0.7, + "learning_rate": 4.473545186832919e-05, + "loss": 0.4921, + "step": 23371 + }, + { + "epoch": 0.7, + "learning_rate": 4.47274159846115e-05, + "loss": 0.5891, + "step": 23372 + }, + { + "epoch": 0.7, + "learning_rate": 4.471938061479625e-05, + "loss": 0.358, + "step": 23373 + }, + { + "epoch": 0.7, + "learning_rate": 4.4711345758958214e-05, + "loss": 0.5651, + "step": 23374 + }, + { + "epoch": 0.7, + "learning_rate": 4.470331141717205e-05, + "loss": 0.399, + "step": 23375 + }, + { + "epoch": 0.7, + "learning_rate": 4.469527758951241e-05, + "loss": 0.4278, + "step": 23376 + }, + { + "epoch": 0.7, + "learning_rate": 4.468724427605404e-05, + "loss": 0.5617, + "step": 23377 + }, + { + "epoch": 0.7, + "learning_rate": 4.467921147687166e-05, + "loss": 0.508, + "step": 23378 + }, + { + "epoch": 0.7, + "learning_rate": 4.4671179192039936e-05, + "loss": 0.5477, + "step": 23379 + }, + { + "epoch": 0.7, + "learning_rate": 4.466314742163349e-05, + "loss": 0.4871, + "step": 23380 + }, + { + "epoch": 0.7, + "learning_rate": 4.465511616572704e-05, + "loss": 0.4771, + "step": 23381 + }, + { + "epoch": 0.7, + "learning_rate": 4.464708542439531e-05, + "loss": 0.6039, + "step": 23382 + }, + { + "epoch": 0.7, + "learning_rate": 4.463905519771287e-05, + "loss": 0.532, + "step": 23383 + }, + { + "epoch": 0.7, + "learning_rate": 4.463102548575447e-05, + "loss": 0.5968, + "step": 23384 + }, + { + "epoch": 0.7, + "learning_rate": 4.46229962885947e-05, + "loss": 0.5451, + "step": 23385 + }, + { + "epoch": 0.7, + "learning_rate": 4.4614967606308266e-05, + "loss": 0.5436, + "step": 23386 + }, + { + "epoch": 0.7, + "learning_rate": 4.460693943896975e-05, + "loss": 0.4926, + "step": 23387 + }, + { + "epoch": 0.7, + "learning_rate": 4.4598911786653885e-05, + "loss": 0.4939, + "step": 23388 + }, + { + "epoch": 0.7, + "learning_rate": 4.459088464943521e-05, + "loss": 0.6368, + "step": 23389 + }, + { + "epoch": 0.7, + "learning_rate": 4.458285802738844e-05, + "loss": 0.5204, + "step": 23390 + }, + { + "epoch": 0.7, + "learning_rate": 4.457483192058815e-05, + "loss": 0.5358, + "step": 23391 + }, + { + "epoch": 0.7, + "learning_rate": 4.456680632910896e-05, + "loss": 0.5007, + "step": 23392 + }, + { + "epoch": 0.7, + "learning_rate": 4.455878125302555e-05, + "loss": 0.6604, + "step": 23393 + }, + { + "epoch": 0.7, + "learning_rate": 4.455075669241251e-05, + "loss": 0.5301, + "step": 23394 + }, + { + "epoch": 0.7, + "learning_rate": 4.4542732647344385e-05, + "loss": 0.5898, + "step": 23395 + }, + { + "epoch": 0.7, + "learning_rate": 4.4534709117895836e-05, + "loss": 0.4761, + "step": 23396 + }, + { + "epoch": 0.7, + "learning_rate": 4.4526686104141444e-05, + "loss": 0.605, + "step": 23397 + }, + { + "epoch": 0.7, + "learning_rate": 4.451866360615591e-05, + "loss": 0.553, + "step": 23398 + }, + { + "epoch": 0.7, + "learning_rate": 4.451064162401364e-05, + "loss": 0.6411, + "step": 23399 + }, + { + "epoch": 0.7, + "learning_rate": 4.450262015778931e-05, + "loss": 0.6443, + "step": 23400 + }, + { + "epoch": 0.7, + "learning_rate": 4.449459920755753e-05, + "loss": 0.5821, + "step": 23401 + }, + { + "epoch": 0.7, + "learning_rate": 4.448657877339282e-05, + "loss": 0.4622, + "step": 23402 + }, + { + "epoch": 0.7, + "learning_rate": 4.44785588553698e-05, + "loss": 0.4863, + "step": 23403 + }, + { + "epoch": 0.7, + "learning_rate": 4.447053945356299e-05, + "loss": 0.3964, + "step": 23404 + }, + { + "epoch": 0.7, + "learning_rate": 4.4462520568047e-05, + "loss": 0.5155, + "step": 23405 + }, + { + "epoch": 0.7, + "learning_rate": 4.445450219889632e-05, + "loss": 0.7273, + "step": 23406 + }, + { + "epoch": 0.7, + "learning_rate": 4.4446484346185535e-05, + "loss": 0.6607, + "step": 23407 + }, + { + "epoch": 0.7, + "learning_rate": 4.443846700998925e-05, + "loss": 0.623, + "step": 23408 + }, + { + "epoch": 0.7, + "learning_rate": 4.4430450190381935e-05, + "loss": 0.4579, + "step": 23409 + }, + { + "epoch": 0.7, + "learning_rate": 4.442243388743812e-05, + "loss": 0.5446, + "step": 23410 + }, + { + "epoch": 0.7, + "learning_rate": 4.441441810123237e-05, + "loss": 0.4893, + "step": 23411 + }, + { + "epoch": 0.7, + "learning_rate": 4.440640283183923e-05, + "loss": 0.5546, + "step": 23412 + }, + { + "epoch": 0.7, + "learning_rate": 4.4398388079333206e-05, + "loss": 0.4862, + "step": 23413 + }, + { + "epoch": 0.7, + "learning_rate": 4.439037384378877e-05, + "loss": 0.527, + "step": 23414 + }, + { + "epoch": 0.7, + "learning_rate": 4.438236012528048e-05, + "loss": 0.4799, + "step": 23415 + }, + { + "epoch": 0.7, + "learning_rate": 4.437434692388283e-05, + "loss": 0.4252, + "step": 23416 + }, + { + "epoch": 0.7, + "learning_rate": 4.436633423967037e-05, + "loss": 0.5174, + "step": 23417 + }, + { + "epoch": 0.7, + "learning_rate": 4.435832207271756e-05, + "loss": 0.5439, + "step": 23418 + }, + { + "epoch": 0.7, + "learning_rate": 4.4350310423098875e-05, + "loss": 0.4842, + "step": 23419 + }, + { + "epoch": 0.7, + "learning_rate": 4.43422992908888e-05, + "loss": 0.4905, + "step": 23420 + }, + { + "epoch": 0.7, + "learning_rate": 4.433428867616189e-05, + "loss": 0.5053, + "step": 23421 + }, + { + "epoch": 0.7, + "learning_rate": 4.4326278578992586e-05, + "loss": 0.5011, + "step": 23422 + }, + { + "epoch": 0.7, + "learning_rate": 4.4318268999455304e-05, + "loss": 0.6224, + "step": 23423 + }, + { + "epoch": 0.7, + "learning_rate": 4.4310259937624574e-05, + "loss": 0.5804, + "step": 23424 + }, + { + "epoch": 0.7, + "learning_rate": 4.430225139357488e-05, + "loss": 0.5827, + "step": 23425 + }, + { + "epoch": 0.7, + "learning_rate": 4.429424336738063e-05, + "loss": 0.4548, + "step": 23426 + }, + { + "epoch": 0.7, + "learning_rate": 4.428623585911632e-05, + "loss": 0.4987, + "step": 23427 + }, + { + "epoch": 0.7, + "learning_rate": 4.427822886885637e-05, + "loss": 0.5891, + "step": 23428 + }, + { + "epoch": 0.7, + "learning_rate": 4.4270222396675255e-05, + "loss": 0.535, + "step": 23429 + }, + { + "epoch": 0.7, + "learning_rate": 4.426221644264738e-05, + "loss": 0.5984, + "step": 23430 + }, + { + "epoch": 0.7, + "learning_rate": 4.425421100684722e-05, + "loss": 0.5462, + "step": 23431 + }, + { + "epoch": 0.7, + "learning_rate": 4.424620608934921e-05, + "loss": 0.5382, + "step": 23432 + }, + { + "epoch": 0.7, + "learning_rate": 4.4238201690227696e-05, + "loss": 0.5447, + "step": 23433 + }, + { + "epoch": 0.7, + "learning_rate": 4.4230197809557175e-05, + "loss": 0.5487, + "step": 23434 + }, + { + "epoch": 0.7, + "learning_rate": 4.4222194447412034e-05, + "loss": 0.4897, + "step": 23435 + }, + { + "epoch": 0.7, + "learning_rate": 4.421419160386675e-05, + "loss": 0.4573, + "step": 23436 + }, + { + "epoch": 0.7, + "learning_rate": 4.420618927899567e-05, + "loss": 0.5059, + "step": 23437 + }, + { + "epoch": 0.7, + "learning_rate": 4.419818747287316e-05, + "loss": 0.5771, + "step": 23438 + }, + { + "epoch": 0.7, + "learning_rate": 4.419018618557368e-05, + "loss": 0.3717, + "step": 23439 + }, + { + "epoch": 0.7, + "learning_rate": 4.418218541717163e-05, + "loss": 0.5813, + "step": 23440 + }, + { + "epoch": 0.7, + "learning_rate": 4.417418516774138e-05, + "loss": 0.4871, + "step": 23441 + }, + { + "epoch": 0.7, + "learning_rate": 4.4166185437357266e-05, + "loss": 0.5185, + "step": 23442 + }, + { + "epoch": 0.7, + "learning_rate": 4.41581862260937e-05, + "loss": 0.4813, + "step": 23443 + }, + { + "epoch": 0.7, + "learning_rate": 4.415018753402511e-05, + "loss": 0.4807, + "step": 23444 + }, + { + "epoch": 0.7, + "learning_rate": 4.4142189361225775e-05, + "loss": 0.6133, + "step": 23445 + }, + { + "epoch": 0.7, + "learning_rate": 4.413419170777014e-05, + "loss": 0.4721, + "step": 23446 + }, + { + "epoch": 0.7, + "learning_rate": 4.412619457373248e-05, + "loss": 0.451, + "step": 23447 + }, + { + "epoch": 0.7, + "learning_rate": 4.4118197959187244e-05, + "loss": 0.5053, + "step": 23448 + }, + { + "epoch": 0.7, + "learning_rate": 4.411020186420869e-05, + "loss": 0.6199, + "step": 23449 + }, + { + "epoch": 0.7, + "learning_rate": 4.410220628887123e-05, + "loss": 0.5099, + "step": 23450 + }, + { + "epoch": 0.7, + "learning_rate": 4.409421123324915e-05, + "loss": 0.488, + "step": 23451 + }, + { + "epoch": 0.7, + "learning_rate": 4.408621669741685e-05, + "loss": 0.5017, + "step": 23452 + }, + { + "epoch": 0.7, + "learning_rate": 4.407822268144858e-05, + "loss": 0.4785, + "step": 23453 + }, + { + "epoch": 0.7, + "learning_rate": 4.407022918541871e-05, + "loss": 0.6136, + "step": 23454 + }, + { + "epoch": 0.7, + "learning_rate": 4.4062236209401586e-05, + "loss": 0.6218, + "step": 23455 + }, + { + "epoch": 0.7, + "learning_rate": 4.4054243753471504e-05, + "loss": 0.4734, + "step": 23456 + }, + { + "epoch": 0.7, + "learning_rate": 4.4046251817702735e-05, + "loss": 0.545, + "step": 23457 + }, + { + "epoch": 0.7, + "learning_rate": 4.403826040216961e-05, + "loss": 0.5689, + "step": 23458 + }, + { + "epoch": 0.7, + "learning_rate": 4.403026950694643e-05, + "loss": 0.4942, + "step": 23459 + }, + { + "epoch": 0.7, + "learning_rate": 4.402227913210757e-05, + "loss": 0.5548, + "step": 23460 + }, + { + "epoch": 0.7, + "learning_rate": 4.401428927772718e-05, + "loss": 0.6387, + "step": 23461 + }, + { + "epoch": 0.7, + "learning_rate": 4.4006299943879614e-05, + "loss": 0.5874, + "step": 23462 + }, + { + "epoch": 0.7, + "learning_rate": 4.399831113063919e-05, + "loss": 0.4015, + "step": 23463 + }, + { + "epoch": 0.7, + "learning_rate": 4.399032283808011e-05, + "loss": 0.5579, + "step": 23464 + }, + { + "epoch": 0.7, + "learning_rate": 4.3982335066276726e-05, + "loss": 0.537, + "step": 23465 + }, + { + "epoch": 0.7, + "learning_rate": 4.397434781530323e-05, + "loss": 0.5371, + "step": 23466 + }, + { + "epoch": 0.7, + "learning_rate": 4.3966361085233944e-05, + "loss": 0.4922, + "step": 23467 + }, + { + "epoch": 0.7, + "learning_rate": 4.395837487614307e-05, + "loss": 0.5159, + "step": 23468 + }, + { + "epoch": 0.7, + "learning_rate": 4.395038918810489e-05, + "loss": 0.5139, + "step": 23469 + }, + { + "epoch": 0.7, + "learning_rate": 4.394240402119369e-05, + "loss": 0.5602, + "step": 23470 + }, + { + "epoch": 0.7, + "learning_rate": 4.393441937548367e-05, + "loss": 0.5503, + "step": 23471 + }, + { + "epoch": 0.7, + "learning_rate": 4.392643525104905e-05, + "loss": 0.4659, + "step": 23472 + }, + { + "epoch": 0.7, + "learning_rate": 4.391845164796408e-05, + "loss": 0.5892, + "step": 23473 + }, + { + "epoch": 0.7, + "learning_rate": 4.391046856630302e-05, + "loss": 0.5799, + "step": 23474 + }, + { + "epoch": 0.7, + "learning_rate": 4.390248600614008e-05, + "loss": 0.5448, + "step": 23475 + }, + { + "epoch": 0.7, + "learning_rate": 4.3894503967549425e-05, + "loss": 0.5409, + "step": 23476 + }, + { + "epoch": 0.7, + "learning_rate": 4.388652245060531e-05, + "loss": 0.5656, + "step": 23477 + }, + { + "epoch": 0.7, + "learning_rate": 4.387854145538195e-05, + "loss": 0.5522, + "step": 23478 + }, + { + "epoch": 0.7, + "learning_rate": 4.3870560981953577e-05, + "loss": 0.4993, + "step": 23479 + }, + { + "epoch": 0.7, + "learning_rate": 4.386258103039436e-05, + "loss": 0.4367, + "step": 23480 + }, + { + "epoch": 0.7, + "learning_rate": 4.385460160077844e-05, + "loss": 0.5538, + "step": 23481 + }, + { + "epoch": 0.7, + "learning_rate": 4.384662269318007e-05, + "loss": 0.5237, + "step": 23482 + }, + { + "epoch": 0.7, + "learning_rate": 4.383864430767346e-05, + "loss": 0.5344, + "step": 23483 + }, + { + "epoch": 0.7, + "learning_rate": 4.383066644433275e-05, + "loss": 0.5225, + "step": 23484 + }, + { + "epoch": 0.7, + "learning_rate": 4.382268910323207e-05, + "loss": 0.5469, + "step": 23485 + }, + { + "epoch": 0.7, + "learning_rate": 4.381471228444565e-05, + "loss": 0.5233, + "step": 23486 + }, + { + "epoch": 0.7, + "learning_rate": 4.3806735988047674e-05, + "loss": 0.6343, + "step": 23487 + }, + { + "epoch": 0.7, + "learning_rate": 4.379876021411223e-05, + "loss": 0.7421, + "step": 23488 + }, + { + "epoch": 0.7, + "learning_rate": 4.379078496271356e-05, + "loss": 0.5176, + "step": 23489 + }, + { + "epoch": 0.7, + "learning_rate": 4.378281023392572e-05, + "loss": 0.5535, + "step": 23490 + }, + { + "epoch": 0.7, + "learning_rate": 4.377483602782295e-05, + "loss": 0.4606, + "step": 23491 + }, + { + "epoch": 0.7, + "learning_rate": 4.3766862344479306e-05, + "loss": 0.5521, + "step": 23492 + }, + { + "epoch": 0.7, + "learning_rate": 4.3758889183969e-05, + "loss": 0.4277, + "step": 23493 + }, + { + "epoch": 0.7, + "learning_rate": 4.3750916546366085e-05, + "loss": 0.5019, + "step": 23494 + }, + { + "epoch": 0.7, + "learning_rate": 4.374294443174477e-05, + "loss": 0.446, + "step": 23495 + }, + { + "epoch": 0.7, + "learning_rate": 4.37349728401791e-05, + "loss": 0.5138, + "step": 23496 + }, + { + "epoch": 0.7, + "learning_rate": 4.3727001771743224e-05, + "loss": 0.4847, + "step": 23497 + }, + { + "epoch": 0.7, + "learning_rate": 4.371903122651131e-05, + "loss": 0.4847, + "step": 23498 + }, + { + "epoch": 0.7, + "learning_rate": 4.371106120455739e-05, + "loss": 0.5108, + "step": 23499 + }, + { + "epoch": 0.7, + "learning_rate": 4.3703091705955565e-05, + "loss": 0.5581, + "step": 23500 + }, + { + "epoch": 0.7, + "learning_rate": 4.369512273077995e-05, + "loss": 0.4111, + "step": 23501 + }, + { + "epoch": 0.7, + "learning_rate": 4.368715427910468e-05, + "loss": 0.5159, + "step": 23502 + }, + { + "epoch": 0.7, + "learning_rate": 4.367918635100381e-05, + "loss": 0.5251, + "step": 23503 + }, + { + "epoch": 0.7, + "learning_rate": 4.367121894655138e-05, + "loss": 0.4752, + "step": 23504 + }, + { + "epoch": 0.7, + "learning_rate": 4.36632520658215e-05, + "loss": 0.4613, + "step": 23505 + }, + { + "epoch": 0.7, + "learning_rate": 4.365528570888829e-05, + "loss": 0.5584, + "step": 23506 + }, + { + "epoch": 0.7, + "learning_rate": 4.364731987582574e-05, + "loss": 0.5781, + "step": 23507 + }, + { + "epoch": 0.7, + "learning_rate": 4.363935456670799e-05, + "loss": 0.6084, + "step": 23508 + }, + { + "epoch": 0.7, + "learning_rate": 4.363138978160901e-05, + "loss": 0.5986, + "step": 23509 + }, + { + "epoch": 0.7, + "learning_rate": 4.362342552060296e-05, + "loss": 0.4576, + "step": 23510 + }, + { + "epoch": 0.7, + "learning_rate": 4.361546178376377e-05, + "loss": 0.4776, + "step": 23511 + }, + { + "epoch": 0.7, + "learning_rate": 4.360749857116555e-05, + "loss": 0.5955, + "step": 23512 + }, + { + "epoch": 0.7, + "learning_rate": 4.359953588288237e-05, + "loss": 0.5963, + "step": 23513 + }, + { + "epoch": 0.7, + "learning_rate": 4.3591573718988234e-05, + "loss": 0.5272, + "step": 23514 + }, + { + "epoch": 0.7, + "learning_rate": 4.358361207955712e-05, + "loss": 0.5194, + "step": 23515 + }, + { + "epoch": 0.7, + "learning_rate": 4.3575650964663096e-05, + "loss": 0.599, + "step": 23516 + }, + { + "epoch": 0.7, + "learning_rate": 4.356769037438022e-05, + "loss": 0.6318, + "step": 23517 + }, + { + "epoch": 0.7, + "learning_rate": 4.3559730308782464e-05, + "loss": 0.5447, + "step": 23518 + }, + { + "epoch": 0.7, + "learning_rate": 4.35517707679438e-05, + "loss": 0.6111, + "step": 23519 + }, + { + "epoch": 0.7, + "learning_rate": 4.354381175193828e-05, + "loss": 0.5281, + "step": 23520 + }, + { + "epoch": 0.7, + "learning_rate": 4.3535853260839895e-05, + "loss": 0.5278, + "step": 23521 + }, + { + "epoch": 0.7, + "learning_rate": 4.352789529472272e-05, + "loss": 0.4963, + "step": 23522 + }, + { + "epoch": 0.7, + "learning_rate": 4.3519937853660585e-05, + "loss": 0.6303, + "step": 23523 + }, + { + "epoch": 0.7, + "learning_rate": 4.351198093772757e-05, + "loss": 0.5764, + "step": 23524 + }, + { + "epoch": 0.7, + "learning_rate": 4.350402454699762e-05, + "loss": 0.4432, + "step": 23525 + }, + { + "epoch": 0.7, + "learning_rate": 4.349606868154479e-05, + "loss": 0.4553, + "step": 23526 + }, + { + "epoch": 0.7, + "learning_rate": 4.3488113341442996e-05, + "loss": 0.4603, + "step": 23527 + }, + { + "epoch": 0.7, + "learning_rate": 4.348015852676615e-05, + "loss": 0.48, + "step": 23528 + }, + { + "epoch": 0.7, + "learning_rate": 4.34722042375883e-05, + "loss": 0.5682, + "step": 23529 + }, + { + "epoch": 0.7, + "learning_rate": 4.346425047398333e-05, + "loss": 0.6526, + "step": 23530 + }, + { + "epoch": 0.7, + "learning_rate": 4.3456297236025234e-05, + "loss": 0.5039, + "step": 23531 + }, + { + "epoch": 0.7, + "learning_rate": 4.3448344523787986e-05, + "loss": 0.5534, + "step": 23532 + }, + { + "epoch": 0.7, + "learning_rate": 4.3440392337345484e-05, + "loss": 0.5391, + "step": 23533 + }, + { + "epoch": 0.7, + "learning_rate": 4.343244067677164e-05, + "loss": 0.4359, + "step": 23534 + }, + { + "epoch": 0.7, + "learning_rate": 4.342448954214041e-05, + "loss": 0.5701, + "step": 23535 + }, + { + "epoch": 0.7, + "learning_rate": 4.3416538933525776e-05, + "loss": 0.6272, + "step": 23536 + }, + { + "epoch": 0.7, + "learning_rate": 4.3408588851001605e-05, + "loss": 0.549, + "step": 23537 + }, + { + "epoch": 0.7, + "learning_rate": 4.340063929464178e-05, + "loss": 0.5649, + "step": 23538 + }, + { + "epoch": 0.7, + "learning_rate": 4.339269026452025e-05, + "loss": 0.5109, + "step": 23539 + }, + { + "epoch": 0.7, + "learning_rate": 4.3384741760710936e-05, + "loss": 0.533, + "step": 23540 + }, + { + "epoch": 0.7, + "learning_rate": 4.3376793783287764e-05, + "loss": 0.6253, + "step": 23541 + }, + { + "epoch": 0.7, + "learning_rate": 4.3368846332324584e-05, + "loss": 0.536, + "step": 23542 + }, + { + "epoch": 0.7, + "learning_rate": 4.336089940789527e-05, + "loss": 0.5335, + "step": 23543 + }, + { + "epoch": 0.7, + "learning_rate": 4.335295301007374e-05, + "loss": 0.583, + "step": 23544 + }, + { + "epoch": 0.7, + "learning_rate": 4.3345007138933913e-05, + "loss": 0.6235, + "step": 23545 + }, + { + "epoch": 0.7, + "learning_rate": 4.333706179454963e-05, + "loss": 0.4471, + "step": 23546 + }, + { + "epoch": 0.7, + "learning_rate": 4.332911697699472e-05, + "loss": 0.5181, + "step": 23547 + }, + { + "epoch": 0.7, + "learning_rate": 4.332117268634309e-05, + "loss": 0.4107, + "step": 23548 + }, + { + "epoch": 0.7, + "learning_rate": 4.331322892266866e-05, + "loss": 0.5131, + "step": 23549 + }, + { + "epoch": 0.7, + "learning_rate": 4.330528568604518e-05, + "loss": 0.5297, + "step": 23550 + }, + { + "epoch": 0.7, + "learning_rate": 4.329734297654661e-05, + "loss": 0.54, + "step": 23551 + }, + { + "epoch": 0.7, + "learning_rate": 4.328940079424671e-05, + "loss": 0.5094, + "step": 23552 + }, + { + "epoch": 0.7, + "learning_rate": 4.328145913921937e-05, + "loss": 0.5235, + "step": 23553 + }, + { + "epoch": 0.7, + "learning_rate": 4.3273518011538406e-05, + "loss": 0.5487, + "step": 23554 + }, + { + "epoch": 0.7, + "learning_rate": 4.326557741127769e-05, + "loss": 0.5449, + "step": 23555 + }, + { + "epoch": 0.7, + "learning_rate": 4.325763733851098e-05, + "loss": 0.5255, + "step": 23556 + }, + { + "epoch": 0.7, + "learning_rate": 4.324969779331219e-05, + "loss": 0.4968, + "step": 23557 + }, + { + "epoch": 0.7, + "learning_rate": 4.3241758775755045e-05, + "loss": 0.6414, + "step": 23558 + }, + { + "epoch": 0.7, + "learning_rate": 4.323382028591342e-05, + "loss": 0.5212, + "step": 23559 + }, + { + "epoch": 0.7, + "learning_rate": 4.3225882323861136e-05, + "loss": 0.53, + "step": 23560 + }, + { + "epoch": 0.7, + "learning_rate": 4.321794488967197e-05, + "loss": 0.4829, + "step": 23561 + }, + { + "epoch": 0.7, + "learning_rate": 4.3210007983419686e-05, + "loss": 0.4979, + "step": 23562 + }, + { + "epoch": 0.7, + "learning_rate": 4.320207160517811e-05, + "loss": 0.4967, + "step": 23563 + }, + { + "epoch": 0.7, + "learning_rate": 4.319413575502107e-05, + "loss": 0.4717, + "step": 23564 + }, + { + "epoch": 0.7, + "learning_rate": 4.318620043302233e-05, + "loss": 0.5452, + "step": 23565 + }, + { + "epoch": 0.7, + "learning_rate": 4.31782656392556e-05, + "loss": 0.442, + "step": 23566 + }, + { + "epoch": 0.7, + "learning_rate": 4.3170331373794725e-05, + "loss": 0.6356, + "step": 23567 + }, + { + "epoch": 0.7, + "learning_rate": 4.316239763671348e-05, + "loss": 0.5142, + "step": 23568 + }, + { + "epoch": 0.7, + "learning_rate": 4.315446442808558e-05, + "loss": 0.5254, + "step": 23569 + }, + { + "epoch": 0.7, + "learning_rate": 4.3146531747984854e-05, + "loss": 0.5599, + "step": 23570 + }, + { + "epoch": 0.7, + "learning_rate": 4.313859959648498e-05, + "loss": 0.5616, + "step": 23571 + }, + { + "epoch": 0.7, + "learning_rate": 4.3130667973659765e-05, + "loss": 0.4879, + "step": 23572 + }, + { + "epoch": 0.7, + "learning_rate": 4.312273687958292e-05, + "loss": 0.5731, + "step": 23573 + }, + { + "epoch": 0.7, + "learning_rate": 4.311480631432818e-05, + "loss": 0.5662, + "step": 23574 + }, + { + "epoch": 0.7, + "learning_rate": 4.3106876277969345e-05, + "loss": 0.5699, + "step": 23575 + }, + { + "epoch": 0.7, + "learning_rate": 4.309894677058009e-05, + "loss": 0.4935, + "step": 23576 + }, + { + "epoch": 0.7, + "learning_rate": 4.309101779223412e-05, + "loss": 0.5165, + "step": 23577 + }, + { + "epoch": 0.7, + "learning_rate": 4.3083089343005176e-05, + "loss": 0.6157, + "step": 23578 + }, + { + "epoch": 0.7, + "learning_rate": 4.307516142296704e-05, + "loss": 0.4352, + "step": 23579 + }, + { + "epoch": 0.7, + "learning_rate": 4.306723403219334e-05, + "loss": 0.4718, + "step": 23580 + }, + { + "epoch": 0.7, + "learning_rate": 4.305930717075779e-05, + "loss": 0.3774, + "step": 23581 + }, + { + "epoch": 0.7, + "learning_rate": 4.305138083873409e-05, + "loss": 0.4853, + "step": 23582 + }, + { + "epoch": 0.7, + "learning_rate": 4.3043455036195967e-05, + "loss": 0.5615, + "step": 23583 + }, + { + "epoch": 0.7, + "learning_rate": 4.303552976321713e-05, + "loss": 0.5123, + "step": 23584 + }, + { + "epoch": 0.7, + "learning_rate": 4.302760501987123e-05, + "loss": 0.5858, + "step": 23585 + }, + { + "epoch": 0.7, + "learning_rate": 4.301968080623192e-05, + "loss": 0.5736, + "step": 23586 + }, + { + "epoch": 0.7, + "learning_rate": 4.3011757122372906e-05, + "loss": 0.5512, + "step": 23587 + }, + { + "epoch": 0.7, + "learning_rate": 4.30038339683679e-05, + "loss": 0.6582, + "step": 23588 + }, + { + "epoch": 0.7, + "learning_rate": 4.299591134429053e-05, + "loss": 0.5236, + "step": 23589 + }, + { + "epoch": 0.7, + "learning_rate": 4.298798925021442e-05, + "loss": 0.532, + "step": 23590 + }, + { + "epoch": 0.7, + "learning_rate": 4.2980067686213254e-05, + "loss": 0.5484, + "step": 23591 + }, + { + "epoch": 0.7, + "learning_rate": 4.297214665236075e-05, + "loss": 0.4512, + "step": 23592 + }, + { + "epoch": 0.7, + "learning_rate": 4.2964226148730436e-05, + "loss": 0.4787, + "step": 23593 + }, + { + "epoch": 0.7, + "learning_rate": 4.2956306175396065e-05, + "loss": 0.6114, + "step": 23594 + }, + { + "epoch": 0.7, + "learning_rate": 4.2948386732431225e-05, + "loss": 0.5091, + "step": 23595 + }, + { + "epoch": 0.7, + "learning_rate": 4.2940467819909505e-05, + "loss": 0.5958, + "step": 23596 + }, + { + "epoch": 0.7, + "learning_rate": 4.293254943790458e-05, + "loss": 0.5222, + "step": 23597 + }, + { + "epoch": 0.7, + "learning_rate": 4.2924631586490096e-05, + "loss": 0.5679, + "step": 23598 + }, + { + "epoch": 0.7, + "learning_rate": 4.291671426573963e-05, + "loss": 0.5079, + "step": 23599 + }, + { + "epoch": 0.7, + "learning_rate": 4.290879747572678e-05, + "loss": 0.4759, + "step": 23600 + }, + { + "epoch": 0.7, + "learning_rate": 4.290088121652518e-05, + "loss": 0.614, + "step": 23601 + }, + { + "epoch": 0.7, + "learning_rate": 4.289296548820843e-05, + "loss": 0.5496, + "step": 23602 + }, + { + "epoch": 0.7, + "learning_rate": 4.2885050290850145e-05, + "loss": 0.6188, + "step": 23603 + }, + { + "epoch": 0.7, + "learning_rate": 4.287713562452391e-05, + "loss": 0.5521, + "step": 23604 + }, + { + "epoch": 0.7, + "learning_rate": 4.286922148930327e-05, + "loss": 0.5459, + "step": 23605 + }, + { + "epoch": 0.7, + "learning_rate": 4.286130788526182e-05, + "loss": 0.5333, + "step": 23606 + }, + { + "epoch": 0.7, + "learning_rate": 4.285339481247321e-05, + "loss": 0.6417, + "step": 23607 + }, + { + "epoch": 0.7, + "learning_rate": 4.284548227101096e-05, + "loss": 0.5929, + "step": 23608 + }, + { + "epoch": 0.7, + "learning_rate": 4.283757026094859e-05, + "loss": 0.4853, + "step": 23609 + }, + { + "epoch": 0.7, + "learning_rate": 4.282965878235972e-05, + "loss": 0.5993, + "step": 23610 + }, + { + "epoch": 0.7, + "learning_rate": 4.2821747835317925e-05, + "loss": 0.6445, + "step": 23611 + }, + { + "epoch": 0.7, + "learning_rate": 4.2813837419896696e-05, + "loss": 0.5795, + "step": 23612 + }, + { + "epoch": 0.7, + "learning_rate": 4.280592753616964e-05, + "loss": 0.5125, + "step": 23613 + }, + { + "epoch": 0.7, + "learning_rate": 4.2798018184210254e-05, + "loss": 0.4648, + "step": 23614 + }, + { + "epoch": 0.7, + "learning_rate": 4.2790109364092125e-05, + "loss": 0.4259, + "step": 23615 + }, + { + "epoch": 0.7, + "learning_rate": 4.2782201075888725e-05, + "loss": 0.5164, + "step": 23616 + }, + { + "epoch": 0.7, + "learning_rate": 4.277429331967365e-05, + "loss": 0.4688, + "step": 23617 + }, + { + "epoch": 0.7, + "learning_rate": 4.276638609552035e-05, + "loss": 0.5738, + "step": 23618 + }, + { + "epoch": 0.7, + "learning_rate": 4.2758479403502416e-05, + "loss": 0.5385, + "step": 23619 + }, + { + "epoch": 0.7, + "learning_rate": 4.275057324369329e-05, + "loss": 0.4378, + "step": 23620 + }, + { + "epoch": 0.7, + "learning_rate": 4.274266761616652e-05, + "loss": 0.5259, + "step": 23621 + }, + { + "epoch": 0.7, + "learning_rate": 4.273476252099564e-05, + "loss": 0.4811, + "step": 23622 + }, + { + "epoch": 0.7, + "learning_rate": 4.272685795825412e-05, + "loss": 0.4637, + "step": 23623 + }, + { + "epoch": 0.7, + "learning_rate": 4.27189539280154e-05, + "loss": 0.6161, + "step": 23624 + }, + { + "epoch": 0.7, + "learning_rate": 4.2711050430353014e-05, + "loss": 0.4676, + "step": 23625 + }, + { + "epoch": 0.7, + "learning_rate": 4.270314746534046e-05, + "loss": 0.5907, + "step": 23626 + }, + { + "epoch": 0.7, + "learning_rate": 4.269524503305128e-05, + "loss": 0.5322, + "step": 23627 + }, + { + "epoch": 0.7, + "learning_rate": 4.268734313355879e-05, + "loss": 0.5071, + "step": 23628 + }, + { + "epoch": 0.7, + "learning_rate": 4.2679441766936545e-05, + "loss": 0.5244, + "step": 23629 + }, + { + "epoch": 0.7, + "learning_rate": 4.267154093325804e-05, + "loss": 0.5452, + "step": 23630 + }, + { + "epoch": 0.7, + "learning_rate": 4.266364063259667e-05, + "loss": 0.5615, + "step": 23631 + }, + { + "epoch": 0.7, + "learning_rate": 4.265574086502594e-05, + "loss": 0.5684, + "step": 23632 + }, + { + "epoch": 0.7, + "learning_rate": 4.264784163061924e-05, + "loss": 0.4214, + "step": 23633 + }, + { + "epoch": 0.7, + "learning_rate": 4.263994292945009e-05, + "loss": 0.6721, + "step": 23634 + }, + { + "epoch": 0.7, + "learning_rate": 4.263204476159186e-05, + "loss": 0.5401, + "step": 23635 + }, + { + "epoch": 0.7, + "learning_rate": 4.262414712711801e-05, + "loss": 0.4783, + "step": 23636 + }, + { + "epoch": 0.7, + "learning_rate": 4.261625002610201e-05, + "loss": 0.4894, + "step": 23637 + }, + { + "epoch": 0.7, + "learning_rate": 4.260835345861724e-05, + "loss": 0.5038, + "step": 23638 + }, + { + "epoch": 0.7, + "learning_rate": 4.26004574247371e-05, + "loss": 0.5576, + "step": 23639 + }, + { + "epoch": 0.7, + "learning_rate": 4.259256192453502e-05, + "loss": 0.6307, + "step": 23640 + }, + { + "epoch": 0.7, + "learning_rate": 4.258466695808446e-05, + "loss": 0.5904, + "step": 23641 + }, + { + "epoch": 0.7, + "learning_rate": 4.257677252545877e-05, + "loss": 0.4078, + "step": 23642 + }, + { + "epoch": 0.7, + "learning_rate": 4.2568878626731344e-05, + "loss": 0.5554, + "step": 23643 + }, + { + "epoch": 0.7, + "learning_rate": 4.2560985261975595e-05, + "loss": 0.5333, + "step": 23644 + }, + { + "epoch": 0.7, + "learning_rate": 4.2553092431264896e-05, + "loss": 0.4974, + "step": 23645 + }, + { + "epoch": 0.7, + "learning_rate": 4.254520013467269e-05, + "loss": 0.477, + "step": 23646 + }, + { + "epoch": 0.7, + "learning_rate": 4.253730837227232e-05, + "loss": 0.5104, + "step": 23647 + }, + { + "epoch": 0.7, + "learning_rate": 4.2529417144137116e-05, + "loss": 0.6566, + "step": 23648 + }, + { + "epoch": 0.7, + "learning_rate": 4.252152645034049e-05, + "loss": 0.4755, + "step": 23649 + }, + { + "epoch": 0.7, + "learning_rate": 4.251363629095583e-05, + "loss": 0.4651, + "step": 23650 + }, + { + "epoch": 0.7, + "learning_rate": 4.2505746666056465e-05, + "loss": 0.5114, + "step": 23651 + }, + { + "epoch": 0.7, + "learning_rate": 4.2497857575715725e-05, + "loss": 0.4531, + "step": 23652 + }, + { + "epoch": 0.7, + "learning_rate": 4.248996902000699e-05, + "loss": 0.543, + "step": 23653 + }, + { + "epoch": 0.7, + "learning_rate": 4.2482080999003624e-05, + "loss": 0.5195, + "step": 23654 + }, + { + "epoch": 0.7, + "learning_rate": 4.247419351277893e-05, + "loss": 0.5294, + "step": 23655 + }, + { + "epoch": 0.7, + "learning_rate": 4.2466306561406275e-05, + "loss": 0.5631, + "step": 23656 + }, + { + "epoch": 0.7, + "learning_rate": 4.245842014495894e-05, + "loss": 0.557, + "step": 23657 + }, + { + "epoch": 0.7, + "learning_rate": 4.245053426351031e-05, + "loss": 0.6025, + "step": 23658 + }, + { + "epoch": 0.7, + "learning_rate": 4.2442648917133655e-05, + "loss": 0.5929, + "step": 23659 + }, + { + "epoch": 0.7, + "learning_rate": 4.243476410590234e-05, + "loss": 0.6027, + "step": 23660 + }, + { + "epoch": 0.7, + "learning_rate": 4.2426879829889645e-05, + "loss": 0.5506, + "step": 23661 + }, + { + "epoch": 0.7, + "learning_rate": 4.241899608916883e-05, + "loss": 0.6095, + "step": 23662 + }, + { + "epoch": 0.7, + "learning_rate": 4.241111288381324e-05, + "loss": 0.6396, + "step": 23663 + }, + { + "epoch": 0.7, + "learning_rate": 4.2403230213896185e-05, + "loss": 0.5364, + "step": 23664 + }, + { + "epoch": 0.7, + "learning_rate": 4.239534807949096e-05, + "loss": 0.5185, + "step": 23665 + }, + { + "epoch": 0.7, + "learning_rate": 4.238746648067084e-05, + "loss": 0.5854, + "step": 23666 + }, + { + "epoch": 0.7, + "learning_rate": 4.237958541750906e-05, + "loss": 0.5289, + "step": 23667 + }, + { + "epoch": 0.7, + "learning_rate": 4.237170489007892e-05, + "loss": 0.5238, + "step": 23668 + }, + { + "epoch": 0.7, + "learning_rate": 4.2363824898453744e-05, + "loss": 0.534, + "step": 23669 + }, + { + "epoch": 0.7, + "learning_rate": 4.2355945442706746e-05, + "loss": 0.5226, + "step": 23670 + }, + { + "epoch": 0.7, + "learning_rate": 4.234806652291116e-05, + "loss": 0.4265, + "step": 23671 + }, + { + "epoch": 0.7, + "learning_rate": 4.234018813914027e-05, + "loss": 0.571, + "step": 23672 + }, + { + "epoch": 0.7, + "learning_rate": 4.233231029146737e-05, + "loss": 0.5026, + "step": 23673 + }, + { + "epoch": 0.7, + "learning_rate": 4.232443297996563e-05, + "loss": 0.5627, + "step": 23674 + }, + { + "epoch": 0.7, + "learning_rate": 4.2316556204708355e-05, + "loss": 0.5432, + "step": 23675 + }, + { + "epoch": 0.7, + "learning_rate": 4.230867996576872e-05, + "loss": 0.5419, + "step": 23676 + }, + { + "epoch": 0.7, + "learning_rate": 4.230080426322002e-05, + "loss": 0.484, + "step": 23677 + }, + { + "epoch": 0.7, + "learning_rate": 4.22929290971354e-05, + "loss": 0.4727, + "step": 23678 + }, + { + "epoch": 0.7, + "learning_rate": 4.2285054467588133e-05, + "loss": 0.4685, + "step": 23679 + }, + { + "epoch": 0.7, + "learning_rate": 4.227718037465147e-05, + "loss": 0.5281, + "step": 23680 + }, + { + "epoch": 0.71, + "learning_rate": 4.2269306818398567e-05, + "loss": 0.625, + "step": 23681 + }, + { + "epoch": 0.71, + "learning_rate": 4.2261433798902616e-05, + "loss": 0.468, + "step": 23682 + }, + { + "epoch": 0.71, + "learning_rate": 4.2253561316236835e-05, + "loss": 0.5037, + "step": 23683 + }, + { + "epoch": 0.71, + "learning_rate": 4.224568937047446e-05, + "loss": 0.5564, + "step": 23684 + }, + { + "epoch": 0.71, + "learning_rate": 4.223781796168864e-05, + "loss": 0.5307, + "step": 23685 + }, + { + "epoch": 0.71, + "learning_rate": 4.2229947089952536e-05, + "loss": 0.4136, + "step": 23686 + }, + { + "epoch": 0.71, + "learning_rate": 4.2222076755339365e-05, + "loss": 0.5062, + "step": 23687 + }, + { + "epoch": 0.71, + "learning_rate": 4.2214206957922296e-05, + "loss": 0.6197, + "step": 23688 + }, + { + "epoch": 0.71, + "learning_rate": 4.2206337697774565e-05, + "loss": 0.621, + "step": 23689 + }, + { + "epoch": 0.71, + "learning_rate": 4.2198468974969195e-05, + "loss": 0.5667, + "step": 23690 + }, + { + "epoch": 0.71, + "learning_rate": 4.2190600789579437e-05, + "loss": 0.5511, + "step": 23691 + }, + { + "epoch": 0.71, + "learning_rate": 4.218273314167842e-05, + "loss": 0.4992, + "step": 23692 + }, + { + "epoch": 0.71, + "learning_rate": 4.2174866031339354e-05, + "loss": 0.5675, + "step": 23693 + }, + { + "epoch": 0.71, + "learning_rate": 4.216699945863533e-05, + "loss": 0.5477, + "step": 23694 + }, + { + "epoch": 0.71, + "learning_rate": 4.215913342363946e-05, + "loss": 0.546, + "step": 23695 + }, + { + "epoch": 0.71, + "learning_rate": 4.215126792642495e-05, + "loss": 0.4685, + "step": 23696 + }, + { + "epoch": 0.71, + "learning_rate": 4.214340296706486e-05, + "loss": 0.465, + "step": 23697 + }, + { + "epoch": 0.71, + "learning_rate": 4.213553854563235e-05, + "loss": 0.6913, + "step": 23698 + }, + { + "epoch": 0.71, + "learning_rate": 4.212767466220057e-05, + "loss": 0.4706, + "step": 23699 + }, + { + "epoch": 0.71, + "learning_rate": 4.2119811316842614e-05, + "loss": 0.5447, + "step": 23700 + }, + { + "epoch": 0.71, + "learning_rate": 4.211194850963154e-05, + "loss": 0.6664, + "step": 23701 + }, + { + "epoch": 0.71, + "learning_rate": 4.2104086240640506e-05, + "loss": 0.531, + "step": 23702 + }, + { + "epoch": 0.71, + "learning_rate": 4.209622450994263e-05, + "loss": 0.4737, + "step": 23703 + }, + { + "epoch": 0.71, + "learning_rate": 4.208836331761099e-05, + "loss": 0.5017, + "step": 23704 + }, + { + "epoch": 0.71, + "learning_rate": 4.208050266371862e-05, + "loss": 0.3789, + "step": 23705 + }, + { + "epoch": 0.71, + "learning_rate": 4.207264254833866e-05, + "loss": 0.472, + "step": 23706 + }, + { + "epoch": 0.71, + "learning_rate": 4.206478297154417e-05, + "loss": 0.652, + "step": 23707 + }, + { + "epoch": 0.71, + "learning_rate": 4.205692393340829e-05, + "loss": 0.4893, + "step": 23708 + }, + { + "epoch": 0.71, + "learning_rate": 4.2049065434004023e-05, + "loss": 0.4603, + "step": 23709 + }, + { + "epoch": 0.71, + "learning_rate": 4.204120747340442e-05, + "loss": 0.6075, + "step": 23710 + }, + { + "epoch": 0.71, + "learning_rate": 4.2033350051682576e-05, + "loss": 0.5088, + "step": 23711 + }, + { + "epoch": 0.71, + "learning_rate": 4.202549316891157e-05, + "loss": 0.559, + "step": 23712 + }, + { + "epoch": 0.71, + "learning_rate": 4.201763682516441e-05, + "loss": 0.5009, + "step": 23713 + }, + { + "epoch": 0.71, + "learning_rate": 4.200978102051413e-05, + "loss": 0.484, + "step": 23714 + }, + { + "epoch": 0.71, + "learning_rate": 4.200192575503379e-05, + "loss": 0.5489, + "step": 23715 + }, + { + "epoch": 0.71, + "learning_rate": 4.199407102879647e-05, + "loss": 0.4771, + "step": 23716 + }, + { + "epoch": 0.71, + "learning_rate": 4.198621684187511e-05, + "loss": 0.5315, + "step": 23717 + }, + { + "epoch": 0.71, + "learning_rate": 4.197836319434284e-05, + "loss": 0.6529, + "step": 23718 + }, + { + "epoch": 0.71, + "learning_rate": 4.197051008627258e-05, + "loss": 0.4788, + "step": 23719 + }, + { + "epoch": 0.71, + "learning_rate": 4.1962657517737416e-05, + "loss": 0.4272, + "step": 23720 + }, + { + "epoch": 0.71, + "learning_rate": 4.19548054888103e-05, + "loss": 0.5256, + "step": 23721 + }, + { + "epoch": 0.71, + "learning_rate": 4.19469539995643e-05, + "loss": 0.4905, + "step": 23722 + }, + { + "epoch": 0.71, + "learning_rate": 4.193910305007236e-05, + "loss": 0.5439, + "step": 23723 + }, + { + "epoch": 0.71, + "learning_rate": 4.1931252640407525e-05, + "loss": 0.492, + "step": 23724 + }, + { + "epoch": 0.71, + "learning_rate": 4.192340277064273e-05, + "loss": 0.6193, + "step": 23725 + }, + { + "epoch": 0.71, + "learning_rate": 4.1915553440850976e-05, + "loss": 0.4775, + "step": 23726 + }, + { + "epoch": 0.71, + "learning_rate": 4.1907704651105305e-05, + "loss": 0.5188, + "step": 23727 + }, + { + "epoch": 0.71, + "learning_rate": 4.189985640147863e-05, + "loss": 0.4602, + "step": 23728 + }, + { + "epoch": 0.71, + "learning_rate": 4.1892008692043916e-05, + "loss": 0.5993, + "step": 23729 + }, + { + "epoch": 0.71, + "learning_rate": 4.188416152287413e-05, + "loss": 0.5679, + "step": 23730 + }, + { + "epoch": 0.71, + "learning_rate": 4.187631489404229e-05, + "loss": 0.5009, + "step": 23731 + }, + { + "epoch": 0.71, + "learning_rate": 4.186846880562131e-05, + "loss": 0.578, + "step": 23732 + }, + { + "epoch": 0.71, + "learning_rate": 4.186062325768409e-05, + "loss": 0.6056, + "step": 23733 + }, + { + "epoch": 0.71, + "learning_rate": 4.1852778250303637e-05, + "loss": 0.4628, + "step": 23734 + }, + { + "epoch": 0.71, + "learning_rate": 4.184493378355291e-05, + "loss": 0.4737, + "step": 23735 + }, + { + "epoch": 0.71, + "learning_rate": 4.183708985750477e-05, + "loss": 0.5428, + "step": 23736 + }, + { + "epoch": 0.71, + "learning_rate": 4.1829246472232234e-05, + "loss": 0.5713, + "step": 23737 + }, + { + "epoch": 0.71, + "learning_rate": 4.182140362780813e-05, + "loss": 0.5025, + "step": 23738 + }, + { + "epoch": 0.71, + "learning_rate": 4.181356132430547e-05, + "loss": 0.4715, + "step": 23739 + }, + { + "epoch": 0.71, + "learning_rate": 4.1805719561797086e-05, + "loss": 0.4476, + "step": 23740 + }, + { + "epoch": 0.71, + "learning_rate": 4.179787834035593e-05, + "loss": 0.4693, + "step": 23741 + }, + { + "epoch": 0.71, + "learning_rate": 4.179003766005495e-05, + "loss": 0.5378, + "step": 23742 + }, + { + "epoch": 0.71, + "learning_rate": 4.1782197520966984e-05, + "loss": 0.563, + "step": 23743 + }, + { + "epoch": 0.71, + "learning_rate": 4.1774357923164906e-05, + "loss": 0.4819, + "step": 23744 + }, + { + "epoch": 0.71, + "learning_rate": 4.176651886672165e-05, + "loss": 0.4171, + "step": 23745 + }, + { + "epoch": 0.71, + "learning_rate": 4.175868035171013e-05, + "loss": 0.5277, + "step": 23746 + }, + { + "epoch": 0.71, + "learning_rate": 4.175084237820318e-05, + "loss": 0.5533, + "step": 23747 + }, + { + "epoch": 0.71, + "learning_rate": 4.1743004946273646e-05, + "loss": 0.6171, + "step": 23748 + }, + { + "epoch": 0.71, + "learning_rate": 4.173516805599445e-05, + "loss": 0.5619, + "step": 23749 + }, + { + "epoch": 0.71, + "learning_rate": 4.172733170743841e-05, + "loss": 0.47, + "step": 23750 + }, + { + "epoch": 0.71, + "learning_rate": 4.171949590067848e-05, + "loss": 0.6376, + "step": 23751 + }, + { + "epoch": 0.71, + "learning_rate": 4.171166063578743e-05, + "loss": 0.6122, + "step": 23752 + }, + { + "epoch": 0.71, + "learning_rate": 4.170382591283809e-05, + "loss": 0.4785, + "step": 23753 + }, + { + "epoch": 0.71, + "learning_rate": 4.169599173190335e-05, + "loss": 0.4771, + "step": 23754 + }, + { + "epoch": 0.71, + "learning_rate": 4.168815809305607e-05, + "loss": 0.4689, + "step": 23755 + }, + { + "epoch": 0.71, + "learning_rate": 4.168032499636907e-05, + "loss": 0.5521, + "step": 23756 + }, + { + "epoch": 0.71, + "learning_rate": 4.167249244191511e-05, + "loss": 0.499, + "step": 23757 + }, + { + "epoch": 0.71, + "learning_rate": 4.166466042976709e-05, + "loss": 0.4747, + "step": 23758 + }, + { + "epoch": 0.71, + "learning_rate": 4.165682895999783e-05, + "loss": 0.5056, + "step": 23759 + }, + { + "epoch": 0.71, + "learning_rate": 4.164899803268008e-05, + "loss": 0.5317, + "step": 23760 + }, + { + "epoch": 0.71, + "learning_rate": 4.164116764788673e-05, + "loss": 0.5385, + "step": 23761 + }, + { + "epoch": 0.71, + "learning_rate": 4.163333780569055e-05, + "loss": 0.5002, + "step": 23762 + }, + { + "epoch": 0.71, + "learning_rate": 4.162550850616429e-05, + "loss": 0.4941, + "step": 23763 + }, + { + "epoch": 0.71, + "learning_rate": 4.161767974938079e-05, + "loss": 0.5787, + "step": 23764 + }, + { + "epoch": 0.71, + "learning_rate": 4.160985153541287e-05, + "loss": 0.563, + "step": 23765 + }, + { + "epoch": 0.71, + "learning_rate": 4.160202386433327e-05, + "loss": 0.5092, + "step": 23766 + }, + { + "epoch": 0.71, + "learning_rate": 4.159419673621475e-05, + "loss": 0.5228, + "step": 23767 + }, + { + "epoch": 0.71, + "learning_rate": 4.15863701511301e-05, + "loss": 0.4467, + "step": 23768 + }, + { + "epoch": 0.71, + "learning_rate": 4.157854410915212e-05, + "loss": 0.5787, + "step": 23769 + }, + { + "epoch": 0.71, + "learning_rate": 4.157071861035358e-05, + "loss": 0.5186, + "step": 23770 + }, + { + "epoch": 0.71, + "learning_rate": 4.15628936548072e-05, + "loss": 0.5792, + "step": 23771 + }, + { + "epoch": 0.71, + "learning_rate": 4.155506924258571e-05, + "loss": 0.6216, + "step": 23772 + }, + { + "epoch": 0.71, + "learning_rate": 4.15472453737619e-05, + "loss": 0.4597, + "step": 23773 + }, + { + "epoch": 0.71, + "learning_rate": 4.153942204840854e-05, + "loss": 0.5177, + "step": 23774 + }, + { + "epoch": 0.71, + "learning_rate": 4.153159926659833e-05, + "loss": 0.5546, + "step": 23775 + }, + { + "epoch": 0.71, + "learning_rate": 4.152377702840398e-05, + "loss": 0.5857, + "step": 23776 + }, + { + "epoch": 0.71, + "learning_rate": 4.151595533389824e-05, + "loss": 0.5263, + "step": 23777 + }, + { + "epoch": 0.71, + "learning_rate": 4.150813418315387e-05, + "loss": 0.4779, + "step": 23778 + }, + { + "epoch": 0.71, + "learning_rate": 4.150031357624352e-05, + "loss": 0.6497, + "step": 23779 + }, + { + "epoch": 0.71, + "learning_rate": 4.1492493513239974e-05, + "loss": 0.5078, + "step": 23780 + }, + { + "epoch": 0.71, + "learning_rate": 4.148467399421587e-05, + "loss": 0.4727, + "step": 23781 + }, + { + "epoch": 0.71, + "learning_rate": 4.1476855019243985e-05, + "loss": 0.5464, + "step": 23782 + }, + { + "epoch": 0.71, + "learning_rate": 4.1469036588396936e-05, + "loss": 0.3957, + "step": 23783 + }, + { + "epoch": 0.71, + "learning_rate": 4.1461218701747496e-05, + "loss": 0.5052, + "step": 23784 + }, + { + "epoch": 0.71, + "learning_rate": 4.145340135936827e-05, + "loss": 0.5336, + "step": 23785 + }, + { + "epoch": 0.71, + "learning_rate": 4.144558456133203e-05, + "loss": 0.6418, + "step": 23786 + }, + { + "epoch": 0.71, + "learning_rate": 4.1437768307711366e-05, + "loss": 0.5113, + "step": 23787 + }, + { + "epoch": 0.71, + "learning_rate": 4.142995259857899e-05, + "loss": 0.6241, + "step": 23788 + }, + { + "epoch": 0.71, + "learning_rate": 4.142213743400761e-05, + "loss": 0.4977, + "step": 23789 + }, + { + "epoch": 0.71, + "learning_rate": 4.141432281406985e-05, + "loss": 0.5515, + "step": 23790 + }, + { + "epoch": 0.71, + "learning_rate": 4.140650873883833e-05, + "loss": 0.6697, + "step": 23791 + }, + { + "epoch": 0.71, + "learning_rate": 4.139869520838574e-05, + "loss": 0.5131, + "step": 23792 + }, + { + "epoch": 0.71, + "learning_rate": 4.1390882222784764e-05, + "loss": 0.5427, + "step": 23793 + }, + { + "epoch": 0.71, + "learning_rate": 4.1383069782107996e-05, + "loss": 0.5764, + "step": 23794 + }, + { + "epoch": 0.71, + "learning_rate": 4.1375257886428056e-05, + "loss": 0.483, + "step": 23795 + }, + { + "epoch": 0.71, + "learning_rate": 4.13674465358176e-05, + "loss": 0.5689, + "step": 23796 + }, + { + "epoch": 0.71, + "learning_rate": 4.135963573034929e-05, + "loss": 0.6222, + "step": 23797 + }, + { + "epoch": 0.71, + "learning_rate": 4.1351825470095684e-05, + "loss": 0.6048, + "step": 23798 + }, + { + "epoch": 0.71, + "learning_rate": 4.134401575512946e-05, + "loss": 0.5758, + "step": 23799 + }, + { + "epoch": 0.71, + "learning_rate": 4.1336206585523175e-05, + "loss": 0.5021, + "step": 23800 + }, + { + "epoch": 0.71, + "learning_rate": 4.1328397961349484e-05, + "loss": 0.5205, + "step": 23801 + }, + { + "epoch": 0.71, + "learning_rate": 4.132058988268094e-05, + "loss": 0.5248, + "step": 23802 + }, + { + "epoch": 0.71, + "learning_rate": 4.131278234959015e-05, + "loss": 0.5685, + "step": 23803 + }, + { + "epoch": 0.71, + "learning_rate": 4.130497536214977e-05, + "loss": 0.522, + "step": 23804 + }, + { + "epoch": 0.71, + "learning_rate": 4.129716892043232e-05, + "loss": 0.5484, + "step": 23805 + }, + { + "epoch": 0.71, + "learning_rate": 4.128936302451037e-05, + "loss": 0.4667, + "step": 23806 + }, + { + "epoch": 0.71, + "learning_rate": 4.1281557674456514e-05, + "loss": 0.5459, + "step": 23807 + }, + { + "epoch": 0.71, + "learning_rate": 4.127375287034338e-05, + "loss": 0.5378, + "step": 23808 + }, + { + "epoch": 0.71, + "learning_rate": 4.126594861224349e-05, + "loss": 0.5339, + "step": 23809 + }, + { + "epoch": 0.71, + "learning_rate": 4.1258144900229345e-05, + "loss": 0.4987, + "step": 23810 + }, + { + "epoch": 0.71, + "learning_rate": 4.125034173437357e-05, + "loss": 0.5744, + "step": 23811 + }, + { + "epoch": 0.71, + "learning_rate": 4.12425391147487e-05, + "loss": 0.5895, + "step": 23812 + }, + { + "epoch": 0.71, + "learning_rate": 4.123473704142732e-05, + "loss": 0.6164, + "step": 23813 + }, + { + "epoch": 0.71, + "learning_rate": 4.1226935514481936e-05, + "loss": 0.5027, + "step": 23814 + }, + { + "epoch": 0.71, + "learning_rate": 4.121913453398504e-05, + "loss": 0.4869, + "step": 23815 + }, + { + "epoch": 0.71, + "learning_rate": 4.12113341000092e-05, + "loss": 0.5212, + "step": 23816 + }, + { + "epoch": 0.71, + "learning_rate": 4.1203534212626994e-05, + "loss": 0.5228, + "step": 23817 + }, + { + "epoch": 0.71, + "learning_rate": 4.1195734871910876e-05, + "loss": 0.5363, + "step": 23818 + }, + { + "epoch": 0.71, + "learning_rate": 4.118793607793335e-05, + "loss": 0.4045, + "step": 23819 + }, + { + "epoch": 0.71, + "learning_rate": 4.118013783076695e-05, + "loss": 0.6009, + "step": 23820 + }, + { + "epoch": 0.71, + "learning_rate": 4.1172340130484235e-05, + "loss": 0.61, + "step": 23821 + }, + { + "epoch": 0.71, + "learning_rate": 4.116454297715761e-05, + "loss": 0.469, + "step": 23822 + }, + { + "epoch": 0.71, + "learning_rate": 4.1156746370859645e-05, + "loss": 0.4318, + "step": 23823 + }, + { + "epoch": 0.71, + "learning_rate": 4.114895031166277e-05, + "loss": 0.4323, + "step": 23824 + }, + { + "epoch": 0.71, + "learning_rate": 4.1141154799639535e-05, + "loss": 0.5542, + "step": 23825 + }, + { + "epoch": 0.71, + "learning_rate": 4.1133359834862354e-05, + "loss": 0.5104, + "step": 23826 + }, + { + "epoch": 0.71, + "learning_rate": 4.112556541740376e-05, + "loss": 0.4608, + "step": 23827 + }, + { + "epoch": 0.71, + "learning_rate": 4.111777154733619e-05, + "loss": 0.458, + "step": 23828 + }, + { + "epoch": 0.71, + "learning_rate": 4.110997822473208e-05, + "loss": 0.4844, + "step": 23829 + }, + { + "epoch": 0.71, + "learning_rate": 4.110218544966392e-05, + "loss": 0.601, + "step": 23830 + }, + { + "epoch": 0.71, + "learning_rate": 4.1094393222204164e-05, + "loss": 0.5747, + "step": 23831 + }, + { + "epoch": 0.71, + "learning_rate": 4.108660154242531e-05, + "loss": 0.5114, + "step": 23832 + }, + { + "epoch": 0.71, + "learning_rate": 4.107881041039974e-05, + "loss": 0.5413, + "step": 23833 + }, + { + "epoch": 0.71, + "learning_rate": 4.107101982619987e-05, + "loss": 0.6588, + "step": 23834 + }, + { + "epoch": 0.71, + "learning_rate": 4.1063229789898186e-05, + "loss": 0.4681, + "step": 23835 + }, + { + "epoch": 0.71, + "learning_rate": 4.105544030156713e-05, + "loss": 0.5295, + "step": 23836 + }, + { + "epoch": 0.71, + "learning_rate": 4.104765136127909e-05, + "loss": 0.5733, + "step": 23837 + }, + { + "epoch": 0.71, + "learning_rate": 4.103986296910647e-05, + "loss": 0.4663, + "step": 23838 + }, + { + "epoch": 0.71, + "learning_rate": 4.1032075125121696e-05, + "loss": 0.5479, + "step": 23839 + }, + { + "epoch": 0.71, + "learning_rate": 4.1024287829397224e-05, + "loss": 0.5146, + "step": 23840 + }, + { + "epoch": 0.71, + "learning_rate": 4.101650108200538e-05, + "loss": 0.4779, + "step": 23841 + }, + { + "epoch": 0.71, + "learning_rate": 4.100871488301863e-05, + "loss": 0.5853, + "step": 23842 + }, + { + "epoch": 0.71, + "learning_rate": 4.100092923250931e-05, + "loss": 0.4897, + "step": 23843 + }, + { + "epoch": 0.71, + "learning_rate": 4.099314413054987e-05, + "loss": 0.5297, + "step": 23844 + }, + { + "epoch": 0.71, + "learning_rate": 4.0985359577212625e-05, + "loss": 0.55, + "step": 23845 + }, + { + "epoch": 0.71, + "learning_rate": 4.097757557256998e-05, + "loss": 0.5983, + "step": 23846 + }, + { + "epoch": 0.71, + "learning_rate": 4.096979211669435e-05, + "loss": 0.5656, + "step": 23847 + }, + { + "epoch": 0.71, + "learning_rate": 4.096200920965807e-05, + "loss": 0.5903, + "step": 23848 + }, + { + "epoch": 0.71, + "learning_rate": 4.095422685153345e-05, + "loss": 0.5347, + "step": 23849 + }, + { + "epoch": 0.71, + "learning_rate": 4.094644504239291e-05, + "loss": 0.5655, + "step": 23850 + }, + { + "epoch": 0.71, + "learning_rate": 4.0938663782308816e-05, + "loss": 0.5004, + "step": 23851 + }, + { + "epoch": 0.71, + "learning_rate": 4.0930883071353484e-05, + "loss": 0.4153, + "step": 23852 + }, + { + "epoch": 0.71, + "learning_rate": 4.092310290959923e-05, + "loss": 0.6376, + "step": 23853 + }, + { + "epoch": 0.71, + "learning_rate": 4.091532329711842e-05, + "loss": 0.4647, + "step": 23854 + }, + { + "epoch": 0.71, + "learning_rate": 4.0907544233983387e-05, + "loss": 0.6042, + "step": 23855 + }, + { + "epoch": 0.71, + "learning_rate": 4.089976572026648e-05, + "loss": 0.5934, + "step": 23856 + }, + { + "epoch": 0.71, + "learning_rate": 4.0891987756040005e-05, + "loss": 0.4014, + "step": 23857 + }, + { + "epoch": 0.71, + "learning_rate": 4.088421034137624e-05, + "loss": 0.5577, + "step": 23858 + }, + { + "epoch": 0.71, + "learning_rate": 4.0876433476347556e-05, + "loss": 0.471, + "step": 23859 + }, + { + "epoch": 0.71, + "learning_rate": 4.086865716102618e-05, + "loss": 0.538, + "step": 23860 + }, + { + "epoch": 0.71, + "learning_rate": 4.0860881395484516e-05, + "loss": 0.5795, + "step": 23861 + }, + { + "epoch": 0.71, + "learning_rate": 4.0853106179794764e-05, + "loss": 0.5815, + "step": 23862 + }, + { + "epoch": 0.71, + "learning_rate": 4.0845331514029294e-05, + "loss": 0.5524, + "step": 23863 + }, + { + "epoch": 0.71, + "learning_rate": 4.0837557398260315e-05, + "loss": 0.4815, + "step": 23864 + }, + { + "epoch": 0.71, + "learning_rate": 4.0829783832560156e-05, + "loss": 0.6197, + "step": 23865 + }, + { + "epoch": 0.71, + "learning_rate": 4.0822010817001113e-05, + "loss": 0.4403, + "step": 23866 + }, + { + "epoch": 0.71, + "learning_rate": 4.081423835165542e-05, + "loss": 0.5175, + "step": 23867 + }, + { + "epoch": 0.71, + "learning_rate": 4.080646643659532e-05, + "loss": 0.5679, + "step": 23868 + }, + { + "epoch": 0.71, + "learning_rate": 4.079869507189311e-05, + "loss": 0.5677, + "step": 23869 + }, + { + "epoch": 0.71, + "learning_rate": 4.079092425762107e-05, + "loss": 0.4656, + "step": 23870 + }, + { + "epoch": 0.71, + "learning_rate": 4.078315399385142e-05, + "loss": 0.5865, + "step": 23871 + }, + { + "epoch": 0.71, + "learning_rate": 4.0775384280656356e-05, + "loss": 0.6607, + "step": 23872 + }, + { + "epoch": 0.71, + "learning_rate": 4.0767615118108174e-05, + "loss": 0.5581, + "step": 23873 + }, + { + "epoch": 0.71, + "learning_rate": 4.0759846506279095e-05, + "loss": 0.4938, + "step": 23874 + }, + { + "epoch": 0.71, + "learning_rate": 4.075207844524139e-05, + "loss": 0.6395, + "step": 23875 + }, + { + "epoch": 0.71, + "learning_rate": 4.074431093506725e-05, + "loss": 0.4515, + "step": 23876 + }, + { + "epoch": 0.71, + "learning_rate": 4.073654397582884e-05, + "loss": 0.4666, + "step": 23877 + }, + { + "epoch": 0.71, + "learning_rate": 4.0728777567598444e-05, + "loss": 0.4873, + "step": 23878 + }, + { + "epoch": 0.71, + "learning_rate": 4.072101171044828e-05, + "loss": 0.5841, + "step": 23879 + }, + { + "epoch": 0.71, + "learning_rate": 4.071324640445051e-05, + "loss": 0.5563, + "step": 23880 + }, + { + "epoch": 0.71, + "learning_rate": 4.0705481649677326e-05, + "loss": 0.5198, + "step": 23881 + }, + { + "epoch": 0.71, + "learning_rate": 4.069771744620093e-05, + "loss": 0.4882, + "step": 23882 + }, + { + "epoch": 0.71, + "learning_rate": 4.0689953794093574e-05, + "loss": 0.4209, + "step": 23883 + }, + { + "epoch": 0.71, + "learning_rate": 4.068219069342734e-05, + "loss": 0.5423, + "step": 23884 + }, + { + "epoch": 0.71, + "learning_rate": 4.067442814427449e-05, + "loss": 0.602, + "step": 23885 + }, + { + "epoch": 0.71, + "learning_rate": 4.0666666146707125e-05, + "loss": 0.6013, + "step": 23886 + }, + { + "epoch": 0.71, + "learning_rate": 4.065890470079748e-05, + "loss": 0.5651, + "step": 23887 + }, + { + "epoch": 0.71, + "learning_rate": 4.065114380661768e-05, + "loss": 0.4943, + "step": 23888 + }, + { + "epoch": 0.71, + "learning_rate": 4.0643383464239905e-05, + "loss": 0.4907, + "step": 23889 + }, + { + "epoch": 0.71, + "learning_rate": 4.063562367373626e-05, + "loss": 0.4781, + "step": 23890 + }, + { + "epoch": 0.71, + "learning_rate": 4.0627864435178965e-05, + "loss": 0.5648, + "step": 23891 + }, + { + "epoch": 0.71, + "learning_rate": 4.0620105748640084e-05, + "loss": 0.5473, + "step": 23892 + }, + { + "epoch": 0.71, + "learning_rate": 4.0612347614191795e-05, + "loss": 0.462, + "step": 23893 + }, + { + "epoch": 0.71, + "learning_rate": 4.0604590031906254e-05, + "loss": 0.4811, + "step": 23894 + }, + { + "epoch": 0.71, + "learning_rate": 4.059683300185557e-05, + "loss": 0.572, + "step": 23895 + }, + { + "epoch": 0.71, + "learning_rate": 4.058907652411182e-05, + "loss": 0.4976, + "step": 23896 + }, + { + "epoch": 0.71, + "learning_rate": 4.0581320598747165e-05, + "loss": 0.519, + "step": 23897 + }, + { + "epoch": 0.71, + "learning_rate": 4.057356522583374e-05, + "loss": 0.4968, + "step": 23898 + }, + { + "epoch": 0.71, + "learning_rate": 4.056581040544363e-05, + "loss": 0.5723, + "step": 23899 + }, + { + "epoch": 0.71, + "learning_rate": 4.055805613764887e-05, + "loss": 0.5216, + "step": 23900 + }, + { + "epoch": 0.71, + "learning_rate": 4.0550302422521626e-05, + "loss": 0.5603, + "step": 23901 + }, + { + "epoch": 0.71, + "learning_rate": 4.054254926013401e-05, + "loss": 0.4691, + "step": 23902 + }, + { + "epoch": 0.71, + "learning_rate": 4.053479665055804e-05, + "loss": 0.4865, + "step": 23903 + }, + { + "epoch": 0.71, + "learning_rate": 4.052704459386587e-05, + "loss": 0.62, + "step": 23904 + }, + { + "epoch": 0.71, + "learning_rate": 4.051929309012949e-05, + "loss": 0.4847, + "step": 23905 + }, + { + "epoch": 0.71, + "learning_rate": 4.0511542139421055e-05, + "loss": 0.4928, + "step": 23906 + }, + { + "epoch": 0.71, + "learning_rate": 4.0503791741812555e-05, + "loss": 0.4594, + "step": 23907 + }, + { + "epoch": 0.71, + "learning_rate": 4.049604189737609e-05, + "loss": 0.534, + "step": 23908 + }, + { + "epoch": 0.71, + "learning_rate": 4.0488292606183745e-05, + "loss": 0.5381, + "step": 23909 + }, + { + "epoch": 0.71, + "learning_rate": 4.048054386830754e-05, + "loss": 0.5669, + "step": 23910 + }, + { + "epoch": 0.71, + "learning_rate": 4.047279568381947e-05, + "loss": 0.5458, + "step": 23911 + }, + { + "epoch": 0.71, + "learning_rate": 4.046504805279163e-05, + "loss": 0.5883, + "step": 23912 + }, + { + "epoch": 0.71, + "learning_rate": 4.045730097529607e-05, + "loss": 0.5462, + "step": 23913 + }, + { + "epoch": 0.71, + "learning_rate": 4.044955445140482e-05, + "loss": 0.4692, + "step": 23914 + }, + { + "epoch": 0.71, + "learning_rate": 4.0441808481189814e-05, + "loss": 0.5694, + "step": 23915 + }, + { + "epoch": 0.71, + "learning_rate": 4.043406306472315e-05, + "loss": 0.5635, + "step": 23916 + }, + { + "epoch": 0.71, + "learning_rate": 4.042631820207682e-05, + "loss": 0.3291, + "step": 23917 + }, + { + "epoch": 0.71, + "learning_rate": 4.041857389332288e-05, + "loss": 0.5902, + "step": 23918 + }, + { + "epoch": 0.71, + "learning_rate": 4.0410830138533294e-05, + "loss": 0.4956, + "step": 23919 + }, + { + "epoch": 0.71, + "learning_rate": 4.040308693778002e-05, + "loss": 0.6657, + "step": 23920 + }, + { + "epoch": 0.71, + "learning_rate": 4.0395344291135086e-05, + "loss": 0.5436, + "step": 23921 + }, + { + "epoch": 0.71, + "learning_rate": 4.0387602198670526e-05, + "loss": 0.6035, + "step": 23922 + }, + { + "epoch": 0.71, + "learning_rate": 4.037986066045828e-05, + "loss": 0.596, + "step": 23923 + }, + { + "epoch": 0.71, + "learning_rate": 4.0372119676570295e-05, + "loss": 0.6295, + "step": 23924 + }, + { + "epoch": 0.71, + "learning_rate": 4.036437924707861e-05, + "loss": 0.5254, + "step": 23925 + }, + { + "epoch": 0.71, + "learning_rate": 4.035663937205513e-05, + "loss": 0.4354, + "step": 23926 + }, + { + "epoch": 0.71, + "learning_rate": 4.0348900051571834e-05, + "loss": 0.5028, + "step": 23927 + }, + { + "epoch": 0.71, + "learning_rate": 4.0341161285700736e-05, + "loss": 0.5847, + "step": 23928 + }, + { + "epoch": 0.71, + "learning_rate": 4.033342307451373e-05, + "loss": 0.6427, + "step": 23929 + }, + { + "epoch": 0.71, + "learning_rate": 4.032568541808275e-05, + "loss": 0.6326, + "step": 23930 + }, + { + "epoch": 0.71, + "learning_rate": 4.031794831647976e-05, + "loss": 0.5098, + "step": 23931 + }, + { + "epoch": 0.71, + "learning_rate": 4.031021176977674e-05, + "loss": 0.5, + "step": 23932 + }, + { + "epoch": 0.71, + "learning_rate": 4.030247577804557e-05, + "loss": 0.5819, + "step": 23933 + }, + { + "epoch": 0.71, + "learning_rate": 4.029474034135816e-05, + "loss": 0.5402, + "step": 23934 + }, + { + "epoch": 0.71, + "learning_rate": 4.0287005459786454e-05, + "loss": 0.4566, + "step": 23935 + }, + { + "epoch": 0.71, + "learning_rate": 4.0279271133402376e-05, + "loss": 0.4908, + "step": 23936 + }, + { + "epoch": 0.71, + "learning_rate": 4.0271537362277865e-05, + "loss": 0.5986, + "step": 23937 + }, + { + "epoch": 0.71, + "learning_rate": 4.02638041464848e-05, + "loss": 0.511, + "step": 23938 + }, + { + "epoch": 0.71, + "learning_rate": 4.025607148609504e-05, + "loss": 0.6123, + "step": 23939 + }, + { + "epoch": 0.71, + "learning_rate": 4.024833938118051e-05, + "loss": 0.5212, + "step": 23940 + }, + { + "epoch": 0.71, + "learning_rate": 4.0240607831813135e-05, + "loss": 0.5193, + "step": 23941 + }, + { + "epoch": 0.71, + "learning_rate": 4.023287683806478e-05, + "loss": 0.4994, + "step": 23942 + }, + { + "epoch": 0.71, + "learning_rate": 4.0225146400007286e-05, + "loss": 0.6486, + "step": 23943 + }, + { + "epoch": 0.71, + "learning_rate": 4.021741651771255e-05, + "loss": 0.5199, + "step": 23944 + }, + { + "epoch": 0.71, + "learning_rate": 4.020968719125248e-05, + "loss": 0.5093, + "step": 23945 + }, + { + "epoch": 0.71, + "learning_rate": 4.0201958420698874e-05, + "loss": 0.6208, + "step": 23946 + }, + { + "epoch": 0.71, + "learning_rate": 4.019423020612366e-05, + "loss": 0.5964, + "step": 23947 + }, + { + "epoch": 0.71, + "learning_rate": 4.0186502547598626e-05, + "loss": 0.4639, + "step": 23948 + }, + { + "epoch": 0.71, + "learning_rate": 4.017877544519568e-05, + "loss": 0.5772, + "step": 23949 + }, + { + "epoch": 0.71, + "learning_rate": 4.017104889898662e-05, + "loss": 0.5641, + "step": 23950 + }, + { + "epoch": 0.71, + "learning_rate": 4.0163322909043324e-05, + "loss": 0.5327, + "step": 23951 + }, + { + "epoch": 0.71, + "learning_rate": 4.0155597475437564e-05, + "loss": 0.4705, + "step": 23952 + }, + { + "epoch": 0.71, + "learning_rate": 4.014787259824124e-05, + "loss": 0.5985, + "step": 23953 + }, + { + "epoch": 0.71, + "learning_rate": 4.014014827752613e-05, + "loss": 0.4691, + "step": 23954 + }, + { + "epoch": 0.71, + "learning_rate": 4.0132424513364045e-05, + "loss": 0.5852, + "step": 23955 + }, + { + "epoch": 0.71, + "learning_rate": 4.012470130582685e-05, + "loss": 0.3945, + "step": 23956 + }, + { + "epoch": 0.71, + "learning_rate": 4.0116978654986314e-05, + "loss": 0.5656, + "step": 23957 + }, + { + "epoch": 0.71, + "learning_rate": 4.010925656091422e-05, + "loss": 0.5802, + "step": 23958 + }, + { + "epoch": 0.71, + "learning_rate": 4.010153502368239e-05, + "loss": 0.4681, + "step": 23959 + }, + { + "epoch": 0.71, + "learning_rate": 4.009381404336264e-05, + "loss": 0.4776, + "step": 23960 + }, + { + "epoch": 0.71, + "learning_rate": 4.008609362002673e-05, + "loss": 0.5194, + "step": 23961 + }, + { + "epoch": 0.71, + "learning_rate": 4.007837375374642e-05, + "loss": 0.6054, + "step": 23962 + }, + { + "epoch": 0.71, + "learning_rate": 4.00706544445935e-05, + "loss": 0.465, + "step": 23963 + }, + { + "epoch": 0.71, + "learning_rate": 4.006293569263978e-05, + "loss": 0.3716, + "step": 23964 + }, + { + "epoch": 0.71, + "learning_rate": 4.0055217497956963e-05, + "loss": 0.573, + "step": 23965 + }, + { + "epoch": 0.71, + "learning_rate": 4.0047499860616885e-05, + "loss": 0.5892, + "step": 23966 + }, + { + "epoch": 0.71, + "learning_rate": 4.003978278069121e-05, + "loss": 0.6064, + "step": 23967 + }, + { + "epoch": 0.71, + "learning_rate": 4.003206625825179e-05, + "loss": 0.5135, + "step": 23968 + }, + { + "epoch": 0.71, + "learning_rate": 4.0024350293370274e-05, + "loss": 0.5445, + "step": 23969 + }, + { + "epoch": 0.71, + "learning_rate": 4.001663488611844e-05, + "loss": 0.5026, + "step": 23970 + }, + { + "epoch": 0.71, + "learning_rate": 4.000892003656806e-05, + "loss": 0.4309, + "step": 23971 + }, + { + "epoch": 0.71, + "learning_rate": 4.000120574479084e-05, + "loss": 0.5966, + "step": 23972 + }, + { + "epoch": 0.71, + "learning_rate": 3.9993492010858455e-05, + "loss": 0.623, + "step": 23973 + }, + { + "epoch": 0.71, + "learning_rate": 3.998577883484267e-05, + "loss": 0.5533, + "step": 23974 + }, + { + "epoch": 0.71, + "learning_rate": 3.9978066216815237e-05, + "loss": 0.5531, + "step": 23975 + }, + { + "epoch": 0.71, + "learning_rate": 3.997035415684781e-05, + "loss": 0.5481, + "step": 23976 + }, + { + "epoch": 0.71, + "learning_rate": 3.996264265501207e-05, + "loss": 0.4926, + "step": 23977 + }, + { + "epoch": 0.71, + "learning_rate": 3.995493171137975e-05, + "loss": 0.4913, + "step": 23978 + }, + { + "epoch": 0.71, + "learning_rate": 3.994722132602254e-05, + "loss": 0.5792, + "step": 23979 + }, + { + "epoch": 0.71, + "learning_rate": 3.993951149901217e-05, + "loss": 0.5429, + "step": 23980 + }, + { + "epoch": 0.71, + "learning_rate": 3.993180223042029e-05, + "loss": 0.566, + "step": 23981 + }, + { + "epoch": 0.71, + "learning_rate": 3.9924093520318537e-05, + "loss": 0.4639, + "step": 23982 + }, + { + "epoch": 0.71, + "learning_rate": 3.991638536877861e-05, + "loss": 0.4275, + "step": 23983 + }, + { + "epoch": 0.71, + "learning_rate": 3.990867777587223e-05, + "loss": 0.5509, + "step": 23984 + }, + { + "epoch": 0.71, + "learning_rate": 3.990097074167101e-05, + "loss": 0.5126, + "step": 23985 + }, + { + "epoch": 0.71, + "learning_rate": 3.9893264266246567e-05, + "loss": 0.6105, + "step": 23986 + }, + { + "epoch": 0.71, + "learning_rate": 3.9885558349670605e-05, + "loss": 0.5372, + "step": 23987 + }, + { + "epoch": 0.71, + "learning_rate": 3.9877852992014795e-05, + "loss": 0.5395, + "step": 23988 + }, + { + "epoch": 0.71, + "learning_rate": 3.987014819335071e-05, + "loss": 0.5059, + "step": 23989 + }, + { + "epoch": 0.71, + "learning_rate": 3.986244395375006e-05, + "loss": 0.4847, + "step": 23990 + }, + { + "epoch": 0.71, + "learning_rate": 3.985474027328444e-05, + "loss": 0.4694, + "step": 23991 + }, + { + "epoch": 0.71, + "learning_rate": 3.9847037152025434e-05, + "loss": 0.5647, + "step": 23992 + }, + { + "epoch": 0.71, + "learning_rate": 3.98393345900447e-05, + "loss": 0.5834, + "step": 23993 + }, + { + "epoch": 0.71, + "learning_rate": 3.9831632587413894e-05, + "loss": 0.4739, + "step": 23994 + }, + { + "epoch": 0.71, + "learning_rate": 3.982393114420458e-05, + "loss": 0.5359, + "step": 23995 + }, + { + "epoch": 0.71, + "learning_rate": 3.981623026048835e-05, + "loss": 0.6403, + "step": 23996 + }, + { + "epoch": 0.71, + "learning_rate": 3.980852993633681e-05, + "loss": 0.5909, + "step": 23997 + }, + { + "epoch": 0.71, + "learning_rate": 3.980083017182158e-05, + "loss": 0.5527, + "step": 23998 + }, + { + "epoch": 0.71, + "learning_rate": 3.979313096701427e-05, + "loss": 0.4259, + "step": 23999 + }, + { + "epoch": 0.71, + "learning_rate": 3.978543232198642e-05, + "loss": 0.5445, + "step": 24000 + }, + { + "epoch": 0.71, + "learning_rate": 3.9777734236809606e-05, + "loss": 0.4875, + "step": 24001 + }, + { + "epoch": 0.71, + "learning_rate": 3.97700367115554e-05, + "loss": 0.4879, + "step": 24002 + }, + { + "epoch": 0.71, + "learning_rate": 3.976233974629543e-05, + "loss": 0.5436, + "step": 24003 + }, + { + "epoch": 0.71, + "learning_rate": 3.9754643341101216e-05, + "loss": 0.5052, + "step": 24004 + }, + { + "epoch": 0.71, + "learning_rate": 3.9746947496044285e-05, + "loss": 0.4842, + "step": 24005 + }, + { + "epoch": 0.71, + "learning_rate": 3.973925221119622e-05, + "loss": 0.5633, + "step": 24006 + }, + { + "epoch": 0.71, + "learning_rate": 3.97315574866286e-05, + "loss": 0.6012, + "step": 24007 + }, + { + "epoch": 0.71, + "learning_rate": 3.97238633224129e-05, + "loss": 0.5488, + "step": 24008 + }, + { + "epoch": 0.71, + "learning_rate": 3.9716169718620746e-05, + "loss": 0.5706, + "step": 24009 + }, + { + "epoch": 0.71, + "learning_rate": 3.970847667532357e-05, + "loss": 0.5914, + "step": 24010 + }, + { + "epoch": 0.71, + "learning_rate": 3.9700784192592985e-05, + "loss": 0.5752, + "step": 24011 + }, + { + "epoch": 0.71, + "learning_rate": 3.9693092270500445e-05, + "loss": 0.5556, + "step": 24012 + }, + { + "epoch": 0.71, + "learning_rate": 3.96854009091175e-05, + "loss": 0.5383, + "step": 24013 + }, + { + "epoch": 0.71, + "learning_rate": 3.967771010851568e-05, + "loss": 0.4748, + "step": 24014 + }, + { + "epoch": 0.71, + "learning_rate": 3.967001986876648e-05, + "loss": 0.4566, + "step": 24015 + }, + { + "epoch": 0.71, + "learning_rate": 3.9662330189941355e-05, + "loss": 0.6051, + "step": 24016 + }, + { + "epoch": 0.72, + "learning_rate": 3.965464107211184e-05, + "loss": 0.5603, + "step": 24017 + }, + { + "epoch": 0.72, + "learning_rate": 3.9646952515349445e-05, + "loss": 0.531, + "step": 24018 + }, + { + "epoch": 0.72, + "learning_rate": 3.963926451972564e-05, + "loss": 0.5188, + "step": 24019 + }, + { + "epoch": 0.72, + "learning_rate": 3.9631577085311856e-05, + "loss": 0.5511, + "step": 24020 + }, + { + "epoch": 0.72, + "learning_rate": 3.96238902121796e-05, + "loss": 0.5431, + "step": 24021 + }, + { + "epoch": 0.72, + "learning_rate": 3.9616203900400404e-05, + "loss": 0.4851, + "step": 24022 + }, + { + "epoch": 0.72, + "learning_rate": 3.9608518150045627e-05, + "loss": 0.5101, + "step": 24023 + }, + { + "epoch": 0.72, + "learning_rate": 3.9600832961186815e-05, + "loss": 0.4743, + "step": 24024 + }, + { + "epoch": 0.72, + "learning_rate": 3.9593148333895357e-05, + "loss": 0.4894, + "step": 24025 + }, + { + "epoch": 0.72, + "learning_rate": 3.958546426824277e-05, + "loss": 0.5862, + "step": 24026 + }, + { + "epoch": 0.72, + "learning_rate": 3.957778076430041e-05, + "loss": 0.5541, + "step": 24027 + }, + { + "epoch": 0.72, + "learning_rate": 3.957009782213981e-05, + "loss": 0.5477, + "step": 24028 + }, + { + "epoch": 0.72, + "learning_rate": 3.95624154418323e-05, + "loss": 0.5498, + "step": 24029 + }, + { + "epoch": 0.72, + "learning_rate": 3.955473362344941e-05, + "loss": 0.5303, + "step": 24030 + }, + { + "epoch": 0.72, + "learning_rate": 3.954705236706249e-05, + "loss": 0.445, + "step": 24031 + }, + { + "epoch": 0.72, + "learning_rate": 3.9539371672742966e-05, + "loss": 0.5067, + "step": 24032 + }, + { + "epoch": 0.72, + "learning_rate": 3.953169154056231e-05, + "loss": 0.4573, + "step": 24033 + }, + { + "epoch": 0.72, + "learning_rate": 3.952401197059189e-05, + "loss": 0.6339, + "step": 24034 + }, + { + "epoch": 0.72, + "learning_rate": 3.951633296290306e-05, + "loss": 0.5584, + "step": 24035 + }, + { + "epoch": 0.72, + "learning_rate": 3.950865451756727e-05, + "loss": 0.5381, + "step": 24036 + }, + { + "epoch": 0.72, + "learning_rate": 3.950097663465593e-05, + "loss": 0.6071, + "step": 24037 + }, + { + "epoch": 0.72, + "learning_rate": 3.9493299314240395e-05, + "loss": 0.5518, + "step": 24038 + }, + { + "epoch": 0.72, + "learning_rate": 3.948562255639201e-05, + "loss": 0.4552, + "step": 24039 + }, + { + "epoch": 0.72, + "learning_rate": 3.94779463611822e-05, + "loss": 0.5241, + "step": 24040 + }, + { + "epoch": 0.72, + "learning_rate": 3.9470270728682304e-05, + "loss": 0.6217, + "step": 24041 + }, + { + "epoch": 0.72, + "learning_rate": 3.946259565896375e-05, + "loss": 0.5609, + "step": 24042 + }, + { + "epoch": 0.72, + "learning_rate": 3.9454921152097855e-05, + "loss": 0.5201, + "step": 24043 + }, + { + "epoch": 0.72, + "learning_rate": 3.944724720815595e-05, + "loss": 0.4878, + "step": 24044 + }, + { + "epoch": 0.72, + "learning_rate": 3.943957382720939e-05, + "loss": 0.5736, + "step": 24045 + }, + { + "epoch": 0.72, + "learning_rate": 3.9431901009329574e-05, + "loss": 0.6487, + "step": 24046 + }, + { + "epoch": 0.72, + "learning_rate": 3.94242287545878e-05, + "loss": 0.504, + "step": 24047 + }, + { + "epoch": 0.72, + "learning_rate": 3.941655706305537e-05, + "loss": 0.5282, + "step": 24048 + }, + { + "epoch": 0.72, + "learning_rate": 3.9408885934803654e-05, + "loss": 0.4532, + "step": 24049 + }, + { + "epoch": 0.72, + "learning_rate": 3.940121536990399e-05, + "loss": 0.4748, + "step": 24050 + }, + { + "epoch": 0.72, + "learning_rate": 3.9393545368427644e-05, + "loss": 0.5432, + "step": 24051 + }, + { + "epoch": 0.72, + "learning_rate": 3.938587593044599e-05, + "loss": 0.5185, + "step": 24052 + }, + { + "epoch": 0.72, + "learning_rate": 3.937820705603028e-05, + "loss": 0.4742, + "step": 24053 + }, + { + "epoch": 0.72, + "learning_rate": 3.937053874525186e-05, + "loss": 0.4893, + "step": 24054 + }, + { + "epoch": 0.72, + "learning_rate": 3.936287099818198e-05, + "loss": 0.4761, + "step": 24055 + }, + { + "epoch": 0.72, + "learning_rate": 3.935520381489199e-05, + "loss": 0.5847, + "step": 24056 + }, + { + "epoch": 0.72, + "learning_rate": 3.934753719545314e-05, + "loss": 0.5991, + "step": 24057 + }, + { + "epoch": 0.72, + "learning_rate": 3.9339871139936694e-05, + "loss": 0.4715, + "step": 24058 + }, + { + "epoch": 0.72, + "learning_rate": 3.9332205648413936e-05, + "loss": 0.5178, + "step": 24059 + }, + { + "epoch": 0.72, + "learning_rate": 3.932454072095615e-05, + "loss": 0.5562, + "step": 24060 + }, + { + "epoch": 0.72, + "learning_rate": 3.931687635763465e-05, + "loss": 0.5517, + "step": 24061 + }, + { + "epoch": 0.72, + "learning_rate": 3.9309212558520636e-05, + "loss": 0.4573, + "step": 24062 + }, + { + "epoch": 0.72, + "learning_rate": 3.930154932368534e-05, + "loss": 0.5056, + "step": 24063 + }, + { + "epoch": 0.72, + "learning_rate": 3.929388665320005e-05, + "loss": 0.5385, + "step": 24064 + }, + { + "epoch": 0.72, + "learning_rate": 3.928622454713604e-05, + "loss": 0.5107, + "step": 24065 + }, + { + "epoch": 0.72, + "learning_rate": 3.927856300556452e-05, + "loss": 0.6657, + "step": 24066 + }, + { + "epoch": 0.72, + "learning_rate": 3.9270902028556675e-05, + "loss": 0.5105, + "step": 24067 + }, + { + "epoch": 0.72, + "learning_rate": 3.9263241616183786e-05, + "loss": 0.677, + "step": 24068 + }, + { + "epoch": 0.72, + "learning_rate": 3.9255581768517105e-05, + "loss": 0.6239, + "step": 24069 + }, + { + "epoch": 0.72, + "learning_rate": 3.9247922485627774e-05, + "loss": 0.5134, + "step": 24070 + }, + { + "epoch": 0.72, + "learning_rate": 3.9240263767587094e-05, + "loss": 0.5605, + "step": 24071 + }, + { + "epoch": 0.72, + "learning_rate": 3.9232605614466175e-05, + "loss": 0.5531, + "step": 24072 + }, + { + "epoch": 0.72, + "learning_rate": 3.92249480263363e-05, + "loss": 0.5422, + "step": 24073 + }, + { + "epoch": 0.72, + "learning_rate": 3.921729100326862e-05, + "loss": 0.5943, + "step": 24074 + }, + { + "epoch": 0.72, + "learning_rate": 3.920963454533434e-05, + "loss": 0.4669, + "step": 24075 + }, + { + "epoch": 0.72, + "learning_rate": 3.9201978652604674e-05, + "loss": 0.4567, + "step": 24076 + }, + { + "epoch": 0.72, + "learning_rate": 3.919432332515078e-05, + "loss": 0.5305, + "step": 24077 + }, + { + "epoch": 0.72, + "learning_rate": 3.918666856304381e-05, + "loss": 0.5962, + "step": 24078 + }, + { + "epoch": 0.72, + "learning_rate": 3.917901436635495e-05, + "loss": 0.5584, + "step": 24079 + }, + { + "epoch": 0.72, + "learning_rate": 3.9171360735155405e-05, + "loss": 0.5723, + "step": 24080 + }, + { + "epoch": 0.72, + "learning_rate": 3.91637076695163e-05, + "loss": 0.4633, + "step": 24081 + }, + { + "epoch": 0.72, + "learning_rate": 3.915605516950877e-05, + "loss": 0.6247, + "step": 24082 + }, + { + "epoch": 0.72, + "learning_rate": 3.914840323520398e-05, + "loss": 0.4969, + "step": 24083 + }, + { + "epoch": 0.72, + "learning_rate": 3.914075186667309e-05, + "loss": 0.5189, + "step": 24084 + }, + { + "epoch": 0.72, + "learning_rate": 3.9133101063987263e-05, + "loss": 0.5128, + "step": 24085 + }, + { + "epoch": 0.72, + "learning_rate": 3.9125450827217614e-05, + "loss": 0.5745, + "step": 24086 + }, + { + "epoch": 0.72, + "learning_rate": 3.911780115643522e-05, + "loss": 0.6316, + "step": 24087 + }, + { + "epoch": 0.72, + "learning_rate": 3.911015205171128e-05, + "loss": 0.5644, + "step": 24088 + }, + { + "epoch": 0.72, + "learning_rate": 3.910250351311684e-05, + "loss": 0.445, + "step": 24089 + }, + { + "epoch": 0.72, + "learning_rate": 3.9094855540723087e-05, + "loss": 0.5851, + "step": 24090 + }, + { + "epoch": 0.72, + "learning_rate": 3.908720813460106e-05, + "loss": 0.432, + "step": 24091 + }, + { + "epoch": 0.72, + "learning_rate": 3.907956129482193e-05, + "loss": 0.4748, + "step": 24092 + }, + { + "epoch": 0.72, + "learning_rate": 3.9071915021456716e-05, + "loss": 0.6242, + "step": 24093 + }, + { + "epoch": 0.72, + "learning_rate": 3.9064269314576564e-05, + "loss": 0.4873, + "step": 24094 + }, + { + "epoch": 0.72, + "learning_rate": 3.9056624174252574e-05, + "loss": 0.5613, + "step": 24095 + }, + { + "epoch": 0.72, + "learning_rate": 3.9048979600555813e-05, + "loss": 0.5727, + "step": 24096 + }, + { + "epoch": 0.72, + "learning_rate": 3.9041335593557305e-05, + "loss": 0.5635, + "step": 24097 + }, + { + "epoch": 0.72, + "learning_rate": 3.903369215332816e-05, + "loss": 0.4869, + "step": 24098 + }, + { + "epoch": 0.72, + "learning_rate": 3.902604927993949e-05, + "loss": 0.6352, + "step": 24099 + }, + { + "epoch": 0.72, + "learning_rate": 3.901840697346231e-05, + "loss": 0.5766, + "step": 24100 + }, + { + "epoch": 0.72, + "learning_rate": 3.901076523396764e-05, + "loss": 0.5844, + "step": 24101 + }, + { + "epoch": 0.72, + "learning_rate": 3.900312406152656e-05, + "loss": 0.6365, + "step": 24102 + }, + { + "epoch": 0.72, + "learning_rate": 3.899548345621013e-05, + "loss": 0.55, + "step": 24103 + }, + { + "epoch": 0.72, + "learning_rate": 3.898784341808942e-05, + "loss": 0.5346, + "step": 24104 + }, + { + "epoch": 0.72, + "learning_rate": 3.898020394723542e-05, + "loss": 0.5919, + "step": 24105 + }, + { + "epoch": 0.72, + "learning_rate": 3.897256504371912e-05, + "loss": 0.4818, + "step": 24106 + }, + { + "epoch": 0.72, + "learning_rate": 3.896492670761159e-05, + "loss": 0.5631, + "step": 24107 + }, + { + "epoch": 0.72, + "learning_rate": 3.895728893898389e-05, + "loss": 0.5416, + "step": 24108 + }, + { + "epoch": 0.72, + "learning_rate": 3.894965173790699e-05, + "loss": 0.6279, + "step": 24109 + }, + { + "epoch": 0.72, + "learning_rate": 3.894201510445184e-05, + "loss": 0.6513, + "step": 24110 + }, + { + "epoch": 0.72, + "learning_rate": 3.893437903868953e-05, + "loss": 0.5278, + "step": 24111 + }, + { + "epoch": 0.72, + "learning_rate": 3.892674354069103e-05, + "loss": 0.5901, + "step": 24112 + }, + { + "epoch": 0.72, + "learning_rate": 3.891910861052731e-05, + "loss": 0.4765, + "step": 24113 + }, + { + "epoch": 0.72, + "learning_rate": 3.8911474248269406e-05, + "loss": 0.5067, + "step": 24114 + }, + { + "epoch": 0.72, + "learning_rate": 3.890384045398823e-05, + "loss": 0.3995, + "step": 24115 + }, + { + "epoch": 0.72, + "learning_rate": 3.8896207227754835e-05, + "loss": 0.5461, + "step": 24116 + }, + { + "epoch": 0.72, + "learning_rate": 3.88885745696401e-05, + "loss": 0.5657, + "step": 24117 + }, + { + "epoch": 0.72, + "learning_rate": 3.888094247971509e-05, + "loss": 0.6385, + "step": 24118 + }, + { + "epoch": 0.72, + "learning_rate": 3.887331095805068e-05, + "loss": 0.4749, + "step": 24119 + }, + { + "epoch": 0.72, + "learning_rate": 3.8865680004717894e-05, + "loss": 0.5844, + "step": 24120 + }, + { + "epoch": 0.72, + "learning_rate": 3.885804961978763e-05, + "loss": 0.5506, + "step": 24121 + }, + { + "epoch": 0.72, + "learning_rate": 3.885041980333084e-05, + "loss": 0.5506, + "step": 24122 + }, + { + "epoch": 0.72, + "learning_rate": 3.8842790555418504e-05, + "loss": 0.5996, + "step": 24123 + }, + { + "epoch": 0.72, + "learning_rate": 3.8835161876121526e-05, + "loss": 0.6307, + "step": 24124 + }, + { + "epoch": 0.72, + "learning_rate": 3.8827533765510814e-05, + "loss": 0.4492, + "step": 24125 + }, + { + "epoch": 0.72, + "learning_rate": 3.8819906223657295e-05, + "loss": 0.4704, + "step": 24126 + }, + { + "epoch": 0.72, + "learning_rate": 3.881227925063196e-05, + "loss": 0.4974, + "step": 24127 + }, + { + "epoch": 0.72, + "learning_rate": 3.880465284650565e-05, + "loss": 0.5018, + "step": 24128 + }, + { + "epoch": 0.72, + "learning_rate": 3.879702701134926e-05, + "loss": 0.5639, + "step": 24129 + }, + { + "epoch": 0.72, + "learning_rate": 3.878940174523371e-05, + "loss": 0.5681, + "step": 24130 + }, + { + "epoch": 0.72, + "learning_rate": 3.878177704822995e-05, + "loss": 0.4253, + "step": 24131 + }, + { + "epoch": 0.72, + "learning_rate": 3.877415292040879e-05, + "loss": 0.4484, + "step": 24132 + }, + { + "epoch": 0.72, + "learning_rate": 3.8766529361841194e-05, + "loss": 0.5065, + "step": 24133 + }, + { + "epoch": 0.72, + "learning_rate": 3.875890637259796e-05, + "loss": 0.6087, + "step": 24134 + }, + { + "epoch": 0.72, + "learning_rate": 3.875128395275005e-05, + "loss": 0.5798, + "step": 24135 + }, + { + "epoch": 0.72, + "learning_rate": 3.874366210236825e-05, + "loss": 0.594, + "step": 24136 + }, + { + "epoch": 0.72, + "learning_rate": 3.8736040821523465e-05, + "loss": 0.4939, + "step": 24137 + }, + { + "epoch": 0.72, + "learning_rate": 3.872842011028659e-05, + "loss": 0.4796, + "step": 24138 + }, + { + "epoch": 0.72, + "learning_rate": 3.872079996872844e-05, + "loss": 0.4656, + "step": 24139 + }, + { + "epoch": 0.72, + "learning_rate": 3.871318039691985e-05, + "loss": 0.5541, + "step": 24140 + }, + { + "epoch": 0.72, + "learning_rate": 3.8705561394931664e-05, + "loss": 0.3925, + "step": 24141 + }, + { + "epoch": 0.72, + "learning_rate": 3.8697942962834775e-05, + "loss": 0.5584, + "step": 24142 + }, + { + "epoch": 0.72, + "learning_rate": 3.8690325100699986e-05, + "loss": 0.6314, + "step": 24143 + }, + { + "epoch": 0.72, + "learning_rate": 3.868270780859808e-05, + "loss": 0.5393, + "step": 24144 + }, + { + "epoch": 0.72, + "learning_rate": 3.8675091086599924e-05, + "loss": 0.4729, + "step": 24145 + }, + { + "epoch": 0.72, + "learning_rate": 3.866747493477633e-05, + "loss": 0.6504, + "step": 24146 + }, + { + "epoch": 0.72, + "learning_rate": 3.8659859353198136e-05, + "loss": 0.5726, + "step": 24147 + }, + { + "epoch": 0.72, + "learning_rate": 3.865224434193613e-05, + "loss": 0.5217, + "step": 24148 + }, + { + "epoch": 0.72, + "learning_rate": 3.864462990106106e-05, + "loss": 0.6031, + "step": 24149 + }, + { + "epoch": 0.72, + "learning_rate": 3.863701603064378e-05, + "loss": 0.5348, + "step": 24150 + }, + { + "epoch": 0.72, + "learning_rate": 3.8629402730755095e-05, + "loss": 0.4445, + "step": 24151 + }, + { + "epoch": 0.72, + "learning_rate": 3.862179000146576e-05, + "loss": 0.5604, + "step": 24152 + }, + { + "epoch": 0.72, + "learning_rate": 3.8614177842846535e-05, + "loss": 0.4364, + "step": 24153 + }, + { + "epoch": 0.72, + "learning_rate": 3.8606566254968245e-05, + "loss": 0.5334, + "step": 24154 + }, + { + "epoch": 0.72, + "learning_rate": 3.85989552379016e-05, + "loss": 0.5331, + "step": 24155 + }, + { + "epoch": 0.72, + "learning_rate": 3.85913447917174e-05, + "loss": 0.6511, + "step": 24156 + }, + { + "epoch": 0.72, + "learning_rate": 3.858373491648644e-05, + "loss": 0.4979, + "step": 24157 + }, + { + "epoch": 0.72, + "learning_rate": 3.8576125612279426e-05, + "loss": 0.5568, + "step": 24158 + }, + { + "epoch": 0.72, + "learning_rate": 3.8568516879167086e-05, + "loss": 0.5236, + "step": 24159 + }, + { + "epoch": 0.72, + "learning_rate": 3.856090871722018e-05, + "loss": 0.471, + "step": 24160 + }, + { + "epoch": 0.72, + "learning_rate": 3.85533011265095e-05, + "loss": 0.5962, + "step": 24161 + }, + { + "epoch": 0.72, + "learning_rate": 3.8545694107105733e-05, + "loss": 0.5478, + "step": 24162 + }, + { + "epoch": 0.72, + "learning_rate": 3.853808765907958e-05, + "loss": 0.5467, + "step": 24163 + }, + { + "epoch": 0.72, + "learning_rate": 3.853048178250179e-05, + "loss": 0.4226, + "step": 24164 + }, + { + "epoch": 0.72, + "learning_rate": 3.852287647744308e-05, + "loss": 0.5511, + "step": 24165 + }, + { + "epoch": 0.72, + "learning_rate": 3.85152717439742e-05, + "loss": 0.5954, + "step": 24166 + }, + { + "epoch": 0.72, + "learning_rate": 3.850766758216582e-05, + "loss": 0.4566, + "step": 24167 + }, + { + "epoch": 0.72, + "learning_rate": 3.85000639920886e-05, + "loss": 0.5324, + "step": 24168 + }, + { + "epoch": 0.72, + "learning_rate": 3.849246097381326e-05, + "loss": 0.5262, + "step": 24169 + }, + { + "epoch": 0.72, + "learning_rate": 3.848485852741055e-05, + "loss": 0.5515, + "step": 24170 + }, + { + "epoch": 0.72, + "learning_rate": 3.847725665295111e-05, + "loss": 0.4348, + "step": 24171 + }, + { + "epoch": 0.72, + "learning_rate": 3.846965535050559e-05, + "loss": 0.5066, + "step": 24172 + }, + { + "epoch": 0.72, + "learning_rate": 3.846205462014467e-05, + "loss": 0.4854, + "step": 24173 + }, + { + "epoch": 0.72, + "learning_rate": 3.845445446193909e-05, + "loss": 0.6704, + "step": 24174 + }, + { + "epoch": 0.72, + "learning_rate": 3.844685487595943e-05, + "loss": 0.566, + "step": 24175 + }, + { + "epoch": 0.72, + "learning_rate": 3.84392558622764e-05, + "loss": 0.5343, + "step": 24176 + }, + { + "epoch": 0.72, + "learning_rate": 3.843165742096061e-05, + "loss": 0.5192, + "step": 24177 + }, + { + "epoch": 0.72, + "learning_rate": 3.842405955208277e-05, + "loss": 0.5371, + "step": 24178 + }, + { + "epoch": 0.72, + "learning_rate": 3.8416462255713436e-05, + "loss": 0.5425, + "step": 24179 + }, + { + "epoch": 0.72, + "learning_rate": 3.8408865531923296e-05, + "loss": 0.5173, + "step": 24180 + }, + { + "epoch": 0.72, + "learning_rate": 3.8401269380783014e-05, + "loss": 0.4606, + "step": 24181 + }, + { + "epoch": 0.72, + "learning_rate": 3.839367380236317e-05, + "loss": 0.5198, + "step": 24182 + }, + { + "epoch": 0.72, + "learning_rate": 3.838607879673437e-05, + "loss": 0.4345, + "step": 24183 + }, + { + "epoch": 0.72, + "learning_rate": 3.837848436396725e-05, + "loss": 0.4613, + "step": 24184 + }, + { + "epoch": 0.72, + "learning_rate": 3.837089050413245e-05, + "loss": 0.4385, + "step": 24185 + }, + { + "epoch": 0.72, + "learning_rate": 3.8363297217300554e-05, + "loss": 0.5063, + "step": 24186 + }, + { + "epoch": 0.72, + "learning_rate": 3.8355704503542114e-05, + "loss": 0.652, + "step": 24187 + }, + { + "epoch": 0.72, + "learning_rate": 3.8348112362927766e-05, + "loss": 0.5464, + "step": 24188 + }, + { + "epoch": 0.72, + "learning_rate": 3.834052079552813e-05, + "loss": 0.5514, + "step": 24189 + }, + { + "epoch": 0.72, + "learning_rate": 3.833292980141373e-05, + "loss": 0.6076, + "step": 24190 + }, + { + "epoch": 0.72, + "learning_rate": 3.83253393806552e-05, + "loss": 0.5933, + "step": 24191 + }, + { + "epoch": 0.72, + "learning_rate": 3.8317749533323055e-05, + "loss": 0.4936, + "step": 24192 + }, + { + "epoch": 0.72, + "learning_rate": 3.8310160259487915e-05, + "loss": 0.6379, + "step": 24193 + }, + { + "epoch": 0.72, + "learning_rate": 3.830257155922029e-05, + "loss": 0.5472, + "step": 24194 + }, + { + "epoch": 0.72, + "learning_rate": 3.829498343259079e-05, + "loss": 0.5356, + "step": 24195 + }, + { + "epoch": 0.72, + "learning_rate": 3.828739587966992e-05, + "loss": 0.6865, + "step": 24196 + }, + { + "epoch": 0.72, + "learning_rate": 3.827980890052827e-05, + "loss": 0.5394, + "step": 24197 + }, + { + "epoch": 0.72, + "learning_rate": 3.8272222495236335e-05, + "loss": 0.5274, + "step": 24198 + }, + { + "epoch": 0.72, + "learning_rate": 3.826463666386466e-05, + "loss": 0.527, + "step": 24199 + }, + { + "epoch": 0.72, + "learning_rate": 3.8257051406483836e-05, + "loss": 0.5228, + "step": 24200 + }, + { + "epoch": 0.72, + "learning_rate": 3.824946672316433e-05, + "loss": 0.4458, + "step": 24201 + }, + { + "epoch": 0.72, + "learning_rate": 3.824188261397663e-05, + "loss": 0.5324, + "step": 24202 + }, + { + "epoch": 0.72, + "learning_rate": 3.823429907899131e-05, + "loss": 0.5525, + "step": 24203 + }, + { + "epoch": 0.72, + "learning_rate": 3.822671611827888e-05, + "loss": 0.6801, + "step": 24204 + }, + { + "epoch": 0.72, + "learning_rate": 3.821913373190983e-05, + "loss": 0.5728, + "step": 24205 + }, + { + "epoch": 0.72, + "learning_rate": 3.8211551919954615e-05, + "loss": 0.6027, + "step": 24206 + }, + { + "epoch": 0.72, + "learning_rate": 3.820397068248376e-05, + "loss": 0.5464, + "step": 24207 + }, + { + "epoch": 0.72, + "learning_rate": 3.8196390019567754e-05, + "loss": 0.5092, + "step": 24208 + }, + { + "epoch": 0.72, + "learning_rate": 3.8188809931277124e-05, + "loss": 0.4995, + "step": 24209 + }, + { + "epoch": 0.72, + "learning_rate": 3.81812304176823e-05, + "loss": 0.6323, + "step": 24210 + }, + { + "epoch": 0.72, + "learning_rate": 3.817365147885372e-05, + "loss": 0.6059, + "step": 24211 + }, + { + "epoch": 0.72, + "learning_rate": 3.816607311486189e-05, + "loss": 0.5061, + "step": 24212 + }, + { + "epoch": 0.72, + "learning_rate": 3.81584953257773e-05, + "loss": 0.5091, + "step": 24213 + }, + { + "epoch": 0.72, + "learning_rate": 3.8150918111670374e-05, + "loss": 0.7233, + "step": 24214 + }, + { + "epoch": 0.72, + "learning_rate": 3.814334147261153e-05, + "loss": 0.5482, + "step": 24215 + }, + { + "epoch": 0.72, + "learning_rate": 3.8135765408671245e-05, + "loss": 0.556, + "step": 24216 + }, + { + "epoch": 0.72, + "learning_rate": 3.812818991991998e-05, + "loss": 0.4733, + "step": 24217 + }, + { + "epoch": 0.72, + "learning_rate": 3.8120615006428116e-05, + "loss": 0.5629, + "step": 24218 + }, + { + "epoch": 0.72, + "learning_rate": 3.8113040668266144e-05, + "loss": 0.4425, + "step": 24219 + }, + { + "epoch": 0.72, + "learning_rate": 3.810546690550446e-05, + "loss": 0.4995, + "step": 24220 + }, + { + "epoch": 0.72, + "learning_rate": 3.809789371821344e-05, + "loss": 0.4919, + "step": 24221 + }, + { + "epoch": 0.72, + "learning_rate": 3.809032110646353e-05, + "loss": 0.527, + "step": 24222 + }, + { + "epoch": 0.72, + "learning_rate": 3.8082749070325176e-05, + "loss": 0.4285, + "step": 24223 + }, + { + "epoch": 0.72, + "learning_rate": 3.807517760986874e-05, + "loss": 0.5233, + "step": 24224 + }, + { + "epoch": 0.72, + "learning_rate": 3.8067606725164596e-05, + "loss": 0.4102, + "step": 24225 + }, + { + "epoch": 0.72, + "learning_rate": 3.8060036416283154e-05, + "loss": 0.6022, + "step": 24226 + }, + { + "epoch": 0.72, + "learning_rate": 3.80524666832948e-05, + "loss": 0.5091, + "step": 24227 + }, + { + "epoch": 0.72, + "learning_rate": 3.804489752626996e-05, + "loss": 0.5594, + "step": 24228 + }, + { + "epoch": 0.72, + "learning_rate": 3.803732894527897e-05, + "loss": 0.4558, + "step": 24229 + }, + { + "epoch": 0.72, + "learning_rate": 3.802976094039218e-05, + "loss": 0.4909, + "step": 24230 + }, + { + "epoch": 0.72, + "learning_rate": 3.802219351167996e-05, + "loss": 0.4759, + "step": 24231 + }, + { + "epoch": 0.72, + "learning_rate": 3.801462665921271e-05, + "loss": 0.499, + "step": 24232 + }, + { + "epoch": 0.72, + "learning_rate": 3.8007060383060765e-05, + "loss": 0.5308, + "step": 24233 + }, + { + "epoch": 0.72, + "learning_rate": 3.7999494683294436e-05, + "loss": 0.6523, + "step": 24234 + }, + { + "epoch": 0.72, + "learning_rate": 3.7991929559984083e-05, + "loss": 0.511, + "step": 24235 + }, + { + "epoch": 0.72, + "learning_rate": 3.798436501320009e-05, + "loss": 0.5804, + "step": 24236 + }, + { + "epoch": 0.72, + "learning_rate": 3.7976801043012734e-05, + "loss": 0.5154, + "step": 24237 + }, + { + "epoch": 0.72, + "learning_rate": 3.7969237649492386e-05, + "loss": 0.4544, + "step": 24238 + }, + { + "epoch": 0.72, + "learning_rate": 3.79616748327093e-05, + "loss": 0.4835, + "step": 24239 + }, + { + "epoch": 0.72, + "learning_rate": 3.795411259273389e-05, + "loss": 0.4626, + "step": 24240 + }, + { + "epoch": 0.72, + "learning_rate": 3.794655092963637e-05, + "loss": 0.6031, + "step": 24241 + }, + { + "epoch": 0.72, + "learning_rate": 3.7938989843487085e-05, + "loss": 0.5577, + "step": 24242 + }, + { + "epoch": 0.72, + "learning_rate": 3.793142933435637e-05, + "loss": 0.5644, + "step": 24243 + }, + { + "epoch": 0.72, + "learning_rate": 3.792386940231448e-05, + "loss": 0.5186, + "step": 24244 + }, + { + "epoch": 0.72, + "learning_rate": 3.791631004743169e-05, + "loss": 0.5104, + "step": 24245 + }, + { + "epoch": 0.72, + "learning_rate": 3.79087512697783e-05, + "loss": 0.5531, + "step": 24246 + }, + { + "epoch": 0.72, + "learning_rate": 3.790119306942463e-05, + "loss": 0.5719, + "step": 24247 + }, + { + "epoch": 0.72, + "learning_rate": 3.7893635446440914e-05, + "loss": 0.557, + "step": 24248 + }, + { + "epoch": 0.72, + "learning_rate": 3.788607840089738e-05, + "loss": 0.5261, + "step": 24249 + }, + { + "epoch": 0.72, + "learning_rate": 3.787852193286435e-05, + "loss": 0.5205, + "step": 24250 + }, + { + "epoch": 0.72, + "learning_rate": 3.787096604241206e-05, + "loss": 0.6476, + "step": 24251 + }, + { + "epoch": 0.72, + "learning_rate": 3.786341072961079e-05, + "loss": 0.5676, + "step": 24252 + }, + { + "epoch": 0.72, + "learning_rate": 3.7855855994530773e-05, + "loss": 0.4736, + "step": 24253 + }, + { + "epoch": 0.72, + "learning_rate": 3.784830183724219e-05, + "loss": 0.5378, + "step": 24254 + }, + { + "epoch": 0.72, + "learning_rate": 3.784074825781537e-05, + "loss": 0.5752, + "step": 24255 + }, + { + "epoch": 0.72, + "learning_rate": 3.7833195256320466e-05, + "loss": 0.514, + "step": 24256 + }, + { + "epoch": 0.72, + "learning_rate": 3.7825642832827765e-05, + "loss": 0.5145, + "step": 24257 + }, + { + "epoch": 0.72, + "learning_rate": 3.781809098740743e-05, + "loss": 0.4704, + "step": 24258 + }, + { + "epoch": 0.72, + "learning_rate": 3.7810539720129726e-05, + "loss": 0.5614, + "step": 24259 + }, + { + "epoch": 0.72, + "learning_rate": 3.7802989031064805e-05, + "loss": 0.4152, + "step": 24260 + }, + { + "epoch": 0.72, + "learning_rate": 3.779543892028291e-05, + "loss": 0.5323, + "step": 24261 + }, + { + "epoch": 0.72, + "learning_rate": 3.778788938785426e-05, + "loss": 0.4686, + "step": 24262 + }, + { + "epoch": 0.72, + "learning_rate": 3.778034043384901e-05, + "loss": 0.534, + "step": 24263 + }, + { + "epoch": 0.72, + "learning_rate": 3.777279205833734e-05, + "loss": 0.512, + "step": 24264 + }, + { + "epoch": 0.72, + "learning_rate": 3.776524426138943e-05, + "loss": 0.542, + "step": 24265 + }, + { + "epoch": 0.72, + "learning_rate": 3.7757697043075526e-05, + "loss": 0.4969, + "step": 24266 + }, + { + "epoch": 0.72, + "learning_rate": 3.7750150403465747e-05, + "loss": 0.4658, + "step": 24267 + }, + { + "epoch": 0.72, + "learning_rate": 3.774260434263022e-05, + "loss": 0.5114, + "step": 24268 + }, + { + "epoch": 0.72, + "learning_rate": 3.773505886063914e-05, + "loss": 0.5879, + "step": 24269 + }, + { + "epoch": 0.72, + "learning_rate": 3.7727513957562666e-05, + "loss": 0.5237, + "step": 24270 + }, + { + "epoch": 0.72, + "learning_rate": 3.7719969633470984e-05, + "loss": 0.53, + "step": 24271 + }, + { + "epoch": 0.72, + "learning_rate": 3.77124258884342e-05, + "loss": 0.5334, + "step": 24272 + }, + { + "epoch": 0.72, + "learning_rate": 3.770488272252243e-05, + "loss": 0.5194, + "step": 24273 + }, + { + "epoch": 0.72, + "learning_rate": 3.7697340135805814e-05, + "loss": 0.5637, + "step": 24274 + }, + { + "epoch": 0.72, + "learning_rate": 3.768979812835455e-05, + "loss": 0.5302, + "step": 24275 + }, + { + "epoch": 0.72, + "learning_rate": 3.7682256700238705e-05, + "loss": 0.4472, + "step": 24276 + }, + { + "epoch": 0.72, + "learning_rate": 3.767471585152836e-05, + "loss": 0.52, + "step": 24277 + }, + { + "epoch": 0.72, + "learning_rate": 3.766717558229367e-05, + "loss": 0.57, + "step": 24278 + }, + { + "epoch": 0.72, + "learning_rate": 3.7659635892604773e-05, + "loss": 0.5378, + "step": 24279 + }, + { + "epoch": 0.72, + "learning_rate": 3.76520967825317e-05, + "loss": 0.4669, + "step": 24280 + }, + { + "epoch": 0.72, + "learning_rate": 3.764455825214461e-05, + "loss": 0.4937, + "step": 24281 + }, + { + "epoch": 0.72, + "learning_rate": 3.7637020301513535e-05, + "loss": 0.6335, + "step": 24282 + }, + { + "epoch": 0.72, + "learning_rate": 3.762948293070863e-05, + "loss": 0.4675, + "step": 24283 + }, + { + "epoch": 0.72, + "learning_rate": 3.762194613979989e-05, + "loss": 0.5879, + "step": 24284 + }, + { + "epoch": 0.72, + "learning_rate": 3.761440992885748e-05, + "loss": 0.6342, + "step": 24285 + }, + { + "epoch": 0.72, + "learning_rate": 3.760687429795141e-05, + "loss": 0.5756, + "step": 24286 + }, + { + "epoch": 0.72, + "learning_rate": 3.759933924715173e-05, + "loss": 0.5906, + "step": 24287 + }, + { + "epoch": 0.72, + "learning_rate": 3.759180477652852e-05, + "loss": 0.5312, + "step": 24288 + }, + { + "epoch": 0.72, + "learning_rate": 3.758427088615183e-05, + "loss": 0.5438, + "step": 24289 + }, + { + "epoch": 0.72, + "learning_rate": 3.7576737576091756e-05, + "loss": 0.55, + "step": 24290 + }, + { + "epoch": 0.72, + "learning_rate": 3.7569204846418296e-05, + "loss": 0.6634, + "step": 24291 + }, + { + "epoch": 0.72, + "learning_rate": 3.7561672697201445e-05, + "loss": 0.5931, + "step": 24292 + }, + { + "epoch": 0.72, + "learning_rate": 3.7554141128511275e-05, + "loss": 0.5642, + "step": 24293 + }, + { + "epoch": 0.72, + "learning_rate": 3.754661014041786e-05, + "loss": 0.4464, + "step": 24294 + }, + { + "epoch": 0.72, + "learning_rate": 3.7539079732991166e-05, + "loss": 0.5877, + "step": 24295 + }, + { + "epoch": 0.72, + "learning_rate": 3.7531549906301176e-05, + "loss": 0.584, + "step": 24296 + }, + { + "epoch": 0.72, + "learning_rate": 3.752402066041794e-05, + "loss": 0.5926, + "step": 24297 + }, + { + "epoch": 0.72, + "learning_rate": 3.75164919954115e-05, + "loss": 0.4941, + "step": 24298 + }, + { + "epoch": 0.72, + "learning_rate": 3.7508963911351766e-05, + "loss": 0.4866, + "step": 24299 + }, + { + "epoch": 0.72, + "learning_rate": 3.750143640830882e-05, + "loss": 0.5506, + "step": 24300 + }, + { + "epoch": 0.72, + "learning_rate": 3.749390948635257e-05, + "loss": 0.5118, + "step": 24301 + }, + { + "epoch": 0.72, + "learning_rate": 3.7486383145553086e-05, + "loss": 0.5712, + "step": 24302 + }, + { + "epoch": 0.72, + "learning_rate": 3.747885738598025e-05, + "loss": 0.5094, + "step": 24303 + }, + { + "epoch": 0.72, + "learning_rate": 3.747133220770408e-05, + "loss": 0.6331, + "step": 24304 + }, + { + "epoch": 0.72, + "learning_rate": 3.7463807610794575e-05, + "loss": 0.512, + "step": 24305 + }, + { + "epoch": 0.72, + "learning_rate": 3.745628359532166e-05, + "loss": 0.4601, + "step": 24306 + }, + { + "epoch": 0.72, + "learning_rate": 3.744876016135527e-05, + "loss": 0.5173, + "step": 24307 + }, + { + "epoch": 0.72, + "learning_rate": 3.744123730896537e-05, + "loss": 0.6025, + "step": 24308 + }, + { + "epoch": 0.72, + "learning_rate": 3.743371503822195e-05, + "loss": 0.5788, + "step": 24309 + }, + { + "epoch": 0.72, + "learning_rate": 3.742619334919491e-05, + "loss": 0.5405, + "step": 24310 + }, + { + "epoch": 0.72, + "learning_rate": 3.7418672241954146e-05, + "loss": 0.5267, + "step": 24311 + }, + { + "epoch": 0.72, + "learning_rate": 3.7411151716569624e-05, + "loss": 0.5372, + "step": 24312 + }, + { + "epoch": 0.72, + "learning_rate": 3.740363177311127e-05, + "loss": 0.5646, + "step": 24313 + }, + { + "epoch": 0.72, + "learning_rate": 3.739611241164904e-05, + "loss": 0.5367, + "step": 24314 + }, + { + "epoch": 0.72, + "learning_rate": 3.73885936322528e-05, + "loss": 0.523, + "step": 24315 + }, + { + "epoch": 0.72, + "learning_rate": 3.738107543499243e-05, + "loss": 0.5356, + "step": 24316 + }, + { + "epoch": 0.72, + "learning_rate": 3.7373557819937854e-05, + "loss": 0.6113, + "step": 24317 + }, + { + "epoch": 0.72, + "learning_rate": 3.736604078715903e-05, + "loss": 0.6139, + "step": 24318 + }, + { + "epoch": 0.72, + "learning_rate": 3.7358524336725775e-05, + "loss": 0.5687, + "step": 24319 + }, + { + "epoch": 0.72, + "learning_rate": 3.735100846870797e-05, + "loss": 0.5228, + "step": 24320 + }, + { + "epoch": 0.72, + "learning_rate": 3.734349318317556e-05, + "loss": 0.4318, + "step": 24321 + }, + { + "epoch": 0.72, + "learning_rate": 3.733597848019833e-05, + "loss": 0.5721, + "step": 24322 + }, + { + "epoch": 0.72, + "learning_rate": 3.732846435984621e-05, + "loss": 0.5363, + "step": 24323 + }, + { + "epoch": 0.72, + "learning_rate": 3.732095082218907e-05, + "loss": 0.5308, + "step": 24324 + }, + { + "epoch": 0.72, + "learning_rate": 3.731343786729675e-05, + "loss": 0.5175, + "step": 24325 + }, + { + "epoch": 0.72, + "learning_rate": 3.730592549523907e-05, + "loss": 0.5562, + "step": 24326 + }, + { + "epoch": 0.72, + "learning_rate": 3.729841370608591e-05, + "loss": 0.4126, + "step": 24327 + }, + { + "epoch": 0.72, + "learning_rate": 3.729090249990714e-05, + "loss": 0.5561, + "step": 24328 + }, + { + "epoch": 0.72, + "learning_rate": 3.7283391876772564e-05, + "loss": 0.5439, + "step": 24329 + }, + { + "epoch": 0.72, + "learning_rate": 3.727588183675199e-05, + "loss": 0.6184, + "step": 24330 + }, + { + "epoch": 0.72, + "learning_rate": 3.7268372379915263e-05, + "loss": 0.5395, + "step": 24331 + }, + { + "epoch": 0.72, + "learning_rate": 3.72608635063322e-05, + "loss": 0.5348, + "step": 24332 + }, + { + "epoch": 0.72, + "learning_rate": 3.725335521607266e-05, + "loss": 0.5817, + "step": 24333 + }, + { + "epoch": 0.72, + "learning_rate": 3.724584750920641e-05, + "loss": 0.487, + "step": 24334 + }, + { + "epoch": 0.72, + "learning_rate": 3.723834038580324e-05, + "loss": 0.5975, + "step": 24335 + }, + { + "epoch": 0.72, + "learning_rate": 3.7230833845932956e-05, + "loss": 0.5134, + "step": 24336 + }, + { + "epoch": 0.72, + "learning_rate": 3.7223327889665397e-05, + "loss": 0.4507, + "step": 24337 + }, + { + "epoch": 0.72, + "learning_rate": 3.721582251707031e-05, + "loss": 0.5262, + "step": 24338 + }, + { + "epoch": 0.72, + "learning_rate": 3.7208317728217454e-05, + "loss": 0.5966, + "step": 24339 + }, + { + "epoch": 0.72, + "learning_rate": 3.720081352317662e-05, + "loss": 0.4884, + "step": 24340 + }, + { + "epoch": 0.72, + "learning_rate": 3.719330990201765e-05, + "loss": 0.581, + "step": 24341 + }, + { + "epoch": 0.72, + "learning_rate": 3.71858068648102e-05, + "loss": 0.5423, + "step": 24342 + }, + { + "epoch": 0.72, + "learning_rate": 3.717830441162413e-05, + "loss": 0.5761, + "step": 24343 + }, + { + "epoch": 0.72, + "learning_rate": 3.71708025425291e-05, + "loss": 0.5853, + "step": 24344 + }, + { + "epoch": 0.72, + "learning_rate": 3.716330125759495e-05, + "loss": 0.4987, + "step": 24345 + }, + { + "epoch": 0.72, + "learning_rate": 3.715580055689134e-05, + "loss": 0.5067, + "step": 24346 + }, + { + "epoch": 0.72, + "learning_rate": 3.714830044048806e-05, + "loss": 0.5614, + "step": 24347 + }, + { + "epoch": 0.72, + "learning_rate": 3.714080090845485e-05, + "loss": 0.6339, + "step": 24348 + }, + { + "epoch": 0.72, + "learning_rate": 3.713330196086143e-05, + "loss": 0.5107, + "step": 24349 + }, + { + "epoch": 0.72, + "learning_rate": 3.712580359777749e-05, + "loss": 0.5018, + "step": 24350 + }, + { + "epoch": 0.72, + "learning_rate": 3.711830581927275e-05, + "loss": 0.6533, + "step": 24351 + }, + { + "epoch": 0.72, + "learning_rate": 3.711080862541698e-05, + "loss": 0.5823, + "step": 24352 + }, + { + "epoch": 0.73, + "learning_rate": 3.7103312016279845e-05, + "loss": 0.4495, + "step": 24353 + }, + { + "epoch": 0.73, + "learning_rate": 3.709581599193102e-05, + "loss": 0.4541, + "step": 24354 + }, + { + "epoch": 0.73, + "learning_rate": 3.708832055244022e-05, + "loss": 0.4529, + "step": 24355 + }, + { + "epoch": 0.73, + "learning_rate": 3.7080825697877175e-05, + "loss": 0.4703, + "step": 24356 + }, + { + "epoch": 0.73, + "learning_rate": 3.707333142831151e-05, + "loss": 0.5464, + "step": 24357 + }, + { + "epoch": 0.73, + "learning_rate": 3.706583774381295e-05, + "loss": 0.5971, + "step": 24358 + }, + { + "epoch": 0.73, + "learning_rate": 3.705834464445112e-05, + "loss": 0.5962, + "step": 24359 + }, + { + "epoch": 0.73, + "learning_rate": 3.7050852130295754e-05, + "loss": 0.4989, + "step": 24360 + }, + { + "epoch": 0.73, + "learning_rate": 3.7043360201416434e-05, + "loss": 0.4577, + "step": 24361 + }, + { + "epoch": 0.73, + "learning_rate": 3.703586885788289e-05, + "loss": 0.6008, + "step": 24362 + }, + { + "epoch": 0.73, + "learning_rate": 3.702837809976472e-05, + "loss": 0.4405, + "step": 24363 + }, + { + "epoch": 0.73, + "learning_rate": 3.702088792713162e-05, + "loss": 0.5318, + "step": 24364 + }, + { + "epoch": 0.73, + "learning_rate": 3.7013398340053176e-05, + "loss": 0.4607, + "step": 24365 + }, + { + "epoch": 0.73, + "learning_rate": 3.7005909338599044e-05, + "loss": 0.551, + "step": 24366 + }, + { + "epoch": 0.73, + "learning_rate": 3.6998420922838895e-05, + "loss": 0.5038, + "step": 24367 + }, + { + "epoch": 0.73, + "learning_rate": 3.6990933092842314e-05, + "loss": 0.6846, + "step": 24368 + }, + { + "epoch": 0.73, + "learning_rate": 3.698344584867889e-05, + "loss": 0.4456, + "step": 24369 + }, + { + "epoch": 0.73, + "learning_rate": 3.697595919041827e-05, + "loss": 0.545, + "step": 24370 + }, + { + "epoch": 0.73, + "learning_rate": 3.696847311813011e-05, + "loss": 0.4789, + "step": 24371 + }, + { + "epoch": 0.73, + "learning_rate": 3.696098763188396e-05, + "loss": 0.583, + "step": 24372 + }, + { + "epoch": 0.73, + "learning_rate": 3.695350273174938e-05, + "loss": 0.5952, + "step": 24373 + }, + { + "epoch": 0.73, + "learning_rate": 3.6946018417796004e-05, + "loss": 0.6007, + "step": 24374 + }, + { + "epoch": 0.73, + "learning_rate": 3.693853469009343e-05, + "loss": 0.6308, + "step": 24375 + }, + { + "epoch": 0.73, + "learning_rate": 3.6931051548711245e-05, + "loss": 0.4823, + "step": 24376 + }, + { + "epoch": 0.73, + "learning_rate": 3.6923568993719015e-05, + "loss": 0.5524, + "step": 24377 + }, + { + "epoch": 0.73, + "learning_rate": 3.691608702518626e-05, + "loss": 0.526, + "step": 24378 + }, + { + "epoch": 0.73, + "learning_rate": 3.690860564318259e-05, + "loss": 0.5735, + "step": 24379 + }, + { + "epoch": 0.73, + "learning_rate": 3.69011248477776e-05, + "loss": 0.5777, + "step": 24380 + }, + { + "epoch": 0.73, + "learning_rate": 3.689364463904079e-05, + "loss": 0.478, + "step": 24381 + }, + { + "epoch": 0.73, + "learning_rate": 3.68861650170417e-05, + "loss": 0.6203, + "step": 24382 + }, + { + "epoch": 0.73, + "learning_rate": 3.687868598184988e-05, + "loss": 0.5436, + "step": 24383 + }, + { + "epoch": 0.73, + "learning_rate": 3.6871207533534926e-05, + "loss": 0.463, + "step": 24384 + }, + { + "epoch": 0.73, + "learning_rate": 3.686372967216628e-05, + "loss": 0.4493, + "step": 24385 + }, + { + "epoch": 0.73, + "learning_rate": 3.685625239781354e-05, + "loss": 0.5165, + "step": 24386 + }, + { + "epoch": 0.73, + "learning_rate": 3.6848775710546216e-05, + "loss": 0.5759, + "step": 24387 + }, + { + "epoch": 0.73, + "learning_rate": 3.6841299610433755e-05, + "loss": 0.5411, + "step": 24388 + }, + { + "epoch": 0.73, + "learning_rate": 3.6833824097545736e-05, + "loss": 0.6062, + "step": 24389 + }, + { + "epoch": 0.73, + "learning_rate": 3.6826349171951657e-05, + "loss": 0.5604, + "step": 24390 + }, + { + "epoch": 0.73, + "learning_rate": 3.6818874833721014e-05, + "loss": 0.4669, + "step": 24391 + }, + { + "epoch": 0.73, + "learning_rate": 3.6811401082923266e-05, + "loss": 0.4854, + "step": 24392 + }, + { + "epoch": 0.73, + "learning_rate": 3.680392791962792e-05, + "loss": 0.4268, + "step": 24393 + }, + { + "epoch": 0.73, + "learning_rate": 3.6796455343904466e-05, + "loss": 0.5041, + "step": 24394 + }, + { + "epoch": 0.73, + "learning_rate": 3.678898335582242e-05, + "loss": 0.5691, + "step": 24395 + }, + { + "epoch": 0.73, + "learning_rate": 3.6781511955451196e-05, + "loss": 0.6433, + "step": 24396 + }, + { + "epoch": 0.73, + "learning_rate": 3.677404114286026e-05, + "loss": 0.5833, + "step": 24397 + }, + { + "epoch": 0.73, + "learning_rate": 3.676657091811908e-05, + "loss": 0.4691, + "step": 24398 + }, + { + "epoch": 0.73, + "learning_rate": 3.6759101281297156e-05, + "loss": 0.5744, + "step": 24399 + }, + { + "epoch": 0.73, + "learning_rate": 3.6751632232463906e-05, + "loss": 0.4254, + "step": 24400 + }, + { + "epoch": 0.73, + "learning_rate": 3.674416377168874e-05, + "loss": 0.6016, + "step": 24401 + }, + { + "epoch": 0.73, + "learning_rate": 3.673669589904112e-05, + "loss": 0.5402, + "step": 24402 + }, + { + "epoch": 0.73, + "learning_rate": 3.672922861459053e-05, + "loss": 0.504, + "step": 24403 + }, + { + "epoch": 0.73, + "learning_rate": 3.672176191840632e-05, + "loss": 0.4696, + "step": 24404 + }, + { + "epoch": 0.73, + "learning_rate": 3.6714295810557975e-05, + "loss": 0.4771, + "step": 24405 + }, + { + "epoch": 0.73, + "learning_rate": 3.670683029111486e-05, + "loss": 0.54, + "step": 24406 + }, + { + "epoch": 0.73, + "learning_rate": 3.669936536014643e-05, + "loss": 0.5026, + "step": 24407 + }, + { + "epoch": 0.73, + "learning_rate": 3.669190101772205e-05, + "loss": 0.4989, + "step": 24408 + }, + { + "epoch": 0.73, + "learning_rate": 3.6684437263911145e-05, + "loss": 0.4818, + "step": 24409 + }, + { + "epoch": 0.73, + "learning_rate": 3.667697409878312e-05, + "loss": 0.6899, + "step": 24410 + }, + { + "epoch": 0.73, + "learning_rate": 3.6669511522407376e-05, + "loss": 0.4874, + "step": 24411 + }, + { + "epoch": 0.73, + "learning_rate": 3.6662049534853225e-05, + "loss": 0.5626, + "step": 24412 + }, + { + "epoch": 0.73, + "learning_rate": 3.6654588136190105e-05, + "loss": 0.519, + "step": 24413 + }, + { + "epoch": 0.73, + "learning_rate": 3.66471273264874e-05, + "loss": 0.5538, + "step": 24414 + }, + { + "epoch": 0.73, + "learning_rate": 3.6639667105814457e-05, + "loss": 0.5531, + "step": 24415 + }, + { + "epoch": 0.73, + "learning_rate": 3.663220747424061e-05, + "loss": 0.5569, + "step": 24416 + }, + { + "epoch": 0.73, + "learning_rate": 3.662474843183523e-05, + "loss": 0.5645, + "step": 24417 + }, + { + "epoch": 0.73, + "learning_rate": 3.6617289978667724e-05, + "loss": 0.5507, + "step": 24418 + }, + { + "epoch": 0.73, + "learning_rate": 3.6609832114807365e-05, + "loss": 0.6156, + "step": 24419 + }, + { + "epoch": 0.73, + "learning_rate": 3.660237484032354e-05, + "loss": 0.6388, + "step": 24420 + }, + { + "epoch": 0.73, + "learning_rate": 3.659491815528554e-05, + "loss": 0.3762, + "step": 24421 + }, + { + "epoch": 0.73, + "learning_rate": 3.658746205976275e-05, + "loss": 0.5477, + "step": 24422 + }, + { + "epoch": 0.73, + "learning_rate": 3.6580006553824444e-05, + "loss": 0.6138, + "step": 24423 + }, + { + "epoch": 0.73, + "learning_rate": 3.6572551637539984e-05, + "loss": 0.5237, + "step": 24424 + }, + { + "epoch": 0.73, + "learning_rate": 3.6565097310978624e-05, + "loss": 0.5954, + "step": 24425 + }, + { + "epoch": 0.73, + "learning_rate": 3.655764357420974e-05, + "loss": 0.6487, + "step": 24426 + }, + { + "epoch": 0.73, + "learning_rate": 3.655019042730257e-05, + "loss": 0.514, + "step": 24427 + }, + { + "epoch": 0.73, + "learning_rate": 3.654273787032644e-05, + "loss": 0.5058, + "step": 24428 + }, + { + "epoch": 0.73, + "learning_rate": 3.653528590335068e-05, + "loss": 0.4766, + "step": 24429 + }, + { + "epoch": 0.73, + "learning_rate": 3.652783452644453e-05, + "loss": 0.5519, + "step": 24430 + }, + { + "epoch": 0.73, + "learning_rate": 3.6520383739677244e-05, + "loss": 0.5532, + "step": 24431 + }, + { + "epoch": 0.73, + "learning_rate": 3.651293354311813e-05, + "loss": 0.4527, + "step": 24432 + }, + { + "epoch": 0.73, + "learning_rate": 3.650548393683649e-05, + "loss": 0.5708, + "step": 24433 + }, + { + "epoch": 0.73, + "learning_rate": 3.649803492090156e-05, + "loss": 0.5038, + "step": 24434 + }, + { + "epoch": 0.73, + "learning_rate": 3.649058649538255e-05, + "loss": 0.5105, + "step": 24435 + }, + { + "epoch": 0.73, + "learning_rate": 3.6483138660348756e-05, + "loss": 0.4708, + "step": 24436 + }, + { + "epoch": 0.73, + "learning_rate": 3.647569141586943e-05, + "loss": 0.614, + "step": 24437 + }, + { + "epoch": 0.73, + "learning_rate": 3.6468244762013836e-05, + "loss": 0.5734, + "step": 24438 + }, + { + "epoch": 0.73, + "learning_rate": 3.6460798698851184e-05, + "loss": 0.5216, + "step": 24439 + }, + { + "epoch": 0.73, + "learning_rate": 3.645335322645066e-05, + "loss": 0.5718, + "step": 24440 + }, + { + "epoch": 0.73, + "learning_rate": 3.644590834488154e-05, + "loss": 0.4498, + "step": 24441 + }, + { + "epoch": 0.73, + "learning_rate": 3.6438464054213065e-05, + "loss": 0.5182, + "step": 24442 + }, + { + "epoch": 0.73, + "learning_rate": 3.643102035451441e-05, + "loss": 0.5495, + "step": 24443 + }, + { + "epoch": 0.73, + "learning_rate": 3.642357724585477e-05, + "loss": 0.5991, + "step": 24444 + }, + { + "epoch": 0.73, + "learning_rate": 3.6416134728303355e-05, + "loss": 0.5445, + "step": 24445 + }, + { + "epoch": 0.73, + "learning_rate": 3.640869280192942e-05, + "loss": 0.5329, + "step": 24446 + }, + { + "epoch": 0.73, + "learning_rate": 3.6401251466802086e-05, + "loss": 0.4647, + "step": 24447 + }, + { + "epoch": 0.73, + "learning_rate": 3.6393810722990594e-05, + "loss": 0.659, + "step": 24448 + }, + { + "epoch": 0.73, + "learning_rate": 3.638637057056407e-05, + "loss": 0.4478, + "step": 24449 + }, + { + "epoch": 0.73, + "learning_rate": 3.637893100959174e-05, + "loss": 0.5405, + "step": 24450 + }, + { + "epoch": 0.73, + "learning_rate": 3.637149204014273e-05, + "loss": 0.5485, + "step": 24451 + }, + { + "epoch": 0.73, + "learning_rate": 3.6364053662286237e-05, + "loss": 0.4586, + "step": 24452 + }, + { + "epoch": 0.73, + "learning_rate": 3.635661587609146e-05, + "loss": 0.5522, + "step": 24453 + }, + { + "epoch": 0.73, + "learning_rate": 3.634917868162744e-05, + "loss": 0.5359, + "step": 24454 + }, + { + "epoch": 0.73, + "learning_rate": 3.634174207896339e-05, + "loss": 0.565, + "step": 24455 + }, + { + "epoch": 0.73, + "learning_rate": 3.633430606816846e-05, + "loss": 0.5511, + "step": 24456 + }, + { + "epoch": 0.73, + "learning_rate": 3.632687064931179e-05, + "loss": 0.5747, + "step": 24457 + }, + { + "epoch": 0.73, + "learning_rate": 3.631943582246251e-05, + "loss": 0.6413, + "step": 24458 + }, + { + "epoch": 0.73, + "learning_rate": 3.6312001587689704e-05, + "loss": 0.5307, + "step": 24459 + }, + { + "epoch": 0.73, + "learning_rate": 3.630456794506252e-05, + "loss": 0.6012, + "step": 24460 + }, + { + "epoch": 0.73, + "learning_rate": 3.629713489465011e-05, + "loss": 0.5231, + "step": 24461 + }, + { + "epoch": 0.73, + "learning_rate": 3.628970243652156e-05, + "loss": 0.4695, + "step": 24462 + }, + { + "epoch": 0.73, + "learning_rate": 3.6282270570745926e-05, + "loss": 0.4989, + "step": 24463 + }, + { + "epoch": 0.73, + "learning_rate": 3.627483929739234e-05, + "loss": 0.6296, + "step": 24464 + }, + { + "epoch": 0.73, + "learning_rate": 3.626740861652993e-05, + "loss": 0.5535, + "step": 24465 + }, + { + "epoch": 0.73, + "learning_rate": 3.6259978528227725e-05, + "loss": 0.6409, + "step": 24466 + }, + { + "epoch": 0.73, + "learning_rate": 3.625254903255487e-05, + "loss": 0.5307, + "step": 24467 + }, + { + "epoch": 0.73, + "learning_rate": 3.6245120129580365e-05, + "loss": 0.524, + "step": 24468 + }, + { + "epoch": 0.73, + "learning_rate": 3.6237691819373373e-05, + "loss": 0.6192, + "step": 24469 + }, + { + "epoch": 0.73, + "learning_rate": 3.6230264102002864e-05, + "loss": 0.5363, + "step": 24470 + }, + { + "epoch": 0.73, + "learning_rate": 3.622283697753794e-05, + "loss": 0.5366, + "step": 24471 + }, + { + "epoch": 0.73, + "learning_rate": 3.6215410446047705e-05, + "loss": 0.5816, + "step": 24472 + }, + { + "epoch": 0.73, + "learning_rate": 3.6207984507601156e-05, + "loss": 0.5456, + "step": 24473 + }, + { + "epoch": 0.73, + "learning_rate": 3.62005591622673e-05, + "loss": 0.5798, + "step": 24474 + }, + { + "epoch": 0.73, + "learning_rate": 3.6193134410115224e-05, + "loss": 0.5626, + "step": 24475 + }, + { + "epoch": 0.73, + "learning_rate": 3.618571025121399e-05, + "loss": 0.6079, + "step": 24476 + }, + { + "epoch": 0.73, + "learning_rate": 3.617828668563259e-05, + "loss": 0.5043, + "step": 24477 + }, + { + "epoch": 0.73, + "learning_rate": 3.6170863713440004e-05, + "loss": 0.4936, + "step": 24478 + }, + { + "epoch": 0.73, + "learning_rate": 3.6163441334705285e-05, + "loss": 0.4193, + "step": 24479 + }, + { + "epoch": 0.73, + "learning_rate": 3.6156019549497455e-05, + "loss": 0.4671, + "step": 24480 + }, + { + "epoch": 0.73, + "learning_rate": 3.614859835788553e-05, + "loss": 0.4817, + "step": 24481 + }, + { + "epoch": 0.73, + "learning_rate": 3.61411777599385e-05, + "loss": 0.5198, + "step": 24482 + }, + { + "epoch": 0.73, + "learning_rate": 3.613375775572531e-05, + "loss": 0.5276, + "step": 24483 + }, + { + "epoch": 0.73, + "learning_rate": 3.612633834531503e-05, + "loss": 0.591, + "step": 24484 + }, + { + "epoch": 0.73, + "learning_rate": 3.611891952877654e-05, + "loss": 0.4725, + "step": 24485 + }, + { + "epoch": 0.73, + "learning_rate": 3.611150130617893e-05, + "loss": 0.519, + "step": 24486 + }, + { + "epoch": 0.73, + "learning_rate": 3.610408367759107e-05, + "loss": 0.5604, + "step": 24487 + }, + { + "epoch": 0.73, + "learning_rate": 3.609666664308202e-05, + "loss": 0.5163, + "step": 24488 + }, + { + "epoch": 0.73, + "learning_rate": 3.6089250202720656e-05, + "loss": 0.5669, + "step": 24489 + }, + { + "epoch": 0.73, + "learning_rate": 3.608183435657597e-05, + "loss": 0.6343, + "step": 24490 + }, + { + "epoch": 0.73, + "learning_rate": 3.607441910471695e-05, + "loss": 0.5062, + "step": 24491 + }, + { + "epoch": 0.73, + "learning_rate": 3.60670044472125e-05, + "loss": 0.5495, + "step": 24492 + }, + { + "epoch": 0.73, + "learning_rate": 3.6059590384131524e-05, + "loss": 0.4757, + "step": 24493 + }, + { + "epoch": 0.73, + "learning_rate": 3.6052176915543e-05, + "loss": 0.4408, + "step": 24494 + }, + { + "epoch": 0.73, + "learning_rate": 3.6044764041515876e-05, + "loss": 0.4878, + "step": 24495 + }, + { + "epoch": 0.73, + "learning_rate": 3.6037351762119054e-05, + "loss": 0.6994, + "step": 24496 + }, + { + "epoch": 0.73, + "learning_rate": 3.6029940077421395e-05, + "loss": 0.4646, + "step": 24497 + }, + { + "epoch": 0.73, + "learning_rate": 3.6022528987491867e-05, + "loss": 0.5788, + "step": 24498 + }, + { + "epoch": 0.73, + "learning_rate": 3.601511849239937e-05, + "loss": 0.646, + "step": 24499 + }, + { + "epoch": 0.73, + "learning_rate": 3.600770859221282e-05, + "loss": 0.4998, + "step": 24500 + }, + { + "epoch": 0.73, + "learning_rate": 3.600029928700111e-05, + "loss": 0.5731, + "step": 24501 + }, + { + "epoch": 0.73, + "learning_rate": 3.599289057683305e-05, + "loss": 0.4974, + "step": 24502 + }, + { + "epoch": 0.73, + "learning_rate": 3.5985482461777596e-05, + "loss": 0.637, + "step": 24503 + }, + { + "epoch": 0.73, + "learning_rate": 3.597807494190365e-05, + "loss": 0.6448, + "step": 24504 + }, + { + "epoch": 0.73, + "learning_rate": 3.597066801728004e-05, + "loss": 0.5321, + "step": 24505 + }, + { + "epoch": 0.73, + "learning_rate": 3.596326168797561e-05, + "loss": 0.5526, + "step": 24506 + }, + { + "epoch": 0.73, + "learning_rate": 3.595585595405925e-05, + "loss": 0.5576, + "step": 24507 + }, + { + "epoch": 0.73, + "learning_rate": 3.5948450815599846e-05, + "loss": 0.6229, + "step": 24508 + }, + { + "epoch": 0.73, + "learning_rate": 3.594104627266619e-05, + "loss": 0.4744, + "step": 24509 + }, + { + "epoch": 0.73, + "learning_rate": 3.593364232532718e-05, + "loss": 0.4455, + "step": 24510 + }, + { + "epoch": 0.73, + "learning_rate": 3.5926238973651605e-05, + "loss": 0.5607, + "step": 24511 + }, + { + "epoch": 0.73, + "learning_rate": 3.591883621770835e-05, + "loss": 0.5847, + "step": 24512 + }, + { + "epoch": 0.73, + "learning_rate": 3.5911434057566185e-05, + "loss": 0.4453, + "step": 24513 + }, + { + "epoch": 0.73, + "learning_rate": 3.590403249329396e-05, + "loss": 0.4989, + "step": 24514 + }, + { + "epoch": 0.73, + "learning_rate": 3.589663152496056e-05, + "loss": 0.5409, + "step": 24515 + }, + { + "epoch": 0.73, + "learning_rate": 3.588923115263466e-05, + "loss": 0.5533, + "step": 24516 + }, + { + "epoch": 0.73, + "learning_rate": 3.588183137638515e-05, + "loss": 0.3827, + "step": 24517 + }, + { + "epoch": 0.73, + "learning_rate": 3.587443219628081e-05, + "loss": 0.3963, + "step": 24518 + }, + { + "epoch": 0.73, + "learning_rate": 3.586703361239048e-05, + "loss": 0.3793, + "step": 24519 + }, + { + "epoch": 0.73, + "learning_rate": 3.585963562478291e-05, + "loss": 0.5874, + "step": 24520 + }, + { + "epoch": 0.73, + "learning_rate": 3.5852238233526834e-05, + "loss": 0.5268, + "step": 24521 + }, + { + "epoch": 0.73, + "learning_rate": 3.58448414386911e-05, + "loss": 0.4368, + "step": 24522 + }, + { + "epoch": 0.73, + "learning_rate": 3.583744524034448e-05, + "loss": 0.5786, + "step": 24523 + }, + { + "epoch": 0.73, + "learning_rate": 3.58300496385557e-05, + "loss": 0.4995, + "step": 24524 + }, + { + "epoch": 0.73, + "learning_rate": 3.582265463339357e-05, + "loss": 0.479, + "step": 24525 + }, + { + "epoch": 0.73, + "learning_rate": 3.581526022492678e-05, + "loss": 0.5891, + "step": 24526 + }, + { + "epoch": 0.73, + "learning_rate": 3.580786641322417e-05, + "loss": 0.518, + "step": 24527 + }, + { + "epoch": 0.73, + "learning_rate": 3.580047319835439e-05, + "loss": 0.538, + "step": 24528 + }, + { + "epoch": 0.73, + "learning_rate": 3.579308058038625e-05, + "loss": 0.5144, + "step": 24529 + }, + { + "epoch": 0.73, + "learning_rate": 3.578568855938843e-05, + "loss": 0.5556, + "step": 24530 + }, + { + "epoch": 0.73, + "learning_rate": 3.5778297135429714e-05, + "loss": 0.5552, + "step": 24531 + }, + { + "epoch": 0.73, + "learning_rate": 3.577090630857877e-05, + "loss": 0.6037, + "step": 24532 + }, + { + "epoch": 0.73, + "learning_rate": 3.5763516078904325e-05, + "loss": 0.5204, + "step": 24533 + }, + { + "epoch": 0.73, + "learning_rate": 3.5756126446475145e-05, + "loss": 0.5686, + "step": 24534 + }, + { + "epoch": 0.73, + "learning_rate": 3.57487374113599e-05, + "loss": 0.5117, + "step": 24535 + }, + { + "epoch": 0.73, + "learning_rate": 3.574134897362725e-05, + "loss": 0.5446, + "step": 24536 + }, + { + "epoch": 0.73, + "learning_rate": 3.573396113334594e-05, + "loss": 0.5572, + "step": 24537 + }, + { + "epoch": 0.73, + "learning_rate": 3.572657389058466e-05, + "loss": 0.701, + "step": 24538 + }, + { + "epoch": 0.73, + "learning_rate": 3.571918724541209e-05, + "loss": 0.586, + "step": 24539 + }, + { + "epoch": 0.73, + "learning_rate": 3.5711801197896856e-05, + "loss": 0.5245, + "step": 24540 + }, + { + "epoch": 0.73, + "learning_rate": 3.5704415748107676e-05, + "loss": 0.5833, + "step": 24541 + }, + { + "epoch": 0.73, + "learning_rate": 3.569703089611322e-05, + "loss": 0.5472, + "step": 24542 + }, + { + "epoch": 0.73, + "learning_rate": 3.5689646641982165e-05, + "loss": 0.5849, + "step": 24543 + }, + { + "epoch": 0.73, + "learning_rate": 3.568226298578314e-05, + "loss": 0.5075, + "step": 24544 + }, + { + "epoch": 0.73, + "learning_rate": 3.567487992758477e-05, + "loss": 0.6111, + "step": 24545 + }, + { + "epoch": 0.73, + "learning_rate": 3.5667497467455734e-05, + "loss": 0.4705, + "step": 24546 + }, + { + "epoch": 0.73, + "learning_rate": 3.5660115605464696e-05, + "loss": 0.5327, + "step": 24547 + }, + { + "epoch": 0.73, + "learning_rate": 3.565273434168026e-05, + "loss": 0.482, + "step": 24548 + }, + { + "epoch": 0.73, + "learning_rate": 3.564535367617102e-05, + "loss": 0.4834, + "step": 24549 + }, + { + "epoch": 0.73, + "learning_rate": 3.563797360900567e-05, + "loss": 0.581, + "step": 24550 + }, + { + "epoch": 0.73, + "learning_rate": 3.563059414025276e-05, + "loss": 0.507, + "step": 24551 + }, + { + "epoch": 0.73, + "learning_rate": 3.562321526998091e-05, + "loss": 0.5016, + "step": 24552 + }, + { + "epoch": 0.73, + "learning_rate": 3.56158369982588e-05, + "loss": 0.5374, + "step": 24553 + }, + { + "epoch": 0.73, + "learning_rate": 3.560845932515497e-05, + "loss": 0.4893, + "step": 24554 + }, + { + "epoch": 0.73, + "learning_rate": 3.560108225073799e-05, + "loss": 0.5488, + "step": 24555 + }, + { + "epoch": 0.73, + "learning_rate": 3.559370577507648e-05, + "loss": 0.5311, + "step": 24556 + }, + { + "epoch": 0.73, + "learning_rate": 3.558632989823906e-05, + "loss": 0.5458, + "step": 24557 + }, + { + "epoch": 0.73, + "learning_rate": 3.5578954620294255e-05, + "loss": 0.4946, + "step": 24558 + }, + { + "epoch": 0.73, + "learning_rate": 3.557157994131064e-05, + "loss": 0.4337, + "step": 24559 + }, + { + "epoch": 0.73, + "learning_rate": 3.5564205861356784e-05, + "loss": 0.4722, + "step": 24560 + }, + { + "epoch": 0.73, + "learning_rate": 3.555683238050127e-05, + "loss": 0.5376, + "step": 24561 + }, + { + "epoch": 0.73, + "learning_rate": 3.5549459498812665e-05, + "loss": 0.5394, + "step": 24562 + }, + { + "epoch": 0.73, + "learning_rate": 3.554208721635951e-05, + "loss": 0.4915, + "step": 24563 + }, + { + "epoch": 0.73, + "learning_rate": 3.5534715533210295e-05, + "loss": 0.6407, + "step": 24564 + }, + { + "epoch": 0.73, + "learning_rate": 3.552734444943361e-05, + "loss": 0.4699, + "step": 24565 + }, + { + "epoch": 0.73, + "learning_rate": 3.551997396509801e-05, + "loss": 0.6006, + "step": 24566 + }, + { + "epoch": 0.73, + "learning_rate": 3.551260408027199e-05, + "loss": 0.5093, + "step": 24567 + }, + { + "epoch": 0.73, + "learning_rate": 3.550523479502405e-05, + "loss": 0.5813, + "step": 24568 + }, + { + "epoch": 0.73, + "learning_rate": 3.549786610942273e-05, + "loss": 0.5337, + "step": 24569 + }, + { + "epoch": 0.73, + "learning_rate": 3.549049802353657e-05, + "loss": 0.4695, + "step": 24570 + }, + { + "epoch": 0.73, + "learning_rate": 3.548313053743402e-05, + "loss": 0.4958, + "step": 24571 + }, + { + "epoch": 0.73, + "learning_rate": 3.547576365118364e-05, + "loss": 0.4952, + "step": 24572 + }, + { + "epoch": 0.73, + "learning_rate": 3.546839736485386e-05, + "loss": 0.4738, + "step": 24573 + }, + { + "epoch": 0.73, + "learning_rate": 3.546103167851323e-05, + "loss": 0.4841, + "step": 24574 + }, + { + "epoch": 0.73, + "learning_rate": 3.545366659223017e-05, + "loss": 0.5373, + "step": 24575 + }, + { + "epoch": 0.73, + "learning_rate": 3.54463021060732e-05, + "loss": 0.5595, + "step": 24576 + }, + { + "epoch": 0.73, + "learning_rate": 3.5438938220110816e-05, + "loss": 0.5658, + "step": 24577 + }, + { + "epoch": 0.73, + "learning_rate": 3.543157493441145e-05, + "loss": 0.5952, + "step": 24578 + }, + { + "epoch": 0.73, + "learning_rate": 3.542421224904354e-05, + "loss": 0.4896, + "step": 24579 + }, + { + "epoch": 0.73, + "learning_rate": 3.541685016407557e-05, + "loss": 0.5678, + "step": 24580 + }, + { + "epoch": 0.73, + "learning_rate": 3.540948867957601e-05, + "loss": 0.5914, + "step": 24581 + }, + { + "epoch": 0.73, + "learning_rate": 3.540212779561327e-05, + "loss": 0.5286, + "step": 24582 + }, + { + "epoch": 0.73, + "learning_rate": 3.539476751225579e-05, + "loss": 0.5501, + "step": 24583 + }, + { + "epoch": 0.73, + "learning_rate": 3.538740782957199e-05, + "loss": 0.5433, + "step": 24584 + }, + { + "epoch": 0.73, + "learning_rate": 3.538004874763038e-05, + "loss": 0.5281, + "step": 24585 + }, + { + "epoch": 0.73, + "learning_rate": 3.5372690266499266e-05, + "loss": 0.4825, + "step": 24586 + }, + { + "epoch": 0.73, + "learning_rate": 3.5365332386247165e-05, + "loss": 0.4372, + "step": 24587 + }, + { + "epoch": 0.73, + "learning_rate": 3.5357975106942396e-05, + "loss": 0.5176, + "step": 24588 + }, + { + "epoch": 0.73, + "learning_rate": 3.535061842865346e-05, + "loss": 0.5697, + "step": 24589 + }, + { + "epoch": 0.73, + "learning_rate": 3.534326235144867e-05, + "loss": 0.5073, + "step": 24590 + }, + { + "epoch": 0.73, + "learning_rate": 3.533590687539649e-05, + "loss": 0.5854, + "step": 24591 + }, + { + "epoch": 0.73, + "learning_rate": 3.532855200056524e-05, + "loss": 0.6214, + "step": 24592 + }, + { + "epoch": 0.73, + "learning_rate": 3.532119772702337e-05, + "loss": 0.5494, + "step": 24593 + }, + { + "epoch": 0.73, + "learning_rate": 3.5313844054839184e-05, + "loss": 0.4701, + "step": 24594 + }, + { + "epoch": 0.73, + "learning_rate": 3.530649098408111e-05, + "loss": 0.5072, + "step": 24595 + }, + { + "epoch": 0.73, + "learning_rate": 3.529913851481752e-05, + "loss": 0.6251, + "step": 24596 + }, + { + "epoch": 0.73, + "learning_rate": 3.529178664711677e-05, + "loss": 0.5641, + "step": 24597 + }, + { + "epoch": 0.73, + "learning_rate": 3.528443538104714e-05, + "loss": 0.5941, + "step": 24598 + }, + { + "epoch": 0.73, + "learning_rate": 3.5277084716677054e-05, + "loss": 0.5718, + "step": 24599 + }, + { + "epoch": 0.73, + "learning_rate": 3.526973465407487e-05, + "loss": 0.548, + "step": 24600 + }, + { + "epoch": 0.73, + "learning_rate": 3.526238519330889e-05, + "loss": 0.4998, + "step": 24601 + }, + { + "epoch": 0.73, + "learning_rate": 3.5255036334447414e-05, + "loss": 0.5432, + "step": 24602 + }, + { + "epoch": 0.73, + "learning_rate": 3.5247688077558814e-05, + "loss": 0.4906, + "step": 24603 + }, + { + "epoch": 0.73, + "learning_rate": 3.524034042271139e-05, + "loss": 0.4777, + "step": 24604 + }, + { + "epoch": 0.73, + "learning_rate": 3.523299336997352e-05, + "loss": 0.5116, + "step": 24605 + }, + { + "epoch": 0.73, + "learning_rate": 3.522564691941346e-05, + "loss": 0.4808, + "step": 24606 + }, + { + "epoch": 0.73, + "learning_rate": 3.521830107109947e-05, + "loss": 0.5761, + "step": 24607 + }, + { + "epoch": 0.73, + "learning_rate": 3.521095582509991e-05, + "loss": 0.4652, + "step": 24608 + }, + { + "epoch": 0.73, + "learning_rate": 3.520361118148309e-05, + "loss": 0.6181, + "step": 24609 + }, + { + "epoch": 0.73, + "learning_rate": 3.519626714031727e-05, + "loss": 0.5057, + "step": 24610 + }, + { + "epoch": 0.73, + "learning_rate": 3.5188923701670697e-05, + "loss": 0.6163, + "step": 24611 + }, + { + "epoch": 0.73, + "learning_rate": 3.518158086561169e-05, + "loss": 0.4943, + "step": 24612 + }, + { + "epoch": 0.73, + "learning_rate": 3.517423863220852e-05, + "loss": 0.5171, + "step": 24613 + }, + { + "epoch": 0.73, + "learning_rate": 3.516689700152943e-05, + "loss": 0.4862, + "step": 24614 + }, + { + "epoch": 0.73, + "learning_rate": 3.515955597364271e-05, + "loss": 0.5066, + "step": 24615 + }, + { + "epoch": 0.73, + "learning_rate": 3.5152215548616606e-05, + "loss": 0.5553, + "step": 24616 + }, + { + "epoch": 0.73, + "learning_rate": 3.514487572651932e-05, + "loss": 0.5249, + "step": 24617 + }, + { + "epoch": 0.73, + "learning_rate": 3.513753650741913e-05, + "loss": 0.5109, + "step": 24618 + }, + { + "epoch": 0.73, + "learning_rate": 3.513019789138427e-05, + "loss": 0.4905, + "step": 24619 + }, + { + "epoch": 0.73, + "learning_rate": 3.512285987848305e-05, + "loss": 0.5277, + "step": 24620 + }, + { + "epoch": 0.73, + "learning_rate": 3.511552246878355e-05, + "loss": 0.5401, + "step": 24621 + }, + { + "epoch": 0.73, + "learning_rate": 3.5108185662354065e-05, + "loss": 0.5733, + "step": 24622 + }, + { + "epoch": 0.73, + "learning_rate": 3.51008494592628e-05, + "loss": 0.6321, + "step": 24623 + }, + { + "epoch": 0.73, + "learning_rate": 3.509351385957801e-05, + "loss": 0.51, + "step": 24624 + }, + { + "epoch": 0.73, + "learning_rate": 3.508617886336786e-05, + "loss": 0.5354, + "step": 24625 + }, + { + "epoch": 0.73, + "learning_rate": 3.50788444707005e-05, + "loss": 0.4518, + "step": 24626 + }, + { + "epoch": 0.73, + "learning_rate": 3.507151068164418e-05, + "loss": 0.5379, + "step": 24627 + }, + { + "epoch": 0.73, + "learning_rate": 3.5064177496267116e-05, + "loss": 0.5072, + "step": 24628 + }, + { + "epoch": 0.73, + "learning_rate": 3.50568449146374e-05, + "loss": 0.5837, + "step": 24629 + }, + { + "epoch": 0.73, + "learning_rate": 3.50495129368233e-05, + "loss": 0.635, + "step": 24630 + }, + { + "epoch": 0.73, + "learning_rate": 3.504218156289291e-05, + "loss": 0.4481, + "step": 24631 + }, + { + "epoch": 0.73, + "learning_rate": 3.5034850792914456e-05, + "loss": 0.5323, + "step": 24632 + }, + { + "epoch": 0.73, + "learning_rate": 3.502752062695603e-05, + "loss": 0.5564, + "step": 24633 + }, + { + "epoch": 0.73, + "learning_rate": 3.502019106508586e-05, + "loss": 0.5232, + "step": 24634 + }, + { + "epoch": 0.73, + "learning_rate": 3.501286210737201e-05, + "loss": 0.5438, + "step": 24635 + }, + { + "epoch": 0.73, + "learning_rate": 3.5005533753882714e-05, + "loss": 0.5566, + "step": 24636 + }, + { + "epoch": 0.73, + "learning_rate": 3.4998206004686026e-05, + "loss": 0.5812, + "step": 24637 + }, + { + "epoch": 0.73, + "learning_rate": 3.4990878859850105e-05, + "loss": 0.5204, + "step": 24638 + }, + { + "epoch": 0.73, + "learning_rate": 3.4983552319443125e-05, + "loss": 0.5404, + "step": 24639 + }, + { + "epoch": 0.73, + "learning_rate": 3.497622638353316e-05, + "loss": 0.6129, + "step": 24640 + }, + { + "epoch": 0.73, + "learning_rate": 3.496890105218829e-05, + "loss": 0.5639, + "step": 24641 + }, + { + "epoch": 0.73, + "learning_rate": 3.4961576325476654e-05, + "loss": 0.5659, + "step": 24642 + }, + { + "epoch": 0.73, + "learning_rate": 3.4954252203466395e-05, + "loss": 0.5778, + "step": 24643 + }, + { + "epoch": 0.73, + "learning_rate": 3.494692868622558e-05, + "loss": 0.5481, + "step": 24644 + }, + { + "epoch": 0.73, + "learning_rate": 3.493960577382226e-05, + "loss": 0.5537, + "step": 24645 + }, + { + "epoch": 0.73, + "learning_rate": 3.4932283466324554e-05, + "loss": 0.4587, + "step": 24646 + }, + { + "epoch": 0.73, + "learning_rate": 3.492496176380058e-05, + "loss": 0.5864, + "step": 24647 + }, + { + "epoch": 0.73, + "learning_rate": 3.491764066631833e-05, + "loss": 0.6049, + "step": 24648 + }, + { + "epoch": 0.73, + "learning_rate": 3.4910320173945966e-05, + "loss": 0.5292, + "step": 24649 + }, + { + "epoch": 0.73, + "learning_rate": 3.490300028675145e-05, + "loss": 0.3852, + "step": 24650 + }, + { + "epoch": 0.73, + "learning_rate": 3.489568100480295e-05, + "loss": 0.4746, + "step": 24651 + }, + { + "epoch": 0.73, + "learning_rate": 3.488836232816841e-05, + "loss": 