diff --git "a/checkpoint-3150/trainer_state.json" "b/checkpoint-3150/trainer_state.json" new file mode 100644--- /dev/null +++ "b/checkpoint-3150/trainer_state.json" @@ -0,0 +1,18986 @@ +{ + "best_metric": 0.0019458403112366796, + "best_model_checkpoint": "/tmp/model/checkpoint-3150", + "epoch": 5.0, + "eval_steps": 500, + "global_step": 3150, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 1.5873015873015874e-07, + "loss": 0.7663, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 3.174603174603175e-07, + "loss": 0.7545, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 4.761904761904763e-07, + "loss": 0.7735, + "step": 3 + }, + { + "epoch": 0.01, + "learning_rate": 6.34920634920635e-07, + "loss": 0.7847, + "step": 4 + }, + { + "epoch": 0.01, + "learning_rate": 7.936507936507937e-07, + "loss": 0.7334, + "step": 5 + }, + { + "epoch": 0.01, + "learning_rate": 9.523809523809526e-07, + "loss": 0.7527, + "step": 6 + }, + { + "epoch": 0.01, + "learning_rate": 1.1111111111111112e-06, + "loss": 0.7424, + "step": 7 + }, + { + "epoch": 0.01, + "learning_rate": 1.26984126984127e-06, + "loss": 0.7459, + "step": 8 + }, + { + "epoch": 0.01, + "learning_rate": 1.4285714285714286e-06, + "loss": 0.7326, + "step": 9 + }, + { + "epoch": 0.02, + "learning_rate": 1.5873015873015873e-06, + "loss": 0.6567, + "step": 10 + }, + { + "epoch": 0.02, + "learning_rate": 1.7460317460317462e-06, + "loss": 0.6426, + "step": 11 + }, + { + "epoch": 0.02, + "learning_rate": 1.9047619047619051e-06, + "loss": 0.6626, + "step": 12 + }, + { + "epoch": 0.02, + "learning_rate": 2.0634920634920634e-06, + "loss": 0.6233, + "step": 13 + }, + { + "epoch": 0.02, + "learning_rate": 2.2222222222222225e-06, + "loss": 0.5864, + "step": 14 + }, + { + "epoch": 0.02, + "learning_rate": 2.3809523809523808e-06, + "loss": 0.5446, + "step": 15 + }, + { + "epoch": 0.03, + "learning_rate": 2.53968253968254e-06, + "loss": 0.5413, + "step": 16 + }, + { + "epoch": 0.03, + "learning_rate": 2.6984126984126986e-06, + "loss": 0.4948, + "step": 17 + }, + { + "epoch": 0.03, + "learning_rate": 2.8571428571428573e-06, + "loss": 0.4958, + "step": 18 + }, + { + "epoch": 0.03, + "learning_rate": 3.015873015873016e-06, + "loss": 0.4295, + "step": 19 + }, + { + "epoch": 0.03, + "learning_rate": 3.1746031746031746e-06, + "loss": 0.4114, + "step": 20 + }, + { + "epoch": 0.03, + "learning_rate": 3.3333333333333333e-06, + "loss": 0.388, + "step": 21 + }, + { + "epoch": 0.03, + "learning_rate": 3.4920634920634924e-06, + "loss": 0.3397, + "step": 22 + }, + { + "epoch": 0.04, + "learning_rate": 3.650793650793651e-06, + "loss": 0.3149, + "step": 23 + }, + { + "epoch": 0.04, + "learning_rate": 3.8095238095238102e-06, + "loss": 0.3032, + "step": 24 + }, + { + "epoch": 0.04, + "learning_rate": 3.968253968253968e-06, + "loss": 0.2983, + "step": 25 + }, + { + "epoch": 0.04, + "learning_rate": 4.126984126984127e-06, + "loss": 0.2551, + "step": 26 + }, + { + "epoch": 0.04, + "learning_rate": 4.285714285714286e-06, + "loss": 0.2158, + "step": 27 + }, + { + "epoch": 0.04, + "learning_rate": 4.444444444444445e-06, + "loss": 0.193, + "step": 28 + }, + { + "epoch": 0.05, + "learning_rate": 4.603174603174604e-06, + "loss": 0.1685, + "step": 29 + }, + { + "epoch": 0.05, + "learning_rate": 4.7619047619047615e-06, + "loss": 0.1529, + "step": 30 + }, + { + "epoch": 0.05, + "learning_rate": 4.920634920634921e-06, + "loss": 0.1322, + "step": 31 + }, + { + "epoch": 0.05, + "learning_rate": 5.07936507936508e-06, + "loss": 0.1115, + "step": 32 + }, + { + "epoch": 0.05, + "learning_rate": 5.2380952380952384e-06, + "loss": 0.0971, + "step": 33 + }, + { + "epoch": 0.05, + "learning_rate": 5.396825396825397e-06, + "loss": 0.0826, + "step": 34 + }, + { + "epoch": 0.06, + "learning_rate": 5.555555555555556e-06, + "loss": 0.0743, + "step": 35 + }, + { + "epoch": 0.06, + "learning_rate": 5.7142857142857145e-06, + "loss": 0.0635, + "step": 36 + }, + { + "epoch": 0.06, + "learning_rate": 5.873015873015873e-06, + "loss": 0.0481, + "step": 37 + }, + { + "epoch": 0.06, + "learning_rate": 6.031746031746032e-06, + "loss": 0.1245, + "step": 38 + }, + { + "epoch": 0.06, + "learning_rate": 6.190476190476191e-06, + "loss": 0.0351, + "step": 39 + }, + { + "epoch": 0.06, + "learning_rate": 6.349206349206349e-06, + "loss": 0.0296, + "step": 40 + }, + { + "epoch": 0.07, + "learning_rate": 6.507936507936509e-06, + "loss": 0.025, + "step": 41 + }, + { + "epoch": 0.07, + "learning_rate": 6.666666666666667e-06, + "loss": 0.1199, + "step": 42 + }, + { + "epoch": 0.07, + "learning_rate": 6.825396825396825e-06, + "loss": 0.0169, + "step": 43 + }, + { + "epoch": 0.07, + "learning_rate": 6.984126984126985e-06, + "loss": 0.0157, + "step": 44 + }, + { + "epoch": 0.07, + "learning_rate": 7.142857142857143e-06, + "loss": 0.1047, + "step": 45 + }, + { + "epoch": 0.07, + "learning_rate": 7.301587301587302e-06, + "loss": 0.1361, + "step": 46 + }, + { + "epoch": 0.07, + "learning_rate": 7.460317460317461e-06, + "loss": 0.1214, + "step": 47 + }, + { + "epoch": 0.08, + "learning_rate": 7.6190476190476205e-06, + "loss": 0.0103, + "step": 48 + }, + { + "epoch": 0.08, + "learning_rate": 7.777777777777777e-06, + "loss": 0.0107, + "step": 49 + }, + { + "epoch": 0.08, + "learning_rate": 7.936507936507936e-06, + "loss": 0.0099, + "step": 50 + }, + { + "epoch": 0.08, + "learning_rate": 8.095238095238097e-06, + "loss": 0.009, + "step": 51 + }, + { + "epoch": 0.08, + "learning_rate": 8.253968253968254e-06, + "loss": 0.0087, + "step": 52 + }, + { + "epoch": 0.08, + "learning_rate": 8.412698412698414e-06, + "loss": 0.0081, + "step": 53 + }, + { + "epoch": 0.09, + "learning_rate": 8.571428571428573e-06, + "loss": 0.0074, + "step": 54 + }, + { + "epoch": 0.09, + "learning_rate": 8.73015873015873e-06, + "loss": 0.0094, + "step": 55 + }, + { + "epoch": 0.09, + "learning_rate": 8.88888888888889e-06, + "loss": 0.0095, + "step": 56 + }, + { + "epoch": 0.09, + "learning_rate": 9.047619047619047e-06, + "loss": 0.0069, + "step": 57 + }, + { + "epoch": 0.09, + "learning_rate": 9.206349206349207e-06, + "loss": 0.0061, + "step": 58 + }, + { + "epoch": 0.09, + "learning_rate": 9.365079365079366e-06, + "loss": 0.0055, + "step": 59 + }, + { + "epoch": 0.1, + "learning_rate": 9.523809523809523e-06, + "loss": 0.005, + "step": 60 + }, + { + "epoch": 0.1, + "learning_rate": 9.682539682539683e-06, + "loss": 0.0047, + "step": 61 + }, + { + "epoch": 0.1, + "learning_rate": 9.841269841269842e-06, + "loss": 0.0045, + "step": 62 + }, + { + "epoch": 0.1, + "learning_rate": 1e-05, + "loss": 0.0057, + "step": 63 + }, + { + "epoch": 0.1, + "learning_rate": 1.015873015873016e-05, + "loss": 0.0038, + "step": 64 + }, + { + "epoch": 0.1, + "learning_rate": 1.0317460317460318e-05, + "loss": 0.004, + "step": 65 + }, + { + "epoch": 0.1, + "learning_rate": 1.0476190476190477e-05, + "loss": 0.0033, + "step": 66 + }, + { + "epoch": 0.11, + "learning_rate": 1.0634920634920636e-05, + "loss": 0.1076, + "step": 67 + }, + { + "epoch": 0.11, + "learning_rate": 1.0793650793650794e-05, + "loss": 0.0039, + "step": 68 + }, + { + "epoch": 0.11, + "learning_rate": 1.0952380952380953e-05, + "loss": 0.0032, + "step": 69 + }, + { + "epoch": 0.11, + "learning_rate": 1.1111111111111112e-05, + "loss": 0.0026, + "step": 70 + }, + { + "epoch": 0.11, + "learning_rate": 1.126984126984127e-05, + "loss": 0.0025, + "step": 71 + }, + { + "epoch": 0.11, + "learning_rate": 1.1428571428571429e-05, + "loss": 0.0027, + "step": 72 + }, + { + "epoch": 0.12, + "learning_rate": 1.1587301587301588e-05, + "loss": 0.0024, + "step": 73 + }, + { + "epoch": 0.12, + "learning_rate": 1.1746031746031746e-05, + "loss": 0.0024, + "step": 74 + }, + { + "epoch": 0.12, + "learning_rate": 1.1904761904761905e-05, + "loss": 0.002, + "step": 75 + }, + { + "epoch": 0.12, + "learning_rate": 1.2063492063492064e-05, + "loss": 0.0023, + "step": 76 + }, + { + "epoch": 0.12, + "learning_rate": 1.2222222222222222e-05, + "loss": 0.0021, + "step": 77 + }, + { + "epoch": 0.12, + "learning_rate": 1.2380952380952381e-05, + "loss": 0.0019, + "step": 78 + }, + { + "epoch": 0.13, + "learning_rate": 1.253968253968254e-05, + "loss": 0.0019, + "step": 79 + }, + { + "epoch": 0.13, + "learning_rate": 1.2698412698412699e-05, + "loss": 0.0017, + "step": 80 + }, + { + "epoch": 0.13, + "learning_rate": 1.2857142857142857e-05, + "loss": 0.0017, + "step": 81 + }, + { + "epoch": 0.13, + "learning_rate": 1.3015873015873018e-05, + "loss": 0.0015, + "step": 82 + }, + { + "epoch": 0.13, + "learning_rate": 1.3174603174603176e-05, + "loss": 0.0016, + "step": 83 + }, + { + "epoch": 0.13, + "learning_rate": 1.3333333333333333e-05, + "loss": 0.0014, + "step": 84 + }, + { + "epoch": 0.13, + "learning_rate": 1.3492063492063492e-05, + "loss": 0.0014, + "step": 85 + }, + { + "epoch": 0.14, + "learning_rate": 1.365079365079365e-05, + "loss": 0.0803, + "step": 86 + }, + { + "epoch": 0.14, + "learning_rate": 1.3809523809523811e-05, + "loss": 0.0013, + "step": 87 + }, + { + "epoch": 0.14, + "learning_rate": 1.396825396825397e-05, + "loss": 0.0013, + "step": 88 + }, + { + "epoch": 0.14, + "learning_rate": 1.4126984126984127e-05, + "loss": 0.0018, + "step": 89 + }, + { + "epoch": 0.14, + "learning_rate": 1.4285714285714285e-05, + "loss": 0.0012, + "step": 90 + }, + { + "epoch": 0.14, + "learning_rate": 1.4444444444444444e-05, + "loss": 0.0012, + "step": 91 + }, + { + "epoch": 0.15, + "learning_rate": 1.4603174603174605e-05, + "loss": 0.0013, + "step": 92 + }, + { + "epoch": 0.15, + "learning_rate": 1.4761904761904763e-05, + "loss": 0.0263, + "step": 93 + }, + { + "epoch": 0.15, + "learning_rate": 1.4920634920634922e-05, + "loss": 0.0729, + "step": 94 + }, + { + "epoch": 0.15, + "learning_rate": 1.5079365079365079e-05, + "loss": 0.0012, + "step": 95 + }, + { + "epoch": 0.15, + "learning_rate": 1.5238095238095241e-05, + "loss": 0.0012, + "step": 96 + }, + { + "epoch": 0.15, + "learning_rate": 1.5396825396825398e-05, + "loss": 0.0082, + "step": 97 + }, + { + "epoch": 0.16, + "learning_rate": 1.5555555555555555e-05, + "loss": 0.0014, + "step": 98 + }, + { + "epoch": 0.16, + "learning_rate": 1.5714285714285715e-05, + "loss": 0.0014, + "step": 99 + }, + { + "epoch": 0.16, + "learning_rate": 1.5873015873015872e-05, + "loss": 0.0017, + "step": 100 + }, + { + "epoch": 0.16, + "learning_rate": 1.6031746031746033e-05, + "loss": 0.0013, + "step": 101 + }, + { + "epoch": 0.16, + "learning_rate": 1.6190476190476193e-05, + "loss": 0.0026, + "step": 102 + }, + { + "epoch": 0.16, + "learning_rate": 1.634920634920635e-05, + "loss": 0.0012, + "step": 103 + }, + { + "epoch": 0.17, + "learning_rate": 1.6507936507936507e-05, + "loss": 0.0009, + "step": 104 + }, + { + "epoch": 0.17, + "learning_rate": 1.6666666666666667e-05, + "loss": 0.0012, + "step": 105 + }, + { + "epoch": 0.17, + "learning_rate": 1.6825396825396828e-05, + "loss": 0.0012, + "step": 106 + }, + { + "epoch": 0.17, + "learning_rate": 1.6984126984126985e-05, + "loss": 0.001, + "step": 107 + }, + { + "epoch": 0.17, + "learning_rate": 1.7142857142857145e-05, + "loss": 0.001, + "step": 108 + }, + { + "epoch": 0.17, + "learning_rate": 1.7301587301587302e-05, + "loss": 0.0056, + "step": 109 + }, + { + "epoch": 0.17, + "learning_rate": 1.746031746031746e-05, + "loss": 0.0011, + "step": 110 + }, + { + "epoch": 0.18, + "learning_rate": 1.761904761904762e-05, + "loss": 0.0007, + "step": 111 + }, + { + "epoch": 0.18, + "learning_rate": 1.777777777777778e-05, + "loss": 0.0008, + "step": 112 + }, + { + "epoch": 0.18, + "learning_rate": 1.7936507936507937e-05, + "loss": 0.0007, + "step": 113 + }, + { + "epoch": 0.18, + "learning_rate": 1.8095238095238094e-05, + "loss": 0.0007, + "step": 114 + }, + { + "epoch": 0.18, + "learning_rate": 1.8253968253968254e-05, + "loss": 0.0007, + "step": 115 + }, + { + "epoch": 0.18, + "learning_rate": 1.8412698412698415e-05, + "loss": 0.0007, + "step": 116 + }, + { + "epoch": 0.19, + "learning_rate": 1.8571428571428572e-05, + "loss": 0.0007, + "step": 117 + }, + { + "epoch": 0.19, + "learning_rate": 1.8730158730158732e-05, + "loss": 0.0006, + "step": 118 + }, + { + "epoch": 0.19, + "learning_rate": 1.888888888888889e-05, + "loss": 0.1116, + "step": 119 + }, + { + "epoch": 0.19, + "learning_rate": 1.9047619047619046e-05, + "loss": 0.0007, + "step": 120 + }, + { + "epoch": 0.19, + "learning_rate": 1.920634920634921e-05, + "loss": 0.0006, + "step": 121 + }, + { + "epoch": 0.19, + "learning_rate": 1.9365079365079367e-05, + "loss": 0.0006, + "step": 122 + }, + { + "epoch": 0.2, + "learning_rate": 1.9523809523809524e-05, + "loss": 0.0005, + "step": 123 + }, + { + "epoch": 0.2, + "learning_rate": 1.9682539682539684e-05, + "loss": 0.0005, + "step": 124 + }, + { + "epoch": 0.2, + "learning_rate": 1.984126984126984e-05, + "loss": 0.0005, + "step": 125 + }, + { + "epoch": 0.2, + "learning_rate": 2e-05, + "loss": 0.0005, + "step": 126 + }, + { + "epoch": 0.2, + "learning_rate": 2.015873015873016e-05, + "loss": 0.0005, + "step": 127 + }, + { + "epoch": 0.2, + "learning_rate": 2.031746031746032e-05, + "loss": 0.0005, + "step": 128 + }, + { + "epoch": 0.2, + "learning_rate": 2.0476190476190476e-05, + "loss": 0.0005, + "step": 129 + }, + { + "epoch": 0.21, + "learning_rate": 2.0634920634920636e-05, + "loss": 0.0005, + "step": 130 + }, + { + "epoch": 0.21, + "learning_rate": 2.0793650793650797e-05, + "loss": 0.0004, + "step": 131 + }, + { + "epoch": 0.21, + "learning_rate": 2.0952380952380954e-05, + "loss": 0.0031, + "step": 132 + }, + { + "epoch": 0.21, + "learning_rate": 2.111111111111111e-05, + "loss": 0.0021, + "step": 133 + }, + { + "epoch": 0.21, + "learning_rate": 2.126984126984127e-05, + "loss": 0.0004, + "step": 134 + }, + { + "epoch": 0.21, + "learning_rate": 2.1428571428571428e-05, + "loss": 0.0004, + "step": 135 + }, + { + "epoch": 0.22, + "learning_rate": 2.158730158730159e-05, + "loss": 0.0004, + "step": 136 + }, + { + "epoch": 0.22, + "learning_rate": 2.174603174603175e-05, + "loss": 0.0004, + "step": 137 + }, + { + "epoch": 0.22, + "learning_rate": 2.1904761904761906e-05, + "loss": 0.0004, + "step": 138 + }, + { + "epoch": 0.22, + "learning_rate": 2.2063492063492063e-05, + "loss": 0.0004, + "step": 139 + }, + { + "epoch": 0.22, + "learning_rate": 2.2222222222222223e-05, + "loss": 0.0004, + "step": 140 + }, + { + "epoch": 0.22, + "learning_rate": 2.2380952380952384e-05, + "loss": 0.0004, + "step": 141 + }, + { + "epoch": 0.23, + "learning_rate": 2.253968253968254e-05, + "loss": 0.0004, + "step": 142 + }, + { + "epoch": 0.23, + "learning_rate": 2.2698412698412698e-05, + "loss": 0.0003, + "step": 143 + }, + { + "epoch": 0.23, + "learning_rate": 2.2857142857142858e-05, + "loss": 0.0003, + "step": 144 + }, + { + "epoch": 0.23, + "learning_rate": 2.3015873015873015e-05, + "loss": 0.0004, + "step": 145 + }, + { + "epoch": 0.23, + "learning_rate": 2.3174603174603175e-05, + "loss": 0.0004, + "step": 146 + }, + { + "epoch": 0.23, + "learning_rate": 2.3333333333333336e-05, + "loss": 0.0004, + "step": 147 + }, + { + "epoch": 0.23, + "learning_rate": 2.3492063492063493e-05, + "loss": 0.0003, + "step": 148 + }, + { + "epoch": 0.24, + "learning_rate": 2.365079365079365e-05, + "loss": 0.0003, + "step": 149 + }, + { + "epoch": 0.24, + "learning_rate": 2.380952380952381e-05, + "loss": 0.0003, + "step": 150 + }, + { + "epoch": 0.24, + "learning_rate": 2.396825396825397e-05, + "loss": 0.0003, + "step": 151 + }, + { + "epoch": 0.24, + "learning_rate": 2.4126984126984128e-05, + "loss": 0.0003, + "step": 152 + }, + { + "epoch": 0.24, + "learning_rate": 2.4285714285714288e-05, + "loss": 0.0003, + "step": 153 + }, + { + "epoch": 0.24, + "learning_rate": 2.4444444444444445e-05, + "loss": 0.0003, + "step": 154 + }, + { + "epoch": 0.25, + "learning_rate": 2.4603174603174602e-05, + "loss": 0.001, + "step": 155 + }, + { + "epoch": 0.25, + "learning_rate": 2.4761904761904762e-05, + "loss": 0.0003, + "step": 156 + }, + { + "epoch": 0.25, + "learning_rate": 2.4920634920634923e-05, + "loss": 0.0003, + "step": 157 + }, + { + "epoch": 0.25, + "learning_rate": 2.507936507936508e-05, + "loss": 0.1312, + "step": 158 + }, + { + "epoch": 0.25, + "learning_rate": 2.523809523809524e-05, + "loss": 0.0003, + "step": 159 + }, + { + "epoch": 0.25, + "learning_rate": 2.5396825396825397e-05, + "loss": 0.0003, + "step": 160 + }, + { + "epoch": 0.26, + "learning_rate": 2.5555555555555554e-05, + "loss": 0.0003, + "step": 161 + }, + { + "epoch": 0.26, + "learning_rate": 2.5714285714285714e-05, + "loss": 0.0434, + "step": 162 + }, + { + "epoch": 0.26, + "learning_rate": 2.5873015873015878e-05, + "loss": 0.0004, + "step": 163 + }, + { + "epoch": 0.26, + "learning_rate": 2.6031746031746035e-05, + "loss": 0.0012, + "step": 164 + }, + { + "epoch": 0.26, + "learning_rate": 2.6190476190476192e-05, + "loss": 0.0007, + "step": 165 + }, + { + "epoch": 0.26, + "learning_rate": 2.6349206349206353e-05, + "loss": 0.0009, + "step": 166 + }, + { + "epoch": 0.27, + "learning_rate": 2.650793650793651e-05, + "loss": 0.001, + "step": 167 + }, + { + "epoch": 0.27, + "learning_rate": 2.6666666666666667e-05, + "loss": 0.0032, + "step": 168 + }, + { + "epoch": 0.27, + "learning_rate": 2.6825396825396827e-05, + "loss": 0.0009, + "step": 169 + }, + { + "epoch": 0.27, + "learning_rate": 2.6984126984126984e-05, + "loss": 0.0009, + "step": 170 + }, + { + "epoch": 0.27, + "learning_rate": 2.714285714285714e-05, + "loss": 0.0012, + "step": 171 + }, + { + "epoch": 0.27, + "learning_rate": 2.73015873015873e-05, + "loss": 0.0005, + "step": 172 + }, + { + "epoch": 0.27, + "learning_rate": 2.7460317460317465e-05, + "loss": 0.0004, + "step": 173 + }, + { + "epoch": 0.28, + "learning_rate": 2.7619047619047622e-05, + "loss": 0.0008, + "step": 174 + }, + { + "epoch": 0.28, + "learning_rate": 2.777777777777778e-05, + "loss": 0.0004, + "step": 175 + }, + { + "epoch": 0.28, + "learning_rate": 2.793650793650794e-05, + "loss": 0.0003, + "step": 176 + }, + { + "epoch": 0.28, + "learning_rate": 2.8095238095238096e-05, + "loss": 0.0004, + "step": 177 + }, + { + "epoch": 0.28, + "learning_rate": 2.8253968253968253e-05, + "loss": 0.0004, + "step": 178 + }, + { + "epoch": 0.28, + "learning_rate": 2.8412698412698414e-05, + "loss": 0.0003, + "step": 179 + }, + { + "epoch": 0.29, + "learning_rate": 2.857142857142857e-05, + "loss": 0.0003, + "step": 180 + }, + { + "epoch": 0.29, + "learning_rate": 2.8730158730158728e-05, + "loss": 0.0005, + "step": 181 + }, + { + "epoch": 0.29, + "learning_rate": 2.8888888888888888e-05, + "loss": 0.0003, + "step": 182 + }, + { + "epoch": 0.29, + "learning_rate": 2.9047619047619052e-05, + "loss": 0.0006, + "step": 183 + }, + { + "epoch": 0.29, + "learning_rate": 2.920634920634921e-05, + "loss": 0.0006, + "step": 184 + }, + { + "epoch": 0.29, + "learning_rate": 2.9365079365079366e-05, + "loss": 0.023, + "step": 185 + }, + { + "epoch": 0.3, + "learning_rate": 2.9523809523809526e-05, + "loss": 0.0007, + "step": 186 + }, + { + "epoch": 0.3, + "learning_rate": 2.9682539682539683e-05, + "loss": 0.0003, + "step": 187 + }, + { + "epoch": 0.3, + "learning_rate": 2.9841269841269844e-05, + "loss": 0.0003, + "step": 188 + }, + { + "epoch": 0.3, + "learning_rate": 3e-05, + "loss": 0.0002, + "step": 189 + }, + { + "epoch": 0.3, + "learning_rate": 3.0158730158730158e-05, + "loss": 0.0003, + "step": 190 + }, + { + "epoch": 0.3, + "learning_rate": 3.0317460317460318e-05, + "loss": 0.0003, + "step": 191 + }, + { + "epoch": 0.3, + "learning_rate": 3.0476190476190482e-05, + "loss": 0.0002, + "step": 192 + }, + { + "epoch": 0.31, + "learning_rate": 3.063492063492064e-05, + "loss": 0.0023, + "step": 193 + }, + { + "epoch": 0.31, + "learning_rate": 3.0793650793650796e-05, + "loss": 0.0003, + "step": 194 + }, + { + "epoch": 0.31, + "learning_rate": 3.095238095238095e-05, + "loss": 0.0003, + "step": 195 + }, + { + "epoch": 0.31, + "learning_rate": 3.111111111111111e-05, + "loss": 0.0004, + "step": 196 + }, + { + "epoch": 0.31, + "learning_rate": 3.1269841269841274e-05, + "loss": 0.0002, + "step": 197 + }, + { + "epoch": 0.31, + "learning_rate": 3.142857142857143e-05, + "loss": 0.0003, + "step": 198 + }, + { + "epoch": 0.32, + "learning_rate": 3.158730158730159e-05, + "loss": 0.0002, + "step": 199 + }, + { + "epoch": 0.32, + "learning_rate": 3.1746031746031745e-05, + "loss": 0.0003, + "step": 200 + }, + { + "epoch": 0.32, + "learning_rate": 3.19047619047619e-05, + "loss": 0.0003, + "step": 201 + }, + { + "epoch": 0.32, + "learning_rate": 3.2063492063492065e-05, + "loss": 0.0002, + "step": 202 + }, + { + "epoch": 0.32, + "learning_rate": 3.222222222222223e-05, + "loss": 0.0002, + "step": 203 + }, + { + "epoch": 0.32, + "learning_rate": 3.2380952380952386e-05, + "loss": 0.0003, + "step": 204 + }, + { + "epoch": 0.33, + "learning_rate": 3.253968253968254e-05, + "loss": 0.0002, + "step": 205 + }, + { + "epoch": 0.33, + "learning_rate": 3.26984126984127e-05, + "loss": 0.0002, + "step": 206 + }, + { + "epoch": 0.33, + "learning_rate": 3.285714285714286e-05, + "loss": 0.0002, + "step": 207 + }, + { + "epoch": 0.33, + "learning_rate": 3.3015873015873014e-05, + "loss": 0.0077, + "step": 208 + }, + { + "epoch": 0.33, + "learning_rate": 3.317460317460318e-05, + "loss": 0.0002, + "step": 209 + }, + { + "epoch": 0.33, + "learning_rate": 3.3333333333333335e-05, + "loss": 0.0002, + "step": 210 + }, + { + "epoch": 0.33, + "learning_rate": 3.349206349206349e-05, + "loss": 0.0002, + "step": 211 + }, + { + "epoch": 0.34, + "learning_rate": 3.3650793650793656e-05, + "loss": 0.0002, + "step": 212 + }, + { + "epoch": 0.34, + "learning_rate": 3.380952380952381e-05, + "loss": 0.0004, + "step": 213 + }, + { + "epoch": 0.34, + "learning_rate": 3.396825396825397e-05, + "loss": 0.0002, + "step": 214 + }, + { + "epoch": 0.34, + "learning_rate": 3.412698412698413e-05, + "loss": 0.0002, + "step": 215 + }, + { + "epoch": 0.34, + "learning_rate": 3.428571428571429e-05, + "loss": 0.0002, + "step": 216 + }, + { + "epoch": 0.34, + "learning_rate": 3.444444444444445e-05, + "loss": 0.0002, + "step": 217 + }, + { + "epoch": 0.35, + "learning_rate": 3.4603174603174604e-05, + "loss": 0.0002, + "step": 218 + }, + { + "epoch": 0.35, + "learning_rate": 3.476190476190476e-05, + "loss": 0.0002, + "step": 219 + }, + { + "epoch": 0.35, + "learning_rate": 3.492063492063492e-05, + "loss": 0.0002, + "step": 220 + }, + { + "epoch": 0.35, + "learning_rate": 3.5079365079365075e-05, + "loss": 0.0002, + "step": 221 + }, + { + "epoch": 0.35, + "learning_rate": 3.523809523809524e-05, + "loss": 0.0002, + "step": 222 + }, + { + "epoch": 0.35, + "learning_rate": 3.53968253968254e-05, + "loss": 0.0002, + "step": 223 + }, + { + "epoch": 0.36, + "learning_rate": 3.555555555555556e-05, + "loss": 0.2106, + "step": 224 + }, + { + "epoch": 0.36, + "learning_rate": 3.571428571428572e-05, + "loss": 0.0002, + "step": 225 + }, + { + "epoch": 0.36, + "learning_rate": 3.5873015873015874e-05, + "loss": 0.0002, + "step": 226 + }, + { + "epoch": 0.36, + "learning_rate": 3.603174603174603e-05, + "loss": 0.0003, + "step": 227 + }, + { + "epoch": 0.36, + "learning_rate": 3.619047619047619e-05, + "loss": 0.0004, + "step": 228 + }, + { + "epoch": 0.36, + "learning_rate": 3.634920634920635e-05, + "loss": 0.0006, + "step": 229 + }, + { + "epoch": 0.37, + "learning_rate": 3.650793650793651e-05, + "loss": 0.001, + "step": 230 + }, + { + "epoch": 0.37, + "learning_rate": 3.6666666666666666e-05, + "loss": 0.0011, + "step": 231 + }, + { + "epoch": 0.37, + "learning_rate": 3.682539682539683e-05, + "loss": 0.0022, + "step": 232 + }, + { + "epoch": 0.37, + "learning_rate": 3.6984126984126986e-05, + "loss": 0.0021, + "step": 233 + }, + { + "epoch": 0.37, + "learning_rate": 3.7142857142857143e-05, + "loss": 0.0009, + "step": 234 + }, + { + "epoch": 0.37, + "learning_rate": 3.730158730158731e-05, + "loss": 0.001, + "step": 235 + }, + { + "epoch": 0.37, + "learning_rate": 3.7460317460317464e-05, + "loss": 0.0006, + "step": 236 + }, + { + "epoch": 0.38, + "learning_rate": 3.761904761904762e-05, + "loss": 0.0005, + "step": 237 + }, + { + "epoch": 0.38, + "learning_rate": 3.777777777777778e-05, + "loss": 0.0003, + "step": 238 + }, + { + "epoch": 0.38, + "learning_rate": 3.7936507936507935e-05, + "loss": 0.0003, + "step": 239 + }, + { + "epoch": 0.38, + "learning_rate": 3.809523809523809e-05, + "loss": 0.0002, + "step": 240 + }, + { + "epoch": 0.38, + "learning_rate": 3.8253968253968256e-05, + "loss": 0.0002, + "step": 241 + }, + { + "epoch": 0.38, + "learning_rate": 3.841269841269842e-05, + "loss": 0.0002, + "step": 242 + }, + { + "epoch": 0.39, + "learning_rate": 3.857142857142858e-05, + "loss": 0.0002, + "step": 243 + }, + { + "epoch": 0.39, + "learning_rate": 3.8730158730158734e-05, + "loss": 0.0002, + "step": 244 + }, + { + "epoch": 0.39, + "learning_rate": 3.888888888888889e-05, + "loss": 0.0002, + "step": 245 + }, + { + "epoch": 0.39, + "learning_rate": 3.904761904761905e-05, + "loss": 0.0002, + "step": 246 + }, + { + "epoch": 0.39, + "learning_rate": 3.9206349206349205e-05, + "loss": 0.0002, + "step": 247 + }, + { + "epoch": 0.39, + "learning_rate": 3.936507936507937e-05, + "loss": 0.0004, + "step": 248 + }, + { + "epoch": 0.4, + "learning_rate": 3.9523809523809526e-05, + "loss": 0.0002, + "step": 249 + }, + { + "epoch": 0.4, + "learning_rate": 3.968253968253968e-05, + "loss": 0.0002, + "step": 250 + }, + { + "epoch": 0.4, + "learning_rate": 3.984126984126984e-05, + "loss": 0.0002, + "step": 251 + }, + { + "epoch": 0.4, + "learning_rate": 4e-05, + "loss": 0.0002, + "step": 252 + }, + { + "epoch": 0.4, + "learning_rate": 4.015873015873016e-05, + "loss": 0.0002, + "step": 253 + }, + { + "epoch": 0.4, + "learning_rate": 4.031746031746032e-05, + "loss": 0.0002, + "step": 254 + }, + { + "epoch": 0.4, + "learning_rate": 4.047619047619048e-05, + "loss": 0.0002, + "step": 255 + }, + { + "epoch": 0.41, + "learning_rate": 4.063492063492064e-05, + "loss": 0.0002, + "step": 256 + }, + { + "epoch": 0.41, + "learning_rate": 4.0793650793650795e-05, + "loss": 0.0001, + "step": 257 + }, + { + "epoch": 0.41, + "learning_rate": 4.095238095238095e-05, + "loss": 0.0001, + "step": 258 + }, + { + "epoch": 0.41, + "learning_rate": 4.111111111111111e-05, + "loss": 0.0003, + "step": 259 + }, + { + "epoch": 0.41, + "learning_rate": 4.126984126984127e-05, + "loss": 0.0001, + "step": 260 + }, + { + "epoch": 0.41, + "learning_rate": 4.1428571428571437e-05, + "loss": 0.0002, + "step": 261 + }, + { + "epoch": 0.42, + "learning_rate": 4.1587301587301594e-05, + "loss": 0.0002, + "step": 262 + }, + { + "epoch": 0.42, + "learning_rate": 4.174603174603175e-05, + "loss": 0.0003, + "step": 263 + }, + { + "epoch": 0.42, + "learning_rate": 4.190476190476191e-05, + "loss": 0.0001, + "step": 264 + }, + { + "epoch": 0.42, + "learning_rate": 4.2063492063492065e-05, + "loss": 0.0002, + "step": 265 + }, + { + "epoch": 0.42, + "learning_rate": 4.222222222222222e-05, + "loss": 0.0001, + "step": 266 + }, + { + "epoch": 0.42, + "learning_rate": 4.2380952380952385e-05, + "loss": 0.0001, + "step": 267 + }, + { + "epoch": 0.43, + "learning_rate": 4.253968253968254e-05, + "loss": 0.0001, + "step": 268 + }, + { + "epoch": 0.43, + "learning_rate": 4.26984126984127e-05, + "loss": 0.0001, + "step": 269 + }, + { + "epoch": 0.43, + "learning_rate": 4.2857142857142856e-05, + "loss": 0.0001, + "step": 270 + }, + { + "epoch": 0.43, + "learning_rate": 4.301587301587302e-05, + "loss": 0.0001, + "step": 271 + }, + { + "epoch": 0.43, + "learning_rate": 4.317460317460318e-05, + "loss": 0.0001, + "step": 272 + }, + { + "epoch": 0.43, + "learning_rate": 4.3333333333333334e-05, + "loss": 0.0001, + "step": 273 + }, + { + "epoch": 0.43, + "learning_rate": 4.34920634920635e-05, + "loss": 0.0001, + "step": 274 + }, + { + "epoch": 0.44, + "learning_rate": 4.3650793650793655e-05, + "loss": 0.0008, + "step": 275 + }, + { + "epoch": 0.44, + "learning_rate": 4.380952380952381e-05, + "loss": 0.0001, + "step": 276 + }, + { + "epoch": 0.44, + "learning_rate": 4.396825396825397e-05, + "loss": 0.0001, + "step": 277 + }, + { + "epoch": 0.44, + "learning_rate": 4.4126984126984126e-05, + "loss": 0.0001, + "step": 278 + }, + { + "epoch": 0.44, + "learning_rate": 4.428571428571428e-05, + "loss": 0.0002, + "step": 279 + }, + { + "epoch": 0.44, + "learning_rate": 4.4444444444444447e-05, + "loss": 0.0001, + "step": 280 + }, + { + "epoch": 0.45, + "learning_rate": 4.460317460317461e-05, + "loss": 0.0555, + "step": 281 + }, + { + "epoch": 0.45, + "learning_rate": 4.476190476190477e-05, + "loss": 0.0001, + "step": 282 + }, + { + "epoch": 0.45, + "learning_rate": 4.4920634920634924e-05, + "loss": 0.0001, + "step": 283 + }, + { + "epoch": 0.45, + "learning_rate": 4.507936507936508e-05, + "loss": 0.0001, + "step": 284 + }, + { + "epoch": 0.45, + "learning_rate": 4.523809523809524e-05, + "loss": 0.0001, + "step": 285 + }, + { + "epoch": 0.45, + "learning_rate": 4.5396825396825395e-05, + "loss": 0.0002, + "step": 286 + }, + { + "epoch": 0.46, + "learning_rate": 4.555555555555556e-05, + "loss": 0.0001, + "step": 287 + }, + { + "epoch": 0.46, + "learning_rate": 4.5714285714285716e-05, + "loss": 0.0001, + "step": 288 + }, + { + "epoch": 0.46, + "learning_rate": 4.587301587301587e-05, + "loss": 0.0001, + "step": 289 + }, + { + "epoch": 0.46, + "learning_rate": 4.603174603174603e-05, + "loss": 0.0002, + "step": 290 + }, + { + "epoch": 0.46, + "learning_rate": 4.6190476190476194e-05, + "loss": 0.0001, + "step": 291 + }, + { + "epoch": 0.46, + "learning_rate": 4.634920634920635e-05, + "loss": 0.0001, + "step": 292 + }, + { + "epoch": 0.47, + "learning_rate": 4.6507936507936515e-05, + "loss": 0.0001, + "step": 293 + }, + { + "epoch": 0.47, + "learning_rate": 4.666666666666667e-05, + "loss": 0.0001, + "step": 294 + }, + { + "epoch": 0.47, + "learning_rate": 4.682539682539683e-05, + "loss": 0.0003, + "step": 295 + }, + { + "epoch": 0.47, + "learning_rate": 4.6984126984126986e-05, + "loss": 0.0001, + "step": 296 + }, + { + "epoch": 0.47, + "learning_rate": 4.714285714285714e-05, + "loss": 0.0001, + "step": 297 + }, + { + "epoch": 0.47, + "learning_rate": 4.73015873015873e-05, + "loss": 0.0001, + "step": 298 + }, + { + "epoch": 0.47, + "learning_rate": 4.746031746031746e-05, + "loss": 0.0001, + "step": 299 + }, + { + "epoch": 0.48, + "learning_rate": 4.761904761904762e-05, + "loss": 0.0001, + "step": 300 + }, + { + "epoch": 0.48, + "learning_rate": 4.7777777777777784e-05, + "loss": 0.0001, + "step": 301 + }, + { + "epoch": 0.48, + "learning_rate": 4.793650793650794e-05, + "loss": 0.0001, + "step": 302 + }, + { + "epoch": 0.48, + "learning_rate": 4.80952380952381e-05, + "loss": 0.0001, + "step": 303 + }, + { + "epoch": 0.48, + "learning_rate": 4.8253968253968255e-05, + "loss": 0.0001, + "step": 304 + }, + { + "epoch": 0.48, + "learning_rate": 4.841269841269841e-05, + "loss": 0.0762, + "step": 305 + }, + { + "epoch": 0.49, + "learning_rate": 4.8571428571428576e-05, + "loss": 0.0001, + "step": 306 + }, + { + "epoch": 0.49, + "learning_rate": 4.873015873015873e-05, + "loss": 0.0001, + "step": 307 + }, + { + "epoch": 0.49, + "learning_rate": 4.888888888888889e-05, + "loss": 0.0001, + "step": 308 + }, + { + "epoch": 0.49, + "learning_rate": 4.904761904761905e-05, + "loss": 0.0001, + "step": 309 + }, + { + "epoch": 0.49, + "learning_rate": 4.9206349206349204e-05, + "loss": 0.0003, + "step": 310 + }, + { + "epoch": 0.49, + "learning_rate": 4.936507936507937e-05, + "loss": 0.0003, + "step": 311 + }, + { + "epoch": 0.5, + "learning_rate": 4.9523809523809525e-05, + "loss": 0.0001, + "step": 312 + }, + { + "epoch": 0.5, + "learning_rate": 4.968253968253969e-05, + "loss": 0.0002, + "step": 313 + }, + { + "epoch": 0.5, + "learning_rate": 4.9841269841269845e-05, + "loss": 0.0002, + "step": 314 + }, + { + "epoch": 0.5, + "learning_rate": 5e-05, + "loss": 0.0006, + "step": 315 + }, + { + "epoch": 0.5, + "learning_rate": 4.998236331569665e-05, + "loss": 0.0023, + "step": 316 + }, + { + "epoch": 0.5, + "learning_rate": 4.99647266313933e-05, + "loss": 0.0009, + "step": 317 + }, + { + "epoch": 0.5, + "learning_rate": 4.9947089947089946e-05, + "loss": 0.0002, + "step": 318 + }, + { + "epoch": 0.51, + "learning_rate": 4.992945326278659e-05, + "loss": 0.0001, + "step": 319 + }, + { + "epoch": 0.51, + "learning_rate": 4.991181657848325e-05, + "loss": 0.0001, + "step": 320 + }, + { + "epoch": 0.51, + "learning_rate": 4.9894179894179895e-05, + "loss": 0.0002, + "step": 321 + }, + { + "epoch": 0.51, + "learning_rate": 4.987654320987655e-05, + "loss": 0.0001, + "step": 322 + }, + { + "epoch": 0.51, + "learning_rate": 4.98589065255732e-05, + "loss": 0.0001, + "step": 323 + }, + { + "epoch": 0.51, + "learning_rate": 4.9841269841269845e-05, + "loss": 0.0001, + "step": 324 + }, + { + "epoch": 0.52, + "learning_rate": 4.982363315696649e-05, + "loss": 0.0001, + "step": 325 + }, + { + "epoch": 0.52, + "learning_rate": 4.980599647266314e-05, + "loss": 0.0001, + "step": 326 + }, + { + "epoch": 0.52, + "learning_rate": 4.978835978835979e-05, + "loss": 0.0001, + "step": 327 + }, + { + "epoch": 0.52, + "learning_rate": 4.9770723104056436e-05, + "loss": 0.0001, + "step": 328 + }, + { + "epoch": 0.52, + "learning_rate": 4.9753086419753084e-05, + "loss": 0.0001, + "step": 329 + }, + { + "epoch": 0.52, + "learning_rate": 4.973544973544973e-05, + "loss": 0.0001, + "step": 330 + }, + { + "epoch": 0.53, + "learning_rate": 4.9717813051146386e-05, + "loss": 0.0001, + "step": 331 + }, + { + "epoch": 0.53, + "learning_rate": 4.970017636684304e-05, + "loss": 0.0001, + "step": 332 + }, + { + "epoch": 0.53, + "learning_rate": 4.968253968253969e-05, + "loss": 0.0001, + "step": 333 + }, + { + "epoch": 0.53, + "learning_rate": 4.9664902998236336e-05, + "loss": 0.0001, + "step": 334 + }, + { + "epoch": 0.53, + "learning_rate": 4.9647266313932984e-05, + "loss": 0.0001, + "step": 335 + }, + { + "epoch": 0.53, + "learning_rate": 4.962962962962963e-05, + "loss": 0.0001, + "step": 336 + }, + { + "epoch": 0.53, + "learning_rate": 4.961199294532628e-05, + "loss": 0.0001, + "step": 337 + }, + { + "epoch": 0.54, + "learning_rate": 4.959435626102293e-05, + "loss": 0.0001, + "step": 338 + }, + { + "epoch": 0.54, + "learning_rate": 4.9576719576719575e-05, + "loss": 0.0001, + "step": 339 + }, + { + "epoch": 0.54, + "learning_rate": 4.955908289241622e-05, + "loss": 0.0001, + "step": 340 + }, + { + "epoch": 0.54, + "learning_rate": 4.954144620811288e-05, + "loss": 0.0001, + "step": 341 + }, + { + "epoch": 0.54, + "learning_rate": 4.9523809523809525e-05, + "loss": 0.0001, + "step": 342 + }, + { + "epoch": 0.54, + "learning_rate": 4.950617283950618e-05, + "loss": 0.0001, + "step": 343 + }, + { + "epoch": 0.55, + "learning_rate": 4.948853615520283e-05, + "loss": 0.0001, + "step": 344 + }, + { + "epoch": 0.55, + "learning_rate": 4.9470899470899475e-05, + "loss": 0.0001, + "step": 345 + }, + { + "epoch": 0.55, + "learning_rate": 4.945326278659612e-05, + "loss": 0.0001, + "step": 346 + }, + { + "epoch": 0.55, + "learning_rate": 4.943562610229277e-05, + "loss": 0.0001, + "step": 347 + }, + { + "epoch": 0.55, + "learning_rate": 4.941798941798942e-05, + "loss": 0.0001, + "step": 348 + }, + { + "epoch": 0.55, + "learning_rate": 4.9400352733686065e-05, + "loss": 0.2247, + "step": 349 + }, + { + "epoch": 0.56, + "learning_rate": 4.938271604938271e-05, + "loss": 0.0005, + "step": 350 + }, + { + "epoch": 0.56, + "learning_rate": 4.936507936507937e-05, + "loss": 0.0001, + "step": 351 + }, + { + "epoch": 0.56, + "learning_rate": 4.9347442680776015e-05, + "loss": 0.0001, + "step": 352 + }, + { + "epoch": 0.56, + "learning_rate": 4.932980599647267e-05, + "loss": 0.0001, + "step": 353 + }, + { + "epoch": 0.56, + "learning_rate": 4.931216931216932e-05, + "loss": 0.0001, + "step": 354 + }, + { + "epoch": 0.56, + "learning_rate": 4.9294532627865965e-05, + "loss": 0.0028, + "step": 355 + }, + { + "epoch": 0.57, + "learning_rate": 4.927689594356261e-05, + "loss": 0.0004, + "step": 356 + }, + { + "epoch": 0.57, + "learning_rate": 4.925925925925926e-05, + "loss": 0.0093, + "step": 357 + }, + { + "epoch": 0.57, + "learning_rate": 4.924162257495591e-05, + "loss": 0.0003, + "step": 358 + }, + { + "epoch": 0.57, + "learning_rate": 4.9223985890652556e-05, + "loss": 0.0427, + "step": 359 + }, + { + "epoch": 0.57, + "learning_rate": 4.9206349206349204e-05, + "loss": 0.0001, + "step": 360 + }, + { + "epoch": 0.57, + "learning_rate": 4.918871252204586e-05, + "loss": 0.0001, + "step": 361 + }, + { + "epoch": 0.57, + "learning_rate": 4.9171075837742506e-05, + "loss": 0.0001, + "step": 362 + }, + { + "epoch": 0.58, + "learning_rate": 4.9153439153439154e-05, + "loss": 0.0001, + "step": 363 + }, + { + "epoch": 0.58, + "learning_rate": 4.913580246913581e-05, + "loss": 0.0001, + "step": 364 + }, + { + "epoch": 0.58, + "learning_rate": 4.9118165784832456e-05, + "loss": 0.0001, + "step": 365 + }, + { + "epoch": 0.58, + "learning_rate": 4.9100529100529104e-05, + "loss": 0.0001, + "step": 366 + }, + { + "epoch": 0.58, + "learning_rate": 4.908289241622575e-05, + "loss": 0.0001, + "step": 367 + }, + { + "epoch": 0.58, + "learning_rate": 4.90652557319224e-05, + "loss": 0.0001, + "step": 368 + }, + { + "epoch": 0.59, + "learning_rate": 4.904761904761905e-05, + "loss": 0.0001, + "step": 369 + }, + { + "epoch": 0.59, + "learning_rate": 4.9029982363315695e-05, + "loss": 0.0001, + "step": 370 + }, + { + "epoch": 0.59, + "learning_rate": 4.901234567901235e-05, + "loss": 0.0165, + "step": 371 + }, + { + "epoch": 0.59, + "learning_rate": 4.8994708994709e-05, + "loss": 0.0001, + "step": 372 + }, + { + "epoch": 0.59, + "learning_rate": 4.8977072310405645e-05, + "loss": 0.0001, + "step": 373 + }, + { + "epoch": 0.59, + "learning_rate": 4.89594356261023e-05, + "loss": 0.0001, + "step": 374 + }, + { + "epoch": 0.6, + "learning_rate": 4.894179894179895e-05, + "loss": 0.0001, + "step": 375 + }, + { + "epoch": 0.6, + "learning_rate": 4.8924162257495594e-05, + "loss": 0.0001, + "step": 376 + }, + { + "epoch": 0.6, + "learning_rate": 4.890652557319224e-05, + "loss": 0.0002, + "step": 377 + }, + { + "epoch": 0.6, + "learning_rate": 4.888888888888889e-05, + "loss": 0.0001, + "step": 378 + }, + { + "epoch": 0.6, + "learning_rate": 4.887125220458554e-05, + "loss": 0.0001, + "step": 379 + }, + { + "epoch": 0.6, + "learning_rate": 4.8853615520282185e-05, + "loss": 0.0001, + "step": 380 + }, + { + "epoch": 0.6, + "learning_rate": 4.883597883597884e-05, + "loss": 0.0003, + "step": 381 + }, + { + "epoch": 0.61, + "learning_rate": 4.881834215167549e-05, + "loss": 0.0001, + "step": 382 + }, + { + "epoch": 0.61, + "learning_rate": 4.8800705467372135e-05, + "loss": 0.0001, + "step": 383 + }, + { + "epoch": 0.61, + "learning_rate": 4.878306878306878e-05, + "loss": 0.0001, + "step": 384 + }, + { + "epoch": 0.61, + "learning_rate": 4.876543209876544e-05, + "loss": 0.0001, + "step": 385 + }, + { + "epoch": 0.61, + "learning_rate": 4.8747795414462085e-05, + "loss": 0.0001, + "step": 386 + }, + { + "epoch": 0.61, + "learning_rate": 4.873015873015873e-05, + "loss": 0.0009, + "step": 387 + }, + { + "epoch": 0.62, + "learning_rate": 4.871252204585538e-05, + "loss": 0.0001, + "step": 388 + }, + { + "epoch": 0.62, + "learning_rate": 4.869488536155203e-05, + "loss": 0.0798, + "step": 389 + }, + { + "epoch": 