{ "best_metric": null, "best_model_checkpoint": null, "epoch": 25.0, "global_step": 22300, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 3.0000000000000004e-08, "loss": 10.9963, "step": 1 }, { "epoch": 0.0, "learning_rate": 6.000000000000001e-08, "loss": 10.9998, "step": 2 }, { "epoch": 0.0, "learning_rate": 9e-08, "loss": 11.0021, "step": 3 }, { "epoch": 0.0, "learning_rate": 1.2000000000000002e-07, "loss": 10.9631, "step": 4 }, { "epoch": 0.01, "learning_rate": 1.5000000000000002e-07, "loss": 10.9817, "step": 5 }, { "epoch": 0.01, "learning_rate": 1.8e-07, "loss": 10.9947, "step": 6 }, { "epoch": 0.01, "learning_rate": 2.1e-07, "loss": 10.9827, "step": 7 }, { "epoch": 0.01, "learning_rate": 2.4000000000000003e-07, "loss": 10.994, "step": 8 }, { "epoch": 0.01, "learning_rate": 2.7e-07, "loss": 10.9984, "step": 9 }, { "epoch": 0.01, "learning_rate": 3.0000000000000004e-07, "loss": 10.9665, "step": 10 }, { "epoch": 0.01, "learning_rate": 3.2999999999999996e-07, "loss": 10.9452, "step": 11 }, { "epoch": 0.01, "learning_rate": 3.6e-07, "loss": 10.959, "step": 12 }, { "epoch": 0.01, "learning_rate": 3.8999999999999997e-07, "loss": 10.9485, "step": 13 }, { "epoch": 0.02, "learning_rate": 4.2e-07, "loss": 10.9255, "step": 14 }, { "epoch": 0.02, "learning_rate": 4.5e-07, "loss": 10.9058, "step": 15 }, { "epoch": 0.02, "learning_rate": 4.800000000000001e-07, "loss": 10.8987, "step": 16 }, { "epoch": 0.02, "learning_rate": 5.100000000000001e-07, "loss": 10.897, "step": 17 }, { "epoch": 0.02, "learning_rate": 5.4e-07, "loss": 10.8976, "step": 18 }, { "epoch": 0.02, "learning_rate": 5.7e-07, "loss": 10.9104, "step": 19 }, { "epoch": 0.02, "learning_rate": 6.000000000000001e-07, "loss": 10.9183, "step": 20 }, { "epoch": 0.02, "learning_rate": 6.3e-07, "loss": 10.9019, "step": 21 }, { "epoch": 0.02, "learning_rate": 6.599999999999999e-07, "loss": 10.8414, "step": 22 }, { "epoch": 0.03, "learning_rate": 6.9e-07, "loss": 10.8347, "step": 23 }, { "epoch": 0.03, "learning_rate": 7.2e-07, "loss": 10.8418, "step": 24 }, { "epoch": 0.03, "learning_rate": 7.5e-07, "loss": 10.8078, "step": 25 }, { "epoch": 0.03, "learning_rate": 7.799999999999999e-07, "loss": 10.8402, "step": 26 }, { "epoch": 0.03, "learning_rate": 8.1e-07, "loss": 10.7863, "step": 27 }, { "epoch": 0.03, "learning_rate": 8.4e-07, "loss": 10.7765, "step": 28 }, { "epoch": 0.03, "learning_rate": 8.7e-07, "loss": 10.7305, "step": 29 }, { "epoch": 0.03, "learning_rate": 9e-07, "loss": 10.7609, "step": 30 }, { "epoch": 0.03, "learning_rate": 9.3e-07, "loss": 10.6934, "step": 31 }, { "epoch": 0.04, "learning_rate": 9.600000000000001e-07, "loss": 10.6615, "step": 32 }, { "epoch": 0.04, "learning_rate": 9.9e-07, "loss": 10.6698, "step": 33 }, { "epoch": 0.04, "learning_rate": 1.0200000000000002e-06, "loss": 10.636, "step": 34 }, { "epoch": 0.04, "learning_rate": 1.0500000000000001e-06, "loss": 10.6253, "step": 35 }, { "epoch": 0.04, "learning_rate": 1.08e-06, "loss": 10.6634, "step": 36 }, { "epoch": 0.04, "learning_rate": 1.11e-06, "loss": 10.5809, "step": 37 }, { "epoch": 0.04, "learning_rate": 1.14e-06, "loss": 10.5975, "step": 38 }, { "epoch": 0.04, "learning_rate": 1.17e-06, "loss": 10.5724, "step": 39 }, { "epoch": 0.04, "learning_rate": 1.2000000000000002e-06, "loss": 10.4578, "step": 40 }, { "epoch": 0.05, "learning_rate": 1.23e-06, "loss": 10.4741, "step": 41 }, { "epoch": 0.05, "learning_rate": 1.26e-06, "loss": 10.4915, "step": 42 }, { "epoch": 0.05, "learning_rate": 1.29e-06, "loss": 10.4854, "step": 43 }, { "epoch": 0.05, "learning_rate": 1.3199999999999999e-06, "loss": 10.4809, "step": 44 }, { "epoch": 0.05, "learning_rate": 1.35e-06, "loss": 10.3995, "step": 45 }, { "epoch": 0.05, "learning_rate": 1.38e-06, "loss": 10.4031, "step": 46 }, { "epoch": 0.05, "learning_rate": 1.41e-06, "loss": 10.4729, "step": 47 }, { "epoch": 0.05, "learning_rate": 1.44e-06, "loss": 10.4023, "step": 48 }, { "epoch": 0.05, "learning_rate": 1.4700000000000001e-06, "loss": 10.2837, "step": 49 }, { "epoch": 0.06, "learning_rate": 1.5e-06, "loss": 10.2657, "step": 50 }, { "epoch": 0.06, "learning_rate": 1.53e-06, "loss": 10.3796, "step": 51 }, { "epoch": 0.06, "learning_rate": 1.5599999999999999e-06, "loss": 10.2526, "step": 52 }, { "epoch": 0.06, "learning_rate": 1.59e-06, "loss": 10.2906, "step": 53 }, { "epoch": 0.06, "learning_rate": 1.62e-06, "loss": 10.3816, "step": 54 }, { "epoch": 0.06, "learning_rate": 1.65e-06, "loss": 10.2359, "step": 55 }, { "epoch": 0.06, "learning_rate": 1.68e-06, "loss": 10.2533, "step": 56 }, { "epoch": 0.06, "learning_rate": 1.7100000000000001e-06, "loss": 10.2105, "step": 57 }, { "epoch": 0.07, "learning_rate": 1.74e-06, "loss": 10.2375, "step": 58 }, { "epoch": 0.07, "learning_rate": 1.77e-06, "loss": 10.1644, "step": 59 }, { "epoch": 0.07, "learning_rate": 1.8e-06, "loss": 10.1392, "step": 60 }, { "epoch": 0.07, "learning_rate": 1.83e-06, "loss": 10.121, "step": 61 }, { "epoch": 0.07, "learning_rate": 1.86e-06, "loss": 10.1292, "step": 62 }, { "epoch": 0.07, "learning_rate": 1.8900000000000001e-06, "loss": 10.1207, "step": 63 }, { "epoch": 0.07, "learning_rate": 1.9200000000000003e-06, "loss": 10.1248, "step": 64 }, { "epoch": 0.07, "learning_rate": 1.95e-06, "loss": 10.0755, "step": 65 }, { "epoch": 0.07, "learning_rate": 1.98e-06, "loss": 10.165, "step": 66 }, { "epoch": 0.08, "learning_rate": 2.0100000000000002e-06, "loss": 10.0689, "step": 67 }, { "epoch": 0.08, "learning_rate": 2.0400000000000004e-06, "loss": 10.0401, "step": 68 }, { "epoch": 0.08, "learning_rate": 2.07e-06, "loss": 9.946, "step": 69 }, { "epoch": 0.08, "learning_rate": 2.1000000000000002e-06, "loss": 9.9455, "step": 70 }, { "epoch": 0.08, "learning_rate": 2.13e-06, "loss": 10.0096, "step": 71 }, { "epoch": 0.08, "learning_rate": 2.16e-06, "loss": 9.9436, "step": 72 }, { "epoch": 0.08, "learning_rate": 2.1899999999999998e-06, "loss": 9.929, "step": 73 }, { "epoch": 0.08, "learning_rate": 2.22e-06, "loss": 9.9471, "step": 74 }, { "epoch": 0.08, "learning_rate": 2.25e-06, "loss": 9.8279, "step": 75 }, { "epoch": 0.09, "learning_rate": 2.28e-06, "loss": 9.916, "step": 76 }, { "epoch": 0.09, "learning_rate": 2.31e-06, "loss": 9.8037, "step": 77 }, { "epoch": 0.09, "learning_rate": 2.34e-06, "loss": 9.7625, "step": 78 }, { "epoch": 0.09, "learning_rate": 2.37e-06, "loss": 9.7898, "step": 79 }, { "epoch": 0.09, "learning_rate": 2.4000000000000003e-06, "loss": 9.8063, "step": 80 }, { "epoch": 0.09, "learning_rate": 2.43e-06, "loss": 9.6767, "step": 81 }, { "epoch": 0.09, "learning_rate": 2.46e-06, "loss": 9.6849, "step": 82 }, { "epoch": 0.09, "learning_rate": 2.4900000000000003e-06, "loss": 9.7427, "step": 83 }, { "epoch": 0.09, "learning_rate": 2.52e-06, "loss": 9.6709, "step": 84 }, { "epoch": 0.1, "learning_rate": 2.55e-06, "loss": 9.7299, "step": 85 }, { "epoch": 0.1, "learning_rate": 2.58e-06, "loss": 9.6612, "step": 86 }, { "epoch": 0.1, "learning_rate": 2.61e-06, "loss": 9.6834, "step": 87 }, { "epoch": 0.1, "learning_rate": 2.6399999999999997e-06, "loss": 9.6318, "step": 88 }, { "epoch": 0.1, "learning_rate": 2.67e-06, "loss": 9.6744, "step": 89 }, { "epoch": 0.1, "learning_rate": 2.7e-06, "loss": 9.6683, "step": 90 }, { "epoch": 0.1, "learning_rate": 2.73e-06, "loss": 9.7545, "step": 91 }, { "epoch": 0.1, "learning_rate": 2.76e-06, "loss": 9.5666, "step": 92 }, { "epoch": 0.1, "learning_rate": 2.79e-06, "loss": 9.5076, "step": 93 }, { "epoch": 0.11, "learning_rate": 2.82e-06, "loss": 9.5778, "step": 94 }, { "epoch": 0.11, "learning_rate": 2.8500000000000002e-06, "loss": 9.6056, "step": 95 }, { "epoch": 0.11, "learning_rate": 2.88e-06, "loss": 9.5169, "step": 96 }, { "epoch": 0.11, "learning_rate": 2.91e-06, "loss": 9.5374, "step": 97 }, { "epoch": 0.11, "learning_rate": 2.9400000000000002e-06, "loss": 9.5078, "step": 98 }, { "epoch": 0.11, "learning_rate": 2.9700000000000004e-06, "loss": 9.3497, "step": 99 }, { "epoch": 0.11, "learning_rate": 3e-06, "loss": 9.23, "step": 100 }, { "epoch": 0.11, "learning_rate": 3.0300000000000002e-06, "loss": 9.5742, "step": 101 }, { "epoch": 0.11, "learning_rate": 3.06e-06, "loss": 9.6004, "step": 102 }, { "epoch": 0.12, "learning_rate": 3.09e-06, "loss": 9.5768, "step": 103 }, { "epoch": 0.12, "learning_rate": 3.1199999999999998e-06, "loss": 9.5213, "step": 104 }, { "epoch": 0.12, "learning_rate": 3.15e-06, "loss": 9.4511, "step": 105 }, { "epoch": 0.12, "learning_rate": 3.18e-06, "loss": 9.5211, "step": 106 }, { "epoch": 0.12, "learning_rate": 3.21e-06, "loss": 9.4725, "step": 107 }, { "epoch": 0.12, "learning_rate": 3.24e-06, "loss": 9.4941, "step": 108 }, { "epoch": 0.12, "learning_rate": 3.27e-06, "loss": 9.4422, "step": 109 }, { "epoch": 0.12, "learning_rate": 3.3e-06, "loss": 9.526, "step": 110 }, { "epoch": 0.12, "learning_rate": 3.3300000000000003e-06, "loss": 9.4651, "step": 111 }, { "epoch": 0.13, "learning_rate": 3.36e-06, "loss": 9.3618, "step": 112 }, { "epoch": 0.13, "learning_rate": 3.39e-06, "loss": 9.3668, "step": 113 }, { "epoch": 0.13, "learning_rate": 3.4200000000000003e-06, "loss": 9.4078, "step": 114 }, { "epoch": 0.13, "learning_rate": 3.4500000000000004e-06, "loss": 9.3529, "step": 115 }, { "epoch": 0.13, "learning_rate": 3.48e-06, "loss": 9.2959, "step": 116 }, { "epoch": 0.13, "learning_rate": 3.5100000000000003e-06, "loss": 9.4459, "step": 117 }, { "epoch": 0.13, "learning_rate": 3.54e-06, "loss": 9.4078, "step": 118 }, { "epoch": 0.13, "learning_rate": 3.57e-06, "loss": 9.3646, "step": 119 }, { "epoch": 0.13, "learning_rate": 3.6e-06, "loss": 9.4291, "step": 120 }, { "epoch": 0.14, "learning_rate": 3.63e-06, "loss": 9.308, "step": 121 }, { "epoch": 0.14, "learning_rate": 3.66e-06, "loss": 9.4276, "step": 122 }, { "epoch": 0.14, "learning_rate": 3.6900000000000002e-06, "loss": 9.2579, "step": 123 }, { "epoch": 0.14, "learning_rate": 3.72e-06, "loss": 9.282, "step": 124 }, { "epoch": 0.14, "learning_rate": 3.75e-06, "loss": 9.3681, "step": 125 }, { "epoch": 0.14, "learning_rate": 3.7800000000000002e-06, "loss": 9.3774, "step": 126 }, { "epoch": 0.14, "learning_rate": 3.81e-06, "loss": 9.4799, "step": 127 }, { "epoch": 0.14, "learning_rate": 3.8400000000000005e-06, "loss": 9.43, "step": 128 }, { "epoch": 0.14, "learning_rate": 3.87e-06, "loss": 9.3484, "step": 129 }, { "epoch": 0.15, "learning_rate": 3.9e-06, "loss": 9.2755, "step": 130 }, { "epoch": 0.15, "learning_rate": 3.9300000000000005e-06, "loss": 9.2778, "step": 131 }, { "epoch": 0.15, "learning_rate": 3.96e-06, "loss": 9.3403, "step": 132 }, { "epoch": 0.15, "learning_rate": 3.99e-06, "loss": 9.2517, "step": 133 }, { "epoch": 0.15, "learning_rate": 4.0200000000000005e-06, "loss": 9.2861, "step": 134 }, { "epoch": 0.15, "learning_rate": 4.05e-06, "loss": 9.2408, "step": 135 }, { "epoch": 0.15, "learning_rate": 4.080000000000001e-06, "loss": 9.1979, "step": 136 }, { "epoch": 0.15, "learning_rate": 4.1100000000000005e-06, "loss": 9.2516, "step": 137 }, { "epoch": 0.15, "learning_rate": 4.14e-06, "loss": 9.154, "step": 138 }, { "epoch": 0.16, "learning_rate": 4.170000000000001e-06, "loss": 9.2063, "step": 139 }, { "epoch": 0.16, "learning_rate": 4.2000000000000004e-06, "loss": 9.197, "step": 140 }, { "epoch": 0.16, "learning_rate": 4.229999999999999e-06, "loss": 9.2896, "step": 141 }, { "epoch": 0.16, "learning_rate": 4.26e-06, "loss": 9.1635, "step": 142 }, { "epoch": 0.16, "learning_rate": 4.29e-06, "loss": 9.1342, "step": 143 }, { "epoch": 0.16, "learning_rate": 4.32e-06, "loss": 9.1704, "step": 144 }, { "epoch": 0.16, "learning_rate": 4.35e-06, "loss": 9.0791, "step": 145 }, { "epoch": 0.16, "learning_rate": 4.3799999999999996e-06, "loss": 9.0771, "step": 146 }, { "epoch": 0.16, "learning_rate": 4.41e-06, "loss": 9.1866, "step": 147 }, { "epoch": 0.17, "learning_rate": 4.44e-06, "loss": 8.8435, "step": 148 }, { "epoch": 0.17, "learning_rate": 4.4699999999999996e-06, "loss": 8.9952, "step": 149 }, { "epoch": 0.17, "learning_rate": 4.5e-06, "loss": 8.8216, "step": 150 }, { "epoch": 0.17, "learning_rate": 4.53e-06, "loss": 9.2337, "step": 151 }, { "epoch": 0.17, "learning_rate": 4.56e-06, "loss": 9.12, "step": 152 }, { "epoch": 0.17, "learning_rate": 4.59e-06, "loss": 9.1796, "step": 153 }, { "epoch": 0.17, "learning_rate": 4.62e-06, "loss": 9.1335, "step": 154 }, { "epoch": 0.17, "learning_rate": 4.65e-06, "loss": 9.1247, "step": 155 }, { "epoch": 0.17, "learning_rate": 4.68e-06, "loss": 9.1277, "step": 156 }, { "epoch": 0.18, "learning_rate": 4.71e-06, "loss": 9.2079, "step": 157 }, { "epoch": 0.18, "learning_rate": 4.74e-06, "loss": 9.1566, "step": 158 }, { "epoch": 0.18, "learning_rate": 4.77e-06, "loss": 9.1055, "step": 159 }, { "epoch": 0.18, "learning_rate": 4.800000000000001e-06, "loss": 9.1078, "step": 160 }, { "epoch": 0.18, "learning_rate": 4.83e-06, "loss": 8.9972, "step": 161 }, { "epoch": 0.18, "learning_rate": 4.86e-06, "loss": 9.076, "step": 162 }, { "epoch": 0.18, "learning_rate": 4.890000000000001e-06, "loss": 9.0701, "step": 163 }, { "epoch": 0.18, "learning_rate": 4.92e-06, "loss": 9.0787, "step": 164 }, { "epoch": 0.18, "learning_rate": 4.95e-06, "loss": 9.0121, "step": 165 }, { "epoch": 0.19, "learning_rate": 4.980000000000001e-06, "loss": 9.0557, "step": 166 }, { "epoch": 0.19, "learning_rate": 5.01e-06, "loss": 9.0279, "step": 167 }, { "epoch": 0.19, "learning_rate": 5.04e-06, "loss": 9.0231, "step": 168 }, { "epoch": 0.19, "learning_rate": 5.070000000000001e-06, "loss": 9.0048, "step": 169 }, { "epoch": 0.19, "learning_rate": 5.1e-06, "loss": 8.9363, "step": 170 }, { "epoch": 0.19, "learning_rate": 5.130000000000001e-06, "loss": 9.0671, "step": 171 }, { "epoch": 0.19, "learning_rate": 5.16e-06, "loss": 9.0105, "step": 172 }, { "epoch": 0.19, "learning_rate": 5.1899999999999994e-06, "loss": 8.9963, "step": 173 }, { "epoch": 0.2, "learning_rate": 5.22e-06, "loss": 8.9972, "step": 174 }, { "epoch": 0.2, "learning_rate": 5.25e-06, "loss": 8.8935, "step": 175 }, { "epoch": 0.2, "learning_rate": 5.279999999999999e-06, "loss": 8.8279, "step": 176 }, { "epoch": 0.2, "learning_rate": 5.31e-06, "loss": 8.8836, "step": 177 }, { "epoch": 0.2, "learning_rate": 5.34e-06, "loss": 8.9315, "step": 178 }, { "epoch": 0.2, "learning_rate": 5.37e-06, "loss": 8.99, "step": 179 }, { "epoch": 0.2, "learning_rate": 5.4e-06, "loss": 8.9123, "step": 180 }, { "epoch": 0.2, "learning_rate": 5.43e-06, "loss": 8.9728, "step": 181 }, { "epoch": 0.2, "learning_rate": 5.46e-06, "loss": 8.8425, "step": 182 }, { "epoch": 0.21, "learning_rate": 5.49e-06, "loss": 8.8566, "step": 183 }, { "epoch": 0.21, "learning_rate": 5.52e-06, "loss": 8.9403, "step": 184 }, { "epoch": 0.21, "learning_rate": 5.55e-06, "loss": 8.8112, "step": 185 }, { "epoch": 0.21, "learning_rate": 5.58e-06, "loss": 8.9323, "step": 186 }, { "epoch": 0.21, "learning_rate": 5.6100000000000005e-06, "loss": 8.9165, "step": 187 }, { "epoch": 0.21, "learning_rate": 5.64e-06, "loss": 8.8138, "step": 188 }, { "epoch": 0.21, "learning_rate": 5.67e-06, "loss": 8.8385, "step": 189 }, { "epoch": 0.21, "learning_rate": 5.7000000000000005e-06, "loss": 8.8747, "step": 190 }, { "epoch": 0.21, "learning_rate": 5.73e-06, "loss": 8.8553, "step": 191 }, { "epoch": 0.22, "learning_rate": 5.76e-06, "loss": 8.9839, "step": 192 }, { "epoch": 0.22, "learning_rate": 5.7900000000000005e-06, "loss": 8.7731, "step": 193 }, { "epoch": 0.22, "learning_rate": 5.82e-06, "loss": 8.86, "step": 194 }, { "epoch": 0.22, "learning_rate": 5.850000000000001e-06, "loss": 8.6921, "step": 195 }, { "epoch": 0.22, "learning_rate": 5.8800000000000005e-06, "loss": 8.8648, "step": 196 }, { "epoch": 0.22, "learning_rate": 5.91e-06, "loss": 8.7083, "step": 197 }, { "epoch": 0.22, "learning_rate": 5.940000000000001e-06, "loss": 8.6784, "step": 198 }, { "epoch": 0.22, "learning_rate": 5.9700000000000004e-06, "loss": 8.4155, "step": 199 }, { "epoch": 0.22, "learning_rate": 6e-06, "loss": 8.5263, "step": 200 }, { "epoch": 0.23, "learning_rate": 6.030000000000001e-06, "loss": 8.812, "step": 201 }, { "epoch": 0.23, "learning_rate": 6.0600000000000004e-06, "loss": 8.8858, "step": 202 }, { "epoch": 0.23, "learning_rate": 6.090000000000001e-06, "loss": 8.763, "step": 203 }, { "epoch": 0.23, "learning_rate": 6.12e-06, "loss": 8.7127, "step": 204 }, { "epoch": 0.23, "learning_rate": 6.1499999999999996e-06, "loss": 8.7173, "step": 205 }, { "epoch": 0.23, "learning_rate": 6.18e-06, "loss": 8.7562, "step": 206 }, { "epoch": 0.23, "learning_rate": 6.21e-06, "loss": 8.7974, "step": 207 }, { "epoch": 0.23, "learning_rate": 6.2399999999999995e-06, "loss": 8.6521, "step": 208 }, { "epoch": 0.23, "learning_rate": 6.27e-06, "loss": 8.6856, "step": 209 }, { "epoch": 0.24, "learning_rate": 6.3e-06, "loss": 8.6918, "step": 210 }, { "epoch": 0.24, "learning_rate": 6.3299999999999995e-06, "loss": 8.6384, "step": 211 }, { "epoch": 0.24, "learning_rate": 6.36e-06, "loss": 8.6375, "step": 212 }, { "epoch": 0.24, "learning_rate": 6.39e-06, "loss": 8.7154, "step": 213 }, { "epoch": 0.24, "learning_rate": 6.42e-06, "loss": 8.6438, "step": 214 }, { "epoch": 0.24, "learning_rate": 6.45e-06, "loss": 8.7082, "step": 215 }, { "epoch": 0.24, "learning_rate": 6.48e-06, "loss": 8.6748, "step": 216 }, { "epoch": 0.24, "learning_rate": 6.51e-06, "loss": 8.635, "step": 217 }, { "epoch": 0.24, "learning_rate": 6.54e-06, "loss": 8.5619, "step": 218 }, { "epoch": 0.25, "learning_rate": 6.57e-06, "loss": 8.5128, "step": 219 }, { "epoch": 0.25, "learning_rate": 6.6e-06, "loss": 8.6317, "step": 220 }, { "epoch": 0.25, "learning_rate": 6.63e-06, "loss": 8.5787, "step": 221 }, { "epoch": 0.25, "learning_rate": 6.660000000000001e-06, "loss": 8.5318, "step": 222 }, { "epoch": 0.25, "learning_rate": 6.69e-06, "loss": 8.7262, "step": 223 }, { "epoch": 0.25, "learning_rate": 6.72e-06, "loss": 8.6315, "step": 224 }, { "epoch": 0.25, "learning_rate": 6.750000000000001e-06, "loss": 8.5606, "step": 225 }, { "epoch": 0.25, "learning_rate": 6.78e-06, "loss": 8.6401, "step": 226 }, { "epoch": 0.25, "learning_rate": 6.81e-06, "loss": 8.5672, "step": 227 }, { "epoch": 0.26, "learning_rate": 6.840000000000001e-06, "loss": 8.5178, "step": 228 }, { "epoch": 0.26, "learning_rate": 6.87e-06, "loss": 8.5577, "step": 229 }, { "epoch": 0.26, "learning_rate": 6.900000000000001e-06, "loss": 8.5249, "step": 230 }, { "epoch": 0.26, "learning_rate": 6.9300000000000006e-06, "loss": 8.4545, "step": 231 }, { "epoch": 0.26, "learning_rate": 6.96e-06, "loss": 8.5653, "step": 232 }, { "epoch": 0.26, "learning_rate": 6.990000000000001e-06, "loss": 8.5352, "step": 233 }, { "epoch": 0.26, "learning_rate": 7.0200000000000006e-06, "loss": 8.4682, "step": 234 }, { "epoch": 0.26, "learning_rate": 7.049999999999999e-06, "loss": 8.5334, "step": 235 }, { "epoch": 0.26, "learning_rate": 7.08e-06, "loss": 8.364, "step": 236 }, { "epoch": 0.27, "learning_rate": 7.11e-06, "loss": 8.4418, "step": 237 }, { "epoch": 0.27, "learning_rate": 7.14e-06, "loss": 8.5319, "step": 238 }, { "epoch": 0.27, "learning_rate": 7.17e-06, "loss": 8.4589, "step": 239 }, { "epoch": 0.27, "learning_rate": 7.2e-06, "loss": 8.4172, "step": 240 }, { "epoch": 0.27, "learning_rate": 7.23e-06, "loss": 8.4781, "step": 241 }, { "epoch": 0.27, "learning_rate": 7.26e-06, "loss": 8.402, "step": 242 }, { "epoch": 0.27, "learning_rate": 7.29e-06, "loss": 8.2885, "step": 243 }, { "epoch": 0.27, "learning_rate": 7.32e-06, "loss": 8.2803, "step": 244 }, { "epoch": 0.27, "learning_rate": 7.35e-06, "loss": 8.3376, "step": 245 }, { "epoch": 0.28, "learning_rate": 7.3800000000000005e-06, "loss": 8.2272, "step": 246 }, { "epoch": 0.28, "learning_rate": 7.41e-06, "loss": 8.2853, "step": 247 }, { "epoch": 0.28, "learning_rate": 7.44e-06, "loss": 8.3631, "step": 248 }, { "epoch": 0.28, "learning_rate": 7.4700000000000005e-06, "loss": 8.0905, "step": 249 }, { "epoch": 0.28, "learning_rate": 7.5e-06, "loss": 7.8898, "step": 250 }, { "epoch": 0.28, "learning_rate": 7.53e-06, "loss": 8.4397, "step": 251 }, { "epoch": 0.28, "learning_rate": 7.5600000000000005e-06, "loss": 8.4703, "step": 252 }, { "epoch": 0.28, "learning_rate": 7.59e-06, "loss": 8.372, "step": 253 }, { "epoch": 0.28, "learning_rate": 7.62e-06, "loss": 8.5127, "step": 254 }, { "epoch": 0.29, "learning_rate": 7.65e-06, "loss": 8.3414, "step": 255 }, { "epoch": 0.29, "learning_rate": 7.680000000000001e-06, "loss": 8.3941, "step": 256 }, { "epoch": 0.29, "learning_rate": 7.71e-06, "loss": 8.3436, "step": 257 }, { "epoch": 0.29, "learning_rate": 7.74e-06, "loss": 8.3238, "step": 258 }, { "epoch": 0.29, "learning_rate": 7.77e-06, "loss": 8.2363, "step": 259 }, { "epoch": 0.29, "learning_rate": 7.8e-06, "loss": 8.2773, "step": 260 }, { "epoch": 0.29, "learning_rate": 7.830000000000001e-06, "loss": 8.3266, "step": 261 }, { "epoch": 0.29, "learning_rate": 7.860000000000001e-06, "loss": 8.1988, "step": 262 }, { "epoch": 0.29, "learning_rate": 7.89e-06, "loss": 8.2609, "step": 263 }, { "epoch": 0.3, "learning_rate": 7.92e-06, "loss": 8.3095, "step": 264 }, { "epoch": 0.3, "learning_rate": 7.95e-06, "loss": 8.2975, "step": 265 }, { "epoch": 0.3, "learning_rate": 7.98e-06, "loss": 8.3033, "step": 266 }, { "epoch": 0.3, "learning_rate": 8.010000000000001e-06, "loss": 8.356, "step": 267 }, { "epoch": 0.3, "learning_rate": 8.040000000000001e-06, "loss": 8.3039, "step": 268 }, { "epoch": 0.3, "learning_rate": 8.07e-06, "loss": 8.2249, "step": 269 }, { "epoch": 0.3, "learning_rate": 8.1e-06, "loss": 8.2049, "step": 270 }, { "epoch": 0.3, "learning_rate": 8.13e-06, "loss": 8.1977, "step": 271 }, { "epoch": 0.3, "learning_rate": 8.160000000000001e-06, "loss": 8.1587, "step": 272 }, { "epoch": 0.31, "learning_rate": 8.190000000000001e-06, "loss": 8.1175, "step": 273 }, { "epoch": 0.31, "learning_rate": 8.220000000000001e-06, "loss": 8.0332, "step": 274 }, { "epoch": 0.31, "learning_rate": 8.25e-06, "loss": 8.2667, "step": 275 }, { "epoch": 0.31, "learning_rate": 8.28e-06, "loss": 8.1367, "step": 276 }, { "epoch": 0.31, "learning_rate": 8.310000000000002e-06, "loss": 8.0283, "step": 277 }, { "epoch": 0.31, "learning_rate": 8.340000000000001e-06, "loss": 8.124, "step": 278 }, { "epoch": 0.31, "learning_rate": 8.370000000000001e-06, "loss": 8.1631, "step": 279 }, { "epoch": 0.31, "learning_rate": 8.400000000000001e-06, "loss": 8.1611, "step": 280 }, { "epoch": 0.32, "learning_rate": 8.43e-06, "loss": 8.0191, "step": 281 }, { "epoch": 0.32, "learning_rate": 8.459999999999999e-06, "loss": 8.0795, "step": 282 }, { "epoch": 0.32, "learning_rate": 8.49e-06, "loss": 7.9484, "step": 283 }, { "epoch": 0.32, "learning_rate": 8.52e-06, "loss": 8.1467, "step": 284 }, { "epoch": 0.32, "learning_rate": 8.55e-06, "loss": 8.0431, "step": 285 }, { "epoch": 0.32, "learning_rate": 8.58e-06, "loss": 7.9628, "step": 286 }, { "epoch": 0.32, "learning_rate": 8.609999999999999e-06, "loss": 8.2102, "step": 287 }, { "epoch": 0.32, "learning_rate": 8.64e-06, "loss": 8.0003, "step": 288 }, { "epoch": 0.32, "learning_rate": 8.67e-06, "loss": 8.169, "step": 289 }, { "epoch": 0.33, "learning_rate": 8.7e-06, "loss": 8.1153, "step": 290 }, { "epoch": 0.33, "learning_rate": 8.73e-06, "loss": 8.1166, "step": 291 }, { "epoch": 0.33, "learning_rate": 8.759999999999999e-06, "loss": 7.9297, "step": 292 }, { "epoch": 0.33, "learning_rate": 8.79e-06, "loss": 7.8661, "step": 293 }, { "epoch": 0.33, "learning_rate": 8.82e-06, "loss": 7.9741, "step": 294 }, { "epoch": 0.33, "learning_rate": 8.85e-06, "loss": 7.8919, "step": 295 }, { "epoch": 0.33, "learning_rate": 8.88e-06, "loss": 7.896, "step": 296 }, { "epoch": 0.33, "learning_rate": 8.91e-06, "loss": 7.9141, "step": 297 }, { "epoch": 0.33, "learning_rate": 8.939999999999999e-06, "loss": 7.819, "step": 298 }, { "epoch": 0.34, "learning_rate": 8.97e-06, "loss": 7.7626, "step": 299 }, { "epoch": 0.34, "learning_rate": 9e-06, "loss": 7.5145, "step": 300 }, { "epoch": 0.34, "learning_rate": 9.03e-06, "loss": 8.0349, "step": 301 }, { "epoch": 0.34, "learning_rate": 9.06e-06, "loss": 7.9921, "step": 302 }, { "epoch": 0.34, "learning_rate": 9.09e-06, "loss": 7.9416, "step": 303 }, { "epoch": 0.34, "learning_rate": 9.12e-06, "loss": 7.9991, "step": 304 }, { "epoch": 0.34, "learning_rate": 9.15e-06, "loss": 7.8294, "step": 305 }, { "epoch": 0.34, "learning_rate": 9.18e-06, "loss": 7.956, "step": 306 }, { "epoch": 0.34, "learning_rate": 9.21e-06, "loss": 7.8918, "step": 307 }, { "epoch": 0.35, "learning_rate": 9.24e-06, "loss": 7.973, "step": 308 }, { "epoch": 0.35, "learning_rate": 9.27e-06, "loss": 7.8696, "step": 309 }, { "epoch": 0.35, "learning_rate": 9.3e-06, "loss": 7.9359, "step": 310 }, { "epoch": 0.35, "learning_rate": 9.33e-06, "loss": 7.9476, "step": 311 }, { "epoch": 0.35, "learning_rate": 9.36e-06, "loss": 7.9668, "step": 312 }, { "epoch": 0.35, "learning_rate": 9.39e-06, "loss": 7.8796, "step": 313 }, { "epoch": 0.35, "learning_rate": 9.42e-06, "loss": 7.9165, "step": 314 }, { "epoch": 0.35, "learning_rate": 9.450000000000001e-06, "loss": 7.8171, "step": 315 }, { "epoch": 0.35, "learning_rate": 9.48e-06, "loss": 7.8817, "step": 316 }, { "epoch": 0.36, "learning_rate": 9.51e-06, "loss": 7.8338, "step": 317 }, { "epoch": 0.36, "learning_rate": 9.54e-06, "loss": 7.8629, "step": 318 }, { "epoch": 0.36, "learning_rate": 9.57e-06, "loss": 7.7397, "step": 319 }, { "epoch": 0.36, "learning_rate": 9.600000000000001e-06, "loss": 7.8831, "step": 320 }, { "epoch": 0.36, "learning_rate": 9.630000000000001e-06, "loss": 7.8358, "step": 321 }, { "epoch": 0.36, "learning_rate": 9.66e-06, "loss": 7.7765, "step": 322 }, { "epoch": 0.36, "learning_rate": 9.69e-06, "loss": 7.7984, "step": 323 }, { "epoch": 0.36, "learning_rate": 9.72e-06, "loss": 7.8341, "step": 324 }, { "epoch": 0.36, "learning_rate": 9.75e-06, "loss": 7.7684, "step": 325 }, { "epoch": 0.37, "learning_rate": 9.780000000000001e-06, "loss": 7.7689, "step": 326 }, { "epoch": 0.37, "learning_rate": 9.810000000000001e-06, "loss": 7.7012, "step": 327 }, { "epoch": 0.37, "learning_rate": 9.84e-06, "loss": 7.7769, "step": 328 }, { "epoch": 0.37, "learning_rate": 9.87e-06, "loss": 7.7687, "step": 329 }, { "epoch": 0.37, "learning_rate": 9.9e-06, "loss": 7.769, "step": 330 }, { "epoch": 0.37, "learning_rate": 9.930000000000001e-06, "loss": 7.7026, "step": 331 }, { "epoch": 0.37, "learning_rate": 9.960000000000001e-06, "loss": 7.4947, "step": 332 }, { "epoch": 0.37, "learning_rate": 9.990000000000001e-06, "loss": 7.7253, "step": 333 }, { "epoch": 0.37, "learning_rate": 1.002e-05, "loss": 7.7003, "step": 334 }, { "epoch": 0.38, "learning_rate": 1.005e-05, "loss": 7.762, "step": 335 }, { "epoch": 0.38, "learning_rate": 1.008e-05, "loss": 7.6208, "step": 336 }, { "epoch": 0.38, "learning_rate": 1.0110000000000001e-05, "loss": 7.5637, "step": 337 }, { "epoch": 0.38, "learning_rate": 1.0140000000000001e-05, "loss": 7.7512, "step": 338 }, { "epoch": 0.38, "learning_rate": 1.0170000000000001e-05, "loss": 7.7148, "step": 339 }, { "epoch": 0.38, "learning_rate": 1.02e-05, "loss": 7.6507, "step": 340 }, { "epoch": 0.38, "learning_rate": 1.023e-05, "loss": 7.6943, "step": 341 }, { "epoch": 0.38, "learning_rate": 1.0260000000000002e-05, "loss": 7.6202, "step": 342 }, { "epoch": 0.38, "learning_rate": 1.0290000000000001e-05, "loss": 7.6772, "step": 343 }, { "epoch": 0.39, "learning_rate": 1.032e-05, "loss": 7.619, "step": 344 }, { "epoch": 0.39, "learning_rate": 1.035e-05, "loss": 7.415, "step": 345 }, { "epoch": 0.39, "learning_rate": 1.0379999999999999e-05, "loss": 7.5973, "step": 346 }, { "epoch": 0.39, "learning_rate": 1.041e-05, "loss": 7.4379, "step": 347 }, { "epoch": 0.39, "learning_rate": 1.044e-05, "loss": 7.5475, "step": 348 }, { "epoch": 0.39, "learning_rate": 1.047e-05, "loss": 7.559, "step": 349 }, { "epoch": 0.39, "learning_rate": 1.05e-05, "loss": 7.2969, "step": 350 }, { "epoch": 0.39, "learning_rate": 1.0529999999999999e-05, "loss": 7.6362, "step": 351 }, { "epoch": 0.39, "learning_rate": 1.0559999999999999e-05, "loss": 7.6349, "step": 352 }, { "epoch": 0.4, "learning_rate": 1.059e-05, "loss": 7.6525, "step": 353 }, { "epoch": 0.4, "learning_rate": 1.062e-05, "loss": 7.6345, "step": 354 }, { "epoch": 0.4, "learning_rate": 1.065e-05, "loss": 7.6702, "step": 355 }, { "epoch": 0.4, "learning_rate": 1.068e-05, "loss": 7.5792, "step": 356 }, { "epoch": 0.4, "learning_rate": 1.0709999999999999e-05, "loss": 7.735, "step": 357 }, { "epoch": 0.4, "learning_rate": 1.074e-05, "loss": 7.523, "step": 358 }, { "epoch": 0.4, "learning_rate": 1.077e-05, "loss": 7.523, "step": 359 }, { "epoch": 0.4, "learning_rate": 1.08e-05, "loss": 7.4569, "step": 360 }, { "epoch": 0.4, "learning_rate": 1.083e-05, "loss": 7.4638, "step": 361 }, { "epoch": 0.41, "learning_rate": 1.086e-05, "loss": 7.5615, "step": 362 }, { "epoch": 0.41, "learning_rate": 1.089e-05, "loss": 7.5019, "step": 363 }, { "epoch": 0.41, "learning_rate": 1.092e-05, "loss": 7.5408, "step": 364 }, { "epoch": 0.41, "learning_rate": 1.095e-05, "loss": 7.4969, "step": 365 }, { "epoch": 0.41, "learning_rate": 1.098e-05, "loss": 7.4347, "step": 366 }, { "epoch": 0.41, "learning_rate": 1.101e-05, "loss": 7.4298, "step": 367 }, { "epoch": 0.41, "learning_rate": 1.104e-05, "loss": 7.4546, "step": 368 }, { "epoch": 0.41, "learning_rate": 1.107e-05, "loss": 7.3747, "step": 369 }, { "epoch": 0.41, "learning_rate": 1.11e-05, "loss": 7.5675, "step": 370 }, { "epoch": 0.42, "learning_rate": 1.113e-05, "loss": 7.4561, "step": 371 }, { "epoch": 0.42, "learning_rate": 1.116e-05, "loss": 7.6097, "step": 372 }, { "epoch": 0.42, "learning_rate": 1.119e-05, "loss": 7.458, "step": 373 }, { "epoch": 0.42, "learning_rate": 1.1220000000000001e-05, "loss": 7.5322, "step": 374 }, { "epoch": 0.42, "learning_rate": 1.125e-05, "loss": 7.3106, "step": 375 }, { "epoch": 0.42, "learning_rate": 1.128e-05, "loss": 7.4766, "step": 376 }, { "epoch": 0.42, "learning_rate": 1.131e-05, "loss": 7.3253, "step": 377 }, { "epoch": 0.42, "learning_rate": 1.134e-05, "loss": 7.339, "step": 378 }, { "epoch": 0.42, "learning_rate": 1.137e-05, "loss": 7.3829, "step": 379 }, { "epoch": 0.43, "learning_rate": 1.1400000000000001e-05, "loss": 7.4647, "step": 380 }, { "epoch": 0.43, "learning_rate": 1.143e-05, "loss": 7.441, "step": 381 }, { "epoch": 0.43, "learning_rate": 1.146e-05, "loss": 7.3215, "step": 382 }, { "epoch": 0.43, "learning_rate": 1.149e-05, "loss": 7.4654, "step": 383 }, { "epoch": 0.43, "learning_rate": 1.152e-05, "loss": 7.4314, "step": 384 }, { "epoch": 0.43, "learning_rate": 1.1550000000000001e-05, "loss": 7.3343, "step": 385 }, { "epoch": 0.43, "learning_rate": 1.1580000000000001e-05, "loss": 7.3554, "step": 386 }, { "epoch": 0.43, "learning_rate": 1.161e-05, "loss": 7.3426, "step": 387 }, { "epoch": 0.43, "learning_rate": 1.164e-05, "loss": 7.3292, "step": 388 }, { "epoch": 0.44, "learning_rate": 1.167e-05, "loss": 7.2827, "step": 389 }, { "epoch": 0.44, "learning_rate": 1.1700000000000001e-05, "loss": 7.4665, "step": 390 }, { "epoch": 0.44, "learning_rate": 1.1730000000000001e-05, "loss": 7.416, "step": 391 }, { "epoch": 0.44, "learning_rate": 1.1760000000000001e-05, "loss": 7.3757, "step": 392 }, { "epoch": 0.44, "learning_rate": 1.179e-05, "loss": 7.2848, "step": 393 }, { "epoch": 0.44, "learning_rate": 1.182e-05, "loss": 7.4558, "step": 394 }, { "epoch": 0.44, "learning_rate": 1.185e-05, "loss": 7.254, "step": 395 }, { "epoch": 0.44, "learning_rate": 1.1880000000000001e-05, "loss": 7.0884, "step": 396 }, { "epoch": 0.45, "learning_rate": 1.1910000000000001e-05, "loss": 7.0408, "step": 397 }, { "epoch": 0.45, "learning_rate": 1.1940000000000001e-05, "loss": 7.2409, "step": 398 }, { "epoch": 0.45, "learning_rate": 1.197e-05, "loss": 7.1615, "step": 399 }, { "epoch": 0.45, "learning_rate": 1.2e-05, "loss": 7.0189, "step": 400 }, { "epoch": 0.45, "learning_rate": 1.2030000000000002e-05, "loss": 7.6844, "step": 401 }, { "epoch": 0.45, "learning_rate": 1.2060000000000001e-05, "loss": 7.6816, "step": 402 }, { "epoch": 0.45, "learning_rate": 1.2090000000000001e-05, "loss": 7.6434, "step": 403 }, { "epoch": 0.45, "learning_rate": 1.2120000000000001e-05, "loss": 7.5461, "step": 404 }, { "epoch": 0.45, "learning_rate": 1.215e-05, "loss": 7.6362, "step": 405 }, { "epoch": 0.46, "learning_rate": 1.2180000000000002e-05, "loss": 7.4529, "step": 406 }, { "epoch": 0.46, "learning_rate": 1.221e-05, "loss": 7.3868, "step": 407 }, { "epoch": 0.46, "learning_rate": 1.224e-05, "loss": 7.3136, "step": 408 }, { "epoch": 0.46, "learning_rate": 1.227e-05, "loss": 7.47, "step": 409 }, { "epoch": 0.46, "learning_rate": 1.2299999999999999e-05, "loss": 7.3939, "step": 410 }, { "epoch": 0.46, "learning_rate": 1.2329999999999999e-05, "loss": 7.3262, "step": 411 }, { "epoch": 0.46, "learning_rate": 1.236e-05, "loss": 7.3071, "step": 412 }, { "epoch": 0.46, "learning_rate": 1.239e-05, "loss": 7.1696, "step": 413 }, { "epoch": 0.46, "learning_rate": 1.242e-05, "loss": 7.152, "step": 414 }, { "epoch": 0.47, "learning_rate": 1.245e-05, "loss": 7.2075, "step": 415 }, { "epoch": 0.47, "learning_rate": 1.2479999999999999e-05, "loss": 7.3324, "step": 416 }, { "epoch": 0.47, "learning_rate": 1.251e-05, "loss": 7.2192, "step": 417 }, { "epoch": 0.47, "learning_rate": 1.254e-05, "loss": 7.296, "step": 418 }, { "epoch": 0.47, "learning_rate": 1.257e-05, "loss": 7.2919, "step": 419 }, { "epoch": 0.47, "learning_rate": 1.26e-05, "loss": 7.3075, "step": 420 }, { "epoch": 0.47, "learning_rate": 1.263e-05, "loss": 7.2815, "step": 421 }, { "epoch": 0.47, "learning_rate": 1.2659999999999999e-05, "loss": 7.3214, "step": 422 }, { "epoch": 0.47, "learning_rate": 1.269e-05, "loss": 7.296, "step": 423 }, { "epoch": 0.48, "learning_rate": 1.272e-05, "loss": 7.2566, "step": 424 }, { "epoch": 0.48, "learning_rate": 1.275e-05, "loss": 7.2731, "step": 425 }, { "epoch": 0.48, "learning_rate": 1.278e-05, "loss": 7.3859, "step": 426 }, { "epoch": 0.48, "learning_rate": 1.281e-05, "loss": 7.1349, "step": 427 }, { "epoch": 0.48, "learning_rate": 1.284e-05, "loss": 7.4523, "step": 428 }, { "epoch": 0.48, "learning_rate": 1.287e-05, "loss": 7.213, "step": 429 }, { "epoch": 0.48, "learning_rate": 1.29e-05, "loss": 7.2149, "step": 430 }, { "epoch": 0.48, "learning_rate": 1.293e-05, "loss": 7.3725, "step": 431 }, { "epoch": 0.48, "learning_rate": 1.296e-05, "loss": 7.2863, "step": 432 }, { "epoch": 0.49, "learning_rate": 1.2990000000000001e-05, "loss": 6.9803, "step": 433 }, { "epoch": 0.49, "learning_rate": 1.302e-05, "loss": 7.1407, "step": 434 }, { "epoch": 0.49, "learning_rate": 1.305e-05, "loss": 7.0905, "step": 435 }, { "epoch": 0.49, "learning_rate": 1.308e-05, "loss": 7.246, "step": 436 }, { "epoch": 0.49, "learning_rate": 1.311e-05, "loss": 7.223, "step": 437 }, { "epoch": 0.49, "learning_rate": 1.314e-05, "loss": 7.2302, "step": 438 }, { "epoch": 0.49, "learning_rate": 1.3170000000000001e-05, "loss": 7.0976, "step": 439 }, { "epoch": 0.49, "learning_rate": 1.32e-05, "loss": 7.2214, "step": 440 }, { "epoch": 0.49, "learning_rate": 1.323e-05, "loss": 7.2718, "step": 441 }, { "epoch": 0.5, "learning_rate": 1.326e-05, "loss": 7.1749, "step": 442 }, { "epoch": 0.5, "learning_rate": 1.329e-05, "loss": 7.069, "step": 443 }, { "epoch": 0.5, "learning_rate": 1.3320000000000001e-05, "loss": 7.3068, "step": 444 }, { "epoch": 0.5, "learning_rate": 1.3350000000000001e-05, "loss": 6.9465, "step": 445 }, { "epoch": 0.5, "learning_rate": 1.338e-05, "loss": 6.985, "step": 446 }, { "epoch": 0.5, "learning_rate": 1.341e-05, "loss": 7.0937, "step": 447 }, { "epoch": 0.5, "learning_rate": 1.344e-05, "loss": 7.0551, "step": 448 }, { "epoch": 0.5, "learning_rate": 1.3470000000000001e-05, "loss": 6.9195, "step": 449 }, { "epoch": 0.5, "learning_rate": 1.3500000000000001e-05, "loss": 6.8403, "step": 450 }, { "epoch": 0.51, "learning_rate": 1.3530000000000001e-05, "loss": 7.3704, "step": 451 }, { "epoch": 0.51, "learning_rate": 1.356e-05, "loss": 7.4571, "step": 452 }, { "epoch": 0.51, "learning_rate": 1.359e-05, "loss": 7.5165, "step": 453 }, { "epoch": 0.51, "learning_rate": 1.362e-05, "loss": 7.2796, "step": 454 }, { "epoch": 0.51, "learning_rate": 1.3650000000000001e-05, "loss": 7.362, "step": 455 }, { "epoch": 0.51, "learning_rate": 1.3680000000000001e-05, "loss": 7.4003, "step": 456 }, { "epoch": 0.51, "learning_rate": 1.3710000000000001e-05, "loss": 7.1832, "step": 457 }, { "epoch": 0.51, "learning_rate": 1.374e-05, "loss": 7.1787, "step": 458 }, { "epoch": 0.51, "learning_rate": 1.377e-05, "loss": 7.1757, "step": 459 }, { "epoch": 0.52, "learning_rate": 1.3800000000000002e-05, "loss": 7.242, "step": 460 }, { "epoch": 0.52, "learning_rate": 1.3830000000000001e-05, "loss": 7.0405, "step": 461 }, { "epoch": 0.52, "learning_rate": 1.3860000000000001e-05, "loss": 7.2315, "step": 462 }, { "epoch": 0.52, "learning_rate": 1.389e-05, "loss": 7.0826, "step": 463 }, { "epoch": 0.52, "learning_rate": 1.392e-05, "loss": 7.0607, "step": 464 }, { "epoch": 0.52, "learning_rate": 1.395e-05, "loss": 7.1658, "step": 465 }, { "epoch": 0.52, "learning_rate": 1.3980000000000002e-05, "loss": 7.0979, "step": 466 }, { "epoch": 0.52, "learning_rate": 1.4010000000000001e-05, "loss": 7.1165, "step": 467 }, { "epoch": 0.52, "learning_rate": 1.4040000000000001e-05, "loss": 7.163, "step": 468 }, { "epoch": 0.53, "learning_rate": 1.4069999999999999e-05, "loss": 7.0705, "step": 469 }, { "epoch": 0.53, "learning_rate": 1.4099999999999999e-05, "loss": 7.1478, "step": 470 }, { "epoch": 0.53, "learning_rate": 1.413e-05, "loss": 7.0901, "step": 471 }, { "epoch": 0.53, "learning_rate": 1.416e-05, "loss": 7.0749, "step": 472 }, { "epoch": 0.53, "learning_rate": 1.419e-05, "loss": 7.1195, "step": 473 }, { "epoch": 0.53, "learning_rate": 1.422e-05, "loss": 7.0042, "step": 474 }, { "epoch": 0.53, "learning_rate": 1.4249999999999999e-05, "loss": 7.1632, "step": 475 }, { "epoch": 0.53, "learning_rate": 1.428e-05, "loss": 7.1534, "step": 476 }, { "epoch": 0.53, "learning_rate": 1.431e-05, "loss": 7.2385, "step": 477 }, { "epoch": 0.54, "learning_rate": 1.434e-05, "loss": 7.0036, "step": 478 }, { "epoch": 0.54, "learning_rate": 1.437e-05, "loss": 6.9553, "step": 479 }, { "epoch": 0.54, "learning_rate": 1.44e-05, "loss": 6.9784, "step": 480 }, { "epoch": 0.54, "learning_rate": 1.4429999999999999e-05, "loss": 7.012, "step": 481 }, { "epoch": 0.54, "learning_rate": 1.446e-05, "loss": 7.0209, "step": 482 }, { "epoch": 0.54, "learning_rate": 1.449e-05, "loss": 7.0518, "step": 483 }, { "epoch": 0.54, "learning_rate": 1.452e-05, "loss": 7.0685, "step": 484 }, { "epoch": 0.54, "learning_rate": 1.455e-05, "loss": 7.1275, "step": 485 }, { "epoch": 0.54, "learning_rate": 1.458e-05, "loss": 7.0102, "step": 486 }, { "epoch": 0.55, "learning_rate": 1.461e-05, "loss": 7.0275, "step": 487 }, { "epoch": 0.55, "learning_rate": 1.464e-05, "loss": 6.9669, "step": 488 }, { "epoch": 0.55, "learning_rate": 1.467e-05, "loss": 7.1323, "step": 489 }, { "epoch": 0.55, "learning_rate": 1.47e-05, "loss": 7.1855, "step": 490 }, { "epoch": 0.55, "learning_rate": 1.473e-05, "loss": 6.9329, "step": 491 }, { "epoch": 0.55, "learning_rate": 1.4760000000000001e-05, "loss": 7.0087, "step": 492 }, { "epoch": 0.55, "learning_rate": 1.479e-05, "loss": 7.0379, "step": 493 }, { "epoch": 0.55, "learning_rate": 1.482e-05, "loss": 7.0483, "step": 494 }, { "epoch": 0.55, "learning_rate": 1.485e-05, "loss": 6.9957, "step": 495 }, { "epoch": 0.56, "learning_rate": 1.488e-05, "loss": 6.8874, "step": 496 }, { "epoch": 0.56, "learning_rate": 1.491e-05, "loss": 6.738, "step": 497 }, { "epoch": 0.56, "learning_rate": 1.4940000000000001e-05, "loss": 7.2307, "step": 498 }, { "epoch": 0.56, "learning_rate": 1.497e-05, "loss": 6.994, "step": 499 }, { "epoch": 0.56, "learning_rate": 1.5e-05, "loss": 6.7125, "step": 500 }, { "epoch": 0.56, "learning_rate": 1.503e-05, "loss": 7.6062, "step": 501 }, { "epoch": 0.56, "learning_rate": 1.506e-05, "loss": 7.3811, "step": 502 }, { "epoch": 0.56, "learning_rate": 1.5090000000000001e-05, "loss": 7.4753, "step": 503 }, { "epoch": 0.57, "learning_rate": 1.5120000000000001e-05, "loss": 7.4732, "step": 504 }, { "epoch": 0.57, "learning_rate": 1.515e-05, "loss": 7.3572, "step": 505 }, { "epoch": 0.57, "learning_rate": 1.518e-05, "loss": 7.248, "step": 506 }, { "epoch": 0.57, "learning_rate": 1.521e-05, "loss": 7.1674, "step": 507 }, { "epoch": 0.57, "learning_rate": 1.524e-05, "loss": 7.0337, "step": 508 }, { "epoch": 0.57, "learning_rate": 1.527e-05, "loss": 7.0716, "step": 509 }, { "epoch": 0.57, "learning_rate": 1.53e-05, "loss": 7.0661, "step": 510 }, { "epoch": 0.57, "learning_rate": 1.533e-05, "loss": 7.0423, "step": 511 }, { "epoch": 0.57, "learning_rate": 1.5360000000000002e-05, "loss": 7.0613, "step": 512 }, { "epoch": 0.58, "learning_rate": 1.539e-05, "loss": 7.2284, "step": 513 }, { "epoch": 0.58, "learning_rate": 1.542e-05, "loss": 7.1288, "step": 514 }, { "epoch": 0.58, "learning_rate": 1.545e-05, "loss": 7.124, "step": 515 }, { "epoch": 0.58, "learning_rate": 1.548e-05, "loss": 6.9077, "step": 516 }, { "epoch": 0.58, "learning_rate": 1.5510000000000002e-05, "loss": 7.0841, "step": 517 }, { "epoch": 0.58, "learning_rate": 1.554e-05, "loss": 6.9657, "step": 518 }, { "epoch": 0.58, "learning_rate": 1.5570000000000002e-05, "loss": 7.0967, "step": 519 }, { "epoch": 0.58, "learning_rate": 1.56e-05, "loss": 7.1554, "step": 520 }, { "epoch": 0.58, "learning_rate": 1.563e-05, "loss": 6.9425, "step": 521 }, { "epoch": 0.59, "learning_rate": 1.5660000000000003e-05, "loss": 6.8829, "step": 522 }, { "epoch": 0.59, "learning_rate": 1.569e-05, "loss": 6.8259, "step": 523 }, { "epoch": 0.59, "learning_rate": 1.5720000000000002e-05, "loss": 6.9507, "step": 524 }, { "epoch": 0.59, "learning_rate": 1.575e-05, "loss": 7.0513, "step": 525 }, { "epoch": 0.59, "learning_rate": 1.578e-05, "loss": 6.8547, "step": 526 }, { "epoch": 0.59, "learning_rate": 1.5810000000000003e-05, "loss": 6.9426, "step": 527 }, { "epoch": 0.59, "learning_rate": 1.584e-05, "loss": 7.0087, "step": 528 }, { "epoch": 0.59, "learning_rate": 1.5870000000000002e-05, "loss": 7.0541, "step": 529 }, { "epoch": 0.59, "learning_rate": 1.59e-05, "loss": 7.0746, "step": 530 }, { "epoch": 0.6, "learning_rate": 1.593e-05, "loss": 6.9426, "step": 531 }, { "epoch": 0.6, "learning_rate": 1.596e-05, "loss": 6.9161, "step": 532 }, { "epoch": 0.6, "learning_rate": 1.599e-05, "loss": 6.8707, "step": 533 }, { "epoch": 0.6, "learning_rate": 1.6020000000000002e-05, "loss": 6.877, "step": 534 }, { "epoch": 0.6, "learning_rate": 1.605e-05, "loss": 6.8998, "step": 535 }, { "epoch": 0.6, "learning_rate": 1.6080000000000002e-05, "loss": 7.0667, "step": 536 }, { "epoch": 0.6, "learning_rate": 1.611e-05, "loss": 6.999, "step": 537 }, { "epoch": 0.6, "learning_rate": 1.614e-05, "loss": 7.0087, "step": 538 }, { "epoch": 0.6, "learning_rate": 1.6170000000000003e-05, "loss": 7.012, "step": 539 }, { "epoch": 0.61, "learning_rate": 1.62e-05, "loss": 6.9587, "step": 540 }, { "epoch": 0.61, "learning_rate": 1.6230000000000002e-05, "loss": 7.0052, "step": 541 }, { "epoch": 0.61, "learning_rate": 1.626e-05, "loss": 6.9466, "step": 542 }, { "epoch": 0.61, "learning_rate": 1.629e-05, "loss": 6.8965, "step": 543 }, { "epoch": 0.61, "learning_rate": 1.6320000000000003e-05, "loss": 6.7743, "step": 544 }, { "epoch": 0.61, "learning_rate": 1.635e-05, "loss": 6.8034, "step": 545 }, { "epoch": 0.61, "learning_rate": 1.6380000000000002e-05, "loss": 6.8411, "step": 546 }, { "epoch": 0.61, "learning_rate": 1.641e-05, "loss": 7.0767, "step": 547 }, { "epoch": 0.61, "learning_rate": 1.6440000000000002e-05, "loss": 6.918, "step": 548 }, { "epoch": 0.62, "learning_rate": 1.6470000000000003e-05, "loss": 6.584, "step": 549 }, { "epoch": 0.62, "learning_rate": 1.65e-05, "loss": 6.8046, "step": 550 }, { "epoch": 0.62, "learning_rate": 1.6530000000000003e-05, "loss": 7.3232, "step": 551 }, { "epoch": 0.62, "learning_rate": 1.656e-05, "loss": 7.5494, "step": 552 }, { "epoch": 0.62, "learning_rate": 1.6590000000000002e-05, "loss": 7.3171, "step": 553 }, { "epoch": 0.62, "learning_rate": 1.6620000000000004e-05, "loss": 7.3058, "step": 554 }, { "epoch": 0.62, "learning_rate": 1.665e-05, "loss": 7.1965, "step": 555 }, { "epoch": 0.62, "learning_rate": 1.6680000000000003e-05, "loss": 7.1298, "step": 556 }, { "epoch": 0.62, "learning_rate": 1.671e-05, "loss": 6.9915, "step": 557 }, { "epoch": 0.63, "learning_rate": 1.6740000000000002e-05, "loss": 7.0506, "step": 558 }, { "epoch": 0.63, "learning_rate": 1.677e-05, "loss": 6.9666, "step": 559 }, { "epoch": 0.63, "learning_rate": 1.6800000000000002e-05, "loss": 6.8825, "step": 560 }, { "epoch": 0.63, "learning_rate": 1.6830000000000003e-05, "loss": 6.985, "step": 561 }, { "epoch": 0.63, "learning_rate": 1.686e-05, "loss": 6.838, "step": 562 }, { "epoch": 0.63, "learning_rate": 1.689e-05, "loss": 6.8455, "step": 563 }, { "epoch": 0.63, "learning_rate": 1.6919999999999997e-05, "loss": 6.9067, "step": 564 }, { "epoch": 0.63, "learning_rate": 1.695e-05, "loss": 7.096, "step": 565 }, { "epoch": 0.63, "learning_rate": 1.698e-05, "loss": 6.8064, "step": 566 }, { "epoch": 0.64, "learning_rate": 1.7009999999999998e-05, "loss": 6.8761, "step": 567 }, { "epoch": 0.64, "learning_rate": 1.704e-05, "loss": 6.8555, "step": 568 }, { "epoch": 0.64, "learning_rate": 1.7069999999999998e-05, "loss": 6.8458, "step": 569 }, { "epoch": 0.64, "learning_rate": 1.71e-05, "loss": 6.9565, "step": 570 }, { "epoch": 0.64, "learning_rate": 1.713e-05, "loss": 6.8278, "step": 571 }, { "epoch": 0.64, "learning_rate": 1.716e-05, "loss": 7.0034, "step": 572 }, { "epoch": 0.64, "learning_rate": 1.719e-05, "loss": 6.7617, "step": 573 }, { "epoch": 0.64, "learning_rate": 1.7219999999999998e-05, "loss": 6.7165, "step": 574 }, { "epoch": 0.64, "learning_rate": 1.725e-05, "loss": 6.9117, "step": 575 }, { "epoch": 0.65, "learning_rate": 1.728e-05, "loss": 6.6774, "step": 576 }, { "epoch": 0.65, "learning_rate": 1.731e-05, "loss": 6.8219, "step": 577 }, { "epoch": 0.65, "learning_rate": 1.734e-05, "loss": 6.9196, "step": 578 }, { "epoch": 0.65, "learning_rate": 1.7369999999999998e-05, "loss": 6.7383, "step": 579 }, { "epoch": 0.65, "learning_rate": 1.74e-05, "loss": 7.0127, "step": 580 }, { "epoch": 0.65, "learning_rate": 1.743e-05, "loss": 6.9535, "step": 581 }, { "epoch": 0.65, "learning_rate": 1.746e-05, "loss": 6.8286, "step": 582 }, { "epoch": 0.65, "learning_rate": 1.749e-05, "loss": 6.8007, "step": 583 }, { "epoch": 0.65, "learning_rate": 1.7519999999999998e-05, "loss": 6.7722, "step": 584 }, { "epoch": 0.66, "learning_rate": 1.755e-05, "loss": 6.9072, "step": 585 }, { "epoch": 0.66, "learning_rate": 1.758e-05, "loss": 6.8819, "step": 586 }, { "epoch": 0.66, "learning_rate": 1.761e-05, "loss": 6.91, "step": 587 }, { "epoch": 0.66, "learning_rate": 1.764e-05, "loss": 6.8204, "step": 588 }, { "epoch": 0.66, "learning_rate": 1.767e-05, "loss": 6.7104, "step": 589 }, { "epoch": 0.66, "learning_rate": 1.77e-05, "loss": 6.7947, "step": 590 }, { "epoch": 0.66, "learning_rate": 1.7729999999999998e-05, "loss": 6.8408, "step": 591 }, { "epoch": 0.66, "learning_rate": 1.776e-05, "loss": 6.8955, "step": 592 }, { "epoch": 0.66, "learning_rate": 1.779e-05, "loss": 6.9198, "step": 593 }, { "epoch": 0.67, "learning_rate": 1.782e-05, "loss": 6.6562, "step": 594 }, { "epoch": 0.67, "learning_rate": 1.785e-05, "loss": 6.7913, "step": 595 }, { "epoch": 0.67, "learning_rate": 1.7879999999999998e-05, "loss": 6.5692, "step": 596 }, { "epoch": 0.67, "learning_rate": 1.791e-05, "loss": 6.6054, "step": 597 }, { "epoch": 0.67, "learning_rate": 1.794e-05, "loss": 6.7038, "step": 598 }, { "epoch": 0.67, "learning_rate": 1.797e-05, "loss": 6.5903, "step": 599 }, { "epoch": 0.67, "learning_rate": 1.8e-05, "loss": 6.3211, "step": 600 }, { "epoch": 0.67, "learning_rate": 1.803e-05, "loss": 7.1768, "step": 601 }, { "epoch": 0.67, "learning_rate": 1.806e-05, "loss": 7.0299, "step": 602 }, { "epoch": 0.68, "learning_rate": 1.809e-05, "loss": 7.1771, "step": 603 }, { "epoch": 0.68, "learning_rate": 1.812e-05, "loss": 7.091, "step": 604 }, { "epoch": 0.68, "learning_rate": 1.815e-05, "loss": 7.0872, "step": 605 }, { "epoch": 0.68, "learning_rate": 1.818e-05, "loss": 6.7709, "step": 606 }, { "epoch": 0.68, "learning_rate": 1.821e-05, "loss": 6.8938, "step": 607 }, { "epoch": 0.68, "learning_rate": 1.824e-05, "loss": 6.8887, "step": 608 }, { "epoch": 0.68, "learning_rate": 1.827e-05, "loss": 6.8837, "step": 609 }, { "epoch": 0.68, "learning_rate": 1.83e-05, "loss": 6.8241, "step": 610 }, { "epoch": 0.68, "learning_rate": 1.833e-05, "loss": 6.7565, "step": 611 }, { "epoch": 0.69, "learning_rate": 1.836e-05, "loss": 6.8718, "step": 612 }, { "epoch": 0.69, "learning_rate": 1.8390000000000002e-05, "loss": 6.7659, "step": 613 }, { "epoch": 0.69, "learning_rate": 1.842e-05, "loss": 6.7993, "step": 614 }, { "epoch": 0.69, "learning_rate": 1.845e-05, "loss": 6.996, "step": 615 }, { "epoch": 0.69, "learning_rate": 1.848e-05, "loss": 6.716, "step": 616 }, { "epoch": 0.69, "learning_rate": 1.851e-05, "loss": 6.7879, "step": 617 }, { "epoch": 0.69, "learning_rate": 1.854e-05, "loss": 6.6775, "step": 618 }, { "epoch": 0.69, "learning_rate": 1.857e-05, "loss": 6.7344, "step": 619 }, { "epoch": 0.7, "learning_rate": 1.86e-05, "loss": 6.7928, "step": 620 }, { "epoch": 0.7, "learning_rate": 1.863e-05, "loss": 6.8515, "step": 621 }, { "epoch": 0.7, "learning_rate": 1.866e-05, "loss": 6.7522, "step": 622 }, { "epoch": 0.7, "learning_rate": 1.869e-05, "loss": 6.8367, "step": 623 }, { "epoch": 0.7, "learning_rate": 1.872e-05, "loss": 6.7838, "step": 624 }, { "epoch": 0.7, "learning_rate": 1.8750000000000002e-05, "loss": 6.918, "step": 625 }, { "epoch": 0.7, "learning_rate": 1.878e-05, "loss": 6.7281, "step": 626 }, { "epoch": 0.7, "learning_rate": 1.881e-05, "loss": 6.8747, "step": 627 }, { "epoch": 0.7, "learning_rate": 1.884e-05, "loss": 6.7801, "step": 628 }, { "epoch": 0.71, "learning_rate": 1.887e-05, "loss": 6.77, "step": 629 }, { "epoch": 0.71, "learning_rate": 1.8900000000000002e-05, "loss": 6.9942, "step": 630 }, { "epoch": 0.71, "learning_rate": 1.893e-05, "loss": 6.7125, "step": 631 }, { "epoch": 0.71, "learning_rate": 1.896e-05, "loss": 6.5515, "step": 632 }, { "epoch": 0.71, "learning_rate": 1.899e-05, "loss": 6.7882, "step": 633 }, { "epoch": 0.71, "learning_rate": 1.902e-05, "loss": 6.7016, "step": 634 }, { "epoch": 0.71, "learning_rate": 1.9050000000000002e-05, "loss": 6.8185, "step": 635 }, { "epoch": 0.71, "learning_rate": 1.908e-05, "loss": 6.8425, "step": 636 }, { "epoch": 0.71, "learning_rate": 1.911e-05, "loss": 6.7852, "step": 637 }, { "epoch": 0.72, "learning_rate": 1.914e-05, "loss": 6.5962, "step": 638 }, { "epoch": 0.72, "learning_rate": 1.917e-05, "loss": 6.9415, "step": 639 }, { "epoch": 0.72, "learning_rate": 1.9200000000000003e-05, "loss": 6.6436, "step": 640 }, { "epoch": 0.72, "learning_rate": 1.923e-05, "loss": 6.727, "step": 641 }, { "epoch": 0.72, "learning_rate": 1.9260000000000002e-05, "loss": 6.9343, "step": 642 }, { "epoch": 0.72, "learning_rate": 1.929e-05, "loss": 6.7715, "step": 643 }, { "epoch": 0.72, "learning_rate": 1.932e-05, "loss": 6.5715, "step": 644 }, { "epoch": 0.72, "learning_rate": 1.935e-05, "loss": 6.6077, "step": 645 }, { "epoch": 0.72, "learning_rate": 1.938e-05, "loss": 6.7828, "step": 646 }, { "epoch": 0.73, "learning_rate": 1.9410000000000002e-05, "loss": 6.5736, "step": 647 }, { "epoch": 0.73, "learning_rate": 1.944e-05, "loss": 6.3795, "step": 648 }, { "epoch": 0.73, "learning_rate": 1.947e-05, "loss": 6.6066, "step": 649 }, { "epoch": 0.73, "learning_rate": 1.95e-05, "loss": 6.417, "step": 650 }, { "epoch": 0.73, "learning_rate": 1.953e-05, "loss": 7.1679, "step": 651 }, { "epoch": 0.73, "learning_rate": 1.9560000000000002e-05, "loss": 7.1666, "step": 652 }, { "epoch": 0.73, "learning_rate": 1.959e-05, "loss": 7.1194, "step": 653 }, { "epoch": 0.73, "learning_rate": 1.9620000000000002e-05, "loss": 6.9984, "step": 654 }, { "epoch": 0.73, "learning_rate": 1.965e-05, "loss": 7.0867, "step": 655 }, { "epoch": 0.74, "learning_rate": 1.968e-05, "loss": 6.9124, "step": 656 }, { "epoch": 0.74, "learning_rate": 1.9710000000000003e-05, "loss": 6.7249, "step": 657 }, { "epoch": 0.74, "learning_rate": 1.974e-05, "loss": 7.0205, "step": 658 }, { "epoch": 0.74, "learning_rate": 1.9770000000000002e-05, "loss": 6.7131, "step": 659 }, { "epoch": 0.74, "learning_rate": 1.98e-05, "loss": 6.7116, "step": 660 }, { "epoch": 0.74, "learning_rate": 1.983e-05, "loss": 6.675, "step": 661 }, { "epoch": 0.74, "learning_rate": 1.9860000000000003e-05, "loss": 6.7527, "step": 662 }, { "epoch": 0.74, "learning_rate": 1.989e-05, "loss": 6.8638, "step": 663 }, { "epoch": 0.74, "learning_rate": 1.9920000000000002e-05, "loss": 6.6198, "step": 664 }, { "epoch": 0.75, "learning_rate": 1.995e-05, "loss": 6.6287, "step": 665 }, { "epoch": 0.75, "learning_rate": 1.9980000000000002e-05, "loss": 6.8327, "step": 666 }, { "epoch": 0.75, "learning_rate": 2.0010000000000003e-05, "loss": 6.7063, "step": 667 }, { "epoch": 0.75, "learning_rate": 2.004e-05, "loss": 6.6294, "step": 668 }, { "epoch": 0.75, "learning_rate": 2.0070000000000003e-05, "loss": 6.8264, "step": 669 }, { "epoch": 0.75, "learning_rate": 2.01e-05, "loss": 6.7138, "step": 670 }, { "epoch": 0.75, "learning_rate": 2.0130000000000002e-05, "loss": 6.6042, "step": 671 }, { "epoch": 0.75, "learning_rate": 2.016e-05, "loss": 6.6955, "step": 672 }, { "epoch": 0.75, "learning_rate": 2.019e-05, "loss": 6.6659, "step": 673 }, { "epoch": 0.76, "learning_rate": 2.0220000000000003e-05, "loss": 6.6217, "step": 674 }, { "epoch": 0.76, "learning_rate": 2.025e-05, "loss": 6.6772, "step": 675 }, { "epoch": 0.76, "learning_rate": 2.0280000000000002e-05, "loss": 6.7739, "step": 676 }, { "epoch": 0.76, "learning_rate": 2.031e-05, "loss": 6.613, "step": 677 }, { "epoch": 0.76, "learning_rate": 2.0340000000000002e-05, "loss": 6.4539, "step": 678 }, { "epoch": 0.76, "learning_rate": 2.0370000000000003e-05, "loss": 6.7437, "step": 679 }, { "epoch": 0.76, "learning_rate": 2.04e-05, "loss": 6.6279, "step": 680 }, { "epoch": 0.76, "learning_rate": 2.0430000000000003e-05, "loss": 6.8547, "step": 681 }, { "epoch": 0.76, "learning_rate": 2.046e-05, "loss": 6.6097, "step": 682 }, { "epoch": 0.77, "learning_rate": 2.0490000000000002e-05, "loss": 6.6685, "step": 683 }, { "epoch": 0.77, "learning_rate": 2.0520000000000003e-05, "loss": 6.8237, "step": 684 }, { "epoch": 0.77, "learning_rate": 2.055e-05, "loss": 6.7639, "step": 685 }, { "epoch": 0.77, "learning_rate": 2.0580000000000003e-05, "loss": 6.5686, "step": 686 }, { "epoch": 0.77, "learning_rate": 2.061e-05, "loss": 6.6421, "step": 687 }, { "epoch": 0.77, "learning_rate": 2.064e-05, "loss": 6.8361, "step": 688 }, { "epoch": 0.77, "learning_rate": 2.067e-05, "loss": 6.6915, "step": 689 }, { "epoch": 0.77, "learning_rate": 2.07e-05, "loss": 6.7249, "step": 690 }, { "epoch": 0.77, "learning_rate": 2.073e-05, "loss": 6.6044, "step": 691 }, { "epoch": 0.78, "learning_rate": 2.0759999999999998e-05, "loss": 6.6874, "step": 692 }, { "epoch": 0.78, "learning_rate": 2.079e-05, "loss": 6.5, "step": 693 }, { "epoch": 0.78, "learning_rate": 2.082e-05, "loss": 6.669, "step": 694 }, { "epoch": 0.78, "learning_rate": 2.085e-05, "loss": 6.6466, "step": 695 }, { "epoch": 0.78, "learning_rate": 2.088e-05, "loss": 6.7548, "step": 696 }, { "epoch": 0.78, "learning_rate": 2.0909999999999998e-05, "loss": 6.515, "step": 697 }, { "epoch": 0.78, "learning_rate": 2.094e-05, "loss": 6.3181, "step": 698 }, { "epoch": 0.78, "learning_rate": 2.097e-05, "loss": 6.5157, "step": 699 }, { "epoch": 0.78, "learning_rate": 2.1e-05, "loss": 6.1912, "step": 700 }, { "epoch": 0.79, "learning_rate": 2.103e-05, "loss": 7.1464, "step": 701 }, { "epoch": 0.79, "learning_rate": 2.1059999999999998e-05, "loss": 7.1235, "step": 702 }, { "epoch": 0.79, "learning_rate": 2.109e-05, "loss": 6.9964, "step": 703 }, { "epoch": 0.79, "learning_rate": 2.1119999999999998e-05, "loss": 6.86, "step": 704 }, { "epoch": 0.79, "learning_rate": 2.115e-05, "loss": 6.8704, "step": 705 }, { "epoch": 0.79, "learning_rate": 2.118e-05, "loss": 6.883, "step": 706 }, { "epoch": 0.79, "learning_rate": 2.121e-05, "loss": 6.8516, "step": 707 }, { "epoch": 0.79, "learning_rate": 2.124e-05, "loss": 6.9634, "step": 708 }, { "epoch": 0.79, "learning_rate": 2.1269999999999998e-05, "loss": 6.7539, "step": 709 }, { "epoch": 0.8, "learning_rate": 2.13e-05, "loss": 6.6728, "step": 710 }, { "epoch": 0.8, "learning_rate": 2.133e-05, "loss": 6.6825, "step": 711 }, { "epoch": 0.8, "learning_rate": 2.136e-05, "loss": 6.577, "step": 712 }, { "epoch": 0.8, "learning_rate": 2.139e-05, "loss": 6.578, "step": 713 }, { "epoch": 0.8, "learning_rate": 2.1419999999999998e-05, "loss": 6.7443, "step": 714 }, { "epoch": 0.8, "learning_rate": 2.145e-05, "loss": 6.7634, "step": 715 }, { "epoch": 0.8, "learning_rate": 2.148e-05, "loss": 6.736, "step": 716 }, { "epoch": 0.8, "learning_rate": 2.151e-05, "loss": 6.956, "step": 717 }, { "epoch": 0.8, "learning_rate": 2.154e-05, "loss": 6.7056, "step": 718 }, { "epoch": 0.81, "learning_rate": 2.157e-05, "loss": 6.4774, "step": 719 }, { "epoch": 0.81, "learning_rate": 2.16e-05, "loss": 6.5953, "step": 720 }, { "epoch": 0.81, "learning_rate": 2.163e-05, "loss": 6.5801, "step": 721 }, { "epoch": 0.81, "learning_rate": 2.166e-05, "loss": 6.5822, "step": 722 }, { "epoch": 0.81, "learning_rate": 2.169e-05, "loss": 6.5238, "step": 723 }, { "epoch": 0.81, "learning_rate": 2.172e-05, "loss": 6.713, "step": 724 }, { "epoch": 0.81, "learning_rate": 2.175e-05, "loss": 6.7164, "step": 725 }, { "epoch": 0.81, "learning_rate": 2.178e-05, "loss": 6.7938, "step": 726 }, { "epoch": 0.82, "learning_rate": 2.181e-05, "loss": 6.6696, "step": 727 }, { "epoch": 0.82, "learning_rate": 2.184e-05, "loss": 6.5896, "step": 728 }, { "epoch": 0.82, "learning_rate": 2.187e-05, "loss": 6.5385, "step": 729 }, { "epoch": 0.82, "learning_rate": 2.19e-05, "loss": 6.6886, "step": 730 }, { "epoch": 0.82, "learning_rate": 2.193e-05, "loss": 6.854, "step": 731 }, { "epoch": 0.82, "learning_rate": 2.196e-05, "loss": 6.7223, "step": 732 }, { "epoch": 0.82, "learning_rate": 2.199e-05, "loss": 6.6452, "step": 733 }, { "epoch": 0.82, "learning_rate": 2.202e-05, "loss": 6.8424, "step": 734 }, { "epoch": 0.82, "learning_rate": 2.205e-05, "loss": 6.5514, "step": 735 }, { "epoch": 0.83, "learning_rate": 2.208e-05, "loss": 6.7269, "step": 736 }, { "epoch": 0.83, "learning_rate": 2.211e-05, "loss": 6.6487, "step": 737 }, { "epoch": 0.83, "learning_rate": 2.214e-05, "loss": 6.6926, "step": 738 }, { "epoch": 0.83, "learning_rate": 2.217e-05, "loss": 6.5576, "step": 739 }, { "epoch": 0.83, "learning_rate": 2.22e-05, "loss": 6.6876, "step": 740 }, { "epoch": 0.83, "learning_rate": 2.223e-05, "loss": 6.7458, "step": 741 }, { "epoch": 0.83, "learning_rate": 2.226e-05, "loss": 6.6989, "step": 742 }, { "epoch": 0.83, "learning_rate": 2.2290000000000002e-05, "loss": 6.5419, "step": 743 }, { "epoch": 0.83, "learning_rate": 2.232e-05, "loss": 6.6296, "step": 744 }, { "epoch": 0.84, "learning_rate": 2.235e-05, "loss": 6.5066, "step": 745 }, { "epoch": 0.84, "learning_rate": 2.238e-05, "loss": 6.7757, "step": 746 }, { "epoch": 0.84, "learning_rate": 2.241e-05, "loss": 6.4543, "step": 747 }, { "epoch": 0.84, "learning_rate": 2.2440000000000002e-05, "loss": 6.4333, "step": 748 }, { "epoch": 0.84, "learning_rate": 2.247e-05, "loss": 6.4532, "step": 749 }, { "epoch": 0.84, "learning_rate": 2.25e-05, "loss": 6.2844, "step": 750 }, { "epoch": 0.84, "learning_rate": 2.253e-05, "loss": 7.0454, "step": 751 }, { "epoch": 0.84, "learning_rate": 2.256e-05, "loss": 7.1118, "step": 752 }, { "epoch": 0.84, "learning_rate": 2.2590000000000002e-05, "loss": 7.0478, "step": 753 }, { "epoch": 0.85, "learning_rate": 2.262e-05, "loss": 6.8659, "step": 754 }, { "epoch": 0.85, "learning_rate": 2.265e-05, "loss": 6.7425, "step": 755 }, { "epoch": 0.85, "learning_rate": 2.268e-05, "loss": 6.8628, "step": 756 }, { "epoch": 0.85, "learning_rate": 2.271e-05, "loss": 6.7184, "step": 757 }, { "epoch": 0.85, "learning_rate": 2.274e-05, "loss": 6.6364, "step": 758 }, { "epoch": 0.85, "learning_rate": 2.277e-05, "loss": 6.6891, "step": 759 }, { "epoch": 0.85, "learning_rate": 2.2800000000000002e-05, "loss": 6.6402, "step": 760 }, { "epoch": 0.85, "learning_rate": 2.283e-05, "loss": 6.6746, "step": 761 }, { "epoch": 0.85, "learning_rate": 2.286e-05, "loss": 6.6051, "step": 762 }, { "epoch": 0.86, "learning_rate": 2.289e-05, "loss": 6.4717, "step": 763 }, { "epoch": 0.86, "learning_rate": 2.292e-05, "loss": 6.4113, "step": 764 }, { "epoch": 0.86, "learning_rate": 2.2950000000000002e-05, "loss": 6.6297, "step": 765 }, { "epoch": 0.86, "learning_rate": 2.298e-05, "loss": 6.5547, "step": 766 }, { "epoch": 0.86, "learning_rate": 2.301e-05, "loss": 6.5759, "step": 767 }, { "epoch": 0.86, "learning_rate": 2.304e-05, "loss": 6.5464, "step": 768 }, { "epoch": 0.86, "learning_rate": 2.307e-05, "loss": 6.6716, "step": 769 }, { "epoch": 0.86, "learning_rate": 2.3100000000000002e-05, "loss": 6.531, "step": 770 }, { "epoch": 0.86, "learning_rate": 2.313e-05, "loss": 6.7244, "step": 771 }, { "epoch": 0.87, "learning_rate": 2.3160000000000002e-05, "loss": 6.7236, "step": 772 }, { "epoch": 0.87, "learning_rate": 2.319e-05, "loss": 6.6473, "step": 773 }, { "epoch": 0.87, "learning_rate": 2.322e-05, "loss": 6.4939, "step": 774 }, { "epoch": 0.87, "learning_rate": 2.3250000000000003e-05, "loss": 6.3336, "step": 775 }, { "epoch": 0.87, "learning_rate": 2.328e-05, "loss": 6.4583, "step": 776 }, { "epoch": 0.87, "learning_rate": 2.3310000000000002e-05, "loss": 6.4127, "step": 777 }, { "epoch": 0.87, "learning_rate": 2.334e-05, "loss": 6.6659, "step": 778 }, { "epoch": 0.87, "learning_rate": 2.337e-05, "loss": 6.5701, "step": 779 }, { "epoch": 0.87, "learning_rate": 2.3400000000000003e-05, "loss": 6.5708, "step": 780 }, { "epoch": 0.88, "learning_rate": 2.343e-05, "loss": 6.5875, "step": 781 }, { "epoch": 0.88, "learning_rate": 2.3460000000000002e-05, "loss": 6.5154, "step": 782 }, { "epoch": 0.88, "learning_rate": 2.349e-05, "loss": 6.6474, "step": 783 }, { "epoch": 0.88, "learning_rate": 2.3520000000000002e-05, "loss": 6.4704, "step": 784 }, { "epoch": 0.88, "learning_rate": 2.3550000000000003e-05, "loss": 6.5953, "step": 785 }, { "epoch": 0.88, "learning_rate": 2.358e-05, "loss": 6.5129, "step": 786 }, { "epoch": 0.88, "learning_rate": 2.3610000000000003e-05, "loss": 6.6226, "step": 787 }, { "epoch": 0.88, "learning_rate": 2.364e-05, "loss": 6.6121, "step": 788 }, { "epoch": 0.88, "learning_rate": 2.3670000000000002e-05, "loss": 6.5475, "step": 789 }, { "epoch": 0.89, "learning_rate": 2.37e-05, "loss": 6.5039, "step": 790 }, { "epoch": 0.89, "learning_rate": 2.373e-05, "loss": 6.5055, "step": 791 }, { "epoch": 0.89, "learning_rate": 2.3760000000000003e-05, "loss": 6.3689, "step": 792 }, { "epoch": 0.89, "learning_rate": 2.379e-05, "loss": 6.5519, "step": 793 }, { "epoch": 0.89, "learning_rate": 2.3820000000000002e-05, "loss": 6.4651, "step": 794 }, { "epoch": 0.89, "learning_rate": 2.385e-05, "loss": 6.5821, "step": 795 }, { "epoch": 0.89, "learning_rate": 2.3880000000000002e-05, "loss": 6.4667, "step": 796 }, { "epoch": 0.89, "learning_rate": 2.3910000000000003e-05, "loss": 6.4944, "step": 797 }, { "epoch": 0.89, "learning_rate": 2.394e-05, "loss": 6.4661, "step": 798 }, { "epoch": 0.9, "learning_rate": 2.3970000000000003e-05, "loss": 6.7376, "step": 799 }, { "epoch": 0.9, "learning_rate": 2.4e-05, "loss": 6.1727, "step": 800 }, { "epoch": 0.9, "learning_rate": 2.4030000000000002e-05, "loss": 7.0353, "step": 801 }, { "epoch": 0.9, "learning_rate": 2.4060000000000003e-05, "loss": 6.8983, "step": 802 }, { "epoch": 0.9, "learning_rate": 2.409e-05, "loss": 7.1156, "step": 803 }, { "epoch": 0.9, "learning_rate": 2.4120000000000003e-05, "loss": 6.9282, "step": 804 }, { "epoch": 0.9, "learning_rate": 2.415e-05, "loss": 6.6145, "step": 805 }, { "epoch": 0.9, "learning_rate": 2.4180000000000002e-05, "loss": 6.7737, "step": 806 }, { "epoch": 0.9, "learning_rate": 2.4210000000000004e-05, "loss": 6.727, "step": 807 }, { "epoch": 0.91, "learning_rate": 2.4240000000000002e-05, "loss": 6.4697, "step": 808 }, { "epoch": 0.91, "learning_rate": 2.4270000000000003e-05, "loss": 6.6108, "step": 809 }, { "epoch": 0.91, "learning_rate": 2.43e-05, "loss": 6.5582, "step": 810 }, { "epoch": 0.91, "learning_rate": 2.4330000000000003e-05, "loss": 6.66, "step": 811 }, { "epoch": 0.91, "learning_rate": 2.4360000000000004e-05, "loss": 6.6864, "step": 812 }, { "epoch": 0.91, "learning_rate": 2.439e-05, "loss": 6.4978, "step": 813 }, { "epoch": 0.91, "learning_rate": 2.442e-05, "loss": 6.5428, "step": 814 }, { "epoch": 0.91, "learning_rate": 2.4449999999999998e-05, "loss": 6.5548, "step": 815 }, { "epoch": 0.91, "learning_rate": 2.448e-05, "loss": 6.4841, "step": 816 }, { "epoch": 0.92, "learning_rate": 2.4509999999999997e-05, "loss": 6.6387, "step": 817 }, { "epoch": 0.92, "learning_rate": 2.454e-05, "loss": 6.5544, "step": 818 }, { "epoch": 0.92, "learning_rate": 2.457e-05, "loss": 6.4226, "step": 819 }, { "epoch": 0.92, "learning_rate": 2.4599999999999998e-05, "loss": 6.6169, "step": 820 }, { "epoch": 0.92, "learning_rate": 2.463e-05, "loss": 6.5977, "step": 821 }, { "epoch": 0.92, "learning_rate": 2.4659999999999998e-05, "loss": 6.4566, "step": 822 }, { "epoch": 0.92, "learning_rate": 2.469e-05, "loss": 6.6328, "step": 823 }, { "epoch": 0.92, "learning_rate": 2.472e-05, "loss": 6.5394, "step": 824 }, { "epoch": 0.92, "learning_rate": 2.475e-05, "loss": 6.7847, "step": 825 }, { "epoch": 0.93, "learning_rate": 2.478e-05, "loss": 6.6761, "step": 826 }, { "epoch": 0.93, "learning_rate": 2.4809999999999998e-05, "loss": 6.5714, "step": 827 }, { "epoch": 0.93, "learning_rate": 2.484e-05, "loss": 6.3966, "step": 828 }, { "epoch": 0.93, "learning_rate": 2.487e-05, "loss": 6.4798, "step": 829 }, { "epoch": 0.93, "learning_rate": 2.49e-05, "loss": 6.5346, "step": 830 }, { "epoch": 0.93, "learning_rate": 2.493e-05, "loss": 6.4588, "step": 831 }, { "epoch": 0.93, "learning_rate": 2.4959999999999998e-05, "loss": 6.4755, "step": 832 }, { "epoch": 0.93, "learning_rate": 2.499e-05, "loss": 6.5371, "step": 833 }, { "epoch": 0.93, "learning_rate": 2.502e-05, "loss": 6.4751, "step": 834 }, { "epoch": 0.94, "learning_rate": 2.505e-05, "loss": 6.5485, "step": 835 }, { "epoch": 0.94, "learning_rate": 2.508e-05, "loss": 6.6267, "step": 836 }, { "epoch": 0.94, "learning_rate": 2.511e-05, "loss": 6.5094, "step": 837 }, { "epoch": 0.94, "learning_rate": 2.514e-05, "loss": 6.5696, "step": 838 }, { "epoch": 0.94, "learning_rate": 2.517e-05, "loss": 6.3804, "step": 839 }, { "epoch": 0.94, "learning_rate": 2.52e-05, "loss": 6.3738, "step": 840 }, { "epoch": 0.94, "learning_rate": 2.523e-05, "loss": 6.7085, "step": 841 }, { "epoch": 0.94, "learning_rate": 2.526e-05, "loss": 6.4735, "step": 842 }, { "epoch": 0.95, "learning_rate": 2.529e-05, "loss": 6.7461, "step": 843 }, { "epoch": 0.95, "learning_rate": 2.5319999999999998e-05, "loss": 6.4815, "step": 844 }, { "epoch": 0.95, "learning_rate": 2.535e-05, "loss": 6.5509, "step": 845 }, { "epoch": 0.95, "learning_rate": 2.538e-05, "loss": 6.4504, "step": 846 }, { "epoch": 0.95, "learning_rate": 2.541e-05, "loss": 6.2539, "step": 847 }, { "epoch": 0.95, "learning_rate": 2.544e-05, "loss": 6.5468, "step": 848 }, { "epoch": 0.95, "learning_rate": 2.547e-05, "loss": 6.3551, "step": 849 }, { "epoch": 0.95, "learning_rate": 2.55e-05, "loss": 5.9837, "step": 850 }, { "epoch": 0.95, "learning_rate": 2.553e-05, "loss": 6.9341, "step": 851 }, { "epoch": 0.96, "learning_rate": 2.556e-05, "loss": 6.9014, "step": 852 }, { "epoch": 0.96, "learning_rate": 2.559e-05, "loss": 7.0729, "step": 853 }, { "epoch": 0.96, "learning_rate": 2.562e-05, "loss": 6.9696, "step": 854 }, { "epoch": 0.96, "learning_rate": 2.565e-05, "loss": 6.7086, "step": 855 }, { "epoch": 0.96, "learning_rate": 2.568e-05, "loss": 6.8364, "step": 856 }, { "epoch": 0.96, "learning_rate": 2.571e-05, "loss": 6.7848, "step": 857 }, { "epoch": 0.96, "learning_rate": 2.574e-05, "loss": 6.6633, "step": 858 }, { "epoch": 0.96, "learning_rate": 2.577e-05, "loss": 6.4253, "step": 859 }, { "epoch": 0.96, "learning_rate": 2.58e-05, "loss": 6.5386, "step": 860 }, { "epoch": 0.97, "learning_rate": 2.5830000000000002e-05, "loss": 6.7223, "step": 861 }, { "epoch": 0.97, "learning_rate": 2.586e-05, "loss": 6.6383, "step": 862 }, { "epoch": 0.97, "learning_rate": 2.589e-05, "loss": 6.5371, "step": 863 }, { "epoch": 0.97, "learning_rate": 2.592e-05, "loss": 6.7278, "step": 864 }, { "epoch": 0.97, "learning_rate": 2.595e-05, "loss": 6.6923, "step": 865 }, { "epoch": 0.97, "learning_rate": 2.5980000000000002e-05, "loss": 6.4889, "step": 866 }, { "epoch": 0.97, "learning_rate": 2.601e-05, "loss": 6.6291, "step": 867 }, { "epoch": 0.97, "learning_rate": 2.604e-05, "loss": 6.4489, "step": 868 }, { "epoch": 0.97, "learning_rate": 2.607e-05, "loss": 6.5719, "step": 869 }, { "epoch": 0.98, "learning_rate": 2.61e-05, "loss": 6.4931, "step": 870 }, { "epoch": 0.98, "learning_rate": 2.6130000000000002e-05, "loss": 6.4038, "step": 871 }, { "epoch": 0.98, "learning_rate": 2.616e-05, "loss": 6.5443, "step": 872 }, { "epoch": 0.98, "learning_rate": 2.619e-05, "loss": 6.5067, "step": 873 }, { "epoch": 0.98, "learning_rate": 2.622e-05, "loss": 6.5629, "step": 874 }, { "epoch": 0.98, "learning_rate": 2.625e-05, "loss": 6.6661, "step": 875 }, { "epoch": 0.98, "learning_rate": 2.628e-05, "loss": 6.3605, "step": 876 }, { "epoch": 0.98, "learning_rate": 2.631e-05, "loss": 6.3728, "step": 877 }, { "epoch": 0.98, "learning_rate": 2.6340000000000002e-05, "loss": 6.4918, "step": 878 }, { "epoch": 0.99, "learning_rate": 2.637e-05, "loss": 6.4699, "step": 879 }, { "epoch": 0.99, "learning_rate": 2.64e-05, "loss": 6.68, "step": 880 }, { "epoch": 0.99, "learning_rate": 2.643e-05, "loss": 6.2607, "step": 881 }, { "epoch": 0.99, "learning_rate": 2.646e-05, "loss": 6.43, "step": 882 }, { "epoch": 0.99, "learning_rate": 2.6490000000000002e-05, "loss": 6.3461, "step": 883 }, { "epoch": 0.99, "learning_rate": 2.652e-05, "loss": 6.5306, "step": 884 }, { "epoch": 0.99, "learning_rate": 2.655e-05, "loss": 6.5091, "step": 885 }, { "epoch": 0.99, "learning_rate": 2.658e-05, "loss": 6.5151, "step": 886 }, { "epoch": 0.99, "learning_rate": 2.661e-05, "loss": 6.2351, "step": 887 }, { "epoch": 1.0, "learning_rate": 2.6640000000000002e-05, "loss": 6.4284, "step": 888 }, { "epoch": 1.0, "learning_rate": 2.667e-05, "loss": 6.421, "step": 889 }, { "epoch": 1.0, "learning_rate": 2.6700000000000002e-05, "loss": 6.1901, "step": 890 }, { "epoch": 1.0, "learning_rate": 2.673e-05, "loss": 6.159, "step": 891 }, { "epoch": 1.0, "learning_rate": 2.676e-05, "loss": 5.8712, "step": 892 }, { "epoch": 1.0, "learning_rate": 2.6790000000000003e-05, "loss": 6.9582, "step": 893 }, { "epoch": 1.0, "learning_rate": 2.682e-05, "loss": 7.0123, "step": 894 }, { "epoch": 1.0, "learning_rate": 2.6850000000000002e-05, "loss": 6.6138, "step": 895 }, { "epoch": 1.0, "learning_rate": 2.688e-05, "loss": 6.6295, "step": 896 }, { "epoch": 1.01, "learning_rate": 2.691e-05, "loss": 6.5914, "step": 897 }, { "epoch": 1.01, "learning_rate": 2.6940000000000003e-05, "loss": 6.6843, "step": 898 }, { "epoch": 1.01, "learning_rate": 2.697e-05, "loss": 6.559, "step": 899 }, { "epoch": 1.01, "learning_rate": 2.7000000000000002e-05, "loss": 6.5964, "step": 900 }, { "epoch": 1.01, "learning_rate": 2.703e-05, "loss": 6.6301, "step": 901 }, { "epoch": 1.01, "learning_rate": 2.7060000000000002e-05, "loss": 6.5901, "step": 902 }, { "epoch": 1.01, "learning_rate": 2.709e-05, "loss": 6.2806, "step": 903 }, { "epoch": 1.01, "learning_rate": 2.712e-05, "loss": 6.5409, "step": 904 }, { "epoch": 1.01, "learning_rate": 2.7150000000000003e-05, "loss": 6.3442, "step": 905 }, { "epoch": 1.02, "learning_rate": 2.718e-05, "loss": 6.4226, "step": 906 }, { "epoch": 1.02, "learning_rate": 2.7210000000000002e-05, "loss": 6.6101, "step": 907 }, { "epoch": 1.02, "learning_rate": 2.724e-05, "loss": 6.6256, "step": 908 }, { "epoch": 1.02, "learning_rate": 2.727e-05, "loss": 6.467, "step": 909 }, { "epoch": 1.02, "learning_rate": 2.7300000000000003e-05, "loss": 6.2838, "step": 910 }, { "epoch": 1.02, "learning_rate": 2.733e-05, "loss": 6.3838, "step": 911 }, { "epoch": 1.02, "learning_rate": 2.7360000000000002e-05, "loss": 6.4119, "step": 912 }, { "epoch": 1.02, "learning_rate": 2.739e-05, "loss": 6.412, "step": 913 }, { "epoch": 1.02, "learning_rate": 2.7420000000000002e-05, "loss": 6.4896, "step": 914 }, { "epoch": 1.03, "learning_rate": 2.7450000000000003e-05, "loss": 6.2475, "step": 915 }, { "epoch": 1.03, "learning_rate": 2.748e-05, "loss": 6.4396, "step": 916 }, { "epoch": 1.03, "learning_rate": 2.7510000000000003e-05, "loss": 6.2611, "step": 917 }, { "epoch": 1.03, "learning_rate": 2.754e-05, "loss": 6.4782, "step": 918 }, { "epoch": 1.03, "learning_rate": 2.7570000000000002e-05, "loss": 6.4861, "step": 919 }, { "epoch": 1.03, "learning_rate": 2.7600000000000003e-05, "loss": 6.3341, "step": 920 }, { "epoch": 1.03, "learning_rate": 2.763e-05, "loss": 6.3595, "step": 921 }, { "epoch": 1.03, "learning_rate": 2.7660000000000003e-05, "loss": 6.6259, "step": 922 }, { "epoch": 1.03, "learning_rate": 2.769e-05, "loss": 6.4195, "step": 923 }, { "epoch": 1.04, "learning_rate": 2.7720000000000002e-05, "loss": 6.4816, "step": 924 }, { "epoch": 1.04, "learning_rate": 2.7750000000000004e-05, "loss": 6.3767, "step": 925 }, { "epoch": 1.04, "learning_rate": 2.778e-05, "loss": 6.4373, "step": 926 }, { "epoch": 1.04, "learning_rate": 2.7810000000000003e-05, "loss": 6.438, "step": 927 }, { "epoch": 1.04, "learning_rate": 2.784e-05, "loss": 6.3644, "step": 928 }, { "epoch": 1.04, "learning_rate": 2.7870000000000003e-05, "loss": 6.4318, "step": 929 }, { "epoch": 1.04, "learning_rate": 2.79e-05, "loss": 6.5832, "step": 930 }, { "epoch": 1.04, "learning_rate": 2.7930000000000002e-05, "loss": 6.363, "step": 931 }, { "epoch": 1.04, "learning_rate": 2.7960000000000003e-05, "loss": 6.2379, "step": 932 }, { "epoch": 1.05, "learning_rate": 2.799e-05, "loss": 6.3931, "step": 933 }, { "epoch": 1.05, "learning_rate": 2.8020000000000003e-05, "loss": 6.5303, "step": 934 }, { "epoch": 1.05, "learning_rate": 2.805e-05, "loss": 6.2901, "step": 935 }, { "epoch": 1.05, "learning_rate": 2.8080000000000002e-05, "loss": 6.2308, "step": 936 }, { "epoch": 1.05, "learning_rate": 2.8110000000000004e-05, "loss": 6.411, "step": 937 }, { "epoch": 1.05, "learning_rate": 2.8139999999999998e-05, "loss": 6.1754, "step": 938 }, { "epoch": 1.05, "learning_rate": 2.817e-05, "loss": 6.0475, "step": 939 }, { "epoch": 1.05, "learning_rate": 2.8199999999999998e-05, "loss": 6.3637, "step": 940 }, { "epoch": 1.05, "learning_rate": 2.823e-05, "loss": 6.0, "step": 941 }, { "epoch": 1.06, "learning_rate": 2.826e-05, "loss": 6.3181, "step": 942 }, { "epoch": 1.06, "learning_rate": 2.829e-05, "loss": 6.8344, "step": 943 }, { "epoch": 1.06, "learning_rate": 2.832e-05, "loss": 6.7312, "step": 944 }, { "epoch": 1.06, "learning_rate": 2.8349999999999998e-05, "loss": 6.9842, "step": 945 }, { "epoch": 1.06, "learning_rate": 2.838e-05, "loss": 6.7705, "step": 946 }, { "epoch": 1.06, "learning_rate": 2.841e-05, "loss": 6.8939, "step": 947 }, { "epoch": 1.06, "learning_rate": 2.844e-05, "loss": 6.8511, "step": 948 }, { "epoch": 1.06, "learning_rate": 2.847e-05, "loss": 6.6316, "step": 949 }, { "epoch": 1.07, "learning_rate": 2.8499999999999998e-05, "loss": 6.6692, "step": 950 }, { "epoch": 1.07, "learning_rate": 2.853e-05, "loss": 6.2612, "step": 951 }, { "epoch": 1.07, "learning_rate": 2.856e-05, "loss": 6.5524, "step": 952 }, { "epoch": 1.07, "learning_rate": 2.859e-05, "loss": 6.5752, "step": 953 }, { "epoch": 1.07, "learning_rate": 2.862e-05, "loss": 6.3672, "step": 954 }, { "epoch": 1.07, "learning_rate": 2.865e-05, "loss": 6.6167, "step": 955 }, { "epoch": 1.07, "learning_rate": 2.868e-05, "loss": 6.5005, "step": 956 }, { "epoch": 1.07, "learning_rate": 2.871e-05, "loss": 6.4512, "step": 957 }, { "epoch": 1.07, "learning_rate": 2.874e-05, "loss": 6.4692, "step": 958 }, { "epoch": 1.08, "learning_rate": 2.877e-05, "loss": 6.5474, "step": 959 }, { "epoch": 1.08, "learning_rate": 2.88e-05, "loss": 6.3747, "step": 960 }, { "epoch": 1.08, "learning_rate": 2.883e-05, "loss": 6.3979, "step": 961 }, { "epoch": 1.08, "learning_rate": 2.8859999999999998e-05, "loss": 6.5424, "step": 962 }, { "epoch": 1.08, "learning_rate": 2.889e-05, "loss": 6.5999, "step": 963 }, { "epoch": 1.08, "learning_rate": 2.892e-05, "loss": 6.4115, "step": 964 }, { "epoch": 1.08, "learning_rate": 2.895e-05, "loss": 6.3567, "step": 965 }, { "epoch": 1.08, "learning_rate": 2.898e-05, "loss": 6.4201, "step": 966 }, { "epoch": 1.08, "learning_rate": 2.901e-05, "loss": 6.3718, "step": 967 }, { "epoch": 1.09, "learning_rate": 2.904e-05, "loss": 6.2852, "step": 968 }, { "epoch": 1.09, "learning_rate": 2.907e-05, "loss": 6.5303, "step": 969 }, { "epoch": 1.09, "learning_rate": 2.91e-05, "loss": 6.3467, "step": 970 }, { "epoch": 1.09, "learning_rate": 2.913e-05, "loss": 6.3755, "step": 971 }, { "epoch": 1.09, "learning_rate": 2.916e-05, "loss": 6.3919, "step": 972 }, { "epoch": 1.09, "learning_rate": 2.919e-05, "loss": 6.5368, "step": 973 }, { "epoch": 1.09, "learning_rate": 2.922e-05, "loss": 6.5004, "step": 974 }, { "epoch": 1.09, "learning_rate": 2.925e-05, "loss": 6.3812, "step": 975 }, { "epoch": 1.09, "learning_rate": 2.928e-05, "loss": 6.3642, "step": 976 }, { "epoch": 1.1, "learning_rate": 2.931e-05, "loss": 6.5603, "step": 977 }, { "epoch": 1.1, "learning_rate": 2.934e-05, "loss": 6.3053, "step": 978 }, { "epoch": 1.1, "learning_rate": 2.9370000000000002e-05, "loss": 6.4027, "step": 979 }, { "epoch": 1.1, "learning_rate": 2.94e-05, "loss": 6.3164, "step": 980 }, { "epoch": 1.1, "learning_rate": 2.943e-05, "loss": 6.2913, "step": 981 }, { "epoch": 1.1, "learning_rate": 2.946e-05, "loss": 6.5012, "step": 982 }, { "epoch": 1.1, "learning_rate": 2.949e-05, "loss": 6.6924, "step": 983 }, { "epoch": 1.1, "learning_rate": 2.9520000000000002e-05, "loss": 6.3478, "step": 984 }, { "epoch": 1.1, "learning_rate": 2.955e-05, "loss": 6.2374, "step": 985 }, { "epoch": 1.11, "learning_rate": 2.958e-05, "loss": 6.3753, "step": 986 }, { "epoch": 1.11, "learning_rate": 2.961e-05, "loss": 6.2503, "step": 987 }, { "epoch": 1.11, "learning_rate": 2.964e-05, "loss": 6.1337, "step": 988 }, { "epoch": 1.11, "learning_rate": 2.967e-05, "loss": 6.0494, "step": 989 }, { "epoch": 1.11, "learning_rate": 2.97e-05, "loss": 6.4621, "step": 990 }, { "epoch": 1.11, "learning_rate": 2.973e-05, "loss": 5.8124, "step": 991 }, { "epoch": 1.11, "learning_rate": 2.976e-05, "loss": 5.7165, "step": 992 }, { "epoch": 1.11, "learning_rate": 2.979e-05, "loss": 6.6435, "step": 993 }, { "epoch": 1.11, "learning_rate": 2.982e-05, "loss": 6.7515, "step": 994 }, { "epoch": 1.12, "learning_rate": 2.985e-05, "loss": 6.72, "step": 995 }, { "epoch": 1.12, "learning_rate": 2.9880000000000002e-05, "loss": 6.7765, "step": 996 }, { "epoch": 1.12, "learning_rate": 2.991e-05, "loss": 6.5132, "step": 997 }, { "epoch": 1.12, "learning_rate": 2.994e-05, "loss": 6.7084, "step": 998 }, { "epoch": 1.12, "learning_rate": 2.997e-05, "loss": 6.5142, "step": 999 }, { "epoch": 1.12, "learning_rate": 3e-05, "loss": 6.2787, "step": 1000 }, { "epoch": 1.12, "learning_rate": 2.9998591549295774e-05, "loss": 6.4155, "step": 1001 }, { "epoch": 1.12, "learning_rate": 2.999718309859155e-05, "loss": 6.515, "step": 1002 }, { "epoch": 1.12, "learning_rate": 2.9995774647887323e-05, "loss": 6.3912, "step": 1003 }, { "epoch": 1.13, "learning_rate": 2.99943661971831e-05, "loss": 6.5432, "step": 1004 }, { "epoch": 1.13, "learning_rate": 2.9992957746478872e-05, "loss": 6.5534, "step": 1005 }, { "epoch": 1.13, "learning_rate": 2.999154929577465e-05, "loss": 6.3818, "step": 1006 }, { "epoch": 1.13, "learning_rate": 2.999014084507042e-05, "loss": 6.3068, "step": 1007 }, { "epoch": 1.13, "learning_rate": 2.9988732394366198e-05, "loss": 6.3789, "step": 1008 }, { "epoch": 1.13, "learning_rate": 2.9987323943661974e-05, "loss": 6.3886, "step": 1009 }, { "epoch": 1.13, "learning_rate": 2.9985915492957747e-05, "loss": 6.3875, "step": 1010 }, { "epoch": 1.13, "learning_rate": 2.9984507042253523e-05, "loss": 6.3901, "step": 1011 }, { "epoch": 1.13, "learning_rate": 2.9983098591549296e-05, "loss": 6.5045, "step": 1012 }, { "epoch": 1.14, "learning_rate": 2.9981690140845073e-05, "loss": 6.2532, "step": 1013 }, { "epoch": 1.14, "learning_rate": 2.9980281690140846e-05, "loss": 6.3673, "step": 1014 }, { "epoch": 1.14, "learning_rate": 2.9978873239436622e-05, "loss": 6.2703, "step": 1015 }, { "epoch": 1.14, "learning_rate": 2.9977464788732395e-05, "loss": 6.3101, "step": 1016 }, { "epoch": 1.14, "learning_rate": 2.9976056338028168e-05, "loss": 6.2903, "step": 1017 }, { "epoch": 1.14, "learning_rate": 2.9974647887323944e-05, "loss": 6.4326, "step": 1018 }, { "epoch": 1.14, "learning_rate": 2.9973239436619717e-05, "loss": 6.3883, "step": 1019 }, { "epoch": 1.14, "learning_rate": 2.9971830985915494e-05, "loss": 6.5243, "step": 1020 }, { "epoch": 1.14, "learning_rate": 2.9970422535211266e-05, "loss": 6.34, "step": 1021 }, { "epoch": 1.15, "learning_rate": 2.9969014084507043e-05, "loss": 6.2807, "step": 1022 }, { "epoch": 1.15, "learning_rate": 2.996760563380282e-05, "loss": 6.3336, "step": 1023 }, { "epoch": 1.15, "learning_rate": 2.9966197183098592e-05, "loss": 6.4147, "step": 1024 }, { "epoch": 1.15, "learning_rate": 2.996478873239437e-05, "loss": 6.1611, "step": 1025 }, { "epoch": 1.15, "learning_rate": 2.996338028169014e-05, "loss": 6.237, "step": 1026 }, { "epoch": 1.15, "learning_rate": 2.9961971830985918e-05, "loss": 6.5098, "step": 1027 }, { "epoch": 1.15, "learning_rate": 2.996056338028169e-05, "loss": 6.2212, "step": 1028 }, { "epoch": 1.15, "learning_rate": 2.9959154929577467e-05, "loss": 6.1175, "step": 1029 }, { "epoch": 1.15, "learning_rate": 2.995774647887324e-05, "loss": 6.4027, "step": 1030 }, { "epoch": 1.16, "learning_rate": 2.9956338028169016e-05, "loss": 6.142, "step": 1031 }, { "epoch": 1.16, "learning_rate": 2.995492957746479e-05, "loss": 6.2257, "step": 1032 }, { "epoch": 1.16, "learning_rate": 2.9953521126760562e-05, "loss": 6.2887, "step": 1033 }, { "epoch": 1.16, "learning_rate": 2.995211267605634e-05, "loss": 6.2943, "step": 1034 }, { "epoch": 1.16, "learning_rate": 2.995070422535211e-05, "loss": 6.4215, "step": 1035 }, { "epoch": 1.16, "learning_rate": 2.9949295774647888e-05, "loss": 6.4424, "step": 1036 }, { "epoch": 1.16, "learning_rate": 2.994788732394366e-05, "loss": 6.2669, "step": 1037 }, { "epoch": 1.16, "learning_rate": 2.994647887323944e-05, "loss": 6.3936, "step": 1038 }, { "epoch": 1.16, "learning_rate": 2.9945070422535213e-05, "loss": 6.1754, "step": 1039 }, { "epoch": 1.17, "learning_rate": 2.9943661971830986e-05, "loss": 6.1602, "step": 1040 }, { "epoch": 1.17, "learning_rate": 2.9942253521126763e-05, "loss": 6.0637, "step": 1041 }, { "epoch": 1.17, "learning_rate": 2.9940845070422536e-05, "loss": 5.8476, "step": 1042 }, { "epoch": 1.17, "learning_rate": 2.9939436619718312e-05, "loss": 6.6751, "step": 1043 }, { "epoch": 1.17, "learning_rate": 2.9938028169014085e-05, "loss": 6.5956, "step": 1044 }, { "epoch": 1.17, "learning_rate": 2.993661971830986e-05, "loss": 6.6727, "step": 1045 }, { "epoch": 1.17, "learning_rate": 2.9935211267605634e-05, "loss": 6.6289, "step": 1046 }, { "epoch": 1.17, "learning_rate": 2.993380281690141e-05, "loss": 6.6454, "step": 1047 }, { "epoch": 1.17, "learning_rate": 2.9932394366197183e-05, "loss": 6.4628, "step": 1048 }, { "epoch": 1.18, "learning_rate": 2.9930985915492956e-05, "loss": 6.4952, "step": 1049 }, { "epoch": 1.18, "learning_rate": 2.9929577464788733e-05, "loss": 6.3537, "step": 1050 }, { "epoch": 1.18, "learning_rate": 2.9928169014084506e-05, "loss": 6.368, "step": 1051 }, { "epoch": 1.18, "learning_rate": 2.9926760563380282e-05, "loss": 6.5519, "step": 1052 }, { "epoch": 1.18, "learning_rate": 2.9925352112676058e-05, "loss": 6.3569, "step": 1053 }, { "epoch": 1.18, "learning_rate": 2.9923943661971835e-05, "loss": 6.5364, "step": 1054 }, { "epoch": 1.18, "learning_rate": 2.9922535211267608e-05, "loss": 6.303, "step": 1055 }, { "epoch": 1.18, "learning_rate": 2.992112676056338e-05, "loss": 6.3067, "step": 1056 }, { "epoch": 1.18, "learning_rate": 2.9919718309859157e-05, "loss": 6.4836, "step": 1057 }, { "epoch": 1.19, "learning_rate": 2.991830985915493e-05, "loss": 6.5024, "step": 1058 }, { "epoch": 1.19, "learning_rate": 2.9916901408450706e-05, "loss": 6.3039, "step": 1059 }, { "epoch": 1.19, "learning_rate": 2.991549295774648e-05, "loss": 6.2701, "step": 1060 }, { "epoch": 1.19, "learning_rate": 2.9914084507042255e-05, "loss": 6.385, "step": 1061 }, { "epoch": 1.19, "learning_rate": 2.9912676056338028e-05, "loss": 6.3375, "step": 1062 }, { "epoch": 1.19, "learning_rate": 2.9911267605633805e-05, "loss": 6.2609, "step": 1063 }, { "epoch": 1.19, "learning_rate": 2.9909859154929578e-05, "loss": 6.3331, "step": 1064 }, { "epoch": 1.19, "learning_rate": 2.990845070422535e-05, "loss": 6.4065, "step": 1065 }, { "epoch": 1.2, "learning_rate": 2.9907042253521127e-05, "loss": 6.3662, "step": 1066 }, { "epoch": 1.2, "learning_rate": 2.99056338028169e-05, "loss": 6.1809, "step": 1067 }, { "epoch": 1.2, "learning_rate": 2.9904225352112676e-05, "loss": 6.1951, "step": 1068 }, { "epoch": 1.2, "learning_rate": 2.9902816901408452e-05, "loss": 6.5026, "step": 1069 }, { "epoch": 1.2, "learning_rate": 2.990140845070423e-05, "loss": 6.3484, "step": 1070 }, { "epoch": 1.2, "learning_rate": 2.9900000000000002e-05, "loss": 6.3834, "step": 1071 }, { "epoch": 1.2, "learning_rate": 2.9898591549295775e-05, "loss": 6.4122, "step": 1072 }, { "epoch": 1.2, "learning_rate": 2.989718309859155e-05, "loss": 6.3136, "step": 1073 }, { "epoch": 1.2, "learning_rate": 2.9895774647887324e-05, "loss": 6.1785, "step": 1074 }, { "epoch": 1.21, "learning_rate": 2.98943661971831e-05, "loss": 6.2038, "step": 1075 }, { "epoch": 1.21, "learning_rate": 2.9892957746478873e-05, "loss": 6.224, "step": 1076 }, { "epoch": 1.21, "learning_rate": 2.989154929577465e-05, "loss": 6.1755, "step": 1077 }, { "epoch": 1.21, "learning_rate": 2.9890140845070422e-05, "loss": 6.3848, "step": 1078 }, { "epoch": 1.21, "learning_rate": 2.98887323943662e-05, "loss": 6.3983, "step": 1079 }, { "epoch": 1.21, "learning_rate": 2.9887323943661972e-05, "loss": 6.2985, "step": 1080 }, { "epoch": 1.21, "learning_rate": 2.9885915492957745e-05, "loss": 6.3046, "step": 1081 }, { "epoch": 1.21, "learning_rate": 2.988450704225352e-05, "loss": 6.2889, "step": 1082 }, { "epoch": 1.21, "learning_rate": 2.9883098591549294e-05, "loss": 6.3635, "step": 1083 }, { "epoch": 1.22, "learning_rate": 2.9881690140845074e-05, "loss": 6.1507, "step": 1084 }, { "epoch": 1.22, "learning_rate": 2.9880281690140847e-05, "loss": 6.219, "step": 1085 }, { "epoch": 1.22, "learning_rate": 2.9878873239436623e-05, "loss": 6.246, "step": 1086 }, { "epoch": 1.22, "learning_rate": 2.9877464788732396e-05, "loss": 6.1908, "step": 1087 }, { "epoch": 1.22, "learning_rate": 2.987605633802817e-05, "loss": 6.467, "step": 1088 }, { "epoch": 1.22, "learning_rate": 2.9874647887323945e-05, "loss": 6.2215, "step": 1089 }, { "epoch": 1.22, "learning_rate": 2.9873239436619718e-05, "loss": 6.0667, "step": 1090 }, { "epoch": 1.22, "learning_rate": 2.9871830985915494e-05, "loss": 6.0272, "step": 1091 }, { "epoch": 1.22, "learning_rate": 2.9870422535211267e-05, "loss": 5.8186, "step": 1092 }, { "epoch": 1.23, "learning_rate": 2.9869014084507044e-05, "loss": 6.7718, "step": 1093 }, { "epoch": 1.23, "learning_rate": 2.9867605633802817e-05, "loss": 6.4789, "step": 1094 }, { "epoch": 1.23, "learning_rate": 2.9866197183098593e-05, "loss": 6.6294, "step": 1095 }, { "epoch": 1.23, "learning_rate": 2.9864788732394366e-05, "loss": 6.6414, "step": 1096 }, { "epoch": 1.23, "learning_rate": 2.986338028169014e-05, "loss": 6.6939, "step": 1097 }, { "epoch": 1.23, "learning_rate": 2.9861971830985915e-05, "loss": 6.5149, "step": 1098 }, { "epoch": 1.23, "learning_rate": 2.986056338028169e-05, "loss": 6.5356, "step": 1099 }, { "epoch": 1.23, "learning_rate": 2.9859154929577468e-05, "loss": 6.5936, "step": 1100 }, { "epoch": 1.23, "learning_rate": 2.985774647887324e-05, "loss": 6.3738, "step": 1101 }, { "epoch": 1.24, "learning_rate": 2.9856338028169017e-05, "loss": 6.2633, "step": 1102 }, { "epoch": 1.24, "learning_rate": 2.985492957746479e-05, "loss": 6.1541, "step": 1103 }, { "epoch": 1.24, "learning_rate": 2.9853521126760563e-05, "loss": 6.4302, "step": 1104 }, { "epoch": 1.24, "learning_rate": 2.985211267605634e-05, "loss": 6.3339, "step": 1105 }, { "epoch": 1.24, "learning_rate": 2.9850704225352112e-05, "loss": 6.2754, "step": 1106 }, { "epoch": 1.24, "learning_rate": 2.984929577464789e-05, "loss": 6.2945, "step": 1107 }, { "epoch": 1.24, "learning_rate": 2.984788732394366e-05, "loss": 6.1753, "step": 1108 }, { "epoch": 1.24, "learning_rate": 2.9846478873239438e-05, "loss": 6.1892, "step": 1109 }, { "epoch": 1.24, "learning_rate": 2.984507042253521e-05, "loss": 6.2654, "step": 1110 }, { "epoch": 1.25, "learning_rate": 2.9843661971830987e-05, "loss": 6.3718, "step": 1111 }, { "epoch": 1.25, "learning_rate": 2.984225352112676e-05, "loss": 6.293, "step": 1112 }, { "epoch": 1.25, "learning_rate": 2.9840845070422533e-05, "loss": 6.33, "step": 1113 }, { "epoch": 1.25, "learning_rate": 2.9839436619718313e-05, "loss": 6.3277, "step": 1114 }, { "epoch": 1.25, "learning_rate": 2.9838028169014086e-05, "loss": 6.3241, "step": 1115 }, { "epoch": 1.25, "learning_rate": 2.9836619718309862e-05, "loss": 6.1622, "step": 1116 }, { "epoch": 1.25, "learning_rate": 2.9835211267605635e-05, "loss": 6.1668, "step": 1117 }, { "epoch": 1.25, "learning_rate": 2.983380281690141e-05, "loss": 6.3833, "step": 1118 }, { "epoch": 1.25, "learning_rate": 2.9832394366197184e-05, "loss": 6.2057, "step": 1119 }, { "epoch": 1.26, "learning_rate": 2.9830985915492957e-05, "loss": 6.2075, "step": 1120 }, { "epoch": 1.26, "learning_rate": 2.9829577464788734e-05, "loss": 6.3352, "step": 1121 }, { "epoch": 1.26, "learning_rate": 2.9828169014084507e-05, "loss": 6.3192, "step": 1122 }, { "epoch": 1.26, "learning_rate": 2.9826760563380283e-05, "loss": 6.3066, "step": 1123 }, { "epoch": 1.26, "learning_rate": 2.9825352112676056e-05, "loss": 6.3148, "step": 1124 }, { "epoch": 1.26, "learning_rate": 2.9823943661971832e-05, "loss": 6.3329, "step": 1125 }, { "epoch": 1.26, "learning_rate": 2.9822535211267605e-05, "loss": 6.4241, "step": 1126 }, { "epoch": 1.26, "learning_rate": 2.9821126760563378e-05, "loss": 6.251, "step": 1127 }, { "epoch": 1.26, "learning_rate": 2.9819718309859154e-05, "loss": 6.2758, "step": 1128 }, { "epoch": 1.27, "learning_rate": 2.981830985915493e-05, "loss": 6.3617, "step": 1129 }, { "epoch": 1.27, "learning_rate": 2.9816901408450707e-05, "loss": 6.1587, "step": 1130 }, { "epoch": 1.27, "learning_rate": 2.981549295774648e-05, "loss": 6.2446, "step": 1131 }, { "epoch": 1.27, "learning_rate": 2.9814084507042256e-05, "loss": 6.1706, "step": 1132 }, { "epoch": 1.27, "learning_rate": 2.981267605633803e-05, "loss": 6.3018, "step": 1133 }, { "epoch": 1.27, "learning_rate": 2.9811267605633806e-05, "loss": 6.4065, "step": 1134 }, { "epoch": 1.27, "learning_rate": 2.980985915492958e-05, "loss": 5.7803, "step": 1135 }, { "epoch": 1.27, "learning_rate": 2.980845070422535e-05, "loss": 5.9618, "step": 1136 }, { "epoch": 1.27, "learning_rate": 2.9807042253521128e-05, "loss": 6.3216, "step": 1137 }, { "epoch": 1.28, "learning_rate": 2.98056338028169e-05, "loss": 6.0235, "step": 1138 }, { "epoch": 1.28, "learning_rate": 2.9804225352112677e-05, "loss": 6.0309, "step": 1139 }, { "epoch": 1.28, "learning_rate": 2.980281690140845e-05, "loss": 5.9559, "step": 1140 }, { "epoch": 1.28, "learning_rate": 2.9801408450704226e-05, "loss": 6.2245, "step": 1141 }, { "epoch": 1.28, "learning_rate": 2.98e-05, "loss": 6.1722, "step": 1142 }, { "epoch": 1.28, "learning_rate": 2.9798591549295772e-05, "loss": 6.8418, "step": 1143 }, { "epoch": 1.28, "learning_rate": 2.9797183098591552e-05, "loss": 6.5136, "step": 1144 }, { "epoch": 1.28, "learning_rate": 2.9795774647887325e-05, "loss": 6.5552, "step": 1145 }, { "epoch": 1.28, "learning_rate": 2.97943661971831e-05, "loss": 6.7636, "step": 1146 }, { "epoch": 1.29, "learning_rate": 2.9792957746478874e-05, "loss": 6.5178, "step": 1147 }, { "epoch": 1.29, "learning_rate": 2.979154929577465e-05, "loss": 6.5732, "step": 1148 }, { "epoch": 1.29, "learning_rate": 2.9790140845070423e-05, "loss": 6.4417, "step": 1149 }, { "epoch": 1.29, "learning_rate": 2.97887323943662e-05, "loss": 6.3921, "step": 1150 }, { "epoch": 1.29, "learning_rate": 2.9787323943661973e-05, "loss": 6.2961, "step": 1151 }, { "epoch": 1.29, "learning_rate": 2.9785915492957746e-05, "loss": 6.2402, "step": 1152 }, { "epoch": 1.29, "learning_rate": 2.9784507042253522e-05, "loss": 6.3551, "step": 1153 }, { "epoch": 1.29, "learning_rate": 2.9783098591549295e-05, "loss": 6.3722, "step": 1154 }, { "epoch": 1.29, "learning_rate": 2.978169014084507e-05, "loss": 6.3246, "step": 1155 }, { "epoch": 1.3, "learning_rate": 2.9780281690140844e-05, "loss": 6.245, "step": 1156 }, { "epoch": 1.3, "learning_rate": 2.977887323943662e-05, "loss": 6.3196, "step": 1157 }, { "epoch": 1.3, "learning_rate": 2.9777464788732393e-05, "loss": 6.5105, "step": 1158 }, { "epoch": 1.3, "learning_rate": 2.977605633802817e-05, "loss": 6.3339, "step": 1159 }, { "epoch": 1.3, "learning_rate": 2.9774647887323946e-05, "loss": 6.1188, "step": 1160 }, { "epoch": 1.3, "learning_rate": 2.977323943661972e-05, "loss": 6.3183, "step": 1161 }, { "epoch": 1.3, "learning_rate": 2.9771830985915495e-05, "loss": 6.2606, "step": 1162 }, { "epoch": 1.3, "learning_rate": 2.977042253521127e-05, "loss": 6.183, "step": 1163 }, { "epoch": 1.3, "learning_rate": 2.9769014084507045e-05, "loss": 6.1827, "step": 1164 }, { "epoch": 1.31, "learning_rate": 2.9767605633802818e-05, "loss": 6.3021, "step": 1165 }, { "epoch": 1.31, "learning_rate": 2.9766197183098594e-05, "loss": 6.0902, "step": 1166 }, { "epoch": 1.31, "learning_rate": 2.9764788732394367e-05, "loss": 6.5424, "step": 1167 }, { "epoch": 1.31, "learning_rate": 2.976338028169014e-05, "loss": 6.2943, "step": 1168 }, { "epoch": 1.31, "learning_rate": 2.9761971830985916e-05, "loss": 6.4792, "step": 1169 }, { "epoch": 1.31, "learning_rate": 2.976056338028169e-05, "loss": 6.3929, "step": 1170 }, { "epoch": 1.31, "learning_rate": 2.9759154929577465e-05, "loss": 6.2827, "step": 1171 }, { "epoch": 1.31, "learning_rate": 2.975774647887324e-05, "loss": 6.413, "step": 1172 }, { "epoch": 1.32, "learning_rate": 2.9756338028169015e-05, "loss": 6.2644, "step": 1173 }, { "epoch": 1.32, "learning_rate": 2.975492957746479e-05, "loss": 6.2294, "step": 1174 }, { "epoch": 1.32, "learning_rate": 2.9753521126760564e-05, "loss": 6.2768, "step": 1175 }, { "epoch": 1.32, "learning_rate": 2.975211267605634e-05, "loss": 6.3884, "step": 1176 }, { "epoch": 1.32, "learning_rate": 2.9750704225352113e-05, "loss": 6.3685, "step": 1177 }, { "epoch": 1.32, "learning_rate": 2.974929577464789e-05, "loss": 6.2081, "step": 1178 }, { "epoch": 1.32, "learning_rate": 2.9747887323943663e-05, "loss": 6.3249, "step": 1179 }, { "epoch": 1.32, "learning_rate": 2.974647887323944e-05, "loss": 6.15, "step": 1180 }, { "epoch": 1.32, "learning_rate": 2.9745070422535212e-05, "loss": 6.2148, "step": 1181 }, { "epoch": 1.33, "learning_rate": 2.9743661971830988e-05, "loss": 6.4251, "step": 1182 }, { "epoch": 1.33, "learning_rate": 2.974225352112676e-05, "loss": 6.3884, "step": 1183 }, { "epoch": 1.33, "learning_rate": 2.9740845070422534e-05, "loss": 6.3458, "step": 1184 }, { "epoch": 1.33, "learning_rate": 2.973943661971831e-05, "loss": 6.0537, "step": 1185 }, { "epoch": 1.33, "learning_rate": 2.9738028169014083e-05, "loss": 6.0015, "step": 1186 }, { "epoch": 1.33, "learning_rate": 2.973661971830986e-05, "loss": 6.0147, "step": 1187 }, { "epoch": 1.33, "learning_rate": 2.9735211267605633e-05, "loss": 6.1876, "step": 1188 }, { "epoch": 1.33, "learning_rate": 2.973380281690141e-05, "loss": 5.7878, "step": 1189 }, { "epoch": 1.33, "learning_rate": 2.9732394366197185e-05, "loss": 5.8621, "step": 1190 }, { "epoch": 1.34, "learning_rate": 2.9730985915492958e-05, "loss": 6.041, "step": 1191 }, { "epoch": 1.34, "learning_rate": 2.9729577464788735e-05, "loss": 5.6256, "step": 1192 }, { "epoch": 1.34, "learning_rate": 2.9728169014084507e-05, "loss": 6.7033, "step": 1193 }, { "epoch": 1.34, "learning_rate": 2.9726760563380284e-05, "loss": 6.6923, "step": 1194 }, { "epoch": 1.34, "learning_rate": 2.9725352112676057e-05, "loss": 6.5142, "step": 1195 }, { "epoch": 1.34, "learning_rate": 2.9723943661971833e-05, "loss": 6.626, "step": 1196 }, { "epoch": 1.34, "learning_rate": 2.9722535211267606e-05, "loss": 6.5225, "step": 1197 }, { "epoch": 1.34, "learning_rate": 2.9721126760563382e-05, "loss": 6.3683, "step": 1198 }, { "epoch": 1.34, "learning_rate": 2.9719718309859155e-05, "loss": 6.5126, "step": 1199 }, { "epoch": 1.35, "learning_rate": 2.9718309859154928e-05, "loss": 6.4658, "step": 1200 }, { "epoch": 1.35, "learning_rate": 2.9716901408450705e-05, "loss": 6.3185, "step": 1201 }, { "epoch": 1.35, "learning_rate": 2.9715492957746478e-05, "loss": 6.3775, "step": 1202 }, { "epoch": 1.35, "learning_rate": 2.9714084507042254e-05, "loss": 6.2809, "step": 1203 }, { "epoch": 1.35, "learning_rate": 2.9712676056338027e-05, "loss": 6.2642, "step": 1204 }, { "epoch": 1.35, "learning_rate": 2.9711267605633807e-05, "loss": 6.2291, "step": 1205 }, { "epoch": 1.35, "learning_rate": 2.970985915492958e-05, "loss": 6.2198, "step": 1206 }, { "epoch": 1.35, "learning_rate": 2.9708450704225352e-05, "loss": 6.2646, "step": 1207 }, { "epoch": 1.35, "learning_rate": 2.970704225352113e-05, "loss": 6.2704, "step": 1208 }, { "epoch": 1.36, "learning_rate": 2.9705633802816902e-05, "loss": 6.2019, "step": 1209 }, { "epoch": 1.36, "learning_rate": 2.9704225352112678e-05, "loss": 6.2941, "step": 1210 }, { "epoch": 1.36, "learning_rate": 2.970281690140845e-05, "loss": 6.421, "step": 1211 }, { "epoch": 1.36, "learning_rate": 2.9701408450704227e-05, "loss": 6.247, "step": 1212 }, { "epoch": 1.36, "learning_rate": 2.97e-05, "loss": 6.3664, "step": 1213 }, { "epoch": 1.36, "learning_rate": 2.9698591549295777e-05, "loss": 6.2476, "step": 1214 }, { "epoch": 1.36, "learning_rate": 2.969718309859155e-05, "loss": 6.407, "step": 1215 }, { "epoch": 1.36, "learning_rate": 2.9695774647887322e-05, "loss": 6.177, "step": 1216 }, { "epoch": 1.36, "learning_rate": 2.96943661971831e-05, "loss": 6.2237, "step": 1217 }, { "epoch": 1.37, "learning_rate": 2.9692957746478872e-05, "loss": 6.3802, "step": 1218 }, { "epoch": 1.37, "learning_rate": 2.9691549295774648e-05, "loss": 6.1625, "step": 1219 }, { "epoch": 1.37, "learning_rate": 2.9690140845070424e-05, "loss": 6.2065, "step": 1220 }, { "epoch": 1.37, "learning_rate": 2.96887323943662e-05, "loss": 6.2503, "step": 1221 }, { "epoch": 1.37, "learning_rate": 2.9687323943661974e-05, "loss": 6.2091, "step": 1222 }, { "epoch": 1.37, "learning_rate": 2.9685915492957747e-05, "loss": 6.1216, "step": 1223 }, { "epoch": 1.37, "learning_rate": 2.9684507042253523e-05, "loss": 6.2817, "step": 1224 }, { "epoch": 1.37, "learning_rate": 2.9683098591549296e-05, "loss": 6.3398, "step": 1225 }, { "epoch": 1.37, "learning_rate": 2.9681690140845072e-05, "loss": 6.3739, "step": 1226 }, { "epoch": 1.38, "learning_rate": 2.9680281690140845e-05, "loss": 6.3014, "step": 1227 }, { "epoch": 1.38, "learning_rate": 2.967887323943662e-05, "loss": 6.2711, "step": 1228 }, { "epoch": 1.38, "learning_rate": 2.9677464788732394e-05, "loss": 6.2797, "step": 1229 }, { "epoch": 1.38, "learning_rate": 2.967605633802817e-05, "loss": 6.0434, "step": 1230 }, { "epoch": 1.38, "learning_rate": 2.9674647887323944e-05, "loss": 6.132, "step": 1231 }, { "epoch": 1.38, "learning_rate": 2.9673239436619717e-05, "loss": 6.3005, "step": 1232 }, { "epoch": 1.38, "learning_rate": 2.9671830985915493e-05, "loss": 6.2625, "step": 1233 }, { "epoch": 1.38, "learning_rate": 2.9670422535211266e-05, "loss": 6.3293, "step": 1234 }, { "epoch": 1.38, "learning_rate": 2.9669014084507046e-05, "loss": 6.2742, "step": 1235 }, { "epoch": 1.39, "learning_rate": 2.966760563380282e-05, "loss": 6.1273, "step": 1236 }, { "epoch": 1.39, "learning_rate": 2.9666197183098595e-05, "loss": 6.136, "step": 1237 }, { "epoch": 1.39, "learning_rate": 2.9664788732394368e-05, "loss": 6.1319, "step": 1238 }, { "epoch": 1.39, "learning_rate": 2.966338028169014e-05, "loss": 5.9317, "step": 1239 }, { "epoch": 1.39, "learning_rate": 2.9661971830985917e-05, "loss": 5.9022, "step": 1240 }, { "epoch": 1.39, "learning_rate": 2.966056338028169e-05, "loss": 5.7843, "step": 1241 }, { "epoch": 1.39, "learning_rate": 2.9659154929577466e-05, "loss": 6.1212, "step": 1242 }, { "epoch": 1.39, "learning_rate": 2.965774647887324e-05, "loss": 6.6604, "step": 1243 }, { "epoch": 1.39, "learning_rate": 2.9656338028169016e-05, "loss": 6.4978, "step": 1244 }, { "epoch": 1.4, "learning_rate": 2.965492957746479e-05, "loss": 6.5917, "step": 1245 }, { "epoch": 1.4, "learning_rate": 2.965352112676056e-05, "loss": 6.3559, "step": 1246 }, { "epoch": 1.4, "learning_rate": 2.9652112676056338e-05, "loss": 6.4043, "step": 1247 }, { "epoch": 1.4, "learning_rate": 2.965070422535211e-05, "loss": 6.4181, "step": 1248 }, { "epoch": 1.4, "learning_rate": 2.9649295774647887e-05, "loss": 6.3234, "step": 1249 }, { "epoch": 1.4, "learning_rate": 2.9647887323943664e-05, "loss": 6.2772, "step": 1250 }, { "epoch": 1.4, "learning_rate": 2.964647887323944e-05, "loss": 6.1885, "step": 1251 }, { "epoch": 1.4, "learning_rate": 2.9645070422535213e-05, "loss": 6.0323, "step": 1252 }, { "epoch": 1.4, "learning_rate": 2.964366197183099e-05, "loss": 5.9795, "step": 1253 }, { "epoch": 1.41, "learning_rate": 2.9642253521126762e-05, "loss": 6.3404, "step": 1254 }, { "epoch": 1.41, "learning_rate": 2.9640845070422535e-05, "loss": 6.3313, "step": 1255 }, { "epoch": 1.41, "learning_rate": 2.963943661971831e-05, "loss": 6.1211, "step": 1256 }, { "epoch": 1.41, "learning_rate": 2.9638028169014084e-05, "loss": 6.209, "step": 1257 }, { "epoch": 1.41, "learning_rate": 2.963661971830986e-05, "loss": 6.2449, "step": 1258 }, { "epoch": 1.41, "learning_rate": 2.9635211267605634e-05, "loss": 6.2589, "step": 1259 }, { "epoch": 1.41, "learning_rate": 2.963380281690141e-05, "loss": 6.1962, "step": 1260 }, { "epoch": 1.41, "learning_rate": 2.9632394366197183e-05, "loss": 6.1758, "step": 1261 }, { "epoch": 1.41, "learning_rate": 2.9630985915492956e-05, "loss": 6.3079, "step": 1262 }, { "epoch": 1.42, "learning_rate": 2.9629577464788732e-05, "loss": 5.9786, "step": 1263 }, { "epoch": 1.42, "learning_rate": 2.9628169014084505e-05, "loss": 6.1616, "step": 1264 }, { "epoch": 1.42, "learning_rate": 2.9626760563380285e-05, "loss": 6.0702, "step": 1265 }, { "epoch": 1.42, "learning_rate": 2.9625352112676058e-05, "loss": 6.1419, "step": 1266 }, { "epoch": 1.42, "learning_rate": 2.9623943661971834e-05, "loss": 6.3296, "step": 1267 }, { "epoch": 1.42, "learning_rate": 2.9622535211267607e-05, "loss": 6.1546, "step": 1268 }, { "epoch": 1.42, "learning_rate": 2.9621126760563383e-05, "loss": 6.1684, "step": 1269 }, { "epoch": 1.42, "learning_rate": 2.9619718309859156e-05, "loss": 6.1884, "step": 1270 }, { "epoch": 1.42, "learning_rate": 2.961830985915493e-05, "loss": 6.0297, "step": 1271 }, { "epoch": 1.43, "learning_rate": 2.9616901408450706e-05, "loss": 6.1439, "step": 1272 }, { "epoch": 1.43, "learning_rate": 2.961549295774648e-05, "loss": 6.1178, "step": 1273 }, { "epoch": 1.43, "learning_rate": 2.9614084507042255e-05, "loss": 6.0774, "step": 1274 }, { "epoch": 1.43, "learning_rate": 2.9612676056338028e-05, "loss": 6.1374, "step": 1275 }, { "epoch": 1.43, "learning_rate": 2.9611267605633804e-05, "loss": 6.1075, "step": 1276 }, { "epoch": 1.43, "learning_rate": 2.9609859154929577e-05, "loss": 6.4283, "step": 1277 }, { "epoch": 1.43, "learning_rate": 2.960845070422535e-05, "loss": 6.1142, "step": 1278 }, { "epoch": 1.43, "learning_rate": 2.9607042253521126e-05, "loss": 6.2991, "step": 1279 }, { "epoch": 1.43, "learning_rate": 2.9605633802816903e-05, "loss": 6.0578, "step": 1280 }, { "epoch": 1.44, "learning_rate": 2.960422535211268e-05, "loss": 6.1605, "step": 1281 }, { "epoch": 1.44, "learning_rate": 2.9602816901408452e-05, "loss": 6.088, "step": 1282 }, { "epoch": 1.44, "learning_rate": 2.9601408450704228e-05, "loss": 6.13, "step": 1283 }, { "epoch": 1.44, "learning_rate": 2.96e-05, "loss": 6.0776, "step": 1284 }, { "epoch": 1.44, "learning_rate": 2.9598591549295778e-05, "loss": 6.1847, "step": 1285 }, { "epoch": 1.44, "learning_rate": 2.959718309859155e-05, "loss": 6.2409, "step": 1286 }, { "epoch": 1.44, "learning_rate": 2.9595774647887323e-05, "loss": 5.8963, "step": 1287 }, { "epoch": 1.44, "learning_rate": 2.95943661971831e-05, "loss": 6.1678, "step": 1288 }, { "epoch": 1.45, "learning_rate": 2.9592957746478873e-05, "loss": 6.1462, "step": 1289 }, { "epoch": 1.45, "learning_rate": 2.959154929577465e-05, "loss": 5.9499, "step": 1290 }, { "epoch": 1.45, "learning_rate": 2.9590140845070422e-05, "loss": 6.1849, "step": 1291 }, { "epoch": 1.45, "learning_rate": 2.9588732394366198e-05, "loss": 6.0151, "step": 1292 }, { "epoch": 1.45, "learning_rate": 2.958732394366197e-05, "loss": 6.533, "step": 1293 }, { "epoch": 1.45, "learning_rate": 2.9585915492957744e-05, "loss": 6.5616, "step": 1294 }, { "epoch": 1.45, "learning_rate": 2.9584507042253524e-05, "loss": 6.5221, "step": 1295 }, { "epoch": 1.45, "learning_rate": 2.9583098591549297e-05, "loss": 6.4168, "step": 1296 }, { "epoch": 1.45, "learning_rate": 2.9581690140845073e-05, "loss": 6.2439, "step": 1297 }, { "epoch": 1.46, "learning_rate": 2.9580281690140846e-05, "loss": 6.3584, "step": 1298 }, { "epoch": 1.46, "learning_rate": 2.9578873239436622e-05, "loss": 6.17, "step": 1299 }, { "epoch": 1.46, "learning_rate": 2.9577464788732395e-05, "loss": 6.3574, "step": 1300 }, { "epoch": 1.46, "learning_rate": 2.9576056338028172e-05, "loss": 6.3803, "step": 1301 }, { "epoch": 1.46, "learning_rate": 2.9574647887323945e-05, "loss": 6.1537, "step": 1302 }, { "epoch": 1.46, "learning_rate": 2.9573239436619718e-05, "loss": 6.1912, "step": 1303 }, { "epoch": 1.46, "learning_rate": 2.9571830985915494e-05, "loss": 6.1908, "step": 1304 }, { "epoch": 1.46, "learning_rate": 2.9570422535211267e-05, "loss": 6.2329, "step": 1305 }, { "epoch": 1.46, "learning_rate": 2.9569014084507043e-05, "loss": 6.1863, "step": 1306 }, { "epoch": 1.47, "learning_rate": 2.9567605633802816e-05, "loss": 6.076, "step": 1307 }, { "epoch": 1.47, "learning_rate": 2.9566197183098593e-05, "loss": 6.3223, "step": 1308 }, { "epoch": 1.47, "learning_rate": 2.9564788732394365e-05, "loss": 6.0839, "step": 1309 }, { "epoch": 1.47, "learning_rate": 2.956338028169014e-05, "loss": 6.171, "step": 1310 }, { "epoch": 1.47, "learning_rate": 2.9561971830985918e-05, "loss": 6.2582, "step": 1311 }, { "epoch": 1.47, "learning_rate": 2.956056338028169e-05, "loss": 6.3031, "step": 1312 }, { "epoch": 1.47, "learning_rate": 2.9559154929577467e-05, "loss": 6.0447, "step": 1313 }, { "epoch": 1.47, "learning_rate": 2.955774647887324e-05, "loss": 6.2031, "step": 1314 }, { "epoch": 1.47, "learning_rate": 2.9556338028169017e-05, "loss": 6.2181, "step": 1315 }, { "epoch": 1.48, "learning_rate": 2.955492957746479e-05, "loss": 6.0487, "step": 1316 }, { "epoch": 1.48, "learning_rate": 2.9553521126760566e-05, "loss": 6.1888, "step": 1317 }, { "epoch": 1.48, "learning_rate": 2.955211267605634e-05, "loss": 6.2502, "step": 1318 }, { "epoch": 1.48, "learning_rate": 2.9550704225352112e-05, "loss": 6.3163, "step": 1319 }, { "epoch": 1.48, "learning_rate": 2.9549295774647888e-05, "loss": 6.233, "step": 1320 }, { "epoch": 1.48, "learning_rate": 2.954788732394366e-05, "loss": 6.0955, "step": 1321 }, { "epoch": 1.48, "learning_rate": 2.9546478873239437e-05, "loss": 6.2674, "step": 1322 }, { "epoch": 1.48, "learning_rate": 2.954507042253521e-05, "loss": 6.2027, "step": 1323 }, { "epoch": 1.48, "learning_rate": 2.9543661971830987e-05, "loss": 6.1798, "step": 1324 }, { "epoch": 1.49, "learning_rate": 2.954225352112676e-05, "loss": 6.2686, "step": 1325 }, { "epoch": 1.49, "learning_rate": 2.9540845070422536e-05, "loss": 6.1927, "step": 1326 }, { "epoch": 1.49, "learning_rate": 2.9539436619718312e-05, "loss": 6.2187, "step": 1327 }, { "epoch": 1.49, "learning_rate": 2.9538028169014085e-05, "loss": 6.1811, "step": 1328 }, { "epoch": 1.49, "learning_rate": 2.953661971830986e-05, "loss": 6.3899, "step": 1329 }, { "epoch": 1.49, "learning_rate": 2.9535211267605635e-05, "loss": 6.182, "step": 1330 }, { "epoch": 1.49, "learning_rate": 2.953380281690141e-05, "loss": 6.3192, "step": 1331 }, { "epoch": 1.49, "learning_rate": 2.9532394366197184e-05, "loss": 6.2179, "step": 1332 }, { "epoch": 1.49, "learning_rate": 2.953098591549296e-05, "loss": 6.0658, "step": 1333 }, { "epoch": 1.5, "learning_rate": 2.9529577464788733e-05, "loss": 6.2216, "step": 1334 }, { "epoch": 1.5, "learning_rate": 2.9528169014084506e-05, "loss": 6.1638, "step": 1335 }, { "epoch": 1.5, "learning_rate": 2.9526760563380282e-05, "loss": 6.1537, "step": 1336 }, { "epoch": 1.5, "learning_rate": 2.9525352112676055e-05, "loss": 6.1336, "step": 1337 }, { "epoch": 1.5, "learning_rate": 2.952394366197183e-05, "loss": 6.1129, "step": 1338 }, { "epoch": 1.5, "learning_rate": 2.9522535211267605e-05, "loss": 6.0335, "step": 1339 }, { "epoch": 1.5, "learning_rate": 2.952112676056338e-05, "loss": 6.1841, "step": 1340 }, { "epoch": 1.5, "learning_rate": 2.9519718309859157e-05, "loss": 5.9554, "step": 1341 }, { "epoch": 1.5, "learning_rate": 2.951830985915493e-05, "loss": 5.3781, "step": 1342 }, { "epoch": 1.51, "learning_rate": 2.9516901408450707e-05, "loss": 6.7931, "step": 1343 }, { "epoch": 1.51, "learning_rate": 2.951549295774648e-05, "loss": 6.3238, "step": 1344 }, { "epoch": 1.51, "learning_rate": 2.9514084507042256e-05, "loss": 6.5419, "step": 1345 }, { "epoch": 1.51, "learning_rate": 2.951267605633803e-05, "loss": 6.3065, "step": 1346 }, { "epoch": 1.51, "learning_rate": 2.9511267605633805e-05, "loss": 6.3313, "step": 1347 }, { "epoch": 1.51, "learning_rate": 2.9509859154929578e-05, "loss": 6.3849, "step": 1348 }, { "epoch": 1.51, "learning_rate": 2.950845070422535e-05, "loss": 6.1944, "step": 1349 }, { "epoch": 1.51, "learning_rate": 2.9507042253521127e-05, "loss": 6.1431, "step": 1350 }, { "epoch": 1.51, "learning_rate": 2.95056338028169e-05, "loss": 6.263, "step": 1351 }, { "epoch": 1.52, "learning_rate": 2.9504225352112677e-05, "loss": 6.1507, "step": 1352 }, { "epoch": 1.52, "learning_rate": 2.950281690140845e-05, "loss": 6.1481, "step": 1353 }, { "epoch": 1.52, "learning_rate": 2.9501408450704226e-05, "loss": 6.24, "step": 1354 }, { "epoch": 1.52, "learning_rate": 2.95e-05, "loss": 6.156, "step": 1355 }, { "epoch": 1.52, "learning_rate": 2.949859154929578e-05, "loss": 6.1826, "step": 1356 }, { "epoch": 1.52, "learning_rate": 2.949718309859155e-05, "loss": 6.1456, "step": 1357 }, { "epoch": 1.52, "learning_rate": 2.949718309859155e-05, "loss": 6.2927, "step": 1358 }, { "epoch": 1.52, "learning_rate": 2.9495774647887324e-05, "loss": 6.0679, "step": 1359 }, { "epoch": 1.52, "learning_rate": 2.94943661971831e-05, "loss": 5.9713, "step": 1360 }, { "epoch": 1.53, "learning_rate": 2.9492957746478874e-05, "loss": 5.8458, "step": 1361 }, { "epoch": 1.53, "learning_rate": 2.949154929577465e-05, "loss": 6.3182, "step": 1362 }, { "epoch": 1.53, "learning_rate": 2.9490140845070423e-05, "loss": 6.0503, "step": 1363 }, { "epoch": 1.53, "learning_rate": 2.94887323943662e-05, "loss": 6.201, "step": 1364 }, { "epoch": 1.53, "learning_rate": 2.9487323943661972e-05, "loss": 6.1747, "step": 1365 }, { "epoch": 1.53, "learning_rate": 2.9485915492957745e-05, "loss": 6.2487, "step": 1366 }, { "epoch": 1.53, "learning_rate": 2.948450704225352e-05, "loss": 6.1108, "step": 1367 }, { "epoch": 1.53, "learning_rate": 2.9483098591549294e-05, "loss": 6.1945, "step": 1368 }, { "epoch": 1.53, "learning_rate": 2.948169014084507e-05, "loss": 6.0762, "step": 1369 }, { "epoch": 1.54, "learning_rate": 2.9480281690140844e-05, "loss": 6.0512, "step": 1370 }, { "epoch": 1.54, "learning_rate": 2.947887323943662e-05, "loss": 6.2483, "step": 1371 }, { "epoch": 1.54, "learning_rate": 2.9477464788732396e-05, "loss": 6.2172, "step": 1372 }, { "epoch": 1.54, "learning_rate": 2.9476056338028173e-05, "loss": 6.1367, "step": 1373 }, { "epoch": 1.54, "learning_rate": 2.9474647887323946e-05, "loss": 6.0144, "step": 1374 }, { "epoch": 1.54, "learning_rate": 2.947323943661972e-05, "loss": 6.2348, "step": 1375 }, { "epoch": 1.54, "learning_rate": 2.9471830985915495e-05, "loss": 6.1486, "step": 1376 }, { "epoch": 1.54, "learning_rate": 2.9470422535211268e-05, "loss": 5.8287, "step": 1377 }, { "epoch": 1.54, "learning_rate": 2.9469014084507044e-05, "loss": 6.0736, "step": 1378 }, { "epoch": 1.55, "learning_rate": 2.9467605633802817e-05, "loss": 6.2771, "step": 1379 }, { "epoch": 1.55, "learning_rate": 2.9466197183098593e-05, "loss": 6.3177, "step": 1380 }, { "epoch": 1.55, "learning_rate": 2.9464788732394366e-05, "loss": 6.131, "step": 1381 }, { "epoch": 1.55, "learning_rate": 2.946338028169014e-05, "loss": 6.2355, "step": 1382 }, { "epoch": 1.55, "learning_rate": 2.9461971830985916e-05, "loss": 5.918, "step": 1383 }, { "epoch": 1.55, "learning_rate": 2.946056338028169e-05, "loss": 6.0834, "step": 1384 }, { "epoch": 1.55, "learning_rate": 2.9459154929577465e-05, "loss": 6.2861, "step": 1385 }, { "epoch": 1.55, "learning_rate": 2.9457746478873238e-05, "loss": 6.0045, "step": 1386 }, { "epoch": 1.55, "learning_rate": 2.9456338028169018e-05, "loss": 6.0641, "step": 1387 }, { "epoch": 1.56, "learning_rate": 2.945492957746479e-05, "loss": 6.1697, "step": 1388 }, { "epoch": 1.56, "learning_rate": 2.9453521126760567e-05, "loss": 5.7096, "step": 1389 }, { "epoch": 1.56, "learning_rate": 2.945211267605634e-05, "loss": 5.7731, "step": 1390 }, { "epoch": 1.56, "learning_rate": 2.9450704225352113e-05, "loss": 5.9168, "step": 1391 }, { "epoch": 1.56, "learning_rate": 2.944929577464789e-05, "loss": 5.448, "step": 1392 }, { "epoch": 1.56, "learning_rate": 2.9447887323943662e-05, "loss": 6.4478, "step": 1393 }, { "epoch": 1.56, "learning_rate": 2.944647887323944e-05, "loss": 6.5113, "step": 1394 }, { "epoch": 1.56, "learning_rate": 2.944507042253521e-05, "loss": 6.4274, "step": 1395 }, { "epoch": 1.57, "learning_rate": 2.9443661971830988e-05, "loss": 6.2405, "step": 1396 }, { "epoch": 1.57, "learning_rate": 2.944225352112676e-05, "loss": 6.1301, "step": 1397 }, { "epoch": 1.57, "learning_rate": 2.9440845070422534e-05, "loss": 6.2568, "step": 1398 }, { "epoch": 1.57, "learning_rate": 2.943943661971831e-05, "loss": 6.2239, "step": 1399 }, { "epoch": 1.57, "learning_rate": 2.9438028169014083e-05, "loss": 6.2438, "step": 1400 }, { "epoch": 1.57, "learning_rate": 2.943661971830986e-05, "loss": 6.1611, "step": 1401 }, { "epoch": 1.57, "learning_rate": 2.9435211267605635e-05, "loss": 6.2848, "step": 1402 }, { "epoch": 1.57, "learning_rate": 2.9433802816901412e-05, "loss": 6.1112, "step": 1403 }, { "epoch": 1.57, "learning_rate": 2.9432394366197185e-05, "loss": 6.2344, "step": 1404 }, { "epoch": 1.58, "learning_rate": 2.943098591549296e-05, "loss": 6.0674, "step": 1405 }, { "epoch": 1.58, "learning_rate": 2.9429577464788734e-05, "loss": 6.2329, "step": 1406 }, { "epoch": 1.58, "learning_rate": 2.9428169014084507e-05, "loss": 6.2518, "step": 1407 }, { "epoch": 1.58, "learning_rate": 2.9426760563380283e-05, "loss": 6.0866, "step": 1408 }, { "epoch": 1.58, "learning_rate": 2.9425352112676056e-05, "loss": 6.1608, "step": 1409 }, { "epoch": 1.58, "learning_rate": 2.9423943661971833e-05, "loss": 6.1921, "step": 1410 }, { "epoch": 1.58, "learning_rate": 2.9422535211267606e-05, "loss": 6.1989, "step": 1411 }, { "epoch": 1.58, "learning_rate": 2.9421126760563382e-05, "loss": 6.1989, "step": 1412 }, { "epoch": 1.58, "learning_rate": 2.9419718309859155e-05, "loss": 6.0636, "step": 1413 }, { "epoch": 1.59, "learning_rate": 2.9418309859154928e-05, "loss": 6.1139, "step": 1414 }, { "epoch": 1.59, "learning_rate": 2.9416901408450704e-05, "loss": 6.134, "step": 1415 }, { "epoch": 1.59, "learning_rate": 2.9415492957746477e-05, "loss": 5.998, "step": 1416 }, { "epoch": 1.59, "learning_rate": 2.9414084507042253e-05, "loss": 5.9862, "step": 1417 }, { "epoch": 1.59, "learning_rate": 2.941267605633803e-05, "loss": 6.0246, "step": 1418 }, { "epoch": 1.59, "learning_rate": 2.9411267605633806e-05, "loss": 6.0458, "step": 1419 }, { "epoch": 1.59, "learning_rate": 2.940985915492958e-05, "loss": 6.3279, "step": 1420 }, { "epoch": 1.59, "learning_rate": 2.9408450704225355e-05, "loss": 6.2257, "step": 1421 }, { "epoch": 1.59, "learning_rate": 2.9407042253521128e-05, "loss": 6.2774, "step": 1422 }, { "epoch": 1.6, "learning_rate": 2.94056338028169e-05, "loss": 6.1291, "step": 1423 }, { "epoch": 1.6, "learning_rate": 2.9404225352112678e-05, "loss": 6.0113, "step": 1424 }, { "epoch": 1.6, "learning_rate": 2.940281690140845e-05, "loss": 5.9661, "step": 1425 }, { "epoch": 1.6, "learning_rate": 2.9401408450704227e-05, "loss": 5.9626, "step": 1426 }, { "epoch": 1.6, "learning_rate": 2.94e-05, "loss": 5.9844, "step": 1427 }, { "epoch": 1.6, "learning_rate": 2.9398591549295776e-05, "loss": 6.0673, "step": 1428 }, { "epoch": 1.6, "learning_rate": 2.939718309859155e-05, "loss": 6.0259, "step": 1429 }, { "epoch": 1.6, "learning_rate": 2.9395774647887322e-05, "loss": 6.0812, "step": 1430 }, { "epoch": 1.6, "learning_rate": 2.9394366197183098e-05, "loss": 6.1162, "step": 1431 }, { "epoch": 1.61, "learning_rate": 2.939295774647887e-05, "loss": 6.2233, "step": 1432 }, { "epoch": 1.61, "learning_rate": 2.939154929577465e-05, "loss": 6.2687, "step": 1433 }, { "epoch": 1.61, "learning_rate": 2.9390140845070424e-05, "loss": 5.9497, "step": 1434 }, { "epoch": 1.61, "learning_rate": 2.93887323943662e-05, "loss": 6.1435, "step": 1435 }, { "epoch": 1.61, "learning_rate": 2.9387323943661973e-05, "loss": 6.2364, "step": 1436 }, { "epoch": 1.61, "learning_rate": 2.938591549295775e-05, "loss": 6.2021, "step": 1437 }, { "epoch": 1.61, "learning_rate": 2.9384507042253522e-05, "loss": 5.8919, "step": 1438 }, { "epoch": 1.61, "learning_rate": 2.9383098591549295e-05, "loss": 5.7925, "step": 1439 }, { "epoch": 1.61, "learning_rate": 2.9381690140845072e-05, "loss": 5.8799, "step": 1440 }, { "epoch": 1.62, "learning_rate": 2.9380281690140845e-05, "loss": 5.5088, "step": 1441 }, { "epoch": 1.62, "learning_rate": 2.937887323943662e-05, "loss": 5.8268, "step": 1442 }, { "epoch": 1.62, "learning_rate": 2.9377464788732394e-05, "loss": 6.263, "step": 1443 }, { "epoch": 1.62, "learning_rate": 2.937605633802817e-05, "loss": 6.283, "step": 1444 }, { "epoch": 1.62, "learning_rate": 2.9374647887323943e-05, "loss": 6.471, "step": 1445 }, { "epoch": 1.62, "learning_rate": 2.9373239436619716e-05, "loss": 6.2394, "step": 1446 }, { "epoch": 1.62, "learning_rate": 2.9371830985915492e-05, "loss": 6.2217, "step": 1447 }, { "epoch": 1.62, "learning_rate": 2.937042253521127e-05, "loss": 6.1597, "step": 1448 }, { "epoch": 1.62, "learning_rate": 2.9369014084507045e-05, "loss": 6.154, "step": 1449 }, { "epoch": 1.63, "learning_rate": 2.9367605633802818e-05, "loss": 6.2529, "step": 1450 }, { "epoch": 1.63, "learning_rate": 2.9366197183098594e-05, "loss": 6.1294, "step": 1451 }, { "epoch": 1.63, "learning_rate": 2.9364788732394367e-05, "loss": 6.3784, "step": 1452 }, { "epoch": 1.63, "learning_rate": 2.936338028169014e-05, "loss": 6.0904, "step": 1453 }, { "epoch": 1.63, "learning_rate": 2.9361971830985917e-05, "loss": 6.0391, "step": 1454 }, { "epoch": 1.63, "learning_rate": 2.936056338028169e-05, "loss": 6.1968, "step": 1455 }, { "epoch": 1.63, "learning_rate": 2.9359154929577466e-05, "loss": 6.0566, "step": 1456 }, { "epoch": 1.63, "learning_rate": 2.935774647887324e-05, "loss": 6.0875, "step": 1457 }, { "epoch": 1.63, "learning_rate": 2.9356338028169015e-05, "loss": 6.0705, "step": 1458 }, { "epoch": 1.64, "learning_rate": 2.9354929577464788e-05, "loss": 6.1732, "step": 1459 }, { "epoch": 1.64, "learning_rate": 2.9353521126760564e-05, "loss": 6.0562, "step": 1460 }, { "epoch": 1.64, "learning_rate": 2.9352112676056337e-05, "loss": 6.0472, "step": 1461 }, { "epoch": 1.64, "learning_rate": 2.935070422535211e-05, "loss": 6.1378, "step": 1462 }, { "epoch": 1.64, "learning_rate": 2.934929577464789e-05, "loss": 6.0334, "step": 1463 }, { "epoch": 1.64, "learning_rate": 2.9347887323943663e-05, "loss": 6.3486, "step": 1464 }, { "epoch": 1.64, "learning_rate": 2.934647887323944e-05, "loss": 6.1129, "step": 1465 }, { "epoch": 1.64, "learning_rate": 2.9345070422535212e-05, "loss": 6.1778, "step": 1466 }, { "epoch": 1.64, "learning_rate": 2.934366197183099e-05, "loss": 6.0709, "step": 1467 }, { "epoch": 1.65, "learning_rate": 2.934225352112676e-05, "loss": 6.1215, "step": 1468 }, { "epoch": 1.65, "learning_rate": 2.9340845070422535e-05, "loss": 6.1232, "step": 1469 }, { "epoch": 1.65, "learning_rate": 2.933943661971831e-05, "loss": 6.0483, "step": 1470 }, { "epoch": 1.65, "learning_rate": 2.9338028169014084e-05, "loss": 5.9697, "step": 1471 }, { "epoch": 1.65, "learning_rate": 2.933661971830986e-05, "loss": 6.193, "step": 1472 }, { "epoch": 1.65, "learning_rate": 2.9335211267605633e-05, "loss": 6.0191, "step": 1473 }, { "epoch": 1.65, "learning_rate": 2.933380281690141e-05, "loss": 5.906, "step": 1474 }, { "epoch": 1.65, "learning_rate": 2.9332394366197182e-05, "loss": 5.8394, "step": 1475 }, { "epoch": 1.65, "learning_rate": 2.933098591549296e-05, "loss": 6.0015, "step": 1476 }, { "epoch": 1.66, "learning_rate": 2.932957746478873e-05, "loss": 6.0934, "step": 1477 }, { "epoch": 1.66, "learning_rate": 2.9328169014084508e-05, "loss": 5.9619, "step": 1478 }, { "epoch": 1.66, "learning_rate": 2.9326760563380284e-05, "loss": 6.1109, "step": 1479 }, { "epoch": 1.66, "learning_rate": 2.9325352112676057e-05, "loss": 6.0571, "step": 1480 }, { "epoch": 1.66, "learning_rate": 2.9323943661971834e-05, "loss": 5.8815, "step": 1481 }, { "epoch": 1.66, "learning_rate": 2.9322535211267606e-05, "loss": 5.9836, "step": 1482 }, { "epoch": 1.66, "learning_rate": 2.9321126760563383e-05, "loss": 5.9076, "step": 1483 }, { "epoch": 1.66, "learning_rate": 2.9319718309859156e-05, "loss": 6.1244, "step": 1484 }, { "epoch": 1.66, "learning_rate": 2.931830985915493e-05, "loss": 6.1473, "step": 1485 }, { "epoch": 1.67, "learning_rate": 2.9316901408450705e-05, "loss": 6.0701, "step": 1486 }, { "epoch": 1.67, "learning_rate": 2.9315492957746478e-05, "loss": 6.0622, "step": 1487 }, { "epoch": 1.67, "learning_rate": 2.9314084507042254e-05, "loss": 6.0141, "step": 1488 }, { "epoch": 1.67, "learning_rate": 2.9312676056338027e-05, "loss": 5.7696, "step": 1489 }, { "epoch": 1.67, "learning_rate": 2.9311267605633804e-05, "loss": 6.1497, "step": 1490 }, { "epoch": 1.67, "learning_rate": 2.9309859154929577e-05, "loss": 5.7356, "step": 1491 }, { "epoch": 1.67, "learning_rate": 2.9308450704225353e-05, "loss": 5.7511, "step": 1492 }, { "epoch": 1.67, "learning_rate": 2.930704225352113e-05, "loss": 6.404, "step": 1493 }, { "epoch": 1.67, "learning_rate": 2.9305633802816902e-05, "loss": 6.2661, "step": 1494 }, { "epoch": 1.68, "learning_rate": 2.930422535211268e-05, "loss": 6.3202, "step": 1495 }, { "epoch": 1.68, "learning_rate": 2.930281690140845e-05, "loss": 6.2215, "step": 1496 }, { "epoch": 1.68, "learning_rate": 2.9301408450704228e-05, "loss": 6.3391, "step": 1497 }, { "epoch": 1.68, "learning_rate": 2.93e-05, "loss": 6.1115, "step": 1498 }, { "epoch": 1.68, "learning_rate": 2.9298591549295777e-05, "loss": 6.3253, "step": 1499 }, { "epoch": 1.68, "learning_rate": 2.929718309859155e-05, "loss": 6.1228, "step": 1500 }, { "epoch": 1.68, "eval_loss": 6.049000263214111, "eval_runtime": 744.98, "eval_samples_per_second": 3.546, "eval_steps_per_second": 0.444, "eval_wer": 1.1433359777865926, "step": 1500 }, { "epoch": 1.68, "learning_rate": 2.9295774647887323e-05, "loss": 6.2004, "step": 1501 }, { "epoch": 1.68, "learning_rate": 2.92943661971831e-05, "loss": 6.0471, "step": 1502 }, { "epoch": 1.68, "learning_rate": 2.9292957746478872e-05, "loss": 5.9515, "step": 1503 }, { "epoch": 1.69, "learning_rate": 2.929154929577465e-05, "loss": 6.0629, "step": 1504 }, { "epoch": 1.69, "learning_rate": 2.929014084507042e-05, "loss": 6.0985, "step": 1505 }, { "epoch": 1.69, "learning_rate": 2.9288732394366198e-05, "loss": 6.0924, "step": 1506 }, { "epoch": 1.69, "learning_rate": 2.928732394366197e-05, "loss": 5.9687, "step": 1507 }, { "epoch": 1.69, "learning_rate": 2.928591549295775e-05, "loss": 6.0135, "step": 1508 }, { "epoch": 1.69, "learning_rate": 2.9284507042253523e-05, "loss": 5.9107, "step": 1509 }, { "epoch": 1.69, "learning_rate": 2.9283098591549296e-05, "loss": 5.9253, "step": 1510 }, { "epoch": 1.69, "learning_rate": 2.9281690140845073e-05, "loss": 6.0603, "step": 1511 }, { "epoch": 1.7, "learning_rate": 2.9280281690140846e-05, "loss": 6.0473, "step": 1512 }, { "epoch": 1.7, "learning_rate": 2.9278873239436622e-05, "loss": 6.1673, "step": 1513 }, { "epoch": 1.7, "learning_rate": 2.9277464788732395e-05, "loss": 6.1492, "step": 1514 }, { "epoch": 1.7, "learning_rate": 2.927605633802817e-05, "loss": 6.0785, "step": 1515 }, { "epoch": 1.7, "learning_rate": 2.9274647887323944e-05, "loss": 6.1546, "step": 1516 }, { "epoch": 1.7, "learning_rate": 2.9273239436619717e-05, "loss": 6.0367, "step": 1517 }, { "epoch": 1.7, "learning_rate": 2.9271830985915493e-05, "loss": 6.0377, "step": 1518 }, { "epoch": 1.7, "learning_rate": 2.9270422535211266e-05, "loss": 6.0164, "step": 1519 }, { "epoch": 1.7, "learning_rate": 2.9269014084507043e-05, "loss": 5.9489, "step": 1520 }, { "epoch": 1.71, "learning_rate": 2.9267605633802816e-05, "loss": 6.1456, "step": 1521 }, { "epoch": 1.71, "learning_rate": 2.9266197183098592e-05, "loss": 6.279, "step": 1522 }, { "epoch": 1.71, "learning_rate": 2.926478873239437e-05, "loss": 5.9338, "step": 1523 }, { "epoch": 1.71, "learning_rate": 2.9263380281690145e-05, "loss": 6.0337, "step": 1524 }, { "epoch": 1.71, "learning_rate": 2.9261971830985918e-05, "loss": 6.1372, "step": 1525 }, { "epoch": 1.71, "learning_rate": 2.926056338028169e-05, "loss": 5.9877, "step": 1526 }, { "epoch": 1.71, "learning_rate": 2.9259154929577467e-05, "loss": 6.2077, "step": 1527 }, { "epoch": 1.71, "learning_rate": 2.925774647887324e-05, "loss": 6.1133, "step": 1528 }, { "epoch": 1.71, "learning_rate": 2.9256338028169016e-05, "loss": 6.1261, "step": 1529 }, { "epoch": 1.72, "learning_rate": 2.925492957746479e-05, "loss": 6.0555, "step": 1530 }, { "epoch": 1.72, "learning_rate": 2.9253521126760565e-05, "loss": 5.948, "step": 1531 }, { "epoch": 1.72, "learning_rate": 2.925211267605634e-05, "loss": 6.0647, "step": 1532 }, { "epoch": 1.72, "learning_rate": 2.925070422535211e-05, "loss": 5.9601, "step": 1533 }, { "epoch": 1.72, "learning_rate": 2.9249295774647888e-05, "loss": 6.0147, "step": 1534 }, { "epoch": 1.72, "learning_rate": 2.924788732394366e-05, "loss": 5.9219, "step": 1535 }, { "epoch": 1.72, "learning_rate": 2.9246478873239437e-05, "loss": 6.0848, "step": 1536 }, { "epoch": 1.72, "learning_rate": 2.924507042253521e-05, "loss": 6.0089, "step": 1537 }, { "epoch": 1.72, "learning_rate": 2.9243661971830986e-05, "loss": 5.8692, "step": 1538 }, { "epoch": 1.73, "learning_rate": 2.9242253521126763e-05, "loss": 5.7797, "step": 1539 }, { "epoch": 1.73, "learning_rate": 2.924084507042254e-05, "loss": 5.9839, "step": 1540 }, { "epoch": 1.73, "learning_rate": 2.9239436619718312e-05, "loss": 5.7731, "step": 1541 }, { "epoch": 1.73, "learning_rate": 2.9238028169014085e-05, "loss": 5.46, "step": 1542 }, { "epoch": 1.73, "learning_rate": 2.923661971830986e-05, "loss": 6.4622, "step": 1543 }, { "epoch": 1.73, "learning_rate": 2.9235211267605634e-05, "loss": 6.5215, "step": 1544 }, { "epoch": 1.73, "learning_rate": 2.923380281690141e-05, "loss": 6.4027, "step": 1545 }, { "epoch": 1.73, "learning_rate": 2.9232394366197183e-05, "loss": 6.3182, "step": 1546 }, { "epoch": 1.73, "learning_rate": 2.923098591549296e-05, "loss": 6.1534, "step": 1547 }, { "epoch": 1.74, "learning_rate": 2.9229577464788733e-05, "loss": 6.1147, "step": 1548 }, { "epoch": 1.74, "learning_rate": 2.9228169014084506e-05, "loss": 6.3367, "step": 1549 }, { "epoch": 1.74, "learning_rate": 2.9226760563380282e-05, "loss": 6.1292, "step": 1550 }, { "epoch": 1.74, "learning_rate": 2.9225352112676055e-05, "loss": 6.0025, "step": 1551 }, { "epoch": 1.74, "learning_rate": 2.922394366197183e-05, "loss": 6.2187, "step": 1552 }, { "epoch": 1.74, "learning_rate": 2.9222535211267604e-05, "loss": 6.1783, "step": 1553 }, { "epoch": 1.74, "learning_rate": 2.9221126760563384e-05, "loss": 5.9208, "step": 1554 }, { "epoch": 1.74, "learning_rate": 2.9219718309859157e-05, "loss": 6.0509, "step": 1555 }, { "epoch": 1.74, "learning_rate": 2.9218309859154933e-05, "loss": 5.9957, "step": 1556 }, { "epoch": 1.75, "learning_rate": 2.9216901408450706e-05, "loss": 6.0498, "step": 1557 }, { "epoch": 1.75, "learning_rate": 2.921549295774648e-05, "loss": 6.0713, "step": 1558 }, { "epoch": 1.75, "learning_rate": 2.9214084507042255e-05, "loss": 6.1264, "step": 1559 }, { "epoch": 1.75, "learning_rate": 2.9212676056338028e-05, "loss": 6.1788, "step": 1560 }, { "epoch": 1.75, "learning_rate": 2.9211267605633805e-05, "loss": 6.0089, "step": 1561 }, { "epoch": 1.75, "learning_rate": 2.9209859154929577e-05, "loss": 6.2609, "step": 1562 }, { "epoch": 1.75, "learning_rate": 2.9208450704225354e-05, "loss": 6.1638, "step": 1563 }, { "epoch": 1.75, "learning_rate": 2.9207042253521127e-05, "loss": 6.0382, "step": 1564 }, { "epoch": 1.75, "learning_rate": 2.92056338028169e-05, "loss": 5.8401, "step": 1565 }, { "epoch": 1.76, "learning_rate": 2.9204225352112676e-05, "loss": 6.2542, "step": 1566 }, { "epoch": 1.76, "learning_rate": 2.920281690140845e-05, "loss": 6.1117, "step": 1567 }, { "epoch": 1.76, "learning_rate": 2.9201408450704225e-05, "loss": 5.9185, "step": 1568 }, { "epoch": 1.76, "learning_rate": 2.92e-05, "loss": 5.9253, "step": 1569 }, { "epoch": 1.76, "learning_rate": 2.9198591549295778e-05, "loss": 6.1013, "step": 1570 }, { "epoch": 1.76, "learning_rate": 2.919718309859155e-05, "loss": 6.0516, "step": 1571 }, { "epoch": 1.76, "learning_rate": 2.9195774647887324e-05, "loss": 6.1426, "step": 1572 }, { "epoch": 1.76, "learning_rate": 2.91943661971831e-05, "loss": 6.1057, "step": 1573 }, { "epoch": 1.76, "learning_rate": 2.9192957746478873e-05, "loss": 5.9246, "step": 1574 }, { "epoch": 1.77, "learning_rate": 2.919154929577465e-05, "loss": 6.0455, "step": 1575 }, { "epoch": 1.77, "learning_rate": 2.9190140845070422e-05, "loss": 6.1037, "step": 1576 }, { "epoch": 1.77, "learning_rate": 2.91887323943662e-05, "loss": 5.9483, "step": 1577 }, { "epoch": 1.77, "learning_rate": 2.918732394366197e-05, "loss": 6.0085, "step": 1578 }, { "epoch": 1.77, "learning_rate": 2.9185915492957748e-05, "loss": 6.0449, "step": 1579 }, { "epoch": 1.77, "learning_rate": 2.918450704225352e-05, "loss": 6.0782, "step": 1580 }, { "epoch": 1.77, "learning_rate": 2.9183098591549294e-05, "loss": 6.1216, "step": 1581 }, { "epoch": 1.77, "learning_rate": 2.918169014084507e-05, "loss": 6.1153, "step": 1582 }, { "epoch": 1.77, "learning_rate": 2.9180281690140843e-05, "loss": 6.062, "step": 1583 }, { "epoch": 1.78, "learning_rate": 2.9178873239436623e-05, "loss": 6.0298, "step": 1584 }, { "epoch": 1.78, "learning_rate": 2.9177464788732396e-05, "loss": 5.9663, "step": 1585 }, { "epoch": 1.78, "learning_rate": 2.9176056338028172e-05, "loss": 5.9854, "step": 1586 }, { "epoch": 1.78, "learning_rate": 2.9174647887323945e-05, "loss": 5.9946, "step": 1587 }, { "epoch": 1.78, "learning_rate": 2.9174647887323945e-05, "loss": 5.8764, "step": 1588 }, { "epoch": 1.78, "learning_rate": 2.9173239436619718e-05, "loss": 6.0714, "step": 1589 }, { "epoch": 1.78, "learning_rate": 2.9171830985915494e-05, "loss": 5.8694, "step": 1590 }, { "epoch": 1.78, "learning_rate": 2.9170422535211267e-05, "loss": 5.9904, "step": 1591 }, { "epoch": 1.78, "learning_rate": 2.9169014084507044e-05, "loss": 5.3936, "step": 1592 }, { "epoch": 1.79, "learning_rate": 2.9167605633802817e-05, "loss": 6.3687, "step": 1593 }, { "epoch": 1.79, "learning_rate": 2.9166197183098593e-05, "loss": 6.2445, "step": 1594 }, { "epoch": 1.79, "learning_rate": 2.9164788732394366e-05, "loss": 6.2438, "step": 1595 }, { "epoch": 1.79, "learning_rate": 2.9163380281690142e-05, "loss": 6.2029, "step": 1596 }, { "epoch": 1.79, "learning_rate": 2.9161971830985915e-05, "loss": 6.0679, "step": 1597 }, { "epoch": 1.79, "learning_rate": 2.9160563380281688e-05, "loss": 6.1293, "step": 1598 }, { "epoch": 1.79, "learning_rate": 2.9159154929577464e-05, "loss": 6.0646, "step": 1599 }, { "epoch": 1.79, "learning_rate": 2.915774647887324e-05, "loss": 6.3119, "step": 1600 }, { "epoch": 1.79, "learning_rate": 2.9156338028169017e-05, "loss": 6.1867, "step": 1601 }, { "epoch": 1.8, "learning_rate": 2.915492957746479e-05, "loss": 6.1826, "step": 1602 }, { "epoch": 1.8, "learning_rate": 2.9153521126760566e-05, "loss": 6.1194, "step": 1603 }, { "epoch": 1.8, "learning_rate": 2.9153521126760566e-05, "loss": 6.1019, "step": 1604 }, { "epoch": 1.8, "learning_rate": 2.915211267605634e-05, "loss": 6.096, "step": 1605 }, { "epoch": 1.8, "learning_rate": 2.9150704225352112e-05, "loss": 6.1322, "step": 1606 }, { "epoch": 1.8, "learning_rate": 2.914929577464789e-05, "loss": 6.0409, "step": 1607 }, { "epoch": 1.8, "learning_rate": 2.914788732394366e-05, "loss": 6.1079, "step": 1608 }, { "epoch": 1.8, "learning_rate": 2.914788732394366e-05, "loss": 6.0876, "step": 1609 }, { "epoch": 1.8, "learning_rate": 2.9146478873239438e-05, "loss": 5.981, "step": 1610 }, { "epoch": 1.81, "learning_rate": 2.914507042253521e-05, "loss": 6.2796, "step": 1611 }, { "epoch": 1.81, "learning_rate": 2.9143661971830987e-05, "loss": 6.0683, "step": 1612 }, { "epoch": 1.81, "learning_rate": 2.914225352112676e-05, "loss": 6.0818, "step": 1613 }, { "epoch": 1.81, "learning_rate": 2.9140845070422536e-05, "loss": 6.2159, "step": 1614 }, { "epoch": 1.81, "learning_rate": 2.913943661971831e-05, "loss": 6.1481, "step": 1615 }, { "epoch": 1.81, "learning_rate": 2.9138028169014082e-05, "loss": 6.1143, "step": 1616 }, { "epoch": 1.81, "learning_rate": 2.9136619718309862e-05, "loss": 6.1928, "step": 1617 }, { "epoch": 1.81, "learning_rate": 2.9135211267605635e-05, "loss": 6.4181, "step": 1618 }, { "epoch": 1.82, "learning_rate": 2.913380281690141e-05, "loss": 6.2003, "step": 1619 }, { "epoch": 1.82, "learning_rate": 2.9132394366197184e-05, "loss": 5.9223, "step": 1620 }, { "epoch": 1.82, "learning_rate": 2.913098591549296e-05, "loss": 6.2242, "step": 1621 }, { "epoch": 1.82, "learning_rate": 2.9129577464788734e-05, "loss": 5.9816, "step": 1622 }, { "epoch": 1.82, "learning_rate": 2.9128169014084506e-05, "loss": 6.1214, "step": 1623 }, { "epoch": 1.82, "learning_rate": 2.9126760563380283e-05, "loss": 5.892, "step": 1624 }, { "epoch": 1.82, "learning_rate": 2.9125352112676056e-05, "loss": 5.9336, "step": 1625 }, { "epoch": 1.82, "learning_rate": 2.9123943661971832e-05, "loss": 5.9664, "step": 1626 }, { "epoch": 1.82, "learning_rate": 2.9122535211267605e-05, "loss": 6.0782, "step": 1627 }, { "epoch": 1.83, "learning_rate": 2.912112676056338e-05, "loss": 6.2288, "step": 1628 }, { "epoch": 1.83, "learning_rate": 2.9119718309859154e-05, "loss": 6.1136, "step": 1629 }, { "epoch": 1.83, "learning_rate": 2.911830985915493e-05, "loss": 6.1143, "step": 1630 }, { "epoch": 1.83, "learning_rate": 2.9116901408450704e-05, "loss": 5.7994, "step": 1631 }, { "epoch": 1.83, "learning_rate": 2.911549295774648e-05, "loss": 5.9985, "step": 1632 }, { "epoch": 1.83, "learning_rate": 2.9114084507042256e-05, "loss": 6.1532, "step": 1633 }, { "epoch": 1.83, "learning_rate": 2.911267605633803e-05, "loss": 6.2393, "step": 1634 }, { "epoch": 1.83, "learning_rate": 2.9111267605633806e-05, "loss": 5.994, "step": 1635 }, { "epoch": 1.83, "learning_rate": 2.910985915492958e-05, "loss": 5.9457, "step": 1636 }, { "epoch": 1.84, "learning_rate": 2.9108450704225355e-05, "loss": 5.9513, "step": 1637 }, { "epoch": 1.84, "learning_rate": 2.9107042253521128e-05, "loss": 5.9366, "step": 1638 }, { "epoch": 1.84, "learning_rate": 2.91056338028169e-05, "loss": 5.8267, "step": 1639 }, { "epoch": 1.84, "learning_rate": 2.9104225352112677e-05, "loss": 5.8861, "step": 1640 }, { "epoch": 1.84, "learning_rate": 2.910281690140845e-05, "loss": 5.9422, "step": 1641 }, { "epoch": 1.84, "learning_rate": 2.9101408450704226e-05, "loss": 5.467, "step": 1642 }, { "epoch": 1.84, "learning_rate": 2.91e-05, "loss": 6.0324, "step": 1643 }, { "epoch": 1.84, "learning_rate": 2.9098591549295776e-05, "loss": 6.2842, "step": 1644 }, { "epoch": 1.84, "learning_rate": 2.909718309859155e-05, "loss": 6.1351, "step": 1645 }, { "epoch": 1.85, "learning_rate": 2.9095774647887325e-05, "loss": 6.0895, "step": 1646 }, { "epoch": 1.85, "learning_rate": 2.90943661971831e-05, "loss": 6.0895, "step": 1647 }, { "epoch": 1.85, "learning_rate": 2.9092957746478874e-05, "loss": 5.9978, "step": 1648 }, { "epoch": 1.85, "learning_rate": 2.909154929577465e-05, "loss": 6.1866, "step": 1649 }, { "epoch": 1.85, "learning_rate": 2.9090140845070423e-05, "loss": 6.2649, "step": 1650 }, { "epoch": 1.85, "learning_rate": 2.90887323943662e-05, "loss": 6.065, "step": 1651 }, { "epoch": 1.85, "learning_rate": 2.9087323943661973e-05, "loss": 6.2777, "step": 1652 }, { "epoch": 1.85, "learning_rate": 2.908591549295775e-05, "loss": 5.9167, "step": 1653 }, { "epoch": 1.85, "learning_rate": 2.9084507042253522e-05, "loss": 5.8953, "step": 1654 }, { "epoch": 1.86, "learning_rate": 2.9083098591549295e-05, "loss": 6.0555, "step": 1655 }, { "epoch": 1.86, "learning_rate": 2.908169014084507e-05, "loss": 6.091, "step": 1656 }, { "epoch": 1.86, "learning_rate": 2.9080281690140844e-05, "loss": 6.1498, "step": 1657 }, { "epoch": 1.86, "learning_rate": 2.907887323943662e-05, "loss": 6.2747, "step": 1658 }, { "epoch": 1.86, "learning_rate": 2.9077464788732393e-05, "loss": 5.9149, "step": 1659 }, { "epoch": 1.86, "learning_rate": 2.907605633802817e-05, "loss": 6.0596, "step": 1660 }, { "epoch": 1.86, "learning_rate": 2.9074647887323943e-05, "loss": 6.1016, "step": 1661 }, { "epoch": 1.86, "learning_rate": 2.907323943661972e-05, "loss": 5.988, "step": 1662 }, { "epoch": 1.86, "learning_rate": 2.9071830985915495e-05, "loss": 6.009, "step": 1663 }, { "epoch": 1.87, "learning_rate": 2.9070422535211268e-05, "loss": 6.0478, "step": 1664 }, { "epoch": 1.87, "learning_rate": 2.9069014084507045e-05, "loss": 5.8709, "step": 1665 }, { "epoch": 1.87, "learning_rate": 2.9067605633802818e-05, "loss": 6.2083, "step": 1666 }, { "epoch": 1.87, "learning_rate": 2.9066197183098594e-05, "loss": 5.924, "step": 1667 }, { "epoch": 1.87, "learning_rate": 2.9064788732394367e-05, "loss": 6.1025, "step": 1668 }, { "epoch": 1.87, "learning_rate": 2.9063380281690143e-05, "loss": 6.0255, "step": 1669 }, { "epoch": 1.87, "learning_rate": 2.9061971830985916e-05, "loss": 5.9941, "step": 1670 }, { "epoch": 1.87, "learning_rate": 2.906056338028169e-05, "loss": 5.9321, "step": 1671 }, { "epoch": 1.87, "learning_rate": 2.9059154929577465e-05, "loss": 5.9785, "step": 1672 }, { "epoch": 1.88, "learning_rate": 2.905774647887324e-05, "loss": 6.1911, "step": 1673 }, { "epoch": 1.88, "learning_rate": 2.9056338028169015e-05, "loss": 6.0033, "step": 1674 }, { "epoch": 1.88, "learning_rate": 2.9054929577464788e-05, "loss": 5.9987, "step": 1675 }, { "epoch": 1.88, "learning_rate": 2.9053521126760564e-05, "loss": 6.1161, "step": 1676 }, { "epoch": 1.88, "learning_rate": 2.9052112676056337e-05, "loss": 6.0699, "step": 1677 }, { "epoch": 1.88, "learning_rate": 2.9050704225352113e-05, "loss": 5.8009, "step": 1678 }, { "epoch": 1.88, "learning_rate": 2.904929577464789e-05, "loss": 6.1334, "step": 1679 }, { "epoch": 1.88, "learning_rate": 2.9047887323943662e-05, "loss": 5.9756, "step": 1680 }, { "epoch": 1.88, "learning_rate": 2.904647887323944e-05, "loss": 6.0275, "step": 1681 }, { "epoch": 1.89, "learning_rate": 2.9045070422535212e-05, "loss": 5.866, "step": 1682 }, { "epoch": 1.89, "learning_rate": 2.9043661971830988e-05, "loss": 6.1578, "step": 1683 }, { "epoch": 1.89, "learning_rate": 2.904225352112676e-05, "loss": 5.8021, "step": 1684 }, { "epoch": 1.89, "learning_rate": 2.9040845070422537e-05, "loss": 5.7749, "step": 1685 }, { "epoch": 1.89, "learning_rate": 2.903943661971831e-05, "loss": 5.8918, "step": 1686 }, { "epoch": 1.89, "learning_rate": 2.9038028169014083e-05, "loss": 5.8616, "step": 1687 }, { "epoch": 1.89, "learning_rate": 2.903661971830986e-05, "loss": 5.7828, "step": 1688 }, { "epoch": 1.89, "learning_rate": 2.9035211267605633e-05, "loss": 5.9387, "step": 1689 }, { "epoch": 1.89, "learning_rate": 2.903380281690141e-05, "loss": 5.8401, "step": 1690 }, { "epoch": 1.9, "learning_rate": 2.9032394366197182e-05, "loss": 6.1115, "step": 1691 }, { "epoch": 1.9, "learning_rate": 2.9030985915492958e-05, "loss": 5.473, "step": 1692 }, { "epoch": 1.9, "learning_rate": 2.9029577464788734e-05, "loss": 6.2981, "step": 1693 }, { "epoch": 1.9, "learning_rate": 2.9028169014084507e-05, "loss": 6.2871, "step": 1694 }, { "epoch": 1.9, "learning_rate": 2.9026760563380284e-05, "loss": 6.2449, "step": 1695 }, { "epoch": 1.9, "learning_rate": 2.9025352112676057e-05, "loss": 6.231, "step": 1696 }, { "epoch": 1.9, "learning_rate": 2.9023943661971833e-05, "loss": 6.1089, "step": 1697 }, { "epoch": 1.9, "learning_rate": 2.9022535211267606e-05, "loss": 6.1017, "step": 1698 }, { "epoch": 1.9, "learning_rate": 2.9021126760563382e-05, "loss": 6.1276, "step": 1699 }, { "epoch": 1.91, "learning_rate": 2.9019718309859155e-05, "loss": 6.0436, "step": 1700 }, { "epoch": 1.91, "learning_rate": 2.901830985915493e-05, "loss": 6.069, "step": 1701 }, { "epoch": 1.91, "learning_rate": 2.9016901408450705e-05, "loss": 6.0546, "step": 1702 }, { "epoch": 1.91, "learning_rate": 2.9015492957746477e-05, "loss": 6.0379, "step": 1703 }, { "epoch": 1.91, "learning_rate": 2.9014084507042254e-05, "loss": 6.1858, "step": 1704 }, { "epoch": 1.91, "learning_rate": 2.9012676056338027e-05, "loss": 6.0262, "step": 1705 }, { "epoch": 1.91, "learning_rate": 2.9011267605633803e-05, "loss": 5.8768, "step": 1706 }, { "epoch": 1.91, "learning_rate": 2.9009859154929576e-05, "loss": 5.9884, "step": 1707 }, { "epoch": 1.91, "learning_rate": 2.9008450704225356e-05, "loss": 6.0383, "step": 1708 }, { "epoch": 1.92, "learning_rate": 2.900704225352113e-05, "loss": 6.0388, "step": 1709 }, { "epoch": 1.92, "learning_rate": 2.90056338028169e-05, "loss": 5.9414, "step": 1710 }, { "epoch": 1.92, "learning_rate": 2.9004225352112678e-05, "loss": 5.9026, "step": 1711 }, { "epoch": 1.92, "learning_rate": 2.900281690140845e-05, "loss": 6.014, "step": 1712 }, { "epoch": 1.92, "learning_rate": 2.9001408450704227e-05, "loss": 5.9604, "step": 1713 }, { "epoch": 1.92, "learning_rate": 2.9e-05, "loss": 5.8868, "step": 1714 }, { "epoch": 1.92, "learning_rate": 2.8998591549295777e-05, "loss": 6.0206, "step": 1715 }, { "epoch": 1.92, "learning_rate": 2.899718309859155e-05, "loss": 6.0864, "step": 1716 }, { "epoch": 1.92, "learning_rate": 2.8995774647887326e-05, "loss": 6.1443, "step": 1717 }, { "epoch": 1.93, "learning_rate": 2.89943661971831e-05, "loss": 5.9528, "step": 1718 }, { "epoch": 1.93, "learning_rate": 2.899295774647887e-05, "loss": 5.9281, "step": 1719 }, { "epoch": 1.93, "learning_rate": 2.8991549295774648e-05, "loss": 5.8645, "step": 1720 }, { "epoch": 1.93, "learning_rate": 2.899014084507042e-05, "loss": 6.0346, "step": 1721 }, { "epoch": 1.93, "learning_rate": 2.8988732394366197e-05, "loss": 5.964, "step": 1722 }, { "epoch": 1.93, "learning_rate": 2.8987323943661974e-05, "loss": 5.9747, "step": 1723 }, { "epoch": 1.93, "learning_rate": 2.898591549295775e-05, "loss": 5.976, "step": 1724 }, { "epoch": 1.93, "learning_rate": 2.8984507042253523e-05, "loss": 6.0413, "step": 1725 }, { "epoch": 1.93, "learning_rate": 2.8983098591549296e-05, "loss": 5.775, "step": 1726 }, { "epoch": 1.94, "learning_rate": 2.8981690140845072e-05, "loss": 5.9934, "step": 1727 }, { "epoch": 1.94, "learning_rate": 2.8980281690140845e-05, "loss": 5.9974, "step": 1728 }, { "epoch": 1.94, "learning_rate": 2.897887323943662e-05, "loss": 5.8809, "step": 1729 }, { "epoch": 1.94, "learning_rate": 2.8977464788732394e-05, "loss": 6.0173, "step": 1730 }, { "epoch": 1.94, "learning_rate": 2.897605633802817e-05, "loss": 5.8258, "step": 1731 }, { "epoch": 1.94, "learning_rate": 2.8974647887323944e-05, "loss": 6.185, "step": 1732 }, { "epoch": 1.94, "learning_rate": 2.897323943661972e-05, "loss": 5.8519, "step": 1733 }, { "epoch": 1.94, "learning_rate": 2.8971830985915493e-05, "loss": 5.8446, "step": 1734 }, { "epoch": 1.95, "learning_rate": 2.8970422535211266e-05, "loss": 5.8754, "step": 1735 }, { "epoch": 1.95, "learning_rate": 2.8969014084507042e-05, "loss": 5.8957, "step": 1736 }, { "epoch": 1.95, "learning_rate": 2.8967605633802815e-05, "loss": 5.9727, "step": 1737 }, { "epoch": 1.95, "learning_rate": 2.8966197183098595e-05, "loss": 5.9213, "step": 1738 }, { "epoch": 1.95, "learning_rate": 2.8964788732394368e-05, "loss": 5.9713, "step": 1739 }, { "epoch": 1.95, "learning_rate": 2.8963380281690144e-05, "loss": 5.7956, "step": 1740 }, { "epoch": 1.95, "learning_rate": 2.8961971830985917e-05, "loss": 5.556, "step": 1741 }, { "epoch": 1.95, "learning_rate": 2.896056338028169e-05, "loss": 5.5466, "step": 1742 }, { "epoch": 1.95, "learning_rate": 2.8959154929577466e-05, "loss": 6.2041, "step": 1743 }, { "epoch": 1.96, "learning_rate": 2.895774647887324e-05, "loss": 6.1028, "step": 1744 }, { "epoch": 1.96, "learning_rate": 2.8956338028169016e-05, "loss": 6.2803, "step": 1745 }, { "epoch": 1.96, "learning_rate": 2.895492957746479e-05, "loss": 6.1702, "step": 1746 }, { "epoch": 1.96, "learning_rate": 2.8953521126760565e-05, "loss": 6.1599, "step": 1747 }, { "epoch": 1.96, "learning_rate": 2.8952112676056338e-05, "loss": 6.01, "step": 1748 }, { "epoch": 1.96, "learning_rate": 2.8950704225352114e-05, "loss": 5.98, "step": 1749 }, { "epoch": 1.96, "learning_rate": 2.8949295774647887e-05, "loss": 6.2947, "step": 1750 }, { "epoch": 1.96, "learning_rate": 2.894788732394366e-05, "loss": 6.001, "step": 1751 }, { "epoch": 1.96, "learning_rate": 2.8946478873239436e-05, "loss": 6.0582, "step": 1752 }, { "epoch": 1.97, "learning_rate": 2.8945070422535213e-05, "loss": 6.1292, "step": 1753 }, { "epoch": 1.97, "learning_rate": 2.894366197183099e-05, "loss": 6.0204, "step": 1754 }, { "epoch": 1.97, "learning_rate": 2.8942253521126762e-05, "loss": 5.8626, "step": 1755 }, { "epoch": 1.97, "learning_rate": 2.894084507042254e-05, "loss": 5.9242, "step": 1756 }, { "epoch": 1.97, "learning_rate": 2.893943661971831e-05, "loss": 6.035, "step": 1757 }, { "epoch": 1.97, "learning_rate": 2.8938028169014084e-05, "loss": 6.068, "step": 1758 }, { "epoch": 1.97, "learning_rate": 2.893661971830986e-05, "loss": 6.0139, "step": 1759 }, { "epoch": 1.97, "learning_rate": 2.8935211267605633e-05, "loss": 5.8642, "step": 1760 }, { "epoch": 1.97, "learning_rate": 2.893380281690141e-05, "loss": 6.1216, "step": 1761 }, { "epoch": 1.98, "learning_rate": 2.8932394366197183e-05, "loss": 5.9186, "step": 1762 }, { "epoch": 1.98, "learning_rate": 2.893098591549296e-05, "loss": 6.0292, "step": 1763 }, { "epoch": 1.98, "learning_rate": 2.8929577464788732e-05, "loss": 6.0365, "step": 1764 }, { "epoch": 1.98, "learning_rate": 2.892816901408451e-05, "loss": 5.8717, "step": 1765 }, { "epoch": 1.98, "learning_rate": 2.892676056338028e-05, "loss": 5.9989, "step": 1766 }, { "epoch": 1.98, "learning_rate": 2.8925352112676054e-05, "loss": 5.9163, "step": 1767 }, { "epoch": 1.98, "learning_rate": 2.8923943661971834e-05, "loss": 6.0269, "step": 1768 }, { "epoch": 1.98, "learning_rate": 2.8922535211267607e-05, "loss": 5.8766, "step": 1769 }, { "epoch": 1.98, "learning_rate": 2.8921126760563383e-05, "loss": 6.0761, "step": 1770 }, { "epoch": 1.99, "learning_rate": 2.8919718309859156e-05, "loss": 5.9186, "step": 1771 }, { "epoch": 1.99, "learning_rate": 2.8918309859154933e-05, "loss": 5.9118, "step": 1772 }, { "epoch": 1.99, "learning_rate": 2.8916901408450705e-05, "loss": 5.8939, "step": 1773 }, { "epoch": 1.99, "learning_rate": 2.891549295774648e-05, "loss": 6.0847, "step": 1774 }, { "epoch": 1.99, "learning_rate": 2.8914084507042255e-05, "loss": 5.987, "step": 1775 }, { "epoch": 1.99, "learning_rate": 2.8912676056338028e-05, "loss": 6.0162, "step": 1776 }, { "epoch": 1.99, "learning_rate": 2.8911267605633804e-05, "loss": 6.1322, "step": 1777 }, { "epoch": 1.99, "learning_rate": 2.8909859154929577e-05, "loss": 6.0875, "step": 1778 }, { "epoch": 1.99, "learning_rate": 2.8908450704225353e-05, "loss": 5.9622, "step": 1779 }, { "epoch": 2.0, "learning_rate": 2.8907042253521126e-05, "loss": 5.7307, "step": 1780 }, { "epoch": 2.0, "learning_rate": 2.8905633802816903e-05, "loss": 5.8522, "step": 1781 }, { "epoch": 2.0, "learning_rate": 2.8904225352112676e-05, "loss": 5.9204, "step": 1782 }, { "epoch": 2.0, "learning_rate": 2.890281690140845e-05, "loss": 5.8027, "step": 1783 }, { "epoch": 2.0, "learning_rate": 2.8901408450704228e-05, "loss": 4.9459, "step": 1784 }, { "epoch": 2.0, "learning_rate": 2.89e-05, "loss": 6.1994, "step": 1785 }, { "epoch": 2.0, "learning_rate": 2.8898591549295777e-05, "loss": 5.8571, "step": 1786 }, { "epoch": 2.0, "learning_rate": 2.889718309859155e-05, "loss": 5.9837, "step": 1787 }, { "epoch": 2.0, "learning_rate": 2.8895774647887327e-05, "loss": 6.0824, "step": 1788 }, { "epoch": 2.01, "learning_rate": 2.88943661971831e-05, "loss": 6.0676, "step": 1789 }, { "epoch": 2.01, "learning_rate": 2.8892957746478873e-05, "loss": 5.903, "step": 1790 }, { "epoch": 2.01, "learning_rate": 2.889154929577465e-05, "loss": 6.0827, "step": 1791 }, { "epoch": 2.01, "learning_rate": 2.8890140845070422e-05, "loss": 5.9821, "step": 1792 }, { "epoch": 2.01, "learning_rate": 2.8888732394366198e-05, "loss": 5.9209, "step": 1793 }, { "epoch": 2.01, "learning_rate": 2.888732394366197e-05, "loss": 5.9597, "step": 1794 }, { "epoch": 2.01, "learning_rate": 2.8885915492957748e-05, "loss": 6.0573, "step": 1795 }, { "epoch": 2.01, "learning_rate": 2.888450704225352e-05, "loss": 5.9782, "step": 1796 }, { "epoch": 2.01, "learning_rate": 2.8883098591549297e-05, "loss": 5.934, "step": 1797 }, { "epoch": 2.02, "learning_rate": 2.888169014084507e-05, "loss": 5.8453, "step": 1798 }, { "epoch": 2.02, "learning_rate": 2.8880281690140846e-05, "loss": 6.0885, "step": 1799 }, { "epoch": 2.02, "learning_rate": 2.8878873239436622e-05, "loss": 5.9105, "step": 1800 }, { "epoch": 2.02, "learning_rate": 2.8877464788732395e-05, "loss": 6.0729, "step": 1801 }, { "epoch": 2.02, "learning_rate": 2.887605633802817e-05, "loss": 6.2503, "step": 1802 }, { "epoch": 2.02, "learning_rate": 2.8874647887323945e-05, "loss": 6.0034, "step": 1803 }, { "epoch": 2.02, "learning_rate": 2.887323943661972e-05, "loss": 5.8594, "step": 1804 }, { "epoch": 2.02, "learning_rate": 2.8871830985915494e-05, "loss": 5.776, "step": 1805 }, { "epoch": 2.02, "learning_rate": 2.8870422535211267e-05, "loss": 5.9168, "step": 1806 }, { "epoch": 2.03, "learning_rate": 2.8869014084507043e-05, "loss": 5.9309, "step": 1807 }, { "epoch": 2.03, "learning_rate": 2.8867605633802816e-05, "loss": 5.8879, "step": 1808 }, { "epoch": 2.03, "learning_rate": 2.8866197183098592e-05, "loss": 5.8001, "step": 1809 }, { "epoch": 2.03, "learning_rate": 2.8864788732394365e-05, "loss": 5.8309, "step": 1810 }, { "epoch": 2.03, "learning_rate": 2.8863380281690142e-05, "loss": 5.8391, "step": 1811 }, { "epoch": 2.03, "learning_rate": 2.8861971830985915e-05, "loss": 5.8729, "step": 1812 }, { "epoch": 2.03, "learning_rate": 2.8860563380281688e-05, "loss": 5.8651, "step": 1813 }, { "epoch": 2.03, "learning_rate": 2.8859154929577467e-05, "loss": 5.9036, "step": 1814 }, { "epoch": 2.03, "learning_rate": 2.885774647887324e-05, "loss": 6.0393, "step": 1815 }, { "epoch": 2.04, "learning_rate": 2.8856338028169017e-05, "loss": 5.8909, "step": 1816 }, { "epoch": 2.04, "learning_rate": 2.885492957746479e-05, "loss": 5.8818, "step": 1817 }, { "epoch": 2.04, "learning_rate": 2.8853521126760566e-05, "loss": 5.898, "step": 1818 }, { "epoch": 2.04, "learning_rate": 2.885211267605634e-05, "loss": 5.8306, "step": 1819 }, { "epoch": 2.04, "learning_rate": 2.8850704225352115e-05, "loss": 5.6477, "step": 1820 }, { "epoch": 2.04, "learning_rate": 2.8849295774647888e-05, "loss": 5.889, "step": 1821 }, { "epoch": 2.04, "learning_rate": 2.884788732394366e-05, "loss": 5.8112, "step": 1822 }, { "epoch": 2.04, "learning_rate": 2.8846478873239437e-05, "loss": 5.771, "step": 1823 }, { "epoch": 2.04, "learning_rate": 2.884507042253521e-05, "loss": 5.8213, "step": 1824 }, { "epoch": 2.05, "learning_rate": 2.8843661971830987e-05, "loss": 5.794, "step": 1825 }, { "epoch": 2.05, "learning_rate": 2.884225352112676e-05, "loss": 6.0089, "step": 1826 }, { "epoch": 2.05, "learning_rate": 2.8840845070422536e-05, "loss": 5.773, "step": 1827 }, { "epoch": 2.05, "learning_rate": 2.883943661971831e-05, "loss": 5.9772, "step": 1828 }, { "epoch": 2.05, "learning_rate": 2.8838028169014085e-05, "loss": 5.7183, "step": 1829 }, { "epoch": 2.05, "learning_rate": 2.883661971830986e-05, "loss": 5.8214, "step": 1830 }, { "epoch": 2.05, "learning_rate": 2.8835211267605634e-05, "loss": 5.8222, "step": 1831 }, { "epoch": 2.05, "learning_rate": 2.883380281690141e-05, "loss": 5.5481, "step": 1832 }, { "epoch": 2.05, "learning_rate": 2.8832394366197184e-05, "loss": 5.6639, "step": 1833 }, { "epoch": 2.06, "learning_rate": 2.883098591549296e-05, "loss": 5.162, "step": 1834 }, { "epoch": 2.06, "learning_rate": 2.8829577464788733e-05, "loss": 6.3365, "step": 1835 }, { "epoch": 2.06, "learning_rate": 2.882816901408451e-05, "loss": 6.2642, "step": 1836 }, { "epoch": 2.06, "learning_rate": 2.8826760563380282e-05, "loss": 6.0377, "step": 1837 }, { "epoch": 2.06, "learning_rate": 2.8825352112676055e-05, "loss": 6.1595, "step": 1838 }, { "epoch": 2.06, "learning_rate": 2.882394366197183e-05, "loss": 6.1345, "step": 1839 }, { "epoch": 2.06, "learning_rate": 2.8822535211267604e-05, "loss": 5.9267, "step": 1840 }, { "epoch": 2.06, "learning_rate": 2.882112676056338e-05, "loss": 6.1211, "step": 1841 }, { "epoch": 2.07, "learning_rate": 2.8819718309859154e-05, "loss": 5.8519, "step": 1842 }, { "epoch": 2.07, "learning_rate": 2.881830985915493e-05, "loss": 5.8371, "step": 1843 }, { "epoch": 2.07, "learning_rate": 2.8816901408450706e-05, "loss": 6.0094, "step": 1844 }, { "epoch": 2.07, "learning_rate": 2.881549295774648e-05, "loss": 5.9379, "step": 1845 }, { "epoch": 2.07, "learning_rate": 2.8814084507042256e-05, "loss": 5.8732, "step": 1846 }, { "epoch": 2.07, "learning_rate": 2.881267605633803e-05, "loss": 5.8584, "step": 1847 }, { "epoch": 2.07, "learning_rate": 2.8811267605633805e-05, "loss": 5.8848, "step": 1848 }, { "epoch": 2.07, "learning_rate": 2.8809859154929578e-05, "loss": 5.9606, "step": 1849 }, { "epoch": 2.07, "learning_rate": 2.8808450704225354e-05, "loss": 5.84, "step": 1850 }, { "epoch": 2.08, "learning_rate": 2.8807042253521127e-05, "loss": 6.0211, "step": 1851 }, { "epoch": 2.08, "learning_rate": 2.8805633802816904e-05, "loss": 5.8018, "step": 1852 }, { "epoch": 2.08, "learning_rate": 2.8804225352112676e-05, "loss": 6.0281, "step": 1853 }, { "epoch": 2.08, "learning_rate": 2.880281690140845e-05, "loss": 5.9569, "step": 1854 }, { "epoch": 2.08, "learning_rate": 2.8801408450704226e-05, "loss": 6.1381, "step": 1855 }, { "epoch": 2.08, "learning_rate": 2.88e-05, "loss": 5.9415, "step": 1856 }, { "epoch": 2.08, "learning_rate": 2.8798591549295775e-05, "loss": 5.9612, "step": 1857 }, { "epoch": 2.08, "learning_rate": 2.8797183098591548e-05, "loss": 5.8194, "step": 1858 }, { "epoch": 2.08, "learning_rate": 2.8795774647887328e-05, "loss": 6.1103, "step": 1859 }, { "epoch": 2.09, "learning_rate": 2.87943661971831e-05, "loss": 5.9647, "step": 1860 }, { "epoch": 2.09, "learning_rate": 2.8792957746478874e-05, "loss": 6.0057, "step": 1861 }, { "epoch": 2.09, "learning_rate": 2.879154929577465e-05, "loss": 5.8244, "step": 1862 }, { "epoch": 2.09, "learning_rate": 2.8790140845070423e-05, "loss": 5.7268, "step": 1863 }, { "epoch": 2.09, "learning_rate": 2.87887323943662e-05, "loss": 5.9277, "step": 1864 }, { "epoch": 2.09, "learning_rate": 2.8787323943661972e-05, "loss": 5.9162, "step": 1865 }, { "epoch": 2.09, "learning_rate": 2.878591549295775e-05, "loss": 5.8096, "step": 1866 }, { "epoch": 2.09, "learning_rate": 2.878450704225352e-05, "loss": 6.024, "step": 1867 }, { "epoch": 2.09, "learning_rate": 2.8783098591549298e-05, "loss": 5.9118, "step": 1868 }, { "epoch": 2.1, "learning_rate": 2.878169014084507e-05, "loss": 5.8889, "step": 1869 }, { "epoch": 2.1, "learning_rate": 2.8780281690140844e-05, "loss": 5.8264, "step": 1870 }, { "epoch": 2.1, "learning_rate": 2.877887323943662e-05, "loss": 5.9722, "step": 1871 }, { "epoch": 2.1, "learning_rate": 2.8777464788732393e-05, "loss": 5.7185, "step": 1872 }, { "epoch": 2.1, "learning_rate": 2.877605633802817e-05, "loss": 5.9989, "step": 1873 }, { "epoch": 2.1, "learning_rate": 2.8774647887323946e-05, "loss": 5.883, "step": 1874 }, { "epoch": 2.1, "learning_rate": 2.8773239436619722e-05, "loss": 5.9111, "step": 1875 }, { "epoch": 2.1, "learning_rate": 2.8771830985915495e-05, "loss": 5.8808, "step": 1876 }, { "epoch": 2.1, "learning_rate": 2.8770422535211268e-05, "loss": 5.8345, "step": 1877 }, { "epoch": 2.11, "learning_rate": 2.8769014084507044e-05, "loss": 5.6619, "step": 1878 }, { "epoch": 2.11, "learning_rate": 2.8767605633802817e-05, "loss": 5.9856, "step": 1879 }, { "epoch": 2.11, "learning_rate": 2.8766197183098593e-05, "loss": 5.7821, "step": 1880 }, { "epoch": 2.11, "learning_rate": 2.8764788732394366e-05, "loss": 5.7517, "step": 1881 }, { "epoch": 2.11, "learning_rate": 2.8763380281690143e-05, "loss": 5.7488, "step": 1882 }, { "epoch": 2.11, "learning_rate": 2.8761971830985916e-05, "loss": 5.3296, "step": 1883 }, { "epoch": 2.11, "learning_rate": 2.8760563380281692e-05, "loss": 5.2088, "step": 1884 }, { "epoch": 2.11, "learning_rate": 2.8759154929577465e-05, "loss": 6.1974, "step": 1885 }, { "epoch": 2.11, "learning_rate": 2.8757746478873238e-05, "loss": 6.1445, "step": 1886 }, { "epoch": 2.12, "learning_rate": 2.8756338028169014e-05, "loss": 6.2407, "step": 1887 }, { "epoch": 2.12, "learning_rate": 2.8754929577464787e-05, "loss": 6.0791, "step": 1888 }, { "epoch": 2.12, "learning_rate": 2.8753521126760567e-05, "loss": 6.0178, "step": 1889 }, { "epoch": 2.12, "learning_rate": 2.875211267605634e-05, "loss": 6.0032, "step": 1890 }, { "epoch": 2.12, "learning_rate": 2.8750704225352116e-05, "loss": 5.9402, "step": 1891 }, { "epoch": 2.12, "learning_rate": 2.874929577464789e-05, "loss": 5.8831, "step": 1892 }, { "epoch": 2.12, "learning_rate": 2.8747887323943662e-05, "loss": 6.2141, "step": 1893 }, { "epoch": 2.12, "learning_rate": 2.874647887323944e-05, "loss": 6.0088, "step": 1894 }, { "epoch": 2.12, "learning_rate": 2.874507042253521e-05, "loss": 5.7848, "step": 1895 }, { "epoch": 2.13, "learning_rate": 2.8743661971830988e-05, "loss": 5.9547, "step": 1896 }, { "epoch": 2.13, "learning_rate": 2.874225352112676e-05, "loss": 5.7837, "step": 1897 }, { "epoch": 2.13, "learning_rate": 2.8740845070422537e-05, "loss": 5.7686, "step": 1898 }, { "epoch": 2.13, "learning_rate": 2.873943661971831e-05, "loss": 5.7459, "step": 1899 }, { "epoch": 2.13, "learning_rate": 2.8738028169014086e-05, "loss": 5.67, "step": 1900 }, { "epoch": 2.13, "learning_rate": 2.873661971830986e-05, "loss": 5.8458, "step": 1901 }, { "epoch": 2.13, "learning_rate": 2.8735211267605632e-05, "loss": 5.997, "step": 1902 }, { "epoch": 2.13, "learning_rate": 2.873380281690141e-05, "loss": 5.8628, "step": 1903 }, { "epoch": 2.13, "learning_rate": 2.873239436619718e-05, "loss": 5.8621, "step": 1904 }, { "epoch": 2.14, "learning_rate": 2.873098591549296e-05, "loss": 5.7989, "step": 1905 }, { "epoch": 2.14, "learning_rate": 2.8729577464788734e-05, "loss": 5.8548, "step": 1906 }, { "epoch": 2.14, "learning_rate": 2.872816901408451e-05, "loss": 5.8664, "step": 1907 }, { "epoch": 2.14, "learning_rate": 2.8726760563380283e-05, "loss": 5.9371, "step": 1908 }, { "epoch": 2.14, "learning_rate": 2.8725352112676056e-05, "loss": 5.8656, "step": 1909 }, { "epoch": 2.14, "learning_rate": 2.8723943661971833e-05, "loss": 5.8873, "step": 1910 }, { "epoch": 2.14, "learning_rate": 2.8722535211267605e-05, "loss": 5.7808, "step": 1911 }, { "epoch": 2.14, "learning_rate": 2.8721126760563382e-05, "loss": 5.9482, "step": 1912 }, { "epoch": 2.14, "learning_rate": 2.8719718309859155e-05, "loss": 5.9546, "step": 1913 }, { "epoch": 2.15, "learning_rate": 2.871830985915493e-05, "loss": 5.8413, "step": 1914 }, { "epoch": 2.15, "learning_rate": 2.8716901408450704e-05, "loss": 5.8225, "step": 1915 }, { "epoch": 2.15, "learning_rate": 2.871549295774648e-05, "loss": 5.8411, "step": 1916 }, { "epoch": 2.15, "learning_rate": 2.8714084507042253e-05, "loss": 5.7652, "step": 1917 }, { "epoch": 2.15, "learning_rate": 2.8712676056338026e-05, "loss": 5.8499, "step": 1918 }, { "epoch": 2.15, "learning_rate": 2.8711267605633803e-05, "loss": 5.87, "step": 1919 }, { "epoch": 2.15, "learning_rate": 2.870985915492958e-05, "loss": 5.9712, "step": 1920 }, { "epoch": 2.15, "learning_rate": 2.8708450704225355e-05, "loss": 5.613, "step": 1921 }, { "epoch": 2.15, "learning_rate": 2.8707042253521128e-05, "loss": 5.8087, "step": 1922 }, { "epoch": 2.16, "learning_rate": 2.8705633802816904e-05, "loss": 5.6964, "step": 1923 }, { "epoch": 2.16, "learning_rate": 2.8704225352112677e-05, "loss": 5.7242, "step": 1924 }, { "epoch": 2.16, "learning_rate": 2.870281690140845e-05, "loss": 6.0036, "step": 1925 }, { "epoch": 2.16, "learning_rate": 2.8701408450704227e-05, "loss": 5.767, "step": 1926 }, { "epoch": 2.16, "learning_rate": 2.87e-05, "loss": 5.6924, "step": 1927 }, { "epoch": 2.16, "learning_rate": 2.8698591549295776e-05, "loss": 5.49, "step": 1928 }, { "epoch": 2.16, "learning_rate": 2.869718309859155e-05, "loss": 5.6426, "step": 1929 }, { "epoch": 2.16, "learning_rate": 2.8695774647887325e-05, "loss": 5.6094, "step": 1930 }, { "epoch": 2.16, "learning_rate": 2.8694366197183098e-05, "loss": 5.5626, "step": 1931 }, { "epoch": 2.17, "learning_rate": 2.869295774647887e-05, "loss": 5.3004, "step": 1932 }, { "epoch": 2.17, "learning_rate": 2.8691549295774647e-05, "loss": 5.6664, "step": 1933 }, { "epoch": 2.17, "learning_rate": 2.869014084507042e-05, "loss": 5.5065, "step": 1934 }, { "epoch": 2.17, "learning_rate": 2.86887323943662e-05, "loss": 6.3115, "step": 1935 }, { "epoch": 2.17, "learning_rate": 2.8687323943661973e-05, "loss": 6.2473, "step": 1936 }, { "epoch": 2.17, "learning_rate": 2.868591549295775e-05, "loss": 6.0691, "step": 1937 }, { "epoch": 2.17, "learning_rate": 2.8684507042253522e-05, "loss": 5.9342, "step": 1938 }, { "epoch": 2.17, "learning_rate": 2.86830985915493e-05, "loss": 6.089, "step": 1939 }, { "epoch": 2.17, "learning_rate": 2.868169014084507e-05, "loss": 6.2426, "step": 1940 }, { "epoch": 2.18, "learning_rate": 2.8680281690140845e-05, "loss": 6.0406, "step": 1941 }, { "epoch": 2.18, "learning_rate": 2.867887323943662e-05, "loss": 5.8842, "step": 1942 }, { "epoch": 2.18, "learning_rate": 2.8677464788732394e-05, "loss": 5.896, "step": 1943 }, { "epoch": 2.18, "learning_rate": 2.867605633802817e-05, "loss": 5.7995, "step": 1944 }, { "epoch": 2.18, "learning_rate": 2.8674647887323943e-05, "loss": 6.0995, "step": 1945 }, { "epoch": 2.18, "learning_rate": 2.867323943661972e-05, "loss": 5.8436, "step": 1946 }, { "epoch": 2.18, "learning_rate": 2.8671830985915492e-05, "loss": 5.8292, "step": 1947 }, { "epoch": 2.18, "learning_rate": 2.8670422535211265e-05, "loss": 5.74, "step": 1948 }, { "epoch": 2.18, "learning_rate": 2.866901408450704e-05, "loss": 6.1615, "step": 1949 }, { "epoch": 2.19, "learning_rate": 2.8667605633802818e-05, "loss": 5.9368, "step": 1950 }, { "epoch": 2.19, "learning_rate": 2.8666197183098594e-05, "loss": 5.7045, "step": 1951 }, { "epoch": 2.19, "learning_rate": 2.8664788732394367e-05, "loss": 5.7292, "step": 1952 }, { "epoch": 2.19, "learning_rate": 2.8663380281690144e-05, "loss": 5.8332, "step": 1953 }, { "epoch": 2.19, "learning_rate": 2.8661971830985917e-05, "loss": 5.9824, "step": 1954 }, { "epoch": 2.19, "learning_rate": 2.8660563380281693e-05, "loss": 5.9189, "step": 1955 }, { "epoch": 2.19, "learning_rate": 2.8659154929577466e-05, "loss": 5.8587, "step": 1956 }, { "epoch": 2.19, "learning_rate": 2.865774647887324e-05, "loss": 5.8337, "step": 1957 }, { "epoch": 2.2, "learning_rate": 2.8656338028169015e-05, "loss": 5.7901, "step": 1958 }, { "epoch": 2.2, "learning_rate": 2.8654929577464788e-05, "loss": 5.7059, "step": 1959 }, { "epoch": 2.2, "learning_rate": 2.8653521126760564e-05, "loss": 5.9823, "step": 1960 }, { "epoch": 2.2, "learning_rate": 2.8652112676056337e-05, "loss": 5.7012, "step": 1961 }, { "epoch": 2.2, "learning_rate": 2.8650704225352114e-05, "loss": 5.7561, "step": 1962 }, { "epoch": 2.2, "learning_rate": 2.8649295774647887e-05, "loss": 5.8505, "step": 1963 }, { "epoch": 2.2, "learning_rate": 2.864788732394366e-05, "loss": 5.7051, "step": 1964 }, { "epoch": 2.2, "learning_rate": 2.864647887323944e-05, "loss": 5.7869, "step": 1965 }, { "epoch": 2.2, "learning_rate": 2.8645070422535212e-05, "loss": 5.782, "step": 1966 }, { "epoch": 2.21, "learning_rate": 2.864366197183099e-05, "loss": 5.7431, "step": 1967 }, { "epoch": 2.21, "learning_rate": 2.864225352112676e-05, "loss": 5.7848, "step": 1968 }, { "epoch": 2.21, "learning_rate": 2.8640845070422538e-05, "loss": 5.9563, "step": 1969 }, { "epoch": 2.21, "learning_rate": 2.863943661971831e-05, "loss": 5.7153, "step": 1970 }, { "epoch": 2.21, "learning_rate": 2.8638028169014087e-05, "loss": 5.8247, "step": 1971 }, { "epoch": 2.21, "learning_rate": 2.863661971830986e-05, "loss": 5.7414, "step": 1972 }, { "epoch": 2.21, "learning_rate": 2.8635211267605633e-05, "loss": 5.6961, "step": 1973 }, { "epoch": 2.21, "learning_rate": 2.863380281690141e-05, "loss": 5.8037, "step": 1974 }, { "epoch": 2.21, "learning_rate": 2.8632394366197182e-05, "loss": 5.7879, "step": 1975 }, { "epoch": 2.22, "learning_rate": 2.863098591549296e-05, "loss": 5.6461, "step": 1976 }, { "epoch": 2.22, "learning_rate": 2.862957746478873e-05, "loss": 5.6381, "step": 1977 }, { "epoch": 2.22, "learning_rate": 2.8628169014084508e-05, "loss": 5.7958, "step": 1978 }, { "epoch": 2.22, "learning_rate": 2.862676056338028e-05, "loss": 5.4361, "step": 1979 }, { "epoch": 2.22, "learning_rate": 2.8625352112676057e-05, "loss": 5.6355, "step": 1980 }, { "epoch": 2.22, "learning_rate": 2.8623943661971833e-05, "loss": 5.6566, "step": 1981 }, { "epoch": 2.22, "learning_rate": 2.8622535211267606e-05, "loss": 5.6209, "step": 1982 }, { "epoch": 2.22, "learning_rate": 2.8621126760563383e-05, "loss": 5.6118, "step": 1983 }, { "epoch": 2.22, "learning_rate": 2.8619718309859156e-05, "loss": 5.1321, "step": 1984 }, { "epoch": 2.23, "learning_rate": 2.8618309859154932e-05, "loss": 6.2108, "step": 1985 }, { "epoch": 2.23, "learning_rate": 2.8616901408450705e-05, "loss": 6.2084, "step": 1986 }, { "epoch": 2.23, "learning_rate": 2.861549295774648e-05, "loss": 6.1595, "step": 1987 }, { "epoch": 2.23, "learning_rate": 2.8614084507042254e-05, "loss": 5.9365, "step": 1988 }, { "epoch": 2.23, "learning_rate": 2.8612676056338027e-05, "loss": 5.9935, "step": 1989 }, { "epoch": 2.23, "learning_rate": 2.8611267605633804e-05, "loss": 6.0562, "step": 1990 }, { "epoch": 2.23, "learning_rate": 2.8609859154929576e-05, "loss": 6.0055, "step": 1991 }, { "epoch": 2.23, "learning_rate": 2.8608450704225353e-05, "loss": 6.061, "step": 1992 }, { "epoch": 2.23, "learning_rate": 2.8607042253521126e-05, "loss": 5.8255, "step": 1993 }, { "epoch": 2.24, "learning_rate": 2.8605633802816902e-05, "loss": 5.9146, "step": 1994 }, { "epoch": 2.24, "learning_rate": 2.860422535211268e-05, "loss": 6.0246, "step": 1995 }, { "epoch": 2.24, "learning_rate": 2.860281690140845e-05, "loss": 5.7166, "step": 1996 }, { "epoch": 2.24, "learning_rate": 2.8601408450704228e-05, "loss": 5.7648, "step": 1997 }, { "epoch": 2.24, "learning_rate": 2.86e-05, "loss": 5.7172, "step": 1998 }, { "epoch": 2.24, "learning_rate": 2.8598591549295777e-05, "loss": 5.9561, "step": 1999 }, { "epoch": 2.24, "learning_rate": 2.859718309859155e-05, "loss": 5.6615, "step": 2000 }, { "epoch": 2.24, "learning_rate": 2.8595774647887326e-05, "loss": 5.7574, "step": 2001 }, { "epoch": 2.24, "learning_rate": 2.85943661971831e-05, "loss": 5.6408, "step": 2002 }, { "epoch": 2.25, "learning_rate": 2.8592957746478875e-05, "loss": 5.7802, "step": 2003 }, { "epoch": 2.25, "learning_rate": 2.859154929577465e-05, "loss": 5.8991, "step": 2004 }, { "epoch": 2.25, "learning_rate": 2.859014084507042e-05, "loss": 5.7627, "step": 2005 }, { "epoch": 2.25, "learning_rate": 2.8588732394366198e-05, "loss": 6.0348, "step": 2006 }, { "epoch": 2.25, "learning_rate": 2.858732394366197e-05, "loss": 5.8023, "step": 2007 }, { "epoch": 2.25, "learning_rate": 2.8585915492957747e-05, "loss": 5.8579, "step": 2008 }, { "epoch": 2.25, "learning_rate": 2.858450704225352e-05, "loss": 5.8842, "step": 2009 }, { "epoch": 2.25, "learning_rate": 2.8583098591549296e-05, "loss": 5.8682, "step": 2010 }, { "epoch": 2.25, "learning_rate": 2.8581690140845073e-05, "loss": 5.8831, "step": 2011 }, { "epoch": 2.26, "learning_rate": 2.8580281690140846e-05, "loss": 5.7566, "step": 2012 }, { "epoch": 2.26, "learning_rate": 2.8578873239436622e-05, "loss": 5.8043, "step": 2013 }, { "epoch": 2.26, "learning_rate": 2.8577464788732395e-05, "loss": 5.8309, "step": 2014 }, { "epoch": 2.26, "learning_rate": 2.857605633802817e-05, "loss": 5.9212, "step": 2015 }, { "epoch": 2.26, "learning_rate": 2.8574647887323944e-05, "loss": 5.7546, "step": 2016 }, { "epoch": 2.26, "learning_rate": 2.857323943661972e-05, "loss": 5.9025, "step": 2017 }, { "epoch": 2.26, "learning_rate": 2.8571830985915493e-05, "loss": 5.6982, "step": 2018 }, { "epoch": 2.26, "learning_rate": 2.857042253521127e-05, "loss": 5.9393, "step": 2019 }, { "epoch": 2.26, "learning_rate": 2.8569014084507043e-05, "loss": 5.9644, "step": 2020 }, { "epoch": 2.27, "learning_rate": 2.8567605633802816e-05, "loss": 5.7932, "step": 2021 }, { "epoch": 2.27, "learning_rate": 2.8566197183098592e-05, "loss": 5.862, "step": 2022 }, { "epoch": 2.27, "learning_rate": 2.8564788732394365e-05, "loss": 5.7996, "step": 2023 }, { "epoch": 2.27, "learning_rate": 2.856338028169014e-05, "loss": 5.6835, "step": 2024 }, { "epoch": 2.27, "learning_rate": 2.8561971830985914e-05, "loss": 5.7527, "step": 2025 }, { "epoch": 2.27, "learning_rate": 2.8560563380281694e-05, "loss": 5.7734, "step": 2026 }, { "epoch": 2.27, "learning_rate": 2.8559154929577467e-05, "loss": 5.853, "step": 2027 }, { "epoch": 2.27, "learning_rate": 2.855774647887324e-05, "loss": 5.6165, "step": 2028 }, { "epoch": 2.27, "learning_rate": 2.8556338028169016e-05, "loss": 5.6451, "step": 2029 }, { "epoch": 2.28, "learning_rate": 2.855492957746479e-05, "loss": 5.9104, "step": 2030 }, { "epoch": 2.28, "learning_rate": 2.8553521126760565e-05, "loss": 5.3484, "step": 2031 }, { "epoch": 2.28, "learning_rate": 2.8552112676056338e-05, "loss": 5.6619, "step": 2032 }, { "epoch": 2.28, "learning_rate": 2.8550704225352115e-05, "loss": 5.2984, "step": 2033 }, { "epoch": 2.28, "learning_rate": 2.8549295774647888e-05, "loss": 5.6132, "step": 2034 }, { "epoch": 2.28, "learning_rate": 2.854788732394366e-05, "loss": 6.1405, "step": 2035 }, { "epoch": 2.28, "learning_rate": 2.8546478873239437e-05, "loss": 6.1075, "step": 2036 }, { "epoch": 2.28, "learning_rate": 2.854507042253521e-05, "loss": 6.0858, "step": 2037 }, { "epoch": 2.28, "learning_rate": 2.8543661971830986e-05, "loss": 6.0523, "step": 2038 }, { "epoch": 2.29, "learning_rate": 2.854225352112676e-05, "loss": 5.8812, "step": 2039 }, { "epoch": 2.29, "learning_rate": 2.8540845070422535e-05, "loss": 6.1216, "step": 2040 }, { "epoch": 2.29, "learning_rate": 2.8539436619718312e-05, "loss": 5.9399, "step": 2041 }, { "epoch": 2.29, "learning_rate": 2.8538028169014088e-05, "loss": 6.0216, "step": 2042 }, { "epoch": 2.29, "learning_rate": 2.853661971830986e-05, "loss": 6.0011, "step": 2043 }, { "epoch": 2.29, "learning_rate": 2.8535211267605634e-05, "loss": 5.6948, "step": 2044 }, { "epoch": 2.29, "learning_rate": 2.853380281690141e-05, "loss": 5.8894, "step": 2045 }, { "epoch": 2.29, "learning_rate": 2.8532394366197183e-05, "loss": 5.8007, "step": 2046 }, { "epoch": 2.29, "learning_rate": 2.853098591549296e-05, "loss": 6.0615, "step": 2047 }, { "epoch": 2.3, "learning_rate": 2.8529577464788732e-05, "loss": 5.6508, "step": 2048 }, { "epoch": 2.3, "learning_rate": 2.852816901408451e-05, "loss": 6.143, "step": 2049 }, { "epoch": 2.3, "learning_rate": 2.8526760563380282e-05, "loss": 5.8407, "step": 2050 }, { "epoch": 2.3, "learning_rate": 2.8525352112676055e-05, "loss": 5.8993, "step": 2051 }, { "epoch": 2.3, "learning_rate": 2.852394366197183e-05, "loss": 5.7718, "step": 2052 }, { "epoch": 2.3, "learning_rate": 2.8522535211267604e-05, "loss": 5.7944, "step": 2053 }, { "epoch": 2.3, "learning_rate": 2.852112676056338e-05, "loss": 5.819, "step": 2054 }, { "epoch": 2.3, "learning_rate": 2.8519718309859153e-05, "loss": 5.8964, "step": 2055 }, { "epoch": 2.3, "learning_rate": 2.8518309859154933e-05, "loss": 5.6777, "step": 2056 }, { "epoch": 2.31, "learning_rate": 2.8516901408450706e-05, "loss": 5.8218, "step": 2057 }, { "epoch": 2.31, "learning_rate": 2.8515492957746482e-05, "loss": 5.5697, "step": 2058 }, { "epoch": 2.31, "learning_rate": 2.8514084507042255e-05, "loss": 5.9746, "step": 2059 }, { "epoch": 2.31, "learning_rate": 2.8512676056338028e-05, "loss": 5.8826, "step": 2060 }, { "epoch": 2.31, "learning_rate": 2.8511267605633804e-05, "loss": 6.0874, "step": 2061 }, { "epoch": 2.31, "learning_rate": 2.8509859154929577e-05, "loss": 5.9012, "step": 2062 }, { "epoch": 2.31, "learning_rate": 2.8508450704225354e-05, "loss": 5.7428, "step": 2063 }, { "epoch": 2.31, "learning_rate": 2.8507042253521127e-05, "loss": 6.0235, "step": 2064 }, { "epoch": 2.32, "learning_rate": 2.8505633802816903e-05, "loss": 5.8498, "step": 2065 }, { "epoch": 2.32, "learning_rate": 2.8504225352112676e-05, "loss": 5.6396, "step": 2066 }, { "epoch": 2.32, "learning_rate": 2.850281690140845e-05, "loss": 5.7857, "step": 2067 }, { "epoch": 2.32, "learning_rate": 2.8501408450704225e-05, "loss": 5.682, "step": 2068 }, { "epoch": 2.32, "learning_rate": 2.8499999999999998e-05, "loss": 5.8512, "step": 2069 }, { "epoch": 2.32, "learning_rate": 2.8498591549295775e-05, "loss": 5.7528, "step": 2070 }, { "epoch": 2.32, "learning_rate": 2.849718309859155e-05, "loss": 5.6986, "step": 2071 }, { "epoch": 2.32, "learning_rate": 2.8495774647887327e-05, "loss": 5.77, "step": 2072 }, { "epoch": 2.32, "learning_rate": 2.84943661971831e-05, "loss": 5.6122, "step": 2073 }, { "epoch": 2.33, "learning_rate": 2.8492957746478876e-05, "loss": 5.7251, "step": 2074 }, { "epoch": 2.33, "learning_rate": 2.849154929577465e-05, "loss": 5.7859, "step": 2075 }, { "epoch": 2.33, "learning_rate": 2.8490140845070422e-05, "loss": 5.6906, "step": 2076 }, { "epoch": 2.33, "learning_rate": 2.84887323943662e-05, "loss": 5.7066, "step": 2077 }, { "epoch": 2.33, "learning_rate": 2.848732394366197e-05, "loss": 5.7038, "step": 2078 }, { "epoch": 2.33, "learning_rate": 2.8485915492957748e-05, "loss": 5.4338, "step": 2079 }, { "epoch": 2.33, "learning_rate": 2.848450704225352e-05, "loss": 5.6034, "step": 2080 }, { "epoch": 2.33, "learning_rate": 2.8483098591549297e-05, "loss": 5.3847, "step": 2081 }, { "epoch": 2.33, "learning_rate": 2.848169014084507e-05, "loss": 5.2634, "step": 2082 }, { "epoch": 2.34, "learning_rate": 2.8480281690140843e-05, "loss": 5.2658, "step": 2083 }, { "epoch": 2.34, "learning_rate": 2.847887323943662e-05, "loss": 5.1592, "step": 2084 }, { "epoch": 2.34, "learning_rate": 2.8477464788732392e-05, "loss": 6.0809, "step": 2085 }, { "epoch": 2.34, "learning_rate": 2.8476056338028172e-05, "loss": 6.1528, "step": 2086 }, { "epoch": 2.34, "learning_rate": 2.8474647887323945e-05, "loss": 6.1449, "step": 2087 }, { "epoch": 2.34, "learning_rate": 2.847323943661972e-05, "loss": 6.0439, "step": 2088 }, { "epoch": 2.34, "learning_rate": 2.8471830985915494e-05, "loss": 5.9507, "step": 2089 }, { "epoch": 2.34, "learning_rate": 2.847042253521127e-05, "loss": 5.8721, "step": 2090 }, { "epoch": 2.34, "learning_rate": 2.8469014084507044e-05, "loss": 5.9599, "step": 2091 }, { "epoch": 2.35, "learning_rate": 2.8467605633802817e-05, "loss": 5.8351, "step": 2092 }, { "epoch": 2.35, "learning_rate": 2.8466197183098593e-05, "loss": 5.9343, "step": 2093 }, { "epoch": 2.35, "learning_rate": 2.8464788732394366e-05, "loss": 6.123, "step": 2094 }, { "epoch": 2.35, "learning_rate": 2.8463380281690142e-05, "loss": 5.867, "step": 2095 }, { "epoch": 2.35, "learning_rate": 2.8461971830985915e-05, "loss": 5.9136, "step": 2096 }, { "epoch": 2.35, "learning_rate": 2.846056338028169e-05, "loss": 6.0425, "step": 2097 }, { "epoch": 2.35, "learning_rate": 2.8459154929577464e-05, "loss": 6.0528, "step": 2098 }, { "epoch": 2.35, "learning_rate": 2.8457746478873237e-05, "loss": 5.7986, "step": 2099 }, { "epoch": 2.35, "learning_rate": 2.8456338028169014e-05, "loss": 5.8548, "step": 2100 }, { "epoch": 2.36, "learning_rate": 2.845492957746479e-05, "loss": 5.6312, "step": 2101 }, { "epoch": 2.36, "learning_rate": 2.8453521126760566e-05, "loss": 5.9511, "step": 2102 }, { "epoch": 2.36, "learning_rate": 2.845211267605634e-05, "loss": 5.9284, "step": 2103 }, { "epoch": 2.36, "learning_rate": 2.8450704225352116e-05, "loss": 5.75, "step": 2104 }, { "epoch": 2.36, "learning_rate": 2.844929577464789e-05, "loss": 5.8859, "step": 2105 }, { "epoch": 2.36, "learning_rate": 2.8447887323943665e-05, "loss": 5.9259, "step": 2106 }, { "epoch": 2.36, "learning_rate": 2.8446478873239438e-05, "loss": 5.6451, "step": 2107 }, { "epoch": 2.36, "learning_rate": 2.844507042253521e-05, "loss": 5.8041, "step": 2108 }, { "epoch": 2.36, "learning_rate": 2.8443661971830987e-05, "loss": 5.748, "step": 2109 }, { "epoch": 2.37, "learning_rate": 2.844225352112676e-05, "loss": 5.7324, "step": 2110 }, { "epoch": 2.37, "learning_rate": 2.8440845070422536e-05, "loss": 5.8477, "step": 2111 }, { "epoch": 2.37, "learning_rate": 2.843943661971831e-05, "loss": 5.6478, "step": 2112 }, { "epoch": 2.37, "learning_rate": 2.8438028169014086e-05, "loss": 5.7785, "step": 2113 }, { "epoch": 2.37, "learning_rate": 2.843661971830986e-05, "loss": 5.8865, "step": 2114 }, { "epoch": 2.37, "learning_rate": 2.843521126760563e-05, "loss": 5.6852, "step": 2115 }, { "epoch": 2.37, "learning_rate": 2.843380281690141e-05, "loss": 5.7384, "step": 2116 }, { "epoch": 2.37, "learning_rate": 2.8432394366197184e-05, "loss": 5.5511, "step": 2117 }, { "epoch": 2.37, "learning_rate": 2.843098591549296e-05, "loss": 5.7603, "step": 2118 }, { "epoch": 2.38, "learning_rate": 2.8429577464788733e-05, "loss": 5.7384, "step": 2119 }, { "epoch": 2.38, "learning_rate": 2.842816901408451e-05, "loss": 5.8607, "step": 2120 }, { "epoch": 2.38, "learning_rate": 2.8426760563380283e-05, "loss": 5.6934, "step": 2121 }, { "epoch": 2.38, "learning_rate": 2.842535211267606e-05, "loss": 5.7802, "step": 2122 }, { "epoch": 2.38, "learning_rate": 2.8423943661971832e-05, "loss": 5.7333, "step": 2123 }, { "epoch": 2.38, "learning_rate": 2.8422535211267605e-05, "loss": 5.764, "step": 2124 }, { "epoch": 2.38, "learning_rate": 2.842112676056338e-05, "loss": 5.6891, "step": 2125 }, { "epoch": 2.38, "learning_rate": 2.8419718309859154e-05, "loss": 5.6562, "step": 2126 }, { "epoch": 2.38, "learning_rate": 2.841830985915493e-05, "loss": 5.7912, "step": 2127 }, { "epoch": 2.39, "learning_rate": 2.8416901408450703e-05, "loss": 5.6383, "step": 2128 }, { "epoch": 2.39, "learning_rate": 2.841549295774648e-05, "loss": 5.6606, "step": 2129 }, { "epoch": 2.39, "learning_rate": 2.8414084507042253e-05, "loss": 5.543, "step": 2130 }, { "epoch": 2.39, "learning_rate": 2.8412676056338026e-05, "loss": 5.4169, "step": 2131 }, { "epoch": 2.39, "learning_rate": 2.8411267605633805e-05, "loss": 5.5031, "step": 2132 }, { "epoch": 2.39, "learning_rate": 2.840985915492958e-05, "loss": 5.5936, "step": 2133 }, { "epoch": 2.39, "learning_rate": 2.8408450704225355e-05, "loss": 5.3114, "step": 2134 }, { "epoch": 2.39, "learning_rate": 2.8407042253521128e-05, "loss": 6.0965, "step": 2135 }, { "epoch": 2.39, "learning_rate": 2.8405633802816904e-05, "loss": 6.228, "step": 2136 }, { "epoch": 2.4, "learning_rate": 2.8404225352112677e-05, "loss": 5.9357, "step": 2137 }, { "epoch": 2.4, "learning_rate": 2.840281690140845e-05, "loss": 5.8483, "step": 2138 }, { "epoch": 2.4, "learning_rate": 2.8401408450704226e-05, "loss": 5.9798, "step": 2139 }, { "epoch": 2.4, "learning_rate": 2.84e-05, "loss": 5.8991, "step": 2140 }, { "epoch": 2.4, "learning_rate": 2.8398591549295775e-05, "loss": 5.684, "step": 2141 }, { "epoch": 2.4, "learning_rate": 2.839718309859155e-05, "loss": 6.1346, "step": 2142 }, { "epoch": 2.4, "learning_rate": 2.8395774647887325e-05, "loss": 6.034, "step": 2143 }, { "epoch": 2.4, "learning_rate": 2.8394366197183098e-05, "loss": 5.6537, "step": 2144 }, { "epoch": 2.4, "learning_rate": 2.8392957746478874e-05, "loss": 5.7046, "step": 2145 }, { "epoch": 2.41, "learning_rate": 2.8391549295774647e-05, "loss": 5.8485, "step": 2146 }, { "epoch": 2.41, "learning_rate": 2.8390140845070423e-05, "loss": 5.8418, "step": 2147 }, { "epoch": 2.41, "learning_rate": 2.83887323943662e-05, "loss": 5.8244, "step": 2148 }, { "epoch": 2.41, "learning_rate": 2.8387323943661973e-05, "loss": 5.7888, "step": 2149 }, { "epoch": 2.41, "learning_rate": 2.838591549295775e-05, "loss": 5.9484, "step": 2150 }, { "epoch": 2.41, "learning_rate": 2.8384507042253522e-05, "loss": 5.9085, "step": 2151 }, { "epoch": 2.41, "learning_rate": 2.8383098591549298e-05, "loss": 5.9434, "step": 2152 }, { "epoch": 2.41, "learning_rate": 2.838169014084507e-05, "loss": 5.7778, "step": 2153 }, { "epoch": 2.41, "learning_rate": 2.8380281690140844e-05, "loss": 5.8803, "step": 2154 }, { "epoch": 2.42, "learning_rate": 2.837887323943662e-05, "loss": 5.9469, "step": 2155 }, { "epoch": 2.42, "learning_rate": 2.8377464788732393e-05, "loss": 5.7063, "step": 2156 }, { "epoch": 2.42, "learning_rate": 2.837605633802817e-05, "loss": 5.7086, "step": 2157 }, { "epoch": 2.42, "learning_rate": 2.8374647887323943e-05, "loss": 5.6479, "step": 2158 }, { "epoch": 2.42, "learning_rate": 2.837323943661972e-05, "loss": 5.7333, "step": 2159 }, { "epoch": 2.42, "learning_rate": 2.8371830985915492e-05, "loss": 5.9268, "step": 2160 }, { "epoch": 2.42, "learning_rate": 2.8370422535211268e-05, "loss": 5.7788, "step": 2161 }, { "epoch": 2.42, "learning_rate": 2.8369014084507045e-05, "loss": 5.6174, "step": 2162 }, { "epoch": 2.42, "learning_rate": 2.8367605633802817e-05, "loss": 5.8078, "step": 2163 }, { "epoch": 2.43, "learning_rate": 2.8366197183098594e-05, "loss": 5.7366, "step": 2164 }, { "epoch": 2.43, "learning_rate": 2.8364788732394367e-05, "loss": 5.7979, "step": 2165 }, { "epoch": 2.43, "learning_rate": 2.8363380281690143e-05, "loss": 5.6635, "step": 2166 }, { "epoch": 2.43, "learning_rate": 2.8361971830985916e-05, "loss": 5.6626, "step": 2167 }, { "epoch": 2.43, "learning_rate": 2.8360563380281692e-05, "loss": 5.724, "step": 2168 }, { "epoch": 2.43, "learning_rate": 2.8359154929577465e-05, "loss": 5.6971, "step": 2169 }, { "epoch": 2.43, "learning_rate": 2.8357746478873238e-05, "loss": 5.7653, "step": 2170 }, { "epoch": 2.43, "learning_rate": 2.8356338028169015e-05, "loss": 5.9734, "step": 2171 }, { "epoch": 2.43, "learning_rate": 2.8354929577464788e-05, "loss": 5.7484, "step": 2172 }, { "epoch": 2.44, "learning_rate": 2.8353521126760564e-05, "loss": 5.7015, "step": 2173 }, { "epoch": 2.44, "learning_rate": 2.8352112676056337e-05, "loss": 5.6011, "step": 2174 }, { "epoch": 2.44, "learning_rate": 2.8350704225352113e-05, "loss": 5.7701, "step": 2175 }, { "epoch": 2.44, "learning_rate": 2.8349295774647886e-05, "loss": 5.684, "step": 2176 }, { "epoch": 2.44, "learning_rate": 2.8347887323943666e-05, "loss": 5.9103, "step": 2177 }, { "epoch": 2.44, "learning_rate": 2.834647887323944e-05, "loss": 5.9756, "step": 2178 }, { "epoch": 2.44, "learning_rate": 2.8345070422535212e-05, "loss": 5.706, "step": 2179 }, { "epoch": 2.44, "learning_rate": 2.8343661971830988e-05, "loss": 5.6264, "step": 2180 }, { "epoch": 2.45, "learning_rate": 2.834225352112676e-05, "loss": 5.5747, "step": 2181 }, { "epoch": 2.45, "learning_rate": 2.8340845070422537e-05, "loss": 5.7545, "step": 2182 }, { "epoch": 2.45, "learning_rate": 2.833943661971831e-05, "loss": 5.2311, "step": 2183 }, { "epoch": 2.45, "learning_rate": 2.8338028169014087e-05, "loss": 5.2831, "step": 2184 }, { "epoch": 2.45, "learning_rate": 2.833661971830986e-05, "loss": 6.0746, "step": 2185 }, { "epoch": 2.45, "learning_rate": 2.8335211267605632e-05, "loss": 5.931, "step": 2186 }, { "epoch": 2.45, "learning_rate": 2.833380281690141e-05, "loss": 6.0348, "step": 2187 }, { "epoch": 2.45, "learning_rate": 2.8332394366197182e-05, "loss": 5.9807, "step": 2188 }, { "epoch": 2.45, "learning_rate": 2.8330985915492958e-05, "loss": 5.8757, "step": 2189 }, { "epoch": 2.46, "learning_rate": 2.832957746478873e-05, "loss": 5.7444, "step": 2190 }, { "epoch": 2.46, "learning_rate": 2.8328169014084507e-05, "loss": 6.0013, "step": 2191 }, { "epoch": 2.46, "learning_rate": 2.8326760563380284e-05, "loss": 5.9196, "step": 2192 }, { "epoch": 2.46, "learning_rate": 2.832535211267606e-05, "loss": 5.8775, "step": 2193 }, { "epoch": 2.46, "learning_rate": 2.8323943661971833e-05, "loss": 5.9269, "step": 2194 }, { "epoch": 2.46, "learning_rate": 2.8322535211267606e-05, "loss": 5.7042, "step": 2195 }, { "epoch": 2.46, "learning_rate": 2.8321126760563382e-05, "loss": 5.7475, "step": 2196 }, { "epoch": 2.46, "learning_rate": 2.8319718309859155e-05, "loss": 5.7585, "step": 2197 }, { "epoch": 2.46, "learning_rate": 2.831830985915493e-05, "loss": 5.6354, "step": 2198 }, { "epoch": 2.47, "learning_rate": 2.8316901408450704e-05, "loss": 5.8367, "step": 2199 }, { "epoch": 2.47, "learning_rate": 2.831549295774648e-05, "loss": 5.913, "step": 2200 }, { "epoch": 2.47, "learning_rate": 2.8314084507042254e-05, "loss": 5.8146, "step": 2201 }, { "epoch": 2.47, "learning_rate": 2.8312676056338027e-05, "loss": 5.712, "step": 2202 }, { "epoch": 2.47, "learning_rate": 2.8311267605633803e-05, "loss": 5.6283, "step": 2203 }, { "epoch": 2.47, "learning_rate": 2.8309859154929576e-05, "loss": 5.7563, "step": 2204 }, { "epoch": 2.47, "learning_rate": 2.8308450704225352e-05, "loss": 5.7944, "step": 2205 }, { "epoch": 2.47, "learning_rate": 2.8307042253521125e-05, "loss": 5.6978, "step": 2206 }, { "epoch": 2.47, "learning_rate": 2.8305633802816905e-05, "loss": 5.7106, "step": 2207 }, { "epoch": 2.48, "learning_rate": 2.8304225352112678e-05, "loss": 5.8494, "step": 2208 }, { "epoch": 2.48, "learning_rate": 2.8302816901408454e-05, "loss": 5.7598, "step": 2209 }, { "epoch": 2.48, "learning_rate": 2.8301408450704227e-05, "loss": 5.6048, "step": 2210 }, { "epoch": 2.48, "learning_rate": 2.83e-05, "loss": 5.801, "step": 2211 }, { "epoch": 2.48, "learning_rate": 2.8298591549295776e-05, "loss": 5.6903, "step": 2212 }, { "epoch": 2.48, "learning_rate": 2.829718309859155e-05, "loss": 5.9447, "step": 2213 }, { "epoch": 2.48, "learning_rate": 2.8295774647887326e-05, "loss": 5.8336, "step": 2214 }, { "epoch": 2.48, "learning_rate": 2.82943661971831e-05, "loss": 5.8991, "step": 2215 }, { "epoch": 2.48, "learning_rate": 2.8292957746478875e-05, "loss": 5.6478, "step": 2216 }, { "epoch": 2.49, "learning_rate": 2.8291549295774648e-05, "loss": 5.6424, "step": 2217 }, { "epoch": 2.49, "learning_rate": 2.829014084507042e-05, "loss": 5.7126, "step": 2218 }, { "epoch": 2.49, "learning_rate": 2.8288732394366197e-05, "loss": 5.5857, "step": 2219 }, { "epoch": 2.49, "learning_rate": 2.828732394366197e-05, "loss": 5.846, "step": 2220 }, { "epoch": 2.49, "learning_rate": 2.8285915492957746e-05, "loss": 5.7761, "step": 2221 }, { "epoch": 2.49, "learning_rate": 2.8284507042253523e-05, "loss": 5.6428, "step": 2222 }, { "epoch": 2.49, "learning_rate": 2.82830985915493e-05, "loss": 6.0243, "step": 2223 }, { "epoch": 2.49, "learning_rate": 2.8281690140845072e-05, "loss": 5.6406, "step": 2224 }, { "epoch": 2.49, "learning_rate": 2.828028169014085e-05, "loss": 5.6799, "step": 2225 }, { "epoch": 2.5, "learning_rate": 2.827887323943662e-05, "loss": 5.7015, "step": 2226 }, { "epoch": 2.5, "learning_rate": 2.8277464788732394e-05, "loss": 5.6528, "step": 2227 }, { "epoch": 2.5, "learning_rate": 2.827605633802817e-05, "loss": 5.3903, "step": 2228 }, { "epoch": 2.5, "learning_rate": 2.8274647887323944e-05, "loss": 5.5735, "step": 2229 }, { "epoch": 2.5, "learning_rate": 2.827323943661972e-05, "loss": 5.5092, "step": 2230 }, { "epoch": 2.5, "learning_rate": 2.8271830985915493e-05, "loss": 5.5309, "step": 2231 }, { "epoch": 2.5, "learning_rate": 2.827042253521127e-05, "loss": 5.5968, "step": 2232 }, { "epoch": 2.5, "learning_rate": 2.8269014084507042e-05, "loss": 5.5656, "step": 2233 }, { "epoch": 2.5, "learning_rate": 2.8267605633802815e-05, "loss": 5.2354, "step": 2234 }, { "epoch": 2.51, "learning_rate": 2.826619718309859e-05, "loss": 6.2234, "step": 2235 }, { "epoch": 2.51, "learning_rate": 2.8264788732394364e-05, "loss": 6.0079, "step": 2236 }, { "epoch": 2.51, "learning_rate": 2.8263380281690144e-05, "loss": 6.2418, "step": 2237 }, { "epoch": 2.51, "learning_rate": 2.8261971830985917e-05, "loss": 6.2431, "step": 2238 }, { "epoch": 2.51, "learning_rate": 2.8260563380281693e-05, "loss": 5.899, "step": 2239 }, { "epoch": 2.51, "learning_rate": 2.8259154929577466e-05, "loss": 5.901, "step": 2240 }, { "epoch": 2.51, "learning_rate": 2.8257746478873243e-05, "loss": 5.8162, "step": 2241 }, { "epoch": 2.51, "learning_rate": 2.8256338028169016e-05, "loss": 5.8456, "step": 2242 }, { "epoch": 2.51, "learning_rate": 2.825492957746479e-05, "loss": 5.9214, "step": 2243 }, { "epoch": 2.52, "learning_rate": 2.8253521126760565e-05, "loss": 5.8194, "step": 2244 }, { "epoch": 2.52, "learning_rate": 2.8252112676056338e-05, "loss": 5.7194, "step": 2245 }, { "epoch": 2.52, "learning_rate": 2.8250704225352114e-05, "loss": 5.8196, "step": 2246 }, { "epoch": 2.52, "learning_rate": 2.8249295774647887e-05, "loss": 5.7652, "step": 2247 }, { "epoch": 2.52, "learning_rate": 2.8247887323943663e-05, "loss": 5.9587, "step": 2248 }, { "epoch": 2.52, "learning_rate": 2.8246478873239436e-05, "loss": 5.6588, "step": 2249 }, { "epoch": 2.52, "learning_rate": 2.824507042253521e-05, "loss": 5.5829, "step": 2250 }, { "epoch": 2.52, "learning_rate": 2.8243661971830986e-05, "loss": 5.6256, "step": 2251 }, { "epoch": 2.52, "learning_rate": 2.824225352112676e-05, "loss": 5.834, "step": 2252 }, { "epoch": 2.53, "learning_rate": 2.8240845070422538e-05, "loss": 5.7205, "step": 2253 }, { "epoch": 2.53, "learning_rate": 2.823943661971831e-05, "loss": 5.8234, "step": 2254 }, { "epoch": 2.53, "learning_rate": 2.8238028169014088e-05, "loss": 5.8395, "step": 2255 }, { "epoch": 2.53, "learning_rate": 2.823661971830986e-05, "loss": 5.7666, "step": 2256 }, { "epoch": 2.53, "learning_rate": 2.8235211267605633e-05, "loss": 5.6642, "step": 2257 }, { "epoch": 2.53, "learning_rate": 2.823380281690141e-05, "loss": 5.7016, "step": 2258 }, { "epoch": 2.53, "learning_rate": 2.8232394366197183e-05, "loss": 5.81, "step": 2259 }, { "epoch": 2.53, "learning_rate": 2.823098591549296e-05, "loss": 5.7926, "step": 2260 }, { "epoch": 2.53, "learning_rate": 2.8229577464788732e-05, "loss": 5.697, "step": 2261 }, { "epoch": 2.54, "learning_rate": 2.8228169014084508e-05, "loss": 5.7542, "step": 2262 }, { "epoch": 2.54, "learning_rate": 2.822676056338028e-05, "loss": 5.8097, "step": 2263 }, { "epoch": 2.54, "learning_rate": 2.8225352112676058e-05, "loss": 5.856, "step": 2264 }, { "epoch": 2.54, "learning_rate": 2.822394366197183e-05, "loss": 5.7212, "step": 2265 }, { "epoch": 2.54, "learning_rate": 2.8222535211267603e-05, "loss": 5.639, "step": 2266 }, { "epoch": 2.54, "learning_rate": 2.822112676056338e-05, "loss": 5.8659, "step": 2267 }, { "epoch": 2.54, "learning_rate": 2.8219718309859156e-05, "loss": 5.7632, "step": 2268 }, { "epoch": 2.54, "learning_rate": 2.8218309859154932e-05, "loss": 5.8361, "step": 2269 }, { "epoch": 2.54, "learning_rate": 2.8216901408450705e-05, "loss": 5.78, "step": 2270 }, { "epoch": 2.55, "learning_rate": 2.8215492957746482e-05, "loss": 5.716, "step": 2271 }, { "epoch": 2.55, "learning_rate": 2.8214084507042255e-05, "loss": 5.5012, "step": 2272 }, { "epoch": 2.55, "learning_rate": 2.8212676056338028e-05, "loss": 5.7095, "step": 2273 }, { "epoch": 2.55, "learning_rate": 2.8211267605633804e-05, "loss": 5.7195, "step": 2274 }, { "epoch": 2.55, "learning_rate": 2.8209859154929577e-05, "loss": 5.702, "step": 2275 }, { "epoch": 2.55, "learning_rate": 2.8208450704225353e-05, "loss": 5.626, "step": 2276 }, { "epoch": 2.55, "learning_rate": 2.8207042253521126e-05, "loss": 5.6698, "step": 2277 }, { "epoch": 2.55, "learning_rate": 2.8205633802816903e-05, "loss": 5.603, "step": 2278 }, { "epoch": 2.55, "learning_rate": 2.8204225352112675e-05, "loss": 5.8827, "step": 2279 }, { "epoch": 2.56, "learning_rate": 2.8202816901408452e-05, "loss": 5.6258, "step": 2280 }, { "epoch": 2.56, "learning_rate": 2.8201408450704225e-05, "loss": 5.5015, "step": 2281 }, { "epoch": 2.56, "learning_rate": 2.8199999999999998e-05, "loss": 5.5246, "step": 2282 }, { "epoch": 2.56, "learning_rate": 2.8198591549295777e-05, "loss": 5.2364, "step": 2283 }, { "epoch": 2.56, "learning_rate": 2.819718309859155e-05, "loss": 5.1964, "step": 2284 }, { "epoch": 2.56, "learning_rate": 2.8195774647887327e-05, "loss": 6.1527, "step": 2285 }, { "epoch": 2.56, "learning_rate": 2.81943661971831e-05, "loss": 5.8351, "step": 2286 }, { "epoch": 2.56, "learning_rate": 2.8192957746478876e-05, "loss": 5.9004, "step": 2287 }, { "epoch": 2.57, "learning_rate": 2.819154929577465e-05, "loss": 5.9822, "step": 2288 }, { "epoch": 2.57, "learning_rate": 2.8190140845070422e-05, "loss": 5.771, "step": 2289 }, { "epoch": 2.57, "learning_rate": 2.8188732394366198e-05, "loss": 5.9669, "step": 2290 }, { "epoch": 2.57, "learning_rate": 2.818732394366197e-05, "loss": 5.8817, "step": 2291 }, { "epoch": 2.57, "learning_rate": 2.8185915492957747e-05, "loss": 5.8039, "step": 2292 }, { "epoch": 2.57, "learning_rate": 2.818450704225352e-05, "loss": 5.7584, "step": 2293 }, { "epoch": 2.57, "learning_rate": 2.8183098591549297e-05, "loss": 5.8281, "step": 2294 }, { "epoch": 2.57, "learning_rate": 2.818169014084507e-05, "loss": 5.7773, "step": 2295 }, { "epoch": 2.57, "learning_rate": 2.8180281690140846e-05, "loss": 5.941, "step": 2296 }, { "epoch": 2.58, "learning_rate": 2.817887323943662e-05, "loss": 5.908, "step": 2297 }, { "epoch": 2.58, "learning_rate": 2.8177464788732395e-05, "loss": 5.8274, "step": 2298 }, { "epoch": 2.58, "learning_rate": 2.817605633802817e-05, "loss": 5.689, "step": 2299 }, { "epoch": 2.58, "learning_rate": 2.8174647887323945e-05, "loss": 5.717, "step": 2300 }, { "epoch": 2.58, "learning_rate": 2.817323943661972e-05, "loss": 5.7965, "step": 2301 }, { "epoch": 2.58, "learning_rate": 2.8171830985915494e-05, "loss": 5.6809, "step": 2302 }, { "epoch": 2.58, "learning_rate": 2.817042253521127e-05, "loss": 5.8359, "step": 2303 }, { "epoch": 2.58, "learning_rate": 2.8169014084507043e-05, "loss": 5.7818, "step": 2304 }, { "epoch": 2.58, "learning_rate": 2.8167605633802816e-05, "loss": 5.6071, "step": 2305 }, { "epoch": 2.59, "learning_rate": 2.8166197183098592e-05, "loss": 5.6471, "step": 2306 }, { "epoch": 2.59, "learning_rate": 2.8164788732394365e-05, "loss": 5.7602, "step": 2307 }, { "epoch": 2.59, "learning_rate": 2.816338028169014e-05, "loss": 5.8734, "step": 2308 }, { "epoch": 2.59, "learning_rate": 2.8161971830985915e-05, "loss": 5.8623, "step": 2309 }, { "epoch": 2.59, "learning_rate": 2.816056338028169e-05, "loss": 5.7955, "step": 2310 }, { "epoch": 2.59, "learning_rate": 2.8159154929577464e-05, "loss": 5.6221, "step": 2311 }, { "epoch": 2.59, "learning_rate": 2.815774647887324e-05, "loss": 5.7984, "step": 2312 }, { "epoch": 2.59, "learning_rate": 2.8156338028169017e-05, "loss": 5.7789, "step": 2313 }, { "epoch": 2.59, "learning_rate": 2.815492957746479e-05, "loss": 5.7498, "step": 2314 }, { "epoch": 2.6, "learning_rate": 2.8153521126760566e-05, "loss": 5.5605, "step": 2315 }, { "epoch": 2.6, "learning_rate": 2.815211267605634e-05, "loss": 5.5426, "step": 2316 }, { "epoch": 2.6, "learning_rate": 2.8150704225352115e-05, "loss": 5.6703, "step": 2317 }, { "epoch": 2.6, "learning_rate": 2.8149295774647888e-05, "loss": 5.7036, "step": 2318 }, { "epoch": 2.6, "learning_rate": 2.8147887323943664e-05, "loss": 5.7514, "step": 2319 }, { "epoch": 2.6, "learning_rate": 2.8146478873239437e-05, "loss": 5.8487, "step": 2320 }, { "epoch": 2.6, "learning_rate": 2.814507042253521e-05, "loss": 5.8343, "step": 2321 }, { "epoch": 2.6, "learning_rate": 2.8143661971830987e-05, "loss": 5.7731, "step": 2322 }, { "epoch": 2.6, "learning_rate": 2.814225352112676e-05, "loss": 5.6575, "step": 2323 }, { "epoch": 2.61, "learning_rate": 2.8140845070422536e-05, "loss": 5.6566, "step": 2324 }, { "epoch": 2.61, "learning_rate": 2.813943661971831e-05, "loss": 5.7784, "step": 2325 }, { "epoch": 2.61, "learning_rate": 2.8138028169014085e-05, "loss": 5.7277, "step": 2326 }, { "epoch": 2.61, "learning_rate": 2.8136619718309858e-05, "loss": 5.6061, "step": 2327 }, { "epoch": 2.61, "learning_rate": 2.8135211267605638e-05, "loss": 5.5008, "step": 2328 }, { "epoch": 2.61, "learning_rate": 2.813380281690141e-05, "loss": 5.6817, "step": 2329 }, { "epoch": 2.61, "learning_rate": 2.8132394366197184e-05, "loss": 5.3199, "step": 2330 }, { "epoch": 2.61, "learning_rate": 2.813098591549296e-05, "loss": 5.3417, "step": 2331 }, { "epoch": 2.61, "learning_rate": 2.8129577464788733e-05, "loss": 5.3562, "step": 2332 }, { "epoch": 2.62, "learning_rate": 2.812816901408451e-05, "loss": 5.5677, "step": 2333 }, { "epoch": 2.62, "learning_rate": 2.8126760563380282e-05, "loss": 5.0041, "step": 2334 }, { "epoch": 2.62, "learning_rate": 2.812535211267606e-05, "loss": 6.0551, "step": 2335 }, { "epoch": 2.62, "learning_rate": 2.812394366197183e-05, "loss": 5.9496, "step": 2336 }, { "epoch": 2.62, "learning_rate": 2.8122535211267604e-05, "loss": 5.944, "step": 2337 }, { "epoch": 2.62, "learning_rate": 2.812112676056338e-05, "loss": 5.8294, "step": 2338 }, { "epoch": 2.62, "learning_rate": 2.8119718309859154e-05, "loss": 5.7788, "step": 2339 }, { "epoch": 2.62, "learning_rate": 2.811830985915493e-05, "loss": 5.9267, "step": 2340 }, { "epoch": 2.62, "learning_rate": 2.8116901408450703e-05, "loss": 5.8697, "step": 2341 }, { "epoch": 2.63, "learning_rate": 2.811549295774648e-05, "loss": 5.7988, "step": 2342 }, { "epoch": 2.63, "learning_rate": 2.8114084507042256e-05, "loss": 5.7477, "step": 2343 }, { "epoch": 2.63, "learning_rate": 2.8112676056338032e-05, "loss": 5.757, "step": 2344 }, { "epoch": 2.63, "learning_rate": 2.8111267605633805e-05, "loss": 5.6647, "step": 2345 }, { "epoch": 2.63, "learning_rate": 2.8109859154929578e-05, "loss": 5.7407, "step": 2346 }, { "epoch": 2.63, "learning_rate": 2.8108450704225354e-05, "loss": 5.7853, "step": 2347 }, { "epoch": 2.63, "learning_rate": 2.8107042253521127e-05, "loss": 5.7882, "step": 2348 }, { "epoch": 2.63, "learning_rate": 2.8105633802816903e-05, "loss": 5.6109, "step": 2349 }, { "epoch": 2.63, "learning_rate": 2.8104225352112676e-05, "loss": 5.8261, "step": 2350 }, { "epoch": 2.64, "learning_rate": 2.8102816901408453e-05, "loss": 5.6776, "step": 2351 }, { "epoch": 2.64, "learning_rate": 2.8101408450704226e-05, "loss": 5.6085, "step": 2352 }, { "epoch": 2.64, "learning_rate": 2.81e-05, "loss": 5.7469, "step": 2353 }, { "epoch": 2.64, "learning_rate": 2.8098591549295775e-05, "loss": 5.6612, "step": 2354 }, { "epoch": 2.64, "learning_rate": 2.8097183098591548e-05, "loss": 5.6548, "step": 2355 }, { "epoch": 2.64, "learning_rate": 2.8095774647887324e-05, "loss": 5.7668, "step": 2356 }, { "epoch": 2.64, "learning_rate": 2.8094366197183097e-05, "loss": 5.7211, "step": 2357 }, { "epoch": 2.64, "learning_rate": 2.8092957746478877e-05, "loss": 5.6407, "step": 2358 }, { "epoch": 2.64, "learning_rate": 2.809154929577465e-05, "loss": 5.8428, "step": 2359 }, { "epoch": 2.65, "learning_rate": 2.8090140845070423e-05, "loss": 5.6122, "step": 2360 }, { "epoch": 2.65, "learning_rate": 2.80887323943662e-05, "loss": 5.7323, "step": 2361 }, { "epoch": 2.65, "learning_rate": 2.8087323943661972e-05, "loss": 5.7675, "step": 2362 }, { "epoch": 2.65, "learning_rate": 2.808591549295775e-05, "loss": 5.7105, "step": 2363 }, { "epoch": 2.65, "learning_rate": 2.808450704225352e-05, "loss": 5.7395, "step": 2364 }, { "epoch": 2.65, "learning_rate": 2.8083098591549298e-05, "loss": 5.6453, "step": 2365 }, { "epoch": 2.65, "learning_rate": 2.808169014084507e-05, "loss": 5.7347, "step": 2366 }, { "epoch": 2.65, "learning_rate": 2.8080281690140847e-05, "loss": 5.7945, "step": 2367 }, { "epoch": 2.65, "learning_rate": 2.807887323943662e-05, "loss": 5.5934, "step": 2368 }, { "epoch": 2.66, "learning_rate": 2.8077464788732393e-05, "loss": 5.7464, "step": 2369 }, { "epoch": 2.66, "learning_rate": 2.807605633802817e-05, "loss": 5.7336, "step": 2370 }, { "epoch": 2.66, "learning_rate": 2.8074647887323942e-05, "loss": 5.7658, "step": 2371 }, { "epoch": 2.66, "learning_rate": 2.807323943661972e-05, "loss": 5.8412, "step": 2372 }, { "epoch": 2.66, "learning_rate": 2.807183098591549e-05, "loss": 5.6434, "step": 2373 }, { "epoch": 2.66, "learning_rate": 2.807042253521127e-05, "loss": 5.6429, "step": 2374 }, { "epoch": 2.66, "learning_rate": 2.8069014084507044e-05, "loss": 5.8194, "step": 2375 }, { "epoch": 2.66, "learning_rate": 2.8067605633802817e-05, "loss": 5.7194, "step": 2376 }, { "epoch": 2.66, "learning_rate": 2.8066197183098593e-05, "loss": 5.6101, "step": 2377 }, { "epoch": 2.67, "learning_rate": 2.8064788732394366e-05, "loss": 5.5595, "step": 2378 }, { "epoch": 2.67, "learning_rate": 2.8063380281690143e-05, "loss": 5.4239, "step": 2379 }, { "epoch": 2.67, "learning_rate": 2.8061971830985916e-05, "loss": 5.5543, "step": 2380 }, { "epoch": 2.67, "learning_rate": 2.8060563380281692e-05, "loss": 5.4819, "step": 2381 }, { "epoch": 2.67, "learning_rate": 2.8059154929577465e-05, "loss": 5.4334, "step": 2382 }, { "epoch": 2.67, "learning_rate": 2.805774647887324e-05, "loss": 5.5228, "step": 2383 }, { "epoch": 2.67, "learning_rate": 2.8056338028169014e-05, "loss": 4.8885, "step": 2384 }, { "epoch": 2.67, "learning_rate": 2.8054929577464787e-05, "loss": 6.0503, "step": 2385 }, { "epoch": 2.67, "learning_rate": 2.8053521126760563e-05, "loss": 5.9355, "step": 2386 }, { "epoch": 2.68, "learning_rate": 2.8052112676056336e-05, "loss": 5.8382, "step": 2387 }, { "epoch": 2.68, "learning_rate": 2.8050704225352113e-05, "loss": 5.833, "step": 2388 }, { "epoch": 2.68, "learning_rate": 2.804929577464789e-05, "loss": 5.7725, "step": 2389 }, { "epoch": 2.68, "learning_rate": 2.8047887323943665e-05, "loss": 5.8524, "step": 2390 }, { "epoch": 2.68, "learning_rate": 2.8046478873239438e-05, "loss": 5.7466, "step": 2391 }, { "epoch": 2.68, "learning_rate": 2.804507042253521e-05, "loss": 5.8391, "step": 2392 }, { "epoch": 2.68, "learning_rate": 2.8043661971830988e-05, "loss": 5.8834, "step": 2393 }, { "epoch": 2.68, "learning_rate": 2.804225352112676e-05, "loss": 5.5149, "step": 2394 }, { "epoch": 2.68, "learning_rate": 2.8040845070422537e-05, "loss": 5.5413, "step": 2395 }, { "epoch": 2.69, "learning_rate": 2.803943661971831e-05, "loss": 5.7998, "step": 2396 }, { "epoch": 2.69, "learning_rate": 2.8038028169014086e-05, "loss": 5.5432, "step": 2397 }, { "epoch": 2.69, "learning_rate": 2.803661971830986e-05, "loss": 5.5504, "step": 2398 }, { "epoch": 2.69, "learning_rate": 2.8035211267605635e-05, "loss": 5.6585, "step": 2399 }, { "epoch": 2.69, "learning_rate": 2.8033802816901408e-05, "loss": 5.55, "step": 2400 }, { "epoch": 2.69, "learning_rate": 2.803239436619718e-05, "loss": 5.7817, "step": 2401 }, { "epoch": 2.69, "learning_rate": 2.8030985915492958e-05, "loss": 5.6482, "step": 2402 }, { "epoch": 2.69, "learning_rate": 2.802957746478873e-05, "loss": 5.7672, "step": 2403 }, { "epoch": 2.7, "learning_rate": 2.802816901408451e-05, "loss": 5.6727, "step": 2404 }, { "epoch": 2.7, "learning_rate": 2.8026760563380283e-05, "loss": 5.7006, "step": 2405 }, { "epoch": 2.7, "learning_rate": 2.802535211267606e-05, "loss": 5.902, "step": 2406 }, { "epoch": 2.7, "learning_rate": 2.8023943661971832e-05, "loss": 5.8426, "step": 2407 }, { "epoch": 2.7, "learning_rate": 2.8022535211267605e-05, "loss": 5.7642, "step": 2408 }, { "epoch": 2.7, "learning_rate": 2.8021126760563382e-05, "loss": 5.5587, "step": 2409 }, { "epoch": 2.7, "learning_rate": 2.8019718309859155e-05, "loss": 5.5818, "step": 2410 }, { "epoch": 2.7, "learning_rate": 2.801830985915493e-05, "loss": 5.7981, "step": 2411 }, { "epoch": 2.7, "learning_rate": 2.8016901408450704e-05, "loss": 5.6255, "step": 2412 }, { "epoch": 2.71, "learning_rate": 2.801549295774648e-05, "loss": 5.6772, "step": 2413 }, { "epoch": 2.71, "learning_rate": 2.8014084507042253e-05, "loss": 5.5972, "step": 2414 }, { "epoch": 2.71, "learning_rate": 2.801267605633803e-05, "loss": 5.6789, "step": 2415 }, { "epoch": 2.71, "learning_rate": 2.8011267605633802e-05, "loss": 5.8302, "step": 2416 }, { "epoch": 2.71, "learning_rate": 2.8009859154929575e-05, "loss": 5.5367, "step": 2417 }, { "epoch": 2.71, "learning_rate": 2.8008450704225352e-05, "loss": 5.6729, "step": 2418 }, { "epoch": 2.71, "learning_rate": 2.8007042253521128e-05, "loss": 5.4426, "step": 2419 }, { "epoch": 2.71, "learning_rate": 2.8005633802816904e-05, "loss": 5.6593, "step": 2420 }, { "epoch": 2.71, "learning_rate": 2.8004225352112677e-05, "loss": 5.7091, "step": 2421 }, { "epoch": 2.72, "learning_rate": 2.8002816901408454e-05, "loss": 5.7341, "step": 2422 }, { "epoch": 2.72, "learning_rate": 2.8001408450704227e-05, "loss": 5.476, "step": 2423 }, { "epoch": 2.72, "learning_rate": 2.8e-05, "loss": 5.5516, "step": 2424 }, { "epoch": 2.72, "learning_rate": 2.7998591549295776e-05, "loss": 5.669, "step": 2425 }, { "epoch": 2.72, "learning_rate": 2.799718309859155e-05, "loss": 5.5051, "step": 2426 }, { "epoch": 2.72, "learning_rate": 2.7995774647887325e-05, "loss": 5.5403, "step": 2427 }, { "epoch": 2.72, "learning_rate": 2.7994366197183098e-05, "loss": 5.5317, "step": 2428 }, { "epoch": 2.72, "learning_rate": 2.7992957746478874e-05, "loss": 5.6651, "step": 2429 }, { "epoch": 2.72, "learning_rate": 2.7991549295774647e-05, "loss": 5.3641, "step": 2430 }, { "epoch": 2.73, "learning_rate": 2.7990140845070424e-05, "loss": 5.54, "step": 2431 }, { "epoch": 2.73, "learning_rate": 2.7988732394366197e-05, "loss": 5.36, "step": 2432 }, { "epoch": 2.73, "learning_rate": 2.798732394366197e-05, "loss": 5.2864, "step": 2433 }, { "epoch": 2.73, "learning_rate": 2.798591549295775e-05, "loss": 4.9371, "step": 2434 }, { "epoch": 2.73, "learning_rate": 2.7984507042253522e-05, "loss": 6.0766, "step": 2435 }, { "epoch": 2.73, "learning_rate": 2.79830985915493e-05, "loss": 5.8073, "step": 2436 }, { "epoch": 2.73, "learning_rate": 2.798169014084507e-05, "loss": 5.9915, "step": 2437 }, { "epoch": 2.73, "learning_rate": 2.7980281690140848e-05, "loss": 5.9455, "step": 2438 }, { "epoch": 2.73, "learning_rate": 2.797887323943662e-05, "loss": 5.8183, "step": 2439 }, { "epoch": 2.74, "learning_rate": 2.7977464788732394e-05, "loss": 5.9879, "step": 2440 }, { "epoch": 2.74, "learning_rate": 2.797605633802817e-05, "loss": 5.9121, "step": 2441 }, { "epoch": 2.74, "learning_rate": 2.7974647887323943e-05, "loss": 5.7478, "step": 2442 }, { "epoch": 2.74, "learning_rate": 2.797323943661972e-05, "loss": 5.7507, "step": 2443 }, { "epoch": 2.74, "learning_rate": 2.7971830985915492e-05, "loss": 5.7551, "step": 2444 }, { "epoch": 2.74, "learning_rate": 2.797042253521127e-05, "loss": 5.5979, "step": 2445 }, { "epoch": 2.74, "learning_rate": 2.796901408450704e-05, "loss": 5.628, "step": 2446 }, { "epoch": 2.74, "learning_rate": 2.7967605633802818e-05, "loss": 5.6582, "step": 2447 }, { "epoch": 2.74, "learning_rate": 2.796619718309859e-05, "loss": 5.8853, "step": 2448 }, { "epoch": 2.75, "learning_rate": 2.7964788732394367e-05, "loss": 5.7402, "step": 2449 }, { "epoch": 2.75, "learning_rate": 2.7963380281690144e-05, "loss": 5.8396, "step": 2450 }, { "epoch": 2.75, "learning_rate": 2.7961971830985916e-05, "loss": 5.83, "step": 2451 }, { "epoch": 2.75, "learning_rate": 2.7960563380281693e-05, "loss": 5.7517, "step": 2452 }, { "epoch": 2.75, "learning_rate": 2.7959154929577466e-05, "loss": 5.5882, "step": 2453 }, { "epoch": 2.75, "learning_rate": 2.7957746478873242e-05, "loss": 5.8737, "step": 2454 }, { "epoch": 2.75, "learning_rate": 2.7956338028169015e-05, "loss": 5.598, "step": 2455 }, { "epoch": 2.75, "learning_rate": 2.7954929577464788e-05, "loss": 5.8423, "step": 2456 }, { "epoch": 2.75, "learning_rate": 2.7953521126760564e-05, "loss": 5.6835, "step": 2457 }, { "epoch": 2.76, "learning_rate": 2.7952112676056337e-05, "loss": 5.6505, "step": 2458 }, { "epoch": 2.76, "learning_rate": 2.7950704225352114e-05, "loss": 5.6356, "step": 2459 }, { "epoch": 2.76, "learning_rate": 2.7949295774647887e-05, "loss": 5.6206, "step": 2460 }, { "epoch": 2.76, "learning_rate": 2.7947887323943663e-05, "loss": 5.6917, "step": 2461 }, { "epoch": 2.76, "learning_rate": 2.7946478873239436e-05, "loss": 5.6362, "step": 2462 }, { "epoch": 2.76, "learning_rate": 2.7945070422535212e-05, "loss": 5.7117, "step": 2463 }, { "epoch": 2.76, "learning_rate": 2.794366197183099e-05, "loss": 5.6936, "step": 2464 }, { "epoch": 2.76, "learning_rate": 2.794225352112676e-05, "loss": 5.7734, "step": 2465 }, { "epoch": 2.76, "learning_rate": 2.7940845070422538e-05, "loss": 5.575, "step": 2466 }, { "epoch": 2.77, "learning_rate": 2.793943661971831e-05, "loss": 5.6555, "step": 2467 }, { "epoch": 2.77, "learning_rate": 2.7938028169014087e-05, "loss": 5.7263, "step": 2468 }, { "epoch": 2.77, "learning_rate": 2.793661971830986e-05, "loss": 5.6385, "step": 2469 }, { "epoch": 2.77, "learning_rate": 2.7935211267605636e-05, "loss": 5.579, "step": 2470 }, { "epoch": 2.77, "learning_rate": 2.793380281690141e-05, "loss": 5.6018, "step": 2471 }, { "epoch": 2.77, "learning_rate": 2.7932394366197182e-05, "loss": 5.6562, "step": 2472 }, { "epoch": 2.77, "learning_rate": 2.793098591549296e-05, "loss": 5.5973, "step": 2473 }, { "epoch": 2.77, "learning_rate": 2.792957746478873e-05, "loss": 5.5781, "step": 2474 }, { "epoch": 2.77, "learning_rate": 2.7928169014084508e-05, "loss": 5.6546, "step": 2475 }, { "epoch": 2.78, "learning_rate": 2.792676056338028e-05, "loss": 5.7238, "step": 2476 }, { "epoch": 2.78, "learning_rate": 2.7925352112676057e-05, "loss": 5.4016, "step": 2477 }, { "epoch": 2.78, "learning_rate": 2.792394366197183e-05, "loss": 5.6336, "step": 2478 }, { "epoch": 2.78, "learning_rate": 2.7922535211267606e-05, "loss": 5.421, "step": 2479 }, { "epoch": 2.78, "learning_rate": 2.7921126760563383e-05, "loss": 5.405, "step": 2480 }, { "epoch": 2.78, "learning_rate": 2.7919718309859156e-05, "loss": 5.1707, "step": 2481 }, { "epoch": 2.78, "learning_rate": 2.7918309859154932e-05, "loss": 5.282, "step": 2482 }, { "epoch": 2.78, "learning_rate": 2.7916901408450705e-05, "loss": 5.2574, "step": 2483 }, { "epoch": 2.78, "learning_rate": 2.791549295774648e-05, "loss": 5.1088, "step": 2484 }, { "epoch": 2.79, "learning_rate": 2.7914084507042254e-05, "loss": 5.8417, "step": 2485 }, { "epoch": 2.79, "learning_rate": 2.791267605633803e-05, "loss": 5.8681, "step": 2486 }, { "epoch": 2.79, "learning_rate": 2.7911267605633803e-05, "loss": 5.8169, "step": 2487 }, { "epoch": 2.79, "learning_rate": 2.7909859154929576e-05, "loss": 5.9287, "step": 2488 }, { "epoch": 2.79, "learning_rate": 2.7908450704225353e-05, "loss": 5.9287, "step": 2489 }, { "epoch": 2.79, "learning_rate": 2.7907042253521126e-05, "loss": 5.8013, "step": 2490 }, { "epoch": 2.79, "learning_rate": 2.7905633802816902e-05, "loss": 5.7922, "step": 2491 }, { "epoch": 2.79, "learning_rate": 2.7904225352112675e-05, "loss": 5.8515, "step": 2492 }, { "epoch": 2.79, "learning_rate": 2.790281690140845e-05, "loss": 5.6763, "step": 2493 }, { "epoch": 2.8, "learning_rate": 2.7901408450704224e-05, "loss": 5.8258, "step": 2494 }, { "epoch": 2.8, "learning_rate": 2.79e-05, "loss": 5.6405, "step": 2495 }, { "epoch": 2.8, "learning_rate": 2.7898591549295777e-05, "loss": 5.71, "step": 2496 }, { "epoch": 2.8, "learning_rate": 2.789718309859155e-05, "loss": 5.6849, "step": 2497 }, { "epoch": 2.8, "learning_rate": 2.7895774647887326e-05, "loss": 5.4637, "step": 2498 }, { "epoch": 2.8, "learning_rate": 2.78943661971831e-05, "loss": 5.5308, "step": 2499 }, { "epoch": 2.8, "learning_rate": 2.7892957746478875e-05, "loss": 5.5638, "step": 2500 }, { "epoch": 2.8, "learning_rate": 2.789154929577465e-05, "loss": 5.7663, "step": 2501 }, { "epoch": 2.8, "learning_rate": 2.7890140845070425e-05, "loss": 5.457, "step": 2502 }, { "epoch": 2.81, "learning_rate": 2.7888732394366198e-05, "loss": 5.4773, "step": 2503 }, { "epoch": 2.81, "learning_rate": 2.788732394366197e-05, "loss": 5.6711, "step": 2504 }, { "epoch": 2.81, "learning_rate": 2.7885915492957747e-05, "loss": 5.7387, "step": 2505 }, { "epoch": 2.81, "learning_rate": 2.788450704225352e-05, "loss": 5.6963, "step": 2506 }, { "epoch": 2.81, "learning_rate": 2.7883098591549296e-05, "loss": 5.6568, "step": 2507 }, { "epoch": 2.81, "learning_rate": 2.788169014084507e-05, "loss": 5.603, "step": 2508 }, { "epoch": 2.81, "learning_rate": 2.7880281690140845e-05, "loss": 5.7038, "step": 2509 }, { "epoch": 2.81, "learning_rate": 2.7878873239436622e-05, "loss": 5.6849, "step": 2510 }, { "epoch": 2.82, "learning_rate": 2.7877464788732395e-05, "loss": 5.8184, "step": 2511 }, { "epoch": 2.82, "learning_rate": 2.787605633802817e-05, "loss": 5.5234, "step": 2512 }, { "epoch": 2.82, "learning_rate": 2.7874647887323944e-05, "loss": 5.7395, "step": 2513 }, { "epoch": 2.82, "learning_rate": 2.787323943661972e-05, "loss": 5.6433, "step": 2514 }, { "epoch": 2.82, "learning_rate": 2.7871830985915493e-05, "loss": 5.7222, "step": 2515 }, { "epoch": 2.82, "learning_rate": 2.787042253521127e-05, "loss": 5.5362, "step": 2516 }, { "epoch": 2.82, "learning_rate": 2.7869014084507043e-05, "loss": 5.6451, "step": 2517 }, { "epoch": 2.82, "learning_rate": 2.786760563380282e-05, "loss": 5.656, "step": 2518 }, { "epoch": 2.82, "learning_rate": 2.7866197183098592e-05, "loss": 5.5741, "step": 2519 }, { "epoch": 2.83, "learning_rate": 2.7864788732394365e-05, "loss": 5.6033, "step": 2520 }, { "epoch": 2.83, "learning_rate": 2.786338028169014e-05, "loss": 5.6367, "step": 2521 }, { "epoch": 2.83, "learning_rate": 2.7861971830985914e-05, "loss": 5.6778, "step": 2522 }, { "epoch": 2.83, "learning_rate": 2.786056338028169e-05, "loss": 5.7316, "step": 2523 }, { "epoch": 2.83, "learning_rate": 2.7859154929577463e-05, "loss": 5.4934, "step": 2524 }, { "epoch": 2.83, "learning_rate": 2.7857746478873243e-05, "loss": 5.5772, "step": 2525 }, { "epoch": 2.83, "learning_rate": 2.7856338028169016e-05, "loss": 5.6694, "step": 2526 }, { "epoch": 2.83, "learning_rate": 2.785492957746479e-05, "loss": 5.7479, "step": 2527 }, { "epoch": 2.83, "learning_rate": 2.7853521126760565e-05, "loss": 5.6437, "step": 2528 }, { "epoch": 2.84, "learning_rate": 2.7852112676056338e-05, "loss": 5.4111, "step": 2529 }, { "epoch": 2.84, "learning_rate": 2.7850704225352115e-05, "loss": 5.4943, "step": 2530 }, { "epoch": 2.84, "learning_rate": 2.7849295774647887e-05, "loss": 5.3034, "step": 2531 }, { "epoch": 2.84, "learning_rate": 2.7847887323943664e-05, "loss": 5.4324, "step": 2532 }, { "epoch": 2.84, "learning_rate": 2.7846478873239437e-05, "loss": 5.2858, "step": 2533 }, { "epoch": 2.84, "learning_rate": 2.7845070422535213e-05, "loss": 4.9984, "step": 2534 }, { "epoch": 2.84, "learning_rate": 2.7843661971830986e-05, "loss": 5.9877, "step": 2535 }, { "epoch": 2.84, "learning_rate": 2.784225352112676e-05, "loss": 5.8959, "step": 2536 }, { "epoch": 2.84, "learning_rate": 2.7840845070422535e-05, "loss": 5.763, "step": 2537 }, { "epoch": 2.85, "learning_rate": 2.7839436619718308e-05, "loss": 5.9719, "step": 2538 }, { "epoch": 2.85, "learning_rate": 2.7838028169014085e-05, "loss": 5.7494, "step": 2539 }, { "epoch": 2.85, "learning_rate": 2.783661971830986e-05, "loss": 5.7277, "step": 2540 }, { "epoch": 2.85, "learning_rate": 2.7835211267605637e-05, "loss": 5.8642, "step": 2541 }, { "epoch": 2.85, "learning_rate": 2.783380281690141e-05, "loss": 5.849, "step": 2542 }, { "epoch": 2.85, "learning_rate": 2.7832394366197183e-05, "loss": 5.7068, "step": 2543 }, { "epoch": 2.85, "learning_rate": 2.783098591549296e-05, "loss": 5.7638, "step": 2544 }, { "epoch": 2.85, "learning_rate": 2.7829577464788732e-05, "loss": 5.6719, "step": 2545 }, { "epoch": 2.85, "learning_rate": 2.782816901408451e-05, "loss": 5.7483, "step": 2546 }, { "epoch": 2.86, "learning_rate": 2.782676056338028e-05, "loss": 5.7158, "step": 2547 }, { "epoch": 2.86, "learning_rate": 2.7825352112676058e-05, "loss": 5.4833, "step": 2548 }, { "epoch": 2.86, "learning_rate": 2.782394366197183e-05, "loss": 5.694, "step": 2549 }, { "epoch": 2.86, "learning_rate": 2.7822535211267607e-05, "loss": 5.5874, "step": 2550 }, { "epoch": 2.86, "learning_rate": 2.782112676056338e-05, "loss": 5.6684, "step": 2551 }, { "epoch": 2.86, "learning_rate": 2.7819718309859153e-05, "loss": 5.6867, "step": 2552 }, { "epoch": 2.86, "learning_rate": 2.781830985915493e-05, "loss": 5.5873, "step": 2553 }, { "epoch": 2.86, "learning_rate": 2.7816901408450702e-05, "loss": 5.6552, "step": 2554 }, { "epoch": 2.86, "learning_rate": 2.7815492957746482e-05, "loss": 5.404, "step": 2555 }, { "epoch": 2.87, "learning_rate": 2.7814084507042255e-05, "loss": 5.5642, "step": 2556 }, { "epoch": 2.87, "learning_rate": 2.781267605633803e-05, "loss": 5.6958, "step": 2557 }, { "epoch": 2.87, "learning_rate": 2.7811267605633804e-05, "loss": 5.5959, "step": 2558 }, { "epoch": 2.87, "learning_rate": 2.7809859154929577e-05, "loss": 5.7749, "step": 2559 }, { "epoch": 2.87, "learning_rate": 2.7808450704225354e-05, "loss": 5.4652, "step": 2560 }, { "epoch": 2.87, "learning_rate": 2.7807042253521127e-05, "loss": 5.4928, "step": 2561 }, { "epoch": 2.87, "learning_rate": 2.7805633802816903e-05, "loss": 5.4733, "step": 2562 }, { "epoch": 2.87, "learning_rate": 2.7804225352112676e-05, "loss": 5.5133, "step": 2563 }, { "epoch": 2.87, "learning_rate": 2.7802816901408452e-05, "loss": 5.6387, "step": 2564 }, { "epoch": 2.88, "learning_rate": 2.7801408450704225e-05, "loss": 5.4843, "step": 2565 }, { "epoch": 2.88, "learning_rate": 2.78e-05, "loss": 5.5344, "step": 2566 }, { "epoch": 2.88, "learning_rate": 2.7798591549295774e-05, "loss": 5.4469, "step": 2567 }, { "epoch": 2.88, "learning_rate": 2.7797183098591547e-05, "loss": 5.5419, "step": 2568 }, { "epoch": 2.88, "learning_rate": 2.7795774647887324e-05, "loss": 5.7793, "step": 2569 }, { "epoch": 2.88, "learning_rate": 2.77943661971831e-05, "loss": 5.4858, "step": 2570 }, { "epoch": 2.88, "learning_rate": 2.7792957746478876e-05, "loss": 5.6724, "step": 2571 }, { "epoch": 2.88, "learning_rate": 2.779154929577465e-05, "loss": 5.5382, "step": 2572 }, { "epoch": 2.88, "learning_rate": 2.7790140845070426e-05, "loss": 5.5104, "step": 2573 }, { "epoch": 2.89, "learning_rate": 2.77887323943662e-05, "loss": 5.6066, "step": 2574 }, { "epoch": 2.89, "learning_rate": 2.778732394366197e-05, "loss": 5.6297, "step": 2575 }, { "epoch": 2.89, "learning_rate": 2.7785915492957748e-05, "loss": 5.5579, "step": 2576 }, { "epoch": 2.89, "learning_rate": 2.778450704225352e-05, "loss": 5.5108, "step": 2577 }, { "epoch": 2.89, "learning_rate": 2.7783098591549297e-05, "loss": 5.5665, "step": 2578 }, { "epoch": 2.89, "learning_rate": 2.778169014084507e-05, "loss": 5.3076, "step": 2579 }, { "epoch": 2.89, "learning_rate": 2.7780281690140846e-05, "loss": 5.4312, "step": 2580 }, { "epoch": 2.89, "learning_rate": 2.777887323943662e-05, "loss": 5.3215, "step": 2581 }, { "epoch": 2.89, "learning_rate": 2.7777464788732396e-05, "loss": 5.3403, "step": 2582 }, { "epoch": 2.9, "learning_rate": 2.777605633802817e-05, "loss": 5.0783, "step": 2583 }, { "epoch": 2.9, "learning_rate": 2.777464788732394e-05, "loss": 4.8549, "step": 2584 }, { "epoch": 2.9, "learning_rate": 2.777323943661972e-05, "loss": 6.0113, "step": 2585 }, { "epoch": 2.9, "learning_rate": 2.7771830985915494e-05, "loss": 5.7576, "step": 2586 }, { "epoch": 2.9, "learning_rate": 2.777042253521127e-05, "loss": 5.7873, "step": 2587 }, { "epoch": 2.9, "learning_rate": 2.7769014084507044e-05, "loss": 5.851, "step": 2588 }, { "epoch": 2.9, "learning_rate": 2.776760563380282e-05, "loss": 5.803, "step": 2589 }, { "epoch": 2.9, "learning_rate": 2.7766197183098593e-05, "loss": 5.908, "step": 2590 }, { "epoch": 2.9, "learning_rate": 2.7764788732394366e-05, "loss": 5.7334, "step": 2591 }, { "epoch": 2.91, "learning_rate": 2.7763380281690142e-05, "loss": 5.7843, "step": 2592 }, { "epoch": 2.91, "learning_rate": 2.7761971830985915e-05, "loss": 5.6909, "step": 2593 }, { "epoch": 2.91, "learning_rate": 2.776056338028169e-05, "loss": 5.5913, "step": 2594 }, { "epoch": 2.91, "learning_rate": 2.7759154929577464e-05, "loss": 5.6534, "step": 2595 }, { "epoch": 2.91, "learning_rate": 2.775774647887324e-05, "loss": 5.5699, "step": 2596 }, { "epoch": 2.91, "learning_rate": 2.7756338028169014e-05, "loss": 5.6838, "step": 2597 }, { "epoch": 2.91, "learning_rate": 2.775492957746479e-05, "loss": 5.6114, "step": 2598 }, { "epoch": 2.91, "learning_rate": 2.7753521126760563e-05, "loss": 5.5513, "step": 2599 }, { "epoch": 2.91, "learning_rate": 2.775211267605634e-05, "loss": 5.6324, "step": 2600 }, { "epoch": 2.92, "learning_rate": 2.7750704225352116e-05, "loss": 5.6424, "step": 2601 }, { "epoch": 2.92, "learning_rate": 2.774929577464789e-05, "loss": 5.6626, "step": 2602 }, { "epoch": 2.92, "learning_rate": 2.7747887323943665e-05, "loss": 5.6062, "step": 2603 }, { "epoch": 2.92, "learning_rate": 2.7746478873239438e-05, "loss": 5.7219, "step": 2604 }, { "epoch": 2.92, "learning_rate": 2.7745070422535214e-05, "loss": 5.3733, "step": 2605 }, { "epoch": 2.92, "learning_rate": 2.7743661971830987e-05, "loss": 5.5156, "step": 2606 }, { "epoch": 2.92, "learning_rate": 2.774225352112676e-05, "loss": 5.6222, "step": 2607 }, { "epoch": 2.92, "learning_rate": 2.7740845070422536e-05, "loss": 5.3706, "step": 2608 }, { "epoch": 2.92, "learning_rate": 2.773943661971831e-05, "loss": 5.5939, "step": 2609 }, { "epoch": 2.93, "learning_rate": 2.7738028169014086e-05, "loss": 5.6031, "step": 2610 }, { "epoch": 2.93, "learning_rate": 2.773661971830986e-05, "loss": 5.5898, "step": 2611 }, { "epoch": 2.93, "learning_rate": 2.7735211267605635e-05, "loss": 5.4355, "step": 2612 }, { "epoch": 2.93, "learning_rate": 2.7733802816901408e-05, "loss": 5.7506, "step": 2613 }, { "epoch": 2.93, "learning_rate": 2.773239436619718e-05, "loss": 5.4992, "step": 2614 }, { "epoch": 2.93, "learning_rate": 2.7730985915492957e-05, "loss": 5.6942, "step": 2615 }, { "epoch": 2.93, "learning_rate": 2.7729577464788733e-05, "loss": 5.7021, "step": 2616 }, { "epoch": 2.93, "learning_rate": 2.772816901408451e-05, "loss": 5.5195, "step": 2617 }, { "epoch": 2.93, "learning_rate": 2.7726760563380283e-05, "loss": 5.4501, "step": 2618 }, { "epoch": 2.94, "learning_rate": 2.772535211267606e-05, "loss": 5.5725, "step": 2619 }, { "epoch": 2.94, "learning_rate": 2.7723943661971832e-05, "loss": 5.4995, "step": 2620 }, { "epoch": 2.94, "learning_rate": 2.7722535211267608e-05, "loss": 5.3801, "step": 2621 }, { "epoch": 2.94, "learning_rate": 2.772112676056338e-05, "loss": 5.4785, "step": 2622 }, { "epoch": 2.94, "learning_rate": 2.7719718309859154e-05, "loss": 5.5644, "step": 2623 }, { "epoch": 2.94, "learning_rate": 2.771830985915493e-05, "loss": 5.6688, "step": 2624 }, { "epoch": 2.94, "learning_rate": 2.7716901408450703e-05, "loss": 5.4444, "step": 2625 }, { "epoch": 2.94, "learning_rate": 2.771549295774648e-05, "loss": 5.3823, "step": 2626 }, { "epoch": 2.95, "learning_rate": 2.7714084507042253e-05, "loss": 5.1955, "step": 2627 }, { "epoch": 2.95, "learning_rate": 2.771267605633803e-05, "loss": 5.1879, "step": 2628 }, { "epoch": 2.95, "learning_rate": 2.7711267605633802e-05, "loss": 5.5572, "step": 2629 }, { "epoch": 2.95, "learning_rate": 2.7709859154929575e-05, "loss": 5.5976, "step": 2630 }, { "epoch": 2.95, "learning_rate": 2.7708450704225355e-05, "loss": 5.0735, "step": 2631 }, { "epoch": 2.95, "learning_rate": 2.7707042253521128e-05, "loss": 5.0322, "step": 2632 }, { "epoch": 2.95, "learning_rate": 2.7705633802816904e-05, "loss": 5.0061, "step": 2633 }, { "epoch": 2.95, "learning_rate": 2.7704225352112677e-05, "loss": 4.8947, "step": 2634 }, { "epoch": 2.95, "learning_rate": 2.7702816901408453e-05, "loss": 5.9352, "step": 2635 }, { "epoch": 2.96, "learning_rate": 2.7701408450704226e-05, "loss": 5.8543, "step": 2636 }, { "epoch": 2.96, "learning_rate": 2.7700000000000002e-05, "loss": 5.9398, "step": 2637 }, { "epoch": 2.96, "learning_rate": 2.7698591549295775e-05, "loss": 5.625, "step": 2638 }, { "epoch": 2.96, "learning_rate": 2.769718309859155e-05, "loss": 5.7823, "step": 2639 }, { "epoch": 2.96, "learning_rate": 2.7695774647887325e-05, "loss": 5.5466, "step": 2640 }, { "epoch": 2.96, "learning_rate": 2.7694366197183098e-05, "loss": 5.6185, "step": 2641 }, { "epoch": 2.96, "learning_rate": 2.7692957746478874e-05, "loss": 5.7074, "step": 2642 }, { "epoch": 2.96, "learning_rate": 2.7691549295774647e-05, "loss": 5.7114, "step": 2643 }, { "epoch": 2.96, "learning_rate": 2.7690140845070423e-05, "loss": 5.3789, "step": 2644 }, { "epoch": 2.97, "learning_rate": 2.7688732394366196e-05, "loss": 5.4808, "step": 2645 }, { "epoch": 2.97, "learning_rate": 2.7687323943661972e-05, "loss": 5.6949, "step": 2646 }, { "epoch": 2.97, "learning_rate": 2.768591549295775e-05, "loss": 5.6868, "step": 2647 }, { "epoch": 2.97, "learning_rate": 2.7684507042253522e-05, "loss": 5.6716, "step": 2648 }, { "epoch": 2.97, "learning_rate": 2.7683098591549298e-05, "loss": 5.6738, "step": 2649 }, { "epoch": 2.97, "learning_rate": 2.768169014084507e-05, "loss": 5.4585, "step": 2650 }, { "epoch": 2.97, "learning_rate": 2.7680281690140847e-05, "loss": 5.672, "step": 2651 }, { "epoch": 2.97, "learning_rate": 2.767887323943662e-05, "loss": 5.5443, "step": 2652 }, { "epoch": 2.97, "learning_rate": 2.7677464788732397e-05, "loss": 5.6285, "step": 2653 }, { "epoch": 2.98, "learning_rate": 2.767605633802817e-05, "loss": 5.46, "step": 2654 }, { "epoch": 2.98, "learning_rate": 2.7674647887323943e-05, "loss": 5.5451, "step": 2655 }, { "epoch": 2.98, "learning_rate": 2.767323943661972e-05, "loss": 5.5886, "step": 2656 }, { "epoch": 2.98, "learning_rate": 2.7671830985915492e-05, "loss": 5.608, "step": 2657 }, { "epoch": 2.98, "learning_rate": 2.7670422535211268e-05, "loss": 5.7211, "step": 2658 }, { "epoch": 2.98, "learning_rate": 2.766901408450704e-05, "loss": 5.5374, "step": 2659 }, { "epoch": 2.98, "learning_rate": 2.7667605633802817e-05, "loss": 5.477, "step": 2660 }, { "epoch": 2.98, "learning_rate": 2.7666197183098594e-05, "loss": 5.6128, "step": 2661 }, { "epoch": 2.98, "learning_rate": 2.7664788732394367e-05, "loss": 5.7499, "step": 2662 }, { "epoch": 2.99, "learning_rate": 2.7663380281690143e-05, "loss": 5.4866, "step": 2663 }, { "epoch": 2.99, "learning_rate": 2.7661971830985916e-05, "loss": 5.72, "step": 2664 }, { "epoch": 2.99, "learning_rate": 2.7660563380281692e-05, "loss": 5.1649, "step": 2665 }, { "epoch": 2.99, "learning_rate": 2.7659154929577465e-05, "loss": 5.3756, "step": 2666 }, { "epoch": 2.99, "learning_rate": 2.765774647887324e-05, "loss": 5.5378, "step": 2667 }, { "epoch": 2.99, "learning_rate": 2.7656338028169015e-05, "loss": 5.5316, "step": 2668 }, { "epoch": 2.99, "learning_rate": 2.765492957746479e-05, "loss": 5.4148, "step": 2669 }, { "epoch": 2.99, "learning_rate": 2.7653521126760564e-05, "loss": 5.4294, "step": 2670 }, { "epoch": 2.99, "learning_rate": 2.7652112676056337e-05, "loss": 5.4528, "step": 2671 }, { "epoch": 3.0, "learning_rate": 2.7650704225352113e-05, "loss": 5.3467, "step": 2672 }, { "epoch": 3.0, "learning_rate": 2.7649295774647886e-05, "loss": 5.2414, "step": 2673 }, { "epoch": 3.0, "learning_rate": 2.7647887323943662e-05, "loss": 5.0751, "step": 2674 }, { "epoch": 3.0, "learning_rate": 2.7646478873239435e-05, "loss": 5.1968, "step": 2675 }, { "epoch": 3.0, "learning_rate": 2.7645070422535215e-05, "loss": 4.792, "step": 2676 }, { "epoch": 3.0, "learning_rate": 2.7643661971830988e-05, "loss": 5.7013, "step": 2677 }, { "epoch": 3.0, "learning_rate": 2.764225352112676e-05, "loss": 5.6952, "step": 2678 }, { "epoch": 3.0, "learning_rate": 2.7640845070422537e-05, "loss": 5.7713, "step": 2679 }, { "epoch": 3.0, "learning_rate": 2.763943661971831e-05, "loss": 5.6176, "step": 2680 }, { "epoch": 3.01, "learning_rate": 2.7638028169014087e-05, "loss": 5.8076, "step": 2681 }, { "epoch": 3.01, "learning_rate": 2.763661971830986e-05, "loss": 5.5453, "step": 2682 }, { "epoch": 3.01, "learning_rate": 2.7635211267605636e-05, "loss": 5.4657, "step": 2683 }, { "epoch": 3.01, "learning_rate": 2.763380281690141e-05, "loss": 5.6471, "step": 2684 }, { "epoch": 3.01, "learning_rate": 2.7632394366197185e-05, "loss": 5.5475, "step": 2685 }, { "epoch": 3.01, "learning_rate": 2.7630985915492958e-05, "loss": 5.4929, "step": 2686 }, { "epoch": 3.01, "learning_rate": 2.762957746478873e-05, "loss": 5.6232, "step": 2687 }, { "epoch": 3.01, "learning_rate": 2.7628169014084507e-05, "loss": 5.4537, "step": 2688 }, { "epoch": 3.01, "learning_rate": 2.762676056338028e-05, "loss": 5.5382, "step": 2689 }, { "epoch": 3.02, "learning_rate": 2.7625352112676057e-05, "loss": 5.5003, "step": 2690 }, { "epoch": 3.02, "learning_rate": 2.7623943661971833e-05, "loss": 5.5356, "step": 2691 }, { "epoch": 3.02, "learning_rate": 2.762253521126761e-05, "loss": 5.6451, "step": 2692 }, { "epoch": 3.02, "learning_rate": 2.7621126760563382e-05, "loss": 5.5723, "step": 2693 }, { "epoch": 3.02, "learning_rate": 2.7619718309859155e-05, "loss": 5.4755, "step": 2694 }, { "epoch": 3.02, "learning_rate": 2.761830985915493e-05, "loss": 5.5362, "step": 2695 }, { "epoch": 3.02, "learning_rate": 2.7616901408450704e-05, "loss": 5.343, "step": 2696 }, { "epoch": 3.02, "learning_rate": 2.761549295774648e-05, "loss": 5.4554, "step": 2697 }, { "epoch": 3.02, "learning_rate": 2.7614084507042254e-05, "loss": 5.3489, "step": 2698 }, { "epoch": 3.03, "learning_rate": 2.761267605633803e-05, "loss": 5.5748, "step": 2699 }, { "epoch": 3.03, "learning_rate": 2.7611267605633803e-05, "loss": 5.5954, "step": 2700 }, { "epoch": 3.03, "learning_rate": 2.760985915492958e-05, "loss": 5.5094, "step": 2701 }, { "epoch": 3.03, "learning_rate": 2.7608450704225352e-05, "loss": 5.4326, "step": 2702 }, { "epoch": 3.03, "learning_rate": 2.7607042253521125e-05, "loss": 5.703, "step": 2703 }, { "epoch": 3.03, "learning_rate": 2.76056338028169e-05, "loss": 5.2898, "step": 2704 }, { "epoch": 3.03, "learning_rate": 2.7604225352112674e-05, "loss": 5.5208, "step": 2705 }, { "epoch": 3.03, "learning_rate": 2.7602816901408454e-05, "loss": 5.5139, "step": 2706 }, { "epoch": 3.03, "learning_rate": 2.7601408450704227e-05, "loss": 5.3379, "step": 2707 }, { "epoch": 3.04, "learning_rate": 2.7600000000000003e-05, "loss": 5.2163, "step": 2708 }, { "epoch": 3.04, "learning_rate": 2.7598591549295776e-05, "loss": 5.2784, "step": 2709 }, { "epoch": 3.04, "learning_rate": 2.759718309859155e-05, "loss": 5.4137, "step": 2710 }, { "epoch": 3.04, "learning_rate": 2.7595774647887326e-05, "loss": 5.3316, "step": 2711 }, { "epoch": 3.04, "learning_rate": 2.75943661971831e-05, "loss": 5.5045, "step": 2712 }, { "epoch": 3.04, "learning_rate": 2.7592957746478875e-05, "loss": 5.4127, "step": 2713 }, { "epoch": 3.04, "learning_rate": 2.7591549295774648e-05, "loss": 5.4343, "step": 2714 }, { "epoch": 3.04, "learning_rate": 2.7590140845070424e-05, "loss": 5.3557, "step": 2715 }, { "epoch": 3.04, "learning_rate": 2.7588732394366197e-05, "loss": 5.3358, "step": 2716 }, { "epoch": 3.05, "learning_rate": 2.758732394366197e-05, "loss": 5.3474, "step": 2717 }, { "epoch": 3.05, "learning_rate": 2.7585915492957746e-05, "loss": 5.3478, "step": 2718 }, { "epoch": 3.05, "learning_rate": 2.758450704225352e-05, "loss": 5.3237, "step": 2719 }, { "epoch": 3.05, "learning_rate": 2.7583098591549296e-05, "loss": 5.4654, "step": 2720 }, { "epoch": 3.05, "learning_rate": 2.758169014084507e-05, "loss": 5.093, "step": 2721 }, { "epoch": 3.05, "learning_rate": 2.758028169014085e-05, "loss": 5.1985, "step": 2722 }, { "epoch": 3.05, "learning_rate": 2.757887323943662e-05, "loss": 5.1496, "step": 2723 }, { "epoch": 3.05, "learning_rate": 2.7577464788732398e-05, "loss": 5.0891, "step": 2724 }, { "epoch": 3.05, "learning_rate": 2.757605633802817e-05, "loss": 5.2755, "step": 2725 }, { "epoch": 3.06, "learning_rate": 2.7574647887323943e-05, "loss": 4.6618, "step": 2726 }, { "epoch": 3.06, "learning_rate": 2.757323943661972e-05, "loss": 5.7287, "step": 2727 }, { "epoch": 3.06, "learning_rate": 2.7571830985915493e-05, "loss": 5.7024, "step": 2728 }, { "epoch": 3.06, "learning_rate": 2.757042253521127e-05, "loss": 5.5971, "step": 2729 }, { "epoch": 3.06, "learning_rate": 2.7569014084507042e-05, "loss": 5.4607, "step": 2730 }, { "epoch": 3.06, "learning_rate": 2.756760563380282e-05, "loss": 5.6877, "step": 2731 }, { "epoch": 3.06, "learning_rate": 2.756619718309859e-05, "loss": 5.6774, "step": 2732 }, { "epoch": 3.06, "learning_rate": 2.7564788732394364e-05, "loss": 5.7797, "step": 2733 }, { "epoch": 3.07, "learning_rate": 2.756338028169014e-05, "loss": 5.4228, "step": 2734 }, { "epoch": 3.07, "learning_rate": 2.7561971830985914e-05, "loss": 5.4179, "step": 2735 }, { "epoch": 3.07, "learning_rate": 2.756056338028169e-05, "loss": 5.5022, "step": 2736 }, { "epoch": 3.07, "learning_rate": 2.7559154929577466e-05, "loss": 5.4024, "step": 2737 }, { "epoch": 3.07, "learning_rate": 2.7557746478873243e-05, "loss": 5.4886, "step": 2738 }, { "epoch": 3.07, "learning_rate": 2.7556338028169015e-05, "loss": 5.5164, "step": 2739 }, { "epoch": 3.07, "learning_rate": 2.7554929577464792e-05, "loss": 5.3964, "step": 2740 }, { "epoch": 3.07, "learning_rate": 2.7553521126760565e-05, "loss": 5.6581, "step": 2741 }, { "epoch": 3.07, "learning_rate": 2.7552112676056338e-05, "loss": 5.5589, "step": 2742 }, { "epoch": 3.08, "learning_rate": 2.7550704225352114e-05, "loss": 5.4021, "step": 2743 }, { "epoch": 3.08, "learning_rate": 2.7549295774647887e-05, "loss": 5.4646, "step": 2744 }, { "epoch": 3.08, "learning_rate": 2.7547887323943663e-05, "loss": 5.4412, "step": 2745 }, { "epoch": 3.08, "learning_rate": 2.7546478873239436e-05, "loss": 5.511, "step": 2746 }, { "epoch": 3.08, "learning_rate": 2.7545070422535213e-05, "loss": 5.3518, "step": 2747 }, { "epoch": 3.08, "learning_rate": 2.7543661971830986e-05, "loss": 5.4101, "step": 2748 }, { "epoch": 3.08, "learning_rate": 2.754225352112676e-05, "loss": 5.5455, "step": 2749 }, { "epoch": 3.08, "learning_rate": 2.7540845070422535e-05, "loss": 5.597, "step": 2750 }, { "epoch": 3.08, "learning_rate": 2.7539436619718308e-05, "loss": 5.4826, "step": 2751 }, { "epoch": 3.09, "learning_rate": 2.7538028169014087e-05, "loss": 5.4764, "step": 2752 }, { "epoch": 3.09, "learning_rate": 2.753661971830986e-05, "loss": 5.5434, "step": 2753 }, { "epoch": 3.09, "learning_rate": 2.7535211267605637e-05, "loss": 5.3409, "step": 2754 }, { "epoch": 3.09, "learning_rate": 2.753380281690141e-05, "loss": 5.5827, "step": 2755 }, { "epoch": 3.09, "learning_rate": 2.7532394366197186e-05, "loss": 5.4738, "step": 2756 }, { "epoch": 3.09, "learning_rate": 2.753098591549296e-05, "loss": 5.4158, "step": 2757 }, { "epoch": 3.09, "learning_rate": 2.7529577464788732e-05, "loss": 5.4007, "step": 2758 }, { "epoch": 3.09, "learning_rate": 2.7528169014084508e-05, "loss": 5.4755, "step": 2759 }, { "epoch": 3.09, "learning_rate": 2.752676056338028e-05, "loss": 5.342, "step": 2760 }, { "epoch": 3.1, "learning_rate": 2.7525352112676058e-05, "loss": 5.3881, "step": 2761 }, { "epoch": 3.1, "learning_rate": 2.752394366197183e-05, "loss": 5.2653, "step": 2762 }, { "epoch": 3.1, "learning_rate": 2.7522535211267607e-05, "loss": 5.4605, "step": 2763 }, { "epoch": 3.1, "learning_rate": 2.752112676056338e-05, "loss": 5.3996, "step": 2764 }, { "epoch": 3.1, "learning_rate": 2.7519718309859153e-05, "loss": 5.3856, "step": 2765 }, { "epoch": 3.1, "learning_rate": 2.751830985915493e-05, "loss": 5.3884, "step": 2766 }, { "epoch": 3.1, "learning_rate": 2.7516901408450705e-05, "loss": 5.429, "step": 2767 }, { "epoch": 3.1, "learning_rate": 2.751549295774648e-05, "loss": 5.4097, "step": 2768 }, { "epoch": 3.1, "learning_rate": 2.7514084507042255e-05, "loss": 5.1265, "step": 2769 }, { "epoch": 3.11, "learning_rate": 2.751267605633803e-05, "loss": 5.371, "step": 2770 }, { "epoch": 3.11, "learning_rate": 2.7511267605633804e-05, "loss": 5.0224, "step": 2771 }, { "epoch": 3.11, "learning_rate": 2.750985915492958e-05, "loss": 5.194, "step": 2772 }, { "epoch": 3.11, "learning_rate": 2.7508450704225353e-05, "loss": 5.1228, "step": 2773 }, { "epoch": 3.11, "learning_rate": 2.7507042253521126e-05, "loss": 4.9252, "step": 2774 }, { "epoch": 3.11, "learning_rate": 2.7505633802816902e-05, "loss": 5.1188, "step": 2775 }, { "epoch": 3.11, "learning_rate": 2.7504225352112675e-05, "loss": 4.8875, "step": 2776 }, { "epoch": 3.11, "learning_rate": 2.7502816901408452e-05, "loss": 5.8303, "step": 2777 }, { "epoch": 3.11, "learning_rate": 2.7501408450704225e-05, "loss": 5.6425, "step": 2778 }, { "epoch": 3.12, "learning_rate": 2.75e-05, "loss": 5.4299, "step": 2779 }, { "epoch": 3.12, "learning_rate": 2.7498591549295774e-05, "loss": 5.646, "step": 2780 }, { "epoch": 3.12, "learning_rate": 2.7497183098591547e-05, "loss": 5.5982, "step": 2781 }, { "epoch": 3.12, "learning_rate": 2.7495774647887327e-05, "loss": 5.6583, "step": 2782 }, { "epoch": 3.12, "learning_rate": 2.74943661971831e-05, "loss": 5.6388, "step": 2783 }, { "epoch": 3.12, "learning_rate": 2.7492957746478876e-05, "loss": 5.4251, "step": 2784 }, { "epoch": 3.12, "learning_rate": 2.749154929577465e-05, "loss": 5.5509, "step": 2785 }, { "epoch": 3.12, "learning_rate": 2.7490140845070425e-05, "loss": 5.48, "step": 2786 }, { "epoch": 3.12, "learning_rate": 2.7488732394366198e-05, "loss": 5.4533, "step": 2787 }, { "epoch": 3.13, "learning_rate": 2.7487323943661974e-05, "loss": 5.6917, "step": 2788 }, { "epoch": 3.13, "learning_rate": 2.7485915492957747e-05, "loss": 5.5326, "step": 2789 }, { "epoch": 3.13, "learning_rate": 2.748450704225352e-05, "loss": 5.499, "step": 2790 }, { "epoch": 3.13, "learning_rate": 2.7483098591549297e-05, "loss": 5.5817, "step": 2791 }, { "epoch": 3.13, "learning_rate": 2.748169014084507e-05, "loss": 5.3945, "step": 2792 }, { "epoch": 3.13, "learning_rate": 2.7480281690140846e-05, "loss": 5.4868, "step": 2793 }, { "epoch": 3.13, "learning_rate": 2.747887323943662e-05, "loss": 5.3528, "step": 2794 }, { "epoch": 3.13, "learning_rate": 2.7477464788732395e-05, "loss": 5.4097, "step": 2795 }, { "epoch": 3.13, "learning_rate": 2.7476056338028168e-05, "loss": 5.3707, "step": 2796 }, { "epoch": 3.14, "learning_rate": 2.7474647887323944e-05, "loss": 5.5157, "step": 2797 }, { "epoch": 3.14, "learning_rate": 2.747323943661972e-05, "loss": 5.5047, "step": 2798 }, { "epoch": 3.14, "learning_rate": 2.7471830985915494e-05, "loss": 5.3661, "step": 2799 }, { "epoch": 3.14, "learning_rate": 2.747042253521127e-05, "loss": 5.4985, "step": 2800 }, { "epoch": 3.14, "learning_rate": 2.7469014084507043e-05, "loss": 5.3448, "step": 2801 }, { "epoch": 3.14, "learning_rate": 2.746760563380282e-05, "loss": 5.3967, "step": 2802 }, { "epoch": 3.14, "learning_rate": 2.7466197183098592e-05, "loss": 5.4215, "step": 2803 }, { "epoch": 3.14, "learning_rate": 2.746478873239437e-05, "loss": 5.4258, "step": 2804 }, { "epoch": 3.14, "learning_rate": 2.746338028169014e-05, "loss": 5.2885, "step": 2805 }, { "epoch": 3.15, "learning_rate": 2.7461971830985914e-05, "loss": 5.3478, "step": 2806 }, { "epoch": 3.15, "learning_rate": 2.746056338028169e-05, "loss": 5.2254, "step": 2807 }, { "epoch": 3.15, "learning_rate": 2.7459154929577464e-05, "loss": 5.2296, "step": 2808 }, { "epoch": 3.15, "learning_rate": 2.745774647887324e-05, "loss": 5.3741, "step": 2809 }, { "epoch": 3.15, "learning_rate": 2.7456338028169013e-05, "loss": 5.3875, "step": 2810 }, { "epoch": 3.15, "learning_rate": 2.745492957746479e-05, "loss": 5.3158, "step": 2811 }, { "epoch": 3.15, "learning_rate": 2.7453521126760566e-05, "loss": 5.3014, "step": 2812 }, { "epoch": 3.15, "learning_rate": 2.745211267605634e-05, "loss": 5.1913, "step": 2813 }, { "epoch": 3.15, "learning_rate": 2.7450704225352115e-05, "loss": 5.4527, "step": 2814 }, { "epoch": 3.16, "learning_rate": 2.7449295774647888e-05, "loss": 5.2578, "step": 2815 }, { "epoch": 3.16, "learning_rate": 2.7447887323943664e-05, "loss": 5.3978, "step": 2816 }, { "epoch": 3.16, "learning_rate": 2.7446478873239437e-05, "loss": 5.3023, "step": 2817 }, { "epoch": 3.16, "learning_rate": 2.7445070422535214e-05, "loss": 5.2486, "step": 2818 }, { "epoch": 3.16, "learning_rate": 2.7443661971830986e-05, "loss": 5.3082, "step": 2819 }, { "epoch": 3.16, "learning_rate": 2.744225352112676e-05, "loss": 5.2013, "step": 2820 }, { "epoch": 3.16, "learning_rate": 2.7440845070422536e-05, "loss": 5.2998, "step": 2821 }, { "epoch": 3.16, "learning_rate": 2.743943661971831e-05, "loss": 5.1176, "step": 2822 }, { "epoch": 3.16, "learning_rate": 2.7438028169014085e-05, "loss": 5.0891, "step": 2823 }, { "epoch": 3.17, "learning_rate": 2.7436619718309858e-05, "loss": 5.0252, "step": 2824 }, { "epoch": 3.17, "learning_rate": 2.7435211267605634e-05, "loss": 4.836, "step": 2825 }, { "epoch": 3.17, "learning_rate": 2.7433802816901407e-05, "loss": 4.6971, "step": 2826 }, { "epoch": 3.17, "learning_rate": 2.7432394366197187e-05, "loss": 5.881, "step": 2827 }, { "epoch": 3.17, "learning_rate": 2.743098591549296e-05, "loss": 5.595, "step": 2828 }, { "epoch": 3.17, "learning_rate": 2.7429577464788733e-05, "loss": 5.6698, "step": 2829 }, { "epoch": 3.17, "learning_rate": 2.742816901408451e-05, "loss": 5.752, "step": 2830 }, { "epoch": 3.17, "learning_rate": 2.7426760563380282e-05, "loss": 5.656, "step": 2831 }, { "epoch": 3.17, "learning_rate": 2.742535211267606e-05, "loss": 5.7363, "step": 2832 }, { "epoch": 3.18, "learning_rate": 2.742394366197183e-05, "loss": 5.4768, "step": 2833 }, { "epoch": 3.18, "learning_rate": 2.7422535211267608e-05, "loss": 5.7295, "step": 2834 }, { "epoch": 3.18, "learning_rate": 2.742112676056338e-05, "loss": 5.4822, "step": 2835 }, { "epoch": 3.18, "learning_rate": 2.7419718309859154e-05, "loss": 5.4004, "step": 2836 }, { "epoch": 3.18, "learning_rate": 2.741830985915493e-05, "loss": 5.5026, "step": 2837 }, { "epoch": 3.18, "learning_rate": 2.7416901408450703e-05, "loss": 5.4272, "step": 2838 }, { "epoch": 3.18, "learning_rate": 2.741549295774648e-05, "loss": 5.6285, "step": 2839 }, { "epoch": 3.18, "learning_rate": 2.7414084507042252e-05, "loss": 5.6141, "step": 2840 }, { "epoch": 3.18, "learning_rate": 2.741267605633803e-05, "loss": 5.4623, "step": 2841 }, { "epoch": 3.19, "learning_rate": 2.74112676056338e-05, "loss": 5.2836, "step": 2842 }, { "epoch": 3.19, "learning_rate": 2.740985915492958e-05, "loss": 5.4177, "step": 2843 }, { "epoch": 3.19, "learning_rate": 2.7408450704225354e-05, "loss": 5.4073, "step": 2844 }, { "epoch": 3.19, "learning_rate": 2.7407042253521127e-05, "loss": 5.4872, "step": 2845 }, { "epoch": 3.19, "learning_rate": 2.7405633802816903e-05, "loss": 5.4659, "step": 2846 }, { "epoch": 3.19, "learning_rate": 2.7404225352112676e-05, "loss": 5.4848, "step": 2847 }, { "epoch": 3.19, "learning_rate": 2.7402816901408453e-05, "loss": 5.5281, "step": 2848 }, { "epoch": 3.19, "learning_rate": 2.7401408450704226e-05, "loss": 5.464, "step": 2849 }, { "epoch": 3.2, "learning_rate": 2.7400000000000002e-05, "loss": 5.3048, "step": 2850 }, { "epoch": 3.2, "learning_rate": 2.7398591549295775e-05, "loss": 5.3645, "step": 2851 }, { "epoch": 3.2, "learning_rate": 2.7397183098591548e-05, "loss": 5.3738, "step": 2852 }, { "epoch": 3.2, "learning_rate": 2.7395774647887324e-05, "loss": 5.4127, "step": 2853 }, { "epoch": 3.2, "learning_rate": 2.7394366197183097e-05, "loss": 5.3761, "step": 2854 }, { "epoch": 3.2, "learning_rate": 2.7392957746478873e-05, "loss": 5.3829, "step": 2855 }, { "epoch": 3.2, "learning_rate": 2.7391549295774646e-05, "loss": 5.4175, "step": 2856 }, { "epoch": 3.2, "learning_rate": 2.7390140845070423e-05, "loss": 5.5984, "step": 2857 }, { "epoch": 3.2, "learning_rate": 2.73887323943662e-05, "loss": 5.2833, "step": 2858 }, { "epoch": 3.21, "learning_rate": 2.7387323943661975e-05, "loss": 5.4304, "step": 2859 }, { "epoch": 3.21, "learning_rate": 2.738591549295775e-05, "loss": 5.3256, "step": 2860 }, { "epoch": 3.21, "learning_rate": 2.738450704225352e-05, "loss": 5.2476, "step": 2861 }, { "epoch": 3.21, "learning_rate": 2.7383098591549298e-05, "loss": 5.4746, "step": 2862 }, { "epoch": 3.21, "learning_rate": 2.738169014084507e-05, "loss": 5.1882, "step": 2863 }, { "epoch": 3.21, "learning_rate": 2.7380281690140847e-05, "loss": 5.3123, "step": 2864 }, { "epoch": 3.21, "learning_rate": 2.737887323943662e-05, "loss": 5.202, "step": 2865 }, { "epoch": 3.21, "learning_rate": 2.7377464788732396e-05, "loss": 5.2325, "step": 2866 }, { "epoch": 