0.6275, + "step": 24652 + }, + { + "epoch": 0.73, + "learning_rate": 3.488104425691596e-05, + "loss": 0.4916, + "step": 24653 + }, + { + "epoch": 0.73, + "learning_rate": 3.487372679111358e-05, + "loss": 0.5141, + "step": 24654 + }, + { + "epoch": 0.73, + "learning_rate": 3.486640993082936e-05, + "loss": 0.5904, + "step": 24655 + }, + { + "epoch": 0.73, + "learning_rate": 3.485909367613127e-05, + "loss": 0.5251, + "step": 24656 + }, + { + "epoch": 0.73, + "learning_rate": 3.4851778027087365e-05, + "loss": 0.5229, + "step": 24657 + }, + { + "epoch": 0.73, + "learning_rate": 3.484446298376569e-05, + "loss": 0.5004, + "step": 24658 + }, + { + "epoch": 0.73, + "learning_rate": 3.4837148546234245e-05, + "loss": 0.4482, + "step": 24659 + }, + { + "epoch": 0.73, + "learning_rate": 3.4829834714560975e-05, + "loss": 0.6366, + "step": 24660 + }, + { + "epoch": 0.73, + "learning_rate": 3.482252148881393e-05, + "loss": 0.6001, + "step": 24661 + }, + { + "epoch": 0.73, + "learning_rate": 3.481520886906113e-05, + "loss": 0.454, + "step": 24662 + }, + { + "epoch": 0.73, + "learning_rate": 3.480789685537055e-05, + "loss": 0.5071, + "step": 24663 + }, + { + "epoch": 0.73, + "learning_rate": 3.4800585447810116e-05, + "loss": 0.5151, + "step": 24664 + }, + { + "epoch": 0.73, + "learning_rate": 3.479327464644786e-05, + "loss": 0.4684, + "step": 24665 + }, + { + "epoch": 0.73, + "learning_rate": 3.478596445135175e-05, + "loss": 0.4724, + "step": 24666 + }, + { + "epoch": 0.73, + "learning_rate": 3.477865486258978e-05, + "loss": 0.4759, + "step": 24667 + }, + { + "epoch": 0.73, + "learning_rate": 3.477134588022989e-05, + "loss": 0.4369, + "step": 24668 + }, + { + "epoch": 0.73, + "learning_rate": 3.476403750433999e-05, + "loss": 0.6587, + "step": 24669 + }, + { + "epoch": 0.73, + "learning_rate": 3.475672973498806e-05, + "loss": 0.5781, + "step": 24670 + }, + { + "epoch": 0.73, + "learning_rate": 3.47494225722421e-05, + "loss": 0.6985, + "step": 24671 + }, + { + "epoch": 0.73, + "learning_rate": 3.474211601616999e-05, + "loss": 0.5072, + "step": 24672 + }, + { + "epoch": 0.73, + "learning_rate": 3.473481006683965e-05, + "loss": 0.6143, + "step": 24673 + }, + { + "epoch": 0.73, + "learning_rate": 3.4727504724319036e-05, + "loss": 0.4468, + "step": 24674 + }, + { + "epoch": 0.73, + "learning_rate": 3.472019998867609e-05, + "loss": 0.5185, + "step": 24675 + }, + { + "epoch": 0.73, + "learning_rate": 3.471289585997869e-05, + "loss": 0.484, + "step": 24676 + }, + { + "epoch": 0.73, + "learning_rate": 3.470559233829479e-05, + "loss": 0.6284, + "step": 24677 + }, + { + "epoch": 0.73, + "learning_rate": 3.469828942369223e-05, + "loss": 0.4883, + "step": 24678 + }, + { + "epoch": 0.73, + "learning_rate": 3.469098711623899e-05, + "loss": 0.5708, + "step": 24679 + }, + { + "epoch": 0.73, + "learning_rate": 3.4683685416002885e-05, + "loss": 0.471, + "step": 24680 + }, + { + "epoch": 0.73, + "learning_rate": 3.4676384323051836e-05, + "loss": 0.5358, + "step": 24681 + }, + { + "epoch": 0.73, + "learning_rate": 3.466908383745381e-05, + "loss": 0.4921, + "step": 24682 + }, + { + "epoch": 0.73, + "learning_rate": 3.4661783959276516e-05, + "loss": 0.4511, + "step": 24683 + }, + { + "epoch": 0.73, + "learning_rate": 3.465448468858793e-05, + "loss": 0.5417, + "step": 24684 + }, + { + "epoch": 0.73, + "learning_rate": 3.46471860254559e-05, + "loss": 0.5907, + "step": 24685 + }, + { + "epoch": 0.73, + "learning_rate": 3.4639887969948315e-05, + "loss": 0.6192, + "step": 24686 + }, + { + "epoch": 0.73, + "learning_rate": 3.4632590522133e-05, + "loss": 0.4952, + "step": 24687 + }, + { + "epoch": 0.74, + "learning_rate": 3.462529368207778e-05, + "loss": 0.3333, + "step": 24688 + }, + { + "epoch": 0.74, + "learning_rate": 3.461799744985052e-05, + "loss": 0.4556, + "step": 24689 + }, + { + "epoch": 0.74, + "learning_rate": 3.4610701825519096e-05, + "loss": 0.4859, + "step": 24690 + }, + { + "epoch": 0.74, + "learning_rate": 3.4603406809151274e-05, + "loss": 0.6509, + "step": 24691 + }, + { + "epoch": 0.74, + "learning_rate": 3.459611240081494e-05, + "loss": 0.5763, + "step": 24692 + }, + { + "epoch": 0.74, + "learning_rate": 3.458881860057784e-05, + "loss": 0.533, + "step": 24693 + }, + { + "epoch": 0.74, + "learning_rate": 3.4581525408507886e-05, + "loss": 0.5924, + "step": 24694 + }, + { + "epoch": 0.74, + "learning_rate": 3.45742328246728e-05, + "loss": 0.4475, + "step": 24695 + }, + { + "epoch": 0.74, + "learning_rate": 3.456694084914045e-05, + "loss": 0.6125, + "step": 24696 + }, + { + "epoch": 0.74, + "learning_rate": 3.4559649481978573e-05, + "loss": 0.5683, + "step": 24697 + }, + { + "epoch": 0.74, + "learning_rate": 3.455235872325503e-05, + "loss": 0.504, + "step": 24698 + }, + { + "epoch": 0.74, + "learning_rate": 3.454506857303753e-05, + "loss": 0.4188, + "step": 24699 + }, + { + "epoch": 0.74, + "learning_rate": 3.453777903139389e-05, + "loss": 0.6081, + "step": 24700 + }, + { + "epoch": 0.74, + "learning_rate": 3.4530490098391933e-05, + "loss": 0.4875, + "step": 24701 + }, + { + "epoch": 0.74, + "learning_rate": 3.452320177409939e-05, + "loss": 0.569, + "step": 24702 + }, + { + "epoch": 0.74, + "learning_rate": 3.4515914058583974e-05, + "loss": 0.4844, + "step": 24703 + }, + { + "epoch": 0.74, + "learning_rate": 3.45086269519135e-05, + "loss": 0.5312, + "step": 24704 + }, + { + "epoch": 0.74, + "learning_rate": 3.450134045415574e-05, + "loss": 0.5127, + "step": 24705 + }, + { + "epoch": 0.74, + "learning_rate": 3.4494054565378405e-05, + "loss": 0.4527, + "step": 24706 + }, + { + "epoch": 0.74, + "learning_rate": 3.448676928564922e-05, + "loss": 0.6341, + "step": 24707 + }, + { + "epoch": 0.74, + "learning_rate": 3.447948461503594e-05, + "loss": 0.4402, + "step": 24708 + }, + { + "epoch": 0.74, + "learning_rate": 3.447220055360629e-05, + "loss": 0.3987, + "step": 24709 + }, + { + "epoch": 0.74, + "learning_rate": 3.446491710142805e-05, + "loss": 0.4479, + "step": 24710 + }, + { + "epoch": 0.74, + "learning_rate": 3.445763425856888e-05, + "loss": 0.5557, + "step": 24711 + }, + { + "epoch": 0.74, + "learning_rate": 3.4450352025096475e-05, + "loss": 0.4922, + "step": 24712 + }, + { + "epoch": 0.74, + "learning_rate": 3.4443070401078605e-05, + "loss": 0.4717, + "step": 24713 + }, + { + "epoch": 0.74, + "learning_rate": 3.4435789386582906e-05, + "loss": 0.589, + "step": 24714 + }, + { + "epoch": 0.74, + "learning_rate": 3.4428508981677134e-05, + "loss": 0.5628, + "step": 24715 + }, + { + "epoch": 0.74, + "learning_rate": 3.4421229186428924e-05, + "loss": 0.5369, + "step": 24716 + }, + { + "epoch": 0.74, + "learning_rate": 3.4413950000906016e-05, + "loss": 0.5395, + "step": 24717 + }, + { + "epoch": 0.74, + "learning_rate": 3.440667142517603e-05, + "loss": 0.5805, + "step": 24718 + }, + { + "epoch": 0.74, + "learning_rate": 3.4399393459306674e-05, + "loss": 0.5707, + "step": 24719 + }, + { + "epoch": 0.74, + "learning_rate": 3.4392116103365636e-05, + "loss": 0.4424, + "step": 24720 + }, + { + "epoch": 0.74, + "learning_rate": 3.4384839357420565e-05, + "loss": 0.6024, + "step": 24721 + }, + { + "epoch": 0.74, + "learning_rate": 3.437756322153907e-05, + "loss": 0.5302, + "step": 24722 + }, + { + "epoch": 0.74, + "learning_rate": 3.437028769578883e-05, + "loss": 0.5683, + "step": 24723 + }, + { + "epoch": 0.74, + "learning_rate": 3.4363012780237524e-05, + "loss": 0.5086, + "step": 24724 + }, + { + "epoch": 0.74, + "learning_rate": 3.4355738474952784e-05, + "loss": 0.4612, + "step": 24725 + }, + { + "epoch": 0.74, + "learning_rate": 3.4348464780002175e-05, + "loss": 0.5702, + "step": 24726 + }, + { + "epoch": 0.74, + "learning_rate": 3.434119169545339e-05, + "loss": 0.49, + "step": 24727 + }, + { + "epoch": 0.74, + "learning_rate": 3.4333919221374025e-05, + "loss": 0.5056, + "step": 24728 + }, + { + "epoch": 0.74, + "learning_rate": 3.4326647357831745e-05, + "loss": 0.4496, + "step": 24729 + }, + { + "epoch": 0.74, + "learning_rate": 3.431937610489412e-05, + "loss": 0.5409, + "step": 24730 + }, + { + "epoch": 0.74, + "learning_rate": 3.431210546262873e-05, + "loss": 0.6074, + "step": 24731 + }, + { + "epoch": 0.74, + "learning_rate": 3.43048354311032e-05, + "loss": 0.5722, + "step": 24732 + }, + { + "epoch": 0.74, + "learning_rate": 3.4297566010385155e-05, + "loss": 0.4783, + "step": 24733 + }, + { + "epoch": 0.74, + "learning_rate": 3.429029720054216e-05, + "loss": 0.6184, + "step": 24734 + }, + { + "epoch": 0.74, + "learning_rate": 3.4283029001641756e-05, + "loss": 0.5323, + "step": 24735 + }, + { + "epoch": 0.74, + "learning_rate": 3.427576141375156e-05, + "loss": 0.6404, + "step": 24736 + }, + { + "epoch": 0.74, + "learning_rate": 3.426849443693917e-05, + "loss": 0.4613, + "step": 24737 + }, + { + "epoch": 0.74, + "learning_rate": 3.426122807127209e-05, + "loss": 0.4982, + "step": 24738 + }, + { + "epoch": 0.74, + "learning_rate": 3.4253962316817946e-05, + "loss": 0.4312, + "step": 24739 + }, + { + "epoch": 0.74, + "learning_rate": 3.424669717364422e-05, + "loss": 0.5075, + "step": 24740 + }, + { + "epoch": 0.74, + "learning_rate": 3.4239432641818525e-05, + "loss": 0.5668, + "step": 24741 + }, + { + "epoch": 0.74, + "learning_rate": 3.423216872140834e-05, + "loss": 0.5841, + "step": 24742 + }, + { + "epoch": 0.74, + "learning_rate": 3.4224905412481246e-05, + "loss": 0.5447, + "step": 24743 + }, + { + "epoch": 0.74, + "learning_rate": 3.4217642715104805e-05, + "loss": 0.5046, + "step": 24744 + }, + { + "epoch": 0.74, + "learning_rate": 3.4210380629346495e-05, + "loss": 0.4627, + "step": 24745 + }, + { + "epoch": 0.74, + "learning_rate": 3.420311915527381e-05, + "loss": 0.5661, + "step": 24746 + }, + { + "epoch": 0.74, + "learning_rate": 3.419585829295431e-05, + "loss": 0.6265, + "step": 24747 + }, + { + "epoch": 0.74, + "learning_rate": 3.418859804245551e-05, + "loss": 0.4907, + "step": 24748 + }, + { + "epoch": 0.74, + "learning_rate": 3.418133840384491e-05, + "loss": 0.5739, + "step": 24749 + }, + { + "epoch": 0.74, + "learning_rate": 3.417407937718995e-05, + "loss": 0.5775, + "step": 24750 + }, + { + "epoch": 0.74, + "learning_rate": 3.4166820962558166e-05, + "loss": 0.6555, + "step": 24751 + }, + { + "epoch": 0.74, + "learning_rate": 3.415956316001708e-05, + "loss": 0.448, + "step": 24752 + }, + { + "epoch": 0.74, + "learning_rate": 3.4152305969634104e-05, + "loss": 0.6342, + "step": 24753 + }, + { + "epoch": 0.74, + "learning_rate": 3.414504939147677e-05, + "loss": 0.5993, + "step": 24754 + }, + { + "epoch": 0.74, + "learning_rate": 3.4137793425612485e-05, + "loss": 0.518, + "step": 24755 + }, + { + "epoch": 0.74, + "learning_rate": 3.4130538072108774e-05, + "loss": 0.4543, + "step": 24756 + }, + { + "epoch": 0.74, + "learning_rate": 3.412328333103304e-05, + "loss": 0.6101, + "step": 24757 + }, + { + "epoch": 0.74, + "learning_rate": 3.41160292024528e-05, + "loss": 0.5693, + "step": 24758 + }, + { + "epoch": 0.74, + "learning_rate": 3.4108775686435425e-05, + "loss": 0.5618, + "step": 24759 + }, + { + "epoch": 0.74, + "learning_rate": 3.410152278304842e-05, + "loss": 0.5091, + "step": 24760 + }, + { + "epoch": 0.74, + "learning_rate": 3.409427049235917e-05, + "loss": 0.5407, + "step": 24761 + }, + { + "epoch": 0.74, + "learning_rate": 3.4087018814435125e-05, + "loss": 0.5541, + "step": 24762 + }, + { + "epoch": 0.74, + "learning_rate": 3.4079767749343724e-05, + "loss": 0.481, + "step": 24763 + }, + { + "epoch": 0.74, + "learning_rate": 3.407251729715238e-05, + "loss": 0.5298, + "step": 24764 + }, + { + "epoch": 0.74, + "learning_rate": 3.406526745792847e-05, + "loss": 0.4765, + "step": 24765 + }, + { + "epoch": 0.74, + "learning_rate": 3.405801823173942e-05, + "loss": 0.6178, + "step": 24766 + }, + { + "epoch": 0.74, + "learning_rate": 3.4050769618652656e-05, + "loss": 0.5373, + "step": 24767 + }, + { + "epoch": 0.74, + "learning_rate": 3.404352161873556e-05, + "loss": 0.4221, + "step": 24768 + }, + { + "epoch": 0.74, + "learning_rate": 3.403627423205548e-05, + "loss": 0.5736, + "step": 24769 + }, + { + "epoch": 0.74, + "learning_rate": 3.402902745867983e-05, + "loss": 0.4105, + "step": 24770 + }, + { + "epoch": 0.74, + "learning_rate": 3.402178129867599e-05, + "loss": 0.4987, + "step": 24771 + }, + { + "epoch": 0.74, + "learning_rate": 3.4014535752111366e-05, + "loss": 0.616, + "step": 24772 + }, + { + "epoch": 0.74, + "learning_rate": 3.400729081905328e-05, + "loss": 0.4383, + "step": 24773 + }, + { + "epoch": 0.74, + "learning_rate": 3.400004649956908e-05, + "loss": 0.5716, + "step": 24774 + }, + { + "epoch": 0.74, + "learning_rate": 3.3992802793726145e-05, + "loss": 0.6563, + "step": 24775 + }, + { + "epoch": 0.74, + "learning_rate": 3.3985559701591844e-05, + "loss": 0.4174, + "step": 24776 + }, + { + "epoch": 0.74, + "learning_rate": 3.39783172232335e-05, + "loss": 0.4975, + "step": 24777 + }, + { + "epoch": 0.74, + "learning_rate": 3.397107535871842e-05, + "loss": 0.4985, + "step": 24778 + }, + { + "epoch": 0.74, + "learning_rate": 3.396383410811399e-05, + "loss": 0.5553, + "step": 24779 + }, + { + "epoch": 0.74, + "learning_rate": 3.395659347148749e-05, + "loss": 0.5022, + "step": 24780 + }, + { + "epoch": 0.74, + "learning_rate": 3.3949353448906254e-05, + "loss": 0.5346, + "step": 24781 + }, + { + "epoch": 0.74, + "learning_rate": 3.394211404043763e-05, + "loss": 0.5361, + "step": 24782 + }, + { + "epoch": 0.74, + "learning_rate": 3.3934875246148924e-05, + "loss": 0.5998, + "step": 24783 + }, + { + "epoch": 0.74, + "learning_rate": 3.392763706610736e-05, + "loss": 0.5402, + "step": 24784 + }, + { + "epoch": 0.74, + "learning_rate": 3.392039950038031e-05, + "loss": 0.6244, + "step": 24785 + }, + { + "epoch": 0.74, + "learning_rate": 3.3913162549035035e-05, + "loss": 0.5652, + "step": 24786 + }, + { + "epoch": 0.74, + "learning_rate": 3.3905926212138915e-05, + "loss": 0.5786, + "step": 24787 + }, + { + "epoch": 0.74, + "learning_rate": 3.389869048975908e-05, + "loss": 0.578, + "step": 24788 + }, + { + "epoch": 0.74, + "learning_rate": 3.389145538196287e-05, + "loss": 0.5324, + "step": 24789 + }, + { + "epoch": 0.74, + "learning_rate": 3.388422088881756e-05, + "loss": 0.6021, + "step": 24790 + }, + { + "epoch": 0.74, + "learning_rate": 3.387698701039045e-05, + "loss": 0.5922, + "step": 24791 + }, + { + "epoch": 0.74, + "learning_rate": 3.386975374674877e-05, + "loss": 0.4857, + "step": 24792 + }, + { + "epoch": 0.74, + "learning_rate": 3.3862521097959707e-05, + "loss": 0.5231, + "step": 24793 + }, + { + "epoch": 0.74, + "learning_rate": 3.3855289064090576e-05, + "loss": 0.4497, + "step": 24794 + }, + { + "epoch": 0.74, + "learning_rate": 3.384805764520864e-05, + "loss": 0.489, + "step": 24795 + }, + { + "epoch": 0.74, + "learning_rate": 3.3840826841381056e-05, + "loss": 0.6599, + "step": 24796 + }, + { + "epoch": 0.74, + "learning_rate": 3.3833596652675135e-05, + "loss": 0.5182, + "step": 24797 + }, + { + "epoch": 0.74, + "learning_rate": 3.3826367079158036e-05, + "loss": 0.5081, + "step": 24798 + }, + { + "epoch": 0.74, + "learning_rate": 3.381913812089702e-05, + "loss": 0.5674, + "step": 24799 + }, + { + "epoch": 0.74, + "learning_rate": 3.381190977795926e-05, + "loss": 0.5122, + "step": 24800 + }, + { + "epoch": 0.74, + "learning_rate": 3.380468205041201e-05, + "loss": 0.6032, + "step": 24801 + }, + { + "epoch": 0.74, + "learning_rate": 3.3797454938322415e-05, + "loss": 0.5399, + "step": 24802 + }, + { + "epoch": 0.74, + "learning_rate": 3.379022844175772e-05, + "loss": 0.5154, + "step": 24803 + }, + { + "epoch": 0.74, + "learning_rate": 3.378300256078507e-05, + "loss": 0.6242, + "step": 24804 + }, + { + "epoch": 0.74, + "learning_rate": 3.377577729547167e-05, + "loss": 0.5907, + "step": 24805 + }, + { + "epoch": 0.74, + "learning_rate": 3.376855264588472e-05, + "loss": 0.5924, + "step": 24806 + }, + { + "epoch": 0.74, + "learning_rate": 3.376132861209138e-05, + "loss": 0.5273, + "step": 24807 + }, + { + "epoch": 0.74, + "learning_rate": 3.375410519415877e-05, + "loss": 0.5723, + "step": 24808 + }, + { + "epoch": 0.74, + "learning_rate": 3.374688239215409e-05, + "loss": 0.5042, + "step": 24809 + }, + { + "epoch": 0.74, + "learning_rate": 3.373966020614453e-05, + "loss": 0.6516, + "step": 24810 + }, + { + "epoch": 0.74, + "learning_rate": 3.37324386361972e-05, + "loss": 0.4657, + "step": 24811 + }, + { + "epoch": 0.74, + "learning_rate": 3.3725217682379195e-05, + "loss": 0.4891, + "step": 24812 + }, + { + "epoch": 0.74, + "learning_rate": 3.3717997344757714e-05, + "loss": 0.5483, + "step": 24813 + }, + { + "epoch": 0.74, + "learning_rate": 3.3710777623399904e-05, + "loss": 0.5552, + "step": 24814 + }, + { + "epoch": 0.74, + "learning_rate": 3.370355851837283e-05, + "loss": 0.6325, + "step": 24815 + }, + { + "epoch": 0.74, + "learning_rate": 3.369634002974369e-05, + "loss": 0.5337, + "step": 24816 + }, + { + "epoch": 0.74, + "learning_rate": 3.368912215757951e-05, + "loss": 0.5229, + "step": 24817 + }, + { + "epoch": 0.74, + "learning_rate": 3.368190490194748e-05, + "loss": 0.4736, + "step": 24818 + }, + { + "epoch": 0.74, + "learning_rate": 3.3674688262914634e-05, + "loss": 0.578, + "step": 24819 + }, + { + "epoch": 0.74, + "learning_rate": 3.366747224054813e-05, + "loss": 0.5445, + "step": 24820 + }, + { + "epoch": 0.74, + "learning_rate": 3.3660256834915e-05, + "loss": 0.6333, + "step": 24821 + }, + { + "epoch": 0.74, + "learning_rate": 3.365304204608239e-05, + "loss": 0.4354, + "step": 24822 + }, + { + "epoch": 0.74, + "learning_rate": 3.364582787411733e-05, + "loss": 0.5124, + "step": 24823 + }, + { + "epoch": 0.74, + "learning_rate": 3.363861431908691e-05, + "loss": 0.6447, + "step": 24824 + }, + { + "epoch": 0.74, + "learning_rate": 3.363140138105825e-05, + "loss": 0.5762, + "step": 24825 + }, + { + "epoch": 0.74, + "learning_rate": 3.3624189060098355e-05, + "loss": 0.5897, + "step": 24826 + }, + { + "epoch": 0.74, + "learning_rate": 3.361697735627427e-05, + "loss": 0.527, + "step": 24827 + }, + { + "epoch": 0.74, + "learning_rate": 3.3609766269653066e-05, + "loss": 0.4949, + "step": 24828 + }, + { + "epoch": 0.74, + "learning_rate": 3.360255580030183e-05, + "loss": 0.5615, + "step": 24829 + }, + { + "epoch": 0.74, + "learning_rate": 3.359534594828756e-05, + "loss": 0.4986, + "step": 24830 + }, + { + "epoch": 0.74, + "learning_rate": 3.3588136713677276e-05, + "loss": 0.5939, + "step": 24831 + }, + { + "epoch": 0.74, + "learning_rate": 3.3580928096538024e-05, + "loss": 0.6024, + "step": 24832 + }, + { + "epoch": 0.74, + "learning_rate": 3.3573720096936824e-05, + "loss": 0.503, + "step": 24833 + }, + { + "epoch": 0.74, + "learning_rate": 3.356651271494073e-05, + "loss": 0.458, + "step": 24834 + }, + { + "epoch": 0.74, + "learning_rate": 3.355930595061674e-05, + "loss": 0.489, + "step": 24835 + }, + { + "epoch": 0.74, + "learning_rate": 3.355209980403179e-05, + "loss": 0.5988, + "step": 24836 + }, + { + "epoch": 0.74, + "learning_rate": 3.3544894275252946e-05, + "loss": 0.5463, + "step": 24837 + }, + { + "epoch": 0.74, + "learning_rate": 3.3537689364347216e-05, + "loss": 0.5347, + "step": 24838 + }, + { + "epoch": 0.74, + "learning_rate": 3.3530485071381566e-05, + "loss": 0.6059, + "step": 24839 + }, + { + "epoch": 0.74, + "learning_rate": 3.352328139642293e-05, + "loss": 0.5725, + "step": 24840 + }, + { + "epoch": 0.74, + "learning_rate": 3.3516078339538336e-05, + "loss": 0.5619, + "step": 24841 + }, + { + "epoch": 0.74, + "learning_rate": 3.350887590079478e-05, + "loss": 0.4797, + "step": 24842 + }, + { + "epoch": 0.74, + "learning_rate": 3.350167408025917e-05, + "loss": 0.549, + "step": 24843 + }, + { + "epoch": 0.74, + "learning_rate": 3.349447287799852e-05, + "loss": 0.5226, + "step": 24844 + }, + { + "epoch": 0.74, + "learning_rate": 3.348727229407974e-05, + "loss": 0.515, + "step": 24845 + }, + { + "epoch": 0.74, + "learning_rate": 3.348007232856978e-05, + "loss": 0.6144, + "step": 24846 + }, + { + "epoch": 0.74, + "learning_rate": 3.3472872981535585e-05, + "loss": 0.4859, + "step": 24847 + }, + { + "epoch": 0.74, + "learning_rate": 3.3465674253044096e-05, + "loss": 0.5567, + "step": 24848 + }, + { + "epoch": 0.74, + "learning_rate": 3.345847614316232e-05, + "loss": 0.6468, + "step": 24849 + }, + { + "epoch": 0.74, + "learning_rate": 3.3451278651957054e-05, + "loss": 0.5291, + "step": 24850 + }, + { + "epoch": 0.74, + "learning_rate": 3.344408177949527e-05, + "loss": 0.5608, + "step": 24851 + }, + { + "epoch": 0.74, + "learning_rate": 3.3436885525843885e-05, + "loss": 0.538, + "step": 24852 + }, + { + "epoch": 0.74, + "learning_rate": 3.342968989106985e-05, + "loss": 0.519, + "step": 24853 + }, + { + "epoch": 0.74, + "learning_rate": 3.3422494875240016e-05, + "loss": 0.4616, + "step": 24854 + }, + { + "epoch": 0.74, + "learning_rate": 3.341530047842126e-05, + "loss": 0.5812, + "step": 24855 + }, + { + "epoch": 0.74, + "learning_rate": 3.340810670068051e-05, + "loss": 0.5051, + "step": 24856 + }, + { + "epoch": 0.74, + "learning_rate": 3.3400913542084664e-05, + "loss": 0.4134, + "step": 24857 + }, + { + "epoch": 0.74, + "learning_rate": 3.3393721002700554e-05, + "loss": 0.4489, + "step": 24858 + }, + { + "epoch": 0.74, + "learning_rate": 3.3386529082595107e-05, + "loss": 0.4946, + "step": 24859 + }, + { + "epoch": 0.74, + "learning_rate": 3.337933778183514e-05, + "loss": 0.494, + "step": 24860 + }, + { + "epoch": 0.74, + "learning_rate": 3.337214710048756e-05, + "loss": 0.6291, + "step": 24861 + }, + { + "epoch": 0.74, + "learning_rate": 3.336495703861917e-05, + "loss": 0.6099, + "step": 24862 + }, + { + "epoch": 0.74, + "learning_rate": 3.335776759629689e-05, + "loss": 0.5664, + "step": 24863 + }, + { + "epoch": 0.74, + "learning_rate": 3.335057877358749e-05, + "loss": 0.5455, + "step": 24864 + }, + { + "epoch": 0.74, + "learning_rate": 3.334339057055787e-05, + "loss": 0.6672, + "step": 24865 + }, + { + "epoch": 0.74, + "learning_rate": 3.3336202987274823e-05, + "loss": 0.5476, + "step": 24866 + }, + { + "epoch": 0.74, + "learning_rate": 3.332901602380517e-05, + "loss": 0.5086, + "step": 24867 + }, + { + "epoch": 0.74, + "learning_rate": 3.33218296802158e-05, + "loss": 0.5439, + "step": 24868 + }, + { + "epoch": 0.74, + "learning_rate": 3.3314643956573476e-05, + "loss": 0.5156, + "step": 24869 + }, + { + "epoch": 0.74, + "learning_rate": 3.3307458852945e-05, + "loss": 0.4472, + "step": 24870 + }, + { + "epoch": 0.74, + "learning_rate": 3.3300274369397166e-05, + "loss": 0.508, + "step": 24871 + }, + { + "epoch": 0.74, + "learning_rate": 3.3293090505996836e-05, + "loss": 0.4879, + "step": 24872 + }, + { + "epoch": 0.74, + "learning_rate": 3.328590726281077e-05, + "loss": 0.5809, + "step": 24873 + }, + { + "epoch": 0.74, + "learning_rate": 3.327872463990571e-05, + "loss": 0.5404, + "step": 24874 + }, + { + "epoch": 0.74, + "learning_rate": 3.327154263734849e-05, + "loss": 0.5606, + "step": 24875 + }, + { + "epoch": 0.74, + "learning_rate": 3.3264361255205856e-05, + "loss": 0.523, + "step": 24876 + }, + { + "epoch": 0.74, + "learning_rate": 3.325718049354463e-05, + "loss": 0.5955, + "step": 24877 + }, + { + "epoch": 0.74, + "learning_rate": 3.325000035243153e-05, + "loss": 0.536, + "step": 24878 + }, + { + "epoch": 0.74, + "learning_rate": 3.32428208319333e-05, + "loss": 0.5521, + "step": 24879 + }, + { + "epoch": 0.74, + "learning_rate": 3.3235641932116746e-05, + "loss": 0.5655, + "step": 24880 + }, + { + "epoch": 0.74, + "learning_rate": 3.3228463653048546e-05, + "loss": 0.4878, + "step": 24881 + }, + { + "epoch": 0.74, + "learning_rate": 3.3221285994795526e-05, + "loss": 0.5914, + "step": 24882 + }, + { + "epoch": 0.74, + "learning_rate": 3.3214108957424325e-05, + "loss": 0.5648, + "step": 24883 + }, + { + "epoch": 0.74, + "learning_rate": 3.3206932541001756e-05, + "loss": 0.5783, + "step": 24884 + }, + { + "epoch": 0.74, + "learning_rate": 3.319975674559448e-05, + "loss": 0.4782, + "step": 24885 + }, + { + "epoch": 0.74, + "learning_rate": 3.319258157126924e-05, + "loss": 0.5806, + "step": 24886 + }, + { + "epoch": 0.74, + "learning_rate": 3.318540701809278e-05, + "loss": 0.5479, + "step": 24887 + }, + { + "epoch": 0.74, + "learning_rate": 3.317823308613178e-05, + "loss": 0.4968, + "step": 24888 + }, + { + "epoch": 0.74, + "learning_rate": 3.317105977545291e-05, + "loss": 0.5633, + "step": 24889 + }, + { + "epoch": 0.74, + "learning_rate": 3.316388708612287e-05, + "loss": 0.5377, + "step": 24890 + }, + { + "epoch": 0.74, + "learning_rate": 3.315671501820842e-05, + "loss": 0.4743, + "step": 24891 + }, + { + "epoch": 0.74, + "learning_rate": 3.314954357177619e-05, + "loss": 0.5238, + "step": 24892 + }, + { + "epoch": 0.74, + "learning_rate": 3.314237274689282e-05, + "loss": 0.5321, + "step": 24893 + }, + { + "epoch": 0.74, + "learning_rate": 3.313520254362502e-05, + "loss": 0.5786, + "step": 24894 + }, + { + "epoch": 0.74, + "learning_rate": 3.312803296203947e-05, + "loss": 0.7228, + "step": 24895 + }, + { + "epoch": 0.74, + "learning_rate": 3.3120864002202846e-05, + "loss": 0.5288, + "step": 24896 + }, + { + "epoch": 0.74, + "learning_rate": 3.3113695664181764e-05, + "loss": 0.5592, + "step": 24897 + }, + { + "epoch": 0.74, + "learning_rate": 3.310652794804285e-05, + "loss": 0.5374, + "step": 24898 + }, + { + "epoch": 0.74, + "learning_rate": 3.309936085385279e-05, + "loss": 0.4622, + "step": 24899 + }, + { + "epoch": 0.74, + "learning_rate": 3.3092194381678235e-05, + "loss": 0.5292, + "step": 24900 + }, + { + "epoch": 0.74, + "learning_rate": 3.3085028531585785e-05, + "loss": 0.5178, + "step": 24901 + }, + { + "epoch": 0.74, + "learning_rate": 3.307786330364204e-05, + "loss": 0.5373, + "step": 24902 + }, + { + "epoch": 0.74, + "learning_rate": 3.307069869791365e-05, + "loss": 0.5632, + "step": 24903 + }, + { + "epoch": 0.74, + "learning_rate": 3.306353471446726e-05, + "loss": 0.4066, + "step": 24904 + }, + { + "epoch": 0.74, + "learning_rate": 3.305637135336942e-05, + "loss": 0.4835, + "step": 24905 + }, + { + "epoch": 0.74, + "learning_rate": 3.304920861468678e-05, + "loss": 0.4974, + "step": 24906 + }, + { + "epoch": 0.74, + "learning_rate": 3.304204649848589e-05, + "loss": 0.6186, + "step": 24907 + }, + { + "epoch": 0.74, + "learning_rate": 3.303488500483339e-05, + "loss": 0.4505, + "step": 24908 + }, + { + "epoch": 0.74, + "learning_rate": 3.3027724133795814e-05, + "loss": 0.5371, + "step": 24909 + }, + { + "epoch": 0.74, + "learning_rate": 3.302056388543977e-05, + "loss": 0.5275, + "step": 24910 + }, + { + "epoch": 0.74, + "learning_rate": 3.301340425983188e-05, + "loss": 0.5698, + "step": 24911 + }, + { + "epoch": 0.74, + "learning_rate": 3.3006245257038605e-05, + "loss": 0.5168, + "step": 24912 + }, + { + "epoch": 0.74, + "learning_rate": 3.299908687712655e-05, + "loss": 0.6175, + "step": 24913 + }, + { + "epoch": 0.74, + "learning_rate": 3.299192912016229e-05, + "loss": 0.5842, + "step": 24914 + }, + { + "epoch": 0.74, + "learning_rate": 3.298477198621239e-05, + "loss": 0.4949, + "step": 24915 + }, + { + "epoch": 0.74, + "learning_rate": 3.297761547534338e-05, + "loss": 0.5567, + "step": 24916 + }, + { + "epoch": 0.74, + "learning_rate": 3.297045958762176e-05, + "loss": 0.5309, + "step": 24917 + }, + { + "epoch": 0.74, + "learning_rate": 3.296330432311409e-05, + "loss": 0.4883, + "step": 24918 + }, + { + "epoch": 0.74, + "learning_rate": 3.2956149681886915e-05, + "loss": 0.5527, + "step": 24919 + }, + { + "epoch": 0.74, + "learning_rate": 3.2948995664006724e-05, + "loss": 0.5281, + "step": 24920 + }, + { + "epoch": 0.74, + "learning_rate": 3.294184226954007e-05, + "loss": 0.5401, + "step": 24921 + }, + { + "epoch": 0.74, + "learning_rate": 3.2934689498553406e-05, + "loss": 0.5544, + "step": 24922 + }, + { + "epoch": 0.74, + "learning_rate": 3.2927537351113304e-05, + "loss": 0.4871, + "step": 24923 + }, + { + "epoch": 0.74, + "learning_rate": 3.292038582728621e-05, + "loss": 0.5264, + "step": 24924 + }, + { + "epoch": 0.74, + "learning_rate": 3.291323492713865e-05, + "loss": 0.5679, + "step": 24925 + }, + { + "epoch": 0.74, + "learning_rate": 3.2906084650737056e-05, + "loss": 0.4684, + "step": 24926 + }, + { + "epoch": 0.74, + "learning_rate": 3.289893499814799e-05, + "loss": 0.5587, + "step": 24927 + }, + { + "epoch": 0.74, + "learning_rate": 3.2891785969437836e-05, + "loss": 0.5174, + "step": 24928 + }, + { + "epoch": 0.74, + "learning_rate": 3.2884637564673127e-05, + "loss": 0.5855, + "step": 24929 + }, + { + "epoch": 0.74, + "learning_rate": 3.2877489783920334e-05, + "loss": 0.5051, + "step": 24930 + }, + { + "epoch": 0.74, + "learning_rate": 3.2870342627245886e-05, + "loss": 0.6009, + "step": 24931 + }, + { + "epoch": 0.74, + "learning_rate": 3.286319609471621e-05, + "loss": 0.538, + "step": 24932 + }, + { + "epoch": 0.74, + "learning_rate": 3.285605018639777e-05, + "loss": 0.4823, + "step": 24933 + }, + { + "epoch": 0.74, + "learning_rate": 3.284890490235706e-05, + "loss": 0.4843, + "step": 24934 + }, + { + "epoch": 0.74, + "learning_rate": 3.284176024266046e-05, + "loss": 0.6749, + "step": 24935 + }, + { + "epoch": 0.74, + "learning_rate": 3.283461620737437e-05, + "loss": 0.5087, + "step": 24936 + }, + { + "epoch": 0.74, + "learning_rate": 3.282747279656525e-05, + "loss": 0.5335, + "step": 24937 + }, + { + "epoch": 0.74, + "learning_rate": 3.282033001029952e-05, + "loss": 0.4927, + "step": 24938 + }, + { + "epoch": 0.74, + "learning_rate": 3.281318784864362e-05, + "loss": 0.4335, + "step": 24939 + }, + { + "epoch": 0.74, + "learning_rate": 3.280604631166392e-05, + "loss": 0.4818, + "step": 24940 + }, + { + "epoch": 0.74, + "learning_rate": 3.2798905399426784e-05, + "loss": 0.5172, + "step": 24941 + }, + { + "epoch": 0.74, + "learning_rate": 3.279176511199865e-05, + "loss": 0.4741, + "step": 24942 + }, + { + "epoch": 0.74, + "learning_rate": 3.2784625449445914e-05, + "loss": 0.4655, + "step": 24943 + }, + { + "epoch": 0.74, + "learning_rate": 3.277748641183496e-05, + "loss": 0.5019, + "step": 24944 + }, + { + "epoch": 0.74, + "learning_rate": 3.277034799923211e-05, + "loss": 0.486, + "step": 24945 + }, + { + "epoch": 0.74, + "learning_rate": 3.276321021170379e-05, + "loss": 0.4943, + "step": 24946 + }, + { + "epoch": 0.74, + "learning_rate": 3.2756073049316314e-05, + "loss": 0.4806, + "step": 24947 + }, + { + "epoch": 0.74, + "learning_rate": 3.2748936512136075e-05, + "loss": 0.5472, + "step": 24948 + }, + { + "epoch": 0.74, + "learning_rate": 3.274180060022945e-05, + "loss": 0.5271, + "step": 24949 + }, + { + "epoch": 0.74, + "learning_rate": 3.273466531366276e-05, + "loss": 0.5538, + "step": 24950 + }, + { + "epoch": 0.74, + "learning_rate": 3.2727530652502314e-05, + "loss": 0.4586, + "step": 24951 + }, + { + "epoch": 0.74, + "learning_rate": 3.272039661681446e-05, + "loss": 0.5368, + "step": 24952 + }, + { + "epoch": 0.74, + "learning_rate": 3.2713263206665566e-05, + "loss": 0.4892, + "step": 24953 + }, + { + "epoch": 0.74, + "learning_rate": 3.2706130422121984e-05, + "loss": 0.5764, + "step": 24954 + }, + { + "epoch": 0.74, + "learning_rate": 3.269899826324991e-05, + "loss": 0.5439, + "step": 24955 + }, + { + "epoch": 0.74, + "learning_rate": 3.269186673011574e-05, + "loss": 0.4983, + "step": 24956 + }, + { + "epoch": 0.74, + "learning_rate": 3.268473582278576e-05, + "loss": 0.4577, + "step": 24957 + }, + { + "epoch": 0.74, + "learning_rate": 3.267760554132631e-05, + "loss": 0.4497, + "step": 24958 + }, + { + "epoch": 0.74, + "learning_rate": 3.2670475885803654e-05, + "loss": 0.5006, + "step": 24959 + }, + { + "epoch": 0.74, + "learning_rate": 3.266334685628404e-05, + "loss": 0.4109, + "step": 24960 + }, + { + "epoch": 0.74, + "learning_rate": 3.26562184528338e-05, + "loss": 0.5611, + "step": 24961 + }, + { + "epoch": 0.74, + "learning_rate": 3.264909067551922e-05, + "loss": 0.5702, + "step": 24962 + }, + { + "epoch": 0.74, + "learning_rate": 3.264196352440653e-05, + "loss": 0.6115, + "step": 24963 + }, + { + "epoch": 0.74, + "learning_rate": 3.263483699956203e-05, + "loss": 0.5373, + "step": 24964 + }, + { + "epoch": 0.74, + "learning_rate": 3.262771110105195e-05, + "loss": 0.6227, + "step": 24965 + }, + { + "epoch": 0.74, + "learning_rate": 3.2620585828942597e-05, + "loss": 0.5301, + "step": 24966 + }, + { + "epoch": 0.74, + "learning_rate": 3.261346118330014e-05, + "loss": 0.5457, + "step": 24967 + }, + { + "epoch": 0.74, + "learning_rate": 3.26063371641909e-05, + "loss": 0.5444, + "step": 24968 + }, + { + "epoch": 0.74, + "learning_rate": 3.2599213771681045e-05, + "loss": 0.5017, + "step": 24969 + }, + { + "epoch": 0.74, + "learning_rate": 3.259209100583687e-05, + "loss": 0.4458, + "step": 24970 + }, + { + "epoch": 0.74, + "learning_rate": 3.258496886672453e-05, + "loss": 0.4971, + "step": 24971 + }, + { + "epoch": 0.74, + "learning_rate": 3.257784735441029e-05, + "loss": 0.506, + "step": 24972 + }, + { + "epoch": 0.74, + "learning_rate": 3.257072646896037e-05, + "loss": 0.481, + "step": 24973 + }, + { + "epoch": 0.74, + "learning_rate": 3.256360621044097e-05, + "loss": 0.4943, + "step": 24974 + }, + { + "epoch": 0.74, + "learning_rate": 3.255648657891824e-05, + "loss": 0.5577, + "step": 24975 + }, + { + "epoch": 0.74, + "learning_rate": 3.2549367574458436e-05, + "loss": 0.5028, + "step": 24976 + }, + { + "epoch": 0.74, + "learning_rate": 3.254224919712775e-05, + "loss": 0.5256, + "step": 24977 + }, + { + "epoch": 0.74, + "learning_rate": 3.253513144699234e-05, + "loss": 0.4928, + "step": 24978 + }, + { + "epoch": 0.74, + "learning_rate": 3.252801432411837e-05, + "loss": 0.5523, + "step": 24979 + }, + { + "epoch": 0.74, + "learning_rate": 3.2520897828572014e-05, + "loss": 0.601, + "step": 24980 + }, + { + "epoch": 0.74, + "learning_rate": 3.251378196041951e-05, + "loss": 0.5949, + "step": 24981 + }, + { + "epoch": 0.74, + "learning_rate": 3.2506666719726916e-05, + "loss": 0.6226, + "step": 24982 + }, + { + "epoch": 0.74, + "learning_rate": 3.249955210656047e-05, + "loss": 0.4261, + "step": 24983 + }, + { + "epoch": 0.74, + "learning_rate": 3.2492438120986254e-05, + "loss": 0.4818, + "step": 24984 + }, + { + "epoch": 0.74, + "learning_rate": 3.2485324763070465e-05, + "loss": 0.6375, + "step": 24985 + }, + { + "epoch": 0.74, + "learning_rate": 3.2478212032879206e-05, + "loss": 0.5081, + "step": 24986 + }, + { + "epoch": 0.74, + "learning_rate": 3.247109993047864e-05, + "loss": 0.4253, + "step": 24987 + }, + { + "epoch": 0.74, + "learning_rate": 3.246398845593484e-05, + "loss": 0.5521, + "step": 24988 + }, + { + "epoch": 0.74, + "learning_rate": 3.2456877609314e-05, + "loss": 0.6008, + "step": 24989 + }, + { + "epoch": 0.74, + "learning_rate": 3.244976739068214e-05, + "loss": 0.5012, + "step": 24990 + }, + { + "epoch": 0.74, + "learning_rate": 3.2442657800105434e-05, + "loss": 0.521, + "step": 24991 + }, + { + "epoch": 0.74, + "learning_rate": 3.243554883765e-05, + "loss": 0.6231, + "step": 24992 + }, + { + "epoch": 0.74, + "learning_rate": 3.242844050338191e-05, + "loss": 0.5132, + "step": 24993 + }, + { + "epoch": 0.74, + "learning_rate": 3.242133279736721e-05, + "loss": 0.4302, + "step": 24994 + }, + { + "epoch": 0.74, + "learning_rate": 3.241422571967202e-05, + "loss": 0.6604, + "step": 24995 + }, + { + "epoch": 0.74, + "learning_rate": 3.240711927036246e-05, + "loss": 0.607, + "step": 24996 + }, + { + "epoch": 0.74, + "learning_rate": 3.240001344950456e-05, + "loss": 0.4731, + "step": 24997 + }, + { + "epoch": 0.74, + "learning_rate": 3.239290825716437e-05, + "loss": 0.559, + "step": 24998 + }, + { + "epoch": 0.74, + "learning_rate": 3.238580369340797e-05, + "loss": 0.5048, + "step": 24999 + }, + { + "epoch": 0.74, + "learning_rate": 3.2378699758301414e-05, + "loss": 0.6094, + "step": 25000 + }, + { + "epoch": 0.74, + "learning_rate": 3.237159645191079e-05, + "loss": 0.5418, + "step": 25001 + }, + { + "epoch": 0.74, + "learning_rate": 3.236449377430211e-05, + "loss": 0.5556, + "step": 25002 + }, + { + "epoch": 0.74, + "learning_rate": 3.235739172554138e-05, + "loss": 0.5199, + "step": 25003 + }, + { + "epoch": 0.74, + "learning_rate": 3.235029030569465e-05, + "loss": 0.5466, + "step": 25004 + }, + { + "epoch": 0.74, + "learning_rate": 3.2343189514827996e-05, + "loss": 0.4608, + "step": 25005 + }, + { + "epoch": 0.74, + "learning_rate": 3.2336089353007405e-05, + "loss": 0.5684, + "step": 25006 + }, + { + "epoch": 0.74, + "learning_rate": 3.232898982029886e-05, + "loss": 0.4565, + "step": 25007 + }, + { + "epoch": 0.74, + "learning_rate": 3.232189091676838e-05, + "loss": 0.4658, + "step": 25008 + }, + { + "epoch": 0.74, + "learning_rate": 3.231479264248204e-05, + "loss": 0.6185, + "step": 25009 + }, + { + "epoch": 0.74, + "learning_rate": 3.2307694997505734e-05, + "loss": 0.5471, + "step": 25010 + }, + { + "epoch": 0.74, + "learning_rate": 3.2300597981905535e-05, + "loss": 0.5441, + "step": 25011 + }, + { + "epoch": 0.74, + "learning_rate": 3.229350159574739e-05, + "loss": 0.4812, + "step": 25012 + }, + { + "epoch": 0.74, + "learning_rate": 3.2286405839097255e-05, + "loss": 0.5891, + "step": 25013 + }, + { + "epoch": 0.74, + "learning_rate": 3.227931071202113e-05, + "loss": 0.4368, + "step": 25014 + }, + { + "epoch": 0.74, + "learning_rate": 3.2272216214584985e-05, + "loss": 0.4828, + "step": 25015 + }, + { + "epoch": 0.74, + "learning_rate": 3.226512234685484e-05, + "loss": 0.5758, + "step": 25016 + }, + { + "epoch": 0.74, + "learning_rate": 3.225802910889653e-05, + "loss": 0.535, + "step": 25017 + }, + { + "epoch": 0.74, + "learning_rate": 3.2250936500776055e-05, + "loss": 0.515, + "step": 25018 + }, + { + "epoch": 0.74, + "learning_rate": 3.2243844522559384e-05, + "loss": 0.5704, + "step": 25019 + }, + { + "epoch": 0.74, + "learning_rate": 3.223675317431246e-05, + "loss": 0.5363, + "step": 25020 + }, + { + "epoch": 0.74, + "learning_rate": 3.222966245610122e-05, + "loss": 0.5304, + "step": 25021 + }, + { + "epoch": 0.74, + "learning_rate": 3.222257236799151e-05, + "loss": 0.5093, + "step": 25022 + }, + { + "epoch": 0.74, + "learning_rate": 3.221548291004932e-05, + "loss": 0.6111, + "step": 25023 + }, + { + "epoch": 0.75, + "learning_rate": 3.2208394082340586e-05, + "loss": 0.5416, + "step": 25024 + }, + { + "epoch": 0.75, + "learning_rate": 3.220130588493115e-05, + "loss": 0.535, + "step": 25025 + }, + { + "epoch": 0.75, + "learning_rate": 3.219421831788699e-05, + "loss": 0.5907, + "step": 25026 + }, + { + "epoch": 0.75, + "learning_rate": 3.2187131381273924e-05, + "loss": 0.5061, + "step": 25027 + }, + { + "epoch": 0.75, + "learning_rate": 3.218004507515793e-05, + "loss": 0.5395, + "step": 25028 + }, + { + "epoch": 0.75, + "learning_rate": 3.2172959399604794e-05, + "loss": 0.6013, + "step": 25029 + }, + { + "epoch": 0.75, + "learning_rate": 3.216587435468049e-05, + "loss": 0.5989, + "step": 25030 + }, + { + "epoch": 0.75, + "learning_rate": 3.215878994045082e-05, + "loss": 0.6023, + "step": 25031 + }, + { + "epoch": 0.75, + "learning_rate": 3.215170615698171e-05, + "loss": 0.7274, + "step": 25032 + }, + { + "epoch": 0.75, + "learning_rate": 3.214462300433896e-05, + "loss": 0.499, + "step": 25033 + }, + { + "epoch": 0.75, + "learning_rate": 3.213754048258847e-05, + "loss": 0.6279, + "step": 25034 + }, + { + "epoch": 0.75, + "learning_rate": 3.213045859179611e-05, + "loss": 0.526, + "step": 25035 + }, + { + "epoch": 0.75, + "learning_rate": 3.2123377332027695e-05, + "loss": 0.5337, + "step": 25036 + }, + { + "epoch": 0.75, + "learning_rate": 3.2116296703349026e-05, + "loss": 0.633, + "step": 25037 + }, + { + "epoch": 0.75, + "learning_rate": 3.210921670582597e-05, + "loss": 0.6388, + "step": 25038 + }, + { + "epoch": 0.75, + "learning_rate": 3.210213733952441e-05, + "loss": 0.4865, + "step": 25039 + }, + { + "epoch": 0.75, + "learning_rate": 3.2095058604510097e-05, + "loss": 0.5289, + "step": 25040 + }, + { + "epoch": 0.75, + "learning_rate": 3.208798050084885e-05, + "loss": 0.4581, + "step": 25041 + }, + { + "epoch": 0.75, + "learning_rate": 3.208090302860648e-05, + "loss": 0.5465, + "step": 25042 + }, + { + "epoch": 0.75, + "learning_rate": 3.2073826187848844e-05, + "loss": 0.5544, + "step": 25043 + }, + { + "epoch": 0.75, + "learning_rate": 3.2066749978641664e-05, + "loss": 0.471, + "step": 25044 + }, + { + "epoch": 0.75, + "learning_rate": 3.20596744010508e-05, + "loss": 0.6105, + "step": 25045 + }, + { + "epoch": 0.75, + "learning_rate": 3.2052599455141976e-05, + "loss": 0.4295, + "step": 25046 + }, + { + "epoch": 0.75, + "learning_rate": 3.204552514098102e-05, + "loss": 0.5664, + "step": 25047 + }, + { + "epoch": 0.75, + "learning_rate": 3.203845145863367e-05, + "loss": 0.4763, + "step": 25048 + }, + { + "epoch": 0.75, + "learning_rate": 3.2031378408165736e-05, + "loss": 0.5054, + "step": 25049 + }, + { + "epoch": 0.75, + "learning_rate": 3.202430598964292e-05, + "loss": 0.4512, + "step": 25050 + }, + { + "epoch": 0.75, + "learning_rate": 3.201723420313105e-05, + "loss": 0.5407, + "step": 25051 + }, + { + "epoch": 0.75, + "learning_rate": 3.2010163048695815e-05, + "loss": 0.5939, + "step": 25052 + }, + { + "epoch": 0.75, + "learning_rate": 3.200309252640299e-05, + "loss": 0.5928, + "step": 25053 + }, + { + "epoch": 0.75, + "learning_rate": 3.199602263631833e-05, + "loss": 0.6245, + "step": 25054 + }, + { + "epoch": 0.75, + "learning_rate": 3.1988953378507555e-05, + "loss": 0.5804, + "step": 25055 + }, + { + "epoch": 0.75, + "learning_rate": 3.198188475303635e-05, + "loss": 0.5565, + "step": 25056 + }, + { + "epoch": 0.75, + "learning_rate": 3.197481675997047e-05, + "loss": 0.4572, + "step": 25057 + }, + { + "epoch": 0.75, + "learning_rate": 3.1967749399375665e-05, + "loss": 0.529, + "step": 25058 + }, + { + "epoch": 0.75, + "learning_rate": 3.196068267131761e-05, + "loss": 0.4836, + "step": 25059 + }, + { + "epoch": 0.75, + "learning_rate": 3.195361657586198e-05, + "loss": 0.4049, + "step": 25060 + }, + { + "epoch": 0.75, + "learning_rate": 3.1946551113074497e-05, + "loss": 0.6192, + "step": 25061 + }, + { + "epoch": 0.75, + "learning_rate": 3.193948628302087e-05, + "loss": 0.531, + "step": 25062 + }, + { + "epoch": 0.75, + "learning_rate": 3.193242208576679e-05, + "loss": 0.6293, + "step": 25063 + }, + { + "epoch": 0.75, + "learning_rate": 3.192535852137793e-05, + "loss": 0.5376, + "step": 25064 + }, + { + "epoch": 0.75, + "learning_rate": 3.191829558991992e-05, + "loss": 0.5229, + "step": 25065 + }, + { + "epoch": 0.75, + "learning_rate": 3.1911233291458464e-05, + "loss": 0.454, + "step": 25066 + }, + { + "epoch": 0.75, + "learning_rate": 3.1904171626059255e-05, + "loss": 0.528, + "step": 25067 + }, + { + "epoch": 0.75, + "learning_rate": 3.1897110593787915e-05, + "loss": 0.4944, + "step": 25068 + }, + { + "epoch": 0.75, + "learning_rate": 3.189005019471008e-05, + "loss": 0.4989, + "step": 25069 + }, + { + "epoch": 0.75, + "learning_rate": 3.188299042889139e-05, + "loss": 0.515, + "step": 25070 + }, + { + "epoch": 0.75, + "learning_rate": 3.187593129639755e-05, + "loss": 0.5412, + "step": 25071 + }, + { + "epoch": 0.75, + "learning_rate": 3.1868872797294115e-05, + "loss": 0.4974, + "step": 25072 + }, + { + "epoch": 0.75, + "learning_rate": 3.186181493164678e-05, + "loss": 0.4899, + "step": 25073 + }, + { + "epoch": 0.75, + "learning_rate": 3.185475769952113e-05, + "loss": 0.5006, + "step": 25074 + }, + { + "epoch": 0.75, + "learning_rate": 3.1847701100982766e-05, + "loss": 0.6009, + "step": 25075 + }, + { + "epoch": 0.75, + "learning_rate": 3.1840645136097294e-05, + "loss": 0.5797, + "step": 25076 + }, + { + "epoch": 0.75, + "learning_rate": 3.1833589804930344e-05, + "loss": 0.5749, + "step": 25077 + }, + { + "epoch": 0.75, + "learning_rate": 3.182653510754757e-05, + "loss": 0.4062, + "step": 25078 + }, + { + "epoch": 0.75, + "learning_rate": 3.1819481044014445e-05, + "loss": 0.4971, + "step": 25079 + }, + { + "epoch": 0.75, + "learning_rate": 3.18124276143966e-05, + "loss": 0.5549, + "step": 25080 + }, + { + "epoch": 0.75, + "learning_rate": 3.180537481875964e-05, + "loss": 0.5962, + "step": 25081 + }, + { + "epoch": 0.75, + "learning_rate": 3.179832265716915e-05, + "loss": 0.4994, + "step": 25082 + }, + { + "epoch": 0.75, + "learning_rate": 3.179127112969067e-05, + "loss": 0.5483, + "step": 25083 + }, + { + "epoch": 0.75, + "learning_rate": 3.178422023638974e-05, + "loss": 0.4359, + "step": 25084 + }, + { + "epoch": 0.75, + "learning_rate": 3.177716997733193e-05, + "loss": 0.6023, + "step": 25085 + }, + { + "epoch": 0.75, + "learning_rate": 3.1770120352582855e-05, + "loss": 0.5166, + "step": 25086 + }, + { + "epoch": 0.75, + "learning_rate": 3.1763071362207966e-05, + "loss": 0.5639, + "step": 25087 + }, + { + "epoch": 0.75, + "learning_rate": 3.1756023006272864e-05, + "loss": 0.5929, + "step": 25088 + }, + { + "epoch": 0.75, + "learning_rate": 3.174897528484304e-05, + "loss": 0.483, + "step": 25089 + }, + { + "epoch": 0.75, + "learning_rate": 3.174192819798407e-05, + "loss": 0.4847, + "step": 25090 + }, + { + "epoch": 0.75, + "learning_rate": 3.173488174576141e-05, + "loss": 0.5291, + "step": 25091 + }, + { + "epoch": 0.75, + "learning_rate": 3.172783592824066e-05, + "loss": 0.6386, + "step": 25092 + }, + { + "epoch": 0.75, + "learning_rate": 3.172079074548724e-05, + "loss": 0.4687, + "step": 25093 + }, + { + "epoch": 0.75, + "learning_rate": 3.1713746197566715e-05, + "loss": 0.5548, + "step": 25094 + }, + { + "epoch": 0.75, + "learning_rate": 3.170670228454454e-05, + "loss": 0.6023, + "step": 25095 + }, + { + "epoch": 0.75, + "learning_rate": 3.1699659006486225e-05, + "loss": 0.5277, + "step": 25096 + }, + { + "epoch": 0.75, + "learning_rate": 3.16926163634573e-05, + "loss": 0.5434, + "step": 25097 + }, + { + "epoch": 0.75, + "learning_rate": 3.16855743555232e-05, + "loss": 0.5777, + "step": 25098 + }, + { + "epoch": 0.75, + "learning_rate": 3.167853298274936e-05, + "loss": 0.3774, + "step": 25099 + }, + { + "epoch": 0.75, + "learning_rate": 3.1671492245201294e-05, + "loss": 0.5946, + "step": 25100 + }, + { + "epoch": 0.75, + "learning_rate": 3.16644521429445e-05, + "loss": 0.5252, + "step": 25101 + }, + { + "epoch": 0.75, + "learning_rate": 3.1657412676044384e-05, + "loss": 0.5233, + "step": 25102 + }, + { + "epoch": 0.75, + "learning_rate": 3.1650373844566384e-05, + "loss": 0.5321, + "step": 25103 + }, + { + "epoch": 0.75, + "learning_rate": 3.164333564857596e-05, + "loss": 0.5097, + "step": 25104 + }, + { + "epoch": 0.75, + "learning_rate": 3.163629808813856e-05, + "loss": 0.4715, + "step": 25105 + }, + { + "epoch": 0.75, + "learning_rate": 3.162926116331965e-05, + "loss": 0.4997, + "step": 25106 + }, + { + "epoch": 0.75, + "learning_rate": 3.1622224874184605e-05, + "loss": 0.621, + "step": 25107 + }, + { + "epoch": 0.75, + "learning_rate": 3.1615189220798835e-05, + "loss": 0.4459, + "step": 25108 + }, + { + "epoch": 0.75, + "learning_rate": 3.160815420322782e-05, + "loss": 0.5612, + "step": 25109 + }, + { + "epoch": 0.75, + "learning_rate": 3.160111982153689e-05, + "loss": 0.579, + "step": 25110 + }, + { + "epoch": 0.75, + "learning_rate": 3.159408607579152e-05, + "loss": 0.5064, + "step": 25111 + }, + { + "epoch": 0.75, + "learning_rate": 3.158705296605704e-05, + "loss": 0.5187, + "step": 25112 + }, + { + "epoch": 0.75, + "learning_rate": 3.158002049239891e-05, + "loss": 0.4813, + "step": 25113 + }, + { + "epoch": 0.75, + "learning_rate": 3.157298865488245e-05, + "loss": 0.4769, + "step": 25114 + }, + { + "epoch": 0.75, + "learning_rate": 3.1565957453573056e-05, + "loss": 0.5404, + "step": 25115 + }, + { + "epoch": 0.75, + "learning_rate": 3.1558926888536164e-05, + "loss": 0.4452, + "step": 25116 + }, + { + "epoch": 0.75, + "learning_rate": 3.155189695983708e-05, + "loss": 0.6678, + "step": 25117 + }, + { + "epoch": 0.75, + "learning_rate": 3.1544867667541157e-05, + "loss": 0.5401, + "step": 25118 + }, + { + "epoch": 0.75, + "learning_rate": 3.1537839011713756e-05, + "loss": 0.4842, + "step": 25119 + }, + { + "epoch": 0.75, + "learning_rate": 3.153081099242026e-05, + "loss": 0.592, + "step": 25120 + }, + { + "epoch": 0.75, + "learning_rate": 3.152378360972604e-05, + "loss": 0.5062, + "step": 25121 + }, + { + "epoch": 0.75, + "learning_rate": 3.151675686369633e-05, + "loss": 0.4759, + "step": 25122 + }, + { + "epoch": 0.75, + "learning_rate": 3.150973075439653e-05, + "loss": 0.5323, + "step": 25123 + }, + { + "epoch": 0.75, + "learning_rate": 3.150270528189194e-05, + "loss": 0.4294, + "step": 25124 + }, + { + "epoch": 0.75, + "learning_rate": 3.1495680446247934e-05, + "loss": 0.471, + "step": 25125 + }, + { + "epoch": 0.75, + "learning_rate": 3.148865624752979e-05, + "loss": 0.4708, + "step": 25126 + }, + { + "epoch": 0.75, + "learning_rate": 3.1481632685802775e-05, + "loss": 0.5473, + "step": 25127 + }, + { + "epoch": 0.75, + "learning_rate": 3.147460976113224e-05, + "loss": 0.621, + "step": 25128 + }, + { + "epoch": 0.75, + "learning_rate": 3.14675874735835e-05, + "loss": 0.5656, + "step": 25129 + }, + { + "epoch": 0.75, + "learning_rate": 3.1460565823221786e-05, + "loss": 0.4922, + "step": 25130 + }, + { + "epoch": 0.75, + "learning_rate": 3.145354481011245e-05, + "loss": 0.5946, + "step": 25131 + }, + { + "epoch": 0.75, + "learning_rate": 3.14465244343207e-05, + "loss": 0.5529, + "step": 25132 + }, + { + "epoch": 0.75, + "learning_rate": 3.143950469591187e-05, + "loss": 0.514, + "step": 25133 + }, + { + "epoch": 0.75, + "learning_rate": 3.143248559495118e-05, + "loss": 0.4767, + "step": 25134 + }, + { + "epoch": 0.75, + "learning_rate": 3.142546713150393e-05, + "loss": 0.498, + "step": 25135 + }, + { + "epoch": 0.75, + "learning_rate": 3.141844930563532e-05, + "loss": 0.4812, + "step": 25136 + }, + { + "epoch": 0.75, + "learning_rate": 3.1411432117410675e-05, + "loss": 0.4549, + "step": 25137 + }, + { + "epoch": 0.75, + "learning_rate": 3.140441556689517e-05, + "loss": 0.5463, + "step": 25138 + }, + { + "epoch": 0.75, + "learning_rate": 3.139739965415406e-05, + "loss": 0.5435, + "step": 25139 + }, + { + "epoch": 0.75, + "learning_rate": 3.1390384379252625e-05, + "loss": 0.4184, + "step": 25140 + }, + { + "epoch": 0.75, + "learning_rate": 3.138336974225605e-05, + "loss": 0.5496, + "step": 25141 + }, + { + "epoch": 0.75, + "learning_rate": 3.137635574322951e-05, + "loss": 0.5196, + "step": 25142 + }, + { + "epoch": 0.75, + "learning_rate": 3.136934238223829e-05, + "loss": 0.5233, + "step": 25143 + }, + { + "epoch": 0.75, + "learning_rate": 3.136232965934758e-05, + "loss": 0.4963, + "step": 25144 + }, + { + "epoch": 0.75, + "learning_rate": 3.135531757462258e-05, + "loss": 0.5167, + "step": 25145 + }, + { + "epoch": 0.75, + "learning_rate": 3.134830612812845e-05, + "loss": 0.5615, + "step": 25146 + }, + { + "epoch": 0.75, + "learning_rate": 3.13412953199304e-05, + "loss": 0.5628, + "step": 25147 + }, + { + "epoch": 0.75, + "learning_rate": 3.133428515009366e-05, + "loss": 0.5204, + "step": 25148 + }, + { + "epoch": 0.75, + "learning_rate": 3.1327275618683326e-05, + "loss": 0.4894, + "step": 25149 + }, + { + "epoch": 0.75, + "learning_rate": 3.1320266725764656e-05, + "loss": 0.4403, + "step": 25150 + }, + { + "epoch": 0.75, + "learning_rate": 3.131325847140274e-05, + "loss": 0.5552, + "step": 25151 + }, + { + "epoch": 0.75, + "learning_rate": 3.13062508556628e-05, + "loss": 0.5266, + "step": 25152 + }, + { + "epoch": 0.75, + "learning_rate": 3.129924387860994e-05, + "loss": 0.5659, + "step": 25153 + }, + { + "epoch": 0.75, + "learning_rate": 3.129223754030935e-05, + "loss": 0.3611, + "step": 25154 + }, + { + "epoch": 0.75, + "learning_rate": 3.1285231840826115e-05, + "loss": 0.5431, + "step": 25155 + }, + { + "epoch": 0.75, + "learning_rate": 3.127822678022545e-05, + "loss": 0.4845, + "step": 25156 + }, + { + "epoch": 0.75, + "learning_rate": 3.1271222358572404e-05, + "loss": 0.6217, + "step": 25157 + }, + { + "epoch": 0.75, + "learning_rate": 3.126421857593214e-05, + "loss": 0.5563, + "step": 25158 + }, + { + "epoch": 0.75, + "learning_rate": 3.1257215432369815e-05, + "loss": 0.5312, + "step": 25159 + }, + { + "epoch": 0.75, + "learning_rate": 3.125021292795049e-05, + "loss": 0.5518, + "step": 25160 + }, + { + "epoch": 0.75, + "learning_rate": 3.1243211062739266e-05, + "loss": 0.5279, + "step": 25161 + }, + { + "epoch": 0.75, + "learning_rate": 3.123620983680126e-05, + "loss": 0.4435, + "step": 25162 + }, + { + "epoch": 0.75, + "learning_rate": 3.1229209250201586e-05, + "loss": 0.4315, + "step": 25163 + }, + { + "epoch": 0.75, + "learning_rate": 3.122220930300532e-05, + "loss": 0.464, + "step": 25164 + }, + { + "epoch": 0.75, + "learning_rate": 3.121520999527752e-05, + "loss": 0.5655, + "step": 25165 + }, + { + "epoch": 0.75, + "learning_rate": 3.1208211327083284e-05, + "loss": 0.5846, + "step": 25166 + }, + { + "epoch": 0.75, + "learning_rate": 3.1201213298487666e-05, + "loss": 0.4529, + "step": 25167 + }, + { + "epoch": 0.75, + "learning_rate": 3.119421590955578e-05, + "loss": 0.5879, + "step": 25168 + }, + { + "epoch": 0.75, + "learning_rate": 3.1187219160352665e-05, + "loss": 0.4606, + "step": 25169 + }, + { + "epoch": 0.75, + "learning_rate": 3.118022305094331e-05, + "loss": 0.6416, + "step": 25170 + }, + { + "epoch": 0.75, + "learning_rate": 3.1173227581392826e-05, + "loss": 0.4919, + "step": 25171 + }, + { + "epoch": 0.75, + "learning_rate": 3.1166232751766265e-05, + "loss": 0.6029, + "step": 25172 + }, + { + "epoch": 0.75, + "learning_rate": 3.115923856212865e-05, + "loss": 0.5826, + "step": 25173 + }, + { + "epoch": 0.75, + "learning_rate": 3.1152245012544944e-05, + "loss": 0.6575, + "step": 25174 + }, + { + "epoch": 0.75, + "learning_rate": 3.114525210308027e-05, + "loss": 0.5206, + "step": 25175 + }, + { + "epoch": 0.75, + "learning_rate": 3.113825983379956e-05, + "loss": 0.5642, + "step": 25176 + }, + { + "epoch": 0.75, + "learning_rate": 3.113126820476788e-05, + "loss": 0.4787, + "step": 25177 + }, + { + "epoch": 0.75, + "learning_rate": 3.1124277216050234e-05, + "loss": 0.4999, + "step": 25178 + }, + { + "epoch": 0.75, + "learning_rate": 3.111728686771161e-05, + "loss": 0.5517, + "step": 25179 + }, + { + "epoch": 0.75, + "learning_rate": 3.111029715981698e-05, + "loss": 0.489, + "step": 25180 + }, + { + "epoch": 0.75, + "learning_rate": 3.110330809243134e-05, + "loss": 0.5459, + "step": 25181 + }, + { + "epoch": 0.75, + "learning_rate": 3.109631966561969e-05, + "loss": 0.4929, + "step": 25182 + }, + { + "epoch": 0.75, + "learning_rate": 3.108933187944705e-05, + "loss": 0.5152, + "step": 25183 + }, + { + "epoch": 0.75, + "learning_rate": 3.108234473397827e-05, + "loss": 0.4588, + "step": 25184 + }, + { + "epoch": 0.75, + "learning_rate": 3.10753582292784e-05, + "loss": 0.408, + "step": 25185 + }, + { + "epoch": 0.75, + "learning_rate": 3.106837236541237e-05, + "loss": 0.6165, + "step": 25186 + }, + { + "epoch": 0.75, + "learning_rate": 3.106138714244516e-05, + "loss": 0.4199, + "step": 25187 + }, + { + "epoch": 0.75, + "learning_rate": 3.105440256044171e-05, + "loss": 0.5886, + "step": 25188 + }, + { + "epoch": 0.75, + "learning_rate": 3.104741861946692e-05, + "loss": 0.5005, + "step": 25189 + }, + { + "epoch": 0.75, + "learning_rate": 3.104043531958574e-05, + "loss": 0.5904, + "step": 25190 + }, + { + "epoch": 0.75, + "learning_rate": 3.1033452660863124e-05, + "loss": 0.5348, + "step": 25191 + }, + { + "epoch": 0.75, + "learning_rate": 3.1026470643363966e-05, + "loss": 0.5311, + "step": 25192 + }, + { + "epoch": 0.75, + "learning_rate": 3.101948926715322e-05, + "loss": 0.5071, + "step": 25193 + }, + { + "epoch": 0.75, + "learning_rate": 3.101250853229575e-05, + "loss": 0.5278, + "step": 25194 + }, + { + "epoch": 0.75, + "learning_rate": 3.100552843885649e-05, + "loss": 0.6521, + "step": 25195 + }, + { + "epoch": 0.75, + "learning_rate": 3.099854898690031e-05, + "loss": 0.5017, + "step": 25196 + }, + { + "epoch": 0.75, + "learning_rate": 3.099157017649214e-05, + "loss": 0.5201, + "step": 25197 + }, + { + "epoch": 0.75, + "learning_rate": 3.098459200769683e-05, + "loss": 0.4484, + "step": 25198 + }, + { + "epoch": 0.75, + "learning_rate": 3.0977614480579306e-05, + "loss": 0.5038, + "step": 25199 + }, + { + "epoch": 0.75, + "learning_rate": 3.097063759520438e-05, + "loss": 0.4978, + "step": 25200 + }, + { + "epoch": 0.75, + "learning_rate": 3.096366135163695e-05, + "loss": 0.601, + "step": 25201 + }, + { + "epoch": 0.75, + "learning_rate": 3.095668574994192e-05, + "loss": 0.4999, + "step": 25202 + }, + { + "epoch": 0.75, + "learning_rate": 3.0949710790184094e-05, + "loss": 0.598, + "step": 25203 + }, + { + "epoch": 0.75, + "learning_rate": 3.094273647242831e-05, + "loss": 0.5459, + "step": 25204 + }, + { + "epoch": 0.75, + "learning_rate": 3.093576279673944e-05, + "loss": 0.415, + "step": 25205 + }, + { + "epoch": 0.75, + "learning_rate": 3.092878976318236e-05, + "loss": 0.6077, + "step": 25206 + }, + { + "epoch": 0.75, + "learning_rate": 3.092181737182186e-05, + "loss": 0.4358, + "step": 25207 + }, + { + "epoch": 0.75, + "learning_rate": 3.091484562272273e-05, + "loss": 0.4457, + "step": 25208 + }, + { + "epoch": 0.75, + "learning_rate": 3.0907874515949843e-05, + "loss": 0.4946, + "step": 25209 + }, + { + "epoch": 0.75, + "learning_rate": 3.090090405156802e-05, + "loss": 0.4202, + "step": 25210 + }, + { + "epoch": 0.75, + "learning_rate": 3.089393422964203e-05, + "loss": 0.5128, + "step": 25211 + }, + { + "epoch": 0.75, + "learning_rate": 3.0886965050236726e-05, + "loss": 0.5288, + "step": 25212 + }, + { + "epoch": 0.75, + "learning_rate": 3.087999651341684e-05, + "loss": 0.4865, + "step": 25213 + }, + { + "epoch": 0.75, + "learning_rate": 3.087302861924724e-05, + "loss": 0.5386, + "step": 25214 + }, + { + "epoch": 0.75, + "learning_rate": 3.086606136779263e-05, + "loss": 0.499, + "step": 25215 + }, + { + "epoch": 0.75, + "learning_rate": 3.0859094759117854e-05, + "loss": 0.538, + "step": 25216 + }, + { + "epoch": 0.75, + "learning_rate": 3.085212879328763e-05, + "loss": 0.4273, + "step": 25217 + }, + { + "epoch": 0.75, + "learning_rate": 3.084516347036679e-05, + "loss": 0.4716, + "step": 25218 + }, + { + "epoch": 0.75, + "learning_rate": 3.0838198790420016e-05, + "loss": 0.52, + "step": 25219 + }, + { + "epoch": 0.75, + "learning_rate": 3.0831234753512104e-05, + "loss": 0.4708, + "step": 25220 + }, + { + "epoch": 0.75, + "learning_rate": 3.0824271359707845e-05, + "loss": 0.4595, + "step": 25221 + }, + { + "epoch": 0.75, + "learning_rate": 3.081730860907194e-05, + "loss": 0.6198, + "step": 25222 + }, + { + "epoch": 0.75, + "learning_rate": 3.0810346501669084e-05, + "loss": 0.486, + "step": 25223 + }, + { + "epoch": 0.75, + "learning_rate": 3.080338503756406e-05, + "loss": 0.6149, + "step": 25224 + }, + { + "epoch": 0.75, + "learning_rate": 3.0796424216821584e-05, + "loss": 0.5473, + "step": 25225 + }, + { + "epoch": 0.75, + "learning_rate": 3.078946403950643e-05, + "loss": 0.5944, + "step": 25226 + }, + { + "epoch": 0.75, + "learning_rate": 3.07825045056832e-05, + "loss": 0.5452, + "step": 25227 + }, + { + "epoch": 0.75, + "learning_rate": 3.077554561541663e-05, + "loss": 0.4714, + "step": 25228 + }, + { + "epoch": 0.75, + "learning_rate": 3.0768587368771474e-05, + "loss": 0.5091, + "step": 25229 + }, + { + "epoch": 0.75, + "learning_rate": 3.076162976581243e-05, + "loss": 0.5526, + "step": 25230 + }, + { + "epoch": 0.75, + "learning_rate": 3.075467280660415e-05, + "loss": 0.5941, + "step": 25231 + }, + { + "epoch": 0.75, + "learning_rate": 3.074771649121129e-05, + "loss": 0.5149, + "step": 25232 + }, + { + "epoch": 0.75, + "learning_rate": 3.0740760819698556e-05, + "loss": 0.5147, + "step": 25233 + }, + { + "epoch": 0.75, + "learning_rate": 3.0733805792130664e-05, + "loss": 0.5399, + "step": 25234 + }, + { + "epoch": 0.75, + "learning_rate": 3.072685140857224e-05, + "loss": 0.5369, + "step": 25235 + }, + { + "epoch": 0.75, + "learning_rate": 3.07198976690879e-05, + "loss": 0.5115, + "step": 25236 + }, + { + "epoch": 0.75, + "learning_rate": 3.071294457374235e-05, + "loss": 0.4442, + "step": 25237 + }, + { + "epoch": 0.75, + "learning_rate": 3.070599212260026e-05, + "loss": 0.5149, + "step": 25238 + }, + { + "epoch": 0.75, + "learning_rate": 3.06990403157262e-05, + "loss": 0.5178, + "step": 25239 + }, + { + "epoch": 0.75, + "learning_rate": 3.069208915318488e-05, + "loss": 0.4597, + "step": 25240 + }, + { + "epoch": 0.75, + "learning_rate": 3.0685138635040874e-05, + "loss": 0.522, + "step": 25241 + }, + { + "epoch": 0.75, + "learning_rate": 3.067818876135882e-05, + "loss": 0.4795, + "step": 25242 + }, + { + "epoch": 0.75, + "learning_rate": 3.067123953220331e-05, + "loss": 0.5542, + "step": 25243 + }, + { + "epoch": 0.75, + "learning_rate": 3.0664290947639e-05, + "loss": 0.5335, + "step": 25244 + }, + { + "epoch": 0.75, + "learning_rate": 3.065734300773053e-05, + "loss": 0.3863, + "step": 25245 + }, + { + "epoch": 0.75, + "learning_rate": 3.065039571254239e-05, + "loss": 0.5489, + "step": 25246 + }, + { + "epoch": 0.75, + "learning_rate": 3.064344906213924e-05, + "loss": 0.5452, + "step": 25247 + }, + { + "epoch": 0.75, + "learning_rate": 3.063650305658564e-05, + "loss": 0.6383, + "step": 25248 + }, + { + "epoch": 0.75, + "learning_rate": 3.062955769594623e-05, + "loss": 0.501, + "step": 25249 + }, + { + "epoch": 0.75, + "learning_rate": 3.062261298028555e-05, + "loss": 0.6023, + "step": 25250 + }, + { + "epoch": 0.75, + "learning_rate": 3.0615668909668125e-05, + "loss": 0.5104, + "step": 25251 + }, + { + "epoch": 0.75, + "learning_rate": 3.060872548415855e-05, + "loss": 0.5559, + "step": 25252 + }, + { + "epoch": 0.75, + "learning_rate": 3.0601782703821434e-05, + "loss": 0.4953, + "step": 25253 + }, + { + "epoch": 0.75, + "learning_rate": 3.059484056872125e-05, + "loss": 0.5207, + "step": 25254 + }, + { + "epoch": 0.75, + "learning_rate": 3.05878990789226e-05, + "loss": 0.5127, + "step": 25255 + }, + { + "epoch": 0.75, + "learning_rate": 3.058095823448997e-05, + "loss": 0.4299, + "step": 25256 + }, + { + "epoch": 0.75, + "learning_rate": 3.057401803548795e-05, + "loss": 0.5459, + "step": 25257 + }, + { + "epoch": 0.75, + "learning_rate": 3.056707848198101e-05, + "loss": 0.5889, + "step": 25258 + }, + { + "epoch": 0.75, + "learning_rate": 3.056013957403374e-05, + "loss": 0.496, + "step": 25259 + }, + { + "epoch": 0.75, + "learning_rate": 3.055320131171058e-05, + "loss": 0.596, + "step": 25260 + }, + { + "epoch": 0.75, + "learning_rate": 3.054626369507612e-05, + "loss": 0.4767, + "step": 25261 + }, + { + "epoch": 0.75, + "learning_rate": 3.0539326724194764e-05, + "loss": 0.4606, + "step": 25262 + }, + { + "epoch": 0.75, + "learning_rate": 3.053239039913107e-05, + "loss": 0.4859, + "step": 25263 + }, + { + "epoch": 0.75, + "learning_rate": 3.0525454719949564e-05, + "loss": 0.5894, + "step": 25264 + }, + { + "epoch": 0.75, + "learning_rate": 3.0518519686714684e-05, + "loss": 0.5403, + "step": 25265 + }, + { + "epoch": 0.75, + "learning_rate": 3.0511585299490887e-05, + "loss": 0.5511, + "step": 25266 + }, + { + "epoch": 0.75, + "learning_rate": 3.0504651558342667e-05, + "loss": 0.4349, + "step": 25267 + }, + { + "epoch": 0.75, + "learning_rate": 3.0497718463334548e-05, + "loss": 0.573, + "step": 25268 + }, + { + "epoch": 0.75, + "learning_rate": 3.0490786014530926e-05, + "loss": 0.5622, + "step": 25269 + }, + { + "epoch": 0.75, + "learning_rate": 3.0483854211996242e-05, + "loss": 0.5317, + "step": 25270 + }, + { + "epoch": 0.75, + "learning_rate": 3.047692305579497e-05, + "loss": 0.4393, + "step": 25271 + }, + { + "epoch": 0.75, + "learning_rate": 3.04699925459916e-05, + "loss": 0.4222, + "step": 25272 + }, + { + "epoch": 0.75, + "learning_rate": 3.04630626826505e-05, + "loss": 0.5464, + "step": 25273 + }, + { + "epoch": 0.75, + "learning_rate": 3.0456133465836145e-05, + "loss": 0.5142, + "step": 25274 + }, + { + "epoch": 0.75, + "learning_rate": 3.0449204895612927e-05, + "loss": 0.5837, + "step": 25275 + }, + { + "epoch": 0.75, + "learning_rate": 3.0442276972045303e-05, + "loss": 0.4777, + "step": 25276 + }, + { + "epoch": 0.75, + "learning_rate": 3.0435349695197633e-05, + "loss": 0.5626, + "step": 25277 + }, + { + "epoch": 0.75, + "learning_rate": 3.0428423065134392e-05, + "loss": 0.5591, + "step": 25278 + }, + { + "epoch": 0.75, + "learning_rate": 3.0421497081919912e-05, + "loss": 0.5764, + "step": 25279 + }, + { + "epoch": 0.75, + "learning_rate": 3.0414571745618657e-05, + "loss": 0.51, + "step": 25280 + }, + { + "epoch": 0.75, + "learning_rate": 3.040764705629494e-05, + "loss": 0.5138, + "step": 25281 + }, + { + "epoch": 0.75, + "learning_rate": 3.040072301401319e-05, + "loss": 0.5434, + "step": 25282 + }, + { + "epoch": 0.75, + "learning_rate": 3.0393799618837803e-05, + "loss": 0.5716, + "step": 25283 + }, + { + "epoch": 0.75, + "learning_rate": 3.0386876870833136e-05, + "loss": 0.4149, + "step": 25284 + }, + { + "epoch": 0.75, + "learning_rate": 3.0379954770063502e-05, + "loss": 0.5516, + "step": 25285 + }, + { + "epoch": 0.75, + "learning_rate": 3.03730333165933e-05, + "loss": 0.5942, + "step": 25286 + }, + { + "epoch": 0.75, + "learning_rate": 3.036611251048689e-05, + "loss": 0.5378, + "step": 25287 + }, + { + "epoch": 0.75, + "learning_rate": 3.0359192351808674e-05, + "loss": 0.5746, + "step": 25288 + }, + { + "epoch": 0.75, + "learning_rate": 3.0352272840622864e-05, + "loss": 0.5988, + "step": 25289 + }, + { + "epoch": 0.75, + "learning_rate": 3.0345353976993864e-05, + "loss": 0.6839, + "step": 25290 + }, + { + "epoch": 0.75, + "learning_rate": 3.0338435760985994e-05, + "loss": 0.4798, + "step": 25291 + }, + { + "epoch": 0.75, + "learning_rate": 3.0331518192663632e-05, + "loss": 0.4972, + "step": 25292 + }, + { + "epoch": 0.75, + "learning_rate": 3.0324601272091035e-05, + "loss": 0.5246, + "step": 25293 + }, + { + "epoch": 0.75, + "learning_rate": 3.0317684999332496e-05, + "loss": 0.5436, + "step": 25294 + }, + { + "epoch": 0.75, + "learning_rate": 3.031076937445235e-05, + "loss": 0.5753, + "step": 25295 + }, + { + "epoch": 0.75, + "learning_rate": 3.030385439751493e-05, + "loss": 0.4908, + "step": 25296 + }, + { + "epoch": 0.75, + "learning_rate": 3.0296940068584455e-05, + "loss": 0.6043, + "step": 25297 + }, + { + "epoch": 0.75, + "learning_rate": 3.029002638772529e-05, + "loss": 0.5358, + "step": 25298 + }, + { + "epoch": 0.75, + "learning_rate": 3.0283113355001636e-05, + "loss": 0.5965, + "step": 25299 + }, + { + "epoch": 0.75, + "learning_rate": 3.027620097047784e-05, + "loss": 0.5823, + "step": 25300 + }, + { + "epoch": 0.75, + "learning_rate": 3.0269289234218113e-05, + "loss": 0.4771, + "step": 25301 + }, + { + "epoch": 0.75, + "learning_rate": 3.0262378146286774e-05, + "loss": 0.4724, + "step": 25302 + }, + { + "epoch": 0.75, + "learning_rate": 3.025546770674802e-05, + "loss": 0.6033, + "step": 25303 + }, + { + "epoch": 0.75, + "learning_rate": 3.0248557915666153e-05, + "loss": 0.519, + "step": 25304 + }, + { + "epoch": 0.75, + "learning_rate": 3.024164877310537e-05, + "loss": 0.5596, + "step": 25305 + }, + { + "epoch": 0.75, + "learning_rate": 3.0234740279129926e-05, + "loss": 0.4678, + "step": 25306 + }, + { + "epoch": 0.75, + "learning_rate": 3.022783243380409e-05, + "loss": 0.6864, + "step": 25307 + }, + { + "epoch": 0.75, + "learning_rate": 3.0220925237192067e-05, + "loss": 0.484, + "step": 25308 + }, + { + "epoch": 0.75, + "learning_rate": 3.0214018689358037e-05, + "loss": 0.5583, + "step": 25309 + }, + { + "epoch": 0.75, + "learning_rate": 3.0207112790366244e-05, + "loss": 0.4393, + "step": 25310 + }, + { + "epoch": 0.75, + "learning_rate": 3.0200207540280933e-05, + "loss": 0.436, + "step": 25311 + }, + { + "epoch": 0.75, + "learning_rate": 3.019330293916628e-05, + "loss": 0.4665, + "step": 25312 + }, + { + "epoch": 0.75, + "learning_rate": 3.018639898708644e-05, + "loss": 0.5135, + "step": 25313 + }, + { + "epoch": 0.75, + "learning_rate": 3.017949568410563e-05, + "loss": 0.4907, + "step": 25314 + }, + { + "epoch": 0.75, + "learning_rate": 3.0172593030288076e-05, + "loss": 0.5161, + "step": 25315 + }, + { + "epoch": 0.75, + "learning_rate": 3.0165691025697896e-05, + "loss": 0.5816, + "step": 25316 + }, + { + "epoch": 0.75, + "learning_rate": 3.0158789670399313e-05, + "loss": 0.5099, + "step": 25317 + }, + { + "epoch": 0.75, + "learning_rate": 3.0151888964456432e-05, + "loss": 0.4506, + "step": 25318 + }, + { + "epoch": 0.75, + "learning_rate": 3.0144988907933492e-05, + "loss": 0.4134, + "step": 25319 + }, + { + "epoch": 0.75, + "learning_rate": 3.013808950089456e-05, + "loss": 0.5518, + "step": 25320 + }, + { + "epoch": 0.75, + "learning_rate": 3.013119074340387e-05, + "loss": 0.6163, + "step": 25321 + }, + { + "epoch": 0.75, + "learning_rate": 3.0124292635525487e-05, + "loss": 0.6303, + "step": 25322 + }, + { + "epoch": 0.75, + "learning_rate": 3.0117395177323603e-05, + "loss": 0.4584, + "step": 25323 + }, + { + "epoch": 0.75, + "learning_rate": 3.0110498368862307e-05, + "loss": 0.6654, + "step": 25324 + }, + { + "epoch": 0.75, + "learning_rate": 3.010360221020574e-05, + "loss": 0.4628, + "step": 25325 + }, + { + "epoch": 0.75, + "learning_rate": 3.0096706701418042e-05, + "loss": 0.551, + "step": 25326 + }, + { + "epoch": 0.75, + "learning_rate": 3.008981184256332e-05, + "loss": 0.5428, + "step": 25327 + }, + { + "epoch": 0.75, + "learning_rate": 3.0082917633705622e-05, + "loss": 0.583, + "step": 25328 + }, + { + "epoch": 0.75, + "learning_rate": 3.0076024074909094e-05, + "loss": 0.5852, + "step": 25329 + }, + { + "epoch": 0.75, + "learning_rate": 3.006913116623785e-05, + "loss": 0.6022, + "step": 25330 + }, + { + "epoch": 0.75, + "learning_rate": 3.006223890775596e-05, + "loss": 0.5221, + "step": 25331 + }, + { + "epoch": 0.75, + "learning_rate": 3.005534729952747e-05, + "loss": 0.4355, + "step": 25332 + }, + { + "epoch": 0.75, + "learning_rate": 3.004845634161647e-05, + "loss": 0.52, + "step": 25333 + }, + { + "epoch": 0.75, + "learning_rate": 3.0041566034087053e-05, + "loss": 0.4429, + "step": 25334 + }, + { + "epoch": 0.75, + "learning_rate": 3.0034676377003313e-05, + "loss": 0.5204, + "step": 25335 + }, + { + "epoch": 0.75, + "learning_rate": 3.002778737042925e-05, + "loss": 0.6247, + "step": 25336 + }, + { + "epoch": 0.75, + "learning_rate": 3.0020899014428904e-05, + "loss": 0.4332, + "step": 25337 + }, + { + "epoch": 0.75, + "learning_rate": 3.0014011309066382e-05, + "loss": 0.521, + "step": 25338 + }, + { + "epoch": 0.75, + "learning_rate": 3.000712425440566e-05, + "loss": 0.5749, + "step": 25339 + }, + { + "epoch": 0.75, + "learning_rate": 3.0000237850510826e-05, + "loss": 0.4168, + "step": 25340 + }, + { + "epoch": 0.75, + "learning_rate": 2.999335209744586e-05, + "loss": 0.5301, + "step": 25341 + }, + { + "epoch": 0.75, + "learning_rate": 2.998646699527483e-05, + "loss": 0.5045, + "step": 25342 + }, + { + "epoch": 0.75, + "learning_rate": 2.997958254406169e-05, + "loss": 0.5494, + "step": 25343 + }, + { + "epoch": 0.75, + "learning_rate": 2.9972698743870487e-05, + "loss": 0.4285, + "step": 25344 + }, + { + "epoch": 0.75, + "learning_rate": 2.9965815594765256e-05, + "loss": 0.5813, + "step": 25345 + }, + { + "epoch": 0.75, + "learning_rate": 2.995893309680995e-05, + "loss": 0.5783, + "step": 25346 + }, + { + "epoch": 0.75, + "learning_rate": 2.9952051250068548e-05, + "loss": 0.5201, + "step": 25347 + }, + { + "epoch": 0.75, + "learning_rate": 2.9945170054605043e-05, + "loss": 0.4915, + "step": 25348 + }, + { + "epoch": 0.75, + "learning_rate": 2.993828951048343e-05, + "loss": 0.489, + "step": 25349 + }, + { + "epoch": 0.75, + "learning_rate": 2.9931409617767746e-05, + "loss": 0.5272, + "step": 25350 + }, + { + "epoch": 0.75, + "learning_rate": 2.992453037652182e-05, + "loss": 0.44, + "step": 25351 + }, + { + "epoch": 0.75, + "learning_rate": 2.9917651786809687e-05, + "loss": 0.5759, + "step": 25352 + }, + { + "epoch": 0.75, + "learning_rate": 2.9910773848695295e-05, + "loss": 0.4582, + "step": 25353 + }, + { + "epoch": 0.75, + "learning_rate": 2.9903896562242617e-05, + "loss": 0.5513, + "step": 25354 + }, + { + "epoch": 0.75, + "learning_rate": 2.989701992751558e-05, + "loss": 0.5144, + "step": 25355 + }, + { + "epoch": 0.75, + "learning_rate": 2.9890143944578076e-05, + "loss": 0.5448, + "step": 25356 + }, + { + "epoch": 0.75, + "learning_rate": 2.9883268613494063e-05, + "loss": 0.5639, + "step": 25357 + }, + { + "epoch": 0.75, + "learning_rate": 2.9876393934327517e-05, + "loss": 0.5453, + "step": 25358 + }, + { + "epoch": 0.75, + "learning_rate": 2.9869519907142272e-05, + "loss": 0.4961, + "step": 25359 + }, + { + "epoch": 0.76, + "learning_rate": 2.9862646532002316e-05, + "loss": 0.5605, + "step": 25360 + }, + { + "epoch": 0.76, + "learning_rate": 2.9855773808971487e-05, + "loss": 0.5686, + "step": 25361 + }, + { + "epoch": 0.76, + "learning_rate": 2.9848901738113744e-05, + "loss": 0.5315, + "step": 25362 + }, + { + "epoch": 0.76, + "learning_rate": 2.984203031949292e-05, + "loss": 0.5041, + "step": 25363 + }, + { + "epoch": 0.76, + "learning_rate": 2.9835159553172976e-05, + "loss": 0.519, + "step": 25364 + }, + { + "epoch": 0.76, + "learning_rate": 2.9828289439217717e-05, + "loss": 0.4483, + "step": 25365 + }, + { + "epoch": 0.76, + "learning_rate": 2.982141997769108e-05, + "loss": 0.5693, + "step": 25366 + }, + { + "epoch": 0.76, + "learning_rate": 2.981455116865689e-05, + "loss": 0.5591, + "step": 25367 + }, + { + "epoch": 0.76, + "learning_rate": 2.9807683012179022e-05, + "loss": 0.5492, + "step": 25368 + }, + { + "epoch": 0.76, + "learning_rate": 2.9800815508321377e-05, + "loss": 0.5498, + "step": 25369 + }, + { + "epoch": 0.76, + "learning_rate": 2.979394865714776e-05, + "loss": 0.5921, + "step": 25370 + }, + { + "epoch": 0.76, + "learning_rate": 2.9787082458722015e-05, + "loss": 0.532, + "step": 25371 + }, + { + "epoch": 0.76, + "learning_rate": 2.9780216913107982e-05, + "loss": 0.445, + "step": 25372 + }, + { + "epoch": 0.76, + "learning_rate": 2.9773352020369537e-05, + "loss": 0.4738, + "step": 25373 + }, + { + "epoch": 0.76, + "learning_rate": 2.9766487780570475e-05, + "loss": 0.5811, + "step": 25374 + }, + { + "epoch": 0.76, + "learning_rate": 2.9759624193774583e-05, + "loss": 0.5826, + "step": 25375 + }, + { + "epoch": 0.76, + "learning_rate": 2.9752761260045712e-05, + "loss": 0.5995, + "step": 25376 + }, + { + "epoch": 0.76, + "learning_rate": 2.9745898979447708e-05, + "loss": 0.4921, + "step": 25377 + }, + { + "epoch": 0.76, + "learning_rate": 2.9739037352044297e-05, + "loss": 0.5625, + "step": 25378 + }, + { + "epoch": 0.76, + "learning_rate": 2.9732176377899346e-05, + "loss": 0.5023, + "step": 25379 + }, + { + "epoch": 0.76, + "learning_rate": 2.972531605707658e-05, + "loss": 0.5047, + "step": 25380 + }, + { + "epoch": 0.76, + "learning_rate": 2.9718456389639848e-05, + "loss": 0.5911, + "step": 25381 + }, + { + "epoch": 0.76, + "learning_rate": 2.9711597375652868e-05, + "loss": 0.5655, + "step": 25382 + }, + { + "epoch": 0.76, + "learning_rate": 2.970473901517947e-05, + "loss": 0.5724, + "step": 25383 + }, + { + "epoch": 0.76, + "learning_rate": 2.969788130828337e-05, + "loss": 0.5242, + "step": 25384 + }, + { + "epoch": 0.76, + "learning_rate": 2.9691024255028366e-05, + "loss": 0.4418, + "step": 25385 + }, + { + "epoch": 0.76, + "learning_rate": 2.9684167855478173e-05, + "loss": 0.522, + "step": 25386 + }, + { + "epoch": 0.76, + "learning_rate": 2.9677312109696565e-05, + "loss": 0.4903, + "step": 25387 + }, + { + "epoch": 0.76, + "learning_rate": 2.967045701774731e-05, + "loss": 0.6862, + "step": 25388 + }, + { + "epoch": 0.76, + "learning_rate": 2.9663602579694117e-05, + "loss": 0.6023, + "step": 25389 + }, + { + "epoch": 0.76, + "learning_rate": 2.9656748795600675e-05, + "loss": 0.6092, + "step": 25390 + }, + { + "epoch": 0.76, + "learning_rate": 2.9649895665530746e-05, + "loss": 0.4585, + "step": 25391 + }, + { + "epoch": 0.76, + "learning_rate": 2.964304318954806e-05, + "loss": 0.4307, + "step": 25392 + }, + { + "epoch": 0.76, + "learning_rate": 2.963619136771637e-05, + "loss": 0.506, + "step": 25393 + }, + { + "epoch": 0.76, + "learning_rate": 2.9629340200099266e-05, + "loss": 0.5696, + "step": 25394 + }, + { + "epoch": 0.76, + "learning_rate": 2.9622489686760513e-05, + "loss": 0.5073, + "step": 25395 + }, + { + "epoch": 0.76, + "learning_rate": 2.9615639827763806e-05, + "loss": 0.6327, + "step": 25396 + }, + { + "epoch": 0.76, + "learning_rate": 2.9608790623172855e-05, + "loss": 0.602, + "step": 25397 + }, + { + "epoch": 0.76, + "learning_rate": 2.9601942073051303e-05, + "loss": 0.4597, + "step": 25398 + }, + { + "epoch": 0.76, + "learning_rate": 2.9595094177462822e-05, + "loss": 0.422, + "step": 25399 + }, + { + "epoch": 0.76, + "learning_rate": 2.9588246936471087e-05, + "loss": 0.568, + "step": 25400 + }, + { + "epoch": 0.76, + "learning_rate": 2.95814003501398e-05, + "loss": 0.5354, + "step": 25401 + }, + { + "epoch": 0.76, + "learning_rate": 2.9574554418532587e-05, + "loss": 0.5208, + "step": 25402 + }, + { + "epoch": 0.76, + "learning_rate": 2.9567709141713064e-05, + "loss": 0.4612, + "step": 25403 + }, + { + "epoch": 0.76, + "learning_rate": 2.9560864519744934e-05, + "loss": 0.5535, + "step": 25404 + }, + { + "epoch": 0.76, + "learning_rate": 2.9554020552691797e-05, + "loss": 0.5352, + "step": 25405 + }, + { + "epoch": 0.76, + "learning_rate": 2.9547177240617284e-05, + "loss": 0.5054, + "step": 25406 + }, + { + "epoch": 0.76, + "learning_rate": 2.9540334583585083e-05, + "loss": 0.5172, + "step": 25407 + }, + { + "epoch": 0.76, + "learning_rate": 2.953349258165876e-05, + "loss": 0.4722, + "step": 25408 + }, + { + "epoch": 0.76, + "learning_rate": 2.9526651234901903e-05, + "loss": 0.5928, + "step": 25409 + }, + { + "epoch": 0.76, + "learning_rate": 2.9519810543378167e-05, + "loss": 0.4152, + "step": 25410 + }, + { + "epoch": 0.76, + "learning_rate": 2.9512970507151126e-05, + "loss": 0.4857, + "step": 25411 + }, + { + "epoch": 0.76, + "learning_rate": 2.9506131126284474e-05, + "loss": 0.3827, + "step": 25412 + }, + { + "epoch": 0.76, + "learning_rate": 2.9499292400841638e-05, + "loss": 0.5471, + "step": 25413 + }, + { + "epoch": 0.76, + "learning_rate": 2.94924543308863e-05, + "loss": 0.5475, + "step": 25414 + }, + { + "epoch": 0.76, + "learning_rate": 2.9485616916482007e-05, + "loss": 0.6181, + "step": 25415 + }, + { + "epoch": 0.76, + "learning_rate": 2.9478780157692376e-05, + "loss": 0.487, + "step": 25416 + }, + { + "epoch": 0.76, + "learning_rate": 2.947194405458095e-05, + "loss": 0.598, + "step": 25417 + }, + { + "epoch": 0.76, + "learning_rate": 2.9465108607211234e-05, + "loss": 0.5532, + "step": 25418 + }, + { + "epoch": 0.76, + "learning_rate": 2.945827381564683e-05, + "loss": 0.5932, + "step": 25419 + }, + { + "epoch": 0.76, + "learning_rate": 2.9451439679951308e-05, + "loss": 0.5246, + "step": 25420 + }, + { + "epoch": 0.76, + "learning_rate": 2.9444606200188162e-05, + "loss": 0.5716, + "step": 25421 + }, + { + "epoch": 0.76, + "learning_rate": 2.9437773376420975e-05, + "loss": 0.5028, + "step": 25422 + }, + { + "epoch": 0.76, + "learning_rate": 2.9430941208713213e-05, + "loss": 0.4104, + "step": 25423 + }, + { + "epoch": 0.76, + "learning_rate": 2.9424109697128464e-05, + "loss": 0.4992, + "step": 25424 + }, + { + "epoch": 0.76, + "learning_rate": 2.941727884173019e-05, + "loss": 0.6383, + "step": 25425 + }, + { + "epoch": 0.76, + "learning_rate": 2.9410448642581945e-05, + "loss": 0.4256, + "step": 25426 + }, + { + "epoch": 0.76, + "learning_rate": 2.9403619099747192e-05, + "loss": 0.5135, + "step": 25427 + }, + { + "epoch": 0.76, + "learning_rate": 2.939679021328948e-05, + "loss": 0.5828, + "step": 25428 + }, + { + "epoch": 0.76, + "learning_rate": 2.9389961983272242e-05, + "loss": 0.4771, + "step": 25429 + }, + { + "epoch": 0.76, + "learning_rate": 2.9383134409758995e-05, + "loss": 0.5592, + "step": 25430 + }, + { + "epoch": 0.76, + "learning_rate": 2.937630749281325e-05, + "loss": 0.5972, + "step": 25431 + }, + { + "epoch": 0.76, + "learning_rate": 2.9369481232498454e-05, + "loss": 0.5617, + "step": 25432 + }, + { + "epoch": 0.76, + "learning_rate": 2.9362655628878032e-05, + "loss": 0.5328, + "step": 25433 + }, + { + "epoch": 0.76, + "learning_rate": 2.9355830682015496e-05, + "loss": 0.458, + "step": 25434 + }, + { + "epoch": 0.76, + "learning_rate": 2.934900639197431e-05, + "loss": 0.4365, + "step": 25435 + }, + { + "epoch": 0.76, + "learning_rate": 2.9342182758817914e-05, + "loss": 0.5318, + "step": 25436 + }, + { + "epoch": 0.76, + "learning_rate": 2.9335359782609707e-05, + "loss": 0.5498, + "step": 25437 + }, + { + "epoch": 0.76, + "learning_rate": 2.9328537463413152e-05, + "loss": 0.449, + "step": 25438 + }, + { + "epoch": 0.76, + "learning_rate": 2.9321715801291727e-05, + "loss": 0.4826, + "step": 25439 + }, + { + "epoch": 0.76, + "learning_rate": 2.931489479630879e-05, + "loss": 0.5453, + "step": 25440 + }, + { + "epoch": 0.76, + "learning_rate": 2.9308074448527822e-05, + "loss": 0.4577, + "step": 25441 + }, + { + "epoch": 0.76, + "learning_rate": 2.9301254758012164e-05, + "loss": 0.4808, + "step": 25442 + }, + { + "epoch": 0.76, + "learning_rate": 2.9294435724825298e-05, + "loss": 0.469, + "step": 25443 + }, + { + "epoch": 0.76, + "learning_rate": 2.9287617349030548e-05, + "loss": 0.522, + "step": 25444 + }, + { + "epoch": 0.76, + "learning_rate": 2.928079963069139e-05, + "loss": 0.4699, + "step": 25445 + }, + { + "epoch": 0.76, + "learning_rate": 2.9273982569871138e-05, + "loss": 0.498, + "step": 25446 + }, + { + "epoch": 0.76, + "learning_rate": 2.926716616663323e-05, + "loss": 0.5303, + "step": 25447 + }, + { + "epoch": 0.76, + "learning_rate": 2.9260350421041005e-05, + "loss": 0.5298, + "step": 25448 + }, + { + "epoch": 0.76, + "learning_rate": 2.9253535333157834e-05, + "loss": 0.5744, + "step": 25449 + }, + { + "epoch": 0.76, + "learning_rate": 2.924672090304713e-05, + "loss": 0.5116, + "step": 25450 + }, + { + "epoch": 0.76, + "learning_rate": 2.9239907130772214e-05, + "loss": 0.5749, + "step": 25451 + }, + { + "epoch": 0.76, + "learning_rate": 2.9233094016396413e-05, + "loss": 0.434, + "step": 25452 + }, + { + "epoch": 0.76, + "learning_rate": 2.9226281559983093e-05, + "loss": 0.5806, + "step": 25453 + }, + { + "epoch": 0.76, + "learning_rate": 2.921946976159561e-05, + "loss": 0.5174, + "step": 25454 + }, + { + "epoch": 0.76, + "learning_rate": 2.921265862129734e-05, + "loss": 0.5878, + "step": 25455 + }, + { + "epoch": 0.76, + "learning_rate": 2.9205848139151493e-05, + "loss": 0.5452, + "step": 25456 + }, + { + "epoch": 0.76, + "learning_rate": 2.919903831522146e-05, + "loss": 0.5563, + "step": 25457 + }, + { + "epoch": 0.76, + "learning_rate": 2.919222914957055e-05, + "loss": 0.5167, + "step": 25458 + }, + { + "epoch": 0.76, + "learning_rate": 2.9185420642262097e-05, + "loss": 0.5903, + "step": 25459 + }, + { + "epoch": 0.76, + "learning_rate": 2.9178612793359385e-05, + "loss": 0.5564, + "step": 25460 + }, + { + "epoch": 0.76, + "learning_rate": 2.9171805602925673e-05, + "loss": 0.5039, + "step": 25461 + }, + { + "epoch": 0.76, + "learning_rate": 2.916499907102428e-05, + "loss": 0.4885, + "step": 25462 + }, + { + "epoch": 0.76, + "learning_rate": 2.915819319771853e-05, + "loss": 0.5451, + "step": 25463 + }, + { + "epoch": 0.76, + "learning_rate": 2.915138798307162e-05, + "loss": 0.6348, + "step": 25464 + }, + { + "epoch": 0.76, + "learning_rate": 2.914458342714691e-05, + "loss": 0.507, + "step": 25465 + }, + { + "epoch": 0.76, + "learning_rate": 2.913777953000759e-05, + "loss": 0.6341, + "step": 25466 + }, + { + "epoch": 0.76, + "learning_rate": 2.9130976291716992e-05, + "loss": 0.7273, + "step": 25467 + }, + { + "epoch": 0.76, + "learning_rate": 2.9124173712338287e-05, + "loss": 0.4095, + "step": 25468 + }, + { + "epoch": 0.76, + "learning_rate": 2.911737179193481e-05, + "loss": 0.6083, + "step": 25469 + }, + { + "epoch": 0.76, + "learning_rate": 2.9110570530569746e-05, + "loss": 0.4885, + "step": 25470 + }, + { + "epoch": 0.76, + "learning_rate": 2.9103769928306324e-05, + "loss": 0.474, + "step": 25471 + }, + { + "epoch": 0.76, + "learning_rate": 2.909696998520778e-05, + "loss": 0.5529, + "step": 25472 + }, + { + "epoch": 0.76, + "learning_rate": 2.909017070133736e-05, + "loss": 0.5655, + "step": 25473 + }, + { + "epoch": 0.76, + "learning_rate": 2.90833720767583e-05, + "loss": 0.4742, + "step": 25474 + }, + { + "epoch": 0.76, + "learning_rate": 2.9076574111533783e-05, + "loss": 0.6067, + "step": 25475 + }, + { + "epoch": 0.76, + "learning_rate": 2.9069776805726977e-05, + "loss": 0.5919, + "step": 25476 + }, + { + "epoch": 0.76, + "learning_rate": 2.906298015940112e-05, + "loss": 0.3811, + "step": 25477 + }, + { + "epoch": 0.76, + "learning_rate": 2.9056184172619427e-05, + "loss": 0.56, + "step": 25478 + }, + { + "epoch": 0.76, + "learning_rate": 2.904938884544507e-05, + "loss": 0.5244, + "step": 25479 + }, + { + "epoch": 0.76, + "learning_rate": 2.9042594177941174e-05, + "loss": 0.6191, + "step": 25480 + }, + { + "epoch": 0.76, + "learning_rate": 2.9035800170170956e-05, + "loss": 0.5241, + "step": 25481 + }, + { + "epoch": 0.76, + "learning_rate": 2.9029006822197625e-05, + "loss": 0.6118, + "step": 25482 + }, + { + "epoch": 0.76, + "learning_rate": 2.9022214134084268e-05, + "loss": 0.6076, + "step": 25483 + }, + { + "epoch": 0.76, + "learning_rate": 2.9015422105894098e-05, + "loss": 0.5962, + "step": 25484 + }, + { + "epoch": 0.76, + "learning_rate": 2.9008630737690224e-05, + "loss": 0.5734, + "step": 25485 + }, + { + "epoch": 0.76, + "learning_rate": 2.9001840029535832e-05, + "loss": 0.495, + "step": 25486 + }, + { + "epoch": 0.76, + "learning_rate": 2.8995049981493993e-05, + "loss": 0.4702, + "step": 25487 + }, + { + "epoch": 0.76, + "learning_rate": 2.8988260593627925e-05, + "loss": 0.5391, + "step": 25488 + }, + { + "epoch": 0.76, + "learning_rate": 2.8981471866000665e-05, + "loss": 0.5929, + "step": 25489 + }, + { + "epoch": 0.76, + "learning_rate": 2.897468379867542e-05, + "loss": 0.6089, + "step": 25490 + }, + { + "epoch": 0.76, + "learning_rate": 2.896789639171521e-05, + "loss": 0.4827, + "step": 25491 + }, + { + "epoch": 0.76, + "learning_rate": 2.8961109645183204e-05, + "loss": 0.5043, + "step": 25492 + }, + { + "epoch": 0.76, + "learning_rate": 2.8954323559142506e-05, + "loss": 0.5267, + "step": 25493 + }, + { + "epoch": 0.76, + "learning_rate": 2.8947538133656193e-05, + "loss": 0.5074, + "step": 25494 + }, + { + "epoch": 0.76, + "learning_rate": 2.894075336878733e-05, + "loss": 0.5502, + "step": 25495 + }, + { + "epoch": 0.76, + "learning_rate": 2.8933969264599015e-05, + "loss": 0.4239, + "step": 25496 + }, + { + "epoch": 0.76, + "learning_rate": 2.8927185821154358e-05, + "loss": 0.4888, + "step": 25497 + }, + { + "epoch": 0.76, + "learning_rate": 2.8920403038516408e-05, + "loss": 0.4861, + "step": 25498 + }, + { + "epoch": 0.76, + "learning_rate": 2.8913620916748186e-05, + "loss": 0.5534, + "step": 25499 + }, + { + "epoch": 0.76, + "learning_rate": 2.8906839455912783e-05, + "loss": 0.5562, + "step": 25500 + }, + { + "epoch": 0.76, + "learning_rate": 2.890005865607325e-05, + "loss": 0.6265, + "step": 25501 + }, + { + "epoch": 0.76, + "learning_rate": 2.8893278517292667e-05, + "loss": 0.509, + "step": 25502 + }, + { + "epoch": 0.76, + "learning_rate": 2.888649903963404e-05, + "loss": 0.6147, + "step": 25503 + }, + { + "epoch": 0.76, + "learning_rate": 2.8879720223160366e-05, + "loss": 0.5996, + "step": 25504 + }, + { + "epoch": 0.76, + "learning_rate": 2.887294206793475e-05, + "loss": 0.5259, + "step": 25505 + }, + { + "epoch": 0.76, + "learning_rate": 2.8866164574020127e-05, + "loss": 0.5012, + "step": 25506 + }, + { + "epoch": 0.76, + "learning_rate": 2.8859387741479592e-05, + "loss": 0.408, + "step": 25507 + }, + { + "epoch": 0.76, + "learning_rate": 2.885261157037609e-05, + "loss": 0.5223, + "step": 25508 + }, + { + "epoch": 0.76, + "learning_rate": 2.884583606077267e-05, + "loss": 0.5391, + "step": 25509 + }, + { + "epoch": 0.76, + "learning_rate": 2.8839061212732287e-05, + "loss": 0.5505, + "step": 25510 + }, + { + "epoch": 0.76, + "learning_rate": 2.883228702631794e-05, + "loss": 0.6129, + "step": 25511 + }, + { + "epoch": 0.76, + "learning_rate": 2.882551350159266e-05, + "loss": 0.4843, + "step": 25512 + }, + { + "epoch": 0.76, + "learning_rate": 2.881874063861939e-05, + "loss": 0.535, + "step": 25513 + }, + { + "epoch": 0.76, + "learning_rate": 2.881196843746107e-05, + "loss": 0.4987, + "step": 25514 + }, + { + "epoch": 0.76, + "learning_rate": 2.8805196898180685e-05, + "loss": 0.5165, + "step": 25515 + }, + { + "epoch": 0.76, + "learning_rate": 2.8798426020841208e-05, + "loss": 0.6681, + "step": 25516 + }, + { + "epoch": 0.76, + "learning_rate": 2.8791655805505647e-05, + "loss": 0.402, + "step": 25517 + }, + { + "epoch": 0.76, + "learning_rate": 2.878488625223683e-05, + "loss": 0.5605, + "step": 25518 + }, + { + "epoch": 0.76, + "learning_rate": 2.8778117361097767e-05, + "loss": 0.4646, + "step": 25519 + }, + { + "epoch": 0.76, + "learning_rate": 2.877134913215137e-05, + "loss": 0.4438, + "step": 25520 + }, + { + "epoch": 0.76, + "learning_rate": 2.8764581565460612e-05, + "loss": 0.521, + "step": 25521 + }, + { + "epoch": 0.76, + "learning_rate": 2.8757814661088388e-05, + "loss": 0.5508, + "step": 25522 + }, + { + "epoch": 0.76, + "learning_rate": 2.875104841909758e-05, + "loss": 0.5746, + "step": 25523 + }, + { + "epoch": 0.76, + "learning_rate": 2.8744282839551118e-05, + "loss": 0.5178, + "step": 25524 + }, + { + "epoch": 0.76, + "learning_rate": 2.873751792251195e-05, + "loss": 0.4975, + "step": 25525 + }, + { + "epoch": 0.76, + "learning_rate": 2.873075366804291e-05, + "loss": 0.6577, + "step": 25526 + }, + { + "epoch": 0.76, + "learning_rate": 2.8723990076206942e-05, + "loss": 0.7341, + "step": 25527 + }, + { + "epoch": 0.76, + "learning_rate": 2.8717227147066882e-05, + "loss": 0.6085, + "step": 25528 + }, + { + "epoch": 0.76, + "learning_rate": 2.871046488068566e-05, + "loss": 0.5316, + "step": 25529 + }, + { + "epoch": 0.76, + "learning_rate": 2.8703703277126093e-05, + "loss": 0.5783, + "step": 25530 + }, + { + "epoch": 0.76, + "learning_rate": 2.8696942336451103e-05, + "loss": 0.5562, + "step": 25531 + }, + { + "epoch": 0.76, + "learning_rate": 2.8690182058723503e-05, + "loss": 0.5102, + "step": 25532 + }, + { + "epoch": 0.76, + "learning_rate": 2.868342244400619e-05, + "loss": 0.5796, + "step": 25533 + }, + { + "epoch": 0.76, + "learning_rate": 2.8676663492361965e-05, + "loss": 0.5711, + "step": 25534 + }, + { + "epoch": 0.76, + "learning_rate": 2.8669905203853697e-05, + "loss": 0.5215, + "step": 25535 + }, + { + "epoch": 0.76, + "learning_rate": 2.8663147578544257e-05, + "loss": 0.5547, + "step": 25536 + }, + { + "epoch": 0.76, + "learning_rate": 2.8656390616496432e-05, + "loss": 0.6131, + "step": 25537 + }, + { + "epoch": 0.76, + "learning_rate": 2.8649634317773023e-05, + "loss": 0.5543, + "step": 25538 + }, + { + "epoch": 0.76, + "learning_rate": 2.864287868243688e-05, + "loss": 0.4307, + "step": 25539 + }, + { + "epoch": 0.76, + "learning_rate": 2.863612371055083e-05, + "loss": 0.5245, + "step": 25540 + }, + { + "epoch": 0.76, + "learning_rate": 2.8629369402177673e-05, + "loss": 0.6061, + "step": 25541 + }, + { + "epoch": 0.76, + "learning_rate": 2.8622615757380155e-05, + "loss": 0.5859, + "step": 25542 + }, + { + "epoch": 0.76, + "learning_rate": 2.8615862776221102e-05, + "loss": 0.4801, + "step": 25543 + }, + { + "epoch": 0.76, + "learning_rate": 2.8609110458763345e-05, + "loss": 0.5064, + "step": 25544 + }, + { + "epoch": 0.76, + "learning_rate": 2.8602358805069594e-05, + "loss": 0.5051, + "step": 25545 + }, + { + "epoch": 0.76, + "learning_rate": 2.8595607815202686e-05, + "loss": 0.5747, + "step": 25546 + }, + { + "epoch": 0.76, + "learning_rate": 2.8588857489225328e-05, + "loss": 0.5513, + "step": 25547 + }, + { + "epoch": 0.76, + "learning_rate": 2.8582107827200343e-05, + "loss": 0.4772, + "step": 25548 + }, + { + "epoch": 0.76, + "learning_rate": 2.8575358829190425e-05, + "loss": 0.4929, + "step": 25549 + }, + { + "epoch": 0.76, + "learning_rate": 2.8568610495258376e-05, + "loss": 0.5369, + "step": 25550 + }, + { + "epoch": 0.76, + "learning_rate": 2.856186282546689e-05, + "loss": 0.6316, + "step": 25551 + }, + { + "epoch": 0.76, + "learning_rate": 2.8555115819878763e-05, + "loss": 0.6154, + "step": 25552 + }, + { + "epoch": 0.76, + "learning_rate": 2.854836947855666e-05, + "loss": 0.5562, + "step": 25553 + }, + { + "epoch": 0.76, + "learning_rate": 2.8541623801563345e-05, + "loss": 0.623, + "step": 25554 + }, + { + "epoch": 0.76, + "learning_rate": 2.853487878896156e-05, + "loss": 0.6371, + "step": 25555 + }, + { + "epoch": 0.76, + "learning_rate": 2.852813444081398e-05, + "loss": 0.4811, + "step": 25556 + }, + { + "epoch": 0.76, + "learning_rate": 2.8521390757183308e-05, + "loss": 0.596, + "step": 25557 + }, + { + "epoch": 0.76, + "learning_rate": 2.8514647738132237e-05, + "loss": 0.5261, + "step": 25558 + }, + { + "epoch": 0.76, + "learning_rate": 2.8507905383723488e-05, + "loss": 0.5576, + "step": 25559 + }, + { + "epoch": 0.76, + "learning_rate": 2.8501163694019806e-05, + "loss": 0.4364, + "step": 25560 + }, + { + "epoch": 0.76, + "learning_rate": 2.8494422669083742e-05, + "loss": 0.5376, + "step": 25561 + }, + { + "epoch": 0.76, + "learning_rate": 2.8487682308978035e-05, + "loss": 0.5771, + "step": 25562 + }, + { + "epoch": 0.76, + "learning_rate": 2.848094261376536e-05, + "loss": 0.5039, + "step": 25563 + }, + { + "epoch": 0.76, + "learning_rate": 2.8474203583508408e-05, + "loss": 0.5577, + "step": 25564 + }, + { + "epoch": 0.76, + "learning_rate": 2.8467465218269805e-05, + "loss": 0.6696, + "step": 25565 + }, + { + "epoch": 0.76, + "learning_rate": 2.8460727518112162e-05, + "loss": 0.56, + "step": 25566 + }, + { + "epoch": 0.76, + "learning_rate": 2.845399048309817e-05, + "loss": 0.6591, + "step": 25567 + }, + { + "epoch": 0.76, + "learning_rate": 2.844725411329048e-05, + "loss": 0.622, + "step": 25568 + }, + { + "epoch": 0.76, + "learning_rate": 2.8440518408751682e-05, + "loss": 0.4853, + "step": 25569 + }, + { + "epoch": 0.76, + "learning_rate": 2.843378336954444e-05, + "loss": 0.5537, + "step": 25570 + }, + { + "epoch": 0.76, + "learning_rate": 2.8427048995731376e-05, + "loss": 0.5726, + "step": 25571 + }, + { + "epoch": 0.76, + "learning_rate": 2.8420315287375044e-05, + "loss": 0.5879, + "step": 25572 + }, + { + "epoch": 0.76, + "learning_rate": 2.8413582244538082e-05, + "loss": 0.4988, + "step": 25573 + }, + { + "epoch": 0.76, + "learning_rate": 2.8406849867283148e-05, + "loss": 0.5646, + "step": 25574 + }, + { + "epoch": 0.76, + "learning_rate": 2.8400118155672782e-05, + "loss": 0.4931, + "step": 25575 + }, + { + "epoch": 0.76, + "learning_rate": 2.8393387109769554e-05, + "loss": 0.4321, + "step": 25576 + }, + { + "epoch": 0.76, + "learning_rate": 2.8386656729636065e-05, + "loss": 0.508, + "step": 25577 + }, + { + "epoch": 0.76, + "learning_rate": 2.8379927015334908e-05, + "loss": 0.5456, + "step": 25578 + }, + { + "epoch": 0.76, + "learning_rate": 2.83731979669287e-05, + "loss": 0.5017, + "step": 25579 + }, + { + "epoch": 0.76, + "learning_rate": 2.8366469584479895e-05, + "loss": 0.5189, + "step": 25580 + }, + { + "epoch": 0.76, + "learning_rate": 2.8359741868051092e-05, + "loss": 0.4198, + "step": 25581 + }, + { + "epoch": 0.76, + "learning_rate": 2.835301481770487e-05, + "loss": 0.4521, + "step": 25582 + }, + { + "epoch": 0.76, + "learning_rate": 2.8346288433503785e-05, + "loss": 0.5214, + "step": 25583 + }, + { + "epoch": 0.76, + "learning_rate": 2.833956271551035e-05, + "loss": 0.476, + "step": 25584 + }, + { + "epoch": 0.76, + "learning_rate": 2.833283766378707e-05, + "loss": 0.5505, + "step": 25585 + }, + { + "epoch": 0.76, + "learning_rate": 2.8326113278396503e-05, + "loss": 0.5854, + "step": 25586 + }, + { + "epoch": 0.76, + "learning_rate": 2.83193895594012e-05, + "loss": 0.6401, + "step": 25587 + }, + { + "epoch": 0.76, + "learning_rate": 2.8312666506863616e-05, + "loss": 0.5573, + "step": 25588 + }, + { + "epoch": 0.76, + "learning_rate": 2.8305944120846316e-05, + "loss": 0.4961, + "step": 25589 + }, + { + "epoch": 0.76, + "learning_rate": 2.829922240141174e-05, + "loss": 0.5391, + "step": 25590 + }, + { + "epoch": 0.76, + "learning_rate": 2.8292501348622447e-05, + "loss": 0.5002, + "step": 25591 + }, + { + "epoch": 0.76, + "learning_rate": 2.8285780962540863e-05, + "loss": 0.4916, + "step": 25592 + }, + { + "epoch": 0.76, + "learning_rate": 2.8279061243229544e-05, + "loss": 0.4655, + "step": 25593 + }, + { + "epoch": 0.76, + "learning_rate": 2.82723421907509e-05, + "loss": 0.526, + "step": 25594 + }, + { + "epoch": 0.76, + "learning_rate": 2.826562380516745e-05, + "loss": 0.6272, + "step": 25595 + }, + { + "epoch": 0.76, + "learning_rate": 2.8258906086541615e-05, + "loss": 0.5836, + "step": 25596 + }, + { + "epoch": 0.76, + "learning_rate": 2.8252189034935884e-05, + "loss": 0.4428, + "step": 25597 + }, + { + "epoch": 0.76, + "learning_rate": 2.824547265041272e-05, + "loss": 0.5494, + "step": 25598 + }, + { + "epoch": 0.76, + "learning_rate": 2.823875693303456e-05, + "loss": 0.5365, + "step": 25599 + }, + { + "epoch": 0.76, + "learning_rate": 2.8232041882863802e-05, + "loss": 0.5807, + "step": 25600 + }, + { + "epoch": 0.76, + "learning_rate": 2.8225327499962928e-05, + "loss": 0.4199, + "step": 25601 + }, + { + "epoch": 0.76, + "learning_rate": 2.8218613784394366e-05, + "loss": 0.5385, + "step": 25602 + }, + { + "epoch": 0.76, + "learning_rate": 2.8211900736220533e-05, + "loss": 0.5003, + "step": 25603 + }, + { + "epoch": 0.76, + "learning_rate": 2.820518835550381e-05, + "loss": 0.6925, + "step": 25604 + }, + { + "epoch": 0.76, + "learning_rate": 2.8198476642306628e-05, + "loss": 0.5623, + "step": 25605 + }, + { + "epoch": 0.76, + "learning_rate": 2.8191765596691412e-05, + "loss": 0.588, + "step": 25606 + }, + { + "epoch": 0.76, + "learning_rate": 2.818505521872051e-05, + "loss": 0.5817, + "step": 25607 + }, + { + "epoch": 0.76, + "learning_rate": 2.8178345508456383e-05, + "loss": 0.4595, + "step": 25608 + }, + { + "epoch": 0.76, + "learning_rate": 2.8171636465961347e-05, + "loss": 0.5061, + "step": 25609 + }, + { + "epoch": 0.76, + "learning_rate": 2.8164928091297828e-05, + "loss": 0.5234, + "step": 25610 + }, + { + "epoch": 0.76, + "learning_rate": 2.8158220384528146e-05, + "loss": 0.5459, + "step": 25611 + }, + { + "epoch": 0.76, + "learning_rate": 2.815151334571473e-05, + "loss": 0.4539, + "step": 25612 + }, + { + "epoch": 0.76, + "learning_rate": 2.814480697491988e-05, + "loss": 0.5009, + "step": 25613 + }, + { + "epoch": 0.76, + "learning_rate": 2.8138101272206008e-05, + "loss": 0.4752, + "step": 25614 + }, + { + "epoch": 0.76, + "learning_rate": 2.8131396237635387e-05, + "loss": 0.5164, + "step": 25615 + }, + { + "epoch": 0.76, + "learning_rate": 2.8124691871270413e-05, + "loss": 0.5674, + "step": 25616 + }, + { + "epoch": 0.76, + "learning_rate": 2.8117988173173428e-05, + "loss": 0.6366, + "step": 25617 + }, + { + "epoch": 0.76, + "learning_rate": 2.8111285143406753e-05, + "loss": 0.6216, + "step": 25618 + }, + { + "epoch": 0.76, + "learning_rate": 2.810458278203265e-05, + "loss": 0.4759, + "step": 25619 + }, + { + "epoch": 0.76, + "learning_rate": 2.8097881089113486e-05, + "loss": 0.5747, + "step": 25620 + }, + { + "epoch": 0.76, + "learning_rate": 2.8091180064711574e-05, + "loss": 0.5079, + "step": 25621 + }, + { + "epoch": 0.76, + "learning_rate": 2.8084479708889268e-05, + "loss": 0.5703, + "step": 25622 + }, + { + "epoch": 0.76, + "learning_rate": 2.8077780021708742e-05, + "loss": 0.7515, + "step": 25623 + }, + { + "epoch": 0.76, + "learning_rate": 2.8071081003232346e-05, + "loss": 0.4503, + "step": 25624 + }, + { + "epoch": 0.76, + "learning_rate": 2.806438265352238e-05, + "loss": 0.5745, + "step": 25625 + }, + { + "epoch": 0.76, + "learning_rate": 2.805768497264115e-05, + "loss": 0.605, + "step": 25626 + }, + { + "epoch": 0.76, + "learning_rate": 2.8050987960650878e-05, + "loss": 0.4345, + "step": 25627 + }, + { + "epoch": 0.76, + "learning_rate": 2.8044291617613827e-05, + "loss": 0.4651, + "step": 25628 + }, + { + "epoch": 0.76, + "learning_rate": 2.8037595943592264e-05, + "loss": 0.508, + "step": 25629 + }, + { + "epoch": 0.76, + "learning_rate": 2.8030900938648484e-05, + "loss": 0.5398, + "step": 25630 + }, + { + "epoch": 0.76, + "learning_rate": 2.802420660284467e-05, + "loss": 0.4278, + "step": 25631 + }, + { + "epoch": 0.76, + "learning_rate": 2.8017512936243128e-05, + "loss": 0.4245, + "step": 25632 + }, + { + "epoch": 0.76, + "learning_rate": 2.8010819938906076e-05, + "loss": 0.4865, + "step": 25633 + }, + { + "epoch": 0.76, + "learning_rate": 2.800412761089568e-05, + "loss": 0.5303, + "step": 25634 + }, + { + "epoch": 0.76, + "learning_rate": 2.799743595227422e-05, + "loss": 0.4666, + "step": 25635 + }, + { + "epoch": 0.76, + "learning_rate": 2.7990744963103922e-05, + "loss": 0.4773, + "step": 25636 + }, + { + "epoch": 0.76, + "learning_rate": 2.798405464344699e-05, + "loss": 0.5152, + "step": 25637 + }, + { + "epoch": 0.76, + "learning_rate": 2.7977364993365572e-05, + "loss": 0.4583, + "step": 25638 + }, + { + "epoch": 0.76, + "learning_rate": 2.7970676012921925e-05, + "loss": 0.4286, + "step": 25639 + }, + { + "epoch": 0.76, + "learning_rate": 2.7963987702178217e-05, + "loss": 0.4645, + "step": 25640 + }, + { + "epoch": 0.76, + "learning_rate": 2.7957300061196668e-05, + "loss": 0.5584, + "step": 25641 + }, + { + "epoch": 0.76, + "learning_rate": 2.7950613090039435e-05, + "loss": 0.4321, + "step": 25642 + }, + { + "epoch": 0.76, + "learning_rate": 2.7943926788768647e-05, + "loss": 0.4584, + "step": 25643 + }, + { + "epoch": 0.76, + "learning_rate": 2.7937241157446514e-05, + "loss": 0.6224, + "step": 25644 + }, + { + "epoch": 0.76, + "learning_rate": 2.7930556196135225e-05, + "loss": 0.5354, + "step": 25645 + }, + { + "epoch": 0.76, + "learning_rate": 2.792387190489689e-05, + "loss": 0.4474, + "step": 25646 + }, + { + "epoch": 0.76, + "learning_rate": 2.791718828379365e-05, + "loss": 0.5214, + "step": 25647 + }, + { + "epoch": 0.76, + "learning_rate": 2.791050533288766e-05, + "loss": 0.4972, + "step": 25648 + }, + { + "epoch": 0.76, + "learning_rate": 2.7903823052241084e-05, + "loss": 0.5691, + "step": 25649 + }, + { + "epoch": 0.76, + "learning_rate": 2.7897141441915998e-05, + "loss": 0.4845, + "step": 25650 + }, + { + "epoch": 0.76, + "learning_rate": 2.789046050197459e-05, + "loss": 0.5212, + "step": 25651 + }, + { + "epoch": 0.76, + "learning_rate": 2.7883780232478907e-05, + "loss": 0.4741, + "step": 25652 + }, + { + "epoch": 0.76, + "learning_rate": 2.7877100633491128e-05, + "loss": 0.505, + "step": 25653 + }, + { + "epoch": 0.76, + "learning_rate": 2.787042170507328e-05, + "loss": 0.5201, + "step": 25654 + }, + { + "epoch": 0.76, + "learning_rate": 2.786374344728754e-05, + "loss": 0.6191, + "step": 25655 + }, + { + "epoch": 0.76, + "learning_rate": 2.785706586019593e-05, + "loss": 0.5541, + "step": 25656 + }, + { + "epoch": 0.76, + "learning_rate": 2.7850388943860594e-05, + "loss": 0.564, + "step": 25657 + }, + { + "epoch": 0.76, + "learning_rate": 2.784371269834356e-05, + "loss": 0.5015, + "step": 25658 + }, + { + "epoch": 0.76, + "learning_rate": 2.7837037123706932e-05, + "loss": 0.6186, + "step": 25659 + }, + { + "epoch": 0.76, + "learning_rate": 2.7830362220012796e-05, + "loss": 0.4352, + "step": 25660 + }, + { + "epoch": 0.76, + "learning_rate": 2.782368798732319e-05, + "loss": 0.544, + "step": 25661 + }, + { + "epoch": 0.76, + "learning_rate": 2.7817014425700138e-05, + "loss": 0.3654, + "step": 25662 + }, + { + "epoch": 0.76, + "learning_rate": 2.781034153520572e-05, + "loss": 0.657, + "step": 25663 + }, + { + "epoch": 0.76, + "learning_rate": 2.7803669315901993e-05, + "loss": 0.5625, + "step": 25664 + }, + { + "epoch": 0.76, + "learning_rate": 2.7796997767850986e-05, + "loss": 0.4583, + "step": 25665 + }, + { + "epoch": 0.76, + "learning_rate": 2.779032689111467e-05, + "loss": 0.5972, + "step": 25666 + }, + { + "epoch": 0.76, + "learning_rate": 2.7783656685755132e-05, + "loss": 0.4518, + "step": 25667 + }, + { + "epoch": 0.76, + "learning_rate": 2.777698715183439e-05, + "loss": 0.5131, + "step": 25668 + }, + { + "epoch": 0.76, + "learning_rate": 2.777031828941441e-05, + "loss": 0.6342, + "step": 25669 + }, + { + "epoch": 0.76, + "learning_rate": 2.776365009855726e-05, + "loss": 0.5753, + "step": 25670 + }, + { + "epoch": 0.76, + "learning_rate": 2.7756982579324863e-05, + "loss": 0.5709, + "step": 25671 + }, + { + "epoch": 0.76, + "learning_rate": 2.7750315731779273e-05, + "loss": 0.5067, + "step": 25672 + }, + { + "epoch": 0.76, + "learning_rate": 2.7743649555982432e-05, + "loss": 0.4473, + "step": 25673 + }, + { + "epoch": 0.76, + "learning_rate": 2.7736984051996373e-05, + "loss": 0.471, + "step": 25674 + }, + { + "epoch": 0.76, + "learning_rate": 2.7730319219882994e-05, + "loss": 0.5109, + "step": 25675 + }, + { + "epoch": 0.76, + "learning_rate": 2.7723655059704334e-05, + "loss": 0.4951, + "step": 25676 + }, + { + "epoch": 0.76, + "learning_rate": 2.77169915715223e-05, + "loss": 0.5061, + "step": 25677 + }, + { + "epoch": 0.76, + "learning_rate": 2.771032875539886e-05, + "loss": 0.5099, + "step": 25678 + }, + { + "epoch": 0.76, + "learning_rate": 2.7703666611396005e-05, + "loss": 0.5703, + "step": 25679 + }, + { + "epoch": 0.76, + "learning_rate": 2.769700513957565e-05, + "loss": 0.4928, + "step": 25680 + }, + { + "epoch": 0.76, + "learning_rate": 2.7690344339999684e-05, + "loss": 0.5046, + "step": 25681 + }, + { + "epoch": 0.76, + "learning_rate": 2.768368421273009e-05, + "loss": 0.4993, + "step": 25682 + }, + { + "epoch": 0.76, + "learning_rate": 2.7677024757828762e-05, + "loss": 0.5181, + "step": 25683 + }, + { + "epoch": 0.76, + "learning_rate": 2.7670365975357703e-05, + "loss": 0.5771, + "step": 25684 + }, + { + "epoch": 0.76, + "learning_rate": 2.7663707865378685e-05, + "loss": 0.5424, + "step": 25685 + }, + { + "epoch": 0.76, + "learning_rate": 2.765705042795369e-05, + "loss": 0.4782, + "step": 25686 + }, + { + "epoch": 0.76, + "learning_rate": 2.76503936631446e-05, + "loss": 0.5571, + "step": 25687 + }, + { + "epoch": 0.76, + "learning_rate": 2.7643737571013352e-05, + "loss": 0.5277, + "step": 25688 + }, + { + "epoch": 0.76, + "learning_rate": 2.7637082151621785e-05, + "loss": 0.5285, + "step": 25689 + }, + { + "epoch": 0.76, + "learning_rate": 2.7630427405031754e-05, + "loss": 0.5868, + "step": 25690 + }, + { + "epoch": 0.76, + "learning_rate": 2.762377333130518e-05, + "loss": 0.5222, + "step": 25691 + }, + { + "epoch": 0.76, + "learning_rate": 2.7617119930503933e-05, + "loss": 0.5836, + "step": 25692 + }, + { + "epoch": 0.76, + "learning_rate": 2.7610467202689827e-05, + "loss": 0.4388, + "step": 25693 + }, + { + "epoch": 0.76, + "learning_rate": 2.7603815147924783e-05, + "loss": 0.605, + "step": 25694 + }, + { + "epoch": 0.76, + "learning_rate": 2.7597163766270574e-05, + "loss": 0.4296, + "step": 25695 + }, + { + "epoch": 0.77, + "learning_rate": 2.7590513057789113e-05, + "loss": 0.4951, + "step": 25696 + }, + { + "epoch": 0.77, + "learning_rate": 2.7583863022542177e-05, + "loss": 0.4717, + "step": 25697 + }, + { + "epoch": 0.77, + "learning_rate": 2.7577213660591637e-05, + "loss": 0.5291, + "step": 25698 + }, + { + "epoch": 0.77, + "learning_rate": 2.7570564971999314e-05, + "loss": 0.5525, + "step": 25699 + }, + { + "epoch": 0.77, + "learning_rate": 2.7563916956826973e-05, + "loss": 0.5213, + "step": 25700 + }, + { + "epoch": 0.77, + "learning_rate": 2.755726961513646e-05, + "loss": 0.5247, + "step": 25701 + }, + { + "epoch": 0.77, + "learning_rate": 2.7550622946989578e-05, + "loss": 0.6041, + "step": 25702 + }, + { + "epoch": 0.77, + "learning_rate": 2.754397695244817e-05, + "loss": 0.5403, + "step": 25703 + }, + { + "epoch": 0.77, + "learning_rate": 2.7537331631573982e-05, + "loss": 0.5651, + "step": 25704 + }, + { + "epoch": 0.77, + "learning_rate": 2.7530686984428765e-05, + "loss": 0.5087, + "step": 25705 + }, + { + "epoch": 0.77, + "learning_rate": 2.7524043011074353e-05, + "loss": 0.6126, + "step": 25706 + }, + { + "epoch": 0.77, + "learning_rate": 2.7517399711572524e-05, + "loss": 0.4954, + "step": 25707 + }, + { + "epoch": 0.77, + "learning_rate": 2.7510757085985018e-05, + "loss": 0.5113, + "step": 25708 + }, + { + "epoch": 0.77, + "learning_rate": 2.7504115134373575e-05, + "loss": 0.5628, + "step": 25709 + }, + { + "epoch": 0.77, + "learning_rate": 2.749747385679998e-05, + "loss": 0.5661, + "step": 25710 + }, + { + "epoch": 0.77, + "learning_rate": 2.749083325332601e-05, + "loss": 0.5204, + "step": 25711 + }, + { + "epoch": 0.77, + "learning_rate": 2.748419332401333e-05, + "loss": 0.586, + "step": 25712 + }, + { + "epoch": 0.77, + "learning_rate": 2.7477554068923762e-05, + "loss": 0.4237, + "step": 25713 + }, + { + "epoch": 0.77, + "learning_rate": 2.7470915488118954e-05, + "loss": 0.5906, + "step": 25714 + }, + { + "epoch": 0.77, + "learning_rate": 2.7464277581660704e-05, + "loss": 0.5248, + "step": 25715 + }, + { + "epoch": 0.77, + "learning_rate": 2.7457640349610646e-05, + "loss": 0.5024, + "step": 25716 + }, + { + "epoch": 0.77, + "learning_rate": 2.745100379203057e-05, + "loss": 0.5522, + "step": 25717 + }, + { + "epoch": 0.77, + "learning_rate": 2.7444367908982126e-05, + "loss": 0.5203, + "step": 25718 + }, + { + "epoch": 0.77, + "learning_rate": 2.7437732700527053e-05, + "loss": 0.4579, + "step": 25719 + }, + { + "epoch": 0.77, + "learning_rate": 2.7431098166726998e-05, + "loss": 0.524, + "step": 25720 + }, + { + "epoch": 0.77, + "learning_rate": 2.742446430764366e-05, + "loss": 0.5894, + "step": 25721 + }, + { + "epoch": 0.77, + "learning_rate": 2.7417831123338766e-05, + "loss": 0.4637, + "step": 25722 + }, + { + "epoch": 0.77, + "learning_rate": 2.7411198613873946e-05, + "loss": 0.6368, + "step": 25723 + }, + { + "epoch": 0.77, + "learning_rate": 2.7404566779310836e-05, + "loss": 0.5835, + "step": 25724 + }, + { + "epoch": 0.77, + "learning_rate": 2.7397935619711134e-05, + "loss": 0.4682, + "step": 25725 + }, + { + "epoch": 0.77, + "learning_rate": 2.739130513513649e-05, + "loss": 0.5606, + "step": 25726 + }, + { + "epoch": 0.77, + "learning_rate": 2.7384675325648614e-05, + "loss": 0.4853, + "step": 25727 + }, + { + "epoch": 0.77, + "learning_rate": 2.7378046191309025e-05, + "loss": 0.4742, + "step": 25728 + }, + { + "epoch": 0.77, + "learning_rate": 2.7371417732179417e-05, + "loss": 0.6057, + "step": 25729 + }, + { + "epoch": 0.77, + "learning_rate": 2.7364789948321424e-05, + "loss": 0.5084, + "step": 25730 + }, + { + "epoch": 0.77, + "learning_rate": 2.7358162839796685e-05, + "loss": 0.4978, + "step": 25731 + }, + { + "epoch": 0.77, + "learning_rate": 2.73515364066668e-05, + "loss": 0.3894, + "step": 25732 + }, + { + "epoch": 0.77, + "learning_rate": 2.7344910648993348e-05, + "loss": 0.3841, + "step": 25733 + }, + { + "epoch": 0.77, + "learning_rate": 2.7338285566837983e-05, + "loss": 0.5271, + "step": 25734 + }, + { + "epoch": 0.77, + "learning_rate": 2.7331661160262257e-05, + "loss": 0.5128, + "step": 25735 + }, + { + "epoch": 0.77, + "learning_rate": 2.732503742932778e-05, + "loss": 0.5169, + "step": 25736 + }, + { + "epoch": 0.77, + "learning_rate": 2.7318414374096168e-05, + "loss": 0.4992, + "step": 25737 + }, + { + "epoch": 0.77, + "learning_rate": 2.7311791994628967e-05, + "loss": 0.5386, + "step": 25738 + }, + { + "epoch": 0.77, + "learning_rate": 2.7305170290987725e-05, + "loss": 0.621, + "step": 25739 + }, + { + "epoch": 0.77, + "learning_rate": 2.7298549263234042e-05, + "loss": 0.635, + "step": 25740 + }, + { + "epoch": 0.77, + "learning_rate": 2.72919289114295e-05, + "loss": 0.5505, + "step": 25741 + }, + { + "epoch": 0.77, + "learning_rate": 2.7285309235635627e-05, + "loss": 0.5059, + "step": 25742 + }, + { + "epoch": 0.77, + "learning_rate": 2.727869023591393e-05, + "loss": 0.6237, + "step": 25743 + }, + { + "epoch": 0.77, + "learning_rate": 2.727207191232599e-05, + "loss": 0.5365, + "step": 25744 + }, + { + "epoch": 0.77, + "learning_rate": 2.7265454264933345e-05, + "loss": 0.5109, + "step": 25745 + }, + { + "epoch": 0.77, + "learning_rate": 2.7258837293797533e-05, + "loss": 0.5473, + "step": 25746 + }, + { + "epoch": 0.77, + "learning_rate": 2.725222099898007e-05, + "loss": 0.5928, + "step": 25747 + }, + { + "epoch": 0.77, + "learning_rate": 2.724560538054244e-05, + "loss": 0.4443, + "step": 25748 + }, + { + "epoch": 0.77, + "learning_rate": 2.7238990438546162e-05, + "loss": 0.5215, + "step": 25749 + }, + { + "epoch": 0.77, + "learning_rate": 2.7232376173052786e-05, + "loss": 0.5206, + "step": 25750 + }, + { + "epoch": 0.77, + "learning_rate": 2.7225762584123772e-05, + "loss": 0.6264, + "step": 25751 + }, + { + "epoch": 0.77, + "learning_rate": 2.7219149671820586e-05, + "loss": 0.5052, + "step": 25752 + }, + { + "epoch": 0.77, + "learning_rate": 2.721253743620473e-05, + "loss": 0.5321, + "step": 25753 + }, + { + "epoch": 0.77, + "learning_rate": 2.7205925877337724e-05, + "loss": 0.5114, + "step": 25754 + }, + { + "epoch": 0.77, + "learning_rate": 2.7199314995280976e-05, + "loss": 0.4631, + "step": 25755 + }, + { + "epoch": 0.77, + "learning_rate": 2.7192704790096025e-05, + "loss": 0.4497, + "step": 25756 + }, + { + "epoch": 0.77, + "learning_rate": 2.7186095261844247e-05, + "loss": 0.4872, + "step": 25757 + }, + { + "epoch": 0.77, + "learning_rate": 2.7179486410587164e-05, + "loss": 0.5087, + "step": 25758 + }, + { + "epoch": 0.77, + "learning_rate": 2.7172878236386155e-05, + "loss": 0.4924, + "step": 25759 + }, + { + "epoch": 0.77, + "learning_rate": 2.7166270739302736e-05, + "loss": 0.5393, + "step": 25760 + }, + { + "epoch": 0.77, + "learning_rate": 2.715966391939827e-05, + "loss": 0.4982, + "step": 25761 + }, + { + "epoch": 0.77, + "learning_rate": 2.715305777673426e-05, + "loss": 0.4838, + "step": 25762 + }, + { + "epoch": 0.77, + "learning_rate": 2.714645231137204e-05, + "loss": 0.4999, + "step": 25763 + }, + { + "epoch": 0.77, + "learning_rate": 2.7139847523373075e-05, + "loss": 0.59, + "step": 25764 + }, + { + "epoch": 0.77, + "learning_rate": 2.7133243412798804e-05, + "loss": 0.6239, + "step": 25765 + }, + { + "epoch": 0.77, + "learning_rate": 2.7126639979710588e-05, + "loss": 0.4987, + "step": 25766 + }, + { + "epoch": 0.77, + "learning_rate": 2.7120037224169793e-05, + "loss": 0.5816, + "step": 25767 + }, + { + "epoch": 0.77, + "learning_rate": 2.7113435146237852e-05, + "loss": 0.5183, + "step": 25768 + }, + { + "epoch": 0.77, + "learning_rate": 2.7106833745976167e-05, + "loss": 0.5855, + "step": 25769 + }, + { + "epoch": 0.77, + "learning_rate": 2.7100233023446098e-05, + "loss": 0.5417, + "step": 25770 + }, + { + "epoch": 0.77, + "learning_rate": 2.709363297870897e-05, + "loss": 0.4423, + "step": 25771 + }, + { + "epoch": 0.77, + "learning_rate": 2.7087033611826196e-05, + "loss": 0.586, + "step": 25772 + }, + { + "epoch": 0.77, + "learning_rate": 2.7080434922859142e-05, + "loss": 0.5498, + "step": 25773 + }, + { + "epoch": 0.77, + "learning_rate": 2.707383691186911e-05, + "loss": 0.503, + "step": 25774 + }, + { + "epoch": 0.77, + "learning_rate": 2.7067239578917513e-05, + "loss": 0.5047, + "step": 25775 + }, + { + "epoch": 0.77, + "learning_rate": 2.7060642924065628e-05, + "loss": 0.5609, + "step": 25776 + }, + { + "epoch": 0.77, + "learning_rate": 2.7054046947374833e-05, + "loss": 0.4976, + "step": 25777 + }, + { + "epoch": 0.77, + "learning_rate": 2.704745164890641e-05, + "loss": 0.5237, + "step": 25778 + }, + { + "epoch": 0.77, + "learning_rate": 2.7040857028721735e-05, + "loss": 0.6005, + "step": 25779 + }, + { + "epoch": 0.77, + "learning_rate": 2.7034263086882072e-05, + "loss": 0.6005, + "step": 25780 + }, + { + "epoch": 0.77, + "learning_rate": 2.7027669823448766e-05, + "loss": 0.5892, + "step": 25781 + }, + { + "epoch": 0.77, + "learning_rate": 2.7021077238483083e-05, + "loss": 0.5014, + "step": 25782 + }, + { + "epoch": 0.77, + "learning_rate": 2.7014485332046337e-05, + "loss": 0.4645, + "step": 25783 + }, + { + "epoch": 0.77, + "learning_rate": 2.7007894104199838e-05, + "loss": 0.5337, + "step": 25784 + }, + { + "epoch": 0.77, + "learning_rate": 2.7001303555004854e-05, + "loss": 0.5891, + "step": 25785 + }, + { + "epoch": 0.77, + "learning_rate": 2.6994713684522622e-05, + "loss": 0.5564, + "step": 25786 + }, + { + "epoch": 0.77, + "learning_rate": 2.6988124492814447e-05, + "loss": 0.608, + "step": 25787 + }, + { + "epoch": 0.77, + "learning_rate": 2.6981535979941586e-05, + "loss": 0.5431, + "step": 25788 + }, + { + "epoch": 0.77, + "learning_rate": 2.6974948145965372e-05, + "loss": 0.6395, + "step": 25789 + }, + { + "epoch": 0.77, + "learning_rate": 2.6968360990946916e-05, + "loss": 0.6818, + "step": 25790 + }, + { + "epoch": 0.77, + "learning_rate": 2.6961774514947536e-05, + "loss": 0.5803, + "step": 25791 + }, + { + "epoch": 0.77, + "learning_rate": 2.6955188718028467e-05, + "loss": 0.6917, + "step": 25792 + }, + { + "epoch": 0.77, + "learning_rate": 2.6948603600250965e-05, + "loss": 0.5715, + "step": 25793 + }, + { + "epoch": 0.77, + "learning_rate": 2.6942019161676234e-05, + "loss": 0.5582, + "step": 25794 + }, + { + "epoch": 0.77, + "learning_rate": 2.6935435402365473e-05, + "loss": 0.4577, + "step": 25795 + }, + { + "epoch": 0.77, + "learning_rate": 2.6928852322379894e-05, + "loss": 0.4599, + "step": 25796 + }, + { + "epoch": 0.77, + "learning_rate": 2.6922269921780762e-05, + "loss": 0.5414, + "step": 25797 + }, + { + "epoch": 0.77, + "learning_rate": 2.6915688200629207e-05, + "loss": 0.4845, + "step": 25798 + }, + { + "epoch": 0.77, + "learning_rate": 2.690910715898649e-05, + "loss": 0.5384, + "step": 25799 + }, + { + "epoch": 0.77, + "learning_rate": 2.6902526796913753e-05, + "loss": 0.5288, + "step": 25800 + }, + { + "epoch": 0.77, + "learning_rate": 2.6895947114472166e-05, + "loss": 0.4646, + "step": 25801 + }, + { + "epoch": 0.77, + "learning_rate": 2.6889368111722924e-05, + "loss": 0.5316, + "step": 25802 + }, + { + "epoch": 0.77, + "learning_rate": 2.6882789788727224e-05, + "loss": 0.4393, + "step": 25803 + }, + { + "epoch": 0.77, + "learning_rate": 2.6876212145546208e-05, + "loss": 0.5607, + "step": 25804 + }, + { + "epoch": 0.77, + "learning_rate": 2.6869635182241006e-05, + "loss": 0.5615, + "step": 25805 + }, + { + "epoch": 0.77, + "learning_rate": 2.6863058898872772e-05, + "loss": 0.6003, + "step": 25806 + }, + { + "epoch": 0.77, + "learning_rate": 2.6856483295502676e-05, + "loss": 0.5129, + "step": 25807 + }, + { + "epoch": 0.77, + "learning_rate": 2.6849908372191868e-05, + "loss": 0.4742, + "step": 25808 + }, + { + "epoch": 0.77, + "learning_rate": 2.6843334129001475e-05, + "loss": 0.6999, + "step": 25809 + }, + { + "epoch": 0.77, + "learning_rate": 2.683676056599256e-05, + "loss": 0.5209, + "step": 25810 + }, + { + "epoch": 0.77, + "learning_rate": 2.6830187683226283e-05, + "loss": 0.6193, + "step": 25811 + }, + { + "epoch": 0.77, + "learning_rate": 2.6823615480763798e-05, + "loss": 0.5885, + "step": 25812 + }, + { + "epoch": 0.77, + "learning_rate": 2.6817043958666166e-05, + "loss": 0.5153, + "step": 25813 + }, + { + "epoch": 0.77, + "learning_rate": 2.681047311699446e-05, + "loss": 0.4043, + "step": 25814 + }, + { + "epoch": 0.77, + "learning_rate": 2.68039029558098e-05, + "loss": 0.6332, + "step": 25815 + }, + { + "epoch": 0.77, + "learning_rate": 2.6797333475173315e-05, + "loss": 0.5234, + "step": 25816 + }, + { + "epoch": 0.77, + "learning_rate": 2.6790764675146018e-05, + "loss": 0.5089, + "step": 25817 + }, + { + "epoch": 0.77, + "learning_rate": 2.6784196555789043e-05, + "loss": 0.4649, + "step": 25818 + }, + { + "epoch": 0.77, + "learning_rate": 2.6777629117163395e-05, + "loss": 0.5348, + "step": 25819 + }, + { + "epoch": 0.77, + "learning_rate": 2.6771062359330202e-05, + "loss": 0.6358, + "step": 25820 + }, + { + "epoch": 0.77, + "learning_rate": 2.6764496282350447e-05, + "loss": 0.6722, + "step": 25821 + }, + { + "epoch": 0.77, + "learning_rate": 2.6757930886285252e-05, + "loss": 0.5523, + "step": 25822 + }, + { + "epoch": 0.77, + "learning_rate": 2.675136617119559e-05, + "loss": 0.5357, + "step": 25823 + }, + { + "epoch": 0.77, + "learning_rate": 2.6744802137142566e-05, + "loss": 0.4662, + "step": 25824 + }, + { + "epoch": 0.77, + "learning_rate": 2.6738238784187143e-05, + "loss": 0.4758, + "step": 25825 + }, + { + "epoch": 0.77, + "learning_rate": 2.673167611239038e-05, + "loss": 0.5514, + "step": 25826 + }, + { + "epoch": 0.77, + "learning_rate": 2.672511412181331e-05, + "loss": 0.5076, + "step": 25827 + }, + { + "epoch": 0.77, + "learning_rate": 2.6718552812516938e-05, + "loss": 0.5203, + "step": 25828 + }, + { + "epoch": 0.77, + "learning_rate": 2.6711992184562217e-05, + "loss": 0.4688, + "step": 25829 + }, + { + "epoch": 0.77, + "learning_rate": 2.6705432238010188e-05, + "loss": 0.5376, + "step": 25830 + }, + { + "epoch": 0.77, + "learning_rate": 2.6698872972921863e-05, + "loss": 0.519, + "step": 25831 + }, + { + "epoch": 0.77, + "learning_rate": 2.6692314389358196e-05, + "loss": 0.4983, + "step": 25832 + }, + { + "epoch": 0.77, + "learning_rate": 2.6685756487380153e-05, + "loss": 0.4844, + "step": 25833 + }, + { + "epoch": 0.77, + "learning_rate": 2.6679199267048727e-05, + "loss": 0.4385, + "step": 25834 + }, + { + "epoch": 0.77, + "learning_rate": 2.6672642728424913e-05, + "loss": 0.5444, + "step": 25835 + }, + { + "epoch": 0.77, + "learning_rate": 2.6666086871569605e-05, + "loss": 0.5368, + "step": 25836 + }, + { + "epoch": 0.77, + "learning_rate": 2.6659531696543828e-05, + "loss": 0.5276, + "step": 25837 + }, + { + "epoch": 0.77, + "learning_rate": 2.6652977203408458e-05, + "loss": 0.5141, + "step": 25838 + }, + { + "epoch": 0.77, + "learning_rate": 2.6646423392224507e-05, + "loss": 0.6181, + "step": 25839 + }, + { + "epoch": 0.77, + "learning_rate": 2.6639870263052845e-05, + "loss": 0.5435, + "step": 25840 + }, + { + "epoch": 0.77, + "learning_rate": 2.6633317815954463e-05, + "loss": 0.4277, + "step": 25841 + }, + { + "epoch": 0.77, + "learning_rate": 2.6626766050990226e-05, + "loss": 0.47, + "step": 25842 + }, + { + "epoch": 0.77, + "learning_rate": 2.662021496822109e-05, + "loss": 0.6118, + "step": 25843 + }, + { + "epoch": 0.77, + "learning_rate": 2.661366456770793e-05, + "loss": 0.5304, + "step": 25844 + }, + { + "epoch": 0.77, + "learning_rate": 2.660711484951166e-05, + "loss": 0.5822, + "step": 25845 + }, + { + "epoch": 0.77, + "learning_rate": 2.6600565813693224e-05, + "loss": 0.5392, + "step": 25846 + }, + { + "epoch": 0.77, + "learning_rate": 2.659401746031347e-05, + "loss": 0.5198, + "step": 25847 + }, + { + "epoch": 0.77, + "learning_rate": 2.658746978943325e-05, + "loss": 0.5757, + "step": 25848 + }, + { + "epoch": 0.77, + "learning_rate": 2.6580922801113484e-05, + "loss": 0.5733, + "step": 25849 + }, + { + "epoch": 0.77, + "learning_rate": 2.657437649541503e-05, + "loss": 0.6414, + "step": 25850 + }, + { + "epoch": 0.77, + "learning_rate": 2.6567830872398825e-05, + "loss": 0.6089, + "step": 25851 + }, + { + "epoch": 0.77, + "learning_rate": 2.656128593212559e-05, + "loss": 0.5386, + "step": 25852 + }, + { + "epoch": 0.77, + "learning_rate": 2.6554741674656258e-05, + "loss": 0.4783, + "step": 25853 + }, + { + "epoch": 0.77, + "learning_rate": 2.654819810005167e-05, + "loss": 0.5698, + "step": 25854 + }, + { + "epoch": 0.77, + "learning_rate": 2.6541655208372684e-05, + "loss": 0.5676, + "step": 25855 + }, + { + "epoch": 0.77, + "learning_rate": 2.653511299968011e-05, + "loss": 0.4996, + "step": 25856 + }, + { + "epoch": 0.77, + "learning_rate": 2.6528571474034748e-05, + "loss": 0.6513, + "step": 25857 + }, + { + "epoch": 0.77, + "learning_rate": 2.652203063149744e-05, + "loss": 0.5895, + "step": 25858 + }, + { + "epoch": 0.77, + "learning_rate": 2.6515490472129033e-05, + "loss": 0.4456, + "step": 25859 + }, + { + "epoch": 0.77, + "learning_rate": 2.650895099599029e-05, + "loss": 0.5281, + "step": 25860 + }, + { + "epoch": 0.77, + "learning_rate": 2.6502412203142057e-05, + "loss": 0.4594, + "step": 25861 + }, + { + "epoch": 0.77, + "learning_rate": 2.6495874093645066e-05, + "loss": 0.5987, + "step": 25862 + }, + { + "epoch": 0.77, + "learning_rate": 2.6489336667560173e-05, + "loss": 0.5476, + "step": 25863 + }, + { + "epoch": 0.77, + "learning_rate": 2.64827999249481e-05, + "loss": 0.4511, + "step": 25864 + }, + { + "epoch": 0.77, + "learning_rate": 2.6476263865869687e-05, + "loss": 0.5642, + "step": 25865 + }, + { + "epoch": 0.77, + "learning_rate": 2.646972849038566e-05, + "loss": 0.6022, + "step": 25866 + }, + { + "epoch": 0.77, + "learning_rate": 2.6463193798556762e-05, + "loss": 0.561, + "step": 25867 + }, + { + "epoch": 0.77, + "learning_rate": 2.6456659790443796e-05, + "loss": 0.6326, + "step": 25868 + }, + { + "epoch": 0.77, + "learning_rate": 2.6450126466107473e-05, + "loss": 0.4992, + "step": 25869 + }, + { + "epoch": 0.77, + "learning_rate": 2.6443593825608604e-05, + "loss": 0.673, + "step": 25870 + }, + { + "epoch": 0.77, + "learning_rate": 2.6437061869007884e-05, + "loss": 0.567, + "step": 25871 + }, + { + "epoch": 0.77, + "learning_rate": 2.6430530596366e-05, + "loss": 0.6049, + "step": 25872 + }, + { + "epoch": 0.77, + "learning_rate": 2.642400000774373e-05, + "loss": 0.5668, + "step": 25873 + }, + { + "epoch": 0.77, + "learning_rate": 2.6417470103201814e-05, + "loss": 0.5122, + "step": 25874 + }, + { + "epoch": 0.77, + "learning_rate": 2.6410940882800937e-05, + "loss": 0.5567, + "step": 25875 + }, + { + "epoch": 0.77, + "learning_rate": 2.640441234660176e-05, + "loss": 0.6348, + "step": 25876 + }, + { + "epoch": 0.77, + "learning_rate": 2.639788449466504e-05, + "loss": 0.4929, + "step": 25877 + }, + { + "epoch": 0.77, + "learning_rate": 2.6391357327051468e-05, + "loss": 0.4789, + "step": 25878 + }, + { + "epoch": 0.77, + "learning_rate": 2.6384830843821707e-05, + "loss": 0.5246, + "step": 25879 + }, + { + "epoch": 0.77, + "learning_rate": 2.6378305045036466e-05, + "loss": 0.6322, + "step": 25880 + }, + { + "epoch": 0.77, + "learning_rate": 2.6371779930756367e-05, + "loss": 0.5057, + "step": 25881 + }, + { + "epoch": 0.77, + "learning_rate": 2.6365255501042153e-05, + "loss": 0.4702, + "step": 25882 + }, + { + "epoch": 0.77, + "learning_rate": 2.6358731755954412e-05, + "loss": 0.5495, + "step": 25883 + }, + { + "epoch": 0.77, + "learning_rate": 2.635220869555387e-05, + "loss": 0.5442, + "step": 25884 + }, + { + "epoch": 0.77, + "learning_rate": 2.6345686319901097e-05, + "loss": 0.4405, + "step": 25885 + }, + { + "epoch": 0.77, + "learning_rate": 2.6339164629056813e-05, + "loss": 0.4792, + "step": 25886 + }, + { + "epoch": 0.77, + "learning_rate": 2.6332643623081587e-05, + "loss": 0.4956, + "step": 25887 + }, + { + "epoch": 0.77, + "learning_rate": 2.6326123302036087e-05, + "loss": 0.5456, + "step": 25888 + }, + { + "epoch": 0.77, + "learning_rate": 2.6319603665980952e-05, + "loss": 0.5347, + "step": 25889 + }, + { + "epoch": 0.77, + "learning_rate": 2.6313084714976777e-05, + "loss": 0.5409, + "step": 25890 + }, + { + "epoch": 0.77, + "learning_rate": 2.6306566449084136e-05, + "loss": 0.4259, + "step": 25891 + }, + { + "epoch": 0.77, + "learning_rate": 2.630004886836367e-05, + "loss": 0.6376, + "step": 25892 + }, + { + "epoch": 0.77, + "learning_rate": 2.6293531972875985e-05, + "loss": 0.5303, + "step": 25893 + }, + { + "epoch": 0.77, + "learning_rate": 2.6287015762681712e-05, + "loss": 0.5904, + "step": 25894 + }, + { + "epoch": 0.77, + "learning_rate": 2.6280500237841333e-05, + "loss": 0.521, + "step": 25895 + }, + { + "epoch": 0.77, + "learning_rate": 2.627398539841548e-05, + "loss": 0.5262, + "step": 25896 + }, + { + "epoch": 0.77, + "learning_rate": 2.626747124446476e-05, + "loss": 0.5754, + "step": 25897 + }, + { + "epoch": 0.77, + "learning_rate": 2.6260957776049676e-05, + "loss": 0.4544, + "step": 25898 + }, + { + "epoch": 0.77, + "learning_rate": 2.625444499323084e-05, + "loss": 0.5129, + "step": 25899 + }, + { + "epoch": 0.77, + "learning_rate": 2.624793289606876e-05, + "loss": 0.6129, + "step": 25900 + }, + { + "epoch": 0.77, + "learning_rate": 2.624142148462403e-05, + "loss": 0.5249, + "step": 25901 + }, + { + "epoch": 0.77, + "learning_rate": 2.6234910758957132e-05, + "loss": 0.4818, + "step": 25902 + }, + { + "epoch": 0.77, + "learning_rate": 2.6228400719128642e-05, + "loss": 0.5019, + "step": 25903 + }, + { + "epoch": 0.77, + "learning_rate": 2.62218913651991e-05, + "loss": 0.4775, + "step": 25904 + }, + { + "epoch": 0.77, + "learning_rate": 2.6215382697229018e-05, + "loss": 0.6862, + "step": 25905 + }, + { + "epoch": 0.77, + "learning_rate": 2.6208874715278874e-05, + "loss": 0.5412, + "step": 25906 + }, + { + "epoch": 0.77, + "learning_rate": 2.6202367419409192e-05, + "loss": 0.7101, + "step": 25907 + }, + { + "epoch": 0.77, + "learning_rate": 2.6195860809680517e-05, + "loss": 0.5649, + "step": 25908 + }, + { + "epoch": 0.77, + "learning_rate": 2.6189354886153318e-05, + "loss": 0.5655, + "step": 25909 + }, + { + "epoch": 0.77, + "learning_rate": 2.618284964888804e-05, + "loss": 0.4441, + "step": 25910 + }, + { + "epoch": 0.77, + "learning_rate": 2.617634509794521e-05, + "loss": 0.5206, + "step": 25911 + }, + { + "epoch": 0.77, + "learning_rate": 2.6169841233385295e-05, + "loss": 0.5658, + "step": 25912 + }, + { + "epoch": 0.77, + "learning_rate": 2.6163338055268803e-05, + "loss": 0.5936, + "step": 25913 + }, + { + "epoch": 0.77, + "learning_rate": 2.6156835563656166e-05, + "loss": 0.5694, + "step": 25914 + }, + { + "epoch": 0.77, + "learning_rate": 2.6150333758607813e-05, + "loss": 0.5434, + "step": 25915 + }, + { + "epoch": 0.77, + "learning_rate": 2.6143832640184228e-05, + "loss": 0.4659, + "step": 25916 + }, + { + "epoch": 0.77, + "learning_rate": 2.6137332208445876e-05, + "loss": 0.4966, + "step": 25917 + }, + { + "epoch": 0.77, + "learning_rate": 2.613083246345317e-05, + "loss": 0.5502, + "step": 25918 + }, + { + "epoch": 0.77, + "learning_rate": 2.612433340526651e-05, + "loss": 0.5325, + "step": 25919 + }, + { + "epoch": 0.77, + "learning_rate": 2.6117835033946358e-05, + "loss": 0.5057, + "step": 25920 + }, + { + "epoch": 0.77, + "learning_rate": 2.6111337349553157e-05, + "loss": 0.4588, + "step": 25921 + }, + { + "epoch": 0.77, + "learning_rate": 2.6104840352147264e-05, + "loss": 0.5471, + "step": 25922 + }, + { + "epoch": 0.77, + "learning_rate": 2.6098344041789147e-05, + "loss": 0.4699, + "step": 25923 + }, + { + "epoch": 0.77, + "learning_rate": 2.6091848418539133e-05, + "loss": 0.4654, + "step": 25924 + }, + { + "epoch": 0.77, + "learning_rate": 2.6085353482457696e-05, + "loss": 0.5234, + "step": 25925 + }, + { + "epoch": 0.77, + "learning_rate": 2.6078859233605148e-05, + "loss": 0.4378, + "step": 25926 + }, + { + "epoch": 0.77, + "learning_rate": 2.6072365672041933e-05, + "loss": 0.5413, + "step": 25927 + }, + { + "epoch": 