0.62, + "learning_rate": 4.8677248677248676e-05, + "loss": 0.0002, + "step": 390 + }, + { + "epoch": 0.62, + "learning_rate": 4.865961199294533e-05, + "loss": 0.0001, + "step": 391 + }, + { + "epoch": 0.62, + "learning_rate": 4.864197530864198e-05, + "loss": 0.0002, + "step": 392 + }, + { + "epoch": 0.62, + "learning_rate": 4.8624338624338626e-05, + "loss": 0.0002, + "step": 393 + }, + { + "epoch": 0.63, + "learning_rate": 4.8606701940035274e-05, + "loss": 0.0003, + "step": 394 + }, + { + "epoch": 0.63, + "learning_rate": 4.858906525573193e-05, + "loss": 0.0001, + "step": 395 + }, + { + "epoch": 0.63, + "learning_rate": 4.8571428571428576e-05, + "loss": 0.0028, + "step": 396 + }, + { + "epoch": 0.63, + "learning_rate": 4.8553791887125224e-05, + "loss": 0.0001, + "step": 397 + }, + { + "epoch": 0.63, + "learning_rate": 4.853615520282187e-05, + "loss": 0.0002, + "step": 398 + }, + { + "epoch": 0.63, + "learning_rate": 4.851851851851852e-05, + "loss": 0.0001, + "step": 399 + }, + { + "epoch": 0.63, + "learning_rate": 4.850088183421517e-05, + "loss": 0.0002, + "step": 400 + }, + { + "epoch": 0.64, + "learning_rate": 4.848324514991182e-05, + "loss": 0.0001, + "step": 401 + }, + { + "epoch": 0.64, + "learning_rate": 4.846560846560847e-05, + "loss": 0.0014, + "step": 402 + }, + { + "epoch": 0.64, + "learning_rate": 4.844797178130512e-05, + "loss": 0.0001, + "step": 403 + }, + { + "epoch": 0.64, + "learning_rate": 4.8430335097001764e-05, + "loss": 0.0001, + "step": 404 + }, + { + "epoch": 0.64, + "learning_rate": 4.841269841269841e-05, + "loss": 0.0001, + "step": 405 + }, + { + "epoch": 0.64, + "learning_rate": 4.8395061728395067e-05, + "loss": 0.0001, + "step": 406 + }, + { + "epoch": 0.65, + "learning_rate": 4.8377425044091714e-05, + "loss": 0.0001, + "step": 407 + }, + { + "epoch": 0.65, + "learning_rate": 4.835978835978836e-05, + "loss": 0.0001, + "step": 408 + }, + { + "epoch": 0.65, + "learning_rate": 4.834215167548501e-05, + "loss": 0.1846, + "step": 409 + }, + { + "epoch": 0.65, + "learning_rate": 4.832451499118166e-05, + "loss": 0.0001, + "step": 410 + }, + { + "epoch": 0.65, + "learning_rate": 4.830687830687831e-05, + "loss": 0.0002, + "step": 411 + }, + { + "epoch": 0.65, + "learning_rate": 4.828924162257496e-05, + "loss": 0.0001, + "step": 412 + }, + { + "epoch": 0.66, + "learning_rate": 4.827160493827161e-05, + "loss": 0.0001, + "step": 413 + }, + { + "epoch": 0.66, + "learning_rate": 4.8253968253968255e-05, + "loss": 0.0001, + "step": 414 + }, + { + "epoch": 0.66, + "learning_rate": 4.82363315696649e-05, + "loss": 0.0343, + "step": 415 + }, + { + "epoch": 0.66, + "learning_rate": 4.821869488536155e-05, + "loss": 0.0202, + "step": 416 + }, + { + "epoch": 0.66, + "learning_rate": 4.8201058201058205e-05, + "loss": 0.0001, + "step": 417 + }, + { + "epoch": 0.66, + "learning_rate": 4.818342151675485e-05, + "loss": 0.0001, + "step": 418 + }, + { + "epoch": 0.67, + "learning_rate": 4.81657848324515e-05, + "loss": 0.0001, + "step": 419 + }, + { + "epoch": 0.67, + "learning_rate": 4.814814814814815e-05, + "loss": 0.0001, + "step": 420 + }, + { + "epoch": 0.67, + "learning_rate": 4.81305114638448e-05, + "loss": 0.0001, + "step": 421 + }, + { + "epoch": 0.67, + "learning_rate": 4.811287477954145e-05, + "loss": 0.0001, + "step": 422 + }, + { + "epoch": 0.67, + "learning_rate": 4.80952380952381e-05, + "loss": 0.0001, + "step": 423 + }, + { + "epoch": 0.67, + "learning_rate": 4.8077601410934746e-05, + "loss": 0.0054, + "step": 424 + }, + { + "epoch": 0.67, + "learning_rate": 4.8059964726631394e-05, + "loss": 0.0001, + "step": 425 + }, + { + "epoch": 0.68, + "learning_rate": 4.804232804232804e-05, + "loss": 0.0001, + "step": 426 + }, + { + "epoch": 0.68, + "learning_rate": 4.8024691358024696e-05, + "loss": 0.0001, + "step": 427 + }, + { + "epoch": 0.68, + "learning_rate": 4.8007054673721343e-05, + "loss": 0.0001, + "step": 428 + }, + { + "epoch": 0.68, + "learning_rate": 4.798941798941799e-05, + "loss": 0.0001, + "step": 429 + }, + { + "epoch": 0.68, + "learning_rate": 4.797178130511464e-05, + "loss": 0.0018, + "step": 430 + }, + { + "epoch": 0.68, + "learning_rate": 4.7954144620811293e-05, + "loss": 0.0001, + "step": 431 + }, + { + "epoch": 0.69, + "learning_rate": 4.793650793650794e-05, + "loss": 0.2972, + "step": 432 + }, + { + "epoch": 0.69, + "learning_rate": 4.791887125220459e-05, + "loss": 0.0004, + "step": 433 + }, + { + "epoch": 0.69, + "learning_rate": 4.7901234567901237e-05, + "loss": 0.002, + "step": 434 + }, + { + "epoch": 0.69, + "learning_rate": 4.7883597883597884e-05, + "loss": 0.0043, + "step": 435 + }, + { + "epoch": 0.69, + "learning_rate": 4.786596119929453e-05, + "loss": 0.0053, + "step": 436 + }, + { + "epoch": 0.69, + "learning_rate": 4.784832451499118e-05, + "loss": 0.0035, + "step": 437 + }, + { + "epoch": 0.7, + "learning_rate": 4.7830687830687834e-05, + "loss": 0.0026, + "step": 438 + }, + { + "epoch": 0.7, + "learning_rate": 4.781305114638448e-05, + "loss": 0.0019, + "step": 439 + }, + { + "epoch": 0.7, + "learning_rate": 4.779541446208113e-05, + "loss": 0.0024, + "step": 440 + }, + { + "epoch": 0.7, + "learning_rate": 4.7777777777777784e-05, + "loss": 0.0012, + "step": 441 + }, + { + "epoch": 0.7, + "learning_rate": 4.776014109347443e-05, + "loss": 0.0009, + "step": 442 + }, + { + "epoch": 0.7, + "learning_rate": 4.774250440917108e-05, + "loss": 0.0009, + "step": 443 + }, + { + "epoch": 0.7, + "learning_rate": 4.772486772486773e-05, + "loss": 0.0008, + "step": 444 + }, + { + "epoch": 0.71, + "learning_rate": 4.7707231040564375e-05, + "loss": 0.0006, + "step": 445 + }, + { + "epoch": 0.71, + "learning_rate": 4.768959435626102e-05, + "loss": 0.0006, + "step": 446 + }, + { + "epoch": 0.71, + "learning_rate": 4.767195767195767e-05, + "loss": 0.0005, + "step": 447 + }, + { + "epoch": 0.71, + "learning_rate": 4.7654320987654325e-05, + "loss": 0.0005, + "step": 448 + }, + { + "epoch": 0.71, + "learning_rate": 4.763668430335097e-05, + "loss": 0.0005, + "step": 449 + }, + { + "epoch": 0.71, + "learning_rate": 4.761904761904762e-05, + "loss": 0.0006, + "step": 450 + }, + { + "epoch": 0.72, + "learning_rate": 4.7601410934744275e-05, + "loss": 0.0004, + "step": 451 + }, + { + "epoch": 0.72, + "learning_rate": 4.758377425044092e-05, + "loss": 0.0003, + "step": 452 + }, + { + "epoch": 0.72, + "learning_rate": 4.756613756613757e-05, + "loss": 0.0003, + "step": 453 + }, + { + "epoch": 0.72, + "learning_rate": 4.754850088183422e-05, + "loss": 0.0006, + "step": 454 + }, + { + "epoch": 0.72, + "learning_rate": 4.7530864197530866e-05, + "loss": 0.0003, + "step": 455 + }, + { + "epoch": 0.72, + "learning_rate": 4.7513227513227513e-05, + "loss": 0.0006, + "step": 456 + }, + { + "epoch": 0.73, + "learning_rate": 4.749559082892416e-05, + "loss": 0.0002, + "step": 457 + }, + { + "epoch": 0.73, + "learning_rate": 4.747795414462081e-05, + "loss": 0.0002, + "step": 458 + }, + { + "epoch": 0.73, + "learning_rate": 4.746031746031746e-05, + "loss": 0.0002, + "step": 459 + }, + { + "epoch": 0.73, + "learning_rate": 4.744268077601411e-05, + "loss": 0.0002, + "step": 460 + }, + { + "epoch": 0.73, + "learning_rate": 4.7425044091710766e-05, + "loss": 0.0002, + "step": 461 + }, + { + "epoch": 0.73, + "learning_rate": 4.740740740740741e-05, + "loss": 0.0002, + "step": 462 + }, + { + "epoch": 0.73, + "learning_rate": 4.738977072310406e-05, + "loss": 0.0003, + "step": 463 + }, + { + "epoch": 0.74, + "learning_rate": 4.737213403880071e-05, + "loss": 0.0001, + "step": 464 + }, + { + "epoch": 0.74, + "learning_rate": 4.7354497354497356e-05, + "loss": 0.0001, + "step": 465 + }, + { + "epoch": 0.74, + "learning_rate": 4.7336860670194004e-05, + "loss": 0.0001, + "step": 466 + }, + { + "epoch": 0.74, + "learning_rate": 4.731922398589065e-05, + "loss": 0.0001, + "step": 467 + }, + { + "epoch": 0.74, + "learning_rate": 4.73015873015873e-05, + "loss": 0.0001, + "step": 468 + }, + { + "epoch": 0.74, + "learning_rate": 4.7283950617283954e-05, + "loss": 0.0001, + "step": 469 + }, + { + "epoch": 0.75, + "learning_rate": 4.72663139329806e-05, + "loss": 0.0001, + "step": 470 + }, + { + "epoch": 0.75, + "learning_rate": 4.7248677248677256e-05, + "loss": 0.0001, + "step": 471 + }, + { + "epoch": 0.75, + "learning_rate": 4.7231040564373904e-05, + "loss": 0.0001, + "step": 472 + }, + { + "epoch": 0.75, + "learning_rate": 4.721340388007055e-05, + "loss": 0.0001, + "step": 473 + }, + { + "epoch": 0.75, + "learning_rate": 4.71957671957672e-05, + "loss": 0.0001, + "step": 474 + }, + { + "epoch": 0.75, + "learning_rate": 4.717813051146385e-05, + "loss": 0.0001, + "step": 475 + }, + { + "epoch": 0.76, + "learning_rate": 4.7160493827160495e-05, + "loss": 0.0001, + "step": 476 + }, + { + "epoch": 0.76, + "learning_rate": 4.714285714285714e-05, + "loss": 0.0001, + "step": 477 + }, + { + "epoch": 0.76, + "learning_rate": 4.712522045855379e-05, + "loss": 0.0002, + "step": 478 + }, + { + "epoch": 0.76, + "learning_rate": 4.710758377425044e-05, + "loss": 0.0001, + "step": 479 + }, + { + "epoch": 0.76, + "learning_rate": 4.708994708994709e-05, + "loss": 0.2853, + "step": 480 + }, + { + "epoch": 0.76, + "learning_rate": 4.707231040564374e-05, + "loss": 0.0002, + "step": 481 + }, + { + "epoch": 0.77, + "learning_rate": 4.7054673721340395e-05, + "loss": 0.0008, + "step": 482 + }, + { + "epoch": 0.77, + "learning_rate": 4.703703703703704e-05, + "loss": 0.0006, + "step": 483 + }, + { + "epoch": 0.77, + "learning_rate": 4.701940035273369e-05, + "loss": 0.0009, + "step": 484 + }, + { + "epoch": 0.77, + "learning_rate": 4.700176366843034e-05, + "loss": 0.0014, + "step": 485 + }, + { + "epoch": 0.77, + "learning_rate": 4.6984126984126986e-05, + "loss": 0.1527, + "step": 486 + }, + { + "epoch": 0.77, + "learning_rate": 4.696649029982363e-05, + "loss": 0.003, + "step": 487 + }, + { + "epoch": 0.77, + "learning_rate": 4.694885361552028e-05, + "loss": 0.004, + "step": 488 + }, + { + "epoch": 0.78, + "learning_rate": 4.693121693121693e-05, + "loss": 0.0046, + "step": 489 + }, + { + "epoch": 0.78, + "learning_rate": 4.691358024691358e-05, + "loss": 0.0033, + "step": 490 + }, + { + "epoch": 0.78, + "learning_rate": 4.689594356261023e-05, + "loss": 0.0022, + "step": 491 + }, + { + "epoch": 0.78, + "learning_rate": 4.6878306878306885e-05, + "loss": 0.0016, + "step": 492 + }, + { + "epoch": 0.78, + "learning_rate": 4.686067019400353e-05, + "loss": 0.0084, + "step": 493 + }, + { + "epoch": 0.78, + "learning_rate": 4.684303350970018e-05, + "loss": 0.0014, + "step": 494 + }, + { + "epoch": 0.79, + "learning_rate": 4.682539682539683e-05, + "loss": 0.0011, + "step": 495 + }, + { + "epoch": 0.79, + "learning_rate": 4.6807760141093476e-05, + "loss": 0.001, + "step": 496 + }, + { + "epoch": 0.79, + "learning_rate": 4.6790123456790124e-05, + "loss": 0.0113, + "step": 497 + }, + { + "epoch": 0.79, + "learning_rate": 4.677248677248677e-05, + "loss": 0.0009, + "step": 498 + }, + { + "epoch": 0.79, + "learning_rate": 4.675485008818342e-05, + "loss": 0.0008, + "step": 499 + }, + { + "epoch": 0.79, + "learning_rate": 4.673721340388007e-05, + "loss": 0.0008, + "step": 500 + }, + { + "epoch": 0.8, + "learning_rate": 4.671957671957672e-05, + "loss": 0.0007, + "step": 501 + }, + { + "epoch": 0.8, + "learning_rate": 4.670194003527337e-05, + "loss": 0.0007, + "step": 502 + }, + { + "epoch": 0.8, + "learning_rate": 4.6684303350970024e-05, + "loss": 0.0008, + "step": 503 + }, + { + "epoch": 0.8, + "learning_rate": 4.666666666666667e-05, + "loss": 0.0006, + "step": 504 + }, + { + "epoch": 0.8, + "learning_rate": 4.664902998236332e-05, + "loss": 0.0005, + "step": 505 + }, + { + "epoch": 0.8, + "learning_rate": 4.663139329805997e-05, + "loss": 0.0005, + "step": 506 + }, + { + "epoch": 0.8, + "learning_rate": 4.6613756613756615e-05, + "loss": 0.0006, + "step": 507 + }, + { + "epoch": 0.81, + "learning_rate": 4.659611992945326e-05, + "loss": 0.0004, + "step": 508 + }, + { + "epoch": 0.81, + "learning_rate": 4.657848324514991e-05, + "loss": 0.0004, + "step": 509 + }, + { + "epoch": 0.81, + "learning_rate": 4.656084656084656e-05, + "loss": 0.0004, + "step": 510 + }, + { + "epoch": 0.81, + "learning_rate": 4.654320987654321e-05, + "loss": 0.0005, + "step": 511 + }, + { + "epoch": 0.81, + "learning_rate": 4.652557319223986e-05, + "loss": 0.0003, + "step": 512 + }, + { + "epoch": 0.81, + "learning_rate": 4.6507936507936515e-05, + "loss": 0.0003, + "step": 513 + }, + { + "epoch": 0.82, + "learning_rate": 4.649029982363316e-05, + "loss": 0.0003, + "step": 514 + }, + { + "epoch": 0.82, + "learning_rate": 4.647266313932981e-05, + "loss": 0.0003, + "step": 515 + }, + { + "epoch": 0.82, + "learning_rate": 4.645502645502646e-05, + "loss": 0.0005, + "step": 516 + }, + { + "epoch": 0.82, + "learning_rate": 4.6437389770723105e-05, + "loss": 0.0003, + "step": 517 + }, + { + "epoch": 0.82, + "learning_rate": 4.641975308641975e-05, + "loss": 0.0003, + "step": 518 + }, + { + "epoch": 0.82, + "learning_rate": 4.64021164021164e-05, + "loss": 0.0003, + "step": 519 + }, + { + "epoch": 0.83, + "learning_rate": 4.638447971781305e-05, + "loss": 0.0002, + "step": 520 + }, + { + "epoch": 0.83, + "learning_rate": 4.63668430335097e-05, + "loss": 0.0002, + "step": 521 + }, + { + "epoch": 0.83, + "learning_rate": 4.634920634920635e-05, + "loss": 0.0002, + "step": 522 + }, + { + "epoch": 0.83, + "learning_rate": 4.6331569664903e-05, + "loss": 0.0002, + "step": 523 + }, + { + "epoch": 0.83, + "learning_rate": 4.631393298059965e-05, + "loss": 0.0002, + "step": 524 + }, + { + "epoch": 0.83, + "learning_rate": 4.62962962962963e-05, + "loss": 0.0002, + "step": 525 + }, + { + "epoch": 0.83, + "learning_rate": 4.627865961199295e-05, + "loss": 0.0002, + "step": 526 + }, + { + "epoch": 0.84, + "learning_rate": 4.6261022927689596e-05, + "loss": 0.0003, + "step": 527 + }, + { + "epoch": 0.84, + "learning_rate": 4.6243386243386244e-05, + "loss": 0.0002, + "step": 528 + }, + { + "epoch": 0.84, + "learning_rate": 4.622574955908289e-05, + "loss": 0.0002, + "step": 529 + }, + { + "epoch": 0.84, + "learning_rate": 4.620811287477954e-05, + "loss": 0.0002, + "step": 530 + }, + { + "epoch": 0.84, + "learning_rate": 4.6190476190476194e-05, + "loss": 0.0002, + "step": 531 + }, + { + "epoch": 0.84, + "learning_rate": 4.617283950617284e-05, + "loss": 0.1655, + "step": 532 + }, + { + "epoch": 0.85, + "learning_rate": 4.615520282186949e-05, + "loss": 0.0002, + "step": 533 + }, + { + "epoch": 0.85, + "learning_rate": 4.6137566137566144e-05, + "loss": 0.0007, + "step": 534 + }, + { + "epoch": 0.85, + "learning_rate": 4.611992945326279e-05, + "loss": 0.0002, + "step": 535 + }, + { + "epoch": 0.85, + "learning_rate": 4.610229276895944e-05, + "loss": 0.0002, + "step": 536 + }, + { + "epoch": 0.85, + "learning_rate": 4.608465608465609e-05, + "loss": 0.0002, + "step": 537 + }, + { + "epoch": 0.85, + "learning_rate": 4.6067019400352735e-05, + "loss": 0.0002, + "step": 538 + }, + { + "epoch": 0.86, + "learning_rate": 4.604938271604938e-05, + "loss": 0.0002, + "step": 539 + }, + { + "epoch": 0.86, + "learning_rate": 4.603174603174603e-05, + "loss": 0.0002, + "step": 540 + }, + { + "epoch": 0.86, + "learning_rate": 4.6014109347442685e-05, + "loss": 0.0002, + "step": 541 + }, + { + "epoch": 0.86, + "learning_rate": 4.599647266313933e-05, + "loss": 0.0002, + "step": 542 + }, + { + "epoch": 0.86, + "learning_rate": 4.597883597883598e-05, + "loss": 0.0002, + "step": 543 + }, + { + "epoch": 0.86, + "learning_rate": 4.596119929453263e-05, + "loss": 0.0002, + "step": 544 + }, + { + "epoch": 0.87, + "learning_rate": 4.594356261022928e-05, + "loss": 0.0002, + "step": 545 + }, + { + "epoch": 0.87, + "learning_rate": 4.592592592592593e-05, + "loss": 0.0041, + "step": 546 + }, + { + "epoch": 0.87, + "learning_rate": 4.590828924162258e-05, + "loss": 0.0002, + "step": 547 + }, + { + "epoch": 0.87, + "learning_rate": 4.5890652557319225e-05, + "loss": 0.0002, + "step": 548 + }, + { + "epoch": 0.87, + "learning_rate": 4.587301587301587e-05, + "loss": 0.0002, + "step": 549 + }, + { + "epoch": 0.87, + "learning_rate": 4.585537918871252e-05, + "loss": 0.0002, + "step": 550 + }, + { + "epoch": 0.87, + "learning_rate": 4.5837742504409175e-05, + "loss": 0.0002, + "step": 551 + }, + { + "epoch": 0.88, + "learning_rate": 4.582010582010582e-05, + "loss": 0.0002, + "step": 552 + }, + { + "epoch": 0.88, + "learning_rate": 4.580246913580247e-05, + "loss": 0.0002, + "step": 553 + }, + { + "epoch": 0.88, + "learning_rate": 4.578483245149912e-05, + "loss": 0.0002, + "step": 554 + }, + { + "epoch": 0.88, + "learning_rate": 4.576719576719577e-05, + "loss": 0.0002, + "step": 555 + }, + { + "epoch": 0.88, + "learning_rate": 4.574955908289242e-05, + "loss": 0.0002, + "step": 556 + }, + { + "epoch": 0.88, + "learning_rate": 4.573192239858907e-05, + "loss": 0.0594, + "step": 557 + }, + { + "epoch": 0.89, + "learning_rate": 4.5714285714285716e-05, + "loss": 0.236, + "step": 558 + }, + { + "epoch": 0.89, + "learning_rate": 4.5696649029982364e-05, + "loss": 0.0002, + "step": 559 + }, + { + "epoch": 0.89, + "learning_rate": 4.567901234567901e-05, + "loss": 0.0068, + "step": 560 + }, + { + "epoch": 0.89, + "learning_rate": 4.5661375661375666e-05, + "loss": 0.0002, + "step": 561 + }, + { + "epoch": 0.89, + "learning_rate": 4.5643738977072314e-05, + "loss": 0.0003, + "step": 562 + }, + { + "epoch": 0.89, + "learning_rate": 4.562610229276896e-05, + "loss": 0.0004, + "step": 563 + }, + { + "epoch": 0.9, + "learning_rate": 4.560846560846561e-05, + "loss": 0.0003, + "step": 564 + }, + { + "epoch": 0.9, + "learning_rate": 4.559082892416226e-05, + "loss": 0.0129, + "step": 565 + }, + { + "epoch": 0.9, + "learning_rate": 4.557319223985891e-05, + "loss": 0.0004, + "step": 566 + }, + { + "epoch": 0.9, + "learning_rate": 4.555555555555556e-05, + "loss": 0.0004, + "step": 567 + }, + { + "epoch": 0.9, + "learning_rate": 4.553791887125221e-05, + "loss": 0.0004, + "step": 568 + }, + { + "epoch": 0.9, + "learning_rate": 4.5520282186948855e-05, + "loss": 0.0005, + "step": 569 + }, + { + "epoch": 0.9, + "learning_rate": 4.55026455026455e-05, + "loss": 0.0005, + "step": 570 + }, + { + "epoch": 0.91, + "learning_rate": 4.548500881834216e-05, + "loss": 0.0005, + "step": 571 + }, + { + "epoch": 0.91, + "learning_rate": 4.5467372134038804e-05, + "loss": 0.0004, + "step": 572 + }, + { + "epoch": 0.91, + "learning_rate": 4.544973544973545e-05, + "loss": 0.0003, + "step": 573 + }, + { + "epoch": 0.91, + "learning_rate": 4.54320987654321e-05, + "loss": 0.0003, + "step": 574 + }, + { + "epoch": 0.91, + "learning_rate": 4.541446208112875e-05, + "loss": 0.0046, + "step": 575 + }, + { + "epoch": 0.91, + "learning_rate": 4.5396825396825395e-05, + "loss": 0.0004, + "step": 576 + }, + { + "epoch": 0.92, + "learning_rate": 4.537918871252205e-05, + "loss": 0.0606, + "step": 577 + }, + { + "epoch": 0.92, + "learning_rate": 4.53615520282187e-05, + "loss": 0.0003, + "step": 578 + }, + { + "epoch": 0.92, + "learning_rate": 4.5343915343915345e-05, + "loss": 0.0002, + "step": 579 + }, + { + "epoch": 0.92, + "learning_rate": 4.532627865961199e-05, + "loss": 0.0002, + "step": 580 + }, + { + "epoch": 0.92, + "learning_rate": 4.530864197530865e-05, + "loss": 0.0073, + "step": 581 + }, + { + "epoch": 0.92, + "learning_rate": 4.5291005291005295e-05, + "loss": 0.0003, + "step": 582 + }, + { + "epoch": 0.93, + "learning_rate": 4.527336860670194e-05, + "loss": 0.0002, + "step": 583 + }, + { + "epoch": 0.93, + "learning_rate": 4.525573192239859e-05, + "loss": 0.0002, + "step": 584 + }, + { + "epoch": 0.93, + "learning_rate": 4.523809523809524e-05, + "loss": 0.0002, + "step": 585 + }, + { + "epoch": 0.93, + "learning_rate": 4.5220458553791886e-05, + "loss": 0.0002, + "step": 586 + }, + { + "epoch": 0.93, + "learning_rate": 4.520282186948854e-05, + "loss": 0.0079, + "step": 587 + }, + { + "epoch": 0.93, + "learning_rate": 4.518518518518519e-05, + "loss": 0.0051, + "step": 588 + }, + { + "epoch": 0.93, + "learning_rate": 4.5167548500881836e-05, + "loss": 0.0002, + "step": 589 + }, + { + "epoch": 0.94, + "learning_rate": 4.5149911816578484e-05, + "loss": 0.0045, + "step": 590 + }, + { + "epoch": 0.94, + "learning_rate": 4.513227513227514e-05, + "loss": 0.0002, + "step": 591 + }, + { + "epoch": 0.94, + "learning_rate": 4.5114638447971786e-05, + "loss": 0.0055, + "step": 592 + }, + { + "epoch": 0.94, + "learning_rate": 4.5097001763668434e-05, + "loss": 0.0002, + "step": 593 + }, + { + "epoch": 0.94, + "learning_rate": 4.507936507936508e-05, + "loss": 0.0002, + "step": 594 + }, + { + "epoch": 0.94, + "learning_rate": 4.506172839506173e-05, + "loss": 0.0014, + "step": 595 + }, + { + "epoch": 0.95, + "learning_rate": 4.504409171075838e-05, + "loss": 0.0001, + "step": 596 + }, + { + "epoch": 0.95, + "learning_rate": 4.5026455026455024e-05, + "loss": 0.0001, + "step": 597 + }, + { + "epoch": 0.95, + "learning_rate": 4.500881834215168e-05, + "loss": 0.0001, + "step": 598 + }, + { + "epoch": 0.95, + "learning_rate": 4.499118165784833e-05, + "loss": 0.0001, + "step": 599 + }, + { + "epoch": 0.95, + "learning_rate": 4.4973544973544974e-05, + "loss": 0.0001, + "step": 600 + }, + { + "epoch": 0.95, + "learning_rate": 4.495590828924163e-05, + "loss": 0.0003, + "step": 601 + }, + { + "epoch": 0.96, + "learning_rate": 4.493827160493828e-05, + "loss": 0.0001, + "step": 602 + }, + { + "epoch": 0.96, + "learning_rate": 4.4920634920634924e-05, + "loss": 0.0001, + "step": 603 + }, + { + "epoch": 0.96, + "learning_rate": 4.490299823633157e-05, + "loss": 0.0005, + "step": 604 + }, + { + "epoch": 0.96, + "learning_rate": 4.488536155202822e-05, + "loss": 0.0001, + "step": 605 + }, + { + "epoch": 0.96, + "learning_rate": 4.486772486772487e-05, + "loss": 0.0001, + "step": 606 + }, + { + "epoch": 0.96, + "learning_rate": 4.4850088183421515e-05, + "loss": 0.0001, + "step": 607 + }, + { + "epoch": 0.97, + "learning_rate": 4.483245149911817e-05, + "loss": 0.0243, + "step": 608 + }, + { + "epoch": 0.97, + "learning_rate": 4.481481481481482e-05, + "loss": 0.0001, + "step": 609 + }, + { + "epoch": 0.97, + "learning_rate": 4.4797178130511465e-05, + "loss": 0.0004, + "step": 610 + }, + { + "epoch": 0.97, + "learning_rate": 4.477954144620812e-05, + "loss": 0.0001, + "step": 611 + }, + { + "epoch": 0.97, + "learning_rate": 4.476190476190477e-05, + "loss": 0.0001, + "step": 612 + }, + { + "epoch": 0.97, + "learning_rate": 4.4744268077601415e-05, + "loss": 0.0001, + "step": 613 + }, + { + "epoch": 0.97, + "learning_rate": 4.472663139329806e-05, + "loss": 0.0001, + "step": 614 + }, + { + "epoch": 0.98, + "learning_rate": 4.470899470899471e-05, + "loss": 0.0001, + "step": 615 + }, + { + "epoch": 0.98, + "learning_rate": 4.469135802469136e-05, + "loss": 0.0001, + "step": 616 + }, + { + "epoch": 0.98, + "learning_rate": 4.4673721340388006e-05, + "loss": 0.0001, + "step": 617 + }, + { + "epoch": 0.98, + "learning_rate": 4.4656084656084654e-05, + "loss": 0.001, + "step": 618 + }, + { + "epoch": 0.98, + "learning_rate": 4.463844797178131e-05, + "loss": 0.0001, + "step": 619 + }, + { + "epoch": 0.98, + "learning_rate": 4.4620811287477956e-05, + "loss": 0.0001, + "step": 620 + }, + { + "epoch": 0.99, + "learning_rate": 4.460317460317461e-05, + "loss": 0.0001, + "step": 621 + }, + { + "epoch": 0.99, + "learning_rate": 4.458553791887126e-05, + "loss": 0.0001, + "step": 622 + }, + { + "epoch": 0.99, + "learning_rate": 4.4567901234567906e-05, + "loss": 0.0001, + "step": 623 + }, + { + "epoch": 0.99, + "learning_rate": 4.4550264550264553e-05, + "loss": 0.0001, + "step": 624 + }, + { + "epoch": 0.99, + "learning_rate": 4.45326278659612e-05, + "loss": 0.0001, + "step": 625 + }, + { + "epoch": 0.99, + "learning_rate": 4.451499118165785e-05, + "loss": 0.0001, + "step": 626 + }, + { + "epoch": 1.0, + "learning_rate": 4.44973544973545e-05, + "loss": 0.0001, + "step": 627 + }, + { + "epoch": 1.0, + "learning_rate": 4.4479717813051144e-05, + "loss": 0.0003, + "step": 628 + }, + { + "epoch": 1.0, + "learning_rate": 4.44620811287478e-05, + "loss": 0.0001, + "step": 629 + }, + { + "epoch": 1.0, + "learning_rate": 4.4444444444444447e-05, + "loss": 0.0001, + "step": 630 + }, + { + "epoch": 1.0, + "eval_accuracy": 0.9473684210526315, + "eval_auc": 0.9958333333333333, + "eval_f1": 0.8571428571428571, + "eval_loss": 0.134189635515213, + "eval_precision": 1.0, + "eval_recall": 0.75, + "eval_runtime": 0.2025, + "eval_samples_per_second": 187.654, + "eval_steps_per_second": 4.938, + "step": 630 + }, + { + "epoch": 1.0, + "learning_rate": 4.44268077601411e-05, + "loss": 0.0001, + "step": 631 + }, + { + "epoch": 1.0, + "learning_rate": 4.440917107583775e-05, + "loss": 0.0001, + "step": 632 + }, + { + "epoch": 1.0, + "learning_rate": 4.4391534391534397e-05, + "loss": 0.1211, + "step": 633 + }, + { + "epoch": 1.01, + "learning_rate": 4.4373897707231044e-05, + "loss": 0.0001, + "step": 634 + }, + { + "epoch": 1.01, + "learning_rate": 4.435626102292769e-05, + "loss": 0.0001, + "step": 635 + }, + { + "epoch": 1.01, + "learning_rate": 4.433862433862434e-05, + "loss": 0.0001, + "step": 636 + }, + { + "epoch": 1.01, + "learning_rate": 4.432098765432099e-05, + "loss": 0.0012, + "step": 637 + }, + { + "epoch": 1.01, + "learning_rate": 4.4303350970017635e-05, + "loss": 0.0001, + "step": 638 + }, + { + "epoch": 1.01, + "learning_rate": 4.428571428571428e-05, + "loss": 0.0004, + "step": 639 + }, + { + "epoch": 1.02, + "learning_rate": 4.426807760141094e-05, + "loss": 0.0001, + "step": 640 + }, + { + "epoch": 1.02, + "learning_rate": 4.425044091710759e-05, + "loss": 0.0001, + "step": 641 + }, + { + "epoch": 1.02, + "learning_rate": 4.423280423280424e-05, + "loss": 0.0001, + "step": 642 + }, + { + "epoch": 1.02, + "learning_rate": 4.421516754850089e-05, + "loss": 0.0001, + "step": 643 + }, + { + "epoch": 1.02, + "learning_rate": 4.4197530864197535e-05, + "loss": 0.0005, + "step": 644 + }, + { + "epoch": 1.02, + "learning_rate": 4.417989417989418e-05, + "loss": 0.0001, + "step": 645 + }, + { + "epoch": 1.03, + "learning_rate": 4.416225749559083e-05, + "loss": 0.0001, + "step": 646 + }, + { + "epoch": 1.03, + "learning_rate": 4.414462081128748e-05, + "loss": 0.212, + "step": 647 + }, + { + "epoch": 1.03, + "learning_rate": 4.4126984126984126e-05, + "loss": 0.0001, + "step": 648 + }, + { + "epoch": 1.03, + "learning_rate": 4.4109347442680774e-05, + "loss": 0.0005, + "step": 649 + }, + { + "epoch": 1.03, + "learning_rate": 4.409171075837743e-05, + "loss": 0.0001, + "step": 650 + }, + { + "epoch": 1.03, + "learning_rate": 4.4074074074074076e-05, + "loss": 0.0001, + "step": 651 + }, + { + "epoch": 1.03, + "learning_rate": 4.405643738977073e-05, + "loss": 0.0001, + "step": 652 + }, + { + "epoch": 1.04, + "learning_rate": 4.403880070546738e-05, + "loss": 0.0001, + "step": 653 + }, + { + "epoch": 1.04, + "learning_rate": 4.4021164021164026e-05, + "loss": 0.0002, + "step": 654 + }, + { + "epoch": 1.04, + "learning_rate": 4.400352733686067e-05, + "loss": 0.0002, + "step": 655 + }, + { + "epoch": 1.04, + "learning_rate": 4.398589065255732e-05, + "loss": 0.0002, + "step": 656 + }, + { + "epoch": 1.04, + "learning_rate": 4.396825396825397e-05, + "loss": 0.0019, + "step": 657 + }, + { + "epoch": 1.04, + "learning_rate": 4.3950617283950617e-05, + "loss": 0.0003, + "step": 658 + }, + { + "epoch": 1.05, + "learning_rate": 4.3932980599647264e-05, + "loss": 0.0002, + "step": 659 + }, + { + "epoch": 1.05, + "learning_rate": 4.391534391534391e-05, + "loss": 0.0003, + "step": 660 + }, + { + "epoch": 1.05, + "learning_rate": 4.3897707231040566e-05, + "loss": 0.0003, + "step": 661 + }, + { + "epoch": 1.05, + "learning_rate": 4.3880070546737214e-05, + "loss": 0.0003, + "step": 662 + }, + { + "epoch": 1.05, + "learning_rate": 4.386243386243387e-05, + "loss": 0.0003, + "step": 663 + }, + { + "epoch": 1.05, + "learning_rate": 4.3844797178130516e-05, + "loss": 0.0003, + "step": 664 + }, + { + "epoch": 1.06, + "learning_rate": 4.3827160493827164e-05, + "loss": 0.0005, + "step": 665 + }, + { + "epoch": 1.06, + "learning_rate": 4.380952380952381e-05, + "loss": 0.0027, + "step": 666 + }, + { + "epoch": 1.06, + "learning_rate": 4.379188712522046e-05, + "loss": 0.0002, + "step": 667 + }, + { + "epoch": 1.06, + "learning_rate": 4.377425044091711e-05, + "loss": 0.0002, + "step": 668 + }, + { + "epoch": 1.06, + "learning_rate": 4.3756613756613755e-05, + "loss": 0.0007, + "step": 669 + }, + { + "epoch": 1.06, + "learning_rate": 4.37389770723104e-05, + "loss": 0.0001, + "step": 670 + }, + { + "epoch": 1.07, + "learning_rate": 4.372134038800706e-05, + "loss": 0.0002, + "step": 671 + }, + { + "epoch": 1.07, + "learning_rate": 4.3703703703703705e-05, + "loss": 0.0001, + "step": 672 + }, + { + "epoch": 1.07, + "learning_rate": 4.368606701940036e-05, + "loss": 0.0008, + "step": 673 + }, + { + "epoch": 1.07, + "learning_rate": 4.366843033509701e-05, + "loss": 0.0001, + "step": 674 + }, + { + "epoch": 1.07, + "learning_rate": 4.3650793650793655e-05, + "loss": 0.0001, + "step": 675 + }, + { + "epoch": 1.07, + "learning_rate": 4.36331569664903e-05, + "loss": 0.0001, + "step": 676 + }, + { + "epoch": 1.07, + "learning_rate": 4.361552028218695e-05, + "loss": 0.0001, + "step": 677 + }, + { + "epoch": 1.08, + "learning_rate": 4.35978835978836e-05, + "loss": 0.0001, + "step": 678 + }, + { + "epoch": 1.08, + "learning_rate": 4.3580246913580246e-05, + "loss": 0.0005, + "step": 679 + }, + { + "epoch": 1.08, + "learning_rate": 4.3562610229276893e-05, + "loss": 0.0001, + "step": 680 + }, + { + "epoch": 1.08, + "learning_rate": 4.354497354497355e-05, + "loss": 0.0001, + "step": 681 + }, + { + "epoch": 1.08, + "learning_rate": 4.3527336860670196e-05, + "loss": 0.0001, + "step": 682 + }, + { + "epoch": 1.08, + "learning_rate": 4.350970017636684e-05, + "loss": 0.0001, + "step": 683 + }, + { + "epoch": 1.09, + "learning_rate": 4.34920634920635e-05, + "loss": 0.0001, + "step": 684 + }, + { + "epoch": 1.09, + "learning_rate": 4.3474426807760146e-05, + "loss": 0.0001, + "step": 685 + }, + { + "epoch": 1.09, + "learning_rate": 4.345679012345679e-05, + "loss": 0.0001, + "step": 686 + }, + { + "epoch": 1.09, + "learning_rate": 4.343915343915344e-05, + "loss": 0.0003, + "step": 687 + }, + { + "epoch": 1.09, + "learning_rate": 4.342151675485009e-05, + "loss": 0.0001, + "step": 688 + }, + { + "epoch": 1.09, + "learning_rate": 4.3403880070546736e-05, + "loss": 0.0004, + "step": 689 + }, + { + "epoch": 1.1, + "learning_rate": 4.3386243386243384e-05, + "loss": 0.0001, + "step": 690 + }, + { + "epoch": 1.1, + "learning_rate": 4.336860670194004e-05, + "loss": 0.0001, + "step": 691 + }, + { + "epoch": 1.1, + "learning_rate": 4.3350970017636686e-05, + "loss": 0.0001, + "step": 692 + }, + { + "epoch": 1.1, + "learning_rate": 4.3333333333333334e-05, + "loss": 0.0001, + "step": 693 + }, + { + "epoch": 1.1, + "learning_rate": 4.331569664902999e-05, + "loss": 0.0005, + "step": 694 + }, + { + "epoch": 1.1, + "learning_rate": 4.3298059964726636e-05, + "loss": 0.0001, + "step": 695 + }, + { + "epoch": 1.1, + "learning_rate": 4.3280423280423284e-05, + "loss": 0.0001, + "step": 696 + }, + { + "epoch": 1.11, + "learning_rate": 4.326278659611993e-05, + "loss": 0.0003, + "step": 697 + }, + { + "epoch": 1.11, + "learning_rate": 4.324514991181658e-05, + "loss": 0.0123, + "step": 698 + }, + { + "epoch": 1.11, + "learning_rate": 4.322751322751323e-05, + "loss": 0.0001, + "step": 699 + }, + { + "epoch": 1.11, + "learning_rate": 4.3209876543209875e-05, + "loss": 0.0001, + "step": 700 + }, + { + "epoch": 1.11, + "learning_rate": 4.319223985890653e-05, + "loss": 0.0001, + "step": 701 + }, + { + "epoch": 1.11, + "learning_rate": 4.317460317460318e-05, + "loss": 0.0011, + "step": 702 + }, + { + "epoch": 1.12, + "learning_rate": 4.3156966490299825e-05, + "loss": 0.0001, + "step": 703 + }, + { + "epoch": 1.12, + "learning_rate": 4.313932980599647e-05, + "loss": 0.0001, + "step": 704 + }, + { + "epoch": 1.12, + "learning_rate": 4.312169312169313e-05, + "loss": 0.0001, + "step": 705 + }, + { + "epoch": 1.12, + "learning_rate": 4.3104056437389775e-05, + "loss": 0.0001, + "step": 706 + }, + { + "epoch": 1.12, + "learning_rate": 4.308641975308642e-05, + "loss": 0.0001, + "step": 707 + }, + { + "epoch": 1.12, + "learning_rate": 4.306878306878307e-05, + "loss": 0.0001, + "step": 708 + }, + { + "epoch": 1.13, + "learning_rate": 4.305114638447972e-05, + "loss": 0.0001, + "step": 709 + }, + { + "epoch": 1.13, + "learning_rate": 4.3033509700176366e-05, + "loss": 0.0001, + "step": 710 + }, + { + "epoch": 1.13, + "learning_rate": 4.301587301587302e-05, + "loss": 0.0001, + "step": 711 + }, + { + "epoch": 1.13, + "learning_rate": 4.299823633156967e-05, + "loss": 0.0002, + "step": 712 + }, + { + "epoch": 1.13, + "learning_rate": 4.2980599647266315e-05, + "loss": 0.0001, + "step": 713 + }, + { + "epoch": 1.13, + "learning_rate": 4.296296296296296e-05, + "loss": 0.0874, + "step": 714 + }, + { + "epoch": 1.13, + "learning_rate": 4.294532627865962e-05, + "loss": 0.0001, + "step": 715 + }, + { + "epoch": 1.14, + "learning_rate": 4.2927689594356265e-05, + "loss": 0.1328, + "step": 716 + }, + { + "epoch": 1.14, + "learning_rate": 4.291005291005291e-05, + "loss": 0.0001, + "step": 717 + }, + { + "epoch": 1.14, + "learning_rate": 4.289241622574956e-05, + "loss": 0.0002, + "step": 718 + }, + { + "epoch": 1.14, + "learning_rate": 4.287477954144621e-05, + "loss": 0.0003, + "step": 719 + }, + { + "epoch": 1.14, + "learning_rate": 4.2857142857142856e-05, + "loss": 0.0001, + "step": 720 + }, + { + "epoch": 1.14, + "learning_rate": 4.283950617283951e-05, + "loss": 0.0001, + "step": 721 + }, + { + "epoch": 1.15, + "learning_rate": 4.282186948853616e-05, + "loss": 0.0001, + "step": 722 + }, + { + "epoch": 1.15, + "learning_rate": 4.2804232804232806e-05, + "loss": 0.0001, + "step": 723 + }, + { + "epoch": 1.15, + "learning_rate": 4.2786596119929454e-05, + "loss": 0.0001, + "step": 724 + }, + { + "epoch": 1.15, + "learning_rate": 4.27689594356261e-05, + "loss": 0.0001, + "step": 725 + }, + { + "epoch": 1.15, + "learning_rate": 4.2751322751322756e-05, + "loss": 0.0001, + "step": 726 + }, + { + "epoch": 1.15, + "learning_rate": 4.2733686067019404e-05, + "loss": 0.0001, + "step": 727 + }, + { + "epoch": 1.16, + "learning_rate": 4.271604938271605e-05, + "loss": 0.0002, + "step": 728 + }, + { + "epoch": 1.16, + "learning_rate": 4.26984126984127e-05, + "loss": 0.0003, + "step": 729 + }, + { + "epoch": 1.16, + "learning_rate": 4.268077601410935e-05, + "loss": 0.0004, + "step": 730 + }, + { + "epoch": 1.16, + "learning_rate": 4.268077601410935e-05, + "loss": 0.106, + "step": 731 + }, + { + "epoch": 1.16, + "learning_rate": 4.2663139329806e-05, + "loss": 0.0001, + "step": 732 + }, + { + "epoch": 1.16, + "learning_rate": 4.264550264550265e-05, + "loss": 0.0001, + "step": 733 + }, + { + "epoch": 1.17, + "learning_rate": 4.26278659611993e-05, + "loss": 0.0001, + "step": 734 + }, + { + "epoch": 1.17, + "learning_rate": 4.2610229276895945e-05, + "loss": 0.0001, + "step": 735 + }, + { + "epoch": 1.17, + "learning_rate": 4.259259259259259e-05, + "loss": 0.1685, + "step": 736 + }, + { + "epoch": 1.17, + "learning_rate": 4.257495590828925e-05, + "loss": 0.0004, + "step": 737 + }, + { + "epoch": 1.17, + "learning_rate": 4.2557319223985895e-05, + "loss": 0.0058, + "step": 738 + }, + { + "epoch": 1.17, + "learning_rate": 4.253968253968254e-05, + "loss": 0.0003, + "step": 739 + }, + { + "epoch": 1.17, + "learning_rate": 4.252204585537919e-05, + "loss": 0.0017, + "step": 740 + }, + { + "epoch": 1.18, + "learning_rate": 4.250440917107584e-05, + "loss": 0.0001, + "step": 741 + }, + { + "epoch": 1.18, + "learning_rate": 4.248677248677249e-05, + "loss": 0.0063, + "step": 742 + }, + { + "epoch": 1.18, + "learning_rate": 4.246913580246914e-05, + "loss": 0.0025, + "step": 743 + }, + { + "epoch": 1.18, + "learning_rate": 4.245149911816579e-05, + "loss": 0.0001, + "step": 744 + }, + { + "epoch": 1.18, + "learning_rate": 4.2433862433862435e-05, + "loss": 0.0001, + "step": 745 + }, + { + "epoch": 1.18, + "learning_rate": 4.241622574955908e-05, + "loss": 0.0007, + "step": 746 + }, + { + "epoch": 1.19, + "learning_rate": 4.239858906525573e-05, + "loss": 0.0001, + "step": 747 + }, + { + "epoch": 1.19, + "learning_rate": 4.2380952380952385e-05, + "loss": 0.0001, + "step": 748 + }, + { + "epoch": 1.19, + "learning_rate": 4.236331569664903e-05, + "loss": 0.0001, + "step": 749 + }, + { + "epoch": 1.19, + "learning_rate": 4.234567901234568e-05, + "loss": 0.0001, + "step": 750 + }, + { + "epoch": 1.19, + "learning_rate": 4.232804232804233e-05, + "loss": 0.0001, + "step": 751 + }, + { + "epoch": 1.19, + "learning_rate": 4.231040564373898e-05, + "loss": 0.0001, + "step": 752 + }, + { + "epoch": 1.2, + "learning_rate": 4.229276895943563e-05, + "loss": 0.0009, + "step": 753 + }, + { + "epoch": 1.2, + "learning_rate": 4.227513227513228e-05, + "loss": 0.0001, + "step": 754 + }, + { + "epoch": 1.2, + "learning_rate": 4.2257495590828926e-05, + "loss": 0.0001, + "step": 755 + }, + { + "epoch": 1.2, + "learning_rate": 4.2239858906525574e-05, + "loss": 0.0001, + "step": 756 + }, + { + "epoch": 1.2, + "learning_rate": 4.222222222222222e-05, + "loss": 0.0001, + "step": 757 + }, + { + "epoch": 1.2, + "learning_rate": 4.220458553791887e-05, + "loss": 0.0001, + "step": 758 + }, + { + "epoch": 1.2, + "learning_rate": 4.2186948853615524e-05, + "loss": 0.0001, + "step": 759 + }, + { + "epoch": 1.21, + "learning_rate": 4.216931216931217e-05, + "loss": 0.0001, + "step": 760 + }, + { + "epoch": 1.21, + "learning_rate": 4.215167548500882e-05, + "loss": 0.0001, + "step": 761 + }, + { + "epoch": 1.21, + "learning_rate": 4.2134038800705474e-05, + "loss": 0.0001, + "step": 762 + }, + { + "epoch": 1.21, + "learning_rate": 4.211640211640212e-05, + "loss": 0.0007, + "step": 763 + }, + { + "epoch": 1.21, + "learning_rate": 4.209876543209877e-05, + "loss": 0.0007, + "step": 764 + }, + { + "epoch": 1.21, + "learning_rate": 4.208112874779542e-05, + "loss": 0.0001, + "step": 765 + }, + { + "epoch": 1.22, + "learning_rate": 4.2063492063492065e-05, + "loss": 0.0007, + "step": 766 + }, + { + "epoch": 1.22, + "learning_rate": 4.204585537918871e-05, + "loss": 0.0002, + "step": 767 + }, + { + "epoch": 1.22, + "learning_rate": 4.202821869488536e-05, + "loss": 0.0001, + "step": 768 + }, + { + "epoch": 1.22, + "learning_rate": 4.2010582010582014e-05, + "loss": 0.0001, + "step": 769 + }, + { + "epoch": 1.22, + "learning_rate": 4.199294532627866e-05, + "loss": 0.0001, + "step": 770 + }, + { + "epoch": 1.22, + "learning_rate": 4.197530864197531e-05, + "loss": 0.0001, + "step": 771 + }, + { + "epoch": 1.23, + "learning_rate": 4.1957671957671964e-05, + "loss": 0.0001, + "step": 772 + }, + { + "epoch": 1.23, + "learning_rate": 4.194003527336861e-05, + "loss": 0.0001, + "step": 773 + }, + { + "epoch": 1.23, + "learning_rate": 4.192239858906526e-05, + "loss": 0.0005, + "step": 774 + }, + { + "epoch": 1.23, + "learning_rate": 4.190476190476191e-05, + "loss": 0.0001, + "step": 775 + }, + { + "epoch": 1.23, + "learning_rate": 4.1887125220458555e-05, + "loss": 0.0001, + "step": 776 + }, + { + "epoch": 1.23, + "learning_rate": 4.18694885361552e-05, + "loss": 0.0006, + "step": 777 + }, + { + "epoch": 1.23, + "learning_rate": 