3.21, "learning_rate": 2.737605633802817e-05, "loss": 5.3083, "step": 2867 }, { "epoch": 3.22, "learning_rate": 2.7374647887323942e-05, "loss": 5.4266, "step": 2868 }, { "epoch": 3.22, "learning_rate": 2.737323943661972e-05, "loss": 5.1618, "step": 2869 }, { "epoch": 3.22, "learning_rate": 2.737183098591549e-05, "loss": 5.0335, "step": 2870 }, { "epoch": 3.22, "learning_rate": 2.7370422535211268e-05, "loss": 5.0405, "step": 2871 }, { "epoch": 3.22, "learning_rate": 2.736901408450704e-05, "loss": 4.9187, "step": 2872 }, { "epoch": 3.22, "learning_rate": 2.736760563380282e-05, "loss": 4.9263, "step": 2873 }, { "epoch": 3.22, "learning_rate": 2.7366197183098593e-05, "loss": 4.9038, "step": 2874 }, { "epoch": 3.22, "learning_rate": 2.736478873239437e-05, "loss": 4.7398, "step": 2875 }, { "epoch": 3.22, "learning_rate": 2.7363380281690143e-05, "loss": 4.5768, "step": 2876 }, { "epoch": 3.23, "learning_rate": 2.7361971830985915e-05, "loss": 5.6687, "step": 2877 }, { "epoch": 3.23, "learning_rate": 2.7360563380281692e-05, "loss": 5.6694, "step": 2878 }, { "epoch": 3.23, "learning_rate": 2.7359154929577465e-05, "loss": 5.7942, "step": 2879 }, { "epoch": 3.23, "learning_rate": 2.735774647887324e-05, "loss": 5.65, "step": 2880 }, { "epoch": 3.23, "learning_rate": 2.7356338028169014e-05, "loss": 5.4595, "step": 2881 }, { "epoch": 3.23, "learning_rate": 2.735492957746479e-05, "loss": 5.478, "step": 2882 }, { "epoch": 3.23, "learning_rate": 2.7353521126760563e-05, "loss": 5.5174, "step": 2883 }, { "epoch": 3.23, "learning_rate": 2.7352112676056336e-05, "loss": 5.6533, "step": 2884 }, { "epoch": 3.23, "learning_rate": 2.7350704225352113e-05, "loss": 5.5154, "step": 2885 }, { "epoch": 3.24, "learning_rate": 2.7349295774647885e-05, "loss": 5.3442, "step": 2886 }, { "epoch": 3.24, "learning_rate": 2.7347887323943662e-05, "loss": 5.2943, "step": 2887 }, { "epoch": 3.24, "learning_rate": 2.7346478873239438e-05, "loss": 5.6235, "step": 2888 }, { "epoch": 3.24, "learning_rate": 2.7345070422535214e-05, "loss": 5.6287, "step": 2889 }, { "epoch": 3.24, "learning_rate": 2.7343661971830987e-05, "loss": 5.3187, "step": 2890 }, { "epoch": 3.24, "learning_rate": 2.7342253521126764e-05, "loss": 5.4847, "step": 2891 }, { "epoch": 3.24, "learning_rate": 2.7340845070422537e-05, "loss": 5.5352, "step": 2892 }, { "epoch": 3.24, "learning_rate": 2.733943661971831e-05, "loss": 5.4769, "step": 2893 }, { "epoch": 3.24, "learning_rate": 2.7338028169014086e-05, "loss": 5.4112, "step": 2894 }, { "epoch": 3.25, "learning_rate": 2.733661971830986e-05, "loss": 5.4573, "step": 2895 }, { "epoch": 3.25, "learning_rate": 2.7335211267605635e-05, "loss": 5.3242, "step": 2896 }, { "epoch": 3.25, "learning_rate": 2.7333802816901408e-05, "loss": 5.4701, "step": 2897 }, { "epoch": 3.25, "learning_rate": 2.7332394366197185e-05, "loss": 5.4336, "step": 2898 }, { "epoch": 3.25, "learning_rate": 2.7330985915492957e-05, "loss": 5.3325, "step": 2899 }, { "epoch": 3.25, "learning_rate": 2.732957746478873e-05, "loss": 5.3905, "step": 2900 }, { "epoch": 3.25, "learning_rate": 2.7328169014084507e-05, "loss": 5.262, "step": 2901 }, { "epoch": 3.25, "learning_rate": 2.732676056338028e-05, "loss": 5.3514, "step": 2902 }, { "epoch": 3.25, "learning_rate": 2.732535211267606e-05, "loss": 5.4273, "step": 2903 }, { "epoch": 3.26, "learning_rate": 2.7323943661971832e-05, "loss": 5.2823, "step": 2904 }, { "epoch": 3.26, "learning_rate": 2.732253521126761e-05, "loss": 5.4945, "step": 2905 }, { "epoch": 3.26, "learning_rate": 2.732112676056338e-05, "loss": 5.533, "step": 2906 }, { "epoch": 3.26, "learning_rate": 2.7319718309859158e-05, "loss": 5.3622, "step": 2907 }, { "epoch": 3.26, "learning_rate": 2.731830985915493e-05, "loss": 5.4556, "step": 2908 }, { "epoch": 3.26, "learning_rate": 2.7316901408450704e-05, "loss": 5.4676, "step": 2909 }, { "epoch": 3.26, "learning_rate": 2.731549295774648e-05, "loss": 5.462, "step": 2910 }, { "epoch": 3.26, "learning_rate": 2.7314084507042253e-05, "loss": 5.3701, "step": 2911 }, { "epoch": 3.26, "learning_rate": 2.731267605633803e-05, "loss": 5.2563, "step": 2912 }, { "epoch": 3.27, "learning_rate": 2.7311267605633802e-05, "loss": 5.3266, "step": 2913 }, { "epoch": 3.27, "learning_rate": 2.730985915492958e-05, "loss": 5.2503, "step": 2914 }, { "epoch": 3.27, "learning_rate": 2.730845070422535e-05, "loss": 5.232, "step": 2915 }, { "epoch": 3.27, "learning_rate": 2.7307042253521125e-05, "loss": 5.2801, "step": 2916 }, { "epoch": 3.27, "learning_rate": 2.73056338028169e-05, "loss": 5.3687, "step": 2917 }, { "epoch": 3.27, "learning_rate": 2.7304225352112677e-05, "loss": 5.3384, "step": 2918 }, { "epoch": 3.27, "learning_rate": 2.7302816901408454e-05, "loss": 5.3763, "step": 2919 }, { "epoch": 3.27, "learning_rate": 2.7301408450704227e-05, "loss": 5.134, "step": 2920 }, { "epoch": 3.27, "learning_rate": 2.7300000000000003e-05, "loss": 5.0577, "step": 2921 }, { "epoch": 3.28, "learning_rate": 2.7298591549295776e-05, "loss": 5.1197, "step": 2922 }, { "epoch": 3.28, "learning_rate": 2.7297183098591552e-05, "loss": 5.1701, "step": 2923 }, { "epoch": 3.28, "learning_rate": 2.7295774647887325e-05, "loss": 4.8645, "step": 2924 }, { "epoch": 3.28, "learning_rate": 2.7294366197183098e-05, "loss": 5.0511, "step": 2925 }, { "epoch": 3.28, "learning_rate": 2.7292957746478874e-05, "loss": 4.6376, "step": 2926 }, { "epoch": 3.28, "learning_rate": 2.7291549295774647e-05, "loss": 5.9568, "step": 2927 }, { "epoch": 3.28, "learning_rate": 2.7290140845070424e-05, "loss": 5.6481, "step": 2928 }, { "epoch": 3.28, "learning_rate": 2.7288732394366197e-05, "loss": 5.7197, "step": 2929 }, { "epoch": 3.28, "learning_rate": 2.7287323943661973e-05, "loss": 5.6232, "step": 2930 }, { "epoch": 3.29, "learning_rate": 2.7285915492957746e-05, "loss": 5.4996, "step": 2931 }, { "epoch": 3.29, "learning_rate": 2.728450704225352e-05, "loss": 5.5295, "step": 2932 }, { "epoch": 3.29, "learning_rate": 2.72830985915493e-05, "loss": 5.5133, "step": 2933 }, { "epoch": 3.29, "learning_rate": 2.728169014084507e-05, "loss": 5.5258, "step": 2934 }, { "epoch": 3.29, "learning_rate": 2.7280281690140848e-05, "loss": 5.5767, "step": 2935 }, { "epoch": 3.29, "learning_rate": 2.727887323943662e-05, "loss": 5.2551, "step": 2936 }, { "epoch": 3.29, "learning_rate": 2.7277464788732397e-05, "loss": 5.4128, "step": 2937 }, { "epoch": 3.29, "learning_rate": 2.727605633802817e-05, "loss": 5.4068, "step": 2938 }, { "epoch": 3.29, "learning_rate": 2.7274647887323943e-05, "loss": 5.4142, "step": 2939 }, { "epoch": 3.3, "learning_rate": 2.727323943661972e-05, "loss": 5.433, "step": 2940 }, { "epoch": 3.3, "learning_rate": 2.7271830985915492e-05, "loss": 5.5776, "step": 2941 }, { "epoch": 3.3, "learning_rate": 2.727042253521127e-05, "loss": 5.4446, "step": 2942 }, { "epoch": 3.3, "learning_rate": 2.726901408450704e-05, "loss": 5.4531, "step": 2943 }, { "epoch": 3.3, "learning_rate": 2.7267605633802818e-05, "loss": 5.4362, "step": 2944 }, { "epoch": 3.3, "learning_rate": 2.726619718309859e-05, "loss": 5.4565, "step": 2945 }, { "epoch": 3.3, "learning_rate": 2.7264788732394367e-05, "loss": 5.468, "step": 2946 }, { "epoch": 3.3, "learning_rate": 2.726338028169014e-05, "loss": 5.3392, "step": 2947 }, { "epoch": 3.3, "learning_rate": 2.7261971830985916e-05, "loss": 5.5057, "step": 2948 }, { "epoch": 3.31, "learning_rate": 2.7260563380281693e-05, "loss": 5.2589, "step": 2949 }, { "epoch": 3.31, "learning_rate": 2.7259154929577466e-05, "loss": 5.3073, "step": 2950 }, { "epoch": 3.31, "learning_rate": 2.7257746478873242e-05, "loss": 5.3511, "step": 2951 }, { "epoch": 3.31, "learning_rate": 2.7256338028169015e-05, "loss": 5.363, "step": 2952 }, { "epoch": 3.31, "learning_rate": 2.725492957746479e-05, "loss": 5.2608, "step": 2953 }, { "epoch": 3.31, "learning_rate": 2.7253521126760564e-05, "loss": 5.3033, "step": 2954 }, { "epoch": 3.31, "learning_rate": 2.7252112676056337e-05, "loss": 5.3978, "step": 2955 }, { "epoch": 3.31, "learning_rate": 2.7250704225352114e-05, "loss": 5.5217, "step": 2956 }, { "epoch": 3.32, "learning_rate": 2.7249295774647886e-05, "loss": 5.3345, "step": 2957 }, { "epoch": 3.32, "learning_rate": 2.7247887323943663e-05, "loss": 5.4852, "step": 2958 }, { "epoch": 3.32, "learning_rate": 2.7246478873239436e-05, "loss": 5.401, "step": 2959 }, { "epoch": 3.32, "learning_rate": 2.7245070422535212e-05, "loss": 5.4103, "step": 2960 }, { "epoch": 3.32, "learning_rate": 2.7243661971830985e-05, "loss": 5.436, "step": 2961 }, { "epoch": 3.32, "learning_rate": 2.724225352112676e-05, "loss": 5.7013, "step": 2962 }, { "epoch": 3.32, "learning_rate": 2.7240845070422534e-05, "loss": 5.4047, "step": 2963 }, { "epoch": 3.32, "learning_rate": 2.723943661971831e-05, "loss": 5.3386, "step": 2964 }, { "epoch": 3.32, "learning_rate": 2.7238028169014087e-05, "loss": 5.3221, "step": 2965 }, { "epoch": 3.33, "learning_rate": 2.723661971830986e-05, "loss": 5.3635, "step": 2966 }, { "epoch": 3.33, "learning_rate": 2.7235211267605636e-05, "loss": 5.325, "step": 2967 }, { "epoch": 3.33, "learning_rate": 2.723380281690141e-05, "loss": 5.1472, "step": 2968 }, { "epoch": 3.33, "learning_rate": 2.7232394366197185e-05, "loss": 5.2585, "step": 2969 }, { "epoch": 3.33, "learning_rate": 2.723098591549296e-05, "loss": 5.2614, "step": 2970 }, { "epoch": 3.33, "learning_rate": 2.722957746478873e-05, "loss": 5.2092, "step": 2971 }, { "epoch": 3.33, "learning_rate": 2.7228169014084508e-05, "loss": 5.0524, "step": 2972 }, { "epoch": 3.33, "learning_rate": 2.722676056338028e-05, "loss": 4.8993, "step": 2973 }, { "epoch": 3.33, "learning_rate": 2.7225352112676057e-05, "loss": 4.8778, "step": 2974 }, { "epoch": 3.34, "learning_rate": 2.722394366197183e-05, "loss": 4.8738, "step": 2975 }, { "epoch": 3.34, "learning_rate": 2.7222535211267606e-05, "loss": 4.4569, "step": 2976 }, { "epoch": 3.34, "learning_rate": 2.722112676056338e-05, "loss": 5.8836, "step": 2977 }, { "epoch": 3.34, "learning_rate": 2.7219718309859156e-05, "loss": 5.5092, "step": 2978 }, { "epoch": 3.34, "learning_rate": 2.7218309859154932e-05, "loss": 5.4719, "step": 2979 }, { "epoch": 3.34, "learning_rate": 2.7216901408450705e-05, "loss": 5.5257, "step": 2980 }, { "epoch": 3.34, "learning_rate": 2.721549295774648e-05, "loss": 5.5899, "step": 2981 }, { "epoch": 3.34, "learning_rate": 2.7214084507042254e-05, "loss": 5.6315, "step": 2982 }, { "epoch": 3.34, "learning_rate": 2.721267605633803e-05, "loss": 5.4529, "step": 2983 }, { "epoch": 3.35, "learning_rate": 2.7211267605633803e-05, "loss": 5.4598, "step": 2984 }, { "epoch": 3.35, "learning_rate": 2.720985915492958e-05, "loss": 5.543, "step": 2985 }, { "epoch": 3.35, "learning_rate": 2.7208450704225353e-05, "loss": 5.3402, "step": 2986 }, { "epoch": 3.35, "learning_rate": 2.7207042253521126e-05, "loss": 5.4273, "step": 2987 }, { "epoch": 3.35, "learning_rate": 2.7205633802816902e-05, "loss": 5.2001, "step": 2988 }, { "epoch": 3.35, "learning_rate": 2.7204225352112675e-05, "loss": 5.4845, "step": 2989 }, { "epoch": 3.35, "learning_rate": 2.720281690140845e-05, "loss": 5.375, "step": 2990 }, { "epoch": 3.35, "learning_rate": 2.7201408450704224e-05, "loss": 5.351, "step": 2991 }, { "epoch": 3.35, "learning_rate": 2.72e-05, "loss": 5.4019, "step": 2992 }, { "epoch": 3.36, "learning_rate": 2.7198591549295773e-05, "loss": 5.4052, "step": 2993 }, { "epoch": 3.36, "learning_rate": 2.7197183098591553e-05, "loss": 5.475, "step": 2994 }, { "epoch": 3.36, "learning_rate": 2.7195774647887326e-05, "loss": 5.3412, "step": 2995 }, { "epoch": 3.36, "learning_rate": 2.71943661971831e-05, "loss": 5.4946, "step": 2996 }, { "epoch": 3.36, "learning_rate": 2.7192957746478875e-05, "loss": 5.4233, "step": 2997 }, { "epoch": 3.36, "learning_rate": 2.7191549295774648e-05, "loss": 5.2527, "step": 2998 }, { "epoch": 3.36, "learning_rate": 2.7190140845070425e-05, "loss": 5.3168, "step": 2999 }, { "epoch": 3.36, "learning_rate": 2.7188732394366198e-05, "loss": 5.4173, "step": 3000 }, { "epoch": 3.36, "eval_loss": 5.345333576202393, "eval_runtime": 881.6995, "eval_samples_per_second": 2.996, "eval_steps_per_second": 0.375, "eval_wer": 1.4877826259420865, "step": 3000 }, { "epoch": 3.36, "learning_rate": 2.7187323943661974e-05, "loss": 5.2776, "step": 3001 }, { "epoch": 3.37, "learning_rate": 2.7185915492957747e-05, "loss": 5.24, "step": 3002 }, { "epoch": 3.37, "learning_rate": 2.718450704225352e-05, "loss": 5.4067, "step": 3003 }, { "epoch": 3.37, "learning_rate": 2.7183098591549296e-05, "loss": 5.4417, "step": 3004 }, { "epoch": 3.37, "learning_rate": 2.718169014084507e-05, "loss": 5.3418, "step": 3005 }, { "epoch": 3.37, "learning_rate": 2.7180281690140845e-05, "loss": 5.4046, "step": 3006 }, { "epoch": 3.37, "learning_rate": 2.717887323943662e-05, "loss": 5.2891, "step": 3007 }, { "epoch": 3.37, "learning_rate": 2.7177464788732395e-05, "loss": 5.1754, "step": 3008 }, { "epoch": 3.37, "learning_rate": 2.717605633802817e-05, "loss": 5.2134, "step": 3009 }, { "epoch": 3.37, "learning_rate": 2.7174647887323947e-05, "loss": 5.3235, "step": 3010 }, { "epoch": 3.38, "learning_rate": 2.717323943661972e-05, "loss": 5.4653, "step": 3011 }, { "epoch": 3.38, "learning_rate": 2.7171830985915493e-05, "loss": 5.2607, "step": 3012 }, { "epoch": 3.38, "learning_rate": 2.717042253521127e-05, "loss": 5.3031, "step": 3013 }, { "epoch": 3.38, "learning_rate": 2.7169014084507042e-05, "loss": 5.3279, "step": 3014 }, { "epoch": 3.38, "learning_rate": 2.716760563380282e-05, "loss": 5.2795, "step": 3015 }, { "epoch": 3.38, "learning_rate": 2.7166197183098592e-05, "loss": 5.2644, "step": 3016 }, { "epoch": 3.38, "learning_rate": 2.7164788732394368e-05, "loss": 5.3668, "step": 3017 }, { "epoch": 3.38, "learning_rate": 2.716338028169014e-05, "loss": 4.9283, "step": 3018 }, { "epoch": 3.38, "learning_rate": 2.7161971830985914e-05, "loss": 5.4063, "step": 3019 }, { "epoch": 3.39, "learning_rate": 2.716056338028169e-05, "loss": 5.1779, "step": 3020 }, { "epoch": 3.39, "learning_rate": 2.7159154929577463e-05, "loss": 5.334, "step": 3021 }, { "epoch": 3.39, "learning_rate": 2.715774647887324e-05, "loss": 5.1132, "step": 3022 }, { "epoch": 3.39, "learning_rate": 2.7156338028169013e-05, "loss": 4.7108, "step": 3023 }, { "epoch": 3.39, "learning_rate": 2.7154929577464792e-05, "loss": 4.7171, "step": 3024 }, { "epoch": 3.39, "learning_rate": 2.7153521126760565e-05, "loss": 4.55, "step": 3025 }, { "epoch": 3.39, "learning_rate": 2.715211267605634e-05, "loss": 4.4849, "step": 3026 }, { "epoch": 3.39, "learning_rate": 2.7150704225352114e-05, "loss": 5.8648, "step": 3027 }, { "epoch": 3.39, "learning_rate": 2.7149295774647887e-05, "loss": 5.6025, "step": 3028 }, { "epoch": 3.4, "learning_rate": 2.7147887323943664e-05, "loss": 5.6692, "step": 3029 }, { "epoch": 3.4, "learning_rate": 2.7146478873239437e-05, "loss": 5.6135, "step": 3030 }, { "epoch": 3.4, "learning_rate": 2.7145070422535213e-05, "loss": 5.3867, "step": 3031 }, { "epoch": 3.4, "learning_rate": 2.7143661971830986e-05, "loss": 5.4987, "step": 3032 }, { "epoch": 3.4, "learning_rate": 2.7142253521126762e-05, "loss": 5.6155, "step": 3033 }, { "epoch": 3.4, "learning_rate": 2.7140845070422535e-05, "loss": 5.3885, "step": 3034 }, { "epoch": 3.4, "learning_rate": 2.7139436619718308e-05, "loss": 5.4349, "step": 3035 }, { "epoch": 3.4, "learning_rate": 2.7138028169014085e-05, "loss": 5.6053, "step": 3036 }, { "epoch": 3.4, "learning_rate": 2.7136619718309857e-05, "loss": 5.3847, "step": 3037 }, { "epoch": 3.41, "learning_rate": 2.7135211267605634e-05, "loss": 5.4099, "step": 3038 }, { "epoch": 3.41, "learning_rate": 2.713380281690141e-05, "loss": 5.4819, "step": 3039 }, { "epoch": 3.41, "learning_rate": 2.7132394366197186e-05, "loss": 5.4586, "step": 3040 }, { "epoch": 3.41, "learning_rate": 2.713098591549296e-05, "loss": 5.2453, "step": 3041 }, { "epoch": 3.41, "learning_rate": 2.7129577464788732e-05, "loss": 5.277, "step": 3042 }, { "epoch": 3.41, "learning_rate": 2.712816901408451e-05, "loss": 5.337, "step": 3043 }, { "epoch": 3.41, "learning_rate": 2.712676056338028e-05, "loss": 5.5333, "step": 3044 }, { "epoch": 3.41, "learning_rate": 2.7125352112676058e-05, "loss": 5.2031, "step": 3045 }, { "epoch": 3.41, "learning_rate": 2.712394366197183e-05, "loss": 5.3513, "step": 3046 }, { "epoch": 3.42, "learning_rate": 2.7122535211267607e-05, "loss": 5.1505, "step": 3047 }, { "epoch": 3.42, "learning_rate": 2.712112676056338e-05, "loss": 5.4156, "step": 3048 }, { "epoch": 3.42, "learning_rate": 2.7119718309859156e-05, "loss": 5.2115, "step": 3049 }, { "epoch": 3.42, "learning_rate": 2.711830985915493e-05, "loss": 5.3715, "step": 3050 }, { "epoch": 3.42, "learning_rate": 2.7116901408450702e-05, "loss": 5.3846, "step": 3051 }, { "epoch": 3.42, "learning_rate": 2.711549295774648e-05, "loss": 5.4057, "step": 3052 }, { "epoch": 3.42, "learning_rate": 2.711408450704225e-05, "loss": 5.3626, "step": 3053 }, { "epoch": 3.42, "learning_rate": 2.711267605633803e-05, "loss": 5.4492, "step": 3054 }, { "epoch": 3.42, "learning_rate": 2.7111267605633804e-05, "loss": 5.3319, "step": 3055 }, { "epoch": 3.43, "learning_rate": 2.710985915492958e-05, "loss": 5.1216, "step": 3056 }, { "epoch": 3.43, "learning_rate": 2.7108450704225354e-05, "loss": 5.4389, "step": 3057 }, { "epoch": 3.43, "learning_rate": 2.7107042253521127e-05, "loss": 5.3765, "step": 3058 }, { "epoch": 3.43, "learning_rate": 2.7105633802816903e-05, "loss": 5.1628, "step": 3059 }, { "epoch": 3.43, "learning_rate": 2.7104225352112676e-05, "loss": 5.2874, "step": 3060 }, { "epoch": 3.43, "learning_rate": 2.7102816901408452e-05, "loss": 5.566, "step": 3061 }, { "epoch": 3.43, "learning_rate": 2.7101408450704225e-05, "loss": 5.3033, "step": 3062 }, { "epoch": 3.43, "learning_rate": 2.71e-05, "loss": 5.1201, "step": 3063 }, { "epoch": 3.43, "learning_rate": 2.7098591549295774e-05, "loss": 5.4784, "step": 3064 }, { "epoch": 3.44, "learning_rate": 2.709718309859155e-05, "loss": 5.341, "step": 3065 }, { "epoch": 3.44, "learning_rate": 2.7095774647887324e-05, "loss": 5.1305, "step": 3066 }, { "epoch": 3.44, "learning_rate": 2.7094366197183097e-05, "loss": 5.2466, "step": 3067 }, { "epoch": 3.44, "learning_rate": 2.7092957746478873e-05, "loss": 5.3437, "step": 3068 }, { "epoch": 3.44, "learning_rate": 2.709154929577465e-05, "loss": 5.3496, "step": 3069 }, { "epoch": 3.44, "learning_rate": 2.7090140845070426e-05, "loss": 5.0908, "step": 3070 }, { "epoch": 3.44, "learning_rate": 2.70887323943662e-05, "loss": 5.1028, "step": 3071 }, { "epoch": 3.44, "learning_rate": 2.7087323943661975e-05, "loss": 5.0997, "step": 3072 }, { "epoch": 3.45, "learning_rate": 2.7085915492957748e-05, "loss": 4.8056, "step": 3073 }, { "epoch": 3.45, "learning_rate": 2.708450704225352e-05, "loss": 4.9495, "step": 3074 }, { "epoch": 3.45, "learning_rate": 2.7083098591549297e-05, "loss": 4.8461, "step": 3075 }, { "epoch": 3.45, "learning_rate": 2.708169014084507e-05, "loss": 4.6203, "step": 3076 }, { "epoch": 3.45, "learning_rate": 2.7080281690140846e-05, "loss": 5.6264, "step": 3077 }, { "epoch": 3.45, "learning_rate": 2.707887323943662e-05, "loss": 5.7099, "step": 3078 }, { "epoch": 3.45, "learning_rate": 2.7077464788732396e-05, "loss": 5.7446, "step": 3079 }, { "epoch": 3.45, "learning_rate": 2.707605633802817e-05, "loss": 5.5944, "step": 3080 }, { "epoch": 3.45, "learning_rate": 2.7074647887323945e-05, "loss": 5.4882, "step": 3081 }, { "epoch": 3.46, "learning_rate": 2.7073239436619718e-05, "loss": 5.5899, "step": 3082 }, { "epoch": 3.46, "learning_rate": 2.707183098591549e-05, "loss": 5.4802, "step": 3083 }, { "epoch": 3.46, "learning_rate": 2.7070422535211267e-05, "loss": 5.37, "step": 3084 }, { "epoch": 3.46, "learning_rate": 2.7069014084507043e-05, "loss": 5.4338, "step": 3085 }, { "epoch": 3.46, "learning_rate": 2.706760563380282e-05, "loss": 5.4084, "step": 3086 }, { "epoch": 3.46, "learning_rate": 2.7066197183098593e-05, "loss": 5.4637, "step": 3087 }, { "epoch": 3.46, "learning_rate": 2.706478873239437e-05, "loss": 5.3796, "step": 3088 }, { "epoch": 3.46, "learning_rate": 2.7063380281690142e-05, "loss": 5.3749, "step": 3089 }, { "epoch": 3.46, "learning_rate": 2.7061971830985915e-05, "loss": 5.272, "step": 3090 }, { "epoch": 3.47, "learning_rate": 2.706056338028169e-05, "loss": 5.3496, "step": 3091 }, { "epoch": 3.47, "learning_rate": 2.7059154929577464e-05, "loss": 5.3032, "step": 3092 }, { "epoch": 3.47, "learning_rate": 2.705774647887324e-05, "loss": 5.322, "step": 3093 }, { "epoch": 3.47, "learning_rate": 2.7056338028169013e-05, "loss": 5.3979, "step": 3094 }, { "epoch": 3.47, "learning_rate": 2.705492957746479e-05, "loss": 5.1783, "step": 3095 }, { "epoch": 3.47, "learning_rate": 2.7053521126760563e-05, "loss": 5.3788, "step": 3096 }, { "epoch": 3.47, "learning_rate": 2.705211267605634e-05, "loss": 5.4311, "step": 3097 }, { "epoch": 3.47, "learning_rate": 2.7050704225352112e-05, "loss": 5.4168, "step": 3098 }, { "epoch": 3.47, "learning_rate": 2.7049295774647885e-05, "loss": 5.3201, "step": 3099 }, { "epoch": 3.48, "learning_rate": 2.7047887323943665e-05, "loss": 5.1767, "step": 3100 }, { "epoch": 3.48, "learning_rate": 2.7046478873239438e-05, "loss": 5.3795, "step": 3101 }, { "epoch": 3.48, "learning_rate": 2.7045070422535214e-05, "loss": 5.3605, "step": 3102 }, { "epoch": 3.48, "learning_rate": 2.7043661971830987e-05, "loss": 5.3501, "step": 3103 }, { "epoch": 3.48, "learning_rate": 2.7042253521126763e-05, "loss": 5.2579, "step": 3104 }, { "epoch": 3.48, "learning_rate": 2.7040845070422536e-05, "loss": 5.3407, "step": 3105 }, { "epoch": 3.48, "learning_rate": 2.703943661971831e-05, "loss": 5.2363, "step": 3106 }, { "epoch": 3.48, "learning_rate": 2.7038028169014085e-05, "loss": 5.3502, "step": 3107 }, { "epoch": 3.48, "learning_rate": 2.703661971830986e-05, "loss": 5.1771, "step": 3108 }, { "epoch": 3.49, "learning_rate": 2.7035211267605635e-05, "loss": 5.3144, "step": 3109 }, { "epoch": 3.49, "learning_rate": 2.7033802816901408e-05, "loss": 5.3631, "step": 3110 }, { "epoch": 3.49, "learning_rate": 2.7032394366197184e-05, "loss": 5.2457, "step": 3111 }, { "epoch": 3.49, "learning_rate": 2.7030985915492957e-05, "loss": 5.2859, "step": 3112 }, { "epoch": 3.49, "learning_rate": 2.7029577464788733e-05, "loss": 5.3168, "step": 3113 }, { "epoch": 3.49, "learning_rate": 2.7028169014084506e-05, "loss": 5.2774, "step": 3114 }, { "epoch": 3.49, "learning_rate": 2.7026760563380283e-05, "loss": 5.1403, "step": 3115 }, { "epoch": 3.49, "learning_rate": 2.702535211267606e-05, "loss": 5.1179, "step": 3116 }, { "epoch": 3.49, "learning_rate": 2.7023943661971832e-05, "loss": 5.1616, "step": 3117 }, { "epoch": 3.5, "learning_rate": 2.7022535211267608e-05, "loss": 5.184, "step": 3118 }, { "epoch": 3.5, "learning_rate": 2.702112676056338e-05, "loss": 5.0349, "step": 3119 }, { "epoch": 3.5, "learning_rate": 2.7019718309859157e-05, "loss": 5.1133, "step": 3120 }, { "epoch": 3.5, "learning_rate": 2.701830985915493e-05, "loss": 5.0319, "step": 3121 }, { "epoch": 3.5, "learning_rate": 2.7016901408450703e-05, "loss": 4.9884, "step": 3122 }, { "epoch": 3.5, "learning_rate": 2.701549295774648e-05, "loss": 5.2658, "step": 3123 }, { "epoch": 3.5, "learning_rate": 2.7014084507042253e-05, "loss": 4.7904, "step": 3124 }, { "epoch": 3.5, "learning_rate": 2.701267605633803e-05, "loss": 4.5897, "step": 3125 }, { "epoch": 3.5, "learning_rate": 2.7011267605633802e-05, "loss": 4.7472, "step": 3126 }, { "epoch": 3.51, "learning_rate": 2.7009859154929578e-05, "loss": 5.6785, "step": 3127 }, { "epoch": 3.51, "learning_rate": 2.700845070422535e-05, "loss": 5.6781, "step": 3128 }, { "epoch": 3.51, "learning_rate": 2.7007042253521127e-05, "loss": 5.7199, "step": 3129 }, { "epoch": 3.51, "learning_rate": 2.7005633802816904e-05, "loss": 5.5444, "step": 3130 }, { "epoch": 3.51, "learning_rate": 2.7004225352112677e-05, "loss": 5.3224, "step": 3131 }, { "epoch": 3.51, "learning_rate": 2.7002816901408453e-05, "loss": 5.4416, "step": 3132 }, { "epoch": 3.51, "learning_rate": 2.7001408450704226e-05, "loss": 5.6123, "step": 3133 }, { "epoch": 3.51, "learning_rate": 2.7000000000000002e-05, "loss": 5.4644, "step": 3134 }, { "epoch": 3.51, "learning_rate": 2.6998591549295775e-05, "loss": 5.3418, "step": 3135 }, { "epoch": 3.52, "learning_rate": 2.699718309859155e-05, "loss": 5.4517, "step": 3136 }, { "epoch": 3.52, "learning_rate": 2.6995774647887325e-05, "loss": 5.5456, "step": 3137 }, { "epoch": 3.52, "learning_rate": 2.6994366197183098e-05, "loss": 5.3807, "step": 3138 }, { "epoch": 3.52, "learning_rate": 2.6992957746478874e-05, "loss": 5.1675, "step": 3139 }, { "epoch": 3.52, "learning_rate": 2.6991549295774647e-05, "loss": 5.3487, "step": 3140 }, { "epoch": 3.52, "learning_rate": 2.6990140845070423e-05, "loss": 5.3873, "step": 3141 }, { "epoch": 3.52, "learning_rate": 2.6988732394366196e-05, "loss": 5.4512, "step": 3142 }, { "epoch": 3.52, "learning_rate": 2.6987323943661972e-05, "loss": 5.4976, "step": 3143 }, { "epoch": 3.52, "learning_rate": 2.6985915492957745e-05, "loss": 5.3522, "step": 3144 }, { "epoch": 3.53, "learning_rate": 2.6984507042253522e-05, "loss": 5.2938, "step": 3145 }, { "epoch": 3.53, "learning_rate": 2.6983098591549298e-05, "loss": 5.3915, "step": 3146 }, { "epoch": 3.53, "learning_rate": 2.698169014084507e-05, "loss": 5.2682, "step": 3147 }, { "epoch": 3.53, "learning_rate": 2.6980281690140847e-05, "loss": 5.4008, "step": 3148 }, { "epoch": 3.53, "learning_rate": 2.697887323943662e-05, "loss": 5.3786, "step": 3149 }, { "epoch": 3.53, "learning_rate": 2.6977464788732397e-05, "loss": 5.299, "step": 3150 }, { "epoch": 3.53, "learning_rate": 2.697605633802817e-05, "loss": 5.2095, "step": 3151 }, { "epoch": 3.53, "learning_rate": 2.6974647887323946e-05, "loss": 5.2901, "step": 3152 }, { "epoch": 3.53, "learning_rate": 2.697323943661972e-05, "loss": 5.2191, "step": 3153 }, { "epoch": 3.54, "learning_rate": 2.6971830985915492e-05, "loss": 5.3375, "step": 3154 }, { "epoch": 3.54, "learning_rate": 2.6970422535211268e-05, "loss": 5.3385, "step": 3155 }, { "epoch": 3.54, "learning_rate": 2.696901408450704e-05, "loss": 5.3681, "step": 3156 }, { "epoch": 3.54, "learning_rate": 2.6967605633802817e-05, "loss": 5.2611, "step": 3157 }, { "epoch": 3.54, "learning_rate": 2.696619718309859e-05, "loss": 5.2198, "step": 3158 }, { "epoch": 3.54, "learning_rate": 2.6964788732394367e-05, "loss": 5.2643, "step": 3159 }, { "epoch": 3.54, "learning_rate": 2.6963380281690143e-05, "loss": 5.3243, "step": 3160 }, { "epoch": 3.54, "learning_rate": 2.6961971830985916e-05, "loss": 5.3634, "step": 3161 }, { "epoch": 3.54, "learning_rate": 2.6960563380281692e-05, "loss": 5.0472, "step": 3162 }, { "epoch": 3.55, "learning_rate": 2.6959154929577465e-05, "loss": 5.1813, "step": 3163 }, { "epoch": 3.55, "learning_rate": 2.695774647887324e-05, "loss": 5.0075, "step": 3164 }, { "epoch": 3.55, "learning_rate": 2.6956338028169014e-05, "loss": 5.282, "step": 3165 }, { "epoch": 3.55, "learning_rate": 2.695492957746479e-05, "loss": 5.257, "step": 3166 }, { "epoch": 3.55, "learning_rate": 2.6953521126760564e-05, "loss": 5.2641, "step": 3167 }, { "epoch": 3.55, "learning_rate": 2.695211267605634e-05, "loss": 5.0278, "step": 3168 }, { "epoch": 3.55, "learning_rate": 2.6950704225352113e-05, "loss": 5.028, "step": 3169 }, { "epoch": 3.55, "learning_rate": 2.6949295774647886e-05, "loss": 5.0472, "step": 3170 }, { "epoch": 3.55, "learning_rate": 2.6947887323943662e-05, "loss": 5.0503, "step": 3171 }, { "epoch": 3.56, "learning_rate": 2.6946478873239435e-05, "loss": 5.1312, "step": 3172 }, { "epoch": 3.56, "learning_rate": 2.694507042253521e-05, "loss": 4.8723, "step": 3173 }, { "epoch": 3.56, "learning_rate": 2.6943661971830984e-05, "loss": 4.9849, "step": 3174 }, { "epoch": 3.56, "learning_rate": 2.6942253521126764e-05, "loss": 4.6898, "step": 3175 }, { "epoch": 3.56, "learning_rate": 2.6940845070422537e-05, "loss": 4.3414, "step": 3176 }, { "epoch": 3.56, "learning_rate": 2.693943661971831e-05, "loss": 5.5601, "step": 3177 }, { "epoch": 3.56, "learning_rate": 2.6938028169014086e-05, "loss": 5.5501, "step": 3178 }, { "epoch": 3.56, "learning_rate": 2.693661971830986e-05, "loss": 5.472, "step": 3179 }, { "epoch": 3.57, "learning_rate": 2.6935211267605636e-05, "loss": 5.6708, "step": 3180 }, { "epoch": 3.57, "learning_rate": 2.693380281690141e-05, "loss": 5.6321, "step": 3181 }, { "epoch": 3.57, "learning_rate": 2.6932394366197185e-05, "loss": 5.5212, "step": 3182 }, { "epoch": 3.57, "learning_rate": 2.6930985915492958e-05, "loss": 5.4937, "step": 3183 }, { "epoch": 3.57, "learning_rate": 2.6929577464788734e-05, "loss": 5.311, "step": 3184 }, { "epoch": 3.57, "learning_rate": 2.6928169014084507e-05, "loss": 5.379, "step": 3185 }, { "epoch": 3.57, "learning_rate": 2.692676056338028e-05, "loss": 5.3097, "step": 3186 }, { "epoch": 3.57, "learning_rate": 2.6925352112676056e-05, "loss": 5.3649, "step": 3187 }, { "epoch": 3.57, "learning_rate": 2.692394366197183e-05, "loss": 5.292, "step": 3188 }, { "epoch": 3.58, "learning_rate": 2.6922535211267606e-05, "loss": 5.3489, "step": 3189 }, { "epoch": 3.58, "learning_rate": 2.6921126760563382e-05, "loss": 5.2364, "step": 3190 }, { "epoch": 3.58, "learning_rate": 2.691971830985916e-05, "loss": 5.3869, "step": 3191 }, { "epoch": 3.58, "learning_rate": 2.691830985915493e-05, "loss": 5.4194, "step": 3192 }, { "epoch": 3.58, "learning_rate": 2.6916901408450704e-05, "loss": 5.4237, "step": 3193 }, { "epoch": 3.58, "learning_rate": 2.691549295774648e-05, "loss": 5.2395, "step": 3194 }, { "epoch": 3.58, "learning_rate": 2.6914084507042254e-05, "loss": 5.1673, "step": 3195 }, { "epoch": 3.58, "learning_rate": 2.691267605633803e-05, "loss": 5.3294, "step": 3196 }, { "epoch": 3.58, "learning_rate": 2.6911267605633803e-05, "loss": 5.3424, "step": 3197 }, { "epoch": 3.59, "learning_rate": 2.690985915492958e-05, "loss": 5.4849, "step": 3198 }, { "epoch": 3.59, "learning_rate": 2.6908450704225352e-05, "loss": 5.2405, "step": 3199 }, { "epoch": 3.59, "learning_rate": 2.690704225352113e-05, "loss": 5.2368, "step": 3200 }, { "epoch": 3.59, "learning_rate": 2.69056338028169e-05, "loss": 5.2781, "step": 3201 }, { "epoch": 3.59, "learning_rate": 2.6904225352112674e-05, "loss": 5.315, "step": 3202 }, { "epoch": 3.59, "learning_rate": 2.690281690140845e-05, "loss": 5.2424, "step": 3203 }, { "epoch": 3.59, "learning_rate": 2.6901408450704224e-05, "loss": 5.1412, "step": 3204 }, { "epoch": 3.59, "learning_rate": 2.69e-05, "loss": 5.0895, "step": 3205 }, { "epoch": 3.59, "learning_rate": 2.6898591549295776e-05, "loss": 5.3938, "step": 3206 }, { "epoch": 3.6, "learning_rate": 2.6897183098591553e-05, "loss": 5.3482, "step": 3207 }, { "epoch": 3.6, "learning_rate": 2.6895774647887326e-05, "loss": 5.3179, "step": 3208 }, { "epoch": 3.6, "learning_rate": 2.68943661971831e-05, "loss": 5.2586, "step": 3209 }, { "epoch": 3.6, "learning_rate": 2.6892957746478875e-05, "loss": 5.1268, "step": 3210 }, { "epoch": 3.6, "learning_rate": 2.6891549295774648e-05, "loss": 5.1943, "step": 3211 }, { "epoch": 3.6, "learning_rate": 2.6890140845070424e-05, "loss": 5.2978, "step": 3212 }, { "epoch": 3.6, "learning_rate": 2.6888732394366197e-05, "loss": 5.1597, "step": 3213 }, { "epoch": 3.6, "learning_rate": 2.6887323943661973e-05, "loss": 5.1847, "step": 3214 }, { "epoch": 3.6, "learning_rate": 2.6885915492957746e-05, "loss": 5.1909, "step": 3215 }, { "epoch": 3.61, "learning_rate": 2.6884507042253523e-05, "loss": 5.4311, "step": 3216 }, { "epoch": 3.61, "learning_rate": 2.6883098591549296e-05, "loss": 5.149, "step": 3217 }, { "epoch": 3.61, "learning_rate": 2.688169014084507e-05, "loss": 5.0373, "step": 3218 }, { "epoch": 3.61, "learning_rate": 2.6880281690140845e-05, "loss": 5.1738, "step": 3219 }, { "epoch": 3.61, "learning_rate": 2.6878873239436618e-05, "loss": 4.9159, "step": 3220 }, { "epoch": 3.61, "learning_rate": 2.6877464788732398e-05, "loss": 5.0831, "step": 3221 }, { "epoch": 3.61, "learning_rate": 2.687605633802817e-05, "loss": 4.8899, "step": 3222 }, { "epoch": 3.61, "learning_rate": 2.6874647887323947e-05, "loss": 4.7628, "step": 3223 }, { "epoch": 3.61, "learning_rate": 2.687323943661972e-05, "loss": 4.6876, "step": 3224 }, { "epoch": 3.62, "learning_rate": 2.6871830985915493e-05, "loss": 4.9007, "step": 3225 }, { "epoch": 3.62, "learning_rate": 2.687042253521127e-05, "loss": 4.3788, "step": 3226 }, { "epoch": 3.62, "learning_rate": 2.6869014084507042e-05, "loss": 5.7212, "step": 3227 }, { "epoch": 3.62, "learning_rate": 2.6867605633802818e-05, "loss": 5.8154, "step": 3228 }, { "epoch": 3.62, "learning_rate": 2.686619718309859e-05, "loss": 5.5962, "step": 3229 }, { "epoch": 3.62, "learning_rate": 2.6864788732394368e-05, "loss": 5.2695, "step": 3230 }, { "epoch": 3.62, "learning_rate": 2.686338028169014e-05, "loss": 5.3825, "step": 3231 }, { "epoch": 3.62, "learning_rate": 2.6861971830985917e-05, "loss": 5.4339, "step": 3232 }, { "epoch": 3.62, "learning_rate": 2.686056338028169e-05, "loss": 5.3802, "step": 3233 }, { "epoch": 3.63, "learning_rate": 2.6859154929577463e-05, "loss": 5.3241, "step": 3234 }, { "epoch": 3.63, "learning_rate": 2.685774647887324e-05, "loss": 5.4268, "step": 3235 }, { "epoch": 3.63, "learning_rate": 2.6856338028169015e-05, "loss": 5.359, "step": 3236 }, { "epoch": 3.63, "learning_rate": 2.6854929577464792e-05, "loss": 5.432, "step": 3237 }, { "epoch": 3.63, "learning_rate": 2.6853521126760565e-05, "loss": 5.3548, "step": 3238 }, { "epoch": 3.63, "learning_rate": 2.685211267605634e-05, "loss": 5.3699, "step": 3239 }, { "epoch": 3.63, "learning_rate": 2.6850704225352114e-05, "loss": 5.1775, "step": 3240 }, { "epoch": 3.63, "learning_rate": 2.6849295774647887e-05, "loss": 5.3467, "step": 3241 }, { "epoch": 3.63, "learning_rate": 2.6847887323943663e-05, "loss": 5.2572, "step": 3242 }, { "epoch": 3.64, "learning_rate": 2.6846478873239436e-05, "loss": 5.1293, "step": 3243 }, { "epoch": 3.64, "learning_rate": 2.6845070422535213e-05, "loss": 5.2385, "step": 3244 }, { "epoch": 3.64, "learning_rate": 2.6843661971830985e-05, "loss": 5.2703, "step": 3245 }, { "epoch": 3.64, "learning_rate": 2.6842253521126762e-05, "loss": 5.2097, "step": 3246 }, { "epoch": 3.64, "learning_rate": 2.6840845070422535e-05, "loss": 5.3361, "step": 3247 }, { "epoch": 3.64, "learning_rate": 2.683943661971831e-05, "loss": 5.124, "step": 3248 }, { "epoch": 3.64, "learning_rate": 2.6838028169014084e-05, "loss": 5.3656, "step": 3249 }, { "epoch": 3.64, "learning_rate": 2.6836619718309857e-05, "loss": 5.4617, "step": 3250 }, { "epoch": 3.64, "learning_rate": 2.6835211267605637e-05, "loss": 5.3475, "step": 3251 }, { "epoch": 3.65, "learning_rate": 2.683380281690141e-05, "loss": 5.2684, "step": 3252 }, { "epoch": 3.65, "learning_rate": 2.6832394366197186e-05, "loss": 5.1428, "step": 3253 }, { "epoch": 3.65, "learning_rate": 2.683098591549296e-05, "loss": 5.2564, "step": 3254 }, { "epoch": 3.65, "learning_rate": 2.6829577464788735e-05, "loss": 5.463, "step": 3255 }, { "epoch": 3.65, "learning_rate": 2.6828169014084508e-05, "loss": 5.1175, "step": 3256 }, { "epoch": 3.65, "learning_rate": 2.682676056338028e-05, "loss": 5.3776, "step": 3257 }, { "epoch": 3.65, "learning_rate": 2.6825352112676057e-05, "loss": 5.2486, "step": 3258 }, { "epoch": 3.65, "learning_rate": 2.682394366197183e-05, "loss": 5.2609, "step": 3259 }, { "epoch": 3.65, "learning_rate": 2.6822535211267607e-05, "loss": 5.1934, "step": 3260 }, { "epoch": 3.66, "learning_rate": 2.682112676056338e-05, "loss": 5.3584, "step": 3261 }, { "epoch": 3.66, "learning_rate": 2.6819718309859156e-05, "loss": 5.1008, "step": 3262 }, { "epoch": 3.66, "learning_rate": 2.681830985915493e-05, "loss": 5.1427, "step": 3263 }, { "epoch": 3.66, "learning_rate": 2.6816901408450705e-05, "loss": 5.2889, "step": 3264 }, { "epoch": 3.66, "learning_rate": 2.6815492957746478e-05, "loss": 5.2197, "step": 3265 }, { "epoch": 3.66, "learning_rate": 2.6814084507042255e-05, "loss": 5.0106, "step": 3266 }, { "epoch": 3.66, "learning_rate": 2.681267605633803e-05, "loss": 5.4212, "step": 3267 }, { "epoch": 3.66, "learning_rate": 2.6811267605633804e-05, "loss": 4.7648, "step": 3268 }, { "epoch": 3.66, "learning_rate": 2.680985915492958e-05, "loss": 5.1272, "step": 3269 }, { "epoch": 3.67, "learning_rate": 2.6808450704225353e-05, "loss": 5.2341, "step": 3270 }, { "epoch": 3.67, "learning_rate": 2.680704225352113e-05, "loss": 5.1492, "step": 3271 }, { "epoch": 3.67, "learning_rate": 2.6805633802816902e-05, "loss": 4.7626, "step": 3272 }, { "epoch": 3.67, "learning_rate": 2.6804225352112675e-05, "loss": 4.8248, "step": 3273 }, { "epoch": 3.67, "learning_rate": 2.680281690140845e-05, "loss": 4.7999, "step": 3274 }, { "epoch": 3.67, "learning_rate": 2.6801408450704225e-05, "loss": 4.9112, "step": 3275 }, { "epoch": 3.67, "learning_rate": 2.68e-05, "loss": 4.5837, "step": 3276 }, { "epoch": 3.67, "learning_rate": 2.6798591549295774e-05, "loss": 5.8708, "step": 3277 }, { "epoch": 3.67, "learning_rate": 2.679718309859155e-05, "loss": 5.4242, "step": 3278 }, { "epoch": 3.68, "learning_rate": 2.6795774647887323e-05, "loss": 5.4539, "step": 3279 }, { "epoch": 3.68, "learning_rate": 2.67943661971831e-05, "loss": 5.5004, "step": 3280 }, { "epoch": 3.68, "learning_rate": 2.6792957746478876e-05, "loss": 5.4329, "step": 3281 }, { "epoch": 3.68, "learning_rate": 2.679154929577465e-05, "loss": 5.2209, "step": 3282 }, { "epoch": 3.68, "learning_rate": 2.6790140845070425e-05, "loss": 5.364, "step": 3283 }, { "epoch": 3.68, "learning_rate": 2.6788732394366198e-05, "loss": 5.3391, "step": 3284 }, { "epoch": 3.68, "learning_rate": 2.6787323943661974e-05, "loss": 5.3153, "step": 3285 }, { "epoch": 3.68, "learning_rate": 2.6785915492957747e-05, "loss": 5.3588, "step": 3286 }, { "epoch": 3.68, "learning_rate": 2.6784507042253524e-05, "loss": 5.4233, "step": 3287 }, { "epoch": 3.69, "learning_rate": 2.6783098591549297e-05, "loss": 5.2165, "step": 3288 }, { "epoch": 3.69, "learning_rate": 2.678169014084507e-05, "loss": 5.3299, "step": 3289 }, { "epoch": 3.69, "learning_rate": 2.6780281690140846e-05, "loss": 5.2316, "step": 3290 }, { "epoch": 3.69, "learning_rate": 2.677887323943662e-05, "loss": 5.0231, "step": 3291 }, { "epoch": 3.69, "learning_rate": 2.6777464788732395e-05, "loss": 5.441, "step": 3292 }, { "epoch": 3.69, "learning_rate": 2.6776056338028168e-05, "loss": 5.2679, "step": 3293 }, { "epoch": 3.69, "learning_rate": 2.6774647887323944e-05, "loss": 5.2396, "step": 3294 }, { "epoch": 3.69, "learning_rate": 2.6773239436619717e-05, "loss": 5.301, "step": 3295 }, { "epoch": 3.7, "learning_rate": 2.6771830985915494e-05, "loss": 5.1742, "step": 3296 }, { "epoch": 3.7, "learning_rate": 2.677042253521127e-05, "loss": 5.1835, "step": 3297 }, { "epoch": 3.7, "learning_rate": 2.6769014084507043e-05, "loss": 5.4293, "step": 3298 }, { "epoch": 3.7, "learning_rate": 2.676760563380282e-05, "loss": 5.2514, "step": 3299 }, { "epoch": 3.7, "learning_rate": 2.6766197183098592e-05, "loss": 5.2126, "step": 3300 }, { "epoch": 3.7, "learning_rate": 2.676478873239437e-05, "loss": 5.1803, "step": 3301 }, { "epoch": 3.7, "learning_rate": 2.676338028169014e-05, "loss": 5.3314, "step": 3302 }, { "epoch": 3.7, "learning_rate": 2.6761971830985918e-05, "loss": 5.3469, "step": 3303 }, { "epoch": 3.7, "learning_rate": 2.676056338028169e-05, "loss": 5.2601, "step": 3304 }, { "epoch": 3.71, "learning_rate": 2.6759154929577464e-05, "loss": 5.1821, "step": 3305 }, { "epoch": 3.71, "learning_rate": 2.675774647887324e-05, "loss": 5.187, "step": 3306 }, { "epoch": 3.71, "learning_rate": 2.6756338028169013e-05, "loss": 5.3822, "step": 3307 }, { "epoch": 3.71, "learning_rate": 2.675492957746479e-05, "loss": 5.2406, "step": 3308 }, { "epoch": 3.71, "learning_rate": 2.6753521126760562e-05, "loss": 5.0446, "step": 3309 }, { "epoch": 3.71, "learning_rate": 2.675211267605634e-05, "loss": 5.0239, "step": 3310 }, { "epoch": 3.71, "learning_rate": 2.6750704225352115e-05, "loss": 5.2138, "step": 3311 }, { "epoch": 3.71, "learning_rate": 2.6749295774647888e-05, "loss": 5.2618, "step": 3312 }, { "epoch": 3.71, "learning_rate": 2.6747887323943664e-05, "loss": 5.1742, "step": 3313 }, { "epoch": 3.72, "learning_rate": 2.6746478873239437e-05, "loss": 5.2376, "step": 3314 }, { "epoch": 3.72, "learning_rate": 2.6745070422535213e-05, "loss": 5.1824, "step": 3315 }, { "epoch": 3.72, "learning_rate": 2.6743661971830986e-05, "loss": 5.1216, "step": 3316 }, { "epoch": 3.72, "learning_rate": 2.6742253521126763e-05, "loss": 5.0719, "step": 3317 }, { "epoch": 3.72, "learning_rate": 2.6740845070422536e-05, "loss": 5.0451, "step": 3318 }, { "epoch": 3.72, "learning_rate": 2.6739436619718312e-05, "loss": 5.245, "step": 3319 }, { "epoch": 3.72, "learning_rate": 2.6738028169014085e-05, "loss": 5.0591, "step": 3320 }, { "epoch": 3.72, "learning_rate": 2.6736619718309858e-05, "loss": 4.9071, "step": 3321 }, { "epoch": 3.72, "learning_rate": 2.6735211267605634e-05, "loss": 4.9047, "step": 3322 }, { "epoch": 3.73, "learning_rate": 2.6733802816901407e-05, "loss": 4.5326, "step": 3323 }, { "epoch": 3.73, "learning_rate": 2.6732394366197184e-05, "loss": 5.0724, "step": 3324 }, { "epoch": 3.73, "learning_rate": 2.6730985915492956e-05, "loss": 4.7521, "step": 3325 }, { "epoch": 3.73, "learning_rate": 2.6729577464788733e-05, "loss": 4.2008, "step": 3326 }, { "epoch": 3.73, "learning_rate": 2.672816901408451e-05, "loss": 5.5437, "step": 3327 }, { "epoch": 3.73, "learning_rate": 2.6726760563380282e-05, "loss": 5.3795, "step": 3328 }, { "epoch": 3.73, "learning_rate": 2.672535211267606e-05, "loss": 5.6519, "step": 3329 }, { "epoch": 3.73, "learning_rate": 2.672394366197183e-05, "loss": 5.6062, "step": 3330 }, { "epoch": 3.73, "learning_rate": 2.6722535211267608e-05, "loss": 5.3543, "step": 3331 }, { "epoch": 3.74, "learning_rate": 2.672112676056338e-05, "loss": 5.5117, "step": 3332 }, { "epoch": 3.74, "learning_rate": 2.6719718309859157e-05, "loss": 5.4627, "step": 3333 }, { "epoch": 3.74, "learning_rate": 2.671830985915493e-05, "loss": 5.3789, "step": 3334 }, { "epoch": 3.74, "learning_rate": 2.6716901408450706e-05, "loss": 5.3528, "step": 3335 }, { "epoch": 3.74, "learning_rate": 2.671549295774648e-05, "loss": 5.3414, "step": 3336 }, { "epoch": 3.74, "learning_rate": 2.6714084507042252e-05, "loss": 5.2881, "step": 3337 }, { "epoch": 3.74, "learning_rate": 2.671267605633803e-05, "loss": 5.1847, "step": 3338 }, { "epoch": 3.74, "learning_rate": 2.67112676056338e-05, "loss": 5.345, "step": 3339 }, { "epoch": 3.74, "learning_rate": 2.6709859154929578e-05, "loss": 5.2555, "step": 3340 }, { "epoch": 3.75, "learning_rate": 2.670845070422535e-05, "loss": 5.3415, "step": 3341 }, { "epoch": 3.75, "learning_rate": 2.670704225352113e-05, "loss": 5.178, "step": 3342 }, { "epoch": 3.75, "learning_rate": 2.6705633802816903e-05, "loss": 5.1852, "step": 3343 }, { "epoch": 3.75, "learning_rate": 2.6704225352112676e-05, "loss": 5.1235, "step": 3344 }, { "epoch": 3.75, "learning_rate": 2.6702816901408453e-05, "loss": 5.3732, "step": 3345 }, { "epoch": 3.75, "learning_rate": 2.6701408450704226e-05, "loss": 5.192, "step": 3346 }, { "epoch": 3.75, "learning_rate": 2.6700000000000002e-05, "loss": 5.1638, "step": 3347 }, { "epoch": 3.75, "learning_rate": 2.6698591549295775e-05, "loss": 5.1055, "step": 3348 }, { "epoch": 3.75, "learning_rate": 2.669718309859155e-05, "loss": 5.0231, "step": 3349 }, { "epoch": 3.76, "learning_rate": 2.6695774647887324e-05, "loss": 5.1712, "step": 3350 }, { "epoch": 3.76, "learning_rate": 2.66943661971831e-05, "loss": 5.199, "step": 3351 }, { "epoch": 3.76, "learning_rate": 2.6692957746478873e-05, "loss": 5.2875, "step": 3352 }, { "epoch": 3.76, "learning_rate": 2.6691549295774646e-05, "loss": 5.2164, "step": 3353 }, { "epoch": 3.76, "learning_rate": 2.6690140845070423e-05, "loss": 5.3624, "step": 3354 }, { "epoch": 3.76, "learning_rate": 2.6688732394366196e-05, "loss": 5.3335, "step": 3355 }, { "epoch": 3.76, "learning_rate": 2.6687323943661972e-05, "loss": 5.1924, "step": 3356 }, { "epoch": 3.76, "learning_rate": 2.6685915492957748e-05, "loss": 5.3714, "step": 3357 }, { "epoch": 3.76, "learning_rate": 2.6684507042253525e-05, "loss": 5.0983, "step": 3358 }, { "epoch": 3.77, "learning_rate": 2.6683098591549298e-05, "loss": 5.1528, "step": 3359 }, { "epoch": 3.77, "learning_rate": 2.668169014084507e-05, "loss": 5.1497, "step": 3360 }, { "epoch": 3.77, "learning_rate": 2.6680281690140847e-05, "loss": 5.1206, "step": 3361 }, { "epoch": 3.77, "learning_rate": 2.667887323943662e-05, "loss": 5.1632, "step": 3362 }, { "epoch": 3.77, "learning_rate": 2.6677464788732396e-05, "loss": 5.1695, "step": 3363 }, { "epoch": 3.77, "learning_rate": 2.667605633802817e-05, "loss": 5.1974, "step": 3364 }, { "epoch": 3.77, "learning_rate": 2.6674647887323945e-05, "loss": 5.2894, "step": 3365 }, { "epoch": 3.77, "learning_rate": 2.6673239436619718e-05, "loss": 5.1635, "step": 3366 }, { "epoch": 3.77, "learning_rate": 2.6671830985915495e-05, "loss": 5.2477, "step": 3367 }, { "epoch": 3.78, "learning_rate": 2.6670422535211268e-05, "loss": 5.1001, "step": 3368 }, { "epoch": 3.78, "learning_rate": 2.666901408450704e-05, "loss": 5.1521, "step": 3369 }, { "epoch": 3.78, "learning_rate": 2.6667605633802817e-05, "loss": 5.0003, "step": 3370 }, { "epoch": 3.78, "learning_rate": 2.666619718309859e-05, "loss": 4.7922, "step": 3371 }, { "epoch": 3.78, "learning_rate": 2.666478873239437e-05, "loss": 4.4286, "step": 3372 }, { "epoch": 3.78, "learning_rate": 2.6663380281690142e-05, "loss": 4.8517, "step": 3373 }, { "epoch": 3.78, "learning_rate": 2.666197183098592e-05, "loss": 4.6345, "step": 3374 }, { "epoch": 3.78, "learning_rate": 2.6660563380281692e-05, "loss": 4.749, "step": 3375 }, { "epoch": 3.78, "learning_rate": 2.6659154929577465e-05, "loss": 4.3196, "step": 3376 }, { "epoch": 3.79, "learning_rate": 2.665774647887324e-05, "loss": 5.4565, "step": 3377 }, { "epoch": 3.79, "learning_rate": 2.6656338028169014e-05, "loss": 5.6257, "step": 3378 }, { "epoch": 3.79, "learning_rate": 2.665492957746479e-05, "loss": 5.5673, "step": 3379 }, { "epoch": 3.79, "learning_rate": 2.6653521126760563e-05, "loss": 5.6474, "step": 3380 }, { "epoch": 3.79, "learning_rate": 2.665211267605634e-05, "loss": 5.4491, "step": 3381 }, { "epoch": 3.79, "learning_rate": 2.6650704225352112e-05, "loss": 5.3844, "step": 3382 }, { "epoch": 3.79, "learning_rate": 2.664929577464789e-05, "loss": 5.5364, "step": 3383 }, { "epoch": 3.79, "learning_rate": 2.6647887323943662e-05, "loss": 5.3637, "step": 3384 }, { "epoch": 3.79, "learning_rate": 2.6646478873239435e-05, "loss": 5.378, "step": 3385 }, { "epoch": 3.8, "learning_rate": 2.664507042253521e-05, "loss": 5.2606, "step": 3386 }, { "epoch": 3.8, "learning_rate": 2.6643661971830987e-05, "loss": 5.2463, "step": 3387 }, { "epoch": 3.8, "learning_rate": 2.6642253521126764e-05, "loss": 5.2937, "step": 3388 }, { "epoch": 3.8, "learning_rate": 2.6640845070422537e-05, "loss": 5.0887, "step": 3389 }, { "epoch": 3.8, "learning_rate": 2.6639436619718313e-05, "loss": 5.1171, "step": 3390 }, { "epoch": 3.8, "learning_rate": 2.6638028169014086e-05, "loss": 5.2537, "step": 3391 }, { "epoch": 3.8, "learning_rate": 2.663661971830986e-05, "loss": 5.3082, "step": 3392 }, { "epoch": 3.8, "learning_rate": 2.6635211267605635e-05, "loss": 5.1567, "step": 3393 }, { "epoch": 3.8, "learning_rate": 2.6633802816901408e-05, "loss": 4.9999, "step": 3394 }, { "epoch": 3.81, "learning_rate": 2.6632394366197184e-05, "loss": 5.277, "step": 3395 }, { "epoch": 3.81, "learning_rate": 2.6630985915492957e-05, "loss": 5.3076, "step": 3396 }, { "epoch": 3.81, "learning_rate": 2.6629577464788734e-05, "loss": 5.2528, "step": 3397 }, { "epoch": 3.81, "learning_rate": 2.6628169014084507e-05, "loss": 5.2202, "step": 3398 }, { "epoch": 3.81, "learning_rate": 2.662676056338028e-05, "loss": 5.1282, "step": 3399 }, { "epoch": 3.81, "learning_rate": 2.6625352112676056e-05, "loss": 5.174, "step": 3400 }, { "epoch": 3.81, "learning_rate": 2.662394366197183e-05, "loss": 5.2275, "step": 3401 }, { "epoch": 3.81, "learning_rate": 2.662253521126761e-05, "loss": 5.0826, "step": 3402 }, { "epoch": 3.82, "learning_rate": 2.662112676056338e-05, "loss": 5.1962, "step": 3403 }, { "epoch": 3.82, "learning_rate": 2.6619718309859158e-05, "loss": 5.0358, "step": 3404 }, { "epoch": 3.82, "learning_rate": 2.661830985915493e-05, "loss": 4.9937, "step": 3405 }, { "epoch": 3.82, "learning_rate": 2.6616901408450707e-05, "loss": 5.2552, "step": 3406 }, { "epoch": 3.82, "learning_rate": 2.661549295774648e-05, "loss": 5.2319, "step": 3407 }, { "epoch": 3.82, "learning_rate": 2.6614084507042253e-05, "loss": 5.1588, "step": 3408 }, { "epoch": 3.82, "learning_rate": 2.661267605633803e-05, "loss": 5.2903, "step": 3409 }, { "epoch": 3.82, "learning_rate": 2.6611267605633802e-05, "loss": 5.1715, "step": 3410 }, { "epoch": 3.82, "learning_rate": 2.660985915492958e-05, "loss": 4.9293, "step": 3411 }, { "epoch": 3.83, "learning_rate": 2.660845070422535e-05, "loss": 5.0787, "step": 3412 }, { "epoch": 3.83, "learning_rate": 2.6607042253521128e-05, "loss": 5.2679, "step": 3413 }, { "epoch": 3.83, "learning_rate": 2.66056338028169e-05, "loss": 5.1531, "step": 3414 }, { "epoch": 3.83, "learning_rate": 2.6604225352112674e-05, "loss": 4.9575, "step": 3415 }, { "epoch": 3.83, "learning_rate": 2.660281690140845e-05, "loss": 4.9664, "step": 3416 }, { "epoch": 3.83, "learning_rate": 2.6601408450704226e-05, "loss": 4.9756, "step": 3417 }, { "epoch": 3.83, "learning_rate": 2.6600000000000003e-05, "loss": 4.899, "step": 3418 }, { "epoch": 3.83, "learning_rate": 2.6598591549295776e-05, "loss": 4.917, "step": 3419 }, { "epoch": 3.83, "learning_rate": 2.6597183098591552e-05, "loss": 5.1343, "step": 3420 }, { "epoch": 3.84, "learning_rate": 2.6595774647887325e-05, "loss": 4.8069, "step": 3421 }, { "epoch": 3.84, "learning_rate": 2.65943661971831e-05, "loss": 5.126, "step": 3422 }, { "epoch": 3.84, "learning_rate": 2.6592957746478874e-05, "loss": 4.8097, "step": 3423 }, { "epoch": 3.84, "learning_rate": 2.6591549295774647e-05, "loss": 4.7802, "step": 3424 }, { "epoch": 3.84, "learning_rate": 2.6590140845070424e-05, "loss": 4.4835, "step": 3425 }, { "epoch": 3.84, "learning_rate": 2.6588732394366197e-05, "loss": 4.5163, "step": 3426 }, { "epoch": 3.84, "learning_rate": 2.6587323943661973e-05, "loss": 5.6291, "step": 3427 }, { "epoch": 3.84, "learning_rate": 2.6585915492957746e-05, "loss": 5.4827, "step": 3428 }, { "epoch": 3.84, "learning_rate": 2.6584507042253522e-05, "loss": 5.4117, "step": 3429 }, { "epoch": 3.85, "learning_rate": 2.6583098591549295e-05, "loss": 5.5129, "step": 3430 }, { "epoch": 3.85, "learning_rate": 2.6581690140845068e-05, "loss": 5.3209, "step": 3431 }, { "epoch": 3.85, "learning_rate": 2.6580281690140844e-05, "loss": 5.4073, "step": 3432 }, { "epoch": 3.85, "learning_rate": 2.657887323943662e-05, "loss": 5.3063, "step": 3433 }, { "epoch": 3.85, "learning_rate": 2.6577464788732397e-05, "loss": 5.4481, "step": 3434 }, { "epoch": 3.85, "learning_rate": 2.657605633802817e-05, "loss": 5.3217, "step": 3435 }, { "epoch": 3.85, "learning_rate": 2.6574647887323946e-05, "loss": 5.1328, "step": 3436 }, { "epoch": 3.85, "learning_rate": 2.657323943661972e-05, "loss": 5.0812, "step": 3437 }, { "epoch": 3.85, "learning_rate": 2.6571830985915496e-05, "loss": 5.1995, "step": 3438 }, { "epoch": 3.86, "learning_rate": 2.657042253521127e-05, "loss": 5.3, "step": 3439 }, { "epoch": 3.86, "learning_rate": 2.656901408450704e-05, "loss": 5.3387, "step": 3440 }, { "epoch": 3.86, "learning_rate": 2.6567605633802818e-05, "loss": 5.1001, "step": 3441 }, { "epoch": 3.86, "learning_rate": 2.656619718309859e-05, "loss": 5.1481, "step": 3442 }, { "epoch": 3.86, "learning_rate": 2.6564788732394367e-05, "loss": 5.3107, "step": 3443 }, { "epoch": 3.86, "learning_rate": 2.656338028169014e-05, "loss": 5.3284, "step": 3444 }, { "epoch": 3.86, "learning_rate": 2.6561971830985916e-05, "loss": 5.116, "step": 3445 }, { "epoch": 3.86, "learning_rate": 2.656056338028169e-05, "loss": 5.1851, "step": 3446 }, { "epoch": 3.86, "learning_rate": 2.6559154929577462e-05, "loss": 5.3007, "step": 3447 }, { "epoch": 3.87, "learning_rate": 2.6557746478873242e-05, "loss": 5.1377, "step": 3448 }, { "epoch": 3.87, "learning_rate": 2.6556338028169015e-05, "loss": 5.3108, "step": 3449 }, { "epoch": 3.87, "learning_rate": 2.655492957746479e-05, "loss": 5.0772, "step": 3450 }, { "epoch": 3.87, "learning_rate": 2.6553521126760564e-05, "loss": 5.1976, "step": 3451 }, { "epoch": 3.87, "learning_rate": 2.655211267605634e-05, "loss": 5.1544, "step": 3452 }, { "epoch": 3.87, "learning_rate": 2.6550704225352113e-05, "loss": 5.0453, "step": 3453 }, { "epoch": 3.87, "learning_rate": 2.654929577464789e-05, "loss": 5.0131, "step": 3454 }, { "epoch": 3.87, "learning_rate": 2.6547887323943663e-05, "loss": 5.1386, "step": 3455 }, { "epoch": 3.87, "learning_rate": 2.6546478873239436e-05, "loss": 5.2299, "step": 3456 }, { "epoch": 3.88, "learning_rate": 2.6545070422535212e-05, "loss": 4.9761, "step": 3457 }, { "epoch": 3.88, "learning_rate": 2.6543661971830985e-05, "loss": 5.1266, "step": 3458 }, { "epoch": 3.88, "learning_rate": 2.654225352112676e-05, "loss": 5.1232, "step": 3459 }, { "epoch": 3.88, "learning_rate": 2.6540845070422534e-05, "loss": 5.0761, "step": 3460 }, { "epoch": 3.88, "learning_rate": 2.653943661971831e-05, "loss": 5.016, "step": 3461 }, { "epoch": 3.88, "learning_rate": 2.6538028169014083e-05, "loss": 4.957, "step": 3462 }, { "epoch": 3.88, "learning_rate": 2.653661971830986e-05, "loss": 5.2514, "step": 3463 }, { "epoch": 3.88, "learning_rate": 2.6535211267605636e-05, "loss": 5.1263, "step": 3464 }, { "epoch": 3.88, "learning_rate": 2.653380281690141e-05, "loss": 5.1086, "step": 3465 }, { "epoch": 3.89, "learning_rate": 2.6532394366197185e-05, "loss": 4.9663, "step": 3466 }, { "epoch": 3.89, "learning_rate": 2.653098591549296e-05, "loss": 5.1947, "step": 3467 }, { "epoch": 3.89, "learning_rate": 2.6529577464788735e-05, "loss": 5.0966, "step": 3468 }, { "epoch": 3.89, "learning_rate": 2.6528169014084508e-05, "loss": 4.6818, "step": 3469 }, { "epoch": 3.89, "learning_rate": 2.6526760563380284e-05, "loss": 4.7858, "step": 3470 }, { "epoch": 3.89, "learning_rate": 2.6525352112676057e-05, "loss": 4.8394, "step": 3471 }, { "epoch": 3.89, "learning_rate": 2.652394366197183e-05, "loss": 4.4851, "step": 3472 }, { "epoch": 3.89, "learning_rate": 2.6522535211267606e-05, "loss": 4.8166, "step": 3473 }, { "epoch": 3.89, "learning_rate": 2.652112676056338e-05, "loss": 4.599, "step": 3474 }, { "epoch": 3.9, "learning_rate": 2.6519718309859155e-05, "loss": 4.2225, "step": 3475 }, { "epoch": 3.9, "learning_rate": 2.651830985915493e-05, "loss": 4.1761, "step": 3476 }, { "epoch": 3.9, "learning_rate": 2.6516901408450705e-05, "loss": 5.6583, "step": 3477 }, { "epoch": 3.9, "learning_rate": 2.651549295774648e-05, "loss": 5.4928, "step": 3478 }, { "epoch": 3.9, "learning_rate": 2.6514084507042254e-05, "loss": 5.5418, "step": 3479 }, { "epoch": 3.9, "learning_rate": 2.651267605633803e-05, "loss": 5.4358, "step": 3480 }, { "epoch": 3.9, "learning_rate": 2.6511267605633803e-05, "loss": 5.3967, "step": 3481 }, { "epoch": 3.9, "learning_rate": 2.650985915492958e-05, "loss": 5.1991, "step": 3482 }, { "epoch": 3.9, "learning_rate": 2.6508450704225353e-05, "loss": 5.2027, "step": 3483 }, { "epoch": 3.91, "learning_rate": 2.650704225352113e-05, "loss": 5.4039, "step": 3484 }, { "epoch": 3.91, "learning_rate": 2.6505633802816902e-05, "loss": 5.374, "step": 3485 }, { "epoch": 3.91, "learning_rate": 2.6504225352112678e-05, "loss": 5.2204, "step": 3486 }, { "epoch": 3.91, "learning_rate": 2.650281690140845e-05, "loss": 5.2218, "step": 3487 }, { "epoch": 3.91, "learning_rate": 2.6501408450704224e-05, "loss": 5.1468, "step": 3488 }, { "epoch": 3.91, "learning_rate": 2.65e-05, "loss": 5.298, "step": 3489 }, { "epoch": 3.91, "learning_rate": 2.6498591549295773e-05, "loss": 5.4073, "step": 3490 }, { "epoch": 3.91, "learning_rate": 2.649718309859155e-05, "loss": 5.115, "step": 3491 }, { "epoch": 3.91, "learning_rate": 2.6495774647887323e-05, "loss": 5.2296, "step": 3492 }, { "epoch": 3.92, "learning_rate": 2.6494366197183102e-05, "loss": 5.2828, "step": 3493 }, { "epoch": 3.92, "learning_rate": 2.6492957746478875e-05, "loss": 5.1133, "step": 3494 }, { "epoch": 3.92, "learning_rate": 2.6491549295774648e-05, "loss": 5.386, "step": 3495 }, { "epoch": 3.92, "learning_rate": 2.6490140845070425e-05, "loss": 5.0124, "step": 3496 }, { "epoch": 3.92, "learning_rate": 2.6488732394366197e-05, "loss": 5.1112, "step": 3497 }, { "epoch": 3.92, "learning_rate": 2.6487323943661974e-05, "loss": 5.2227, "step": 3498 }, { "epoch": 3.92, "learning_rate": 2.6485915492957747e-05, "loss": 5.2654, "step": 3499 }, { "epoch": 3.92, "learning_rate": 2.6484507042253523e-05, "loss": 5.2477, "step": 3500 }, { "epoch": 3.92, "learning_rate": 2.6483098591549296e-05, "loss": 5.2017, "step": 3501 }, { "epoch": 3.93, "learning_rate": 2.648169014084507e-05, "loss": 5.1798, "step": 3502 }, { "epoch": 3.93, "learning_rate": 2.6480281690140845e-05, "loss": 5.1344, "step": 3503 }, { "epoch": 3.93, "learning_rate": 2.6478873239436618e-05, "loss": 5.118, "step": 3504 }, { "epoch": 3.93, "learning_rate": 2.6477464788732395e-05, "loss": 5.1493, "step": 3505 }, { "epoch": 3.93, "learning_rate": 2.6476056338028168e-05, "loss": 5.0036, "step": 3506 }, { "epoch": 3.93, "learning_rate": 2.6474647887323944e-05, "loss": 5.1022, "step": 3507 }, { "epoch": 3.93, "learning_rate": 2.647323943661972e-05, "loss": 4.925, "step": 3508 }, { "epoch": 3.93, "learning_rate": 2.6471830985915497e-05, "loss": 4.9776, "step": 3509 }, { "epoch": 3.93, "learning_rate": 2.647042253521127e-05, "loss": 5.0862, "step": 3510 }, { "epoch": 3.94, "learning_rate": 2.6469014084507042e-05, "loss": 5.195, "step": 3511 }, { "epoch": 3.94, "learning_rate": 2.646760563380282e-05, "loss": 5.0817, "step": 3512 }, { "epoch": 3.94, "learning_rate": 2.646619718309859e-05, "loss": 5.0294, "step": 3513 }, { "epoch": 3.94, "learning_rate": 2.6464788732394368e-05, "loss": 4.8848, "step": 3514 }, { "epoch": 3.94, "learning_rate": 2.646338028169014e-05, "loss": 5.1299, "step": 3515 }, { "epoch": 3.94, "learning_rate": 2.6461971830985917e-05, "loss": 4.9979, "step": 3516 }, { "epoch": 3.94, "learning_rate": 2.646056338028169e-05, "loss": 4.9102, "step": 3517 }, { "epoch": 3.94, "learning_rate": 2.6459154929577463e-05, "loss": 5.0511, "step": 3518 }, { "epoch": 3.95, "learning_rate": 2.645774647887324e-05, "loss": 4.9384, "step": 3519 }, { "epoch": 3.95, "learning_rate": 2.6456338028169012e-05, "loss": 4.7504, "step": 3520 }, { "epoch": 3.95, "learning_rate": 2.645492957746479e-05, "loss": 4.7952, "step": 3521 }, { "epoch": 3.95, "learning_rate": 2.6453521126760562e-05, "loss": 4.8398, "step": 3522 }, { "epoch": 3.95, "learning_rate": 2.645211267605634e-05, "loss": 4.6847, "step": 3523 }, { "epoch": 3.95, "learning_rate": 2.6450704225352114e-05, "loss": 4.3903, "step": 3524 }, { "epoch": 3.95, "learning_rate": 2.644929577464789e-05, "loss": 4.2589, "step": 3525 }, { "epoch": 3.95, "learning_rate": 2.6447887323943664e-05, "loss": 4.3444, "step": 3526 }, { "epoch": 3.95, "learning_rate": 2.6446478873239437e-05, "loss": 5.4966, "step": 3527 }, { "epoch": 3.96, "learning_rate": 2.6445070422535213e-05, "loss": 5.333, "step": 3528 }, { "epoch": 3.96, "learning_rate": 2.6443661971830986e-05, "loss": 5.3665, "step": 3529 }, { "epoch": 3.96, "learning_rate": 2.6442253521126762e-05, "loss": 5.4107, "step": 3530 }, { "epoch": 3.96, "learning_rate": 2.6440845070422535e-05, "loss": 5.2227, "step": 3531 }, { "epoch": 3.96, "learning_rate": 2.643943661971831e-05, "loss": 5.2931, "step": 3532 }, { "epoch": 3.96, "learning_rate": 2.6438028169014084e-05, "loss": 5.2368, "step": 3533 }, { "epoch": 3.96, "learning_rate": 2.6436619718309857e-05, "loss": 5.2856, "step": 3534 }, { "epoch": 3.96, "learning_rate": 2.6435211267605634e-05, "loss": 5.0643, "step": 3535 }, { "epoch": 3.96, "learning_rate": 2.6433802816901407e-05, "loss": 5.1928, "step": 3536 }, { "epoch": 3.97, "learning_rate": 2.6432394366197183e-05, "loss": 5.19, "step": 3537 }, { "epoch": 3.97, "learning_rate": 2.643098591549296e-05, "loss": 5.1892, "step": 3538 }, { "epoch": 3.97, "learning_rate": 2.6429577464788736e-05, "loss": 5.1551, "step": 3539 }, { "epoch": 3.97, "learning_rate": 2.642816901408451e-05, "loss": 5.1539, "step": 3540 }, { "epoch": 3.97, "learning_rate": 2.6426760563380285e-05, "loss": 5.1561, "step": 3541 }, { "epoch": 3.97, "learning_rate": 2.6425352112676058e-05, "loss": 5.2503, "step": 3542 }, { "epoch": 3.97, "learning_rate": 2.642394366197183e-05, "loss": 5.286, "step": 3543 }, { "epoch": 3.97, "learning_rate": 2.6422535211267607e-05, "loss": 5.2715, "step": 3544 }, { "epoch": 3.97, "learning_rate": 2.642112676056338e-05, "loss": 5.1241, "step": 3545 }, { "epoch": 3.98, "learning_rate": 2.6419718309859156e-05, "loss": 5.1281, "step": 3546 }, { "epoch": 3.98, "learning_rate": 2.641830985915493e-05, "loss": 5.1372, "step": 3547 }, { "epoch": 3.98, "learning_rate": 2.6416901408450706e-05, "loss": 5.0842, "step": 3548 }, { "epoch": 3.98, "learning_rate": 2.641549295774648e-05, "loss": 5.0802, "step": 3549 }, { "epoch": 3.98, "learning_rate": 2.641408450704225e-05, "loss": 5.0442, "step": 3550 }, { "epoch": 3.98, "learning_rate": 2.6412676056338028e-05, "loss": 5.3268, "step": 3551 }, { "epoch": 3.98, "learning_rate": 2.64112676056338e-05, "loss": 5.2392, "step": 3552 }, { "epoch": 3.98, "learning_rate": 2.6409859154929577e-05, "loss": 5.0833, "step": 3553 }, { "epoch": 3.98, "learning_rate": 2.6408450704225354e-05, "loss": 5.0618, "step": 3554 }, { "epoch": 3.99, "learning_rate": 2.640704225352113e-05, "loss": 5.2824, "step": 3555 }, { "epoch": 3.99, "learning_rate": 2.6405633802816903e-05, "loss": 5.0839, "step": 3556 }, { "epoch": 3.99, "learning_rate": 2.640422535211268e-05, "loss": 5.0539, "step": 3557 }, { "epoch": 3.99, "learning_rate": 2.6402816901408452e-05, "loss": 5.0078, "step": 3558 }, { "epoch": 3.99, "learning_rate": 2.6401408450704225e-05, "loss": 5.1238, "step": 3559 }, { "epoch": 3.99, "learning_rate": 2.64e-05, "loss": 5.1942, "step": 3560 }, { "epoch": 3.99, "learning_rate": 2.6398591549295774e-05, "loss": 5.0647, "step": 3561 }, { "epoch": 3.99, "learning_rate": 2.639718309859155e-05, "loss": 5.0878, "step": 3562 }, { "epoch": 3.99, "learning_rate": 2.6395774647887324e-05, "loss": 4.6735, "step": 3563 }, { "epoch": 4.0, "learning_rate": 2.63943661971831e-05, "loss": 4.8315, "step": 3564 }, { "epoch": 4.0, "learning_rate": 2.6392957746478873e-05, "loss": 4.6357, "step": 3565 }, { "epoch": 4.0, "learning_rate": 2.6391549295774646e-05, "loss": 4.5227, "step": 3566 }, { "epoch": 4.0, "learning_rate": 2.6390140845070422e-05, "loss": 4.4827, "step": 3567 }, { "epoch": 4.0, "learning_rate": 2.6388732394366195e-05, "loss": 4.3361, "step": 3568 }, { "epoch": 4.0, "learning_rate": 2.6387323943661975e-05, "loss": 5.5523, "step": 3569 }, { "epoch": 4.0, "learning_rate": 2.6385915492957748e-05, "loss": 5.4169, "step": 3570 }, { "epoch": 4.0, "learning_rate": 2.6384507042253524e-05, "loss": 5.296, "step": 3571 }, { "epoch": 4.0, "learning_rate": 2.6383098591549297e-05, "loss": 5.3073, "step": 3572 }, { "epoch": 4.01, "learning_rate": 2.6381690140845073e-05, "loss": 5.2211, "step": 3573 }, { "epoch": 4.01, "learning_rate": 2.6380281690140846e-05, "loss": 5.3032, "step": 3574 }, { "epoch": 4.01, "learning_rate": 2.637887323943662e-05, "loss": 5.1939, "step": 3575 }, { "epoch": 4.01, "learning_rate": 2.6377464788732396e-05, "loss": 5.1493, "step": 3576 }, { "epoch": 4.01, "learning_rate": 2.637605633802817e-05, "loss": 5.1907, "step": 3577 }, { "epoch": 4.01, "learning_rate": 2.6374647887323945e-05, "loss": 5.0118, "step": 3578 }, { "epoch": 4.01, "learning_rate": 2.6373239436619718e-05, "loss": 4.9748, "step": 3579 }, { "epoch": 4.01, "learning_rate": 2.6371830985915494e-05, "loss": 4.9889, "step": 3580 }, { "epoch": 4.01, "learning_rate": 2.6370422535211267e-05, "loss": 4.9844, "step": 3581 }, { "epoch": 4.02, "learning_rate": 2.636901408450704e-05, "loss": 4.9817, "step": 3582 }, { "epoch": 4.02, "learning_rate": 2.6367605633802816e-05, "loss": 5.0172, "step": 3583 }, { "epoch": 4.02, "learning_rate": 2.6366197183098593e-05, "loss": 5.203, "step": 3584 }, { "epoch": 4.02, "learning_rate": 2.636478873239437e-05, "loss": 5.1028, "step": 3585 }, { "epoch": 4.02, "learning_rate": 2.6363380281690142e-05, "loss": 5.1391, "step": 3586 }, { "epoch": 4.02, "learning_rate": 2.6361971830985918e-05, "loss": 5.0435, "step": 3587 }, { "epoch": 4.02, "learning_rate": 2.636056338028169e-05, "loss": 4.953, "step": 3588 }, { "epoch": 4.02, "learning_rate": 2.6359154929577468e-05, "loss": 4.9169, "step": 3589 }, { "epoch": 4.02, "learning_rate": 2.635774647887324e-05, "loss": 5.053, "step": 3590 }, { "epoch": 4.03, "learning_rate": 2.6356338028169013e-05, "loss": 5.1838, "step": 3591 }, { "epoch": 4.03, "learning_rate": 2.635492957746479e-05, "loss": 4.9412, "step": 3592 }, { "epoch": 4.03, "learning_rate": 2.6353521126760563e-05, "loss": 4.9583, "step": 3593 }, { "epoch": 4.03, "learning_rate": 2.635211267605634e-05, "loss": 4.9084, "step": 3594 }, { "epoch": 4.03, "learning_rate": 2.6350704225352112e-05, "loss": 5.1063, "step": 3595 }, { "epoch": 4.03, "learning_rate": 2.6349295774647888e-05, "loss": 4.7811, "step": 3596 }, { "epoch": 4.03, "learning_rate": 2.634788732394366e-05, "loss": 5.0915, "step": 3597 }, { "epoch": 4.03, "learning_rate": 2.6346478873239434e-05, "loss": 5.0695, "step": 3598 }, { "epoch": 4.03, "learning_rate": 2.6345070422535214e-05, "loss": 4.903, "step": 3599 }, { "epoch": 4.04, "learning_rate": 2.6343661971830987e-05, "loss": 5.0135, "step": 3600 }, { "epoch": 4.04, "learning_rate": 2.6342253521126763e-05, "loss": 4.7338, "step": 3601 }, { "epoch": 4.04, "learning_rate": 2.6340845070422536e-05, "loss": 4.8303, "step": 3602 }, { "epoch": 4.04, "learning_rate": 2.6339436619718312e-05, "loss": 5.0318, "step": 3603 }, { "epoch": 4.04, "learning_rate": 2.6338028169014085e-05, "loss": 4.9617, "step": 3604 }, { "epoch": 4.04, "learning_rate": 2.6336619718309862e-05, "loss": 4.8744, "step": 3605 }, { "epoch": 4.04, "learning_rate": 2.6335211267605635e-05, "loss": 5.1239, "step": 3606 }, { "epoch": 4.04, "learning_rate": 2.6333802816901408e-05, "loss": 4.8226, "step": 3607 }, { "epoch": 4.04, "learning_rate": 2.6332394366197184e-05, "loss": 4.8314, "step": 3608 }, { "epoch": 4.05, "learning_rate": 2.6330985915492957e-05, "loss": 4.7503, "step": 3609 }, { "epoch": 4.05, "learning_rate": 2.6329577464788733e-05, "loss": 4.9581, "step": 3610 }, { "epoch": 4.05, "learning_rate": 2.6328169014084506e-05, "loss": 4.7251, "step": 3611 }, { "epoch": 4.05, "learning_rate": 2.6326760563380282e-05, "loss": 4.7167, "step": 3612 }, { "epoch": 4.05, "learning_rate": 2.6325352112676055e-05, "loss": 4.6528, "step": 3613 }, { "epoch": 4.05, "learning_rate": 2.6323943661971832e-05, "loss": 4.6613, "step": 3614 }, { "epoch": 4.05, "learning_rate": 2.6322535211267608e-05, "loss": 4.4081, "step": 3615 }, { "epoch": 4.05, "learning_rate": 2.632112676056338e-05, "loss": 4.3929, "step": 3616 }, { "epoch": 4.05, "learning_rate": 2.6319718309859157e-05, "loss": 4.5607, "step": 3617 }, { "epoch": 4.06, "learning_rate": 2.631830985915493e-05, "loss": 3.8184, "step": 3618 }, { "epoch": 4.06, "learning_rate": 2.6316901408450707e-05, "loss": 5.4135, "step": 3619 }, { "epoch": 4.06, "learning_rate": 2.631549295774648e-05, "loss": 5.2552, "step": 3620 }, { "epoch": 4.06, "learning_rate": 2.6314084507042253e-05, "loss": 5.2366, "step": 3621 }, { "epoch": 4.06, "learning_rate": 2.631267605633803e-05, "loss": 5.1782, "step": 3622 }, { "epoch": 4.06, "learning_rate": 2.6311267605633802e-05, "loss": 5.181, "step": 3623 }, { "epoch": 4.06, "learning_rate": 2.6309859154929578e-05, "loss": 5.2778, "step": 3624 }, { "epoch": 4.06, "learning_rate": 2.630845070422535e-05, "loss": 5.2159, "step": 3625 }, { "epoch": 4.07, "learning_rate": 2.6307042253521127e-05, "loss": 5.1524, "step": 3626 }, { "epoch": 4.07, "learning_rate": 2.63056338028169e-05, "loss": 5.0992, "step": 3627 }, { "epoch": 4.07, "learning_rate": 2.6304225352112677e-05, "loss": 5.1742, "step": 3628 }, { "epoch": 4.07, "learning_rate": 2.6302816901408453e-05, "loss": 4.9915, "step": 3629 }, { "epoch": 4.07, "learning_rate": 2.6301408450704226e-05, "loss": 5.0124, "step": 3630 }, { "epoch": 4.07, "learning_rate": 2.6300000000000002e-05, "loss": 5.07, "step": 3631 }, { "epoch": 4.07, "learning_rate": 2.6298591549295775e-05, "loss": 5.2045, "step": 3632 }, { "epoch": 4.07, "learning_rate": 2.629718309859155e-05, "loss": 5.1801, "step": 3633 }, { "epoch": 4.07, "learning_rate": 2.6295774647887325e-05, "loss": 5.1384, "step": 3634 }, { "epoch": 4.08, "learning_rate": 2.62943661971831e-05, "loss": 4.9736, "step": 3635 }, { "epoch": 4.08, "learning_rate": 2.6292957746478874e-05, "loss": 4.8266, "step": 3636 }, { "epoch": 4.08, "learning_rate": 2.6291549295774647e-05, "loss": 4.8389, "step": 3637 }, { "epoch": 4.08, "learning_rate": 2.6290140845070423e-05, "loss": 5.015, "step": 3638 }, { "epoch": 4.08, "learning_rate": 2.6288732394366196e-05, "loss": 5.0688, "step": 3639 }, { "epoch": 4.08, "learning_rate": 2.6287323943661972e-05, "loss": 4.9097, "step": 3640 }, { "epoch": 4.08, "learning_rate": 2.6285915492957745e-05, "loss": 4.7833, "step": 3641 }, { "epoch": 4.08, "learning_rate": 2.628450704225352e-05, "loss": 4.9314, "step": 3642 }, { "epoch": 4.08, "learning_rate": 2.6283098591549295e-05, "loss": 4.945, "step": 3643 }, { "epoch": 4.09, "learning_rate": 2.6281690140845074e-05, "loss": 4.9761, "step": 3644 }, { "epoch": 4.09, "learning_rate": 2.6280281690140847e-05, "loss": 4.9023, "step": 3645 }, { "epoch": 4.09, "learning_rate": 2.627887323943662e-05, "loss": 4.9523, "step": 3646 }, { "epoch": 4.09, "learning_rate": 2.6277464788732397e-05, "loss": 5.0228, "step": 3647 }, { "epoch": 4.09, "learning_rate": 2.627605633802817e-05, "loss": 4.8285, "step": 3648 }, { "epoch": 4.09, "learning_rate": 2.6274647887323946e-05, "loss": 4.8482, "step": 3649 }, { "epoch": 4.09, "learning_rate": 2.627323943661972e-05, "loss": 5.0135, "step": 3650 }, { "epoch": 4.09, "learning_rate": 2.6271830985915495e-05, "loss": 4.9594, "step": 3651 }, { "epoch": 4.09, "learning_rate": 2.6270422535211268e-05, "loss": 4.8401, "step": 3652 }, { "epoch": 4.1, "learning_rate": 2.626901408450704e-05, "loss": 4.7935, "step": 3653 }, { "epoch": 4.1, "learning_rate": 2.6267605633802817e-05, "loss": 4.9934, "step": 3654 }, { "epoch": 4.1, "learning_rate": 2.626619718309859e-05, "loss": 4.8101, "step": 3655 }, { "epoch": 4.1, "learning_rate": 2.6264788732394367e-05, "loss": 4.9119, "step": 3656 }, { "epoch": 4.1, "learning_rate": 2.626338028169014e-05, "loss": 4.9023, "step": 3657 }, { "epoch": 4.1, "learning_rate": 2.6261971830985916e-05, "loss": 4.906, "step": 3658 }, { "epoch": 4.1, "learning_rate": 2.6260563380281692e-05, "loss": 4.6904, "step": 3659 }, { "epoch": 4.1, "learning_rate": 2.625915492957747e-05, "loss": 4.757, "step": 3660 }, { "epoch": 4.1, "learning_rate": 2.625774647887324e-05, "loss": 4.6002, "step": 3661 }, { "epoch": 4.11, "learning_rate": 2.6256338028169014e-05, "loss": 4.6876, "step": 3662 }, { "epoch": 4.11, "learning_rate": 2.625492957746479e-05, "loss": 4.655, "step": 3663 }, { "epoch": 4.11, "learning_rate": 2.6253521126760564e-05, "loss": 4.2724, "step": 3664 }, { "epoch": 4.11, "learning_rate": 2.625211267605634e-05, "loss": 4.2814, "step": 3665 }, { "epoch": 4.11, "learning_rate": 2.6250704225352113e-05, "loss": 4.1844, "step": 3666 }, { "epoch": 4.11, "learning_rate": 2.624929577464789e-05, "loss": 4.4725, "step": 3667 }, { "epoch": 4.11, "learning_rate": 2.6247887323943662e-05, "loss": 3.9039, "step": 3668 }, { "epoch": 4.11, "learning_rate": 2.6246478873239435e-05, "loss": 5.3863, "step": 3669 }, { "epoch": 4.11, "learning_rate": 2.624507042253521e-05, "loss": 5.3843, "step": 3670 }, { "epoch": 4.12, "learning_rate": 2.6243661971830984e-05, "loss": 5.3282, "step": 3671 }, { "epoch": 4.12, "learning_rate": 2.624225352112676e-05, "loss": 5.2749, "step": 3672 }, { "epoch": 4.12, "learning_rate": 2.6240845070422534e-05, "loss": 5.0988, "step": 3673 }, { "epoch": 4.12, "learning_rate": 2.623943661971831e-05, "loss": 5.0581, "step": 3674 }, { "epoch": 4.12, "learning_rate": 2.6238028169014086e-05, "loss": 5.2945, "step": 3675 }, { "epoch": 4.12, "learning_rate": 2.6236619718309863e-05, "loss": 5.2076, "step": 3676 }, { "epoch": 4.12, "learning_rate": 2.6235211267605636e-05, "loss": 5.0455, "step": 3677 }, { "epoch": 4.12, "learning_rate": 2.623380281690141e-05, "loss": 4.9307, "step": 3678 }, { "epoch": 4.12, "learning_rate": 2.6232394366197185e-05, "loss": 5.0641, "step": 3679 }, { "epoch": 4.13, "learning_rate": 2.6230985915492958e-05, "loss": 4.9005, "step": 3680 }, { "epoch": 4.13, "learning_rate": 2.6229577464788734e-05, "loss": 5.078, "step": 3681 }, { "epoch": 4.13, "learning_rate": 2.6228169014084507e-05, "loss": 4.9082, "step": 3682 }, { "epoch": 4.13, "learning_rate": 2.6226760563380283e-05, "loss": 5.025, "step": 3683 }, { "epoch": 4.13, "learning_rate": 2.6225352112676056e-05, "loss": 4.9392, "step": 3684 }, { "epoch": 4.13, "learning_rate": 2.622394366197183e-05, "loss": 5.0353, "step": 3685 }, { "epoch": 4.13, "learning_rate": 2.6222535211267606e-05, "loss": 5.0877, "step": 3686 }, { "epoch": 4.13, "learning_rate": 2.622112676056338e-05, "loss": 4.9942, "step": 3687 }, { "epoch": 4.13, "learning_rate": 2.6219718309859155e-05, "loss": 4.887, "step": 3688 }, { "epoch": 4.14, "learning_rate": 2.6218309859154928e-05, "loss": 4.922, "step": 3689 }, { "epoch": 4.14, "learning_rate": 2.6216901408450708e-05, "loss": 4.8702, "step": 3690 }, { "epoch": 4.14, "learning_rate": 2.621549295774648e-05, "loss": 5.068, "step": 3691 }, { "epoch": 4.14, "learning_rate": 2.6214084507042257e-05, "loss": 4.7891, "step": 3692 }, { "epoch": 4.14, "learning_rate": 2.621267605633803e-05, "loss": 4.8443, "step": 3693 }, { "epoch": 4.14, "learning_rate": 2.6211267605633803e-05, "loss": 4.8827, "step": 3694 }, { "epoch": 4.14, "learning_rate": 2.620985915492958e-05, "loss": 4.8307, "step": 3695 }, { "epoch": 4.14, "learning_rate": 2.6208450704225352e-05, "loss": 4.7892, "step": 3696 }, { "epoch": 4.14, "learning_rate": 2.620704225352113e-05, "loss": 4.9136, "step": 3697 }, { "epoch": 4.15, "learning_rate": 2.62056338028169e-05, "loss": 5.1553, "step": 3698 }, { "epoch": 4.15, "learning_rate": 2.6204225352112678e-05, "loss": 4.9447, "step": 3699 }, { "epoch": 4.15, "learning_rate": 2.620281690140845e-05, "loss": 4.9537, "step": 3700 }, { "epoch": 4.15, "learning_rate": 2.6201408450704224e-05, "loss": 4.9477, "step": 3701 }, { "epoch": 4.15, "learning_rate": 2.62e-05, "loss": 4.8941, "step": 3702 }, { "epoch": 4.15, "learning_rate": 2.6198591549295773e-05, "loss": 4.8907, "step": 3703 }, { "epoch": 4.15, "learning_rate": 2.619718309859155e-05, "loss": 4.7911, "step": 3704 }, { "epoch": 4.15, "learning_rate": 2.6195774647887325e-05, "loss": 4.9353, "step": 3705 }, { "epoch": 4.15, "learning_rate": 2.6194366197183102e-05, "loss": 4.8492, "step": 3706 }, { "epoch": 4.16, "learning_rate": 2.6192957746478875e-05, "loss": 4.9132, "step": 3707 }, { "epoch": 4.16, "learning_rate": 2.619154929577465e-05, "loss": 4.8633, "step": 3708 }, { "epoch": 4.16, "learning_rate": 2.6190140845070424e-05, "loss": 4.616, "step": 3709 }, { "epoch": 4.16, "learning_rate": 2.6188732394366197e-05, "loss": 4.7864, "step": 3710 }, { "epoch": 4.16, "learning_rate": 2.6187323943661973e-05, "loss": 4.6641, "step": 3711 }, { "epoch": 4.16, "learning_rate": 2.6185915492957746e-05, "loss": 4.9506, "step": 3712 }, { "epoch": 4.16, "learning_rate": 2.6184507042253523e-05, "loss": 4.6483, "step": 3713 }, { "epoch": 4.16, "learning_rate": 2.6183098591549296e-05, "loss": 4.4026, "step": 3714 }, { "epoch": 4.16, "learning_rate": 2.6181690140845072e-05, "loss": 4.3311, "step": 3715 }, { "epoch": 4.17, "learning_rate": 2.6180281690140845e-05, "loss": 3.9966, "step": 3716 }, { "epoch": 4.17, "learning_rate": 2.6178873239436618e-05, "loss": 4.1032, "step": 3717 }, { "epoch": 4.17, "learning_rate": 2.6177464788732394e-05, "loss": 4.0153, "step": 3718 }, { "epoch": 4.17, "learning_rate": 2.6176056338028167e-05, "loss": 5.4733, "step": 3719 }, { "epoch": 4.17, "learning_rate": 2.6174647887323947e-05, "loss": 5.4063, "step": 3720 }, { "epoch": 4.17, "learning_rate": 2.617323943661972e-05, "loss": 5.2603, "step": 3721 }, { "epoch": 4.17, "learning_rate": 2.6171830985915496e-05, "loss": 5.4486, "step": 3722 }, { "epoch": 4.17, "learning_rate": 2.617042253521127e-05, "loss": 5.3073, "step": 3723 }, { "epoch": 4.17, "learning_rate": 2.6169014084507042e-05, "loss": 5.0946, "step": 3724 }, { "epoch": 4.18, "learning_rate": 2.6167605633802818e-05, "loss": 5.068, "step": 3725 }, { "epoch": 4.18, "learning_rate": 2.616619718309859e-05, "loss": 5.0753, "step": 3726 }, { "epoch": 4.18, "learning_rate": 2.6164788732394368e-05, "loss": 4.9769, "step": 3727 }, { "epoch": 4.18, "learning_rate": 2.616338028169014e-05, "loss": 5.0928, "step": 3728 }, { "epoch": 4.18, "learning_rate": 2.6161971830985917e-05, "loss": 4.9964, "step": 3729 }, { "epoch": 4.18, "learning_rate": 2.616056338028169e-05, "loss": 4.9622, "step": 3730 }, { "epoch": 4.18, "learning_rate": 2.6159154929577466e-05, "loss": 4.9714, "step": 3731 }, { "epoch": 4.18, "learning_rate": 2.615774647887324e-05, "loss": 5.0355, "step": 3732 }, { "epoch": 4.18, "learning_rate": 2.6156338028169012e-05, "loss": 4.9695, "step": 3733 }, { "epoch": 4.19, "learning_rate": 2.6154929577464788e-05, "loss": 4.9185, "step": 3734 }, { "epoch": 4.19, "learning_rate": 2.6153521126760565e-05, "loss": 4.8687, "step": 3735 }, { "epoch": 4.19, "learning_rate": 2.615211267605634e-05, "loss": 4.9234, "step": 3736 }, { "epoch": 4.19, "learning_rate": 2.6150704225352114e-05, "loss": 5.0301, "step": 3737 }, { "epoch": 4.19, "learning_rate": 2.614929577464789e-05, "loss": 4.8353, "step": 3738 }, { "epoch": 4.19, "learning_rate": 2.6147887323943663e-05, "loss": 4.9818, "step": 3739 }, { "epoch": 4.19, "learning_rate": 2.6146478873239436e-05, "loss": 4.8368, "step": 3740 }, { "epoch": 4.19, "learning_rate": 2.6145070422535212e-05, "loss": 4.9275, "step": 3741 }, { "epoch": 4.2, "learning_rate": 2.6143661971830985e-05, "loss": 4.9293, "step": 3742 }, { "epoch": 4.2, "learning_rate": 2.6142253521126762e-05, "loss": 5.0276, "step": 3743 }, { "epoch": 4.2, "learning_rate": 2.6140845070422535e-05, "loss": 5.1105, "step": 3744 }, { "epoch": 4.2, "learning_rate": 2.613943661971831e-05, "loss": 4.7197, "step": 3745 }, { "epoch": 4.2, "learning_rate": 2.6138028169014084e-05, "loss": 4.74, "step": 3746 }, { "epoch": 4.2, "learning_rate": 2.613661971830986e-05, "loss": 4.8619, "step": 3747 }, { "epoch": 4.2, "learning_rate": 2.6135211267605633e-05, "loss": 4.9968, "step": 3748 }, { "epoch": 4.2, "learning_rate": 2.6133802816901406e-05, "loss": 4.6999, "step": 3749 }, { "epoch": 4.2, "learning_rate": 2.6132394366197186e-05, "loss": 5.0018, "step": 3750 }, { "epoch": 4.21, "learning_rate": 2.613098591549296e-05, "loss": 4.9208, "step": 3751 }, { "epoch": 4.21, "learning_rate": 2.6129577464788735e-05, "loss": 5.005, "step": 3752 }, { "epoch": 4.21, "learning_rate": 2.6128169014084508e-05, "loss": 4.9064, "step": 3753 }, { "epoch": 4.21, "learning_rate": 2.6126760563380284e-05, "loss": 4.8292, "step": 3754 }, { "epoch": 4.21, "learning_rate": 2.6125352112676057e-05, "loss": 4.7737, "step": 3755 }, { "epoch": 4.21, "learning_rate": 2.612394366197183e-05, "loss": 4.8782, "step": 3756 }, { "epoch": 4.21, "learning_rate": 2.6122535211267607e-05, "loss": 4.8092, "step": 3757 }, { "epoch": 4.21, "learning_rate": 2.612112676056338e-05, "loss": 4.8289, "step": 3758 }, { "epoch": 4.21, "learning_rate": 2.6119718309859156e-05, "loss": 4.6082, "step": 3759 }, { "epoch": 4.22, "learning_rate": 2.611830985915493e-05, "loss": 4.793, "step": 3760 }, { "epoch": 4.22, "learning_rate": 2.6116901408450705e-05, "loss": 4.5903, "step": 3761 }, { "epoch": 4.22, "learning_rate": 2.6115492957746478e-05, "loss": 4.6617, "step": 3762 }, { "epoch": 4.22, "learning_rate": 2.6114084507042254e-05, "loss": 4.3848, "step": 3763 }, { "epoch": 4.22, "learning_rate": 2.6112676056338027e-05, "loss": 4.5219, "step": 3764 }, { "epoch": 4.22, "learning_rate": 2.6111267605633804e-05, "loss": 4.3516, "step": 3765 }, { "epoch": 4.22, "learning_rate": 2.610985915492958e-05, "loss": 4.1298, "step": 3766 }, { "epoch": 4.22, "learning_rate": 2.6108450704225353e-05, "loss": 4.2988, "step": 3767 }, { "epoch": 4.22, "learning_rate": 2.610704225352113e-05, "loss": 3.7578, "step": 3768 }, { "epoch": 4.23, "learning_rate": 2.6105633802816902e-05, "loss": 5.6225, "step": 3769 }, { "epoch": 4.23, "learning_rate": 2.610422535211268e-05, "loss": 5.4369, "step": 3770 }, { "epoch": 4.23, "learning_rate": 2.610281690140845e-05, "loss": 5.365, "step": 3771 }, { "epoch": 4.23, "learning_rate": 2.6101408450704224e-05, "loss": 5.26, "step": 3772 }, { "epoch": 4.23, "learning_rate": 2.61e-05, "loss": 5.1827, "step": 3773 }, { "epoch": 4.23, "learning_rate": 2.6098591549295774e-05, "loss": 5.0331, "step": 3774 }, { "epoch": 4.23, "learning_rate": 2.609718309859155e-05, "loss": 5.2548, "step": 3775 }, { "epoch": 4.23, "learning_rate": 2.6095774647887323e-05, "loss": 5.2086, "step": 3776 }, { "epoch": 4.23, "learning_rate": 2.60943661971831e-05, "loss": 4.9003, "step": 3777 }, { "epoch": 4.24, "learning_rate": 2.6092957746478872e-05, "loss": 5.2088, "step": 3778 }, { "epoch": 4.24, "learning_rate": 2.609154929577465e-05, "loss": 4.9277, "step": 3779 }, { "epoch": 4.24, "learning_rate": 2.6090140845070425e-05, "loss": 4.9921, "step": 3780 }, { "epoch": 4.24, "learning_rate": 2.6088732394366198e-05, "loss": 5.0888, "step": 3781 }, { "epoch": 4.24, "learning_rate": 2.6087323943661974e-05, "loss": 4.8871, "step": 3782 }, { "epoch": 4.24, "learning_rate": 2.6085915492957747e-05, "loss": 4.8161, "step": 3783 }, { "epoch": 4.24, "learning_rate": 2.6084507042253524e-05, "loss": 4.8269, "step": 3784 }, { "epoch": 4.24, "learning_rate": 2.6083098591549296e-05, "loss": 4.9843, "step": 3785 }, { "epoch": 4.24, "learning_rate": 2.6081690140845073e-05, "loss": 4.9217, "step": 3786 }, { "epoch": 4.25, "learning_rate": 2.6080281690140846e-05, "loss": 4.9473, "step": 3787 }, { "epoch": 4.25, "learning_rate": 2.607887323943662e-05, "loss": 5.1155, "step": 3788 }, { "epoch": 4.25, "learning_rate": 2.6077464788732395e-05, "loss": 5.1249, "step": 3789 }, { "epoch": 4.25, "learning_rate": 2.6076056338028168e-05, "loss": 4.7676, "step": 3790 }, { "epoch": 4.25, "learning_rate": 2.6074647887323944e-05, "loss": 5.1001, "step": 3791 }, { "epoch": 4.25, "learning_rate": 2.6073239436619717e-05, "loss": 4.9167, "step": 3792 }, { "epoch": 4.25, "learning_rate": 2.6071830985915494e-05, "loss": 4.7461, "step": 3793 }, { "epoch": 4.25, "learning_rate": 2.6070422535211267e-05, "loss": 4.7887, "step": 3794 }, { "epoch": 4.25, "learning_rate": 2.6069014084507043e-05, "loss": 4.898, "step": 3795 }, { "epoch": 4.26, "learning_rate": 2.606760563380282e-05, "loss": 5.0129, "step": 3796 }, { "epoch": 4.26, "learning_rate": 2.6066197183098592e-05, "loss": 4.8591, "step": 3797 }, { "epoch": 4.26, "learning_rate": 2.606478873239437e-05, "loss": 4.7191, "step": 3798 }, { "epoch": 4.26, "learning_rate": 2.606338028169014e-05, "loss": 4.8395, "step": 3799 }, { "epoch": 4.26, "learning_rate": 2.6061971830985918e-05, "loss": 4.7358, "step": 3800 }, { "epoch": 4.26, "learning_rate": 2.606056338028169e-05, "loss": 4.8128, "step": 3801 }, { "epoch": 4.26, "learning_rate": 2.6059154929577467e-05, "loss": 4.9085, "step": 3802 }, { "epoch": 4.26, "learning_rate": 2.605774647887324e-05, "loss": 4.745, "step": 3803 }, { "epoch": 4.26, "learning_rate": 2.6056338028169013e-05, "loss": 4.8462, "step": 3804 }, { "epoch": 4.27, "learning_rate": 2.605492957746479e-05, "loss": 4.9604, "step": 3805 }, { "epoch": 4.27, "learning_rate": 2.6053521126760562e-05, "loss": 4.6451, "step": 3806 }, { "epoch": 4.27, "learning_rate": 2.605211267605634e-05, "loss": 4.6447, "step": 3807 }, { "epoch": 4.27, "learning_rate": 2.605070422535211e-05, "loss": 4.8505, "step": 3808 }, { "epoch": 4.27, "learning_rate": 2.6049295774647888e-05, "loss": 4.473, "step": 3809 }, { "epoch": 4.27, "learning_rate": 2.604788732394366e-05, "loss": 4.9291, "step": 3810 }, { "epoch": 4.27, "learning_rate": 2.604647887323944e-05, "loss": 4.3975, "step": 3811 }, { "epoch": 4.27, "learning_rate": 2.6045070422535213e-05, "loss": 4.6694, "step": 3812 }, { "epoch": 4.27, "learning_rate": 2.6043661971830986e-05, "loss": 4.5661, "step": 3813 }, { "epoch": 4.28, "learning_rate": 2.6042253521126763e-05, "loss": 4.2724, "step": 3814 }, { "epoch": 4.28, "learning_rate": 2.6040845070422536e-05, "loss": 4.252, "step": 3815 }, { "epoch": 4.28, "learning_rate": 2.6039436619718312e-05, "loss": 4.4218, "step": 3816 }, { "epoch": 4.28, "learning_rate": 2.6038028169014085e-05, "loss": 4.0757, "step": 3817 }, { "epoch": 4.28, "learning_rate": 2.603661971830986e-05, "loss": 3.9109, "step": 3818 }, { "epoch": 4.28, "learning_rate": 2.6035211267605634e-05, "loss": 5.4193, "step": 3819 }, { "epoch": 4.28, "learning_rate": 2.6033802816901407e-05, "loss": 5.3448, "step": 3820 }, { "epoch": 4.28, "learning_rate": 2.6032394366197183e-05, "loss": 5.154, "step": 3821 }, { "epoch": 4.28, "learning_rate": 2.6030985915492956e-05, "loss": 5.1777, "step": 3822 }, { "epoch": 4.29, "learning_rate": 2.6029577464788733e-05, "loss": 5.1877, "step": 3823 }, { "epoch": 4.29, "learning_rate": 2.6028169014084506e-05, "loss": 4.9054, "step": 3824 }, { "epoch": 4.29, "learning_rate": 2.6026760563380282e-05, "loss": 4.8409, "step": 3825 }, { "epoch": 4.29, "learning_rate": 2.602535211267606e-05, "loss": 5.0699, "step": 3826 }, { "epoch": 4.29, "learning_rate": 2.602394366197183e-05, "loss": 4.8195, "step": 3827 }, { "epoch": 4.29, "learning_rate": 2.6022535211267608e-05, "loss": 4.8969, "step": 3828 }, { "epoch": 4.29, "learning_rate": 2.602112676056338e-05, "loss": 4.9797, "step": 3829 }, { "epoch": 4.29, "learning_rate": 2.6019718309859157e-05, "loss": 5.1077, "step": 3830 }, { "epoch": 4.29, "learning_rate": 2.601830985915493e-05, "loss": 5.0934, "step": 3831 }, { "epoch": 4.3, "learning_rate": 2.6016901408450706e-05, "loss": 5.1721, "step": 3832 }, { "epoch": 4.3, "learning_rate": 2.601549295774648e-05, "loss": 5.1484, "step": 3833 }, { "epoch": 4.3, "learning_rate": 2.6014084507042255e-05, "loss": 4.9322, "step": 3834 }, { "epoch": 4.3, "learning_rate": 2.601267605633803e-05, "loss": 4.8308, "step": 3835 }, { "epoch": 4.3, "learning_rate": 2.60112676056338e-05, "loss": 4.9247, "step": 3836 }, { "epoch": 4.3, "learning_rate": 2.6009859154929578e-05, "loss": 4.8623, "step": 3837 }, { "epoch": 4.3, "learning_rate": 2.600845070422535e-05, "loss": 4.9316, "step": 3838 }, { "epoch": 4.3, "learning_rate": 2.6007042253521127e-05, "loss": 4.8615, "step": 3839 }, { "epoch": 4.3, "learning_rate": 2.60056338028169e-05, "loss": 4.7684, "step": 3840 }, { "epoch": 4.31, "learning_rate": 2.600422535211268e-05, "loss": 4.9575, "step": 3841 }, { "epoch": 4.31, "learning_rate": 2.6002816901408453e-05, "loss": 4.8855, "step": 3842 }, { "epoch": 4.31, "learning_rate": 2.6001408450704225e-05, "loss": 4.769, "step": 3843 }, { "epoch": 4.31, "learning_rate": 2.6000000000000002e-05, "loss": 4.9535, "step": 3844 }, { "epoch": 4.31, "learning_rate": 2.5998591549295775e-05, "loss": 4.8071, "step": 3845 }, { "epoch": 4.31, "learning_rate": 2.599718309859155e-05, "loss": 4.7856, "step": 3846 }, { "epoch": 4.31, "learning_rate": 2.5995774647887324e-05, "loss": 4.7961, "step": 3847 }, { "epoch": 4.31, "learning_rate": 2.59943661971831e-05, "loss": 4.7983, "step": 3848 }, { "epoch": 4.32, "learning_rate": 2.5992957746478873e-05, "loss": 4.8399, "step": 3849 }, { "epoch": 4.32, "learning_rate": 2.599154929577465e-05, "loss": 5.0458, "step": 3850 }, { "epoch": 4.32, "learning_rate": 2.5990140845070423e-05, "loss": 4.7592, "step": 3851 }, { "epoch": 4.32, "learning_rate": 2.5988732394366196e-05, "loss": 4.81, "step": 3852 }, { "epoch": 4.32, "learning_rate": 2.5987323943661972e-05, "loss": 4.8402, "step": 3853 }, { "epoch": 4.32, "learning_rate": 2.5985915492957745e-05, "loss": 4.7044, "step": 3854 }, { "epoch": 4.32, "learning_rate": 2.598450704225352e-05, "loss": 4.6136, "step": 3855 }, { "epoch": 4.32, "learning_rate": 2.5983098591549297e-05, "loss": 4.6961, "step": 3856 }, { "epoch": 4.32, "learning_rate": 2.5981690140845074e-05, "loss": 4.6545, "step": 3857 }, { "epoch": 4.33, "learning_rate": 2.5980281690140847e-05, "loss": 4.6612, "step": 3858 }, { "epoch": 4.33, "learning_rate": 2.597887323943662e-05, "loss": 4.9217, "step": 3859 }, { "epoch": 4.33, "learning_rate": 2.5977464788732396e-05, "loss": 4.618, "step": 3860 }, { "epoch": 4.33, "learning_rate": 2.597605633802817e-05, "loss": 4.5158, "step": 3861 }, { "epoch": 4.33, "learning_rate": 2.5974647887323945e-05, "loss": 4.5217, "step": 3862 }, { "epoch": 4.33, "learning_rate": 2.5973239436619718e-05, "loss": 4.3234, "step": 3863 }, { "epoch": 4.33, "learning_rate": 2.5971830985915495e-05, "loss": 4.2867, "step": 3864 }, { "epoch": 4.33, "learning_rate": 2.5970422535211267e-05, "loss": 4.087, "step": 3865 }, { "epoch": 4.33, "learning_rate": 2.5969014084507044e-05, "loss": 4.7247, "step": 3866 }, { "epoch": 4.34, "learning_rate": 2.5967605633802817e-05, "loss": 4.0801, "step": 3867 }, { "epoch": 4.34, "learning_rate": 2.596619718309859e-05, "loss": 3.7179, "step": 3868 }, { "epoch": 4.34, "learning_rate": 2.5964788732394366e-05, "loss": 5.608, "step": 3869 }, { "epoch": 4.34, "learning_rate": 2.596338028169014e-05, "loss": 5.4262, "step": 3870 }, { "epoch": 4.34, "learning_rate": 2.596197183098592e-05, "loss": 5.3583, "step": 3871 }, { "epoch": 4.34, "learning_rate": 2.596056338028169e-05, "loss": 5.146, "step": 3872 }, { "epoch": 4.34, "learning_rate": 2.5959154929577468e-05, "loss": 5.1943, "step": 3873 }, { "epoch": 4.34, "learning_rate": 2.595774647887324e-05, "loss": 5.095, "step": 3874 }, { "epoch": 4.34, "learning_rate": 2.5956338028169014e-05, "loss": 5.1092, "step": 3875 }, { "epoch": 4.35, "learning_rate": 2.595492957746479e-05, "loss": 5.1645, "step": 3876 }, { "epoch": 4.35, "learning_rate": 2.5953521126760563e-05, "loss": 4.9034, "step": 3877 }, { "epoch": 4.35, "learning_rate": 2.595211267605634e-05, "loss": 5.0217, "step": 3878 }, { "epoch": 4.35, "learning_rate": 2.5950704225352112e-05, "loss": 4.8514, "step": 3879 }, { "epoch": 4.35, "learning_rate": 2.594929577464789e-05, "loss": 5.1082, "step": 3880 }, { "epoch": 4.35, "learning_rate": 2.594788732394366e-05, "loss": 4.92, "step": 3881 }, { "epoch": 4.35, "learning_rate": 2.5946478873239438e-05, "loss": 4.7668, "step": 3882 }, { "epoch": 4.35, "learning_rate": 2.594507042253521e-05, "loss": 4.9191, "step": 3883 }, { "epoch": 4.35, "learning_rate": 2.5943661971830984e-05, "loss": 4.7994, "step": 3884 }, { "epoch": 4.36, "learning_rate": 2.594225352112676e-05, "loss": 4.7578, "step": 3885 }, { "epoch": 4.36, "learning_rate": 2.5940845070422537e-05, "loss": 4.9126, "step": 3886 }, { "epoch": 4.36, "learning_rate": 2.5939436619718313e-05, "loss": 4.7533, "step": 3887 }, { "epoch": 4.36, "learning_rate": 2.5938028169014086e-05, "loss": 4.8556, "step": 3888 }, { "epoch": 4.36, "learning_rate": 2.5936619718309862e-05, "loss": 4.8988, "step": 3889 }, { "epoch": 4.36, "learning_rate": 2.5935211267605635e-05, "loss": 4.6723, "step": 3890 }, { "epoch": 4.36, "learning_rate": 2.5933802816901408e-05, "loss": 5.0371, "step": 3891 }, { "epoch": 4.36, "learning_rate": 2.5932394366197184e-05, "loss": 4.806, "step": 3892 }, { "epoch": 4.36, "learning_rate": 2.5930985915492957e-05, "loss": 4.8174, "step": 3893 }, { "epoch": 4.37, "learning_rate": 2.5929577464788734e-05, "loss": 4.9353, "step": 3894 }, { "epoch": 4.37, "learning_rate": 2.5928169014084507e-05, "loss": 4.7083, "step": 3895 }, { "epoch": 4.37, "learning_rate": 2.5926760563380283e-05, "loss": 4.6926, "step": 3896 }, { "epoch": 4.37, "learning_rate": 2.5925352112676056e-05, "loss": 4.775, "step": 3897 }, { "epoch": 4.37, "learning_rate": 2.5923943661971832e-05, "loss": 4.8797, "step": 3898 }, { "epoch": 4.37, "learning_rate": 2.5922535211267605e-05, "loss": 4.7645, "step": 3899 }, { "epoch": 4.37, "learning_rate": 2.5921126760563378e-05, "loss": 4.5831, "step": 3900 }, { "epoch": 4.37, "learning_rate": 2.5919718309859158e-05, "loss": 4.7539, "step": 3901 }, { "epoch": 4.37, "learning_rate": 2.591830985915493e-05, "loss": 4.7283, "step": 3902 }, { "epoch": 4.38, "learning_rate": 2.5916901408450707e-05, "loss": 4.9037, "step": 3903 }, { "epoch": 4.38, "learning_rate": 2.591549295774648e-05, "loss": 4.7311, "step": 3904 }, { "epoch": 4.38, "learning_rate": 2.5914084507042256e-05, "loss": 4.8597, "step": 3905 }, { "epoch": 4.38, "learning_rate": 2.591267605633803e-05, "loss": 4.7569, "step": 3906 }, { "epoch": 4.38, "learning_rate": 2.5911267605633802e-05, "loss": 4.6087, "step": 3907 }, { "epoch": 4.38, "learning_rate": 2.590985915492958e-05, "loss": 4.6718, "step": 3908 }, { "epoch": 4.38, "learning_rate": 2.590845070422535e-05, "loss": 4.7522, "step": 3909 }, { "epoch": 4.38, "learning_rate": 2.5907042253521128e-05, "loss": 4.6554, "step": 3910 }, { "epoch": 4.38, "learning_rate": 2.59056338028169e-05, "loss": 4.5846, "step": 3911 }, { "epoch": 4.39, "learning_rate": 2.5904225352112677e-05, "loss": 4.4974, "step": 3912 }, { "epoch": 4.39, "learning_rate": 2.590281690140845e-05, "loss": 4.1727, "step": 3913 }, { "epoch": 4.39, "learning_rate": 2.5901408450704226e-05, "loss": 4.3157, "step": 3914 }, { "epoch": 4.39, "learning_rate": 2.59e-05, "loss": 4.2459, "step": 3915 }, { "epoch": 4.39, "learning_rate": 2.5898591549295772e-05, "loss": 4.2879, "step": 3916 }, { "epoch": 4.39, "learning_rate": 2.5897183098591552e-05, "loss": 4.0678, "step": 3917 }, { "epoch": 4.39, "learning_rate": 2.5895774647887325e-05, "loss": 3.5995, "step": 3918 }, { "epoch": 4.39, "learning_rate": 2.58943661971831e-05, "loss": 5.5826, "step": 3919 }, { "epoch": 4.39, "learning_rate": 2.5892957746478874e-05, "loss": 5.4562, "step": 3920 }, { "epoch": 4.4, "learning_rate": 2.589154929577465e-05, "loss": 5.169, "step": 3921 }, { "epoch": 4.4, "learning_rate": 2.5890140845070424e-05, "loss": 5.3786, "step": 3922 }, { "epoch": 4.4, "learning_rate": 2.5888732394366196e-05, "loss": 5.2302, "step": 3923 }, { "epoch": 4.4, "learning_rate": 2.5887323943661973e-05, "loss": 5.0571, "step": 3924 }, { "epoch": 4.4, "learning_rate": 2.5885915492957746e-05, "loss": 5.087, "step": 3925 }, { "epoch": 4.4, "learning_rate": 2.5884507042253522e-05, "loss": 5.1098, "step": 3926 }, { "epoch": 4.4, "learning_rate": 2.5883098591549295e-05, "loss": 4.8416, "step": 3927 }, { "epoch": 4.4, "learning_rate": 2.588169014084507e-05, "loss": 5.1023, "step": 3928 }, { "epoch": 4.4, "learning_rate": 2.5880281690140844e-05, "loss": 4.8329, "step": 3929 }, { "epoch": 4.41, "learning_rate": 2.587887323943662e-05, "loss": 4.9049, "step": 3930 }, { "epoch": 4.41, "learning_rate": 2.5877464788732394e-05, "loss": 4.94, "step": 3931 }, { "epoch": 4.41, "learning_rate": 2.587605633802817e-05, "loss": 4.7981, "step": 3932 }, { "epoch": 4.41, "learning_rate": 2.5874647887323946e-05, "loss": 4.8326, "step": 3933 }, { "epoch": 4.41, "learning_rate": 2.587323943661972e-05, "loss": 4.8335, "step": 3934 }, { "epoch": 4.41, "learning_rate": 2.5871830985915495e-05, "loss": 4.8182, "step": 3935 }, { "epoch": 4.41, "learning_rate": 2.587042253521127e-05, "loss": 4.8107, "step": 3936 }, { "epoch": 4.41, "learning_rate": 2.5869014084507045e-05, "loss": 4.6898, "step": 3937 }, { "epoch": 4.41, "learning_rate": 2.5867605633802818e-05, "loss": 5.0505, "step": 3938 }, { "epoch": 4.42, "learning_rate": 2.586619718309859e-05, "loss": 4.9829, "step": 3939 }, { "epoch": 4.42, "learning_rate": 2.5864788732394367e-05, "loss": 4.8331, "step": 3940 }, { "epoch": 4.42, "learning_rate": 2.586338028169014e-05, "loss": 4.948, "step": 3941 }, { "epoch": 4.42, "learning_rate": 2.5861971830985916e-05, "loss": 4.8234, "step": 3942 }, { "epoch": 4.42, "learning_rate": 2.586056338028169e-05, "loss": 4.9283, "step": 3943 }, { "epoch": 4.42, "learning_rate": 2.5859154929577466e-05, "loss": 4.8823, "step": 3944 }, { "epoch": 4.42, "learning_rate": 2.585774647887324e-05, "loss": 4.804, "step": 3945 }, { "epoch": 4.42, "learning_rate": 2.5856338028169015e-05, "loss": 4.9214, "step": 3946 }, { "epoch": 4.42, "learning_rate": 2.585492957746479e-05, "loss": 4.664, "step": 3947 }, { "epoch": 4.43, "learning_rate": 2.5853521126760564e-05, "loss": 4.6584, "step": 3948 }, { "epoch": 4.43, "learning_rate": 2.585211267605634e-05, "loss": 4.8066, "step": 3949 }, { "epoch": 4.43, "learning_rate": 2.5850704225352113e-05, "loss": 4.6281, "step": 3950 }, { "epoch": 4.43, "learning_rate": 2.584929577464789e-05, "loss": 4.6853, "step": 3951 }, { "epoch": 4.43, "learning_rate": 2.5847887323943663e-05, "loss": 4.7722, "step": 3952 }, { "epoch": 4.43, "learning_rate": 2.584647887323944e-05, "loss": 4.86, "step": 3953 }, { "epoch": 4.43, "learning_rate": 2.5845070422535212e-05, "loss": 4.8406, "step": 3954 }, { "epoch": 4.43, "learning_rate": 2.5843661971830985e-05, "loss": 4.8688, "step": 3955 }, { "epoch": 4.43, "learning_rate": 2.584225352112676e-05, "loss": 4.8076, "step": 3956 }, { "epoch": 4.44, "learning_rate": 2.5840845070422534e-05, "loss": 4.563, "step": 3957 }, { "epoch": 4.44, "learning_rate": 2.583943661971831e-05, "loss": 4.5965, "step": 3958 }, { "epoch": 4.44, "learning_rate": 2.5838028169014083e-05, "loss": 4.5641, "step": 3959 }, { "epoch": 4.44, "learning_rate": 2.583661971830986e-05, "loss": 4.5951, "step": 3960 }, { "epoch": 4.44, "learning_rate": 2.5835211267605633e-05, "loss": 4.409, "step": 3961 }, { "epoch": 4.44, "learning_rate": 2.583380281690141e-05, "loss": 4.5279, "step": 3962 }, { "epoch": 4.44, "learning_rate": 2.5832394366197185e-05, "loss": 4.337, "step": 3963 }, { "epoch": 4.44, "learning_rate": 2.5830985915492958e-05, "loss": 4.1529, "step": 3964 }, { "epoch": 4.45, "learning_rate": 2.5829577464788735e-05, "loss": 4.2832, "step": 3965 }, { "epoch": 4.45, "learning_rate": 2.5828169014084508e-05, "loss": 4.1164, "step": 3966 }, { "epoch": 4.45, "learning_rate": 2.5826760563380284e-05, "loss": 3.9915, "step": 3967 }, { "epoch": 4.45, "learning_rate": 2.5825352112676057e-05, "loss": 3.8712, "step": 3968 }, { "epoch": 4.45, "learning_rate": 2.5823943661971833e-05, "loss": 5.4794, "step": 3969 }, { "epoch": 4.45, "learning_rate": 2.5822535211267606e-05, "loss": 5.5516, "step": 3970 }, { "epoch": 4.45, "learning_rate": 2.582112676056338e-05, "loss": 5.384, "step": 3971 }, { "epoch": 4.45, "learning_rate": 2.5819718309859155e-05, "loss": 5.0833, "step": 3972 }, { "epoch": 4.45, "learning_rate": 2.581830985915493e-05, "loss": 5.2092, "step": 3973 }, { "epoch": 4.46, "learning_rate": 2.5816901408450705e-05, "loss": 5.0177, "step": 3974 }, { "epoch": 4.46, "learning_rate": 2.5815492957746478e-05, "loss": 5.0173, "step": 3975 }, { "epoch": 4.46, "learning_rate": 2.5814084507042254e-05, "loss": 5.0168, "step": 3976 }, { "epoch": 4.46, "learning_rate": 2.581267605633803e-05, "loss": 4.9668, "step": 3977 }, { "epoch": 4.46, "learning_rate": 2.5811267605633803e-05, "loss": 4.8325, "step": 3978 }, { "epoch": 4.46, "learning_rate": 2.580985915492958e-05, "loss": 4.6794, "step": 3979 }, { "epoch": 4.46, "learning_rate": 2.5808450704225352e-05, "loss": 4.9075, "step": 3980 }, { "epoch": 4.46, "learning_rate": 2.580704225352113e-05, "loss": 4.6986, "step": 3981 }, { "epoch": 4.46, "learning_rate": 2.5805633802816902e-05, "loss": 4.8238, "step": 3982 }, { "epoch": 4.47, "learning_rate": 2.5804225352112678e-05, "loss": 4.9795, "step": 3983 }, { "epoch": 4.47, "learning_rate": 2.580281690140845e-05, "loss": 4.6851, "step": 3984 }, { "epoch": 4.47, "learning_rate": 2.5801408450704227e-05, "loss": 4.7885, "step": 3985 }, { "epoch": 4.47, "learning_rate": 2.58e-05, "loss": 4.7232, "step": 3986 }, { "epoch": 4.47, "learning_rate": 2.5798591549295773e-05, "loss": 4.8585, "step": 3987 }, { "epoch": 4.47, "learning_rate": 2.579718309859155e-05, "loss": 4.9019, "step": 3988 }, { "epoch": 4.47, "learning_rate": 2.5795774647887323e-05, "loss": 4.8046, "step": 3989 }, { "epoch": 4.47, "learning_rate": 2.57943661971831e-05, "loss": 4.8588, "step": 3990 }, { "epoch": 4.47, "learning_rate": 2.5792957746478872e-05, "loss": 4.7349, "step": 3991 }, { "epoch": 4.48, "learning_rate": 2.579154929577465e-05, "loss": 4.7131, "step": 3992 }, { "epoch": 4.48, "learning_rate": 2.5790140845070424e-05, "loss": 4.697, "step": 3993 }, { "epoch": 4.48, "learning_rate": 2.5788732394366197e-05, "loss": 4.7299, "step": 3994 }, { "epoch": 4.48, "learning_rate": 2.5787323943661974e-05, "loss": 4.8003, "step": 3995 }, { "epoch": 4.48, "learning_rate": 2.5785915492957747e-05, "loss": 4.7559, "step": 3996 }, { "epoch": 4.48, "learning_rate": 2.5784507042253523e-05, "loss": 4.628, "step": 3997 }, { "epoch": 4.48, "learning_rate": 2.5783098591549296e-05, "loss": 4.6655, "step": 3998 }, { "epoch": 4.48, "learning_rate": 2.5781690140845072e-05, "loss": 4.4832, "step": 3999 }, { "epoch": 4.48, "learning_rate": 2.5780281690140845e-05, "loss": 4.6773, "step": 4000 }, { "epoch": 4.49, "learning_rate": 2.577887323943662e-05, "loss": 4.6091, "step": 4001 }, { "epoch": 4.49, "learning_rate": 2.5777464788732395e-05, "loss": 4.6025, "step": 4002 }, { "epoch": 4.49, "learning_rate": 2.5776056338028167e-05, "loss": 4.8039, "step": 4003 }, { "epoch": 4.49, "learning_rate": 2.5774647887323944e-05, "loss": 4.5195, "step": 4004 }, { "epoch": 4.49, "learning_rate": 2.5773239436619717e-05, "loss": 4.6383, "step": 4005 }, { "epoch": 4.49, "learning_rate": 2.5771830985915493e-05, "loss": 4.5266, "step": 4006 }, { "epoch": 4.49, "learning_rate": 2.577042253521127e-05, "loss": 4.6192, "step": 4007 }, { "epoch": 4.49, "learning_rate": 2.5769014084507046e-05, "loss": 4.652, "step": 4008 }, { "epoch": 4.49, "learning_rate": 2.576760563380282e-05, "loss": 4.5985, "step": 4009 }, { "epoch": 4.5, "learning_rate": 2.576619718309859e-05, "loss": 4.392, "step": 4010 }, { "epoch": 4.5, "learning_rate": 2.5764788732394368e-05, "loss": 4.6059, "step": 4011 }, { "epoch": 4.5, "learning_rate": 2.576338028169014e-05, "loss": 4.6727, "step": 4012 }, { "epoch": 4.5, "learning_rate": 2.5761971830985917e-05, "loss": 4.1768, "step": 4013 }, { "epoch": 4.5, "learning_rate": 2.576056338028169e-05, "loss": 4.0687, "step": 4014 }, { "epoch": 4.5, "learning_rate": 2.5759154929577466e-05, "loss": 3.9711, "step": 4015 }, { "epoch": 4.5, "learning_rate": 2.575774647887324e-05, "loss": 3.8315, "step": 4016 }, { "epoch": 4.5, "learning_rate": 2.5756338028169016e-05, "loss": 3.7569, "step": 4017 }, { "epoch": 4.5, "learning_rate": 2.575492957746479e-05, "loss": 3.8431, "step": 4018 }, { "epoch": 4.51, "learning_rate": 2.575352112676056e-05, "loss": 5.4262, "step": 4019 }, { "epoch": 4.51, "learning_rate": 2.5752112676056338e-05, "loss": 5.4479, "step": 4020 }, { "epoch": 4.51, "learning_rate": 2.575070422535211e-05, "loss": 5.4761, "step": 4021 }, { "epoch": 4.51, "learning_rate": 2.5749295774647887e-05, "loss": 5.2565, "step": 4022 }, { "epoch": 4.51, "learning_rate": 2.5747887323943664e-05, "loss": 5.049, "step": 4023 }, { "epoch": 4.51, "learning_rate": 2.574647887323944e-05, "loss": 5.0563, "step": 4024 }, { "epoch": 4.51, "learning_rate": 2.5745070422535213e-05, "loss": 5.1052, "step": 4025 }, { "epoch": 4.51, "learning_rate": 2.5743661971830986e-05, "loss": 4.9886, "step": 4026 }, { "epoch": 4.51, "learning_rate": 2.5742253521126762e-05, "loss": 4.9068, "step": 4027 }, { "epoch": 4.52, "learning_rate": 2.5740845070422535e-05, "loss": 4.7961, "step": 4028 }, { "epoch": 4.52, "learning_rate": 2.573943661971831e-05, "loss": 4.7625, "step": 4029 }, { "epoch": 4.52, "learning_rate": 2.5738028169014084e-05, "loss": 4.7036, "step": 4030 }, { "epoch": 4.52, "learning_rate": 2.573661971830986e-05, "loss": 5.0672, "step": 4031 }, { "epoch": 4.52, "learning_rate": 2.5735211267605634e-05, "loss": 5.0565, "step": 4032 }, { "epoch": 4.52, "learning_rate": 2.573380281690141e-05, "loss": 4.8043, "step": 4033 }, { "epoch": 4.52, "learning_rate": 2.5732394366197183e-05, "loss": 4.8689, "step": 4034 }, { "epoch": 4.52, "learning_rate": 2.5730985915492956e-05, "loss": 4.6725, "step": 4035 }, { "epoch": 4.52, "learning_rate": 2.5729577464788732e-05, "loss": 4.7555, "step": 4036 }, { "epoch": 4.53, "learning_rate": 2.5728169014084505e-05, "loss": 4.8676, "step": 4037 }, { "epoch": 4.53, "learning_rate": 2.5726760563380285e-05, "loss": 4.7515, "step": 4038 }, { "epoch": 4.53, "learning_rate": 2.5725352112676058e-05, "loss": 4.9014, "step": 4039 }, { "epoch": 4.53, "learning_rate": 2.5723943661971834e-05, "loss": 4.7306, "step": 4040 }, { "epoch": 4.53, "learning_rate": 2.5722535211267607e-05, "loss": 4.7836, "step": 4041 }, { "epoch": 4.53, "learning_rate": 2.572112676056338e-05, "loss": 4.8609, "step": 4042 }, { "epoch": 4.53, "learning_rate": 2.5719718309859156e-05, "loss": 4.862, "step": 4043 }, { "epoch": 4.53, "learning_rate": 2.571830985915493e-05, "loss": 4.7031, "step": 4044 }, { "epoch": 4.53, "learning_rate": 2.5716901408450706e-05, "loss": 4.6654, "step": 4045 }, { "epoch": 4.54, "learning_rate": 2.571549295774648e-05, "loss": 4.6561, "step": 4046 }, { "epoch": 4.54, "learning_rate": 2.5714084507042255e-05, "loss": 4.7105, "step": 4047 }, { "epoch": 4.54, "learning_rate": 2.5712676056338028e-05, "loss": 4.7029, "step": 4048 }, { "epoch": 4.54, "learning_rate": 2.5711267605633804e-05, "loss": 4.9324, "step": 4049 }, { "epoch": 4.54, "learning_rate": 2.5709859154929577e-05, "loss": 4.678, "step": 4050 }, { "epoch": 4.54, "learning_rate": 2.570845070422535e-05, "loss": 4.646, "step": 4051 }, { "epoch": 4.54, "learning_rate": 2.5707042253521126e-05, "loss": 4.8489, "step": 4052 }, { "epoch": 4.54, "learning_rate": 2.5705633802816903e-05, "loss": 4.8387, "step": 4053 }, { "epoch": 4.54, "learning_rate": 2.570422535211268e-05, "loss": 4.565, "step": 4054 }, { "epoch": 4.55, "learning_rate": 2.5702816901408452e-05, "loss": 4.8447, "step": 4055 }, { "epoch": 4.55, "learning_rate": 2.570140845070423e-05, "loss": 4.7193, "step": 4056 }, { "epoch": 4.55, "learning_rate": 2.57e-05, "loss": 4.4336, "step": 4057 }, { "epoch": 4.55, "learning_rate": 2.5698591549295774e-05, "loss": 4.6135, "step": 4058 }, { "epoch": 4.55, "learning_rate": 2.569718309859155e-05, "loss": 4.5698, "step": 4059 }, { "epoch": 4.55, "learning_rate": 2.5695774647887323e-05, "loss": 4.4563, "step": 4060 }, { "epoch": 4.55, "learning_rate": 2.56943661971831e-05, "loss": 4.6802, "step": 4061 }, { "epoch": 4.55, "learning_rate": 2.5692957746478873e-05, "loss": 4.4834, "step": 4062 }, { "epoch": 4.55, "learning_rate": 2.569154929577465e-05, "loss": 4.471, "step": 4063 }, { "epoch": 4.56, "learning_rate": 2.5690140845070422e-05, "loss": 4.1438, "step": 4064 }, { "epoch": 4.56, "learning_rate": 2.56887323943662e-05, "loss": 4.3233, "step": 4065 }, { "epoch": 4.56, "learning_rate": 2.568732394366197e-05, "loss": 4.214, "step": 4066 }, { "epoch": 4.56, "learning_rate": 2.5685915492957744e-05, "loss": 4.0974, "step": 4067 }, { "epoch": 4.56, "learning_rate": 2.5684507042253524e-05, "loss": 3.715, "step": 4068 }, { "epoch": 4.56, "learning_rate": 2.5683098591549297e-05, "loss": 5.4752, "step": 4069 }, { "epoch": 4.56, "learning_rate": 2.5681690140845073e-05, "loss": 5.4681, "step": 4070 }, { "epoch": 4.56, "learning_rate": 2.5680281690140846e-05, "loss": 5.4592, "step": 4071 }, { "epoch": 4.57, "learning_rate": 2.5678873239436623e-05, "loss": 5.2666, "step": 4072 }, { "epoch": 4.57, "learning_rate": 2.5677464788732395e-05, "loss": 5.3277, "step": 4073 }, { "epoch": 4.57, "learning_rate": 2.567605633802817e-05, "loss": 5.1118, "step": 4074 }, { "epoch": 4.57, "learning_rate": 2.5674647887323945e-05, "loss": 4.9976, "step": 4075 }, { "epoch": 4.57, "learning_rate": 2.5673239436619718e-05, "loss": 4.8379, "step": 4076 }, { "epoch": 4.57, "learning_rate": 2.5671830985915494e-05, "loss": 5.0186, "step": 4077 }, { "epoch": 4.57, "learning_rate": 2.5670422535211267e-05, "loss": 4.7657, "step": 4078 }, { "epoch": 4.57, "learning_rate": 2.5669014084507043e-05, "loss": 4.8031, "step": 4079 }, { "epoch": 4.57, "learning_rate": 2.5667605633802816e-05, "loss": 4.7563, "step": 4080 }, { "epoch": 4.58, "learning_rate": 2.566619718309859e-05, "loss": 4.9523, "step": 4081 }, { "epoch": 4.58, "learning_rate": 2.5664788732394366e-05, "loss": 4.6952, "step": 4082 }, { "epoch": 4.58, "learning_rate": 2.5663380281690142e-05, "loss": 4.6882, "step": 4083 }, { "epoch": 4.58, "learning_rate": 2.5661971830985918e-05, "loss": 4.7432, "step": 4084 }, { "epoch": 4.58, "learning_rate": 2.566056338028169e-05, "loss": 4.8637, "step": 4085 }, { "epoch": 4.58, "learning_rate": 2.5659154929577467e-05, "loss": 4.6988, "step": 4086 }, { "epoch": 4.58, "learning_rate": 2.565774647887324e-05, "loss": 4.6743, "step": 4087 }, { "epoch": 4.58, "learning_rate": 2.5656338028169017e-05, "loss": 4.6048, "step": 4088 }, { "epoch": 4.58, "learning_rate": 2.565492957746479e-05, "loss": 4.8475, "step": 4089 }, { "epoch": 4.59, "learning_rate": 2.5653521126760563e-05, "loss": 4.6543, "step": 4090 }, { "epoch": 4.59, "learning_rate": 2.565211267605634e-05, "loss": 4.7806, "step": 4091 }, { "epoch": 4.59, "learning_rate": 2.5650704225352112e-05, "loss": 4.7768, "step": 4092 }, { "epoch": 4.59, "learning_rate": 2.5649295774647888e-05, "loss": 4.5065, "step": 4093 }, { "epoch": 4.59, "learning_rate": 2.564788732394366e-05, "loss": 4.6529, "step": 4094 }, { "epoch": 4.59, "learning_rate": 2.5646478873239437e-05, "loss": 4.739, "step": 4095 }, { "epoch": 4.59, "learning_rate": 2.564507042253521e-05, "loss": 4.4413, "step": 4096 }, { "epoch": 4.59, "learning_rate": 2.5643661971830983e-05, "loss": 4.6624, "step": 4097 }, { "epoch": 4.59, "learning_rate": 2.5642253521126763e-05, "loss": 4.7273, "step": 4098 }, { "epoch": 4.6, "learning_rate": 2.5640845070422536e-05, "loss": 4.7617, "step": 4099 }, { "epoch": 4.6, "learning_rate": 2.5639436619718312e-05, "loss": 4.4233, "step": 4100 }, { "epoch": 4.6, "learning_rate": 2.5638028169014085e-05, "loss": 4.4215, "step": 4101 }, { "epoch": 4.6, "learning_rate": 2.563661971830986e-05, "loss": 4.5006, "step": 4102 }, { "epoch": 4.6, "learning_rate": 2.5635211267605635e-05, "loss": 4.6532, "step": 4103 }, { "epoch": 4.6, "learning_rate": 2.563380281690141e-05, "loss": 4.5514, "step": 4104 }, { "epoch": 4.6, "learning_rate": 2.5632394366197184e-05, "loss": 4.3715, "step": 4105 }, { "epoch": 4.6, "learning_rate": 2.5630985915492957e-05, "loss": 4.4236, "step": 4106 }, { "epoch": 4.6, "learning_rate": 2.5629577464788733e-05, "loss": 4.5391, "step": 4107 }, { "epoch": 4.61, "learning_rate": 2.5628169014084506e-05, "loss": 4.6251, "step": 4108 }, { "epoch": 4.61, "learning_rate": 2.5626760563380282e-05, "loss": 4.6863, "step": 4109 }, { "epoch": 4.61, "learning_rate": 2.5625352112676055e-05, "loss": 4.4915, "step": 4110 }, { "epoch": 4.61, "learning_rate": 2.5623943661971832e-05, "loss": 4.3183, "step": 4111 }, { "epoch": 4.61, "learning_rate": 2.5622535211267605e-05, "loss": 4.2948, "step": 4112 }, { "epoch": 4.61, "learning_rate": 2.562112676056338e-05, "loss": 4.1384, "step": 4113 }, { "epoch": 4.61, "learning_rate": 2.5619718309859157e-05, "loss": 4.1391, "step": 4114 }, { "epoch": 4.61, "learning_rate": 2.561830985915493e-05, "loss": 4.3696, "step": 4115 }, { "epoch": 4.61, "learning_rate": 2.5616901408450707e-05, "loss": 4.1234, "step": 4116 }, { "epoch": 4.62, "learning_rate": 2.561549295774648e-05, "loss": 3.9943, "step": 4117 }, { "epoch": 4.62, "learning_rate": 2.5614084507042256e-05, "loss": 3.7997, "step": 4118 }, { "epoch": 4.62, "learning_rate": 2.561267605633803e-05, "loss": 5.3105, "step": 4119 }, { "epoch": 4.62, "learning_rate": 2.5611267605633805e-05, "loss": 5.4078, "step": 4120 }, { "epoch": 4.62, "learning_rate": 2.5609859154929578e-05, "loss": 5.3784, "step": 4121 }, { "epoch": 4.62, "learning_rate": 2.560845070422535e-05, "loss": 5.2749, "step": 4122 }, { "epoch": 4.62, "learning_rate": 2.5607042253521127e-05, "loss": 5.1009, "step": 4123 }, { "epoch": 4.62, "learning_rate": 2.56056338028169e-05, "loss": 5.0133, "step": 4124 }, { "epoch": 4.62, "learning_rate": 2.5604225352112677e-05, "loss": 5.078, "step": 4125 }, { "epoch": 4.63, "learning_rate": 2.560281690140845e-05, "loss": 4.8067, "step": 4126 }, { "epoch": 4.63, "learning_rate": 2.5601408450704226e-05, "loss": 5.093, "step": 4127 }, { "epoch": 4.63, "learning_rate": 2.5600000000000002e-05, "loss": 4.769, "step": 4128 }, { "epoch": 4.63, "learning_rate": 2.5598591549295775e-05, "loss": 4.871, "step": 4129 }, { "epoch": 4.63, "learning_rate": 2.559718309859155e-05, "loss": 4.8854, "step": 4130 }, { "epoch": 4.63, "learning_rate": 2.5595774647887324e-05, "loss": 4.9136, "step": 4131 }, { "epoch": 4.63, "learning_rate": 2.55943661971831e-05, "loss": 4.8314, "step": 4132 }, { "epoch": 4.63, "learning_rate": 2.5592957746478874e-05, "loss": 4.7401, "step": 4133 }, { "epoch": 4.63, "learning_rate": 2.559154929577465e-05, "loss": 4.9807, "step": 4134 }, { "epoch": 4.64, "learning_rate": 2.5590140845070423e-05, "loss": 4.7788, "step": 4135 }, { "epoch": 4.64, "learning_rate": 2.55887323943662e-05, "loss": 4.7407, "step": 4136 }, { "epoch": 4.64, "learning_rate": 2.5587323943661972e-05, "loss": 4.7006, "step": 4137 }, { "epoch": 4.64, "learning_rate": 2.5585915492957745e-05, "loss": 4.7574, "step": 4138 }, { "epoch": 4.64, "learning_rate": 2.558450704225352e-05, "loss": 4.7143, "step": 4139 }, { "epoch": 4.64, "learning_rate": 2.5583098591549294e-05, "loss": 4.6466, "step": 4140 }, { "epoch": 4.64, "learning_rate": 2.558169014084507e-05, "loss": 4.6134, "step": 4141 }, { "epoch": 4.64, "learning_rate": 2.5580281690140844e-05, "loss": 4.5582, "step": 4142 }, { "epoch": 4.64, "learning_rate": 2.557887323943662e-05, "loss": 4.8235, "step": 4143 }, { "epoch": 4.65, "learning_rate": 2.5577464788732396e-05, "loss": 4.5535, "step": 4144 }, { "epoch": 4.65, "learning_rate": 2.557605633802817e-05, "loss": 4.8461, "step": 4145 }, { "epoch": 4.65, "learning_rate": 2.5574647887323946e-05, "loss": 4.617, "step": 4146 }, { "epoch": 4.65, "learning_rate": 2.557323943661972e-05, "loss": 4.4752, "step": 4147 }, { "epoch": 4.65, "learning_rate": 2.5571830985915495e-05, "loss": 4.6713, "step": 4148 }, { "epoch": 4.65, "learning_rate": 2.5570422535211268e-05, "loss": 4.6049, "step": 4149 }, { "epoch": 4.65, "learning_rate": 2.5569014084507044e-05, "loss": 4.5022, "step": 4150 }, { "epoch": 4.65, "learning_rate": 2.5567605633802817e-05, "loss": 4.5093, "step": 4151 }, { "epoch": 4.65, "learning_rate": 2.5566197183098594e-05, "loss": 4.4099, "step": 4152 }, { "epoch": 4.66, "learning_rate": 2.5564788732394366e-05, "loss": 4.4801, "step": 4153 }, { "epoch": 4.66, "learning_rate": 2.556338028169014e-05, "loss": 4.526, "step": 4154 }, { "epoch": 4.66, "learning_rate": 2.5561971830985916e-05, "loss": 4.4081, "step": 4155 }, { "epoch": 4.66, "learning_rate": 2.556056338028169e-05, "loss": 4.3911, "step": 4156 }, { "epoch": 4.66, "learning_rate": 2.5559154929577465e-05, "loss": 4.3287, "step": 4157 }, { "epoch": 4.66, "learning_rate": 2.5557746478873238e-05, "loss": 4.4536, "step": 4158 }, { "epoch": 4.66, "learning_rate": 2.5556338028169018e-05, "loss": 4.2793, "step": 4159 }, { "epoch": 4.66, "learning_rate": 2.555492957746479e-05, "loss": 4.3769, "step": 4160 }, { "epoch": 4.66, "learning_rate": 2.5553521126760564e-05, "loss": 4.5831, "step": 4161 }, { "epoch": 4.67, "learning_rate": 2.555211267605634e-05, "loss": 4.3799, "step": 4162 }, { "epoch": 4.67, "learning_rate": 2.5550704225352113e-05, "loss": 4.5624, "step": 4163 }, { "epoch": 4.67, "learning_rate": 2.554929577464789e-05, "loss": 4.3079, "step": 4164 }, { "epoch": 4.67, "learning_rate": 2.5547887323943662e-05, "loss": 4.372, "step": 4165 }, { "epoch": 4.67, "learning_rate": 2.554647887323944e-05, "loss": 3.9188, "step": 4166 }, { "epoch": 4.67, "learning_rate": 2.554507042253521e-05, "loss": 3.8892, "step": 4167 }, { "epoch": 4.67, "learning_rate": 2.5543661971830988e-05, "loss": 4.0508, "step": 4168 }, { "epoch": 4.67, "learning_rate": 2.554225352112676e-05, "loss": 5.3688, "step": 4169 }, { "epoch": 4.67, "learning_rate": 2.5540845070422534e-05, "loss": 5.3245, "step": 4170 }, { "epoch": 4.68, "learning_rate": 2.553943661971831e-05, "loss": 5.337, "step": 4171 }, { "epoch": 4.68, "learning_rate": 2.5538028169014083e-05, "loss": 5.2399, "step": 4172 }, { "epoch": 4.68, "learning_rate": 2.553661971830986e-05, "loss": 5.2962, "step": 4173 }, { "epoch": 4.68, "learning_rate": 2.5535211267605636e-05, "loss": 5.0756, "step": 4174 }, { "epoch": 4.68, "learning_rate": 2.5533802816901412e-05, "loss": 5.0978, "step": 4175 }, { "epoch": 4.68, "learning_rate": 2.5532394366197185e-05, "loss": 4.96, "step": 4176 }, { "epoch": 4.68, "learning_rate": 2.5530985915492958e-05, "loss": 4.8173, "step": 4177 }, { "epoch": 4.68, "learning_rate": 2.5529577464788734e-05, "loss": 4.817, "step": 4178 }, { "epoch": 4.68, "learning_rate": 2.5528169014084507e-05, "loss": 4.749, "step": 4179 }, { "epoch": 4.69, "learning_rate": 2.5526760563380283e-05, "loss": 4.9175, "step": 4180 }, { "epoch": 4.69, "learning_rate": 2.5525352112676056e-05, "loss": 4.781, "step": 4181 }, { "epoch": 4.69, "learning_rate": 2.5523943661971833e-05, "loss": 4.7639, "step": 4182 }, { "epoch": 4.69, "learning_rate": 2.5522535211267606e-05, "loss": 4.7502, "step": 4183 }, { "epoch": 4.69, "learning_rate": 2.552112676056338e-05, "loss": 4.7867, "step": 4184 }, { "epoch": 4.69, "learning_rate": 2.5519718309859155e-05, "loss": 4.6658, "step": 4185 }, { "epoch": 4.69, "learning_rate": 2.5518309859154928e-05, "loss": 4.6749, "step": 4186 }, { "epoch": 4.69, "learning_rate": 2.5516901408450704e-05, "loss": 4.7731, "step": 4187 }, { "epoch": 4.7, "learning_rate": 2.5515492957746477e-05, "loss": 4.9924, "step": 4188 }, { "epoch": 4.7, "learning_rate": 2.5514084507042257e-05, "loss": 4.6283, "step": 4189 }, { "epoch": 4.7, "learning_rate": 2.551267605633803e-05, "loss": 4.6695, "step": 4190 }, { "epoch": 4.7, "learning_rate": 2.5511267605633806e-05, "loss": 4.6509, "step": 4191 }, { "epoch": 4.7, "learning_rate": 2.550985915492958e-05, "loss": 4.6845, "step": 4192 }, { "epoch": 4.7, "learning_rate": 2.5508450704225352e-05, "loss": 4.8075, "step": 4193 }, { "epoch": 4.7, "learning_rate": 2.5507042253521128e-05, "loss": 4.5747, "step": 4194 }, { "epoch": 4.7, "learning_rate": 2.55056338028169e-05, "loss": 4.8769, "step": 4195 }, { "epoch": 4.7, "learning_rate": 2.5504225352112678e-05, "loss": 4.562, "step": 4196 }, { "epoch": 4.71, "learning_rate": 2.550281690140845e-05, "loss": 4.6546, "step": 4197 }, { "epoch": 4.71, "learning_rate": 2.5501408450704227e-05, "loss": 4.7696, "step": 4198 }, { "epoch": 4.71, "learning_rate": 2.55e-05, "loss": 4.6921, "step": 4199 }, { "epoch": 4.71, "learning_rate": 2.5498591549295773e-05, "loss": 4.7507, "step": 4200 }, { "epoch": 4.71, "learning_rate": 2.549718309859155e-05, "loss": 4.3793, "step": 4201 }, { "epoch": 4.71, "learning_rate": 2.5495774647887322e-05, "loss": 4.5702, "step": 4202 }, { "epoch": 4.71, "learning_rate": 2.54943661971831e-05, "loss": 4.5833, "step": 4203 }, { "epoch": 4.71, "learning_rate": 2.5492957746478875e-05, "loss": 4.4802, "step": 4204 }, { "epoch": 4.71, "learning_rate": 2.549154929577465e-05, "loss": 4.4622, "step": 4205 }, { "epoch": 4.72, "learning_rate": 2.5490140845070424e-05, "loss": 4.5289, "step": 4206 }, { "epoch": 4.72, "learning_rate": 2.54887323943662e-05, "loss": 4.4879, "step": 4207 }, { "epoch": 4.72, "learning_rate": 2.5487323943661973e-05, "loss": 4.369, "step": 4208 }, { "epoch": 4.72, "learning_rate": 2.5485915492957746e-05, "loss": 4.5203, "step": 4209 }, { "epoch": 4.72, "learning_rate": 2.5484507042253523e-05, "loss": 4.6213, "step": 4210 }, { "epoch": 4.72, "learning_rate": 2.5483098591549295e-05, "loss": 4.2354, "step": 4211 }, { "epoch": 4.72, "learning_rate": 2.5481690140845072e-05, "loss": 4.3223, "step": 4212 }, { "epoch": 4.72, "learning_rate": 2.5480281690140845e-05, "loss": 4.0859, "step": 4213 }, { "epoch": 4.72, "learning_rate": 2.547887323943662e-05, "loss": 4.1328, "step": 4214 }, { "epoch": 4.73, "learning_rate": 2.5477464788732394e-05, "loss": 4.0892, "step": 4215 }, { "epoch": 4.73, "learning_rate": 2.5476056338028167e-05, "loss": 3.9395, "step": 4216 }, { "epoch": 4.73, "learning_rate": 2.5474647887323943e-05, "loss": 3.9962, "step": 4217 }, { "epoch": 4.73, "learning_rate": 2.5473239436619716e-05, "loss": 3.6347, "step": 4218 }, { "epoch": 4.73, "learning_rate": 2.5471830985915496e-05, "loss": 5.6374, "step": 4219 }, { "epoch": 4.73, "learning_rate": 2.547042253521127e-05, "loss": 5.5514, "step": 4220 }, { "epoch": 4.73, "learning_rate": 2.5469014084507045e-05, "loss": 5.4365, "step": 4221 }, { "epoch": 4.73, "learning_rate": 2.5467605633802818e-05, "loss": 5.4114, "step": 4222 }, { "epoch": 4.73, "learning_rate": 2.5466197183098594e-05, "loss": 5.0541, "step": 4223 }, { "epoch": 4.74, "learning_rate": 2.5464788732394367e-05, "loss": 4.9598, "step": 4224 }, { "epoch": 4.74, "learning_rate": 2.546338028169014e-05, "loss": 5.1155, "step": 4225 }, { "epoch": 4.74, "learning_rate": 2.5461971830985917e-05, "loss": 4.8778, "step": 4226 }, { "epoch": 4.74, "learning_rate": 2.546056338028169e-05, "loss": 4.8349, "step": 4227 }, { "epoch": 4.74, "learning_rate": 2.5459154929577466e-05, "loss": 4.7684, "step": 4228 }, { "epoch": 4.74, "learning_rate": 2.545774647887324e-05, "loss": 4.7931, "step": 4229 }, { "epoch": 4.74, "learning_rate": 2.5456338028169015e-05, "loss": 4.6889, "step": 4230 }, { "epoch": 4.74, "learning_rate": 2.5454929577464788e-05, "loss": 4.6578, "step": 4231 }, { "epoch": 4.74, "learning_rate": 2.545352112676056e-05, "loss": 4.5848, "step": 4232 }, { "epoch": 4.75, "learning_rate": 2.5452112676056337e-05, "loss": 4.6038, "step": 4233 }, { "epoch": 4.75, "learning_rate": 2.5450704225352114e-05, "loss": 4.6188, "step": 4234 }, { "epoch": 4.75, "learning_rate": 2.544929577464789e-05, "loss": 4.7022, "step": 4235 }, { "epoch": 4.75, "learning_rate": 2.5447887323943663e-05, "loss": 4.6262, "step": 4236 }, { "epoch": 4.75, "learning_rate": 2.544647887323944e-05, "loss": 4.6646, "step": 4237 }, { "epoch": 4.75, "learning_rate": 2.5445070422535212e-05, "loss": 4.6681, "step": 4238 }, { "epoch": 4.75, "learning_rate": 2.544366197183099e-05, "loss": 4.5551, "step": 4239 }, { "epoch": 4.75, "learning_rate": 2.544225352112676e-05, "loss": 4.6427, "step": 4240 }, { "epoch": 4.75, "learning_rate": 2.5440845070422535e-05, "loss": 4.337, "step": 4241 }, { "epoch": 4.76, "learning_rate": 2.543943661971831e-05, "loss": 4.5667, "step": 