0.77, + "learning_rate": 2.6065872797828374e-05, + "loss": 0.5632, + "step": 25928 + }, + { + "epoch": 0.77, + "learning_rate": 2.6059380611024887e-05, + "loss": 0.5013, + "step": 25929 + }, + { + "epoch": 0.77, + "learning_rate": 2.605288911169178e-05, + "loss": 0.4434, + "step": 25930 + }, + { + "epoch": 0.77, + "learning_rate": 2.604639829988943e-05, + "loss": 0.5268, + "step": 25931 + }, + { + "epoch": 0.77, + "learning_rate": 2.6039908175678228e-05, + "loss": 0.4887, + "step": 25932 + }, + { + "epoch": 0.77, + "learning_rate": 2.6033418739118486e-05, + "loss": 0.5389, + "step": 25933 + }, + { + "epoch": 0.77, + "learning_rate": 2.602692999027051e-05, + "loss": 0.5434, + "step": 25934 + }, + { + "epoch": 0.77, + "learning_rate": 2.602044192919465e-05, + "loss": 0.3524, + "step": 25935 + }, + { + "epoch": 0.77, + "learning_rate": 2.601395455595127e-05, + "loss": 0.5219, + "step": 25936 + }, + { + "epoch": 0.77, + "learning_rate": 2.6007467870600654e-05, + "loss": 0.5215, + "step": 25937 + }, + { + "epoch": 0.77, + "learning_rate": 2.600098187320308e-05, + "loss": 0.6603, + "step": 25938 + }, + { + "epoch": 0.77, + "learning_rate": 2.599449656381887e-05, + "loss": 0.5857, + "step": 25939 + }, + { + "epoch": 0.77, + "learning_rate": 2.5988011942508383e-05, + "loss": 0.5674, + "step": 25940 + }, + { + "epoch": 0.77, + "learning_rate": 2.5981528009331825e-05, + "loss": 0.517, + "step": 25941 + }, + { + "epoch": 0.77, + "learning_rate": 2.5975044764349543e-05, + "loss": 0.5695, + "step": 25942 + }, + { + "epoch": 0.77, + "learning_rate": 2.5968562207621772e-05, + "loss": 0.5532, + "step": 25943 + }, + { + "epoch": 0.77, + "learning_rate": 2.5962080339208827e-05, + "loss": 0.5289, + "step": 25944 + }, + { + "epoch": 0.77, + "learning_rate": 2.5955599159170928e-05, + "loss": 0.5301, + "step": 25945 + }, + { + "epoch": 0.77, + "learning_rate": 2.5949118667568385e-05, + "loss": 0.4543, + "step": 25946 + }, + { + "epoch": 0.77, + "learning_rate": 2.5942638864461387e-05, + "loss": 0.5331, + "step": 25947 + }, + { + "epoch": 0.77, + "learning_rate": 2.593615974991025e-05, + "loss": 0.4645, + "step": 25948 + }, + { + "epoch": 0.77, + "learning_rate": 2.5929681323975142e-05, + "loss": 0.5333, + "step": 25949 + }, + { + "epoch": 0.77, + "learning_rate": 2.592320358671635e-05, + "loss": 0.4597, + "step": 25950 + }, + { + "epoch": 0.77, + "learning_rate": 2.5916726538194103e-05, + "loss": 0.5499, + "step": 25951 + }, + { + "epoch": 0.77, + "learning_rate": 2.591025017846862e-05, + "loss": 0.5028, + "step": 25952 + }, + { + "epoch": 0.77, + "learning_rate": 2.5903774507600065e-05, + "loss": 0.6128, + "step": 25953 + }, + { + "epoch": 0.77, + "learning_rate": 2.5897299525648678e-05, + "loss": 0.5785, + "step": 25954 + }, + { + "epoch": 0.77, + "learning_rate": 2.5890825232674666e-05, + "loss": 0.5608, + "step": 25955 + }, + { + "epoch": 0.77, + "learning_rate": 2.588435162873829e-05, + "loss": 0.5893, + "step": 25956 + }, + { + "epoch": 0.77, + "learning_rate": 2.5877878713899617e-05, + "loss": 0.5275, + "step": 25957 + }, + { + "epoch": 0.77, + "learning_rate": 2.5871406488218875e-05, + "loss": 0.5601, + "step": 25958 + }, + { + "epoch": 0.77, + "learning_rate": 2.586493495175626e-05, + "loss": 0.5317, + "step": 25959 + }, + { + "epoch": 0.77, + "learning_rate": 2.5858464104571955e-05, + "loss": 0.4961, + "step": 25960 + }, + { + "epoch": 0.77, + "learning_rate": 2.5851993946726106e-05, + "loss": 0.5378, + "step": 25961 + }, + { + "epoch": 0.77, + "learning_rate": 2.5845524478278825e-05, + "loss": 0.4718, + "step": 25962 + }, + { + "epoch": 0.77, + "learning_rate": 2.583905569929035e-05, + "loss": 0.4779, + "step": 25963 + }, + { + "epoch": 0.77, + "learning_rate": 2.5832587609820735e-05, + "loss": 0.604, + "step": 25964 + }, + { + "epoch": 0.77, + "learning_rate": 2.582612020993016e-05, + "loss": 0.4687, + "step": 25965 + }, + { + "epoch": 0.77, + "learning_rate": 2.5819653499678787e-05, + "loss": 0.5204, + "step": 25966 + }, + { + "epoch": 0.77, + "learning_rate": 2.581318747912671e-05, + "loss": 0.4534, + "step": 25967 + }, + { + "epoch": 0.77, + "learning_rate": 2.5806722148334017e-05, + "loss": 0.5149, + "step": 25968 + }, + { + "epoch": 0.77, + "learning_rate": 2.580025750736086e-05, + "loss": 0.4757, + "step": 25969 + }, + { + "epoch": 0.77, + "learning_rate": 2.5793793556267364e-05, + "loss": 0.5765, + "step": 25970 + }, + { + "epoch": 0.77, + "learning_rate": 2.578733029511359e-05, + "loss": 0.5186, + "step": 25971 + }, + { + "epoch": 0.77, + "learning_rate": 2.578086772395961e-05, + "loss": 0.4232, + "step": 25972 + }, + { + "epoch": 0.77, + "learning_rate": 2.5774405842865557e-05, + "loss": 0.5098, + "step": 25973 + }, + { + "epoch": 0.77, + "learning_rate": 2.5767944651891486e-05, + "loss": 0.5283, + "step": 25974 + }, + { + "epoch": 0.77, + "learning_rate": 2.5761484151097505e-05, + "loss": 0.4815, + "step": 25975 + }, + { + "epoch": 0.77, + "learning_rate": 2.5755024340543666e-05, + "loss": 0.5838, + "step": 25976 + }, + { + "epoch": 0.77, + "learning_rate": 2.5748565220289977e-05, + "loss": 0.4995, + "step": 25977 + }, + { + "epoch": 0.77, + "learning_rate": 2.574210679039655e-05, + "loss": 0.4475, + "step": 25978 + }, + { + "epoch": 0.77, + "learning_rate": 2.5735649050923426e-05, + "loss": 0.5421, + "step": 25979 + }, + { + "epoch": 0.77, + "learning_rate": 2.5729192001930658e-05, + "loss": 0.5016, + "step": 25980 + }, + { + "epoch": 0.77, + "learning_rate": 2.5722735643478224e-05, + "loss": 0.5567, + "step": 25981 + }, + { + "epoch": 0.77, + "learning_rate": 2.5716279975626188e-05, + "loss": 0.5011, + "step": 25982 + }, + { + "epoch": 0.77, + "learning_rate": 2.5709824998434606e-05, + "loss": 0.5948, + "step": 25983 + }, + { + "epoch": 0.77, + "learning_rate": 2.5703370711963437e-05, + "loss": 0.5171, + "step": 25984 + }, + { + "epoch": 0.77, + "learning_rate": 2.5696917116272746e-05, + "loss": 0.5423, + "step": 25985 + }, + { + "epoch": 0.77, + "learning_rate": 2.5690464211422462e-05, + "loss": 0.5794, + "step": 25986 + }, + { + "epoch": 0.77, + "learning_rate": 2.5684011997472667e-05, + "loss": 0.53, + "step": 25987 + }, + { + "epoch": 0.77, + "learning_rate": 2.5677560474483276e-05, + "loss": 0.4418, + "step": 25988 + }, + { + "epoch": 0.77, + "learning_rate": 2.5671109642514335e-05, + "loss": 0.5131, + "step": 25989 + }, + { + "epoch": 0.77, + "learning_rate": 2.5664659501625766e-05, + "loss": 0.5384, + "step": 25990 + }, + { + "epoch": 0.77, + "learning_rate": 2.5658210051877585e-05, + "loss": 0.482, + "step": 25991 + }, + { + "epoch": 0.77, + "learning_rate": 2.5651761293329723e-05, + "loss": 0.4925, + "step": 25992 + }, + { + "epoch": 0.77, + "learning_rate": 2.5645313226042135e-05, + "loss": 0.5786, + "step": 25993 + }, + { + "epoch": 0.77, + "learning_rate": 2.5638865850074833e-05, + "loss": 0.445, + "step": 25994 + }, + { + "epoch": 0.77, + "learning_rate": 2.5632419165487708e-05, + "loss": 0.5328, + "step": 25995 + }, + { + "epoch": 0.77, + "learning_rate": 2.562597317234069e-05, + "loss": 0.5564, + "step": 25996 + }, + { + "epoch": 0.77, + "learning_rate": 2.5619527870693727e-05, + "loss": 0.4964, + "step": 25997 + }, + { + "epoch": 0.77, + "learning_rate": 2.5613083260606773e-05, + "loss": 0.5527, + "step": 25998 + }, + { + "epoch": 0.77, + "learning_rate": 2.5606639342139738e-05, + "loss": 0.5381, + "step": 25999 + }, + { + "epoch": 0.77, + "learning_rate": 2.5600196115352482e-05, + "loss": 0.5542, + "step": 26000 + }, + { + "epoch": 0.77, + "learning_rate": 2.5593753580304946e-05, + "loss": 0.6907, + "step": 26001 + }, + { + "epoch": 0.77, + "learning_rate": 2.5587311737057064e-05, + "loss": 0.5006, + "step": 26002 + }, + { + "epoch": 0.77, + "learning_rate": 2.5580870585668672e-05, + "loss": 0.5309, + "step": 26003 + }, + { + "epoch": 0.77, + "learning_rate": 2.5574430126199712e-05, + "loss": 0.4945, + "step": 26004 + }, + { + "epoch": 0.77, + "learning_rate": 2.5567990358710016e-05, + "loss": 0.5541, + "step": 26005 + }, + { + "epoch": 0.77, + "learning_rate": 2.55615512832595e-05, + "loss": 0.5196, + "step": 26006 + }, + { + "epoch": 0.77, + "learning_rate": 2.5555112899907984e-05, + "loss": 0.5166, + "step": 26007 + }, + { + "epoch": 0.77, + "learning_rate": 2.554867520871539e-05, + "loss": 0.4545, + "step": 26008 + }, + { + "epoch": 0.77, + "learning_rate": 2.554223820974151e-05, + "loss": 0.5102, + "step": 26009 + }, + { + "epoch": 0.77, + "learning_rate": 2.5535801903046253e-05, + "loss": 0.5192, + "step": 26010 + }, + { + "epoch": 0.77, + "learning_rate": 2.5529366288689405e-05, + "loss": 0.5989, + "step": 26011 + }, + { + "epoch": 0.77, + "learning_rate": 2.552293136673083e-05, + "loss": 0.7258, + "step": 26012 + }, + { + "epoch": 0.77, + "learning_rate": 2.5516497137230377e-05, + "loss": 0.4717, + "step": 26013 + }, + { + "epoch": 0.77, + "learning_rate": 2.5510063600247845e-05, + "loss": 0.5953, + "step": 26014 + }, + { + "epoch": 0.77, + "learning_rate": 2.5503630755843033e-05, + "loss": 0.6266, + "step": 26015 + }, + { + "epoch": 0.77, + "learning_rate": 2.5497198604075768e-05, + "loss": 0.4906, + "step": 26016 + }, + { + "epoch": 0.77, + "learning_rate": 2.5490767145005857e-05, + "loss": 0.5587, + "step": 26017 + }, + { + "epoch": 0.77, + "learning_rate": 2.548433637869315e-05, + "loss": 0.5115, + "step": 26018 + }, + { + "epoch": 0.77, + "learning_rate": 2.547790630519733e-05, + "loss": 0.5237, + "step": 26019 + }, + { + "epoch": 0.77, + "learning_rate": 2.547147692457823e-05, + "loss": 0.4287, + "step": 26020 + }, + { + "epoch": 0.77, + "learning_rate": 2.546504823689564e-05, + "loss": 0.4578, + "step": 26021 + }, + { + "epoch": 0.77, + "learning_rate": 2.545862024220935e-05, + "loss": 0.5444, + "step": 26022 + }, + { + "epoch": 0.77, + "learning_rate": 2.54521929405791e-05, + "loss": 0.5868, + "step": 26023 + }, + { + "epoch": 0.77, + "learning_rate": 2.5445766332064623e-05, + "loss": 0.5025, + "step": 26024 + }, + { + "epoch": 0.77, + "learning_rate": 2.54393404167257e-05, + "loss": 0.4873, + "step": 26025 + }, + { + "epoch": 0.77, + "learning_rate": 2.54329151946221e-05, + "loss": 0.504, + "step": 26026 + }, + { + "epoch": 0.77, + "learning_rate": 2.5426490665813495e-05, + "loss": 0.5213, + "step": 26027 + }, + { + "epoch": 0.77, + "learning_rate": 2.542006683035969e-05, + "loss": 0.5386, + "step": 26028 + }, + { + "epoch": 0.77, + "learning_rate": 2.5413643688320387e-05, + "loss": 0.4626, + "step": 26029 + }, + { + "epoch": 0.77, + "learning_rate": 2.5407221239755265e-05, + "loss": 0.5072, + "step": 26030 + }, + { + "epoch": 0.77, + "learning_rate": 2.540079948472407e-05, + "loss": 0.5238, + "step": 26031 + }, + { + "epoch": 0.78, + "learning_rate": 2.5394378423286537e-05, + "loss": 0.526, + "step": 26032 + }, + { + "epoch": 0.78, + "learning_rate": 2.538795805550235e-05, + "loss": 0.543, + "step": 26033 + }, + { + "epoch": 0.78, + "learning_rate": 2.538153838143115e-05, + "loss": 0.5657, + "step": 26034 + }, + { + "epoch": 0.78, + "learning_rate": 2.5375119401132665e-05, + "loss": 0.5473, + "step": 26035 + }, + { + "epoch": 0.78, + "learning_rate": 2.5368701114666583e-05, + "loss": 0.4962, + "step": 26036 + }, + { + "epoch": 0.78, + "learning_rate": 2.536228352209259e-05, + "loss": 0.5806, + "step": 26037 + }, + { + "epoch": 0.78, + "learning_rate": 2.5355866623470347e-05, + "loss": 0.4048, + "step": 26038 + }, + { + "epoch": 0.78, + "learning_rate": 2.5349450418859478e-05, + "loss": 0.5, + "step": 26039 + }, + { + "epoch": 0.78, + "learning_rate": 2.534303490831966e-05, + "loss": 0.524, + "step": 26040 + }, + { + "epoch": 0.78, + "learning_rate": 2.5336620091910592e-05, + "loss": 0.5633, + "step": 26041 + }, + { + "epoch": 0.78, + "learning_rate": 2.5330205969691857e-05, + "loss": 0.399, + "step": 26042 + }, + { + "epoch": 0.78, + "learning_rate": 2.532379254172309e-05, + "loss": 0.514, + "step": 26043 + }, + { + "epoch": 0.78, + "learning_rate": 2.5317379808063924e-05, + "loss": 0.5404, + "step": 26044 + }, + { + "epoch": 0.78, + "learning_rate": 2.531096776877403e-05, + "loss": 0.5288, + "step": 26045 + }, + { + "epoch": 0.78, + "learning_rate": 2.530455642391296e-05, + "loss": 0.5105, + "step": 26046 + }, + { + "epoch": 0.78, + "learning_rate": 2.529814577354038e-05, + "loss": 0.5835, + "step": 26047 + }, + { + "epoch": 0.78, + "learning_rate": 2.5291735817715846e-05, + "loss": 0.5341, + "step": 26048 + }, + { + "epoch": 0.78, + "learning_rate": 2.5285326556499e-05, + "loss": 0.5905, + "step": 26049 + }, + { + "epoch": 0.78, + "learning_rate": 2.527891798994938e-05, + "loss": 0.5451, + "step": 26050 + }, + { + "epoch": 0.78, + "learning_rate": 2.5272510118126635e-05, + "loss": 0.535, + "step": 26051 + }, + { + "epoch": 0.78, + "learning_rate": 2.5266102941090275e-05, + "loss": 0.5143, + "step": 26052 + }, + { + "epoch": 0.78, + "learning_rate": 2.5259696458899927e-05, + "loss": 0.5218, + "step": 26053 + }, + { + "epoch": 0.78, + "learning_rate": 2.5253290671615104e-05, + "loss": 0.5151, + "step": 26054 + }, + { + "epoch": 0.78, + "learning_rate": 2.5246885579295397e-05, + "loss": 0.4119, + "step": 26055 + }, + { + "epoch": 0.78, + "learning_rate": 2.5240481182000385e-05, + "loss": 0.5619, + "step": 26056 + }, + { + "epoch": 0.78, + "learning_rate": 2.5234077479789585e-05, + "loss": 0.5649, + "step": 26057 + }, + { + "epoch": 0.78, + "learning_rate": 2.5227674472722506e-05, + "loss": 0.5631, + "step": 26058 + }, + { + "epoch": 0.78, + "learning_rate": 2.52212721608587e-05, + "loss": 0.463, + "step": 26059 + }, + { + "epoch": 0.78, + "learning_rate": 2.5214870544257706e-05, + "loss": 0.557, + "step": 26060 + }, + { + "epoch": 0.78, + "learning_rate": 2.520846962297909e-05, + "loss": 0.4663, + "step": 26061 + }, + { + "epoch": 0.78, + "learning_rate": 2.5202069397082272e-05, + "loss": 0.6041, + "step": 26062 + }, + { + "epoch": 0.78, + "learning_rate": 2.5195669866626793e-05, + "loss": 0.4973, + "step": 26063 + }, + { + "epoch": 0.78, + "learning_rate": 2.5189271031672192e-05, + "loss": 0.6413, + "step": 26064 + }, + { + "epoch": 0.78, + "learning_rate": 2.51828728922779e-05, + "loss": 0.5541, + "step": 26065 + }, + { + "epoch": 0.78, + "learning_rate": 2.5176475448503468e-05, + "loss": 0.6066, + "step": 26066 + }, + { + "epoch": 0.78, + "learning_rate": 2.5170078700408306e-05, + "loss": 0.518, + "step": 26067 + }, + { + "epoch": 0.78, + "learning_rate": 2.5163682648051967e-05, + "loss": 0.4651, + "step": 26068 + }, + { + "epoch": 0.78, + "learning_rate": 2.5157287291493858e-05, + "loss": 0.5793, + "step": 26069 + }, + { + "epoch": 0.78, + "learning_rate": 2.5150892630793454e-05, + "loss": 0.507, + "step": 26070 + }, + { + "epoch": 0.78, + "learning_rate": 2.514449866601024e-05, + "loss": 0.562, + "step": 26071 + }, + { + "epoch": 0.78, + "learning_rate": 2.513810539720366e-05, + "loss": 0.4941, + "step": 26072 + }, + { + "epoch": 0.78, + "learning_rate": 2.5131712824433097e-05, + "loss": 0.5177, + "step": 26073 + }, + { + "epoch": 0.78, + "learning_rate": 2.5125320947758036e-05, + "loss": 0.4949, + "step": 26074 + }, + { + "epoch": 0.78, + "learning_rate": 2.511892976723792e-05, + "loss": 0.4939, + "step": 26075 + }, + { + "epoch": 0.78, + "learning_rate": 2.5112539282932156e-05, + "loss": 0.562, + "step": 26076 + }, + { + "epoch": 0.78, + "learning_rate": 2.5106149494900122e-05, + "loss": 0.4533, + "step": 26077 + }, + { + "epoch": 0.78, + "learning_rate": 2.5099760403201266e-05, + "loss": 0.435, + "step": 26078 + }, + { + "epoch": 0.78, + "learning_rate": 2.5093372007894977e-05, + "loss": 0.4943, + "step": 26079 + }, + { + "epoch": 0.78, + "learning_rate": 2.5086984309040694e-05, + "loss": 0.5527, + "step": 26080 + }, + { + "epoch": 0.78, + "learning_rate": 2.5080597306697772e-05, + "loss": 0.523, + "step": 26081 + }, + { + "epoch": 0.78, + "learning_rate": 2.5074211000925575e-05, + "loss": 0.5498, + "step": 26082 + }, + { + "epoch": 0.78, + "learning_rate": 2.5067825391783495e-05, + "loss": 0.6113, + "step": 26083 + }, + { + "epoch": 0.78, + "learning_rate": 2.5061440479330944e-05, + "loss": 0.4797, + "step": 26084 + }, + { + "epoch": 0.78, + "learning_rate": 2.505505626362724e-05, + "loss": 0.5662, + "step": 26085 + }, + { + "epoch": 0.78, + "learning_rate": 2.504867274473174e-05, + "loss": 0.5272, + "step": 26086 + }, + { + "epoch": 0.78, + "learning_rate": 2.5042289922703787e-05, + "loss": 0.5121, + "step": 26087 + }, + { + "epoch": 0.78, + "learning_rate": 2.50359077976028e-05, + "loss": 0.5863, + "step": 26088 + }, + { + "epoch": 0.78, + "learning_rate": 2.5029526369488022e-05, + "loss": 0.5947, + "step": 26089 + }, + { + "epoch": 0.78, + "learning_rate": 2.5023145638418853e-05, + "loss": 0.6006, + "step": 26090 + }, + { + "epoch": 0.78, + "learning_rate": 2.501676560445457e-05, + "loss": 0.5762, + "step": 26091 + }, + { + "epoch": 0.78, + "learning_rate": 2.501038626765454e-05, + "loss": 0.5967, + "step": 26092 + }, + { + "epoch": 0.78, + "learning_rate": 2.500400762807803e-05, + "loss": 0.4577, + "step": 26093 + }, + { + "epoch": 0.78, + "learning_rate": 2.4997629685784385e-05, + "loss": 0.5829, + "step": 26094 + }, + { + "epoch": 0.78, + "learning_rate": 2.499125244083289e-05, + "loss": 0.4712, + "step": 26095 + }, + { + "epoch": 0.78, + "learning_rate": 2.498487589328281e-05, + "loss": 0.5289, + "step": 26096 + }, + { + "epoch": 0.78, + "learning_rate": 2.4978500043193454e-05, + "loss": 0.603, + "step": 26097 + }, + { + "epoch": 0.78, + "learning_rate": 2.4972124890624105e-05, + "loss": 0.4387, + "step": 26098 + }, + { + "epoch": 0.78, + "learning_rate": 2.496575043563406e-05, + "loss": 0.4616, + "step": 26099 + }, + { + "epoch": 0.78, + "learning_rate": 2.4959376678282566e-05, + "loss": 0.4975, + "step": 26100 + }, + { + "epoch": 0.78, + "learning_rate": 2.4953003618628844e-05, + "loss": 0.5683, + "step": 26101 + }, + { + "epoch": 0.78, + "learning_rate": 2.494663125673219e-05, + "loss": 0.5071, + "step": 26102 + }, + { + "epoch": 0.78, + "learning_rate": 2.494025959265187e-05, + "loss": 0.5545, + "step": 26103 + }, + { + "epoch": 0.78, + "learning_rate": 2.4933888626447112e-05, + "loss": 0.5445, + "step": 26104 + }, + { + "epoch": 0.78, + "learning_rate": 2.4927518358177095e-05, + "loss": 0.5407, + "step": 26105 + }, + { + "epoch": 0.78, + "learning_rate": 2.4921148787901095e-05, + "loss": 0.5676, + "step": 26106 + }, + { + "epoch": 0.78, + "learning_rate": 2.4914779915678364e-05, + "loss": 0.655, + "step": 26107 + }, + { + "epoch": 0.78, + "learning_rate": 2.490841174156805e-05, + "loss": 0.6001, + "step": 26108 + }, + { + "epoch": 0.78, + "learning_rate": 2.490204426562943e-05, + "loss": 0.4948, + "step": 26109 + }, + { + "epoch": 0.78, + "learning_rate": 2.489567748792163e-05, + "loss": 0.5297, + "step": 26110 + }, + { + "epoch": 0.78, + "learning_rate": 2.4889311408503935e-05, + "loss": 0.6186, + "step": 26111 + }, + { + "epoch": 0.78, + "learning_rate": 2.488294602743544e-05, + "loss": 0.5339, + "step": 26112 + }, + { + "epoch": 0.78, + "learning_rate": 2.4876581344775406e-05, + "loss": 0.4983, + "step": 26113 + }, + { + "epoch": 0.78, + "learning_rate": 2.487021736058296e-05, + "loss": 0.5464, + "step": 26114 + }, + { + "epoch": 0.78, + "learning_rate": 2.4863854074917315e-05, + "loss": 0.4986, + "step": 26115 + }, + { + "epoch": 0.78, + "learning_rate": 2.485749148783757e-05, + "loss": 0.6259, + "step": 26116 + }, + { + "epoch": 0.78, + "learning_rate": 2.4851129599402922e-05, + "loss": 0.5846, + "step": 26117 + }, + { + "epoch": 0.78, + "learning_rate": 2.484476840967256e-05, + "loss": 0.5679, + "step": 26118 + }, + { + "epoch": 0.78, + "learning_rate": 2.4838407918705574e-05, + "loss": 0.6154, + "step": 26119 + }, + { + "epoch": 0.78, + "learning_rate": 2.483204812656109e-05, + "loss": 0.4828, + "step": 26120 + }, + { + "epoch": 0.78, + "learning_rate": 2.482568903329826e-05, + "loss": 0.6222, + "step": 26121 + }, + { + "epoch": 0.78, + "learning_rate": 2.4819330638976224e-05, + "loss": 0.46, + "step": 26122 + }, + { + "epoch": 0.78, + "learning_rate": 2.4812972943654132e-05, + "loss": 0.5364, + "step": 26123 + }, + { + "epoch": 0.78, + "learning_rate": 2.4806615947390998e-05, + "loss": 0.4615, + "step": 26124 + }, + { + "epoch": 0.78, + "learning_rate": 2.4800259650245982e-05, + "loss": 0.6149, + "step": 26125 + }, + { + "epoch": 0.78, + "learning_rate": 2.4793904052278204e-05, + "loss": 0.5549, + "step": 26126 + }, + { + "epoch": 0.78, + "learning_rate": 2.4787549153546707e-05, + "loss": 0.459, + "step": 26127 + }, + { + "epoch": 0.78, + "learning_rate": 2.478119495411063e-05, + "loss": 0.6025, + "step": 26128 + }, + { + "epoch": 0.78, + "learning_rate": 2.4774841454028997e-05, + "loss": 0.4439, + "step": 26129 + }, + { + "epoch": 0.78, + "learning_rate": 2.4768488653360933e-05, + "loss": 0.5088, + "step": 26130 + }, + { + "epoch": 0.78, + "learning_rate": 2.476213655216545e-05, + "loss": 0.4572, + "step": 26131 + }, + { + "epoch": 0.78, + "learning_rate": 2.475578515050163e-05, + "loss": 0.5075, + "step": 26132 + }, + { + "epoch": 0.78, + "learning_rate": 2.4749434448428567e-05, + "loss": 0.4203, + "step": 26133 + }, + { + "epoch": 0.78, + "learning_rate": 2.474308444600526e-05, + "loss": 0.454, + "step": 26134 + }, + { + "epoch": 0.78, + "learning_rate": 2.4736735143290733e-05, + "loss": 0.5213, + "step": 26135 + }, + { + "epoch": 0.78, + "learning_rate": 2.4730386540344053e-05, + "loss": 0.4481, + "step": 26136 + }, + { + "epoch": 0.78, + "learning_rate": 2.4724038637224268e-05, + "loss": 0.4372, + "step": 26137 + }, + { + "epoch": 0.78, + "learning_rate": 2.4717691433990366e-05, + "loss": 0.524, + "step": 26138 + }, + { + "epoch": 0.78, + "learning_rate": 2.471134493070133e-05, + "loss": 0.6374, + "step": 26139 + }, + { + "epoch": 0.78, + "learning_rate": 2.4704999127416207e-05, + "loss": 0.5488, + "step": 26140 + }, + { + "epoch": 0.78, + "learning_rate": 2.469865402419399e-05, + "loss": 0.5934, + "step": 26141 + }, + { + "epoch": 0.78, + "learning_rate": 2.46923096210937e-05, + "loss": 0.5194, + "step": 26142 + }, + { + "epoch": 0.78, + "learning_rate": 2.4685965918174314e-05, + "loss": 0.532, + "step": 26143 + }, + { + "epoch": 0.78, + "learning_rate": 2.467962291549476e-05, + "loss": 0.4988, + "step": 26144 + }, + { + "epoch": 0.78, + "learning_rate": 2.4673280613114057e-05, + "loss": 0.518, + "step": 26145 + }, + { + "epoch": 0.78, + "learning_rate": 2.4666939011091194e-05, + "loss": 0.6249, + "step": 26146 + }, + { + "epoch": 0.78, + "learning_rate": 2.466059810948511e-05, + "loss": 0.628, + "step": 26147 + }, + { + "epoch": 0.78, + "learning_rate": 2.465425790835474e-05, + "loss": 0.5673, + "step": 26148 + }, + { + "epoch": 0.78, + "learning_rate": 2.464791840775903e-05, + "loss": 0.4896, + "step": 26149 + }, + { + "epoch": 0.78, + "learning_rate": 2.4641579607756982e-05, + "loss": 0.5893, + "step": 26150 + }, + { + "epoch": 0.78, + "learning_rate": 2.4635241508407457e-05, + "loss": 0.6302, + "step": 26151 + }, + { + "epoch": 0.78, + "learning_rate": 2.4628904109769458e-05, + "loss": 0.5776, + "step": 26152 + }, + { + "epoch": 0.78, + "learning_rate": 2.4622567411901833e-05, + "loss": 0.4394, + "step": 26153 + }, + { + "epoch": 0.78, + "learning_rate": 2.4616231414863566e-05, + "loss": 0.466, + "step": 26154 + }, + { + "epoch": 0.78, + "learning_rate": 2.4609896118713492e-05, + "loss": 0.4659, + "step": 26155 + }, + { + "epoch": 0.78, + "learning_rate": 2.46035615235106e-05, + "loss": 0.574, + "step": 26156 + }, + { + "epoch": 0.78, + "learning_rate": 2.4597227629313712e-05, + "loss": 0.5901, + "step": 26157 + }, + { + "epoch": 0.78, + "learning_rate": 2.4590894436181777e-05, + "loss": 0.4334, + "step": 26158 + }, + { + "epoch": 0.78, + "learning_rate": 2.458456194417362e-05, + "loss": 0.4066, + "step": 26159 + }, + { + "epoch": 0.78, + "learning_rate": 2.4578230153348146e-05, + "loss": 0.5356, + "step": 26160 + }, + { + "epoch": 0.78, + "learning_rate": 2.457189906376426e-05, + "loss": 0.4883, + "step": 26161 + }, + { + "epoch": 0.78, + "learning_rate": 2.456556867548079e-05, + "loss": 0.4893, + "step": 26162 + }, + { + "epoch": 0.78, + "learning_rate": 2.4559238988556575e-05, + "loss": 0.4979, + "step": 26163 + }, + { + "epoch": 0.78, + "learning_rate": 2.4552910003050488e-05, + "loss": 0.529, + "step": 26164 + }, + { + "epoch": 0.78, + "learning_rate": 2.45465817190214e-05, + "loss": 0.5367, + "step": 26165 + }, + { + "epoch": 0.78, + "learning_rate": 2.4540254136528118e-05, + "loss": 0.4571, + "step": 26166 + }, + { + "epoch": 0.78, + "learning_rate": 2.453392725562945e-05, + "loss": 0.4974, + "step": 26167 + }, + { + "epoch": 0.78, + "learning_rate": 2.4527601076384254e-05, + "loss": 0.5291, + "step": 26168 + }, + { + "epoch": 0.78, + "learning_rate": 2.452127559885137e-05, + "loss": 0.5363, + "step": 26169 + }, + { + "epoch": 0.78, + "learning_rate": 2.4514950823089543e-05, + "loss": 0.4918, + "step": 26170 + }, + { + "epoch": 0.78, + "learning_rate": 2.450862674915766e-05, + "loss": 0.5193, + "step": 26171 + }, + { + "epoch": 0.78, + "learning_rate": 2.4502303377114443e-05, + "loss": 0.5047, + "step": 26172 + }, + { + "epoch": 0.78, + "learning_rate": 2.449598070701875e-05, + "loss": 0.4915, + "step": 26173 + }, + { + "epoch": 0.78, + "learning_rate": 2.4489658738929312e-05, + "loss": 0.6066, + "step": 26174 + }, + { + "epoch": 0.78, + "learning_rate": 2.4483337472904953e-05, + "loss": 0.4857, + "step": 26175 + }, + { + "epoch": 0.78, + "learning_rate": 2.4477016909004392e-05, + "loss": 0.4588, + "step": 26176 + }, + { + "epoch": 0.78, + "learning_rate": 2.4470697047286462e-05, + "loss": 0.6044, + "step": 26177 + }, + { + "epoch": 0.78, + "learning_rate": 2.4464377887809865e-05, + "loss": 0.5807, + "step": 26178 + }, + { + "epoch": 0.78, + "learning_rate": 2.4458059430633363e-05, + "loss": 0.5039, + "step": 26179 + }, + { + "epoch": 0.78, + "learning_rate": 2.4451741675815754e-05, + "loss": 0.5478, + "step": 26180 + }, + { + "epoch": 0.78, + "learning_rate": 2.4445424623415735e-05, + "loss": 0.529, + "step": 26181 + }, + { + "epoch": 0.78, + "learning_rate": 2.4439108273492027e-05, + "loss": 0.5589, + "step": 26182 + }, + { + "epoch": 0.78, + "learning_rate": 2.4432792626103362e-05, + "loss": 0.5634, + "step": 26183 + }, + { + "epoch": 0.78, + "learning_rate": 2.442647768130848e-05, + "loss": 0.4267, + "step": 26184 + }, + { + "epoch": 0.78, + "learning_rate": 2.4420163439166145e-05, + "loss": 0.5258, + "step": 26185 + }, + { + "epoch": 0.78, + "learning_rate": 2.441384989973495e-05, + "loss": 0.5437, + "step": 26186 + }, + { + "epoch": 0.78, + "learning_rate": 2.440753706307365e-05, + "loss": 0.4448, + "step": 26187 + }, + { + "epoch": 0.78, + "learning_rate": 2.440122492924094e-05, + "loss": 0.5285, + "step": 26188 + }, + { + "epoch": 0.78, + "learning_rate": 2.4394913498295546e-05, + "loss": 0.4882, + "step": 26189 + }, + { + "epoch": 0.78, + "learning_rate": 2.438860277029611e-05, + "loss": 0.4426, + "step": 26190 + }, + { + "epoch": 0.78, + "learning_rate": 2.4382292745301282e-05, + "loss": 0.5857, + "step": 26191 + }, + { + "epoch": 0.78, + "learning_rate": 2.4375983423369786e-05, + "loss": 0.5207, + "step": 26192 + }, + { + "epoch": 0.78, + "learning_rate": 2.4369674804560215e-05, + "loss": 0.5413, + "step": 26193 + }, + { + "epoch": 0.78, + "learning_rate": 2.436336688893128e-05, + "loss": 0.5368, + "step": 26194 + }, + { + "epoch": 0.78, + "learning_rate": 2.435705967654164e-05, + "loss": 0.5438, + "step": 26195 + }, + { + "epoch": 0.78, + "learning_rate": 2.43507531674499e-05, + "loss": 0.5709, + "step": 26196 + }, + { + "epoch": 0.78, + "learning_rate": 2.4344447361714695e-05, + "loss": 0.5349, + "step": 26197 + }, + { + "epoch": 0.78, + "learning_rate": 2.4338142259394646e-05, + "loss": 0.5042, + "step": 26198 + }, + { + "epoch": 0.78, + "learning_rate": 2.4331837860548433e-05, + "loss": 0.5177, + "step": 26199 + }, + { + "epoch": 0.78, + "learning_rate": 2.4325534165234643e-05, + "loss": 0.4185, + "step": 26200 + }, + { + "epoch": 0.78, + "learning_rate": 2.4319231173511837e-05, + "loss": 0.4871, + "step": 26201 + }, + { + "epoch": 0.78, + "learning_rate": 2.4312928885438657e-05, + "loss": 0.5749, + "step": 26202 + }, + { + "epoch": 0.78, + "learning_rate": 2.4306627301073693e-05, + "loss": 0.5782, + "step": 26203 + }, + { + "epoch": 0.78, + "learning_rate": 2.430032642047557e-05, + "loss": 0.4701, + "step": 26204 + }, + { + "epoch": 0.78, + "learning_rate": 2.429402624370285e-05, + "loss": 0.55, + "step": 26205 + }, + { + "epoch": 0.78, + "learning_rate": 2.428772677081407e-05, + "loss": 0.5301, + "step": 26206 + }, + { + "epoch": 0.78, + "learning_rate": 2.4281428001867824e-05, + "loss": 0.4584, + "step": 26207 + }, + { + "epoch": 0.78, + "learning_rate": 2.4275129936922713e-05, + "loss": 0.4353, + "step": 26208 + }, + { + "epoch": 0.78, + "learning_rate": 2.426883257603726e-05, + "loss": 0.4218, + "step": 26209 + }, + { + "epoch": 0.78, + "learning_rate": 2.426253591926999e-05, + "loss": 0.6221, + "step": 26210 + }, + { + "epoch": 0.78, + "learning_rate": 2.4256239966679485e-05, + "loss": 0.559, + "step": 26211 + }, + { + "epoch": 0.78, + "learning_rate": 2.4249944718324292e-05, + "loss": 0.5081, + "step": 26212 + }, + { + "epoch": 0.78, + "learning_rate": 2.424365017426289e-05, + "loss": 0.5012, + "step": 26213 + }, + { + "epoch": 0.78, + "learning_rate": 2.4237356334553873e-05, + "loss": 0.5661, + "step": 26214 + }, + { + "epoch": 0.78, + "learning_rate": 2.423106319925569e-05, + "loss": 0.6011, + "step": 26215 + }, + { + "epoch": 0.78, + "learning_rate": 2.4224770768426908e-05, + "loss": 0.5522, + "step": 26216 + }, + { + "epoch": 0.78, + "learning_rate": 2.4218479042125975e-05, + "loss": 0.4628, + "step": 26217 + }, + { + "epoch": 0.78, + "learning_rate": 2.4212188020411453e-05, + "loss": 0.4639, + "step": 26218 + }, + { + "epoch": 0.78, + "learning_rate": 2.4205897703341773e-05, + "loss": 0.4812, + "step": 26219 + }, + { + "epoch": 0.78, + "learning_rate": 2.419960809097548e-05, + "loss": 0.5087, + "step": 26220 + }, + { + "epoch": 0.78, + "learning_rate": 2.4193319183370988e-05, + "loss": 0.4418, + "step": 26221 + }, + { + "epoch": 0.78, + "learning_rate": 2.41870309805868e-05, + "loss": 0.4545, + "step": 26222 + }, + { + "epoch": 0.78, + "learning_rate": 2.41807434826814e-05, + "loss": 0.7098, + "step": 26223 + }, + { + "epoch": 0.78, + "learning_rate": 2.4174456689713244e-05, + "loss": 0.572, + "step": 26224 + }, + { + "epoch": 0.78, + "learning_rate": 2.4168170601740735e-05, + "loss": 0.552, + "step": 26225 + }, + { + "epoch": 0.78, + "learning_rate": 2.4161885218822355e-05, + "loss": 0.5307, + "step": 26226 + }, + { + "epoch": 0.78, + "learning_rate": 2.4155600541016565e-05, + "loss": 0.4683, + "step": 26227 + }, + { + "epoch": 0.78, + "learning_rate": 2.4149316568381774e-05, + "loss": 0.6405, + "step": 26228 + }, + { + "epoch": 0.78, + "learning_rate": 2.414303330097637e-05, + "loss": 0.5287, + "step": 26229 + }, + { + "epoch": 0.78, + "learning_rate": 2.4136750738858816e-05, + "loss": 0.5068, + "step": 26230 + }, + { + "epoch": 0.78, + "learning_rate": 2.4130468882087543e-05, + "loss": 0.4581, + "step": 26231 + }, + { + "epoch": 0.78, + "learning_rate": 2.4124187730720917e-05, + "loss": 0.5427, + "step": 26232 + }, + { + "epoch": 0.78, + "learning_rate": 2.4117907284817364e-05, + "loss": 0.5652, + "step": 26233 + }, + { + "epoch": 0.78, + "learning_rate": 2.4111627544435243e-05, + "loss": 0.4888, + "step": 26234 + }, + { + "epoch": 0.78, + "learning_rate": 2.4105348509632984e-05, + "loss": 0.6752, + "step": 26235 + }, + { + "epoch": 0.78, + "learning_rate": 2.409907018046893e-05, + "loss": 0.527, + "step": 26236 + }, + { + "epoch": 0.78, + "learning_rate": 2.4092792557001465e-05, + "loss": 0.5547, + "step": 26237 + }, + { + "epoch": 0.78, + "learning_rate": 2.4086515639288988e-05, + "loss": 0.5406, + "step": 26238 + }, + { + "epoch": 0.78, + "learning_rate": 2.408023942738984e-05, + "loss": 0.584, + "step": 26239 + }, + { + "epoch": 0.78, + "learning_rate": 2.407396392136233e-05, + "loss": 0.5883, + "step": 26240 + }, + { + "epoch": 0.78, + "learning_rate": 2.406768912126485e-05, + "loss": 0.5442, + "step": 26241 + }, + { + "epoch": 0.78, + "learning_rate": 2.4061415027155755e-05, + "loss": 0.54, + "step": 26242 + }, + { + "epoch": 0.78, + "learning_rate": 2.405514163909336e-05, + "loss": 0.5819, + "step": 26243 + }, + { + "epoch": 0.78, + "learning_rate": 2.404886895713595e-05, + "loss": 0.5574, + "step": 26244 + }, + { + "epoch": 0.78, + "learning_rate": 2.40425969813419e-05, + "loss": 0.5132, + "step": 26245 + }, + { + "epoch": 0.78, + "learning_rate": 2.4036325711769504e-05, + "loss": 0.6185, + "step": 26246 + }, + { + "epoch": 0.78, + "learning_rate": 2.4030055148477106e-05, + "loss": 0.5383, + "step": 26247 + }, + { + "epoch": 0.78, + "learning_rate": 2.402378529152297e-05, + "loss": 0.5405, + "step": 26248 + }, + { + "epoch": 0.78, + "learning_rate": 2.401751614096538e-05, + "loss": 0.62, + "step": 26249 + }, + { + "epoch": 0.78, + "learning_rate": 2.4011247696862625e-05, + "loss": 0.5919, + "step": 26250 + }, + { + "epoch": 0.78, + "learning_rate": 2.4004979959273043e-05, + "loss": 0.6497, + "step": 26251 + }, + { + "epoch": 0.78, + "learning_rate": 2.3998712928254873e-05, + "loss": 0.6265, + "step": 26252 + }, + { + "epoch": 0.78, + "learning_rate": 2.3992446603866346e-05, + "loss": 0.6857, + "step": 26253 + }, + { + "epoch": 0.78, + "learning_rate": 2.3986180986165752e-05, + "loss": 0.4884, + "step": 26254 + }, + { + "epoch": 0.78, + "learning_rate": 2.3979916075211384e-05, + "loss": 0.6766, + "step": 26255 + }, + { + "epoch": 0.78, + "learning_rate": 2.3973651871061432e-05, + "loss": 0.5428, + "step": 26256 + }, + { + "epoch": 0.78, + "learning_rate": 2.396738837377418e-05, + "loss": 0.5453, + "step": 26257 + }, + { + "epoch": 0.78, + "learning_rate": 2.3961125583407863e-05, + "loss": 0.4484, + "step": 26258 + }, + { + "epoch": 0.78, + "learning_rate": 2.395486350002065e-05, + "loss": 0.5645, + "step": 26259 + }, + { + "epoch": 0.78, + "learning_rate": 2.3948602123670817e-05, + "loss": 0.5927, + "step": 26260 + }, + { + "epoch": 0.78, + "learning_rate": 2.3942341454416606e-05, + "loss": 0.5119, + "step": 26261 + }, + { + "epoch": 0.78, + "learning_rate": 2.3936081492316177e-05, + "loss": 0.4681, + "step": 26262 + }, + { + "epoch": 0.78, + "learning_rate": 2.3929822237427722e-05, + "loss": 0.3927, + "step": 26263 + }, + { + "epoch": 0.78, + "learning_rate": 2.392356368980947e-05, + "loss": 0.4996, + "step": 26264 + }, + { + "epoch": 0.78, + "learning_rate": 2.3917305849519588e-05, + "loss": 0.4233, + "step": 26265 + }, + { + "epoch": 0.78, + "learning_rate": 2.3911048716616314e-05, + "loss": 0.5134, + "step": 26266 + }, + { + "epoch": 0.78, + "learning_rate": 2.3904792291157774e-05, + "loss": 0.487, + "step": 26267 + }, + { + "epoch": 0.78, + "learning_rate": 2.3898536573202113e-05, + "loss": 0.5756, + "step": 26268 + }, + { + "epoch": 0.78, + "learning_rate": 2.3892281562807538e-05, + "loss": 0.5345, + "step": 26269 + }, + { + "epoch": 0.78, + "learning_rate": 2.388602726003222e-05, + "loss": 0.5253, + "step": 26270 + }, + { + "epoch": 0.78, + "learning_rate": 2.387977366493429e-05, + "loss": 0.5977, + "step": 26271 + }, + { + "epoch": 0.78, + "learning_rate": 2.3873520777571857e-05, + "loss": 0.6376, + "step": 26272 + }, + { + "epoch": 0.78, + "learning_rate": 2.386726859800308e-05, + "loss": 0.567, + "step": 26273 + }, + { + "epoch": 0.78, + "learning_rate": 2.386101712628612e-05, + "loss": 0.5659, + "step": 26274 + }, + { + "epoch": 0.78, + "learning_rate": 2.3854766362479064e-05, + "loss": 0.5984, + "step": 26275 + }, + { + "epoch": 0.78, + "learning_rate": 2.3848516306640056e-05, + "loss": 0.4875, + "step": 26276 + }, + { + "epoch": 0.78, + "learning_rate": 2.3842266958827175e-05, + "loss": 0.537, + "step": 26277 + }, + { + "epoch": 0.78, + "learning_rate": 2.3836018319098564e-05, + "loss": 0.4352, + "step": 26278 + }, + { + "epoch": 0.78, + "learning_rate": 2.3829770387512272e-05, + "loss": 0.5552, + "step": 26279 + }, + { + "epoch": 0.78, + "learning_rate": 2.382352316412645e-05, + "loss": 0.5843, + "step": 26280 + }, + { + "epoch": 0.78, + "learning_rate": 2.3817276648999108e-05, + "loss": 0.6141, + "step": 26281 + }, + { + "epoch": 0.78, + "learning_rate": 2.381103084218841e-05, + "loss": 0.5964, + "step": 26282 + }, + { + "epoch": 0.78, + "learning_rate": 2.3804785743752335e-05, + "loss": 0.5123, + "step": 26283 + }, + { + "epoch": 0.78, + "learning_rate": 2.3798541353749004e-05, + "loss": 0.5174, + "step": 26284 + }, + { + "epoch": 0.78, + "learning_rate": 2.3792297672236485e-05, + "loss": 0.5637, + "step": 26285 + }, + { + "epoch": 0.78, + "learning_rate": 2.378605469927282e-05, + "loss": 0.6001, + "step": 26286 + }, + { + "epoch": 0.78, + "learning_rate": 2.3779812434916004e-05, + "loss": 0.5536, + "step": 26287 + }, + { + "epoch": 0.78, + "learning_rate": 2.3773570879224126e-05, + "loss": 0.5998, + "step": 26288 + }, + { + "epoch": 0.78, + "learning_rate": 2.3767330032255198e-05, + "loss": 0.5535, + "step": 26289 + }, + { + "epoch": 0.78, + "learning_rate": 2.37610898940673e-05, + "loss": 0.5452, + "step": 26290 + }, + { + "epoch": 0.78, + "learning_rate": 2.375485046471836e-05, + "loss": 0.6236, + "step": 26291 + }, + { + "epoch": 0.78, + "learning_rate": 2.3748611744266426e-05, + "loss": 0.4407, + "step": 26292 + }, + { + "epoch": 0.78, + "learning_rate": 2.374237373276953e-05, + "loss": 0.5349, + "step": 26293 + }, + { + "epoch": 0.78, + "learning_rate": 2.373613643028564e-05, + "loss": 0.5022, + "step": 26294 + }, + { + "epoch": 0.78, + "learning_rate": 2.3729899836872772e-05, + "loss": 0.5247, + "step": 26295 + }, + { + "epoch": 0.78, + "learning_rate": 2.3723663952588872e-05, + "loss": 0.4715, + "step": 26296 + }, + { + "epoch": 0.78, + "learning_rate": 2.3717428777491967e-05, + "loss": 0.5333, + "step": 26297 + }, + { + "epoch": 0.78, + "learning_rate": 2.3711194311639984e-05, + "loss": 0.587, + "step": 26298 + }, + { + "epoch": 0.78, + "learning_rate": 2.3704960555090904e-05, + "loss": 0.6018, + "step": 26299 + }, + { + "epoch": 0.78, + "learning_rate": 2.369872750790273e-05, + "loss": 0.5296, + "step": 26300 + }, + { + "epoch": 0.78, + "learning_rate": 2.3692495170133378e-05, + "loss": 0.653, + "step": 26301 + }, + { + "epoch": 0.78, + "learning_rate": 2.368626354184076e-05, + "loss": 0.5179, + "step": 26302 + }, + { + "epoch": 0.78, + "learning_rate": 2.3680032623082838e-05, + "loss": 0.4653, + "step": 26303 + }, + { + "epoch": 0.78, + "learning_rate": 2.367380241391759e-05, + "loss": 0.5372, + "step": 26304 + }, + { + "epoch": 0.78, + "learning_rate": 2.3667572914402912e-05, + "loss": 0.5024, + "step": 26305 + }, + { + "epoch": 0.78, + "learning_rate": 2.3661344124596683e-05, + "loss": 0.4279, + "step": 26306 + }, + { + "epoch": 0.78, + "learning_rate": 2.3655116044556857e-05, + "loss": 0.4134, + "step": 26307 + }, + { + "epoch": 0.78, + "learning_rate": 2.364888867434133e-05, + "loss": 0.4062, + "step": 26308 + }, + { + "epoch": 0.78, + "learning_rate": 2.3642662014008034e-05, + "loss": 0.5196, + "step": 26309 + }, + { + "epoch": 0.78, + "learning_rate": 2.3636436063614842e-05, + "loss": 0.6795, + "step": 26310 + }, + { + "epoch": 0.78, + "learning_rate": 2.363021082321959e-05, + "loss": 0.5095, + "step": 26311 + }, + { + "epoch": 0.78, + "learning_rate": 2.3623986292880206e-05, + "loss": 0.6009, + "step": 26312 + }, + { + "epoch": 0.78, + "learning_rate": 2.361776247265458e-05, + "loss": 0.4982, + "step": 26313 + }, + { + "epoch": 0.78, + "learning_rate": 2.3611539362600567e-05, + "loss": 0.4663, + "step": 26314 + }, + { + "epoch": 0.78, + "learning_rate": 2.360531696277598e-05, + "loss": 0.4999, + "step": 26315 + }, + { + "epoch": 0.78, + "learning_rate": 2.3599095273238702e-05, + "loss": 0.655, + "step": 26316 + }, + { + "epoch": 0.78, + "learning_rate": 2.359287429404663e-05, + "loss": 0.4929, + "step": 26317 + }, + { + "epoch": 0.78, + "learning_rate": 2.3586654025257516e-05, + "loss": 0.4907, + "step": 26318 + }, + { + "epoch": 0.78, + "learning_rate": 2.358043446692928e-05, + "loss": 0.5336, + "step": 26319 + }, + { + "epoch": 0.78, + "learning_rate": 2.3574215619119668e-05, + "loss": 0.5892, + "step": 26320 + }, + { + "epoch": 0.78, + "learning_rate": 2.356799748188657e-05, + "loss": 0.4433, + "step": 26321 + }, + { + "epoch": 0.78, + "learning_rate": 2.3561780055287742e-05, + "loss": 0.6272, + "step": 26322 + }, + { + "epoch": 0.78, + "learning_rate": 2.3555563339381037e-05, + "loss": 0.4915, + "step": 26323 + }, + { + "epoch": 0.78, + "learning_rate": 2.3549347334224246e-05, + "loss": 0.5462, + "step": 26324 + }, + { + "epoch": 0.78, + "learning_rate": 2.3543132039875125e-05, + "loss": 0.5047, + "step": 26325 + }, + { + "epoch": 0.78, + "learning_rate": 2.3536917456391484e-05, + "loss": 0.564, + "step": 26326 + }, + { + "epoch": 0.78, + "learning_rate": 2.3530703583831116e-05, + "loss": 0.6183, + "step": 26327 + }, + { + "epoch": 0.78, + "learning_rate": 2.352449042225181e-05, + "loss": 0.4819, + "step": 26328 + }, + { + "epoch": 0.78, + "learning_rate": 2.351827797171131e-05, + "loss": 0.5489, + "step": 26329 + }, + { + "epoch": 0.78, + "learning_rate": 2.351206623226736e-05, + "loss": 0.5258, + "step": 26330 + }, + { + "epoch": 0.78, + "learning_rate": 2.3505855203977733e-05, + "loss": 0.6004, + "step": 26331 + }, + { + "epoch": 0.78, + "learning_rate": 2.3499644886900198e-05, + "loss": 0.5184, + "step": 26332 + }, + { + "epoch": 0.78, + "learning_rate": 2.3493435281092478e-05, + "loss": 0.6044, + "step": 26333 + }, + { + "epoch": 0.78, + "learning_rate": 2.348722638661227e-05, + "loss": 0.4602, + "step": 26334 + }, + { + "epoch": 0.78, + "learning_rate": 2.3481018203517345e-05, + "loss": 0.5683, + "step": 26335 + }, + { + "epoch": 0.78, + "learning_rate": 2.3474810731865437e-05, + "loss": 0.4225, + "step": 26336 + }, + { + "epoch": 0.78, + "learning_rate": 2.346860397171422e-05, + "loss": 0.47, + "step": 26337 + }, + { + "epoch": 0.78, + "learning_rate": 2.346239792312144e-05, + "loss": 0.5898, + "step": 26338 + }, + { + "epoch": 0.78, + "learning_rate": 2.3456192586144755e-05, + "loss": 0.5708, + "step": 26339 + }, + { + "epoch": 0.78, + "learning_rate": 2.3449987960841912e-05, + "loss": 0.4329, + "step": 26340 + }, + { + "epoch": 0.78, + "learning_rate": 2.3443784047270534e-05, + "loss": 0.4397, + "step": 26341 + }, + { + "epoch": 0.78, + "learning_rate": 2.3437580845488348e-05, + "loss": 0.4451, + "step": 26342 + }, + { + "epoch": 0.78, + "learning_rate": 2.3431378355553047e-05, + "loss": 0.5559, + "step": 26343 + }, + { + "epoch": 0.78, + "learning_rate": 2.3425176577522278e-05, + "loss": 0.4527, + "step": 26344 + }, + { + "epoch": 0.78, + "learning_rate": 2.3418975511453668e-05, + "loss": 0.5278, + "step": 26345 + }, + { + "epoch": 0.78, + "learning_rate": 2.3412775157404897e-05, + "loss": 0.4533, + "step": 26346 + }, + { + "epoch": 0.78, + "learning_rate": 2.340657551543365e-05, + "loss": 0.6014, + "step": 26347 + }, + { + "epoch": 0.78, + "learning_rate": 2.3400376585597527e-05, + "loss": 0.6352, + "step": 26348 + }, + { + "epoch": 0.78, + "learning_rate": 2.3394178367954155e-05, + "loss": 0.4964, + "step": 26349 + }, + { + "epoch": 0.78, + "learning_rate": 2.3387980862561178e-05, + "loss": 0.5113, + "step": 26350 + }, + { + "epoch": 0.78, + "learning_rate": 2.338178406947622e-05, + "loss": 0.4008, + "step": 26351 + }, + { + "epoch": 0.78, + "learning_rate": 2.3375587988756953e-05, + "loss": 0.5795, + "step": 26352 + }, + { + "epoch": 0.78, + "learning_rate": 2.336939262046087e-05, + "loss": 0.5547, + "step": 26353 + }, + { + "epoch": 0.78, + "learning_rate": 2.3363197964645646e-05, + "loss": 0.543, + "step": 26354 + }, + { + "epoch": 0.78, + "learning_rate": 2.3357004021368843e-05, + "loss": 0.5681, + "step": 26355 + }, + { + "epoch": 0.78, + "learning_rate": 2.3350810790688117e-05, + "loss": 0.6058, + "step": 26356 + }, + { + "epoch": 0.78, + "learning_rate": 2.334461827266099e-05, + "loss": 0.5302, + "step": 26357 + }, + { + "epoch": 0.78, + "learning_rate": 2.333842646734502e-05, + "loss": 0.4966, + "step": 26358 + }, + { + "epoch": 0.78, + "learning_rate": 2.333223537479785e-05, + "loss": 0.5851, + "step": 26359 + }, + { + "epoch": 0.78, + "learning_rate": 2.3326044995076956e-05, + "loss": 0.5887, + "step": 26360 + }, + { + "epoch": 0.78, + "learning_rate": 2.3319855328239938e-05, + "loss": 0.5425, + "step": 26361 + }, + { + "epoch": 0.78, + "learning_rate": 2.331366637434438e-05, + "loss": 0.5082, + "step": 26362 + }, + { + "epoch": 0.78, + "learning_rate": 2.330747813344778e-05, + "loss": 0.529, + "step": 26363 + }, + { + "epoch": 0.78, + "learning_rate": 2.3301290605607662e-05, + "loss": 0.5865, + "step": 26364 + }, + { + "epoch": 0.78, + "learning_rate": 2.3295103790881566e-05, + "loss": 0.5387, + "step": 26365 + }, + { + "epoch": 0.78, + "learning_rate": 2.328891768932705e-05, + "loss": 0.6988, + "step": 26366 + }, + { + "epoch": 0.78, + "learning_rate": 2.328273230100161e-05, + "loss": 0.5698, + "step": 26367 + }, + { + "epoch": 0.79, + "learning_rate": 2.327654762596272e-05, + "loss": 0.6438, + "step": 26368 + }, + { + "epoch": 0.79, + "learning_rate": 2.3270363664267915e-05, + "loss": 0.3565, + "step": 26369 + }, + { + "epoch": 0.79, + "learning_rate": 2.326418041597468e-05, + "loss": 0.3616, + "step": 26370 + }, + { + "epoch": 0.79, + "learning_rate": 2.3257997881140536e-05, + "loss": 0.6191, + "step": 26371 + }, + { + "epoch": 0.79, + "learning_rate": 2.325181605982295e-05, + "loss": 0.5067, + "step": 26372 + }, + { + "epoch": 0.79, + "learning_rate": 2.3245634952079364e-05, + "loss": 0.5815, + "step": 26373 + }, + { + "epoch": 0.79, + "learning_rate": 2.3239454557967268e-05, + "loss": 0.6313, + "step": 26374 + }, + { + "epoch": 0.79, + "learning_rate": 2.3233274877544154e-05, + "loss": 0.5249, + "step": 26375 + }, + { + "epoch": 0.79, + "learning_rate": 2.3227095910867457e-05, + "loss": 0.5848, + "step": 26376 + }, + { + "epoch": 0.79, + "learning_rate": 2.322091765799459e-05, + "loss": 0.5657, + "step": 26377 + }, + { + "epoch": 0.79, + "learning_rate": 2.3214740118983036e-05, + "loss": 0.5366, + "step": 26378 + }, + { + "epoch": 0.79, + "learning_rate": 2.320856329389025e-05, + "loss": 0.5603, + "step": 26379 + }, + { + "epoch": 0.79, + "learning_rate": 2.320238718277362e-05, + "loss": 0.4235, + "step": 26380 + }, + { + "epoch": 0.79, + "learning_rate": 2.31962117856906e-05, + "loss": 0.4374, + "step": 26381 + }, + { + "epoch": 0.79, + "learning_rate": 2.319003710269857e-05, + "loss": 0.5927, + "step": 26382 + }, + { + "epoch": 0.79, + "learning_rate": 2.318386313385499e-05, + "loss": 0.5769, + "step": 26383 + }, + { + "epoch": 0.79, + "learning_rate": 2.31776898792172e-05, + "loss": 0.5361, + "step": 26384 + }, + { + "epoch": 0.79, + "learning_rate": 2.3171517338842662e-05, + "loss": 0.5088, + "step": 26385 + }, + { + "epoch": 0.79, + "learning_rate": 2.3165345512788704e-05, + "loss": 0.4414, + "step": 26386 + }, + { + "epoch": 0.79, + "learning_rate": 2.3159174401112783e-05, + "loss": 0.5261, + "step": 26387 + }, + { + "epoch": 0.79, + "learning_rate": 2.3153004003872192e-05, + "loss": 0.6187, + "step": 26388 + }, + { + "epoch": 0.79, + "learning_rate": 2.3146834321124346e-05, + "loss": 0.4938, + "step": 26389 + }, + { + "epoch": 0.79, + "learning_rate": 2.3140665352926626e-05, + "loss": 0.5615, + "step": 26390 + }, + { + "epoch": 0.79, + "learning_rate": 2.313449709933637e-05, + "loss": 0.542, + "step": 26391 + }, + { + "epoch": 0.79, + "learning_rate": 2.31283295604109e-05, + "loss": 0.5691, + "step": 26392 + }, + { + "epoch": 0.79, + "learning_rate": 2.3122162736207576e-05, + "loss": 0.4643, + "step": 26393 + }, + { + "epoch": 0.79, + "learning_rate": 2.3115996626783777e-05, + "loss": 0.4699, + "step": 26394 + }, + { + "epoch": 0.79, + "learning_rate": 2.3109831232196788e-05, + "loss": 0.544, + "step": 26395 + }, + { + "epoch": 0.79, + "learning_rate": 2.3103666552503922e-05, + "loss": 0.4413, + "step": 26396 + }, + { + "epoch": 0.79, + "learning_rate": 2.309750258776252e-05, + "loss": 0.437, + "step": 26397 + }, + { + "epoch": 0.79, + "learning_rate": 2.3091339338029914e-05, + "loss": 0.5713, + "step": 26398 + }, + { + "epoch": 0.79, + "learning_rate": 2.3085176803363352e-05, + "loss": 0.5845, + "step": 26399 + }, + { + "epoch": 0.79, + "learning_rate": 2.307901498382019e-05, + "loss": 0.5501, + "step": 26400 + }, + { + "epoch": 0.79, + "learning_rate": 2.307285387945767e-05, + "loss": 0.4472, + "step": 26401 + }, + { + "epoch": 0.79, + "learning_rate": 2.3066693490333114e-05, + "loss": 0.4366, + "step": 26402 + }, + { + "epoch": 0.79, + "learning_rate": 2.306053381650375e-05, + "loss": 0.5528, + "step": 26403 + }, + { + "epoch": 0.79, + "learning_rate": 2.3054374858026882e-05, + "loss": 0.6481, + "step": 26404 + }, + { + "epoch": 0.79, + "learning_rate": 2.3048216614959795e-05, + "loss": 0.5233, + "step": 26405 + }, + { + "epoch": 0.79, + "learning_rate": 2.3042059087359714e-05, + "loss": 0.5238, + "step": 26406 + }, + { + "epoch": 0.79, + "learning_rate": 2.303590227528387e-05, + "loss": 0.5673, + "step": 26407 + }, + { + "epoch": 0.79, + "learning_rate": 2.3029746178789536e-05, + "loss": 0.5445, + "step": 26408 + }, + { + "epoch": 0.79, + "learning_rate": 2.3023590797933968e-05, + "loss": 0.5369, + "step": 26409 + }, + { + "epoch": 0.79, + "learning_rate": 2.3017436132774382e-05, + "loss": 0.5332, + "step": 26410 + }, + { + "epoch": 0.79, + "learning_rate": 2.3011282183367954e-05, + "loss": 0.4918, + "step": 26411 + }, + { + "epoch": 0.79, + "learning_rate": 2.300512894977195e-05, + "loss": 0.5502, + "step": 26412 + }, + { + "epoch": 0.79, + "learning_rate": 2.2998976432043563e-05, + "loss": 0.4012, + "step": 26413 + }, + { + "epoch": 0.79, + "learning_rate": 2.2992824630240028e-05, + "loss": 0.5099, + "step": 26414 + }, + { + "epoch": 0.79, + "learning_rate": 2.2986673544418525e-05, + "loss": 0.4846, + "step": 26415 + }, + { + "epoch": 0.79, + "learning_rate": 2.2980523174636202e-05, + "loss": 0.4813, + "step": 26416 + }, + { + "epoch": 0.79, + "learning_rate": 2.2974373520950288e-05, + "loss": 0.5865, + "step": 26417 + }, + { + "epoch": 0.79, + "learning_rate": 2.2968224583417965e-05, + "loss": 0.4871, + "step": 26418 + }, + { + "epoch": 0.79, + "learning_rate": 2.2962076362096396e-05, + "loss": 0.4108, + "step": 26419 + }, + { + "epoch": 0.79, + "learning_rate": 2.2955928857042706e-05, + "loss": 0.4876, + "step": 26420 + }, + { + "epoch": 0.79, + "learning_rate": 2.2949782068314084e-05, + "loss": 0.6355, + "step": 26421 + }, + { + "epoch": 0.79, + "learning_rate": 2.2943635995967707e-05, + "loss": 0.5204, + "step": 26422 + }, + { + "epoch": 0.79, + "learning_rate": 2.2937490640060665e-05, + "loss": 0.5063, + "step": 26423 + }, + { + "epoch": 0.79, + "learning_rate": 2.2931346000650146e-05, + "loss": 0.6405, + "step": 26424 + }, + { + "epoch": 0.79, + "learning_rate": 2.2925202077793262e-05, + "loss": 0.5188, + "step": 26425 + }, + { + "epoch": 0.79, + "learning_rate": 2.2919058871547094e-05, + "loss": 0.548, + "step": 26426 + }, + { + "epoch": 0.79, + "learning_rate": 2.2912916381968795e-05, + "loss": 0.4139, + "step": 26427 + }, + { + "epoch": 0.79, + "learning_rate": 2.29067746091155e-05, + "loss": 0.4719, + "step": 26428 + }, + { + "epoch": 0.79, + "learning_rate": 2.2900633553044304e-05, + "loss": 0.6326, + "step": 26429 + }, + { + "epoch": 0.79, + "learning_rate": 2.289449321381225e-05, + "loss": 0.5025, + "step": 26430 + }, + { + "epoch": 0.79, + "learning_rate": 2.2888353591476464e-05, + "loss": 0.6085, + "step": 26431 + }, + { + "epoch": 0.79, + "learning_rate": 2.288221468609404e-05, + "loss": 0.6366, + "step": 26432 + }, + { + "epoch": 0.79, + "learning_rate": 2.2876076497722066e-05, + "loss": 0.4324, + "step": 26433 + }, + { + "epoch": 0.79, + "learning_rate": 2.28699390264176e-05, + "loss": 0.5761, + "step": 26434 + }, + { + "epoch": 0.79, + "learning_rate": 2.286380227223768e-05, + "loss": 0.4701, + "step": 26435 + }, + { + "epoch": 0.79, + "learning_rate": 2.2857666235239373e-05, + "loss": 0.4674, + "step": 26436 + }, + { + "epoch": 0.79, + "learning_rate": 2.2851530915479767e-05, + "loss": 0.4965, + "step": 26437 + }, + { + "epoch": 0.79, + "learning_rate": 2.2845396313015878e-05, + "loss": 0.4958, + "step": 26438 + }, + { + "epoch": 0.79, + "learning_rate": 2.2839262427904717e-05, + "loss": 0.5422, + "step": 26439 + }, + { + "epoch": 0.79, + "learning_rate": 2.2833129260203344e-05, + "loss": 0.4811, + "step": 26440 + }, + { + "epoch": 0.79, + "learning_rate": 2.2826996809968804e-05, + "loss": 0.593, + "step": 26441 + }, + { + "epoch": 0.79, + "learning_rate": 2.282086507725807e-05, + "loss": 0.5421, + "step": 26442 + }, + { + "epoch": 0.79, + "learning_rate": 2.2814734062128186e-05, + "loss": 0.4908, + "step": 26443 + }, + { + "epoch": 0.79, + "learning_rate": 2.2808603764636115e-05, + "loss": 0.4677, + "step": 26444 + }, + { + "epoch": 0.79, + "learning_rate": 2.280247418483892e-05, + "loss": 0.4589, + "step": 26445 + }, + { + "epoch": 0.79, + "learning_rate": 2.2796345322793523e-05, + "loss": 0.44, + "step": 26446 + }, + { + "epoch": 0.79, + "learning_rate": 2.2790217178556948e-05, + "loss": 0.5355, + "step": 26447 + }, + { + "epoch": 0.79, + "learning_rate": 2.2784089752186156e-05, + "loss": 0.6223, + "step": 26448 + }, + { + "epoch": 0.79, + "learning_rate": 2.277796304373814e-05, + "loss": 0.5024, + "step": 26449 + }, + { + "epoch": 0.79, + "learning_rate": 2.2771837053269816e-05, + "loss": 0.6026, + "step": 26450 + }, + { + "epoch": 0.79, + "learning_rate": 2.2765711780838174e-05, + "loss": 0.5946, + "step": 26451 + }, + { + "epoch": 0.79, + "learning_rate": 2.2759587226500177e-05, + "loss": 0.4672, + "step": 26452 + }, + { + "epoch": 0.79, + "learning_rate": 2.275346339031277e-05, + "loss": 0.5829, + "step": 26453 + }, + { + "epoch": 0.79, + "learning_rate": 2.274734027233284e-05, + "loss": 0.5874, + "step": 26454 + }, + { + "epoch": 0.79, + "learning_rate": 2.274121787261735e-05, + "loss": 0.4604, + "step": 26455 + }, + { + "epoch": 0.79, + "learning_rate": 2.273509619122325e-05, + "loss": 0.4407, + "step": 26456 + }, + { + "epoch": 0.79, + "learning_rate": 2.2728975228207427e-05, + "loss": 0.6343, + "step": 26457 + }, + { + "epoch": 0.79, + "learning_rate": 2.2722854983626774e-05, + "loss": 0.5566, + "step": 26458 + }, + { + "epoch": 0.79, + "learning_rate": 2.2716735457538207e-05, + "loss": 0.6014, + "step": 26459 + }, + { + "epoch": 0.79, + "learning_rate": 2.2710616649998673e-05, + "loss": 0.5948, + "step": 26460 + }, + { + "epoch": 0.79, + "learning_rate": 2.270449856106498e-05, + "loss": 0.4964, + "step": 26461 + }, + { + "epoch": 0.79, + "learning_rate": 2.2698381190794094e-05, + "loss": 0.4923, + "step": 26462 + }, + { + "epoch": 0.79, + "learning_rate": 2.2692264539242812e-05, + "loss": 0.4298, + "step": 26463 + }, + { + "epoch": 0.79, + "learning_rate": 2.2686148606468073e-05, + "loss": 0.5919, + "step": 26464 + }, + { + "epoch": 0.79, + "learning_rate": 2.268003339252668e-05, + "loss": 0.5529, + "step": 26465 + }, + { + "epoch": 0.79, + "learning_rate": 2.267391889747552e-05, + "loss": 0.62, + "step": 26466 + }, + { + "epoch": 0.79, + "learning_rate": 2.2667805121371477e-05, + "loss": 0.5976, + "step": 26467 + }, + { + "epoch": 0.79, + "learning_rate": 2.2661692064271357e-05, + "loss": 0.5271, + "step": 26468 + }, + { + "epoch": 0.79, + "learning_rate": 2.2655579726231968e-05, + "loss": 0.5191, + "step": 26469 + }, + { + "epoch": 0.79, + "learning_rate": 2.264946810731018e-05, + "loss": 0.4437, + "step": 26470 + }, + { + "epoch": 0.79, + "learning_rate": 2.264335720756283e-05, + "loss": 0.6228, + "step": 26471 + }, + { + "epoch": 0.79, + "learning_rate": 2.2637247027046717e-05, + "loss": 0.5707, + "step": 26472 + }, + { + "epoch": 0.79, + "learning_rate": 2.2631137565818617e-05, + "loss": 0.5597, + "step": 26473 + }, + { + "epoch": 0.79, + "learning_rate": 2.262502882393537e-05, + "loss": 0.5282, + "step": 26474 + }, + { + "epoch": 0.79, + "learning_rate": 2.2618920801453758e-05, + "loss": 0.4623, + "step": 26475 + }, + { + "epoch": 0.79, + "learning_rate": 2.261281349843061e-05, + "loss": 0.4462, + "step": 26476 + }, + { + "epoch": 0.79, + "learning_rate": 2.260670691492267e-05, + "loss": 0.509, + "step": 26477 + }, + { + "epoch": 0.79, + "learning_rate": 2.2600601050986712e-05, + "loss": 0.5135, + "step": 26478 + }, + { + "epoch": 0.79, + "learning_rate": 2.25944959066795e-05, + "loss": 0.5242, + "step": 26479 + }, + { + "epoch": 0.79, + "learning_rate": 2.258839148205786e-05, + "loss": 0.4858, + "step": 26480 + }, + { + "epoch": 0.79, + "learning_rate": 2.258228777717849e-05, + "loss": 0.5056, + "step": 26481 + }, + { + "epoch": 0.79, + "learning_rate": 2.2576184792098133e-05, + "loss": 0.5958, + "step": 26482 + }, + { + "epoch": 0.79, + "learning_rate": 2.2570082526873548e-05, + "loss": 0.4793, + "step": 26483 + }, + { + "epoch": 0.79, + "learning_rate": 2.2563980981561506e-05, + "loss": 0.5445, + "step": 26484 + }, + { + "epoch": 0.79, + "learning_rate": 2.2557880156218668e-05, + "loss": 0.5416, + "step": 26485 + }, + { + "epoch": 0.79, + "learning_rate": 2.2551780050901827e-05, + "loss": 0.5485, + "step": 26486 + }, + { + "epoch": 0.79, + "learning_rate": 2.2545680665667647e-05, + "loss": 0.6102, + "step": 26487 + }, + { + "epoch": 0.79, + "learning_rate": 2.253958200057288e-05, + "loss": 0.4575, + "step": 26488 + }, + { + "epoch": 0.79, + "learning_rate": 2.253348405567418e-05, + "loss": 0.5109, + "step": 26489 + }, + { + "epoch": 0.79, + "learning_rate": 2.2527386831028296e-05, + "loss": 0.4978, + "step": 26490 + }, + { + "epoch": 0.79, + "learning_rate": 2.25212903266919e-05, + "loss": 0.5403, + "step": 26491 + }, + { + "epoch": 0.79, + "learning_rate": 2.251519454272163e-05, + "loss": 0.5023, + "step": 26492 + }, + { + "epoch": 0.79, + "learning_rate": 2.25090994791742e-05, + "loss": 0.6195, + "step": 26493 + }, + { + "epoch": 0.79, + "learning_rate": 2.250300513610628e-05, + "loss": 0.5433, + "step": 26494 + }, + { + "epoch": 0.79, + "learning_rate": 2.249691151357456e-05, + "loss": 0.554, + "step": 26495 + }, + { + "epoch": 0.79, + "learning_rate": 2.249081861163567e-05, + "loss": 0.4462, + "step": 26496 + }, + { + "epoch": 0.79, + "learning_rate": 2.2484726430346215e-05, + "loss": 0.5758, + "step": 26497 + }, + { + "epoch": 0.79, + "learning_rate": 2.247863496976289e-05, + "loss": 0.6206, + "step": 26498 + }, + { + "epoch": 0.79, + "learning_rate": 2.247254422994235e-05, + "loss": 0.467, + "step": 26499 + }, + { + "epoch": 0.79, + "learning_rate": 2.246645421094119e-05, + "loss": 0.5073, + "step": 26500 + }, + { + "epoch": 0.79, + "learning_rate": 2.2460364912816e-05, + "loss": 0.474, + "step": 26501 + }, + { + "epoch": 0.79, + "learning_rate": 2.2454276335623446e-05, + "loss": 0.4452, + "step": 26502 + }, + { + "epoch": 0.79, + "learning_rate": 2.244818847942015e-05, + "loss": 0.5521, + "step": 26503 + }, + { + "epoch": 0.79, + "learning_rate": 2.2442101344262656e-05, + "loss": 0.5448, + "step": 26504 + }, + { + "epoch": 0.79, + "learning_rate": 2.2436014930207627e-05, + "loss": 0.6106, + "step": 26505 + }, + { + "epoch": 0.79, + "learning_rate": 2.2429929237311577e-05, + "loss": 0.5292, + "step": 26506 + }, + { + "epoch": 0.79, + "learning_rate": 2.2423844265631166e-05, + "loss": 0.481, + "step": 26507 + }, + { + "epoch": 0.79, + "learning_rate": 2.2417760015222898e-05, + "loss": 0.5885, + "step": 26508 + }, + { + "epoch": 0.79, + "learning_rate": 2.2411676486143385e-05, + "loss": 0.4893, + "step": 26509 + }, + { + "epoch": 0.79, + "learning_rate": 2.240559367844921e-05, + "loss": 0.5626, + "step": 26510 + }, + { + "epoch": 0.79, + "learning_rate": 2.2399511592196886e-05, + "loss": 0.536, + "step": 26511 + }, + { + "epoch": 0.79, + "learning_rate": 2.2393430227442956e-05, + "loss": 0.5288, + "step": 26512 + }, + { + "epoch": 0.79, + "learning_rate": 2.2387349584243976e-05, + "loss": 0.5107, + "step": 26513 + }, + { + "epoch": 0.79, + "learning_rate": 2.2381269662656525e-05, + "loss": 0.6455, + "step": 26514 + }, + { + "epoch": 0.79, + "learning_rate": 2.2375190462737082e-05, + "loss": 0.5598, + "step": 26515 + }, + { + "epoch": 0.79, + "learning_rate": 2.2369111984542155e-05, + "loss": 0.6203, + "step": 26516 + }, + { + "epoch": 0.79, + "learning_rate": 2.2363034228128276e-05, + "loss": 0.5742, + "step": 26517 + }, + { + "epoch": 0.79, + "learning_rate": 2.2356957193551974e-05, + "loss": 0.579, + "step": 26518 + }, + { + "epoch": 0.79, + "learning_rate": 2.2350880880869784e-05, + "loss": 0.428, + "step": 26519 + }, + { + "epoch": 0.79, + "learning_rate": 2.2344805290138116e-05, + "loss": 0.4983, + "step": 26520 + }, + { + "epoch": 0.79, + "learning_rate": 2.233873042141348e-05, + "loss": 0.5071, + "step": 26521 + }, + { + "epoch": 0.79, + "learning_rate": 2.2332656274752416e-05, + "loss": 0.5434, + "step": 26522 + }, + { + "epoch": 0.79, + "learning_rate": 2.2326582850211318e-05, + "loss": 0.497, + "step": 26523 + }, + { + "epoch": 0.79, + "learning_rate": 2.232051014784674e-05, + "loss": 0.5921, + "step": 26524 + }, + { + "epoch": 0.79, + "learning_rate": 2.2314438167715056e-05, + "loss": 0.4309, + "step": 26525 + }, + { + "epoch": 0.79, + "learning_rate": 2.2308366909872802e-05, + "loss": 0.598, + "step": 26526 + }, + { + "epoch": 0.79, + "learning_rate": 2.2302296374376353e-05, + "loss": 0.5111, + "step": 26527 + }, + { + "epoch": 0.79, + "learning_rate": 2.2296226561282184e-05, + "loss": 0.6283, + "step": 26528 + }, + { + "epoch": 0.79, + "learning_rate": 2.2290157470646767e-05, + "loss": 0.4803, + "step": 26529 + }, + { + "epoch": 0.79, + "learning_rate": 2.2284089102526483e-05, + "loss": 0.533, + "step": 26530 + }, + { + "epoch": 0.79, + "learning_rate": 2.2278021456977748e-05, + "loss": 0.4985, + "step": 26531 + }, + { + "epoch": 0.79, + "learning_rate": 2.227195453405698e-05, + "loss": 0.6474, + "step": 26532 + }, + { + "epoch": 0.79, + "learning_rate": 2.2265888333820638e-05, + "loss": 0.4958, + "step": 26533 + }, + { + "epoch": 0.79, + "learning_rate": 2.2259822856325085e-05, + "loss": 0.6547, + "step": 26534 + }, + { + "epoch": 0.79, + "learning_rate": 2.2253758101626676e-05, + "loss": 0.5517, + "step": 26535 + }, + { + "epoch": 0.79, + "learning_rate": 2.2247694069781844e-05, + "loss": 0.575, + "step": 26536 + }, + { + "epoch": 0.79, + "learning_rate": 2.2241630760846965e-05, + "loss": 0.5037, + "step": 26537 + }, + { + "epoch": 0.79, + "learning_rate": 2.223556817487844e-05, + "loss": 0.5429, + "step": 26538 + }, + { + "epoch": 0.79, + "learning_rate": 2.2229506311932603e-05, + "loss": 0.5208, + "step": 26539 + }, + { + "epoch": 0.79, + "learning_rate": 2.2223445172065793e-05, + "loss": 0.551, + "step": 26540 + }, + { + "epoch": 0.79, + "learning_rate": 2.2217384755334402e-05, + "loss": 0.5011, + "step": 26541 + }, + { + "epoch": 0.79, + "learning_rate": 2.2211325061794784e-05, + "loss": 0.5647, + "step": 26542 + }, + { + "epoch": 0.79, + "learning_rate": 2.2205266091503264e-05, + "loss": 0.4761, + "step": 26543 + }, + { + "epoch": 0.79, + "learning_rate": 2.2199207844516146e-05, + "loss": 0.6007, + "step": 26544 + }, + { + "epoch": 0.79, + "learning_rate": 2.219315032088979e-05, + "loss": 0.5553, + "step": 26545 + }, + { + "epoch": 0.79, + "learning_rate": 2.2187093520680525e-05, + "loss": 0.5509, + "step": 26546 + }, + { + "epoch": 0.79, + "learning_rate": 2.2181037443944643e-05, + "loss": 0.4753, + "step": 26547 + }, + { + "epoch": 0.79, + "learning_rate": 2.2174982090738473e-05, + "loss": 0.4877, + "step": 26548 + }, + { + "epoch": 0.79, + "learning_rate": 2.2168927461118273e-05, + "loss": 0.4667, + "step": 26549 + }, + { + "epoch": 0.79, + "learning_rate": 2.2162873555140396e-05, + "loss": 0.5218, + "step": 26550 + }, + { + "epoch": 0.79, + "learning_rate": 2.2156820372861077e-05, + "loss": 0.4002, + "step": 26551 + }, + { + "epoch": 0.79, + "learning_rate": 2.2150767914336635e-05, + "loss": 0.5952, + "step": 26552 + }, + { + "epoch": 0.79, + "learning_rate": 2.2144716179623303e-05, + "loss": 0.5328, + "step": 26553 + }, + { + "epoch": 0.79, + "learning_rate": 2.2138665168777396e-05, + "loss": 0.548, + "step": 26554 + }, + { + "epoch": 0.79, + "learning_rate": 2.213261488185512e-05, + "loss": 0.5236, + "step": 26555 + }, + { + "epoch": 0.79, + "learning_rate": 2.2126565318912763e-05, + "loss": 0.5239, + "step": 26556 + }, + { + "epoch": 0.79, + "learning_rate": 2.212051648000658e-05, + "loss": 0.467, + "step": 26557 + }, + { + "epoch": 0.79, + "learning_rate": 2.211446836519281e-05, + "loss": 0.5095, + "step": 26558 + }, + { + "epoch": 0.79, + "learning_rate": 2.2108420974527633e-05, + "loss": 0.5288, + "step": 26559 + }, + { + "epoch": 0.79, + "learning_rate": 2.210237430806731e-05, + "loss": 0.5217, + "step": 26560 + }, + { + "epoch": 0.79, + "learning_rate": 2.209632836586809e-05, + "loss": 0.4583, + "step": 26561 + }, + { + "epoch": 0.79, + "learning_rate": 2.2090283147986167e-05, + "loss": 0.5302, + "step": 26562 + }, + { + "epoch": 0.79, + "learning_rate": 2.2084238654477718e-05, + "loss": 0.5951, + "step": 26563 + }, + { + "epoch": 0.79, + "learning_rate": 2.207819488539895e-05, + "loss": 0.5029, + "step": 26564 + }, + { + "epoch": 0.79, + "learning_rate": 2.207215184080611e-05, + "loss": 0.5599, + "step": 26565 + }, + { + "epoch": 0.79, + "learning_rate": 2.2066109520755307e-05, + "loss": 0.5254, + "step": 26566 + }, + { + "epoch": 0.79, + "learning_rate": 2.2060067925302785e-05, + "loss": 0.486, + "step": 26567 + }, + { + "epoch": 0.79, + "learning_rate": 2.2054027054504656e-05, + "loss": 0.6063, + "step": 26568 + }, + { + "epoch": 0.79, + "learning_rate": 2.2047986908417152e-05, + "loss": 0.5376, + "step": 26569 + }, + { + "epoch": 0.79, + "learning_rate": 2.204194748709637e-05, + "loss": 0.5012, + "step": 26570 + }, + { + "epoch": 0.79, + "learning_rate": 2.203590879059848e-05, + "loss": 0.5913, + "step": 26571 + }, + { + "epoch": 0.79, + "learning_rate": 2.202987081897967e-05, + "loss": 0.5576, + "step": 26572 + }, + { + "epoch": 0.79, + "learning_rate": 2.202383357229605e-05, + "loss": 0.5227, + "step": 26573 + }, + { + "epoch": 0.79, + "learning_rate": 2.2017797050603716e-05, + "loss": 0.5712, + "step": 26574 + }, + { + "epoch": 0.79, + "learning_rate": 2.201176125395882e-05, + "loss": 0.5481, + "step": 26575 + }, + { + "epoch": 0.79, + "learning_rate": 2.200572618241752e-05, + "loss": 0.6418, + "step": 26576 + }, + { + "epoch": 0.79, + "learning_rate": 2.1999691836035885e-05, + "loss": 0.6178, + "step": 26577 + }, + { + "epoch": 0.79, + "learning_rate": 2.1993658214870006e-05, + "loss": 0.469, + "step": 26578 + }, + { + "epoch": 0.79, + "learning_rate": 2.1987625318976e-05, + "loss": 0.5815, + "step": 26579 + }, + { + "epoch": 0.79, + "learning_rate": 2.198159314840996e-05, + "loss": 0.5707, + "step": 26580 + }, + { + "epoch": 0.79, + "learning_rate": 2.1975561703227997e-05, + "loss": 0.5168, + "step": 26581 + }, + { + "epoch": 0.79, + "learning_rate": 2.1969530983486176e-05, + "loss": 0.524, + "step": 26582 + }, + { + "epoch": 0.79, + "learning_rate": 2.1963500989240517e-05, + "loss": 0.5004, + "step": 26583 + }, + { + "epoch": 0.79, + "learning_rate": 2.195747172054713e-05, + "loss": 0.5029, + "step": 26584 + }, + { + "epoch": 0.79, + "learning_rate": 2.1951443177462082e-05, + "loss": 0.506, + "step": 26585 + }, + { + "epoch": 0.79, + "learning_rate": 2.194541536004142e-05, + "loss": 0.48, + "step": 26586 + }, + { + "epoch": 0.79, + "learning_rate": 2.1939388268341145e-05, + "loss": 0.5455, + "step": 26587 + }, + { + "epoch": 0.79, + "learning_rate": 2.193336190241736e-05, + "loss": 0.5535, + "step": 26588 + }, + { + "epoch": 0.79, + "learning_rate": 2.1927336262326027e-05, + "loss": 0.423, + "step": 26589 + }, + { + "epoch": 0.79, + "learning_rate": 2.192131134812321e-05, + "loss": 0.5693, + "step": 26590 + }, + { + "epoch": 0.79, + "learning_rate": 2.191528715986494e-05, + "loss": 0.5009, + "step": 26591 + }, + { + "epoch": 0.79, + "learning_rate": 2.1909263697607207e-05, + "loss": 0.5644, + "step": 26592 + }, + { + "epoch": 0.79, + "learning_rate": 2.1903240961406e-05, + "loss": 0.4878, + "step": 26593 + }, + { + "epoch": 0.79, + "learning_rate": 2.1897218951317323e-05, + "loss": 0.4341, + "step": 26594 + }, + { + "epoch": 0.79, + "learning_rate": 2.1891197667397212e-05, + "loss": 0.5398, + "step": 26595 + }, + { + "epoch": 0.79, + "learning_rate": 2.1885177109701604e-05, + "loss": 0.4838, + "step": 26596 + }, + { + "epoch": 0.79, + "learning_rate": 2.1879157278286466e-05, + "loss": 0.5528, + "step": 26597 + }, + { + "epoch": 0.79, + "learning_rate": 2.1873138173207785e-05, + "loss": 0.5388, + "step": 26598 + }, + { + "epoch": 0.79, + "learning_rate": 2.1867119794521527e-05, + "loss": 0.4842, + "step": 26599 + }, + { + "epoch": 0.79, + "learning_rate": 2.1861102142283675e-05, + "loss": 0.4963, + "step": 26600 + }, + { + "epoch": 0.79, + "learning_rate": 2.1855085216550152e-05, + "loss": 0.5637, + "step": 26601 + }, + { + "epoch": 0.79, + "learning_rate": 2.1849069017376877e-05, + "loss": 0.5286, + "step": 26602 + }, + { + "epoch": 0.79, + "learning_rate": 2.1843053544819814e-05, + "loss": 0.4994, + "step": 26603 + }, + { + "epoch": 0.79, + "learning_rate": 2.1837038798934905e-05, + "loss": 0.5373, + "step": 26604 + }, + { + "epoch": 0.79, + "learning_rate": 2.1831024779778064e-05, + "loss": 0.5297, + "step": 26605 + }, + { + "epoch": 0.79, + "learning_rate": 2.1825011487405166e-05, + "loss": 0.5874, + "step": 26606 + }, + { + "epoch": 0.79, + "learning_rate": 2.1818998921872157e-05, + "loss": 0.5464, + "step": 26607 + }, + { + "epoch": 0.79, + "learning_rate": 2.181298708323497e-05, + "loss": 0.5469, + "step": 26608 + }, + { + "epoch": 0.79, + "learning_rate": 2.1806975971549438e-05, + "loss": 0.5487, + "step": 26609 + }, + { + "epoch": 0.79, + "learning_rate": 2.18009655868715e-05, + "loss": 0.6776, + "step": 26610 + }, + { + "epoch": 0.79, + "learning_rate": 2.1794955929256987e-05, + "loss": 0.6285, + "step": 26611 + }, + { + "epoch": 0.79, + "learning_rate": 2.178894699876184e-05, + "loss": 0.506, + "step": 26612 + }, + { + "epoch": 0.79, + "learning_rate": 2.1782938795441853e-05, + "loss": 0.5072, + "step": 26613 + }, + { + "epoch": 0.79, + "learning_rate": 2.1776931319352966e-05, + "loss": 0.546, + "step": 26614 + }, + { + "epoch": 0.79, + "learning_rate": 2.1770924570550965e-05, + "loss": 0.6026, + "step": 26615 + }, + { + "epoch": 0.79, + "learning_rate": 2.1764918549091752e-05, + "loss": 0.5268, + "step": 26616 + }, + { + "epoch": 0.79, + "learning_rate": 2.1758913255031122e-05, + "loss": 0.4735, + "step": 26617 + }, + { + "epoch": 0.79, + "learning_rate": 2.1752908688424932e-05, + "loss": 0.6304, + "step": 26618 + }, + { + "epoch": 0.79, + "learning_rate": 2.174690484932903e-05, + "loss": 0.4836, + "step": 26619 + }, + { + "epoch": 0.79, + "learning_rate": 2.1740901737799236e-05, + "loss": 0.5003, + "step": 26620 + }, + { + "epoch": 0.79, + "learning_rate": 2.1734899353891302e-05, + "loss": 0.4972, + "step": 26621 + }, + { + "epoch": 0.79, + "learning_rate": 2.172889769766109e-05, + "loss": 0.56, + "step": 26622 + }, + { + "epoch": 0.79, + "learning_rate": 2.1722896769164413e-05, + "loss": 0.5142, + "step": 26623 + }, + { + "epoch": 0.79, + "learning_rate": 2.1716896568457046e-05, + "loss": 0.462, + "step": 26624 + }, + { + "epoch": 0.79, + "learning_rate": 2.171089709559475e-05, + "loss": 0.5164, + "step": 26625 + }, + { + "epoch": 0.79, + "learning_rate": 2.1704898350633318e-05, + "loss": 0.4992, + "step": 26626 + }, + { + "epoch": 0.79, + "learning_rate": 2.169890033362858e-05, + "loss": 0.5653, + "step": 26627 + }, + { + "epoch": 0.79, + "learning_rate": 2.1692903044636226e-05, + "loss": 0.4768, + "step": 26628 + }, + { + "epoch": 0.79, + "learning_rate": 2.1686906483712066e-05, + "loss": 0.4615, + "step": 26629 + }, + { + "epoch": 0.79, + "learning_rate": 2.1680910650911825e-05, + "loss": 0.6407, + "step": 26630 + }, + { + "epoch": 0.79, + "learning_rate": 2.1674915546291276e-05, + "loss": 0.6168, + "step": 26631 + }, + { + "epoch": 0.79, + "learning_rate": 2.166892116990612e-05, + "loss": 0.52, + "step": 26632 + }, + { + "epoch": 0.79, + "learning_rate": 2.1662927521812126e-05, + "loss": 0.5004, + "step": 26633 + }, + { + "epoch": 0.79, + "learning_rate": 2.1656934602065028e-05, + "loss": 0.5754, + "step": 26634 + }, + { + "epoch": 0.79, + "learning_rate": 2.165094241072053e-05, + "loss": 0.5632, + "step": 26635 + }, + { + "epoch": 0.79, + "learning_rate": 2.164495094783432e-05, + "loss": 0.4143, + "step": 26636 + }, + { + "epoch": 0.79, + "learning_rate": 2.1638960213462113e-05, + "loss": 0.4805, + "step": 26637 + }, + { + "epoch": 0.79, + "learning_rate": 2.1632970207659664e-05, + "loss": 0.5205, + "step": 26638 + }, + { + "epoch": 0.79, + "learning_rate": 2.162698093048262e-05, + "loss": 0.5439, + "step": 26639 + }, + { + "epoch": 0.79, + "learning_rate": 2.1620992381986638e-05, + "loss": 0.5334, + "step": 26640 + }, + { + "epoch": 0.79, + "learning_rate": 2.1615004562227436e-05, + "loss": 0.3949, + "step": 26641 + }, + { + "epoch": 0.79, + "learning_rate": 2.1609017471260674e-05, + "loss": 0.4907, + "step": 26642 + }, + { + "epoch": 0.79, + "learning_rate": 2.1603031109142057e-05, + "loss": 0.5317, + "step": 26643 + }, + { + "epoch": 0.79, + "learning_rate": 2.1597045475927213e-05, + "loss": 0.4484, + "step": 26644 + }, + { + "epoch": 0.79, + "learning_rate": 2.1591060571671763e-05, + "loss": 0.4388, + "step": 26645 + }, + { + "epoch": 0.79, + "learning_rate": 2.158507639643137e-05, + "loss": 0.62, + "step": 26646 + }, + { + "epoch": 0.79, + "learning_rate": 2.1579092950261726e-05, + "loss": 0.4207, + "step": 26647 + }, + { + "epoch": 0.79, + "learning_rate": 2.157311023321842e-05, + "loss": 0.5475, + "step": 26648 + }, + { + "epoch": 0.79, + "learning_rate": 2.1567128245357038e-05, + "loss": 0.6129, + "step": 26649 + }, + { + "epoch": 0.79, + "learning_rate": 2.156114698673324e-05, + "loss": 0.5726, + "step": 26650 + }, + { + "epoch": 0.79, + "learning_rate": 2.1555166457402664e-05, + "loss": 0.5526, + "step": 26651 + }, + { + "epoch": 0.79, + "learning_rate": 2.154918665742086e-05, + "loss": 0.5685, + "step": 26652 + }, + { + "epoch": 0.79, + "learning_rate": 2.1543207586843484e-05, + "loss": 0.5821, + "step": 26653 + }, + { + "epoch": 0.79, + "learning_rate": 2.1537229245726088e-05, + "loss": 0.4812, + "step": 26654 + }, + { + "epoch": 0.79, + "learning_rate": 2.1531251634124227e-05, + "loss": 0.6008, + "step": 26655 + }, + { + "epoch": 0.79, + "learning_rate": 2.1525274752093528e-05, + "loss": 0.5453, + "step": 26656 + }, + { + "epoch": 0.79, + "learning_rate": 2.1519298599689562e-05, + "loss": 0.4708, + "step": 26657 + }, + { + "epoch": 0.79, + "learning_rate": 2.1513323176967892e-05, + "loss": 0.5107, + "step": 26658 + }, + { + "epoch": 0.79, + "learning_rate": 2.1507348483984026e-05, + "loss": 0.5525, + "step": 26659 + }, + { + "epoch": 0.79, + "learning_rate": 2.1501374520793548e-05, + "loss": 0.4645, + "step": 26660 + }, + { + "epoch": 0.79, + "learning_rate": 2.149540128745201e-05, + "loss": 0.4291, + "step": 26661 + }, + { + "epoch": 0.79, + "learning_rate": 2.1489428784014964e-05, + "loss": 0.5415, + "step": 26662 + }, + { + "epoch": 0.79, + "learning_rate": 2.148345701053791e-05, + "loss": 0.5676, + "step": 26663 + }, + { + "epoch": 0.79, + "learning_rate": 2.1477485967076372e-05, + "loss": 0.4486, + "step": 26664 + }, + { + "epoch": 0.79, + "learning_rate": 2.147151565368586e-05, + "loss": 0.4521, + "step": 26665 + }, + { + "epoch": 0.79, + "learning_rate": 2.146554607042193e-05, + "loss": 0.5865, + "step": 26666 + }, + { + "epoch": 0.79, + "learning_rate": 2.1459577217340042e-05, + "loss": 0.5311, + "step": 26667 + }, + { + "epoch": 0.79, + "learning_rate": 2.1453609094495685e-05, + "loss": 0.3935, + "step": 26668 + }, + { + "epoch": 0.79, + "learning_rate": 2.1447641701944355e-05, + "loss": 0.4312, + "step": 26669 + }, + { + "epoch": 0.79, + "learning_rate": 2.1441675039741583e-05, + "loss": 0.5584, + "step": 26670 + }, + { + "epoch": 0.79, + "learning_rate": 2.143570910794277e-05, + "loss": 0.5757, + "step": 26671 + }, + { + "epoch": 0.79, + "learning_rate": 2.142974390660345e-05, + "loss": 0.4661, + "step": 26672 + }, + { + "epoch": 0.79, + "learning_rate": 2.1423779435779033e-05, + "loss": 0.5032, + "step": 26673 + }, + { + "epoch": 0.79, + "learning_rate": 2.1417815695525022e-05, + "loss": 0.5234, + "step": 26674 + }, + { + "epoch": 0.79, + "learning_rate": 2.1411852685896817e-05, + "loss": 0.5107, + "step": 26675 + }, + { + "epoch": 0.79, + "learning_rate": 2.140589040694988e-05, + "loss": 0.4221, + "step": 26676 + }, + { + "epoch": 0.79, + "learning_rate": 2.1399928858739683e-05, + "loss": 0.4144, + "step": 26677 + }, + { + "epoch": 0.79, + "learning_rate": 2.139396804132161e-05, + "loss": 0.5255, + "step": 26678 + }, + { + "epoch": 0.79, + "learning_rate": 2.1388007954751078e-05, + "loss": 0.5335, + "step": 26679 + }, + { + "epoch": 0.79, + "learning_rate": 2.1382048599083504e-05, + "loss": 0.5366, + "step": 26680 + }, + { + "epoch": 0.79, + "learning_rate": 2.1376089974374336e-05, + "loss": 0.457, + "step": 26681 + }, + { + "epoch": 0.79, + "learning_rate": 2.137013208067896e-05, + "loss": 0.5595, + "step": 26682 + }, + { + "epoch": 0.79, + "learning_rate": 2.1364174918052716e-05, + "loss": 0.5001, + "step": 26683 + }, + { + "epoch": 0.79, + "learning_rate": 2.135821848655104e-05, + "loss": 0.457, + "step": 26684 + }, + { + "epoch": 0.79, + "learning_rate": 2.1352262786229326e-05, + "loss": 0.5284, + "step": 26685 + }, + { + "epoch": 0.79, + "learning_rate": 2.1346307817142908e-05, + "loss": 0.4861, + "step": 26686 + }, + { + "epoch": 0.79, + "learning_rate": 2.1340353579347193e-05, + "loss": 0.5807, + "step": 26687 + }, + { + "epoch": 0.79, + "learning_rate": 2.13344000728975e-05, + "loss": 0.5348, + "step": 26688 + }, + { + "epoch": 0.79, + "learning_rate": 2.1328447297849218e-05, + "loss": 0.5501, + "step": 26689 + }, + { + "epoch": 0.79, + "learning_rate": 2.132249525425766e-05, + "loss": 0.4955, + "step": 26690 + }, + { + "epoch": 0.79, + "learning_rate": 2.1316543942178203e-05, + "loss": 0.6335, + "step": 26691 + }, + { + "epoch": 0.79, + "learning_rate": 2.131059336166614e-05, + "loss": 0.5939, + "step": 26692 + }, + { + "epoch": 0.79, + "learning_rate": 2.1304643512776847e-05, + "loss": 0.5078, + "step": 26693 + }, + { + "epoch": 0.79, + "learning_rate": 2.1298694395565576e-05, + "loss": 0.5148, + "step": 26694 + }, + { + "epoch": 0.79, + "learning_rate": 2.1292746010087684e-05, + "loss": 0.4706, + "step": 26695 + }, + { + "epoch": 0.79, + "learning_rate": 2.1286798356398496e-05, + "loss": 0.6815, + "step": 26696 + }, + { + "epoch": 0.79, + "learning_rate": 2.1280851434553283e-05, + "loss": 0.5428, + "step": 26697 + }, + { + "epoch": 0.79, + "learning_rate": 2.1274905244607312e-05, + "loss": 0.5323, + "step": 26698 + }, + { + "epoch": 0.79, + "learning_rate": 2.1268959786615893e-05, + "loss": 0.5724, + "step": 26699 + }, + { + "epoch": 0.79, + "learning_rate": 2.126301506063434e-05, + "loss": 0.5594, + "step": 26700 + }, + { + "epoch": 0.79, + "learning_rate": 2.1257071066717893e-05, + "loss": 0.5111, + "step": 26701 + }, + { + "epoch": 0.79, + "learning_rate": 2.1251127804921777e-05, + "loss": 0.6091, + "step": 26702 + }, + { + "epoch": 0.79, + "learning_rate": 2.1245185275301293e-05, + "loss": 0.4866, + "step": 26703 + }, + { + "epoch": 0.8, + "learning_rate": 2.1239243477911695e-05, + "loss": 0.5794, + "step": 26704 + }, + { + "epoch": 0.8, + "learning_rate": 2.1233302412808243e-05, + "loss": 0.6611, + "step": 26705 + }, + { + "epoch": 0.8, + "learning_rate": 2.1227362080046154e-05, + "loss": 0.4404, + "step": 26706 + }, + { + "epoch": 0.8, + "learning_rate": 2.122142247968062e-05, + "loss": 0.4763, + "step": 26707 + }, + { + "epoch": 0.8, + "learning_rate": 2.1215483611766918e-05, + "loss": 0.4446, + "step": 26708 + }, + { + "epoch": 0.8, + "learning_rate": 2.120954547636026e-05, + "loss": 0.5948, + "step": 26709 + }, + { + "epoch": 0.8, + "learning_rate": 2.1203608073515856e-05, + "loss": 0.4803, + "step": 26710 + }, + { + "epoch": 0.8, + "learning_rate": 2.1197671403288865e-05, + "loss": 0.5407, + "step": 26711 + }, + { + "epoch": 0.8, + "learning_rate": 2.1191735465734518e-05, + "loss": 0.6282, + "step": 26712 + }, + { + "epoch": 0.8, + "learning_rate": 2.1185800260908038e-05, + "loss": 0.4967, + "step": 26713 + }, + { + "epoch": 0.8, + "learning_rate": 2.117986578886455e-05, + "loss": 0.548, + "step": 26714 + }, + { + "epoch": 0.8, + "learning_rate": 2.1173932049659273e-05, + "loss": 0.6209, + "step": 26715 + }, + { + "epoch": 0.8, + "learning_rate": 2.116799904334733e-05, + "loss": 0.5573, + "step": 26716 + }, + { + "epoch": 0.8, + "learning_rate": 2.1162066769983946e-05, + "loss": 0.615, + "step": 26717 + }, + { + "epoch": 0.8, + "learning_rate": 2.115613522962422e-05, + "loss": 0.5023, + "step": 26718 + }, + { + "epoch": 0.8, + "learning_rate": 2.115020442232335e-05, + "loss": 0.5535, + "step": 26719 + }, + { + "epoch": 0.8, + "learning_rate": 2.114427434813645e-05, + "loss": 0.5693, + "step": 26720 + }, + { + "epoch": 0.8, + "learning_rate": 2.113834500711863e-05, + "loss": 0.4457, + "step": 26721 + }, + { + "epoch": 0.8, + "learning_rate": 2.113241639932505e-05, + "loss": 0.476, + "step": 26722 + }, + { + "epoch": 0.8, + "learning_rate": 2.112648852481083e-05, + "loss": 0.4746, + "step": 26723 + }, + { + "epoch": 0.8, + "learning_rate": 2.11205613836311e-05, + "loss": 0.5906, + "step": 26724 + }, + { + "epoch": 0.8, + "learning_rate": 2.1114634975840962e-05, + "loss": 0.4972, + "step": 26725 + }, + { + "epoch": 0.8, + "learning_rate": 2.1108709301495477e-05, + "loss": 0.536, + "step": 26726 + }, + { + "epoch": 0.8, + "learning_rate": 2.1102784360649773e-05, + "loss": 0.5984, + "step": 26727 + }, + { + "epoch": 0.8, + "learning_rate": 2.1096860153358966e-05, + "loss": 0.5681, + "step": 26728 + }, + { + "epoch": 0.8, + "learning_rate": 2.10909366796781e-05, + "loss": 0.478, + "step": 26729 + }, + { + "epoch": 0.8, + "learning_rate": 2.1085013939662225e-05, + "loss": 0.5724, + "step": 26730 + }, + { + "epoch": 0.8, + "learning_rate": 2.1079091933366436e-05, + "loss": 0.5238, + "step": 26731 + }, + { + "epoch": 0.8, + "learning_rate": 2.107317066084583e-05, + "loss": 0.5756, + "step": 26732 + }, + { + "epoch": 0.8, + "learning_rate": 2.1067250122155403e-05, + "loss": 0.5991, + "step": 26733 + }, + { + "epoch": 0.8, + "learning_rate": 2.1061330317350238e-05, + "loss": 0.5617, + "step": 26734 + }, + { + "epoch": 0.8, + "learning_rate": 2.1055411246485346e-05, + "loss": 0.5898, + "step": 26735 + }, + { + "epoch": 0.8, + "learning_rate": 2.1049492909615797e-05, + "loss": 0.5597, + "step": 26736 + }, + { + "epoch": 0.8, + "learning_rate": 2.104357530679657e-05, + "loss": 0.5716, + "step": 26737 + }, + { + "epoch": 0.8, + "learning_rate": 2.1037658438082708e-05, + "loss": 0.5473, + "step": 26738 + }, + { + "epoch": 0.8, + "learning_rate": 2.1031742303529255e-05, + "loss": 0.5248, + "step": 26739 + }, + { + "epoch": 0.8, + "learning_rate": 2.1025826903191193e-05, + "loss": 0.4482, + "step": 26740 + }, + { + "epoch": 0.8, + "learning_rate": 2.1019912237123497e-05, + "loss": 0.4896, + "step": 26741 + }, + { + "epoch": 0.8, + "learning_rate": 2.1013998305381165e-05, + "loss": 0.5724, + "step": 26742 + }, + { + "epoch": 0.8, + "learning_rate": 2.1008085108019227e-05, + "loss": 0.6554, + "step": 26743 + }, + { + "epoch": 0.8, + "learning_rate": 2.1002172645092632e-05, + "loss": 0.4367, + "step": 26744 + }, + { + "epoch": 0.8, + "learning_rate": 2.099626091665632e-05, + "loss": 0.5552, + "step": 26745 + }, + { + "epoch": 0.8, + "learning_rate": 2.0990349922765284e-05, + "loss": 0.5061, + "step": 26746 + }, + { + "epoch": 0.8, + "learning_rate": 2.09844396634745e-05, + "loss": 0.4751, + "step": 26747 + }, + { + "epoch": 0.8, + "learning_rate": 2.0978530138838913e-05, + "loss": 0.4404, + "step": 26748 + }, + { + "epoch": 0.8, + "learning_rate": 2.0972621348913458e-05, + "loss": 0.4539, + "step": 26749 + }, + { + "epoch": 0.8, + "learning_rate": 2.0966713293753058e-05, + "loss": 0.6445, + "step": 26750 + }, + { + "epoch": 0.8, + "learning_rate": 2.0960805973412677e-05, + "loss": 0.4902, + "step": 26751 + }, + { + "epoch": 0.8, + "learning_rate": 2.095489938794719e-05, + "loss": 0.4599, + "step": 26752 + }, + { + "epoch": 0.8, + "learning_rate": 2.0948993537411576e-05, + "loss": 0.5886, + "step": 26753 + }, + { + "epoch": 0.8, + "learning_rate": 2.094308842186069e-05, + "loss": 0.4623, + "step": 26754 + }, + { + "epoch": 0.8, + "learning_rate": 2.093718404134949e-05, + "loss": 0.5424, + "step": 26755 + }, + { + "epoch": 0.8, + "learning_rate": 2.0931280395932794e-05, + "loss": 0.5194, + "step": 26756 + }, + { + "epoch": 0.8, + "learning_rate": 2.0925377485665553e-05, + "loss": 0.5416, + "step": 26757 + }, + { + "epoch": 0.8, + "learning_rate": 2.0919475310602665e-05, + "loss": 0.4253, + "step": 26758 + }, + { + "epoch": 0.8, + "learning_rate": 2.0913573870798976e-05, + "loss": 0.5876, + "step": 26759 + }, + { + "epoch": 0.8, + "learning_rate": 2.0907673166309328e-05, + "loss": 0.5081, + "step": 26760 + }, + { + "epoch": 0.8, + "learning_rate": 2.0901773197188613e-05, + "loss": 0.5076, + "step": 26761 + }, + { + "epoch": 0.8, + "learning_rate": 2.0895873963491708e-05, + "loss": 0.4482, + "step": 26762 + }, + { + "epoch": 0.8, + "learning_rate": 2.088997546527345e-05, + "loss": 0.4635, + "step": 26763 + }, + { + "epoch": 0.8, + "learning_rate": 2.0884077702588635e-05, + "loss": 0.521, + "step": 26764 + }, + { + "epoch": 0.8, + "learning_rate": 2.087818067549213e-05, + "loss": 0.5608, + "step": 26765 + }, + { + "epoch": 0.8, + "learning_rate": 2.0872284384038765e-05, + "loss": 0.589, + "step": 26766 + }, + { + "epoch": 0.8, + "learning_rate": 2.08663888282834e-05, + "loss": 0.6032, + "step": 26767 + }, + { + "epoch": 0.8, + "learning_rate": 2.0860494008280806e-05, + "loss": 0.5813, + "step": 26768 + }, + { + "epoch": 0.8, + "learning_rate": 2.0854599924085772e-05, + "loss": 0.5166, + "step": 26769 + }, + { + "epoch": 0.8, + "learning_rate": 2.0848706575753118e-05, + "loss": 0.5891, + "step": 26770 + }, + { + "epoch": 0.8, + "learning_rate": 2.084281396333767e-05, + "loss": 0.5686, + "step": 26771 + }, + { + "epoch": 0.8, + "learning_rate": 2.0836922086894193e-05, + "loss": 0.5567, + "step": 26772 + }, + { + "epoch": 0.8, + "learning_rate": 2.0831030946477437e-05, + "loss": 0.511, + "step": 26773 + }, + { + "epoch": 0.8, + "learning_rate": 2.082514054214221e-05, + "loss": 0.4462, + "step": 26774 + }, + { + "epoch": 0.8, + "learning_rate": 2.0819250873943274e-05, + "loss": 0.4977, + "step": 26775 + }, + { + "epoch": 0.8, + "learning_rate": 2.0813361941935382e-05, + "loss": 0.6093, + "step": 26776 + }, + { + "epoch": 0.8, + "learning_rate": 2.0807473746173302e-05, + "loss": 0.5485, + "step": 26777 + }, + { + "epoch": 0.8, + "learning_rate": 2.0801586286711748e-05, + "loss": 0.532, + "step": 26778 + }, + { + "epoch": 0.8, + "learning_rate": 2.079569956360551e-05, + "loss": 0.5791, + "step": 26779 + }, + { + "epoch": 0.8, + "learning_rate": 2.0789813576909256e-05, + "loss": 0.5298, + "step": 26780 + }, + { + "epoch": 0.8, + "learning_rate": 2.078392832667777e-05, + "loss": 0.5286, + "step": 26781 + }, + { + "epoch": 0.8, + "learning_rate": 2.077804381296572e-05, + "loss": 0.5901, + "step": 26782 + }, + { + "epoch": 0.8, + "learning_rate": 2.077216003582787e-05, + "loss": 0.5416, + "step": 26783 + }, + { + "epoch": 0.8, + "learning_rate": 2.0766276995318878e-05, + "loss": 0.6631, + "step": 26784 + }, + { + "epoch": 0.8, + "learning_rate": 2.0760394691493455e-05, + "loss": 0.5286, + "step": 26785 + }, + { + "epoch": 0.8, + "learning_rate": 2.075451312440634e-05, + "loss": 0.4928, + "step": 26786 + }, + { + "epoch": 0.8, + "learning_rate": 2.0748632294112168e-05, + "loss": 0.4559, + "step": 26787 + }, + { + "epoch": 0.8, + "learning_rate": 2.0742752200665605e-05, + "loss": 0.443, + "step": 26788 + }, + { + "epoch": 0.8, + "learning_rate": 2.0736872844121337e-05, + "loss": 0.6162, + "step": 26789 + }, + { + "epoch": 0.8, + "learning_rate": 2.073099422453406e-05, + "loss": 0.499, + "step": 26790 + }, + { + "epoch": 0.8, + "learning_rate": 2.072511634195842e-05, + "loss": 0.525, + "step": 26791 + }, + { + "epoch": 0.8, + "learning_rate": 2.0719239196449014e-05, + "loss": 0.4655, + "step": 26792 + }, + { + "epoch": 0.8, + "learning_rate": 2.071336278806052e-05, + "loss": 0.5154, + "step": 26793 + }, + { + "epoch": 0.8, + "learning_rate": 2.070748711684761e-05, + "loss": 0.5111, + "step": 26794 + }, + { + "epoch": 0.8, + "learning_rate": 2.070161218286486e-05, + "loss": 0.476, + "step": 26795 + }, + { + "epoch": 0.8, + "learning_rate": 2.069573798616693e-05, + "loss": 0.5836, + "step": 26796 + }, + { + "epoch": 0.8, + "learning_rate": 2.06898645268084e-05, + "loss": 0.4631, + "step": 26797 + }, + { + "epoch": 0.8, + "learning_rate": 2.0683991804843926e-05, + "loss": 0.5312, + "step": 26798 + }, + { + "epoch": 0.8, + "learning_rate": 2.067811982032807e-05, + "loss": 0.4424, + "step": 26799 + }, + { + "epoch": 0.8, + "learning_rate": 2.0672248573315422e-05, + "loss": 0.5481, + "step": 26800 + }, + { + "epoch": 0.8, + "learning_rate": 2.0666378063860625e-05, + "loss": 0.6266, + "step": 26801 + }, + { + "epoch": 0.8, + "learning_rate": 2.0660508292018222e-05, + "loss": 0.5354, + "step": 26802 + }, + { + "epoch": 0.8, + "learning_rate": 2.0654639257842766e-05, + "loss": 0.536, + "step": 26803 + }, + { + "epoch": 0.8, + "learning_rate": 2.0648770961388852e-05, + "loss": 0.5989, + "step": 26804 + }, + { + "epoch": 0.8, + "learning_rate": 2.0642903402711068e-05, + "loss": 0.5396, + "step": 26805 + }, + { + "epoch": 0.8, + "learning_rate": 2.0637036581863934e-05, + "loss": 0.5135, + "step": 26806 + }, + { + "epoch": 0.8, + "learning_rate": 2.0631170498901987e-05, + "loss": 0.5619, + "step": 26807 + }, + { + "epoch": 0.8, + "learning_rate": 2.062530515387977e-05, + "loss": 0.5342, + "step": 26808 + }, + { + "epoch": 0.8, + "learning_rate": 2.061944054685183e-05, + "loss": 0.5203, + "step": 26809 + }, + { + "epoch": 0.8, + "learning_rate": 2.0613576677872725e-05, + "loss": 0.6093, + "step": 26810 + }, + { + "epoch": 0.8, + "learning_rate": 2.0607713546996955e-05, + "loss": 0.4609, + "step": 26811 + }, + { + "epoch": 0.8, + "learning_rate": 2.0601851154278984e-05, + "loss": 0.4618, + "step": 26812 + }, + { + "epoch": 0.8, + "learning_rate": 2.059598949977335e-05, + "loss": 0.4621, + "step": 26813 + }, + { + "epoch": 0.8, + "learning_rate": 2.0590128583534595e-05, + "loss": 0.4112, + "step": 26814 + }, + { + "epoch": 0.8, + "learning_rate": 2.0584268405617168e-05, + "loss": 0.649, + "step": 26815 + }, + { + "epoch": 0.8, + "learning_rate": 2.0578408966075545e-05, + "loss": 0.4517, + "step": 26816 + }, + { + "epoch": 0.8, + "learning_rate": 2.0572550264964242e-05, + "loss": 0.5931, + "step": 26817 + }, + { + "epoch": 0.8, + "learning_rate": 2.0566692302337676e-05, + "loss": 0.5202, + "step": 26818 + }, + { + "epoch": 0.8, + "learning_rate": 2.056083507825035e-05, + "loss": 0.5831, + "step": 26819 + }, + { + "epoch": 0.8, + "learning_rate": 2.055497859275676e-05, + "loss": 0.6398, + "step": 26820 + }, + { + "epoch": 0.8, + "learning_rate": 2.0549122845911296e-05, + "loss": 0.5581, + "step": 26821 + }, + { + "epoch": 0.8, + "learning_rate": 2.05432678377684e-05, + "loss": 0.5213, + "step": 26822 + }, + { + "epoch": 0.8, + "learning_rate": 2.0537413568382536e-05, + "loss": 0.6099, + "step": 26823 + }, + { + "epoch": 0.8, + "learning_rate": 2.053156003780815e-05, + "loss": 0.5017, + "step": 26824 + }, + { + "epoch": 0.8, + "learning_rate": 2.0525707246099658e-05, + "loss": 0.5753, + "step": 26825 + }, + { + "epoch": 0.8, + "learning_rate": 2.051985519331142e-05, + "loss": 0.5533, + "step": 26826 + }, + { + "epoch": 0.8, + "learning_rate": 2.05140038794979e-05, + "loss": 0.5208, + "step": 26827 + }, + { + "epoch": 0.8, + "learning_rate": 2.0508153304713495e-05, + "loss": 0.5066, + "step": 26828 + }, + { + "epoch": 0.8, + "learning_rate": 2.050230346901262e-05, + "loss": 0.5073, + "step": 26829 + }, + { + "epoch": 0.8, + "learning_rate": 2.0496454372449647e-05, + "loss": 0.6234, + "step": 26830 + }, + { + "epoch": 0.8, + "learning_rate": 2.0490606015078928e-05, + "loss": 0.4702, + "step": 26831 + }, + { + "epoch": 0.8, + "learning_rate": 2.048475839695486e-05, + "loss": 0.5368, + "step": 26832 + }, + { + "epoch": 0.8, + "learning_rate": 2.0478911518131848e-05, + "loss": 0.4451, + "step": 26833 + }, + { + "epoch": 0.8, + "learning_rate": 2.0473065378664212e-05, + "loss": 0.5201, + "step": 26834 + }, + { + "epoch": 0.8, + "learning_rate": 2.04672199786063e-05, + "loss": 0.5135, + "step": 26835 + }, + { + "epoch": 0.8, + "learning_rate": 2.0461375318012488e-05, + "loss": 0.5561, + "step": 26836 + }, + { + "epoch": 0.8, + "learning_rate": 2.045553139693712e-05, + "loss": 0.5322, + "step": 26837 + }, + { + "epoch": 0.8, + "learning_rate": 2.0449688215434494e-05, + "loss": 0.4405, + "step": 26838 + }, + { + "epoch": 0.8, + "learning_rate": 2.044384577355899e-05, + "loss": 0.5532, + "step": 26839 + }, + { + "epoch": 0.8, + "learning_rate": 2.0438004071364868e-05, + "loss": 0.4975, + "step": 26840 + }, + { + "epoch": 0.8, + "learning_rate": 2.0432163108906512e-05, + "loss": 0.563, + "step": 26841 + }, + { + "epoch": 0.8, + "learning_rate": 2.0426322886238146e-05, + "loss": 0.4442, + "step": 26842 + }, + { + "epoch": 0.8, + "learning_rate": 2.0420483403414138e-05, + "loss": 0.5334, + "step": 26843 + }, + { + "epoch": 0.8, + "learning_rate": 2.041464466048877e-05, + "loss": 0.6182, + "step": 26844 + }, + { + "epoch": 0.8, + "learning_rate": 2.040880665751631e-05, + "loss": 0.4389, + "step": 26845 + }, + { + "epoch": 0.8, + "learning_rate": 2.040296939455103e-05, + "loss": 0.4681, + "step": 26846 + }, + { + "epoch": 0.8, + "learning_rate": 2.039713287164722e-05, + "loss": 0.4652, + "step": 26847 + }, + { + "epoch": 0.8, + "learning_rate": 2.039129708885915e-05, + "loss": 0.5485, + "step": 26848 + }, + { + "epoch": 0.8, + "learning_rate": 2.038546204624109e-05, + "loss": 0.3607, + "step": 26849 + }, + { + "epoch": 0.8, + "learning_rate": 2.037962774384723e-05, + "loss": 0.4895, + "step": 26850 + }, + { + "epoch": 0.8, + "learning_rate": 2.0373794181731866e-05, + "loss": 0.5329, + "step": 26851 + }, + { + "epoch": 0.8, + "learning_rate": 2.0367961359949262e-05, + "loss": 0.5533, + "step": 26852 + }, + { + "epoch": 0.8, + "learning_rate": 2.0362129278553577e-05, + "loss": 0.5722, + "step": 26853 + }, + { + "epoch": 0.8, + "learning_rate": 2.03562979375991e-05, + "loss": 0.6284, + "step": 26854 + }, + { + "epoch": 0.8, + "learning_rate": 2.035046733714e-05, + "loss": 0.4711, + "step": 26855 + }, + { + "epoch": 0.8, + "learning_rate": 2.0344637477230532e-05, + "loss": 0.5133, + "step": 26856 + }, + { + "epoch": 0.8, + "learning_rate": 2.0338808357924855e-05, + "loss": 0.4065, + "step": 26857 + }, + { + "epoch": 0.8, + "learning_rate": 2.0332979979277212e-05, + "loss": 0.5151, + "step": 26858 + }, + { + "epoch": 0.8, + "learning_rate": 2.0327152341341748e-05, + "loss": 0.5554, + "step": 26859 + }, + { + "epoch": 0.8, + "learning_rate": 2.032132544417269e-05, + "loss": 0.6923, + "step": 26860 + }, + { + "epoch": 0.8, + "learning_rate": 2.0315499287824156e-05, + "loss": 0.5603, + "step": 26861 + }, + { + "epoch": 0.8, + "learning_rate": 2.0309673872350365e-05, + "loss": 0.5138, + "step": 26862 + }, + { + "epoch": 0.8, + "learning_rate": 2.030384919780548e-05, + "loss": 0.5126, + "step": 26863 + }, + { + "epoch": 0.8, + "learning_rate": 2.0298025264243647e-05, + "loss": 0.5836, + "step": 26864 + }, + { + "epoch": 0.8, + "learning_rate": 2.029220207171898e-05, + "loss": 0.6233, + "step": 26865 + }, + { + "epoch": 0.8, + "learning_rate": 2.0286379620285655e-05, + "loss": 0.53, + "step": 26866 + }, + { + "epoch": 0.8, + "learning_rate": 2.0280557909997822e-05, + "loss": 0.5771, + "step": 26867 + }, + { + "epoch": 0.8, + "learning_rate": 2.0274736940909588e-05, + "loss": 0.4215, + "step": 26868 + }, + { + "epoch": 0.8, + "learning_rate": 2.0268916713075058e-05, + "loss": 0.5858, + "step": 26869 + }, + { + "epoch": 0.8, + "learning_rate": 2.0263097226548355e-05, + "loss": 0.4877, + "step": 26870 + }, + { + "epoch": 0.8, + "learning_rate": 2.025727848138359e-05, + "loss": 0.5508, + "step": 26871 + }, + { + "epoch": 0.8, + "learning_rate": 2.0251460477634897e-05, + "loss": 0.4025, + "step": 26872 + }, + { + "epoch": 0.8, + "learning_rate": 2.0245643215356347e-05, + "loss": 0.5014, + "step": 26873 + }, + { + "epoch": 0.8, + "learning_rate": 2.0239826694601982e-05, + "loss": 0.5669, + "step": 26874 + }, + { + "epoch": 0.8, + "learning_rate": 2.023401091542593e-05, + "loss": 0.4898, + "step": 26875 + }, + { + "epoch": 0.8, + "learning_rate": 2.0228195877882274e-05, + "loss": 0.5193, + "step": 26876 + }, + { + "epoch": 0.8, + "learning_rate": 2.0222381582025062e-05, + "loss": 0.525, + "step": 26877 + }, + { + "epoch": 0.8, + "learning_rate": 2.021656802790831e-05, + "loss": 0.648, + "step": 26878 + }, + { + "epoch": 0.8, + "learning_rate": 2.021075521558613e-05, + "loss": 0.5065, + "step": 26879 + }, + { + "epoch": 0.8, + "learning_rate": 2.0204943145112564e-05, + "loss": 0.6239, + "step": 26880 + }, + { + "epoch": 0.8, + "learning_rate": 2.0199131816541607e-05, + "loss": 0.5296, + "step": 26881 + }, + { + "epoch": 0.8, + "learning_rate": 2.0193321229927344e-05, + "loss": 0.5242, + "step": 26882 + }, + { + "epoch": 0.8, + "learning_rate": 2.0187511385323764e-05, + "loss": 0.5667, + "step": 26883 + }, + { + "epoch": 0.8, + "learning_rate": 2.0181702282784876e-05, + "loss": 0.6525, + "step": 26884 + }, + { + "epoch": 0.8, + "learning_rate": 2.017589392236471e-05, + "loss": 0.5516, + "step": 26885 + }, + { + "epoch": 0.8, + "learning_rate": 2.0170086304117287e-05, + "loss": 0.5652, + "step": 26886 + }, + { + "epoch": 0.8, + "learning_rate": 2.016427942809659e-05, + "loss": 0.4615, + "step": 26887 + }, + { + "epoch": 0.8, + "learning_rate": 2.0158473294356573e-05, + "loss": 0.4776, + "step": 26888 + }, + { + "epoch": 0.8, + "learning_rate": 2.015266790295125e-05, + "loss": 0.5158, + "step": 26889 + }, + { + "epoch": 0.8, + "learning_rate": 2.01468632539346e-05, + "loss": 0.4739, + "step": 26890 + }, + { + "epoch": 0.8, + "learning_rate": 2.0141059347360613e-05, + "loss": 0.5292, + "step": 26891 + }, + { + "epoch": 0.8, + "learning_rate": 2.0135256183283225e-05, + "loss": 0.5501, + "step": 26892 + }, + { + "epoch": 0.8, + "learning_rate": 2.0129453761756366e-05, + "loss": 0.537, + "step": 26893 + }, + { + "epoch": 0.8, + "learning_rate": 2.0123652082834022e-05, + "loss": 0.5541, + "step": 26894 + }, + { + "epoch": 0.8, + "learning_rate": 2.0117851146570143e-05, + "loss": 0.4988, + "step": 26895 + }, + { + "epoch": 0.8, + "learning_rate": 2.0112050953018656e-05, + "loss": 0.4771, + "step": 26896 + }, + { + "epoch": 0.8, + "learning_rate": 2.010625150223344e-05, + "loss": 0.5413, + "step": 26897 + }, + { + "epoch": 0.8, + "learning_rate": 2.0100452794268463e-05, + "loss": 0.4723, + "step": 26898 + }, + { + "epoch": 0.8, + "learning_rate": 2.0094654829177662e-05, + "loss": 0.5812, + "step": 26899 + }, + { + "epoch": 0.8, + "learning_rate": 2.0088857607014878e-05, + "loss": 0.5978, + "step": 26900 + }, + { + "epoch": 0.8, + "learning_rate": 2.008306112783408e-05, + "loss": 0.4801, + "step": 26901 + }, + { + "epoch": 0.8, + "learning_rate": 2.007726539168909e-05, + "loss": 0.5313, + "step": 26902 + }, + { + "epoch": 0.8, + "learning_rate": 2.0071470398633875e-05, + "loss": 0.5289, + "step": 26903 + }, + { + "epoch": 0.8, + "learning_rate": 2.0065676148722236e-05, + "loss": 0.5452, + "step": 26904 + }, + { + "epoch": 0.8, + "learning_rate": 2.0059882642008075e-05, + "loss": 0.4773, + "step": 26905 + }, + { + "epoch": 0.8, + "learning_rate": 2.0054089878545302e-05, + "loss": 0.5914, + "step": 26906 + }, + { + "epoch": 0.8, + "learning_rate": 2.0048297858387733e-05, + "loss": 0.5669, + "step": 26907 + }, + { + "epoch": 0.8, + "learning_rate": 2.0042506581589203e-05, + "loss": 0.5165, + "step": 26908 + }, + { + "epoch": 0.8, + "learning_rate": 2.0036716048203576e-05, + "loss": 0.4625, + "step": 26909 + }, + { + "epoch": 0.8, + "learning_rate": 2.003092625828471e-05, + "loss": 0.4531, + "step": 26910 + }, + { + "epoch": 0.8, + "learning_rate": 2.0025137211886434e-05, + "loss": 0.6069, + "step": 26911 + }, + { + "epoch": 0.8, + "learning_rate": 2.001934890906252e-05, + "loss": 0.4964, + "step": 26912 + }, + { + "epoch": 0.8, + "learning_rate": 2.0013561349866828e-05, + "loss": 0.4722, + "step": 26913 + }, + { + "epoch": 0.8, + "learning_rate": 2.0007774534353165e-05, + "loss": 0.5513, + "step": 26914 + }, + { + "epoch": 0.8, + "learning_rate": 2.000198846257536e-05, + "loss": 0.4631, + "step": 26915 + }, + { + "epoch": 0.8, + "learning_rate": 1.9996203134587177e-05, + "loss": 0.606, + "step": 26916 + }, + { + "epoch": 0.8, + "learning_rate": 1.999041855044238e-05, + "loss": 0.5787, + "step": 26917 + }, + { + "epoch": 0.8, + "learning_rate": 1.9984634710194817e-05, + "loss": 0.6427, + "step": 26918 + }, + { + "epoch": 0.8, + "learning_rate": 1.997885161389821e-05, + "loss": 0.502, + "step": 26919 + }, + { + "epoch": 0.8, + "learning_rate": 1.9973069261606358e-05, + "loss": 0.4837, + "step": 26920 + }, + { + "epoch": 0.8, + "learning_rate": 1.996728765337299e-05, + "loss": 0.4841, + "step": 26921 + }, + { + "epoch": 0.8, + "learning_rate": 1.9961506789251915e-05, + "loss": 0.6271, + "step": 26922 + }, + { + "epoch": 0.8, + "learning_rate": 1.9955726669296826e-05, + "loss": 0.5392, + "step": 26923 + }, + { + "epoch": 0.8, + "learning_rate": 1.9949947293561476e-05, + "loss": 0.4932, + "step": 26924 + }, + { + "epoch": 0.8, + "learning_rate": 1.9944168662099648e-05, + "loss": 0.6038, + "step": 26925 + }, + { + "epoch": 0.8, + "learning_rate": 1.993839077496503e-05, + "loss": 0.363, + "step": 26926 + }, + { + "epoch": 0.8, + "learning_rate": 1.9932613632211316e-05, + "loss": 0.5923, + "step": 26927 + }, + { + "epoch": 0.8, + "learning_rate": 1.992683723389226e-05, + "loss": 0.4561, + "step": 26928 + }, + { + "epoch": 0.8, + "learning_rate": 1.9921061580061574e-05, + "loss": 0.5831, + "step": 26929 + }, + { + "epoch": 0.8, + "learning_rate": 1.9915286670772937e-05, + "loss": 0.3849, + "step": 26930 + }, + { + "epoch": 0.8, + "learning_rate": 1.990951250608003e-05, + "loss": 0.529, + "step": 26931 + }, + { + "epoch": 0.8, + "learning_rate": 1.9903739086036542e-05, + "loss": 0.5534, + "step": 26932 + }, + { + "epoch": 0.8, + "learning_rate": 1.9897966410696178e-05, + "loss": 0.5125, + "step": 26933 + }, + { + "epoch": 0.8, + "learning_rate": 1.9892194480112612e-05, + "loss": 0.6869, + "step": 26934 + }, + { + "epoch": 0.8, + "learning_rate": 1.9886423294339507e-05, + "loss": 0.5691, + "step": 26935 + }, + { + "epoch": 0.8, + "learning_rate": 1.988065285343047e-05, + "loss": 0.4911, + "step": 26936 + }, + { + "epoch": 0.8, + "learning_rate": 1.9874883157439195e-05, + "loss": 0.5948, + "step": 26937 + }, + { + "epoch": 0.8, + "learning_rate": 1.986911420641935e-05, + "loss": 0.5857, + "step": 26938 + }, + { + "epoch": 0.8, + "learning_rate": 1.9863346000424542e-05, + "loss": 0.6111, + "step": 26939 + }, + { + "epoch": 0.8, + "learning_rate": 1.985757853950837e-05, + "loss": 0.4404, + "step": 26940 + }, + { + "epoch": 0.8, + "learning_rate": 1.9851811823724497e-05, + "loss": 0.5231, + "step": 26941 + }, + { + "epoch": 0.8, + "learning_rate": 1.984604585312656e-05, + "loss": 0.4752, + "step": 26942 + }, + { + "epoch": 0.8, + "learning_rate": 1.9840280627768127e-05, + "loss": 0.4993, + "step": 26943 + }, + { + "epoch": 0.8, + "learning_rate": 1.9834516147702832e-05, + "loss": 0.5367, + "step": 26944 + }, + { + "epoch": 0.8, + "learning_rate": 1.982875241298423e-05, + "loss": 0.5269, + "step": 26945 + }, + { + "epoch": 0.8, + "learning_rate": 1.982298942366596e-05, + "loss": 0.5842, + "step": 26946 + }, + { + "epoch": 0.8, + "learning_rate": 1.9817227179801545e-05, + "loss": 0.5856, + "step": 26947 + }, + { + "epoch": 0.8, + "learning_rate": 1.981146568144463e-05, + "loss": 0.6207, + "step": 26948 + }, + { + "epoch": 0.8, + "learning_rate": 1.980570492864875e-05, + "loss": 0.447, + "step": 26949 + }, + { + "epoch": 0.8, + "learning_rate": 1.9799944921467428e-05, + "loss": 0.5277, + "step": 26950 + }, + { + "epoch": 0.8, + "learning_rate": 1.9794185659954255e-05, + "loss": 0.4448, + "step": 26951 + }, + { + "epoch": 0.8, + "learning_rate": 1.978842714416279e-05, + "loss": 0.4823, + "step": 26952 + }, + { + "epoch": 0.8, + "learning_rate": 1.9782669374146567e-05, + "loss": 0.6153, + "step": 26953 + }, + { + "epoch": 0.8, + "learning_rate": 1.9776912349959133e-05, + "loss": 0.5729, + "step": 26954 + }, + { + "epoch": 0.8, + "learning_rate": 1.977115607165395e-05, + "loss": 0.4435, + "step": 26955 + }, + { + "epoch": 0.8, + "learning_rate": 1.9765400539284595e-05, + "loss": 0.5908, + "step": 26956 + }, + { + "epoch": 0.8, + "learning_rate": 1.9759645752904588e-05, + "loss": 0.5403, + "step": 26957 + }, + { + "epoch": 0.8, + "learning_rate": 1.9753891712567408e-05, + "loss": 0.558, + "step": 26958 + }, + { + "epoch": 0.8, + "learning_rate": 1.9748138418326546e-05, + "loss": 0.3937, + "step": 26959 + }, + { + "epoch": 0.8, + "learning_rate": 1.9742385870235492e-05, + "loss": 0.39, + "step": 26960 + }, + { + "epoch": 0.8, + "learning_rate": 1.9736634068347783e-05, + "loss": 0.6363, + "step": 26961 + }, + { + "epoch": 0.8, + "learning_rate": 1.973088301271684e-05, + "loss": 0.5433, + "step": 26962 + }, + { + "epoch": 0.8, + "learning_rate": 1.9725132703396166e-05, + "loss": 0.5429, + "step": 26963 + }, + { + "epoch": 0.8, + "learning_rate": 1.9719383140439197e-05, + "loss": 0.586, + "step": 26964 + }, + { + "epoch": 0.8, + "learning_rate": 1.9713634323899433e-05, + "loss": 0.5441, + "step": 26965 + }, + { + "epoch": 0.8, + "learning_rate": 1.9707886253830275e-05, + "loss": 0.508, + "step": 26966 + }, + { + "epoch": 0.8, + "learning_rate": 1.9702138930285186e-05, + "loss": 0.4955, + "step": 26967 + }, + { + "epoch": 0.8, + "learning_rate": 1.9696392353317627e-05, + "loss": 0.5665, + "step": 26968 + }, + { + "epoch": 0.8, + "learning_rate": 1.9690646522981015e-05, + "loss": 0.5264, + "step": 26969 + }, + { + "epoch": 0.8, + "learning_rate": 1.968490143932873e-05, + "loss": 0.4932, + "step": 26970 + }, + { + "epoch": 0.8, + "learning_rate": 1.9679157102414224e-05, + "loss": 0.4689, + "step": 26971 + }, + { + "epoch": 0.8, + "learning_rate": 1.9673413512290926e-05, + "loss": 0.5349, + "step": 26972 + }, + { + "epoch": 0.8, + "learning_rate": 1.966767066901222e-05, + "loss": 0.4377, + "step": 26973 + }, + { + "epoch": 0.8, + "learning_rate": 1.966192857263146e-05, + "loss": 0.4783, + "step": 26974 + }, + { + "epoch": 0.8, + "learning_rate": 1.9656187223202073e-05, + "loss": 0.5057, + "step": 26975 + }, + { + "epoch": 0.8, + "learning_rate": 1.965044662077744e-05, + "loss": 0.4581, + "step": 26976 + }, + { + "epoch": 0.8, + "learning_rate": 1.9644706765410947e-05, + "loss": 0.4437, + "step": 26977 + }, + { + "epoch": 0.8, + "learning_rate": 1.963896765715595e-05, + "loss": 0.5468, + "step": 26978 + }, + { + "epoch": 0.8, + "learning_rate": 1.963322929606578e-05, + "loss": 0.5427, + "step": 26979 + }, + { + "epoch": 0.8, + "learning_rate": 1.96274916821938e-05, + "loss": 0.4542, + "step": 26980 + }, + { + "epoch": 0.8, + "learning_rate": 1.962175481559341e-05, + "loss": 0.5466, + "step": 26981 + }, + { + "epoch": 0.8, + "learning_rate": 1.9616018696317894e-05, + "loss": 0.5196, + "step": 26982 + }, + { + "epoch": 0.8, + "learning_rate": 1.9610283324420586e-05, + "loss": 0.4887, + "step": 26983 + }, + { + "epoch": 0.8, + "learning_rate": 1.960454869995484e-05, + "loss": 0.5147, + "step": 26984 + }, + { + "epoch": 0.8, + "learning_rate": 1.9598814822973943e-05, + "loss": 0.4597, + "step": 26985 + }, + { + "epoch": 0.8, + "learning_rate": 1.9593081693531213e-05, + "loss": 0.492, + "step": 26986 + }, + { + "epoch": 0.8, + "learning_rate": 1.958734931167998e-05, + "loss": 0.585, + "step": 26987 + }, + { + "epoch": 0.8, + "learning_rate": 1.9581617677473542e-05, + "loss": 0.4999, + "step": 26988 + }, + { + "epoch": 0.8, + "learning_rate": 1.9575886790965138e-05, + "loss": 0.5304, + "step": 26989 + }, + { + "epoch": 0.8, + "learning_rate": 1.9570156652208072e-05, + "loss": 0.5369, + "step": 26990 + }, + { + "epoch": 0.8, + "learning_rate": 1.956442726125567e-05, + "loss": 0.5429, + "step": 26991 + }, + { + "epoch": 0.8, + "learning_rate": 1.9558698618161163e-05, + "loss": 0.497, + "step": 26992 + }, + { + "epoch": 0.8, + "learning_rate": 1.955297072297779e-05, + "loss": 0.5259, + "step": 26993 + }, + { + "epoch": 0.8, + "learning_rate": 1.9547243575758844e-05, + "loss": 0.5261, + "step": 26994 + }, + { + "epoch": 0.8, + "learning_rate": 1.9541517176557546e-05, + "loss": 0.6471, + "step": 26995 + }, + { + "epoch": 0.8, + "learning_rate": 1.9535791525427184e-05, + "loss": 0.4997, + "step": 26996 + }, + { + "epoch": 0.8, + "learning_rate": 1.953006662242096e-05, + "loss": 0.5816, + "step": 26997 + }, + { + "epoch": 0.8, + "learning_rate": 1.9524342467592084e-05, + "loss": 0.5428, + "step": 26998 + }, + { + "epoch": 0.8, + "learning_rate": 1.9518619060993804e-05, + "loss": 0.5948, + "step": 26999 + }, + { + "epoch": 0.8, + "learning_rate": 1.9512896402679338e-05, + "loss": 0.562, + "step": 27000 + }, + { + "epoch": 0.8, + "learning_rate": 1.9507174492701885e-05, + "loss": 0.5962, + "step": 27001 + }, + { + "epoch": 0.8, + "learning_rate": 1.9501453331114626e-05, + "loss": 0.48, + "step": 27002 + }, + { + "epoch": 0.8, + "learning_rate": 1.9495732917970765e-05, + "loss": 0.5751, + "step": 27003 + }, + { + "epoch": 0.8, + "learning_rate": 1.9490013253323513e-05, + "loss": 0.5558, + "step": 27004 + }, + { + "epoch": 0.8, + "learning_rate": 1.9484294337226018e-05, + "loss": 0.4709, + "step": 27005 + }, + { + "epoch": 0.8, + "learning_rate": 1.947857616973148e-05, + "loss": 0.4792, + "step": 27006 + }, + { + "epoch": 0.8, + "learning_rate": 1.947285875089302e-05, + "loss": 0.4975, + "step": 27007 + }, + { + "epoch": 0.8, + "learning_rate": 1.9467142080763857e-05, + "loss": 0.5387, + "step": 27008 + }, + { + "epoch": 0.8, + "learning_rate": 1.9461426159397077e-05, + "loss": 0.5065, + "step": 27009 + }, + { + "epoch": 0.8, + "learning_rate": 1.9455710986845864e-05, + "loss": 0.4198, + "step": 27010 + }, + { + "epoch": 0.8, + "learning_rate": 1.9449996563163374e-05, + "loss": 0.5538, + "step": 27011 + }, + { + "epoch": 0.8, + "learning_rate": 1.944428288840271e-05, + "loss": 0.4948, + "step": 27012 + }, + { + "epoch": 0.8, + "learning_rate": 1.9438569962616972e-05, + "loss": 0.5118, + "step": 27013 + }, + { + "epoch": 0.8, + "learning_rate": 1.9432857785859293e-05, + "loss": 0.5008, + "step": 27014 + }, + { + "epoch": 0.8, + "learning_rate": 1.9427146358182825e-05, + "loss": 0.6308, + "step": 27015 + }, + { + "epoch": 0.8, + "learning_rate": 1.9421435679640632e-05, + "loss": 0.5686, + "step": 27016 + }, + { + "epoch": 0.8, + "learning_rate": 1.9415725750285796e-05, + "loss": 0.6157, + "step": 27017 + }, + { + "epoch": 0.8, + "learning_rate": 1.941001657017142e-05, + "loss": 0.5065, + "step": 27018 + }, + { + "epoch": 0.8, + "learning_rate": 1.9404308139350623e-05, + "loss": 0.4847, + "step": 27019 + }, + { + "epoch": 0.8, + "learning_rate": 1.9398600457876413e-05, + "loss": 0.5692, + "step": 27020 + }, + { + "epoch": 0.8, + "learning_rate": 1.9392893525801925e-05, + "loss": 0.4841, + "step": 27021 + }, + { + "epoch": 0.8, + "learning_rate": 1.9387187343180146e-05, + "loss": 0.4497, + "step": 27022 + }, + { + "epoch": 0.8, + "learning_rate": 1.9381481910064205e-05, + "loss": 0.4131, + "step": 27023 + }, + { + "epoch": 0.8, + "learning_rate": 1.9375777226507085e-05, + "loss": 0.5264, + "step": 27024 + }, + { + "epoch": 0.8, + "learning_rate": 1.9370073292561886e-05, + "loss": 0.5423, + "step": 27025 + }, + { + "epoch": 0.8, + "learning_rate": 1.9364370108281583e-05, + "loss": 0.5637, + "step": 27026 + }, + { + "epoch": 0.8, + "learning_rate": 1.9358667673719254e-05, + "loss": 0.6175, + "step": 27027 + }, + { + "epoch": 0.8, + "learning_rate": 1.9352965988927863e-05, + "loss": 0.4661, + "step": 27028 + }, + { + "epoch": 0.8, + "learning_rate": 1.9347265053960452e-05, + "loss": 0.3577, + "step": 27029 + }, + { + "epoch": 0.8, + "learning_rate": 1.9341564868870056e-05, + "loss": 0.5287, + "step": 27030 + }, + { + "epoch": 0.8, + "learning_rate": 1.9335865433709644e-05, + "loss": 0.4881, + "step": 27031 + }, + { + "epoch": 0.8, + "learning_rate": 1.9330166748532174e-05, + "loss": 0.5909, + "step": 27032 + }, + { + "epoch": 0.8, + "learning_rate": 1.9324468813390662e-05, + "loss": 0.5525, + "step": 27033 + }, + { + "epoch": 0.8, + "learning_rate": 1.9318771628338106e-05, + "loss": 0.5926, + "step": 27034 + }, + { + "epoch": 0.8, + "learning_rate": 1.9313075193427466e-05, + "loss": 0.4858, + "step": 27035 + }, + { + "epoch": 0.8, + "learning_rate": 1.9307379508711664e-05, + "loss": 0.465, + "step": 27036 + }, + { + "epoch": 0.8, + "learning_rate": 1.9301684574243684e-05, + "loss": 0.6034, + "step": 27037 + }, + { + "epoch": 0.8, + "learning_rate": 1.929599039007648e-05, + "loss": 0.4844, + "step": 27038 + }, + { + "epoch": 0.8, + "learning_rate": 1.929029695626301e-05, + "loss": 0.5828, + "step": 27039 + }, + { + "epoch": 0.81, + "learning_rate": 1.92846042728562e-05, + "loss": 0.5996, + "step": 27040 + }, + { + "epoch": 0.81, + "learning_rate": 1.927891233990894e-05, + "loss": 0.4451, + "step": 27041 + }, + { + "epoch": 0.81, + "learning_rate": 1.9273221157474187e-05, + "loss": 0.4705, + "step": 27042 + }, + { + "epoch": 0.81, + "learning_rate": 1.9267530725604866e-05, + "loss": 0.5611, + "step": 27043 + }, + { + "epoch": 0.81, + "learning_rate": 1.926184104435388e-05, + "loss": 0.6417, + "step": 27044 + }, + { + "epoch": 0.81, + "learning_rate": 1.9256152113774074e-05, + "loss": 0.4347, + "step": 27045 + }, + { + "epoch": 0.81, + "learning_rate": 1.9250463933918396e-05, + "loss": 0.5226, + "step": 27046 + }, + { + "epoch": 0.81, + "learning_rate": 1.9244776504839745e-05, + "loss": 0.5219, + "step": 27047 + }, + { + "epoch": 0.81, + "learning_rate": 1.923908982659095e-05, + "loss": 0.5847, + "step": 27048 + }, + { + "epoch": 0.81, + "learning_rate": 1.9233403899224935e-05, + "loss": 0.5423, + "step": 27049 + }, + { + "epoch": 0.81, + "learning_rate": 1.922771872279454e-05, + "loss": 0.5407, + "step": 27050 + }, + { + "epoch": 0.81, + "learning_rate": 1.92220342973526e-05, + "loss": 0.4712, + "step": 27051 + }, + { + "epoch": 0.81, + "learning_rate": 1.9216350622951985e-05, + "loss": 0.553, + "step": 27052 + }, + { + "epoch": 0.81, + "learning_rate": 1.9210667699645578e-05, + "loss": 0.5946, + "step": 27053 + }, + { + "epoch": 0.81, + "learning_rate": 1.9204985527486175e-05, + "loss": 0.526, + "step": 27054 + }, + { + "epoch": 0.81, + "learning_rate": 1.9199304106526596e-05, + "loss": 0.5321, + "step": 27055 + }, + { + "epoch": 0.81, + "learning_rate": 1.9193623436819673e-05, + "loss": 0.4743, + "step": 27056 + }, + { + "epoch": 0.81, + "learning_rate": 1.918794351841825e-05, + "loss": 0.5981, + "step": 27057 + }, + { + "epoch": 0.81, + "learning_rate": 1.9182264351375134e-05, + "loss": 0.5423, + "step": 27058 + }, + { + "epoch": 0.81, + "learning_rate": 1.917658593574311e-05, + "loss": 0.5155, + "step": 27059 + }, + { + "epoch": 0.81, + "learning_rate": 1.917090827157496e-05, + "loss": 0.5836, + "step": 27060 + }, + { + "epoch": 0.81, + "learning_rate": 1.9165231358923485e-05, + "loss": 0.5184, + "step": 27061 + }, + { + "epoch": 0.81, + "learning_rate": 1.9159555197841495e-05, + "loss": 0.5096, + "step": 27062 + }, + { + "epoch": 0.81, + "learning_rate": 1.9153879788381746e-05, + "loss": 0.5067, + "step": 27063 + }, + { + "epoch": 0.81, + "learning_rate": 1.9148205130596974e-05, + "loss": 0.5054, + "step": 27064 + }, + { + "epoch": 0.81, + "learning_rate": 1.914253122453996e-05, + "loss": 0.6307, + "step": 27065 + }, + { + "epoch": 0.81, + "learning_rate": 1.91368580702635e-05, + "loss": 0.5619, + "step": 27066 + }, + { + "epoch": 0.81, + "learning_rate": 1.9131185667820274e-05, + "loss": 0.4754, + "step": 27067 + }, + { + "epoch": 0.81, + "learning_rate": 1.912551401726308e-05, + "loss": 0.5051, + "step": 27068 + }, + { + "epoch": 0.81, + "learning_rate": 1.9119843118644597e-05, + "loss": 0.5704, + "step": 27069 + }, + { + "epoch": 0.81, + "learning_rate": 1.9114172972017596e-05, + "loss": 0.52, + "step": 27070 + }, + { + "epoch": 0.81, + "learning_rate": 1.9108503577434754e-05, + "loss": 0.5273, + "step": 27071 + }, + { + "epoch": 0.81, + "learning_rate": 1.910283493494881e-05, + "loss": 0.4738, + "step": 27072 + }, + { + "epoch": 0.81, + "learning_rate": 1.9097167044612497e-05, + "loss": 0.4912, + "step": 27073 + }, + { + "epoch": 0.81, + "learning_rate": 1.909149990647847e-05, + "loss": 0.5384, + "step": 27074 + }, + { + "epoch": 0.81, + "learning_rate": 1.9085833520599405e-05, + "loss": 0.5159, + "step": 27075 + }, + { + "epoch": 0.81, + "learning_rate": 1.908016788702801e-05, + "loss": 0.5472, + "step": 27076 + }, + { + "epoch": 0.81, + "learning_rate": 1.9074503005816992e-05, + "loss": 0.5106, + "step": 27077 + }, + { + "epoch": 0.81, + "learning_rate": 1.906883887701899e-05, + "loss": 0.417, + "step": 27078 + }, + { + "epoch": 0.81, + "learning_rate": 1.9063175500686636e-05, + "loss": 0.5274, + "step": 27079 + }, + { + "epoch": 0.81, + "learning_rate": 1.9057512876872618e-05, + "loss": 0.5208, + "step": 27080 + }, + { + "epoch": 0.81, + "learning_rate": 1.9051851005629617e-05, + "loss": 0.5664, + "step": 27081 + }, + { + "epoch": 0.81, + "learning_rate": 1.9046189887010223e-05, + "loss": 0.3811, + "step": 27082 + }, + { + "epoch": 0.81, + "learning_rate": 1.9040529521067108e-05, + "loss": 0.5126, + "step": 27083 + }, + { + "epoch": 0.81, + "learning_rate": 1.903486990785286e-05, + "loss": 0.4683, + "step": 27084 + }, + { + "epoch": 0.81, + "learning_rate": 1.9029211047420137e-05, + "loss": 0.542, + "step": 27085 + }, + { + "epoch": 0.81, + "learning_rate": 1.902355293982152e-05, + "loss": 0.5142, + "step": 27086 + }, + { + "epoch": 0.81, + "learning_rate": 1.9017895585109668e-05, + "loss": 0.5209, + "step": 27087 + }, + { + "epoch": 0.81, + "learning_rate": 1.9012238983337104e-05, + "loss": 0.4799, + "step": 27088 + }, + { + "epoch": 0.81, + "learning_rate": 1.90065831345565e-05, + "loss": 0.5216, + "step": 27089 + }, + { + "epoch": 0.81, + "learning_rate": 1.9000928038820377e-05, + "loss": 0.649, + "step": 27090 + }, + { + "epoch": 0.81, + "learning_rate": 1.8995273696181347e-05, + "loss": 0.5015, + "step": 27091 + }, + { + "epoch": 0.81, + "learning_rate": 1.8989620106691996e-05, + "loss": 0.5843, + "step": 27092 + }, + { + "epoch": 0.81, + "learning_rate": 1.8983967270404866e-05, + "loss": 0.4406, + "step": 27093 + }, + { + "epoch": 0.81, + "learning_rate": 1.897831518737251e-05, + "loss": 0.515, + "step": 27094 + }, + { + "epoch": 0.81, + "learning_rate": 1.8972663857647467e-05, + "loss": 0.5017, + "step": 27095 + }, + { + "epoch": 0.81, + "learning_rate": 1.896701328128234e-05, + "loss": 0.5731, + "step": 27096 + }, + { + "epoch": 0.81, + "learning_rate": 1.8961363458329616e-05, + "loss": 0.4338, + "step": 27097 + }, + { + "epoch": 0.81, + "learning_rate": 1.895571438884183e-05, + "loss": 0.4749, + "step": 27098 + }, + { + "epoch": 0.81, + "learning_rate": 1.8950066072871485e-05, + "loss": 0.5948, + "step": 27099 + }, + { + "epoch": 0.81, + "learning_rate": 1.894441851047114e-05, + "loss": 0.5452, + "step": 27100 + }, + { + "epoch": 0.81, + "learning_rate": 1.8938771701693305e-05, + "loss": 0.5332, + "step": 27101 + }, + { + "epoch": 0.81, + "learning_rate": 1.8933125646590466e-05, + "loss": 0.4881, + "step": 27102 + }, + { + "epoch": 0.81, + "learning_rate": 1.8927480345215088e-05, + "loss": 0.4795, + "step": 27103 + }, + { + "epoch": 0.81, + "learning_rate": 1.892183579761968e-05, + "loss": 0.57, + "step": 27104 + }, + { + "epoch": 0.81, + "learning_rate": 1.891619200385676e-05, + "loss": 0.5454, + "step": 27105 + }, + { + "epoch": 0.81, + "learning_rate": 1.891054896397877e-05, + "loss": 0.3765, + "step": 27106 + }, + { + "epoch": 0.81, + "learning_rate": 1.8904906678038147e-05, + "loss": 0.541, + "step": 27107 + }, + { + "epoch": 0.81, + "learning_rate": 1.8899265146087388e-05, + "loss": 0.5568, + "step": 27108 + }, + { + "epoch": 0.81, + "learning_rate": 1.889362436817895e-05, + "loss": 0.4725, + "step": 27109 + }, + { + "epoch": 0.81, + "learning_rate": 1.8887984344365252e-05, + "loss": 0.5702, + "step": 27110 + }, + { + "epoch": 0.81, + "learning_rate": 1.8882345074698772e-05, + "loss": 0.572, + "step": 27111 + }, + { + "epoch": 0.81, + "learning_rate": 1.8876706559231892e-05, + "loss": 0.4904, + "step": 27112 + }, + { + "epoch": 0.81, + "learning_rate": 1.8871068798017087e-05, + "loss": 0.4622, + "step": 27113 + }, + { + "epoch": 0.81, + "learning_rate": 1.8865431791106725e-05, + "loss": 0.5491, + "step": 27114 + }, + { + "epoch": 0.81, + "learning_rate": 1.885979553855326e-05, + "loss": 0.4891, + "step": 27115 + }, + { + "epoch": 0.81, + "learning_rate": 1.8854160040409086e-05, + "loss": 0.5918, + "step": 27116 + }, + { + "epoch": 0.81, + "learning_rate": 1.884852529672656e-05, + "loss": 0.5339, + "step": 27117 + }, + { + "epoch": 0.81, + "learning_rate": 1.8842891307558098e-05, + "loss": 0.4455, + "step": 27118 + }, + { + "epoch": 0.81, + "learning_rate": 1.8837258072956088e-05, + "loss": 0.482, + "step": 27119 + }, + { + "epoch": 0.81, + "learning_rate": 1.8831625592972936e-05, + "loss": 0.5591, + "step": 27120 + }, + { + "epoch": 0.81, + "learning_rate": 1.8825993867660974e-05, + "loss": 0.6348, + "step": 27121 + }, + { + "epoch": 0.81, + "learning_rate": 1.8820362897072553e-05, + "loss": 0.5069, + "step": 27122 + }, + { + "epoch": 0.81, + "learning_rate": 1.8814732681260026e-05, + "loss": 0.6317, + "step": 27123 + }, + { + "epoch": 0.81, + "learning_rate": 1.880910322027579e-05, + "loss": 0.6862, + "step": 27124 + }, + { + "epoch": 0.81, + "learning_rate": 1.880347451417216e-05, + "loss": 0.5052, + "step": 27125 + }, + { + "epoch": 0.81, + "learning_rate": 1.8797846563001432e-05, + "loss": 0.6661, + "step": 27126 + }, + { + "epoch": 0.81, + "learning_rate": 1.8792219366815954e-05, + "loss": 0.5072, + "step": 27127 + }, + { + "epoch": 0.81, + "learning_rate": 1.878659292566809e-05, + "loss": 0.5733, + "step": 27128 + }, + { + "epoch": 0.81, + "learning_rate": 1.878096723961008e-05, + "loss": 0.551, + "step": 27129 + }, + { + "epoch": 0.81, + "learning_rate": 1.8775342308694298e-05, + "loss": 0.4461, + "step": 27130 + }, + { + "epoch": 0.81, + "learning_rate": 1.8769718132972992e-05, + "loss": 0.5283, + "step": 27131 + }, + { + "epoch": 0.81, + "learning_rate": 1.8764094712498493e-05, + "loss": 0.4712, + "step": 27132 + }, + { + "epoch": 0.81, + "learning_rate": 1.875847204732303e-05, + "loss": 0.6197, + "step": 27133 + }, + { + "epoch": 0.81, + "learning_rate": 1.8752850137498924e-05, + "loss": 0.4623, + "step": 27134 + }, + { + "epoch": 0.81, + "learning_rate": 1.874722898307846e-05, + "loss": 0.4398, + "step": 27135 + }, + { + "epoch": 0.81, + "learning_rate": 1.874160858411387e-05, + "loss": 0.6582, + "step": 27136 + }, + { + "epoch": 0.81, + "learning_rate": 1.8735988940657388e-05, + "loss": 0.5254, + "step": 27137 + }, + { + "epoch": 0.81, + "learning_rate": 1.8730370052761304e-05, + "loss": 0.5282, + "step": 27138 + }, + { + "epoch": 0.81, + "learning_rate": 1.8724751920477866e-05, + "loss": 0.4777, + "step": 27139 + }, + { + "epoch": 0.81, + "learning_rate": 1.8719134543859294e-05, + "loss": 0.4928, + "step": 27140 + }, + { + "epoch": 0.81, + "learning_rate": 1.8713517922957802e-05, + "loss": 0.5944, + "step": 27141 + }, + { + "epoch": 0.81, + "learning_rate": 1.87079020578256e-05, + "loss": 0.449, + "step": 27142 + }, + { + "epoch": 0.81, + "learning_rate": 1.8702286948514935e-05, + "loss": 0.4786, + "step": 27143 + }, + { + "epoch": 0.81, + "learning_rate": 1.8696672595078034e-05, + "loss": 0.5887, + "step": 27144 + }, + { + "epoch": 0.81, + "learning_rate": 1.8691058997567056e-05, + "loss": 0.5753, + "step": 27145 + }, + { + "epoch": 0.81, + "learning_rate": 1.868544615603419e-05, + "loss": 0.5026, + "step": 27146 + }, + { + "epoch": 0.81, + "learning_rate": 1.8679834070531654e-05, + "loss": 0.574, + "step": 27147 + }, + { + "epoch": 0.81, + "learning_rate": 1.867422274111159e-05, + "loss": 0.5348, + "step": 27148 + }, + { + "epoch": 0.81, + "learning_rate": 1.8668612167826215e-05, + "loss": 0.495, + "step": 27149 + }, + { + "epoch": 0.81, + "learning_rate": 1.8663002350727642e-05, + "loss": 0.5093, + "step": 27150 + }, + { + "epoch": 0.81, + "learning_rate": 1.8657393289868084e-05, + "loss": 0.4247, + "step": 27151 + }, + { + "epoch": 0.81, + "learning_rate": 1.8651784985299637e-05, + "loss": 0.4747, + "step": 27152 + }, + { + "epoch": 0.81, + "learning_rate": 1.8646177437074464e-05, + "loss": 0.4961, + "step": 27153 + }, + { + "epoch": 0.81, + "learning_rate": 1.8640570645244736e-05, + "loss": 0.4996, + "step": 27154 + }, + { + "epoch": 0.81, + "learning_rate": 1.863496460986256e-05, + "loss": 0.5862, + "step": 27155 + }, + { + "epoch": 0.81, + "learning_rate": 1.8629359330980024e-05, + "loss": 0.6368, + "step": 27156 + }, + { + "epoch": 0.81, + "learning_rate": 1.8623754808649273e-05, + "loss": 0.4611, + "step": 27157 + }, + { + "epoch": 0.81, + "learning_rate": 1.861815104292245e-05, + "loss": 0.4298, + "step": 27158 + }, + { + "epoch": 0.81, + "learning_rate": 1.8612548033851617e-05, + "loss": 0.4435, + "step": 27159 + }, + { + "epoch": 0.81, + "learning_rate": 1.860694578148885e-05, + "loss": 0.486, + "step": 27160 + }, + { + "epoch": 0.81, + "learning_rate": 1.860134428588626e-05, + "loss": 0.4155, + "step": 27161 + }, + { + "epoch": 0.81, + "learning_rate": 1.8595743547095934e-05, + "loss": 0.5798, + "step": 27162 + }, + { + "epoch": 0.81, + "learning_rate": 1.859014356516996e-05, + "loss": 0.582, + "step": 27163 + }, + { + "epoch": 0.81, + "learning_rate": 1.858454434016038e-05, + "loss": 0.5489, + "step": 27164 + }, + { + "epoch": 0.81, + "learning_rate": 1.8578945872119247e-05, + "loss": 0.4855, + "step": 27165 + }, + { + "epoch": 0.81, + "learning_rate": 1.8573348161098614e-05, + "loss": 0.5277, + "step": 27166 + }, + { + "epoch": 0.81, + "learning_rate": 1.8567751207150562e-05, + "loss": 0.5191, + "step": 27167 + }, + { + "epoch": 0.81, + "learning_rate": 1.8562155010327097e-05, + "loss": 0.5651, + "step": 27168 + }, + { + "epoch": 0.81, + "learning_rate": 1.8556559570680245e-05, + "loss": 0.5742, + "step": 27169 + }, + { + "epoch": 0.81, + "learning_rate": 1.8550964888262036e-05, + "loss": 0.512, + "step": 27170 + }, + { + "epoch": 0.81, + "learning_rate": 1.854537096312452e-05, + "loss": 0.6033, + "step": 27171 + }, + { + "epoch": 0.81, + "learning_rate": 1.8539777795319646e-05, + "loss": 0.4002, + "step": 27172 + }, + { + "epoch": 0.81, + "learning_rate": 1.8534185384899484e-05, + "loss": 0.5509, + "step": 27173 + }, + { + "epoch": 0.81, + "learning_rate": 1.8528593731915977e-05, + "loss": 0.5505, + "step": 27174 + }, + { + "epoch": 0.81, + "learning_rate": 1.8523002836421153e-05, + "loss": 0.4744, + "step": 27175 + }, + { + "epoch": 0.81, + "learning_rate": 1.8517412698466952e-05, + "loss": 0.5322, + "step": 27176 + }, + { + "epoch": 0.81, + "learning_rate": 1.851182331810537e-05, + "loss": 0.5249, + "step": 27177 + }, + { + "epoch": 0.81, + "learning_rate": 1.8506234695388404e-05, + "loss": 0.5302, + "step": 27178 + }, + { + "epoch": 0.81, + "learning_rate": 1.8500646830367984e-05, + "loss": 0.542, + "step": 27179 + }, + { + "epoch": 0.81, + "learning_rate": 1.849505972309604e-05, + "loss": 0.5046, + "step": 27180 + }, + { + "epoch": 0.81, + "learning_rate": 1.8489473373624555e-05, + "loss": 0.5235, + "step": 27181 + }, + { + "epoch": 0.81, + "learning_rate": 1.848388778200548e-05, + "loss": 0.4892, + "step": 27182 + }, + { + "epoch": 0.81, + "learning_rate": 1.8478302948290726e-05, + "loss": 0.5527, + "step": 27183 + }, + { + "epoch": 0.81, + "learning_rate": 1.8472718872532193e-05, + "loss": 0.6442, + "step": 27184 + }, + { + "epoch": 0.81, + "learning_rate": 1.8467135554781834e-05, + "loss": 0.4846, + "step": 27185 + }, + { + "epoch": 0.81, + "learning_rate": 1.846155299509157e-05, + "loss": 0.4831, + "step": 27186 + }, + { + "epoch": 0.81, + "learning_rate": 1.8455971193513267e-05, + "loss": 0.4754, + "step": 27187 + }, + { + "epoch": 0.81, + "learning_rate": 1.8450390150098875e-05, + "loss": 0.5521, + "step": 27188 + }, + { + "epoch": 0.81, + "learning_rate": 1.8444809864900214e-05, + "loss": 0.4708, + "step": 27189 + }, + { + "epoch": 0.81, + "learning_rate": 1.843923033796924e-05, + "loss": 0.5313, + "step": 27190 + }, + { + "epoch": 0.81, + "learning_rate": 1.8433651569357768e-05, + "loss": 0.6341, + "step": 27191 + }, + { + "epoch": 0.81, + "learning_rate": 1.842807355911772e-05, + "loss": 0.5546, + "step": 27192 + }, + { + "epoch": 0.81, + "learning_rate": 1.8422496307300908e-05, + "loss": 0.5373, + "step": 27193 + }, + { + "epoch": 0.81, + "learning_rate": 1.8416919813959243e-05, + "loss": 0.6039, + "step": 27194 + }, + { + "epoch": 0.81, + "learning_rate": 1.841134407914451e-05, + "loss": 0.5887, + "step": 27195 + }, + { + "epoch": 0.81, + "learning_rate": 1.8405769102908575e-05, + "loss": 0.4522, + "step": 27196 + }, + { + "epoch": 0.81, + "learning_rate": 1.8400194885303313e-05, + "loss": 0.5636, + "step": 27197 + }, + { + "epoch": 0.81, + "learning_rate": 1.839462142638052e-05, + "loss": 0.7342, + "step": 27198 + }, + { + "epoch": 0.81, + "learning_rate": 1.8389048726191982e-05, + "loss": 0.5162, + "step": 27199 + }, + { + "epoch": 0.81, + "learning_rate": 1.8383476784789543e-05, + "loss": 0.5456, + "step": 27200 + }, + { + "epoch": 0.81, + "learning_rate": 1.8377905602225033e-05, + "loss": 0.5752, + "step": 27201 + }, + { + "epoch": 0.81, + "learning_rate": 1.8372335178550215e-05, + "loss": 0.5382, + "step": 27202 + }, + { + "epoch": 0.81, + "learning_rate": 1.8366765513816873e-05, + "loss": 0.6012, + "step": 27203 + }, + { + "epoch": 0.81, + "learning_rate": 1.8361196608076813e-05, + "loss": 0.5499, + "step": 27204 + }, + { + "epoch": 0.81, + "learning_rate": 1.8355628461381812e-05, + "loss": 