4.185185185185185e-05, + "loss": 0.0001, + "step": 778 + }, + { + "epoch": 1.24, + "learning_rate": 4.18342151675485e-05, + "loss": 0.0001, + "step": 779 + }, + { + "epoch": 1.24, + "learning_rate": 4.181657848324515e-05, + "loss": 0.0376, + "step": 780 + }, + { + "epoch": 1.24, + "learning_rate": 4.17989417989418e-05, + "loss": 0.0001, + "step": 781 + }, + { + "epoch": 1.24, + "learning_rate": 4.1781305114638455e-05, + "loss": 0.0001, + "step": 782 + }, + { + "epoch": 1.24, + "learning_rate": 4.17636684303351e-05, + "loss": 0.001, + "step": 783 + }, + { + "epoch": 1.24, + "learning_rate": 4.174603174603175e-05, + "loss": 0.0001, + "step": 784 + }, + { + "epoch": 1.25, + "learning_rate": 4.17283950617284e-05, + "loss": 0.0151, + "step": 785 + }, + { + "epoch": 1.25, + "learning_rate": 4.1710758377425046e-05, + "loss": 0.0001, + "step": 786 + }, + { + "epoch": 1.25, + "learning_rate": 4.1693121693121694e-05, + "loss": 0.0001, + "step": 787 + }, + { + "epoch": 1.25, + "learning_rate": 4.167548500881834e-05, + "loss": 0.0001, + "step": 788 + }, + { + "epoch": 1.25, + "learning_rate": 4.165784832451499e-05, + "loss": 0.0001, + "step": 789 + }, + { + "epoch": 1.25, + "learning_rate": 4.1640211640211644e-05, + "loss": 0.0001, + "step": 790 + }, + { + "epoch": 1.26, + "learning_rate": 4.162257495590829e-05, + "loss": 0.0001, + "step": 791 + }, + { + "epoch": 1.26, + "learning_rate": 4.1604938271604946e-05, + "loss": 0.0001, + "step": 792 + }, + { + "epoch": 1.26, + "learning_rate": 4.1587301587301594e-05, + "loss": 0.0001, + "step": 793 + }, + { + "epoch": 1.26, + "learning_rate": 4.156966490299824e-05, + "loss": 0.0001, + "step": 794 + }, + { + "epoch": 1.26, + "learning_rate": 4.155202821869489e-05, + "loss": 0.0001, + "step": 795 + }, + { + "epoch": 1.26, + "learning_rate": 4.153439153439154e-05, + "loss": 0.0001, + "step": 796 + }, + { + "epoch": 1.27, + "learning_rate": 4.1516754850088184e-05, + "loss": 0.0145, + "step": 797 + }, + { + "epoch": 1.27, + "learning_rate": 4.149911816578483e-05, + "loss": 0.0001, + "step": 798 + }, + { + "epoch": 1.27, + "learning_rate": 4.148148148148148e-05, + "loss": 0.0001, + "step": 799 + }, + { + "epoch": 1.27, + "learning_rate": 4.146384479717813e-05, + "loss": 0.0001, + "step": 800 + }, + { + "epoch": 1.27, + "learning_rate": 4.144620811287478e-05, + "loss": 0.0001, + "step": 801 + }, + { + "epoch": 1.27, + "learning_rate": 4.1428571428571437e-05, + "loss": 0.0001, + "step": 802 + }, + { + "epoch": 1.27, + "learning_rate": 4.1410934744268084e-05, + "loss": 0.0001, + "step": 803 + }, + { + "epoch": 1.28, + "learning_rate": 4.139329805996473e-05, + "loss": 0.0005, + "step": 804 + }, + { + "epoch": 1.28, + "learning_rate": 4.137566137566138e-05, + "loss": 0.0001, + "step": 805 + }, + { + "epoch": 1.28, + "learning_rate": 4.135802469135803e-05, + "loss": 0.0004, + "step": 806 + }, + { + "epoch": 1.28, + "learning_rate": 4.1340388007054675e-05, + "loss": 0.1308, + "step": 807 + }, + { + "epoch": 1.28, + "learning_rate": 4.132275132275132e-05, + "loss": 0.0001, + "step": 808 + }, + { + "epoch": 1.28, + "learning_rate": 4.130511463844797e-05, + "loss": 0.0001, + "step": 809 + }, + { + "epoch": 1.29, + "learning_rate": 4.128747795414462e-05, + "loss": 0.0001, + "step": 810 + }, + { + "epoch": 1.29, + "learning_rate": 4.126984126984127e-05, + "loss": 0.0001, + "step": 811 + }, + { + "epoch": 1.29, + "learning_rate": 4.125220458553792e-05, + "loss": 0.0001, + "step": 812 + }, + { + "epoch": 1.29, + "learning_rate": 4.1234567901234575e-05, + "loss": 0.0001, + "step": 813 + }, + { + "epoch": 1.29, + "learning_rate": 4.121693121693122e-05, + "loss": 0.002, + "step": 814 + }, + { + "epoch": 1.29, + "learning_rate": 4.119929453262787e-05, + "loss": 0.0001, + "step": 815 + }, + { + "epoch": 1.3, + "learning_rate": 4.118165784832452e-05, + "loss": 0.0001, + "step": 816 + }, + { + "epoch": 1.3, + "learning_rate": 4.1164021164021166e-05, + "loss": 0.0001, + "step": 817 + }, + { + "epoch": 1.3, + "learning_rate": 4.1146384479717814e-05, + "loss": 0.064, + "step": 818 + }, + { + "epoch": 1.3, + "learning_rate": 4.112874779541446e-05, + "loss": 0.0001, + "step": 819 + }, + { + "epoch": 1.3, + "learning_rate": 4.111111111111111e-05, + "loss": 0.0001, + "step": 820 + }, + { + "epoch": 1.3, + "learning_rate": 4.109347442680776e-05, + "loss": 0.0001, + "step": 821 + }, + { + "epoch": 1.3, + "learning_rate": 4.107583774250441e-05, + "loss": 0.0001, + "step": 822 + }, + { + "epoch": 1.31, + "learning_rate": 4.105820105820106e-05, + "loss": 0.0001, + "step": 823 + }, + { + "epoch": 1.31, + "learning_rate": 4.1040564373897713e-05, + "loss": 0.0001, + "step": 824 + }, + { + "epoch": 1.31, + "learning_rate": 4.102292768959436e-05, + "loss": 0.0001, + "step": 825 + }, + { + "epoch": 1.31, + "learning_rate": 4.100529100529101e-05, + "loss": 0.0001, + "step": 826 + }, + { + "epoch": 1.31, + "learning_rate": 4.0987654320987657e-05, + "loss": 0.0001, + "step": 827 + }, + { + "epoch": 1.31, + "learning_rate": 4.0970017636684304e-05, + "loss": 0.0001, + "step": 828 + }, + { + "epoch": 1.32, + "learning_rate": 4.095238095238095e-05, + "loss": 0.0001, + "step": 829 + }, + { + "epoch": 1.32, + "learning_rate": 4.09347442680776e-05, + "loss": 0.0001, + "step": 830 + }, + { + "epoch": 1.32, + "learning_rate": 4.091710758377425e-05, + "loss": 0.0001, + "step": 831 + }, + { + "epoch": 1.32, + "learning_rate": 4.08994708994709e-05, + "loss": 0.0001, + "step": 832 + }, + { + "epoch": 1.32, + "learning_rate": 4.088183421516755e-05, + "loss": 0.0001, + "step": 833 + }, + { + "epoch": 1.32, + "learning_rate": 4.0864197530864204e-05, + "loss": 0.0001, + "step": 834 + }, + { + "epoch": 1.33, + "learning_rate": 4.084656084656085e-05, + "loss": 0.0001, + "step": 835 + }, + { + "epoch": 1.33, + "learning_rate": 4.08289241622575e-05, + "loss": 0.0001, + "step": 836 + }, + { + "epoch": 1.33, + "learning_rate": 4.081128747795415e-05, + "loss": 0.0001, + "step": 837 + }, + { + "epoch": 1.33, + "learning_rate": 4.0793650793650795e-05, + "loss": 0.0001, + "step": 838 + }, + { + "epoch": 1.33, + "learning_rate": 4.077601410934744e-05, + "loss": 0.0041, + "step": 839 + }, + { + "epoch": 1.33, + "learning_rate": 4.075837742504409e-05, + "loss": 0.0001, + "step": 840 + }, + { + "epoch": 1.33, + "learning_rate": 4.074074074074074e-05, + "loss": 0.0001, + "step": 841 + }, + { + "epoch": 1.34, + "learning_rate": 4.072310405643739e-05, + "loss": 0.0001, + "step": 842 + }, + { + "epoch": 1.34, + "learning_rate": 4.070546737213404e-05, + "loss": 0.0001, + "step": 843 + }, + { + "epoch": 1.34, + "learning_rate": 4.068783068783069e-05, + "loss": 0.0026, + "step": 844 + }, + { + "epoch": 1.34, + "learning_rate": 4.067019400352734e-05, + "loss": 0.0001, + "step": 845 + }, + { + "epoch": 1.34, + "learning_rate": 4.065255731922399e-05, + "loss": 0.0001, + "step": 846 + }, + { + "epoch": 1.34, + "learning_rate": 4.063492063492064e-05, + "loss": 0.0001, + "step": 847 + }, + { + "epoch": 1.35, + "learning_rate": 4.0617283950617286e-05, + "loss": 0.0001, + "step": 848 + }, + { + "epoch": 1.35, + "learning_rate": 4.0599647266313933e-05, + "loss": 0.0018, + "step": 849 + }, + { + "epoch": 1.35, + "learning_rate": 4.058201058201058e-05, + "loss": 0.0001, + "step": 850 + }, + { + "epoch": 1.35, + "learning_rate": 4.056437389770723e-05, + "loss": 0.0001, + "step": 851 + }, + { + "epoch": 1.35, + "learning_rate": 4.0546737213403883e-05, + "loss": 0.0001, + "step": 852 + }, + { + "epoch": 1.35, + "learning_rate": 4.052910052910053e-05, + "loss": 0.0001, + "step": 853 + }, + { + "epoch": 1.36, + "learning_rate": 4.051146384479718e-05, + "loss": 0.0009, + "step": 854 + }, + { + "epoch": 1.36, + "learning_rate": 4.049382716049383e-05, + "loss": 0.0001, + "step": 855 + }, + { + "epoch": 1.36, + "learning_rate": 4.047619047619048e-05, + "loss": 0.0001, + "step": 856 + }, + { + "epoch": 1.36, + "learning_rate": 4.045855379188713e-05, + "loss": 0.0001, + "step": 857 + }, + { + "epoch": 1.36, + "learning_rate": 4.0440917107583776e-05, + "loss": 0.0001, + "step": 858 + }, + { + "epoch": 1.36, + "learning_rate": 4.0423280423280424e-05, + "loss": 0.0001, + "step": 859 + }, + { + "epoch": 1.37, + "learning_rate": 4.040564373897707e-05, + "loss": 0.0, + "step": 860 + }, + { + "epoch": 1.37, + "learning_rate": 4.038800705467372e-05, + "loss": 0.0001, + "step": 861 + }, + { + "epoch": 1.37, + "learning_rate": 4.0370370370370374e-05, + "loss": 0.0, + "step": 862 + }, + { + "epoch": 1.37, + "learning_rate": 4.035273368606702e-05, + "loss": 0.0, + "step": 863 + }, + { + "epoch": 1.37, + "learning_rate": 4.033509700176367e-05, + "loss": 0.0, + "step": 864 + }, + { + "epoch": 1.37, + "learning_rate": 4.031746031746032e-05, + "loss": 0.0015, + "step": 865 + }, + { + "epoch": 1.37, + "learning_rate": 4.029982363315697e-05, + "loss": 0.0, + "step": 866 + }, + { + "epoch": 1.38, + "learning_rate": 4.028218694885362e-05, + "loss": 0.0, + "step": 867 + }, + { + "epoch": 1.38, + "learning_rate": 4.026455026455027e-05, + "loss": 0.0, + "step": 868 + }, + { + "epoch": 1.38, + "learning_rate": 4.0246913580246915e-05, + "loss": 0.0, + "step": 869 + }, + { + "epoch": 1.38, + "learning_rate": 4.022927689594356e-05, + "loss": 0.0, + "step": 870 + }, + { + "epoch": 1.38, + "learning_rate": 4.021164021164021e-05, + "loss": 0.0, + "step": 871 + }, + { + "epoch": 1.38, + "learning_rate": 4.0194003527336865e-05, + "loss": 0.0, + "step": 872 + }, + { + "epoch": 1.39, + "learning_rate": 4.017636684303351e-05, + "loss": 0.0, + "step": 873 + }, + { + "epoch": 1.39, + "learning_rate": 4.015873015873016e-05, + "loss": 0.0, + "step": 874 + }, + { + "epoch": 1.39, + "learning_rate": 4.014109347442681e-05, + "loss": 0.0, + "step": 875 + }, + { + "epoch": 1.39, + "learning_rate": 4.012345679012346e-05, + "loss": 0.001, + "step": 876 + }, + { + "epoch": 1.39, + "learning_rate": 4.010582010582011e-05, + "loss": 0.001, + "step": 877 + }, + { + "epoch": 1.39, + "learning_rate": 4.008818342151676e-05, + "loss": 0.0, + "step": 878 + }, + { + "epoch": 1.4, + "learning_rate": 4.0070546737213406e-05, + "loss": 0.0013, + "step": 879 + }, + { + "epoch": 1.4, + "learning_rate": 4.005291005291005e-05, + "loss": 0.0, + "step": 880 + }, + { + "epoch": 1.4, + "learning_rate": 4.00352733686067e-05, + "loss": 0.0, + "step": 881 + }, + { + "epoch": 1.4, + "learning_rate": 4.0017636684303356e-05, + "loss": 0.0, + "step": 882 + }, + { + "epoch": 1.4, + "learning_rate": 4e-05, + "loss": 0.0, + "step": 883 + }, + { + "epoch": 1.4, + "learning_rate": 3.998236331569665e-05, + "loss": 0.0, + "step": 884 + }, + { + "epoch": 1.4, + "learning_rate": 3.99647266313933e-05, + "loss": 0.0, + "step": 885 + }, + { + "epoch": 1.41, + "learning_rate": 3.9947089947089946e-05, + "loss": 0.0, + "step": 886 + }, + { + "epoch": 1.41, + "learning_rate": 3.99294532627866e-05, + "loss": 0.0, + "step": 887 + }, + { + "epoch": 1.41, + "learning_rate": 3.991181657848325e-05, + "loss": 0.0, + "step": 888 + }, + { + "epoch": 1.41, + "learning_rate": 3.9894179894179896e-05, + "loss": 0.0, + "step": 889 + }, + { + "epoch": 1.41, + "learning_rate": 3.9876543209876544e-05, + "loss": 0.0, + "step": 890 + }, + { + "epoch": 1.41, + "learning_rate": 3.985890652557319e-05, + "loss": 0.0, + "step": 891 + }, + { + "epoch": 1.42, + "learning_rate": 3.984126984126984e-05, + "loss": 0.0, + "step": 892 + }, + { + "epoch": 1.42, + "learning_rate": 3.9823633156966494e-05, + "loss": 0.0015, + "step": 893 + }, + { + "epoch": 1.42, + "learning_rate": 3.980599647266314e-05, + "loss": 0.0, + "step": 894 + }, + { + "epoch": 1.42, + "learning_rate": 3.978835978835979e-05, + "loss": 0.0, + "step": 895 + }, + { + "epoch": 1.42, + "learning_rate": 3.977072310405644e-05, + "loss": 0.0, + "step": 896 + }, + { + "epoch": 1.42, + "learning_rate": 3.975308641975309e-05, + "loss": 0.0, + "step": 897 + }, + { + "epoch": 1.43, + "learning_rate": 3.973544973544974e-05, + "loss": 0.0, + "step": 898 + }, + { + "epoch": 1.43, + "learning_rate": 3.971781305114639e-05, + "loss": 0.0, + "step": 899 + }, + { + "epoch": 1.43, + "learning_rate": 3.9700176366843035e-05, + "loss": 0.0, + "step": 900 + }, + { + "epoch": 1.43, + "learning_rate": 3.968253968253968e-05, + "loss": 0.0, + "step": 901 + }, + { + "epoch": 1.43, + "learning_rate": 3.966490299823633e-05, + "loss": 0.0, + "step": 902 + }, + { + "epoch": 1.43, + "learning_rate": 3.9647266313932985e-05, + "loss": 0.0, + "step": 903 + }, + { + "epoch": 1.43, + "learning_rate": 3.962962962962963e-05, + "loss": 0.0043, + "step": 904 + }, + { + "epoch": 1.44, + "learning_rate": 3.961199294532628e-05, + "loss": 0.0, + "step": 905 + }, + { + "epoch": 1.44, + "learning_rate": 3.959435626102293e-05, + "loss": 0.0, + "step": 906 + }, + { + "epoch": 1.44, + "learning_rate": 3.9576719576719576e-05, + "loss": 0.0005, + "step": 907 + }, + { + "epoch": 1.44, + "learning_rate": 3.955908289241623e-05, + "loss": 0.0, + "step": 908 + }, + { + "epoch": 1.44, + "learning_rate": 3.954144620811288e-05, + "loss": 0.0, + "step": 909 + }, + { + "epoch": 1.44, + "learning_rate": 3.9523809523809526e-05, + "loss": 0.0007, + "step": 910 + }, + { + "epoch": 1.45, + "learning_rate": 3.950617283950617e-05, + "loss": 0.0006, + "step": 911 + }, + { + "epoch": 1.45, + "learning_rate": 3.948853615520282e-05, + "loss": 0.0, + "step": 912 + }, + { + "epoch": 1.45, + "learning_rate": 3.9470899470899475e-05, + "loss": 0.0006, + "step": 913 + }, + { + "epoch": 1.45, + "learning_rate": 3.945326278659612e-05, + "loss": 0.0, + "step": 914 + }, + { + "epoch": 1.45, + "learning_rate": 3.943562610229277e-05, + "loss": 0.0007, + "step": 915 + }, + { + "epoch": 1.45, + "learning_rate": 3.941798941798942e-05, + "loss": 0.0, + "step": 916 + }, + { + "epoch": 1.46, + "learning_rate": 3.9400352733686066e-05, + "loss": 0.0, + "step": 917 + }, + { + "epoch": 1.46, + "learning_rate": 3.938271604938272e-05, + "loss": 0.0, + "step": 918 + }, + { + "epoch": 1.46, + "learning_rate": 3.936507936507937e-05, + "loss": 0.0, + "step": 919 + }, + { + "epoch": 1.46, + "learning_rate": 3.9347442680776016e-05, + "loss": 0.0, + "step": 920 + }, + { + "epoch": 1.46, + "learning_rate": 3.9329805996472664e-05, + "loss": 0.0, + "step": 921 + }, + { + "epoch": 1.46, + "learning_rate": 3.931216931216931e-05, + "loss": 0.0, + "step": 922 + }, + { + "epoch": 1.47, + "learning_rate": 3.9294532627865966e-05, + "loss": 0.0, + "step": 923 + }, + { + "epoch": 1.47, + "learning_rate": 3.9276895943562614e-05, + "loss": 0.0, + "step": 924 + }, + { + "epoch": 1.47, + "learning_rate": 3.925925925925926e-05, + "loss": 0.0, + "step": 925 + }, + { + "epoch": 1.47, + "learning_rate": 3.924162257495591e-05, + "loss": 0.0, + "step": 926 + }, + { + "epoch": 1.47, + "learning_rate": 3.922398589065256e-05, + "loss": 0.0011, + "step": 927 + }, + { + "epoch": 1.47, + "learning_rate": 3.9206349206349205e-05, + "loss": 0.0, + "step": 928 + }, + { + "epoch": 1.47, + "learning_rate": 3.918871252204586e-05, + "loss": 0.0, + "step": 929 + }, + { + "epoch": 1.48, + "learning_rate": 3.917107583774251e-05, + "loss": 0.0, + "step": 930 + }, + { + "epoch": 1.48, + "learning_rate": 3.9153439153439155e-05, + "loss": 0.0, + "step": 931 + }, + { + "epoch": 1.48, + "learning_rate": 3.91358024691358e-05, + "loss": 0.0, + "step": 932 + }, + { + "epoch": 1.48, + "learning_rate": 3.911816578483246e-05, + "loss": 0.0, + "step": 933 + }, + { + "epoch": 1.48, + "learning_rate": 3.9100529100529105e-05, + "loss": 0.0, + "step": 934 + }, + { + "epoch": 1.48, + "learning_rate": 3.908289241622575e-05, + "loss": 0.0005, + "step": 935 + }, + { + "epoch": 1.49, + "learning_rate": 3.90652557319224e-05, + "loss": 0.0, + "step": 936 + }, + { + "epoch": 1.49, + "learning_rate": 3.904761904761905e-05, + "loss": 0.0, + "step": 937 + }, + { + "epoch": 1.49, + "learning_rate": 3.9029982363315695e-05, + "loss": 0.0, + "step": 938 + }, + { + "epoch": 1.49, + "learning_rate": 3.901234567901234e-05, + "loss": 0.0, + "step": 939 + }, + { + "epoch": 1.49, + "learning_rate": 3.8994708994709e-05, + "loss": 0.0, + "step": 940 + }, + { + "epoch": 1.49, + "learning_rate": 3.8977072310405645e-05, + "loss": 0.0, + "step": 941 + }, + { + "epoch": 1.5, + "learning_rate": 3.895943562610229e-05, + "loss": 0.0, + "step": 942 + }, + { + "epoch": 1.5, + "learning_rate": 3.894179894179895e-05, + "loss": 0.0, + "step": 943 + }, + { + "epoch": 1.5, + "learning_rate": 3.8924162257495595e-05, + "loss": 0.0, + "step": 944 + }, + { + "epoch": 1.5, + "learning_rate": 3.890652557319224e-05, + "loss": 0.0, + "step": 945 + }, + { + "epoch": 1.5, + "learning_rate": 3.888888888888889e-05, + "loss": 0.0, + "step": 946 + }, + { + "epoch": 1.5, + "learning_rate": 3.887125220458554e-05, + "loss": 0.0, + "step": 947 + }, + { + "epoch": 1.5, + "learning_rate": 3.8853615520282186e-05, + "loss": 0.0, + "step": 948 + }, + { + "epoch": 1.51, + "learning_rate": 3.8835978835978834e-05, + "loss": 0.0, + "step": 949 + }, + { + "epoch": 1.51, + "learning_rate": 3.881834215167549e-05, + "loss": 0.0, + "step": 950 + }, + { + "epoch": 1.51, + "learning_rate": 3.8800705467372136e-05, + "loss": 0.0, + "step": 951 + }, + { + "epoch": 1.51, + "learning_rate": 3.8783068783068784e-05, + "loss": 0.0, + "step": 952 + }, + { + "epoch": 1.51, + "learning_rate": 3.876543209876544e-05, + "loss": 0.0, + "step": 953 + }, + { + "epoch": 1.51, + "learning_rate": 3.8747795414462086e-05, + "loss": 0.0, + "step": 954 + }, + { + "epoch": 1.52, + "learning_rate": 3.8730158730158734e-05, + "loss": 0.0, + "step": 955 + }, + { + "epoch": 1.52, + "learning_rate": 3.871252204585538e-05, + "loss": 0.0005, + "step": 956 + }, + { + "epoch": 1.52, + "learning_rate": 3.869488536155203e-05, + "loss": 0.0, + "step": 957 + }, + { + "epoch": 1.52, + "learning_rate": 3.867724867724868e-05, + "loss": 0.0004, + "step": 958 + }, + { + "epoch": 1.52, + "learning_rate": 3.8659611992945325e-05, + "loss": 0.0, + "step": 959 + }, + { + "epoch": 1.52, + "learning_rate": 3.864197530864197e-05, + "loss": 0.0, + "step": 960 + }, + { + "epoch": 1.53, + "learning_rate": 3.862433862433863e-05, + "loss": 0.0, + "step": 961 + }, + { + "epoch": 1.53, + "learning_rate": 3.8606701940035275e-05, + "loss": 0.0, + "step": 962 + }, + { + "epoch": 1.53, + "learning_rate": 3.858906525573193e-05, + "loss": 0.0, + "step": 963 + }, + { + "epoch": 1.53, + "learning_rate": 3.857142857142858e-05, + "loss": 0.0, + "step": 964 + }, + { + "epoch": 1.53, + "learning_rate": 3.8553791887125224e-05, + "loss": 0.0005, + "step": 965 + }, + { + "epoch": 1.53, + "learning_rate": 3.853615520282187e-05, + "loss": 0.0003, + "step": 966 + }, + { + "epoch": 1.53, + "learning_rate": 3.851851851851852e-05, + "loss": 0.0, + "step": 967 + }, + { + "epoch": 1.54, + "learning_rate": 3.850088183421517e-05, + "loss": 0.0, + "step": 968 + }, + { + "epoch": 1.54, + "learning_rate": 3.8483245149911815e-05, + "loss": 0.0006, + "step": 969 + }, + { + "epoch": 1.54, + "learning_rate": 3.846560846560846e-05, + "loss": 0.0, + "step": 970 + }, + { + "epoch": 1.54, + "learning_rate": 3.844797178130512e-05, + "loss": 0.0, + "step": 971 + }, + { + "epoch": 1.54, + "learning_rate": 3.8430335097001765e-05, + "loss": 0.0, + "step": 972 + }, + { + "epoch": 1.54, + "learning_rate": 3.841269841269842e-05, + "loss": 0.0007, + "step": 973 + }, + { + "epoch": 1.55, + "learning_rate": 3.839506172839507e-05, + "loss": 0.0, + "step": 974 + }, + { + "epoch": 1.55, + "learning_rate": 3.8377425044091715e-05, + "loss": 0.0, + "step": 975 + }, + { + "epoch": 1.55, + "learning_rate": 3.835978835978836e-05, + "loss": 0.0, + "step": 976 + }, + { + "epoch": 1.55, + "learning_rate": 3.834215167548501e-05, + "loss": 0.0, + "step": 977 + }, + { + "epoch": 1.55, + "learning_rate": 3.832451499118166e-05, + "loss": 0.0, + "step": 978 + }, + { + "epoch": 1.55, + "learning_rate": 3.8306878306878306e-05, + "loss": 0.0003, + "step": 979 + }, + { + "epoch": 1.56, + "learning_rate": 3.8289241622574954e-05, + "loss": 0.0, + "step": 980 + }, + { + "epoch": 1.56, + "learning_rate": 3.82716049382716e-05, + "loss": 0.0, + "step": 981 + }, + { + "epoch": 1.56, + "learning_rate": 3.8253968253968256e-05, + "loss": 0.0003, + "step": 982 + }, + { + "epoch": 1.56, + "learning_rate": 3.823633156966491e-05, + "loss": 0.0, + "step": 983 + }, + { + "epoch": 1.56, + "learning_rate": 3.821869488536156e-05, + "loss": 0.0, + "step": 984 + }, + { + "epoch": 1.56, + "learning_rate": 3.8201058201058206e-05, + "loss": 0.0003, + "step": 985 + }, + { + "epoch": 1.57, + "learning_rate": 3.8183421516754854e-05, + "loss": 0.0, + "step": 986 + }, + { + "epoch": 1.57, + "learning_rate": 3.81657848324515e-05, + "loss": 0.0, + "step": 987 + }, + { + "epoch": 1.57, + "learning_rate": 3.814814814814815e-05, + "loss": 0.0, + "step": 988 + }, + { + "epoch": 1.57, + "learning_rate": 3.81305114638448e-05, + "loss": 0.0, + "step": 989 + }, + { + "epoch": 1.57, + "learning_rate": 3.8112874779541445e-05, + "loss": 0.0, + "step": 990 + }, + { + "epoch": 1.57, + "learning_rate": 3.809523809523809e-05, + "loss": 0.0, + "step": 991 + }, + { + "epoch": 1.57, + "learning_rate": 3.807760141093475e-05, + "loss": 0.0, + "step": 992 + }, + { + "epoch": 1.58, + "learning_rate": 3.8059964726631394e-05, + "loss": 0.0, + "step": 993 + }, + { + "epoch": 1.58, + "learning_rate": 3.804232804232805e-05, + "loss": 0.0, + "step": 994 + }, + { + "epoch": 1.58, + "learning_rate": 3.80246913580247e-05, + "loss": 0.0002, + "step": 995 + }, + { + "epoch": 1.58, + "learning_rate": 3.8007054673721344e-05, + "loss": 0.0, + "step": 996 + }, + { + "epoch": 1.58, + "learning_rate": 3.798941798941799e-05, + "loss": 0.0, + "step": 997 + }, + { + "epoch": 1.58, + "learning_rate": 3.797178130511464e-05, + "loss": 0.0, + "step": 998 + }, + { + "epoch": 1.59, + "learning_rate": 3.795414462081129e-05, + "loss": 0.0, + "step": 999 + }, + { + "epoch": 1.59, + "learning_rate": 3.7936507936507935e-05, + "loss": 0.0, + "step": 1000 + }, + { + "epoch": 1.59, + "learning_rate": 3.791887125220458e-05, + "loss": 0.0, + "step": 1001 + }, + { + "epoch": 1.59, + "learning_rate": 3.790123456790123e-05, + "loss": 0.0, + "step": 1002 + }, + { + "epoch": 1.59, + "learning_rate": 3.7883597883597885e-05, + "loss": 0.0, + "step": 1003 + }, + { + "epoch": 1.59, + "learning_rate": 3.786596119929453e-05, + "loss": 0.0, + "step": 1004 + }, + { + "epoch": 1.6, + "learning_rate": 3.784832451499119e-05, + "loss": 0.0, + "step": 1005 + }, + { + "epoch": 1.6, + "learning_rate": 3.7830687830687835e-05, + "loss": 0.0, + "step": 1006 + }, + { + "epoch": 1.6, + "learning_rate": 3.781305114638448e-05, + "loss": 0.0, + "step": 1007 + }, + { + "epoch": 1.6, + "learning_rate": 3.779541446208113e-05, + "loss": 0.0, + "step": 1008 + }, + { + "epoch": 1.6, + "learning_rate": 3.777777777777778e-05, + "loss": 0.0, + "step": 1009 + }, + { + "epoch": 1.6, + "learning_rate": 3.7760141093474426e-05, + "loss": 0.0006, + "step": 1010 + }, + { + "epoch": 1.6, + "learning_rate": 3.7742504409171074e-05, + "loss": 0.0, + "step": 1011 + }, + { + "epoch": 1.61, + "learning_rate": 3.772486772486772e-05, + "loss": 0.0, + "step": 1012 + }, + { + "epoch": 1.61, + "learning_rate": 3.7707231040564376e-05, + "loss": 0.0, + "step": 1013 + }, + { + "epoch": 1.61, + "learning_rate": 3.7689594356261024e-05, + "loss": 0.0002, + "step": 1014 + }, + { + "epoch": 1.61, + "learning_rate": 3.767195767195768e-05, + "loss": 0.0, + "step": 1015 + }, + { + "epoch": 1.61, + "learning_rate": 3.7654320987654326e-05, + "loss": 0.0, + "step": 1016 + }, + { + "epoch": 1.61, + "learning_rate": 3.7636684303350974e-05, + "loss": 0.0, + "step": 1017 + }, + { + "epoch": 1.62, + "learning_rate": 3.761904761904762e-05, + "loss": 0.0, + "step": 1018 + }, + { + "epoch": 1.62, + "learning_rate": 3.760141093474427e-05, + "loss": 0.0, + "step": 1019 + }, + { + "epoch": 1.62, + "learning_rate": 3.758377425044092e-05, + "loss": 0.0, + "step": 1020 + }, + { + "epoch": 1.62, + "learning_rate": 3.7566137566137564e-05, + "loss": 0.0, + "step": 1021 + }, + { + "epoch": 1.62, + "learning_rate": 3.754850088183421e-05, + "loss": 0.0003, + "step": 1022 + }, + { + "epoch": 1.62, + "learning_rate": 3.7530864197530867e-05, + "loss": 0.0, + "step": 1023 + }, + { + "epoch": 1.63, + "learning_rate": 3.7513227513227514e-05, + "loss": 0.0, + "step": 1024 + }, + { + "epoch": 1.63, + "learning_rate": 3.749559082892416e-05, + "loss": 0.0, + "step": 1025 + }, + { + "epoch": 1.63, + "learning_rate": 3.7477954144620817e-05, + "loss": 0.0, + "step": 1026 + }, + { + "epoch": 1.63, + "learning_rate": 3.7460317460317464e-05, + "loss": 0.0002, + "step": 1027 + }, + { + "epoch": 1.63, + "learning_rate": 3.744268077601411e-05, + "loss": 0.0, + "step": 1028 + }, + { + "epoch": 1.63, + "learning_rate": 3.742504409171076e-05, + "loss": 0.0, + "step": 1029 + }, + { + "epoch": 1.63, + "learning_rate": 3.740740740740741e-05, + "loss": 0.0, + "step": 1030 + }, + { + "epoch": 1.64, + "learning_rate": 3.7389770723104055e-05, + "loss": 0.0, + "step": 1031 + }, + { + "epoch": 1.64, + "learning_rate": 3.73721340388007e-05, + "loss": 0.0, + "step": 1032 + }, + { + "epoch": 1.64, + "learning_rate": 3.735449735449736e-05, + "loss": 0.0, + "step": 1033 + }, + { + "epoch": 1.64, + "learning_rate": 3.7336860670194005e-05, + "loss": 0.0, + "step": 1034 + }, + { + "epoch": 1.64, + "learning_rate": 3.731922398589065e-05, + "loss": 0.0, + "step": 1035 + }, + { + "epoch": 1.64, + "learning_rate": 3.730158730158731e-05, + "loss": 0.0, + "step": 1036 + }, + { + "epoch": 1.65, + "learning_rate": 3.7283950617283955e-05, + "loss": 0.0, + "step": 1037 + }, + { + "epoch": 1.65, + "learning_rate": 3.72663139329806e-05, + "loss": 0.0, + "step": 1038 + }, + { + "epoch": 1.65, + "learning_rate": 3.724867724867725e-05, + "loss": 0.0, + "step": 1039 + }, + { + "epoch": 1.65, + "learning_rate": 3.72310405643739e-05, + "loss": 0.0002, + "step": 1040 + }, + { + "epoch": 1.65, + "learning_rate": 3.7213403880070546e-05, + "loss": 0.0, + "step": 1041 + }, + { + "epoch": 1.65, + "learning_rate": 3.7195767195767194e-05, + "loss": 0.0, + "step": 1042 + }, + { + "epoch": 1.66, + "learning_rate": 3.717813051146385e-05, + "loss": 0.0, + "step": 1043 + }, + { + "epoch": 1.66, + "learning_rate": 3.7160493827160496e-05, + "loss": 0.0, + "step": 1044 + }, + { + "epoch": 1.66, + "learning_rate": 3.7142857142857143e-05, + "loss": 0.0, + "step": 1045 + }, + { + "epoch": 1.66, + "learning_rate": 3.712522045855379e-05, + "loss": 0.0, + "step": 1046 + }, + { + "epoch": 1.66, + "learning_rate": 3.7107583774250446e-05, + "loss": 0.0, + "step": 1047 + }, + { + "epoch": 1.66, + "learning_rate": 3.7089947089947093e-05, + "loss": 0.0, + "step": 1048 + }, + { + "epoch": 1.67, + "learning_rate": 3.707231040564374e-05, + "loss": 0.0, + "step": 1049 + }, + { + "epoch": 1.67, + "learning_rate": 3.705467372134039e-05, + "loss": 0.0, + "step": 1050 + }, + { + "epoch": 1.67, + "learning_rate": 3.7037037037037037e-05, + "loss": 0.0, + "step": 1051 + }, + { + "epoch": 1.67, + "learning_rate": 3.7019400352733684e-05, + "loss": 0.0, + "step": 1052 + }, + { + "epoch": 1.67, + "learning_rate": 3.700176366843034e-05, + "loss": 0.0, + "step": 1053 + }, + { + "epoch": 1.67, + "learning_rate": 3.6984126984126986e-05, + "loss": 0.0, + "step": 1054 + }, + { + "epoch": 1.67, + "learning_rate": 3.6966490299823634e-05, + "loss": 0.0, + "step": 1055 + }, + { + "epoch": 1.68, + "learning_rate": 3.694885361552028e-05, + "loss": 0.0, + "step": 1056 + }, + { + "epoch": 1.68, + "learning_rate": 3.6931216931216936e-05, + "loss": 0.0, + "step": 1057 + }, + { + "epoch": 1.68, + "learning_rate": 3.6913580246913584e-05, + "loss": 0.0, + "step": 1058 + }, + { + "epoch": 1.68, + "learning_rate": 3.689594356261023e-05, + "loss": 0.0, + "step": 1059 + }, + { + "epoch": 1.68, + "learning_rate": 3.687830687830688e-05, + "loss": 0.0, + "step": 1060 + }, + { + "epoch": 1.68, + "learning_rate": 3.686067019400353e-05, + "loss": 0.0, + "step": 1061 + }, + { + "epoch": 1.69, + "learning_rate": 3.6843033509700175e-05, + "loss": 0.0, + "step": 1062 + }, + { + "epoch": 1.69, + "learning_rate": 3.682539682539683e-05, + "loss": 0.0, + "step": 1063 + }, + { + "epoch": 1.69, + "learning_rate": 3.680776014109348e-05, + "loss": 0.0, + "step": 1064 + }, + { + "epoch": 1.69, + "learning_rate": 3.6790123456790125e-05, + "loss": 0.0, + "step": 1065 + }, + { + "epoch": 1.69, + "learning_rate": 3.677248677248677e-05, + "loss": 0.0, + "step": 1066 + }, + { + "epoch": 1.69, + "learning_rate": 3.675485008818342e-05, + "loss": 0.0, + "step": 1067 + }, + { + "epoch": 1.7, + "learning_rate": 3.6737213403880075e-05, + "loss": 0.1229, + "step": 1068 + }, + { + "epoch": 1.7, + "learning_rate": 3.671957671957672e-05, + "loss": 0.004, + "step": 1069 + }, + { + "epoch": 1.7, + "learning_rate": 3.670194003527337e-05, + "loss": 0.0, + "step": 1070 + }, + { + "epoch": 1.7, + "learning_rate": 3.668430335097002e-05, + "loss": 0.0, + "step": 1071 + }, + { + "epoch": 1.7, + "learning_rate": 3.6666666666666666e-05, + "loss": 0.0, + "step": 1072 + }, + { + "epoch": 1.7, + "learning_rate": 3.664902998236332e-05, + "loss": 0.1125, + "step": 1073 + }, + { + "epoch": 1.7, + "learning_rate": 3.663139329805997e-05, + "loss": 0.0, + "step": 1074 + }, + { + "epoch": 1.71, + "learning_rate": 3.6613756613756616e-05, + "loss": 0.0, + "step": 1075 + }, + { + "epoch": 1.71, + "learning_rate": 3.659611992945326e-05, + "loss": 0.0, + "step": 1076 + }, + { + "epoch": 1.71, + "learning_rate": 3.657848324514991e-05, + "loss": 0.0, + "step": 1077 + }, + { + "epoch": 1.71, + "learning_rate": 3.6560846560846566e-05, + "loss": 0.0, + "step": 1078 + }, + { + "epoch": 1.71, + "learning_rate": 3.654320987654321e-05, + "loss": 0.0028, + "step": 1079 + }, + { + "epoch": 1.71, + "learning_rate": 3.652557319223986e-05, + "loss": 0.0, + "step": 1080 + }, + { + "epoch": 1.72, + "learning_rate": 3.650793650793651e-05, + "loss": 0.0, + "step": 1081 + }, + { + "epoch": 1.72, + "learning_rate": 3.6490299823633156e-05, + "loss": 0.0, + "step": 1082 + }, + { + "epoch": 1.72, + "learning_rate": 3.647266313932981e-05, + "loss": 0.0, + "step": 1083 + }, + { + "epoch": 1.72, + "learning_rate": 3.645502645502646e-05, + "loss": 0.0, + "step": 1084 + }, + { + "epoch": 1.72, + "learning_rate": 3.6437389770723106e-05, + "loss": 0.0, + "step": 1085 + }, + { + "epoch": 1.72, + "learning_rate": 3.6419753086419754e-05, + "loss": 0.0, + "step": 1086 + }, + { + "epoch": 1.73, + "learning_rate": 3.64021164021164e-05, + "loss": 0.0003, + "step": 1087 + }, + { + "epoch": 1.73, + "learning_rate": 3.638447971781305e-05, + "loss": 0.0, + "step": 1088 + }, + { + "epoch": 1.73, + "learning_rate": 3.6366843033509704e-05, + "loss": 0.0, + "step": 1089 + }, + { + "epoch": 1.73, + "learning_rate": 3.634920634920635e-05, + "loss": 0.0, + "step": 1090 + }, + { + "epoch": 1.73, + "learning_rate": 3.6331569664903e-05, + "loss": 0.0, + "step": 1091 + }, + { + "epoch": 1.73, + "learning_rate": 3.631393298059965e-05, + "loss": 0.0, + "step": 1092 + }, + { + "epoch": 1.73, + "learning_rate": 3.62962962962963e-05, + "loss": 0.0, + "step": 1093 + }, + { + "epoch": 1.74, + "learning_rate": 3.627865961199295e-05, + "loss": 0.0001, + "step": 1094 + }, + { + "epoch": 1.74, + "learning_rate": 3.62610229276896e-05, + "loss": 0.0, + "step": 1095 + }, + { + "epoch": 1.74, + "learning_rate": 3.6243386243386245e-05, + "loss": 0.0, + "step": 1096 + }, + { + "epoch": 1.74, + "learning_rate": 3.622574955908289e-05, + "loss": 0.0, + "step": 1097 + }, + { + "epoch": 1.74, + "learning_rate": 3.620811287477954e-05, + "loss": 0.0, + "step": 1098 + }, + { + "epoch": 1.74, + "learning_rate": 3.619047619047619e-05, + "loss": 0.0, + "step": 1099 + }, + { + "epoch": 1.75, + "learning_rate": 3.617283950617284e-05, + "loss": 0.0002, + "step": 1100 + }, + { + "epoch": 1.75, + "learning_rate": 3.615520282186949e-05, + "loss": 0.0, + "step": 1101 + }, + { + "epoch": 1.75, + "learning_rate": 3.613756613756614e-05, + "loss": 0.0001, + "step": 1102 + }, + { + "epoch": 1.75, + "learning_rate": 3.611992945326279e-05, + "loss": 0.0002, + "step": 1103 + }, + { + "epoch": 1.75, + "learning_rate": 3.610229276895944e-05, + "loss": 0.0, + "step": 1104 + }, + { + "epoch": 1.75, + "learning_rate": 3.608465608465609e-05, + "loss": 0.0, + "step": 1105 + }, + { + "epoch": 1.76, + "learning_rate": 3.6067019400352736e-05, + "loss": 0.0, + "step": 1106 + }, + { + "epoch": 1.76, + "learning_rate": 3.604938271604938e-05, + "loss": 0.0, + "step": 1107 + }, + { + "epoch": 1.76, + "learning_rate": 3.603174603174603e-05, + "loss": 0.0, + "step": 1108 + }, + { + "epoch": 1.76, + "learning_rate": 3.601410934744268e-05, + "loss": 0.0006, + "step": 1109 + }, + { + "epoch": 1.76, + "learning_rate": 3.599647266313933e-05, + "loss": 0.0, + "step": 1110 + }, + { + "epoch": 1.76, + "learning_rate": 3.597883597883598e-05, + "loss": 0.0001, + "step": 1111 + }, + { + "epoch": 1.77, + "learning_rate": 3.596119929453263e-05, + "loss": 0.0004, + "step": 1112 + }, + { + "epoch": 1.77, + "learning_rate": 3.594356261022928e-05, + "loss": 0.0, + "step": 1113 + }, + { + "epoch": 1.77, + "learning_rate": 3.592592592592593e-05, + "loss": 0.0002, + "step": 1114 + }, + { + "epoch": 1.77, + "learning_rate": 3.590828924162258e-05, + "loss": 0.0, + "step": 1115 + }, + { + "epoch": 1.77, + "learning_rate": 3.5890652557319226e-05, + "loss": 0.0002, + "step": 1116 + }, + { + "epoch": 1.77, + "learning_rate": 3.5873015873015874e-05, + "loss": 0.0, + "step": 1117 + }, + { + "epoch": 1.77, + "learning_rate": 3.585537918871252e-05, + "loss": 0.0, + "step": 1118 + }, + { + "epoch": 1.78, + "learning_rate": 3.583774250440917e-05, + "loss": 0.0001, + "step": 1119 + }, + { + "epoch": 1.78, + "learning_rate": 3.582010582010582e-05, + "loss": 0.0, + "step": 1120 + }, + { + "epoch": 1.78, + "learning_rate": 3.580246913580247e-05, + "loss": 0.0001, + "step": 1121 + }, + { + "epoch": 1.78, + "learning_rate": 3.578483245149912e-05, + "loss": 0.0, + "step": 1122 + }, + { + "epoch": 1.78, + "learning_rate": 3.5767195767195774e-05, + "loss": 0.0, + "step": 1123 + }, + { + "epoch": 1.78, + "learning_rate": 3.574955908289242e-05, + "loss": 0.0, + "step": 1124 + }, + { + "epoch": 1.79, + "learning_rate": 3.573192239858907e-05, + "loss": 0.0, + "step": 1125 + }, + { + "epoch": 1.79, + "learning_rate": 3.571428571428572e-05, + "loss": 0.0, + "step": 1126 + }, + { + "epoch": 1.79, + "learning_rate": 3.5696649029982365e-05, + "loss": 0.0, + "step": 1127 + }, + { + "epoch": 1.79, + "learning_rate": 3.567901234567901e-05, + "loss": 0.0, + "step": 1128 + }, + { + "epoch": 1.79, + "learning_rate": 3.566137566137566e-05, + "loss": 0.0, + "step": 1129 + }, + { + "epoch": 1.79, + "learning_rate": 3.564373897707231e-05, + "loss": 0.0, + "step": 1130 + }, + { + "epoch": 1.8, + "learning_rate": 3.562610229276896e-05, + "loss": 0.0, + "step": 1131 + }, + { + "epoch": 1.8, + "learning_rate": 3.560846560846561e-05, + "loss": 0.0, + "step": 1132 + }, + { + "epoch": 1.8, + "learning_rate": 3.5590828924162265e-05, + "loss": 0.0, + "step": 1133 + }, + { + "epoch": 1.8, + "learning_rate": 3.557319223985891e-05, + "loss": 0.0, + "step": 1134 + }, + { + "epoch": 1.8, + "learning_rate": 3.555555555555556e-05, + "loss": 0.0, + "step": 1135 + }, + { + "epoch": 1.8, + "learning_rate": 3.553791887125221e-05, + "loss": 0.0, + "step": 1136 + }, + { + "epoch": 1.8, + "learning_rate": 3.5520282186948855e-05, + "loss": 0.0, + "step": 1137 + }, + { + "epoch": 1.81, + "learning_rate": 3.55026455026455e-05, + "loss": 0.0, + "step": 1138 + }, + { + "epoch": 1.81, + "learning_rate": 3.548500881834215e-05, + "loss": 0.0, + "step": 1139 + }, + { + "epoch": 1.81, + "learning_rate": 3.54673721340388e-05, + "loss": 0.0, + "step": 1140 + }, + { + "epoch": 1.81, + "learning_rate": 3.5449735449735446e-05, + "loss": 0.0, + "step": 1141 + }, + { + "epoch": 1.81, + "learning_rate": 3.54320987654321e-05, + "loss": 0.0, + "step": 1142 + }, + { + "epoch": 1.81, + "learning_rate": 3.5414462081128755e-05, + "loss": 0.0, + "step": 1143 + }, + { + "epoch": 1.82, + "learning_rate": 3.53968253968254e-05, + "loss": 0.0, + "step": 1144 + }, + { + "epoch": 1.82, + "learning_rate": 3.537918871252205e-05, + "loss": 0.0001, + "step": 1145 + }, + { + "epoch": 1.82, + "learning_rate": 3.53615520282187e-05, + "loss": 0.0, + "step": 1146 + }, + { + "epoch": 1.82, + "learning_rate": 3.5343915343915346e-05, + "loss": 0.0, + "step": 1147 + }, + { + "epoch": 1.82, + "learning_rate": 3.5326278659611994e-05, + "loss": 0.0, + "step": 1148 + }, + { + "epoch": 1.82, + "learning_rate": 3.530864197530864e-05, + "loss": 0.0, + "step": 1149 + }, + { + "epoch": 1.83, + "learning_rate": 3.529100529100529e-05, + "loss": 0.0, + "step": 1150 + }, + { + "epoch": 1.83, + "learning_rate": 3.527336860670194e-05, + "loss": 0.0, + "step": 1151 + }, + { + "epoch": 1.83, + "learning_rate": 3.525573192239859e-05, + "loss": 0.0, + "step": 1152 + }, + { + "epoch": 1.83, + "learning_rate": 3.523809523809524e-05, + "loss": 0.0, + "step": 1153 + }, + { + "epoch": 1.83, + "learning_rate": 3.5220458553791894e-05, + "loss": 0.0, + "step": 1154 + }, + { + "epoch": 1.83, + "learning_rate": 3.520282186948854e-05, + "loss": 0.0, + "step": 1155 + }, + { + "epoch": 1.83, + "learning_rate": 3.518518518518519e-05, + "loss": 0.0, + "step": 1156 + }, + { + "epoch": 1.84, + "learning_rate": 3.516754850088184e-05, + "loss": 0.0, + "step": 1157 + }, + { + "epoch": 1.84, + "learning_rate": 3.5149911816578485e-05, + "loss": 0.0, + "step": 1158 + }, + { + "epoch": 1.84, + "learning_rate": 3.513227513227513e-05, + "loss": 0.0, + "step": 1159 + }, + { + "epoch": 1.84, + "learning_rate": 3.511463844797178e-05, + "loss": 0.0, + "step": 1160 + }, + { + "epoch": 1.84, + "learning_rate": 3.509700176366843e-05, + "loss": 0.0, + "step": 1161 + }, + { + "epoch": 1.84, + "learning_rate": 3.5079365079365075e-05, + "loss": 0.0, + "step": 1162 + }, + { + "epoch": 1.85, + "learning_rate": 3.506172839506173e-05, + "loss": 0.0, + "step": 1163 + }, + { + "epoch": 1.85, + "learning_rate": 3.5044091710758384e-05, + "loss": 0.0, + "step": 1164 + }, + { + "epoch": 1.85, + "learning_rate": 3.502645502645503e-05, + "loss": 0.0, + "step": 1165 + }, + { + "epoch": 1.85, + "learning_rate": 3.500881834215168e-05, + "loss": 0.0, + "step": 1166 + }, + { + "epoch": 1.85, + "learning_rate": 3.499118165784833e-05, + "loss": 0.0, + "step": 1167 + }, + { + "epoch": 1.85, + "learning_rate": 3.4973544973544975e-05, + "loss": 0.0, + "step": 1168 + }, + { + "epoch": 1.86, + "learning_rate": 3.495590828924162e-05, + "loss": 0.0002, + "step": 1169 + }, + { + "epoch": 1.86, + "learning_rate": 3.493827160493827e-05, + "loss": 0.0, + "step": 1170 + }, + { + "epoch": 1.86, + "learning_rate": 3.492063492063492e-05, + "loss": 0.0, + "step": 1171 + }, + { + "epoch": 1.86, + "learning_rate": 3.4902998236331566e-05, + "loss": 0.0, + "step": 1172 + }, + { + "epoch": 1.86, + "learning_rate": 3.488536155202822e-05, + "loss": 0.0, + "step": 1173 + }, + { + "epoch": 1.86, + "learning_rate": 3.486772486772487e-05, + "loss": 0.0, + "step": 1174 + }, + { + "epoch": 1.87, + "learning_rate": 3.485008818342152e-05, + "loss": 0.0, + "step": 1175 + }, + { + "epoch": 