4242 }, { "epoch": 4.76, "learning_rate": 2.5438028169014084e-05, "loss": 4.7007, "step": 4243 }, { "epoch": 4.76, "learning_rate": 2.543661971830986e-05, "loss": 4.6934, "step": 4244 }, { "epoch": 4.76, "learning_rate": 2.5435211267605633e-05, "loss": 4.662, "step": 4245 }, { "epoch": 4.76, "learning_rate": 2.543380281690141e-05, "loss": 4.6251, "step": 4246 }, { "epoch": 4.76, "learning_rate": 2.5432394366197182e-05, "loss": 4.6418, "step": 4247 }, { "epoch": 4.76, "learning_rate": 2.5430985915492955e-05, "loss": 4.5528, "step": 4248 }, { "epoch": 4.76, "learning_rate": 2.5429577464788735e-05, "loss": 4.3914, "step": 4249 }, { "epoch": 4.76, "learning_rate": 2.5428169014084508e-05, "loss": 4.4628, "step": 4250 }, { "epoch": 4.77, "learning_rate": 2.5426760563380284e-05, "loss": 4.5529, "step": 4251 }, { "epoch": 4.77, "learning_rate": 2.5425352112676057e-05, "loss": 4.5995, "step": 4252 }, { "epoch": 4.77, "learning_rate": 2.5423943661971834e-05, "loss": 4.5593, "step": 4253 }, { "epoch": 4.77, "learning_rate": 2.5422535211267607e-05, "loss": 4.2896, "step": 4254 }, { "epoch": 4.77, "learning_rate": 2.5421126760563383e-05, "loss": 4.4223, "step": 4255 }, { "epoch": 4.77, "learning_rate": 2.5419718309859156e-05, "loss": 4.4026, "step": 4256 }, { "epoch": 4.77, "learning_rate": 2.541830985915493e-05, "loss": 4.317, "step": 4257 }, { "epoch": 4.77, "learning_rate": 2.5416901408450705e-05, "loss": 4.3703, "step": 4258 }, { "epoch": 4.77, "learning_rate": 2.5415492957746478e-05, "loss": 4.4447, "step": 4259 }, { "epoch": 4.78, "learning_rate": 2.5414084507042254e-05, "loss": 4.2911, "step": 4260 }, { "epoch": 4.78, "learning_rate": 2.5412676056338027e-05, "loss": 4.4786, "step": 4261 }, { "epoch": 4.78, "learning_rate": 2.5411267605633804e-05, "loss": 3.959, "step": 4262 }, { "epoch": 4.78, "learning_rate": 2.5409859154929577e-05, "loss": 4.0245, "step": 4263 }, { "epoch": 4.78, "learning_rate": 2.540845070422535e-05, "loss": 4.0744, "step": 4264 }, { "epoch": 4.78, "learning_rate": 2.540704225352113e-05, "loss": 3.8679, "step": 4265 }, { "epoch": 4.78, "learning_rate": 2.5405633802816902e-05, "loss": 4.0271, "step": 4266 }, { "epoch": 4.78, "learning_rate": 2.540422535211268e-05, "loss": 3.8775, "step": 4267 }, { "epoch": 4.78, "learning_rate": 2.540281690140845e-05, "loss": 3.5268, "step": 4268 }, { "epoch": 4.79, "learning_rate": 2.5401408450704228e-05, "loss": 5.303, "step": 4269 }, { "epoch": 4.79, "learning_rate": 2.54e-05, "loss": 5.3831, "step": 4270 }, { "epoch": 4.79, "learning_rate": 2.5398591549295777e-05, "loss": 5.2878, "step": 4271 }, { "epoch": 4.79, "learning_rate": 2.539718309859155e-05, "loss": 5.0805, "step": 4272 }, { "epoch": 4.79, "learning_rate": 2.5395774647887323e-05, "loss": 5.1407, "step": 4273 }, { "epoch": 4.79, "learning_rate": 2.53943661971831e-05, "loss": 5.0126, "step": 4274 }, { "epoch": 4.79, "learning_rate": 2.5392957746478872e-05, "loss": 4.9877, "step": 4275 }, { "epoch": 4.79, "learning_rate": 2.539154929577465e-05, "loss": 4.9527, "step": 4276 }, { "epoch": 4.79, "learning_rate": 2.539014084507042e-05, "loss": 4.6919, "step": 4277 }, { "epoch": 4.8, "learning_rate": 2.5388732394366198e-05, "loss": 4.7146, "step": 4278 }, { "epoch": 4.8, "learning_rate": 2.538732394366197e-05, "loss": 4.6118, "step": 4279 }, { "epoch": 4.8, "learning_rate": 2.5385915492957747e-05, "loss": 4.8078, "step": 4280 }, { "epoch": 4.8, "learning_rate": 2.5384507042253523e-05, "loss": 4.654, "step": 4281 }, { "epoch": 4.8, "learning_rate": 2.5383098591549296e-05, "loss": 4.603, "step": 4282 }, { "epoch": 4.8, "learning_rate": 2.5381690140845073e-05, "loss": 4.8975, "step": 4283 }, { "epoch": 4.8, "learning_rate": 2.5380281690140846e-05, "loss": 4.5766, "step": 4284 }, { "epoch": 4.8, "learning_rate": 2.5378873239436622e-05, "loss": 4.8352, "step": 4285 }, { "epoch": 4.8, "learning_rate": 2.5377464788732395e-05, "loss": 4.5732, "step": 4286 }, { "epoch": 4.81, "learning_rate": 2.537605633802817e-05, "loss": 4.6915, "step": 4287 }, { "epoch": 4.81, "learning_rate": 2.5374647887323944e-05, "loss": 4.657, "step": 4288 }, { "epoch": 4.81, "learning_rate": 2.5373239436619717e-05, "loss": 4.7203, "step": 4289 }, { "epoch": 4.81, "learning_rate": 2.5371830985915494e-05, "loss": 4.5952, "step": 4290 }, { "epoch": 4.81, "learning_rate": 2.5370422535211266e-05, "loss": 4.3997, "step": 4291 }, { "epoch": 4.81, "learning_rate": 2.5369014084507043e-05, "loss": 4.6, "step": 4292 }, { "epoch": 4.81, "learning_rate": 2.5367605633802816e-05, "loss": 4.5403, "step": 4293 }, { "epoch": 4.81, "learning_rate": 2.5366197183098592e-05, "loss": 4.4192, "step": 4294 }, { "epoch": 4.82, "learning_rate": 2.536478873239437e-05, "loss": 4.4559, "step": 4295 }, { "epoch": 4.82, "learning_rate": 2.536338028169014e-05, "loss": 4.6334, "step": 4296 }, { "epoch": 4.82, "learning_rate": 2.5361971830985918e-05, "loss": 4.5, "step": 4297 }, { "epoch": 4.82, "learning_rate": 2.536056338028169e-05, "loss": 4.5838, "step": 4298 }, { "epoch": 4.82, "learning_rate": 2.5359154929577467e-05, "loss": 4.5092, "step": 4299 }, { "epoch": 4.82, "learning_rate": 2.535774647887324e-05, "loss": 4.3288, "step": 4300 }, { "epoch": 4.82, "learning_rate": 2.5356338028169016e-05, "loss": 4.3348, "step": 4301 }, { "epoch": 4.82, "learning_rate": 2.535492957746479e-05, "loss": 4.408, "step": 4302 }, { "epoch": 4.82, "learning_rate": 2.5353521126760562e-05, "loss": 4.6112, "step": 4303 }, { "epoch": 4.83, "learning_rate": 2.535211267605634e-05, "loss": 4.2527, "step": 4304 }, { "epoch": 4.83, "learning_rate": 2.535070422535211e-05, "loss": 4.4032, "step": 4305 }, { "epoch": 4.83, "learning_rate": 2.5349295774647888e-05, "loss": 4.5288, "step": 4306 }, { "epoch": 4.83, "learning_rate": 2.534788732394366e-05, "loss": 4.2969, "step": 4307 }, { "epoch": 4.83, "learning_rate": 2.5346478873239437e-05, "loss": 4.3288, "step": 4308 }, { "epoch": 4.83, "learning_rate": 2.534507042253521e-05, "loss": 4.265, "step": 4309 }, { "epoch": 4.83, "learning_rate": 2.534366197183099e-05, "loss": 4.1433, "step": 4310 }, { "epoch": 4.83, "learning_rate": 2.5342253521126763e-05, "loss": 4.2292, "step": 4311 }, { "epoch": 4.83, "learning_rate": 2.5340845070422536e-05, "loss": 4.145, "step": 4312 }, { "epoch": 4.84, "learning_rate": 2.5339436619718312e-05, "loss": 4.0954, "step": 4313 }, { "epoch": 4.84, "learning_rate": 2.5338028169014085e-05, "loss": 4.3295, "step": 4314 }, { "epoch": 4.84, "learning_rate": 2.533661971830986e-05, "loss": 3.7898, "step": 4315 }, { "epoch": 4.84, "learning_rate": 2.5335211267605634e-05, "loss": 3.7544, "step": 4316 }, { "epoch": 4.84, "learning_rate": 2.533380281690141e-05, "loss": 3.8458, "step": 4317 }, { "epoch": 4.84, "learning_rate": 2.5332394366197183e-05, "loss": 3.6618, "step": 4318 }, { "epoch": 4.84, "learning_rate": 2.5330985915492956e-05, "loss": 5.5149, "step": 4319 }, { "epoch": 4.84, "learning_rate": 2.5329577464788733e-05, "loss": 5.257, "step": 4320 }, { "epoch": 4.84, "learning_rate": 2.5328169014084506e-05, "loss": 5.3685, "step": 4321 }, { "epoch": 4.85, "learning_rate": 2.5326760563380282e-05, "loss": 5.4429, "step": 4322 }, { "epoch": 4.85, "learning_rate": 2.5325352112676055e-05, "loss": 5.1264, "step": 4323 }, { "epoch": 4.85, "learning_rate": 2.532394366197183e-05, "loss": 5.1492, "step": 4324 }, { "epoch": 4.85, "learning_rate": 2.5322535211267608e-05, "loss": 5.0448, "step": 4325 }, { "epoch": 4.85, "learning_rate": 2.5321126760563384e-05, "loss": 4.8158, "step": 4326 }, { "epoch": 4.85, "learning_rate": 2.5319718309859157e-05, "loss": 4.818, "step": 4327 }, { "epoch": 4.85, "learning_rate": 2.531830985915493e-05, "loss": 4.5857, "step": 4328 }, { "epoch": 4.85, "learning_rate": 2.5316901408450706e-05, "loss": 4.7512, "step": 4329 }, { "epoch": 4.85, "learning_rate": 2.531549295774648e-05, "loss": 4.8476, "step": 4330 }, { "epoch": 4.86, "learning_rate": 2.5314084507042255e-05, "loss": 4.6392, "step": 4331 }, { "epoch": 4.86, "learning_rate": 2.5312676056338028e-05, "loss": 4.8961, "step": 4332 }, { "epoch": 4.86, "learning_rate": 2.5311267605633805e-05, "loss": 4.8845, "step": 4333 }, { "epoch": 4.86, "learning_rate": 2.5309859154929578e-05, "loss": 4.634, "step": 4334 }, { "epoch": 4.86, "learning_rate": 2.530845070422535e-05, "loss": 4.6237, "step": 4335 }, { "epoch": 4.86, "learning_rate": 2.5307042253521127e-05, "loss": 4.4056, "step": 4336 }, { "epoch": 4.86, "learning_rate": 2.53056338028169e-05, "loss": 4.7218, "step": 4337 }, { "epoch": 4.86, "learning_rate": 2.5304225352112676e-05, "loss": 4.6864, "step": 4338 }, { "epoch": 4.86, "learning_rate": 2.530281690140845e-05, "loss": 4.4281, "step": 4339 }, { "epoch": 4.87, "learning_rate": 2.530140845070423e-05, "loss": 4.6817, "step": 4340 }, { "epoch": 4.87, "learning_rate": 2.5300000000000002e-05, "loss": 4.5376, "step": 4341 }, { "epoch": 4.87, "learning_rate": 2.5298591549295778e-05, "loss": 4.5562, "step": 4342 }, { "epoch": 4.87, "learning_rate": 2.529718309859155e-05, "loss": 4.5507, "step": 4343 }, { "epoch": 4.87, "learning_rate": 2.5295774647887324e-05, "loss": 4.4243, "step": 4344 }, { "epoch": 4.87, "learning_rate": 2.52943661971831e-05, "loss": 4.6536, "step": 4345 }, { "epoch": 4.87, "learning_rate": 2.5292957746478873e-05, "loss": 4.6425, "step": 4346 }, { "epoch": 4.87, "learning_rate": 2.529154929577465e-05, "loss": 4.6012, "step": 4347 }, { "epoch": 4.87, "learning_rate": 2.5290140845070422e-05, "loss": 4.5201, "step": 4348 }, { "epoch": 4.88, "learning_rate": 2.52887323943662e-05, "loss": 4.634, "step": 4349 }, { "epoch": 4.88, "learning_rate": 2.5287323943661972e-05, "loss": 4.4704, "step": 4350 }, { "epoch": 4.88, "learning_rate": 2.5285915492957745e-05, "loss": 4.5225, "step": 4351 }, { "epoch": 4.88, "learning_rate": 2.528450704225352e-05, "loss": 4.4527, "step": 4352 }, { "epoch": 4.88, "learning_rate": 2.5283098591549294e-05, "loss": 4.409, "step": 4353 }, { "epoch": 4.88, "learning_rate": 2.528169014084507e-05, "loss": 4.3972, "step": 4354 }, { "epoch": 4.88, "learning_rate": 2.5280281690140847e-05, "loss": 4.4777, "step": 4355 }, { "epoch": 4.88, "learning_rate": 2.5278873239436623e-05, "loss": 4.2236, "step": 4356 }, { "epoch": 4.88, "learning_rate": 2.5277464788732396e-05, "loss": 4.3449, "step": 4357 }, { "epoch": 4.89, "learning_rate": 2.5276056338028172e-05, "loss": 4.2204, "step": 4358 }, { "epoch": 4.89, "learning_rate": 2.5274647887323945e-05, "loss": 4.2529, "step": 4359 }, { "epoch": 4.89, "learning_rate": 2.5273239436619718e-05, "loss": 4.1792, "step": 4360 }, { "epoch": 4.89, "learning_rate": 2.5271830985915494e-05, "loss": 4.2221, "step": 4361 }, { "epoch": 4.89, "learning_rate": 2.5270422535211267e-05, "loss": 3.9136, "step": 4362 }, { "epoch": 4.89, "learning_rate": 2.5269014084507044e-05, "loss": 3.9897, "step": 4363 }, { "epoch": 4.89, "learning_rate": 2.5267605633802817e-05, "loss": 3.9635, "step": 4364 }, { "epoch": 4.89, "learning_rate": 2.5266197183098593e-05, "loss": 4.2591, "step": 4365 }, { "epoch": 4.89, "learning_rate": 2.5264788732394366e-05, "loss": 3.8099, "step": 4366 }, { "epoch": 4.9, "learning_rate": 2.526338028169014e-05, "loss": 3.8723, "step": 4367 }, { "epoch": 4.9, "learning_rate": 2.5261971830985915e-05, "loss": 3.6873, "step": 4368 }, { "epoch": 4.9, "learning_rate": 2.5260563380281688e-05, "loss": 5.4044, "step": 4369 }, { "epoch": 4.9, "learning_rate": 2.5259154929577468e-05, "loss": 5.2612, "step": 4370 }, { "epoch": 4.9, "learning_rate": 2.525774647887324e-05, "loss": 5.2118, "step": 4371 }, { "epoch": 4.9, "learning_rate": 2.5256338028169017e-05, "loss": 5.1285, "step": 4372 }, { "epoch": 4.9, "learning_rate": 2.525492957746479e-05, "loss": 5.0505, "step": 4373 }, { "epoch": 4.9, "learning_rate": 2.5253521126760566e-05, "loss": 4.94, "step": 4374 }, { "epoch": 4.9, "learning_rate": 2.525211267605634e-05, "loss": 4.9563, "step": 4375 }, { "epoch": 4.91, "learning_rate": 2.5250704225352112e-05, "loss": 4.6422, "step": 4376 }, { "epoch": 4.91, "learning_rate": 2.524929577464789e-05, "loss": 4.6742, "step": 4377 }, { "epoch": 4.91, "learning_rate": 2.524788732394366e-05, "loss": 4.6706, "step": 4378 }, { "epoch": 4.91, "learning_rate": 2.5246478873239438e-05, "loss": 4.8068, "step": 4379 }, { "epoch": 4.91, "learning_rate": 2.524507042253521e-05, "loss": 4.6545, "step": 4380 }, { "epoch": 4.91, "learning_rate": 2.5243661971830987e-05, "loss": 4.4751, "step": 4381 }, { "epoch": 4.91, "learning_rate": 2.524225352112676e-05, "loss": 4.7216, "step": 4382 }, { "epoch": 4.91, "learning_rate": 2.5240845070422533e-05, "loss": 4.5653, "step": 4383 }, { "epoch": 4.91, "learning_rate": 2.523943661971831e-05, "loss": 4.8358, "step": 4384 }, { "epoch": 4.92, "learning_rate": 2.5238028169014082e-05, "loss": 4.6188, "step": 4385 }, { "epoch": 4.92, "learning_rate": 2.5236619718309862e-05, "loss": 4.5912, "step": 4386 }, { "epoch": 4.92, "learning_rate": 2.5235211267605635e-05, "loss": 4.568, "step": 4387 }, { "epoch": 4.92, "learning_rate": 2.523380281690141e-05, "loss": 4.3676, "step": 4388 }, { "epoch": 4.92, "learning_rate": 2.5232394366197184e-05, "loss": 4.4255, "step": 4389 }, { "epoch": 4.92, "learning_rate": 2.523098591549296e-05, "loss": 4.5439, "step": 4390 }, { "epoch": 4.92, "learning_rate": 2.5229577464788734e-05, "loss": 4.5205, "step": 4391 }, { "epoch": 4.92, "learning_rate": 2.5228169014084507e-05, "loss": 4.586, "step": 4392 }, { "epoch": 4.92, "learning_rate": 2.5226760563380283e-05, "loss": 4.4982, "step": 4393 }, { "epoch": 4.93, "learning_rate": 2.5225352112676056e-05, "loss": 4.4209, "step": 4394 }, { "epoch": 4.93, "learning_rate": 2.5223943661971832e-05, "loss": 4.5379, "step": 4395 }, { "epoch": 4.93, "learning_rate": 2.5222535211267605e-05, "loss": 4.4286, "step": 4396 }, { "epoch": 4.93, "learning_rate": 2.522112676056338e-05, "loss": 4.2499, "step": 4397 }, { "epoch": 4.93, "learning_rate": 2.5219718309859154e-05, "loss": 4.4073, "step": 4398 }, { "epoch": 4.93, "learning_rate": 2.5218309859154927e-05, "loss": 4.3925, "step": 4399 }, { "epoch": 4.93, "learning_rate": 2.5216901408450704e-05, "loss": 4.4456, "step": 4400 }, { "epoch": 4.93, "learning_rate": 2.521549295774648e-05, "loss": 4.394, "step": 4401 }, { "epoch": 4.93, "learning_rate": 2.5214084507042256e-05, "loss": 4.2883, "step": 4402 }, { "epoch": 4.94, "learning_rate": 2.521267605633803e-05, "loss": 4.1904, "step": 4403 }, { "epoch": 4.94, "learning_rate": 2.5211267605633806e-05, "loss": 4.2889, "step": 4404 }, { "epoch": 4.94, "learning_rate": 2.520985915492958e-05, "loss": 4.4031, "step": 4405 }, { "epoch": 4.94, "learning_rate": 2.520845070422535e-05, "loss": 4.2963, "step": 4406 }, { "epoch": 4.94, "learning_rate": 2.5207042253521128e-05, "loss": 4.4124, "step": 4407 }, { "epoch": 4.94, "learning_rate": 2.52056338028169e-05, "loss": 4.2831, "step": 4408 }, { "epoch": 4.94, "learning_rate": 2.5204225352112677e-05, "loss": 4.3441, "step": 4409 }, { "epoch": 4.94, "learning_rate": 2.520281690140845e-05, "loss": 4.1125, "step": 4410 }, { "epoch": 4.95, "learning_rate": 2.5201408450704226e-05, "loss": 4.194, "step": 4411 }, { "epoch": 4.95, "learning_rate": 2.52e-05, "loss": 4.1517, "step": 4412 }, { "epoch": 4.95, "learning_rate": 2.5198591549295776e-05, "loss": 4.0182, "step": 4413 }, { "epoch": 4.95, "learning_rate": 2.519718309859155e-05, "loss": 4.2307, "step": 4414 }, { "epoch": 4.95, "learning_rate": 2.519577464788732e-05, "loss": 3.7525, "step": 4415 }, { "epoch": 4.95, "learning_rate": 2.51943661971831e-05, "loss": 4.0247, "step": 4416 }, { "epoch": 4.95, "learning_rate": 2.5192957746478874e-05, "loss": 3.9597, "step": 4417 }, { "epoch": 4.95, "learning_rate": 2.519154929577465e-05, "loss": 3.7992, "step": 4418 }, { "epoch": 4.95, "learning_rate": 2.5190140845070423e-05, "loss": 5.2253, "step": 4419 }, { "epoch": 4.96, "learning_rate": 2.51887323943662e-05, "loss": 5.2453, "step": 4420 }, { "epoch": 4.96, "learning_rate": 2.5187323943661973e-05, "loss": 5.3615, "step": 4421 }, { "epoch": 4.96, "learning_rate": 2.5185915492957746e-05, "loss": 5.0519, "step": 4422 }, { "epoch": 4.96, "learning_rate": 2.5184507042253522e-05, "loss": 4.9824, "step": 4423 }, { "epoch": 4.96, "learning_rate": 2.5183098591549295e-05, "loss": 5.0812, "step": 4424 }, { "epoch": 4.96, "learning_rate": 2.518169014084507e-05, "loss": 4.7386, "step": 4425 }, { "epoch": 4.96, "learning_rate": 2.5180281690140844e-05, "loss": 4.6989, "step": 4426 }, { "epoch": 4.96, "learning_rate": 2.517887323943662e-05, "loss": 4.6304, "step": 4427 }, { "epoch": 4.96, "learning_rate": 2.5177464788732393e-05, "loss": 4.4752, "step": 4428 }, { "epoch": 4.97, "learning_rate": 2.517605633802817e-05, "loss": 4.6751, "step": 4429 }, { "epoch": 4.97, "learning_rate": 2.5174647887323943e-05, "loss": 4.619, "step": 4430 }, { "epoch": 4.97, "learning_rate": 2.517323943661972e-05, "loss": 4.6885, "step": 4431 }, { "epoch": 4.97, "learning_rate": 2.5171830985915495e-05, "loss": 4.6616, "step": 4432 }, { "epoch": 4.97, "learning_rate": 2.517042253521127e-05, "loss": 4.624, "step": 4433 }, { "epoch": 4.97, "learning_rate": 2.5169014084507045e-05, "loss": 4.5831, "step": 4434 }, { "epoch": 4.97, "learning_rate": 2.5167605633802818e-05, "loss": 4.6196, "step": 4435 }, { "epoch": 4.97, "learning_rate": 2.5166197183098594e-05, "loss": 4.5078, "step": 4436 }, { "epoch": 4.97, "learning_rate": 2.5164788732394367e-05, "loss": 4.434, "step": 4437 }, { "epoch": 4.98, "learning_rate": 2.516338028169014e-05, "loss": 4.6611, "step": 4438 }, { "epoch": 4.98, "learning_rate": 2.5161971830985916e-05, "loss": 4.4755, "step": 4439 }, { "epoch": 4.98, "learning_rate": 2.516056338028169e-05, "loss": 4.6299, "step": 4440 }, { "epoch": 4.98, "learning_rate": 2.5159154929577465e-05, "loss": 4.5842, "step": 4441 }, { "epoch": 4.98, "learning_rate": 2.515774647887324e-05, "loss": 4.5274, "step": 4442 }, { "epoch": 4.98, "learning_rate": 2.5156338028169015e-05, "loss": 4.5347, "step": 4443 }, { "epoch": 4.98, "learning_rate": 2.5154929577464788e-05, "loss": 4.3147, "step": 4444 }, { "epoch": 4.98, "learning_rate": 2.5153521126760564e-05, "loss": 4.3562, "step": 4445 }, { "epoch": 4.98, "learning_rate": 2.515211267605634e-05, "loss": 4.4786, "step": 4446 }, { "epoch": 4.99, "learning_rate": 2.5150704225352113e-05, "loss": 4.5249, "step": 4447 }, { "epoch": 4.99, "learning_rate": 2.514929577464789e-05, "loss": 4.633, "step": 4448 }, { "epoch": 4.99, "learning_rate": 2.5147887323943663e-05, "loss": 4.4011, "step": 4449 }, { "epoch": 4.99, "learning_rate": 2.514647887323944e-05, "loss": 4.3563, "step": 4450 }, { "epoch": 4.99, "learning_rate": 2.5145070422535212e-05, "loss": 4.2371, "step": 4451 }, { "epoch": 4.99, "learning_rate": 2.5143661971830988e-05, "loss": 4.3243, "step": 4452 }, { "epoch": 4.99, "learning_rate": 2.514225352112676e-05, "loss": 4.405, "step": 4453 }, { "epoch": 4.99, "learning_rate": 2.5140845070422534e-05, "loss": 4.1306, "step": 4454 }, { "epoch": 4.99, "learning_rate": 2.513943661971831e-05, "loss": 4.0316, "step": 4455 }, { "epoch": 5.0, "learning_rate": 2.5138028169014083e-05, "loss": 4.1055, "step": 4456 }, { "epoch": 5.0, "learning_rate": 2.513661971830986e-05, "loss": 3.9497, "step": 4457 }, { "epoch": 5.0, "learning_rate": 2.5135211267605633e-05, "loss": 3.7287, "step": 4458 }, { "epoch": 5.0, "learning_rate": 2.513380281690141e-05, "loss": 4.1499, "step": 4459 }, { "epoch": 5.0, "learning_rate": 2.5132394366197182e-05, "loss": 3.2272, "step": 4460 }, { "epoch": 5.0, "learning_rate": 2.513098591549296e-05, "loss": 5.2525, "step": 4461 }, { "epoch": 5.0, "learning_rate": 2.5129577464788735e-05, "loss": 5.1816, "step": 4462 }, { "epoch": 5.0, "learning_rate": 2.5128169014084507e-05, "loss": 5.0649, "step": 4463 }, { "epoch": 5.0, "learning_rate": 2.5126760563380284e-05, "loss": 5.1471, "step": 4464 }, { "epoch": 5.01, "learning_rate": 2.5125352112676057e-05, "loss": 4.9135, "step": 4465 }, { "epoch": 5.01, "learning_rate": 2.5123943661971833e-05, "loss": 4.799, "step": 4466 }, { "epoch": 5.01, "learning_rate": 2.5122535211267606e-05, "loss": 4.696, "step": 4467 }, { "epoch": 5.01, "learning_rate": 2.5121126760563382e-05, "loss": 4.6792, "step": 4468 }, { "epoch": 5.01, "learning_rate": 2.5119718309859155e-05, "loss": 4.5265, "step": 4469 }, { "epoch": 5.01, "learning_rate": 2.5118309859154928e-05, "loss": 4.573, "step": 4470 }, { "epoch": 5.01, "learning_rate": 2.5116901408450705e-05, "loss": 4.4882, "step": 4471 }, { "epoch": 5.01, "learning_rate": 2.5115492957746478e-05, "loss": 4.5198, "step": 4472 }, { "epoch": 5.01, "learning_rate": 2.5114084507042254e-05, "loss": 4.5515, "step": 4473 }, { "epoch": 5.02, "learning_rate": 2.5112676056338027e-05, "loss": 4.5577, "step": 4474 }, { "epoch": 5.02, "learning_rate": 2.5111267605633803e-05, "loss": 4.3693, "step": 4475 }, { "epoch": 5.02, "learning_rate": 2.510985915492958e-05, "loss": 4.4566, "step": 4476 }, { "epoch": 5.02, "learning_rate": 2.5108450704225356e-05, "loss": 4.3851, "step": 4477 }, { "epoch": 5.02, "learning_rate": 2.510704225352113e-05, "loss": 4.4042, "step": 4478 }, { "epoch": 5.02, "learning_rate": 2.51056338028169e-05, "loss": 4.3433, "step": 4479 }, { "epoch": 5.02, "learning_rate": 2.5104225352112678e-05, "loss": 4.2528, "step": 4480 }, { "epoch": 5.02, "learning_rate": 2.510281690140845e-05, "loss": 4.4295, "step": 4481 }, { "epoch": 5.02, "learning_rate": 2.5101408450704227e-05, "loss": 4.3751, "step": 4482 }, { "epoch": 5.03, "learning_rate": 2.51e-05, "loss": 4.3081, "step": 4483 }, { "epoch": 5.03, "learning_rate": 2.5098591549295777e-05, "loss": 4.4219, "step": 4484 }, { "epoch": 5.03, "learning_rate": 2.509718309859155e-05, "loss": 4.4139, "step": 4485 }, { "epoch": 5.03, "learning_rate": 2.5095774647887322e-05, "loss": 4.339, "step": 4486 }, { "epoch": 5.03, "learning_rate": 2.50943661971831e-05, "loss": 4.4406, "step": 4487 }, { "epoch": 5.03, "learning_rate": 2.5092957746478872e-05, "loss": 4.1018, "step": 4488 }, { "epoch": 5.03, "learning_rate": 2.5091549295774648e-05, "loss": 4.3477, "step": 4489 }, { "epoch": 5.03, "learning_rate": 2.509014084507042e-05, "loss": 4.1556, "step": 4490 }, { "epoch": 5.03, "learning_rate": 2.50887323943662e-05, "loss": 4.3966, "step": 4491 }, { "epoch": 5.04, "learning_rate": 2.5087323943661974e-05, "loss": 4.1944, "step": 4492 }, { "epoch": 5.04, "learning_rate": 2.508591549295775e-05, "loss": 4.3668, "step": 4493 }, { "epoch": 5.04, "learning_rate": 2.5084507042253523e-05, "loss": 4.143, "step": 4494 }, { "epoch": 5.04, "learning_rate": 2.5083098591549296e-05, "loss": 4.2563, "step": 4495 }, { "epoch": 5.04, "learning_rate": 2.5081690140845072e-05, "loss": 4.1186, "step": 4496 }, { "epoch": 5.04, "learning_rate": 2.5080281690140845e-05, "loss": 4.2421, "step": 4497 }, { "epoch": 5.04, "learning_rate": 2.507887323943662e-05, "loss": 4.3225, "step": 4498 }, { "epoch": 5.04, "learning_rate": 2.5077464788732394e-05, "loss": 4.1002, "step": 4499 }, { "epoch": 5.04, "learning_rate": 2.507605633802817e-05, "loss": 4.1635, "step": 4500 }, { "epoch": 5.04, "eval_loss": 4.418454647064209, "eval_runtime": 850.6586, "eval_samples_per_second": 3.106, "eval_steps_per_second": 0.389, "eval_wer": 0.9643990479968266, "step": 4500 }, { "epoch": 5.05, "learning_rate": 2.5074647887323944e-05, "loss": 4.1334, "step": 4501 }, { "epoch": 5.05, "learning_rate": 2.5073239436619717e-05, "loss": 4.1327, "step": 4502 }, { "epoch": 5.05, "learning_rate": 2.5071830985915493e-05, "loss": 3.9927, "step": 4503 }, { "epoch": 5.05, "learning_rate": 2.5070422535211266e-05, "loss": 3.933, "step": 4504 }, { "epoch": 5.05, "learning_rate": 2.5069014084507042e-05, "loss": 3.7663, "step": 4505 }, { "epoch": 5.05, "learning_rate": 2.5067605633802815e-05, "loss": 3.6219, "step": 4506 }, { "epoch": 5.05, "learning_rate": 2.5066197183098595e-05, "loss": 3.5125, "step": 4507 }, { "epoch": 5.05, "learning_rate": 2.5064788732394368e-05, "loss": 3.7067, "step": 4508 }, { "epoch": 5.05, "learning_rate": 2.506338028169014e-05, "loss": 3.4432, "step": 4509 }, { "epoch": 5.06, "learning_rate": 2.5061971830985917e-05, "loss": 3.2361, "step": 4510 }, { "epoch": 5.06, "learning_rate": 2.506056338028169e-05, "loss": 5.1154, "step": 4511 }, { "epoch": 5.06, "learning_rate": 2.5059154929577466e-05, "loss": 5.01, "step": 4512 }, { "epoch": 5.06, "learning_rate": 2.505774647887324e-05, "loss": 5.1598, "step": 4513 }, { "epoch": 5.06, "learning_rate": 2.5056338028169016e-05, "loss": 5.1109, "step": 4514 }, { "epoch": 5.06, "learning_rate": 2.505492957746479e-05, "loss": 4.8225, "step": 4515 }, { "epoch": 5.06, "learning_rate": 2.5053521126760565e-05, "loss": 4.9806, "step": 4516 }, { "epoch": 5.06, "learning_rate": 2.5052112676056338e-05, "loss": 4.728, "step": 4517 }, { "epoch": 5.07, "learning_rate": 2.505070422535211e-05, "loss": 4.7577, "step": 4518 }, { "epoch": 5.07, "learning_rate": 2.5049295774647887e-05, "loss": 4.5486, "step": 4519 }, { "epoch": 5.07, "learning_rate": 2.504788732394366e-05, "loss": 4.4715, "step": 4520 }, { "epoch": 5.07, "learning_rate": 2.5046478873239436e-05, "loss": 4.3655, "step": 4521 }, { "epoch": 5.07, "learning_rate": 2.5045070422535213e-05, "loss": 4.4885, "step": 4522 }, { "epoch": 5.07, "learning_rate": 2.504366197183099e-05, "loss": 4.514, "step": 4523 }, { "epoch": 5.07, "learning_rate": 2.5042253521126762e-05, "loss": 4.4834, "step": 4524 }, { "epoch": 5.07, "learning_rate": 2.5040845070422535e-05, "loss": 4.4108, "step": 4525 }, { "epoch": 5.07, "learning_rate": 2.503943661971831e-05, "loss": 4.3736, "step": 4526 }, { "epoch": 5.08, "learning_rate": 2.5038028169014084e-05, "loss": 4.4629, "step": 4527 }, { "epoch": 5.08, "learning_rate": 2.503661971830986e-05, "loss": 4.2302, "step": 4528 }, { "epoch": 5.08, "learning_rate": 2.5035211267605634e-05, "loss": 4.451, "step": 4529 }, { "epoch": 5.08, "learning_rate": 2.503380281690141e-05, "loss": 4.4127, "step": 4530 }, { "epoch": 5.08, "learning_rate": 2.5032394366197183e-05, "loss": 4.1755, "step": 4531 }, { "epoch": 5.08, "learning_rate": 2.503098591549296e-05, "loss": 4.4088, "step": 4532 }, { "epoch": 5.08, "learning_rate": 2.5029577464788732e-05, "loss": 4.4484, "step": 4533 }, { "epoch": 5.08, "learning_rate": 2.5028169014084505e-05, "loss": 4.3483, "step": 4534 }, { "epoch": 5.08, "learning_rate": 2.502676056338028e-05, "loss": 4.1664, "step": 4535 }, { "epoch": 5.09, "learning_rate": 2.5025352112676054e-05, "loss": 4.4078, "step": 4536 }, { "epoch": 5.09, "learning_rate": 2.5023943661971834e-05, "loss": 4.3507, "step": 4537 }, { "epoch": 5.09, "learning_rate": 2.5022535211267607e-05, "loss": 4.2445, "step": 4538 }, { "epoch": 5.09, "learning_rate": 2.5021126760563383e-05, "loss": 4.2374, "step": 4539 }, { "epoch": 5.09, "learning_rate": 2.5019718309859156e-05, "loss": 4.2539, "step": 4540 }, { "epoch": 5.09, "learning_rate": 2.501830985915493e-05, "loss": 4.2062, "step": 4541 }, { "epoch": 5.09, "learning_rate": 2.5016901408450706e-05, "loss": 4.1707, "step": 4542 }, { "epoch": 5.09, "learning_rate": 2.501549295774648e-05, "loss": 4.2305, "step": 4543 }, { "epoch": 5.09, "learning_rate": 2.5014084507042255e-05, "loss": 4.088, "step": 4544 }, { "epoch": 5.1, "learning_rate": 2.5012676056338028e-05, "loss": 4.1011, "step": 4545 }, { "epoch": 5.1, "learning_rate": 2.5011267605633804e-05, "loss": 4.3019, "step": 4546 }, { "epoch": 5.1, "learning_rate": 2.5009859154929577e-05, "loss": 4.2498, "step": 4547 }, { "epoch": 5.1, "learning_rate": 2.5008450704225353e-05, "loss": 4.2734, "step": 4548 }, { "epoch": 5.1, "learning_rate": 2.5007042253521126e-05, "loss": 4.0884, "step": 4549 }, { "epoch": 5.1, "learning_rate": 2.50056338028169e-05, "loss": 3.8887, "step": 4550 }, { "epoch": 5.1, "learning_rate": 2.5004225352112676e-05, "loss": 4.1619, "step": 4551 }, { "epoch": 5.1, "learning_rate": 2.5002816901408452e-05, "loss": 3.8936, "step": 4552 }, { "epoch": 5.1, "learning_rate": 2.5001408450704228e-05, "loss": 3.9273, "step": 4553 }, { "epoch": 5.11, "learning_rate": 2.5e-05, "loss": 3.8228, "step": 4554 }, { "epoch": 5.11, "learning_rate": 2.4998591549295778e-05, "loss": 3.7416, "step": 4555 }, { "epoch": 5.11, "learning_rate": 2.499718309859155e-05, "loss": 3.5315, "step": 4556 }, { "epoch": 5.11, "learning_rate": 2.4995774647887323e-05, "loss": 3.6455, "step": 4557 }, { "epoch": 5.11, "learning_rate": 2.49943661971831e-05, "loss": 3.5294, "step": 4558 }, { "epoch": 5.11, "learning_rate": 2.4992957746478873e-05, "loss": 3.4158, "step": 4559 }, { "epoch": 5.11, "learning_rate": 2.499154929577465e-05, "loss": 3.0523, "step": 4560 }, { "epoch": 5.11, "learning_rate": 2.4990140845070422e-05, "loss": 5.3069, "step": 4561 }, { "epoch": 5.11, "learning_rate": 2.4988732394366198e-05, "loss": 5.04, "step": 4562 }, { "epoch": 5.12, "learning_rate": 2.498732394366197e-05, "loss": 4.9776, "step": 4563 }, { "epoch": 5.12, "learning_rate": 2.4985915492957748e-05, "loss": 5.0097, "step": 4564 }, { "epoch": 5.12, "learning_rate": 2.498450704225352e-05, "loss": 5.1329, "step": 4565 }, { "epoch": 5.12, "learning_rate": 2.4983098591549293e-05, "loss": 4.6403, "step": 4566 }, { "epoch": 5.12, "learning_rate": 2.4981690140845073e-05, "loss": 4.6498, "step": 4567 }, { "epoch": 5.12, "learning_rate": 2.4980281690140846e-05, "loss": 4.5418, "step": 4568 }, { "epoch": 5.12, "learning_rate": 2.4978873239436622e-05, "loss": 4.6831, "step": 4569 }, { "epoch": 5.12, "learning_rate": 2.4977464788732395e-05, "loss": 4.4512, "step": 4570 }, { "epoch": 5.12, "learning_rate": 2.4976056338028172e-05, "loss": 4.2152, "step": 4571 }, { "epoch": 5.13, "learning_rate": 2.4974647887323945e-05, "loss": 4.5446, "step": 4572 }, { "epoch": 5.13, "learning_rate": 2.4973239436619718e-05, "loss": 4.5208, "step": 4573 }, { "epoch": 5.13, "learning_rate": 2.4971830985915494e-05, "loss": 4.382, "step": 4574 }, { "epoch": 5.13, "learning_rate": 2.4970422535211267e-05, "loss": 4.3897, "step": 4575 }, { "epoch": 5.13, "learning_rate": 2.4969014084507043e-05, "loss": 4.2322, "step": 4576 }, { "epoch": 5.13, "learning_rate": 2.4967605633802816e-05, "loss": 4.4172, "step": 4577 }, { "epoch": 5.13, "learning_rate": 2.4966197183098592e-05, "loss": 4.3674, "step": 4578 }, { "epoch": 5.13, "learning_rate": 2.4964788732394365e-05, "loss": 4.3678, "step": 4579 }, { "epoch": 5.13, "learning_rate": 2.4963380281690142e-05, "loss": 4.4197, "step": 4580 }, { "epoch": 5.14, "learning_rate": 2.4961971830985915e-05, "loss": 4.2387, "step": 4581 }, { "epoch": 5.14, "learning_rate": 2.496056338028169e-05, "loss": 4.4758, "step": 4582 }, { "epoch": 5.14, "learning_rate": 2.4959154929577467e-05, "loss": 4.2924, "step": 4583 }, { "epoch": 5.14, "learning_rate": 2.495774647887324e-05, "loss": 4.4601, "step": 4584 }, { "epoch": 5.14, "learning_rate": 2.4956338028169017e-05, "loss": 4.3315, "step": 4585 }, { "epoch": 5.14, "learning_rate": 2.495492957746479e-05, "loss": 4.1834, "step": 4586 }, { "epoch": 5.14, "learning_rate": 2.4953521126760566e-05, "loss": 4.307, "step": 4587 }, { "epoch": 5.14, "learning_rate": 2.495211267605634e-05, "loss": 4.1576, "step": 4588 }, { "epoch": 5.14, "learning_rate": 2.4950704225352112e-05, "loss": 4.1996, "step": 4589 }, { "epoch": 5.15, "learning_rate": 2.4949295774647888e-05, "loss": 4.2496, "step": 4590 }, { "epoch": 5.15, "learning_rate": 2.494788732394366e-05, "loss": 4.216, "step": 4591 }, { "epoch": 5.15, "learning_rate": 2.4946478873239437e-05, "loss": 4.0771, "step": 4592 }, { "epoch": 5.15, "learning_rate": 2.494507042253521e-05, "loss": 4.2069, "step": 4593 }, { "epoch": 5.15, "learning_rate": 2.4943661971830987e-05, "loss": 4.0195, "step": 4594 }, { "epoch": 5.15, "learning_rate": 2.494225352112676e-05, "loss": 4.1923, "step": 4595 }, { "epoch": 5.15, "learning_rate": 2.4940845070422536e-05, "loss": 4.0917, "step": 4596 }, { "epoch": 5.15, "learning_rate": 2.4939436619718312e-05, "loss": 4.1583, "step": 4597 }, { "epoch": 5.15, "learning_rate": 2.4938028169014085e-05, "loss": 4.407, "step": 4598 }, { "epoch": 5.16, "learning_rate": 2.493661971830986e-05, "loss": 3.912, "step": 4599 }, { "epoch": 5.16, "learning_rate": 2.4935211267605635e-05, "loss": 4.0649, "step": 4600 }, { "epoch": 5.16, "learning_rate": 2.493380281690141e-05, "loss": 4.0782, "step": 4601 }, { "epoch": 5.16, "learning_rate": 2.4932394366197184e-05, "loss": 4.0893, "step": 4602 }, { "epoch": 5.16, "learning_rate": 2.493098591549296e-05, "loss": 3.9703, "step": 4603 }, { "epoch": 5.16, "learning_rate": 2.4929577464788733e-05, "loss": 3.9818, "step": 4604 }, { "epoch": 5.16, "learning_rate": 2.4928169014084506e-05, "loss": 3.8726, "step": 4605 }, { "epoch": 5.16, "learning_rate": 2.4926760563380282e-05, "loss": 3.8501, "step": 4606 }, { "epoch": 5.16, "learning_rate": 2.4925352112676055e-05, "loss": 3.5653, "step": 4607 }, { "epoch": 5.17, "learning_rate": 2.492394366197183e-05, "loss": 3.5419, "step": 4608 }, { "epoch": 5.17, "learning_rate": 2.4922535211267605e-05, "loss": 4.0125, "step": 4609 }, { "epoch": 5.17, "learning_rate": 2.492112676056338e-05, "loss": 3.3073, "step": 4610 }, { "epoch": 5.17, "learning_rate": 2.4919718309859154e-05, "loss": 5.3195, "step": 4611 }, { "epoch": 5.17, "learning_rate": 2.4918309859154934e-05, "loss": 5.1036, "step": 4612 }, { "epoch": 5.17, "learning_rate": 2.4916901408450707e-05, "loss": 4.9902, "step": 4613 }, { "epoch": 5.17, "learning_rate": 2.491549295774648e-05, "loss": 5.0747, "step": 4614 }, { "epoch": 5.17, "learning_rate": 2.4914084507042256e-05, "loss": 4.7824, "step": 4615 }, { "epoch": 5.17, "learning_rate": 2.491267605633803e-05, "loss": 4.8564, "step": 4616 }, { "epoch": 5.18, "learning_rate": 2.4911267605633805e-05, "loss": 4.6721, "step": 4617 }, { "epoch": 5.18, "learning_rate": 2.4909859154929578e-05, "loss": 4.4741, "step": 4618 }, { "epoch": 5.18, "learning_rate": 2.4908450704225354e-05, "loss": 4.4435, "step": 4619 }, { "epoch": 5.18, "learning_rate": 2.4907042253521127e-05, "loss": 4.3953, "step": 4620 }, { "epoch": 5.18, "learning_rate": 2.49056338028169e-05, "loss": 4.47, "step": 4621 }, { "epoch": 5.18, "learning_rate": 2.4904225352112677e-05, "loss": 4.2493, "step": 4622 }, { "epoch": 5.18, "learning_rate": 2.490281690140845e-05, "loss": 4.2916, "step": 4623 }, { "epoch": 5.18, "learning_rate": 2.4901408450704226e-05, "loss": 4.3498, "step": 4624 }, { "epoch": 5.18, "learning_rate": 2.49e-05, "loss": 4.2033, "step": 4625 }, { "epoch": 5.19, "learning_rate": 2.4898591549295775e-05, "loss": 4.2374, "step": 4626 }, { "epoch": 5.19, "learning_rate": 2.4897183098591548e-05, "loss": 4.2635, "step": 4627 }, { "epoch": 5.19, "learning_rate": 2.4895774647887324e-05, "loss": 4.3289, "step": 4628 }, { "epoch": 5.19, "learning_rate": 2.48943661971831e-05, "loss": 4.2622, "step": 4629 }, { "epoch": 5.19, "learning_rate": 2.4892957746478874e-05, "loss": 4.2837, "step": 4630 }, { "epoch": 5.19, "learning_rate": 2.489154929577465e-05, "loss": 4.1303, "step": 4631 }, { "epoch": 5.19, "learning_rate": 2.4890140845070423e-05, "loss": 4.2885, "step": 4632 }, { "epoch": 5.19, "learning_rate": 2.48887323943662e-05, "loss": 4.5202, "step": 4633 }, { "epoch": 5.2, "learning_rate": 2.4887323943661972e-05, "loss": 4.2143, "step": 4634 }, { "epoch": 5.2, "learning_rate": 2.488591549295775e-05, "loss": 4.2537, "step": 4635 }, { "epoch": 5.2, "learning_rate": 2.488450704225352e-05, "loss": 4.1915, "step": 4636 }, { "epoch": 5.2, "learning_rate": 2.4883098591549294e-05, "loss": 4.0755, "step": 4637 }, { "epoch": 5.2, "learning_rate": 2.488169014084507e-05, "loss": 4.2314, "step": 4638 }, { "epoch": 5.2, "learning_rate": 2.4880281690140844e-05, "loss": 4.0243, "step": 4639 }, { "epoch": 5.2, "learning_rate": 2.487887323943662e-05, "loss": 4.1255, "step": 4640 }, { "epoch": 5.2, "learning_rate": 2.4877464788732393e-05, "loss": 4.0216, "step": 4641 }, { "epoch": 5.2, "learning_rate": 2.487605633802817e-05, "loss": 4.205, "step": 4642 }, { "epoch": 5.21, "learning_rate": 2.4874647887323946e-05, "loss": 4.224, "step": 4643 }, { "epoch": 5.21, "learning_rate": 2.487323943661972e-05, "loss": 4.2059, "step": 4644 }, { "epoch": 5.21, "learning_rate": 2.4871830985915495e-05, "loss": 4.1086, "step": 4645 }, { "epoch": 5.21, "learning_rate": 2.4870422535211268e-05, "loss": 4.0094, "step": 4646 }, { "epoch": 5.21, "learning_rate": 2.4869014084507044e-05, "loss": 4.1503, "step": 4647 }, { "epoch": 5.21, "learning_rate": 2.4867605633802817e-05, "loss": 4.1736, "step": 4648 }, { "epoch": 5.21, "learning_rate": 2.4866197183098593e-05, "loss": 4.0309, "step": 4649 }, { "epoch": 5.21, "learning_rate": 2.4864788732394366e-05, "loss": 3.9635, "step": 4650 }, { "epoch": 5.21, "learning_rate": 2.4863380281690143e-05, "loss": 3.9651, "step": 4651 }, { "epoch": 5.22, "learning_rate": 2.4861971830985916e-05, "loss": 3.9507, "step": 4652 }, { "epoch": 5.22, "learning_rate": 2.486056338028169e-05, "loss": 3.8449, "step": 4653 }, { "epoch": 5.22, "learning_rate": 2.4859154929577465e-05, "loss": 3.708, "step": 4654 }, { "epoch": 5.22, "learning_rate": 2.4857746478873238e-05, "loss": 3.9665, "step": 4655 }, { "epoch": 5.22, "learning_rate": 2.4856338028169014e-05, "loss": 3.7656, "step": 4656 }, { "epoch": 5.22, "learning_rate": 2.4854929577464787e-05, "loss": 3.5953, "step": 4657 }, { "epoch": 5.22, "learning_rate": 2.4853521126760567e-05, "loss": 3.3628, "step": 4658 }, { "epoch": 5.22, "learning_rate": 2.485211267605634e-05, "loss": 3.4784, "step": 4659 }, { "epoch": 5.22, "learning_rate": 2.4850704225352113e-05, "loss": 3.8376, "step": 4660 }, { "epoch": 5.23, "learning_rate": 2.484929577464789e-05, "loss": 5.3596, "step": 4661 }, { "epoch": 5.23, "learning_rate": 2.4847887323943662e-05, "loss": 5.1607, "step": 4662 }, { "epoch": 5.23, "learning_rate": 2.484647887323944e-05, "loss": 5.1714, "step": 4663 }, { "epoch": 5.23, "learning_rate": 2.484507042253521e-05, "loss": 5.0269, "step": 4664 }, { "epoch": 5.23, "learning_rate": 2.4843661971830988e-05, "loss": 4.8215, "step": 4665 }, { "epoch": 5.23, "learning_rate": 2.484225352112676e-05, "loss": 4.8848, "step": 4666 }, { "epoch": 5.23, "learning_rate": 2.4840845070422537e-05, "loss": 4.5637, "step": 4667 }, { "epoch": 5.23, "learning_rate": 2.483943661971831e-05, "loss": 4.681, "step": 4668 }, { "epoch": 5.23, "learning_rate": 2.4838028169014083e-05, "loss": 4.5157, "step": 4669 }, { "epoch": 5.24, "learning_rate": 2.483661971830986e-05, "loss": 4.527, "step": 4670 }, { "epoch": 5.24, "learning_rate": 2.4835211267605632e-05, "loss": 4.148, "step": 4671 }, { "epoch": 5.24, "learning_rate": 2.483380281690141e-05, "loss": 4.4895, "step": 4672 }, { "epoch": 5.24, "learning_rate": 2.4832394366197185e-05, "loss": 4.4116, "step": 4673 }, { "epoch": 5.24, "learning_rate": 2.483098591549296e-05, "loss": 4.4635, "step": 4674 }, { "epoch": 5.24, "learning_rate": 2.4829577464788734e-05, "loss": 4.2091, "step": 4675 }, { "epoch": 5.24, "learning_rate": 2.4828169014084507e-05, "loss": 4.1104, "step": 4676 }, { "epoch": 5.24, "learning_rate": 2.4826760563380283e-05, "loss": 4.0761, "step": 4677 }, { "epoch": 5.24, "learning_rate": 2.4825352112676056e-05, "loss": 4.136, "step": 4678 }, { "epoch": 5.25, "learning_rate": 2.4823943661971833e-05, "loss": 4.2416, "step": 4679 }, { "epoch": 5.25, "learning_rate": 2.4822535211267606e-05, "loss": 4.1372, "step": 4680 }, { "epoch": 5.25, "learning_rate": 2.4821126760563382e-05, "loss": 4.4599, "step": 4681 }, { "epoch": 5.25, "learning_rate": 2.4819718309859155e-05, "loss": 4.4934, "step": 4682 }, { "epoch": 5.25, "learning_rate": 2.481830985915493e-05, "loss": 4.3268, "step": 4683 }, { "epoch": 5.25, "learning_rate": 2.4816901408450704e-05, "loss": 4.1015, "step": 4684 }, { "epoch": 5.25, "learning_rate": 2.4815492957746477e-05, "loss": 4.1813, "step": 4685 }, { "epoch": 5.25, "learning_rate": 2.4814084507042253e-05, "loss": 4.2299, "step": 4686 }, { "epoch": 5.25, "learning_rate": 2.4812676056338026e-05, "loss": 3.9813, "step": 4687 }, { "epoch": 5.26, "learning_rate": 2.4811267605633806e-05, "loss": 4.2897, "step": 4688 }, { "epoch": 5.26, "learning_rate": 2.480985915492958e-05, "loss": 4.2225, "step": 4689 }, { "epoch": 5.26, "learning_rate": 2.4808450704225355e-05, "loss": 4.2462, "step": 4690 }, { "epoch": 5.26, "learning_rate": 2.4807042253521128e-05, "loss": 4.162, "step": 4691 }, { "epoch": 5.26, "learning_rate": 2.48056338028169e-05, "loss": 4.5163, "step": 4692 }, { "epoch": 5.26, "learning_rate": 2.4804225352112678e-05, "loss": 3.9518, "step": 4693 }, { "epoch": 5.26, "learning_rate": 2.480281690140845e-05, "loss": 4.1034, "step": 4694 }, { "epoch": 5.26, "learning_rate": 2.4801408450704227e-05, "loss": 4.2296, "step": 4695 }, { "epoch": 5.26, "learning_rate": 2.48e-05, "loss": 4.1954, "step": 4696 }, { "epoch": 5.27, "learning_rate": 2.4798591549295776e-05, "loss": 4.1308, "step": 4697 }, { "epoch": 5.27, "learning_rate": 2.479718309859155e-05, "loss": 4.1578, "step": 4698 }, { "epoch": 5.27, "learning_rate": 2.4795774647887325e-05, "loss": 3.7449, "step": 4699 }, { "epoch": 5.27, "learning_rate": 2.4794366197183098e-05, "loss": 4.1651, "step": 4700 }, { "epoch": 5.27, "learning_rate": 2.479295774647887e-05, "loss": 4.2356, "step": 4701 }, { "epoch": 5.27, "learning_rate": 2.4791549295774648e-05, "loss": 3.7482, "step": 4702 }, { "epoch": 5.27, "learning_rate": 2.4790140845070424e-05, "loss": 3.8756, "step": 4703 }, { "epoch": 5.27, "learning_rate": 2.47887323943662e-05, "loss": 4.088, "step": 4704 }, { "epoch": 5.27, "learning_rate": 2.4787323943661973e-05, "loss": 3.8199, "step": 4705 }, { "epoch": 5.28, "learning_rate": 2.478591549295775e-05, "loss": 3.5607, "step": 4706 }, { "epoch": 5.28, "learning_rate": 2.4784507042253522e-05, "loss": 3.4516, "step": 4707 }, { "epoch": 5.28, "learning_rate": 2.4783098591549295e-05, "loss": 3.7031, "step": 4708 }, { "epoch": 5.28, "learning_rate": 2.4781690140845072e-05, "loss": 3.5856, "step": 4709 }, { "epoch": 5.28, "learning_rate": 2.4780281690140845e-05, "loss": 3.209, "step": 4710 }, { "epoch": 5.28, "learning_rate": 2.477887323943662e-05, "loss": 5.134, "step": 4711 }, { "epoch": 5.28, "learning_rate": 2.4777464788732394e-05, "loss": 5.075, "step": 4712 }, { "epoch": 5.28, "learning_rate": 2.477605633802817e-05, "loss": 4.902, "step": 4713 }, { "epoch": 5.28, "learning_rate": 2.4774647887323943e-05, "loss": 4.8853, "step": 4714 }, { "epoch": 5.29, "learning_rate": 2.477323943661972e-05, "loss": 4.6947, "step": 4715 }, { "epoch": 5.29, "learning_rate": 2.4771830985915492e-05, "loss": 4.8889, "step": 4716 }, { "epoch": 5.29, "learning_rate": 2.4770422535211265e-05, "loss": 4.6843, "step": 4717 }, { "epoch": 5.29, "learning_rate": 2.4769014084507045e-05, "loss": 4.3619, "step": 4718 }, { "epoch": 5.29, "learning_rate": 2.4767605633802818e-05, "loss": 4.492, "step": 4719 }, { "epoch": 5.29, "learning_rate": 2.4766197183098594e-05, "loss": 4.2434, "step": 4720 }, { "epoch": 5.29, "learning_rate": 2.4764788732394367e-05, "loss": 4.3966, "step": 4721 }, { "epoch": 5.29, "learning_rate": 2.4763380281690144e-05, "loss": 4.4935, "step": 4722 }, { "epoch": 5.29, "learning_rate": 2.4761971830985917e-05, "loss": 4.4618, "step": 4723 }, { "epoch": 5.3, "learning_rate": 2.476056338028169e-05, "loss": 4.2283, "step": 4724 }, { "epoch": 5.3, "learning_rate": 2.4759154929577466e-05, "loss": 4.3293, "step": 4725 }, { "epoch": 5.3, "learning_rate": 2.475774647887324e-05, "loss": 4.2158, "step": 4726 }, { "epoch": 5.3, "learning_rate": 2.4756338028169015e-05, "loss": 4.4622, "step": 4727 }, { "epoch": 5.3, "learning_rate": 2.4754929577464788e-05, "loss": 4.2442, "step": 4728 }, { "epoch": 5.3, "learning_rate": 2.4753521126760564e-05, "loss": 4.1969, "step": 4729 }, { "epoch": 5.3, "learning_rate": 2.4752112676056337e-05, "loss": 4.2549, "step": 4730 }, { "epoch": 5.3, "learning_rate": 2.4750704225352114e-05, "loss": 4.3229, "step": 4731 }, { "epoch": 5.3, "learning_rate": 2.4749295774647887e-05, "loss": 4.2482, "step": 4732 }, { "epoch": 5.31, "learning_rate": 2.474788732394366e-05, "loss": 4.0791, "step": 4733 }, { "epoch": 5.31, "learning_rate": 2.474647887323944e-05, "loss": 4.0869, "step": 4734 }, { "epoch": 5.31, "learning_rate": 2.4745070422535212e-05, "loss": 4.1902, "step": 4735 }, { "epoch": 5.31, "learning_rate": 2.474366197183099e-05, "loss": 4.3307, "step": 4736 }, { "epoch": 5.31, "learning_rate": 2.474225352112676e-05, "loss": 4.1978, "step": 4737 }, { "epoch": 5.31, "learning_rate": 2.4740845070422538e-05, "loss": 4.0053, "step": 4738 }, { "epoch": 5.31, "learning_rate": 2.473943661971831e-05, "loss": 4.1445, "step": 4739 }, { "epoch": 5.31, "learning_rate": 2.4738028169014084e-05, "loss": 4.2694, "step": 4740 }, { "epoch": 5.32, "learning_rate": 2.473661971830986e-05, "loss": 4.2849, "step": 4741 }, { "epoch": 5.32, "learning_rate": 2.4735211267605633e-05, "loss": 4.2264, "step": 4742 }, { "epoch": 5.32, "learning_rate": 2.473380281690141e-05, "loss": 4.1794, "step": 4743 }, { "epoch": 5.32, "learning_rate": 2.4732394366197182e-05, "loss": 4.189, "step": 4744 }, { "epoch": 5.32, "learning_rate": 2.473098591549296e-05, "loss": 4.0718, "step": 4745 }, { "epoch": 5.32, "learning_rate": 2.472957746478873e-05, "loss": 4.1875, "step": 4746 }, { "epoch": 5.32, "learning_rate": 2.4728169014084508e-05, "loss": 4.0729, "step": 4747 }, { "epoch": 5.32, "learning_rate": 2.472676056338028e-05, "loss": 4.2393, "step": 4748 }, { "epoch": 5.32, "learning_rate": 2.4725352112676057e-05, "loss": 4.0051, "step": 4749 }, { "epoch": 5.33, "learning_rate": 2.4723943661971834e-05, "loss": 3.9728, "step": 4750 }, { "epoch": 5.33, "learning_rate": 2.4722535211267606e-05, "loss": 4.0947, "step": 4751 }, { "epoch": 5.33, "learning_rate": 2.4721126760563383e-05, "loss": 3.7952, "step": 4752 }, { "epoch": 5.33, "learning_rate": 2.4719718309859156e-05, "loss": 3.8407, "step": 4753 }, { "epoch": 5.33, "learning_rate": 2.4718309859154932e-05, "loss": 3.721, "step": 4754 }, { "epoch": 5.33, "learning_rate": 2.4716901408450705e-05, "loss": 3.8303, "step": 4755 }, { "epoch": 5.33, "learning_rate": 2.4715492957746478e-05, "loss": 3.82, "step": 4756 }, { "epoch": 5.33, "learning_rate": 2.4714084507042254e-05, "loss": 3.6875, "step": 4757 }, { "epoch": 5.33, "learning_rate": 2.4712676056338027e-05, "loss": 3.4169, "step": 4758 }, { "epoch": 5.34, "learning_rate": 2.4711267605633804e-05, "loss": 3.5786, "step": 4759 }, { "epoch": 5.34, "learning_rate": 2.4709859154929577e-05, "loss": 2.8432, "step": 4760 }, { "epoch": 5.34, "learning_rate": 2.4708450704225353e-05, "loss": 5.1349, "step": 4761 }, { "epoch": 5.34, "learning_rate": 2.4707042253521126e-05, "loss": 5.2673, "step": 4762 }, { "epoch": 5.34, "learning_rate": 2.47056338028169e-05, "loss": 5.1259, "step": 4763 }, { "epoch": 5.34, "learning_rate": 2.470422535211268e-05, "loss": 4.9802, "step": 4764 }, { "epoch": 5.34, "learning_rate": 2.470281690140845e-05, "loss": 4.8268, "step": 4765 }, { "epoch": 5.34, "learning_rate": 2.4701408450704228e-05, "loss": 5.0426, "step": 4766 }, { "epoch": 5.34, "learning_rate": 2.47e-05, "loss": 4.5654, "step": 4767 }, { "epoch": 5.35, "learning_rate": 2.4698591549295777e-05, "loss": 4.5883, "step": 4768 }, { "epoch": 5.35, "learning_rate": 2.469718309859155e-05, "loss": 4.3003, "step": 4769 }, { "epoch": 5.35, "learning_rate": 2.4695774647887326e-05, "loss": 4.3592, "step": 4770 }, { "epoch": 5.35, "learning_rate": 2.46943661971831e-05, "loss": 4.2487, "step": 4771 }, { "epoch": 5.35, "learning_rate": 2.4692957746478872e-05, "loss": 4.3982, "step": 4772 }, { "epoch": 5.35, "learning_rate": 2.469154929577465e-05, "loss": 4.6553, "step": 4773 }, { "epoch": 5.35, "learning_rate": 2.469014084507042e-05, "loss": 4.4295, "step": 4774 }, { "epoch": 5.35, "learning_rate": 2.4688732394366198e-05, "loss": 4.2349, "step": 4775 }, { "epoch": 5.35, "learning_rate": 2.468732394366197e-05, "loss": 4.3864, "step": 4776 }, { "epoch": 5.36, "learning_rate": 2.4685915492957747e-05, "loss": 4.2188, "step": 4777 }, { "epoch": 5.36, "learning_rate": 2.468450704225352e-05, "loss": 4.0894, "step": 4778 }, { "epoch": 5.36, "learning_rate": 2.4683098591549296e-05, "loss": 4.3218, "step": 4779 }, { "epoch": 5.36, "learning_rate": 2.4681690140845073e-05, "loss": 4.2099, "step": 4780 }, { "epoch": 5.36, "learning_rate": 2.4680281690140846e-05, "loss": 4.3102, "step": 4781 }, { "epoch": 5.36, "learning_rate": 2.4678873239436622e-05, "loss": 4.2502, "step": 4782 }, { "epoch": 5.36, "learning_rate": 2.4677464788732395e-05, "loss": 3.9, "step": 4783 }, { "epoch": 5.36, "learning_rate": 2.467605633802817e-05, "loss": 4.1113, "step": 4784 }, { "epoch": 5.36, "learning_rate": 2.4674647887323944e-05, "loss": 4.0521, "step": 4785 }, { "epoch": 5.37, "learning_rate": 2.467323943661972e-05, "loss": 4.212, "step": 4786 }, { "epoch": 5.37, "learning_rate": 2.4671830985915493e-05, "loss": 4.165, "step": 4787 }, { "epoch": 5.37, "learning_rate": 2.4670422535211266e-05, "loss": 3.9865, "step": 4788 }, { "epoch": 5.37, "learning_rate": 2.4669014084507043e-05, "loss": 4.0467, "step": 4789 }, { "epoch": 5.37, "learning_rate": 2.4667605633802816e-05, "loss": 4.0373, "step": 4790 }, { "epoch": 5.37, "learning_rate": 2.4666197183098592e-05, "loss": 4.1226, "step": 4791 }, { "epoch": 5.37, "learning_rate": 2.4664788732394365e-05, "loss": 4.0377, "step": 4792 }, { "epoch": 5.37, "learning_rate": 2.466338028169014e-05, "loss": 4.0125, "step": 4793 }, { "epoch": 5.37, "learning_rate": 2.4661971830985918e-05, "loss": 3.9188, "step": 4794 }, { "epoch": 5.38, "learning_rate": 2.466056338028169e-05, "loss": 4.1644, "step": 4795 }, { "epoch": 5.38, "learning_rate": 2.4659154929577467e-05, "loss": 3.8308, "step": 4796 }, { "epoch": 5.38, "learning_rate": 2.465774647887324e-05, "loss": 3.9867, "step": 4797 }, { "epoch": 5.38, "learning_rate": 2.4656338028169016e-05, "loss": 3.8561, "step": 4798 }, { "epoch": 5.38, "learning_rate": 2.465492957746479e-05, "loss": 4.0219, "step": 4799 }, { "epoch": 5.38, "learning_rate": 2.4653521126760565e-05, "loss": 3.7464, "step": 4800 }, { "epoch": 5.38, "learning_rate": 2.465211267605634e-05, "loss": 3.9383, "step": 4801 }, { "epoch": 5.38, "learning_rate": 2.4650704225352115e-05, "loss": 3.8427, "step": 4802 }, { "epoch": 5.38, "learning_rate": 2.4649295774647888e-05, "loss": 3.7123, "step": 4803 }, { "epoch": 5.39, "learning_rate": 2.464788732394366e-05, "loss": 3.9672, "step": 4804 }, { "epoch": 5.39, "learning_rate": 2.4646478873239437e-05, "loss": 3.9982, "step": 4805 }, { "epoch": 5.39, "learning_rate": 2.464507042253521e-05, "loss": 3.7624, "step": 4806 }, { "epoch": 5.39, "learning_rate": 2.4643661971830986e-05, "loss": 3.3904, "step": 4807 }, { "epoch": 5.39, "learning_rate": 2.464225352112676e-05, "loss": 4.0025, "step": 4808 }, { "epoch": 5.39, "learning_rate": 2.464084507042254e-05, "loss": 3.3092, "step": 4809 }, { "epoch": 5.39, "learning_rate": 2.4639436619718312e-05, "loss": 2.9306, "step": 4810 }, { "epoch": 5.39, "learning_rate": 2.4638028169014085e-05, "loss": 5.2561, "step": 4811 }, { "epoch": 5.39, "learning_rate": 2.463661971830986e-05, "loss": 5.1258, "step": 4812 }, { "epoch": 5.4, "learning_rate": 2.4635211267605634e-05, "loss": 5.0519, "step": 4813 }, { "epoch": 5.4, "learning_rate": 2.463380281690141e-05, "loss": 5.2001, "step": 4814 }, { "epoch": 5.4, "learning_rate": 2.4632394366197183e-05, "loss": 4.812, "step": 4815 }, { "epoch": 5.4, "learning_rate": 2.463098591549296e-05, "loss": 4.6996, "step": 4816 }, { "epoch": 5.4, "learning_rate": 2.4629577464788733e-05, "loss": 4.5247, "step": 4817 }, { "epoch": 5.4, "learning_rate": 2.462816901408451e-05, "loss": 4.3261, "step": 4818 }, { "epoch": 5.4, "learning_rate": 2.4626760563380282e-05, "loss": 4.3001, "step": 4819 }, { "epoch": 5.4, "learning_rate": 2.4625352112676055e-05, "loss": 4.4169, "step": 4820 }, { "epoch": 5.4, "learning_rate": 2.462394366197183e-05, "loss": 4.2883, "step": 4821 }, { "epoch": 5.41, "learning_rate": 2.4622535211267604e-05, "loss": 4.2282, "step": 4822 }, { "epoch": 5.41, "learning_rate": 2.462112676056338e-05, "loss": 4.5256, "step": 4823 }, { "epoch": 5.41, "learning_rate": 2.4619718309859157e-05, "loss": 4.3816, "step": 4824 }, { "epoch": 5.41, "learning_rate": 2.4618309859154933e-05, "loss": 4.1438, "step": 4825 }, { "epoch": 5.41, "learning_rate": 2.4616901408450706e-05, "loss": 4.3709, "step": 4826 }, { "epoch": 5.41, "learning_rate": 2.461549295774648e-05, "loss": 4.4191, "step": 4827 }, { "epoch": 5.41, "learning_rate": 2.4614084507042255e-05, "loss": 4.203, "step": 4828 }, { "epoch": 5.41, "learning_rate": 2.4612676056338028e-05, "loss": 4.2163, "step": 4829 }, { "epoch": 5.41, "learning_rate": 2.4611267605633805e-05, "loss": 4.1596, "step": 4830 }, { "epoch": 5.42, "learning_rate": 2.4609859154929577e-05, "loss": 4.2884, "step": 4831 }, { "epoch": 5.42, "learning_rate": 2.4608450704225354e-05, "loss": 4.1539, "step": 4832 }, { "epoch": 5.42, "learning_rate": 2.4607042253521127e-05, "loss": 4.0992, "step": 4833 }, { "epoch": 5.42, "learning_rate": 2.4605633802816903e-05, "loss": 3.8978, "step": 4834 }, { "epoch": 5.42, "learning_rate": 2.4604225352112676e-05, "loss": 4.1001, "step": 4835 }, { "epoch": 5.42, "learning_rate": 2.460281690140845e-05, "loss": 4.251, "step": 4836 }, { "epoch": 5.42, "learning_rate": 2.4601408450704225e-05, "loss": 4.1142, "step": 4837 }, { "epoch": 5.42, "learning_rate": 2.4599999999999998e-05, "loss": 3.983, "step": 4838 }, { "epoch": 5.42, "learning_rate": 2.4598591549295778e-05, "loss": 3.9722, "step": 4839 }, { "epoch": 5.43, "learning_rate": 2.459718309859155e-05, "loss": 4.2524, "step": 4840 }, { "epoch": 5.43, "learning_rate": 2.4595774647887327e-05, "loss": 3.9139, "step": 4841 }, { "epoch": 5.43, "learning_rate": 2.45943661971831e-05, "loss": 3.8969, "step": 4842 }, { "epoch": 5.43, "learning_rate": 2.4592957746478873e-05, "loss": 3.9677, "step": 4843 }, { "epoch": 5.43, "learning_rate": 2.459154929577465e-05, "loss": 4.022, "step": 4844 }, { "epoch": 5.43, "learning_rate": 2.4590140845070422e-05, "loss": 4.1922, "step": 4845 }, { "epoch": 5.43, "learning_rate": 2.45887323943662e-05, "loss": 3.9275, "step": 4846 }, { "epoch": 5.43, "learning_rate": 2.458732394366197e-05, "loss": 3.8427, "step": 4847 }, { "epoch": 5.43, "learning_rate": 2.4585915492957748e-05, "loss": 3.8508, "step": 4848 }, { "epoch": 5.44, "learning_rate": 2.458450704225352e-05, "loss": 3.8406, "step": 4849 }, { "epoch": 5.44, "learning_rate": 2.4583098591549297e-05, "loss": 3.7417, "step": 4850 }, { "epoch": 5.44, "learning_rate": 2.458169014084507e-05, "loss": 4.0745, "step": 4851 }, { "epoch": 5.44, "learning_rate": 2.4580281690140843e-05, "loss": 3.8002, "step": 4852 }, { "epoch": 5.44, "learning_rate": 2.457887323943662e-05, "loss": 3.9155, "step": 4853 }, { "epoch": 5.44, "learning_rate": 2.4577464788732392e-05, "loss": 3.7018, "step": 4854 }, { "epoch": 5.44, "learning_rate": 2.4576056338028172e-05, "loss": 3.5881, "step": 4855 }, { "epoch": 5.44, "learning_rate": 2.4574647887323945e-05, "loss": 3.6656, "step": 4856 }, { "epoch": 5.45, "learning_rate": 2.457323943661972e-05, "loss": 3.6339, "step": 4857 }, { "epoch": 5.45, "learning_rate": 2.4571830985915494e-05, "loss": 3.22, "step": 4858 }, { "epoch": 5.45, "learning_rate": 2.4570422535211267e-05, "loss": 3.6768, "step": 4859 }, { "epoch": 5.45, "learning_rate": 2.4569014084507044e-05, "loss": 3.7178, "step": 4860 }, { "epoch": 5.45, "learning_rate": 2.4567605633802817e-05, "loss": 5.1511, "step": 4861 }, { "epoch": 5.45, "learning_rate": 2.4566197183098593e-05, "loss": 5.0291, "step": 4862 }, { "epoch": 5.45, "learning_rate": 2.4564788732394366e-05, "loss": 5.1045, "step": 4863 }, { "epoch": 5.45, "learning_rate": 2.4563380281690142e-05, "loss": 4.748, "step": 4864 }, { "epoch": 5.45, "learning_rate": 2.4561971830985915e-05, "loss": 4.6915, "step": 4865 }, { "epoch": 5.46, "learning_rate": 2.4560563380281688e-05, "loss": 4.6066, "step": 4866 }, { "epoch": 5.46, "learning_rate": 2.4559154929577464e-05, "loss": 4.4134, "step": 4867 }, { "epoch": 5.46, "learning_rate": 2.4557746478873237e-05, "loss": 4.3013, "step": 4868 }, { "epoch": 5.46, "learning_rate": 2.4556338028169014e-05, "loss": 4.2143, "step": 4869 }, { "epoch": 5.46, "learning_rate": 2.455492957746479e-05, "loss": 4.4263, "step": 4870 }, { "epoch": 5.46, "learning_rate": 2.4553521126760566e-05, "loss": 4.2612, "step": 4871 }, { "epoch": 5.46, "learning_rate": 2.455211267605634e-05, "loss": 4.1951, "step": 4872 }, { "epoch": 5.46, "learning_rate": 2.4550704225352116e-05, "loss": 4.1402, "step": 4873 }, { "epoch": 5.46, "learning_rate": 2.454929577464789e-05, "loss": 4.2851, "step": 4874 }, { "epoch": 5.47, "learning_rate": 2.454788732394366e-05, "loss": 4.1005, "step": 4875 }, { "epoch": 5.47, "learning_rate": 2.4546478873239438e-05, "loss": 4.391, "step": 4876 }, { "epoch": 5.47, "learning_rate": 2.454507042253521e-05, "loss": 4.2119, "step": 4877 }, { "epoch": 5.47, "learning_rate": 2.4543661971830987e-05, "loss": 4.262, "step": 4878 }, { "epoch": 5.47, "learning_rate": 2.454225352112676e-05, "loss": 4.0468, "step": 4879 }, { "epoch": 5.47, "learning_rate": 2.4540845070422536e-05, "loss": 4.3569, "step": 4880 }, { "epoch": 5.47, "learning_rate": 2.453943661971831e-05, "loss": 4.1075, "step": 4881 }, { "epoch": 5.47, "learning_rate": 2.4538028169014082e-05, "loss": 4.0553, "step": 4882 }, { "epoch": 5.47, "learning_rate": 2.453661971830986e-05, "loss": 4.1433, "step": 4883 }, { "epoch": 5.48, "learning_rate": 2.453521126760563e-05, "loss": 4.2237, "step": 4884 }, { "epoch": 5.48, "learning_rate": 2.453380281690141e-05, "loss": 4.1379, "step": 4885 }, { "epoch": 5.48, "learning_rate": 2.4532394366197184e-05, "loss": 3.9542, "step": 4886 }, { "epoch": 5.48, "learning_rate": 2.453098591549296e-05, "loss": 4.1651, "step": 4887 }, { "epoch": 5.48, "learning_rate": 2.4529577464788734e-05, "loss": 3.9599, "step": 4888 }, { "epoch": 5.48, "learning_rate": 2.452816901408451e-05, "loss": 3.9256, "step": 4889 }, { "epoch": 5.48, "learning_rate": 2.4526760563380283e-05, "loss": 4.1189, "step": 4890 }, { "epoch": 5.48, "learning_rate": 2.4525352112676056e-05, "loss": 3.8898, "step": 4891 }, { "epoch": 5.48, "learning_rate": 2.4523943661971832e-05, "loss": 3.7968, "step": 4892 }, { "epoch": 5.49, "learning_rate": 2.4522535211267605e-05, "loss": 3.9123, "step": 4893 }, { "epoch": 5.49, "learning_rate": 2.452112676056338e-05, "loss": 3.9743, "step": 4894 }, { "epoch": 5.49, "learning_rate": 2.4519718309859154e-05, "loss": 3.8728, "step": 4895 }, { "epoch": 5.49, "learning_rate": 2.451830985915493e-05, "loss": 4.0503, "step": 4896 }, { "epoch": 5.49, "learning_rate": 2.4516901408450704e-05, "loss": 3.74, "step": 4897 }, { "epoch": 5.49, "learning_rate": 2.4515492957746477e-05, "loss": 3.8018, "step": 4898 }, { "epoch": 5.49, "learning_rate": 2.4514084507042253e-05, "loss": 3.7016, "step": 4899 }, { "epoch": 5.49, "learning_rate": 2.451267605633803e-05, "loss": 3.9504, "step": 4900 }, { "epoch": 5.49, "learning_rate": 2.4511267605633805e-05, "loss": 3.7613, "step": 4901 }, { "epoch": 5.5, "learning_rate": 2.450985915492958e-05, "loss": 3.8323, "step": 4902 }, { "epoch": 5.5, "learning_rate": 2.4508450704225355e-05, "loss": 3.8122, "step": 4903 }, { "epoch": 5.5, "learning_rate": 2.4507042253521128e-05, "loss": 3.6966, "step": 4904 }, { "epoch": 5.5, "learning_rate": 2.4505633802816904e-05, "loss": 3.1552, "step": 4905 }, { "epoch": 5.5, "learning_rate": 2.4504225352112677e-05, "loss": 3.5123, "step": 4906 }, { "epoch": 5.5, "learning_rate": 2.450281690140845e-05, "loss": 3.2232, "step": 4907 }, { "epoch": 5.5, "learning_rate": 2.4501408450704226e-05, "loss": 3.1765, "step": 4908 }, { "epoch": 5.5, "learning_rate": 2.45e-05, "loss": 3.4051, "step": 4909 }, { "epoch": 5.5, "learning_rate": 2.4498591549295776e-05, "loss": 2.7972, "step": 4910 }, { "epoch": 5.51, "learning_rate": 2.449718309859155e-05, "loss": 5.2149, "step": 4911 }, { "epoch": 5.51, "learning_rate": 2.4495774647887325e-05, "loss": 5.1626, "step": 4912 }, { "epoch": 5.51, "learning_rate": 2.4494366197183098e-05, "loss": 4.9578, "step": 4913 }, { "epoch": 5.51, "learning_rate": 2.449295774647887e-05, "loss": 5.0559, "step": 4914 }, { "epoch": 5.51, "learning_rate": 2.449154929577465e-05, "loss": 4.8697, "step": 4915 }, { "epoch": 5.51, "learning_rate": 2.4490140845070423e-05, "loss": 4.7343, "step": 4916 }, { "epoch": 5.51, "learning_rate": 2.44887323943662e-05, "loss": 4.6573, "step": 4917 }, { "epoch": 5.51, "learning_rate": 2.4487323943661973e-05, "loss": 4.3713, "step": 4918 }, { "epoch": 5.51, "learning_rate": 2.448591549295775e-05, "loss": 4.1661, "step": 4919 }, { "epoch": 5.52, "learning_rate": 2.4484507042253522e-05, "loss": 4.1634, "step": 4920 }, { "epoch": 5.52, "learning_rate": 2.4483098591549298e-05, "loss": 4.2291, "step": 4921 }, { "epoch": 5.52, "learning_rate": 2.448169014084507e-05, "loss": 4.3146, "step": 4922 }, { "epoch": 5.52, "learning_rate": 2.4480281690140844e-05, "loss": 4.1965, "step": 4923 }, { "epoch": 5.52, "learning_rate": 2.447887323943662e-05, "loss": 4.2426, "step": 4924 }, { "epoch": 5.52, "learning_rate": 2.4477464788732393e-05, "loss": 4.1001, "step": 4925 }, { "epoch": 5.52, "learning_rate": 2.447605633802817e-05, "loss": 4.1212, "step": 4926 }, { "epoch": 5.52, "learning_rate": 2.4474647887323943e-05, "loss": 4.124, "step": 4927 }, { "epoch": 5.52, "learning_rate": 2.447323943661972e-05, "loss": 4.2481, "step": 4928 }, { "epoch": 5.53, "learning_rate": 2.4471830985915492e-05, "loss": 4.0192, "step": 4929 }, { "epoch": 5.53, "learning_rate": 2.4470422535211268e-05, "loss": 4.2763, "step": 4930 }, { "epoch": 5.53, "learning_rate": 2.4469014084507045e-05, "loss": 3.8975, "step": 4931 }, { "epoch": 5.53, "learning_rate": 2.4467605633802818e-05, "loss": 4.2057, "step": 4932 }, { "epoch": 5.53, "learning_rate": 2.4466197183098594e-05, "loss": 3.9265, "step": 4933 }, { "epoch": 5.53, "learning_rate": 2.4464788732394367e-05, "loss": 4.0758, "step": 4934 }, { "epoch": 5.53, "learning_rate": 2.4463380281690143e-05, "loss": 4.0735, "step": 4935 }, { "epoch": 5.53, "learning_rate": 2.4461971830985916e-05, "loss": 4.2196, "step": 4936 }, { "epoch": 5.53, "learning_rate": 2.4460563380281692e-05, "loss": 3.8827, "step": 4937 }, { "epoch": 5.54, "learning_rate": 2.4459154929577465e-05, "loss": 4.0607, "step": 4938 }, { "epoch": 5.54, "learning_rate": 2.445774647887324e-05, "loss": 3.8866, "step": 4939 }, { "epoch": 5.54, "learning_rate": 2.4456338028169015e-05, "loss": 4.1534, "step": 4940 }, { "epoch": 5.54, "learning_rate": 2.4454929577464788e-05, "loss": 3.9622, "step": 4941 }, { "epoch": 5.54, "learning_rate": 2.4453521126760564e-05, "loss": 3.7653, "step": 4942 }, { "epoch": 5.54, "learning_rate": 2.4452112676056337e-05, "loss": 3.8662, "step": 4943 }, { "epoch": 5.54, "learning_rate": 2.4450704225352113e-05, "loss": 3.8741, "step": 4944 }, { "epoch": 5.54, "learning_rate": 2.444929577464789e-05, "loss": 3.9898, "step": 4945 }, { "epoch": 5.54, "learning_rate": 2.4447887323943662e-05, "loss": 3.8565, "step": 4946 }, { "epoch": 5.55, "learning_rate": 2.444647887323944e-05, "loss": 3.6008, "step": 4947 }, { "epoch": 5.55, "learning_rate": 2.4445070422535212e-05, "loss": 3.7099, "step": 4948 }, { "epoch": 5.55, "learning_rate": 2.4443661971830988e-05, "loss": 3.6069, "step": 4949 }, { "epoch": 5.55, "learning_rate": 2.444225352112676e-05, "loss": 3.9441, "step": 4950 }, { "epoch": 5.55, "learning_rate": 2.4440845070422537e-05, "loss": 3.7236, "step": 4951 }, { "epoch": 5.55, "learning_rate": 2.443943661971831e-05, "loss": 3.7063, "step": 4952 }, { "epoch": 5.55, "learning_rate": 2.4438028169014087e-05, "loss": 3.6128, "step": 4953 }, { "epoch": 5.55, "learning_rate": 2.443661971830986e-05, "loss": 3.7266, "step": 4954 }, { "epoch": 5.55, "learning_rate": 2.4435211267605633e-05, "loss": 3.4546, "step": 4955 }, { "epoch": 5.56, "learning_rate": 2.443380281690141e-05, "loss": 3.4511, "step": 4956 }, { "epoch": 5.56, "learning_rate": 2.4432394366197182e-05, "loss": 3.2852, "step": 4957 }, { "epoch": 5.56, "learning_rate": 2.4430985915492958e-05, "loss": 3.5884, "step": 4958 }, { "epoch": 5.56, "learning_rate": 2.442957746478873e-05, "loss": 3.4117, "step": 4959 }, { "epoch": 5.56, "learning_rate": 2.442816901408451e-05, "loss": 3.0155, "step": 4960 }, { "epoch": 5.56, "learning_rate": 2.4426760563380284e-05, "loss": 5.152, "step": 4961 }, { "epoch": 5.56, "learning_rate": 2.4425352112676057e-05, "loss": 5.0012, "step": 4962 }, { "epoch": 5.56, "learning_rate": 2.4423943661971833e-05, "loss": 4.8071, "step": 4963 }, { "epoch": 5.57, "learning_rate": 2.4422535211267606e-05, "loss": 4.8056, "step": 4964 }, { "epoch": 5.57, "learning_rate": 2.4421126760563382e-05, "loss": 4.7273, "step": 4965 }, { "epoch": 5.57, "learning_rate": 2.4419718309859155e-05, "loss": 4.5205, "step": 4966 }, { "epoch": 5.57, "learning_rate": 2.441830985915493e-05, "loss": 4.4241, "step": 4967 }, { "epoch": 5.57, "learning_rate": 2.4416901408450705e-05, "loss": 4.5379, "step": 4968 }, { "epoch": 5.57, "learning_rate": 2.441549295774648e-05, "loss": 4.3153, "step": 4969 }, { "epoch": 5.57, "learning_rate": 2.4414084507042254e-05, "loss": 4.5608, "step": 4970 }, { "epoch": 5.57, "learning_rate": 2.4412676056338027e-05, "loss": 4.2939, "step": 4971 }, { "epoch": 5.57, "learning_rate": 2.4411267605633803e-05, "loss": 4.1505, "step": 4972 }, { "epoch": 5.58, "learning_rate": 2.4409859154929576e-05, "loss": 4.239, "step": 4973 }, { "epoch": 5.58, "learning_rate": 2.4408450704225352e-05, "loss": 4.2594, "step": 4974 }, { "epoch": 5.58, "learning_rate": 2.4407042253521125e-05, "loss": 4.3118, "step": 4975 }, { "epoch": 5.58, "learning_rate": 2.4405633802816905e-05, "loss": 4.1442, "step": 4976 }, { "epoch": 5.58, "learning_rate": 2.4404225352112678e-05, "loss": 4.0096, "step": 4977 }, { "epoch": 5.58, "learning_rate": 2.440281690140845e-05, "loss": 4.0561, "step": 4978 }, { "epoch": 5.58, "learning_rate": 2.4401408450704227e-05, "loss": 4.1218, "step": 4979 }, { "epoch": 5.58, "learning_rate": 2.44e-05, "loss": 4.0756, "step": 4980 }, { "epoch": 5.58, "learning_rate": 2.4398591549295776e-05, "loss": 4.1877, "step": 4981 }, { "epoch": 5.59, "learning_rate": 2.439718309859155e-05, "loss": 4.1154, "step": 4982 }, { "epoch": 5.59, "learning_rate": 2.4395774647887326e-05, "loss": 4.0746, "step": 4983 }, { "epoch": 5.59, "learning_rate": 2.43943661971831e-05, "loss": 3.8163, "step": 4984 }, { "epoch": 5.59, "learning_rate": 2.439295774647887e-05, "loss": 4.1105, "step": 4985 }, { "epoch": 5.59, "learning_rate": 2.4391549295774648e-05, "loss": 4.0595, "step": 4986 }, { "epoch": 5.59, "learning_rate": 2.439014084507042e-05, "loss": 4.1911, "step": 4987 }, { "epoch": 5.59, "learning_rate": 2.4388732394366197e-05, "loss": 4.0296, "step": 4988 }, { "epoch": 5.59, "learning_rate": 2.438732394366197e-05, "loss": 4.0925, "step": 4989 }, { "epoch": 5.59, "learning_rate": 2.4385915492957747e-05, "loss": 3.9368, "step": 4990 }, { "epoch": 5.6, "learning_rate": 2.4384507042253523e-05, "loss": 3.9593, "step": 4991 }, { "epoch": 5.6, "learning_rate": 2.43830985915493e-05, "loss": 3.7112, "step": 4992 }, { "epoch": 5.6, "learning_rate": 2.4381690140845072e-05, "loss": 3.8616, "step": 4993 }, { "epoch": 5.6, "learning_rate": 2.4380281690140845e-05, "loss": 3.9836, "step": 4994 }, { "epoch": 5.6, "learning_rate": 2.437887323943662e-05, "loss": 3.6975, "step": 4995 }, { "epoch": 5.6, "learning_rate": 2.4377464788732394e-05, "loss": 3.8224, "step": 4996 }, { "epoch": 5.6, "learning_rate": 2.437605633802817e-05, "loss": 3.8715, "step": 4997 }, { "epoch": 5.6, "learning_rate": 2.4374647887323944e-05, "loss": 4.0222, "step": 4998 }, { "epoch": 5.6, "learning_rate": 2.437323943661972e-05, "loss": 3.6565, "step": 4999 }, { "epoch": 5.61, "learning_rate": 2.4371830985915493e-05, "loss": 3.7058, "step": 5000 }, { "epoch": 5.61, "learning_rate": 2.4370422535211266e-05, "loss": 3.641, "step": 5001 }, { "epoch": 5.61, "learning_rate": 2.4369014084507042e-05, "loss": 3.9482, "step": 5002 }, { "epoch": 5.61, "learning_rate": 2.4367605633802815e-05, "loss": 3.7642, "step": 5003 }, { "epoch": 5.61, "learning_rate": 2.436619718309859e-05, "loss": 3.5764, "step": 5004 }, { "epoch": 5.61, "learning_rate": 2.4364788732394364e-05, "loss": 3.559, "step": 5005 }, { "epoch": 5.61, "learning_rate": 2.4363380281690144e-05, "loss": 3.4822, "step": 5006 }, { "epoch": 5.61, "learning_rate": 2.4361971830985917e-05, "loss": 3.5247, "step": 5007 }, { "epoch": 5.61, "learning_rate": 2.4360563380281693e-05, "loss": 3.3135, "step": 5008 }, { "epoch": 5.62, "learning_rate": 2.4359154929577466e-05, "loss": 3.4929, "step": 5009 }, { "epoch": 5.62, "learning_rate": 2.435774647887324e-05, "loss": 3.4262, "step": 5010 }, { "epoch": 5.62, "learning_rate": 2.4356338028169016e-05, "loss": 5.0657, "step": 5011 }, { "epoch": 5.62, "learning_rate": 2.435492957746479e-05, "loss": 4.7491, "step": 5012 }, { "epoch": 5.62, "learning_rate": 2.4353521126760565e-05, "loss": 4.8214, "step": 5013 }, { "epoch": 5.62, "learning_rate": 2.4352112676056338e-05, "loss": 4.5989, "step": 5014 }, { "epoch": 5.62, "learning_rate": 2.4350704225352114e-05, "loss": 4.4556, "step": 5015 }, { "epoch": 5.62, "learning_rate": 2.4349295774647887e-05, "loss": 4.2842, "step": 5016 }, { "epoch": 5.62, "learning_rate": 2.434788732394366e-05, "loss": 4.3291, "step": 5017 }, { "epoch": 5.63, "learning_rate": 2.4346478873239436e-05, "loss": 4.3404, "step": 5018 }, { "epoch": 5.63, "learning_rate": 2.434507042253521e-05, "loss": 3.9853, "step": 5019 }, { "epoch": 5.63, "learning_rate": 2.4343661971830986e-05, "loss": 4.3611, "step": 5020 }, { "epoch": 5.63, "learning_rate": 2.4342253521126762e-05, "loss": 4.2163, "step": 5021 }, { "epoch": 5.63, "learning_rate": 2.434084507042254e-05, "loss": 4.2896, "step": 5022 }, { "epoch": 5.63, "learning_rate": 2.433943661971831e-05, "loss": 4.1822, "step": 5023 }, { "epoch": 5.63, "learning_rate": 2.4338028169014088e-05, "loss": 3.9191, "step": 5024 }, { "epoch": 5.63, "learning_rate": 2.433661971830986e-05, "loss": 4.0231, "step": 5025 }, { "epoch": 5.63, "learning_rate": 2.4335211267605633e-05, "loss": 4.1093, "step": 5026 }, { "epoch": 5.64, "learning_rate": 2.433380281690141e-05, "loss": 4.3343, "step": 5027 }, { "epoch": 5.64, "learning_rate": 2.4332394366197183e-05, "loss": 3.8167, "step": 5028 }, { "epoch": 5.64, "learning_rate": 2.433098591549296e-05, "loss": 4.1198, "step": 5029 }, { "epoch": 5.64, "learning_rate": 2.4329577464788732e-05, "loss": 4.131, "step": 5030 }, { "epoch": 5.64, "learning_rate": 2.432816901408451e-05, "loss": 3.8388, "step": 5031 }, { "epoch": 5.64, "learning_rate": 2.432676056338028e-05, "loss": 4.1103, "step": 5032 }, { "epoch": 5.64, "learning_rate": 2.4325352112676054e-05, "loss": 3.9452, "step": 5033 }, { "epoch": 5.64, "learning_rate": 2.432394366197183e-05, "loss": 3.8659, "step": 5034 }, { "epoch": 5.64, "learning_rate": 2.4322535211267604e-05, "loss": 4.1364, "step": 5035 }, { "epoch": 5.65, "learning_rate": 2.4321126760563383e-05, "loss": 3.9103, "step": 5036 }, { "epoch": 5.65, "learning_rate": 2.4319718309859156e-05, "loss": 3.8219, "step": 5037 }, { "epoch": 5.65, "learning_rate": 2.4318309859154933e-05, "loss": 3.9381, "step": 5038 }, { "epoch": 5.65, "learning_rate": 2.4316901408450705e-05, "loss": 3.6976, "step": 5039 }, { "epoch": 5.65, "learning_rate": 2.4315492957746482e-05, "loss": 3.8953, "step": 5040 }, { "epoch": 5.65, "learning_rate": 2.4314084507042255e-05, "loss": 3.7336, "step": 5041 }, { "epoch": 5.65, "learning_rate": 2.4312676056338028e-05, "loss": 3.7851, "step": 5042 }, { "epoch": 5.65, "learning_rate": 2.4311267605633804e-05, "loss": 3.9642, "step": 5043 }, { "epoch": 5.65, "learning_rate": 2.4309859154929577e-05, "loss": 3.8595, "step": 5044 }, { "epoch": 5.66, "learning_rate": 2.4308450704225353e-05, "loss": 3.8485, "step": 5045 }, { "epoch": 5.66, "learning_rate": 2.4307042253521126e-05, "loss": 3.6739, "step": 5046 }, { "epoch": 5.66, "learning_rate": 2.4305633802816903e-05, "loss": 3.8458, "step": 5047 }, { "epoch": 5.66, "learning_rate": 2.4304225352112676e-05, "loss": 3.885, "step": 5048 }, { "epoch": 5.66, "learning_rate": 2.430281690140845e-05, "loss": 3.7322, "step": 5049 }, { "epoch": 5.66, "learning_rate": 2.4301408450704225e-05, "loss": 3.638, "step": 5050 }, { "epoch": 5.66, "learning_rate": 2.43e-05, "loss": 3.7113, "step": 5051 }, { "epoch": 5.66, "learning_rate": 2.4298591549295777e-05, "loss": 3.6373, "step": 5052 }, { "epoch": 5.66, "learning_rate": 2.429718309859155e-05, "loss": 3.8107, "step": 5053 }, { "epoch": 5.67, "learning_rate": 2.4295774647887327e-05, "loss": 3.4683, "step": 5054 }, { "epoch": 5.67, "learning_rate": 2.42943661971831e-05, "loss": 3.1562, "step": 5055 }, { "epoch": 5.67, "learning_rate": 2.4292957746478876e-05, "loss": 3.6171, "step": 5056 }, { "epoch": 5.67, "learning_rate": 2.429154929577465e-05, "loss": 3.3495, "step": 5057 }, { "epoch": 5.67, "learning_rate": 2.4290140845070422e-05, "loss": 3.4204, "step": 5058 }, { "epoch": 5.67, "learning_rate": 2.4288732394366198e-05, "loss": 3.034, "step": 5059 }, { "epoch": 5.67, "learning_rate": 2.428732394366197e-05, "loss": 2.9802, "step": 5060 }, { "epoch": 5.67, "learning_rate": 2.4285915492957747e-05, "loss": 5.0752, "step": 5061 }, { "epoch": 5.67, "learning_rate": 2.428450704225352e-05, "loss": 4.9958, "step": 5062 }, { "epoch": 5.68, "learning_rate": 2.4283098591549297e-05, "loss": 4.9892, "step": 5063 }, { "epoch": 5.68, "learning_rate": 2.428169014084507e-05, "loss": 4.7825, "step": 5064 }, { "epoch": 5.68, "learning_rate": 2.4280281690140843e-05, "loss": 4.5577, "step": 5065 }, { "epoch": 5.68, "learning_rate": 2.4278873239436622e-05, "loss": 4.5881, "step": 5066 }, { "epoch": 5.68, "learning_rate": 2.4277464788732395e-05, "loss": 4.377, "step": 5067 }, { "epoch": 5.68, "learning_rate": 2.427605633802817e-05, "loss": 4.316, "step": 5068 }, { "epoch": 5.68, "learning_rate": 2.4274647887323945e-05, "loss": 4.1193, "step": 5069 }, { "epoch": 5.68, "learning_rate": 2.427323943661972e-05, "loss": 3.9219, "step": 5070 }, { "epoch": 5.68, "learning_rate": 2.4271830985915494e-05, "loss": 4.1482, "step": 5071 }, { "epoch": 5.69, "learning_rate": 2.427042253521127e-05, "loss": 4.2801, "step": 5072 }, { "epoch": 5.69, "learning_rate": 2.4269014084507043e-05, "loss": 4.254, "step": 5073 }, { "epoch": 5.69, "learning_rate": 2.4267605633802816e-05, "loss": 4.2358, "step": 5074 }, { "epoch": 5.69, "learning_rate": 2.4266197183098592e-05, "loss": 4.1682, "step": 5075 }, { "epoch": 5.69, "learning_rate": 2.4264788732394365e-05, "loss": 4.0874, "step": 5076 }, { "epoch": 5.69, "learning_rate": 2.4263380281690142e-05, "loss": 3.9502, "step": 5077 }, { "epoch": 5.69, "learning_rate": 2.4261971830985915e-05, "loss": 4.1371, "step": 5078 }, { "epoch": 5.69, "learning_rate": 2.426056338028169e-05, "loss": 3.9344, "step": 5079 }, { "epoch": 5.7, "learning_rate": 2.4259154929577464e-05, "loss": 3.9915, "step": 5080 }, { "epoch": 5.7, "learning_rate": 2.425774647887324e-05, "loss": 3.9574, "step": 5081 }, { "epoch": 5.7, "learning_rate": 2.4256338028169017e-05, "loss": 3.8709, "step": 5082 }, { "epoch": 5.7, "learning_rate": 2.425492957746479e-05, "loss": 4.0405, "step": 5083 }, { "epoch": 5.7, "learning_rate": 2.4253521126760566e-05, "loss": 3.8843, "step": 5084 }, { "epoch": 5.7, "learning_rate": 2.425211267605634e-05, "loss": 3.979, "step": 5085 }, { "epoch": 5.7, "learning_rate": 2.4250704225352115e-05, "loss": 3.9554, "step": 5086 }, { "epoch": 5.7, "learning_rate": 2.4249295774647888e-05, "loss": 3.9154, "step": 5087 }, { "epoch": 5.7, "learning_rate": 2.424788732394366e-05, "loss": 3.993, "step": 5088 }, { "epoch": 5.71, "learning_rate": 2.4246478873239437e-05, "loss": 4.001, "step": 5089 }, { "epoch": 5.71, "learning_rate": 2.424507042253521e-05, "loss": 3.871, "step": 5090 }, { "epoch": 5.71, "learning_rate": 2.4243661971830987e-05, "loss": 3.9607, "step": 5091 }, { "epoch": 5.71, "learning_rate": 2.424225352112676e-05, "loss": 3.6451, "step": 5092 }, { "epoch": 5.71, "learning_rate": 2.4240845070422536e-05, "loss": 3.6168, "step": 5093 }, { "epoch": 5.71, "learning_rate": 2.423943661971831e-05, "loss": 3.8628, "step": 5094 }, { "epoch": 5.71, "learning_rate": 2.4238028169014085e-05, "loss": 3.8508, "step": 5095 }, { "epoch": 5.71, "learning_rate": 2.4236619718309858e-05, "loss": 3.6365, "step": 5096 }, { "epoch": 5.71, "learning_rate": 2.4235211267605634e-05, "loss": 3.9044, "step": 5097 }, { "epoch": 5.72, "learning_rate": 2.423380281690141e-05, "loss": 3.7977, "step": 5098 }, { "epoch": 5.72, "learning_rate": 2.4232394366197184e-05, "loss": 3.569, "step": 5099 }, { "epoch": 5.72, "learning_rate": 2.423098591549296e-05, "loss": 3.8248, "step": 5100 }, { "epoch": 5.72, "learning_rate": 2.4229577464788733e-05, "loss": 3.6216, "step": 5101 }, { "epoch": 5.72, "learning_rate": 2.422816901408451e-05, "loss": 3.6538, "step": 5102 }, { "epoch": 5.72, "learning_rate": 2.4226760563380282e-05, "loss": 3.6525, "step": 5103 }, { "epoch": 5.72, "learning_rate": 2.4225352112676055e-05, "loss": 3.4743, "step": 5104 }, { "epoch": 5.72, "learning_rate": 2.422394366197183e-05, "loss": 3.5715, "step": 5105 }, { "epoch": 5.72, "learning_rate": 2.4222535211267604e-05, "loss": 3.6023, "step": 5106 }, { "epoch": 5.73, "learning_rate": 2.422112676056338e-05, "loss": 3.2675, "step": 5107 }, { "epoch": 5.73, "learning_rate": 2.4219718309859154e-05, "loss": 3.6872, "step": 5108 }, { "epoch": 5.73, "learning_rate": 2.421830985915493e-05, "loss": 3.4163, "step": 5109 }, { "epoch": 5.73, "learning_rate": 2.4216901408450703e-05, "loss": 3.3757, "step": 5110 }, { "epoch": 5.73, "learning_rate": 2.421549295774648e-05, "loss": 4.9276, "step": 5111 }, { "epoch": 5.73, "learning_rate": 2.4214084507042256e-05, "loss": 5.0382, "step": 5112 }, { "epoch": 5.73, "learning_rate": 2.421267605633803e-05, "loss": 4.9112, "step": 5113 }, { "epoch": 5.73, "learning_rate": 2.4211267605633805e-05, "loss": 4.6963, "step": 5114 }, { "epoch": 5.73, "learning_rate": 2.4209859154929578e-05, "loss": 4.5422, "step": 5115 }, { "epoch": 5.74, "learning_rate": 2.4208450704225354e-05, "loss": 4.3341, "step": 5116 }, { "epoch": 5.74, "learning_rate": 2.4207042253521127e-05, "loss": 4.3975, "step": 5117 }, { "epoch": 5.74, "learning_rate": 2.4205633802816904e-05, "loss": 4.2519, "step": 5118 }, { "epoch": 5.74, "learning_rate": 2.4204225352112676e-05, "loss": 4.0381, "step": 5119 }, { "epoch": 5.74, "learning_rate": 2.420281690140845e-05, "loss": 3.9928, "step": 5120 }, { "epoch": 5.74, "learning_rate": 2.4201408450704226e-05, "loss": 4.0847, "step": 5121 }, { "epoch": 5.74, "learning_rate": 2.42e-05, "loss": 3.9772, "step": 5122 }, { "epoch": 5.74, "learning_rate": 2.4198591549295775e-05, "loss": 4.0742, "step": 5123 }, { "epoch": 5.74, "learning_rate": 2.4197183098591548e-05, "loss": 4.0156, "step": 5124 }, { "epoch": 5.75, "learning_rate": 2.4195774647887324e-05, "loss": 4.0368, "step": 5125 }, { "epoch": 5.75, "learning_rate": 2.4194366197183097e-05, "loss": 3.9326, "step": 5126 }, { "epoch": 5.75, "learning_rate": 2.4192957746478877e-05, "loss": 3.9479, "step": 5127 }, { "epoch": 5.75, "learning_rate": 2.419154929577465e-05, "loss": 4.0254, "step": 5128 }, { "epoch": 5.75, "learning_rate": 2.4190140845070423e-05, "loss": 3.8436, "step": 5129 }, { "epoch": 5.75, "learning_rate": 2.41887323943662e-05, "loss": 4.0395, "step": 5130 }, { "epoch": 5.75, "learning_rate": 2.4187323943661972e-05, "loss": 3.9754, "step": 5131 }, { "epoch": 5.75, "learning_rate": 2.418591549295775e-05, "loss": 4.0155, "step": 5132 }, { "epoch": 5.75, "learning_rate": 2.418450704225352e-05, "loss": 3.7747, "step": 5133 }, { "epoch": 5.76, "learning_rate": 2.4183098591549298e-05, "loss": 3.99, "step": 5134 }, { "epoch": 5.76, "learning_rate": 2.418169014084507e-05, "loss": 3.6149, "step": 5135 }, { "epoch": 5.76, "learning_rate": 2.4180281690140844e-05, "loss": 3.7316, "step": 5136 }, { "epoch": 5.76, "learning_rate": 2.417887323943662e-05, "loss": 3.7426, "step": 5137 }, { "epoch": 5.76, "learning_rate": 2.4177464788732393e-05, "loss": 3.696, "step": 5138 }, { "epoch": 5.76, "learning_rate": 2.417605633802817e-05, "loss": 3.9848, "step": 5139 }, { "epoch": 5.76, "learning_rate": 2.4174647887323942e-05, "loss": 4.1071, "step": 5140 }, { "epoch": 5.76, "learning_rate": 2.417323943661972e-05, "loss": 3.7386, "step": 5141 }, { "epoch": 5.76, "learning_rate": 2.4171830985915495e-05, "loss": 3.7086, "step": 5142 }, { "epoch": 5.77, "learning_rate": 2.417042253521127e-05, "loss": 3.8901, "step": 5143 }, { "epoch": 5.77, "learning_rate": 2.4169014084507044e-05, "loss": 3.7985, "step": 5144 }, { "epoch": 5.77, "learning_rate": 2.4167605633802817e-05, "loss": 3.8193, "step": 5145 }, { "epoch": 5.77, "learning_rate": 2.4166197183098593e-05, "loss": 3.7704, "step": 5146 }, { "epoch": 5.77, "learning_rate": 2.4164788732394366e-05, "loss": 3.6901, "step": 5147 }, { "epoch": 5.77, "learning_rate": 2.4163380281690143e-05, "loss": 3.5481, "step": 5148 }, { "epoch": 5.77, "learning_rate": 2.4161971830985916e-05, "loss": 3.8929, "step": 5149 }, { "epoch": 5.77, "learning_rate": 2.4160563380281692e-05, "loss": 3.652, "step": 5150 }, { "epoch": 5.77, "learning_rate": 2.4159154929577465e-05, "loss": 3.7853, "step": 5151 }, { "epoch": 5.78, "learning_rate": 2.4157746478873238e-05, "loss": 3.7046, "step": 5152 }, { "epoch": 5.78, "learning_rate": 2.4156338028169014e-05, "loss": 3.6112, "step": 5153 }, { "epoch": 5.78, "learning_rate": 2.4154929577464787e-05, "loss": 3.4797, "step": 5154 }, { "epoch": 5.78, "learning_rate": 2.4153521126760563e-05, "loss": 3.5588, "step": 5155 }, { "epoch": 5.78, "learning_rate": 2.4152112676056336e-05, "loss": 3.1706, "step": 5156 }, { "epoch": 5.78, "learning_rate": 2.4150704225352116e-05, "loss": 3.2644, "step": 5157 }, { "epoch": 5.78, "learning_rate": 2.414929577464789e-05, "loss": 3.2535, "step": 5158 }, { "epoch": 5.78, "learning_rate": 2.4147887323943665e-05, "loss": 3.1084, "step": 5159 }, { "epoch": 5.78, "learning_rate": 2.414647887323944e-05, "loss": 2.9282, "step": 5160 }, { "epoch": 5.79, "learning_rate": 2.414507042253521e-05, "loss": 4.8711, "step": 5161 }, { "epoch": 5.79, "learning_rate": 2.4143661971830988e-05, "loss": 4.9786, "step": 5162 }, { "epoch": 5.79, "learning_rate": 2.414225352112676e-05, "loss": 4.6868, "step": 5163 }, { "epoch": 5.79, "learning_rate": 2.4140845070422537e-05, "loss": 4.7359, "step": 5164 }, { "epoch": 5.79, "learning_rate": 2.413943661971831e-05, "loss": 4.669, "step": 5165 }, { "epoch": 5.79, "learning_rate": 2.4138028169014086e-05, "loss": 4.3517, "step": 5166 }, { "epoch": 5.79, "learning_rate": 2.413661971830986e-05, "loss": 4.2893, "step": 5167 }, { "epoch": 5.79, "learning_rate": 2.4135211267605632e-05, "loss": 3.9477, "step": 5168 }, { "epoch": 5.79, "learning_rate": 2.413380281690141e-05, "loss": 4.1746, "step": 5169 }, { "epoch": 5.8, "learning_rate": 2.413239436619718e-05, "loss": 4.2819, "step": 5170 }, { "epoch": 5.8, "learning_rate": 2.4130985915492958e-05, "loss": 3.9176, "step": 5171 }, { "epoch": 5.8, "learning_rate": 2.4129577464788734e-05, "loss": 3.8586, "step": 5172 }, { "epoch": 5.8, "learning_rate": 2.412816901408451e-05, "loss": 4.1483, "step": 5173 }, { "epoch": 5.8, "learning_rate": 2.4126760563380283e-05, "loss": 3.9774, "step": 5174 }, { "epoch": 5.8, "learning_rate": 2.412535211267606e-05, "loss": 3.9336, "step": 5175 }, { "epoch": 5.8, "learning_rate": 2.4123943661971833e-05, "loss": 3.9992, "step": 5176 }, { "epoch": 5.8, "learning_rate": 2.4122535211267605e-05, "loss": 3.9051, "step": 5177 }, { "epoch": 5.8, "learning_rate": 2.4121126760563382e-05, "loss": 4.1692, "step": 5178 }, { "epoch": 5.81, "learning_rate": 2.4119718309859155e-05, "loss": 4.0495, "step": 5179 }, { "epoch": 5.81, "learning_rate": 2.411830985915493e-05, "loss": 4.0289, "step": 5180 }, { "epoch": 5.81, "learning_rate": 2.4116901408450704e-05, "loss": 3.8371, "step": 5181 }, { "epoch": 5.81, "learning_rate": 2.411549295774648e-05, "loss": 3.7318, "step": 5182 }, { "epoch": 5.81, "learning_rate": 2.4114084507042253e-05, "loss": 3.612, "step": 5183 }, { "epoch": 5.81, "learning_rate": 2.4112676056338026e-05, "loss": 3.734, "step": 5184 }, { "epoch": 5.81, "learning_rate": 2.4111267605633803e-05, "loss": 3.7208, "step": 5185 }, { "epoch": 5.81, "learning_rate": 2.4109859154929575e-05, "loss": 3.8519, "step": 5186 }, { "epoch": 5.82, "learning_rate": 2.4108450704225355e-05, "loss": 3.9308, "step": 5187 }, { "epoch": 5.82, "learning_rate": 2.4107042253521128e-05, "loss": 3.7582, "step": 5188 }, { "epoch": 5.82, "learning_rate": 2.4105633802816904e-05, "loss": 3.5669, "step": 5189 }, { "epoch": 5.82, "learning_rate": 2.4104225352112677e-05, "loss": 3.7871, "step": 5190 }, { "epoch": 5.82, "learning_rate": 2.410281690140845e-05, "loss": 3.6946, "step": 5191 }, { "epoch": 5.82, "learning_rate": 2.4101408450704227e-05, "loss": 3.6973, "step": 5192 }, { "epoch": 5.82, "learning_rate": 2.41e-05, "loss": 3.6979, "step": 5193 }, { "epoch": 5.82, "learning_rate": 2.4098591549295776e-05, "loss": 3.6591, "step": 5194 }, { "epoch": 5.82, "learning_rate": 2.409718309859155e-05, "loss": 3.731, "step": 5195 }, { "epoch": 5.83, "learning_rate": 2.4095774647887325e-05, "loss": 3.9144, "step": 5196 }, { "epoch": 5.83, "learning_rate": 2.4094366197183098e-05, "loss": 3.6129, "step": 5197 }, { "epoch": 5.83, "learning_rate": 2.4092957746478875e-05, "loss": 3.6213, "step": 5198 }, { "epoch": 5.83, "learning_rate": 2.4091549295774647e-05, "loss": 3.7212, "step": 5199 }, { "epoch": 5.83, "learning_rate": 2.409014084507042e-05, "loss": 3.7322, "step": 5200 }, { "epoch": 5.83, "learning_rate": 2.4088732394366197e-05, "loss": 3.574, "step": 5201 }, { "epoch": 5.83, "learning_rate": 2.4087323943661973e-05, "loss": 3.8352, "step": 5202 }, { "epoch": 5.83, "learning_rate": 2.408591549295775e-05, "loss": 3.3715, "step": 5203 }, { "epoch": 5.83, "learning_rate": 2.4084507042253522e-05, "loss": 3.7183, "step": 5204 }, { "epoch": 5.84, "learning_rate": 2.40830985915493e-05, "loss": 3.3335, "step": 5205 }, { "epoch": 5.84, "learning_rate": 2.408169014084507e-05, "loss": 3.3524, "step": 5206 }, { "epoch": 5.84, "learning_rate": 2.4080281690140845e-05, "loss": 3.5667, "step": 5207 }, { "epoch": 5.84, "learning_rate": 2.407887323943662e-05, "loss": 3.3841, "step": 5208 }, { "epoch": 5.84, "learning_rate": 2.4077464788732394e-05, "loss": 2.9933, "step": 5209 }, { "epoch": 5.84, "learning_rate": 2.407605633802817e-05, "loss": 2.9998, "step": 5210 }, { "epoch": 5.84, "learning_rate": 2.4074647887323943e-05, "loss": 4.8876, "step": 5211 }, { "epoch": 5.84, "learning_rate": 2.407323943661972e-05, "loss": 5.0221, "step": 5212 }, { "epoch": 5.84, "learning_rate": 2.4071830985915492e-05, "loss": 4.8534, "step": 5213 }, { "epoch": 5.85, "learning_rate": 2.407042253521127e-05, "loss": 4.6534, "step": 5214 }, { "epoch": 5.85, "learning_rate": 2.406901408450704e-05, "loss": 4.4469, "step": 5215 }, { "epoch": 5.85, "learning_rate": 2.4067605633802815e-05, "loss": 4.3562, "step": 5216 }, { "epoch": 5.85, "learning_rate": 2.406619718309859e-05, "loss": 4.0416, "step": 5217 }, { "epoch": 5.85, "learning_rate": 2.4064788732394367e-05, "loss": 4.197, "step": 5218 }, { "epoch": 5.85, "learning_rate": 2.4063380281690144e-05, "loss": 3.773, "step": 5219 }, { "epoch": 5.85, "learning_rate": 2.4061971830985917e-05, "loss": 3.9968, "step": 5220 }, { "epoch": 5.85, "learning_rate": 2.4060563380281693e-05, "loss": 4.0729, "step": 5221 }, { "epoch": 5.85, "learning_rate": 2.4059154929577466e-05, "loss": 3.8794, "step": 5222 }, { "epoch": 5.86, "learning_rate": 2.405774647887324e-05, "loss": 3.9009, "step": 5223 }, { "epoch": 5.86, "learning_rate": 2.4056338028169015e-05, "loss": 3.7737, "step": 5224 }, { "epoch": 5.86, "learning_rate": 2.4054929577464788e-05, "loss": 3.9829, "step": 5225 }, { "epoch": 5.86, "learning_rate": 2.4053521126760564e-05, "loss": 4.029, "step": 5226 }, { "epoch": 5.86, "learning_rate": 2.4052112676056337e-05, "loss": 3.8302, "step": 5227 }, { "epoch": 5.86, "learning_rate": 2.4050704225352114e-05, "loss": 3.8841, "step": 5228 }, { "epoch": 5.86, "learning_rate": 2.4049295774647887e-05, "loss": 3.9586, "step": 5229 }, { "epoch": 5.86, "learning_rate": 2.4047887323943663e-05, "loss": 3.8371, "step": 5230 }, { "epoch": 5.86, "learning_rate": 2.4046478873239436e-05, "loss": 3.9261, "step": 5231 }, { "epoch": 5.87, "learning_rate": 2.404507042253521e-05, "loss": 3.7876, "step": 5232 }, { "epoch": 5.87, "learning_rate": 2.404366197183099e-05, "loss": 3.6697, "step": 5233 }, { "epoch": 5.87, "learning_rate": 2.404225352112676e-05, "loss": 3.7496, "step": 5234 }, { "epoch": 5.87, "learning_rate": 2.4040845070422538e-05, "loss": 3.7403, "step": 5235 }, { "epoch": 5.87, "learning_rate": 2.403943661971831e-05, "loss": 3.7393, "step": 5236 }, { "epoch": 5.87, "learning_rate": 2.4038028169014087e-05, "loss": 3.9016, "step": 5237 }, { "epoch": 5.87, "learning_rate": 2.403661971830986e-05, "loss": 3.6946, "step": 5238 }, { "epoch": 5.87, "learning_rate": 2.4035211267605633e-05, "loss": 3.7056, "step": 5239 }, { "epoch": 5.87, "learning_rate": 2.403380281690141e-05, "loss": 3.7164, "step": 5240 }, { "epoch": 5.88, "learning_rate": 2.4032394366197182e-05, "loss": 3.7976, "step": 5241 }, { "epoch": 5.88, "learning_rate": 2.403098591549296e-05, "loss": 3.781, "step": 5242 }, { "epoch": 5.88, "learning_rate": 2.402957746478873e-05, "loss": 3.9625, "step": 5243 }, { "epoch": 5.88, "learning_rate": 2.4028169014084508e-05, "loss": 3.6903, "step": 5244 }, { "epoch": 5.88, "learning_rate": 2.402676056338028e-05, "loss": 3.7483, "step": 5245 }, { "epoch": 5.88, "learning_rate": 2.4025352112676057e-05, "loss": 3.4099, "step": 5246 }, { "epoch": 5.88, "learning_rate": 2.402394366197183e-05, "loss": 3.809, "step": 5247 }, { "epoch": 5.88, "learning_rate": 2.4022535211267606e-05, "loss": 3.4909, "step": 5248 }, { "epoch": 5.88, "learning_rate": 2.4021126760563383e-05, "loss": 3.6567, "step": 5249 }, { "epoch": 5.89, "learning_rate": 2.4019718309859156e-05, "loss": 3.5121, "step": 5250 }, { "epoch": 5.89, "learning_rate": 2.4018309859154932e-05, "loss": 3.1651, "step": 5251 }, { "epoch": 5.89, "learning_rate": 2.4016901408450705e-05, "loss": 3.6818, "step": 5252 }, { "epoch": 5.89, "learning_rate": 2.401549295774648e-05, "loss": 3.313, "step": 5253 }, { "epoch": 5.89, "learning_rate": 2.4014084507042254e-05, "loss": 3.4195, "step": 5254 }, { "epoch": 5.89, "learning_rate": 2.4012676056338027e-05, "loss": 3.3233, "step": 5255 }, { "epoch": 5.89, "learning_rate": 2.4011267605633804e-05, "loss": 3.4719, "step": 5256 }, { "epoch": 5.89, "learning_rate": 2.4009859154929576e-05, "loss": 2.9775, "step": 5257 }, { "epoch": 5.89, "learning_rate": 2.4008450704225353e-05, "loss": 3.2037, "step": 5258 }, { "epoch": 5.9, "learning_rate": 2.4007042253521126e-05, "loss": 3.0728, "step": 5259 }, { "epoch": 5.9, "learning_rate": 2.4005633802816902e-05, "loss": 3.3712, "step": 5260 }, { "epoch": 5.9, "learning_rate": 2.4004225352112675e-05, "loss": 5.0948, "step": 5261 }, { "epoch": 5.9, "learning_rate": 2.400281690140845e-05, "loss": 4.9959, "step": 5262 }, { "epoch": 5.9, "learning_rate": 2.4001408450704228e-05, "loss": 4.8439, "step": 5263 }, { "epoch": 5.9, "learning_rate": 2.4e-05, "loss": 4.5695, "step": 5264 }, { "epoch": 5.9, "learning_rate": 2.3998591549295777e-05, "loss": 4.4759, "step": 5265 }, { "epoch": 5.9, "learning_rate": 2.399718309859155e-05, "loss": 4.2196, "step": 5266 }, { "epoch": 5.9, "learning_rate": 2.3995774647887326e-05, "loss": 4.1768, "step": 5267 }, { "epoch": 5.91, "learning_rate": 2.39943661971831e-05, "loss": 4.0939, "step": 5268 }, { "epoch": 5.91, "learning_rate": 2.3992957746478875e-05, "loss": 3.9335, "step": 5269 }, { "epoch": 5.91, "learning_rate": 2.399154929577465e-05, "loss": 4.048, "step": 5270 }, { "epoch": 5.91, "learning_rate": 2.399014084507042e-05, "loss": 3.9116, "step": 5271 }, { "epoch": 5.91, "learning_rate": 2.3988732394366198e-05, "loss": 4.001, "step": 5272 }, { "epoch": 5.91, "learning_rate": 2.398732394366197e-05, "loss": 3.8966, "step": 5273 }, { "epoch": 5.91, "learning_rate": 2.3985915492957747e-05, "loss": 3.8162, "step": 5274 }, { "epoch": 5.91, "learning_rate": 2.398450704225352e-05, "loss": 3.8624, "step": 5275 }, { "epoch": 5.91, "learning_rate": 2.3983098591549296e-05, "loss": 3.7266, "step": 5276 }, { "epoch": 5.92, "learning_rate": 2.398169014084507e-05, "loss": 3.7544, "step": 5277 }, { "epoch": 5.92, "learning_rate": 2.398028169014085e-05, "loss": 3.8685, "step": 5278 }, { "epoch": 5.92, "learning_rate": 2.3978873239436622e-05, "loss": 3.6481, "step": 5279 }, { "epoch": 5.92, "learning_rate": 2.3977464788732395e-05, "loss": 3.7229, "step": 5280 }, { "epoch": 5.92, "learning_rate": 2.397605633802817e-05, "loss": 3.9982, "step": 5281 }, { "epoch": 5.92, "learning_rate": 2.3974647887323944e-05, "loss": 3.5931, "step": 5282 }, { "epoch": 5.92, "learning_rate": 2.397323943661972e-05, "loss": 3.7659, "step": 5283 }, { "epoch": 5.92, "learning_rate": 2.3971830985915493e-05, "loss": 3.7253, "step": 5284 }, { "epoch": 5.92, "learning_rate": 2.397042253521127e-05, "loss": 3.7197, "step": 5285 }, { "epoch": 5.93, "learning_rate": 2.3969014084507043e-05, "loss": 3.6606, "step": 5286 }, { "epoch": 5.93, "learning_rate": 2.3967605633802816e-05, "loss": 3.8419, "step": 5287 }, { "epoch": 5.93, "learning_rate": 2.3966197183098592e-05, "loss": 3.7767, "step": 5288 }, { "epoch": 5.93, "learning_rate": 2.3964788732394365e-05, "loss": 3.622, "step": 5289 }, { "epoch": 5.93, "learning_rate": 2.396338028169014e-05, "loss": 3.7016, "step": 5290 }, { "epoch": 5.93, "learning_rate": 2.3961971830985914e-05, "loss": 3.7218, "step": 5291 }, { "epoch": 5.93, "learning_rate": 2.396056338028169e-05, "loss": 3.4422, "step": 5292 }, { "epoch": 5.93, "learning_rate": 2.3959154929577467e-05, "loss": 3.558, "step": 5293 }, { "epoch": 5.93, "learning_rate": 2.3957746478873243e-05, "loss": 3.5592, "step": 5294 }, { "epoch": 5.94, "learning_rate": 2.3956338028169016e-05, "loss": 3.4933, "step": 5295 }, { "epoch": 5.94, "learning_rate": 2.395492957746479e-05, "loss": 3.5899, "step": 5296 }, { "epoch": 5.94, "learning_rate": 2.3953521126760565e-05, "loss": 3.6054, "step": 5297 }, { "epoch": 5.94, "learning_rate": 2.3952112676056338e-05, "loss": 3.2973, "step": 5298 }, { "epoch": 5.94, "learning_rate": 2.3950704225352115e-05, "loss": 3.4866, "step": 5299 }, { "epoch": 5.94, "learning_rate": 2.3949295774647888e-05, "loss": 3.7538, "step": 5300 }, { "epoch": 5.94, "learning_rate": 2.3947887323943664e-05, "loss": 3.5482, "step": 5301 }, { "epoch": 5.94, "learning_rate": 2.3946478873239437e-05, "loss": 3.4883, "step": 5302 }, { "epoch": 5.95, "learning_rate": 2.394507042253521e-05, "loss": 3.4709, "step": 5303 }, { "epoch": 5.95, "learning_rate": 2.3943661971830986e-05, "loss": 3.1639, "step": 5304 }, { "epoch": 5.95, "learning_rate": 2.394225352112676e-05, "loss": 3.2833, "step": 5305 }, { "epoch": 5.95, "learning_rate": 2.3940845070422535e-05, "loss": 2.8809, "step": 5306 }, { "epoch": 5.95, "learning_rate": 2.393943661971831e-05, "loss": 3.6022, "step": 5307 }, { "epoch": 5.95, "learning_rate": 2.3938028169014088e-05, "loss": 3.1454, "step": 5308 }, { "epoch": 5.95, "learning_rate": 2.393661971830986e-05, "loss": 3.2413, "step": 5309 }, { "epoch": 5.95, "learning_rate": 2.3935211267605634e-05, "loss": 3.0149, "step": 5310 }, { "epoch": 5.95, "learning_rate": 2.393380281690141e-05, "loss": 4.8728, "step": 5311 }, { "epoch": 5.96, "learning_rate": 2.3932394366197183e-05, "loss": 4.9812, "step": 5312 }, { "epoch": 5.96, "learning_rate": 2.393098591549296e-05, "loss": 4.8105, "step": 5313 }, { "epoch": 5.96, "learning_rate": 2.3929577464788732e-05, "loss": 4.5237, "step": 5314 }, { "epoch": 5.96, "learning_rate": 2.392816901408451e-05, "loss": 4.2661, "step": 5315 }, { "epoch": 5.96, "learning_rate": 2.3926760563380282e-05, "loss": 4.3063, "step": 5316 }, { "epoch": 5.96, "learning_rate": 2.3925352112676058e-05, "loss": 3.8446, "step": 5317 }, { "epoch": 5.96, "learning_rate": 2.392394366197183e-05, "loss": 3.9587, "step": 5318 }, { "epoch": 5.96, "learning_rate": 2.3922535211267604e-05, "loss": 4.0946, "step": 5319 }, { "epoch": 5.96, "learning_rate": 2.392112676056338e-05, "loss": 4.1643, "step": 5320 }, { "epoch": 5.97, "learning_rate": 2.3919718309859153e-05, "loss": 4.2482, "step": 5321 }, { "epoch": 5.97, "learning_rate": 2.391830985915493e-05, "loss": 4.1565, "step": 5322 }, { "epoch": 5.97, "learning_rate": 2.3916901408450706e-05, "loss": 4.1892, "step": 5323 }, { "epoch": 5.97, "learning_rate": 2.3915492957746482e-05, "loss": 3.938, "step": 5324 }, { "epoch": 5.97, "learning_rate": 2.3914084507042255e-05, "loss": 3.7922, "step": 5325 }, { "epoch": 5.97, "learning_rate": 2.3912676056338028e-05, "loss": 3.941, "step": 5326 }, { "epoch": 5.97, "learning_rate": 2.3911267605633804e-05, "loss": 3.6748, "step": 5327 }, { "epoch": 5.97, "learning_rate": 2.3909859154929577e-05, "loss": 3.7668, "step": 5328 }, { "epoch": 5.97, "learning_rate": 2.3908450704225354e-05, "loss": 3.9112, "step": 5329 }, { "epoch": 5.98, "learning_rate": 2.3907042253521127e-05, "loss": 3.6619, "step": 5330 }, { "epoch": 5.98, "learning_rate": 2.3905633802816903e-05, "loss": 3.8502, "step": 5331 }, { "epoch": 5.98, "learning_rate": 2.3904225352112676e-05, "loss": 3.6198, "step": 5332 }, { "epoch": 5.98, "learning_rate": 2.3902816901408452e-05, "loss": 3.773, "step": 5333 }, { "epoch": 5.98, "learning_rate": 2.3901408450704225e-05, "loss": 3.616, "step": 5334 }, { "epoch": 5.98, "learning_rate": 2.3899999999999998e-05, "loss": 3.7229, "step": 5335 }, { "epoch": 5.98, "learning_rate": 2.3898591549295775e-05, "loss": 3.8098, "step": 5336 }, { "epoch": 5.98, "learning_rate": 2.3897183098591547e-05, "loss": 3.3736, "step": 5337 }, { "epoch": 5.98, "learning_rate": 2.3895774647887324e-05, "loss": 3.8415, "step": 5338 }, { "epoch": 5.99, "learning_rate": 2.38943661971831e-05, "loss": 3.6628, "step": 5339 }, { "epoch": 5.99, "learning_rate": 2.3892957746478876e-05, "loss": 3.3489, "step": 5340 }, { "epoch": 5.99, "learning_rate": 2.389154929577465e-05, "loss": 3.5916, "step": 5341 }, { "epoch": 5.99, "learning_rate": 2.3890140845070422e-05, "loss": 3.4368, "step": 5342 }, { "epoch": 5.99, "learning_rate": 2.38887323943662e-05, "loss": 3.4178, "step": 5343 }, { "epoch": 5.99, "learning_rate": 2.388732394366197e-05, "loss": 3.6133, "step": 5344 }, { "epoch": 5.99, "learning_rate": 2.3885915492957748e-05, "loss": 3.5591, "step": 5345 }, { "epoch": 5.99, "learning_rate": 2.388450704225352e-05, "loss": 3.4906, "step": 5346 }, { "epoch": 5.99, "learning_rate": 2.3883098591549297e-05, "loss": 3.58, "step": 5347 }, { "epoch": 6.0, "learning_rate": 2.388169014084507e-05, "loss": 3.2565, "step": 5348 }, { "epoch": 6.0, "learning_rate": 2.3880281690140846e-05, "loss": 3.063, "step": 5349 }, { "epoch": 6.0, "learning_rate": 2.387887323943662e-05, "loss": 2.9821, "step": 5350 }, { "epoch": 6.0, "learning_rate": 2.3877464788732392e-05, "loss": 3.0465, "step": 5351 }, { "epoch": 6.0, "learning_rate": 2.387605633802817e-05, "loss": 3.4441, "step": 5352 }, { "epoch": 6.0, "learning_rate": 2.387464788732394e-05, "loss": 4.9694, "step": 5353 }, { "epoch": 6.0, "learning_rate": 2.387323943661972e-05, "loss": 4.8158, "step": 5354 }, { "epoch": 6.0, "learning_rate": 2.3871830985915494e-05, "loss": 4.693, "step": 5355 }, { "epoch": 6.0, "learning_rate": 2.387042253521127e-05, "loss": 4.4646, "step": 5356 }, { "epoch": 6.01, "learning_rate": 2.3869014084507044e-05, "loss": 4.5067, "step": 5357 }, { "epoch": 6.01, "learning_rate": 2.3867605633802817e-05, "loss": 4.3752, "step": 5358 }, { "epoch": 6.01, "learning_rate": 2.3866197183098593e-05, "loss": 4.0213, "step": 5359 }, { "epoch": 6.01, "learning_rate": 2.3864788732394366e-05, "loss": 4.2366, "step": 5360 }, { "epoch": 6.01, "learning_rate": 2.3863380281690142e-05, "loss": 3.9175, "step": 5361 }, { "epoch": 6.01, "learning_rate": 2.3861971830985915e-05, "loss": 3.9286, "step": 5362 }, { "epoch": 6.01, "learning_rate": 2.386056338028169e-05, "loss": 3.8406, "step": 5363 }, { "epoch": 6.01, "learning_rate": 2.3859154929577464e-05, "loss": 3.9705, "step": 5364 }, { "epoch": 6.01, "learning_rate": 2.385774647887324e-05, "loss": 4.0768, "step": 5365 }, { "epoch": 6.02, "learning_rate": 2.3856338028169014e-05, "loss": 4.01, "step": 5366 }, { "epoch": 6.02, "learning_rate": 2.3854929577464787e-05, "loss": 3.7156, "step": 5367 }, { "epoch": 6.02, "learning_rate": 2.3853521126760563e-05, "loss": 3.7625, "step": 5368 }, { "epoch": 6.02, "learning_rate": 2.385211267605634e-05, "loss": 3.8824, "step": 5369 }, { "epoch": 6.02, "learning_rate": 2.3850704225352116e-05, "loss": 3.464, "step": 5370 }, { "epoch": 6.02, "learning_rate": 2.384929577464789e-05, "loss": 3.4764, "step": 5371 }, { "epoch": 6.02, "learning_rate": 2.3847887323943665e-05, "loss": 3.5362, "step": 5372 }, { "epoch": 6.02, "learning_rate": 2.3846478873239438e-05, "loss": 3.6112, "step": 5373 }, { "epoch": 6.02, "learning_rate": 2.384507042253521e-05, "loss": 3.7115, "step": 5374 }, { "epoch": 6.03, "learning_rate": 2.3843661971830987e-05, "loss": 3.4083, "step": 5375 }, { "epoch": 6.03, "learning_rate": 2.384225352112676e-05, "loss": 3.4961, "step": 5376 }, { "epoch": 6.03, "learning_rate": 2.3840845070422536e-05, "loss": 3.6358, "step": 5377 }, { "epoch": 6.03, "learning_rate": 2.383943661971831e-05, "loss": 3.5701, "step": 5378 }, { "epoch": 6.03, "learning_rate": 2.3838028169014086e-05, "loss": 3.4776, "step": 5379 }, { "epoch": 6.03, "learning_rate": 2.383661971830986e-05, "loss": 3.4578, "step": 5380 }, { "epoch": 6.03, "learning_rate": 2.3835211267605635e-05, "loss": 3.6086, "step": 5381 }, { "epoch": 6.03, "learning_rate": 2.3833802816901408e-05, "loss": 3.6297, "step": 5382 }, { "epoch": 6.03, "learning_rate": 2.383239436619718e-05, "loss": 3.5811, "step": 5383 }, { "epoch": 6.04, "learning_rate": 2.383098591549296e-05, "loss": 3.3355, "step": 5384 }, { "epoch": 6.04, "learning_rate": 2.3829577464788733e-05, "loss": 3.2801, "step": 5385 }, { "epoch": 6.04, "learning_rate": 2.382816901408451e-05, "loss": 3.4001, "step": 5386 }, { "epoch": 6.04, "learning_rate": 2.3826760563380283e-05, "loss": 3.4289, "step": 5387 }, { "epoch": 6.04, "learning_rate": 2.382535211267606e-05, "loss": 3.4138, "step": 5388 }, { "epoch": 6.04, "learning_rate": 2.3823943661971832e-05, "loss": 3.3643, "step": 5389 }, { "epoch": 6.04, "learning_rate": 2.3822535211267605e-05, "loss": 2.9166, "step": 5390 }, { "epoch": 6.04, "learning_rate": 2.382112676056338e-05, "loss": 3.1367, "step": 5391 }, { "epoch": 6.04, "learning_rate": 2.3819718309859154e-05, "loss": 3.2081, "step": 5392 }, { "epoch": 6.05, "learning_rate": 2.381830985915493e-05, "loss": 3.4812, "step": 5393 }, { "epoch": 6.05, "learning_rate": 2.3816901408450703e-05, "loss": 3.3823, "step": 5394 }, { "epoch": 6.05, "learning_rate": 2.381549295774648e-05, "loss": 3.2537, "step": 5395 }, { "epoch": 6.05, "learning_rate": 2.3814084507042253e-05, "loss": 3.0175, "step": 5396 }, { "epoch": 6.05, "learning_rate": 2.381267605633803e-05, "loss": 2.9268, "step": 5397 }, { "epoch": 6.05, "learning_rate": 2.3811267605633802e-05, "loss": 3.177, "step": 5398 }, { "epoch": 6.05, "learning_rate": 2.380985915492958e-05, "loss": 3.0892, "step": 5399 }, { "epoch": 6.05, "learning_rate": 2.3808450704225355e-05, "loss": 2.8593, "step": 5400 }, { "epoch": 6.05, "learning_rate": 2.3807042253521128e-05, "loss": 3.0077, "step": 5401 }, { "epoch": 6.06, "learning_rate": 2.3805633802816904e-05, "loss": 3.1672, "step": 5402 }, { "epoch": 6.06, "learning_rate": 2.3804225352112677e-05, "loss": 4.5168, "step": 5403 }, { "epoch": 6.06, "learning_rate": 2.3802816901408453e-05, "loss": 4.4492, "step": 5404 }, { "epoch": 6.06, "learning_rate": 2.3801408450704226e-05, "loss": 4.5544, "step": 5405 }, { "epoch": 6.06, "learning_rate": 2.38e-05, "loss": 4.279, "step": 5406 }, { "epoch": 6.06, "learning_rate": 2.3798591549295775e-05, "loss": 4.3404, "step": 5407 }, { "epoch": 6.06, "learning_rate": 2.379718309859155e-05, "loss": 3.6833, "step": 5408 }, { "epoch": 6.06, "learning_rate": 2.3795774647887325e-05, "loss": 3.9128, "step": 5409 }, { "epoch": 6.07, "learning_rate": 2.3794366197183098e-05, "loss": 3.8242, "step": 5410 }, { "epoch": 6.07, "learning_rate": 2.3792957746478874e-05, "loss": 3.7433, "step": 5411 }, { "epoch": 6.07, "learning_rate": 2.3791549295774647e-05, "loss": 3.6185, "step": 5412 }, { "epoch": 6.07, "learning_rate": 2.3790140845070423e-05, "loss": 3.7925, "step": 5413 }, { "epoch": 6.07, "learning_rate": 2.37887323943662e-05, "loss": 3.8606, "step": 5414 }, { "epoch": 6.07, "learning_rate": 2.3787323943661973e-05, "loss": 3.7729, "step": 5415 }, { "epoch": 6.07, "learning_rate": 2.378591549295775e-05, "loss": 3.486, "step": 5416 }, { "epoch": 6.07, "learning_rate": 2.3784507042253522e-05, "loss": 3.722, "step": 5417 }, { "epoch": 6.07, "learning_rate": 2.3783098591549298e-05, "loss": 3.6177, "step": 5418 }, { "epoch": 6.08, "learning_rate": 2.378169014084507e-05, "loss": 3.5336, "step": 5419 }, { "epoch": 6.08, "learning_rate": 2.3780281690140847e-05, "loss": 3.6962, "step": 5420 }, { "epoch": 6.08, "learning_rate": 2.377887323943662e-05, "loss": 3.5972, "step": 5421 }, { "epoch": 6.08, "learning_rate": 2.3777464788732393e-05, "loss": 3.6507, "step": 5422 }, { "epoch": 6.08, "learning_rate": 2.377605633802817e-05, "loss": 3.5137, "step": 5423 }, { "epoch": 6.08, "learning_rate": 2.3774647887323943e-05, "loss": 3.5916, "step": 5424 }, { "epoch": 6.08, "learning_rate": 2.377323943661972e-05, "loss": 3.4354, "step": 5425 }, { "epoch": 6.08, "learning_rate": 2.3771830985915492e-05, "loss": 3.2763, "step": 5426 }, { "epoch": 6.08, "learning_rate": 2.3770422535211268e-05, "loss": 3.4724, "step": 5427 }, { "epoch": 6.09, "learning_rate": 2.376901408450704e-05, "loss": 3.3295, "step": 5428 }, { "epoch": 6.09, "learning_rate": 2.3767605633802817e-05, "loss": 3.3237, "step": 5429 }, { "epoch": 6.09, "learning_rate": 2.3766197183098594e-05, "loss": 3.5903, "step": 5430 }, { "epoch": 6.09, "learning_rate": 2.3764788732394367e-05, "loss": 3.5168, "step": 5431 }, { "epoch": 6.09, "learning_rate": 2.3763380281690143e-05, "loss": 3.4929, "step": 5432 }, { "epoch": 6.09, "learning_rate": 2.3761971830985916e-05, "loss": 3.6756, "step": 5433 }, { "epoch": 6.09, "learning_rate": 2.3760563380281692e-05, "loss": 3.2534, "step": 5434 }, { "epoch": 6.09, "learning_rate": 2.3759154929577465e-05, "loss": 3.4058, "step": 5435 }, { "epoch": 6.09, "learning_rate": 2.375774647887324e-05, "loss": 3.2951, "step": 5436 }, { "epoch": 6.1, "learning_rate": 2.3756338028169015e-05, "loss": 3.2902, "step": 5437 }, { "epoch": 6.1, "learning_rate": 2.3754929577464788e-05, "loss": 3.2037, "step": 5438 }, { "epoch": 6.1, "learning_rate": 2.3753521126760564e-05, "loss": 3.3175, "step": 5439 }, { "epoch": 6.1, "learning_rate": 2.3752112676056337e-05, "loss": 3.4539, "step": 5440 }, { "epoch": 6.1, "learning_rate": 2.3750704225352113e-05, "loss": 3.3846, "step": 5441 }, { "epoch": 6.1, "learning_rate": 2.3749295774647886e-05, "loss": 3.5199, "step": 5442 }, { "epoch": 6.1, "learning_rate": 2.3747887323943662e-05, "loss": 3.1895, "step": 5443 }, { "epoch": 6.1, "learning_rate": 2.3746478873239435e-05, "loss": 3.1384, "step": 5444 }, { "epoch": 6.1, "learning_rate": 2.374507042253521e-05, "loss": 3.1318, "step": 5445 }, { "epoch": 6.11, "learning_rate": 2.3743661971830988e-05, "loss": 3.0714, "step": 5446 }, { "epoch": 6.11, "learning_rate": 2.374225352112676e-05, "loss": 2.9544, "step": 5447 }, { "epoch": 6.11, "learning_rate": 2.3740845070422537e-05, "loss": 3.0632, "step": 5448 }, { "epoch": 6.11, "learning_rate": 2.373943661971831e-05, "loss": 2.7881, "step": 5449 }, { "epoch": 6.11, "learning_rate": 2.3738028169014087e-05, "loss": 3.0255, "step": 5450 }, { "epoch": 6.11, "learning_rate": 2.373661971830986e-05, "loss": 2.827, "step": 5451 }, { "epoch": 6.11, "learning_rate": 2.3735211267605636e-05, "loss": 2.8089, "step": 5452 }, { "epoch": 6.11, "learning_rate": 2.373380281690141e-05, "loss": 4.7112, "step": 5453 }, { "epoch": 6.11, "learning_rate": 2.3732394366197182e-05, "loss": 4.3636, "step": 5454 }, { "epoch": 6.12, "learning_rate": 2.3730985915492958e-05, "loss": 4.6045, "step": 5455 }, { "epoch": 6.12, "learning_rate": 2.372957746478873e-05, "loss": 4.4773, "step": 5456 }, { "epoch": 6.12, "learning_rate": 2.3728169014084507e-05, "loss": 4.2619, "step": 5457 }, { "epoch": 6.12, "learning_rate": 2.372676056338028e-05, "loss": 4.1489, "step": 5458 }, { "epoch": 6.12, "learning_rate": 2.3725352112676057e-05, "loss": 3.7772, "step": 5459 }, { "epoch": 6.12, "learning_rate": 2.3723943661971833e-05, "loss": 3.7389, "step": 5460 }, { "epoch": 6.12, "learning_rate": 2.3722535211267606e-05, "loss": 3.7752, "step": 5461 }, { "epoch": 6.12, "learning_rate": 2.3721126760563382e-05, "loss": 3.8405, "step": 5462 }, { "epoch": 6.12, "learning_rate": 2.3719718309859155e-05, "loss": 3.7487, "step": 5463 }, { "epoch": 6.13, "learning_rate": 2.371830985915493e-05, "loss": 3.9276, "step": 5464 }, { "epoch": 6.13, "learning_rate": 2.3716901408450704e-05, "loss": 3.7164, "step": 5465 }, { "epoch": 6.13, "learning_rate": 2.371549295774648e-05, "loss": 3.848, "step": 5466 }, { "epoch": 6.13, "learning_rate": 2.3714084507042254e-05, "loss": 3.7617, "step": 5467 }, { "epoch": 6.13, "learning_rate": 2.371267605633803e-05, "loss": 3.6534, "step": 5468 }, { "epoch": 6.13, "learning_rate": 2.3711267605633803e-05, "loss": 3.5428, "step": 5469 }, { "epoch": 6.13, "learning_rate": 2.3709859154929576e-05, "loss": 3.4074, "step": 5470 }, { "epoch": 6.13, "learning_rate": 2.3708450704225352e-05, "loss": 3.5859, "step": 5471 }, { "epoch": 6.13, "learning_rate": 2.3707042253521125e-05, "loss": 3.5721, "step": 5472 }, { "epoch": 6.14, "learning_rate": 2.37056338028169e-05, "loss": 3.411, "step": 5473 }, { "epoch": 6.14, "learning_rate": 2.3704225352112674e-05, "loss": 3.2242, "step": 5474 }, { "epoch": 6.14, "learning_rate": 2.3702816901408454e-05, "loss": 3.4767, "step": 5475 }, { "epoch": 6.14, "learning_rate": 2.3701408450704227e-05, "loss": 3.3659, "step": 5476 }, { "epoch": 6.14, "learning_rate": 2.37e-05, "loss": 3.2483, "step": 5477 }, { "epoch": 6.14, "learning_rate": 2.3698591549295776e-05, "loss": 3.3769, "step": 5478 }, { "epoch": 6.14, "learning_rate": 2.369718309859155e-05, "loss": 3.3533, "step": 5479 }, { "epoch": 6.14, "learning_rate": 2.3695774647887326e-05, "loss": 3.4189, "step": 5480 }, { "epoch": 6.14, "learning_rate": 2.36943661971831e-05, "loss": 3.4452, "step": 5481 }, { "epoch": 6.15, "learning_rate": 2.3692957746478875e-05, "loss": 3.4971, "step": 5482 }, { "epoch": 6.15, "learning_rate": 2.3691549295774648e-05, "loss": 3.265, "step": 5483 }, { "epoch": 6.15, "learning_rate": 2.3690140845070424e-05, "loss": 3.2828, "step": 5484 }, { "epoch": 6.15, "learning_rate": 2.3688732394366197e-05, "loss": 3.4621, "step": 5485 }, { "epoch": 6.15, "learning_rate": 2.368732394366197e-05, "loss": 3.164, "step": 5486 }, { "epoch": 6.15, "learning_rate": 2.3685915492957746e-05, "loss": 3.5942, "step": 5487 }, { "epoch": 6.15, "learning_rate": 2.368450704225352e-05, "loss": 3.1363, "step": 5488 }, { "epoch": 6.15, "learning_rate": 2.3683098591549296e-05, "loss": 3.3697, "step": 5489 }, { "epoch": 6.15, "learning_rate": 2.3681690140845072e-05, "loss": 3.2271, "step": 5490 }, { "epoch": 6.16, "learning_rate": 2.368028169014085e-05, "loss": 3.2021, "step": 5491 }, { "epoch": 6.16, "learning_rate": 2.367887323943662e-05, "loss": 3.2052, "step": 5492 }, { "epoch": 6.16, "learning_rate": 2.3677464788732394e-05, "loss": 3.1943, "step": 5493 }, { "epoch": 6.16, "learning_rate": 2.367605633802817e-05, "loss": 3.3088, "step": 5494 }, { "epoch": 6.16, "learning_rate": 2.3674647887323944e-05, "loss": 2.9888, "step": 5495 }, { "epoch": 6.16, "learning_rate": 2.367323943661972e-05, "loss": 3.1506, "step": 5496 }, { "epoch": 6.16, "learning_rate": 2.3671830985915493e-05, "loss": 3.1426, "step": 5497 }, { "epoch": 6.16, "learning_rate": 2.367042253521127e-05, "loss": 3.2706, "step": 5498 }, { "epoch": 6.16, "learning_rate": 2.3669014084507042e-05, "loss": 2.9105, "step": 5499 }, { "epoch": 6.17, "learning_rate": 2.366760563380282e-05, "loss": 2.7499, "step": 5500 }, { "epoch": 6.17, "learning_rate": 2.366619718309859e-05, "loss": 2.7157, "step": 5501 }, { "epoch": 6.17, "learning_rate": 2.3664788732394364e-05, "loss": 2.8871, "step": 5502 }, { "epoch": 6.17, "learning_rate": 2.366338028169014e-05, "loss": 4.9091, "step": 5503 }, { "epoch": 6.17, "learning_rate": 2.3661971830985914e-05, "loss": 4.7668, "step": 5504 }, { "epoch": 6.17, "learning_rate": 2.3660563380281693e-05, "loss": 4.4606, "step": 5505 }, { "epoch": 6.17, "learning_rate": 2.3659154929577466e-05, "loss": 4.4873, "step": 5506 }, { "epoch": 6.17, "learning_rate": 2.3657746478873243e-05, "loss": 4.1592, "step": 5507 }, { "epoch": 6.17, "learning_rate": 2.3656338028169016e-05, "loss": 4.0082, "step": 5508 }, { "epoch": 6.18, "learning_rate": 2.365492957746479e-05, "loss": 3.8736, "step": 5509 }, { "epoch": 6.18, "learning_rate": 2.3653521126760565e-05, "loss": 3.9051, "step": 5510 }, { "epoch": 6.18, "learning_rate": 2.3652112676056338e-05, "loss": 3.6912, "step": 5511 }, { "epoch": 6.18, "learning_rate": 2.3650704225352114e-05, "loss": 3.5647, "step": 5512 }, { "epoch": 6.18, "learning_rate": 2.3649295774647887e-05, "loss": 3.7094, "step": 5513 }, { "epoch": 6.18, "learning_rate": 2.3647887323943663e-05, "loss": 3.5573, "step": 5514 }, { "epoch": 6.18, "learning_rate": 2.3646478873239436e-05, "loss": 3.4994, "step": 5515 }, { "epoch": 6.18, "learning_rate": 2.3645070422535213e-05, "loss": 3.4824, "step": 5516 }, { "epoch": 6.18, "learning_rate": 2.3643661971830986e-05, "loss": 3.6783, "step": 5517 }, { "epoch": 6.19, "learning_rate": 2.364225352112676e-05, "loss": 3.5266, "step": 5518 }, { "epoch": 6.19, "learning_rate": 2.3640845070422535e-05, "loss": 3.6082, "step": 5519 }, { "epoch": 6.19, "learning_rate": 2.363943661971831e-05, "loss": 3.3943, "step": 5520 }, { "epoch": 6.19, "learning_rate": 2.3638028169014088e-05, "loss": 3.3532, "step": 5521 }, { "epoch": 6.19, "learning_rate": 2.363661971830986e-05, "loss": 3.5859, "step": 5522 }, { "epoch": 6.19, "learning_rate": 2.3635211267605637e-05, "loss": 3.4807, "step": 5523 }, { "epoch": 6.19, "learning_rate": 2.363380281690141e-05, "loss": 3.2862, "step": 5524 }, { "epoch": 6.19, "learning_rate": 2.3632394366197183e-05, "loss": 3.3967, "step": 5525 }, { "epoch": 6.2, "learning_rate": 2.363098591549296e-05, "loss": 3.3602, "step": 5526 }, { "epoch": 6.2, "learning_rate": 2.3629577464788732e-05, "loss": 3.4102, "step": 5527 }, { "epoch": 6.2, "learning_rate": 2.3628169014084508e-05, "loss": 3.3108, "step": 5528 }, { "epoch": 6.2, "learning_rate": 2.362676056338028e-05, "loss": 3.3602, "step": 5529 }, { "epoch": 6.2, "learning_rate": 2.3625352112676058e-05, "loss": 3.3001, "step": 5530 }, { "epoch": 6.2, "learning_rate": 2.362394366197183e-05, "loss": 3.3366, "step": 5531 }, { "epoch": 6.2, "learning_rate": 2.3622535211267607e-05, "loss": 3.1723, "step": 5532 }, { "epoch": 6.2, "learning_rate": 2.362112676056338e-05, "loss": 3.3881, "step": 5533 }, { "epoch": 6.2, "learning_rate": 2.3619718309859153e-05, "loss": 3.2668, "step": 5534 }, { "epoch": 6.21, "learning_rate": 2.3618309859154932e-05, "loss": 3.1694, "step": 5535 }, { "epoch": 6.21, "learning_rate": 2.3616901408450705e-05, "loss": 3.392, "step": 5536 }, { "epoch": 6.21, "learning_rate": 2.3615492957746482e-05, "loss": 3.3049, "step": 5537 }, { "epoch": 6.21, "learning_rate": 2.3614084507042255e-05, "loss": 3.5167, "step": 5538 }, { "epoch": 6.21, "learning_rate": 2.361267605633803e-05, "loss": 3.3331, "step": 5539 }, { "epoch": 6.21, "learning_rate": 2.3611267605633804e-05, "loss": 3.3582, "step": 5540 }, { "epoch": 6.21, "learning_rate": 2.3609859154929577e-05, "loss": 3.3451, "step": 5541 }, { "epoch": 6.21, "learning_rate": 2.3608450704225353e-05, "loss": 3.1509, "step": 5542 }, { "epoch": 6.21, "learning_rate": 2.3607042253521126e-05, "loss": 3.4027, "step": 5543 }, { "epoch": 6.22, "learning_rate": 2.3605633802816902e-05, "loss": 3.0919, "step": 5544 }, { "epoch": 6.22, "learning_rate": 2.3604225352112675e-05, "loss": 3.1222, "step": 5545 }, { "epoch": 6.22, "learning_rate": 2.3602816901408452e-05, "loss": 2.8138, "step": 5546 }, { "epoch": 6.22, "learning_rate": 2.3601408450704225e-05, "loss": 3.0522, "step": 5547 }, { "epoch": 6.22, "learning_rate": 2.3599999999999998e-05, "loss": 3.1423, "step": 5548 }, { "epoch": 6.22, "learning_rate": 2.3598591549295774e-05, "loss": 2.7162, "step": 5549 }, { "epoch": 6.22, "learning_rate": 2.359718309859155e-05, "loss": 2.6226, "step": 5550 }, { "epoch": 6.22, "learning_rate": 2.3595774647887327e-05, "loss": 2.9092, "step": 5551 }, { "epoch": 6.22, "learning_rate": 2.35943661971831e-05, "loss": 2.5482, "step": 5552 }, { "epoch": 6.23, "learning_rate": 2.3592957746478876e-05, "loss": 4.6555, "step": 5553 }, { "epoch": 6.23, "learning_rate": 2.359154929577465e-05, "loss": 4.6007, "step": 5554 }, { "epoch": 6.23, "learning_rate": 2.3590140845070425e-05, "loss": 4.4728, "step": 5555 }, { "epoch": 6.23, "learning_rate": 2.3588732394366198e-05, "loss": 4.209, "step": 5556 }, { "epoch": 6.23, "learning_rate": 2.358732394366197e-05, "loss": 4.1935, "step": 5557 }, { "epoch": 6.23, "learning_rate": 2.3585915492957747e-05, "loss": 3.8245, "step": 5558 }, { "epoch": 6.23, "learning_rate": 2.358450704225352e-05, "loss": 3.8989, "step": 5559 }, { "epoch": 6.23, "learning_rate": 2.3583098591549297e-05, "loss": 3.3165, "step": 5560 }, { "epoch": 6.23, "learning_rate": 2.358169014084507e-05, "loss": 3.6415, "step": 5561 }, { "epoch": 6.24, "learning_rate": 2.3580281690140846e-05, "loss": 3.4202, "step": 5562 }, { "epoch": 6.24, "learning_rate": 2.357887323943662e-05, "loss": 3.6795, "step": 5563 }, { "epoch": 6.24, "learning_rate": 2.3577464788732392e-05, "loss": 3.8427, "step": 5564 }, { "epoch": 6.24, "learning_rate": 2.3576056338028168e-05, "loss": 3.4898, "step": 5565 }, { "epoch": 6.24, "learning_rate": 2.3574647887323945e-05, "loss": 3.3327, "step": 5566 }, { "epoch": 6.24, "learning_rate": 2.357323943661972e-05, "loss": 3.4158, "step": 5567 }, { "epoch": 6.24, "learning_rate": 2.3571830985915494e-05, "loss": 3.2577, "step": 5568 }, { "epoch": 6.24, "learning_rate": 2.357042253521127e-05, "loss": 3.1082, "step": 5569 }, { "epoch": 6.24, "learning_rate": 2.3569014084507043e-05, "loss": 3.3148, "step": 5570 }, { "epoch": 6.25, "learning_rate": 2.356760563380282e-05, "loss": 3.0952, "step": 5571 }, { "epoch": 6.25, "learning_rate": 2.3566197183098592e-05, "loss": 3.3844, "step": 5572 }, { "epoch": 6.25, "learning_rate": 2.3564788732394365e-05, "loss": 3.6547, "step": 5573 }, { "epoch": 6.25, "learning_rate": 2.356338028169014e-05, "loss": 3.4373, "step": 5574 }, { "epoch": 6.25, "learning_rate": 2.3561971830985915e-05, "loss": 3.2984, "step": 5575 }, { "epoch": 6.25, "learning_rate": 2.356056338028169e-05, "loss": 3.2542, "step": 5576 }, { "epoch": 6.25, "learning_rate": 2.3559154929577464e-05, "loss": 3.396, "step": 5577 }, { "epoch": 6.25, "learning_rate": 2.355774647887324e-05, "loss": 3.4495, "step": 5578 }, { "epoch": 6.25, "learning_rate": 2.3556338028169013e-05, "loss": 3.1795, "step": 5579 }, { "epoch": 6.26, "learning_rate": 2.3554929577464786e-05, "loss": 3.1951, "step": 5580 }, { "epoch": 6.26, "learning_rate": 2.3553521126760566e-05, "loss": 3.5687, "step": 5581 }, { "epoch": 6.26, "learning_rate": 2.355211267605634e-05, "loss": 3.3735, "step": 5582 }, { "epoch": 6.26, "learning_rate": 2.3550704225352115e-05, "loss": 3.2926, "step": 5583 }, { "epoch": 6.26, "learning_rate": 2.3549295774647888e-05, "loss": 3.2519, "step": 5584 }, { "epoch": 6.26, "learning_rate": 2.3547887323943664e-05, "loss": 3.3743, "step": 5585 }, { "epoch": 6.26, "learning_rate": 2.3546478873239437e-05, "loss": 3.3695, "step": 5586 }, { "epoch": 6.26, "learning_rate": 2.3545070422535214e-05, "loss": 3.2113, "step": 5587 }, { "epoch": 6.26, "learning_rate": 2.3543661971830987e-05, "loss": 3.1254, "step": 5588 }, { "epoch": 6.27, "learning_rate": 2.354225352112676e-05, "loss": 3.3141, "step": 5589 }, { "epoch": 6.27, "learning_rate": 2.3540845070422536e-05, "loss": 3.0385, "step": 5590 }, { "epoch": 6.27, "learning_rate": 2.353943661971831e-05, "loss": 3.3995, "step": 5591 }, { "epoch": 6.27, "learning_rate": 2.3538028169014085e-05, "loss": 3.0947, "step": 5592 }, { "epoch": 6.27, "learning_rate": 2.3536619718309858e-05, "loss": 3.2809, "step": 5593 }, { "epoch": 6.27, "learning_rate": 2.3535211267605634e-05, "loss": 3.0968, "step": 5594 }, { "epoch": 6.27, "learning_rate": 2.3533802816901407e-05, "loss": 3.2536, "step": 5595 }, { "epoch": 6.27, "learning_rate": 2.3532394366197184e-05, "loss": 3.1152, "step": 5596 }, { "epoch": 6.27, "learning_rate": 2.353098591549296e-05, "loss": 2.8365, "step": 5597 }, { "epoch": 6.28, "learning_rate": 2.3529577464788733e-05, "loss": 2.8742, "step": 5598 }, { "epoch": 6.28, "learning_rate": 2.352816901408451e-05, "loss": 3.0892, "step": 5599 }, { "epoch": 6.28, "learning_rate": 2.3526760563380282e-05, "loss": 2.6088, "step": 5600 }, { "epoch": 6.28, "learning_rate": 2.352535211267606e-05, "loss": 3.1313, "step": 5601 }, { "epoch": 6.28, "learning_rate": 2.352394366197183e-05, "loss": 2.6165, "step": 5602 }, { "epoch": 6.28, "learning_rate": 2.3522535211267608e-05, "loss": 4.394, "step": 5603 }, { "epoch": 6.28, "learning_rate": 2.352112676056338e-05, "loss": 4.3799, "step": 5604 }, { "epoch": 6.28, "learning_rate": 2.3519718309859154e-05, "loss": 4.2592, "step": 5605 }, { "epoch": 6.28, "learning_rate": 2.351830985915493e-05, "loss": 4.3643, "step": 5606 }, { "epoch": 6.29, "learning_rate": 2.3516901408450703e-05, "loss": 3.9459, "step": 5607 }, { "epoch": 6.29, "learning_rate": 2.351549295774648e-05, "loss": 3.8354, "step": 5608 }, { "epoch": 6.29, "learning_rate": 2.3514084507042252e-05, "loss": 3.5162, "step": 5609 }, { "epoch": 6.29, "learning_rate": 2.351267605633803e-05, "loss": 3.6009, "step": 5610 }, { "epoch": 6.29, "learning_rate": 2.3511267605633805e-05, "loss": 3.584, "step": 5611 }, { "epoch": 6.29, "learning_rate": 2.3509859154929578e-05, "loss": 3.5123, "step": 5612 }, { "epoch": 6.29, "learning_rate": 2.3508450704225354e-05, "loss": 3.783, "step": 5613 }, { "epoch": 6.29, "learning_rate": 2.3507042253521127e-05, "loss": 3.6244, "step": 5614 }, { "epoch": 6.29, "learning_rate": 2.3505633802816903e-05, "loss": 3.6072, "step": 5615 }, { "epoch": 6.3, "learning_rate": 2.3504225352112676e-05, "loss": 3.3014, "step": 5616 }, { "epoch": 6.3, "learning_rate": 2.3502816901408453e-05, "loss": 3.219, "step": 5617 }, { "epoch": 6.3, "learning_rate": 2.3501408450704226e-05, "loss": 3.5637, "step": 