0.5768, + "step": 27205 + }, + { + "epoch": 0.81, + "learning_rate": 1.8350061073783653e-05, + "loss": 0.547, + "step": 27206 + }, + { + "epoch": 0.81, + "learning_rate": 1.834449444533409e-05, + "loss": 0.4666, + "step": 27207 + }, + { + "epoch": 0.81, + "learning_rate": 1.8338928576084846e-05, + "loss": 0.5872, + "step": 27208 + }, + { + "epoch": 0.81, + "learning_rate": 1.8333363466087694e-05, + "loss": 0.6191, + "step": 27209 + }, + { + "epoch": 0.81, + "learning_rate": 1.8327799115394417e-05, + "loss": 0.5572, + "step": 27210 + }, + { + "epoch": 0.81, + "learning_rate": 1.8322235524056697e-05, + "loss": 0.4752, + "step": 27211 + }, + { + "epoch": 0.81, + "learning_rate": 1.8316672692126258e-05, + "loss": 0.4516, + "step": 27212 + }, + { + "epoch": 0.81, + "learning_rate": 1.831111061965487e-05, + "loss": 0.5952, + "step": 27213 + }, + { + "epoch": 0.81, + "learning_rate": 1.830554930669419e-05, + "loss": 0.5715, + "step": 27214 + }, + { + "epoch": 0.81, + "learning_rate": 1.829998875329595e-05, + "loss": 0.5336, + "step": 27215 + }, + { + "epoch": 0.81, + "learning_rate": 1.8294428959511878e-05, + "loss": 0.5784, + "step": 27216 + }, + { + "epoch": 0.81, + "learning_rate": 1.8288869925393638e-05, + "loss": 0.5065, + "step": 27217 + }, + { + "epoch": 0.81, + "learning_rate": 1.8283311650992896e-05, + "loss": 0.5178, + "step": 27218 + }, + { + "epoch": 0.81, + "learning_rate": 1.8277754136361358e-05, + "loss": 0.5461, + "step": 27219 + }, + { + "epoch": 0.81, + "learning_rate": 1.8272197381550714e-05, + "loss": 0.5, + "step": 27220 + }, + { + "epoch": 0.81, + "learning_rate": 1.8266641386612604e-05, + "loss": 0.5675, + "step": 27221 + }, + { + "epoch": 0.81, + "learning_rate": 1.8261086151598662e-05, + "loss": 0.5418, + "step": 27222 + }, + { + "epoch": 0.81, + "learning_rate": 1.8255531676560556e-05, + "loss": 0.6085, + "step": 27223 + }, + { + "epoch": 0.81, + "learning_rate": 1.8249977961549945e-05, + "loss": 0.5504, + "step": 27224 + }, + { + "epoch": 0.81, + "learning_rate": 1.824442500661848e-05, + "loss": 0.647, + "step": 27225 + }, + { + "epoch": 0.81, + "learning_rate": 1.8238872811817765e-05, + "loss": 0.5483, + "step": 27226 + }, + { + "epoch": 0.81, + "learning_rate": 1.82333213771994e-05, + "loss": 0.4491, + "step": 27227 + }, + { + "epoch": 0.81, + "learning_rate": 1.8227770702815017e-05, + "loss": 0.52, + "step": 27228 + }, + { + "epoch": 0.81, + "learning_rate": 1.822222078871627e-05, + "loss": 0.5225, + "step": 27229 + }, + { + "epoch": 0.81, + "learning_rate": 1.8216671634954707e-05, + "loss": 0.4488, + "step": 27230 + }, + { + "epoch": 0.81, + "learning_rate": 1.8211123241581917e-05, + "loss": 0.5641, + "step": 27231 + }, + { + "epoch": 0.81, + "learning_rate": 1.8205575608649506e-05, + "loss": 0.5185, + "step": 27232 + }, + { + "epoch": 0.81, + "learning_rate": 1.8200028736209073e-05, + "loss": 0.5312, + "step": 27233 + }, + { + "epoch": 0.81, + "learning_rate": 1.8194482624312147e-05, + "loss": 0.4482, + "step": 27234 + }, + { + "epoch": 0.81, + "learning_rate": 1.8188937273010353e-05, + "loss": 0.6134, + "step": 27235 + }, + { + "epoch": 0.81, + "learning_rate": 1.818339268235517e-05, + "loss": 0.4688, + "step": 27236 + }, + { + "epoch": 0.81, + "learning_rate": 1.817784885239824e-05, + "loss": 0.5767, + "step": 27237 + }, + { + "epoch": 0.81, + "learning_rate": 1.8172305783191013e-05, + "loss": 0.5442, + "step": 27238 + }, + { + "epoch": 0.81, + "learning_rate": 1.8166763474785087e-05, + "loss": 0.4666, + "step": 27239 + }, + { + "epoch": 0.81, + "learning_rate": 1.8161221927231998e-05, + "loss": 0.5518, + "step": 27240 + }, + { + "epoch": 0.81, + "learning_rate": 1.815568114058326e-05, + "loss": 0.5122, + "step": 27241 + }, + { + "epoch": 0.81, + "learning_rate": 1.815014111489034e-05, + "loss": 0.4147, + "step": 27242 + }, + { + "epoch": 0.81, + "learning_rate": 1.814460185020479e-05, + "loss": 0.5755, + "step": 27243 + }, + { + "epoch": 0.81, + "learning_rate": 1.8139063346578133e-05, + "loss": 0.6679, + "step": 27244 + }, + { + "epoch": 0.81, + "learning_rate": 1.813352560406184e-05, + "loss": 0.5272, + "step": 27245 + }, + { + "epoch": 0.81, + "learning_rate": 1.812798862270737e-05, + "loss": 0.6029, + "step": 27246 + }, + { + "epoch": 0.81, + "learning_rate": 1.8122452402566235e-05, + "loss": 0.4948, + "step": 27247 + }, + { + "epoch": 0.81, + "learning_rate": 1.8116916943689933e-05, + "loss": 0.4787, + "step": 27248 + }, + { + "epoch": 0.81, + "learning_rate": 1.811138224612987e-05, + "loss": 0.3426, + "step": 27249 + }, + { + "epoch": 0.81, + "learning_rate": 1.8105848309937558e-05, + "loss": 0.4724, + "step": 27250 + }, + { + "epoch": 0.81, + "learning_rate": 1.8100315135164413e-05, + "loss": 0.561, + "step": 27251 + }, + { + "epoch": 0.81, + "learning_rate": 1.8094782721861914e-05, + "loss": 0.565, + "step": 27252 + }, + { + "epoch": 0.81, + "learning_rate": 1.808925107008146e-05, + "loss": 0.463, + "step": 27253 + }, + { + "epoch": 0.81, + "learning_rate": 1.808372017987453e-05, + "loss": 0.4839, + "step": 27254 + }, + { + "epoch": 0.81, + "learning_rate": 1.807819005129249e-05, + "loss": 0.5631, + "step": 27255 + }, + { + "epoch": 0.81, + "learning_rate": 1.807266068438681e-05, + "loss": 0.4091, + "step": 27256 + }, + { + "epoch": 0.81, + "learning_rate": 1.806713207920886e-05, + "loss": 0.5581, + "step": 27257 + }, + { + "epoch": 0.81, + "learning_rate": 1.806160423581006e-05, + "loss": 0.5368, + "step": 27258 + }, + { + "epoch": 0.81, + "learning_rate": 1.8056077154241835e-05, + "loss": 0.4966, + "step": 27259 + }, + { + "epoch": 0.81, + "learning_rate": 1.8050550834555545e-05, + "loss": 0.4389, + "step": 27260 + }, + { + "epoch": 0.81, + "learning_rate": 1.804502527680254e-05, + "loss": 0.5479, + "step": 27261 + }, + { + "epoch": 0.81, + "learning_rate": 1.803950048103423e-05, + "loss": 0.6799, + "step": 27262 + }, + { + "epoch": 0.81, + "learning_rate": 1.8033976447302004e-05, + "loss": 0.5458, + "step": 27263 + }, + { + "epoch": 0.81, + "learning_rate": 1.802845317565719e-05, + "loss": 0.4774, + "step": 27264 + }, + { + "epoch": 0.81, + "learning_rate": 1.8022930666151127e-05, + "loss": 0.5473, + "step": 27265 + }, + { + "epoch": 0.81, + "learning_rate": 1.8017408918835178e-05, + "loss": 0.439, + "step": 27266 + }, + { + "epoch": 0.81, + "learning_rate": 1.8011887933760695e-05, + "loss": 0.5348, + "step": 27267 + }, + { + "epoch": 0.81, + "learning_rate": 1.8006367710979023e-05, + "loss": 0.4962, + "step": 27268 + }, + { + "epoch": 0.81, + "learning_rate": 1.8000848250541456e-05, + "loss": 0.5279, + "step": 27269 + }, + { + "epoch": 0.81, + "learning_rate": 1.7995329552499306e-05, + "loss": 0.4887, + "step": 27270 + }, + { + "epoch": 0.81, + "learning_rate": 1.7989811616903896e-05, + "loss": 0.4289, + "step": 27271 + }, + { + "epoch": 0.81, + "learning_rate": 1.798429444380655e-05, + "loss": 0.5567, + "step": 27272 + }, + { + "epoch": 0.81, + "learning_rate": 1.7978778033258547e-05, + "loss": 0.5051, + "step": 27273 + }, + { + "epoch": 0.81, + "learning_rate": 1.7973262385311163e-05, + "loss": 0.5599, + "step": 27274 + }, + { + "epoch": 0.81, + "learning_rate": 1.7967747500015674e-05, + "loss": 0.5103, + "step": 27275 + }, + { + "epoch": 0.81, + "learning_rate": 1.796223337742341e-05, + "loss": 0.5982, + "step": 27276 + }, + { + "epoch": 0.81, + "learning_rate": 1.795672001758558e-05, + "loss": 0.563, + "step": 27277 + }, + { + "epoch": 0.81, + "learning_rate": 1.7951207420553486e-05, + "loss": 0.5744, + "step": 27278 + }, + { + "epoch": 0.81, + "learning_rate": 1.7945695586378364e-05, + "loss": 0.5954, + "step": 27279 + }, + { + "epoch": 0.81, + "learning_rate": 1.7940184515111446e-05, + "loss": 0.5321, + "step": 27280 + }, + { + "epoch": 0.81, + "learning_rate": 1.793467420680398e-05, + "loss": 0.4494, + "step": 27281 + }, + { + "epoch": 0.81, + "learning_rate": 1.79291646615072e-05, + "loss": 0.6349, + "step": 27282 + }, + { + "epoch": 0.81, + "learning_rate": 1.7923655879272393e-05, + "loss": 0.4006, + "step": 27283 + }, + { + "epoch": 0.81, + "learning_rate": 1.7918147860150668e-05, + "loss": 0.5073, + "step": 27284 + }, + { + "epoch": 0.81, + "learning_rate": 1.79126406041933e-05, + "loss": 0.5331, + "step": 27285 + }, + { + "epoch": 0.81, + "learning_rate": 1.7907134111451473e-05, + "loss": 0.6, + "step": 27286 + }, + { + "epoch": 0.81, + "learning_rate": 1.790162838197642e-05, + "loss": 0.4448, + "step": 27287 + }, + { + "epoch": 0.81, + "learning_rate": 1.7896123415819298e-05, + "loss": 0.4413, + "step": 27288 + }, + { + "epoch": 0.81, + "learning_rate": 1.789061921303128e-05, + "loss": 0.58, + "step": 27289 + }, + { + "epoch": 0.81, + "learning_rate": 1.7885115773663564e-05, + "loss": 0.5137, + "step": 27290 + }, + { + "epoch": 0.81, + "learning_rate": 1.7879613097767334e-05, + "loss": 0.5499, + "step": 27291 + }, + { + "epoch": 0.81, + "learning_rate": 1.787411118539373e-05, + "loss": 0.5199, + "step": 27292 + }, + { + "epoch": 0.81, + "learning_rate": 1.786861003659388e-05, + "loss": 0.5892, + "step": 27293 + }, + { + "epoch": 0.81, + "learning_rate": 1.786310965141896e-05, + "loss": 0.4685, + "step": 27294 + }, + { + "epoch": 0.81, + "learning_rate": 1.7857610029920145e-05, + "loss": 0.5815, + "step": 27295 + }, + { + "epoch": 0.81, + "learning_rate": 1.78521111721485e-05, + "loss": 0.5468, + "step": 27296 + }, + { + "epoch": 0.81, + "learning_rate": 1.7846613078155217e-05, + "loss": 0.5057, + "step": 27297 + }, + { + "epoch": 0.81, + "learning_rate": 1.7841115747991345e-05, + "loss": 0.602, + "step": 27298 + }, + { + "epoch": 0.81, + "learning_rate": 1.7835619181708064e-05, + "loss": 0.4419, + "step": 27299 + }, + { + "epoch": 0.81, + "learning_rate": 1.783012337935642e-05, + "loss": 0.5671, + "step": 27300 + }, + { + "epoch": 0.81, + "learning_rate": 1.782462834098755e-05, + "loss": 0.6058, + "step": 27301 + }, + { + "epoch": 0.81, + "learning_rate": 1.7819134066652555e-05, + "loss": 0.4632, + "step": 27302 + }, + { + "epoch": 0.81, + "learning_rate": 1.781364055640249e-05, + "loss": 0.4817, + "step": 27303 + }, + { + "epoch": 0.81, + "learning_rate": 1.7808147810288423e-05, + "loss": 0.6089, + "step": 27304 + }, + { + "epoch": 0.81, + "learning_rate": 1.7802655828361436e-05, + "loss": 0.6166, + "step": 27305 + }, + { + "epoch": 0.81, + "learning_rate": 1.7797164610672613e-05, + "loss": 0.5392, + "step": 27306 + }, + { + "epoch": 0.81, + "learning_rate": 1.7791674157273e-05, + "loss": 0.5228, + "step": 27307 + }, + { + "epoch": 0.81, + "learning_rate": 1.7786184468213608e-05, + "loss": 0.6402, + "step": 27308 + }, + { + "epoch": 0.81, + "learning_rate": 1.77806955435455e-05, + "loss": 0.5738, + "step": 27309 + }, + { + "epoch": 0.81, + "learning_rate": 1.7775207383319747e-05, + "loss": 0.5724, + "step": 27310 + }, + { + "epoch": 0.81, + "learning_rate": 1.776971998758732e-05, + "loss": 0.4239, + "step": 27311 + }, + { + "epoch": 0.81, + "learning_rate": 1.7764233356399272e-05, + "loss": 0.6418, + "step": 27312 + }, + { + "epoch": 0.81, + "learning_rate": 1.7758747489806605e-05, + "loss": 0.4857, + "step": 27313 + }, + { + "epoch": 0.81, + "learning_rate": 1.7753262387860338e-05, + "loss": 0.6529, + "step": 27314 + }, + { + "epoch": 0.81, + "learning_rate": 1.774777805061143e-05, + "loss": 0.6081, + "step": 27315 + }, + { + "epoch": 0.81, + "learning_rate": 1.7742294478110933e-05, + "loss": 0.478, + "step": 27316 + }, + { + "epoch": 0.81, + "learning_rate": 1.773681167040976e-05, + "loss": 0.6001, + "step": 27317 + }, + { + "epoch": 0.81, + "learning_rate": 1.7731329627558955e-05, + "loss": 0.4949, + "step": 27318 + }, + { + "epoch": 0.81, + "learning_rate": 1.772584834960944e-05, + "loss": 0.4879, + "step": 27319 + }, + { + "epoch": 0.81, + "learning_rate": 1.7720367836612196e-05, + "loss": 0.4278, + "step": 27320 + }, + { + "epoch": 0.81, + "learning_rate": 1.77148880886182e-05, + "loss": 0.4554, + "step": 27321 + }, + { + "epoch": 0.81, + "learning_rate": 1.7709409105678377e-05, + "loss": 0.5848, + "step": 27322 + }, + { + "epoch": 0.81, + "learning_rate": 1.7703930887843646e-05, + "loss": 0.479, + "step": 27323 + }, + { + "epoch": 0.81, + "learning_rate": 1.7698453435164964e-05, + "loss": 0.5981, + "step": 27324 + }, + { + "epoch": 0.81, + "learning_rate": 1.7692976747693278e-05, + "loss": 0.5581, + "step": 27325 + }, + { + "epoch": 0.81, + "learning_rate": 1.7687500825479496e-05, + "loss": 0.4834, + "step": 27326 + }, + { + "epoch": 0.81, + "learning_rate": 1.768202566857451e-05, + "loss": 0.5594, + "step": 27327 + }, + { + "epoch": 0.81, + "learning_rate": 1.767655127702923e-05, + "loss": 0.5372, + "step": 27328 + }, + { + "epoch": 0.81, + "learning_rate": 1.7671077650894562e-05, + "loss": 0.6075, + "step": 27329 + }, + { + "epoch": 0.81, + "learning_rate": 1.7665604790221434e-05, + "loss": 0.555, + "step": 27330 + }, + { + "epoch": 0.81, + "learning_rate": 1.7660132695060706e-05, + "loss": 0.4552, + "step": 27331 + }, + { + "epoch": 0.81, + "learning_rate": 1.765466136546321e-05, + "loss": 0.4776, + "step": 27332 + }, + { + "epoch": 0.81, + "learning_rate": 1.7649190801479854e-05, + "loss": 0.4703, + "step": 27333 + }, + { + "epoch": 0.81, + "learning_rate": 1.7643721003161516e-05, + "loss": 0.4713, + "step": 27334 + }, + { + "epoch": 0.81, + "learning_rate": 1.763825197055905e-05, + "loss": 0.5194, + "step": 27335 + }, + { + "epoch": 0.81, + "learning_rate": 1.7632783703723265e-05, + "loss": 0.4746, + "step": 27336 + }, + { + "epoch": 0.81, + "learning_rate": 1.7627316202705024e-05, + "loss": 0.6212, + "step": 27337 + }, + { + "epoch": 0.81, + "learning_rate": 1.7621849467555186e-05, + "loss": 0.6439, + "step": 27338 + }, + { + "epoch": 0.81, + "learning_rate": 1.761638349832454e-05, + "loss": 0.6484, + "step": 27339 + }, + { + "epoch": 0.81, + "learning_rate": 1.761091829506395e-05, + "loss": 0.5307, + "step": 27340 + }, + { + "epoch": 0.81, + "learning_rate": 1.7605453857824185e-05, + "loss": 0.4748, + "step": 27341 + }, + { + "epoch": 0.81, + "learning_rate": 1.759999018665609e-05, + "loss": 0.5002, + "step": 27342 + }, + { + "epoch": 0.81, + "learning_rate": 1.7594527281610418e-05, + "loss": 0.5199, + "step": 27343 + }, + { + "epoch": 0.81, + "learning_rate": 1.7589065142737993e-05, + "loss": 0.5565, + "step": 27344 + }, + { + "epoch": 0.81, + "learning_rate": 1.758360377008964e-05, + "loss": 0.4944, + "step": 27345 + }, + { + "epoch": 0.81, + "learning_rate": 1.7578143163716054e-05, + "loss": 0.6263, + "step": 27346 + }, + { + "epoch": 0.81, + "learning_rate": 1.7572683323668025e-05, + "loss": 0.5587, + "step": 27347 + }, + { + "epoch": 0.81, + "learning_rate": 1.756722424999634e-05, + "loss": 0.5513, + "step": 27348 + }, + { + "epoch": 0.81, + "learning_rate": 1.7561765942751783e-05, + "loss": 0.6158, + "step": 27349 + }, + { + "epoch": 0.81, + "learning_rate": 1.7556308401985054e-05, + "loss": 0.5181, + "step": 27350 + }, + { + "epoch": 0.81, + "learning_rate": 1.75508516277469e-05, + "loss": 0.4066, + "step": 27351 + }, + { + "epoch": 0.81, + "learning_rate": 1.7545395620088056e-05, + "loss": 0.7033, + "step": 27352 + }, + { + "epoch": 0.81, + "learning_rate": 1.7539940379059293e-05, + "loss": 0.5218, + "step": 27353 + }, + { + "epoch": 0.81, + "learning_rate": 1.753448590471127e-05, + "loss": 0.4976, + "step": 27354 + }, + { + "epoch": 0.81, + "learning_rate": 1.7529032197094764e-05, + "loss": 0.5473, + "step": 27355 + }, + { + "epoch": 0.81, + "learning_rate": 1.7523579256260415e-05, + "loss": 0.4977, + "step": 27356 + }, + { + "epoch": 0.81, + "learning_rate": 1.7518127082258973e-05, + "loss": 0.3902, + "step": 27357 + }, + { + "epoch": 0.81, + "learning_rate": 1.751267567514109e-05, + "loss": 0.4819, + "step": 27358 + }, + { + "epoch": 0.81, + "learning_rate": 1.7507225034957496e-05, + "loss": 0.4767, + "step": 27359 + }, + { + "epoch": 0.81, + "learning_rate": 1.7501775161758828e-05, + "loss": 0.491, + "step": 27360 + }, + { + "epoch": 0.81, + "learning_rate": 1.7496326055595792e-05, + "loss": 0.3939, + "step": 27361 + }, + { + "epoch": 0.81, + "learning_rate": 1.7490877716519017e-05, + "loss": 0.579, + "step": 27362 + }, + { + "epoch": 0.81, + "learning_rate": 1.7485430144579162e-05, + "loss": 0.4559, + "step": 27363 + }, + { + "epoch": 0.81, + "learning_rate": 1.747998333982692e-05, + "loss": 0.5918, + "step": 27364 + }, + { + "epoch": 0.81, + "learning_rate": 1.747453730231291e-05, + "loss": 0.5569, + "step": 27365 + }, + { + "epoch": 0.81, + "learning_rate": 1.746909203208773e-05, + "loss": 0.5102, + "step": 27366 + }, + { + "epoch": 0.81, + "learning_rate": 1.7463647529202044e-05, + "loss": 0.5689, + "step": 27367 + }, + { + "epoch": 0.81, + "learning_rate": 1.7458203793706497e-05, + "loss": 0.5445, + "step": 27368 + }, + { + "epoch": 0.81, + "learning_rate": 1.745276082565166e-05, + "loss": 0.439, + "step": 27369 + }, + { + "epoch": 0.81, + "learning_rate": 1.7447318625088137e-05, + "loss": 0.5175, + "step": 27370 + }, + { + "epoch": 0.81, + "learning_rate": 1.7441877192066537e-05, + "loss": 0.5448, + "step": 27371 + }, + { + "epoch": 0.81, + "learning_rate": 1.7436436526637466e-05, + "loss": 0.5436, + "step": 27372 + }, + { + "epoch": 0.81, + "learning_rate": 1.7430996628851516e-05, + "loss": 0.6251, + "step": 27373 + }, + { + "epoch": 0.81, + "learning_rate": 1.742555749875926e-05, + "loss": 0.5356, + "step": 27374 + }, + { + "epoch": 0.81, + "learning_rate": 1.742011913641123e-05, + "loss": 0.4406, + "step": 27375 + }, + { + "epoch": 0.82, + "learning_rate": 1.7414681541858048e-05, + "loss": 0.4097, + "step": 27376 + }, + { + "epoch": 0.82, + "learning_rate": 1.740924471515021e-05, + "loss": 0.56, + "step": 27377 + }, + { + "epoch": 0.82, + "learning_rate": 1.740380865633833e-05, + "loss": 0.4823, + "step": 27378 + }, + { + "epoch": 0.82, + "learning_rate": 1.7398373365472886e-05, + "loss": 0.4561, + "step": 27379 + }, + { + "epoch": 0.82, + "learning_rate": 1.7392938842604477e-05, + "loss": 0.5385, + "step": 27380 + }, + { + "epoch": 0.82, + "learning_rate": 1.7387505087783563e-05, + "loss": 0.5834, + "step": 27381 + }, + { + "epoch": 0.82, + "learning_rate": 1.7382072101060708e-05, + "loss": 0.4517, + "step": 27382 + }, + { + "epoch": 0.82, + "learning_rate": 1.7376639882486445e-05, + "loss": 0.6538, + "step": 27383 + }, + { + "epoch": 0.82, + "learning_rate": 1.7371208432111252e-05, + "loss": 0.5877, + "step": 27384 + }, + { + "epoch": 0.82, + "learning_rate": 1.73657777499856e-05, + "loss": 0.4175, + "step": 27385 + }, + { + "epoch": 0.82, + "learning_rate": 1.7360347836160018e-05, + "loss": 0.5386, + "step": 27386 + }, + { + "epoch": 0.82, + "learning_rate": 1.735491869068501e-05, + "loss": 0.5185, + "step": 27387 + }, + { + "epoch": 0.82, + "learning_rate": 1.7349490313611028e-05, + "loss": 0.555, + "step": 27388 + }, + { + "epoch": 0.82, + "learning_rate": 1.734406270498852e-05, + "loss": 0.6172, + "step": 27389 + }, + { + "epoch": 0.82, + "learning_rate": 1.733863586486797e-05, + "loss": 0.4592, + "step": 27390 + }, + { + "epoch": 0.82, + "learning_rate": 1.7333209793299833e-05, + "loss": 0.4892, + "step": 27391 + }, + { + "epoch": 0.82, + "learning_rate": 1.7327784490334596e-05, + "loss": 0.4961, + "step": 27392 + }, + { + "epoch": 0.82, + "learning_rate": 1.7322359956022672e-05, + "loss": 0.5673, + "step": 27393 + }, + { + "epoch": 0.82, + "learning_rate": 1.731693619041447e-05, + "loss": 0.5535, + "step": 27394 + }, + { + "epoch": 0.82, + "learning_rate": 1.7311513193560426e-05, + "loss": 0.3927, + "step": 27395 + }, + { + "epoch": 0.82, + "learning_rate": 1.7306090965511013e-05, + "loss": 0.499, + "step": 27396 + }, + { + "epoch": 0.82, + "learning_rate": 1.73006695063166e-05, + "loss": 0.5655, + "step": 27397 + }, + { + "epoch": 0.82, + "learning_rate": 1.729524881602759e-05, + "loss": 0.5231, + "step": 27398 + }, + { + "epoch": 0.82, + "learning_rate": 1.728982889469438e-05, + "loss": 0.4188, + "step": 27399 + }, + { + "epoch": 0.82, + "learning_rate": 1.72844097423674e-05, + "loss": 0.6321, + "step": 27400 + }, + { + "epoch": 0.82, + "learning_rate": 1.7278991359096996e-05, + "loss": 0.6157, + "step": 27401 + }, + { + "epoch": 0.82, + "learning_rate": 1.727357374493358e-05, + "loss": 0.5108, + "step": 27402 + }, + { + "epoch": 0.82, + "learning_rate": 1.726815689992748e-05, + "loss": 0.4555, + "step": 27403 + }, + { + "epoch": 0.82, + "learning_rate": 1.7262740824129108e-05, + "loss": 0.4954, + "step": 27404 + }, + { + "epoch": 0.82, + "learning_rate": 1.7257325517588773e-05, + "loss": 0.4178, + "step": 27405 + }, + { + "epoch": 0.82, + "learning_rate": 1.7251910980356855e-05, + "loss": 0.5785, + "step": 27406 + }, + { + "epoch": 0.82, + "learning_rate": 1.7246497212483704e-05, + "loss": 0.5634, + "step": 27407 + }, + { + "epoch": 0.82, + "learning_rate": 1.724108421401963e-05, + "loss": 0.4851, + "step": 27408 + }, + { + "epoch": 0.82, + "learning_rate": 1.7235671985014966e-05, + "loss": 0.5003, + "step": 27409 + }, + { + "epoch": 0.82, + "learning_rate": 1.7230260525520024e-05, + "loss": 0.5338, + "step": 27410 + }, + { + "epoch": 0.82, + "learning_rate": 1.7224849835585154e-05, + "loss": 0.4425, + "step": 27411 + }, + { + "epoch": 0.82, + "learning_rate": 1.7219439915260648e-05, + "loss": 0.4833, + "step": 27412 + }, + { + "epoch": 0.82, + "learning_rate": 1.7214030764596767e-05, + "loss": 0.5739, + "step": 27413 + }, + { + "epoch": 0.82, + "learning_rate": 1.7208622383643834e-05, + "loss": 0.5764, + "step": 27414 + }, + { + "epoch": 0.82, + "learning_rate": 1.720321477245216e-05, + "loss": 0.573, + "step": 27415 + }, + { + "epoch": 0.82, + "learning_rate": 1.7197807931071974e-05, + "loss": 0.5483, + "step": 27416 + }, + { + "epoch": 0.82, + "learning_rate": 1.7192401859553586e-05, + "loss": 0.5515, + "step": 27417 + }, + { + "epoch": 0.82, + "learning_rate": 1.718699655794721e-05, + "loss": 0.5423, + "step": 27418 + }, + { + "epoch": 0.82, + "learning_rate": 1.718159202630317e-05, + "loss": 0.5663, + "step": 27419 + }, + { + "epoch": 0.82, + "learning_rate": 1.7176188264671657e-05, + "loss": 0.534, + "step": 27420 + }, + { + "epoch": 0.82, + "learning_rate": 1.7170785273102952e-05, + "loss": 0.6224, + "step": 27421 + }, + { + "epoch": 0.82, + "learning_rate": 1.7165383051647245e-05, + "loss": 0.5686, + "step": 27422 + }, + { + "epoch": 0.82, + "learning_rate": 1.7159981600354825e-05, + "loss": 0.502, + "step": 27423 + }, + { + "epoch": 0.82, + "learning_rate": 1.7154580919275854e-05, + "loss": 0.5279, + "step": 27424 + }, + { + "epoch": 0.82, + "learning_rate": 1.7149181008460568e-05, + "loss": 0.4885, + "step": 27425 + }, + { + "epoch": 0.82, + "learning_rate": 1.7143781867959196e-05, + "loss": 0.5378, + "step": 27426 + }, + { + "epoch": 0.82, + "learning_rate": 1.713838349782192e-05, + "loss": 0.4775, + "step": 27427 + }, + { + "epoch": 0.82, + "learning_rate": 1.713298589809891e-05, + "loss": 0.5257, + "step": 27428 + }, + { + "epoch": 0.82, + "learning_rate": 1.7127589068840365e-05, + "loss": 0.5866, + "step": 27429 + }, + { + "epoch": 0.82, + "learning_rate": 1.7122193010096498e-05, + "loss": 0.5048, + "step": 27430 + }, + { + "epoch": 0.82, + "learning_rate": 1.7116797721917443e-05, + "loss": 0.3813, + "step": 27431 + }, + { + "epoch": 0.82, + "learning_rate": 1.7111403204353338e-05, + "loss": 0.5936, + "step": 27432 + }, + { + "epoch": 0.82, + "learning_rate": 1.7106009457454377e-05, + "loss": 0.5076, + "step": 27433 + }, + { + "epoch": 0.82, + "learning_rate": 1.7100616481270704e-05, + "loss": 0.5854, + "step": 27434 + }, + { + "epoch": 0.82, + "learning_rate": 1.7095224275852472e-05, + "loss": 0.5395, + "step": 27435 + }, + { + "epoch": 0.82, + "learning_rate": 1.708983284124982e-05, + "loss": 0.5852, + "step": 27436 + }, + { + "epoch": 0.82, + "learning_rate": 1.7084442177512817e-05, + "loss": 0.3997, + "step": 27437 + }, + { + "epoch": 0.82, + "learning_rate": 1.7079052284691632e-05, + "loss": 0.517, + "step": 27438 + }, + { + "epoch": 0.82, + "learning_rate": 1.7073663162836396e-05, + "loss": 0.5523, + "step": 27439 + }, + { + "epoch": 0.82, + "learning_rate": 1.7068274811997177e-05, + "loss": 0.6403, + "step": 27440 + }, + { + "epoch": 0.82, + "learning_rate": 1.7062887232224067e-05, + "loss": 0.4472, + "step": 27441 + }, + { + "epoch": 0.82, + "learning_rate": 1.7057500423567208e-05, + "loss": 0.5027, + "step": 27442 + }, + { + "epoch": 0.82, + "learning_rate": 1.7052114386076622e-05, + "loss": 0.5519, + "step": 27443 + }, + { + "epoch": 0.82, + "learning_rate": 1.704672911980242e-05, + "loss": 0.4833, + "step": 27444 + }, + { + "epoch": 0.82, + "learning_rate": 1.7041344624794686e-05, + "loss": 0.5468, + "step": 27445 + }, + { + "epoch": 0.82, + "learning_rate": 1.7035960901103476e-05, + "loss": 0.4686, + "step": 27446 + }, + { + "epoch": 0.82, + "learning_rate": 1.7030577948778802e-05, + "loss": 0.5323, + "step": 27447 + }, + { + "epoch": 0.82, + "learning_rate": 1.702519576787075e-05, + "loss": 0.5037, + "step": 27448 + }, + { + "epoch": 0.82, + "learning_rate": 1.7019814358429355e-05, + "loss": 0.568, + "step": 27449 + }, + { + "epoch": 0.82, + "learning_rate": 1.7014433720504698e-05, + "loss": 0.6252, + "step": 27450 + }, + { + "epoch": 0.82, + "learning_rate": 1.7009053854146717e-05, + "loss": 0.5128, + "step": 27451 + }, + { + "epoch": 0.82, + "learning_rate": 1.7003674759405475e-05, + "loss": 0.6381, + "step": 27452 + }, + { + "epoch": 0.82, + "learning_rate": 1.6998296436330996e-05, + "loss": 0.4268, + "step": 27453 + }, + { + "epoch": 0.82, + "learning_rate": 1.6992918884973285e-05, + "loss": 0.5861, + "step": 27454 + }, + { + "epoch": 0.82, + "learning_rate": 1.6987542105382338e-05, + "loss": 0.4033, + "step": 27455 + }, + { + "epoch": 0.82, + "learning_rate": 1.698216609760812e-05, + "loss": 0.4733, + "step": 27456 + }, + { + "epoch": 0.82, + "learning_rate": 1.697679086170063e-05, + "loss": 0.5321, + "step": 27457 + }, + { + "epoch": 0.82, + "learning_rate": 1.6971416397709872e-05, + "loss": 0.4725, + "step": 27458 + }, + { + "epoch": 0.82, + "learning_rate": 1.696604270568576e-05, + "loss": 0.5421, + "step": 27459 + }, + { + "epoch": 0.82, + "learning_rate": 1.696066978567833e-05, + "loss": 0.4896, + "step": 27460 + }, + { + "epoch": 0.82, + "learning_rate": 1.6955297637737454e-05, + "loss": 0.5683, + "step": 27461 + }, + { + "epoch": 0.82, + "learning_rate": 1.694992626191316e-05, + "loss": 0.6115, + "step": 27462 + }, + { + "epoch": 0.82, + "learning_rate": 1.6944555658255325e-05, + "loss": 0.6065, + "step": 27463 + }, + { + "epoch": 0.82, + "learning_rate": 1.693918582681393e-05, + "loss": 0.478, + "step": 27464 + }, + { + "epoch": 0.82, + "learning_rate": 1.6933816767638856e-05, + "loss": 0.4549, + "step": 27465 + }, + { + "epoch": 0.82, + "learning_rate": 1.692844848078007e-05, + "loss": 0.5364, + "step": 27466 + }, + { + "epoch": 0.82, + "learning_rate": 1.692308096628743e-05, + "loss": 0.4921, + "step": 27467 + }, + { + "epoch": 0.82, + "learning_rate": 1.691771422421089e-05, + "loss": 0.5569, + "step": 27468 + }, + { + "epoch": 0.82, + "learning_rate": 1.6912348254600342e-05, + "loss": 0.5416, + "step": 27469 + }, + { + "epoch": 0.82, + "learning_rate": 1.6906983057505666e-05, + "loss": 0.6001, + "step": 27470 + }, + { + "epoch": 0.82, + "learning_rate": 1.6901618632976724e-05, + "loss": 0.4948, + "step": 27471 + }, + { + "epoch": 0.82, + "learning_rate": 1.68962549810634e-05, + "loss": 0.628, + "step": 27472 + }, + { + "epoch": 0.82, + "learning_rate": 1.6890892101815613e-05, + "loss": 0.4335, + "step": 27473 + }, + { + "epoch": 0.82, + "learning_rate": 1.6885529995283177e-05, + "loss": 0.5507, + "step": 27474 + }, + { + "epoch": 0.82, + "learning_rate": 1.6880168661515937e-05, + "loss": 0.5578, + "step": 27475 + }, + { + "epoch": 0.82, + "learning_rate": 1.6874808100563765e-05, + "loss": 0.4135, + "step": 27476 + }, + { + "epoch": 0.82, + "learning_rate": 1.6869448312476522e-05, + "loss": 0.5855, + "step": 27477 + }, + { + "epoch": 0.82, + "learning_rate": 1.6864089297303987e-05, + "loss": 0.6522, + "step": 27478 + }, + { + "epoch": 0.82, + "learning_rate": 1.6858731055096034e-05, + "loss": 0.4126, + "step": 27479 + }, + { + "epoch": 0.82, + "learning_rate": 1.6853373585902444e-05, + "loss": 0.4572, + "step": 27480 + }, + { + "epoch": 0.82, + "learning_rate": 1.6848016889773077e-05, + "loss": 0.4766, + "step": 27481 + }, + { + "epoch": 0.82, + "learning_rate": 1.6842660966757674e-05, + "loss": 0.5139, + "step": 27482 + }, + { + "epoch": 0.82, + "learning_rate": 1.6837305816906092e-05, + "loss": 0.441, + "step": 27483 + }, + { + "epoch": 0.82, + "learning_rate": 1.6831951440268078e-05, + "loss": 0.5488, + "step": 27484 + }, + { + "epoch": 0.82, + "learning_rate": 1.6826597836893456e-05, + "loss": 0.6102, + "step": 27485 + }, + { + "epoch": 0.82, + "learning_rate": 1.682124500683194e-05, + "loss": 0.4864, + "step": 27486 + }, + { + "epoch": 0.82, + "learning_rate": 1.681589295013335e-05, + "loss": 0.4147, + "step": 27487 + }, + { + "epoch": 0.82, + "learning_rate": 1.6810541666847457e-05, + "loss": 0.5322, + "step": 27488 + }, + { + "epoch": 0.82, + "learning_rate": 1.680519115702398e-05, + "loss": 0.5739, + "step": 27489 + }, + { + "epoch": 0.82, + "learning_rate": 1.679984142071267e-05, + "loss": 0.4968, + "step": 27490 + }, + { + "epoch": 0.82, + "learning_rate": 1.6794492457963262e-05, + "loss": 0.5343, + "step": 27491 + }, + { + "epoch": 0.82, + "learning_rate": 1.678914426882553e-05, + "loss": 0.5132, + "step": 27492 + }, + { + "epoch": 0.82, + "learning_rate": 1.6783796853349166e-05, + "loss": 0.571, + "step": 27493 + }, + { + "epoch": 0.82, + "learning_rate": 1.6778450211583862e-05, + "loss": 0.6176, + "step": 27494 + }, + { + "epoch": 0.82, + "learning_rate": 1.677310434357937e-05, + "loss": 0.5099, + "step": 27495 + }, + { + "epoch": 0.82, + "learning_rate": 1.6767759249385385e-05, + "loss": 0.5209, + "step": 27496 + }, + { + "epoch": 0.82, + "learning_rate": 1.6762414929051617e-05, + "loss": 0.4766, + "step": 27497 + }, + { + "epoch": 0.82, + "learning_rate": 1.6757071382627744e-05, + "loss": 0.4753, + "step": 27498 + }, + { + "epoch": 0.82, + "learning_rate": 1.675172861016342e-05, + "loss": 0.5417, + "step": 27499 + }, + { + "epoch": 0.82, + "learning_rate": 1.674638661170834e-05, + "loss": 0.6427, + "step": 27500 + }, + { + "epoch": 0.82, + "learning_rate": 1.6741045387312203e-05, + "loss": 0.5593, + "step": 27501 + }, + { + "epoch": 0.82, + "learning_rate": 1.6735704937024642e-05, + "loss": 0.473, + "step": 27502 + }, + { + "epoch": 0.82, + "learning_rate": 1.6730365260895288e-05, + "loss": 0.5247, + "step": 27503 + }, + { + "epoch": 0.82, + "learning_rate": 1.6725026358973804e-05, + "loss": 0.4752, + "step": 27504 + }, + { + "epoch": 0.82, + "learning_rate": 1.671968823130987e-05, + "loss": 0.5093, + "step": 27505 + }, + { + "epoch": 0.82, + "learning_rate": 1.6714350877953045e-05, + "loss": 0.6452, + "step": 27506 + }, + { + "epoch": 0.82, + "learning_rate": 1.670901429895303e-05, + "loss": 0.562, + "step": 27507 + }, + { + "epoch": 0.82, + "learning_rate": 1.6703678494359386e-05, + "loss": 0.5087, + "step": 27508 + }, + { + "epoch": 0.82, + "learning_rate": 1.6698343464221733e-05, + "loss": 0.5232, + "step": 27509 + }, + { + "epoch": 0.82, + "learning_rate": 1.6693009208589672e-05, + "loss": 0.4969, + "step": 27510 + }, + { + "epoch": 0.82, + "learning_rate": 1.6687675727512808e-05, + "loss": 0.4187, + "step": 27511 + }, + { + "epoch": 0.82, + "learning_rate": 1.6682343021040782e-05, + "loss": 0.5447, + "step": 27512 + }, + { + "epoch": 0.82, + "learning_rate": 1.6677011089223073e-05, + "loss": 0.4769, + "step": 27513 + }, + { + "epoch": 0.82, + "learning_rate": 1.6671679932109307e-05, + "loss": 0.5199, + "step": 27514 + }, + { + "epoch": 0.82, + "learning_rate": 1.666634954974905e-05, + "loss": 0.5148, + "step": 27515 + }, + { + "epoch": 0.82, + "learning_rate": 1.6661019942191892e-05, + "loss": 0.4906, + "step": 27516 + }, + { + "epoch": 0.82, + "learning_rate": 1.665569110948735e-05, + "loss": 0.446, + "step": 27517 + }, + { + "epoch": 0.82, + "learning_rate": 1.665036305168495e-05, + "loss": 0.4484, + "step": 27518 + }, + { + "epoch": 0.82, + "learning_rate": 1.664503576883425e-05, + "loss": 0.5113, + "step": 27519 + }, + { + "epoch": 0.82, + "learning_rate": 1.6639709260984804e-05, + "loss": 0.6266, + "step": 27520 + }, + { + "epoch": 0.82, + "learning_rate": 1.6634383528186105e-05, + "loss": 0.5129, + "step": 27521 + }, + { + "epoch": 0.82, + "learning_rate": 1.66290585704877e-05, + "loss": 0.4982, + "step": 27522 + }, + { + "epoch": 0.82, + "learning_rate": 1.662373438793906e-05, + "loss": 0.5286, + "step": 27523 + }, + { + "epoch": 0.82, + "learning_rate": 1.661841098058973e-05, + "loss": 0.5174, + "step": 27524 + }, + { + "epoch": 0.82, + "learning_rate": 1.6613088348489157e-05, + "loss": 0.5104, + "step": 27525 + }, + { + "epoch": 0.82, + "learning_rate": 1.6607766491686882e-05, + "loss": 0.5578, + "step": 27526 + }, + { + "epoch": 0.82, + "learning_rate": 1.6602445410232326e-05, + "loss": 0.5046, + "step": 27527 + }, + { + "epoch": 0.82, + "learning_rate": 1.6597125104175026e-05, + "loss": 0.5324, + "step": 27528 + }, + { + "epoch": 0.82, + "learning_rate": 1.659180557356439e-05, + "loss": 0.5336, + "step": 27529 + }, + { + "epoch": 0.82, + "learning_rate": 1.6586486818449898e-05, + "loss": 0.5155, + "step": 27530 + }, + { + "epoch": 0.82, + "learning_rate": 1.6581168838881023e-05, + "loss": 0.5697, + "step": 27531 + }, + { + "epoch": 0.82, + "learning_rate": 1.65758516349072e-05, + "loss": 0.5944, + "step": 27532 + }, + { + "epoch": 0.82, + "learning_rate": 1.6570535206577843e-05, + "loss": 0.509, + "step": 27533 + }, + { + "epoch": 0.82, + "learning_rate": 1.656521955394239e-05, + "loss": 0.4896, + "step": 27534 + }, + { + "epoch": 0.82, + "learning_rate": 1.65599046770503e-05, + "loss": 0.603, + "step": 27535 + }, + { + "epoch": 0.82, + "learning_rate": 1.6554590575950957e-05, + "loss": 0.5274, + "step": 27536 + }, + { + "epoch": 0.82, + "learning_rate": 1.654927725069375e-05, + "loss": 0.432, + "step": 27537 + }, + { + "epoch": 0.82, + "learning_rate": 1.6543964701328096e-05, + "loss": 0.5216, + "step": 27538 + }, + { + "epoch": 0.82, + "learning_rate": 1.6538652927903408e-05, + "loss": 0.581, + "step": 27539 + }, + { + "epoch": 0.82, + "learning_rate": 1.653334193046907e-05, + "loss": 0.6237, + "step": 27540 + }, + { + "epoch": 0.82, + "learning_rate": 1.6528031709074465e-05, + "loss": 0.5499, + "step": 27541 + }, + { + "epoch": 0.82, + "learning_rate": 1.652272226376893e-05, + "loss": 0.4858, + "step": 27542 + }, + { + "epoch": 0.82, + "learning_rate": 1.651741359460187e-05, + "loss": 0.5399, + "step": 27543 + }, + { + "epoch": 0.82, + "learning_rate": 1.6512105701622606e-05, + "loss": 0.4829, + "step": 27544 + }, + { + "epoch": 0.82, + "learning_rate": 1.6506798584880523e-05, + "loss": 0.4419, + "step": 27545 + }, + { + "epoch": 0.82, + "learning_rate": 1.6501492244424932e-05, + "loss": 0.4874, + "step": 27546 + }, + { + "epoch": 0.82, + "learning_rate": 1.6496186680305203e-05, + "loss": 0.5673, + "step": 27547 + }, + { + "epoch": 0.82, + "learning_rate": 1.649088189257064e-05, + "loss": 0.5756, + "step": 27548 + }, + { + "epoch": 0.82, + "learning_rate": 1.6485577881270565e-05, + "loss": 0.5805, + "step": 27549 + }, + { + "epoch": 0.82, + "learning_rate": 1.648027464645433e-05, + "loss": 0.5562, + "step": 27550 + }, + { + "epoch": 0.82, + "learning_rate": 1.64749721881712e-05, + "loss": 0.5861, + "step": 27551 + }, + { + "epoch": 0.82, + "learning_rate": 1.646967050647048e-05, + "loss": 0.5385, + "step": 27552 + }, + { + "epoch": 0.82, + "learning_rate": 1.6464369601401463e-05, + "loss": 0.4951, + "step": 27553 + }, + { + "epoch": 0.82, + "learning_rate": 1.645906947301348e-05, + "loss": 0.5488, + "step": 27554 + }, + { + "epoch": 0.82, + "learning_rate": 1.6453770121355762e-05, + "loss": 0.6122, + "step": 27555 + }, + { + "epoch": 0.82, + "learning_rate": 1.644847154647756e-05, + "loss": 0.5501, + "step": 27556 + }, + { + "epoch": 0.82, + "learning_rate": 1.6443173748428187e-05, + "loss": 0.513, + "step": 27557 + }, + { + "epoch": 0.82, + "learning_rate": 1.6437876727256873e-05, + "loss": 0.5122, + "step": 27558 + }, + { + "epoch": 0.82, + "learning_rate": 1.6432580483012892e-05, + "loss": 0.6233, + "step": 27559 + }, + { + "epoch": 0.82, + "learning_rate": 1.6427285015745485e-05, + "loss": 0.5101, + "step": 27560 + }, + { + "epoch": 0.82, + "learning_rate": 1.6421990325503844e-05, + "loss": 0.5211, + "step": 27561 + }, + { + "epoch": 0.82, + "learning_rate": 1.641669641233723e-05, + "loss": 0.5445, + "step": 27562 + }, + { + "epoch": 0.82, + "learning_rate": 1.6411403276294878e-05, + "loss": 0.5093, + "step": 27563 + }, + { + "epoch": 0.82, + "learning_rate": 1.6406110917425975e-05, + "loss": 0.545, + "step": 27564 + }, + { + "epoch": 0.82, + "learning_rate": 1.6400819335779726e-05, + "loss": 0.6083, + "step": 27565 + }, + { + "epoch": 0.82, + "learning_rate": 1.6395528531405336e-05, + "loss": 0.6334, + "step": 27566 + }, + { + "epoch": 0.82, + "learning_rate": 1.6390238504352028e-05, + "loss": 0.5688, + "step": 27567 + }, + { + "epoch": 0.82, + "learning_rate": 1.6384949254668923e-05, + "loss": 0.5038, + "step": 27568 + }, + { + "epoch": 0.82, + "learning_rate": 1.637966078240527e-05, + "loss": 0.4465, + "step": 27569 + }, + { + "epoch": 0.82, + "learning_rate": 1.6374373087610173e-05, + "loss": 0.5242, + "step": 27570 + }, + { + "epoch": 0.82, + "learning_rate": 1.6369086170332848e-05, + "loss": 0.4454, + "step": 27571 + }, + { + "epoch": 0.82, + "learning_rate": 1.636380003062241e-05, + "loss": 0.5342, + "step": 27572 + }, + { + "epoch": 0.82, + "learning_rate": 1.6358514668528012e-05, + "loss": 0.4312, + "step": 27573 + }, + { + "epoch": 0.82, + "learning_rate": 1.635323008409887e-05, + "loss": 0.5071, + "step": 27574 + }, + { + "epoch": 0.82, + "learning_rate": 1.6347946277384007e-05, + "loss": 0.4568, + "step": 27575 + }, + { + "epoch": 0.82, + "learning_rate": 1.63426632484326e-05, + "loss": 0.3892, + "step": 27576 + }, + { + "epoch": 0.82, + "learning_rate": 1.6337380997293762e-05, + "loss": 0.606, + "step": 27577 + }, + { + "epoch": 0.82, + "learning_rate": 1.633209952401663e-05, + "loss": 0.468, + "step": 27578 + }, + { + "epoch": 0.82, + "learning_rate": 1.6326818828650303e-05, + "loss": 0.5451, + "step": 27579 + }, + { + "epoch": 0.82, + "learning_rate": 1.632153891124384e-05, + "loss": 0.6213, + "step": 27580 + }, + { + "epoch": 0.82, + "learning_rate": 1.631625977184634e-05, + "loss": 0.5271, + "step": 27581 + }, + { + "epoch": 0.82, + "learning_rate": 1.6310981410506944e-05, + "loss": 0.628, + "step": 27582 + }, + { + "epoch": 0.82, + "learning_rate": 1.6305703827274664e-05, + "loss": 0.5658, + "step": 27583 + }, + { + "epoch": 0.82, + "learning_rate": 1.6300427022198607e-05, + "loss": 0.4892, + "step": 27584 + }, + { + "epoch": 0.82, + "learning_rate": 1.6295150995327802e-05, + "loss": 0.5043, + "step": 27585 + }, + { + "epoch": 0.82, + "learning_rate": 1.6289875746711347e-05, + "loss": 0.5438, + "step": 27586 + }, + { + "epoch": 0.82, + "learning_rate": 1.6284601276398237e-05, + "loss": 0.4699, + "step": 27587 + }, + { + "epoch": 0.82, + "learning_rate": 1.6279327584437564e-05, + "loss": 0.5321, + "step": 27588 + }, + { + "epoch": 0.82, + "learning_rate": 1.6274054670878312e-05, + "loss": 0.3988, + "step": 27589 + }, + { + "epoch": 0.82, + "learning_rate": 1.6268782535769556e-05, + "loss": 0.4136, + "step": 27590 + }, + { + "epoch": 0.82, + "learning_rate": 1.6263511179160263e-05, + "loss": 0.5727, + "step": 27591 + }, + { + "epoch": 0.82, + "learning_rate": 1.625824060109946e-05, + "loss": 0.5167, + "step": 27592 + }, + { + "epoch": 0.82, + "learning_rate": 1.6252970801636192e-05, + "loss": 0.5135, + "step": 27593 + }, + { + "epoch": 0.82, + "learning_rate": 1.6247701780819423e-05, + "loss": 0.5689, + "step": 27594 + }, + { + "epoch": 0.82, + "learning_rate": 1.624243353869812e-05, + "loss": 0.5876, + "step": 27595 + }, + { + "epoch": 0.82, + "learning_rate": 1.6237166075321285e-05, + "loss": 0.5607, + "step": 27596 + }, + { + "epoch": 0.82, + "learning_rate": 1.6231899390737915e-05, + "loss": 0.5112, + "step": 27597 + }, + { + "epoch": 0.82, + "learning_rate": 1.6226633484996966e-05, + "loss": 0.5528, + "step": 27598 + }, + { + "epoch": 0.82, + "learning_rate": 1.622136835814736e-05, + "loss": 0.5839, + "step": 27599 + }, + { + "epoch": 0.82, + "learning_rate": 1.6216104010238077e-05, + "loss": 0.5759, + "step": 27600 + }, + { + "epoch": 0.82, + "learning_rate": 1.6210840441318077e-05, + "loss": 0.58, + "step": 27601 + }, + { + "epoch": 0.82, + "learning_rate": 1.62055776514363e-05, + "loss": 0.4537, + "step": 27602 + }, + { + "epoch": 0.82, + "learning_rate": 1.6200315640641673e-05, + "loss": 0.4709, + "step": 27603 + }, + { + "epoch": 0.82, + "learning_rate": 1.619505440898309e-05, + "loss": 0.5471, + "step": 27604 + }, + { + "epoch": 0.82, + "learning_rate": 1.6189793956509482e-05, + "loss": 0.4787, + "step": 27605 + }, + { + "epoch": 0.82, + "learning_rate": 1.6184534283269793e-05, + "loss": 0.5499, + "step": 27606 + }, + { + "epoch": 0.82, + "learning_rate": 1.6179275389312898e-05, + "loss": 0.4796, + "step": 27607 + }, + { + "epoch": 0.82, + "learning_rate": 1.617401727468767e-05, + "loss": 0.5317, + "step": 27608 + }, + { + "epoch": 0.82, + "learning_rate": 1.6168759939443047e-05, + "loss": 0.4373, + "step": 27609 + }, + { + "epoch": 0.82, + "learning_rate": 1.6163503383627853e-05, + "loss": 0.5424, + "step": 27610 + }, + { + "epoch": 0.82, + "learning_rate": 1.6158247607290987e-05, + "loss": 0.4573, + "step": 27611 + }, + { + "epoch": 0.82, + "learning_rate": 1.6152992610481356e-05, + "loss": 0.4806, + "step": 27612 + }, + { + "epoch": 0.82, + "learning_rate": 1.614773839324777e-05, + "loss": 0.5609, + "step": 27613 + }, + { + "epoch": 0.82, + "learning_rate": 1.614248495563907e-05, + "loss": 0.5259, + "step": 27614 + }, + { + "epoch": 0.82, + "learning_rate": 1.6137232297704118e-05, + "loss": 0.5372, + "step": 27615 + }, + { + "epoch": 0.82, + "learning_rate": 1.6131980419491765e-05, + "loss": 0.4835, + "step": 27616 + }, + { + "epoch": 0.82, + "learning_rate": 1.6126729321050883e-05, + "loss": 0.5799, + "step": 27617 + }, + { + "epoch": 0.82, + "learning_rate": 1.6121479002430186e-05, + "loss": 0.5627, + "step": 27618 + }, + { + "epoch": 0.82, + "learning_rate": 1.6116229463678543e-05, + "loss": 0.5746, + "step": 27619 + }, + { + "epoch": 0.82, + "learning_rate": 1.611098070484478e-05, + "loss": 0.6149, + "step": 27620 + }, + { + "epoch": 0.82, + "learning_rate": 1.6105732725977697e-05, + "loss": 0.5485, + "step": 27621 + }, + { + "epoch": 0.82, + "learning_rate": 1.6100485527126074e-05, + "loss": 0.6411, + "step": 27622 + }, + { + "epoch": 0.82, + "learning_rate": 1.6095239108338676e-05, + "loss": 0.4421, + "step": 27623 + }, + { + "epoch": 0.82, + "learning_rate": 1.608999346966431e-05, + "loss": 0.4285, + "step": 27624 + }, + { + "epoch": 0.82, + "learning_rate": 1.6084748611151758e-05, + "loss": 0.5256, + "step": 27625 + }, + { + "epoch": 0.82, + "learning_rate": 1.6079504532849743e-05, + "loss": 0.4521, + "step": 27626 + }, + { + "epoch": 0.82, + "learning_rate": 1.6074261234807077e-05, + "loss": 0.4715, + "step": 27627 + }, + { + "epoch": 0.82, + "learning_rate": 1.606901871707245e-05, + "loss": 0.562, + "step": 27628 + }, + { + "epoch": 0.82, + "learning_rate": 1.6063776979694666e-05, + "loss": 0.5455, + "step": 27629 + }, + { + "epoch": 0.82, + "learning_rate": 1.6058536022722405e-05, + "loss": 0.5555, + "step": 27630 + }, + { + "epoch": 0.82, + "learning_rate": 1.605329584620444e-05, + "loss": 0.6127, + "step": 27631 + }, + { + "epoch": 0.82, + "learning_rate": 1.6048056450189452e-05, + "loss": 0.5255, + "step": 27632 + }, + { + "epoch": 0.82, + "learning_rate": 1.6042817834726197e-05, + "loss": 0.587, + "step": 27633 + }, + { + "epoch": 0.82, + "learning_rate": 1.603757999986334e-05, + "loss": 0.5958, + "step": 27634 + }, + { + "epoch": 0.82, + "learning_rate": 1.6032342945649592e-05, + "loss": 0.6643, + "step": 27635 + }, + { + "epoch": 0.82, + "learning_rate": 1.6027106672133685e-05, + "loss": 0.4932, + "step": 27636 + }, + { + "epoch": 0.82, + "learning_rate": 1.602187117936427e-05, + "loss": 0.5215, + "step": 27637 + }, + { + "epoch": 0.82, + "learning_rate": 1.601663646738999e-05, + "loss": 0.588, + "step": 27638 + }, + { + "epoch": 0.82, + "learning_rate": 1.6011402536259567e-05, + "loss": 0.4801, + "step": 27639 + }, + { + "epoch": 0.82, + "learning_rate": 1.6006169386021664e-05, + "loss": 0.5386, + "step": 27640 + }, + { + "epoch": 0.82, + "learning_rate": 1.6000937016724926e-05, + "loss": 0.5314, + "step": 27641 + }, + { + "epoch": 0.82, + "learning_rate": 1.5995705428417972e-05, + "loss": 0.5083, + "step": 27642 + }, + { + "epoch": 0.82, + "learning_rate": 1.5990474621149475e-05, + "loss": 0.4813, + "step": 27643 + }, + { + "epoch": 0.82, + "learning_rate": 1.5985244594968073e-05, + "loss": 0.629, + "step": 27644 + }, + { + "epoch": 0.82, + "learning_rate": 1.5980015349922372e-05, + "loss": 0.5829, + "step": 27645 + }, + { + "epoch": 0.82, + "learning_rate": 1.597478688606101e-05, + "loss": 0.5332, + "step": 27646 + }, + { + "epoch": 0.82, + "learning_rate": 1.5969559203432582e-05, + "loss": 0.5457, + "step": 27647 + }, + { + "epoch": 0.82, + "learning_rate": 1.5964332302085715e-05, + "loss": 0.6268, + "step": 27648 + }, + { + "epoch": 0.82, + "learning_rate": 1.5959106182068974e-05, + "loss": 0.4455, + "step": 27649 + }, + { + "epoch": 0.82, + "learning_rate": 1.595388084343099e-05, + "loss": 0.5021, + "step": 27650 + }, + { + "epoch": 0.82, + "learning_rate": 1.59486562862203e-05, + "loss": 0.5379, + "step": 27651 + }, + { + "epoch": 0.82, + "learning_rate": 1.5943432510485522e-05, + "loss": 0.5189, + "step": 27652 + }, + { + "epoch": 0.82, + "learning_rate": 1.5938209516275195e-05, + "loss": 0.5818, + "step": 27653 + }, + { + "epoch": 0.82, + "learning_rate": 1.5932987303637882e-05, + "loss": 0.5664, + "step": 27654 + }, + { + "epoch": 0.82, + "learning_rate": 1.592776587262218e-05, + "loss": 0.487, + "step": 27655 + }, + { + "epoch": 0.82, + "learning_rate": 1.5922545223276596e-05, + "loss": 0.5438, + "step": 27656 + }, + { + "epoch": 0.82, + "learning_rate": 1.591732535564966e-05, + "loss": 0.4915, + "step": 27657 + }, + { + "epoch": 0.82, + "learning_rate": 1.5912106269789907e-05, + "loss": 0.5264, + "step": 27658 + }, + { + "epoch": 0.82, + "learning_rate": 1.590688796574591e-05, + "loss": 0.486, + "step": 27659 + }, + { + "epoch": 0.82, + "learning_rate": 1.5901670443566152e-05, + "loss": 0.3779, + "step": 27660 + }, + { + "epoch": 0.82, + "learning_rate": 1.5896453703299117e-05, + "loss": 0.4606, + "step": 27661 + }, + { + "epoch": 0.82, + "learning_rate": 1.5891237744993337e-05, + "loss": 0.5377, + "step": 27662 + }, + { + "epoch": 0.82, + "learning_rate": 1.588602256869731e-05, + "loss": 0.4927, + "step": 27663 + }, + { + "epoch": 0.82, + "learning_rate": 1.5880808174459538e-05, + "loss": 0.4979, + "step": 27664 + }, + { + "epoch": 0.82, + "learning_rate": 1.587559456232849e-05, + "loss": 0.5392, + "step": 27665 + }, + { + "epoch": 0.82, + "learning_rate": 1.5870381732352612e-05, + "loss": 0.5109, + "step": 27666 + }, + { + "epoch": 0.82, + "learning_rate": 1.5865169684580383e-05, + "loss": 0.454, + "step": 27667 + }, + { + "epoch": 0.82, + "learning_rate": 1.5859958419060307e-05, + "loss": 0.5273, + "step": 27668 + }, + { + "epoch": 0.82, + "learning_rate": 1.585474793584081e-05, + "loss": 0.6492, + "step": 27669 + }, + { + "epoch": 0.82, + "learning_rate": 1.5849538234970296e-05, + "loss": 0.5531, + "step": 27670 + }, + { + "epoch": 0.82, + "learning_rate": 1.584432931649723e-05, + "loss": 0.5892, + "step": 27671 + }, + { + "epoch": 0.82, + "learning_rate": 1.5839121180470086e-05, + "loss": 0.5868, + "step": 27672 + }, + { + "epoch": 0.82, + "learning_rate": 1.583391382693722e-05, + "loss": 0.4607, + "step": 27673 + }, + { + "epoch": 0.82, + "learning_rate": 1.5828707255947094e-05, + "loss": 0.5527, + "step": 27674 + }, + { + "epoch": 0.82, + "learning_rate": 1.5823501467548117e-05, + "loss": 0.5798, + "step": 27675 + }, + { + "epoch": 0.82, + "learning_rate": 1.581829646178864e-05, + "loss": 0.4711, + "step": 27676 + }, + { + "epoch": 0.82, + "learning_rate": 1.5813092238717087e-05, + "loss": 0.4446, + "step": 27677 + }, + { + "epoch": 0.82, + "learning_rate": 1.5807888798381854e-05, + "loss": 0.5587, + "step": 27678 + }, + { + "epoch": 0.82, + "learning_rate": 1.580268614083137e-05, + "loss": 0.5026, + "step": 27679 + }, + { + "epoch": 0.82, + "learning_rate": 1.5797484266113895e-05, + "loss": 0.5167, + "step": 27680 + }, + { + "epoch": 0.82, + "learning_rate": 1.5792283174277857e-05, + "loss": 0.4, + "step": 27681 + }, + { + "epoch": 0.82, + "learning_rate": 1.5787082865371604e-05, + "loss": 0.5159, + "step": 27682 + }, + { + "epoch": 0.82, + "learning_rate": 1.5781883339443516e-05, + "loss": 0.7243, + "step": 27683 + }, + { + "epoch": 0.82, + "learning_rate": 1.5776684596541914e-05, + "loss": 0.5542, + "step": 27684 + }, + { + "epoch": 0.82, + "learning_rate": 1.5771486636715105e-05, + "loss": 0.6173, + "step": 27685 + }, + { + "epoch": 0.82, + "learning_rate": 1.5766289460011442e-05, + "loss": 0.5683, + "step": 27686 + }, + { + "epoch": 0.82, + "learning_rate": 1.576109306647927e-05, + "loss": 0.6042, + "step": 27687 + }, + { + "epoch": 0.82, + "learning_rate": 1.5755897456166857e-05, + "loss": 0.5536, + "step": 27688 + }, + { + "epoch": 0.82, + "learning_rate": 1.5750702629122548e-05, + "loss": 0.4797, + "step": 27689 + }, + { + "epoch": 0.82, + "learning_rate": 1.574550858539462e-05, + "loss": 0.5358, + "step": 27690 + }, + { + "epoch": 0.82, + "learning_rate": 1.5740315325031384e-05, + "loss": 0.4465, + "step": 27691 + }, + { + "epoch": 0.82, + "learning_rate": 1.5735122848081098e-05, + "loss": 0.488, + "step": 27692 + }, + { + "epoch": 0.82, + "learning_rate": 1.5729931154592083e-05, + "loss": 0.5506, + "step": 27693 + }, + { + "epoch": 0.82, + "learning_rate": 1.5724740244612546e-05, + "loss": 0.4477, + "step": 27694 + }, + { + "epoch": 0.82, + "learning_rate": 1.5719550118190818e-05, + "loss": 0.5141, + "step": 27695 + }, + { + "epoch": 0.82, + "learning_rate": 1.5714360775375092e-05, + "loss": 0.5781, + "step": 27696 + }, + { + "epoch": 0.82, + "learning_rate": 1.5709172216213642e-05, + "loss": 0.5328, + "step": 27697 + }, + { + "epoch": 0.82, + "learning_rate": 1.570398444075475e-05, + "loss": 0.4087, + "step": 27698 + }, + { + "epoch": 0.82, + "learning_rate": 1.569879744904661e-05, + "loss": 0.5338, + "step": 27699 + }, + { + "epoch": 0.82, + "learning_rate": 1.5693611241137417e-05, + "loss": 0.5828, + "step": 27700 + }, + { + "epoch": 0.82, + "learning_rate": 1.5688425817075437e-05, + "loss": 0.5478, + "step": 27701 + }, + { + "epoch": 0.82, + "learning_rate": 1.568324117690889e-05, + "loss": 0.4744, + "step": 27702 + }, + { + "epoch": 0.82, + "learning_rate": 1.567805732068597e-05, + "loss": 0.5404, + "step": 27703 + }, + { + "epoch": 0.82, + "learning_rate": 1.5672874248454837e-05, + "loss": 0.5201, + "step": 27704 + }, + { + "epoch": 0.82, + "learning_rate": 1.566769196026371e-05, + "loss": 0.558, + "step": 27705 + }, + { + "epoch": 0.82, + "learning_rate": 1.5662510456160783e-05, + "loss": 0.5118, + "step": 27706 + }, + { + "epoch": 0.82, + "learning_rate": 1.5657329736194216e-05, + "loss": 0.4735, + "step": 27707 + }, + { + "epoch": 0.82, + "learning_rate": 1.5652149800412186e-05, + "loss": 0.5189, + "step": 27708 + }, + { + "epoch": 0.82, + "learning_rate": 1.5646970648862836e-05, + "loss": 0.5332, + "step": 27709 + }, + { + "epoch": 0.82, + "learning_rate": 1.564179228159436e-05, + "loss": 0.5654, + "step": 27710 + }, + { + "epoch": 0.83, + "learning_rate": 1.563661469865485e-05, + "loss": 0.5214, + "step": 27711 + }, + { + "epoch": 0.83, + "learning_rate": 1.5631437900092492e-05, + "loss": 0.5252, + "step": 27712 + }, + { + "epoch": 0.83, + "learning_rate": 1.5626261885955385e-05, + "loss": 0.625, + "step": 27713 + }, + { + "epoch": 0.83, + "learning_rate": 1.5621086656291685e-05, + "loss": 0.5376, + "step": 27714 + }, + { + "epoch": 0.83, + "learning_rate": 1.5615912211149464e-05, + "loss": 0.5193, + "step": 27715 + }, + { + "epoch": 0.83, + "learning_rate": 1.5610738550576864e-05, + "loss": 0.5924, + "step": 27716 + }, + { + "epoch": 0.83, + "learning_rate": 1.5605565674622002e-05, + "loss": 0.5284, + "step": 27717 + }, + { + "epoch": 0.83, + "learning_rate": 1.560039358333296e-05, + "loss": 0.5499, + "step": 27718 + }, + { + "epoch": 0.83, + "learning_rate": 1.5595222276757793e-05, + "loss": 0.5397, + "step": 27719 + }, + { + "epoch": 0.83, + "learning_rate": 1.5590051754944614e-05, + "loss": 0.5438, + "step": 27720 + }, + { + "epoch": 0.83, + "learning_rate": 1.5584882017941504e-05, + "loss": 0.4985, + "step": 27721 + }, + { + "epoch": 0.83, + "learning_rate": 1.5579713065796518e-05, + "loss": 0.4499, + "step": 27722 + }, + { + "epoch": 0.83, + "learning_rate": 1.5574544898557696e-05, + "loss": 0.5929, + "step": 27723 + }, + { + "epoch": 0.83, + "learning_rate": 1.5569377516273108e-05, + "loss": 0.5283, + "step": 27724 + }, + { + "epoch": 0.83, + "learning_rate": 1.5564210918990795e-05, + "loss": 0.4043, + "step": 27725 + }, + { + "epoch": 0.83, + "learning_rate": 1.5559045106758817e-05, + "loss": 0.49, + "step": 27726 + }, + { + "epoch": 0.83, + "learning_rate": 1.5553880079625182e-05, + "loss": 0.52, + "step": 27727 + }, + { + "epoch": 0.83, + "learning_rate": 1.5548715837637895e-05, + "loss": 0.4977, + "step": 27728 + }, + { + "epoch": 0.83, + "learning_rate": 1.5543552380844983e-05, + "loss": 0.462, + "step": 27729 + }, + { + "epoch": 0.83, + "learning_rate": 1.5538389709294487e-05, + "loss": 0.5022, + "step": 27730 + }, + { + "epoch": 0.83, + "learning_rate": 1.553322782303438e-05, + "loss": 0.4948, + "step": 27731 + }, + { + "epoch": 0.83, + "learning_rate": 1.552806672211263e-05, + "loss": 0.5622, + "step": 27732 + }, + { + "epoch": 0.83, + "learning_rate": 1.5522906406577254e-05, + "loss": 0.5405, + "step": 27733 + }, + { + "epoch": 0.83, + "learning_rate": 1.5517746876476248e-05, + "loss": 0.5328, + "step": 27734 + }, + { + "epoch": 0.83, + "learning_rate": 1.551258813185753e-05, + "loss": 0.5225, + "step": 27735 + }, + { + "epoch": 0.83, + "learning_rate": 1.5507430172769123e-05, + "loss": 0.5367, + "step": 27736 + }, + { + "epoch": 0.83, + "learning_rate": 1.550227299925896e-05, + "loss": 0.6219, + "step": 27737 + }, + { + "epoch": 0.83, + "learning_rate": 1.5497116611374952e-05, + "loss": 0.4705, + "step": 27738 + }, + { + "epoch": 0.83, + "learning_rate": 1.549196100916508e-05, + "loss": 0.5559, + "step": 27739 + }, + { + "epoch": 0.83, + "learning_rate": 1.548680619267726e-05, + "loss": 0.5074, + "step": 27740 + }, + { + "epoch": 0.83, + "learning_rate": 1.5481652161959493e-05, + "loss": 0.516, + "step": 27741 + }, + { + "epoch": 0.83, + "learning_rate": 1.5476498917059588e-05, + "loss": 0.4711, + "step": 27742 + }, + { + "epoch": 0.83, + "learning_rate": 1.5471346458025515e-05, + "loss": 0.5455, + "step": 27743 + }, + { + "epoch": 0.83, + "learning_rate": 1.5466194784905162e-05, + "loss": 0.49, + "step": 27744 + }, + { + "epoch": 0.83, + "learning_rate": 1.5461043897746464e-05, + "loss": 0.5093, + "step": 27745 + }, + { + "epoch": 0.83, + "learning_rate": 1.5455893796597288e-05, + "loss": 0.4895, + "step": 27746 + }, + { + "epoch": 0.83, + "learning_rate": 1.5450744481505496e-05, + "loss": 0.5562, + "step": 27747 + }, + { + "epoch": 0.83, + "learning_rate": 1.5445595952518978e-05, + "loss": 0.555, + "step": 27748 + }, + { + "epoch": 0.83, + "learning_rate": 1.5440448209685632e-05, + "loss": 0.4604, + "step": 27749 + }, + { + "epoch": 0.83, + "learning_rate": 1.543530125305328e-05, + "loss": 0.4914, + "step": 27750 + }, + { + "epoch": 0.83, + "learning_rate": 1.5430155082669806e-05, + "loss": 0.5593, + "step": 27751 + }, + { + "epoch": 0.83, + "learning_rate": 1.5425009698583027e-05, + "loss": 0.6191, + "step": 27752 + }, + { + "epoch": 0.83, + "learning_rate": 1.5419865100840823e-05, + "loss": 0.5621, + "step": 27753 + }, + { + "epoch": 0.83, + "learning_rate": 1.5414721289490973e-05, + "loss": 0.5313, + "step": 27754 + }, + { + "epoch": 0.83, + "learning_rate": 1.5409578264581358e-05, + "loss": 0.5848, + "step": 27755 + }, + { + "epoch": 0.83, + "learning_rate": 1.540443602615975e-05, + "loss": 0.5184, + "step": 27756 + }, + { + "epoch": 0.83, + "learning_rate": 1.5399294574274003e-05, + "loss": 0.4535, + "step": 27757 + }, + { + "epoch": 0.83, + "learning_rate": 1.5394153908971876e-05, + "loss": 0.4827, + "step": 27758 + }, + { + "epoch": 0.83, + "learning_rate": 1.5389014030301173e-05, + "loss": 0.4638, + "step": 27759 + }, + { + "epoch": 0.83, + "learning_rate": 1.538387493830973e-05, + "loss": 0.4639, + "step": 27760 + }, + { + "epoch": 0.83, + "learning_rate": 1.5378736633045287e-05, + "loss": 0.4898, + "step": 27761 + }, + { + "epoch": 0.83, + "learning_rate": 1.5373599114555614e-05, + "loss": 0.5219, + "step": 27762 + }, + { + "epoch": 0.83, + "learning_rate": 1.536846238288847e-05, + "loss": 0.4343, + "step": 27763 + }, + { + "epoch": 0.83, + "learning_rate": 1.5363326438091665e-05, + "loss": 0.5022, + "step": 27764 + }, + { + "epoch": 0.83, + "learning_rate": 1.535819128021292e-05, + "loss": 0.4773, + "step": 27765 + }, + { + "epoch": 0.83, + "learning_rate": 1.5353056909299945e-05, + "loss": 0.5474, + "step": 27766 + }, + { + "epoch": 0.83, + "learning_rate": 1.5347923325400516e-05, + "loss": 0.5949, + "step": 27767 + }, + { + "epoch": 0.83, + "learning_rate": 1.5342790528562356e-05, + "loss": 0.5524, + "step": 27768 + }, + { + "epoch": 0.83, + "learning_rate": 1.533765851883321e-05, + "loss": 0.5, + "step": 27769 + }, + { + "epoch": 0.83, + "learning_rate": 1.5332527296260768e-05, + "loss": 0.4712, + "step": 27770 + }, + { + "epoch": 0.83, + "learning_rate": 1.5327396860892728e-05, + "loss": 0.5574, + "step": 27771 + }, + { + "epoch": 0.83, + "learning_rate": 1.532226721277683e-05, + "loss": 0.6882, + "step": 27772 + }, + { + "epoch": 0.83, + "learning_rate": 1.5317138351960713e-05, + "loss": 0.485, + "step": 27773 + }, + { + "epoch": 0.83, + "learning_rate": 1.5312010278492106e-05, + "loss": 0.5163, + "step": 27774 + }, + { + "epoch": 0.83, + "learning_rate": 1.530688299241867e-05, + "loss": 0.5581, + "step": 27775 + }, + { + "epoch": 0.83, + "learning_rate": 1.5301756493788088e-05, + "loss": 0.5757, + "step": 27776 + }, + { + "epoch": 0.83, + "learning_rate": 1.5296630782648e-05, + "loss": 0.6222, + "step": 27777 + }, + { + "epoch": 0.83, + "learning_rate": 1.5291505859046085e-05, + "loss": 0.4557, + "step": 27778 + }, + { + "epoch": 0.83, + "learning_rate": 1.528638172303001e-05, + "loss": 0.4855, + "step": 27779 + }, + { + "epoch": 0.83, + "learning_rate": 1.528125837464738e-05, + "loss": 0.478, + "step": 27780 + }, + { + "epoch": 0.83, + "learning_rate": 1.5276135813945835e-05, + "loss": 0.5335, + "step": 27781 + }, + { + "epoch": 0.83, + "learning_rate": 1.5271014040973008e-05, + "loss": 0.5242, + "step": 27782 + }, + { + "epoch": 0.83, + "learning_rate": 1.5265893055776513e-05, + "loss": 0.5696, + "step": 27783 + }, + { + "epoch": 0.83, + "learning_rate": 1.5260772858404027e-05, + "loss": 0.6243, + "step": 27784 + }, + { + "epoch": 0.83, + "learning_rate": 1.5255653448903062e-05, + "loss": 0.5755, + "step": 27785 + }, + { + "epoch": 0.83, + "learning_rate": 1.525053482732125e-05, + "loss": 0.6104, + "step": 27786 + }, + { + "epoch": 0.83, + "learning_rate": 1.5245416993706186e-05, + "loss": 0.5573, + "step": 27787 + }, + { + "epoch": 0.83, + "learning_rate": 1.5240299948105475e-05, + "loss": 0.5191, + "step": 27788 + }, + { + "epoch": 0.83, + "learning_rate": 1.523518369056668e-05, + "loss": 0.5243, + "step": 27789 + }, + { + "epoch": 0.83, + "learning_rate": 1.5230068221137329e-05, + "loss": 0.5288, + "step": 27790 + }, + { + "epoch": 0.83, + "learning_rate": 1.5224953539865028e-05, + "loss": 0.5182, + "step": 27791 + }, + { + "epoch": 0.83, + "learning_rate": 1.521983964679734e-05, + "loss": 0.7614, + "step": 27792 + }, + { + "epoch": 0.83, + "learning_rate": 1.5214726541981771e-05, + "loss": 0.4839, + "step": 27793 + }, + { + "epoch": 0.83, + "learning_rate": 1.5209614225465907e-05, + "loss": 0.4948, + "step": 27794 + }, + { + "epoch": 0.83, + "learning_rate": 1.520450269729723e-05, + "loss": 0.5362, + "step": 27795 + }, + { + "epoch": 0.83, + "learning_rate": 1.5199391957523324e-05, + "loss": 0.6123, + "step": 27796 + }, + { + "epoch": 0.83, + "learning_rate": 1.519428200619164e-05, + "loss": 0.5167, + "step": 27797 + }, + { + "epoch": 0.83, + "learning_rate": 1.5189172843349742e-05, + "loss": 0.5776, + "step": 27798 + }, + { + "epoch": 0.83, + "learning_rate": 1.518406446904509e-05, + "loss": 0.5396, + "step": 27799 + }, + { + "epoch": 0.83, + "learning_rate": 1.5178956883325235e-05, + "loss": 0.5196, + "step": 27800 + }, + { + "epoch": 0.83, + "learning_rate": 1.5173850086237607e-05, + "loss": 0.5933, + "step": 27801 + }, + { + "epoch": 0.83, + "learning_rate": 1.51687440778297e-05, + "loss": 0.4589, + "step": 27802 + }, + { + "epoch": 0.83, + "learning_rate": 1.5163638858149032e-05, + "loss": 0.4277, + "step": 27803 + }, + { + "epoch": 0.83, + "learning_rate": 1.5158534427243043e-05, + "loss": 0.5809, + "step": 27804 + }, + { + "epoch": 0.83, + "learning_rate": 1.5153430785159162e-05, + "loss": 0.6372, + "step": 27805 + }, + { + "epoch": 0.83, + "learning_rate": 1.5148327931944862e-05, + "loss": 0.5851, + "step": 27806 + }, + { + "epoch": 0.83, + "learning_rate": 1.5143225867647614e-05, + "loss": 0.5912, + "step": 27807 + }, + { + "epoch": 0.83, + "learning_rate": 1.5138124592314828e-05, + "loss": 0.6581, + "step": 27808 + }, + { + "epoch": 0.83, + "learning_rate": 1.513302410599391e-05, + "loss": 0.5396, + "step": 27809 + }, + { + "epoch": 0.83, + "learning_rate": 1.5127924408732318e-05, + "loss": 0.5103, + "step": 27810 + }, + { + "epoch": 0.83, + "learning_rate": 1.5122825500577475e-05, + "loss": 0.5741, + "step": 27811 + }, + { + "epoch": 0.83, + "learning_rate": 1.5117727381576752e-05, + "loss": 0.5451, + "step": 27812 + }, + { + "epoch": 0.83, + "learning_rate": 1.5112630051777598e-05, + "loss": 0.6534, + "step": 27813 + }, + { + "epoch": 0.83, + "learning_rate": 1.5107533511227346e-05, + "loss": 0.5096, + "step": 27814 + }, + { + "epoch": 0.83, + "learning_rate": 1.5102437759973443e-05, + "loss": 0.5869, + "step": 27815 + }, + { + "epoch": 0.83, + "learning_rate": 1.5097342798063208e-05, + "loss": 0.561, + "step": 27816 + }, + { + "epoch": 0.83, + "learning_rate": 1.509224862554406e-05, + "loss": 0.4957, + "step": 27817 + }, + { + "epoch": 0.83, + "learning_rate": 1.5087155242463324e-05, + "loss": 0.5538, + "step": 27818 + }, + { + "epoch": 0.83, + "learning_rate": 1.50820626488684e-05, + "loss": 0.4722, + "step": 27819 + }, + { + "epoch": 0.83, + "learning_rate": 1.5076970844806593e-05, + "loss": 0.4635, + "step": 27820 + }, + { + "epoch": 0.83, + "learning_rate": 1.5071879830325264e-05, + "loss": 0.5316, + "step": 27821 + }, + { + "epoch": 0.83, + "learning_rate": 1.5066789605471766e-05, + "loss": 0.6025, + "step": 27822 + }, + { + "epoch": 0.83, + "learning_rate": 1.5061700170293403e-05, + "loss": 0.5781, + "step": 27823 + }, + { + "epoch": 0.83, + "learning_rate": 1.5056611524837483e-05, + "loss": 0.5134, + "step": 27824 + }, + { + "epoch": 0.83, + "learning_rate": 1.5051523669151335e-05, + "loss": 0.5671, + "step": 27825 + }, + { + "epoch": 0.83, + "learning_rate": 1.5046436603282277e-05, + "loss": 0.4068, + "step": 27826 + }, + { + "epoch": 0.83, + "learning_rate": 1.5041350327277603e-05, + "loss": 0.5769, + "step": 27827 + }, + { + "epoch": 0.83, + "learning_rate": 1.5036264841184556e-05, + "loss": 0.6169, + "step": 27828 + }, + { + "epoch": 0.83, + "learning_rate": 1.5031180145050472e-05, + "loss": 0.5226, + "step": 27829 + }, + { + "epoch": 0.83, + "learning_rate": 1.5026096238922593e-05, + "loss": 0.5828, + "step": 27830 + }, + { + "epoch": 0.83, + "learning_rate": 1.5021013122848237e-05, + "loss": 0.5707, + "step": 27831 + }, + { + "epoch": 0.83, + "learning_rate": 1.5015930796874633e-05, + "loss": 0.506, + "step": 27832 + }, + { + "epoch": 0.83, + "learning_rate": 1.501084926104901e-05, + "loss": 0.4656, + "step": 27833 + }, + { + "epoch": 0.83, + "learning_rate": 1.5005768515418627e-05, + "loss": 0.5214, + "step": 27834 + }, + { + "epoch": 0.83, + "learning_rate": 1.5000688560030762e-05, + "loss": 0.4302, + "step": 27835 + }, + { + "epoch": 0.83, + "learning_rate": 1.4995609394932619e-05, + "loss": 0.5154, + "step": 27836 + }, + { + "epoch": 0.83, + "learning_rate": 1.4990531020171405e-05, + "loss": 0.4291, + "step": 27837 + }, + { + "epoch": 0.83, + "learning_rate": 1.498545343579436e-05, + "loss": 0.5739, + "step": 27838 + }, + { + "epoch": 0.83, + "learning_rate": 1.498037664184867e-05, + "loss": 0.4612, + "step": 27839 + }, + { + "epoch": 0.83, + "learning_rate": 1.4975300638381551e-05, + "loss": 0.519, + "step": 27840 + }, + { + "epoch": 0.83, + "learning_rate": 1.4970225425440232e-05, + "loss": 0.5111, + "step": 27841 + }, + { + "epoch": 0.83, + "learning_rate": 1.4965151003071853e-05, + "loss": 0.552, + "step": 27842 + }, + { + "epoch": 0.83, + "learning_rate": 1.4960077371323589e-05, + "loss": 0.5388, + "step": 27843 + }, + { + "epoch": 0.83, + "learning_rate": 1.4955004530242633e-05, + "loss": 0.5597, + "step": 27844 + }, + { + "epoch": 0.83, + "learning_rate": 1.494993247987615e-05, + "loss": 0.656, + "step": 27845 + }, + { + "epoch": 0.83, + "learning_rate": 1.4944861220271345e-05, + "loss": 0.3995, + "step": 27846 + }, + { + "epoch": 0.83, + "learning_rate": 1.493979075147528e-05, + "loss": 0.5339, + "step": 27847 + }, + { + "epoch": 0.83, + "learning_rate": 1.493472107353513e-05, + "loss": 0.7162, + "step": 27848 + }, + { + "epoch": 0.83, + "learning_rate": 1.4929652186498045e-05, + "loss": 0.5953, + "step": 27849 + }, + { + "epoch": 0.83, + "learning_rate": 1.4924584090411164e-05, + "loss": 0.6264, + "step": 27850 + }, + { + "epoch": 0.83, + "learning_rate": 1.4919516785321597e-05, + "loss": 0.4583, + "step": 27851 + }, + { + "epoch": 0.83, + "learning_rate": 1.4914450271276437e-05, + "loss": 0.5347, + "step": 27852 + }, + { + "epoch": 0.83, + "learning_rate": 1.4909384548322792e-05, + "loss": 0.5221, + "step": 27853 + }, + { + "epoch": 0.83, + "learning_rate": 1.4904319616507812e-05, + "loss": 0.5185, + "step": 27854 + }, + { + "epoch": 0.83, + "learning_rate": 1.4899255475878526e-05, + "loss": 0.5582, + "step": 27855 + }, + { + "epoch": 0.83, + "learning_rate": 1.4894192126482065e-05, + "loss": 0.4749, + "step": 27856 + }, + { + "epoch": 0.83, + "learning_rate": 1.4889129568365456e-05, + "loss": 0.5143, + "step": 27857 + }, + { + "epoch": 0.83, + "learning_rate": 1.488406780157583e-05, + "loss": 0.636, + "step": 27858 + }, + { + "epoch": 0.83, + "learning_rate": 1.4879006826160191e-05, + "loss": 0.6648, + "step": 27859 + }, + { + "epoch": 0.83, + "learning_rate": 1.4873946642165648e-05, + "loss": 0.5362, + "step": 27860 + }, + { + "epoch": 0.83, + "learning_rate": 1.4868887249639195e-05, + "loss": 0.6455, + "step": 27861 + }, + { + "epoch": 0.83, + "learning_rate": 1.486382864862793e-05, + "loss": 0.4966, + "step": 27862 + }, + { + "epoch": 0.83, + "learning_rate": 1.4858770839178826e-05, + "loss": 0.5061, + "step": 27863 + }, + { + "epoch": 0.83, + "learning_rate": 1.4853713821338933e-05, + "loss": 0.636, + "step": 27864 + }, + { + "epoch": 0.83, + "learning_rate": 1.4848657595155291e-05, + "loss": 0.6313, + "step": 27865 + }, + { + "epoch": 0.83, + "learning_rate": 1.4843602160674908e-05, + "loss": 0.4129, + "step": 27866 + }, + { + "epoch": 0.83, + "learning_rate": 1.4838547517944734e-05, + "loss": 0.4373, + "step": 27867 + }, + { + "epoch": 0.83, + "learning_rate": 1.483349366701181e-05, + "loss": 0.4712, + "step": 27868 + }, + { + "epoch": 0.83, + "learning_rate": 1.4828440607923144e-05, + "loss": 0.526, + "step": 27869 + }, + { + "epoch": 0.83, + "learning_rate": 1.4823388340725686e-05, + "loss": 0.453, + "step": 27870 + }, + { + "epoch": 0.83, + "learning_rate": 1.4818336865466399e-05, + "loss": 0.5852, + "step": 27871 + }, + { + "epoch": 0.83, + "learning_rate": 1.4813286182192255e-05, + "loss": 0.5171, + "step": 27872 + }, + { + "epoch": 0.83, + "learning_rate": 1.4808236290950261e-05, + "loss": 0.5248, + "step": 27873 + }, + { + "epoch": 0.83, + "learning_rate": 1.4803187191787294e-05, + "loss": 0.5845, + "step": 27874 + }, + { + "epoch": 0.83, + "learning_rate": 1.4798138884750368e-05, + "loss": 0.4743, + "step": 27875 + }, + { + "epoch": 0.83, + "learning_rate": 1.4793091369886369e-05, + "loss": 0.6069, + "step": 27876 + }, + { + "epoch": 0.83, + "learning_rate": 1.4788044647242261e-05, + "loss": 0.52, + "step": 27877 + }, + { + "epoch": 0.83, + "learning_rate": 1.4782998716864927e-05, + "loss": 0.5528, + "step": 27878 + }, + { + "epoch": 0.83, + "learning_rate": 1.477795357880134e-05, + "loss": 0.5935, + "step": 27879 + }, + { + "epoch": 0.83, + "learning_rate": 1.4772909233098343e-05, + "loss": 0.5181, + "step": 27880 + }, + { + "epoch": 0.83, + "learning_rate": 1.4767865679802894e-05, + "loss": 0.472, + "step": 27881 + }, + { + "epoch": 0.83, + "learning_rate": 1.4762822918961839e-05, + "loss": 0.5956, + "step": 27882 + }, + { + "epoch": 0.83, + "learning_rate": 1.475778095062208e-05, + "loss": 0.6225, + "step": 27883 + }, + { + "epoch": 0.83, + "learning_rate": 1.4752739774830526e-05, + "loss": 0.6849, + "step": 27884 + }, + { + "epoch": 0.83, + "learning_rate": 1.4747699391634018e-05, + "loss": 0.5527, + "step": 27885 + }, + { + "epoch": 0.83, + "learning_rate": 1.4742659801079406e-05, + "loss": 0.5446, + "step": 27886 + }, + { + "epoch": 0.83, + "learning_rate": 1.4737621003213553e-05, + "loss": 0.4871, + "step": 27887 + }, + { + "epoch": 0.83, + "learning_rate": 1.4732582998083344e-05, + "loss": 0.459, + "step": 27888 + }, + { + "epoch": 0.83, + "learning_rate": 1.4727545785735598e-05, + "loss": 0.4207, + "step": 27889 + }, + { + "epoch": 0.83, + "learning_rate": 1.4722509366217119e-05, + "loss": 0.658, + "step": 27890 + }, + { + "epoch": 0.83, + "learning_rate": 1.4717473739574762e-05, + "loss": 0.4247, + "step": 27891 + }, + { + "epoch": 0.83, + "learning_rate": 1.471243890585534e-05, + "loss": 0.4538, + "step": 27892 + }, + { + "epoch": 0.83, + "learning_rate": 1.4707404865105701e-05, + "loss": 0.4525, + "step": 27893 + }, + { + "epoch": 0.83, + "learning_rate": 1.4702371617372612e-05, + "loss": 0.4833, + "step": 27894 + }, + { + "epoch": 0.83, + "learning_rate": 1.469733916270285e-05, + "loss": 0.5381, + "step": 27895 + }, + { + "epoch": 0.83, + "learning_rate": 1.4692307501143232e-05, + "loss": 0.5269, + "step": 27896 + }, + { + "epoch": 0.83, + "learning_rate": 1.4687276632740566e-05, + "loss": 0.5375, + "step": 27897 + }, + { + "epoch": 0.83, + "learning_rate": 1.468224655754159e-05, + "loss": 0.5039, + "step": 27898 + }, + { + "epoch": 0.83, + "learning_rate": 1.4677217275593058e-05, + "loss": 0.4655, + "step": 27899 + }, + { + "epoch": 0.83, + "learning_rate": 1.4672188786941766e-05, + "loss": 0.5078, + "step": 27900 + }, + { + "epoch": 0.83, + "learning_rate": 1.4667161091634463e-05, + "loss": 0.5529, + "step": 27901 + }, + { + "epoch": 0.83, + "learning_rate": 1.4662134189717868e-05, + "loss": 0.5699, + "step": 27902 + }, + { + "epoch": 0.83, + "learning_rate": 1.4657108081238768e-05, + "loss": 0.6061, + "step": 27903 + }, + { + "epoch": 0.83, + "learning_rate": 1.4652082766243857e-05, + "loss": 0.4966, + "step": 27904 + }, + { + "epoch": 0.83, + "learning_rate": 1.4647058244779843e-05, + "loss": 0.4541, + "step": 27905 + }, + { + "epoch": 0.83, + "learning_rate": 1.4642034516893455e-05, + "loss": 0.5931, + "step": 27906 + }, + { + "epoch": 0.83, + "learning_rate": 1.4637011582631422e-05, + "loss": 0.5014, + "step": 27907 + }, + { + "epoch": 0.83, + "learning_rate": 1.463198944204046e-05, + "loss": 0.4921, + "step": 27908 + }, + { + "epoch": 0.83, + "learning_rate": 1.4626968095167204e-05, + "loss": 0.551, + "step": 27909 + }, + { + "epoch": 0.83, + "learning_rate": 1.4621947542058379e-05, + "loss": 0.6255, + "step": 27910 + }, + { + "epoch": 0.83, + "learning_rate": 1.4616927782760647e-05, + "loss": 0.4904, + "step": 27911 + }, + { + "epoch": 0.83, + "learning_rate": 1.4611908817320707e-05, + "loss": 0.544, + "step": 27912 + }, + { + "epoch": 0.83, + "learning_rate": 1.460689064578522e-05, + "loss": 0.4624, + "step": 27913 + }, + { + "epoch": 0.83, + "learning_rate": 1.4601873268200805e-05, + "loss": 0.5284, + "step": 27914 + }, + { + "epoch": 0.83, + "learning_rate": 1.4596856684614124e-05, + "loss": 0.5422, + "step": 27915 + }, + { + "epoch": 0.83, + "learning_rate": 1.4591840895071863e-05, + "loss": 0.4813, + "step": 27916 + }, + { + "epoch": 0.83, + "learning_rate": 1.4586825899620604e-05, + "loss": 0.517, + "step": 27917 + }, + { + "epoch": 0.83, + "learning_rate": 1.4581811698307013e-05, + "loss": 0.5355, + "step": 27918 + }, + { + "epoch": 0.83, + "learning_rate": 1.4576798291177674e-05, + "loss": 0.4791, + "step": 27919 + }, + { + "epoch": 0.83, + "learning_rate": 1.4571785678279238e-05, + "loss": 0.4842, + "step": 27920 + }, + { + "epoch": 0.83, + "learning_rate": 1.4566773859658267e-05, + "loss": 0.5203, + "step": 27921 + }, + { + "epoch": 0.83, + "learning_rate": 1.4561762835361414e-05, + "loss": 0.5843, + "step": 27922 + }, + { + "epoch": 0.83, + "learning_rate": 1.4556752605435208e-05, + "loss": 0.4286, + "step": 27923 + }, + { + "epoch": 0.83, + "learning_rate": 1.4551743169926279e-05, + "loss": 0.5052, + "step": 27924 + }, + { + "epoch": 0.83, + "learning_rate": 1.4546734528881167e-05, + "loss": 0.5501, + "step": 27925 + }, + { + "epoch": 0.83, + "learning_rate": 1.4541726682346469e-05, + "loss": 0.6102, + "step": 27926 + }, + { + "epoch": 0.83, + "learning_rate": 1.4536719630368744e-05, + "loss": 0.4836, + "step": 27927 + }, + { + "epoch": 0.83, + "learning_rate": 1.4531713372994549e-05, + "loss": 0.5686, + "step": 27928 + }, + { + "epoch": 0.83, + "learning_rate": 1.4526707910270388e-05, + "loss": 0.5752, + "step": 27929 + }, + { + "epoch": 0.83, + "learning_rate": 1.4521703242242834e-05, + "loss": 0.4542, + "step": 27930 + }, + { + "epoch": 0.83, + "learning_rate": 1.4516699368958441e-05, + "loss": 0.4807, + "step": 27931 + }, + { + "epoch": 0.83, + "learning_rate": 1.4511696290463706e-05, + "loss": 0.687, + "step": 27932 + }, + { + "epoch": 0.83, + "learning_rate": 1.450669400680511e-05, + "loss": 0.5242, + "step": 27933 + }, + { + "epoch": 0.83, + "learning_rate": 1.450169251802922e-05, + "loss": 0.5163, + "step": 27934 + }, + { + "epoch": 0.83, + "learning_rate": 1.449669182418253e-05, + "loss": 0.5002, + "step": 27935 + }, + { + "epoch": 0.83, + "learning_rate": 1.449169192531149e-05, + "loss": 0.5164, + "step": 27936 + }, + { + "epoch": 0.83, + "learning_rate": 1.4486692821462655e-05, + "loss": 0.5256, + "step": 27937 + }, + { + "epoch": 0.83, + "learning_rate": 1.4481694512682432e-05, + "loss": 0.4704, + "step": 27938 + }, + { + "epoch": 0.83, + "learning_rate": 1.4476696999017359e-05, + "loss": 0.4904, + "step": 27939 + }, + { + "epoch": 0.83, + "learning_rate": 1.4471700280513855e-05, + "loss": 0.5339, + "step": 27940 + }, + { + "epoch": 0.83, + "learning_rate": 1.4466704357218408e-05, + "loss": 0.5191, + "step": 27941 + }, + { + "epoch": 0.83, + "learning_rate": 1.4461709229177434e-05, + "loss": 0.4857, + "step": 27942 + }, + { + "epoch": 0.83, + "learning_rate": 1.4456714896437418e-05, + "loss": 0.4753, + "step": 27943 + }, + { + "epoch": 0.83, + "learning_rate": 1.4451721359044746e-05, + "loss": 0.5753, + "step": 27944 + }, + { + "epoch": 0.83, + "learning_rate": 1.4446728617045879e-05, + "loss": 0.5252, + "step": 27945 + }, + { + "epoch": 0.83, + "learning_rate": 1.4441736670487248e-05, + "loss": 0.5332, + "step": 27946 + }, + { + "epoch": 0.83, + "learning_rate": 1.4436745519415252e-05, + "loss": 0.5293, + "step": 27947 + }, + { + "epoch": 0.83, + "learning_rate": 1.443175516387627e-05, + "loss": 0.4599, + "step": 27948 + }, + { + "epoch": 0.83, + "learning_rate": 1.4426765603916725e-05, + "loss": 0.5504, + "step": 27949 + }, + { + "epoch": 0.83, + "learning_rate": 1.4421776839583002e-05, + "loss": 0.3956, + "step": 27950 + }, + { + "epoch": 0.83, + "learning_rate": 1.4416788870921537e-05, + "loss": 0.467, + "step": 27951 + }, + { + "epoch": 0.83, + "learning_rate": 1.4411801697978622e-05, + "loss": 0.5977, + "step": 27952 + }, + { + "epoch": 0.83, + "learning_rate": 1.4406815320800659e-05, + "loss": 0.4938, + "step": 27953 + }, + { + "epoch": 0.83, + "learning_rate": 1.4401829739434014e-05, + "loss": 0.5019, + "step": 27954 + }, + { + "epoch": 0.83, + "learning_rate": 1.4396844953925059e-05, + "loss": 0.5937, + "step": 27955 + }, + { + "epoch": 0.83, + "learning_rate": 1.4391860964320125e-05, + "loss": 0.5164, + "step": 27956 + }, + { + "epoch": 0.83, + "learning_rate": 1.4386877770665519e-05, + "loss": 0.5573, + "step": 27957 + }, + { + "epoch": 0.83, + "learning_rate": 1.4381895373007604e-05, + "loss": 0.472, + "step": 27958 + }, + { + "epoch": 0.83, + "learning_rate": 1.437691377139273e-05, + "loss": 0.5334, + "step": 27959 + }, + { + "epoch": 0.83, + "learning_rate": 1.4371932965867163e-05, + "loss": 0.5529, + "step": 27960 + }, + { + "epoch": 0.83, + "learning_rate": 1.4366952956477253e-05, + "loss": 0.5408, + "step": 27961 + }, + { + "epoch": 0.83, + "learning_rate": 1.4361973743269263e-05, + "loss": 0.4976, + "step": 27962 + }, + { + "epoch": 0.83, + "learning_rate": 1.4356995326289535e-05, + "loss": 0.4145, + "step": 27963 + }, + { + "epoch": 0.83, + "learning_rate": 1.4352017705584297e-05, + "loss": 0.4738, + "step": 27964 + }, + { + "epoch": 0.83, + "learning_rate": 1.4347040881199891e-05, + "loss": 0.5222, + "step": 27965 + }, + { + "epoch": 0.83, + "learning_rate": 1.4342064853182546e-05, + "loss": 0.4988, + "step": 27966 + }, + { + "epoch": 0.83, + "learning_rate": 1.4337089621578559e-05, + "loss": 0.4751, + "step": 27967 + }, + { + "epoch": 0.83, + "learning_rate": 1.4332115186434148e-05, + "loss": 0.5779, + "step": 27968 + }, + { + "epoch": 0.83, + "learning_rate": 1.4327141547795586e-05, + "loss": 0.5433, + "step": 27969 + }, + { + "epoch": 0.83, + "learning_rate": 1.4322168705709138e-05, + "loss": 0.6905, + "step": 27970 + }, + { + "epoch": 0.83, + "learning_rate": 1.4317196660221021e-05, + "loss": 0.5563, + "step": 27971 + }, + { + "epoch": 0.83, + "learning_rate": 1.4312225411377433e-05, + "loss": 0.5655, + "step": 27972 + }, + { + "epoch": 0.83, + "learning_rate": 1.4307254959224625e-05, + "loss": 0.5348, + "step": 27973 + }, + { + "epoch": 0.83, + "learning_rate": 1.4302285303808827e-05, + "loss": 0.6097, + "step": 27974 + }, + { + "epoch": 0.83, + "learning_rate": 1.4297316445176222e-05, + "loss": 0.4516, + "step": 27975 + }, + { + "epoch": 0.83, + "learning_rate": 1.4292348383372989e-05, + "loss": 0.4893, + "step": 27976 + }, + { + "epoch": 0.83, + "learning_rate": 1.4287381118445343e-05, + "loss": 0.4654, + "step": 27977 + }, + { + "epoch": 0.83, + "learning_rate": 1.4282414650439491e-05, + "loss": 0.5156, + "step": 27978 + }, + { + "epoch": 0.83, + "learning_rate": 1.4277448979401564e-05, + "loss": 0.5373, + "step": 27979 + }, + { + "epoch": 0.83, + "learning_rate": 1.4272484105377758e-05, + "loss": 0.5021, + "step": 27980 + }, + { + "epoch": 0.83, + "learning_rate": 1.4267520028414216e-05, + "loss": 0.5241, + "step": 27981 + }, + { + "epoch": 0.83, + "learning_rate": 1.4262556748557132e-05, + "loss": 0.5073, + "step": 27982 + }, + { + "epoch": 0.83, + "learning_rate": 1.4257594265852592e-05, + "loss": 0.5793, + "step": 27983 + }, + { + "epoch": 0.83, + "learning_rate": 1.4252632580346791e-05, + "loss": 0.5527, + "step": 27984 + }, + { + "epoch": 0.83, + "learning_rate": 1.4247671692085818e-05, + "loss": 0.5409, + "step": 27985 + }, + { + "epoch": 0.83, + "learning_rate": 1.4242711601115832e-05, + "loss": 0.5201, + "step": 27986 + }, + { + "epoch": 0.83, + "learning_rate": 1.423775230748291e-05, + "loss": 0.5829, + "step": 27987 + }, + { + "epoch": 0.83, + "learning_rate": 1.423279381123319e-05, + "loss": 0.5517, + "step": 27988 + }, + { + "epoch": 0.83, + "learning_rate": 1.4227836112412797e-05, + "loss": 0.5155, + "step": 27989 + }, + { + "epoch": 0.83, + "learning_rate": 1.4222879211067796e-05, + "loss": 0.4628, + "step": 27990 + }, + { + "epoch": 0.83, + "learning_rate": 1.4217923107244247e-05, + "loss": 0.4968, + "step": 27991 + }, + { + "epoch": 0.83, + "learning_rate": 1.4212967800988264e-05, + "loss": 0.6188, + "step": 27992 + }, + { + "epoch": 0.83, + "learning_rate": 1.4208013292345945e-05, + "loss": 0.4954, + "step": 27993 + }, + { + "epoch": 0.83, + "learning_rate": 1.4203059581363309e-05, + "loss": 0.4897, + "step": 27994 + }, + { + "epoch": 0.83, + "learning_rate": 1.4198106668086419e-05, + "loss": 0.5931, + "step": 27995 + }, + { + "epoch": 0.83, + "learning_rate": 1.4193154552561339e-05, + "loss": 0.4988, + "step": 27996 + }, + { + "epoch": 0.83, + "learning_rate": 1.4188203234834097e-05, + "loss": 0.5389, + "step": 27997 + }, + { + "epoch": 0.83, + "learning_rate": 1.4183252714950767e-05, + "loss": 0.576, + "step": 27998 + }, + { + "epoch": 0.83, + "learning_rate": 1.4178302992957359e-05, + "loss": 0.5151, + "step": 27999 + }, + { + "epoch": 0.83, + "learning_rate": 1.4173354068899847e-05, + "loss": 0.6393, + "step": 28000 + }, + { + "epoch": 0.83, + "learning_rate": 1.4168405942824315e-05, + "loss": 0.4573, + "step": 28001 + }, + { + "epoch": 0.83, + "learning_rate": 1.4163458614776714e-05, + "loss": 0.5917, + "step": 28002 + }, + { + "epoch": 0.83, + "learning_rate": 1.4158512084803076e-05, + "loss": 0.5633, + "step": 28003 + }, + { + "epoch": 0.83, + "learning_rate": 1.4153566352949366e-05, + "loss": 0.5827, + "step": 28004 + }, + { + "epoch": 0.83, + "learning_rate": 1.4148621419261598e-05, + "loss": 0.4718, + "step": 28005 + }, + { + "epoch": 0.83, + "learning_rate": 1.4143677283785718e-05, + "loss": 0.5354, + "step": 28006 + }, + { + "epoch": 0.83, + "learning_rate": 1.4138733946567695e-05, + "loss": 0.4759, + "step": 28007 + }, + { + "epoch": 0.83, + "learning_rate": 1.4133791407653541e-05, + "loss": 0.5583, + "step": 28008 + }, + { + "epoch": 0.83, + "learning_rate": 1.4128849667089161e-05, + "loss": 0.5563, + "step": 28009 + }, + { + "epoch": 0.83, + "learning_rate": 1.4123908724920487e-05, + "loss": 0.5028, + "step": 28010 + }, + { + "epoch": 0.83, + "learning_rate": 1.4118968581193482e-05, + "loss": 0.5646, + "step": 28011 + }, + { + "epoch": 0.83, + "learning_rate": 1.4114029235954085e-05, + "loss": 0.5475, + "step": 28012 + }, + { + "epoch": 0.83, + "learning_rate": 1.4109090689248261e-05, + "loss": 0.5803, + "step": 28013 + }, + { + "epoch": 0.83, + "learning_rate": 1.410415294112183e-05, + "loss": 0.5409, + "step": 28014 + }, + { + "epoch": 0.83, + "learning_rate": 1.4099215991620751e-05, + "loss": 0.5872, + "step": 28015 + }, + { + "epoch": 0.83, + "learning_rate": 1.4094279840790913e-05, + "loss": 0.7515, + "step": 28016 + }, + { + "epoch": 0.83, + "learning_rate": 1.4089344488678257e-05, + "loss": 0.5183, + "step": 28017 + }, + { + "epoch": 0.83, + "learning_rate": 1.4084409935328636e-05, + "loss": 0.6456, + "step": 28018 + }, + { + "epoch": 0.83, + "learning_rate": 1.407947618078791e-05, + "loss": 0.5851, + "step": 28019 + }, + { + "epoch": 0.83, + "learning_rate": 1.4074543225101955e-05, + "loss": 0.5339, + "step": 28020 + }, + { + "epoch": 0.83, + "learning_rate": 1.406961106831669e-05, + "loss": 0.4487, + "step": 28021 + }, + { + "epoch": 0.83, + "learning_rate": 1.40646797104779e-05, + "loss": 0.5276, + "step": 28022 + }, + { + "epoch": 0.83, + "learning_rate": 1.4059749151631496e-05, + "loss": 0.5305, + "step": 28023 + }, + { + "epoch": 0.83, + "learning_rate": 1.405481939182326e-05, + "loss": 0.4938, + "step": 28024 + }, + { + "epoch": 0.83, + "learning_rate": 1.4049890431099089e-05, + "loss": 0.5531, + "step": 28025 + }, + { + "epoch": 0.83, + "learning_rate": 1.4044962269504758e-05, + "loss": 0.4917, + "step": 28026 + }, + { + "epoch": 0.83, + "learning_rate": 1.404003490708613e-05, + "loss": 0.4674, + "step": 28027 + }, + { + "epoch": 0.83, + "learning_rate": 1.403510834388897e-05, + "loss": 0.5635, + "step": 28028 + }, + { + "epoch": 0.83, + "learning_rate": 1.403018257995914e-05, + "loss": 0.4075, + "step": 28029 + }, + { + "epoch": 0.83, + "learning_rate": 1.402525761534238e-05, + "loss": 0.4022, + "step": 28030 + }, + { + "epoch": 0.83, + "learning_rate": 1.4020333450084521e-05, + "loss": 0.5746, + "step": 28031 + }, + { + "epoch": 0.83, + "learning_rate": 1.401541008423134e-05, + "loss": 0.4695, + "step": 28032 + }, + { + "epoch": 0.83, + "learning_rate": 1.4010487517828619e-05, + "loss": 0.5439, + "step": 28033 + }, + { + "epoch": 0.83, + "learning_rate": 1.4005565750922078e-05, + "loss": 0.4606, + "step": 28034 + }, + { + "epoch": 0.83, + "learning_rate": 1.4000644783557526e-05, + "loss": 0.5101, + "step": 28035 + }, + { + "epoch": 0.83, + "learning_rate": 1.3995724615780714e-05, + "loss": 0.4725, + "step": 28036 + }, + { + "epoch": 0.83, + "learning_rate": 1.3990805247637384e-05, + "loss": 0.575, + "step": 28037 + }, + { + "epoch": 0.83, + "learning_rate": 1.3985886679173243e-05, + "loss": 0.6104, + "step": 28038 + }, + { + "epoch": 0.83, + "learning_rate": 1.3980968910434044e-05, + "loss": 0.5073, + "step": 28039 + }, + { + "epoch": 0.83, + "learning_rate": 1.397605194146553e-05, + "loss": 0.5763, + "step": 28040 + }, + { + "epoch": 0.83, + "learning_rate": 1.3971135772313382e-05, + "loss": 0.5192, + "step": 28041 + }, + { + "epoch": 0.83, + "learning_rate": 1.3966220403023355e-05, + "loss": 0.6248, + "step": 28042 + }, + { + "epoch": 0.83, + "learning_rate": 1.3961305833641092e-05, + "loss": 0.4863, + "step": 28043 + }, + { + "epoch": 0.83, + "learning_rate": 1.3956392064212332e-05, + "loss": 0.5846, + "step": 28044 + }, + { + "epoch": 0.83, + "learning_rate": 1.3951479094782727e-05, + "loss": 0.5414, + "step": 28045 + }, + { + "epoch": 0.83, + "learning_rate": 1.3946566925397998e-05, + "loss": 0.5072, + "step": 28046 + }, + { + "epoch": 0.84, + "learning_rate": 1.3941655556103772e-05, + "loss": 0.5076, + "step": 28047 + }, + { + "epoch": 0.84, + "learning_rate": 1.393674498694576e-05, + "loss": 0.4871, + "step": 28048 + }, + { + "epoch": 0.84, + "learning_rate": 1.3931835217969558e-05, + "loss": 0.6004, + "step": 28049 + }, + { + "epoch": 0.84, + "learning_rate": 1.392692624922085e-05, + "loss": 0.459, + "step": 28050 + }, + { + "epoch": 0.84, + "learning_rate": 1.3922018080745302e-05, + "loss": 0.5456, + "step": 28051 + }, + { + "epoch": 0.84, + "learning_rate": 1.3917110712588522e-05, + "loss": 0.504, + "step": 28052 + }, + { + "epoch": 0.84, + "learning_rate": 1.3912204144796115e-05, + "loss": 0.4907, + "step": 28053 + }, + { + "epoch": 0.84, + "learning_rate": 1.3907298377413725e-05, + "loss": 0.5663, + "step": 28054 + }, + { + "epoch": 0.84, + "learning_rate": 1.3902393410486957e-05, + "loss": 0.5949, + "step": 28055 + }, + { + "epoch": 0.84, + "learning_rate": 1.3897489244061468e-05, + "loss": 0.5433, + "step": 28056 + }, + { + "epoch": 0.84, + "learning_rate": 1.3892585878182762e-05, + "loss": 0.5948, + "step": 28057 + }, + { + "epoch": 0.84, + "learning_rate": 1.3887683312896482e-05, + "loss": 0.5126, + "step": 28058 + }, + { + "epoch": 0.84, + "learning_rate": 1.3882781548248192e-05, + "loss": 0.4961, + "step": 28059 + }, + { + "epoch": 0.84, + "learning_rate": 1.3877880584283498e-05, + "loss": 0.5653, + "step": 28060 + }, + { + "epoch": 0.84, + "learning_rate": 1.3872980421047943e-05, + "loss": 0.5286, + "step": 28061 + }, + { + "epoch": 0.84, + "learning_rate": 1.3868081058587068e-05, + "loss": 0.5192, + "step": 28062 + }, + { + "epoch": 0.84, + "learning_rate": 1.3863182496946458e-05, + "loss": 0.6583, + "step": 28063 + }, + { + "epoch": 0.84, + "learning_rate": 1.3858284736171656e-05, + "loss": 0.6058, + "step": 28064 + }, + { + "epoch": 0.84, + "learning_rate": 1.3853387776308191e-05, + "loss": 0.5181, + "step": 28065 + }, + { + "epoch": 0.84, + "learning_rate": 1.3848491617401572e-05, + "loss": 0.5511, + "step": 28066 + }, + { + "epoch": 0.84, + "learning_rate": 1.3843596259497371e-05, + "loss": 0.5778, + "step": 28067 + }, + { + "epoch": 0.84, + "learning_rate": 1.3838701702641044e-05, + "loss": 0.5653, + "step": 28068 + }, + { + "epoch": 0.84, + "learning_rate": 1.3833807946878118e-05, + "loss": 0.5789, + "step": 28069 + }, + { + "epoch": 0.84, + "learning_rate": 1.3828914992254138e-05, + "loss": 0.5134, + "step": 28070 + }, + { + "epoch": 0.84, + "learning_rate": 1.3824022838814565e-05, + "loss": 0.4887, + "step": 28071 + }, + { + "epoch": 0.84, + "learning_rate": 1.3819131486604852e-05, + "loss": 0.5331, + "step": 28072 + }, + { + "epoch": 0.84, + "learning_rate": 1.3814240935670508e-05, + "loss": 0.5916, + "step": 28073 + }, + { + "epoch": 0.84, + "learning_rate": 1.3809351186056995e-05, + "loss": 0.5094, + "step": 28074 + }, + { + "epoch": 0.84, + "learning_rate": 1.3804462237809834e-05, + "loss": 0.5717, + "step": 28075 + }, + { + "epoch": 0.84, + "learning_rate": 1.3799574090974388e-05, + "loss": 0.6215, + "step": 28076 + }, + { + "epoch": 0.84, + "learning_rate": 1.3794686745596141e-05, + "loss": 0.5443, + "step": 28077 + }, + { + "epoch": 0.84, + "learning_rate": 1.3789800201720537e-05, + "loss": 0.4976, + "step": 28078 + }, + { + "epoch": 0.84, + "learning_rate": 1.378491445939304e-05, + "loss": 0.617, + "step": 28079 + }, + { + "epoch": 0.84, + "learning_rate": 1.3780029518659044e-05, + "loss": 0.4722, + "step": 28080 + }, + { + "epoch": 0.84, + "learning_rate": 1.3775145379563937e-05, + "loss": 0.5899, + "step": 28081 + }, + { + "epoch": 0.84, + "learning_rate": 1.3770262042153171e-05, + "loss": 0.4822, + "step": 28082 + }, + { + "epoch": 0.84, + "learning_rate": 1.3765379506472164e-05, + "loss": 0.5238, + "step": 28083 + }, + { + "epoch": 0.84, + "learning_rate": 1.376049777256626e-05, + "loss": 0.5586, + "step": 28084 + }, + { + "epoch": 0.84, + "learning_rate": 1.3755616840480912e-05, + "loss": 0.4008, + "step": 28085 + }, + { + "epoch": 0.84, + "learning_rate": 1.3750736710261437e-05, + "loss": 0.4187, + "step": 28086 + }, + { + "epoch": 0.84, + "learning_rate": 1.3745857381953254e-05, + "loss": 0.5596, + "step": 28087 + }, + { + "epoch": 0.84, + "learning_rate": 1.3740978855601683e-05, + "loss": 0.5301, + "step": 28088 + }, + { + "epoch": 0.84, + "learning_rate": 1.3736101131252144e-05, + "loss": 0.499, + "step": 28089 + }, + { + "epoch": 0.84, + "learning_rate": 1.3731224208949934e-05, + "loss": 0.5415, + "step": 28090 + }, + { + "epoch": 0.84, + "learning_rate": 1.3726348088740437e-05, + "loss": 0.4714, + "step": 28091 + }, + { + "epoch": 0.84, + "learning_rate": 1.3721472770668942e-05, + "loss": 0.4868, + "step": 28092 + }, + { + "epoch": 0.84, + "learning_rate": 1.371659825478081e-05, + "loss": 0.5725, + "step": 28093 + }, + { + "epoch": 0.84, + "learning_rate": 1.3711724541121384e-05, + "loss": 0.5692, + "step": 28094 + }, + { + "epoch": 0.84, + "learning_rate": 1.370685162973595e-05, + "loss": 0.6632, + "step": 28095 + }, + { + "epoch": 0.84, + "learning_rate": 1.3701979520669794e-05, + "loss": 0.5114, + "step": 28096 + }, + { + "epoch": 0.84, + "learning_rate": 1.3697108213968235e-05, + "loss": 0.5126, + "step": 28097 + }, + { + "epoch": 0.84, + "learning_rate": 1.3692237709676592e-05, + "loss": 0.5523, + "step": 28098 + }, + { + "epoch": 0.84, + "learning_rate": 1.368736800784013e-05, + "loss": 0.4956, + "step": 28099 + }, + { + "epoch": 0.84, + "learning_rate": 1.3682499108504088e-05, + "loss": 0.4848, + "step": 28100 + }, + { + "epoch": 0.84, + "learning_rate": 1.3677631011713777e-05, + "loss": 0.4461, + "step": 28101 + }, + { + "epoch": 0.84, + "learning_rate": 1.3672763717514458e-05, + "loss": 0.4691, + "step": 28102 + }, + { + "epoch": 0.84, + "learning_rate": 1.3667897225951353e-05, + "loss": 0.5485, + "step": 28103 + }, + { + "epoch": 0.84, + "learning_rate": 1.3663031537069759e-05, + "loss": 0.5169, + "step": 28104 + }, + { + "epoch": 0.84, + "learning_rate": 1.3658166650914872e-05, + "loss": 0.423, + "step": 28105 + }, + { + "epoch": 0.84, + "learning_rate": 1.3653302567531945e-05, + "loss": 0.5738, + "step": 28106 + }, + { + "epoch": 0.84, + "learning_rate": 1.3648439286966185e-05, + "loss": 0.5254, + "step": 28107 + }, + { + "epoch": 0.84, + "learning_rate": 1.3643576809262837e-05, + "loss": 0.6386, + "step": 28108 + }, + { + "epoch": 0.84, + "learning_rate": 1.3638715134467083e-05, + "loss": 0.5088, + "step": 28109 + }, + { + "epoch": 0.84, + "learning_rate": 1.3633854262624147e-05, + "loss": 0.5471, + "step": 28110 + }, + { + "epoch": 0.84, + "learning_rate": 1.3628994193779188e-05, + "loss": 0.5466, + "step": 28111 + }, + { + "epoch": 0.84, + "learning_rate": 1.3624134927977428e-05, + "loss": 0.5411, + "step": 28112 + }, + { + "epoch": 0.84, + "learning_rate": 1.3619276465264042e-05, + "loss": 0.5319, + "step": 28113 + }, + { + "epoch": 0.84, + "learning_rate": 1.3614418805684204e-05, + "loss": 0.5405, + "step": 28114 + }, + { + "epoch": 0.84, + "learning_rate": 1.3609561949283045e-05, + "loss": 0.6079, + "step": 28115 + }, + { + "epoch": 0.84, + "learning_rate": 1.3604705896105751e-05, + "loss": 0.4788, + "step": 28116 + }, + { + "epoch": 0.84, + "learning_rate": 1.3599850646197476e-05, + "loss": 0.5441, + "step": 28117 + }, + { + "epoch": 0.84, + "learning_rate": 1.3594996199603382e-05, + "loss": 0.514, + "step": 28118 + }, + { + "epoch": 0.84, + "learning_rate": 1.3590142556368535e-05, + "loss": 0.5913, + "step": 28119 + }, + { + "epoch": 0.84, + "learning_rate": 1.3585289716538108e-05, + "loss": 0.5493, + "step": 28120 + }, + { + "epoch": 0.84, + "learning_rate": 1.358043768015722e-05, + "loss": 0.5053, + "step": 28121 + }, + { + "epoch": 0.84, + "learning_rate": 1.3575586447270994e-05, + "loss": 0.555, + "step": 28122 + }, + { + "epoch": 0.84, + "learning_rate": 1.3570736017924524e-05, + "loss": 0.5826, + "step": 28123 + }, + { + "epoch": 0.84, + "learning_rate": 1.3565886392162875e-05, + "loss": 0.5072, + "step": 28124 + }, + { + "epoch": 0.84, + "learning_rate": 1.3561037570031165e-05, + "loss": 0.4462, + "step": 28125 + }, + { + "epoch": 0.84, + "learning_rate": 1.3556189551574505e-05, + "loss": 0.5024, + "step": 28126 + }, + { + "epoch": 0.84, + "learning_rate": 1.3551342336837913e-05, + "loss": 0.6267, + "step": 28127 + }, + { + "epoch": 0.84, + "learning_rate": 1.354649592586651e-05, + "loss": 0.5837, + "step": 28128 + }, + { + "epoch": 0.84, + "learning_rate": 1.3541650318705312e-05, + "loss": 0.5481, + "step": 28129 + }, + { + "epoch": 0.84, + "learning_rate": 1.3536805515399409e-05, + "loss": 0.5924, + "step": 28130 + }, + { + "epoch": 0.84, + "learning_rate": 1.3531961515993808e-05, + "loss": 0.5443, + "step": 28131 + }, + { + "epoch": 0.84, + "learning_rate": 1.3527118320533583e-05, + "loss": 0.5109, + "step": 28132 + }, + { + "epoch": 0.84, + "learning_rate": 1.3522275929063754e-05, + "loss": 0.5604, + "step": 28133 + }, + { + "epoch": 0.84, + "learning_rate": 1.351743434162931e-05, + "loss": 0.4518, + "step": 28134 + }, + { + "epoch": 0.84, + "learning_rate": 1.3512593558275288e-05, + "loss": 0.4327, + "step": 28135 + }, + { + "epoch": 0.84, + "learning_rate": 1.3507753579046711e-05, + "loss": 0.4815, + "step": 28136 + }, + { + "epoch": 0.84, + "learning_rate": 1.3502914403988587e-05, + "loss": 0.6264, + "step": 28137 + }, + { + "epoch": 0.84, + "learning_rate": 1.349807603314589e-05, + "loss": 0.5997, + "step": 28138 + }, + { + "epoch": 0.84, + "learning_rate": 1.3493238466563585e-05, + "loss": 0.4745, + "step": 28139 + }, + { + "epoch": 0.84, + "learning_rate": 1.3488401704286668e-05, + "loss": 0.5312, + "step": 28140 + }, + { + "epoch": 0.84, + "learning_rate": 1.3483565746360127e-05, + "loss": 0.4968, + "step": 28141 + }, + { + "epoch": 0.84, + "learning_rate": 1.3478730592828926e-05, + "loss": 0.4724, + "step": 28142 + }, + { + "epoch": 0.84, + "learning_rate": 1.3473896243737971e-05, + "loss": 0.5313, + "step": 28143 + }, + { + "epoch": 0.84, + "learning_rate": 1.346906269913224e-05, + "loss": 0.5038, + "step": 28144 + }, + { + "epoch": 0.84, + "learning_rate": 1.3464229959056707e-05, + "loss": 0.6975, + "step": 28145 + }, + { + "epoch": 0.84, + "learning_rate": 1.3459398023556247e-05, + "loss": 0.4866, + "step": 28146 + }, + { + "epoch": 0.84, + "learning_rate": 1.3454566892675835e-05, + "loss": 0.5643, + "step": 28147 + }, + { + "epoch": 0.84, + "learning_rate": 1.3449736566460347e-05, + "loss": 0.5158, + "step": 28148 + }, + { + "epoch": 0.84, + "learning_rate": 1.3444907044954735e-05, + "loss": 0.6828, + "step": 28149 + }, + { + "epoch": 0.84, + "learning_rate": 1.3440078328203854e-05, + "loss": 0.4993, + "step": 28150 + }, + { + "epoch": 0.84, + "learning_rate": 1.3435250416252654e-05, + "loss": 0.464, + "step": 28151 + }, + { + "epoch": 0.84, + "learning_rate": 1.3430423309145967e-05, + "loss": 0.4742, + "step": 28152 + }, + { + "epoch": 0.84, + "learning_rate": 1.3425597006928725e-05, + "loss": 0.54, + "step": 28153 + }, + { + "epoch": 0.84, + "learning_rate": 1.3420771509645746e-05, + "loss": 0.5717, + "step": 28154 + }, + { + "epoch": 0.84, + "learning_rate": 1.341594681734194e-05, + "loss": 0.4617, + "step": 28155 + }, + { + "epoch": 0.84, + "learning_rate": 1.341112293006217e-05, + "loss": 0.5662, + "step": 28156 + }, + { + "epoch": 0.84, + "learning_rate": 1.3406299847851267e-05, + "loss": 0.544, + "step": 28157 + }, + { + "epoch": 0.84, + "learning_rate": 1.340147757075405e-05, + "loss": 0.5965, + "step": 28158 + }, + { + "epoch": 0.84, + "learning_rate": 1.3396656098815375e-05, + "loss": 0.5851, + "step": 28159 + }, + { + "epoch": 0.84, + "learning_rate": 1.3391835432080112e-05, + "loss": 0.6036, + "step": 28160 + }, + { + "epoch": 0.84, + "learning_rate": 1.3387015570593032e-05, + "loss": 0.4954, + "step": 28161 + }, + { + "epoch": 0.84, + "learning_rate": 1.3382196514398948e-05, + "loss": 0.5299, + "step": 28162 + }, + { + "epoch": 0.84, + "learning_rate": 1.3377378263542673e-05, + "loss": 0.5817, + "step": 28163 + }, + { + "epoch": 0.84, + "learning_rate": 1.3372560818069001e-05, + "loss": 0.5338, + "step": 28164 + }, + { + "epoch": 0.84, + "learning_rate": 1.3367744178022767e-05, + "loss": 0.4714, + "step": 28165 + }, + { + "epoch": 0.84, + "learning_rate": 1.3362928343448721e-05, + "loss": 0.4325, + "step": 28166 + }, + { + "epoch": 0.84, + "learning_rate": 1.3358113314391607e-05, + "loss": 0.4795, + "step": 28167 + }, + { + "epoch": 0.84, + "learning_rate": 1.3353299090896254e-05, + "loss": 0.4273, + "step": 28168 + }, + { + "epoch": 0.84, + "learning_rate": 1.334848567300736e-05, + "loss": 0.5516, + "step": 28169 + }, + { + "epoch": 0.84, + "learning_rate": 1.3343673060769735e-05, + "loss": 0.4757, + "step": 28170 + }, + { + "epoch": 0.84, + "learning_rate": 1.3338861254228086e-05, + "loss": 0.5403, + "step": 28171 + }, + { + "epoch": 0.84, + "learning_rate": 1.3334050253427177e-05, + "loss": 0.5178, + "step": 28172 + }, + { + "epoch": 0.84, + "learning_rate": 1.3329240058411708e-05, + "loss": 0.533, + "step": 28173 + }, + { + "epoch": 0.84, + "learning_rate": 1.3324430669226428e-05, + "loss": 0.569, + "step": 28174 + }, + { + "epoch": 0.84, + "learning_rate": 1.3319622085916062e-05, + "loss": 0.4699, + "step": 28175 + }, + { + "epoch": 0.84, + "learning_rate": 1.3314814308525303e-05, + "loss": 0.4567, + "step": 28176 + }, + { + "epoch": 0.84, + "learning_rate": 1.3310007337098829e-05, + "loss": 0.7033, + "step": 28177 + }, + { + "epoch": 0.84, + "learning_rate": 1.3305201171681358e-05, + "loss": 0.5915, + "step": 28178 + }, + { + "epoch": 0.84, + "learning_rate": 1.3300395812317568e-05, + "loss": 0.4151, + "step": 28179 + }, + { + "epoch": 0.84, + "learning_rate": 1.3295591259052198e-05, + "loss": 0.5536, + "step": 28180 + }, + { + "epoch": 0.84, + "learning_rate": 1.3290787511929814e-05, + "loss": 0.6179, + "step": 28181 + }, + { + "epoch": 0.84, + "learning_rate": 1.3285984570995124e-05, + "loss": 0.3779, + "step": 28182 + }, + { + "epoch": 0.84, + "learning_rate": 1.3281182436292793e-05, + "loss": 0.5401, + "step": 28183 + }, + { + "epoch": 0.84, + "learning_rate": 1.3276381107867486e-05, + "loss": 0.5327, + "step": 28184 + }, + { + "epoch": 0.84, + "learning_rate": 1.327158058576382e-05, + "loss": 0.5272, + "step": 28185 + }, + { + "epoch": 0.84, + "learning_rate": 1.3266780870026407e-05, + "loss": 0.4911, + "step": 28186 + }, + { + "epoch": 0.84, + "learning_rate": 1.326198196069991e-05, + "loss": 0.5297, + "step": 28187 + }, + { + "epoch": 0.84, + "learning_rate": 1.3257183857828947e-05, + "loss": 0.6355, + "step": 28188 + }, + { + "epoch": 0.84, + "learning_rate": 1.3252386561458085e-05, + "loss": 0.5254, + "step": 28189 + }, + { + "epoch": 0.84, + "learning_rate": 1.3247590071631998e-05, + "loss": 0.557, + "step": 28190 + }, + { + "epoch": 0.84, + "learning_rate": 1.3242794388395208e-05, + "loss": 0.538, + "step": 28191 + }, + { + "epoch": 0.84, + "learning_rate": 1.3237999511792365e-05, + "loss": 0.5457, + "step": 28192 + }, + { + "epoch": 0.84, + "learning_rate": 1.3233205441868002e-05, + "loss": 0.5223, + "step": 28193 + }, + { + "epoch": 0.84, + "learning_rate": 1.3228412178666738e-05, + "loss": 0.5282, + "step": 28194 + }, + { + "epoch": 0.84, + "learning_rate": 1.3223619722233083e-05, + "loss": 0.4904, + "step": 28195 + }, + { + "epoch": 0.84, + "learning_rate": 1.3218828072611655e-05, + "loss": 0.5816, + "step": 28196 + }, + { + "epoch": 0.84, + "learning_rate": 1.3214037229846954e-05, + "loss": 0.505, + "step": 28197 + }, + { + "epoch": 0.84, + "learning_rate": 1.3209247193983543e-05, + "loss": 0.4758, + "step": 28198 + }, + { + "epoch": 0.84, + "learning_rate": 1.3204457965065975e-05, + "loss": 0.5063, + "step": 28199 + }, + { + "epoch": 0.84, + "learning_rate": 1.3199669543138771e-05, + "loss": 0.6109, + "step": 28200 + }, + { + "epoch": 0.84, + "learning_rate": 1.3194881928246428e-05, + "loss": 0.5011, + "step": 28201 + }, + { + "epoch": 0.84, + "learning_rate": 1.3190095120433466e-05, + "loss": 0.3931, + "step": 28202 + }, + { + "epoch": 0.84, + "learning_rate": 1.3185309119744426e-05, + "loss": 0.5483, + "step": 28203 + }, + { + "epoch": 0.84, + "learning_rate": 1.3180523926223787e-05, + "loss": 0.4647, + "step": 28204 + }, + { + "epoch": 0.84, + "learning_rate": 1.3175739539915999e-05, + "loss": 0.5735, + "step": 28205 + }, + { + "epoch": 0.84, + "learning_rate": 1.3170955960865595e-05, + "loss": 0.5501, + "step": 28206 + }, + { + "epoch": 0.84, + "learning_rate": 1.3166173189117048e-05, + "loss": 0.6553, + "step": 28207 + }, + { + "epoch": 0.84, + "learning_rate": 1.3161391224714793e-05, + "loss": 0.6428, + "step": 28208 + }, + { + "epoch": 0.84, + "learning_rate": 1.3156610067703334e-05, + "loss": 0.4241, + "step": 28209 + }, + { + "epoch": 0.84, + "learning_rate": 1.3151829718127085e-05, + "loss": 0.6906, + "step": 28210 + }, + { + "epoch": 0.84, + "learning_rate": 1.314705017603053e-05, + "loss": 0.4266, + "step": 28211 + }, + { + "epoch": 0.84, + "learning_rate": 1.3142271441458055e-05, + "loss": 0.5318, + "step": 28212 + }, + { + "epoch": 0.84, + "learning_rate": 1.3137493514454146e-05, + "loss": 0.5562, + "step": 28213 + }, + { + "epoch": 0.84, + "learning_rate": 1.3132716395063195e-05, + "loss": 0.5418, + "step": 28214 + }, + { + "epoch": 0.84, + "learning_rate": 1.3127940083329627e-05, + "loss": 0.56, + "step": 28215 + }, + { + "epoch": 0.84, + "learning_rate": 1.3123164579297842e-05, + "loss": 0.5325, + "step": 28216 + }, + { + "epoch": 0.84, + "learning_rate": 1.3118389883012238e-05, + "loss": 0.619, + "step": 28217 + }, + { + "epoch": 0.84, + "learning_rate": 1.3113615994517247e-05, + "loss": 0.5442, + "step": 28218 + }, + { + "epoch": 0.84, + "learning_rate": 1.3108842913857222e-05, + "loss": 0.5183, + "step": 28219 + }, + { + "epoch": 0.84, + "learning_rate": 1.3104070641076515e-05, + "loss": 0.532, + "step": 28220 + }, + { + "epoch": 0.84, + "learning_rate": 1.3099299176219537e-05, + "loss": 0.4802, + "step": 28221 + }, + { + "epoch": 0.84, + "learning_rate": 1.309452851933063e-05, + "loss": 0.4978, + "step": 28222 + }, + { + "epoch": 0.84, + "learning_rate": 1.3089758670454211e-05, + "loss": 0.5195, + "step": 28223 + }, + { + "epoch": 0.84, + "learning_rate": 1.3084989629634526e-05, + "loss": 0.5322, + "step": 28224 + }, + { + "epoch": 0.84, + "learning_rate": 1.3080221396915971e-05, + "loss": 0.4389, + "step": 28225 + }, + { + "epoch": 0.84, + "learning_rate": 1.3075453972342876e-05, + "loss": 0.4741, + "step": 28226 + }, + { + "epoch": 0.84, + "learning_rate": 1.3070687355959588e-05, + "loss": 0.6014, + "step": 28227 + }, + { + "epoch": 0.84, + "learning_rate": 1.30659215478104e-05, + "loss": 0.4923, + "step": 28228 + }, + { + "epoch": 0.84, + "learning_rate": 1.3061156547939602e-05, + "loss": 0.5554, + "step": 28229 + }, + { + "epoch": 0.84, + "learning_rate": 1.3056392356391522e-05, + "loss": 0.4797, + "step": 28230 + }, + { + "epoch": 0.84, + "learning_rate": 1.3051628973210472e-05, + "loss": 0.5816, + "step": 28231 + }, + { + "epoch": 0.84, + "learning_rate": 1.3046866398440726e-05, + "loss": 0.5727, + "step": 28232 + }, + { + "epoch": 0.84, + "learning_rate": 1.3042104632126539e-05, + "loss": 0.5907, + "step": 28233 + }, + { + "epoch": 0.84, + "learning_rate": 1.3037343674312218e-05, + "loss": 0.4695, + "step": 28234 + }, + { + "epoch": 0.84, + "learning_rate": 1.3032583525042008e-05, + "loss": 0.6318, + "step": 28235 + }, + { + "epoch": 0.84, + "learning_rate": 1.302782418436016e-05, + "loss": 0.5207, + "step": 28236 + }, + { + "epoch": 0.84, + "learning_rate": 1.3023065652310972e-05, + "loss": 0.5378, + "step": 28237 + }, + { + "epoch": 0.84, + "learning_rate": 1.3018307928938644e-05, + "loss": 0.5712, + "step": 28238 + }, + { + "epoch": 0.84, + "learning_rate": 1.3013551014287406e-05, + "loss": 0.5048, + "step": 28239 + }, + { + "epoch": 0.84, + "learning_rate": 1.30087949084015e-05, + "loss": 0.585, + "step": 28240 + }, + { + "epoch": 0.84, + "learning_rate": 1.3004039611325146e-05, + "loss": 0.5156, + "step": 28241 + }, + { + "epoch": 0.84, + "learning_rate": 1.2999285123102612e-05, + "loss": 0.6403, + "step": 28242 + }, + { + "epoch": 0.84, + "learning_rate": 1.2994531443778002e-05, + "loss": 0.4602, + "step": 28243 + }, + { + "epoch": 0.84, + "learning_rate": 1.298977857339555e-05, + "loss": 0.5475, + "step": 28244 + }, + { + "epoch": 0.84, + "learning_rate": 1.2985026511999464e-05, + "loss": 0.5027, + "step": 28245 + }, + { + "epoch": 0.84, + "learning_rate": 1.2980275259633945e-05, + "loss": 0.6454, + "step": 28246 + }, + { + "epoch": 0.84, + "learning_rate": 1.2975524816343155e-05, + "loss": 0.5119, + "step": 28247 + }, + { + "epoch": 0.84, + "learning_rate": 1.2970775182171213e-05, + "loss": 0.4695, + "step": 28248 + }, + { + "epoch": 0.84, + "learning_rate": 1.296602635716232e-05, + "loss": 0.5723, + "step": 28249 + }, + { + "epoch": 0.84, + "learning_rate": 1.296127834136065e-05, + "loss": 0.5695, + "step": 28250 + }, + { + "epoch": 0.84, + "learning_rate": 1.2956531134810301e-05, + "loss": 0.5199, + "step": 28251 + }, + { + "epoch": 0.84, + "learning_rate": 1.295178473755545e-05, + "loss": 0.511, + "step": 28252 + }, + { + "epoch": 0.84, + "learning_rate": 1.2947039149640195e-05, + "loss": 0.4542, + "step": 28253 + }, + { + "epoch": 0.84, + "learning_rate": 1.2942294371108699e-05, + "loss": 0.6237, + "step": 28254 + }, + { + "epoch": 0.84, + "learning_rate": 1.2937550402005028e-05, + "loss": 0.5642, + "step": 28255 + }, + { + "epoch": 0.84, + "learning_rate": 1.2932807242373335e-05, + "loss": 0.4505, + "step": 28256 + }, + { + "epoch": 0.84, + "learning_rate": 1.2928064892257673e-05, + "loss": 0.5992, + "step": 28257 + }, + { + "epoch": 0.84, + "learning_rate": 1.2923323351702188e-05, + "loss": 0.5739, + "step": 28258 + }, + { + "epoch": 0.84, + "learning_rate": 1.2918582620750907e-05, + "loss": 0.4988, + "step": 28259 + }, + { + "epoch": 0.84, + "learning_rate": 1.2913842699447942e-05, + "loss": 0.4934, + "step": 28260 + }, + { + "epoch": 0.84, + "learning_rate": 1.2909103587837379e-05, + "loss": 0.5342, + "step": 28261 + }, + { + "epoch": 0.84, + "learning_rate": 1.290436528596326e-05, + "loss": 0.6917, + "step": 28262 + }, + { + "epoch": 0.84, + "learning_rate": 1.2899627793869629e-05, + "loss": 0.5466, + "step": 28263 + }, + { + "epoch": 0.84, + "learning_rate": 1.2894891111600527e-05, + "loss": 0.6168, + "step": 28264 + }, + { + "epoch": 0.84, + "learning_rate": 1.2890155239200042e-05, + "loss": 0.4717, + "step": 28265 + }, + { + "epoch": 0.84, + "learning_rate": 1.288542017671217e-05, + "loss": 0.4654, + "step": 28266 + }, + { + "epoch": 0.84, + "learning_rate": 1.2880685924180924e-05, + "loss": 0.6208, + "step": 28267 + }, + { + "epoch": 0.84, + "learning_rate": 1.2875952481650332e-05, + "loss": 0.5398, + "step": 28268 + }, + { + "epoch": 0.84, + "learning_rate": 1.287121984916444e-05, + "loss": 0.5145, + "step": 28269 + }, + { + "epoch": 0.84, + "learning_rate": 1.2866488026767198e-05, + "loss": 0.507, + "step": 28270 + }, + { + "epoch": 0.84, + "learning_rate": 1.286175701450264e-05, + "loss": 0.5126, + "step": 28271 + }, + { + "epoch": 0.84, + "learning_rate": 1.2857026812414707e-05, + "loss": 0.4475, + "step": 28272 + }, + { + "epoch": 0.84, + "learning_rate": 1.2852297420547432e-05, + "loss": 0.5561, + "step": 28273 + }, + { + "epoch": 0.84, + "learning_rate": 1.2847568838944746e-05, + "loss": 0.5936, + "step": 28274 + }, + { + "epoch": 0.84, + "learning_rate": 1.2842841067650647e-05, + "loss": 0.5694, + "step": 28275 + }, + { + "epoch": 0.84, + "learning_rate": 1.2838114106709053e-05, + "loss": 0.5128, + "step": 28276 + }, + { + "epoch": 0.84, + "learning_rate": 1.2833387956163957e-05, + "loss": 0.4931, + "step": 28277 + }, + { + "epoch": 0.84, + "learning_rate": 1.2828662616059262e-05, + "loss": 0.5347, + "step": 28278 + }, + { + "epoch": 0.84, + "learning_rate": 1.2823938086438914e-05, + "loss": 0.5686, + "step": 28279 + }, + { + "epoch": 0.84, + "learning_rate": 1.2819214367346865e-05, + "loss": 0.5152, + "step": 28280 + }, + { + "epoch": 0.84, + "learning_rate": 1.2814491458827017e-05, + "loss": 0.5919, + "step": 28281 + }, + { + "epoch": 0.84, + "learning_rate": 1.2809769360923252e-05, + "loss": 0.434, + "step": 28282 + }, + { + "epoch": 0.84, + "learning_rate": 1.2805048073679505e-05, + "loss": 0.5611, + "step": 28283 + }, + { + "epoch": 0.84, + "learning_rate": 1.2800327597139672e-05, + "loss": 0.5737, + "step": 28284 + }, + { + "epoch": 0.84, + "learning_rate": 1.2795607931347675e-05, + "loss": 0.5007, + "step": 28285 + }, + { + "epoch": 0.84, + "learning_rate": 1.2790889076347313e-05, + "loss": 0.5759, + "step": 28286 + }, + { + "epoch": 0.84, + "learning_rate": 1.2786171032182514e-05, + "loss": 0.4399, + "step": 28287 + }, + { + "epoch": 0.84, + "learning_rate": 1.2781453798897136e-05, + "loss": 0.6108, + "step": 28288 + }, + { + "epoch": 0.84, + "learning_rate": 1.2776737376535063e-05, + "loss": 0.549, + "step": 28289 + }, + { + "epoch": 0.84, + "learning_rate": 1.2772021765140118e-05, + "loss": 0.4668, + "step": 28290 + }, + { + "epoch": 0.84, + "learning_rate": 1.2767306964756121e-05, + "loss": 0.5682, + "step": 28291 + }, + { + "epoch": 0.84, + "learning_rate": 1.2762592975426945e-05, + "loss": 0.498, + "step": 28292 + }, + { + "epoch": 0.84, + "learning_rate": 1.2757879797196427e-05, + "loss": 0.4058, + "step": 28293 + }, + { + "epoch": 0.84, + "learning_rate": 1.275316743010836e-05, + "loss": 0.5699, + "step": 28294 + }, + { + "epoch": 0.84, + "learning_rate": 1.2748455874206578e-05, + "loss": 0.456, + "step": 28295 + }, + { + "epoch": 0.84, + "learning_rate": 1.274374512953488e-05, + "loss": 0.4535, + "step": 28296 + }, + { + "epoch": 0.84, + "learning_rate": 1.273903519613705e-05, + "loss": 0.5761, + "step": 28297 + }, + { + "epoch": 0.84, + "learning_rate": 1.2734326074056879e-05, + "loss": 0.5136, + "step": 28298 + }, + { + "epoch": 0.84, + "learning_rate": 1.2729617763338197e-05, + "loss": 0.5872, + "step": 28299 + }, + { + "epoch": 0.84, + "learning_rate": 1.2724910264024747e-05, + "loss": 0.5414, + "step": 28300 + }, + { + "epoch": 0.84, + "learning_rate": 1.2720203576160273e-05, + "loss": 0.5331, + "step": 28301 + }, + { + "epoch": 0.84, + "learning_rate": 1.2715497699788549e-05, + "loss": 0.5764, + "step": 28302 + }, + { + "epoch": 0.84, + "learning_rate": 1.271079263495335e-05, + "loss": 0.5655, + "step": 28303 + }, + { + "epoch": 0.84, + "learning_rate": 1.2706088381698434e-05, + "loss": 0.5106, + "step": 28304 + }, + { + "epoch": 0.84, + "learning_rate": 1.2701384940067507e-05, + "loss": 0.5192, + "step": 28305 + }, + { + "epoch": 0.84, + "learning_rate": 1.269668231010429e-05, + "loss": 0.5311, + "step": 28306 + }, + { + "epoch": 0.84, + "learning_rate": 1.2691980491852529e-05, + "loss": 0.6144, + "step": 28307 + }, + { + "epoch": 0.84, + "learning_rate": 1.2687279485355963e-05, + "loss": 0.4782, + "step": 28308 + }, + { + "epoch": 0.84, + "learning_rate": 1.2682579290658258e-05, + "loss": 0.5559, + "step": 28309 + }, + { + "epoch": 0.84, + "learning_rate": 1.2677879907803125e-05, + "loss": 0.4189, + "step": 28310 + }, + { + "epoch": 0.84, + "learning_rate": 1.267318133683425e-05, + "loss": 0.5212, + "step": 28311 + }, + { + "epoch": 0.84, + "learning_rate": 1.2668483577795354e-05, + "loss": 0.4979, + "step": 28312 + }, + { + "epoch": 0.84, + "learning_rate": 1.2663786630730068e-05, + "loss": 0.536, + "step": 28313 + }, + { + "epoch": 0.84, + "learning_rate": 1.2659090495682114e-05, + "loss": 0.5057, + "step": 28314 + }, + { + "epoch": 0.84, + "learning_rate": 1.2654395172695088e-05, + "loss": 0.5914, + "step": 28315 + }, + { + "epoch": 0.84, + "learning_rate": 1.2649700661812724e-05, + "loss": 0.5117, + "step": 28316 + }, + { + "epoch": 0.84, + "learning_rate": 1.2645006963078599e-05, + "loss": 0.5747, + "step": 28317 + }, + { + "epoch": 0.84, + "learning_rate": 1.2640314076536398e-05, + "loss": 0.5592, + "step": 28318 + }, + { + "epoch": 0.84, + "learning_rate": 1.263562200222973e-05, + "loss": 0.4711, + "step": 28319 + }, + { + "epoch": 0.84, + "learning_rate": 1.263093074020224e-05, + "loss": 0.5767, + "step": 28320 + }, + { + "epoch": 0.84, + "learning_rate": 1.2626240290497516e-05, + "loss": 0.5561, + "step": 28321 + }, + { + "epoch": 0.84, + "learning_rate": 1.2621550653159187e-05, + "loss": 0.4889, + "step": 28322 + }, + { + "epoch": 0.84, + "learning_rate": 1.2616861828230875e-05, + "loss": 0.6113, + "step": 28323 + }, + { + "epoch": 0.84, + "learning_rate": 1.2612173815756146e-05, + "loss": 0.5018, + "step": 28324 + }, + { + "epoch": 0.84, + "learning_rate": 1.2607486615778586e-05, + "loss": 0.5039, + "step": 28325 + }, + { + "epoch": 0.84, + "learning_rate": 1.2602800228341782e-05, + "loss": 0.5118, + "step": 28326 + }, + { + "epoch": 0.84, + "learning_rate": 1.2598114653489323e-05, + "loss": 0.5332, + "step": 28327 + }, + { + "epoch": 0.84, + "learning_rate": 1.2593429891264764e-05, + "loss": 0.4601, + "step": 28328 + }, + { + "epoch": 0.84, + "learning_rate": 1.2588745941711644e-05, + "loss": 0.5317, + "step": 28329 + }, + { + "epoch": 0.84, + "learning_rate": 1.2584062804873508e-05, + "loss": 0.5315, + "step": 28330 + }, + { + "epoch": 0.84, + "learning_rate": 1.2579380480793957e-05, + "loss": 0.6342, + "step": 28331 + }, + { + "epoch": 0.84, + "learning_rate": 1.257469896951644e-05, + "loss": 0.4727, + "step": 28332 + }, + { + "epoch": 0.84, + "learning_rate": 1.2570018271084571e-05, + "loss": 0.5376, + "step": 28333 + }, + { + "epoch": 0.84, + "learning_rate": 1.2565338385541792e-05, + "loss": 0.5112, + "step": 28334 + }, + { + "epoch": 0.84, + "learning_rate": 1.2560659312931666e-05, + "loss": 0.4649, + "step": 28335 + }, + { + "epoch": 0.84, + "learning_rate": 1.2555981053297672e-05, + "loss": 0.5316, + "step": 28336 + }, + { + "epoch": 0.84, + "learning_rate": 1.255130360668333e-05, + "loss": 0.4906, + "step": 28337 + }, + { + "epoch": 0.84, + "learning_rate": 1.2546626973132091e-05, + "loss": 0.5121, + "step": 28338 + }, + { + "epoch": 0.84, + "learning_rate": 1.2541951152687481e-05, + "loss": 0.6578, + "step": 28339 + }, + { + "epoch": 0.84, + "learning_rate": 1.253727614539293e-05, + "loss": 0.4913, + "step": 28340 + }, + { + "epoch": 0.84, + "learning_rate": 1.2532601951291922e-05, + "loss": 0.5983, + "step": 28341 + }, + { + "epoch": 0.84, + "learning_rate": 1.2527928570427939e-05, + "loss": 0.5508, + "step": 28342 + }, + { + "epoch": 0.84, + "learning_rate": 1.2523256002844418e-05, + "loss": 0.5799, + "step": 28343 + }, + { + "epoch": 0.84, + "learning_rate": 1.2518584248584787e-05, + "loss": 0.4428, + "step": 28344 + }, + { + "epoch": 0.84, + "learning_rate": 1.2513913307692481e-05, + "loss": 0.4632, + "step": 28345 + }, + { + "epoch": 0.84, + "learning_rate": 1.2509243180210938e-05, + "loss": 0.5547, + "step": 28346 + }, + { + "epoch": 0.84, + "learning_rate": 1.2504573866183634e-05, + "loss": 0.5932, + "step": 28347 + }, + { + "epoch": 0.84, + "learning_rate": 1.2499905365653885e-05, + "loss": 0.4789, + "step": 28348 + }, + { + "epoch": 0.84, + "learning_rate": 1.2495237678665138e-05, + "loss": 0.4852, + "step": 28349 + }, + { + "epoch": 0.84, + "learning_rate": 1.2490570805260792e-05, + "loss": 0.5041, + "step": 28350 + }, + { + "epoch": 0.84, + "learning_rate": 1.2485904745484267e-05, + "loss": 0.5181, + "step": 28351 + }, + { + "epoch": 0.84, + "learning_rate": 1.2481239499378926e-05, + "loss": 0.5183, + "step": 28352 + }, + { + "epoch": 0.84, + "learning_rate": 1.2476575066988106e-05, + "loss": 0.515, + "step": 28353 + }, + { + "epoch": 0.84, + "learning_rate": 1.2471911448355212e-05, + "loss": 0.5166, + "step": 28354 + }, + { + "epoch": 0.84, + "learning_rate": 1.2467248643523622e-05, + "loss": 0.4862, + "step": 28355 + }, + { + "epoch": 0.84, + "learning_rate": 1.2462586652536646e-05, + "loss": 0.5048, + "step": 28356 + }, + { + "epoch": 0.84, + "learning_rate": 1.245792547543767e-05, + "loss": 0.6125, + "step": 28357 + }, + { + "epoch": 0.84, + "learning_rate": 1.2453265112269996e-05, + "loss": 0.5514, + "step": 28358 + }, + { + "epoch": 0.84, + "learning_rate": 1.2448605563076986e-05, + "loss": 0.586, + "step": 28359 + }, + { + "epoch": 0.84, + "learning_rate": 1.2443946827901931e-05, + "loss": 0.593, + "step": 28360 + }, + { + "epoch": 0.84, + "learning_rate": 1.2439288906788183e-05, + "loss": 0.4841, + "step": 28361 + }, + { + "epoch": 0.84, + "learning_rate": 1.2434631799779028e-05, + "loss": 0.5468, + "step": 28362 + }, + { + "epoch": 0.84, + "learning_rate": 1.2429975506917734e-05, + "loss": 0.5803, + "step": 28363 + }, + { + "epoch": 0.84, + "learning_rate": 1.2425320028247633e-05, + "loss": 0.6088, + "step": 28364 + }, + { + "epoch": 0.84, + "learning_rate": 1.2420665363812012e-05, + "loss": 0.5765, + "step": 28365 + }, + { + "epoch": 0.84, + "learning_rate": 1.2416011513654147e-05, + "loss": 0.5297, + "step": 28366 + }, + { + "epoch": 0.84, + "learning_rate": 1.2411358477817302e-05, + "loss": 0.5254, + "step": 28367 + }, + { + "epoch": 0.84, + "learning_rate": 1.2406706256344724e-05, + "loss": 0.5587, + "step": 28368 + }, + { + "epoch": 0.84, + "learning_rate": 1.2402054849279677e-05, + "loss": 0.5079, + "step": 28369 + }, + { + "epoch": 0.84, + "learning_rate": 1.2397404256665423e-05, + "loss": 0.5053, + "step": 28370 + }, + { + "epoch": 0.84, + "learning_rate": 1.2392754478545199e-05, + "loss": 0.6014, + "step": 28371 + }, + { + "epoch": 0.84, + "learning_rate": 1.2388105514962201e-05, + "loss": 0.6343, + "step": 28372 + }, + { + "epoch": 0.84, + "learning_rate": 1.2383457365959672e-05, + "loss": 0.591, + "step": 28373 + }, + { + "epoch": 0.84, + "learning_rate": 1.2378810031580868e-05, + "loss": 0.6278, + "step": 28374 + }, + { + "epoch": 0.84, + "learning_rate": 1.2374163511868942e-05, + "loss": 0.4632, + "step": 28375 + }, + { + "epoch": 0.84, + "learning_rate": 1.2369517806867149e-05, + "loss": 0.5451, + "step": 28376 + }, + { + "epoch": 0.84, + "learning_rate": 1.2364872916618619e-05, + "loss": 0.5527, + "step": 28377 + }, + { + "epoch": 0.84, + "learning_rate": 1.23602288411666e-05, + "loss": 0.5536, + "step": 28378 + }, + { + "epoch": 0.84, + "learning_rate": 1.2355585580554218e-05, + "loss": 0.5216, + "step": 28379 + }, + { + "epoch": 0.84, + "learning_rate": 1.2350943134824699e-05, + "loss": 0.5689, + "step": 28380 + }, + { + "epoch": 0.84, + "learning_rate": 1.2346301504021141e-05, + "loss": 0.528, + "step": 28381 + }, + { + "epoch": 0.84, + "learning_rate": 1.2341660688186773e-05, + "loss": 0.4732, + "step": 28382 + }, + { + "epoch": 0.85, + "learning_rate": 1.2337020687364676e-05, + "loss": 0.6426, + "step": 28383 + }, + { + "epoch": 0.85, + "learning_rate": 1.2332381501598034e-05, + "loss": 0.4836, + "step": 28384 + }, + { + "epoch": 0.85, + "learning_rate": 1.232774313092998e-05, + "loss": 0.5671, + "step": 28385 + }, + { + "epoch": 0.85, + "learning_rate": 1.2323105575403626e-05, + "loss": 0.5277, + "step": 28386 + }, + { + "epoch": 0.85, + "learning_rate": 1.231846883506207e-05, + "loss": 0.5382, + "step": 28387 + }, + { + "epoch": 0.85, + "learning_rate": 1.2313832909948452e-05, + "loss": 0.5379, + "step": 28388 + }, + { + "epoch": 0.85, + "learning_rate": 1.2309197800105854e-05, + "loss": 0.4343, + "step": 28389 + }, + { + "epoch": 0.85, + "learning_rate": 1.2304563505577438e-05, + "loss": 0.5002, + "step": 28390 + }, + { + "epoch": 0.85, + "learning_rate": 1.2299930026406192e-05, + "loss": 0.4416, + "step": 28391 + }, + { + "epoch": 0.85, + "learning_rate": 1.229529736263525e-05, + "loss": 0.5332, + "step": 28392 + }, + { + "epoch": 0.85, + "learning_rate": 1.2290665514307675e-05, + "loss": 0.4386, + "step": 28393 + }, + { + "epoch": 0.85, + "learning_rate": 1.2286034481466546e-05, + "loss": 0.4969, + "step": 28394 + }, + { + "epoch": 0.85, + "learning_rate": 1.2281404264154917e-05, + "loss": 0.5731, + "step": 28395 + }, + { + "epoch": 0.85, + "learning_rate": 1.2276774862415807e-05, + "loss": 0.6692, + "step": 28396 + }, + { + "epoch": 0.85, + "learning_rate": 1.2272146276292306e-05, + "loss": 0.5712, + "step": 28397 + }, + { + "epoch": 0.85, + "learning_rate": 1.2267518505827402e-05, + "loss": 0.5554, + "step": 28398 + }, + { + "epoch": 0.85, + "learning_rate": 1.2262891551064138e-05, + "loss": 0.4813, + "step": 28399 + }, + { + "epoch": 0.85, + "learning_rate": 1.2258265412045566e-05, + "loss": 0.4837, + "step": 28400 + }, + { + "epoch": 0.85, + "learning_rate": 1.2253640088814678e-05, + "loss": 0.5961, + "step": 28401 + }, + { + "epoch": 0.85, + "learning_rate": 1.2249015581414435e-05, + "loss": 0.5001, + "step": 28402 + }, + { + "epoch": 0.85, + "learning_rate": 1.2244391889887886e-05, + "loss": 0.5058, + "step": 28403 + }, + { + "epoch": 0.85, + "learning_rate": 1.2239769014278024e-05, + "loss": 0.3928, + "step": 28404 + }, + { + "epoch": 0.85, + "learning_rate": 1.2235146954627807e-05, + "loss": 0.4966, + "step": 28405 + }, + { + "epoch": 0.85, + "learning_rate": 1.223052571098019e-05, + "loss": 0.5555, + "step": 28406 + }, + { + "epoch": 0.85, + "learning_rate": 1.2225905283378169e-05, + "loss": 0.611, + "step": 28407 + }, + { + "epoch": 0.85, + "learning_rate": 1.2221285671864702e-05, + "loss": 0.4677, + "step": 28408 + }, + { + "epoch": 0.85, + "learning_rate": 1.2216666876482775e-05, + "loss": 0.4388, + "step": 28409 + }, + { + "epoch": 0.85, + "learning_rate": 1.221204889727524e-05, + "loss": 0.4628, + "step": 28410 + }, + { + "epoch": 0.85, + "learning_rate": 1.220743173428509e-05, + "loss": 0.6482, + "step": 28411 + }, + { + "epoch": 0.85, + "learning_rate": 1.2202815387555255e-05, + "loss": 0.6174, + "step": 28412 + }, + { + "epoch": 0.85, + "learning_rate": 1.2198199857128667e-05, + "loss": 0.5878, + "step": 28413 + }, + { + "epoch": 0.85, + "learning_rate": 1.2193585143048225e-05, + "loss": 0.5165, + "step": 28414 + }, + { + "epoch": 0.85, + "learning_rate": 1.2188971245356805e-05, + "loss": 0.614, + "step": 28415 + }, + { + "epoch": 0.85, + "learning_rate": 1.2184358164097343e-05, + "loss": 0.5604, + "step": 28416 + }, + { + "epoch": 0.85, + "learning_rate": 1.2179745899312733e-05, + "loss": 0.4959, + "step": 28417 + }, + { + "epoch": 0.85, + "learning_rate": 1.2175134451045821e-05, + "loss": 0.4932, + "step": 28418 + }, + { + "epoch": 0.85, + "learning_rate": 1.217052381933953e-05, + "loss": 0.5442, + "step": 28419 + }, + { + "epoch": 0.85, + "learning_rate": 1.2165914004236677e-05, + "loss": 0.6402, + "step": 28420 + }, + { + "epoch": 0.85, + "learning_rate": 1.2161305005780177e-05, + "loss": 0.5604, + "step": 28421 + }, + { + "epoch": 0.85, + "learning_rate": 1.2156696824012837e-05, + "loss": 0.5477, + "step": 28422 + }, + { + "epoch": 0.85, + "learning_rate": 1.2152089458977533e-05, + "loss": 0.6067, + "step": 28423 + }, + { + "epoch": 0.85, + "learning_rate": 1.2147482910717068e-05, + "loss": 0.4853, + "step": 28424 + }, + { + "epoch": 0.85, + "learning_rate": 1.2142877179274304e-05, + "loss": 0.5342, + "step": 28425 + }, + { + "epoch": 0.85, + "learning_rate": 1.2138272264692042e-05, + "loss": 0.5225, + "step": 28426 + }, + { + "epoch": 0.85, + "learning_rate": 1.2133668167013101e-05, + "loss": 0.6675, + "step": 28427 + }, + { + "epoch": 0.85, + "learning_rate": 1.2129064886280317e-05, + "loss": 0.4748, + "step": 28428 + }, + { + "epoch": 0.85, + "learning_rate": 1.2124462422536465e-05, + "loss": 0.4864, + "step": 28429 + }, + { + "epoch": 0.85, + "learning_rate": 1.2119860775824322e-05, + "loss": 0.5444, + "step": 28430 + }, + { + "epoch": 0.85, + "learning_rate": 1.2115259946186685e-05, + "loss": 0.536, + "step": 28431 + }, + { + "epoch": 0.85, + "learning_rate": 1.2110659933666346e-05, + "loss": 0.4645, + "step": 28432 + }, + { + "epoch": 0.85, + "learning_rate": 1.2106060738306068e-05, + "loss": 0.4942, + "step": 28433 + }, + { + "epoch": 0.85, + "learning_rate": 1.2101462360148586e-05, + "loss": 0.6041, + "step": 28434 + }, + { + "epoch": 0.85, + "learning_rate": 1.2096864799236673e-05, + "loss": 0.5721, + "step": 28435 + }, + { + "epoch": 0.85, + "learning_rate": 1.2092268055613099e-05, + "loss": 0.5632, + "step": 28436 + }, + { + "epoch": 0.85, + "learning_rate": 1.208767212932056e-05, + "loss": 0.6612, + "step": 28437 + }, + { + "epoch": 0.85, + "learning_rate": 1.2083077020401834e-05, + "loss": 0.4829, + "step": 28438 + }, + { + "epoch": 0.85, + "learning_rate": 1.2078482728899587e-05, + "loss": 0.5493, + "step": 28439 + }, + { + "epoch": 0.85, + "learning_rate": 1.2073889254856586e-05, + "loss": 0.5653, + "step": 28440 + }, + { + "epoch": 0.85, + "learning_rate": 1.2069296598315504e-05, + "loss": 0.6013, + "step": 28441 + }, + { + "epoch": 0.85, + "learning_rate": 1.2064704759319068e-05, + "loss": 0.562, + "step": 28442 + }, + { + "epoch": 0.85, + "learning_rate": 1.206011373790995e-05, + "loss": 0.5638, + "step": 28443 + }, + { + "epoch": 0.85, + "learning_rate": 1.2055523534130853e-05, + "loss": 0.5725, + "step": 28444 + }, + { + "epoch": 0.85, + "learning_rate": 1.205093414802444e-05, + "loss": 0.5884, + "step": 28445 + }, + { + "epoch": 0.85, + "learning_rate": 1.2046345579633379e-05, + "loss": 0.5675, + "step": 28446 + }, + { + "epoch": 0.85, + "learning_rate": 1.2041757829000366e-05, + "loss": 0.5501, + "step": 28447 + }, + { + "epoch": 0.85, + "learning_rate": 1.2037170896168026e-05, + "loss": 0.5497, + "step": 28448 + }, + { + "epoch": 0.85, + "learning_rate": 1.2032584781179002e-05, + "loss": 0.6147, + "step": 28449 + }, + { + "epoch": 0.85, + "learning_rate": 1.2027999484075936e-05, + "loss": 0.56, + "step": 28450 + }, + { + "epoch": 0.85, + "learning_rate": 1.2023415004901473e-05, + "loss": 0.52, + "step": 28451 + }, + { + "epoch": 0.85, + "learning_rate": 1.2018831343698278e-05, + "loss": 0.4408, + "step": 28452 + }, + { + "epoch": 0.85, + "learning_rate": 1.2014248500508874e-05, + "loss": 0.5088, + "step": 28453 + }, + { + "epoch": 0.85, + "learning_rate": 1.2009666475375925e-05, + "loss": 0.5286, + "step": 28454 + }, + { + "epoch": 0.85, + "learning_rate": 1.2005085268342032e-05, + "loss": 0.4681, + "step": 28455 + }, + { + "epoch": 0.85, + "learning_rate": 1.2000504879449792e-05, + "loss": 0.4579, + "step": 28456 + }, + { + "epoch": 0.85, + "learning_rate": 1.1995925308741795e-05, + "loss": 0.5137, + "step": 28457 + }, + { + "epoch": 0.85, + "learning_rate": 1.1991346556260586e-05, + "loss": 0.5713, + "step": 28458 + }, + { + "epoch": 0.85, + "learning_rate": 1.1986768622048761e-05, + "loss": 0.474, + "step": 28459 + }, + { + "epoch": 0.85, + "learning_rate": 1.1982191506148899e-05, + "loss": 0.4321, + "step": 28460 + }, + { + "epoch": 0.85, + "learning_rate": 1.1977615208603522e-05, + "loss": 0.5591, + "step": 28461 + }, + { + "epoch": 0.85, + "learning_rate": 1.1973039729455227e-05, + "loss": 0.6108, + "step": 28462 + }, + { + "epoch": 0.85, + "learning_rate": 1.1968465068746515e-05, + "loss": 0.5247, + "step": 28463 + }, + { + "epoch": 0.85, + "learning_rate": 1.1963891226519908e-05, + "loss": 0.5074, + "step": 28464 + }, + { + "epoch": 0.85, + "learning_rate": 1.1959318202817948e-05, + "loss": 0.5455, + "step": 28465 + }, + { + "epoch": 0.85, + "learning_rate": 1.195474599768318e-05, + "loss": 0.566, + "step": 28466 + }, + { + "epoch": 0.85, + "learning_rate": 1.1950174611158105e-05, + "loss": 0.5102, + "step": 28467 + }, + { + "epoch": 0.85, + "learning_rate": 1.1945604043285185e-05, + "loss": 0.4182, + "step": 28468 + }, + { + "epoch": 0.85, + "learning_rate": 1.1941034294106933e-05, + "loss": 0.5661, + "step": 28469 + }, + { + "epoch": 0.85, + "learning_rate": 1.193646536366586e-05, + "loss": 0.473, + "step": 28470 + }, + { + "epoch": 0.85, + "learning_rate": 1.1931897252004454e-05, + "loss": 0.4444, + "step": 28471 + }, + { + "epoch": 0.85, + "learning_rate": 1.1927329959165167e-05, + "loss": 0.5633, + "step": 28472 + }, + { + "epoch": 0.85, + "learning_rate": 1.1922763485190436e-05, + "loss": 0.6275, + "step": 28473 + }, + { + "epoch": 0.85, + "learning_rate": 1.1918197830122747e-05, + "loss": 0.4591, + "step": 28474 + }, + { + "epoch": 0.85, + "learning_rate": 1.1913632994004575e-05, + "loss": 0.545, + "step": 28475 + }, + { + "epoch": 0.85, + "learning_rate": 1.1909068976878334e-05, + "loss": 0.5598, + "step": 28476 + }, + { + "epoch": 0.85, + "learning_rate": 1.1904505778786435e-05, + "loss": 0.5234, + "step": 28477 + }, + { + "epoch": 0.85, + "learning_rate": 1.189994339977133e-05, + "loss": 0.5272, + "step": 28478 + }, + { + "epoch": 0.85, + "learning_rate": 1.1895381839875464e-05, + "loss": 0.5292, + "step": 28479 + }, + { + "epoch": 0.85, + "learning_rate": 1.1890821099141202e-05, + "loss": 0.4747, + "step": 28480 + }, + { + "epoch": 0.85, + "learning_rate": 1.188626117761098e-05, + "loss": 0.5338, + "step": 28481 + }, + { + "epoch": 0.85, + "learning_rate": 1.1881702075327173e-05, + "loss": 0.5054, + "step": 28482 + }, + { + "epoch": 0.85, + "learning_rate": 1.1877143792332201e-05, + "loss": 0.4388, + "step": 28483 + }, + { + "epoch": 0.85, + "learning_rate": 1.187258632866839e-05, + "loss": 0.559, + "step": 28484 + }, + { + "epoch": 0.85, + "learning_rate": 1.186802968437818e-05, + "loss": 0.5149, + "step": 28485 + }, + { + "epoch": 0.85, + "learning_rate": 1.1863473859503881e-05, + "loss": 0.7027, + "step": 28486 + }, + { + "epoch": 0.85, + "learning_rate": 1.1858918854087897e-05, + "loss": 0.6251, + "step": 28487 + }, + { + "epoch": 0.85, + "learning_rate": 1.1854364668172524e-05, + "loss": 0.6347, + "step": 28488 + }, + { + "epoch": 0.85, + "learning_rate": 1.1849811301800151e-05, + "loss": 0.5417, + "step": 28489 + }, + { + "epoch": 0.85, + "learning_rate": 1.184525875501311e-05, + "loss": 0.5806, + "step": 28490 + }, + { + "epoch": 0.85, + "learning_rate": 1.1840707027853726e-05, + "loss": 0.5738, + "step": 28491 + }, + { + "epoch": 0.85, + "learning_rate": 1.1836156120364283e-05, + "loss": 0.4833, + "step": 28492 + }, + { + "epoch": 0.85, + "learning_rate": 1.1831606032587128e-05, + "loss": 0.4953, + "step": 28493 + }, + { + "epoch": 0.85, + "learning_rate": 1.182705676456457e-05, + "loss": 0.5142, + "step": 28494 + }, + { + "epoch": 0.85, + "learning_rate": 1.1822508316338909e-05, + "loss": 0.4381, + "step": 28495 + }, + { + "epoch": 0.85, + "learning_rate": 1.1817960687952389e-05, + "loss": 0.5088, + "step": 28496 + }, + { + "epoch": 0.85, + "learning_rate": 1.1813413879447333e-05, + "loss": 0.4865, + "step": 28497 + }, + { + "epoch": 0.85, + "learning_rate": 1.1808867890866026e-05, + "loss": 0.5352, + "step": 28498 + }, + { + "epoch": 0.85, + "learning_rate": 1.1804322722250693e-05, + "loss": 0.5352, + "step": 28499 + }, + { + "epoch": 0.85, + "learning_rate": 1.1799778373643645e-05, + "loss": 0.6766, + "step": 28500 + }, + { + "epoch": 0.85, + "learning_rate": 1.1795234845087077e-05, + "loss": 0.5646, + "step": 28501 + }, + { + "epoch": 0.85, + "learning_rate": 1.1790692136623282e-05, + "loss": 0.5261, + "step": 28502 + }, + { + "epoch": 0.85, + "learning_rate": 1.1786150248294448e-05, + "loss": 0.458, + "step": 28503 + }, + { + "epoch": 0.85, + "learning_rate": 1.1781609180142861e-05, + "loss": 0.5326, + "step": 28504 + }, + { + "epoch": 0.85, + "learning_rate": 1.1777068932210689e-05, + "loss": 0.5879, + "step": 28505 + }, + { + "epoch": 0.85, + "learning_rate": 1.1772529504540187e-05, + "loss": 0.4609, + "step": 28506 + }, + { + "epoch": 0.85, + "learning_rate": 1.1767990897173519e-05, + "loss": 0.5337, + "step": 28507 + }, + { + "epoch": 0.85, + "learning_rate": 1.17634531101529e-05, + "loss": 0.5052, + "step": 28508 + }, + { + "epoch": 0.85, + "learning_rate": 1.175891614352056e-05, + "loss": 0.5633, + "step": 28509 + }, + { + "epoch": 0.85, + "learning_rate": 1.1754379997318643e-05, + "loss": 0.5949, + "step": 28510 + }, + { + "epoch": 0.85, + "learning_rate": 1.1749844671589306e-05, + "loss": 0.451, + "step": 28511 + }, + { + "epoch": 0.85, + "learning_rate": 1.1745310166374745e-05, + "loss": 0.5897, + "step": 28512 + }, + { + "epoch": 0.85, + "learning_rate": 1.1740776481717109e-05, + "loss": 0.6345, + "step": 28513 + }, + { + "epoch": 0.85, + "learning_rate": 1.1736243617658604e-05, + "loss": 0.463, + "step": 28514 + }, + { + "epoch": 0.85, + "learning_rate": 1.1731711574241277e-05, + "loss": 0.4649, + "step": 28515 + }, + { + "epoch": 0.85, + "learning_rate": 1.1727180351507327e-05, + "loss": 0.4684, + "step": 28516 + }, + { + "epoch": 0.85, + "learning_rate": 1.1722649949498864e-05, + "loss": 0.5514, + "step": 28517 + }, + { + "epoch": 0.85, + "learning_rate": 1.1718120368258045e-05, + "loss": 0.5699, + "step": 28518 + }, + { + "epoch": 0.85, + "learning_rate": 1.1713591607826956e-05, + "loss": 0.549, + "step": 28519 + }, + { + "epoch": 0.85, + "learning_rate": 1.1709063668247677e-05, + "loss": 0.5614, + "step": 28520 + }, + { + "epoch": 0.85, + "learning_rate": 1.170453654956234e-05, + "loss": 0.5716, + "step": 28521 + }, + { + "epoch": 0.85, + "learning_rate": 1.1700010251813053e-05, + "loss": 0.4795, + "step": 28522 + }, + { + "epoch": 0.85, + "learning_rate": 1.1695484775041853e-05, + "loss": 0.5968, + "step": 28523 + }, + { + "epoch": 0.85, + "learning_rate": 1.1690960119290861e-05, + "loss": 0.5639, + "step": 28524 + }, + { + "epoch": 0.85, + "learning_rate": 1.1686436284602109e-05, + "loss": 0.5279, + "step": 28525 + }, + { + "epoch": 0.85, + "learning_rate": 1.1681913271017697e-05, + "loss": 0.4786, + "step": 28526 + }, + { + "epoch": 0.85, + "learning_rate": 1.1677391078579624e-05, + "loss": 0.5422, + "step": 28527 + }, + { + "epoch": 0.85, + "learning_rate": 1.1672869707330002e-05, + "loss": 0.5361, + "step": 28528 + }, + { + "epoch": 0.85, + "learning_rate": 1.166834915731082e-05, + "loss": 0.4122, + "step": 28529 + }, + { + "epoch": 0.85, + "learning_rate": 1.166382942856411e-05, + "loss": 0.5544, + "step": 28530 + }, + { + "epoch": 0.85, + "learning_rate": 1.1659310521131906e-05, + "loss": 0.45, + "step": 28531 + }, + { + "epoch": 0.85, + "learning_rate": 1.1654792435056227e-05, + "loss": 0.5287, + "step": 28532 + }, + { + "epoch": 0.85, + "learning_rate": 1.1650275170379087e-05, + "loss": 0.548, + "step": 28533 + }, + { + "epoch": 0.85, + "learning_rate": 1.1645758727142486e-05, + "loss": 0.5657, + "step": 28534 + }, + { + "epoch": 0.85, + "learning_rate": 1.1641243105388377e-05, + "loss": 0.5148, + "step": 28535 + }, + { + "epoch": 0.85, + "learning_rate": 1.1636728305158772e-05, + "loss": 0.5255, + "step": 28536 + }, + { + "epoch": 0.85, + "learning_rate": 1.1632214326495684e-05, + "loss": 0.6126, + "step": 28537 + }, + { + "epoch": 0.85, + "learning_rate": 1.162770116944103e-05, + "loss": 0.4952, + "step": 28538 + }, + { + "epoch": 0.85, + "learning_rate": 1.162318883403678e-05, + "loss": 0.6158, + "step": 28539 + }, + { + "epoch": 0.85, + "learning_rate": 1.1618677320324889e-05, + "loss": 0.4986, + "step": 28540 + }, + { + "epoch": 0.85, + "learning_rate": 1.1614166628347335e-05, + "loss": 0.5504, + "step": 28541 + }, + { + "epoch": 0.85, + "learning_rate": 1.1609656758146014e-05, + "loss": 0.5841, + "step": 28542 + }, + { + "epoch": 0.85, + "learning_rate": 1.1605147709762898e-05, + "loss": 0.5554, + "step": 28543 + }, + { + "epoch": 0.85, + "learning_rate": 1.160063948323986e-05, + "loss": 0.5382, + "step": 28544 + }, + { + "epoch": 0.85, + "learning_rate": 1.159613207861887e-05, + "loss": 0.5888, + "step": 28545 + }, + { + "epoch": 0.85, + "learning_rate": 1.1591625495941783e-05, + "loss": 0.4342, + "step": 28546 + }, + { + "epoch": 0.85, + "learning_rate": 1.1587119735250551e-05, + "loss": 0.6197, + "step": 28547 + }, + { + "epoch": 0.85, + "learning_rate": 1.158261479658702e-05, + "loss": 0.3933, + "step": 28548 + }, + { + "epoch": 0.85, + "learning_rate": 1.1578110679993114e-05, + "loss": 0.366, + "step": 28549 + }, + { + "epoch": 0.85, + "learning_rate": 1.1573607385510688e-05, + "loss": 0.6324, + "step": 28550 + }, + { + "epoch": 0.85, + "learning_rate": 1.1569104913181606e-05, + "loss": 0.566, + "step": 28551 + }, + { + "epoch": 0.85, + "learning_rate": 1.1564603263047758e-05, + "loss": 0.4649, + "step": 28552 + }, + { + "epoch": 0.85, + "learning_rate": 1.1560102435150988e-05, + "loss": 0.5782, + "step": 28553 + }, + { + "epoch": 0.85, + "learning_rate": 1.1555602429533108e-05, + "loss": 0.431, + "step": 28554 + }, + { + "epoch": 0.85, + "learning_rate": 1.1551103246235995e-05, + "loss": 0.4671, + "step": 28555 + }, + { + "epoch": 0.85, + "learning_rate": 1.154660488530147e-05, + "loss": 0.5253, + "step": 28556 + }, + { + "epoch": 0.85, + "learning_rate": 1.1542107346771402e-05, + "loss": 0.5327, + "step": 28557 + }, + { + "epoch": 0.85, + "learning_rate": 1.1537610630687522e-05, + "loss": 0.4984, + "step": 28558 + }, + { + "epoch": 0.85, + "learning_rate": 1.1533114737091677e-05, + "loss": 0.4359, + "step": 28559 + }, + { + "epoch": 0.85, + "learning_rate": 1.1528619666025698e-05, + "loss": 0.4391, + "step": 28560 + }, + { + "epoch": 0.85, + "learning_rate": 1.1524125417531317e-05, + "loss": 0.5904, + "step": 28561 + }, + { + "epoch": 0.85, + "learning_rate": 1.1519631991650393e-05, + "loss": 0.5632, + "step": 28562 + }, + { + "epoch": 0.85, + "learning_rate": 1.1515139388424633e-05, + "loss": 0.5127, + "step": 28563 + }, + { + "epoch": 0.85, + "learning_rate": 1.1510647607895875e-05, + "loss": 0.4933, + "step": 28564 + }, + { + "epoch": 0.85, + "learning_rate": 1.1506156650105815e-05, + "loss": 0.4792, + "step": 28565 + }, + { + "epoch": 0.85, + "learning_rate": 1.1501666515096243e-05, + "loss": 0.6498, + "step": 28566 + }, + { + "epoch": 0.85, + "learning_rate": 1.1497177202908927e-05, + "loss": 0.5063, + "step": 28567 + }, + { + "epoch": 0.85, + "learning_rate": 1.1492688713585576e-05, + "loss": 0.5373, + "step": 28568 + }, + { + "epoch": 0.85, + "learning_rate": 1.1488201047167913e-05, + "loss": 0.5256, + "step": 28569 + }, + { + "epoch": 0.85, + "learning_rate": 1.1483714203697681e-05, + "loss": 0.5067, + "step": 28570 + }, + { + "epoch": 0.85, + "learning_rate": 1.1479228183216616e-05, + "loss": 0.578, + "step": 28571 + }, + { + "epoch": 0.85, + "learning_rate": 1.1474742985766407e-05, + "loss": 0.4698, + "step": 28572 + }, + { + "epoch": 0.85, + "learning_rate": 1.147025861138874e-05, + "loss": 0.4307, + "step": 28573 + }, + { + "epoch": 0.85, + "learning_rate": 1.1465775060125317e-05, + "loss": 0.4747, + "step": 28574 + }, + { + "epoch": 0.85, + "learning_rate": 1.1461292332017826e-05, + "loss": 0.5352, + "step": 28575 + }, + { + "epoch": 0.85, + "learning_rate": 1.1456810427108001e-05, + "loss": 0.5004, + "step": 28576 + }, + { + "epoch": 0.85, + "learning_rate": 1.1452329345437418e-05, + "loss": 0.5876, + "step": 28577 + }, + { + "epoch": 0.85, + "learning_rate": 1.144784908704779e-05, + "loss": 0.5697, + "step": 28578 + }, + { + "epoch": 0.85, + "learning_rate": 1.1443369651980761e-05, + "loss": 0.4615, + "step": 28579 + }, + { + "epoch": 0.85, + "learning_rate": 1.1438891040278011e-05, + "loss": 0.4857, + "step": 28580 + }, + { + "epoch": 0.85, + "learning_rate": 1.1434413251981147e-05, + "loss": 0.5034, + "step": 28581 + }, + { + "epoch": 0.85, + "learning_rate": 1.1429936287131803e-05, + "loss": 0.4238, + "step": 28582 + }, + { + "epoch": 0.85, + "learning_rate": 1.1425460145771594e-05, + "loss": 0.4997, + "step": 28583 + }, + { + "epoch": 0.85, + "learning_rate": 1.1420984827942194e-05, + "loss": 0.6922, + "step": 28584 + }, + { + "epoch": 0.85, + "learning_rate": 1.1416510333685137e-05, + "loss": 0.5207, + "step": 28585 + }, + { + "epoch": 0.85, + "learning_rate": 1.1412036663042092e-05, + "loss": 0.4592, + "step": 28586 + }, + { + "epoch": 0.85, + "learning_rate": 1.140756381605459e-05, + "loss": 0.5448, + "step": 28587 + }, + { + "epoch": 0.85, + "learning_rate": 1.1403091792764276e-05, + "loss": 0.4743, + "step": 28588 + }, + { + "epoch": 0.85, + "learning_rate": 1.1398620593212683e-05, + "loss": 0.4817, + "step": 28589 + }, + { + "epoch": 0.85, + "learning_rate": 1.1394150217441425e-05, + "loss": 0.5725, + "step": 28590 + }, + { + "epoch": 0.85, + "learning_rate": 1.138968066549203e-05, + "loss": 0.4699, + "step": 28591 + }, + { + "epoch": 0.85, + "learning_rate": 1.1385211937406081e-05, + "loss": 0.621, + "step": 28592 + }, + { + "epoch": 0.85, + "learning_rate": 1.1380744033225087e-05, + "loss": 0.505, + "step": 28593 + }, + { + "epoch": 0.85, + "learning_rate": 1.1376276952990617e-05, + "loss": 0.5495, + "step": 28594 + }, + { + "epoch": 0.85, + "learning_rate": 1.1371810696744234e-05, + "loss": 0.3917, + "step": 28595 + }, + { + "epoch": 0.85, + "learning_rate": 1.1367345264527419e-05, + "loss": 0.551, + "step": 28596 + }, + { + "epoch": 0.85, + "learning_rate": 1.136288065638168e-05, + "loss": 0.4787, + "step": 28597 + }, + { + "epoch": 0.85, + "learning_rate": 1.1358416872348565e-05, + "loss": 0.5401, + "step": 28598 + }, + { + "epoch": 0.85, + "learning_rate": 1.1353953912469561e-05, + "loss": 0.5833, + "step": 28599 + }, + { + "epoch": 0.85, + "learning_rate": 1.1349491776786181e-05, + "loss": 0.4674, + "step": 28600 + }, + { + "epoch": 0.85, + "learning_rate": 1.1345030465339868e-05, + "loss": 0.5728, + "step": 28601 + }, + { + "epoch": 0.85, + "learning_rate": 1.1340569978172123e-05, + "loss": 0.5407, + "step": 28602 + }, + { + "epoch": 0.85, + "learning_rate": 1.1336110315324444e-05, + "loss": 0.5437, + "step": 28603 + }, + { + "epoch": 0.85, + "learning_rate": 1.1331651476838246e-05, + "loss": 0.6097, + "step": 28604 + }, + { + "epoch": 0.85, + "learning_rate": 1.1327193462755048e-05, + "loss": 0.5922, + "step": 28605 + }, + { + "epoch": 0.85, + "learning_rate": 1.132273627311623e-05, + "loss": 0.6637, + "step": 28606 + }, + { + "epoch": 0.85, + "learning_rate": 1.1318279907963292e-05, + "loss": 0.4258, + "step": 28607 + }, + { + "epoch": 0.85, + "learning_rate": 1.1313824367337622e-05, + "loss": 0.4137, + "step": 28608 + }, + { + "epoch": 0.85, + "learning_rate": 1.1309369651280688e-05, + "loss": 0.3665, + "step": 28609 + }, + { + "epoch": 0.85, + "learning_rate": 1.1304915759833867e-05, + "loss": 0.5068, + "step": 28610 + }, + { + "epoch": 0.85, + "learning_rate": 1.1300462693038604e-05, + "loss": 0.4394, + "step": 28611 + }, + { + "epoch": 0.85, + "learning_rate": 1.129601045093628e-05, + "loss": 0.5283, + "step": 28612 + }, + { + "epoch": 0.85, + "learning_rate": 1.1291559033568288e-05, + "loss": 0.5283, + "step": 28613 + }, + { + "epoch": 0.85, + "learning_rate": 1.1287108440976036e-05, + "loss": 0.5481, + "step": 28614 + }, + { + "epoch": 0.85, + "learning_rate": 1.1282658673200907e-05, + "loss": 0.4691, + "step": 28615 + }, + { + "epoch": 0.85, + "learning_rate": 1.1278209730284229e-05, + "loss": 0.4534, + "step": 28616 + }, + { + "epoch": 0.85, + "learning_rate": 1.1273761612267397e-05, + "loss": 0.5483, + "step": 28617 + }, + { + "epoch": 0.85, + "learning_rate": 1.126931431919177e-05, + "loss": 0.5291, + "step": 28618 + }, + { + "epoch": 0.85, + "learning_rate": 1.1264867851098727e-05, + "loss": 0.4976, + "step": 28619 + }, + { + "epoch": 0.85, + "learning_rate": 1.1260422208029542e-05, + "loss": 0.5493, + "step": 28620 + }, + { + "epoch": 0.85, + "learning_rate": 1.1255977390025584e-05, + "loss": 0.51, + "step": 28621 + }, + { + "epoch": 0.85, + "learning_rate": 1.1251533397128178e-05, + "loss": 0.4523, + "step": 28622 + }, + { + "epoch": 0.85, + "learning_rate": 1.1247090229378666e-05, + "loss": 0.5778, + "step": 28623 + }, + { + "epoch": 0.85, + "learning_rate": 1.1242647886818325e-05, + "loss": 0.4208, + "step": 28624 + }, + { + "epoch": 0.85, + "learning_rate": 1.1238206369488458e-05, + "loss": 0.4917, + "step": 28625 + }, + { + "epoch": 0.85, + "learning_rate": 1.1233765677430385e-05, + "loss": 0.4933, + "step": 28626 + }, + { + "epoch": 0.85, + "learning_rate": 1.1229325810685365e-05, + "loss": 0.5796, + "step": 28627 + }, + { + "epoch": 0.85, + "learning_rate": 1.1224886769294696e-05, + "loss": 0.5305, + "step": 28628 + }, + { + "epoch": 0.85, + "learning_rate": 1.1220448553299656e-05, + "loss": 0.4881, + "step": 28629 + }, + { + "epoch": 0.85, + "learning_rate": 1.1216011162741513e-05, + "loss": 0.507, + "step": 28630 + }, + { + "epoch": 0.85, + "learning_rate": 1.1211574597661489e-05, + "loss": 0.5653, + "step": 28631 + }, + { + "epoch": 0.85, + "learning_rate": 1.1207138858100851e-05, + "loss": 0.5012, + "step": 28632 + }, + { + "epoch": 0.85, + "learning_rate": 1.1202703944100878e-05, + "loss": 0.6376, + "step": 28633 + }, + { + "epoch": 0.85, + "learning_rate": 1.1198269855702759e-05, + "loss": 0.4976, + "step": 28634 + }, + { + "epoch": 0.85, + "learning_rate": 1.1193836592947727e-05, + "loss": 0.4339, + "step": 28635 + }, + { + "epoch": 0.85, + "learning_rate": 1.1189404155876992e-05, + "loss": 0.5485, + "step": 28636 + }, + { + "epoch": 0.85, + "learning_rate": 1.1184972544531792e-05, + "loss": 0.6054, + "step": 28637 + }, + { + "epoch": 0.85, + "learning_rate": 1.1180541758953345e-05, + "loss": 0.5661, + "step": 28638 + }, + { + "epoch": 0.85, + "learning_rate": 1.1176111799182809e-05, + "loss": 0.4919, + "step": 28639 + }, + { + "epoch": 0.85, + "learning_rate": 1.1171682665261373e-05, + "loss": 0.4755, + "step": 28640 + }, + { + "epoch": 0.85, + "learning_rate": 1.1167254357230216e-05, + "loss": 0.6137, + "step": 28641 + }, + { + "epoch": 0.85, + "learning_rate": 1.116282687513055e-05, + "loss": 0.4766, + "step": 28642 + }, + { + "epoch": 0.85, + "learning_rate": 1.1158400219003518e-05, + "loss": 0.4745, + "step": 28643 + }, + { + "epoch": 0.85, + "learning_rate": 1.1153974388890242e-05, + "loss": 0.4723, + "step": 28644 + }, + { + "epoch": 0.85, + "learning_rate": 1.1149549384831903e-05, + "loss": 0.5492, + "step": 28645 + }, + { + "epoch": 0.85, + "learning_rate": 1.1145125206869667e-05, + "loss": 0.5365, + "step": 28646 + }, + { + "epoch": 0.85, + "learning_rate": 1.1140701855044621e-05, + "loss": 0.4592, + "step": 28647 + }, + { + "epoch": 0.85, + "learning_rate": 1.1136279329397936e-05, + "loss": 0.5667, + "step": 28648 + }, + { + "epoch": 0.85, + "learning_rate": 1.1131857629970688e-05, + "loss": 0.6004, + "step": 28649 + }, + { + "epoch": 0.85, + "learning_rate": 1.1127436756804021e-05, + "loss": 0.5415, + "step": 28650 + }, + { + "epoch": 0.85, + "learning_rate": 1.1123016709939005e-05, + "loss": 0.4727, + "step": 28651 + }, + { + "epoch": 0.85, + "learning_rate": 1.1118597489416783e-05, + "loss": 0.4972, + "step": 28652 + }, + { + "epoch": 0.85, + "learning_rate": 1.1114179095278398e-05, + "loss": 0.5117, + "step": 28653 + }, + { + "epoch": 0.85, + "learning_rate": 1.1109761527564966e-05, + "loss": 0.6297, + "step": 28654 + }, + { + "epoch": 0.85, + "learning_rate": 1.1105344786317528e-05, + "loss": 0.5752, + "step": 28655 + }, + { + "epoch": 0.85, + "learning_rate": 1.1100928871577154e-05, + "loss": 0.5536, + "step": 28656 + }, + { + "epoch": 0.85, + "learning_rate": 1.1096513783384944e-05, + "loss": 0.4381, + "step": 28657 + }, + { + "epoch": 0.85, + "learning_rate": 1.1092099521781907e-05, + "loss": 0.5218, + "step": 28658 + }, + { + "epoch": 0.85, + "learning_rate": 1.1087686086809079e-05, + "loss": 0.5944, + "step": 28659 + }, + { + "epoch": 0.85, + "learning_rate": 1.1083273478507505e-05, + "loss": 0.5686, + "step": 28660 + }, + { + "epoch": 0.85, + "learning_rate": 1.1078861696918241e-05, + "loss": 0.4536, + "step": 28661 + }, + { + "epoch": 0.85, + "learning_rate": 1.1074450742082276e-05, + "loss": 0.5514, + "step": 28662 + }, + { + "epoch": 0.85, + "learning_rate": 1.10700406140406e-05, + "loss": 0.4359, + "step": 28663 + }, + { + "epoch": 0.85, + "learning_rate": 1.1065631312834246e-05, + "loss": 0.5233, + "step": 28664 + }, + { + "epoch": 0.85, + "learning_rate": 1.1061222838504237e-05, + "loss": 0.5491, + "step": 28665 + }, + { + "epoch": 0.85, + "learning_rate": 1.1056815191091497e-05, + "loss": 0.4638, + "step": 28666 + }, + { + "epoch": 0.85, + "learning_rate": 1.1052408370637069e-05, + "loss": 0.4199, + "step": 28667 + }, + { + "epoch": 0.85, + "learning_rate": 1.1048002377181866e-05, + "loss": 0.6081, + "step": 28668 + }, + { + "epoch": 0.85, + "learning_rate": 1.1043597210766921e-05, + "loss": 0.5096, + "step": 28669 + }, + { + "epoch": 0.85, + "learning_rate": 1.1039192871433125e-05, + "loss": 0.4924, + "step": 28670 + }, + { + "epoch": 0.85, + "learning_rate": 1.1034789359221475e-05, + "loss": 0.5091, + "step": 28671 + }, + { + "epoch": 0.85, + "learning_rate": 1.1030386674172876e-05, + "loss": 0.4399, + "step": 28672 + }, + { + "epoch": 0.85, + "learning_rate": 1.1025984816328305e-05, + "loss": 0.6451, + "step": 28673 + }, + { + "epoch": 0.85, + "learning_rate": 1.102158378572864e-05, + "loss": 0.4922, + "step": 28674 + }, + { + "epoch": 0.85, + "learning_rate": 1.1017183582414835e-05, + "loss": 0.48, + "step": 28675 + }, + { + "epoch": 0.85, + "learning_rate": 1.1012784206427806e-05, + "loss": 0.5288, + "step": 28676 + }, + { + "epoch": 0.85, + "learning_rate": 1.100838565780844e-05, + "loss": 0.5229, + "step": 28677 + }, + { + "epoch": 0.85, + "learning_rate": 1.1003987936597614e-05, + "loss": 0.5646, + "step": 28678 + }, + { + "epoch": 0.85, + "learning_rate": 1.0999591042836244e-05, + "loss": 0.4845, + "step": 28679 + }, + { + "epoch": 0.85, + "learning_rate": 1.0995194976565205e-05, + "loss": 0.5809, + "step": 28680 + }, + { + "epoch": 0.85, + "learning_rate": 1.0990799737825408e-05, + "loss": 0.4599, + "step": 28681 + }, + { + "epoch": 0.85, + "learning_rate": 1.0986405326657645e-05, + "loss": 0.5326, + "step": 28682 + }, + { + "epoch": 0.85, + "learning_rate": 1.0982011743102805e-05, + "loss": 0.5202, + "step": 28683 + }, + { + "epoch": 0.85, + "learning_rate": 1.0977618987201732e-05, + "loss": 0.6262, + "step": 28684 + }, + { + "epoch": 0.85, + "learning_rate": 1.0973227058995305e-05, + "loss": 0.5212, + "step": 28685 + }, + { + "epoch": 0.85, + "learning_rate": 1.0968835958524326e-05, + "loss": 0.4048, + "step": 28686 + }, + { + "epoch": 0.85, + "learning_rate": 1.0964445685829605e-05, + "loss": 0.5407, + "step": 28687 + }, + { + "epoch": 0.85, + "learning_rate": 1.0960056240951988e-05, + "loss": 0.5555, + "step": 28688 + }, + { + "epoch": 0.85, + "learning_rate": 1.0955667623932298e-05, + "loss": 0.6389, + "step": 28689 + }, + { + "epoch": 0.85, + "learning_rate": 1.09512798348113e-05, + "loss": 0.5057, + "step": 28690 + }, + { + "epoch": 0.85, + "learning_rate": 1.0946892873629833e-05, + "loss": 0.5655, + "step": 28691 + }, + { + "epoch": 0.85, + "learning_rate": 1.0942506740428659e-05, + "loss": 0.5563, + "step": 28692 + }, + { + "epoch": 0.85, + "learning_rate": 1.0938121435248538e-05, + "loss": 0.5878, + "step": 28693 + }, + { + "epoch": 0.85, + "learning_rate": 1.093373695813028e-05, + "loss": 0.5968, + "step": 28694 + }, + { + "epoch": 0.85, + "learning_rate": 1.0929353309114654e-05, + "loss": 0.5491, + "step": 28695 + }, + { + "epoch": 0.85, + "learning_rate": 1.0924970488242392e-05, + "loss": 0.5778, + "step": 28696 + }, + { + "epoch": 0.85, + "learning_rate": 1.0920588495554241e-05, + "loss": 0.4869, + "step": 28697 + }, + { + "epoch": 0.85, + "learning_rate": 1.0916207331090955e-05, + "loss": 0.4733, + "step": 28698 + }, + { + "epoch": 0.85, + "learning_rate": 1.0911826994893259e-05, + "loss": 0.6367, + "step": 28699 + }, + { + "epoch": 0.85, + "learning_rate": 1.090744748700192e-05, + "loss": 0.5064, + "step": 28700 + }, + { + "epoch": 0.85, + "learning_rate": 1.0903068807457618e-05, + "loss": 0.4864, + "step": 28701 + }, + { + "epoch": 0.85, + "learning_rate": 1.089869095630105e-05, + "loss": 0.4434, + "step": 28702 + }, + { + "epoch": 0.85, + "learning_rate": 1.0894313933572952e-05, + "loss": 0.4787, + "step": 28703 + }, + { + "epoch": 0.85, + "learning_rate": 1.0889937739314016e-05, + "loss": 0.5594, + "step": 28704 + }, + { + "epoch": 0.85, + "learning_rate": 1.0885562373564929e-05, + "loss": 0.5544, + "step": 28705 + }, + { + "epoch": 0.85, + "learning_rate": 1.0881187836366347e-05, + "loss": 0.5783, + "step": 28706 + }, + { + "epoch": 0.85, + "learning_rate": 1.0876814127758961e-05, + "loss": 0.5092, + "step": 28707 + }, + { + "epoch": 0.85, + "learning_rate": 1.0872441247783461e-05, + "loss": 0.5733, + "step": 28708 + }, + { + "epoch": 0.85, + "learning_rate": 1.0868069196480446e-05, + "loss": 0.53, + "step": 28709 + }, + { + "epoch": 0.85, + "learning_rate": 1.086369797389063e-05, + "loss": 0.4998, + "step": 28710 + }, + { + "epoch": 0.85, + "learning_rate": 1.0859327580054601e-05, + "loss": 0.5273, + "step": 28711 + }, + { + "epoch": 0.85, + "learning_rate": 1.085495801501304e-05, + "loss": 0.42, + "step": 28712 + }, + { + "epoch": 0.85, + "learning_rate": 1.0850589278806523e-05, + "loss": 0.4255, + "step": 28713 + }, + { + "epoch": 0.85, + "learning_rate": 1.0846221371475718e-05, + "loss": 0.6401, + "step": 28714 + }, + { + "epoch": 0.85, + "learning_rate": 1.0841854293061193e-05, + "loss": 0.4914, + "step": 28715 + }, + { + "epoch": 0.85, + "learning_rate": 1.0837488043603594e-05, + "loss": 0.4557, + "step": 28716 + }, + { + "epoch": 0.85, + "learning_rate": 1.0833122623143477e-05, + "loss": 0.4375, + "step": 28717 + }, + { + "epoch": 0.85, + "learning_rate": 1.0828758031721442e-05, + "loss": 0.5216, + "step": 28718 + }, + { + "epoch": 0.86, + "learning_rate": 1.0824394269378091e-05, + "loss": 0.488, + "step": 28719 + }, + { + "epoch": 0.86, + "learning_rate": 1.0820031336153991e-05, + "loss": 0.4745, + "step": 28720 + }, + { + "epoch": 0.86, + "learning_rate": 1.0815669232089664e-05, + "loss": 0.528, + "step": 28721 + }, + { + "epoch": 0.86, + "learning_rate": 1.0811307957225713e-05, + "loss": 0.5837, + "step": 28722 + }, + { + "epoch": 0.86, + "learning_rate": 1.080694751160266e-05, + "loss": 0.5357, + "step": 28723 + }, + { + "epoch": 0.86, + "learning_rate": 1.0802587895261108e-05, + "loss": 0.4985, + "step": 28724 + }, + { + "epoch": 0.86, + "learning_rate": 1.0798229108241497e-05, + "loss": 0.6585, + "step": 28725 + }, + { + "epoch": 0.86, + "learning_rate": 1.0793871150584412e-05, + "loss": 0.5075, + "step": 28726 + }, + { + "epoch": 0.86, + "learning_rate": 1.078951402233037e-05, + "loss": 0.5311, + "step": 28727 + }, + { + "epoch": 0.86, + "learning_rate": 1.0785157723519845e-05, + "loss": 0.5837, + "step": 28728 + }, + { + "epoch": 0.86, + "learning_rate": 1.07808022541934e-05, + "loss": 0.6409, + "step": 28729 + }, + { + "epoch": 0.86, + "learning_rate": 1.0776447614391461e-05, + "loss": 0.578, + "step": 28730 + }, + { + "epoch": 0.86, + "learning_rate": 1.0772093804154582e-05, + "loss": 0.613, + "step": 28731 + }, + { + "epoch": 0.86, + "learning_rate": 1.0767740823523187e-05, + "loss": 0.4989, + "step": 28732 + }, + { + "epoch": 0.86, + "learning_rate": 1.0763388672537777e-05, + "loss": 0.5184, + "step": 28733 + }, + { + "epoch": 0.86, + "learning_rate": 1.0759037351238831e-05, + "loss": 0.5168, + "step": 28734 + }, + { + "epoch": 0.86, + "learning_rate": 1.0754686859666797e-05, + "loss": 0.5009, + "step": 28735 + }, + { + "epoch": 0.86, + "learning_rate": 1.0750337197862081e-05, + "loss": 0.5271, + "step": 28736 + }, + { + "epoch": 0.86, + "learning_rate": 1.0745988365865167e-05, + "loss": 0.4894, + "step": 28737 + }, + { + "epoch": 0.86, + "learning_rate": 1.0741640363716488e-05, + "loss": 0.4649, + "step": 28738 + }, + { + "epoch": 0.86, + "learning_rate": 1.0737293191456477e-05, + "loss": 0.564, + "step": 28739 + }, + { + "epoch": 0.86, + "learning_rate": 1.0732946849125503e-05, + "loss": 0.5363, + "step": 28740 + }, + { + "epoch": 0.86, + "learning_rate": 1.0728601336764022e-05, + "loss": 0.4969, + "step": 28741 + }, + { + "epoch": 0.86, + "learning_rate": 1.0724256654412423e-05, + "loss": 0.4547, + "step": 28742 + }, + { + "epoch": 0.86, + "learning_rate": 1.071991280211112e-05, + "loss": 0.5614, + "step": 28743 + }, + { + "epoch": 0.86, + "learning_rate": 1.0715569779900492e-05, + "loss": 0.5536, + "step": 28744 + }, + { + "epoch": 0.86, + "learning_rate": 1.0711227587820883e-05, + "loss": 0.6056, + "step": 28745 + }, + { + "epoch": 0.86, + "learning_rate": 1.0706886225912705e-05, + "loss": 0.4245, + "step": 28746 + }, + { + "epoch": 0.86, + "learning_rate": 1.0702545694216316e-05, + "loss": 0.5457, + "step": 28747 + }, + { + "epoch": 0.86, + "learning_rate": 1.069820599277207e-05, + "loss": 0.5086, + "step": 28748 + }, + { + "epoch": 0.86, + "learning_rate": 1.0693867121620293e-05, + "loss": 0.4608, + "step": 28749 + }, + { + "epoch": 0.86, + "learning_rate": 1.0689529080801352e-05, + "loss": 0.6757, + "step": 28750 + }, + { + "epoch": 0.86, + "learning_rate": 1.0685191870355594e-05, + "loss": 0.5061, + "step": 28751 + }, + { + "epoch": 0.86, + "learning_rate": 1.0680855490323294e-05, + "loss": 0.5546, + "step": 28752 + }, + { + "epoch": 0.86, + "learning_rate": 1.0676519940744822e-05, + "loss": 0.5562, + "step": 28753 + }, + { + "epoch": 0.86, + "learning_rate": 1.0672185221660458e-05, + "loss": 0.483, + "step": 28754 + }, + { + "epoch": 0.86, + "learning_rate": 1.0667851333110524e-05, + "loss": 0.4709, + "step": 28755 + }, + { + "epoch": 0.86, + "learning_rate": 1.0663518275135286e-05, + "loss": 0.4329, + "step": 28756 + }, + { + "epoch": 0.86, + "learning_rate": 1.0659186047775072e-05, + "loss": 0.5669, + "step": 28757 + }, + { + "epoch": 0.86, + "learning_rate": 1.0654854651070135e-05, + "loss": 0.5856, + "step": 28758 + }, + { + "epoch": 0.86, + "learning_rate": 1.065052408506073e-05, + "loss": 0.5932, + "step": 28759 + }, + { + "epoch": 0.86, + "learning_rate": 1.0646194349787141e-05, + "loss": 0.5126, + "step": 28760 + }, + { + "epoch": 0.86, + "learning_rate": 1.0641865445289623e-05, + "loss": 0.5711, + "step": 28761 + }, + { + "epoch": 0.86, + "learning_rate": 1.0637537371608452e-05, + "loss": 0.6026, + "step": 28762 + }, + { + "epoch": 0.86, + "learning_rate": 1.0633210128783832e-05, + "loss": 0.4977, + "step": 28763 + }, + { + "epoch": 0.86, + "learning_rate": 1.0628883716855997e-05, + "loss": 0.5474, + "step": 28764 + }, + { + "epoch": 0.86, + "learning_rate": 1.062455813586517e-05, + "loss": 0.5813, + "step": 28765 + }, + { + "epoch": 0.86, + "learning_rate": 1.0620233385851608e-05, + "loss": 0.5184, + "step": 28766 + }, + { + "epoch": 0.86, + "learning_rate": 1.061590946685549e-05, + "loss": 0.482, + "step": 28767 + }, + { + "epoch": 0.86, + "learning_rate": 1.0611586378917004e-05, + "loss": 0.7102, + "step": 28768 + }, + { + "epoch": 0.86, + "learning_rate": 1.0607264122076354e-05, + "loss": 0.5803, + "step": 28769 + }, + { + "epoch": 0.86, + "learning_rate": 1.0602942696373763e-05, + "loss": 0.5684, + "step": 28770 + }, + { + "epoch": 0.86, + "learning_rate": 1.0598622101849353e-05, + "loss": 0.4657, + "step": 28771 + }, + { + "epoch": 0.86, + "learning_rate": 1.059430233854335e-05, + "loss": 0.5458, + "step": 28772 + }, + { + "epoch": 0.86, + "learning_rate": 1.0589983406495862e-05, + "loss": 0.5153, + "step": 28773 + }, + { + "epoch": 0.86, + "learning_rate": 1.0585665305747095e-05, + "loss": 0.5711, + "step": 28774 + }, + { + "epoch": 0.86, + "learning_rate": 1.0581348036337157e-05, + "loss": 0.6317, + "step": 28775 + }, + { + "epoch": 0.86, + "learning_rate": 1.057703159830623e-05, + "loss": 0.5409, + "step": 28776 + }, + { + "epoch": 0.86, + "learning_rate": 1.0572715991694393e-05, + "loss": 0.5824, + "step": 28777 + }, + { + "epoch": 0.86, + "learning_rate": 1.0568401216541835e-05, + "loss": 0.4619, + "step": 28778 + }, + { + "epoch": 0.86, + "learning_rate": 1.0564087272888602e-05, + "loss": 0.5734, + "step": 28779 + }, + { + "epoch": 0.86, + "learning_rate": 1.0559774160774848e-05, + "loss": 0.5931, + "step": 28780 + }, + { + "epoch": 0.86, + "learning_rate": 1.0555461880240692e-05, + "loss": 0.6048, + "step": 28781 + }, + { + "epoch": 0.86, + "learning_rate": 1.0551150431326195e-05, + "loss": 0.5361, + "step": 28782 + }, + { + "epoch": 0.86, + "learning_rate": 1.054683981407143e-05, + "loss": 0.5363, + "step": 28783 + }, + { + "epoch": 0.86, + "learning_rate": 1.0542530028516496e-05, + "loss": 0.5083, + "step": 28784 + }, + { + "epoch": 0.86, + "learning_rate": 1.0538221074701471e-05, + "loss": 0.4663, + "step": 28785 + }, + { + "epoch": 0.86, + "learning_rate": 1.0533912952666447e-05, + "loss": 0.6428, + "step": 28786 + }, + { + "epoch": 0.86, + "learning_rate": 1.0529605662451392e-05, + "loss": 0.6697, + "step": 28787 + }, + { + "epoch": 0.86, + "learning_rate": 1.0525299204096417e-05, + "loss": 0.5931, + "step": 28788 + }, + { + "epoch": 0.86, + "learning_rate": 1.052099357764157e-05, + "loss": 0.566, + "step": 28789 + }, + { + "epoch": 0.86, + "learning_rate": 1.0516688783126826e-05, + "loss": 0.4663, + "step": 28790 + }, + { + "epoch": 0.86, + "learning_rate": 1.0512384820592281e-05, + "loss": 0.5978, + "step": 28791 + }, + { + "epoch": 0.86, + "learning_rate": 1.0508081690077886e-05, + "loss": 0.4977, + "step": 28792 + }, + { + "epoch": 0.86, + "learning_rate": 1.0503779391623713e-05, + "loss": 0.554, + "step": 28793 + }, + { + "epoch": 0.86, + "learning_rate": 1.0499477925269696e-05, + "loss": 0.4168, + "step": 28794 + }, + { + "epoch": 0.86, + "learning_rate": 1.0495177291055868e-05, + "loss": 0.4525, + "step": 28795 + }, + { + "epoch": 0.86, + "learning_rate": 1.049087748902222e-05, + "loss": 0.5506, + "step": 28796 + }, + { + "epoch": 0.86, + "learning_rate": 1.0486578519208734e-05, + "loss": 0.5253, + "step": 28797 + }, + { + "epoch": 0.86, + "learning_rate": 1.048228038165533e-05, + "loss": 0.4826, + "step": 28798 + }, + { + "epoch": 0.86, + "learning_rate": 1.047798307640201e-05, + "loss": 0.5132, + "step": 28799 + }, + { + "epoch": 0.86, + "learning_rate": 1.0473686603488753e-05, + "loss": 0.4842, + "step": 28800 + }, + { + "epoch": 0.86, + "learning_rate": 1.0469390962955462e-05, + "loss": 0.6438, + "step": 28801 + }, + { + "epoch": 0.86, + "learning_rate": 1.046509615484208e-05, + "loss": 0.5407, + "step": 28802 + }, + { + "epoch": 0.86, + "learning_rate": 1.0460802179188545e-05, + "loss": 0.4499, + "step": 28803 + }, + { + "epoch": 0.86, + "learning_rate": 1.0456509036034779e-05, + "loss": 0.5329, + "step": 28804 + }, + { + "epoch": 0.86, + "learning_rate": 1.0452216725420727e-05, + "loss": 0.4383, + "step": 28805 + }, + { + "epoch": 0.86, + "learning_rate": 1.0447925247386281e-05, + "loss": 0.5412, + "step": 28806 + }, + { + "epoch": 0.86, + "learning_rate": 1.0443634601971309e-05, + "loss": 0.5204, + "step": 28807 + }, + { + "epoch": 0.86, + "learning_rate": 1.0439344789215722e-05, + "loss": 0.4826, + "step": 28808 + }, + { + "epoch": 0.86, + "learning_rate": 1.0435055809159444e-05, + "loss": 0.5554, + "step": 28809 + }, + { + "epoch": 0.86, + "learning_rate": 1.043076766184231e-05, + "loss": 0.5211, + "step": 28810 + }, + { + "epoch": 0.86, + "learning_rate": 1.0426480347304179e-05, + "loss": 0.5105, + "step": 28811 + }, + { + "epoch": 0.86, + "learning_rate": 1.042219386558494e-05, + "loss": 0.4347, + "step": 28812 + }, + { + "epoch": 0.86, + "learning_rate": 1.041790821672446e-05, + "loss": 0.5515, + "step": 28813 + }, + { + "epoch": 0.86, + "learning_rate": 1.041362340076254e-05, + "loss": 0.4804, + "step": 28814 + }, + { + "epoch": 0.86, + "learning_rate": 1.0409339417739072e-05, + "loss": 0.3698, + "step": 28815 + }, + { + "epoch": 0.86, + "learning_rate": 1.0405056267693836e-05, + "loss": 0.5615, + "step": 28816 + }, + { + "epoch": 0.86, + "learning_rate": 1.0400773950666686e-05, + "loss": 0.5284, + "step": 28817 + }, + { + "epoch": 0.86, + "learning_rate": 1.0396492466697426e-05, + "loss": 0.5991, + "step": 28818 + }, + { + "epoch": 0.86, + "learning_rate": 1.039221181582588e-05, + "loss": 0.6025, + "step": 28819 + }, + { + "epoch": 0.86, + "learning_rate": 1.0387931998091816e-05, + "loss": 0.5147, + "step": 28820 + }, + { + "epoch": 0.86, + "learning_rate": 1.0383653013535055e-05, + "loss": 0.4273, + "step": 28821 + }, + { + "epoch": 0.86, + "learning_rate": 1.0379374862195369e-05, + "loss": 0.6243, + "step": 28822 + }, + { + "epoch": 0.86, + "learning_rate": 1.0375097544112522e-05, + "loss": 0.4717, + "step": 28823 + }, + { + "epoch": 0.86, + "learning_rate": 1.037082105932632e-05, + "loss": 0.5057, + "step": 28824 + }, + { + "epoch": 0.86, + "learning_rate": 1.0366545407876504e-05, + "loss": 0.5663, + "step": 28825 + }, + { + "epoch": 0.86, + "learning_rate": 1.0362270589802802e-05, + "loss": 0.4763, + "step": 28826 + }, + { + "epoch": 0.86, + "learning_rate": 1.0357996605144982e-05, + "loss": 0.6072, + "step": 28827 + }, + { + "epoch": 0.86, + "learning_rate": 1.0353723453942798e-05, + "loss": 0.513, + "step": 28828 + }, + { + "epoch": 0.86, + "learning_rate": 1.0349451136235965e-05, + "loss": 0.5551, + "step": 28829 + }, + { + "epoch": 0.86, + "learning_rate": 1.0345179652064186e-05, + "loss": 0.583, + "step": 28830 + }, + { + "epoch": 0.86, + "learning_rate": 1.0340909001467203e-05, + "loss": 0.5885, + "step": 28831 + }, + { + "epoch": 0.86, + "learning_rate": 1.033663918448472e-05, + "loss": 0.5527, + "step": 28832 + }, + { + "epoch": 0.86, + "learning_rate": 1.0332370201156416e-05, + "loss": 0.4772, + "step": 28833 + }, + { + "epoch": 0.86, + "learning_rate": 1.0328102051522015e-05, + "loss": 0.5495, + "step": 28834 + }, + { + "epoch": 0.86, + "learning_rate": 1.0323834735621163e-05, + "loss": 0.4734, + "step": 28835 + }, + { + "epoch": 0.86, + "learning_rate": 1.0319568253493572e-05, + "loss": 0.5748, + "step": 28836 + }, + { + "epoch": 0.86, + "learning_rate": 1.0315302605178867e-05, + "loss": 0.5245, + "step": 28837 + }, + { + "epoch": 0.86, + "learning_rate": 1.031103779071676e-05, + "loss": 0.4969, + "step": 28838 + }, + { + "epoch": 0.86, + "learning_rate": 1.0306773810146853e-05, + "loss": 0.5289, + "step": 28839 + }, + { + "epoch": 0.86, + "learning_rate": 1.0302510663508847e-05, + "loss": 0.4434, + "step": 28840 + }, + { + "epoch": 0.86, + "learning_rate": 1.029824835084231e-05, + "loss": 0.5135, + "step": 28841 + }, + { + "epoch": 0.86, + "learning_rate": 1.0293986872186922e-05, + "loss": 0.4654, + "step": 28842 + }, + { + "epoch": 0.86, + "learning_rate": 1.0289726227582308e-05, + "loss": 0.5513, + "step": 28843 + }, + { + "epoch": 0.86, + "learning_rate": 1.0285466417068056e-05, + "loss": 0.5475, + "step": 28844 + }, + { + "epoch": 0.86, + "learning_rate": 1.0281207440683771e-05, + "loss": 0.5649, + "step": 28845 + }, + { + "epoch": 0.86, + "learning_rate": 1.0276949298469053e-05, + "loss": 0.5881, + "step": 28846 + }, + { + "epoch": 0.86, + "learning_rate": 1.02726919904635e-05, + "loss": 0.6073, + "step": 28847 + }, + { + "epoch": 0.86, + "learning_rate": 1.0268435516706743e-05, + "loss": 0.5574, + "step": 28848 + }, + { + "epoch": 0.86, + "learning_rate": 1.0264179877238267e-05, + "loss": 0.5559, + "step": 28849 + }, + { + "epoch": 0.86, + "learning_rate": 1.0259925072097676e-05, + "loss": 0.5088, + "step": 28850 + }, + { + "epoch": 0.86, + "learning_rate": 1.0255671101324537e-05, + "loss": 0.4575, + "step": 28851 + }, + { + "epoch": 0.86, + "learning_rate": 1.025141796495842e-05, + "loss": 0.5428, + "step": 28852 + }, + { + "epoch": 0.86, + "learning_rate": 1.024716566303885e-05, + "loss": 0.5033, + "step": 28853 + }, + { + "epoch": 0.86, + "learning_rate": 1.0242914195605335e-05, + "loss": 0.5604, + "step": 28854 + }, + { + "epoch": 0.86, + "learning_rate": 1.023866356269746e-05, + "loss": 0.5446, + "step": 28855 + }, + { + "epoch": 0.86, + "learning_rate": 1.0234413764354688e-05, + "loss": 0.5762, + "step": 28856 + }, + { + "epoch": 0.86, + "learning_rate": 1.023016480061657e-05, + "loss": 0.6252, + "step": 28857 + }, + { + "epoch": 0.86, + "learning_rate": 1.0225916671522617e-05, + "loss": 0.5357, + "step": 28858 + }, + { + "epoch": 0.86, + "learning_rate": 1.0221669377112298e-05, + "loss": 0.4453, + "step": 28859 + }, + { + "epoch": 0.86, + "learning_rate": 1.0217422917425112e-05, + "loss": 0.6134, + "step": 28860 + }, + { + "epoch": 0.86, + "learning_rate": 1.021317729250053e-05, + "loss": 0.4279, + "step": 28861 + }, + { + "epoch": 0.86, + "learning_rate": 1.0208932502378066e-05, + "loss": 0.6196, + "step": 28862 + }, + { + "epoch": 0.86, + "learning_rate": 1.0204688547097164e-05, + "loss": 0.564, + "step": 28863 + }, + { + "epoch": 0.86, + "learning_rate": 1.0200445426697258e-05, + "loss": 0.5506, + "step": 28864 + }, + { + "epoch": 0.86, + "learning_rate": 1.019620314121782e-05, + "loss": 0.4367, + "step": 28865 + }, + { + "epoch": 0.86, + "learning_rate": 1.0191961690698294e-05, + "loss": 0.6236, + "step": 28866 + }, + { + "epoch": 0.86, + "learning_rate": 1.0187721075178124e-05, + "loss": 0.5705, + "step": 28867 + }, + { + "epoch": 0.86, + "learning_rate": 1.018348129469674e-05, + "loss": 0.4304, + "step": 28868 + }, + { + "epoch": 0.86, + "learning_rate": 1.0179242349293516e-05, + "loss": 0.4315, + "step": 28869 + }, + { + "epoch": 0.86, + "learning_rate": 1.0175004239007913e-05, + "loss": 0.6315, + "step": 28870 + }, + { + "epoch": 0.86, + "learning_rate": 1.0170766963879331e-05, + "loss": 0.4847, + "step": 28871 + }, + { + "epoch": 0.86, + "learning_rate": 1.016653052394716e-05, + "loss": 0.5547, + "step": 28872 + }, + { + "epoch": 0.86, + "learning_rate": 1.0162294919250759e-05, + "loss": 0.506, + "step": 28873 + }, + { + "epoch": 0.86, + "learning_rate": 1.015806014982954e-05, + "loss": 0.5171, + "step": 28874 + }, + { + "epoch": 0.86, + "learning_rate": 1.0153826215722883e-05, + "loss": 0.439, + "step": 28875 + }, + { + "epoch": 0.86, + "learning_rate": 1.0149593116970125e-05, + "loss": 0.5055, + "step": 28876 + }, + { + "epoch": 0.86, + "learning_rate": 1.0145360853610663e-05, + "loss": 0.5203, + "step": 28877 + }, + { + "epoch": 0.86, + "learning_rate": 1.0141129425683793e-05, + "loss": 0.507, + "step": 28878 + }, + { + "epoch": 0.86, + "learning_rate": 1.0136898833228915e-05, + "loss": 0.4549, + "step": 28879 + }, + { + "epoch": 0.86, + "learning_rate": 1.0132669076285318e-05, + "loss": 0.5287, + "step": 28880 + }, + { + "epoch": 0.86, + "learning_rate": 1.012844015489236e-05, + "loss": 0.6162, + "step": 28881 + }, + { + "epoch": 0.86, + "learning_rate": 1.0124212069089333e-05, + "loss": 0.4965, + "step": 28882 + }, + { + "epoch": 0.86, + "learning_rate": 1.0119984818915573e-05, + "loss": 0.4334, + "step": 28883 + }, + { + "epoch": 0.86, + "learning_rate": 1.0115758404410359e-05, + "loss": 0.633, + "step": 28884 + }, + { + "epoch": 0.86, + "learning_rate": 1.011153282561299e-05, + "loss": 0.6117, + "step": 28885 + }, + { + "epoch": 0.86, + "learning_rate": 1.0107308082562794e-05, + "loss": 0.567, + "step": 28886 + }, + { + "epoch": 0.86, + "learning_rate": 1.0103084175299016e-05, + "loss": 0.5188, + "step": 28887 + }, + { + "epoch": 0.86, + "learning_rate": 1.0098861103860902e-05, + "loss": 0.4725, + "step": 28888 + }, + { + "epoch": 0.86, + "learning_rate": 1.0094638868287754e-05, + "loss": 0.4855, + "step": 28889 + }, + { + "epoch": 0.86, + "learning_rate": 1.009041746861884e-05, + "loss": 0.681, + "step": 28890 + }, + { + "epoch": 0.86, + "learning_rate": 1.0086196904893396e-05, + "loss": 0.5973, + "step": 28891 + }, + { + "epoch": 0.86, + "learning_rate": 1.0081977177150637e-05, + "loss": 0.4822, + "step": 28892 + }, + { + "epoch": 0.86, + "learning_rate": 1.0077758285429805e-05, + "loss": 0.5593, + "step": 28893 + }, + { + "epoch": 0.86, + "learning_rate": 1.0073540229770163e-05, + "loss": 0.5783, + "step": 28894 + }, + { + "epoch": 0.86, + "learning_rate": 1.0069323010210886e-05, + "loss": 0.4647, + "step": 28895 + }, + { + "epoch": 0.86, + "learning_rate": 1.006510662679122e-05, + "loss": 0.6255, + "step": 28896 + }, + { + "epoch": 0.86, + "learning_rate": 1.0060891079550328e-05, + "loss": 0.4702, + "step": 28897 + }, + { + "epoch": 0.86, + "learning_rate": 1.005667636852744e-05, + "loss": 0.6754, + "step": 28898 + }, + { + "epoch": 0.86, + "learning_rate": 1.0052462493761716e-05, + "loss": 0.5548, + "step": 28899 + }, + { + "epoch": 0.86, + "learning_rate": 1.0048249455292336e-05, + "loss": 0.4854, + "step": 28900 + }, + { + "epoch": 0.86, + "learning_rate": 1.004403725315851e-05, + "loss": 0.4894, + "step": 28901 + }, + { + "epoch": 0.86, + "learning_rate": 1.0039825887399368e-05, + "loss": 0.5306, + "step": 28902 + }, + { + "epoch": 0.86, + "learning_rate": 1.0035615358054051e-05, + "loss": 0.4223, + "step": 28903 + }, + { + "epoch": 0.86, + "learning_rate": 1.003140566516172e-05, + "loss": 0.5512, + "step": 28904 + }, + { + "epoch": 0.86, + "learning_rate": 1.0027196808761553e-05, + "loss": 0.5239, + "step": 28905 + }, + { + "epoch": 0.86, + "learning_rate": 1.002298878889264e-05, + "loss": 0.4733, + "step": 28906 + }, + { + "epoch": 0.86, + "learning_rate": 1.0018781605594108e-05, + "loss": 0.5287, + "step": 28907 + }, + { + "epoch": 0.86, + "learning_rate": 1.0014575258905067e-05, + "loss": 0.5007, + "step": 28908 + }, + { + "epoch": 0.86, + "learning_rate": 1.0010369748864656e-05, + "loss": 0.446, + "step": 28909 + }, + { + "epoch": 0.86, + "learning_rate": 1.0006165075511964e-05, + "loss": 0.5597, + "step": 28910 + }, + { + "epoch": 0.86, + "learning_rate": 1.0001961238886093e-05, + "loss": 0.513, + "step": 28911 + }, + { + "epoch": 0.86, + "learning_rate": 9.997758239026089e-06, + "loss": 0.5386, + "step": 28912 + }, + { + "epoch": 0.86, + "learning_rate": 9.993556075971066e-06, + "loss": 0.508, + "step": 28913 + }, + { + "epoch": 0.86, + "learning_rate": 9.989354749760104e-06, + "loss": 0.5844, + "step": 28914 + }, + { + "epoch": 0.86, + "learning_rate": 9.98515426043224e-06, + "loss": 0.5554, + "step": 28915 + }, + { + "epoch": 0.86, + "learning_rate": 9.980954608026527e-06, + "loss": 0.5577, + "step": 28916 + }, + { + "epoch": 0.86, + "learning_rate": 9.976755792582004e-06, + "loss": 0.5781, + "step": 28917 + }, + { + "epoch": 0.86, + "learning_rate": 9.972557814137761e-06, + "loss": 0.5255, + "step": 28918 + }, + { + "epoch": 0.86, + "learning_rate": 9.968360672732769e-06, + "loss": 0.4989, + "step": 28919 + }, + { + "epoch": 0.86, + "learning_rate": 9.964164368406093e-06, + "loss": 0.5953, + "step": 28920 + }, + { + "epoch": 0.86, + "learning_rate": 9.959968901196737e-06, + "loss": 0.489, + "step": 28921 + }, + { + "epoch": 0.86, + "learning_rate": 9.95577427114368e-06, + "loss": 0.5289, + "step": 28922 + }, + { + "epoch": 0.86, + "learning_rate": 9.95158047828595e-06, + "loss": 0.6466, + "step": 28923 + }, + { + "epoch": 0.86, + "learning_rate": 9.947387522662544e-06, + "loss": 0.3703, + "step": 28924 + }, + { + "epoch": 0.86, + "learning_rate": 9.943195404312444e-06, + "loss": 0.4713, + "step": 28925 + }, + { + "epoch": 0.86, + "learning_rate": 9.939004123274609e-06, + "loss": 0.5504, + "step": 28926 + }, + { + "epoch": 0.86, + "learning_rate": 9.934813679588007e-06, + "loss": 0.5283, + "step": 28927 + }, + { + "epoch": 0.86, + "learning_rate": 9.930624073291616e-06, + "loss": 0.5473, + "step": 28928 + }, + { + "epoch": 0.86, + "learning_rate": 9.926435304424398e-06, + "loss": 0.461, + "step": 28929 + }, + { + "epoch": 0.86, + "learning_rate": 9.922247373025295e-06, + "loss": 0.6901, + "step": 28930 + }, + { + "epoch": 0.86, + "learning_rate": 9.918060279133212e-06, + "loss": 0.5047, + "step": 28931 + }, + { + "epoch": 0.86, + "learning_rate": 9.913874022787094e-06, + "loss": 0.6387, + "step": 28932 + }, + { + "epoch": 0.86, + "learning_rate": 9.909688604025891e-06, + "loss": 0.5029, + "step": 28933 + }, + { + "epoch": 0.86, + "learning_rate": 9.905504022888502e-06, + "loss": 0.4915, + "step": 28934 + }, + { + "epoch": 0.86, + "learning_rate": 9.901320279413817e-06, + "loss": 0.47, + "step": 28935 + }, + { + "epoch": 0.86, + "learning_rate": 9.897137373640731e-06, + "loss": 0.6552, + "step": 28936 + }, + { + "epoch": 0.86, + "learning_rate": 9.892955305608176e-06, + "loss": 0.4993, + "step": 28937 + }, + { + "epoch": 0.86, + "learning_rate": 9.88877407535499e-06, + "loss": 0.6118, + "step": 28938 + }, + { + "epoch": 0.86, + "learning_rate": 9.884593682920084e-06, + "loss": 0.5183, + "step": 28939 + }, + { + "epoch": 0.86, + "learning_rate": 9.880414128342296e-06, + "loss": 0.5176, + "step": 28940 + }, + { + "epoch": 0.86, + "learning_rate": 9.876235411660517e-06, + "loss": 0.4166, + "step": 28941 + }, + { + "epoch": 0.86, + "learning_rate": 9.87205753291356e-06, + "loss": 0.5694, + "step": 28942 + }, + { + "epoch": 0.86, + "learning_rate": 9.867880492140313e-06, + "loss": 0.5896, + "step": 28943 + }, + { + "epoch": 0.86, + "learning_rate": 9.863704289379571e-06, + "loss": 0.5219, + "step": 28944 + }, + { + "epoch": 0.86, + "learning_rate": 9.85952892467019e-06, + "loss": 0.4915, + "step": 28945 + }, + { + "epoch": 0.86, + "learning_rate": 9.855354398050975e-06, + "loss": 0.5231, + "step": 28946 + }, + { + "epoch": 0.86, + "learning_rate": 9.851180709560749e-06, + "loss": 0.4971, + "step": 28947 + }, + { + "epoch": 0.86, + "learning_rate": 9.847007859238323e-06, + "loss": 0.5944, + "step": 28948 + }, + { + "epoch": 0.86, + "learning_rate": 9.8428358471225e-06, + "loss": 0.481, + "step": 28949 + }, + { + "epoch": 0.86, + "learning_rate": 9.83866467325203e-06, + "loss": 0.5322, + "step": 28950 + }, + { + "epoch": 0.86, + "learning_rate": 9.834494337665713e-06, + "loss": 0.542, + "step": 28951 + }, + { + "epoch": 0.86, + "learning_rate": 9.830324840402338e-06, + "loss": 0.6031, + "step": 28952 + }, + { + "epoch": 0.86, + "learning_rate": 9.826156181500713e-06, + "loss": 0.6182, + "step": 28953 + }, + { + "epoch": 0.86, + "learning_rate": 9.82198836099949e-06, + "loss": 0.4903, + "step": 28954 + }, + { + "epoch": 0.86, + "learning_rate": 9.817821378937497e-06, + "loss": 0.4588, + "step": 28955 + }, + { + "epoch": 0.86, + "learning_rate": 9.813655235353458e-06, + "loss": 0.7003, + "step": 28956 + }, + { + "epoch": 0.86, + "learning_rate": 9.80948993028611e-06, + "loss": 0.5201, + "step": 28957 + }, + { + "epoch": 0.86, + "learning_rate": 9.805325463774184e-06, + "loss": 0.4669, + "step": 28958 + }, + { + "epoch": 0.86, + "learning_rate": 9.801161835856377e-06, + "loss": 0.4893, + "step": 28959 + }, + { + "epoch": 0.86, + "learning_rate": 9.796999046571432e-06, + "loss": 0.5335, + "step": 28960 + }, + { + "epoch": 0.86, + "learning_rate": 9.79283709595803e-06, + "loss": 0.5945, + "step": 28961 + }, + { + "epoch": 0.86, + "learning_rate": 9.788675984054863e-06, + "loss": 0.4683, + "step": 28962 + }, + { + "epoch": 0.86, + "learning_rate": 9.784515710900655e-06, + "loss": 0.5741, + "step": 28963 + }, + { + "epoch": 0.86, + "learning_rate": 9.780356276534075e-06, + "loss": 0.5729, + "step": 28964 + }, + { + "epoch": 0.86, + "learning_rate": 9.776197680993749e-06, + "loss": 0.5396, + "step": 28965 + }, + { + "epoch": 0.86, + "learning_rate": 9.772039924318387e-06, + "loss": 0.4975, + "step": 28966 + }, + { + "epoch": 0.86, + "learning_rate": 9.767883006546652e-06, + "loss": 0.4657, + "step": 28967 + }, + { + "epoch": 0.86, + "learning_rate": 9.763726927717188e-06, + "loss": 0.5015, + "step": 28968 + }, + { + "epoch": 0.86, + "learning_rate": 9.759571687868597e-06, + "loss": 0.4833, + "step": 28969 + }, + { + "epoch": 0.86, + "learning_rate": 9.755417287039536e-06, + "loss": 0.4812, + "step": 28970 + }, + { + "epoch": 0.86, + "learning_rate": 9.751263725268644e-06, + "loss": 0.5366, + "step": 28971 + }, + { + "epoch": 0.86, + "learning_rate": 9.747111002594555e-06, + "loss": 0.529, + "step": 28972 + }, + { + "epoch": 0.86, + "learning_rate": 9.74295911905585e-06, + "loss": 0.5519, + "step": 28973 + }, + { + "epoch": 0.86, + "learning_rate": 9.73880807469113e-06, + "loss": 0.5252, + "step": 28974 + }, + { + "epoch": 0.86, + "learning_rate": 9.734657869538988e-06, + "loss": 0.4659, + "step": 28975 + }, + { + "epoch": 0.86, + "learning_rate": 9.730508503638047e-06, + "loss": 0.5172, + "step": 28976 + }, + { + "epoch": 0.86, + "learning_rate": 9.726359977026855e-06, + "loss": 0.4641, + "step": 28977 + }, + { + "epoch": 0.86, + "learning_rate": 9.72221228974397e-06, + "loss": 0.4973, + "step": 28978 + }, + { + "epoch": 0.86, + "learning_rate": 9.718065441827973e-06, + "loss": 0.5144, + "step": 28979 + }, + { + "epoch": 0.86, + "learning_rate": 9.713919433317442e-06, + "loss": 0.4404, + "step": 28980 + }, + { + "epoch": 0.86, + "learning_rate": 9.709774264250881e-06, + "loss": 0.5888, + "step": 28981 + }, + { + "epoch": 0.86, + "learning_rate": 9.705629934666871e-06, + "loss": 0.4374, + "step": 28982 + }, + { + "epoch": 0.86, + "learning_rate": 9.701486444603913e-06, + "loss": 0.5577, + "step": 28983 + }, + { + "epoch": 0.86, + "learning_rate": 9.697343794100566e-06, + "loss": 0.4498, + "step": 28984 + }, + { + "epoch": 0.86, + "learning_rate": 9.6932019831953e-06, + "loss": 0.5054, + "step": 28985 + }, + { + "epoch": 0.86, + "learning_rate": 9.68906101192667e-06, + "loss": 0.5226, + "step": 28986 + }, + { + "epoch": 0.86, + "learning_rate": 9.684920880333159e-06, + "loss": 0.5297, + "step": 28987 + }, + { + "epoch": 0.86, + "learning_rate": 9.680781588453237e-06, + "loss": 0.5051, + "step": 28988 + }, + { + "epoch": 0.86, + "learning_rate": 9.676643136325403e-06, + "loss": 0.5352, + "step": 28989 + }, + { + "epoch": 0.86, + "learning_rate": 9.672505523988152e-06, + "loss": 0.4358, + "step": 28990 + }, + { + "epoch": 0.86, + "learning_rate": 9.668368751479951e-06, + "loss": 0.4996, + "step": 28991 + }, + { + "epoch": 0.86, + "learning_rate": 9.664232818839258e-06, + "loss": 0.5726, + "step": 28992 + }, + { + "epoch": 0.86, + "learning_rate": 9.66009772610451e-06, + "loss": 0.4889, + "step": 28993 + }, + { + "epoch": 0.86, + "learning_rate": 9.655963473314167e-06, + "loss": 0.4761, + "step": 28994 + }, + { + "epoch": 0.86, + "learning_rate": 9.651830060506694e-06, + "loss": 0.5608, + "step": 28995 + }, + { + "epoch": 0.86, + "learning_rate": 9.647697487720486e-06, + "loss": 0.5598, + "step": 28996 + }, + { + "epoch": 0.86, + "learning_rate": 9.643565754993955e-06, + "loss": 0.4809, + "step": 28997 + }, + { + "epoch": 0.86, + "learning_rate": 9.639434862365549e-06, + "loss": 0.5807, + "step": 28998 + }, + { + "epoch": 0.86, + "learning_rate": 9.63530480987367e-06, + "loss": 0.5619, + "step": 28999 + }, + { + "epoch": 0.86, + "learning_rate": 9.631175597556696e-06, + "loss": 0.5307, + "step": 29000 + }, + { + "epoch": 0.86, + "learning_rate": 9.627047225453057e-06, + "loss": 0.455, + "step": 29001 + }, + { + "epoch": 0.86, + "learning_rate": 9.622919693601085e-06, + "loss": 0.4482, + "step": 29002 + }, + { + "epoch": 0.86, + "learning_rate": 9.618793002039218e-06, + "loss": 0.4548, + "step": 29003 + }, + { + "epoch": 0.86, + "learning_rate": 9.614667150805755e-06, + "loss": 0.5293, + "step": 29004 + }, + { + "epoch": 0.86, + "learning_rate": 9.610542139939127e-06, + "loss": 0.5778, + "step": 29005 + }, + { + "epoch": 0.86, + "learning_rate": 9.606417969477621e-06, + "loss": 0.6565, + "step": 29006 + }, + { + "epoch": 0.86, + "learning_rate": 9.60229463945964e-06, + "loss": 0.5374, + "step": 29007 + }, + { + "epoch": 0.86, + "learning_rate": 9.598172149923479e-06, + "loss": 0.5726, + "step": 29008 + }, + { + "epoch": 0.86, + "learning_rate": 9.59405050090748e-06, + "loss": 0.6869, + "step": 29009 + }, + { + "epoch": 0.86, + "learning_rate": 9.589929692449984e-06, + "loss": 0.5044, + "step": 29010 + }, + { + "epoch": 0.86, + "learning_rate": 9.58580972458929e-06, + "loss": 0.5172, + "step": 29011 + }, + { + "epoch": 0.86, + "learning_rate": 9.58169059736368e-06, + "loss": 0.5416, + "step": 29012 + }, + { + "epoch": 0.86, + "learning_rate": 9.577572310811477e-06, + "loss": 0.6145, + "step": 29013 + }, + { + "epoch": 0.86, + "learning_rate": 9.573454864970965e-06, + "loss": 0.6323, + "step": 29014 + }, + { + "epoch": 0.86, + "learning_rate": 9.569338259880468e-06, + "loss": 0.5726, + "step": 29015 + }, + { + "epoch": 0.86, + "learning_rate": 9.565222495578174e-06, + "loss": 0.5571, + "step": 29016 + }, + { + "epoch": 0.86, + "learning_rate": 9.561107572102412e-06, + "loss": 0.5221, + "step": 29017 + }, + { + "epoch": 0.86, + "learning_rate": 9.556993489491417e-06, + "loss": 0.4668, + "step": 29018 + }, + { + "epoch": 0.86, + "learning_rate": 9.55288024778348e-06, + "loss": 0.4255, + "step": 29019 + }, + { + "epoch": 0.86, + "learning_rate": 9.548767847016803e-06, + "loss": 0.5388, + "step": 29020 + }, + { + "epoch": 0.86, + "learning_rate": 9.544656287229614e-06, + "loss": 0.507, + "step": 29021 + }, + { + "epoch": 0.86, + "learning_rate": 9.54054556846018e-06, + "loss": 0.6046, + "step": 29022 + }, + { + "epoch": 0.86, + "learning_rate": 9.536435690746681e-06, + "loss": 0.5502, + "step": 29023 + }, + { + "epoch": 0.86, + "learning_rate": 9.532326654127354e-06, + "loss": 0.5363, + "step": 29024 + }, + { + "epoch": 0.86, + "learning_rate": 9.528218458640415e-06, + "loss": 0.5595, + "step": 29025 + }, + { + "epoch": 0.86, + "learning_rate": 9.52411110432403e-06, + "loss": 0.5026, + "step": 29026 + }, + { + "epoch": 0.86, + "learning_rate": 9.520004591216391e-06, + "loss": 0.4694, + "step": 29027 + }, + { + "epoch": 0.86, + "learning_rate": 9.51589891935568e-06, + "loss": 0.499, + "step": 29028 + }, + { + "epoch": 0.86, + "learning_rate": 9.511794088780101e-06, + "loss": 0.5472, + "step": 29029 + }, + { + "epoch": 0.86, + "learning_rate": 9.5076900995278e-06, + "loss": 0.4706, + "step": 29030 + }, + { + "epoch": 0.86, + "learning_rate": 9.503586951636912e-06, + "loss": 0.491, + "step": 29031 + }, + { + "epoch": 0.86, + "learning_rate": 9.499484645145596e-06, + "loss": 0.5602, + "step": 29032 + }, + { + "epoch": 0.86, + "learning_rate": 9.495383180091999e-06, + "loss": 0.4469, + "step": 29033 + }, + { + "epoch": 0.86, + "learning_rate": 9.491282556514281e-06, + "loss": 0.4776, + "step": 29034 + }, + { + "epoch": 0.86, + "learning_rate": 9.487182774450554e-06, + "loss": 0.5426, + "step": 29035 + }, + { + "epoch": 0.86, + "learning_rate": 9.4830838339389e-06, + "loss": 0.441, + "step": 29036 + }, + { + "epoch": 0.86, + "learning_rate": 9.478985735017454e-06, + "loss": 0.5838, + "step": 29037 + }, + { + "epoch": 0.86, + "learning_rate": 9.474888477724341e-06, + "loss": 0.4943, + "step": 29038 + }, + { + "epoch": 0.86, + "learning_rate": 9.470792062097633e-06, + "loss": 0.5723, + "step": 29039 + }, + { + "epoch": 0.86, + "learning_rate": 9.466696488175397e-06, + "loss": 0.5591, + "step": 29040 + }, + { + "epoch": 0.86, + "learning_rate": 9.462601755995748e-06, + "loss": 0.4046, + "step": 29041 + }, + { + "epoch": 0.86, + "learning_rate": 9.458507865596755e-06, + "loss": 0.4634, + "step": 29042 + }, + { + "epoch": 0.86, + "learning_rate": 9.454414817016443e-06, + "loss": 0.4924, + "step": 29043 + }, + { + "epoch": 0.86, + "learning_rate": 9.450322610292917e-06, + "loss": 0.5513, + "step": 29044 + }, + { + "epoch": 0.86, + "learning_rate": 9.446231245464188e-06, + "loss": 0.4532, + "step": 29045 + }, + { + "epoch": 0.86, + "learning_rate": 9.442140722568327e-06, + "loss": 0.4838, + "step": 29046 + }, + { + "epoch": 0.86, + "learning_rate": 9.438051041643326e-06, + "loss": 0.4681, + "step": 29047 + }, + { + "epoch": 0.86, + "learning_rate": 9.433962202727254e-06, + "loss": 0.4908, + "step": 29048 + }, + { + "epoch": 0.86, + "learning_rate": 9.429874205858091e-06, + "loss": 0.3814, + "step": 29049 + }, + { + "epoch": 0.86, + "learning_rate": 9.425787051073876e-06, + "loss": 0.5277, + "step": 29050 + }, + { + "epoch": 0.86, + "learning_rate": 9.421700738412564e-06, + "loss": 0.5463, + "step": 29051 + }, + { + "epoch": 0.86, + "learning_rate": 9.417615267912195e-06, + "loss": 0.4822, + "step": 29052 + }, + { + "epoch": 0.86, + "learning_rate": 9.413530639610746e-06, + "loss": 0.6774, + "step": 29053 + }, + { + "epoch": 0.86, + "learning_rate": 9.409446853546178e-06, + "loss": 0.4882, + "step": 29054 + }, + { + "epoch": 0.87, + "learning_rate": 9.405363909756459e-06, + "loss": 0.7056, + "step": 29055 + }, + { + "epoch": 0.87, + "learning_rate": 9.401281808279561e-06, + "loss": 0.4876, + "step": 29056 + }, + { + "epoch": 0.87, + "learning_rate": 9.39720054915344e-06, + "loss": 0.5087, + "step": 29057 + }, + { + "epoch": 0.87, + "learning_rate": 9.393120132416055e-06, + "loss": 0.5079, + "step": 29058 + }, + { + "epoch": 0.87, + "learning_rate": 9.3890405581053e-06, + "loss": 0.5054, + "step": 29059 + }, + { + "epoch": 0.87, + "learning_rate": 9.384961826259131e-06, + "loss": 0.552, + "step": 29060 + }, + { + "epoch": 0.87, + "learning_rate": 9.380883936915486e-06, + "loss": 0.4995, + "step": 29061 + }, + { + "epoch": 0.87, + "learning_rate": 9.376806890112255e-06, + "loss": 0.5082, + "step": 29062 + }, + { + "epoch": 0.87, + "learning_rate": 9.372730685887365e-06, + "loss": 0.3732, + "step": 29063 + }, + { + "epoch": 0.87, + "learning_rate": 9.368655324278697e-06, + "loss": 0.5814, + "step": 29064 + }, + { + "epoch": 0.87, + "learning_rate": 9.364580805324163e-06, + "loss": 0.5435, + "step": 29065 + }, + { + "epoch": 0.87, + "learning_rate": 9.360507129061613e-06, + "loss": 0.4222, + "step": 29066 + }, + { + "epoch": 0.87, + "learning_rate": 9.356434295528937e-06, + "loss": 0.5471, + "step": 29067 + }, + { + "epoch": 0.87, + "learning_rate": 9.352362304764028e-06, + "loss": 0.5144, + "step": 29068 + }, + { + "epoch": 0.87, + "learning_rate": 9.348291156804733e-06, + "loss": 0.5016, + "step": 29069 + }, + { + "epoch": 0.87, + "learning_rate": 9.344220851688878e-06, + "loss": 0.5868, + "step": 29070 + }, + { + "epoch": 0.87, + "learning_rate": 9.340151389454321e-06, + "loss": 0.5433, + "step": 29071 + }, + { + "epoch": 0.87, + "learning_rate": 9.336082770138921e-06, + "loss": 0.5089, + "step": 29072 + }, + { + "epoch": 0.87, + "learning_rate": 9.332014993780492e-06, + "loss": 0.6025, + "step": 29073 + }, + { + "epoch": 0.87, + "learning_rate": 9.327948060416836e-06, + "loss": 0.505, + "step": 29074 + }, + { + "epoch": 0.87, + "learning_rate": 9.323881970085767e-06, + "loss": 0.5044, + "step": 29075 + }, + { + "epoch": 0.87, + "learning_rate": 9.319816722825114e-06, + "loss": 0.5664, + "step": 29076 + }, + { + "epoch": 0.87, + "learning_rate": 9.315752318672677e-06, + "loss": 0.522, + "step": 29077 + }, + { + "epoch": 0.87, + "learning_rate": 9.311688757666236e-06, + "loss": 0.507, + "step": 29078 + }, + { + "epoch": 0.87, + "learning_rate": 9.307626039843554e-06, + "loss": 0.4792, + "step": 29079 + }, + { + "epoch": 0.87, + "learning_rate": 9.303564165242406e-06, + "loss": 0.5189, + "step": 29080 + }, + { + "epoch": 0.87, + "learning_rate": 9.2995031339006e-06, + "loss": 0.4797, + "step": 29081 + }, + { + "epoch": 0.87, + "learning_rate": 9.295442945855858e-06, + "loss": 0.526, + "step": 29082 + }, + { + "epoch": 0.87, + "learning_rate": 9.29138360114592e-06, + "loss": 0.5411, + "step": 29083 + }, + { + "epoch": 0.87, + "learning_rate": 9.287325099808552e-06, + "loss": 0.5771, + "step": 29084 + }, + { + "epoch": 0.87, + "learning_rate": 9.283267441881482e-06, + "loss": 0.5122, + "step": 29085 + }, + { + "epoch": 0.87, + "learning_rate": 9.279210627402435e-06, + "loss": 0.5122, + "step": 29086 + }, + { + "epoch": 0.87, + "learning_rate": 9.275154656409136e-06, + "loss": 0.5485, + "step": 29087 + }, + { + "epoch": 0.87, + "learning_rate": 9.271099528939298e-06, + "loss": 0.4363, + "step": 29088 + }, + { + "epoch": 0.87, + "learning_rate": 9.267045245030592e-06, + "loss": 0.4192, + "step": 29089 + }, + { + "epoch": 0.87, + "learning_rate": 9.262991804720734e-06, + "loss": 0.7049, + "step": 29090 + }, + { + "epoch": 0.87, + "learning_rate": 9.258939208047435e-06, + "loss": 0.4923, + "step": 29091 + }, + { + "epoch": 0.87, + "learning_rate": 9.254887455048355e-06, + "loss": 0.4281, + "step": 29092 + }, + { + "epoch": 0.87, + "learning_rate": 9.250836545761144e-06, + "loss": 0.6224, + "step": 29093 + }, + { + "epoch": 0.87, + "learning_rate": 9.246786480223479e-06, + "loss": 0.5717, + "step": 29094 + }, + { + "epoch": 0.87, + "learning_rate": 9.242737258473034e-06, + "loss": 0.5458, + "step": 29095 + }, + { + "epoch": 0.87, + "learning_rate": 9.238688880547463e-06, + "loss": 0.5622, + "step": 29096 + }, + { + "epoch": 0.87, + "learning_rate": 9.234641346484385e-06, + "loss": 0.5625, + "step": 29097 + }, + { + "epoch": 0.87, + "learning_rate": 9.230594656321422e-06, + "loss": 0.5478, + "step": 29098 + }, + { + "epoch": 0.87, + "learning_rate": 9.226548810096202e-06, + "loss": 0.5678, + "step": 29099 + }, + { + "epoch": 0.87, + "learning_rate": 9.222503807846383e-06, + "loss": 0.5497, + "step": 29100 + }, + { + "epoch": 0.87, + "learning_rate": 9.218459649609535e-06, + "loss": 0.4437, + "step": 29101 + }, + { + "epoch": 0.87, + "learning_rate": 9.21441633542326e-06, + "loss": 0.4863, + "step": 29102 + }, + { + "epoch": 0.87, + "learning_rate": 9.210373865325151e-06, + "loss": 0.6434, + "step": 29103 + }, + { + "epoch": 0.87, + "learning_rate": 9.206332239352811e-06, + "loss": 0.6411, + "step": 29104 + }, + { + "epoch": 0.87, + "learning_rate": 9.202291457543798e-06, + "loss": 0.4781, + "step": 29105 + }, + { + "epoch": 0.87, + "learning_rate": 9.198251519935708e-06, + "loss": 0.5034, + "step": 29106 + }, + { + "epoch": 0.87, + "learning_rate": 9.194212426566062e-06, + "loss": 0.6257, + "step": 29107 + }, + { + "epoch": 0.87, + "learning_rate": 9.190174177472465e-06, + "loss": 0.4471, + "step": 29108 + }, + { + "epoch": 0.87, + "learning_rate": 9.186136772692411e-06, + "loss": 0.5841, + "step": 29109 + }, + { + "epoch": 0.87, + "learning_rate": 9.182100212263478e-06, + "loss": 0.4465, + "step": 29110 + }, + { + "epoch": 0.87, + "learning_rate": 9.17806449622317e-06, + "loss": 0.4453, + "step": 29111 + }, + { + "epoch": 0.87, + "learning_rate": 9.174029624609038e-06, + "loss": 0.4574, + "step": 29112 + }, + { + "epoch": 0.87, + "learning_rate": 9.169995597458559e-06, + "loss": 0.5228, + "step": 29113 + }, + { + "epoch": 0.87, + "learning_rate": 9.165962414809259e-06, + "loss": 0.5591, + "step": 29114 + }, + { + "epoch": 0.87, + "learning_rate": 9.161930076698655e-06, + "loss": 0.4282, + "step": 29115 + }, + { + "epoch": 0.87, + "learning_rate": 9.157898583164216e-06, + "loss": 0.6554, + "step": 29116 + }, + { + "epoch": 0.87, + "learning_rate": 9.15386793424342e-06, + "loss": 0.4502, + "step": 29117 + }, + { + "epoch": 0.87, + "learning_rate": 9.149838129973742e-06, + "loss": 0.547, + "step": 29118 + }, + { + "epoch": 0.87, + "learning_rate": 9.145809170392683e-06, + "loss": 0.4467, + "step": 29119 + }, + { + "epoch": 0.87, + "learning_rate": 9.14178105553769e-06, + "loss": 0.485, + "step": 29120 + }, + { + "epoch": 0.87, + "learning_rate": 9.137753785446168e-06, + "loss": 0.4974, + "step": 29121 + }, + { + "epoch": 0.87, + "learning_rate": 9.133727360155608e-06, + "loss": 0.4848, + "step": 29122 + }, + { + "epoch": 0.87, + "learning_rate": 9.129701779703458e-06, + "loss": 0.4974, + "step": 29123 + }, + { + "epoch": 0.87, + "learning_rate": 9.12567704412709e-06, + "loss": 0.4724, + "step": 29124 + }, + { + "epoch": 0.87, + "learning_rate": 9.121653153463994e-06, + "loss": 0.5228, + "step": 29125 + }, + { + "epoch": 0.87, + "learning_rate": 9.117630107751518e-06, + "loss": 0.5995, + "step": 29126 + }, + { + "epoch": 0.87, + "learning_rate": 9.113607907027111e-06, + "loss": 0.4779, + "step": 29127 + }, + { + "epoch": 0.87, + "learning_rate": 9.109586551328141e-06, + "loss": 0.5909, + "step": 29128 + }, + { + "epoch": 0.87, + "learning_rate": 9.105566040692003e-06, + "loss": 0.5681, + "step": 29129 + }, + { + "epoch": 0.87, + "learning_rate": 9.101546375156112e-06, + "loss": 0.4895, + "step": 29130 + }, + { + "epoch": 0.87, + "learning_rate": 9.097527554757813e-06, + "loss": 0.4863, + "step": 29131 + }, + { + "epoch": 0.87, + "learning_rate": 9.093509579534454e-06, + "loss": 0.5598, + "step": 29132 + }, + { + "epoch": 0.87, + "learning_rate": 9.089492449523407e-06, + "loss": 0.5044, + "step": 29133 + }, + { + "epoch": 0.87, + "learning_rate": 9.08547616476204e-06, + "loss": 0.6423, + "step": 29134 + }, + { + "epoch": 0.87, + "learning_rate": 9.081460725287694e-06, + "loss": 0.5874, + "step": 29135 + }, + { + "epoch": 0.87, + "learning_rate": 9.077446131137668e-06, + "loss": 0.5831, + "step": 29136 + }, + { + "epoch": 0.87, + "learning_rate": 9.073432382349301e-06, + "loss": 0.4445, + "step": 29137 + }, + { + "epoch": 0.87, + "learning_rate": 9.069419478959929e-06, + "loss": 0.5312, + "step": 29138 + }, + { + "epoch": 0.87, + "learning_rate": 9.065407421006866e-06, + "loss": 0.588, + "step": 29139 + }, + { + "epoch": 0.87, + "learning_rate": 9.061396208527406e-06, + "loss": 0.461, + "step": 29140 + }, + { + "epoch": 0.87, + "learning_rate": 9.057385841558819e-06, + "loss": 0.4426, + "step": 29141 + }, + { + "epoch": 0.87, + "learning_rate": 9.053376320138408e-06, + "loss": 0.5047, + "step": 29142 + }, + { + "epoch": 0.87, + "learning_rate": 9.049367644303475e-06, + "loss": 0.4601, + "step": 29143 + }, + { + "epoch": 0.87, + "learning_rate": 9.045359814091281e-06, + "loss": 0.5057, + "step": 29144 + }, + { + "epoch": 0.87, + "learning_rate": 9.041352829539052e-06, + "loss": 0.4801, + "step": 29145 + }, + { + "epoch": 0.87, + "learning_rate": 9.037346690684079e-06, + "loss": 0.4475, + "step": 29146 + }, + { + "epoch": 0.87, + "learning_rate": 9.033341397563622e-06, + "loss": 0.43, + "step": 29147 + }, + { + "epoch": 0.87, + "learning_rate": 9.029336950214872e-06, + "loss": 0.6683, + "step": 29148 + }, + { + "epoch": 0.87, + "learning_rate": 9.025333348675124e-06, + "loss": 0.5694, + "step": 29149 + }, + { + "epoch": 0.87, + "learning_rate": 9.021330592981535e-06, + "loss": 0.558, + "step": 29150 + }, + { + "epoch": 0.87, + "learning_rate": 9.017328683171378e-06, + "loss": 0.5394, + "step": 29151 + }, + { + "epoch": 0.87, + "learning_rate": 9.01332761928182e-06, + "loss": 0.449, + "step": 29152 + }, + { + "epoch": 0.87, + "learning_rate": 9.009327401350098e-06, + "loss": 0.3761, + "step": 29153 + }, + { + "epoch": 0.87, + "learning_rate": 9.005328029413373e-06, + "loss": 0.5009, + "step": 29154 + }, + { + "epoch": 0.87, + "learning_rate": 9.001329503508838e-06, + "loss": 0.4889, + "step": 29155 + }, + { + "epoch": 0.87, + "learning_rate": 8.99733182367366e-06, + "loss": 0.5537, + "step": 29156 + }, + { + "epoch": 0.87, + "learning_rate": 8.993334989945024e-06, + "loss": 0.5264, + "step": 29157 + }, + { + "epoch": 0.87, + "learning_rate": 8.989339002360109e-06, + "loss": 0.508, + "step": 29158 + }, + { + "epoch": 0.87, + "learning_rate": 8.985343860956052e-06, + "loss": 0.5453, + "step": 29159 + }, + { + "epoch": 0.87, + "learning_rate": 8.981349565769969e-06, + "loss": 0.5439, + "step": 29160 + }, + { + "epoch": 0.87, + "learning_rate": 8.97735611683902e-06, + "loss": 0.4198, + "step": 29161 + }, + { + "epoch": 0.87, + "learning_rate": 8.97336351420035e-06, + "loss": 0.4555, + "step": 29162 + }, + { + "epoch": 0.87, + "learning_rate": 8.969371757891076e-06, + "loss": 0.4305, + "step": 29163 + }, + { + "epoch": 0.87, + "learning_rate": 8.965380847948279e-06, + "loss": 0.4624, + "step": 29164 + }, + { + "epoch": 0.87, + "learning_rate": 8.961390784409085e-06, + "loss": 0.5791, + "step": 29165 + }, + { + "epoch": 0.87, + "learning_rate": 8.957401567310619e-06, + "loss": 0.5241, + "step": 29166 + }, + { + "epoch": 0.87, + "learning_rate": 8.953413196689919e-06, + "loss": 0.5964, + "step": 29167 + }, + { + "epoch": 0.87, + "learning_rate": 8.949425672584123e-06, + "loss": 0.4397, + "step": 29168 + }, + { + "epoch": 0.87, + "learning_rate": 8.945438995030264e-06, + "loss": 0.5307, + "step": 29169 + }, + { + "epoch": 0.87, + "learning_rate": 8.941453164065427e-06, + "loss": 0.4776, + "step": 29170 + }, + { + "epoch": 0.87, + "learning_rate": 8.93746817972666e-06, + "loss": 0.479, + "step": 29171 + }, + { + "epoch": 0.87, + "learning_rate": 8.93348404205101e-06, + "loss": 0.6058, + "step": 29172 + }, + { + "epoch": 0.87, + "learning_rate": 8.92950075107556e-06, + "loss": 0.5297, + "step": 29173 + }, + { + "epoch": 0.87, + "learning_rate": 8.925518306837311e-06, + "loss": 0.4444, + "step": 29174 + }, + { + "epoch": 0.87, + "learning_rate": 8.921536709373269e-06, + "loss": 0.6472, + "step": 29175 + }, + { + "epoch": 0.87, + "learning_rate": 8.91755595872048e-06, + "loss": 0.6419, + "step": 29176 + }, + { + "epoch": 0.87, + "learning_rate": 8.913576054915973e-06, + "loss": 0.5427, + "step": 29177 + }, + { + "epoch": 0.87, + "learning_rate": 8.909596997996739e-06, + "loss": 0.4526, + "step": 29178 + }, + { + "epoch": 0.87, + "learning_rate": 8.905618787999747e-06, + "loss": 0.564, + "step": 29179 + }, + { + "epoch": 0.87, + "learning_rate": 8.901641424961993e-06, + "loss": 0.4924, + "step": 29180 + }, + { + "epoch": 0.87, + "learning_rate": 8.897664908920477e-06, + "loss": 0.5881, + "step": 29181 + }, + { + "epoch": 0.87, + "learning_rate": 8.893689239912207e-06, + "loss": 0.5719, + "step": 29182 + }, + { + "epoch": 0.87, + "learning_rate": 8.88971441797406e-06, + "loss": 0.5019, + "step": 29183 + }, + { + "epoch": 0.87, + "learning_rate": 8.885740443143031e-06, + "loss": 0.6048, + "step": 29184 + }, + { + "epoch": 0.87, + "learning_rate": 8.881767315456091e-06, + "loss": 0.4407, + "step": 29185 + }, + { + "epoch": 0.87, + "learning_rate": 8.877795034950154e-06, + "loss": 0.6564, + "step": 29186 + }, + { + "epoch": 0.87, + "learning_rate": 8.873823601662167e-06, + "loss": 0.5576, + "step": 29187 + }, + { + "epoch": 0.87, + "learning_rate": 8.869853015629038e-06, + "loss": 0.6457, + "step": 29188 + }, + { + "epoch": 0.87, + "learning_rate": 8.86588327688771e-06, + "loss": 0.4855, + "step": 29189 + }, + { + "epoch": 0.87, + "learning_rate": 8.861914385475068e-06, + "loss": 0.4774, + "step": 29190 + }, + { + "epoch": 0.87, + "learning_rate": 8.857946341428013e-06, + "loss": 0.4284, + "step": 29191 + }, + { + "epoch": 0.87, + "learning_rate": 8.853979144783465e-06, + "loss": 0.6627, + "step": 29192 + }, + { + "epoch": 0.87, + "learning_rate": 8.850012795578289e-06, + "loss": 0.529, + "step": 29193 + }, + { + "epoch": 0.87, + "learning_rate": 8.846047293849358e-06, + "loss": 0.4529, + "step": 29194 + }, + { + "epoch": 0.87, + "learning_rate": 8.842082639633542e-06, + "loss": 0.6291, + "step": 29195 + }, + { + "epoch": 0.87, + "learning_rate": 8.838118832967723e-06, + "loss": 0.4412, + "step": 29196 + }, + { + "epoch": 0.87, + "learning_rate": 8.83415587388875e-06, + "loss": 0.6274, + "step": 29197 + }, + { + "epoch": 0.87, + "learning_rate": 8.830193762433436e-06, + "loss": 0.5405, + "step": 29198 + }, + { + "epoch": 0.87, + "learning_rate": 8.826232498638653e-06, + "loss": 0.4988, + "step": 29199 + }, + { + "epoch": 0.87, + "learning_rate": 8.822272082541216e-06, + "loss": 0.592, + "step": 29200 + }, + { + "epoch": 0.87, + "learning_rate": 8.818312514177974e-06, + "loss": 0.5495, + "step": 29201 + }, + { + "epoch": 0.87, + "learning_rate": 8.814353793585717e-06, + "loss": 0.5896, + "step": 29202 + }, + { + "epoch": 0.87, + "learning_rate": 8.81039592080124e-06, + "loss": 0.6025, + "step": 29203 + }, + { + "epoch": 0.87, + "learning_rate": 8.806438895861347e-06, + "loss": 0.4085, + "step": 29204 + }, + { + "epoch": 0.87, + "learning_rate": 8.802482718802874e-06, + "loss": 0.6654, + "step": 29205 + }, + { + "epoch": 0.87, + "learning_rate": 8.798527389662558e-06, + "loss": 0.5869, + "step": 29206 + }, + { + "epoch": 0.87, + "learning_rate": 8.794572908477172e-06, + "loss": 0.5343, + "step": 29207 + }, + { + "epoch": 0.87, + "learning_rate": 8.790619275283484e-06, + "loss": 0.4609, + "step": 29208 + }, + { + "epoch": 0.87, + "learning_rate": 8.7866664901183e-06, + "loss": 0.5846, + "step": 29209 + }, + { + "epoch": 0.87, + "learning_rate": 8.782714553018312e-06, + "loss": 0.4913, + "step": 29210 + }, + { + "epoch": 0.87, + "learning_rate": 8.7787634640203e-06, + "loss": 0.6206, + "step": 29211 + }, + { + "epoch": 0.87, + "learning_rate": 8.774813223160982e-06, + "loss": 0.5059, + "step": 29212 + }, + { + "epoch": 0.87, + "learning_rate": 8.770863830477105e-06, + "loss": 0.5665, + "step": 29213 + }, + { + "epoch": 0.87, + "learning_rate": 8.766915286005362e-06, + "loss": 0.5127, + "step": 29214 + }, + { + "epoch": 0.87, + "learning_rate": 8.762967589782489e-06, + "loss": 0.5236, + "step": 29215 + }, + { + "epoch": 0.87, + "learning_rate": 8.759020741845159e-06, + "loss": 0.6018, + "step": 29216 + }, + { + "epoch": 0.87, + "learning_rate": 8.755074742230119e-06, + "loss": 0.5528, + "step": 29217 + }, + { + "epoch": 0.87, + "learning_rate": 8.751129590974006e-06, + "loss": 0.3834, + "step": 29218 + }, + { + "epoch": 0.87, + "learning_rate": 8.747185288113512e-06, + "loss": 0.545, + "step": 29219 + }, + { + "epoch": 0.87, + "learning_rate": 8.743241833685345e-06, + "loss": 0.5523, + "step": 29220 + }, + { + "epoch": 0.87, + "learning_rate": 8.73929922772615e-06, + "loss": 0.5293, + "step": 29221 + }, + { + "epoch": 0.87, + "learning_rate": 8.735357470272554e-06, + "loss": 0.4647, + "step": 29222 + }, + { + "epoch": 0.87, + "learning_rate": 8.731416561361228e-06, + "loss": 0.4912, + "step": 29223 + }, + { + "epoch": 0.87, + "learning_rate": 8.727476501028831e-06, + "loss": 0.3909, + "step": 29224 + }, + { + "epoch": 0.87, + "learning_rate": 8.723537289311989e-06, + "loss": 0.4221, + "step": 29225 + }, + { + "epoch": 0.87, + "learning_rate": 8.719598926247285e-06, + "loss": 0.5671, + "step": 29226 + }, + { + "epoch": 0.87, + "learning_rate": 8.715661411871379e-06, + "loss": 0.6507, + "step": 29227 + }, + { + "epoch": 0.87, + "learning_rate": 8.711724746220896e-06, + "loss": 0.5152, + "step": 29228 + }, + { + "epoch": 0.87, + "learning_rate": 8.707788929332383e-06, + "loss": 0.4869, + "step": 29229 + }, + { + "epoch": 0.87, + "learning_rate": 8.70385396124248e-06, + "loss": 0.5072, + "step": 29230 + }, + { + "epoch": 0.87, + "learning_rate": 8.699919841987748e-06, + "loss": 0.5256, + "step": 29231 + }, + { + "epoch": 0.87, + "learning_rate": 8.695986571604787e-06, + "loss": 0.5134, + "step": 29232 + }, + { + "epoch": 0.87, + "learning_rate": 8.692054150130136e-06, + "loss": 0.457, + "step": 29233 + }, + { + "epoch": 0.87, + "learning_rate": 8.688122577600367e-06, + "loss": 0.4968, + "step": 29234 + }, + { + "epoch": 0.87, + "learning_rate": 8.68419185405206e-06, + "loss": 0.5642, + "step": 29235 + }, + { + "epoch": 0.87, + "learning_rate": 8.680261979521742e-06, + "loss": 0.5621, + "step": 29236 + }, + { + "epoch": 0.87, + "learning_rate": 8.67633295404594e-06, + "loss": 0.518, + "step": 29237 + }, + { + "epoch": 0.87, + "learning_rate": 8.672404777661192e-06, + "loss": 0.592, + "step": 29238 + }, + { + "epoch": 0.87, + "learning_rate": 8.668477450404044e-06, + "loss": 0.4843, + "step": 29239 + }, + { + "epoch": 0.87, + "learning_rate": 8.664550972310992e-06, + "loss": 0.5249, + "step": 29240 + }, + { + "epoch": 0.87, + "learning_rate": 8.660625343418527e-06, + "loss": 0.7324, + "step": 29241 + }, + { + "epoch": 0.87, + "learning_rate": 8.656700563763154e-06, + "loss": 0.6295, + "step": 29242 + }, + { + "epoch": 0.87, + "learning_rate": 8.652776633381377e-06, + "loss": 0.5452, + "step": 29243 + }, + { + "epoch": 0.87, + "learning_rate": 8.6488535523097e-06, + "loss": 0.5323, + "step": 29244 + }, + { + "epoch": 0.87, + "learning_rate": 8.64493132058457e-06, + "loss": 0.5685, + "step": 29245 + }, + { + "epoch": 0.87, + "learning_rate": 8.64100993824244e-06, + "loss": 0.6475, + "step": 29246 + }, + { + "epoch": 0.87, + "learning_rate": 8.637089405319787e-06, + "loss": 0.544, + "step": 29247 + }, + { + "epoch": 0.87, + "learning_rate": 8.633169721853074e-06, + "loss": 0.4737, + "step": 29248 + }, + { + "epoch": 0.87, + "learning_rate": 8.629250887878748e-06, + "loss": 0.5262, + "step": 29249 + }, + { + "epoch": 0.87, + "learning_rate": 8.625332903433203e-06, + "loss": 0.6009, + "step": 29250 + }, + { + "epoch": 0.87, + "learning_rate": 8.62141576855291e-06, + "loss": 0.6469, + "step": 29251 + }, + { + "epoch": 0.87, + "learning_rate": 8.61749948327426e-06, + "loss": 0.4894, + "step": 29252 + }, + { + "epoch": 0.87, + "learning_rate": 8.613584047633683e-06, + "loss": 0.616, + "step": 29253 + }, + { + "epoch": 0.87, + "learning_rate": 8.609669461667592e-06, + "loss": 0.5098, + "step": 29254 + }, + { + "epoch": 0.87, + "learning_rate": 8.605755725412367e-06, + "loss": 0.5269, + "step": 29255 + }, + { + "epoch": 0.87, + "learning_rate": 8.601842838904383e-06, + "loss": 0.6228, + "step": 29256 + }, + { + "epoch": 0.87, + "learning_rate": 8.597930802180032e-06, + "loss": 0.4545, + "step": 29257 + }, + { + "epoch": 0.87, + "learning_rate": 8.594019615275717e-06, + "loss": 0.5808, + "step": 29258 + }, + { + "epoch": 0.87, + "learning_rate": 8.590109278227776e-06, + "loss": 0.4996, + "step": 29259 + }, + { + "epoch": 0.87, + "learning_rate": 8.586199791072547e-06, + "loss": 0.59, + "step": 29260 + }, + { + "epoch": 0.87, + "learning_rate": 8.5822911538464e-06, + "loss": 0.3883, + "step": 29261 + }, + { + "epoch": 0.87, + "learning_rate": 8.578383366585662e-06, + "loss": 0.5957, + "step": 29262 + }, + { + "epoch": 0.87, + "learning_rate": 8.574476429326716e-06, + "loss": 0.5951, + "step": 29263 + }, + { + "epoch": 0.87, + "learning_rate": 8.570570342105833e-06, + "loss": 0.5484, + "step": 29264 + }, + { + "epoch": 0.87, + "learning_rate": 8.566665104959337e-06, + "loss": 0.5047, + "step": 29265 + }, + { + "epoch": 0.87, + "learning_rate": 8.562760717923546e-06, + "loss": 0.4659, + "step": 29266 + }, + { + "epoch": 0.87, + "learning_rate": 8.558857181034786e-06, + "loss": 0.5566, + "step": 29267 + }, + { + "epoch": 0.87, + "learning_rate": 8.554954494329315e-06, + "loss": 0.4292, + "step": 29268 + }, + { + "epoch": 0.87, + "learning_rate": 8.551052657843418e-06, + "loss": 0.4496, + "step": 29269 + }, + { + "epoch": 0.87, + "learning_rate": 8.547151671613385e-06, + "loss": 0.4671, + "step": 29270 + }, + { + "epoch": 0.87, + "learning_rate": 8.543251535675512e-06, + "loss": 0.5576, + "step": 29271 + }, + { + "epoch": 0.87, + "learning_rate": 8.539352250066012e-06, + "loss": 0.5773, + "step": 29272 + }, + { + "epoch": 0.87, + "learning_rate": 8.535453814821182e-06, + "loss": 0.4813, + "step": 29273 + }, + { + "epoch": 0.87, + "learning_rate": 8.531556229977222e-06, + "loss": 0.5512, + "step": 29274 + }, + { + "epoch": 0.87, + "learning_rate": 8.527659495570416e-06, + "loss": 0.6156, + "step": 29275 + }, + { + "epoch": 0.87, + "learning_rate": 8.52376361163697e-06, + "loss": 0.5074, + "step": 29276 + }, + { + "epoch": 0.87, + "learning_rate": 8.51986857821312e-06, + "loss": 0.5688, + "step": 29277 + }, + { + "epoch": 0.87, + "learning_rate": 8.515974395335047e-06, + "loss": 0.6163, + "step": 29278 + }, + { + "epoch": 0.87, + "learning_rate": 8.512081063039012e-06, + "loss": 0.5224, + "step": 29279 + }, + { + "epoch": 0.87, + "learning_rate": 8.508188581361155e-06, + "loss": 0.3987, + "step": 29280 + }, + { + "epoch": 0.87, + "learning_rate": 8.504296950337699e-06, + "loss": 0.5349, + "step": 29281 + }, + { + "epoch": 0.87, + "learning_rate": 8.50040617000485e-06, + "loss": 0.5862, + "step": 29282 + }, + { + "epoch": 0.87, + "learning_rate": 8.496516240398744e-06, + "loss": 0.4622, + "step": 29283 + }, + { + "epoch": 0.87, + "learning_rate": 8.492627161555544e-06, + "loss": 0.4058, + "step": 29284 + }, + { + "epoch": 0.87, + "learning_rate": 8.488738933511431e-06, + "loss": 0.6945, + "step": 29285 + }, + { + "epoch": 0.87, + "learning_rate": 8.484851556302576e-06, + "loss": 0.5025, + "step": 29286 + }, + { + "epoch": 0.87, + "learning_rate": 8.480965029965083e-06, + "loss": 0.5009, + "step": 29287 + }, + { + "epoch": 0.87, + "learning_rate": 8.47707935453509e-06, + "loss": 0.4368, + "step": 29288 + }, + { + "epoch": 0.87, + "learning_rate": 8.473194530048733e-06, + "loss": 0.5713, + "step": 29289 + }, + { + "epoch": 0.87, + "learning_rate": 8.469310556542154e-06, + "loss": 0.648, + "step": 29290 + }, + { + "epoch": 0.87, + "learning_rate": 8.465427434051421e-06, + "loss": 0.5378, + "step": 29291 + }, + { + "epoch": 0.87, + "learning_rate": 8.461545162612682e-06, + "loss": 0.5606, + "step": 29292 + }, + { + "epoch": 0.87, + "learning_rate": 8.457663742262001e-06, + "loss": 0.6401, + "step": 29293 + }, + { + "epoch": 0.87, + "learning_rate": 8.453783173035489e-06, + "loss": 0.6451, + "step": 29294 + }, + { + "epoch": 0.87, + "learning_rate": 8.449903454969199e-06, + "loss": 0.4699, + "step": 29295 + }, + { + "epoch": 0.87, + "learning_rate": 8.446024588099221e-06, + "loss": 0.6082, + "step": 29296 + }, + { + "epoch": 0.87, + "learning_rate": 8.44214657246163e-06, + "loss": 0.4767, + "step": 29297 + }, + { + "epoch": 0.87, + "learning_rate": 8.43826940809247e-06, + "loss": 0.5145, + "step": 29298 + }, + { + "epoch": 0.87, + "learning_rate": 8.43439309502777e-06, + "loss": 0.5467, + "step": 29299 + }, + { + "epoch": 0.87, + "learning_rate": 8.430517633303592e-06, + "loss": 0.4788, + "step": 29300 + }, + { + "epoch": 0.87, + "learning_rate": 8.426643022955983e-06, + "loss": 0.6091, + "step": 29301 + }, + { + "epoch": 0.87, + "learning_rate": 8.422769264020946e-06, + "loss": 0.5349, + "step": 29302 + }, + { + "epoch": 0.87, + "learning_rate": 8.418896356534478e-06, + "loss": 0.5763, + "step": 29303 + }, + { + "epoch": 0.87, + "learning_rate": 8.415024300532615e-06, + "loss": 0.4401, + "step": 29304 + }, + { + "epoch": 0.87, + "learning_rate": 8.41115309605136e-06, + "loss": 0.5683, + "step": 29305 + }, + { + "epoch": 0.87, + "learning_rate": 8.40728274312671e-06, + "loss": 0.4988, + "step": 29306 + }, + { + "epoch": 0.87, + "learning_rate": 8.403413241794644e-06, + "loss": 0.5909, + "step": 29307 + }, + { + "epoch": 0.87, + "learning_rate": 8.399544592091124e-06, + "loss": 0.6624, + "step": 29308 + }, + { + "epoch": 0.87, + "learning_rate": 8.395676794052122e-06, + "loss": 0.5062, + "step": 29309 + }, + { + "epoch": 0.87, + "learning_rate": 8.391809847713627e-06, + "loss": 0.4058, + "step": 29310 + }, + { + "epoch": 0.87, + "learning_rate": 8.387943753111572e-06, + "loss": 0.4475, + "step": 29311 + }, + { + "epoch": 0.87, + "learning_rate": 8.384078510281878e-06, + "loss": 0.5773, + "step": 29312 + }, + { + "epoch": 0.87, + "learning_rate": 8.380214119260521e-06, + "loss": 0.5106, + "step": 29313 + }, + { + "epoch": 0.87, + "learning_rate": 8.376350580083426e-06, + "loss": 0.3993, + "step": 29314 + }, + { + "epoch": 0.87, + "learning_rate": 8.372487892786485e-06, + "loss": 0.4681, + "step": 29315 + }, + { + "epoch": 0.87, + "learning_rate": 8.36862605740566e-06, + "loss": 0.5148, + "step": 29316 + }, + { + "epoch": 0.87, + "learning_rate": 8.364765073976821e-06, + "loss": 0.5213, + "step": 29317 + }, + { + "epoch": 0.87, + "learning_rate": 8.360904942535864e-06, + "loss": 0.53, + "step": 29318 + }, + { + "epoch": 0.87, + "learning_rate": 8.357045663118679e-06, + "loss": 0.4867, + "step": 29319 + }, + { + "epoch": 0.87, + "learning_rate": 8.353187235761185e-06, + "loss": 0.451, + "step": 29320 + }, + { + "epoch": 0.87, + "learning_rate": 8.349329660499227e-06, + "loss": 0.5925, + "step": 29321 + }, + { + "epoch": 0.87, + "learning_rate": 8.345472937368649e-06, + "loss": 0.3901, + "step": 29322 + }, + { + "epoch": 0.87, + "learning_rate": 8.341617066405338e-06, + "loss": 0.4438, + "step": 29323 + }, + { + "epoch": 0.87, + "learning_rate": 8.337762047645149e-06, + "loss": 0.6211, + "step": 29324 + }, + { + "epoch": 0.87, + "learning_rate": 8.333907881123925e-06, + "loss": 0.5173, + "step": 29325 + }, + { + "epoch": 0.87, + "learning_rate": 8.330054566877487e-06, + "loss": 0.386, + "step": 29326 + }, + { + "epoch": 0.87, + "learning_rate": 8.32620210494166e-06, + "loss": 0.5673, + "step": 29327 + }, + { + "epoch": 0.87, + "learning_rate": 8.32235049535226e-06, + "loss": 0.5432, + "step": 29328 + }, + { + "epoch": 0.87, + "learning_rate": 8.318499738145124e-06, + "loss": 0.5012, + "step": 29329 + }, + { + "epoch": 0.87, + "learning_rate": 8.314649833356036e-06, + "loss": 0.477, + "step": 29330 + }, + { + "epoch": 0.87, + "learning_rate": 8.310800781020778e-06, + "loss": 0.5628, + "step": 29331 + }, + { + "epoch": 0.87, + "learning_rate": 8.306952581175142e-06, + "loss": 0.511, + "step": 29332 + }, + { + "epoch": 0.87, + "learning_rate": 8.303105233854947e-06, + "loss": 0.5065, + "step": 29333 + }, + { + "epoch": 0.87, + "learning_rate": 8.299258739095905e-06, + "loss": 0.5977, + "step": 29334 + }, + { + "epoch": 0.87, + "learning_rate": 8.295413096933824e-06, + "loss": 0.5228, + "step": 29335 + }, + { + "epoch": 0.87, + "learning_rate": 8.29156830740443e-06, + "loss": 0.5421, + "step": 29336 + }, + { + "epoch": 0.87, + "learning_rate": 8.287724370543493e-06, + "loss": 0.5273, + "step": 29337 + }, + { + "epoch": 0.87, + "learning_rate": 8.28388128638673e-06, + "loss": 0.6055, + "step": 29338 + }, + { + "epoch": 0.87, + "learning_rate": 8.28003905496989e-06, + "loss": 0.6021, + "step": 29339 + }, + { + "epoch": 0.87, + "learning_rate": 8.2761976763287e-06, + "loss": 0.5344, + "step": 29340 + }, + { + "epoch": 0.87, + "learning_rate": 8.272357150498866e-06, + "loss": 0.4453, + "step": 29341 + }, + { + "epoch": 0.87, + "learning_rate": 8.26851747751608e-06, + "loss": 0.5308, + "step": 29342 + }, + { + "epoch": 0.87, + "learning_rate": 8.264678657416059e-06, + "loss": 0.6412, + "step": 29343 + }, + { + "epoch": 0.87, + "learning_rate": 8.260840690234518e-06, + "loss": 0.4859, + "step": 29344 + }, + { + "epoch": 0.87, + "learning_rate": 8.257003576007106e-06, + "loss": 0.494, + "step": 29345 + }, + { + "epoch": 0.87, + "learning_rate": 8.253167314769494e-06, + "loss": 0.4403, + "step": 29346 + }, + { + "epoch": 0.87, + "learning_rate": 8.249331906557367e-06, + "loss": 0.5496, + "step": 29347 + }, + { + "epoch": 0.87, + "learning_rate": 8.245497351406407e-06, + "loss": 0.5398, + "step": 29348 + }, + { + "epoch": 0.87, + "learning_rate": 8.241663649352239e-06, + "loss": 0.5106, + "step": 29349 + }, + { + "epoch": 0.87, + "learning_rate": 8.23783080043049e-06, + "loss": 0.6501, + "step": 29350 + }, + { + "epoch": 0.87, + "learning_rate": 8.233998804676823e-06, + "loss": 0.4984, + "step": 29351 + }, + { + "epoch": 0.87, + "learning_rate": 8.230167662126876e-06, + "loss": 0.5875, + "step": 29352 + }, + { + "epoch": 0.87, + "learning_rate": 8.226337372816228e-06, + "loss": 0.4746, + "step": 29353 + }, + { + "epoch": 0.87, + "learning_rate": 8.222507936780544e-06, + "loss": 0.4681, + "step": 29354 + }, + { + "epoch": 0.87, + "learning_rate": 8.218679354055382e-06, + "loss": 0.4865, + "step": 29355 + }, + { + "epoch": 0.87, + "learning_rate": 8.21485162467638e-06, + "loss": 0.4635, + "step": 29356 + }, + { + "epoch": 0.87, + "learning_rate": 8.211024748679085e-06, + "loss": 0.5227, + "step": 29357 + }, + { + "epoch": 0.87, + "learning_rate": 8.207198726099109e-06, + "loss": 0.5251, + "step": 29358 + }, + { + "epoch": 0.87, + "learning_rate": 8.203373556972016e-06, + "loss": 0.5544, + "step": 29359 + }, + { + "epoch": 0.87, + "learning_rate": 8.199549241333382e-06, + "loss": 0.5267, + "step": 29360 + }, + { + "epoch": 0.87, + "learning_rate": 8.19572577921872e-06, + "loss": 0.5469, + "step": 29361 + }, + { + "epoch": 0.87, + "learning_rate": 8.191903170663617e-06, + "loss": 0.5835, + "step": 29362 + }, + { + "epoch": 0.87, + "learning_rate": 8.18808141570363e-06, + "loss": 0.4818, + "step": 29363 + }, + { + "epoch": 0.87, + "learning_rate": 8.184260514374265e-06, + "loss": 0.4812, + "step": 29364 + }, + { + "epoch": 0.87, + "learning_rate": 8.180440466711025e-06, + "loss": 0.5044, + "step": 29365 + }, + { + "epoch": 0.87, + "learning_rate": 8.176621272749474e-06, + "loss": 0.5253, + "step": 29366 + }, + { + "epoch": 0.87, + "learning_rate": 8.172802932525082e-06, + "loss": 0.5418, + "step": 29367 + }, + { + "epoch": 0.87, + "learning_rate": 8.168985446073396e-06, + "loss": 0.4975, + "step": 29368 + }, + { + "epoch": 0.87, + "learning_rate": 8.16516881342988e-06, + "loss": 0.491, + "step": 29369 + }, + { + "epoch": 0.87, + "learning_rate": 8.161353034630015e-06, + "loss": 0.5501, + "step": 29370 + }, + { + "epoch": 0.87, + "learning_rate": 8.157538109709273e-06, + "loss": 0.5745, + "step": 29371 + }, + { + "epoch": 0.87, + "learning_rate": 8.153724038703158e-06, + "loss": 0.4138, + "step": 29372 + }, + { + "epoch": 0.87, + "learning_rate": 8.149910821647111e-06, + "loss": 0.5537, + "step": 29373 + }, + { + "epoch": 0.87, + "learning_rate": 8.146098458576568e-06, + "loss": 0.4618, + "step": 29374 + }, + { + "epoch": 0.87, + "learning_rate": 8.142286949527e-06, + "loss": 0.5326, + "step": 29375 + }, + { + "epoch": 0.87, + "learning_rate": 8.138476294533848e-06, + "loss": 0.6104, + "step": 29376 + }, + { + "epoch": 0.87, + "learning_rate": 8.134666493632515e-06, + "loss": 0.4977, + "step": 29377 + }, + { + "epoch": 0.87, + "learning_rate": 8.13085754685845e-06, + "loss": 0.5735, + "step": 29378 + }, + { + "epoch": 0.87, + "learning_rate": 8.127049454247048e-06, + "loss": 0.5468, + "step": 29379 + }, + { + "epoch": 0.87, + "learning_rate": 8.123242215833738e-06, + "loss": 0.4692, + "step": 29380 + }, + { + "epoch": 0.87, + "learning_rate": 8.119435831653876e-06, + "loss": 0.6037, + "step": 29381 + }, + { + "epoch": 0.87, + "learning_rate": 8.115630301742905e-06, + "loss": 0.6052, + "step": 29382 + }, + { + "epoch": 0.87, + "learning_rate": 8.111825626136183e-06, + "loss": 0.5142, + "step": 29383 + }, + { + "epoch": 0.87, + "learning_rate": 8.10802180486907e-06, + "loss": 0.5251, + "step": 29384 + }, + { + "epoch": 0.87, + "learning_rate": 8.10421883797694e-06, + "loss": 0.4861, + "step": 29385 + }, + { + "epoch": 0.87, + "learning_rate": 8.100416725495153e-06, + "loss": 0.5957, + "step": 29386 + }, + { + "epoch": 0.87, + "learning_rate": 8.09661546745909e-06, + "loss": 0.4436, + "step": 29387 + }, + { + "epoch": 0.87, + "learning_rate": 8.09281506390407e-06, + "loss": 0.5288, + "step": 29388 + }, + { + "epoch": 0.87, + "learning_rate": 8.089015514865395e-06, + "loss": 0.623, + "step": 29389 + }, + { + "epoch": 0.87, + "learning_rate": 8.085216820378427e-06, + "loss": 0.5087, + "step": 29390 + }, + { + "epoch": 0.88, + "learning_rate": 8.081418980478495e-06, + "loss": 0.4975, + "step": 29391 + }, + { + "epoch": 0.88, + "learning_rate": 8.077621995200902e-06, + "loss": 0.5305, + "step": 29392 + }, + { + "epoch": 0.88, + "learning_rate": 8.07382586458092e-06, + "loss": 0.5754, + "step": 29393 + }, + { + "epoch": 0.88, + "learning_rate": 8.070030588653866e-06, + "loss": 0.5831, + "step": 29394 + }, + { + "epoch": 0.88, + "learning_rate": 8.066236167455043e-06, + "loss": 0.6603, + "step": 29395 + }, + { + "epoch": 0.88, + "learning_rate": 8.062442601019704e-06, + "loss": 0.4907, + "step": 29396 + }, + { + "epoch": 0.88, + "learning_rate": 8.058649889383151e-06, + "loss": 0.5727, + "step": 29397 + }, + { + "epoch": 0.88, + "learning_rate": 8.054858032580604e-06, + "loss": 0.5712, + "step": 29398 + }, + { + "epoch": 0.88, + "learning_rate": 8.051067030647352e-06, + "loss": 0.5402, + "step": 29399 + }, + { + "epoch": 0.88, + "learning_rate": 8.047276883618626e-06, + "loss": 0.5647, + "step": 29400 + }, + { + "epoch": 0.88, + "learning_rate": 8.043487591529663e-06, + "loss": 0.6024, + "step": 29401 + }, + { + "epoch": 0.88, + "learning_rate": 8.039699154415725e-06, + "loss": 0.5491, + "step": 29402 + }, + { + "epoch": 0.88, + "learning_rate": 8.035911572312004e-06, + "loss": 0.4827, + "step": 29403 + }, + { + "epoch": 0.88, + "learning_rate": 8.032124845253719e-06, + "loss": 0.5274, + "step": 29404 + }, + { + "epoch": 0.88, + "learning_rate": 8.028338973276072e-06, + "loss": 0.5205, + "step": 29405 + }, + { + "epoch": 0.88, + "learning_rate": 8.024553956414294e-06, + "loss": 0.5121, + "step": 29406 + }, + { + "epoch": 0.88, + "learning_rate": 8.020769794703553e-06, + "loss": 0.4616, + "step": 29407 + }, + { + "epoch": 0.88, + "learning_rate": 8.016986488179023e-06, + "loss": 0.6132, + "step": 29408 + }, + { + "epoch": 0.88, + "learning_rate": 8.013204036875888e-06, + "loss": 0.5795, + "step": 29409 + }, + { + "epoch": 0.88, + "learning_rate": 8.009422440829329e-06, + "loss": 0.5236, + "step": 29410 + }, + { + "epoch": 0.88, + "learning_rate": 8.005641700074495e-06, + "loss": 0.4784, + "step": 29411 + }, + { + "epoch": 0.88, + "learning_rate": 8.00186181464656e-06, + "loss": 0.5479, + "step": 29412 + }, + { + "epoch": 0.88, + "learning_rate": 7.998082784580618e-06, + "loss": 0.5062, + "step": 29413 + }, + { + "epoch": 0.88, + "learning_rate": 7.994304609911852e-06, + "loss": 0.5074, + "step": 29414 + }, + { + "epoch": 0.88, + "learning_rate": 7.990527290675354e-06, + "loss": 0.4813, + "step": 29415 + }, + { + "epoch": 0.88, + "learning_rate": 7.986750826906275e-06, + "loss": 0.5755, + "step": 29416 + }, + { + "epoch": 0.88, + "learning_rate": 7.98297521863971e-06, + "loss": 0.5587, + "step": 29417 + }, + { + "epoch": 0.88, + "learning_rate": 7.979200465910774e-06, + "loss": 0.529, + "step": 29418 + }, + { + "epoch": 0.88, + "learning_rate": 7.975426568754541e-06, + "loss": 0.5897, + "step": 29419 + }, + { + "epoch": 0.88, + "learning_rate": 7.971653527206114e-06, + "loss": 0.4598, + "step": 29420 + }, + { + "epoch": 0.88, + "learning_rate": 7.967881341300588e-06, + "loss": 0.4915, + "step": 29421 + }, + { + "epoch": 0.88, + "learning_rate": 7.964110011073022e-06, + "loss": 0.598, + "step": 29422 + }, + { + "epoch": 0.88, + "learning_rate": 7.96033953655847e-06, + "loss": 0.5758, + "step": 29423 + }, + { + "epoch": 0.88, + "learning_rate": 7.956569917791989e-06, + "loss": 0.6136, + "step": 29424 + }, + { + "epoch": 0.88, + "learning_rate": 7.952801154808665e-06, + "loss": 0.5186, + "step": 29425 + }, + { + "epoch": 0.88, + "learning_rate": 7.9490332476435e-06, + "loss": 0.446, + "step": 29426 + }, + { + "epoch": 0.88, + "learning_rate": 7.945266196331514e-06, + "loss": 0.4675, + "step": 29427 + }, + { + "epoch": 0.88, + "learning_rate": 7.941500000907764e-06, + "loss": 0.5604, + "step": 29428 + }, + { + "epoch": 0.88, + "learning_rate": 7.937734661407247e-06, + "loss": 0.3842, + "step": 29429 + }, + { + "epoch": 0.88, + "learning_rate": 7.933970177865001e-06, + "loss": 0.5602, + "step": 29430 + }, + { + "epoch": 0.88, + "learning_rate": 7.93020655031601e-06, + "loss": 0.6193, + "step": 29431 + }, + { + "epoch": 0.88, + "learning_rate": 7.926443778795245e-06, + "loss": 0.4356, + "step": 29432 + }, + { + "epoch": 0.88, + "learning_rate": 7.9226818633377e-06, + "loss": 0.5039, + "step": 29433 + }, + { + "epoch": 0.88, + "learning_rate": 7.91892080397839e-06, + "loss": 0.5353, + "step": 29434 + }, + { + "epoch": 0.88, + "learning_rate": 7.915160600752247e-06, + "loss": 0.6469, + "step": 29435 + }, + { + "epoch": 0.88, + "learning_rate": 7.911401253694228e-06, + "loss": 0.5574, + "step": 29436 + }, + { + "epoch": 0.88, + "learning_rate": 7.907642762839296e-06, + "loss": 0.5485, + "step": 29437 + }, + { + "epoch": 0.88, + "learning_rate": 7.90388512822241e-06, + "loss": 0.5548, + "step": 29438 + }, + { + "epoch": 0.88, + "learning_rate": 7.900128349878478e-06, + "loss": 0.4906, + "step": 29439 + }, + { + "epoch": 0.88, + "learning_rate": 7.896372427842469e-06, + "loss": 0.4478, + "step": 29440 + }, + { + "epoch": 0.88, + "learning_rate": 7.892617362149258e-06, + "loss": 0.4974, + "step": 29441 + }, + { + "epoch": 0.88, + "learning_rate": 7.888863152833792e-06, + "loss": 0.5326, + "step": 29442 + }, + { + "epoch": 0.88, + "learning_rate": 7.885109799930956e-06, + "loss": 0.6264, + "step": 29443 + }, + { + "epoch": 0.88, + "learning_rate": 7.881357303475667e-06, + "loss": 0.5258, + "step": 29444 + }, + { + "epoch": 0.88, + "learning_rate": 7.877605663502786e-06, + "loss": 0.4626, + "step": 29445 + }, + { + "epoch": 0.88, + "learning_rate": 7.873854880047227e-06, + "loss": 0.6928, + "step": 29446 + }, + { + "epoch": 0.88, + "learning_rate": 7.870104953143841e-06, + "loss": 0.52, + "step": 29447 + }, + { + "epoch": 0.88, + "learning_rate": 7.866355882827492e-06, + "loss": 0.5964, + "step": 29448 + }, + { + "epoch": 0.88, + "learning_rate": 7.86260766913306e-06, + "loss": 0.4381, + "step": 29449 + }, + { + "epoch": 0.88, + "learning_rate": 7.858860312095384e-06, + "loss": 0.5042, + "step": 29450 + }, + { + "epoch": 0.88, + "learning_rate": 7.855113811749281e-06, + "loss": 0.5499, + "step": 29451 + }, + { + "epoch": 0.88, + "learning_rate": 7.851368168129602e-06, + "loss": 0.5161, + "step": 29452 + }, + { + "epoch": 0.88, + "learning_rate": 7.847623381271207e-06, + "loss": 0.5061, + "step": 29453 + }, + { + "epoch": 0.88, + "learning_rate": 7.84387945120887e-06, + "loss": 0.3962, + "step": 29454 + }, + { + "epoch": 0.88, + "learning_rate": 7.840136377977392e-06, + "loss": 0.5506, + "step": 29455 + }, + { + "epoch": 0.88, + "learning_rate": 7.836394161611593e-06, + "loss": 0.4773, + "step": 29456 + }, + { + "epoch": 0.88, + "learning_rate": 7.83265280214629e-06, + "loss": 0.5758, + "step": 29457 + }, + { + "epoch": 0.88, + "learning_rate": 7.828912299616231e-06, + "loss": 0.6016, + "step": 29458 + }, + { + "epoch": 0.88, + "learning_rate": 7.825172654056234e-06, + "loss": 0.4751, + "step": 29459 + }, + { + "epoch": 0.88, + "learning_rate": 7.821433865501016e-06, + "loss": 0.531, + "step": 29460 + }, + { + "epoch": 0.88, + "learning_rate": 7.817695933985392e-06, + "loss": 0.5722, + "step": 29461 + }, + { + "epoch": 0.88, + "learning_rate": 7.813958859544068e-06, + "loss": 0.4675, + "step": 29462 + }, + { + "epoch": 0.88, + "learning_rate": 7.810222642211806e-06, + "loss": 0.5136, + "step": 29463 + }, + { + "epoch": 0.88, + "learning_rate": 7.806487282023379e-06, + "loss": 0.591, + "step": 29464 + }, + { + "epoch": 0.88, + "learning_rate": 7.802752779013489e-06, + "loss": 0.568, + "step": 29465 + }, + { + "epoch": 0.88, + "learning_rate": 7.799019133216834e-06, + "loss": 0.6029, + "step": 29466 + }, + { + "epoch": 0.88, + "learning_rate": 7.795286344668152e-06, + "loss": 0.4883, + "step": 29467 + }, + { + "epoch": 0.88, + "learning_rate": 7.79155441340217e-06, + "loss": 0.563, + "step": 29468 + }, + { + "epoch": 0.88, + "learning_rate": 7.787823339453571e-06, + "loss": 0.5878, + "step": 29469 + }, + { + "epoch": 0.88, + "learning_rate": 7.784093122857006e-06, + "loss": 0.4419, + "step": 29470 + }, + { + "epoch": 0.88, + "learning_rate": 7.780363763647192e-06, + "loss": 0.5352, + "step": 29471 + }, + { + "epoch": 0.88, + "learning_rate": 7.776635261858811e-06, + "loss": 0.5526, + "step": 29472 + }, + { + "epoch": 0.88, + "learning_rate": 7.772907617526526e-06, + "loss": 0.4018, + "step": 29473 + }, + { + "epoch": 0.88, + "learning_rate": 7.769180830684997e-06, + "loss": 0.6485, + "step": 29474 + }, + { + "epoch": 0.88, + "learning_rate": 7.76545490136884e-06, + "loss": 0.5228, + "step": 29475 + }, + { + "epoch": 0.88, + "learning_rate": 7.761729829612729e-06, + "loss": 0.4825, + "step": 29476 + }, + { + "epoch": 0.88, + "learning_rate": 7.758005615451303e-06, + "loss": 0.484, + "step": 29477 + }, + { + "epoch": 0.88, + "learning_rate": 7.754282258919177e-06, + "loss": 0.5182, + "step": 29478 + }, + { + "epoch": 0.88, + "learning_rate": 7.750559760050945e-06, + "loss": 0.4888, + "step": 29479 + }, + { + "epoch": 0.88, + "learning_rate": 7.746838118881271e-06, + "loss": 0.5328, + "step": 29480 + }, + { + "epoch": 0.88, + "learning_rate": 7.743117335444705e-06, + "loss": 0.4789, + "step": 29481 + }, + { + "epoch": 0.88, + "learning_rate": 7.739397409775872e-06, + "loss": 0.5473, + "step": 29482 + }, + { + "epoch": 0.88, + "learning_rate": 7.735678341909358e-06, + "loss": 0.5997, + "step": 29483 + }, + { + "epoch": 0.88, + "learning_rate": 7.731960131879734e-06, + "loss": 0.517, + "step": 29484 + }, + { + "epoch": 0.88, + "learning_rate": 7.728242779721561e-06, + "loss": 0.5426, + "step": 29485 + }, + { + "epoch": 0.88, + "learning_rate": 7.724526285469402e-06, + "loss": 0.6805, + "step": 29486 + }, + { + "epoch": 0.88, + "learning_rate": 7.72081064915784e-06, + "loss": 0.6036, + "step": 29487 + }, + { + "epoch": 0.88, + "learning_rate": 7.7170958708214e-06, + "loss": 0.4534, + "step": 29488 + }, + { + "epoch": 0.88, + "learning_rate": 7.713381950494614e-06, + "loss": 0.5486, + "step": 29489 + }, + { + "epoch": 0.88, + "learning_rate": 7.709668888212007e-06, + "loss": 0.4707, + "step": 29490 + }, + { + "epoch": 0.88, + "learning_rate": 7.705956684008131e-06, + "loss": 0.4563, + "step": 29491 + }, + { + "epoch": 0.88, + "learning_rate": 7.702245337917491e-06, + "loss": 0.5704, + "step": 29492 + }, + { + "epoch": 0.88, + "learning_rate": 7.698534849974592e-06, + "loss": 0.6093, + "step": 29493 + }, + { + "epoch": 0.88, + "learning_rate": 7.694825220213909e-06, + "loss": 0.5814, + "step": 29494 + }, + { + "epoch": 0.88, + "learning_rate": 7.691116448669954e-06, + "loss": 0.5853, + "step": 29495 + }, + { + "epoch": 0.88, + "learning_rate": 7.687408535377216e-06, + "loss": 0.4282, + "step": 29496 + }, + { + "epoch": 0.88, + "learning_rate": 7.683701480370176e-06, + "loss": 0.5302, + "step": 29497 + }, + { + "epoch": 0.88, + "learning_rate": 7.679995283683261e-06, + "loss": 0.5824, + "step": 29498 + }, + { + "epoch": 0.88, + "learning_rate": 7.676289945350956e-06, + "loss": 0.4665, + "step": 29499 + }, + { + "epoch": 0.88, + "learning_rate": 7.672585465407722e-06, + "loss": 0.5038, + "step": 29500 + }, + { + "epoch": 0.88, + "learning_rate": 7.668881843887977e-06, + "loss": 0.5388, + "step": 29501 + }, + { + "epoch": 0.88, + "learning_rate": 7.665179080826191e-06, + "loss": 0.5572, + "step": 29502 + }, + { + "epoch": 0.88, + "learning_rate": 7.661477176256748e-06, + "loss": 0.5065, + "step": 29503 + }, + { + "epoch": 0.88, + "learning_rate": 7.6577761302141e-06, + "loss": 0.5962, + "step": 29504 + }, + { + "epoch": 0.88, + "learning_rate": 7.65407594273263e-06, + "loss": 0.6499, + "step": 29505 + }, + { + "epoch": 0.88, + "learning_rate": 7.650376613846766e-06, + "loss": 0.6056, + "step": 29506 + }, + { + "epoch": 0.88, + "learning_rate": 7.646678143590901e-06, + "loss": 0.5878, + "step": 29507 + }, + { + "epoch": 0.88, + "learning_rate": 7.642980531999422e-06, + "loss": 0.4618, + "step": 29508 + }, + { + "epoch": 0.88, + "learning_rate": 7.639283779106677e-06, + "loss": 0.5438, + "step": 29509 + }, + { + "epoch": 0.88, + "learning_rate": 7.63558788494706e-06, + "loss": 0.5953, + "step": 29510 + }, + { + "epoch": 0.88, + "learning_rate": 7.631892849554956e-06, + "loss": 0.4841, + "step": 29511 + }, + { + "epoch": 0.88, + "learning_rate": 7.628198672964693e-06, + "loss": 0.5467, + "step": 29512 + }, + { + "epoch": 0.88, + "learning_rate": 7.624505355210609e-06, + "loss": 0.4401, + "step": 29513 + }, + { + "epoch": 0.88, + "learning_rate": 7.6208128963270455e-06, + "loss": 0.5265, + "step": 29514 + }, + { + "epoch": 0.88, + "learning_rate": 7.617121296348373e-06, + "loss": 0.4159, + "step": 29515 + }, + { + "epoch": 0.88, + "learning_rate": 7.6134305553088754e-06, + "loss": 0.5123, + "step": 29516 + }, + { + "epoch": 0.88, + "learning_rate": 7.609740673242882e-06, + "loss": 0.419, + "step": 29517 + }, + { + "epoch": 0.88, + "learning_rate": 7.606051650184687e-06, + "loss": 0.4142, + "step": 29518 + }, + { + "epoch": 0.88, + "learning_rate": 7.602363486168617e-06, + "loss": 0.5935, + "step": 29519 + }, + { + "epoch": 0.88, + "learning_rate": 7.5986761812289256e-06, + "loss": 0.5131, + "step": 29520 + }, + { + "epoch": 0.88, + "learning_rate": 7.5949897353999375e-06, + "loss": 0.491, + "step": 29521 + }, + { + "epoch": 0.88, + "learning_rate": 7.591304148715883e-06, + "loss": 0.5255, + "step": 29522 + }, + { + "epoch": 0.88, + "learning_rate": 7.587619421211067e-06, + "loss": 0.473, + "step": 29523 + }, + { + "epoch": 0.88, + "learning_rate": 7.58393555291973e-06, + "loss": 0.5642, + "step": 29524 + }, + { + "epoch": 0.88, + "learning_rate": 7.5802525438761095e-06, + "loss": 0.369, + "step": 29525 + }, + { + "epoch": 0.88, + "learning_rate": 7.5765703941144905e-06, + "loss": 0.6663, + "step": 29526 + }, + { + "epoch": 0.88, + "learning_rate": 7.57288910366909e-06, + "loss": 0.6355, + "step": 29527 + }, + { + "epoch": 0.88, + "learning_rate": 7.5692086725741016e-06, + "loss": 0.4972, + "step": 29528 + }, + { + "epoch": 0.88, + "learning_rate": 7.565529100863777e-06, + "loss": 0.5107, + "step": 29529 + }, + { + "epoch": 0.88, + "learning_rate": 7.561850388572345e-06, + "loss": 0.5086, + "step": 29530 + }, + { + "epoch": 0.88, + "learning_rate": 7.558172535733976e-06, + "loss": 0.5231, + "step": 29531 + }, + { + "epoch": 0.88, + "learning_rate": 7.554495542382856e-06, + "loss": 0.5211, + "step": 29532 + }, + { + "epoch": 0.88, + "learning_rate": 7.5508194085532005e-06, + "loss": 0.4305, + "step": 29533 + }, + { + "epoch": 0.88, + "learning_rate": 7.547144134279171e-06, + "loss": 0.589, + "step": 29534 + }, + { + "epoch": 0.88, + "learning_rate": 7.543469719594975e-06, + "loss": 0.4434, + "step": 29535 + }, + { + "epoch": 0.88, + "learning_rate": 7.539796164534741e-06, + "loss": 0.5625, + "step": 29536 + }, + { + "epoch": 0.88, + "learning_rate": 7.536123469132617e-06, + "loss": 0.5813, + "step": 29537 + }, + { + "epoch": 0.88, + "learning_rate": 7.532451633422755e-06, + "loss": 0.5943, + "step": 29538 + }, + { + "epoch": 0.88, + "learning_rate": 7.5287806574393385e-06, + "loss": 0.4409, + "step": 29539 + }, + { + "epoch": 0.88, + "learning_rate": 7.525110541216451e-06, + "loss": 0.5876, + "step": 29540 + }, + { + "epoch": 0.88, + "learning_rate": 7.521441284788211e-06, + "loss": 0.4224, + "step": 29541 + }, + { + "epoch": 0.88, + "learning_rate": 7.517772888188756e-06, + "loss": 0.56, + "step": 29542 + }, + { + "epoch": 0.88, + "learning_rate": 7.514105351452206e-06, + "loss": 0.5243, + "step": 29543 + }, + { + "epoch": 0.88, + "learning_rate": 7.51043867461263e-06, + "loss": 0.5027, + "step": 29544 + }, + { + "epoch": 0.88, + "learning_rate": 7.506772857704136e-06, + "loss": 0.5427, + "step": 29545 + }, + { + "epoch": 0.88, + "learning_rate": 7.50310790076082e-06, + "loss": 0.4876, + "step": 29546 + }, + { + "epoch": 0.88, + "learning_rate": 7.499443803816719e-06, + "loss": 0.4849, + "step": 29547 + }, + { + "epoch": 0.88, + "learning_rate": 7.495780566905919e-06, + "loss": 0.5284, + "step": 29548 + }, + { + "epoch": 0.88, + "learning_rate": 7.4921181900625025e-06, + "loss": 0.5003, + "step": 29549 + }, + { + "epoch": 0.88, + "learning_rate": 7.488456673320499e-06, + "loss": 0.5467, + "step": 29550 + }, + { + "epoch": 0.88, + "learning_rate": 7.484796016713935e-06, + "loss": 0.5125, + "step": 29551 + }, + { + "epoch": 0.88, + "learning_rate": 7.481136220276863e-06, + "loss": 0.6078, + "step": 29552 + }, + { + "epoch": 0.88, + "learning_rate": 7.477477284043322e-06, + "loss": 0.5274, + "step": 29553 + }, + { + "epoch": 0.88, + "learning_rate": 7.473819208047328e-06, + "loss": 0.6193, + "step": 29554 + }, + { + "epoch": 0.88, + "learning_rate": 7.470161992322888e-06, + "loss": 0.6261, + "step": 29555 + }, + { + "epoch": 0.88, + "learning_rate": 7.466505636903987e-06, + "loss": 0.4425, + "step": 29556 + }, + { + "epoch": 0.88, + "learning_rate": 7.4628501418246395e-06, + "loss": 0.5641, + "step": 29557 + }, + { + "epoch": 0.88, + "learning_rate": 7.459195507118843e-06, + "loss": 0.5993, + "step": 29558 + }, + { + "epoch": 0.88, + "learning_rate": 7.45554173282057e-06, + "loss": 0.528, + "step": 29559 + }, + { + "epoch": 0.88, + "learning_rate": 7.451888818963771e-06, + "loss": 0.4207, + "step": 29560 + }, + { + "epoch": 0.88, + "learning_rate": 7.448236765582406e-06, + "loss": 0.4974, + "step": 29561 + }, + { + "epoch": 0.88, + "learning_rate": 7.444585572710483e-06, + "loss": 0.5327, + "step": 29562 + }, + { + "epoch": 0.88, + "learning_rate": 7.440935240381886e-06, + "loss": 0.4513, + "step": 29563 + }, + { + "epoch": 0.88, + "learning_rate": 7.437285768630608e-06, + "loss": 0.5748, + "step": 29564 + }, + { + "epoch": 0.88, + "learning_rate": 7.4336371574905365e-06, + "loss": 0.5423, + "step": 29565 + }, + { + "epoch": 0.88, + "learning_rate": 7.4299894069956185e-06, + "loss": 0.621, + "step": 29566 + }, + { + "epoch": 0.88, + "learning_rate": 7.426342517179763e-06, + "loss": 0.5297, + "step": 29567 + }, + { + "epoch": 0.88, + "learning_rate": 7.422696488076874e-06, + "loss": 0.4454, + "step": 29568 + }, + { + "epoch": 0.88, + "learning_rate": 7.419051319720871e-06, + "loss": 0.6758, + "step": 29569 + }, + { + "epoch": 0.88, + "learning_rate": 7.415407012145625e-06, + "loss": 0.5827, + "step": 29570 + }, + { + "epoch": 0.88, + "learning_rate": 7.41176356538501e-06, + "loss": 0.4339, + "step": 29571 + }, + { + "epoch": 0.88, + "learning_rate": 7.408120979472921e-06, + "loss": 0.5636, + "step": 29572 + }, + { + "epoch": 0.88, + "learning_rate": 7.404479254443231e-06, + "loss": 0.5355, + "step": 29573 + }, + { + "epoch": 0.88, + "learning_rate": 7.40083839032979e-06, + "loss": 0.5367, + "step": 29574 + }, + { + "epoch": 0.88, + "learning_rate": 7.397198387166426e-06, + "loss": 0.5096, + "step": 29575 + }, + { + "epoch": 0.88, + "learning_rate": 7.393559244987002e-06, + "loss": 0.504, + "step": 29576 + }, + { + "epoch": 0.88, + "learning_rate": 7.389920963825369e-06, + "loss": 0.6161, + "step": 29577 + }, + { + "epoch": 0.88, + "learning_rate": 7.386283543715345e-06, + "loss": 0.5422, + "step": 29578 + }, + { + "epoch": 0.88, + "learning_rate": 7.382646984690744e-06, + "loss": 0.6035, + "step": 29579 + }, + { + "epoch": 0.88, + "learning_rate": 7.379011286785364e-06, + "loss": 0.5349, + "step": 29580 + }, + { + "epoch": 0.88, + "learning_rate": 7.375376450033045e-06, + "loss": 0.4521, + "step": 29581 + }, + { + "epoch": 0.88, + "learning_rate": 7.371742474467536e-06, + "loss": 0.5552, + "step": 29582 + }, + { + "epoch": 0.88, + "learning_rate": 7.3681093601226655e-06, + "loss": 0.4142, + "step": 29583 + }, + { + "epoch": 0.88, + "learning_rate": 7.364477107032175e-06, + "loss": 0.6045, + "step": 29584 + }, + { + "epoch": 0.88, + "learning_rate": 7.360845715229869e-06, + "loss": 0.518, + "step": 29585 + }, + { + "epoch": 0.88, + "learning_rate": 7.357215184749488e-06, + "loss": 0.5125, + "step": 29586 + }, + { + "epoch": 0.88, + "learning_rate": 7.353585515624795e-06, + "loss": 0.5952, + "step": 29587 + }, + { + "epoch": 0.88, + "learning_rate": 7.3499567078895605e-06, + "loss": 0.6125, + "step": 29588 + }, + { + "epoch": 0.88, + "learning_rate": 7.3463287615774926e-06, + "loss": 0.4307, + "step": 29589 + }, + { + "epoch": 0.88, + "learning_rate": 7.342701676722319e-06, + "loss": 0.6502, + "step": 29590 + }, + { + "epoch": 0.88, + "learning_rate": 7.3390754533577685e-06, + "loss": 0.6002, + "step": 29591 + }, + { + "epoch": 0.88, + "learning_rate": 7.335450091517593e-06, + "loss": 0.4156, + "step": 29592 + }, + { + "epoch": 0.88, + "learning_rate": 7.331825591235464e-06, + "loss": 0.4946, + "step": 29593 + }, + { + "epoch": 0.88, + "learning_rate": 7.328201952545066e-06, + "loss": 0.6156, + "step": 29594 + }, + { + "epoch": 0.88, + "learning_rate": 7.324579175480117e-06, + "loss": 0.6206, + "step": 29595 + }, + { + "epoch": 0.88, + "learning_rate": 7.32095726007429e-06, + "loss": 0.4857, + "step": 29596 + }, + { + "epoch": 0.88, + "learning_rate": 7.317336206361291e-06, + "loss": 0.6058, + "step": 29597 + }, + { + "epoch": 0.88, + "learning_rate": 7.313716014374749e-06, + "loss": 0.542, + "step": 29598 + }, + { + "epoch": 0.88, + "learning_rate": 7.310096684148315e-06, + "loss": 0.563, + "step": 29599 + }, + { + "epoch": 0.88, + "learning_rate": 7.306478215715673e-06, + "loss": 0.5151, + "step": 29600 + }, + { + "epoch": 0.88, + "learning_rate": 7.302860609110462e-06, + "loss": 0.5355, + "step": 29601 + }, + { + "epoch": 0.88, + "learning_rate": 7.2992438643663005e-06, + "loss": 0.624, + "step": 29602 + }, + { + "epoch": 0.88, + "learning_rate": 7.295627981516817e-06, + "loss": 0.4975, + "step": 29603 + }, + { + "epoch": 0.88, + "learning_rate": 7.292012960595629e-06, + "loss": 0.5189, + "step": 29604 + }, + { + "epoch": 0.88, + "learning_rate": 7.288398801636376e-06, + "loss": 0.5443, + "step": 29605 + }, + { + "epoch": 0.88, + "learning_rate": 7.28478550467262e-06, + "loss": 0.5278, + "step": 29606 + }, + { + "epoch": 0.88, + "learning_rate": 7.28117306973799e-06, + "loss": 0.5399, + "step": 29607 + }, + { + "epoch": 0.88, + "learning_rate": 7.277561496866048e-06, + "loss": 0.5593, + "step": 29608 + }, + { + "epoch": 0.88, + "learning_rate": 7.2739507860904e-06, + "loss": 0.5035, + "step": 29609 + }, + { + "epoch": 0.88, + "learning_rate": 7.270340937444587e-06, + "loss": 0.5043, + "step": 29610 + }, + { + "epoch": 0.88, + "learning_rate": 7.266731950962191e-06, + "loss": 0.5146, + "step": 29611 + }, + { + "epoch": 0.88, + "learning_rate": 7.2631238266767656e-06, + "loss": 0.5245, + "step": 29612 + }, + { + "epoch": 0.88, + "learning_rate": 7.2595165646218485e-06, + "loss": 0.5536, + "step": 29613 + }, + { + "epoch": 0.88, + "learning_rate": 7.25591016483097e-06, + "loss": 0.5411, + "step": 29614 + }, + { + "epoch": 0.88, + "learning_rate": 7.252304627337681e-06, + "loss": 0.5541, + "step": 29615 + }, + { + "epoch": 0.88, + "learning_rate": 7.248699952175508e-06, + "loss": 0.6174, + "step": 29616 + }, + { + "epoch": 0.88, + "learning_rate": 7.2450961393779605e-06, + "loss": 0.5479, + "step": 29617 + }, + { + "epoch": 0.88, + "learning_rate": 7.241493188978532e-06, + "loss": 0.5638, + "step": 29618 + }, + { + "epoch": 0.88, + "learning_rate": 7.237891101010719e-06, + "loss": 0.56, + "step": 29619 + }, + { + "epoch": 0.88, + "learning_rate": 7.234289875508038e-06, + "loss": 0.4841, + "step": 29620 + }, + { + "epoch": 0.88, + "learning_rate": 7.230689512503963e-06, + "loss": 0.5749, + "step": 29621 + }, + { + "epoch": 0.88, + "learning_rate": 7.227090012031945e-06, + "loss": 0.5377, + "step": 29622 + }, + { + "epoch": 0.88, + "learning_rate": 7.223491374125468e-06, + "loss": 0.4546, + "step": 29623 + }, + { + "epoch": 0.88, + "learning_rate": 7.219893598818006e-06, + "loss": 0.5969, + "step": 29624 + }, + { + "epoch": 0.88, + "learning_rate": 7.216296686142987e-06, + "loss": 0.6738, + "step": 29625 + }, + { + "epoch": 0.88, + "learning_rate": 7.212700636133873e-06, + "loss": 0.4986, + "step": 29626 + }, + { + "epoch": 0.88, + "learning_rate": 7.209105448824061e-06, + "loss": 0.5196, + "step": 29627 + }, + { + "epoch": 0.88, + "learning_rate": 7.205511124247033e-06, + "loss": 0.4452, + "step": 29628 + }, + { + "epoch": 0.88, + "learning_rate": 7.201917662436153e-06, + "loss": 0.5697, + "step": 29629 + }, + { + "epoch": 0.88, + "learning_rate": 7.19832506342486e-06, + "loss": 0.4999, + "step": 29630 + }, + { + "epoch": 0.88, + "learning_rate": 7.194733327246572e-06, + "loss": 0.6187, + "step": 29631 + }, + { + "epoch": 0.88, + "learning_rate": 7.191142453934663e-06, + "loss": 0.5045, + "step": 29632 + }, + { + "epoch": 0.88, + "learning_rate": 7.187552443522505e-06, + "loss": 0.5227, + "step": 29633 + }, + { + "epoch": 0.88, + "learning_rate": 7.183963296043494e-06, + "loss": 0.4528, + "step": 29634 + }, + { + "epoch": 0.88, + "learning_rate": 7.180375011531015e-06, + "loss": 0.5586, + "step": 29635 + }, + { + "epoch": 0.88, + "learning_rate": 7.176787590018419e-06, + "loss": 0.5784, + "step": 29636 + }, + { + "epoch": 0.88, + "learning_rate": 7.173201031539034e-06, + "loss": 0.5326, + "step": 29637 + }, + { + "epoch": 0.88, + "learning_rate": 7.169615336126234e-06, + "loss": 0.4775, + "step": 29638 + }, + { + "epoch": 0.88, + "learning_rate": 7.166030503813348e-06, + "loss": 0.606, + "step": 29639 + }, + { + "epoch": 0.88, + "learning_rate": 7.162446534633727e-06, + "loss": 0.4012, + "step": 29640 + }, + { + "epoch": 0.88, + "learning_rate": 7.158863428620677e-06, + "loss": 0.6672, + "step": 29641 + }, + { + "epoch": 0.88, + "learning_rate": 7.155281185807505e-06, + "loss": 0.4747, + "step": 29642 + }, + { + "epoch": 0.88, + "learning_rate": 7.151699806227508e-06, + "loss": 0.5029, + "step": 29643 + }, + { + "epoch": 0.88, + "learning_rate": 7.148119289914023e-06, + "loss": 0.4958, + "step": 29644 + }, + { + "epoch": 0.88, + "learning_rate": 7.144539636900327e-06, + "loss": 0.528, + "step": 29645 + }, + { + "epoch": 0.88, + "learning_rate": 7.1409608472196685e-06, + "loss": 0.5794, + "step": 29646 + }, + { + "epoch": 0.88, + "learning_rate": 7.137382920905356e-06, + "loss": 0.6483, + "step": 29647 + }, + { + "epoch": 0.88, + "learning_rate": 7.133805857990639e-06, + "loss": 0.6012, + "step": 29648 + }, + { + "epoch": 0.88, + "learning_rate": 7.13022965850878e-06, + "loss": 0.5124, + "step": 29649 + }, + { + "epoch": 0.88, + "learning_rate": 7.126654322493043e-06, + "loss": 0.5706, + "step": 29650 + }, + { + "epoch": 0.88, + "learning_rate": 7.123079849976666e-06, + "loss": 0.6544, + "step": 29651 + }, + { + "epoch": 0.88, + "learning_rate": 7.119506240992857e-06, + "loss": 0.5238, + "step": 29652 + }, + { + "epoch": 0.88, + "learning_rate": 7.115933495574856e-06, + "loss": 0.6082, + "step": 29653 + }, + { + "epoch": 0.88, + "learning_rate": 7.112361613755914e-06, + "loss": 0.5518, + "step": 29654 + }, + { + "epoch": 0.88, + "learning_rate": 7.108790595569204e-06, + "loss": 0.4604, + "step": 29655 + }, + { + "epoch": 0.88, + "learning_rate": 7.105220441047911e-06, + "loss": 0.4157, + "step": 29656 + }, + { + "epoch": 0.88, + "learning_rate": 7.1016511502252635e-06, + "loss": 0.5249, + "step": 29657 + }, + { + "epoch": 0.88, + "learning_rate": 7.098082723134436e-06, + "loss": 0.5251, + "step": 29658 + }, + { + "epoch": 0.88, + "learning_rate": 7.0945151598086234e-06, + "loss": 0.6222, + "step": 29659 + }, + { + "epoch": 0.88, + "learning_rate": 7.090948460280988e-06, + "loss": 0.5363, + "step": 29660 + }, + { + "epoch": 0.88, + "learning_rate": 7.08738262458466e-06, + "loss": 0.6147, + "step": 29661 + }, + { + "epoch": 0.88, + "learning_rate": 7.0838176527528225e-06, + "loss": 0.4979, + "step": 29662 + }, + { + "epoch": 0.88, + "learning_rate": 7.080253544818627e-06, + "loss": 0.5712, + "step": 29663 + }, + { + "epoch": 0.88, + "learning_rate": 7.076690300815203e-06, + "loss": 0.5548, + "step": 29664 + }, + { + "epoch": 0.88, + "learning_rate": 7.0731279207756574e-06, + "loss": 0.5466, + "step": 29665 + }, + { + "epoch": 0.88, + "learning_rate": 7.0695664047331305e-06, + "loss": 0.4962, + "step": 29666 + }, + { + "epoch": 0.88, + "learning_rate": 7.066005752720761e-06, + "loss": 0.4956, + "step": 29667 + }, + { + "epoch": 0.88, + "learning_rate": 7.062445964771603e-06, + "loss": 0.488, + "step": 29668 + }, + { + "epoch": 0.88, + "learning_rate": 7.0588870409188045e-06, + "loss": 0.5417, + "step": 29669 + }, + { + "epoch": 0.88, + "learning_rate": 7.055328981195419e-06, + "loss": 0.7466, + "step": 29670 + }, + { + "epoch": 0.88, + "learning_rate": 7.0517717856345535e-06, + "loss": 0.5713, + "step": 29671 + }, + { + "epoch": 0.88, + "learning_rate": 7.048215454269247e-06, + "loss": 0.604, + "step": 29672 + }, + { + "epoch": 0.88, + "learning_rate": 7.044659987132585e-06, + "loss": 0.4926, + "step": 29673 + }, + { + "epoch": 0.88, + "learning_rate": 7.041105384257652e-06, + "loss": 0.465, + "step": 29674 + }, + { + "epoch": 0.88, + "learning_rate": 7.0375516456774646e-06, + "loss": 0.4784, + "step": 29675 + }, + { + "epoch": 0.88, + "learning_rate": 7.0339987714250545e-06, + "loss": 0.6116, + "step": 29676 + }, + { + "epoch": 0.88, + "learning_rate": 7.030446761533483e-06, + "loss": 0.5499, + "step": 29677 + }, + { + "epoch": 0.88, + "learning_rate": 7.026895616035767e-06, + "loss": 0.4762, + "step": 29678 + }, + { + "epoch": 0.88, + "learning_rate": 7.023345334964926e-06, + "loss": 0.6574, + "step": 29679 + }, + { + "epoch": 0.88, + "learning_rate": 7.0197959183539445e-06, + "loss": 0.421, + "step": 29680 + }, + { + "epoch": 0.88, + "learning_rate": 7.016247366235839e-06, + "loss": 0.5377, + "step": 29681 + }, + { + "epoch": 0.88, + "learning_rate": 7.012699678643619e-06, + "loss": 0.4595, + "step": 29682 + }, + { + "epoch": 0.88, + "learning_rate": 7.009152855610246e-06, + "loss": 0.4892, + "step": 29683 + }, + { + "epoch": 0.88, + "learning_rate": 7.005606897168726e-06, + "loss": 0.5055, + "step": 29684 + }, + { + "epoch": 0.88, + "learning_rate": 7.0020618033519895e-06, + "loss": 0.5008, + "step": 29685 + }, + { + "epoch": 0.88, + "learning_rate": 6.998517574193031e-06, + "loss": 0.5346, + "step": 29686 + }, + { + "epoch": 0.88, + "learning_rate": 6.9949742097247806e-06, + "loss": 0.4997, + "step": 29687 + }, + { + "epoch": 0.88, + "learning_rate": 6.9914317099802e-06, + "loss": 0.4621, + "step": 29688 + }, + { + "epoch": 0.88, + "learning_rate": 6.987890074992209e-06, + "loss": 0.6358, + "step": 29689 + }, + { + "epoch": 0.88, + "learning_rate": 6.984349304793758e-06, + "loss": 0.4352, + "step": 29690 + }, + { + "epoch": 0.88, + "learning_rate": 6.980809399417742e-06, + "loss": 0.5822, + "step": 29691 + }, + { + "epoch": 0.88, + "learning_rate": 6.977270358897081e-06, + "loss": 0.5025, + "step": 29692 + }, + { + "epoch": 0.88, + "learning_rate": 6.973732183264703e-06, + "loss": 0.5471, + "step": 29693 + }, + { + "epoch": 0.88, + "learning_rate": 6.970194872553482e-06, + "loss": 0.4917, + "step": 29694 + }, + { + "epoch": 0.88, + "learning_rate": 6.966658426796302e-06, + "loss": 0.4124, + "step": 29695 + }, + { + "epoch": 0.88, + "learning_rate": 6.963122846026049e-06, + "loss": 0.4913, + "step": 29696 + }, + { + "epoch": 0.88, + "learning_rate": 6.959588130275618e-06, + "loss": 0.4784, + "step": 29697 + }, + { + "epoch": 0.88, + "learning_rate": 6.95605427957785e-06, + "loss": 0.5106, + "step": 29698 + }, + { + "epoch": 0.88, + "learning_rate": 6.952521293965597e-06, + "loss": 0.553, + "step": 29699 + }, + { + "epoch": 0.88, + "learning_rate": 6.9489891734717096e-06, + "loss": 0.4699, + "step": 29700 + }, + { + "epoch": 0.88, + "learning_rate": 6.945457918129039e-06, + "loss": 0.4863, + "step": 29701 + }, + { + "epoch": 0.88, + "learning_rate": 6.941927527970415e-06, + "loss": 0.5344, + "step": 29702 + }, + { + "epoch": 0.88, + "learning_rate": 6.938398003028679e-06, + "loss": 0.468, + "step": 29703 + }, + { + "epoch": 0.88, + "learning_rate": 6.934869343336603e-06, + "loss": 0.5484, + "step": 29704 + }, + { + "epoch": 0.88, + "learning_rate": 6.931341548927017e-06, + "loss": 0.4955, + "step": 29705 + }, + { + "epoch": 0.88, + "learning_rate": 6.92781461983274e-06, + "loss": 0.564, + "step": 29706 + }, + { + "epoch": 0.88, + "learning_rate": 6.924288556086555e-06, + "loss": 0.601, + "step": 29707 + }, + { + "epoch": 0.88, + "learning_rate": 6.920763357721216e-06, + "loss": 0.6574, + "step": 29708 + }, + { + "epoch": 0.88, + "learning_rate": 6.917239024769517e-06, + "loss": 0.6729, + "step": 29709 + }, + { + "epoch": 0.88, + "learning_rate": 6.913715557264256e-06, + "loss": 0.5756, + "step": 29710 + }, + { + "epoch": 0.88, + "learning_rate": 6.910192955238149e-06, + "loss": 0.6024, + "step": 29711 + }, + { + "epoch": 0.88, + "learning_rate": 6.906671218723981e-06, + "loss": 0.5957, + "step": 29712 + }, + { + "epoch": 0.88, + "learning_rate": 6.903150347754483e-06, + "loss": 0.5154, + "step": 29713 + }, + { + "epoch": 0.88, + "learning_rate": 6.899630342362362e-06, + "loss": 0.4883, + "step": 29714 + }, + { + "epoch": 0.88, + "learning_rate": 6.89611120258038e-06, + "loss": 0.5871, + "step": 29715 + }, + { + "epoch": 0.88, + "learning_rate": 6.892592928441266e-06, + "loss": 0.4566, + "step": 29716 + }, + { + "epoch": 0.88, + "learning_rate": 6.889075519977717e-06, + "loss": 0.5314, + "step": 29717 + }, + { + "epoch": 0.88, + "learning_rate": 6.885558977222406e-06, + "loss": 0.6249, + "step": 29718 + }, + { + "epoch": 0.88, + "learning_rate": 6.882043300208063e-06, + "loss": 0.4609, + "step": 29719 + }, + { + "epoch": 0.88, + "learning_rate": 6.878528488967373e-06, + "loss": 0.4255, + "step": 29720 + }, + { + "epoch": 0.88, + "learning_rate": 6.87501454353302e-06, + "loss": 0.5059, + "step": 29721 + }, + { + "epoch": 0.88, + "learning_rate": 6.871501463937668e-06, + "loss": 0.6079, + "step": 29722 + }, + { + "epoch": 0.88, + "learning_rate": 6.867989250213958e-06, + "loss": 0.5248, + "step": 29723 + }, + { + "epoch": 0.88, + "learning_rate": 6.864477902394572e-06, + "loss": 0.5362, + "step": 29724 + }, + { + "epoch": 0.88, + "learning_rate": 6.860967420512165e-06, + "loss": 0.6121, + "step": 29725 + }, + { + "epoch": 0.88, + "learning_rate": 6.857457804599354e-06, + "loss": 0.586, + "step": 29726 + }, + { + "epoch": 0.89, + "learning_rate": 6.853949054688768e-06, + "loss": 0.4792, + "step": 29727 + }, + { + "epoch": 0.89, + "learning_rate": 6.850441170813038e-06, + "loss": 0.5365, + "step": 29728 + }, + { + "epoch": 0.89, + "learning_rate": 6.846934153004792e-06, + "loss": 0.4685, + "step": 29729 + }, + { + "epoch": 0.89, + "learning_rate": 6.843428001296615e-06, + "loss": 0.4839, + "step": 29730 + }, + { + "epoch": 0.89, + "learning_rate": 6.839922715721137e-06, + "loss": 0.5797, + "step": 29731 + }, + { + "epoch": 0.89, + "learning_rate": 6.836418296310898e-06, + "loss": 0.4907, + "step": 29732 + }, + { + "epoch": 0.89, + "learning_rate": 6.83291474309854e-06, + "loss": 0.4514, + "step": 29733 + }, + { + "epoch": 0.89, + "learning_rate": 6.82941205611658e-06, + "loss": 0.5395, + "step": 29734 + }, + { + "epoch": 0.89, + "learning_rate": 6.825910235397614e-06, + "loss": 0.4843, + "step": 29735 + }, + { + "epoch": 0.89, + "learning_rate": 6.822409280974218e-06, + "loss": 0.4794, + "step": 29736 + }, + { + "epoch": 0.89, + "learning_rate": 6.818909192878931e-06, + "loss": 0.6507, + "step": 29737 + }, + { + "epoch": 0.89, + "learning_rate": 6.81540997114426e-06, + "loss": 0.4834, + "step": 29738 + }, + { + "epoch": 0.89, + "learning_rate": 6.81191161580278e-06, + "loss": 0.6609, + "step": 29739 + }, + { + "epoch": 0.89, + "learning_rate": 6.808414126887008e-06, + "loss": 0.5981, + "step": 29740 + }, + { + "epoch": 0.89, + "learning_rate": 6.804917504429464e-06, + "loss": 0.4962, + "step": 29741 + }, + { + "epoch": 0.89, + "learning_rate": 6.801421748462644e-06, + "loss": 0.5365, + "step": 29742 + }, + { + "epoch": 0.89, + "learning_rate": 6.797926859019044e-06, + "loss": 0.5361, + "step": 29743 + }, + { + "epoch": 0.89, + "learning_rate": 6.7944328361312035e-06, + "loss": 0.5349, + "step": 29744 + }, + { + "epoch": 0.89, + "learning_rate": 6.7909396798315535e-06, + "loss": 0.5101, + "step": 29745 + }, + { + "epoch": 0.89, + "learning_rate": 6.787447390152623e-06, + "loss": 0.5467, + "step": 29746 + }, + { + "epoch": 0.89, + "learning_rate": 6.783955967126832e-06, + "loss": 0.53, + "step": 29747 + }, + { + "epoch": 0.89, + "learning_rate": 6.780465410786685e-06, + "loss": 0.6509, + "step": 29748 + }, + { + "epoch": 0.89, + "learning_rate": 6.776975721164591e-06, + "loss": 0.5604, + "step": 29749 + }, + { + "epoch": 0.89, + "learning_rate": 6.7734868982930464e-06, + "loss": 0.5122, + "step": 29750 + }, + { + "epoch": 0.89, + "learning_rate": 6.769998942204448e-06, + "loss": 0.5284, + "step": 29751 + }, + { + "epoch": 0.89, + "learning_rate": 6.7665118529312565e-06, + "loss": 0.5178, + "step": 29752 + }, + { + "epoch": 0.89, + "learning_rate": 6.76302563050587e-06, + "loss": 0.5297, + "step": 29753 + }, + { + "epoch": 0.89, + "learning_rate": 6.759540274960707e-06, + "loss": 0.4707, + "step": 29754 + }, + { + "epoch": 0.89, + "learning_rate": 6.756055786328186e-06, + "loss": 0.56, + "step": 29755 + }, + { + "epoch": 0.89, + "learning_rate": 6.752572164640702e-06, + "loss": 0.5613, + "step": 29756 + }, + { + "epoch": 