1.87, + "learning_rate": 3.483245149911817e-05, + "loss": 0.0001, + "step": 1176 + }, + { + "epoch": 1.87, + "learning_rate": 3.481481481481482e-05, + "loss": 0.0, + "step": 1177 + }, + { + "epoch": 1.87, + "learning_rate": 3.4797178130511466e-05, + "loss": 0.0, + "step": 1178 + }, + { + "epoch": 1.87, + "learning_rate": 3.4779541446208114e-05, + "loss": 0.0001, + "step": 1179 + }, + { + "epoch": 1.87, + "learning_rate": 3.476190476190476e-05, + "loss": 0.0, + "step": 1180 + }, + { + "epoch": 1.87, + "learning_rate": 3.474426807760141e-05, + "loss": 0.0, + "step": 1181 + }, + { + "epoch": 1.88, + "learning_rate": 3.472663139329806e-05, + "loss": 0.0, + "step": 1182 + }, + { + "epoch": 1.88, + "learning_rate": 3.470899470899471e-05, + "loss": 0.0, + "step": 1183 + }, + { + "epoch": 1.88, + "learning_rate": 3.469135802469136e-05, + "loss": 0.0, + "step": 1184 + }, + { + "epoch": 1.88, + "learning_rate": 3.467372134038801e-05, + "loss": 0.0, + "step": 1185 + }, + { + "epoch": 1.88, + "learning_rate": 3.465608465608466e-05, + "loss": 0.0, + "step": 1186 + }, + { + "epoch": 1.88, + "learning_rate": 3.463844797178131e-05, + "loss": 0.0, + "step": 1187 + }, + { + "epoch": 1.89, + "learning_rate": 3.462081128747796e-05, + "loss": 0.0, + "step": 1188 + }, + { + "epoch": 1.89, + "learning_rate": 3.4603174603174604e-05, + "loss": 0.0, + "step": 1189 + }, + { + "epoch": 1.89, + "learning_rate": 3.458553791887125e-05, + "loss": 0.0, + "step": 1190 + }, + { + "epoch": 1.89, + "learning_rate": 3.45679012345679e-05, + "loss": 0.0001, + "step": 1191 + }, + { + "epoch": 1.89, + "learning_rate": 3.455026455026455e-05, + "loss": 0.0, + "step": 1192 + }, + { + "epoch": 1.89, + "learning_rate": 3.45326278659612e-05, + "loss": 0.0, + "step": 1193 + }, + { + "epoch": 1.9, + "learning_rate": 3.451499118165785e-05, + "loss": 0.0, + "step": 1194 + }, + { + "epoch": 1.9, + "learning_rate": 3.44973544973545e-05, + "loss": 0.0, + "step": 1195 + }, + { + "epoch": 1.9, + "learning_rate": 3.447971781305115e-05, + "loss": 0.0001, + "step": 1196 + }, + { + "epoch": 1.9, + "learning_rate": 3.44620811287478e-05, + "loss": 0.0, + "step": 1197 + }, + { + "epoch": 1.9, + "learning_rate": 3.444444444444445e-05, + "loss": 0.0, + "step": 1198 + }, + { + "epoch": 1.9, + "learning_rate": 3.4426807760141095e-05, + "loss": 0.0003, + "step": 1199 + }, + { + "epoch": 1.9, + "learning_rate": 3.440917107583774e-05, + "loss": 0.0, + "step": 1200 + }, + { + "epoch": 1.91, + "learning_rate": 3.439153439153439e-05, + "loss": 0.0, + "step": 1201 + }, + { + "epoch": 1.91, + "learning_rate": 3.437389770723104e-05, + "loss": 0.0, + "step": 1202 + }, + { + "epoch": 1.91, + "learning_rate": 3.435626102292769e-05, + "loss": 0.0001, + "step": 1203 + }, + { + "epoch": 1.91, + "learning_rate": 3.433862433862434e-05, + "loss": 0.0001, + "step": 1204 + }, + { + "epoch": 1.91, + "learning_rate": 3.432098765432099e-05, + "loss": 0.0, + "step": 1205 + }, + { + "epoch": 1.91, + "learning_rate": 3.4303350970017636e-05, + "loss": 0.0, + "step": 1206 + }, + { + "epoch": 1.92, + "learning_rate": 3.428571428571429e-05, + "loss": 0.0, + "step": 1207 + }, + { + "epoch": 1.92, + "learning_rate": 3.426807760141094e-05, + "loss": 0.0, + "step": 1208 + }, + { + "epoch": 1.92, + "learning_rate": 3.4250440917107586e-05, + "loss": 0.0, + "step": 1209 + }, + { + "epoch": 1.92, + "learning_rate": 3.4232804232804234e-05, + "loss": 0.0, + "step": 1210 + }, + { + "epoch": 1.92, + "learning_rate": 3.421516754850088e-05, + "loss": 0.0, + "step": 1211 + }, + { + "epoch": 1.92, + "learning_rate": 3.419753086419753e-05, + "loss": 0.0, + "step": 1212 + }, + { + "epoch": 1.93, + "learning_rate": 3.4179894179894184e-05, + "loss": 0.0, + "step": 1213 + }, + { + "epoch": 1.93, + "learning_rate": 3.416225749559083e-05, + "loss": 0.0, + "step": 1214 + }, + { + "epoch": 1.93, + "learning_rate": 3.414462081128748e-05, + "loss": 0.0, + "step": 1215 + }, + { + "epoch": 1.93, + "learning_rate": 3.412698412698413e-05, + "loss": 0.0, + "step": 1216 + }, + { + "epoch": 1.93, + "learning_rate": 3.410934744268078e-05, + "loss": 0.0, + "step": 1217 + }, + { + "epoch": 1.93, + "learning_rate": 3.409171075837743e-05, + "loss": 0.0, + "step": 1218 + }, + { + "epoch": 1.93, + "learning_rate": 3.4074074074074077e-05, + "loss": 0.0, + "step": 1219 + }, + { + "epoch": 1.94, + "learning_rate": 3.4056437389770724e-05, + "loss": 0.0, + "step": 1220 + }, + { + "epoch": 1.94, + "learning_rate": 3.403880070546737e-05, + "loss": 0.0, + "step": 1221 + }, + { + "epoch": 1.94, + "learning_rate": 3.402116402116402e-05, + "loss": 0.0, + "step": 1222 + }, + { + "epoch": 1.94, + "learning_rate": 3.4003527336860674e-05, + "loss": 0.0, + "step": 1223 + }, + { + "epoch": 1.94, + "learning_rate": 3.398589065255732e-05, + "loss": 0.0, + "step": 1224 + }, + { + "epoch": 1.94, + "learning_rate": 3.396825396825397e-05, + "loss": 0.0, + "step": 1225 + }, + { + "epoch": 1.95, + "learning_rate": 3.395061728395062e-05, + "loss": 0.0, + "step": 1226 + }, + { + "epoch": 1.95, + "learning_rate": 3.3932980599647265e-05, + "loss": 0.0, + "step": 1227 + }, + { + "epoch": 1.95, + "learning_rate": 3.391534391534392e-05, + "loss": 0.0, + "step": 1228 + }, + { + "epoch": 1.95, + "learning_rate": 3.389770723104057e-05, + "loss": 0.0, + "step": 1229 + }, + { + "epoch": 1.95, + "learning_rate": 3.3880070546737215e-05, + "loss": 0.0, + "step": 1230 + }, + { + "epoch": 1.95, + "learning_rate": 3.386243386243386e-05, + "loss": 0.0, + "step": 1231 + }, + { + "epoch": 1.96, + "learning_rate": 3.384479717813051e-05, + "loss": 0.0, + "step": 1232 + }, + { + "epoch": 1.96, + "learning_rate": 3.3827160493827165e-05, + "loss": 0.0, + "step": 1233 + }, + { + "epoch": 1.96, + "learning_rate": 3.380952380952381e-05, + "loss": 0.0001, + "step": 1234 + }, + { + "epoch": 1.96, + "learning_rate": 3.379188712522046e-05, + "loss": 0.0, + "step": 1235 + }, + { + "epoch": 1.96, + "learning_rate": 3.377425044091711e-05, + "loss": 0.0, + "step": 1236 + }, + { + "epoch": 1.96, + "learning_rate": 3.3756613756613756e-05, + "loss": 0.0, + "step": 1237 + }, + { + "epoch": 1.97, + "learning_rate": 3.373897707231041e-05, + "loss": 0.0, + "step": 1238 + }, + { + "epoch": 1.97, + "learning_rate": 3.372134038800706e-05, + "loss": 0.0001, + "step": 1239 + }, + { + "epoch": 1.97, + "learning_rate": 3.3703703703703706e-05, + "loss": 0.0, + "step": 1240 + }, + { + "epoch": 1.97, + "learning_rate": 3.3686067019400353e-05, + "loss": 0.0, + "step": 1241 + }, + { + "epoch": 1.97, + "learning_rate": 3.3668430335097e-05, + "loss": 0.0, + "step": 1242 + }, + { + "epoch": 1.97, + "learning_rate": 3.3650793650793656e-05, + "loss": 0.0, + "step": 1243 + }, + { + "epoch": 1.97, + "learning_rate": 3.3633156966490303e-05, + "loss": 0.0, + "step": 1244 + }, + { + "epoch": 1.98, + "learning_rate": 3.361552028218695e-05, + "loss": 0.0, + "step": 1245 + }, + { + "epoch": 1.98, + "learning_rate": 3.35978835978836e-05, + "loss": 0.0, + "step": 1246 + }, + { + "epoch": 1.98, + "learning_rate": 3.3580246913580247e-05, + "loss": 0.0, + "step": 1247 + }, + { + "epoch": 1.98, + "learning_rate": 3.3562610229276894e-05, + "loss": 0.0, + "step": 1248 + }, + { + "epoch": 1.98, + "learning_rate": 3.354497354497355e-05, + "loss": 0.0, + "step": 1249 + }, + { + "epoch": 1.98, + "learning_rate": 3.3527336860670196e-05, + "loss": 0.0, + "step": 1250 + }, + { + "epoch": 1.99, + "learning_rate": 3.3509700176366844e-05, + "loss": 0.0, + "step": 1251 + }, + { + "epoch": 1.99, + "learning_rate": 3.349206349206349e-05, + "loss": 0.0, + "step": 1252 + }, + { + "epoch": 1.99, + "learning_rate": 3.3474426807760146e-05, + "loss": 0.0001, + "step": 1253 + }, + { + "epoch": 1.99, + "learning_rate": 3.3456790123456794e-05, + "loss": 0.0, + "step": 1254 + }, + { + "epoch": 1.99, + "learning_rate": 3.343915343915344e-05, + "loss": 0.0, + "step": 1255 + }, + { + "epoch": 1.99, + "learning_rate": 3.342151675485009e-05, + "loss": 0.0, + "step": 1256 + }, + { + "epoch": 2.0, + "learning_rate": 3.340388007054674e-05, + "loss": 0.0, + "step": 1257 + }, + { + "epoch": 2.0, + "learning_rate": 3.3386243386243385e-05, + "loss": 0.0, + "step": 1258 + }, + { + "epoch": 2.0, + "learning_rate": 3.336860670194004e-05, + "loss": 0.0, + "step": 1259 + }, + { + "epoch": 2.0, + "learning_rate": 3.335097001763669e-05, + "loss": 0.0, + "step": 1260 + }, + { + "epoch": 2.0, + "eval_accuracy": 0.9210526315789473, + "eval_auc": 0.9708333333333334, + "eval_f1": 0.7999999999999999, + "eval_loss": 0.6265161633491516, + "eval_precision": 0.8571428571428571, + "eval_recall": 0.75, + "eval_runtime": 0.2016, + "eval_samples_per_second": 188.455, + "eval_steps_per_second": 4.959, + "step": 1260 + }, + { + "epoch": 2.0, + "learning_rate": 3.3333333333333335e-05, + "loss": 0.0, + "step": 1261 + }, + { + "epoch": 2.0, + "learning_rate": 3.331569664902998e-05, + "loss": 0.0, + "step": 1262 + }, + { + "epoch": 2.0, + "learning_rate": 3.329805996472664e-05, + "loss": 0.0, + "step": 1263 + }, + { + "epoch": 2.01, + "learning_rate": 3.3280423280423285e-05, + "loss": 0.0, + "step": 1264 + }, + { + "epoch": 2.01, + "learning_rate": 3.326278659611993e-05, + "loss": 0.0, + "step": 1265 + }, + { + "epoch": 2.01, + "learning_rate": 3.324514991181658e-05, + "loss": 0.0, + "step": 1266 + }, + { + "epoch": 2.01, + "learning_rate": 3.322751322751323e-05, + "loss": 0.0, + "step": 1267 + }, + { + "epoch": 2.01, + "learning_rate": 3.3209876543209876e-05, + "loss": 0.0, + "step": 1268 + }, + { + "epoch": 2.01, + "learning_rate": 3.3192239858906523e-05, + "loss": 0.0, + "step": 1269 + }, + { + "epoch": 2.02, + "learning_rate": 3.317460317460318e-05, + "loss": 0.0, + "step": 1270 + }, + { + "epoch": 2.02, + "learning_rate": 3.3156966490299826e-05, + "loss": 0.0, + "step": 1271 + }, + { + "epoch": 2.02, + "learning_rate": 3.313932980599647e-05, + "loss": 0.0, + "step": 1272 + }, + { + "epoch": 2.02, + "learning_rate": 3.312169312169313e-05, + "loss": 0.0, + "step": 1273 + }, + { + "epoch": 2.02, + "learning_rate": 3.3104056437389776e-05, + "loss": 0.0, + "step": 1274 + }, + { + "epoch": 2.02, + "learning_rate": 3.308641975308642e-05, + "loss": 0.0, + "step": 1275 + }, + { + "epoch": 2.03, + "learning_rate": 3.306878306878307e-05, + "loss": 0.0, + "step": 1276 + }, + { + "epoch": 2.03, + "learning_rate": 3.305114638447972e-05, + "loss": 0.0, + "step": 1277 + }, + { + "epoch": 2.03, + "learning_rate": 3.3033509700176366e-05, + "loss": 0.0001, + "step": 1278 + }, + { + "epoch": 2.03, + "learning_rate": 3.3015873015873014e-05, + "loss": 0.0, + "step": 1279 + }, + { + "epoch": 2.03, + "learning_rate": 3.299823633156966e-05, + "loss": 0.0, + "step": 1280 + }, + { + "epoch": 2.03, + "learning_rate": 3.2980599647266316e-05, + "loss": 0.0, + "step": 1281 + }, + { + "epoch": 2.03, + "learning_rate": 3.2962962962962964e-05, + "loss": 0.0, + "step": 1282 + }, + { + "epoch": 2.04, + "learning_rate": 3.294532627865962e-05, + "loss": 0.0, + "step": 1283 + }, + { + "epoch": 2.04, + "learning_rate": 3.2927689594356266e-05, + "loss": 0.0, + "step": 1284 + }, + { + "epoch": 2.04, + "learning_rate": 3.2910052910052914e-05, + "loss": 0.0, + "step": 1285 + }, + { + "epoch": 2.04, + "learning_rate": 3.289241622574956e-05, + "loss": 0.0, + "step": 1286 + }, + { + "epoch": 2.04, + "learning_rate": 3.287477954144621e-05, + "loss": 0.0, + "step": 1287 + }, + { + "epoch": 2.04, + "learning_rate": 3.285714285714286e-05, + "loss": 0.0, + "step": 1288 + }, + { + "epoch": 2.05, + "learning_rate": 3.2839506172839505e-05, + "loss": 0.0, + "step": 1289 + }, + { + "epoch": 2.05, + "learning_rate": 3.282186948853615e-05, + "loss": 0.0, + "step": 1290 + }, + { + "epoch": 2.05, + "learning_rate": 3.280423280423281e-05, + "loss": 0.0, + "step": 1291 + }, + { + "epoch": 2.05, + "learning_rate": 3.2786596119929455e-05, + "loss": 0.0, + "step": 1292 + }, + { + "epoch": 2.05, + "learning_rate": 3.276895943562611e-05, + "loss": 0.0, + "step": 1293 + }, + { + "epoch": 2.05, + "learning_rate": 3.275132275132276e-05, + "loss": 0.0, + "step": 1294 + }, + { + "epoch": 2.06, + "learning_rate": 3.2733686067019405e-05, + "loss": 0.0, + "step": 1295 + }, + { + "epoch": 2.06, + "learning_rate": 3.271604938271605e-05, + "loss": 0.0, + "step": 1296 + }, + { + "epoch": 2.06, + "learning_rate": 3.26984126984127e-05, + "loss": 0.0002, + "step": 1297 + }, + { + "epoch": 2.06, + "learning_rate": 3.268077601410935e-05, + "loss": 0.0, + "step": 1298 + }, + { + "epoch": 2.06, + "learning_rate": 3.2663139329805996e-05, + "loss": 0.0001, + "step": 1299 + }, + { + "epoch": 2.06, + "learning_rate": 3.264550264550264e-05, + "loss": 0.0, + "step": 1300 + }, + { + "epoch": 2.07, + "learning_rate": 3.262786596119929e-05, + "loss": 0.0, + "step": 1301 + }, + { + "epoch": 2.07, + "learning_rate": 3.2610229276895946e-05, + "loss": 0.0, + "step": 1302 + }, + { + "epoch": 2.07, + "learning_rate": 3.25925925925926e-05, + "loss": 0.0, + "step": 1303 + }, + { + "epoch": 2.07, + "learning_rate": 3.257495590828925e-05, + "loss": 0.0, + "step": 1304 + }, + { + "epoch": 2.07, + "learning_rate": 3.2557319223985895e-05, + "loss": 0.0, + "step": 1305 + }, + { + "epoch": 2.07, + "learning_rate": 3.253968253968254e-05, + "loss": 0.0, + "step": 1306 + }, + { + "epoch": 2.07, + "learning_rate": 3.252204585537919e-05, + "loss": 0.0, + "step": 1307 + }, + { + "epoch": 2.08, + "learning_rate": 3.250440917107584e-05, + "loss": 0.0, + "step": 1308 + }, + { + "epoch": 2.08, + "learning_rate": 3.2486772486772486e-05, + "loss": 0.0, + "step": 1309 + }, + { + "epoch": 2.08, + "learning_rate": 3.2469135802469134e-05, + "loss": 0.0, + "step": 1310 + }, + { + "epoch": 2.08, + "learning_rate": 3.245149911816578e-05, + "loss": 0.0, + "step": 1311 + }, + { + "epoch": 2.08, + "learning_rate": 3.2433862433862436e-05, + "loss": 0.0, + "step": 1312 + }, + { + "epoch": 2.08, + "learning_rate": 3.2416225749559084e-05, + "loss": 0.0001, + "step": 1313 + }, + { + "epoch": 2.09, + "learning_rate": 3.239858906525574e-05, + "loss": 0.0, + "step": 1314 + }, + { + "epoch": 2.09, + "learning_rate": 3.2380952380952386e-05, + "loss": 0.0, + "step": 1315 + }, + { + "epoch": 2.09, + "learning_rate": 3.2363315696649034e-05, + "loss": 0.0, + "step": 1316 + }, + { + "epoch": 2.09, + "learning_rate": 3.234567901234568e-05, + "loss": 0.0, + "step": 1317 + }, + { + "epoch": 2.09, + "learning_rate": 3.232804232804233e-05, + "loss": 0.0001, + "step": 1318 + }, + { + "epoch": 2.09, + "learning_rate": 3.231040564373898e-05, + "loss": 0.0, + "step": 1319 + }, + { + "epoch": 2.1, + "learning_rate": 3.2292768959435625e-05, + "loss": 0.0, + "step": 1320 + }, + { + "epoch": 2.1, + "learning_rate": 3.227513227513227e-05, + "loss": 0.0, + "step": 1321 + }, + { + "epoch": 2.1, + "learning_rate": 3.225749559082892e-05, + "loss": 0.0, + "step": 1322 + }, + { + "epoch": 2.1, + "learning_rate": 3.2239858906525575e-05, + "loss": 0.0, + "step": 1323 + }, + { + "epoch": 2.1, + "learning_rate": 3.222222222222223e-05, + "loss": 0.0, + "step": 1324 + }, + { + "epoch": 2.1, + "learning_rate": 3.220458553791888e-05, + "loss": 0.0, + "step": 1325 + }, + { + "epoch": 2.1, + "learning_rate": 3.2186948853615525e-05, + "loss": 0.0, + "step": 1326 + }, + { + "epoch": 2.11, + "learning_rate": 3.216931216931217e-05, + "loss": 0.0, + "step": 1327 + }, + { + "epoch": 2.11, + "learning_rate": 3.215167548500882e-05, + "loss": 0.0, + "step": 1328 + }, + { + "epoch": 2.11, + "learning_rate": 3.213403880070547e-05, + "loss": 0.0, + "step": 1329 + }, + { + "epoch": 2.11, + "learning_rate": 3.2116402116402115e-05, + "loss": 0.0, + "step": 1330 + }, + { + "epoch": 2.11, + "learning_rate": 3.209876543209876e-05, + "loss": 0.0, + "step": 1331 + }, + { + "epoch": 2.11, + "learning_rate": 3.208112874779541e-05, + "loss": 0.0, + "step": 1332 + }, + { + "epoch": 2.12, + "learning_rate": 3.2063492063492065e-05, + "loss": 0.0, + "step": 1333 + }, + { + "epoch": 2.12, + "learning_rate": 3.204585537918871e-05, + "loss": 0.0, + "step": 1334 + }, + { + "epoch": 2.12, + "learning_rate": 3.202821869488537e-05, + "loss": 0.0001, + "step": 1335 + }, + { + "epoch": 2.12, + "learning_rate": 3.2010582010582015e-05, + "loss": 0.0001, + "step": 1336 + }, + { + "epoch": 2.12, + "learning_rate": 3.199294532627866e-05, + "loss": 0.0, + "step": 1337 + }, + { + "epoch": 2.12, + "learning_rate": 3.197530864197531e-05, + "loss": 0.0, + "step": 1338 + }, + { + "epoch": 2.13, + "learning_rate": 3.195767195767196e-05, + "loss": 0.0001, + "step": 1339 + }, + { + "epoch": 2.13, + "learning_rate": 3.1940035273368606e-05, + "loss": 0.0001, + "step": 1340 + }, + { + "epoch": 2.13, + "learning_rate": 3.1922398589065254e-05, + "loss": 0.0, + "step": 1341 + }, + { + "epoch": 2.13, + "learning_rate": 3.19047619047619e-05, + "loss": 0.0, + "step": 1342 + }, + { + "epoch": 2.13, + "learning_rate": 3.1887125220458556e-05, + "loss": 0.0, + "step": 1343 + }, + { + "epoch": 2.13, + "learning_rate": 3.1869488536155204e-05, + "loss": 0.0, + "step": 1344 + }, + { + "epoch": 2.13, + "learning_rate": 3.185185185185185e-05, + "loss": 0.0, + "step": 1345 + }, + { + "epoch": 2.14, + "learning_rate": 3.1834215167548506e-05, + "loss": 0.0, + "step": 1346 + }, + { + "epoch": 2.14, + "learning_rate": 3.1816578483245154e-05, + "loss": 0.0, + "step": 1347 + }, + { + "epoch": 2.14, + "learning_rate": 3.17989417989418e-05, + "loss": 0.0, + "step": 1348 + }, + { + "epoch": 2.14, + "learning_rate": 3.178130511463845e-05, + "loss": 0.0, + "step": 1349 + }, + { + "epoch": 2.14, + "learning_rate": 3.17636684303351e-05, + "loss": 0.0, + "step": 1350 + }, + { + "epoch": 2.14, + "learning_rate": 3.1746031746031745e-05, + "loss": 0.0, + "step": 1351 + }, + { + "epoch": 2.15, + "learning_rate": 3.172839506172839e-05, + "loss": 0.0, + "step": 1352 + }, + { + "epoch": 2.15, + "learning_rate": 3.171075837742505e-05, + "loss": 0.0, + "step": 1353 + }, + { + "epoch": 2.15, + "learning_rate": 3.1693121693121695e-05, + "loss": 0.0, + "step": 1354 + }, + { + "epoch": 2.15, + "learning_rate": 3.167548500881834e-05, + "loss": 0.0, + "step": 1355 + }, + { + "epoch": 2.15, + "learning_rate": 3.1657848324515e-05, + "loss": 0.0, + "step": 1356 + }, + { + "epoch": 2.15, + "learning_rate": 3.1640211640211645e-05, + "loss": 0.0, + "step": 1357 + }, + { + "epoch": 2.16, + "learning_rate": 3.162257495590829e-05, + "loss": 0.0, + "step": 1358 + }, + { + "epoch": 2.16, + "learning_rate": 3.160493827160494e-05, + "loss": 0.0, + "step": 1359 + }, + { + "epoch": 2.16, + "learning_rate": 3.158730158730159e-05, + "loss": 0.0, + "step": 1360 + }, + { + "epoch": 2.16, + "learning_rate": 3.1569664902998235e-05, + "loss": 0.0, + "step": 1361 + }, + { + "epoch": 2.16, + "learning_rate": 3.155202821869488e-05, + "loss": 0.0, + "step": 1362 + }, + { + "epoch": 2.16, + "learning_rate": 3.153439153439154e-05, + "loss": 0.0, + "step": 1363 + }, + { + "epoch": 2.17, + "learning_rate": 3.1516754850088185e-05, + "loss": 0.0001, + "step": 1364 + }, + { + "epoch": 2.17, + "learning_rate": 3.149911816578483e-05, + "loss": 0.0, + "step": 1365 + }, + { + "epoch": 2.17, + "learning_rate": 3.148148148148148e-05, + "loss": 0.0001, + "step": 1366 + }, + { + "epoch": 2.17, + "learning_rate": 3.1463844797178135e-05, + "loss": 0.0, + "step": 1367 + }, + { + "epoch": 2.17, + "learning_rate": 3.144620811287478e-05, + "loss": 0.0, + "step": 1368 + }, + { + "epoch": 2.17, + "learning_rate": 3.142857142857143e-05, + "loss": 0.0, + "step": 1369 + }, + { + "epoch": 2.17, + "learning_rate": 3.141093474426808e-05, + "loss": 0.0, + "step": 1370 + }, + { + "epoch": 2.18, + "learning_rate": 3.1393298059964726e-05, + "loss": 0.0, + "step": 1371 + }, + { + "epoch": 2.18, + "learning_rate": 3.1375661375661374e-05, + "loss": 0.0, + "step": 1372 + }, + { + "epoch": 2.18, + "learning_rate": 3.135802469135803e-05, + "loss": 0.0001, + "step": 1373 + }, + { + "epoch": 2.18, + "learning_rate": 3.1340388007054676e-05, + "loss": 0.0, + "step": 1374 + }, + { + "epoch": 2.18, + "learning_rate": 3.1322751322751324e-05, + "loss": 0.0, + "step": 1375 + }, + { + "epoch": 2.18, + "learning_rate": 3.130511463844797e-05, + "loss": 0.0, + "step": 1376 + }, + { + "epoch": 2.19, + "learning_rate": 3.1287477954144626e-05, + "loss": 0.0, + "step": 1377 + }, + { + "epoch": 2.19, + "learning_rate": 3.1269841269841274e-05, + "loss": 0.0, + "step": 1378 + }, + { + "epoch": 2.19, + "learning_rate": 3.125220458553792e-05, + "loss": 0.0, + "step": 1379 + }, + { + "epoch": 2.19, + "learning_rate": 3.123456790123457e-05, + "loss": 0.0001, + "step": 1380 + }, + { + "epoch": 2.19, + "learning_rate": 3.121693121693122e-05, + "loss": 0.0, + "step": 1381 + }, + { + "epoch": 2.19, + "learning_rate": 3.1199294532627865e-05, + "loss": 0.0, + "step": 1382 + }, + { + "epoch": 2.2, + "learning_rate": 3.118165784832452e-05, + "loss": 0.0, + "step": 1383 + }, + { + "epoch": 2.2, + "learning_rate": 3.116402116402117e-05, + "loss": 0.0, + "step": 1384 + }, + { + "epoch": 2.2, + "learning_rate": 3.1146384479717814e-05, + "loss": 0.0, + "step": 1385 + }, + { + "epoch": 2.2, + "learning_rate": 3.112874779541446e-05, + "loss": 0.0, + "step": 1386 + }, + { + "epoch": 2.2, + "learning_rate": 3.111111111111111e-05, + "loss": 0.0, + "step": 1387 + }, + { + "epoch": 2.2, + "learning_rate": 3.1093474426807764e-05, + "loss": 0.0, + "step": 1388 + }, + { + "epoch": 2.2, + "learning_rate": 3.107583774250441e-05, + "loss": 0.0, + "step": 1389 + }, + { + "epoch": 2.21, + "learning_rate": 3.105820105820106e-05, + "loss": 0.0, + "step": 1390 + }, + { + "epoch": 2.21, + "learning_rate": 3.104056437389771e-05, + "loss": 0.0, + "step": 1391 + }, + { + "epoch": 2.21, + "learning_rate": 3.1022927689594355e-05, + "loss": 0.0, + "step": 1392 + }, + { + "epoch": 2.21, + "learning_rate": 3.100529100529101e-05, + "loss": 0.0, + "step": 1393 + }, + { + "epoch": 2.21, + "learning_rate": 3.098765432098766e-05, + "loss": 0.0, + "step": 1394 + }, + { + "epoch": 2.21, + "learning_rate": 3.0970017636684305e-05, + "loss": 0.0, + "step": 1395 + }, + { + "epoch": 2.22, + "learning_rate": 3.095238095238095e-05, + "loss": 0.0, + "step": 1396 + }, + { + "epoch": 2.22, + "learning_rate": 3.09347442680776e-05, + "loss": 0.0, + "step": 1397 + }, + { + "epoch": 2.22, + "learning_rate": 3.0917107583774255e-05, + "loss": 0.0, + "step": 1398 + }, + { + "epoch": 2.22, + "learning_rate": 3.08994708994709e-05, + "loss": 0.0, + "step": 1399 + }, + { + "epoch": 2.22, + "learning_rate": 3.088183421516755e-05, + "loss": 0.0, + "step": 1400 + }, + { + "epoch": 2.22, + "learning_rate": 3.08641975308642e-05, + "loss": 0.0, + "step": 1401 + }, + { + "epoch": 2.23, + "learning_rate": 3.0846560846560846e-05, + "loss": 0.0, + "step": 1402 + }, + { + "epoch": 2.23, + "learning_rate": 3.08289241622575e-05, + "loss": 0.0, + "step": 1403 + }, + { + "epoch": 2.23, + "learning_rate": 3.081128747795415e-05, + "loss": 0.0, + "step": 1404 + }, + { + "epoch": 2.23, + "learning_rate": 3.0793650793650796e-05, + "loss": 0.0001, + "step": 1405 + }, + { + "epoch": 2.23, + "learning_rate": 3.0776014109347444e-05, + "loss": 0.0001, + "step": 1406 + }, + { + "epoch": 2.23, + "learning_rate": 3.075837742504409e-05, + "loss": 0.0, + "step": 1407 + }, + { + "epoch": 2.23, + "learning_rate": 3.074074074074074e-05, + "loss": 0.0, + "step": 1408 + }, + { + "epoch": 2.24, + "learning_rate": 3.0723104056437394e-05, + "loss": 0.0, + "step": 1409 + }, + { + "epoch": 2.24, + "learning_rate": 3.070546737213404e-05, + "loss": 0.0, + "step": 1410 + }, + { + "epoch": 2.24, + "learning_rate": 3.068783068783069e-05, + "loss": 0.0, + "step": 1411 + }, + { + "epoch": 2.24, + "learning_rate": 3.067019400352734e-05, + "loss": 0.0, + "step": 1412 + }, + { + "epoch": 2.24, + "learning_rate": 3.065255731922399e-05, + "loss": 0.0, + "step": 1413 + }, + { + "epoch": 2.24, + "learning_rate": 3.063492063492064e-05, + "loss": 0.0001, + "step": 1414 + }, + { + "epoch": 2.25, + "learning_rate": 3.061728395061729e-05, + "loss": 0.0, + "step": 1415 + }, + { + "epoch": 2.25, + "learning_rate": 3.0599647266313934e-05, + "loss": 0.0, + "step": 1416 + }, + { + "epoch": 2.25, + "learning_rate": 3.058201058201058e-05, + "loss": 0.0, + "step": 1417 + }, + { + "epoch": 2.25, + "learning_rate": 3.056437389770723e-05, + "loss": 0.0, + "step": 1418 + }, + { + "epoch": 2.25, + "learning_rate": 3.0546737213403884e-05, + "loss": 0.0, + "step": 1419 + }, + { + "epoch": 2.25, + "learning_rate": 3.052910052910053e-05, + "loss": 0.0, + "step": 1420 + }, + { + "epoch": 2.26, + "learning_rate": 3.0511463844797176e-05, + "loss": 0.0, + "step": 1421 + }, + { + "epoch": 2.26, + "learning_rate": 3.0493827160493827e-05, + "loss": 0.0, + "step": 1422 + }, + { + "epoch": 2.26, + "learning_rate": 3.0476190476190482e-05, + "loss": 0.0, + "step": 1423 + }, + { + "epoch": 2.26, + "learning_rate": 3.045855379188713e-05, + "loss": 0.0, + "step": 1424 + }, + { + "epoch": 2.26, + "learning_rate": 3.0440917107583777e-05, + "loss": 0.0001, + "step": 1425 + }, + { + "epoch": 2.26, + "learning_rate": 3.0423280423280425e-05, + "loss": 0.0, + "step": 1426 + }, + { + "epoch": 2.27, + "learning_rate": 3.0405643738977073e-05, + "loss": 0.0, + "step": 1427 + }, + { + "epoch": 2.27, + "learning_rate": 3.0388007054673724e-05, + "loss": 0.0, + "step": 1428 + }, + { + "epoch": 2.27, + "learning_rate": 3.037037037037037e-05, + "loss": 0.0, + "step": 1429 + }, + { + "epoch": 2.27, + "learning_rate": 3.035273368606702e-05, + "loss": 0.0, + "step": 1430 + }, + { + "epoch": 2.27, + "learning_rate": 3.0335097001763667e-05, + "loss": 0.0, + "step": 1431 + }, + { + "epoch": 2.27, + "learning_rate": 3.0317460317460318e-05, + "loss": 0.0001, + "step": 1432 + }, + { + "epoch": 2.27, + "learning_rate": 3.029982363315697e-05, + "loss": 0.0, + "step": 1433 + }, + { + "epoch": 2.28, + "learning_rate": 3.028218694885362e-05, + "loss": 0.0, + "step": 1434 + }, + { + "epoch": 2.28, + "learning_rate": 3.0264550264550268e-05, + "loss": 0.0, + "step": 1435 + }, + { + "epoch": 2.28, + "learning_rate": 3.0246913580246916e-05, + "loss": 0.0, + "step": 1436 + }, + { + "epoch": 2.28, + "learning_rate": 3.0229276895943563e-05, + "loss": 0.0, + "step": 1437 + }, + { + "epoch": 2.28, + "learning_rate": 3.0211640211640215e-05, + "loss": 0.0, + "step": 1438 + }, + { + "epoch": 2.28, + "learning_rate": 3.0194003527336862e-05, + "loss": 0.0001, + "step": 1439 + }, + { + "epoch": 2.29, + "learning_rate": 3.017636684303351e-05, + "loss": 0.0, + "step": 1440 + }, + { + "epoch": 2.29, + "learning_rate": 3.0158730158730158e-05, + "loss": 0.0, + "step": 1441 + }, + { + "epoch": 2.29, + "learning_rate": 3.0141093474426805e-05, + "loss": 0.0, + "step": 1442 + }, + { + "epoch": 2.29, + "learning_rate": 3.012345679012346e-05, + "loss": 0.0, + "step": 1443 + }, + { + "epoch": 2.29, + "learning_rate": 3.010582010582011e-05, + "loss": 0.0, + "step": 1444 + }, + { + "epoch": 2.29, + "learning_rate": 3.008818342151676e-05, + "loss": 0.0, + "step": 1445 + }, + { + "epoch": 2.3, + "learning_rate": 3.0070546737213407e-05, + "loss": 0.0, + "step": 1446 + }, + { + "epoch": 2.3, + "learning_rate": 3.0052910052910054e-05, + "loss": 0.0, + "step": 1447 + }, + { + "epoch": 2.3, + "learning_rate": 3.0035273368606702e-05, + "loss": 0.0, + "step": 1448 + }, + { + "epoch": 2.3, + "learning_rate": 3.0017636684303353e-05, + "loss": 0.0, + "step": 1449 + }, + { + "epoch": 2.3, + "learning_rate": 3e-05, + "loss": 0.0, + "step": 1450 + }, + { + "epoch": 2.3, + "learning_rate": 2.998236331569665e-05, + "loss": 0.0, + "step": 1451 + }, + { + "epoch": 2.3, + "learning_rate": 2.9964726631393296e-05, + "loss": 0.0, + "step": 1452 + }, + { + "epoch": 2.31, + "learning_rate": 2.9947089947089947e-05, + "loss": 0.0, + "step": 1453 + }, + { + "epoch": 2.31, + "learning_rate": 2.99294532627866e-05, + "loss": 0.0, + "step": 1454 + }, + { + "epoch": 2.31, + "learning_rate": 2.991181657848325e-05, + "loss": 0.0, + "step": 1455 + }, + { + "epoch": 2.31, + "learning_rate": 2.9894179894179897e-05, + "loss": 0.0, + "step": 1456 + }, + { + "epoch": 2.31, + "learning_rate": 2.9876543209876545e-05, + "loss": 0.0, + "step": 1457 + }, + { + "epoch": 2.31, + "learning_rate": 2.9858906525573193e-05, + "loss": 0.0, + "step": 1458 + }, + { + "epoch": 2.32, + "learning_rate": 2.9841269841269844e-05, + "loss": 0.0, + "step": 1459 + }, + { + "epoch": 2.32, + "learning_rate": 2.982363315696649e-05, + "loss": 0.0001, + "step": 1460 + }, + { + "epoch": 2.32, + "learning_rate": 2.980599647266314e-05, + "loss": 0.0, + "step": 1461 + }, + { + "epoch": 2.32, + "learning_rate": 2.9788359788359787e-05, + "loss": 0.0, + "step": 1462 + }, + { + "epoch": 2.32, + "learning_rate": 2.9770723104056435e-05, + "loss": 0.0, + "step": 1463 + }, + { + "epoch": 2.32, + "learning_rate": 2.975308641975309e-05, + "loss": 0.0, + "step": 1464 + }, + { + "epoch": 2.33, + "learning_rate": 2.973544973544974e-05, + "loss": 0.0, + "step": 1465 + }, + { + "epoch": 2.33, + "learning_rate": 2.9717813051146388e-05, + "loss": 0.0, + "step": 1466 + }, + { + "epoch": 2.33, + "learning_rate": 2.9700176366843036e-05, + "loss": 0.0, + "step": 1467 + }, + { + "epoch": 2.33, + "learning_rate": 2.9682539682539683e-05, + "loss": 0.0, + "step": 1468 + }, + { + "epoch": 2.33, + "learning_rate": 2.966490299823633e-05, + "loss": 0.0, + "step": 1469 + }, + { + "epoch": 2.33, + "learning_rate": 2.9647266313932982e-05, + "loss": 0.0, + "step": 1470 + }, + { + "epoch": 2.33, + "learning_rate": 2.962962962962963e-05, + "loss": 0.0, + "step": 1471 + }, + { + "epoch": 2.34, + "learning_rate": 2.9611992945326278e-05, + "loss": 0.0001, + "step": 1472 + }, + { + "epoch": 2.34, + "learning_rate": 2.9594356261022925e-05, + "loss": 0.0, + "step": 1473 + }, + { + "epoch": 2.34, + "learning_rate": 2.957671957671958e-05, + "loss": 0.0, + "step": 1474 + }, + { + "epoch": 2.34, + "learning_rate": 2.9559082892416228e-05, + "loss": 0.0, + "step": 1475 + }, + { + "epoch": 2.34, + "learning_rate": 2.954144620811288e-05, + "loss": 0.0, + "step": 1476 + }, + { + "epoch": 2.34, + "learning_rate": 2.9523809523809526e-05, + "loss": 0.0, + "step": 1477 + }, + { + "epoch": 2.35, + "learning_rate": 2.9506172839506174e-05, + "loss": 0.0, + "step": 1478 + }, + { + "epoch": 2.35, + "learning_rate": 2.9488536155202822e-05, + "loss": 0.0, + "step": 1479 + }, + { + "epoch": 2.35, + "learning_rate": 2.9470899470899473e-05, + "loss": 0.0, + "step": 1480 + }, + { + "epoch": 2.35, + "learning_rate": 2.945326278659612e-05, + "loss": 0.0, + "step": 1481 + }, + { + "epoch": 2.35, + "learning_rate": 2.943562610229277e-05, + "loss": 0.0, + "step": 1482 + }, + { + "epoch": 2.35, + "learning_rate": 2.9417989417989416e-05, + "loss": 0.0, + "step": 1483 + }, + { + "epoch": 2.36, + "learning_rate": 2.940035273368607e-05, + "loss": 0.0, + "step": 1484 + }, + { + "epoch": 2.36, + "learning_rate": 2.9382716049382718e-05, + "loss": 0.0, + "step": 1485 + }, + { + "epoch": 2.36, + "learning_rate": 2.9365079365079366e-05, + "loss": 0.0, + "step": 1486 + }, + { + "epoch": 2.36, + "learning_rate": 2.9347442680776017e-05, + "loss": 0.0, + "step": 1487 + }, + { + "epoch": 2.36, + "learning_rate": 2.9329805996472665e-05, + "loss": 0.0, + "step": 1488 + }, + { + "epoch": 2.36, + "learning_rate": 2.9312169312169313e-05, + "loss": 0.0, + "step": 1489 + }, + { + "epoch": 2.37, + "learning_rate": 2.929453262786596e-05, + "loss": 0.0, + "step": 1490 + }, + { + "epoch": 2.37, + "learning_rate": 2.927689594356261e-05, + "loss": 0.0, + "step": 1491 + }, + { + "epoch": 2.37, + "learning_rate": 2.925925925925926e-05, + "loss": 0.0, + "step": 1492 + }, + { + "epoch": 2.37, + "learning_rate": 2.9241622574955907e-05, + "loss": 0.0, + "step": 1493 + }, + { + "epoch": 2.37, + "learning_rate": 2.922398589065256e-05, + "loss": 0.0, + "step": 1494 + }, + { + "epoch": 2.37, + "learning_rate": 2.920634920634921e-05, + "loss": 0.0001, + "step": 1495 + }, + { + "epoch": 2.37, + "learning_rate": 2.9188712522045857e-05, + "loss": 0.0, + "step": 1496 + }, + { + "epoch": 2.38, + "learning_rate": 2.9171075837742508e-05, + "loss": 0.0, + "step": 1497 + }, + { + "epoch": 2.38, + "learning_rate": 2.9153439153439156e-05, + "loss": 0.0, + "step": 1498 + }, + { + "epoch": 2.38, + "learning_rate": 2.9135802469135803e-05, + "loss": 0.0, + "step": 1499 + }, + { + "epoch": 2.38, + "learning_rate": 2.911816578483245e-05, + "loss": 0.0, + "step": 1500 + }, + { + "epoch": 2.38, + "learning_rate": 2.91005291005291e-05, + "loss": 0.0, + "step": 1501 + }, + { + "epoch": 2.38, + "learning_rate": 2.908289241622575e-05, + "loss": 0.0, + "step": 1502 + }, + { + "epoch": 2.39, + "learning_rate": 2.9065255731922398e-05, + "loss": 0.0, + "step": 1503 + }, + { + "epoch": 2.39, + "learning_rate": 2.9047619047619052e-05, + "loss": 0.0, + "step": 1504 + }, + { + "epoch": 2.39, + "learning_rate": 2.90299823633157e-05, + "loss": 0.0, + "step": 1505 + }, + { + "epoch": 2.39, + "learning_rate": 2.9012345679012347e-05, + "loss": 0.0, + "step": 1506 + }, + { + "epoch": 2.39, + "learning_rate": 2.8994708994708995e-05, + "loss": 0.0, + "step": 1507 + }, + { + "epoch": 2.39, + "learning_rate": 2.8977072310405646e-05, + "loss": 0.0, + "step": 1508 + }, + { + "epoch": 2.4, + "learning_rate": 2.8959435626102294e-05, + "loss": 0.0, + "step": 1509 + }, + { + "epoch": 2.4, + "learning_rate": 2.8941798941798942e-05, + "loss": 0.0, + "step": 1510 + }, + { + "epoch": 2.4, + "learning_rate": 2.892416225749559e-05, + "loss": 0.0, + "step": 1511 + }, + { + "epoch": 2.4, + "learning_rate": 2.890652557319224e-05, + "loss": 0.0, + "step": 1512 + }, + { + "epoch": 2.4, + "learning_rate": 2.8888888888888888e-05, + "loss": 0.0, + "step": 1513 + }, + { + "epoch": 2.4, + "learning_rate": 2.8871252204585543e-05, + "loss": 0.0, + "step": 1514 + }, + { + "epoch": 2.4, + "learning_rate": 2.885361552028219e-05, + "loss": 0.0, + "step": 1515 + }, + { + "epoch": 2.41, + "learning_rate": 2.8835978835978838e-05, + "loss": 0.0, + "step": 1516 + }, + { + "epoch": 2.41, + "learning_rate": 2.8818342151675486e-05, + "loss": 0.0, + "step": 1517 + }, + { + "epoch": 2.41, + "learning_rate": 2.8800705467372137e-05, + "loss": 0.0, + "step": 1518 + }, + { + "epoch": 2.41, + "learning_rate": 2.8783068783068785e-05, + "loss": 0.0, + "step": 1519 + }, + { + "epoch": 2.41, + "learning_rate": 2.8765432098765432e-05, + "loss": 0.0, + "step": 1520 + }, + { + "epoch": 2.41, + "learning_rate": 2.874779541446208e-05, + "loss": 0.0, + "step": 1521 + }, + { + "epoch": 2.42, + "learning_rate": 2.8730158730158728e-05, + "loss": 0.0, + "step": 1522 + }, + { + "epoch": 2.42, + "learning_rate": 2.871252204585538e-05, + "loss": 0.0, + "step": 1523 + }, + { + "epoch": 2.42, + "learning_rate": 2.8694885361552033e-05, + "loss": 0.0, + "step": 1524 + }, + { + "epoch": 2.42, + "learning_rate": 2.867724867724868e-05, + "loss": 0.0, + "step": 1525 + }, + { + "epoch": 2.42, + "learning_rate": 2.865961199294533e-05, + "loss": 0.0, + "step": 1526 + }, + { + "epoch": 2.42, + "learning_rate": 2.8641975308641977e-05, + "loss": 0.0, + "step": 1527 + }, + { + "epoch": 2.43, + "learning_rate": 2.8624338624338624e-05, + "loss": 0.0, + "step": 1528 + }, + { + "epoch": 2.43, + "learning_rate": 2.8606701940035275e-05, + "loss": 0.0, + "step": 1529 + }, + { + "epoch": 2.43, + "learning_rate": 2.8589065255731923e-05, + "loss": 0.0, + "step": 1530 + }, + { + "epoch": 2.43, + "learning_rate": 2.857142857142857e-05, + "loss": 0.0, + "step": 1531 + }, + { + "epoch": 2.43, + "learning_rate": 2.855379188712522e-05, + "loss": 0.0, + "step": 1532 + }, + { + "epoch": 2.43, + "learning_rate": 2.853615520282187e-05, + "loss": 0.0, + "step": 1533 + }, + { + "epoch": 2.43, + "learning_rate": 2.851851851851852e-05, + "loss": 0.0, + "step": 1534 + }, + { + "epoch": 2.44, + "learning_rate": 2.8500881834215172e-05, + "loss": 0.0, + "step": 1535 + }, + { + "epoch": 2.44, + "learning_rate": 2.848324514991182e-05, + "loss": 0.0, + "step": 1536 + }, + { + "epoch": 2.44, + "learning_rate": 2.8465608465608467e-05, + "loss": 0.0, + "step": 1537 + }, + { + "epoch": 2.44, + "learning_rate": 2.8447971781305115e-05, + "loss": 0.0, + "step": 1538 + }, + { + "epoch": 2.44, + "learning_rate": 2.8430335097001766e-05, + "loss": 0.0, + "step": 1539 + }, + { + "epoch": 2.44, + "learning_rate": 2.8412698412698414e-05, + "loss": 0.0, + "step": 1540 + }, + { + "epoch": 2.45, + "learning_rate": 2.839506172839506e-05, + "loss": 0.0, + "step": 1541 + }, + { + "epoch": 2.45, + "learning_rate": 2.837742504409171e-05, + "loss": 0.0, + "step": 1542 + }, + { + "epoch": 2.45, + "learning_rate": 2.8359788359788357e-05, + "loss": 0.0, + "step": 1543 + }, + { + "epoch": 2.45, + "learning_rate": 2.834215167548501e-05, + "loss": 0.0, + "step": 1544 + }, + { + "epoch": 2.45, + "learning_rate": 2.8324514991181663e-05, + "loss": 0.0, + "step": 1545 + }, + { + "epoch": 2.45, + "learning_rate": 2.830687830687831e-05, + "loss": 0.0, + "step": 1546 + }, + { + "epoch": 2.46, + "learning_rate": 2.8289241622574958e-05, + "loss": 0.0, + "step": 1547 + }, + { + "epoch": 2.46, + "learning_rate": 2.8271604938271606e-05, + "loss": 0.0, + "step": 1548 + }, + { + "epoch": 2.46, + "learning_rate": 2.8253968253968253e-05, + "loss": 0.0, + "step": 1549 + }, + { + "epoch": 2.46, + "learning_rate": 2.8236331569664905e-05, + "loss": 0.0, + "step": 1550 + }, + { + "epoch": 2.46, + "learning_rate": 2.8218694885361552e-05, + "loss": 0.0, + "step": 1551 + }, + { + "epoch": 2.46, + "learning_rate": 2.82010582010582e-05, + "loss": 0.0, + "step": 1552 + }, + { + "epoch": 2.47, + "learning_rate": 2.8183421516754848e-05, + "loss": 0.0, + "step": 1553 + }, + { + "epoch": 2.47, + "learning_rate": 2.8165784832451502e-05, + "loss": 0.0, + "step": 1554 + }, + { + "epoch": 2.47, + "learning_rate": 2.814814814814815e-05, + "loss": 0.0, + "step": 1555 + }, + { + "epoch": 2.47, + "learning_rate": 2.81305114638448e-05, + "loss": 0.0, + "step": 1556 + }, + { + "epoch": 2.47, + "learning_rate": 2.811287477954145e-05, + "loss": 0.0001, + "step": 1557 + }, + { + "epoch": 2.47, + "learning_rate": 2.8095238095238096e-05, + "loss": 0.0, + "step": 1558 + }, + { + "epoch": 2.47, + "learning_rate": 2.8077601410934744e-05, + "loss": 0.0, + "step": 1559 + }, + { + "epoch": 2.48, + "learning_rate": 2.8059964726631395e-05, + "loss": 0.0001, + "step": 1560 + }, + { + "epoch": 2.48, + "learning_rate": 2.8042328042328043e-05, + "loss": 0.0, + "step": 1561 + }, + { + "epoch": 2.48, + "learning_rate": 2.802469135802469e-05, + "loss": 0.0, + "step": 1562 + }, + { + "epoch": 2.48, + "learning_rate": 2.800705467372134e-05, + "loss": 0.0, + "step": 1563 + }, + { + "epoch": 2.48, + "learning_rate": 2.7989417989417993e-05, + "loss": 0.0, + "step": 1564 + }, + { + "epoch": 2.48, + "learning_rate": 2.797178130511464e-05, + "loss": 0.0, + "step": 1565 + }, + { + "epoch": 2.49, + "learning_rate": 2.795414462081129e-05, + "loss": 0.0, + "step": 1566 + }, + { + "epoch": 2.49, + "learning_rate": 2.793650793650794e-05, + "loss": 0.0, + "step": 1567 + }, + { + "epoch": 2.49, + "learning_rate": 2.7918871252204587e-05, + "loss": 0.0, + "step": 1568 + }, + { + "epoch": 2.49, + "learning_rate": 2.7901234567901235e-05, + "loss": 0.0, + "step": 1569 + }, + { + "epoch": 2.49, + "learning_rate": 2.7883597883597883e-05, + "loss": 