5618 }, { "epoch": 6.3, "learning_rate": 2.3500000000000002e-05, "loss": 3.4358, "step": 5619 }, { "epoch": 6.3, "learning_rate": 2.3498591549295775e-05, "loss": 3.3241, "step": 5620 }, { "epoch": 6.3, "learning_rate": 2.3497183098591548e-05, "loss": 3.5454, "step": 5621 }, { "epoch": 6.3, "learning_rate": 2.3495774647887324e-05, "loss": 3.4213, "step": 5622 }, { "epoch": 6.3, "learning_rate": 2.3494366197183097e-05, "loss": 3.4683, "step": 5623 }, { "epoch": 6.3, "learning_rate": 2.3492957746478874e-05, "loss": 3.2183, "step": 5624 }, { "epoch": 6.31, "learning_rate": 2.3491549295774646e-05, "loss": 3.5249, "step": 5625 }, { "epoch": 6.31, "learning_rate": 2.3490140845070426e-05, "loss": 3.1961, "step": 5626 }, { "epoch": 6.31, "learning_rate": 2.34887323943662e-05, "loss": 3.2144, "step": 5627 }, { "epoch": 6.31, "learning_rate": 2.3487323943661972e-05, "loss": 3.1929, "step": 5628 }, { "epoch": 6.31, "learning_rate": 2.348591549295775e-05, "loss": 3.3613, "step": 5629 }, { "epoch": 6.31, "learning_rate": 2.348450704225352e-05, "loss": 3.0766, "step": 5630 }, { "epoch": 6.31, "learning_rate": 2.3483098591549298e-05, "loss": 3.2805, "step": 5631 }, { "epoch": 6.31, "learning_rate": 2.348169014084507e-05, "loss": 3.3276, "step": 5632 }, { "epoch": 6.32, "learning_rate": 2.3480281690140847e-05, "loss": 3.2675, "step": 5633 }, { "epoch": 6.32, "learning_rate": 2.347887323943662e-05, "loss": 3.2386, "step": 5634 }, { "epoch": 6.32, "learning_rate": 2.3477464788732396e-05, "loss": 3.1971, "step": 5635 }, { "epoch": 6.32, "learning_rate": 2.347605633802817e-05, "loss": 3.0514, "step": 5636 }, { "epoch": 6.32, "learning_rate": 2.3474647887323942e-05, "loss": 3.2216, "step": 5637 }, { "epoch": 6.32, "learning_rate": 2.347323943661972e-05, "loss": 3.2112, "step": 5638 }, { "epoch": 6.32, "learning_rate": 2.347183098591549e-05, "loss": 3.2008, "step": 5639 }, { "epoch": 6.32, "learning_rate": 2.3470422535211268e-05, "loss": 3.2596, "step": 5640 }, { "epoch": 6.32, "learning_rate": 2.3469014084507044e-05, "loss": 2.9232, "step": 5641 }, { "epoch": 6.33, "learning_rate": 2.346760563380282e-05, "loss": 3.0523, "step": 5642 }, { "epoch": 6.33, "learning_rate": 2.3466197183098593e-05, "loss": 3.0611, "step": 5643 }, { "epoch": 6.33, "learning_rate": 2.3464788732394366e-05, "loss": 2.8672, "step": 5644 }, { "epoch": 6.33, "learning_rate": 2.3463380281690143e-05, "loss": 2.9203, "step": 5645 }, { "epoch": 6.33, "learning_rate": 2.3461971830985916e-05, "loss": 3.4617, "step": 5646 }, { "epoch": 6.33, "learning_rate": 2.3460563380281692e-05, "loss": 2.8467, "step": 5647 }, { "epoch": 6.33, "learning_rate": 2.3459154929577465e-05, "loss": 3.0635, "step": 5648 }, { "epoch": 6.33, "learning_rate": 2.345774647887324e-05, "loss": 2.8945, "step": 5649 }, { "epoch": 6.33, "learning_rate": 2.3456338028169014e-05, "loss": 3.1089, "step": 5650 }, { "epoch": 6.34, "learning_rate": 2.345492957746479e-05, "loss": 2.8154, "step": 5651 }, { "epoch": 6.34, "learning_rate": 2.3453521126760563e-05, "loss": 2.5715, "step": 5652 }, { "epoch": 6.34, "learning_rate": 2.3452112676056336e-05, "loss": 4.2344, "step": 5653 }, { "epoch": 6.34, "learning_rate": 2.3450704225352113e-05, "loss": 4.2479, "step": 5654 }, { "epoch": 6.34, "learning_rate": 2.3449295774647886e-05, "loss": 4.3803, "step": 5655 }, { "epoch": 6.34, "learning_rate": 2.3447887323943665e-05, "loss": 4.0466, "step": 5656 }, { "epoch": 6.34, "learning_rate": 2.3446478873239438e-05, "loss": 3.9693, "step": 5657 }, { "epoch": 6.34, "learning_rate": 2.3445070422535215e-05, "loss": 3.5082, "step": 5658 }, { "epoch": 6.34, "learning_rate": 2.3443661971830988e-05, "loss": 3.3387, "step": 5659 }, { "epoch": 6.35, "learning_rate": 2.344225352112676e-05, "loss": 3.3231, "step": 5660 }, { "epoch": 6.35, "learning_rate": 2.3440845070422537e-05, "loss": 3.5595, "step": 5661 }, { "epoch": 6.35, "learning_rate": 2.343943661971831e-05, "loss": 3.697, "step": 5662 }, { "epoch": 6.35, "learning_rate": 2.3438028169014086e-05, "loss": 3.7329, "step": 5663 }, { "epoch": 6.35, "learning_rate": 2.343661971830986e-05, "loss": 3.3485, "step": 5664 }, { "epoch": 6.35, "learning_rate": 2.3435211267605635e-05, "loss": 3.4931, "step": 5665 }, { "epoch": 6.35, "learning_rate": 2.3433802816901408e-05, "loss": 3.4233, "step": 5666 }, { "epoch": 6.35, "learning_rate": 2.343239436619718e-05, "loss": 3.234, "step": 5667 }, { "epoch": 6.35, "learning_rate": 2.3430985915492958e-05, "loss": 3.2479, "step": 5668 }, { "epoch": 6.36, "learning_rate": 2.342957746478873e-05, "loss": 3.2981, "step": 5669 }, { "epoch": 6.36, "learning_rate": 2.3428169014084507e-05, "loss": 3.1855, "step": 5670 }, { "epoch": 6.36, "learning_rate": 2.3426760563380283e-05, "loss": 3.0802, "step": 5671 }, { "epoch": 6.36, "learning_rate": 2.342535211267606e-05, "loss": 3.1197, "step": 5672 }, { "epoch": 6.36, "learning_rate": 2.3423943661971832e-05, "loss": 3.315, "step": 5673 }, { "epoch": 6.36, "learning_rate": 2.342253521126761e-05, "loss": 3.3986, "step": 5674 }, { "epoch": 6.36, "learning_rate": 2.3421126760563382e-05, "loss": 3.4919, "step": 5675 }, { "epoch": 6.36, "learning_rate": 2.3419718309859155e-05, "loss": 3.196, "step": 5676 }, { "epoch": 6.36, "learning_rate": 2.341830985915493e-05, "loss": 3.2533, "step": 5677 }, { "epoch": 6.37, "learning_rate": 2.3416901408450704e-05, "loss": 3.1118, "step": 5678 }, { "epoch": 6.37, "learning_rate": 2.341549295774648e-05, "loss": 3.2082, "step": 5679 }, { "epoch": 6.37, "learning_rate": 2.3414084507042253e-05, "loss": 3.054, "step": 5680 }, { "epoch": 6.37, "learning_rate": 2.341267605633803e-05, "loss": 3.2717, "step": 5681 }, { "epoch": 6.37, "learning_rate": 2.3411267605633802e-05, "loss": 3.0441, "step": 5682 }, { "epoch": 6.37, "learning_rate": 2.3409859154929575e-05, "loss": 3.3951, "step": 5683 }, { "epoch": 6.37, "learning_rate": 2.3408450704225352e-05, "loss": 3.1913, "step": 5684 }, { "epoch": 6.37, "learning_rate": 2.3407042253521125e-05, "loss": 3.3069, "step": 5685 }, { "epoch": 6.37, "learning_rate": 2.34056338028169e-05, "loss": 3.1427, "step": 5686 }, { "epoch": 6.38, "learning_rate": 2.3404225352112677e-05, "loss": 3.1283, "step": 5687 }, { "epoch": 6.38, "learning_rate": 2.3402816901408454e-05, "loss": 3.1044, "step": 5688 }, { "epoch": 6.38, "learning_rate": 2.3401408450704227e-05, "loss": 3.3142, "step": 5689 }, { "epoch": 6.38, "learning_rate": 2.3400000000000003e-05, "loss": 3.1338, "step": 5690 }, { "epoch": 6.38, "learning_rate": 2.3398591549295776e-05, "loss": 3.1377, "step": 5691 }, { "epoch": 6.38, "learning_rate": 2.339718309859155e-05, "loss": 2.9681, "step": 5692 }, { "epoch": 6.38, "learning_rate": 2.3395774647887325e-05, "loss": 3.1896, "step": 5693 }, { "epoch": 6.38, "learning_rate": 2.3394366197183098e-05, "loss": 2.8937, "step": 5694 }, { "epoch": 6.38, "learning_rate": 2.3392957746478874e-05, "loss": 3.0745, "step": 5695 }, { "epoch": 6.39, "learning_rate": 2.3391549295774647e-05, "loss": 3.0477, "step": 5696 }, { "epoch": 6.39, "learning_rate": 2.3390140845070424e-05, "loss": 3.0451, "step": 5697 }, { "epoch": 6.39, "learning_rate": 2.3388732394366197e-05, "loss": 2.8174, "step": 5698 }, { "epoch": 6.39, "learning_rate": 2.338732394366197e-05, "loss": 2.6473, "step": 5699 }, { "epoch": 6.39, "learning_rate": 2.3385915492957746e-05, "loss": 2.7738, "step": 5700 }, { "epoch": 6.39, "learning_rate": 2.338450704225352e-05, "loss": 2.9949, "step": 5701 }, { "epoch": 6.39, "learning_rate": 2.33830985915493e-05, "loss": 2.5429, "step": 5702 }, { "epoch": 6.39, "learning_rate": 2.338169014084507e-05, "loss": 4.756, "step": 5703 }, { "epoch": 6.39, "learning_rate": 2.3380281690140848e-05, "loss": 4.3761, "step": 5704 }, { "epoch": 6.4, "learning_rate": 2.337887323943662e-05, "loss": 4.3328, "step": 5705 }, { "epoch": 6.4, "learning_rate": 2.3377464788732397e-05, "loss": 4.2193, "step": 5706 }, { "epoch": 6.4, "learning_rate": 2.337605633802817e-05, "loss": 3.9833, "step": 5707 }, { "epoch": 6.4, "learning_rate": 2.3374647887323943e-05, "loss": 3.9107, "step": 5708 }, { "epoch": 6.4, "learning_rate": 2.337323943661972e-05, "loss": 3.5215, "step": 5709 }, { "epoch": 6.4, "learning_rate": 2.3371830985915492e-05, "loss": 3.5696, "step": 5710 }, { "epoch": 6.4, "learning_rate": 2.337042253521127e-05, "loss": 3.4475, "step": 5711 }, { "epoch": 6.4, "learning_rate": 2.336901408450704e-05, "loss": 3.3935, "step": 5712 }, { "epoch": 6.4, "learning_rate": 2.3367605633802818e-05, "loss": 3.5406, "step": 5713 }, { "epoch": 6.41, "learning_rate": 2.336619718309859e-05, "loss": 3.634, "step": 5714 }, { "epoch": 6.41, "learning_rate": 2.3364788732394364e-05, "loss": 3.4829, "step": 5715 }, { "epoch": 6.41, "learning_rate": 2.336338028169014e-05, "loss": 3.3891, "step": 5716 }, { "epoch": 6.41, "learning_rate": 2.3361971830985916e-05, "loss": 3.1541, "step": 5717 }, { "epoch": 6.41, "learning_rate": 2.3360563380281693e-05, "loss": 3.4782, "step": 5718 }, { "epoch": 6.41, "learning_rate": 2.3359154929577466e-05, "loss": 3.5859, "step": 5719 }, { "epoch": 6.41, "learning_rate": 2.3357746478873242e-05, "loss": 3.3886, "step": 5720 }, { "epoch": 6.41, "learning_rate": 2.3356338028169015e-05, "loss": 3.3048, "step": 5721 }, { "epoch": 6.41, "learning_rate": 2.335492957746479e-05, "loss": 3.3626, "step": 5722 }, { "epoch": 6.42, "learning_rate": 2.3353521126760564e-05, "loss": 3.2145, "step": 5723 }, { "epoch": 6.42, "learning_rate": 2.3352112676056337e-05, "loss": 3.2501, "step": 5724 }, { "epoch": 6.42, "learning_rate": 2.3350704225352114e-05, "loss": 3.2819, "step": 5725 }, { "epoch": 6.42, "learning_rate": 2.3349295774647887e-05, "loss": 3.0639, "step": 5726 }, { "epoch": 6.42, "learning_rate": 2.3347887323943663e-05, "loss": 3.3265, "step": 5727 }, { "epoch": 6.42, "learning_rate": 2.3346478873239436e-05, "loss": 2.973, "step": 5728 }, { "epoch": 6.42, "learning_rate": 2.3345070422535212e-05, "loss": 3.1683, "step": 5729 }, { "epoch": 6.42, "learning_rate": 2.3343661971830985e-05, "loss": 3.147, "step": 5730 }, { "epoch": 6.42, "learning_rate": 2.3342253521126758e-05, "loss": 3.3946, "step": 5731 }, { "epoch": 6.43, "learning_rate": 2.3340845070422538e-05, "loss": 2.9165, "step": 5732 }, { "epoch": 6.43, "learning_rate": 2.333943661971831e-05, "loss": 3.1404, "step": 5733 }, { "epoch": 6.43, "learning_rate": 2.3338028169014087e-05, "loss": 3.0055, "step": 5734 }, { "epoch": 6.43, "learning_rate": 2.333661971830986e-05, "loss": 3.0109, "step": 5735 }, { "epoch": 6.43, "learning_rate": 2.3335211267605636e-05, "loss": 2.9165, "step": 5736 }, { "epoch": 6.43, "learning_rate": 2.333380281690141e-05, "loss": 2.9525, "step": 5737 }, { "epoch": 6.43, "learning_rate": 2.3332394366197186e-05, "loss": 2.9152, "step": 5738 }, { "epoch": 6.43, "learning_rate": 2.333098591549296e-05, "loss": 3.0982, "step": 5739 }, { "epoch": 6.43, "learning_rate": 2.332957746478873e-05, "loss": 3.1108, "step": 5740 }, { "epoch": 6.44, "learning_rate": 2.3328169014084508e-05, "loss": 3.1338, "step": 5741 }, { "epoch": 6.44, "learning_rate": 2.332676056338028e-05, "loss": 2.8554, "step": 5742 }, { "epoch": 6.44, "learning_rate": 2.3325352112676057e-05, "loss": 3.1674, "step": 5743 }, { "epoch": 6.44, "learning_rate": 2.332394366197183e-05, "loss": 2.9741, "step": 5744 }, { "epoch": 6.44, "learning_rate": 2.3322535211267606e-05, "loss": 2.7686, "step": 5745 }, { "epoch": 6.44, "learning_rate": 2.332112676056338e-05, "loss": 2.8658, "step": 5746 }, { "epoch": 6.44, "learning_rate": 2.3319718309859156e-05, "loss": 2.6218, "step": 5747 }, { "epoch": 6.44, "learning_rate": 2.3318309859154932e-05, "loss": 3.0323, "step": 5748 }, { "epoch": 6.45, "learning_rate": 2.3316901408450705e-05, "loss": 2.4892, "step": 5749 }, { "epoch": 6.45, "learning_rate": 2.331549295774648e-05, "loss": 3.0287, "step": 5750 }, { "epoch": 6.45, "learning_rate": 2.3314084507042254e-05, "loss": 2.228, "step": 5751 }, { "epoch": 6.45, "learning_rate": 2.331267605633803e-05, "loss": 2.9016, "step": 5752 }, { "epoch": 6.45, "learning_rate": 2.3311267605633803e-05, "loss": 4.5567, "step": 5753 }, { "epoch": 6.45, "learning_rate": 2.330985915492958e-05, "loss": 4.6221, "step": 5754 }, { "epoch": 6.45, "learning_rate": 2.3308450704225353e-05, "loss": 4.3542, "step": 5755 }, { "epoch": 6.45, "learning_rate": 2.3307042253521126e-05, "loss": 4.3949, "step": 5756 }, { "epoch": 6.45, "learning_rate": 2.3305633802816902e-05, "loss": 4.0249, "step": 5757 }, { "epoch": 6.46, "learning_rate": 2.3304225352112675e-05, "loss": 3.9196, "step": 5758 }, { "epoch": 6.46, "learning_rate": 2.330281690140845e-05, "loss": 3.7849, "step": 5759 }, { "epoch": 6.46, "learning_rate": 2.3301408450704224e-05, "loss": 3.4834, "step": 5760 }, { "epoch": 6.46, "learning_rate": 2.33e-05, "loss": 3.4796, "step": 5761 }, { "epoch": 6.46, "learning_rate": 2.3298591549295777e-05, "loss": 3.6627, "step": 5762 }, { "epoch": 6.46, "learning_rate": 2.329718309859155e-05, "loss": 3.6494, "step": 5763 }, { "epoch": 6.46, "learning_rate": 2.3295774647887326e-05, "loss": 3.4391, "step": 5764 }, { "epoch": 6.46, "learning_rate": 2.32943661971831e-05, "loss": 3.5137, "step": 5765 }, { "epoch": 6.46, "learning_rate": 2.3292957746478875e-05, "loss": 3.4087, "step": 5766 }, { "epoch": 6.47, "learning_rate": 2.329154929577465e-05, "loss": 3.2557, "step": 5767 }, { "epoch": 6.47, "learning_rate": 2.3290140845070425e-05, "loss": 3.2642, "step": 5768 }, { "epoch": 6.47, "learning_rate": 2.3288732394366198e-05, "loss": 3.5753, "step": 5769 }, { "epoch": 6.47, "learning_rate": 2.328732394366197e-05, "loss": 3.3113, "step": 5770 }, { "epoch": 6.47, "learning_rate": 2.3285915492957747e-05, "loss": 3.2309, "step": 5771 }, { "epoch": 6.47, "learning_rate": 2.328450704225352e-05, "loss": 3.1687, "step": 5772 }, { "epoch": 6.47, "learning_rate": 2.3283098591549296e-05, "loss": 3.3191, "step": 5773 }, { "epoch": 6.47, "learning_rate": 2.328169014084507e-05, "loss": 3.029, "step": 5774 }, { "epoch": 6.47, "learning_rate": 2.3280281690140845e-05, "loss": 3.147, "step": 5775 }, { "epoch": 6.48, "learning_rate": 2.327887323943662e-05, "loss": 3.1578, "step": 5776 }, { "epoch": 6.48, "learning_rate": 2.3277464788732398e-05, "loss": 3.2065, "step": 5777 }, { "epoch": 6.48, "learning_rate": 2.327605633802817e-05, "loss": 3.6999, "step": 5778 }, { "epoch": 6.48, "learning_rate": 2.3274647887323944e-05, "loss": 3.3509, "step": 5779 }, { "epoch": 6.48, "learning_rate": 2.327323943661972e-05, "loss": 3.2884, "step": 5780 }, { "epoch": 6.48, "learning_rate": 2.3271830985915493e-05, "loss": 2.7936, "step": 5781 }, { "epoch": 6.48, "learning_rate": 2.327042253521127e-05, "loss": 2.9043, "step": 5782 }, { "epoch": 6.48, "learning_rate": 2.3269014084507043e-05, "loss": 3.2201, "step": 5783 }, { "epoch": 6.48, "learning_rate": 2.326760563380282e-05, "loss": 3.1818, "step": 5784 }, { "epoch": 6.49, "learning_rate": 2.3266197183098592e-05, "loss": 2.9879, "step": 5785 }, { "epoch": 6.49, "learning_rate": 2.3264788732394365e-05, "loss": 3.2407, "step": 5786 }, { "epoch": 6.49, "learning_rate": 2.326338028169014e-05, "loss": 3.103, "step": 5787 }, { "epoch": 6.49, "learning_rate": 2.3261971830985914e-05, "loss": 2.7721, "step": 5788 }, { "epoch": 6.49, "learning_rate": 2.326056338028169e-05, "loss": 3.1578, "step": 5789 }, { "epoch": 6.49, "learning_rate": 2.3259154929577463e-05, "loss": 2.878, "step": 5790 }, { "epoch": 6.49, "learning_rate": 2.325774647887324e-05, "loss": 2.961, "step": 5791 }, { "epoch": 6.49, "learning_rate": 2.3256338028169016e-05, "loss": 3.0601, "step": 5792 }, { "epoch": 6.49, "learning_rate": 2.3254929577464792e-05, "loss": 3.0078, "step": 5793 }, { "epoch": 6.5, "learning_rate": 2.3253521126760565e-05, "loss": 3.3198, "step": 5794 }, { "epoch": 6.5, "learning_rate": 2.3252112676056338e-05, "loss": 3.0117, "step": 5795 }, { "epoch": 6.5, "learning_rate": 2.3250704225352115e-05, "loss": 2.8363, "step": 5796 }, { "epoch": 6.5, "learning_rate": 2.3249295774647887e-05, "loss": 2.7877, "step": 5797 }, { "epoch": 6.5, "learning_rate": 2.3247887323943664e-05, "loss": 2.9114, "step": 5798 }, { "epoch": 6.5, "learning_rate": 2.3246478873239437e-05, "loss": 2.7695, "step": 5799 }, { "epoch": 6.5, "learning_rate": 2.3245070422535213e-05, "loss": 2.9361, "step": 5800 }, { "epoch": 6.5, "learning_rate": 2.3243661971830986e-05, "loss": 2.8659, "step": 5801 }, { "epoch": 6.5, "learning_rate": 2.324225352112676e-05, "loss": 2.519, "step": 5802 }, { "epoch": 6.51, "learning_rate": 2.3240845070422535e-05, "loss": 4.7973, "step": 5803 }, { "epoch": 6.51, "learning_rate": 2.3239436619718308e-05, "loss": 4.5449, "step": 5804 }, { "epoch": 6.51, "learning_rate": 2.3238028169014085e-05, "loss": 4.2807, "step": 5805 }, { "epoch": 6.51, "learning_rate": 2.3236619718309858e-05, "loss": 4.4273, "step": 5806 }, { "epoch": 6.51, "learning_rate": 2.3235211267605634e-05, "loss": 3.9665, "step": 5807 }, { "epoch": 6.51, "learning_rate": 2.323380281690141e-05, "loss": 3.7331, "step": 5808 }, { "epoch": 6.51, "learning_rate": 2.3232394366197187e-05, "loss": 3.573, "step": 5809 }, { "epoch": 6.51, "learning_rate": 2.323098591549296e-05, "loss": 3.6708, "step": 5810 }, { "epoch": 6.51, "learning_rate": 2.3229577464788732e-05, "loss": 3.2742, "step": 5811 }, { "epoch": 6.52, "learning_rate": 2.322816901408451e-05, "loss": 3.429, "step": 5812 }, { "epoch": 6.52, "learning_rate": 2.322676056338028e-05, "loss": 3.3951, "step": 5813 }, { "epoch": 6.52, "learning_rate": 2.3225352112676058e-05, "loss": 3.2143, "step": 5814 }, { "epoch": 6.52, "learning_rate": 2.322394366197183e-05, "loss": 3.4647, "step": 5815 }, { "epoch": 6.52, "learning_rate": 2.3222535211267607e-05, "loss": 3.1389, "step": 5816 }, { "epoch": 6.52, "learning_rate": 2.322112676056338e-05, "loss": 3.2872, "step": 5817 }, { "epoch": 6.52, "learning_rate": 2.3219718309859153e-05, "loss": 3.2529, "step": 5818 }, { "epoch": 6.52, "learning_rate": 2.321830985915493e-05, "loss": 3.1092, "step": 5819 }, { "epoch": 6.52, "learning_rate": 2.3216901408450702e-05, "loss": 2.9894, "step": 5820 }, { "epoch": 6.53, "learning_rate": 2.321549295774648e-05, "loss": 3.1947, "step": 5821 }, { "epoch": 6.53, "learning_rate": 2.3214084507042252e-05, "loss": 3.364, "step": 5822 }, { "epoch": 6.53, "learning_rate": 2.321267605633803e-05, "loss": 3.4075, "step": 5823 }, { "epoch": 6.53, "learning_rate": 2.3211267605633804e-05, "loss": 2.8188, "step": 5824 }, { "epoch": 6.53, "learning_rate": 2.320985915492958e-05, "loss": 3.1994, "step": 5825 }, { "epoch": 6.53, "learning_rate": 2.3208450704225354e-05, "loss": 3.1586, "step": 5826 }, { "epoch": 6.53, "learning_rate": 2.3207042253521127e-05, "loss": 3.4002, "step": 5827 }, { "epoch": 6.53, "learning_rate": 2.3205633802816903e-05, "loss": 3.0601, "step": 5828 }, { "epoch": 6.53, "learning_rate": 2.3204225352112676e-05, "loss": 3.3633, "step": 5829 }, { "epoch": 6.54, "learning_rate": 2.3202816901408452e-05, "loss": 3.0445, "step": 5830 }, { "epoch": 6.54, "learning_rate": 2.3201408450704225e-05, "loss": 3.1345, "step": 5831 }, { "epoch": 6.54, "learning_rate": 2.32e-05, "loss": 3.1869, "step": 5832 }, { "epoch": 6.54, "learning_rate": 2.3198591549295774e-05, "loss": 3.0415, "step": 5833 }, { "epoch": 6.54, "learning_rate": 2.3197183098591547e-05, "loss": 3.1525, "step": 5834 }, { "epoch": 6.54, "learning_rate": 2.3195774647887324e-05, "loss": 3.037, "step": 5835 }, { "epoch": 6.54, "learning_rate": 2.3194366197183097e-05, "loss": 2.8616, "step": 5836 }, { "epoch": 6.54, "learning_rate": 2.3192957746478873e-05, "loss": 2.9775, "step": 5837 }, { "epoch": 6.54, "learning_rate": 2.319154929577465e-05, "loss": 2.9628, "step": 5838 }, { "epoch": 6.55, "learning_rate": 2.3190140845070426e-05, "loss": 2.8434, "step": 5839 }, { "epoch": 6.55, "learning_rate": 2.31887323943662e-05, "loss": 2.9436, "step": 5840 }, { "epoch": 6.55, "learning_rate": 2.3187323943661975e-05, "loss": 2.8933, "step": 5841 }, { "epoch": 6.55, "learning_rate": 2.3185915492957748e-05, "loss": 2.7315, "step": 5842 }, { "epoch": 6.55, "learning_rate": 2.318450704225352e-05, "loss": 3.0014, "step": 5843 }, { "epoch": 6.55, "learning_rate": 2.3183098591549297e-05, "loss": 2.8889, "step": 5844 }, { "epoch": 6.55, "learning_rate": 2.318169014084507e-05, "loss": 3.0149, "step": 5845 }, { "epoch": 6.55, "learning_rate": 2.3180281690140846e-05, "loss": 2.8929, "step": 5846 }, { "epoch": 6.55, "learning_rate": 2.317887323943662e-05, "loss": 2.7866, "step": 5847 }, { "epoch": 6.56, "learning_rate": 2.3177464788732396e-05, "loss": 2.7714, "step": 5848 }, { "epoch": 6.56, "learning_rate": 2.317605633802817e-05, "loss": 2.1455, "step": 5849 }, { "epoch": 6.56, "learning_rate": 2.317464788732394e-05, "loss": 2.7297, "step": 5850 }, { "epoch": 6.56, "learning_rate": 2.3173239436619718e-05, "loss": 2.7649, "step": 5851 }, { "epoch": 6.56, "learning_rate": 2.317183098591549e-05, "loss": 2.1813, "step": 5852 }, { "epoch": 6.56, "learning_rate": 2.317042253521127e-05, "loss": 4.3577, "step": 5853 }, { "epoch": 6.56, "learning_rate": 2.3169014084507044e-05, "loss": 4.4328, "step": 5854 }, { "epoch": 6.56, "learning_rate": 2.316760563380282e-05, "loss": 4.3222, "step": 5855 }, { "epoch": 6.57, "learning_rate": 2.3166197183098593e-05, "loss": 4.1323, "step": 5856 }, { "epoch": 6.57, "learning_rate": 2.316478873239437e-05, "loss": 3.9979, "step": 5857 }, { "epoch": 6.57, "learning_rate": 2.3163380281690142e-05, "loss": 3.8293, "step": 5858 }, { "epoch": 6.57, "learning_rate": 2.3161971830985915e-05, "loss": 3.5417, "step": 5859 }, { "epoch": 6.57, "learning_rate": 2.316056338028169e-05, "loss": 3.5085, "step": 5860 }, { "epoch": 6.57, "learning_rate": 2.3159154929577464e-05, "loss": 3.2398, "step": 5861 }, { "epoch": 6.57, "learning_rate": 2.315774647887324e-05, "loss": 3.3534, "step": 5862 }, { "epoch": 6.57, "learning_rate": 2.3156338028169014e-05, "loss": 3.551, "step": 5863 }, { "epoch": 6.57, "learning_rate": 2.315492957746479e-05, "loss": 3.3829, "step": 5864 }, { "epoch": 6.58, "learning_rate": 2.3153521126760563e-05, "loss": 3.2308, "step": 5865 }, { "epoch": 6.58, "learning_rate": 2.3152112676056336e-05, "loss": 3.4735, "step": 5866 }, { "epoch": 6.58, "learning_rate": 2.3150704225352112e-05, "loss": 3.2498, "step": 5867 }, { "epoch": 6.58, "learning_rate": 2.314929577464789e-05, "loss": 3.0961, "step": 5868 }, { "epoch": 6.58, "learning_rate": 2.3147887323943665e-05, "loss": 3.0181, "step": 5869 }, { "epoch": 6.58, "learning_rate": 2.3146478873239438e-05, "loss": 3.3411, "step": 5870 }, { "epoch": 6.58, "learning_rate": 2.3145070422535214e-05, "loss": 3.1921, "step": 5871 }, { "epoch": 6.58, "learning_rate": 2.3143661971830987e-05, "loss": 3.1162, "step": 5872 }, { "epoch": 6.58, "learning_rate": 2.314225352112676e-05, "loss": 2.8908, "step": 5873 }, { "epoch": 6.59, "learning_rate": 2.3140845070422536e-05, "loss": 3.1247, "step": 5874 }, { "epoch": 6.59, "learning_rate": 2.313943661971831e-05, "loss": 3.09, "step": 5875 }, { "epoch": 6.59, "learning_rate": 2.3138028169014086e-05, "loss": 2.9104, "step": 5876 }, { "epoch": 6.59, "learning_rate": 2.313661971830986e-05, "loss": 3.0208, "step": 5877 }, { "epoch": 6.59, "learning_rate": 2.3135211267605635e-05, "loss": 3.1238, "step": 5878 }, { "epoch": 6.59, "learning_rate": 2.3133802816901408e-05, "loss": 3.0475, "step": 5879 }, { "epoch": 6.59, "learning_rate": 2.3132394366197184e-05, "loss": 3.1274, "step": 5880 }, { "epoch": 6.59, "learning_rate": 2.3130985915492957e-05, "loss": 3.0204, "step": 5881 }, { "epoch": 6.59, "learning_rate": 2.312957746478873e-05, "loss": 3.0056, "step": 5882 }, { "epoch": 6.6, "learning_rate": 2.312816901408451e-05, "loss": 3.0284, "step": 5883 }, { "epoch": 6.6, "learning_rate": 2.3126760563380283e-05, "loss": 2.8606, "step": 5884 }, { "epoch": 6.6, "learning_rate": 2.312535211267606e-05, "loss": 3.1398, "step": 5885 }, { "epoch": 6.6, "learning_rate": 2.3123943661971832e-05, "loss": 2.9096, "step": 5886 }, { "epoch": 6.6, "learning_rate": 2.3122535211267608e-05, "loss": 3.1287, "step": 5887 }, { "epoch": 6.6, "learning_rate": 2.312112676056338e-05, "loss": 2.8443, "step": 5888 }, { "epoch": 6.6, "learning_rate": 2.3119718309859154e-05, "loss": 3.0999, "step": 5889 }, { "epoch": 6.6, "learning_rate": 2.311830985915493e-05, "loss": 2.8588, "step": 5890 }, { "epoch": 6.6, "learning_rate": 2.3116901408450703e-05, "loss": 2.9229, "step": 5891 }, { "epoch": 6.61, "learning_rate": 2.311549295774648e-05, "loss": 3.0037, "step": 5892 }, { "epoch": 6.61, "learning_rate": 2.3114084507042253e-05, "loss": 2.9862, "step": 5893 }, { "epoch": 6.61, "learning_rate": 2.311267605633803e-05, "loss": 3.2303, "step": 5894 }, { "epoch": 6.61, "learning_rate": 2.3111267605633802e-05, "loss": 3.1273, "step": 5895 }, { "epoch": 6.61, "learning_rate": 2.3109859154929578e-05, "loss": 2.6959, "step": 5896 }, { "epoch": 6.61, "learning_rate": 2.310845070422535e-05, "loss": 2.8265, "step": 5897 }, { "epoch": 6.61, "learning_rate": 2.3107042253521128e-05, "loss": 2.8733, "step": 5898 }, { "epoch": 6.61, "learning_rate": 2.3105633802816904e-05, "loss": 2.7309, "step": 5899 }, { "epoch": 6.61, "learning_rate": 2.3104225352112677e-05, "loss": 2.6425, "step": 5900 }, { "epoch": 6.62, "learning_rate": 2.3102816901408453e-05, "loss": 2.411, "step": 5901 }, { "epoch": 6.62, "learning_rate": 2.3101408450704226e-05, "loss": 2.5389, "step": 5902 }, { "epoch": 6.62, "learning_rate": 2.3100000000000002e-05, "loss": 4.5676, "step": 5903 }, { "epoch": 6.62, "learning_rate": 2.3098591549295775e-05, "loss": 4.0558, "step": 5904 }, { "epoch": 6.62, "learning_rate": 2.309718309859155e-05, "loss": 3.999, "step": 5905 }, { "epoch": 6.62, "learning_rate": 2.3095774647887325e-05, "loss": 3.9417, "step": 5906 }, { "epoch": 6.62, "learning_rate": 2.3094366197183098e-05, "loss": 3.5868, "step": 5907 }, { "epoch": 6.62, "learning_rate": 2.3092957746478874e-05, "loss": 3.4757, "step": 5908 }, { "epoch": 6.62, "learning_rate": 2.3091549295774647e-05, "loss": 3.4893, "step": 5909 }, { "epoch": 6.63, "learning_rate": 2.3090140845070423e-05, "loss": 3.1428, "step": 5910 }, { "epoch": 6.63, "learning_rate": 2.3088732394366196e-05, "loss": 3.0924, "step": 5911 }, { "epoch": 6.63, "learning_rate": 2.3087323943661972e-05, "loss": 3.3437, "step": 5912 }, { "epoch": 6.63, "learning_rate": 2.308591549295775e-05, "loss": 3.4065, "step": 5913 }, { "epoch": 6.63, "learning_rate": 2.3084507042253522e-05, "loss": 3.3747, "step": 5914 }, { "epoch": 6.63, "learning_rate": 2.3083098591549298e-05, "loss": 3.3114, "step": 5915 }, { "epoch": 6.63, "learning_rate": 2.308169014084507e-05, "loss": 3.035, "step": 5916 }, { "epoch": 6.63, "learning_rate": 2.3080281690140847e-05, "loss": 3.1346, "step": 5917 }, { "epoch": 6.63, "learning_rate": 2.307887323943662e-05, "loss": 3.0602, "step": 5918 }, { "epoch": 6.64, "learning_rate": 2.3077464788732397e-05, "loss": 3.2389, "step": 5919 }, { "epoch": 6.64, "learning_rate": 2.307605633802817e-05, "loss": 3.1063, "step": 5920 }, { "epoch": 6.64, "learning_rate": 2.3074647887323943e-05, "loss": 3.073, "step": 5921 }, { "epoch": 6.64, "learning_rate": 2.307323943661972e-05, "loss": 3.1951, "step": 5922 }, { "epoch": 6.64, "learning_rate": 2.3071830985915492e-05, "loss": 3.1539, "step": 5923 }, { "epoch": 6.64, "learning_rate": 2.3070422535211268e-05, "loss": 3.114, "step": 5924 }, { "epoch": 6.64, "learning_rate": 2.306901408450704e-05, "loss": 3.0978, "step": 5925 }, { "epoch": 6.64, "learning_rate": 2.3067605633802817e-05, "loss": 3.1353, "step": 5926 }, { "epoch": 6.64, "learning_rate": 2.306619718309859e-05, "loss": 2.985, "step": 5927 }, { "epoch": 6.65, "learning_rate": 2.3064788732394367e-05, "loss": 2.9931, "step": 5928 }, { "epoch": 6.65, "learning_rate": 2.3063380281690143e-05, "loss": 3.0822, "step": 5929 }, { "epoch": 6.65, "learning_rate": 2.3061971830985916e-05, "loss": 3.1545, "step": 5930 }, { "epoch": 6.65, "learning_rate": 2.3060563380281692e-05, "loss": 2.8368, "step": 5931 }, { "epoch": 6.65, "learning_rate": 2.3059154929577465e-05, "loss": 2.7992, "step": 5932 }, { "epoch": 6.65, "learning_rate": 2.305774647887324e-05, "loss": 2.7472, "step": 5933 }, { "epoch": 6.65, "learning_rate": 2.3056338028169015e-05, "loss": 2.8515, "step": 5934 }, { "epoch": 6.65, "learning_rate": 2.305492957746479e-05, "loss": 2.8936, "step": 5935 }, { "epoch": 6.65, "learning_rate": 2.3053521126760564e-05, "loss": 2.8775, "step": 5936 }, { "epoch": 6.66, "learning_rate": 2.3052112676056337e-05, "loss": 2.9268, "step": 5937 }, { "epoch": 6.66, "learning_rate": 2.3050704225352113e-05, "loss": 2.9397, "step": 5938 }, { "epoch": 6.66, "learning_rate": 2.3049295774647886e-05, "loss": 2.7748, "step": 5939 }, { "epoch": 6.66, "learning_rate": 2.3047887323943662e-05, "loss": 2.7393, "step": 5940 }, { "epoch": 6.66, "learning_rate": 2.3046478873239435e-05, "loss": 2.7065, "step": 5941 }, { "epoch": 6.66, "learning_rate": 2.304507042253521e-05, "loss": 2.9234, "step": 5942 }, { "epoch": 6.66, "learning_rate": 2.3043661971830985e-05, "loss": 2.6635, "step": 5943 }, { "epoch": 6.66, "learning_rate": 2.3042253521126764e-05, "loss": 2.6708, "step": 5944 }, { "epoch": 6.66, "learning_rate": 2.3040845070422537e-05, "loss": 3.0, "step": 5945 }, { "epoch": 6.67, "learning_rate": 2.303943661971831e-05, "loss": 2.822, "step": 5946 }, { "epoch": 6.67, "learning_rate": 2.3038028169014086e-05, "loss": 2.7267, "step": 5947 }, { "epoch": 6.67, "learning_rate": 2.303661971830986e-05, "loss": 2.5733, "step": 5948 }, { "epoch": 6.67, "learning_rate": 2.3035211267605636e-05, "loss": 2.603, "step": 5949 }, { "epoch": 6.67, "learning_rate": 2.303380281690141e-05, "loss": 2.7354, "step": 5950 }, { "epoch": 6.67, "learning_rate": 2.3032394366197185e-05, "loss": 2.7629, "step": 5951 }, { "epoch": 6.67, "learning_rate": 2.3030985915492958e-05, "loss": 2.0627, "step": 5952 }, { "epoch": 6.67, "learning_rate": 2.302957746478873e-05, "loss": 4.5808, "step": 5953 }, { "epoch": 6.67, "learning_rate": 2.3028169014084507e-05, "loss": 4.2959, "step": 5954 }, { "epoch": 6.68, "learning_rate": 2.302676056338028e-05, "loss": 4.185, "step": 5955 }, { "epoch": 6.68, "learning_rate": 2.3025352112676057e-05, "loss": 4.0443, "step": 5956 }, { "epoch": 6.68, "learning_rate": 2.302394366197183e-05, "loss": 3.8888, "step": 5957 }, { "epoch": 6.68, "learning_rate": 2.3022535211267606e-05, "loss": 3.5635, "step": 5958 }, { "epoch": 6.68, "learning_rate": 2.3021126760563382e-05, "loss": 3.3531, "step": 5959 }, { "epoch": 6.68, "learning_rate": 2.301971830985916e-05, "loss": 3.3276, "step": 5960 }, { "epoch": 6.68, "learning_rate": 2.301830985915493e-05, "loss": 3.0733, "step": 5961 }, { "epoch": 6.68, "learning_rate": 2.3016901408450704e-05, "loss": 3.2954, "step": 5962 }, { "epoch": 6.68, "learning_rate": 2.301549295774648e-05, "loss": 3.1756, "step": 5963 }, { "epoch": 6.69, "learning_rate": 2.3014084507042254e-05, "loss": 3.2211, "step": 5964 }, { "epoch": 6.69, "learning_rate": 2.301267605633803e-05, "loss": 3.3135, "step": 5965 }, { "epoch": 6.69, "learning_rate": 2.3011267605633803e-05, "loss": 3.1279, "step": 5966 }, { "epoch": 6.69, "learning_rate": 2.300985915492958e-05, "loss": 2.9758, "step": 5967 }, { "epoch": 6.69, "learning_rate": 2.3008450704225352e-05, "loss": 3.0082, "step": 5968 }, { "epoch": 6.69, "learning_rate": 2.3007042253521125e-05, "loss": 3.0599, "step": 5969 }, { "epoch": 6.69, "learning_rate": 2.30056338028169e-05, "loss": 3.0247, "step": 5970 }, { "epoch": 6.69, "learning_rate": 2.3004225352112674e-05, "loss": 3.0734, "step": 5971 }, { "epoch": 6.7, "learning_rate": 2.300281690140845e-05, "loss": 3.2487, "step": 5972 }, { "epoch": 6.7, "learning_rate": 2.3001408450704224e-05, "loss": 3.0314, "step": 5973 }, { "epoch": 6.7, "learning_rate": 2.3000000000000003e-05, "loss": 2.865, "step": 5974 }, { "epoch": 6.7, "learning_rate": 2.2998591549295776e-05, "loss": 3.2138, "step": 5975 }, { "epoch": 6.7, "learning_rate": 2.2997183098591553e-05, "loss": 3.0808, "step": 5976 }, { "epoch": 6.7, "learning_rate": 2.2995774647887326e-05, "loss": 2.8276, "step": 5977 }, { "epoch": 6.7, "learning_rate": 2.29943661971831e-05, "loss": 2.9211, "step": 5978 }, { "epoch": 6.7, "learning_rate": 2.2992957746478875e-05, "loss": 2.9207, "step": 5979 }, { "epoch": 6.7, "learning_rate": 2.2991549295774648e-05, "loss": 2.7748, "step": 5980 }, { "epoch": 6.71, "learning_rate": 2.2990140845070424e-05, "loss": 3.2209, "step": 5981 }, { "epoch": 6.71, "learning_rate": 2.2988732394366197e-05, "loss": 3.1881, "step": 5982 }, { "epoch": 6.71, "learning_rate": 2.2987323943661973e-05, "loss": 2.7361, "step": 5983 }, { "epoch": 6.71, "learning_rate": 2.2985915492957746e-05, "loss": 3.0961, "step": 5984 }, { "epoch": 6.71, "learning_rate": 2.298450704225352e-05, "loss": 3.0407, "step": 5985 }, { "epoch": 6.71, "learning_rate": 2.2983098591549296e-05, "loss": 2.8887, "step": 5986 }, { "epoch": 6.71, "learning_rate": 2.298169014084507e-05, "loss": 2.6543, "step": 5987 }, { "epoch": 6.71, "learning_rate": 2.2980281690140845e-05, "loss": 2.8851, "step": 5988 }, { "epoch": 6.71, "learning_rate": 2.297887323943662e-05, "loss": 2.9247, "step": 5989 }, { "epoch": 6.72, "learning_rate": 2.2977464788732398e-05, "loss": 2.9428, "step": 5990 }, { "epoch": 6.72, "learning_rate": 2.297605633802817e-05, "loss": 2.7281, "step": 5991 }, { "epoch": 6.72, "learning_rate": 2.2974647887323943e-05, "loss": 2.863, "step": 5992 }, { "epoch": 6.72, "learning_rate": 2.297323943661972e-05, "loss": 2.8039, "step": 5993 }, { "epoch": 6.72, "learning_rate": 2.2971830985915493e-05, "loss": 3.0105, "step": 5994 }, { "epoch": 6.72, "learning_rate": 2.297042253521127e-05, "loss": 2.6008, "step": 5995 }, { "epoch": 6.72, "learning_rate": 2.2969014084507042e-05, "loss": 2.7667, "step": 5996 }, { "epoch": 6.72, "learning_rate": 2.296760563380282e-05, "loss": 2.6081, "step": 5997 }, { "epoch": 6.72, "learning_rate": 2.296619718309859e-05, "loss": 2.5924, "step": 5998 }, { "epoch": 6.73, "learning_rate": 2.2964788732394368e-05, "loss": 2.4398, "step": 5999 }, { "epoch": 6.73, "learning_rate": 2.296338028169014e-05, "loss": 2.1246, "step": 6000 }, { "epoch": 6.73, "eval_loss": 3.2089157104492188, "eval_runtime": 757.9959, "eval_samples_per_second": 3.486, "eval_steps_per_second": 0.437, "eval_wer": 0.502598175327251, "step": 6000 }, { "epoch": 6.73, "learning_rate": 2.2961971830985914e-05, "loss": 3.0225, "step": 6001 }, { "epoch": 6.73, "learning_rate": 2.296056338028169e-05, "loss": 2.2798, "step": 6002 }, { "epoch": 6.73, "learning_rate": 2.2959154929577463e-05, "loss": 4.3444, "step": 6003 }, { "epoch": 6.73, "learning_rate": 2.2957746478873243e-05, "loss": 4.0374, "step": 6004 }, { "epoch": 6.73, "learning_rate": 2.2956338028169015e-05, "loss": 3.9237, "step": 6005 }, { "epoch": 6.73, "learning_rate": 2.2954929577464792e-05, "loss": 3.6773, "step": 6006 }, { "epoch": 6.73, "learning_rate": 2.2953521126760565e-05, "loss": 3.4425, "step": 6007 }, { "epoch": 6.74, "learning_rate": 2.2952112676056338e-05, "loss": 3.2029, "step": 6008 }, { "epoch": 6.74, "learning_rate": 2.2950704225352114e-05, "loss": 3.1057, "step": 6009 }, { "epoch": 6.74, "learning_rate": 2.2949295774647887e-05, "loss": 3.1217, "step": 6010 }, { "epoch": 6.74, "learning_rate": 2.2947887323943663e-05, "loss": 3.0333, "step": 6011 }, { "epoch": 6.74, "learning_rate": 2.2946478873239436e-05, "loss": 3.0456, "step": 6012 }, { "epoch": 6.74, "learning_rate": 2.2945070422535213e-05, "loss": 3.1044, "step": 6013 }, { "epoch": 6.74, "learning_rate": 2.2943661971830986e-05, "loss": 3.0421, "step": 6014 }, { "epoch": 6.74, "learning_rate": 2.2942253521126762e-05, "loss": 3.0946, "step": 6015 }, { "epoch": 6.74, "learning_rate": 2.2940845070422535e-05, "loss": 3.3784, "step": 6016 }, { "epoch": 6.75, "learning_rate": 2.2939436619718308e-05, "loss": 3.0396, "step": 6017 }, { "epoch": 6.75, "learning_rate": 2.2938028169014084e-05, "loss": 3.0264, "step": 6018 }, { "epoch": 6.75, "learning_rate": 2.293661971830986e-05, "loss": 3.1114, "step": 6019 }, { "epoch": 6.75, "learning_rate": 2.2935211267605637e-05, "loss": 2.7579, "step": 6020 }, { "epoch": 6.75, "learning_rate": 2.293380281690141e-05, "loss": 3.0125, "step": 6021 }, { "epoch": 6.75, "learning_rate": 2.2932394366197186e-05, "loss": 3.0458, "step": 6022 }, { "epoch": 6.75, "learning_rate": 2.293098591549296e-05, "loss": 3.1184, "step": 6023 }, { "epoch": 6.75, "learning_rate": 2.2929577464788732e-05, "loss": 3.0331, "step": 6024 }, { "epoch": 6.75, "learning_rate": 2.2928169014084508e-05, "loss": 3.2833, "step": 6025 }, { "epoch": 6.76, "learning_rate": 2.292676056338028e-05, "loss": 3.1842, "step": 6026 }, { "epoch": 6.76, "learning_rate": 2.2925352112676058e-05, "loss": 3.1542, "step": 6027 }, { "epoch": 6.76, "learning_rate": 2.292394366197183e-05, "loss": 2.7982, "step": 6028 }, { "epoch": 6.76, "learning_rate": 2.2922535211267607e-05, "loss": 2.8931, "step": 6029 }, { "epoch": 6.76, "learning_rate": 2.292112676056338e-05, "loss": 2.7372, "step": 6030 }, { "epoch": 6.76, "learning_rate": 2.2919718309859156e-05, "loss": 3.1037, "step": 6031 }, { "epoch": 6.76, "learning_rate": 2.291830985915493e-05, "loss": 3.2674, "step": 6032 }, { "epoch": 6.76, "learning_rate": 2.2916901408450702e-05, "loss": 2.9403, "step": 6033 }, { "epoch": 6.76, "learning_rate": 2.2915492957746478e-05, "loss": 2.5826, "step": 6034 }, { "epoch": 6.77, "learning_rate": 2.2914084507042255e-05, "loss": 2.8286, "step": 6035 }, { "epoch": 6.77, "learning_rate": 2.291267605633803e-05, "loss": 2.6977, "step": 6036 }, { "epoch": 6.77, "learning_rate": 2.2911267605633804e-05, "loss": 2.9278, "step": 6037 }, { "epoch": 6.77, "learning_rate": 2.290985915492958e-05, "loss": 2.9188, "step": 6038 }, { "epoch": 6.77, "learning_rate": 2.2908450704225353e-05, "loss": 2.8492, "step": 6039 }, { "epoch": 6.77, "learning_rate": 2.2907042253521126e-05, "loss": 2.9643, "step": 6040 }, { "epoch": 6.77, "learning_rate": 2.2905633802816902e-05, "loss": 3.0241, "step": 6041 }, { "epoch": 6.77, "learning_rate": 2.2904225352112675e-05, "loss": 2.7395, "step": 6042 }, { "epoch": 6.77, "learning_rate": 2.2902816901408452e-05, "loss": 2.7231, "step": 6043 }, { "epoch": 6.78, "learning_rate": 2.2901408450704225e-05, "loss": 2.7694, "step": 6044 }, { "epoch": 6.78, "learning_rate": 2.29e-05, "loss": 2.5256, "step": 6045 }, { "epoch": 6.78, "learning_rate": 2.2898591549295774e-05, "loss": 2.7164, "step": 6046 }, { "epoch": 6.78, "learning_rate": 2.289718309859155e-05, "loss": 2.5342, "step": 6047 }, { "epoch": 6.78, "learning_rate": 2.2895774647887323e-05, "loss": 2.6325, "step": 6048 }, { "epoch": 6.78, "learning_rate": 2.2894366197183096e-05, "loss": 2.6991, "step": 6049 }, { "epoch": 6.78, "learning_rate": 2.2892957746478876e-05, "loss": 2.7911, "step": 6050 }, { "epoch": 6.78, "learning_rate": 2.289154929577465e-05, "loss": 2.5171, "step": 6051 }, { "epoch": 6.78, "learning_rate": 2.2890140845070425e-05, "loss": 2.7146, "step": 6052 }, { "epoch": 6.79, "learning_rate": 2.2888732394366198e-05, "loss": 4.4663, "step": 6053 }, { "epoch": 6.79, "learning_rate": 2.2887323943661974e-05, "loss": 4.3886, "step": 6054 }, { "epoch": 6.79, "learning_rate": 2.2885915492957747e-05, "loss": 3.7192, "step": 6055 }, { "epoch": 6.79, "learning_rate": 2.288450704225352e-05, "loss": 4.0114, "step": 6056 }, { "epoch": 6.79, "learning_rate": 2.2883098591549297e-05, "loss": 3.5237, "step": 6057 }, { "epoch": 6.79, "learning_rate": 2.288169014084507e-05, "loss": 3.3973, "step": 6058 }, { "epoch": 6.79, "learning_rate": 2.2880281690140846e-05, "loss": 3.3278, "step": 6059 }, { "epoch": 6.79, "learning_rate": 2.287887323943662e-05, "loss": 3.3812, "step": 6060 }, { "epoch": 6.79, "learning_rate": 2.2877464788732395e-05, "loss": 3.0489, "step": 6061 }, { "epoch": 6.8, "learning_rate": 2.2876056338028168e-05, "loss": 3.2093, "step": 6062 }, { "epoch": 6.8, "learning_rate": 2.2874647887323944e-05, "loss": 3.117, "step": 6063 }, { "epoch": 6.8, "learning_rate": 2.2873239436619717e-05, "loss": 2.8643, "step": 6064 }, { "epoch": 6.8, "learning_rate": 2.2871830985915494e-05, "loss": 2.8848, "step": 6065 }, { "epoch": 6.8, "learning_rate": 2.287042253521127e-05, "loss": 2.9299, "step": 6066 }, { "epoch": 6.8, "learning_rate": 2.2869014084507043e-05, "loss": 2.934, "step": 6067 }, { "epoch": 6.8, "learning_rate": 2.286760563380282e-05, "loss": 2.9693, "step": 6068 }, { "epoch": 6.8, "learning_rate": 2.2866197183098592e-05, "loss": 3.1748, "step": 6069 }, { "epoch": 6.8, "learning_rate": 2.286478873239437e-05, "loss": 2.9334, "step": 6070 }, { "epoch": 6.81, "learning_rate": 2.286338028169014e-05, "loss": 2.9286, "step": 6071 }, { "epoch": 6.81, "learning_rate": 2.2861971830985914e-05, "loss": 2.9602, "step": 6072 }, { "epoch": 6.81, "learning_rate": 2.286056338028169e-05, "loss": 3.0425, "step": 6073 }, { "epoch": 6.81, "learning_rate": 2.2859154929577464e-05, "loss": 2.6178, "step": 6074 }, { "epoch": 6.81, "learning_rate": 2.285774647887324e-05, "loss": 2.9567, "step": 6075 }, { "epoch": 6.81, "learning_rate": 2.2856338028169013e-05, "loss": 3.2265, "step": 6076 }, { "epoch": 6.81, "learning_rate": 2.285492957746479e-05, "loss": 2.947, "step": 6077 }, { "epoch": 6.81, "learning_rate": 2.2853521126760562e-05, "loss": 2.806, "step": 6078 }, { "epoch": 6.82, "learning_rate": 2.285211267605634e-05, "loss": 3.053, "step": 6079 }, { "epoch": 6.82, "learning_rate": 2.2850704225352115e-05, "loss": 2.8953, "step": 6080 }, { "epoch": 6.82, "learning_rate": 2.2849295774647888e-05, "loss": 3.0744, "step": 6081 }, { "epoch": 6.82, "learning_rate": 2.2847887323943664e-05, "loss": 2.6536, "step": 6082 }, { "epoch": 6.82, "learning_rate": 2.2846478873239437e-05, "loss": 2.6975, "step": 6083 }, { "epoch": 6.82, "learning_rate": 2.2845070422535214e-05, "loss": 2.7851, "step": 6084 }, { "epoch": 6.82, "learning_rate": 2.2843661971830986e-05, "loss": 2.9705, "step": 6085 }, { "epoch": 6.82, "learning_rate": 2.2842253521126763e-05, "loss": 2.8045, "step": 6086 }, { "epoch": 6.82, "learning_rate": 2.2840845070422536e-05, "loss": 2.9216, "step": 6087 }, { "epoch": 6.83, "learning_rate": 2.283943661971831e-05, "loss": 2.6241, "step": 6088 }, { "epoch": 6.83, "learning_rate": 2.2838028169014085e-05, "loss": 2.8547, "step": 6089 }, { "epoch": 6.83, "learning_rate": 2.2836619718309858e-05, "loss": 2.8935, "step": 6090 }, { "epoch": 6.83, "learning_rate": 2.2835211267605634e-05, "loss": 2.6688, "step": 6091 }, { "epoch": 6.83, "learning_rate": 2.2833802816901407e-05, "loss": 2.8672, "step": 6092 }, { "epoch": 6.83, "learning_rate": 2.2832394366197184e-05, "loss": 2.8608, "step": 6093 }, { "epoch": 6.83, "learning_rate": 2.2830985915492957e-05, "loss": 2.6427, "step": 6094 }, { "epoch": 6.83, "learning_rate": 2.2829577464788733e-05, "loss": 2.5692, "step": 6095 }, { "epoch": 6.83, "learning_rate": 2.282816901408451e-05, "loss": 2.693, "step": 6096 }, { "epoch": 6.84, "learning_rate": 2.2826760563380282e-05, "loss": 2.437, "step": 6097 }, { "epoch": 6.84, "learning_rate": 2.282535211267606e-05, "loss": 2.309, "step": 6098 }, { "epoch": 6.84, "learning_rate": 2.282394366197183e-05, "loss": 2.5459, "step": 6099 }, { "epoch": 6.84, "learning_rate": 2.2822535211267608e-05, "loss": 2.5526, "step": 6100 }, { "epoch": 6.84, "learning_rate": 2.282112676056338e-05, "loss": 2.5057, "step": 6101 }, { "epoch": 6.84, "learning_rate": 2.2819718309859157e-05, "loss": 2.2641, "step": 6102 }, { "epoch": 6.84, "learning_rate": 2.281830985915493e-05, "loss": 4.0621, "step": 6103 }, { "epoch": 6.84, "learning_rate": 2.2816901408450703e-05, "loss": 4.0526, "step": 6104 }, { "epoch": 6.84, "learning_rate": 2.281549295774648e-05, "loss": 3.8222, "step": 6105 }, { "epoch": 6.85, "learning_rate": 2.2814084507042252e-05, "loss": 3.98, "step": 6106 }, { "epoch": 6.85, "learning_rate": 2.281267605633803e-05, "loss": 3.3882, "step": 6107 }, { "epoch": 6.85, "learning_rate": 2.28112676056338e-05, "loss": 3.4555, "step": 6108 }, { "epoch": 6.85, "learning_rate": 2.2809859154929578e-05, "loss": 3.2423, "step": 6109 }, { "epoch": 6.85, "learning_rate": 2.2808450704225354e-05, "loss": 3.0877, "step": 6110 }, { "epoch": 6.85, "learning_rate": 2.2807042253521127e-05, "loss": 2.8998, "step": 6111 }, { "epoch": 6.85, "learning_rate": 2.2805633802816903e-05, "loss": 3.0331, "step": 6112 }, { "epoch": 6.85, "learning_rate": 2.2804225352112676e-05, "loss": 2.9169, "step": 6113 }, { "epoch": 6.85, "learning_rate": 2.2802816901408453e-05, "loss": 2.9031, "step": 6114 }, { "epoch": 6.86, "learning_rate": 2.2801408450704226e-05, "loss": 3.2681, "step": 6115 }, { "epoch": 6.86, "learning_rate": 2.2800000000000002e-05, "loss": 2.8868, "step": 6116 }, { "epoch": 6.86, "learning_rate": 2.2798591549295775e-05, "loss": 2.8943, "step": 6117 }, { "epoch": 6.86, "learning_rate": 2.279718309859155e-05, "loss": 2.9313, "step": 6118 }, { "epoch": 6.86, "learning_rate": 2.2795774647887324e-05, "loss": 2.9416, "step": 6119 }, { "epoch": 6.86, "learning_rate": 2.2794366197183097e-05, "loss": 2.8569, "step": 6120 }, { "epoch": 6.86, "learning_rate": 2.2792957746478873e-05, "loss": 3.0039, "step": 6121 }, { "epoch": 6.86, "learning_rate": 2.2791549295774646e-05, "loss": 2.8095, "step": 6122 }, { "epoch": 6.86, "learning_rate": 2.2790140845070423e-05, "loss": 2.9063, "step": 6123 }, { "epoch": 6.87, "learning_rate": 2.2788732394366196e-05, "loss": 3.1485, "step": 6124 }, { "epoch": 6.87, "learning_rate": 2.2787323943661975e-05, "loss": 2.9193, "step": 6125 }, { "epoch": 6.87, "learning_rate": 2.278591549295775e-05, "loss": 2.5417, "step": 6126 }, { "epoch": 6.87, "learning_rate": 2.278450704225352e-05, "loss": 2.7373, "step": 6127 }, { "epoch": 6.87, "learning_rate": 2.2783098591549298e-05, "loss": 2.9732, "step": 6128 }, { "epoch": 6.87, "learning_rate": 2.278169014084507e-05, "loss": 2.8853, "step": 6129 }, { "epoch": 6.87, "learning_rate": 2.2780281690140847e-05, "loss": 2.8312, "step": 6130 }, { "epoch": 6.87, "learning_rate": 2.277887323943662e-05, "loss": 3.0107, "step": 6131 }, { "epoch": 6.87, "learning_rate": 2.2777464788732396e-05, "loss": 2.913, "step": 6132 }, { "epoch": 6.88, "learning_rate": 2.277605633802817e-05, "loss": 2.9051, "step": 6133 }, { "epoch": 6.88, "learning_rate": 2.2774647887323945e-05, "loss": 2.846, "step": 6134 }, { "epoch": 6.88, "learning_rate": 2.277323943661972e-05, "loss": 2.6275, "step": 6135 }, { "epoch": 6.88, "learning_rate": 2.277183098591549e-05, "loss": 2.8521, "step": 6136 }, { "epoch": 6.88, "learning_rate": 2.2770422535211268e-05, "loss": 3.0793, "step": 6137 }, { "epoch": 6.88, "learning_rate": 2.276901408450704e-05, "loss": 2.7113, "step": 6138 }, { "epoch": 6.88, "learning_rate": 2.2767605633802817e-05, "loss": 2.7392, "step": 6139 }, { "epoch": 6.88, "learning_rate": 2.2766197183098593e-05, "loss": 2.8631, "step": 6140 }, { "epoch": 6.88, "learning_rate": 2.276478873239437e-05, "loss": 2.6469, "step": 6141 }, { "epoch": 6.89, "learning_rate": 2.2763380281690143e-05, "loss": 2.5768, "step": 6142 }, { "epoch": 6.89, "learning_rate": 2.2761971830985915e-05, "loss": 2.729, "step": 6143 }, { "epoch": 6.89, "learning_rate": 2.2760563380281692e-05, "loss": 2.5929, "step": 6144 }, { "epoch": 6.89, "learning_rate": 2.2759154929577465e-05, "loss": 2.4253, "step": 6145 }, { "epoch": 6.89, "learning_rate": 2.275774647887324e-05, "loss": 2.8139, "step": 6146 }, { "epoch": 6.89, "learning_rate": 2.2756338028169014e-05, "loss": 2.6757, "step": 6147 }, { "epoch": 6.89, "learning_rate": 2.275492957746479e-05, "loss": 2.597, "step": 6148 }, { "epoch": 6.89, "learning_rate": 2.2753521126760563e-05, "loss": 2.5337, "step": 6149 }, { "epoch": 6.89, "learning_rate": 2.275211267605634e-05, "loss": 2.3712, "step": 6150 }, { "epoch": 6.9, "learning_rate": 2.2750704225352113e-05, "loss": 2.3932, "step": 6151 }, { "epoch": 6.9, "learning_rate": 2.2749295774647885e-05, "loss": 2.2498, "step": 6152 }, { "epoch": 6.9, "learning_rate": 2.2747887323943662e-05, "loss": 3.9766, "step": 6153 }, { "epoch": 6.9, "learning_rate": 2.2746478873239435e-05, "loss": 4.008, "step": 6154 }, { "epoch": 6.9, "learning_rate": 2.274507042253521e-05, "loss": 3.5352, "step": 6155 }, { "epoch": 6.9, "learning_rate": 2.2743661971830987e-05, "loss": 3.7403, "step": 6156 }, { "epoch": 6.9, "learning_rate": 2.2742253521126764e-05, "loss": 3.1619, "step": 6157 }, { "epoch": 6.9, "learning_rate": 2.2740845070422537e-05, "loss": 2.9545, "step": 6158 }, { "epoch": 6.9, "learning_rate": 2.273943661971831e-05, "loss": 3.1332, "step": 6159 }, { "epoch": 6.91, "learning_rate": 2.2738028169014086e-05, "loss": 3.1118, "step": 6160 }, { "epoch": 6.91, "learning_rate": 2.273661971830986e-05, "loss": 3.0142, "step": 6161 }, { "epoch": 6.91, "learning_rate": 2.2735211267605635e-05, "loss": 3.089, "step": 6162 }, { "epoch": 6.91, "learning_rate": 2.2733802816901408e-05, "loss": 3.0185, "step": 6163 }, { "epoch": 6.91, "learning_rate": 2.2732394366197185e-05, "loss": 2.9828, "step": 6164 }, { "epoch": 6.91, "learning_rate": 2.2730985915492957e-05, "loss": 3.1457, "step": 6165 }, { "epoch": 6.91, "learning_rate": 2.2729577464788734e-05, "loss": 2.827, "step": 6166 }, { "epoch": 6.91, "learning_rate": 2.2728169014084507e-05, "loss": 2.9493, "step": 6167 }, { "epoch": 6.91, "learning_rate": 2.272676056338028e-05, "loss": 2.894, "step": 6168 }, { "epoch": 6.92, "learning_rate": 2.2725352112676056e-05, "loss": 2.8426, "step": 6169 }, { "epoch": 6.92, "learning_rate": 2.272394366197183e-05, "loss": 2.8087, "step": 6170 }, { "epoch": 6.92, "learning_rate": 2.272253521126761e-05, "loss": 2.9362, "step": 6171 }, { "epoch": 6.92, "learning_rate": 2.272112676056338e-05, "loss": 3.1023, "step": 6172 }, { "epoch": 6.92, "learning_rate": 2.2719718309859158e-05, "loss": 2.8828, "step": 6173 }, { "epoch": 6.92, "learning_rate": 2.271830985915493e-05, "loss": 2.8676, "step": 6174 }, { "epoch": 6.92, "learning_rate": 2.2716901408450704e-05, "loss": 2.7214, "step": 6175 }, { "epoch": 6.92, "learning_rate": 2.271549295774648e-05, "loss": 2.6873, "step": 6176 }, { "epoch": 6.92, "learning_rate": 2.2714084507042253e-05, "loss": 2.6767, "step": 6177 }, { "epoch": 6.93, "learning_rate": 2.271267605633803e-05, "loss": 2.8443, "step": 6178 }, { "epoch": 6.93, "learning_rate": 2.2711267605633802e-05, "loss": 2.5998, "step": 6179 }, { "epoch": 6.93, "learning_rate": 2.270985915492958e-05, "loss": 2.7091, "step": 6180 }, { "epoch": 6.93, "learning_rate": 2.270845070422535e-05, "loss": 2.6224, "step": 6181 }, { "epoch": 6.93, "learning_rate": 2.2707042253521128e-05, "loss": 2.6662, "step": 6182 }, { "epoch": 6.93, "learning_rate": 2.27056338028169e-05, "loss": 2.739, "step": 6183 }, { "epoch": 6.93, "learning_rate": 2.2704225352112674e-05, "loss": 2.6897, "step": 6184 }, { "epoch": 6.93, "learning_rate": 2.270281690140845e-05, "loss": 2.64, "step": 6185 }, { "epoch": 6.93, "learning_rate": 2.2701408450704227e-05, "loss": 2.7232, "step": 6186 }, { "epoch": 6.94, "learning_rate": 2.2700000000000003e-05, "loss": 2.6733, "step": 6187 }, { "epoch": 6.94, "learning_rate": 2.2698591549295776e-05, "loss": 2.8416, "step": 6188 }, { "epoch": 6.94, "learning_rate": 2.2697183098591552e-05, "loss": 2.4966, "step": 6189 }, { "epoch": 6.94, "learning_rate": 2.2695774647887325e-05, "loss": 2.6169, "step": 6190 }, { "epoch": 6.94, "learning_rate": 2.2694366197183098e-05, "loss": 2.8284, "step": 6191 }, { "epoch": 6.94, "learning_rate": 2.2692957746478874e-05, "loss": 2.71, "step": 6192 }, { "epoch": 6.94, "learning_rate": 2.2691549295774647e-05, "loss": 2.714, "step": 6193 }, { "epoch": 6.94, "learning_rate": 2.2690140845070424e-05, "loss": 2.7228, "step": 6194 }, { "epoch": 6.95, "learning_rate": 2.2688732394366197e-05, "loss": 2.7043, "step": 6195 }, { "epoch": 6.95, "learning_rate": 2.2687323943661973e-05, "loss": 2.5404, "step": 6196 }, { "epoch": 6.95, "learning_rate": 2.2685915492957746e-05, "loss": 2.5894, "step": 6197 }, { "epoch": 6.95, "learning_rate": 2.2684507042253522e-05, "loss": 2.3896, "step": 6198 }, { "epoch": 6.95, "learning_rate": 2.2683098591549295e-05, "loss": 2.4522, "step": 6199 }, { "epoch": 6.95, "learning_rate": 2.2681690140845068e-05, "loss": 2.3048, "step": 6200 }, { "epoch": 6.95, "learning_rate": 2.2680281690140848e-05, "loss": 2.7091, "step": 6201 }, { "epoch": 6.95, "learning_rate": 2.267887323943662e-05, "loss": 2.572, "step": 6202 }, { "epoch": 6.95, "learning_rate": 2.2677464788732397e-05, "loss": 4.1767, "step": 6203 }, { "epoch": 6.96, "learning_rate": 2.267605633802817e-05, "loss": 4.1412, "step": 6204 }, { "epoch": 6.96, "learning_rate": 2.2674647887323946e-05, "loss": 3.9944, "step": 6205 }, { "epoch": 6.96, "learning_rate": 2.267323943661972e-05, "loss": 3.6106, "step": 6206 }, { "epoch": 6.96, "learning_rate": 2.2671830985915492e-05, "loss": 3.6757, "step": 6207 }, { "epoch": 6.96, "learning_rate": 2.267042253521127e-05, "loss": 3.2515, "step": 6208 }, { "epoch": 6.96, "learning_rate": 2.266901408450704e-05, "loss": 2.9833, "step": 6209 }, { "epoch": 6.96, "learning_rate": 2.2667605633802818e-05, "loss": 2.9323, "step": 6210 }, { "epoch": 6.96, "learning_rate": 2.266619718309859e-05, "loss": 2.6882, "step": 6211 }, { "epoch": 6.96, "learning_rate": 2.2664788732394367e-05, "loss": 2.9945, "step": 6212 }, { "epoch": 6.97, "learning_rate": 2.266338028169014e-05, "loss": 2.887, "step": 6213 }, { "epoch": 6.97, "learning_rate": 2.2661971830985916e-05, "loss": 2.7273, "step": 6214 }, { "epoch": 6.97, "learning_rate": 2.266056338028169e-05, "loss": 3.2419, "step": 6215 }, { "epoch": 6.97, "learning_rate": 2.2659154929577466e-05, "loss": 3.1205, "step": 6216 }, { "epoch": 6.97, "learning_rate": 2.2657746478873242e-05, "loss": 2.8663, "step": 6217 }, { "epoch": 6.97, "learning_rate": 2.2656338028169015e-05, "loss": 3.0583, "step": 6218 }, { "epoch": 6.97, "learning_rate": 2.265492957746479e-05, "loss": 2.7261, "step": 6219 }, { "epoch": 6.97, "learning_rate": 2.2653521126760564e-05, "loss": 2.5909, "step": 6220 }, { "epoch": 6.97, "learning_rate": 2.265211267605634e-05, "loss": 2.5291, "step": 6221 }, { "epoch": 6.98, "learning_rate": 2.2650704225352114e-05, "loss": 2.7746, "step": 6222 }, { "epoch": 6.98, "learning_rate": 2.2649295774647886e-05, "loss": 2.6935, "step": 6223 }, { "epoch": 6.98, "learning_rate": 2.2647887323943663e-05, "loss": 2.8828, "step": 6224 }, { "epoch": 6.98, "learning_rate": 2.2646478873239436e-05, "loss": 2.6538, "step": 6225 }, { "epoch": 6.98, "learning_rate": 2.2645070422535212e-05, "loss": 2.9116, "step": 6226 }, { "epoch": 6.98, "learning_rate": 2.2643661971830985e-05, "loss": 2.6235, "step": 6227 }, { "epoch": 6.98, "learning_rate": 2.264225352112676e-05, "loss": 2.5785, "step": 6228 }, { "epoch": 6.98, "learning_rate": 2.2640845070422534e-05, "loss": 2.8152, "step": 6229 }, { "epoch": 6.98, "learning_rate": 2.2639436619718307e-05, "loss": 2.6865, "step": 6230 }, { "epoch": 6.99, "learning_rate": 2.2638028169014087e-05, "loss": 2.4765, "step": 6231 }, { "epoch": 6.99, "learning_rate": 2.263661971830986e-05, "loss": 2.7304, "step": 6232 }, { "epoch": 6.99, "learning_rate": 2.2635211267605636e-05, "loss": 2.8857, "step": 6233 }, { "epoch": 6.99, "learning_rate": 2.263380281690141e-05, "loss": 2.6727, "step": 6234 }, { "epoch": 6.99, "learning_rate": 2.2632394366197185e-05, "loss": 2.6945, "step": 6235 }, { "epoch": 6.99, "learning_rate": 2.263098591549296e-05, "loss": 2.6728, "step": 6236 }, { "epoch": 6.99, "learning_rate": 2.2629577464788735e-05, "loss": 2.395, "step": 6237 }, { "epoch": 6.99, "learning_rate": 2.2628169014084508e-05, "loss": 2.7139, "step": 6238 }, { "epoch": 6.99, "learning_rate": 2.262676056338028e-05, "loss": 2.4341, "step": 6239 }, { "epoch": 7.0, "learning_rate": 2.2625352112676057e-05, "loss": 2.763, "step": 6240 }, { "epoch": 7.0, "learning_rate": 2.262394366197183e-05, "loss": 2.3162, "step": 6241 }, { "epoch": 7.0, "learning_rate": 2.2622535211267606e-05, "loss": 2.5839, "step": 6242 }, { "epoch": 7.0, "learning_rate": 2.262112676056338e-05, "loss": 2.6279, "step": 6243 }, { "epoch": 7.0, "learning_rate": 2.2619718309859156e-05, "loss": 2.2494, "step": 6244 }, { "epoch": 7.0, "learning_rate": 2.261830985915493e-05, "loss": 3.803, "step": 6245 }, { "epoch": 7.0, "learning_rate": 2.2616901408450705e-05, "loss": 3.6515, "step": 6246 }, { "epoch": 7.0, "learning_rate": 2.261549295774648e-05, "loss": 3.459, "step": 6247 }, { "epoch": 7.0, "learning_rate": 2.2614084507042254e-05, "loss": 3.1065, "step": 6248 }, { "epoch": 7.01, "learning_rate": 2.261267605633803e-05, "loss": 3.512, "step": 6249 }, { "epoch": 7.01, "learning_rate": 2.2611267605633803e-05, "loss": 3.0518, "step": 6250 }, { "epoch": 7.01, "learning_rate": 2.260985915492958e-05, "loss": 2.889, "step": 6251 }, { "epoch": 7.01, "learning_rate": 2.2608450704225353e-05, "loss": 3.1375, "step": 6252 }, { "epoch": 7.01, "learning_rate": 2.260704225352113e-05, "loss": 2.728, "step": 6253 }, { "epoch": 7.01, "learning_rate": 2.2605633802816902e-05, "loss": 3.0106, "step": 6254 }, { "epoch": 7.01, "learning_rate": 2.2604225352112675e-05, "loss": 2.9049, "step": 6255 }, { "epoch": 7.01, "learning_rate": 2.260281690140845e-05, "loss": 2.8725, "step": 6256 }, { "epoch": 7.01, "learning_rate": 2.2601408450704224e-05, "loss": 2.807, "step": 6257 }, { "epoch": 7.02, "learning_rate": 2.26e-05, "loss": 2.6873, "step": 6258 }, { "epoch": 7.02, "learning_rate": 2.2598591549295773e-05, "loss": 2.7466, "step": 6259 }, { "epoch": 7.02, "learning_rate": 2.259718309859155e-05, "loss": 2.7302, "step": 6260 }, { "epoch": 7.02, "learning_rate": 2.2595774647887326e-05, "loss": 2.5968, "step": 6261 }, { "epoch": 7.02, "learning_rate": 2.25943661971831e-05, "loss": 2.4105, "step": 6262 }, { "epoch": 7.02, "learning_rate": 2.2592957746478875e-05, "loss": 2.6812, "step": 6263 }, { "epoch": 7.02, "learning_rate": 2.2591549295774648e-05, "loss": 2.9578, "step": 6264 }, { "epoch": 7.02, "learning_rate": 2.2590140845070425e-05, "loss": 2.8214, "step": 6265 }, { "epoch": 7.02, "learning_rate": 2.2588732394366198e-05, "loss": 2.8248, "step": 6266 }, { "epoch": 7.03, "learning_rate": 2.2587323943661974e-05, "loss": 2.618, "step": 6267 }, { "epoch": 7.03, "learning_rate": 2.2585915492957747e-05, "loss": 2.5584, "step": 6268 }, { "epoch": 7.03, "learning_rate": 2.2584507042253523e-05, "loss": 2.6478, "step": 6269 }, { "epoch": 7.03, "learning_rate": 2.2583098591549296e-05, "loss": 2.598, "step": 6270 }, { "epoch": 7.03, "learning_rate": 2.258169014084507e-05, "loss": 2.7033, "step": 6271 }, { "epoch": 7.03, "learning_rate": 2.2580281690140845e-05, "loss": 2.602, "step": 6272 }, { "epoch": 7.03, "learning_rate": 2.257887323943662e-05, "loss": 2.6225, "step": 6273 }, { "epoch": 7.03, "learning_rate": 2.2577464788732395e-05, "loss": 2.7823, "step": 6274 }, { "epoch": 7.03, "learning_rate": 2.2576056338028168e-05, "loss": 2.4824, "step": 6275 }, { "epoch": 7.04, "learning_rate": 2.2574647887323944e-05, "loss": 2.6002, "step": 6276 }, { "epoch": 7.04, "learning_rate": 2.257323943661972e-05, "loss": 2.291, "step": 6277 }, { "epoch": 7.04, "learning_rate": 2.2571830985915493e-05, "loss": 2.3979, "step": 6278 }, { "epoch": 7.04, "learning_rate": 2.257042253521127e-05, "loss": 2.366, "step": 6279 }, { "epoch": 7.04, "learning_rate": 2.2569014084507042e-05, "loss": 2.425, "step": 6280 }, { "epoch": 7.04, "learning_rate": 2.256760563380282e-05, "loss": 2.7472, "step": 6281 }, { "epoch": 7.04, "learning_rate": 2.2566197183098592e-05, "loss": 2.4172, "step": 6282 }, { "epoch": 7.04, "learning_rate": 2.2564788732394368e-05, "loss": 2.6691, "step": 6283 }, { "epoch": 7.04, "learning_rate": 2.256338028169014e-05, "loss": 2.559, "step": 6284 }, { "epoch": 7.05, "learning_rate": 2.2561971830985917e-05, "loss": 2.5095, "step": 6285 }, { "epoch": 7.05, "learning_rate": 2.256056338028169e-05, "loss": 2.7059, "step": 6286 }, { "epoch": 7.05, "learning_rate": 2.2559154929577463e-05, "loss": 2.459, "step": 6287 }, { "epoch": 7.05, "learning_rate": 2.255774647887324e-05, "loss": 2.4101, "step": 6288 }, { "epoch": 7.05, "learning_rate": 2.2556338028169013e-05, "loss": 2.2745, "step": 6289 }, { "epoch": 7.05, "learning_rate": 2.255492957746479e-05, "loss": 2.4138, "step": 6290 }, { "epoch": 7.05, "learning_rate": 2.2553521126760562e-05, "loss": 2.2739, "step": 6291 }, { "epoch": 7.05, "learning_rate": 2.255211267605634e-05, "loss": 2.3396, "step": 6292 }, { "epoch": 7.05, "learning_rate": 2.2550704225352114e-05, "loss": 2.4977, "step": 6293 }, { "epoch": 7.06, "learning_rate": 2.2549295774647887e-05, "loss": 2.0216, "step": 6294 }, { "epoch": 7.06, "learning_rate": 2.2547887323943664e-05, "loss": 3.9428, "step": 6295 }, { "epoch": 7.06, "learning_rate": 2.2546478873239437e-05, "loss": 3.5864, "step": 6296 }, { "epoch": 7.06, "learning_rate": 2.2545070422535213e-05, "loss": 3.5315, "step": 6297 }, { "epoch": 7.06, "learning_rate": 2.2543661971830986e-05, "loss": 3.3234, "step": 6298 }, { "epoch": 7.06, "learning_rate": 2.2542253521126762e-05, "loss": 3.0737, "step": 6299 }, { "epoch": 7.06, "learning_rate": 2.2540845070422535e-05, "loss": 3.3436, "step": 6300 }, { "epoch": 7.06, "learning_rate": 2.253943661971831e-05, "loss": 2.9594, "step": 6301 }, { "epoch": 7.07, "learning_rate": 2.2538028169014085e-05, "loss": 2.7665, "step": 6302 }, { "epoch": 7.07, "learning_rate": 2.2536619718309857e-05, "loss": 2.6284, "step": 6303 }, { "epoch": 7.07, "learning_rate": 2.2535211267605634e-05, "loss": 2.8141, "step": 6304 }, { "epoch": 7.07, "learning_rate": 2.2533802816901407e-05, "loss": 2.8031, "step": 6305 }, { "epoch": 7.07, "learning_rate": 2.2532394366197183e-05, "loss": 2.6467, "step": 6306 }, { "epoch": 7.07, "learning_rate": 2.253098591549296e-05, "loss": 2.8379, "step": 6307 }, { "epoch": 7.07, "learning_rate": 2.2529577464788736e-05, "loss": 2.553, "step": 6308 }, { "epoch": 7.07, "learning_rate": 2.252816901408451e-05, "loss": 2.5508, "step": 6309 }, { "epoch": 7.07, "learning_rate": 2.252676056338028e-05, "loss": 2.8062, "step": 6310 }, { "epoch": 7.08, "learning_rate": 2.2525352112676058e-05, "loss": 2.5646, "step": 6311 }, { "epoch": 7.08, "learning_rate": 2.252394366197183e-05, "loss": 2.5575, "step": 6312 }, { "epoch": 7.08, "learning_rate": 2.2522535211267607e-05, "loss": 2.7852, "step": 6313 }, { "epoch": 7.08, "learning_rate": 2.252112676056338e-05, "loss": 2.7565, "step": 6314 }, { "epoch": 7.08, "learning_rate": 2.2519718309859156e-05, "loss": 2.5378, "step": 6315 }, { "epoch": 7.08, "learning_rate": 2.251830985915493e-05, "loss": 2.6104, "step": 6316 }, { "epoch": 7.08, "learning_rate": 2.2516901408450706e-05, "loss": 2.5171, "step": 6317 }, { "epoch": 7.08, "learning_rate": 2.251549295774648e-05, "loss": 2.7668, "step": 6318 }, { "epoch": 7.08, "learning_rate": 2.251408450704225e-05, "loss": 2.3995, "step": 6319 }, { "epoch": 7.09, "learning_rate": 2.2512676056338028e-05, "loss": 2.5673, "step": 6320 }, { "epoch": 7.09, "learning_rate": 2.25112676056338e-05, "loss": 2.5231, "step": 6321 }, { "epoch": 7.09, "learning_rate": 2.250985915492958e-05, "loss": 2.5422, "step": 6322 }, { "epoch": 7.09, "learning_rate": 2.2508450704225354e-05, "loss": 2.6894, "step": 6323 }, { "epoch": 7.09, "learning_rate": 2.250704225352113e-05, "loss": 2.5311, "step": 6324 }, { "epoch": 7.09, "learning_rate": 2.2505633802816903e-05, "loss": 2.6435, "step": 6325 }, { "epoch": 7.09, "learning_rate": 2.2504225352112676e-05, "loss": 2.457, "step": 6326 }, { "epoch": 7.09, "learning_rate": 2.2502816901408452e-05, "loss": 2.5383, "step": 6327 }, { "epoch": 7.09, "learning_rate": 2.2501408450704225e-05, "loss": 2.3463, "step": 6328 }, { "epoch": 7.1, "learning_rate": 2.25e-05, "loss": 2.6157, "step": 6329 }, { "epoch": 7.1, "learning_rate": 2.2498591549295774e-05, "loss": 2.5149, "step": 6330 }, { "epoch": 7.1, "learning_rate": 2.249718309859155e-05, "loss": 2.7059, "step": 6331 }, { "epoch": 7.1, "learning_rate": 2.2495774647887324e-05, "loss": 2.4136, "step": 6332 }, { "epoch": 7.1, "learning_rate": 2.24943661971831e-05, "loss": 2.6137, "step": 6333 }, { "epoch": 7.1, "learning_rate": 2.2492957746478873e-05, "loss": 2.5281, "step": 6334 }, { "epoch": 7.1, "learning_rate": 2.2491549295774646e-05, "loss": 2.399, "step": 6335 }, { "epoch": 7.1, "learning_rate": 2.2490140845070422e-05, "loss": 2.5146, "step": 6336 }, { "epoch": 7.1, "learning_rate": 2.24887323943662e-05, "loss": 2.4762, "step": 6337 }, { "epoch": 7.11, "learning_rate": 2.2487323943661975e-05, "loss": 2.4102, "step": 6338 }, { "epoch": 7.11, "learning_rate": 2.2485915492957748e-05, "loss": 2.7208, "step": 6339 }, { "epoch": 7.11, "learning_rate": 2.2484507042253524e-05, "loss": 2.1473, "step": 6340 }, { "epoch": 7.11, "learning_rate": 2.2483098591549297e-05, "loss": 2.563, "step": 6341 }, { "epoch": 7.11, "learning_rate": 2.248169014084507e-05, "loss": 2.1267, "step": 6342 }, { "epoch": 7.11, "learning_rate": 2.2480281690140846e-05, "loss": 2.578, "step": 6343 }, { "epoch": 7.11, "learning_rate": 2.247887323943662e-05, "loss": 2.2673, "step": 6344 }, { "epoch": 7.11, "learning_rate": 2.2477464788732396e-05, "loss": 3.6311, "step": 6345 }, { "epoch": 7.11, "learning_rate": 2.247605633802817e-05, "loss": 3.8113, "step": 6346 }, { "epoch": 7.12, "learning_rate": 2.2474647887323945e-05, "loss": 3.5033, "step": 6347 }, { "epoch": 7.12, "learning_rate": 2.2473239436619718e-05, "loss": 3.398, "step": 6348 }, { "epoch": 7.12, "learning_rate": 2.247183098591549e-05, "loss": 2.9048, "step": 6349 }, { "epoch": 7.12, "learning_rate": 2.2470422535211267e-05, "loss": 2.8806, "step": 6350 }, { "epoch": 7.12, "learning_rate": 2.246901408450704e-05, "loss": 2.8499, "step": 6351 }, { "epoch": 7.12, "learning_rate": 2.246760563380282e-05, "loss": 2.6089, "step": 6352 }, { "epoch": 7.12, "learning_rate": 2.2466197183098593e-05, "loss": 2.6761, "step": 6353 }, { "epoch": 7.12, "learning_rate": 2.246478873239437e-05, "loss": 2.7617, "step": 6354 }, { "epoch": 7.12, "learning_rate": 2.2463380281690142e-05, "loss": 2.6458, "step": 6355 }, { "epoch": 7.13, "learning_rate": 2.246197183098592e-05, "loss": 3.0154, "step": 6356 }, { "epoch": 7.13, "learning_rate": 2.246056338028169e-05, "loss": 2.6533, "step": 6357 }, { "epoch": 7.13, "learning_rate": 2.2459154929577464e-05, "loss": 2.6597, "step": 6358 }, { "epoch": 7.13, "learning_rate": 2.245774647887324e-05, "loss": 2.8976, "step": 6359 }, { "epoch": 7.13, "learning_rate": 2.2456338028169013e-05, "loss": 2.875, "step": 6360 }, { "epoch": 7.13, "learning_rate": 2.245492957746479e-05, "loss": 2.649, "step": 6361 }, { "epoch": 7.13, "learning_rate": 2.2453521126760563e-05, "loss": 2.6123, "step": 6362 }, { "epoch": 7.13, "learning_rate": 2.245211267605634e-05, "loss": 2.4879, "step": 6363 }, { "epoch": 7.13, "learning_rate": 2.2450704225352112e-05, "loss": 2.6135, "step": 6364 }, { "epoch": 7.14, "learning_rate": 2.2449295774647885e-05, "loss": 2.5418, "step": 6365 }, { "epoch": 7.14, "learning_rate": 2.244788732394366e-05, "loss": 2.3888, "step": 6366 }, { "epoch": 7.14, "learning_rate": 2.2446478873239438e-05, "loss": 2.6559, "step": 6367 }, { "epoch": 7.14, "learning_rate": 2.2445070422535214e-05, "loss": 2.8052, "step": 6368 }, { "epoch": 7.14, "learning_rate": 2.2443661971830987e-05, "loss": 2.5399, "step": 6369 }, { "epoch": 7.14, "learning_rate": 2.2442253521126763e-05, "loss": 2.4524, "step": 6370 }, { "epoch": 7.14, "learning_rate": 2.2440845070422536e-05, "loss": 2.4099, "step": 6371 }, { "epoch": 7.14, "learning_rate": 2.2439436619718313e-05, "loss": 2.3791, "step": 6372 }, { "epoch": 7.14, "learning_rate": 2.2438028169014085e-05, "loss": 2.4493, "step": 6373 }, { "epoch": 7.15, "learning_rate": 2.243661971830986e-05, "loss": 2.5409, "step": 6374 }, { "epoch": 7.15, "learning_rate": 2.2435211267605635e-05, "loss": 2.5153, "step": 6375 }, { "epoch": 7.15, "learning_rate": 2.2433802816901408e-05, "loss": 2.1258, "step": 6376 }, { "epoch": 7.15, "learning_rate": 2.2432394366197184e-05, "loss": 2.5673, "step": 6377 }, { "epoch": 7.15, "learning_rate": 2.2430985915492957e-05, "loss": 2.6454, "step": 6378 }, { "epoch": 7.15, "learning_rate": 2.2429577464788733e-05, "loss": 2.6581, "step": 6379 }, { "epoch": 7.15, "learning_rate": 2.2428169014084506e-05, "loss": 2.5563, "step": 6380 }, { "epoch": 7.15, "learning_rate": 2.242676056338028e-05, "loss": 2.5519, "step": 6381 }, { "epoch": 7.15, "learning_rate": 2.242535211267606e-05, "loss": 2.4792, "step": 6382 }, { "epoch": 7.16, "learning_rate": 2.2423943661971832e-05, "loss": 2.4805, "step": 6383 }, { "epoch": 7.16, "learning_rate": 2.2422535211267608e-05, "loss": 2.5434, "step": 6384 }, { "epoch": 7.16, "learning_rate": 2.242112676056338e-05, "loss": 2.3256, "step": 6385 }, { "epoch": 7.16, "learning_rate": 2.2419718309859157e-05, "loss": 2.4143, "step": 6386 }, { "epoch": 7.16, "learning_rate": 2.241830985915493e-05, "loss": 2.2416, "step": 6387 }, { "epoch": 7.16, "learning_rate": 2.2416901408450707e-05, "loss": 2.2356, "step": 6388 }, { "epoch": 7.16, "learning_rate": 2.241549295774648e-05, "loss": 2.2979, "step": 6389 }, { "epoch": 7.16, "learning_rate": 2.2414084507042253e-05, "loss": 2.2718, "step": 6390 }, { "epoch": 7.16, "learning_rate": 2.241267605633803e-05, "loss": 1.987, "step": 6391 }, { "epoch": 7.17, "learning_rate": 2.2411267605633802e-05, "loss": 2.3953, "step": 6392 }, { "epoch": 7.17, "learning_rate": 2.2409859154929578e-05, "loss": 2.1043, "step": 6393 }, { "epoch": 7.17, "learning_rate": 2.240845070422535e-05, "loss": 1.874, "step": 6394 }, { "epoch": 7.17, "learning_rate": 2.2407042253521127e-05, "loss": 3.5156, "step": 6395 }, { "epoch": 7.17, "learning_rate": 2.24056338028169e-05, "loss": 3.3998, "step": 6396 }, { "epoch": 7.17, "learning_rate": 2.2404225352112673e-05, "loss": 3.478, "step": 6397 }, { "epoch": 7.17, "learning_rate": 2.2402816901408453e-05, "loss": 3.4032, "step": 6398 }, { "epoch": 7.17, "learning_rate": 2.2401408450704226e-05, "loss": 3.1007, "step": 6399 }, { "epoch": 7.17, "learning_rate": 2.2400000000000002e-05, "loss": 2.9453, "step": 6400 }, { "epoch": 7.18, "learning_rate": 2.2398591549295775e-05, "loss": 2.8728, "step": 6401 }, { "epoch": 7.18, "learning_rate": 2.239718309859155e-05, "loss": 2.8415, "step": 6402 }, { "epoch": 7.18, "learning_rate": 2.2395774647887325e-05, "loss": 2.5963, "step": 6403 }, { "epoch": 7.18, "learning_rate": 2.23943661971831e-05, "loss": 2.9727, "step": 6404 }, { "epoch": 7.18, "learning_rate": 2.2392957746478874e-05, "loss": 2.8798, "step": 6405 }, { "epoch": 7.18, "learning_rate": 2.2391549295774647e-05, "loss": 2.7892, "step": 6406 }, { "epoch": 7.18, "learning_rate": 2.2390140845070423e-05, "loss": 2.7245, "step": 6407 }, { "epoch": 7.18, "learning_rate": 2.2388732394366196e-05, "loss": 2.7705, "step": 6408 }, { "epoch": 7.18, "learning_rate": 2.2387323943661972e-05, "loss": 2.5712, "step": 6409 }, { "epoch": 7.19, "learning_rate": 2.2385915492957745e-05, "loss": 2.4972, "step": 6410 }, { "epoch": 7.19, "learning_rate": 2.238450704225352e-05, "loss": 2.5234, "step": 6411 }, { "epoch": 7.19, "learning_rate": 2.2383098591549295e-05, "loss": 2.8828, "step": 6412 }, { "epoch": 7.19, "learning_rate": 2.238169014084507e-05, "loss": 2.4272, "step": 6413 }, { "epoch": 7.19, "learning_rate": 2.2380281690140847e-05, "loss": 2.5232, "step": 6414 }, { "epoch": 7.19, "learning_rate": 2.237887323943662e-05, "loss": 2.5471, "step": 6415 }, { "epoch": 7.19, "learning_rate": 2.2377464788732397e-05, "loss": 2.4623, "step": 6416 }, { "epoch": 7.19, "learning_rate": 2.237605633802817e-05, "loss": 2.7818, "step": 6417 }, { "epoch": 7.2, "learning_rate": 2.2374647887323946e-05, "loss": 2.4753, "step": 6418 }, { "epoch": 7.2, "learning_rate": 2.237323943661972e-05, "loss": 2.4408, "step": 6419 }, { "epoch": 7.2, "learning_rate": 2.2371830985915495e-05, "loss": 2.4734, "step": 6420 }, { "epoch": 7.2, "learning_rate": 2.2370422535211268e-05, "loss": 2.6552, "step": 6421 }, { "epoch": 7.2, "learning_rate": 2.236901408450704e-05, "loss": 2.5474, "step": 6422 }, { "epoch": 7.2, "learning_rate": 2.2367605633802817e-05, "loss": 2.6698, "step": 6423 }, { "epoch": 7.2, "learning_rate": 2.236619718309859e-05, "loss": 2.545, "step": 6424 }, { "epoch": 7.2, "learning_rate": 2.2364788732394367e-05, "loss": 2.6418, "step": 6425 }, { "epoch": 7.2, "learning_rate": 2.236338028169014e-05, "loss": 2.457, "step": 6426 }, { "epoch": 7.21, "learning_rate": 2.2361971830985916e-05, "loss": 2.2431, "step": 6427 }, { "epoch": 7.21, "learning_rate": 2.2360563380281692e-05, "loss": 2.4346, "step": 6428 }, { "epoch": 7.21, "learning_rate": 2.2359154929577465e-05, "loss": 2.313, "step": 6429 }, { "epoch": 7.21, "learning_rate": 2.235774647887324e-05, "loss": 2.4625, "step": 6430 }, { "epoch": 7.21, "learning_rate": 2.2356338028169014e-05, "loss": 2.4888, "step": 6431 }, { "epoch": 7.21, "learning_rate": 2.235492957746479e-05, "loss": 2.4332, "step": 6432 }, { "epoch": 7.21, "learning_rate": 2.235492957746479e-05, "loss": 2.7022, "step": 6433 }, { "epoch": 7.21, "learning_rate": 2.2353521126760564e-05, "loss": 2.3755, "step": 6434 }, { "epoch": 7.21, "learning_rate": 2.235211267605634e-05, "loss": 2.4187, "step": 6435 }, { "epoch": 7.22, "learning_rate": 2.2350704225352113e-05, "loss": 2.3641, "step": 6436 }, { "epoch": 7.22, "learning_rate": 2.234929577464789e-05, "loss": 2.4544, "step": 6437 }, { "epoch": 7.22, "learning_rate": 2.2347887323943662e-05, "loss": 2.3226, "step": 6438 }, { "epoch": 7.22, "learning_rate": 2.2346478873239435e-05, "loss": 2.2854, "step": 6439 }, { "epoch": 7.22, "learning_rate": 2.234507042253521e-05, "loss": 2.3297, "step": 6440 }, { "epoch": 7.22, "learning_rate": 2.2343661971830984e-05, "loss": 2.1275, "step": 6441 }, { "epoch": 7.22, "learning_rate": 2.234225352112676e-05, "loss": 2.184, "step": 6442 }, { "epoch": 7.22, "learning_rate": 2.2340845070422534e-05, "loss": 2.3814, "step": 6443 }, { "epoch": 7.22, "learning_rate": 2.2339436619718313e-05, "loss": 2.1175, "step": 6444 }, { "epoch": 7.23, "learning_rate": 2.2338028169014086e-05, "loss": 3.6618, "step": 6445 }, { "epoch": 7.23, "learning_rate": 2.233661971830986e-05, "loss": 3.8079, "step": 6446 }, { "epoch": 7.23, "learning_rate": 2.2335211267605636e-05, "loss": 3.5224, "step": 6447 }, { "epoch": 7.23, "learning_rate": 2.233380281690141e-05, "loss": 3.3804, "step": 6448 }, { "epoch": 7.23, "learning_rate": 2.2332394366197185e-05, "loss": 3.197, "step": 6449 }, { "epoch": 7.23, "learning_rate": 2.2330985915492958e-05, "loss": 2.865, "step": 6450 }, { "epoch": 7.23, "learning_rate": 2.2329577464788734e-05, "loss": 2.7375, "step": 6451 }, { "epoch": 7.23, "learning_rate": 2.2328169014084507e-05, "loss": 2.6869, "step": 6452 }, { "epoch": 7.23, "learning_rate": 2.232676056338028e-05, "loss": 2.5776, "step": 6453 }, { "epoch": 7.24, "learning_rate": 2.2325352112676056e-05, "loss": 2.8183, "step": 6454 }, { "epoch": 7.24, "learning_rate": 2.232394366197183e-05, "loss": 3.0706, "step": 6455 }, { "epoch": 7.24, "learning_rate": 2.2322535211267606e-05, "loss": 2.6379, "step": 6456 }, { "epoch": 7.24, "learning_rate": 2.232112676056338e-05, "loss": 2.817, "step": 6457 }, { "epoch": 7.24, "learning_rate": 2.2319718309859155e-05, "loss": 2.5018, "step": 6458 }, { "epoch": 7.24, "learning_rate": 2.231830985915493e-05, "loss": 2.4805, "step": 6459 }, { "epoch": 7.24, "learning_rate": 2.2316901408450708e-05, "loss": 2.7095, "step": 6460 }, { "epoch": 7.24, "learning_rate": 2.231549295774648e-05, "loss": 2.638, "step": 6461 }, { "epoch": 7.24, "learning_rate": 2.2314084507042254e-05, "loss": 2.6134, "step": 6462 }, { "epoch": 7.25, "learning_rate": 2.231267605633803e-05, "loss": 2.725, "step": 6463 }, { "epoch": 7.25, "learning_rate": 2.2311267605633803e-05, "loss": 2.4532, "step": 6464 }, { "epoch": 7.25, "learning_rate": 2.230985915492958e-05, "loss": 2.44, "step": 6465 }, { "epoch": 7.25, "learning_rate": 2.2308450704225352e-05, "loss": 2.5821, "step": 6466 }, { "epoch": 7.25, "learning_rate": 2.230704225352113e-05, "loss": 2.5337, "step": 6467 }, { "epoch": 7.25, "learning_rate": 2.23056338028169e-05, "loss": 2.2532, "step": 6468 }, { "epoch": 7.25, "learning_rate": 2.2304225352112674e-05, "loss": 2.4827, "step": 6469 }, { "epoch": 7.25, "learning_rate": 2.230281690140845e-05, "loss": 2.6887, "step": 6470 }, { "epoch": 7.25, "learning_rate": 2.2301408450704224e-05, "loss": 2.4392, "step": 6471 }, { "epoch": 7.26, "learning_rate": 2.23e-05, "loss": 2.6988, "step": 6472 }, { "epoch": 7.26, "learning_rate": 2.2298591549295773e-05, "loss": 2.5642, "step": 6473 }, { "epoch": 7.26, "learning_rate": 2.2297183098591553e-05, "loss": 2.4951, "step": 6474 }, { "epoch": 7.26, "learning_rate": 2.2295774647887326e-05, "loss": 2.8216, "step": 6475 }, { "epoch": 7.26, "learning_rate": 2.2294366197183102e-05, "loss": 2.3879, "step": 6476 }, { "epoch": 7.26, "learning_rate": 2.2292957746478875e-05, "loss": 2.6652, "step": 6477 }, { "epoch": 7.26, "learning_rate": 2.2291549295774648e-05, "loss": 2.6694, "step": 6478 }, { "epoch": 7.26, "learning_rate": 2.2290140845070424e-05, "loss": 2.4172, "step": 6479 }, { "epoch": 7.26, "learning_rate": 2.2288732394366197e-05, "loss": 2.3932, "step": 6480 }, { "epoch": 7.27, "learning_rate": 2.2287323943661973e-05, "loss": 2.5574, "step": 6481 }, { "epoch": 7.27, "learning_rate": 2.2285915492957746e-05, "loss": 2.3659, "step": 6482 }, { "epoch": 7.27, "learning_rate": 2.2284507042253523e-05, "loss": 2.7066, "step": 6483 }, { "epoch": 7.27, "learning_rate": 2.2283098591549296e-05, "loss": 2.3058, "step": 6484 }, { "epoch": 7.27, "learning_rate": 2.228169014084507e-05, "loss": 2.359, "step": 6485 }, { "epoch": 7.27, "learning_rate": 2.2280281690140845e-05, "loss": 2.3823, "step": 6486 }, { "epoch": 7.27, "learning_rate": 2.2278873239436618e-05, "loss": 2.6139, "step": 6487 }, { "epoch": 7.27, "learning_rate": 2.2277464788732394e-05, "loss": 2.3893, "step": 6488 }, { "epoch": 7.27, "learning_rate": 2.227605633802817e-05, "loss": 2.4542, "step": 6489 }, { "epoch": 7.28, "learning_rate": 2.2274647887323947e-05, "loss": 2.0707, "step": 6490 }, { "epoch": 7.28, "learning_rate": 2.227323943661972e-05, "loss": 2.2654, "step": 6491 }, { "epoch": 7.28, "learning_rate": 2.2271830985915496e-05, "loss": 1.9905, "step": 6492 }, { "epoch": 7.28, "learning_rate": 2.227042253521127e-05, "loss": 2.3421, "step": 6493 }, { "epoch": 7.28, "learning_rate": 2.2269014084507042e-05, "loss": 2.0067, "step": 6494 }, { "epoch": 7.28, "learning_rate": 2.2267605633802818e-05, "loss": 3.8731, "step": 6495 }, { "epoch": 7.28, "learning_rate": 2.226619718309859e-05, "loss": 3.5257, "step": 6496 }, { "epoch": 7.28, "learning_rate": 2.2264788732394368e-05, "loss": 3.5473, "step": 6497 }, { "epoch": 7.28, "learning_rate": 2.226338028169014e-05, "loss": 3.3031, "step": 6498 }, { "epoch": 7.29, "learning_rate": 2.2261971830985917e-05, "loss": 2.903, "step": 6499 }, { "epoch": 7.29, "learning_rate": 2.226056338028169e-05, "loss": 2.8636, "step": 6500 }, { "epoch": 7.29, "learning_rate": 2.2259154929577463e-05, "loss": 2.663, "step": 6501 }, { "epoch": 7.29, "learning_rate": 2.225774647887324e-05, "loss": 2.6908, "step": 6502 }, { "epoch": 7.29, "learning_rate": 2.2256338028169012e-05, "loss": 2.7018, "step": 6503 }, { "epoch": 7.29, "learning_rate": 2.2254929577464792e-05, "loss": 2.6841, "step": 6504 }, { "epoch": 7.29, "learning_rate": 2.2253521126760565e-05, "loss": 2.4506, "step": 6505 }, { "epoch": 7.29, "learning_rate": 2.225211267605634e-05, "loss": 2.7078, "step": 6506 }, { "epoch": 7.29, "learning_rate": 2.2250704225352114e-05, "loss": 2.7858, "step": 6507 }, { "epoch": 7.3, "learning_rate": 2.224929577464789e-05, "loss": 2.7935, "step": 6508 }, { "epoch": 7.3, "learning_rate": 2.2247887323943663e-05, "loss": 2.7744, "step": 6509 }, { "epoch": 7.3, "learning_rate": 2.2246478873239436e-05, "loss": 2.6345, "step": 6510 }, { "epoch": 7.3, "learning_rate": 2.2245070422535213e-05, "loss": 2.5051, "step": 6511 }, { "epoch": 7.3, "learning_rate": 2.2243661971830985e-05, "loss": 2.4348, "step": 6512 }, { "epoch": 7.3, "learning_rate": 2.2242253521126762e-05, "loss": 2.5623, "step": 6513 }, { "epoch": 7.3, "learning_rate": 2.2240845070422535e-05, "loss": 2.5752, "step": 6514 }, { "epoch": 7.3, "learning_rate": 2.223943661971831e-05, "loss": 2.316, "step": 6515 }, { "epoch": 7.3, "learning_rate": 2.2238028169014084e-05, "loss": 2.5219, "step": 6516 }, { "epoch": 7.31, "learning_rate": 2.2236619718309857e-05, "loss": 2.4746, "step": 6517 }, { "epoch": 7.31, "learning_rate": 2.2235211267605633e-05, "loss": 2.5326, "step": 6518 }, { "epoch": 7.31, "learning_rate": 2.2233802816901406e-05, "loss": 2.2311, "step": 6519 }, { "epoch": 7.31, "learning_rate": 2.2232394366197186e-05, "loss": 2.526, "step": 6520 }, { "epoch": 7.31, "learning_rate": 2.223098591549296e-05, "loss": 2.6992, "step": 6521 }, { "epoch": 7.31, "learning_rate": 2.2229577464788735e-05, "loss": 2.4358, "step": 6522 }, { "epoch": 7.31, "learning_rate": 2.2228169014084508e-05, "loss": 2.6674, "step": 6523 }, { "epoch": 7.31, "learning_rate": 2.2226760563380284e-05, "loss": 2.6917, "step": 6524 }, { "epoch": 7.32, "learning_rate": 2.2225352112676057e-05, "loss": 2.5671, "step": 6525 }, { "epoch": 7.32, "learning_rate": 2.222394366197183e-05, "loss": 2.1891, "step": 6526 }, { "epoch": 7.32, "learning_rate": 2.2222535211267607e-05, "loss": 2.2184, "step": 6527 }, { "epoch": 7.32, "learning_rate": 2.222112676056338e-05, "loss": 2.7432, "step": 6528 }, { "epoch": 7.32, "learning_rate": 2.2219718309859156e-05, "loss": 2.226, "step": 6529 }, { "epoch": 7.32, "learning_rate": 2.221830985915493e-05, "loss": 2.3262, "step": 6530 }, { "epoch": 7.32, "learning_rate": 2.2216901408450705e-05, "loss": 2.4303, "step": 6531 }, { "epoch": 7.32, "learning_rate": 2.2215492957746478e-05, "loss": 2.3869, "step": 6532 }, { "epoch": 7.32, "learning_rate": 2.221408450704225e-05, "loss": 2.3891, "step": 6533 }, { "epoch": 7.33, "learning_rate": 2.2212676056338027e-05, "loss": 2.2446, "step": 6534 }, { "epoch": 7.33, "learning_rate": 2.2211267605633804e-05, "loss": 2.266, "step": 6535 }, { "epoch": 7.33, "learning_rate": 2.220985915492958e-05, "loss": 2.4821, "step": 6536 }, { "epoch": 7.33, "learning_rate": 2.2208450704225353e-05, "loss": 2.3381, "step": 6537 }, { "epoch": 7.33, "learning_rate": 2.220704225352113e-05, "loss": 2.3928, "step": 6538 }, { "epoch": 7.33, "learning_rate": 2.2205633802816902e-05, "loss": 2.0414, "step": 6539 }, { "epoch": 7.33, "learning_rate": 2.220422535211268e-05, "loss": 1.9801, "step": 6540 }, { "epoch": 7.33, "learning_rate": 2.220281690140845e-05, "loss": 2.0712, "step": 6541 }, { "epoch": 7.33, "learning_rate": 2.2201408450704225e-05, "loss": 1.9899, "step": 6542 }, { "epoch": 7.34, "learning_rate": 2.22e-05, "loss": 2.1696, "step": 6543 }, { "epoch": 7.34, "learning_rate": 2.2198591549295774e-05, "loss": 1.7556, "step": 6544 }, { "epoch": 7.34, "learning_rate": 2.219718309859155e-05, "loss": 3.4504, "step": 6545 }, { "epoch": 7.34, "learning_rate": 2.2195774647887323e-05, "loss": 3.5903, "step": 6546 }, { "epoch": 7.34, "learning_rate": 2.21943661971831e-05, "loss": 3.1817, "step": 6547 }, { "epoch": 7.34, "learning_rate": 2.2192957746478872e-05, "loss": 3.0067, "step": 6548 }, { "epoch": 7.34, "learning_rate": 2.2191549295774645e-05, "loss": 2.957, "step": 6549 }, { "epoch": 7.34, "learning_rate": 2.2190140845070425e-05, "loss": 2.8529, "step": 6550 }, { "epoch": 7.34, "learning_rate": 2.2188732394366198e-05, "loss": 2.6649, "step": 6551 }, { "epoch": 7.35, "learning_rate": 2.2187323943661974e-05, "loss": 2.4913, "step": 6552 }, { "epoch": 7.35, "learning_rate": 2.2185915492957747e-05, "loss": 2.6083, "step": 6553 }, { "epoch": 7.35, "learning_rate": 2.2184507042253524e-05, "loss": 2.712, "step": 6554 }, { "epoch": 7.35, "learning_rate": 2.2183098591549297e-05, "loss": 2.4488, "step": 6555 }, { "epoch": 7.35, "learning_rate": 2.218169014084507e-05, "loss": 2.5939, "step": 6556 }, { "epoch": 7.35, "learning_rate": 2.2180281690140846e-05, "loss": 2.843, "step": 6557 }, { "epoch": 7.35, "learning_rate": 2.217887323943662e-05, "loss": 2.4162, "step": 6558 }, { "epoch": 7.35, "learning_rate": 2.2177464788732395e-05, "loss": 2.4134, "step": 6559 }, { "epoch": 7.35, "learning_rate": 2.2176056338028168e-05, "loss": 2.6108, "step": 6560 }, { "epoch": 7.36, "learning_rate": 2.2174647887323944e-05, "loss": 2.6161, "step": 6561 }, { "epoch": 7.36, "learning_rate": 2.2173239436619717e-05, "loss": 2.4993, "step": 6562 }, { "epoch": 7.36, "learning_rate": 2.2171830985915494e-05, "loss": 2.5693, "step": 6563 }, { "epoch": 7.36, "learning_rate": 2.2170422535211267e-05, "loss": 2.3896, "step": 6564 }, { "epoch": 7.36, "learning_rate": 2.2169014084507043e-05, "loss": 2.2668, "step": 6565 }, { "epoch": 7.36, "learning_rate": 2.216760563380282e-05, "loss": 2.4726, "step": 6566 }, { "epoch": 7.36, "learning_rate": 2.2166197183098592e-05, "loss": 2.3958, "step": 6567 }, { "epoch": 7.36, "learning_rate": 2.216478873239437e-05, "loss": 2.4003, "step": 6568 }, { "epoch": 7.36, "learning_rate": 2.216338028169014e-05, "loss": 2.5491, "step": 6569 }, { "epoch": 7.37, "learning_rate": 2.2161971830985918e-05, "loss": 2.3957, "step": 6570 }, { "epoch": 7.37, "learning_rate": 2.216056338028169e-05, "loss": 2.4551, "step": 6571 }, { "epoch": 7.37, "learning_rate": 2.2159154929577464e-05, "loss": 2.3808, "step": 6572 }, { "epoch": 7.37, "learning_rate": 2.215774647887324e-05, "loss": 2.4016, "step": 6573 }, { "epoch": 7.37, "learning_rate": 2.2156338028169013e-05, "loss": 2.2296, "step": 6574 }, { "epoch": 7.37, "learning_rate": 2.215492957746479e-05, "loss": 2.7256, "step": 6575 }, { "epoch": 7.37, "learning_rate": 2.2153521126760562e-05, "loss": 2.6035, "step": 6576 }, { "epoch": 7.37, "learning_rate": 2.215211267605634e-05, "loss": 2.5093, "step": 6577 }, { "epoch": 7.37, "learning_rate": 2.215070422535211e-05, "loss": 2.4067, "step": 6578 }, { "epoch": 7.38, "learning_rate": 2.2149295774647888e-05, "loss": 2.2437, "step": 6579 }, { "epoch": 7.38, "learning_rate": 2.2147887323943664e-05, "loss": 2.4331, "step": 6580 }, { "epoch": 7.38, "learning_rate": 2.2146478873239437e-05, "loss": 2.2022, "step": 6581 }, { "epoch": 7.38, "learning_rate": 2.2145070422535213e-05, "loss": 2.3885, "step": 6582 }, { "epoch": 7.38, "learning_rate": 2.2143661971830986e-05, "loss": 2.3008, "step": 6583 }, { "epoch": 7.38, "learning_rate": 2.2142253521126763e-05, "loss": 2.3283, "step": 6584 }, { "epoch": 7.38, "learning_rate": 2.2140845070422536e-05, "loss": 2.7676, "step": 6585 }, { "epoch": 7.38, "learning_rate": 2.2139436619718312e-05, "loss": 2.009, "step": 6586 }, { "epoch": 7.38, "learning_rate": 2.2138028169014085e-05, "loss": 2.1738, "step": 6587 }, { "epoch": 7.39, "learning_rate": 2.2136619718309858e-05, "loss": 2.1945, "step": 6588 }, { "epoch": 7.39, "learning_rate": 2.2135211267605634e-05, "loss": 2.019, "step": 6589 }, { "epoch": 7.39, "learning_rate": 2.2133802816901407e-05, "loss": 1.9523, "step": 6590 }, { "epoch": 7.39, "learning_rate": 2.2132394366197184e-05, "loss": 2.0034, "step": 6591 }, { "epoch": 7.39, "learning_rate": 2.2130985915492956e-05, "loss": 1.8698, "step": 6592 }, { "epoch": 7.39, "learning_rate": 2.2129577464788733e-05, "loss": 2.2971, "step": 6593 }, { "epoch": 7.39, "learning_rate": 2.2128169014084506e-05, "loss": 1.7398, "step": 6594 }, { "epoch": 7.39, "learning_rate": 2.2126760563380285e-05, "loss": 3.4157, "step": 6595 }, { "epoch": 7.39, "learning_rate": 2.212535211267606e-05, "loss": 3.2292, "step": 6596 }, { "epoch": 7.4, "learning_rate": 2.212394366197183e-05, "loss": 3.0813, "step": 6597 }, { "epoch": 7.4, "learning_rate": 2.2122535211267608e-05, "loss": 3.0457, "step": 6598 }, { "epoch": 7.4, "learning_rate": 2.212112676056338e-05, "loss": 2.8997, "step": 6599 }, { "epoch": 7.4, "learning_rate": 2.2119718309859157e-05, "loss": 2.447, "step": 6600 }, { "epoch": 7.4, "learning_rate": 2.211830985915493e-05, "loss": 2.6229, "step": 6601 }, { "epoch": 7.4, "learning_rate": 2.2116901408450706e-05, "loss": 2.7213, "step": 6602 }, { "epoch": 7.4, "learning_rate": 2.211549295774648e-05, "loss": 2.4019, "step": 6603 }, { "epoch": 7.4, "learning_rate": 2.2114084507042252e-05, "loss": 2.3375, "step": 6604 }, { "epoch": 7.4, "learning_rate": 2.211267605633803e-05, "loss": 2.6179, "step": 6605 }, { "epoch": 7.41, "learning_rate": 2.21112676056338e-05, "loss": 2.6565, "step": 6606 }, { "epoch": 7.41, "learning_rate": 2.2109859154929578e-05, "loss": 2.4646, "step": 6607 }, { "epoch": 7.41, "learning_rate": 2.210845070422535e-05, "loss": 2.4463, "step": 6608 }, { "epoch": 7.41, "learning_rate": 2.2107042253521127e-05, "loss": 2.7411, "step": 6609 }, { "epoch": 7.41, "learning_rate": 2.2105633802816903e-05, "loss": 2.4522, "step": 6610 }, { "epoch": 7.41, "learning_rate": 2.210422535211268e-05, "loss": 2.2004, "step": 6611 }, { "epoch": 7.41, "learning_rate": 2.2102816901408453e-05, "loss": 2.4466, "step": 6612 }, { "epoch": 7.41, "learning_rate": 2.2101408450704226e-05, "loss": 2.5273, "step": 6613 }, { "epoch": 7.41, "learning_rate": 2.2100000000000002e-05, "loss": 2.5017, "step": 6614 }, { "epoch": 7.42, "learning_rate": 2.2098591549295775e-05, "loss": 2.4449, "step": 6615 }, { "epoch": 7.42, "learning_rate": 2.209718309859155e-05, "loss": 2.412, "step": 6616 }, { "epoch": 7.42, "learning_rate": 2.2095774647887324e-05, "loss": 2.4191, "step": 6617 }, { "epoch": 7.42, "learning_rate": 2.20943661971831e-05, "loss": 2.1923, "step": 6618 }, { "epoch": 7.42, "learning_rate": 2.2092957746478873e-05, "loss": 2.3451, "step": 6619 }, { "epoch": 7.42, "learning_rate": 2.2091549295774646e-05, "loss": 2.4596, "step": 6620 }, { "epoch": 7.42, "learning_rate": 2.2090140845070423e-05, "loss": 2.2366, "step": 6621 }, { "epoch": 7.42, "learning_rate": 2.2088732394366196e-05, "loss": 2.4445, "step": 6622 }, { "epoch": 7.42, "learning_rate": 2.2087323943661972e-05, "loss": 2.1361, "step": 6623 }, { "epoch": 7.43, "learning_rate": 2.2085915492957745e-05, "loss": 2.6028, "step": 6624 }, { "epoch": 7.43, "learning_rate": 2.2084507042253525e-05, "loss": 2.4036, "step": 6625 }, { "epoch": 7.43, "learning_rate": 2.2083098591549298e-05, "loss": 2.3098, "step": 6626 }, { "epoch": 7.43, "learning_rate": 2.2081690140845074e-05, "loss": 2.2048, "step": 6627 }, { "epoch": 7.43, "learning_rate": 2.2080281690140847e-05, "loss": 2.2581, "step": 6628 }, { "epoch": 7.43, "learning_rate": 2.207887323943662e-05, "loss": 2.381, "step": 6629 }, { "epoch": 7.43, "learning_rate": 2.2077464788732396e-05, "loss": 2.3342, "step": 6630 }, { "epoch": 7.43, "learning_rate": 2.207605633802817e-05, "loss": 2.3635, "step": 6631 }, { "epoch": 7.43, "learning_rate": 2.2074647887323945e-05, "loss": 2.2607, "step": 6632 }, { "epoch": 7.44, "learning_rate": 2.2073239436619718e-05, "loss": 2.2357, "step": 6633 }, { "epoch": 7.44, "learning_rate": 2.2071830985915495e-05, "loss": 2.4575, "step": 6634 }, { "epoch": 7.44, "learning_rate": 2.2070422535211268e-05, "loss": 2.202, "step": 6635 }, { "epoch": 7.44, "learning_rate": 2.206901408450704e-05, "loss": 2.342, "step": 6636 }, { "epoch": 7.44, "learning_rate": 2.2067605633802817e-05, "loss": 2.1513, "step": 6637 }, { "epoch": 7.44, "learning_rate": 2.206619718309859e-05, "loss": 2.1276, "step": 6638 }, { "epoch": 7.44, "learning_rate": 2.2064788732394366e-05, "loss": 1.9158, "step": 6639 }, { "epoch": 7.44, "learning_rate": 2.206338028169014e-05, "loss": 2.1383, "step": 6640 }, { "epoch": 7.45, "learning_rate": 2.206197183098592e-05, "loss": 2.0073, "step": 6641 }, { "epoch": 7.45, "learning_rate": 2.2060563380281692e-05, "loss": 2.2035, "step": 6642 }, { "epoch": 7.45, "learning_rate": 2.2059154929577468e-05, "loss": 1.8715, "step": 6643 }, { "epoch": 7.45, "learning_rate": 2.205774647887324e-05, "loss": 2.0049, "step": 6644 }, { "epoch": 7.45, "learning_rate": 2.2056338028169014e-05, "loss": 3.6486, "step": 6645 }, { "epoch": 7.45, "learning_rate": 2.205492957746479e-05, "loss": 3.362, "step": 6646 }, { "epoch": 7.45, "learning_rate": 2.2053521126760563e-05, "loss": 3.1666, "step": 6647 }, { "epoch": 7.45, "learning_rate": 2.205211267605634e-05, "loss": 2.818, "step": 6648 }, { "epoch": 7.45, "learning_rate": 2.2050704225352112e-05, "loss": 2.7656, "step": 6649 }, { "epoch": 7.46, "learning_rate": 2.204929577464789e-05, "loss": 2.7104, "step": 6650 }, { "epoch": 7.46, "learning_rate": 2.2047887323943662e-05, "loss": 2.4363, "step": 6651 }, { "epoch": 7.46, "learning_rate": 2.2046478873239435e-05, "loss": 2.4418, "step": 6652 }, { "epoch": 7.46, "learning_rate": 2.204507042253521e-05, "loss": 2.6648, "step": 6653 }, { "epoch": 7.46, "learning_rate": 2.2043661971830984e-05, "loss": 2.5806, "step": 6654 }, { "epoch": 7.46, "learning_rate": 2.204225352112676e-05, "loss": 2.5432, "step": 6655 }, { "epoch": 7.46, "learning_rate": 2.2040845070422537e-05, "loss": 2.6207, "step": 6656 }, { "epoch": 7.46, "learning_rate": 2.2039436619718313e-05, "loss": 2.4649, "step": 6657 }, { "epoch": 7.46, "learning_rate": 2.2038028169014086e-05, "loss": 2.6535, "step": 6658 }, { "epoch": 7.47, "learning_rate": 2.2036619718309862e-05, "loss": 2.4929, "step": 6659 }, { "epoch": 7.47, "learning_rate": 2.2035211267605635e-05, "loss": 2.1877, "step": 6660 }, { "epoch": 7.47, "learning_rate": 2.2033802816901408e-05, "loss": 2.5525, "step": 6661 }, { "epoch": 7.47, "learning_rate": 2.2032394366197184e-05, "loss": 2.5251, "step": 6662 }, { "epoch": 7.47, "learning_rate": 2.2030985915492957e-05, "loss": 2.3539, "step": 6663 }, { "epoch": 7.47, "learning_rate": 2.2029577464788734e-05, "loss": 2.631, "step": 6664 }, { "epoch": 7.47, "learning_rate": 2.2028169014084507e-05, "loss": 2.4144, "step": 6665 }, { "epoch": 7.47, "learning_rate": 2.2026760563380283e-05, "loss": 2.57, "step": 6666 }, { "epoch": 7.47, "learning_rate": 2.2025352112676056e-05, "loss": 2.4756, "step": 6667 }, { "epoch": 7.48, "learning_rate": 2.202394366197183e-05, "loss": 2.4046, "step": 6668 }, { "epoch": 7.48, "learning_rate": 2.2022535211267605e-05, "loss": 2.2831, "step": 6669 }, { "epoch": 7.48, "learning_rate": 2.2021126760563378e-05, "loss": 2.4098, "step": 6670 }, { "epoch": 7.48, "learning_rate": 2.2019718309859158e-05, "loss": 2.3188, "step": 6671 }, { "epoch": 7.48, "learning_rate": 2.201830985915493e-05, "loss": 2.4559, "step": 6672 }, { "epoch": 7.48, "learning_rate": 2.2016901408450707e-05, "loss": 2.4549, "step": 6673 }, { "epoch": 7.48, "learning_rate": 2.201549295774648e-05, "loss": 2.5455, "step": 6674 }, { "epoch": 7.48, "learning_rate": 2.2014084507042253e-05, "loss": 2.132, "step": 6675 }, { "epoch": 7.48, "learning_rate": 2.201267605633803e-05, "loss": 2.1771, "step": 6676 }, { "epoch": 7.49, "learning_rate": 2.2011267605633802e-05, "loss": 2.3893, "step": 6677 }, { "epoch": 7.49, "learning_rate": 2.200985915492958e-05, "loss": 2.1893, "step": 6678 }, { "epoch": 7.49, "learning_rate": 2.200845070422535e-05, "loss": 2.4938, "step": 6679 }, { "epoch": 7.49, "learning_rate": 2.2007042253521128e-05, "loss": 2.1746, "step": 6680 }, { "epoch": 7.49, "learning_rate": 2.20056338028169e-05, "loss": 2.2534, "step": 6681 }, { "epoch": 7.49, "learning_rate": 2.2004225352112677e-05, "loss": 2.1009, "step": 6682 }, { "epoch": 7.49, "learning_rate": 2.200281690140845e-05, "loss": 2.3206, "step": 6683 }, { "epoch": 7.49, "learning_rate": 2.2001408450704223e-05, "loss": 2.1933, "step": 6684 }, { "epoch": 7.49, "learning_rate": 2.2e-05, "loss": 2.3928, "step": 6685 }, { "epoch": 7.5, "learning_rate": 2.1998591549295776e-05, "loss": 2.4374, "step": 6686 }, { "epoch": 7.5, "learning_rate": 2.1997183098591552e-05, "loss": 2.2061, "step": 6687 }, { "epoch": 7.5, "learning_rate": 2.1995774647887325e-05, "loss": 2.4088, "step": 6688 }, { "epoch": 7.5, "learning_rate": 2.19943661971831e-05, "loss": 2.2953, "step": 6689 }, { "epoch": 7.5, "learning_rate": 2.1992957746478874e-05, "loss": 2.0377, "step": 6690 }, { "epoch": 7.5, "learning_rate": 2.1991549295774647e-05, "loss": 2.0913, "step": 6691 }, { "epoch": 7.5, "learning_rate": 2.1990140845070424e-05, "loss": 2.3531, "step": 6692 }, { "epoch": 7.5, "learning_rate": 2.1988732394366197e-05, "loss": 2.0464, "step": 6693 }, { "epoch": 7.5, "learning_rate": 2.1987323943661973e-05, "loss": 2.0296, "step": 6694 }, { "epoch": 7.51, "learning_rate": 2.1985915492957746e-05, "loss": 3.5119, "step": 6695 }, { "epoch": 7.51, "learning_rate": 2.1984507042253522e-05, "loss": 3.3576, "step": 6696 }, { "epoch": 7.51, "learning_rate": 2.1983098591549295e-05, "loss": 3.1048, "step": 6697 }, { "epoch": 7.51, "learning_rate": 2.198169014084507e-05, "loss": 2.9161, "step": 6698 }, { "epoch": 7.51, "learning_rate": 2.1980281690140844e-05, "loss": 3.0151, "step": 6699 }, { "epoch": 7.51, "learning_rate": 2.1978873239436617e-05, "loss": 2.5209, "step": 6700 }, { "epoch": 7.51, "learning_rate": 2.1977464788732397e-05, "loss": 2.5629, "step": 6701 }, { "epoch": 7.51, "learning_rate": 2.197605633802817e-05, "loss": 2.4357, "step": 6702 }, { "epoch": 7.51, "learning_rate": 2.1974647887323946e-05, "loss": 2.6423, "step": 6703 }, { "epoch": 7.52, "learning_rate": 2.197323943661972e-05, "loss": 2.6282, "step": 6704 }, { "epoch": 7.52, "learning_rate": 2.1971830985915496e-05, "loss": 2.3843, "step": 6705 }, { "epoch": 7.52, "learning_rate": 2.197042253521127e-05, "loss": 2.8061, "step": 6706 }, { "epoch": 7.52, "learning_rate": 2.196901408450704e-05, "loss": 2.5323, "step": 6707 }, { "epoch": 7.52, "learning_rate": 2.1967605633802818e-05, "loss": 2.5436, "step": 6708 }, { "epoch": 7.52, "learning_rate": 2.196619718309859e-05, "loss": 2.2751, "step": 6709 }, { "epoch": 7.52, "learning_rate": 2.1964788732394367e-05, "loss": 2.324, "step": 6710 }, { "epoch": 7.52, "learning_rate": 2.196338028169014e-05, "loss": 2.3551, "step": 6711 }, { "epoch": 7.52, "learning_rate": 2.1961971830985916e-05, "loss": 2.2575, "step": 6712 }, { "epoch": 7.53, "learning_rate": 2.196056338028169e-05, "loss": 2.3215, "step": 6713 }, { "epoch": 7.53, "learning_rate": 2.1959154929577466e-05, "loss": 2.6229, "step": 6714 }, { "epoch": 7.53, "learning_rate": 2.195774647887324e-05, "loss": 2.501, "step": 6715 }, { "epoch": 7.53, "learning_rate": 2.1956338028169015e-05, "loss": 2.2271, "step": 6716 }, { "epoch": 7.53, "learning_rate": 2.195492957746479e-05, "loss": 2.3195, "step": 6717 }, { "epoch": 7.53, "learning_rate": 2.1953521126760564e-05, "loss": 2.1428, "step": 6718 }, { "epoch": 7.53, "learning_rate": 2.195211267605634e-05, "loss": 2.3571, "step": 6719 }, { "epoch": 7.53, "learning_rate": 2.1950704225352113e-05, "loss": 2.2709, "step": 6720 }, { "epoch": 7.53, "learning_rate": 2.194929577464789e-05, "loss": 2.3963, "step": 6721 }, { "epoch": 7.54, "learning_rate": 2.1947887323943663e-05, "loss": 2.1819, "step": 6722 }, { "epoch": 7.54, "learning_rate": 2.1946478873239436e-05, "loss": 2.2104, "step": 6723 }, { "epoch": 7.54, "learning_rate": 2.1945070422535212e-05, "loss": 2.1375, "step": 6724 }, { "epoch": 7.54, "learning_rate": 2.1943661971830985e-05, "loss": 2.1206, "step": 6725 }, { "epoch": 7.54, "learning_rate": 2.194225352112676e-05, "loss": 2.4038, "step": 6726 }, { "epoch": 7.54, "learning_rate": 2.1940845070422534e-05, "loss": 2.3317, "step": 6727 }, { "epoch": 7.54, "learning_rate": 2.193943661971831e-05, "loss": 2.2686, "step": 6728 }, { "epoch": 7.54, "learning_rate": 2.1938028169014083e-05, "loss": 2.3219, "step": 6729 }, { "epoch": 7.54, "learning_rate": 2.193661971830986e-05, "loss": 2.4732, "step": 6730 }, { "epoch": 7.55, "learning_rate": 2.1935211267605636e-05, "loss": 2.2731, "step": 6731 }, { "epoch": 7.55, "learning_rate": 2.193380281690141e-05, "loss": 2.3415, "step": 6732 }, { "epoch": 7.55, "learning_rate": 2.1932394366197185e-05, "loss": 2.225, "step": 6733 }, { "epoch": 7.55, "learning_rate": 2.193098591549296e-05, "loss": 2.2436, "step": 6734 }, { "epoch": 7.55, "learning_rate": 2.1929577464788735e-05, "loss": 2.1109, "step": 6735 }, { "epoch": 7.55, "learning_rate": 2.1928169014084508e-05, "loss": 2.2296, "step": 6736 }, { "epoch": 7.55, "learning_rate": 2.1926760563380284e-05, "loss": 2.0396, "step": 6737 }, { "epoch": 7.55, "learning_rate": 2.1925352112676057e-05, "loss": 2.1338, "step": 6738 }, { "epoch": 7.55, "learning_rate": 2.192394366197183e-05, "loss": 2.0474, "step": 6739 }, { "epoch": 7.56, "learning_rate": 2.1922535211267606e-05, "loss": 2.1735, "step": 6740 }, { "epoch": 7.56, "learning_rate": 2.192112676056338e-05, "loss": 2.0979, "step": 6741 }, { "epoch": 7.56, "learning_rate": 2.1919718309859155e-05, "loss": 2.1837, "step": 6742 }, { "epoch": 7.56, "learning_rate": 2.191830985915493e-05, "loss": 2.071, "step": 6743 }, { "epoch": 7.56, "learning_rate": 2.1916901408450705e-05, "loss": 1.9346, "step": 6744 }, { "epoch": 7.56, "learning_rate": 2.1915492957746478e-05, "loss": 3.3807, "step": 6745 }, { "epoch": 7.56, "learning_rate": 2.1914084507042254e-05, "loss": 3.4227, "step": 6746 }, { "epoch": 7.56, "learning_rate": 2.191267605633803e-05, "loss": 3.1595, "step": 6747 }, { "epoch": 7.57, "learning_rate": 2.1911267605633803e-05, "loss": 2.8326, "step": 6748 }, { "epoch": 7.57, "learning_rate": 2.190985915492958e-05, "loss": 2.6185, "step": 6749 }, { "epoch": 7.57, "learning_rate": 2.1908450704225353e-05, "loss": 2.6545, "step": 6750 }, { "epoch": 7.57, "learning_rate": 2.190704225352113e-05, "loss": 2.411, "step": 6751 }, { "epoch": 7.57, "learning_rate": 2.1905633802816902e-05, "loss": 2.4374, "step": 6752 }, { "epoch": 7.57, "learning_rate": 2.1904225352112678e-05, "loss": 2.3963, "step": 6753 }, { "epoch": 7.57, "learning_rate": 2.190281690140845e-05, "loss": 2.6266, "step": 6754 }, { "epoch": 7.57, "learning_rate": 2.1901408450704224e-05, "loss": 2.2344, "step": 6755 }, { "epoch": 7.57, "learning_rate": 2.19e-05, "loss": 2.7412, "step": 6756 }, { "epoch": 7.58, "learning_rate": 2.1898591549295773e-05, "loss": 2.5012, "step": 6757 }, { "epoch": 7.58, "learning_rate": 2.189718309859155e-05, "loss": 2.4245, "step": 6758 }, { "epoch": 7.58, "learning_rate": 2.1895774647887323e-05, "loss": 2.2695, "step": 6759 }, { "epoch": 7.58, "learning_rate": 2.18943661971831e-05, "loss": 2.3753, "step": 6760 }, { "epoch": 7.58, "learning_rate": 2.1892957746478872e-05, "loss": 2.3853, "step": 6761 }, { "epoch": 7.58, "learning_rate": 2.189154929577465e-05, "loss": 2.1499, "step": 6762 }, { "epoch": 7.58, "learning_rate": 2.1890140845070425e-05, "loss": 2.2581, "step": 6763 }, { "epoch": 7.58, "learning_rate": 2.1888732394366197e-05, "loss": 2.3845, "step": 6764 }, { "epoch": 7.58, "learning_rate": 2.1887323943661974e-05, "loss": 2.4232, "step": 6765 }, { "epoch": 7.59, "learning_rate": 2.1885915492957747e-05, "loss": 2.1362, "step": 6766 }, { "epoch": 7.59, "learning_rate": 2.1884507042253523e-05, "loss": 2.3477, "step": 6767 }, { "epoch": 7.59, "learning_rate": 2.1883098591549296e-05, "loss": 2.3782, "step": 6768 }, { "epoch": 7.59, "learning_rate": 2.1881690140845072e-05, "loss": 2.3303, "step": 6769 }, { "epoch": 7.59, "learning_rate": 2.1880281690140845e-05, "loss": 2.1817, "step": 6770 }, { "epoch": 7.59, "learning_rate": 2.1878873239436618e-05, "loss": 2.2886, "step": 6771 }, { "epoch": 7.59, "learning_rate": 2.1877464788732395e-05, "loss": 2.3469, "step": 6772 }, { "epoch": 7.59, "learning_rate": 2.1876056338028168e-05, "loss": 2.1861, "step": 6773 }, { "epoch": 7.59, "learning_rate": 2.1874647887323944e-05, "loss": 2.1759, "step": 6774 }, { "epoch": 7.6, "learning_rate": 2.1873239436619717e-05, "loss": 2.1176, "step": 6775 }, { "epoch": 7.6, "learning_rate": 2.1871830985915493e-05, "loss": 2.3328, "step": 6776 }, { "epoch": 7.6, "learning_rate": 2.187042253521127e-05, "loss": 2.2979, "step": 6777 }, { "epoch": 7.6, "learning_rate": 2.1869014084507042e-05, "loss": 2.2479, "step": 6778 }, { "epoch": 7.6, "learning_rate": 2.186760563380282e-05, "loss": 2.2261, "step": 6779 }, { "epoch": 7.6, "learning_rate": 2.186619718309859e-05, "loss": 2.3645, "step": 6780 }, { "epoch": 7.6, "learning_rate": 2.1864788732394368e-05, "loss": 2.2462, "step": 6781 }, { "epoch": 7.6, "learning_rate": 2.186338028169014e-05, "loss": 2.0285, "step": 6782 }, { "epoch": 7.6, "learning_rate": 2.1861971830985917e-05, "loss": 1.9688, "step": 6783 }, { "epoch": 7.61, "learning_rate": 2.186056338028169e-05, "loss": 2.5209, "step": 6784 }, { "epoch": 7.61, "learning_rate": 2.1859154929577467e-05, "loss": 2.1019, "step": 6785 }, { "epoch": 7.61, "learning_rate": 2.185774647887324e-05, "loss": 2.2823, "step": 6786 }, { "epoch": 7.61, "learning_rate": 2.1856338028169012e-05, "loss": 1.9522, "step": 6787 }, { "epoch": 7.61, "learning_rate": 2.185492957746479e-05, "loss": 2.1692, "step": 6788 }, { "epoch": 7.61, "learning_rate": 2.1853521126760562e-05, "loss": 1.9157, "step": 6789 }, { "epoch": 7.61, "learning_rate": 2.1852112676056338e-05, "loss": 2.4127, "step": 6790 }, { "epoch": 7.61, "learning_rate": 2.185070422535211e-05, "loss": 2.0949, "step": 6791 }, { "epoch": 7.61, "learning_rate": 2.184929577464789e-05, "loss": 2.0793, "step": 6792 }, { "epoch": 7.62, "learning_rate": 2.1847887323943664e-05, "loss": 2.0536, "step": 6793 }, { "epoch": 7.62, "learning_rate": 2.1846478873239437e-05, "loss": 1.8285, "step": 6794 }, { "epoch": 7.62, "learning_rate": 2.1845070422535213e-05, "loss": 3.2425, "step": 6795 }, { "epoch": 7.62, "learning_rate": 2.1843661971830986e-05, "loss": 3.0386, "step": 6796 }, { "epoch": 7.62, "learning_rate": 2.1842253521126762e-05, "loss": 3.1489, "step": 6797 }, { "epoch": 7.62, "learning_rate": 2.1840845070422535e-05, "loss": 2.6618, "step": 6798 }, { "epoch": 7.62, "learning_rate": 2.183943661971831e-05, "loss": 2.5762, "step": 6799 }, { "epoch": 7.62, "learning_rate": 2.1838028169014084e-05, "loss": 2.4982, "step": 6800 }, { "epoch": 7.62, "learning_rate": 2.183661971830986e-05, "loss": 2.469, "step": 6801 }, { "epoch": 7.63, "learning_rate": 2.1835211267605634e-05, "loss": 2.5218, "step": 6802 }, { "epoch": 7.63, "learning_rate": 2.1833802816901407e-05, "loss": 2.3995, "step": 6803 }, { "epoch": 7.63, "learning_rate": 2.1832394366197183e-05, "loss": 2.5534, "step": 6804 }, { "epoch": 7.63, "learning_rate": 2.1830985915492956e-05, "loss": 2.4597, "step": 6805 }, { "epoch": 7.63, "learning_rate": 2.1829577464788732e-05, "loss": 2.4168, "step": 6806 }, { "epoch": 7.63, "learning_rate": 2.182816901408451e-05, "loss": 2.2277, "step": 6807 }, { "epoch": 7.63, "learning_rate": 2.1826760563380285e-05, "loss": 2.3615, "step": 6808 }, { "epoch": 7.63, "learning_rate": 2.1825352112676058e-05, "loss": 2.3514, "step": 6809 }, { "epoch": 7.63, "learning_rate": 2.182394366197183e-05, "loss": 2.349, "step": 6810 }, { "epoch": 7.64, "learning_rate": 2.1822535211267607e-05, "loss": 2.316, "step": 6811 }, { "epoch": 7.64, "learning_rate": 2.182112676056338e-05, "loss": 2.0386, "step": 6812 }, { "epoch": 7.64, "learning_rate": 2.1819718309859156e-05, "loss": 2.262, "step": 6813 }, { "epoch": 7.64, "learning_rate": 2.181830985915493e-05, "loss": 2.2321, "step": 6814 }, { "epoch": 7.64, "learning_rate": 2.1816901408450706e-05, "loss": 2.5786, "step": 6815 }, { "epoch": 7.64, "learning_rate": 2.181549295774648e-05, "loss": 2.4604, "step": 6816 }, { "epoch": 7.64, "learning_rate": 2.1814084507042255e-05, "loss": 2.5995, "step": 6817 }, { "epoch": 7.64, "learning_rate": 2.1812676056338028e-05, "loss": 2.1603, "step": 6818 }, { "epoch": 7.64, "learning_rate": 2.18112676056338e-05, "loss": 2.181, "step": 6819 }, { "epoch": 7.65, "learning_rate": 2.1809859154929577e-05, "loss": 2.4447, "step": 6820 }, { "epoch": 7.65, "learning_rate": 2.180845070422535e-05, "loss": 2.0674, "step": 6821 }, { "epoch": 7.65, "learning_rate": 2.180704225352113e-05, "loss": 2.2796, "step": 6822 }, { "epoch": 7.65, "learning_rate": 2.1805633802816903e-05, "loss": 2.1353, "step": 6823 }, { "epoch": 7.65, "learning_rate": 2.180422535211268e-05, "loss": 2.1583, "step": 6824 }, { "epoch": 7.65, "learning_rate": 2.1802816901408452e-05, "loss": 2.4385, "step": 6825 }, { "epoch": 7.65, "learning_rate": 2.1801408450704225e-05, "loss": 2.2535, "step": 6826 }, { "epoch": 7.65, "learning_rate": 2.18e-05, "loss": 2.1202, "step": 6827 }, { "epoch": 7.65, "learning_rate": 2.1798591549295774e-05, "loss": 2.3117, "step": 6828 }, { "epoch": 7.66, "learning_rate": 2.179718309859155e-05, "loss": 2.4525, "step": 6829 }, { "epoch": 7.66, "learning_rate": 2.1795774647887324e-05, "loss": 2.0921, "step": 6830 }, { "epoch": 7.66, "learning_rate": 2.17943661971831e-05, "loss": 2.2902, "step": 6831 }, { "epoch": 7.66, "learning_rate": 2.1792957746478873e-05, "loss": 2.1264, "step": 6832 }, { "epoch": 7.66, "learning_rate": 2.179154929577465e-05, "loss": 2.1387, "step": 6833 }, { "epoch": 7.66, "learning_rate": 2.1790140845070422e-05, "loss": 2.2204, "step": 6834 }, { "epoch": 7.66, "learning_rate": 2.1788732394366195e-05, "loss": 2.2274, "step": 6835 }, { "epoch": 7.66, "learning_rate": 2.178732394366197e-05, "loss": 2.3262, "step": 6836 }, { "epoch": 7.66, "learning_rate": 2.1785915492957748e-05, "loss": 1.9537, "step": 6837 }, { "epoch": 7.67, "learning_rate": 2.1784507042253524e-05, "loss": 2.2176, "step": 6838 }, { "epoch": 7.67, "learning_rate": 2.1783098591549297e-05, "loss": 2.1151, "step": 6839 }, { "epoch": 7.67, "learning_rate": 2.1781690140845073e-05, "loss": 2.0858, "step": 6840 }, { "epoch": 7.67, "learning_rate": 2.1780281690140846e-05, "loss": 2.1401, "step": 6841 }, { "epoch": 7.67, "learning_rate": 2.177887323943662e-05, "loss": 2.1871, "step": 6842 }, { "epoch": 7.67, "learning_rate": 2.1777464788732396e-05, "loss": 1.8746, "step": 6843 }, { "epoch": 7.67, "learning_rate": 2.177605633802817e-05, "loss": 2.248, "step": 6844 }, { "epoch": 7.67, "learning_rate": 2.1774647887323945e-05, "loss": 3.1379, "step": 6845 }, { "epoch": 7.67, "learning_rate": 2.1773239436619718e-05, "loss": 2.9258, "step": 6846 }, { "epoch": 7.68, "learning_rate": 2.1771830985915494e-05, "loss": 2.8921, "step": 6847 }, { "epoch": 7.68, "learning_rate": 2.1770422535211267e-05, "loss": 2.8456, "step": 6848 }, { "epoch": 7.68, "learning_rate": 2.1769014084507043e-05, "loss": 2.3218, "step": 6849 }, { "epoch": 7.68, "learning_rate": 2.1767605633802816e-05, "loss": 2.4239, "step": 6850 }, { "epoch": 7.68, "learning_rate": 2.176619718309859e-05, "loss": 2.4908, "step": 6851 }, { "epoch": 7.68, "learning_rate": 2.176478873239437e-05, "loss": 2.3127, "step": 6852 }, { "epoch": 7.68, "learning_rate": 2.1763380281690142e-05, "loss": 2.3386, "step": 6853 }, { "epoch": 7.68, "learning_rate": 2.1761971830985918e-05, "loss": 2.6138, "step": 6854 }, { "epoch": 7.68, "learning_rate": 2.176056338028169e-05, "loss": 2.5721, "step": 6855 }, { "epoch": 7.69, "learning_rate": 2.1759154929577468e-05, "loss": 2.3243, "step": 6856 }, { "epoch": 7.69, "learning_rate": 2.175774647887324e-05, "loss": 2.3632, "step": 6857 }, { "epoch": 7.69, "learning_rate": 2.1756338028169013e-05, "loss": 2.4069, "step": 6858 }, { "epoch": 7.69, "learning_rate": 2.175492957746479e-05, "loss": 2.3792, "step": 6859 }, { "epoch": 7.69, "learning_rate": 2.1753521126760563e-05, "loss": 2.2131, "step": 6860 }, { "epoch": 7.69, "learning_rate": 2.175211267605634e-05, "loss": 2.401, "step": 6861 }, { "epoch": 7.69, "learning_rate": 2.1750704225352112e-05, "loss": 2.2487, "step": 6862 }, { "epoch": 7.69, "learning_rate": 2.1749295774647888e-05, "loss": 2.3026, "step": 6863 }, { "epoch": 7.7, "learning_rate": 2.174788732394366e-05, "loss": 2.4887, "step": 6864 }, { "epoch": 7.7, "learning_rate": 2.1746478873239438e-05, "loss": 2.331, "step": 6865 }, { "epoch": 7.7, "learning_rate": 2.174507042253521e-05, "loss": 2.3177, "step": 6866 }, { "epoch": 7.7, "learning_rate": 2.1743661971830983e-05, "loss": 2.1822, "step": 6867 }, { "epoch": 7.7, "learning_rate": 2.1742253521126763e-05, "loss": 2.4108, "step": 6868 }, { "epoch": 7.7, "learning_rate": 2.1740845070422536e-05, "loss": 2.3544, "step": 6869 }, { "epoch": 7.7, "learning_rate": 2.1739436619718312e-05, "loss": 2.4078, "step": 6870 }, { "epoch": 7.7, "learning_rate": 2.1738028169014085e-05, "loss": 2.0562, "step": 6871 }, { "epoch": 7.7, "learning_rate": 2.1736619718309862e-05, "loss": 2.1584, "step": 6872 }, { "epoch": 7.71, "learning_rate": 2.1735211267605635e-05, "loss": 2.3469, "step": 6873 }, { "epoch": 7.71, "learning_rate": 2.1733802816901408e-05, "loss": 2.3905, "step": 6874 }, { "epoch": 7.71, "learning_rate": 2.1732394366197184e-05, "loss": 2.2639, "step": 6875 }, { "epoch": 7.71, "learning_rate": 2.1730985915492957e-05, "loss": 2.2406, "step": 6876 }, { "epoch": 7.71, "learning_rate": 2.1729577464788733e-05, "loss": 2.1491, "step": 6877 }, { "epoch": 7.71, "learning_rate": 2.1728169014084506e-05, "loss": 2.2012, "step": 6878 }, { "epoch": 7.71, "learning_rate": 2.1726760563380282e-05, "loss": 2.1474, "step": 6879 }, { "epoch": 7.71, "learning_rate": 2.1725352112676055e-05, "loss": 1.9671, "step": 6880 }, { "epoch": 7.71, "learning_rate": 2.1723943661971832e-05, "loss": 2.0524, "step": 6881 }, { "epoch": 7.72, "learning_rate": 2.1722535211267605e-05, "loss": 2.1066, "step": 6882 }, { "epoch": 7.72, "learning_rate": 2.172112676056338e-05, "loss": 2.151, "step": 6883 }, { "epoch": 7.72, "learning_rate": 2.1719718309859157e-05, "loss": 2.0231, "step": 6884 }, { "epoch": 7.72, "learning_rate": 2.171830985915493e-05, "loss": 2.0493, "step": 6885 }, { "epoch": 7.72, "learning_rate": 2.1716901408450707e-05, "loss": 2.2862, "step": 6886 }, { "epoch": 7.72, "learning_rate": 2.171549295774648e-05, "loss": 1.9135, "step": 6887 }, { "epoch": 7.72, "learning_rate": 2.1714084507042256e-05, "loss": 1.7882, "step": 6888 }, { "epoch": 7.72, "learning_rate": 2.171267605633803e-05, "loss": 2.3272, "step": 6889 }, { "epoch": 7.72, "learning_rate": 2.1711267605633802e-05, "loss": 2.1987, "step": 6890 }, { "epoch": 7.73, "learning_rate": 2.1709859154929578e-05, "loss": 2.0698, "step": 6891 }, { "epoch": 7.73, "learning_rate": 2.170845070422535e-05, "loss": 2.071, "step": 6892 }, { "epoch": 7.73, "learning_rate": 2.1707042253521127e-05, "loss": 2.0876, "step": 6893 }, { "epoch": 7.73, "learning_rate": 2.17056338028169e-05, "loss": 1.9254, "step": 6894 }, { "epoch": 7.73, "learning_rate": 2.1704225352112677e-05, "loss": 3.2858, "step": 6895 }, { "epoch": 7.73, "learning_rate": 2.170281690140845e-05, "loss": 3.192, "step": 6896 }, { "epoch": 7.73, "learning_rate": 2.1701408450704226e-05, "loss": 2.8232, "step": 6897 }, { "epoch": 7.73, "learning_rate": 2.1700000000000002e-05, "loss": 2.8357, "step": 6898 }, { "epoch": 7.73, "learning_rate": 2.1698591549295775e-05, "loss": 2.8037, "step": 6899 }, { "epoch": 7.74, "learning_rate": 2.169718309859155e-05, "loss": 2.4146, "step": 6900 }, { "epoch": 7.74, "learning_rate": 2.1695774647887325e-05, "loss": 2.1394, "step": 6901 }, { "epoch": 7.74, "learning_rate": 2.16943661971831e-05, "loss": 2.2446, "step": 6902 }, { "epoch": 7.74, "learning_rate": 2.1692957746478874e-05, "loss": 2.3348, "step": 6903 }, { "epoch": 7.74, "learning_rate": 2.169154929577465e-05, "loss": 2.2759, "step": 6904 }, { "epoch": 7.74, "learning_rate": 2.1690140845070423e-05, "loss": 2.3175, "step": 6905 }, { "epoch": 7.74, "learning_rate": 2.1688732394366196e-05, "loss": 2.619, "step": 6906 }, { "epoch": 7.74, "learning_rate": 2.1687323943661972e-05, "loss": 2.4341, "step": 6907 }, { "epoch": 7.74, "learning_rate": 2.1685915492957745e-05, "loss": 2.3884, "step": 6908 }, { "epoch": 7.75, "learning_rate": 2.168450704225352e-05, "loss": 2.1557, "step": 6909 }, { "epoch": 7.75, "learning_rate": 2.1683098591549295e-05, "loss": 2.305, "step": 6910 }, { "epoch": 7.75, "learning_rate": 2.168169014084507e-05, "loss": 2.0761, "step": 6911 }, { "epoch": 7.75, "learning_rate": 2.1680281690140844e-05, "loss": 2.2208, "step": 6912 }, { "epoch": 7.75, "learning_rate": 2.167887323943662e-05, "loss": 2.2062, "step": 6913 }, { "epoch": 7.75, "learning_rate": 2.1677464788732397e-05, "loss": 2.2864, "step": 6914 }, { "epoch": 7.75, "learning_rate": 2.167605633802817e-05, "loss": 2.1582, "step": 6915 }, { "epoch": 7.75, "learning_rate": 2.1674647887323946e-05, "loss": 2.2932, "step": 6916 }, { "epoch": 7.75, "learning_rate": 2.167323943661972e-05, "loss": 2.0992, "step": 6917 }, { "epoch": 7.76, "learning_rate": 2.1671830985915495e-05, "loss": 2.2813, "step": 6918 }, { "epoch": 7.76, "learning_rate": 2.1670422535211268e-05, "loss": 2.0891, "step": 6919 }, { "epoch": 7.76, "learning_rate": 2.1669014084507044e-05, "loss": 2.1551, "step": 6920 }, { "epoch": 7.76, "learning_rate": 2.1667605633802817e-05, "loss": 2.3675, "step": 6921 }, { "epoch": 7.76, "learning_rate": 2.166619718309859e-05, "loss": 2.1424, "step": 6922 }, { "epoch": 7.76, "learning_rate": 2.1664788732394367e-05, "loss": 2.2546, "step": 6923 }, { "epoch": 7.76, "learning_rate": 2.166338028169014e-05, "loss": 2.1601, "step": 6924 }, { "epoch": 7.76, "learning_rate": 2.1661971830985916e-05, "loss": 2.2736, "step": 6925 }, { "epoch": 7.76, "learning_rate": 2.166056338028169e-05, "loss": 2.4315, "step": 6926 }, { "epoch": 7.77, "learning_rate": 2.1659154929577465e-05, "loss": 2.0777, "step": 6927 }, { "epoch": 7.77, "learning_rate": 2.165774647887324e-05, "loss": 2.3052, "step": 6928 }, { "epoch": 7.77, "learning_rate": 2.1656338028169014e-05, "loss": 1.9644, "step": 6929 }, { "epoch": 7.77, "learning_rate": 2.165492957746479e-05, "loss": 1.9863, "step": 6930 }, { "epoch": 7.77, "learning_rate": 2.1653521126760564e-05, "loss": 2.136, "step": 6931 }, { "epoch": 7.77, "learning_rate": 2.165211267605634e-05, "loss": 2.1281, "step": 6932 }, { "epoch": 7.77, "learning_rate": 2.1650704225352113e-05, "loss": 2.222, "step": 6933 }, { "epoch": 7.77, "learning_rate": 2.164929577464789e-05, "loss": 1.8647, "step": 6934 }, { "epoch": 7.77, "learning_rate": 2.1647887323943662e-05, "loss": 2.0431, "step": 6935 }, { "epoch": 7.78, "learning_rate": 2.164647887323944e-05, "loss": 2.3755, "step": 6936 }, { "epoch": 7.78, "learning_rate": 2.164507042253521e-05, "loss": 2.1234, "step": 6937 }, { "epoch": 7.78, "learning_rate": 2.1643661971830984e-05, "loss": 2.4024, "step": 6938 }, { "epoch": 7.78, "learning_rate": 2.164225352112676e-05, "loss": 2.101, "step": 6939 }, { "epoch": 7.78, "learning_rate": 2.1640845070422534e-05, "loss": 1.8734, "step": 6940 }, { "epoch": 7.78, "learning_rate": 2.163943661971831e-05, "loss": 1.9136, "step": 6941 }, { "epoch": 7.78, "learning_rate": 2.1638028169014083e-05, "loss": 1.8324, "step": 6942 }, { "epoch": 7.78, "learning_rate": 2.1636619718309863e-05, "loss": 1.828, "step": 6943 }, { "epoch": 7.78, "learning_rate": 2.1635211267605636e-05, "loss": 1.7279, "step": 6944 }, { "epoch": 7.79, "learning_rate": 2.163380281690141e-05, "loss": 2.8768, "step": 6945 }, { "epoch": 7.79, "learning_rate": 2.1632394366197185e-05, "loss": 2.7629, "step": 6946 }, { "epoch": 7.79, "learning_rate": 2.1630985915492958e-05, "loss": 2.7948, "step": 6947 }, { "epoch": 7.79, "learning_rate": 2.1629577464788734e-05, "loss": 2.9873, "step": 6948 }, { "epoch": 7.79, "learning_rate": 2.1628169014084507e-05, "loss": 2.4781, "step": 6949 }, { "epoch": 7.79, "learning_rate": 2.1626760563380283e-05, "loss": 2.4783, "step": 6950 }, { "epoch": 7.79, "learning_rate": 2.1625352112676056e-05, "loss": 2.1679, "step": 6951 }, { "epoch": 7.79, "learning_rate": 2.1623943661971833e-05, "loss": 2.6306, "step": 6952 }, { "epoch": 7.79, "learning_rate": 2.1622535211267606e-05, "loss": 2.4691, "step": 6953 }, { "epoch": 7.8, "learning_rate": 2.162112676056338e-05, "loss": 2.4042, "step": 6954 }, { "epoch": 7.8, "learning_rate": 2.1619718309859155e-05, "loss": 2.3445, "step": 6955 }, { "epoch": 7.8, "learning_rate": 2.1618309859154928e-05, "loss": 2.5143, "step": 6956 }, { "epoch": 7.8, "learning_rate": 2.1616901408450704e-05, "loss": 2.2345, "step": 6957 }, { "epoch": 7.8, "learning_rate": 2.161549295774648e-05, "loss": 2.4687, "step": 6958 }, { "epoch": 7.8, "learning_rate": 2.1614084507042257e-05, "loss": 2.0577, "step": 6959 }, { "epoch": 7.8, "learning_rate": 2.161267605633803e-05, "loss": 1.9408, "step": 6960 }, { "epoch": 7.8, "learning_rate": 2.1611267605633803e-05, "loss": 2.2557, "step": 6961 }, { "epoch": 7.8, "learning_rate": 2.160985915492958e-05, "loss": 2.1999, "step": 6962 }, { "epoch": 7.81, "learning_rate": 2.1608450704225352e-05, "loss": 2.0507, "step": 6963 }, { "epoch": 7.81, "learning_rate": 2.160704225352113e-05, "loss": 2.4278, "step": 6964 }, { "epoch": 7.81, "learning_rate": 2.16056338028169e-05, "loss": 2.3812, "step": 6965 }, { "epoch": 7.81, "learning_rate": 2.1604225352112678e-05, "loss": 2.1759, "step": 6966 }, { "epoch": 7.81, "learning_rate": 2.160281690140845e-05, "loss": 2.1889, "step": 6967 }, { "epoch": 7.81, "learning_rate": 2.1601408450704227e-05, "loss": 2.2784, "step": 6968 }, { "epoch": 7.81, "learning_rate": 2.16e-05, "loss": 2.2607, "step": 6969 }, { "epoch": 7.81, "learning_rate": 2.1598591549295773e-05, "loss": 2.2525, "step": 6970 }, { "epoch": 7.82, "learning_rate": 2.159718309859155e-05, "loss": 2.3917, "step": 6971 }, { "epoch": 7.82, "learning_rate": 2.1595774647887322e-05, "loss": 2.2137, "step": 6972 }, { "epoch": 7.82, "learning_rate": 2.1594366197183102e-05, "loss": 2.0389, "step": 6973 }, { "epoch": 7.82, "learning_rate": 2.1592957746478875e-05, "loss": 2.0057, "step": 6974 }, { "epoch": 7.82, "learning_rate": 2.159154929577465e-05, "loss": 2.3299, "step": 6975 }, { "epoch": 7.82, "learning_rate": 2.1590140845070424e-05, "loss": 1.9035, "step": 6976 }, { "epoch": 7.82, "learning_rate": 2.1588732394366197e-05, "loss": 2.2144, "step": 6977 }, { "epoch": 7.82, "learning_rate": 2.1587323943661973e-05, "loss": 2.2997, "step": 6978 }, { "epoch": 7.82, "learning_rate": 2.1585915492957746e-05, "loss": 2.1458, "step": 6979 }, { "epoch": 7.83, "learning_rate": 2.1584507042253523e-05, "loss": 2.2711, "step": 6980 }, { "epoch": 7.83, "learning_rate": 2.1583098591549296e-05, "loss": 2.2209, "step": 6981 }, { "epoch": 7.83, "learning_rate": 2.1581690140845072e-05, "loss": 1.8433, "step": 6982 }, { "epoch": 7.83, "learning_rate": 2.1580281690140845e-05, "loss": 1.9725, "step": 6983 }, { "epoch": 7.83, "learning_rate": 2.157887323943662e-05, "loss": 2.0323, "step": 6984 }, { "epoch": 7.83, "learning_rate": 2.1577464788732394e-05, "loss": 2.0681, "step": 6985 }, { "epoch": 7.83, "learning_rate": 2.1576056338028167e-05, "loss": 2.0801, "step": 6986 }, { "epoch": 7.83, "learning_rate": 2.1574647887323943e-05, "loss": 1.9036, "step": 6987 }, { "epoch": 7.83, "learning_rate": 2.1573239436619716e-05, "loss": 1.9299, "step": 6988 }, { "epoch": 7.84, "learning_rate": 2.1571830985915496e-05, "loss": 1.9661, "step": 6989 }, { "epoch": 7.84, "learning_rate": 2.157042253521127e-05, "loss": 2.0646, "step": 6990 }, { "epoch": 7.84, "learning_rate": 2.1569014084507045e-05, "loss": 2.2165, "step": 6991 }, { "epoch": 7.84, "learning_rate": 2.1567605633802818e-05, "loss": 2.1303, "step": 6992 }, { "epoch": 7.84, "learning_rate": 2.156619718309859e-05, "loss": 1.6993, "step": 6993 }, { "epoch": 