0.89, + "learning_rate": 6.749089409930609e-06, + "loss": 0.4824, + "step": 29757 + }, + { + "epoch": 0.89, + "learning_rate": 6.745607522230335e-06, + "loss": 0.5994, + "step": 29758 + }, + { + "epoch": 0.89, + "learning_rate": 6.742126501572232e-06, + "loss": 0.4938, + "step": 29759 + }, + { + "epoch": 0.89, + "learning_rate": 6.738646347988686e-06, + "loss": 0.4621, + "step": 29760 + }, + { + "epoch": 0.89, + "learning_rate": 6.7351670615120045e-06, + "loss": 0.5695, + "step": 29761 + }, + { + "epoch": 0.89, + "learning_rate": 6.731688642174571e-06, + "loss": 0.5583, + "step": 29762 + }, + { + "epoch": 0.89, + "learning_rate": 6.728211090008729e-06, + "loss": 0.4416, + "step": 29763 + }, + { + "epoch": 0.89, + "learning_rate": 6.7247344050468176e-06, + "loss": 0.3923, + "step": 29764 + }, + { + "epoch": 0.89, + "learning_rate": 6.721258587321155e-06, + "loss": 0.6217, + "step": 29765 + }, + { + "epoch": 0.89, + "learning_rate": 6.717783636864039e-06, + "loss": 0.5034, + "step": 29766 + }, + { + "epoch": 0.89, + "learning_rate": 6.714309553707787e-06, + "loss": 0.5261, + "step": 29767 + }, + { + "epoch": 0.89, + "learning_rate": 6.7108363378847185e-06, + "loss": 0.5146, + "step": 29768 + }, + { + "epoch": 0.89, + "learning_rate": 6.707363989427118e-06, + "loss": 0.4745, + "step": 29769 + }, + { + "epoch": 0.89, + "learning_rate": 6.703892508367238e-06, + "loss": 0.5605, + "step": 29770 + }, + { + "epoch": 0.89, + "learning_rate": 6.700421894737386e-06, + "loss": 0.4555, + "step": 29771 + }, + { + "epoch": 0.89, + "learning_rate": 6.696952148569846e-06, + "loss": 0.5298, + "step": 29772 + }, + { + "epoch": 0.89, + "learning_rate": 6.693483269896838e-06, + "loss": 0.5412, + "step": 29773 + }, + { + "epoch": 0.89, + "learning_rate": 6.69001525875066e-06, + "loss": 0.4538, + "step": 29774 + }, + { + "epoch": 0.89, + "learning_rate": 6.686548115163505e-06, + "loss": 0.5716, + "step": 29775 + }, + { + "epoch": 0.89, + "learning_rate": 6.6830818391676485e-06, + "loss": 0.4543, + "step": 29776 + }, + { + "epoch": 0.89, + "learning_rate": 6.679616430795299e-06, + "loss": 0.6165, + "step": 29777 + }, + { + "epoch": 0.89, + "learning_rate": 6.6761518900786854e-06, + "loss": 0.5641, + "step": 29778 + }, + { + "epoch": 0.89, + "learning_rate": 6.672688217050027e-06, + "loss": 0.4348, + "step": 29779 + }, + { + "epoch": 0.89, + "learning_rate": 6.669225411741497e-06, + "loss": 0.4964, + "step": 29780 + }, + { + "epoch": 0.89, + "learning_rate": 6.665763474185305e-06, + "loss": 0.5781, + "step": 29781 + }, + { + "epoch": 0.89, + "learning_rate": 6.6623024044136566e-06, + "loss": 0.4864, + "step": 29782 + }, + { + "epoch": 0.89, + "learning_rate": 6.658842202458726e-06, + "loss": 0.5747, + "step": 29783 + }, + { + "epoch": 0.89, + "learning_rate": 6.655382868352677e-06, + "loss": 0.5124, + "step": 29784 + }, + { + "epoch": 0.89, + "learning_rate": 6.651924402127663e-06, + "loss": 0.5219, + "step": 29785 + }, + { + "epoch": 0.89, + "learning_rate": 6.648466803815856e-06, + "loss": 0.6523, + "step": 29786 + }, + { + "epoch": 0.89, + "learning_rate": 6.645010073449409e-06, + "loss": 0.5053, + "step": 29787 + }, + { + "epoch": 0.89, + "learning_rate": 6.641554211060452e-06, + "loss": 0.5006, + "step": 29788 + }, + { + "epoch": 0.89, + "learning_rate": 6.638099216681093e-06, + "loss": 0.5451, + "step": 29789 + }, + { + "epoch": 0.89, + "learning_rate": 6.634645090343483e-06, + "loss": 0.5512, + "step": 29790 + }, + { + "epoch": 0.89, + "learning_rate": 6.6311918320797525e-06, + "loss": 0.4795, + "step": 29791 + }, + { + "epoch": 0.89, + "learning_rate": 6.627739441921965e-06, + "loss": 0.5492, + "step": 29792 + }, + { + "epoch": 0.89, + "learning_rate": 6.624287919902261e-06, + "loss": 0.5756, + "step": 29793 + }, + { + "epoch": 0.89, + "learning_rate": 6.620837266052693e-06, + "loss": 0.5899, + "step": 29794 + }, + { + "epoch": 0.89, + "learning_rate": 6.61738748040538e-06, + "loss": 0.5865, + "step": 29795 + }, + { + "epoch": 0.89, + "learning_rate": 6.613938562992361e-06, + "loss": 0.5627, + "step": 29796 + }, + { + "epoch": 0.89, + "learning_rate": 6.610490513845724e-06, + "loss": 0.5548, + "step": 29797 + }, + { + "epoch": 0.89, + "learning_rate": 6.607043332997531e-06, + "loss": 0.4532, + "step": 29798 + }, + { + "epoch": 0.89, + "learning_rate": 6.603597020479834e-06, + "loss": 0.5566, + "step": 29799 + }, + { + "epoch": 0.89, + "learning_rate": 6.600151576324654e-06, + "loss": 0.5576, + "step": 29800 + }, + { + "epoch": 0.89, + "learning_rate": 6.59670700056404e-06, + "loss": 0.4954, + "step": 29801 + }, + { + "epoch": 0.89, + "learning_rate": 6.5932632932300344e-06, + "loss": 0.6376, + "step": 29802 + }, + { + "epoch": 0.89, + "learning_rate": 6.589820454354634e-06, + "loss": 0.5014, + "step": 29803 + }, + { + "epoch": 0.89, + "learning_rate": 6.586378483969835e-06, + "loss": 0.5344, + "step": 29804 + }, + { + "epoch": 0.89, + "learning_rate": 6.582937382107668e-06, + "loss": 0.5849, + "step": 29805 + }, + { + "epoch": 0.89, + "learning_rate": 6.579497148800107e-06, + "loss": 0.5843, + "step": 29806 + }, + { + "epoch": 0.89, + "learning_rate": 6.57605778407917e-06, + "loss": 0.5383, + "step": 29807 + }, + { + "epoch": 0.89, + "learning_rate": 6.572619287976811e-06, + "loss": 0.4575, + "step": 29808 + }, + { + "epoch": 0.89, + "learning_rate": 6.569181660524981e-06, + "loss": 0.5538, + "step": 29809 + }, + { + "epoch": 0.89, + "learning_rate": 6.565744901755677e-06, + "loss": 0.4432, + "step": 29810 + }, + { + "epoch": 0.89, + "learning_rate": 6.562309011700829e-06, + "loss": 0.5682, + "step": 29811 + }, + { + "epoch": 0.89, + "learning_rate": 6.5588739903924e-06, + "loss": 0.4325, + "step": 29812 + }, + { + "epoch": 0.89, + "learning_rate": 6.5554398378623095e-06, + "loss": 0.6195, + "step": 29813 + }, + { + "epoch": 0.89, + "learning_rate": 6.552006554142509e-06, + "loss": 0.4066, + "step": 29814 + }, + { + "epoch": 0.89, + "learning_rate": 6.548574139264896e-06, + "loss": 0.5394, + "step": 29815 + }, + { + "epoch": 0.89, + "learning_rate": 6.5451425932613885e-06, + "loss": 0.4766, + "step": 29816 + }, + { + "epoch": 0.89, + "learning_rate": 6.541711916163917e-06, + "loss": 0.465, + "step": 29817 + }, + { + "epoch": 0.89, + "learning_rate": 6.538282108004357e-06, + "loss": 0.5011, + "step": 29818 + }, + { + "epoch": 0.89, + "learning_rate": 6.534853168814581e-06, + "loss": 0.4592, + "step": 29819 + }, + { + "epoch": 0.89, + "learning_rate": 6.531425098626498e-06, + "loss": 0.4899, + "step": 29820 + }, + { + "epoch": 0.89, + "learning_rate": 6.527997897471982e-06, + "loss": 0.5444, + "step": 29821 + }, + { + "epoch": 0.89, + "learning_rate": 6.524571565382886e-06, + "loss": 0.493, + "step": 29822 + }, + { + "epoch": 0.89, + "learning_rate": 6.521146102391051e-06, + "loss": 0.4865, + "step": 29823 + }, + { + "epoch": 0.89, + "learning_rate": 6.51772150852834e-06, + "loss": 0.5858, + "step": 29824 + }, + { + "epoch": 0.89, + "learning_rate": 6.5142977838266065e-06, + "loss": 0.5491, + "step": 29825 + }, + { + "epoch": 0.89, + "learning_rate": 6.510874928317678e-06, + "loss": 0.5132, + "step": 29826 + }, + { + "epoch": 0.89, + "learning_rate": 6.507452942033376e-06, + "loss": 0.5553, + "step": 29827 + }, + { + "epoch": 0.89, + "learning_rate": 6.504031825005497e-06, + "loss": 0.5474, + "step": 29828 + }, + { + "epoch": 0.89, + "learning_rate": 6.500611577265869e-06, + "loss": 0.6008, + "step": 29829 + }, + { + "epoch": 0.89, + "learning_rate": 6.497192198846314e-06, + "loss": 0.4817, + "step": 29830 + }, + { + "epoch": 0.89, + "learning_rate": 6.493773689778582e-06, + "loss": 0.5365, + "step": 29831 + }, + { + "epoch": 0.89, + "learning_rate": 6.490356050094471e-06, + "loss": 0.4682, + "step": 29832 + }, + { + "epoch": 0.89, + "learning_rate": 6.486939279825754e-06, + "loss": 0.4908, + "step": 29833 + }, + { + "epoch": 0.89, + "learning_rate": 6.48352337900423e-06, + "loss": 0.4928, + "step": 29834 + }, + { + "epoch": 0.89, + "learning_rate": 6.480108347661606e-06, + "loss": 0.6057, + "step": 29835 + }, + { + "epoch": 0.89, + "learning_rate": 6.4766941858296795e-06, + "loss": 0.4943, + "step": 29836 + }, + { + "epoch": 0.89, + "learning_rate": 6.473280893540168e-06, + "loss": 0.4898, + "step": 29837 + }, + { + "epoch": 0.89, + "learning_rate": 6.4698684708248244e-06, + "loss": 0.6012, + "step": 29838 + }, + { + "epoch": 0.89, + "learning_rate": 6.466456917715347e-06, + "loss": 0.5361, + "step": 29839 + }, + { + "epoch": 0.89, + "learning_rate": 6.463046234243486e-06, + "loss": 0.5558, + "step": 29840 + }, + { + "epoch": 0.89, + "learning_rate": 6.4596364204409394e-06, + "loss": 0.4621, + "step": 29841 + }, + { + "epoch": 0.89, + "learning_rate": 6.456227476339427e-06, + "loss": 0.591, + "step": 29842 + }, + { + "epoch": 0.89, + "learning_rate": 6.4528194019706115e-06, + "loss": 0.5313, + "step": 29843 + }, + { + "epoch": 0.89, + "learning_rate": 6.449412197366189e-06, + "loss": 0.5456, + "step": 29844 + }, + { + "epoch": 0.89, + "learning_rate": 6.446005862557869e-06, + "loss": 0.451, + "step": 29845 + }, + { + "epoch": 0.89, + "learning_rate": 6.442600397577292e-06, + "loss": 0.4845, + "step": 29846 + }, + { + "epoch": 0.89, + "learning_rate": 6.439195802456121e-06, + "loss": 0.6375, + "step": 29847 + }, + { + "epoch": 0.89, + "learning_rate": 6.435792077226011e-06, + "loss": 0.53, + "step": 29848 + }, + { + "epoch": 0.89, + "learning_rate": 6.432389221918633e-06, + "loss": 0.4774, + "step": 29849 + }, + { + "epoch": 0.89, + "learning_rate": 6.428987236565598e-06, + "loss": 0.5486, + "step": 29850 + }, + { + "epoch": 0.89, + "learning_rate": 6.425586121198557e-06, + "loss": 0.4604, + "step": 29851 + }, + { + "epoch": 0.89, + "learning_rate": 6.422185875849096e-06, + "loss": 0.5584, + "step": 29852 + }, + { + "epoch": 0.89, + "learning_rate": 6.418786500548879e-06, + "loss": 0.5928, + "step": 29853 + }, + { + "epoch": 0.89, + "learning_rate": 6.415387995329481e-06, + "loss": 0.5714, + "step": 29854 + }, + { + "epoch": 0.89, + "learning_rate": 6.411990360222508e-06, + "loss": 0.462, + "step": 29855 + }, + { + "epoch": 0.89, + "learning_rate": 6.408593595259537e-06, + "loss": 0.4824, + "step": 29856 + }, + { + "epoch": 0.89, + "learning_rate": 6.405197700472177e-06, + "loss": 0.5566, + "step": 29857 + }, + { + "epoch": 0.89, + "learning_rate": 6.401802675891965e-06, + "loss": 0.5409, + "step": 29858 + }, + { + "epoch": 0.89, + "learning_rate": 6.398408521550492e-06, + "loss": 0.5014, + "step": 29859 + }, + { + "epoch": 0.89, + "learning_rate": 6.395015237479329e-06, + "loss": 0.5622, + "step": 29860 + }, + { + "epoch": 0.89, + "learning_rate": 6.391622823710009e-06, + "loss": 0.477, + "step": 29861 + }, + { + "epoch": 0.89, + "learning_rate": 6.388231280274048e-06, + "loss": 0.5274, + "step": 29862 + }, + { + "epoch": 0.89, + "learning_rate": 6.384840607203002e-06, + "loss": 0.4371, + "step": 29863 + }, + { + "epoch": 0.89, + "learning_rate": 6.38145080452841e-06, + "loss": 0.4979, + "step": 29864 + }, + { + "epoch": 0.89, + "learning_rate": 6.378061872281782e-06, + "loss": 0.5698, + "step": 29865 + }, + { + "epoch": 0.89, + "learning_rate": 6.374673810494614e-06, + "loss": 0.5919, + "step": 29866 + }, + { + "epoch": 0.89, + "learning_rate": 6.371286619198402e-06, + "loss": 0.4739, + "step": 29867 + }, + { + "epoch": 0.89, + "learning_rate": 6.367900298424645e-06, + "loss": 0.6355, + "step": 29868 + }, + { + "epoch": 0.89, + "learning_rate": 6.364514848204861e-06, + "loss": 0.5615, + "step": 29869 + }, + { + "epoch": 0.89, + "learning_rate": 6.361130268570492e-06, + "loss": 0.5899, + "step": 29870 + }, + { + "epoch": 0.89, + "learning_rate": 6.357746559553002e-06, + "loss": 0.6663, + "step": 29871 + }, + { + "epoch": 0.89, + "learning_rate": 6.354363721183853e-06, + "loss": 0.5175, + "step": 29872 + }, + { + "epoch": 0.89, + "learning_rate": 6.3509817534945334e-06, + "loss": 0.5089, + "step": 29873 + }, + { + "epoch": 0.89, + "learning_rate": 6.34760065651645e-06, + "loss": 0.5582, + "step": 29874 + }, + { + "epoch": 0.89, + "learning_rate": 6.344220430281045e-06, + "loss": 0.4453, + "step": 29875 + }, + { + "epoch": 0.89, + "learning_rate": 6.3408410748197585e-06, + "loss": 0.6528, + "step": 29876 + }, + { + "epoch": 0.89, + "learning_rate": 6.33746259016399e-06, + "loss": 0.6586, + "step": 29877 + }, + { + "epoch": 0.89, + "learning_rate": 6.334084976345167e-06, + "loss": 0.5217, + "step": 29878 + }, + { + "epoch": 0.89, + "learning_rate": 6.3307082333946995e-06, + "loss": 0.4092, + "step": 29879 + }, + { + "epoch": 0.89, + "learning_rate": 6.327332361343974e-06, + "loss": 0.5092, + "step": 29880 + }, + { + "epoch": 0.89, + "learning_rate": 6.323957360224375e-06, + "loss": 0.5559, + "step": 29881 + }, + { + "epoch": 0.89, + "learning_rate": 6.320583230067268e-06, + "loss": 0.5449, + "step": 29882 + }, + { + "epoch": 0.89, + "learning_rate": 6.31720997090407e-06, + "loss": 0.4101, + "step": 29883 + }, + { + "epoch": 0.89, + "learning_rate": 6.313837582766102e-06, + "loss": 0.6071, + "step": 29884 + }, + { + "epoch": 0.89, + "learning_rate": 6.310466065684728e-06, + "loss": 0.5306, + "step": 29885 + }, + { + "epoch": 0.89, + "learning_rate": 6.3070954196913e-06, + "loss": 0.4724, + "step": 29886 + }, + { + "epoch": 0.89, + "learning_rate": 6.3037256448171485e-06, + "loss": 0.4071, + "step": 29887 + }, + { + "epoch": 0.89, + "learning_rate": 6.300356741093638e-06, + "loss": 0.4864, + "step": 29888 + }, + { + "epoch": 0.89, + "learning_rate": 6.296988708552054e-06, + "loss": 0.5862, + "step": 29889 + }, + { + "epoch": 0.89, + "learning_rate": 6.2936215472237155e-06, + "loss": 0.6016, + "step": 29890 + }, + { + "epoch": 0.89, + "learning_rate": 6.290255257139932e-06, + "loss": 0.4635, + "step": 29891 + }, + { + "epoch": 0.89, + "learning_rate": 6.286889838332022e-06, + "loss": 0.6154, + "step": 29892 + }, + { + "epoch": 0.89, + "learning_rate": 6.2835252908312605e-06, + "loss": 0.6831, + "step": 29893 + }, + { + "epoch": 0.89, + "learning_rate": 6.2801616146689116e-06, + "loss": 0.5008, + "step": 29894 + }, + { + "epoch": 0.89, + "learning_rate": 6.276798809876272e-06, + "loss": 0.4443, + "step": 29895 + }, + { + "epoch": 0.89, + "learning_rate": 6.2734368764846175e-06, + "loss": 0.4414, + "step": 29896 + }, + { + "epoch": 0.89, + "learning_rate": 6.270075814525178e-06, + "loss": 0.5054, + "step": 29897 + }, + { + "epoch": 0.89, + "learning_rate": 6.266715624029229e-06, + "loss": 0.4654, + "step": 29898 + }, + { + "epoch": 0.89, + "learning_rate": 6.263356305027979e-06, + "loss": 0.4101, + "step": 29899 + }, + { + "epoch": 0.89, + "learning_rate": 6.259997857552702e-06, + "loss": 0.645, + "step": 29900 + }, + { + "epoch": 0.89, + "learning_rate": 6.256640281634585e-06, + "loss": 0.4871, + "step": 29901 + }, + { + "epoch": 0.89, + "learning_rate": 6.253283577304869e-06, + "loss": 0.4589, + "step": 29902 + }, + { + "epoch": 0.89, + "learning_rate": 6.249927744594775e-06, + "loss": 0.4814, + "step": 29903 + }, + { + "epoch": 0.89, + "learning_rate": 6.246572783535487e-06, + "loss": 0.4659, + "step": 29904 + }, + { + "epoch": 0.89, + "learning_rate": 6.243218694158182e-06, + "loss": 0.5913, + "step": 29905 + }, + { + "epoch": 0.89, + "learning_rate": 6.2398654764940556e-06, + "loss": 0.5114, + "step": 29906 + }, + { + "epoch": 0.89, + "learning_rate": 6.2365131305743175e-06, + "loss": 0.4721, + "step": 29907 + }, + { + "epoch": 0.89, + "learning_rate": 6.233161656430098e-06, + "loss": 0.4831, + "step": 29908 + }, + { + "epoch": 0.89, + "learning_rate": 6.22981105409256e-06, + "loss": 0.5145, + "step": 29909 + }, + { + "epoch": 0.89, + "learning_rate": 6.226461323592869e-06, + "loss": 0.414, + "step": 29910 + }, + { + "epoch": 0.89, + "learning_rate": 6.223112464962178e-06, + "loss": 0.4831, + "step": 29911 + }, + { + "epoch": 0.89, + "learning_rate": 6.2197644782316045e-06, + "loss": 0.5999, + "step": 29912 + }, + { + "epoch": 0.89, + "learning_rate": 6.216417363432292e-06, + "loss": 0.4226, + "step": 29913 + }, + { + "epoch": 0.89, + "learning_rate": 6.213071120595337e-06, + "loss": 0.6137, + "step": 29914 + }, + { + "epoch": 0.89, + "learning_rate": 6.2097257497518825e-06, + "loss": 0.5952, + "step": 29915 + }, + { + "epoch": 0.89, + "learning_rate": 6.2063812509330025e-06, + "loss": 0.5604, + "step": 29916 + }, + { + "epoch": 0.89, + "learning_rate": 6.203037624169816e-06, + "loss": 0.4515, + "step": 29917 + }, + { + "epoch": 0.89, + "learning_rate": 6.199694869493389e-06, + "loss": 0.5639, + "step": 29918 + }, + { + "epoch": 0.89, + "learning_rate": 6.196352986934839e-06, + "loss": 0.6054, + "step": 29919 + }, + { + "epoch": 0.89, + "learning_rate": 6.193011976525187e-06, + "loss": 0.5205, + "step": 29920 + }, + { + "epoch": 0.89, + "learning_rate": 6.189671838295519e-06, + "loss": 0.5594, + "step": 29921 + }, + { + "epoch": 0.89, + "learning_rate": 6.186332572276921e-06, + "loss": 0.4995, + "step": 29922 + }, + { + "epoch": 0.89, + "learning_rate": 6.182994178500401e-06, + "loss": 0.4684, + "step": 29923 + }, + { + "epoch": 0.89, + "learning_rate": 6.179656656997002e-06, + "loss": 0.4434, + "step": 29924 + }, + { + "epoch": 0.89, + "learning_rate": 6.176320007797764e-06, + "loss": 0.5664, + "step": 29925 + }, + { + "epoch": 0.89, + "learning_rate": 6.17298423093371e-06, + "loss": 0.4717, + "step": 29926 + }, + { + "epoch": 0.89, + "learning_rate": 6.169649326435867e-06, + "loss": 0.5187, + "step": 29927 + }, + { + "epoch": 0.89, + "learning_rate": 6.166315294335201e-06, + "loss": 0.486, + "step": 29928 + }, + { + "epoch": 0.89, + "learning_rate": 6.162982134662742e-06, + "loss": 0.5482, + "step": 29929 + }, + { + "epoch": 0.89, + "learning_rate": 6.1596498474494776e-06, + "loss": 0.5652, + "step": 29930 + }, + { + "epoch": 0.89, + "learning_rate": 6.156318432726404e-06, + "loss": 0.5017, + "step": 29931 + }, + { + "epoch": 0.89, + "learning_rate": 6.152987890524475e-06, + "loss": 0.4884, + "step": 29932 + }, + { + "epoch": 0.89, + "learning_rate": 6.149658220874654e-06, + "loss": 0.5433, + "step": 29933 + }, + { + "epoch": 0.89, + "learning_rate": 6.146329423807895e-06, + "loss": 0.4627, + "step": 29934 + }, + { + "epoch": 0.89, + "learning_rate": 6.143001499355183e-06, + "loss": 0.6324, + "step": 29935 + }, + { + "epoch": 0.89, + "learning_rate": 6.139674447547428e-06, + "loss": 0.4324, + "step": 29936 + }, + { + "epoch": 0.89, + "learning_rate": 6.13634826841556e-06, + "loss": 0.5775, + "step": 29937 + }, + { + "epoch": 0.89, + "learning_rate": 6.133022961990509e-06, + "loss": 0.4613, + "step": 29938 + }, + { + "epoch": 0.89, + "learning_rate": 6.129698528303218e-06, + "loss": 0.5653, + "step": 29939 + }, + { + "epoch": 0.89, + "learning_rate": 6.126374967384563e-06, + "loss": 0.5563, + "step": 29940 + }, + { + "epoch": 0.89, + "learning_rate": 6.123052279265473e-06, + "loss": 0.6043, + "step": 29941 + }, + { + "epoch": 0.89, + "learning_rate": 6.119730463976836e-06, + "loss": 0.6459, + "step": 29942 + }, + { + "epoch": 0.89, + "learning_rate": 6.116409521549505e-06, + "loss": 0.5514, + "step": 29943 + }, + { + "epoch": 0.89, + "learning_rate": 6.113089452014375e-06, + "loss": 0.4896, + "step": 29944 + }, + { + "epoch": 0.89, + "learning_rate": 6.109770255402336e-06, + "loss": 0.3894, + "step": 29945 + }, + { + "epoch": 0.89, + "learning_rate": 6.106451931744239e-06, + "loss": 0.4825, + "step": 29946 + }, + { + "epoch": 0.89, + "learning_rate": 6.103134481070915e-06, + "loss": 0.517, + "step": 29947 + }, + { + "epoch": 0.89, + "learning_rate": 6.099817903413219e-06, + "loss": 0.5437, + "step": 29948 + }, + { + "epoch": 0.89, + "learning_rate": 6.09650219880199e-06, + "loss": 0.4377, + "step": 29949 + }, + { + "epoch": 0.89, + "learning_rate": 6.0931873672680715e-06, + "loss": 0.4974, + "step": 29950 + }, + { + "epoch": 0.89, + "learning_rate": 6.089873408842273e-06, + "loss": 0.6189, + "step": 29951 + }, + { + "epoch": 0.89, + "learning_rate": 6.086560323555391e-06, + "loss": 0.4331, + "step": 29952 + }, + { + "epoch": 0.89, + "learning_rate": 6.083248111438233e-06, + "loss": 0.589, + "step": 29953 + }, + { + "epoch": 0.89, + "learning_rate": 6.0799367725216215e-06, + "loss": 0.5523, + "step": 29954 + }, + { + "epoch": 0.89, + "learning_rate": 6.076626306836331e-06, + "loss": 0.4895, + "step": 29955 + }, + { + "epoch": 0.89, + "learning_rate": 6.0733167144131134e-06, + "loss": 0.5396, + "step": 29956 + }, + { + "epoch": 0.89, + "learning_rate": 6.070007995282756e-06, + "loss": 0.5588, + "step": 29957 + }, + { + "epoch": 0.89, + "learning_rate": 6.0667001494760455e-06, + "loss": 0.5087, + "step": 29958 + }, + { + "epoch": 0.89, + "learning_rate": 6.063393177023713e-06, + "loss": 0.4868, + "step": 29959 + }, + { + "epoch": 0.89, + "learning_rate": 6.060087077956522e-06, + "loss": 0.4185, + "step": 29960 + }, + { + "epoch": 0.89, + "learning_rate": 6.056781852305183e-06, + "loss": 0.5068, + "step": 29961 + }, + { + "epoch": 0.89, + "learning_rate": 6.053477500100457e-06, + "loss": 0.4713, + "step": 29962 + }, + { + "epoch": 0.89, + "learning_rate": 6.050174021373045e-06, + "loss": 0.6484, + "step": 29963 + }, + { + "epoch": 0.89, + "learning_rate": 6.0468714161536655e-06, + "loss": 0.603, + "step": 29964 + }, + { + "epoch": 0.89, + "learning_rate": 6.043569684473049e-06, + "loss": 0.5515, + "step": 29965 + }, + { + "epoch": 0.89, + "learning_rate": 6.040268826361862e-06, + "loss": 0.4987, + "step": 29966 + }, + { + "epoch": 0.89, + "learning_rate": 6.03696884185081e-06, + "loss": 0.4483, + "step": 29967 + }, + { + "epoch": 0.89, + "learning_rate": 6.03366973097056e-06, + "loss": 0.4877, + "step": 29968 + }, + { + "epoch": 0.89, + "learning_rate": 6.0303714937518185e-06, + "loss": 0.4425, + "step": 29969 + }, + { + "epoch": 0.89, + "learning_rate": 6.027074130225219e-06, + "loss": 0.5146, + "step": 29970 + }, + { + "epoch": 0.89, + "learning_rate": 6.023777640421424e-06, + "loss": 0.3742, + "step": 29971 + }, + { + "epoch": 0.89, + "learning_rate": 6.020482024371088e-06, + "loss": 0.4715, + "step": 29972 + }, + { + "epoch": 0.89, + "learning_rate": 6.0171872821048745e-06, + "loss": 0.6057, + "step": 29973 + }, + { + "epoch": 0.89, + "learning_rate": 6.013893413653371e-06, + "loss": 0.5927, + "step": 29974 + }, + { + "epoch": 0.89, + "learning_rate": 6.010600419047252e-06, + "loss": 0.448, + "step": 29975 + }, + { + "epoch": 0.89, + "learning_rate": 6.007308298317083e-06, + "loss": 0.4924, + "step": 29976 + }, + { + "epoch": 0.89, + "learning_rate": 6.004017051493527e-06, + "loss": 0.5362, + "step": 29977 + }, + { + "epoch": 0.89, + "learning_rate": 6.000726678607138e-06, + "loss": 0.5761, + "step": 29978 + }, + { + "epoch": 0.89, + "learning_rate": 5.997437179688537e-06, + "loss": 0.6343, + "step": 29979 + }, + { + "epoch": 0.89, + "learning_rate": 5.9941485547682865e-06, + "loss": 0.4725, + "step": 29980 + }, + { + "epoch": 0.89, + "learning_rate": 5.990860803876997e-06, + "loss": 0.5883, + "step": 29981 + }, + { + "epoch": 0.89, + "learning_rate": 5.987573927045198e-06, + "loss": 0.5371, + "step": 29982 + }, + { + "epoch": 0.89, + "learning_rate": 5.984287924303467e-06, + "loss": 0.4666, + "step": 29983 + }, + { + "epoch": 0.89, + "learning_rate": 5.9810027956823664e-06, + "loss": 0.5799, + "step": 29984 + }, + { + "epoch": 0.89, + "learning_rate": 5.977718541212429e-06, + "loss": 0.4662, + "step": 29985 + }, + { + "epoch": 0.89, + "learning_rate": 5.974435160924185e-06, + "loss": 0.5149, + "step": 29986 + }, + { + "epoch": 0.89, + "learning_rate": 5.971152654848166e-06, + "loss": 0.5534, + "step": 29987 + }, + { + "epoch": 0.89, + "learning_rate": 5.967871023014904e-06, + "loss": 0.5238, + "step": 29988 + }, + { + "epoch": 0.89, + "learning_rate": 5.964590265454895e-06, + "loss": 0.623, + "step": 29989 + }, + { + "epoch": 0.89, + "learning_rate": 5.9613103821986374e-06, + "loss": 0.573, + "step": 29990 + }, + { + "epoch": 0.89, + "learning_rate": 5.95803137327664e-06, + "loss": 0.4488, + "step": 29991 + }, + { + "epoch": 0.89, + "learning_rate": 5.954753238719379e-06, + "loss": 0.5676, + "step": 29992 + }, + { + "epoch": 0.89, + "learning_rate": 5.951475978557363e-06, + "loss": 0.4239, + "step": 29993 + }, + { + "epoch": 0.89, + "learning_rate": 5.948199592821024e-06, + "loss": 0.4468, + "step": 29994 + }, + { + "epoch": 0.89, + "learning_rate": 5.9449240815408365e-06, + "loss": 0.4881, + "step": 29995 + }, + { + "epoch": 0.89, + "learning_rate": 5.941649444747255e-06, + "loss": 0.584, + "step": 29996 + }, + { + "epoch": 0.89, + "learning_rate": 5.938375682470754e-06, + "loss": 0.5396, + "step": 29997 + }, + { + "epoch": 0.89, + "learning_rate": 5.935102794741731e-06, + "loss": 0.584, + "step": 29998 + }, + { + "epoch": 0.89, + "learning_rate": 5.93183078159063e-06, + "loss": 0.5803, + "step": 29999 + }, + { + "epoch": 0.89, + "learning_rate": 5.92855964304786e-06, + "loss": 0.5339, + "step": 30000 + }, + { + "epoch": 0.89, + "learning_rate": 5.925289379143872e-06, + "loss": 0.4429, + "step": 30001 + }, + { + "epoch": 0.89, + "learning_rate": 5.922019989909033e-06, + "loss": 0.5612, + "step": 30002 + }, + { + "epoch": 0.89, + "learning_rate": 5.918751475373774e-06, + "loss": 0.4764, + "step": 30003 + }, + { + "epoch": 0.89, + "learning_rate": 5.9154838355684475e-06, + "loss": 0.6829, + "step": 30004 + }, + { + "epoch": 0.89, + "learning_rate": 5.912217070523473e-06, + "loss": 0.5515, + "step": 30005 + }, + { + "epoch": 0.89, + "learning_rate": 5.9089511802691956e-06, + "loss": 0.5493, + "step": 30006 + }, + { + "epoch": 0.89, + "learning_rate": 5.905686164835978e-06, + "loss": 0.5633, + "step": 30007 + }, + { + "epoch": 0.89, + "learning_rate": 5.902422024254217e-06, + "loss": 0.4672, + "step": 30008 + }, + { + "epoch": 0.89, + "learning_rate": 5.899158758554213e-06, + "loss": 0.6084, + "step": 30009 + }, + { + "epoch": 0.89, + "learning_rate": 5.895896367766318e-06, + "loss": 0.5321, + "step": 30010 + }, + { + "epoch": 0.89, + "learning_rate": 5.892634851920875e-06, + "loss": 0.5412, + "step": 30011 + }, + { + "epoch": 0.89, + "learning_rate": 5.889374211048215e-06, + "loss": 0.5506, + "step": 30012 + }, + { + "epoch": 0.89, + "learning_rate": 5.886114445178659e-06, + "loss": 0.5859, + "step": 30013 + }, + { + "epoch": 0.89, + "learning_rate": 5.88285555434247e-06, + "loss": 0.5556, + "step": 30014 + }, + { + "epoch": 0.89, + "learning_rate": 5.879597538569992e-06, + "loss": 0.4698, + "step": 30015 + }, + { + "epoch": 0.89, + "learning_rate": 5.876340397891511e-06, + "loss": 0.4392, + "step": 30016 + }, + { + "epoch": 0.89, + "learning_rate": 5.873084132337281e-06, + "loss": 0.5137, + "step": 30017 + }, + { + "epoch": 0.89, + "learning_rate": 5.8698287419376215e-06, + "loss": 0.5517, + "step": 30018 + }, + { + "epoch": 0.89, + "learning_rate": 5.866574226722765e-06, + "loss": 0.3623, + "step": 30019 + }, + { + "epoch": 0.89, + "learning_rate": 5.863320586722998e-06, + "loss": 0.5343, + "step": 30020 + }, + { + "epoch": 0.89, + "learning_rate": 5.86006782196854e-06, + "loss": 0.4969, + "step": 30021 + }, + { + "epoch": 0.89, + "learning_rate": 5.856815932489667e-06, + "loss": 0.5237, + "step": 30022 + }, + { + "epoch": 0.89, + "learning_rate": 5.85356491831659e-06, + "loss": 0.7208, + "step": 30023 + }, + { + "epoch": 0.89, + "learning_rate": 5.850314779479549e-06, + "loss": 0.6039, + "step": 30024 + }, + { + "epoch": 0.89, + "learning_rate": 5.847065516008743e-06, + "loss": 0.5044, + "step": 30025 + }, + { + "epoch": 0.89, + "learning_rate": 5.843817127934392e-06, + "loss": 0.5563, + "step": 30026 + }, + { + "epoch": 0.89, + "learning_rate": 5.840569615286728e-06, + "loss": 0.5101, + "step": 30027 + }, + { + "epoch": 0.89, + "learning_rate": 5.837322978095905e-06, + "loss": 0.602, + "step": 30028 + }, + { + "epoch": 0.89, + "learning_rate": 5.83407721639212e-06, + "loss": 0.5251, + "step": 30029 + }, + { + "epoch": 0.89, + "learning_rate": 5.830832330205549e-06, + "loss": 0.5819, + "step": 30030 + }, + { + "epoch": 0.89, + "learning_rate": 5.82758831956639e-06, + "loss": 0.4498, + "step": 30031 + }, + { + "epoch": 0.89, + "learning_rate": 5.824345184504776e-06, + "loss": 0.5493, + "step": 30032 + }, + { + "epoch": 0.89, + "learning_rate": 5.821102925050848e-06, + "loss": 0.4794, + "step": 30033 + }, + { + "epoch": 0.89, + "learning_rate": 5.81786154123477e-06, + "loss": 0.5521, + "step": 30034 + }, + { + "epoch": 0.89, + "learning_rate": 5.814621033086687e-06, + "loss": 0.5206, + "step": 30035 + }, + { + "epoch": 0.89, + "learning_rate": 5.8113814006367175e-06, + "loss": 0.5361, + "step": 30036 + }, + { + "epoch": 0.89, + "learning_rate": 5.808142643914993e-06, + "loss": 0.5705, + "step": 30037 + }, + { + "epoch": 0.89, + "learning_rate": 5.804904762951602e-06, + "loss": 0.4852, + "step": 30038 + }, + { + "epoch": 0.89, + "learning_rate": 5.801667757776686e-06, + "loss": 0.5165, + "step": 30039 + }, + { + "epoch": 0.89, + "learning_rate": 5.798431628420298e-06, + "loss": 0.495, + "step": 30040 + }, + { + "epoch": 0.89, + "learning_rate": 5.795196374912559e-06, + "loss": 0.3627, + "step": 30041 + }, + { + "epoch": 0.89, + "learning_rate": 5.791961997283524e-06, + "loss": 0.5255, + "step": 30042 + }, + { + "epoch": 0.89, + "learning_rate": 5.7887284955633005e-06, + "loss": 0.4558, + "step": 30043 + }, + { + "epoch": 0.89, + "learning_rate": 5.78549586978191e-06, + "loss": 0.6395, + "step": 30044 + }, + { + "epoch": 0.89, + "learning_rate": 5.782264119969427e-06, + "loss": 0.5404, + "step": 30045 + }, + { + "epoch": 0.89, + "learning_rate": 5.779033246155918e-06, + "loss": 0.4729, + "step": 30046 + }, + { + "epoch": 0.89, + "learning_rate": 5.775803248371403e-06, + "loss": 0.5342, + "step": 30047 + }, + { + "epoch": 0.89, + "learning_rate": 5.772574126645902e-06, + "loss": 0.476, + "step": 30048 + }, + { + "epoch": 0.89, + "learning_rate": 5.769345881009447e-06, + "loss": 0.5305, + "step": 30049 + }, + { + "epoch": 0.89, + "learning_rate": 5.76611851149208e-06, + "loss": 0.5253, + "step": 30050 + }, + { + "epoch": 0.89, + "learning_rate": 5.762892018123789e-06, + "loss": 0.4696, + "step": 30051 + }, + { + "epoch": 0.89, + "learning_rate": 5.759666400934538e-06, + "loss": 0.5593, + "step": 30052 + }, + { + "epoch": 0.89, + "learning_rate": 5.756441659954359e-06, + "loss": 0.6195, + "step": 30053 + }, + { + "epoch": 0.89, + "learning_rate": 5.753217795213228e-06, + "loss": 0.4876, + "step": 30054 + }, + { + "epoch": 0.89, + "learning_rate": 5.749994806741122e-06, + "loss": 0.5623, + "step": 30055 + }, + { + "epoch": 0.89, + "learning_rate": 5.746772694568003e-06, + "loss": 0.4917, + "step": 30056 + }, + { + "epoch": 0.89, + "learning_rate": 5.743551458723806e-06, + "loss": 0.4982, + "step": 30057 + }, + { + "epoch": 0.89, + "learning_rate": 5.740331099238505e-06, + "loss": 0.5933, + "step": 30058 + }, + { + "epoch": 0.89, + "learning_rate": 5.7371116161420545e-06, + "loss": 0.5071, + "step": 30059 + }, + { + "epoch": 0.89, + "learning_rate": 5.733893009464364e-06, + "loss": 0.5609, + "step": 30060 + }, + { + "epoch": 0.89, + "learning_rate": 5.730675279235354e-06, + "loss": 0.6709, + "step": 30061 + }, + { + "epoch": 0.89, + "learning_rate": 5.727458425484955e-06, + "loss": 0.6175, + "step": 30062 + }, + { + "epoch": 0.9, + "learning_rate": 5.724242448243089e-06, + "loss": 0.4777, + "step": 30063 + }, + { + "epoch": 0.9, + "learning_rate": 5.721027347539631e-06, + "loss": 0.5662, + "step": 30064 + }, + { + "epoch": 0.9, + "learning_rate": 5.717813123404492e-06, + "loss": 0.4627, + "step": 30065 + }, + { + "epoch": 0.9, + "learning_rate": 5.714599775867535e-06, + "loss": 0.475, + "step": 30066 + }, + { + "epoch": 0.9, + "learning_rate": 5.711387304958671e-06, + "loss": 0.5785, + "step": 30067 + }, + { + "epoch": 0.9, + "learning_rate": 5.708175710707731e-06, + "loss": 0.4908, + "step": 30068 + }, + { + "epoch": 0.9, + "learning_rate": 5.70496499314459e-06, + "loss": 0.43, + "step": 30069 + }, + { + "epoch": 0.9, + "learning_rate": 5.7017551522991154e-06, + "loss": 0.6494, + "step": 30070 + }, + { + "epoch": 0.9, + "learning_rate": 5.698546188201148e-06, + "loss": 0.6039, + "step": 30071 + }, + { + "epoch": 0.9, + "learning_rate": 5.695338100880487e-06, + "loss": 0.5074, + "step": 30072 + }, + { + "epoch": 0.9, + "learning_rate": 5.6921308903669865e-06, + "loss": 0.5035, + "step": 30073 + }, + { + "epoch": 0.9, + "learning_rate": 5.688924556690489e-06, + "loss": 0.5125, + "step": 30074 + }, + { + "epoch": 0.9, + "learning_rate": 5.685719099880782e-06, + "loss": 0.5551, + "step": 30075 + }, + { + "epoch": 0.9, + "learning_rate": 5.682514519967641e-06, + "loss": 0.4338, + "step": 30076 + }, + { + "epoch": 0.9, + "learning_rate": 5.679310816980898e-06, + "loss": 0.4697, + "step": 30077 + }, + { + "epoch": 0.9, + "learning_rate": 5.6761079909503525e-06, + "loss": 0.5649, + "step": 30078 + }, + { + "epoch": 0.9, + "learning_rate": 5.672906041905735e-06, + "loss": 0.5186, + "step": 30079 + }, + { + "epoch": 0.9, + "learning_rate": 5.669704969876865e-06, + "loss": 0.6162, + "step": 30080 + }, + { + "epoch": 0.9, + "learning_rate": 5.666504774893455e-06, + "loss": 0.6151, + "step": 30081 + }, + { + "epoch": 0.9, + "learning_rate": 5.663305456985313e-06, + "loss": 0.5336, + "step": 30082 + }, + { + "epoch": 0.9, + "learning_rate": 5.660107016182137e-06, + "loss": 0.5902, + "step": 30083 + }, + { + "epoch": 0.9, + "learning_rate": 5.6569094525137034e-06, + "loss": 0.6058, + "step": 30084 + }, + { + "epoch": 0.9, + "learning_rate": 5.653712766009711e-06, + "loss": 0.5755, + "step": 30085 + }, + { + "epoch": 0.9, + "learning_rate": 5.650516956699903e-06, + "loss": 0.5362, + "step": 30086 + }, + { + "epoch": 0.9, + "learning_rate": 5.6473220246139765e-06, + "loss": 0.394, + "step": 30087 + }, + { + "epoch": 0.9, + "learning_rate": 5.644127969781643e-06, + "loss": 0.5332, + "step": 30088 + }, + { + "epoch": 0.9, + "learning_rate": 5.640934792232611e-06, + "loss": 0.6272, + "step": 30089 + }, + { + "epoch": 0.9, + "learning_rate": 5.637742491996556e-06, + "loss": 0.4985, + "step": 30090 + }, + { + "epoch": 0.9, + "learning_rate": 5.6345510691031556e-06, + "loss": 0.6123, + "step": 30091 + }, + { + "epoch": 0.9, + "learning_rate": 5.631360523582096e-06, + "loss": 0.5167, + "step": 30092 + }, + { + "epoch": 0.9, + "learning_rate": 5.628170855463033e-06, + "loss": 0.4392, + "step": 30093 + }, + { + "epoch": 0.9, + "learning_rate": 5.624982064775641e-06, + "loss": 0.4543, + "step": 30094 + }, + { + "epoch": 0.9, + "learning_rate": 5.62179415154952e-06, + "loss": 0.5499, + "step": 30095 + }, + { + "epoch": 0.9, + "learning_rate": 5.618607115814356e-06, + "loss": 0.571, + "step": 30096 + }, + { + "epoch": 0.9, + "learning_rate": 5.61542095759976e-06, + "loss": 0.4911, + "step": 30097 + }, + { + "epoch": 0.9, + "learning_rate": 5.612235676935373e-06, + "loss": 0.492, + "step": 30098 + }, + { + "epoch": 0.9, + "learning_rate": 5.609051273850807e-06, + "loss": 0.6045, + "step": 30099 + }, + { + "epoch": 0.9, + "learning_rate": 5.605867748375637e-06, + "loss": 0.5241, + "step": 30100 + }, + { + "epoch": 0.9, + "learning_rate": 5.602685100539495e-06, + "loss": 0.4738, + "step": 30101 + }, + { + "epoch": 0.9, + "learning_rate": 5.59950333037198e-06, + "loss": 0.6531, + "step": 30102 + }, + { + "epoch": 0.9, + "learning_rate": 5.596322437902657e-06, + "loss": 0.4509, + "step": 30103 + }, + { + "epoch": 0.9, + "learning_rate": 5.5931424231610795e-06, + "loss": 0.4858, + "step": 30104 + }, + { + "epoch": 0.9, + "learning_rate": 5.589963286176858e-06, + "loss": 0.6025, + "step": 30105 + }, + { + "epoch": 0.9, + "learning_rate": 5.586785026979513e-06, + "loss": 0.5242, + "step": 30106 + }, + { + "epoch": 0.9, + "learning_rate": 5.583607645598621e-06, + "loss": 0.6047, + "step": 30107 + }, + { + "epoch": 0.9, + "learning_rate": 5.5804311420637245e-06, + "loss": 0.5186, + "step": 30108 + }, + { + "epoch": 0.9, + "learning_rate": 5.577255516404334e-06, + "loss": 0.5161, + "step": 30109 + }, + { + "epoch": 0.9, + "learning_rate": 5.574080768649992e-06, + "loss": 0.5374, + "step": 30110 + }, + { + "epoch": 0.9, + "learning_rate": 5.570906898830197e-06, + "loss": 0.6465, + "step": 30111 + }, + { + "epoch": 0.9, + "learning_rate": 5.567733906974493e-06, + "loss": 0.5719, + "step": 30112 + }, + { + "epoch": 0.9, + "learning_rate": 5.564561793112377e-06, + "loss": 0.4957, + "step": 30113 + }, + { + "epoch": 0.9, + "learning_rate": 5.561390557273294e-06, + "loss": 0.4831, + "step": 30114 + }, + { + "epoch": 0.9, + "learning_rate": 5.558220199486774e-06, + "loss": 0.5248, + "step": 30115 + }, + { + "epoch": 0.9, + "learning_rate": 5.555050719782273e-06, + "loss": 0.5215, + "step": 30116 + }, + { + "epoch": 0.9, + "learning_rate": 5.551882118189289e-06, + "loss": 0.4803, + "step": 30117 + }, + { + "epoch": 0.9, + "learning_rate": 5.5487143947372645e-06, + "loss": 0.4302, + "step": 30118 + }, + { + "epoch": 0.9, + "learning_rate": 5.545547549455632e-06, + "loss": 0.4397, + "step": 30119 + }, + { + "epoch": 0.9, + "learning_rate": 5.542381582373846e-06, + "loss": 0.5192, + "step": 30120 + }, + { + "epoch": 0.9, + "learning_rate": 5.539216493521371e-06, + "loss": 0.4488, + "step": 30121 + }, + { + "epoch": 0.9, + "learning_rate": 5.536052282927606e-06, + "loss": 0.4973, + "step": 30122 + }, + { + "epoch": 0.9, + "learning_rate": 5.532888950621962e-06, + "loss": 0.6065, + "step": 30123 + }, + { + "epoch": 0.9, + "learning_rate": 5.52972649663388e-06, + "loss": 0.5387, + "step": 30124 + }, + { + "epoch": 0.9, + "learning_rate": 5.526564920992749e-06, + "loss": 0.5533, + "step": 30125 + }, + { + "epoch": 0.9, + "learning_rate": 5.523404223727957e-06, + "loss": 0.4997, + "step": 30126 + }, + { + "epoch": 0.9, + "learning_rate": 5.5202444048689015e-06, + "loss": 0.5814, + "step": 30127 + }, + { + "epoch": 0.9, + "learning_rate": 5.51708546444496e-06, + "loss": 0.5625, + "step": 30128 + }, + { + "epoch": 0.9, + "learning_rate": 5.513927402485497e-06, + "loss": 0.4432, + "step": 30129 + }, + { + "epoch": 0.9, + "learning_rate": 5.5107702190198785e-06, + "loss": 0.4796, + "step": 30130 + }, + { + "epoch": 0.9, + "learning_rate": 5.5076139140774475e-06, + "loss": 0.5663, + "step": 30131 + }, + { + "epoch": 0.9, + "learning_rate": 5.50445848768758e-06, + "loss": 0.5481, + "step": 30132 + }, + { + "epoch": 0.9, + "learning_rate": 5.501303939879598e-06, + "loss": 0.5316, + "step": 30133 + }, + { + "epoch": 0.9, + "learning_rate": 5.4981502706828224e-06, + "loss": 0.6418, + "step": 30134 + }, + { + "epoch": 0.9, + "learning_rate": 5.494997480126562e-06, + "loss": 0.4797, + "step": 30135 + }, + { + "epoch": 0.9, + "learning_rate": 5.491845568240173e-06, + "loss": 0.5502, + "step": 30136 + }, + { + "epoch": 0.9, + "learning_rate": 5.48869453505294e-06, + "loss": 0.5236, + "step": 30137 + }, + { + "epoch": 0.9, + "learning_rate": 5.4855443805941435e-06, + "loss": 0.5979, + "step": 30138 + }, + { + "epoch": 0.9, + "learning_rate": 5.482395104893079e-06, + "loss": 0.4461, + "step": 30139 + }, + { + "epoch": 0.9, + "learning_rate": 5.4792467079790465e-06, + "loss": 0.5358, + "step": 30140 + }, + { + "epoch": 0.9, + "learning_rate": 5.476099189881301e-06, + "loss": 0.4567, + "step": 30141 + }, + { + "epoch": 0.9, + "learning_rate": 5.472952550629118e-06, + "loss": 0.4474, + "step": 30142 + }, + { + "epoch": 0.9, + "learning_rate": 5.4698067902517306e-06, + "loss": 0.5294, + "step": 30143 + }, + { + "epoch": 0.9, + "learning_rate": 5.466661908778426e-06, + "loss": 0.5092, + "step": 30144 + }, + { + "epoch": 0.9, + "learning_rate": 5.463517906238402e-06, + "loss": 0.5486, + "step": 30145 + }, + { + "epoch": 0.9, + "learning_rate": 5.460374782660927e-06, + "loss": 0.5323, + "step": 30146 + }, + { + "epoch": 0.9, + "learning_rate": 5.457232538075196e-06, + "loss": 0.4969, + "step": 30147 + }, + { + "epoch": 0.9, + "learning_rate": 5.454091172510445e-06, + "loss": 0.5585, + "step": 30148 + }, + { + "epoch": 0.9, + "learning_rate": 5.450950685995859e-06, + "loss": 0.616, + "step": 30149 + }, + { + "epoch": 0.9, + "learning_rate": 5.44781107856065e-06, + "loss": 0.5211, + "step": 30150 + }, + { + "epoch": 0.9, + "learning_rate": 5.4446723502340255e-06, + "loss": 0.5491, + "step": 30151 + }, + { + "epoch": 0.9, + "learning_rate": 5.441534501045154e-06, + "loss": 0.5782, + "step": 30152 + }, + { + "epoch": 0.9, + "learning_rate": 5.4383975310231875e-06, + "loss": 0.5638, + "step": 30153 + }, + { + "epoch": 0.9, + "learning_rate": 5.435261440197315e-06, + "loss": 0.5437, + "step": 30154 + }, + { + "epoch": 0.9, + "learning_rate": 5.432126228596702e-06, + "loss": 0.5265, + "step": 30155 + }, + { + "epoch": 0.9, + "learning_rate": 5.428991896250491e-06, + "loss": 0.428, + "step": 30156 + }, + { + "epoch": 0.9, + "learning_rate": 5.4258584431878034e-06, + "loss": 0.6294, + "step": 30157 + }, + { + "epoch": 0.9, + "learning_rate": 5.422725869437794e-06, + "loss": 0.4513, + "step": 30158 + }, + { + "epoch": 0.9, + "learning_rate": 5.4195941750295855e-06, + "loss": 0.5303, + "step": 30159 + }, + { + "epoch": 0.9, + "learning_rate": 5.416463359992308e-06, + "loss": 0.4759, + "step": 30160 + }, + { + "epoch": 0.9, + "learning_rate": 5.4133334243550606e-06, + "loss": 0.5845, + "step": 30161 + }, + { + "epoch": 0.9, + "learning_rate": 5.4102043681469205e-06, + "loss": 0.4799, + "step": 30162 + }, + { + "epoch": 0.9, + "learning_rate": 5.407076191396998e-06, + "loss": 0.5492, + "step": 30163 + }, + { + "epoch": 0.9, + "learning_rate": 5.403948894134403e-06, + "loss": 0.7016, + "step": 30164 + }, + { + "epoch": 0.9, + "learning_rate": 5.400822476388179e-06, + "loss": 0.5485, + "step": 30165 + }, + { + "epoch": 0.9, + "learning_rate": 5.397696938187391e-06, + "loss": 0.6374, + "step": 30166 + }, + { + "epoch": 0.9, + "learning_rate": 5.394572279561116e-06, + "loss": 0.48, + "step": 30167 + }, + { + "epoch": 0.9, + "learning_rate": 5.391448500538421e-06, + "loss": 0.5161, + "step": 30168 + }, + { + "epoch": 0.9, + "learning_rate": 5.388325601148314e-06, + "loss": 0.4824, + "step": 30169 + }, + { + "epoch": 0.9, + "learning_rate": 5.385203581419862e-06, + "loss": 0.4707, + "step": 30170 + }, + { + "epoch": 0.9, + "learning_rate": 5.382082441382074e-06, + "loss": 0.6981, + "step": 30171 + }, + { + "epoch": 0.9, + "learning_rate": 5.378962181063951e-06, + "loss": 0.5476, + "step": 30172 + }, + { + "epoch": 0.9, + "learning_rate": 5.375842800494535e-06, + "loss": 0.5025, + "step": 30173 + }, + { + "epoch": 0.9, + "learning_rate": 5.372724299702814e-06, + "loss": 0.5534, + "step": 30174 + }, + { + "epoch": 0.9, + "learning_rate": 5.36960667871782e-06, + "loss": 0.5952, + "step": 30175 + }, + { + "epoch": 0.9, + "learning_rate": 5.366489937568487e-06, + "loss": 0.5718, + "step": 30176 + }, + { + "epoch": 0.9, + "learning_rate": 5.363374076283801e-06, + "loss": 0.6458, + "step": 30177 + }, + { + "epoch": 0.9, + "learning_rate": 5.36025909489275e-06, + "loss": 0.5622, + "step": 30178 + }, + { + "epoch": 0.9, + "learning_rate": 5.357144993424301e-06, + "loss": 0.525, + "step": 30179 + }, + { + "epoch": 0.9, + "learning_rate": 5.354031771907398e-06, + "loss": 0.5282, + "step": 30180 + }, + { + "epoch": 0.9, + "learning_rate": 5.350919430370971e-06, + "loss": 0.5228, + "step": 30181 + }, + { + "epoch": 0.9, + "learning_rate": 5.347807968843965e-06, + "loss": 0.53, + "step": 30182 + }, + { + "epoch": 0.9, + "learning_rate": 5.344697387355346e-06, + "loss": 0.4351, + "step": 30183 + }, + { + "epoch": 0.9, + "learning_rate": 5.34158768593398e-06, + "loss": 0.4645, + "step": 30184 + }, + { + "epoch": 0.9, + "learning_rate": 5.33847886460882e-06, + "loss": 0.5194, + "step": 30185 + }, + { + "epoch": 0.9, + "learning_rate": 5.3353709234087445e-06, + "loss": 0.556, + "step": 30186 + }, + { + "epoch": 0.9, + "learning_rate": 5.332263862362663e-06, + "loss": 0.5326, + "step": 30187 + }, + { + "epoch": 0.9, + "learning_rate": 5.329157681499464e-06, + "loss": 0.5141, + "step": 30188 + }, + { + "epoch": 0.9, + "learning_rate": 5.326052380848023e-06, + "loss": 0.5585, + "step": 30189 + }, + { + "epoch": 0.9, + "learning_rate": 5.322947960437208e-06, + "loss": 0.5443, + "step": 30190 + }, + { + "epoch": 0.9, + "learning_rate": 5.319844420295905e-06, + "loss": 0.5391, + "step": 30191 + }, + { + "epoch": 0.9, + "learning_rate": 5.316741760452937e-06, + "loss": 0.6063, + "step": 30192 + }, + { + "epoch": 0.9, + "learning_rate": 5.313639980937168e-06, + "loss": 0.5302, + "step": 30193 + }, + { + "epoch": 0.9, + "learning_rate": 5.310539081777443e-06, + "loss": 0.4558, + "step": 30194 + }, + { + "epoch": 0.9, + "learning_rate": 5.307439063002595e-06, + "loss": 0.6547, + "step": 30195 + }, + { + "epoch": 0.9, + "learning_rate": 5.304339924641421e-06, + "loss": 0.4047, + "step": 30196 + }, + { + "epoch": 0.9, + "learning_rate": 5.301241666722745e-06, + "loss": 0.4726, + "step": 30197 + }, + { + "epoch": 0.9, + "learning_rate": 5.298144289275409e-06, + "loss": 0.4315, + "step": 30198 + }, + { + "epoch": 0.9, + "learning_rate": 5.2950477923281675e-06, + "loss": 0.5161, + "step": 30199 + }, + { + "epoch": 0.9, + "learning_rate": 5.29195217590982e-06, + "loss": 0.4046, + "step": 30200 + }, + { + "epoch": 0.9, + "learning_rate": 5.288857440049155e-06, + "loss": 0.4794, + "step": 30201 + }, + { + "epoch": 0.9, + "learning_rate": 5.28576358477495e-06, + "loss": 0.5325, + "step": 30202 + }, + { + "epoch": 0.9, + "learning_rate": 5.282670610115969e-06, + "loss": 0.5629, + "step": 30203 + }, + { + "epoch": 0.9, + "learning_rate": 5.2795785161009695e-06, + "loss": 0.5493, + "step": 30204 + }, + { + "epoch": 0.9, + "learning_rate": 5.276487302758693e-06, + "loss": 0.5594, + "step": 30205 + }, + { + "epoch": 0.9, + "learning_rate": 5.273396970117883e-06, + "loss": 0.6416, + "step": 30206 + }, + { + "epoch": 0.9, + "learning_rate": 5.270307518207274e-06, + "loss": 0.6025, + "step": 30207 + }, + { + "epoch": 0.9, + "learning_rate": 5.267218947055596e-06, + "loss": 0.5862, + "step": 30208 + }, + { + "epoch": 0.9, + "learning_rate": 5.264131256691541e-06, + "loss": 0.4702, + "step": 30209 + }, + { + "epoch": 0.9, + "learning_rate": 5.261044447143859e-06, + "loss": 0.5598, + "step": 30210 + }, + { + "epoch": 0.9, + "learning_rate": 5.257958518441208e-06, + "loss": 0.6577, + "step": 30211 + }, + { + "epoch": 0.9, + "learning_rate": 5.254873470612298e-06, + "loss": 0.5152, + "step": 30212 + }, + { + "epoch": 0.9, + "learning_rate": 5.251789303685828e-06, + "loss": 0.5207, + "step": 30213 + }, + { + "epoch": 0.9, + "learning_rate": 5.2487060176904525e-06, + "loss": 0.5076, + "step": 30214 + }, + { + "epoch": 0.9, + "learning_rate": 5.245623612654838e-06, + "loss": 0.3587, + "step": 30215 + }, + { + "epoch": 0.9, + "learning_rate": 5.242542088607649e-06, + "loss": 0.5559, + "step": 30216 + }, + { + "epoch": 0.9, + "learning_rate": 5.239461445577554e-06, + "loss": 0.5822, + "step": 30217 + }, + { + "epoch": 0.9, + "learning_rate": 5.236381683593172e-06, + "loss": 0.4768, + "step": 30218 + }, + { + "epoch": 0.9, + "learning_rate": 5.233302802683126e-06, + "loss": 0.4832, + "step": 30219 + }, + { + "epoch": 0.9, + "learning_rate": 5.230224802876071e-06, + "loss": 0.3632, + "step": 30220 + }, + { + "epoch": 0.9, + "learning_rate": 5.227147684200606e-06, + "loss": 0.4624, + "step": 30221 + }, + { + "epoch": 0.9, + "learning_rate": 5.224071446685364e-06, + "loss": 0.488, + "step": 30222 + }, + { + "epoch": 0.9, + "learning_rate": 5.2209960903589426e-06, + "loss": 0.5253, + "step": 30223 + }, + { + "epoch": 0.9, + "learning_rate": 5.217921615249899e-06, + "loss": 0.4846, + "step": 30224 + }, + { + "epoch": 0.9, + "learning_rate": 5.2148480213868425e-06, + "loss": 0.484, + "step": 30225 + }, + { + "epoch": 0.9, + "learning_rate": 5.211775308798372e-06, + "loss": 0.3849, + "step": 30226 + }, + { + "epoch": 0.9, + "learning_rate": 5.208703477513033e-06, + "loss": 0.4574, + "step": 30227 + }, + { + "epoch": 0.9, + "learning_rate": 5.2056325275593786e-06, + "loss": 0.4353, + "step": 30228 + }, + { + "epoch": 0.9, + "learning_rate": 5.202562458965965e-06, + "loss": 0.3789, + "step": 30229 + }, + { + "epoch": 0.9, + "learning_rate": 5.199493271761369e-06, + "loss": 0.4851, + "step": 30230 + }, + { + "epoch": 0.9, + "learning_rate": 5.196424965974078e-06, + "loss": 0.534, + "step": 30231 + }, + { + "epoch": 0.9, + "learning_rate": 5.193357541632659e-06, + "loss": 0.4938, + "step": 30232 + }, + { + "epoch": 0.9, + "learning_rate": 5.19029099876559e-06, + "loss": 0.4077, + "step": 30233 + }, + { + "epoch": 0.9, + "learning_rate": 5.187225337401436e-06, + "loss": 0.445, + "step": 30234 + }, + { + "epoch": 0.9, + "learning_rate": 5.1841605575686516e-06, + "loss": 0.4701, + "step": 30235 + }, + { + "epoch": 0.9, + "learning_rate": 5.181096659295759e-06, + "loss": 0.5099, + "step": 30236 + }, + { + "epoch": 0.9, + "learning_rate": 5.178033642611257e-06, + "loss": 0.489, + "step": 30237 + }, + { + "epoch": 0.9, + "learning_rate": 5.1749715075435915e-06, + "loss": 0.511, + "step": 30238 + }, + { + "epoch": 0.9, + "learning_rate": 5.171910254121237e-06, + "loss": 0.568, + "step": 30239 + }, + { + "epoch": 0.9, + "learning_rate": 5.168849882372661e-06, + "loss": 0.601, + "step": 30240 + }, + { + "epoch": 0.9, + "learning_rate": 5.165790392326353e-06, + "loss": 0.5647, + "step": 30241 + }, + { + "epoch": 0.9, + "learning_rate": 5.1627317840107216e-06, + "loss": 0.5939, + "step": 30242 + }, + { + "epoch": 0.9, + "learning_rate": 5.159674057454189e-06, + "loss": 0.5121, + "step": 30243 + }, + { + "epoch": 0.9, + "learning_rate": 5.156617212685222e-06, + "loss": 0.5215, + "step": 30244 + }, + { + "epoch": 0.9, + "learning_rate": 5.153561249732231e-06, + "loss": 0.4857, + "step": 30245 + }, + { + "epoch": 0.9, + "learning_rate": 5.150506168623615e-06, + "loss": 0.5437, + "step": 30246 + }, + { + "epoch": 0.9, + "learning_rate": 5.147451969387795e-06, + "loss": 0.5819, + "step": 30247 + }, + { + "epoch": 0.9, + "learning_rate": 5.1443986520531615e-06, + "loss": 0.601, + "step": 30248 + }, + { + "epoch": 0.9, + "learning_rate": 5.141346216648113e-06, + "loss": 0.4288, + "step": 30249 + }, + { + "epoch": 0.9, + "learning_rate": 5.1382946632010045e-06, + "loss": 0.4873, + "step": 30250 + }, + { + "epoch": 0.9, + "learning_rate": 5.135243991740235e-06, + "loss": 0.6479, + "step": 30251 + }, + { + "epoch": 0.9, + "learning_rate": 5.13219420229415e-06, + "loss": 0.6255, + "step": 30252 + }, + { + "epoch": 0.9, + "learning_rate": 5.129145294891124e-06, + "loss": 0.523, + "step": 30253 + }, + { + "epoch": 0.9, + "learning_rate": 5.126097269559482e-06, + "loss": 0.5972, + "step": 30254 + }, + { + "epoch": 0.9, + "learning_rate": 5.123050126327566e-06, + "loss": 0.5594, + "step": 30255 + }, + { + "epoch": 0.9, + "learning_rate": 5.120003865223733e-06, + "loss": 0.5995, + "step": 30256 + }, + { + "epoch": 0.9, + "learning_rate": 5.116958486276291e-06, + "loss": 0.606, + "step": 30257 + }, + { + "epoch": 0.9, + "learning_rate": 5.113913989513541e-06, + "loss": 0.5073, + "step": 30258 + }, + { + "epoch": 0.9, + "learning_rate": 5.110870374963806e-06, + "loss": 0.5444, + "step": 30259 + }, + { + "epoch": 0.9, + "learning_rate": 5.107827642655383e-06, + "loss": 0.5523, + "step": 30260 + }, + { + "epoch": 0.9, + "learning_rate": 5.104785792616562e-06, + "loss": 0.5304, + "step": 30261 + }, + { + "epoch": 0.9, + "learning_rate": 5.101744824875598e-06, + "loss": 0.5, + "step": 30262 + }, + { + "epoch": 0.9, + "learning_rate": 5.098704739460802e-06, + "loss": 0.4788, + "step": 30263 + }, + { + "epoch": 0.9, + "learning_rate": 5.095665536400418e-06, + "loss": 0.4793, + "step": 30264 + }, + { + "epoch": 0.9, + "learning_rate": 5.092627215722723e-06, + "loss": 0.4625, + "step": 30265 + }, + { + "epoch": 0.9, + "learning_rate": 5.089589777455961e-06, + "loss": 0.5609, + "step": 30266 + }, + { + "epoch": 0.9, + "learning_rate": 5.086553221628343e-06, + "loss": 0.569, + "step": 30267 + }, + { + "epoch": 0.9, + "learning_rate": 5.083517548268135e-06, + "loss": 0.5087, + "step": 30268 + }, + { + "epoch": 0.9, + "learning_rate": 5.0804827574035594e-06, + "loss": 0.5432, + "step": 30269 + }, + { + "epoch": 0.9, + "learning_rate": 5.0774488490628155e-06, + "loss": 0.4945, + "step": 30270 + }, + { + "epoch": 0.9, + "learning_rate": 5.074415823274115e-06, + "loss": 0.4856, + "step": 30271 + }, + { + "epoch": 0.9, + "learning_rate": 5.071383680065678e-06, + "loss": 0.6633, + "step": 30272 + }, + { + "epoch": 0.9, + "learning_rate": 5.0683524194656605e-06, + "loss": 0.4531, + "step": 30273 + }, + { + "epoch": 0.9, + "learning_rate": 5.065322041502274e-06, + "loss": 0.4468, + "step": 30274 + }, + { + "epoch": 0.9, + "learning_rate": 5.062292546203695e-06, + "loss": 0.6872, + "step": 30275 + }, + { + "epoch": 0.9, + "learning_rate": 5.059263933598079e-06, + "loss": 0.6493, + "step": 30276 + }, + { + "epoch": 0.9, + "learning_rate": 5.056236203713582e-06, + "loss": 0.4626, + "step": 30277 + }, + { + "epoch": 0.9, + "learning_rate": 5.053209356578348e-06, + "loss": 0.562, + "step": 30278 + }, + { + "epoch": 0.9, + "learning_rate": 5.050183392220542e-06, + "loss": 0.488, + "step": 30279 + }, + { + "epoch": 0.9, + "learning_rate": 5.047158310668309e-06, + "loss": 0.6922, + "step": 30280 + }, + { + "epoch": 0.9, + "learning_rate": 5.0441341119497275e-06, + "loss": 0.5764, + "step": 30281 + }, + { + "epoch": 0.9, + "learning_rate": 5.04111079609294e-06, + "loss": 0.5471, + "step": 30282 + }, + { + "epoch": 0.9, + "learning_rate": 5.038088363126059e-06, + "loss": 0.677, + "step": 30283 + }, + { + "epoch": 0.9, + "learning_rate": 5.035066813077205e-06, + "loss": 0.5562, + "step": 30284 + }, + { + "epoch": 0.9, + "learning_rate": 5.032046145974445e-06, + "loss": 0.5876, + "step": 30285 + }, + { + "epoch": 0.9, + "learning_rate": 5.029026361845845e-06, + "loss": 0.4804, + "step": 30286 + }, + { + "epoch": 0.9, + "learning_rate": 5.026007460719517e-06, + "loss": 0.5122, + "step": 30287 + }, + { + "epoch": 0.9, + "learning_rate": 5.022989442623538e-06, + "loss": 0.502, + "step": 30288 + }, + { + "epoch": 0.9, + "learning_rate": 5.019972307585941e-06, + "loss": 0.4401, + "step": 30289 + }, + { + "epoch": 0.9, + "learning_rate": 5.01695605563477e-06, + "loss": 0.5557, + "step": 30290 + }, + { + "epoch": 0.9, + "learning_rate": 5.013940686798091e-06, + "loss": 0.5561, + "step": 30291 + }, + { + "epoch": 0.9, + "learning_rate": 5.010926201103949e-06, + "loss": 0.518, + "step": 30292 + }, + { + "epoch": 0.9, + "learning_rate": 5.007912598580333e-06, + "loss": 0.4448, + "step": 30293 + }, + { + "epoch": 0.9, + "learning_rate": 5.0048998792553095e-06, + "loss": 0.5948, + "step": 30294 + }, + { + "epoch": 0.9, + "learning_rate": 5.001888043156844e-06, + "loss": 0.5114, + "step": 30295 + }, + { + "epoch": 0.9, + "learning_rate": 4.9988770903129814e-06, + "loss": 0.4991, + "step": 30296 + }, + { + "epoch": 0.9, + "learning_rate": 4.995867020751677e-06, + "loss": 0.5502, + "step": 30297 + }, + { + "epoch": 0.9, + "learning_rate": 4.9928578345009414e-06, + "loss": 0.458, + "step": 30298 + }, + { + "epoch": 0.9, + "learning_rate": 4.9898495315887525e-06, + "loss": 0.5882, + "step": 30299 + }, + { + "epoch": 0.9, + "learning_rate": 4.986842112043077e-06, + "loss": 0.5077, + "step": 30300 + }, + { + "epoch": 0.9, + "learning_rate": 4.983835575891871e-06, + "loss": 0.541, + "step": 30301 + }, + { + "epoch": 0.9, + "learning_rate": 4.980829923163089e-06, + "loss": 0.4096, + "step": 30302 + }, + { + "epoch": 0.9, + "learning_rate": 4.977825153884685e-06, + "loss": 0.6199, + "step": 30303 + }, + { + "epoch": 0.9, + "learning_rate": 4.974821268084595e-06, + "loss": 0.5787, + "step": 30304 + }, + { + "epoch": 0.9, + "learning_rate": 4.971818265790729e-06, + "loss": 0.5258, + "step": 30305 + }, + { + "epoch": 0.9, + "learning_rate": 4.96881614703103e-06, + "loss": 0.5835, + "step": 30306 + }, + { + "epoch": 0.9, + "learning_rate": 4.9658149118334105e-06, + "loss": 0.5666, + "step": 30307 + }, + { + "epoch": 0.9, + "learning_rate": 4.962814560225748e-06, + "loss": 0.5538, + "step": 30308 + }, + { + "epoch": 0.9, + "learning_rate": 4.959815092235975e-06, + "loss": 0.5305, + "step": 30309 + }, + { + "epoch": 0.9, + "learning_rate": 4.956816507891959e-06, + "loss": 0.3971, + "step": 30310 + }, + { + "epoch": 0.9, + "learning_rate": 4.953818807221589e-06, + "loss": 0.4796, + "step": 30311 + }, + { + "epoch": 0.9, + "learning_rate": 4.950821990252708e-06, + "loss": 0.4719, + "step": 30312 + }, + { + "epoch": 0.9, + "learning_rate": 4.9478260570132275e-06, + "loss": 0.564, + "step": 30313 + }, + { + "epoch": 0.9, + "learning_rate": 4.944831007530948e-06, + "loss": 0.5769, + "step": 30314 + }, + { + "epoch": 0.9, + "learning_rate": 4.941836841833769e-06, + "loss": 0.5376, + "step": 30315 + }, + { + "epoch": 0.9, + "learning_rate": 4.938843559949491e-06, + "loss": 0.5979, + "step": 30316 + }, + { + "epoch": 0.9, + "learning_rate": 4.9358511619059574e-06, + "loss": 0.4766, + "step": 30317 + }, + { + "epoch": 0.9, + "learning_rate": 4.932859647731003e-06, + "loss": 0.5713, + "step": 30318 + }, + { + "epoch": 0.9, + "learning_rate": 4.929869017452426e-06, + "loss": 0.5448, + "step": 30319 + }, + { + "epoch": 0.9, + "learning_rate": 4.926879271098028e-06, + "loss": 0.5742, + "step": 30320 + }, + { + "epoch": 0.9, + "learning_rate": 4.923890408695608e-06, + "loss": 0.5235, + "step": 30321 + }, + { + "epoch": 0.9, + "learning_rate": 4.920902430272978e-06, + "loss": 0.6037, + "step": 30322 + }, + { + "epoch": 0.9, + "learning_rate": 4.917915335857903e-06, + "loss": 0.6173, + "step": 30323 + }, + { + "epoch": 0.9, + "learning_rate": 4.914929125478152e-06, + "loss": 0.4481, + "step": 30324 + }, + { + "epoch": 0.9, + "learning_rate": 4.911943799161478e-06, + "loss": 0.5528, + "step": 30325 + }, + { + "epoch": 0.9, + "learning_rate": 4.908959356935661e-06, + "loss": 0.5073, + "step": 30326 + }, + { + "epoch": 0.9, + "learning_rate": 4.905975798828444e-06, + "loss": 0.5304, + "step": 30327 + }, + { + "epoch": 0.9, + "learning_rate": 4.902993124867572e-06, + "loss": 0.5793, + "step": 30328 + }, + { + "epoch": 0.9, + "learning_rate": 4.900011335080745e-06, + "loss": 0.6109, + "step": 30329 + }, + { + "epoch": 0.9, + "learning_rate": 4.897030429495719e-06, + "loss": 0.4476, + "step": 30330 + }, + { + "epoch": 0.9, + "learning_rate": 4.894050408140216e-06, + "loss": 0.5751, + "step": 30331 + }, + { + "epoch": 0.9, + "learning_rate": 4.891071271041914e-06, + "loss": 0.5036, + "step": 30332 + }, + { + "epoch": 0.9, + "learning_rate": 4.888093018228523e-06, + "loss": 0.5139, + "step": 30333 + }, + { + "epoch": 0.9, + "learning_rate": 4.885115649727723e-06, + "loss": 0.5057, + "step": 30334 + }, + { + "epoch": 0.9, + "learning_rate": 4.882139165567234e-06, + "loss": 0.5796, + "step": 30335 + }, + { + "epoch": 0.9, + "learning_rate": 4.879163565774692e-06, + "loss": 0.653, + "step": 30336 + }, + { + "epoch": 0.9, + "learning_rate": 4.8761888503777835e-06, + "loss": 0.6181, + "step": 30337 + }, + { + "epoch": 0.9, + "learning_rate": 4.873215019404154e-06, + "loss": 0.5576, + "step": 30338 + }, + { + "epoch": 0.9, + "learning_rate": 4.870242072881448e-06, + "loss": 0.5615, + "step": 30339 + }, + { + "epoch": 0.9, + "learning_rate": 4.867270010837321e-06, + "loss": 0.4474, + "step": 30340 + }, + { + "epoch": 0.9, + "learning_rate": 4.864298833299397e-06, + "loss": 0.5102, + "step": 30341 + }, + { + "epoch": 0.9, + "learning_rate": 4.861328540295329e-06, + "loss": 0.5884, + "step": 30342 + }, + { + "epoch": 0.9, + "learning_rate": 4.858359131852685e-06, + "loss": 0.4813, + "step": 30343 + }, + { + "epoch": 0.9, + "learning_rate": 4.8553906079990995e-06, + "loss": 0.5631, + "step": 30344 + }, + { + "epoch": 0.9, + "learning_rate": 4.852422968762171e-06, + "loss": 0.6106, + "step": 30345 + }, + { + "epoch": 0.9, + "learning_rate": 4.849456214169512e-06, + "loss": 0.4942, + "step": 30346 + }, + { + "epoch": 0.9, + "learning_rate": 4.846490344248678e-06, + "loss": 0.5183, + "step": 30347 + }, + { + "epoch": 0.9, + "learning_rate": 4.843525359027234e-06, + "loss": 0.4434, + "step": 30348 + }, + { + "epoch": 0.9, + "learning_rate": 4.840561258532772e-06, + "loss": 0.5014, + "step": 30349 + }, + { + "epoch": 0.9, + "learning_rate": 4.837598042792857e-06, + "loss": 0.5605, + "step": 30350 + }, + { + "epoch": 0.9, + "learning_rate": 4.834635711835023e-06, + "loss": 0.548, + "step": 30351 + }, + { + "epoch": 0.9, + "learning_rate": 4.831674265686814e-06, + "loss": 0.5638, + "step": 30352 + }, + { + "epoch": 0.9, + "learning_rate": 4.828713704375765e-06, + "loss": 0.419, + "step": 30353 + }, + { + "epoch": 0.9, + "learning_rate": 4.8257540279294186e-06, + "loss": 0.5297, + "step": 30354 + }, + { + "epoch": 0.9, + "learning_rate": 4.822795236375255e-06, + "loss": 0.5492, + "step": 30355 + }, + { + "epoch": 0.9, + "learning_rate": 4.819837329740829e-06, + "loss": 0.5034, + "step": 30356 + }, + { + "epoch": 0.9, + "learning_rate": 4.816880308053606e-06, + "loss": 0.5013, + "step": 30357 + }, + { + "epoch": 0.9, + "learning_rate": 4.813924171341111e-06, + "loss": 0.3884, + "step": 30358 + }, + { + "epoch": 0.9, + "learning_rate": 4.810968919630787e-06, + "loss": 0.6004, + "step": 30359 + }, + { + "epoch": 0.9, + "learning_rate": 4.808014552950146e-06, + "loss": 0.5763, + "step": 30360 + }, + { + "epoch": 0.9, + "learning_rate": 4.805061071326656e-06, + "loss": 0.5391, + "step": 30361 + }, + { + "epoch": 0.9, + "learning_rate": 4.80210847478777e-06, + "loss": 0.5648, + "step": 30362 + }, + { + "epoch": 0.9, + "learning_rate": 4.799156763360913e-06, + "loss": 0.5284, + "step": 30363 + }, + { + "epoch": 0.9, + "learning_rate": 4.796205937073561e-06, + "loss": 0.495, + "step": 30364 + }, + { + "epoch": 0.9, + "learning_rate": 4.79325599595315e-06, + "loss": 0.47, + "step": 30365 + }, + { + "epoch": 0.9, + "learning_rate": 4.790306940027112e-06, + "loss": 0.5441, + "step": 30366 + }, + { + "epoch": 0.9, + "learning_rate": 4.787358769322825e-06, + "loss": 0.4912, + "step": 30367 + }, + { + "epoch": 0.9, + "learning_rate": 4.784411483867734e-06, + "loss": 0.5072, + "step": 30368 + }, + { + "epoch": 0.9, + "learning_rate": 4.781465083689251e-06, + "loss": 0.5338, + "step": 30369 + }, + { + "epoch": 0.9, + "learning_rate": 4.778519568814743e-06, + "loss": 0.4647, + "step": 30370 + }, + { + "epoch": 0.9, + "learning_rate": 4.775574939271621e-06, + "loss": 0.404, + "step": 30371 + }, + { + "epoch": 0.9, + "learning_rate": 4.77263119508723e-06, + "loss": 0.5713, + "step": 30372 + }, + { + "epoch": 0.9, + "learning_rate": 4.76968833628898e-06, + "loss": 0.4825, + "step": 30373 + }, + { + "epoch": 0.9, + "learning_rate": 4.766746362904207e-06, + "loss": 0.5979, + "step": 30374 + }, + { + "epoch": 0.9, + "learning_rate": 4.763805274960276e-06, + "loss": 0.4425, + "step": 30375 + }, + { + "epoch": 0.9, + "learning_rate": 4.76086507248451e-06, + "loss": 0.4911, + "step": 30376 + }, + { + "epoch": 0.9, + "learning_rate": 4.757925755504278e-06, + "loss": 0.6183, + "step": 30377 + }, + { + "epoch": 0.9, + "learning_rate": 4.754987324046878e-06, + "loss": 0.4794, + "step": 30378 + }, + { + "epoch": 0.9, + "learning_rate": 4.752049778139645e-06, + "loss": 0.4304, + "step": 30379 + }, + { + "epoch": 0.9, + "learning_rate": 4.7491131178099e-06, + "loss": 0.5778, + "step": 30380 + }, + { + "epoch": 0.9, + "learning_rate": 4.746177343084946e-06, + "loss": 0.5431, + "step": 30381 + }, + { + "epoch": 0.9, + "learning_rate": 4.743242453992047e-06, + "loss": 0.5982, + "step": 30382 + }, + { + "epoch": 0.9, + "learning_rate": 4.7403084505585286e-06, + "loss": 0.4577, + "step": 30383 + }, + { + "epoch": 0.9, + "learning_rate": 4.737375332811655e-06, + "loss": 0.6076, + "step": 30384 + }, + { + "epoch": 0.9, + "learning_rate": 4.734443100778707e-06, + "loss": 0.4186, + "step": 30385 + }, + { + "epoch": 0.9, + "learning_rate": 4.731511754486928e-06, + "loss": 0.5049, + "step": 30386 + }, + { + "epoch": 0.9, + "learning_rate": 4.728581293963574e-06, + "loss": 0.5169, + "step": 30387 + }, + { + "epoch": 0.9, + "learning_rate": 4.7256517192359014e-06, + "loss": 0.5372, + "step": 30388 + }, + { + "epoch": 0.9, + "learning_rate": 4.722723030331156e-06, + "loss": 0.548, + "step": 30389 + }, + { + "epoch": 0.9, + "learning_rate": 4.719795227276569e-06, + "loss": 0.4444, + "step": 30390 + }, + { + "epoch": 0.9, + "learning_rate": 4.716868310099321e-06, + "loss": 0.5473, + "step": 30391 + }, + { + "epoch": 0.9, + "learning_rate": 4.713942278826666e-06, + "loss": 0.4895, + "step": 30392 + }, + { + "epoch": 0.9, + "learning_rate": 4.711017133485818e-06, + "loss": 0.6106, + "step": 30393 + }, + { + "epoch": 0.9, + "learning_rate": 4.708092874103942e-06, + "loss": 0.4487, + "step": 30394 + }, + { + "epoch": 0.9, + "learning_rate": 4.705169500708229e-06, + "loss": 0.6104, + "step": 30395 + }, + { + "epoch": 0.9, + "learning_rate": 4.702247013325878e-06, + "loss": 0.6006, + "step": 30396 + }, + { + "epoch": 0.9, + "learning_rate": 4.699325411984057e-06, + "loss": 0.5008, + "step": 30397 + }, + { + "epoch": 0.9, + "learning_rate": 4.696404696709911e-06, + "loss": 0.5535, + "step": 30398 + }, + { + "epoch": 0.91, + "learning_rate": 4.693484867530628e-06, + "loss": 0.4951, + "step": 30399 + }, + { + "epoch": 0.91, + "learning_rate": 4.690565924473334e-06, + "loss": 0.4229, + "step": 30400 + }, + { + "epoch": 0.91, + "learning_rate": 4.687647867565159e-06, + "loss": 0.6427, + "step": 30401 + }, + { + "epoch": 0.91, + "learning_rate": 4.684730696833251e-06, + "loss": 0.5084, + "step": 30402 + }, + { + "epoch": 0.91, + "learning_rate": 4.681814412304719e-06, + "loss": 0.512, + "step": 30403 + }, + { + "epoch": 0.91, + "learning_rate": 4.6788990140067214e-06, + "loss": 0.4578, + "step": 30404 + }, + { + "epoch": 0.91, + "learning_rate": 4.675984501966302e-06, + "loss": 0.5616, + "step": 30405 + }, + { + "epoch": 0.91, + "learning_rate": 4.673070876210583e-06, + "loss": 0.5505, + "step": 30406 + }, + { + "epoch": 0.91, + "learning_rate": 4.670158136766656e-06, + "loss": 0.5787, + "step": 30407 + }, + { + "epoch": 0.91, + "learning_rate": 4.667246283661619e-06, + "loss": 0.5692, + "step": 30408 + }, + { + "epoch": 0.91, + "learning_rate": 4.66433531692253e-06, + "loss": 0.5579, + "step": 30409 + }, + { + "epoch": 0.91, + "learning_rate": 4.6614252365764445e-06, + "loss": 0.4488, + "step": 30410 + }, + { + "epoch": 0.91, + "learning_rate": 4.658516042650418e-06, + "loss": 0.6752, + "step": 30411 + }, + { + "epoch": 0.91, + "learning_rate": 4.65560773517153e-06, + "loss": 0.5641, + "step": 30412 + }, + { + "epoch": 0.91, + "learning_rate": 4.652700314166791e-06, + "loss": 0.4879, + "step": 30413 + }, + { + "epoch": 0.91, + "learning_rate": 4.6497937796632585e-06, + "loss": 0.5556, + "step": 30414 + }, + { + "epoch": 0.91, + "learning_rate": 4.646888131687921e-06, + "loss": 0.5807, + "step": 30415 + }, + { + "epoch": 0.91, + "learning_rate": 4.643983370267824e-06, + "loss": 0.5361, + "step": 30416 + }, + { + "epoch": 0.91, + "learning_rate": 4.641079495429957e-06, + "loss": 0.5043, + "step": 30417 + }, + { + "epoch": 0.91, + "learning_rate": 4.638176507201342e-06, + "loss": 0.635, + "step": 30418 + }, + { + "epoch": 0.91, + "learning_rate": 4.635274405608936e-06, + "loss": 0.5742, + "step": 30419 + }, + { + "epoch": 0.91, + "learning_rate": 4.632373190679751e-06, + "loss": 0.5173, + "step": 30420 + }, + { + "epoch": 0.91, + "learning_rate": 4.629472862440742e-06, + "loss": 0.5743, + "step": 30421 + }, + { + "epoch": 0.91, + "learning_rate": 4.626573420918878e-06, + "loss": 0.5745, + "step": 30422 + }, + { + "epoch": 0.91, + "learning_rate": 4.6236748661411474e-06, + "loss": 0.5064, + "step": 30423 + }, + { + "epoch": 0.91, + "learning_rate": 4.620777198134463e-06, + "loss": 0.5091, + "step": 30424 + }, + { + "epoch": 0.91, + "learning_rate": 4.6178804169257684e-06, + "loss": 0.6323, + "step": 30425 + }, + { + "epoch": 0.91, + "learning_rate": 4.61498452254201e-06, + "loss": 0.5976, + "step": 30426 + }, + { + "epoch": 0.91, + "learning_rate": 4.61208951501011e-06, + "loss": 0.5265, + "step": 30427 + }, + { + "epoch": 0.91, + "learning_rate": 4.609195394357002e-06, + "loss": 0.6019, + "step": 30428 + }, + { + "epoch": 0.91, + "learning_rate": 4.606302160609543e-06, + "loss": 0.4741, + "step": 30429 + }, + { + "epoch": 0.91, + "learning_rate": 4.603409813794679e-06, + "loss": 0.5893, + "step": 30430 + }, + { + "epoch": 0.91, + "learning_rate": 4.600518353939287e-06, + "loss": 0.3943, + "step": 30431 + }, + { + "epoch": 0.91, + "learning_rate": 4.597627781070269e-06, + "loss": 0.5252, + "step": 30432 + }, + { + "epoch": 0.91, + "learning_rate": 4.594738095214479e-06, + "loss": 0.4578, + "step": 30433 + }, + { + "epoch": 0.91, + "learning_rate": 4.591849296398775e-06, + "loss": 0.4928, + "step": 30434 + }, + { + "epoch": 0.91, + "learning_rate": 4.588961384650048e-06, + "loss": 0.5735, + "step": 30435 + }, + { + "epoch": 0.91, + "learning_rate": 4.586074359995119e-06, + "loss": 0.5303, + "step": 30436 + }, + { + "epoch": 0.91, + "learning_rate": 4.583188222460855e-06, + "loss": 0.4828, + "step": 30437 + }, + { + "epoch": 0.91, + "learning_rate": 4.580302972074058e-06, + "loss": 0.4932, + "step": 30438 + }, + { + "epoch": 0.91, + "learning_rate": 4.5774186088615945e-06, + "loss": 0.5849, + "step": 30439 + }, + { + "epoch": 0.91, + "learning_rate": 4.574535132850244e-06, + "loss": 0.4679, + "step": 30440 + }, + { + "epoch": 0.91, + "learning_rate": 4.571652544066829e-06, + "loss": 0.4756, + "step": 30441 + }, + { + "epoch": 0.91, + "learning_rate": 4.568770842538173e-06, + "loss": 0.5514, + "step": 30442 + }, + { + "epoch": 0.91, + "learning_rate": 4.565890028291054e-06, + "loss": 0.5171, + "step": 30443 + }, + { + "epoch": 0.91, + "learning_rate": 4.563010101352227e-06, + "loss": 0.5314, + "step": 30444 + }, + { + "epoch": 0.91, + "learning_rate": 4.5601310617485074e-06, + "loss": 0.4894, + "step": 30445 + }, + { + "epoch": 0.91, + "learning_rate": 4.557252909506648e-06, + "loss": 0.6375, + "step": 30446 + }, + { + "epoch": 0.91, + "learning_rate": 4.55437564465343e-06, + "loss": 0.4749, + "step": 30447 + }, + { + "epoch": 0.91, + "learning_rate": 4.551499267215575e-06, + "loss": 0.4887, + "step": 30448 + }, + { + "epoch": 0.91, + "learning_rate": 4.548623777219829e-06, + "loss": 0.5262, + "step": 30449 + }, + { + "epoch": 0.91, + "learning_rate": 4.545749174692926e-06, + "loss": 0.5848, + "step": 30450 + }, + { + "epoch": 0.91, + "learning_rate": 4.542875459661633e-06, + "loss": 0.5785, + "step": 30451 + }, + { + "epoch": 0.91, + "learning_rate": 4.540002632152629e-06, + "loss": 0.4825, + "step": 30452 + }, + { + "epoch": 0.91, + "learning_rate": 4.537130692192626e-06, + "loss": 0.5068, + "step": 30453 + }, + { + "epoch": 0.91, + "learning_rate": 4.534259639808325e-06, + "loss": 0.5005, + "step": 30454 + }, + { + "epoch": 0.91, + "learning_rate": 4.531389475026448e-06, + "loss": 0.5386, + "step": 30455 + }, + { + "epoch": 0.91, + "learning_rate": 4.528520197873654e-06, + "loss": 0.4397, + "step": 30456 + }, + { + "epoch": 0.91, + "learning_rate": 4.525651808376641e-06, + "loss": 0.5401, + "step": 30457 + }, + { + "epoch": 0.91, + "learning_rate": 4.522784306562044e-06, + "loss": 0.4958, + "step": 30458 + }, + { + "epoch": 0.91, + "learning_rate": 4.519917692456554e-06, + "loss": 0.5417, + "step": 30459 + }, + { + "epoch": 0.91, + "learning_rate": 4.517051966086816e-06, + "loss": 0.4568, + "step": 30460 + }, + { + "epoch": 0.91, + "learning_rate": 4.514187127479474e-06, + "loss": 0.5722, + "step": 30461 + }, + { + "epoch": 0.91, + "learning_rate": 4.511323176661153e-06, + "loss": 0.4208, + "step": 30462 + }, + { + "epoch": 0.91, + "learning_rate": 4.508460113658508e-06, + "loss": 0.4754, + "step": 30463 + }, + { + "epoch": 0.91, + "learning_rate": 4.5055979384981295e-06, + "loss": 0.4511, + "step": 30464 + }, + { + "epoch": 0.91, + "learning_rate": 4.502736651206641e-06, + "loss": 0.4558, + "step": 30465 + }, + { + "epoch": 0.91, + "learning_rate": 4.499876251810664e-06, + "loss": 0.5287, + "step": 30466 + }, + { + "epoch": 0.91, + "learning_rate": 4.497016740336757e-06, + "loss": 0.4569, + "step": 30467 + }, + { + "epoch": 0.91, + "learning_rate": 4.49415811681152e-06, + "loss": 0.4533, + "step": 30468 + }, + { + "epoch": 0.91, + "learning_rate": 4.491300381261543e-06, + "loss": 0.5092, + "step": 30469 + }, + { + "epoch": 0.91, + "learning_rate": 4.4884435337133935e-06, + "loss": 0.461, + "step": 30470 + }, + { + "epoch": 0.91, + "learning_rate": 4.485587574193628e-06, + "loss": 0.5487, + "step": 30471 + }, + { + "epoch": 0.91, + "learning_rate": 4.482732502728793e-06, + "loss": 0.3908, + "step": 30472 + }, + { + "epoch": 0.91, + "learning_rate": 4.479878319345443e-06, + "loss": 0.5028, + "step": 30473 + }, + { + "epoch": 0.91, + "learning_rate": 4.477025024070125e-06, + "loss": 0.4966, + "step": 30474 + }, + { + "epoch": 0.91, + "learning_rate": 4.474172616929351e-06, + "loss": 0.613, + "step": 30475 + }, + { + "epoch": 0.91, + "learning_rate": 4.471321097949643e-06, + "loss": 0.7031, + "step": 30476 + }, + { + "epoch": 0.91, + "learning_rate": 4.468470467157515e-06, + "loss": 0.5221, + "step": 30477 + }, + { + "epoch": 0.91, + "learning_rate": 4.465620724579489e-06, + "loss": 0.6234, + "step": 30478 + }, + { + "epoch": 0.91, + "learning_rate": 4.4627718702420326e-06, + "loss": 0.525, + "step": 30479 + }, + { + "epoch": 0.91, + "learning_rate": 4.459923904171648e-06, + "loss": 0.4721, + "step": 30480 + }, + { + "epoch": 0.91, + "learning_rate": 4.457076826394813e-06, + "loss": 0.5378, + "step": 30481 + }, + { + "epoch": 0.91, + "learning_rate": 4.454230636938006e-06, + "loss": 0.5474, + "step": 30482 + }, + { + "epoch": 0.91, + "learning_rate": 4.451385335827662e-06, + "loss": 0.5125, + "step": 30483 + }, + { + "epoch": 0.91, + "learning_rate": 4.448540923090261e-06, + "loss": 0.5487, + "step": 30484 + }, + { + "epoch": 0.91, + "learning_rate": 4.445697398752257e-06, + "loss": 0.6191, + "step": 30485 + }, + { + "epoch": 0.91, + "learning_rate": 4.4428547628400765e-06, + "loss": 0.537, + "step": 30486 + }, + { + "epoch": 0.91, + "learning_rate": 4.440013015380129e-06, + "loss": 0.6132, + "step": 30487 + }, + { + "epoch": 0.91, + "learning_rate": 4.43717215639885e-06, + "loss": 0.5554, + "step": 30488 + }, + { + "epoch": 0.91, + "learning_rate": 4.434332185922685e-06, + "loss": 0.6325, + "step": 30489 + }, + { + "epoch": 0.91, + "learning_rate": 4.431493103977991e-06, + "loss": 0.4187, + "step": 30490 + }, + { + "epoch": 0.91, + "learning_rate": 4.42865491059119e-06, + "loss": 0.4641, + "step": 30491 + }, + { + "epoch": 0.91, + "learning_rate": 4.4258176057886515e-06, + "loss": 0.5699, + "step": 30492 + }, + { + "epoch": 0.91, + "learning_rate": 4.4229811895967755e-06, + "loss": 0.6021, + "step": 30493 + }, + { + "epoch": 0.91, + "learning_rate": 4.420145662041941e-06, + "loss": 0.4803, + "step": 30494 + }, + { + "epoch": 0.91, + "learning_rate": 4.417311023150495e-06, + "loss": 0.466, + "step": 30495 + }, + { + "epoch": 0.91, + "learning_rate": 4.414477272948781e-06, + "loss": 0.4894, + "step": 30496 + }, + { + "epoch": 0.91, + "learning_rate": 4.4116444114631564e-06, + "loss": 0.5316, + "step": 30497 + }, + { + "epoch": 0.91, + "learning_rate": 4.4088124387199895e-06, + "loss": 0.5041, + "step": 30498 + }, + { + "epoch": 0.91, + "learning_rate": 4.405981354745581e-06, + "loss": 0.5995, + "step": 30499 + }, + { + "epoch": 0.91, + "learning_rate": 4.403151159566243e-06, + "loss": 0.4436, + "step": 30500 + }, + { + "epoch": 0.91, + "learning_rate": 4.400321853208311e-06, + "loss": 0.4784, + "step": 30501 + }, + { + "epoch": 0.91, + "learning_rate": 4.3974934356980854e-06, + "loss": 0.5104, + "step": 30502 + }, + { + "epoch": 0.91, + "learning_rate": 4.394665907061846e-06, + "loss": 0.489, + "step": 30503 + }, + { + "epoch": 0.91, + "learning_rate": 4.391839267325926e-06, + "loss": 0.5083, + "step": 30504 + }, + { + "epoch": 0.91, + "learning_rate": 4.389013516516572e-06, + "loss": 0.4713, + "step": 30505 + }, + { + "epoch": 0.91, + "learning_rate": 4.386188654660051e-06, + "loss": 0.562, + "step": 30506 + }, + { + "epoch": 0.91, + "learning_rate": 4.383364681782643e-06, + "loss": 0.4664, + "step": 30507 + }, + { + "epoch": 0.91, + "learning_rate": 4.3805415979105924e-06, + "loss": 0.5019, + "step": 30508 + }, + { + "epoch": 0.91, + "learning_rate": 4.377719403070191e-06, + "loss": 0.5148, + "step": 30509 + }, + { + "epoch": 0.91, + "learning_rate": 4.374898097287616e-06, + "loss": 0.5538, + "step": 30510 + }, + { + "epoch": 0.91, + "learning_rate": 4.372077680589126e-06, + "loss": 0.491, + "step": 30511 + }, + { + "epoch": 0.91, + "learning_rate": 4.3692581530009436e-06, + "loss": 0.5665, + "step": 30512 + }, + { + "epoch": 0.91, + "learning_rate": 4.366439514549292e-06, + "loss": 0.5137, + "step": 30513 + }, + { + "epoch": 0.91, + "learning_rate": 4.363621765260373e-06, + "loss": 0.477, + "step": 30514 + }, + { + "epoch": 0.91, + "learning_rate": 4.3608049051603765e-06, + "loss": 0.4989, + "step": 30515 + }, + { + "epoch": 0.91, + "learning_rate": 4.357988934275492e-06, + "loss": 0.4884, + "step": 30516 + }, + { + "epoch": 0.91, + "learning_rate": 4.355173852631922e-06, + "loss": 0.6121, + "step": 30517 + }, + { + "epoch": 0.91, + "learning_rate": 4.3523596602558e-06, + "loss": 0.4817, + "step": 30518 + }, + { + "epoch": 0.91, + "learning_rate": 4.349546357173339e-06, + "loss": 0.504, + "step": 30519 + }, + { + "epoch": 0.91, + "learning_rate": 4.346733943410664e-06, + "loss": 0.4439, + "step": 30520 + }, + { + "epoch": 0.91, + "learning_rate": 4.343922418993951e-06, + "loss": 0.488, + "step": 30521 + }, + { + "epoch": 0.91, + "learning_rate": 4.341111783949304e-06, + "loss": 0.5545, + "step": 30522 + }, + { + "epoch": 0.91, + "learning_rate": 4.338302038302877e-06, + "loss": 0.468, + "step": 30523 + }, + { + "epoch": 0.91, + "learning_rate": 4.335493182080785e-06, + "loss": 0.5952, + "step": 30524 + }, + { + "epoch": 0.91, + "learning_rate": 4.332685215309162e-06, + "loss": 0.4783, + "step": 30525 + }, + { + "epoch": 0.91, + "learning_rate": 4.329878138014099e-06, + "loss": 0.4904, + "step": 30526 + }, + { + "epoch": 0.91, + "learning_rate": 4.3270719502216835e-06, + "loss": 0.6284, + "step": 30527 + }, + { + "epoch": 0.91, + "learning_rate": 4.324266651958042e-06, + "loss": 0.5841, + "step": 30528 + }, + { + "epoch": 0.91, + "learning_rate": 4.3214622432492415e-06, + "loss": 0.555, + "step": 30529 + }, + { + "epoch": 0.91, + "learning_rate": 4.318658724121327e-06, + "loss": 0.6745, + "step": 30530 + }, + { + "epoch": 0.91, + "learning_rate": 4.31585609460039e-06, + "loss": 0.5733, + "step": 30531 + }, + { + "epoch": 0.91, + "learning_rate": 4.313054354712509e-06, + "loss": 0.4502, + "step": 30532 + }, + { + "epoch": 0.91, + "learning_rate": 4.3102535044837074e-06, + "loss": 0.4157, + "step": 30533 + }, + { + "epoch": 0.91, + "learning_rate": 4.307453543940021e-06, + "loss": 0.4154, + "step": 30534 + }, + { + "epoch": 0.91, + "learning_rate": 4.304654473107483e-06, + "loss": 0.454, + "step": 30535 + }, + { + "epoch": 0.91, + "learning_rate": 4.301856292012152e-06, + "loss": 0.5448, + "step": 30536 + }, + { + "epoch": 0.91, + "learning_rate": 4.299059000679995e-06, + "loss": 0.4417, + "step": 30537 + }, + { + "epoch": 0.91, + "learning_rate": 4.296262599137069e-06, + "loss": 0.5771, + "step": 30538 + }, + { + "epoch": 0.91, + "learning_rate": 4.2934670874093306e-06, + "loss": 0.6005, + "step": 30539 + }, + { + "epoch": 0.91, + "learning_rate": 4.290672465522805e-06, + "loss": 0.5133, + "step": 30540 + }, + { + "epoch": 0.91, + "learning_rate": 4.2878787335034365e-06, + "loss": 0.5764, + "step": 30541 + }, + { + "epoch": 0.91, + "learning_rate": 4.2850858913772494e-06, + "loss": 0.6095, + "step": 30542 + }, + { + "epoch": 0.91, + "learning_rate": 4.282293939170179e-06, + "loss": 0.493, + "step": 30543 + }, + { + "epoch": 0.91, + "learning_rate": 4.279502876908192e-06, + "loss": 0.5386, + "step": 30544 + }, + { + "epoch": 0.91, + "learning_rate": 4.276712704617225e-06, + "loss": 0.4825, + "step": 30545 + }, + { + "epoch": 0.91, + "learning_rate": 4.273923422323234e-06, + "loss": 0.504, + "step": 30546 + }, + { + "epoch": 0.91, + "learning_rate": 4.271135030052164e-06, + "loss": 0.6392, + "step": 30547 + }, + { + "epoch": 0.91, + "learning_rate": 4.268347527829919e-06, + "loss": 0.4555, + "step": 30548 + }, + { + "epoch": 0.91, + "learning_rate": 4.265560915682421e-06, + "loss": 0.5959, + "step": 30549 + }, + { + "epoch": 0.91, + "learning_rate": 4.262775193635582e-06, + "loss": 0.5539, + "step": 30550 + }, + { + "epoch": 0.91, + "learning_rate": 4.259990361715305e-06, + "loss": 0.5446, + "step": 30551 + }, + { + "epoch": 0.91, + "learning_rate": 4.25720641994749e-06, + "loss": 0.6352, + "step": 30552 + }, + { + "epoch": 0.91, + "learning_rate": 4.254423368357996e-06, + "loss": 0.4823, + "step": 30553 + }, + { + "epoch": 0.91, + "learning_rate": 4.251641206972701e-06, + "loss": 0.5612, + "step": 30554 + }, + { + "epoch": 0.91, + "learning_rate": 4.248859935817495e-06, + "loss": 0.4946, + "step": 30555 + }, + { + "epoch": 0.91, + "learning_rate": 4.246079554918236e-06, + "loss": 0.4421, + "step": 30556 + }, + { + "epoch": 0.91, + "learning_rate": 4.24330006430077e-06, + "loss": 0.4791, + "step": 30557 + }, + { + "epoch": 0.91, + "learning_rate": 4.240521463990921e-06, + "loss": 0.5579, + "step": 30558 + }, + { + "epoch": 0.91, + "learning_rate": 4.237743754014534e-06, + "loss": 0.5232, + "step": 30559 + }, + { + "epoch": 0.91, + "learning_rate": 4.234966934397455e-06, + "loss": 0.5168, + "step": 30560 + }, + { + "epoch": 0.91, + "learning_rate": 4.232191005165487e-06, + "loss": 0.5415, + "step": 30561 + }, + { + "epoch": 0.91, + "learning_rate": 4.229415966344419e-06, + "loss": 0.5594, + "step": 30562 + }, + { + "epoch": 0.91, + "learning_rate": 4.226641817960075e-06, + "loss": 0.6489, + "step": 30563 + }, + { + "epoch": 0.91, + "learning_rate": 4.223868560038258e-06, + "loss": 0.481, + "step": 30564 + }, + { + "epoch": 0.91, + "learning_rate": 4.221096192604723e-06, + "loss": 0.4859, + "step": 30565 + }, + { + "epoch": 0.91, + "learning_rate": 4.218324715685273e-06, + "loss": 0.4795, + "step": 30566 + }, + { + "epoch": 0.91, + "learning_rate": 4.215554129305665e-06, + "loss": 0.6281, + "step": 30567 + }, + { + "epoch": 0.91, + "learning_rate": 4.212784433491656e-06, + "loss": 0.5637, + "step": 30568 + }, + { + "epoch": 0.91, + "learning_rate": 4.210015628268993e-06, + "loss": 0.5881, + "step": 30569 + }, + { + "epoch": 0.91, + "learning_rate": 4.207247713663421e-06, + "loss": 0.6185, + "step": 30570 + }, + { + "epoch": 0.91, + "learning_rate": 4.204480689700707e-06, + "loss": 0.5149, + "step": 30571 + }, + { + "epoch": 0.91, + "learning_rate": 4.201714556406533e-06, + "loss": 0.485, + "step": 30572 + }, + { + "epoch": 0.91, + "learning_rate": 4.198949313806633e-06, + "loss": 0.4208, + "step": 30573 + }, + { + "epoch": 0.91, + "learning_rate": 4.196184961926719e-06, + "loss": 0.495, + "step": 30574 + }, + { + "epoch": 0.91, + "learning_rate": 4.193421500792505e-06, + "loss": 0.4596, + "step": 30575 + }, + { + "epoch": 0.91, + "learning_rate": 4.190658930429681e-06, + "loss": 0.5747, + "step": 30576 + }, + { + "epoch": 0.91, + "learning_rate": 4.1878972508639035e-06, + "loss": 0.4559, + "step": 30577 + }, + { + "epoch": 0.91, + "learning_rate": 4.185136462120864e-06, + "loss": 0.5825, + "step": 30578 + }, + { + "epoch": 0.91, + "learning_rate": 4.182376564226265e-06, + "loss": 0.4291, + "step": 30579 + }, + { + "epoch": 0.91, + "learning_rate": 4.179617557205717e-06, + "loss": 0.4595, + "step": 30580 + }, + { + "epoch": 0.91, + "learning_rate": 4.176859441084913e-06, + "loss": 0.5115, + "step": 30581 + }, + { + "epoch": 0.91, + "learning_rate": 4.174102215889464e-06, + "loss": 0.5772, + "step": 30582 + }, + { + "epoch": 0.91, + "learning_rate": 4.1713458816450275e-06, + "loss": 0.67, + "step": 30583 + }, + { + "epoch": 0.91, + "learning_rate": 4.168590438377218e-06, + "loss": 0.4485, + "step": 30584 + }, + { + "epoch": 0.91, + "learning_rate": 4.1658358861116685e-06, + "loss": 0.4745, + "step": 30585 + }, + { + "epoch": 0.91, + "learning_rate": 4.163082224873971e-06, + "loss": 0.5253, + "step": 30586 + }, + { + "epoch": 0.91, + "learning_rate": 4.160329454689759e-06, + "loss": 0.5873, + "step": 30587 + }, + { + "epoch": 0.91, + "learning_rate": 4.157577575584582e-06, + "loss": 0.4822, + "step": 30588 + }, + { + "epoch": 0.91, + "learning_rate": 4.154826587584048e-06, + "loss": 0.5469, + "step": 30589 + }, + { + "epoch": 0.91, + "learning_rate": 4.152076490713752e-06, + "loss": 0.5457, + "step": 30590 + }, + { + "epoch": 0.91, + "learning_rate": 4.1493272849992495e-06, + "loss": 0.4814, + "step": 30591 + }, + { + "epoch": 0.91, + "learning_rate": 4.146578970466087e-06, + "loss": 0.4882, + "step": 30592 + }, + { + "epoch": 0.91, + "learning_rate": 4.143831547139832e-06, + "loss": 0.5241, + "step": 30593 + }, + { + "epoch": 0.91, + "learning_rate": 4.141085015046032e-06, + "loss": 0.4464, + "step": 30594 + }, + { + "epoch": 0.91, + "learning_rate": 4.138339374210221e-06, + "loss": 0.4969, + "step": 30595 + }, + { + "epoch": 0.91, + "learning_rate": 4.135594624657912e-06, + "loss": 0.5358, + "step": 30596 + }, + { + "epoch": 0.91, + "learning_rate": 4.132850766414642e-06, + "loss": 0.5266, + "step": 30597 + }, + { + "epoch": 0.91, + "learning_rate": 4.130107799505922e-06, + "loss": 0.5544, + "step": 30598 + }, + { + "epoch": 0.91, + "learning_rate": 4.127365723957244e-06, + "loss": 0.5953, + "step": 30599 + }, + { + "epoch": 0.91, + "learning_rate": 4.124624539794119e-06, + "loss": 0.5343, + "step": 30600 + }, + { + "epoch": 0.91, + "learning_rate": 4.121884247042007e-06, + "loss": 0.5095, + "step": 30601 + }, + { + "epoch": 0.91, + "learning_rate": 4.119144845726408e-06, + "loss": 0.7008, + "step": 30602 + }, + { + "epoch": 0.91, + "learning_rate": 4.11640633587278e-06, + "loss": 0.4548, + "step": 30603 + }, + { + "epoch": 0.91, + "learning_rate": 4.113668717506603e-06, + "loss": 0.5691, + "step": 30604 + }, + { + "epoch": 0.91, + "learning_rate": 4.110931990653311e-06, + "loss": 0.4784, + "step": 30605 + }, + { + "epoch": 0.91, + "learning_rate": 4.108196155338362e-06, + "loss": 0.4849, + "step": 30606 + }, + { + "epoch": 0.91, + "learning_rate": 4.10546121158718e-06, + "loss": 0.5389, + "step": 30607 + }, + { + "epoch": 0.91, + "learning_rate": 4.102727159425191e-06, + "loss": 0.6, + "step": 30608 + }, + { + "epoch": 0.91, + "learning_rate": 4.099993998877838e-06, + "loss": 0.4444, + "step": 30609 + }, + { + "epoch": 0.91, + "learning_rate": 4.097261729970525e-06, + "loss": 0.5294, + "step": 30610 + }, + { + "epoch": 0.91, + "learning_rate": 4.094530352728632e-06, + "loss": 0.4238, + "step": 30611 + }, + { + "epoch": 0.91, + "learning_rate": 4.09179986717757e-06, + "loss": 0.5807, + "step": 30612 + }, + { + "epoch": 0.91, + "learning_rate": 4.089070273342732e-06, + "loss": 0.518, + "step": 30613 + }, + { + "epoch": 0.91, + "learning_rate": 4.086341571249519e-06, + "loss": 0.558, + "step": 30614 + }, + { + "epoch": 0.91, + "learning_rate": 4.083613760923244e-06, + "loss": 0.5045, + "step": 30615 + }, + { + "epoch": 0.91, + "learning_rate": 4.080886842389298e-06, + "loss": 0.5499, + "step": 30616 + }, + { + "epoch": 0.91, + "learning_rate": 4.078160815673038e-06, + "loss": 0.4375, + "step": 30617 + }, + { + "epoch": 0.91, + "learning_rate": 4.075435680799833e-06, + "loss": 0.3765, + "step": 30618 + }, + { + "epoch": 0.91, + "learning_rate": 4.072711437794985e-06, + "loss": 0.547, + "step": 30619 + }, + { + "epoch": 0.91, + "learning_rate": 4.069988086683818e-06, + "loss": 0.5204, + "step": 30620 + }, + { + "epoch": 0.91, + "learning_rate": 4.067265627491668e-06, + "loss": 0.5259, + "step": 30621 + }, + { + "epoch": 0.91, + "learning_rate": 4.064544060243869e-06, + "loss": 0.546, + "step": 30622 + }, + { + "epoch": 0.91, + "learning_rate": 4.061823384965679e-06, + "loss": 0.494, + "step": 30623 + }, + { + "epoch": 0.91, + "learning_rate": 4.059103601682435e-06, + "loss": 0.515, + "step": 30624 + }, + { + "epoch": 0.91, + "learning_rate": 4.056384710419392e-06, + "loss": 0.6519, + "step": 30625 + }, + { + "epoch": 0.91, + "learning_rate": 4.053666711201865e-06, + "loss": 0.5219, + "step": 30626 + }, + { + "epoch": 0.91, + "learning_rate": 4.050949604055087e-06, + "loss": 0.4993, + "step": 30627 + }, + { + "epoch": 0.91, + "learning_rate": 4.048233389004352e-06, + "loss": 0.4731, + "step": 30628 + }, + { + "epoch": 0.91, + "learning_rate": 4.0455180660748935e-06, + "loss": 0.5425, + "step": 30629 + }, + { + "epoch": 0.91, + "learning_rate": 4.04280363529197e-06, + "loss": 0.4791, + "step": 30630 + }, + { + "epoch": 0.91, + "learning_rate": 4.040090096680793e-06, + "loss": 0.4581, + "step": 30631 + }, + { + "epoch": 0.91, + "learning_rate": 4.037377450266622e-06, + "loss": 0.4753, + "step": 30632 + }, + { + "epoch": 0.91, + "learning_rate": 4.03466569607468e-06, + "loss": 0.5447, + "step": 30633 + }, + { + "epoch": 0.91, + "learning_rate": 4.0319548341301694e-06, + "loss": 0.4952, + "step": 30634 + }, + { + "epoch": 0.91, + "learning_rate": 4.029244864458282e-06, + "loss": 0.4941, + "step": 30635 + }, + { + "epoch": 0.91, + "learning_rate": 4.026535787084229e-06, + "loss": 0.547, + "step": 30636 + }, + { + "epoch": 0.91, + "learning_rate": 4.023827602033203e-06, + "loss": 0.4626, + "step": 30637 + }, + { + "epoch": 0.91, + "learning_rate": 4.021120309330373e-06, + "loss": 0.6639, + "step": 30638 + }, + { + "epoch": 0.91, + "learning_rate": 4.018413909000895e-06, + "loss": 0.5166, + "step": 30639 + }, + { + "epoch": 0.91, + "learning_rate": 4.015708401069962e-06, + "loss": 0.4425, + "step": 30640 + }, + { + "epoch": 0.91, + "learning_rate": 4.01300378556273e-06, + "loss": 0.5368, + "step": 30641 + }, + { + "epoch": 0.91, + "learning_rate": 4.010300062504313e-06, + "loss": 0.517, + "step": 30642 + }, + { + "epoch": 0.91, + "learning_rate": 4.00759723191988e-06, + "loss": 0.4379, + "step": 30643 + }, + { + "epoch": 0.91, + "learning_rate": 4.004895293834532e-06, + "loss": 0.5155, + "step": 30644 + }, + { + "epoch": 0.91, + "learning_rate": 4.002194248273427e-06, + "loss": 0.523, + "step": 30645 + }, + { + "epoch": 0.91, + "learning_rate": 3.999494095261647e-06, + "loss": 0.5109, + "step": 30646 + }, + { + "epoch": 0.91, + "learning_rate": 3.996794834824324e-06, + "loss": 0.4997, + "step": 30647 + }, + { + "epoch": 0.91, + "learning_rate": 3.99409646698653e-06, + "loss": 0.4488, + "step": 30648 + }, + { + "epoch": 0.91, + "learning_rate": 3.9913989917733655e-06, + "loss": 0.4893, + "step": 30649 + }, + { + "epoch": 0.91, + "learning_rate": 3.988702409209899e-06, + "loss": 0.5823, + "step": 30650 + }, + { + "epoch": 0.91, + "learning_rate": 3.986006719321212e-06, + "loss": 0.5529, + "step": 30651 + }, + { + "epoch": 0.91, + "learning_rate": 3.983311922132382e-06, + "loss": 0.5143, + "step": 30652 + }, + { + "epoch": 0.91, + "learning_rate": 3.980618017668447e-06, + "loss": 0.516, + "step": 30653 + }, + { + "epoch": 0.91, + "learning_rate": 3.977925005954441e-06, + "loss": 0.4863, + "step": 30654 + }, + { + "epoch": 0.91, + "learning_rate": 3.975232887015412e-06, + "loss": 0.5784, + "step": 30655 + }, + { + "epoch": 0.91, + "learning_rate": 3.972541660876417e-06, + "loss": 0.5832, + "step": 30656 + }, + { + "epoch": 0.91, + "learning_rate": 3.969851327562457e-06, + "loss": 0.4788, + "step": 30657 + }, + { + "epoch": 0.91, + "learning_rate": 3.9671618870985364e-06, + "loss": 0.5385, + "step": 30658 + }, + { + "epoch": 0.91, + "learning_rate": 3.964473339509666e-06, + "loss": 0.5728, + "step": 30659 + }, + { + "epoch": 0.91, + "learning_rate": 3.96178568482084e-06, + "loss": 0.5823, + "step": 30660 + }, + { + "epoch": 0.91, + "learning_rate": 3.95909892305707e-06, + "loss": 0.5911, + "step": 30661 + }, + { + "epoch": 0.91, + "learning_rate": 3.956413054243324e-06, + "loss": 0.4275, + "step": 30662 + }, + { + "epoch": 0.91, + "learning_rate": 3.953728078404562e-06, + "loss": 0.5223, + "step": 30663 + }, + { + "epoch": 0.91, + "learning_rate": 3.951043995565762e-06, + "loss": 0.5283, + "step": 30664 + }, + { + "epoch": 0.91, + "learning_rate": 3.9483608057518605e-06, + "loss": 0.5121, + "step": 30665 + }, + { + "epoch": 0.91, + "learning_rate": 3.9456785089878265e-06, + "loss": 0.4371, + "step": 30666 + }, + { + "epoch": 0.91, + "learning_rate": 3.942997105298584e-06, + "loss": 0.5892, + "step": 30667 + }, + { + "epoch": 0.91, + "learning_rate": 3.940316594709082e-06, + "loss": 0.4326, + "step": 30668 + }, + { + "epoch": 0.91, + "learning_rate": 3.937636977244219e-06, + "loss": 0.4628, + "step": 30669 + }, + { + "epoch": 0.91, + "learning_rate": 3.934958252928911e-06, + "loss": 0.5248, + "step": 30670 + }, + { + "epoch": 0.91, + "learning_rate": 3.932280421788093e-06, + "loss": 0.5936, + "step": 30671 + }, + { + "epoch": 0.91, + "learning_rate": 3.929603483846645e-06, + "loss": 0.4742, + "step": 30672 + }, + { + "epoch": 0.91, + "learning_rate": 3.926927439129447e-06, + "loss": 0.457, + "step": 30673 + }, + { + "epoch": 0.91, + "learning_rate": 3.924252287661379e-06, + "loss": 0.5582, + "step": 30674 + }, + { + "epoch": 0.91, + "learning_rate": 3.921578029467321e-06, + "loss": 0.6691, + "step": 30675 + }, + { + "epoch": 0.91, + "learning_rate": 3.918904664572154e-06, + "loss": 0.4376, + "step": 30676 + }, + { + "epoch": 0.91, + "learning_rate": 3.9162321930007014e-06, + "loss": 0.4612, + "step": 30677 + }, + { + "epoch": 0.91, + "learning_rate": 3.9135606147778335e-06, + "loss": 0.5976, + "step": 30678 + }, + { + "epoch": 0.91, + "learning_rate": 3.910889929928374e-06, + "loss": 0.6518, + "step": 30679 + }, + { + "epoch": 0.91, + "learning_rate": 3.908220138477181e-06, + "loss": 0.5272, + "step": 30680 + }, + { + "epoch": 0.91, + "learning_rate": 3.905551240449046e-06, + "loss": 0.5114, + "step": 30681 + }, + { + "epoch": 0.91, + "learning_rate": 3.902883235868793e-06, + "loss": 0.5006, + "step": 30682 + }, + { + "epoch": 0.91, + "learning_rate": 3.900216124761235e-06, + "loss": 0.4675, + "step": 30683 + }, + { + "epoch": 0.91, + "learning_rate": 3.897549907151177e-06, + "loss": 0.5024, + "step": 30684 + }, + { + "epoch": 0.91, + "learning_rate": 3.8948845830633744e-06, + "loss": 0.5643, + "step": 30685 + }, + { + "epoch": 0.91, + "learning_rate": 3.892220152522652e-06, + "loss": 0.5919, + "step": 30686 + }, + { + "epoch": 0.91, + "learning_rate": 3.889556615553747e-06, + "loss": 0.5828, + "step": 30687 + }, + { + "epoch": 0.91, + "learning_rate": 3.8868939721814605e-06, + "loss": 0.5288, + "step": 30688 + }, + { + "epoch": 0.91, + "learning_rate": 3.884232222430506e-06, + "loss": 0.5259, + "step": 30689 + }, + { + "epoch": 0.91, + "learning_rate": 3.881571366325665e-06, + "loss": 0.4817, + "step": 30690 + }, + { + "epoch": 0.91, + "learning_rate": 3.87891140389165e-06, + "loss": 0.5146, + "step": 30691 + }, + { + "epoch": 0.91, + "learning_rate": 3.876252335153219e-06, + "loss": 0.5956, + "step": 30692 + }, + { + "epoch": 0.91, + "learning_rate": 3.873594160135074e-06, + "loss": 0.5358, + "step": 30693 + }, + { + "epoch": 0.91, + "learning_rate": 3.870936878861941e-06, + "loss": 0.519, + "step": 30694 + }, + { + "epoch": 0.91, + "learning_rate": 3.868280491358522e-06, + "loss": 0.4876, + "step": 30695 + }, + { + "epoch": 0.91, + "learning_rate": 3.865624997649531e-06, + "loss": 0.5596, + "step": 30696 + }, + { + "epoch": 0.91, + "learning_rate": 3.8629703977596244e-06, + "loss": 0.5433, + "step": 30697 + }, + { + "epoch": 0.91, + "learning_rate": 3.860316691713506e-06, + "loss": 0.5574, + "step": 30698 + }, + { + "epoch": 0.91, + "learning_rate": 3.857663879535855e-06, + "loss": 0.4566, + "step": 30699 + }, + { + "epoch": 0.91, + "learning_rate": 3.855011961251332e-06, + "loss": 0.5384, + "step": 30700 + }, + { + "epoch": 0.91, + "learning_rate": 3.8523609368845696e-06, + "loss": 0.5651, + "step": 30701 + }, + { + "epoch": 0.91, + "learning_rate": 3.849710806460238e-06, + "loss": 0.452, + "step": 30702 + }, + { + "epoch": 0.91, + "learning_rate": 3.847061570002986e-06, + "loss": 0.5712, + "step": 30703 + }, + { + "epoch": 0.91, + "learning_rate": 3.844413227537424e-06, + "loss": 0.4952, + "step": 30704 + }, + { + "epoch": 0.91, + "learning_rate": 3.841765779088191e-06, + "loss": 0.5531, + "step": 30705 + }, + { + "epoch": 0.91, + "learning_rate": 3.839119224679877e-06, + "loss": 0.4435, + "step": 30706 + }, + { + "epoch": 0.91, + "learning_rate": 3.836473564337128e-06, + "loss": 0.5045, + "step": 30707 + }, + { + "epoch": 0.91, + "learning_rate": 3.833828798084504e-06, + "loss": 0.6047, + "step": 30708 + }, + { + "epoch": 0.91, + "learning_rate": 3.831184925946629e-06, + "loss": 0.5978, + "step": 30709 + }, + { + "epoch": 0.91, + "learning_rate": 3.82854194794805e-06, + "loss": 0.6113, + "step": 30710 + }, + { + "epoch": 0.91, + "learning_rate": 3.825899864113369e-06, + "loss": 0.3888, + "step": 30711 + }, + { + "epoch": 0.91, + "learning_rate": 3.823258674467134e-06, + "loss": 0.5172, + "step": 30712 + }, + { + "epoch": 0.91, + "learning_rate": 3.820618379033903e-06, + "loss": 0.5697, + "step": 30713 + }, + { + "epoch": 0.91, + "learning_rate": 3.817978977838243e-06, + "loss": 0.5622, + "step": 30714 + }, + { + "epoch": 0.91, + "learning_rate": 3.815340470904682e-06, + "loss": 0.4759, + "step": 30715 + }, + { + "epoch": 0.91, + "learning_rate": 3.812702858257733e-06, + "loss": 0.5795, + "step": 30716 + }, + { + "epoch": 0.91, + "learning_rate": 3.810066139921942e-06, + "loss": 0.5184, + "step": 30717 + }, + { + "epoch": 0.91, + "learning_rate": 3.8074303159218227e-06, + "loss": 0.5318, + "step": 30718 + }, + { + "epoch": 0.91, + "learning_rate": 3.8047953862818786e-06, + "loss": 0.4658, + "step": 30719 + }, + { + "epoch": 0.91, + "learning_rate": 3.802161351026601e-06, + "loss": 0.4556, + "step": 30720 + }, + { + "epoch": 0.91, + "learning_rate": 3.799528210180492e-06, + "loss": 0.5086, + "step": 30721 + }, + { + "epoch": 0.91, + "learning_rate": 3.79689596376801e-06, + "loss": 0.5035, + "step": 30722 + }, + { + "epoch": 0.91, + "learning_rate": 3.7942646118136695e-06, + "loss": 0.5422, + "step": 30723 + }, + { + "epoch": 0.91, + "learning_rate": 3.7916341543419166e-06, + "loss": 0.514, + "step": 30724 + }, + { + "epoch": 0.91, + "learning_rate": 3.7890045913771875e-06, + "loss": 0.5468, + "step": 30725 + }, + { + "epoch": 0.91, + "learning_rate": 3.786375922943952e-06, + "loss": 0.5182, + "step": 30726 + }, + { + "epoch": 0.91, + "learning_rate": 3.783748149066657e-06, + "loss": 0.5862, + "step": 30727 + }, + { + "epoch": 0.91, + "learning_rate": 3.7811212697697163e-06, + "loss": 0.5292, + "step": 30728 + }, + { + "epoch": 0.91, + "learning_rate": 3.7784952850775544e-06, + "loss": 0.5766, + "step": 30729 + }, + { + "epoch": 0.91, + "learning_rate": 3.775870195014608e-06, + "loss": 0.4835, + "step": 30730 + }, + { + "epoch": 0.91, + "learning_rate": 3.773245999605246e-06, + "loss": 0.4584, + "step": 30731 + }, + { + "epoch": 0.91, + "learning_rate": 3.7706226988739046e-06, + "loss": 0.3817, + "step": 30732 + }, + { + "epoch": 0.91, + "learning_rate": 3.7680002928449645e-06, + "loss": 0.4711, + "step": 30733 + }, + { + "epoch": 0.92, + "learning_rate": 3.7653787815428054e-06, + "loss": 0.5223, + "step": 30734 + }, + { + "epoch": 0.92, + "learning_rate": 3.7627581649917863e-06, + "loss": 0.5923, + "step": 30735 + }, + { + "epoch": 0.92, + "learning_rate": 3.7601384432162878e-06, + "loss": 0.5224, + "step": 30736 + }, + { + "epoch": 0.92, + "learning_rate": 3.757519616240657e-06, + "loss": 0.4276, + "step": 30737 + }, + { + "epoch": 0.92, + "learning_rate": 3.754901684089285e-06, + "loss": 0.4371, + "step": 30738 + }, + { + "epoch": 0.92, + "learning_rate": 3.7522846467864524e-06, + "loss": 0.4881, + "step": 30739 + }, + { + "epoch": 0.92, + "learning_rate": 3.7496685043565072e-06, + "loss": 0.528, + "step": 30740 + }, + { + "epoch": 0.92, + "learning_rate": 3.747053256823785e-06, + "loss": 0.504, + "step": 30741 + }, + { + "epoch": 0.92, + "learning_rate": 3.7444389042126216e-06, + "loss": 0.4939, + "step": 30742 + }, + { + "epoch": 0.92, + "learning_rate": 3.7418254465472867e-06, + "loss": 0.5996, + "step": 30743 + }, + { + "epoch": 0.92, + "learning_rate": 3.7392128838520944e-06, + "loss": 0.507, + "step": 30744 + }, + { + "epoch": 0.92, + "learning_rate": 3.7366012161513255e-06, + "loss": 0.4679, + "step": 30745 + }, + { + "epoch": 0.92, + "learning_rate": 3.733990443469293e-06, + "loss": 0.4669, + "step": 30746 + }, + { + "epoch": 0.92, + "learning_rate": 3.731380565830234e-06, + "loss": 0.4942, + "step": 30747 + }, + { + "epoch": 0.92, + "learning_rate": 3.728771583258439e-06, + "loss": 0.5789, + "step": 30748 + }, + { + "epoch": 0.92, + "learning_rate": 3.7261634957781457e-06, + "loss": 0.7069, + "step": 30749 + }, + { + "epoch": 0.92, + "learning_rate": 3.723556303413622e-06, + "loss": 0.5021, + "step": 30750 + }, + { + "epoch": 0.92, + "learning_rate": 3.7209500061890943e-06, + "loss": 0.6754, + "step": 30751 + }, + { + "epoch": 0.92, + "learning_rate": 3.7183446041288096e-06, + "loss": 0.6047, + "step": 30752 + }, + { + "epoch": 0.92, + "learning_rate": 3.7157400972569702e-06, + "loss": 0.5799, + "step": 30753 + }, + { + "epoch": 0.92, + "learning_rate": 3.7131364855978123e-06, + "loss": 0.5649, + "step": 30754 + }, + { + "epoch": 0.92, + "learning_rate": 3.7105337691755284e-06, + "loss": 0.5475, + "step": 30755 + }, + { + "epoch": 0.92, + "learning_rate": 3.7079319480143203e-06, + "loss": 0.5692, + "step": 30756 + }, + { + "epoch": 0.92, + "learning_rate": 3.705331022138403e-06, + "loss": 0.4771, + "step": 30757 + }, + { + "epoch": 0.92, + "learning_rate": 3.7027309915719344e-06, + "loss": 0.5206, + "step": 30758 + }, + { + "epoch": 0.92, + "learning_rate": 3.700131856339073e-06, + "loss": 0.5894, + "step": 30759 + }, + { + "epoch": 0.92, + "learning_rate": 3.6975336164640107e-06, + "loss": 0.4268, + "step": 30760 + }, + { + "epoch": 0.92, + "learning_rate": 3.694936271970917e-06, + "loss": 0.4107, + "step": 30761 + }, + { + "epoch": 0.92, + "learning_rate": 3.6923398228839056e-06, + "loss": 0.5568, + "step": 30762 + }, + { + "epoch": 0.92, + "learning_rate": 3.6897442692271357e-06, + "loss": 0.4924, + "step": 30763 + }, + { + "epoch": 0.92, + "learning_rate": 3.6871496110247316e-06, + "loss": 0.5651, + "step": 30764 + }, + { + "epoch": 0.92, + "learning_rate": 3.6845558483008413e-06, + "loss": 0.5912, + "step": 30765 + }, + { + "epoch": 0.92, + "learning_rate": 3.6819629810795452e-06, + "loss": 0.5355, + "step": 30766 + }, + { + "epoch": 0.92, + "learning_rate": 3.679371009384991e-06, + "loss": 0.5051, + "step": 30767 + }, + { + "epoch": 0.92, + "learning_rate": 3.6767799332412367e-06, + "loss": 0.621, + "step": 30768 + }, + { + "epoch": 0.92, + "learning_rate": 3.674189752672408e-06, + "loss": 0.5215, + "step": 30769 + }, + { + "epoch": 0.92, + "learning_rate": 3.671600467702552e-06, + "loss": 0.5879, + "step": 30770 + }, + { + "epoch": 0.92, + "learning_rate": 3.669012078355771e-06, + "loss": 0.5995, + "step": 30771 + }, + { + "epoch": 0.92, + "learning_rate": 3.6664245846561253e-06, + "loss": 0.5268, + "step": 30772 + }, + { + "epoch": 0.92, + "learning_rate": 3.6638379866276717e-06, + "loss": 0.6426, + "step": 30773 + }, + { + "epoch": 0.92, + "learning_rate": 3.6612522842944364e-06, + "loss": 0.4817, + "step": 30774 + }, + { + "epoch": 0.92, + "learning_rate": 3.6586674776804884e-06, + "loss": 0.548, + "step": 30775 + }, + { + "epoch": 0.92, + "learning_rate": 3.6560835668098646e-06, + "loss": 0.5231, + "step": 30776 + }, + { + "epoch": 0.92, + "learning_rate": 3.653500551706579e-06, + "loss": 0.6217, + "step": 30777 + }, + { + "epoch": 0.92, + "learning_rate": 3.6509184323946344e-06, + "loss": 0.557, + "step": 30778 + }, + { + "epoch": 0.92, + "learning_rate": 3.648337208898045e-06, + "loss": 0.5019, + "step": 30779 + }, + { + "epoch": 0.92, + "learning_rate": 3.6457568812408137e-06, + "loss": 0.4443, + "step": 30780 + }, + { + "epoch": 0.92, + "learning_rate": 3.643177449446966e-06, + "loss": 0.4934, + "step": 30781 + }, + { + "epoch": 0.92, + "learning_rate": 3.640598913540416e-06, + "loss": 0.5323, + "step": 30782 + }, + { + "epoch": 0.92, + "learning_rate": 3.638021273545178e-06, + "loss": 0.562, + "step": 30783 + }, + { + "epoch": 0.92, + "learning_rate": 3.63544452948521e-06, + "loss": 0.5221, + "step": 30784 + }, + { + "epoch": 0.92, + "learning_rate": 3.6328686813844824e-06, + "loss": 0.6069, + "step": 30785 + }, + { + "epoch": 0.92, + "learning_rate": 3.6302937292669316e-06, + "loss": 0.4334, + "step": 30786 + }, + { + "epoch": 0.92, + "learning_rate": 3.6277196731564933e-06, + "loss": 0.5989, + "step": 30787 + }, + { + "epoch": 0.92, + "learning_rate": 3.6251465130771044e-06, + "loss": 0.5185, + "step": 30788 + }, + { + "epoch": 0.92, + "learning_rate": 3.6225742490527015e-06, + "loss": 0.5222, + "step": 30789 + }, + { + "epoch": 0.92, + "learning_rate": 3.620002881107176e-06, + "loss": 0.3709, + "step": 30790 + }, + { + "epoch": 0.92, + "learning_rate": 3.6174324092644763e-06, + "loss": 0.4844, + "step": 30791 + }, + { + "epoch": 0.92, + "learning_rate": 3.6148628335484492e-06, + "loss": 0.6186, + "step": 30792 + }, + { + "epoch": 0.92, + "learning_rate": 3.6122941539830313e-06, + "loss": 0.5685, + "step": 30793 + }, + { + "epoch": 0.92, + "learning_rate": 3.60972637059207e-06, + "loss": 0.4894, + "step": 30794 + }, + { + "epoch": 0.92, + "learning_rate": 3.6071594833994693e-06, + "loss": 0.384, + "step": 30795 + }, + { + "epoch": 0.92, + "learning_rate": 3.604593492429087e-06, + "loss": 0.5409, + "step": 30796 + }, + { + "epoch": 0.92, + "learning_rate": 3.6020283977047598e-06, + "loss": 0.5643, + "step": 30797 + }, + { + "epoch": 0.92, + "learning_rate": 3.5994641992503464e-06, + "loss": 0.4784, + "step": 30798 + }, + { + "epoch": 0.92, + "learning_rate": 3.5969008970897054e-06, + "loss": 0.5906, + "step": 30799 + }, + { + "epoch": 0.92, + "learning_rate": 3.5943384912466628e-06, + "loss": 0.5101, + "step": 30800 + }, + { + "epoch": 0.92, + "learning_rate": 3.591776981745032e-06, + "loss": 0.5336, + "step": 30801 + }, + { + "epoch": 0.92, + "learning_rate": 3.589216368608628e-06, + "loss": 0.5355, + "step": 30802 + }, + { + "epoch": 0.92, + "learning_rate": 3.5866566518612643e-06, + "loss": 0.4969, + "step": 30803 + }, + { + "epoch": 0.92, + "learning_rate": 3.584097831526756e-06, + "loss": 0.6971, + "step": 30804 + }, + { + "epoch": 0.92, + "learning_rate": 3.5815399076288724e-06, + "loss": 0.4506, + "step": 30805 + }, + { + "epoch": 0.92, + "learning_rate": 3.578982880191395e-06, + "loss": 0.5487, + "step": 30806 + }, + { + "epoch": 0.92, + "learning_rate": 3.576426749238093e-06, + "loss": 0.382, + "step": 30807 + }, + { + "epoch": 0.92, + "learning_rate": 3.5738715147927703e-06, + "loss": 0.5396, + "step": 30808 + }, + { + "epoch": 0.92, + "learning_rate": 3.571317176879141e-06, + "loss": 0.5438, + "step": 30809 + }, + { + "epoch": 0.92, + "learning_rate": 3.5687637355209857e-06, + "loss": 0.6533, + "step": 30810 + }, + { + "epoch": 0.92, + "learning_rate": 3.566211190742019e-06, + "loss": 0.5528, + "step": 30811 + }, + { + "epoch": 0.92, + "learning_rate": 3.5636595425660004e-06, + "loss": 0.5979, + "step": 30812 + }, + { + "epoch": 0.92, + "learning_rate": 3.561108791016621e-06, + "loss": 0.5845, + "step": 30813 + }, + { + "epoch": 0.92, + "learning_rate": 3.5585589361176285e-06, + "loss": 0.5094, + "step": 30814 + }, + { + "epoch": 0.92, + "learning_rate": 3.556009977892716e-06, + "loss": 0.5991, + "step": 30815 + }, + { + "epoch": 0.92, + "learning_rate": 3.5534619163655856e-06, + "loss": 0.5861, + "step": 30816 + }, + { + "epoch": 0.92, + "learning_rate": 3.5509147515599083e-06, + "loss": 0.5432, + "step": 30817 + }, + { + "epoch": 0.92, + "learning_rate": 3.548368483499398e-06, + "loss": 0.5247, + "step": 30818 + }, + { + "epoch": 0.92, + "learning_rate": 3.5458231122077246e-06, + "loss": 0.4813, + "step": 30819 + }, + { + "epoch": 0.92, + "learning_rate": 3.543278637708547e-06, + "loss": 0.553, + "step": 30820 + }, + { + "epoch": 0.92, + "learning_rate": 3.540735060025502e-06, + "loss": 0.5456, + "step": 30821 + }, + { + "epoch": 0.92, + "learning_rate": 3.5381923791822593e-06, + "loss": 0.5097, + "step": 30822 + }, + { + "epoch": 0.92, + "learning_rate": 3.5356505952024666e-06, + "loss": 0.5127, + "step": 30823 + }, + { + "epoch": 0.92, + "learning_rate": 3.533109708109761e-06, + "loss": 0.5398, + "step": 30824 + }, + { + "epoch": 0.92, + "learning_rate": 3.5305697179277343e-06, + "loss": 0.467, + "step": 30825 + }, + { + "epoch": 0.92, + "learning_rate": 3.528030624680012e-06, + "loss": 0.4499, + "step": 30826 + }, + { + "epoch": 0.92, + "learning_rate": 3.52549242839022e-06, + "loss": 0.5897, + "step": 30827 + }, + { + "epoch": 0.92, + "learning_rate": 3.5229551290819616e-06, + "loss": 0.5222, + "step": 30828 + }, + { + "epoch": 0.92, + "learning_rate": 3.520418726778818e-06, + "loss": 0.4166, + "step": 30829 + }, + { + "epoch": 0.92, + "learning_rate": 3.517883221504348e-06, + "loss": 0.5652, + "step": 30830 + }, + { + "epoch": 0.92, + "learning_rate": 3.5153486132821544e-06, + "loss": 0.476, + "step": 30831 + }, + { + "epoch": 0.92, + "learning_rate": 3.512814902135797e-06, + "loss": 0.4835, + "step": 30832 + }, + { + "epoch": 0.92, + "learning_rate": 3.510282088088823e-06, + "loss": 0.4316, + "step": 30833 + }, + { + "epoch": 0.92, + "learning_rate": 3.507750171164792e-06, + "loss": 0.4611, + "step": 30834 + }, + { + "epoch": 0.92, + "learning_rate": 3.5052191513872512e-06, + "loss": 0.5064, + "step": 30835 + }, + { + "epoch": 0.92, + "learning_rate": 3.5026890287797154e-06, + "loss": 0.5286, + "step": 30836 + }, + { + "epoch": 0.92, + "learning_rate": 3.5001598033657102e-06, + "loss": 0.6301, + "step": 30837 + }, + { + "epoch": 0.92, + "learning_rate": 3.4976314751687723e-06, + "loss": 0.6475, + "step": 30838 + }, + { + "epoch": 0.92, + "learning_rate": 3.495104044212394e-06, + "loss": 0.5846, + "step": 30839 + }, + { + "epoch": 0.92, + "learning_rate": 3.4925775105200674e-06, + "loss": 0.5402, + "step": 30840 + }, + { + "epoch": 0.92, + "learning_rate": 3.4900518741152854e-06, + "loss": 0.4624, + "step": 30841 + }, + { + "epoch": 0.92, + "learning_rate": 3.48752713502154e-06, + "loss": 0.5206, + "step": 30842 + }, + { + "epoch": 0.92, + "learning_rate": 3.485003293262323e-06, + "loss": 0.5276, + "step": 30843 + }, + { + "epoch": 0.92, + "learning_rate": 3.48248034886105e-06, + "loss": 0.4006, + "step": 30844 + }, + { + "epoch": 0.92, + "learning_rate": 3.479958301841213e-06, + "loss": 0.509, + "step": 30845 + }, + { + "epoch": 0.92, + "learning_rate": 3.4774371522262485e-06, + "loss": 0.5198, + "step": 30846 + }, + { + "epoch": 0.92, + "learning_rate": 3.4749169000396266e-06, + "loss": 0.582, + "step": 30847 + }, + { + "epoch": 0.92, + "learning_rate": 3.472397545304751e-06, + "loss": 0.3717, + "step": 30848 + }, + { + "epoch": 0.92, + "learning_rate": 3.469879088045036e-06, + "loss": 0.561, + "step": 30849 + }, + { + "epoch": 0.92, + "learning_rate": 3.467361528283908e-06, + "loss": 0.4951, + "step": 30850 + }, + { + "epoch": 0.92, + "learning_rate": 3.464844866044803e-06, + "loss": 0.4409, + "step": 30851 + }, + { + "epoch": 0.92, + "learning_rate": 3.4623291013510696e-06, + "loss": 0.5721, + "step": 30852 + }, + { + "epoch": 0.92, + "learning_rate": 3.4598142342261443e-06, + "loss": 0.5399, + "step": 30853 + }, + { + "epoch": 0.92, + "learning_rate": 3.457300264693375e-06, + "loss": 0.501, + "step": 30854 + }, + { + "epoch": 0.92, + "learning_rate": 3.4547871927761656e-06, + "loss": 0.6185, + "step": 30855 + }, + { + "epoch": 0.92, + "learning_rate": 3.4522750184978527e-06, + "loss": 0.653, + "step": 30856 + }, + { + "epoch": 0.92, + "learning_rate": 3.4497637418818063e-06, + "loss": 0.4927, + "step": 30857 + }, + { + "epoch": 0.92, + "learning_rate": 3.4472533629513747e-06, + "loss": 0.5924, + "step": 30858 + }, + { + "epoch": 0.92, + "learning_rate": 3.444743881729917e-06, + "loss": 0.5584, + "step": 30859 + }, + { + "epoch": 0.92, + "learning_rate": 3.4422352982407257e-06, + "loss": 0.5373, + "step": 30860 + }, + { + "epoch": 0.92, + "learning_rate": 3.4397276125071488e-06, + "loss": 0.4411, + "step": 30861 + }, + { + "epoch": 0.92, + "learning_rate": 3.4372208245525116e-06, + "loss": 0.52, + "step": 30862 + }, + { + "epoch": 0.92, + "learning_rate": 3.434714934400107e-06, + "loss": 0.4245, + "step": 30863 + }, + { + "epoch": 0.92, + "learning_rate": 3.432209942073228e-06, + "loss": 0.5149, + "step": 30864 + }, + { + "epoch": 0.92, + "learning_rate": 3.429705847595166e-06, + "loss": 0.5269, + "step": 30865 + }, + { + "epoch": 0.92, + "learning_rate": 3.4272026509892363e-06, + "loss": 0.6687, + "step": 30866 + }, + { + "epoch": 0.92, + "learning_rate": 3.424700352278676e-06, + "loss": 0.5397, + "step": 30867 + }, + { + "epoch": 0.92, + "learning_rate": 3.422198951486744e-06, + "loss": 0.5111, + "step": 30868 + }, + { + "epoch": 0.92, + "learning_rate": 3.4196984486367214e-06, + "loss": 0.4787, + "step": 30869 + }, + { + "epoch": 0.92, + "learning_rate": 3.4171988437518563e-06, + "loss": 0.5459, + "step": 30870 + }, + { + "epoch": 0.92, + "learning_rate": 3.4147001368553755e-06, + "loss": 0.4814, + "step": 30871 + }, + { + "epoch": 0.92, + "learning_rate": 3.4122023279705263e-06, + "loss": 0.4993, + "step": 30872 + }, + { + "epoch": 0.92, + "learning_rate": 3.4097054171205122e-06, + "loss": 0.4566, + "step": 30873 + }, + { + "epoch": 0.92, + "learning_rate": 3.407209404328582e-06, + "loss": 0.6365, + "step": 30874 + }, + { + "epoch": 0.92, + "learning_rate": 3.4047142896178943e-06, + "loss": 0.6954, + "step": 30875 + }, + { + "epoch": 0.92, + "learning_rate": 3.402220073011697e-06, + "loss": 0.6526, + "step": 30876 + }, + { + "epoch": 0.92, + "learning_rate": 3.399726754533139e-06, + "loss": 0.5576, + "step": 30877 + }, + { + "epoch": 0.92, + "learning_rate": 3.397234334205435e-06, + "loss": 0.517, + "step": 30878 + }, + { + "epoch": 0.92, + "learning_rate": 3.3947428120517213e-06, + "loss": 0.636, + "step": 30879 + }, + { + "epoch": 0.92, + "learning_rate": 3.3922521880952017e-06, + "loss": 0.5511, + "step": 30880 + }, + { + "epoch": 0.92, + "learning_rate": 3.389762462359025e-06, + "loss": 0.5018, + "step": 30881 + }, + { + "epoch": 0.92, + "learning_rate": 3.387273634866328e-06, + "loss": 0.5059, + "step": 30882 + }, + { + "epoch": 0.92, + "learning_rate": 3.3847857056402365e-06, + "loss": 0.5791, + "step": 30883 + }, + { + "epoch": 0.92, + "learning_rate": 3.3822986747039096e-06, + "loss": 0.5057, + "step": 30884 + }, + { + "epoch": 0.92, + "learning_rate": 3.379812542080463e-06, + "loss": 0.5986, + "step": 30885 + }, + { + "epoch": 0.92, + "learning_rate": 3.3773273077930324e-06, + "loss": 0.5914, + "step": 30886 + }, + { + "epoch": 0.92, + "learning_rate": 3.374842971864678e-06, + "loss": 0.5097, + "step": 30887 + }, + { + "epoch": 0.92, + "learning_rate": 3.3723595343185255e-06, + "loss": 0.6124, + "step": 30888 + }, + { + "epoch": 0.92, + "learning_rate": 3.3698769951776565e-06, + "loss": 0.4552, + "step": 30889 + }, + { + "epoch": 0.92, + "learning_rate": 3.367395354465175e-06, + "loss": 0.521, + "step": 30890 + }, + { + "epoch": 0.92, + "learning_rate": 3.364914612204129e-06, + "loss": 0.5288, + "step": 30891 + }, + { + "epoch": 0.92, + "learning_rate": 3.362434768417588e-06, + "loss": 0.6362, + "step": 30892 + }, + { + "epoch": 0.92, + "learning_rate": 3.3599558231286134e-06, + "loss": 0.5516, + "step": 30893 + }, + { + "epoch": 0.92, + "learning_rate": 3.357477776360263e-06, + "loss": 0.4943, + "step": 30894 + }, + { + "epoch": 0.92, + "learning_rate": 3.3550006281355628e-06, + "loss": 0.5215, + "step": 30895 + }, + { + "epoch": 0.92, + "learning_rate": 3.3525243784775396e-06, + "loss": 0.5361, + "step": 30896 + }, + { + "epoch": 0.92, + "learning_rate": 3.350049027409241e-06, + "loss": 0.6059, + "step": 30897 + }, + { + "epoch": 0.92, + "learning_rate": 3.347574574953638e-06, + "loss": 0.5558, + "step": 30898 + }, + { + "epoch": 0.92, + "learning_rate": 3.3451010211337784e-06, + "loss": 0.6356, + "step": 30899 + }, + { + "epoch": 0.92, + "learning_rate": 3.3426283659726554e-06, + "loss": 0.5137, + "step": 30900 + }, + { + "epoch": 0.92, + "learning_rate": 3.34015660949325e-06, + "loss": 0.5282, + "step": 30901 + }, + { + "epoch": 0.92, + "learning_rate": 3.3376857517185333e-06, + "loss": 0.5058, + "step": 30902 + }, + { + "epoch": 0.92, + "learning_rate": 3.3352157926714866e-06, + "loss": 0.463, + "step": 30903 + }, + { + "epoch": 0.92, + "learning_rate": 3.3327467323750806e-06, + "loss": 0.6488, + "step": 30904 + }, + { + "epoch": 0.92, + "learning_rate": 3.3302785708522855e-06, + "loss": 0.526, + "step": 30905 + }, + { + "epoch": 0.92, + "learning_rate": 3.3278113081260055e-06, + "loss": 0.5641, + "step": 30906 + }, + { + "epoch": 0.92, + "learning_rate": 3.325344944219222e-06, + "loss": 0.5814, + "step": 30907 + }, + { + "epoch": 0.92, + "learning_rate": 3.322879479154839e-06, + "loss": 0.5717, + "step": 30908 + }, + { + "epoch": 0.92, + "learning_rate": 3.3204149129557936e-06, + "loss": 0.4551, + "step": 30909 + }, + { + "epoch": 0.92, + "learning_rate": 3.3179512456450125e-06, + "loss": 0.6007, + "step": 30910 + }, + { + "epoch": 0.92, + "learning_rate": 3.3154884772453654e-06, + "loss": 0.5798, + "step": 30911 + }, + { + "epoch": 0.92, + "learning_rate": 3.313026607779779e-06, + "loss": 0.6131, + "step": 30912 + }, + { + "epoch": 0.92, + "learning_rate": 3.3105656372711457e-06, + "loss": 0.5728, + "step": 30913 + }, + { + "epoch": 0.92, + "learning_rate": 3.3081055657423256e-06, + "loss": 0.4744, + "step": 30914 + }, + { + "epoch": 0.92, + "learning_rate": 3.305646393216211e-06, + "loss": 0.5373, + "step": 30915 + }, + { + "epoch": 0.92, + "learning_rate": 3.30318811971565e-06, + "loss": 0.5232, + "step": 30916 + }, + { + "epoch": 0.92, + "learning_rate": 3.3007307452635137e-06, + "loss": 0.6284, + "step": 30917 + }, + { + "epoch": 0.92, + "learning_rate": 3.2982742698826286e-06, + "loss": 0.4674, + "step": 30918 + }, + { + "epoch": 0.92, + "learning_rate": 3.2958186935958646e-06, + "loss": 0.5763, + "step": 30919 + }, + { + "epoch": 0.92, + "learning_rate": 3.293364016426026e-06, + "loss": 0.4865, + "step": 30920 + }, + { + "epoch": 0.92, + "learning_rate": 3.29091023839595e-06, + "loss": 0.5249, + "step": 30921 + }, + { + "epoch": 0.92, + "learning_rate": 3.2884573595284405e-06, + "loss": 0.4634, + "step": 30922 + }, + { + "epoch": 0.92, + "learning_rate": 3.286005379846313e-06, + "loss": 0.4989, + "step": 30923 + }, + { + "epoch": 0.92, + "learning_rate": 3.28355429937236e-06, + "loss": 0.4647, + "step": 30924 + }, + { + "epoch": 0.92, + "learning_rate": 3.2811041181293857e-06, + "loss": 0.5227, + "step": 30925 + }, + { + "epoch": 0.92, + "learning_rate": 3.2786548361401383e-06, + "loss": 0.4968, + "step": 30926 + }, + { + "epoch": 0.92, + "learning_rate": 3.2762064534273994e-06, + "loss": 0.577, + "step": 30927 + }, + { + "epoch": 0.92, + "learning_rate": 3.2737589700139624e-06, + "loss": 0.5822, + "step": 30928 + }, + { + "epoch": 0.92, + "learning_rate": 3.271312385922565e-06, + "loss": 0.5666, + "step": 30929 + }, + { + "epoch": 0.92, + "learning_rate": 3.268866701175943e-06, + "loss": 0.5015, + "step": 30930 + }, + { + "epoch": 0.92, + "learning_rate": 3.266421915796847e-06, + "loss": 0.5205, + "step": 30931 + }, + { + "epoch": 0.92, + "learning_rate": 3.2639780298080125e-06, + "loss": 0.4862, + "step": 30932 + }, + { + "epoch": 0.92, + "learning_rate": 3.2615350432321444e-06, + "loss": 0.4852, + "step": 30933 + }, + { + "epoch": 0.92, + "learning_rate": 3.2590929560919803e-06, + "loss": 0.6203, + "step": 30934 + }, + { + "epoch": 0.92, + "learning_rate": 3.2566517684101906e-06, + "loss": 0.6291, + "step": 30935 + }, + { + "epoch": 0.92, + "learning_rate": 3.2542114802095236e-06, + "loss": 0.5815, + "step": 30936 + }, + { + "epoch": 0.92, + "learning_rate": 3.251772091512617e-06, + "loss": 0.5441, + "step": 30937 + }, + { + "epoch": 0.92, + "learning_rate": 3.249333602342186e-06, + "loss": 0.4937, + "step": 30938 + }, + { + "epoch": 0.92, + "learning_rate": 3.2468960127208794e-06, + "loss": 0.545, + "step": 30939 + }, + { + "epoch": 0.92, + "learning_rate": 3.2444593226713893e-06, + "loss": 0.5596, + "step": 30940 + }, + { + "epoch": 0.92, + "learning_rate": 3.2420235322163316e-06, + "loss": 0.5382, + "step": 30941 + }, + { + "epoch": 0.92, + "learning_rate": 3.2395886413783773e-06, + "loss": 0.5004, + "step": 30942 + }, + { + "epoch": 0.92, + "learning_rate": 3.237154650180174e-06, + "loss": 0.5419, + "step": 30943 + }, + { + "epoch": 0.92, + "learning_rate": 3.234721558644349e-06, + "loss": 0.5185, + "step": 30944 + }, + { + "epoch": 0.92, + "learning_rate": 3.232289366793495e-06, + "loss": 0.434, + "step": 30945 + }, + { + "epoch": 0.92, + "learning_rate": 3.2298580746502604e-06, + "loss": 0.4394, + "step": 30946 + }, + { + "epoch": 0.92, + "learning_rate": 3.2274276822372273e-06, + "loss": 0.4989, + "step": 30947 + }, + { + "epoch": 0.92, + "learning_rate": 3.224998189577022e-06, + "loss": 0.5995, + "step": 30948 + }, + { + "epoch": 0.92, + "learning_rate": 3.2225695966921933e-06, + "loss": 0.5094, + "step": 30949 + }, + { + "epoch": 0.92, + "learning_rate": 3.2201419036053337e-06, + "loss": 0.4667, + "step": 30950 + }, + { + "epoch": 0.92, + "learning_rate": 3.217715110339037e-06, + "loss": 0.6256, + "step": 30951 + }, + { + "epoch": 0.92, + "learning_rate": 3.215289216915851e-06, + "loss": 0.5159, + "step": 30952 + }, + { + "epoch": 0.92, + "learning_rate": 3.212864223358336e-06, + "loss": 0.4883, + "step": 30953 + }, + { + "epoch": 0.92, + "learning_rate": 3.2104401296890185e-06, + "loss": 0.5462, + "step": 30954 + }, + { + "epoch": 0.92, + "learning_rate": 3.208016935930458e-06, + "loss": 0.6022, + "step": 30955 + }, + { + "epoch": 0.92, + "learning_rate": 3.205594642105192e-06, + "loss": 0.5263, + "step": 30956 + }, + { + "epoch": 0.92, + "learning_rate": 3.2031732482357026e-06, + "loss": 0.6187, + "step": 30957 + }, + { + "epoch": 0.92, + "learning_rate": 3.200752754344549e-06, + "loss": 0.6059, + "step": 30958 + }, + { + "epoch": 0.92, + "learning_rate": 3.1983331604542256e-06, + "loss": 0.6579, + "step": 30959 + }, + { + "epoch": 0.92, + "learning_rate": 3.195914466587191e-06, + "loss": 0.5345, + "step": 30960 + }, + { + "epoch": 0.92, + "learning_rate": 3.193496672765972e-06, + "loss": 0.6104, + "step": 30961 + }, + { + "epoch": 0.92, + "learning_rate": 3.191079779013051e-06, + "loss": 0.4323, + "step": 30962 + }, + { + "epoch": 0.92, + "learning_rate": 3.188663785350876e-06, + "loss": 0.5062, + "step": 30963 + }, + { + "epoch": 0.92, + "learning_rate": 3.1862486918019185e-06, + "loss": 0.5247, + "step": 30964 + }, + { + "epoch": 0.92, + "learning_rate": 3.183834498388627e-06, + "loss": 0.4645, + "step": 30965 + }, + { + "epoch": 0.92, + "learning_rate": 3.1814212051334614e-06, + "loss": 0.5384, + "step": 30966 + }, + { + "epoch": 0.92, + "learning_rate": 3.1790088120588592e-06, + "loss": 0.5359, + "step": 30967 + }, + { + "epoch": 0.92, + "learning_rate": 3.1765973191872467e-06, + "loss": 0.6253, + "step": 30968 + }, + { + "epoch": 0.92, + "learning_rate": 3.1741867265410287e-06, + "loss": 0.4668, + "step": 30969 + }, + { + "epoch": 0.92, + "learning_rate": 3.1717770341426422e-06, + "loss": 0.5441, + "step": 30970 + }, + { + "epoch": 0.92, + "learning_rate": 3.169368242014481e-06, + "loss": 0.4321, + "step": 30971 + }, + { + "epoch": 0.92, + "learning_rate": 3.166960350178938e-06, + "loss": 0.635, + "step": 30972 + }, + { + "epoch": 0.92, + "learning_rate": 3.1645533586584065e-06, + "loss": 0.5685, + "step": 30973 + }, + { + "epoch": 0.92, + "learning_rate": 3.1621472674752463e-06, + "loss": 0.5906, + "step": 30974 + }, + { + "epoch": 0.92, + "learning_rate": 3.1597420766518725e-06, + "loss": 0.5435, + "step": 30975 + }, + { + "epoch": 0.92, + "learning_rate": 3.157337786210601e-06, + "loss": 0.5693, + "step": 30976 + }, + { + "epoch": 0.92, + "learning_rate": 3.1549343961738144e-06, + "loss": 0.4633, + "step": 30977 + }, + { + "epoch": 0.92, + "learning_rate": 3.1525319065638383e-06, + "loss": 0.5062, + "step": 30978 + }, + { + "epoch": 0.92, + "learning_rate": 3.1501303174030326e-06, + "loss": 0.5088, + "step": 30979 + }, + { + "epoch": 0.92, + "learning_rate": 3.1477296287137026e-06, + "loss": 0.4665, + "step": 30980 + }, + { + "epoch": 0.92, + "learning_rate": 3.145329840518196e-06, + "loss": 0.5634, + "step": 30981 + }, + { + "epoch": 0.92, + "learning_rate": 3.142930952838796e-06, + "loss": 0.4534, + "step": 30982 + }, + { + "epoch": 0.92, + "learning_rate": 3.14053296569784e-06, + "loss": 0.6284, + "step": 30983 + }, + { + "epoch": 0.92, + "learning_rate": 3.1381358791175765e-06, + "loss": 0.5051, + "step": 30984 + }, + { + "epoch": 0.92, + "learning_rate": 3.135739693120332e-06, + "loss": 0.5644, + "step": 30985 + }, + { + "epoch": 0.92, + "learning_rate": 3.133344407728389e-06, + "loss": 0.4612, + "step": 30986 + }, + { + "epoch": 0.92, + "learning_rate": 3.1309500229639967e-06, + "loss": 0.4794, + "step": 30987 + }, + { + "epoch": 0.92, + "learning_rate": 3.128556538849403e-06, + "loss": 0.537, + "step": 30988 + }, + { + "epoch": 0.92, + "learning_rate": 3.126163955406891e-06, + "loss": 0.4762, + "step": 30989 + }, + { + "epoch": 0.92, + "learning_rate": 3.1237722726587094e-06, + "loss": 0.4641, + "step": 30990 + }, + { + "epoch": 0.92, + "learning_rate": 3.121381490627073e-06, + "loss": 0.536, + "step": 30991 + }, + { + "epoch": 0.92, + "learning_rate": 3.118991609334221e-06, + "loss": 0.4446, + "step": 30992 + }, + { + "epoch": 0.92, + "learning_rate": 3.1166026288023564e-06, + "loss": 0.6094, + "step": 30993 + }, + { + "epoch": 0.92, + "learning_rate": 3.114214549053729e-06, + "loss": 0.5047, + "step": 30994 + }, + { + "epoch": 0.92, + "learning_rate": 3.1118273701105094e-06, + "loss": 0.5511, + "step": 30995 + }, + { + "epoch": 0.92, + "learning_rate": 3.1094410919949134e-06, + "loss": 0.5407, + "step": 30996 + }, + { + "epoch": 0.92, + "learning_rate": 3.1070557147291013e-06, + "loss": 0.5505, + "step": 30997 + }, + { + "epoch": 0.92, + "learning_rate": 3.104671238335277e-06, + "loss": 0.4981, + "step": 30998 + }, + { + "epoch": 0.92, + "learning_rate": 3.1022876628356012e-06, + "loss": 0.4423, + "step": 30999 + }, + { + "epoch": 0.92, + "learning_rate": 3.0999049882522333e-06, + "loss": 0.4753, + "step": 31000 + }, + { + "epoch": 0.92, + "learning_rate": 3.097523214607323e-06, + "loss": 0.5089, + "step": 31001 + }, + { + "epoch": 0.92, + "learning_rate": 3.0951423419230296e-06, + "loss": 0.6726, + "step": 31002 + }, + { + "epoch": 0.92, + "learning_rate": 3.0927623702214694e-06, + "loss": 0.4276, + "step": 31003 + }, + { + "epoch": 0.92, + "learning_rate": 3.0903832995247905e-06, + "loss": 0.5123, + "step": 31004 + }, + { + "epoch": 0.92, + "learning_rate": 3.088005129855098e-06, + "loss": 0.5729, + "step": 31005 + }, + { + "epoch": 0.92, + "learning_rate": 3.085627861234519e-06, + "loss": 0.5501, + "step": 31006 + }, + { + "epoch": 0.92, + "learning_rate": 3.083251493685135e-06, + "loss": 0.5107, + "step": 31007 + }, + { + "epoch": 0.92, + "learning_rate": 3.0808760272290517e-06, + "loss": 0.5088, + "step": 31008 + }, + { + "epoch": 0.92, + "learning_rate": 3.07850146188835e-06, + "loss": 0.4691, + "step": 31009 + }, + { + "epoch": 0.92, + "learning_rate": 3.0761277976851353e-06, + "loss": 0.5536, + "step": 31010 + }, + { + "epoch": 0.92, + "learning_rate": 3.0737550346414346e-06, + "loss": 0.4829, + "step": 31011 + }, + { + "epoch": 0.92, + "learning_rate": 3.0713831727793295e-06, + "loss": 0.4469, + "step": 31012 + }, + { + "epoch": 0.92, + "learning_rate": 3.06901221212087e-06, + "loss": 0.5863, + "step": 31013 + }, + { + "epoch": 0.92, + "learning_rate": 3.066642152688115e-06, + "loss": 0.5969, + "step": 31014 + }, + { + "epoch": 0.92, + "learning_rate": 3.0642729945030814e-06, + "loss": 0.5894, + "step": 31015 + }, + { + "epoch": 0.92, + "learning_rate": 3.061904737587806e-06, + "loss": 0.459, + "step": 31016 + }, + { + "epoch": 0.92, + "learning_rate": 3.0595373819642946e-06, + "loss": 0.4821, + "step": 31017 + }, + { + "epoch": 0.92, + "learning_rate": 3.0571709276545733e-06, + "loss": 0.5041, + "step": 31018 + }, + { + "epoch": 0.92, + "learning_rate": 3.054805374680636e-06, + "loss": 0.5279, + "step": 31019 + }, + { + "epoch": 0.92, + "learning_rate": 3.052440723064487e-06, + "loss": 0.6198, + "step": 31020 + }, + { + "epoch": 0.92, + "learning_rate": 3.0500769728280977e-06, + "loss": 0.5017, + "step": 31021 + }, + { + "epoch": 0.92, + "learning_rate": 3.0477141239934614e-06, + "loss": 0.5019, + "step": 31022 + }, + { + "epoch": 0.92, + "learning_rate": 3.0453521765825277e-06, + "loss": 0.6692, + "step": 31023 + }, + { + "epoch": 0.92, + "learning_rate": 3.042991130617279e-06, + "loss": 0.4688, + "step": 31024 + }, + { + "epoch": 0.92, + "learning_rate": 3.0406309861196524e-06, + "loss": 0.5332, + "step": 31025 + }, + { + "epoch": 0.92, + "learning_rate": 3.0382717431115757e-06, + "loss": 0.5063, + "step": 31026 + }, + { + "epoch": 0.92, + "learning_rate": 3.0359134016150203e-06, + "loss": 0.5085, + "step": 31027 + }, + { + "epoch": 0.92, + "learning_rate": 3.0335559616518904e-06, + "loss": 0.4789, + "step": 31028 + }, + { + "epoch": 0.92, + "learning_rate": 3.031199423244113e-06, + "loss": 0.4738, + "step": 31029 + }, + { + "epoch": 0.92, + "learning_rate": 3.0288437864136044e-06, + "loss": 0.4323, + "step": 31030 + }, + { + "epoch": 0.92, + "learning_rate": 3.0264890511822353e-06, + "loss": 0.5153, + "step": 31031 + }, + { + "epoch": 0.92, + "learning_rate": 3.024135217571933e-06, + "loss": 0.4385, + "step": 31032 + }, + { + "epoch": 0.92, + "learning_rate": 3.0217822856045795e-06, + "loss": 0.6353, + "step": 31033 + }, + { + "epoch": 0.92, + "learning_rate": 3.019430255302036e-06, + "loss": 0.4918, + "step": 31034 + }, + { + "epoch": 0.92, + "learning_rate": 3.017079126686173e-06, + "loss": 0.5868, + "step": 31035 + }, + { + "epoch": 0.92, + "learning_rate": 3.0147288997788515e-06, + "loss": 0.5342, + "step": 31036 + }, + { + "epoch": 0.92, + "learning_rate": 3.0123795746019422e-06, + "loss": 0.5707, + "step": 31037 + }, + { + "epoch": 0.92, + "learning_rate": 3.0100311511772616e-06, + "loss": 0.5565, + "step": 31038 + }, + { + "epoch": 0.92, + "learning_rate": 3.0076836295266586e-06, + "loss": 0.4783, + "step": 31039 + }, + { + "epoch": 0.92, + "learning_rate": 3.0053370096719493e-06, + "loss": 0.6161, + "step": 31040 + }, + { + "epoch": 0.92, + "learning_rate": 3.00299129163496e-06, + "loss": 0.479, + "step": 31041 + }, + { + "epoch": 0.92, + "learning_rate": 3.000646475437496e-06, + "loss": 0.5308, + "step": 31042 + }, + { + "epoch": 0.92, + "learning_rate": 2.9983025611013738e-06, + "loss": 0.5176, + "step": 31043 + }, + { + "epoch": 0.92, + "learning_rate": 2.995959548648364e-06, + "loss": 0.5385, + "step": 31044 + }, + { + "epoch": 0.92, + "learning_rate": 2.9936174381002713e-06, + "loss": 0.4735, + "step": 31045 + }, + { + "epoch": 0.92, + "learning_rate": 2.9912762294788456e-06, + "loss": 0.59, + "step": 31046 + }, + { + "epoch": 0.92, + "learning_rate": 2.988935922805869e-06, + "loss": 0.5184, + "step": 31047 + }, + { + "epoch": 0.92, + "learning_rate": 2.986596518103113e-06, + "loss": 0.5133, + "step": 31048 + }, + { + "epoch": 0.92, + "learning_rate": 2.9842580153923162e-06, + "loss": 0.4074, + "step": 31049 + }, + { + "epoch": 0.92, + "learning_rate": 2.981920414695205e-06, + "loss": 0.4373, + "step": 31050 + }, + { + "epoch": 0.92, + "learning_rate": 2.9795837160335404e-06, + "loss": 0.442, + "step": 31051 + }, + { + "epoch": 0.92, + "learning_rate": 2.977247919429038e-06, + "loss": 0.5095, + "step": 31052 + }, + { + "epoch": 0.92, + "learning_rate": 2.9749130249034253e-06, + "loss": 0.4931, + "step": 31053 + }, + { + "epoch": 0.92, + "learning_rate": 2.972579032478395e-06, + "loss": 0.5094, + "step": 31054 + }, + { + "epoch": 0.92, + "learning_rate": 2.9702459421756424e-06, + "loss": 0.5652, + "step": 31055 + }, + { + "epoch": 0.92, + "learning_rate": 2.967913754016871e-06, + "loss": 0.4597, + "step": 31056 + }, + { + "epoch": 0.92, + "learning_rate": 2.9655824680237753e-06, + "loss": 0.5252, + "step": 31057 + }, + { + "epoch": 0.92, + "learning_rate": 2.9632520842180266e-06, + "loss": 0.534, + "step": 31058 + }, + { + "epoch": 0.92, + "learning_rate": 2.9609226026212745e-06, + "loss": 0.4437, + "step": 31059 + }, + { + "epoch": 0.92, + "learning_rate": 2.9585940232551902e-06, + "loss": 0.4608, + "step": 31060 + }, + { + "epoch": 0.92, + "learning_rate": 2.956266346141412e-06, + "loss": 0.5618, + "step": 31061 + }, + { + "epoch": 0.92, + "learning_rate": 2.9539395713016118e-06, + "loss": 0.4522, + "step": 31062 + }, + { + "epoch": 0.92, + "learning_rate": 2.9516136987573827e-06, + "loss": 0.537, + "step": 31063 + }, + { + "epoch": 0.92, + "learning_rate": 2.9492887285303863e-06, + "loss": 0.5154, + "step": 31064 + }, + { + "epoch": 0.92, + "learning_rate": 2.9469646606422043e-06, + "loss": 0.645, + "step": 31065 + }, + { + "epoch": 0.92, + "learning_rate": 2.9446414951144752e-06, + "loss": 0.4413, + "step": 31066 + }, + { + "epoch": 0.92, + "learning_rate": 2.942319231968782e-06, + "loss": 0.5326, + "step": 31067 + }, + { + "epoch": 0.92, + "learning_rate": 2.9399978712267294e-06, + "loss": 0.6262, + "step": 31068 + }, + { + "epoch": 0.92, + "learning_rate": 2.937677412909889e-06, + "loss": 0.4772, + "step": 31069 + }, + { + "epoch": 0.93, + "learning_rate": 2.935357857039833e-06, + "loss": 0.5117, + "step": 31070 + }, + { + "epoch": 0.93, + "learning_rate": 2.9330392036381327e-06, + "loss": 0.5895, + "step": 31071 + }, + { + "epoch": 0.93, + "learning_rate": 2.930721452726359e-06, + "loss": 0.5428, + "step": 31072 + }, + { + "epoch": 0.93, + "learning_rate": 2.9284046043260403e-06, + "loss": 0.5325, + "step": 31073 + }, + { + "epoch": 0.93, + "learning_rate": 2.9260886584587255e-06, + "loss": 0.5664, + "step": 31074 + }, + { + "epoch": 0.93, + "learning_rate": 2.9237736151459416e-06, + "loss": 0.5312, + "step": 31075 + }, + { + "epoch": 0.93, + "learning_rate": 2.921459474409238e-06, + "loss": 0.5573, + "step": 31076 + }, + { + "epoch": 0.93, + "learning_rate": 2.9191462362701092e-06, + "loss": 0.4676, + "step": 31077 + }, + { + "epoch": 0.93, + "learning_rate": 2.91683390075006e-06, + "loss": 0.5979, + "step": 31078 + }, + { + "epoch": 0.93, + "learning_rate": 2.914522467870584e-06, + "loss": 0.5527, + "step": 31079 + }, + { + "epoch": 0.93, + "learning_rate": 2.9122119376532088e-06, + "loss": 0.4612, + "step": 31080 + }, + { + "epoch": 0.93, + "learning_rate": 2.9099023101193723e-06, + "loss": 0.5617, + "step": 31081 + }, + { + "epoch": 0.93, + "learning_rate": 2.90759358529058e-06, + "loss": 0.5121, + "step": 31082 + }, + { + "epoch": 0.93, + "learning_rate": 2.90528576318827e-06, + "loss": 0.4949, + "step": 31083 + }, + { + "epoch": 0.93, + "learning_rate": 2.9029788438339366e-06, + "loss": 0.4658, + "step": 31084 + }, + { + "epoch": 0.93, + "learning_rate": 2.9006728272489847e-06, + "loss": 0.6091, + "step": 31085 + }, + { + "epoch": 0.93, + "learning_rate": 2.898367713454897e-06, + "loss": 0.4159, + "step": 31086 + }, + { + "epoch": 0.93, + "learning_rate": 2.8960635024730675e-06, + "loss": 0.5038, + "step": 31087 + }, + { + "epoch": 0.93, + "learning_rate": 2.893760194324957e-06, + "loss": 0.5356, + "step": 31088 + }, + { + "epoch": 0.93, + "learning_rate": 2.8914577890319372e-06, + "loss": 0.5023, + "step": 31089 + }, + { + "epoch": 0.93, + "learning_rate": 2.8891562866154463e-06, + "loss": 0.5296, + "step": 31090 + }, + { + "epoch": 0.93, + "learning_rate": 2.88685568709689e-06, + "loss": 0.62, + "step": 31091 + }, + { + "epoch": 0.93, + "learning_rate": 2.884555990497639e-06, + "loss": 0.632, + "step": 31092 + }, + { + "epoch": 0.93, + "learning_rate": 2.8822571968390777e-06, + "loss": 0.5497, + "step": 31093 + }, + { + "epoch": 0.93, + "learning_rate": 2.8799593061425768e-06, + "loss": 0.5066, + "step": 31094 + }, + { + "epoch": 0.93, + "learning_rate": 2.8776623184295083e-06, + "loss": 0.5414, + "step": 31095 + }, + { + "epoch": 0.93, + "learning_rate": 2.8753662337212327e-06, + "loss": 0.5222, + "step": 31096 + }, + { + "epoch": 0.93, + "learning_rate": 2.873071052039089e-06, + "loss": 0.5865, + "step": 31097 + }, + { + "epoch": 0.93, + "learning_rate": 2.8707767734044155e-06, + "loss": 0.633, + "step": 31098 + }, + { + "epoch": 0.93, + "learning_rate": 2.8684833978385507e-06, + "loss": 0.5124, + "step": 31099 + }, + { + "epoch": 0.93, + "learning_rate": 2.8661909253628107e-06, + "loss": 0.5283, + "step": 31100 + }, + { + "epoch": 0.93, + "learning_rate": 2.8638993559985228e-06, + "loss": 0.5362, + "step": 31101 + }, + { + "epoch": 0.93, + "learning_rate": 2.8616086897669705e-06, + "loss": 0.5669, + "step": 31102 + }, + { + "epoch": 0.93, + "learning_rate": 2.8593189266894803e-06, + "loss": 0.6176, + "step": 31103 + }, + { + "epoch": 0.93, + "learning_rate": 2.857030066787314e-06, + "loss": 0.5847, + "step": 31104 + }, + { + "epoch": 0.93, + "learning_rate": 2.854742110081776e-06, + "loss": 0.5976, + "step": 31105 + }, + { + "epoch": 0.93, + "learning_rate": 2.8524550565941168e-06, + "loss": 0.5466, + "step": 31106 + }, + { + "epoch": 0.93, + "learning_rate": 2.850168906345618e-06, + "loss": 0.5347, + "step": 31107 + }, + { + "epoch": 0.93, + "learning_rate": 2.84788365935752e-06, + "loss": 0.5759, + "step": 31108 + }, + { + "epoch": 0.93, + "learning_rate": 2.8455993156510817e-06, + "loss": 0.5442, + "step": 31109 + }, + { + "epoch": 0.93, + "learning_rate": 2.843315875247543e-06, + "loss": 0.4856, + "step": 31110 + }, + { + "epoch": 0.93, + "learning_rate": 2.841033338168142e-06, + "loss": 0.5211, + "step": 31111 + }, + { + "epoch": 0.93, + "learning_rate": 2.8387517044340726e-06, + "loss": 0.5258, + "step": 31112 + }, + { + "epoch": 0.93, + "learning_rate": 2.836470974066563e-06, + "loss": 0.388, + "step": 31113 + }, + { + "epoch": 0.93, + "learning_rate": 2.8341911470868176e-06, + "loss": 0.4759, + "step": 31114 + }, + { + "epoch": 0.93, + "learning_rate": 2.8319122235160645e-06, + "loss": 0.5338, + "step": 31115 + }, + { + "epoch": 0.93, + "learning_rate": 2.829634203375442e-06, + "loss": 0.4772, + "step": 31116 + }, + { + "epoch": 0.93, + "learning_rate": 2.8273570866861553e-06, + "loss": 0.6194, + "step": 31117 + }, + { + "epoch": 0.93, + "learning_rate": 2.8250808734693658e-06, + "loss": 0.64, + "step": 31118 + }, + { + "epoch": 0.93, + "learning_rate": 2.8228055637462558e-06, + "loss": 0.5772, + "step": 31119 + }, + { + "epoch": 0.93, + "learning_rate": 2.8205311575379754e-06, + "loss": 0.4933, + "step": 31120 + }, + { + "epoch": 0.93, + "learning_rate": 2.8182576548656415e-06, + "loss": 0.7203, + "step": 31121 + }, + { + "epoch": 0.93, + "learning_rate": 2.8159850557504254e-06, + "loss": 0.5294, + "step": 31122 + }, + { + "epoch": 0.93, + "learning_rate": 2.813713360213455e-06, + "loss": 0.5384, + "step": 31123 + }, + { + "epoch": 0.93, + "learning_rate": 2.811442568275835e-06, + "loss": 0.514, + "step": 31124 + }, + { + "epoch": 0.93, + "learning_rate": 2.8091726799586937e-06, + "loss": 0.4549, + "step": 31125 + }, + { + "epoch": 0.93, + "learning_rate": 2.806903695283125e-06, + "loss": 0.5061, + "step": 31126 + }, + { + "epoch": 0.93, + "learning_rate": 2.8046356142702234e-06, + "loss": 0.5962, + "step": 31127 + }, + { + "epoch": 0.93, + "learning_rate": 2.802368436941083e-06, + "loss": 0.4841, + "step": 31128 + }, + { + "epoch": 0.93, + "learning_rate": 2.800102163316787e-06, + "loss": 0.5093, + "step": 31129 + }, + { + "epoch": 0.93, + "learning_rate": 2.797836793418407e-06, + "loss": 0.3906, + "step": 31130 + }, + { + "epoch": 0.93, + "learning_rate": 2.795572327266982e-06, + "loss": 0.5941, + "step": 31131 + }, + { + "epoch": 0.93, + "learning_rate": 2.793308764883584e-06, + "loss": 0.573, + "step": 31132 + }, + { + "epoch": 0.93, + "learning_rate": 2.7910461062892632e-06, + "loss": 0.6144, + "step": 31133 + }, + { + "epoch": 0.93, + "learning_rate": 2.788784351505058e-06, + "loss": 0.5209, + "step": 31134 + }, + { + "epoch": 0.93, + "learning_rate": 2.786523500551996e-06, + "loss": 0.5462, + "step": 31135 + }, + { + "epoch": 0.93, + "learning_rate": 2.784263553451083e-06, + "loss": 0.4772, + "step": 31136 + }, + { + "epoch": 0.93, + "learning_rate": 2.782004510223335e-06, + "loss": 0.4331, + "step": 31137 + }, + { + "epoch": 0.93, + "learning_rate": 2.77974637088978e-06, + "loss": 0.5469, + "step": 31138 + }, + { + "epoch": 0.93, + "learning_rate": 2.7774891354713893e-06, + "loss": 0.5634, + "step": 31139 + }, + { + "epoch": 0.93, + "learning_rate": 2.7752328039891475e-06, + "loss": 0.5295, + "step": 31140 + }, + { + "epoch": 0.93, + "learning_rate": 2.7729773764640365e-06, + "loss": 0.4986, + "step": 31141 + }, + { + "epoch": 0.93, + "learning_rate": 2.7707228529170515e-06, + "loss": 0.5282, + "step": 31142 + }, + { + "epoch": 0.93, + "learning_rate": 2.7684692333691085e-06, + "loss": 0.5636, + "step": 31143 + }, + { + "epoch": 0.93, + "learning_rate": 2.766216517841214e-06, + "loss": 0.5181, + "step": 31144 + }, + { + "epoch": 0.93, + "learning_rate": 2.7639647063542607e-06, + "loss": 0.5778, + "step": 31145 + }, + { + "epoch": 0.93, + "learning_rate": 2.761713798929222e-06, + "loss": 0.5272, + "step": 31146 + }, + { + "epoch": 0.93, + "learning_rate": 2.759463795587003e-06, + "loss": 0.5385, + "step": 31147 + }, + { + "epoch": 0.93, + "learning_rate": 2.7572146963485536e-06, + "loss": 0.4965, + "step": 31148 + }, + { + "epoch": 0.93, + "learning_rate": 2.7549665012347457e-06, + "loss": 0.4674, + "step": 31149 + }, + { + "epoch": 0.93, + "learning_rate": 2.752719210266508e-06, + "loss": 0.552, + "step": 31150 + }, + { + "epoch": 0.93, + "learning_rate": 2.7504728234647225e-06, + "loss": 0.425, + "step": 31151 + }, + { + "epoch": 0.93, + "learning_rate": 2.7482273408502846e-06, + "loss": 0.5385, + "step": 31152 + }, + { + "epoch": 0.93, + "learning_rate": 2.745982762444066e-06, + "loss": 0.6169, + "step": 31153 + }, + { + "epoch": 0.93, + "learning_rate": 2.74373908826695e-06, + "loss": 0.53, + "step": 31154 + }, + { + "epoch": 0.93, + "learning_rate": 2.7414963183397757e-06, + "loss": 0.4781, + "step": 31155 + }, + { + "epoch": 0.93, + "learning_rate": 2.739254452683393e-06, + "loss": 0.4776, + "step": 31156 + }, + { + "epoch": 0.93, + "learning_rate": 2.737013491318674e-06, + "loss": 0.6132, + "step": 31157 + }, + { + "epoch": 0.93, + "learning_rate": 2.7347734342664465e-06, + "loss": 0.4581, + "step": 31158 + }, + { + "epoch": 0.93, + "learning_rate": 2.7325342815475164e-06, + "loss": 0.5997, + "step": 31159 + }, + { + "epoch": 0.93, + "learning_rate": 2.730296033182711e-06, + "loss": 0.5641, + "step": 31160 + }, + { + "epoch": 0.93, + "learning_rate": 2.7280586891928585e-06, + "loss": 0.4445, + "step": 31161 + }, + { + "epoch": 0.93, + "learning_rate": 2.7258222495987307e-06, + "loss": 0.4526, + "step": 31162 + }, + { + "epoch": 0.93, + "learning_rate": 2.7235867144211557e-06, + "loss": 0.4941, + "step": 31163 + }, + { + "epoch": 0.93, + "learning_rate": 2.721352083680884e-06, + "loss": 0.5761, + "step": 31164 + }, + { + "epoch": 0.93, + "learning_rate": 2.719118357398731e-06, + "loss": 0.6204, + "step": 31165 + }, + { + "epoch": 0.93, + "learning_rate": 2.716885535595426e-06, + "loss": 0.5623, + "step": 31166 + }, + { + "epoch": 0.93, + "learning_rate": 2.7146536182917513e-06, + "loss": 0.5668, + "step": 31167 + }, + { + "epoch": 0.93, + "learning_rate": 2.7124226055084466e-06, + "loss": 0.6158, + "step": 31168 + }, + { + "epoch": 0.93, + "learning_rate": 2.710192497266273e-06, + "loss": 0.4491, + "step": 31169 + }, + { + "epoch": 0.93, + "learning_rate": 2.7079632935859468e-06, + "loss": 0.4609, + "step": 31170 + }, + { + "epoch": 0.93, + "learning_rate": 2.705734994488196e-06, + "loss": 0.5169, + "step": 31171 + }, + { + "epoch": 0.93, + "learning_rate": 2.7035075999937485e-06, + "loss": 0.5071, + "step": 31172 + }, + { + "epoch": 0.93, + "learning_rate": 2.7012811101233214e-06, + "loss": 0.4988, + "step": 31173 + }, + { + "epoch": 0.93, + "learning_rate": 2.699055524897587e-06, + "loss": 0.5279, + "step": 31174 + }, + { + "epoch": 0.93, + "learning_rate": 2.6968308443372502e-06, + "loss": 0.5821, + "step": 31175 + }, + { + "epoch": 0.93, + "learning_rate": 2.6946070684630066e-06, + "loss": 0.3973, + "step": 31176 + }, + { + "epoch": 0.93, + "learning_rate": 2.6923841972955387e-06, + "loss": 0.4863, + "step": 31177 + }, + { + "epoch": 0.93, + "learning_rate": 2.6901622308554754e-06, + "loss": 0.5703, + "step": 31178 + }, + { + "epoch": 0.93, + "learning_rate": 2.6879411691635103e-06, + "loss": 0.618, + "step": 31179 + }, + { + "epoch": 0.93, + "learning_rate": 2.6857210122402833e-06, + "loss": 0.5037, + "step": 31180 + }, + { + "epoch": 0.93, + "learning_rate": 2.683501760106444e-06, + "loss": 0.6007, + "step": 31181 + }, + { + "epoch": 0.93, + "learning_rate": 2.6812834127826093e-06, + "loss": 0.4648, + "step": 31182 + }, + { + "epoch": 0.93, + "learning_rate": 2.6790659702894184e-06, + "loss": 0.4766, + "step": 31183 + }, + { + "epoch": 0.93, + "learning_rate": 2.6768494326474767e-06, + "loss": 0.5582, + "step": 31184 + }, + { + "epoch": 0.93, + "learning_rate": 2.6746337998774017e-06, + "loss": 0.5768, + "step": 31185 + }, + { + "epoch": 0.93, + "learning_rate": 2.6724190719997986e-06, + "loss": 0.6823, + "step": 31186 + }, + { + "epoch": 0.93, + "learning_rate": 2.6702052490352516e-06, + "loss": 0.5107, + "step": 31187 + }, + { + "epoch": 0.93, + "learning_rate": 2.667992331004332e-06, + "loss": 0.5379, + "step": 31188 + }, + { + "epoch": 0.93, + "learning_rate": 2.6657803179276464e-06, + "loss": 0.4859, + "step": 31189 + }, + { + "epoch": 0.93, + "learning_rate": 2.6635692098257227e-06, + "loss": 0.5104, + "step": 31190 + }, + { + "epoch": 0.93, + "learning_rate": 2.6613590067191553e-06, + "loss": 0.5829, + "step": 31191 + }, + { + "epoch": 0.93, + "learning_rate": 2.659149708628472e-06, + "loss": 0.5244, + "step": 31192 + }, + { + "epoch": 0.93, + "learning_rate": 2.656941315574213e-06, + "loss": 0.5187, + "step": 31193 + }, + { + "epoch": 0.93, + "learning_rate": 2.6547338275769164e-06, + "loss": 0.4611, + "step": 31194 + }, + { + "epoch": 0.93, + "learning_rate": 2.652527244657099e-06, + "loss": 0.5872, + "step": 31195 + }, + { + "epoch": 0.93, + "learning_rate": 2.650321566835301e-06, + "loss": 0.5752, + "step": 31196 + }, + { + "epoch": 0.93, + "learning_rate": 2.6481167941320053e-06, + "loss": 0.5861, + "step": 31197 + }, + { + "epoch": 0.93, + "learning_rate": 2.6459129265677175e-06, + "loss": 0.5459, + "step": 31198 + }, + { + "epoch": 0.93, + "learning_rate": 2.6437099641629326e-06, + "loss": 0.4976, + "step": 31199 + }, + { + "epoch": 0.93, + "learning_rate": 2.6415079069381344e-06, + "loss": 0.5175, + "step": 31200 + }, + { + "epoch": 0.93, + "learning_rate": 2.639306754913784e-06, + "loss": 0.4186, + "step": 31201 + }, + { + "epoch": 0.93, + "learning_rate": 2.637106508110354e-06, + "loss": 0.6217, + "step": 31202 + }, + { + "epoch": 0.93, + "learning_rate": 2.634907166548306e-06, + "loss": 0.5636, + "step": 31203 + }, + { + "epoch": 0.93, + "learning_rate": 2.6327087302480904e-06, + "loss": 0.5705, + "step": 31204 + }, + { + "epoch": 0.93, + "learning_rate": 2.6305111992301347e-06, + "loss": 0.5434, + "step": 31205 + }, + { + "epoch": 0.93, + "learning_rate": 2.6283145735148894e-06, + "loss": 0.4613, + "step": 31206 + }, + { + "epoch": 0.93, + "learning_rate": 2.6261188531227497e-06, + "loss": 0.5241, + "step": 31207 + }, + { + "epoch": 0.93, + "learning_rate": 2.623924038074166e-06, + "loss": 0.559, + "step": 31208 + }, + { + "epoch": 0.93, + "learning_rate": 2.62173012838951e-06, + "loss": 0.5866, + "step": 31209 + }, + { + "epoch": 0.93, + "learning_rate": 2.619537124089211e-06, + "loss": 0.6244, + "step": 31210 + }, + { + "epoch": 0.93, + "learning_rate": 2.6173450251936293e-06, + "loss": 0.4587, + "step": 31211 + }, + { + "epoch": 0.93, + "learning_rate": 2.6151538317231826e-06, + "loss": 0.6629, + "step": 31212 + }, + { + "epoch": 0.93, + "learning_rate": 2.6129635436981993e-06, + "loss": 0.6065, + "step": 31213 + }, + { + "epoch": 0.93, + "learning_rate": 2.6107741611390623e-06, + "loss": 0.5117, + "step": 31214 + }, + { + "epoch": 0.93, + "learning_rate": 2.608585684066156e-06, + "loss": 0.5779, + "step": 31215 + }, + { + "epoch": 0.93, + "learning_rate": 2.606398112499797e-06, + "loss": 0.5149, + "step": 31216 + }, + { + "epoch": 0.93, + "learning_rate": 2.604211446460314e-06, + "loss": 0.6957, + "step": 31217 + }, + { + "epoch": 0.93, + "learning_rate": 2.602025685968068e-06, + "loss": 0.5998, + "step": 31218 + }, + { + "epoch": 0.93, + "learning_rate": 2.599840831043354e-06, + "loss": 0.6234, + "step": 31219 + }, + { + "epoch": 0.93, + "learning_rate": 2.597656881706534e-06, + "loss": 0.5207, + "step": 31220 + }, + { + "epoch": 0.93, + "learning_rate": 2.5954738379778464e-06, + "loss": 0.5285, + "step": 31221 + }, + { + "epoch": 0.93, + "learning_rate": 2.5932916998776315e-06, + "loss": 0.4864, + "step": 31222 + }, + { + "epoch": 0.93, + "learning_rate": 2.591110467426172e-06, + "loss": 0.5507, + "step": 31223 + }, + { + "epoch": 0.93, + "learning_rate": 2.5889301406437416e-06, + "loss": 0.4909, + "step": 31224 + }, + { + "epoch": 0.93, + "learning_rate": 2.5867507195506235e-06, + "loss": 0.5639, + "step": 31225 + }, + { + "epoch": 0.93, + "learning_rate": 2.584572204167057e-06, + "loss": 0.5228, + "step": 31226 + }, + { + "epoch": 0.93, + "learning_rate": 2.5823945945133156e-06, + "loss": 0.4702, + "step": 31227 + }, + { + "epoch": 0.93, + "learning_rate": 2.5802178906096488e-06, + "loss": 0.4149, + "step": 31228 + }, + { + "epoch": 0.93, + "learning_rate": 2.578042092476274e-06, + "loss": 0.4902, + "step": 31229 + }, + { + "epoch": 0.93, + "learning_rate": 2.575867200133453e-06, + "loss": 0.6118, + "step": 31230 + }, + { + "epoch": 0.93, + "learning_rate": 2.57369321360138e-06, + "loss": 0.4297, + "step": 31231 + }, + { + "epoch": 0.93, + "learning_rate": 2.5715201329002847e-06, + "loss": 0.4953, + "step": 31232 + }, + { + "epoch": 0.93, + "learning_rate": 2.56934795805035e-06, + "loss": 0.4439, + "step": 31233 + }, + { + "epoch": 0.93, + "learning_rate": 2.567176689071804e-06, + "loss": 0.5221, + "step": 31234 + }, + { + "epoch": 0.93, + "learning_rate": 2.5650063259848088e-06, + "loss": 0.6584, + "step": 31235 + }, + { + "epoch": 0.93, + "learning_rate": 2.562836868809537e-06, + "loss": 0.4997, + "step": 31236 + }, + { + "epoch": 0.93, + "learning_rate": 2.560668317566184e-06, + "loss": 0.5126, + "step": 31237 + }, + { + "epoch": 0.93, + "learning_rate": 2.5585006722748995e-06, + "loss": 0.5865, + "step": 31238 + }, + { + "epoch": 0.93, + "learning_rate": 2.5563339329558566e-06, + "loss": 0.5725, + "step": 31239 + }, + { + "epoch": 0.93, + "learning_rate": 2.5541680996291617e-06, + "loss": 0.4817, + "step": 31240 + }, + { + "epoch": 0.93, + "learning_rate": 2.5520031723149874e-06, + "loss": 0.6664, + "step": 31241 + }, + { + "epoch": 0.93, + "learning_rate": 2.5498391510334285e-06, + "loss": 0.5539, + "step": 31242 + }, + { + "epoch": 0.93, + "learning_rate": 2.547676035804647e-06, + "loss": 0.5516, + "step": 31243 + }, + { + "epoch": 0.93, + "learning_rate": 2.545513826648738e-06, + "loss": 0.4938, + "step": 31244 + }, + { + "epoch": 0.93, + "learning_rate": 2.5433525235857736e-06, + "loss": 0.6237, + "step": 31245 + }, + { + "epoch": 0.93, + "learning_rate": 2.541192126635894e-06, + "loss": 0.502, + "step": 31246 + }, + { + "epoch": 0.93, + "learning_rate": 2.5390326358191606e-06, + "loss": 0.6216, + "step": 31247 + }, + { + "epoch": 0.93, + "learning_rate": 2.5368740511556578e-06, + "loss": 0.4572, + "step": 31248 + }, + { + "epoch": 0.93, + "learning_rate": 2.5347163726654576e-06, + "loss": 0.5081, + "step": 31249 + }, + { + "epoch": 0.93, + "learning_rate": 2.5325596003686115e-06, + "loss": 0.6043, + "step": 31250 + }, + { + "epoch": 0.93, + "learning_rate": 2.5304037342851915e-06, + "loss": 0.5128, + "step": 31251 + }, + { + "epoch": 0.93, + "learning_rate": 2.5282487744352266e-06, + "loss": 0.4561, + "step": 31252 + }, + { + "epoch": 0.93, + "learning_rate": 2.5260947208387563e-06, + "loss": 0.4872, + "step": 31253 + }, + { + "epoch": 0.93, + "learning_rate": 2.5239415735157978e-06, + "loss": 0.4777, + "step": 31254 + }, + { + "epoch": 0.93, + "learning_rate": 2.5217893324863904e-06, + "loss": 0.5602, + "step": 31255 + }, + { + "epoch": 0.93, + "learning_rate": 2.51963799777053e-06, + "loss": 0.5076, + "step": 31256 + }, + { + "epoch": 0.93, + "learning_rate": 2.5174875693882216e-06, + "loss": 0.5855, + "step": 31257 + }, + { + "epoch": 0.93, + "learning_rate": 2.5153380473594722e-06, + "loss": 0.6028, + "step": 31258 + }, + { + "epoch": 0.93, + "learning_rate": 2.5131894317042548e-06, + "loss": 0.5936, + "step": 31259 + }, + { + "epoch": 0.93, + "learning_rate": 2.511041722442542e-06, + "loss": 0.4988, + "step": 31260 + }, + { + "epoch": 0.93, + "learning_rate": 2.508894919594296e-06, + "loss": 0.5871, + "step": 31261 + }, + { + "epoch": 0.93, + "learning_rate": 2.506749023179511e-06, + "loss": 0.4902, + "step": 31262 + }, + { + "epoch": 0.93, + "learning_rate": 2.5046040332181163e-06, + "loss": 0.5816, + "step": 31263 + }, + { + "epoch": 0.93, + "learning_rate": 2.5024599497300404e-06, + "loss": 0.5086, + "step": 31264 + }, + { + "epoch": 0.93, + "learning_rate": 2.5003167727352338e-06, + "loss": 0.4917, + "step": 31265 + }, + { + "epoch": 0.93, + "learning_rate": 2.498174502253625e-06, + "loss": 0.4667, + "step": 31266 + }, + { + "epoch": 0.93, + "learning_rate": 2.496033138305132e-06, + "loss": 0.513, + "step": 31267 + }, + { + "epoch": 0.93, + "learning_rate": 2.4938926809096707e-06, + "loss": 0.5995, + "step": 31268 + }, + { + "epoch": 0.93, + "learning_rate": 2.4917531300871157e-06, + "loss": 0.4591, + "step": 31269 + }, + { + "epoch": 0.93, + "learning_rate": 2.4896144858573945e-06, + "loss": 0.5631, + "step": 31270 + }, + { + "epoch": 0.93, + "learning_rate": 2.487476748240358e-06, + "loss": 0.4474, + "step": 31271 + }, + { + "epoch": 0.93, + "learning_rate": 2.4853399172559022e-06, + "loss": 0.5121, + "step": 31272 + }, + { + "epoch": 0.93, + "learning_rate": 2.4832039929238884e-06, + "loss": 0.5193, + "step": 31273 + }, + { + "epoch": 0.93, + "learning_rate": 2.4810689752641892e-06, + "loss": 0.4823, + "step": 31274 + }, + { + "epoch": 0.93, + "learning_rate": 2.478934864296634e-06, + "loss": 0.5341, + "step": 31275 + }, + { + "epoch": 0.93, + "learning_rate": 2.4768016600410725e-06, + "loss": 0.5625, + "step": 31276 + }, + { + "epoch": 0.93, + "learning_rate": 2.4746693625173455e-06, + "loss": 0.4696, + "step": 31277 + }, + { + "epoch": 0.93, + "learning_rate": 2.4725379717452812e-06, + "loss": 0.4445, + "step": 31278 + }, + { + "epoch": 0.93, + "learning_rate": 2.4704074877446748e-06, + "loss": 0.5243, + "step": 31279 + }, + { + "epoch": 0.93, + "learning_rate": 2.4682779105353437e-06, + "loss": 0.4696, + "step": 31280 + }, + { + "epoch": 0.93, + "learning_rate": 2.4661492401370943e-06, + "loss": 0.4686, + "step": 31281 + }, + { + "epoch": 0.93, + "learning_rate": 2.4640214765697444e-06, + "loss": 0.5008, + "step": 31282 + }, + { + "epoch": 0.93, + "learning_rate": 2.4618946198530225e-06, + "loss": 0.5887, + "step": 31283 + }, + { + "epoch": 0.93, + "learning_rate": 2.4597686700067235e-06, + "loss": 0.5584, + "step": 31284 + }, + { + "epoch": 0.93, + "learning_rate": 2.457643627050632e-06, + "loss": 0.4652, + "step": 31285 + }, + { + "epoch": 0.93, + "learning_rate": 2.4555194910044987e-06, + "loss": 0.4936, + "step": 31286 + }, + { + "epoch": 0.93, + "learning_rate": 2.4533962618880636e-06, + "loss": 0.5075, + "step": 31287 + }, + { + "epoch": 0.93, + "learning_rate": 2.451273939721066e-06, + "loss": 0.5688, + "step": 31288 + }, + { + "epoch": 0.93, + "learning_rate": 2.449152524523246e-06, + "loss": 0.6069, + "step": 31289 + }, + { + "epoch": 0.93, + "learning_rate": 2.4470320163143213e-06, + "loss": 0.4855, + "step": 31290 + }, + { + "epoch": 0.93, + "learning_rate": 2.4449124151140203e-06, + "loss": 0.5554, + "step": 31291 + }, + { + "epoch": 0.93, + "learning_rate": 2.442793720942038e-06, + "loss": 0.4861, + "step": 31292 + }, + { + "epoch": 0.93, + "learning_rate": 2.440675933818093e-06, + "loss": 0.5141, + "step": 31293 + }, + { + "epoch": 0.93, + "learning_rate": 2.4385590537618352e-06, + "loss": 0.5189, + "step": 31294 + }, + { + "epoch": 0.93, + "learning_rate": 2.4364430807929826e-06, + "loss": 0.6048, + "step": 31295 + }, + { + "epoch": 0.93, + "learning_rate": 2.4343280149311974e-06, + "loss": 0.5006, + "step": 31296 + }, + { + "epoch": 0.93, + "learning_rate": 2.4322138561961526e-06, + "loss": 0.5518, + "step": 31297 + }, + { + "epoch": 0.93, + "learning_rate": 2.430100604607477e-06, + "loss": 0.4636, + "step": 31298 + }, + { + "epoch": 0.93, + "learning_rate": 2.4279882601848434e-06, + "loss": 0.5372, + "step": 31299 + }, + { + "epoch": 0.93, + "learning_rate": 2.4258768229478925e-06, + "loss": 0.4723, + "step": 31300 + }, + { + "epoch": 0.93, + "learning_rate": 2.4237662929162517e-06, + "loss": 0.5281, + "step": 31301 + }, + { + "epoch": 0.93, + "learning_rate": 2.421656670109551e-06, + "loss": 0.5877, + "step": 31302 + }, + { + "epoch": 0.93, + "learning_rate": 2.4195479545473743e-06, + "loss": 0.4543, + "step": 31303 + }, + { + "epoch": 0.93, + "learning_rate": 2.417440146249361e-06, + "loss": 0.4942, + "step": 31304 + }, + { + "epoch": 0.93, + "learning_rate": 2.4153332452350963e-06, + "loss": 0.4649, + "step": 31305 + }, + { + "epoch": 0.93, + "learning_rate": 2.413227251524175e-06, + "loss": 0.5109, + "step": 31306 + }, + { + "epoch": 0.93, + "learning_rate": 2.411122165136159e-06, + "loss": 0.5846, + "step": 31307 + }, + { + "epoch": 0.93, + "learning_rate": 2.4090179860906338e-06, + "loss": 0.4092, + "step": 31308 + }, + { + "epoch": 0.93, + "learning_rate": 2.4069147144071713e-06, + "loss": 0.4903, + "step": 31309 + }, + { + "epoch": 0.93, + "learning_rate": 2.4048123501053234e-06, + "loss": 0.4934, + "step": 31310 + }, + { + "epoch": 0.93, + "learning_rate": 2.4027108932046295e-06, + "loss": 0.5284, + "step": 31311 + }, + { + "epoch": 0.93, + "learning_rate": 2.4006103437246188e-06, + "loss": 0.4618, + "step": 31312 + }, + { + "epoch": 0.93, + "learning_rate": 2.398510701684853e-06, + "loss": 0.5032, + "step": 31313 + }, + { + "epoch": 0.93, + "learning_rate": 2.396411967104817e-06, + "loss": 0.4995, + "step": 31314 + }, + { + "epoch": 0.93, + "learning_rate": 2.3943141400040612e-06, + "loss": 0.4963, + "step": 31315 + }, + { + "epoch": 0.93, + "learning_rate": 2.3922172204020488e-06, + "loss": 0.5751, + "step": 31316 + }, + { + "epoch": 0.93, + "learning_rate": 2.390121208318319e-06, + "loss": 0.4513, + "step": 31317 + }, + { + "epoch": 0.93, + "learning_rate": 2.3880261037723227e-06, + "loss": 0.6628, + "step": 31318 + }, + { + "epoch": 0.93, + "learning_rate": 2.385931906783556e-06, + "loss": 0.6267, + "step": 31319 + }, + { + "epoch": 0.93, + "learning_rate": 2.3838386173715033e-06, + "loss": 0.4769, + "step": 31320 + }, + { + "epoch": 0.93, + "learning_rate": 2.381746235555604e-06, + "loss": 0.437, + "step": 31321 + }, + { + "epoch": 0.93, + "learning_rate": 2.379654761355321e-06, + "loss": 0.5552, + "step": 31322 + }, + { + "epoch": 0.93, + "learning_rate": 2.377564194790094e-06, + "loss": 0.4566, + "step": 31323 + }, + { + "epoch": 0.93, + "learning_rate": 2.375474535879385e-06, + "loss": 0.5889, + "step": 31324 + }, + { + "epoch": 0.93, + "learning_rate": 2.3733857846426012e-06, + "loss": 0.5529, + "step": 31325 + }, + { + "epoch": 0.93, + "learning_rate": 2.3712979410991487e-06, + "loss": 0.4195, + "step": 31326 + }, + { + "epoch": 0.93, + "learning_rate": 2.369211005268457e-06, + "loss": 0.5952, + "step": 31327 + }, + { + "epoch": 0.93, + "learning_rate": 2.3671249771699542e-06, + "loss": 0.5628, + "step": 31328 + }, + { + "epoch": 0.93, + "learning_rate": 2.3650398568229814e-06, + "loss": 0.4592, + "step": 31329 + }, + { + "epoch": 0.93, + "learning_rate": 2.362955644246978e-06, + "loss": 0.5275, + "step": 31330 + }, + { + "epoch": 0.93, + "learning_rate": 2.3608723394612842e-06, + "loss": 0.5529, + "step": 31331 + }, + { + "epoch": 0.93, + "learning_rate": 2.3587899424852956e-06, + "loss": 0.5567, + "step": 31332 + }, + { + "epoch": 0.93, + "learning_rate": 2.3567084533383523e-06, + "loss": 0.5324, + "step": 31333 + }, + { + "epoch": 0.93, + "learning_rate": 2.3546278720398272e-06, + "loss": 0.4596, + "step": 31334 + }, + { + "epoch": 0.93, + "learning_rate": 2.35254819860905e-06, + "loss": 0.5112, + "step": 31335 + }, + { + "epoch": 0.93, + "learning_rate": 2.350469433065361e-06, + "loss": 0.4023, + "step": 31336 + }, + { + "epoch": 0.93, + "learning_rate": 2.348391575428077e-06, + "loss": 0.6581, + "step": 31337 + }, + { + "epoch": 0.93, + "learning_rate": 2.346314625716539e-06, + "loss": 0.5134, + "step": 31338 + }, + { + "epoch": 0.93, + "learning_rate": 2.3442385839500536e-06, + "loss": 0.6267, + "step": 31339 + }, + { + "epoch": 0.93, + "learning_rate": 2.342163450147905e-06, + "loss": 0.5164, + "step": 31340 + }, + { + "epoch": 0.93, + "learning_rate": 2.3400892243294002e-06, + "loss": 0.472, + "step": 31341 + }, + { + "epoch": 0.93, + "learning_rate": 2.338015906513813e-06, + "loss": 0.5326, + "step": 31342 + }, + { + "epoch": 0.93, + "learning_rate": 2.3359434967204385e-06, + "loss": 0.5678, + "step": 31343 + }, + { + "epoch": 0.93, + "learning_rate": 2.3338719949685505e-06, + "loss": 0.6088, + "step": 31344 + }, + { + "epoch": 0.93, + "learning_rate": 2.331801401277378e-06, + "loss": 0.601, + "step": 31345 + }, + { + "epoch": 0.93, + "learning_rate": 2.3297317156661947e-06, + "loss": 0.5196, + "step": 31346 + }, + { + "epoch": 0.93, + "learning_rate": 2.3276629381542292e-06, + "loss": 0.5894, + "step": 31347 + }, + { + "epoch": 0.93, + "learning_rate": 2.325595068760744e-06, + "loss": 0.4414, + "step": 31348 + }, + { + "epoch": 0.93, + "learning_rate": 2.3235281075049353e-06, + "loss": 0.4338, + "step": 31349 + }, + { + "epoch": 0.93, + "learning_rate": 2.3214620544060315e-06, + "loss": 0.6526, + "step": 31350 + }, + { + "epoch": 0.93, + "learning_rate": 2.3193969094832402e-06, + "loss": 0.4447, + "step": 31351 + }, + { + "epoch": 0.93, + "learning_rate": 2.3173326727557674e-06, + "loss": 0.554, + "step": 31352 + }, + { + "epoch": 0.93, + "learning_rate": 2.3152693442427985e-06, + "loss": 0.6043, + "step": 31353 + }, + { + "epoch": 0.93, + "learning_rate": 2.31320692396354e-06, + "loss": 0.6216, + "step": 31354 + }, + { + "epoch": 0.93, + "learning_rate": 2.3111454119371436e-06, + "loss": 0.6713, + "step": 31355 + }, + { + "epoch": 0.93, + "learning_rate": 2.309084808182771e-06, + "loss": 0.5164, + "step": 31356 + }, + { + "epoch": 0.93, + "learning_rate": 2.307025112719585e-06, + "loss": 0.4742, + "step": 31357 + }, + { + "epoch": 0.93, + "learning_rate": 2.3049663255667596e-06, + "loss": 0.4452, + "step": 31358 + }, + { + "epoch": 0.93, + "learning_rate": 2.302908446743424e-06, + "loss": 0.469, + "step": 31359 + }, + { + "epoch": 0.93, + "learning_rate": 2.300851476268695e-06, + "loss": 0.5196, + "step": 31360 + }, + { + "epoch": 0.93, + "learning_rate": 2.298795414161703e-06, + "loss": 0.5531, + "step": 31361 + }, + { + "epoch": 0.93, + "learning_rate": 2.296740260441577e-06, + "loss": 0.5439, + "step": 31362 + }, + { + "epoch": 0.93, + "learning_rate": 2.294686015127434e-06, + "loss": 0.4803, + "step": 31363 + }, + { + "epoch": 0.93, + "learning_rate": 2.2926326782383488e-06, + "loss": 0.4391, + "step": 31364 + }, + { + "epoch": 0.93, + "learning_rate": 2.290580249793428e-06, + "loss": 0.5917, + "step": 31365 + }, + { + "epoch": 0.93, + "learning_rate": 2.2885287298117342e-06, + "loss": 0.5067, + "step": 31366 + }, + { + "epoch": 0.93, + "learning_rate": 2.286478118312374e-06, + "loss": 0.5183, + "step": 31367 + }, + { + "epoch": 0.93, + "learning_rate": 2.284428415314388e-06, + "loss": 0.4558, + "step": 31368 + }, + { + "epoch": 0.93, + "learning_rate": 2.2823796208368386e-06, + "loss": 0.3753, + "step": 31369 + }, + { + "epoch": 0.93, + "learning_rate": 2.2803317348987775e-06, + "loss": 0.4869, + "step": 31370 + }, + { + "epoch": 0.93, + "learning_rate": 2.2782847575192447e-06, + "loss": 0.5317, + "step": 31371 + }, + { + "epoch": 0.93, + "learning_rate": 2.2762386887172693e-06, + "loss": 0.4574, + "step": 31372 + }, + { + "epoch": 0.93, + "learning_rate": 2.274193528511892e-06, + "loss": 0.5207, + "step": 31373 + }, + { + "epoch": 0.93, + "learning_rate": 2.2721492769220977e-06, + "loss": 0.4215, + "step": 31374 + }, + { + "epoch": 0.93, + "learning_rate": 2.270105933966915e-06, + "loss": 0.501, + "step": 31375 + }, + { + "epoch": 0.93, + "learning_rate": 2.2680634996653403e-06, + "loss": 0.6543, + "step": 31376 + }, + { + "epoch": 0.93, + "learning_rate": 2.2660219740363588e-06, + "loss": 0.6523, + "step": 31377 + }, + { + "epoch": 0.93, + "learning_rate": 2.263981357098943e-06, + "loss": 0.5917, + "step": 31378 + }, + { + "epoch": 0.93, + "learning_rate": 2.261941648872079e-06, + "loss": 0.6331, + "step": 31379 + }, + { + "epoch": 0.93, + "learning_rate": 2.259902849374729e-06, + "loss": 0.5546, + "step": 31380 + }, + { + "epoch": 0.93, + "learning_rate": 2.2578649586258436e-06, + "loss": 0.542, + "step": 31381 + }, + { + "epoch": 0.93, + "learning_rate": 2.2558279766443756e-06, + "loss": 0.4862, + "step": 31382 + }, + { + "epoch": 0.93, + "learning_rate": 2.253791903449265e-06, + "loss": 0.6296, + "step": 31383 + }, + { + "epoch": 0.93, + "learning_rate": 2.2517567390594296e-06, + "loss": 0.5007, + "step": 31384 + }, + { + "epoch": 0.93, + "learning_rate": 2.2497224834938102e-06, + "loss": 0.4981, + "step": 31385 + }, + { + "epoch": 0.93, + "learning_rate": 2.2476891367713028e-06, + "loss": 0.535, + "step": 31386 + }, + { + "epoch": 0.93, + "learning_rate": 2.2456566989108473e-06, + "loss": 0.5005, + "step": 31387 + }, + { + "epoch": 0.93, + "learning_rate": 2.243625169931285e-06, + "loss": 0.5411, + "step": 31388 + }, + { + "epoch": 0.93, + "learning_rate": 2.2415945498515445e-06, + "loss": 0.5548, + "step": 31389 + }, + { + "epoch": 0.93, + "learning_rate": 2.2395648386905e-06, + "loss": 0.5472, + "step": 31390 + }, + { + "epoch": 0.93, + "learning_rate": 2.2375360364670138e-06, + "loss": 0.4916, + "step": 31391 + }, + { + "epoch": 0.93, + "learning_rate": 2.2355081431999604e-06, + "loss": 0.4686, + "step": 31392 + }, + { + "epoch": 0.93, + "learning_rate": 2.2334811589081683e-06, + "loss": 0.496, + "step": 31393 + }, + { + "epoch": 0.93, + "learning_rate": 2.231455083610523e-06, + "loss": 0.6149, + "step": 31394 + }, + { + "epoch": 0.93, + "learning_rate": 2.2294299173258313e-06, + "loss": 0.5816, + "step": 31395 + }, + { + "epoch": 0.93, + "learning_rate": 2.2274056600729345e-06, + "loss": 0.5783, + "step": 31396 + }, + { + "epoch": 0.93, + "learning_rate": 2.2253823118706497e-06, + "loss": 0.6517, + "step": 31397 + }, + { + "epoch": 0.93, + "learning_rate": 2.2233598727377956e-06, + "loss": 0.4719, + "step": 31398 + }, + { + "epoch": 0.93, + "learning_rate": 2.2213383426931688e-06, + "loss": 0.5254, + "step": 31399 + }, + { + "epoch": 0.93, + "learning_rate": 2.219317721755565e-06, + "loss": 0.5688, + "step": 31400 + }, + { + "epoch": 0.93, + "learning_rate": 2.21729800994378e-06, + "loss": 0.5765, + "step": 31401 + }, + { + "epoch": 0.93, + "learning_rate": 2.2152792072765883e-06, + "loss": 0.5834, + "step": 31402 + }, + { + "epoch": 0.93, + "learning_rate": 2.2132613137727408e-06, + "loss": 0.5271, + "step": 31403 + }, + { + "epoch": 0.93, + "learning_rate": 2.2112443294510233e-06, + "loss": 0.362, + "step": 31404 + }, + { + "epoch": 0.93, + "learning_rate": 2.2092282543301867e-06, + "loss": 0.4342, + "step": 31405 + }, + { + "epoch": 0.94, + "learning_rate": 2.207213088428972e-06, + "loss": 0.5758, + "step": 31406 + }, + { + "epoch": 0.94, + "learning_rate": 2.205198831766109e-06, + "loss": 0.5692, + "step": 31407 + }, + { + "epoch": 0.94, + "learning_rate": 2.2031854843603263e-06, + "loss": 0.4737, + "step": 31408 + }, + { + "epoch": 0.94, + "learning_rate": 2.2011730462303426e-06, + "loss": 0.4442, + "step": 31409 + }, + { + "epoch": 0.94, + "learning_rate": 2.1991615173948877e-06, + "loss": 0.5364, + "step": 31410 + }, + { + "epoch": 0.94, + "learning_rate": 2.1971508978726464e-06, + "loss": 0.5191, + "step": 31411 + }, + { + "epoch": 0.94, + "learning_rate": 2.1951411876823147e-06, + "loss": 0.5634, + "step": 31412 + }, + { + "epoch": 0.94, + "learning_rate": 2.1931323868425667e-06, + "loss": 0.5274, + "step": 31413 + }, + { + "epoch": 0.94, + "learning_rate": 2.19112449537211e-06, + "loss": 0.5618, + "step": 31414 + }, + { + "epoch": 0.94, + "learning_rate": 2.1891175132895847e-06, + "loss": 0.5615, + "step": 31415 + }, + { + "epoch": 0.94, + "learning_rate": 2.187111440613665e-06, + "loss": 0.4419, + "step": 31416 + }, + { + "epoch": 0.94, + "learning_rate": 2.1851062773629916e-06, + "loss": 0.4553, + "step": 31417 + }, + { + "epoch": 0.94, + "learning_rate": 2.183102023556227e-06, + "loss": 0.4248, + "step": 31418 + }, + { + "epoch": 0.94, + "learning_rate": 2.181098679211979e-06, + "loss": 0.5534, + "step": 31419 + }, + { + "epoch": 0.94, + "learning_rate": 2.1790962443489106e-06, + "loss": 0.4638, + "step": 31420 + }, + { + "epoch": 0.94, + "learning_rate": 2.1770947189856063e-06, + "loss": 0.4426, + "step": 31421 + }, + { + "epoch": 0.94, + "learning_rate": 2.1750941031406846e-06, + "loss": 0.5691, + "step": 31422 + }, + { + "epoch": 0.94, + "learning_rate": 2.1730943968327422e-06, + "loss": 0.586, + "step": 31423 + }, + { + "epoch": 0.94, + "learning_rate": 2.171095600080375e-06, + "loss": 0.538, + "step": 31424 + }, + { + "epoch": 0.94, + "learning_rate": 2.1690977129021908e-06, + "loss": 0.4867, + "step": 31425 + }, + { + "epoch": 0.94, + "learning_rate": 2.1671007353167405e-06, + "loss": 0.5227, + "step": 31426 + }, + { + "epoch": 0.94, + "learning_rate": 2.165104667342577e-06, + "loss": 0.5415, + "step": 31427 + }, + { + "epoch": 0.94, + "learning_rate": 2.1631095089982845e-06, + "loss": 0.4697, + "step": 31428 + }, + { + "epoch": 0.94, + "learning_rate": 2.1611152603024156e-06, + "loss": 0.499, + "step": 31429 + }, + { + "epoch": 0.94, + "learning_rate": 2.1591219212734993e-06, + "loss": 0.5913, + "step": 31430 + }, + { + "epoch": 0.94, + "learning_rate": 2.1571294919300655e-06, + "loss": 0.5592, + "step": 31431 + }, + { + "epoch": 0.94, + "learning_rate": 2.155137972290644e-06, + "loss": 0.4994, + "step": 31432 + }, + { + "epoch": 0.94, + "learning_rate": 2.153147362373764e-06, + "loss": 0.5459, + "step": 31433 + }, + { + "epoch": 0.94, + "learning_rate": 2.151157662197911e-06, + "loss": 0.6018, + "step": 31434 + }, + { + "epoch": 0.94, + "learning_rate": 2.149168871781604e-06, + "loss": 0.4495, + "step": 31435 + }, + { + "epoch": 0.94, + "learning_rate": 2.147180991143316e-06, + "loss": 0.5212, + "step": 31436 + }, + { + "epoch": 0.94, + "learning_rate": 2.1451940203015553e-06, + "loss": 0.5535, + "step": 31437 + }, + { + "epoch": 0.94, + "learning_rate": 2.143207959274762e-06, + "loss": 0.5088, + "step": 31438 + }, + { + "epoch": 0.94, + "learning_rate": 2.1412228080814334e-06, + "loss": 0.4687, + "step": 31439 + }, + { + "epoch": 0.94, + "learning_rate": 2.139238566739998e-06, + "loss": 0.5475, + "step": 31440 + }, + { + "epoch": 0.94, + "learning_rate": 2.137255235268931e-06, + "loss": 0.5035, + "step": 31441 + }, + { + "epoch": 0.94, + "learning_rate": 2.1352728136866507e-06, + "loss": 0.5296, + "step": 31442 + }, + { + "epoch": 0.94, + "learning_rate": 2.1332913020115975e-06, + "loss": 0.5668, + "step": 31443 + }, + { + "epoch": 0.94, + "learning_rate": 2.1313107002622013e-06, + "loss": 0.4916, + "step": 31444 + }, + { + "epoch": 0.94, + "learning_rate": 2.1293310084568806e-06, + "loss": 0.5913, + "step": 31445 + }, + { + "epoch": 0.94, + "learning_rate": 2.1273522266140213e-06, + "loss": 0.5959, + "step": 31446 + }, + { + "epoch": 0.94, + "learning_rate": 2.12537435475203e-06, + "loss": 0.5862, + "step": 31447 + }, + { + "epoch": 0.94, + "learning_rate": 2.1233973928893034e-06, + "loss": 0.4356, + "step": 31448 + }, + { + "epoch": 0.94, + "learning_rate": 2.1214213410442273e-06, + "loss": 0.4488, + "step": 31449 + }, + { + "epoch": 0.94, + "learning_rate": 2.1194461992351533e-06, + "loss": 0.5274, + "step": 31450 + }, + { + "epoch": 0.94, + "learning_rate": 2.1174719674804554e-06, + "loss": 0.402, + "step": 31451 + }, + { + "epoch": 0.94, + "learning_rate": 2.115498645798497e-06, + "loss": 0.4953, + "step": 31452 + }, + { + "epoch": 0.94, + "learning_rate": 2.113526234207619e-06, + "loss": 0.5343, + "step": 31453 + }, + { + "epoch": 0.94, + "learning_rate": 2.111554732726162e-06, + "loss": 0.6411, + "step": 31454 + }, + { + "epoch": 0.94, + "learning_rate": 2.109584141372456e-06, + "loss": 0.4658, + "step": 31455 + }, + { + "epoch": 0.94, + "learning_rate": 2.1076144601648194e-06, + "loss": 0.6311, + "step": 31456 + }, + { + "epoch": 0.94, + "learning_rate": 2.1056456891215605e-06, + "loss": 0.4837, + "step": 31457 + }, + { + "epoch": 0.94, + "learning_rate": 2.103677828260997e-06, + "loss": 0.5064, + "step": 31458 + }, + { + "epoch": 0.94, + "learning_rate": 2.101710877601426e-06, + "loss": 0.4899, + "step": 31459 + }, + { + "epoch": 0.94, + "learning_rate": 2.0997448371611217e-06, + "loss": 0.4711, + "step": 31460 + }, + { + "epoch": 0.94, + "learning_rate": 2.0977797069583694e-06, + "loss": 0.4903, + "step": 31461 + }, + { + "epoch": 0.94, + "learning_rate": 2.0958154870114433e-06, + "loss": 0.493, + "step": 31462 + }, + { + "epoch": 0.94, + "learning_rate": 2.0938521773386065e-06, + "loss": 0.4674, + "step": 31463 + }, + { + "epoch": 0.94, + "learning_rate": 2.0918897779581115e-06, + "loss": 0.46, + "step": 31464 + }, + { + "epoch": 0.94, + "learning_rate": 2.089928288888199e-06, + "loss": 0.447, + "step": 31465 + }, + { + "epoch": 0.94, + "learning_rate": 2.0879677101470984e-06, + "loss": 0.5317, + "step": 31466 + }, + { + "epoch": 0.94, + "learning_rate": 2.0860080417530625e-06, + "loss": 0.5226, + "step": 31467 + }, + { + "epoch": 0.94, + "learning_rate": 2.0840492837242987e-06, + "loss": 0.5069, + "step": 31468 + }, + { + "epoch": 0.94, + "learning_rate": 2.0820914360790254e-06, + "loss": 0.5443, + "step": 31469 + }, + { + "epoch": 0.94, + "learning_rate": 2.080134498835429e-06, + "loss": 0.5139, + "step": 31470 + }, + { + "epoch": 0.94, + "learning_rate": 2.078178472011716e-06, + "loss": 0.4979, + "step": 31471 + }, + { + "epoch": 0.94, + "learning_rate": 2.0762233556260834e-06, + "loss": 0.5315, + "step": 31472 + }, + { + "epoch": 0.94, + "learning_rate": 2.0742691496966948e-06, + "loss": 0.4792, + "step": 31473 + }, + { + "epoch": 0.94, + "learning_rate": 2.0723158542417133e-06, + "loss": 0.5925, + "step": 31474 + }, + { + "epoch": 0.94, + "learning_rate": 2.070363469279313e-06, + "loss": 0.5274, + "step": 31475 + }, + { + "epoch": 0.94, + "learning_rate": 2.068411994827646e-06, + "loss": 0.5894, + "step": 31476 + }, + { + "epoch": 0.94, + "learning_rate": 2.0664614309048426e-06, + "loss": 0.5656, + "step": 31477 + }, + { + "epoch": 0.94, + "learning_rate": 2.0645117775290544e-06, + "loss": 0.5588, + "step": 31478 + }, + { + "epoch": 0.94, + "learning_rate": 2.062563034718401e-06, + "loss": 0.4523, + "step": 31479 + }, + { + "epoch": 0.94, + "learning_rate": 2.060615202491012e-06, + "loss": 0.578, + "step": 31480 + }, + { + "epoch": 0.94, + "learning_rate": 2.0586682808649727e-06, + "loss": 0.4978, + "step": 31481 + }, + { + "epoch": 0.94, + "learning_rate": 2.056722269858413e-06, + "loss": 0.4712, + "step": 31482 + }, + { + "epoch": 0.94, + "learning_rate": 2.054777169489408e-06, + "loss": 0.5095, + "step": 31483 + }, + { + "epoch": 0.94, + "learning_rate": 2.0528329797760426e-06, + "loss": 0.477, + "step": 31484 + }, + { + "epoch": 0.94, + "learning_rate": 2.0508897007364024e-06, + "loss": 0.4517, + "step": 31485 + }, + { + "epoch": 0.94, + "learning_rate": 2.048947332388551e-06, + "loss": 0.3926, + "step": 31486 + }, + { + "epoch": 0.94, + "learning_rate": 2.0470058747505516e-06, + "loss": 0.5191, + "step": 31487 + }, + { + "epoch": 0.94, + "learning_rate": 2.0450653278404564e-06, + "loss": 0.4856, + "step": 31488 + }, + { + "epoch": 0.94, + "learning_rate": 2.0431256916762843e-06, + "loss": 0.5134, + "step": 31489 + }, + { + "epoch": 0.94, + "learning_rate": 2.0411869662760984e-06, + "loss": 0.5714, + "step": 31490 + }, + { + "epoch": 0.94, + "learning_rate": 2.039249151657918e-06, + "loss": 0.6017, + "step": 31491 + }, + { + "epoch": 0.94, + "learning_rate": 2.037312247839762e-06, + "loss": 0.5945, + "step": 31492 + }, + { + "epoch": 0.94, + "learning_rate": 2.035376254839616e-06, + "loss": 0.5143, + "step": 31493 + }, + { + "epoch": 0.94, + "learning_rate": 2.0334411726754988e-06, + "loss": 0.4512, + "step": 31494 + }, + { + "epoch": 0.94, + "learning_rate": 2.031507001365407e-06, + "loss": 0.3879, + "step": 31495 + }, + { + "epoch": 0.94, + "learning_rate": 2.029573740927304e-06, + "loss": 0.5925, + "step": 31496 + }, + { + "epoch": 0.94, + "learning_rate": 2.027641391379198e-06, + "loss": 0.4974, + "step": 31497 + }, + { + "epoch": 0.94, + "learning_rate": 2.025709952739019e-06, + "loss": 0.4965, + "step": 31498 + }, + { + "epoch": 0.94, + "learning_rate": 2.023779425024741e-06, + "loss": 0.4477, + "step": 31499 + }, + { + "epoch": 0.94, + "learning_rate": 2.021849808254317e-06, + "loss": 0.6648, + "step": 31500 + }, + { + "epoch": 0.94, + "learning_rate": 2.0199211024456876e-06, + "loss": 0.4147, + "step": 31501 + }, + { + "epoch": 0.94, + "learning_rate": 2.0179933076167725e-06, + "loss": 0.5606, + "step": 31502 + }, + { + "epoch": 0.94, + "learning_rate": 2.016066423785512e-06, + "loss": 0.5484, + "step": 31503 + }, + { + "epoch": 0.94, + "learning_rate": 2.0141404509698035e-06, + "loss": 0.4043, + "step": 31504 + }, + { + "epoch": 0.94, + "learning_rate": 2.0122153891875663e-06, + "loss": 0.5113, + "step": 31505 + }, + { + "epoch": 0.94, + "learning_rate": 2.0102912384566963e-06, + "loss": 0.5639, + "step": 31506 + }, + { + "epoch": 0.94, + "learning_rate": 2.008367998795091e-06, + "loss": 0.5262, + "step": 31507 + }, + { + "epoch": 0.94, + "learning_rate": 2.006445670220614e-06, + "loss": 0.5618, + "step": 31508 + }, + { + "epoch": 0.94, + "learning_rate": 2.00452425275115e-06, + "loss": 0.5664, + "step": 31509 + }, + { + "epoch": 0.94, + "learning_rate": 2.002603746404563e-06, + "loss": 0.6258, + "step": 31510 + }, + { + "epoch": 0.94, + "learning_rate": 2.0006841511987283e-06, + "loss": 0.435, + "step": 31511 + }, + { + "epoch": 0.94, + "learning_rate": 1.9987654671514533e-06, + "loss": 0.5705, + "step": 31512 + }, + { + "epoch": 0.94, + "learning_rate": 1.9968476942805903e-06, + "loss": 0.438, + "step": 31513 + }, + { + "epoch": 0.94, + "learning_rate": 1.994930832603992e-06, + "loss": 0.5408, + "step": 31514 + }, + { + "epoch": 0.94, + "learning_rate": 1.9930148821394655e-06, + "loss": 0.5738, + "step": 31515 + }, + { + "epoch": 0.94, + "learning_rate": 1.99109984290482e-06, + "loss": 0.5901, + "step": 31516 + }, + { + "epoch": 0.94, + "learning_rate": 1.9891857149178627e-06, + "loss": 0.5373, + "step": 31517 + }, + { + "epoch": 0.94, + "learning_rate": 1.987272498196402e-06, + "loss": 0.6622, + "step": 31518 + }, + { + "epoch": 0.94, + "learning_rate": 1.9853601927582123e-06, + "loss": 0.4909, + "step": 31519 + }, + { + "epoch": 0.94, + "learning_rate": 1.9834487986210794e-06, + "loss": 0.6274, + "step": 31520 + }, + { + "epoch": 0.94, + "learning_rate": 1.9815383158027778e-06, + "loss": 0.567, + "step": 31521 + }, + { + "epoch": 0.94, + "learning_rate": 1.9796287443210604e-06, + "loss": 0.458, + "step": 31522 + }, + { + "epoch": 0.94, + "learning_rate": 1.9777200841936903e-06, + "loss": 0.5935, + "step": 31523 + }, + { + "epoch": 0.94, + "learning_rate": 1.9758123354384096e-06, + "loss": 0.4802, + "step": 31524 + }, + { + "epoch": 0.94, + "learning_rate": 1.973905498072959e-06, + "loss": 0.4783, + "step": 31525 + }, + { + "epoch": 0.94, + "learning_rate": 1.9719995721150686e-06, + "loss": 0.4667, + "step": 31526 + }, + { + "epoch": 0.94, + "learning_rate": 1.970094557582458e-06, + "loss": 0.4555, + "step": 31527 + }, + { + "epoch": 0.94, + "learning_rate": 1.968190454492824e-06, + "loss": 0.4143, + "step": 31528 + }, + { + "epoch": 0.94, + "learning_rate": 1.9662872628638862e-06, + "loss": 0.5048, + "step": 31529 + }, + { + "epoch": 0.94, + "learning_rate": 1.9643849827133522e-06, + "loss": 0.5442, + "step": 31530 + }, + { + "epoch": 0.94, + "learning_rate": 1.9624836140588852e-06, + "loss": 0.5537, + "step": 31531 + }, + { + "epoch": 0.94, + "learning_rate": 1.960583156918172e-06, + "loss": 0.6056, + "step": 31532 + }, + { + "epoch": 0.94, + "learning_rate": 1.9586836113088756e-06, + "loss": 0.5314, + "step": 31533 + }, + { + "epoch": 0.94, + "learning_rate": 1.956784977248671e-06, + "loss": 0.5366, + "step": 31534 + }, + { + "epoch": 0.94, + "learning_rate": 1.9548872547552e-06, + "loss": 0.384, + "step": 31535 + }, + { + "epoch": 0.94, + "learning_rate": 1.9529904438461143e-06, + "loss": 0.458, + "step": 31536 + }, + { + "epoch": 0.94, + "learning_rate": 1.951094544539034e-06, + "loss": 0.5493, + "step": 31537 + }, + { + "epoch": 0.94, + "learning_rate": 1.9491995568516107e-06, + "loss": 0.6052, + "step": 31538 + }, + { + "epoch": 0.94, + "learning_rate": 1.947305480801442e-06, + "loss": 0.4735, + "step": 31539 + }, + { + "epoch": 0.94, + "learning_rate": 1.945412316406159e-06, + "loss": 0.5183, + "step": 31540 + }, + { + "epoch": 0.94, + "learning_rate": 1.943520063683346e-06, + "loss": 0.6206, + "step": 31541 + }, + { + "epoch": 0.94, + "learning_rate": 1.9416287226506013e-06, + "loss": 0.5159, + "step": 31542 + }, + { + "epoch": 0.94, + "learning_rate": 1.9397382933255102e-06, + "loss": 0.4767, + "step": 31543 + }, + { + "epoch": 0.94, + "learning_rate": 1.937848775725648e-06, + "loss": 0.463, + "step": 31544 + }, + { + "epoch": 0.94, + "learning_rate": 1.935960169868589e-06, + "loss": 0.5417, + "step": 31545 + }, + { + "epoch": 0.94, + "learning_rate": 1.9340724757718863e-06, + "loss": 0.6275, + "step": 31546 + }, + { + "epoch": 0.94, + "learning_rate": 1.9321856934530813e-06, + "loss": 0.6072, + "step": 31547 + }, + { + "epoch": 0.94, + "learning_rate": 1.9302998229297376e-06, + "loss": 0.4801, + "step": 31548 + }, + { + "epoch": 0.94, + "learning_rate": 1.9284148642193855e-06, + "loss": 0.652, + "step": 31549 + }, + { + "epoch": 0.94, + "learning_rate": 1.9265308173395336e-06, + "loss": 0.5465, + "step": 31550 + }, + { + "epoch": 0.94, + "learning_rate": 1.924647682307712e-06, + "loss": 0.4562, + "step": 31551 + }, + { + "epoch": 0.94, + "learning_rate": 1.922765459141429e-06, + "loss": 0.533, + "step": 31552 + }, + { + "epoch": 0.94, + "learning_rate": 1.920884147858182e-06, + "loss": 0.4613, + "step": 31553 + }, + { + "epoch": 0.94, + "learning_rate": 1.9190037484754788e-06, + "loss": 0.4385, + "step": 31554 + }, + { + "epoch": 0.94, + "learning_rate": 1.9171242610107608e-06, + "loss": 0.4324, + "step": 31555 + }, + { + "epoch": 0.94, + "learning_rate": 1.9152456854815483e-06, + "loss": 0.4583, + "step": 31556 + }, + { + "epoch": 0.94, + "learning_rate": 1.913368021905282e-06, + "loss": 0.4413, + "step": 31557 + }, + { + "epoch": 0.94, + "learning_rate": 1.911491270299426e-06, + "loss": 0.5355, + "step": 31558 + }, + { + "epoch": 0.94, + "learning_rate": 1.909615430681433e-06, + "loss": 0.5031, + "step": 31559 + }, + { + "epoch": 0.94, + "learning_rate": 1.9077405030687334e-06, + "loss": 0.6635, + "step": 31560 + }, + { + "epoch": 0.94, + "learning_rate": 1.9058664874787802e-06, + "loss": 0.5055, + "step": 31561 + }, + { + "epoch": 0.94, + "learning_rate": 1.9039933839289704e-06, + "loss": 0.5902, + "step": 31562 + }, + { + "epoch": 0.94, + "learning_rate": 1.9021211924367344e-06, + "loss": 0.5345, + "step": 31563 + }, + { + "epoch": 0.94, + "learning_rate": 1.9002499130194916e-06, + "loss": 0.436, + "step": 31564 + }, + { + "epoch": 0.94, + "learning_rate": 1.8983795456946173e-06, + "loss": 0.5427, + "step": 31565 + }, + { + "epoch": 0.94, + "learning_rate": 1.8965100904795085e-06, + "loss": 0.4564, + "step": 31566 + }, + { + "epoch": 0.94, + "learning_rate": 1.8946415473915402e-06, + "loss": 0.4955, + "step": 31567 + }, + { + "epoch": 0.94, + "learning_rate": 1.8927739164481096e-06, + "loss": 0.5664, + "step": 31568 + }, + { + "epoch": 0.94, + "learning_rate": 1.8909071976665693e-06, + "loss": 0.5445, + "step": 31569 + }, + { + "epoch": 0.94, + "learning_rate": 1.8890413910642501e-06, + "loss": 0.442, + "step": 31570 + }, + { + "epoch": 0.94, + "learning_rate": 1.887176496658538e-06, + "loss": 0.5378, + "step": 31571 + }, + { + "epoch": 0.94, + "learning_rate": 1.8853125144667417e-06, + "loss": 0.5785, + "step": 31572 + }, + { + "epoch": 0.94, + "learning_rate": 1.8834494445062134e-06, + "loss": 0.6067, + "step": 31573 + }, + { + "epoch": 0.94, + "learning_rate": 1.881587286794273e-06, + "loss": 0.5498, + "step": 31574 + }, + { + "epoch": 0.94, + "learning_rate": 1.8797260413482177e-06, + "loss": 0.5516, + "step": 31575 + }, + { + "epoch": 0.94, + "learning_rate": 1.8778657081853668e-06, + "loss": 0.5491, + "step": 31576 + }, + { + "epoch": 0.94, + "learning_rate": 1.8760062873230066e-06, + "loss": 0.4957, + "step": 31577 + }, + { + "epoch": 0.94, + "learning_rate": 1.8741477787784456e-06, + "loss": 0.5826, + "step": 31578 + }, + { + "epoch": 0.94, + "learning_rate": 1.8722901825689366e-06, + "loss": 0.6137, + "step": 31579 + }, + { + "epoch": 0.94, + "learning_rate": 1.8704334987117655e-06, + "loss": 0.3891, + "step": 31580 + }, + { + "epoch": 0.94, + "learning_rate": 1.8685777272241967e-06, + "loss": 0.4839, + "step": 31581 + }, + { + "epoch": 0.94, + "learning_rate": 1.8667228681234828e-06, + "loss": 0.5439, + "step": 31582 + }, + { + "epoch": 0.94, + "learning_rate": 1.8648689214268656e-06, + "loss": 0.5463, + "step": 31583 + }, + { + "epoch": 0.94, + "learning_rate": 1.8630158871515868e-06, + "loss": 0.4603, + "step": 31584 + }, + { + "epoch": 0.94, + "learning_rate": 1.861163765314866e-06, + "loss": 0.4661, + "step": 31585 + }, + { + "epoch": 0.94, + "learning_rate": 1.8593125559339341e-06, + "loss": 0.444, + "step": 31586 + }, + { + "epoch": 0.94, + "learning_rate": 1.8574622590259993e-06, + "loss": 0.5275, + "step": 31587 + }, + { + "epoch": 0.94, + "learning_rate": 1.8556128746082702e-06, + "loss": 0.6096, + "step": 31588 + }, + { + "epoch": 0.94, + "learning_rate": 1.8537644026979217e-06, + "loss": 0.3835, + "step": 31589 + }, + { + "epoch": 0.94, + "learning_rate": 1.8519168433121626e-06, + "loss": 0.5012, + "step": 31590 + }, + { + "epoch": 0.94, + "learning_rate": 1.8500701964681566e-06, + "loss": 0.5694, + "step": 31591 + }, + { + "epoch": 0.94, + "learning_rate": 1.8482244621830901e-06, + "loss": 0.4797, + "step": 31592 + }, + { + "epoch": 0.94, + "learning_rate": 1.8463796404741162e-06, + "loss": 0.5929, + "step": 31593 + }, + { + "epoch": 0.94, + "learning_rate": 1.8445357313583654e-06, + "loss": 0.4916, + "step": 31594 + }, + { + "epoch": 0.94, + "learning_rate": 1.8426927348530133e-06, + "loss": 0.4317, + "step": 31595 + }, + { + "epoch": 0.94, + "learning_rate": 1.8408506509751788e-06, + "loss": 0.5056, + "step": 31596 + }, + { + "epoch": 0.94, + "learning_rate": 1.8390094797420043e-06, + "loss": 0.5697, + "step": 31597 + }, + { + "epoch": 0.94, + "learning_rate": 1.8371692211705872e-06, + "loss": 0.5831, + "step": 31598 + }, + { + "epoch": 0.94, + "learning_rate": 1.8353298752780357e-06, + "loss": 0.4554, + "step": 31599 + }, + { + "epoch": 0.94, + "learning_rate": 1.8334914420814807e-06, + "loss": 0.5405, + "step": 31600 + }, + { + "epoch": 0.94, + "learning_rate": 1.8316539215979866e-06, + "loss": 0.6039, + "step": 31601 + }, + { + "epoch": 0.94, + "learning_rate": 1.8298173138446505e-06, + "loss": 0.5535, + "step": 31602 + }, + { + "epoch": 0.94, + "learning_rate": 1.827981618838548e-06, + "loss": 0.4841, + "step": 31603 + }, + { + "epoch": 0.94, + "learning_rate": 1.8261468365967426e-06, + "loss": 0.552, + "step": 31604 + }, + { + "epoch": 0.94, + "learning_rate": 1.824312967136299e-06, + "loss": 0.5846, + "step": 31605 + }, + { + "epoch": 0.94, + "learning_rate": 1.822480010474259e-06, + "loss": 0.4996, + "step": 31606 + }, + { + "epoch": 0.94, + "learning_rate": 1.8206479666276755e-06, + "loss": 0.4238, + "step": 31607 + }, + { + "epoch": 0.94, + "learning_rate": 1.8188168356135793e-06, + "loss": 0.4643, + "step": 31608 + }, + { + "epoch": 0.94, + "learning_rate": 1.8169866174489791e-06, + "loss": 0.5004, + "step": 31609 + }, + { + "epoch": 0.94, + "learning_rate": 1.8151573121509169e-06, + "loss": 0.524, + "step": 31610 + }, + { + "epoch": 0.94, + "learning_rate": 1.8133289197363902e-06, + "loss": 0.4983, + "step": 31611 + }, + { + "epoch": 0.94, + "learning_rate": 1.8115014402223962e-06, + "loss": 0.4887, + "step": 31612 + }, + { + "epoch": 0.94, + "learning_rate": 1.8096748736259218e-06, + "loss": 0.5172, + "step": 31613 + }, + { + "epoch": 0.94, + "learning_rate": 1.8078492199639642e-06, + "loss": 0.4426, + "step": 31614 + }, + { + "epoch": 0.94, + "learning_rate": 1.8060244792534764e-06, + "loss": 0.6361, + "step": 31615 + }, + { + "epoch": 0.94, + "learning_rate": 1.8042006515114562e-06, + "loss": 0.5021, + "step": 31616 + }, + { + "epoch": 0.94, + "learning_rate": 1.8023777367548345e-06, + "loss": 0.5914, + "step": 31617 + }, + { + "epoch": 0.94, + "learning_rate": 1.8005557350005531e-06, + "loss": 0.5623, + "step": 31618 + }, + { + "epoch": 0.94, + "learning_rate": 1.7987346462655763e-06, + "loss": 0.5563, + "step": 31619 + }, + { + "epoch": 0.94, + "learning_rate": 1.796914470566824e-06, + "loss": 0.5571, + "step": 31620 + }, + { + "epoch": 0.94, + "learning_rate": 1.795095207921227e-06, + "loss": 0.4977, + "step": 31621 + }, + { + "epoch": 0.94, + "learning_rate": 1.7932768583456828e-06, + "loss": 0.5281, + "step": 31622 + }, + { + "epoch": 0.94, + "learning_rate": 1.7914594218571224e-06, + "loss": 0.4901, + "step": 31623 + }, + { + "epoch": 0.94, + "learning_rate": 1.7896428984724211e-06, + "loss": 0.5042, + "step": 31624 + }, + { + "epoch": 0.94, + "learning_rate": 1.7878272882084768e-06, + "loss": 0.5934, + "step": 31625 + }, + { + "epoch": 0.94, + "learning_rate": 1.7860125910821757e-06, + "loss": 0.5525, + "step": 31626 + }, + { + "epoch": 0.94, + "learning_rate": 1.7841988071103932e-06, + "loss": 0.4763, + "step": 31627 + }, + { + "epoch": 0.94, + "learning_rate": 1.7823859363099716e-06, + "loss": 0.4617, + "step": 31628 + }, + { + "epoch": 0.94, + "learning_rate": 1.780573978697775e-06, + "loss": 0.4839, + "step": 31629 + }, + { + "epoch": 0.94, + "learning_rate": 1.7787629342906674e-06, + "loss": 0.4766, + "step": 31630 + }, + { + "epoch": 0.94, + "learning_rate": 1.7769528031054805e-06, + "loss": 0.5602, + "step": 31631 + }, + { + "epoch": 0.94, + "learning_rate": 1.7751435851590226e-06, + "loss": 0.5684, + "step": 31632 + }, + { + "epoch": 0.94, + "learning_rate": 1.7733352804681357e-06, + "loss": 0.6045, + "step": 31633 + }, + { + "epoch": 0.94, + "learning_rate": 1.7715278890496289e-06, + "loss": 0.6232, + "step": 31634 + }, + { + "epoch": 0.94, + "learning_rate": 1.7697214109203107e-06, + "loss": 0.5539, + "step": 31635 + }, + { + "epoch": 0.94, + "learning_rate": 1.767915846096968e-06, + "loss": 0.5624, + "step": 31636 + }, + { + "epoch": 0.94, + "learning_rate": 1.766111194596387e-06, + "loss": 0.5083, + "step": 31637 + }, + { + "epoch": 0.94, + "learning_rate": 1.7643074564353545e-06, + "loss": 0.5784, + "step": 31638 + }, + { + "epoch": 0.94, + "learning_rate": 1.762504631630646e-06, + "loss": 0.526, + "step": 31639 + }, + { + "epoch": 0.94, + "learning_rate": 1.7607027201990257e-06, + "loss": 0.5399, + "step": 31640 + }, + { + "epoch": 0.94, + "learning_rate": 1.7589017221572135e-06, + "loss": 0.5671, + "step": 31641 + }, + { + "epoch": 0.94, + "learning_rate": 1.757101637521985e-06, + "loss": 0.6087, + "step": 31642 + }, + { + "epoch": 0.94, + "learning_rate": 1.7553024663100714e-06, + "loss": 0.5128, + "step": 31643 + }, + { + "epoch": 0.94, + "learning_rate": 1.7535042085382037e-06, + "loss": 0.531, + "step": 31644 + }, + { + "epoch": 0.94, + "learning_rate": 1.7517068642230905e-06, + "loss": 0.6278, + "step": 31645 + }, + { + "epoch": 0.94, + "learning_rate": 1.749910433381452e-06, + "loss": 0.4874, + "step": 31646 + }, + { + "epoch": 0.94, + "learning_rate": 1.748114916029997e-06, + "loss": 0.4799, + "step": 31647 + }, + { + "epoch": 0.94, + "learning_rate": 1.74632031218539e-06, + "loss": 0.6246, + "step": 31648 + }, + { + "epoch": 0.94, + "learning_rate": 1.7445266218643508e-06, + "loss": 0.4631, + "step": 31649 + }, + { + "epoch": 0.94, + "learning_rate": 1.742733845083544e-06, + "loss": 0.6282, + "step": 31650 + }, + { + "epoch": 0.94, + "learning_rate": 1.7409419818596229e-06, + "loss": 0.5159, + "step": 31651 + }, + { + "epoch": 0.94, + "learning_rate": 1.7391510322092741e-06, + "loss": 0.5173, + "step": 31652 + }, + { + "epoch": 0.94, + "learning_rate": 1.7373609961491177e-06, + "loss": 0.4394, + "step": 31653 + }, + { + "epoch": 0.94, + "learning_rate": 1.7355718736958404e-06, + "loss": 0.5764, + "step": 31654 + }, + { + "epoch": 0.94, + "learning_rate": 1.73378366486604e-06, + "loss": 0.4878, + "step": 31655 + }, + { + "epoch": 0.94, + "learning_rate": 1.7319963696763475e-06, + "loss": 0.5154, + "step": 31656 + }, + { + "epoch": 0.94, + "learning_rate": 1.7302099881433942e-06, + "loss": 0.5357, + "step": 31657 + }, + { + "epoch": 0.94, + "learning_rate": 1.728424520283789e-06, + "loss": 0.551, + "step": 31658 + }, + { + "epoch": 0.94, + "learning_rate": 1.7266399661141187e-06, + "loss": 0.4714, + "step": 31659 + }, + { + "epoch": 0.94, + "learning_rate": 1.7248563256509699e-06, + "loss": 0.588, + "step": 31660 + }, + { + "epoch": 0.94, + "learning_rate": 1.7230735989109513e-06, + "loss": 0.4696, + "step": 31661 + }, + { + "epoch": 0.94, + "learning_rate": 1.721291785910628e-06, + "loss": 0.4911, + "step": 31662 + }, + { + "epoch": 0.94, + "learning_rate": 1.7195108866665532e-06, + "loss": 0.3986, + "step": 31663 + }, + { + "epoch": 0.94, + "learning_rate": 1.7177309011953023e-06, + "loss": 0.459, + "step": 31664 + }, + { + "epoch": 0.94, + "learning_rate": 1.715951829513418e-06, + "loss": 0.5248, + "step": 31665 + }, + { + "epoch": 0.94, + "learning_rate": 1.7141736716374424e-06, + "loss": 0.4108, + "step": 31666 + }, + { + "epoch": 0.94, + "learning_rate": 1.7123964275838956e-06, + "loss": 0.5225, + "step": 31667 + }, + { + "epoch": 0.94, + "learning_rate": 1.7106200973693309e-06, + "loss": 0.6449, + "step": 31668 + }, + { + "epoch": 0.94, + "learning_rate": 1.7088446810102354e-06, + "loss": 0.5134, + "step": 31669 + }, + { + "epoch": 0.94, + "learning_rate": 1.7070701785231292e-06, + "loss": 0.4918, + "step": 31670 + }, + { + "epoch": 0.94, + "learning_rate": 1.70529658992451e-06, + "loss": 0.5849, + "step": 31671 + }, + { + "epoch": 0.94, + "learning_rate": 1.7035239152308647e-06, + "loss": 0.6629, + "step": 31672 + }, + { + "epoch": 0.94, + "learning_rate": 1.70175215445868e-06, + "loss": 0.5128, + "step": 31673 + }, + { + "epoch": 0.94, + "learning_rate": 1.699981307624421e-06, + "loss": 0.5944, + "step": 31674 + }, + { + "epoch": 0.94, + "learning_rate": 1.6982113747445628e-06, + "loss": 0.5731, + "step": 31675 + }, + { + "epoch": 0.94, + "learning_rate": 1.696442355835548e-06, + "loss": 0.4933, + "step": 31676 + }, + { + "epoch": 0.94, + "learning_rate": 1.6946742509138302e-06, + "loss": 0.6148, + "step": 31677 + }, + { + "epoch": 0.94, + "learning_rate": 1.692907059995874e-06, + "loss": 0.5467, + "step": 31678 + }, + { + "epoch": 0.94, + "learning_rate": 1.6911407830980663e-06, + "loss": 0.4164, + "step": 31679 + }, + { + "epoch": 0.94, + "learning_rate": 1.6893754202368495e-06, + "loss": 0.5486, + "step": 31680 + }, + { + "epoch": 0.94, + "learning_rate": 1.6876109714286436e-06, + "loss": 0.6568, + "step": 31681 + }, + { + "epoch": 0.94, + "learning_rate": 1.6858474366898469e-06, + "loss": 0.5682, + "step": 31682 + }, + { + "epoch": 0.94, + "learning_rate": 1.6840848160368572e-06, + "loss": 0.5577, + "step": 31683 + }, + { + "epoch": 0.94, + "learning_rate": 1.6823231094860614e-06, + "loss": 0.5437, + "step": 31684 + }, + { + "epoch": 0.94, + "learning_rate": 1.6805623170538353e-06, + "loss": 0.4948, + "step": 31685 + }, + { + "epoch": 0.94, + "learning_rate": 1.6788024387565549e-06, + "loss": 0.4467, + "step": 31686 + }, + { + "epoch": 0.94, + "learning_rate": 1.6770434746105845e-06, + "loss": 0.5121, + "step": 31687 + }, + { + "epoch": 0.94, + "learning_rate": 1.675285424632278e-06, + "loss": 0.4558, + "step": 31688 + }, + { + "epoch": 0.94, + "learning_rate": 1.6735282888379888e-06, + "loss": 0.5306, + "step": 31689 + }, + { + "epoch": 0.94, + "learning_rate": 1.6717720672440262e-06, + "loss": 0.4603, + "step": 31690 + }, + { + "epoch": 0.94, + "learning_rate": 1.6700167598667438e-06, + "loss": 0.51, + "step": 31691 + }, + { + "epoch": 0.94, + "learning_rate": 1.6682623667224618e-06, + "loss": 0.5046, + "step": 31692 + }, + { + "epoch": 0.94, + "learning_rate": 1.6665088878274781e-06, + "loss": 0.4657, + "step": 31693 + }, + { + "epoch": 0.94, + "learning_rate": 1.6647563231981024e-06, + "loss": 0.5571, + "step": 31694 + }, + { + "epoch": 0.94, + "learning_rate": 1.663004672850621e-06, + "loss": 0.6095, + "step": 31695 + }, + { + "epoch": 0.94, + "learning_rate": 1.6612539368013436e-06, + "loss": 0.5807, + "step": 31696 + }, + { + "epoch": 0.94, + "learning_rate": 1.6595041150665346e-06, + "loss": 0.525, + "step": 31697 + }, + { + "epoch": 0.94, + "learning_rate": 1.6577552076624592e-06, + "loss": 0.5526, + "step": 31698 + }, + { + "epoch": 0.94, + "learning_rate": 1.6560072146053707e-06, + "loss": 0.4891, + "step": 31699 + }, + { + "epoch": 0.94, + "learning_rate": 1.654260135911534e-06, + "loss": 0.4772, + "step": 31700 + }, + { + "epoch": 0.94, + "learning_rate": 1.6525139715972027e-06, + "loss": 0.5849, + "step": 31701 + }, + { + "epoch": 0.94, + "learning_rate": 1.6507687216785861e-06, + "loss": 0.5378, + "step": 31702 + }, + { + "epoch": 0.94, + "learning_rate": 1.6490243861719267e-06, + "loss": 0.4644, + "step": 31703 + }, + { + "epoch": 0.94, + "learning_rate": 1.647280965093434e-06, + "loss": 0.5568, + "step": 31704 + }, + { + "epoch": 0.94, + "learning_rate": 1.6455384584593391e-06, + "loss": 0.5312, + "step": 31705 + }, + { + "epoch": 0.94, + "learning_rate": 1.6437968662858182e-06, + "loss": 0.4403, + "step": 31706 + }, + { + "epoch": 0.94, + "learning_rate": 1.6420561885890695e-06, + "loss": 0.4518, + "step": 31707 + }, + { + "epoch": 0.94, + "learning_rate": 1.6403164253852799e-06, + "loss": 0.5534, + "step": 31708 + }, + { + "epoch": 0.94, + "learning_rate": 1.6385775766906364e-06, + "loss": 0.5439, + "step": 31709 + }, + { + "epoch": 0.94, + "learning_rate": 1.6368396425212928e-06, + "loss": 0.5245, + "step": 31710 + }, + { + "epoch": 0.94, + "learning_rate": 1.6351026228934141e-06, + "loss": 0.5353, + "step": 31711 + }, + { + "epoch": 0.94, + "learning_rate": 1.6333665178231317e-06, + "loss": 0.4703, + "step": 31712 + }, + { + "epoch": 0.94, + "learning_rate": 1.6316313273266104e-06, + "loss": 0.5482, + "step": 31713 + }, + { + "epoch": 0.94, + "learning_rate": 1.6298970514199706e-06, + "loss": 0.4823, + "step": 31714 + }, + { + "epoch": 0.94, + "learning_rate": 1.6281636901193441e-06, + "loss": 0.4142, + "step": 31715 + }, + { + "epoch": 0.94, + "learning_rate": 1.6264312434408514e-06, + "loss": 0.5245, + "step": 31716 + }, + { + "epoch": 0.94, + "learning_rate": 1.6246997114006013e-06, + "loss": 0.5798, + "step": 31717 + }, + { + "epoch": 0.94, + "learning_rate": 1.622969094014659e-06, + "loss": 0.5259, + "step": 31718 + }, + { + "epoch": 0.94, + "learning_rate": 1.6212393912991565e-06, + "loss": 0.5598, + "step": 31719 + }, + { + "epoch": 0.94, + "learning_rate": 1.619510603270158e-06, + "loss": 0.4408, + "step": 31720 + }, + { + "epoch": 0.94, + "learning_rate": 1.617782729943751e-06, + "loss": 0.5275, + "step": 31721 + }, + { + "epoch": 0.94, + "learning_rate": 1.616055771335967e-06, + "loss": 0.4251, + "step": 31722 + }, + { + "epoch": 0.94, + "learning_rate": 1.6143297274628932e-06, + "loss": 0.6472, + "step": 31723 + }, + { + "epoch": 0.94, + "learning_rate": 1.6126045983405724e-06, + "loss": 0.5755, + "step": 31724 + }, + { + "epoch": 0.94, + "learning_rate": 1.6108803839850362e-06, + "loss": 0.6285, + "step": 31725 + }, + { + "epoch": 0.94, + "learning_rate": 1.6091570844123271e-06, + "loss": 0.4847, + "step": 31726 + }, + { + "epoch": 0.94, + "learning_rate": 1.607434699638455e-06, + "loss": 0.5458, + "step": 31727 + }, + { + "epoch": 0.94, + "learning_rate": 1.6057132296794398e-06, + "loss": 0.6254, + "step": 31728 + }, + { + "epoch": 0.94, + "learning_rate": 1.6039926745512913e-06, + "loss": 0.4606, + "step": 31729 + }, + { + "epoch": 0.94, + "learning_rate": 1.6022730342699855e-06, + "loss": 0.528, + "step": 31730 + }, + { + "epoch": 0.94, + "learning_rate": 1.600554308851554e-06, + "loss": 0.5818, + "step": 31731 + }, + { + "epoch": 0.94, + "learning_rate": 1.5988364983119397e-06, + "loss": 0.6, + "step": 31732 + }, + { + "epoch": 0.94, + "learning_rate": 1.5971196026671188e-06, + "loss": 0.3893, + "step": 31733 + }, + { + "epoch": 0.94, + "learning_rate": 1.5954036219330558e-06, + "loss": 0.4153, + "step": 31734 + }, + { + "epoch": 0.94, + "learning_rate": 1.5936885561257275e-06, + "loss": 0.5736, + "step": 31735 + }, + { + "epoch": 0.94, + "learning_rate": 1.591974405261054e-06, + "loss": 0.4221, + "step": 31736 + }, + { + "epoch": 0.94, + "learning_rate": 1.5902611693549785e-06, + "loss": 0.5641, + "step": 31737 + }, + { + "epoch": 0.94, + "learning_rate": 1.5885488484234213e-06, + "loss": 0.5355, + "step": 31738 + }, + { + "epoch": 0.94, + "learning_rate": 1.5868374424823252e-06, + "loss": 0.621, + "step": 31739 + }, + { + "epoch": 0.94, + "learning_rate": 1.5851269515475997e-06, + "loss": 0.5228, + "step": 31740 + }, + { + "epoch": 0.94, + "learning_rate": 1.5834173756351322e-06, + "loss": 0.482, + "step": 31741 + }, + { + "epoch": 0.95, + "learning_rate": 1.5817087147608212e-06, + "loss": 0.5434, + "step": 31742 + }, + { + "epoch": 0.95, + "learning_rate": 1.5800009689405537e-06, + "loss": 0.5345, + "step": 31743 + }, + { + "epoch": 0.95, + "learning_rate": 1.5782941381902283e-06, + "loss": 0.5404, + "step": 31744 + }, + { + "epoch": 0.95, + "learning_rate": 1.5765882225256878e-06, + "loss": 0.5603, + "step": 31745 + }, + { + "epoch": 0.95, + "learning_rate": 1.5748832219627973e-06, + "loss": 0.5839, + "step": 31746 + }, + { + "epoch": 0.95, + "learning_rate": 1.5731791365174109e-06, + "loss": 0.6554, + "step": 31747 + }, + { + "epoch": 0.95, + "learning_rate": 1.5714759662053824e-06, + "loss": 0.4769, + "step": 31748 + }, + { + "epoch": 0.95, + "learning_rate": 1.5697737110425437e-06, + "loss": 0.5119, + "step": 31749 + }, + { + "epoch": 0.95, + "learning_rate": 1.5680723710447154e-06, + "loss": 0.4986, + "step": 31750 + }, + { + "epoch": 0.95, + "learning_rate": 1.5663719462277183e-06, + "loss": 0.5437, + "step": 31751 + }, + { + "epoch": 0.95, + "learning_rate": 1.5646724366073618e-06, + "loss": 0.4797, + "step": 31752 + }, + { + "epoch": 0.95, + "learning_rate": 1.5629738421994444e-06, + "loss": 0.5378, + "step": 31753 + }, + { + "epoch": 0.95, + "learning_rate": 1.561276163019776e-06, + "loss": 0.5106, + "step": 31754 + }, + { + "epoch": 0.95, + "learning_rate": 1.5595793990841324e-06, + "loss": 0.4833, + "step": 31755 + }, + { + "epoch": 0.95, + "learning_rate": 1.5578835504082677e-06, + "loss": 0.5958, + "step": 31756 + }, + { + "epoch": 0.95, + "learning_rate": 1.5561886170079699e-06, + "loss": 0.5163, + "step": 31757 + }, + { + "epoch": 0.95, + "learning_rate": 1.5544945988989923e-06, + "loss": 0.4894, + "step": 31758 + }, + { + "epoch": 0.95, + "learning_rate": 1.5528014960970894e-06, + "loss": 0.5948, + "step": 31759 + }, + { + "epoch": 0.95, + "learning_rate": 1.551109308618015e-06, + "loss": 0.6071, + "step": 31760 + }, + { + "epoch": 0.95, + "learning_rate": 1.549418036477468e-06, + "loss": 0.6736, + "step": 31761 + }, + { + "epoch": 0.95, + "learning_rate": 1.5477276796912021e-06, + "loss": 0.528, + "step": 31762 + }, + { + "epoch": 0.95, + "learning_rate": 1.5460382382749273e-06, + "loss": 0.6179, + "step": 31763 + }, + { + "epoch": 0.95, + "learning_rate": 1.5443497122443528e-06, + "loss": 0.5261, + "step": 31764 + }, + { + "epoch": 0.95, + "learning_rate": 1.5426621016151665e-06, + "loss": 0.4766, + "step": 31765 + }, + { + "epoch": 0.95, + "learning_rate": 1.5409754064030669e-06, + "loss": 0.5457, + "step": 31766 + }, + { + "epoch": 0.95, + "learning_rate": 1.5392896266237412e-06, + "loss": 0.517, + "step": 31767 + }, + { + "epoch": 0.95, + "learning_rate": 1.5376047622928547e-06, + "loss": 0.4085, + "step": 31768 + }, + { + "epoch": 0.95, + "learning_rate": 1.5359208134260838e-06, + "loss": 0.5743, + "step": 31769 + }, + { + "epoch": 0.95, + "learning_rate": 1.5342377800390606e-06, + "loss": 0.5506, + "step": 31770 + }, + { + "epoch": 0.95, + "learning_rate": 1.5325556621474613e-06, + "loss": 0.4963, + "step": 31771 + }, + { + "epoch": 0.95, + "learning_rate": 1.5308744597669178e-06, + "loss": 0.6818, + "step": 31772 + }, + { + "epoch": 0.95, + "learning_rate": 1.529194172913051e-06, + "loss": 0.5663, + "step": 31773 + }, + { + "epoch": 0.95, + "learning_rate": 1.5275148016014817e-06, + "loss": 0.54, + "step": 31774 + }, + { + "epoch": 0.95, + "learning_rate": 1.525836345847853e-06, + "loss": 0.5188, + "step": 31775 + }, + { + "epoch": 0.95, + "learning_rate": 1.5241588056677304e-06, + "loss": 0.5197, + "step": 31776 + }, + { + "epoch": 0.95, + "learning_rate": 1.5224821810767343e-06, + "loss": 0.597, + "step": 31777 + }, + { + "epoch": 0.95, + "learning_rate": 1.5208064720904636e-06, + "loss": 0.4944, + "step": 31778 + }, + { + "epoch": 0.95, + "learning_rate": 1.5191316787244835e-06, + "loss": 0.4775, + "step": 31779 + }, + { + "epoch": 0.95, + "learning_rate": 1.5174578009943486e-06, + "loss": 0.4564, + "step": 31780 + }, + { + "epoch": 0.95, + "learning_rate": 1.515784838915646e-06, + "loss": 0.5927, + "step": 31781 + }, + { + "epoch": 0.95, + "learning_rate": 1.51411279250393e-06, + "loss": 0.4979, + "step": 31782 + }, + { + "epoch": 0.95, + "learning_rate": 1.512441661774744e-06, + "loss": 0.5467, + "step": 31783 + }, + { + "epoch": 0.95, + "learning_rate": 1.5107714467436195e-06, + "loss": 0.4822, + "step": 31784 + }, + { + "epoch": 0.95, + "learning_rate": 1.509102147426078e-06, + "loss": 0.534, + "step": 31785 + }, + { + "epoch": 0.95, + "learning_rate": 1.5074337638376624e-06, + "loss": 0.4589, + "step": 31786 + }, + { + "epoch": 0.95, + "learning_rate": 1.50576629599386e-06, + "loss": 0.4935, + "step": 31787 + }, + { + "epoch": 0.95, + "learning_rate": 1.5040997439102033e-06, + "loss": 0.4901, + "step": 31788 + }, + { + "epoch": 0.95, + "learning_rate": 1.5024341076021576e-06, + "loss": 0.6243, + "step": 31789 + }, + { + "epoch": 0.95, + "learning_rate": 1.5007693870852214e-06, + "loss": 0.5244, + "step": 31790 + }, + { + "epoch": 0.95, + "learning_rate": 1.4991055823748823e-06, + "loss": 0.608, + "step": 31791 + }, + { + "epoch": 0.95, + "learning_rate": 1.4974426934865838e-06, + "loss": 0.5784, + "step": 31792 + }, + { + "epoch": 0.95, + "learning_rate": 1.4957807204358244e-06, + "loss": 0.5492, + "step": 31793 + }, + { + "epoch": 0.95, + "learning_rate": 1.494119663238036e-06, + "loss": 0.677, + "step": 31794 + }, + { + "epoch": 0.95, + "learning_rate": 1.4924595219086513e-06, + "loss": 0.4749, + "step": 31795 + }, + { + "epoch": 0.95, + "learning_rate": 1.490800296463113e-06, + "loss": 0.4735, + "step": 31796 + }, + { + "epoch": 0.95, + "learning_rate": 1.4891419869168644e-06, + "loss": 0.5811, + "step": 31797 + }, + { + "epoch": 0.95, + "learning_rate": 1.4874845932853043e-06, + "loss": 0.4857, + "step": 31798 + }, + { + "epoch": 0.95, + "learning_rate": 1.4858281155838537e-06, + "loss": 0.5176, + "step": 31799 + }, + { + "epoch": 0.95, + "learning_rate": 1.4841725538279006e-06, + "loss": 0.5731, + "step": 31800 + }, + { + "epoch": 0.95, + "learning_rate": 1.4825179080328432e-06, + "loss": 0.4468, + "step": 31801 + }, + { + "epoch": 0.95, + "learning_rate": 1.4808641782140808e-06, + "loss": 0.4565, + "step": 31802 + }, + { + "epoch": 0.95, + "learning_rate": 1.4792113643869787e-06, + "loss": 0.4105, + "step": 31803 + }, + { + "epoch": 0.95, + "learning_rate": 1.4775594665669024e-06, + "loss": 0.5799, + "step": 31804 + }, + { + "epoch": 0.95, + "learning_rate": 1.4759084847692063e-06, + "loss": 0.5199, + "step": 31805 + }, + { + "epoch": 0.95, + "learning_rate": 1.4742584190092446e-06, + "loss": 0.5772, + "step": 31806 + }, + { + "epoch": 0.95, + "learning_rate": 1.4726092693023608e-06, + "loss": 0.5144, + "step": 31807 + }, + { + "epoch": 0.95, + "learning_rate": 1.470961035663887e-06, + "loss": 0.5002, + "step": 31808 + }, + { + "epoch": 0.95, + "learning_rate": 1.4693137181091443e-06, + "loss": 0.5168, + "step": 31809 + }, + { + "epoch": 0.95, + "learning_rate": 1.467667316653465e-06, + "loss": 0.4783, + "step": 31810 + }, + { + "epoch": 0.95, + "learning_rate": 1.4660218313121254e-06, + "loss": 0.4719, + "step": 31811 + }, + { + "epoch": 0.95, + "learning_rate": 1.4643772621004581e-06, + "loss": 0.4922, + "step": 31812 + }, + { + "epoch": 0.95, + "learning_rate": 1.4627336090337285e-06, + "loss": 0.6538, + "step": 31813 + }, + { + "epoch": 0.95, + "learning_rate": 1.4610908721272354e-06, + "loss": 0.5455, + "step": 31814 + }, + { + "epoch": 0.95, + "learning_rate": 1.4594490513962445e-06, + "loss": 0.5461, + "step": 31815 + }, + { + "epoch": 0.95, + "learning_rate": 1.4578081468560212e-06, + "loss": 0.5954, + "step": 31816 + }, + { + "epoch": 0.95, + "learning_rate": 1.4561681585218312e-06, + "loss": 0.453, + "step": 31817 + }, + { + "epoch": 0.95, + "learning_rate": 1.4545290864088957e-06, + "loss": 0.4911, + "step": 31818 + }, + { + "epoch": 0.95, + "learning_rate": 1.45289093053248e-06, + "loss": 0.5391, + "step": 31819 + }, + { + "epoch": 0.95, + "learning_rate": 1.4512536909078056e-06, + "loss": 0.467, + "step": 31820 + }, + { + "epoch": 0.95, + "learning_rate": 1.4496173675501046e-06, + "loss": 0.4949, + "step": 31821 + }, + { + "epoch": 0.95, + "learning_rate": 1.447981960474587e-06, + "loss": 0.5206, + "step": 31822 + }, + { + "epoch": 0.95, + "learning_rate": 1.4463474696964408e-06, + "loss": 0.5157, + "step": 31823 + }, + { + "epoch": 0.95, + "learning_rate": 1.4447138952308758e-06, + "loss": 0.4184, + "step": 31824 + }, + { + "epoch": 0.95, + "learning_rate": 1.4430812370930913e-06, + "loss": 0.5347, + "step": 31825 + }, + { + "epoch": 0.95, + "learning_rate": 1.4414494952982527e-06, + "loss": 0.4524, + "step": 31826 + }, + { + "epoch": 0.95, + "learning_rate": 1.4398186698615367e-06, + "loss": 0.6026, + "step": 31827 + }, + { + "epoch": 0.95, + "learning_rate": 1.438188760798098e-06, + "loss": 0.472, + "step": 31828 + }, + { + "epoch": 0.95, + "learning_rate": 1.4365597681231024e-06, + "loss": 0.5277, + "step": 31829 + }, + { + "epoch": 0.95, + "learning_rate": 1.434931691851682e-06, + "loss": 0.471, + "step": 31830 + }, + { + "epoch": 0.95, + "learning_rate": 1.4333045319990024e-06, + "loss": 0.444, + "step": 31831 + }, + { + "epoch": 0.95, + "learning_rate": 1.4316782885801517e-06, + "loss": 0.3655, + "step": 31832 + }, + { + "epoch": 0.95, + "learning_rate": 1.4300529616102843e-06, + "loss": 0.5086, + "step": 31833 + }, + { + "epoch": 0.95, + "learning_rate": 1.4284285511044882e-06, + "loss": 0.5237, + "step": 31834 + }, + { + "epoch": 0.95, + "learning_rate": 1.4268050570778845e-06, + "loss": 0.4863, + "step": 31835 + }, + { + "epoch": 0.95, + "learning_rate": 1.4251824795455504e-06, + "loss": 0.5215, + "step": 31836 + }, + { + "epoch": 0.95, + "learning_rate": 1.4235608185225958e-06, + "loss": 0.5583, + "step": 31837 + }, + { + "epoch": 0.95, + "learning_rate": 1.4219400740240752e-06, + "loss": 0.5107, + "step": 31838 + }, + { + "epoch": 0.95, + "learning_rate": 1.4203202460650656e-06, + "loss": 0.5884, + "step": 31839 + }, + { + "epoch": 0.95, + "learning_rate": 1.4187013346606326e-06, + "loss": 0.5559, + "step": 31840 + }, + { + "epoch": 0.95, + "learning_rate": 1.4170833398258198e-06, + "loss": 0.5657, + "step": 31841 + }, + { + "epoch": 0.95, + "learning_rate": 1.4154662615756819e-06, + "loss": 0.5822, + "step": 31842 + }, + { + "epoch": 0.95, + "learning_rate": 1.4138500999252292e-06, + "loss": 0.4507, + "step": 31843 + }, + { + "epoch": 0.95, + "learning_rate": 1.412234854889516e-06, + "loss": 0.5032, + "step": 31844 + }, + { + "epoch": 0.95, + "learning_rate": 1.4106205264835637e-06, + "loss": 0.4359, + "step": 31845 + }, + { + "epoch": 0.95, + "learning_rate": 1.4090071147223604e-06, + "loss": 0.4834, + "step": 31846 + }, + { + "epoch": 0.95, + "learning_rate": 1.4073946196209054e-06, + "loss": 0.4703, + "step": 31847 + }, + { + "epoch": 0.95, + "learning_rate": 1.4057830411942085e-06, + "loss": 0.5519, + "step": 31848 + }, + { + "epoch": 0.95, + "learning_rate": 1.404172379457236e-06, + "loss": 0.4807, + "step": 31849 + }, + { + "epoch": 0.95, + "learning_rate": 1.4025626344249864e-06, + "loss": 0.5202, + "step": 31850 + }, + { + "epoch": 0.95, + "learning_rate": 1.4009538061123927e-06, + "loss": 0.562, + "step": 31851 + }, + { + "epoch": 0.95, + "learning_rate": 1.399345894534454e-06, + "loss": 0.5442, + "step": 31852 + }, + { + "epoch": 0.95, + "learning_rate": 1.3977388997060803e-06, + "loss": 0.5161, + "step": 31853 + }, + { + "epoch": 0.95, + "learning_rate": 1.3961328216422375e-06, + "loss": 0.4885, + "step": 31854 + }, + { + "epoch": 0.95, + "learning_rate": 1.3945276603578584e-06, + "loss": 0.6344, + "step": 31855 + }, + { + "epoch": 0.95, + "learning_rate": 1.392923415867864e-06, + "loss": 0.5187, + "step": 31856 + }, + { + "epoch": 0.95, + "learning_rate": 1.3913200881871536e-06, + "loss": 0.5304, + "step": 31857 + }, + { + "epoch": 0.95, + "learning_rate": 1.3897176773306487e-06, + "loss": 0.4576, + "step": 31858 + }, + { + "epoch": 0.95, + "learning_rate": 1.3881161833132595e-06, + "loss": 0.5341, + "step": 31859 + }, + { + "epoch": 0.95, + "learning_rate": 1.3865156061498518e-06, + "loss": 0.5051, + "step": 31860 + }, + { + "epoch": 0.95, + "learning_rate": 1.3849159458553252e-06, + "loss": 0.611, + "step": 31861 + }, + { + "epoch": 0.95, + "learning_rate": 1.383317202444534e-06, + "loss": 0.4492, + "step": 31862 + }, + { + "epoch": 0.95, + "learning_rate": 1.3817193759323666e-06, + "loss": 0.545, + "step": 31863 + }, + { + "epoch": 0.95, + "learning_rate": 1.3801224663336665e-06, + "loss": 0.5492, + "step": 31864 + }, + { + "epoch": 0.95, + "learning_rate": 1.3785264736632774e-06, + "loss": 0.4399, + "step": 31865 + }, + { + "epoch": 0.95, + "learning_rate": 1.3769313979360432e-06, + "loss": 0.6492, + "step": 31866 + }, + { + "epoch": 0.95, + "learning_rate": 1.3753372391667963e-06, + "loss": 0.5378, + "step": 31867 + }, + { + "epoch": 0.95, + "learning_rate": 1.373743997370358e-06, + "loss": 0.4832, + "step": 31868 + }, + { + "epoch": 0.95, + "learning_rate": 1.372151672561539e-06, + "loss": 0.5183, + "step": 31869 + }, + { + "epoch": 0.95, + "learning_rate": 1.3705602647551385e-06, + "loss": 0.5722, + "step": 31870 + }, + { + "epoch": 0.95, + "learning_rate": 1.3689697739659556e-06, + "loss": 0.4824, + "step": 31871 + }, + { + "epoch": 0.95, + "learning_rate": 1.3673802002087898e-06, + "loss": 0.5569, + "step": 31872 + }, + { + "epoch": 0.95, + "learning_rate": 1.365791543498407e-06, + "loss": 0.6303, + "step": 31873 + }, + { + "epoch": 0.95, + "learning_rate": 1.3642038038495841e-06, + "loss": 0.5308, + "step": 31874 + }, + { + "epoch": 0.95, + "learning_rate": 1.3626169812770873e-06, + "loss": 0.413, + "step": 31875 + }, + { + "epoch": 0.95, + "learning_rate": 1.3610310757956601e-06, + "loss": 0.6287, + "step": 31876 + }, + { + "epoch": 0.95, + "learning_rate": 1.3594460874200465e-06, + "loss": 0.4486, + "step": 31877 + }, + { + "epoch": 0.95, + "learning_rate": 1.3578620161650014e-06, + "loss": 0.4395, + "step": 31878 + }, + { + "epoch": 0.95, + "learning_rate": 1.356278862045235e-06, + "loss": 0.6601, + "step": 31879 + }, + { + "epoch": 0.95, + "learning_rate": 1.354696625075469e-06, + "loss": 0.3758, + "step": 31880 + }, + { + "epoch": 0.95, + "learning_rate": 1.353115305270425e-06, + "loss": 0.5936, + "step": 31881 + }, + { + "epoch": 0.95, + "learning_rate": 1.35153490264478e-06, + "loss": 0.488, + "step": 31882 + }, + { + "epoch": 0.95, + "learning_rate": 1.3499554172132667e-06, + "loss": 0.4279, + "step": 31883 + }, + { + "epoch": 0.95, + "learning_rate": 1.3483768489905512e-06, + "loss": 0.5202, + "step": 31884 + }, + { + "epoch": 0.95, + "learning_rate": 1.3467991979912887e-06, + "loss": 0.5396, + "step": 31885 + }, + { + "epoch": 0.95, + "learning_rate": 1.345222464230178e-06, + "loss": 0.5819, + "step": 31886 + }, + { + "epoch": 0.95, + "learning_rate": 1.3436466477218746e-06, + "loss": 0.5034, + "step": 31887 + }, + { + "epoch": 0.95, + "learning_rate": 1.342071748481022e-06, + "loss": 0.4867, + "step": 31888 + }, + { + "epoch": 0.95, + "learning_rate": 1.340497766522264e-06, + "loss": 0.6387, + "step": 31889 + }, + { + "epoch": 0.95, + "learning_rate": 1.3389247018602335e-06, + "loss": 0.539, + "step": 31890 + }, + { + "epoch": 0.95, + "learning_rate": 1.3373525545095633e-06, + "loss": 0.6015, + "step": 31891 + }, + { + "epoch": 0.95, + "learning_rate": 1.3357813244848639e-06, + "loss": 0.611, + "step": 31892 + }, + { + "epoch": 0.95, + "learning_rate": 1.3342110118007457e-06, + "loss": 0.4758, + "step": 31893 + }, + { + "epoch": 0.95, + "learning_rate": 1.332641616471808e-06, + "loss": 0.4461, + "step": 31894 + }, + { + "epoch": 0.95, + "learning_rate": 1.3310731385126396e-06, + "loss": 0.5156, + "step": 31895 + }, + { + "epoch": 0.95, + "learning_rate": 1.3295055779378285e-06, + "loss": 0.6108, + "step": 31896 + }, + { + "epoch": 0.95, + "learning_rate": 1.3279389347619519e-06, + "loss": 0.5444, + "step": 31897 + }, + { + "epoch": 0.95, + "learning_rate": 1.326373208999576e-06, + "loss": 0.6144, + "step": 31898 + }, + { + "epoch": 0.95, + "learning_rate": 1.324808400665256e-06, + "loss": 0.4942, + "step": 31899 + }, + { + "epoch": 0.95, + "learning_rate": 1.3232445097735358e-06, + "loss": 0.4549, + "step": 31900 + }, + { + "epoch": 0.95, + "learning_rate": 1.3216815363389478e-06, + "loss": 0.5109, + "step": 31901 + }, + { + "epoch": 0.95, + "learning_rate": 1.3201194803760475e-06, + "loss": 0.5251, + "step": 31902 + }, + { + "epoch": 0.95, + "learning_rate": 1.3185583418993564e-06, + "loss": 0.4014, + "step": 31903 + }, + { + "epoch": 0.95, + "learning_rate": 1.316998120923363e-06, + "loss": 0.4303, + "step": 31904 + }, + { + "epoch": 0.95, + "learning_rate": 1.3154388174625886e-06, + "loss": 0.5887, + "step": 31905 + }, + { + "epoch": 0.95, + "learning_rate": 1.3138804315315334e-06, + "loss": 0.4521, + "step": 31906 + }, + { + "epoch": 0.95, + "learning_rate": 1.3123229631446964e-06, + "loss": 0.4331, + "step": 31907 + }, + { + "epoch": 0.95, + "learning_rate": 1.3107664123165442e-06, + "loss": 0.5141, + "step": 31908 + }, + { + "epoch": 0.95, + "learning_rate": 1.3092107790615426e-06, + "loss": 0.4059, + "step": 31909 + }, + { + "epoch": 0.95, + "learning_rate": 1.3076560633941693e-06, + "loss": 0.4172, + "step": 31910 + }, + { + "epoch": 0.95, + "learning_rate": 1.3061022653288901e-06, + "loss": 0.6236, + "step": 31911 + }, + { + "epoch": 0.95, + "learning_rate": 1.3045493848801271e-06, + "loss": 0.5606, + "step": 31912 + }, + { + "epoch": 0.95, + "learning_rate": 1.302997422062313e-06, + "loss": 0.6065, + "step": 31913 + }, + { + "epoch": 0.95, + "learning_rate": 1.3014463768899142e-06, + "loss": 0.6829, + "step": 31914 + }, + { + "epoch": 0.95, + "learning_rate": 1.2998962493773192e-06, + "loss": 0.5784, + "step": 31915 + }, + { + "epoch": 0.95, + "learning_rate": 1.2983470395389385e-06, + "loss": 0.434, + "step": 31916 + }, + { + "epoch": 0.95, + "learning_rate": 1.296798747389205e-06, + "loss": 0.5397, + "step": 31917 + }, + { + "epoch": 0.95, + "learning_rate": 1.2952513729424964e-06, + "loss": 0.5837, + "step": 31918 + }, + { + "epoch": 0.95, + "learning_rate": 1.293704916213201e-06, + "loss": 0.4471, + "step": 31919 + }, + { + "epoch": 0.95, + "learning_rate": 1.2921593772156849e-06, + "loss": 0.6447, + "step": 31920 + }, + { + "epoch": 0.95, + "learning_rate": 1.2906147559643477e-06, + "loss": 0.5687, + "step": 31921 + }, + { + "epoch": 0.95, + "learning_rate": 1.2890710524735229e-06, + "loss": 0.5335, + "step": 31922 + }, + { + "epoch": 0.95, + "learning_rate": 1.287528266757576e-06, + "loss": 0.5085, + "step": 31923 + }, + { + "epoch": 0.95, + "learning_rate": 1.2859863988308407e-06, + "loss": 0.5033, + "step": 31924 + }, + { + "epoch": 0.95, + "learning_rate": 1.2844454487076607e-06, + "loss": 0.6308, + "step": 31925 + }, + { + "epoch": 0.95, + "learning_rate": 1.282905416402369e-06, + "loss": 0.5524, + "step": 31926 + }, + { + "epoch": 0.95, + "learning_rate": 1.2813663019292876e-06, + "loss": 0.518, + "step": 31927 + }, + { + "epoch": 0.95, + "learning_rate": 1.279828105302705e-06, + "loss": 0.5659, + "step": 31928 + }, + { + "epoch": 0.95, + "learning_rate": 1.278290826536932e-06, + "loss": 0.4596, + "step": 31929 + }, + { + "epoch": 0.95, + "learning_rate": 1.2767544656462682e-06, + "loss": 0.4455, + "step": 31930 + }, + { + "epoch": 0.95, + "learning_rate": 1.2752190226450023e-06, + "loss": 0.652, + "step": 31931 + }, + { + "epoch": 0.95, + "learning_rate": 1.2736844975473893e-06, + "loss": 0.6201, + "step": 31932 + }, + { + "epoch": 0.95, + "learning_rate": 1.272150890367707e-06, + "loss": 0.5003, + "step": 31933 + }, + { + "epoch": 0.95, + "learning_rate": 1.2706182011202327e-06, + "loss": 0.4788, + "step": 31934 + }, + { + "epoch": 0.95, + "learning_rate": 1.2690864298191885e-06, + "loss": 0.4, + "step": 31935 + }, + { + "epoch": 0.95, + "learning_rate": 1.2675555764788294e-06, + "loss": 0.541, + "step": 31936 + }, + { + "epoch": 0.95, + "learning_rate": 1.2660256411133887e-06, + "loss": 0.4969, + "step": 31937 + }, + { + "epoch": 0.95, + "learning_rate": 1.2644966237370991e-06, + "loss": 0.5363, + "step": 31938 + }, + { + "epoch": 0.95, + "learning_rate": 1.2629685243641498e-06, + "loss": 0.4314, + "step": 31939 + }, + { + "epoch": 0.95, + "learning_rate": 1.2614413430087846e-06, + "loss": 0.5577, + "step": 31940 + }, + { + "epoch": 0.95, + "learning_rate": 1.25991507968517e-06, + "loss": 0.527, + "step": 31941 + }, + { + "epoch": 0.95, + "learning_rate": 1.258389734407528e-06, + "loss": 0.6066, + "step": 31942 + }, + { + "epoch": 0.95, + "learning_rate": 1.2568653071900139e-06, + "loss": 0.4578, + "step": 31943 + }, + { + "epoch": 0.95, + "learning_rate": 1.2553417980468052e-06, + "loss": 0.4742, + "step": 31944 + }, + { + "epoch": 0.95, + "learning_rate": 1.2538192069920795e-06, + "loss": 0.4146, + "step": 31945 + }, + { + "epoch": 0.95, + "learning_rate": 1.2522975340399922e-06, + "loss": 0.6727, + "step": 31946 + }, + { + "epoch": 0.95, + "learning_rate": 1.2507767792046765e-06, + "loss": 0.5132, + "step": 31947 + }, + { + "epoch": 0.95, + "learning_rate": 1.2492569425002764e-06, + "loss": 0.5005, + "step": 31948 + }, + { + "epoch": 0.95, + "learning_rate": 1.2477380239409364e-06, + "loss": 0.4905, + "step": 31949 + }, + { + "epoch": 0.95, + "learning_rate": 1.2462200235407674e-06, + "loss": 0.5117, + "step": 31950 + }, + { + "epoch": 0.95, + "learning_rate": 1.24470294131388e-06, + "loss": 0.4851, + "step": 31951 + }, + { + "epoch": 0.95, + "learning_rate": 1.2431867772743855e-06, + "loss": 0.5604, + "step": 31952 + }, + { + "epoch": 0.95, + "learning_rate": 1.2416715314363946e-06, + "loss": 0.496, + "step": 31953 + }, + { + "epoch": 0.95, + "learning_rate": 1.240157203813963e-06, + "loss": 0.4977, + "step": 31954 + }, + { + "epoch": 0.95, + "learning_rate": 1.2386437944212015e-06, + "loss": 0.4567, + "step": 31955 + }, + { + "epoch": 0.95, + "learning_rate": 1.2371313032721544e-06, + "loss": 0.5123, + "step": 31956 + }, + { + "epoch": 0.95, + "learning_rate": 1.2356197303809104e-06, + "loss": 0.4985, + "step": 31957 + }, + { + "epoch": 0.95, + "learning_rate": 1.2341090757614915e-06, + "loss": 0.7119, + "step": 31958 + }, + { + "epoch": 0.95, + "learning_rate": 1.2325993394279756e-06, + "loss": 0.563, + "step": 31959 + }, + { + "epoch": 0.95, + "learning_rate": 1.2310905213943847e-06, + "loss": 0.4654, + "step": 31960 + }, + { + "epoch": 0.95, + "learning_rate": 1.229582621674752e-06, + "loss": 0.5295, + "step": 31961 + }, + { + "epoch": 0.95, + "learning_rate": 1.2280756402830885e-06, + "loss": 0.563, + "step": 31962 + }, + { + "epoch": 0.95, + "learning_rate": 1.2265695772334163e-06, + "loss": 0.6267, + "step": 31963 + }, + { + "epoch": 0.95, + "learning_rate": 1.2250644325397354e-06, + "loss": 0.4179, + "step": 31964 + }, + { + "epoch": 0.95, + "learning_rate": 1.2235602062160346e-06, + "loss": 0.6356, + "step": 31965 + }, + { + "epoch": 0.95, + "learning_rate": 1.2220568982762914e-06, + "loss": 0.6135, + "step": 31966 + }, + { + "epoch": 0.95, + "learning_rate": 1.220554508734506e-06, + "loss": 0.5741, + "step": 31967 + }, + { + "epoch": 0.95, + "learning_rate": 1.2190530376046338e-06, + "loss": 0.4868, + "step": 31968 + }, + { + "epoch": 0.95, + "learning_rate": 1.2175524849006304e-06, + "loss": 0.4915, + "step": 31969 + }, + { + "epoch": 0.95, + "learning_rate": 1.2160528506364622e-06, + "loss": 0.5769, + "step": 31970 + }, + { + "epoch": 0.95, + "learning_rate": 1.2145541348260626e-06, + "loss": 0.5598, + "step": 31971 + }, + { + "epoch": 0.95, + "learning_rate": 1.2130563374833536e-06, + "loss": 0.3574, + "step": 31972 + }, + { + "epoch": 0.95, + "learning_rate": 1.211559458622291e-06, + "loss": 0.5276, + "step": 31973 + }, + { + "epoch": 0.95, + "learning_rate": 1.2100634982567639e-06, + "loss": 0.6493, + "step": 31974 + }, + { + "epoch": 0.95, + "learning_rate": 1.208568456400694e-06, + "loss": 0.4653, + "step": 31975 + }, + { + "epoch": 0.95, + "learning_rate": 1.2070743330679812e-06, + "loss": 0.4968, + "step": 31976 + }, + { + "epoch": 0.95, + "learning_rate": 1.205581128272526e-06, + "loss": 0.5021, + "step": 31977 + }, + { + "epoch": 0.95, + "learning_rate": 1.2040888420281838e-06, + "loss": 0.5542, + "step": 31978 + }, + { + "epoch": 0.95, + "learning_rate": 1.2025974743488654e-06, + "loss": 0.5672, + "step": 31979 + }, + { + "epoch": 0.95, + "learning_rate": 1.2011070252484157e-06, + "loss": 0.4764, + "step": 31980 + }, + { + "epoch": 0.95, + "learning_rate": 1.1996174947406901e-06, + "loss": 0.4415, + "step": 31981 + }, + { + "epoch": 0.95, + "learning_rate": 1.198128882839533e-06, + "loss": 0.5243, + "step": 31982 + }, + { + "epoch": 0.95, + "learning_rate": 1.1966411895588115e-06, + "loss": 0.5427, + "step": 31983 + }, + { + "epoch": 0.95, + "learning_rate": 1.1951544149123363e-06, + "loss": 0.5811, + "step": 31984 + }, + { + "epoch": 0.95, + "learning_rate": 1.1936685589139407e-06, + "loss": 0.4242, + "step": 31985 + }, + { + "epoch": 0.95, + "learning_rate": 1.192183621577425e-06, + "loss": 0.5566, + "step": 31986 + }, + { + "epoch": 0.95, + "learning_rate": 1.1906996029166007e-06, + "loss": 0.5088, + "step": 31987 + }, + { + "epoch": 0.95, + "learning_rate": 1.1892165029452785e-06, + "loss": 0.5287, + "step": 31988 + }, + { + "epoch": 0.95, + "learning_rate": 1.1877343216772475e-06, + "loss": 0.5778, + "step": 31989 + }, + { + "epoch": 0.95, + "learning_rate": 1.1862530591262632e-06, + "loss": 0.5012, + "step": 31990 + }, + { + "epoch": 0.95, + "learning_rate": 1.184772715306126e-06, + "loss": 0.5834, + "step": 31991 + }, + { + "epoch": 0.95, + "learning_rate": 1.1832932902305803e-06, + "loss": 0.4969, + "step": 31992 + }, + { + "epoch": 0.95, + "learning_rate": 1.1818147839134041e-06, + "loss": 0.6319, + "step": 31993 + }, + { + "epoch": 0.95, + "learning_rate": 1.1803371963683085e-06, + "loss": 0.5149, + "step": 31994 + }, + { + "epoch": 0.95, + "learning_rate": 1.1788605276090602e-06, + "loss": 0.3812, + "step": 31995 + }, + { + "epoch": 0.95, + "learning_rate": 1.1773847776493817e-06, + "loss": 0.503, + "step": 31996 + }, + { + "epoch": 0.95, + "learning_rate": 1.1759099465029954e-06, + "loss": 0.5478, + "step": 31997 + }, + { + "epoch": 0.95, + "learning_rate": 1.1744360341836124e-06, + "loss": 0.5031, + "step": 31998 + }, + { + "epoch": 0.95, + "learning_rate": 1.172963040704933e-06, + "loss": 0.415, + "step": 31999 + }, + { + "epoch": 0.95, + "learning_rate": 1.1714909660806572e-06, + "loss": 0.5816, + "step": 32000 + }, + { + "epoch": 0.95, + "learning_rate": 1.170019810324452e-06, + "loss": 0.4651, + "step": 32001 + }, + { + "epoch": 0.95, + "learning_rate": 1.1685495734500397e-06, + "loss": 0.5738, + "step": 32002 + }, + { + "epoch": 0.95, + "learning_rate": 1.1670802554710425e-06, + "loss": 0.4877, + "step": 32003 + }, + { + "epoch": 0.95, + "learning_rate": 1.1656118564011499e-06, + "loss": 0.5335, + "step": 32004 + }, + { + "epoch": 0.95, + "learning_rate": 1.1641443762540062e-06, + "loss": 0.6004, + "step": 32005 + }, + { + "epoch": 0.95, + "learning_rate": 1.162677815043256e-06, + "loss": 0.6112, + "step": 32006 + }, + { + "epoch": 0.95, + "learning_rate": 1.1612121727825332e-06, + "loss": 0.4749, + "step": 32007 + }, + { + "epoch": 0.95, + "learning_rate": 1.159747449485471e-06, + "loss": 0.6259, + "step": 32008 + }, + { + "epoch": 0.95, + "learning_rate": 1.158283645165681e-06, + "loss": 0.5319, + "step": 32009 + }, + { + "epoch": 0.95, + "learning_rate": 1.1568207598367742e-06, + "loss": 0.5177, + "step": 32010 + }, + { + "epoch": 0.95, + "learning_rate": 1.155358793512351e-06, + "loss": 0.6357, + "step": 32011 + }, + { + "epoch": 0.95, + "learning_rate": 1.153897746206012e-06, + "loss": 0.5041, + "step": 32012 + }, + { + "epoch": 0.95, + "learning_rate": 1.1524376179313345e-06, + "loss": 0.4978, + "step": 32013 + }, + { + "epoch": 0.95, + "learning_rate": 1.150978408701886e-06, + "loss": 0.6738, + "step": 32014 + }, + { + "epoch": 0.95, + "learning_rate": 1.1495201185312553e-06, + "loss": 0.5678, + "step": 32015 + }, + { + "epoch": 0.95, + "learning_rate": 1.1480627474329763e-06, + "loss": 0.5157, + "step": 32016 + }, + { + "epoch": 0.95, + "learning_rate": 1.1466062954206268e-06, + "loss": 0.5187, + "step": 32017 + }, + { + "epoch": 0.95, + "learning_rate": 1.1451507625077185e-06, + "loss": 0.5814, + "step": 32018 + }, + { + "epoch": 0.95, + "learning_rate": 1.1436961487078067e-06, + "loss": 0.492, + "step": 32019 + }, + { + "epoch": 0.95, + "learning_rate": 1.1422424540344034e-06, + "loss": 0.5367, + "step": 32020 + }, + { + "epoch": 0.95, + "learning_rate": 1.1407896785010308e-06, + "loss": 0.6493, + "step": 32021 + }, + { + "epoch": 0.95, + "learning_rate": 1.1393378221211893e-06, + "loss": 0.48, + "step": 32022 + }, + { + "epoch": 0.95, + "learning_rate": 1.1378868849084011e-06, + "loss": 0.4957, + "step": 32023 + }, + { + "epoch": 0.95, + "learning_rate": 1.1364368668761115e-06, + "loss": 0.5314, + "step": 32024 + }, + { + "epoch": 0.95, + "learning_rate": 1.134987768037843e-06, + "loss": 0.5141, + "step": 32025 + }, + { + "epoch": 0.95, + "learning_rate": 1.133539588407062e-06, + "loss": 0.542, + "step": 32026 + }, + { + "epoch": 0.95, + "learning_rate": 1.132092327997214e-06, + "loss": 0.4354, + "step": 32027 + }, + { + "epoch": 0.95, + "learning_rate": 1.1306459868217767e-06, + "loss": 0.6063, + "step": 32028 + }, + { + "epoch": 0.95, + "learning_rate": 1.1292005648941727e-06, + "loss": 0.5998, + "step": 32029 + }, + { + "epoch": 0.95, + "learning_rate": 1.1277560622278583e-06, + "loss": 0.5908, + "step": 32030 + }, + { + "epoch": 0.95, + "learning_rate": 1.1263124788362778e-06, + "loss": 0.4785, + "step": 32031 + }, + { + "epoch": 0.95, + "learning_rate": 1.1248698147328207e-06, + "loss": 0.4356, + "step": 32032 + }, + { + "epoch": 0.95, + "learning_rate": 1.1234280699309207e-06, + "loss": 0.4789, + "step": 32033 + }, + { + "epoch": 0.95, + "learning_rate": 1.1219872444439784e-06, + "loss": 0.5656, + "step": 32034 + }, + { + "epoch": 0.95, + "learning_rate": 1.1205473382853937e-06, + "loss": 0.6165, + "step": 32035 + }, + { + "epoch": 0.95, + "learning_rate": 1.1191083514685563e-06, + "loss": 0.4737, + "step": 32036 + }, + { + "epoch": 0.95, + "learning_rate": 1.1176702840068221e-06, + "loss": 0.5497, + "step": 32037 + }, + { + "epoch": 0.95, + "learning_rate": 1.1162331359135913e-06, + "loss": 0.431, + "step": 32038 + }, + { + "epoch": 0.95, + "learning_rate": 1.114796907202209e-06, + "loss": 0.5697, + "step": 32039 + }, + { + "epoch": 0.95, + "learning_rate": 1.1133615978860313e-06, + "loss": 0.421, + "step": 32040 + }, + { + "epoch": 0.95, + "learning_rate": 1.1119272079784137e-06, + "loss": 0.6045, + "step": 32041 + }, + { + "epoch": 0.95, + "learning_rate": 1.1104937374926683e-06, + "loss": 0.487, + "step": 32042 + }, + { + "epoch": 0.95, + "learning_rate": 1.109061186442162e-06, + "loss": 0.4862, + "step": 32043 + }, + { + "epoch": 0.95, + "learning_rate": 1.1076295548401727e-06, + "loss": 0.5182, + "step": 32044 + }, + { + "epoch": 0.95, + "learning_rate": 1.1061988427000347e-06, + "loss": 0.4417, + "step": 32045 + }, + { + "epoch": 0.95, + "learning_rate": 1.1047690500350483e-06, + "loss": 0.5365, + "step": 32046 + }, + { + "epoch": 0.95, + "learning_rate": 1.1033401768584917e-06, + "loss": 0.5232, + "step": 32047 + }, + { + "epoch": 0.95, + "learning_rate": 1.1019122231836653e-06, + "loss": 0.5236, + "step": 32048 + }, + { + "epoch": 0.95, + "learning_rate": 1.1004851890238476e-06, + "loss": 0.5058, + "step": 32049 + }, + { + "epoch": 0.95, + "learning_rate": 1.0990590743923056e-06, + "loss": 0.5758, + "step": 32050 + }, + { + "epoch": 0.95, + "learning_rate": 1.0976338793022845e-06, + "loss": 0.6384, + "step": 32051 + }, + { + "epoch": 0.95, + "learning_rate": 1.096209603767051e-06, + "loss": 0.6476, + "step": 32052 + }, + { + "epoch": 0.95, + "learning_rate": 1.0947862477998393e-06, + "loss": 0.5279, + "step": 32053 + }, + { + "epoch": 0.95, + "learning_rate": 1.0933638114138833e-06, + "loss": 0.5156, + "step": 32054 + }, + { + "epoch": 0.95, + "learning_rate": 1.0919422946224167e-06, + "loss": 0.4955, + "step": 32055 + }, + { + "epoch": 0.95, + "learning_rate": 1.09052169743864e-06, + "loss": 0.6516, + "step": 32056 + }, + { + "epoch": 0.95, + "learning_rate": 1.0891020198757763e-06, + "loss": 0.591, + "step": 32057 + }, + { + "epoch": 0.95, + "learning_rate": 1.0876832619470255e-06, + "loss": 0.5768, + "step": 32058 + }, + { + "epoch": 0.95, + "learning_rate": 1.0862654236655666e-06, + "loss": 0.5568, + "step": 32059 + }, + { + "epoch": 0.95, + "learning_rate": 1.084848505044589e-06, + "loss": 0.5703, + "step": 32060 + }, + { + "epoch": 0.95, + "learning_rate": 1.0834325060972706e-06, + "loss": 0.6327, + "step": 32061 + }, + { + "epoch": 0.95, + "learning_rate": 1.0820174268367678e-06, + "loss": 0.6199, + "step": 32062 + }, + { + "epoch": 0.95, + "learning_rate": 1.0806032672762367e-06, + "loss": 0.5892, + "step": 32063 + }, + { + "epoch": 0.95, + "learning_rate": 1.0791900274288336e-06, + "loss": 0.4864, + "step": 32064 + }, + { + "epoch": 0.95, + "learning_rate": 1.0777777073077034e-06, + "loss": 0.4093, + "step": 32065 + }, + { + "epoch": 0.95, + "learning_rate": 1.07636630692598e-06, + "loss": 0.5581, + "step": 32066 + }, + { + "epoch": 0.95, + "learning_rate": 1.074955826296753e-06, + "loss": 0.54, + "step": 32067 + }, + { + "epoch": 0.95, + "learning_rate": 1.0735462654331673e-06, + "loss": 0.4544, + "step": 32068 + }, + { + "epoch": 0.95, + "learning_rate": 1.0721376243483239e-06, + "loss": 0.4664, + "step": 32069 + }, + { + "epoch": 0.95, + "learning_rate": 1.070729903055312e-06, + "loss": 0.4315, + "step": 32070 + }, + { + "epoch": 0.95, + "learning_rate": 1.0693231015672212e-06, + "loss": 0.4123, + "step": 32071 + }, + { + "epoch": 0.95, + "learning_rate": 1.067917219897141e-06, + "loss": 0.642, + "step": 32072 + }, + { + "epoch": 0.95, + "learning_rate": 1.0665122580581277e-06, + "loss": 0.6236, + "step": 32073 + }, + { + "epoch": 0.95, + "learning_rate": 1.0651082160632597e-06, + "loss": 0.4911, + "step": 32074 + }, + { + "epoch": 0.95, + "learning_rate": 1.0637050939255932e-06, + "loss": 0.4995, + "step": 32075 + }, + { + "epoch": 0.95, + "learning_rate": 1.0623028916581511e-06, + "loss": 0.4948, + "step": 32076 + }, + { + "epoch": 0.95, + "learning_rate": 1.0609016092739898e-06, + "loss": 0.6728, + "step": 32077 + }, + { + "epoch": 0.96, + "learning_rate": 1.059501246786121e-06, + "loss": 0.5664, + "step": 32078 + }, + { + "epoch": 0.96, + "learning_rate": 1.0581018042075896e-06, + "loss": 0.5759, + "step": 32079 + }, + { + "epoch": 0.96, + "learning_rate": 1.0567032815513855e-06, + "loss": 0.5198, + "step": 32080 + }, + { + "epoch": 0.96, + "learning_rate": 1.0553056788305204e-06, + "loss": 0.5269, + "step": 32081 + }, + { + "epoch": 0.96, + "learning_rate": 1.053908996057984e-06, + "loss": 0.5497, + "step": 32082 + }, + { + "epoch": 0.96, + "learning_rate": 1.0525132332467658e-06, + "loss": 0.5508, + "step": 32083 + }, + { + "epoch": 0.96, + "learning_rate": 1.0511183904098553e-06, + "loss": 0.4385, + "step": 32084 + }, + { + "epoch": 0.96, + "learning_rate": 1.0497244675602092e-06, + "loss": 0.444, + "step": 32085 + }, + { + "epoch": 0.96, + "learning_rate": 1.0483314647107722e-06, + "loss": 0.5642, + "step": 32086 + }, + { + "epoch": 0.96, + "learning_rate": 1.046939381874512e-06, + "loss": 0.5195, + "step": 32087 + }, + { + "epoch": 0.96, + "learning_rate": 1.045548219064385e-06, + "loss": 0.5623, + "step": 32088 + }, + { + "epoch": 0.96, + "learning_rate": 1.0441579762933029e-06, + "loss": 0.5713, + "step": 32089 + }, + { + "epoch": 0.96, + "learning_rate": 1.0427686535742e-06, + "loss": 0.514, + "step": 32090 + }, + { + "epoch": 0.96, + "learning_rate": 1.0413802509199877e-06, + "loss": 0.5382, + "step": 32091 + }, + { + "epoch": 0.96, + "learning_rate": 1.0399927683435894e-06, + "loss": 0.6943, + "step": 32092 + }, + { + "epoch": 0.96, + "learning_rate": 1.038606205857895e-06, + "loss": 0.5087, + "step": 32093 + }, + { + "epoch": 0.96, + "learning_rate": 1.0372205634758048e-06, + "loss": 0.4492, + "step": 32094 + }, + { + "epoch": 0.96, + "learning_rate": 1.0358358412101753e-06, + "loss": 0.545, + "step": 32095 + }, + { + "epoch": 0.96, + "learning_rate": 1.0344520390739186e-06, + "loss": 0.6418, + "step": 32096 + }, + { + "epoch": 0.96, + "learning_rate": 1.03306915707988e-06, + "loss": 0.5958, + "step": 32097 + }, + { + "epoch": 0.96, + "learning_rate": 1.0316871952409157e-06, + "loss": 0.6438, + "step": 32098 + }, + { + "epoch": 0.96, + "learning_rate": 1.0303061535698822e-06, + "loss": 0.4635, + "step": 32099 + }, + { + "epoch": 0.96, + "learning_rate": 1.0289260320796135e-06, + "loss": 0.5539, + "step": 32100 + }, + { + "epoch": 0.96, + "learning_rate": 1.0275468307829438e-06, + "loss": 0.5745, + "step": 32101 + }, + { + "epoch": 0.96, + "learning_rate": 1.0261685496926965e-06, + "loss": 0.4892, + "step": 32102 + }, + { + "epoch": 0.96, + "learning_rate": 1.0247911888216943e-06, + "loss": 0.419, + "step": 32103 + }, + { + "epoch": 0.96, + "learning_rate": 1.0234147481827272e-06, + "loss": 0.5009, + "step": 32104 + }, + { + "epoch": 0.96, + "learning_rate": 1.0220392277886182e-06, + "loss": 0.5212, + "step": 32105 + }, + { + "epoch": 0.96, + "learning_rate": 1.0206646276521236e-06, + "loss": 0.4953, + "step": 32106 + }, + { + "epoch": 0.96, + "learning_rate": 1.0192909477860447e-06, + "loss": 0.5288, + "step": 32107 + }, + { + "epoch": 0.96, + "learning_rate": 1.0179181882031485e-06, + "loss": 0.644, + "step": 32108 + }, + { + "epoch": 0.96, + "learning_rate": 1.016546348916203e-06, + "loss": 0.551, + "step": 32109 + }, + { + "epoch": 0.96, + "learning_rate": 1.0151754299379534e-06, + "loss": 0.606, + "step": 32110 + }, + { + "epoch": 0.96, + "learning_rate": 1.013805431281145e-06, + "loss": 0.5009, + "step": 32111 + }, + { + "epoch": 0.96, + "learning_rate": 1.0124363529585347e-06, + "loss": 0.6878, + "step": 32112 + }, + { + "epoch": 0.96, + "learning_rate": 1.0110681949828337e-06, + "loss": 0.3687, + "step": 32113 + }, + { + "epoch": 0.96, + "learning_rate": 1.009700957366766e-06, + "loss": 0.5987, + "step": 32114 + }, + { + "epoch": 0.96, + "learning_rate": 1.0083346401230432e-06, + "loss": 0.569, + "step": 32115 + }, + { + "epoch": 0.96, + "learning_rate": 1.0069692432643774e-06, + "loss": 0.6369, + "step": 32116 + }, + { + "epoch": 0.96, + "learning_rate": 1.0056047668034475e-06, + "loss": 0.5993, + "step": 32117 + }, + { + "epoch": 0.96, + "learning_rate": 1.004241210752954e-06, + "loss": 0.5091, + "step": 32118 + }, + { + "epoch": 0.96, + "learning_rate": 1.0028785751255653e-06, + "loss": 0.5853, + "step": 32119 + }, + { + "epoch": 0.96, + "learning_rate": 1.0015168599339708e-06, + "loss": 0.4551, + "step": 32120 + }, + { + "epoch": 0.96, + "learning_rate": 1.0001560651907937e-06, + "loss": 0.619, + "step": 32121 + }, + { + "epoch": 0.96, + "learning_rate": 9.98796190908724e-07, + "loss": 0.6198, + "step": 32122 + }, + { + "epoch": 0.96, + "learning_rate": 9.974372371003849e-07, + "loss": 0.5523, + "step": 32123 + }, + { + "epoch": 0.96, + "learning_rate": 9.960792037784217e-07, + "loss": 0.5048, + "step": 32124 + }, + { + "epoch": 0.96, + "learning_rate": 9.947220909554467e-07, + "loss": 0.4726, + "step": 32125 + }, + { + "epoch": 0.96, + "learning_rate": 9.93365898644083e-07, + "loss": 0.5755, + "step": 32126 + }, + { + "epoch": 0.96, + "learning_rate": 9.92010626856954e-07, + "loss": 0.6156, + "step": 32127 + }, + { + "epoch": 0.96, + "learning_rate": 9.906562756066384e-07, + "loss": 0.6469, + "step": 32128 + }, + { + "epoch": 0.96, + "learning_rate": 9.893028449057484e-07, + "loss": 0.4653, + "step": 32129 + }, + { + "epoch": 0.96, + "learning_rate": 9.879503347668517e-07, + "loss": 0.5693, + "step": 32130 + }, + { + "epoch": 0.96, + "learning_rate": 9.865987452025383e-07, + "loss": 0.5179, + "step": 32131 + }, + { + "epoch": 0.96, + "learning_rate": 9.852480762253646e-07, + "loss": 0.6577, + "step": 32132 + }, + { + "epoch": 0.96, + "learning_rate": 9.838983278478876e-07, + "loss": 0.5187, + "step": 32133 + }, + { + "epoch": 0.96, + "learning_rate": 9.825495000826523e-07, + "loss": 0.5314, + "step": 32134 + }, + { + "epoch": 0.96, + "learning_rate": 9.812015929422158e-07, + "loss": 0.5335, + "step": 32135 + }, + { + "epoch": 0.96, + "learning_rate": 9.798546064391123e-07, + "loss": 0.3958, + "step": 32136 + }, + { + "epoch": 0.96, + "learning_rate": 9.78508540585843e-07, + "loss": 0.5254, + "step": 32137 + }, + { + "epoch": 0.96, + "learning_rate": 9.771633953949423e-07, + "loss": 0.5, + "step": 32138 + }, + { + "epoch": 0.96, + "learning_rate": 9.758191708789e-07, + "loss": 0.5924, + "step": 32139 + }, + { + "epoch": 0.96, + "learning_rate": 9.744758670502286e-07, + "loss": 0.5416, + "step": 32140 + }, + { + "epoch": 0.96, + "learning_rate": 9.73133483921418e-07, + "loss": 0.6318, + "step": 32141 + }, + { + "epoch": 0.96, + "learning_rate": 9.71792021504936e-07, + "loss": 0.5603, + "step": 32142 + }, + { + "epoch": 0.96, + "learning_rate": 9.704514798132725e-07, + "loss": 0.5489, + "step": 32143 + }, + { + "epoch": 0.96, + "learning_rate": 9.691118588588732e-07, + "loss": 0.6114, + "step": 32144 + }, + { + "epoch": 0.96, + "learning_rate": 9.67773158654195e-07, + "loss": 0.5386, + "step": 32145 + }, + { + "epoch": 0.96, + "learning_rate": 9.664353792117054e-07, + "loss": 0.5202, + "step": 32146 + }, + { + "epoch": 0.96, + "learning_rate": 9.650985205438278e-07, + "loss": 0.4357, + "step": 32147 + }, + { + "epoch": 0.96, + "learning_rate": 9.637625826629858e-07, + "loss": 0.6188, + "step": 32148 + }, + { + "epoch": 0.96, + "learning_rate": 9.62427565581614e-07, + "loss": 0.5615, + "step": 32149 + }, + { + "epoch": 0.96, + "learning_rate": 9.610934693121131e-07, + "loss": 0.6086, + "step": 32150 + }, + { + "epoch": 0.96, + "learning_rate": 9.597602938668958e-07, + "loss": 0.4977, + "step": 32151 + }, + { + "epoch": 0.96, + "learning_rate": 9.584280392583522e-07, + "loss": 0.4716, + "step": 32152 + }, + { + "epoch": 0.96, + "learning_rate": 9.570967054988723e-07, + "loss": 0.4411, + "step": 32153 + }, + { + "epoch": 0.96, + "learning_rate": 9.557662926008237e-07, + "loss": 0.561, + "step": 32154 + }, + { + "epoch": 0.96, + "learning_rate": 9.54436800576597e-07, + "loss": 0.5158, + "step": 32155 + }, + { + "epoch": 0.96, + "learning_rate": 9.531082294385374e-07, + "loss": 0.5065, + "step": 32156 + }, + { + "epoch": 0.96, + "learning_rate": 9.51780579199002e-07, + "loss": 0.4981, + "step": 32157 + }, + { + "epoch": 0.96, + "learning_rate": 9.504538498703252e-07, + "loss": 0.5044, + "step": 32158 + }, + { + "epoch": 0.96, + "learning_rate": 9.49128041464864e-07, + "loss": 0.619, + "step": 32159 + }, + { + "epoch": 0.96, + "learning_rate": 9.478031539949306e-07, + "loss": 0.4678, + "step": 32160 + }, + { + "epoch": 0.96, + "learning_rate": 9.464791874728374e-07, + "loss": 0.4886, + "step": 32161 + }, + { + "epoch": 0.96, + "learning_rate": 9.451561419109078e-07, + "loss": 0.5268, + "step": 32162 + }, + { + "epoch": 0.96, + "learning_rate": 9.438340173214322e-07, + "loss": 0.5209, + "step": 32163 + }, + { + "epoch": 0.96, + "learning_rate": 9.425128137167116e-07, + "loss": 0.5253, + "step": 32164 + }, + { + "epoch": 0.96, + "learning_rate": 9.411925311090253e-07, + "loss": 0.5174, + "step": 32165 + }, + { + "epoch": 0.96, + "learning_rate": 9.398731695106522e-07, + "loss": 0.4669, + "step": 32166 + }, + { + "epoch": 0.96, + "learning_rate": 9.385547289338603e-07, + "loss": 0.5646, + "step": 32167 + }, + { + "epoch": 0.96, + "learning_rate": 9.372372093908954e-07, + "loss": 0.5083, + "step": 32168 + }, + { + "epoch": 0.96, + "learning_rate": 9.359206108940144e-07, + "loss": 0.5651, + "step": 32169 + }, + { + "epoch": 0.96, + "learning_rate": 9.346049334554741e-07, + "loss": 0.5788, + "step": 32170 + }, + { + "epoch": 0.96, + "learning_rate": 9.332901770874869e-07, + "loss": 0.5412, + "step": 32171 + }, + { + "epoch": 0.96, + "learning_rate": 9.319763418022764e-07, + "loss": 0.5608, + "step": 32172 + }, + { + "epoch": 0.96, + "learning_rate": 9.306634276120663e-07, + "loss": 0.4717, + "step": 32173 + }, + { + "epoch": 0.96, + "learning_rate": 9.293514345290799e-07, + "loss": 0.5911, + "step": 32174 + }, + { + "epoch": 0.96, + "learning_rate": 9.280403625654855e-07, + "loss": 0.6135, + "step": 32175 + }, + { + "epoch": 0.96, + "learning_rate": 9.267302117334842e-07, + "loss": 0.5542, + "step": 32176 + }, + { + "epoch": 0.96, + "learning_rate": 9.254209820452553e-07, + "loss": 0.517, + "step": 32177 + }, + { + "epoch": 0.96, + "learning_rate": 9.24112673512989e-07, + "loss": 0.5982, + "step": 32178 + }, + { + "epoch": 0.96, + "learning_rate": 9.228052861488201e-07, + "loss": 0.5394, + "step": 32179 + }, + { + "epoch": 0.96, + "learning_rate": 9.214988199649277e-07, + "loss": 0.4607, + "step": 32180 + }, + { + "epoch": 0.96, + "learning_rate": 9.201932749734465e-07, + "loss": 0.5391, + "step": 32181 + }, + { + "epoch": 0.96, + "learning_rate": 9.188886511865225e-07, + "loss": 0.4769, + "step": 32182 + }, + { + "epoch": 0.96, + "learning_rate": 9.175849486162791e-07, + "loss": 0.5021, + "step": 32183 + }, + { + "epoch": 0.96, + "learning_rate": 9.162821672748401e-07, + "loss": 0.4393, + "step": 32184 + }, + { + "epoch": 0.96, + "learning_rate": 9.14980307174318e-07, + "loss": 0.5095, + "step": 32185 + }, + { + "epoch": 0.96, + "learning_rate": 9.136793683268252e-07, + "loss": 0.4523, + "step": 32186 + }, + { + "epoch": 0.96, + "learning_rate": 9.123793507444412e-07, + "loss": 0.6924, + "step": 32187 + }, + { + "epoch": 0.96, + "learning_rate": 9.110802544392672e-07, + "loss": 0.5729, + "step": 32188 + }, + { + "epoch": 0.96, + "learning_rate": 9.097820794233714e-07, + "loss": 0.5687, + "step": 32189 + }, + { + "epoch": 0.96, + "learning_rate": 9.084848257088329e-07, + "loss": 0.5217, + "step": 32190 + }, + { + "epoch": 0.96, + "learning_rate": 9.07188493307709e-07, + "loss": 0.5105, + "step": 32191 + }, + { + "epoch": 0.96, + "learning_rate": 9.058930822320455e-07, + "loss": 0.5377, + "step": 32192 + }, + { + "epoch": 0.96, + "learning_rate": 9.045985924938994e-07, + "loss": 0.5819, + "step": 32193 + }, + { + "epoch": 0.96, + "learning_rate": 9.033050241053054e-07, + "loss": 0.4927, + "step": 32194 + }, + { + "epoch": 0.96, + "learning_rate": 9.020123770782762e-07, + "loss": 0.4518, + "step": 32195 + }, + { + "epoch": 0.96, + "learning_rate": 9.007206514248357e-07, + "loss": 0.4467, + "step": 32196 + }, + { + "epoch": 0.96, + "learning_rate": 8.994298471569962e-07, + "loss": 0.478, + "step": 32197 + }, + { + "epoch": 0.96, + "learning_rate": 8.981399642867705e-07, + "loss": 0.5362, + "step": 32198 + }, + { + "epoch": 0.96, + "learning_rate": 8.968510028261379e-07, + "loss": 0.4229, + "step": 32199 + }, + { + "epoch": 0.96, + "learning_rate": 8.955629627870887e-07, + "loss": 0.5684, + "step": 32200 + }, + { + "epoch": 0.96, + "learning_rate": 8.942758441815913e-07, + "loss": 0.5443, + "step": 32201 + }, + { + "epoch": 0.96, + "learning_rate": 8.929896470216137e-07, + "loss": 0.5325, + "step": 32202 + }, + { + "epoch": 0.96, + "learning_rate": 8.917043713191353e-07, + "loss": 0.3998, + "step": 32203 + }, + { + "epoch": 0.96, + "learning_rate": 8.904200170860688e-07, + "loss": 0.5843, + "step": 32204 + }, + { + "epoch": 0.96, + "learning_rate": 8.891365843343824e-07, + "loss": 0.6588, + "step": 32205 + }, + { + "epoch": 0.96, + "learning_rate": 8.87854073076011e-07, + "loss": 0.4529, + "step": 32206 + }, + { + "epoch": 0.96, + "learning_rate": 8.865724833228562e-07, + "loss": 0.5414, + "step": 32207 + }, + { + "epoch": 0.96, + "learning_rate": 8.852918150868527e-07, + "loss": 0.54, + "step": 32208 + }, + { + "epoch": 0.96, + "learning_rate": 8.840120683799136e-07, + "loss": 0.5301, + "step": 32209 + }, + { + "epoch": 0.96, + "learning_rate": 8.827332432139179e-07, + "loss": 0.5263, + "step": 32210 + }, + { + "epoch": 0.96, + "learning_rate": 8.814553396007564e-07, + "loss": 0.5665, + "step": 32211 + }, + { + "epoch": 0.96, + "learning_rate": 8.801783575523304e-07, + "loss": 0.485, + "step": 32212 + }, + { + "epoch": 0.96, + "learning_rate": 8.789022970804972e-07, + "loss": 0.5765, + "step": 32213 + }, + { + "epoch": 0.96, + "learning_rate": 8.776271581971251e-07, + "loss": 0.4707, + "step": 32214 + }, + { + "epoch": 0.96, + "learning_rate": 8.763529409140603e-07, + "loss": 0.4311, + "step": 32215 + }, + { + "epoch": 0.96, + "learning_rate": 8.750796452431708e-07, + "loss": 0.5348, + "step": 32216 + }, + { + "epoch": 0.96, + "learning_rate": 8.738072711962808e-07, + "loss": 0.4856, + "step": 32217 + }, + { + "epoch": 0.96, + "learning_rate": 8.725358187852251e-07, + "loss": 0.6344, + "step": 32218 + }, + { + "epoch": 0.96, + "learning_rate": 8.712652880218164e-07, + "loss": 0.5777, + "step": 32219 + }, + { + "epoch": 0.96, + "learning_rate": 8.699956789178787e-07, + "loss": 0.5489, + "step": 32220 + }, + { + "epoch": 0.96, + "learning_rate": 8.687269914852136e-07, + "loss": 0.5578, + "step": 32221 + }, + { + "epoch": 0.96, + "learning_rate": 8.674592257356118e-07, + "loss": 0.434, + "step": 32222 + }, + { + "epoch": 0.96, + "learning_rate": 8.661923816808637e-07, + "loss": 0.5056, + "step": 32223 + }, + { + "epoch": 0.96, + "learning_rate": 8.649264593327489e-07, + "loss": 0.4957, + "step": 32224 + }, + { + "epoch": 0.96, + "learning_rate": 8.636614587030356e-07, + "loss": 0.6816, + "step": 32225 + }, + { + "epoch": 0.96, + "learning_rate": 8.623973798034813e-07, + "loss": 0.5099, + "step": 32226 + }, + { + "epoch": 0.96, + "learning_rate": 8.611342226458541e-07, + "loss": 0.4232, + "step": 32227 + }, + { + "epoch": 0.96, + "learning_rate": 8.598719872418781e-07, + "loss": 0.6472, + "step": 32228 + }, + { + "epoch": 0.96, + "learning_rate": 8.586106736033106e-07, + "loss": 0.5035, + "step": 32229 + }, + { + "epoch": 0.96, + "learning_rate": 8.573502817418644e-07, + "loss": 0.4916, + "step": 32230 + }, + { + "epoch": 0.96, + "learning_rate": 8.560908116692523e-07, + "loss": 0.5147, + "step": 32231 + }, + { + "epoch": 0.96, + "learning_rate": 8.548322633971983e-07, + "loss": 0.4939, + "step": 32232 + }, + { + "epoch": 0.96, + "learning_rate": 8.535746369374042e-07, + "loss": 0.4857, + "step": 32233 + }, + { + "epoch": 0.96, + "learning_rate": 8.523179323015496e-07, + "loss": 0.531, + "step": 32234 + }, + { + "epoch": 0.96, + "learning_rate": 8.510621495013249e-07, + "loss": 0.5853, + "step": 32235 + }, + { + "epoch": 0.96, + "learning_rate": 8.498072885484099e-07, + "loss": 0.5394, + "step": 32236 + }, + { + "epoch": 0.96, + "learning_rate": 8.485533494544729e-07, + "loss": 0.4449, + "step": 32237 + }, + { + "epoch": 0.96, + "learning_rate": 8.473003322311601e-07, + "loss": 0.5335, + "step": 32238 + }, + { + "epoch": 0.96, + "learning_rate": 8.460482368901401e-07, + "loss": 0.4723, + "step": 32239 + }, + { + "epoch": 0.96, + "learning_rate": 8.447970634430369e-07, + "loss": 0.5376, + "step": 32240 + }, + { + "epoch": 0.96, + "learning_rate": 8.435468119014967e-07, + "loss": 0.4026, + "step": 32241 + }, + { + "epoch": 0.96, + "learning_rate": 8.422974822771434e-07, + "loss": 0.6731, + "step": 32242 + }, + { + "epoch": 0.96, + "learning_rate": 8.410490745815791e-07, + "loss": 0.5914, + "step": 32243 + }, + { + "epoch": 0.96, + "learning_rate": 8.398015888264276e-07, + "loss": 0.4702, + "step": 32244 + }, + { + "epoch": 0.96, + "learning_rate": 8.385550250232688e-07, + "loss": 0.4906, + "step": 32245 + }, + { + "epoch": 0.96, + "learning_rate": 8.373093831837154e-07, + "loss": 0.5105, + "step": 32246 + }, + { + "epoch": 0.96, + "learning_rate": 8.360646633193247e-07, + "loss": 0.5825, + "step": 32247 + }, + { + "epoch": 0.96, + "learning_rate": 8.348208654416878e-07, + "loss": 0.4975, + "step": 32248 + }, + { + "epoch": 0.96, + "learning_rate": 8.335779895623618e-07, + "loss": 0.5636, + "step": 32249 + }, + { + "epoch": 0.96, + "learning_rate": 8.323360356929044e-07, + "loss": 0.5164, + "step": 32250 + }, + { + "epoch": 0.96, + "learning_rate": 8.310950038448617e-07, + "loss": 0.4029, + "step": 32251 + }, + { + "epoch": 0.96, + "learning_rate": 8.298548940297801e-07, + "loss": 0.5017, + "step": 32252 + }, + { + "epoch": 0.96, + "learning_rate": 8.286157062591726e-07, + "loss": 0.57, + "step": 32253 + }, + { + "epoch": 0.96, + "learning_rate": 8.273774405445744e-07, + "loss": 0.5275, + "step": 32254 + }, + { + "epoch": 0.96, + "learning_rate": 8.261400968974986e-07, + "loss": 0.6157, + "step": 32255 + }, + { + "epoch": 0.96, + "learning_rate": 8.24903675329447e-07, + "loss": 0.5109, + "step": 32256 + }, + { + "epoch": 0.96, + "learning_rate": 8.236681758518994e-07, + "loss": 0.5767, + "step": 32257 + }, + { + "epoch": 0.96, + "learning_rate": 8.224335984763686e-07, + "loss": 0.4173, + "step": 32258 + }, + { + "epoch": 0.96, + "learning_rate": 8.211999432143124e-07, + "loss": 0.5251, + "step": 32259 + }, + { + "epoch": 0.96, + "learning_rate": 8.199672100772215e-07, + "loss": 0.4964, + "step": 32260 + }, + { + "epoch": 0.96, + "learning_rate": 8.187353990765423e-07, + "loss": 0.5245, + "step": 32261 + }, + { + "epoch": 0.96, + "learning_rate": 8.175045102237322e-07, + "loss": 0.5333, + "step": 32262 + }, + { + "epoch": 0.96, + "learning_rate": 8.162745435302265e-07, + "loss": 0.4589, + "step": 32263 + }, + { + "epoch": 0.96, + "learning_rate": 8.150454990074829e-07, + "loss": 0.5049, + "step": 32264 + }, + { + "epoch": 0.96, + "learning_rate": 8.138173766669143e-07, + "loss": 0.4949, + "step": 32265 + }, + { + "epoch": 0.96, + "learning_rate": 8.125901765199339e-07, + "loss": 0.4974, + "step": 32266 + }, + { + "epoch": 0.96, + "learning_rate": 8.113638985779548e-07, + "loss": 0.522, + "step": 32267 + }, + { + "epoch": 0.96, + "learning_rate": 8.101385428523789e-07, + "loss": 0.4632, + "step": 32268 + }, + { + "epoch": 0.96, + "learning_rate": 8.089141093546082e-07, + "loss": 0.5026, + "step": 32269 + }, + { + "epoch": 0.96, + "learning_rate": 8.076905980960225e-07, + "loss": 0.5534, + "step": 32270 + }, + { + "epoch": 0.96, + "learning_rate": 8.064680090879905e-07, + "loss": 0.5821, + "step": 32271 + }, + { + "epoch": 0.96, + "learning_rate": 8.052463423418921e-07, + "loss": 0.5867, + "step": 32272 + }, + { + "epoch": 0.96, + "learning_rate": 8.040255978690736e-07, + "loss": 0.5754, + "step": 32273 + }, + { + "epoch": 0.96, + "learning_rate": 8.028057756808927e-07, + "loss": 0.5843, + "step": 32274 + }, + { + "epoch": 0.96, + "learning_rate": 8.015868757886847e-07, + "loss": 0.4694, + "step": 32275 + }, + { + "epoch": 0.96, + "learning_rate": 8.003688982037849e-07, + "loss": 0.4131, + "step": 32276 + }, + { + "epoch": 0.96, + "learning_rate": 7.991518429375067e-07, + "loss": 0.408, + "step": 32277 + }, + { + "epoch": 0.96, + "learning_rate": 7.979357100011853e-07, + "loss": 0.5128, + "step": 32278 + }, + { + "epoch": 0.96, + "learning_rate": 7.967204994061339e-07, + "loss": 0.4792, + "step": 32279 + }, + { + "epoch": 0.96, + "learning_rate": 7.955062111636214e-07, + "loss": 0.4385, + "step": 32280 + }, + { + "epoch": 0.96, + "learning_rate": 7.942928452849608e-07, + "loss": 0.5638, + "step": 32281 + }, + { + "epoch": 0.96, + "learning_rate": 7.930804017814097e-07, + "loss": 0.6058, + "step": 32282 + }, + { + "epoch": 0.96, + "learning_rate": 7.918688806642815e-07, + "loss": 0.49, + "step": 32283 + }, + { + "epoch": 0.96, + "learning_rate": 7.906582819448005e-07, + "loss": 0.5124, + "step": 32284 + }, + { + "epoch": 0.96, + "learning_rate": 7.894486056342354e-07, + "loss": 0.6111, + "step": 32285 + }, + { + "epoch": 0.96, + "learning_rate": 7.882398517438439e-07, + "loss": 0.5616, + "step": 32286 + }, + { + "epoch": 0.96, + "learning_rate": 7.870320202848502e-07, + "loss": 0.5629, + "step": 32287 + }, + { + "epoch": 0.96, + "learning_rate": 7.858251112685011e-07, + "loss": 0.5231, + "step": 32288 + }, + { + "epoch": 0.96, + "learning_rate": 7.846191247059987e-07, + "loss": 0.4699, + "step": 32289 + }, + { + "epoch": 0.96, + "learning_rate": 7.834140606085672e-07, + "loss": 0.5247, + "step": 32290 + }, + { + "epoch": 0.96, + "learning_rate": 7.822099189874088e-07, + "loss": 0.427, + "step": 32291 + }, + { + "epoch": 0.96, + "learning_rate": 7.810066998537146e-07, + "loss": 0.5882, + "step": 32292 + }, + { + "epoch": 0.96, + "learning_rate": 7.798044032186646e-07, + "loss": 0.7406, + "step": 32293 + }, + { + "epoch": 0.96, + "learning_rate": 7.78603029093472e-07, + "loss": 0.6257, + "step": 32294 + }, + { + "epoch": 0.96, + "learning_rate": 7.774025774892724e-07, + "loss": 0.5736, + "step": 32295 + }, + { + "epoch": 0.96, + "learning_rate": 7.762030484172345e-07, + "loss": 0.4897, + "step": 32296 + }, + { + "epoch": 0.96, + "learning_rate": 7.75004441888505e-07, + "loss": 0.4176, + "step": 32297 + }, + { + "epoch": 0.96, + "learning_rate": 7.738067579142527e-07, + "loss": 0.5839, + "step": 32298 + }, + { + "epoch": 0.96, + "learning_rate": 7.726099965055911e-07, + "loss": 0.5381, + "step": 32299 + }, + { + "epoch": 0.96, + "learning_rate": 7.714141576736555e-07, + "loss": 0.4791, + "step": 32300 + }, + { + "epoch": 0.96, + "learning_rate": 7.702192414295595e-07, + "loss": 0.6025, + "step": 32301 + }, + { + "epoch": 0.96, + "learning_rate": 7.690252477844162e-07, + "loss": 0.4511, + "step": 32302 + }, + { + "epoch": 0.96, + "learning_rate": 7.678321767493279e-07, + "loss": 0.4272, + "step": 32303 + }, + { + "epoch": 0.96, + "learning_rate": 7.666400283353969e-07, + "loss": 0.5656, + "step": 32304 + }, + { + "epoch": 0.96, + "learning_rate": 7.65448802553681e-07, + "loss": 0.4577, + "step": 32305 + }, + { + "epoch": 0.96, + "learning_rate": 7.642584994152713e-07, + "loss": 0.504, + "step": 32306 + }, + { + "epoch": 0.96, + "learning_rate": 7.630691189312478e-07, + "loss": 0.5024, + "step": 32307 + }, + { + "epoch": 0.96, + "learning_rate": 7.618806611126461e-07, + "loss": 0.5029, + "step": 32308 + }, + { + "epoch": 0.96, + "learning_rate": 7.606931259705241e-07, + "loss": 0.5814, + "step": 32309 + }, + { + "epoch": 0.96, + "learning_rate": 7.595065135159285e-07, + "loss": 0.5206, + "step": 32310 + }, + { + "epoch": 0.96, + "learning_rate": 7.583208237598838e-07, + "loss": 0.4811, + "step": 32311 + }, + { + "epoch": 0.96, + "learning_rate": 7.571360567134145e-07, + "loss": 0.4255, + "step": 32312 + }, + { + "epoch": 0.96, + "learning_rate": 7.559522123875562e-07, + "loss": 0.4626, + "step": 32313 + }, + { + "epoch": 0.96, + "learning_rate": 7.547692907932891e-07, + "loss": 0.5591, + "step": 32314 + }, + { + "epoch": 0.96, + "learning_rate": 7.535872919416264e-07, + "loss": 0.5017, + "step": 32315 + }, + { + "epoch": 0.96, + "learning_rate": 7.524062158435374e-07, + "loss": 0.4992, + "step": 32316 + }, + { + "epoch": 0.96, + "learning_rate": 7.512260625100354e-07, + "loss": 0.5617, + "step": 32317 + }, + { + "epoch": 0.96, + "learning_rate": 7.500468319520781e-07, + "loss": 0.4506, + "step": 32318 + }, + { + "epoch": 0.96, + "learning_rate": 7.488685241806237e-07, + "loss": 0.4895, + "step": 32319 + }, + { + "epoch": 0.96, + "learning_rate": 7.476911392066299e-07, + "loss": 0.517, + "step": 32320 + }, + { + "epoch": 0.96, + "learning_rate": 7.465146770410436e-07, + "loss": 0.4296, + "step": 32321 + }, + { + "epoch": 0.96, + "learning_rate": 7.453391376948116e-07, + "loss": 0.549, + "step": 32322 + }, + { + "epoch": 0.96, + "learning_rate": 7.441645211788584e-07, + "loss": 0.5543, + "step": 32323 + }, + { + "epoch": 0.96, + "learning_rate": 7.429908275040976e-07, + "loss": 0.5374, + "step": 32324 + }, + { + "epoch": 0.96, + "learning_rate": 7.418180566814426e-07, + "loss": 0.6173, + "step": 32325 + }, + { + "epoch": 0.96, + "learning_rate": 7.40646208721818e-07, + "loss": 0.4671, + "step": 32326 + }, + { + "epoch": 0.96, + "learning_rate": 7.394752836360929e-07, + "loss": 0.5708, + "step": 32327 + }, + { + "epoch": 0.96, + "learning_rate": 7.383052814351698e-07, + "loss": 0.5251, + "step": 32328 + }, + { + "epoch": 0.96, + "learning_rate": 7.371362021299177e-07, + "loss": 0.5482, + "step": 32329 + }, + { + "epoch": 0.96, + "learning_rate": 7.359680457312279e-07, + "loss": 0.5402, + "step": 32330 + }, + { + "epoch": 0.96, + "learning_rate": 7.34800812249925e-07, + "loss": 0.5312, + "step": 32331 + }, + { + "epoch": 0.96, + "learning_rate": 7.336345016968893e-07, + "loss": 0.4649, + "step": 32332 + }, + { + "epoch": 0.96, + "learning_rate": 7.324691140829565e-07, + "loss": 0.5643, + "step": 32333 + }, + { + "epoch": 0.96, + "learning_rate": 7.313046494189624e-07, + "loss": 0.647, + "step": 32334 + }, + { + "epoch": 0.96, + "learning_rate": 7.301411077157205e-07, + "loss": 0.7683, + "step": 32335 + }, + { + "epoch": 0.96, + "learning_rate": 7.289784889840778e-07, + "loss": 0.435, + "step": 32336 + }, + { + "epoch": 0.96, + "learning_rate": 7.278167932348257e-07, + "loss": 0.6397, + "step": 32337 + }, + { + "epoch": 0.96, + "learning_rate": 7.266560204787554e-07, + "loss": 0.4667, + "step": 32338 + }, + { + "epoch": 0.96, + "learning_rate": 7.254961707266805e-07, + "loss": 0.4566, + "step": 32339 + }, + { + "epoch": 0.96, + "learning_rate": 7.243372439893703e-07, + "loss": 0.4887, + "step": 32340 + }, + { + "epoch": 0.96, + "learning_rate": 7.231792402776161e-07, + "loss": 0.5852, + "step": 32341 + }, + { + "epoch": 0.96, + "learning_rate": 7.220221596021648e-07, + "loss": 0.5281, + "step": 32342 + }, + { + "epoch": 0.96, + "learning_rate": 7.208660019737856e-07, + "loss": 0.5072, + "step": 32343 + }, + { + "epoch": 0.96, + "learning_rate": 7.197107674032144e-07, + "loss": 0.5217, + "step": 32344 + }, + { + "epoch": 0.96, + "learning_rate": 7.185564559012203e-07, + "loss": 0.6333, + "step": 32345 + }, + { + "epoch": 0.96, + "learning_rate": 7.17403067478506e-07, + "loss": 0.544, + "step": 32346 + }, + { + "epoch": 0.96, + "learning_rate": 7.162506021458182e-07, + "loss": 0.5365, + "step": 32347 + }, + { + "epoch": 0.96, + "learning_rate": 7.150990599138485e-07, + "loss": 0.5928, + "step": 32348 + }, + { + "epoch": 0.96, + "learning_rate": 7.139484407933328e-07, + "loss": 0.4504, + "step": 32349 + }, + { + "epoch": 0.96, + "learning_rate": 7.127987447949403e-07, + "loss": 0.5525, + "step": 32350 + }, + { + "epoch": 0.96, + "learning_rate": 7.116499719293846e-07, + "loss": 0.4245, + "step": 32351 + }, + { + "epoch": 0.96, + "learning_rate": 7.105021222073238e-07, + "loss": 0.6352, + "step": 32352 + }, + { + "epoch": 0.96, + "learning_rate": 7.093551956394496e-07, + "loss": 0.652, + "step": 32353 + }, + { + "epoch": 0.96, + "learning_rate": 7.082091922364087e-07, + "loss": 0.5092, + "step": 32354 + }, + { + "epoch": 0.96, + "learning_rate": 7.070641120088595e-07, + "loss": 0.5135, + "step": 32355 + }, + { + "epoch": 0.96, + "learning_rate": 7.059199549674711e-07, + "loss": 0.559, + "step": 32356 + }, + { + "epoch": 0.96, + "learning_rate": 7.047767211228573e-07, + "loss": 0.4688, + "step": 32357 + }, + { + "epoch": 0.96, + "learning_rate": 7.036344104856541e-07, + "loss": 0.4274, + "step": 32358 + }, + { + "epoch": 0.96, + "learning_rate": 7.02493023066475e-07, + "loss": 0.5172, + "step": 32359 + }, + { + "epoch": 0.96, + "learning_rate": 7.013525588759562e-07, + "loss": 0.5942, + "step": 32360 + }, + { + "epoch": 0.96, + "learning_rate": 7.00213017924678e-07, + "loss": 0.4727, + "step": 32361 + }, + { + "epoch": 0.96, + "learning_rate": 6.990744002232319e-07, + "loss": 0.4197, + "step": 32362 + }, + { + "epoch": 0.96, + "learning_rate": 6.979367057822206e-07, + "loss": 0.6458, + "step": 32363 + }, + { + "epoch": 0.96, + "learning_rate": 6.967999346122245e-07, + "loss": 0.6293, + "step": 32364 + }, + { + "epoch": 0.96, + "learning_rate": 6.956640867238017e-07, + "loss": 0.5575, + "step": 32365 + }, + { + "epoch": 0.96, + "learning_rate": 6.945291621275107e-07, + "loss": 0.3714, + "step": 32366 + }, + { + "epoch": 0.96, + "learning_rate": 6.933951608339096e-07, + "loss": 0.4898, + "step": 32367 + }, + { + "epoch": 0.96, + "learning_rate": 6.922620828535343e-07, + "loss": 0.492, + "step": 32368 + }, + { + "epoch": 0.96, + "learning_rate": 6.911299281969319e-07, + "loss": 0.5166, + "step": 32369 + }, + { + "epoch": 0.96, + "learning_rate": 6.899986968746275e-07, + "loss": 0.5668, + "step": 32370 + }, + { + "epoch": 0.96, + "learning_rate": 6.888683888971237e-07, + "loss": 0.495, + "step": 32371 + }, + { + "epoch": 0.96, + "learning_rate": 6.877390042749454e-07, + "loss": 0.5479, + "step": 32372 + }, + { + "epoch": 0.96, + "learning_rate": 6.866105430185954e-07, + "loss": 0.5101, + "step": 32373 + }, + { + "epoch": 0.96, + "learning_rate": 6.854830051385431e-07, + "loss": 0.5016, + "step": 32374 + }, + { + "epoch": 0.96, + "learning_rate": 6.843563906453022e-07, + "loss": 0.5779, + "step": 32375 + }, + { + "epoch": 0.96, + "learning_rate": 6.832306995493199e-07, + "loss": 0.5821, + "step": 32376 + }, + { + "epoch": 0.96, + "learning_rate": 6.821059318610879e-07, + "loss": 0.5322, + "step": 32377 + }, + { + "epoch": 0.96, + "learning_rate": 6.809820875910311e-07, + "loss": 0.5339, + "step": 32378 + }, + { + "epoch": 0.96, + "learning_rate": 6.798591667496413e-07, + "loss": 0.4476, + "step": 32379 + }, + { + "epoch": 0.96, + "learning_rate": 6.78737169347321e-07, + "loss": 0.5604, + "step": 32380 + }, + { + "epoch": 0.96, + "learning_rate": 6.776160953945177e-07, + "loss": 0.5328, + "step": 32381 + }, + { + "epoch": 0.96, + "learning_rate": 6.764959449016561e-07, + "loss": 0.4791, + "step": 32382 + }, + { + "epoch": 0.96, + "learning_rate": 6.753767178791504e-07, + "loss": 0.4528, + "step": 32383 + }, + { + "epoch": 0.96, + "learning_rate": 6.742584143374032e-07, + "loss": 0.6269, + "step": 32384 + }, + { + "epoch": 0.96, + "learning_rate": 6.731410342868172e-07, + "loss": 0.6248, + "step": 32385 + }, + { + "epoch": 0.96, + "learning_rate": 6.720245777377732e-07, + "loss": 0.5162, + "step": 32386 + }, + { + "epoch": 0.96, + "learning_rate": 6.709090447006517e-07, + "loss": 0.6847, + "step": 32387 + }, + { + "epoch": 0.96, + "learning_rate": 6.697944351858443e-07, + "loss": 0.5893, + "step": 32388 + }, + { + "epoch": 0.96, + "learning_rate": 6.686807492036873e-07, + "loss": 0.5019, + "step": 32389 + }, + { + "epoch": 0.96, + "learning_rate": 6.675679867645501e-07, + "loss": 0.5135, + "step": 32390 + }, + { + "epoch": 0.96, + "learning_rate": 6.664561478787801e-07, + "loss": 0.5474, + "step": 32391 + }, + { + "epoch": 0.96, + "learning_rate": 6.653452325567022e-07, + "loss": 0.531, + "step": 32392 + }, + { + "epoch": 0.96, + "learning_rate": 6.642352408086638e-07, + "loss": 0.577, + "step": 32393 + }, + { + "epoch": 0.96, + "learning_rate": 6.631261726449678e-07, + "loss": 0.5383, + "step": 32394 + }, + { + "epoch": 0.96, + "learning_rate": 6.620180280759281e-07, + "loss": 0.339, + "step": 32395 + }, + { + "epoch": 0.96, + "learning_rate": 6.609108071118697e-07, + "loss": 0.4875, + "step": 32396 + }, + { + "epoch": 0.96, + "learning_rate": 6.598045097630512e-07, + "loss": 0.522, + "step": 32397 + }, + { + "epoch": 0.96, + "learning_rate": 6.586991360397865e-07, + "loss": 0.5351, + "step": 32398 + }, + { + "epoch": 0.96, + "learning_rate": 6.575946859523451e-07, + "loss": 0.5245, + "step": 32399 + }, + { + "epoch": 0.96, + "learning_rate": 6.564911595109969e-07, + "loss": 0.4732, + "step": 32400 + }, + { + "epoch": 0.96, + "learning_rate": 6.553885567259888e-07, + "loss": 0.5088, + "step": 32401 + }, + { + "epoch": 0.96, + "learning_rate": 6.542868776075795e-07, + "loss": 0.4742, + "step": 32402 + }, + { + "epoch": 0.96, + "learning_rate": 6.531861221660274e-07, + "loss": 0.5274, + "step": 32403 + }, + { + "epoch": 0.96, + "learning_rate": 6.520862904115576e-07, + "loss": 0.5079, + "step": 32404 + }, + { + "epoch": 0.96, + "learning_rate": 6.509873823543844e-07, + "loss": 0.4421, + "step": 32405 + }, + { + "epoch": 0.96, + "learning_rate": 6.498893980047327e-07, + "loss": 0.4794, + "step": 32406 + }, + { + "epoch": 0.96, + "learning_rate": 6.487923373728278e-07, + "loss": 0.6681, + "step": 32407 + }, + { + "epoch": 0.96, + "learning_rate": 6.476962004688391e-07, + "loss": 0.5577, + "step": 32408 + }, + { + "epoch": 0.96, + "learning_rate": 6.46600987302981e-07, + "loss": 0.5521, + "step": 32409 + }, + { + "epoch": 0.96, + "learning_rate": 6.45506697885423e-07, + "loss": 0.6077, + "step": 32410 + }, + { + "epoch": 0.96, + "learning_rate": 6.444133322263457e-07, + "loss": 0.5411, + "step": 32411 + }, + { + "epoch": 0.96, + "learning_rate": 6.433208903359078e-07, + "loss": 0.5297, + "step": 32412 + }, + { + "epoch": 0.96, + "learning_rate": 6.4222937222429e-07, + "loss": 0.568, + "step": 32413 + }, + { + "epoch": 0.97, + "learning_rate": 6.411387779016065e-07, + "loss": 0.5734, + "step": 32414 + }, + { + "epoch": 0.97, + "learning_rate": 6.400491073780268e-07, + "loss": 0.5813, + "step": 32415 + }, + { + "epoch": 0.97, + "learning_rate": 6.389603606636652e-07, + "loss": 0.5998, + "step": 32416 + }, + { + "epoch": 0.97, + "learning_rate": 6.378725377686357e-07, + "loss": 0.6015, + "step": 32417 + }, + { + "epoch": 0.97, + "learning_rate": 6.367856387030857e-07, + "loss": 0.5154, + "step": 32418 + }, + { + "epoch": 0.97, + "learning_rate": 6.356996634770962e-07, + "loss": 0.649, + "step": 32419 + }, + { + "epoch": 0.97, + "learning_rate": 6.34614612100759e-07, + "loss": 0.5086, + "step": 32420 + }, + { + "epoch": 0.97, + "learning_rate": 6.335304845841772e-07, + "loss": 0.5231, + "step": 32421 + }, + { + "epoch": 0.97, + "learning_rate": 6.324472809374204e-07, + "loss": 0.5468, + "step": 32422 + }, + { + "epoch": 0.97, + "learning_rate": 6.313650011705807e-07, + "loss": 0.4782, + "step": 32423 + }, + { + "epoch": 0.97, + "learning_rate": 6.302836452936834e-07, + "loss": 0.5791, + "step": 32424 + }, + { + "epoch": 0.97, + "learning_rate": 6.292032133168202e-07, + "loss": 0.5126, + "step": 32425 + }, + { + "epoch": 0.97, + "learning_rate": 6.281237052500056e-07, + "loss": 0.4805, + "step": 32426 + }, + { + "epoch": 0.97, + "learning_rate": 6.270451211033091e-07, + "loss": 0.5314, + "step": 32427 + }, + { + "epoch": 0.97, + "learning_rate": 6.259674608867338e-07, + "loss": 0.5735, + "step": 32428 + }, + { + "epoch": 0.97, + "learning_rate": 6.248907246102942e-07, + "loss": 0.5391, + "step": 32429 + }, + { + "epoch": 0.97, + "learning_rate": 6.238149122840264e-07, + "loss": 0.5272, + "step": 32430 + }, + { + "epoch": 0.97, + "learning_rate": 6.227400239179226e-07, + "loss": 0.4987, + "step": 32431 + }, + { + "epoch": 0.97, + "learning_rate": 6.216660595219748e-07, + "loss": 0.4834, + "step": 32432 + }, + { + "epoch": 0.97, + "learning_rate": 6.205930191061526e-07, + "loss": 0.4164, + "step": 32433 + }, + { + "epoch": 0.97, + "learning_rate": 6.195209026804593e-07, + "loss": 0.4953, + "step": 32434 + }, + { + "epoch": 0.97, + "learning_rate": 6.184497102548537e-07, + "loss": 0.5522, + "step": 32435 + }, + { + "epoch": 0.97, + "learning_rate": 6.173794418392831e-07, + "loss": 0.5958, + "step": 32436 + }, + { + "epoch": 0.97, + "learning_rate": 6.163100974437175e-07, + "loss": 0.5582, + "step": 32437 + }, + { + "epoch": 0.97, + "learning_rate": 6.152416770780822e-07, + "loss": 0.5507, + "step": 32438 + }, + { + "epoch": 0.97, + "learning_rate": 6.141741807523249e-07, + "loss": 0.5625, + "step": 32439 + }, + { + "epoch": 0.97, + "learning_rate": 6.131076084763599e-07, + "loss": 0.4981, + "step": 32440 + }, + { + "epoch": 0.97, + "learning_rate": 6.120419602601235e-07, + "loss": 0.4969, + "step": 32441 + }, + { + "epoch": 0.97, + "learning_rate": 6.109772361134969e-07, + "loss": 0.6491, + "step": 32442 + }, + { + "epoch": 0.97, + "learning_rate": 6.099134360463832e-07, + "loss": 0.5746, + "step": 32443 + }, + { + "epoch": 0.97, + "learning_rate": 6.088505600686967e-07, + "loss": 0.5023, + "step": 32444 + }, + { + "epoch": 0.97, + "learning_rate": 6.077886081902961e-07, + "loss": 0.6079, + "step": 32445 + }, + { + "epoch": 0.97, + "learning_rate": 6.067275804210626e-07, + "loss": 0.54, + "step": 32446 + }, + { + "epoch": 0.97, + "learning_rate": 6.056674767708659e-07, + "loss": 0.5027, + "step": 32447 + }, + { + "epoch": 0.97, + "learning_rate": 6.046082972495537e-07, + "loss": 0.4913, + "step": 32448 + }, + { + "epoch": 0.97, + "learning_rate": 6.035500418669849e-07, + "loss": 0.549, + "step": 32449 + }, + { + "epoch": 0.97, + "learning_rate": 6.024927106329847e-07, + "loss": 0.4906, + "step": 32450 + }, + { + "epoch": 0.97, + "learning_rate": 6.014363035574011e-07, + "loss": 0.4878, + "step": 32451 + }, + { + "epoch": 0.97, + "learning_rate": 6.003808206500372e-07, + "loss": 0.4917, + "step": 32452 + }, + { + "epoch": 0.97, + "learning_rate": 5.993262619207185e-07, + "loss": 0.5014, + "step": 32453 + }, + { + "epoch": 0.97, + "learning_rate": 5.982726273792483e-07, + "loss": 0.5569, + "step": 32454 + }, + { + "epoch": 0.97, + "learning_rate": 5.972199170354187e-07, + "loss": 0.6346, + "step": 32455 + }, + { + "epoch": 0.97, + "learning_rate": 5.961681308990219e-07, + "loss": 0.6, + "step": 32456 + }, + { + "epoch": 0.97, + "learning_rate": 5.95117268979839e-07, + "loss": 0.5558, + "step": 32457 + }, + { + "epoch": 0.97, + "learning_rate": 5.940673312876399e-07, + "loss": 0.5492, + "step": 32458 + }, + { + "epoch": 0.97, + "learning_rate": 5.930183178321724e-07, + "loss": 0.4743, + "step": 32459 + }, + { + "epoch": 0.97, + "learning_rate": 5.919702286232065e-07, + "loss": 0.4183, + "step": 32460 + }, + { + "epoch": 0.97, + "learning_rate": 5.909230636704788e-07, + "loss": 0.4963, + "step": 32461 + }, + { + "epoch": 0.97, + "learning_rate": 5.898768229837259e-07, + "loss": 0.5974, + "step": 32462 + }, + { + "epoch": 0.97, + "learning_rate": 5.888315065726846e-07, + "loss": 0.4387, + "step": 32463 + }, + { + "epoch": 0.97, + "learning_rate": 5.877871144470581e-07, + "loss": 0.4454, + "step": 32464 + }, + { + "epoch": 0.97, + "learning_rate": 5.867436466165721e-07, + "loss": 0.572, + "step": 32465 + }, + { + "epoch": 0.97, + "learning_rate": 5.857011030909299e-07, + "loss": 0.536, + "step": 32466 + }, + { + "epoch": 0.97, + "learning_rate": 5.846594838798014e-07, + "loss": 0.5484, + "step": 32467 + }, + { + "epoch": 0.97, + "learning_rate": 5.836187889928901e-07, + "loss": 0.5245, + "step": 32468 + }, + { + "epoch": 0.97, + "learning_rate": 5.825790184398661e-07, + "loss": 0.4951, + "step": 32469 + }, + { + "epoch": 0.97, + "learning_rate": 5.815401722303992e-07, + "loss": 0.5768, + "step": 32470 + }, + { + "epoch": 0.97, + "learning_rate": 5.805022503741486e-07, + "loss": 0.5471, + "step": 32471 + }, + { + "epoch": 0.97, + "learning_rate": 5.794652528807621e-07, + "loss": 0.5455, + "step": 32472 + }, + { + "epoch": 0.97, + "learning_rate": 5.784291797598874e-07, + "loss": 0.4724, + "step": 32473 + }, + { + "epoch": 0.97, + "learning_rate": 5.773940310211501e-07, + "loss": 0.4476, + "step": 32474 + }, + { + "epoch": 0.97, + "learning_rate": 5.76359806674176e-07, + "loss": 0.4207, + "step": 32475 + }, + { + "epoch": 0.97, + "learning_rate": 5.753265067285796e-07, + "loss": 0.6925, + "step": 32476 + }, + { + "epoch": 0.97, + "learning_rate": 5.742941311939754e-07, + "loss": 0.5043, + "step": 32477 + }, + { + "epoch": 0.97, + "learning_rate": 5.732626800799557e-07, + "loss": 0.6016, + "step": 32478 + }, + { + "epoch": 0.97, + "learning_rate": 5.722321533961128e-07, + "loss": 0.4314, + "step": 32479 + }, + { + "epoch": 0.97, + "learning_rate": 5.71202551152028e-07, + "loss": 0.5598, + "step": 32480 + }, + { + "epoch": 0.97, + "learning_rate": 5.701738733572715e-07, + "loss": 0.5095, + "step": 32481 + }, + { + "epoch": 0.97, + "learning_rate": 5.691461200214021e-07, + "loss": 0.5272, + "step": 32482 + }, + { + "epoch": 0.97, + "learning_rate": 5.681192911539901e-07, + "loss": 0.4524, + "step": 32483 + }, + { + "epoch": 0.97, + "learning_rate": 5.670933867645834e-07, + "loss": 0.5722, + "step": 32484 + }, + { + "epoch": 0.97, + "learning_rate": 5.660684068627075e-07, + "loss": 0.4968, + "step": 32485 + }, + { + "epoch": 0.97, + "learning_rate": 5.650443514578885e-07, + "loss": 0.5153, + "step": 32486 + }, + { + "epoch": 0.97, + "learning_rate": 5.640212205596629e-07, + "loss": 0.5824, + "step": 32487 + }, + { + "epoch": 0.97, + "learning_rate": 5.629990141775343e-07, + "loss": 0.6635, + "step": 32488 + }, + { + "epoch": 0.97, + "learning_rate": 5.619777323210063e-07, + "loss": 0.6351, + "step": 32489 + }, + { + "epoch": 0.97, + "learning_rate": 5.609573749995933e-07, + "loss": 0.4759, + "step": 32490 + }, + { + "epoch": 0.97, + "learning_rate": 5.599379422227435e-07, + "loss": 0.5385, + "step": 32491 + }, + { + "epoch": 0.97, + "learning_rate": 5.589194339999715e-07, + "loss": 0.5058, + "step": 32492 + }, + { + "epoch": 0.97, + "learning_rate": 5.579018503407363e-07, + "loss": 0.6062, + "step": 32493 + }, + { + "epoch": 0.97, + "learning_rate": 5.568851912544859e-07, + "loss": 0.5361, + "step": 32494 + }, + { + "epoch": 0.97, + "learning_rate": 5.558694567506906e-07, + "loss": 0.5013, + "step": 32495 + }, + { + "epoch": 0.97, + "learning_rate": 5.548546468387761e-07, + "loss": 0.4351, + "step": 32496 + }, + { + "epoch": 0.97, + "learning_rate": 5.538407615282015e-07, + "loss": 0.5193, + "step": 32497 + }, + { + "epoch": 0.97, + "learning_rate": 5.528278008283705e-07, + "loss": 0.4466, + "step": 32498 + }, + { + "epoch": 0.97, + "learning_rate": 5.518157647487199e-07, + "loss": 0.4937, + "step": 32499 + }, + { + "epoch": 0.97, + "learning_rate": 5.508046532986421e-07, + "loss": 0.5495, + "step": 32500 + }, + { + "epoch": 0.97, + "learning_rate": 5.497944664875521e-07, + "loss": 0.6176, + "step": 32501 + }, + { + "epoch": 0.97, + "learning_rate": 5.487852043248309e-07, + "loss": 0.5621, + "step": 32502 + }, + { + "epoch": 0.97, + "learning_rate": 5.477768668198601e-07, + "loss": 0.4821, + "step": 32503 + }, + { + "epoch": 0.97, + "learning_rate": 5.467694539820323e-07, + "loss": 0.4828, + "step": 32504 + }, + { + "epoch": 0.97, + "learning_rate": 5.457629658207064e-07, + "loss": 0.5607, + "step": 32505 + }, + { + "epoch": 0.97, + "learning_rate": 5.447574023452306e-07, + "loss": 0.6155, + "step": 32506 + }, + { + "epoch": 0.97, + "learning_rate": 5.437527635649641e-07, + "loss": 0.4836, + "step": 32507 + }, + { + "epoch": 0.97, + "learning_rate": 5.427490494892551e-07, + "loss": 0.5345, + "step": 32508 + }, + { + "epoch": 0.97, + "learning_rate": 5.417462601274181e-07, + "loss": 0.4562, + "step": 32509 + }, + { + "epoch": 0.97, + "learning_rate": 5.407443954887792e-07, + "loss": 0.57, + "step": 32510 + }, + { + "epoch": 0.97, + "learning_rate": 5.397434555826642e-07, + "loss": 0.5696, + "step": 32511 + }, + { + "epoch": 0.97, + "learning_rate": 5.387434404183766e-07, + "loss": 0.5307, + "step": 32512 + }, + { + "epoch": 0.97, + "learning_rate": 5.377443500051981e-07, + "loss": 0.4199, + "step": 32513 + }, + { + "epoch": 0.97, + "learning_rate": 5.367461843524435e-07, + "loss": 0.4055, + "step": 32514 + }, + { + "epoch": 0.97, + "learning_rate": 5.357489434693719e-07, + "loss": 0.6232, + "step": 32515 + }, + { + "epoch": 0.97, + "learning_rate": 5.347526273652759e-07, + "loss": 0.5466, + "step": 32516 + }, + { + "epoch": 0.97, + "learning_rate": 5.337572360493926e-07, + "loss": 0.4259, + "step": 32517 + }, + { + "epoch": 0.97, + "learning_rate": 5.327627695310033e-07, + "loss": 0.53, + "step": 32518 + }, + { + "epoch": 0.97, + "learning_rate": 5.317692278193342e-07, + "loss": 0.574, + "step": 32519 + }, + { + "epoch": 0.97, + "learning_rate": 5.307766109236334e-07, + "loss": 0.5017, + "step": 32520 + }, + { + "epoch": 0.97, + "learning_rate": 5.297849188531267e-07, + "loss": 0.6527, + "step": 32521 + }, + { + "epoch": 0.97, + "learning_rate": 5.287941516170292e-07, + "loss": 0.5178, + "step": 32522 + }, + { + "epoch": 0.97, + "learning_rate": 5.278043092245666e-07, + "loss": 0.5246, + "step": 32523 + }, + { + "epoch": 0.97, + "learning_rate": 5.268153916849427e-07, + "loss": 0.4501, + "step": 32524 + }, + { + "epoch": 0.97, + "learning_rate": 5.258273990073281e-07, + "loss": 0.5556, + "step": 32525 + }, + { + "epoch": 0.97, + "learning_rate": 5.248403312009265e-07, + "loss": 0.5534, + "step": 32526 + }, + { + "epoch": 0.97, + "learning_rate": 5.238541882749193e-07, + "loss": 0.5681, + "step": 32527 + }, + { + "epoch": 0.97, + "learning_rate": 5.228689702384659e-07, + "loss": 0.4552, + "step": 32528 + }, + { + "epoch": 0.97, + "learning_rate": 5.218846771007257e-07, + "loss": 0.4381, + "step": 32529 + }, + { + "epoch": 0.97, + "learning_rate": 5.20901308870847e-07, + "loss": 0.5342, + "step": 32530 + }, + { + "epoch": 0.97, + "learning_rate": 5.19918865557989e-07, + "loss": 0.648, + "step": 32531 + }, + { + "epoch": 0.97, + "learning_rate": 5.189373471712778e-07, + "loss": 0.5095, + "step": 32532 + }, + { + "epoch": 0.97, + "learning_rate": 5.179567537198282e-07, + "loss": 0.4998, + "step": 32533 + }, + { + "epoch": 0.97, + "learning_rate": 5.169770852127775e-07, + "loss": 0.4494, + "step": 32534 + }, + { + "epoch": 0.97, + "learning_rate": 5.159983416592185e-07, + "loss": 0.4658, + "step": 32535 + }, + { + "epoch": 0.97, + "learning_rate": 5.150205230682547e-07, + "loss": 0.5056, + "step": 32536 + }, + { + "epoch": 0.97, + "learning_rate": 5.140436294489903e-07, + "loss": 0.4091, + "step": 32537 + }, + { + "epoch": 0.97, + "learning_rate": 5.130676608104845e-07, + "loss": 0.5734, + "step": 32538 + }, + { + "epoch": 0.97, + "learning_rate": 5.120926171618301e-07, + "loss": 0.6131, + "step": 32539 + }, + { + "epoch": 0.97, + "learning_rate": 5.111184985120865e-07, + "loss": 0.5432, + "step": 32540 + }, + { + "epoch": 0.97, + "learning_rate": 5.101453048703131e-07, + "loss": 0.4391, + "step": 32541 + }, + { + "epoch": 0.97, + "learning_rate": 5.091730362455472e-07, + "loss": 0.5631, + "step": 32542 + }, + { + "epoch": 0.97, + "learning_rate": 5.08201692646848e-07, + "loss": 0.5267, + "step": 32543 + }, + { + "epoch": 0.97, + "learning_rate": 5.072312740832308e-07, + "loss": 0.4954, + "step": 32544 + }, + { + "epoch": 0.97, + "learning_rate": 5.062617805637216e-07, + "loss": 0.5161, + "step": 32545 + }, + { + "epoch": 0.97, + "learning_rate": 5.052932120973353e-07, + "loss": 0.5835, + "step": 32546 + }, + { + "epoch": 0.97, + "learning_rate": 5.04325568693087e-07, + "loss": 0.4683, + "step": 32547 + }, + { + "epoch": 0.97, + "learning_rate": 5.033588503599473e-07, + "loss": 0.5004, + "step": 32548 + }, + { + "epoch": 0.97, + "learning_rate": 5.023930571069313e-07, + "loss": 0.4671, + "step": 32549 + }, + { + "epoch": 0.97, + "learning_rate": 5.014281889430095e-07, + "loss": 0.5717, + "step": 32550 + }, + { + "epoch": 0.97, + "learning_rate": 5.004642458771525e-07, + "loss": 0.5403, + "step": 32551 + }, + { + "epoch": 0.97, + "learning_rate": 4.995012279183308e-07, + "loss": 0.4716, + "step": 32552 + }, + { + "epoch": 0.97, + "learning_rate": 4.985391350754709e-07, + "loss": 0.6179, + "step": 32553 + }, + { + "epoch": 0.97, + "learning_rate": 4.975779673575542e-07, + "loss": 0.4948, + "step": 32554 + }, + { + "epoch": 0.97, + "learning_rate": 4.966177247734849e-07, + "loss": 0.5629, + "step": 32555 + }, + { + "epoch": 0.97, + "learning_rate": 4.956584073322224e-07, + "loss": 0.5449, + "step": 32556 + }, + { + "epoch": 0.97, + "learning_rate": 4.947000150426595e-07, + "loss": 0.5395, + "step": 32557 + }, + { + "epoch": 0.97, + "learning_rate": 4.937425479137115e-07, + "loss": 0.4837, + "step": 32558 + }, + { + "epoch": 0.97, + "learning_rate": 4.927860059542933e-07, + "loss": 0.4867, + "step": 32559 + }, + { + "epoch": 0.97, + "learning_rate": 4.918303891732868e-07, + "loss": 0.5387, + "step": 32560 + }, + { + "epoch": 0.97, + "learning_rate": 4.908756975795958e-07, + "loss": 0.4568, + "step": 32561 + }, + { + "epoch": 0.97, + "learning_rate": 4.899219311820691e-07, + "loss": 0.4611, + "step": 32562 + }, + { + "epoch": 0.97, + "learning_rate": 4.889690899895883e-07, + "loss": 0.4892, + "step": 32563 + }, + { + "epoch": 0.97, + "learning_rate": 4.880171740110129e-07, + "loss": 0.5712, + "step": 32564 + }, + { + "epoch": 0.97, + "learning_rate": 4.870661832551914e-07, + "loss": 0.5509, + "step": 32565 + }, + { + "epoch": 0.97, + "learning_rate": 4.861161177309615e-07, + "loss": 0.4856, + "step": 32566 + }, + { + "epoch": 0.97, + "learning_rate": 4.851669774471712e-07, + "loss": 0.4716, + "step": 32567 + }, + { + "epoch": 0.97, + "learning_rate": 4.84218762412636e-07, + "loss": 0.4141, + "step": 32568 + }, + { + "epoch": 0.97, + "learning_rate": 4.832714726361598e-07, + "loss": 0.4939, + "step": 32569 + }, + { + "epoch": 0.97, + "learning_rate": 4.8232510812658e-07, + "loss": 0.4422, + "step": 32570 + }, + { + "epoch": 0.97, + "learning_rate": 4.813796688926675e-07, + "loss": 0.5345, + "step": 32571 + }, + { + "epoch": 0.97, + "learning_rate": 4.804351549432152e-07, + "loss": 0.4746, + "step": 32572 + }, + { + "epoch": 0.97, + "learning_rate": 4.794915662870159e-07, + "loss": 0.4892, + "step": 32573 + }, + { + "epoch": 0.97, + "learning_rate": 4.785489029328516e-07, + "loss": 0.5122, + "step": 32574 + }, + { + "epoch": 0.97, + "learning_rate": 4.776071648894598e-07, + "loss": 0.62, + "step": 32575 + }, + { + "epoch": 0.97, + "learning_rate": 4.7666635216562225e-07, + "loss": 0.605, + "step": 32576 + }, + { + "epoch": 0.97, + "learning_rate": 4.757264647700654e-07, + "loss": 0.5179, + "step": 32577 + }, + { + "epoch": 0.97, + "learning_rate": 4.7478750271154894e-07, + "loss": 0.4481, + "step": 32578 + }, + { + "epoch": 0.97, + "learning_rate": 4.738494659987769e-07, + "loss": 0.5678, + "step": 32579 + }, + { + "epoch": 0.97, + "learning_rate": 4.7291235464049787e-07, + "loss": 0.4928, + "step": 32580 + }, + { + "epoch": 0.97, + "learning_rate": 4.7197616864540495e-07, + "loss": 0.6477, + "step": 32581 + }, + { + "epoch": 0.97, + "learning_rate": 4.7104090802221336e-07, + "loss": 0.5069, + "step": 32582 + }, + { + "epoch": 0.97, + "learning_rate": 4.70106572779605e-07, + "loss": 0.5537, + "step": 32583 + }, + { + "epoch": 0.97, + "learning_rate": 4.691731629262841e-07, + "loss": 0.4536, + "step": 32584 + }, + { + "epoch": 0.97, + "learning_rate": 4.6824067847092145e-07, + "loss": 0.4771, + "step": 32585 + }, + { + "epoch": 0.97, + "learning_rate": 4.67309119422199e-07, + "loss": 0.583, + "step": 32586 + }, + { + "epoch": 0.97, + "learning_rate": 4.663784857887543e-07, + "loss": 0.5216, + "step": 32587 + }, + { + "epoch": 0.97, + "learning_rate": 4.654487775792471e-07, + "loss": 0.5924, + "step": 32588 + }, + { + "epoch": 0.97, + "learning_rate": 4.645199948023371e-07, + "loss": 0.5724, + "step": 32589 + }, + { + "epoch": 0.97, + "learning_rate": 4.6359213746665073e-07, + "loss": 0.548, + "step": 32590 + }, + { + "epoch": 0.97, + "learning_rate": 4.626652055808034e-07, + "loss": 0.4837, + "step": 32591 + }, + { + "epoch": 0.97, + "learning_rate": 4.6173919915343256e-07, + "loss": 0.4862, + "step": 32592 + }, + { + "epoch": 0.97, + "learning_rate": 4.6081411819313136e-07, + "loss": 0.5644, + "step": 32593 + }, + { + "epoch": 0.97, + "learning_rate": 4.5988996270851515e-07, + "loss": 0.5433, + "step": 32594 + }, + { + "epoch": 0.97, + "learning_rate": 4.589667327081659e-07, + "loss": 0.5536, + "step": 32595 + }, + { + "epoch": 0.97, + "learning_rate": 4.580444282006657e-07, + "loss": 0.5659, + "step": 32596 + }, + { + "epoch": 0.97, + "learning_rate": 4.571230491945966e-07, + "loss": 0.5126, + "step": 32597 + }, + { + "epoch": 0.97, + "learning_rate": 4.562025956985294e-07, + "loss": 0.5376, + "step": 32598 + }, + { + "epoch": 0.97, + "learning_rate": 4.552830677210129e-07, + "loss": 0.4442, + "step": 32599 + }, + { + "epoch": 0.97, + "learning_rate": 4.5436446527059584e-07, + "loss": 0.536, + "step": 32600 + }, + { + "epoch": 0.97, + "learning_rate": 4.5344678835582686e-07, + "loss": 0.6028, + "step": 32601 + }, + { + "epoch": 0.97, + "learning_rate": 4.5253003698524365e-07, + "loss": 0.609, + "step": 32602 + }, + { + "epoch": 0.97, + "learning_rate": 4.5161421116735045e-07, + "loss": 0.5414, + "step": 32603 + }, + { + "epoch": 0.97, + "learning_rate": 4.5069931091067383e-07, + "loss": 0.5247, + "step": 32604 + }, + { + "epoch": 0.97, + "learning_rate": 4.497853362237181e-07, + "loss": 0.6097, + "step": 32605 + }, + { + "epoch": 0.97, + "learning_rate": 4.488722871149764e-07, + "loss": 0.5831, + "step": 32606 + }, + { + "epoch": 0.97, + "learning_rate": 4.4796016359295314e-07, + "loss": 0.5689, + "step": 32607 + }, + { + "epoch": 0.97, + "learning_rate": 4.470489656661192e-07, + "loss": 0.4924, + "step": 32608 + }, + { + "epoch": 0.97, + "learning_rate": 4.4613869334293454e-07, + "loss": 0.5264, + "step": 32609 + }, + { + "epoch": 0.97, + "learning_rate": 4.452293466318813e-07, + "loss": 0.5663, + "step": 32610 + }, + { + "epoch": 0.97, + "learning_rate": 4.4432092554140823e-07, + "loss": 0.5367, + "step": 32611 + }, + { + "epoch": 0.97, + "learning_rate": 4.4341343007995305e-07, + "loss": 0.4396, + "step": 32612 + }, + { + "epoch": 0.97, + "learning_rate": 4.4250686025596457e-07, + "loss": 0.4805, + "step": 32613 + }, + { + "epoch": 0.97, + "learning_rate": 4.4160121607786933e-07, + "loss": 0.5286, + "step": 32614 + }, + { + "epoch": 0.97, + "learning_rate": 4.406964975540828e-07, + "loss": 0.5705, + "step": 32615 + }, + { + "epoch": 0.97, + "learning_rate": 4.397927046930206e-07, + "loss": 0.5757, + "step": 32616 + }, + { + "epoch": 0.97, + "learning_rate": 4.38889837503087e-07, + "loss": 0.467, + "step": 32617 + }, + { + "epoch": 0.97, + "learning_rate": 4.379878959926864e-07, + "loss": 0.4512, + "step": 32618 + }, + { + "epoch": 0.97, + "learning_rate": 4.3708688017016777e-07, + "loss": 0.5673, + "step": 32619 + }, + { + "epoch": 0.97, + "learning_rate": 4.3618679004394644e-07, + "loss": 0.4947, + "step": 32620 + }, + { + "epoch": 0.97, + "learning_rate": 4.3528762562238254e-07, + "loss": 0.5804, + "step": 32621 + }, + { + "epoch": 0.97, + "learning_rate": 4.343893869138249e-07, + "loss": 0.5075, + "step": 32622 + }, + { + "epoch": 0.97, + "learning_rate": 4.334920739266335e-07, + "loss": 0.5691, + "step": 32623 + }, + { + "epoch": 0.97, + "learning_rate": 4.325956866691461e-07, + "loss": 0.5219, + "step": 32624 + }, + { + "epoch": 0.97, + "learning_rate": 4.3170022514971154e-07, + "loss": 0.561, + "step": 32625 + }, + { + "epoch": 0.97, + "learning_rate": 4.308056893766343e-07, + "loss": 0.5557, + "step": 32626 + }, + { + "epoch": 0.97, + "learning_rate": 4.299120793582412e-07, + "loss": 0.5559, + "step": 32627 + }, + { + "epoch": 0.97, + "learning_rate": 4.2901939510284763e-07, + "loss": 0.5935, + "step": 32628 + }, + { + "epoch": 0.97, + "learning_rate": 4.28127636618747e-07, + "loss": 0.5323, + "step": 32629 + }, + { + "epoch": 0.97, + "learning_rate": 4.272368039142216e-07, + "loss": 0.5537, + "step": 32630 + }, + { + "epoch": 0.97, + "learning_rate": 4.263468969975648e-07, + "loss": 0.6315, + "step": 32631 + }, + { + "epoch": 0.97, + "learning_rate": 4.254579158770478e-07, + "loss": 0.6216, + "step": 32632 + }, + { + "epoch": 0.97, + "learning_rate": 4.245698605609416e-07, + "loss": 0.4658, + "step": 32633 + }, + { + "epoch": 0.97, + "learning_rate": 4.236827310574953e-07, + "loss": 0.5896, + "step": 32634 + }, + { + "epoch": 0.97, + "learning_rate": 4.227965273749579e-07, + "loss": 0.4961, + "step": 32635 + }, + { + "epoch": 0.97, + "learning_rate": 4.2191124952157825e-07, + "loss": 0.5946, + "step": 32636 + }, + { + "epoch": 0.97, + "learning_rate": 4.21026897505572e-07, + "loss": 0.5618, + "step": 32637 + }, + { + "epoch": 0.97, + "learning_rate": 4.2014347133517707e-07, + "loss": 0.46, + "step": 32638 + }, + { + "epoch": 0.97, + "learning_rate": 4.1926097101858685e-07, + "loss": 0.5111, + "step": 32639 + }, + { + "epoch": 0.97, + "learning_rate": 4.1837939656403924e-07, + "loss": 0.4799, + "step": 32640 + }, + { + "epoch": 0.97, + "learning_rate": 4.1749874797969433e-07, + "loss": 0.5551, + "step": 32641 + }, + { + "epoch": 0.97, + "learning_rate": 4.1661902527375675e-07, + "loss": 0.5634, + "step": 32642 + }, + { + "epoch": 0.97, + "learning_rate": 4.1574022845440874e-07, + "loss": 0.5232, + "step": 32643 + }, + { + "epoch": 0.97, + "learning_rate": 4.148623575298216e-07, + "loss": 0.5461, + "step": 32644 + }, + { + "epoch": 0.97, + "learning_rate": 4.139854125081444e-07, + "loss": 0.4667, + "step": 32645 + }, + { + "epoch": 0.97, + "learning_rate": 4.131093933975483e-07, + "loss": 0.4879, + "step": 32646 + }, + { + "epoch": 0.97, + "learning_rate": 4.122343002061713e-07, + "loss": 0.5896, + "step": 32647 + }, + { + "epoch": 0.97, + "learning_rate": 4.1136013294214014e-07, + "loss": 0.6463, + "step": 32648 + }, + { + "epoch": 0.97, + "learning_rate": 4.104868916135929e-07, + "loss": 0.5762, + "step": 32649 + }, + { + "epoch": 0.97, + "learning_rate": 4.0961457622864516e-07, + "loss": 0.6288, + "step": 32650 + }, + { + "epoch": 0.97, + "learning_rate": 4.087431867954128e-07, + "loss": 0.6133, + "step": 32651 + }, + { + "epoch": 0.97, + "learning_rate": 4.0787272332198924e-07, + "loss": 0.4921, + "step": 32652 + }, + { + "epoch": 0.97, + "learning_rate": 4.0700318581647914e-07, + "loss": 0.4525, + "step": 32653 + }, + { + "epoch": 0.97, + "learning_rate": 4.061345742869427e-07, + "loss": 0.5073, + "step": 32654 + }, + { + "epoch": 0.97, + "learning_rate": 4.0526688874148456e-07, + "loss": 0.5107, + "step": 32655 + }, + { + "epoch": 0.97, + "learning_rate": 4.0440012918816497e-07, + "loss": 0.4457, + "step": 32656 + }, + { + "epoch": 0.97, + "learning_rate": 4.0353429563503297e-07, + "loss": 0.5075, + "step": 32657 + }, + { + "epoch": 0.97, + "learning_rate": 4.0266938809013775e-07, + "loss": 0.5141, + "step": 32658 + }, + { + "epoch": 0.97, + "learning_rate": 4.0180540656152846e-07, + "loss": 0.558, + "step": 32659 + }, + { + "epoch": 0.97, + "learning_rate": 4.0094235105724296e-07, + "loss": 0.562, + "step": 32660 + }, + { + "epoch": 0.97, + "learning_rate": 4.0008022158529725e-07, + "loss": 0.5059, + "step": 32661 + }, + { + "epoch": 0.97, + "learning_rate": 3.99219018153707e-07, + "loss": 0.5168, + "step": 32662 + }, + { + "epoch": 0.97, + "learning_rate": 3.9835874077048807e-07, + "loss": 0.496, + "step": 32663 + }, + { + "epoch": 0.97, + "learning_rate": 3.97499389443623e-07, + "loss": 0.5072, + "step": 32664 + }, + { + "epoch": 0.97, + "learning_rate": 3.966409641811164e-07, + "loss": 0.5623, + "step": 32665 + }, + { + "epoch": 0.97, + "learning_rate": 3.9578346499093975e-07, + "loss": 0.5753, + "step": 32666 + }, + { + "epoch": 0.97, + "learning_rate": 3.949268918810756e-07, + "loss": 0.5603, + "step": 32667 + }, + { + "epoch": 0.97, + "learning_rate": 3.940712448594841e-07, + "loss": 0.692, + "step": 32668 + }, + { + "epoch": 0.97, + "learning_rate": 3.9321652393411456e-07, + "loss": 0.5234, + "step": 32669 + }, + { + "epoch": 0.97, + "learning_rate": 3.9236272911291616e-07, + "loss": 0.5915, + "step": 32670 + }, + { + "epoch": 0.97, + "learning_rate": 3.9150986040383806e-07, + "loss": 0.5892, + "step": 32671 + }, + { + "epoch": 0.97, + "learning_rate": 3.906579178147851e-07, + "loss": 0.5245, + "step": 32672 + }, + { + "epoch": 0.97, + "learning_rate": 3.8980690135370646e-07, + "loss": 0.6373, + "step": 32673 + }, + { + "epoch": 0.97, + "learning_rate": 3.889568110284958e-07, + "loss": 0.543, + "step": 32674 + }, + { + "epoch": 0.97, + "learning_rate": 3.8810764684706904e-07, + "loss": 0.5239, + "step": 32675 + }, + { + "epoch": 0.97, + "learning_rate": 3.8725940881731984e-07, + "loss": 0.5573, + "step": 32676 + }, + { + "epoch": 0.97, + "learning_rate": 3.8641209694713076e-07, + "loss": 0.5397, + "step": 32677 + }, + { + "epoch": 0.97, + "learning_rate": 3.8556571124437333e-07, + "loss": 0.6376, + "step": 32678 + }, + { + "epoch": 0.97, + "learning_rate": 3.8472025171693015e-07, + "loss": 0.5383, + "step": 32679 + }, + { + "epoch": 0.97, + "learning_rate": 3.8387571837265045e-07, + "loss": 0.5207, + "step": 32680 + }, + { + "epoch": 0.97, + "learning_rate": 3.8303211121939465e-07, + "loss": 0.4619, + "step": 32681 + }, + { + "epoch": 0.97, + "learning_rate": 3.8218943026498976e-07, + "loss": 0.5438, + "step": 32682 + }, + { + "epoch": 0.97, + "learning_rate": 3.813476755172962e-07, + "loss": 0.5646, + "step": 32683 + }, + { + "epoch": 0.97, + "learning_rate": 3.8050684698412995e-07, + "loss": 0.5648, + "step": 32684 + }, + { + "epoch": 0.97, + "learning_rate": 3.79666944673307e-07, + "loss": 0.5465, + "step": 32685 + }, + { + "epoch": 0.97, + "learning_rate": 3.7882796859262104e-07, + "loss": 0.5817, + "step": 32686 + }, + { + "epoch": 0.97, + "learning_rate": 3.779899187498992e-07, + "loss": 0.5136, + "step": 32687 + }, + { + "epoch": 0.97, + "learning_rate": 3.771527951529241e-07, + "loss": 0.4967, + "step": 32688 + }, + { + "epoch": 0.97, + "learning_rate": 3.7631659780946737e-07, + "loss": 0.5849, + "step": 32689 + }, + { + "epoch": 0.97, + "learning_rate": 3.754813267273227e-07, + "loss": 0.5165, + "step": 32690 + }, + { + "epoch": 0.97, + "learning_rate": 3.746469819142506e-07, + "loss": 0.4946, + "step": 32691 + }, + { + "epoch": 0.97, + "learning_rate": 3.7381356337798934e-07, + "loss": 0.5731, + "step": 32692 + }, + { + "epoch": 0.97, + "learning_rate": 3.729810711263104e-07, + "loss": 0.5465, + "step": 32693 + }, + { + "epoch": 0.97, + "learning_rate": 3.7214950516694105e-07, + "loss": 0.4941, + "step": 32694 + }, + { + "epoch": 0.97, + "learning_rate": 3.7131886550763053e-07, + "loss": 0.5466, + "step": 32695 + }, + { + "epoch": 0.97, + "learning_rate": 3.7048915215607273e-07, + "loss": 0.4688, + "step": 32696 + }, + { + "epoch": 0.97, + "learning_rate": 3.696603651200059e-07, + "loss": 0.5777, + "step": 32697 + }, + { + "epoch": 0.97, + "learning_rate": 3.6883250440712393e-07, + "loss": 0.6403, + "step": 32698 + }, + { + "epoch": 0.97, + "learning_rate": 3.680055700251317e-07, + "loss": 0.4683, + "step": 32699 + }, + { + "epoch": 0.97, + "learning_rate": 3.671795619817231e-07, + "loss": 0.5553, + "step": 32700 + }, + { + "epoch": 0.97, + "learning_rate": 3.6635448028455865e-07, + "loss": 0.4973, + "step": 32701 + }, + { + "epoch": 0.97, + "learning_rate": 3.6553032494132114e-07, + "loss": 0.428, + "step": 32702 + }, + { + "epoch": 0.97, + "learning_rate": 3.64707095959671e-07, + "loss": 0.5097, + "step": 32703 + }, + { + "epoch": 0.97, + "learning_rate": 3.6388479334726887e-07, + "loss": 0.5302, + "step": 32704 + }, + { + "epoch": 0.97, + "learning_rate": 3.6306341711175306e-07, + "loss": 0.5757, + "step": 32705 + }, + { + "epoch": 0.97, + "learning_rate": 3.622429672607619e-07, + "loss": 0.4926, + "step": 32706 + }, + { + "epoch": 0.97, + "learning_rate": 3.6142344380192263e-07, + "loss": 0.4879, + "step": 32707 + }, + { + "epoch": 0.97, + "learning_rate": 3.606048467428513e-07, + "loss": 0.4528, + "step": 32708 + }, + { + "epoch": 0.97, + "learning_rate": 3.597871760911753e-07, + "loss": 0.5019, + "step": 32709 + }, + { + "epoch": 0.97, + "learning_rate": 3.5897043185447733e-07, + "loss": 0.5808, + "step": 32710 + }, + { + "epoch": 0.97, + "learning_rate": 3.581546140403624e-07, + "loss": 0.4454, + "step": 32711 + }, + { + "epoch": 0.97, + "learning_rate": 3.5733972265641347e-07, + "loss": 0.6432, + "step": 32712 + }, + { + "epoch": 0.97, + "learning_rate": 3.565257577102021e-07, + "loss": 0.4677, + "step": 32713 + }, + { + "epoch": 0.97, + "learning_rate": 3.557127192093113e-07, + "loss": 0.5348, + "step": 32714 + }, + { + "epoch": 0.97, + "learning_rate": 3.5490060716126815e-07, + "loss": 0.425, + "step": 32715 + }, + { + "epoch": 0.97, + "learning_rate": 3.5408942157365564e-07, + "loss": 0.4456, + "step": 32716 + }, + { + "epoch": 0.97, + "learning_rate": 3.532791624540122e-07, + "loss": 0.4808, + "step": 32717 + }, + { + "epoch": 0.97, + "learning_rate": 3.5246982980985386e-07, + "loss": 0.5156, + "step": 32718 + }, + { + "epoch": 0.97, + "learning_rate": 3.5166142364871923e-07, + "loss": 0.5257, + "step": 32719 + }, + { + "epoch": 0.97, + "learning_rate": 3.5085394397812443e-07, + "loss": 0.5302, + "step": 32720 + }, + { + "epoch": 0.97, + "learning_rate": 3.500473908055635e-07, + "loss": 0.515, + "step": 32721 + }, + { + "epoch": 0.97, + "learning_rate": 3.492417641385637e-07, + "loss": 0.4568, + "step": 32722 + }, + { + "epoch": 0.97, + "learning_rate": 3.4843706398458575e-07, + "loss": 0.5815, + "step": 32723 + }, + { + "epoch": 0.97, + "learning_rate": 3.476332903511126e-07, + "loss": 0.4826, + "step": 32724 + }, + { + "epoch": 0.97, + "learning_rate": 3.468304432456382e-07, + "loss": 0.5649, + "step": 32725 + }, + { + "epoch": 0.97, + "learning_rate": 3.4602852267562325e-07, + "loss": 0.6904, + "step": 32726 + }, + { + "epoch": 0.97, + "learning_rate": 3.4522752864850627e-07, + "loss": 0.6138, + "step": 32727 + }, + { + "epoch": 0.97, + "learning_rate": 3.4442746117174797e-07, + "loss": 0.5389, + "step": 32728 + }, + { + "epoch": 0.97, + "learning_rate": 3.4362832025278677e-07, + "loss": 0.5277, + "step": 32729 + }, + { + "epoch": 0.97, + "learning_rate": 3.4283010589905016e-07, + "loss": 0.4618, + "step": 32730 + }, + { + "epoch": 0.97, + "learning_rate": 3.420328181179544e-07, + "loss": 0.6434, + "step": 32731 + }, + { + "epoch": 0.97, + "learning_rate": 3.4123645691691577e-07, + "loss": 0.4756, + "step": 32732 + }, + { + "epoch": 0.97, + "learning_rate": 3.404410223033505e-07, + "loss": 0.4968, + "step": 32733 + }, + { + "epoch": 0.97, + "learning_rate": 3.3964651428464167e-07, + "loss": 0.456, + "step": 32734 + }, + { + "epoch": 0.97, + "learning_rate": 3.3885293286817223e-07, + "loss": 0.4632, + "step": 32735 + }, + { + "epoch": 0.97, + "learning_rate": 3.3806027806132513e-07, + "loss": 0.5211, + "step": 32736 + }, + { + "epoch": 0.97, + "learning_rate": 3.372685498714834e-07, + "loss": 0.4129, + "step": 32737 + }, + { + "epoch": 0.97, + "learning_rate": 3.364777483059967e-07, + "loss": 0.4669, + "step": 32738 + }, + { + "epoch": 0.97, + "learning_rate": 3.356878733722146e-07, + "loss": 0.5055, + "step": 32739 + }, + { + "epoch": 0.97, + "learning_rate": 3.348989250774759e-07, + "loss": 0.4962, + "step": 32740 + }, + { + "epoch": 0.97, + "learning_rate": 3.341109034291412e-07, + "loss": 0.447, + "step": 32741 + }, + { + "epoch": 0.97, + "learning_rate": 3.3332380843450473e-07, + "loss": 0.5171, + "step": 32742 + }, + { + "epoch": 0.97, + "learning_rate": 3.32537640100905e-07, + "loss": 0.6201, + "step": 32743 + }, + { + "epoch": 0.97, + "learning_rate": 3.317523984356474e-07, + "loss": 0.5758, + "step": 32744 + }, + { + "epoch": 0.97, + "learning_rate": 3.3096808344603716e-07, + "loss": 0.5727, + "step": 32745 + }, + { + "epoch": 0.97, + "learning_rate": 3.3018469513935724e-07, + "loss": 0.623, + "step": 32746 + }, + { + "epoch": 0.97, + "learning_rate": 3.294022335229019e-07, + "loss": 0.5616, + "step": 32747 + }, + { + "epoch": 0.97, + "learning_rate": 3.28620698603932e-07, + "loss": 0.5633, + "step": 32748 + }, + { + "epoch": 0.97, + "learning_rate": 3.278400903897194e-07, + "loss": 0.5025, + "step": 32749 + }, + { + "epoch": 0.98, + "learning_rate": 3.270604088875251e-07, + "loss": 0.5424, + "step": 32750 + }, + { + "epoch": 0.98, + "learning_rate": 3.262816541046099e-07, + "loss": 0.4907, + "step": 32751 + }, + { + "epoch": 0.98, + "learning_rate": 3.255038260481902e-07, + "loss": 0.55, + "step": 32752 + }, + { + "epoch": 0.98, + "learning_rate": 3.247269247255158e-07, + "loss": 0.6538, + "step": 32753 + }, + { + "epoch": 0.98, + "learning_rate": 3.2395095014379204e-07, + "loss": 0.5774, + "step": 32754 + }, + { + "epoch": 0.98, + "learning_rate": 3.231759023102576e-07, + "loss": 0.4105, + "step": 32755 + }, + { + "epoch": 0.98, + "learning_rate": 3.2240178123210675e-07, + "loss": 0.5691, + "step": 32756 + }, + { + "epoch": 0.98, + "learning_rate": 3.2162858691653365e-07, + "loss": 0.5087, + "step": 32757 + }, + { + "epoch": 0.98, + "learning_rate": 3.2085631937072146e-07, + "loss": 0.5624, + "step": 32758 + }, + { + "epoch": 0.98, + "learning_rate": 3.200849786018645e-07, + "loss": 0.4827, + "step": 32759 + }, + { + "epoch": 0.98, + "learning_rate": 3.193145646171347e-07, + "loss": 0.494, + "step": 32760 + }, + { + "epoch": 0.98, + "learning_rate": 3.18545077423682e-07, + "loss": 0.5391, + "step": 32761 + }, + { + "epoch": 0.98, + "learning_rate": 3.177765170286784e-07, + "loss": 0.6189, + "step": 32762 + }, + { + "epoch": 0.98, + "learning_rate": 3.170088834392404e-07, + "loss": 0.5255, + "step": 32763 + }, + { + "epoch": 0.98, + "learning_rate": 3.162421766625401e-07, + "loss": 0.6139, + "step": 32764 + }, + { + "epoch": 0.98, + "learning_rate": 3.1547639670568286e-07, + "loss": 0.5419, + "step": 32765 + }, + { + "epoch": 0.98, + "learning_rate": 3.1471154357580747e-07, + "loss": 0.4682, + "step": 32766 + }, + { + "epoch": 0.98, + "learning_rate": 3.1394761727999713e-07, + "loss": 0.4605, + "step": 32767 + }, + { + "epoch": 0.98, + "learning_rate": 3.131846178253794e-07, + "loss": 0.4702, + "step": 32768 + }, + { + "epoch": 0.98, + "learning_rate": 3.124225452190377e-07, + "loss": 0.5092, + "step": 32769 + }, + { + "epoch": 0.98, + "learning_rate": 3.1166139946806614e-07, + "loss": 0.5168, + "step": 32770 + }, + { + "epoch": 0.98, + "learning_rate": 3.1090118057952586e-07, + "loss": 0.504, + "step": 32771 + }, + { + "epoch": 0.98, + "learning_rate": 3.101418885605001e-07, + "loss": 0.5002, + "step": 32772 + }, + { + "epoch": 0.98, + "learning_rate": 3.0938352341803866e-07, + "loss": 0.3764, + "step": 32773 + }, + { + "epoch": 0.98, + "learning_rate": 3.0862608515920267e-07, + "loss": 0.4301, + "step": 32774 + }, + { + "epoch": 0.98, + "learning_rate": 3.0786957379101977e-07, + "loss": 0.5488, + "step": 32775 + }, + { + "epoch": 0.98, + "learning_rate": 3.0711398932053994e-07, + "loss": 0.39, + "step": 32776 + }, + { + "epoch": 0.98, + "learning_rate": 3.063593317547797e-07, + "loss": 0.4636, + "step": 32777 + }, + { + "epoch": 0.98, + "learning_rate": 3.0560560110074464e-07, + "loss": 0.5227, + "step": 32778 + }, + { + "epoch": 0.98, + "learning_rate": 3.0485279736546244e-07, + "loss": 0.4655, + "step": 32779 + }, + { + "epoch": 0.98, + "learning_rate": 3.0410092055592755e-07, + "loss": 0.6192, + "step": 32780 + }, + { + "epoch": 0.98, + "learning_rate": 3.033499706791343e-07, + "loss": 0.5234, + "step": 32781 + }, + { + "epoch": 0.98, + "learning_rate": 3.025999477420438e-07, + "loss": 0.5449, + "step": 32782 + }, + { + "epoch": 0.98, + "learning_rate": 3.018508517516505e-07, + "loss": 0.5539, + "step": 32783 + }, + { + "epoch": 0.98, + "learning_rate": 3.011026827149044e-07, + "loss": 0.5953, + "step": 32784 + }, + { + "epoch": 0.98, + "learning_rate": 3.003554406387776e-07, + "loss": 0.5477, + "step": 32785 + }, + { + "epoch": 0.98, + "learning_rate": 2.9960912553020914e-07, + "loss": 0.4597, + "step": 32786 + }, + { + "epoch": 0.98, + "learning_rate": 2.988637373961378e-07, + "loss": 0.5344, + "step": 32787 + }, + { + "epoch": 0.98, + "learning_rate": 2.9811927624349147e-07, + "loss": 0.5145, + "step": 32788 + }, + { + "epoch": 0.98, + "learning_rate": 2.9737574207919784e-07, + "loss": 0.4865, + "step": 32789 + }, + { + "epoch": 0.98, + "learning_rate": 2.9663313491016254e-07, + "loss": 0.6213, + "step": 32790 + }, + { + "epoch": 0.98, + "learning_rate": 2.9589145474330224e-07, + "loss": 0.5219, + "step": 32791 + }, + { + "epoch": 0.98, + "learning_rate": 2.951507015855004e-07, + "loss": 0.4929, + "step": 32792 + }, + { + "epoch": 0.98, + "learning_rate": 2.944108754436514e-07, + "loss": 0.5874, + "step": 32793 + }, + { + "epoch": 0.98, + "learning_rate": 2.9367197632462763e-07, + "loss": 0.522, + "step": 32794 + }, + { + "epoch": 0.98, + "learning_rate": 2.929340042353013e-07, + "loss": 0.504, + "step": 32795 + }, + { + "epoch": 0.98, + "learning_rate": 2.921969591825446e-07, + "loss": 0.5749, + "step": 32796 + }, + { + "epoch": 0.98, + "learning_rate": 2.914608411731856e-07, + "loss": 0.5133, + "step": 32797 + }, + { + "epoch": 0.98, + "learning_rate": 2.907256502140965e-07, + "loss": 0.5083, + "step": 32798 + }, + { + "epoch": 0.98, + "learning_rate": 2.89991386312094e-07, + "loss": 0.4513, + "step": 32799 + }, + { + "epoch": 0.98, + "learning_rate": 2.892580494740171e-07, + "loss": 0.5589, + "step": 32800 + }, + { + "epoch": 0.98, + "learning_rate": 2.8852563970667155e-07, + "loss": 0.4986, + "step": 32801 + }, + { + "epoch": 0.98, + "learning_rate": 2.8779415701687406e-07, + "loss": 0.6495, + "step": 32802 + }, + { + "epoch": 0.98, + "learning_rate": 2.870636014114303e-07, + "loss": 0.524, + "step": 32803 + }, + { + "epoch": 0.98, + "learning_rate": 2.8633397289712373e-07, + "loss": 0.5309, + "step": 32804 + }, + { + "epoch": 0.98, + "learning_rate": 2.856052714807489e-07, + "loss": 0.5746, + "step": 32805 + }, + { + "epoch": 0.98, + "learning_rate": 2.848774971690671e-07, + "loss": 0.497, + "step": 32806 + }, + { + "epoch": 0.98, + "learning_rate": 2.8415064996885067e-07, + "loss": 0.5757, + "step": 32807 + }, + { + "epoch": 0.98, + "learning_rate": 2.8342472988686084e-07, + "loss": 0.5746, + "step": 32808 + }, + { + "epoch": 0.98, + "learning_rate": 2.8269973692984784e-07, + "loss": 0.4749, + "step": 32809 + }, + { + "epoch": 0.98, + "learning_rate": 2.819756711045396e-07, + "loss": 0.4935, + "step": 32810 + }, + { + "epoch": 0.98, + "learning_rate": 2.812525324176862e-07, + "loss": 0.5084, + "step": 32811 + }, + { + "epoch": 0.98, + "learning_rate": 2.8053032087600463e-07, + "loss": 0.5397, + "step": 32812 + }, + { + "epoch": 0.98, + "learning_rate": 2.7980903648620047e-07, + "loss": 0.5649, + "step": 32813 + }, + { + "epoch": 0.98, + "learning_rate": 2.790886792549907e-07, + "loss": 0.439, + "step": 32814 + }, + { + "epoch": 0.98, + "learning_rate": 2.7836924918906993e-07, + "loss": 0.6206, + "step": 32815 + }, + { + "epoch": 0.98, + "learning_rate": 2.7765074629513276e-07, + "loss": 0.4293, + "step": 32816 + }, + { + "epoch": 0.98, + "learning_rate": 2.769331705798406e-07, + "loss": 0.5503, + "step": 32817 + }, + { + "epoch": 0.98, + "learning_rate": 2.76216522049888e-07, + "loss": 0.4689, + "step": 32818 + }, + { + "epoch": 0.98, + "learning_rate": 2.755008007119253e-07, + "loss": 0.5544, + "step": 32819 + }, + { + "epoch": 0.98, + "learning_rate": 2.7478600657261376e-07, + "loss": 0.5381, + "step": 32820 + }, + { + "epoch": 0.98, + "learning_rate": 2.740721396385926e-07, + "loss": 0.5614, + "step": 32821 + }, + { + "epoch": 0.98, + "learning_rate": 2.7335919991650083e-07, + "loss": 0.4819, + "step": 32822 + }, + { + "epoch": 0.98, + "learning_rate": 2.726471874129666e-07, + "loss": 0.4823, + "step": 32823 + }, + { + "epoch": 0.98, + "learning_rate": 2.7193610213461787e-07, + "loss": 0.575, + "step": 32824 + }, + { + "epoch": 0.98, + "learning_rate": 2.7122594408804937e-07, + "loss": 0.5175, + "step": 32825 + }, + { + "epoch": 0.98, + "learning_rate": 2.7051671327987806e-07, + "loss": 0.4424, + "step": 32826 + }, + { + "epoch": 0.98, + "learning_rate": 2.698084097166986e-07, + "loss": 0.5729, + "step": 32827 + }, + { + "epoch": 0.98, + "learning_rate": 2.6910103340508364e-07, + "loss": 0.5073, + "step": 32828 + }, + { + "epoch": 0.98, + "learning_rate": 2.6839458435162777e-07, + "loss": 0.5325, + "step": 32829 + }, + { + "epoch": 0.98, + "learning_rate": 2.676890625628814e-07, + "loss": 0.5073, + "step": 32830 + }, + { + "epoch": 0.98, + "learning_rate": 2.66984468045417e-07, + "loss": 0.5813, + "step": 32831 + }, + { + "epoch": 0.98, + "learning_rate": 2.662808008057849e-07, + "loss": 0.5991, + "step": 32832 + }, + { + "epoch": 0.98, + "learning_rate": 2.655780608505243e-07, + "loss": 0.5093, + "step": 32833 + }, + { + "epoch": 0.98, + "learning_rate": 2.648762481861633e-07, + "loss": 0.5358, + "step": 32834 + }, + { + "epoch": 0.98, + "learning_rate": 2.6417536281923003e-07, + "loss": 0.4503, + "step": 32835 + }, + { + "epoch": 0.98, + "learning_rate": 2.6347540475625265e-07, + "loss": 0.5663, + "step": 32836 + }, + { + "epoch": 0.98, + "learning_rate": 2.627763740037259e-07, + "loss": 0.689, + "step": 32837 + }, + { + "epoch": 0.98, + "learning_rate": 2.620782705681557e-07, + "loss": 0.5994, + "step": 32838 + }, + { + "epoch": 0.98, + "learning_rate": 2.613810944560258e-07, + "loss": 0.4997, + "step": 32839 + }, + { + "epoch": 0.98, + "learning_rate": 2.6068484567381977e-07, + "loss": 0.5015, + "step": 32840 + }, + { + "epoch": 0.98, + "learning_rate": 2.5998952422801037e-07, + "loss": 0.434, + "step": 32841 + }, + { + "epoch": 0.98, + "learning_rate": 2.592951301250812e-07, + "loss": 0.4692, + "step": 32842 + }, + { + "epoch": 0.98, + "learning_rate": 2.586016633714605e-07, + "loss": 0.5443, + "step": 32843 + }, + { + "epoch": 0.98, + "learning_rate": 2.579091239736098e-07, + "loss": 0.5737, + "step": 32844 + }, + { + "epoch": 0.98, + "learning_rate": 2.572175119379683e-07, + "loss": 0.5312, + "step": 32845 + }, + { + "epoch": 0.98, + "learning_rate": 2.5652682727097533e-07, + "loss": 0.4704, + "step": 32846 + }, + { + "epoch": 0.98, + "learning_rate": 2.558370699790258e-07, + "loss": 0.5441, + "step": 32847 + }, + { + "epoch": 0.98, + "learning_rate": 2.55148240068559e-07, + "loss": 0.477, + "step": 32848 + }, + { + "epoch": 0.98, + "learning_rate": 2.5446033754596977e-07, + "loss": 0.5152, + "step": 32849 + }, + { + "epoch": 0.98, + "learning_rate": 2.5377336241765304e-07, + "loss": 0.5388, + "step": 32850 + }, + { + "epoch": 0.98, + "learning_rate": 2.5308731468999256e-07, + "loss": 0.4861, + "step": 32851 + }, + { + "epoch": 0.98, + "learning_rate": 2.524021943693722e-07, + "loss": 0.5801, + "step": 32852 + }, + { + "epoch": 0.98, + "learning_rate": 2.517180014621645e-07, + "loss": 0.4869, + "step": 32853 + }, + { + "epoch": 0.98, + "learning_rate": 2.5103473597472006e-07, + "loss": 0.5383, + "step": 32854 + }, + { + "epoch": 0.98, + "learning_rate": 2.503523979134004e-07, + "loss": 0.6007, + "step": 32855 + }, + { + "epoch": 0.98, + "learning_rate": 2.4967098728455597e-07, + "loss": 0.627, + "step": 32856 + }, + { + "epoch": 0.98, + "learning_rate": 2.48990504094504e-07, + "loss": 0.5446, + "step": 32857 + }, + { + "epoch": 0.98, + "learning_rate": 2.483109483495949e-07, + "loss": 0.5065, + "step": 32858 + }, + { + "epoch": 0.98, + "learning_rate": 2.476323200561237e-07, + "loss": 0.5429, + "step": 32859 + }, + { + "epoch": 0.98, + "learning_rate": 2.469546192204075e-07, + "loss": 0.6094, + "step": 32860 + }, + { + "epoch": 0.98, + "learning_rate": 2.4627784584875245e-07, + "loss": 0.5076, + "step": 32861 + }, + { + "epoch": 0.98, + "learning_rate": 2.456019999474535e-07, + "loss": 0.4727, + "step": 32862 + }, + { + "epoch": 0.98, + "learning_rate": 2.4492708152278334e-07, + "loss": 0.6625, + "step": 32863 + }, + { + "epoch": 0.98, + "learning_rate": 2.4425309058101475e-07, + "loss": 0.4822, + "step": 32864 + }, + { + "epoch": 0.98, + "learning_rate": 2.4358002712843164e-07, + "loss": 0.4716, + "step": 32865 + }, + { + "epoch": 0.98, + "learning_rate": 2.4290789117128454e-07, + "loss": 0.6278, + "step": 32866 + }, + { + "epoch": 0.98, + "learning_rate": 2.422366827158129e-07, + "loss": 0.4944, + "step": 32867 + }, + { + "epoch": 0.98, + "learning_rate": 2.415664017682673e-07, + "loss": 0.5466, + "step": 32868 + }, + { + "epoch": 0.98, + "learning_rate": 2.408970483348871e-07, + "loss": 0.6079, + "step": 32869 + }, + { + "epoch": 0.98, + "learning_rate": 2.402286224218786e-07, + "loss": 0.5262, + "step": 32870 + }, + { + "epoch": 0.98, + "learning_rate": 2.3956112403547004e-07, + "loss": 0.5949, + "step": 32871 + }, + { + "epoch": 0.98, + "learning_rate": 2.388945531818454e-07, + "loss": 0.4704, + "step": 32872 + }, + { + "epoch": 0.98, + "learning_rate": 2.38228909867233e-07, + "loss": 0.4764, + "step": 32873 + }, + { + "epoch": 0.98, + "learning_rate": 2.3756419409780571e-07, + "loss": 0.487, + "step": 32874 + }, + { + "epoch": 0.98, + "learning_rate": 2.3690040587973638e-07, + "loss": 0.6366, + "step": 32875 + }, + { + "epoch": 0.98, + "learning_rate": 2.3623754521922003e-07, + "loss": 0.4657, + "step": 32876 + }, + { + "epoch": 0.98, + "learning_rate": 2.355756121223962e-07, + "loss": 0.5465, + "step": 32877 + }, + { + "epoch": 0.98, + "learning_rate": 2.3491460659542664e-07, + "loss": 0.5824, + "step": 32878 + }, + { + "epoch": 0.98, + "learning_rate": 2.3425452864446196e-07, + "loss": 0.5998, + "step": 32879 + }, + { + "epoch": 0.98, + "learning_rate": 2.335953782756306e-07, + "loss": 0.4719, + "step": 32880 + }, + { + "epoch": 0.98, + "learning_rate": 2.3293715549507212e-07, + "loss": 0.568, + "step": 32881 + }, + { + "epoch": 0.98, + "learning_rate": 2.322798603088927e-07, + "loss": 0.5595, + "step": 32882 + }, + { + "epoch": 0.98, + "learning_rate": 2.316234927232208e-07, + "loss": 0.5357, + "step": 32883 + }, + { + "epoch": 0.98, + "learning_rate": 2.3096805274414045e-07, + "loss": 0.5671, + "step": 32884 + }, + { + "epoch": 0.98, + "learning_rate": 2.3031354037776897e-07, + "loss": 0.4937, + "step": 32885 + }, + { + "epoch": 0.98, + "learning_rate": 2.296599556301682e-07, + "loss": 0.4818, + "step": 32886 + }, + { + "epoch": 0.98, + "learning_rate": 2.290072985074332e-07, + "loss": 0.4734, + "step": 32887 + }, + { + "epoch": 0.98, + "learning_rate": 2.2835556901561472e-07, + "loss": 0.3861, + "step": 32888 + }, + { + "epoch": 0.98, + "learning_rate": 2.2770476716078571e-07, + "loss": 0.573, + "step": 32889 + }, + { + "epoch": 0.98, + "learning_rate": 2.2705489294898575e-07, + "loss": 0.5732, + "step": 32890 + }, + { + "epoch": 0.98, + "learning_rate": 2.2640594638626556e-07, + "loss": 0.5106, + "step": 32891 + }, + { + "epoch": 0.98, + "learning_rate": 2.2575792747866475e-07, + "loss": 0.4881, + "step": 32892 + }, + { + "epoch": 0.98, + "learning_rate": 2.2511083623220074e-07, + "loss": 0.5761, + "step": 32893 + }, + { + "epoch": 0.98, + "learning_rate": 2.2446467265287986e-07, + "loss": 0.4487, + "step": 32894 + }, + { + "epoch": 0.98, + "learning_rate": 2.2381943674673056e-07, + "loss": 0.5456, + "step": 32895 + }, + { + "epoch": 0.98, + "learning_rate": 2.231751285197259e-07, + "loss": 0.4711, + "step": 32896 + }, + { + "epoch": 0.98, + "learning_rate": 2.2253174797788324e-07, + "loss": 0.5191, + "step": 32897 + }, + { + "epoch": 0.98, + "learning_rate": 2.218892951271756e-07, + "loss": 0.5117, + "step": 32898 + }, + { + "epoch": 0.98, + "learning_rate": 2.2124776997356488e-07, + "loss": 0.5933, + "step": 32899 + }, + { + "epoch": 0.98, + "learning_rate": 2.206071725230352e-07, + "loss": 0.515, + "step": 32900 + }, + { + "epoch": 0.98, + "learning_rate": 2.1996750278152622e-07, + "loss": 0.3775, + "step": 32901 + }, + { + "epoch": 0.98, + "learning_rate": 2.1932876075498876e-07, + "loss": 0.5135, + "step": 32902 + }, + { + "epoch": 0.98, + "learning_rate": 2.186909464493625e-07, + "loss": 0.5158, + "step": 32903 + }, + { + "epoch": 0.98, + "learning_rate": 2.1805405987058714e-07, + "loss": 0.4724, + "step": 32904 + }, + { + "epoch": 0.98, + "learning_rate": 2.1741810102456906e-07, + "loss": 0.5223, + "step": 32905 + }, + { + "epoch": 0.98, + "learning_rate": 2.1678306991722575e-07, + "loss": 0.5443, + "step": 32906 + }, + { + "epoch": 0.98, + "learning_rate": 2.161489665544747e-07, + "loss": 0.6502, + "step": 32907 + }, + { + "epoch": 0.98, + "learning_rate": 2.15515790942189e-07, + "loss": 0.6155, + "step": 32908 + }, + { + "epoch": 0.98, + "learning_rate": 2.14883543086275e-07, + "loss": 0.5813, + "step": 32909 + }, + { + "epoch": 0.98, + "learning_rate": 2.1425222299260585e-07, + "loss": 0.6095, + "step": 32910 + }, + { + "epoch": 0.98, + "learning_rate": 2.1362183066703233e-07, + "loss": 0.5053, + "step": 32911 + }, + { + "epoch": 0.98, + "learning_rate": 2.129923661154498e-07, + "loss": 0.5227, + "step": 32912 + }, + { + "epoch": 0.98, + "learning_rate": 2.1236382934368692e-07, + "loss": 0.5443, + "step": 32913 + }, + { + "epoch": 0.98, + "learning_rate": 2.1173622035759456e-07, + "loss": 0.6227, + "step": 32914 + }, + { + "epoch": 0.98, + "learning_rate": 2.1110953916300136e-07, + "loss": 0.5283, + "step": 32915 + }, + { + "epoch": 0.98, + "learning_rate": 2.1048378576574712e-07, + "loss": 0.5587, + "step": 32916 + }, + { + "epoch": 0.98, + "learning_rate": 2.098589601716383e-07, + "loss": 0.3956, + "step": 32917 + }, + { + "epoch": 0.98, + "learning_rate": 2.092350623864925e-07, + "loss": 0.5245, + "step": 32918 + }, + { + "epoch": 0.98, + "learning_rate": 2.0861209241610502e-07, + "loss": 0.482, + "step": 32919 + }, + { + "epoch": 0.98, + "learning_rate": 2.0799005026627129e-07, + "loss": 0.5375, + "step": 32920 + }, + { + "epoch": 0.98, + "learning_rate": 2.073689359427644e-07, + "loss": 0.5367, + "step": 32921 + }, + { + "epoch": 0.98, + "learning_rate": 2.0674874945136867e-07, + "loss": 0.5229, + "step": 32922 + }, + { + "epoch": 0.98, + "learning_rate": 2.0612949079785727e-07, + "loss": 0.4329, + "step": 32923 + }, + { + "epoch": 0.98, + "learning_rate": 2.055111599879811e-07, + "loss": 0.5057, + "step": 32924 + }, + { + "epoch": 0.98, + "learning_rate": 2.0489375702748003e-07, + "loss": 0.5159, + "step": 32925 + }, + { + "epoch": 0.98, + "learning_rate": 2.0427728192210504e-07, + "loss": 0.5261, + "step": 32926 + }, + { + "epoch": 0.98, + "learning_rate": 2.0366173467758488e-07, + "loss": 0.5243, + "step": 32927 + }, + { + "epoch": 0.98, + "learning_rate": 2.0304711529964825e-07, + "loss": 0.6437, + "step": 32928 + }, + { + "epoch": 0.98, + "learning_rate": 2.0243342379400177e-07, + "loss": 0.533, + "step": 32929 + }, + { + "epoch": 0.98, + "learning_rate": 2.0182066016635193e-07, + "loss": 0.5307, + "step": 32930 + }, + { + "epoch": 0.98, + "learning_rate": 2.012088244223942e-07, + "loss": 0.5496, + "step": 32931 + }, + { + "epoch": 0.98, + "learning_rate": 2.0059791656782401e-07, + "loss": 0.4727, + "step": 32932 + }, + { + "epoch": 0.98, + "learning_rate": 1.999879366083257e-07, + "loss": 0.5447, + "step": 32933 + }, + { + "epoch": 0.98, + "learning_rate": 1.9937888454955035e-07, + "loss": 0.6384, + "step": 32934 + }, + { + "epoch": 0.98, + "learning_rate": 1.9877076039718222e-07, + "loss": 0.5412, + "step": 32935 + }, + { + "epoch": 0.98, + "learning_rate": 1.9816356415686132e-07, + "loss": 0.6714, + "step": 32936 + }, + { + "epoch": 0.98, + "learning_rate": 1.9755729583424975e-07, + "loss": 0.4784, + "step": 32937 + }, + { + "epoch": 0.98, + "learning_rate": 1.9695195543496526e-07, + "loss": 0.4491, + "step": 32938 + }, + { + "epoch": 0.98, + "learning_rate": 1.9634754296464775e-07, + "loss": 0.5276, + "step": 32939 + }, + { + "epoch": 0.98, + "learning_rate": 1.9574405842890388e-07, + "loss": 0.5328, + "step": 32940 + }, + { + "epoch": 0.98, + "learning_rate": 1.9514150183335133e-07, + "loss": 0.6214, + "step": 32941 + }, + { + "epoch": 0.98, + "learning_rate": 1.945398731836079e-07, + "loss": 0.4927, + "step": 32942 + }, + { + "epoch": 0.98, + "learning_rate": 1.9393917248524685e-07, + "loss": 0.5648, + "step": 32943 + }, + { + "epoch": 0.98, + "learning_rate": 1.9333939974386373e-07, + "loss": 0.508, + "step": 32944 + }, + { + "epoch": 0.98, + "learning_rate": 1.9274055496503185e-07, + "loss": 0.4595, + "step": 32945 + }, + { + "epoch": 0.98, + "learning_rate": 1.9214263815431343e-07, + "loss": 0.5692, + "step": 32946 + }, + { + "epoch": 0.98, + "learning_rate": 1.915456493172818e-07, + "loss": 0.5215, + "step": 32947 + }, + { + "epoch": 0.98, + "learning_rate": 1.9094958845947698e-07, + "loss": 0.5821, + "step": 32948 + }, + { + "epoch": 0.98, + "learning_rate": 1.9035445558643894e-07, + "loss": 0.4919, + "step": 32949 + }, + { + "epoch": 0.98, + "learning_rate": 1.8976025070370773e-07, + "loss": 0.5251, + "step": 32950 + }, + { + "epoch": 0.98, + "learning_rate": 1.8916697381680115e-07, + "loss": 0.5273, + "step": 32951 + }, + { + "epoch": 0.98, + "learning_rate": 1.885746249312481e-07, + "loss": 0.3724, + "step": 32952 + }, + { + "epoch": 0.98, + "learning_rate": 1.879832040525442e-07, + "loss": 0.5353, + "step": 32953 + }, + { + "epoch": 0.98, + "learning_rate": 1.8739271118618507e-07, + "loss": 0.4529, + "step": 32954 + }, + { + "epoch": 0.98, + "learning_rate": 1.8680314633766628e-07, + "loss": 0.5892, + "step": 32955 + }, + { + "epoch": 0.98, + "learning_rate": 1.8621450951247232e-07, + "loss": 0.6197, + "step": 32956 + }, + { + "epoch": 0.98, + "learning_rate": 1.8562680071608773e-07, + "loss": 0.4399, + "step": 32957 + }, + { + "epoch": 0.98, + "learning_rate": 1.8504001995394149e-07, + "loss": 0.4779, + "step": 32958 + }, + { + "epoch": 0.98, + "learning_rate": 1.844541672315292e-07, + "loss": 0.5465, + "step": 32959 + }, + { + "epoch": 0.98, + "learning_rate": 1.8386924255426874e-07, + "loss": 0.5993, + "step": 32960 + }, + { + "epoch": 0.98, + "learning_rate": 1.832852459276113e-07, + "loss": 0.4942, + "step": 32961 + }, + { + "epoch": 0.98, + "learning_rate": 1.8270217735698592e-07, + "loss": 0.5527, + "step": 32962 + }, + { + "epoch": 0.98, + "learning_rate": 1.8212003684781043e-07, + "loss": 0.5351, + "step": 32963 + }, + { + "epoch": 0.98, + "learning_rate": 1.8153882440550275e-07, + "loss": 0.4704, + "step": 32964 + }, + { + "epoch": 0.98, + "learning_rate": 1.8095854003545854e-07, + "loss": 0.5078, + "step": 32965 + }, + { + "epoch": 0.98, + "learning_rate": 1.803791837430846e-07, + "loss": 0.5338, + "step": 32966 + }, + { + "epoch": 0.98, + "learning_rate": 1.7980075553375442e-07, + "loss": 0.4953, + "step": 32967 + }, + { + "epoch": 0.98, + "learning_rate": 1.7922325541285256e-07, + "loss": 0.471, + "step": 32968 + }, + { + "epoch": 0.98, + "learning_rate": 1.7864668338575253e-07, + "loss": 0.4974, + "step": 32969 + }, + { + "epoch": 0.98, + "learning_rate": 1.7807103945781667e-07, + "loss": 0.439, + "step": 32970 + }, + { + "epoch": 0.98, + "learning_rate": 1.7749632363437407e-07, + "loss": 0.5425, + "step": 32971 + }, + { + "epoch": 0.98, + "learning_rate": 1.7692253592079822e-07, + "loss": 0.5946, + "step": 32972 + }, + { + "epoch": 0.98, + "learning_rate": 1.7634967632240705e-07, + "loss": 0.5885, + "step": 32973 + }, + { + "epoch": 0.98, + "learning_rate": 1.757777448445297e-07, + "loss": 0.424, + "step": 32974 + }, + { + "epoch": 0.98, + "learning_rate": 1.7520674149249516e-07, + "loss": 0.5619, + "step": 32975 + }, + { + "epoch": 0.98, + "learning_rate": 1.7463666627158815e-07, + "loss": 0.5794, + "step": 32976 + }, + { + "epoch": 0.98, + "learning_rate": 1.740675191871266e-07, + "loss": 0.517, + "step": 32977 + }, + { + "epoch": 0.98, + "learning_rate": 1.7349930024440631e-07, + "loss": 0.5064, + "step": 32978 + }, + { + "epoch": 0.98, + "learning_rate": 1.729320094486897e-07, + "loss": 0.4951, + "step": 32979 + }, + { + "epoch": 0.98, + "learning_rate": 1.7236564680527256e-07, + "loss": 0.5899, + "step": 32980 + }, + { + "epoch": 0.98, + "learning_rate": 1.7180021231941733e-07, + "loss": 0.5085, + "step": 32981 + }, + { + "epoch": 0.98, + "learning_rate": 1.712357059963643e-07, + "loss": 0.4961, + "step": 32982 + }, + { + "epoch": 0.98, + "learning_rate": 1.7067212784137587e-07, + "loss": 0.4856, + "step": 32983 + }, + { + "epoch": 0.98, + "learning_rate": 1.7010947785969235e-07, + "loss": 0.5762, + "step": 32984 + }, + { + "epoch": 0.98, + "learning_rate": 1.6954775605655392e-07, + "loss": 0.5957, + "step": 32985 + }, + { + "epoch": 0.98, + "learning_rate": 1.689869624371565e-07, + "loss": 0.4803, + "step": 32986 + }, + { + "epoch": 0.98, + "learning_rate": 1.684270970067403e-07, + "loss": 0.5225, + "step": 32987 + }, + { + "epoch": 0.98, + "learning_rate": 1.6786815977049007e-07, + "loss": 0.5406, + "step": 32988 + }, + { + "epoch": 0.98, + "learning_rate": 1.6731015073361277e-07, + "loss": 0.5657, + "step": 32989 + }, + { + "epoch": 0.98, + "learning_rate": 1.6675306990130425e-07, + "loss": 0.4911, + "step": 32990 + }, + { + "epoch": 0.98, + "learning_rate": 1.66196917278727e-07, + "loss": 0.4399, + "step": 32991 + }, + { + "epoch": 0.98, + "learning_rate": 1.656416928710658e-07, + "loss": 0.6224, + "step": 32992 + }, + { + "epoch": 0.98, + "learning_rate": 1.6508739668348318e-07, + "loss": 0.5267, + "step": 32993 + }, + { + "epoch": 0.98, + "learning_rate": 1.6453402872111944e-07, + "loss": 0.5645, + "step": 32994 + }, + { + "epoch": 0.98, + "learning_rate": 1.6398158898913718e-07, + "loss": 0.417, + "step": 32995 + }, + { + "epoch": 0.98, + "learning_rate": 1.634300774926545e-07, + "loss": 0.5829, + "step": 32996 + }, + { + "epoch": 0.98, + "learning_rate": 1.628794942368117e-07, + "loss": 0.5638, + "step": 32997 + }, + { + "epoch": 0.98, + "learning_rate": 1.6232983922672695e-07, + "loss": 0.4959, + "step": 32998 + }, + { + "epoch": 0.98, + "learning_rate": 1.6178111246749617e-07, + "loss": 0.502, + "step": 32999 + }, + { + "epoch": 0.98, + "learning_rate": 1.6123331396424857e-07, + "loss": 0.5207, + "step": 33000 + }, + { + "epoch": 0.98, + "learning_rate": 1.6068644372205787e-07, + "loss": 0.4955, + "step": 33001 + }, + { + "epoch": 0.98, + "learning_rate": 1.601405017460089e-07, + "loss": 0.5864, + "step": 33002 + }, + { + "epoch": 0.98, + "learning_rate": 1.595954880411754e-07, + "loss": 0.6368, + "step": 33003 + }, + { + "epoch": 0.98, + "learning_rate": 1.5905140261263107e-07, + "loss": 0.4416, + "step": 33004 + }, + { + "epoch": 0.98, + "learning_rate": 1.5850824546544963e-07, + "loss": 0.5416, + "step": 33005 + }, + { + "epoch": 0.98, + "learning_rate": 1.5796601660464926e-07, + "loss": 0.4857, + "step": 33006 + }, + { + "epoch": 0.98, + "learning_rate": 1.5742471603530372e-07, + "loss": 0.5584, + "step": 33007 + }, + { + "epoch": 0.98, + "learning_rate": 1.5688434376240903e-07, + "loss": 0.5302, + "step": 33008 + }, + { + "epoch": 0.98, + "learning_rate": 1.5634489979102773e-07, + "loss": 0.5438, + "step": 33009 + }, + { + "epoch": 0.98, + "learning_rate": 1.558063841261559e-07, + "loss": 0.6763, + "step": 33010 + }, + { + "epoch": 0.98, + "learning_rate": 1.552687967728006e-07, + "loss": 0.4149, + "step": 33011 + }, + { + "epoch": 0.98, + "learning_rate": 1.5473213773595784e-07, + "loss": 0.5745, + "step": 33012 + }, + { + "epoch": 0.98, + "learning_rate": 1.5419640702062364e-07, + "loss": 0.5553, + "step": 33013 + }, + { + "epoch": 0.98, + "learning_rate": 1.5366160463177181e-07, + "loss": 0.6314, + "step": 33014 + }, + { + "epoch": 0.98, + "learning_rate": 1.5312773057438722e-07, + "loss": 0.4658, + "step": 33015 + }, + { + "epoch": 0.98, + "learning_rate": 1.525947848534215e-07, + "loss": 0.5386, + "step": 33016 + }, + { + "epoch": 0.98, + "learning_rate": 1.520627674738373e-07, + "loss": 0.4147, + "step": 33017 + }, + { + "epoch": 0.98, + "learning_rate": 1.5153167844057514e-07, + "loss": 0.4022, + "step": 33018 + }, + { + "epoch": 0.98, + "learning_rate": 1.510015177585755e-07, + "loss": 0.6797, + "step": 33019 + }, + { + "epoch": 0.98, + "learning_rate": 1.5047228543276782e-07, + "loss": 0.4984, + "step": 33020 + }, + { + "epoch": 0.98, + "learning_rate": 1.4994398146807032e-07, + "loss": 0.5866, + "step": 33021 + }, + { + "epoch": 0.98, + "learning_rate": 1.494166058694013e-07, + "loss": 0.4408, + "step": 33022 + }, + { + "epoch": 0.98, + "learning_rate": 1.488901586416569e-07, + "loss": 0.6601, + "step": 33023 + }, + { + "epoch": 0.98, + "learning_rate": 1.4836463978974425e-07, + "loss": 0.526, + "step": 33024 + }, + { + "epoch": 0.98, + "learning_rate": 1.4784004931853722e-07, + "loss": 0.4876, + "step": 33025 + }, + { + "epoch": 0.98, + "learning_rate": 1.4731638723290976e-07, + "loss": 0.5194, + "step": 33026 + }, + { + "epoch": 0.98, + "learning_rate": 1.4679365353773566e-07, + "loss": 0.4983, + "step": 33027 + }, + { + "epoch": 0.98, + "learning_rate": 1.4627184823788886e-07, + "loss": 0.5357, + "step": 33028 + }, + { + "epoch": 0.98, + "learning_rate": 1.457509713381988e-07, + "loss": 0.4684, + "step": 33029 + }, + { + "epoch": 0.98, + "learning_rate": 1.452310228435172e-07, + "loss": 0.5678, + "step": 33030 + }, + { + "epoch": 0.98, + "learning_rate": 1.4471200275868457e-07, + "loss": 0.5816, + "step": 33031 + }, + { + "epoch": 0.98, + "learning_rate": 1.441939110885193e-07, + "loss": 0.4731, + "step": 33032 + }, + { + "epoch": 0.98, + "learning_rate": 1.4367674783783979e-07, + "loss": 0.5037, + "step": 33033 + }, + { + "epoch": 0.98, + "learning_rate": 1.431605130114533e-07, + "loss": 0.5739, + "step": 33034 + }, + { + "epoch": 0.98, + "learning_rate": 1.4264520661416702e-07, + "loss": 0.5724, + "step": 33035 + }, + { + "epoch": 0.98, + "learning_rate": 1.421308286507661e-07, + "loss": 0.6271, + "step": 33036 + }, + { + "epoch": 0.98, + "learning_rate": 1.4161737912602446e-07, + "loss": 0.5366, + "step": 33037 + }, + { + "epoch": 0.98, + "learning_rate": 1.4110485804473827e-07, + "loss": 0.4896, + "step": 33038 + }, + { + "epoch": 0.98, + "learning_rate": 1.405932654116482e-07, + "loss": 0.6418, + "step": 33039 + }, + { + "epoch": 0.98, + "learning_rate": 1.4008260123152818e-07, + "loss": 0.532, + "step": 33040 + }, + { + "epoch": 0.98, + "learning_rate": 1.3957286550911886e-07, + "loss": 0.4998, + "step": 33041 + }, + { + "epoch": 0.98, + "learning_rate": 1.390640582491498e-07, + "loss": 0.5222, + "step": 33042 + }, + { + "epoch": 0.98, + "learning_rate": 1.3855617945637278e-07, + "loss": 0.5828, + "step": 33043 + }, + { + "epoch": 0.98, + "learning_rate": 1.3804922913550622e-07, + "loss": 0.4982, + "step": 33044 + }, + { + "epoch": 0.98, + "learning_rate": 1.3754320729124636e-07, + "loss": 0.5403, + "step": 33045 + }, + { + "epoch": 0.98, + "learning_rate": 1.3703811392831166e-07, + "loss": 0.4719, + "step": 33046 + }, + { + "epoch": 0.98, + "learning_rate": 1.3653394905139838e-07, + "loss": 0.5477, + "step": 33047 + }, + { + "epoch": 0.98, + "learning_rate": 1.3603071266518053e-07, + "loss": 0.5498, + "step": 33048 + }, + { + "epoch": 0.98, + "learning_rate": 1.3552840477435435e-07, + "loss": 0.6136, + "step": 33049 + }, + { + "epoch": 0.98, + "learning_rate": 1.350270253835828e-07, + "loss": 0.4926, + "step": 33050 + }, + { + "epoch": 0.98, + "learning_rate": 1.3452657449752882e-07, + "loss": 0.554, + "step": 33051 + }, + { + "epoch": 0.98, + "learning_rate": 1.3402705212084422e-07, + "loss": 0.5276, + "step": 33052 + }, + { + "epoch": 0.98, + "learning_rate": 1.3352845825816974e-07, + "loss": 0.5505, + "step": 33053 + }, + { + "epoch": 0.98, + "learning_rate": 1.330307929141461e-07, + "loss": 0.5006, + "step": 33054 + }, + { + "epoch": 0.98, + "learning_rate": 1.3253405609340296e-07, + "loss": 0.5603, + "step": 33055 + }, + { + "epoch": 0.98, + "learning_rate": 1.3203824780055885e-07, + "loss": 0.4919, + "step": 33056 + }, + { + "epoch": 0.98, + "learning_rate": 1.3154336804021007e-07, + "loss": 0.4117, + "step": 33057 + }, + { + "epoch": 0.98, + "learning_rate": 1.3104941681696404e-07, + "loss": 0.5791, + "step": 33058 + }, + { + "epoch": 0.98, + "learning_rate": 1.305563941354282e-07, + "loss": 0.5454, + "step": 33059 + }, + { + "epoch": 0.98, + "learning_rate": 1.3006430000016556e-07, + "loss": 0.467, + "step": 33060 + }, + { + "epoch": 0.98, + "learning_rate": 1.2957313441577245e-07, + "loss": 0.5006, + "step": 33061 + }, + { + "epoch": 0.98, + "learning_rate": 1.290828973867897e-07, + "loss": 0.5641, + "step": 33062 + }, + { + "epoch": 0.98, + "learning_rate": 1.2859358891780248e-07, + "loss": 0.4372, + "step": 33063 + }, + { + "epoch": 0.98, + "learning_rate": 1.2810520901334054e-07, + "loss": 0.5356, + "step": 33064 + }, + { + "epoch": 0.98, + "learning_rate": 1.2761775767795582e-07, + "loss": 0.4405, + "step": 33065 + }, + { + "epoch": 0.98, + "learning_rate": 1.2713123491616685e-07, + "loss": 0.5976, + "step": 33066 + }, + { + "epoch": 0.98, + "learning_rate": 1.2664564073251452e-07, + "loss": 0.4281, + "step": 33067 + }, + { + "epoch": 0.98, + "learning_rate": 1.2616097513150626e-07, + "loss": 0.572, + "step": 33068 + }, + { + "epoch": 0.98, + "learning_rate": 1.2567723811763854e-07, + "loss": 0.4991, + "step": 33069 + }, + { + "epoch": 0.98, + "learning_rate": 1.2519442969541884e-07, + "loss": 0.4676, + "step": 33070 + }, + { + "epoch": 0.98, + "learning_rate": 1.2471254986934355e-07, + "loss": 0.5733, + "step": 33071 + }, + { + "epoch": 0.98, + "learning_rate": 1.2423159864388689e-07, + "loss": 0.4299, + "step": 33072 + }, + { + "epoch": 0.98, + "learning_rate": 1.2375157602350086e-07, + "loss": 0.4604, + "step": 33073 + }, + { + "epoch": 0.98, + "learning_rate": 1.2327248201268183e-07, + "loss": 0.5569, + "step": 33074 + }, + { + "epoch": 0.98, + "learning_rate": 1.227943166158596e-07, + "loss": 0.4212, + "step": 33075 + }, + { + "epoch": 0.98, + "learning_rate": 1.223170798374973e-07, + "loss": 0.5437, + "step": 33076 + }, + { + "epoch": 0.98, + "learning_rate": 1.2184077168201357e-07, + "loss": 0.4852, + "step": 33077 + }, + { + "epoch": 0.98, + "learning_rate": 1.2136539215386046e-07, + "loss": 0.513, + "step": 33078 + }, + { + "epoch": 0.98, + "learning_rate": 1.2089094125743437e-07, + "loss": 0.555, + "step": 33079 + }, + { + "epoch": 0.98, + "learning_rate": 1.204174189971541e-07, + "loss": 0.5054, + "step": 33080 + }, + { + "epoch": 0.98, + "learning_rate": 1.1994482537742711e-07, + "loss": 0.5686, + "step": 33081 + }, + { + "epoch": 0.98, + "learning_rate": 1.1947316040265e-07, + "loss": 0.6201, + "step": 33082 + }, + { + "epoch": 0.98, + "learning_rate": 1.1900242407719697e-07, + "loss": 0.5153, + "step": 33083 + }, + { + "epoch": 0.98, + "learning_rate": 1.1853261640545343e-07, + "loss": 0.528, + "step": 33084 + }, + { + "epoch": 0.98, + "learning_rate": 1.1806373739178256e-07, + "loss": 0.5588, + "step": 33085 + }, + { + "epoch": 0.99, + "learning_rate": 1.1759578704054752e-07, + "loss": 0.5798, + "step": 33086 + }, + { + "epoch": 0.99, + "learning_rate": 1.171287653561004e-07, + "loss": 0.5634, + "step": 33087 + }, + { + "epoch": 0.99, + "learning_rate": 1.1666267234278217e-07, + "loss": 0.5676, + "step": 33088 + }, + { + "epoch": 0.99, + "learning_rate": 1.1619750800491158e-07, + "loss": 0.4425, + "step": 33089 + }, + { + "epoch": 0.99, + "learning_rate": 1.1573327234684072e-07, + "loss": 0.4889, + "step": 33090 + }, + { + "epoch": 0.99, + "learning_rate": 1.1526996537286616e-07, + "loss": 0.4632, + "step": 33091 + }, + { + "epoch": 0.99, + "learning_rate": 1.1480758708730665e-07, + "loss": 0.3765, + "step": 33092 + }, + { + "epoch": 0.99, + "learning_rate": 1.1434613749444767e-07, + "loss": 0.4339, + "step": 33093 + }, + { + "epoch": 0.99, + "learning_rate": 1.1388561659859686e-07, + "loss": 0.5416, + "step": 33094 + }, + { + "epoch": 0.99, + "learning_rate": 1.134260244040175e-07, + "loss": 0.5288, + "step": 33095 + }, + { + "epoch": 0.99, + "learning_rate": 1.1296736091499505e-07, + "loss": 0.5851, + "step": 33096 + }, + { + "epoch": 0.99, + "learning_rate": 1.1250962613579275e-07, + "loss": 0.4158, + "step": 33097 + }, + { + "epoch": 0.99, + "learning_rate": 1.1205282007066276e-07, + "loss": 0.5396, + "step": 33098 + }, + { + "epoch": 0.99, + "learning_rate": 1.1159694272385724e-07, + "loss": 0.6208, + "step": 33099 + }, + { + "epoch": 0.99, + "learning_rate": 1.1114199409961722e-07, + "loss": 0.5342, + "step": 33100 + }, + { + "epoch": 0.99, + "learning_rate": 1.106879742021616e-07, + "loss": 0.4746, + "step": 33101 + }, + { + "epoch": 0.99, + "learning_rate": 1.1023488303570917e-07, + "loss": 0.5338, + "step": 33102 + }, + { + "epoch": 0.99, + "learning_rate": 1.0978272060448991e-07, + "loss": 0.6448, + "step": 33103 + }, + { + "epoch": 0.99, + "learning_rate": 1.0933148691270045e-07, + "loss": 0.5241, + "step": 33104 + }, + { + "epoch": 0.99, + "learning_rate": 1.0888118196452635e-07, + "loss": 0.6194, + "step": 33105 + }, + { + "epoch": 0.99, + "learning_rate": 1.0843180576416423e-07, + "loss": 0.4937, + "step": 33106 + }, + { + "epoch": 0.99, + "learning_rate": 1.0798335831578854e-07, + "loss": 0.4397, + "step": 33107 + }, + { + "epoch": 0.99, + "learning_rate": 1.0753583962357372e-07, + "loss": 0.496, + "step": 33108 + }, + { + "epoch": 0.99, + "learning_rate": 1.070892496916831e-07, + "loss": 0.637, + "step": 33109 + }, + { + "epoch": 0.99, + "learning_rate": 1.0664358852425782e-07, + "loss": 0.5376, + "step": 33110 + }, + { + "epoch": 0.99, + "learning_rate": 1.0619885612545011e-07, + "loss": 0.6783, + "step": 33111 + }, + { + "epoch": 0.99, + "learning_rate": 1.0575505249938999e-07, + "loss": 0.5706, + "step": 33112 + }, + { + "epoch": 0.99, + "learning_rate": 1.053121776501964e-07, + "loss": 0.5387, + "step": 33113 + }, + { + "epoch": 0.99, + "learning_rate": 1.0487023158201048e-07, + "loss": 0.4735, + "step": 33114 + }, + { + "epoch": 0.99, + "learning_rate": 1.0442921429891783e-07, + "loss": 0.5512, + "step": 33115 + }, + { + "epoch": 0.99, + "learning_rate": 1.0398912580502629e-07, + "loss": 0.6024, + "step": 33116 + }, + { + "epoch": 0.99, + "learning_rate": 1.0354996610443258e-07, + "loss": 0.6049, + "step": 33117 + }, + { + "epoch": 0.99, + "learning_rate": 1.0311173520122231e-07, + "loss": 0.5926, + "step": 33118 + }, + { + "epoch": 0.99, + "learning_rate": 1.026744330994589e-07, + "loss": 0.5339, + "step": 33119 + }, + { + "epoch": 0.99, + "learning_rate": 1.0223805980321688e-07, + "loss": 0.5042, + "step": 33120 + }, + { + "epoch": 0.99, + "learning_rate": 1.0180261531654856e-07, + "loss": 0.5651, + "step": 33121 + }, + { + "epoch": 0.99, + "learning_rate": 1.0136809964349514e-07, + "loss": 0.6158, + "step": 33122 + }, + { + "epoch": 0.99, + "learning_rate": 1.0093451278812005e-07, + "loss": 0.5825, + "step": 33123 + }, + { + "epoch": 0.99, + "learning_rate": 1.0050185475443118e-07, + "loss": 0.6027, + "step": 33124 + }, + { + "epoch": 0.99, + "learning_rate": 1.0007012554645867e-07, + "loss": 0.563, + "step": 33125 + }, + { + "epoch": 0.99, + "learning_rate": 9.96393251682215e-08, + "loss": 0.579, + "step": 33126 + }, + { + "epoch": 0.99, + "learning_rate": 9.920945362371647e-08, + "loss": 0.4957, + "step": 33127 + }, + { + "epoch": 0.99, + "learning_rate": 9.87805109169515e-08, + "loss": 0.3629, + "step": 33128 + }, + { + "epoch": 0.99, + "learning_rate": 9.835249705191229e-08, + "loss": 0.517, + "step": 33129 + }, + { + "epoch": 0.99, + "learning_rate": 9.792541203256234e-08, + "loss": 0.5659, + "step": 33130 + }, + { + "epoch": 0.99, + "learning_rate": 9.749925586289843e-08, + "loss": 0.5425, + "step": 33131 + }, + { + "epoch": 0.99, + "learning_rate": 9.707402854686187e-08, + "loss": 0.4781, + "step": 33132 + }, + { + "epoch": 0.99, + "learning_rate": 9.664973008841615e-08, + "loss": 0.4763, + "step": 33133 + }, + { + "epoch": 0.99, + "learning_rate": 9.622636049150258e-08, + "loss": 0.5093, + "step": 33134 + }, + { + "epoch": 0.99, + "learning_rate": 9.580391976005132e-08, + "loss": 0.589, + "step": 33135 + }, + { + "epoch": 0.99, + "learning_rate": 9.538240789800368e-08, + "loss": 0.5554, + "step": 33136 + }, + { + "epoch": 0.99, + "learning_rate": 9.496182490927874e-08, + "loss": 0.4883, + "step": 33137 + }, + { + "epoch": 0.99, + "learning_rate": 9.454217079777338e-08, + "loss": 0.4545, + "step": 33138 + }, + { + "epoch": 0.99, + "learning_rate": 9.412344556739561e-08, + "loss": 0.52, + "step": 33139 + }, + { + "epoch": 0.99, + "learning_rate": 9.370564922204228e-08, + "loss": 0.5356, + "step": 33140 + }, + { + "epoch": 0.99, + "learning_rate": 9.328878176559918e-08, + "loss": 0.5184, + "step": 33141 + }, + { + "epoch": 0.99, + "learning_rate": 9.28728432019299e-08, + "loss": 0.6025, + "step": 33142 + }, + { + "epoch": 0.99, + "learning_rate": 9.245783353490911e-08, + "loss": 0.5664, + "step": 33143 + }, + { + "epoch": 0.99, + "learning_rate": 9.204375276840038e-08, + "loss": 0.4791, + "step": 33144 + }, + { + "epoch": 0.99, + "learning_rate": 9.16306009062562e-08, + "loss": 0.6856, + "step": 33145 + }, + { + "epoch": 0.99, + "learning_rate": 9.121837795229571e-08, + "loss": 0.3941, + "step": 33146 + }, + { + "epoch": 0.99, + "learning_rate": 9.080708391038251e-08, + "loss": 0.591, + "step": 33147 + }, + { + "epoch": 0.99, + "learning_rate": 9.039671878431355e-08, + "loss": 0.5837, + "step": 33148 + }, + { + "epoch": 0.99, + "learning_rate": 8.99872825779191e-08, + "loss": 0.4625, + "step": 33149 + }, + { + "epoch": 0.99, + "learning_rate": 8.957877529500724e-08, + "loss": 0.5199, + "step": 33150 + }, + { + "epoch": 0.99, + "learning_rate": 8.917119693936382e-08, + "loss": 0.448, + "step": 33151 + }, + { + "epoch": 0.99, + "learning_rate": 8.87645475147969e-08, + "loss": 0.6024, + "step": 33152 + }, + { + "epoch": 0.99, + "learning_rate": 8.835882702507015e-08, + "loss": 0.4354, + "step": 33153 + }, + { + "epoch": 0.99, + "learning_rate": 8.795403547395831e-08, + "loss": 0.5339, + "step": 33154 + }, + { + "epoch": 0.99, + "learning_rate": 8.755017286523615e-08, + "loss": 0.5167, + "step": 33155 + }, + { + "epoch": 0.99, + "learning_rate": 8.714723920264511e-08, + "loss": 0.639, + "step": 33156 + }, + { + "epoch": 0.99, + "learning_rate": 8.674523448994887e-08, + "loss": 0.4951, + "step": 33157 + }, + { + "epoch": 0.99, + "learning_rate": 8.634415873086665e-08, + "loss": 0.4396, + "step": 33158 + }, + { + "epoch": 0.99, + "learning_rate": 8.594401192912882e-08, + "loss": 0.5991, + "step": 33159 + }, + { + "epoch": 0.99, + "learning_rate": 8.554479408846572e-08, + "loss": 0.55, + "step": 33160 + }, + { + "epoch": 0.99, + "learning_rate": 8.514650521258549e-08, + "loss": 0.5801, + "step": 33161 + }, + { + "epoch": 0.99, + "learning_rate": 8.474914530519629e-08, + "loss": 0.56, + "step": 33162 + }, + { + "epoch": 0.99, + "learning_rate": 8.435271436998404e-08, + "loss": 0.4388, + "step": 33163 + }, + { + "epoch": 0.99, + "learning_rate": 8.39572124106347e-08, + "loss": 0.6586, + "step": 33164 + }, + { + "epoch": 0.99, + "learning_rate": 8.356263943082309e-08, + "loss": 0.5256, + "step": 33165 + }, + { + "epoch": 0.99, + "learning_rate": 8.316899543423518e-08, + "loss": 0.5156, + "step": 33166 + }, + { + "epoch": 0.99, + "learning_rate": 8.277628042451246e-08, + "loss": 0.6845, + "step": 33167 + }, + { + "epoch": 0.99, + "learning_rate": 8.23844944053187e-08, + "loss": 0.4802, + "step": 33168 + }, + { + "epoch": 0.99, + "learning_rate": 8.199363738028432e-08, + "loss": 0.5101, + "step": 33169 + }, + { + "epoch": 0.99, + "learning_rate": 8.160370935305083e-08, + "loss": 0.5057, + "step": 33170 + }, + { + "epoch": 0.99, + "learning_rate": 8.12147103272487e-08, + "loss": 0.4596, + "step": 33171 + }, + { + "epoch": 0.99, + "learning_rate": 8.082664030648612e-08, + "loss": 0.4248, + "step": 33172 + }, + { + "epoch": 0.99, + "learning_rate": 8.043949929437134e-08, + "loss": 0.458, + "step": 33173 + }, + { + "epoch": 0.99, + "learning_rate": 8.005328729450146e-08, + "loss": 0.5417, + "step": 33174 + }, + { + "epoch": 0.99, + "learning_rate": 7.966800431047361e-08, + "loss": 0.5886, + "step": 33175 + }, + { + "epoch": 0.99, + "learning_rate": 7.928365034587382e-08, + "loss": 0.4599, + "step": 33176 + }, + { + "epoch": 0.99, + "learning_rate": 7.89002254042659e-08, + "loss": 0.5225, + "step": 33177 + }, + { + "epoch": 0.99, + "learning_rate": 7.851772948922476e-08, + "loss": 0.4493, + "step": 33178 + }, + { + "epoch": 0.99, + "learning_rate": 7.813616260429202e-08, + "loss": 0.4869, + "step": 33179 + }, + { + "epoch": 0.99, + "learning_rate": 7.77555247530315e-08, + "loss": 0.5761, + "step": 33180 + }, + { + "epoch": 0.99, + "learning_rate": 7.73758159389737e-08, + "loss": 0.653, + "step": 33181 + }, + { + "epoch": 0.99, + "learning_rate": 7.699703616564913e-08, + "loss": 0.4121, + "step": 33182 + }, + { + "epoch": 0.99, + "learning_rate": 7.661918543657721e-08, + "loss": 0.4941, + "step": 33183 + }, + { + "epoch": 0.99, + "learning_rate": 7.624226375527733e-08, + "loss": 0.5679, + "step": 33184 + }, + { + "epoch": 0.99, + "learning_rate": 7.586627112525779e-08, + "loss": 0.489, + "step": 33185 + }, + { + "epoch": 0.99, + "learning_rate": 7.549120755000472e-08, + "loss": 0.3705, + "step": 33186 + }, + { + "epoch": 0.99, + "learning_rate": 7.51170730330042e-08, + "loss": 0.5491, + "step": 33187 + }, + { + "epoch": 0.99, + "learning_rate": 7.474386757774232e-08, + "loss": 0.4741, + "step": 33188 + }, + { + "epoch": 0.99, + "learning_rate": 7.4371591187683e-08, + "loss": 0.5973, + "step": 33189 + }, + { + "epoch": 0.99, + "learning_rate": 7.400024386629012e-08, + "loss": 0.5127, + "step": 33190 + }, + { + "epoch": 0.99, + "learning_rate": 7.362982561701648e-08, + "loss": 0.4986, + "step": 33191 + }, + { + "epoch": 0.99, + "learning_rate": 7.326033644330377e-08, + "loss": 0.4584, + "step": 33192 + }, + { + "epoch": 0.99, + "learning_rate": 7.289177634859368e-08, + "loss": 0.4791, + "step": 33193 + }, + { + "epoch": 0.99, + "learning_rate": 7.252414533630569e-08, + "loss": 0.5972, + "step": 33194 + }, + { + "epoch": 0.99, + "learning_rate": 7.215744340987041e-08, + "loss": 0.5134, + "step": 33195 + }, + { + "epoch": 0.99, + "learning_rate": 7.179167057267399e-08, + "loss": 0.6058, + "step": 33196 + }, + { + "epoch": 0.99, + "learning_rate": 7.142682682813595e-08, + "loss": 0.5876, + "step": 33197 + }, + { + "epoch": 0.99, + "learning_rate": 7.106291217965355e-08, + "loss": 0.4465, + "step": 33198 + }, + { + "epoch": 0.99, + "learning_rate": 7.069992663059078e-08, + "loss": 0.6266, + "step": 33199 + }, + { + "epoch": 0.99, + "learning_rate": 7.03378701843338e-08, + "loss": 0.6431, + "step": 33200 + }, + { + "epoch": 0.99, + "learning_rate": 6.99767428442466e-08, + "loss": 0.5545, + "step": 33201 + }, + { + "epoch": 0.99, + "learning_rate": 6.961654461369315e-08, + "loss": 0.5611, + "step": 33202 + }, + { + "epoch": 0.99, + "learning_rate": 6.925727549600414e-08, + "loss": 0.4963, + "step": 33203 + }, + { + "epoch": 0.99, + "learning_rate": 6.889893549454351e-08, + "loss": 0.4226, + "step": 33204 + }, + { + "epoch": 0.99, + "learning_rate": 6.854152461263086e-08, + "loss": 0.5138, + "step": 33205 + }, + { + "epoch": 0.99, + "learning_rate": 6.818504285358573e-08, + "loss": 0.449, + "step": 33206 + }, + { + "epoch": 0.99, + "learning_rate": 6.782949022072772e-08, + "loss": 0.5021, + "step": 33207 + }, + { + "epoch": 0.99, + "learning_rate": 6.747486671736524e-08, + "loss": 0.5574, + "step": 33208 + }, + { + "epoch": 0.99, + "learning_rate": 6.712117234679571e-08, + "loss": 0.596, + "step": 33209 + }, + { + "epoch": 0.99, + "learning_rate": 6.676840711229427e-08, + "loss": 0.537, + "step": 33210 + }, + { + "epoch": 0.99, + "learning_rate": 6.641657101715826e-08, + "loss": 0.5809, + "step": 33211 + }, + { + "epoch": 0.99, + "learning_rate": 6.606566406465175e-08, + "loss": 0.5552, + "step": 33212 + }, + { + "epoch": 0.99, + "learning_rate": 6.57156862580277e-08, + "loss": 0.4837, + "step": 33213 + }, + { + "epoch": 0.99, + "learning_rate": 6.536663760056127e-08, + "loss": 0.5095, + "step": 33214 + }, + { + "epoch": 0.99, + "learning_rate": 6.501851809548321e-08, + "loss": 0.5185, + "step": 33215 + }, + { + "epoch": 0.99, + "learning_rate": 6.467132774602425e-08, + "loss": 0.5557, + "step": 33216 + }, + { + "epoch": 0.99, + "learning_rate": 6.432506655542625e-08, + "loss": 0.4977, + "step": 33217 + }, + { + "epoch": 0.99, + "learning_rate": 6.397973452690886e-08, + "loss": 0.4886, + "step": 33218 + }, + { + "epoch": 0.99, + "learning_rate": 6.363533166366953e-08, + "loss": 0.5153, + "step": 33219 + }, + { + "epoch": 0.99, + "learning_rate": 6.329185796892789e-08, + "loss": 0.4173, + "step": 33220 + }, + { + "epoch": 0.99, + "learning_rate": 6.294931344585919e-08, + "loss": 0.5242, + "step": 33221 + }, + { + "epoch": 0.99, + "learning_rate": 6.260769809766088e-08, + "loss": 0.582, + "step": 33222 + }, + { + "epoch": 0.99, + "learning_rate": 6.226701192749706e-08, + "loss": 0.4871, + "step": 33223 + }, + { + "epoch": 0.99, + "learning_rate": 6.192725493855411e-08, + "loss": 0.5764, + "step": 33224 + }, + { + "epoch": 0.99, + "learning_rate": 6.158842713397395e-08, + "loss": 0.6014, + "step": 33225 + }, + { + "epoch": 0.99, + "learning_rate": 6.12505285169096e-08, + "loss": 0.5308, + "step": 33226 + }, + { + "epoch": 0.99, + "learning_rate": 6.09135590905141e-08, + "loss": 0.4709, + "step": 33227 + }, + { + "epoch": 0.99, + "learning_rate": 6.05775188579072e-08, + "loss": 0.5622, + "step": 33228 + }, + { + "epoch": 0.99, + "learning_rate": 6.024240782221968e-08, + "loss": 0.4396, + "step": 33229 + }, + { + "epoch": 0.99, + "learning_rate": 5.990822598657131e-08, + "loss": 0.52, + "step": 33230 + }, + { + "epoch": 0.99, + "learning_rate": 5.957497335405959e-08, + "loss": 0.6081, + "step": 33231 + }, + { + "epoch": 0.99, + "learning_rate": 5.9242649927782054e-08, + "loss": 0.5075, + "step": 33232 + }, + { + "epoch": 0.99, + "learning_rate": 5.8911255710847323e-08, + "loss": 0.4868, + "step": 33233 + }, + { + "epoch": 0.99, + "learning_rate": 5.858079070631961e-08, + "loss": 0.544, + "step": 33234 + }, + { + "epoch": 0.99, + "learning_rate": 5.8251254917263134e-08, + "loss": 0.5207, + "step": 33235 + }, + { + "epoch": 0.99, + "learning_rate": 5.7922648346764306e-08, + "loss": 0.4399, + "step": 33236 + }, + { + "epoch": 0.99, + "learning_rate": 5.759497099786515e-08, + "loss": 0.5639, + "step": 33237 + }, + { + "epoch": 0.99, + "learning_rate": 5.726822287360767e-08, + "loss": 0.503, + "step": 33238 + }, + { + "epoch": 0.99, + "learning_rate": 5.6942403977044975e-08, + "loss": 0.6803, + "step": 33239 + }, + { + "epoch": 0.99, + "learning_rate": 5.661751431119688e-08, + "loss": 0.4598, + "step": 33240 + }, + { + "epoch": 0.99, + "learning_rate": 5.6293553879094297e-08, + "loss": 0.4583, + "step": 33241 + }, + { + "epoch": 0.99, + "learning_rate": 5.5970522683734814e-08, + "loss": 0.499, + "step": 33242 + }, + { + "epoch": 0.99, + "learning_rate": 5.5648420728116044e-08, + "loss": 0.5024, + "step": 33243 + }, + { + "epoch": 0.99, + "learning_rate": 5.532724801525779e-08, + "loss": 0.5828, + "step": 33244 + }, + { + "epoch": 0.99, + "learning_rate": 5.500700454812435e-08, + "loss": 0.6214, + "step": 33245 + }, + { + "epoch": 0.99, + "learning_rate": 5.4687690329713325e-08, + "loss": 0.544, + "step": 33246 + }, + { + "epoch": 0.99, + "learning_rate": 5.43693053629668e-08, + "loss": 0.5314, + "step": 33247 + }, + { + "epoch": 0.99, + "learning_rate": 5.405184965087129e-08, + "loss": 0.5058, + "step": 33248 + }, + { + "epoch": 0.99, + "learning_rate": 5.373532319636887e-08, + "loss": 0.5463, + "step": 33249 + }, + { + "epoch": 0.99, + "learning_rate": 5.3419726002390535e-08, + "loss": 0.5306, + "step": 33250 + }, + { + "epoch": 0.99, + "learning_rate": 5.310505807188948e-08, + "loss": 0.4707, + "step": 33251 + }, + { + "epoch": 0.99, + "learning_rate": 5.2791319407774484e-08, + "loss": 0.5144, + "step": 33252 + }, + { + "epoch": 0.99, + "learning_rate": 5.2478510012976544e-08, + "loss": 0.6078, + "step": 33253 + }, + { + "epoch": 0.99, + "learning_rate": 5.2166629890393335e-08, + "loss": 0.5389, + "step": 33254 + }, + { + "epoch": 0.99, + "learning_rate": 5.185567904293365e-08, + "loss": 0.5289, + "step": 33255 + }, + { + "epoch": 0.99, + "learning_rate": 5.1545657473484056e-08, + "loss": 0.4967, + "step": 33256 + }, + { + "epoch": 0.99, + "learning_rate": 5.123656518493114e-08, + "loss": 0.5431, + "step": 33257 + }, + { + "epoch": 0.99, + "learning_rate": 5.092840218013928e-08, + "loss": 0.5797, + "step": 33258 + }, + { + "epoch": 0.99, + "learning_rate": 5.0621168461983947e-08, + "loss": 0.4616, + "step": 33259 + }, + { + "epoch": 0.99, + "learning_rate": 5.031486403330732e-08, + "loss": 0.5546, + "step": 33260 + }, + { + "epoch": 0.99, + "learning_rate": 5.000948889698487e-08, + "loss": 0.4621, + "step": 33261 + }, + { + "epoch": 0.99, + "learning_rate": 4.9705043055825465e-08, + "loss": 0.5275, + "step": 33262 + }, + { + "epoch": 0.99, + "learning_rate": 4.940152651267127e-08, + "loss": 0.514, + "step": 33263 + }, + { + "epoch": 0.99, + "learning_rate": 4.909893927034226e-08, + "loss": 0.5586, + "step": 33264 + }, + { + "epoch": 0.99, + "learning_rate": 4.87972813316695e-08, + "loss": 0.5651, + "step": 33265 + }, + { + "epoch": 0.99, + "learning_rate": 4.8496552699428546e-08, + "loss": 0.5461, + "step": 33266 + }, + { + "epoch": 0.99, + "learning_rate": 4.8196753376439364e-08, + "loss": 0.5066, + "step": 33267 + }, + { + "epoch": 0.99, + "learning_rate": 4.7897883365477514e-08, + "loss": 0.4355, + "step": 33268 + }, + { + "epoch": 0.99, + "learning_rate": 4.759994266931855e-08, + "loss": 0.6016, + "step": 33269 + }, + { + "epoch": 0.99, + "learning_rate": 4.730293129074914e-08, + "loss": 0.4911, + "step": 33270 + }, + { + "epoch": 0.99, + "learning_rate": 4.700684923251153e-08, + "loss": 0.4481, + "step": 33271 + }, + { + "epoch": 0.99, + "learning_rate": 4.6711696497370174e-08, + "loss": 0.4948, + "step": 33272 + }, + { + "epoch": 0.99, + "learning_rate": 4.6417473088056215e-08, + "loss": 0.5692, + "step": 33273 + }, + { + "epoch": 0.99, + "learning_rate": 4.612417900732302e-08, + "loss": 0.4834, + "step": 33274 + }, + { + "epoch": 0.99, + "learning_rate": 4.5831814257890626e-08, + "loss": 0.4286, + "step": 33275 + }, + { + "epoch": 0.99, + "learning_rate": 4.5540378842479084e-08, + "loss": 0.6196, + "step": 33276 + }, + { + "epoch": 0.99, + "learning_rate": 4.524987276379733e-08, + "loss": 0.5647, + "step": 33277 + }, + { + "epoch": 0.99, + "learning_rate": 4.4960296024532114e-08, + "loss": 0.5408, + "step": 33278 + }, + { + "epoch": 0.99, + "learning_rate": 4.467164862739237e-08, + "loss": 0.4516, + "step": 33279 + }, + { + "epoch": 0.99, + "learning_rate": 4.438393057506485e-08, + "loss": 0.4675, + "step": 33280 + }, + { + "epoch": 0.99, + "learning_rate": 4.4097141870202974e-08, + "loss": 0.5737, + "step": 33281 + }, + { + "epoch": 0.99, + "learning_rate": 4.3811282515504594e-08, + "loss": 0.7221, + "step": 33282 + }, + { + "epoch": 0.99, + "learning_rate": 4.352635251360093e-08, + "loss": 0.5086, + "step": 33283 + }, + { + "epoch": 0.99, + "learning_rate": 4.324235186714543e-08, + "loss": 0.4802, + "step": 33284 + }, + { + "epoch": 0.99, + "learning_rate": 4.295928057879151e-08, + "loss": 0.4007, + "step": 33285 + }, + { + "epoch": 0.99, + "learning_rate": 4.26771386511593e-08, + "loss": 0.572, + "step": 33286 + }, + { + "epoch": 0.99, + "learning_rate": 4.239592608688003e-08, + "loss": 0.5422, + "step": 33287 + }, + { + "epoch": 0.99, + "learning_rate": 4.211564288856273e-08, + "loss": 0.5808, + "step": 33288 + }, + { + "epoch": 0.99, + "learning_rate": 4.183628905881643e-08, + "loss": 0.4614, + "step": 33289 + }, + { + "epoch": 0.99, + "learning_rate": 4.1557864600227925e-08, + "loss": 0.5114, + "step": 33290 + }, + { + "epoch": 0.99, + "learning_rate": 4.128036951540626e-08, + "loss": 0.6527, + "step": 33291 + }, + { + "epoch": 0.99, + "learning_rate": 4.100380380691604e-08, + "loss": 0.5078, + "step": 33292 + }, + { + "epoch": 0.99, + "learning_rate": 4.072816747733299e-08, + "loss": 0.5901, + "step": 33293 + }, + { + "epoch": 0.99, + "learning_rate": 4.045346052922172e-08, + "loss": 0.4333, + "step": 33294 + }, + { + "epoch": 0.99, + "learning_rate": 4.0179682965124644e-08, + "loss": 0.5038, + "step": 33295 + }, + { + "epoch": 0.99, + "learning_rate": 3.9906834787606374e-08, + "loss": 0.5243, + "step": 33296 + }, + { + "epoch": 0.99, + "learning_rate": 3.963491599919822e-08, + "loss": 0.5329, + "step": 33297 + }, + { + "epoch": 0.99, + "learning_rate": 3.9363926602409287e-08, + "loss": 0.5983, + "step": 33298 + }, + { + "epoch": 0.99, + "learning_rate": 3.909386659978198e-08, + "loss": 0.5934, + "step": 33299 + }, + { + "epoch": 0.99, + "learning_rate": 3.8824735993825415e-08, + "loss": 0.5236, + "step": 33300 + }, + { + "epoch": 0.99, + "learning_rate": 3.8556534787026476e-08, + "loss": 0.4382, + "step": 33301 + }, + { + "epoch": 0.99, + "learning_rate": 3.828926298189428e-08, + "loss": 0.5825, + "step": 33302 + }, + { + "epoch": 0.99, + "learning_rate": 3.802292058090462e-08, + "loss": 0.5293, + "step": 33303 + }, + { + "epoch": 0.99, + "learning_rate": 3.7757507586544393e-08, + "loss": 0.4916, + "step": 33304 + }, + { + "epoch": 0.99, + "learning_rate": 3.74930240012672e-08, + "loss": 0.4749, + "step": 33305 + }, + { + "epoch": 0.99, + "learning_rate": 3.722946982753772e-08, + "loss": 0.5546, + "step": 33306 + }, + { + "epoch": 0.99, + "learning_rate": 3.696684506780956e-08, + "loss": 0.5371, + "step": 33307 + }, + { + "epoch": 0.99, + "learning_rate": 3.670514972451411e-08, + "loss": 0.3389, + "step": 33308 + }, + { + "epoch": 0.99, + "learning_rate": 3.644438380010495e-08, + "loss": 0.4755, + "step": 33309 + }, + { + "epoch": 0.99, + "learning_rate": 3.618454729699128e-08, + "loss": 0.5964, + "step": 33310 + }, + { + "epoch": 0.99, + "learning_rate": 3.592564021758227e-08, + "loss": 0.3873, + "step": 33311 + }, + { + "epoch": 0.99, + "learning_rate": 3.5667662564309314e-08, + "loss": 0.5956, + "step": 33312 + }, + { + "epoch": 0.99, + "learning_rate": 3.5410614339548286e-08, + "loss": 0.4659, + "step": 33313 + }, + { + "epoch": 0.99, + "learning_rate": 3.515449554569728e-08, + "loss": 0.4214, + "step": 33314 + }, + { + "epoch": 0.99, + "learning_rate": 3.4899306185143256e-08, + "loss": 0.4824, + "step": 33315 + }, + { + "epoch": 0.99, + "learning_rate": 3.4645046260251004e-08, + "loss": 0.5757, + "step": 33316 + }, + { + "epoch": 0.99, + "learning_rate": 3.439171577338529e-08, + "loss": 0.5326, + "step": 33317 + }, + { + "epoch": 0.99, + "learning_rate": 3.413931472689979e-08, + "loss": 0.4654, + "step": 33318 + }, + { + "epoch": 0.99, + "learning_rate": 3.3887843123148186e-08, + "loss": 0.5477, + "step": 33319 + }, + { + "epoch": 0.99, + "learning_rate": 3.363730096446194e-08, + "loss": 0.5017, + "step": 33320 + }, + { + "epoch": 0.99, + "learning_rate": 3.3387688253183615e-08, + "loss": 0.4869, + "step": 33321 + }, + { + "epoch": 0.99, + "learning_rate": 3.313900499161138e-08, + "loss": 0.5619, + "step": 33322 + }, + { + "epoch": 0.99, + "learning_rate": 3.28912511820767e-08, + "loss": 0.5997, + "step": 33323 + }, + { + "epoch": 0.99, + "learning_rate": 3.2644426826866636e-08, + "loss": 0.4968, + "step": 33324 + }, + { + "epoch": 0.99, + "learning_rate": 3.2398531928301554e-08, + "loss": 0.5167, + "step": 33325 + }, + { + "epoch": 0.99, + "learning_rate": 3.2153566488635214e-08, + "loss": 0.6227, + "step": 33326 + }, + { + "epoch": 0.99, + "learning_rate": 3.1909530510176866e-08, + "loss": 0.6237, + "step": 33327 + }, + { + "epoch": 0.99, + "learning_rate": 3.166642399516917e-08, + "loss": 0.4722, + "step": 33328 + }, + { + "epoch": 0.99, + "learning_rate": 3.142424694588808e-08, + "loss": 0.5131, + "step": 33329 + }, + { + "epoch": 0.99, + "learning_rate": 3.118299936458735e-08, + "loss": 0.4265, + "step": 33330 + }, + { + "epoch": 0.99, + "learning_rate": 3.094268125349853e-08, + "loss": 0.5353, + "step": 33331 + }, + { + "epoch": 0.99, + "learning_rate": 3.0703292614853164e-08, + "loss": 0.5784, + "step": 33332 + }, + { + "epoch": 0.99, + "learning_rate": 3.046483345089391e-08, + "loss": 0.4631, + "step": 33333 + }, + { + "epoch": 0.99, + "learning_rate": 3.022730376381899e-08, + "loss": 0.4588, + "step": 33334 + }, + { + "epoch": 0.99, + "learning_rate": 2.999070355585998e-08, + "loss": 0.5531, + "step": 33335 + }, + { + "epoch": 0.99, + "learning_rate": 2.9755032829192896e-08, + "loss": 0.6129, + "step": 33336 + }, + { + "epoch": 0.99, + "learning_rate": 2.9520291586027092e-08, + "loss": 0.4217, + "step": 33337 + }, + { + "epoch": 0.99, + "learning_rate": 2.9286479828527503e-08, + "loss": 0.5172, + "step": 33338 + }, + { + "epoch": 0.99, + "learning_rate": 2.9053597558881263e-08, + "loss": 0.4096, + "step": 33339 + }, + { + "epoch": 0.99, + "learning_rate": 2.882164477925331e-08, + "loss": 0.5808, + "step": 33340 + }, + { + "epoch": 0.99, + "learning_rate": 2.8590621491797476e-08, + "loss": 0.4563, + "step": 33341 + }, + { + "epoch": 0.99, + "learning_rate": 2.836052769865649e-08, + "loss": 0.5933, + "step": 33342 + }, + { + "epoch": 0.99, + "learning_rate": 2.813136340198419e-08, + "loss": 0.4447, + "step": 33343 + }, + { + "epoch": 0.99, + "learning_rate": 2.7903128603889995e-08, + "loss": 0.5931, + "step": 33344 + }, + { + "epoch": 0.99, + "learning_rate": 2.7675823306516635e-08, + "loss": 0.5184, + "step": 33345 + }, + { + "epoch": 0.99, + "learning_rate": 2.7449447511962433e-08, + "loss": 0.5192, + "step": 33346 + }, + { + "epoch": 0.99, + "learning_rate": 2.7224001222347915e-08, + "loss": 0.4849, + "step": 33347 + }, + { + "epoch": 0.99, + "learning_rate": 2.69994844397492e-08, + "loss": 0.4437, + "step": 33348 + }, + { + "epoch": 0.99, + "learning_rate": 2.6775897166264608e-08, + "loss": 0.5249, + "step": 33349 + }, + { + "epoch": 0.99, + "learning_rate": 2.6553239403981356e-08, + "loss": 0.4348, + "step": 33350 + }, + { + "epoch": 0.99, + "learning_rate": 2.633151115495336e-08, + "loss": 0.4248, + "step": 33351 + }, + { + "epoch": 0.99, + "learning_rate": 2.611071242124563e-08, + "loss": 0.4654, + "step": 33352 + }, + { + "epoch": 0.99, + "learning_rate": 2.5890843204923187e-08, + "loss": 0.5865, + "step": 33353 + }, + { + "epoch": 0.99, + "learning_rate": 2.5671903508017736e-08, + "loss": 0.4814, + "step": 33354 + }, + { + "epoch": 0.99, + "learning_rate": 2.5453893332560984e-08, + "loss": 0.5478, + "step": 33355 + }, + { + "epoch": 0.99, + "learning_rate": 2.5236812680595744e-08, + "loss": 0.4612, + "step": 33356 + }, + { + "epoch": 0.99, + "learning_rate": 2.5020661554142623e-08, + "loss": 0.6806, + "step": 33357 + }, + { + "epoch": 0.99, + "learning_rate": 2.4805439955188914e-08, + "loss": 0.5127, + "step": 33358 + }, + { + "epoch": 0.99, + "learning_rate": 2.4591147885744125e-08, + "loss": 0.5131, + "step": 33359 + }, + { + "epoch": 0.99, + "learning_rate": 2.437778534781776e-08, + "loss": 0.5288, + "step": 33360 + }, + { + "epoch": 0.99, + "learning_rate": 2.4165352343374913e-08, + "loss": 0.473, + "step": 33361 + }, + { + "epoch": 0.99, + "learning_rate": 2.3953848874391782e-08, + "loss": 0.4017, + "step": 33362 + }, + { + "epoch": 0.99, + "learning_rate": 2.3743274942844564e-08, + "loss": 0.4972, + "step": 33363 + }, + { + "epoch": 0.99, + "learning_rate": 2.3533630550687246e-08, + "loss": 0.5515, + "step": 33364 + }, + { + "epoch": 0.99, + "learning_rate": 2.3324915699862724e-08, + "loss": 0.4783, + "step": 33365 + }, + { + "epoch": 0.99, + "learning_rate": 2.311713039231389e-08, + "loss": 0.4962, + "step": 33366 + }, + { + "epoch": 0.99, + "learning_rate": 2.291027462998363e-08, + "loss": 0.5587, + "step": 33367 + }, + { + "epoch": 0.99, + "learning_rate": 2.270434841478153e-08, + "loss": 0.5551, + "step": 33368 + }, + { + "epoch": 0.99, + "learning_rate": 2.2499351748628274e-08, + "loss": 0.4864, + "step": 33369 + }, + { + "epoch": 0.99, + "learning_rate": 2.2295284633422342e-08, + "loss": 0.4478, + "step": 33370 + }, + { + "epoch": 0.99, + "learning_rate": 2.209214707107332e-08, + "loss": 0.4785, + "step": 33371 + }, + { + "epoch": 0.99, + "learning_rate": 2.188993906346859e-08, + "loss": 0.5398, + "step": 33372 + }, + { + "epoch": 0.99, + "learning_rate": 2.1688660612484423e-08, + "loss": 0.525, + "step": 33373 + }, + { + "epoch": 0.99, + "learning_rate": 2.1488311719986e-08, + "loss": 0.6322, + "step": 33374 + }, + { + "epoch": 0.99, + "learning_rate": 2.1288892387838488e-08, + "loss": 0.4765, + "step": 33375 + }, + { + "epoch": 0.99, + "learning_rate": 2.109040261790707e-08, + "loss": 0.4741, + "step": 33376 + }, + { + "epoch": 0.99, + "learning_rate": 2.0892842412023606e-08, + "loss": 0.5164, + "step": 33377 + }, + { + "epoch": 0.99, + "learning_rate": 2.0696211772042172e-08, + "loss": 0.5551, + "step": 33378 + }, + { + "epoch": 0.99, + "learning_rate": 2.0500510699761333e-08, + "loss": 0.5947, + "step": 33379 + }, + { + "epoch": 0.99, + "learning_rate": 2.0305739197035156e-08, + "loss": 0.4736, + "step": 33380 + }, + { + "epoch": 0.99, + "learning_rate": 2.0111897265651104e-08, + "loss": 0.5676, + "step": 33381 + }, + { + "epoch": 0.99, + "learning_rate": 1.9918984907418835e-08, + "loss": 0.5865, + "step": 33382 + }, + { + "epoch": 0.99, + "learning_rate": 1.9727002124125815e-08, + "loss": 0.4804, + "step": 33383 + }, + { + "epoch": 0.99, + "learning_rate": 1.9535948917570602e-08, + "loss": 0.5984, + "step": 33384 + }, + { + "epoch": 0.99, + "learning_rate": 1.934582528951845e-08, + "loss": 0.5415, + "step": 33385 + }, + { + "epoch": 0.99, + "learning_rate": 1.9156631241745713e-08, + "loss": 0.5761, + "step": 33386 + }, + { + "epoch": 0.99, + "learning_rate": 1.8968366776006554e-08, + "loss": 0.5697, + "step": 33387 + }, + { + "epoch": 0.99, + "learning_rate": 1.878103189404401e-08, + "loss": 0.5152, + "step": 33388 + }, + { + "epoch": 0.99, + "learning_rate": 1.8594626597612243e-08, + "loss": 0.5213, + "step": 33389 + }, + { + "epoch": 0.99, + "learning_rate": 1.840915088843209e-08, + "loss": 0.5162, + "step": 33390 + }, + { + "epoch": 0.99, + "learning_rate": 1.822460476824661e-08, + "loss": 0.4652, + "step": 33391 + }, + { + "epoch": 0.99, + "learning_rate": 1.8040988238754443e-08, + "loss": 0.5597, + "step": 33392 + }, + { + "epoch": 0.99, + "learning_rate": 1.7858301301665325e-08, + "loss": 0.6244, + "step": 33393 + }, + { + "epoch": 0.99, + "learning_rate": 1.7676543958677905e-08, + "loss": 0.5792, + "step": 33394 + }, + { + "epoch": 0.99, + "learning_rate": 1.749571621149082e-08, + "loss": 0.5742, + "step": 33395 + }, + { + "epoch": 0.99, + "learning_rate": 1.7315818061780508e-08, + "loss": 0.5646, + "step": 33396 + }, + { + "epoch": 0.99, + "learning_rate": 1.7136849511212304e-08, + "loss": 0.5906, + "step": 33397 + }, + { + "epoch": 0.99, + "learning_rate": 1.6958810561462647e-08, + "loss": 0.5322, + "step": 33398 + }, + { + "epoch": 0.99, + "learning_rate": 1.678170121417466e-08, + "loss": 0.5773, + "step": 33399 + }, + { + "epoch": 0.99, + "learning_rate": 1.6605521471002583e-08, + "loss": 0.5554, + "step": 33400 + }, + { + "epoch": 0.99, + "learning_rate": 1.6430271333589543e-08, + "loss": 0.3949, + "step": 33401 + }, + { + "epoch": 0.99, + "learning_rate": 1.6255950803556464e-08, + "loss": 0.5863, + "step": 33402 + }, + { + "epoch": 0.99, + "learning_rate": 1.6082559882524274e-08, + "loss": 0.4585, + "step": 33403 + }, + { + "epoch": 0.99, + "learning_rate": 1.5910098572102794e-08, + "loss": 0.4837, + "step": 33404 + }, + { + "epoch": 0.99, + "learning_rate": 1.573856687390185e-08, + "loss": 0.5189, + "step": 33405 + }, + { + "epoch": 0.99, + "learning_rate": 1.556796478950906e-08, + "loss": 0.4331, + "step": 33406 + }, + { + "epoch": 0.99, + "learning_rate": 1.5398292320523144e-08, + "loss": 0.4839, + "step": 33407 + }, + { + "epoch": 0.99, + "learning_rate": 1.5229549468509517e-08, + "loss": 0.6509, + "step": 33408 + }, + { + "epoch": 0.99, + "learning_rate": 1.50617362350447e-08, + "loss": 0.5544, + "step": 33409 + }, + { + "epoch": 0.99, + "learning_rate": 1.4894852621682998e-08, + "loss": 0.4352, + "step": 33410 + }, + { + "epoch": 0.99, + "learning_rate": 1.4728898629989828e-08, + "loss": 0.4131, + "step": 33411 + }, + { + "epoch": 0.99, + "learning_rate": 1.45638742614862e-08, + "loss": 0.669, + "step": 33412 + }, + { + "epoch": 0.99, + "learning_rate": 1.4399779517715318e-08, + "loss": 0.6358, + "step": 33413 + }, + { + "epoch": 0.99, + "learning_rate": 1.4236614400209291e-08, + "loss": 0.4591, + "step": 33414 + }, + { + "epoch": 0.99, + "learning_rate": 1.4074378910478025e-08, + "loss": 0.4987, + "step": 33415 + }, + { + "epoch": 0.99, + "learning_rate": 1.3913073050042525e-08, + "loss": 0.5097, + "step": 33416 + }, + { + "epoch": 0.99, + "learning_rate": 1.3752696820379385e-08, + "loss": 0.4543, + "step": 33417 + }, + { + "epoch": 0.99, + "learning_rate": 1.359325022299851e-08, + "loss": 0.6206, + "step": 33418 + }, + { + "epoch": 0.99, + "learning_rate": 1.3434733259376497e-08, + "loss": 0.4982, + "step": 33419 + }, + { + "epoch": 0.99, + "learning_rate": 1.3277145930989943e-08, + "loss": 0.4292, + "step": 33420 + }, + { + "epoch": 0.99, + "learning_rate": 1.312048823930434e-08, + "loss": 0.5024, + "step": 33421 + }, + { + "epoch": 1.0, + "learning_rate": 1.2964760185762981e-08, + "loss": 0.4925, + "step": 33422 + }, + { + "epoch": 1.0, + "learning_rate": 1.2809961771831358e-08, + "loss": 0.5774, + "step": 33423 + }, + { + "epoch": 1.0, + "learning_rate": 1.265609299894166e-08, + "loss": 0.4329, + "step": 33424 + }, + { + "epoch": 1.0, + "learning_rate": 1.2503153868526074e-08, + "loss": 0.5085, + "step": 33425 + }, + { + "epoch": 1.0, + "learning_rate": 1.2351144382005686e-08, + "loss": 0.5946, + "step": 33426 + }, + { + "epoch": 1.0, + "learning_rate": 1.2200064540790479e-08, + "loss": 0.4672, + "step": 33427 + }, + { + "epoch": 1.0, + "learning_rate": 1.2049914346290436e-08, + "loss": 0.4337, + "step": 33428 + }, + { + "epoch": 1.0, + "learning_rate": 1.1900693799893337e-08, + "loss": 0.5526, + "step": 33429 + }, + { + "epoch": 1.0, + "learning_rate": 1.1752402902998061e-08, + "loss": 0.5997, + "step": 33430 + }, + { + "epoch": 1.0, + "learning_rate": 1.1605041656970184e-08, + "loss": 0.5424, + "step": 33431 + }, + { + "epoch": 1.0, + "learning_rate": 1.1458610063197484e-08, + "loss": 0.5384, + "step": 33432 + }, + { + "epoch": 1.0, + "learning_rate": 1.131310812302333e-08, + "loss": 0.5227, + "step": 33433 + }, + { + "epoch": 1.0, + "learning_rate": 1.11685358378133e-08, + "loss": 0.5157, + "step": 33434 + }, + { + "epoch": 1.0, + "learning_rate": 1.1024893208899656e-08, + "loss": 0.4933, + "step": 33435 + }, + { + "epoch": 1.0, + "learning_rate": 1.0882180237636874e-08, + "loss": 0.5638, + "step": 33436 + }, + { + "epoch": 1.0, + "learning_rate": 1.0740396925323914e-08, + "loss": 0.554, + "step": 33437 + }, + { + "epoch": 1.0, + "learning_rate": 1.0599543273304146e-08, + "loss": 0.5496, + "step": 33438 + }, + { + "epoch": 1.0, + "learning_rate": 1.0459619282876531e-08, + "loss": 0.4018, + "step": 33439 + }, + { + "epoch": 1.0, + "learning_rate": 1.0320624955351133e-08, + "loss": 0.5375, + "step": 33440 + }, + { + "epoch": 1.0, + "learning_rate": 1.0182560292004705e-08, + "loss": 0.5904, + "step": 33441 + }, + { + "epoch": 1.0, + "learning_rate": 1.004542529412511e-08, + "loss": 0.4046, + "step": 33442 + }, + { + "epoch": 1.0, + "learning_rate": 9.909219963000205e-09, + "loss": 0.5869, + "step": 33443 + }, + { + "epoch": 1.0, + "learning_rate": 9.773944299884541e-09, + "loss": 0.6212, + "step": 33444 + }, + { + "epoch": 1.0, + "learning_rate": 9.639598306032671e-09, + "loss": 0.4643, + "step": 33445 + }, + { + "epoch": 1.0, + "learning_rate": 9.506181982710249e-09, + "loss": 0.5399, + "step": 33446 + }, + { + "epoch": 1.0, + "learning_rate": 9.373695331138522e-09, + "loss": 0.4909, + "step": 33447 + }, + { + "epoch": 1.0, + "learning_rate": 9.242138352560936e-09, + "loss": 0.5961, + "step": 33448 + }, + { + "epoch": 1.0, + "learning_rate": 9.11151104819874e-09, + "loss": 0.4331, + "step": 33449 + }, + { + "epoch": 1.0, + "learning_rate": 8.981813419262076e-09, + "loss": 0.6015, + "step": 33450 + }, + { + "epoch": 1.0, + "learning_rate": 8.853045466961086e-09, + "loss": 0.5355, + "step": 33451 + }, + { + "epoch": 1.0, + "learning_rate": 8.72520719248371e-09, + "loss": 0.471, + "step": 33452 + }, + { + "epoch": 1.0, + "learning_rate": 8.59829859704009e-09, + "loss": 0.4918, + "step": 33453 + }, + { + "epoch": 1.0, + "learning_rate": 8.472319681795959e-09, + "loss": 0.6203, + "step": 33454 + }, + { + "epoch": 1.0, + "learning_rate": 8.347270447917055e-09, + "loss": 0.5348, + "step": 33455 + }, + { + "epoch": 1.0, + "learning_rate": 8.22315089656911e-09, + "loss": 0.5615, + "step": 33456 + }, + { + "epoch": 1.0, + "learning_rate": 8.099961028917858e-09, + "loss": 0.6785, + "step": 33457 + }, + { + "epoch": 1.0, + "learning_rate": 7.977700846095726e-09, + "loss": 0.5692, + "step": 33458 + }, + { + "epoch": 1.0, + "learning_rate": 7.856370349246244e-09, + "loss": 0.4727, + "step": 33459 + }, + { + "epoch": 1.0, + "learning_rate": 7.735969539501842e-09, + "loss": 0.6077, + "step": 33460 + }, + { + "epoch": 1.0, + "learning_rate": 7.616498417961637e-09, + "loss": 0.5307, + "step": 33461 + }, + { + "epoch": 1.0, + "learning_rate": 7.497956985758059e-09, + "loss": 0.6093, + "step": 33462 + }, + { + "epoch": 1.0, + "learning_rate": 7.380345243990227e-09, + "loss": 0.5019, + "step": 33463 + }, + { + "epoch": 1.0, + "learning_rate": 7.26366319373506e-09, + "loss": 0.5681, + "step": 33464 + }, + { + "epoch": 1.0, + "learning_rate": 7.1479108360916754e-09, + "loss": 0.5697, + "step": 33465 + }, + { + "epoch": 1.0, + "learning_rate": 7.033088172136993e-09, + "loss": 0.6111, + "step": 33466 + }, + { + "epoch": 1.0, + "learning_rate": 6.919195202936824e-09, + "loss": 0.6251, + "step": 33467 + }, + { + "epoch": 1.0, + "learning_rate": 6.806231929556983e-09, + "loss": 0.5101, + "step": 33468 + }, + { + "epoch": 1.0, + "learning_rate": 6.694198353029979e-09, + "loss": 0.586, + "step": 33469 + }, + { + "epoch": 1.0, + "learning_rate": 6.583094474399421e-09, + "loss": 0.5807, + "step": 33470 + }, + { + "epoch": 1.0, + "learning_rate": 6.47292029472002e-09, + "loss": 0.457, + "step": 33471 + }, + { + "epoch": 1.0, + "learning_rate": 6.363675814990977e-09, + "loss": 0.5811, + "step": 33472 + }, + { + "epoch": 1.0, + "learning_rate": 6.2553610362448e-09, + "loss": 0.5339, + "step": 33473 + }, + { + "epoch": 1.0, + "learning_rate": 6.14797595948069e-09, + "loss": 0.4413, + "step": 33474 + }, + { + "epoch": 1.0, + "learning_rate": 6.041520585697846e-09, + "loss": 0.4277, + "step": 33475 + }, + { + "epoch": 1.0, + "learning_rate": 5.9359949158843685e-09, + "loss": 0.5633, + "step": 33476 + }, + { + "epoch": 1.0, + "learning_rate": 5.8313989510394575e-09, + "loss": 0.4538, + "step": 33477 + }, + { + "epoch": 1.0, + "learning_rate": 5.727732692106802e-09, + "loss": 0.6144, + "step": 33478 + }, + { + "epoch": 1.0, + "learning_rate": 5.624996140063399e-09, + "loss": 0.5428, + "step": 33479 + }, + { + "epoch": 1.0, + "learning_rate": 5.523189295875142e-09, + "loss": 0.5863, + "step": 33480 + }, + { + "epoch": 1.0, + "learning_rate": 5.422312160463516e-09, + "loss": 0.5475, + "step": 33481 + }, + { + "epoch": 1.0, + "learning_rate": 5.322364734794416e-09, + "loss": 0.5185, + "step": 33482 + }, + { + "epoch": 1.0, + "learning_rate": 5.223347019778224e-09, + "loss": 0.5427, + "step": 33483 + }, + { + "epoch": 1.0, + "learning_rate": 5.125259016336425e-09, + "loss": 0.6105, + "step": 33484 + }, + { + "epoch": 1.0, + "learning_rate": 5.028100725401608e-09, + "loss": 0.4927, + "step": 33485 + }, + { + "epoch": 1.0, + "learning_rate": 4.931872147850846e-09, + "loss": 0.5357, + "step": 33486 + }, + { + "epoch": 1.0, + "learning_rate": 4.836573284583423e-09, + "loss": 0.5237, + "step": 33487 + }, + { + "epoch": 1.0, + "learning_rate": 4.742204136498618e-09, + "loss": 0.6305, + "step": 33488 + }, + { + "epoch": 1.0, + "learning_rate": 4.648764704473507e-09, + "loss": 0.5791, + "step": 33489 + }, + { + "epoch": 1.0, + "learning_rate": 4.5562549893629625e-09, + "loss": 0.4695, + "step": 33490 + }, + { + "epoch": 1.0, + "learning_rate": 4.464674992032957e-09, + "loss": 0.5659, + "step": 33491 + }, + { + "epoch": 1.0, + "learning_rate": 4.374024713338365e-09, + "loss": 0.5259, + "step": 33492 + }, + { + "epoch": 1.0, + "learning_rate": 4.284304154111851e-09, + "loss": 0.465, + "step": 33493 + }, + { + "epoch": 1.0, + "learning_rate": 4.195513315208288e-09, + "loss": 0.54, + "step": 33494 + }, + { + "epoch": 1.0, + "learning_rate": 4.107652197427037e-09, + "loss": 0.4997, + "step": 33495 + }, + { + "epoch": 1.0, + "learning_rate": 4.020720801611866e-09, + "loss": 0.4547, + "step": 33496 + }, + { + "epoch": 1.0, + "learning_rate": 3.934719128551035e-09, + "loss": 0.5, + "step": 33497 + }, + { + "epoch": 1.0, + "learning_rate": 3.8496471790439026e-09, + "loss": 0.4532, + "step": 33498 + }, + { + "epoch": 1.0, + "learning_rate": 3.765504953900933e-09, + "loss": 0.5121, + "step": 33499 + }, + { + "epoch": 1.0, + "learning_rate": 3.6822924538881807e-09, + "loss": 0.4898, + "step": 33500 + }, + { + "epoch": 1.0, + "learning_rate": 3.600009679782801e-09, + "loss": 0.4625, + "step": 33501 + }, + { + "epoch": 1.0, + "learning_rate": 3.5186566323397453e-09, + "loss": 0.4817, + "step": 33502 + }, + { + "epoch": 1.0, + "learning_rate": 3.43823331233617e-09, + "loss": 0.4378, + "step": 33503 + }, + { + "epoch": 1.0, + "learning_rate": 3.358739720504822e-09, + "loss": 0.5746, + "step": 33504 + }, + { + "epoch": 1.0, + "learning_rate": 3.2801758575895513e-09, + "loss": 0.4916, + "step": 33505 + }, + { + "epoch": 1.0, + "learning_rate": 3.202541724323105e-09, + "loss": 0.6304, + "step": 33506 + }, + { + "epoch": 1.0, + "learning_rate": 3.1258373214271274e-09, + "loss": 0.5406, + "step": 33507 + }, + { + "epoch": 1.0, + "learning_rate": 3.050062649612162e-09, + "loss": 0.5612, + "step": 33508 + }, + { + "epoch": 1.0, + "learning_rate": 2.975217709577649e-09, + "loss": 0.6292, + "step": 33509 + }, + { + "epoch": 1.0, + "learning_rate": 2.901302502023029e-09, + "loss": 0.5845, + "step": 33510 + }, + { + "epoch": 1.0, + "learning_rate": 2.828317027647742e-09, + "loss": 0.3898, + "step": 33511 + }, + { + "epoch": 1.0, + "learning_rate": 2.7562612871068203e-09, + "loss": 0.5244, + "step": 33512 + }, + { + "epoch": 1.0, + "learning_rate": 2.685135281088602e-09, + "loss": 0.5503, + "step": 33513 + }, + { + "epoch": 1.0, + "learning_rate": 2.6149390102481186e-09, + "loss": 0.5854, + "step": 33514 + }, + { + "epoch": 1.0, + "learning_rate": 2.5456724752404014e-09, + "loss": 0.5338, + "step": 33515 + }, + { + "epoch": 1.0, + "learning_rate": 2.4773356767093802e-09, + "loss": 0.4973, + "step": 33516 + }, + { + "epoch": 1.0, + "learning_rate": 2.409928615287882e-09, + "loss": 0.6243, + "step": 33517 + }, + { + "epoch": 1.0, + "learning_rate": 2.3434512916087335e-09, + "loss": 0.6627, + "step": 33518 + }, + { + "epoch": 1.0, + "learning_rate": 2.277903706271456e-09, + "loss": 0.4886, + "step": 33519 + }, + { + "epoch": 1.0, + "learning_rate": 2.213285859908876e-09, + "loss": 0.477, + "step": 33520 + }, + { + "epoch": 1.0, + "learning_rate": 2.1495977531094114e-09, + "loss": 0.4121, + "step": 33521 + }, + { + "epoch": 1.0, + "learning_rate": 2.086839386472583e-09, + "loss": 0.3822, + "step": 33522 + }, + { + "epoch": 1.0, + "learning_rate": 2.0250107605757073e-09, + "loss": 0.4939, + "step": 33523 + }, + { + "epoch": 1.0, + "learning_rate": 1.9641118759849975e-09, + "loss": 0.4658, + "step": 33524 + }, + { + "epoch": 1.0, + "learning_rate": 1.9041427332888717e-09, + "loss": 0.4957, + "step": 33525 + }, + { + "epoch": 1.0, + "learning_rate": 1.8451033330313393e-09, + "loss": 0.475, + "step": 33526 + }, + { + "epoch": 1.0, + "learning_rate": 1.78699367575641e-09, + "loss": 0.5294, + "step": 33527 + }, + { + "epoch": 1.0, + "learning_rate": 1.7298137620080922e-09, + "loss": 0.4005, + "step": 33528 + }, + { + "epoch": 1.0, + "learning_rate": 1.6735635923303961e-09, + "loss": 0.5067, + "step": 33529 + }, + { + "epoch": 1.0, + "learning_rate": 1.6182431672340237e-09, + "loss": 0.5434, + "step": 33530 + }, + { + "epoch": 1.0, + "learning_rate": 1.5638524872296779e-09, + "loss": 0.509, + "step": 33531 + }, + { + "epoch": 1.0, + "learning_rate": 1.5103915528391633e-09, + "loss": 0.455, + "step": 33532 + }, + { + "epoch": 1.0, + "learning_rate": 1.4578603645398759e-09, + "loss": 0.3975, + "step": 33533 + }, + { + "epoch": 1.0, + "learning_rate": 1.406258922831416e-09, + "loss": 0.445, + "step": 33534 + }, + { + "epoch": 1.0, + "learning_rate": 1.355587228202282e-09, + "loss": 0.4968, + "step": 33535 + }, + { + "epoch": 1.0, + "learning_rate": 1.305845281107665e-09, + "loss": 0.5305, + "step": 33536 + }, + { + "epoch": 1.0, + "learning_rate": 1.2570330820138587e-09, + "loss": 0.6009, + "step": 33537 + }, + { + "epoch": 1.0, + "learning_rate": 1.2091506313760547e-09, + "loss": 0.5711, + "step": 33538 + }, + { + "epoch": 1.0, + "learning_rate": 1.1621979296383422e-09, + "loss": 0.6095, + "step": 33539 + }, + { + "epoch": 1.0, + "learning_rate": 1.1161749772448104e-09, + "loss": 0.5431, + "step": 33540 + }, + { + "epoch": 1.0, + "learning_rate": 1.071081774606242e-09, + "loss": 0.5656, + "step": 33541 + }, + { + "epoch": 1.0, + "learning_rate": 1.0269183221667255e-09, + "loss": 0.3756, + "step": 33542 + }, + { + "epoch": 1.0, + "learning_rate": 9.836846203148398e-10, + "loss": 0.5166, + "step": 33543 + }, + { + "epoch": 1.0, + "learning_rate": 9.413806694613669e-10, + "loss": 0.3969, + "step": 33544 + }, + { + "epoch": 1.0, + "learning_rate": 9.000064700059874e-10, + "loss": 0.5271, + "step": 33545 + }, + { + "epoch": 1.0, + "learning_rate": 8.595620223150746e-10, + "loss": 0.5735, + "step": 33546 + }, + { + "epoch": 1.0, + "learning_rate": 8.200473267883091e-10, + "loss": 0.4584, + "step": 33547 + }, + { + "epoch": 1.0, + "learning_rate": 7.814623837698598e-10, + "loss": 0.5068, + "step": 33548 + }, + { + "epoch": 1.0, + "learning_rate": 7.438071936372026e-10, + "loss": 0.5122, + "step": 33549 + }, + { + "epoch": 1.0, + "learning_rate": 7.070817567345067e-10, + "loss": 0.5478, + "step": 33550 + }, + { + "epoch": 1.0, + "learning_rate": 6.712860733948389e-10, + "loss": 0.5507, + "step": 33551 + }, + { + "epoch": 1.0, + "learning_rate": 6.364201439623685e-10, + "loss": 0.5724, + "step": 33552 + }, + { + "epoch": 1.0, + "learning_rate": 6.0248396875906e-10, + "loss": 0.522, + "step": 33553 + }, + { + "epoch": 1.0, + "learning_rate": 5.694775480957759e-10, + "loss": 0.4876, + "step": 33554 + }, + { + "epoch": 1.0, + "learning_rate": 5.374008822722764e-10, + "loss": 0.5655, + "step": 33555 + }, + { + "epoch": 1.0, + "learning_rate": 5.062539716105263e-10, + "loss": 0.4188, + "step": 33556 + }, + { + "epoch": 1.0, + "learning_rate": 4.76036816376979e-10, + "loss": 0.4022, + "step": 33557 + }, + { + "epoch": 1.0, + "learning_rate": 4.467494168602926e-10, + "loss": 0.4582, + "step": 33558 + }, + { + "epoch": 1.0, + "learning_rate": 4.183917733380227e-10, + "loss": 0.4958, + "step": 33559 + }, + { + "epoch": 1.0, + "learning_rate": 3.909638860766229e-10, + "loss": 0.5079, + "step": 33560 + }, + { + "epoch": 1.0, + "learning_rate": 3.644657553092401e-10, + "loss": 0.5582, + "step": 33561 + }, + { + "epoch": 1.0, + "learning_rate": 3.3889738130232774e-10, + "loss": 0.4594, + "step": 33562 + }, + { + "epoch": 1.0, + "learning_rate": 3.142587643001349e-10, + "loss": 0.4234, + "step": 33563 + }, + { + "epoch": 1.0, + "learning_rate": 2.9054990450250177e-10, + "loss": 0.4638, + "step": 33564 + }, + { + "epoch": 1.0, + "learning_rate": 2.677708021536773e-10, + "loss": 0.4779, + "step": 33565 + }, + { + "epoch": 1.0, + "learning_rate": 2.4592145746460406e-10, + "loss": 0.5828, + "step": 33566 + }, + { + "epoch": 1.0, + "learning_rate": 2.2500187063512202e-10, + "loss": 0.5369, + "step": 33567 + }, + { + "epoch": 1.0, + "learning_rate": 2.0501204184286694e-10, + "loss": 0.5541, + "step": 33568 + }, + { + "epoch": 1.0, + "learning_rate": 1.8595197129878117e-10, + "loss": 0.4888, + "step": 33569 + }, + { + "epoch": 1.0, + "learning_rate": 1.678216591693982e-10, + "loss": 0.5564, + "step": 33570 + }, + { + "epoch": 1.0, + "learning_rate": 1.5062110562125142e-10, + "loss": 0.5088, + "step": 33571 + }, + { + "epoch": 1.0, + "learning_rate": 1.3435031082087436e-10, + "loss": 0.5087, + "step": 33572 + }, + { + "epoch": 1.0, + "learning_rate": 1.1900927491259595e-10, + "loss": 0.5243, + "step": 33573 + }, + { + "epoch": 1.0, + "learning_rate": 1.0459799804074521e-10, + "loss": 0.5291, + "step": 33574 + }, + { + "epoch": 1.0, + "learning_rate": 9.111648033854891e-11, + "loss": 0.4467, + "step": 33575 + }, + { + "epoch": 1.0, + "learning_rate": 7.856472193923381e-11, + "loss": 0.4397, + "step": 33576 + }, + { + "epoch": 1.0, + "learning_rate": 6.694272294271997e-11, + "loss": 0.4963, + "step": 33577 + }, + { + "epoch": 1.0, + "learning_rate": 5.625048348223416e-11, + "loss": 0.5123, + "step": 33578 + }, + { + "epoch": 1.0, + "learning_rate": 4.6488003635492e-11, + "loss": 0.5301, + "step": 33579 + }, + { + "epoch": 1.0, + "learning_rate": 3.765528349131131e-11, + "loss": 0.5026, + "step": 33580 + }, + { + "epoch": 1.0, + "learning_rate": 2.975232314961218e-11, + "loss": 0.5166, + "step": 33581 + }, + { + "epoch": 1.0, + "learning_rate": 2.277912267700799e-11, + "loss": 0.4988, + "step": 33582 + }, + { + "epoch": 1.0, + "learning_rate": 1.6735682140112118e-11, + "loss": 0.6224, + "step": 33583 + }, + { + "epoch": 1.0, + "learning_rate": 1.1622001583333486e-11, + "loss": 0.5363, + "step": 33584 + }, + { + "epoch": 1.0, + "learning_rate": 7.438081062183244e-12, + "loss": 0.5358, + "step": 33585 + }, + { + "epoch": 1.0, + "learning_rate": 4.183920621070314e-12, + "loss": 0.3719, + "step": 33586 + }, + { + "epoch": 1.0, + "learning_rate": 1.8595202821991563e-12, + "loss": 0.5685, + "step": 33587 + }, + { + "epoch": 1.0, + "learning_rate": 4.648800677742316e-13, + "loss": 0.6227, + "step": 33588 + }, + { + "epoch": 1.0, + "learning_rate": 0.0, + "loss": 0.5332, + "step": 33589 + }, + { + "epoch": 1.0, + "step": 33589, + "total_flos": 4291503404843008.0, + "train_loss": 0.5519348401742088, + "train_runtime": 110010.7773, + "train_samples_per_second": 19.541, + "train_steps_per_second": 0.305 + } + ], + "logging_steps": 1.0, + "max_steps": 33589, + "num_input_tokens_seen": 0, + "num_train_epochs": 1, + "save_steps": 5000, + "total_flos": 4291503404843008.0, + "train_batch_size": 8, + "trial_name": null, + "trial_params": null +}