0.0, + "step": 1570 + }, + { + "epoch": 2.49, + "learning_rate": 2.7865961199294534e-05, + "loss": 0.0, + "step": 1571 + }, + { + "epoch": 2.5, + "learning_rate": 2.784832451499118e-05, + "loss": 0.0, + "step": 1572 + }, + { + "epoch": 2.5, + "learning_rate": 2.783068783068783e-05, + "loss": 0.0, + "step": 1573 + }, + { + "epoch": 2.5, + "learning_rate": 2.7813051146384484e-05, + "loss": 0.0, + "step": 1574 + }, + { + "epoch": 2.5, + "learning_rate": 2.779541446208113e-05, + "loss": 0.0, + "step": 1575 + }, + { + "epoch": 2.5, + "learning_rate": 2.777777777777778e-05, + "loss": 0.0001, + "step": 1576 + }, + { + "epoch": 2.5, + "learning_rate": 2.776014109347443e-05, + "loss": 0.0, + "step": 1577 + }, + { + "epoch": 2.5, + "learning_rate": 2.7742504409171078e-05, + "loss": 0.0, + "step": 1578 + }, + { + "epoch": 2.51, + "learning_rate": 2.7724867724867726e-05, + "loss": 0.0, + "step": 1579 + }, + { + "epoch": 2.51, + "learning_rate": 2.7707231040564373e-05, + "loss": 0.0, + "step": 1580 + }, + { + "epoch": 2.51, + "learning_rate": 2.768959435626102e-05, + "loss": 0.0, + "step": 1581 + }, + { + "epoch": 2.51, + "learning_rate": 2.7671957671957672e-05, + "loss": 0.0, + "step": 1582 + }, + { + "epoch": 2.51, + "learning_rate": 2.765432098765432e-05, + "loss": 0.0, + "step": 1583 + }, + { + "epoch": 2.51, + "learning_rate": 2.7636684303350974e-05, + "loss": 0.001, + "step": 1584 + }, + { + "epoch": 2.52, + "learning_rate": 2.7619047619047622e-05, + "loss": 0.0, + "step": 1585 + }, + { + "epoch": 2.52, + "learning_rate": 2.760141093474427e-05, + "loss": 0.0, + "step": 1586 + }, + { + "epoch": 2.52, + "learning_rate": 2.7583774250440918e-05, + "loss": 0.0, + "step": 1587 + }, + { + "epoch": 2.52, + "learning_rate": 2.756613756613757e-05, + "loss": 0.0, + "step": 1588 + }, + { + "epoch": 2.52, + "learning_rate": 2.7548500881834216e-05, + "loss": 0.0, + "step": 1589 + }, + { + "epoch": 2.52, + "learning_rate": 2.7530864197530864e-05, + "loss": 0.0, + "step": 1590 + }, + { + "epoch": 2.53, + "learning_rate": 2.7513227513227512e-05, + "loss": 0.0, + "step": 1591 + }, + { + "epoch": 2.53, + "learning_rate": 2.7495590828924163e-05, + "loss": 0.0001, + "step": 1592 + }, + { + "epoch": 2.53, + "learning_rate": 2.747795414462081e-05, + "loss": 0.0, + "step": 1593 + }, + { + "epoch": 2.53, + "learning_rate": 2.7460317460317465e-05, + "loss": 0.0, + "step": 1594 + }, + { + "epoch": 2.53, + "learning_rate": 2.7442680776014113e-05, + "loss": 0.0, + "step": 1595 + }, + { + "epoch": 2.53, + "learning_rate": 2.742504409171076e-05, + "loss": 0.2011, + "step": 1596 + }, + { + "epoch": 2.53, + "learning_rate": 2.7407407407407408e-05, + "loss": 0.0, + "step": 1597 + }, + { + "epoch": 2.54, + "learning_rate": 2.738977072310406e-05, + "loss": 0.0, + "step": 1598 + }, + { + "epoch": 2.54, + "learning_rate": 2.7372134038800707e-05, + "loss": 0.0, + "step": 1599 + }, + { + "epoch": 2.54, + "learning_rate": 2.7354497354497355e-05, + "loss": 0.0, + "step": 1600 + }, + { + "epoch": 2.54, + "learning_rate": 2.7336860670194003e-05, + "loss": 0.0, + "step": 1601 + }, + { + "epoch": 2.54, + "learning_rate": 2.731922398589065e-05, + "loss": 0.0, + "step": 1602 + }, + { + "epoch": 2.54, + "learning_rate": 2.73015873015873e-05, + "loss": 0.0, + "step": 1603 + }, + { + "epoch": 2.55, + "learning_rate": 2.7283950617283956e-05, + "loss": 0.0, + "step": 1604 + }, + { + "epoch": 2.55, + "learning_rate": 2.7266313932980604e-05, + "loss": 0.002, + "step": 1605 + }, + { + "epoch": 2.55, + "learning_rate": 2.724867724867725e-05, + "loss": 0.0, + "step": 1606 + }, + { + "epoch": 2.55, + "learning_rate": 2.72310405643739e-05, + "loss": 0.0001, + "step": 1607 + }, + { + "epoch": 2.55, + "learning_rate": 2.7213403880070547e-05, + "loss": 0.0, + "step": 1608 + }, + { + "epoch": 2.55, + "learning_rate": 2.7195767195767198e-05, + "loss": 0.0, + "step": 1609 + }, + { + "epoch": 2.56, + "learning_rate": 2.7178130511463846e-05, + "loss": 0.0, + "step": 1610 + }, + { + "epoch": 2.56, + "learning_rate": 2.7160493827160493e-05, + "loss": 0.116, + "step": 1611 + }, + { + "epoch": 2.56, + "learning_rate": 2.714285714285714e-05, + "loss": 0.0, + "step": 1612 + }, + { + "epoch": 2.56, + "learning_rate": 2.7125220458553792e-05, + "loss": 0.0, + "step": 1613 + }, + { + "epoch": 2.56, + "learning_rate": 2.7107583774250443e-05, + "loss": 0.0, + "step": 1614 + }, + { + "epoch": 2.56, + "learning_rate": 2.7089947089947094e-05, + "loss": 0.0, + "step": 1615 + }, + { + "epoch": 2.57, + "learning_rate": 2.7072310405643742e-05, + "loss": 0.0, + "step": 1616 + }, + { + "epoch": 2.57, + "learning_rate": 2.705467372134039e-05, + "loss": 0.0, + "step": 1617 + }, + { + "epoch": 2.57, + "learning_rate": 2.7037037037037037e-05, + "loss": 0.0, + "step": 1618 + }, + { + "epoch": 2.57, + "learning_rate": 2.701940035273369e-05, + "loss": 0.0, + "step": 1619 + }, + { + "epoch": 2.57, + "learning_rate": 2.7001763668430336e-05, + "loss": 0.0, + "step": 1620 + }, + { + "epoch": 2.57, + "learning_rate": 2.6984126984126984e-05, + "loss": 0.0, + "step": 1621 + }, + { + "epoch": 2.57, + "learning_rate": 2.696649029982363e-05, + "loss": 0.0, + "step": 1622 + }, + { + "epoch": 2.58, + "learning_rate": 2.694885361552028e-05, + "loss": 0.0, + "step": 1623 + }, + { + "epoch": 2.58, + "learning_rate": 2.6931216931216934e-05, + "loss": 0.0, + "step": 1624 + }, + { + "epoch": 2.58, + "learning_rate": 2.6913580246913585e-05, + "loss": 0.0, + "step": 1625 + }, + { + "epoch": 2.58, + "learning_rate": 2.6895943562610233e-05, + "loss": 0.0, + "step": 1626 + }, + { + "epoch": 2.58, + "learning_rate": 2.687830687830688e-05, + "loss": 0.0, + "step": 1627 + }, + { + "epoch": 2.58, + "learning_rate": 2.6860670194003528e-05, + "loss": 0.0076, + "step": 1628 + }, + { + "epoch": 2.59, + "learning_rate": 2.6843033509700176e-05, + "loss": 0.0, + "step": 1629 + }, + { + "epoch": 2.59, + "learning_rate": 2.6825396825396827e-05, + "loss": 0.0, + "step": 1630 + }, + { + "epoch": 2.59, + "learning_rate": 2.6807760141093475e-05, + "loss": 0.0001, + "step": 1631 + }, + { + "epoch": 2.59, + "learning_rate": 2.6790123456790122e-05, + "loss": 0.0, + "step": 1632 + }, + { + "epoch": 2.59, + "learning_rate": 2.677248677248677e-05, + "loss": 0.0, + "step": 1633 + }, + { + "epoch": 2.59, + "learning_rate": 2.6754850088183425e-05, + "loss": 0.0089, + "step": 1634 + }, + { + "epoch": 2.6, + "learning_rate": 2.6737213403880072e-05, + "loss": 0.0, + "step": 1635 + }, + { + "epoch": 2.6, + "learning_rate": 2.6719576719576723e-05, + "loss": 0.0, + "step": 1636 + }, + { + "epoch": 2.6, + "learning_rate": 2.670194003527337e-05, + "loss": 0.0, + "step": 1637 + }, + { + "epoch": 2.6, + "learning_rate": 2.668430335097002e-05, + "loss": 0.0, + "step": 1638 + }, + { + "epoch": 2.6, + "learning_rate": 2.6666666666666667e-05, + "loss": 0.0, + "step": 1639 + }, + { + "epoch": 2.6, + "learning_rate": 2.6649029982363318e-05, + "loss": 0.0, + "step": 1640 + }, + { + "epoch": 2.6, + "learning_rate": 2.6631393298059965e-05, + "loss": 0.0, + "step": 1641 + }, + { + "epoch": 2.61, + "learning_rate": 2.6613756613756613e-05, + "loss": 0.0001, + "step": 1642 + }, + { + "epoch": 2.61, + "learning_rate": 2.659611992945326e-05, + "loss": 0.0, + "step": 1643 + }, + { + "epoch": 2.61, + "learning_rate": 2.6578483245149915e-05, + "loss": 0.0041, + "step": 1644 + }, + { + "epoch": 2.61, + "learning_rate": 2.6560846560846563e-05, + "loss": 0.0, + "step": 1645 + }, + { + "epoch": 2.61, + "learning_rate": 2.654320987654321e-05, + "loss": 0.0, + "step": 1646 + }, + { + "epoch": 2.61, + "learning_rate": 2.6525573192239862e-05, + "loss": 0.0, + "step": 1647 + }, + { + "epoch": 2.62, + "learning_rate": 2.650793650793651e-05, + "loss": 0.0, + "step": 1648 + }, + { + "epoch": 2.62, + "learning_rate": 2.6490299823633157e-05, + "loss": 0.0, + "step": 1649 + }, + { + "epoch": 2.62, + "learning_rate": 2.6472663139329805e-05, + "loss": 0.0, + "step": 1650 + }, + { + "epoch": 2.62, + "learning_rate": 2.6455026455026456e-05, + "loss": 0.0, + "step": 1651 + }, + { + "epoch": 2.62, + "learning_rate": 2.6437389770723104e-05, + "loss": 0.0, + "step": 1652 + }, + { + "epoch": 2.62, + "learning_rate": 2.641975308641975e-05, + "loss": 0.0, + "step": 1653 + }, + { + "epoch": 2.63, + "learning_rate": 2.6402116402116406e-05, + "loss": 0.0, + "step": 1654 + }, + { + "epoch": 2.63, + "learning_rate": 2.6384479717813054e-05, + "loss": 0.0, + "step": 1655 + }, + { + "epoch": 2.63, + "learning_rate": 2.63668430335097e-05, + "loss": 0.0, + "step": 1656 + }, + { + "epoch": 2.63, + "learning_rate": 2.6349206349206353e-05, + "loss": 0.0001, + "step": 1657 + }, + { + "epoch": 2.63, + "learning_rate": 2.6331569664903e-05, + "loss": 0.0, + "step": 1658 + }, + { + "epoch": 2.63, + "learning_rate": 2.6313932980599648e-05, + "loss": 0.0, + "step": 1659 + }, + { + "epoch": 2.63, + "learning_rate": 2.6296296296296296e-05, + "loss": 0.0, + "step": 1660 + }, + { + "epoch": 2.64, + "learning_rate": 2.6278659611992943e-05, + "loss": 0.0, + "step": 1661 + }, + { + "epoch": 2.64, + "learning_rate": 2.6261022927689595e-05, + "loss": 0.0, + "step": 1662 + }, + { + "epoch": 2.64, + "learning_rate": 2.6243386243386242e-05, + "loss": 0.0, + "step": 1663 + }, + { + "epoch": 2.64, + "learning_rate": 2.6225749559082897e-05, + "loss": 0.0, + "step": 1664 + }, + { + "epoch": 2.64, + "learning_rate": 2.6208112874779544e-05, + "loss": 0.0, + "step": 1665 + }, + { + "epoch": 2.64, + "learning_rate": 2.6190476190476192e-05, + "loss": 0.0, + "step": 1666 + }, + { + "epoch": 2.65, + "learning_rate": 2.617283950617284e-05, + "loss": 0.0, + "step": 1667 + }, + { + "epoch": 2.65, + "learning_rate": 2.615520282186949e-05, + "loss": 0.0001, + "step": 1668 + }, + { + "epoch": 2.65, + "learning_rate": 2.613756613756614e-05, + "loss": 0.0, + "step": 1669 + }, + { + "epoch": 2.65, + "learning_rate": 2.6119929453262786e-05, + "loss": 0.0001, + "step": 1670 + }, + { + "epoch": 2.65, + "learning_rate": 2.6102292768959434e-05, + "loss": 0.0001, + "step": 1671 + }, + { + "epoch": 2.65, + "learning_rate": 2.6084656084656085e-05, + "loss": 0.0, + "step": 1672 + }, + { + "epoch": 2.66, + "learning_rate": 2.6067019400352733e-05, + "loss": 0.0001, + "step": 1673 + }, + { + "epoch": 2.66, + "learning_rate": 2.6049382716049388e-05, + "loss": 0.0, + "step": 1674 + }, + { + "epoch": 2.66, + "learning_rate": 2.6031746031746035e-05, + "loss": 0.0, + "step": 1675 + }, + { + "epoch": 2.66, + "learning_rate": 2.6014109347442683e-05, + "loss": 0.0, + "step": 1676 + }, + { + "epoch": 2.66, + "learning_rate": 2.599647266313933e-05, + "loss": 0.0, + "step": 1677 + }, + { + "epoch": 2.66, + "learning_rate": 2.5978835978835982e-05, + "loss": 0.0, + "step": 1678 + }, + { + "epoch": 2.67, + "learning_rate": 2.596119929453263e-05, + "loss": 0.0, + "step": 1679 + }, + { + "epoch": 2.67, + "learning_rate": 2.5943562610229277e-05, + "loss": 0.0, + "step": 1680 + }, + { + "epoch": 2.67, + "learning_rate": 2.5925925925925925e-05, + "loss": 0.0, + "step": 1681 + }, + { + "epoch": 2.67, + "learning_rate": 2.5908289241622573e-05, + "loss": 0.0, + "step": 1682 + }, + { + "epoch": 2.67, + "learning_rate": 2.5890652557319224e-05, + "loss": 0.0, + "step": 1683 + }, + { + "epoch": 2.67, + "learning_rate": 2.5873015873015878e-05, + "loss": 0.0001, + "step": 1684 + }, + { + "epoch": 2.67, + "learning_rate": 2.5855379188712526e-05, + "loss": 0.0, + "step": 1685 + }, + { + "epoch": 2.68, + "learning_rate": 2.5837742504409174e-05, + "loss": 0.0, + "step": 1686 + }, + { + "epoch": 2.68, + "learning_rate": 2.582010582010582e-05, + "loss": 0.0, + "step": 1687 + }, + { + "epoch": 2.68, + "learning_rate": 2.580246913580247e-05, + "loss": 0.0, + "step": 1688 + }, + { + "epoch": 2.68, + "learning_rate": 2.578483245149912e-05, + "loss": 0.0, + "step": 1689 + }, + { + "epoch": 2.68, + "learning_rate": 2.5767195767195768e-05, + "loss": 0.0, + "step": 1690 + }, + { + "epoch": 2.68, + "learning_rate": 2.5749559082892416e-05, + "loss": 0.0, + "step": 1691 + }, + { + "epoch": 2.69, + "learning_rate": 2.5731922398589063e-05, + "loss": 0.0, + "step": 1692 + }, + { + "epoch": 2.69, + "learning_rate": 2.5714285714285714e-05, + "loss": 0.0, + "step": 1693 + }, + { + "epoch": 2.69, + "learning_rate": 2.5696649029982366e-05, + "loss": 0.0, + "step": 1694 + }, + { + "epoch": 2.69, + "learning_rate": 2.5679012345679017e-05, + "loss": 0.0, + "step": 1695 + }, + { + "epoch": 2.69, + "learning_rate": 2.5661375661375664e-05, + "loss": 0.0, + "step": 1696 + }, + { + "epoch": 2.69, + "learning_rate": 2.5643738977072312e-05, + "loss": 0.0, + "step": 1697 + }, + { + "epoch": 2.7, + "learning_rate": 2.562610229276896e-05, + "loss": 0.0, + "step": 1698 + }, + { + "epoch": 2.7, + "learning_rate": 2.560846560846561e-05, + "loss": 0.0, + "step": 1699 + }, + { + "epoch": 2.7, + "learning_rate": 2.559082892416226e-05, + "loss": 0.0, + "step": 1700 + }, + { + "epoch": 2.7, + "learning_rate": 2.5573192239858906e-05, + "loss": 0.0, + "step": 1701 + }, + { + "epoch": 2.7, + "learning_rate": 2.5555555555555554e-05, + "loss": 0.0, + "step": 1702 + }, + { + "epoch": 2.7, + "learning_rate": 2.5537918871252202e-05, + "loss": 0.0, + "step": 1703 + }, + { + "epoch": 2.7, + "learning_rate": 2.5520282186948856e-05, + "loss": 0.0, + "step": 1704 + }, + { + "epoch": 2.71, + "learning_rate": 2.5502645502645507e-05, + "loss": 0.0, + "step": 1705 + }, + { + "epoch": 2.71, + "learning_rate": 2.5485008818342155e-05, + "loss": 0.0, + "step": 1706 + }, + { + "epoch": 2.71, + "learning_rate": 2.5467372134038803e-05, + "loss": 0.0, + "step": 1707 + }, + { + "epoch": 2.71, + "learning_rate": 2.544973544973545e-05, + "loss": 0.0, + "step": 1708 + }, + { + "epoch": 2.71, + "learning_rate": 2.5432098765432098e-05, + "loss": 0.0, + "step": 1709 + }, + { + "epoch": 2.71, + "learning_rate": 2.541446208112875e-05, + "loss": 0.0, + "step": 1710 + }, + { + "epoch": 2.72, + "learning_rate": 2.5396825396825397e-05, + "loss": 0.0, + "step": 1711 + }, + { + "epoch": 2.72, + "learning_rate": 2.5379188712522045e-05, + "loss": 0.0001, + "step": 1712 + }, + { + "epoch": 2.72, + "learning_rate": 2.5361552028218693e-05, + "loss": 0.0013, + "step": 1713 + }, + { + "epoch": 2.72, + "learning_rate": 2.5343915343915347e-05, + "loss": 0.1339, + "step": 1714 + }, + { + "epoch": 2.72, + "learning_rate": 2.5326278659611995e-05, + "loss": 0.0, + "step": 1715 + }, + { + "epoch": 2.72, + "learning_rate": 2.5308641975308646e-05, + "loss": 0.0001, + "step": 1716 + }, + { + "epoch": 2.73, + "learning_rate": 2.5291005291005294e-05, + "loss": 0.0, + "step": 1717 + }, + { + "epoch": 2.73, + "learning_rate": 2.527336860670194e-05, + "loss": 0.0, + "step": 1718 + }, + { + "epoch": 2.73, + "learning_rate": 2.525573192239859e-05, + "loss": 0.0, + "step": 1719 + }, + { + "epoch": 2.73, + "learning_rate": 2.523809523809524e-05, + "loss": 0.0, + "step": 1720 + }, + { + "epoch": 2.73, + "learning_rate": 2.5220458553791888e-05, + "loss": 0.0, + "step": 1721 + }, + { + "epoch": 2.73, + "learning_rate": 2.5202821869488536e-05, + "loss": 0.0, + "step": 1722 + }, + { + "epoch": 2.73, + "learning_rate": 2.5185185185185183e-05, + "loss": 0.0001, + "step": 1723 + }, + { + "epoch": 2.74, + "learning_rate": 2.5167548500881838e-05, + "loss": 0.0, + "step": 1724 + }, + { + "epoch": 2.74, + "learning_rate": 2.5149911816578485e-05, + "loss": 0.0, + "step": 1725 + }, + { + "epoch": 2.74, + "learning_rate": 2.5132275132275137e-05, + "loss": 0.0, + "step": 1726 + }, + { + "epoch": 2.74, + "learning_rate": 2.5114638447971784e-05, + "loss": 0.0001, + "step": 1727 + }, + { + "epoch": 2.74, + "learning_rate": 2.5097001763668432e-05, + "loss": 0.0, + "step": 1728 + }, + { + "epoch": 2.74, + "learning_rate": 2.507936507936508e-05, + "loss": 0.0, + "step": 1729 + }, + { + "epoch": 2.75, + "learning_rate": 2.5061728395061727e-05, + "loss": 0.0, + "step": 1730 + }, + { + "epoch": 2.75, + "learning_rate": 2.504409171075838e-05, + "loss": 0.0, + "step": 1731 + }, + { + "epoch": 2.75, + "learning_rate": 2.5026455026455026e-05, + "loss": 0.0, + "step": 1732 + }, + { + "epoch": 2.75, + "learning_rate": 2.5008818342151674e-05, + "loss": 0.0, + "step": 1733 + }, + { + "epoch": 2.75, + "learning_rate": 2.4991181657848325e-05, + "loss": 0.0, + "step": 1734 + }, + { + "epoch": 2.75, + "learning_rate": 2.4973544973544973e-05, + "loss": 0.0, + "step": 1735 + }, + { + "epoch": 2.76, + "learning_rate": 2.4955908289241624e-05, + "loss": 0.0, + "step": 1736 + }, + { + "epoch": 2.76, + "learning_rate": 2.4938271604938275e-05, + "loss": 0.0, + "step": 1737 + }, + { + "epoch": 2.76, + "learning_rate": 2.4920634920634923e-05, + "loss": 0.0, + "step": 1738 + }, + { + "epoch": 2.76, + "learning_rate": 2.490299823633157e-05, + "loss": 0.0, + "step": 1739 + }, + { + "epoch": 2.76, + "learning_rate": 2.4885361552028218e-05, + "loss": 0.0, + "step": 1740 + }, + { + "epoch": 2.76, + "learning_rate": 2.4867724867724866e-05, + "loss": 0.0001, + "step": 1741 + }, + { + "epoch": 2.77, + "learning_rate": 2.485008818342152e-05, + "loss": 0.0, + "step": 1742 + }, + { + "epoch": 2.77, + "learning_rate": 2.4832451499118168e-05, + "loss": 0.0, + "step": 1743 + }, + { + "epoch": 2.77, + "learning_rate": 2.4814814814814816e-05, + "loss": 0.0, + "step": 1744 + }, + { + "epoch": 2.77, + "learning_rate": 2.4797178130511463e-05, + "loss": 0.0, + "step": 1745 + }, + { + "epoch": 2.77, + "learning_rate": 2.477954144620811e-05, + "loss": 0.0, + "step": 1746 + }, + { + "epoch": 2.77, + "learning_rate": 2.4761904761904762e-05, + "loss": 0.0, + "step": 1747 + }, + { + "epoch": 2.77, + "learning_rate": 2.4744268077601413e-05, + "loss": 0.0001, + "step": 1748 + }, + { + "epoch": 2.78, + "learning_rate": 2.472663139329806e-05, + "loss": 0.0, + "step": 1749 + }, + { + "epoch": 2.78, + "learning_rate": 2.470899470899471e-05, + "loss": 0.0, + "step": 1750 + }, + { + "epoch": 2.78, + "learning_rate": 2.4691358024691357e-05, + "loss": 0.0, + "step": 1751 + }, + { + "epoch": 2.78, + "learning_rate": 2.4673721340388008e-05, + "loss": 0.0, + "step": 1752 + }, + { + "epoch": 2.78, + "learning_rate": 2.465608465608466e-05, + "loss": 0.0, + "step": 1753 + }, + { + "epoch": 2.78, + "learning_rate": 2.4638447971781306e-05, + "loss": 0.0, + "step": 1754 + }, + { + "epoch": 2.79, + "learning_rate": 2.4620811287477954e-05, + "loss": 0.0, + "step": 1755 + }, + { + "epoch": 2.79, + "learning_rate": 2.4603174603174602e-05, + "loss": 0.0001, + "step": 1756 + }, + { + "epoch": 2.79, + "learning_rate": 2.4585537918871253e-05, + "loss": 0.0, + "step": 1757 + }, + { + "epoch": 2.79, + "learning_rate": 2.4567901234567904e-05, + "loss": 0.0, + "step": 1758 + }, + { + "epoch": 2.79, + "learning_rate": 2.4550264550264552e-05, + "loss": 0.0, + "step": 1759 + }, + { + "epoch": 2.79, + "learning_rate": 2.45326278659612e-05, + "loss": 0.0, + "step": 1760 + }, + { + "epoch": 2.8, + "learning_rate": 2.4514991181657847e-05, + "loss": 0.0, + "step": 1761 + }, + { + "epoch": 2.8, + "learning_rate": 2.44973544973545e-05, + "loss": 0.0, + "step": 1762 + }, + { + "epoch": 2.8, + "learning_rate": 2.447971781305115e-05, + "loss": 0.0, + "step": 1763 + }, + { + "epoch": 2.8, + "learning_rate": 2.4462081128747797e-05, + "loss": 0.0, + "step": 1764 + }, + { + "epoch": 2.8, + "learning_rate": 2.4444444444444445e-05, + "loss": 0.0, + "step": 1765 + }, + { + "epoch": 2.8, + "learning_rate": 2.4426807760141093e-05, + "loss": 0.0, + "step": 1766 + }, + { + "epoch": 2.8, + "learning_rate": 2.4409171075837744e-05, + "loss": 0.0, + "step": 1767 + }, + { + "epoch": 2.81, + "learning_rate": 2.439153439153439e-05, + "loss": 0.0, + "step": 1768 + }, + { + "epoch": 2.81, + "learning_rate": 2.4373897707231043e-05, + "loss": 0.0, + "step": 1769 + }, + { + "epoch": 2.81, + "learning_rate": 2.435626102292769e-05, + "loss": 0.0, + "step": 1770 + }, + { + "epoch": 2.81, + "learning_rate": 2.4338624338624338e-05, + "loss": 0.0, + "step": 1771 + }, + { + "epoch": 2.81, + "learning_rate": 2.432098765432099e-05, + "loss": 0.0, + "step": 1772 + }, + { + "epoch": 2.81, + "learning_rate": 2.4303350970017637e-05, + "loss": 0.0, + "step": 1773 + }, + { + "epoch": 2.82, + "learning_rate": 2.4285714285714288e-05, + "loss": 0.0, + "step": 1774 + }, + { + "epoch": 2.82, + "learning_rate": 2.4268077601410936e-05, + "loss": 0.0, + "step": 1775 + }, + { + "epoch": 2.82, + "learning_rate": 2.4250440917107583e-05, + "loss": 0.0, + "step": 1776 + }, + { + "epoch": 2.82, + "learning_rate": 2.4232804232804234e-05, + "loss": 0.0002, + "step": 1777 + }, + { + "epoch": 2.82, + "learning_rate": 2.4215167548500882e-05, + "loss": 0.0, + "step": 1778 + }, + { + "epoch": 2.82, + "learning_rate": 2.4197530864197533e-05, + "loss": 0.0, + "step": 1779 + }, + { + "epoch": 2.83, + "learning_rate": 2.417989417989418e-05, + "loss": 0.0, + "step": 1780 + }, + { + "epoch": 2.83, + "learning_rate": 2.416225749559083e-05, + "loss": 0.0, + "step": 1781 + }, + { + "epoch": 2.83, + "learning_rate": 2.414462081128748e-05, + "loss": 0.0, + "step": 1782 + }, + { + "epoch": 2.83, + "learning_rate": 2.4126984126984128e-05, + "loss": 0.0, + "step": 1783 + }, + { + "epoch": 2.83, + "learning_rate": 2.4109347442680775e-05, + "loss": 0.0, + "step": 1784 + }, + { + "epoch": 2.83, + "learning_rate": 2.4091710758377426e-05, + "loss": 0.0, + "step": 1785 + }, + { + "epoch": 2.83, + "learning_rate": 2.4074074074074074e-05, + "loss": 0.0, + "step": 1786 + }, + { + "epoch": 2.84, + "learning_rate": 2.4056437389770725e-05, + "loss": 0.0, + "step": 1787 + }, + { + "epoch": 2.84, + "learning_rate": 2.4038800705467373e-05, + "loss": 0.0, + "step": 1788 + }, + { + "epoch": 2.84, + "learning_rate": 2.402116402116402e-05, + "loss": 0.0, + "step": 1789 + }, + { + "epoch": 2.84, + "learning_rate": 2.4003527336860672e-05, + "loss": 0.0, + "step": 1790 + }, + { + "epoch": 2.84, + "learning_rate": 2.398589065255732e-05, + "loss": 0.0, + "step": 1791 + }, + { + "epoch": 2.84, + "learning_rate": 2.396825396825397e-05, + "loss": 0.0, + "step": 1792 + }, + { + "epoch": 2.85, + "learning_rate": 2.3950617283950618e-05, + "loss": 0.0, + "step": 1793 + }, + { + "epoch": 2.85, + "learning_rate": 2.3932980599647266e-05, + "loss": 0.0001, + "step": 1794 + }, + { + "epoch": 2.85, + "learning_rate": 2.3915343915343917e-05, + "loss": 0.0, + "step": 1795 + }, + { + "epoch": 2.85, + "learning_rate": 2.3897707231040565e-05, + "loss": 0.0, + "step": 1796 + }, + { + "epoch": 2.85, + "learning_rate": 2.3880070546737216e-05, + "loss": 0.0, + "step": 1797 + }, + { + "epoch": 2.85, + "learning_rate": 2.3862433862433864e-05, + "loss": 0.0, + "step": 1798 + }, + { + "epoch": 2.86, + "learning_rate": 2.384479717813051e-05, + "loss": 0.0, + "step": 1799 + }, + { + "epoch": 2.86, + "learning_rate": 2.3827160493827162e-05, + "loss": 0.0, + "step": 1800 + }, + { + "epoch": 2.86, + "learning_rate": 2.380952380952381e-05, + "loss": 0.0, + "step": 1801 + }, + { + "epoch": 2.86, + "learning_rate": 2.379188712522046e-05, + "loss": 0.0, + "step": 1802 + }, + { + "epoch": 2.86, + "learning_rate": 2.377425044091711e-05, + "loss": 0.0, + "step": 1803 + }, + { + "epoch": 2.86, + "learning_rate": 2.3756613756613757e-05, + "loss": 0.0, + "step": 1804 + }, + { + "epoch": 2.87, + "learning_rate": 2.3738977072310404e-05, + "loss": 0.0, + "step": 1805 + }, + { + "epoch": 2.87, + "learning_rate": 2.3721340388007056e-05, + "loss": 0.0, + "step": 1806 + }, + { + "epoch": 2.87, + "learning_rate": 2.3703703703703707e-05, + "loss": 0.0, + "step": 1807 + }, + { + "epoch": 2.87, + "learning_rate": 2.3686067019400354e-05, + "loss": 0.0, + "step": 1808 + }, + { + "epoch": 2.87, + "learning_rate": 2.3668430335097002e-05, + "loss": 0.0, + "step": 1809 + }, + { + "epoch": 2.87, + "learning_rate": 2.365079365079365e-05, + "loss": 0.0, + "step": 1810 + }, + { + "epoch": 2.87, + "learning_rate": 2.36331569664903e-05, + "loss": 0.0, + "step": 1811 + }, + { + "epoch": 2.88, + "learning_rate": 2.3615520282186952e-05, + "loss": 0.0, + "step": 1812 + }, + { + "epoch": 2.88, + "learning_rate": 2.35978835978836e-05, + "loss": 0.0, + "step": 1813 + }, + { + "epoch": 2.88, + "learning_rate": 2.3580246913580247e-05, + "loss": 0.0, + "step": 1814 + }, + { + "epoch": 2.88, + "learning_rate": 2.3562610229276895e-05, + "loss": 0.0, + "step": 1815 + }, + { + "epoch": 2.88, + "learning_rate": 2.3544973544973546e-05, + "loss": 0.0, + "step": 1816 + }, + { + "epoch": 2.88, + "learning_rate": 2.3527336860670197e-05, + "loss": 0.0, + "step": 1817 + }, + { + "epoch": 2.89, + "learning_rate": 2.3509700176366845e-05, + "loss": 0.0, + "step": 1818 + }, + { + "epoch": 2.89, + "learning_rate": 2.3492063492063493e-05, + "loss": 0.0, + "step": 1819 + }, + { + "epoch": 2.89, + "learning_rate": 2.347442680776014e-05, + "loss": 0.0, + "step": 1820 + }, + { + "epoch": 2.89, + "learning_rate": 2.345679012345679e-05, + "loss": 0.0, + "step": 1821 + }, + { + "epoch": 2.89, + "learning_rate": 2.3439153439153443e-05, + "loss": 0.0, + "step": 1822 + }, + { + "epoch": 2.89, + "learning_rate": 2.342151675485009e-05, + "loss": 0.0, + "step": 1823 + }, + { + "epoch": 2.9, + "learning_rate": 2.3403880070546738e-05, + "loss": 0.0, + "step": 1824 + }, + { + "epoch": 2.9, + "learning_rate": 2.3386243386243386e-05, + "loss": 0.0, + "step": 1825 + }, + { + "epoch": 2.9, + "learning_rate": 2.3368606701940034e-05, + "loss": 0.0, + "step": 1826 + }, + { + "epoch": 2.9, + "learning_rate": 2.3350970017636685e-05, + "loss": 0.0, + "step": 1827 + }, + { + "epoch": 2.9, + "learning_rate": 2.3333333333333336e-05, + "loss": 0.0, + "step": 1828 + }, + { + "epoch": 2.9, + "learning_rate": 2.3315696649029984e-05, + "loss": 0.0, + "step": 1829 + }, + { + "epoch": 2.9, + "learning_rate": 2.329805996472663e-05, + "loss": 0.0, + "step": 1830 + }, + { + "epoch": 2.91, + "learning_rate": 2.328042328042328e-05, + "loss": 0.0, + "step": 1831 + }, + { + "epoch": 2.91, + "learning_rate": 2.326278659611993e-05, + "loss": 0.0, + "step": 1832 + }, + { + "epoch": 2.91, + "learning_rate": 2.324514991181658e-05, + "loss": 0.0, + "step": 1833 + }, + { + "epoch": 2.91, + "learning_rate": 2.322751322751323e-05, + "loss": 0.0, + "step": 1834 + }, + { + "epoch": 2.91, + "learning_rate": 2.3209876543209877e-05, + "loss": 0.0, + "step": 1835 + }, + { + "epoch": 2.91, + "learning_rate": 2.3192239858906524e-05, + "loss": 0.0479, + "step": 1836 + }, + { + "epoch": 2.92, + "learning_rate": 2.3174603174603175e-05, + "loss": 0.0, + "step": 1837 + }, + { + "epoch": 2.92, + "learning_rate": 2.3156966490299827e-05, + "loss": 0.0, + "step": 1838 + }, + { + "epoch": 2.92, + "learning_rate": 2.3139329805996474e-05, + "loss": 0.0, + "step": 1839 + }, + { + "epoch": 2.92, + "learning_rate": 2.3121693121693122e-05, + "loss": 0.0, + "step": 1840 + }, + { + "epoch": 2.92, + "learning_rate": 2.310405643738977e-05, + "loss": 0.0, + "step": 1841 + }, + { + "epoch": 2.92, + "learning_rate": 2.308641975308642e-05, + "loss": 0.0, + "step": 1842 + }, + { + "epoch": 2.93, + "learning_rate": 2.3068783068783072e-05, + "loss": 0.0227, + "step": 1843 + }, + { + "epoch": 2.93, + "learning_rate": 2.305114638447972e-05, + "loss": 0.0, + "step": 1844 + }, + { + "epoch": 2.93, + "learning_rate": 2.3033509700176367e-05, + "loss": 0.0, + "step": 1845 + }, + { + "epoch": 2.93, + "learning_rate": 2.3015873015873015e-05, + "loss": 0.0, + "step": 1846 + }, + { + "epoch": 2.93, + "learning_rate": 2.2998236331569666e-05, + "loss": 0.0001, + "step": 1847 + }, + { + "epoch": 2.93, + "learning_rate": 2.2980599647266314e-05, + "loss": 0.0, + "step": 1848 + }, + { + "epoch": 2.93, + "learning_rate": 2.2962962962962965e-05, + "loss": 0.0, + "step": 1849 + }, + { + "epoch": 2.94, + "learning_rate": 2.2945326278659613e-05, + "loss": 0.0, + "step": 1850 + }, + { + "epoch": 2.94, + "learning_rate": 2.292768959435626e-05, + "loss": 0.0, + "step": 1851 + }, + { + "epoch": 2.94, + "learning_rate": 2.291005291005291e-05, + "loss": 0.0, + "step": 1852 + }, + { + "epoch": 2.94, + "learning_rate": 2.289241622574956e-05, + "loss": 0.0, + "step": 1853 + }, + { + "epoch": 2.94, + "learning_rate": 2.287477954144621e-05, + "loss": 0.0, + "step": 1854 + }, + { + "epoch": 2.94, + "learning_rate": 2.2857142857142858e-05, + "loss": 0.0, + "step": 1855 + }, + { + "epoch": 2.95, + "learning_rate": 2.2839506172839506e-05, + "loss": 0.0, + "step": 1856 + }, + { + "epoch": 2.95, + "learning_rate": 2.2821869488536157e-05, + "loss": 0.0, + "step": 1857 + }, + { + "epoch": 2.95, + "learning_rate": 2.2804232804232805e-05, + "loss": 0.0, + "step": 1858 + }, + { + "epoch": 2.95, + "learning_rate": 2.2786596119929456e-05, + "loss": 0.0, + "step": 1859 + }, + { + "epoch": 2.95, + "learning_rate": 2.2768959435626103e-05, + "loss": 0.0, + "step": 1860 + }, + { + "epoch": 2.95, + "learning_rate": 2.275132275132275e-05, + "loss": 0.0, + "step": 1861 + }, + { + "epoch": 2.96, + "learning_rate": 2.2733686067019402e-05, + "loss": 0.0, + "step": 1862 + }, + { + "epoch": 2.96, + "learning_rate": 2.271604938271605e-05, + "loss": 0.0, + "step": 1863 + }, + { + "epoch": 2.96, + "learning_rate": 2.2698412698412698e-05, + "loss": 0.0, + "step": 1864 + }, + { + "epoch": 2.96, + "learning_rate": 2.268077601410935e-05, + "loss": 0.0, + "step": 1865 + }, + { + "epoch": 2.96, + "learning_rate": 2.2663139329805996e-05, + "loss": 0.0005, + "step": 1866 + }, + { + "epoch": 2.96, + "learning_rate": 2.2645502645502648e-05, + "loss": 0.0, + "step": 1867 + }, + { + "epoch": 2.97, + "learning_rate": 2.2627865961199295e-05, + "loss": 0.0, + "step": 1868 + }, + { + "epoch": 2.97, + "learning_rate": 2.2610229276895943e-05, + "loss": 0.0, + "step": 1869 + }, + { + "epoch": 2.97, + "learning_rate": 2.2592592592592594e-05, + "loss": 0.0, + "step": 1870 + }, + { + "epoch": 2.97, + "learning_rate": 2.2574955908289242e-05, + "loss": 0.0, + "step": 1871 + }, + { + "epoch": 2.97, + "learning_rate": 2.2557319223985893e-05, + "loss": 0.0001, + "step": 1872 + }, + { + "epoch": 2.97, + "learning_rate": 2.253968253968254e-05, + "loss": 0.0, + "step": 1873 + }, + { + "epoch": 2.97, + "learning_rate": 2.252204585537919e-05, + "loss": 0.0, + "step": 1874 + }, + { + "epoch": 2.98, + "learning_rate": 2.250440917107584e-05, + "loss": 0.0, + "step": 1875 + }, + { + "epoch": 2.98, + "learning_rate": 2.2486772486772487e-05, + "loss": 0.0, + "step": 1876 + }, + { + "epoch": 2.98, + "learning_rate": 2.246913580246914e-05, + "loss": 0.0004, + "step": 1877 + }, + { + "epoch": 2.98, + "learning_rate": 2.2451499118165786e-05, + "loss": 0.0, + "step": 1878 + }, + { + "epoch": 2.98, + "learning_rate": 2.2433862433862434e-05, + "loss": 0.0, + "step": 1879 + }, + { + "epoch": 2.98, + "learning_rate": 2.2416225749559085e-05, + "loss": 0.0, + "step": 1880 + }, + { + "epoch": 2.99, + "learning_rate": 2.2398589065255733e-05, + "loss": 0.0, + "step": 1881 + }, + { + "epoch": 2.99, + "learning_rate": 2.2380952380952384e-05, + "loss": 0.0, + "step": 1882 + }, + { + "epoch": 2.99, + "learning_rate": 2.236331569664903e-05, + "loss": 0.0, + "step": 1883 + }, + { + "epoch": 2.99, + "learning_rate": 2.234567901234568e-05, + "loss": 0.0, + "step": 1884 + }, + { + "epoch": 2.99, + "learning_rate": 2.2328042328042327e-05, + "loss": 0.0, + "step": 1885 + }, + { + "epoch": 2.99, + "learning_rate": 2.2310405643738978e-05, + "loss": 0.0, + "step": 1886 + }, + { + "epoch": 3.0, + "learning_rate": 2.229276895943563e-05, + "loss": 0.0, + "step": 1887 + }, + { + "epoch": 3.0, + "learning_rate": 2.2275132275132277e-05, + "loss": 0.0, + "step": 1888 + }, + { + "epoch": 3.0, + "learning_rate": 2.2257495590828924e-05, + "loss": 0.0, + "step": 1889 + }, + { + "epoch": 3.0, + "learning_rate": 2.2239858906525572e-05, + "loss": 0.0001, + "step": 1890 + }, + { + "epoch": 3.0, + "eval_accuracy": 0.9736842105263158, + "eval_auc": 0.9958333333333333, + "eval_f1": 0.9333333333333333, + "eval_loss": 0.09085065126419067, + "eval_precision": 1.0, + "eval_recall": 0.875, + "eval_runtime": 0.2019, + "eval_samples_per_second": 188.206, + "eval_steps_per_second": 4.953, + "step": 1890 + }, + { + "epoch": 3.0, + "learning_rate": 2.2222222222222223e-05, + "loss": 0.0, + "step": 1891 + }, + { + "epoch": 3.0, + "learning_rate": 2.2204585537918874e-05, + "loss": 0.0, + "step": 1892 + }, + { + "epoch": 3.0, + "learning_rate": 2.2186948853615522e-05, + "loss": 0.0, + "step": 1893 + }, + { + "epoch": 3.01, + "learning_rate": 2.216931216931217e-05, + "loss": 0.0, + "step": 1894 + }, + { + "epoch": 3.01, + "learning_rate": 2.2151675485008818e-05, + "loss": 0.0, + "step": 1895 + }, + { + "epoch": 3.01, + "learning_rate": 2.213403880070547e-05, + "loss": 0.0, + "step": 1896 + }, + { + "epoch": 3.01, + "learning_rate": 2.211640211640212e-05, + "loss": 0.0, + "step": 1897 + }, + { + "epoch": 3.01, + "learning_rate": 2.2098765432098767e-05, + "loss": 0.0, + "step": 1898 + }, + { + "epoch": 3.01, + "learning_rate": 2.2081128747795415e-05, + "loss": 0.0, + "step": 1899 + }, + { + "epoch": 3.02, + "learning_rate": 2.2063492063492063e-05, + "loss": 0.0, + "step": 1900 + }, + { + "epoch": 3.02, + "learning_rate": 2.2045855379188714e-05, + "loss": 0.0, + "step": 1901 + }, + { + "epoch": 3.02, + "learning_rate": 2.2028218694885365e-05, + "loss": 0.0, + "step": 1902 + }, + { + "epoch": 3.02, + "learning_rate": 2.2010582010582013e-05, + "loss": 0.0, + "step": 1903 + }, + { + "epoch": 3.02, + "learning_rate": 2.199294532627866e-05, + "loss": 0.0, + "step": 1904 + }, + { + "epoch": 3.02, + "learning_rate": 2.1975308641975308e-05, + "loss": 0.0, + "step": 1905 + }, + { + "epoch": 3.03, + "learning_rate": 2.1957671957671956e-05, + "loss": 0.0, + "step": 1906 + }, + { + "epoch": 3.03, + "learning_rate": 2.1940035273368607e-05, + "loss": 0.0, + "step": 1907 + }, + { + "epoch": 3.03, + "learning_rate": 2.1922398589065258e-05, + "loss": 0.0, + "step": 1908 + }, + { + "epoch": 3.03, + "learning_rate": 2.1904761904761906e-05, + "loss": 0.0, + "step": 1909 + }, + { + "epoch": 3.03, + "learning_rate": 2.1887125220458554e-05, + "loss": 0.0, + "step": 1910 + }, + { + "epoch": 3.03, + "learning_rate": 2.18694885361552e-05, + "loss": 0.0, + "step": 1911 + }, + { + "epoch": 3.03, + "learning_rate": 2.1851851851851852e-05, + "loss": 0.0, + "step": 1912 + }, + { + "epoch": 3.04, + "learning_rate": 2.1834215167548504e-05, + "loss": 0.0, + "step": 1913 + }, + { + "epoch": 3.04, + "learning_rate": 2.181657848324515e-05, + "loss": 0.0, + "step": 1914 + }, + { + "epoch": 3.04, + "learning_rate": 2.17989417989418e-05, + "loss": 0.0, + "step": 1915 + }, + { + "epoch": 3.04, + "learning_rate": 2.1781305114638447e-05, + "loss": 0.0, + "step": 1916 + }, + { + "epoch": 3.04, + "learning_rate": 2.1763668430335098e-05, + "loss": 0.0, + "step": 1917 + }, + { + "epoch": 3.04, + "learning_rate": 2.174603174603175e-05, + "loss": 0.0, + "step": 1918 + }, + { + "epoch": 3.05, + "learning_rate": 2.1728395061728397e-05, + "loss": 0.0, + "step": 1919 + }, + { + "epoch": 3.05, + "learning_rate": 2.1710758377425044e-05, + "loss": 0.0, + "step": 1920 + }, + { + "epoch": 3.05, + "learning_rate": 2.1693121693121692e-05, + "loss": 0.0, + "step": 1921 + }, + { + "epoch": 3.05, + "learning_rate": 2.1675485008818343e-05, + "loss": 0.0, + "step": 1922 + }, + { + "epoch": 3.05, + "learning_rate": 2.1657848324514994e-05, + "loss": 0.0, + "step": 1923 + }, + { + "epoch": 3.05, + "learning_rate": 2.1640211640211642e-05, + "loss": 0.0, + "step": 1924 + }, + { + "epoch": 3.06, + "learning_rate": 2.162257495590829e-05, + "loss": 0.0, + "step": 1925 + }, + { + "epoch": 3.06, + "learning_rate": 2.1604938271604937e-05, + "loss": 0.0, + "step": 1926 + }, + { + "epoch": 3.06, + "learning_rate": 2.158730158730159e-05, + "loss": 0.0, + "step": 1927 + }, + { + "epoch": 3.06, + "learning_rate": 2.1569664902998236e-05, + "loss": 0.0, + "step": 1928 + }, + { + "epoch": 3.06, + "learning_rate": 2.1552028218694887e-05, + "loss": 0.0, + "step": 1929 + }, + { + "epoch": 3.06, + "learning_rate": 2.1534391534391535e-05, + "loss": 0.0, + "step": 1930 + }, + { + "epoch": 3.07, + "learning_rate": 2.1516754850088183e-05, + "loss": 0.0, + "step": 1931 + }, + { + "epoch": 3.07, + "learning_rate": 2.1499118165784834e-05, + "loss": 0.0, + "step": 1932 + }, + { + "epoch": 3.07, + "learning_rate": 2.148148148148148e-05, + "loss": 0.0, + "step": 1933 + }, + { + "epoch": 3.07, + "learning_rate": 2.1463844797178133e-05, + "loss": 0.0, + "step": 1934 + }, + { + "epoch": 3.07, + "learning_rate": 2.144620811287478e-05, + "loss": 0.0, + "step": 1935 + }, + { + "epoch": 3.07, + "learning_rate": 2.1428571428571428e-05, + "loss": 0.0, + "step": 1936 + }, + { + "epoch": 3.07, + "learning_rate": 2.141093474426808e-05, + "loss": 0.0, + "step": 1937 + }, + { + "epoch": 3.08, + "learning_rate": 2.1393298059964727e-05, + "loss": 0.0, + "step": 1938 + }, + { + "epoch": 3.08, + "learning_rate": 2.1375661375661378e-05, + "loss": 0.0, + "step": 1939 + }, + { + "epoch": 3.08, + "learning_rate": 2.1358024691358026e-05, + "loss": 0.0, + "step": 1940 + }, + { + "epoch": 3.08, + "learning_rate": 2.1340388007054674e-05, + "loss": 0.0, + "step": 1941 + }, + { + "epoch": 3.08, + "learning_rate": 2.1322751322751325e-05, + "loss": 0.0, + "step": 1942 + }, + { + "epoch": 3.08, + "learning_rate": 2.1305114638447972e-05, + "loss": 0.0, + "step": 1943 + }, + { + "epoch": 3.09, + "learning_rate": 2.1287477954144623e-05, + "loss": 0.0, + "step": 1944 + }, + { + "epoch": 3.09, + "learning_rate": 2.126984126984127e-05, + "loss": 0.0, + "step": 1945 + }, + { + "epoch": 3.09, + "learning_rate": 2.125220458553792e-05, + "loss": 0.0, + "step": 1946 + }, + { + "epoch": 3.09, + "learning_rate": 2.123456790123457e-05, + "loss": 0.0, + "step": 1947 + }, + { + "epoch": 3.09, + "learning_rate": 2.1216931216931218e-05, + "loss": 0.0, + "step": 1948 + }, + { + "epoch": 3.09, + "learning_rate": 2.1199294532627865e-05, + "loss": 0.0, + "step": 1949 + }, + { + "epoch": 3.1, + "learning_rate": 2.1181657848324517e-05, + "loss": 0.0, + "step": 1950 + }, + { + "epoch": 3.1, + "learning_rate": 2.1164021164021164e-05, + "loss": 0.0, + "step": 1951 + }, + { + "epoch": 3.1, + "learning_rate": 2.1146384479717815e-05, + "loss": 0.0, + "step": 1952 + }, + { + "epoch": 3.1, + "learning_rate": 2.1128747795414463e-05, + "loss": 0.0, + "step": 1953 + }, + { + "epoch": 3.1, + "learning_rate": 2.111111111111111e-05, + "loss": 0.0, + "step": 1954 + }, + { + "epoch": 3.1, + "learning_rate": 2.1093474426807762e-05, + "loss": 0.0, + "step": 1955 + }, + { + "epoch": 3.1, + "learning_rate": 2.107583774250441e-05, + "loss": 0.0, + "step": 1956 + }, + { + "epoch": 3.11, + "learning_rate": 2.105820105820106e-05, + "loss": 0.0, + "step": 1957 + }, + { + "epoch": 3.11, + "learning_rate": 2.104056437389771e-05, + "loss": 0.0, + "step": 1958 + }, + { + "epoch": 3.11, + "learning_rate": 2.1022927689594356e-05, + "loss": 0.0, + "step": 1959 + }, + { + "epoch": 3.11, + "learning_rate": 2.1005291005291007e-05, + "loss": 0.0, + "step": 1960 + }, + { + "epoch": 3.11, + "learning_rate": 2.0987654320987655e-05, + "loss": 0.0, + "step": 1961 + }, + { + "epoch": 3.11, + "learning_rate": 2.0970017636684306e-05, + "loss": 0.0, + "step": 1962 + }, + { + "epoch": 3.12, + "learning_rate": 2.0952380952380954e-05, + "loss": 0.0, + "step": 