7.84, "learning_rate": 2.1564788732394368e-05, "loss": 1.5544, "step": 6994 }, { "epoch": 7.84, "learning_rate": 2.156338028169014e-05, "loss": 3.4347, "step": 6995 }, { "epoch": 7.84, "learning_rate": 2.1561971830985917e-05, "loss": 2.8264, "step": 6996 }, { "epoch": 7.84, "learning_rate": 2.156056338028169e-05, "loss": 2.9127, "step": 6997 }, { "epoch": 7.85, "learning_rate": 2.1559154929577466e-05, "loss": 2.5187, "step": 6998 }, { "epoch": 7.85, "learning_rate": 2.155774647887324e-05, "loss": 2.282, "step": 6999 }, { "epoch": 7.85, "learning_rate": 2.1556338028169015e-05, "loss": 2.1923, "step": 7000 }, { "epoch": 7.85, "learning_rate": 2.1554929577464788e-05, "loss": 2.4254, "step": 7001 }, { "epoch": 7.85, "learning_rate": 2.155352112676056e-05, "loss": 2.1335, "step": 7002 }, { "epoch": 7.85, "learning_rate": 2.1552112676056338e-05, "loss": 2.1682, "step": 7003 }, { "epoch": 7.85, "learning_rate": 2.1550704225352114e-05, "loss": 2.4376, "step": 7004 }, { "epoch": 7.85, "learning_rate": 2.154929577464789e-05, "loss": 2.3918, "step": 7005 }, { "epoch": 7.85, "learning_rate": 2.1547887323943663e-05, "loss": 2.3699, "step": 7006 }, { "epoch": 7.86, "learning_rate": 2.154647887323944e-05, "loss": 2.3865, "step": 7007 }, { "epoch": 7.86, "learning_rate": 2.1545070422535212e-05, "loss": 2.2675, "step": 7008 }, { "epoch": 7.86, "learning_rate": 2.1543661971830985e-05, "loss": 2.1088, "step": 7009 }, { "epoch": 7.86, "learning_rate": 2.1542253521126762e-05, "loss": 2.261, "step": 7010 }, { "epoch": 7.86, "learning_rate": 2.1540845070422535e-05, "loss": 2.3649, "step": 7011 }, { "epoch": 7.86, "learning_rate": 2.153943661971831e-05, "loss": 2.2595, "step": 7012 }, { "epoch": 7.86, "learning_rate": 2.1538028169014084e-05, "loss": 2.1261, "step": 7013 }, { "epoch": 7.86, "learning_rate": 2.153661971830986e-05, "loss": 2.0751, "step": 7014 }, { "epoch": 7.86, "learning_rate": 2.1535211267605633e-05, "loss": 2.1878, "step": 7015 }, { "epoch": 7.87, "learning_rate": 2.153380281690141e-05, "loss": 2.3453, "step": 7016 }, { "epoch": 7.87, "learning_rate": 2.1532394366197182e-05, "loss": 2.1321, "step": 7017 }, { "epoch": 7.87, "learning_rate": 2.1530985915492955e-05, "loss": 2.1897, "step": 7018 }, { "epoch": 7.87, "learning_rate": 2.1529577464788735e-05, "loss": 2.2863, "step": 7019 }, { "epoch": 7.87, "learning_rate": 2.1528169014084508e-05, "loss": 2.179, "step": 7020 }, { "epoch": 7.87, "learning_rate": 2.1526760563380284e-05, "loss": 2.1752, "step": 7021 }, { "epoch": 7.87, "learning_rate": 2.1525352112676057e-05, "loss": 2.0804, "step": 7022 }, { "epoch": 7.87, "learning_rate": 2.1523943661971834e-05, "loss": 2.1943, "step": 7023 }, { "epoch": 7.87, "learning_rate": 2.1522535211267607e-05, "loss": 2.2726, "step": 7024 }, { "epoch": 7.88, "learning_rate": 2.152112676056338e-05, "loss": 1.9935, "step": 7025 }, { "epoch": 7.88, "learning_rate": 2.1519718309859156e-05, "loss": 2.0171, "step": 7026 }, { "epoch": 7.88, "learning_rate": 2.151830985915493e-05, "loss": 2.4781, "step": 7027 }, { "epoch": 7.88, "learning_rate": 2.1516901408450705e-05, "loss": 2.0737, "step": 7028 }, { "epoch": 7.88, "learning_rate": 2.1515492957746478e-05, "loss": 2.4198, "step": 7029 }, { "epoch": 7.88, "learning_rate": 2.1514084507042254e-05, "loss": 2.0415, "step": 7030 }, { "epoch": 7.88, "learning_rate": 2.1512676056338027e-05, "loss": 2.1463, "step": 7031 }, { "epoch": 7.88, "learning_rate": 2.15112676056338e-05, "loss": 1.9597, "step": 7032 }, { "epoch": 7.88, "learning_rate": 2.1509859154929577e-05, "loss": 1.9874, "step": 7033 }, { "epoch": 7.89, "learning_rate": 2.1508450704225353e-05, "loss": 2.1977, "step": 7034 }, { "epoch": 7.89, "learning_rate": 2.150704225352113e-05, "loss": 2.1801, "step": 7035 }, { "epoch": 7.89, "learning_rate": 2.1505633802816902e-05, "loss": 2.1997, "step": 7036 }, { "epoch": 7.89, "learning_rate": 2.150422535211268e-05, "loss": 2.0134, "step": 7037 }, { "epoch": 7.89, "learning_rate": 2.150281690140845e-05, "loss": 2.1805, "step": 7038 }, { "epoch": 7.89, "learning_rate": 2.1501408450704228e-05, "loss": 2.3369, "step": 7039 }, { "epoch": 7.89, "learning_rate": 2.15e-05, "loss": 2.1432, "step": 7040 }, { "epoch": 7.89, "learning_rate": 2.1498591549295774e-05, "loss": 1.6823, "step": 7041 }, { "epoch": 7.89, "learning_rate": 2.149718309859155e-05, "loss": 1.9925, "step": 7042 }, { "epoch": 7.9, "learning_rate": 2.1495774647887323e-05, "loss": 1.9649, "step": 7043 }, { "epoch": 7.9, "learning_rate": 2.14943661971831e-05, "loss": 1.8011, "step": 7044 }, { "epoch": 7.9, "learning_rate": 2.1492957746478872e-05, "loss": 3.247, "step": 7045 }, { "epoch": 7.9, "learning_rate": 2.149154929577465e-05, "loss": 2.8947, "step": 7046 }, { "epoch": 7.9, "learning_rate": 2.149014084507042e-05, "loss": 2.4269, "step": 7047 }, { "epoch": 7.9, "learning_rate": 2.1488732394366195e-05, "loss": 2.6737, "step": 7048 }, { "epoch": 7.9, "learning_rate": 2.1487323943661974e-05, "loss": 2.3004, "step": 7049 }, { "epoch": 7.9, "learning_rate": 2.1485915492957747e-05, "loss": 2.4435, "step": 7050 }, { "epoch": 7.9, "learning_rate": 2.1484507042253524e-05, "loss": 2.1899, "step": 7051 }, { "epoch": 7.91, "learning_rate": 2.1483098591549296e-05, "loss": 2.3508, "step": 7052 }, { "epoch": 7.91, "learning_rate": 2.1481690140845073e-05, "loss": 2.3897, "step": 7053 }, { "epoch": 7.91, "learning_rate": 2.1480281690140846e-05, "loss": 2.2821, "step": 7054 }, { "epoch": 7.91, "learning_rate": 2.1478873239436622e-05, "loss": 2.1888, "step": 7055 }, { "epoch": 7.91, "learning_rate": 2.1477464788732395e-05, "loss": 2.2507, "step": 7056 }, { "epoch": 7.91, "learning_rate": 2.1476056338028168e-05, "loss": 2.0515, "step": 7057 }, { "epoch": 7.91, "learning_rate": 2.1474647887323944e-05, "loss": 2.1673, "step": 7058 }, { "epoch": 7.91, "learning_rate": 2.1473239436619717e-05, "loss": 2.1569, "step": 7059 }, { "epoch": 7.91, "learning_rate": 2.1471830985915494e-05, "loss": 2.2522, "step": 7060 }, { "epoch": 7.92, "learning_rate": 2.1470422535211267e-05, "loss": 2.2473, "step": 7061 }, { "epoch": 7.92, "learning_rate": 2.1469014084507043e-05, "loss": 2.2284, "step": 7062 }, { "epoch": 7.92, "learning_rate": 2.1467605633802816e-05, "loss": 2.3012, "step": 7063 }, { "epoch": 7.92, "learning_rate": 2.1466197183098592e-05, "loss": 2.0172, "step": 7064 }, { "epoch": 7.92, "learning_rate": 2.146478873239437e-05, "loss": 2.0642, "step": 7065 }, { "epoch": 7.92, "learning_rate": 2.146338028169014e-05, "loss": 2.1714, "step": 7066 }, { "epoch": 7.92, "learning_rate": 2.1461971830985918e-05, "loss": 2.0, "step": 7067 }, { "epoch": 7.92, "learning_rate": 2.146056338028169e-05, "loss": 2.0672, "step": 7068 }, { "epoch": 7.92, "learning_rate": 2.1459154929577467e-05, "loss": 2.1107, "step": 7069 }, { "epoch": 7.93, "learning_rate": 2.145774647887324e-05, "loss": 2.4292, "step": 7070 }, { "epoch": 7.93, "learning_rate": 2.1456338028169016e-05, "loss": 1.9885, "step": 7071 }, { "epoch": 7.93, "learning_rate": 2.145492957746479e-05, "loss": 1.8756, "step": 7072 }, { "epoch": 7.93, "learning_rate": 2.1453521126760562e-05, "loss": 2.2338, "step": 7073 }, { "epoch": 7.93, "learning_rate": 2.145211267605634e-05, "loss": 2.5096, "step": 7074 }, { "epoch": 7.93, "learning_rate": 2.145070422535211e-05, "loss": 2.0169, "step": 7075 }, { "epoch": 7.93, "learning_rate": 2.1449295774647888e-05, "loss": 2.0716, "step": 7076 }, { "epoch": 7.93, "learning_rate": 2.144788732394366e-05, "loss": 1.7968, "step": 7077 }, { "epoch": 7.93, "learning_rate": 2.1446478873239437e-05, "loss": 2.1001, "step": 7078 }, { "epoch": 7.94, "learning_rate": 2.1445070422535213e-05, "loss": 2.1546, "step": 7079 }, { "epoch": 7.94, "learning_rate": 2.1443661971830986e-05, "loss": 2.0841, "step": 7080 }, { "epoch": 7.94, "learning_rate": 2.1442253521126763e-05, "loss": 2.1381, "step": 7081 }, { "epoch": 7.94, "learning_rate": 2.1440845070422536e-05, "loss": 2.1393, "step": 7082 }, { "epoch": 7.94, "learning_rate": 2.1439436619718312e-05, "loss": 1.9302, "step": 7083 }, { "epoch": 7.94, "learning_rate": 2.1438028169014085e-05, "loss": 2.0344, "step": 7084 }, { "epoch": 7.94, "learning_rate": 2.143661971830986e-05, "loss": 1.987, "step": 7085 }, { "epoch": 7.94, "learning_rate": 2.1435211267605634e-05, "loss": 2.0961, "step": 7086 }, { "epoch": 7.95, "learning_rate": 2.143380281690141e-05, "loss": 2.0729, "step": 7087 }, { "epoch": 7.95, "learning_rate": 2.1432394366197183e-05, "loss": 1.8667, "step": 7088 }, { "epoch": 7.95, "learning_rate": 2.1430985915492956e-05, "loss": 2.044, "step": 7089 }, { "epoch": 7.95, "learning_rate": 2.1429577464788733e-05, "loss": 1.7509, "step": 7090 }, { "epoch": 7.95, "learning_rate": 2.1428169014084506e-05, "loss": 1.9398, "step": 7091 }, { "epoch": 7.95, "learning_rate": 2.1426760563380282e-05, "loss": 1.9819, "step": 7092 }, { "epoch": 7.95, "learning_rate": 2.1425352112676055e-05, "loss": 1.9159, "step": 7093 }, { "epoch": 7.95, "learning_rate": 2.1423943661971835e-05, "loss": 1.56, "step": 7094 }, { "epoch": 7.95, "learning_rate": 2.1422535211267608e-05, "loss": 2.8832, "step": 7095 }, { "epoch": 7.96, "learning_rate": 2.142112676056338e-05, "loss": 2.746, "step": 7096 }, { "epoch": 7.96, "learning_rate": 2.1419718309859157e-05, "loss": 2.6213, "step": 7097 }, { "epoch": 7.96, "learning_rate": 2.141830985915493e-05, "loss": 2.5198, "step": 7098 }, { "epoch": 7.96, "learning_rate": 2.1416901408450706e-05, "loss": 2.4723, "step": 7099 }, { "epoch": 7.96, "learning_rate": 2.141549295774648e-05, "loss": 2.417, "step": 7100 }, { "epoch": 7.96, "learning_rate": 2.1414084507042255e-05, "loss": 2.2135, "step": 7101 }, { "epoch": 7.96, "learning_rate": 2.141267605633803e-05, "loss": 2.3136, "step": 7102 }, { "epoch": 7.96, "learning_rate": 2.1411267605633805e-05, "loss": 2.205, "step": 7103 }, { "epoch": 7.96, "learning_rate": 2.1409859154929578e-05, "loss": 2.2752, "step": 7104 }, { "epoch": 7.97, "learning_rate": 2.140845070422535e-05, "loss": 2.2887, "step": 7105 }, { "epoch": 7.97, "learning_rate": 2.1407042253521127e-05, "loss": 2.0473, "step": 7106 }, { "epoch": 7.97, "learning_rate": 2.14056338028169e-05, "loss": 2.13, "step": 7107 }, { "epoch": 7.97, "learning_rate": 2.1404225352112676e-05, "loss": 2.333, "step": 7108 }, { "epoch": 7.97, "learning_rate": 2.140281690140845e-05, "loss": 2.4434, "step": 7109 }, { "epoch": 7.97, "learning_rate": 2.140140845070423e-05, "loss": 2.1152, "step": 7110 }, { "epoch": 7.97, "learning_rate": 2.1400000000000002e-05, "loss": 2.0006, "step": 7111 }, { "epoch": 7.97, "learning_rate": 2.1398591549295775e-05, "loss": 2.3255, "step": 7112 }, { "epoch": 7.97, "learning_rate": 2.139718309859155e-05, "loss": 2.026, "step": 7113 }, { "epoch": 7.98, "learning_rate": 2.1395774647887324e-05, "loss": 1.9333, "step": 7114 }, { "epoch": 7.98, "learning_rate": 2.13943661971831e-05, "loss": 2.1686, "step": 7115 }, { "epoch": 7.98, "learning_rate": 2.1392957746478873e-05, "loss": 2.0571, "step": 7116 }, { "epoch": 7.98, "learning_rate": 2.139154929577465e-05, "loss": 1.9907, "step": 7117 }, { "epoch": 7.98, "learning_rate": 2.1390140845070423e-05, "loss": 2.0603, "step": 7118 }, { "epoch": 7.98, "learning_rate": 2.13887323943662e-05, "loss": 2.1758, "step": 7119 }, { "epoch": 7.98, "learning_rate": 2.1387323943661972e-05, "loss": 1.9883, "step": 7120 }, { "epoch": 7.98, "learning_rate": 2.1385915492957745e-05, "loss": 2.1863, "step": 7121 }, { "epoch": 7.98, "learning_rate": 2.138450704225352e-05, "loss": 2.1236, "step": 7122 }, { "epoch": 7.99, "learning_rate": 2.1383098591549294e-05, "loss": 2.2821, "step": 7123 }, { "epoch": 7.99, "learning_rate": 2.138169014084507e-05, "loss": 1.7849, "step": 7124 }, { "epoch": 7.99, "learning_rate": 2.1380281690140847e-05, "loss": 2.1679, "step": 7125 }, { "epoch": 7.99, "learning_rate": 2.1378873239436623e-05, "loss": 2.1873, "step": 7126 }, { "epoch": 7.99, "learning_rate": 2.1377464788732396e-05, "loss": 1.7239, "step": 7127 }, { "epoch": 7.99, "learning_rate": 2.137605633802817e-05, "loss": 1.9274, "step": 7128 }, { "epoch": 7.99, "learning_rate": 2.1374647887323945e-05, "loss": 2.0503, "step": 7129 }, { "epoch": 7.99, "learning_rate": 2.1373239436619718e-05, "loss": 2.066, "step": 7130 }, { "epoch": 7.99, "learning_rate": 2.1371830985915495e-05, "loss": 1.8156, "step": 7131 }, { "epoch": 8.0, "learning_rate": 2.1370422535211267e-05, "loss": 2.0676, "step": 7132 }, { "epoch": 8.0, "learning_rate": 2.1369014084507044e-05, "loss": 1.6394, "step": 7133 }, { "epoch": 8.0, "learning_rate": 2.1367605633802817e-05, "loss": 1.7507, "step": 7134 }, { "epoch": 8.0, "learning_rate": 2.136619718309859e-05, "loss": 2.3336, "step": 7135 }, { "epoch": 8.0, "learning_rate": 2.1364788732394366e-05, "loss": 1.7271, "step": 7136 }, { "epoch": 8.0, "learning_rate": 2.136338028169014e-05, "loss": 2.8102, "step": 7137 }, { "epoch": 8.0, "learning_rate": 2.1361971830985915e-05, "loss": 2.5996, "step": 7138 }, { "epoch": 8.0, "learning_rate": 2.1360563380281688e-05, "loss": 2.3414, "step": 7139 }, { "epoch": 8.0, "learning_rate": 2.1359154929577468e-05, "loss": 2.1437, "step": 7140 }, { "epoch": 8.01, "learning_rate": 2.135774647887324e-05, "loss": 2.1797, "step": 7141 }, { "epoch": 8.01, "learning_rate": 2.1356338028169017e-05, "loss": 2.1043, "step": 7142 }, { "epoch": 8.01, "learning_rate": 2.135492957746479e-05, "loss": 1.9561, "step": 7143 }, { "epoch": 8.01, "learning_rate": 2.1353521126760563e-05, "loss": 2.1314, "step": 7144 }, { "epoch": 8.01, "learning_rate": 2.135211267605634e-05, "loss": 2.0614, "step": 7145 }, { "epoch": 8.01, "learning_rate": 2.1350704225352112e-05, "loss": 2.304, "step": 7146 }, { "epoch": 8.01, "learning_rate": 2.134929577464789e-05, "loss": 1.9863, "step": 7147 }, { "epoch": 8.01, "learning_rate": 2.134788732394366e-05, "loss": 2.2279, "step": 7148 }, { "epoch": 8.01, "learning_rate": 2.1346478873239438e-05, "loss": 2.3515, "step": 7149 }, { "epoch": 8.02, "learning_rate": 2.134507042253521e-05, "loss": 2.0984, "step": 7150 }, { "epoch": 8.02, "learning_rate": 2.1343661971830984e-05, "loss": 2.1493, "step": 7151 }, { "epoch": 8.02, "learning_rate": 2.134225352112676e-05, "loss": 2.1182, "step": 7152 }, { "epoch": 8.02, "learning_rate": 2.1340845070422533e-05, "loss": 2.0679, "step": 7153 }, { "epoch": 8.02, "learning_rate": 2.133943661971831e-05, "loss": 2.0108, "step": 7154 }, { "epoch": 8.02, "learning_rate": 2.1338028169014086e-05, "loss": 1.9599, "step": 7155 }, { "epoch": 8.02, "learning_rate": 2.1336619718309862e-05, "loss": 2.0038, "step": 7156 }, { "epoch": 8.02, "learning_rate": 2.1335211267605635e-05, "loss": 1.9565, "step": 7157 }, { "epoch": 8.02, "learning_rate": 2.133380281690141e-05, "loss": 2.047, "step": 7158 }, { "epoch": 8.03, "learning_rate": 2.1332394366197184e-05, "loss": 2.066, "step": 7159 }, { "epoch": 8.03, "learning_rate": 2.1330985915492957e-05, "loss": 2.1645, "step": 7160 }, { "epoch": 8.03, "learning_rate": 2.1329577464788734e-05, "loss": 2.0456, "step": 7161 }, { "epoch": 8.03, "learning_rate": 2.1328169014084507e-05, "loss": 1.8568, "step": 7162 }, { "epoch": 8.03, "learning_rate": 2.1326760563380283e-05, "loss": 2.1658, "step": 7163 }, { "epoch": 8.03, "learning_rate": 2.1325352112676056e-05, "loss": 1.8766, "step": 7164 }, { "epoch": 8.03, "learning_rate": 2.1323943661971832e-05, "loss": 2.0728, "step": 7165 }, { "epoch": 8.03, "learning_rate": 2.1322535211267605e-05, "loss": 1.9827, "step": 7166 }, { "epoch": 8.03, "learning_rate": 2.1321126760563378e-05, "loss": 1.8866, "step": 7167 }, { "epoch": 8.04, "learning_rate": 2.1319718309859154e-05, "loss": 1.83, "step": 7168 }, { "epoch": 8.04, "learning_rate": 2.1318309859154927e-05, "loss": 1.7719, "step": 7169 }, { "epoch": 8.04, "learning_rate": 2.1316901408450707e-05, "loss": 2.1412, "step": 7170 }, { "epoch": 8.04, "learning_rate": 2.131549295774648e-05, "loss": 2.0605, "step": 7171 }, { "epoch": 8.04, "learning_rate": 2.1314084507042256e-05, "loss": 2.1876, "step": 7172 }, { "epoch": 8.04, "learning_rate": 2.131267605633803e-05, "loss": 1.6758, "step": 7173 }, { "epoch": 8.04, "learning_rate": 2.1311267605633806e-05, "loss": 2.102, "step": 7174 }, { "epoch": 8.04, "learning_rate": 2.130985915492958e-05, "loss": 2.0745, "step": 7175 }, { "epoch": 8.04, "learning_rate": 2.130845070422535e-05, "loss": 2.0373, "step": 7176 }, { "epoch": 8.05, "learning_rate": 2.1307042253521128e-05, "loss": 2.187, "step": 7177 }, { "epoch": 8.05, "learning_rate": 2.13056338028169e-05, "loss": 1.9651, "step": 7178 }, { "epoch": 8.05, "learning_rate": 2.1304225352112677e-05, "loss": 1.7662, "step": 7179 }, { "epoch": 8.05, "learning_rate": 2.130281690140845e-05, "loss": 1.7478, "step": 7180 }, { "epoch": 8.05, "learning_rate": 2.1301408450704226e-05, "loss": 1.6478, "step": 7181 }, { "epoch": 8.05, "learning_rate": 2.13e-05, "loss": 1.7533, "step": 7182 }, { "epoch": 8.05, "learning_rate": 2.1298591549295772e-05, "loss": 1.8579, "step": 7183 }, { "epoch": 8.05, "learning_rate": 2.129718309859155e-05, "loss": 1.906, "step": 7184 }, { "epoch": 8.05, "learning_rate": 2.1295774647887325e-05, "loss": 1.7722, "step": 7185 }, { "epoch": 8.06, "learning_rate": 2.12943661971831e-05, "loss": 1.5037, "step": 7186 }, { "epoch": 8.06, "learning_rate": 2.1292957746478874e-05, "loss": 2.9472, "step": 7187 }, { "epoch": 8.06, "learning_rate": 2.129154929577465e-05, "loss": 2.6267, "step": 7188 }, { "epoch": 8.06, "learning_rate": 2.1290140845070424e-05, "loss": 2.4199, "step": 7189 }, { "epoch": 8.06, "learning_rate": 2.12887323943662e-05, "loss": 2.4332, "step": 7190 }, { "epoch": 8.06, "learning_rate": 2.1287323943661973e-05, "loss": 2.2884, "step": 7191 }, { "epoch": 8.06, "learning_rate": 2.1285915492957746e-05, "loss": 2.2218, "step": 7192 }, { "epoch": 8.06, "learning_rate": 2.1284507042253522e-05, "loss": 2.0204, "step": 7193 }, { "epoch": 8.07, "learning_rate": 2.1283098591549295e-05, "loss": 1.9816, "step": 7194 }, { "epoch": 8.07, "learning_rate": 2.128169014084507e-05, "loss": 2.1416, "step": 7195 }, { "epoch": 8.07, "learning_rate": 2.1280281690140844e-05, "loss": 2.2048, "step": 7196 }, { "epoch": 8.07, "learning_rate": 2.127887323943662e-05, "loss": 2.1064, "step": 7197 }, { "epoch": 8.07, "learning_rate": 2.1277464788732394e-05, "loss": 2.16, "step": 7198 }, { "epoch": 8.07, "learning_rate": 2.1276056338028166e-05, "loss": 1.8924, "step": 7199 }, { "epoch": 8.07, "learning_rate": 2.1274647887323946e-05, "loss": 1.9068, "step": 7200 }, { "epoch": 8.07, "learning_rate": 2.127323943661972e-05, "loss": 2.001, "step": 7201 }, { "epoch": 8.07, "learning_rate": 2.1271830985915495e-05, "loss": 2.1334, "step": 7202 }, { "epoch": 8.08, "learning_rate": 2.127042253521127e-05, "loss": 1.8844, "step": 7203 }, { "epoch": 8.08, "learning_rate": 2.1269014084507045e-05, "loss": 2.0085, "step": 7204 }, { "epoch": 8.08, "learning_rate": 2.1267605633802818e-05, "loss": 2.0837, "step": 7205 }, { "epoch": 8.08, "learning_rate": 2.1266197183098594e-05, "loss": 2.2414, "step": 7206 }, { "epoch": 8.08, "learning_rate": 2.1264788732394367e-05, "loss": 1.9049, "step": 7207 }, { "epoch": 8.08, "learning_rate": 2.126338028169014e-05, "loss": 1.9791, "step": 7208 }, { "epoch": 8.08, "learning_rate": 2.1261971830985916e-05, "loss": 1.7805, "step": 7209 }, { "epoch": 8.08, "learning_rate": 2.126056338028169e-05, "loss": 1.6996, "step": 7210 }, { "epoch": 8.08, "learning_rate": 2.1259154929577466e-05, "loss": 1.8253, "step": 7211 }, { "epoch": 8.09, "learning_rate": 2.125774647887324e-05, "loss": 1.8423, "step": 7212 }, { "epoch": 8.09, "learning_rate": 2.1256338028169015e-05, "loss": 1.7998, "step": 7213 }, { "epoch": 8.09, "learning_rate": 2.1254929577464788e-05, "loss": 1.8959, "step": 7214 }, { "epoch": 8.09, "learning_rate": 2.1253521126760564e-05, "loss": 2.0239, "step": 7215 }, { "epoch": 8.09, "learning_rate": 2.125211267605634e-05, "loss": 2.0859, "step": 7216 }, { "epoch": 8.09, "learning_rate": 2.1250704225352113e-05, "loss": 1.9545, "step": 7217 }, { "epoch": 8.09, "learning_rate": 2.124929577464789e-05, "loss": 2.0681, "step": 7218 }, { "epoch": 8.09, "learning_rate": 2.1247887323943663e-05, "loss": 1.8306, "step": 7219 }, { "epoch": 8.09, "learning_rate": 2.124647887323944e-05, "loss": 1.8851, "step": 7220 }, { "epoch": 8.1, "learning_rate": 2.1245070422535212e-05, "loss": 2.057, "step": 7221 }, { "epoch": 8.1, "learning_rate": 2.1243661971830988e-05, "loss": 1.9464, "step": 7222 }, { "epoch": 8.1, "learning_rate": 2.124225352112676e-05, "loss": 1.9286, "step": 7223 }, { "epoch": 8.1, "learning_rate": 2.1240845070422534e-05, "loss": 1.8938, "step": 7224 }, { "epoch": 8.1, "learning_rate": 2.123943661971831e-05, "loss": 1.8136, "step": 7225 }, { "epoch": 8.1, "learning_rate": 2.1238028169014083e-05, "loss": 1.9273, "step": 7226 }, { "epoch": 8.1, "learning_rate": 2.123661971830986e-05, "loss": 2.0101, "step": 7227 }, { "epoch": 8.1, "learning_rate": 2.1235211267605633e-05, "loss": 1.5795, "step": 7228 }, { "epoch": 8.1, "learning_rate": 2.123380281690141e-05, "loss": 2.2039, "step": 7229 }, { "epoch": 8.11, "learning_rate": 2.1232394366197182e-05, "loss": 1.8505, "step": 7230 }, { "epoch": 8.11, "learning_rate": 2.1230985915492958e-05, "loss": 2.1185, "step": 7231 }, { "epoch": 8.11, "learning_rate": 2.1229577464788735e-05, "loss": 1.9268, "step": 7232 }, { "epoch": 8.11, "learning_rate": 2.1228169014084508e-05, "loss": 1.6359, "step": 7233 }, { "epoch": 8.11, "learning_rate": 2.1226760563380284e-05, "loss": 1.6069, "step": 7234 }, { "epoch": 8.11, "learning_rate": 2.1225352112676057e-05, "loss": 1.9815, "step": 7235 }, { "epoch": 8.11, "learning_rate": 2.1223943661971833e-05, "loss": 1.5955, "step": 7236 }, { "epoch": 8.11, "learning_rate": 2.1222535211267606e-05, "loss": 2.7307, "step": 7237 }, { "epoch": 8.11, "learning_rate": 2.122112676056338e-05, "loss": 2.4326, "step": 7238 }, { "epoch": 8.12, "learning_rate": 2.1219718309859155e-05, "loss": 2.4616, "step": 7239 }, { "epoch": 8.12, "learning_rate": 2.121830985915493e-05, "loss": 2.3018, "step": 7240 }, { "epoch": 8.12, "learning_rate": 2.1216901408450705e-05, "loss": 2.3127, "step": 7241 }, { "epoch": 8.12, "learning_rate": 2.1215492957746478e-05, "loss": 1.9474, "step": 7242 }, { "epoch": 8.12, "learning_rate": 2.1214084507042254e-05, "loss": 2.3581, "step": 7243 }, { "epoch": 8.12, "learning_rate": 2.1212676056338027e-05, "loss": 1.9239, "step": 7244 }, { "epoch": 8.12, "learning_rate": 2.1211267605633803e-05, "loss": 2.1708, "step": 7245 }, { "epoch": 8.12, "learning_rate": 2.120985915492958e-05, "loss": 1.9293, "step": 7246 }, { "epoch": 8.12, "learning_rate": 2.1208450704225352e-05, "loss": 1.9128, "step": 7247 }, { "epoch": 8.13, "learning_rate": 2.120704225352113e-05, "loss": 1.925, "step": 7248 }, { "epoch": 8.13, "learning_rate": 2.1205633802816902e-05, "loss": 2.2127, "step": 7249 }, { "epoch": 8.13, "learning_rate": 2.1204225352112678e-05, "loss": 1.9701, "step": 7250 }, { "epoch": 8.13, "learning_rate": 2.120281690140845e-05, "loss": 1.8836, "step": 7251 }, { "epoch": 8.13, "learning_rate": 2.1201408450704227e-05, "loss": 2.1537, "step": 7252 }, { "epoch": 8.13, "learning_rate": 2.12e-05, "loss": 2.0553, "step": 7253 }, { "epoch": 8.13, "learning_rate": 2.1198591549295773e-05, "loss": 1.8706, "step": 7254 }, { "epoch": 8.13, "learning_rate": 2.119718309859155e-05, "loss": 2.2258, "step": 7255 }, { "epoch": 8.13, "learning_rate": 2.1195774647887323e-05, "loss": 1.7954, "step": 7256 }, { "epoch": 8.14, "learning_rate": 2.11943661971831e-05, "loss": 1.9594, "step": 7257 }, { "epoch": 8.14, "learning_rate": 2.1192957746478872e-05, "loss": 2.2601, "step": 7258 }, { "epoch": 8.14, "learning_rate": 2.1191549295774648e-05, "loss": 1.9537, "step": 7259 }, { "epoch": 8.14, "learning_rate": 2.119014084507042e-05, "loss": 2.0978, "step": 7260 }, { "epoch": 8.14, "learning_rate": 2.11887323943662e-05, "loss": 2.1465, "step": 7261 }, { "epoch": 8.14, "learning_rate": 2.1187323943661974e-05, "loss": 1.9403, "step": 7262 }, { "epoch": 8.14, "learning_rate": 2.1185915492957747e-05, "loss": 1.9563, "step": 7263 }, { "epoch": 8.14, "learning_rate": 2.1184507042253523e-05, "loss": 1.7461, "step": 7264 }, { "epoch": 8.14, "learning_rate": 2.1183098591549296e-05, "loss": 1.9424, "step": 7265 }, { "epoch": 8.15, "learning_rate": 2.1181690140845072e-05, "loss": 1.7057, "step": 7266 }, { "epoch": 8.15, "learning_rate": 2.1180281690140845e-05, "loss": 1.8533, "step": 7267 }, { "epoch": 8.15, "learning_rate": 2.117887323943662e-05, "loss": 2.2203, "step": 7268 }, { "epoch": 8.15, "learning_rate": 2.1177464788732395e-05, "loss": 1.9713, "step": 7269 }, { "epoch": 8.15, "learning_rate": 2.1176056338028167e-05, "loss": 1.9144, "step": 7270 }, { "epoch": 8.15, "learning_rate": 2.1174647887323944e-05, "loss": 2.0655, "step": 7271 }, { "epoch": 8.15, "learning_rate": 2.1173239436619717e-05, "loss": 2.0333, "step": 7272 }, { "epoch": 8.15, "learning_rate": 2.1171830985915493e-05, "loss": 2.0884, "step": 7273 }, { "epoch": 8.15, "learning_rate": 2.1170422535211266e-05, "loss": 1.9221, "step": 7274 }, { "epoch": 8.16, "learning_rate": 2.1169014084507042e-05, "loss": 1.8455, "step": 7275 }, { "epoch": 8.16, "learning_rate": 2.116760563380282e-05, "loss": 1.9071, "step": 7276 }, { "epoch": 8.16, "learning_rate": 2.1166197183098595e-05, "loss": 2.0471, "step": 7277 }, { "epoch": 8.16, "learning_rate": 2.1164788732394368e-05, "loss": 1.6779, "step": 7278 }, { "epoch": 8.16, "learning_rate": 2.116338028169014e-05, "loss": 1.572, "step": 7279 }, { "epoch": 8.16, "learning_rate": 2.1161971830985917e-05, "loss": 1.8724, "step": 7280 }, { "epoch": 8.16, "learning_rate": 2.116056338028169e-05, "loss": 1.6555, "step": 7281 }, { "epoch": 8.16, "learning_rate": 2.1159154929577466e-05, "loss": 1.7221, "step": 7282 }, { "epoch": 8.16, "learning_rate": 2.115774647887324e-05, "loss": 1.5564, "step": 7283 }, { "epoch": 8.17, "learning_rate": 2.1156338028169016e-05, "loss": 1.4241, "step": 7284 }, { "epoch": 8.17, "learning_rate": 2.115492957746479e-05, "loss": 1.7544, "step": 7285 }, { "epoch": 8.17, "learning_rate": 2.115352112676056e-05, "loss": 2.0158, "step": 7286 }, { "epoch": 8.17, "learning_rate": 2.1152112676056338e-05, "loss": 2.8005, "step": 7287 }, { "epoch": 8.17, "learning_rate": 2.115070422535211e-05, "loss": 2.8936, "step": 7288 }, { "epoch": 8.17, "learning_rate": 2.1149295774647887e-05, "loss": 2.705, "step": 7289 }, { "epoch": 8.17, "learning_rate": 2.114788732394366e-05, "loss": 2.494, "step": 7290 }, { "epoch": 8.17, "learning_rate": 2.114647887323944e-05, "loss": 2.108, "step": 7291 }, { "epoch": 8.17, "learning_rate": 2.1145070422535213e-05, "loss": 2.2079, "step": 7292 }, { "epoch": 8.18, "learning_rate": 2.114366197183099e-05, "loss": 2.1838, "step": 7293 }, { "epoch": 8.18, "learning_rate": 2.1142253521126762e-05, "loss": 1.9835, "step": 7294 }, { "epoch": 8.18, "learning_rate": 2.1140845070422535e-05, "loss": 2.1601, "step": 7295 }, { "epoch": 8.18, "learning_rate": 2.113943661971831e-05, "loss": 2.0364, "step": 7296 }, { "epoch": 8.18, "learning_rate": 2.1138028169014084e-05, "loss": 2.1864, "step": 7297 }, { "epoch": 8.18, "learning_rate": 2.113661971830986e-05, "loss": 2.1104, "step": 7298 }, { "epoch": 8.18, "learning_rate": 2.1135211267605634e-05, "loss": 2.112, "step": 7299 }, { "epoch": 8.18, "learning_rate": 2.113380281690141e-05, "loss": 2.0935, "step": 7300 }, { "epoch": 8.18, "learning_rate": 2.1132394366197183e-05, "loss": 1.953, "step": 7301 }, { "epoch": 8.19, "learning_rate": 2.1130985915492956e-05, "loss": 1.7853, "step": 7302 }, { "epoch": 8.19, "learning_rate": 2.1129577464788732e-05, "loss": 1.9684, "step": 7303 }, { "epoch": 8.19, "learning_rate": 2.1128169014084505e-05, "loss": 1.994, "step": 7304 }, { "epoch": 8.19, "learning_rate": 2.112676056338028e-05, "loss": 2.0937, "step": 7305 }, { "epoch": 8.19, "learning_rate": 2.1125352112676058e-05, "loss": 2.0463, "step": 7306 }, { "epoch": 8.19, "learning_rate": 2.1123943661971834e-05, "loss": 1.8279, "step": 7307 }, { "epoch": 8.19, "learning_rate": 2.1122535211267607e-05, "loss": 1.719, "step": 7308 }, { "epoch": 8.19, "learning_rate": 2.1121126760563383e-05, "loss": 1.8004, "step": 7309 }, { "epoch": 8.2, "learning_rate": 2.1119718309859156e-05, "loss": 1.7372, "step": 7310 }, { "epoch": 8.2, "learning_rate": 2.111830985915493e-05, "loss": 1.8424, "step": 7311 }, { "epoch": 8.2, "learning_rate": 2.1116901408450706e-05, "loss": 1.9893, "step": 7312 }, { "epoch": 8.2, "learning_rate": 2.111549295774648e-05, "loss": 1.8422, "step": 7313 }, { "epoch": 8.2, "learning_rate": 2.1114084507042255e-05, "loss": 1.8005, "step": 7314 }, { "epoch": 8.2, "learning_rate": 2.1112676056338028e-05, "loss": 1.7752, "step": 7315 }, { "epoch": 8.2, "learning_rate": 2.1111267605633804e-05, "loss": 1.7664, "step": 7316 }, { "epoch": 8.2, "learning_rate": 2.1109859154929577e-05, "loss": 1.8878, "step": 7317 }, { "epoch": 8.2, "learning_rate": 2.110845070422535e-05, "loss": 2.1051, "step": 7318 }, { "epoch": 8.21, "learning_rate": 2.1107042253521126e-05, "loss": 1.8377, "step": 7319 }, { "epoch": 8.21, "learning_rate": 2.11056338028169e-05, "loss": 1.7916, "step": 7320 }, { "epoch": 8.21, "learning_rate": 2.110422535211268e-05, "loss": 1.8793, "step": 7321 }, { "epoch": 8.21, "learning_rate": 2.1102816901408452e-05, "loss": 1.7964, "step": 7322 }, { "epoch": 8.21, "learning_rate": 2.110140845070423e-05, "loss": 1.8281, "step": 7323 }, { "epoch": 8.21, "learning_rate": 2.11e-05, "loss": 1.8836, "step": 7324 }, { "epoch": 8.21, "learning_rate": 2.1098591549295778e-05, "loss": 2.1253, "step": 7325 }, { "epoch": 8.21, "learning_rate": 2.109718309859155e-05, "loss": 1.9532, "step": 7326 }, { "epoch": 8.21, "learning_rate": 2.1095774647887323e-05, "loss": 1.9923, "step": 7327 }, { "epoch": 8.22, "learning_rate": 2.10943661971831e-05, "loss": 1.8652, "step": 7328 }, { "epoch": 8.22, "learning_rate": 2.1092957746478873e-05, "loss": 1.6862, "step": 7329 }, { "epoch": 8.22, "learning_rate": 2.109154929577465e-05, "loss": 1.8304, "step": 7330 }, { "epoch": 8.22, "learning_rate": 2.1090140845070422e-05, "loss": 1.8424, "step": 7331 }, { "epoch": 8.22, "learning_rate": 2.10887323943662e-05, "loss": 1.973, "step": 7332 }, { "epoch": 8.22, "learning_rate": 2.108732394366197e-05, "loss": 1.6613, "step": 7333 }, { "epoch": 8.22, "learning_rate": 2.1085915492957744e-05, "loss": 1.6527, "step": 7334 }, { "epoch": 8.22, "learning_rate": 2.108450704225352e-05, "loss": 1.6903, "step": 7335 }, { "epoch": 8.22, "learning_rate": 2.1083098591549294e-05, "loss": 1.3148, "step": 7336 }, { "epoch": 8.23, "learning_rate": 2.1081690140845073e-05, "loss": 2.4713, "step": 7337 }, { "epoch": 8.23, "learning_rate": 2.1080281690140846e-05, "loss": 2.4947, "step": 7338 }, { "epoch": 8.23, "learning_rate": 2.1078873239436623e-05, "loss": 2.3591, "step": 7339 }, { "epoch": 8.23, "learning_rate": 2.1077464788732395e-05, "loss": 2.4778, "step": 7340 }, { "epoch": 8.23, "learning_rate": 2.1076056338028172e-05, "loss": 2.2519, "step": 7341 }, { "epoch": 8.23, "learning_rate": 2.1074647887323945e-05, "loss": 2.0793, "step": 7342 }, { "epoch": 8.23, "learning_rate": 2.1073239436619718e-05, "loss": 1.8848, "step": 7343 }, { "epoch": 8.23, "learning_rate": 2.1071830985915494e-05, "loss": 2.2435, "step": 7344 }, { "epoch": 8.23, "learning_rate": 2.1070422535211267e-05, "loss": 2.1409, "step": 7345 }, { "epoch": 8.24, "learning_rate": 2.1069014084507043e-05, "loss": 2.1032, "step": 7346 }, { "epoch": 8.24, "learning_rate": 2.1067605633802816e-05, "loss": 2.2802, "step": 7347 }, { "epoch": 8.24, "learning_rate": 2.1066197183098593e-05, "loss": 1.9786, "step": 7348 }, { "epoch": 8.24, "learning_rate": 2.1064788732394366e-05, "loss": 2.0941, "step": 7349 }, { "epoch": 8.24, "learning_rate": 2.106338028169014e-05, "loss": 1.9752, "step": 7350 }, { "epoch": 8.24, "learning_rate": 2.1061971830985915e-05, "loss": 1.8911, "step": 7351 }, { "epoch": 8.24, "learning_rate": 2.106056338028169e-05, "loss": 1.9772, "step": 7352 }, { "epoch": 8.24, "learning_rate": 2.1059154929577467e-05, "loss": 2.1512, "step": 7353 }, { "epoch": 8.24, "learning_rate": 2.105774647887324e-05, "loss": 2.0881, "step": 7354 }, { "epoch": 8.25, "learning_rate": 2.1056338028169017e-05, "loss": 1.8676, "step": 7355 }, { "epoch": 8.25, "learning_rate": 2.105492957746479e-05, "loss": 1.838, "step": 7356 }, { "epoch": 8.25, "learning_rate": 2.1053521126760563e-05, "loss": 1.8149, "step": 7357 }, { "epoch": 8.25, "learning_rate": 2.105211267605634e-05, "loss": 1.7392, "step": 7358 }, { "epoch": 8.25, "learning_rate": 2.1050704225352112e-05, "loss": 1.8418, "step": 7359 }, { "epoch": 8.25, "learning_rate": 2.1049295774647888e-05, "loss": 1.8154, "step": 7360 }, { "epoch": 8.25, "learning_rate": 2.104788732394366e-05, "loss": 1.9142, "step": 7361 }, { "epoch": 8.25, "learning_rate": 2.1046478873239437e-05, "loss": 1.8819, "step": 7362 }, { "epoch": 8.25, "learning_rate": 2.104507042253521e-05, "loss": 2.0331, "step": 7363 }, { "epoch": 8.26, "learning_rate": 2.1043661971830987e-05, "loss": 1.9045, "step": 7364 }, { "epoch": 8.26, "learning_rate": 2.104225352112676e-05, "loss": 1.8417, "step": 7365 }, { "epoch": 8.26, "learning_rate": 2.1040845070422533e-05, "loss": 2.1181, "step": 7366 }, { "epoch": 8.26, "learning_rate": 2.1039436619718312e-05, "loss": 1.7921, "step": 7367 }, { "epoch": 8.26, "learning_rate": 2.1038028169014085e-05, "loss": 1.5868, "step": 7368 }, { "epoch": 8.26, "learning_rate": 2.103661971830986e-05, "loss": 2.0659, "step": 7369 }, { "epoch": 8.26, "learning_rate": 2.1035211267605635e-05, "loss": 1.9764, "step": 7370 }, { "epoch": 8.26, "learning_rate": 2.103380281690141e-05, "loss": 1.9277, "step": 7371 }, { "epoch": 8.26, "learning_rate": 2.1032394366197184e-05, "loss": 1.6653, "step": 7372 }, { "epoch": 8.27, "learning_rate": 2.1030985915492957e-05, "loss": 2.0857, "step": 7373 }, { "epoch": 8.27, "learning_rate": 2.1029577464788733e-05, "loss": 1.8774, "step": 7374 }, { "epoch": 8.27, "learning_rate": 2.1028169014084506e-05, "loss": 1.7004, "step": 7375 }, { "epoch": 8.27, "learning_rate": 2.1026760563380282e-05, "loss": 1.8174, "step": 7376 }, { "epoch": 8.27, "learning_rate": 2.1025352112676055e-05, "loss": 1.732, "step": 7377 }, { "epoch": 8.27, "learning_rate": 2.1023943661971832e-05, "loss": 1.9784, "step": 7378 }, { "epoch": 8.27, "learning_rate": 2.1022535211267605e-05, "loss": 1.6796, "step": 7379 }, { "epoch": 8.27, "learning_rate": 2.102112676056338e-05, "loss": 1.7388, "step": 7380 }, { "epoch": 8.27, "learning_rate": 2.1019718309859154e-05, "loss": 1.6382, "step": 7381 }, { "epoch": 8.28, "learning_rate": 2.101830985915493e-05, "loss": 1.7631, "step": 7382 }, { "epoch": 8.28, "learning_rate": 2.1016901408450707e-05, "loss": 1.6923, "step": 7383 }, { "epoch": 8.28, "learning_rate": 2.101549295774648e-05, "loss": 1.5405, "step": 7384 }, { "epoch": 8.28, "learning_rate": 2.1014084507042256e-05, "loss": 1.6625, "step": 7385 }, { "epoch": 8.28, "learning_rate": 2.101267605633803e-05, "loss": 1.4779, "step": 7386 }, { "epoch": 8.28, "learning_rate": 2.1011267605633805e-05, "loss": 2.8655, "step": 7387 }, { "epoch": 8.28, "learning_rate": 2.1009859154929578e-05, "loss": 2.7766, "step": 7388 }, { "epoch": 8.28, "learning_rate": 2.100845070422535e-05, "loss": 2.5701, "step": 7389 }, { "epoch": 8.28, "learning_rate": 2.1007042253521127e-05, "loss": 2.0969, "step": 7390 }, { "epoch": 8.29, "learning_rate": 2.10056338028169e-05, "loss": 2.2585, "step": 7391 }, { "epoch": 8.29, "learning_rate": 2.1004225352112677e-05, "loss": 2.1319, "step": 7392 }, { "epoch": 8.29, "learning_rate": 2.100281690140845e-05, "loss": 1.8194, "step": 7393 }, { "epoch": 8.29, "learning_rate": 2.1001408450704226e-05, "loss": 2.1053, "step": 7394 }, { "epoch": 8.29, "learning_rate": 2.1e-05, "loss": 1.8215, "step": 7395 }, { "epoch": 8.29, "learning_rate": 2.0998591549295775e-05, "loss": 2.0458, "step": 7396 }, { "epoch": 8.29, "learning_rate": 2.099718309859155e-05, "loss": 1.955, "step": 7397 }, { "epoch": 8.29, "learning_rate": 2.0995774647887324e-05, "loss": 1.8577, "step": 7398 }, { "epoch": 8.29, "learning_rate": 2.09943661971831e-05, "loss": 2.1365, "step": 7399 }, { "epoch": 8.3, "learning_rate": 2.0992957746478874e-05, "loss": 1.8866, "step": 7400 }, { "epoch": 8.3, "learning_rate": 2.099154929577465e-05, "loss": 1.8142, "step": 7401 }, { "epoch": 8.3, "learning_rate": 2.0990140845070423e-05, "loss": 1.975, "step": 7402 }, { "epoch": 8.3, "learning_rate": 2.09887323943662e-05, "loss": 2.0441, "step": 7403 }, { "epoch": 8.3, "learning_rate": 2.0987323943661972e-05, "loss": 1.9419, "step": 7404 }, { "epoch": 8.3, "learning_rate": 2.0985915492957745e-05, "loss": 1.8416, "step": 7405 }, { "epoch": 8.3, "learning_rate": 2.098450704225352e-05, "loss": 1.7742, "step": 7406 }, { "epoch": 8.3, "learning_rate": 2.0983098591549294e-05, "loss": 1.7949, "step": 7407 }, { "epoch": 8.3, "learning_rate": 2.098169014084507e-05, "loss": 1.6894, "step": 7408 }, { "epoch": 8.31, "learning_rate": 2.0980281690140844e-05, "loss": 1.8483, "step": 7409 }, { "epoch": 8.31, "learning_rate": 2.097887323943662e-05, "loss": 1.8854, "step": 7410 }, { "epoch": 8.31, "learning_rate": 2.0977464788732393e-05, "loss": 1.7413, "step": 7411 }, { "epoch": 8.31, "learning_rate": 2.0976056338028173e-05, "loss": 2.1174, "step": 7412 }, { "epoch": 8.31, "learning_rate": 2.0974647887323946e-05, "loss": 1.8334, "step": 7413 }, { "epoch": 8.31, "learning_rate": 2.097323943661972e-05, "loss": 1.9621, "step": 7414 }, { "epoch": 8.31, "learning_rate": 2.0971830985915495e-05, "loss": 1.8239, "step": 7415 }, { "epoch": 8.31, "learning_rate": 2.0970422535211268e-05, "loss": 1.6192, "step": 7416 }, { "epoch": 8.32, "learning_rate": 2.0969014084507044e-05, "loss": 1.9582, "step": 7417 }, { "epoch": 8.32, "learning_rate": 2.0967605633802817e-05, "loss": 2.0277, "step": 7418 }, { "epoch": 8.32, "learning_rate": 2.0966197183098594e-05, "loss": 1.8184, "step": 7419 }, { "epoch": 8.32, "learning_rate": 2.0964788732394366e-05, "loss": 1.8367, "step": 7420 }, { "epoch": 8.32, "learning_rate": 2.096338028169014e-05, "loss": 1.8982, "step": 7421 }, { "epoch": 8.32, "learning_rate": 2.0961971830985916e-05, "loss": 1.9251, "step": 7422 }, { "epoch": 8.32, "learning_rate": 2.096056338028169e-05, "loss": 1.9389, "step": 7423 }, { "epoch": 8.32, "learning_rate": 2.0959154929577465e-05, "loss": 1.9683, "step": 7424 }, { "epoch": 8.32, "learning_rate": 2.0957746478873238e-05, "loss": 1.7609, "step": 7425 }, { "epoch": 8.33, "learning_rate": 2.0956338028169014e-05, "loss": 1.7444, "step": 7426 }, { "epoch": 8.33, "learning_rate": 2.095492957746479e-05, "loss": 1.6056, "step": 7427 }, { "epoch": 8.33, "learning_rate": 2.0953521126760567e-05, "loss": 2.0176, "step": 7428 }, { "epoch": 8.33, "learning_rate": 2.095211267605634e-05, "loss": 1.7466, "step": 7429 }, { "epoch": 8.33, "learning_rate": 2.0950704225352113e-05, "loss": 1.5989, "step": 7430 }, { "epoch": 8.33, "learning_rate": 2.094929577464789e-05, "loss": 1.6303, "step": 7431 }, { "epoch": 8.33, "learning_rate": 2.0947887323943662e-05, "loss": 1.5208, "step": 7432 }, { "epoch": 8.33, "learning_rate": 2.094647887323944e-05, "loss": 1.5765, "step": 7433 }, { "epoch": 8.33, "learning_rate": 2.094507042253521e-05, "loss": 1.951, "step": 7434 }, { "epoch": 8.34, "learning_rate": 2.0943661971830988e-05, "loss": 1.8676, "step": 7435 }, { "epoch": 8.34, "learning_rate": 2.094225352112676e-05, "loss": 1.6178, "step": 7436 }, { "epoch": 8.34, "learning_rate": 2.0940845070422534e-05, "loss": 2.589, "step": 7437 }, { "epoch": 8.34, "learning_rate": 2.093943661971831e-05, "loss": 2.8676, "step": 7438 }, { "epoch": 8.34, "learning_rate": 2.0938028169014083e-05, "loss": 2.3211, "step": 7439 }, { "epoch": 8.34, "learning_rate": 2.093661971830986e-05, "loss": 2.1282, "step": 7440 }, { "epoch": 8.34, "learning_rate": 2.0935211267605632e-05, "loss": 2.0899, "step": 7441 }, { "epoch": 8.34, "learning_rate": 2.0933802816901412e-05, "loss": 2.1358, "step": 7442 }, { "epoch": 8.34, "learning_rate": 2.0932394366197185e-05, "loss": 1.9908, "step": 7443 }, { "epoch": 8.35, "learning_rate": 2.093098591549296e-05, "loss": 1.7472, "step": 7444 }, { "epoch": 8.35, "learning_rate": 2.0929577464788734e-05, "loss": 2.0396, "step": 7445 }, { "epoch": 8.35, "learning_rate": 2.0928169014084507e-05, "loss": 1.9351, "step": 7446 }, { "epoch": 8.35, "learning_rate": 2.0926760563380283e-05, "loss": 2.0196, "step": 7447 }, { "epoch": 8.35, "learning_rate": 2.0925352112676056e-05, "loss": 1.9568, "step": 7448 }, { "epoch": 8.35, "learning_rate": 2.0923943661971833e-05, "loss": 1.9355, "step": 7449 }, { "epoch": 8.35, "learning_rate": 2.0922535211267606e-05, "loss": 1.8463, "step": 7450 }, { "epoch": 8.35, "learning_rate": 2.0921126760563382e-05, "loss": 1.8377, "step": 7451 }, { "epoch": 8.35, "learning_rate": 2.0919718309859155e-05, "loss": 2.055, "step": 7452 }, { "epoch": 8.36, "learning_rate": 2.0918309859154928e-05, "loss": 2.1258, "step": 7453 }, { "epoch": 8.36, "learning_rate": 2.0916901408450704e-05, "loss": 1.6486, "step": 7454 }, { "epoch": 8.36, "learning_rate": 2.0915492957746477e-05, "loss": 2.0961, "step": 7455 }, { "epoch": 8.36, "learning_rate": 2.0914084507042253e-05, "loss": 2.3201, "step": 7456 }, { "epoch": 8.36, "learning_rate": 2.0912676056338026e-05, "loss": 1.6484, "step": 7457 }, { "epoch": 8.36, "learning_rate": 2.0911267605633806e-05, "loss": 1.9481, "step": 7458 }, { "epoch": 8.36, "learning_rate": 2.090985915492958e-05, "loss": 1.6262, "step": 7459 }, { "epoch": 8.36, "learning_rate": 2.0908450704225352e-05, "loss": 1.7178, "step": 7460 }, { "epoch": 8.36, "learning_rate": 2.0907042253521128e-05, "loss": 1.7686, "step": 7461 }, { "epoch": 8.37, "learning_rate": 2.09056338028169e-05, "loss": 1.8389, "step": 7462 }, { "epoch": 8.37, "learning_rate": 2.0904225352112678e-05, "loss": 1.8172, "step": 7463 }, { "epoch": 8.37, "learning_rate": 2.090281690140845e-05, "loss": 1.9575, "step": 7464 }, { "epoch": 8.37, "learning_rate": 2.0901408450704227e-05, "loss": 1.7822, "step": 7465 }, { "epoch": 8.37, "learning_rate": 2.09e-05, "loss": 1.8224, "step": 7466 }, { "epoch": 8.37, "learning_rate": 2.0898591549295776e-05, "loss": 1.6286, "step": 7467 }, { "epoch": 8.37, "learning_rate": 2.089718309859155e-05, "loss": 1.7047, "step": 7468 }, { "epoch": 8.37, "learning_rate": 2.0895774647887322e-05, "loss": 2.1318, "step": 7469 }, { "epoch": 8.37, "learning_rate": 2.08943661971831e-05, "loss": 2.0886, "step": 7470 }, { "epoch": 8.38, "learning_rate": 2.089295774647887e-05, "loss": 1.5126, "step": 7471 }, { "epoch": 8.38, "learning_rate": 2.0891549295774648e-05, "loss": 1.933, "step": 7472 }, { "epoch": 8.38, "learning_rate": 2.0890140845070424e-05, "loss": 1.9979, "step": 7473 }, { "epoch": 8.38, "learning_rate": 2.08887323943662e-05, "loss": 1.7657, "step": 7474 }, { "epoch": 8.38, "learning_rate": 2.0887323943661973e-05, "loss": 1.7029, "step": 7475 }, { "epoch": 8.38, "learning_rate": 2.0885915492957746e-05, "loss": 1.8104, "step": 7476 }, { "epoch": 8.38, "learning_rate": 2.0884507042253523e-05, "loss": 1.9523, "step": 7477 }, { "epoch": 8.38, "learning_rate": 2.0883098591549295e-05, "loss": 1.9344, "step": 7478 }, { "epoch": 8.38, "learning_rate": 2.0881690140845072e-05, "loss": 1.4665, "step": 7479 }, { "epoch": 8.39, "learning_rate": 2.0880281690140845e-05, "loss": 1.8033, "step": 7480 }, { "epoch": 8.39, "learning_rate": 2.087887323943662e-05, "loss": 1.578, "step": 7481 }, { "epoch": 8.39, "learning_rate": 2.0877464788732394e-05, "loss": 1.7684, "step": 7482 }, { "epoch": 8.39, "learning_rate": 2.087605633802817e-05, "loss": 1.7508, "step": 7483 }, { "epoch": 8.39, "learning_rate": 2.0874647887323943e-05, "loss": 1.7391, "step": 7484 }, { "epoch": 8.39, "learning_rate": 2.0873239436619716e-05, "loss": 1.8315, "step": 7485 }, { "epoch": 8.39, "learning_rate": 2.0871830985915493e-05, "loss": 1.5988, "step": 7486 }, { "epoch": 8.39, "learning_rate": 2.0870422535211265e-05, "loss": 2.6162, "step": 7487 }, { "epoch": 8.39, "learning_rate": 2.0869014084507045e-05, "loss": 2.3508, "step": 7488 }, { "epoch": 8.4, "learning_rate": 2.0867605633802818e-05, "loss": 2.2789, "step": 7489 }, { "epoch": 8.4, "learning_rate": 2.0866197183098594e-05, "loss": 2.0158, "step": 7490 }, { "epoch": 8.4, "learning_rate": 2.0864788732394367e-05, "loss": 2.0931, "step": 7491 }, { "epoch": 8.4, "learning_rate": 2.086338028169014e-05, "loss": 1.8738, "step": 7492 }, { "epoch": 8.4, "learning_rate": 2.0861971830985917e-05, "loss": 1.8817, "step": 7493 }, { "epoch": 8.4, "learning_rate": 2.086056338028169e-05, "loss": 2.1243, "step": 7494 }, { "epoch": 8.4, "learning_rate": 2.0859154929577466e-05, "loss": 2.2406, "step": 7495 }, { "epoch": 8.4, "learning_rate": 2.085774647887324e-05, "loss": 2.1076, "step": 7496 }, { "epoch": 8.4, "learning_rate": 2.0856338028169015e-05, "loss": 1.9095, "step": 7497 }, { "epoch": 8.41, "learning_rate": 2.0854929577464788e-05, "loss": 2.0107, "step": 7498 }, { "epoch": 8.41, "learning_rate": 2.0853521126760565e-05, "loss": 1.9532, "step": 7499 }, { "epoch": 8.41, "learning_rate": 2.0852112676056337e-05, "loss": 1.88, "step": 7500 }, { "epoch": 8.41, "eval_loss": 1.9886143207550049, "eval_runtime": 837.3432, "eval_samples_per_second": 3.155, "eval_steps_per_second": 0.395, "eval_wer": 0.3438119793732646, "step": 7500 }, { "epoch": 8.41, "learning_rate": 2.085070422535211e-05, "loss": 1.6842, "step": 7501 }, { "epoch": 8.41, "learning_rate": 2.0849295774647887e-05, "loss": 1.7961, "step": 7502 }, { "epoch": 8.41, "learning_rate": 2.0847887323943663e-05, "loss": 1.9229, "step": 7503 }, { "epoch": 8.41, "learning_rate": 2.084647887323944e-05, "loss": 1.8836, "step": 7504 }, { "epoch": 8.41, "learning_rate": 2.0845070422535212e-05, "loss": 1.5749, "step": 7505 }, { "epoch": 8.41, "learning_rate": 2.084366197183099e-05, "loss": 1.8908, "step": 7506 }, { "epoch": 8.42, "learning_rate": 2.084225352112676e-05, "loss": 1.8384, "step": 7507 }, { "epoch": 8.42, "learning_rate": 2.0840845070422535e-05, "loss": 1.9776, "step": 7508 }, { "epoch": 8.42, "learning_rate": 2.083943661971831e-05, "loss": 1.8293, "step": 7509 }, { "epoch": 8.42, "learning_rate": 2.0838028169014084e-05, "loss": 1.9215, "step": 7510 }, { "epoch": 8.42, "learning_rate": 2.083661971830986e-05, "loss": 1.924, "step": 7511 }, { "epoch": 8.42, "learning_rate": 2.0835211267605633e-05, "loss": 1.832, "step": 7512 }, { "epoch": 8.42, "learning_rate": 2.083380281690141e-05, "loss": 1.896, "step": 7513 }, { "epoch": 8.42, "learning_rate": 2.0832394366197182e-05, "loss": 1.838, "step": 7514 }, { "epoch": 8.42, "learning_rate": 2.083098591549296e-05, "loss": 1.5689, "step": 7515 }, { "epoch": 8.43, "learning_rate": 2.082957746478873e-05, "loss": 1.9124, "step": 7516 }, { "epoch": 8.43, "learning_rate": 2.0828169014084505e-05, "loss": 1.5403, "step": 7517 }, { "epoch": 8.43, "learning_rate": 2.0826760563380284e-05, "loss": 1.7816, "step": 7518 }, { "epoch": 8.43, "learning_rate": 2.0825352112676057e-05, "loss": 1.6958, "step": 7519 }, { "epoch": 8.43, "learning_rate": 2.0823943661971834e-05, "loss": 1.7361, "step": 7520 }, { "epoch": 8.43, "learning_rate": 2.0822535211267607e-05, "loss": 1.7934, "step": 7521 }, { "epoch": 8.43, "learning_rate": 2.0821126760563383e-05, "loss": 1.7162, "step": 7522 }, { "epoch": 8.43, "learning_rate": 2.0819718309859156e-05, "loss": 1.6997, "step": 7523 }, { "epoch": 8.43, "learning_rate": 2.081830985915493e-05, "loss": 1.7608, "step": 7524 }, { "epoch": 8.44, "learning_rate": 2.0816901408450705e-05, "loss": 1.7515, "step": 7525 }, { "epoch": 8.44, "learning_rate": 2.0815492957746478e-05, "loss": 1.6938, "step": 7526 }, { "epoch": 8.44, "learning_rate": 2.0814084507042254e-05, "loss": 1.6319, "step": 7527 }, { "epoch": 8.44, "learning_rate": 2.0812676056338027e-05, "loss": 1.9774, "step": 7528 }, { "epoch": 8.44, "learning_rate": 2.0811267605633804e-05, "loss": 2.0282, "step": 7529 }, { "epoch": 8.44, "learning_rate": 2.0809859154929577e-05, "loss": 1.5287, "step": 7530 }, { "epoch": 8.44, "learning_rate": 2.0808450704225353e-05, "loss": 1.614, "step": 7531 }, { "epoch": 8.44, "learning_rate": 2.0807042253521126e-05, "loss": 1.5358, "step": 7532 }, { "epoch": 8.45, "learning_rate": 2.0805633802816902e-05, "loss": 1.5575, "step": 7533 }, { "epoch": 8.45, "learning_rate": 2.080422535211268e-05, "loss": 1.7989, "step": 7534 }, { "epoch": 8.45, "learning_rate": 2.080281690140845e-05, "loss": 1.5582, "step": 7535 }, { "epoch": 8.45, "learning_rate": 2.0801408450704228e-05, "loss": 1.6089, "step": 7536 }, { "epoch": 8.45, "learning_rate": 2.08e-05, "loss": 2.6379, "step": 7537 }, { "epoch": 8.45, "learning_rate": 2.0798591549295777e-05, "loss": 2.3949, "step": 7538 }, { "epoch": 8.45, "learning_rate": 2.079718309859155e-05, "loss": 2.1352, "step": 7539 }, { "epoch": 8.45, "learning_rate": 2.0795774647887323e-05, "loss": 2.1484, "step": 7540 }, { "epoch": 8.45, "learning_rate": 2.07943661971831e-05, "loss": 1.8534, "step": 7541 }, { "epoch": 8.46, "learning_rate": 2.0792957746478872e-05, "loss": 1.725, "step": 7542 }, { "epoch": 8.46, "learning_rate": 2.079154929577465e-05, "loss": 2.0553, "step": 7543 }, { "epoch": 8.46, "learning_rate": 2.079014084507042e-05, "loss": 2.0524, "step": 7544 }, { "epoch": 8.46, "learning_rate": 2.0788732394366198e-05, "loss": 1.909, "step": 7545 }, { "epoch": 8.46, "learning_rate": 2.078732394366197e-05, "loss": 2.0669, "step": 7546 }, { "epoch": 8.46, "learning_rate": 2.0785915492957747e-05, "loss": 1.9484, "step": 7547 }, { "epoch": 8.46, "learning_rate": 2.0784507042253523e-05, "loss": 1.8101, "step": 7548 }, { "epoch": 8.46, "learning_rate": 2.0783098591549296e-05, "loss": 1.8026, "step": 7549 }, { "epoch": 8.46, "learning_rate": 2.0781690140845073e-05, "loss": 2.0029, "step": 7550 }, { "epoch": 8.47, "learning_rate": 2.0780281690140846e-05, "loss": 1.772, "step": 7551 }, { "epoch": 8.47, "learning_rate": 2.0778873239436622e-05, "loss": 1.9534, "step": 7552 }, { "epoch": 8.47, "learning_rate": 2.0777464788732395e-05, "loss": 1.7715, "step": 7553 }, { "epoch": 8.47, "learning_rate": 2.077605633802817e-05, "loss": 1.8481, "step": 7554 }, { "epoch": 8.47, "learning_rate": 2.0774647887323944e-05, "loss": 1.6654, "step": 7555 }, { "epoch": 8.47, "learning_rate": 2.0773239436619717e-05, "loss": 1.7345, "step": 7556 }, { "epoch": 8.47, "learning_rate": 2.0771830985915494e-05, "loss": 1.7319, "step": 7557 }, { "epoch": 8.47, "learning_rate": 2.0770422535211266e-05, "loss": 1.8048, "step": 7558 }, { "epoch": 8.47, "learning_rate": 2.0769014084507043e-05, "loss": 1.8101, "step": 7559 }, { "epoch": 8.48, "learning_rate": 2.0767605633802816e-05, "loss": 1.5874, "step": 7560 }, { "epoch": 8.48, "learning_rate": 2.0766197183098592e-05, "loss": 1.7951, "step": 7561 }, { "epoch": 8.48, "learning_rate": 2.0764788732394365e-05, "loss": 1.7527, "step": 7562 }, { "epoch": 8.48, "learning_rate": 2.076338028169014e-05, "loss": 1.7797, "step": 7563 }, { "epoch": 8.48, "learning_rate": 2.0761971830985918e-05, "loss": 2.0164, "step": 7564 }, { "epoch": 8.48, "learning_rate": 2.076056338028169e-05, "loss": 1.7824, "step": 7565 }, { "epoch": 8.48, "learning_rate": 2.0759154929577467e-05, "loss": 1.6179, "step": 7566 }, { "epoch": 8.48, "learning_rate": 2.075774647887324e-05, "loss": 1.873, "step": 7567 }, { "epoch": 8.48, "learning_rate": 2.0756338028169016e-05, "loss": 1.7364, "step": 7568 }, { "epoch": 8.49, "learning_rate": 2.075492957746479e-05, "loss": 1.5907, "step": 7569 }, { "epoch": 8.49, "learning_rate": 2.0753521126760565e-05, "loss": 1.7535, "step": 7570 }, { "epoch": 8.49, "learning_rate": 2.075211267605634e-05, "loss": 1.6631, "step": 7571 }, { "epoch": 8.49, "learning_rate": 2.075070422535211e-05, "loss": 1.5341, "step": 7572 }, { "epoch": 8.49, "learning_rate": 2.0749295774647888e-05, "loss": 1.841, "step": 7573 }, { "epoch": 8.49, "learning_rate": 2.074788732394366e-05, "loss": 1.764, "step": 7574 }, { "epoch": 8.49, "learning_rate": 2.0746478873239437e-05, "loss": 1.7715, "step": 7575 }, { "epoch": 8.49, "learning_rate": 2.074507042253521e-05, "loss": 1.8971, "step": 7576 }, { "epoch": 8.49, "learning_rate": 2.0743661971830986e-05, "loss": 1.9871, "step": 7577 }, { "epoch": 8.5, "learning_rate": 2.074225352112676e-05, "loss": 1.5237, "step": 7578 }, { "epoch": 8.5, "learning_rate": 2.0740845070422536e-05, "loss": 1.77, "step": 7579 }, { "epoch": 8.5, "learning_rate": 2.0739436619718312e-05, "loss": 1.4337, "step": 7580 }, { "epoch": 8.5, "learning_rate": 2.0738028169014085e-05, "loss": 1.5791, "step": 7581 }, { "epoch": 8.5, "learning_rate": 2.073661971830986e-05, "loss": 1.621, "step": 7582 }, { "epoch": 8.5, "learning_rate": 2.0735211267605634e-05, "loss": 1.5585, "step": 7583 }, { "epoch": 8.5, "learning_rate": 2.073380281690141e-05, "loss": 1.3604, "step": 7584 }, { "epoch": 8.5, "learning_rate": 2.0732394366197183e-05, "loss": 1.5833, "step": 7585 }, { "epoch": 8.5, "learning_rate": 2.073098591549296e-05, "loss": 1.4253, "step": 7586 }, { "epoch": 8.51, "learning_rate": 2.0729577464788733e-05, "loss": 2.1551, "step": 7587 }, { "epoch": 8.51, "learning_rate": 2.0728169014084506e-05, "loss": 2.207, "step": 7588 }, { "epoch": 8.51, "learning_rate": 2.0726760563380282e-05, "loss": 2.1768, "step": 7589 }, { "epoch": 8.51, "learning_rate": 2.0725352112676055e-05, "loss": 2.0875, "step": 7590 }, { "epoch": 8.51, "learning_rate": 2.072394366197183e-05, "loss": 1.7876, "step": 7591 }, { "epoch": 8.51, "learning_rate": 2.0722535211267604e-05, "loss": 2.0517, "step": 7592 }, { "epoch": 8.51, "learning_rate": 2.072112676056338e-05, "loss": 2.1793, "step": 7593 }, { "epoch": 8.51, "learning_rate": 2.0719718309859157e-05, "loss": 1.7866, "step": 7594 }, { "epoch": 8.51, "learning_rate": 2.071830985915493e-05, "loss": 1.8839, "step": 7595 }, { "epoch": 8.52, "learning_rate": 2.0716901408450706e-05, "loss": 1.8407, "step": 7596 }, { "epoch": 8.52, "learning_rate": 2.071549295774648e-05, "loss": 2.0046, "step": 7597 }, { "epoch": 8.52, "learning_rate": 2.0714084507042255e-05, "loss": 1.7703, "step": 7598 }, { "epoch": 8.52, "learning_rate": 2.0712676056338028e-05, "loss": 2.0538, "step": 7599 }, { "epoch": 8.52, "learning_rate": 2.0711267605633805e-05, "loss": 1.8063, "step": 7600 }, { "epoch": 8.52, "learning_rate": 2.0709859154929578e-05, "loss": 1.8416, "step": 7601 }, { "epoch": 8.52, "learning_rate": 2.0708450704225354e-05, "loss": 1.7829, "step": 7602 }, { "epoch": 8.52, "learning_rate": 2.0707042253521127e-05, "loss": 1.701, "step": 7603 }, { "epoch": 8.52, "learning_rate": 2.07056338028169e-05, "loss": 1.6986, "step": 7604 }, { "epoch": 8.53, "learning_rate": 2.0704225352112676e-05, "loss": 1.6766, "step": 7605 }, { "epoch": 8.53, "learning_rate": 2.070281690140845e-05, "loss": 1.8597, "step": 7606 }, { "epoch": 8.53, "learning_rate": 2.0701408450704225e-05, "loss": 1.7575, "step": 7607 }, { "epoch": 8.53, "learning_rate": 2.07e-05, "loss": 1.7975, "step": 7608 }, { "epoch": 8.53, "learning_rate": 2.0698591549295778e-05, "loss": 1.6532, "step": 7609 }, { "epoch": 8.53, "learning_rate": 2.069718309859155e-05, "loss": 1.7056, "step": 7610 }, { "epoch": 8.53, "learning_rate": 2.0695774647887324e-05, "loss": 1.7396, "step": 7611 }, { "epoch": 8.53, "learning_rate": 2.06943661971831e-05, "loss": 1.616, "step": 7612 }, { "epoch": 8.53, "learning_rate": 2.0692957746478873e-05, "loss": 1.8009, "step": 7613 }, { "epoch": 8.54, "learning_rate": 2.069154929577465e-05, "loss": 1.7586, "step": 7614 }, { "epoch": 8.54, "learning_rate": 2.0690140845070422e-05, "loss": 1.6187, "step": 7615 }, { "epoch": 8.54, "learning_rate": 2.06887323943662e-05, "loss": 1.9234, "step": 7616 }, { "epoch": 8.54, "learning_rate": 2.0687323943661972e-05, "loss": 1.7045, "step": 7617 }, { "epoch": 8.54, "learning_rate": 2.0685915492957748e-05, "loss": 1.8264, "step": 7618 }, { "epoch": 8.54, "learning_rate": 2.068450704225352e-05, "loss": 1.8043, "step": 7619 }, { "epoch": 8.54, "learning_rate": 2.0683098591549294e-05, "loss": 1.384, "step": 7620 }, { "epoch": 8.54, "learning_rate": 2.068169014084507e-05, "loss": 1.7906, "step": 7621 }, { "epoch": 8.54, "learning_rate": 2.0680281690140843e-05, "loss": 1.6527, "step": 7622 }, { "epoch": 8.55, "learning_rate": 2.067887323943662e-05, "loss": 1.7947, "step": 7623 }, { "epoch": 8.55, "learning_rate": 2.0677464788732396e-05, "loss": 1.6699, "step": 7624 }, { "epoch": 8.55, "learning_rate": 2.0676056338028172e-05, "loss": 1.7901, "step": 7625 }, { "epoch": 8.55, "learning_rate": 2.0674647887323945e-05, "loss": 1.5742, "step": 7626 }, { "epoch": 8.55, "learning_rate": 2.0673239436619718e-05, "loss": 1.4426, "step": 7627 }, { "epoch": 8.55, "learning_rate": 2.0671830985915494e-05, "loss": 1.5573, "step": 7628 }, { "epoch": 8.55, "learning_rate": 2.0670422535211267e-05, "loss": 1.8698, "step": 7629 }, { "epoch": 8.55, "learning_rate": 2.0669014084507044e-05, "loss": 1.713, "step": 7630 }, { "epoch": 8.55, "learning_rate": 2.0667605633802817e-05, "loss": 1.6217, "step": 7631 }, { "epoch": 8.56, "learning_rate": 2.0666197183098593e-05, "loss": 1.4388, "step": 7632 }, { "epoch": 8.56, "learning_rate": 2.0664788732394366e-05, "loss": 1.4036, "step": 7633 }, { "epoch": 8.56, "learning_rate": 2.0663380281690142e-05, "loss": 1.8446, "step": 7634 }, { "epoch": 8.56, "learning_rate": 2.0661971830985915e-05, "loss": 1.5498, "step": 7635 }, { "epoch": 8.56, "learning_rate": 2.0660563380281688e-05, "loss": 1.9195, "step": 7636 }, { "epoch": 8.56, "learning_rate": 2.0659154929577465e-05, "loss": 2.4821, "step": 7637 }, { "epoch": 8.56, "learning_rate": 2.0657746478873237e-05, "loss": 2.258, "step": 7638 }, { "epoch": 8.56, "learning_rate": 2.0656338028169017e-05, "loss": 1.988, "step": 7639 }, { "epoch": 8.57, "learning_rate": 2.065492957746479e-05, "loss": 1.9177, "step": 7640 }, { "epoch": 8.57, "learning_rate": 2.0653521126760566e-05, "loss": 1.8456, "step": 7641 }, { "epoch": 8.57, "learning_rate": 2.065211267605634e-05, "loss": 1.8957, "step": 7642 }, { "epoch": 8.57, "learning_rate": 2.0650704225352112e-05, "loss": 1.8466, "step": 7643 }, { "epoch": 8.57, "learning_rate": 2.064929577464789e-05, "loss": 1.934, "step": 7644 }, { "epoch": 8.57, "learning_rate": 2.064788732394366e-05, "loss": 1.821, "step": 7645 }, { "epoch": 8.57, "learning_rate": 2.0646478873239438e-05, "loss": 1.7192, "step": 7646 }, { "epoch": 8.57, "learning_rate": 2.064507042253521e-05, "loss": 1.7791, "step": 7647 }, { "epoch": 8.57, "learning_rate": 2.0643661971830987e-05, "loss": 1.7864, "step": 7648 }, { "epoch": 8.58, "learning_rate": 2.064225352112676e-05, "loss": 1.8572, "step": 7649 }, { "epoch": 8.58, "learning_rate": 2.0640845070422536e-05, "loss": 1.704, "step": 7650 }, { "epoch": 8.58, "learning_rate": 2.063943661971831e-05, "loss": 1.8532, "step": 7651 }, { "epoch": 8.58, "learning_rate": 2.0638028169014082e-05, "loss": 1.8071, "step": 7652 }, { "epoch": 8.58, "learning_rate": 2.063661971830986e-05, "loss": 1.9035, "step": 7653 }, { "epoch": 8.58, "learning_rate": 2.0635211267605635e-05, "loss": 1.8111, "step": 7654 }, { "epoch": 8.58, "learning_rate": 2.063380281690141e-05, "loss": 1.799, "step": 7655 }, { "epoch": 8.58, "learning_rate": 2.0632394366197184e-05, "loss": 1.6538, "step": 7656 }, { "epoch": 8.58, "learning_rate": 2.063098591549296e-05, "loss": 1.7782, "step": 7657 }, { "epoch": 8.59, "learning_rate": 2.0629577464788734e-05, "loss": 1.6875, "step": 7658 }, { "epoch": 8.59, "learning_rate": 2.0628169014084507e-05, "loss": 1.8838, "step": 7659 }, { "epoch": 8.59, "learning_rate": 2.0626760563380283e-05, "loss": 1.77, "step": 7660 }, { "epoch": 8.59, "learning_rate": 2.0625352112676056e-05, "loss": 1.5577, "step": 7661 }, { "epoch": 8.59, "learning_rate": 2.0623943661971832e-05, "loss": 1.7983, "step": 7662 }, { "epoch": 8.59, "learning_rate": 2.0622535211267605e-05, "loss": 1.5671, "step": 7663 }, { "epoch": 8.59, "learning_rate": 2.062112676056338e-05, "loss": 1.6268, "step": 7664 }, { "epoch": 8.59, "learning_rate": 2.0619718309859154e-05, "loss": 1.9556, "step": 7665 }, { "epoch": 8.59, "learning_rate": 2.061830985915493e-05, "loss": 1.6635, "step": 7666 }, { "epoch": 8.6, "learning_rate": 2.0616901408450704e-05, "loss": 1.935, "step": 7667 }, { "epoch": 8.6, "learning_rate": 2.0615492957746477e-05, "loss": 1.764, "step": 7668 }, { "epoch": 8.6, "learning_rate": 2.0614084507042256e-05, "loss": 1.7115, "step": 7669 }, { "epoch": 8.6, "learning_rate": 2.061267605633803e-05, "loss": 1.5737, "step": 7670 }, { "epoch": 8.6, "learning_rate": 2.0611267605633806e-05, "loss": 1.5992, "step": 7671 }, { "epoch": 8.6, "learning_rate": 2.060985915492958e-05, "loss": 1.8929, "step": 7672 }, { "epoch": 8.6, "learning_rate": 2.0608450704225355e-05, "loss": 1.6677, "step": 7673 }, { "epoch": 8.6, "learning_rate": 2.0607042253521128e-05, "loss": 1.646, "step": 7674 }, { "epoch": 8.6, "learning_rate": 2.06056338028169e-05, "loss": 1.5384, "step": 7675 }, { "epoch": 8.61, "learning_rate": 2.0604225352112677e-05, "loss": 1.7386, "step": 7676 }, { "epoch": 8.61, "learning_rate": 2.060281690140845e-05, "loss": 1.5987, "step": 7677 }, { "epoch": 8.61, "learning_rate": 2.0601408450704226e-05, "loss": 1.7348, "step": 7678 }, { "epoch": 8.61, "learning_rate": 2.06e-05, "loss": 1.4725, "step": 7679 }, { "epoch": 8.61, "learning_rate": 2.0598591549295776e-05, "loss": 1.603, "step": 7680 }, { "epoch": 8.61, "learning_rate": 2.059718309859155e-05, "loss": 1.8268, "step": 7681 }, { "epoch": 8.61, "learning_rate": 2.0595774647887325e-05, "loss": 2.0707, "step": 7682 }, { "epoch": 8.61, "learning_rate": 2.0594366197183098e-05, "loss": 1.0376, "step": 7683 }, { "epoch": 8.61, "learning_rate": 2.0592957746478874e-05, "loss": 1.7691, "step": 7684 }, { "epoch": 8.62, "learning_rate": 2.059154929577465e-05, "loss": 1.3712, "step": 7685 }, { "epoch": 8.62, "learning_rate": 2.0590140845070423e-05, "loss": 1.6298, "step": 7686 }, { "epoch": 8.62, "learning_rate": 2.05887323943662e-05, "loss": 2.3246, "step": 7687 }, { "epoch": 8.62, "learning_rate": 2.0587323943661973e-05, "loss": 2.2471, "step": 7688 }, { "epoch": 8.62, "learning_rate": 2.058591549295775e-05, "loss": 2.2541, "step": 7689 }, { "epoch": 8.62, "learning_rate": 2.0584507042253522e-05, "loss": 2.1749, "step": 7690 }, { "epoch": 8.62, "learning_rate": 2.0583098591549295e-05, "loss": 1.915, "step": 7691 }, { "epoch": 8.62, "learning_rate": 2.058169014084507e-05, "loss": 1.8474, "step": 7692 }, { "epoch": 8.62, "learning_rate": 2.0580281690140844e-05, "loss": 1.8776, "step": 7693 }, { "epoch": 8.63, "learning_rate": 2.057887323943662e-05, "loss": 1.7921, "step": 7694 }, { "epoch": 8.63, "learning_rate": 2.0577464788732393e-05, "loss": 1.7134, "step": 7695 }, { "epoch": 8.63, "learning_rate": 2.057605633802817e-05, "loss": 1.8949, "step": 7696 }, { "epoch": 8.63, "learning_rate": 2.0574647887323943e-05, "loss": 1.8138, "step": 7697 }, { "epoch": 8.63, "learning_rate": 2.057323943661972e-05, "loss": 1.9366, "step": 7698 }, { "epoch": 8.63, "learning_rate": 2.0571830985915492e-05, "loss": 1.8912, "step": 7699 }, { "epoch": 8.63, "learning_rate": 2.057042253521127e-05, "loss": 1.603, "step": 7700 }, { "epoch": 8.63, "learning_rate": 2.0569014084507045e-05, "loss": 2.0101, "step": 7701 }, { "epoch": 8.63, "learning_rate": 2.0567605633802818e-05, "loss": 1.7905, "step": 7702 }, { "epoch": 8.64, "learning_rate": 2.0566197183098594e-05, "loss": 1.6905, "step": 7703 }, { "epoch": 8.64, "learning_rate": 2.0564788732394367e-05, "loss": 1.6832, "step": 7704 }, { "epoch": 8.64, "learning_rate": 2.0563380281690143e-05, "loss": 1.6842, "step": 7705 }, { "epoch": 8.64, "learning_rate": 2.0561971830985916e-05, "loss": 2.1065, "step": 7706 }, { "epoch": 8.64, "learning_rate": 2.056056338028169e-05, "loss": 1.7446, "step": 7707 }, { "epoch": 8.64, "learning_rate": 2.0559154929577465e-05, "loss": 1.66, "step": 7708 }, { "epoch": 8.64, "learning_rate": 2.055774647887324e-05, "loss": 1.7168, "step": 7709 }, { "epoch": 8.64, "learning_rate": 2.0556338028169015e-05, "loss": 1.7996, "step": 7710 }, { "epoch": 8.64, "learning_rate": 2.0554929577464788e-05, "loss": 1.7485, "step": 7711 }, { "epoch": 8.65, "learning_rate": 2.0553521126760564e-05, "loss": 1.6265, "step": 7712 }, { "epoch": 8.65, "learning_rate": 2.0552112676056337e-05, "loss": 1.8978, "step": 7713 }, { "epoch": 8.65, "learning_rate": 2.055070422535211e-05, "loss": 1.6997, "step": 7714 }, { "epoch": 8.65, "learning_rate": 2.054929577464789e-05, "loss": 1.7787, "step": 7715 }, { "epoch": 8.65, "learning_rate": 2.0547887323943663e-05, "loss": 1.5592, "step": 7716 }, { "epoch": 8.65, "learning_rate": 2.054647887323944e-05, "loss": 1.985, "step": 7717 }, { "epoch": 8.65, "learning_rate": 2.0545070422535212e-05, "loss": 1.591, "step": 7718 }, { "epoch": 8.65, "learning_rate": 2.0543661971830988e-05, "loss": 1.8019, "step": 7719 }, { "epoch": 8.65, "learning_rate": 2.054225352112676e-05, "loss": 1.5954, "step": 7720 }, { "epoch": 8.66, "learning_rate": 2.0540845070422537e-05, "loss": 1.6788, "step": 7721 }, { "epoch": 8.66, "learning_rate": 2.053943661971831e-05, "loss": 1.7371, "step": 7722 }, { "epoch": 8.66, "learning_rate": 2.0538028169014083e-05, "loss": 1.4326, "step": 7723 }, { "epoch": 8.66, "learning_rate": 2.053661971830986e-05, "loss": 1.5951, "step": 7724 }, { "epoch": 8.66, "learning_rate": 2.0535211267605633e-05, "loss": 1.5789, "step": 7725 }, { "epoch": 8.66, "learning_rate": 2.053380281690141e-05, "loss": 1.6984, "step": 7726 }, { "epoch": 8.66, "learning_rate": 2.0532394366197182e-05, "loss": 1.7526, "step": 7727 }, { "epoch": 8.66, "learning_rate": 2.0530985915492958e-05, "loss": 1.7677, "step": 7728 }, { "epoch": 8.66, "learning_rate": 2.052957746478873e-05, "loss": 1.6765, "step": 7729 }, { "epoch": 8.67, "learning_rate": 2.0528169014084507e-05, "loss": 1.8274, "step": 7730 }, { "epoch": 8.67, "learning_rate": 2.0526760563380284e-05, "loss": 1.5012, "step": 7731 }, { "epoch": 8.67, "learning_rate": 2.0525352112676057e-05, "loss": 1.8068, "step": 7732 }, { "epoch": 8.67, "learning_rate": 2.0523943661971833e-05, "loss": 1.3473, "step": 7733 }, { "epoch": 8.67, "learning_rate": 2.0522535211267606e-05, "loss": 1.8208, "step": 7734 }, { "epoch": 8.67, "learning_rate": 2.0521126760563382e-05, "loss": 1.4143, "step": 7735 }, { "epoch": 8.67, "learning_rate": 2.0519718309859155e-05, "loss": 1.5128, "step": 7736 }, { "epoch": 8.67, "learning_rate": 2.051830985915493e-05, "loss": 2.6032, "step": 7737 }, { "epoch": 8.67, "learning_rate": 2.0516901408450705e-05, "loss": 2.3196, "step": 7738 }, { "epoch": 8.68, "learning_rate": 2.0515492957746478e-05, "loss": 2.1741, "step": 7739 }, { "epoch": 8.68, "learning_rate": 2.0514084507042254e-05, "loss": 1.7849, "step": 7740 }, { "epoch": 8.68, "learning_rate": 2.0512676056338027e-05, "loss": 1.8152, "step": 7741 }, { "epoch": 8.68, "learning_rate": 2.0511267605633803e-05, "loss": 1.8539, "step": 7742 }, { "epoch": 8.68, "learning_rate": 2.0509859154929576e-05, "loss": 1.5551, "step": 7743 }, { "epoch": 8.68, "learning_rate": 2.0508450704225352e-05, "loss": 1.7757, "step": 7744 }, { "epoch": 8.68, "learning_rate": 2.050704225352113e-05, "loss": 1.855, "step": 7745 }, { "epoch": 8.68, "learning_rate": 2.05056338028169e-05, "loss": 1.843, "step": 7746 }, { "epoch": 8.68, "learning_rate": 2.0504225352112678e-05, "loss": 1.8541, "step": 7747 }, { "epoch": 8.69, "learning_rate": 2.050281690140845e-05, "loss": 1.7614, "step": 7748 }, { "epoch": 8.69, "learning_rate": 2.0501408450704227e-05, "loss": 1.6421, "step": 7749 }, { "epoch": 8.69, "learning_rate": 2.05e-05, "loss": 1.6263, "step": 7750 }, { "epoch": 8.69, "learning_rate": 2.0498591549295777e-05, "loss": 1.8007, "step": 7751 }, { "epoch": 8.69, "learning_rate": 2.049718309859155e-05, "loss": 1.7921, "step": 7752 }, { "epoch": 8.69, "learning_rate": 2.0495774647887326e-05, "loss": 1.6322, "step": 7753 }, { "epoch": 8.69, "learning_rate": 2.04943661971831e-05, "loss": 1.8266, "step": 7754 }, { "epoch": 8.69, "learning_rate": 2.0492957746478872e-05, "loss": 2.0622, "step": 7755 }, { "epoch": 8.7, "learning_rate": 2.0491549295774648e-05, "loss": 1.6843, "step": 7756 }, { "epoch": 8.7, "learning_rate": 2.049014084507042e-05, "loss": 1.8809, "step": 7757 }, { "epoch": 8.7, "learning_rate": 2.0488732394366197e-05, "loss": 2.0105, "step": 7758 }, { "epoch": 8.7, "learning_rate": 2.048732394366197e-05, "loss": 1.6597, "step": 7759 }, { "epoch": 8.7, "learning_rate": 2.048591549295775e-05, "loss": 1.6693, "step": 7760 }, { "epoch": 8.7, "learning_rate": 2.0484507042253523e-05, "loss": 1.8415, "step": 7761 }, { "epoch": 8.7, "learning_rate": 2.0483098591549296e-05, "loss": 1.8764, "step": 7762 }, { "epoch": 8.7, "learning_rate": 2.0481690140845072e-05, "loss": 1.6868, "step": 7763 }, { "epoch": 8.7, "learning_rate": 2.0480281690140845e-05, "loss": 1.7714, "step": 7764 }, { "epoch": 8.71, "learning_rate": 2.047887323943662e-05, "loss": 1.666, "step": 7765 }, { "epoch": 8.71, "learning_rate": 2.0477464788732394e-05, "loss": 1.7384, "step": 7766 }, { "epoch": 8.71, "learning_rate": 2.047605633802817e-05, "loss": 1.8563, "step": 7767 }, { "epoch": 8.71, "learning_rate": 2.0474647887323944e-05, "loss": 1.9796, "step": 7768 }, { "epoch": 8.71, "learning_rate": 2.047323943661972e-05, "loss": 1.4123, "step": 7769 }, { "epoch": 8.71, "learning_rate": 2.0471830985915493e-05, "loss": 1.6899, "step": 7770 }, { "epoch": 8.71, "learning_rate": 2.0470422535211266e-05, "loss": 1.9991, "step": 7771 }, { "epoch": 8.71, "learning_rate": 2.0469014084507042e-05, "loss": 1.6897, "step": 7772 }, { "epoch": 8.71, "learning_rate": 2.0467605633802815e-05, "loss": 1.8484, "step": 7773 }, { "epoch": 8.72, "learning_rate": 2.046619718309859e-05, "loss": 1.6226, "step": 7774 }, { "epoch": 8.72, "learning_rate": 2.0464788732394368e-05, "loss": 1.6073, "step": 7775 }, { "epoch": 8.72, "learning_rate": 2.0463380281690144e-05, "loss": 1.7905, "step": 7776 }, { "epoch": 8.72, "learning_rate": 2.0461971830985917e-05, "loss": 1.678, "step": 7777 }, { "epoch": 8.72, "learning_rate": 2.046056338028169e-05, "loss": 1.7289, "step": 7778 }, { "epoch": 8.72, "learning_rate": 2.0459154929577466e-05, "loss": 1.5548, "step": 7779 }, { "epoch": 8.72, "learning_rate": 2.045774647887324e-05, "loss": 1.572, "step": 7780 }, { "epoch": 8.72, "learning_rate": 2.0456338028169016e-05, "loss": 1.8214, "step": 7781 }, { "epoch": 8.72, "learning_rate": 2.045492957746479e-05, "loss": 1.4736, "step": 7782 }, { "epoch": 8.73, "learning_rate": 2.0453521126760565e-05, "loss": 1.7331, "step": 7783 }, { "epoch": 8.73, "learning_rate": 2.0452112676056338e-05, "loss": 1.3974, "step": 7784 }, { "epoch": 8.73, "learning_rate": 2.0450704225352114e-05, "loss": 1.5041, "step": 7785 }, { "epoch": 8.73, "learning_rate": 2.0449295774647887e-05, "loss": 1.5681, "step": 7786 }, { "epoch": 8.73, "learning_rate": 2.044788732394366e-05, "loss": 2.6206, "step": 7787 }, { "epoch": 8.73, "learning_rate": 2.0446478873239436e-05, "loss": 2.0717, "step": 7788 }, { "epoch": 8.73, "learning_rate": 2.044507042253521e-05, "loss": 2.082, "step": 7789 }, { "epoch": 8.73, "learning_rate": 2.044366197183099e-05, "loss": 1.8976, "step": 7790 }, { "epoch": 8.73, "learning_rate": 2.0442253521126762e-05, "loss": 1.8776, "step": 7791 }, { "epoch": 8.74, "learning_rate": 2.044084507042254e-05, "loss": 1.8991, "step": 7792 }, { "epoch": 8.74, "learning_rate": 2.043943661971831e-05, "loss": 1.9888, "step": 7793 }, { "epoch": 8.74, "learning_rate": 2.0438028169014084e-05, "loss": 1.8375, "step": 7794 }, { "epoch": 8.74, "learning_rate": 2.043661971830986e-05, "loss": 1.9093, "step": 7795 }, { "epoch": 8.74, "learning_rate": 2.0435211267605634e-05, "loss": 1.5852, "step": 7796 }, { "epoch": 8.74, "learning_rate": 2.043380281690141e-05, "loss": 1.5819, "step": 7797 }, { "epoch": 8.74, "learning_rate": 2.0432394366197183e-05, "loss": 1.7277, "step": 7798 }, { "epoch": 8.74, "learning_rate": 2.043098591549296e-05, "loss": 1.6807, "step": 7799 }, { "epoch": 8.74, "learning_rate": 2.0429577464788732e-05, "loss": 1.7585, "step": 7800 }, { "epoch": 8.75, "learning_rate": 2.042816901408451e-05, "loss": 1.6848, "step": 7801 }, { "epoch": 8.75, "learning_rate": 2.042676056338028e-05, "loss": 1.7436, "step": 7802 }, { "epoch": 8.75, "learning_rate": 2.0425352112676054e-05, "loss": 1.6873, "step": 7803 }, { "epoch": 8.75, "learning_rate": 2.042394366197183e-05, "loss": 1.7607, "step": 7804 }, { "epoch": 8.75, "learning_rate": 2.0422535211267607e-05, "loss": 1.8954, "step": 7805 }, { "epoch": 8.75, "learning_rate": 2.0421126760563383e-05, "loss": 1.6554, "step": 7806 }, { "epoch": 8.75, "learning_rate": 2.0419718309859156e-05, "loss": 1.6177, "step": 7807 }, { "epoch": 8.75, "learning_rate": 2.0418309859154933e-05, "loss": 1.7425, "step": 7808 }, { "epoch": 8.75, "learning_rate": 2.0416901408450706e-05, "loss": 1.5457, "step": 7809 }, { "epoch": 8.76, "learning_rate": 2.041549295774648e-05, "loss": 1.7024, "step": 7810 }, { "epoch": 8.76, "learning_rate": 2.0414084507042255e-05, "loss": 1.8964, "step": 7811 }, { "epoch": 8.76, "learning_rate": 2.0412676056338028e-05, "loss": 1.8131, "step": 7812 }, { "epoch": 8.76, "learning_rate": 2.0411267605633804e-05, "loss": 1.518, "step": 7813 }, { "epoch": 8.76, "learning_rate": 2.0409859154929577e-05, "loss": 1.8124, "step": 7814 }, { "epoch": 8.76, "learning_rate": 2.0408450704225353e-05, "loss": 1.7564, "step": 7815 }, { "epoch": 8.76, "learning_rate": 2.0407042253521126e-05, "loss": 1.5952, "step": 7816 }, { "epoch": 8.76, "learning_rate": 2.04056338028169e-05, "loss": 1.9531, "step": 7817 }, { "epoch": 8.76, "learning_rate": 2.0404225352112676e-05, "loss": 1.7223, "step": 7818 }, { "epoch": 8.77, "learning_rate": 2.040281690140845e-05, "loss": 1.6531, "step": 7819 }, { "epoch": 8.77, "learning_rate": 2.0401408450704225e-05, "loss": 1.8408, "step": 7820 }, { "epoch": 8.77, "learning_rate": 2.04e-05, "loss": 1.7655, "step": 7821 }, { "epoch": 8.77, "learning_rate": 2.0398591549295778e-05, "loss": 1.5562, "step": 7822 }, { "epoch": 8.77, "learning_rate": 2.039718309859155e-05, "loss": 1.7487, "step": 7823 }, { "epoch": 8.77, "learning_rate": 2.0395774647887327e-05, "loss": 1.5692, "step": 7824 }, { "epoch": 8.77, "learning_rate": 2.03943661971831e-05, "loss": 1.6987, "step": 7825 }, { "epoch": 8.77, "learning_rate": 2.0392957746478873e-05, "loss": 1.6039, "step": 7826 }, { "epoch": 8.77, "learning_rate": 2.039154929577465e-05, "loss": 1.5126, "step": 7827 }, { "epoch": 8.78, "learning_rate": 2.0390140845070422e-05, "loss": 1.7544, "step": 7828 }, { "epoch": 8.78, "learning_rate": 2.0388732394366198e-05, "loss": 1.5675, "step": 7829 }, { "epoch": 8.78, "learning_rate": 2.038732394366197e-05, "loss": 1.7088, "step": 7830 }, { "epoch": 8.78, "learning_rate": 2.0385915492957748e-05, "loss": 1.4613, "step": 7831 }, { "epoch": 8.78, "learning_rate": 2.038450704225352e-05, "loss": 1.6754, "step": 7832 }, { "epoch": 8.78, "learning_rate": 2.0383098591549293e-05, "loss": 1.6857, "step": 7833 }, { "epoch": 8.78, "learning_rate": 2.038169014084507e-05, "loss": 1.4946, "step": 7834 }, { "epoch": 8.78, "learning_rate": 2.0380281690140843e-05, "loss": 1.4885, "step": 7835 }, { "epoch": 8.78, "learning_rate": 2.0378873239436622e-05, "loss": 1.4573, "step": 7836 }, { "epoch": 8.79, "learning_rate": 2.0377464788732395e-05, "loss": 2.6177, "step": 7837 }, { "epoch": 8.79, "learning_rate": 2.0376056338028172e-05, "loss": 2.1318, "step": 7838 }, { "epoch": 8.79, "learning_rate": 2.0374647887323945e-05, "loss": 2.1622, "step": 7839 }, { "epoch": 8.79, "learning_rate": 2.037323943661972e-05, "loss": 1.9601, "step": 7840 }, { "epoch": 8.79, "learning_rate": 2.0371830985915494e-05, "loss": 1.9121, "step": 7841 }, { "epoch": 8.79, "learning_rate": 2.0370422535211267e-05, "loss": 1.9612, "step": 7842 }, { "epoch": 8.79, "learning_rate": 2.0369014084507043e-05, "loss": 1.6782, "step": 7843 }, { "epoch": 8.79, "learning_rate": 2.0367605633802816e-05, "loss": 1.881, "step": 7844 }, { "epoch": 8.79, "learning_rate": 2.0366197183098592e-05, "loss": 1.7385, "step": 7845 }, { "epoch": 8.8, "learning_rate": 2.0364788732394365e-05, "loss": 1.8092, "step": 7846 }, { "epoch": 8.8, "learning_rate": 2.0363380281690142e-05, "loss": 1.9218, "step": 7847 }, { "epoch": 8.8, "learning_rate": 2.0361971830985915e-05, "loss": 1.8612, "step": 7848 }, { "epoch": 8.8, "learning_rate": 2.0360563380281688e-05, "loss": 1.7784, "step": 7849 }, { "epoch": 8.8, "learning_rate": 2.0359154929577464e-05, "loss": 1.8387, "step": 7850 }, { "epoch": 8.8, "learning_rate": 2.035774647887324e-05, "loss": 1.6939, "step": 7851 }, { "epoch": 8.8, "learning_rate": 2.0356338028169017e-05, "loss": 1.7463, "step": 7852 }, { "epoch": 8.8, "learning_rate": 2.035492957746479e-05, "loss": 1.6731, "step": 7853 }, { "epoch": 8.8, "learning_rate": 2.0353521126760566e-05, "loss": 1.6937, "step": 7854 }, { "epoch": 8.81, "learning_rate": 2.035211267605634e-05, "loss": 1.6543, "step": 7855 }, { "epoch": 8.81, "learning_rate": 2.0350704225352115e-05, "loss": 1.7895, "step": 7856 }, { "epoch": 8.81, "learning_rate": 2.0349295774647888e-05, "loss": 1.7045, "step": 7857 }, { "epoch": 8.81, "learning_rate": 2.034788732394366e-05, "loss": 1.7761, "step": 7858 }, { "epoch": 8.81, "learning_rate": 2.0346478873239437e-05, "loss": 1.76, "step": 7859 }, { "epoch": 8.81, "learning_rate": 2.034507042253521e-05, "loss": 1.6399, "step": 7860 }, { "epoch": 8.81, "learning_rate": 2.0343661971830987e-05, "loss": 1.7003, "step": 7861 }, { "epoch": 8.81, "learning_rate": 2.034225352112676e-05, "loss": 1.6141, "step": 7862 }, { "epoch": 8.82, "learning_rate": 2.0340845070422536e-05, "loss": 1.5946, "step": 7863 }, { "epoch": 8.82, "learning_rate": 2.033943661971831e-05, "loss": 1.825, "step": 7864 }, { "epoch": 8.82, "learning_rate": 2.0338028169014082e-05, "loss": 1.7668, "step": 7865 }, { "epoch": 8.82, "learning_rate": 2.033661971830986e-05, "loss": 1.6289, "step": 7866 }, { "epoch": 8.82, "learning_rate": 2.0335211267605635e-05, "loss": 1.7231, "step": 7867 }, { "epoch": 8.82, "learning_rate": 2.033380281690141e-05, "loss": 1.5693, "step": 7868 }, { "epoch": 8.82, "learning_rate": 2.0332394366197184e-05, "loss": 1.3982, "step": 7869 }, { "epoch": 8.82, "learning_rate": 2.033098591549296e-05, "loss": 1.4701, "step": 7870 }, { "epoch": 8.82, "learning_rate": 2.0329577464788733e-05, "loss": 1.605, "step": 7871 }, { "epoch": 8.83, "learning_rate": 2.032816901408451e-05, "loss": 1.6522, "step": 7872 }, { "epoch": 8.83, "learning_rate": 2.0326760563380282e-05, "loss": 1.5679, "step": 7873 }, { "epoch": 8.83, "learning_rate": 2.0325352112676055e-05, "loss": 1.5575, "step": 7874 }, { "epoch": 8.83, "learning_rate": 2.032394366197183e-05, "loss": 1.7388, "step": 7875 }, { "epoch": 8.83, "learning_rate": 2.0322535211267605e-05, "loss": 1.5483, "step": 7876 }, { "epoch": 8.83, "learning_rate": 2.032112676056338e-05, "loss": 1.4363, "step": 7877 }, { "epoch": 8.83, "learning_rate": 2.0319718309859154e-05, "loss": 1.3517, "step": 7878 }, { "epoch": 8.83, "learning_rate": 2.031830985915493e-05, "loss": 1.5717, "step": 7879 }, { "epoch": 8.83, "learning_rate": 2.0316901408450703e-05, "loss": 1.7757, "step": 7880 }, { "epoch": 8.84, "learning_rate": 2.031549295774648e-05, "loss": 1.4249, "step": 7881 }, { "epoch": 8.84, "learning_rate": 2.0314084507042256e-05, "loss": 1.6628, "step": 7882 }, { "epoch": 8.84, "learning_rate": 2.031267605633803e-05, "loss": 1.7718, "step": 7883 }, { "epoch": 8.84, "learning_rate": 2.0311267605633805e-05, "loss": 1.6108, "step": 7884 }, { "epoch": 8.84, "learning_rate": 2.0309859154929578e-05, "loss": 1.6527, "step": 7885 }, { "epoch": 8.84, "learning_rate": 2.0308450704225354e-05, "loss": 1.4773, "step": 7886 }, { "epoch": 8.84, "learning_rate": 2.0307042253521127e-05, "loss": 2.101, "step": 7887 }, { "epoch": 8.84, "learning_rate": 2.0305633802816904e-05, "loss": 1.9611, "step": 7888 }, { "epoch": 8.84, "learning_rate": 2.0304225352112677e-05, "loss": 2.0877, "step": 7889 }, { "epoch": 8.85, "learning_rate": 2.030281690140845e-05, "loss": 1.814, "step": 7890 }, { "epoch": 8.85, "learning_rate": 2.0301408450704226e-05, "loss": 1.7423, "step": 7891 }, { "epoch": 8.85, "learning_rate": 2.03e-05, "loss": 1.842, "step": 7892 }, { "epoch": 8.85, "learning_rate": 2.0298591549295775e-05, "loss": 1.5804, "step": 7893 }, { "epoch": 8.85, "learning_rate": 2.0297183098591548e-05, "loss": 1.6399, "step": 7894 }, { "epoch": 8.85, "learning_rate": 2.0295774647887324e-05, "loss": 1.727, "step": 7895 }, { "epoch": 8.85, "learning_rate": 2.02943661971831e-05, "loss": 1.8933, "step": 7896 }, { "epoch": 8.85, "learning_rate": 2.0292957746478874e-05, "loss": 1.7112, "step": 7897 }, { "epoch": 8.85, "learning_rate": 2.029154929577465e-05, "loss": 1.7014, "step": 7898 }, { "epoch": 8.86, "learning_rate": 2.0290140845070423e-05, "loss": 1.6662, "step": 7899 }, { "epoch": 8.86, "learning_rate": 2.02887323943662e-05, "loss": 1.8571, "step": 7900 }, { "epoch": 8.86, "learning_rate": 2.0287323943661972e-05, "loss": 1.5543, "step": 7901 }, { "epoch": 8.86, "learning_rate": 2.028591549295775e-05, "loss": 1.6755, "step": 7902 }, { "epoch": 8.86, "learning_rate": 2.028450704225352e-05, "loss": 1.5962, "step": 7903 }, { "epoch": 8.86, "learning_rate": 2.0283098591549298e-05, "loss": 1.6682, "step": 7904 }, { "epoch": 8.86, "learning_rate": 2.028169014084507e-05, "loss": 1.8658, "step": 7905 }, { "epoch": 8.86, "learning_rate": 2.0280281690140844e-05, "loss": 1.6271, "step": 7906 }, { "epoch": 8.86, "learning_rate": 2.027887323943662e-05, "loss": 1.7882, "step": 7907 }, { "epoch": 8.87, "learning_rate": 2.0277464788732393e-05, "loss": 1.7374, "step": 7908 }, { "epoch": 8.87, "learning_rate": 2.027605633802817e-05, "loss": 1.5022, "step": 7909 }, { "epoch": 8.87, "learning_rate": 2.0274647887323942e-05, "loss": 1.6018, "step": 7910 }, { "epoch": 8.87, "learning_rate": 2.0273239436619722e-05, "loss": 1.635, "step": 7911 }, { "epoch": 8.87, "learning_rate": 2.0271830985915495e-05, "loss": 1.718, "step": 7912 }, { "epoch": 8.87, "learning_rate": 2.0270422535211268e-05, "loss": 1.5943, "step": 7913 }, { "epoch": 8.87, "learning_rate": 2.0269014084507044e-05, "loss": 1.5235, "step": 7914 }, { "epoch": 8.87, "learning_rate": 2.0267605633802817e-05, "loss": 1.7187, "step": 7915 }, { "epoch": 8.87, "learning_rate": 2.0266197183098593e-05, "loss": 1.8589, "step": 7916 }, { "epoch": 8.88, "learning_rate": 2.0264788732394366e-05, "loss": 1.6609, "step": 7917 }, { "epoch": 8.88, "learning_rate": 2.0263380281690143e-05, "loss": 1.6179, "step": 7918 }, { "epoch": 8.88, "learning_rate": 2.0261971830985916e-05, "loss": 1.6533, "step": 7919 }, { "epoch": 8.88, "learning_rate": 2.026056338028169e-05, "loss": 1.7225, "step": 7920 }, { "epoch": 8.88, "learning_rate": 2.0259154929577465e-05, "loss": 1.7795, "step": 7921 }, { "epoch": 8.88, "learning_rate": 2.0257746478873238e-05, "loss": 1.4253, "step": 7922 }, { "epoch": 8.88, "learning_rate": 2.0256338028169014e-05, "loss": 1.4498, "step": 7923 }, { "epoch": 8.88, "learning_rate": 2.0254929577464787e-05, "loss": 1.4249, "step": 7924 }, { "epoch": 8.88, "learning_rate": 2.0253521126760563e-05, "loss": 1.4361, "step": 7925 }, { "epoch": 8.89, "learning_rate": 2.025211267605634e-05, "loss": 1.716, "step": 7926 }, { "epoch": 8.89, "learning_rate": 2.0250704225352116e-05, "loss": 1.5084, "step": 7927 }, { "epoch": 8.89, "learning_rate": 2.024929577464789e-05, "loss": 1.3299, "step": 7928 }, { "epoch": 8.89, "learning_rate": 2.0247887323943662e-05, "loss": 1.5731, "step": 7929 }, { "epoch": 8.89, "learning_rate": 2.024647887323944e-05, "loss": 1.4862, "step": 7930 }, { "epoch": 8.89, "learning_rate": 2.024507042253521e-05, "loss": 1.7644, "step": 7931 }, { "epoch": 8.89, "learning_rate": 2.0243661971830988e-05, "loss": 1.6747, "step": 7932 }, { "epoch": 8.89, "learning_rate": 2.024225352112676e-05, "loss": 1.1206, "step": 7933 }, { "epoch": 8.89, "learning_rate": 2.0240845070422537e-05, "loss": 1.3137, "step": 7934 }, { "epoch": 8.9, "learning_rate": 2.023943661971831e-05, "loss": 1.7534, "step": 7935 }, { "epoch": 8.9, "learning_rate": 2.0238028169014083e-05, "loss": 1.4301, "step": 7936 }, { "epoch": 8.9, "learning_rate": 2.023661971830986e-05, "loss": 2.284, "step": 7937 }, { "epoch": 8.9, "learning_rate": 2.0235211267605632e-05, "loss": 2.3586, "step": 7938 }, { "epoch": 8.9, "learning_rate": 2.023380281690141e-05, "loss": 1.6628, "step": 7939 }, { "epoch": 8.9, "learning_rate": 2.023239436619718e-05, "loss": 1.8591, "step": 7940 }, { "epoch": 8.9, "learning_rate": 2.0230985915492958e-05, "loss": 1.7581, "step": 7941 }, { "epoch": 8.9, "learning_rate": 2.0229577464788734e-05, "loss": 2.0346, "step": 7942 }, { "epoch": 8.9, "learning_rate": 2.022816901408451e-05, "loss": 1.8582, "step": 7943 }, { "epoch": 8.91, "learning_rate": 2.0226760563380283e-05, "loss": 1.7362, "step": 7944 }, { "epoch": 8.91, "learning_rate": 2.0225352112676056e-05, "loss": 1.7093, "step": 7945 }, { "epoch": 8.91, "learning_rate": 2.0223943661971833e-05, "loss": 1.8185, "step": 7946 }, { "epoch": 8.91, "learning_rate": 2.0222535211267606e-05, "loss": 1.9268, "step": 7947 }, { "epoch": 8.91, "learning_rate": 2.0221126760563382e-05, "loss": 1.6183, "step": 7948 }, { "epoch": 8.91, "learning_rate": 2.0219718309859155e-05, "loss": 1.687, "step": 7949 }, { "epoch": 8.91, "learning_rate": 2.021830985915493e-05, "loss": 1.6612, "step": 7950 }, { "epoch": 8.91, "learning_rate": 2.0216901408450704e-05, "loss": 1.6835, "step": 7951 }, { "epoch": 8.91, "learning_rate": 2.0215492957746477e-05, "loss": 1.7513, "step": 7952 }, { "epoch": 8.92, "learning_rate": 2.0214084507042253e-05, "loss": 1.6304, "step": 7953 }, { "epoch": 8.92, "learning_rate": 2.0212676056338026e-05, "loss": 1.4747, "step": 7954 }, { "epoch": 8.92, "learning_rate": 2.0211267605633803e-05, "loss": 1.6211, "step": 7955 }, { "epoch": 8.92, "learning_rate": 2.0209859154929576e-05, "loss": 1.6613, "step": 7956 }, { "epoch": 8.92, "learning_rate": 2.0208450704225355e-05, "loss": 1.5641, "step": 7957 }, { "epoch": 8.92, "learning_rate": 2.0207042253521128e-05, "loss": 1.5409, "step": 7958 }, { "epoch": 8.92, "learning_rate": 2.0205633802816905e-05, "loss": 1.6258, "step": 7959 }, { "epoch": 8.92, "learning_rate": 2.0204225352112678e-05, "loss": 1.5643, "step": 7960 }, { "epoch": 8.92, "learning_rate": 2.020281690140845e-05, "loss": 1.5659, "step": 7961 }, { "epoch": 8.93, "learning_rate": 2.0201408450704227e-05, "loss": 1.6118, "step": 7962 }, { "epoch": 8.93, "learning_rate": 2.02e-05, "loss": 1.6073, "step": 7963 }, { "epoch": 8.93, "learning_rate": 2.0198591549295776e-05, "loss": 1.6399, "step": 7964 }, { "epoch": 8.93, "learning_rate": 2.019718309859155e-05, "loss": 1.7221, "step": 7965 }, { "epoch": 8.93, "learning_rate": 2.0195774647887325e-05, "loss": 1.8338, "step": 7966 }, { "epoch": 8.93, "learning_rate": 2.0194366197183098e-05, "loss": 1.7507, "step": 7967 }, { "epoch": 8.93, "learning_rate": 2.019295774647887e-05, "loss": 1.5701, "step": 7968 }, { "epoch": 8.93, "learning_rate": 2.0191549295774648e-05, "loss": 1.5175, "step": 7969 }, { "epoch": 8.93, "learning_rate": 2.019014084507042e-05, "loss": 1.5095, "step": 7970 }, { "epoch": 8.94, "learning_rate": 2.0188732394366197e-05, "loss": 1.5596, "step": 7971 }, { "epoch": 8.94, "learning_rate": 2.0187323943661973e-05, "loss": 1.5183, "step": 7972 }, { "epoch": 8.94, "learning_rate": 2.018591549295775e-05, "loss": 1.4964, "step": 7973 }, { "epoch": 8.94, "learning_rate": 2.0184507042253522e-05, "loss": 1.4887, "step": 7974 }, { "epoch": 8.94, "learning_rate": 2.01830985915493e-05, "loss": 1.5783, "step": 7975 }, { "epoch": 8.94, "learning_rate": 2.0181690140845072e-05, "loss": 1.5087, "step": 7976 }, { "epoch": 8.94, "learning_rate": 2.0180281690140845e-05, "loss": 1.6373, "step": 7977 }, { "epoch": 8.94, "learning_rate": 2.017887323943662e-05, "loss": 1.7616, "step": 7978 }, { "epoch": 8.95, "learning_rate": 2.0177464788732394e-05, "loss": 1.8225, "step": 7979 }, { "epoch": 8.95, "learning_rate": 2.017605633802817e-05, "loss": 1.4509, "step": 7980 }, { "epoch": 8.95, "learning_rate": 2.0174647887323943e-05, "loss": 1.583, "step": 7981 }, { "epoch": 8.95, "learning_rate": 2.017323943661972e-05, "loss": 1.3568, "step": 7982 }, { "epoch": 8.95, "learning_rate": 2.0171830985915492e-05, "loss": 1.5772, "step": 7983 }, { "epoch": 8.95, "learning_rate": 2.0170422535211265e-05, "loss": 1.4959, "step": 7984 }, { "epoch": 8.95, "learning_rate": 2.0169014084507042e-05, "loss": 1.5773, "step": 7985 }, { "epoch": 8.95, "learning_rate": 2.0167605633802815e-05, "loss": 1.5315, "step": 7986 }, { "epoch": 8.95, "learning_rate": 2.0166197183098594e-05, "loss": 1.9956, "step": 7987 }, { "epoch": 8.96, "learning_rate": 2.0164788732394367e-05, "loss": 1.9931, "step": 7988 }, { "epoch": 8.96, "learning_rate": 2.0163380281690144e-05, "loss": 1.7288, "step": 7989 }, { "epoch": 8.96, "learning_rate": 2.0161971830985917e-05, "loss": 1.7861, "step": 7990 }, { "epoch": 8.96, "learning_rate": 2.0160563380281693e-05, "loss": 1.8682, "step": 7991 }, { "epoch": 8.96, "learning_rate": 2.0159154929577466e-05, "loss": 1.8861, "step": 7992 }, { "epoch": 8.96, "learning_rate": 2.015774647887324e-05, "loss": 1.561, "step": 7993 }, { "epoch": 8.96, "learning_rate": 2.0156338028169015e-05, "loss": 1.7188, "step": 7994 }, { "epoch": 8.96, "learning_rate": 2.0154929577464788e-05, "loss": 1.6546, "step": 7995 }, { "epoch": 8.96, "learning_rate": 2.0153521126760564e-05, "loss": 1.7711, "step": 7996 }, { "epoch": 8.97, "learning_rate": 2.0152112676056337e-05, "loss": 1.8204, "step": 7997 }, { "epoch": 8.97, "learning_rate": 2.0150704225352114e-05, "loss": 1.5193, "step": 7998 }, { "epoch": 8.97, "learning_rate": 2.0149295774647887e-05, "loss": 1.5889, "step": 7999 }, { "epoch": 8.97, "learning_rate": 2.014788732394366e-05, "loss": 1.6327, "step": 8000 }, { "epoch": 8.97, "learning_rate": 2.0146478873239436e-05, "loss": 1.6584, "step": 8001 }, { "epoch": 8.97, "learning_rate": 2.0145070422535212e-05, "loss": 1.5551, "step": 8002 }, { "epoch": 8.97, "learning_rate": 2.014366197183099e-05, "loss": 1.9346, "step": 8003 }, { "epoch": 8.97, "learning_rate": 2.014225352112676e-05, "loss": 1.5635, "step": 8004 }, { "epoch": 8.97, "learning_rate": 2.0140845070422538e-05, "loss": 1.7932, "step": 8005 }, { "epoch": 8.98, "learning_rate": 2.013943661971831e-05, "loss": 1.6527, "step": 8006 }, { "epoch": 8.98, "learning_rate": 2.0138028169014087e-05, "loss": 1.4408, "step": 8007 }, { "epoch": 8.98, "learning_rate": 2.013661971830986e-05, "loss": 1.8334, "step": 8008 }, { "epoch": 8.98, "learning_rate": 2.0135211267605633e-05, "loss": 1.5373, "step": 8009 }, { "epoch": 8.98, "learning_rate": 2.013380281690141e-05, "loss": 1.5604, "step": 8010 }, { "epoch": 8.98, "learning_rate": 2.0132394366197182e-05, "loss": 1.7893, "step": 8011 }, { "epoch": 8.98, "learning_rate": 2.013098591549296e-05, "loss": 1.6153, "step": 8012 }, { "epoch": 8.98, "learning_rate": 2.012957746478873e-05, "loss": 1.4752, "step": 8013 }, { "epoch": 8.98, "learning_rate": 2.0128169014084508e-05, "loss": 1.6364, "step": 8014 }, { "epoch": 8.99, "learning_rate": 2.012676056338028e-05, "loss": 1.4649, "step": 8015 }, { "epoch": 8.99, "learning_rate": 2.0125352112676054e-05, "loss": 1.6388, "step": 8016 }, { "epoch": 8.99, "learning_rate": 2.0123943661971834e-05, "loss": 1.6056, "step": 8017 }, { "epoch": 8.99, "learning_rate": 2.0122535211267606e-05, "loss": 1.3435, "step": 8018 }, { "epoch": 8.99, "learning_rate": 2.0121126760563383e-05, "loss": 1.5799, "step": 8019 }, { "epoch": 8.99, "learning_rate": 2.0119718309859156e-05, "loss": 1.7717, "step": 8020 }, { "epoch": 8.99, "learning_rate": 2.0118309859154932e-05, "loss": 1.8534, "step": 8021 }, { "epoch": 8.99, "learning_rate": 2.0116901408450705e-05, "loss": 1.4026, "step": 8022 }, { "epoch": 8.99, "learning_rate": 2.011549295774648e-05, "loss": 1.6833, "step": 8023 }, { "epoch": 9.0, "learning_rate": 2.0114084507042254e-05, "loss": 1.2933, "step": 8024 }, { "epoch": 9.0, "learning_rate": 2.0112676056338027e-05, "loss": 1.9502, "step": 8025 }, { "epoch": 9.0, "learning_rate": 2.0111267605633804e-05, "loss": 1.4664, "step": 8026 }, { "epoch": 9.0, "learning_rate": 2.0109859154929577e-05, "loss": 1.2964, "step": 8027 }, { "epoch": 9.0, "learning_rate": 2.0108450704225353e-05, "loss": 1.4696, "step": 8028 }, { "epoch": 9.0, "learning_rate": 2.0107042253521126e-05, "loss": 2.1308, "step": 8029 }, { "epoch": 9.0, "learning_rate": 2.0105633802816902e-05, "loss": 1.9281, "step": 8030 }, { "epoch": 9.0, "learning_rate": 2.0104225352112675e-05, "loss": 2.023, "step": 8031 }, { "epoch": 9.0, "learning_rate": 2.010281690140845e-05, "loss": 1.9562, "step": 8032 }, { "epoch": 9.01, "learning_rate": 2.0101408450704228e-05, "loss": 1.6526, "step": 8033 }, { "epoch": 9.01, "learning_rate": 2.01e-05, "loss": 1.786, "step": 8034 }, { "epoch": 9.01, "learning_rate": 2.0098591549295777e-05, "loss": 1.6703, "step": 8035 }, { "epoch": 9.01, "learning_rate": 2.009718309859155e-05, "loss": 1.5588, "step": 8036 }, { "epoch": 9.01, "learning_rate": 2.0095774647887326e-05, "loss": 1.6219, "step": 8037 }, { "epoch": 9.01, "learning_rate": 2.00943661971831e-05, "loss": 1.6016, "step": 8038 }, { "epoch": 9.01, "learning_rate": 2.0092957746478872e-05, "loss": 1.7419, "step": 8039 }, { "epoch": 9.01, "learning_rate": 2.009154929577465e-05, "loss": 1.7367, "step": 8040 }, { "epoch": 9.01, "learning_rate": 2.009014084507042e-05, "loss": 1.5177, "step": 8041 }, { "epoch": 9.02, "learning_rate": 2.0088732394366198e-05, "loss": 1.6565, "step": 8042 }, { "epoch": 9.02, "learning_rate": 2.008732394366197e-05, "loss": 1.6188, "step": 8043 }, { "epoch": 9.02, "learning_rate": 2.0085915492957747e-05, "loss": 1.5391, "step": 8044 }, { "epoch": 9.02, "learning_rate": 2.008450704225352e-05, "loss": 1.6033, "step": 8045 }, { "epoch": 9.02, "learning_rate": 2.0083098591549296e-05, "loss": 1.6072, "step": 8046 }, { "epoch": 9.02, "learning_rate": 2.008169014084507e-05, "loss": 1.5438, "step": 8047 }, { "epoch": 9.02, "learning_rate": 2.0080281690140846e-05, "loss": 1.5315, "step": 8048 }, { "epoch": 9.02, "learning_rate": 2.0078873239436622e-05, "loss": 1.574, "step": 8049 }, { "epoch": 9.02, "learning_rate": 2.0077464788732395e-05, "loss": 1.7205, "step": 8050 }, { "epoch": 9.03, "learning_rate": 2.007605633802817e-05, "loss": 1.4953, "step": 8051 }, { "epoch": 9.03, "learning_rate": 2.0074647887323944e-05, "loss": 1.4824, "step": 8052 }, { "epoch": 9.03, "learning_rate": 2.007323943661972e-05, "loss": 1.4859, "step": 8053 }, { "epoch": 9.03, "learning_rate": 2.0071830985915493e-05, "loss": 1.5475, "step": 8054 }, { "epoch": 9.03, "learning_rate": 2.0070422535211266e-05, "loss": 1.5285, "step": 8055 }, { "epoch": 9.03, "learning_rate": 2.0069014084507043e-05, "loss": 1.3851, "step": 8056 }, { "epoch": 9.03, "learning_rate": 2.0067605633802816e-05, "loss": 1.7544, "step": 8057 }, { "epoch": 9.03, "learning_rate": 2.0066197183098592e-05, "loss": 1.7004, "step": 8058 }, { "epoch": 9.03, "learning_rate": 2.0064788732394365e-05, "loss": 1.4593, "step": 8059 }, { "epoch": 9.04, "learning_rate": 2.006338028169014e-05, "loss": 1.44, "step": 8060 }, { "epoch": 9.04, "learning_rate": 2.0061971830985914e-05, "loss": 1.7718, "step": 8061 }, { "epoch": 9.04, "learning_rate": 2.006056338028169e-05, "loss": 1.7153, "step": 8062 }, { "epoch": 9.04, "learning_rate": 2.0059154929577467e-05, "loss": 1.367, "step": 8063 }, { "epoch": 9.04, "learning_rate": 2.005774647887324e-05, "loss": 1.2944, "step": 8064 }, { "epoch": 9.04, "learning_rate": 2.0056338028169016e-05, "loss": 1.3261, "step": 8065 }, { "epoch": 9.04, "learning_rate": 2.005492957746479e-05, "loss": 1.7726, "step": 8066 }, { "epoch": 9.04, "learning_rate": 2.0053521126760565e-05, "loss": 1.6756, "step": 8067 }, { "epoch": 9.04, "learning_rate": 2.005211267605634e-05, "loss": 1.472, "step": 8068 }, { "epoch": 9.05, "learning_rate": 2.0050704225352115e-05, "loss": 1.6077, "step": 8069 }, { "epoch": 9.05, "learning_rate": 2.0049295774647888e-05, "loss": 1.5332, "step": 8070 }, { "epoch": 9.05, "learning_rate": 2.004788732394366e-05, "loss": 1.396, "step": 8071 }, { "epoch": 9.05, "learning_rate": 2.0046478873239437e-05, "loss": 1.2778, "step": 8072 }, { "epoch": 9.05, "learning_rate": 2.004507042253521e-05, "loss": 1.4119, "step": 8073 }, { "epoch": 9.05, "learning_rate": 2.0043661971830986e-05, "loss": 1.3008, "step": 8074 }, { "epoch": 9.05, "learning_rate": 2.004225352112676e-05, "loss": 1.1951, "step": 8075 }, { "epoch": 9.05, "learning_rate": 2.0040845070422535e-05, "loss": 1.4096, "step": 8076 }, { "epoch": 9.05, "learning_rate": 2.003943661971831e-05, "loss": 1.3591, "step": 8077 }, { "epoch": 9.06, "learning_rate": 2.0038028169014088e-05, "loss": 1.1175, "step": 8078 }, { "epoch": 9.06, "learning_rate": 2.003661971830986e-05, "loss": 2.1943, "step": 8079 }, { "epoch": 9.06, "learning_rate": 2.0035211267605634e-05, "loss": 1.9071, "step": 8080 }, { "epoch": 9.06, "learning_rate": 2.003380281690141e-05, "loss": 1.7401, "step": 8081 }, { "epoch": 9.06, "learning_rate": 2.0032394366197183e-05, "loss": 1.8251, "step": 8082 }, { "epoch": 9.06, "learning_rate": 2.003098591549296e-05, "loss": 1.6813, "step": 8083 }, { "epoch": 9.06, "learning_rate": 2.0029577464788733e-05, "loss": 1.7545, "step": 8084 }, { "epoch": 9.06, "learning_rate": 2.002816901408451e-05, "loss": 1.7249, "step": 8085 }, { "epoch": 9.07, "learning_rate": 2.0026760563380282e-05, "loss": 1.7301, "step": 8086 }, { "epoch": 9.07, "learning_rate": 2.0025352112676055e-05, "loss": 1.5815, "step": 8087 }, { "epoch": 9.07, "learning_rate": 2.002394366197183e-05, "loss": 1.7845, "step": 8088 }, { "epoch": 9.07, "learning_rate": 2.0022535211267604e-05, "loss": 1.5161, "step": 8089 }, { "epoch": 9.07, "learning_rate": 2.002112676056338e-05, "loss": 1.5039, "step": 8090 }, { "epoch": 9.07, "learning_rate": 2.0019718309859153e-05, "loss": 1.5382, "step": 8091 }, { "epoch": 9.07, "learning_rate": 2.001830985915493e-05, "loss": 1.6564, "step": 8092 }, { "epoch": 9.07, "learning_rate": 2.0016901408450706e-05, "loss": 1.2948, "step": 8093 }, { "epoch": 9.07, "learning_rate": 2.0015492957746482e-05, "loss": 1.515, "step": 8094 }, { "epoch": 9.08, "learning_rate": 2.0014084507042255e-05, "loss": 1.5629, "step": 8095 }, { "epoch": 9.08, "learning_rate": 2.0012676056338028e-05, "loss": 1.4656, "step": 8096 }, { "epoch": 9.08, "learning_rate": 2.0011267605633805e-05, "loss": 1.465, "step": 8097 }, { "epoch": 9.08, "learning_rate": 2.0009859154929577e-05, "loss": 1.6888, "step": 8098 }, { "epoch": 9.08, "learning_rate": 2.0008450704225354e-05, "loss": 1.4685, "step": 8099 }, { "epoch": 9.08, "learning_rate": 2.0007042253521127e-05, "loss": 1.4905, "step": 8100 }, { "epoch": 9.08, "learning_rate": 2.0005633802816903e-05, "loss": 1.4478, "step": 8101 }, { "epoch": 9.08, "learning_rate": 2.0004225352112676e-05, "loss": 1.4882, "step": 8102 }, { "epoch": 9.08, "learning_rate": 2.000281690140845e-05, "loss": 1.5415, "step": 8103 }, { "epoch": 9.09, "learning_rate": 2.0001408450704225e-05, "loss": 1.4605, "step": 8104 }, { "epoch": 9.09, "learning_rate": 1.9999999999999998e-05, "loss": 1.4886, "step": 8105 }, { "epoch": 9.09, "learning_rate": 1.9998591549295775e-05, "loss": 1.4231, "step": 8106 }, { "epoch": 9.09, "learning_rate": 1.9997183098591548e-05, "loss": 1.4733, "step": 8107 }, { "epoch": 9.09, "learning_rate": 1.9995774647887327e-05, "loss": 1.3217, "step": 8108 }, { "epoch": 9.09, "learning_rate": 1.99943661971831e-05, "loss": 1.6767, "step": 8109 }, { "epoch": 9.09, "learning_rate": 1.9992957746478877e-05, "loss": 1.5384, "step": 8110 }, { "epoch": 9.09, "learning_rate": 1.999154929577465e-05, "loss": 1.2394, "step": 8111 }, { "epoch": 9.09, "learning_rate": 1.9990140845070422e-05, "loss": 1.4365, "step": 8112 }, { "epoch": 9.1, "learning_rate": 1.99887323943662e-05, "loss": 1.5725, "step": 8113 }, { "epoch": 9.1, "learning_rate": 1.998732394366197e-05, "loss": 1.1946, "step": 8114 }, { "epoch": 9.1, "learning_rate": 1.9985915492957748e-05, "loss": 1.4768, "step": 8115 }, { "epoch": 9.1, "learning_rate": 1.998450704225352e-05, "loss": 1.4959, "step": 8116 }, { "epoch": 9.1, "learning_rate": 1.9983098591549297e-05, "loss": 1.294, "step": 8117 }, { "epoch": 9.1, "learning_rate": 1.998169014084507e-05, "loss": 1.4702, "step": 8118 }, { "epoch": 9.1, "learning_rate": 1.9980281690140843e-05, "loss": 1.305, "step": 8119 }, { "epoch": 9.1, "learning_rate": 1.997887323943662e-05, "loss": 1.3841, "step": 8120 }, { "epoch": 9.1, "learning_rate": 1.9977464788732392e-05, "loss": 1.2583, "step": 8121 }, { "epoch": 9.11, "learning_rate": 1.997605633802817e-05, "loss": 1.3463, "step": 8122 }, { "epoch": 9.11, "learning_rate": 1.9974647887323945e-05, "loss": 1.2002, "step": 8123 }, { "epoch": 9.11, "learning_rate": 1.997323943661972e-05, "loss": 1.262, "step": 8124 }, { "epoch": 9.11, "learning_rate": 1.9971830985915494e-05, "loss": 1.4172, "step": 8125 }, { "epoch": 9.11, "learning_rate": 1.997042253521127e-05, "loss": 1.5473, "step": 8126 }, { "epoch": 9.11, "learning_rate": 1.9969014084507044e-05, "loss": 1.1461, "step": 8127 }, { "epoch": 9.11, "learning_rate": 1.9967605633802817e-05, "loss": 1.5511, "step": 8128 }, { "epoch": 9.11, "learning_rate": 1.9966197183098593e-05, "loss": 1.9541, "step": 8129 }, { "epoch": 9.11, "learning_rate": 1.9964788732394366e-05, "loss": 1.8286, "step": 8130 }, { "epoch": 9.12, "learning_rate": 1.9963380281690142e-05, "loss": 1.7744, "step": 8131 }, { "epoch": 9.12, "learning_rate": 1.9961971830985915e-05, "loss": 1.8526, "step": 8132 }, { "epoch": 9.12, "learning_rate": 1.996056338028169e-05, "loss": 1.4634, "step": 8133 }, { "epoch": 9.12, "learning_rate": 1.9959154929577464e-05, "loss": 1.3563, "step": 8134 }, { "epoch": 9.12, "learning_rate": 1.9957746478873237e-05, "loss": 1.7426, "step": 8135 }, { "epoch": 9.12, "learning_rate": 1.9956338028169014e-05, "loss": 1.5144, "step": 8136 }, { "epoch": 9.12, "learning_rate": 1.9954929577464787e-05, "loss": 1.6197, "step": 8137 }, { "epoch": 9.12, "learning_rate": 1.9953521126760566e-05, "loss": 1.4988, "step": 8138 }, { "epoch": 9.12, "learning_rate": 1.995211267605634e-05, "loss": 1.5175, "step": 8139 }, { "epoch": 9.13, "learning_rate": 1.9950704225352116e-05, "loss": 1.5568, "step": 8140 }, { "epoch": 9.13, "learning_rate": 1.994929577464789e-05, "loss": 1.4094, "step": 8141 }, { "epoch": 9.13, "learning_rate": 1.994788732394366e-05, "loss": 1.4367, "step": 8142 }, { "epoch": 9.13, "learning_rate": 1.9946478873239438e-05, "loss": 1.604, "step": 8143 }, { "epoch": 9.13, "learning_rate": 1.994507042253521e-05, "loss": 1.3386, "step": 8144 }, { "epoch": 9.13, "learning_rate": 1.9943661971830987e-05, "loss": 1.5649, "step": 8145 }, { "epoch": 9.13, "learning_rate": 1.994225352112676e-05, "loss": 1.5865, "step": 8146 }, { "epoch": 9.13, "learning_rate": 1.9940845070422536e-05, "loss": 1.4737, "step": 8147 }, { "epoch": 9.13, "learning_rate": 1.993943661971831e-05, "loss": 1.6028, "step": 8148 }, { "epoch": 9.14, "learning_rate": 1.9938028169014086e-05, "loss": 1.6359, "step": 8149 }, { "epoch": 9.14, "learning_rate": 1.993661971830986e-05, "loss": 1.436, "step": 8150 }, { "epoch": 9.14, "learning_rate": 1.993521126760563e-05, "loss": 1.3828, "step": 8151 }, { "epoch": 9.14, "learning_rate": 1.9933802816901408e-05, "loss": 1.5062, "step": 8152 }, { "epoch": 9.14, "learning_rate": 1.9932394366197184e-05, "loss": 1.3828, "step": 8153 }, { "epoch": 9.14, "learning_rate": 1.993098591549296e-05, "loss": 1.4317, "step": 8154 }, { "epoch": 9.14, "learning_rate": 1.9929577464788734e-05, "loss": 1.3683, "step": 8155 }, { "epoch": 9.14, "learning_rate": 1.992816901408451e-05, "loss": 1.1994, "step": 8156 }, { "epoch": 9.14, "learning_rate": 1.9926760563380283e-05, "loss": 1.6386, "step": 8157 }, { "epoch": 9.15, "learning_rate": 1.9925352112676056e-05, "loss": 1.4668, "step": 8158 }, { "epoch": 9.15, "learning_rate": 1.9923943661971832e-05, "loss": 1.4749, "step": 8159 }, { "epoch": 9.15, "learning_rate": 1.9922535211267605e-05, "loss": 1.4916, "step": 8160 }, { "epoch": 9.15, "learning_rate": 1.992112676056338e-05, "loss": 1.5315, "step": 8161 }, { "epoch": 9.15, "learning_rate": 1.9919718309859154e-05, "loss": 1.3674, "step": 8162 }, { "epoch": 9.15, "learning_rate": 1.991830985915493e-05, "loss": 1.4855, "step": 8163 }, { "epoch": 9.15, "learning_rate": 1.9916901408450704e-05, "loss": 1.7717, "step": 8164 }, { "epoch": 9.15, "learning_rate": 1.991549295774648e-05, "loss": 1.5492, "step": 8165 }, { "epoch": 9.15, "learning_rate": 1.9914084507042253e-05, "loss": 1.4262, "step": 8166 }, { "epoch": 9.16, "learning_rate": 1.9912676056338026e-05, "loss": 1.6173, "step": 8167 }, { "epoch": 9.16, "learning_rate": 1.9911267605633802e-05, "loss": 1.3378, "step": 8168 }, { "epoch": 9.16, "learning_rate": 1.990985915492958e-05, "loss": 1.4245, "step": 8169 }, { "epoch": 9.16, "learning_rate": 1.9908450704225355e-05, "loss": 1.3518, "step": 8170 }, { "epoch": 9.16, "learning_rate": 1.9907042253521128e-05, "loss": 1.42, "step": 8171 }, { "epoch": 9.16, "learning_rate": 1.9905633802816904e-05, "loss": 1.6099, "step": 8172 }, { "epoch": 9.16, "learning_rate": 1.9904225352112677e-05, "loss": 1.3051, "step": 8173 }, { "epoch": 9.16, "learning_rate": 1.990281690140845e-05, "loss": 1.4267, "step": 8174 }, { "epoch": 9.16, "learning_rate": 1.9901408450704226e-05, "loss": 1.1599, "step": 8175 }, { "epoch": 9.17, "learning_rate": 1.99e-05, "loss": 1.4401, "step": 8176 }, { "epoch": 9.17, "learning_rate": 1.9898591549295776e-05, "loss": 1.7944, "step": 8177 }, { "epoch": 9.17, "learning_rate": 1.989718309859155e-05, "loss": 1.2827, "step": 8178 }, { "epoch": 9.17, "learning_rate": 1.9895774647887325e-05, "loss": 2.2034, "step": 8179 }, { "epoch": 9.17, "learning_rate": 1.9894366197183098e-05, "loss": 1.8208, "step": 8180 }, { "epoch": 9.17, "learning_rate": 1.9892957746478874e-05, "loss": 1.7057, "step": 8181 }, { "epoch": 9.17, "learning_rate": 1.9891549295774647e-05, "loss": 1.468, "step": 8182 }, { "epoch": 9.17, "learning_rate": 1.989014084507042e-05, "loss": 1.8173, "step": 8183 }, { "epoch": 9.17, "learning_rate": 1.98887323943662e-05, "loss": 1.6853, "step": 8184 }, { "epoch": 9.18, "learning_rate": 1.9887323943661973e-05, "loss": 1.5182, "step": 8185 }, { "epoch": 9.18, "learning_rate": 1.988591549295775e-05, "loss": 1.5978, "step": 8186 }, { "epoch": 9.18, "learning_rate": 1.9884507042253522e-05, "loss": 1.6858, "step": 8187 }, { "epoch": 9.18, "learning_rate": 1.9883098591549298e-05, "loss": 1.6284, "step": 8188 }, { "epoch": 9.18, "learning_rate": 1.988169014084507e-05, "loss": 1.7754, "step": 8189 }, { "epoch": 9.18, "learning_rate": 1.9880281690140844e-05, "loss": 1.5459, "step": 8190 }, { "epoch": 9.18, "learning_rate": 1.987887323943662e-05, "loss": 1.6707, "step": 8191 }, { "epoch": 9.18, "learning_rate": 1.9877464788732393e-05, "loss": 1.2816, "step": 8192 }, { "epoch": 9.18, "learning_rate": 1.987605633802817e-05, "loss": 1.5363, "step": 8193 }, { "epoch": 9.19, "learning_rate": 1.9874647887323943e-05, "loss": 1.434, "step": 8194 }, { "epoch": 9.19, "learning_rate": 1.987323943661972e-05, "loss": 1.4313, "step": 8195 }, { "epoch": 9.19, "learning_rate": 1.9871830985915492e-05, "loss": 1.3988, "step": 8196 }, { "epoch": 9.19, "learning_rate": 1.9870422535211268e-05, "loss": 1.4314, "step": 8197 }, { "epoch": 9.19, "learning_rate": 1.986901408450704e-05, "loss": 1.2413, "step": 8198 }, { "epoch": 9.19, "learning_rate": 1.9867605633802818e-05, "loss": 1.2043, "step": 8199 }, { "epoch": 9.19, "learning_rate": 1.9866197183098594e-05, "loss": 1.7969, "step": 8200 }, { "epoch": 9.19, "learning_rate": 1.9864788732394367e-05, "loss": 1.3972, "step": 8201 }, { "epoch": 9.2, "learning_rate": 1.9863380281690143e-05, "loss": 1.2525, "step": 8202 }, { "epoch": 9.2, "learning_rate": 1.9861971830985916e-05, "loss": 1.2966, "step": 8203 }, { "epoch": 9.2, "learning_rate": 1.9860563380281692e-05, "loss": 1.359, "step": 8204 }, { "epoch": 9.2, "learning_rate": 1.9859154929577465e-05, "loss": 1.3073, "step": 8205 }, { "epoch": 9.2, "learning_rate": 1.985774647887324e-05, "loss": 1.5572, "step": 8206 }, { "epoch": 9.2, "learning_rate": 1.9856338028169015e-05, "loss": 1.4392, "step": 8207 }, { "epoch": 9.2, "learning_rate": 1.9854929577464788e-05, "loss": 1.4204, "step": 8208 }, { "epoch": 9.2, "learning_rate": 1.9853521126760564e-05, "loss": 1.5306, "step": 8209 }, { "epoch": 9.2, "learning_rate": 1.9852112676056337e-05, "loss": 1.4563, "step": 8210 }, { "epoch": 9.21, "learning_rate": 1.9850704225352113e-05, "loss": 1.5613, "step": 8211 }, { "epoch": 9.21, "learning_rate": 1.9849295774647886e-05, "loss": 1.2619, "step": 8212 }, { "epoch": 9.21, "learning_rate": 1.9847887323943662e-05, "loss": 1.4494, "step": 8213 }, { "epoch": 9.21, "learning_rate": 1.984647887323944e-05, "loss": 1.6179, "step": 8214 }, { "epoch": 9.21, "learning_rate": 1.9845070422535212e-05, "loss": 1.6077, "step": 8215 }, { "epoch": 9.21, "learning_rate": 1.9843661971830988e-05, "loss": 1.493, "step": 8216 }, { "epoch": 9.21, "learning_rate": 1.984225352112676e-05, "loss": 1.3912, "step": 8217 }, { "epoch": 9.21, "learning_rate": 1.9840845070422537e-05, "loss": 1.4095, "step": 8218 }, { "epoch": 9.21, "learning_rate": 1.983943661971831e-05, "loss": 1.5667, "step": 8219 }, { "epoch": 9.22, "learning_rate": 1.9838028169014087e-05, "loss": 1.5105, "step": 8220 }, { "epoch": 9.22, "learning_rate": 1.983661971830986e-05, "loss": 1.1887, "step": 8221 }, { "epoch": 9.22, "learning_rate": 1.9835211267605633e-05, "loss": 1.7121, "step": 8222 }, { "epoch": 9.22, "learning_rate": 1.983380281690141e-05, "loss": 1.4209, "step": 8223 }, { "epoch": 9.22, "learning_rate": 1.9832394366197182e-05, "loss": 1.3082, "step": 8224 }, { "epoch": 9.22, "learning_rate": 1.9830985915492958e-05, "loss": 1.3848, "step": 8225 }, { "epoch": 9.22, "learning_rate": 1.982957746478873e-05, "loss": 1.553, "step": 8226 }, { "epoch": 9.22, "learning_rate": 1.9828169014084507e-05, "loss": 1.5238, "step": 8227 }, { "epoch": 9.22, "learning_rate": 1.982676056338028e-05, "loss": 1.1396, "step": 8228 }, { "epoch": 9.23, "learning_rate": 1.982535211267606e-05, "loss": 2.0814, "step": 8229 }, { "epoch": 9.23, "learning_rate": 1.9823943661971833e-05, "loss": 1.9337, "step": 8230 }, { "epoch": 9.23, "learning_rate": 1.9822535211267606e-05, "loss": 1.7683, "step": 8231 }, { "epoch": 9.23, "learning_rate": 1.9821126760563382e-05, "loss": 1.6776, "step": 8232 }, { "epoch": 9.23, "learning_rate": 1.9819718309859155e-05, "loss": 1.5401, "step": 8233 }, { "epoch": 9.23, "learning_rate": 1.981830985915493e-05, "loss": 1.5799, "step": 8234 }, { "epoch": 9.23, "learning_rate": 1.9816901408450705e-05, "loss": 1.4657, "step": 8235 }, { "epoch": 9.23, "learning_rate": 1.981549295774648e-05, "loss": 1.5851, "step": 8236 }, { "epoch": 9.23, "learning_rate": 1.9814084507042254e-05, "loss": 1.7568, "step": 8237 }, { "epoch": 9.24, "learning_rate": 1.9812676056338027e-05, "loss": 1.6326, "step": 8238 }, { "epoch": 9.24, "learning_rate": 1.9811267605633803e-05, "loss": 1.6446, "step": 8239 }, { "epoch": 9.24, "learning_rate": 1.9809859154929576e-05, "loss": 1.3161, "step": 8240 }, { "epoch": 9.24, "learning_rate": 1.9808450704225352e-05, "loss": 1.3226, "step": 8241 }, { "epoch": 9.24, "learning_rate": 1.9807042253521125e-05, "loss": 1.4377, "step": 8242 }, { "epoch": 9.24, "learning_rate": 1.98056338028169e-05, "loss": 1.5401, "step": 8243 }, { "epoch": 9.24, "learning_rate": 1.9804225352112678e-05, "loss": 1.5611, "step": 8244 }, { "epoch": 9.24, "learning_rate": 1.980281690140845e-05, "loss": 1.617, "step": 8245 }, { "epoch": 9.24, "learning_rate": 1.9801408450704227e-05, "loss": 1.2942, "step": 8246 }, { "epoch": 9.25, "learning_rate": 1.98e-05, "loss": 1.5551, "step": 8247 }, { "epoch": 9.25, "learning_rate": 1.9798591549295776e-05, "loss": 1.6091, "step": 8248 }, { "epoch": 9.25, "learning_rate": 1.979718309859155e-05, "loss": 1.47, "step": 8249 }, { "epoch": 9.25, "learning_rate": 1.9795774647887326e-05, "loss": 1.4409, "step": 8250 }, { "epoch": 9.25, "learning_rate": 1.97943661971831e-05, "loss": 1.5022, "step": 8251 }, { "epoch": 9.25, "learning_rate": 1.9792957746478875e-05, "loss": 1.5528, "step": 8252 }, { "epoch": 9.25, "learning_rate": 1.9791549295774648e-05, "loss": 1.7078, "step": 8253 }, { "epoch": 9.25, "learning_rate": 1.979014084507042e-05, "loss": 1.6896, "step": 8254 }, { "epoch": 9.25, "learning_rate": 1.9788732394366197e-05, "loss": 1.4777, "step": 8255 }, { "epoch": 9.26, "learning_rate": 1.978732394366197e-05, "loss": 1.4768, "step": 8256 }, { "epoch": 9.26, "learning_rate": 1.9785915492957747e-05, "loss": 1.378, "step": 8257 }, { "epoch": 9.26, "learning_rate": 1.978450704225352e-05, "loss": 1.4038, "step": 8258 }, { "epoch": 9.26, "learning_rate": 1.97830985915493e-05, "loss": 1.6871, "step": 8259 }, { "epoch": 9.26, "learning_rate": 1.9781690140845072e-05, "loss": 1.4788, "step": 8260 }, { "epoch": 9.26, "learning_rate": 1.9780281690140845e-05, "loss": 1.499, "step": 8261 }, { "epoch": 9.26, "learning_rate": 1.977887323943662e-05, "loss": 1.3768, "step": 8262 }, { "epoch": 9.26, "learning_rate": 1.9777464788732394e-05, "loss": 1.3321, "step": 8263 }, { "epoch": 9.26, "learning_rate": 1.977605633802817e-05, "loss": 1.4302, "step": 8264 }, { "epoch": 9.27, "learning_rate": 1.9774647887323944e-05, "loss": 1.6654, "step": 8265 }, { "epoch": 9.27, "learning_rate": 1.977323943661972e-05, "loss": 1.4488, "step": 8266 }, { "epoch": 9.27, "learning_rate": 1.9771830985915493e-05, "loss": 1.2135, "step": 8267 }, { "epoch": 9.27, "learning_rate": 1.977042253521127e-05, "loss": 1.1686, "step": 8268 }, { "epoch": 9.27, "learning_rate": 1.9769014084507042e-05, "loss": 1.4404, "step": 8269 }, { "epoch": 9.27, "learning_rate": 1.9767605633802815e-05, "loss": 1.44, "step": 8270 }, { "epoch": 9.27, "learning_rate": 1.976619718309859e-05, "loss": 1.4353, "step": 8271 }, { "epoch": 9.27, "learning_rate": 1.9764788732394364e-05, "loss": 1.2702, "step": 8272 }, { "epoch": 9.27, "learning_rate": 1.976338028169014e-05, "loss": 1.4258, "step": 8273 }, { "epoch": 9.28, "learning_rate": 1.9761971830985917e-05, "loss": 1.2186, "step": 8274 }, { "epoch": 9.28, "learning_rate": 1.9760563380281693e-05, "loss": 1.2694, "step": 8275 }, { "epoch": 9.28, "learning_rate": 1.9759154929577466e-05, "loss": 1.1684, "step": 8276 }, { "epoch": 9.28, "learning_rate": 1.975774647887324e-05, "loss": 1.3411, "step": 8277 }, { "epoch": 9.28, "learning_rate": 1.9756338028169016e-05, "loss": 1.2089, "step": 8278 }, { "epoch": 9.28, "learning_rate": 1.975492957746479e-05, "loss": 1.9976, "step": 8279 }, { "epoch": 9.28, "learning_rate": 1.9753521126760565e-05, "loss": 1.6192, "step": 8280 }, { "epoch": 9.28, "learning_rate": 1.9752112676056338e-05, "loss": 1.9391, "step": 8281 }, { "epoch": 9.28, "learning_rate": 1.9750704225352114e-05, "loss": 1.6009, "step": 8282 }, { "epoch": 9.29, "learning_rate": 1.9749295774647887e-05, "loss": 1.4844, "step": 8283 }, { "epoch": 9.29, "learning_rate": 1.9747887323943663e-05, "loss": 1.5405, "step": 8284 }, { "epoch": 9.29, "learning_rate": 1.9746478873239436e-05, "loss": 1.6879, "step": 8285 }, { "epoch": 9.29, "learning_rate": 1.974507042253521e-05, "loss": 1.5016, "step": 8286 }, { "epoch": 9.29, "learning_rate": 1.9743661971830986e-05, "loss": 1.5665, "step": 8287 }, { "epoch": 9.29, "learning_rate": 1.974225352112676e-05, "loss": 1.3185, "step": 8288 }, { "epoch": 9.29, "learning_rate": 1.9740845070422535e-05, "loss": 1.5502, "step": 8289 }, { "epoch": 9.29, "learning_rate": 1.973943661971831e-05, "loss": 1.6821, "step": 8290 }, { "epoch": 9.29, "learning_rate": 1.9738028169014088e-05, "loss": 1.4686, "step": 8291 }, { "epoch": 9.3, "learning_rate": 1.973661971830986e-05, "loss": 1.688, "step": 8292 }, { "epoch": 9.3, "learning_rate": 1.9735211267605633e-05, "loss": 1.6719, "step": 8293 }, { "epoch": 9.3, "learning_rate": 1.973380281690141e-05, "loss": 1.4457, "step": 8294 }, { "epoch": 9.3, "learning_rate": 1.9732394366197183e-05, "loss": 1.2574, "step": 8295 }, { "epoch": 9.3, "learning_rate": 1.973098591549296e-05, "loss": 1.6056, "step": 8296 }, { "epoch": 9.3, "learning_rate": 1.9729577464788732e-05, "loss": 1.4842, "step": 8297 }, { "epoch": 9.3, "learning_rate": 1.972816901408451e-05, "loss": 1.4853, "step": 8298 }, { "epoch": 9.3, "learning_rate": 1.972676056338028e-05, "loss": 1.3312, "step": 8299 }, { "epoch": 9.3, "learning_rate": 1.9725352112676058e-05, "loss": 1.6399, "step": 8300 }, { "epoch": 9.31, "learning_rate": 1.972394366197183e-05, "loss": 1.4423, "step": 8301 }, { "epoch": 9.31, "learning_rate": 1.9722535211267604e-05, "loss": 1.5433, "step": 8302 }, { "epoch": 9.31, "learning_rate": 1.972112676056338e-05, "loss": 1.4427, "step": 8303 }, { "epoch": 9.31, "learning_rate": 1.9719718309859153e-05, "loss": 1.4578, "step": 8304 }, { "epoch": 9.31, "learning_rate": 1.9718309859154933e-05, "loss": 1.3631, "step": 8305 }, { "epoch": 9.31, "learning_rate": 1.9716901408450705e-05, "loss": 1.5115, "step": 8306 }, { "epoch": 9.31, "learning_rate": 1.9715492957746482e-05, "loss": 1.5931, "step": 8307 }, { "epoch": 9.31, "learning_rate": 1.9714084507042255e-05, "loss": 1.6422, "step": 8308 }, { "epoch": 9.32, "learning_rate": 1.9712676056338028e-05, "loss": 1.5051, "step": 8309 }, { "epoch": 9.32, "learning_rate": 1.9711267605633804e-05, "loss": 1.13, "step": 8310 }, { "epoch": 9.32, "learning_rate": 1.9709859154929577e-05, "loss": 1.3231, "step": 8311 }, { "epoch": 9.32, "learning_rate": 1.9708450704225353e-05, "loss": 1.3466, "step": 8312 }, { "epoch": 9.32, "learning_rate": 1.9707042253521126e-05, "loss": 1.5529, "step": 8313 }, { "epoch": 9.32, "learning_rate": 1.9705633802816903e-05, "loss": 1.2844, "step": 8314 }, { "epoch": 9.32, "learning_rate": 1.9704225352112676e-05, "loss": 1.3336, "step": 8315 }, { "epoch": 9.32, "learning_rate": 1.9702816901408452e-05, "loss": 1.2069, "step": 8316 }, { "epoch": 9.32, "learning_rate": 1.9701408450704225e-05, "loss": 1.3412, "step": 8317 }, { "epoch": 9.33, "learning_rate": 1.9699999999999998e-05, "loss": 1.4256, "step": 8318 }, { "epoch": 9.33, "learning_rate": 1.9698591549295774e-05, "loss": 1.5301, "step": 8319 }, { "epoch": 9.33, "learning_rate": 1.969718309859155e-05, "loss": 1.438, "step": 8320 }, { "epoch": 9.33, "learning_rate": 1.9695774647887327e-05, "loss": 1.2503, "step": 8321 }, { "epoch": 9.33, "learning_rate": 1.96943661971831e-05, "loss": 1.4089, "step": 8322 }, { "epoch": 9.33, "learning_rate": 1.9692957746478876e-05, "loss": 1.2835, "step": 8323 }, { "epoch": 9.33, "learning_rate": 1.969154929577465e-05, "loss": 1.6125, "step": 8324 }, { "epoch": 9.33, "learning_rate": 1.9690140845070422e-05, "loss": 1.1565, "step": 8325 }, { "epoch": 9.33, "learning_rate": 1.9688732394366198e-05, "loss": 1.4361, "step": 8326 }, { "epoch": 9.34, "learning_rate": 1.968732394366197e-05, "loss": 1.1708, "step": 8327 }, { "epoch": 9.34, "learning_rate": 1.9685915492957747e-05, "loss": 1.6319, "step": 8328 }, { "epoch": 9.34, "learning_rate": 1.968450704225352e-05, "loss": 2.005, "step": 8329 }, { "epoch": 9.34, "learning_rate": 1.9683098591549297e-05, "loss": 1.8215, "step": 8330 }, { "epoch": 9.34, "learning_rate": 1.968169014084507e-05, "loss": 1.8311, "step": 8331 }, { "epoch": 9.34, "learning_rate": 1.9680281690140846e-05, "loss": 1.4853, "step": 8332 }, { "epoch": 9.34, "learning_rate": 1.967887323943662e-05, "loss": 1.7019, "step": 8333 }, { "epoch": 9.34, "learning_rate": 1.9677464788732392e-05, "loss": 1.334, "step": 8334 }, { "epoch": 9.34, "learning_rate": 1.967605633802817e-05, "loss": 1.6293, "step": 8335 }, { "epoch": 9.35, "learning_rate": 1.9674647887323945e-05, "loss": 1.319, "step": 8336 }, { "epoch": 9.35, "learning_rate": 1.967323943661972e-05, "loss": 1.6137, "step": 8337 }, { "epoch": 9.35, "learning_rate": 1.9671830985915494e-05, "loss": 1.4042, "step": 8338 }, { "epoch": 9.35, "learning_rate": 1.967042253521127e-05, "loss": 1.4752, "step": 8339 }, { "epoch": 9.35, "learning_rate": 1.9669014084507043e-05, "loss": 1.4577, "step": 8340 }, { "epoch": 9.35, "learning_rate": 1.9667605633802816e-05, "loss": 1.4751, "step": 8341 }, { "epoch": 9.35, "learning_rate": 1.9666197183098592e-05, "loss": 1.5455, "step": 8342 }, { "epoch": 9.35, "learning_rate": 1.9664788732394365e-05, "loss": 1.4564, "step": 8343 }, { "epoch": 9.35, "learning_rate": 1.9663380281690142e-05, "loss": 1.3813, "step": 8344 }, { "epoch": 9.36, "learning_rate": 1.9661971830985915e-05, "loss": 1.5497, "step": 8345 }, { "epoch": 9.36, "learning_rate": 1.966056338028169e-05, "loss": 1.2689, "step": 8346 }, { "epoch": 9.36, "learning_rate": 1.9659154929577464e-05, "loss": 1.4535, "step": 8347 }, { "epoch": 9.36, "learning_rate": 1.965774647887324e-05, "loss": 1.4373, "step": 8348 }, { "epoch": 9.36, "learning_rate": 1.9656338028169013e-05, "loss": 1.2827, "step": 8349 }, { "epoch": 9.36, "learning_rate": 1.965492957746479e-05, "loss": 1.4609, "step": 8350 }, { "epoch": 9.36, "learning_rate": 1.9653521126760566e-05, "loss": 1.4238, "step": 8351 }, { "epoch": 9.36, "learning_rate": 1.965211267605634e-05, "loss": 1.3052, "step": 8352 }, { "epoch": 9.36, "learning_rate": 1.9650704225352115e-05, "loss": 1.36, "step": 8353 }, { "epoch": 9.37, "learning_rate": 1.9649295774647888e-05, "loss": 1.4444, "step": 8354 }, { "epoch": 9.37, "learning_rate": 1.9647887323943664e-05, "loss": 1.4877, "step": 8355 }, { "epoch": 9.37, "learning_rate": 1.9646478873239437e-05, "loss": 1.332, "step": 8356 }, { "epoch": 9.37, "learning_rate": 1.964507042253521e-05, "loss": 1.5029, "step": 8357 }, { "epoch": 9.37, "learning_rate": 1.9643661971830987e-05, "loss": 1.2788, "step": 8358 }, { "epoch": 9.37, "learning_rate": 1.964225352112676e-05, "loss": 1.5543, "step": 8359 }, { "epoch": 9.37, "learning_rate": 1.9640845070422536e-05, "loss": 1.4939, "step": 8360 }, { "epoch": 9.37, "learning_rate": 1.963943661971831e-05, "loss": 1.2668, "step": 8361 }, { "epoch": 9.37, "learning_rate": 1.9638028169014085e-05, "loss": 1.356, "step": 8362 }, { "epoch": 9.38, "learning_rate": 1.9636619718309858e-05, "loss": 1.6373, "step": 8363 }, { "epoch": 9.38, "learning_rate": 1.9635211267605634e-05, "loss": 1.6231, "step": 8364 }, { "epoch": 9.38, "learning_rate": 1.963380281690141e-05, "loss": 1.3733, "step": 8365 }, { "epoch": 9.38, "learning_rate": 1.9632394366197184e-05, "loss": 1.3112, "step": 8366 }, { "epoch": 9.38, "learning_rate": 1.963098591549296e-05, "loss": 1.3649, "step": 8367 }, { "epoch": 9.38, "learning_rate": 1.9629577464788733e-05, "loss": 1.3248, "step": 8368 }, { "epoch": 9.38, "learning_rate": 1.962816901408451e-05, "loss": 1.6341, "step": 8369 }, { "epoch": 9.38, "learning_rate": 