1963 + }, + { + "epoch": 3.12, + "learning_rate": 2.09347442680776e-05, + "loss": 0.0, + "step": 1964 + }, + { + "epoch": 3.12, + "learning_rate": 2.091710758377425e-05, + "loss": 0.0, + "step": 1965 + }, + { + "epoch": 3.12, + "learning_rate": 2.08994708994709e-05, + "loss": 0.0, + "step": 1966 + }, + { + "epoch": 3.12, + "learning_rate": 2.088183421516755e-05, + "loss": 0.0, + "step": 1967 + }, + { + "epoch": 3.12, + "learning_rate": 2.08641975308642e-05, + "loss": 0.0, + "step": 1968 + }, + { + "epoch": 3.13, + "learning_rate": 2.0846560846560847e-05, + "loss": 0.0, + "step": 1969 + }, + { + "epoch": 3.13, + "learning_rate": 2.0828924162257495e-05, + "loss": 0.0, + "step": 1970 + }, + { + "epoch": 3.13, + "learning_rate": 2.0811287477954146e-05, + "loss": 0.0, + "step": 1971 + }, + { + "epoch": 3.13, + "learning_rate": 2.0793650793650797e-05, + "loss": 0.0, + "step": 1972 + }, + { + "epoch": 3.13, + "learning_rate": 2.0776014109347444e-05, + "loss": 0.0, + "step": 1973 + }, + { + "epoch": 3.13, + "learning_rate": 2.0758377425044092e-05, + "loss": 0.0, + "step": 1974 + }, + { + "epoch": 3.13, + "learning_rate": 2.074074074074074e-05, + "loss": 0.0, + "step": 1975 + }, + { + "epoch": 3.14, + "learning_rate": 2.072310405643739e-05, + "loss": 0.0, + "step": 1976 + }, + { + "epoch": 3.14, + "learning_rate": 2.0705467372134042e-05, + "loss": 0.0, + "step": 1977 + }, + { + "epoch": 3.14, + "learning_rate": 2.068783068783069e-05, + "loss": 0.0, + "step": 1978 + }, + { + "epoch": 3.14, + "learning_rate": 2.0670194003527338e-05, + "loss": 0.0, + "step": 1979 + }, + { + "epoch": 3.14, + "learning_rate": 2.0652557319223985e-05, + "loss": 0.0, + "step": 1980 + }, + { + "epoch": 3.14, + "learning_rate": 2.0634920634920636e-05, + "loss": 0.0, + "step": 1981 + }, + { + "epoch": 3.15, + "learning_rate": 2.0617283950617287e-05, + "loss": 0.0, + "step": 1982 + }, + { + "epoch": 3.15, + "learning_rate": 2.0599647266313935e-05, + "loss": 0.0, + "step": 1983 + }, + { + "epoch": 3.15, + "learning_rate": 2.0582010582010583e-05, + "loss": 0.0, + "step": 1984 + }, + { + "epoch": 3.15, + "learning_rate": 2.056437389770723e-05, + "loss": 0.0, + "step": 1985 + }, + { + "epoch": 3.15, + "learning_rate": 2.054673721340388e-05, + "loss": 0.0, + "step": 1986 + }, + { + "epoch": 3.15, + "learning_rate": 2.052910052910053e-05, + "loss": 0.0, + "step": 1987 + }, + { + "epoch": 3.16, + "learning_rate": 2.051146384479718e-05, + "loss": 0.0, + "step": 1988 + }, + { + "epoch": 3.16, + "learning_rate": 2.0493827160493828e-05, + "loss": 0.0, + "step": 1989 + }, + { + "epoch": 3.16, + "learning_rate": 2.0476190476190476e-05, + "loss": 0.0, + "step": 1990 + }, + { + "epoch": 3.16, + "learning_rate": 2.0458553791887124e-05, + "loss": 0.0, + "step": 1991 + }, + { + "epoch": 3.16, + "learning_rate": 2.0440917107583775e-05, + "loss": 0.0, + "step": 1992 + }, + { + "epoch": 3.16, + "learning_rate": 2.0423280423280426e-05, + "loss": 0.0, + "step": 1993 + }, + { + "epoch": 3.17, + "learning_rate": 2.0405643738977074e-05, + "loss": 0.0, + "step": 1994 + }, + { + "epoch": 3.17, + "learning_rate": 2.038800705467372e-05, + "loss": 0.0, + "step": 1995 + }, + { + "epoch": 3.17, + "learning_rate": 2.037037037037037e-05, + "loss": 0.0, + "step": 1996 + }, + { + "epoch": 3.17, + "learning_rate": 2.035273368606702e-05, + "loss": 0.0, + "step": 1997 + }, + { + "epoch": 3.17, + "learning_rate": 2.033509700176367e-05, + "loss": 0.0, + "step": 1998 + }, + { + "epoch": 3.17, + "learning_rate": 2.031746031746032e-05, + "loss": 0.0, + "step": 1999 + }, + { + "epoch": 3.17, + "learning_rate": 2.0299823633156967e-05, + "loss": 0.0, + "step": 2000 + }, + { + "epoch": 3.18, + "learning_rate": 2.0282186948853614e-05, + "loss": 0.0, + "step": 2001 + }, + { + "epoch": 3.18, + "learning_rate": 2.0264550264550266e-05, + "loss": 0.0, + "step": 2002 + }, + { + "epoch": 3.18, + "learning_rate": 2.0246913580246917e-05, + "loss": 0.0, + "step": 2003 + }, + { + "epoch": 3.18, + "learning_rate": 2.0229276895943564e-05, + "loss": 0.0, + "step": 2004 + }, + { + "epoch": 3.18, + "learning_rate": 2.0211640211640212e-05, + "loss": 0.0, + "step": 2005 + }, + { + "epoch": 3.18, + "learning_rate": 2.019400352733686e-05, + "loss": 0.0, + "step": 2006 + }, + { + "epoch": 3.19, + "learning_rate": 2.017636684303351e-05, + "loss": 0.0, + "step": 2007 + }, + { + "epoch": 3.19, + "learning_rate": 2.015873015873016e-05, + "loss": 0.0001, + "step": 2008 + }, + { + "epoch": 3.19, + "learning_rate": 2.014109347442681e-05, + "loss": 0.0, + "step": 2009 + }, + { + "epoch": 3.19, + "learning_rate": 2.0123456790123457e-05, + "loss": 0.0, + "step": 2010 + }, + { + "epoch": 3.19, + "learning_rate": 2.0105820105820105e-05, + "loss": 0.0, + "step": 2011 + }, + { + "epoch": 3.19, + "learning_rate": 2.0088183421516756e-05, + "loss": 0.0, + "step": 2012 + }, + { + "epoch": 3.2, + "learning_rate": 2.0070546737213404e-05, + "loss": 0.0, + "step": 2013 + }, + { + "epoch": 3.2, + "learning_rate": 2.0052910052910055e-05, + "loss": 0.0, + "step": 2014 + }, + { + "epoch": 3.2, + "learning_rate": 2.0035273368606703e-05, + "loss": 0.0, + "step": 2015 + }, + { + "epoch": 3.2, + "learning_rate": 2.001763668430335e-05, + "loss": 0.0, + "step": 2016 + }, + { + "epoch": 3.2, + "learning_rate": 2e-05, + "loss": 0.0, + "step": 2017 + }, + { + "epoch": 3.2, + "learning_rate": 1.998236331569665e-05, + "loss": 0.0, + "step": 2018 + }, + { + "epoch": 3.2, + "learning_rate": 1.99647266313933e-05, + "loss": 0.0, + "step": 2019 + }, + { + "epoch": 3.21, + "learning_rate": 1.9947089947089948e-05, + "loss": 0.0, + "step": 2020 + }, + { + "epoch": 3.21, + "learning_rate": 1.9929453262786596e-05, + "loss": 0.0, + "step": 2021 + }, + { + "epoch": 3.21, + "learning_rate": 1.9911816578483247e-05, + "loss": 0.0, + "step": 2022 + }, + { + "epoch": 3.21, + "learning_rate": 1.9894179894179895e-05, + "loss": 0.0, + "step": 2023 + }, + { + "epoch": 3.21, + "learning_rate": 1.9876543209876546e-05, + "loss": 0.0, + "step": 2024 + }, + { + "epoch": 3.21, + "learning_rate": 1.9858906525573194e-05, + "loss": 0.0, + "step": 2025 + }, + { + "epoch": 3.22, + "learning_rate": 1.984126984126984e-05, + "loss": 0.0, + "step": 2026 + }, + { + "epoch": 3.22, + "learning_rate": 1.9823633156966492e-05, + "loss": 0.0, + "step": 2027 + }, + { + "epoch": 3.22, + "learning_rate": 1.980599647266314e-05, + "loss": 0.0, + "step": 2028 + }, + { + "epoch": 3.22, + "learning_rate": 1.9788359788359788e-05, + "loss": 0.0, + "step": 2029 + }, + { + "epoch": 3.22, + "learning_rate": 1.977072310405644e-05, + "loss": 0.0, + "step": 2030 + }, + { + "epoch": 3.22, + "learning_rate": 1.9753086419753087e-05, + "loss": 0.0, + "step": 2031 + }, + { + "epoch": 3.23, + "learning_rate": 1.9735449735449738e-05, + "loss": 0.0, + "step": 2032 + }, + { + "epoch": 3.23, + "learning_rate": 1.9717813051146385e-05, + "loss": 0.0, + "step": 2033 + }, + { + "epoch": 3.23, + "learning_rate": 1.9700176366843033e-05, + "loss": 0.0, + "step": 2034 + }, + { + "epoch": 3.23, + "learning_rate": 1.9682539682539684e-05, + "loss": 0.0, + "step": 2035 + }, + { + "epoch": 3.23, + "learning_rate": 1.9664902998236332e-05, + "loss": 0.0, + "step": 2036 + }, + { + "epoch": 3.23, + "learning_rate": 1.9647266313932983e-05, + "loss": 0.0, + "step": 2037 + }, + { + "epoch": 3.23, + "learning_rate": 1.962962962962963e-05, + "loss": 0.0, + "step": 2038 + }, + { + "epoch": 3.24, + "learning_rate": 1.961199294532628e-05, + "loss": 0.0, + "step": 2039 + }, + { + "epoch": 3.24, + "learning_rate": 1.959435626102293e-05, + "loss": 0.0, + "step": 2040 + }, + { + "epoch": 3.24, + "learning_rate": 1.9576719576719577e-05, + "loss": 0.0, + "step": 2041 + }, + { + "epoch": 3.24, + "learning_rate": 1.955908289241623e-05, + "loss": 0.0, + "step": 2042 + }, + { + "epoch": 3.24, + "learning_rate": 1.9541446208112876e-05, + "loss": 0.0, + "step": 2043 + }, + { + "epoch": 3.24, + "learning_rate": 1.9523809523809524e-05, + "loss": 0.0, + "step": 2044 + }, + { + "epoch": 3.25, + "learning_rate": 1.950617283950617e-05, + "loss": 0.0, + "step": 2045 + }, + { + "epoch": 3.25, + "learning_rate": 1.9488536155202823e-05, + "loss": 0.0, + "step": 2046 + }, + { + "epoch": 3.25, + "learning_rate": 1.9470899470899474e-05, + "loss": 0.0, + "step": 2047 + }, + { + "epoch": 3.25, + "learning_rate": 1.945326278659612e-05, + "loss": 0.0, + "step": 2048 + }, + { + "epoch": 3.25, + "learning_rate": 1.943562610229277e-05, + "loss": 0.0, + "step": 2049 + }, + { + "epoch": 3.25, + "learning_rate": 1.9417989417989417e-05, + "loss": 0.0, + "step": 2050 + }, + { + "epoch": 3.26, + "learning_rate": 1.9400352733686068e-05, + "loss": 0.0, + "step": 2051 + }, + { + "epoch": 3.26, + "learning_rate": 1.938271604938272e-05, + "loss": 0.0, + "step": 2052 + }, + { + "epoch": 3.26, + "learning_rate": 1.9365079365079367e-05, + "loss": 0.0, + "step": 2053 + }, + { + "epoch": 3.26, + "learning_rate": 1.9347442680776015e-05, + "loss": 0.0, + "step": 2054 + }, + { + "epoch": 3.26, + "learning_rate": 1.9329805996472662e-05, + "loss": 0.0, + "step": 2055 + }, + { + "epoch": 3.26, + "learning_rate": 1.9312169312169313e-05, + "loss": 0.0, + "step": 2056 + }, + { + "epoch": 3.27, + "learning_rate": 1.9294532627865965e-05, + "loss": 0.0, + "step": 2057 + }, + { + "epoch": 3.27, + "learning_rate": 1.9276895943562612e-05, + "loss": 0.0, + "step": 2058 + }, + { + "epoch": 3.27, + "learning_rate": 1.925925925925926e-05, + "loss": 0.0, + "step": 2059 + }, + { + "epoch": 3.27, + "learning_rate": 1.9241622574955908e-05, + "loss": 0.0, + "step": 2060 + }, + { + "epoch": 3.27, + "learning_rate": 1.922398589065256e-05, + "loss": 0.0, + "step": 2061 + }, + { + "epoch": 3.27, + "learning_rate": 1.920634920634921e-05, + "loss": 0.0, + "step": 2062 + }, + { + "epoch": 3.27, + "learning_rate": 1.9188712522045858e-05, + "loss": 0.0, + "step": 2063 + }, + { + "epoch": 3.28, + "learning_rate": 1.9171075837742505e-05, + "loss": 0.0, + "step": 2064 + }, + { + "epoch": 3.28, + "learning_rate": 1.9153439153439153e-05, + "loss": 0.0, + "step": 2065 + }, + { + "epoch": 3.28, + "learning_rate": 1.91358024691358e-05, + "loss": 0.0, + "step": 2066 + }, + { + "epoch": 3.28, + "learning_rate": 1.9118165784832455e-05, + "loss": 0.0, + "step": 2067 + }, + { + "epoch": 3.28, + "learning_rate": 1.9100529100529103e-05, + "loss": 0.0, + "step": 2068 + }, + { + "epoch": 3.28, + "learning_rate": 1.908289241622575e-05, + "loss": 0.0, + "step": 2069 + }, + { + "epoch": 3.29, + "learning_rate": 1.90652557319224e-05, + "loss": 0.0, + "step": 2070 + }, + { + "epoch": 3.29, + "learning_rate": 1.9047619047619046e-05, + "loss": 0.0, + "step": 2071 + }, + { + "epoch": 3.29, + "learning_rate": 1.9029982363315697e-05, + "loss": 0.0, + "step": 2072 + }, + { + "epoch": 3.29, + "learning_rate": 1.901234567901235e-05, + "loss": 0.0, + "step": 2073 + }, + { + "epoch": 3.29, + "learning_rate": 1.8994708994708996e-05, + "loss": 0.0, + "step": 2074 + }, + { + "epoch": 3.29, + "learning_rate": 1.8977072310405644e-05, + "loss": 0.0, + "step": 2075 + }, + { + "epoch": 3.3, + "learning_rate": 1.895943562610229e-05, + "loss": 0.0, + "step": 2076 + }, + { + "epoch": 3.3, + "learning_rate": 1.8941798941798943e-05, + "loss": 0.0, + "step": 2077 + }, + { + "epoch": 3.3, + "learning_rate": 1.8924162257495594e-05, + "loss": 0.0, + "step": 2078 + }, + { + "epoch": 3.3, + "learning_rate": 1.890652557319224e-05, + "loss": 0.0, + "step": 2079 + }, + { + "epoch": 3.3, + "learning_rate": 1.888888888888889e-05, + "loss": 0.0, + "step": 2080 + }, + { + "epoch": 3.3, + "learning_rate": 1.8871252204585537e-05, + "loss": 0.0, + "step": 2081 + }, + { + "epoch": 3.3, + "learning_rate": 1.8853615520282188e-05, + "loss": 0.0, + "step": 2082 + }, + { + "epoch": 3.31, + "learning_rate": 1.883597883597884e-05, + "loss": 0.0, + "step": 2083 + }, + { + "epoch": 3.31, + "learning_rate": 1.8818342151675487e-05, + "loss": 0.0, + "step": 2084 + }, + { + "epoch": 3.31, + "learning_rate": 1.8800705467372134e-05, + "loss": 0.0, + "step": 2085 + }, + { + "epoch": 3.31, + "learning_rate": 1.8783068783068782e-05, + "loss": 0.0, + "step": 2086 + }, + { + "epoch": 3.31, + "learning_rate": 1.8765432098765433e-05, + "loss": 0.0, + "step": 2087 + }, + { + "epoch": 3.31, + "learning_rate": 1.874779541446208e-05, + "loss": 0.0, + "step": 2088 + }, + { + "epoch": 3.32, + "learning_rate": 1.8730158730158732e-05, + "loss": 0.0, + "step": 2089 + }, + { + "epoch": 3.32, + "learning_rate": 1.871252204585538e-05, + "loss": 0.0, + "step": 2090 + }, + { + "epoch": 3.32, + "learning_rate": 1.8694885361552028e-05, + "loss": 0.0, + "step": 2091 + }, + { + "epoch": 3.32, + "learning_rate": 1.867724867724868e-05, + "loss": 0.0, + "step": 2092 + }, + { + "epoch": 3.32, + "learning_rate": 1.8659611992945326e-05, + "loss": 0.0, + "step": 2093 + }, + { + "epoch": 3.32, + "learning_rate": 1.8641975308641977e-05, + "loss": 0.0, + "step": 2094 + }, + { + "epoch": 3.33, + "learning_rate": 1.8624338624338625e-05, + "loss": 0.0, + "step": 2095 + }, + { + "epoch": 3.33, + "learning_rate": 1.8606701940035273e-05, + "loss": 0.0, + "step": 2096 + }, + { + "epoch": 3.33, + "learning_rate": 1.8589065255731924e-05, + "loss": 0.0, + "step": 2097 + }, + { + "epoch": 3.33, + "learning_rate": 1.8571428571428572e-05, + "loss": 0.0, + "step": 2098 + }, + { + "epoch": 3.33, + "learning_rate": 1.8553791887125223e-05, + "loss": 0.0, + "step": 2099 + }, + { + "epoch": 3.33, + "learning_rate": 1.853615520282187e-05, + "loss": 0.0, + "step": 2100 + }, + { + "epoch": 3.33, + "learning_rate": 1.8518518518518518e-05, + "loss": 0.0, + "step": 2101 + }, + { + "epoch": 3.34, + "learning_rate": 1.850088183421517e-05, + "loss": 0.0, + "step": 2102 + }, + { + "epoch": 3.34, + "learning_rate": 1.8483245149911817e-05, + "loss": 0.0, + "step": 2103 + }, + { + "epoch": 3.34, + "learning_rate": 1.8465608465608468e-05, + "loss": 0.0, + "step": 2104 + }, + { + "epoch": 3.34, + "learning_rate": 1.8447971781305116e-05, + "loss": 0.0, + "step": 2105 + }, + { + "epoch": 3.34, + "learning_rate": 1.8430335097001764e-05, + "loss": 0.0, + "step": 2106 + }, + { + "epoch": 3.34, + "learning_rate": 1.8412698412698415e-05, + "loss": 0.0, + "step": 2107 + }, + { + "epoch": 3.35, + "learning_rate": 1.8395061728395062e-05, + "loss": 0.0, + "step": 2108 + }, + { + "epoch": 3.35, + "learning_rate": 1.837742504409171e-05, + "loss": 0.0, + "step": 2109 + }, + { + "epoch": 3.35, + "learning_rate": 1.835978835978836e-05, + "loss": 0.0, + "step": 2110 + }, + { + "epoch": 3.35, + "learning_rate": 1.834215167548501e-05, + "loss": 0.0, + "step": 2111 + }, + { + "epoch": 3.35, + "learning_rate": 1.832451499118166e-05, + "loss": 0.0, + "step": 2112 + }, + { + "epoch": 3.35, + "learning_rate": 1.8306878306878308e-05, + "loss": 0.0, + "step": 2113 + }, + { + "epoch": 3.36, + "learning_rate": 1.8289241622574956e-05, + "loss": 0.0, + "step": 2114 + }, + { + "epoch": 3.36, + "learning_rate": 1.8271604938271607e-05, + "loss": 0.0, + "step": 2115 + }, + { + "epoch": 3.36, + "learning_rate": 1.8253968253968254e-05, + "loss": 0.0, + "step": 2116 + }, + { + "epoch": 3.36, + "learning_rate": 1.8236331569664905e-05, + "loss": 0.0, + "step": 2117 + }, + { + "epoch": 3.36, + "learning_rate": 1.8218694885361553e-05, + "loss": 0.0, + "step": 2118 + }, + { + "epoch": 3.36, + "learning_rate": 1.82010582010582e-05, + "loss": 0.0, + "step": 2119 + }, + { + "epoch": 3.37, + "learning_rate": 1.8183421516754852e-05, + "loss": 0.0, + "step": 2120 + }, + { + "epoch": 3.37, + "learning_rate": 1.81657848324515e-05, + "loss": 0.0, + "step": 2121 + }, + { + "epoch": 3.37, + "learning_rate": 1.814814814814815e-05, + "loss": 0.0, + "step": 2122 + }, + { + "epoch": 3.37, + "learning_rate": 1.81305114638448e-05, + "loss": 0.0, + "step": 2123 + }, + { + "epoch": 3.37, + "learning_rate": 1.8112874779541446e-05, + "loss": 0.0, + "step": 2124 + }, + { + "epoch": 3.37, + "learning_rate": 1.8095238095238094e-05, + "loss": 0.0, + "step": 2125 + }, + { + "epoch": 3.37, + "learning_rate": 1.8077601410934745e-05, + "loss": 0.0, + "step": 2126 + }, + { + "epoch": 3.38, + "learning_rate": 1.8059964726631396e-05, + "loss": 0.0, + "step": 2127 + }, + { + "epoch": 3.38, + "learning_rate": 1.8042328042328044e-05, + "loss": 0.0, + "step": 2128 + }, + { + "epoch": 3.38, + "learning_rate": 1.802469135802469e-05, + "loss": 0.0, + "step": 2129 + }, + { + "epoch": 3.38, + "learning_rate": 1.800705467372134e-05, + "loss": 0.0, + "step": 2130 + }, + { + "epoch": 3.38, + "learning_rate": 1.798941798941799e-05, + "loss": 0.0, + "step": 2131 + }, + { + "epoch": 3.38, + "learning_rate": 1.797178130511464e-05, + "loss": 0.0, + "step": 2132 + }, + { + "epoch": 3.39, + "learning_rate": 1.795414462081129e-05, + "loss": 0.0, + "step": 2133 + }, + { + "epoch": 3.39, + "learning_rate": 1.7936507936507937e-05, + "loss": 0.0, + "step": 2134 + }, + { + "epoch": 3.39, + "learning_rate": 1.7918871252204585e-05, + "loss": 0.0, + "step": 2135 + }, + { + "epoch": 3.39, + "learning_rate": 1.7901234567901236e-05, + "loss": 0.0, + "step": 2136 + }, + { + "epoch": 3.39, + "learning_rate": 1.7883597883597887e-05, + "loss": 0.0, + "step": 2137 + }, + { + "epoch": 3.39, + "learning_rate": 1.7865961199294535e-05, + "loss": 0.0, + "step": 2138 + }, + { + "epoch": 3.4, + "learning_rate": 1.7848324514991182e-05, + "loss": 0.0, + "step": 2139 + }, + { + "epoch": 3.4, + "learning_rate": 1.783068783068783e-05, + "loss": 0.0, + "step": 2140 + }, + { + "epoch": 3.4, + "learning_rate": 1.781305114638448e-05, + "loss": 0.0, + "step": 2141 + }, + { + "epoch": 3.4, + "learning_rate": 1.7795414462081132e-05, + "loss": 0.0, + "step": 2142 + }, + { + "epoch": 3.4, + "learning_rate": 1.777777777777778e-05, + "loss": 0.0, + "step": 2143 + }, + { + "epoch": 3.4, + "learning_rate": 1.7760141093474428e-05, + "loss": 0.0, + "step": 2144 + }, + { + "epoch": 3.4, + "learning_rate": 1.7742504409171075e-05, + "loss": 0.0, + "step": 2145 + }, + { + "epoch": 3.41, + "learning_rate": 1.7724867724867723e-05, + "loss": 0.0, + "step": 2146 + }, + { + "epoch": 3.41, + "learning_rate": 1.7707231040564378e-05, + "loss": 0.0, + "step": 2147 + }, + { + "epoch": 3.41, + "learning_rate": 1.7689594356261025e-05, + "loss": 0.0, + "step": 2148 + }, + { + "epoch": 3.41, + "learning_rate": 1.7671957671957673e-05, + "loss": 0.0, + "step": 2149 + }, + { + "epoch": 3.41, + "learning_rate": 1.765432098765432e-05, + "loss": 0.0, + "step": 2150 + }, + { + "epoch": 3.41, + "learning_rate": 1.763668430335097e-05, + "loss": 0.0, + "step": 2151 + }, + { + "epoch": 3.42, + "learning_rate": 1.761904761904762e-05, + "loss": 0.0, + "step": 2152 + }, + { + "epoch": 3.42, + "learning_rate": 1.760141093474427e-05, + "loss": 0.0, + "step": 2153 + }, + { + "epoch": 3.42, + "learning_rate": 1.758377425044092e-05, + "loss": 0.0, + "step": 2154 + }, + { + "epoch": 3.42, + "learning_rate": 1.7566137566137566e-05, + "loss": 0.0001, + "step": 2155 + }, + { + "epoch": 3.42, + "learning_rate": 1.7548500881834214e-05, + "loss": 0.0, + "step": 2156 + }, + { + "epoch": 3.42, + "learning_rate": 1.7530864197530865e-05, + "loss": 0.0, + "step": 2157 + }, + { + "epoch": 3.43, + "learning_rate": 1.7513227513227516e-05, + "loss": 0.0, + "step": 2158 + }, + { + "epoch": 3.43, + "learning_rate": 1.7495590828924164e-05, + "loss": 0.0, + "step": 2159 + }, + { + "epoch": 3.43, + "learning_rate": 1.747795414462081e-05, + "loss": 0.0, + "step": 2160 + }, + { + "epoch": 3.43, + "learning_rate": 1.746031746031746e-05, + "loss": 0.1389, + "step": 2161 + }, + { + "epoch": 3.43, + "learning_rate": 1.744268077601411e-05, + "loss": 0.0, + "step": 2162 + }, + { + "epoch": 3.43, + "learning_rate": 1.742504409171076e-05, + "loss": 0.0, + "step": 2163 + }, + { + "epoch": 3.43, + "learning_rate": 1.740740740740741e-05, + "loss": 0.0, + "step": 2164 + }, + { + "epoch": 3.44, + "learning_rate": 1.7389770723104057e-05, + "loss": 0.0, + "step": 2165 + }, + { + "epoch": 3.44, + "learning_rate": 1.7372134038800705e-05, + "loss": 0.0, + "step": 2166 + }, + { + "epoch": 3.44, + "learning_rate": 1.7354497354497356e-05, + "loss": 0.0, + "step": 2167 + }, + { + "epoch": 3.44, + "learning_rate": 1.7336860670194003e-05, + "loss": 0.0, + "step": 2168 + }, + { + "epoch": 3.44, + "learning_rate": 1.7319223985890655e-05, + "loss": 0.0, + "step": 2169 + }, + { + "epoch": 3.44, + "learning_rate": 1.7301587301587302e-05, + "loss": 0.0, + "step": 2170 + }, + { + "epoch": 3.45, + "learning_rate": 1.728395061728395e-05, + "loss": 0.0, + "step": 2171 + }, + { + "epoch": 3.45, + "learning_rate": 1.72663139329806e-05, + "loss": 0.0, + "step": 2172 + }, + { + "epoch": 3.45, + "learning_rate": 1.724867724867725e-05, + "loss": 0.0, + "step": 2173 + }, + { + "epoch": 3.45, + "learning_rate": 1.72310405643739e-05, + "loss": 0.0, + "step": 2174 + }, + { + "epoch": 3.45, + "learning_rate": 1.7213403880070548e-05, + "loss": 0.0, + "step": 2175 + }, + { + "epoch": 3.45, + "learning_rate": 1.7195767195767195e-05, + "loss": 0.0, + "step": 2176 + }, + { + "epoch": 3.46, + "learning_rate": 1.7178130511463846e-05, + "loss": 0.0, + "step": 2177 + }, + { + "epoch": 3.46, + "learning_rate": 1.7160493827160494e-05, + "loss": 0.0, + "step": 2178 + }, + { + "epoch": 3.46, + "learning_rate": 1.7142857142857145e-05, + "loss": 0.0, + "step": 2179 + }, + { + "epoch": 3.46, + "learning_rate": 1.7125220458553793e-05, + "loss": 0.0, + "step": 2180 + }, + { + "epoch": 3.46, + "learning_rate": 1.710758377425044e-05, + "loss": 0.0, + "step": 2181 + }, + { + "epoch": 3.46, + "learning_rate": 1.7089947089947092e-05, + "loss": 0.0, + "step": 2182 + }, + { + "epoch": 3.47, + "learning_rate": 1.707231040564374e-05, + "loss": 0.0, + "step": 2183 + }, + { + "epoch": 3.47, + "learning_rate": 1.705467372134039e-05, + "loss": 0.0, + "step": 2184 + }, + { + "epoch": 3.47, + "learning_rate": 1.7037037037037038e-05, + "loss": 0.0, + "step": 2185 + }, + { + "epoch": 3.47, + "learning_rate": 1.7019400352733686e-05, + "loss": 0.0, + "step": 2186 + }, + { + "epoch": 3.47, + "learning_rate": 1.7001763668430337e-05, + "loss": 0.0, + "step": 2187 + }, + { + "epoch": 3.47, + "learning_rate": 1.6984126984126985e-05, + "loss": 0.0, + "step": 2188 + }, + { + "epoch": 3.47, + "learning_rate": 1.6966490299823633e-05, + "loss": 0.0, + "step": 2189 + }, + { + "epoch": 3.48, + "learning_rate": 1.6948853615520284e-05, + "loss": 0.0, + "step": 2190 + }, + { + "epoch": 3.48, + "learning_rate": 1.693121693121693e-05, + "loss": 0.0, + "step": 2191 + }, + { + "epoch": 3.48, + "learning_rate": 1.6913580246913582e-05, + "loss": 0.0, + "step": 2192 + }, + { + "epoch": 3.48, + "learning_rate": 1.689594356261023e-05, + "loss": 0.0, + "step": 2193 + }, + { + "epoch": 3.48, + "learning_rate": 1.6878306878306878e-05, + "loss": 0.0, + "step": 2194 + }, + { + "epoch": 3.48, + "learning_rate": 1.686067019400353e-05, + "loss": 0.0, + "step": 2195 + }, + { + "epoch": 3.49, + "learning_rate": 1.6843033509700177e-05, + "loss": 0.0, + "step": 2196 + }, + { + "epoch": 3.49, + "learning_rate": 1.6825396825396828e-05, + "loss": 0.0, + "step": 2197 + }, + { + "epoch": 3.49, + "learning_rate": 1.6807760141093476e-05, + "loss": 0.0, + "step": 2198 + }, + { + "epoch": 3.49, + "learning_rate": 1.6790123456790123e-05, + "loss": 0.0, + "step": 2199 + }, + { + "epoch": 3.49, + "learning_rate": 1.6772486772486774e-05, + "loss": 0.0, + "step": 2200 + }, + { + "epoch": 3.49, + "learning_rate": 1.6754850088183422e-05, + "loss": 0.0, + "step": 2201 + }, + { + "epoch": 3.5, + "learning_rate": 1.6737213403880073e-05, + "loss": 0.0007, + "step": 2202 + }, + { + "epoch": 3.5, + "learning_rate": 1.671957671957672e-05, + "loss": 0.0, + "step": 2203 + }, + { + "epoch": 3.5, + "learning_rate": 1.670194003527337e-05, + "loss": 0.0, + "step": 2204 + }, + { + "epoch": 3.5, + "learning_rate": 1.668430335097002e-05, + "loss": 0.0, + "step": 2205 + }, + { + "epoch": 3.5, + "learning_rate": 1.6666666666666667e-05, + "loss": 0.0, + "step": 2206 + }, + { + "epoch": 3.5, + "learning_rate": 1.664902998236332e-05, + "loss": 0.0, + "step": 2207 + }, + { + "epoch": 3.5, + "learning_rate": 1.6631393298059966e-05, + "loss": 0.0, + "step": 2208 + }, + { + "epoch": 3.51, + "learning_rate": 1.6613756613756614e-05, + "loss": 0.0, + "step": 2209 + }, + { + "epoch": 3.51, + "learning_rate": 1.6596119929453262e-05, + "loss": 0.0, + "step": 2210 + }, + { + "epoch": 3.51, + "learning_rate": 1.6578483245149913e-05, + "loss": 0.0, + "step": 2211 + }, + { + "epoch": 3.51, + "learning_rate": 1.6560846560846564e-05, + "loss": 0.0, + "step": 2212 + }, + { + "epoch": 3.51, + "learning_rate": 1.654320987654321e-05, + "loss": 0.0, + "step": 2213 + }, + { + "epoch": 3.51, + "learning_rate": 1.652557319223986e-05, + "loss": 0.0, + "step": 2214 + }, + { + "epoch": 3.52, + "learning_rate": 1.6507936507936507e-05, + "loss": 0.0, + "step": 2215 + }, + { + "epoch": 3.52, + "learning_rate": 1.6490299823633158e-05, + "loss": 0.0, + "step": 2216 + }, + { + "epoch": 3.52, + "learning_rate": 1.647266313932981e-05, + "loss": 0.0, + "step": 2217 + }, + { + "epoch": 3.52, + "learning_rate": 1.6455026455026457e-05, + "loss": 0.0, + "step": 2218 + }, + { + "epoch": 3.52, + "learning_rate": 1.6437389770723105e-05, + "loss": 0.0, + "step": 2219 + }, + { + "epoch": 3.52, + "learning_rate": 1.6419753086419752e-05, + "loss": 0.0, + "step": 2220 + }, + { + "epoch": 3.53, + "learning_rate": 1.6402116402116404e-05, + "loss": 0.0, + "step": 2221 + }, + { + "epoch": 3.53, + "learning_rate": 1.6384479717813055e-05, + "loss": 0.0008, + "step": 2222 + }, + { + "epoch": 3.53, + "learning_rate": 1.6366843033509702e-05, + "loss": 0.0, + "step": 2223 + }, + { + "epoch": 3.53, + "learning_rate": 1.634920634920635e-05, + "loss": 0.0, + "step": 2224 + }, + { + "epoch": 3.53, + "learning_rate": 1.6331569664902998e-05, + "loss": 0.0, + "step": 2225 + }, + { + "epoch": 3.53, + "learning_rate": 1.6313932980599646e-05, + "loss": 0.0, + "step": 2226 + }, + { + "epoch": 3.53, + "learning_rate": 1.62962962962963e-05, + "loss": 0.0, + "step": 2227 + }, + { + "epoch": 3.54, + "learning_rate": 1.6278659611992948e-05, + "loss": 0.0, + "step": 2228 + }, + { + "epoch": 3.54, + "learning_rate": 1.6261022927689595e-05, + "loss": 0.0, + "step": 2229 + }, + { + "epoch": 3.54, + "learning_rate": 1.6243386243386243e-05, + "loss": 0.0, + "step": 2230 + }, + { + "epoch": 3.54, + "learning_rate": 1.622574955908289e-05, + "loss": 0.0001, + "step": 2231 + }, + { + "epoch": 3.54, + "learning_rate": 1.6208112874779542e-05, + "loss": 0.0, + "step": 2232 + }, + { + "epoch": 3.54, + "learning_rate": 1.6190476190476193e-05, + "loss": 0.0, + "step": 2233 + }, + { + "epoch": 3.55, + "learning_rate": 1.617283950617284e-05, + "loss": 0.0, + "step": 2234 + }, + { + "epoch": 3.55, + "learning_rate": 1.615520282186949e-05, + "loss": 0.0, + "step": 2235 + }, + { + "epoch": 3.55, + "learning_rate": 1.6137566137566136e-05, + "loss": 0.0, + "step": 2236 + }, + { + "epoch": 3.55, + "learning_rate": 1.6119929453262787e-05, + "loss": 0.0, + "step": 2237 + }, + { + "epoch": 3.55, + "learning_rate": 1.610229276895944e-05, + "loss": 0.0, + "step": 2238 + }, + { + "epoch": 3.55, + "learning_rate": 1.6084656084656086e-05, + "loss": 0.0, + "step": 2239 + }, + { + "epoch": 3.56, + "learning_rate": 1.6067019400352734e-05, + "loss": 0.0, + "step": 2240 + }, + { + "epoch": 3.56, + "learning_rate": 1.604938271604938e-05, + "loss": 0.0, + "step": 2241 + }, + { + "epoch": 3.56, + "learning_rate": 1.6031746031746033e-05, + "loss": 0.0, + "step": 2242 + }, + { + "epoch": 3.56, + "learning_rate": 1.6014109347442684e-05, + "loss": 0.0, + "step": 2243 + }, + { + "epoch": 3.56, + "learning_rate": 1.599647266313933e-05, + "loss": 0.0, + "step": 2244 + }, + { + "epoch": 3.56, + "learning_rate": 1.597883597883598e-05, + "loss": 0.0, + "step": 2245 + }, + { + "epoch": 3.57, + "learning_rate": 1.5961199294532627e-05, + "loss": 0.0, + "step": 2246 + }, + { + "epoch": 3.57, + "learning_rate": 1.5943562610229278e-05, + "loss": 0.0, + "step": 2247 + }, + { + "epoch": 3.57, + "learning_rate": 1.5925925925925926e-05, + "loss": 0.0, + "step": 2248 + }, + { + "epoch": 3.57, + "learning_rate": 1.5908289241622577e-05, + "loss": 0.0, + "step": 2249 + }, + { + "epoch": 3.57, + "learning_rate": 1.5890652557319225e-05, + "loss": 0.0, + "step": 2250 + }, + { + "epoch": 3.57, + "learning_rate": 1.5873015873015872e-05, + "loss": 0.0, + "step": 2251 + }, + { + "epoch": 3.57, + "learning_rate": 1.5855379188712523e-05, + "loss": 0.0, + "step": 2252 + }, + { + "epoch": 3.58, + "learning_rate": 1.583774250440917e-05, + "loss": 0.0, + "step": 2253 + }, + { + "epoch": 3.58, + "learning_rate": 1.5820105820105822e-05, + "loss": 0.0, + "step": 2254 + }, + { + "epoch": 3.58, + "learning_rate": 1.580246913580247e-05, + "loss": 0.0, + "step": 2255 + }, + { + "epoch": 3.58, + "learning_rate": 1.5784832451499118e-05, + "loss": 0.0, + "step": 2256 + }, + { + "epoch": 3.58, + "learning_rate": 1.576719576719577e-05, + "loss": 0.0, + "step": 2257 + }, + { + "epoch": 3.58, + "learning_rate": 1.5749559082892417e-05, + "loss": 0.0, + "step": 2258 + }, + { + "epoch": 3.59, + "learning_rate": 1.5731922398589068e-05, + "loss": 0.0, + "step": 2259 + }, + { + "epoch": 3.59, + "learning_rate": 1.5714285714285715e-05, + "loss": 0.0, + "step": 2260 + }, + { + "epoch": 3.59, + "learning_rate": 1.5696649029982363e-05, + "loss": 0.0, + "step": 2261 + }, + { + "epoch": 3.59, + "learning_rate": 1.5679012345679014e-05, + "loss": 0.0, + "step": 2262 + }, + { + "epoch": 3.59, + "learning_rate": 1.5661375661375662e-05, + "loss": 0.0, + "step": 2263 + }, + { + "epoch": 3.59, + "learning_rate": 1.5643738977072313e-05, + "loss": 0.0001, + "step": 2264 + }, + { + "epoch": 3.6, + "learning_rate": 1.562610229276896e-05, + "loss": 0.0, + "step": 2265 + }, + { + "epoch": 3.6, + "learning_rate": 1.560846560846561e-05, + "loss": 0.0, + "step": 2266 + }, + { + "epoch": 3.6, + "learning_rate": 1.559082892416226e-05, + "loss": 0.0, + "step": 2267 + }, + { + "epoch": 3.6, + "learning_rate": 1.5573192239858907e-05, + "loss": 0.0, + "step": 2268 + }, + { + "epoch": 3.6, + "learning_rate": 1.5555555555555555e-05, + "loss": 0.0004, + "step": 2269 + }, + { + "epoch": 3.6, + "learning_rate": 1.5537918871252206e-05, + "loss": 0.0, + "step": 2270 + }, + { + "epoch": 3.6, + "learning_rate": 1.5520282186948854e-05, + "loss": 0.0, + "step": 2271 + }, + { + "epoch": 3.61, + "learning_rate": 1.5502645502645505e-05, + "loss": 0.0, + "step": 2272 + }, + { + "epoch": 3.61, + "learning_rate": 1.5485008818342153e-05, + "loss": 0.0, + "step": 2273 + }, + { + "epoch": 3.61, + "learning_rate": 1.54673721340388e-05, + "loss": 0.0, + "step": 2274 + }, + { + "epoch": 3.61, + "learning_rate": 1.544973544973545e-05, + "loss": 0.0, + "step": 2275 + }, + { + "epoch": 3.61, + "learning_rate": 1.54320987654321e-05, + "loss": 0.0, + "step": 2276 + }, + { + "epoch": 3.61, + "learning_rate": 1.541446208112875e-05, + "loss": 0.0, + "step": 2277 + }, + { + "epoch": 3.62, + "learning_rate": 1.5396825396825398e-05, + "loss": 0.0, + "step": 2278 + }, + { + "epoch": 3.62, + "learning_rate": 1.5379188712522046e-05, + "loss": 0.0, + "step": 2279 + }, + { + "epoch": 3.62, + "learning_rate": 1.5361552028218697e-05, + "loss": 0.0, + "step": 2280 + }, + { + "epoch": 3.62, + "learning_rate": 1.5343915343915344e-05, + "loss": 0.0, + "step": 2281 + }, + { + "epoch": 3.62, + "learning_rate": 1.5326278659611996e-05, + "loss": 0.0, + "step": 2282 + }, + { + "epoch": 3.62, + "learning_rate": 1.5308641975308643e-05, + "loss": 0.0, + "step": 2283 + }, + { + "epoch": 3.63, + "learning_rate": 1.529100529100529e-05, + "loss": 0.0, + "step": 2284 + }, + { + "epoch": 3.63, + "learning_rate": 1.5273368606701942e-05, + "loss": 0.0, + "step": 2285 + }, + { + "epoch": 3.63, + "learning_rate": 1.5255731922398588e-05, + "loss": 0.0, + "step": 2286 + }, + { + "epoch": 3.63, + "learning_rate": 1.5238095238095241e-05, + "loss": 0.0, + "step": 2287 + }, + { + "epoch": 3.63, + "learning_rate": 1.5220458553791889e-05, + "loss": 0.0, + "step": 2288 + }, + { + "epoch": 3.63, + "learning_rate": 1.5202821869488536e-05, + "loss": 0.0, + "step": 2289 + }, + { + "epoch": 3.63, + "learning_rate": 1.5185185185185186e-05, + "loss": 0.0, + "step": 2290 + }, + { + "epoch": 3.64, + "learning_rate": 1.5167548500881834e-05, + "loss": 0.0, + "step": 2291 + }, + { + "epoch": 3.64, + "learning_rate": 1.5149911816578485e-05, + "loss": 0.0, + "step": 2292 + }, + { + "epoch": 3.64, + "learning_rate": 1.5132275132275134e-05, + "loss": 0.0, + "step": 2293 + }, + { + "epoch": 3.64, + "learning_rate": 1.5114638447971782e-05, + "loss": 0.0, + "step": 2294 + }, + { + "epoch": 3.64, + "learning_rate": 1.5097001763668431e-05, + "loss": 0.0, + "step": 2295 + }, + { + "epoch": 3.64, + "learning_rate": 1.5079365079365079e-05, + "loss": 0.0, + "step": 2296 + }, + { + "epoch": 3.65, + "learning_rate": 1.506172839506173e-05, + "loss": 0.0, + "step": 2297 + }, + { + "epoch": 3.65, + "learning_rate": 1.504409171075838e-05, + "loss": 0.0, + "step": 2298 + }, + { + "epoch": 3.65, + "learning_rate": 1.5026455026455027e-05, + "loss": 0.0, + "step": 2299 + }, + { + "epoch": 3.65, + "learning_rate": 1.5008818342151677e-05, + "loss": 0.0, + "step": 2300 + }, + { + "epoch": 3.65, + "learning_rate": 1.4991181657848324e-05, + "loss": 0.0, + "step": 2301 + }, + { + "epoch": 3.65, + "learning_rate": 1.4973544973544974e-05, + "loss": 0.0, + "step": 2302 + }, + { + "epoch": 3.66, + "learning_rate": 1.4955908289241625e-05, + "loss": 0.0, + "step": 2303 + }, + { + "epoch": 3.66, + "learning_rate": 1.4938271604938272e-05, + "loss": 0.0, + "step": 2304 + }, + { + "epoch": 3.66, + "learning_rate": 1.4920634920634922e-05, + "loss": 0.0, + "step": 2305 + }, + { + "epoch": 3.66, + "learning_rate": 1.490299823633157e-05, + "loss": 0.0, + "step": 2306 + }, + { + "epoch": 3.66, + "learning_rate": 1.4885361552028217e-05, + "loss": 0.0056, + "step": 2307 + }, + { + "epoch": 3.66, + "learning_rate": 1.486772486772487e-05, + "loss": 0.0, + "step": 2308 + }, + { + "epoch": 3.67, + "learning_rate": 1.4850088183421518e-05, + "loss": 0.0, + "step": 2309 + }, + { + "epoch": 3.67, + "learning_rate": 1.4832451499118166e-05, + "loss": 0.0, + "step": 2310 + }, + { + "epoch": 3.67, + "learning_rate": 1.4814814814814815e-05, + "loss": 0.0, + "step": 2311 + }, + { + "epoch": 3.67, + "learning_rate": 1.4797178130511463e-05, + "loss": 0.0, + "step": 2312 + }, + { + "epoch": 3.67, + "learning_rate": 1.4779541446208114e-05, + "loss": 0.0, + "step": 2313 + }, + { + "epoch": 3.67, + "learning_rate": 1.4761904761904763e-05, + "loss": 0.0, + "step": 2314 + }, + { + "epoch": 3.67, + "learning_rate": 1.4744268077601411e-05, + "loss": 0.0, + "step": 2315 + }, + { + "epoch": 3.68, + "learning_rate": 1.472663139329806e-05, + "loss": 0.0, + "step": 2316 + }, + { + "epoch": 3.68, + "learning_rate": 1.4708994708994708e-05, + "loss": 0.0, + "step": 2317 + }, + { + "epoch": 3.68, + "learning_rate": 1.4691358024691359e-05, + "loss": 0.0, + "step": 2318 + }, + { + "epoch": 3.68, + "learning_rate": 1.4673721340388009e-05, + "loss": 0.0, + "step": 2319 + }, + { + "epoch": 3.68, + "learning_rate": 1.4656084656084656e-05, + "loss": 0.0, + "step": 2320 + }, + { + "epoch": 3.68, + "learning_rate": 1.4638447971781306e-05, + "loss": 0.0, + "step": 2321 + }, + { + "epoch": 3.69, + "learning_rate": 1.4620811287477953e-05, + "loss": 0.0, + "step": 2322 + }, + { + "epoch": 3.69, + "learning_rate": 1.4603174603174605e-05, + "loss": 0.0, + "step": 2323 + }, + { + "epoch": 3.69, + "learning_rate": 1.4585537918871254e-05, + "loss": 0.0, + "step": 2324 + }, + { + "epoch": 3.69, + "learning_rate": 1.4567901234567902e-05, + "loss": 0.0, + "step": 2325 + }, + { + "epoch": 3.69, + "learning_rate": 1.455026455026455e-05, + "loss": 0.0, + "step": 2326 + }, + { + "epoch": 3.69, + "learning_rate": 1.4532627865961199e-05, + "loss": 0.0, + "step": 2327 + }, + { + "epoch": 3.7, + "learning_rate": 1.451499118165785e-05, + "loss": 0.0, + "step": 2328 + }, + { + "epoch": 3.7, + "learning_rate": 1.4497354497354498e-05, + "loss": 0.0, + "step": 2329 + }, + { + "epoch": 3.7, + "learning_rate": 1.4479717813051147e-05, + "loss": 0.0, + "step": 2330 + }, + { + "epoch": 3.7, + "learning_rate": 1.4462081128747795e-05, + "loss": 0.0, + "step": 2331 + }, + { + "epoch": 3.7, + "learning_rate": 1.4444444444444444e-05, + "loss": 0.0, + "step": 2332 + }, + { + "epoch": 3.7, + "learning_rate": 1.4426807760141095e-05, + "loss": 0.0, + "step": 2333 + }, + { + "epoch": 3.7, + "learning_rate": 1.4409171075837743e-05, + "loss": 0.0, + "step": 2334 + }, + { + "epoch": 3.71, + "learning_rate": 1.4391534391534392e-05, + "loss": 0.0, + "step": 2335 + }, + { + "epoch": 3.71, + "learning_rate": 1.437389770723104e-05, + "loss": 0.0, + "step": 2336 + }, + { + "epoch": 3.71, + "learning_rate": 1.435626102292769e-05, + "loss": 0.0, + "step": 2337 + }, + { + "epoch": 3.71, + "learning_rate": 1.433862433862434e-05, + "loss": 0.0, + "step": 2338 + }, + { + "epoch": 3.71, + "learning_rate": 1.4320987654320988e-05, + "loss": 0.0, + "step": 2339 + }, + { + "epoch": 3.71, + "learning_rate": 1.4303350970017638e-05, + "loss": 0.0, + "step": 2340 + }, + { + "epoch": 3.72, + "learning_rate": 1.4285714285714285e-05, + "loss": 0.0, + "step": 2341 + }, + { + "epoch": 3.72, + "learning_rate": 1.4268077601410935e-05, + "loss": 0.0, + "step": 2342 + }, + { + "epoch": 3.72, + "learning_rate": 1.4250440917107586e-05, + "loss": 0.0, + "step": 2343 + }, + { + "epoch": 3.72, + "learning_rate": 1.4232804232804234e-05, + "loss": 0.0, + "step": 2344 + }, + { + "epoch": 3.72, + "learning_rate": 1.4215167548500883e-05, + "loss": 0.0, + "step": 2345 + }, + { + "epoch": 3.72, + "learning_rate": 1.419753086419753e-05, + "loss": 0.0, + "step": 2346 + }, + { + "epoch": 3.73, + "learning_rate": 1.4179894179894179e-05, + "loss": 0.0, + "step": 2347 + }, + { + "epoch": 3.73, + "learning_rate": 1.4162257495590831e-05, + "loss": 0.0, + "step": 2348 + }, + { + "epoch": 3.73, + "learning_rate": 1.4144620811287479e-05, + "loss": 0.0, + "step": 2349 + }, + { + "epoch": 3.73, + "learning_rate": 1.4126984126984127e-05, + "loss": 0.0, + "step": 2350 + }, + { + "epoch": 3.73, + "learning_rate": 1.4109347442680776e-05, + "loss": 0.0, + "step": 2351 + }, + { + "epoch": 3.73, + "learning_rate": 1.4091710758377424e-05, + "loss": 0.0, + "step": 2352 + }, + { + "epoch": 3.73, + "learning_rate": 1.4074074074074075e-05, + "loss": 0.0, + "step": 2353 + }, + { + "epoch": 3.74, + "learning_rate": 1.4056437389770724e-05, + "loss": 0.0, + "step": 2354 + }, + { + "epoch": 3.74, + "learning_rate": 1.4038800705467372e-05, + "loss": 0.0, + "step": 2355 + }, + { + "epoch": 3.74, + "learning_rate": 1.4021164021164022e-05, + "loss": 0.0, + "step": 2356 + }, + { + "epoch": 3.74, + "learning_rate": 1.400352733686067e-05, + "loss": 0.0, + "step": 2357 + }, + { + "epoch": 3.74, + "learning_rate": 1.398589065255732e-05, + "loss": 0.0058, + "step": 2358 + }, + { + "epoch": 3.74, + "learning_rate": 1.396825396825397e-05, + "loss": 0.0, + "step": 2359 + }, + { + "epoch": 3.75, + "learning_rate": 1.3950617283950617e-05, + "loss": 0.0, + "step": 2360 + }, + { + "epoch": 3.75, + "learning_rate": 