1.9626760563380282e-05, "loss": 1.4723, "step": 8370 }, { "epoch": 9.38, "learning_rate": 1.962535211267606e-05, "loss": 1.2921, "step": 8371 }, { "epoch": 9.39, "learning_rate": 1.962394366197183e-05, "loss": 1.2097, "step": 8372 }, { "epoch": 9.39, "learning_rate": 1.9622535211267604e-05, "loss": 1.3494, "step": 8373 }, { "epoch": 9.39, "learning_rate": 1.962112676056338e-05, "loss": 1.3372, "step": 8374 }, { "epoch": 9.39, "learning_rate": 1.9619718309859154e-05, "loss": 1.2577, "step": 8375 }, { "epoch": 9.39, "learning_rate": 1.961830985915493e-05, "loss": 1.4171, "step": 8376 }, { "epoch": 9.39, "learning_rate": 1.9616901408450703e-05, "loss": 1.3532, "step": 8377 }, { "epoch": 9.39, "learning_rate": 1.961549295774648e-05, "loss": 1.2255, "step": 8378 }, { "epoch": 9.39, "learning_rate": 1.9614084507042252e-05, "loss": 2.0285, "step": 8379 }, { "epoch": 9.39, "learning_rate": 1.961267605633803e-05, "loss": 1.7786, "step": 8380 }, { "epoch": 9.4, "learning_rate": 1.9611267605633805e-05, "loss": 1.5476, "step": 8381 }, { "epoch": 9.4, "learning_rate": 1.9609859154929578e-05, "loss": 1.75, "step": 8382 }, { "epoch": 9.4, "learning_rate": 1.9608450704225354e-05, "loss": 1.5925, "step": 8383 }, { "epoch": 9.4, "learning_rate": 1.9607042253521127e-05, "loss": 1.5499, "step": 8384 }, { "epoch": 9.4, "learning_rate": 1.9605633802816904e-05, "loss": 1.3868, "step": 8385 }, { "epoch": 9.4, "learning_rate": 1.9604225352112676e-05, "loss": 1.3449, "step": 8386 }, { "epoch": 9.4, "learning_rate": 1.9602816901408453e-05, "loss": 1.5301, "step": 8387 }, { "epoch": 9.4, "learning_rate": 1.9601408450704226e-05, "loss": 1.4656, "step": 8388 }, { "epoch": 9.4, "learning_rate": 1.96e-05, "loss": 1.606, "step": 8389 }, { "epoch": 9.41, "learning_rate": 1.9598591549295775e-05, "loss": 1.5543, "step": 8390 }, { "epoch": 9.41, "learning_rate": 1.9597183098591548e-05, "loss": 1.6863, "step": 8391 }, { "epoch": 9.41, "learning_rate": 1.9595774647887324e-05, "loss": 1.8482, "step": 8392 }, { "epoch": 9.41, "learning_rate": 1.9594366197183097e-05, "loss": 1.3804, "step": 8393 }, { "epoch": 9.41, "learning_rate": 1.9592957746478874e-05, "loss": 1.5685, "step": 8394 }, { "epoch": 9.41, "learning_rate": 1.959154929577465e-05, "loss": 1.2646, "step": 8395 }, { "epoch": 9.41, "learning_rate": 1.9590140845070423e-05, "loss": 1.248, "step": 8396 }, { "epoch": 9.41, "learning_rate": 1.95887323943662e-05, "loss": 1.5046, "step": 8397 }, { "epoch": 9.41, "learning_rate": 1.9587323943661972e-05, "loss": 1.3885, "step": 8398 }, { "epoch": 9.42, "learning_rate": 1.958591549295775e-05, "loss": 1.573, "step": 8399 }, { "epoch": 9.42, "learning_rate": 1.958450704225352e-05, "loss": 1.5613, "step": 8400 }, { "epoch": 9.42, "learning_rate": 1.9583098591549298e-05, "loss": 1.3897, "step": 8401 }, { "epoch": 9.42, "learning_rate": 1.958169014084507e-05, "loss": 1.3921, "step": 8402 }, { "epoch": 9.42, "learning_rate": 1.9580281690140847e-05, "loss": 1.4026, "step": 8403 }, { "epoch": 9.42, "learning_rate": 1.957887323943662e-05, "loss": 1.5873, "step": 8404 }, { "epoch": 9.42, "learning_rate": 1.9577464788732393e-05, "loss": 1.5166, "step": 8405 }, { "epoch": 9.42, "learning_rate": 1.957605633802817e-05, "loss": 1.2287, "step": 8406 }, { "epoch": 9.42, "learning_rate": 1.9574647887323942e-05, "loss": 1.4712, "step": 8407 }, { "epoch": 9.43, "learning_rate": 1.957323943661972e-05, "loss": 1.5198, "step": 8408 }, { "epoch": 9.43, "learning_rate": 1.957183098591549e-05, "loss": 1.4896, "step": 8409 }, { "epoch": 9.43, "learning_rate": 1.9570422535211268e-05, "loss": 1.3645, "step": 8410 }, { "epoch": 9.43, "learning_rate": 1.9569014084507044e-05, "loss": 1.4983, "step": 8411 }, { "epoch": 9.43, "learning_rate": 1.9567605633802817e-05, "loss": 1.4136, "step": 8412 }, { "epoch": 9.43, "learning_rate": 1.9566197183098593e-05, "loss": 1.3508, "step": 8413 }, { "epoch": 9.43, "learning_rate": 1.9564788732394366e-05, "loss": 1.434, "step": 8414 }, { "epoch": 9.43, "learning_rate": 1.9563380281690143e-05, "loss": 1.1904, "step": 8415 }, { "epoch": 9.43, "learning_rate": 1.9561971830985916e-05, "loss": 1.3264, "step": 8416 }, { "epoch": 9.44, "learning_rate": 1.9560563380281692e-05, "loss": 1.4277, "step": 8417 }, { "epoch": 9.44, "learning_rate": 1.9559154929577465e-05, "loss": 1.0933, "step": 8418 }, { "epoch": 9.44, "learning_rate": 1.955774647887324e-05, "loss": 1.2854, "step": 8419 }, { "epoch": 9.44, "learning_rate": 1.9556338028169014e-05, "loss": 1.5825, "step": 8420 }, { "epoch": 9.44, "learning_rate": 1.9554929577464787e-05, "loss": 1.4633, "step": 8421 }, { "epoch": 9.44, "learning_rate": 1.9553521126760563e-05, "loss": 1.3658, "step": 8422 }, { "epoch": 9.44, "learning_rate": 1.9552112676056336e-05, "loss": 1.2959, "step": 8423 }, { "epoch": 9.44, "learning_rate": 1.9550704225352113e-05, "loss": 1.2547, "step": 8424 }, { "epoch": 9.45, "learning_rate": 1.9549295774647886e-05, "loss": 1.2712, "step": 8425 }, { "epoch": 9.45, "learning_rate": 1.9547887323943665e-05, "loss": 1.1258, "step": 8426 }, { "epoch": 9.45, "learning_rate": 1.9546478873239438e-05, "loss": 1.4752, "step": 8427 }, { "epoch": 9.45, "learning_rate": 1.954507042253521e-05, "loss": 1.3982, "step": 8428 }, { "epoch": 9.45, "learning_rate": 1.9543661971830988e-05, "loss": 1.9598, "step": 8429 }, { "epoch": 9.45, "learning_rate": 1.954225352112676e-05, "loss": 1.8341, "step": 8430 }, { "epoch": 9.45, "learning_rate": 1.9540845070422537e-05, "loss": 1.7447, "step": 8431 }, { "epoch": 9.45, "learning_rate": 1.953943661971831e-05, "loss": 1.6453, "step": 8432 }, { "epoch": 9.45, "learning_rate": 1.9538028169014086e-05, "loss": 1.5986, "step": 8433 }, { "epoch": 9.46, "learning_rate": 1.953661971830986e-05, "loss": 1.5802, "step": 8434 }, { "epoch": 9.46, "learning_rate": 1.9535211267605635e-05, "loss": 1.4965, "step": 8435 }, { "epoch": 9.46, "learning_rate": 1.953380281690141e-05, "loss": 1.5922, "step": 8436 }, { "epoch": 9.46, "learning_rate": 1.953239436619718e-05, "loss": 1.5711, "step": 8437 }, { "epoch": 9.46, "learning_rate": 1.9530985915492958e-05, "loss": 1.3812, "step": 8438 }, { "epoch": 9.46, "learning_rate": 1.952957746478873e-05, "loss": 1.6344, "step": 8439 }, { "epoch": 9.46, "learning_rate": 1.9528169014084507e-05, "loss": 1.4957, "step": 8440 }, { "epoch": 9.46, "learning_rate": 1.9526760563380283e-05, "loss": 1.5369, "step": 8441 }, { "epoch": 9.46, "learning_rate": 1.952535211267606e-05, "loss": 1.5379, "step": 8442 }, { "epoch": 9.47, "learning_rate": 1.9523943661971833e-05, "loss": 1.3298, "step": 8443 }, { "epoch": 9.47, "learning_rate": 1.9522535211267605e-05, "loss": 1.456, "step": 8444 }, { "epoch": 9.47, "learning_rate": 1.9521126760563382e-05, "loss": 1.5254, "step": 8445 }, { "epoch": 9.47, "learning_rate": 1.9519718309859155e-05, "loss": 1.6763, "step": 8446 }, { "epoch": 9.47, "learning_rate": 1.951830985915493e-05, "loss": 1.3363, "step": 8447 }, { "epoch": 9.47, "learning_rate": 1.9516901408450704e-05, "loss": 1.5441, "step": 8448 }, { "epoch": 9.47, "learning_rate": 1.951549295774648e-05, "loss": 1.6167, "step": 8449 }, { "epoch": 9.47, "learning_rate": 1.9514084507042253e-05, "loss": 1.3493, "step": 8450 }, { "epoch": 9.47, "learning_rate": 1.951267605633803e-05, "loss": 1.4945, "step": 8451 }, { "epoch": 9.48, "learning_rate": 1.9511267605633803e-05, "loss": 1.3386, "step": 8452 }, { "epoch": 9.48, "learning_rate": 1.9509859154929575e-05, "loss": 1.553, "step": 8453 }, { "epoch": 9.48, "learning_rate": 1.9508450704225352e-05, "loss": 1.4111, "step": 8454 }, { "epoch": 9.48, "learning_rate": 1.9507042253521125e-05, "loss": 1.4757, "step": 8455 }, { "epoch": 9.48, "learning_rate": 1.9505633802816904e-05, "loss": 1.3688, "step": 8456 }, { "epoch": 9.48, "learning_rate": 1.9504225352112677e-05, "loss": 1.2568, "step": 8457 }, { "epoch": 9.48, "learning_rate": 1.9502816901408454e-05, "loss": 1.4215, "step": 8458 }, { "epoch": 9.48, "learning_rate": 1.9501408450704227e-05, "loss": 1.3186, "step": 8459 }, { "epoch": 9.48, "learning_rate": 1.95e-05, "loss": 1.2821, "step": 8460 }, { "epoch": 9.49, "learning_rate": 1.9498591549295776e-05, "loss": 1.3389, "step": 8461 }, { "epoch": 9.49, "learning_rate": 1.949718309859155e-05, "loss": 1.1702, "step": 8462 }, { "epoch": 9.49, "learning_rate": 1.9495774647887325e-05, "loss": 1.6157, "step": 8463 }, { "epoch": 9.49, "learning_rate": 1.9494366197183098e-05, "loss": 1.4263, "step": 8464 }, { "epoch": 9.49, "learning_rate": 1.9492957746478875e-05, "loss": 1.3143, "step": 8465 }, { "epoch": 9.49, "learning_rate": 1.9491549295774647e-05, "loss": 1.2279, "step": 8466 }, { "epoch": 9.49, "learning_rate": 1.9490140845070424e-05, "loss": 1.2777, "step": 8467 }, { "epoch": 9.49, "learning_rate": 1.9488732394366197e-05, "loss": 1.1021, "step": 8468 }, { "epoch": 9.49, "learning_rate": 1.948732394366197e-05, "loss": 1.3874, "step": 8469 }, { "epoch": 9.5, "learning_rate": 1.9485915492957746e-05, "loss": 1.1715, "step": 8470 }, { "epoch": 9.5, "learning_rate": 1.9484507042253522e-05, "loss": 1.2232, "step": 8471 }, { "epoch": 9.5, "learning_rate": 1.94830985915493e-05, "loss": 1.3755, "step": 8472 }, { "epoch": 9.5, "learning_rate": 1.948169014084507e-05, "loss": 1.1577, "step": 8473 }, { "epoch": 9.5, "learning_rate": 1.9480281690140848e-05, "loss": 1.3543, "step": 8474 }, { "epoch": 9.5, "learning_rate": 1.947887323943662e-05, "loss": 1.1369, "step": 8475 }, { "epoch": 9.5, "learning_rate": 1.9477464788732394e-05, "loss": 1.4234, "step": 8476 }, { "epoch": 9.5, "learning_rate": 1.947605633802817e-05, "loss": 1.5498, "step": 8477 }, { "epoch": 9.5, "learning_rate": 1.9474647887323943e-05, "loss": 1.4319, "step": 8478 }, { "epoch": 9.51, "learning_rate": 1.947323943661972e-05, "loss": 1.6767, "step": 8479 }, { "epoch": 9.51, "learning_rate": 1.9471830985915492e-05, "loss": 1.5494, "step": 8480 }, { "epoch": 9.51, "learning_rate": 1.947042253521127e-05, "loss": 1.6372, "step": 8481 }, { "epoch": 9.51, "learning_rate": 1.946901408450704e-05, "loss": 1.5183, "step": 8482 }, { "epoch": 9.51, "learning_rate": 1.9467605633802818e-05, "loss": 1.6457, "step": 8483 }, { "epoch": 9.51, "learning_rate": 1.946619718309859e-05, "loss": 1.6103, "step": 8484 }, { "epoch": 9.51, "learning_rate": 1.9464788732394364e-05, "loss": 1.4994, "step": 8485 }, { "epoch": 9.51, "learning_rate": 1.9463380281690144e-05, "loss": 1.6203, "step": 8486 }, { "epoch": 9.51, "learning_rate": 1.9461971830985917e-05, "loss": 1.3255, "step": 8487 }, { "epoch": 9.52, "learning_rate": 1.9460563380281693e-05, "loss": 1.4851, "step": 8488 }, { "epoch": 9.52, "learning_rate": 1.9459154929577466e-05, "loss": 1.4945, "step": 8489 }, { "epoch": 9.52, "learning_rate": 1.9457746478873242e-05, "loss": 1.6554, "step": 8490 }, { "epoch": 9.52, "learning_rate": 1.9456338028169015e-05, "loss": 1.4205, "step": 8491 }, { "epoch": 9.52, "learning_rate": 1.9454929577464788e-05, "loss": 1.1945, "step": 8492 }, { "epoch": 9.52, "learning_rate": 1.9453521126760564e-05, "loss": 1.4877, "step": 8493 }, { "epoch": 9.52, "learning_rate": 1.9452112676056337e-05, "loss": 1.4001, "step": 8494 }, { "epoch": 9.52, "learning_rate": 1.9450704225352114e-05, "loss": 1.4474, "step": 8495 }, { "epoch": 9.52, "learning_rate": 1.9449295774647887e-05, "loss": 1.4127, "step": 8496 }, { "epoch": 9.53, "learning_rate": 1.9447887323943663e-05, "loss": 1.3342, "step": 8497 }, { "epoch": 9.53, "learning_rate": 1.9446478873239436e-05, "loss": 1.3872, "step": 8498 }, { "epoch": 9.53, "learning_rate": 1.944507042253521e-05, "loss": 1.2469, "step": 8499 }, { "epoch": 9.53, "learning_rate": 1.9443661971830985e-05, "loss": 1.3629, "step": 8500 }, { "epoch": 9.53, "learning_rate": 1.944225352112676e-05, "loss": 1.4361, "step": 8501 }, { "epoch": 9.53, "learning_rate": 1.9440845070422538e-05, "loss": 1.3841, "step": 8502 }, { "epoch": 9.53, "learning_rate": 1.943943661971831e-05, "loss": 1.471, "step": 8503 }, { "epoch": 9.53, "learning_rate": 1.9438028169014087e-05, "loss": 1.4267, "step": 8504 }, { "epoch": 9.53, "learning_rate": 1.943661971830986e-05, "loss": 1.1438, "step": 8505 }, { "epoch": 9.54, "learning_rate": 1.9435211267605636e-05, "loss": 1.473, "step": 8506 }, { "epoch": 9.54, "learning_rate": 1.943380281690141e-05, "loss": 1.3933, "step": 8507 }, { "epoch": 9.54, "learning_rate": 1.9432394366197182e-05, "loss": 1.3678, "step": 8508 }, { "epoch": 9.54, "learning_rate": 1.943098591549296e-05, "loss": 1.3037, "step": 8509 }, { "epoch": 9.54, "learning_rate": 1.942957746478873e-05, "loss": 1.3786, "step": 8510 }, { "epoch": 9.54, "learning_rate": 1.9428169014084508e-05, "loss": 1.171, "step": 8511 }, { "epoch": 9.54, "learning_rate": 1.942676056338028e-05, "loss": 1.205, "step": 8512 }, { "epoch": 9.54, "learning_rate": 1.9425352112676057e-05, "loss": 1.3856, "step": 8513 }, { "epoch": 9.54, "learning_rate": 1.942394366197183e-05, "loss": 1.2886, "step": 8514 }, { "epoch": 9.55, "learning_rate": 1.9422535211267603e-05, "loss": 1.5192, "step": 8515 }, { "epoch": 9.55, "learning_rate": 1.9421126760563383e-05, "loss": 1.3452, "step": 8516 }, { "epoch": 9.55, "learning_rate": 1.9419718309859156e-05, "loss": 1.1066, "step": 8517 }, { "epoch": 9.55, "learning_rate": 1.9418309859154932e-05, "loss": 1.2419, "step": 8518 }, { "epoch": 9.55, "learning_rate": 1.9416901408450705e-05, "loss": 1.1209, "step": 8519 }, { "epoch": 9.55, "learning_rate": 1.941549295774648e-05, "loss": 1.4795, "step": 8520 }, { "epoch": 9.55, "learning_rate": 1.9414084507042254e-05, "loss": 1.6565, "step": 8521 }, { "epoch": 9.55, "learning_rate": 1.941267605633803e-05, "loss": 1.1248, "step": 8522 }, { "epoch": 9.55, "learning_rate": 1.9411267605633804e-05, "loss": 1.7139, "step": 8523 }, { "epoch": 9.56, "learning_rate": 1.9409859154929576e-05, "loss": 1.4436, "step": 8524 }, { "epoch": 9.56, "learning_rate": 1.9408450704225353e-05, "loss": 1.3653, "step": 8525 }, { "epoch": 9.56, "learning_rate": 1.9407042253521126e-05, "loss": 1.3852, "step": 8526 }, { "epoch": 9.56, "learning_rate": 1.9405633802816902e-05, "loss": 1.0893, "step": 8527 }, { "epoch": 9.56, "learning_rate": 1.9404225352112675e-05, "loss": 1.8018, "step": 8528 }, { "epoch": 9.56, "learning_rate": 1.940281690140845e-05, "loss": 1.6587, "step": 8529 }, { "epoch": 9.56, "learning_rate": 1.9401408450704224e-05, "loss": 1.6887, "step": 8530 }, { "epoch": 9.56, "learning_rate": 1.9399999999999997e-05, "loss": 1.4775, "step": 8531 }, { "epoch": 9.57, "learning_rate": 1.9398591549295777e-05, "loss": 1.5307, "step": 8532 }, { "epoch": 9.57, "learning_rate": 1.939718309859155e-05, "loss": 1.4164, "step": 8533 }, { "epoch": 9.57, "learning_rate": 1.9395774647887326e-05, "loss": 1.3292, "step": 8534 }, { "epoch": 9.57, "learning_rate": 1.93943661971831e-05, "loss": 1.539, "step": 8535 }, { "epoch": 9.57, "learning_rate": 1.9392957746478875e-05, "loss": 1.578, "step": 8536 }, { "epoch": 9.57, "learning_rate": 1.939154929577465e-05, "loss": 1.714, "step": 8537 }, { "epoch": 9.57, "learning_rate": 1.9390140845070425e-05, "loss": 1.4492, "step": 8538 }, { "epoch": 9.57, "learning_rate": 1.9388732394366198e-05, "loss": 1.4678, "step": 8539 }, { "epoch": 9.57, "learning_rate": 1.938732394366197e-05, "loss": 1.2978, "step": 8540 }, { "epoch": 9.58, "learning_rate": 1.9385915492957747e-05, "loss": 1.4903, "step": 8541 }, { "epoch": 9.58, "learning_rate": 1.938450704225352e-05, "loss": 1.378, "step": 8542 }, { "epoch": 9.58, "learning_rate": 1.9383098591549296e-05, "loss": 1.2561, "step": 8543 }, { "epoch": 9.58, "learning_rate": 1.938169014084507e-05, "loss": 1.5879, "step": 8544 }, { "epoch": 9.58, "learning_rate": 1.9380281690140846e-05, "loss": 1.5837, "step": 8545 }, { "epoch": 9.58, "learning_rate": 1.937887323943662e-05, "loss": 1.4159, "step": 8546 }, { "epoch": 9.58, "learning_rate": 1.9377464788732395e-05, "loss": 1.3545, "step": 8547 }, { "epoch": 9.58, "learning_rate": 1.937605633802817e-05, "loss": 1.6052, "step": 8548 }, { "epoch": 9.58, "learning_rate": 1.9374647887323944e-05, "loss": 1.2752, "step": 8549 }, { "epoch": 9.59, "learning_rate": 1.937323943661972e-05, "loss": 1.3044, "step": 8550 }, { "epoch": 9.59, "learning_rate": 1.9371830985915493e-05, "loss": 1.1169, "step": 8551 }, { "epoch": 9.59, "learning_rate": 1.937042253521127e-05, "loss": 1.0027, "step": 8552 }, { "epoch": 9.59, "learning_rate": 1.9369014084507043e-05, "loss": 1.4489, "step": 8553 }, { "epoch": 9.59, "learning_rate": 1.936760563380282e-05, "loss": 1.2424, "step": 8554 }, { "epoch": 9.59, "learning_rate": 1.9366197183098592e-05, "loss": 1.3672, "step": 8555 }, { "epoch": 9.59, "learning_rate": 1.9364788732394365e-05, "loss": 1.4173, "step": 8556 }, { "epoch": 9.59, "learning_rate": 1.936338028169014e-05, "loss": 1.4097, "step": 8557 }, { "epoch": 9.59, "learning_rate": 1.9361971830985914e-05, "loss": 1.3046, "step": 8558 }, { "epoch": 9.6, "learning_rate": 1.936056338028169e-05, "loss": 1.3221, "step": 8559 }, { "epoch": 9.6, "learning_rate": 1.9359154929577463e-05, "loss": 1.4421, "step": 8560 }, { "epoch": 9.6, "learning_rate": 1.935774647887324e-05, "loss": 1.2523, "step": 8561 }, { "epoch": 9.6, "learning_rate": 1.9356338028169016e-05, "loss": 1.3528, "step": 8562 }, { "epoch": 9.6, "learning_rate": 1.935492957746479e-05, "loss": 1.3601, "step": 8563 }, { "epoch": 9.6, "learning_rate": 1.9353521126760565e-05, "loss": 1.4986, "step": 8564 }, { "epoch": 9.6, "learning_rate": 1.9352112676056338e-05, "loss": 1.3205, "step": 8565 }, { "epoch": 9.6, "learning_rate": 1.9350704225352115e-05, "loss": 1.4494, "step": 8566 }, { "epoch": 9.6, "learning_rate": 1.9349295774647888e-05, "loss": 1.3105, "step": 8567 }, { "epoch": 9.61, "learning_rate": 1.9347887323943664e-05, "loss": 1.3315, "step": 8568 }, { "epoch": 9.61, "learning_rate": 1.9346478873239437e-05, "loss": 1.3962, "step": 8569 }, { "epoch": 9.61, "learning_rate": 1.9345070422535213e-05, "loss": 1.418, "step": 8570 }, { "epoch": 9.61, "learning_rate": 1.9343661971830986e-05, "loss": 1.3236, "step": 8571 }, { "epoch": 9.61, "learning_rate": 1.934225352112676e-05, "loss": 1.2262, "step": 8572 }, { "epoch": 9.61, "learning_rate": 1.9340845070422535e-05, "loss": 1.4776, "step": 8573 }, { "epoch": 9.61, "learning_rate": 1.933943661971831e-05, "loss": 1.3009, "step": 8574 }, { "epoch": 9.61, "learning_rate": 1.9338028169014085e-05, "loss": 1.3863, "step": 8575 }, { "epoch": 9.61, "learning_rate": 1.9336619718309858e-05, "loss": 1.0212, "step": 8576 }, { "epoch": 9.62, "learning_rate": 1.9335211267605637e-05, "loss": 0.9113, "step": 8577 }, { "epoch": 9.62, "learning_rate": 1.933380281690141e-05, "loss": 1.1799, "step": 8578 }, { "epoch": 9.62, "learning_rate": 1.9332394366197183e-05, "loss": 1.8828, "step": 8579 }, { "epoch": 9.62, "learning_rate": 1.933098591549296e-05, "loss": 1.8, "step": 8580 }, { "epoch": 9.62, "learning_rate": 1.9329577464788732e-05, "loss": 1.4275, "step": 8581 }, { "epoch": 9.62, "learning_rate": 1.932816901408451e-05, "loss": 1.5849, "step": 8582 }, { "epoch": 9.62, "learning_rate": 1.9326760563380282e-05, "loss": 1.4472, "step": 8583 }, { "epoch": 9.62, "learning_rate": 1.9325352112676058e-05, "loss": 1.672, "step": 8584 }, { "epoch": 9.62, "learning_rate": 1.932394366197183e-05, "loss": 1.4919, "step": 8585 }, { "epoch": 9.63, "learning_rate": 1.9322535211267607e-05, "loss": 1.4672, "step": 8586 }, { "epoch": 9.63, "learning_rate": 1.932112676056338e-05, "loss": 1.4448, "step": 8587 }, { "epoch": 9.63, "learning_rate": 1.9319718309859153e-05, "loss": 1.4606, "step": 8588 }, { "epoch": 9.63, "learning_rate": 1.931830985915493e-05, "loss": 1.5213, "step": 8589 }, { "epoch": 9.63, "learning_rate": 1.9316901408450703e-05, "loss": 1.3469, "step": 8590 }, { "epoch": 9.63, "learning_rate": 1.931549295774648e-05, "loss": 1.3644, "step": 8591 }, { "epoch": 9.63, "learning_rate": 1.9314084507042255e-05, "loss": 1.6188, "step": 8592 }, { "epoch": 9.63, "learning_rate": 1.931267605633803e-05, "loss": 1.3592, "step": 8593 }, { "epoch": 9.63, "learning_rate": 1.9311267605633804e-05, "loss": 1.4692, "step": 8594 }, { "epoch": 9.64, "learning_rate": 1.9309859154929577e-05, "loss": 1.5169, "step": 8595 }, { "epoch": 9.64, "learning_rate": 1.9308450704225354e-05, "loss": 1.3935, "step": 8596 }, { "epoch": 9.64, "learning_rate": 1.9307042253521127e-05, "loss": 1.4832, "step": 8597 }, { "epoch": 9.64, "learning_rate": 1.9305633802816903e-05, "loss": 1.153, "step": 8598 }, { "epoch": 9.64, "learning_rate": 1.9304225352112676e-05, "loss": 1.2244, "step": 8599 }, { "epoch": 9.64, "learning_rate": 1.9302816901408452e-05, "loss": 1.2861, "step": 8600 }, { "epoch": 9.64, "learning_rate": 1.9301408450704225e-05, "loss": 1.2217, "step": 8601 }, { "epoch": 9.64, "learning_rate": 1.9299999999999998e-05, "loss": 1.5319, "step": 8602 }, { "epoch": 9.64, "learning_rate": 1.9298591549295775e-05, "loss": 1.18, "step": 8603 }, { "epoch": 9.65, "learning_rate": 1.9297183098591547e-05, "loss": 1.4882, "step": 8604 }, { "epoch": 9.65, "learning_rate": 1.9295774647887324e-05, "loss": 1.3853, "step": 8605 }, { "epoch": 9.65, "learning_rate": 1.9294366197183097e-05, "loss": 1.4068, "step": 8606 }, { "epoch": 9.65, "learning_rate": 1.9292957746478876e-05, "loss": 1.2707, "step": 8607 }, { "epoch": 9.65, "learning_rate": 1.929154929577465e-05, "loss": 1.3926, "step": 8608 }, { "epoch": 9.65, "learning_rate": 1.9290140845070426e-05, "loss": 1.3158, "step": 8609 }, { "epoch": 9.65, "learning_rate": 1.92887323943662e-05, "loss": 1.3575, "step": 8610 }, { "epoch": 9.65, "learning_rate": 1.928732394366197e-05, "loss": 1.2906, "step": 8611 }, { "epoch": 9.65, "learning_rate": 1.9285915492957748e-05, "loss": 1.3343, "step": 8612 }, { "epoch": 9.66, "learning_rate": 1.928450704225352e-05, "loss": 1.4448, "step": 8613 }, { "epoch": 9.66, "learning_rate": 1.9283098591549297e-05, "loss": 1.2986, "step": 8614 }, { "epoch": 9.66, "learning_rate": 1.928169014084507e-05, "loss": 1.3223, "step": 8615 }, { "epoch": 9.66, "learning_rate": 1.9280281690140846e-05, "loss": 1.3567, "step": 8616 }, { "epoch": 9.66, "learning_rate": 1.927887323943662e-05, "loss": 1.2284, "step": 8617 }, { "epoch": 9.66, "learning_rate": 1.9277464788732392e-05, "loss": 1.4879, "step": 8618 }, { "epoch": 9.66, "learning_rate": 1.927605633802817e-05, "loss": 1.2993, "step": 8619 }, { "epoch": 9.66, "learning_rate": 1.927464788732394e-05, "loss": 1.3125, "step": 8620 }, { "epoch": 9.66, "learning_rate": 1.9273239436619718e-05, "loss": 1.3334, "step": 8621 }, { "epoch": 9.67, "learning_rate": 1.9271830985915494e-05, "loss": 1.23, "step": 8622 }, { "epoch": 9.67, "learning_rate": 1.927042253521127e-05, "loss": 1.263, "step": 8623 }, { "epoch": 9.67, "learning_rate": 1.9269014084507044e-05, "loss": 1.4408, "step": 8624 }, { "epoch": 9.67, "learning_rate": 1.926760563380282e-05, "loss": 1.2176, "step": 8625 }, { "epoch": 9.67, "learning_rate": 1.9266197183098593e-05, "loss": 1.2201, "step": 8626 }, { "epoch": 9.67, "learning_rate": 1.9264788732394366e-05, "loss": 1.4872, "step": 8627 }, { "epoch": 9.67, "learning_rate": 1.9263380281690142e-05, "loss": 0.9755, "step": 8628 }, { "epoch": 9.67, "learning_rate": 1.9261971830985915e-05, "loss": 1.897, "step": 8629 }, { "epoch": 9.67, "learning_rate": 1.926056338028169e-05, "loss": 1.6427, "step": 8630 }, { "epoch": 9.68, "learning_rate": 1.9259154929577464e-05, "loss": 1.5272, "step": 8631 }, { "epoch": 9.68, "learning_rate": 1.925774647887324e-05, "loss": 1.3047, "step": 8632 }, { "epoch": 9.68, "learning_rate": 1.9256338028169014e-05, "loss": 1.449, "step": 8633 }, { "epoch": 9.68, "learning_rate": 1.9254929577464787e-05, "loss": 1.2811, "step": 8634 }, { "epoch": 9.68, "learning_rate": 1.9253521126760563e-05, "loss": 1.5593, "step": 8635 }, { "epoch": 9.68, "learning_rate": 1.9252112676056336e-05, "loss": 1.4289, "step": 8636 }, { "epoch": 9.68, "learning_rate": 1.9250704225352116e-05, "loss": 1.1535, "step": 8637 }, { "epoch": 9.68, "learning_rate": 1.924929577464789e-05, "loss": 1.343, "step": 8638 }, { "epoch": 9.68, "learning_rate": 1.9247887323943665e-05, "loss": 1.3103, "step": 8639 }, { "epoch": 9.69, "learning_rate": 1.9246478873239438e-05, "loss": 1.3481, "step": 8640 }, { "epoch": 9.69, "learning_rate": 1.9245070422535214e-05, "loss": 1.1715, "step": 8641 }, { "epoch": 9.69, "learning_rate": 1.9243661971830987e-05, "loss": 1.2342, "step": 8642 }, { "epoch": 9.69, "learning_rate": 1.924225352112676e-05, "loss": 1.4027, "step": 8643 }, { "epoch": 9.69, "learning_rate": 1.9240845070422536e-05, "loss": 1.3736, "step": 8644 }, { "epoch": 9.69, "learning_rate": 1.923943661971831e-05, "loss": 1.34, "step": 8645 }, { "epoch": 9.69, "learning_rate": 1.9238028169014086e-05, "loss": 1.2681, "step": 8646 }, { "epoch": 9.69, "learning_rate": 1.923661971830986e-05, "loss": 1.3802, "step": 8647 }, { "epoch": 9.7, "learning_rate": 1.9235211267605635e-05, "loss": 1.5432, "step": 8648 }, { "epoch": 9.7, "learning_rate": 1.9233802816901408e-05, "loss": 1.2677, "step": 8649 }, { "epoch": 9.7, "learning_rate": 1.923239436619718e-05, "loss": 1.3615, "step": 8650 }, { "epoch": 9.7, "learning_rate": 1.9230985915492957e-05, "loss": 1.1578, "step": 8651 }, { "epoch": 9.7, "learning_rate": 1.922957746478873e-05, "loss": 1.4619, "step": 8652 }, { "epoch": 9.7, "learning_rate": 1.922816901408451e-05, "loss": 1.1841, "step": 8653 }, { "epoch": 9.7, "learning_rate": 1.9226760563380283e-05, "loss": 1.2546, "step": 8654 }, { "epoch": 9.7, "learning_rate": 1.922535211267606e-05, "loss": 1.4378, "step": 8655 }, { "epoch": 9.7, "learning_rate": 1.9223943661971832e-05, "loss": 1.2229, "step": 8656 }, { "epoch": 9.71, "learning_rate": 1.922253521126761e-05, "loss": 1.2808, "step": 8657 }, { "epoch": 9.71, "learning_rate": 1.922112676056338e-05, "loss": 1.4692, "step": 8658 }, { "epoch": 9.71, "learning_rate": 1.9219718309859154e-05, "loss": 1.2907, "step": 8659 }, { "epoch": 9.71, "learning_rate": 1.921830985915493e-05, "loss": 1.2413, "step": 8660 }, { "epoch": 9.71, "learning_rate": 1.9216901408450703e-05, "loss": 1.2448, "step": 8661 }, { "epoch": 9.71, "learning_rate": 1.921549295774648e-05, "loss": 1.3752, "step": 8662 }, { "epoch": 9.71, "learning_rate": 1.9214084507042253e-05, "loss": 1.5537, "step": 8663 }, { "epoch": 9.71, "learning_rate": 1.921267605633803e-05, "loss": 1.2205, "step": 8664 }, { "epoch": 9.71, "learning_rate": 1.9211267605633802e-05, "loss": 1.276, "step": 8665 }, { "epoch": 9.72, "learning_rate": 1.9209859154929575e-05, "loss": 1.3036, "step": 8666 }, { "epoch": 9.72, "learning_rate": 1.920845070422535e-05, "loss": 1.238, "step": 8667 }, { "epoch": 9.72, "learning_rate": 1.9207042253521128e-05, "loss": 1.4006, "step": 8668 }, { "epoch": 9.72, "learning_rate": 1.9205633802816904e-05, "loss": 1.3944, "step": 8669 }, { "epoch": 9.72, "learning_rate": 1.9204225352112677e-05, "loss": 1.4835, "step": 8670 }, { "epoch": 9.72, "learning_rate": 1.9202816901408453e-05, "loss": 1.1457, "step": 8671 }, { "epoch": 9.72, "learning_rate": 1.9201408450704226e-05, "loss": 1.254, "step": 8672 }, { "epoch": 9.72, "learning_rate": 1.9200000000000003e-05, "loss": 1.0965, "step": 8673 }, { "epoch": 9.72, "learning_rate": 1.9198591549295775e-05, "loss": 1.3033, "step": 8674 }, { "epoch": 9.73, "learning_rate": 1.919718309859155e-05, "loss": 0.9893, "step": 8675 }, { "epoch": 9.73, "learning_rate": 1.9195774647887325e-05, "loss": 1.2359, "step": 8676 }, { "epoch": 9.73, "learning_rate": 1.9194366197183098e-05, "loss": 0.9705, "step": 8677 }, { "epoch": 9.73, "learning_rate": 1.9192957746478874e-05, "loss": 0.9731, "step": 8678 }, { "epoch": 9.73, "learning_rate": 1.9191549295774647e-05, "loss": 1.7056, "step": 8679 }, { "epoch": 9.73, "learning_rate": 1.9190140845070423e-05, "loss": 1.577, "step": 8680 }, { "epoch": 9.73, "learning_rate": 1.9188732394366196e-05, "loss": 1.5942, "step": 8681 }, { "epoch": 9.73, "learning_rate": 1.918732394366197e-05, "loss": 1.4915, "step": 8682 }, { "epoch": 9.73, "learning_rate": 1.918591549295775e-05, "loss": 1.508, "step": 8683 }, { "epoch": 9.74, "learning_rate": 1.9184507042253522e-05, "loss": 1.2277, "step": 8684 }, { "epoch": 9.74, "learning_rate": 1.9183098591549298e-05, "loss": 1.2942, "step": 8685 }, { "epoch": 9.74, "learning_rate": 1.918169014084507e-05, "loss": 1.5756, "step": 8686 }, { "epoch": 9.74, "learning_rate": 1.9180281690140847e-05, "loss": 1.359, "step": 8687 }, { "epoch": 9.74, "learning_rate": 1.917887323943662e-05, "loss": 1.6414, "step": 8688 }, { "epoch": 9.74, "learning_rate": 1.9177464788732397e-05, "loss": 1.3347, "step": 8689 }, { "epoch": 9.74, "learning_rate": 1.917605633802817e-05, "loss": 1.3961, "step": 8690 }, { "epoch": 9.74, "learning_rate": 1.9174647887323943e-05, "loss": 1.4515, "step": 8691 }, { "epoch": 9.74, "learning_rate": 1.917323943661972e-05, "loss": 1.1676, "step": 8692 }, { "epoch": 9.75, "learning_rate": 1.9171830985915492e-05, "loss": 1.4983, "step": 8693 }, { "epoch": 9.75, "learning_rate": 1.9170422535211268e-05, "loss": 1.3104, "step": 8694 }, { "epoch": 9.75, "learning_rate": 1.916901408450704e-05, "loss": 1.3979, "step": 8695 }, { "epoch": 9.75, "learning_rate": 1.9167605633802817e-05, "loss": 1.3942, "step": 8696 }, { "epoch": 9.75, "learning_rate": 1.916619718309859e-05, "loss": 1.4912, "step": 8697 }, { "epoch": 9.75, "learning_rate": 1.9164788732394367e-05, "loss": 1.5952, "step": 8698 }, { "epoch": 9.75, "learning_rate": 1.9163380281690143e-05, "loss": 1.5127, "step": 8699 }, { "epoch": 9.75, "learning_rate": 1.9161971830985916e-05, "loss": 1.3906, "step": 8700 }, { "epoch": 9.75, "learning_rate": 1.9160563380281692e-05, "loss": 1.2767, "step": 8701 }, { "epoch": 9.76, "learning_rate": 1.9159154929577465e-05, "loss": 1.383, "step": 8702 }, { "epoch": 9.76, "learning_rate": 1.915774647887324e-05, "loss": 1.4118, "step": 8703 }, { "epoch": 9.76, "learning_rate": 1.9156338028169015e-05, "loss": 1.317, "step": 8704 }, { "epoch": 9.76, "learning_rate": 1.915492957746479e-05, "loss": 1.3883, "step": 8705 }, { "epoch": 9.76, "learning_rate": 1.9153521126760564e-05, "loss": 1.4434, "step": 8706 }, { "epoch": 9.76, "learning_rate": 1.9152112676056337e-05, "loss": 1.3693, "step": 8707 }, { "epoch": 9.76, "learning_rate": 1.9150704225352113e-05, "loss": 1.2684, "step": 8708 }, { "epoch": 9.76, "learning_rate": 1.9149295774647886e-05, "loss": 1.3017, "step": 8709 }, { "epoch": 9.76, "learning_rate": 1.9147887323943662e-05, "loss": 1.2549, "step": 8710 }, { "epoch": 9.77, "learning_rate": 1.9146478873239435e-05, "loss": 1.2643, "step": 8711 }, { "epoch": 9.77, "learning_rate": 1.914507042253521e-05, "loss": 1.3189, "step": 8712 }, { "epoch": 9.77, "learning_rate": 1.9143661971830988e-05, "loss": 1.3241, "step": 8713 }, { "epoch": 9.77, "learning_rate": 1.914225352112676e-05, "loss": 1.3587, "step": 8714 }, { "epoch": 9.77, "learning_rate": 1.9140845070422537e-05, "loss": 1.235, "step": 8715 }, { "epoch": 9.77, "learning_rate": 1.913943661971831e-05, "loss": 1.2895, "step": 8716 }, { "epoch": 9.77, "learning_rate": 1.9138028169014087e-05, "loss": 1.404, "step": 8717 }, { "epoch": 9.77, "learning_rate": 1.913661971830986e-05, "loss": 1.1988, "step": 8718 }, { "epoch": 9.77, "learning_rate": 1.9135211267605636e-05, "loss": 1.4974, "step": 8719 }, { "epoch": 9.78, "learning_rate": 1.913380281690141e-05, "loss": 1.2786, "step": 8720 }, { "epoch": 9.78, "learning_rate": 1.9132394366197182e-05, "loss": 1.1977, "step": 8721 }, { "epoch": 9.78, "learning_rate": 1.9130985915492958e-05, "loss": 1.1641, "step": 8722 }, { "epoch": 9.78, "learning_rate": 1.912957746478873e-05, "loss": 1.2788, "step": 8723 }, { "epoch": 9.78, "learning_rate": 1.9128169014084507e-05, "loss": 1.1097, "step": 8724 }, { "epoch": 9.78, "learning_rate": 1.912676056338028e-05, "loss": 1.2837, "step": 8725 }, { "epoch": 9.78, "learning_rate": 1.9125352112676057e-05, "loss": 1.1096, "step": 8726 }, { "epoch": 9.78, "learning_rate": 1.912394366197183e-05, "loss": 1.274, "step": 8727 }, { "epoch": 9.78, "learning_rate": 1.912253521126761e-05, "loss": 1.044, "step": 8728 }, { "epoch": 9.79, "learning_rate": 1.9121126760563382e-05, "loss": 1.6414, "step": 8729 }, { "epoch": 9.79, "learning_rate": 1.9119718309859155e-05, "loss": 1.6186, "step": 8730 }, { "epoch": 9.79, "learning_rate": 1.911830985915493e-05, "loss": 1.5882, "step": 8731 }, { "epoch": 9.79, "learning_rate": 1.9116901408450704e-05, "loss": 1.5073, "step": 8732 }, { "epoch": 9.79, "learning_rate": 1.911549295774648e-05, "loss": 1.3681, "step": 8733 }, { "epoch": 9.79, "learning_rate": 1.9114084507042254e-05, "loss": 1.3244, "step": 8734 }, { "epoch": 9.79, "learning_rate": 1.911267605633803e-05, "loss": 1.2442, "step": 8735 }, { "epoch": 9.79, "learning_rate": 1.9111267605633803e-05, "loss": 1.5604, "step": 8736 }, { "epoch": 9.79, "learning_rate": 1.9109859154929576e-05, "loss": 1.4143, "step": 8737 }, { "epoch": 9.8, "learning_rate": 1.9108450704225352e-05, "loss": 1.3593, "step": 8738 }, { "epoch": 9.8, "learning_rate": 1.9107042253521125e-05, "loss": 1.1157, "step": 8739 }, { "epoch": 9.8, "learning_rate": 1.91056338028169e-05, "loss": 1.6688, "step": 8740 }, { "epoch": 9.8, "learning_rate": 1.9104225352112674e-05, "loss": 1.2867, "step": 8741 }, { "epoch": 9.8, "learning_rate": 1.910281690140845e-05, "loss": 1.169, "step": 8742 }, { "epoch": 9.8, "learning_rate": 1.9101408450704227e-05, "loss": 1.3254, "step": 8743 }, { "epoch": 9.8, "learning_rate": 1.9100000000000003e-05, "loss": 1.1239, "step": 8744 }, { "epoch": 9.8, "learning_rate": 1.9098591549295776e-05, "loss": 1.2495, "step": 8745 }, { "epoch": 9.8, "learning_rate": 1.909718309859155e-05, "loss": 1.2949, "step": 8746 }, { "epoch": 9.81, "learning_rate": 1.9095774647887326e-05, "loss": 1.1914, "step": 8747 }, { "epoch": 9.81, "learning_rate": 1.90943661971831e-05, "loss": 1.5459, "step": 8748 }, { "epoch": 9.81, "learning_rate": 1.9092957746478875e-05, "loss": 1.2321, "step": 8749 }, { "epoch": 9.81, "learning_rate": 1.9091549295774648e-05, "loss": 1.284, "step": 8750 }, { "epoch": 9.81, "learning_rate": 1.9090140845070424e-05, "loss": 1.1993, "step": 8751 }, { "epoch": 9.81, "learning_rate": 1.9088732394366197e-05, "loss": 1.262, "step": 8752 }, { "epoch": 9.81, "learning_rate": 1.908732394366197e-05, "loss": 1.2362, "step": 8753 }, { "epoch": 9.81, "learning_rate": 1.9085915492957746e-05, "loss": 1.1699, "step": 8754 }, { "epoch": 9.82, "learning_rate": 1.908450704225352e-05, "loss": 1.2229, "step": 8755 }, { "epoch": 9.82, "learning_rate": 1.9083098591549296e-05, "loss": 1.4112, "step": 8756 }, { "epoch": 9.82, "learning_rate": 1.908169014084507e-05, "loss": 1.5002, "step": 8757 }, { "epoch": 9.82, "learning_rate": 1.9080281690140845e-05, "loss": 1.3114, "step": 8758 }, { "epoch": 9.82, "learning_rate": 1.907887323943662e-05, "loss": 1.1886, "step": 8759 }, { "epoch": 9.82, "learning_rate": 1.9077464788732398e-05, "loss": 1.3677, "step": 8760 }, { "epoch": 9.82, "learning_rate": 1.907605633802817e-05, "loss": 1.3375, "step": 8761 }, { "epoch": 9.82, "learning_rate": 1.9074647887323944e-05, "loss": 1.3593, "step": 8762 }, { "epoch": 9.82, "learning_rate": 1.907323943661972e-05, "loss": 1.4232, "step": 8763 }, { "epoch": 9.83, "learning_rate": 1.9071830985915493e-05, "loss": 1.2188, "step": 8764 }, { "epoch": 9.83, "learning_rate": 1.907042253521127e-05, "loss": 1.4604, "step": 8765 }, { "epoch": 9.83, "learning_rate": 1.9069014084507042e-05, "loss": 1.1506, "step": 8766 }, { "epoch": 9.83, "learning_rate": 1.906760563380282e-05, "loss": 1.365, "step": 8767 }, { "epoch": 9.83, "learning_rate": 1.906619718309859e-05, "loss": 1.1956, "step": 8768 }, { "epoch": 9.83, "learning_rate": 1.9064788732394364e-05, "loss": 1.0739, "step": 8769 }, { "epoch": 9.83, "learning_rate": 1.906338028169014e-05, "loss": 1.2516, "step": 8770 }, { "epoch": 9.83, "learning_rate": 1.9061971830985914e-05, "loss": 1.3632, "step": 8771 }, { "epoch": 9.83, "learning_rate": 1.906056338028169e-05, "loss": 1.3409, "step": 8772 }, { "epoch": 9.84, "learning_rate": 1.9059154929577463e-05, "loss": 1.2192, "step": 8773 }, { "epoch": 9.84, "learning_rate": 1.9057746478873243e-05, "loss": 1.2163, "step": 8774 }, { "epoch": 9.84, "learning_rate": 1.9056338028169016e-05, "loss": 1.1622, "step": 8775 }, { "epoch": 9.84, "learning_rate": 1.9054929577464792e-05, "loss": 1.037, "step": 8776 }, { "epoch": 9.84, "learning_rate": 1.9053521126760565e-05, "loss": 1.3462, "step": 8777 }, { "epoch": 9.84, "learning_rate": 1.9052112676056338e-05, "loss": 1.261, "step": 8778 }, { "epoch": 9.84, "learning_rate": 1.9050704225352114e-05, "loss": 1.9024, "step": 8779 }, { "epoch": 9.84, "learning_rate": 1.9049295774647887e-05, "loss": 1.7848, "step": 8780 }, { "epoch": 9.84, "learning_rate": 1.9047887323943663e-05, "loss": 1.6188, "step": 8781 }, { "epoch": 9.85, "learning_rate": 1.9046478873239436e-05, "loss": 1.4472, "step": 8782 }, { "epoch": 9.85, "learning_rate": 1.9045070422535213e-05, "loss": 1.2861, "step": 8783 }, { "epoch": 9.85, "learning_rate": 1.9043661971830986e-05, "loss": 1.2351, "step": 8784 }, { "epoch": 9.85, "learning_rate": 1.904225352112676e-05, "loss": 1.3833, "step": 8785 }, { "epoch": 9.85, "learning_rate": 1.9040845070422535e-05, "loss": 1.402, "step": 8786 }, { "epoch": 9.85, "learning_rate": 1.9039436619718308e-05, "loss": 1.3911, "step": 8787 }, { "epoch": 9.85, "learning_rate": 1.9038028169014084e-05, "loss": 1.5502, "step": 8788 }, { "epoch": 9.85, "learning_rate": 1.903661971830986e-05, "loss": 1.4548, "step": 8789 }, { "epoch": 9.85, "learning_rate": 1.9035211267605637e-05, "loss": 1.4044, "step": 8790 }, { "epoch": 9.86, "learning_rate": 1.903380281690141e-05, "loss": 1.3506, "step": 8791 }, { "epoch": 9.86, "learning_rate": 1.9032394366197186e-05, "loss": 1.4931, "step": 8792 }, { "epoch": 9.86, "learning_rate": 1.903098591549296e-05, "loss": 1.5329, "step": 8793 }, { "epoch": 9.86, "learning_rate": 1.9029577464788732e-05, "loss": 1.2193, "step": 8794 }, { "epoch": 9.86, "learning_rate": 1.9028169014084508e-05, "loss": 1.2359, "step": 8795 }, { "epoch": 9.86, "learning_rate": 1.902676056338028e-05, "loss": 1.3406, "step": 8796 }, { "epoch": 9.86, "learning_rate": 1.9025352112676058e-05, "loss": 1.3851, "step": 8797 }, { "epoch": 9.86, "learning_rate": 1.902394366197183e-05, "loss": 1.6301, "step": 8798 }, { "epoch": 9.86, "learning_rate": 1.9022535211267607e-05, "loss": 1.2903, "step": 8799 }, { "epoch": 9.87, "learning_rate": 1.902112676056338e-05, "loss": 1.2494, "step": 8800 }, { "epoch": 9.87, "learning_rate": 1.9019718309859153e-05, "loss": 1.383, "step": 8801 }, { "epoch": 9.87, "learning_rate": 1.901830985915493e-05, "loss": 1.1931, "step": 8802 }, { "epoch": 9.87, "learning_rate": 1.9016901408450702e-05, "loss": 1.5056, "step": 8803 }, { "epoch": 9.87, "learning_rate": 1.9015492957746482e-05, "loss": 1.3001, "step": 8804 }, { "epoch": 9.87, "learning_rate": 1.9014084507042255e-05, "loss": 1.4736, "step": 8805 }, { "epoch": 9.87, "learning_rate": 1.901267605633803e-05, "loss": 1.3399, "step": 8806 }, { "epoch": 9.87, "learning_rate": 1.9011267605633804e-05, "loss": 1.1558, "step": 8807 }, { "epoch": 9.87, "learning_rate": 1.900985915492958e-05, "loss": 1.1516, "step": 8808 }, { "epoch": 9.88, "learning_rate": 1.9008450704225353e-05, "loss": 1.3348, "step": 8809 }, { "epoch": 9.88, "learning_rate": 1.9007042253521126e-05, "loss": 1.2545, "step": 8810 }, { "epoch": 9.88, "learning_rate": 1.9005633802816902e-05, "loss": 1.1333, "step": 8811 }, { "epoch": 9.88, "learning_rate": 1.9004225352112675e-05, "loss": 1.382, "step": 8812 }, { "epoch": 9.88, "learning_rate": 1.9002816901408452e-05, "loss": 1.2587, "step": 8813 }, { "epoch": 9.88, "learning_rate": 1.9001408450704225e-05, "loss": 1.3406, "step": 8814 }, { "epoch": 9.88, "learning_rate": 1.9e-05, "loss": 1.3357, "step": 8815 }, { "epoch": 9.88, "learning_rate": 1.8998591549295774e-05, "loss": 1.257, "step": 8816 }, { "epoch": 9.88, "learning_rate": 1.8997183098591547e-05, "loss": 1.1518, "step": 8817 }, { "epoch": 9.89, "learning_rate": 1.8995774647887323e-05, "loss": 1.266, "step": 8818 }, { "epoch": 9.89, "learning_rate": 1.89943661971831e-05, "loss": 1.3254, "step": 8819 }, { "epoch": 9.89, "learning_rate": 1.8992957746478876e-05, "loss": 1.3391, "step": 8820 }, { "epoch": 9.89, "learning_rate": 1.899154929577465e-05, "loss": 1.2798, "step": 8821 }, { "epoch": 9.89, "learning_rate": 1.8990140845070425e-05, "loss": 1.1182, "step": 8822 }, { "epoch": 9.89, "learning_rate": 1.8988732394366198e-05, "loss": 1.336, "step": 8823 }, { "epoch": 9.89, "learning_rate": 1.898732394366197e-05, "loss": 1.1167, "step": 8824 }, { "epoch": 9.89, "learning_rate": 1.8985915492957747e-05, "loss": 1.3189, "step": 8825 }, { "epoch": 9.89, "learning_rate": 1.898450704225352e-05, "loss": 1.2591, "step": 8826 }, { "epoch": 9.9, "learning_rate": 1.8983098591549297e-05, "loss": 1.1552, "step": 8827 }, { "epoch": 9.9, "learning_rate": 1.898169014084507e-05, "loss": 1.2258, "step": 8828 }, { "epoch": 9.9, "learning_rate": 1.8980281690140846e-05, "loss": 1.8736, "step": 8829 }, { "epoch": 9.9, "learning_rate": 1.897887323943662e-05, "loss": 1.6125, "step": 8830 }, { "epoch": 9.9, "learning_rate": 1.8977464788732395e-05, "loss": 1.6214, "step": 8831 }, { "epoch": 9.9, "learning_rate": 1.8976056338028168e-05, "loss": 1.5583, "step": 8832 }, { "epoch": 9.9, "learning_rate": 1.897464788732394e-05, "loss": 1.4126, "step": 8833 }, { "epoch": 9.9, "learning_rate": 1.897323943661972e-05, "loss": 1.3523, "step": 8834 }, { "epoch": 9.9, "learning_rate": 1.8971830985915494e-05, "loss": 1.4188, "step": 8835 }, { "epoch": 9.91, "learning_rate": 1.897042253521127e-05, "loss": 1.3987, "step": 8836 }, { "epoch": 9.91, "learning_rate": 1.8969014084507043e-05, "loss": 1.4015, "step": 8837 }, { "epoch": 9.91, "learning_rate": 1.896760563380282e-05, "loss": 1.4223, "step": 8838 }, { "epoch": 9.91, "learning_rate": 1.8966197183098592e-05, "loss": 1.2707, "step": 8839 }, { "epoch": 9.91, "learning_rate": 1.8964788732394365e-05, "loss": 1.3093, "step": 8840 }, { "epoch": 9.91, "learning_rate": 1.896338028169014e-05, "loss": 1.2947, "step": 8841 }, { "epoch": 9.91, "learning_rate": 1.8961971830985915e-05, "loss": 1.154, "step": 8842 }, { "epoch": 9.91, "learning_rate": 1.896056338028169e-05, "loss": 1.3044, "step": 8843 }, { "epoch": 9.91, "learning_rate": 1.8959154929577464e-05, "loss": 1.3845, "step": 8844 }, { "epoch": 9.92, "learning_rate": 1.895774647887324e-05, "loss": 1.351, "step": 8845 }, { "epoch": 9.92, "learning_rate": 1.8956338028169013e-05, "loss": 1.3158, "step": 8846 }, { "epoch": 9.92, "learning_rate": 1.895492957746479e-05, "loss": 1.195, "step": 8847 }, { "epoch": 9.92, "learning_rate": 1.8953521126760562e-05, "loss": 1.3333, "step": 8848 }, { "epoch": 9.92, "learning_rate": 1.895211267605634e-05, "loss": 1.2887, "step": 8849 }, { "epoch": 9.92, "learning_rate": 1.8950704225352115e-05, "loss": 1.4413, "step": 8850 }, { "epoch": 9.92, "learning_rate": 1.8949295774647888e-05, "loss": 1.432, "step": 8851 }, { "epoch": 9.92, "learning_rate": 1.8947887323943664e-05, "loss": 1.2896, "step": 8852 }, { "epoch": 9.92, "learning_rate": 1.8946478873239437e-05, "loss": 1.1988, "step": 8853 }, { "epoch": 9.93, "learning_rate": 1.8945070422535214e-05, "loss": 1.075, "step": 8854 }, { "epoch": 9.93, "learning_rate": 1.8943661971830987e-05, "loss": 1.2109, "step": 8855 }, { "epoch": 9.93, "learning_rate": 1.894225352112676e-05, "loss": 1.4673, "step": 8856 }, { "epoch": 9.93, "learning_rate": 1.8940845070422536e-05, "loss": 1.3554, "step": 8857 }, { "epoch": 9.93, "learning_rate": 1.893943661971831e-05, "loss": 1.4199, "step": 8858 }, { "epoch": 9.93, "learning_rate": 1.8938028169014085e-05, "loss": 1.3366, "step": 8859 }, { "epoch": 9.93, "learning_rate": 1.8936619718309858e-05, "loss": 1.1738, "step": 8860 }, { "epoch": 9.93, "learning_rate": 1.8935211267605634e-05, "loss": 1.2623, "step": 8861 }, { "epoch": 9.93, "learning_rate": 1.8933802816901407e-05, "loss": 1.2559, "step": 8862 }, { "epoch": 9.94, "learning_rate": 1.8932394366197184e-05, "loss": 1.1008, "step": 8863 }, { "epoch": 9.94, "learning_rate": 1.893098591549296e-05, "loss": 1.1603, "step": 8864 }, { "epoch": 9.94, "learning_rate": 1.8929577464788733e-05, "loss": 1.3614, "step": 8865 }, { "epoch": 9.94, "learning_rate": 1.892816901408451e-05, "loss": 1.2848, "step": 8866 }, { "epoch": 9.94, "learning_rate": 1.8926760563380282e-05, "loss": 1.5535, "step": 8867 }, { "epoch": 9.94, "learning_rate": 1.892535211267606e-05, "loss": 1.2155, "step": 8868 }, { "epoch": 9.94, "learning_rate": 1.892394366197183e-05, "loss": 1.3587, "step": 8869 }, { "epoch": 9.94, "learning_rate": 1.8922535211267608e-05, "loss": 1.4325, "step": 8870 }, { "epoch": 9.95, "learning_rate": 1.892112676056338e-05, "loss": 1.1541, "step": 8871 }, { "epoch": 9.95, "learning_rate": 1.8919718309859154e-05, "loss": 1.1402, "step": 8872 }, { "epoch": 9.95, "learning_rate": 1.891830985915493e-05, "loss": 1.2188, "step": 8873 }, { "epoch": 9.95, "learning_rate": 1.8916901408450703e-05, "loss": 1.0075, "step": 8874 }, { "epoch": 9.95, "learning_rate": 1.891549295774648e-05, "loss": 1.1544, "step": 8875 }, { "epoch": 9.95, "learning_rate": 1.8914084507042252e-05, "loss": 1.2195, "step": 8876 }, { "epoch": 9.95, "learning_rate": 1.891267605633803e-05, "loss": 1.1182, "step": 8877 }, { "epoch": 9.95, "learning_rate": 1.89112676056338e-05, "loss": 1.0656, "step": 8878 }, { "epoch": 9.95, "learning_rate": 1.8909859154929578e-05, "loss": 1.8691, "step": 8879 }, { "epoch": 9.96, "learning_rate": 1.8908450704225354e-05, "loss": 1.8291, "step": 8880 }, { "epoch": 9.96, "learning_rate": 1.8907042253521127e-05, "loss": 1.3843, "step": 8881 }, { "epoch": 9.96, "learning_rate": 1.8905633802816903e-05, "loss": 1.4511, "step": 8882 }, { "epoch": 9.96, "learning_rate": 1.8904225352112676e-05, "loss": 1.4674, "step": 8883 }, { "epoch": 9.96, "learning_rate": 1.8902816901408453e-05, "loss": 1.2969, "step": 8884 }, { "epoch": 9.96, "learning_rate": 1.8901408450704226e-05, "loss": 1.3773, "step": 8885 }, { "epoch": 9.96, "learning_rate": 1.8900000000000002e-05, "loss": 1.4383, "step": 8886 }, { "epoch": 9.96, "learning_rate": 1.8898591549295775e-05, "loss": 1.423, "step": 8887 }, { "epoch": 9.96, "learning_rate": 1.8897183098591548e-05, "loss": 1.5221, "step": 8888 }, { "epoch": 9.97, "learning_rate": 1.8895774647887324e-05, "loss": 1.3624, "step": 8889 }, { "epoch": 9.97, "learning_rate": 1.8894366197183097e-05, "loss": 1.4841, "step": 8890 }, { "epoch": 9.97, "learning_rate": 1.8892957746478873e-05, "loss": 1.3182, "step": 8891 }, { "epoch": 9.97, "learning_rate": 1.8891549295774646e-05, "loss": 1.3148, "step": 8892 }, { "epoch": 9.97, "learning_rate": 1.8890140845070423e-05, "loss": 1.2075, "step": 8893 }, { "epoch": 9.97, "learning_rate": 1.8888732394366196e-05, "loss": 1.3267, "step": 8894 }, { "epoch": 9.97, "learning_rate": 1.8887323943661975e-05, "loss": 1.1272, "step": 8895 }, { "epoch": 9.97, "learning_rate": 1.888591549295775e-05, "loss": 1.3014, "step": 8896 }, { "epoch": 9.97, "learning_rate": 1.888450704225352e-05, "loss": 1.3987, "step": 8897 }, { "epoch": 9.98, "learning_rate": 1.8883098591549298e-05, "loss": 1.6137, "step": 8898 }, { "epoch": 9.98, "learning_rate": 1.888169014084507e-05, "loss": 1.3377, "step": 8899 }, { "epoch": 9.98, "learning_rate": 1.8880281690140847e-05, "loss": 1.2793, "step": 8900 }, { "epoch": 9.98, "learning_rate": 1.887887323943662e-05, "loss": 1.4161, "step": 8901 }, { "epoch": 9.98, "learning_rate": 1.8877464788732396e-05, "loss": 1.3416, "step": 8902 }, { "epoch": 9.98, "learning_rate": 1.887605633802817e-05, "loss": 1.4282, "step": 8903 }, { "epoch": 9.98, "learning_rate": 1.8874647887323942e-05, "loss": 1.3371, "step": 8904 }, { "epoch": 9.98, "learning_rate": 1.887323943661972e-05, "loss": 1.2706, "step": 8905 }, { "epoch": 9.98, "learning_rate": 1.887183098591549e-05, "loss": 1.3295, "step": 8906 }, { "epoch": 9.99, "learning_rate": 1.8870422535211268e-05, "loss": 1.2083, "step": 8907 }, { "epoch": 9.99, "learning_rate": 1.886901408450704e-05, "loss": 1.2368, "step": 8908 }, { "epoch": 9.99, "learning_rate": 1.8867605633802817e-05, "loss": 1.4568, "step": 8909 }, { "epoch": 9.99, "learning_rate": 1.8866197183098593e-05, "loss": 1.0838, "step": 8910 }, { "epoch": 9.99, "learning_rate": 1.886478873239437e-05, "loss": 1.1563, "step": 8911 }, { "epoch": 9.99, "learning_rate": 1.8863380281690143e-05, "loss": 1.3162, "step": 8912 }, { "epoch": 9.99, "learning_rate": 1.8861971830985916e-05, "loss": 1.4375, "step": 8913 }, { "epoch": 9.99, "learning_rate": 1.8860563380281692e-05, "loss": 1.2065, "step": 8914 }, { "epoch": 9.99, "learning_rate": 1.8859154929577465e-05, "loss": 1.2324, "step": 8915 }, { "epoch": 10.0, "learning_rate": 1.885774647887324e-05, "loss": 1.1438, "step": 8916 }, { "epoch": 10.0, "learning_rate": 1.8856338028169014e-05, "loss": 1.1235, "step": 8917 }, { "epoch": 10.0, "learning_rate": 1.885492957746479e-05, "loss": 0.9158, "step": 8918 }, { "epoch": 10.0, "learning_rate": 1.8853521126760563e-05, "loss": 1.1468, "step": 8919 }, { "epoch": 10.0, "learning_rate": 1.8852112676056336e-05, "loss": 0.9514, "step": 8920 }, { "epoch": 10.0, "learning_rate": 1.8850704225352113e-05, "loss": 1.5909, "step": 8921 }, { "epoch": 10.0, "learning_rate": 1.8849295774647886e-05, "loss": 1.5151, "step": 8922 }, { "epoch": 10.0, "learning_rate": 1.8847887323943662e-05, "loss": 1.4922, "step": 8923 }, { "epoch": 10.0, "learning_rate": 1.8846478873239435e-05, "loss": 1.5569, "step": 8924 }, { "epoch": 10.01, "learning_rate": 1.8845070422535215e-05, "loss": 1.4185, "step": 8925 }, { "epoch": 10.01, "learning_rate": 1.8843661971830988e-05, "loss": 1.1786, "step": 8926 }, { "epoch": 10.01, "learning_rate": 1.884225352112676e-05, "loss": 1.2896, "step": 8927 }, { "epoch": 10.01, "learning_rate": 1.8840845070422537e-05, "loss": 1.1941, "step": 8928 }, { "epoch": 10.01, "learning_rate": 1.883943661971831e-05, "loss": 1.3978, "step": 8929 }, { "epoch": 10.01, "learning_rate": 1.8838028169014086e-05, "loss": 1.1992, "step": 8930 }, { "epoch": 10.01, "learning_rate": 1.883661971830986e-05, "loss": 1.1308, "step": 8931 }, { "epoch": 10.01, "learning_rate": 1.8835211267605635e-05, "loss": 1.4845, "step": 8932 }, { "epoch": 10.01, "learning_rate": 1.8833802816901408e-05, "loss": 1.1165, "step": 8933 }, { "epoch": 10.02, "learning_rate": 1.8832394366197185e-05, "loss": 1.1827, "step": 8934 }, { "epoch": 10.02, "learning_rate": 1.8830985915492958e-05, "loss": 1.2805, "step": 8935 }, { "epoch": 10.02, "learning_rate": 1.882957746478873e-05, "loss": 1.2466, "step": 8936 }, { "epoch": 10.02, "learning_rate": 1.8828169014084507e-05, "loss": 1.2835, "step": 8937 }, { "epoch": 10.02, "learning_rate": 1.882676056338028e-05, "loss": 1.1538, "step": 8938 }, { "epoch": 10.02, "learning_rate": 1.8825352112676056e-05, "loss": 1.1197, "step": 8939 }, { "epoch": 10.02, "learning_rate": 1.8823943661971832e-05, "loss": 1.253, "step": 8940 }, { "epoch": 10.02, "learning_rate": 1.882253521126761e-05, "loss": 1.1693, "step": 8941 }, { "epoch": 10.02, "learning_rate": 1.8821126760563382e-05, "loss": 1.2447, "step": 8942 }, { "epoch": 10.03, "learning_rate": 1.8819718309859155e-05, "loss": 1.1644, "step": 8943 }, { "epoch": 10.03, "learning_rate": 1.881830985915493e-05, "loss": 1.2138, "step": 8944 }, { "epoch": 10.03, "learning_rate": 1.8816901408450704e-05, "loss": 1.049, "step": 8945 }, { "epoch": 10.03, "learning_rate": 1.881549295774648e-05, "loss": 1.2748, "step": 8946 }, { "epoch": 10.03, "learning_rate": 1.8814084507042253e-05, "loss": 1.3532, "step": 8947 }, { "epoch": 10.03, "learning_rate": 1.881267605633803e-05, "loss": 1.1981, "step": 8948 }, { "epoch": 10.03, "learning_rate": 1.8811267605633802e-05, "loss": 1.075, "step": 8949 }, { "epoch": 10.03, "learning_rate": 1.880985915492958e-05, "loss": 1.2657, "step": 8950 }, { "epoch": 10.03, "learning_rate": 1.8808450704225352e-05, "loss": 1.323, "step": 8951 }, { "epoch": 10.04, "learning_rate": 1.8807042253521125e-05, "loss": 1.202, "step": 8952 }, { "epoch": 10.04, "learning_rate": 1.88056338028169e-05, "loss": 1.2753, "step": 8953 }, { "epoch": 10.04, "learning_rate": 1.8804225352112674e-05, "loss": 1.2248, "step": 8954 }, { "epoch": 10.04, "learning_rate": 1.8802816901408454e-05, "loss": 1.0684, "step": 8955 }, { "epoch": 10.04, "learning_rate": 1.8801408450704227e-05, "loss": 1.252, "step": 8956 }, { "epoch": 10.04, "learning_rate": 1.8800000000000003e-05, "loss": 1.2121, "step": 8957 }, { "epoch": 10.04, "learning_rate": 1.8798591549295776e-05, "loss": 1.0777, "step": 8958 }, { "epoch": 10.04, "learning_rate": 1.879718309859155e-05, "loss": 1.1551, "step": 8959 }, { "epoch": 10.04, "learning_rate": 1.8795774647887325e-05, "loss": 1.1001, "step": 8960 }, { "epoch": 10.05, "learning_rate": 1.8794366197183098e-05, "loss": 1.2212, "step": 8961 }, { "epoch": 10.05, "learning_rate": 1.8792957746478874e-05, "loss": 1.2867, "step": 8962 }, { "epoch": 10.05, "learning_rate": 1.8791549295774647e-05, "loss": 1.2169, "step": 8963 }, { "epoch": 10.05, "learning_rate": 1.8790140845070424e-05, "loss": 0.9713, "step": 8964 }, { "epoch": 10.05, "learning_rate": 1.8788732394366197e-05, "loss": 1.2775, "step": 8965 }, { "epoch": 10.05, "learning_rate": 1.8787323943661973e-05, "loss": 0.9772, "step": 8966 }, { "epoch": 10.05, "learning_rate": 1.8785915492957746e-05, "loss": 1.0939, "step": 8967 }, { "epoch": 10.05, "learning_rate": 1.878450704225352e-05, "loss": 1.0097, "step": 8968 }, { "epoch": 10.05, "learning_rate": 1.8783098591549295e-05, "loss": 1.3041, "step": 8969 }, { "epoch": 10.06, "learning_rate": 1.878169014084507e-05, "loss": 1.1597, "step": 8970 }, { "epoch": 10.06, "learning_rate": 1.8780281690140848e-05, "loss": 1.5565, "step": 8971 }, { "epoch": 10.06, "learning_rate": 1.877887323943662e-05, "loss": 1.4652, "step": 8972 }, { "epoch": 10.06, "learning_rate": 1.8777464788732397e-05, "loss": 1.5594, "step": 8973 }, { "epoch": 10.06, "learning_rate": 1.877605633802817e-05, "loss": 1.3952, "step": 8974 }, { "epoch": 10.06, "learning_rate": 1.8774647887323943e-05, "loss": 1.2926, "step": 8975 }, { "epoch": 10.06, "learning_rate": 1.877323943661972e-05, "loss": 1.2435, "step": 8976 }, { "epoch": 10.06, "learning_rate": 1.8771830985915492e-05, "loss": 1.4079, "step": 8977 }, { "epoch": 10.07, "learning_rate": 1.877042253521127e-05, "loss": 1.2924, "step": 8978 }, { "epoch": 10.07, "learning_rate": 1.876901408450704e-05, "loss": 1.2549, "step": 8979 }, { "epoch": 10.07, "learning_rate": 1.8767605633802818e-05, "loss": 1.3102, "step": 8980 }, { "epoch": 10.07, "learning_rate": 1.876619718309859e-05, "loss": 1.1584, "step": 8981 }, { "epoch": 10.07, "learning_rate": 1.8764788732394367e-05, "loss": 1.3289, "step": 8982 }, { "epoch": 10.07, "learning_rate": 1.876338028169014e-05, "loss": 1.2392, "step": 8983 }, { "epoch": 10.07, "learning_rate": 1.8761971830985913e-05, "loss": 1.0854, "step": 8984 }, { "epoch": 10.07, "learning_rate": 1.8760563380281693e-05, "loss": 1.1505, "step": 8985 }, { "epoch": 10.07, "learning_rate": 1.8759154929577466e-05, "loss": 1.3043, "step": 8986 }, { "epoch": 10.08, "learning_rate": 1.8757746478873242e-05, "loss": 1.1568, "step": 8987 }, { "epoch": 10.08, "learning_rate": 1.8756338028169015e-05, "loss": 1.0146, "step": 8988 }, { "epoch": 10.08, "learning_rate": 1.875492957746479e-05, "loss": 1.2638, "step": 8989 }, { "epoch": 10.08, "learning_rate": 1.8753521126760564e-05, "loss": 1.1323, "step": 8990 }, { "epoch": 10.08, "learning_rate": 1.8752112676056337e-05, "loss": 0.9769, "step": 8991 }, { "epoch": 10.08, "learning_rate": 1.8750704225352114e-05, "loss": 1.0449, "step": 8992 }, { "epoch": 10.08, "learning_rate": 1.8749295774647887e-05, "loss": 1.2671, "step": 8993 }, { "epoch": 10.08, "learning_rate": 1.8747887323943663e-05, "loss": 1.2195, "step": 8994 }, { "epoch": 10.08, "learning_rate": 1.8746478873239436e-05, "loss": 1.1586, "step": 8995 }, { "epoch": 10.09, "learning_rate": 1.8745070422535212e-05, "loss": 1.1027, "step": 8996 }, { "epoch": 10.09, "learning_rate": 1.8743661971830985e-05, "loss": 1.1241, "step": 8997 }, { "epoch": 10.09, "learning_rate": 1.874225352112676e-05, "loss": 1.1465, "step": 8998 }, { "epoch": 10.09, "learning_rate": 1.8740845070422534e-05, "loss": 1.1027, "step": 8999 }, { "epoch": 10.09, "learning_rate": 1.8739436619718307e-05, "loss": 1.2606, "step": 9000 }, { "epoch": 10.09, "eval_loss": 1.4472116231918335, "eval_runtime": 832.1648, "eval_samples_per_second": 3.175, "eval_steps_per_second": 0.398, "eval_wer": 0.24873066243554146, "step": 9000 }, { "epoch": 10.09, "learning_rate": 1.8738028169014087e-05, "loss": 1.2827, "step": 9001 }, { "epoch": 10.09, "learning_rate": 1.873661971830986e-05, "loss": 1.3478, "step": 9002 }, { "epoch": 10.09, "learning_rate": 1.8735211267605636e-05, "loss": 1.1723, "step": 9003 }, { "epoch": 10.09, "learning_rate": 1.873380281690141e-05, "loss": 1.3395, "step": 9004 }, { "epoch": 10.1, "learning_rate": 1.8732394366197186e-05, "loss": 1.2961, "step": 9005 }, { "epoch": 10.1, "learning_rate": 1.873098591549296e-05, "loss": 1.0371, "step": 9006 }, { "epoch": 10.1, "learning_rate": 1.872957746478873e-05, "loss": 1.0171, "step": 9007 }, { "epoch": 10.1, "learning_rate": 1.8728169014084508e-05, "loss": 1.2466, "step": 9008 }, { "epoch": 10.1, "learning_rate": 1.872676056338028e-05, "loss": 0.9554, "step": 9009 }, { "epoch": 10.1, "learning_rate": 1.8725352112676057e-05, "loss": 0.9231, "step": 9010 }, { "epoch": 10.1, "learning_rate": 1.872394366197183e-05, "loss": 1.0929, "step": 9011 }, { "epoch": 10.1, "learning_rate": 1.8722535211267606e-05, "loss": 1.06, "step": 9012 }, { "epoch": 10.1, "learning_rate": 1.872112676056338e-05, "loss": 1.183, "step": 9013 }, { "epoch": 10.11, "learning_rate": 1.8719718309859156e-05, "loss": 1.1263, "step": 9014 }, { "epoch": 10.11, "learning_rate": 1.871830985915493e-05, "loss": 1.0576, "step": 9015 }, { "epoch": 10.11, "learning_rate": 1.8716901408450705e-05, "loss": 1.0275, "step": 9016 }, { "epoch": 10.11, "learning_rate": 1.871549295774648e-05, "loss": 1.0009, "step": 9017 }, { "epoch": 10.11, "learning_rate": 1.8714084507042254e-05, "loss": 1.0115, "step": 9018 }, { "epoch": 10.11, "learning_rate": 1.871267605633803e-05, "loss": 1.063, "step": 9019 }, { "epoch": 10.11, "learning_rate": 1.8711267605633803e-05, "loss": 1.3989, "step": 9020 }, { "epoch": 10.11, "learning_rate": 1.870985915492958e-05, "loss": 1.534, "step": 9021 }, { "epoch": 10.11, "learning_rate": 1.8708450704225353e-05, "loss": 1.4866, "step": 9022 }, { "epoch": 10.12, "learning_rate": 1.8707042253521126e-05, "loss": 1.5636, "step": 9023 }, { "epoch": 10.12, "learning_rate": 1.8705633802816902e-05, "loss": 1.4033, "step": 9024 }, { "epoch": 10.12, "learning_rate": 1.8704225352112675e-05, "loss": 1.2775, "step": 9025 }, { "epoch": 10.12, "learning_rate": 1.870281690140845e-05, "loss": 1.2313, "step": 9026 }, { "epoch": 10.12, "learning_rate": 1.8701408450704224e-05, "loss": 1.223, "step": 9027 }, { "epoch": 10.12, "learning_rate": 1.87e-05, "loss": 1.2786, "step": 9028 }, { "epoch": 10.12, "learning_rate": 1.8698591549295773e-05, "loss": 1.3098, "step": 9029 }, { "epoch": 10.12, "learning_rate": 1.869718309859155e-05, "loss": 1.2006, "step": 9030 }, { "epoch": 10.12, "learning_rate": 1.8695774647887326e-05, "loss": 1.446, "step": 9031 }, { "epoch": 10.13, "learning_rate": 1.86943661971831e-05, "loss": 1.186, "step": 9032 }, { "epoch": 10.13, "learning_rate": 1.8692957746478875e-05, "loss": 1.2389, "step": 9033 }, { "epoch": 10.13, "learning_rate": 1.869154929577465e-05, "loss": 1.3305, "step": 9034 }, { "epoch": 10.13, "learning_rate": 1.8690140845070425e-05, "loss": 1.2234, "step": 9035 }, { "epoch": 10.13, "learning_rate": 1.8688732394366198e-05, "loss": 1.1313, "step": 9036 }, { "epoch": 10.13, "learning_rate": 1.8687323943661974e-05, "loss": 1.087, "step": 9037 }, { "epoch": 10.13, "learning_rate": 1.8685915492957747e-05, "loss": 1.2596, "step": 9038 }, { "epoch": 10.13, "learning_rate": 1.868450704225352e-05, "loss": 1.2318, "step": 9039 }, { "epoch": 10.13, "learning_rate": 1.8683098591549296e-05, "loss": 1.0832, "step": 9040 }, { "epoch": 10.14, "learning_rate": 1.868169014084507e-05, "loss": 0.9594, "step": 9041 }, { "epoch": 10.14, "learning_rate": 1.8680281690140845e-05, "loss": 1.0842, "step": 9042 }, { "epoch": 10.14, "learning_rate": 1.867887323943662e-05, "loss": 1.282, "step": 9043 }, { "epoch": 10.14, "learning_rate": 1.8677464788732395e-05, "loss": 1.1916, "step": 9044 }, { "epoch": 10.14, "learning_rate": 1.8676056338028168e-05, "loss": 1.0426, "step": 9045 }, { "epoch": 10.14, "learning_rate": 1.8674647887323944e-05, "loss": 1.1666, "step": 9046 }, { "epoch": 10.14, "learning_rate": 1.867323943661972e-05, "loss": 1.2237, "step": 9047 }, { "epoch": 10.14, "learning_rate": 1.8671830985915493e-05, "loss": 1.2152, "step": 9048 }, { "epoch": 10.14, "learning_rate": 1.867042253521127e-05, "loss": 1.119, "step": 9049 }, { "epoch": 10.15, "learning_rate": 1.8669014084507043e-05, "loss": 1.4169, "step": 9050 }, { "epoch": 10.15, "learning_rate": 1.866760563380282e-05, "loss": 1.0873, "step": 9051 }, { "epoch": 10.15, "learning_rate": 1.8666197183098592e-05, "loss": 1.1601, "step": 9052 }, { "epoch": 10.15, "learning_rate": 1.8664788732394368e-05, "loss": 1.2969, "step": 9053 }, { "epoch": 10.15, "learning_rate": 1.866338028169014e-05, "loss": 1.1302, "step": 9054 }, { "epoch": 10.15, "learning_rate": 1.8661971830985914e-05, "loss": 1.2812, "step": 9055 }, { "epoch": 10.15, "learning_rate": 1.866056338028169e-05, "loss": 1.142, "step": 9056 }, { "epoch": 10.15, "learning_rate": 1.8659154929577463e-05, "loss": 1.145, "step": 9057 }, { "epoch": 10.15, "learning_rate": 1.865774647887324e-05, "loss": 1.3508, "step": 9058 }, { "epoch": 10.16, "learning_rate": 1.8656338028169013e-05, "loss": 1.1475, "step": 9059 }, { "epoch": 10.16, "learning_rate": 1.865492957746479e-05, "loss": 1.2119, "step": 9060 }, { "epoch": 10.16, "learning_rate": 1.8653521126760565e-05, "loss": 1.1652, "step": 9061 }, { "epoch": 10.16, "learning_rate": 1.8652112676056338e-05, "loss": 1.1035, "step": 9062 }, { "epoch": 10.16, "learning_rate": 1.8650704225352115e-05, "loss": 1.0231, "step": 9063 }, { "epoch": 10.16, "learning_rate": 1.8649295774647887e-05, "loss": 1.1374, "step": 9064 }, { "epoch": 10.16, "learning_rate": 1.8647887323943664e-05, "loss": 1.2264, "step": 9065 }, { "epoch": 10.16, "learning_rate": 1.8646478873239437e-05, "loss": 0.9684, "step": 9066 }, { "epoch": 10.16, "learning_rate": 1.8645070422535213e-05, "loss": 0.9866, "step": 9067 }, { "epoch": 10.17, "learning_rate": 1.8643661971830986e-05, "loss": 1.0588, "step": 9068 }, { "epoch": 10.17, "learning_rate": 1.8642253521126762e-05, "loss": 0.9454, "step": 9069 }, { "epoch": 10.17, "learning_rate": 1.8640845070422535e-05, "loss": 1.004, "step": 9070 }, { "epoch": 10.17, "learning_rate": 1.8639436619718308e-05, "loss": 1.5336, "step": 9071 }, { "epoch": 10.17, "learning_rate": 1.8638028169014085e-05, "loss": 1.4744, "step": 9072 }, { "epoch": 10.17, "learning_rate": 1.8636619718309858e-05, "loss": 1.5373, "step": 9073 }, { "epoch": 10.17, "learning_rate": 1.8635211267605634e-05, "loss": 1.4553, "step": 9074 }, { "epoch": 10.17, "learning_rate": 1.8633802816901407e-05, "loss": 1.2394, "step": 9075 }, { "epoch": 10.17, "learning_rate": 1.8632394366197187e-05, "loss": 1.3432, "step": 9076 }, { "epoch": 10.18, "learning_rate": 1.863098591549296e-05, "loss": 1.4128, "step": 9077 }, { "epoch": 10.18, "learning_rate": 1.8629577464788732e-05, "loss": 1.2859, "step": 9078 }, { "epoch": 10.18, "learning_rate": 1.862816901408451e-05, "loss": 1.1762, "step": 9079 }, { "epoch": 10.18, "learning_rate": 1.862676056338028e-05, "loss": 1.1052, "step": 9080 }, { "epoch": 10.18, "learning_rate": 1.8625352112676058e-05, "loss": 1.3617, "step": 9081 }, { "epoch": 10.18, "learning_rate": 1.862394366197183e-05, "loss": 1.3351, "step": 9082 }, { "epoch": 10.18, "learning_rate": 1.8622535211267607e-05, "loss": 1.4672, "step": 9083 }, { "epoch": 10.18, "learning_rate": 1.862112676056338e-05, "loss": 1.2061, "step": 9084 }, { "epoch": 10.18, "learning_rate": 1.8619718309859157e-05, "loss": 1.2846, "step": 9085 }, { "epoch": 10.19, "learning_rate": 1.861830985915493e-05, "loss": 1.3783, "step": 9086 }, { "epoch": 10.19, "learning_rate": 1.8616901408450702e-05, "loss": 1.0916, "step": 9087 }, { "epoch": 10.19, "learning_rate": 1.861549295774648e-05, "loss": 1.2728, "step": 9088 }, { "epoch": 10.19, "learning_rate": 1.8614084507042252e-05, "loss": 1.3061, "step": 9089 }, { "epoch": 10.19, "learning_rate": 1.8612676056338028e-05, "loss": 1.1756, "step": 9090 }, { "epoch": 10.19, "learning_rate": 1.8611267605633804e-05, "loss": 1.2508, "step": 9091 }, { "epoch": 10.19, "learning_rate": 1.860985915492958e-05, "loss": 1.2359, "step": 9092 }, { "epoch": 10.19, "learning_rate": 1.8608450704225354e-05, "loss": 1.2069, "step": 9093 }, { "epoch": 10.2, "learning_rate": 1.8607042253521127e-05, "loss": 1.1086, "step": 9094 }, { "epoch": 10.2, "learning_rate": 1.8605633802816903e-05, "loss": 1.2892, "step": 9095 }, { "epoch": 10.2, "learning_rate": 1.8604225352112676e-05, "loss": 1.1127, "step": 9096 }, { "epoch": 10.2, "learning_rate": 1.8602816901408452e-05, "loss": 1.1954, "step": 9097 }, { "epoch": 10.2, "learning_rate": 1.8601408450704225e-05, "loss": 1.1323, "step": 9098 }, { "epoch": 10.2, "learning_rate": 1.86e-05, "loss": 1.0322, "step": 9099 }, { "epoch": 10.2, "learning_rate": 1.8598591549295774e-05, "loss": 1.2818, "step": 9100 }, { "epoch": 10.2, "learning_rate": 1.859718309859155e-05, "loss": 1.0078, "step": 9101 }, { "epoch": 10.2, "learning_rate": 1.8595774647887324e-05, "loss": 1.1483, "step": 9102 }, { "epoch": 10.21, "learning_rate": 1.8594366197183097e-05, "loss": 1.1317, "step": 9103 }, { "epoch": 10.21, "learning_rate": 1.8592957746478873e-05, "loss": 1.1793, "step": 9104 }, { "epoch": 10.21, "learning_rate": 1.8591549295774646e-05, "loss": 1.2274, "step": 9105 }, { "epoch": 10.21, "learning_rate": 1.8590140845070426e-05, "loss": 1.0464, "step": 9106 }, { "epoch": 10.21, "learning_rate": 1.85887323943662e-05, "loss": 1.2501, "step": 9107 }, { "epoch": 10.21, "learning_rate": 1.8587323943661975e-05, "loss": 1.1199, "step": 9108 }, { "epoch": 10.21, "learning_rate": 1.8585915492957748e-05, "loss": 1.1868, "step": 9109 }, { "epoch": 10.21, "learning_rate": 1.858450704225352e-05, "loss": 1.1481, "step": 9110 }, { "epoch": 10.21, "learning_rate": 1.8583098591549297e-05, "loss": 1.1389, "step": 9111 }, { "epoch": 10.22, "learning_rate": 1.858169014084507e-05, "loss": 1.0358, "step": 9112 }, { "epoch": 10.22, "learning_rate": 1.8580281690140846e-05, "loss": 0.864, "step": 9113 }, { "epoch": 10.22, "learning_rate": 1.857887323943662e-05, "loss": 1.0382, "step": 9114 }, { "epoch": 10.22, "learning_rate": 1.8577464788732396e-05, "loss": 1.0482, "step": 9115 }, { "epoch": 10.22, "learning_rate": 1.857605633802817e-05, "loss": 1.0864, "step": 9116 }, { "epoch": 10.22, "learning_rate": 1.8574647887323945e-05, "loss": 0.9806, "step": 9117 }, { "epoch": 10.22, "learning_rate": 1.8573239436619718e-05, "loss": 1.2756, "step": 9118 }, { "epoch": 10.22, "learning_rate": 1.857183098591549e-05, "loss": 0.9534, "step": 9119 }, { "epoch": 10.22, "learning_rate": 1.8570422535211267e-05, "loss": 1.15, "step": 9120 }, { "epoch": 10.23, "learning_rate": 1.856901408450704e-05, "loss": 1.5151, "step": 9121 }, { "epoch": 10.23, "learning_rate": 1.856760563380282e-05, "loss": 1.5303, "step": 9122 }, { "epoch": 10.23, "learning_rate": 1.8566197183098593e-05, "loss": 1.256, "step": 9123 }, { "epoch": 10.23, "learning_rate": 1.856478873239437e-05, "loss": 1.1116, "step": 9124 }, { "epoch": 10.23, "learning_rate": 1.8563380281690142e-05, "loss": 1.308, "step": 9125 }, { "epoch": 10.23, "learning_rate": 1.8561971830985915e-05, "loss": 1.2719, "step": 9126 }, { "epoch": 10.23, "learning_rate": 1.856056338028169e-05, "loss": 1.2815, "step": 9127 }, { "epoch": 10.23, "learning_rate": 1.8559154929577464e-05, "loss": 1.371, "step": 9128 }, { "epoch": 10.23, "learning_rate": 1.855774647887324e-05, "loss": 1.1419, "step": 9129 }, { "epoch": 10.24, "learning_rate": 1.8556338028169014e-05, "loss": 1.4819, "step": 9130 }, { "epoch": 10.24, "learning_rate": 1.855492957746479e-05, "loss": 1.207, "step": 9131 }, { "epoch": 10.24, "learning_rate": 1.8553521126760563e-05, "loss": 1.3207, "step": 9132 }, { "epoch": 10.24, "learning_rate": 1.855211267605634e-05, "loss": 1.2315, "step": 9133 }, { "epoch": 10.24, "learning_rate": 1.8550704225352112e-05, "loss": 1.1712, "step": 9134 }, { "epoch": 10.24, "learning_rate": 1.8549295774647885e-05, "loss": 1.0736, "step": 9135 }, { "epoch": 10.24, "learning_rate": 1.854788732394366e-05, "loss": 1.1095, "step": 9136 }, { "epoch": 10.24, "learning_rate": 1.8546478873239438e-05, "loss": 1.1829, "step": 9137 }, { "epoch": 10.24, "learning_rate": 1.8545070422535214e-05, "loss": 1.2668, "step": 9138 }, { "epoch": 10.25, "learning_rate": 1.8543661971830987e-05, "loss": 1.2354, "step": 9139 }, { "epoch": 10.25, "learning_rate": 1.8542253521126763e-05, "loss": 1.0995, "step": 9140 }, { "epoch": 10.25, "learning_rate": 1.8540845070422536e-05, "loss": 1.2768, "step": 9141 }, { "epoch": 10.25, "learning_rate": 1.853943661971831e-05, "loss": 0.9595, "step": 9142 }, { "epoch": 10.25, "learning_rate": 1.8538028169014086e-05, "loss": 1.2464, "step": 9143 }, { "epoch": 10.25, "learning_rate": 1.853661971830986e-05, "loss": 1.2882, "step": 9144 }, { "epoch": 10.25, "learning_rate": 1.8535211267605635e-05, "loss": 1.0928, "step": 9145 }, { "epoch": 10.25, "learning_rate": 1.8533802816901408e-05, "loss": 1.1985, "step": 9146 }, { "epoch": 10.25, "learning_rate": 1.8532394366197184e-05, "loss": 1.1614, "step": 9147 }, { "epoch": 10.26, "learning_rate": 1.8530985915492957e-05, "loss": 1.2943, "step": 9148 }, { "epoch": 10.26, "learning_rate": 1.8529577464788733e-05, "loss": 1.3358, "step": 9149 }, { "epoch": 10.26, "learning_rate": 1.8528169014084506e-05, "loss": 1.0369, "step": 9150 }, { "epoch": 10.26, "learning_rate": 1.852676056338028e-05, "loss": 1.1542, "step": 9151 }, { "epoch": 10.26, "learning_rate": 1.852535211267606e-05, "loss": 1.0088, "step": 9152 }, { "epoch": 10.26, "learning_rate": 1.8523943661971832e-05, "loss": 1.0025, "step": 9153 }, { "epoch": 10.26, "learning_rate": 1.8522535211267608e-05, "loss": 1.088, "step": 9154 }, { "epoch": 10.26, "learning_rate": 1.852112676056338e-05, "loss": 1.1588, "step": 9155 }, { "epoch": 10.26, "learning_rate": 1.8519718309859158e-05, "loss": 1.2959, "step": 9156 }, { "epoch": 10.27, "learning_rate": 1.851830985915493e-05, "loss": 1.1353, "step": 9157 }, { "epoch": 10.27, "learning_rate": 1.8516901408450703e-05, "loss": 1.1334, "step": 9158 }, { "epoch": 10.27, "learning_rate": 1.851549295774648e-05, "loss": 0.9851, "step": 9159 }, { "epoch": 10.27, "learning_rate": 1.8514084507042253e-05, "loss": 1.0721, "step": 9160 }, { "epoch": 10.27, "learning_rate": 1.851267605633803e-05, "loss": 1.283, "step": 9161 }, { "epoch": 10.27, "learning_rate": 1.8511267605633802e-05, "loss": 1.0422, "step": 9162 }, { "epoch": 10.27, "learning_rate": 1.8509859154929578e-05, "loss": 1.0164, "step": 9163 }, { "epoch": 10.27, "learning_rate": 1.850845070422535e-05, "loss": 0.9578, "step": 9164 }, { "epoch": 10.27, "learning_rate": 1.8507042253521128e-05, "loss": 1.1638, "step": 9165 }, { "epoch": 10.28, "learning_rate": 1.85056338028169e-05, "loss": 1.3549, "step": 9166 }, { "epoch": 10.28, "learning_rate": 1.8504225352112677e-05, "loss": 1.1543, "step": 9167 }, { "epoch": 10.28, "learning_rate": 1.8502816901408453e-05, "loss": 1.0532, "step": 9168 }, { "epoch": 10.28, "learning_rate": 1.8501408450704226e-05, "loss": 1.1482, "step": 9169 }, { "epoch": 10.28, "learning_rate": 1.8500000000000002e-05, "loss": 1.0494, "step": 9170 }, { "epoch": 10.28, "learning_rate": 1.8498591549295775e-05, "loss": 1.4067, "step": 9171 }, { "epoch": 10.28, "learning_rate": 1.8497183098591552e-05, "loss": 1.3957, "step": 9172 }, { "epoch": 10.28, "learning_rate": 1.8495774647887325e-05, "loss": 1.3468, "step": 9173 }, { "epoch": 10.28, "learning_rate": 1.8494366197183098e-05, "loss": 1.2201, "step": 9174 }, { "epoch": 10.29, "learning_rate": 1.8492957746478874e-05, "loss": 1.2174, "step": 9175 }, { "epoch": 10.29, "learning_rate": 1.8491549295774647e-05, "loss": 1.3268, "step": 9176 }, { "epoch": 10.29, "learning_rate": 1.8490140845070423e-05, "loss": 1.0057, "step": 9177 }, { "epoch": 10.29, "learning_rate": 1.8488732394366196e-05, "loss": 1.3369, "step": 9178 }, { "epoch": 10.29, "learning_rate": 1.8487323943661972e-05, "loss": 1.2544, "step": 9179 }, { "epoch": 10.29, "learning_rate": 1.8485915492957745e-05, "loss": 1.1861, "step": 9180 }, { "epoch": 10.29, "learning_rate": 1.848450704225352e-05, "loss": 1.3601, "step": 9181 }, { "epoch": 10.29, "learning_rate": 1.8483098591549298e-05, "loss": 1.1504, "step": 9182 }, { "epoch": 10.29, "learning_rate": 1.848169014084507e-05, "loss": 1.4868, "step": 9183 }, { "epoch": 10.3, "learning_rate": 1.8480281690140847e-05, "loss": 1.1275, "step": 9184 }, { "epoch": 10.3, "learning_rate": 1.847887323943662e-05, "loss": 1.1881, "step": 9185 }, { "epoch": 10.3, "learning_rate": 1.8477464788732397e-05, "loss": 1.0447, "step": 9186 }, { "epoch": 10.3, "learning_rate": 1.847605633802817e-05, "loss": 1.294, "step": 9187 }, { "epoch": 10.3, "learning_rate": 1.8474647887323946e-05, "loss": 1.1432, "step": 9188 }, { "epoch": 10.3, "learning_rate": 1.847323943661972e-05, "loss": 1.0325, "step": 9189 }, { "epoch": 10.3, "learning_rate": 1.8471830985915492e-05, "loss": 1.2813, "step": 9190 }, { "epoch": 10.3, "learning_rate": 1.8470422535211268e-05, "loss": 1.1855, "step": 9191 }, { "epoch": 10.3, "learning_rate": 1.846901408450704e-05, "loss": 1.235, "step": 9192 }, { "epoch": 10.31, "learning_rate": 1.8467605633802817e-05, "loss": 1.2099, "step": 9193 }, { "epoch": 10.31, "learning_rate": 1.846619718309859e-05, "loss": 1.2832, "step": 9194 }, { "epoch": 10.31, "learning_rate": 1.8464788732394367e-05, "loss": 1.1799, "step": 9195 }, { "epoch": 10.31, "learning_rate": 1.846338028169014e-05, "loss": 1.2314, "step": 9196 }, { "epoch": 10.31, "learning_rate": 1.8461971830985916e-05, "loss": 1.1126, "step": 9197 }, { "epoch": 10.31, "learning_rate": 1.8460563380281692e-05, "loss": 1.1377, "step": 9198 }, { "epoch": 10.31, "learning_rate": 1.8459154929577465e-05, "loss": 1.2424, "step": 9199 }, { "epoch": 10.31, "learning_rate": 1.845774647887324e-05, "loss": 1.0989, "step": 9200 }, { "epoch": 10.32, "learning_rate": 1.8456338028169015e-05, "loss": 1.1737, "step": 9201 }, { "epoch": 10.32, "learning_rate": 1.845492957746479e-05, "loss": 1.1011, "step": 9202 }, { "epoch": 10.32, "learning_rate": 1.8453521126760564e-05, "loss": 1.1338, "step": 9203 }, { "epoch": 10.32, "learning_rate": 1.845211267605634e-05, "loss": 1.2003, "step": 9204 }, { "epoch": 10.32, "learning_rate": 1.8450704225352113e-05, "loss": 1.3306, "step": 9205 }, { "epoch": 10.32, "learning_rate": 1.8449295774647886e-05, "loss": 1.1673, "step": 9206 }, { "epoch": 10.32, "learning_rate": 1.8447887323943662e-05, "loss": 1.1977, "step": 9207 }, { "epoch": 10.32, "learning_rate": 1.8446478873239435e-05, "loss": 1.0825, "step": 9208 }, { "epoch": 10.32, "learning_rate": 1.844507042253521e-05, "loss": 1.3384, "step": 9209 }, { "epoch": 10.33, "learning_rate": 1.8443661971830985e-05, "loss": 1.1961, "step": 9210 }, { "epoch": 10.33, "learning_rate": 1.844225352112676e-05, "loss": 0.8826, "step": 9211 }, { "epoch": 10.33, "learning_rate": 1.8440845070422537e-05, "loss": 1.2264, "step": 9212 }, { "epoch": 10.33, "learning_rate": 1.843943661971831e-05, "loss": 1.178, "step": 9213 }, { "epoch": 10.33, "learning_rate": 1.8438028169014086e-05, "loss": 1.1655, "step": 9214 }, { "epoch": 10.33, "learning_rate": 1.843661971830986e-05, "loss": 1.1565, "step": 9215 }, { "epoch": 10.33, "learning_rate": 1.8435211267605636e-05, "loss": 1.2296, "step": 9216 }, { "epoch": 10.33, "learning_rate": 1.843380281690141e-05, "loss": 0.8579, "step": 9217 }, { "epoch": 10.33, "learning_rate": 1.8432394366197185e-05, "loss": 1.2721, "step": 9218 }, { "epoch": 10.34, "learning_rate": 1.8430985915492958e-05, "loss": 1.2417, "step": 9219 }, { "epoch": 10.34, "learning_rate": 1.8429577464788734e-05, "loss": 1.0166, "step": 9220 }, { "epoch": 10.34, "learning_rate": 1.8428169014084507e-05, "loss": 1.4874, "step": 9221 }, { "epoch": 10.34, "learning_rate": 1.842676056338028e-05, "loss": 1.4376, "step": 9222 }, { "epoch": 10.34, "learning_rate": 1.8425352112676057e-05, "loss": 1.327, "step": 9223 }, { "epoch": 10.34, "learning_rate": 1.842394366197183e-05, "loss": 1.2702, "step": 9224 }, { "epoch": 10.34, "learning_rate": 1.8422535211267606e-05, "loss": 1.3216, "step": 9225 }, { "epoch": 10.34, "learning_rate": 1.842112676056338e-05, "loss": 1.3244, "step": 9226 }, { "epoch": 10.34, "learning_rate": 1.841971830985916e-05, "loss": 1.1196, "step": 9227 }, { "epoch": 10.35, "learning_rate": 1.841830985915493e-05, "loss": 1.2428, "step": 9228 }, { "epoch": 10.35, "learning_rate": 1.8416901408450704e-05, "loss": 1.4356, "step": 9229 }, { "epoch": 10.35, "learning_rate": 1.841549295774648e-05, "loss": 1.0527, "step": 9230 }, { "epoch": 10.35, "learning_rate": 1.8414084507042254e-05, "loss": 1.2814, "step": 9231 }, { "epoch": 10.35, "learning_rate": 1.841267605633803e-05, "loss": 1.2147, "step": 9232 }, { "epoch": 10.35, "learning_rate": 1.8411267605633803e-05, "loss": 1.337, "step": 9233 }, { "epoch": 10.35, "learning_rate": 1.840985915492958e-05, "loss": 1.2071, "step": 9234 }, { "epoch": 10.35, "learning_rate": 1.8408450704225352e-05, "loss": 1.4012, "step": 9235 }, { "epoch": 10.35, "learning_rate": 1.840704225352113e-05, "loss": 1.1536, "step": 9236 }, { "epoch": 10.36, "learning_rate": 1.84056338028169e-05, "loss": 1.0996, "step": 9237 }, { "epoch": 10.36, "learning_rate": 1.8404225352112674e-05, "loss": 1.2653, "step": 9238 }, { "epoch": 10.36, "learning_rate": 1.840281690140845e-05, "loss": 1.0544, "step": 9239 }, { "epoch": 10.36, "learning_rate": 1.8401408450704224e-05, "loss": 1.0546, "step": 9240 }, { "epoch": 10.36, "learning_rate": 1.84e-05, "loss": 1.0574, "step": 9241 }, { "epoch": 10.36, "learning_rate": 1.8398591549295773e-05, "loss": 0.9466, "step": 9242 }, { "epoch": 10.36, "learning_rate": 1.8397183098591553e-05, "loss": 1.0318, "step": 9243 }, { "epoch": 10.36, "learning_rate": 1.8395774647887326e-05, "loss": 1.0805, "step": 9244 }, { "epoch": 10.36, "learning_rate": 1.83943661971831e-05, "loss": 1.0252, "step": 9245 }, { "epoch": 10.37, "learning_rate": 1.8392957746478875e-05, "loss": 1.1456, "step": 9246 }, { "epoch": 10.37, "learning_rate": 1.8391549295774648e-05, "loss": 1.1022, "step": 9247 }, { "epoch": 10.37, "learning_rate": 1.8390140845070424e-05, "loss": 1.1979, "step": 9248 }, { "epoch": 10.37, "learning_rate": 1.8388732394366197e-05, "loss": 1.076, "step": 9249 }, { "epoch": 10.37, "learning_rate": 1.8387323943661973e-05, "loss": 1.1222, "step": 9250 }, { "epoch": 10.37, "learning_rate": 1.8385915492957746e-05, "loss": 1.2303, "step": 9251 }, { "epoch": 10.37, "learning_rate": 1.8384507042253523e-05, "loss": 1.1256, "step": 9252 }, { "epoch": 10.37, "learning_rate": 1.8383098591549296e-05, "loss": 0.9882, "step": 9253 }, { "epoch": 10.37, "learning_rate": 1.838169014084507e-05, "loss": 1.1262, "step": 9254 }, { "epoch": 10.38, "learning_rate": 1.8380281690140845e-05, "loss": 1.1941, "step": 9255 }, { "epoch": 10.38, "learning_rate": 1.8378873239436618e-05, "loss": 1.2806, "step": 9256 }, { "epoch": 10.38, "learning_rate": 1.8377464788732394e-05, "loss": 1.1759, "step": 9257 }, { "epoch": 10.38, "learning_rate": 1.837605633802817e-05, "loss": 0.9166, "step": 9258 }, { "epoch": 10.38, "learning_rate": 1.8374647887323947e-05, "loss": 1.0325, "step": 9259 }, { "epoch": 10.38, "learning_rate": 1.837323943661972e-05, "loss": 1.2013, "step": 9260 }, { "epoch": 10.38, "learning_rate": 1.8371830985915493e-05, "loss": 1.2604, "step": 9261 }, { "epoch": 10.38, "learning_rate": 1.837042253521127e-05, "loss": 0.9677, "step": 9262 }, { "epoch": 10.38, "learning_rate": 1.8369014084507042e-05, "loss": 1.0568, "step": 9263 }, { "epoch": 10.39, "learning_rate": 1.836760563380282e-05, "loss": 1.2455, "step": 9264 }, { "epoch": 10.39, "learning_rate": 1.836619718309859e-05, "loss": 0.8606, "step": 9265 }, { "epoch": 10.39, "learning_rate": 1.8364788732394368e-05, "loss": 0.9782, "step": 9266 }, { "epoch": 10.39, "learning_rate": 1.836338028169014e-05, "loss": 1.1201, "step": 9267 }, { "epoch": 10.39, "learning_rate": 1.8361971830985917e-05, "loss": 1.3154, "step": 9268 }, { "epoch": 10.39, "learning_rate": 1.836056338028169e-05, "loss": 1.0124, "step": 9269 }, { "epoch": 10.39, "learning_rate": 1.8359154929577463e-05, "loss": 1.0262, "step": 9270 }, { "epoch": 10.39, "learning_rate": 1.835774647887324e-05, "loss": 1.4869, "step": 9271 }, { "epoch": 10.39, "learning_rate": 1.8356338028169012e-05, "loss": 1.4734, "step": 9272 }, { "epoch": 10.4, "learning_rate": 1.8354929577464792e-05, "loss": 1.4858, "step": 9273 }, { "epoch": 10.4, "learning_rate": 1.8353521126760565e-05, "loss": 1.3446, "step": 9274 }, { "epoch": 10.4, "learning_rate": 1.835211267605634e-05, "loss": 1.3187, "step": 9275 }, { "epoch": 10.4, "learning_rate": 1.8350704225352114e-05, "loss": 1.223, "step": 9276 }, { "epoch": 10.4, "learning_rate": 1.8349295774647887e-05, "loss": 1.0408, "step": 9277 }, { "epoch": 10.4, "learning_rate": 1.8347887323943663e-05, "loss": 1.2552, "step": 9278 }, { "epoch": 10.4, "learning_rate": 1.8346478873239436e-05, "loss": 1.096, "step": 9279 }, { "epoch": 10.4, "learning_rate": 1.8345070422535213e-05, "loss": 1.1705, "step": 9280 }, { "epoch": 10.4, "learning_rate": 1.8343661971830986e-05, "loss": 1.1002, "step": 9281 }, { "epoch": 10.41, "learning_rate": 1.8342253521126762e-05, "loss": 1.133, "step": 9282 }, { "epoch": 10.41, "learning_rate": 1.8340845070422535e-05, "loss": 1.3488, "step": 9283 }, { "epoch": 10.41, "learning_rate": 1.833943661971831e-05, "loss": 1.3514, "step": 9284 }, { "epoch": 10.41, "learning_rate": 1.8338028169014084e-05, "loss": 1.1879, "step": 9285 }, { "epoch": 10.41, "learning_rate": 1.8336619718309857e-05, "loss": 1.217, "step": 9286 }, { "epoch": 10.41, "learning_rate": 1.8335211267605633e-05, "loss": 1.1136, "step": 9287 }, { "epoch": 10.41, "learning_rate": 1.833380281690141e-05, "loss": 1.0737, "step": 9288 }, { "epoch": 10.41, "learning_rate": 1.8332394366197186e-05, "loss": 1.086, "step": 9289 }, { "epoch": 10.41, "learning_rate": 1.833098591549296e-05, "loss": 1.3196, "step": 9290 }, { "epoch": 10.42, "learning_rate": 1.8329577464788735e-05, "loss": 1.0807, "step": 9291 }, { "epoch": 10.42, "learning_rate": 1.8328169014084508e-05, "loss": 1.3091, "step": 9292 }, { "epoch": 10.42, "learning_rate": 1.832676056338028e-05, "loss": 0.9437, "step": 9293 }, { "epoch": 10.42, "learning_rate": 1.8325352112676057e-05, "loss": 1.0726, "step": 9294 }, { "epoch": 10.42, "learning_rate": 1.832394366197183e-05, "loss": 1.157, "step": 9295 }, { "epoch": 10.42, "learning_rate": 1.8322535211267607e-05, "loss": 1.1328, "step": 9296 }, { "epoch": 10.42, "learning_rate": 1.832112676056338e-05, "loss": 1.1081, "step": 9297 }, { "epoch": 10.42, "learning_rate": 1.8319718309859156e-05, "loss": 1.2429, "step": 9298 }, { "epoch": 10.42, "learning_rate": 1.831830985915493e-05, "loss": 1.1015, "step": 9299 }, { "epoch": 10.43, "learning_rate": 1.8316901408450702e-05, "loss": 1.2526, "step": 9300 }, { "epoch": 10.43, "learning_rate": 1.8315492957746478e-05, "loss": 1.2734, "step": 9301 }, { "epoch": 10.43, "learning_rate": 1.831408450704225e-05, "loss": 0.9103, "step": 9302 }, { "epoch": 10.43, "learning_rate": 1.831267605633803e-05, "loss": 1.0223, "step": 9303 }, { "epoch": 10.43, "learning_rate": 1.8311267605633804e-05, "loss": 1.1561, "step": 9304 }, { "epoch": 10.43, "learning_rate": 1.830985915492958e-05, "loss": 1.3368, "step": 9305 }, { "epoch": 10.43, "learning_rate": 1.8308450704225353e-05, "loss": 1.1492, "step": 9306 }, { "epoch": 10.43, "learning_rate": 1.830704225352113e-05, "loss": 1.0302, "step": 9307 }, { "epoch": 10.43, "learning_rate": 1.8305633802816902e-05, "loss": 1.0536, "step": 9308 }, { "epoch": 10.44, "learning_rate": 1.8304225352112675e-05, "loss": 1.1207, "step": 9309 }, { "epoch": 10.44, "learning_rate": 1.8302816901408452e-05, "loss": 1.188, "step": 9310 }, { "epoch": 10.44, "learning_rate": 1.8301408450704225e-05, "loss": 1.0978, "step": 9311 }, { "epoch": 10.44, "learning_rate": 1.83e-05, "loss": 1.0029, "step": 9312 }, { "epoch": 10.44, "learning_rate": 1.8298591549295774e-05, "loss": 1.0362, "step": 9313 }, { "epoch": 10.44, "learning_rate": 1.829718309859155e-05, "loss": 1.344, "step": 9314 }, { "epoch": 10.44, "learning_rate": 1.8295774647887323e-05, "loss": 1.0979, "step": 9315 }, { "epoch": 10.44, "learning_rate": 1.8294366197183096e-05, "loss": 0.829, "step": 9316 }, { "epoch": 10.45, "learning_rate": 1.8292957746478872e-05, "loss": 0.7715, "step": 9317 }, { "epoch": 10.45, "learning_rate": 1.829154929577465e-05, "loss": 0.8173, "step": 9318 }, { "epoch": 10.45, "learning_rate": 1.8290140845070425e-05, "loss": 1.3745, "step": 9319 }, { "epoch": 10.45, "learning_rate": 1.8288732394366198e-05, "loss": 1.0043, "step": 9320 }, { "epoch": 10.45, "learning_rate": 1.8287323943661974e-05, "loss": 1.5424, "step": 9321 }, { "epoch": 10.45, "learning_rate": 1.8285915492957747e-05, "loss": 1.587, "step": 9322 }, { "epoch": 10.45, "learning_rate": 1.8284507042253524e-05, "loss": 1.2875, "step": 9323 }, { "epoch": 10.45, "learning_rate": 1.8283098591549297e-05, "loss": 1.3978, "step": 9324 }, { "epoch": 10.45, "learning_rate": 1.828169014084507e-05, "loss": 1.1044, "step": 9325 }, { "epoch": 10.46, "learning_rate": 1.8280281690140846e-05, "loss": 1.064, "step": 9326 }, { "epoch": 10.46, "learning_rate": 1.827887323943662e-05, "loss": 1.1777, "step": 9327 }, { "epoch": 10.46, "learning_rate": 1.8277464788732395e-05, "loss": 1.0588, "step": 9328 }, { "epoch": 10.46, "learning_rate": 1.8276056338028168e-05, "loss": 1.3383, "step": 9329 }, { "epoch": 10.46, "learning_rate": 1.8274647887323944e-05, "loss": 1.1845, "step": 9330 }, { "epoch": 10.46, "learning_rate": 1.8273239436619717e-05, "loss": 1.241, "step": 9331 }, { "epoch": 10.46, "learning_rate": 1.827183098591549e-05, "loss": 1.1459, "step": 9332 }, { "epoch": 10.46, "learning_rate": 1.827042253521127e-05, "loss": 1.0723, "step": 9333 }, { "epoch": 10.46, "learning_rate": 1.8269014084507043e-05, "loss": 1.0301, "step": 9334 }, { "epoch": 10.47, "learning_rate": 1.826760563380282e-05, "loss": 1.143, "step": 9335 }, { "epoch": 10.47, "learning_rate": 1.8266197183098592e-05, "loss": 1.1449, "step": 9336 }, { "epoch": 10.47, "learning_rate": 1.826478873239437e-05, "loss": 1.1452, "step": 9337 }, { "epoch": 10.47, "learning_rate": 1.826338028169014e-05, "loss": 1.0309, "step": 9338 }, { "epoch": 10.47, "learning_rate": 1.8261971830985918e-05, "loss": 1.2616, "step": 9339 }, { "epoch": 10.47, "learning_rate": 1.826056338028169e-05, "loss": 1.3055, "step": 9340 }, { "epoch": 10.47, "learning_rate": 1.8259154929577464e-05, "loss": 1.3707, "step": 9341 }, { "epoch": 10.47, "learning_rate": 1.825774647887324e-05, "loss": 1.1516, "step": 9342 }, { "epoch": 10.47, "learning_rate": 1.8256338028169013e-05, "loss": 1.2711, "step": 9343 }, { "epoch": 10.48, "learning_rate": 1.825492957746479e-05, "loss": 1.2184, "step": 9344 }, { "epoch": 10.48, "learning_rate": 1.8253521126760562e-05, "loss": 1.193, "step": 9345 }, { "epoch": 10.48, "learning_rate": 1.825211267605634e-05, "loss": 1.1591, "step": 9346 }, { "epoch": 10.48, "learning_rate": 1.825070422535211e-05, "loss": 1.0502, "step": 9347 }, { "epoch": 10.48, "learning_rate": 1.8249295774647885e-05, "loss": 0.9427, "step": 9348 }, { "epoch": 10.48, "learning_rate": 1.8247887323943664e-05, "loss": 0.9142, "step": 9349 }, { "epoch": 10.48, "learning_rate": 1.8246478873239437e-05, "loss": 1.2521, "step": 9350 }, { "epoch": 10.48, "learning_rate": 1.8245070422535214e-05, "loss": 1.0911, "step": 9351 }, { "epoch": 10.48, "learning_rate": 1.8243661971830986e-05, "loss": 1.1372, "step": 9352 }, { "epoch": 10.49, "learning_rate": 1.8242253521126763e-05, "loss": 1.121, "step": 9353 }, { "epoch": 10.49, "learning_rate": 1.8240845070422536e-05, "loss": 0.9612, "step": 9354 }, { "epoch": 10.49, "learning_rate": 1.8239436619718312e-05, "loss": 1.0488, "step": 9355 }, { "epoch": 10.49, "learning_rate": 1.8238028169014085e-05, "loss": 0.9789, "step": 9356 }, { "epoch": 10.49, "learning_rate": 1.8236619718309858e-05, "loss": 0.9046, "step": 9357 }, { "epoch": 10.49, "learning_rate": 1.8235211267605634e-05, "loss": 1.2518, "step": 9358 }, { "epoch": 10.49, "learning_rate": 1.8233802816901407e-05, "loss": 1.0869, "step": 9359 }, { "epoch": 10.49, "learning_rate": 1.8232394366197184e-05, "loss": 1.1995, "step": 9360 }, { "epoch": 10.49, "learning_rate": 1.8230985915492957e-05, "loss": 1.1922, "step": 9361 }, { "epoch": 10.5, "learning_rate": 1.8229577464788733e-05, "loss": 0.99, "step": 9362 }, { "epoch": 10.5, "learning_rate": 1.8228169014084506e-05, "loss": 1.2112, "step": 9363 }, { "epoch": 10.5, "learning_rate": 1.8226760563380282e-05, "loss": 1.0165, "step": 9364 }, { "epoch": 10.5, "learning_rate": 1.822535211267606e-05, "loss": 1.2075, "step": 9365 }, { "epoch": 10.5, "learning_rate": 1.822394366197183e-05, "loss": 0.8734, "step": 9366 }, { "epoch": 10.5, "learning_rate": 1.8222535211267608e-05, "loss": 0.9869, "step": 9367 }, { "epoch": 10.5, "learning_rate": 1.822112676056338e-05, "loss": 1.2019, "step": 9368 }, { "epoch": 10.5, "learning_rate": 1.8219718309859157e-05, "loss": 0.8466, "step": 9369 }, { "epoch": 10.5, "learning_rate": 1.821830985915493e-05, "loss": 0.8798, "step": 9370 }, { "epoch": 10.51, "learning_rate": 1.8216901408450706e-05, "loss": 1.4321, "step": 9371 }, { "epoch": 10.51, "learning_rate": 1.821549295774648e-05, "loss": 1.2855, "step": 9372 }, { "epoch": 10.51, "learning_rate": 1.8214084507042252e-05, "loss": 1.2934, "step": 9373 }, { "epoch": 10.51, "learning_rate": 1.821267605633803e-05, "loss": 1.2046, "step": 9374 }, { "epoch": 10.51, "learning_rate": 1.82112676056338e-05, "loss": 1.1856, "step": 9375 }, { "epoch": 10.51, "learning_rate": 1.8209859154929578e-05, "loss": 1.1829, "step": 9376 }, { "epoch": 10.51, "learning_rate": 1.820845070422535e-05, "loss": 1.1236, "step": 9377 }, { "epoch": 10.51, "learning_rate": 1.8207042253521127e-05, "loss": 1.2139, "step": 9378 }, { "epoch": 10.51, "learning_rate": 1.8205633802816903e-05, "loss": 1.1914, "step": 9379 }, { "epoch": 10.52, "learning_rate": 1.8204225352112676e-05, "loss": 1.3173, "step": 9380 }, { "epoch": 10.52, "learning_rate": 1.8202816901408453e-05, "loss": 1.148, "step": 9381 }, { "epoch": 10.52, "learning_rate": 1.8201408450704226e-05, "loss": 1.2704, "step": 9382 }, { "epoch": 10.52, "learning_rate": 1.8200000000000002e-05, "loss": 1.3435, "step": 9383 }, { "epoch": 10.52, "learning_rate": 1.8198591549295775e-05, "loss": 1.1587, "step": 9384 }, { "epoch": 10.52, "learning_rate": 1.819718309859155e-05, "loss": 1.1415, "step": 9385 }, { "epoch": 10.52, "learning_rate": 1.8195774647887324e-05, "loss": 1.21, "step": 9386 }, { "epoch": 10.52, "learning_rate": 1.81943661971831e-05, "loss": 1.1241, "step": 9387 }, { "epoch": 10.52, "learning_rate": 1.8192957746478873e-05, "loss": 1.15, "step": 9388 }, { "epoch": 10.53, "learning_rate": 1.8191549295774646e-05, "loss": 1.0963, "step": 9389 }, { "epoch": 10.53, "learning_rate": 1.8190140845070423e-05, "loss": 1.1237, "step": 9390 }, { "epoch": 10.53, "learning_rate": 1.8188732394366196e-05, "loss": 1.2579, "step": 9391 }, { "epoch": 10.53, "learning_rate": 1.8187323943661972e-05, "loss": 0.9817, "step": 9392 }, { "epoch": 10.53, "learning_rate": 1.8185915492957745e-05, "loss": 1.035, "step": 9393 }, { "epoch": 10.53, "learning_rate": 1.8184507042253525e-05, "loss": 1.0435, "step": 9394 }, { "epoch": 10.53, "learning_rate": 1.8183098591549298e-05, "loss": 1.0521, "step": 9395 }, { "epoch": 10.53, "learning_rate": 1.818169014084507e-05, "loss": 1.084, "step": 9396 }, { "epoch": 10.53, "learning_rate": 1.8180281690140847e-05, "loss": 1.0893, "step": 9397 }, { "epoch": 10.54, "learning_rate": 1.817887323943662e-05, "loss": 1.1395, "step": 9398 }, { "epoch": 10.54, "learning_rate": 1.8177464788732396e-05, "loss": 1.13, "step": 9399 }, { "epoch": 10.54, "learning_rate": 1.817605633802817e-05, "loss": 0.9942, "step": 9400 }, { "epoch": 10.54, "learning_rate": 1.8174647887323945e-05, "loss": 1.003, "step": 9401 }, { "epoch": 10.54, "learning_rate": 1.817323943661972e-05, "loss": 1.0096, "step": 9402 }, { "epoch": 10.54, "learning_rate": 1.817183098591549e-05, "loss": 1.1337, "step": 9403 }, { "epoch": 10.54, "learning_rate": 1.8170422535211268e-05, "loss": 1.1485, "step": 9404 }, { "epoch": 10.54, "learning_rate": 1.816901408450704e-05, "loss": 1.0625, "step": 9405 }, { "epoch": 10.54, "learning_rate": 1.8167605633802817e-05, "loss": 1.0299, "step": 9406 }, { "epoch": 10.55, "learning_rate": 1.816619718309859e-05, "loss": 1.0547, "step": 9407 }, { "epoch": 10.55, "learning_rate": 1.8164788732394366e-05, "loss": 0.9583, "step": 9408 }, { "epoch": 10.55, "learning_rate": 1.8163380281690143e-05, "loss": 0.9879, "step": 9409 }, { "epoch": 10.55, "learning_rate": 1.816197183098592e-05, "loss": 1.1046, "step": 9410 }, { "epoch": 10.55, "learning_rate": 1.8160563380281692e-05, "loss": 1.1316, "step": 9411 }, { "epoch": 10.55, "learning_rate": 1.8159154929577465e-05, "loss": 1.0816, "step": 9412 }, { "epoch": 10.55, "learning_rate": 1.815774647887324e-05, "loss": 1.0159, "step": 9413 }, { "epoch": 10.55, "learning_rate": 1.8156338028169014e-05, "loss": 0.9029, "step": 9414 }, { "epoch": 10.55, "learning_rate": 1.815492957746479e-05, "loss": 1.0874, "step": 9415 }, { "epoch": 10.56, "learning_rate": 1.8153521126760563e-05, "loss": 1.0123, "step": 9416 }, { "epoch": 10.56, "learning_rate": 1.815211267605634e-05, "loss": 0.867, "step": 9417 }, { "epoch": 10.56, "learning_rate": 1.8150704225352113e-05, "loss": 1.3146, "step": 9418 }, { "epoch": 10.56, "learning_rate": 1.8149295774647885e-05, "loss": 1.095, "step": 9419 }, { "epoch": 10.56, "learning_rate": 1.8147887323943662e-05, "loss": 0.9343, "step": 9420 }, { "epoch": 10.56, "learning_rate": 1.8146478873239435e-05, "loss": 1.6818, "step": 9421 }, { "epoch": 10.56, "learning_rate": 1.814507042253521e-05, "loss": 1.2234, "step": 9422 }, { "epoch": 10.56, "learning_rate": 1.8143661971830984e-05, "loss": 1.2763, "step": 9423 }, { "epoch": 10.57, "learning_rate": 1.8142253521126764e-05, "loss": 1.2079, "step": 9424 }, { "epoch": 10.57, "learning_rate": 1.8140845070422537e-05, "loss": 1.3389, "step": 9425 }, { "epoch": 10.57, "learning_rate": 1.8139436619718313e-05, "loss": 1.2585, "step": 9426 }, { "epoch": 10.57, "learning_rate": 1.8138028169014086e-05, "loss": 1.2906, "step": 9427 }, { "epoch": 10.57, "learning_rate": 1.813661971830986e-05, "loss": 1.0924, "step": 9428 }, { "epoch": 10.57, "learning_rate": 1.8135211267605635e-05, "loss": 1.2001, "step": 9429 }, { "epoch": 10.57, "learning_rate": 1.8133802816901408e-05, "loss": 1.2588, "step": 9430 }, { "epoch": 10.57, "learning_rate": 1.8132394366197185e-05, "loss": 1.1355, "step": 9431 }, { "epoch": 10.57, "learning_rate": 1.8130985915492957e-05, "loss": 1.2493, "step": 9432 }, { "epoch": 10.58, "learning_rate": 1.8129577464788734e-05, "loss": 1.1629, "step": 9433 }, { "epoch": 10.58, "learning_rate": 1.8128169014084507e-05, "loss": 1.0652, "step": 9434 }, { "epoch": 10.58, "learning_rate": 1.812676056338028e-05, "loss": 1.1694, "step": 9435 }, { "epoch": 10.58, "learning_rate": 1.8125352112676056e-05, "loss": 1.2614, "step": 9436 }, { "epoch": 10.58, "learning_rate": 1.812394366197183e-05, "loss": 1.1846, "step": 9437 }, { "epoch": 10.58, "learning_rate": 1.8122535211267605e-05, "loss": 1.0757, "step": 9438 }, { "epoch": 10.58, "learning_rate": 1.812112676056338e-05, "loss": 1.0544, "step": 9439 }, { "epoch": 10.58, "learning_rate": 1.8119718309859158e-05, "loss": 1.2505, "step": 9440 }, { "epoch": 10.58, "learning_rate": 1.811830985915493e-05, "loss": 1.1284, "step": 9441 }, { "epoch": 10.59, "learning_rate": 1.8116901408450707e-05, "loss": 0.8481, "step": 9442 }, { "epoch": 10.59, "learning_rate": 1.811549295774648e-05, "loss": 1.2257, "step": 9443 }, { "epoch": 10.59, "learning_rate": 1.8114084507042253e-05, "loss": 1.1721, "step": 9444 }, { "epoch": 10.59, "learning_rate": 1.811267605633803e-05, "loss": 1.125, "step": 9445 }, { "epoch": 10.59, "learning_rate": 1.8111267605633802e-05, "loss": 1.1734, "step": 9446 }, { "epoch": 10.59, "learning_rate": 1.810985915492958e-05, "loss": 1.1804, "step": 9447 }, { "epoch": 10.59, "learning_rate": 1.810845070422535e-05, "loss": 1.056, "step": 9448 }, { "epoch": 10.59, "learning_rate": 1.8107042253521128e-05, "loss": 1.1715, "step": 9449 }, { "epoch": 10.59, "learning_rate": 1.81056338028169e-05, "loss": 1.1528, "step": 9450 }, { "epoch": 10.6, "learning_rate": 1.8104225352112674e-05, "loss": 1.2391, "step": 9451 }, { "epoch": 10.6, "learning_rate": 1.810281690140845e-05, "loss": 1.118, "step": 9452 }, { "epoch": 10.6, "learning_rate": 1.8101408450704223e-05, "loss": 0.9798, "step": 9453 }, { "epoch": 10.6, "learning_rate": 1.8100000000000003e-05, "loss": 1.1589, "step": 9454 }, { "epoch": 10.6, "learning_rate": 1.8098591549295776e-05, "loss": 1.1091, "step": 9455 }, { "epoch": 10.6, "learning_rate": 1.8097183098591552e-05, "loss": 1.1576, "step": 9456 }, { "epoch": 10.6, "learning_rate": 1.8095774647887325e-05, "loss": 1.006, "step": 9457 }, { "epoch": 10.6, "learning_rate": 1.80943661971831e-05, "loss": 1.2501, "step": 9458 }, { "epoch": 10.6, "learning_rate": 1.8092957746478874e-05, "loss": 0.9686, "step": 9459 }, { "epoch": 10.61, "learning_rate": 1.8091549295774647e-05, "loss": 1.213, "step": 9460 }, { "epoch": 10.61, "learning_rate": 1.8090140845070424e-05, "loss": 1.0808, "step": 9461 }, { "epoch": 10.61, "learning_rate": 1.8088732394366197e-05, "loss": 1.195, "step": 9462 }, { "epoch": 10.61, "learning_rate": 1.8087323943661973e-05, "loss": 1.025, "step": 9463 }, { "epoch": 10.61, "learning_rate": 1.8085915492957746e-05, "loss": 0.8984, "step": 9464 }, { "epoch": 10.61, "learning_rate": 1.8084507042253522e-05, "loss": 1.0566, "step": 9465 }, { "epoch": 10.61, "learning_rate": 1.8083098591549295e-05, "loss": 1.2366, "step": 9466 }, { "epoch": 10.61, "learning_rate": 1.8081690140845068e-05, "loss": 0.9526, "step": 9467 }, { "epoch": 10.61, "learning_rate": 1.8080281690140844e-05, "loss": 0.8318, "step": 9468 }, { "epoch": 10.62, "learning_rate": 1.8078873239436617e-05, "loss": 1.0407, "step": 9469 }, { "epoch": 10.62, "learning_rate": 1.8077464788732397e-05, "loss": 1.1361, "step": 9470 }, { "epoch": 10.62, "learning_rate": 1.807605633802817e-05, "loss": 1.3741, "step": 9471 }, { "epoch": 10.62, "learning_rate": 1.8074647887323946e-05, "loss": 1.1736, "step": 9472 }, { "epoch": 10.62, "learning_rate": 1.807323943661972e-05, "loss": 1.1974, "step": 9473 }, { "epoch": 10.62, "learning_rate": 1.8071830985915496e-05, "loss": 1.2711, "step": 9474 }, { "epoch": 10.62, "learning_rate": 1.807042253521127e-05, "loss": 1.2956, "step": 9475 }, { "epoch": 10.62, "learning_rate": 1.806901408450704e-05, "loss": 1.1207, "step": 9476 }, { "epoch": 10.62, "learning_rate": 1.8067605633802818e-05, "loss": 1.1986, "step": 9477 }, { "epoch": 10.63, "learning_rate": 1.806619718309859e-05, "loss": 1.0736, "step": 9478 }, { "epoch": 10.63, "learning_rate": 1.8064788732394367e-05, "loss": 1.2977, "step": 9479 }, { "epoch": 10.63, "learning_rate": 1.806338028169014e-05, "loss": 1.1382, "step": 9480 }, { "epoch": 10.63, "learning_rate": 1.8061971830985916e-05, "loss": 1.042, "step": 9481 }, { "epoch": 10.63, "learning_rate": 1.806056338028169e-05, "loss": 1.0413, "step": 9482 }, { "epoch": 10.63, "learning_rate": 1.8059154929577462e-05, "loss": 1.1982, "step": 9483 }, { "epoch": 10.63, "learning_rate": 1.805774647887324e-05, "loss": 1.0517, "step": 9484 }, { "epoch": 10.63, "learning_rate": 1.8056338028169015e-05, "loss": 1.2008, "step": 9485 }, { "epoch": 10.63, "learning_rate": 1.805492957746479e-05, "loss": 1.0764, "step": 9486 }, { "epoch": 10.64, "learning_rate": 1.8053521126760564e-05, "loss": 1.0842, "step": 9487 }, { "epoch": 10.64, "learning_rate": 1.805211267605634e-05, "loss": 1.1165, "step": 9488 }, { "epoch": 10.64, "learning_rate": 1.8050704225352114e-05, "loss": 1.1393, "step": 9489 }, { "epoch": 10.64, "learning_rate": 1.804929577464789e-05, "loss": 1.0008, "step": 9490 }, { "epoch": 10.64, "learning_rate": 1.8047887323943663e-05, "loss": 1.2052, "step": 9491 }, { "epoch": 10.64, "learning_rate": 1.8046478873239436e-05, "loss": 1.2514, "step": 9492 }, { "epoch": 10.64, "learning_rate": 1.8045070422535212e-05, "loss": 1.1431, "step": 9493 }, { "epoch": 10.64, "learning_rate": 1.8043661971830985e-05, "loss": 1.0108, "step": 9494 }, { "epoch": 10.64, "learning_rate": 1.804225352112676e-05, "loss": 1.1681, "step": 9495 }, { "epoch": 10.65, "learning_rate": 1.8040845070422534e-05, "loss": 1.0917, "step": 9496 }, { "epoch": 10.65, "learning_rate": 1.803943661971831e-05, "loss": 0.9932, "step": 9497 }, { "epoch": 10.65, "learning_rate": 1.8038028169014084e-05, "loss": 1.2585, "step": 9498 }, { "epoch": 10.65, "learning_rate": 1.8036619718309856e-05, "loss": 0.9782, "step": 9499 }, { "epoch": 10.65, "learning_rate": 1.8035211267605636e-05, "loss": 0.8307, "step": 9500 }, { "epoch": 10.65, "learning_rate": 1.803380281690141e-05, "loss": 1.1859, "step": 9501 }, { "epoch": 10.65, "learning_rate": 1.8032394366197185e-05, "loss": 1.0159, "step": 9502 }, { "epoch": 10.65, "learning_rate": 1.803098591549296e-05, "loss": 1.139, "step": 9503 }, { "epoch": 10.65, "learning_rate": 1.8029577464788735e-05, "loss": 1.1373, "step": 9504 }, { "epoch": 10.66, "learning_rate": 1.8028169014084508e-05, "loss": 1.3211, "step": 9505 }, { "epoch": 10.66, "learning_rate": 1.802676056338028e-05, "loss": 1.3105, "step": 9506 }, { "epoch": 10.66, "learning_rate": 1.8025352112676057e-05, "loss": 1.1837, "step": 9507 }, { "epoch": 10.66, "learning_rate": 1.802394366197183e-05, "loss": 1.1804, "step": 9508 }, { "epoch": 10.66, "learning_rate": 1.8022535211267606e-05, "loss": 0.966, "step": 9509 }, { "epoch": 10.66, "learning_rate": 1.802112676056338e-05, "loss": 0.9361, "step": 9510 }, { "epoch": 10.66, "learning_rate": 1.8019718309859156e-05, "loss": 0.8501, "step": 9511 }, { "epoch": 10.66, "learning_rate": 1.801830985915493e-05, "loss": 0.9082, "step": 9512 }, { "epoch": 10.66, "learning_rate": 1.8016901408450705e-05, "loss": 0.8993, "step": 9513 }, { "epoch": 10.67, "learning_rate": 1.8015492957746478e-05, "loss": 0.9357, "step": 9514 }, { "epoch": 10.67, "learning_rate": 1.8014084507042254e-05, "loss": 1.0456, "step": 9515 }, { "epoch": 10.67, "learning_rate": 1.801267605633803e-05, "loss": 0.9528, "step": 9516 }, { "epoch": 10.67, "learning_rate": 1.8011267605633803e-05, "loss": 1.1603, "step": 9517 }, { "epoch": 10.67, "learning_rate": 1.800985915492958e-05, "loss": 1.0319, "step": 9518 }, { "epoch": 10.67, "learning_rate": 1.8008450704225353e-05, "loss": 1.0613, "step": 9519 }, { "epoch": 10.67, "learning_rate": 1.800704225352113e-05, "loss": 1.0634, "step": 9520 }, { "epoch": 10.67, "learning_rate": 1.8005633802816902e-05, "loss": 1.6543, "step": 9521 }, { "epoch": 10.67, "learning_rate": 1.8004225352112675e-05, "loss": 1.4528, "step": 9522 }, { "epoch": 10.68, "learning_rate": 1.800281690140845e-05, "loss": 1.194, "step": 9523 }, { "epoch": 10.68, "learning_rate": 1.8001408450704224e-05, "loss": 1.1753, "step": 9524 }, { "epoch": 10.68, "learning_rate": 1.8e-05, "loss": 1.24, "step": 9525 }, { "epoch": 10.68, "learning_rate": 1.7998591549295773e-05, "loss": 1.4308, "step": 9526 }, { "epoch": 10.68, "learning_rate": 1.799718309859155e-05, "loss": 1.3266, "step": 9527 }, { "epoch": 10.68, "learning_rate": 1.7995774647887323e-05, "loss": 1.2538, "step": 9528 }, { "epoch": 10.68, "learning_rate": 1.79943661971831e-05, "loss": 1.1835, "step": 9529 }, { "epoch": 10.68, "learning_rate": 1.7992957746478875e-05, "loss": 1.1482, "step": 9530 }, { "epoch": 10.68, "learning_rate": 1.7991549295774648e-05, "loss": 1.1513, "step": 9531 }, { "epoch": 10.69, "learning_rate": 1.7990140845070425e-05, "loss": 1.1431, "step": 9532 }, { "epoch": 10.69, "learning_rate": 1.7988732394366198e-05, "loss": 1.1833, "step": 9533 }, { "epoch": 10.69, "learning_rate": 1.7987323943661974e-05, "loss": 1.1453, "step": 9534 }, { "epoch": 10.69, "learning_rate": 1.7985915492957747e-05, "loss": 1.152, "step": 9535 }, { "epoch": 10.69, "learning_rate": 1.7984507042253523e-05, "loss": 1.075, "step": 9536 }, { "epoch": 10.69, "learning_rate": 1.7983098591549296e-05, "loss": 1.0815, "step": 9537 }, { "epoch": 10.69, "learning_rate": 1.798169014084507e-05, "loss": 1.075, "step": 9538 }, { "epoch": 10.69, "learning_rate": 1.7980281690140845e-05, "loss": 1.0898, "step": 9539 }, { "epoch": 10.7, "learning_rate": 1.797887323943662e-05, "loss": 1.1998, "step": 9540 }, { "epoch": 10.7, "learning_rate": 1.7977464788732395e-05, "loss": 1.183, "step": 9541 }, { "epoch": 10.7, "learning_rate": 1.7976056338028168e-05, "loss": 1.1924, "step": 9542 }, { "epoch": 10.7, "learning_rate": 1.7974647887323944e-05, "loss": 0.9858, "step": 9543 }, { "epoch": 10.7, "learning_rate": 1.7973239436619717e-05, "loss": 0.9277, "step": 9544 }, { "epoch": 10.7, "learning_rate": 1.7971830985915497e-05, "loss": 0.9745, "step": 9545 }, { "epoch": 10.7, "learning_rate": 1.797042253521127e-05, "loss": 1.2327, "step": 9546 }, { "epoch": 10.7, "learning_rate": 1.7969014084507042e-05, "loss": 0.9233, "step": 9547 }, { "epoch": 10.7, "learning_rate": 1.796760563380282e-05, "loss": 1.0366, "step": 9548 }, { "epoch": 10.71, "learning_rate": 1.7966197183098592e-05, "loss": 1.165, "step": 9549 }, { "epoch": 10.71, "learning_rate": 1.7964788732394368e-05, "loss": 1.168, "step": 9550 }, { "epoch": 10.71, "learning_rate": 1.796338028169014e-05, "loss": 1.0995, "step": 9551 }, { "epoch": 10.71, "learning_rate": 1.7961971830985917e-05, "loss": 1.1876, "step": 9552 }, { "epoch": 10.71, "learning_rate": 1.796056338028169e-05, "loss": 0.9876, "step": 9553 }, { "epoch": 10.71, "learning_rate": 1.7959154929577463e-05, "loss": 0.958, "step": 9554 }, { "epoch": 10.71, "learning_rate": 1.795774647887324e-05, "loss": 0.958, "step": 9555 }, { "epoch": 10.71, "learning_rate": 1.7956338028169013e-05, "loss": 0.9094, "step": 9556 }, { "epoch": 10.71, "learning_rate": 1.795492957746479e-05, "loss": 1.1437, "step": 9557 }, { "epoch": 10.72, "learning_rate": 1.7953521126760562e-05, "loss": 1.1658, "step": 9558 }, { "epoch": 10.72, "learning_rate": 1.7952112676056338e-05, "loss": 1.0978, "step": 9559 }, { "epoch": 10.72, "learning_rate": 1.7950704225352114e-05, "loss": 0.9313, "step": 9560 }, { "epoch": 10.72, "learning_rate": 1.794929577464789e-05, "loss": 1.1124, "step": 9561 }, { "epoch": 10.72, "learning_rate": 1.7947887323943664e-05, "loss": 1.1379, "step": 9562 }, { "epoch": 10.72, "learning_rate": 1.7946478873239437e-05, "loss": 1.0502, "step": 9563 }, { "epoch": 10.72, "learning_rate": 1.7945070422535213e-05, "loss": 1.0746, "step": 9564 }, { "epoch": 10.72, "learning_rate": 1.7943661971830986e-05, "loss": 0.9406, "step": 9565 }, { "epoch": 10.72, "learning_rate": 1.7942253521126762e-05, "loss": 0.9156, "step": 9566 }, { "epoch": 10.73, "learning_rate": 1.7940845070422535e-05, "loss": 1.201, "step": 9567 }, { "epoch": 10.73, "learning_rate": 1.793943661971831e-05, "loss": 0.9792, "step": 9568 }, { "epoch": 10.73, "learning_rate": 1.7938028169014085e-05, "loss": 1.1128, "step": 9569 }, { "epoch": 10.73, "learning_rate": 1.7936619718309857e-05, "loss": 0.8713, "step": 9570 }, { "epoch": 10.73, "learning_rate": 1.7935211267605634e-05, "loss": 1.4481, "step": 9571 }, { "epoch": 10.73, "learning_rate": 1.7933802816901407e-05, "loss": 1.1456, "step": 9572 }, { "epoch": 10.73, "learning_rate": 1.7932394366197183e-05, "loss": 1.3638, "step": 9573 }, { "epoch": 10.73, "learning_rate": 1.7930985915492956e-05, "loss": 1.2444, "step": 9574 }, { "epoch": 10.73, "learning_rate": 1.7929577464788736e-05, "loss": 1.2028, "step": 9575 }, { "epoch": 10.74, "learning_rate": 1.792816901408451e-05, "loss": 0.9204, "step": 9576 }, { "epoch": 10.74, "learning_rate": 1.7926760563380285e-05, "loss": 1.2346, "step": 9577 }, { "epoch": 10.74, "learning_rate": 1.7925352112676058e-05, "loss": 1.0206, "step": 9578 }, { "epoch": 10.74, "learning_rate": 1.792394366197183e-05, "loss": 0.9647, "step": 9579 }, { "epoch": 10.74, "learning_rate": 1.7922535211267607e-05, "loss": 1.0721, "step": 9580 }, { "epoch": 10.74, "learning_rate": 1.792112676056338e-05, "loss": 1.081, "step": 9581 }, { "epoch": 10.74, "learning_rate": 1.7919718309859156e-05, "loss": 1.0815, "step": 9582 }, { "epoch": 10.74, "learning_rate": 1.791830985915493e-05, "loss": 1.0675, "step": 9583 }, { "epoch": 10.74, "learning_rate": 1.7916901408450706e-05, "loss": 0.9417, "step": 9584 }, { "epoch": 10.75, "learning_rate": 1.791549295774648e-05, "loss": 1.0395, "step": 9585 }, { "epoch": 10.75, "learning_rate": 1.791408450704225e-05, "loss": 1.2244, "step": 9586 }, { "epoch": 10.75, "learning_rate": 1.7912676056338028e-05, "loss": 1.047, "step": 9587 }, { "epoch": 10.75, "learning_rate": 1.79112676056338e-05, "loss": 1.0469, "step": 9588 }, { "epoch": 10.75, "learning_rate": 1.7909859154929577e-05, "loss": 1.1863, "step": 9589 }, { "epoch": 10.75, "learning_rate": 1.790845070422535e-05, "loss": 1.3438, "step": 9590 }, { "epoch": 10.75, "learning_rate": 1.790704225352113e-05, "loss": 1.0065, "step": 9591 }, { "epoch": 10.75, "learning_rate": 1.7905633802816903e-05, "loss": 1.1119, "step": 9592 }, { "epoch": 10.75, "learning_rate": 1.790422535211268e-05, "loss": 1.1528, "step": 9593 }, { "epoch": 10.76, "learning_rate": 1.7902816901408452e-05, "loss": 1.0944, "step": 9594 }, { "epoch": 10.76, "learning_rate": 1.7901408450704225e-05, "loss": 1.2377, "step": 9595 }, { "epoch": 10.76, "learning_rate": 1.79e-05, "loss": 0.9273, "step": 9596 }, { "epoch": 10.76, "learning_rate": 1.7898591549295774e-05, "loss": 1.0653, "step": 9597 }, { "epoch": 10.76, "learning_rate": 1.789718309859155e-05, "loss": 1.0963, "step": 9598 }, { "epoch": 10.76, "learning_rate": 1.7895774647887324e-05, "loss": 1.0484, "step": 9599 }, { "epoch": 10.76, "learning_rate": 1.78943661971831e-05, "loss": 1.0155, "step": 9600 }, { "epoch": 10.76, "learning_rate": 1.7892957746478873e-05, "loss": 1.1493, "step": 9601 }, { "epoch": 10.76, "learning_rate": 1.7891549295774646e-05, "loss": 1.1033, "step": 9602 }, { "epoch": 10.77, "learning_rate": 1.7890140845070422e-05, "loss": 0.9666, "step": 9603 }, { "epoch": 10.77, "learning_rate": 1.7888732394366195e-05, "loss": 1.1401, "step": 9604 }, { "epoch": 10.77, "learning_rate": 1.788732394366197e-05, "loss": 0.9742, "step": 9605 }, { "epoch": 10.77, "learning_rate": 1.7885915492957748e-05, "loss": 1.0602, "step": 9606 }, { "epoch": 10.77, "learning_rate": 1.7884507042253524e-05, "loss": 1.0231, "step": 9607 }, { "epoch": 10.77, "learning_rate": 1.7883098591549297e-05, "loss": 1.0765, "step": 9608 }, { "epoch": 10.77, "learning_rate": 1.788169014084507e-05, "loss": 1.0221, "step": 9609 }, { "epoch": 10.77, "learning_rate": 1.7880281690140846e-05, "loss": 1.0663, "step": 9610 }, { "epoch": 10.77, "learning_rate": 1.787887323943662e-05, "loss": 1.0571, "step": 9611 }, { "epoch": 10.78, "learning_rate": 1.7877464788732396e-05, "loss": 1.2658, "step": 9612 }, { "epoch": 10.78, "learning_rate": 1.787605633802817e-05, "loss": 0.8559, "step": 9613 }, { "epoch": 10.78, "learning_rate": 1.7874647887323945e-05, "loss": 0.9707, "step": 9614 }, { "epoch": 10.78, "learning_rate": 1.7873239436619718e-05, "loss": 1.1317, "step": 9615 }, { "epoch": 10.78, "learning_rate": 1.7871830985915494e-05, "loss": 0.7616, "step": 9616 }, { "epoch": 10.78, "learning_rate": 1.7870422535211267e-05, "loss": 0.8461, "step": 9617 }, { "epoch": 10.78, "learning_rate": 1.786901408450704e-05, "loss": 1.0662, "step": 9618 }, { "epoch": 10.78, "learning_rate": 1.7867605633802816e-05, "loss": 0.9733, "step": 9619 }, { "epoch": 10.78, "learning_rate": 1.786619718309859e-05, "loss": 1.1307, "step": 9620 }, { "epoch": 10.79, "learning_rate": 1.786478873239437e-05, "loss": 1.566, "step": 9621 }, { "epoch": 10.79, "learning_rate": 1.7863380281690142e-05, "loss": 1.0572, "step": 9622 }, { "epoch": 10.79, "learning_rate": 1.786197183098592e-05, "loss": 1.2812, "step": 9623 }, { "epoch": 10.79, "learning_rate": 1.786056338028169e-05, "loss": 1.177, "step": 9624 }, { "epoch": 10.79, "learning_rate": 1.7859154929577464e-05, "loss": 1.0685, "step": 9625 }, { "epoch": 10.79, "learning_rate": 1.785774647887324e-05, "loss": 1.1545, "step": 9626 }, { "epoch": 10.79, "learning_rate": 1.7856338028169013e-05, "loss": 1.2572, "step": 9627 }, { "epoch": 10.79, "learning_rate": 1.785492957746479e-05, "loss": 1.1721, "step": 9628 }, { "epoch": 10.79, "learning_rate": 1.7853521126760563e-05, "loss": 1.048, "step": 9629 }, { "epoch": 10.8, "learning_rate": 1.785211267605634e-05, "loss": 1.1271, "step": 9630 }, { "epoch": 10.8, "learning_rate": 1.7850704225352112e-05, "loss": 1.1786, "step": 9631 }, { "epoch": 10.8, "learning_rate": 1.784929577464789e-05, "loss": 1.2831, "step": 9632 }, { "epoch": 10.8, "learning_rate": 1.784788732394366e-05, "loss": 1.1231, "step": 9633 }, { "epoch": 10.8, "learning_rate": 1.7846478873239434e-05, "loss": 1.0267, "step": 9634 }, { "epoch": 10.8, "learning_rate": 1.784507042253521e-05, "loss": 1.0065, "step": 9635 }, { "epoch": 10.8, "learning_rate": 1.7843661971830987e-05, "loss": 1.1454, "step": 9636 }, { "epoch": 10.8, "learning_rate": 1.7842253521126763e-05, "loss": 1.0343, "step": 9637 }, { "epoch": 10.8, "learning_rate": 1.7840845070422536e-05, "loss": 1.0858, "step": 9638 }, { "epoch": 10.81, "learning_rate": 1.7839436619718313e-05, "loss": 1.1527, "step": 9639 }, { "epoch": 10.81, "learning_rate": 1.7838028169014085e-05, "loss": 1.199, "step": 9640 }, { "epoch": 10.81, "learning_rate": 1.783661971830986e-05, "loss": 1.1068, "step": 9641 }, { "epoch": 10.81, "learning_rate": 1.7835211267605635e-05, "loss": 1.2067, "step": 9642 }, { "epoch": 10.81, "learning_rate": 1.7833802816901408e-05, "loss": 1.0325, "step": 9643 }, { "epoch": 10.81, "learning_rate": 1.7832394366197184e-05, "loss": 1.0038, "step": 9644 }, { "epoch": 10.81, "learning_rate": 1.7830985915492957e-05, "loss": 1.0075, "step": 9645 }, { "epoch": 10.81, "learning_rate": 1.7829577464788733e-05, "loss": 1.1797, "step": 9646 }, { "epoch": 10.82, "learning_rate": 1.7828169014084506e-05, "loss": 1.0569, "step": 9647 }, { "epoch": 10.82, "learning_rate": 1.7826760563380283e-05, "loss": 0.9814, "step": 9648 }, { "epoch": 10.82, "learning_rate": 1.7825352112676056e-05, "loss": 1.1408, "step": 9649 }, { "epoch": 10.82, "learning_rate": 1.782394366197183e-05, "loss": 1.1112, "step": 9650 }, { "epoch": 10.82, "learning_rate": 1.7822535211267608e-05, "loss": 0.993, "step": 9651 }, { "epoch": 10.82, "learning_rate": 1.782112676056338e-05, "loss": 1.2621, "step": 9652 }, { "epoch": 10.82, "learning_rate": 1.7819718309859157e-05, "loss": 1.0579, "step": 9653 }, { "epoch": 10.82, "learning_rate": 1.781830985915493e-05, "loss": 1.139, "step": 9654 }, { "epoch": 10.82, "learning_rate": 1.7816901408450707e-05, "loss": 1.0432, "step": 9655 }, { "epoch": 10.83, "learning_rate": 1.781549295774648e-05, "loss": 0.9978, "step": 9656 }, { "epoch": 10.83, "learning_rate": 1.7814084507042253e-05, "loss": 0.9767, "step": 9657 }, { "epoch": 10.83, "learning_rate": 1.781267605633803e-05, "loss": 0.9029, "step": 9658 }, { "epoch": 10.83, "learning_rate": 1.7811267605633802e-05, "loss": 0.9634, "step": 9659 }, { "epoch": 10.83, "learning_rate": 1.7809859154929578e-05, "loss": 1.1773, "step": 9660 }, { "epoch": 10.83, "learning_rate": 1.780845070422535e-05, "loss": 0.9522, "step": 9661 }, { "epoch": 10.83, "learning_rate": 1.7807042253521127e-05, "loss": 1.1654, "step": 9662 }, { "epoch": 10.83, "learning_rate": 1.78056338028169e-05, "loss": 1.1383, "step": 9663 }, { "epoch": 10.83, "learning_rate": 1.7804225352112677e-05, "loss": 1.0231, "step": 9664 }, { "epoch": 10.84, "learning_rate": 1.780281690140845e-05, "loss": 1.1691, "step": 9665 }, { "epoch": 10.84, "learning_rate": 1.7801408450704226e-05, "loss": 1.0711, "step": 9666 }, { "epoch": 10.84, "learning_rate": 1.7800000000000002e-05, "loss": 0.88, "step": 9667 }, { "epoch": 10.84, "learning_rate": 1.7798591549295775e-05, "loss": 1.0097, "step": 9668 }, { "epoch": 10.84, "learning_rate": 1.779718309859155e-05, "loss": 0.9207, "step": 9669 }, { "epoch": 10.84, "learning_rate": 1.7795774647887325e-05, "loss": 0.8429, "step": 9670 }, { "epoch": 10.84, "learning_rate": 1.77943661971831e-05, "loss": 1.3201, "step": 9671 }, { "epoch": 10.84, "learning_rate": 1.7792957746478874e-05, "loss": 1.3967, "step": 9672 }, { "epoch": 10.84, "learning_rate": 1.7791549295774647e-05, "loss": 1.0992, "step": 9673 }, { "epoch": 10.85, "learning_rate": 1.7790140845070423e-05, "loss": 1.1087, "step": 9674 }, { "epoch": 10.85, "learning_rate": 1.7788732394366196e-05, "loss": 0.8704, "step": 9675 }, { "epoch": 10.85, "learning_rate": 1.7787323943661972e-05, "loss": 1.1843, "step": 9676 }, { "epoch": 10.85, "learning_rate": 1.7785915492957745e-05, "loss": 1.1316, "step": 9677 }, { "epoch": 10.85, "learning_rate": 1.778450704225352e-05, "loss": 1.1216, "step": 9678 }, { "epoch": 10.85, "learning_rate": 1.7783098591549295e-05, "loss": 1.1098, "step": 9679 }, { "epoch": 10.85, "learning_rate": 1.778169014084507e-05, "loss": 1.2637, "step": 9680 }, { "epoch": 10.85, "learning_rate": 1.7780281690140847e-05, "loss": 1.0516, "step": 9681 }, { "epoch": 10.85, "learning_rate": 1.777887323943662e-05, "loss": 1.026, "step": 9682 }, { "epoch": 10.86, "learning_rate": 1.7777464788732397e-05, "loss": 1.2521, "step": 9683 }, { "epoch": 10.86, "learning_rate": 1.777605633802817e-05, "loss": 1.1301, "step": 9684 }, { "epoch": 10.86, "learning_rate": 1.7774647887323946e-05, "loss": 1.0484, "step": 9685 }, { "epoch": 10.86, "learning_rate": 1.777323943661972e-05, "loss": 1.1378, "step": 9686 }, { "epoch": 10.86, "learning_rate": 1.7771830985915495e-05, "loss": 1.0096, "step": 9687 }, { "epoch": 10.86, "learning_rate": 1.7770422535211268e-05, "loss": 1.0865, "step": 9688 }, { "epoch": 10.86, "learning_rate": 1.776901408450704e-05, "loss": 1.0898, "step": 9689 }, { "epoch": 10.86, "learning_rate": 1.7767605633802817e-05, "loss": 0.95, "step": 9690 }, { "epoch": 10.86, "learning_rate": 1.776619718309859e-05, "loss": 1.163, "step": 9691 }, { "epoch": 10.87, "learning_rate": 1.7764788732394367e-05, "loss": 1.0508, "step": 9692 }, { "epoch": 10.87, "learning_rate": 1.776338028169014e-05, "loss": 1.1066, "step": 9693 }, { "epoch": 10.87, "learning_rate": 1.7761971830985916e-05, "loss": 0.7823, "step": 9694 }, { "epoch": 10.87, "learning_rate": 1.776056338028169e-05, "loss": 0.9448, "step": 9695 }, { "epoch": 10.87, "learning_rate": 1.775915492957747e-05, "loss": 1.0152, "step": 9696 }, { "epoch": 10.87, "learning_rate": 1.775774647887324e-05, "loss": 1.1117, "step": 9697 }, { "epoch": 10.87, "learning_rate": 1.7756338028169014e-05, "loss": 1.1246, "step": 9698 }, { "epoch": 10.87, "learning_rate": 1.775492957746479e-05, "loss": 1.0976, "step": 9699 }, { "epoch": 10.87, "learning_rate": 1.7753521126760564e-05, "loss": 0.9821, "step": 9700 }, { "epoch": 10.88, "learning_rate": 1.775211267605634e-05, "loss": 1.0745, "step": 9701 }, { "epoch": 10.88, "learning_rate": 1.7750704225352113e-05, "loss": 0.9925, "step": 9702 }, { "epoch": 10.88, "learning_rate": 1.774929577464789e-05, "loss": 0.9195, "step": 9703 }, { "epoch": 10.88, "learning_rate": 1.7747887323943662e-05, "loss": 0.867, "step": 9704 }, { "epoch": 10.88, "learning_rate": 1.7746478873239435e-05, "loss": 0.9543, "step": 9705 }, { "epoch": 10.88, "learning_rate": 1.774507042253521e-05, "loss": 1.0148, "step": 9706 }, { "epoch": 10.88, "learning_rate": 1.7743661971830984e-05, "loss": 0.9214, "step": 9707 }, { "epoch": 10.88, "learning_rate": 1.774225352112676e-05, "loss": 1.0837, "step": 9708 }, { "epoch": 10.88, "learning_rate": 1.7740845070422534e-05, "loss": 0.9741, "step": 9709 }, { "epoch": 10.89, "learning_rate": 1.773943661971831e-05, "loss": 1.1158, "step": 9710 }, { "epoch": 10.89, "learning_rate": 1.7738028169014083e-05, "loss": 1.1625, "step": 9711 }, { "epoch": 10.89, "learning_rate": 1.7736619718309863e-05, "loss": 1.1687, "step": 9712 }, { "epoch": 10.89, "learning_rate": 1.7735211267605636e-05, "loss": 1.1779, "step": 9713 }, { "epoch": 10.89, "learning_rate": 1.773380281690141e-05, "loss": 0.9321, "step": 9714 }, { "epoch": 10.89, "learning_rate": 1.7732394366197185e-05, "loss": 1.0905, "step": 9715 }, { "epoch": 10.89, "learning_rate": 1.7730985915492958e-05, "loss": 1.0371, "step": 9716 }, { "epoch": 10.89, "learning_rate": 1.7729577464788734e-05, "loss": 0.9618, "step": 9717 }, { "epoch": 10.89, "learning_rate": 1.7728169014084507e-05, "loss": 1.0588, "step": 9718 }, { "epoch": 10.9, "learning_rate": 1.7726760563380284e-05, "loss": 0.9557, "step": 9719 }, { "epoch": 10.9, "learning_rate": 1.7725352112676056e-05, "loss": 1.128, "step": 9720 }, { "epoch": 10.9, "learning_rate": 1.772394366197183e-05, "loss": 1.4084, "step": 9721 }, { "epoch": 10.9, "learning_rate": 1.7722535211267606e-05, "loss": 1.2667, "step": 9722 }, { "epoch": 10.9, "learning_rate": 1.772112676056338e-05, "loss": 1.2601, "step": 9723 }, { "epoch": 10.9, "learning_rate": 1.7719718309859155e-05, "loss": 1.187, "step": 9724 }, { "epoch": 10.9, "learning_rate": 1.7718309859154928e-05, "loss": 1.2611, "step": 9725 }, { "epoch": 10.9, "learning_rate": 1.7716901408450704e-05, "loss": 1.1068, "step": 9726 }, { "epoch": 10.9, "learning_rate": 1.771549295774648e-05, "loss": 0.9461, "step": 9727 }, { "epoch": 10.91, "learning_rate": 1.7714084507042254e-05, "loss": 1.2725, "step": 9728 }, { "epoch": 10.91, "learning_rate": 1.771267605633803e-05, "loss": 1.1926, "step": 9729 }, { "epoch": 10.91, "learning_rate": 1.7711267605633803e-05, "loss": 1.2313, "step": 9730 }, { "epoch": 10.91, "learning_rate": 1.770985915492958e-05, "loss": 0.9268, "step": 9731 }, { "epoch": 10.91, "learning_rate": 1.7708450704225352e-05, "loss": 1.085, "step": 9732 }, { "epoch": 10.91, "learning_rate": 1.770704225352113e-05, "loss": 1.103, "step": 9733 }, { "epoch": 10.91, "learning_rate": 1.77056338028169e-05, "loss": 0.972, "step": 9734 }, { "epoch": 10.91, "learning_rate": 1.7704225352112678e-05, "loss": 1.0053, "step": 9735 }, { "epoch": 10.91, "learning_rate": 