1.3932980599647267e-05, + "loss": 0.0, + "step": 2361 + }, + { + "epoch": 3.75, + "learning_rate": 1.3915343915343915e-05, + "loss": 0.0, + "step": 2362 + }, + { + "epoch": 3.75, + "learning_rate": 1.3897707231040566e-05, + "loss": 0.0, + "step": 2363 + }, + { + "epoch": 3.75, + "learning_rate": 1.3880070546737215e-05, + "loss": 0.0, + "step": 2364 + }, + { + "epoch": 3.75, + "learning_rate": 1.3862433862433863e-05, + "loss": 0.0, + "step": 2365 + }, + { + "epoch": 3.76, + "learning_rate": 1.384479717813051e-05, + "loss": 0.0, + "step": 2366 + }, + { + "epoch": 3.76, + "learning_rate": 1.382716049382716e-05, + "loss": 0.0, + "step": 2367 + }, + { + "epoch": 3.76, + "learning_rate": 1.3809523809523811e-05, + "loss": 0.0, + "step": 2368 + }, + { + "epoch": 3.76, + "learning_rate": 1.3791887125220459e-05, + "loss": 0.0, + "step": 2369 + }, + { + "epoch": 3.76, + "learning_rate": 1.3774250440917108e-05, + "loss": 0.0, + "step": 2370 + }, + { + "epoch": 3.76, + "learning_rate": 1.3756613756613756e-05, + "loss": 0.0, + "step": 2371 + }, + { + "epoch": 3.77, + "learning_rate": 1.3738977072310405e-05, + "loss": 0.0, + "step": 2372 + }, + { + "epoch": 3.77, + "learning_rate": 1.3721340388007056e-05, + "loss": 0.0, + "step": 2373 + }, + { + "epoch": 3.77, + "learning_rate": 1.3703703703703704e-05, + "loss": 0.0, + "step": 2374 + }, + { + "epoch": 3.77, + "learning_rate": 1.3686067019400354e-05, + "loss": 0.0, + "step": 2375 + }, + { + "epoch": 3.77, + "learning_rate": 1.3668430335097001e-05, + "loss": 0.0, + "step": 2376 + }, + { + "epoch": 3.77, + "learning_rate": 1.365079365079365e-05, + "loss": 0.0, + "step": 2377 + }, + { + "epoch": 3.77, + "learning_rate": 1.3633156966490302e-05, + "loss": 0.0, + "step": 2378 + }, + { + "epoch": 3.78, + "learning_rate": 1.361552028218695e-05, + "loss": 0.0, + "step": 2379 + }, + { + "epoch": 3.78, + "learning_rate": 1.3597883597883599e-05, + "loss": 0.0, + "step": 2380 + }, + { + "epoch": 3.78, + "learning_rate": 1.3580246913580247e-05, + "loss": 0.0, + "step": 2381 + }, + { + "epoch": 3.78, + "learning_rate": 1.3562610229276896e-05, + "loss": 0.0, + "step": 2382 + }, + { + "epoch": 3.78, + "learning_rate": 1.3544973544973547e-05, + "loss": 0.0, + "step": 2383 + }, + { + "epoch": 3.78, + "learning_rate": 1.3527336860670195e-05, + "loss": 0.0, + "step": 2384 + }, + { + "epoch": 3.79, + "learning_rate": 1.3509700176366844e-05, + "loss": 0.0, + "step": 2385 + }, + { + "epoch": 3.79, + "learning_rate": 1.3492063492063492e-05, + "loss": 0.0, + "step": 2386 + }, + { + "epoch": 3.79, + "learning_rate": 1.347442680776014e-05, + "loss": 0.0, + "step": 2387 + }, + { + "epoch": 3.79, + "learning_rate": 1.3456790123456793e-05, + "loss": 0.0, + "step": 2388 + }, + { + "epoch": 3.79, + "learning_rate": 1.343915343915344e-05, + "loss": 0.0, + "step": 2389 + }, + { + "epoch": 3.79, + "learning_rate": 1.3421516754850088e-05, + "loss": 0.0, + "step": 2390 + }, + { + "epoch": 3.8, + "learning_rate": 1.3403880070546737e-05, + "loss": 0.0, + "step": 2391 + }, + { + "epoch": 3.8, + "learning_rate": 1.3386243386243385e-05, + "loss": 0.0, + "step": 2392 + }, + { + "epoch": 3.8, + "learning_rate": 1.3368606701940036e-05, + "loss": 0.0, + "step": 2393 + }, + { + "epoch": 3.8, + "learning_rate": 1.3350970017636686e-05, + "loss": 0.0, + "step": 2394 + }, + { + "epoch": 3.8, + "learning_rate": 1.3333333333333333e-05, + "loss": 0.0, + "step": 2395 + }, + { + "epoch": 3.8, + "learning_rate": 1.3315696649029983e-05, + "loss": 0.0, + "step": 2396 + }, + { + "epoch": 3.8, + "learning_rate": 1.329805996472663e-05, + "loss": 0.0, + "step": 2397 + }, + { + "epoch": 3.81, + "learning_rate": 1.3280423280423282e-05, + "loss": 0.0, + "step": 2398 + }, + { + "epoch": 3.81, + "learning_rate": 1.3262786596119931e-05, + "loss": 0.0, + "step": 2399 + }, + { + "epoch": 3.81, + "learning_rate": 1.3245149911816579e-05, + "loss": 0.0, + "step": 2400 + }, + { + "epoch": 3.81, + "learning_rate": 1.3227513227513228e-05, + "loss": 0.0, + "step": 2401 + }, + { + "epoch": 3.81, + "learning_rate": 1.3209876543209876e-05, + "loss": 0.0, + "step": 2402 + }, + { + "epoch": 3.81, + "learning_rate": 1.3192239858906527e-05, + "loss": 0.0, + "step": 2403 + }, + { + "epoch": 3.82, + "learning_rate": 1.3174603174603176e-05, + "loss": 0.0, + "step": 2404 + }, + { + "epoch": 3.82, + "learning_rate": 1.3156966490299824e-05, + "loss": 0.0, + "step": 2405 + }, + { + "epoch": 3.82, + "learning_rate": 1.3139329805996472e-05, + "loss": 0.0, + "step": 2406 + }, + { + "epoch": 3.82, + "learning_rate": 1.3121693121693121e-05, + "loss": 0.0, + "step": 2407 + }, + { + "epoch": 3.82, + "learning_rate": 1.3104056437389772e-05, + "loss": 0.0, + "step": 2408 + }, + { + "epoch": 3.82, + "learning_rate": 1.308641975308642e-05, + "loss": 0.0, + "step": 2409 + }, + { + "epoch": 3.83, + "learning_rate": 1.306878306878307e-05, + "loss": 0.0, + "step": 2410 + }, + { + "epoch": 3.83, + "learning_rate": 1.3051146384479717e-05, + "loss": 0.0, + "step": 2411 + }, + { + "epoch": 3.83, + "learning_rate": 1.3033509700176367e-05, + "loss": 0.0, + "step": 2412 + }, + { + "epoch": 3.83, + "learning_rate": 1.3015873015873018e-05, + "loss": 0.0, + "step": 2413 + }, + { + "epoch": 3.83, + "learning_rate": 1.2998236331569665e-05, + "loss": 0.0, + "step": 2414 + }, + { + "epoch": 3.83, + "learning_rate": 1.2980599647266315e-05, + "loss": 0.0, + "step": 2415 + }, + { + "epoch": 3.83, + "learning_rate": 1.2962962962962962e-05, + "loss": 0.0, + "step": 2416 + }, + { + "epoch": 3.84, + "learning_rate": 1.2945326278659612e-05, + "loss": 0.0, + "step": 2417 + }, + { + "epoch": 3.84, + "learning_rate": 1.2927689594356263e-05, + "loss": 0.0, + "step": 2418 + }, + { + "epoch": 3.84, + "learning_rate": 1.291005291005291e-05, + "loss": 0.0, + "step": 2419 + }, + { + "epoch": 3.84, + "learning_rate": 1.289241622574956e-05, + "loss": 0.0, + "step": 2420 + }, + { + "epoch": 3.84, + "learning_rate": 1.2874779541446208e-05, + "loss": 0.0, + "step": 2421 + }, + { + "epoch": 3.84, + "learning_rate": 1.2857142857142857e-05, + "loss": 0.0, + "step": 2422 + }, + { + "epoch": 3.85, + "learning_rate": 1.2839506172839508e-05, + "loss": 0.0, + "step": 2423 + }, + { + "epoch": 3.85, + "learning_rate": 1.2821869488536156e-05, + "loss": 0.0, + "step": 2424 + }, + { + "epoch": 3.85, + "learning_rate": 1.2804232804232805e-05, + "loss": 0.0, + "step": 2425 + }, + { + "epoch": 3.85, + "learning_rate": 1.2786596119929453e-05, + "loss": 0.0, + "step": 2426 + }, + { + "epoch": 3.85, + "learning_rate": 1.2768959435626101e-05, + "loss": 0.0, + "step": 2427 + }, + { + "epoch": 3.85, + "learning_rate": 1.2751322751322754e-05, + "loss": 0.0, + "step": 2428 + }, + { + "epoch": 3.86, + "learning_rate": 1.2733686067019401e-05, + "loss": 0.0, + "step": 2429 + }, + { + "epoch": 3.86, + "learning_rate": 1.2716049382716049e-05, + "loss": 0.0, + "step": 2430 + }, + { + "epoch": 3.86, + "learning_rate": 1.2698412698412699e-05, + "loss": 0.0, + "step": 2431 + }, + { + "epoch": 3.86, + "learning_rate": 1.2680776014109346e-05, + "loss": 0.0, + "step": 2432 + }, + { + "epoch": 3.86, + "learning_rate": 1.2663139329805997e-05, + "loss": 0.0, + "step": 2433 + }, + { + "epoch": 3.86, + "learning_rate": 1.2645502645502647e-05, + "loss": 0.0, + "step": 2434 + }, + { + "epoch": 3.87, + "learning_rate": 1.2627865961199294e-05, + "loss": 0.0, + "step": 2435 + }, + { + "epoch": 3.87, + "learning_rate": 1.2610229276895944e-05, + "loss": 0.0, + "step": 2436 + }, + { + "epoch": 3.87, + "learning_rate": 1.2592592592592592e-05, + "loss": 0.0, + "step": 2437 + }, + { + "epoch": 3.87, + "learning_rate": 1.2574955908289243e-05, + "loss": 0.0, + "step": 2438 + }, + { + "epoch": 3.87, + "learning_rate": 1.2557319223985892e-05, + "loss": 0.0, + "step": 2439 + }, + { + "epoch": 3.87, + "learning_rate": 1.253968253968254e-05, + "loss": 0.0, + "step": 2440 + }, + { + "epoch": 3.87, + "learning_rate": 1.252204585537919e-05, + "loss": 0.0001, + "step": 2441 + }, + { + "epoch": 3.88, + "learning_rate": 1.2504409171075837e-05, + "loss": 0.0, + "step": 2442 + }, + { + "epoch": 3.88, + "learning_rate": 1.2486772486772486e-05, + "loss": 0.0, + "step": 2443 + }, + { + "epoch": 3.88, + "learning_rate": 1.2469135802469137e-05, + "loss": 0.0, + "step": 2444 + }, + { + "epoch": 3.88, + "learning_rate": 1.2451499118165785e-05, + "loss": 0.0, + "step": 2445 + }, + { + "epoch": 3.88, + "learning_rate": 1.2433862433862433e-05, + "loss": 0.0, + "step": 2446 + }, + { + "epoch": 3.88, + "learning_rate": 1.2416225749559084e-05, + "loss": 0.0, + "step": 2447 + }, + { + "epoch": 3.89, + "learning_rate": 1.2398589065255732e-05, + "loss": 0.0, + "step": 2448 + }, + { + "epoch": 3.89, + "learning_rate": 1.2380952380952381e-05, + "loss": 0.0, + "step": 2449 + }, + { + "epoch": 3.89, + "learning_rate": 1.236331569664903e-05, + "loss": 0.0081, + "step": 2450 + }, + { + "epoch": 3.89, + "learning_rate": 1.2345679012345678e-05, + "loss": 0.0, + "step": 2451 + }, + { + "epoch": 3.89, + "learning_rate": 1.232804232804233e-05, + "loss": 0.0, + "step": 2452 + }, + { + "epoch": 3.89, + "learning_rate": 1.2310405643738977e-05, + "loss": 0.0, + "step": 2453 + }, + { + "epoch": 3.9, + "learning_rate": 1.2292768959435627e-05, + "loss": 0.0, + "step": 2454 + }, + { + "epoch": 3.9, + "learning_rate": 1.2275132275132276e-05, + "loss": 0.0, + "step": 2455 + }, + { + "epoch": 3.9, + "learning_rate": 1.2257495590828924e-05, + "loss": 0.0001, + "step": 2456 + }, + { + "epoch": 3.9, + "learning_rate": 1.2239858906525575e-05, + "loss": 0.0, + "step": 2457 + }, + { + "epoch": 3.9, + "learning_rate": 1.2222222222222222e-05, + "loss": 0.0, + "step": 2458 + }, + { + "epoch": 3.9, + "learning_rate": 1.2204585537918872e-05, + "loss": 0.0, + "step": 2459 + }, + { + "epoch": 3.9, + "learning_rate": 1.2186948853615521e-05, + "loss": 0.0, + "step": 2460 + }, + { + "epoch": 3.91, + "learning_rate": 1.2169312169312169e-05, + "loss": 0.0, + "step": 2461 + }, + { + "epoch": 3.91, + "learning_rate": 1.2151675485008818e-05, + "loss": 0.0, + "step": 2462 + }, + { + "epoch": 3.91, + "learning_rate": 1.2134038800705468e-05, + "loss": 0.0, + "step": 2463 + }, + { + "epoch": 3.91, + "learning_rate": 1.2116402116402117e-05, + "loss": 0.0, + "step": 2464 + }, + { + "epoch": 3.91, + "learning_rate": 1.2098765432098767e-05, + "loss": 0.0, + "step": 2465 + }, + { + "epoch": 3.91, + "learning_rate": 1.2081128747795414e-05, + "loss": 0.0, + "step": 2466 + }, + { + "epoch": 3.92, + "learning_rate": 1.2063492063492064e-05, + "loss": 0.0, + "step": 2467 + }, + { + "epoch": 3.92, + "learning_rate": 1.2045855379188713e-05, + "loss": 0.0, + "step": 2468 + }, + { + "epoch": 3.92, + "learning_rate": 1.2028218694885363e-05, + "loss": 0.0, + "step": 2469 + }, + { + "epoch": 3.92, + "learning_rate": 1.201058201058201e-05, + "loss": 0.0, + "step": 2470 + }, + { + "epoch": 3.92, + "learning_rate": 1.199294532627866e-05, + "loss": 0.0, + "step": 2471 + }, + { + "epoch": 3.92, + "learning_rate": 1.1975308641975309e-05, + "loss": 0.0, + "step": 2472 + }, + { + "epoch": 3.93, + "learning_rate": 1.1957671957671959e-05, + "loss": 0.0, + "step": 2473 + }, + { + "epoch": 3.93, + "learning_rate": 1.1940035273368608e-05, + "loss": 0.0, + "step": 2474 + }, + { + "epoch": 3.93, + "learning_rate": 1.1922398589065256e-05, + "loss": 0.0, + "step": 2475 + }, + { + "epoch": 3.93, + "learning_rate": 1.1904761904761905e-05, + "loss": 0.0, + "step": 2476 + }, + { + "epoch": 3.93, + "learning_rate": 1.1887125220458555e-05, + "loss": 0.0, + "step": 2477 + }, + { + "epoch": 3.93, + "learning_rate": 1.1869488536155202e-05, + "loss": 0.0, + "step": 2478 + }, + { + "epoch": 3.93, + "learning_rate": 1.1851851851851853e-05, + "loss": 0.0, + "step": 2479 + }, + { + "epoch": 3.94, + "learning_rate": 1.1834215167548501e-05, + "loss": 0.0, + "step": 2480 + }, + { + "epoch": 3.94, + "learning_rate": 1.181657848324515e-05, + "loss": 0.0, + "step": 2481 + }, + { + "epoch": 3.94, + "learning_rate": 1.17989417989418e-05, + "loss": 0.0, + "step": 2482 + }, + { + "epoch": 3.94, + "learning_rate": 1.1781305114638448e-05, + "loss": 0.0, + "step": 2483 + }, + { + "epoch": 3.94, + "learning_rate": 1.1763668430335099e-05, + "loss": 0.0, + "step": 2484 + }, + { + "epoch": 3.94, + "learning_rate": 1.1746031746031746e-05, + "loss": 0.0, + "step": 2485 + }, + { + "epoch": 3.95, + "learning_rate": 1.1728395061728396e-05, + "loss": 0.0, + "step": 2486 + }, + { + "epoch": 3.95, + "learning_rate": 1.1710758377425045e-05, + "loss": 0.0, + "step": 2487 + }, + { + "epoch": 3.95, + "learning_rate": 1.1693121693121693e-05, + "loss": 0.0, + "step": 2488 + }, + { + "epoch": 3.95, + "learning_rate": 1.1675485008818342e-05, + "loss": 0.0, + "step": 2489 + }, + { + "epoch": 3.95, + "learning_rate": 1.1657848324514992e-05, + "loss": 0.0, + "step": 2490 + }, + { + "epoch": 3.95, + "learning_rate": 1.164021164021164e-05, + "loss": 0.0, + "step": 2491 + }, + { + "epoch": 3.96, + "learning_rate": 1.162257495590829e-05, + "loss": 0.0, + "step": 2492 + }, + { + "epoch": 3.96, + "learning_rate": 1.1604938271604938e-05, + "loss": 0.0, + "step": 2493 + }, + { + "epoch": 3.96, + "learning_rate": 1.1587301587301588e-05, + "loss": 0.0, + "step": 2494 + }, + { + "epoch": 3.96, + "learning_rate": 1.1569664902998237e-05, + "loss": 0.0, + "step": 2495 + }, + { + "epoch": 3.96, + "learning_rate": 1.1552028218694885e-05, + "loss": 0.0, + "step": 2496 + }, + { + "epoch": 3.96, + "learning_rate": 1.1534391534391536e-05, + "loss": 0.0, + "step": 2497 + }, + { + "epoch": 3.97, + "learning_rate": 1.1516754850088184e-05, + "loss": 0.0, + "step": 2498 + }, + { + "epoch": 3.97, + "learning_rate": 1.1499118165784833e-05, + "loss": 0.0, + "step": 2499 + }, + { + "epoch": 3.97, + "learning_rate": 1.1481481481481482e-05, + "loss": 0.0, + "step": 2500 + }, + { + "epoch": 3.97, + "learning_rate": 1.146384479717813e-05, + "loss": 0.0, + "step": 2501 + }, + { + "epoch": 3.97, + "learning_rate": 1.144620811287478e-05, + "loss": 0.0, + "step": 2502 + }, + { + "epoch": 3.97, + "learning_rate": 1.1428571428571429e-05, + "loss": 0.0, + "step": 2503 + }, + { + "epoch": 3.97, + "learning_rate": 1.1410934744268078e-05, + "loss": 0.0, + "step": 2504 + }, + { + "epoch": 3.98, + "learning_rate": 1.1393298059964728e-05, + "loss": 0.0, + "step": 2505 + }, + { + "epoch": 3.98, + "learning_rate": 1.1375661375661376e-05, + "loss": 0.0, + "step": 2506 + }, + { + "epoch": 3.98, + "learning_rate": 1.1358024691358025e-05, + "loss": 0.0, + "step": 2507 + }, + { + "epoch": 3.98, + "learning_rate": 1.1340388007054674e-05, + "loss": 0.0, + "step": 2508 + }, + { + "epoch": 3.98, + "learning_rate": 1.1322751322751324e-05, + "loss": 0.0, + "step": 2509 + }, + { + "epoch": 3.98, + "learning_rate": 1.1305114638447972e-05, + "loss": 0.0, + "step": 2510 + }, + { + "epoch": 3.99, + "learning_rate": 1.1287477954144621e-05, + "loss": 0.0, + "step": 2511 + }, + { + "epoch": 3.99, + "learning_rate": 1.126984126984127e-05, + "loss": 0.0, + "step": 2512 + }, + { + "epoch": 3.99, + "learning_rate": 1.125220458553792e-05, + "loss": 0.0, + "step": 2513 + }, + { + "epoch": 3.99, + "learning_rate": 1.123456790123457e-05, + "loss": 0.0, + "step": 2514 + }, + { + "epoch": 3.99, + "learning_rate": 1.1216931216931217e-05, + "loss": 0.0, + "step": 2515 + }, + { + "epoch": 3.99, + "learning_rate": 1.1199294532627866e-05, + "loss": 0.0, + "step": 2516 + }, + { + "epoch": 4.0, + "learning_rate": 1.1181657848324516e-05, + "loss": 0.0, + "step": 2517 + }, + { + "epoch": 4.0, + "learning_rate": 1.1164021164021163e-05, + "loss": 0.0, + "step": 2518 + }, + { + "epoch": 4.0, + "learning_rate": 1.1146384479717815e-05, + "loss": 0.0, + "step": 2519 + }, + { + "epoch": 4.0, + "learning_rate": 1.1128747795414462e-05, + "loss": 0.0, + "step": 2520 + }, + { + "epoch": 4.0, + "eval_accuracy": 1.0, + "eval_auc": 1.0, + "eval_f1": 1.0, + "eval_loss": 0.003079816000536084, + "eval_precision": 1.0, + "eval_recall": 1.0, + "eval_runtime": 0.2016, + "eval_samples_per_second": 188.467, + "eval_steps_per_second": 4.96, + "step": 2520 + }, + { + "epoch": 4.0, + "learning_rate": 1.1111111111111112e-05, + "loss": 0.0, + "step": 2521 + }, + { + "epoch": 4.0, + "learning_rate": 1.1093474426807761e-05, + "loss": 0.0, + "step": 2522 + }, + { + "epoch": 4.0, + "learning_rate": 1.1075837742504409e-05, + "loss": 0.0, + "step": 2523 + }, + { + "epoch": 4.01, + "learning_rate": 1.105820105820106e-05, + "loss": 0.0, + "step": 2524 + }, + { + "epoch": 4.01, + "learning_rate": 1.1040564373897708e-05, + "loss": 0.0, + "step": 2525 + }, + { + "epoch": 4.01, + "learning_rate": 1.1022927689594357e-05, + "loss": 0.0, + "step": 2526 + }, + { + "epoch": 4.01, + "learning_rate": 1.1005291005291006e-05, + "loss": 0.0, + "step": 2527 + }, + { + "epoch": 4.01, + "learning_rate": 1.0987654320987654e-05, + "loss": 0.0, + "step": 2528 + }, + { + "epoch": 4.01, + "learning_rate": 1.0970017636684304e-05, + "loss": 0.0, + "step": 2529 + }, + { + "epoch": 4.02, + "learning_rate": 1.0952380952380953e-05, + "loss": 0.0, + "step": 2530 + }, + { + "epoch": 4.02, + "learning_rate": 1.09347442680776e-05, + "loss": 0.0, + "step": 2531 + }, + { + "epoch": 4.02, + "learning_rate": 1.0917107583774252e-05, + "loss": 0.0, + "step": 2532 + }, + { + "epoch": 4.02, + "learning_rate": 1.08994708994709e-05, + "loss": 0.0, + "step": 2533 + }, + { + "epoch": 4.02, + "learning_rate": 1.0881834215167549e-05, + "loss": 0.0, + "step": 2534 + }, + { + "epoch": 4.02, + "learning_rate": 1.0864197530864198e-05, + "loss": 0.0, + "step": 2535 + }, + { + "epoch": 4.03, + "learning_rate": 1.0846560846560846e-05, + "loss": 0.0, + "step": 2536 + }, + { + "epoch": 4.03, + "learning_rate": 1.0828924162257497e-05, + "loss": 0.0, + "step": 2537 + }, + { + "epoch": 4.03, + "learning_rate": 1.0811287477954145e-05, + "loss": 0.0, + "step": 2538 + }, + { + "epoch": 4.03, + "learning_rate": 1.0793650793650794e-05, + "loss": 0.0, + "step": 2539 + }, + { + "epoch": 4.03, + "learning_rate": 1.0776014109347444e-05, + "loss": 0.0, + "step": 2540 + }, + { + "epoch": 4.03, + "learning_rate": 1.0758377425044091e-05, + "loss": 0.0, + "step": 2541 + }, + { + "epoch": 4.03, + "learning_rate": 1.074074074074074e-05, + "loss": 0.0, + "step": 2542 + }, + { + "epoch": 4.04, + "learning_rate": 1.072310405643739e-05, + "loss": 0.0, + "step": 2543 + }, + { + "epoch": 4.04, + "learning_rate": 1.070546737213404e-05, + "loss": 0.0, + "step": 2544 + }, + { + "epoch": 4.04, + "learning_rate": 1.0687830687830689e-05, + "loss": 0.0, + "step": 2545 + }, + { + "epoch": 4.04, + "learning_rate": 1.0670194003527337e-05, + "loss": 0.0, + "step": 2546 + }, + { + "epoch": 4.04, + "learning_rate": 1.0652557319223986e-05, + "loss": 0.0, + "step": 2547 + }, + { + "epoch": 4.04, + "learning_rate": 1.0634920634920636e-05, + "loss": 0.0, + "step": 2548 + }, + { + "epoch": 4.05, + "learning_rate": 1.0617283950617285e-05, + "loss": 0.0, + "step": 2549 + }, + { + "epoch": 4.05, + "learning_rate": 1.0599647266313933e-05, + "loss": 0.0, + "step": 2550 + }, + { + "epoch": 4.05, + "learning_rate": 1.0582010582010582e-05, + "loss": 0.0, + "step": 2551 + }, + { + "epoch": 4.05, + "learning_rate": 1.0564373897707232e-05, + "loss": 0.0, + "step": 2552 + }, + { + "epoch": 4.05, + "learning_rate": 1.0546737213403881e-05, + "loss": 0.0, + "step": 2553 + }, + { + "epoch": 4.05, + "learning_rate": 1.052910052910053e-05, + "loss": 0.0, + "step": 2554 + }, + { + "epoch": 4.06, + "learning_rate": 1.0511463844797178e-05, + "loss": 0.0, + "step": 2555 + }, + { + "epoch": 4.06, + "learning_rate": 1.0493827160493827e-05, + "loss": 0.0, + "step": 2556 + }, + { + "epoch": 4.06, + "learning_rate": 1.0476190476190477e-05, + "loss": 0.0, + "step": 2557 + }, + { + "epoch": 4.06, + "learning_rate": 1.0458553791887125e-05, + "loss": 0.0, + "step": 2558 + }, + { + "epoch": 4.06, + "learning_rate": 1.0440917107583776e-05, + "loss": 0.0, + "step": 2559 + }, + { + "epoch": 4.06, + "learning_rate": 1.0423280423280423e-05, + "loss": 0.0, + "step": 2560 + }, + { + "epoch": 4.07, + "learning_rate": 1.0405643738977073e-05, + "loss": 0.0, + "step": 2561 + }, + { + "epoch": 4.07, + "learning_rate": 1.0388007054673722e-05, + "loss": 0.0, + "step": 2562 + }, + { + "epoch": 4.07, + "learning_rate": 1.037037037037037e-05, + "loss": 0.0, + "step": 2563 + }, + { + "epoch": 4.07, + "learning_rate": 1.0352733686067021e-05, + "loss": 0.0, + "step": 2564 + }, + { + "epoch": 4.07, + "learning_rate": 1.0335097001763669e-05, + "loss": 0.0, + "step": 2565 + }, + { + "epoch": 4.07, + "learning_rate": 1.0317460317460318e-05, + "loss": 0.0, + "step": 2566 + }, + { + "epoch": 4.07, + "learning_rate": 1.0299823633156968e-05, + "loss": 0.0, + "step": 2567 + }, + { + "epoch": 4.08, + "learning_rate": 1.0282186948853615e-05, + "loss": 0.0, + "step": 2568 + }, + { + "epoch": 4.08, + "learning_rate": 1.0264550264550265e-05, + "loss": 0.0, + "step": 2569 + }, + { + "epoch": 4.08, + "learning_rate": 1.0246913580246914e-05, + "loss": 0.0, + "step": 2570 + }, + { + "epoch": 4.08, + "learning_rate": 1.0229276895943562e-05, + "loss": 0.0, + "step": 2571 + }, + { + "epoch": 4.08, + "learning_rate": 1.0211640211640213e-05, + "loss": 0.0, + "step": 2572 + }, + { + "epoch": 4.08, + "learning_rate": 1.019400352733686e-05, + "loss": 0.0, + "step": 2573 + }, + { + "epoch": 4.09, + "learning_rate": 1.017636684303351e-05, + "loss": 0.0, + "step": 2574 + }, + { + "epoch": 4.09, + "learning_rate": 1.015873015873016e-05, + "loss": 0.0, + "step": 2575 + }, + { + "epoch": 4.09, + "learning_rate": 1.0141093474426807e-05, + "loss": 0.0, + "step": 2576 + }, + { + "epoch": 4.09, + "learning_rate": 1.0123456790123458e-05, + "loss": 0.0, + "step": 2577 + }, + { + "epoch": 4.09, + "learning_rate": 1.0105820105820106e-05, + "loss": 0.0, + "step": 2578 + }, + { + "epoch": 4.09, + "learning_rate": 1.0088183421516755e-05, + "loss": 0.0, + "step": 2579 + }, + { + "epoch": 4.1, + "learning_rate": 1.0070546737213405e-05, + "loss": 0.0, + "step": 2580 + }, + { + "epoch": 4.1, + "learning_rate": 1.0052910052910053e-05, + "loss": 0.0, + "step": 2581 + }, + { + "epoch": 4.1, + "learning_rate": 1.0035273368606702e-05, + "loss": 0.0, + "step": 2582 + }, + { + "epoch": 4.1, + "learning_rate": 1.0017636684303351e-05, + "loss": 0.0, + "step": 2583 + }, + { + "epoch": 4.1, + "learning_rate": 1e-05, + "loss": 0.0, + "step": 2584 + }, + { + "epoch": 4.1, + "learning_rate": 9.98236331569665e-06, + "loss": 0.0, + "step": 2585 + }, + { + "epoch": 4.1, + "learning_rate": 9.964726631393298e-06, + "loss": 0.0, + "step": 2586 + }, + { + "epoch": 4.11, + "learning_rate": 9.947089947089947e-06, + "loss": 0.0, + "step": 2587 + }, + { + "epoch": 4.11, + "learning_rate": 9.929453262786597e-06, + "loss": 0.0, + "step": 2588 + }, + { + "epoch": 4.11, + "learning_rate": 9.911816578483246e-06, + "loss": 0.0, + "step": 2589 + }, + { + "epoch": 4.11, + "learning_rate": 9.894179894179894e-06, + "loss": 0.0, + "step": 2590 + }, + { + "epoch": 4.11, + "learning_rate": 9.876543209876543e-06, + "loss": 0.0, + "step": 2591 + }, + { + "epoch": 4.11, + "learning_rate": 9.858906525573193e-06, + "loss": 0.0, + "step": 2592 + }, + { + "epoch": 4.12, + "learning_rate": 9.841269841269842e-06, + "loss": 0.0, + "step": 2593 + }, + { + "epoch": 4.12, + "learning_rate": 9.823633156966492e-06, + "loss": 0.0, + "step": 2594 + }, + { + "epoch": 4.12, + "learning_rate": 9.80599647266314e-06, + "loss": 0.0, + "step": 2595 + }, + { + "epoch": 4.12, + "learning_rate": 9.788359788359789e-06, + "loss": 0.0, + "step": 2596 + }, + { + "epoch": 4.12, + "learning_rate": 9.770723104056438e-06, + "loss": 0.0, + "step": 2597 + }, + { + "epoch": 4.12, + "learning_rate": 9.753086419753086e-06, + "loss": 0.0, + "step": 2598 + }, + { + "epoch": 4.13, + "learning_rate": 9.735449735449737e-06, + "loss": 0.0, + "step": 2599 + }, + { + "epoch": 4.13, + "learning_rate": 9.717813051146385e-06, + "loss": 0.0, + "step": 2600 + }, + { + "epoch": 4.13, + "learning_rate": 9.700176366843034e-06, + "loss": 0.0, + "step": 2601 + }, + { + "epoch": 4.13, + "learning_rate": 9.682539682539683e-06, + "loss": 0.0, + "step": 2602 + }, + { + "epoch": 4.13, + "learning_rate": 9.664902998236331e-06, + "loss": 0.0, + "step": 2603 + }, + { + "epoch": 4.13, + "learning_rate": 9.647266313932982e-06, + "loss": 0.0, + "step": 2604 + }, + { + "epoch": 4.13, + "learning_rate": 9.62962962962963e-06, + "loss": 0.0, + "step": 2605 + }, + { + "epoch": 4.14, + "learning_rate": 9.61199294532628e-06, + "loss": 0.0, + "step": 2606 + }, + { + "epoch": 4.14, + "learning_rate": 9.594356261022929e-06, + "loss": 0.0, + "step": 2607 + }, + { + "epoch": 4.14, + "learning_rate": 9.576719576719577e-06, + "loss": 0.0, + "step": 2608 + }, + { + "epoch": 4.14, + "learning_rate": 9.559082892416228e-06, + "loss": 0.0, + "step": 2609 + }, + { + "epoch": 4.14, + "learning_rate": 9.541446208112875e-06, + "loss": 0.0, + "step": 2610 + }, + { + "epoch": 4.14, + "learning_rate": 9.523809523809523e-06, + "loss": 0.0, + "step": 2611 + }, + { + "epoch": 4.15, + "learning_rate": 9.506172839506174e-06, + "loss": 0.0, + "step": 2612 + }, + { + "epoch": 4.15, + "learning_rate": 9.488536155202822e-06, + "loss": 0.0, + "step": 2613 + }, + { + "epoch": 4.15, + "learning_rate": 9.470899470899471e-06, + "loss": 0.0, + "step": 2614 + }, + { + "epoch": 4.15, + "learning_rate": 9.45326278659612e-06, + "loss": 0.0, + "step": 2615 + }, + { + "epoch": 4.15, + "learning_rate": 9.435626102292768e-06, + "loss": 0.0, + "step": 2616 + }, + { + "epoch": 4.15, + "learning_rate": 9.41798941798942e-06, + "loss": 0.0, + "step": 2617 + }, + { + "epoch": 4.16, + "learning_rate": 9.400352733686067e-06, + "loss": 0.0, + "step": 2618 + }, + { + "epoch": 4.16, + "learning_rate": 9.382716049382717e-06, + "loss": 0.0, + "step": 2619 + }, + { + "epoch": 4.16, + "learning_rate": 9.365079365079366e-06, + "loss": 0.0, + "step": 2620 + }, + { + "epoch": 4.16, + "learning_rate": 9.347442680776014e-06, + "loss": 0.0, + "step": 2621 + }, + { + "epoch": 4.16, + "learning_rate": 9.329805996472663e-06, + "loss": 0.0, + "step": 2622 + }, + { + "epoch": 4.16, + "learning_rate": 9.312169312169313e-06, + "loss": 0.0, + "step": 2623 + }, + { + "epoch": 4.17, + "learning_rate": 9.294532627865962e-06, + "loss": 0.0, + "step": 2624 + }, + { + "epoch": 4.17, + "learning_rate": 9.276895943562611e-06, + "loss": 0.0, + "step": 2625 + }, + { + "epoch": 4.17, + "learning_rate": 9.259259259259259e-06, + "loss": 0.0, + "step": 2626 + }, + { + "epoch": 4.17, + "learning_rate": 9.241622574955909e-06, + "loss": 0.0, + "step": 2627 + }, + { + "epoch": 4.17, + "learning_rate": 9.223985890652558e-06, + "loss": 0.0, + "step": 2628 + }, + { + "epoch": 4.17, + "learning_rate": 9.206349206349207e-06, + "loss": 0.0, + "step": 2629 + }, + { + "epoch": 4.17, + "learning_rate": 9.188712522045855e-06, + "loss": 0.0, + "step": 2630 + }, + { + "epoch": 4.18, + "learning_rate": 9.171075837742504e-06, + "loss": 0.0, + "step": 2631 + }, + { + "epoch": 4.18, + "learning_rate": 9.153439153439154e-06, + "loss": 0.0, + "step": 2632 + }, + { + "epoch": 4.18, + "learning_rate": 9.135802469135803e-06, + "loss": 0.0, + "step": 2633 + }, + { + "epoch": 4.18, + "learning_rate": 9.118165784832453e-06, + "loss": 0.0, + "step": 2634 + }, + { + "epoch": 4.18, + "learning_rate": 9.1005291005291e-06, + "loss": 0.0, + "step": 2635 + }, + { + "epoch": 4.18, + "learning_rate": 9.08289241622575e-06, + "loss": 0.0, + "step": 2636 + }, + { + "epoch": 4.19, + "learning_rate": 9.0652557319224e-06, + "loss": 0.0, + "step": 2637 + }, + { + "epoch": 4.19, + "learning_rate": 9.047619047619047e-06, + "loss": 0.0, + "step": 2638 + }, + { + "epoch": 4.19, + "learning_rate": 9.029982363315698e-06, + "loss": 0.0, + "step": 2639 + }, + { + "epoch": 4.19, + "learning_rate": 9.012345679012346e-06, + "loss": 0.0, + "step": 2640 + }, + { + "epoch": 4.19, + "learning_rate": 8.994708994708995e-06, + "loss": 0.0, + "step": 2641 + }, + { + "epoch": 4.19, + "learning_rate": 8.977072310405645e-06, + "loss": 0.0, + "step": 2642 + }, + { + "epoch": 4.2, + "learning_rate": 8.959435626102292e-06, + "loss": 0.0, + "step": 2643 + }, + { + "epoch": 4.2, + "learning_rate": 8.941798941798943e-06, + "loss": 0.0, + "step": 2644 + }, + { + "epoch": 4.2, + "learning_rate": 8.924162257495591e-06, + "loss": 0.0, + "step": 2645 + }, + { + "epoch": 4.2, + "learning_rate": 8.90652557319224e-06, + "loss": 0.0, + "step": 2646 + }, + { + "epoch": 4.2, + "learning_rate": 8.88888888888889e-06, + "loss": 0.0, + "step": 2647 + }, + { + "epoch": 4.2, + "learning_rate": 8.871252204585538e-06, + "loss": 0.0, + "step": 2648 + }, + { + "epoch": 4.2, + "learning_rate": 8.853615520282189e-06, + "loss": 0.0, + "step": 2649 + }, + { + "epoch": 4.21, + "learning_rate": 8.835978835978837e-06, + "loss": 0.0, + "step": 2650 + }, + { + "epoch": 4.21, + "learning_rate": 8.818342151675484e-06, + "loss": 0.0, + "step": 2651 + }, + { + "epoch": 4.21, + "learning_rate": 8.800705467372135e-06, + "loss": 0.0, + "step": 2652 + }, + { + "epoch": 4.21, + "learning_rate": 8.783068783068783e-06, + "loss": 0.0, + "step": 2653 + }, + { + "epoch": 4.21, + "learning_rate": 8.765432098765432e-06, + "loss": 0.0, + "step": 2654 + }, + { + "epoch": 4.21, + "learning_rate": 8.747795414462082e-06, + "loss": 0.0, + "step": 2655 + }, + { + "epoch": 4.22, + "learning_rate": 8.73015873015873e-06, + "loss": 0.0, + "step": 2656 + }, + { + "epoch": 4.22, + "learning_rate": 8.71252204585538e-06, + "loss": 0.0, + "step": 2657 + }, + { + "epoch": 4.22, + "learning_rate": 8.694885361552028e-06, + "loss": 0.0, + "step": 2658 + }, + { + "epoch": 4.22, + "learning_rate": 8.677248677248678e-06, + "loss": 0.0, + "step": 2659 + }, + { + "epoch": 4.22, + "learning_rate": 8.659611992945327e-06, + "loss": 0.0, + "step": 2660 + }, + { + "epoch": 4.22, + "learning_rate": 8.641975308641975e-06, + "loss": 0.0, + "step": 2661 + }, + { + "epoch": 4.23, + "learning_rate": 8.624338624338624e-06, + "loss": 0.0, + "step": 2662 + }, + { + "epoch": 4.23, + "learning_rate": 8.606701940035274e-06, + "loss": 0.0, + "step": 2663 + }, + { + "epoch": 4.23, + "learning_rate": 8.589065255731923e-06, + "loss": 0.0, + "step": 2664 + }, + { + "epoch": 4.23, + "learning_rate": 8.571428571428573e-06, + "loss": 0.0, + "step": 2665 + }, + { + "epoch": 4.23, + "learning_rate": 8.55379188712522e-06, + "loss": 0.0, + "step": 2666 + }, + { + "epoch": 4.23, + "learning_rate": 8.53615520282187e-06, + "loss": 0.0, + "step": 2667 + }, + { + "epoch": 4.23, + "learning_rate": 8.518518518518519e-06, + "loss": 0.0, + "step": 2668 + }, + { + "epoch": 4.24, + "learning_rate": 8.500881834215169e-06, + "loss": 0.0, + "step": 2669 + }, + { + "epoch": 4.24, + "learning_rate": 8.483245149911816e-06, + "loss": 0.0, + "step": 2670 + }, + { + "epoch": 4.24, + "learning_rate": 8.465608465608466e-06, + "loss": 0.0, + "step": 2671 + }, + { + "epoch": 4.24, + "learning_rate": 8.447971781305115e-06, + "loss": 0.0, + "step": 2672 + }, + { + "epoch": 4.24, + "learning_rate": 8.430335097001765e-06, + "loss": 0.0, + "step": 2673 + }, + { + "epoch": 4.24, + "learning_rate": 8.412698412698414e-06, + "loss": 0.0, + "step": 2674 + }, + { + "epoch": 4.25, + "learning_rate": 8.395061728395062e-06, + "loss": 0.0, + "step": 2675 + }, + { + "epoch": 4.25, + "learning_rate": 8.377425044091711e-06, + "loss": 0.0, + "step": 2676 + }, + { + "epoch": 4.25, + "learning_rate": 8.35978835978836e-06, + "loss": 0.0, + "step": 2677 + }, + { + "epoch": 4.25, + "learning_rate": 8.34215167548501e-06, + "loss": 0.0, + "step": 2678 + }, + { + "epoch": 4.25, + "learning_rate": 8.32451499118166e-06, + "loss": 0.0, + "step": 2679 + }, + { + "epoch": 4.25, + "learning_rate": 8.306878306878307e-06, + "loss": 0.0, + "step": 2680 + }, + { + "epoch": 4.26, + "learning_rate": 8.289241622574956e-06, + "loss": 0.0, + "step": 2681 + }, + { + "epoch": 4.26, + "learning_rate": 8.271604938271606e-06, + "loss": 0.0, + "step": 2682 + }, + { + "epoch": 4.26, + "learning_rate": 8.253968253968254e-06, + "loss": 0.0, + "step": 2683 + }, + { + "epoch": 4.26, + "learning_rate": 8.236331569664905e-06, + "loss": 0.0, + "step": 2684 + }, + { + "epoch": 4.26, + "learning_rate": 8.218694885361552e-06, + "loss": 0.0, + "step": 2685 + }, + { + "epoch": 4.26, + "learning_rate": 8.201058201058202e-06, + "loss": 0.0, + "step": 2686 + }, + { + "epoch": 4.27, + "learning_rate": 8.183421516754851e-06, + "loss": 0.0, + "step": 2687 + }, + { + "epoch": 4.27, + "learning_rate": 8.165784832451499e-06, + "loss": 0.0, + "step": 2688 + }, + { + "epoch": 4.27, + "learning_rate": 8.14814814814815e-06, + "loss": 0.0, + "step": 2689 + }, + { + "epoch": 4.27, + "learning_rate": 8.130511463844798e-06, + "loss": 0.0, + "step": 2690 + }, + { + "epoch": 4.27, + "learning_rate": 8.112874779541445e-06, + "loss": 0.0, + "step": 2691 + }, + { + "epoch": 4.27, + "learning_rate": 8.095238095238097e-06, + "loss": 0.0, + "step": 2692 + }, + { + "epoch": 4.27, + "learning_rate": 8.077601410934744e-06, + "loss": 0.0, + "step": 2693 + }, + { + "epoch": 4.28, + "learning_rate": 8.059964726631394e-06, + "loss": 0.0, + "step": 2694 + }, + { + "epoch": 4.28, + "learning_rate": 8.042328042328043e-06, + "loss": 0.0, + "step": 2695 + }, + { + "epoch": 4.28, + "learning_rate": 8.02469135802469e-06, + "loss": 0.0, + "step": 2696 + }, + { + "epoch": 4.28, + "learning_rate": 8.007054673721342e-06, + "loss": 0.0, + "step": 2697 + }, + { + "epoch": 4.28, + "learning_rate": 7.98941798941799e-06, + "loss": 0.0, + "step": 2698 + }, + { + "epoch": 4.28, + "learning_rate": 7.971781305114639e-06, + "loss": 0.0, + "step": 2699 + }, + { + "epoch": 4.29, + "learning_rate": 7.954144620811288e-06, + "loss": 0.0, + "step": 2700 + }, + { + "epoch": 4.29, + "learning_rate": 7.936507936507936e-06, + "loss": 0.0, + "step": 2701 + }, + { + "epoch": 4.29, + "learning_rate": 7.918871252204586e-06, + "loss": 0.0, + "step": 2702 + }, + { + "epoch": 4.29, + "learning_rate": 7.901234567901235e-06, + "loss": 0.0, + "step": 2703 + }, + { + "epoch": 4.29, + "learning_rate": 7.883597883597884e-06, + "loss": 0.0, + "step": 2704 + }, + { + "epoch": 4.29, + "learning_rate": 7.865961199294534e-06, + "loss": 0.0, + "step": 2705 + }, + { + "epoch": 4.3, + "learning_rate": 7.848324514991182e-06, + "loss": 0.0, + "step": 2706 + }, + { + "epoch": 4.3, + "learning_rate": 7.830687830687831e-06, + "loss": 0.0, + "step": 2707 + }, + { + "epoch": 4.3, + "learning_rate": 7.81305114638448e-06, + "loss": 0.0, + "step": 2708 + }, + { + "epoch": 4.3, + "learning_rate": 7.79541446208113e-06, + "loss": 0.0, + "step": 2709 + }, + { + "epoch": 4.3, + "learning_rate": 7.777777777777777e-06, + "loss": 0.0, + "step": 2710 + }, + { + "epoch": 4.3, + "learning_rate": 7.760141093474427e-06, + "loss": 0.0, + "step": 2711 + }, + { + "epoch": 4.3, + "learning_rate": 7.742504409171076e-06, + "loss": 0.0, + "step": 2712 + }, + { + "epoch": 4.31, + "learning_rate": 7.724867724867726e-06, + "loss": 0.0, + "step": 2713 + }, + { + "epoch": 4.31, + "learning_rate": 7.707231040564375e-06, + "loss": 0.0, + "step": 2714 + }, + { + "epoch": 4.31, + "learning_rate": 7.689594356261023e-06, + "loss": 0.0, + "step": 2715 + }, + { + "epoch": 4.31, + "learning_rate": 7.671957671957672e-06, + "loss": 0.0, + "step": 2716 + }, + { + "epoch": 4.31, + "learning_rate": 7.654320987654322e-06, + "loss": 0.0, + "step": 2717 + }, + { + "epoch": 4.31, + "learning_rate": 7.636684303350971e-06, + "loss": 0.0, + "step": 2718 + }, + { + "epoch": 4.32, + "learning_rate": 7.6190476190476205e-06, + "loss": 0.0, + "step": 2719 + }, + { + "epoch": 4.32, + "learning_rate": 7.601410934744268e-06, + "loss": 0.0, + "step": 2720 + }, + { + "epoch": 4.32, + "learning_rate": 7.583774250440917e-06, + "loss": 0.0, + "step": 2721 + }, + { + "epoch": 4.32, + "learning_rate": 7.566137566137567e-06, + "loss": 0.0, + "step": 2722 + }, + { + "epoch": 4.32, + "learning_rate": 7.548500881834216e-06, + "loss": 0.0, + "step": 2723 + }, + { + "epoch": 4.32, + "learning_rate": 7.530864197530865e-06, + "loss": 0.0, + "step": 2724 + }, + { + "epoch": 4.33, + "learning_rate": 7.5132275132275136e-06, + "loss": 0.0, + "step": 2725 + }, + { + "epoch": 4.33, + "learning_rate": 7.495590828924162e-06, + "loss": 0.0, + "step": 2726 + }, + { + "epoch": 4.33, + "learning_rate": 7.477954144620812e-06, + "loss": 0.0, + "step": 2727 + }, + { + "epoch": 4.33, + "learning_rate": 7.460317460317461e-06, + "loss": 0.0, + "step": 2728 + }, + { + "epoch": 4.33, + "learning_rate": 7.442680776014109e-06, + "loss": 0.0, + "step": 2729 + }, + { + "epoch": 4.33, + "learning_rate": 7.425044091710759e-06, + "loss": 0.0, + "step": 2730 + }, + { + "epoch": 4.33, + "learning_rate": 7.4074074074074075e-06, + "loss": 0.0, + "step": 2731 + }, + { + "epoch": 4.34, + "learning_rate": 7.389770723104057e-06, + "loss": 0.0, + "step": 2732 + }, + { + "epoch": 4.34, + "learning_rate": 7.3721340388007055e-06, + "loss": 0.0, + "step": 2733 + }, + { + "epoch": 4.34, + "learning_rate": 7.354497354497354e-06, + "loss": 0.0, + "step": 2734 + }, + { + "epoch": 4.34, + "learning_rate": 7.336860670194004e-06, + "loss": 0.0, + "step": 2735 + }, + { + "epoch": 4.34, + "learning_rate": 7.319223985890653e-06, + "loss": 0.0, + "step": 2736 + }, + { + "epoch": 4.34, + "learning_rate": 7.301587301587302e-06, + "loss": 0.0, + "step": 2737 + }, + { + "epoch": 4.35, + "learning_rate": 7.283950617283951e-06, + "loss": 0.0, + "step": 2738 + }, + { + "epoch": 4.35, + "learning_rate": 7.266313932980599e-06, + "loss": 0.0, + "step": 2739 + }, + { + "epoch": 4.35, + "learning_rate": 7.248677248677249e-06, + "loss": 0.0, + "step": 2740 + }, + { + "epoch": 4.35, + "learning_rate": 7.231040564373897e-06, + "loss": 0.0, + "step": 2741 + }, + { + "epoch": 4.35, + "learning_rate": 7.213403880070548e-06, + "loss": 0.0, + "step": 2742 + }, + { + "epoch": 4.35, + "learning_rate": 7.195767195767196e-06, + "loss": 0.0, + "step": 2743 + }, + { + "epoch": 4.36, + "learning_rate": 7.178130511463845e-06, + "loss": 0.0, + "step": 2744 + }, + { + "epoch": 4.36, + "learning_rate": 7.160493827160494e-06, + "loss": 0.0, + "step": 2745 + }, + { + "epoch": 4.36, + "learning_rate": 7.142857142857143e-06, + "loss": 0.0, + "step": 2746 + }, + { + "epoch": 4.36, + "learning_rate": 7.125220458553793e-06, + "loss": 0.0, + "step": 2747 + }, + { + "epoch": 4.36, + "learning_rate": 7.1075837742504415e-06, + "loss": 0.0, + "step": 2748 + }, + { + "epoch": 4.36, + "learning_rate": 7.089947089947089e-06, + "loss": 0.0, + "step": 2749 + }, + { + "epoch": 4.37, + "learning_rate": 7.0723104056437395e-06, + "loss": 0.0, + "step": 2750 + }, + { + "epoch": 4.37, + "learning_rate": 7.054673721340388e-06, + "loss": 0.0, + "step": 2751 + }, + { + "epoch": 4.37, + "learning_rate": 7.0370370370370375e-06, + "loss": 0.0, + "step": 2752 + }, + { + "epoch": 4.37, + "learning_rate": 7.019400352733686e-06, + "loss": 0.0, + "step": 2753 + }, + { + "epoch": 4.37, + "learning_rate": 7.001763668430335e-06, + "loss": 0.0, + "step": 2754 + }, + { + "epoch": 4.37, + "learning_rate": 6.984126984126985e-06, + "loss": 0.0, + "step": 2755 + }, + { + "epoch": 4.37, + "learning_rate": 6.9664902998236334e-06, + "loss": 0.0, + "step": 2756 + }, + { + "epoch": 4.38, + "learning_rate": 6.948853615520283e-06, + "loss": 0.0, + "step": 2757 + }, + { + "epoch": 4.38, + "learning_rate": 6.931216931216931e-06, + "loss": 0.0, + "step": 2758 + }, + { + "epoch": 4.38, + "learning_rate": 6.91358024691358e-06, + "loss": 0.0, + "step": 2759 + }, + { + "epoch": 4.38, + "learning_rate": 6.895943562610229e-06, + "loss": 0.0, + "step": 2760 + }, + { + "epoch": 4.38, + "learning_rate": 6.878306878306878e-06, + "loss": 0.0, + "step": 2761 + }, + { + "epoch": 4.38, + "learning_rate": 6.860670194003528e-06, + "loss": 0.0, + "step": 2762 + }, + { + "epoch": 4.39, + "learning_rate": 6.843033509700177e-06, + "loss": 0.0, + "step": 2763 + }, + { + "epoch": 4.39, + "learning_rate": 6.825396825396825e-06, + "loss": 0.0, + "step": 2764 + }, + { + "epoch": 4.39, + "learning_rate": 6.807760141093475e-06, + "loss": 0.0, + "step": 2765 + }, + { + "epoch": 4.39, + "learning_rate": 6.790123456790123e-06, + "loss": 0.0, + "step": 2766 + }, + { + "epoch": 4.39, + "learning_rate": 6.7724867724867736e-06, + "loss": 0.0, + "step": 2767 + }, + { + "epoch": 4.39, + "learning_rate": 6.754850088183422e-06, + "loss": 0.0, + "step": 2768 + }, + { + "epoch": 4.4, + "learning_rate": 6.73721340388007e-06, + "loss": 0.0, + "step": 2769 + }, + { + "epoch": 4.4, + "learning_rate": 6.71957671957672e-06, + "loss": 0.0, + "step": 2770 + }, + { + "epoch": 4.4, + "learning_rate": 6.701940035273369e-06, + "loss": 0.0, + "step": 2771 + }, + { + "epoch": 4.4, + "learning_rate": 6.684303350970018e-06, + "loss": 0.0, + "step": 2772 + }, + { + "epoch": 4.4, + "learning_rate": 6.666666666666667e-06, + "loss": 0.0, + "step": 2773 + }, + { + "epoch": 4.4, + "learning_rate": 6.649029982363315e-06, + "loss": 0.0, + "step": 2774 + }, + { + "epoch": 4.4, + "learning_rate": 6.6313932980599655e-06, + "loss": 0.0, + "step": 2775 + }, + { + "epoch": 4.41, + "learning_rate": 6.613756613756614e-06, + "loss": 0.0, + "step": 2776 + }, + { + "epoch": 4.41, + "learning_rate": 6.5961199294532634e-06, + "loss": 0.0, + "step": 2777 + }, + { + "epoch": 4.41, + "learning_rate": 6.578483245149912e-06, + "loss": 0.0, + "step": 2778 + }, + { + "epoch": 4.41, + "learning_rate": 6.5608465608465606e-06, + "loss": 0.0, + "step": 2779 + }, + { + "epoch": 4.41, + "learning_rate": 6.54320987654321e-06, + "loss": 0.0, + "step": 2780 + }, + { + "epoch": 4.41, + "learning_rate": 6.5255731922398585e-06, + "loss": 0.0, + "step": 2781 + }, + { + "epoch": 4.42, + "learning_rate": 6.507936507936509e-06, + "loss": 0.0, + "step": 2782 + }, + { + "epoch": 4.42, + "learning_rate": 6.490299823633157e-06, + "loss": 0.0, + "step": 2783 + }, + { + "epoch": 4.42, + "learning_rate": 6.472663139329806e-06, + "loss": 0.0, + "step": 2784 + }, + { + "epoch": 4.42, + "learning_rate": 6.455026455026455e-06, + "loss": 0.0, + "step": 2785 + }, + { + "epoch": 4.42, + "learning_rate": 6.437389770723104e-06, + "loss": 0.0, + "step": 2786 + }, + { + "epoch": 4.42, + "learning_rate": 6.419753086419754e-06, + "loss": 0.0, + "step": 2787 + }, + { + "epoch": 4.43, + "learning_rate": 6.402116402116403e-06, + "loss": 0.0, + "step": 2788 + }, + { + "epoch": 4.43, + "learning_rate": 6.3844797178130504e-06, + "loss": 0.0, + "step": 2789 + }, + { + "epoch": 4.43, + "learning_rate": 6.366843033509701e-06, + "loss": 0.0, + "step": 2790 + }, + { + "epoch": 4.43, + "learning_rate": 6.349206349206349e-06, + "loss": 0.0, + "step": 2791 + }, + { + "epoch": 4.43, + "learning_rate": 6.331569664902999e-06, + "loss": 0.0, + "step": 2792 + }, + { + "epoch": 4.43, + "learning_rate": 6.313932980599647e-06, + "loss": 0.0, + "step": 2793 + }, + { + "epoch": 4.43, + "learning_rate": 6.296296296296296e-06, + "loss": 0.0, + "step": 2794 + }, + { + "epoch": 4.44, + "learning_rate": 6.278659611992946e-06, + "loss": 0.0, + "step": 2795 + }, + { + "epoch": 4.44, + "learning_rate": 6.261022927689595e-06, + "loss": 0.0, + "step": 2796 + }, + { + "epoch": 4.44, + "learning_rate": 6.243386243386243e-06, + "loss": 0.0, + "step": 2797 + }, + { + "epoch": 4.44, + "learning_rate": 6.225749559082893e-06, + "loss": 0.0, + "step": 2798 + }, + { + "epoch": 4.44, + "learning_rate": 6.208112874779542e-06, + "loss": 0.0, + "step": 2799 + }, + { + "epoch": 4.44, + "learning_rate": 6.190476190476191e-06, + "loss": 0.0, + "step": 2800 + }, + { + "epoch": 4.45, + "learning_rate": 6.172839506172839e-06, + "loss": 0.0, + "step": 2801 + }, + { + "epoch": 4.45, + "learning_rate": 6.1552028218694886e-06, + "loss": 0.0, + "step": 2802 + }, + { + "epoch": 4.45, + "learning_rate": 6.137566137566138e-06, + "loss": 0.0, + "step": 2803 + }, + { + "epoch": 4.45, + "learning_rate": 6.119929453262787e-06, + "loss": 0.0, + "step": 2804 + }, + { + "epoch": 4.45, + "learning_rate": 6.102292768959436e-06, + "loss": 0.0, + "step": 2805 + }, + { + "epoch": 4.45, + "learning_rate": 6.0846560846560845e-06, + "loss": 0.0, + "step": 2806 + }, + { + "epoch": 4.46, + "learning_rate": 6.067019400352734e-06, + "loss": 0.0, + "step": 2807 + }, + { + "epoch": 4.46, + "learning_rate": 6.049382716049383e-06, + "loss": 0.0, + "step": 2808 + }, + { + "epoch": 4.46, + "learning_rate": 6.031746031746032e-06, + "loss": 0.0, + "step": 2809 + }, + { + "epoch": 4.46, + "learning_rate": 6.014109347442681e-06, + "loss": 0.0, + "step": 2810 + }, + { + "epoch": 4.46, + "learning_rate": 5.99647266313933e-06, + "loss": 0.0, + "step": 2811 + }, + { + "epoch": 4.46, + "learning_rate": 5.978835978835979e-06, + "loss": 0.0, + "step": 2812 + }, + { + "epoch": 4.47, + "learning_rate": 5.961199294532628e-06, + "loss": 0.0, + "step": 2813 + }, + { + "epoch": 4.47, + "learning_rate": 5.943562610229277e-06, + "loss": 0.0, + "step": 2814 + }, + { + "epoch": 4.47, + "learning_rate": 5.925925925925927e-06, + "loss": 0.0, + "step": 2815 + }, + { + "epoch": 4.47, + "learning_rate": 5.908289241622575e-06, + "loss": 0.0, + "step": 2816 + }, + { + "epoch": 4.47, + "learning_rate": 5.890652557319224e-06, + "loss": 0.0, + "step": 2817 + }, + { + "epoch": 4.47, + "learning_rate": 5.873015873015873e-06, + "loss": 0.0, + "step": 2818 + }, + { + "epoch": 4.47, + "learning_rate": 5.855379188712523e-06, + "loss": 0.0, + "step": 2819 + }, + { + "epoch": 4.48, + "learning_rate": 5.837742504409171e-06, + "loss": 0.0, + "step": 2820 + }, + { + "epoch": 4.48, + "learning_rate": 5.82010582010582e-06, + "loss": 0.0, + "step": 2821 + }, + { + "epoch": 4.48, + "learning_rate": 5.802469135802469e-06, + "loss": 0.0, + "step": 2822 + }, + { + "epoch": 4.48, + "learning_rate": 5.7848324514991186e-06, + "loss": 0.0, + "step": 2823 + }, + { + "epoch": 4.48, + "learning_rate": 5.767195767195768e-06, + "loss": 0.0, + "step": 2824 + }, + { + "epoch": 4.48, + "learning_rate": 5.7495590828924165e-06, + "loss": 0.0, + "step": 2825 + }, + { + "epoch": 4.49, + "learning_rate": 5.731922398589065e-06, + "loss": 0.0, + "step": 2826 + }, + { + "epoch": 4.49, + "learning_rate": 5.7142857142857145e-06, + "loss": 0.0, + "step": 2827 + }, + { + "epoch": 4.49, + "learning_rate": 5.696649029982364e-06, + "loss": 0.0, + "step": 2828 + }, + { + "epoch": 4.49, + "learning_rate": 5.6790123456790125e-06, + "loss": 0.0, + "step": 2829 + }, + { + "epoch": 4.49, + "learning_rate": 5.661375661375662e-06, + "loss": 0.0, + "step": 2830 + }, + { + "epoch": 4.49, + "learning_rate": 5.6437389770723105e-06, + "loss": 0.0, + "step": 2831 + }, + { + "epoch": 4.5, + "learning_rate": 5.62610229276896e-06, + "loss": 0.0, + "step": 2832 + }, + { + "epoch": 4.5, + "learning_rate": 5.6084656084656084e-06, + "loss": 0.0, + "step": 2833 + }, + { + "epoch": 4.5, + "learning_rate": 5.590828924162258e-06, + "loss": 0.0, + "step": 2834 + }, + { + "epoch": 4.5, + "learning_rate": 5.573192239858907e-06, + "loss": 0.0, + "step": 2835 + }, + { + "epoch": 4.5, + "learning_rate": 5.555555555555556e-06, + "loss": 0.0, + "step": 2836 + }, + { + "epoch": 4.5, + "learning_rate": 5.537918871252204e-06, + "loss": 0.0, + "step": 2837 + }, + { + "epoch": 4.5, + "learning_rate": 5.520282186948854e-06, + "loss": 0.0, + "step": 2838 + }, + { + "epoch": 4.51, + "learning_rate": 5.502645502645503e-06, + "loss": 0.0, + "step": 2839 + }, + { + "epoch": 4.51, + "learning_rate": 5.485008818342152e-06, + "loss": 0.0, + "step": 2840 + }, + { + "epoch": 4.51, + "learning_rate": 5.4673721340388e-06, + "loss": 0.0, + "step": 2841 + }, + { + "epoch": 4.51, + "learning_rate": 5.44973544973545e-06, + "loss": 0.0, + "step": 2842 + }, + { + "epoch": 4.51, + "learning_rate": 5.432098765432099e-06, + "loss": 0.0, + "step": 2843 + }, + { + "epoch": 4.51, + "learning_rate": 5.4144620811287486e-06, + "loss": 0.0, + "step": 2844 + }, + { + "epoch": 4.52, + "learning_rate": 5.396825396825397e-06, + "loss": 0.0, + "step": 2845 + }, + { + "epoch": 4.52, + "learning_rate": 5.379188712522046e-06, + "loss": 0.0, + "step": 2846 + }, + { + "epoch": 4.52, + "learning_rate": 5.361552028218695e-06, + "loss": 0.0, + "step": 2847 + }, + { + "epoch": 4.52, + "learning_rate": 5.3439153439153445e-06, + "loss": 0.0, + "step": 2848 + }, + { + "epoch": 4.52, + "learning_rate": 5.326278659611993e-06, + "loss": 0.0, + "step": 2849 + }, + { + "epoch": 4.52, + "learning_rate": 5.3086419753086425e-06, + "loss": 0.0, + "step": 2850 + }, + { + "epoch": 4.53, + "learning_rate": 5.291005291005291e-06, + "loss": 0.0, + "step": 2851 + }, + { + "epoch": 4.53, + "learning_rate": 5.2733686067019405e-06, + "loss": 0.0, + "step": 2852 + }, + { + "epoch": 4.53, + "learning_rate": 5.255731922398589e-06, + "loss": 0.0, + "step": 2853 + }, + { + "epoch": 4.53, + "learning_rate": 5.2380952380952384e-06, + "loss": 0.0, + "step": 2854 + }, + { + "epoch": 4.53, + "learning_rate": 5.220458553791888e-06, + "loss": 0.0, + "step": 2855 + }, + { + "epoch": 4.53, + "learning_rate": 5.202821869488536e-06, + "loss": 0.0, + "step": 2856 + }, + { + "epoch": 4.53, + "learning_rate": 5.185185185185185e-06, + "loss": 0.0, + "step": 2857 + }, + { + "epoch": 4.54, + "learning_rate": 5.167548500881834e-06, + "loss": 0.0, + "step": 2858 + }, + { + "epoch": 4.54, + "learning_rate": 5.149911816578484e-06, + "loss": 0.0, + "step": 2859 + }, + { + "epoch": 4.54, + "learning_rate": 5.132275132275132e-06, + "loss": 0.0, + "step": 2860 + }, + { + "epoch": 4.54, + "learning_rate": 5.114638447971781e-06, + "loss": 0.0, + "step": 2861 + }, + { + "epoch": 4.54, + "learning_rate": 5.09700176366843e-06, + "loss": 0.0, + "step": 2862 + }, + { + "epoch": 4.54, + "learning_rate": 5.07936507936508e-06, + "loss": 0.0, + "step": 2863 + }, + { + "epoch": 4.55, + "learning_rate": 5.061728395061729e-06, + "loss": 0.0, + "step": 2864 + }, + { + "epoch": 4.55, + "learning_rate": 5.044091710758378e-06, + "loss": 0.0, + "step": 2865 + }, + { + "epoch": 4.55, + "learning_rate": 5.026455026455026e-06, + "loss": 0.0, + "step": 2866 + }, + { + "epoch": 4.55, + "learning_rate": 5.008818342151676e-06, + "loss": 0.0, + "step": 2867 + }, + { + "epoch": 4.55, + "learning_rate": 4.991181657848325e-06, + "loss": 0.0, + "step": 2868 + }, + { + "epoch": 4.55, + "learning_rate": 4.973544973544974e-06, + "loss": 0.0, + "step": 2869 + }, + { + "epoch": 4.56, + "learning_rate": 4.955908289241623e-06, + "loss": 0.0, + "step": 2870 + }, + { + "epoch": 4.56, + "learning_rate": 4.938271604938272e-06, + "loss": 0.0, + "step": 2871 + }, + { + "epoch": 4.56, + "learning_rate": 4.920634920634921e-06, + "loss": 0.0, + "step": 2872 + }, + { + "epoch": 4.56, + "learning_rate": 4.90299823633157e-06, + "loss": 0.0007, + "step": 2873 + }, + { + "epoch": 4.56, + "learning_rate": 4.885361552028219e-06, + "loss": 0.0, + "step": 2874 + }, + { + "epoch": 4.56, + "learning_rate": 4.8677248677248685e-06, + "loss": 0.0, + "step": 2875 + }, + { + "epoch": 4.57, + "learning_rate": 4.850088183421517e-06, + "loss": 0.0, + "step": 2876 + }, + { + "epoch": 4.57, + "learning_rate": 4.832451499118166e-06, + "loss": 0.0, + "step": 2877 + }, + { + "epoch": 4.57, + "learning_rate": 4.814814814814815e-06, + "loss": 0.0, + "step": 2878 + }, + { + "epoch": 4.57, + "learning_rate": 4.797178130511464e-06, + "loss": 0.0, + "step": 2879 + }, + { + "epoch": 4.57, + "learning_rate": 4.779541446208114e-06, + "loss": 0.0, + "step": 2880 + }, + { + "epoch": 4.57, + "learning_rate": 4.7619047619047615e-06, + "loss": 0.0, + "step": 2881 + }, + { + "epoch": 4.57, + "learning_rate": 4.744268077601411e-06, + "loss": 0.0, + "step": 2882 + }, + { + "epoch": 4.58, + "learning_rate": 4.72663139329806e-06, + "loss": 0.0, + "step": 2883 + }, + { + "epoch": 4.58, + "learning_rate": 4.70899470899471e-06, + "loss": 0.0, + "step": 2884 + }, + { + "epoch": 4.58, + "learning_rate": 4.691358024691358e-06, + "loss": 0.0, + "step": 2885 + }, + { + "epoch": 4.58, + "learning_rate": 4.673721340388007e-06, + "loss": 0.0, + "step": 2886 + }, + { + "epoch": 4.58, + "learning_rate": 4.656084656084656e-06, + "loss": 0.0, + "step": 2887 + }, + { + "epoch": 4.58, + "learning_rate": 4.638447971781306e-06, + "loss": 0.0, + "step": 2888 + }, + { + "epoch": 4.59, + "learning_rate": 4.620811287477954e-06, + "loss": 0.0, + "step": 2889 + }, + { + "epoch": 4.59, + "learning_rate": 4.603174603174604e-06, + "loss": 0.0, + "step": 2890 + }, + { + "epoch": 4.59, + "learning_rate": 4.585537918871252e-06, + "loss": 0.0, + "step": 2891 + }, + { + "epoch": 4.59, + "learning_rate": 4.567901234567902e-06, + "loss": 0.0, + "step": 2892 + }, + { + "epoch": 4.59, + "learning_rate": 4.55026455026455e-06, + "loss": 0.0, + "step": 2893 + }, + { + "epoch": 4.59, + "learning_rate": 4.5326278659612e-06, + "loss": 0.0, + "step": 2894 + }, + { + "epoch": 4.6, + "learning_rate": 4.514991181657849e-06, + "loss": 0.0, + "step": 2895 + }, + { + "epoch": 4.6, + "learning_rate": 4.497354497354498e-06, + "loss": 0.0, + "step": 2896 + }, + { + "epoch": 4.6, + "learning_rate": 4.479717813051146e-06, + "loss": 0.0, + "step": 2897 + }, + { + "epoch": 4.6, + "learning_rate": 4.462081128747796e-06, + "loss": 0.0, + "step": 2898 + }, + { + "epoch": 4.6, + "learning_rate": 4.444444444444445e-06, + "loss": 0.0, + "step": 2899 + }, + { + "epoch": 4.6, + "learning_rate": 4.426807760141094e-06, + "loss": 0.0, + "step": 2900 + }, + { + "epoch": 4.6, + "learning_rate": 4.409171075837742e-06, + "loss": 0.0, + "step": 2901 + }, + { + "epoch": 4.61, + "learning_rate": 4.3915343915343915e-06, + "loss": 0.0, + "step": 2902 + }, + { + "epoch": 4.61, + "learning_rate": 4.373897707231041e-06, + "loss": 0.0, + "step": 2903 + }, + { + "epoch": 4.61, + "learning_rate": 4.35626102292769e-06, + "loss": 0.0, + "step": 2904 + }, + { + "epoch": 4.61, + "learning_rate": 4.338624338624339e-06, + "loss": 0.0, + "step": 2905 + }, + { + "epoch": 4.61, + "learning_rate": 4.3209876543209875e-06, + "loss": 0.0001, + "step": 2906 + }, + { + "epoch": 4.61, + "learning_rate": 4.303350970017637e-06, + "loss": 0.0, + "step": 2907 + }, + { + "epoch": 4.62, + "learning_rate": 4.285714285714286e-06, + "loss": 0.0, + "step": 2908 + }, + { + "epoch": 4.62, + "learning_rate": 4.268077601410935e-06, + "loss": 0.0, + "step": 2909 + }, + { + "epoch": 4.62, + "learning_rate": 4.250440917107584e-06, + "loss": 0.0, + "step": 2910 + }, + { + "epoch": 4.62, + "learning_rate": 4.232804232804233e-06, + "loss": 0.0, + "step": 2911 + }, + { + "epoch": 4.62, + "learning_rate": 4.215167548500882e-06, + "loss": 0.0, + "step": 2912 + }, + { + "epoch": 4.62, + "learning_rate": 4.197530864197531e-06, + "loss": 0.0, + "step": 2913 + }, + { + "epoch": 4.63, + "learning_rate": 4.17989417989418e-06, + "loss": 0.0, + "step": 2914 + }, + { + "epoch": 4.63, + "learning_rate": 4.16225749559083e-06, + "loss": 0.0, + "step": 2915 + }, + { + "epoch": 4.63, + "learning_rate": 4.144620811287478e-06, + "loss": 0.0, + "step": 2916 + }, + { + "epoch": 4.63, + "learning_rate": 4.126984126984127e-06, + "loss": 0.0, + "step": 2917 + }, + { + "epoch": 4.63, + "learning_rate": 4.109347442680776e-06, + "loss": 0.0, + "step": 2918 + }, + { + "epoch": 4.63, + "learning_rate": 4.091710758377426e-06, + "loss": 0.0, + "step": 2919 + }, + { + "epoch": 4.63, + "learning_rate": 4.074074074074075e-06, + "loss": 0.0, + "step": 2920 + }, + { + "epoch": 4.64, + "learning_rate": 4.056437389770723e-06, + "loss": 0.0, + "step": 2921 + }, + { + "epoch": 4.64, + "learning_rate": 4.038800705467372e-06, + "loss": 0.0, + "step": 2922 + }, + { + "epoch": 4.64, + "learning_rate": 4.0211640211640215e-06, + "loss": 0.0, + "step": 2923 + }, + { + "epoch": 4.64, + "learning_rate": 4.003527336860671e-06, + "loss": 0.0, + "step": 2924 + }, + { + "epoch": 4.64, + "learning_rate": 3.9858906525573195e-06, + "loss": 0.0, + "step": 2925 + }, + { + "epoch": 4.64, + "learning_rate": 3.968253968253968e-06, + "loss": 0.0, + "step": 2926 + }, + { + "epoch": 4.65, + "learning_rate": 3.9506172839506175e-06, + "loss": 0.0, + "step": 2927 + }, + { + "epoch": 4.65, + "learning_rate": 3.932980599647267e-06, + "loss": 0.0, + "step": 2928 + }, + { + "epoch": 4.65, + "learning_rate": 3.9153439153439155e-06, + "loss": 0.0, + "step": 2929 + }, + { + "epoch": 4.65, + "learning_rate": 3.897707231040565e-06, + "loss": 0.0, + "step": 2930 + }, + { + "epoch": 4.65, + "learning_rate": 3.8800705467372134e-06, + "loss": 0.0, + "step": 2931 + }, + { + "epoch": 4.65, + "learning_rate": 3.862433862433863e-06, + "loss": 0.0, + "step": 2932 + }, + { + "epoch": 4.66, + "learning_rate": 3.844797178130511e-06, + "loss": 0.0, + "step": 2933 + }, + { + "epoch": 4.66, + "learning_rate": 3.827160493827161e-06, + "loss": 0.0, + "step": 2934 + }, + { + "epoch": 4.66, + "learning_rate": 3.8095238095238102e-06, + "loss": 0.0, + "step": 2935 + }, + { + "epoch": 4.66, + "learning_rate": 3.7918871252204584e-06, + "loss": 0.0, + "step": 2936 + }, + { + "epoch": 4.66, + "learning_rate": 3.774250440917108e-06, + "loss": 0.0, + "step": 2937 + }, + { + "epoch": 4.66, + "learning_rate": 3.7566137566137568e-06, + "loss": 0.0, + "step": 2938 + }, + { + "epoch": 4.67, + "learning_rate": 3.738977072310406e-06, + "loss": 0.0, + "step": 2939 + }, + { + "epoch": 4.67, + "learning_rate": 3.7213403880070543e-06, + "loss": 0.0, + "step": 2940 + }, + { + "epoch": 4.67, + "learning_rate": 3.7037037037037037e-06, + "loss": 0.0, + "step": 2941 + }, + { + "epoch": 4.67, + "learning_rate": 3.6860670194003527e-06, + "loss": 0.0, + "step": 2942 + }, + { + "epoch": 4.67, + "learning_rate": 3.668430335097002e-06, + "loss": 0.0, + "step": 2943 + }, + { + "epoch": 4.67, + "learning_rate": 3.650793650793651e-06, + "loss": 0.0, + "step": 2944 + }, + { + "epoch": 4.67, + "learning_rate": 3.6331569664902997e-06, + "loss": 0.0, + "step": 2945 + }, + { + "epoch": 4.68, + "learning_rate": 3.6155202821869487e-06, + "loss": 0.0, + "step": 2946 + }, + { + "epoch": 4.68, + "learning_rate": 3.597883597883598e-06, + "loss": 0.0007, + "step": 2947 + }, + { + "epoch": 4.68, + "learning_rate": 3.580246913580247e-06, + "loss": 0.0, + "step": 2948 + }, + { + "epoch": 4.68, + "learning_rate": 3.5626102292768965e-06, + "loss": 0.0, + "step": 2949 + }, + { + "epoch": 4.68, + "learning_rate": 3.5449735449735446e-06, + "loss": 0.0, + "step": 2950 + }, + { + "epoch": 4.68, + "learning_rate": 3.527336860670194e-06, + "loss": 0.0, + "step": 2951 + }, + { + "epoch": 4.69, + "learning_rate": 3.509700176366843e-06, + "loss": 0.0, + "step": 2952 + }, + { + "epoch": 4.69, + "learning_rate": 3.4920634920634924e-06, + "loss": 0.0, + "step": 2953 + }, + { + "epoch": 4.69, + "learning_rate": 3.4744268077601414e-06, + "loss": 0.0, + "step": 2954 + }, + { + "epoch": 4.69, + "learning_rate": 3.45679012345679e-06, + "loss": 0.0, + "step": 2955 + }, + { + "epoch": 4.69, + "learning_rate": 3.439153439153439e-06, + "loss": 0.0, + "step": 2956 + }, + { + "epoch": 4.69, + "learning_rate": 3.4215167548500884e-06, + "loss": 0.0, + "step": 2957 + }, + { + "epoch": 4.7, + "learning_rate": 3.4038800705467374e-06, + "loss": 0.0, + "step": 2958 + }, + { + "epoch": 4.7, + "learning_rate": 3.3862433862433868e-06, + "loss": 0.0, + "step": 2959 + }, + { + "epoch": 4.7, + "learning_rate": 3.368606701940035e-06, + "loss": 0.0697, + "step": 2960 + }, + { + "epoch": 4.7, + "learning_rate": 3.3509700176366843e-06, + "loss": 0.0, + "step": 2961 + }, + { + "epoch": 4.7, + "learning_rate": 3.3333333333333333e-06, + "loss": 0.0, + "step": 2962 + }, + { + "epoch": 4.7, + "learning_rate": 3.3156966490299827e-06, + "loss": 0.0, + "step": 2963 + }, + { + "epoch": 4.7, + "learning_rate": 3.2980599647266317e-06, + "loss": 0.0, + "step": 2964 + }, + { + "epoch": 4.71, + "learning_rate": 3.2804232804232803e-06, + "loss": 0.0, + "step": 2965 + }, + { + "epoch": 4.71, + "learning_rate": 3.2627865961199293e-06, + "loss": 0.0, + "step": 2966 + }, + { + "epoch": 4.71, + "learning_rate": 3.2451499118165787e-06, + "loss": 0.0, + "step": 2967 + }, + { + "epoch": 4.71, + "learning_rate": 3.2275132275132277e-06, + "loss": 0.0, + "step": 2968 + }, + { + "epoch": 4.71, + "learning_rate": 3.209876543209877e-06, + "loss": 0.0, + "step": 2969 + }, + { + "epoch": 4.71, + "learning_rate": 3.1922398589065252e-06, + "loss": 0.0, + "step": 2970 + }, + { + "epoch": 4.72, + "learning_rate": 3.1746031746031746e-06, + "loss": 0.0, + "step": 2971 + }, + { + "epoch": 4.72, + "learning_rate": 3.1569664902998236e-06, + "loss": 0.0, + "step": 2972 + }, + { + "epoch": 4.72, + "learning_rate": 3.139329805996473e-06, + "loss": 0.0, + "step": 2973 + }, + { + "epoch": 4.72, + "learning_rate": 3.1216931216931216e-06, + "loss": 0.0, + "step": 2974 + }, + { + "epoch": 4.72, + "learning_rate": 3.104056437389771e-06, + "loss": 0.0, + "step": 2975 + }, + { + "epoch": 4.72, + "learning_rate": 3.0864197530864196e-06, + "loss": 0.0, + "step": 2976 + }, + { + "epoch": 4.73, + "learning_rate": 3.068783068783069e-06, + "loss": 0.0, + "step": 2977 + }, + { + "epoch": 4.73, + "learning_rate": 3.051146384479718e-06, + "loss": 0.0017, + "step": 2978 + }, + { + "epoch": 4.73, + "learning_rate": 3.033509700176367e-06, + "loss": 0.0, + "step": 2979 + }, + { + "epoch": 4.73, + "learning_rate": 3.015873015873016e-06, + "loss": 0.0, + "step": 2980 + }, + { + "epoch": 4.73, + "learning_rate": 2.998236331569665e-06, + "loss": 0.0, + "step": 2981 + }, + { + "epoch": 4.73, + "learning_rate": 2.980599647266314e-06, + "loss": 0.0, + "step": 2982 + }, + { + "epoch": 4.73, + "learning_rate": 2.9629629629629633e-06, + "loss": 0.0, + "step": 2983 + }, + { + "epoch": 4.74, + "learning_rate": 2.945326278659612e-06, + "loss": 0.0, + "step": 2984 + }, + { + "epoch": 4.74, + "learning_rate": 2.9276895943562613e-06, + "loss": 0.0, + "step": 2985 + }, + { + "epoch": 4.74, + "learning_rate": 2.91005291005291e-06, + "loss": 0.0, + "step": 2986 + }, + { + "epoch": 4.74, + "learning_rate": 2.8924162257495593e-06, + "loss": 0.0, + "step": 2987 + }, + { + "epoch": 4.74, + "learning_rate": 2.8747795414462083e-06, + "loss": 0.0, + "step": 2988 + }, + { + "epoch": 4.74, + "learning_rate": 2.8571428571428573e-06, + "loss": 0.0, + "step": 2989 + }, + { + "epoch": 4.75, + "learning_rate": 2.8395061728395062e-06, + "loss": 0.0, + "step": 2990 + }, + { + "epoch": 4.75, + "learning_rate": 2.8218694885361552e-06, + "loss": 0.0, + "step": 2991 + }, + { + "epoch": 4.75, + "learning_rate": 2.8042328042328042e-06, + "loss": 0.0, + "step": 2992 + }, + { + "epoch": 4.75, + "learning_rate": 2.7865961199294536e-06, + "loss": 0.0, + "step": 2993 + }, + { + "epoch": 4.75, + "learning_rate": 2.768959435626102e-06, + "loss": 0.0, + "step": 2994 + }, + { + "epoch": 4.75, + "learning_rate": 2.7513227513227516e-06, + "loss": 0.0, + "step": 2995 + }, + { + "epoch": 4.76, + "learning_rate": 2.7336860670194e-06, + "loss": 0.0, + "step": 2996 + }, + { + "epoch": 4.76, + "learning_rate": 2.7160493827160496e-06, + "loss": 0.0, + "step": 2997 + }, + { + "epoch": 4.76, + "learning_rate": 2.6984126984126986e-06, + "loss": 0.0, + "step": 2998 + }, + { + "epoch": 4.76, + "learning_rate": 2.6807760141093476e-06, + "loss": 0.0, + "step": 2999 + }, + { + "epoch": 4.76, + "learning_rate": 2.6631393298059965e-06, + "loss": 0.0, + "step": 3000 + }, + { + "epoch": 4.76, + "learning_rate": 2.6455026455026455e-06, + "loss": 0.0, + "step": 3001 + }, + { + "epoch": 4.77, + "learning_rate": 2.6278659611992945e-06, + "loss": 0.0, + "step": 3002 + }, + { + "epoch": 4.77, + "learning_rate": 2.610229276895944e-06, + "loss": 0.0, + "step": 3003 + }, + { + "epoch": 4.77, + "learning_rate": 2.5925925925925925e-06, + "loss": 0.0, + "step": 3004 + }, + { + "epoch": 4.77, + "learning_rate": 2.574955908289242e-06, + "loss": 0.0, + "step": 3005 + }, + { + "epoch": 4.77, + "learning_rate": 2.5573192239858905e-06, + "loss": 0.0, + "step": 3006 + }, + { + "epoch": 4.77, + "learning_rate": 2.53968253968254e-06, + "loss": 0.0, + "step": 3007 + }, + { + "epoch": 4.77, + "learning_rate": 2.522045855379189e-06, + "loss": 0.0, + "step": 3008 + }, + { + "epoch": 4.78, + "learning_rate": 2.504409171075838e-06, + "loss": 0.0, + "step": 3009 + }, + { + "epoch": 4.78, + "learning_rate": 2.486772486772487e-06, + "loss": 0.0, + "step": 3010 + }, + { + "epoch": 4.78, + "learning_rate": 2.469135802469136e-06, + "loss": 0.0, + "step": 3011 + }, + { + "epoch": 4.78, + "learning_rate": 2.451499118165785e-06, + "loss": 0.0, + "step": 3012 + }, + { + "epoch": 4.78, + "learning_rate": 2.4338624338624342e-06, + "loss": 0.0, + "step": 3013 + }, + { + "epoch": 4.78, + "learning_rate": 2.416225749559083e-06, + "loss": 0.0, + "step": 3014 + }, + { + "epoch": 4.79, + "learning_rate": 2.398589065255732e-06, + "loss": 0.0, + "step": 3015 + }, + { + "epoch": 4.79, + "learning_rate": 2.3809523809523808e-06, + "loss": 0.0, + "step": 3016 + }, + { + "epoch": 4.79, + "learning_rate": 2.36331569664903e-06, + "loss": 0.0, + "step": 3017 + }, + { + "epoch": 4.79, + "learning_rate": 2.345679012345679e-06, + "loss": 0.0, + "step": 3018 + }, + { + "epoch": 4.79, + "learning_rate": 2.328042328042328e-06, + "loss": 0.0, + "step": 3019 + }, + { + "epoch": 4.79, + "learning_rate": 2.310405643738977e-06, + "loss": 0.0, + "step": 3020 + }, + { + "epoch": 4.8, + "learning_rate": 2.292768959435626e-06, + "loss": 0.0, + "step": 3021 + }, + { + "epoch": 4.8, + "learning_rate": 2.275132275132275e-06, + "loss": 0.0, + "step": 3022 + }, + { + "epoch": 4.8, + "learning_rate": 2.2574955908289245e-06, + "loss": 0.0, + "step": 3023 + }, + { + "epoch": 4.8, + "learning_rate": 2.239858906525573e-06, + "loss": 0.0, + "step": 3024 + }, + { + "epoch": 4.8, + "learning_rate": 2.2222222222222225e-06, + "loss": 0.0, + "step": 3025 + }, + { + "epoch": 4.8, + "learning_rate": 2.204585537918871e-06, + "loss": 0.0, + "step": 3026 + }, + { + "epoch": 4.8, + "learning_rate": 2.1869488536155205e-06, + "loss": 0.0, + "step": 3027 + }, + { + "epoch": 4.81, + "learning_rate": 2.1693121693121695e-06, + "loss": 0.0, + "step": 3028 + }, + { + "epoch": 4.81, + "learning_rate": 2.1516754850088184e-06, + "loss": 0.0, + "step": 3029 + }, + { + "epoch": 4.81, + "learning_rate": 2.1340388007054674e-06, + "loss": 0.0, + "step": 3030 + }, + { + "epoch": 4.81, + "learning_rate": 2.1164021164021164e-06, + "loss": 0.0, + "step": 3031 + }, + { + "epoch": 4.81, + "learning_rate": 2.0987654320987654e-06, + "loss": 0.0, + "step": 3032 + }, + { + "epoch": 4.81, + "learning_rate": 2.081128747795415e-06, + "loss": 0.0, + "step": 3033 + }, + { + "epoch": 4.82, + "learning_rate": 2.0634920634920634e-06, + "loss": 0.0, + "step": 3034 + }, + { + "epoch": 4.82, + "learning_rate": 2.045855379188713e-06, + "loss": 0.0, + "step": 3035 + }, + { + "epoch": 4.82, + "learning_rate": 2.0282186948853614e-06, + "loss": 0.0, + "step": 3036 + }, + { + "epoch": 4.82, + "learning_rate": 2.0105820105820108e-06, + "loss": 0.0, + "step": 3037 + }, + { + "epoch": 4.82, + "learning_rate": 1.9929453262786598e-06, + "loss": 0.0, + "step": 3038 + }, + { + "epoch": 4.82, + "learning_rate": 1.9753086419753087e-06, + "loss": 0.0, + "step": 3039 + }, + { + "epoch": 4.83, + "learning_rate": 1.9576719576719577e-06, + "loss": 0.0, + "step": 3040 + }, + { + "epoch": 4.83, + "learning_rate": 1.9400352733686067e-06, + "loss": 0.0, + "step": 3041 + }, + { + "epoch": 4.83, + "learning_rate": 1.9223985890652557e-06, + "loss": 0.0, + "step": 3042 + }, + { + "epoch": 4.83, + "learning_rate": 1.9047619047619051e-06, + "loss": 0.001, + "step": 3043 + }, + { + "epoch": 4.83, + "learning_rate": 1.887125220458554e-06, + "loss": 0.0, + "step": 3044 + }, + { + "epoch": 4.83, + "learning_rate": 1.869488536155203e-06, + "loss": 0.0, + "step": 3045 + }, + { + "epoch": 4.83, + "learning_rate": 1.8518518518518519e-06, + "loss": 0.0, + "step": 3046 + }, + { + "epoch": 4.84, + "learning_rate": 1.834215167548501e-06, + "loss": 0.0, + "step": 3047 + }, + { + "epoch": 4.84, + "learning_rate": 1.8165784832451498e-06, + "loss": 0.0, + "step": 3048 + }, + { + "epoch": 4.84, + "learning_rate": 1.798941798941799e-06, + "loss": 0.0, + "step": 3049 + }, + { + "epoch": 4.84, + "learning_rate": 1.7813051146384482e-06, + "loss": 0.0, + "step": 3050 + }, + { + "epoch": 4.84, + "learning_rate": 1.763668430335097e-06, + "loss": 0.0, + "step": 3051 + }, + { + "epoch": 4.84, + "learning_rate": 1.7460317460317462e-06, + "loss": 0.0, + "step": 3052 + }, + { + "epoch": 4.85, + "learning_rate": 1.728395061728395e-06, + "loss": 0.0, + "step": 3053 + }, + { + "epoch": 4.85, + "learning_rate": 1.7107583774250442e-06, + "loss": 0.0, + "step": 3054 + }, + { + "epoch": 4.85, + "learning_rate": 1.6931216931216934e-06, + "loss": 0.0, + "step": 3055 + }, + { + "epoch": 4.85, + "learning_rate": 1.6754850088183422e-06, + "loss": 0.0, + "step": 3056 + }, + { + "epoch": 4.85, + "learning_rate": 1.6578483245149914e-06, + "loss": 0.0, + "step": 3057 + }, + { + "epoch": 4.85, + "learning_rate": 1.6402116402116401e-06, + "loss": 0.0, + "step": 3058 + }, + { + "epoch": 4.86, + "learning_rate": 1.6225749559082893e-06, + "loss": 0.0, + "step": 3059 + }, + { + "epoch": 4.86, + "learning_rate": 1.6049382716049385e-06, + "loss": 0.0, + "step": 3060 + }, + { + "epoch": 4.86, + "learning_rate": 1.5873015873015873e-06, + "loss": 0.0, + "step": 3061 + }, + { + "epoch": 4.86, + "learning_rate": 1.5696649029982365e-06, + "loss": 0.0, + "step": 3062 + }, + { + "epoch": 4.86, + "learning_rate": 1.5520282186948855e-06, + "loss": 0.0, + "step": 3063 + }, + { + "epoch": 4.86, + "learning_rate": 1.5343915343915345e-06, + "loss": 0.0, + "step": 3064 + }, + { + "epoch": 4.87, + "learning_rate": 1.5167548500881835e-06, + "loss": 0.0, + "step": 3065 + }, + { + "epoch": 4.87, + "learning_rate": 1.4991181657848325e-06, + "loss": 0.0, + "step": 3066 + }, + { + "epoch": 4.87, + "learning_rate": 1.4814814814814817e-06, + "loss": 0.0, + "step": 3067 + }, + { + "epoch": 4.87, + "learning_rate": 1.4638447971781307e-06, + "loss": 0.0, + "step": 3068 + }, + { + "epoch": 4.87, + "learning_rate": 1.4462081128747796e-06, + "loss": 0.0, + "step": 3069 + }, + { + "epoch": 4.87, + "learning_rate": 1.4285714285714286e-06, + "loss": 0.0, + "step": 3070 + }, + { + "epoch": 4.87, + "learning_rate": 1.4109347442680776e-06, + "loss": 0.0, + "step": 3071 + }, + { + "epoch": 4.88, + "learning_rate": 1.3932980599647268e-06, + "loss": 0.0, + "step": 3072 + }, + { + "epoch": 4.88, + "learning_rate": 1.3756613756613758e-06, + "loss": 0.0, + "step": 3073 + }, + { + "epoch": 4.88, + "learning_rate": 1.3580246913580248e-06, + "loss": 0.0, + "step": 3074 + }, + { + "epoch": 4.88, + "learning_rate": 1.3403880070546738e-06, + "loss": 0.0, + "step": 3075 + }, + { + "epoch": 4.88, + "learning_rate": 1.3227513227513228e-06, + "loss": 0.0, + "step": 3076 + }, + { + "epoch": 4.88, + "learning_rate": 1.305114638447972e-06, + "loss": 0.0, + "step": 3077 + }, + { + "epoch": 4.89, + "learning_rate": 1.287477954144621e-06, + "loss": 0.0, + "step": 3078 + }, + { + "epoch": 4.89, + "learning_rate": 1.26984126984127e-06, + "loss": 0.0, + "step": 3079 + }, + { + "epoch": 4.89, + "learning_rate": 1.252204585537919e-06, + "loss": 0.0, + "step": 3080 + }, + { + "epoch": 4.89, + "learning_rate": 1.234567901234568e-06, + "loss": 0.0, + "step": 3081 + }, + { + "epoch": 4.89, + "learning_rate": 1.2169312169312171e-06, + "loss": 0.0, + "step": 3082 + }, + { + "epoch": 4.89, + "learning_rate": 1.199294532627866e-06, + "loss": 0.0, + "step": 3083 + }, + { + "epoch": 4.9, + "learning_rate": 1.181657848324515e-06, + "loss": 0.0, + "step": 3084 + }, + { + "epoch": 4.9, + "learning_rate": 1.164021164021164e-06, + "loss": 0.0, + "step": 3085 + }, + { + "epoch": 4.9, + "learning_rate": 1.146384479717813e-06, + "loss": 0.0, + "step": 3086 + }, + { + "epoch": 4.9, + "learning_rate": 1.1287477954144623e-06, + "loss": 0.0, + "step": 3087 + }, + { + "epoch": 4.9, + "learning_rate": 1.1111111111111112e-06, + "loss": 0.0, + "step": 3088 + }, + { + "epoch": 4.9, + "learning_rate": 1.0934744268077602e-06, + "loss": 0.0, + "step": 3089 + }, + { + "epoch": 4.9, + "learning_rate": 1.0758377425044092e-06, + "loss": 0.0, + "step": 3090 + }, + { + "epoch": 4.91, + "learning_rate": 1.0582010582010582e-06, + "loss": 0.0, + "step": 3091 + }, + { + "epoch": 4.91, + "learning_rate": 1.0405643738977074e-06, + "loss": 0.0, + "step": 3092 + }, + { + "epoch": 4.91, + "learning_rate": 1.0229276895943564e-06, + "loss": 0.0, + "step": 3093 + }, + { + "epoch": 4.91, + "learning_rate": 1.0052910052910054e-06, + "loss": 0.0, + "step": 3094 + }, + { + "epoch": 4.91, + "learning_rate": 9.876543209876544e-07, + "loss": 0.0, + "step": 3095 + }, + { + "epoch": 4.91, + "learning_rate": 9.700176366843034e-07, + "loss": 0.0, + "step": 3096 + }, + { + "epoch": 4.92, + "learning_rate": 9.523809523809526e-07, + "loss": 0.0, + "step": 3097 + }, + { + "epoch": 4.92, + "learning_rate": 9.347442680776015e-07, + "loss": 0.0, + "step": 3098 + }, + { + "epoch": 4.92, + "learning_rate": 9.171075837742505e-07, + "loss": 0.0, + "step": 3099 + }, + { + "epoch": 4.92, + "learning_rate": 8.994708994708995e-07, + "loss": 0.0, + "step": 3100 + }, + { + "epoch": 4.92, + "learning_rate": 8.818342151675485e-07, + "loss": 0.0, + "step": 3101 + }, + { + "epoch": 4.92, + "learning_rate": 8.641975308641975e-07, + "loss": 0.0, + "step": 3102 + }, + { + "epoch": 4.93, + "learning_rate": 8.465608465608467e-07, + "loss": 0.0, + "step": 3103 + }, + { + "epoch": 4.93, + "learning_rate": 8.289241622574957e-07, + "loss": 0.0, + "step": 3104 + }, + { + "epoch": 4.93, + "learning_rate": 8.112874779541447e-07, + "loss": 0.0, + "step": 3105 + }, + { + "epoch": 4.93, + "learning_rate": 7.936507936507937e-07, + "loss": 0.0, + "step": 3106 + }, + { + "epoch": 4.93, + "learning_rate": 7.760141093474428e-07, + "loss": 0.0, + "step": 3107 + }, + { + "epoch": 4.93, + "learning_rate": 7.583774250440917e-07, + "loss": 0.0, + "step": 3108 + }, + { + "epoch": 4.93, + "learning_rate": 7.407407407407408e-07, + "loss": 0.0, + "step": 3109 + }, + { + "epoch": 4.94, + "learning_rate": 7.231040564373898e-07, + "loss": 0.0, + "step": 3110 + }, + { + "epoch": 4.94, + "learning_rate": 7.054673721340388e-07, + "loss": 0.0, + "step": 3111 + }, + { + "epoch": 4.94, + "learning_rate": 6.878306878306879e-07, + "loss": 0.0, + "step": 3112 + }, + { + "epoch": 4.94, + "learning_rate": 6.701940035273369e-07, + "loss": 0.0, + "step": 3113 + }, + { + "epoch": 4.94, + "learning_rate": 6.52557319223986e-07, + "loss": 0.0, + "step": 3114 + }, + { + "epoch": 4.94, + "learning_rate": 6.34920634920635e-07, + "loss": 0.0, + "step": 3115 + }, + { + "epoch": 4.95, + "learning_rate": 6.17283950617284e-07, + "loss": 0.0, + "step": 3116 + }, + { + "epoch": 4.95, + "learning_rate": 5.99647266313933e-07, + "loss": 0.0, + "step": 3117 + }, + { + "epoch": 4.95, + "learning_rate": 5.82010582010582e-07, + "loss": 0.0, + "step": 3118 + }, + { + "epoch": 4.95, + "learning_rate": 5.643738977072311e-07, + "loss": 0.0, + "step": 3119 + }, + { + "epoch": 4.95, + "learning_rate": 5.467372134038801e-07, + "loss": 0.0, + "step": 3120 + }, + { + "epoch": 4.95, + "learning_rate": 5.291005291005291e-07, + "loss": 0.0, + "step": 3121 + }, + { + "epoch": 4.96, + "learning_rate": 5.114638447971782e-07, + "loss": 0.0, + "step": 3122 + }, + { + "epoch": 4.96, + "learning_rate": 4.938271604938272e-07, + "loss": 0.0, + "step": 3123 + }, + { + "epoch": 4.96, + "learning_rate": 4.761904761904763e-07, + "loss": 0.0, + "step": 3124 + }, + { + "epoch": 4.96, + "learning_rate": 4.5855379188712527e-07, + "loss": 0.0, + "step": 3125 + }, + { + "epoch": 4.96, + "learning_rate": 4.4091710758377425e-07, + "loss": 0.0, + "step": 3126 + }, + { + "epoch": 4.96, + "learning_rate": 4.2328042328042335e-07, + "loss": 0.0, + "step": 3127 + }, + { + "epoch": 4.97, + "learning_rate": 4.0564373897707234e-07, + "loss": 0.0, + "step": 3128 + }, + { + "epoch": 4.97, + "learning_rate": 3.880070546737214e-07, + "loss": 0.0, + "step": 3129 + }, + { + "epoch": 4.97, + "learning_rate": 3.703703703703704e-07, + "loss": 0.0, + "step": 3130 + }, + { + "epoch": 4.97, + "learning_rate": 3.527336860670194e-07, + "loss": 0.0, + "step": 3131 + }, + { + "epoch": 4.97, + "learning_rate": 3.3509700176366844e-07, + "loss": 0.0, + "step": 3132 + }, + { + "epoch": 4.97, + "learning_rate": 3.174603174603175e-07, + "loss": 0.0, + "step": 3133 + }, + { + "epoch": 4.97, + "learning_rate": 2.998236331569665e-07, + "loss": 0.0, + "step": 3134 + }, + { + "epoch": 4.98, + "learning_rate": 2.8218694885361557e-07, + "loss": 0.0, + "step": 3135 + }, + { + "epoch": 4.98, + "learning_rate": 2.6455026455026455e-07, + "loss": 0.0, + "step": 3136 + }, + { + "epoch": 4.98, + "learning_rate": 2.469135802469136e-07, + "loss": 0.0, + "step": 3137 + }, + { + "epoch": 4.98, + "learning_rate": 2.2927689594356263e-07, + "loss": 0.0, + "step": 3138 + }, + { + "epoch": 4.98, + "learning_rate": 2.1164021164021167e-07, + "loss": 0.0, + "step": 3139 + }, + { + "epoch": 4.98, + "learning_rate": 1.940035273368607e-07, + "loss": 0.0, + "step": 3140 + }, + { + "epoch": 4.99, + "learning_rate": 1.763668430335097e-07, + "loss": 0.0, + "step": 3141 + }, + { + "epoch": 4.99, + "learning_rate": 1.5873015873015874e-07, + "loss": 0.0, + "step": 3142 + }, + { + "epoch": 4.99, + "learning_rate": 1.4109347442680778e-07, + "loss": 0.0, + "step": 3143 + }, + { + "epoch": 4.99, + "learning_rate": 1.234567901234568e-07, + "loss": 0.0, + "step": 3144 + }, + { + "epoch": 4.99, + "learning_rate": 1.0582010582010584e-07, + "loss": 0.0, + "step": 3145 + }, + { + "epoch": 4.99, + "learning_rate": 8.818342151675485e-08, + "loss": 0.0, + "step": 3146 + }, + { + "epoch": 5.0, + "learning_rate": 7.054673721340389e-08, + "loss": 0.0, + "step": 3147 + }, + { + "epoch": 5.0, + "learning_rate": 5.291005291005292e-08, + "loss": 0.0, + "step": 3148 + }, + { + "epoch": 5.0, + "learning_rate": 3.5273368606701946e-08, + "loss": 0.0, + "step": 3149 + }, + { + "epoch": 5.0, + "learning_rate": 1.7636684303350973e-08, + "loss": 0.0, + "step": 3150 + }, + { + "epoch": 5.0, + "eval_accuracy": 1.0, + "eval_auc": 1.0, + "eval_f1": 1.0, + "eval_loss": 0.0019458403112366796, + "eval_precision": 1.0, + "eval_recall": 1.0, + "eval_runtime": 0.2025, + "eval_samples_per_second": 187.678, + "eval_steps_per_second": 4.939, + "step": 3150 + } + ], + "logging_steps": 1, + "max_steps": 3150, + "num_input_tokens_seen": 0, + "num_train_epochs": 5, + "save_steps": 500, + "total_flos": 2.64926521641984e+16, + "train_batch_size": 32, + "trial_name": null, + "trial_params": null +}