1.770281690140845e-05, "loss": 1.0609, "step": 9736 }, { "epoch": 10.92, "learning_rate": 1.7701408450704224e-05, "loss": 0.9997, "step": 9737 }, { "epoch": 10.92, "learning_rate": 1.77e-05, "loss": 1.0164, "step": 9738 }, { "epoch": 10.92, "learning_rate": 1.7698591549295773e-05, "loss": 1.1097, "step": 9739 }, { "epoch": 10.92, "learning_rate": 1.769718309859155e-05, "loss": 0.9945, "step": 9740 }, { "epoch": 10.92, "learning_rate": 1.7695774647887322e-05, "loss": 0.9873, "step": 9741 }, { "epoch": 10.92, "learning_rate": 1.7694366197183102e-05, "loss": 1.1824, "step": 9742 }, { "epoch": 10.92, "learning_rate": 1.7692957746478875e-05, "loss": 1.0497, "step": 9743 }, { "epoch": 10.92, "learning_rate": 1.7691549295774648e-05, "loss": 1.0661, "step": 9744 }, { "epoch": 10.92, "learning_rate": 1.7690140845070424e-05, "loss": 1.0881, "step": 9745 }, { "epoch": 10.93, "learning_rate": 1.7688732394366197e-05, "loss": 0.9709, "step": 9746 }, { "epoch": 10.93, "learning_rate": 1.7687323943661973e-05, "loss": 1.1182, "step": 9747 }, { "epoch": 10.93, "learning_rate": 1.7685915492957746e-05, "loss": 1.1795, "step": 9748 }, { "epoch": 10.93, "learning_rate": 1.7684507042253523e-05, "loss": 1.0652, "step": 9749 }, { "epoch": 10.93, "learning_rate": 1.7683098591549296e-05, "loss": 1.146, "step": 9750 }, { "epoch": 10.93, "learning_rate": 1.7681690140845072e-05, "loss": 1.1728, "step": 9751 }, { "epoch": 10.93, "learning_rate": 1.7680281690140845e-05, "loss": 0.8955, "step": 9752 }, { "epoch": 10.93, "learning_rate": 1.7678873239436618e-05, "loss": 1.1076, "step": 9753 }, { "epoch": 10.93, "learning_rate": 1.7677464788732394e-05, "loss": 1.0978, "step": 9754 }, { "epoch": 10.94, "learning_rate": 1.7676056338028167e-05, "loss": 1.1251, "step": 9755 }, { "epoch": 10.94, "learning_rate": 1.7674647887323943e-05, "loss": 1.0765, "step": 9756 }, { "epoch": 10.94, "learning_rate": 1.767323943661972e-05, "loss": 1.0905, "step": 9757 }, { "epoch": 10.94, "learning_rate": 1.7671830985915496e-05, "loss": 0.8472, "step": 9758 }, { "epoch": 10.94, "learning_rate": 1.767042253521127e-05, "loss": 0.9457, "step": 9759 }, { "epoch": 10.94, "learning_rate": 1.7669014084507042e-05, "loss": 1.1698, "step": 9760 }, { "epoch": 10.94, "learning_rate": 1.7667605633802818e-05, "loss": 1.0419, "step": 9761 }, { "epoch": 10.94, "learning_rate": 1.766619718309859e-05, "loss": 1.1968, "step": 9762 }, { "epoch": 10.95, "learning_rate": 1.7664788732394368e-05, "loss": 1.1198, "step": 9763 }, { "epoch": 10.95, "learning_rate": 1.766338028169014e-05, "loss": 1.0163, "step": 9764 }, { "epoch": 10.95, "learning_rate": 1.7661971830985917e-05, "loss": 1.0283, "step": 9765 }, { "epoch": 10.95, "learning_rate": 1.766056338028169e-05, "loss": 1.1217, "step": 9766 }, { "epoch": 10.95, "learning_rate": 1.7659154929577466e-05, "loss": 1.0149, "step": 9767 }, { "epoch": 10.95, "learning_rate": 1.765774647887324e-05, "loss": 1.212, "step": 9768 }, { "epoch": 10.95, "learning_rate": 1.7656338028169012e-05, "loss": 0.8515, "step": 9769 }, { "epoch": 10.95, "learning_rate": 1.765492957746479e-05, "loss": 0.9437, "step": 9770 }, { "epoch": 10.95, "learning_rate": 1.765352112676056e-05, "loss": 1.5196, "step": 9771 }, { "epoch": 10.96, "learning_rate": 1.765211267605634e-05, "loss": 1.1647, "step": 9772 }, { "epoch": 10.96, "learning_rate": 1.7650704225352114e-05, "loss": 1.2257, "step": 9773 }, { "epoch": 10.96, "learning_rate": 1.764929577464789e-05, "loss": 1.2401, "step": 9774 }, { "epoch": 10.96, "learning_rate": 1.7647887323943663e-05, "loss": 1.187, "step": 9775 }, { "epoch": 10.96, "learning_rate": 1.7646478873239436e-05, "loss": 1.1547, "step": 9776 }, { "epoch": 10.96, "learning_rate": 1.7645070422535212e-05, "loss": 1.0016, "step": 9777 }, { "epoch": 10.96, "learning_rate": 1.7643661971830985e-05, "loss": 1.2655, "step": 9778 }, { "epoch": 10.96, "learning_rate": 1.7642253521126762e-05, "loss": 1.0739, "step": 9779 }, { "epoch": 10.96, "learning_rate": 1.7640845070422535e-05, "loss": 1.1139, "step": 9780 }, { "epoch": 10.97, "learning_rate": 1.763943661971831e-05, "loss": 0.9604, "step": 9781 }, { "epoch": 10.97, "learning_rate": 1.7638028169014084e-05, "loss": 0.9977, "step": 9782 }, { "epoch": 10.97, "learning_rate": 1.763661971830986e-05, "loss": 1.0771, "step": 9783 }, { "epoch": 10.97, "learning_rate": 1.7635211267605633e-05, "loss": 1.3479, "step": 9784 }, { "epoch": 10.97, "learning_rate": 1.7633802816901406e-05, "loss": 1.1276, "step": 9785 }, { "epoch": 10.97, "learning_rate": 1.7632394366197183e-05, "loss": 1.1447, "step": 9786 }, { "epoch": 10.97, "learning_rate": 1.763098591549296e-05, "loss": 1.0391, "step": 9787 }, { "epoch": 10.97, "learning_rate": 1.7629577464788735e-05, "loss": 1.0753, "step": 9788 }, { "epoch": 10.97, "learning_rate": 1.7628169014084508e-05, "loss": 1.0962, "step": 9789 }, { "epoch": 10.98, "learning_rate": 1.7626760563380284e-05, "loss": 1.0362, "step": 9790 }, { "epoch": 10.98, "learning_rate": 1.7625352112676057e-05, "loss": 0.9104, "step": 9791 }, { "epoch": 10.98, "learning_rate": 1.762394366197183e-05, "loss": 1.2363, "step": 9792 }, { "epoch": 10.98, "learning_rate": 1.7622535211267607e-05, "loss": 1.0456, "step": 9793 }, { "epoch": 10.98, "learning_rate": 1.762112676056338e-05, "loss": 1.0732, "step": 9794 }, { "epoch": 10.98, "learning_rate": 1.7619718309859156e-05, "loss": 1.1268, "step": 9795 }, { "epoch": 10.98, "learning_rate": 1.761830985915493e-05, "loss": 0.8368, "step": 9796 }, { "epoch": 10.98, "learning_rate": 1.7616901408450705e-05, "loss": 0.8931, "step": 9797 }, { "epoch": 10.98, "learning_rate": 1.7615492957746478e-05, "loss": 0.9713, "step": 9798 }, { "epoch": 10.99, "learning_rate": 1.7614084507042255e-05, "loss": 0.9797, "step": 9799 }, { "epoch": 10.99, "learning_rate": 1.7612676056338027e-05, "loss": 1.0382, "step": 9800 }, { "epoch": 10.99, "learning_rate": 1.76112676056338e-05, "loss": 1.0438, "step": 9801 }, { "epoch": 10.99, "learning_rate": 1.760985915492958e-05, "loss": 0.8533, "step": 9802 }, { "epoch": 10.99, "learning_rate": 1.7608450704225353e-05, "loss": 1.0617, "step": 9803 }, { "epoch": 10.99, "learning_rate": 1.760704225352113e-05, "loss": 1.0828, "step": 9804 }, { "epoch": 10.99, "learning_rate": 1.7605633802816902e-05, "loss": 1.0321, "step": 9805 }, { "epoch": 10.99, "learning_rate": 1.760422535211268e-05, "loss": 0.9826, "step": 9806 }, { "epoch": 10.99, "learning_rate": 1.760281690140845e-05, "loss": 0.8518, "step": 9807 }, { "epoch": 11.0, "learning_rate": 1.7601408450704225e-05, "loss": 1.0009, "step": 9808 }, { "epoch": 11.0, "learning_rate": 1.76e-05, "loss": 1.0954, "step": 9809 }, { "epoch": 11.0, "learning_rate": 1.7598591549295774e-05, "loss": 0.8533, "step": 9810 }, { "epoch": 11.0, "learning_rate": 1.759718309859155e-05, "loss": 0.9785, "step": 9811 }, { "epoch": 11.0, "learning_rate": 1.7595774647887323e-05, "loss": 0.7159, "step": 9812 }, { "epoch": 11.0, "learning_rate": 1.75943661971831e-05, "loss": 1.2801, "step": 9813 }, { "epoch": 11.0, "learning_rate": 1.7592957746478872e-05, "loss": 1.3435, "step": 9814 }, { "epoch": 11.0, "learning_rate": 1.759154929577465e-05, "loss": 0.9866, "step": 9815 }, { "epoch": 11.0, "learning_rate": 1.759014084507042e-05, "loss": 1.092, "step": 9816 }, { "epoch": 11.01, "learning_rate": 1.7588732394366198e-05, "loss": 1.1498, "step": 9817 }, { "epoch": 11.01, "learning_rate": 1.7587323943661974e-05, "loss": 1.023, "step": 9818 }, { "epoch": 11.01, "learning_rate": 1.7585915492957747e-05, "loss": 1.0295, "step": 9819 }, { "epoch": 11.01, "learning_rate": 1.7584507042253524e-05, "loss": 0.9469, "step": 9820 }, { "epoch": 11.01, "learning_rate": 1.7583098591549297e-05, "loss": 0.9878, "step": 9821 }, { "epoch": 11.01, "learning_rate": 1.7581690140845073e-05, "loss": 1.1808, "step": 9822 }, { "epoch": 11.01, "learning_rate": 1.7580281690140846e-05, "loss": 1.1946, "step": 9823 }, { "epoch": 11.01, "learning_rate": 1.757887323943662e-05, "loss": 1.0106, "step": 9824 }, { "epoch": 11.01, "learning_rate": 1.7577464788732395e-05, "loss": 1.1151, "step": 9825 }, { "epoch": 11.02, "learning_rate": 1.7576056338028168e-05, "loss": 0.9354, "step": 9826 }, { "epoch": 11.02, "learning_rate": 1.7574647887323944e-05, "loss": 1.035, "step": 9827 }, { "epoch": 11.02, "learning_rate": 1.7573239436619717e-05, "loss": 1.0337, "step": 9828 }, { "epoch": 11.02, "learning_rate": 1.7571830985915494e-05, "loss": 1.0221, "step": 9829 }, { "epoch": 11.02, "learning_rate": 1.7570422535211267e-05, "loss": 0.9409, "step": 9830 }, { "epoch": 11.02, "learning_rate": 1.7569014084507043e-05, "loss": 0.9453, "step": 9831 }, { "epoch": 11.02, "learning_rate": 1.7567605633802816e-05, "loss": 0.9375, "step": 9832 }, { "epoch": 11.02, "learning_rate": 1.7566197183098592e-05, "loss": 1.0124, "step": 9833 }, { "epoch": 11.02, "learning_rate": 1.756478873239437e-05, "loss": 0.9985, "step": 9834 }, { "epoch": 11.03, "learning_rate": 1.756338028169014e-05, "loss": 0.9725, "step": 9835 }, { "epoch": 11.03, "learning_rate": 1.7561971830985918e-05, "loss": 0.9273, "step": 9836 }, { "epoch": 11.03, "learning_rate": 1.756056338028169e-05, "loss": 0.8149, "step": 9837 }, { "epoch": 11.03, "learning_rate": 1.7559154929577467e-05, "loss": 1.1616, "step": 9838 }, { "epoch": 11.03, "learning_rate": 1.755774647887324e-05, "loss": 1.0509, "step": 9839 }, { "epoch": 11.03, "learning_rate": 1.7556338028169013e-05, "loss": 0.945, "step": 9840 }, { "epoch": 11.03, "learning_rate": 1.755492957746479e-05, "loss": 0.9911, "step": 9841 }, { "epoch": 11.03, "learning_rate": 1.7553521126760562e-05, "loss": 1.0125, "step": 9842 }, { "epoch": 11.03, "learning_rate": 1.755211267605634e-05, "loss": 0.9115, "step": 9843 }, { "epoch": 11.04, "learning_rate": 1.755070422535211e-05, "loss": 0.9082, "step": 9844 }, { "epoch": 11.04, "learning_rate": 1.7549295774647888e-05, "loss": 0.8937, "step": 9845 }, { "epoch": 11.04, "learning_rate": 1.754788732394366e-05, "loss": 1.0856, "step": 9846 }, { "epoch": 11.04, "learning_rate": 1.7546478873239437e-05, "loss": 0.9628, "step": 9847 }, { "epoch": 11.04, "learning_rate": 1.7545070422535213e-05, "loss": 0.7953, "step": 9848 }, { "epoch": 11.04, "learning_rate": 1.7543661971830986e-05, "loss": 0.884, "step": 9849 }, { "epoch": 11.04, "learning_rate": 1.7542253521126763e-05, "loss": 1.0463, "step": 9850 }, { "epoch": 11.04, "learning_rate": 1.7540845070422536e-05, "loss": 0.9918, "step": 9851 }, { "epoch": 11.04, "learning_rate": 1.7539436619718312e-05, "loss": 1.0731, "step": 9852 }, { "epoch": 11.05, "learning_rate": 1.7538028169014085e-05, "loss": 1.0061, "step": 9853 }, { "epoch": 11.05, "learning_rate": 1.753661971830986e-05, "loss": 1.0537, "step": 9854 }, { "epoch": 11.05, "learning_rate": 1.7535211267605634e-05, "loss": 0.9509, "step": 9855 }, { "epoch": 11.05, "learning_rate": 1.7533802816901407e-05, "loss": 0.826, "step": 9856 }, { "epoch": 11.05, "learning_rate": 1.7532394366197183e-05, "loss": 0.8693, "step": 9857 }, { "epoch": 11.05, "learning_rate": 1.7530985915492956e-05, "loss": 1.0972, "step": 9858 }, { "epoch": 11.05, "learning_rate": 1.7529577464788733e-05, "loss": 0.6817, "step": 9859 }, { "epoch": 11.05, "learning_rate": 1.7528169014084506e-05, "loss": 0.952, "step": 9860 }, { "epoch": 11.05, "learning_rate": 1.7526760563380282e-05, "loss": 0.7769, "step": 9861 }, { "epoch": 11.06, "learning_rate": 1.7525352112676055e-05, "loss": 0.6266, "step": 9862 }, { "epoch": 11.06, "learning_rate": 1.752394366197183e-05, "loss": 1.3678, "step": 9863 }, { "epoch": 11.06, "learning_rate": 1.7522535211267608e-05, "loss": 1.1431, "step": 9864 }, { "epoch": 11.06, "learning_rate": 1.752112676056338e-05, "loss": 1.1488, "step": 9865 }, { "epoch": 11.06, "learning_rate": 1.7519718309859157e-05, "loss": 1.0131, "step": 9866 }, { "epoch": 11.06, "learning_rate": 1.751830985915493e-05, "loss": 1.0787, "step": 9867 }, { "epoch": 11.06, "learning_rate": 1.7516901408450706e-05, "loss": 1.0622, "step": 9868 }, { "epoch": 11.06, "learning_rate": 1.751549295774648e-05, "loss": 1.0514, "step": 9869 }, { "epoch": 11.07, "learning_rate": 1.7514084507042255e-05, "loss": 1.1576, "step": 9870 }, { "epoch": 11.07, "learning_rate": 1.751267605633803e-05, "loss": 0.9928, "step": 9871 }, { "epoch": 11.07, "learning_rate": 1.75112676056338e-05, "loss": 1.0541, "step": 9872 }, { "epoch": 11.07, "learning_rate": 1.7509859154929578e-05, "loss": 1.0998, "step": 9873 }, { "epoch": 11.07, "learning_rate": 1.750845070422535e-05, "loss": 1.1397, "step": 9874 }, { "epoch": 11.07, "learning_rate": 1.7507042253521127e-05, "loss": 0.959, "step": 9875 }, { "epoch": 11.07, "learning_rate": 1.75056338028169e-05, "loss": 1.0439, "step": 9876 }, { "epoch": 11.07, "learning_rate": 1.7504225352112676e-05, "loss": 0.9448, "step": 9877 }, { "epoch": 11.07, "learning_rate": 1.7502816901408453e-05, "loss": 1.0925, "step": 9878 }, { "epoch": 11.08, "learning_rate": 1.7501408450704226e-05, "loss": 0.9753, "step": 9879 }, { "epoch": 11.08, "learning_rate": 1.7500000000000002e-05, "loss": 0.9279, "step": 9880 }, { "epoch": 11.08, "learning_rate": 1.7498591549295775e-05, "loss": 0.8879, "step": 9881 }, { "epoch": 11.08, "learning_rate": 1.749718309859155e-05, "loss": 0.9198, "step": 9882 }, { "epoch": 11.08, "learning_rate": 1.7495774647887324e-05, "loss": 1.0603, "step": 9883 }, { "epoch": 11.08, "learning_rate": 1.74943661971831e-05, "loss": 0.9117, "step": 9884 }, { "epoch": 11.08, "learning_rate": 1.7492957746478873e-05, "loss": 0.9632, "step": 9885 }, { "epoch": 11.08, "learning_rate": 1.749154929577465e-05, "loss": 0.7776, "step": 9886 }, { "epoch": 11.08, "learning_rate": 1.7490140845070423e-05, "loss": 1.0469, "step": 9887 }, { "epoch": 11.09, "learning_rate": 1.7488732394366196e-05, "loss": 0.9366, "step": 9888 }, { "epoch": 11.09, "learning_rate": 1.7487323943661972e-05, "loss": 0.9885, "step": 9889 }, { "epoch": 11.09, "learning_rate": 1.7485915492957745e-05, "loss": 0.9055, "step": 9890 }, { "epoch": 11.09, "learning_rate": 1.748450704225352e-05, "loss": 0.925, "step": 9891 }, { "epoch": 11.09, "learning_rate": 1.7483098591549294e-05, "loss": 0.8615, "step": 9892 }, { "epoch": 11.09, "learning_rate": 1.7481690140845074e-05, "loss": 0.9938, "step": 9893 }, { "epoch": 11.09, "learning_rate": 1.7480281690140847e-05, "loss": 0.9868, "step": 9894 }, { "epoch": 11.09, "learning_rate": 1.747887323943662e-05, "loss": 1.013, "step": 9895 }, { "epoch": 11.09, "learning_rate": 1.7477464788732396e-05, "loss": 1.0601, "step": 9896 }, { "epoch": 11.1, "learning_rate": 1.747605633802817e-05, "loss": 1.1235, "step": 9897 }, { "epoch": 11.1, "learning_rate": 1.7474647887323945e-05, "loss": 1.0502, "step": 9898 }, { "epoch": 11.1, "learning_rate": 1.7473239436619718e-05, "loss": 1.0155, "step": 9899 }, { "epoch": 11.1, "learning_rate": 1.7471830985915495e-05, "loss": 0.9316, "step": 9900 }, { "epoch": 11.1, "learning_rate": 1.7470422535211268e-05, "loss": 1.0037, "step": 9901 }, { "epoch": 11.1, "learning_rate": 1.7469014084507044e-05, "loss": 0.8585, "step": 9902 }, { "epoch": 11.1, "learning_rate": 1.7467605633802817e-05, "loss": 1.1196, "step": 9903 }, { "epoch": 11.1, "learning_rate": 1.746619718309859e-05, "loss": 1.0271, "step": 9904 }, { "epoch": 11.1, "learning_rate": 1.7464788732394366e-05, "loss": 1.0046, "step": 9905 }, { "epoch": 11.11, "learning_rate": 1.746338028169014e-05, "loss": 0.8354, "step": 9906 }, { "epoch": 11.11, "learning_rate": 1.7461971830985915e-05, "loss": 0.8962, "step": 9907 }, { "epoch": 11.11, "learning_rate": 1.7460563380281692e-05, "loss": 0.8577, "step": 9908 }, { "epoch": 11.11, "learning_rate": 1.7459154929577468e-05, "loss": 0.772, "step": 9909 }, { "epoch": 11.11, "learning_rate": 1.745774647887324e-05, "loss": 0.873, "step": 9910 }, { "epoch": 11.11, "learning_rate": 1.7456338028169014e-05, "loss": 0.7612, "step": 9911 }, { "epoch": 11.11, "learning_rate": 1.745492957746479e-05, "loss": 0.8975, "step": 9912 }, { "epoch": 11.11, "learning_rate": 1.7453521126760563e-05, "loss": 1.399, "step": 9913 }, { "epoch": 11.11, "learning_rate": 1.745211267605634e-05, "loss": 1.2609, "step": 9914 }, { "epoch": 11.12, "learning_rate": 1.7450704225352112e-05, "loss": 1.1834, "step": 9915 }, { "epoch": 11.12, "learning_rate": 1.744929577464789e-05, "loss": 0.9829, "step": 9916 }, { "epoch": 11.12, "learning_rate": 1.7447887323943662e-05, "loss": 0.9137, "step": 9917 }, { "epoch": 11.12, "learning_rate": 1.7446478873239438e-05, "loss": 1.0544, "step": 9918 }, { "epoch": 11.12, "learning_rate": 1.744507042253521e-05, "loss": 1.0391, "step": 9919 }, { "epoch": 11.12, "learning_rate": 1.7443661971830984e-05, "loss": 0.9004, "step": 9920 }, { "epoch": 11.12, "learning_rate": 1.744225352112676e-05, "loss": 1.1212, "step": 9921 }, { "epoch": 11.12, "learning_rate": 1.7440845070422533e-05, "loss": 0.9712, "step": 9922 }, { "epoch": 11.12, "learning_rate": 1.7439436619718313e-05, "loss": 0.9547, "step": 9923 }, { "epoch": 11.13, "learning_rate": 1.7438028169014086e-05, "loss": 1.075, "step": 9924 }, { "epoch": 11.13, "learning_rate": 1.7436619718309862e-05, "loss": 1.1064, "step": 9925 }, { "epoch": 11.13, "learning_rate": 1.7435211267605635e-05, "loss": 1.0146, "step": 9926 }, { "epoch": 11.13, "learning_rate": 1.7433802816901408e-05, "loss": 0.9964, "step": 9927 }, { "epoch": 11.13, "learning_rate": 1.7432394366197184e-05, "loss": 1.151, "step": 9928 }, { "epoch": 11.13, "learning_rate": 1.7430985915492957e-05, "loss": 1.0773, "step": 9929 }, { "epoch": 11.13, "learning_rate": 1.7429577464788734e-05, "loss": 0.983, "step": 9930 }, { "epoch": 11.13, "learning_rate": 1.7428169014084507e-05, "loss": 1.0243, "step": 9931 }, { "epoch": 11.13, "learning_rate": 1.7426760563380283e-05, "loss": 1.0034, "step": 9932 }, { "epoch": 11.14, "learning_rate": 1.7425352112676056e-05, "loss": 0.7385, "step": 9933 }, { "epoch": 11.14, "learning_rate": 1.7423943661971832e-05, "loss": 0.9854, "step": 9934 }, { "epoch": 11.14, "learning_rate": 1.7422535211267605e-05, "loss": 0.9573, "step": 9935 }, { "epoch": 11.14, "learning_rate": 1.7421126760563378e-05, "loss": 0.8865, "step": 9936 }, { "epoch": 11.14, "learning_rate": 1.7419718309859154e-05, "loss": 1.0787, "step": 9937 }, { "epoch": 11.14, "learning_rate": 1.741830985915493e-05, "loss": 1.0606, "step": 9938 }, { "epoch": 11.14, "learning_rate": 1.7416901408450707e-05, "loss": 1.003, "step": 9939 }, { "epoch": 11.14, "learning_rate": 1.741549295774648e-05, "loss": 0.9469, "step": 9940 }, { "epoch": 11.14, "learning_rate": 1.7414084507042256e-05, "loss": 0.975, "step": 9941 }, { "epoch": 11.15, "learning_rate": 1.741267605633803e-05, "loss": 1.0552, "step": 9942 }, { "epoch": 11.15, "learning_rate": 1.7411267605633802e-05, "loss": 1.0945, "step": 9943 }, { "epoch": 11.15, "learning_rate": 1.740985915492958e-05, "loss": 0.8797, "step": 9944 }, { "epoch": 11.15, "learning_rate": 1.740845070422535e-05, "loss": 0.8165, "step": 9945 }, { "epoch": 11.15, "learning_rate": 1.7407042253521128e-05, "loss": 1.091, "step": 9946 }, { "epoch": 11.15, "learning_rate": 1.74056338028169e-05, "loss": 1.1409, "step": 9947 }, { "epoch": 11.15, "learning_rate": 1.7404225352112677e-05, "loss": 0.8484, "step": 9948 }, { "epoch": 11.15, "learning_rate": 1.740281690140845e-05, "loss": 1.0985, "step": 9949 }, { "epoch": 11.15, "learning_rate": 1.7401408450704226e-05, "loss": 0.9886, "step": 9950 }, { "epoch": 11.16, "learning_rate": 1.74e-05, "loss": 0.8478, "step": 9951 }, { "epoch": 11.16, "learning_rate": 1.7398591549295772e-05, "loss": 0.8216, "step": 9952 }, { "epoch": 11.16, "learning_rate": 1.739718309859155e-05, "loss": 0.9231, "step": 9953 }, { "epoch": 11.16, "learning_rate": 1.7395774647887325e-05, "loss": 0.9964, "step": 9954 }, { "epoch": 11.16, "learning_rate": 1.73943661971831e-05, "loss": 0.7401, "step": 9955 }, { "epoch": 11.16, "learning_rate": 1.7392957746478874e-05, "loss": 0.8875, "step": 9956 }, { "epoch": 11.16, "learning_rate": 1.739154929577465e-05, "loss": 0.9742, "step": 9957 }, { "epoch": 11.16, "learning_rate": 1.7390140845070424e-05, "loss": 1.1276, "step": 9958 }, { "epoch": 11.16, "learning_rate": 1.7388732394366197e-05, "loss": 0.7767, "step": 9959 }, { "epoch": 11.17, "learning_rate": 1.7387323943661973e-05, "loss": 0.5583, "step": 9960 }, { "epoch": 11.17, "learning_rate": 1.7385915492957746e-05, "loss": 0.9703, "step": 9961 }, { "epoch": 11.17, "learning_rate": 1.7384507042253522e-05, "loss": 0.7704, "step": 9962 }, { "epoch": 11.17, "learning_rate": 1.7383098591549295e-05, "loss": 1.3025, "step": 9963 }, { "epoch": 11.17, "learning_rate": 1.738169014084507e-05, "loss": 1.1199, "step": 9964 }, { "epoch": 11.17, "learning_rate": 1.7380281690140844e-05, "loss": 1.0958, "step": 9965 }, { "epoch": 11.17, "learning_rate": 1.737887323943662e-05, "loss": 1.0555, "step": 9966 }, { "epoch": 11.17, "learning_rate": 1.7377464788732394e-05, "loss": 1.0498, "step": 9967 }, { "epoch": 11.17, "learning_rate": 1.7376056338028167e-05, "loss": 1.0055, "step": 9968 }, { "epoch": 11.18, "learning_rate": 1.7374647887323946e-05, "loss": 0.9895, "step": 9969 }, { "epoch": 11.18, "learning_rate": 1.737323943661972e-05, "loss": 1.0433, "step": 9970 }, { "epoch": 11.18, "learning_rate": 1.7371830985915496e-05, "loss": 1.1268, "step": 9971 }, { "epoch": 11.18, "learning_rate": 1.737042253521127e-05, "loss": 0.9795, "step": 9972 }, { "epoch": 11.18, "learning_rate": 1.7369014084507045e-05, "loss": 1.0992, "step": 9973 }, { "epoch": 11.18, "learning_rate": 1.7367605633802818e-05, "loss": 0.996, "step": 9974 }, { "epoch": 11.18, "learning_rate": 1.736619718309859e-05, "loss": 1.0689, "step": 9975 }, { "epoch": 11.18, "learning_rate": 1.7364788732394367e-05, "loss": 0.8219, "step": 9976 }, { "epoch": 11.18, "learning_rate": 1.736338028169014e-05, "loss": 1.0911, "step": 9977 }, { "epoch": 11.19, "learning_rate": 1.7361971830985916e-05, "loss": 0.9773, "step": 9978 }, { "epoch": 11.19, "learning_rate": 1.736056338028169e-05, "loss": 1.1115, "step": 9979 }, { "epoch": 11.19, "learning_rate": 1.7359154929577466e-05, "loss": 0.926, "step": 9980 }, { "epoch": 11.19, "learning_rate": 1.735774647887324e-05, "loss": 0.9851, "step": 9981 }, { "epoch": 11.19, "learning_rate": 1.735633802816901e-05, "loss": 1.1672, "step": 9982 }, { "epoch": 11.19, "learning_rate": 1.7354929577464788e-05, "loss": 0.9204, "step": 9983 }, { "epoch": 11.19, "learning_rate": 1.7353521126760564e-05, "loss": 1.088, "step": 9984 }, { "epoch": 11.19, "learning_rate": 1.735211267605634e-05, "loss": 0.818, "step": 9985 }, { "epoch": 11.2, "learning_rate": 1.7350704225352113e-05, "loss": 0.9348, "step": 9986 }, { "epoch": 11.2, "learning_rate": 1.734929577464789e-05, "loss": 0.8705, "step": 9987 }, { "epoch": 11.2, "learning_rate": 1.7347887323943663e-05, "loss": 0.9377, "step": 9988 }, { "epoch": 11.2, "learning_rate": 1.734647887323944e-05, "loss": 0.8576, "step": 9989 }, { "epoch": 11.2, "learning_rate": 1.7345070422535212e-05, "loss": 0.9805, "step": 9990 }, { "epoch": 11.2, "learning_rate": 1.7343661971830985e-05, "loss": 0.9297, "step": 9991 }, { "epoch": 11.2, "learning_rate": 1.734225352112676e-05, "loss": 0.8642, "step": 9992 }, { "epoch": 11.2, "learning_rate": 1.7340845070422534e-05, "loss": 0.9496, "step": 9993 }, { "epoch": 11.2, "learning_rate": 1.733943661971831e-05, "loss": 1.1164, "step": 9994 }, { "epoch": 11.21, "learning_rate": 1.7338028169014083e-05, "loss": 0.8889, "step": 9995 }, { "epoch": 11.21, "learning_rate": 1.733661971830986e-05, "loss": 0.8411, "step": 9996 }, { "epoch": 11.21, "learning_rate": 1.7335211267605633e-05, "loss": 0.8733, "step": 9997 }, { "epoch": 11.21, "learning_rate": 1.7333802816901406e-05, "loss": 0.9088, "step": 9998 }, { "epoch": 11.21, "learning_rate": 1.7332394366197185e-05, "loss": 0.994, "step": 9999 }, { "epoch": 11.21, "learning_rate": 1.733098591549296e-05, "loss": 1.0585, "step": 10000 }, { "epoch": 11.21, "learning_rate": 1.7329577464788735e-05, "loss": 0.937, "step": 10001 }, { "epoch": 11.21, "learning_rate": 1.7328169014084508e-05, "loss": 1.0859, "step": 10002 }, { "epoch": 11.21, "learning_rate": 1.7326760563380284e-05, "loss": 0.8859, "step": 10003 }, { "epoch": 11.22, "learning_rate": 1.7325352112676057e-05, "loss": 1.0821, "step": 10004 }, { "epoch": 11.22, "learning_rate": 1.7323943661971833e-05, "loss": 0.9039, "step": 10005 }, { "epoch": 11.22, "learning_rate": 1.7322535211267606e-05, "loss": 0.8644, "step": 10006 }, { "epoch": 11.22, "learning_rate": 1.732112676056338e-05, "loss": 0.953, "step": 10007 }, { "epoch": 11.22, "learning_rate": 1.7319718309859155e-05, "loss": 0.7728, "step": 10008 }, { "epoch": 11.22, "learning_rate": 1.731830985915493e-05, "loss": 0.7893, "step": 10009 }, { "epoch": 11.22, "learning_rate": 1.7316901408450705e-05, "loss": 0.8234, "step": 10010 }, { "epoch": 11.22, "learning_rate": 1.7315492957746478e-05, "loss": 0.9607, "step": 10011 }, { "epoch": 11.22, "learning_rate": 1.7314084507042254e-05, "loss": 0.7435, "step": 10012 }, { "epoch": 11.23, "learning_rate": 1.7312676056338027e-05, "loss": 1.168, "step": 10013 }, { "epoch": 11.23, "learning_rate": 1.7311267605633803e-05, "loss": 1.2087, "step": 10014 }, { "epoch": 11.23, "learning_rate": 1.730985915492958e-05, "loss": 1.0455, "step": 10015 }, { "epoch": 11.23, "learning_rate": 1.7308450704225353e-05, "loss": 1.2, "step": 10016 }, { "epoch": 11.23, "learning_rate": 1.730704225352113e-05, "loss": 0.8962, "step": 10017 }, { "epoch": 11.23, "learning_rate": 1.7305633802816902e-05, "loss": 0.9515, "step": 10018 }, { "epoch": 11.23, "learning_rate": 1.7304225352112678e-05, "loss": 1.1641, "step": 10019 }, { "epoch": 11.23, "learning_rate": 1.730281690140845e-05, "loss": 1.001, "step": 10020 }, { "epoch": 11.23, "learning_rate": 1.7301408450704227e-05, "loss": 0.9594, "step": 10021 }, { "epoch": 11.24, "learning_rate": 1.73e-05, "loss": 1.0455, "step": 10022 }, { "epoch": 11.24, "learning_rate": 1.7298591549295773e-05, "loss": 0.8441, "step": 10023 }, { "epoch": 11.24, "learning_rate": 1.729718309859155e-05, "loss": 1.0822, "step": 10024 }, { "epoch": 11.24, "learning_rate": 1.7295774647887323e-05, "loss": 1.0864, "step": 10025 }, { "epoch": 11.24, "learning_rate": 1.72943661971831e-05, "loss": 0.9153, "step": 10026 }, { "epoch": 11.24, "learning_rate": 1.7292957746478872e-05, "loss": 1.0064, "step": 10027 }, { "epoch": 11.24, "learning_rate": 1.7291549295774648e-05, "loss": 0.9218, "step": 10028 }, { "epoch": 11.24, "learning_rate": 1.7290140845070425e-05, "loss": 0.7, "step": 10029 }, { "epoch": 11.24, "learning_rate": 1.7288732394366197e-05, "loss": 1.1126, "step": 10030 }, { "epoch": 11.25, "learning_rate": 1.7287323943661974e-05, "loss": 0.8947, "step": 10031 }, { "epoch": 11.25, "learning_rate": 1.7285915492957747e-05, "loss": 0.9898, "step": 10032 }, { "epoch": 11.25, "learning_rate": 1.7284507042253523e-05, "loss": 0.919, "step": 10033 }, { "epoch": 11.25, "learning_rate": 1.7283098591549296e-05, "loss": 0.9998, "step": 10034 }, { "epoch": 11.25, "learning_rate": 1.7281690140845072e-05, "loss": 0.9923, "step": 10035 }, { "epoch": 11.25, "learning_rate": 1.7280281690140845e-05, "loss": 1.027, "step": 10036 }, { "epoch": 11.25, "learning_rate": 1.727887323943662e-05, "loss": 0.7982, "step": 10037 }, { "epoch": 11.25, "learning_rate": 1.7277464788732395e-05, "loss": 1.047, "step": 10038 }, { "epoch": 11.25, "learning_rate": 1.7276056338028168e-05, "loss": 1.0201, "step": 10039 }, { "epoch": 11.26, "learning_rate": 1.7274647887323944e-05, "loss": 0.8889, "step": 10040 }, { "epoch": 11.26, "learning_rate": 1.7273239436619717e-05, "loss": 1.0098, "step": 10041 }, { "epoch": 11.26, "learning_rate": 1.7271830985915493e-05, "loss": 1.0556, "step": 10042 }, { "epoch": 11.26, "learning_rate": 1.7270422535211266e-05, "loss": 0.8002, "step": 10043 }, { "epoch": 11.26, "learning_rate": 1.7269014084507046e-05, "loss": 0.9788, "step": 10044 }, { "epoch": 11.26, "learning_rate": 1.726760563380282e-05, "loss": 0.9831, "step": 10045 }, { "epoch": 11.26, "learning_rate": 1.726619718309859e-05, "loss": 0.8869, "step": 10046 }, { "epoch": 11.26, "learning_rate": 1.7264788732394368e-05, "loss": 0.9211, "step": 10047 }, { "epoch": 11.26, "learning_rate": 1.726338028169014e-05, "loss": 0.7918, "step": 10048 }, { "epoch": 11.27, "learning_rate": 1.7261971830985917e-05, "loss": 0.8854, "step": 10049 }, { "epoch": 11.27, "learning_rate": 1.726056338028169e-05, "loss": 0.8494, "step": 10050 }, { "epoch": 11.27, "learning_rate": 1.7259154929577467e-05, "loss": 0.8869, "step": 10051 }, { "epoch": 11.27, "learning_rate": 1.725774647887324e-05, "loss": 0.9212, "step": 10052 }, { "epoch": 11.27, "learning_rate": 1.7256338028169016e-05, "loss": 1.1249, "step": 10053 }, { "epoch": 11.27, "learning_rate": 1.725492957746479e-05, "loss": 0.9187, "step": 10054 }, { "epoch": 11.27, "learning_rate": 1.7253521126760562e-05, "loss": 0.9012, "step": 10055 }, { "epoch": 11.27, "learning_rate": 1.7252112676056338e-05, "loss": 1.0817, "step": 10056 }, { "epoch": 11.27, "learning_rate": 1.725070422535211e-05, "loss": 0.8235, "step": 10057 }, { "epoch": 11.28, "learning_rate": 1.7249295774647887e-05, "loss": 0.7475, "step": 10058 }, { "epoch": 11.28, "learning_rate": 1.724788732394366e-05, "loss": 0.7044, "step": 10059 }, { "epoch": 11.28, "learning_rate": 1.724647887323944e-05, "loss": 1.0179, "step": 10060 }, { "epoch": 11.28, "learning_rate": 1.7245070422535213e-05, "loss": 0.853, "step": 10061 }, { "epoch": 11.28, "learning_rate": 1.7243661971830986e-05, "loss": 0.9706, "step": 10062 }, { "epoch": 11.28, "learning_rate": 1.7242253521126762e-05, "loss": 1.3565, "step": 10063 }, { "epoch": 11.28, "learning_rate": 1.7240845070422535e-05, "loss": 1.141, "step": 10064 }, { "epoch": 11.28, "learning_rate": 1.723943661971831e-05, "loss": 1.1171, "step": 10065 }, { "epoch": 11.28, "learning_rate": 1.7238028169014084e-05, "loss": 1.1507, "step": 10066 }, { "epoch": 11.29, "learning_rate": 1.723661971830986e-05, "loss": 1.1572, "step": 10067 }, { "epoch": 11.29, "learning_rate": 1.7235211267605634e-05, "loss": 1.0478, "step": 10068 }, { "epoch": 11.29, "learning_rate": 1.723380281690141e-05, "loss": 1.0627, "step": 10069 }, { "epoch": 11.29, "learning_rate": 1.7232394366197183e-05, "loss": 1.1179, "step": 10070 }, { "epoch": 11.29, "learning_rate": 1.7230985915492956e-05, "loss": 1.1463, "step": 10071 }, { "epoch": 11.29, "learning_rate": 1.7229577464788732e-05, "loss": 1.0107, "step": 10072 }, { "epoch": 11.29, "learning_rate": 1.7228169014084505e-05, "loss": 0.9518, "step": 10073 }, { "epoch": 11.29, "learning_rate": 1.722676056338028e-05, "loss": 0.7866, "step": 10074 }, { "epoch": 11.29, "learning_rate": 1.7225352112676058e-05, "loss": 1.0155, "step": 10075 }, { "epoch": 11.3, "learning_rate": 1.7223943661971834e-05, "loss": 1.046, "step": 10076 }, { "epoch": 11.3, "learning_rate": 1.7222535211267607e-05, "loss": 0.896, "step": 10077 }, { "epoch": 11.3, "learning_rate": 1.722112676056338e-05, "loss": 0.8756, "step": 10078 }, { "epoch": 11.3, "learning_rate": 1.7219718309859156e-05, "loss": 1.0137, "step": 10079 }, { "epoch": 11.3, "learning_rate": 1.721830985915493e-05, "loss": 1.0641, "step": 10080 }, { "epoch": 11.3, "learning_rate": 1.7216901408450706e-05, "loss": 1.011, "step": 10081 }, { "epoch": 11.3, "learning_rate": 1.721549295774648e-05, "loss": 1.1293, "step": 10082 }, { "epoch": 11.3, "learning_rate": 1.7214084507042255e-05, "loss": 0.9052, "step": 10083 }, { "epoch": 11.3, "learning_rate": 1.7212676056338028e-05, "loss": 1.0006, "step": 10084 }, { "epoch": 11.31, "learning_rate": 1.72112676056338e-05, "loss": 0.8729, "step": 10085 }, { "epoch": 11.31, "learning_rate": 1.7209859154929577e-05, "loss": 0.9286, "step": 10086 }, { "epoch": 11.31, "learning_rate": 1.720845070422535e-05, "loss": 0.8685, "step": 10087 }, { "epoch": 11.31, "learning_rate": 1.7207042253521126e-05, "loss": 1.1317, "step": 10088 }, { "epoch": 11.31, "learning_rate": 1.72056338028169e-05, "loss": 1.0547, "step": 10089 }, { "epoch": 11.31, "learning_rate": 1.720422535211268e-05, "loss": 0.8586, "step": 10090 }, { "epoch": 11.31, "learning_rate": 1.7202816901408452e-05, "loss": 0.8639, "step": 10091 }, { "epoch": 11.31, "learning_rate": 1.720140845070423e-05, "loss": 0.9654, "step": 10092 }, { "epoch": 11.32, "learning_rate": 1.72e-05, "loss": 1.1271, "step": 10093 }, { "epoch": 11.32, "learning_rate": 1.7198591549295774e-05, "loss": 0.9217, "step": 10094 }, { "epoch": 11.32, "learning_rate": 1.719718309859155e-05, "loss": 1.0768, "step": 10095 }, { "epoch": 11.32, "learning_rate": 1.7195774647887324e-05, "loss": 0.899, "step": 10096 }, { "epoch": 11.32, "learning_rate": 1.71943661971831e-05, "loss": 0.8896, "step": 10097 }, { "epoch": 11.32, "learning_rate": 1.7192957746478873e-05, "loss": 0.8598, "step": 10098 }, { "epoch": 11.32, "learning_rate": 1.719154929577465e-05, "loss": 0.8821, "step": 10099 }, { "epoch": 11.32, "learning_rate": 1.7190140845070422e-05, "loss": 0.8591, "step": 10100 }, { "epoch": 11.32, "learning_rate": 1.7188732394366195e-05, "loss": 0.8128, "step": 10101 }, { "epoch": 11.33, "learning_rate": 1.718732394366197e-05, "loss": 0.8368, "step": 10102 }, { "epoch": 11.33, "learning_rate": 1.7185915492957744e-05, "loss": 1.0278, "step": 10103 }, { "epoch": 11.33, "learning_rate": 1.718450704225352e-05, "loss": 1.0098, "step": 10104 }, { "epoch": 11.33, "learning_rate": 1.7183098591549297e-05, "loss": 0.8424, "step": 10105 }, { "epoch": 11.33, "learning_rate": 1.7181690140845073e-05, "loss": 1.0006, "step": 10106 }, { "epoch": 11.33, "learning_rate": 1.7180281690140846e-05, "loss": 0.9975, "step": 10107 }, { "epoch": 11.33, "learning_rate": 1.7178873239436623e-05, "loss": 0.9596, "step": 10108 }, { "epoch": 11.33, "learning_rate": 1.7177464788732396e-05, "loss": 0.8563, "step": 10109 }, { "epoch": 11.33, "learning_rate": 1.717605633802817e-05, "loss": 1.0983, "step": 10110 }, { "epoch": 11.34, "learning_rate": 1.7174647887323945e-05, "loss": 1.0789, "step": 10111 }, { "epoch": 11.34, "learning_rate": 1.7173239436619718e-05, "loss": 1.165, "step": 10112 }, { "epoch": 11.34, "learning_rate": 1.7171830985915494e-05, "loss": 1.2657, "step": 10113 }, { "epoch": 11.34, "learning_rate": 1.7170422535211267e-05, "loss": 1.1099, "step": 10114 }, { "epoch": 11.34, "learning_rate": 1.7169014084507043e-05, "loss": 1.1803, "step": 10115 }, { "epoch": 11.34, "learning_rate": 1.7167605633802816e-05, "loss": 0.9623, "step": 10116 }, { "epoch": 11.34, "learning_rate": 1.716619718309859e-05, "loss": 1.0665, "step": 10117 }, { "epoch": 11.34, "learning_rate": 1.7164788732394366e-05, "loss": 1.0421, "step": 10118 }, { "epoch": 11.34, "learning_rate": 1.716338028169014e-05, "loss": 1.1063, "step": 10119 }, { "epoch": 11.35, "learning_rate": 1.7161971830985918e-05, "loss": 0.8315, "step": 10120 }, { "epoch": 11.35, "learning_rate": 1.716056338028169e-05, "loss": 1.0246, "step": 10121 }, { "epoch": 11.35, "learning_rate": 1.7159154929577468e-05, "loss": 0.8399, "step": 10122 }, { "epoch": 11.35, "learning_rate": 1.715774647887324e-05, "loss": 0.9365, "step": 10123 }, { "epoch": 11.35, "learning_rate": 1.7156338028169017e-05, "loss": 1.0646, "step": 10124 }, { "epoch": 11.35, "learning_rate": 1.715492957746479e-05, "loss": 0.888, "step": 10125 }, { "epoch": 11.35, "learning_rate": 1.7153521126760563e-05, "loss": 0.9114, "step": 10126 }, { "epoch": 11.35, "learning_rate": 1.715211267605634e-05, "loss": 0.952, "step": 10127 }, { "epoch": 11.35, "learning_rate": 1.7150704225352112e-05, "loss": 1.0812, "step": 10128 }, { "epoch": 11.36, "learning_rate": 1.7149295774647888e-05, "loss": 0.897, "step": 10129 }, { "epoch": 11.36, "learning_rate": 1.714788732394366e-05, "loss": 0.907, "step": 10130 }, { "epoch": 11.36, "learning_rate": 1.7146478873239438e-05, "loss": 1.1557, "step": 10131 }, { "epoch": 11.36, "learning_rate": 1.714507042253521e-05, "loss": 1.0009, "step": 10132 }, { "epoch": 11.36, "learning_rate": 1.7143661971830983e-05, "loss": 0.9804, "step": 10133 }, { "epoch": 11.36, "learning_rate": 1.714225352112676e-05, "loss": 0.9958, "step": 10134 }, { "epoch": 11.36, "learning_rate": 1.7140845070422536e-05, "loss": 1.1185, "step": 10135 }, { "epoch": 11.36, "learning_rate": 1.7139436619718312e-05, "loss": 0.9061, "step": 10136 }, { "epoch": 11.36, "learning_rate": 1.7138028169014085e-05, "loss": 0.95, "step": 10137 }, { "epoch": 11.37, "learning_rate": 1.7136619718309862e-05, "loss": 0.9279, "step": 10138 }, { "epoch": 11.37, "learning_rate": 1.7135211267605635e-05, "loss": 1.0155, "step": 10139 }, { "epoch": 11.37, "learning_rate": 1.713380281690141e-05, "loss": 0.8419, "step": 10140 }, { "epoch": 11.37, "learning_rate": 1.7132394366197184e-05, "loss": 0.9053, "step": 10141 }, { "epoch": 11.37, "learning_rate": 1.7130985915492957e-05, "loss": 1.0178, "step": 10142 }, { "epoch": 11.37, "learning_rate": 1.7129577464788733e-05, "loss": 0.9164, "step": 10143 }, { "epoch": 11.37, "learning_rate": 1.7128169014084506e-05, "loss": 1.0556, "step": 10144 }, { "epoch": 11.37, "learning_rate": 1.7126760563380282e-05, "loss": 0.8525, "step": 10145 }, { "epoch": 11.37, "learning_rate": 1.7125352112676055e-05, "loss": 0.8892, "step": 10146 }, { "epoch": 11.38, "learning_rate": 1.7123943661971832e-05, "loss": 0.827, "step": 10147 }, { "epoch": 11.38, "learning_rate": 1.7122535211267605e-05, "loss": 0.9858, "step": 10148 }, { "epoch": 11.38, "learning_rate": 1.7121126760563378e-05, "loss": 0.9146, "step": 10149 }, { "epoch": 11.38, "learning_rate": 1.7119718309859157e-05, "loss": 0.9991, "step": 10150 }, { "epoch": 11.38, "learning_rate": 1.711830985915493e-05, "loss": 0.8584, "step": 10151 }, { "epoch": 11.38, "learning_rate": 1.7116901408450707e-05, "loss": 0.9015, "step": 10152 }, { "epoch": 11.38, "learning_rate": 1.711549295774648e-05, "loss": 0.9374, "step": 10153 }, { "epoch": 11.38, "learning_rate": 1.7114084507042256e-05, "loss": 0.9275, "step": 10154 }, { "epoch": 11.38, "learning_rate": 1.711267605633803e-05, "loss": 0.9639, "step": 10155 }, { "epoch": 11.39, "learning_rate": 1.7111267605633805e-05, "loss": 0.8179, "step": 10156 }, { "epoch": 11.39, "learning_rate": 1.7109859154929578e-05, "loss": 0.7431, "step": 10157 }, { "epoch": 11.39, "learning_rate": 1.710845070422535e-05, "loss": 0.9399, "step": 10158 }, { "epoch": 11.39, "learning_rate": 1.7107042253521127e-05, "loss": 0.9939, "step": 10159 }, { "epoch": 11.39, "learning_rate": 1.71056338028169e-05, "loss": 1.1197, "step": 10160 }, { "epoch": 11.39, "learning_rate": 1.7104225352112677e-05, "loss": 0.917, "step": 10161 }, { "epoch": 11.39, "learning_rate": 1.710281690140845e-05, "loss": 1.0054, "step": 10162 }, { "epoch": 11.39, "learning_rate": 1.7101408450704226e-05, "loss": 1.418, "step": 10163 }, { "epoch": 11.39, "learning_rate": 1.71e-05, "loss": 1.0818, "step": 10164 }, { "epoch": 11.4, "learning_rate": 1.7098591549295775e-05, "loss": 1.1604, "step": 10165 }, { "epoch": 11.4, "learning_rate": 1.709718309859155e-05, "loss": 1.0216, "step": 10166 }, { "epoch": 11.4, "learning_rate": 1.7095774647887325e-05, "loss": 1.0835, "step": 10167 }, { "epoch": 11.4, "learning_rate": 1.70943661971831e-05, "loss": 0.9535, "step": 10168 }, { "epoch": 11.4, "learning_rate": 1.7092957746478874e-05, "loss": 0.8335, "step": 10169 }, { "epoch": 11.4, "learning_rate": 1.709154929577465e-05, "loss": 0.9505, "step": 10170 }, { "epoch": 11.4, "learning_rate": 1.7090140845070423e-05, "loss": 0.9654, "step": 10171 }, { "epoch": 11.4, "learning_rate": 1.70887323943662e-05, "loss": 1.1077, "step": 10172 }, { "epoch": 11.4, "learning_rate": 1.7087323943661972e-05, "loss": 1.0219, "step": 10173 }, { "epoch": 11.41, "learning_rate": 1.7085915492957745e-05, "loss": 0.9537, "step": 10174 }, { "epoch": 11.41, "learning_rate": 1.708450704225352e-05, "loss": 0.9274, "step": 10175 }, { "epoch": 11.41, "learning_rate": 1.7083098591549295e-05, "loss": 1.0014, "step": 10176 }, { "epoch": 11.41, "learning_rate": 1.708169014084507e-05, "loss": 1.0801, "step": 10177 }, { "epoch": 11.41, "learning_rate": 1.7080281690140844e-05, "loss": 0.8107, "step": 10178 }, { "epoch": 11.41, "learning_rate": 1.707887323943662e-05, "loss": 1.0468, "step": 10179 }, { "epoch": 11.41, "learning_rate": 1.7077464788732393e-05, "loss": 0.8269, "step": 10180 }, { "epoch": 11.41, "learning_rate": 1.707605633802817e-05, "loss": 0.8574, "step": 10181 }, { "epoch": 11.41, "learning_rate": 1.7074647887323946e-05, "loss": 0.8445, "step": 10182 }, { "epoch": 11.42, "learning_rate": 1.707323943661972e-05, "loss": 1.017, "step": 10183 }, { "epoch": 11.42, "learning_rate": 1.7071830985915495e-05, "loss": 0.971, "step": 10184 }, { "epoch": 11.42, "learning_rate": 1.7070422535211268e-05, "loss": 1.0838, "step": 10185 }, { "epoch": 11.42, "learning_rate": 1.7069014084507044e-05, "loss": 0.9542, "step": 10186 }, { "epoch": 11.42, "learning_rate": 1.7067605633802817e-05, "loss": 0.874, "step": 10187 }, { "epoch": 11.42, "learning_rate": 1.706619718309859e-05, "loss": 0.7236, "step": 10188 }, { "epoch": 11.42, "learning_rate": 1.7064788732394367e-05, "loss": 0.8901, "step": 10189 }, { "epoch": 11.42, "learning_rate": 1.706338028169014e-05, "loss": 0.9778, "step": 10190 }, { "epoch": 11.42, "learning_rate": 1.7061971830985916e-05, "loss": 0.9269, "step": 10191 }, { "epoch": 11.43, "learning_rate": 1.706056338028169e-05, "loss": 1.004, "step": 10192 }, { "epoch": 11.43, "learning_rate": 1.7059154929577465e-05, "loss": 0.9434, "step": 10193 }, { "epoch": 11.43, "learning_rate": 1.7057746478873238e-05, "loss": 0.9748, "step": 10194 }, { "epoch": 11.43, "learning_rate": 1.7056338028169014e-05, "loss": 0.9464, "step": 10195 }, { "epoch": 11.43, "learning_rate": 1.705492957746479e-05, "loss": 0.8783, "step": 10196 }, { "epoch": 11.43, "learning_rate": 1.7053521126760564e-05, "loss": 1.1548, "step": 10197 }, { "epoch": 11.43, "learning_rate": 1.705211267605634e-05, "loss": 0.8237, "step": 10198 }, { "epoch": 11.43, "learning_rate": 1.7050704225352113e-05, "loss": 0.7954, "step": 10199 }, { "epoch": 11.43, "learning_rate": 1.704929577464789e-05, "loss": 0.7747, "step": 10200 }, { "epoch": 11.44, "learning_rate": 1.7047887323943662e-05, "loss": 0.8097, "step": 10201 }, { "epoch": 11.44, "learning_rate": 1.704647887323944e-05, "loss": 1.0878, "step": 10202 }, { "epoch": 11.44, "learning_rate": 1.704507042253521e-05, "loss": 0.8738, "step": 10203 }, { "epoch": 11.44, "learning_rate": 1.7043661971830984e-05, "loss": 1.0956, "step": 10204 }, { "epoch": 11.44, "learning_rate": 1.704225352112676e-05, "loss": 0.9029, "step": 10205 }, { "epoch": 11.44, "learning_rate": 1.7040845070422534e-05, "loss": 0.8179, "step": 10206 }, { "epoch": 11.44, "learning_rate": 1.703943661971831e-05, "loss": 0.763, "step": 10207 }, { "epoch": 11.44, "learning_rate": 1.7038028169014083e-05, "loss": 0.6584, "step": 10208 }, { "epoch": 11.45, "learning_rate": 1.703661971830986e-05, "loss": 0.948, "step": 10209 }, { "epoch": 11.45, "learning_rate": 1.7035211267605632e-05, "loss": 0.9528, "step": 10210 }, { "epoch": 11.45, "learning_rate": 1.7033802816901412e-05, "loss": 0.7704, "step": 10211 }, { "epoch": 11.45, "learning_rate": 1.7032394366197185e-05, "loss": 0.6836, "step": 10212 }, { "epoch": 11.45, "learning_rate": 1.7030985915492958e-05, "loss": 1.1776, "step": 10213 }, { "epoch": 11.45, "learning_rate": 1.7029577464788734e-05, "loss": 1.1477, "step": 10214 }, { "epoch": 11.45, "learning_rate": 1.7028169014084507e-05, "loss": 1.1418, "step": 10215 }, { "epoch": 11.45, "learning_rate": 1.7026760563380283e-05, "loss": 0.9405, "step": 10216 }, { "epoch": 11.45, "learning_rate": 1.7025352112676056e-05, "loss": 0.8513, "step": 10217 }, { "epoch": 11.46, "learning_rate": 1.7023943661971833e-05, "loss": 0.9262, "step": 10218 }, { "epoch": 11.46, "learning_rate": 1.7022535211267606e-05, "loss": 1.1296, "step": 10219 }, { "epoch": 11.46, "learning_rate": 1.702112676056338e-05, "loss": 1.1722, "step": 10220 }, { "epoch": 11.46, "learning_rate": 1.7019718309859155e-05, "loss": 0.8432, "step": 10221 }, { "epoch": 11.46, "learning_rate": 1.7018309859154928e-05, "loss": 0.9615, "step": 10222 }, { "epoch": 11.46, "learning_rate": 1.7016901408450704e-05, "loss": 0.9503, "step": 10223 }, { "epoch": 11.46, "learning_rate": 1.7015492957746477e-05, "loss": 0.967, "step": 10224 }, { "epoch": 11.46, "learning_rate": 1.7014084507042253e-05, "loss": 1.0083, "step": 10225 }, { "epoch": 11.46, "learning_rate": 1.701267605633803e-05, "loss": 1.0239, "step": 10226 }, { "epoch": 11.47, "learning_rate": 1.7011267605633806e-05, "loss": 0.9945, "step": 10227 }, { "epoch": 11.47, "learning_rate": 1.700985915492958e-05, "loss": 0.903, "step": 10228 }, { "epoch": 11.47, "learning_rate": 1.7008450704225352e-05, "loss": 0.9197, "step": 10229 }, { "epoch": 11.47, "learning_rate": 1.700704225352113e-05, "loss": 0.9206, "step": 10230 }, { "epoch": 11.47, "learning_rate": 1.70056338028169e-05, "loss": 1.0953, "step": 10231 }, { "epoch": 11.47, "learning_rate": 1.7004225352112678e-05, "loss": 0.8076, "step": 10232 }, { "epoch": 11.47, "learning_rate": 1.700281690140845e-05, "loss": 0.9414, "step": 10233 }, { "epoch": 11.47, "learning_rate": 1.7001408450704227e-05, "loss": 0.9797, "step": 10234 }, { "epoch": 11.47, "learning_rate": 1.7e-05, "loss": 0.7782, "step": 10235 }, { "epoch": 11.48, "learning_rate": 1.6998591549295773e-05, "loss": 1.084, "step": 10236 }, { "epoch": 11.48, "learning_rate": 1.699718309859155e-05, "loss": 0.9973, "step": 10237 }, { "epoch": 11.48, "learning_rate": 1.6995774647887322e-05, "loss": 0.8819, "step": 10238 }, { "epoch": 11.48, "learning_rate": 1.69943661971831e-05, "loss": 1.1236, "step": 10239 }, { "epoch": 11.48, "learning_rate": 1.699295774647887e-05, "loss": 0.8966, "step": 10240 }, { "epoch": 11.48, "learning_rate": 1.699154929577465e-05, "loss": 0.8857, "step": 10241 }, { "epoch": 11.48, "learning_rate": 1.6990140845070424e-05, "loss": 1.0823, "step": 10242 }, { "epoch": 11.48, "learning_rate": 1.69887323943662e-05, "loss": 0.8225, "step": 10243 }, { "epoch": 11.48, "learning_rate": 1.6987323943661973e-05, "loss": 0.7751, "step": 10244 }, { "epoch": 11.49, "learning_rate": 1.6985915492957746e-05, "loss": 0.8823, "step": 10245 }, { "epoch": 11.49, "learning_rate": 1.6984507042253523e-05, "loss": 0.8167, "step": 10246 }, { "epoch": 11.49, "learning_rate": 1.6983098591549296e-05, "loss": 0.9429, "step": 10247 }, { "epoch": 11.49, "learning_rate": 1.6981690140845072e-05, "loss": 0.9093, "step": 10248 }, { "epoch": 11.49, "learning_rate": 1.6980281690140845e-05, "loss": 0.9582, "step": 10249 }, { "epoch": 11.49, "learning_rate": 1.697887323943662e-05, "loss": 0.8809, "step": 10250 }, { "epoch": 11.49, "learning_rate": 1.6977464788732394e-05, "loss": 0.856, "step": 10251 }, { "epoch": 11.49, "learning_rate": 1.6976056338028167e-05, "loss": 0.9708, "step": 10252 }, { "epoch": 11.49, "learning_rate": 1.6974647887323943e-05, "loss": 0.9894, "step": 10253 }, { "epoch": 11.5, "learning_rate": 1.6973239436619716e-05, "loss": 0.9546, "step": 10254 }, { "epoch": 11.5, "learning_rate": 1.6971830985915493e-05, "loss": 0.908, "step": 10255 }, { "epoch": 11.5, "learning_rate": 1.697042253521127e-05, "loss": 1.0213, "step": 10256 }, { "epoch": 11.5, "learning_rate": 1.6969014084507045e-05, "loss": 0.8861, "step": 10257 }, { "epoch": 11.5, "learning_rate": 1.6967605633802818e-05, "loss": 0.7109, "step": 10258 }, { "epoch": 11.5, "learning_rate": 1.6966197183098595e-05, "loss": 0.7374, "step": 10259 }, { "epoch": 11.5, "learning_rate": 1.6964788732394367e-05, "loss": 0.9972, "step": 10260 }, { "epoch": 11.5, "learning_rate": 1.696338028169014e-05, "loss": 0.8042, "step": 10261 }, { "epoch": 11.5, "learning_rate": 1.6961971830985917e-05, "loss": 0.7866, "step": 10262 }, { "epoch": 11.51, "learning_rate": 1.696056338028169e-05, "loss": 1.3192, "step": 10263 }, { "epoch": 11.51, "learning_rate": 1.6959154929577466e-05, "loss": 1.229, "step": 10264 }, { "epoch": 11.51, "learning_rate": 1.695774647887324e-05, "loss": 1.197, "step": 10265 }, { "epoch": 11.51, "learning_rate": 1.6956338028169015e-05, "loss": 1.1163, "step": 10266 }, { "epoch": 11.51, "learning_rate": 1.6954929577464788e-05, "loss": 0.9126, "step": 10267 }, { "epoch": 11.51, "learning_rate": 1.695352112676056e-05, "loss": 1.1643, "step": 10268 }, { "epoch": 11.51, "learning_rate": 1.6952112676056338e-05, "loss": 0.9372, "step": 10269 }, { "epoch": 11.51, "learning_rate": 1.695070422535211e-05, "loss": 0.9837, "step": 10270 }, { "epoch": 11.51, "learning_rate": 1.694929577464789e-05, "loss": 1.0601, "step": 10271 }, { "epoch": 11.52, "learning_rate": 1.6947887323943663e-05, "loss": 0.9591, "step": 10272 }, { "epoch": 11.52, "learning_rate": 1.694647887323944e-05, "loss": 0.9179, "step": 10273 }, { "epoch": 11.52, "learning_rate": 1.6945070422535212e-05, "loss": 0.9769, "step": 10274 }, { "epoch": 11.52, "learning_rate": 1.694366197183099e-05, "loss": 0.9536, "step": 10275 }, { "epoch": 11.52, "learning_rate": 1.6942253521126762e-05, "loss": 0.8501, "step": 10276 }, { "epoch": 11.52, "learning_rate": 1.6940845070422535e-05, "loss": 1.0173, "step": 10277 }, { "epoch": 11.52, "learning_rate": 1.693943661971831e-05, "loss": 0.8051, "step": 10278 }, { "epoch": 11.52, "learning_rate": 1.6938028169014084e-05, "loss": 0.9355, "step": 10279 }, { "epoch": 11.52, "learning_rate": 1.693661971830986e-05, "loss": 0.9274, "step": 10280 }, { "epoch": 11.53, "learning_rate": 1.6935211267605633e-05, "loss": 0.9435, "step": 10281 }, { "epoch": 11.53, "learning_rate": 1.693380281690141e-05, "loss": 0.9341, "step": 10282 }, { "epoch": 11.53, "learning_rate": 1.6932394366197182e-05, "loss": 1.015, "step": 10283 }, { "epoch": 11.53, "learning_rate": 1.6930985915492955e-05, "loss": 1.0746, "step": 10284 }, { "epoch": 11.53, "learning_rate": 1.6929577464788732e-05, "loss": 0.9222, "step": 10285 }, { "epoch": 11.53, "learning_rate": 1.6928169014084508e-05, "loss": 1.0816, "step": 10286 }, { "epoch": 11.53, "learning_rate": 1.6926760563380284e-05, "loss": 0.686, "step": 10287 }, { "epoch": 11.53, "learning_rate": 1.6925352112676057e-05, "loss": 0.8634, "step": 10288 }, { "epoch": 11.53, "learning_rate": 1.6923943661971834e-05, "loss": 0.9262, "step": 10289 }, { "epoch": 11.54, "learning_rate": 1.6922535211267607e-05, "loss": 1.0021, "step": 10290 }, { "epoch": 11.54, "learning_rate": 1.692112676056338e-05, "loss": 0.941, "step": 10291 }, { "epoch": 11.54, "learning_rate": 1.6919718309859156e-05, "loss": 1.0346, "step": 10292 }, { "epoch": 11.54, "learning_rate": 1.691830985915493e-05, "loss": 0.7694, "step": 10293 }, { "epoch": 11.54, "learning_rate": 1.6916901408450705e-05, "loss": 0.8256, "step": 10294 }, { "epoch": 11.54, "learning_rate": 1.6915492957746478e-05, "loss": 0.8493, "step": 10295 }, { "epoch": 11.54, "learning_rate": 1.6914084507042254e-05, "loss": 0.8303, "step": 10296 }, { "epoch": 11.54, "learning_rate": 1.6912676056338027e-05, "loss": 0.9696, "step": 10297 }, { "epoch": 11.54, "learning_rate": 1.6911267605633804e-05, "loss": 0.9687, "step": 10298 }, { "epoch": 11.55, "learning_rate": 1.6909859154929577e-05, "loss": 0.9692, "step": 10299 }, { "epoch": 11.55, "learning_rate": 1.690845070422535e-05, "loss": 0.7882, "step": 10300 }, { "epoch": 11.55, "learning_rate": 1.6907042253521126e-05, "loss": 0.8637, "step": 10301 }, { "epoch": 11.55, "learning_rate": 1.6905633802816902e-05, "loss": 1.0843, "step": 10302 }, { "epoch": 11.55, "learning_rate": 1.690422535211268e-05, "loss": 0.8248, "step": 10303 }, { "epoch": 11.55, "learning_rate": 1.690281690140845e-05, "loss": 0.9397, "step": 10304 }, { "epoch": 11.55, "learning_rate": 1.6901408450704228e-05, "loss": 0.8381, "step": 10305 }, { "epoch": 11.55, "learning_rate": 1.69e-05, "loss": 0.9654, "step": 10306 }, { "epoch": 11.55, "learning_rate": 1.6898591549295774e-05, "loss": 0.8371, "step": 10307 }, { "epoch": 11.56, "learning_rate": 1.689718309859155e-05, "loss": 0.7198, "step": 10308 }, { "epoch": 11.56, "learning_rate": 1.6895774647887323e-05, "loss": 0.9216, "step": 10309 }, { "epoch": 11.56, "learning_rate": 1.68943661971831e-05, "loss": 1.0342, "step": 10310 }, { "epoch": 11.56, "learning_rate": 1.6892957746478872e-05, "loss": 0.9292, "step": 10311 }, { "epoch": 11.56, "learning_rate": 1.689154929577465e-05, "loss": 0.8038, "step": 10312 }, { "epoch": 11.56, "learning_rate": 1.689014084507042e-05, "loss": 1.4737, "step": 10313 }, { "epoch": 11.56, "learning_rate": 1.6888732394366198e-05, "loss": 1.1833, "step": 10314 }, { "epoch": 11.56, "learning_rate": 1.688732394366197e-05, "loss": 1.2196, "step": 10315 }, { "epoch": 11.57, "learning_rate": 1.6885915492957744e-05, "loss": 1.0589, "step": 10316 }, { "epoch": 11.57, "learning_rate": 1.6884507042253524e-05, "loss": 1.0061, "step": 10317 }, { "epoch": 11.57, "learning_rate": 1.6883098591549296e-05, "loss": 0.9046, "step": 10318 }, { "epoch": 11.57, "learning_rate": 1.6881690140845073e-05, "loss": 1.0311, "step": 10319 }, { "epoch": 11.57, "learning_rate": 1.6880281690140846e-05, "loss": 0.9563, "step": 10320 }, { "epoch": 11.57, "learning_rate": 1.6878873239436622e-05, "loss": 0.9341, "step": 10321 }, { "epoch": 11.57, "learning_rate": 1.6877464788732395e-05, "loss": 0.9877, "step": 10322 }, { "epoch": 11.57, "learning_rate": 1.6876056338028168e-05, "loss": 0.9309, "step": 10323 }, { "epoch": 11.57, "learning_rate": 1.6874647887323944e-05, "loss": 0.9879, "step": 10324 }, { "epoch": 11.58, "learning_rate": 1.6873239436619717e-05, "loss": 1.0261, "step": 10325 }, { "epoch": 11.58, "learning_rate": 1.6871830985915494e-05, "loss": 0.9675, "step": 10326 }, { "epoch": 11.58, "learning_rate": 1.6870422535211267e-05, "loss": 1.0215, "step": 10327 }, { "epoch": 11.58, "learning_rate": 1.6869014084507043e-05, "loss": 0.9336, "step": 10328 }, { "epoch": 11.58, "learning_rate": 1.6867605633802816e-05, "loss": 0.8545, "step": 10329 }, { "epoch": 11.58, "learning_rate": 1.6866197183098592e-05, "loss": 0.9427, "step": 10330 }, { "epoch": 11.58, "learning_rate": 1.6864788732394365e-05, "loss": 0.9095, "step": 10331 }, { "epoch": 11.58, "learning_rate": 1.686338028169014e-05, "loss": 0.8911, "step": 10332 }, { "epoch": 11.58, "learning_rate": 1.6861971830985918e-05, "loss": 1.0631, "step": 10333 }, { "epoch": 11.59, "learning_rate": 1.686056338028169e-05, "loss": 0.7985, "step": 10334 }, { "epoch": 11.59, "learning_rate": 1.6859154929577467e-05, "loss": 0.8083, "step": 10335 }, { "epoch": 11.59, "learning_rate": 1.685774647887324e-05, "loss": 0.8559, "step": 10336 }, { "epoch": 11.59, "learning_rate": 1.6856338028169016e-05, "loss": 0.8928, "step": 10337 }, { "epoch": 11.59, "learning_rate": 1.685492957746479e-05, "loss": 0.8858, "step": 10338 }, { "epoch": 11.59, "learning_rate": 1.6853521126760562e-05, "loss": 0.9233, "step": 10339 }, { "epoch": 11.59, "learning_rate": 1.685211267605634e-05, "loss": 0.8533, "step": 10340 }, { "epoch": 11.59, "learning_rate": 1.685070422535211e-05, "loss": 1.0333, "step": 10341 }, { "epoch": 11.59, "learning_rate": 1.6849295774647888e-05, "loss": 1.0562, "step": 10342 }, { "epoch": 11.6, "learning_rate": 1.684788732394366e-05, "loss": 0.8076, "step": 10343 }, { "epoch": 11.6, "learning_rate": 1.6846478873239437e-05, "loss": 0.8233, "step": 10344 }, { "epoch": 11.6, "learning_rate": 1.684507042253521e-05, "loss": 0.7606, "step": 10345 }, { "epoch": 11.6, "learning_rate": 1.6843661971830986e-05, "loss": 0.9545, "step": 10346 }, { "epoch": 11.6, "learning_rate": 1.6842253521126763e-05, "loss": 0.7175, "step": 10347 }, { "epoch": 11.6, "learning_rate": 1.6840845070422536e-05, "loss": 0.9718, "step": 10348 }, { "epoch": 11.6, "learning_rate": 1.6839436619718312e-05, "loss": 0.7514, "step": 10349 }, { "epoch": 11.6, "learning_rate": 1.6838028169014085e-05, "loss": 0.7221, "step": 10350 }, { "epoch": 11.6, "learning_rate": 1.683661971830986e-05, "loss": 0.8555, "step": 10351 }, { "epoch": 11.61, "learning_rate": 1.6835211267605634e-05, "loss": 0.8119, "step": 10352 }, { "epoch": 11.61, "learning_rate": 1.683380281690141e-05, "loss": 0.7908, "step": 10353 }, { "epoch": 11.61, "learning_rate": 1.6832394366197183e-05, "loss": 0.9449, "step": 10354 }, { "epoch": 11.61, "learning_rate": 1.6830985915492956e-05, "loss": 0.8726, "step": 10355 }, { "epoch": 11.61, "learning_rate": 1.6829577464788733e-05, "loss": 0.8984, "step": 10356 }, { "epoch": 11.61, "learning_rate": 1.6828169014084506e-05, "loss": 0.9046, "step": 10357 }, { "epoch": 11.61, "learning_rate": 1.6826760563380282e-05, "loss": 0.8075, "step": 10358 }, { "epoch": 11.61, "learning_rate": 1.6825352112676055e-05, "loss": 0.7479, "step": 10359 }, { "epoch": 11.61, "learning_rate": 1.682394366197183e-05, "loss": 0.829, "step": 10360 }, { "epoch": 11.62, "learning_rate": 1.6822535211267604e-05, "loss": 0.9575, "step": 10361 }, { "epoch": 11.62, "learning_rate": 1.6821126760563384e-05, "loss": 0.8799, "step": 10362 }, { "epoch": 11.62, "learning_rate": 1.6819718309859157e-05, "loss": 1.197, "step": 10363 }, { "epoch": 11.62, "learning_rate": 1.681830985915493e-05, "loss": 1.0383, "step": 10364 }, { "epoch": 11.62, "learning_rate": 1.6816901408450706e-05, "loss": 1.0753, "step": 10365 }, { "epoch": 11.62, "learning_rate": 1.681549295774648e-05, "loss": 0.9273, "step": 10366 }, { "epoch": 11.62, "learning_rate": 1.6814084507042255e-05, "loss": 1.0225, "step": 10367 }, { "epoch": 11.62, "learning_rate": 1.681267605633803e-05, "loss": 0.8343, "step": 10368 }, { "epoch": 11.62, "learning_rate": 1.6811267605633805e-05, "loss": 1.026, "step": 10369 }, { "epoch": 11.63, "learning_rate": 1.6809859154929578e-05, "loss": 0.9469, "step": 10370 }, { "epoch": 11.63, "learning_rate": 1.680845070422535e-05, "loss": 0.9055, "step": 10371 }, { "epoch": 11.63, "learning_rate": 1.6807042253521127e-05, "loss": 0.9623, "step": 10372 }, { "epoch": 11.63, "learning_rate": 1.68056338028169e-05, "loss": 0.9457, "step": 10373 }, { "epoch": 11.63, "learning_rate": 1.6804225352112676e-05, "loss": 0.9291, "step": 10374 }, { "epoch": 11.63, "learning_rate": 1.680281690140845e-05, "loss": 0.8499, "step": 10375 }, { "epoch": 11.63, "learning_rate": 1.6801408450704225e-05, "loss": 1.1578, "step": 10376 }, { "epoch": 11.63, "learning_rate": 1.6800000000000002e-05, "loss": 0.9377, "step": 10377 }, { "epoch": 11.63, "learning_rate": 1.6798591549295778e-05, "loss": 1.0329, "step": 10378 }, { "epoch": 11.64, "learning_rate": 1.679718309859155e-05, "loss": 0.8285, "step": 10379 }, { "epoch": 11.64, "learning_rate": 1.6795774647887324e-05, "loss": 0.9481, "step": 10380 }, { "epoch": 11.64, "learning_rate": 1.67943661971831e-05, "loss": 0.8547, "step": 10381 }, { "epoch": 11.64, "learning_rate": 1.6792957746478873e-05, "loss": 0.8915, "step": 10382 }, { "epoch": 11.64, "learning_rate": 1.679154929577465e-05, "loss": 0.8151, "step": 10383 }, { "epoch": 11.64, "learning_rate": 1.6790140845070423e-05, "loss": 1.0283, "step": 10384 }, { "epoch": 11.64, "learning_rate": 1.67887323943662e-05, "loss": 0.8153, "step": 10385 }, { "epoch": 11.64, "learning_rate": 1.6787323943661972e-05, "loss": 0.8698, "step": 10386 }, { "epoch": 11.64, "learning_rate": 1.6785915492957745e-05, "loss": 0.9453, "step": 10387 }, { "epoch": 11.65, "learning_rate": 1.678450704225352e-05, "loss": 1.0241, "step": 10388 }, { "epoch": 11.65, "learning_rate": 1.6783098591549294e-05, "loss": 0.7916, "step": 10389 }, { "epoch": 11.65, "learning_rate": 1.678169014084507e-05, "loss": 0.8215, "step": 10390 }, { "epoch": 11.65, "learning_rate": 1.6780281690140843e-05, "loss": 0.9018, "step": 10391 }, { "epoch": 11.65, "learning_rate": 1.6778873239436623e-05, "loss": 0.9108, "step": 10392 }, { "epoch": 11.65, "learning_rate": 1.6777464788732396e-05, "loss": 1.1176, "step": 10393 }, { "epoch": 11.65, "learning_rate": 1.6776056338028172e-05, "loss": 1.0014, "step": 10394 }, { "epoch": 11.65, "learning_rate": 1.6774647887323945e-05, "loss": 0.9742, "step": 10395 }, { "epoch": 11.65, "learning_rate": 1.6773239436619718e-05, "loss": 0.7481, "step": 10396 }, { "epoch": 11.66, "learning_rate": 1.6771830985915495e-05, "loss": 0.9508, "step": 10397 }, { "epoch": 11.66, "learning_rate": 1.6770422535211267e-05, "loss": 0.897, "step": 10398 }, { "epoch": 11.66, "learning_rate": 1.6769014084507044e-05, "loss": 0.949, "step": 10399 }, { "epoch": 11.66, "learning_rate": 1.6767605633802817e-05, "loss": 0.8349, "step": 10400 }, { "epoch": 11.66, "learning_rate": 1.6766197183098593e-05, "loss": 0.7619, "step": 10401 }, { "epoch": 11.66, "learning_rate": 1.6764788732394366e-05, "loss": 0.9868, "step": 10402 }, { "epoch": 11.66, "learning_rate": 1.676338028169014e-05, "loss": 0.7837, "step": 10403 }, { "epoch": 11.66, "learning_rate": 1.6761971830985915e-05, "loss": 0.7669, "step": 10404 }, { "epoch": 11.66, "learning_rate": 1.6760563380281688e-05, "loss": 0.842, "step": 10405 }, { "epoch": 11.67, "learning_rate": 1.6759154929577465e-05, "loss": 0.7434, "step": 10406 }, { "epoch": 11.67, "learning_rate": 1.675774647887324e-05, "loss": 0.7036, "step": 10407 }, { "epoch": 11.67, "learning_rate": 1.6756338028169017e-05, "loss": 0.8457, "step": 10408 }, { "epoch": 11.67, "learning_rate": 1.675492957746479e-05, "loss": 0.5732, "step": 10409 }, { "epoch": 11.67, "learning_rate": 1.6753521126760563e-05, "loss": 1.0498, "step": 10410 }, { "epoch": 11.67, "learning_rate": 1.675211267605634e-05, "loss": 1.0317, "step": 10411 }, { "epoch": 11.67, "learning_rate": 1.6750704225352112e-05, "loss": 1.0992, "step": 10412 }, { "epoch": 11.67, "learning_rate": 1.674929577464789e-05, "loss": 1.2527, "step": 10413 }, { "epoch": 11.67, "learning_rate": 1.674788732394366e-05, "loss": 1.0627, "step": 10414 }, { "epoch": 11.68, "learning_rate": 1.6746478873239438e-05, "loss": 0.8619, "step": 10415 }, { "epoch": 11.68, "learning_rate": 1.674507042253521e-05, "loss": 1.0257, "step": 10416 }, { "epoch": 11.68, "learning_rate": 1.6743661971830987e-05, "loss": 0.9563, "step": 10417 }, { "epoch": 11.68, "learning_rate": 1.674225352112676e-05, "loss": 1.079, "step": 10418 }, { "epoch": 11.68, "learning_rate": 1.6740845070422533e-05, "loss": 0.9165, "step": 10419 }, { "epoch": 11.68, "learning_rate": 1.673943661971831e-05, "loss": 1.0505, "step": 10420 }, { "epoch": 11.68, "learning_rate": 1.6738028169014082e-05, "loss": 0.977, "step": 10421 }, { "epoch": 11.68, "learning_rate": 1.673661971830986e-05, "loss": 0.9965, "step": 10422 }, { "epoch": 11.68, "learning_rate": 1.6735211267605635e-05, "loss": 0.9499, "step": 10423 }, { "epoch": 11.69, "learning_rate": 1.673380281690141e-05, "loss": 1.0535, "step": 10424 }, { "epoch": 11.69, "learning_rate": 1.6732394366197184e-05, "loss": 0.8195, "step": 10425 }, { "epoch": 11.69, "learning_rate": 1.6730985915492957e-05, "loss": 0.9048, "step": 10426 }, { "epoch": 11.69, "learning_rate": 1.6729577464788734e-05, "loss": 0.9631, "step": 10427 }, { "epoch": 11.69, "learning_rate": 1.6728169014084507e-05, "loss": 0.8212, "step": 10428 }, { "epoch": 11.69, "learning_rate": 1.6726760563380283e-05, "loss": 1.0961, "step": 10429 }, { "epoch": 11.69, "learning_rate": 1.6725352112676056e-05, "loss": 0.8774, "step": 10430 }, { "epoch": 11.69, "learning_rate": 1.6723943661971832e-05, "loss": 0.8558, "step": 10431 }, { "epoch": 11.7, "learning_rate": 1.6722535211267605e-05, "loss": 1.0611, "step": 10432 }, { "epoch": 11.7, "learning_rate": 1.672112676056338e-05, "loss": 0.9796, "step": 10433 }, { "epoch": 11.7, "learning_rate": 1.6719718309859154e-05, "loss": 0.9661, "step": 10434 }, { "epoch": 11.7, "learning_rate": 1.6718309859154927e-05, "loss": 0.9516, "step": 10435 }, { "epoch": 11.7, "learning_rate": 1.6716901408450704e-05, "loss": 0.7688, "step": 10436 }, { "epoch": 11.7, "learning_rate": 1.6715492957746477e-05, "loss": 0.8977, "step": 10437 }, { "epoch": 11.7, "learning_rate": 1.6714084507042256e-05, "loss": 0.9327, "step": 10438 }, { "epoch": 11.7, "learning_rate": 1.671267605633803e-05, "loss": 0.9441, "step": 10439 }, { "epoch": 11.7, "learning_rate": 1.6711267605633806e-05, "loss": 0.839, "step": 10440 }, { "epoch": 11.71, "learning_rate": 1.670985915492958e-05, "loss": 0.876, "step": 10441 }, { "epoch": 11.71, "learning_rate": 1.670845070422535e-05, "loss": 0.9731, "step": 10442 }, { "epoch": 11.71, "learning_rate": 1.6707042253521128e-05, "loss": 0.9575, "step": 10443 }, { "epoch": 11.71, "learning_rate": 1.67056338028169e-05, "loss": 0.9195, "step": 10444 }, { "epoch": 11.71, "learning_rate": 1.6704225352112677e-05, "loss": 0.8799, "step": 10445 }, { "epoch": 11.71, "learning_rate": 1.670281690140845e-05, "loss": 0.8701, "step": 10446 }, { "epoch": 11.71, "learning_rate": 1.6701408450704226e-05, "loss": 0.7855, "step": 10447 }, { "epoch": 11.71, "learning_rate": 1.67e-05, "loss": 1.149, "step": 10448 }, { "epoch": 11.71, "learning_rate": 1.6698591549295776e-05, "loss": 0.9383, "step": 10449 }, { "epoch": 11.72, "learning_rate": 1.669718309859155e-05, "loss": 0.9176, "step": 10450 }, { "epoch": 11.72, "learning_rate": 1.669577464788732e-05, "loss": 1.0814, "step": 10451 }, { "epoch": 11.72, "learning_rate": 1.6694366197183098e-05, "loss": 0.8075, "step": 10452 }, { "epoch": 11.72, "learning_rate": 1.6692957746478874e-05, "loss": 1.017, "step": 10453 }, { "epoch": 11.72, "learning_rate": 1.669154929577465e-05, "loss": 0.8057, "step": 10454 }, { "epoch": 11.72, "learning_rate": 1.6690140845070424e-05, "loss": 0.9625, "step": 10455 }, { "epoch": 11.72, "learning_rate": 1.66887323943662e-05, "loss": 1.0564, "step": 10456 }, { "epoch": 11.72, "learning_rate": 1.6687323943661973e-05, "loss": 1.0498, "step": 10457 }, { "epoch": 11.72, "learning_rate": 1.6685915492957746e-05, "loss": 0.8097, "step": 10458 }, { "epoch": 11.73, "learning_rate": 1.6684507042253522e-05, "loss": 0.9343, "step": 10459 }, { "epoch": 11.73, "learning_rate": 1.6683098591549295e-05, "loss": 0.7799, "step": 10460 }, { "epoch": 11.73, "learning_rate": 1.668169014084507e-05, "loss": 0.6758, "step": 10461 }, { "epoch": 11.73, "learning_rate": 1.6680281690140844e-05, "loss": 1.0009, "step": 10462 }, { "epoch": 11.73, "learning_rate": 1.667887323943662e-05, "loss": 1.2079, "step": 10463 }, { "epoch": 11.73, "learning_rate": 1.6677464788732394e-05, "loss": 0.9764, "step": 10464 }, { "epoch": 11.73, "learning_rate": 1.667605633802817e-05, "loss": 0.8132, "step": 10465 }, { "epoch": 11.73, "learning_rate": 1.6674647887323943e-05, "loss": 1.0587, "step": 10466 }, { "epoch": 11.73, "learning_rate": 1.6673239436619716e-05, "loss": 1.1073, "step": 10467 }, { "epoch": 11.74, "learning_rate": 1.6671830985915495e-05, "loss": 0.8654, "step": 10468 }, { "epoch": 11.74, "learning_rate": 1.667042253521127e-05, "loss": 0.9767, "step": 10469 }, { "epoch": 11.74, "learning_rate": 1.6669014084507045e-05, "loss": 0.9663, "step": 10470 }, { "epoch": 11.74, "learning_rate": 1.6667605633802818e-05, "loss": 1.0053, "step": 10471 }, { "epoch": 11.74, "learning_rate": 1.6666197183098594e-05, "loss": 1.0041, "step": 10472 }, { "epoch": 11.74, "learning_rate": 1.6664788732394367e-05, "loss": 0.8963, "step": 10473 }, { "epoch": 11.74, "learning_rate": 1.666338028169014e-05, "loss": 0.944, "step": 10474 }, { "epoch": 11.74, "learning_rate": 1.6661971830985916e-05, "loss": 0.8125, "step": 10475 }, { "epoch": 11.74, "learning_rate": 1.666056338028169e-05, "loss": 0.8726, "step": 10476 }, { "epoch": 11.75, "learning_rate": 1.6659154929577466e-05, "loss": 0.9922, "step": 10477 }, { "epoch": 11.75, "learning_rate": 1.665774647887324e-05, "loss": 0.849, "step": 10478 }, { "epoch": 11.75, "learning_rate": 1.6656338028169015e-05, "loss": 0.9065, "step": 10479 }, { "epoch": 11.75, "learning_rate": 1.6654929577464788e-05, "loss": 0.8228, "step": 10480 }, { "epoch": 11.75, "learning_rate": 1.6653521126760564e-05, "loss": 1.0, "step": 10481 }, { "epoch": 11.75, "learning_rate": 1.6652112676056337e-05, "loss": 0.925, "step": 10482 }, { "epoch": 11.75, "learning_rate": 1.6650704225352113e-05, "loss": 0.904, "step": 10483 }, { "epoch": 11.75, "learning_rate": 1.664929577464789e-05, "loss": 1.0283, "step": 10484 }, { "epoch": 11.75, "learning_rate": 1.6647887323943663e-05, "loss": 1.057, "step": 10485 }, { "epoch": 11.76, "learning_rate": 1.664647887323944e-05, "loss": 0.9346, "step": 10486 }, { "epoch": 11.76, "learning_rate": 1.6645070422535212e-05, "loss": 0.8625, "step": 10487 }, { "epoch": 11.76, "learning_rate": 1.6643661971830988e-05, "loss": 0.9057, "step": 10488 }, { "epoch": 11.76, "learning_rate": 1.664225352112676e-05, "loss": 0.8453, "step": 10489 }, { "epoch": 11.76, "learning_rate": 1.6640845070422534e-05, "loss": 0.8661, "step": 10490 }, { "epoch": 11.76, "learning_rate": 1.663943661971831e-05, "loss": 0.9052, "step": 10491 }, { "epoch": 11.76, "learning_rate": 1.6638028169014083e-05, "loss": 1.1538, "step": 10492 }, { "epoch": 11.76, "learning_rate": 1.663661971830986e-05, "loss": 0.8673, "step": 10493 }, { "epoch": 11.76, "learning_rate": 1.6635211267605633e-05, "loss": 1.0347, "step": 10494 }, { "epoch": 11.77, "learning_rate": 1.663380281690141e-05, "loss": 0.8773, "step": 10495 }, { "epoch": 11.77, "learning_rate": 1.6632394366197182e-05, "loss": 0.9083, "step": 10496 }, { "epoch": 11.77, "learning_rate": 1.6630985915492958e-05, "loss": 1.2145, "step": 10497 }, { "epoch": 11.77, "learning_rate": 1.6629577464788735e-05, "loss": 0.8797, "step": 10498 }, { "epoch": 11.77, "learning_rate": 1.6628169014084508e-05, "loss": 0.8236, "step": 10499 }, { "epoch": 11.77, "learning_rate": 1.6626760563380284e-05, "loss": 0.7492, "step": 10500 }, { "epoch": 11.77, "eval_loss": 1.1715891361236572, "eval_runtime": 828.6483, "eval_samples_per_second": 3.188, "eval_steps_per_second": 0.399, "eval_wer": 0.19486314954383183, "step": 10500 }, { "epoch": 11.77, "learning_rate": 1.6625352112676057e-05, "loss": 0.7943, "step": 10501 }, { "epoch": 11.77, "learning_rate": 1.6623943661971833e-05, "loss": 0.9481, "step": 10502 }, { "epoch": 11.77, "learning_rate": 1.6622535211267606e-05, "loss": 1.0707, "step": 10503 }, { "epoch": 11.78, "learning_rate": 1.6621126760563382e-05, "loss": 0.8931, "step": 10504 }, { "epoch": 11.78, "learning_rate": 1.6619718309859155e-05, "loss": 0.7832, "step": 10505 }, { "epoch": 11.78, "learning_rate": 1.661830985915493e-05, "loss": 0.7574, "step": 10506 }, { "epoch": 11.78, "learning_rate": 1.6616901408450705e-05, "loss": 0.8392, "step": 10507 }, { "epoch": 11.78, "learning_rate": 1.6615492957746478e-05, "loss": 0.9412, "step": 10508 }, { "epoch": 11.78, "learning_rate": 1.6614084507042254e-05, "loss": 0.6211, "step": 10509 }, { "epoch": 11.78, "learning_rate": 1.6612676056338027e-05, "loss": 0.9239, "step": 10510 }, { "epoch": 11.78, "learning_rate": 1.6611267605633803e-05, "loss": 0.919, "step": 10511 }, { "epoch": 11.78, "learning_rate": 1.6609859154929576e-05, "loss": 1.0271, "step": 10512 }, { "epoch": 11.79, "learning_rate": 1.6608450704225352e-05, "loss": 1.1687, "step": 10513 }, { "epoch": 11.79, "learning_rate": 1.660704225352113e-05, "loss": 1.0855, "step": 10514 }, { "epoch": 11.79, "learning_rate": 1.6605633802816902e-05, "loss": 1.1086, "step": 10515 }, { "epoch": 11.79, "learning_rate": 1.6604225352112678e-05, "loss": 1.1194, "step": 10516 }, { "epoch": 11.79, "learning_rate": 1.660281690140845e-05, "loss": 1.0838, "step": 10517 }, { "epoch": 11.79, "learning_rate": 1.6601408450704227e-05, "loss": 1.0368, "step": 10518 }, { "epoch": 11.79, "learning_rate": 1.6601408450704227e-05, "loss": 0.8596, "step": 10519 }, { "epoch": 11.79, "learning_rate": 1.66e-05, "loss": 1.1454, "step": 10520 }, { "epoch": 11.79, "learning_rate": 1.6598591549295777e-05, "loss": 0.9886, "step": 10521 }, { "epoch": 11.8, "learning_rate": 1.659718309859155e-05, "loss": 0.9164, "step": 10522 }, { "epoch": 11.8, "learning_rate": 1.6595774647887323e-05, "loss": 0.8281, "step": 10523 }, { "epoch": 11.8, "learning_rate": 1.65943661971831e-05, "loss": 1.0625, "step": 10524 }, { "epoch": 11.8, "learning_rate": 1.6592957746478872e-05, "loss": 0.9735, "step": 10525 }, { "epoch": 11.8, "learning_rate": 1.6591549295774648e-05, "loss": 0.8383, "step": 10526 }, { "epoch": 11.8, "learning_rate": 1.659014084507042e-05, "loss": 0.9836, "step": 10527 }, { "epoch": 11.8, "learning_rate": 1.6588732394366197e-05, "loss": 0.8972, "step": 10528 }, { "epoch": 11.8, "learning_rate": 1.6587323943661974e-05, "loss": 0.9433, "step": 10529 }, { "epoch": 11.8, "learning_rate": 1.6585915492957747e-05, "loss": 0.9894, "step": 10530 }, { "epoch": 11.81, "learning_rate": 1.6584507042253523e-05, "loss": 0.8855, "step": 10531 }, { "epoch": 11.81, "learning_rate": 1.6583098591549296e-05, "loss": 1.1682, "step": 10532 }, { "epoch": 11.81, "learning_rate": 1.6581690140845072e-05, "loss": 0.856, "step": 10533 }, { "epoch": 11.81, "learning_rate": 1.6580281690140845e-05, "loss": 0.8781, "step": 10534 }, { "epoch": 11.81, "learning_rate": 1.657887323943662e-05, "loss": 0.7253, "step": 10535 }, { "epoch": 11.81, "learning_rate": 1.6577464788732395e-05, "loss": 1.0271, "step": 10536 }, { "epoch": 11.81, "learning_rate": 1.657605633802817e-05, "loss": 0.9639, "step": 10537 }, { "epoch": 11.81, "learning_rate": 1.6574647887323944e-05, "loss": 0.6813, "step": 10538 }, { "epoch": 11.82, "learning_rate": 1.6573239436619717e-05, "loss": 0.8405, "step": 10539 }, { "epoch": 11.82, "learning_rate": 1.6571830985915493e-05, "loss": 0.853, "step": 10540 }, { "epoch": 11.82, "learning_rate": 1.6570422535211266e-05, "loss": 0.8884, "step": 10541 }, { "epoch": 11.82, "learning_rate": 1.6569014084507042e-05, "loss": 0.9346, "step": 10542 }, { "epoch": 11.82, "learning_rate": 1.6567605633802815e-05, "loss": 0.8339, "step": 10543 }, { "epoch": 11.82, "learning_rate": 1.656619718309859e-05, "loss": 0.92, "step": 10544 }, { "epoch": 11.82, "learning_rate": 1.6564788732394368e-05, "loss": 0.8593, "step": 10545 }, { "epoch": 11.82, "learning_rate": 1.656338028169014e-05, "loss": 0.9662, "step": 10546 }, { "epoch": 11.82, "learning_rate": 1.6561971830985917e-05, "loss": 0.9138, "step": 10547 }, { "epoch": 11.83, "learning_rate": 1.656056338028169e-05, "loss": 1.0182, "step": 10548 }, { "epoch": 11.83, "learning_rate": 1.6559154929577466e-05, "loss": 0.9276, "step": 10549 }, { "epoch": 11.83, "learning_rate": 1.655774647887324e-05, "loss": 0.7815, "step": 10550 }, { "epoch": 11.83, "learning_rate": 1.6556338028169016e-05, "loss": 0.8806, "step": 10551 }, { "epoch": 11.83, "learning_rate": 1.655492957746479e-05, "loss": 0.9287, "step": 10552 }, { "epoch": 11.83, "learning_rate": 1.6553521126760565e-05, "loss": 0.7266, "step": 10553 }, { "epoch": 11.83, "learning_rate": 1.6552112676056338e-05, "loss": 0.848, "step": 10554 }, { "epoch": 11.83, "learning_rate": 1.655070422535211e-05, "loss": 0.9568, "step": 10555 }, { "epoch": 11.83, "learning_rate": 1.6549295774647887e-05, "loss": 0.9081, "step": 10556 }, { "epoch": 11.84, "learning_rate": 1.654788732394366e-05, "loss": 0.7242, "step": 10557 }, { "epoch": 11.84, "learning_rate": 1.6546478873239437e-05, "loss": 0.951, "step": 10558 }, { "epoch": 11.84, "learning_rate": 1.654507042253521e-05, "loss": 0.8672, "step": 10559 }, { "epoch": 11.84, "learning_rate": 1.654366197183099e-05, "loss": 1.0206, "step": 10560 }, { "epoch": 11.84, "learning_rate": 1.6542253521126762e-05, "loss": 0.7857, "step": 10561 }, { "epoch": 11.84, "learning_rate": 1.6540845070422535e-05, "loss": 0.8835, "step": 10562 }, { "epoch": 11.84, "learning_rate": 1.653943661971831e-05, "loss": 1.1233, "step": 10563 }, { "epoch": 11.84, "learning_rate": 1.6538028169014084e-05, "loss": 1.1569, "step": 10564 }, { "epoch": 11.84, "learning_rate": 1.653661971830986e-05, "loss": 1.1615, "step": 10565 }, { "epoch": 11.85, "learning_rate": 1.6535211267605634e-05, "loss": 1.0934, "step": 10566 }, { "epoch": 11.85, "learning_rate": 1.653380281690141e-05, "loss": 0.804, "step": 10567 }, { "epoch": 11.85, "learning_rate": 1.6532394366197183e-05, "loss": 0.9781, "step": 10568 }, { "epoch": 11.85, "learning_rate": 1.653098591549296e-05, "loss": 0.9074, "step": 10569 }, { "epoch": 11.85, "learning_rate": 1.6529577464788732e-05, "loss": 0.8921, "step": 10570 }, { "epoch": 11.85, "learning_rate": 1.6528169014084505e-05, "loss": 0.9257, "step": 10571 }, { "epoch": 11.85, "learning_rate": 1.652676056338028e-05, "loss": 1.0616, "step": 10572 }, { "epoch": 11.85, "learning_rate": 1.6525352112676054e-05, "loss": 0.8747, "step": 10573 }, { "epoch": 11.85, "learning_rate": 1.652394366197183e-05, "loss": 0.9755, "step": 10574 }, { "epoch": 11.86, "learning_rate": 1.6522535211267607e-05, "loss": 0.8195, "step": 10575 }, { "epoch": 11.86, "learning_rate": 1.6521126760563383e-05, "loss": 0.9772, "step": 10576 }, { "epoch": 11.86, "learning_rate": 1.6519718309859156e-05, "loss": 0.8743, "step": 10577 }, { "epoch": 11.86, "learning_rate": 1.651830985915493e-05, "loss": 0.6782, "step": 10578 }, { "epoch": 11.86, "learning_rate": 1.6516901408450706e-05, "loss": 0.8312, "step": 10579 }, { "epoch": 11.86, "learning_rate": 1.651549295774648e-05, "loss": 0.8722, "step": 10580 }, { "epoch": 11.86, "learning_rate": 1.6514084507042255e-05, "loss": 0.8989, "step": 10581 }, { "epoch": 11.86, "learning_rate": 1.6512676056338028e-05, "loss": 0.9163, "step": 10582 }, { "epoch": 11.86, "learning_rate": 1.6511267605633804e-05, "loss": 0.697, "step": 10583 }, { "epoch": 11.87, "learning_rate": 1.6509859154929577e-05, "loss": 0.9599, "step": 10584 }, { "epoch": 11.87, "learning_rate": 1.6508450704225353e-05, "loss": 0.9103, "step": 10585 }, { "epoch": 11.87, "learning_rate": 1.6507042253521126e-05, "loss": 1.0127, "step": 10586 }, { "epoch": 11.87, "learning_rate": 1.65056338028169e-05, "loss": 0.8973, "step": 10587 }, { "epoch": 11.87, "learning_rate": 1.6504225352112676e-05, "loss": 1.0086, "step": 10588 }, { "epoch": 11.87, "learning_rate": 1.650281690140845e-05, "loss": 0.9831, "step": 10589 }, { "epoch": 11.87, "learning_rate": 1.650140845070423e-05, "loss": 0.9963, "step": 10590 }, { "epoch": 11.87, "learning_rate": 1.65e-05, "loss": 0.7797, "step": 10591 }, { "epoch": 11.87, "learning_rate": 1.6498591549295778e-05, "loss": 0.9167, "step": 10592 }, { "epoch": 11.88, "learning_rate": 1.649718309859155e-05, "loss": 0.8234, "step": 10593 }, { "epoch": 11.88, "learning_rate": 1.6495774647887323e-05, "loss": 0.8588, "step": 10594 }, { "epoch": 11.88, "learning_rate": 1.64943661971831e-05, "loss": 1.0341, "step": 10595 }, { "epoch": 11.88, "learning_rate": 1.6492957746478873e-05, "loss": 0.9522, "step": 10596 }, { "epoch": 11.88, "learning_rate": 1.649154929577465e-05, "loss": 0.871, "step": 10597 }, { "epoch": 11.88, "learning_rate": 1.6490140845070422e-05, "loss": 0.7769, "step": 10598 }, { "epoch": 11.88, "learning_rate": 1.64887323943662e-05, "loss": 1.1158, "step": 10599 }, { "epoch": 11.88, "learning_rate": 1.648732394366197e-05, "loss": 0.8647, "step": 10600 }, { "epoch": 11.88, "learning_rate": 1.6485915492957748e-05, "loss": 0.8389, "step": 10601 }, { "epoch": 11.89, "learning_rate": 1.648450704225352e-05, "loss": 0.7466, "step": 10602 }, { "epoch": 11.89, "learning_rate": 1.6483098591549294e-05, "loss": 0.9202, "step": 10603 }, { "epoch": 11.89, "learning_rate": 1.648169014084507e-05, "loss": 1.0681, "step": 10604 }, { "epoch": 11.89, "learning_rate": 1.6480281690140846e-05, "loss": 0.9686, "step": 10605 }, { "epoch": 11.89, "learning_rate": 1.6478873239436623e-05, "loss": 0.9811, "step": 10606 }, { "epoch": 11.89, "learning_rate": 1.6477464788732395e-05, "loss": 0.772, "step": 10607 }, { "epoch": 11.89, "learning_rate": 1.6476056338028172e-05, "loss": 0.7872, "step": 10608 }, { "epoch": 11.89, "learning_rate": 1.6474647887323945e-05, "loss": 0.7525, "step": 10609 }, { "epoch": 11.89, "learning_rate": 1.6473239436619718e-05, "loss": 0.5766, "step": 10610 }, { "epoch": 11.9, "learning_rate": 1.6471830985915494e-05, "loss": 0.7408, "step": 10611 }, { "epoch": 11.9, "learning_rate": 1.6470422535211267e-05, "loss": 1.1759, "step": 10612 }, { "epoch": 11.9, "learning_rate": 1.6469014084507043e-05, "loss": 1.1177, "step": 10613 }, { "epoch": 11.9, "learning_rate": 1.6467605633802816e-05, "loss": 1.1451, "step": 10614 }, { "epoch": 11.9, "learning_rate": 1.6466197183098593e-05, "loss": 0.9666, "step": 10615 }, { "epoch": 11.9, "learning_rate": 1.6464788732394366e-05, "loss": 1.2597, "step": 10616 }, { "epoch": 11.9, "learning_rate": 1.6463380281690142e-05, "loss": 1.019, "step": 10617 }, { "epoch": 11.9, "learning_rate": 1.6461971830985915e-05, "loss": 1.0407, "step": 10618 }, { "epoch": 11.9, "learning_rate": 1.6460563380281688e-05, "loss": 0.8832, "step": 10619 }, { "epoch": 11.91, "learning_rate": 1.6459154929577467e-05, "loss": 0.9837, "step": 10620 }, { "epoch": 11.91, "learning_rate": 1.645774647887324e-05, "loss": 1.0228, "step": 10621 }, { "epoch": 11.91, "learning_rate": 1.6456338028169017e-05, "loss": 1.0703, "step": 10622 }, { "epoch": 11.91, "learning_rate": 1.645492957746479e-05, "loss": 0.9569, "step": 10623 }, { "epoch": 11.91, "learning_rate": 1.6453521126760566e-05, "loss": 0.9908, "step": 10624 }, { "epoch": 11.91, "learning_rate": 1.645211267605634e-05, "loss": 0.9394, "step": 10625 }, { "epoch": 11.91, "learning_rate": 1.6450704225352112e-05, "loss": 0.9332, "step": 10626 }, { "epoch": 11.91, "learning_rate": 1.6449295774647888e-05, "loss": 1.0079, "step": 10627 }, { "epoch": 11.91, "learning_rate": 1.644788732394366e-05, "loss": 0.9266, "step": 10628 }, { "epoch": 11.92, "learning_rate": 1.6446478873239437e-05, "loss": 0.798, "step": 10629 }, { "epoch": 11.92, "learning_rate": 1.644507042253521e-05, "loss": 0.9037, "step": 10630 }, { "epoch": 11.92, "learning_rate": 1.6443661971830987e-05, "loss": 0.7467, "step": 10631 }, { "epoch": 11.92, "learning_rate": 1.644225352112676e-05, "loss": 0.9181, "step": 10632 }, { "epoch": 11.92, "learning_rate": 1.6440845070422536e-05, "loss": 0.7283, "step": 10633 }, { "epoch": 11.92, "learning_rate": 1.643943661971831e-05, "loss": 0.9106, "step": 10634 }, { "epoch": 11.92, "learning_rate": 1.6438028169014085e-05, "loss": 0.836, "step": 10635 }, { "epoch": 11.92, "learning_rate": 1.643661971830986e-05, "loss": 0.9776, "step": 10636 }, { "epoch": 11.92, "learning_rate": 1.6435211267605635e-05, "loss": 0.895, "step": 10637 }, { "epoch": 11.93, "learning_rate": 1.643380281690141e-05, "loss": 0.7786, "step": 10638 }, { "epoch": 11.93, "learning_rate": 1.6432394366197184e-05, "loss": 0.7671, "step": 10639 }, { "epoch": 11.93, "learning_rate": 1.643098591549296e-05, "loss": 0.7705, "step": 10640 }, { "epoch": 11.93, "learning_rate": 1.6429577464788733e-05, "loss": 0.8557, "step": 10641 }, { "epoch": 11.93, "learning_rate": 1.6428169014084506e-05, "loss": 0.8368, "step": 10642 }, { "epoch": 11.93, "learning_rate": 1.6426760563380282e-05, "loss": 0.9452, "step": 10643 }, { "epoch": 11.93, "learning_rate": 1.6425352112676055e-05, "loss": 0.8171, "step": 10644 }, { "epoch": 11.93, "learning_rate": 1.642394366197183e-05, "loss": 0.6092, "step": 10645 }, { "epoch": 11.93, "learning_rate": 1.6422535211267605e-05, "loss": 0.9264, "step": 10646 }, { "epoch": 11.94, "learning_rate": 1.642112676056338e-05, "loss": 0.7435, "step": 10647 }, { "epoch": 11.94, "learning_rate": 1.6419718309859154e-05, "loss": 0.8467, "step": 10648 }, { "epoch": 11.94, "learning_rate": 1.641830985915493e-05, "loss": 0.9299, "step": 10649 }, { "epoch": 11.94, "learning_rate": 1.6416901408450707e-05, "loss": 0.821, "step": 10650 }, { "epoch": 11.94, "learning_rate": 1.641549295774648e-05, "loss": 0.7827, "step": 10651 }, { "epoch": 11.94, "learning_rate": 1.6414084507042256e-05, "loss": 0.7829, "step": 10652 }, { "epoch": 11.94, "learning_rate": 1.641267605633803e-05, "loss": 0.6968, "step": 10653 }, { "epoch": 11.94, "learning_rate": 1.6411267605633805e-05, "loss": 0.7736, "step": 10654 }, { "epoch": 11.95, "learning_rate": 1.6409859154929578e-05, "loss": 0.8324, "step": 10655 }, { "epoch": 11.95, "learning_rate": 1.6408450704225354e-05, "loss": 0.8927, "step": 10656 }, { "epoch": 11.95, "learning_rate": 1.6407042253521127e-05, "loss": 0.9714, "step": 10657 }, { "epoch": 11.95, "learning_rate": 1.64056338028169e-05, "loss": 0.8823, "step": 10658 }, { "epoch": 11.95, "learning_rate": 1.6404225352112677e-05, "loss": 0.7999, "step": 10659 }, { "epoch": 11.95, "learning_rate": 1.640281690140845e-05, "loss": 0.7901, "step": 10660 }, { "epoch": 11.95, "learning_rate": 1.6401408450704226e-05, "loss": 0.9246, "step": 10661 }, { "epoch": 11.95, "learning_rate": 1.64e-05, "loss": 0.448, "step": 10662 }, { "epoch": 11.95, "learning_rate": 1.6398591549295775e-05, "loss": 1.2304, "step": 10663 }, { "epoch": 11.96, "learning_rate": 1.6397183098591548e-05, "loss": 0.9386, "step": 10664 }, { "epoch": 11.96, "learning_rate": 1.639577464788732e-05, "loss": 1.0596, "step": 10665 }, { "epoch": 11.96, "learning_rate": 1.63943661971831e-05, "loss": 1.052, "step": 10666 }, { "epoch": 11.96, "learning_rate": 1.6392957746478874e-05, "loss": 0.9248, "step": 10667 }, { "epoch": 11.96, "learning_rate": 1.639154929577465e-05, "loss": 0.9562, "step": 10668 }, { "epoch": 11.96, "learning_rate": 1.6390140845070423e-05, "loss": 1.1072, "step": 10669 }, { "epoch": 11.96, "learning_rate": 1.63887323943662e-05, "loss": 0.7604, "step": 10670 }, { "epoch": 11.96, "learning_rate": 1.6387323943661972e-05, "loss": 0.9482, "step": 10671 }, { "epoch": 11.96, "learning_rate": 1.638591549295775e-05, "loss": 0.9029, "step": 10672 }, { "epoch": 11.97, "learning_rate": 1.638450704225352e-05, "loss": 0.9833, "step": 10673 }, { "epoch": 11.97, "learning_rate": 1.6383098591549294e-05, "loss": 0.853, "step": 10674 }, { "epoch": 11.97, "learning_rate": 1.638169014084507e-05, "loss": 0.9212, "step": 10675 }, { "epoch": 11.97, "learning_rate": 1.6380281690140844e-05, "loss": 0.9943, "step": 10676 }, { "epoch": 11.97, "learning_rate": 1.637887323943662e-05, "loss": 1.0192, "step": 10677 }, { "epoch": 11.97, "learning_rate": 1.6377464788732393e-05, "loss": 0.8505, "step": 10678 }, { "epoch": 11.97, "learning_rate": 1.637605633802817e-05, "loss": 0.9136, "step": 10679 }, { "epoch": 11.97, "learning_rate": 1.6374647887323942e-05, "loss": 0.8385, "step": 10680 }, { "epoch": 11.97, "learning_rate": 1.637323943661972e-05, "loss": 0.7653, "step": 10681 }, { "epoch": 11.98, "learning_rate": 1.6371830985915495e-05, "loss": 0.7571, "step": 10682 }, { "epoch": 11.98, "learning_rate": 1.6370422535211268e-05, "loss": 0.8954, "step": 10683 }, { "epoch": 11.98, "learning_rate": 1.6369014084507044e-05, "loss": 1.0, "step": 10684 }, { "epoch": 11.98, "learning_rate": 1.6367605633802817e-05, "loss": 0.6857, "step": 10685 }, { "epoch": 11.98, "learning_rate": 1.6366197183098594e-05, "loss": 0.8907, "step": 10686 }, { "epoch": 11.98, "learning_rate": 1.6364788732394366e-05, "loss": 0.8116, "step": 10687 }, { "epoch": 11.98, "learning_rate": 1.6363380281690143e-05, "loss": 0.7299, "step": 10688 }, { "epoch": 11.98, "learning_rate": 1.6361971830985916e-05, "loss": 0.8548, "step": 10689 }, { "epoch": 11.98, "learning_rate": 1.636056338028169e-05, "loss": 0.7229, "step": 10690 }, { "epoch": 11.99, "learning_rate": 1.6359154929577465e-05, "loss": 0.8531, "step": 10691 }, { "epoch": 11.99, "learning_rate": 1.6357746478873238e-05, "loss": 1.0131, "step": 10692 }, { "epoch": 11.99, "learning_rate": 1.6356338028169014e-05, "loss": 0.9091, "step": 10693 }, { "epoch": 11.99, "learning_rate": 1.6354929577464787e-05, "loss": 0.9836, "step": 10694 }, { "epoch": 11.99, "learning_rate": 1.6353521126760564e-05, "loss": 0.8251, "step": 10695 }, { "epoch": 11.99, "learning_rate": 1.635211267605634e-05, "loss": 0.8294, "step": 10696 }, { "epoch": 11.99, "learning_rate": 1.6350704225352113e-05, "loss": 0.7826, "step": 10697 }, { "epoch": 11.99, "learning_rate": 1.634929577464789e-05, "loss": 0.8322, "step": 10698 }, { "epoch": 11.99, "learning_rate": 1.6347887323943662e-05, "loss": 0.8345, "step": 10699 }, { "epoch": 12.0, "learning_rate": 1.634647887323944e-05, "loss": 0.6741, "step": 10700 }, { "epoch": 12.0, "learning_rate": 1.634507042253521e-05, "loss": 0.9257, "step": 10701 }, { "epoch": 12.0, "learning_rate": 1.6343661971830988e-05, "loss": 0.7136, "step": 10702 }, { "epoch": 12.0, "learning_rate": 1.634225352112676e-05, "loss": 0.886, "step": 10703 }, { "epoch": 12.0, "learning_rate": 1.6340845070422537e-05, "loss": 0.6715, "step": 10704 }, { "epoch": 12.0, "learning_rate": 1.633943661971831e-05, "loss": 1.2404, "step": 10705 }, { "epoch": 12.0, "learning_rate": 1.6338028169014083e-05, "loss": 1.1353, "step": 10706 }, { "epoch": 12.0, "learning_rate": 1.633661971830986e-05, "loss": 1.1026, "step": 10707 }, { "epoch": 12.0, "learning_rate": 1.6335211267605632e-05, "loss": 0.9345, "step": 10708 }, { "epoch": 12.01, "learning_rate": 1.633380281690141e-05, "loss": 1.0296, "step": 10709 }, { "epoch": 12.01, "learning_rate": 1.633239436619718e-05, "loss": 0.9095, "step": 10710 }, { "epoch": 12.01, "learning_rate": 1.633098591549296e-05, "loss": 0.8411, "step": 10711 }, { "epoch": 12.01, "learning_rate": 1.6329577464788734e-05, "loss": 0.8299, "step": 10712 }, { "epoch": 12.01, "learning_rate": 1.6328169014084507e-05, "loss": 0.7708, "step": 10713 }, { "epoch": 12.01, "learning_rate": 1.6326760563380283e-05, "loss": 0.9364, "step": 10714 }, { "epoch": 12.01, "learning_rate": 1.6325352112676056e-05, "loss": 0.6903, "step": 10715 }, { "epoch": 12.01, "learning_rate": 1.6323943661971833e-05, "loss": 0.9334, "step": 10716 }, { "epoch": 12.01, "learning_rate": 1.6322535211267606e-05, "loss": 0.8961, "step": 10717 }, { "epoch": 12.02, "learning_rate": 1.6321126760563382e-05, "loss": 0.8846, "step": 10718 }, { "epoch": 12.02, "learning_rate": 1.6319718309859155e-05, "loss": 0.8836, "step": 10719 }, { "epoch": 12.02, "learning_rate": 1.631830985915493e-05, "loss": 0.91, "step": 10720 }, { "epoch": 12.02, "learning_rate": 1.6316901408450704e-05, "loss": 1.0822, "step": 10721 }, { "epoch": 12.02, "learning_rate": 1.6315492957746477e-05, "loss": 0.7929, "step": 10722 }, { "epoch": 12.02, "learning_rate": 1.6314084507042253e-05, "loss": 0.9132, "step": 10723 }, { "epoch": 12.02, "learning_rate": 1.6312676056338026e-05, "loss": 0.8799, "step": 10724 }, { "epoch": 12.02, "learning_rate": 1.6311267605633803e-05, "loss": 0.8781, "step": 10725 }, { "epoch": 12.02, "learning_rate": 1.630985915492958e-05, "loss": 0.864, "step": 10726 }, { "epoch": 12.03, "learning_rate": 1.6308450704225355e-05, "loss": 0.8662, "step": 10727 }, { "epoch": 12.03, "learning_rate": 1.6307042253521128e-05, "loss": 0.9514, "step": 10728 }, { "epoch": 12.03, "learning_rate": 1.63056338028169e-05, "loss": 0.7775, "step": 10729 }, { "epoch": 12.03, "learning_rate": 1.6304225352112678e-05, "loss": 0.7664, "step": 10730 }, { "epoch": 12.03, "learning_rate": 1.630281690140845e-05, "loss": 0.8783, "step": 10731 }, { "epoch": 12.03, "learning_rate": 1.6301408450704227e-05, "loss": 0.7386, "step": 10732 }, { "epoch": 12.03, "learning_rate": 1.63e-05, "loss": 0.7374, "step": 10733 }, { "epoch": 12.03, "learning_rate": 1.6298591549295776e-05, "loss": 0.7372, "step": 10734 }, { "epoch": 12.03, "learning_rate": 1.629718309859155e-05, "loss": 0.8929, "step": 10735 }, { "epoch": 12.04, "learning_rate": 1.6295774647887325e-05, "loss": 0.9199, "step": 10736 }, { "epoch": 12.04, "learning_rate": 1.62943661971831e-05, "loss": 0.8755, "step": 10737 }, { "epoch": 12.04, "learning_rate": 1.629295774647887e-05, "loss": 0.8731, "step": 10738 }, { "epoch": 12.04, "learning_rate": 1.6291549295774648e-05, "loss": 0.8509, "step": 10739 }, { "epoch": 12.04, "learning_rate": 1.629014084507042e-05, "loss": 0.7486, "step": 10740 }, { "epoch": 12.04, "learning_rate": 1.62887323943662e-05, "loss": 0.763, "step": 10741 }, { "epoch": 12.04, "learning_rate": 1.6287323943661973e-05, "loss": 0.9359, "step": 10742 }, { "epoch": 12.04, "learning_rate": 1.628591549295775e-05, "loss": 0.8207, "step": 10743 }, { "epoch": 12.04, "learning_rate": 1.6284507042253522e-05, "loss": 0.7447, "step": 10744 }, { "epoch": 12.05, "learning_rate": 1.6283098591549295e-05, "loss": 0.7697, "step": 10745 }, { "epoch": 12.05, "learning_rate": 1.6281690140845072e-05, "loss": 0.8155, "step": 10746 }, { "epoch": 12.05, "learning_rate": 1.6280281690140845e-05, "loss": 1.1605, "step": 10747 }, { "epoch": 12.05, "learning_rate": 1.627887323943662e-05, "loss": 0.8552, "step": 10748 }, { "epoch": 12.05, "learning_rate": 1.6277464788732394e-05, "loss": 0.7976, "step": 10749 }, { "epoch": 12.05, "learning_rate": 1.627605633802817e-05, "loss": 0.7584, "step": 10750 }, { "epoch": 12.05, "learning_rate": 1.6274647887323943e-05, "loss": 0.5572, "step": 10751 }, { "epoch": 12.05, "learning_rate": 1.627323943661972e-05, "loss": 0.7357, "step": 10752 }, { "epoch": 12.05, "learning_rate": 1.6271830985915493e-05, "loss": 0.8453, "step": 10753 }, { "epoch": 12.06, "learning_rate": 1.6270422535211265e-05, "loss": 0.7175, "step": 10754 }, { "epoch": 12.06, "learning_rate": 1.6269014084507042e-05, "loss": 1.2029, "step": 10755 }, { "epoch": 12.06, "learning_rate": 1.6267605633802818e-05, "loss": 0.9231, "step": 10756 }, { "epoch": 12.06, "learning_rate": 1.6266197183098594e-05, "loss": 0.9699, "step": 10757 }, { "epoch": 12.06, "learning_rate": 1.6264788732394367e-05, "loss": 0.9655, "step": 10758 }, { "epoch": 12.06, "learning_rate": 1.6263380281690144e-05, "loss": 0.9079, "step": 10759 }, { "epoch": 12.06, "learning_rate": 1.6261971830985917e-05, "loss": 0.8791, "step": 10760 }, { "epoch": 12.06, "learning_rate": 1.626056338028169e-05, "loss": 0.8211, "step": 10761 }, { "epoch": 12.07, "learning_rate": 1.6259154929577466e-05, "loss": 0.8493, "step": 10762 }, { "epoch": 12.07, "learning_rate": 1.625774647887324e-05, "loss": 0.8602, "step": 10763 }, { "epoch": 12.07, "learning_rate": 1.6256338028169015e-05, "loss": 0.8302, "step": 10764 }, { "epoch": 12.07, "learning_rate": 1.6254929577464788e-05, "loss": 0.8186, "step": 10765 }, { "epoch": 12.07, "learning_rate": 1.6253521126760565e-05, "loss": 0.9111, "step": 10766 }, { "epoch": 12.07, "learning_rate": 1.6252112676056337e-05, "loss": 0.8973, "step": 10767 }, { "epoch": 12.07, "learning_rate": 1.625070422535211e-05, "loss": 0.8587, "step": 10768 }, { "epoch": 12.07, "learning_rate": 1.6249295774647887e-05, "loss": 0.794, "step": 10769 }, { "epoch": 12.07, "learning_rate": 1.624788732394366e-05, "loss": 0.8041, "step": 10770 }, { "epoch": 12.08, "learning_rate": 1.6246478873239436e-05, "loss": 0.9156, "step": 10771 }, { "epoch": 12.08, "learning_rate": 1.6245070422535212e-05, "loss": 0.7098, "step": 10772 }, { "epoch": 12.08, "learning_rate": 1.624366197183099e-05, "loss": 0.8944, "step": 10773 }, { "epoch": 12.08, "learning_rate": 1.624225352112676e-05, "loss": 1.0413, "step": 10774 }, { "epoch": 12.08, "learning_rate": 1.6240845070422538e-05, "loss": 0.8795, "step": 10775 }, { "epoch": 12.08, "learning_rate": 1.623943661971831e-05, "loss": 0.6814, "step": 10776 }, { "epoch": 12.08, "learning_rate": 1.6238028169014084e-05, "loss": 0.6598, "step": 10777 }, { "epoch": 12.08, "learning_rate": 1.623661971830986e-05, "loss": 0.6375, "step": 10778 }, { "epoch": 12.08, "learning_rate": 1.6235211267605633e-05, "loss": 0.8238, "step": 10779 }, { "epoch": 12.09, "learning_rate": 1.623380281690141e-05, "loss": 0.9049, "step": 10780 }, { "epoch": 12.09, "learning_rate": 1.6232394366197182e-05, "loss": 0.8887, "step": 10781 }, { "epoch": 12.09, "learning_rate": 1.623098591549296e-05, "loss": 0.7843, "step": 10782 }, { "epoch": 12.09, "learning_rate": 1.622957746478873e-05, "loss": 0.8525, "step": 10783 }, { "epoch": 12.09, "learning_rate": 1.6228169014084505e-05, "loss": 0.7848, "step": 10784 }, { "epoch": 12.09, "learning_rate": 1.622676056338028e-05, "loss": 0.9463, "step": 10785 }, { "epoch": 12.09, "learning_rate": 1.6225352112676054e-05, "loss": 0.693, "step": 10786 }, { "epoch": 12.09, "learning_rate": 1.6223943661971834e-05, "loss": 0.7796, "step": 10787 }, { "epoch": 12.09, "learning_rate": 1.6222535211267607e-05, "loss": 0.7243, "step": 10788 }, { "epoch": 12.1, "learning_rate": 1.6221126760563383e-05, "loss": 0.8777, "step": 10789 }, { "epoch": 12.1, "learning_rate": 1.6219718309859156e-05, "loss": 0.7756, "step": 10790 }, { "epoch": 12.1, "learning_rate": 1.6218309859154932e-05, "loss": 0.7215, "step": 10791 }, { "epoch": 12.1, "learning_rate": 1.6216901408450705e-05, "loss": 0.752, "step": 10792 }, { "epoch": 12.1, "learning_rate": 1.6215492957746478e-05, "loss": 0.7704, "step": 10793 }, { "epoch": 12.1, "learning_rate": 1.6214084507042254e-05, "loss": 0.7593, "step": 10794 }, { "epoch": 12.1, "learning_rate": 1.6212676056338027e-05, "loss": 0.8378, "step": 10795 }, { "epoch": 12.1, "learning_rate": 1.6211267605633804e-05, "loss": 0.806, "step": 10796 }, { "epoch": 12.1, "learning_rate": 1.6209859154929577e-05, "loss": 0.8066, "step": 10797 }, { "epoch": 12.11, "learning_rate": 1.6208450704225353e-05, "loss": 1.0275, "step": 10798 }, { "epoch": 12.11, "learning_rate": 1.6207042253521126e-05, "loss": 0.6933, "step": 10799 }, { "epoch": 12.11, "learning_rate": 1.62056338028169e-05, "loss": 0.79, "step": 10800 }, { "epoch": 12.11, "learning_rate": 1.6204225352112675e-05, "loss": 0.6954, "step": 10801 }, { "epoch": 12.11, "learning_rate": 1.620281690140845e-05, "loss": 0.9272, "step": 10802 }, { "epoch": 12.11, "learning_rate": 1.6201408450704228e-05, "loss": 0.5991, "step": 10803 }, { "epoch": 12.11, "learning_rate": 1.62e-05, "loss": 0.8506, "step": 10804 }, { "epoch": 12.11, "learning_rate": 1.6198591549295777e-05, "loss": 1.0893, "step": 10805 }, { "epoch": 12.11, "learning_rate": 1.619718309859155e-05, "loss": 0.9641, "step": 10806 }, { "epoch": 12.12, "learning_rate": 1.6195774647887326e-05, "loss": 0.7952, "step": 10807 }, { "epoch": 12.12, "learning_rate": 1.61943661971831e-05, "loss": 0.8606, "step": 10808 }, { "epoch": 12.12, "learning_rate": 1.6192957746478872e-05, "loss": 0.9494, "step": 10809 }, { "epoch": 12.12, "learning_rate": 1.619154929577465e-05, "loss": 0.9434, "step": 10810 }, { "epoch": 12.12, "learning_rate": 1.619014084507042e-05, "loss": 0.8188, "step": 10811 }, { "epoch": 12.12, "learning_rate": 1.6188732394366198e-05, "loss": 0.8831, "step": 10812 }, { "epoch": 12.12, "learning_rate": 1.618732394366197e-05, "loss": 0.8294, "step": 10813 }, { "epoch": 12.12, "learning_rate": 1.6185915492957747e-05, "loss": 0.9623, "step": 10814 }, { "epoch": 12.12, "learning_rate": 1.618450704225352e-05, "loss": 0.8277, "step": 10815 }, { "epoch": 12.13, "learning_rate": 1.6183098591549293e-05, "loss": 0.9278, "step": 10816 }, { "epoch": 12.13, "learning_rate": 1.6181690140845073e-05, "loss": 0.7656, "step": 10817 }, { "epoch": 12.13, "learning_rate": 1.6180281690140846e-05, "loss": 0.912, "step": 10818 }, { "epoch": 12.13, "learning_rate": 1.6178873239436622e-05, "loss": 0.8687, "step": 10819 }, { "epoch": 12.13, "learning_rate": 1.6177464788732395e-05, "loss": 0.9268, "step": 10820 }, { "epoch": 12.13, "learning_rate": 1.617605633802817e-05, "loss": 0.8151, "step": 10821 }, { "epoch": 12.13, "learning_rate": 1.6174647887323944e-05, "loss": 0.7124, "step": 10822 }, { "epoch": 12.13, "learning_rate": 1.617323943661972e-05, "loss": 0.7071, "step": 10823 }, { "epoch": 12.13, "learning_rate": 1.6171830985915493e-05, "loss": 0.7582, "step": 10824 }, { "epoch": 12.14, "learning_rate": 1.6170422535211266e-05, "loss": 0.9056, "step": 10825 }, { "epoch": 12.14, "learning_rate": 1.6169014084507043e-05, "loss": 0.7956, "step": 10826 }, { "epoch": 12.14, "learning_rate": 1.6167605633802816e-05, "loss": 0.8969, "step": 10827 }, { "epoch": 12.14, "learning_rate": 1.6166197183098592e-05, "loss": 0.7526, "step": 10828 }, { "epoch": 12.14, "learning_rate": 1.6164788732394365e-05, "loss": 0.7236, "step": 10829 }, { "epoch": 12.14, "learning_rate": 1.616338028169014e-05, "loss": 0.8357, "step": 10830 }, { "epoch": 12.14, "learning_rate": 1.6161971830985914e-05, "loss": 0.9004, "step": 10831 }, { "epoch": 12.14, "learning_rate": 1.616056338028169e-05, "loss": 0.9093, "step": 10832 }, { "epoch": 12.14, "learning_rate": 1.6159154929577467e-05, "loss": 0.9485, "step": 10833 }, { "epoch": 12.15, "learning_rate": 1.615774647887324e-05, "loss": 0.7057, "step": 10834 }, { "epoch": 12.15, "learning_rate": 1.6156338028169016e-05, "loss": 0.9046, "step": 10835 }, { "epoch": 12.15, "learning_rate": 1.615492957746479e-05, "loss": 0.7917, "step": 10836 }, { "epoch": 12.15, "learning_rate": 1.6153521126760565e-05, "loss": 0.7337, "step": 10837 }, { "epoch": 12.15, "learning_rate": 1.615211267605634e-05, "loss": 0.7623, "step": 10838 }, { "epoch": 12.15, "learning_rate": 1.6150704225352115e-05, "loss": 0.9242, "step": 10839 }, { "epoch": 12.15, "learning_rate": 1.6149295774647888e-05, "loss": 0.7319, "step": 10840 }, { "epoch": 12.15, "learning_rate": 1.614788732394366e-05, "loss": 0.8303, "step": 10841 }, { "epoch": 12.15, "learning_rate": 1.6146478873239437e-05, "loss": 0.6954, "step": 10842 }, { "epoch": 12.16, "learning_rate": 1.614507042253521e-05, "loss": 0.9451, "step": 10843 }, { "epoch": 12.16, "learning_rate": 1.6143661971830986e-05, "loss": 0.9297, "step": 10844 }, { "epoch": 12.16, "learning_rate": 1.614225352112676e-05, "loss": 0.7291, "step": 10845 }, { "epoch": 12.16, "learning_rate": 1.6140845070422536e-05, "loss": 0.832, "step": 10846 }, { "epoch": 12.16, "learning_rate": 1.6139436619718312e-05, "loss": 0.6162, "step": 10847 }, { "epoch": 12.16, "learning_rate": 1.6138028169014085e-05, "loss": 0.74, "step": 10848 }, { "epoch": 12.16, "learning_rate": 1.613661971830986e-05, "loss": 0.6615, "step": 10849 }, { "epoch": 12.16, "learning_rate": 1.6135211267605634e-05, "loss": 0.7416, "step": 10850 }, { "epoch": 12.16, "learning_rate": 1.613380281690141e-05, "loss": 0.5761, "step": 10851 }, { "epoch": 12.17, "learning_rate": 1.6132394366197183e-05, "loss": 0.7346, "step": 10852 }, { "epoch": 12.17, "learning_rate": 1.613098591549296e-05, "loss": 0.834, "step": 10853 }, { "epoch": 12.17, "learning_rate": 1.6129577464788733e-05, "loss": 0.6746, "step": 10854 }, { "epoch": 12.17, "learning_rate": 1.612816901408451e-05, "loss": 1.05, "step": 10855 }, { "epoch": 12.17, "learning_rate": 1.6126760563380282e-05, "loss": 1.0657, "step": 10856 }, { "epoch": 12.17, "learning_rate": 1.6125352112676055e-05, "loss": 1.0015, "step": 10857 }, { "epoch": 12.17, "learning_rate": 1.612394366197183e-05, "loss": 0.9126, "step": 10858 }, { "epoch": 12.17, "learning_rate": 1.6122535211267604e-05, "loss": 0.8484, "step": 10859 }, { "epoch": 12.17, "learning_rate": 1.612112676056338e-05, "loss": 0.913, "step": 10860 }, { "epoch": 12.18, "learning_rate": 1.6119718309859153e-05, "loss": 0.879, "step": 10861 }, { "epoch": 12.18, "learning_rate": 1.6118309859154933e-05, "loss": 1.0267, "step": 10862 }, { "epoch": 12.18, "learning_rate": 1.6116901408450706e-05, "loss": 0.8423, "step": 10863 }, { "epoch": 12.18, "learning_rate": 1.611549295774648e-05, "loss": 0.8434, "step": 10864 }, { "epoch": 12.18, "learning_rate": 1.6114084507042255e-05, "loss": 0.8022, "step": 10865 }, { "epoch": 12.18, "learning_rate": 1.6112676056338028e-05, "loss": 0.8607, "step": 10866 }, { "epoch": 12.18, "learning_rate": 1.6111267605633805e-05, "loss": 0.8532, "step": 10867 }, { "epoch": 12.18, "learning_rate": 1.6109859154929578e-05, "loss": 0.7874, "step": 10868 }, { "epoch": 12.18, "learning_rate": 1.6108450704225354e-05, "loss": 0.8438, "step": 10869 }, { "epoch": 12.19, "learning_rate": 1.6107042253521127e-05, "loss": 0.7069, "step": 10870 }, { "epoch": 12.19, "learning_rate": 1.61056338028169e-05, "loss": 0.7537, "step": 10871 }, { "epoch": 12.19, "learning_rate": 1.6104225352112676e-05, "loss": 0.7792, "step": 10872 }, { "epoch": 12.19, "learning_rate": 1.610281690140845e-05, "loss": 0.8279, "step": 10873 }, { "epoch": 12.19, "learning_rate": 1.6101408450704225e-05, "loss": 0.869, "step": 10874 }, { "epoch": 12.19, "learning_rate": 1.61e-05, "loss": 0.7353, "step": 10875 }, { "epoch": 12.19, "learning_rate": 1.6098591549295775e-05, "loss": 0.9808, "step": 10876 }, { "epoch": 12.19, "learning_rate": 1.609718309859155e-05, "loss": 0.8405, "step": 10877 }, { "epoch": 12.2, "learning_rate": 1.6095774647887327e-05, "loss": 0.7963, "step": 10878 }, { "epoch": 12.2, "learning_rate": 1.60943661971831e-05, "loss": 0.8346, "step": 10879 }, { "epoch": 12.2, "learning_rate": 1.6092957746478873e-05, "loss": 0.7499, "step": 10880 }, { "epoch": 12.2, "learning_rate": 1.609154929577465e-05, "loss": 0.8749, "step": 10881 }, { "epoch": 12.2, "learning_rate": 1.6090140845070422e-05, "loss": 0.7291, "step": 10882 }, { "epoch": 12.2, "learning_rate": 1.60887323943662e-05, "loss": 0.8276, "step": 10883 }, { "epoch": 12.2, "learning_rate": 1.6087323943661972e-05, "loss": 0.8515, "step": 10884 }, { "epoch": 12.2, "learning_rate": 1.6085915492957748e-05, "loss": 0.7633, "step": 10885 }, { "epoch": 12.2, "learning_rate": 1.608450704225352e-05, "loss": 0.8952, "step": 10886 }, { "epoch": 12.21, "learning_rate": 1.6083098591549294e-05, "loss": 0.7724, "step": 10887 }, { "epoch": 12.21, "learning_rate": 1.608169014084507e-05, "loss": 0.7419, "step": 10888 }, { "epoch": 12.21, "learning_rate": 1.6080281690140843e-05, "loss": 0.7511, "step": 10889 }, { "epoch": 12.21, "learning_rate": 1.607887323943662e-05, "loss": 0.7672, "step": 10890 }, { "epoch": 12.21, "learning_rate": 1.6077464788732393e-05, "loss": 1.0101, "step": 10891 }, { "epoch": 12.21, "learning_rate": 1.607605633802817e-05, "loss": 0.7781, "step": 10892 }, { "epoch": 12.21, "learning_rate": 1.6074647887323945e-05, "loss": 0.8596, "step": 10893 }, { "epoch": 12.21, "learning_rate": 1.607323943661972e-05, "loss": 0.8876, "step": 10894 }, { "epoch": 12.21, "learning_rate": 1.6071830985915494e-05, "loss": 0.7663, "step": 10895 }, { "epoch": 12.22, "learning_rate": 1.6070422535211267e-05, "loss": 0.7413, "step": 10896 }, { "epoch": 12.22, "learning_rate": 1.6069014084507044e-05, "loss": 0.7752, "step": 10897 }, { "epoch": 12.22, "learning_rate": 1.6067605633802817e-05, "loss": 0.6504, "step": 10898 }, { "epoch": 12.22, "learning_rate": 1.6066197183098593e-05, "loss": 0.8357, "step": 10899 }, { "epoch": 12.22, "learning_rate": 1.6064788732394366e-05, "loss": 0.735, "step": 10900 }, { "epoch": 12.22, "learning_rate": 1.6063380281690142e-05, "loss": 0.5827, "step": 10901 }, { "epoch": 12.22, "learning_rate": 1.6061971830985915e-05, "loss": 0.7852, "step": 10902 }, { "epoch": 12.22, "learning_rate": 1.6060563380281688e-05, "loss": 0.6508, "step": 10903 }, { "epoch": 12.22, "learning_rate": 1.6059154929577465e-05, "loss": 0.7646, "step": 10904 }, { "epoch": 12.23, "learning_rate": 1.6057746478873237e-05, "loss": 1.1962, "step": 10905 }, { "epoch": 12.23, "learning_rate": 1.6056338028169014e-05, "loss": 0.9542, "step": 10906 }, { "epoch": 12.23, "learning_rate": 1.6054929577464787e-05, "loss": 1.0559, "step": 10907 }, { "epoch": 12.23, "learning_rate": 1.6053521126760566e-05, "loss": 0.8632, "step": 10908 }, { "epoch": 12.23, "learning_rate": 1.605211267605634e-05, "loss": 0.8382, "step": 10909 }, { "epoch": 12.23, "learning_rate": 1.6050704225352116e-05, "loss": 0.8709, "step": 10910 }, { "epoch": 12.23, "learning_rate": 1.604929577464789e-05, "loss": 1.0275, "step": 10911 }, { "epoch": 12.23, "learning_rate": 1.604788732394366e-05, "loss": 0.8179, "step": 10912 }, { "epoch": 12.23, "learning_rate": 1.6046478873239438e-05, "loss": 0.8622, "step": 10913 }, { "epoch": 12.24, "learning_rate": 1.604507042253521e-05, "loss": 0.9473, "step": 10914 }, { "epoch": 12.24, "learning_rate": 1.6043661971830987e-05, "loss": 0.8854, "step": 10915 }, { "epoch": 12.24, "learning_rate": 1.604225352112676e-05, "loss": 0.9561, "step": 10916 }, { "epoch": 12.24, "learning_rate": 1.6040845070422536e-05, "loss": 0.8253, "step": 10917 }, { "epoch": 12.24, "learning_rate": 1.603943661971831e-05, "loss": 0.7176, "step": 10918 }, { "epoch": 12.24, "learning_rate": 1.6038028169014082e-05, "loss": 0.7804, "step": 10919 }, { "epoch": 12.24, "learning_rate": 1.603661971830986e-05, "loss": 0.752, "step": 10920 }, { "epoch": 12.24, "learning_rate": 1.603521126760563e-05, "loss": 0.7179, "step": 10921 }, { "epoch": 12.24, "learning_rate": 1.6033802816901408e-05, "loss": 0.8999, "step": 10922 }, { "epoch": 12.25, "learning_rate": 1.6032394366197184e-05, "loss": 0.8742, "step": 10923 }, { "epoch": 12.25, "learning_rate": 1.603098591549296e-05, "loss": 0.8149, "step": 10924 }, { "epoch": 12.25, "learning_rate": 1.6029577464788734e-05, "loss": 0.8161, "step": 10925 }, { "epoch": 12.25, "learning_rate": 1.602816901408451e-05, "loss": 0.7853, "step": 10926 }, { "epoch": 12.25, "learning_rate": 1.6026760563380283e-05, "loss": 0.6962, "step": 10927 }, { "epoch": 12.25, "learning_rate": 1.6025352112676056e-05, "loss": 0.6747, "step": 10928 }, { "epoch": 12.25, "learning_rate": 1.6023943661971832e-05, "loss": 0.7285, "step": 10929 }, { "epoch": 12.25, "learning_rate": 1.6022535211267605e-05, "loss": 0.7325, "step": 10930 }, { "epoch": 12.25, "learning_rate": 1.602112676056338e-05, "loss": 0.7832, "step": 10931 }, { "epoch": 12.26, "learning_rate": 1.6019718309859154e-05, "loss": 0.7881, "step": 10932 }, { "epoch": 12.26, "learning_rate": 1.601830985915493e-05, "loss": 0.8903, "step": 10933 }, { "epoch": 12.26, "learning_rate": 1.6016901408450704e-05, "loss": 0.8207, "step": 10934 }, { "epoch": 12.26, "learning_rate": 1.6015492957746477e-05, "loss": 1.0251, "step": 10935 }, { "epoch": 12.26, "learning_rate": 1.6014084507042253e-05, "loss": 0.9239, "step": 10936 }, { "epoch": 12.26, "learning_rate": 1.6012676056338026e-05, "loss": 0.6851, "step": 10937 }, { "epoch": 12.26, "learning_rate": 1.6011267605633806e-05, "loss": 0.8445, "step": 10938 }, { "epoch": 12.26, "learning_rate": 1.600985915492958e-05, "loss": 0.65, "step": 10939 }, { "epoch": 12.26, "learning_rate": 1.6008450704225355e-05, "loss": 0.6918, "step": 10940 }, { "epoch": 12.27, "learning_rate": 1.6007042253521128e-05, "loss": 0.7727, "step": 10941 }, { "epoch": 12.27, "learning_rate": 1.6005633802816904e-05, "loss": 0.7833, "step": 10942 }, { "epoch": 12.27, "learning_rate": 1.6004225352112677e-05, "loss": 0.7438, "step": 10943 }, { "epoch": 12.27, "learning_rate": 1.600281690140845e-05, "loss": 0.8274, "step": 10944 }, { "epoch": 12.27, "learning_rate": 1.6001408450704226e-05, "loss": 0.805, "step": 10945 }, { "epoch": 12.27, "learning_rate": 1.6e-05, "loss": 0.8448, "step": 10946 }, { "epoch": 12.27, "learning_rate": 1.5998591549295776e-05, "loss": 0.6665, "step": 10947 }, { "epoch": 12.27, "learning_rate": 1.599718309859155e-05, "loss": 0.9108, "step": 10948 }, { "epoch": 12.27, "learning_rate": 1.5995774647887325e-05, "loss": 0.8573, "step": 10949 }, { "epoch": 12.28, "learning_rate": 1.5994366197183098e-05, "loss": 1.0182, "step": 10950 }, { "epoch": 12.28, "learning_rate": 1.599295774647887e-05, "loss": 0.8043, "step": 10951 }, { "epoch": 12.28, "learning_rate": 1.5991549295774647e-05, "loss": 0.5747, "step": 10952 }, { "epoch": 12.28, "learning_rate": 1.5990140845070423e-05, "loss": 0.659, "step": 10953 }, { "epoch": 12.28, "learning_rate": 1.59887323943662e-05, "loss": 0.9593, "step": 10954 }, { "epoch": 12.28, "learning_rate": 1.5987323943661973e-05, "loss": 1.3247, "step": 10955 }, { "epoch": 12.28, "learning_rate": 1.598591549295775e-05, "loss": 0.993, "step": 10956 }, { "epoch": 12.28, "learning_rate": 1.5984507042253522e-05, "loss": 0.9507, "step": 10957 }, { "epoch": 12.28, "learning_rate": 1.59830985915493e-05, "loss": 0.9158, "step": 10958 }, { "epoch": 12.29, "learning_rate": 1.598169014084507e-05, "loss": 0.9606, "step": 10959 }, { "epoch": 12.29, "learning_rate": 1.5980281690140844e-05, "loss": 0.8477, "step": 10960 }, { "epoch": 12.29, "learning_rate": 1.597887323943662e-05, "loss": 0.8865, "step": 10961 }, { "epoch": 12.29, "learning_rate": 1.5977464788732393e-05, "loss": 0.6908, "step": 10962 }, { "epoch": 12.29, "learning_rate": 1.597605633802817e-05, "loss": 0.9151, "step": 10963 }, { "epoch": 12.29, "learning_rate": 1.5974647887323943e-05, "loss": 0.7975, "step": 10964 }, { "epoch": 12.29, "learning_rate": 1.597323943661972e-05, "loss": 0.7684, "step": 10965 }, { "epoch": 12.29, "learning_rate": 1.5971830985915492e-05, "loss": 0.7748, "step": 10966 }, { "epoch": 12.29, "learning_rate": 1.5970422535211265e-05, "loss": 0.6579, "step": 10967 }, { "epoch": 12.3, "learning_rate": 1.5969014084507045e-05, "loss": 0.9361, "step": 10968 }, { "epoch": 12.3, "learning_rate": 1.5967605633802818e-05, "loss": 0.8821, "step": 10969 }, { "epoch": 12.3, "learning_rate": 1.5966197183098594e-05, "loss": 0.688, "step": 10970 }, { "epoch": 12.3, "learning_rate": 1.5964788732394367e-05, "loss": 0.8483, "step": 10971 }, { "epoch": 12.3, "learning_rate": 1.5963380281690143e-05, "loss": 0.8584, "step": 10972 }, { "epoch": 12.3, "learning_rate": 1.5961971830985916e-05, "loss": 0.7843, "step": 10973 }, { "epoch": 12.3, "learning_rate": 1.596056338028169e-05, "loss": 0.7965, "step": 10974 }, { "epoch": 12.3, "learning_rate": 1.5959154929577465e-05, "loss": 0.7098, "step": 10975 }, { "epoch": 12.3, "learning_rate": 1.595774647887324e-05, "loss": 0.8924, "step": 10976 }, { "epoch": 12.31, "learning_rate": 1.5956338028169015e-05, "loss": 0.7018, "step": 10977 }, { "epoch": 12.31, "learning_rate": 1.5954929577464788e-05, "loss": 0.7258, "step": 10978 }, { "epoch": 12.31, "learning_rate": 1.5953521126760564e-05, "loss": 0.721, "step": 10979 }, { "epoch": 12.31, "learning_rate": 1.5952112676056337e-05, "loss": 0.9056, "step": 10980 }, { "epoch": 12.31, "learning_rate": 1.5950704225352113e-05, "loss": 0.8126, "step": 10981 }, { "epoch": 12.31, "learning_rate": 1.5949295774647886e-05, "loss": 0.7323, "step": 10982 }, { "epoch": 12.31, "learning_rate": 1.5947887323943663e-05, "loss": 0.7311, "step": 10983 }, { "epoch": 12.31, "learning_rate": 1.594647887323944e-05, "loss": 0.6745, "step": 10984 }, { "epoch": 12.32, "learning_rate": 1.5945070422535212e-05, "loss": 1.0069, "step": 10985 }, { "epoch": 12.32, "learning_rate": 1.5943661971830988e-05, "loss": 0.7327, "step": 10986 }, { "epoch": 12.32, "learning_rate": 1.594225352112676e-05, "loss": 0.6581, "step": 10987 }, { "epoch": 12.32, "learning_rate": 1.5940845070422537e-05, "loss": 0.8251, "step": 10988 }, { "epoch": 12.32, "learning_rate": 1.593943661971831e-05, "loss": 0.721, "step": 10989 }, { "epoch": 12.32, "learning_rate": 1.5938028169014083e-05, "loss": 0.8859, "step": 10990 }, { "epoch": 12.32, "learning_rate": 1.593661971830986e-05, "loss": 0.7428, "step": 10991 }, { "epoch": 12.32, "learning_rate": 1.5935211267605633e-05, "loss": 0.7018, "step": 10992 }, { "epoch": 12.32, "learning_rate": 1.593380281690141e-05, "loss": 0.6806, "step": 10993 }, { "epoch": 12.33, "learning_rate": 1.5932394366197182e-05, "loss": 0.676, "step": 10994 }, { "epoch": 12.33, "learning_rate": 1.5930985915492958e-05, "loss": 0.8342, "step": 10995 }, { "epoch": 12.33, "learning_rate": 1.592957746478873e-05, "loss": 0.8881, "step": 10996 }, { "epoch": 12.33, "learning_rate": 1.5928169014084507e-05, "loss": 0.8809, "step": 10997 }, { "epoch": 12.33, "learning_rate": 1.5926760563380284e-05, "loss": 0.7684, "step": 10998 }, { "epoch": 12.33, "learning_rate": 1.5925352112676057e-05, "loss": 0.7532, "step": 10999 }, { "epoch": 12.33, "learning_rate": 1.5923943661971833e-05, "loss": 0.6708, "step": 11000 }, { "epoch": 12.33, "learning_rate": 1.5922535211267606e-05, "loss": 0.6944, "step": 11001 }, { "epoch": 12.33, "learning_rate": 1.5921126760563382e-05, "loss": 0.6184, "step": 11002 }, { "epoch": 12.34, "learning_rate": 1.5919718309859155e-05, "loss": 0.5918, "step": 11003 }, { "epoch": 12.34, "learning_rate": 1.591830985915493e-05, "loss": 0.7276, "step": 11004 }, { "epoch": 12.34, "learning_rate": 1.5916901408450705e-05, "loss": 1.0216, "step": 11005 }, { "epoch": 12.34, "learning_rate": 1.5915492957746478e-05, "loss": 1.0618, "step": 11006 }, { "epoch": 12.34, "learning_rate": 1.5914084507042254e-05, "loss": 0.9596, "step": 11007 }, { "epoch": 12.34, "learning_rate": 1.5912676056338027e-05, "loss": 0.8925, "step": 11008 }, { "epoch": 12.34, "learning_rate": 1.5911267605633803e-05, "loss": 1.046, "step": 11009 }, { "epoch": 12.34, "learning_rate": 1.5909859154929576e-05, "loss": 0.7132, "step": 11010 }, { "epoch": 12.34, "learning_rate": 1.5908450704225352e-05, "loss": 0.9275, "step": 11011 }, { "epoch": 12.35, "learning_rate": 1.5907042253521125e-05, "loss": 0.8007, "step": 11012 }, { "epoch": 12.35, "learning_rate": 1.59056338028169e-05, "loss": 0.9536, "step": 11013 }, { "epoch": 12.35, "learning_rate": 1.5904225352112678e-05, "loss": 0.8952, "step": 11014 }, { "epoch": 12.35, "learning_rate": 1.590281690140845e-05, "loss": 1.004, "step": 11015 }, { "epoch": 12.35, "learning_rate": 1.5901408450704227e-05, "loss": 0.8642, "step": 11016 }, { "epoch": 12.35, "learning_rate": 1.59e-05, "loss": 0.8113, "step": 11017 }, { "epoch": 12.35, "learning_rate": 1.5898591549295777e-05, "loss": 0.6663, "step": 11018 }, { "epoch": 12.35, "learning_rate": 1.589718309859155e-05, "loss": 0.8521, "step": 11019 }, { "epoch": 12.35, "learning_rate": 1.5895774647887326e-05, "loss": 0.8519, "step": 11020 }, { "epoch": 12.36, "learning_rate": 1.58943661971831e-05, "loss": 0.9022, "step": 11021 }, { "epoch": 12.36, "learning_rate": 1.5892957746478872e-05, "loss": 0.7971, "step": 11022 }, { "epoch": 12.36, "learning_rate": 1.5891549295774648e-05, "loss": 0.9282, "step": 11023 }, { "epoch": 12.36, "learning_rate": 1.589014084507042e-05, "loss": 0.7255, "step": 11024 }, { "epoch": 12.36, "learning_rate": 1.5888732394366197e-05, "loss": 0.741, "step": 11025 }, { "epoch": 12.36, "learning_rate": 1.588732394366197e-05, "loss": 0.8954, "step": 11026 }, { "epoch": 12.36, "learning_rate": 1.5885915492957747e-05, "loss": 0.726, "step": 11027 }, { "epoch": 12.36, "learning_rate": 1.588450704225352e-05, "loss": 0.6543, "step": 11028 }, { "epoch": 12.36, "learning_rate": 1.58830985915493e-05, "loss": 0.8273, "step": 11029 }, { "epoch": 12.37, "learning_rate": 1.5881690140845072e-05, "loss": 0.7736, "step": 11030 }, { "epoch": 12.37, "learning_rate": 1.5880281690140845e-05, "loss": 0.8186, "step": 11031 }, { "epoch": 12.37, "learning_rate": 1.587887323943662e-05, "loss": 0.7588, "step": 11032 }, { "epoch": 12.37, "learning_rate": 1.5877464788732394e-05, "loss": 0.6394, "step": 11033 }, { "epoch": 12.37, "learning_rate": 1.587605633802817e-05, "loss": 0.8608, "step": 11034 }, { "epoch": 12.37, "learning_rate": 1.5874647887323944e-05, "loss": 0.8005, "step": 11035 }, { "epoch": 12.37, "learning_rate": 1.587323943661972e-05, "loss": 0.749, "step": 11036 }, { "epoch": 12.37, "learning_rate": 1.5871830985915493e-05, "loss": 0.8462, "step": 11037 }, { "epoch": 12.37, "learning_rate": 1.5870422535211266e-05, "loss": 0.8153, "step": 11038 }, { "epoch": 12.38, "learning_rate": 1.5869014084507042e-05, "loss": 0.7987, "step": 11039 }, { "epoch": 12.38, "learning_rate": 1.5867605633802815e-05, "loss": 0.7055, "step": 11040 }, { "epoch": 12.38, "learning_rate": 1.586619718309859e-05, "loss": 0.5685, "step": 11041 }, { "epoch": 12.38, "learning_rate": 1.5864788732394364e-05, "loss": 0.7857, "step": 11042 }, { "epoch": 12.38, "learning_rate": 1.586338028169014e-05, "loss": 0.8031, "step": 11043 }, { "epoch": 12.38, "learning_rate": 1.5861971830985917e-05, "loss": 0.815, "step": 11044 }, { "epoch": 12.38, "learning_rate": 1.5860563380281693e-05, "loss": 0.6254, "step": 11045 }, { "epoch": 12.38, "learning_rate": 1.5859154929577466e-05, "loss": 0.933, "step": 11046 }, { "epoch": 12.38, "learning_rate": 1.585774647887324e-05, "loss": 0.7616, "step": 11047 }, { "epoch": 12.39, "learning_rate": 1.5856338028169016e-05, "loss": 0.6939, "step": 11048 }, { "epoch": 12.39, "learning_rate": 1.585492957746479e-05, "loss": 0.7355, "step": 11049 }, { "epoch": 12.39, "learning_rate": 1.5853521126760565e-05, "loss": 0.6887, "step": 11050 }, { "epoch": 12.39, "learning_rate": 1.5852112676056338e-05, "loss": 0.6453, "step": 11051 }, { "epoch": 12.39, "learning_rate": 1.5850704225352114e-05, "loss": 0.8792, "step": 11052 }, { "epoch": 12.39, "learning_rate": 1.5849295774647887e-05, "loss": 0.808, "step": 11053 }, { "epoch": 12.39, "learning_rate": 1.584788732394366e-05, "loss": 1.0317, "step": 11054 }, { "epoch": 12.39, "learning_rate": 1.5846478873239436e-05, "loss": 1.0954, "step": 11055 }, { "epoch": 12.39, "learning_rate": 1.584507042253521e-05, "loss": 0.8463, "step": 11056 }, { "epoch": 12.4, "learning_rate": 1.5843661971830986e-05, "loss": 0.9408, "step": 11057 }, { "epoch": 12.4, "learning_rate": 1.584225352112676e-05, "loss": 0.9856, "step": 11058 }, { "epoch": 12.4, "learning_rate": 1.584084507042254e-05, "loss": 1.0458, "step": 11059 }, { "epoch": 12.4, "learning_rate": 1.583943661971831e-05, "loss": 0.8798, "step": 11060 }, { "epoch": 12.4, "learning_rate": 1.5838028169014088e-05, "loss": 0.7755, "step": 11061 }, { "epoch": 12.4, "learning_rate": 1.583661971830986e-05, "loss": 0.9621, "step": 11062 }, { "epoch": 12.4, "learning_rate": 1.5835211267605634e-05, "loss": 0.8209, "step": 11063 }, { "epoch": 12.4, "learning_rate": 1.583380281690141e-05, "loss": 0.7923, "step": 11064 }, { "epoch": 12.4, "learning_rate": 1.5832394366197183e-05, "loss": 0.8493, "step": 11065 }, { "epoch": 12.41, "learning_rate": 1.583098591549296e-05, "loss": 0.7935, "step": 11066 }, { "epoch": 12.41, "learning_rate": 1.5829577464788732e-05, "loss": 0.8183, "step": 11067 }, { "epoch": 12.41, "learning_rate": 1.582816901408451e-05, "loss": 0.7149, "step": 11068 }, { "epoch": 12.41, "learning_rate": 1.582676056338028e-05, "loss": 0.818, "step": 11069 }, { "epoch": 12.41, "learning_rate": 1.5825352112676054e-05, "loss": 0.8931, "step": 11070 }, { "epoch": 12.41, "learning_rate": 1.582394366197183e-05, "loss": 0.8452, "step": 11071 }, { "epoch": 12.41, "learning_rate": 1.5822535211267604e-05, "loss": 0.7707, "step": 11072 }, { "epoch": 12.41, "learning_rate": 1.582112676056338e-05, "loss": 0.7938, "step": 11073 }, { "epoch": 12.41, "learning_rate": 1.5819718309859156e-05, "loss": 0.8368, "step": 11074 }, { "epoch": 12.42, "learning_rate": 1.5818309859154933e-05, "loss": 0.816, "step": 11075 }, { "epoch": 12.42, "learning_rate": 1.5816901408450706e-05, "loss": 0.6863, "step": 11076 }, { "epoch": 12.42, "learning_rate": 1.5815492957746482e-05, "loss": 0.7209, "step": 11077 }, { "epoch": 12.42, "learning_rate": 1.5814084507042255e-05, "loss": 0.7145, "step": 11078 }, { "epoch": 12.42, "learning_rate": 1.5812676056338028e-05, "loss": 0.7227, "step": 11079 }, { "epoch": 12.42, "learning_rate": 1.5811267605633804e-05, "loss": 0.69, "step": 11080 }, { "epoch": 12.42, "learning_rate": 1.5809859154929577e-05, "loss": 0.7682, "step": 11081 }, { "epoch": 12.42, "learning_rate": 1.5808450704225353e-05, "loss": 0.718, "step": 11082 }, { "epoch": 12.42, "learning_rate": 1.5807042253521126e-05, "loss": 0.7631, "step": 11083 }, { "epoch": 12.43, "learning_rate": 1.5805633802816903e-05, "loss": 0.9616, "step": 11084 }, { "epoch": 12.43, "learning_rate": 1.5804225352112676e-05, "loss": 0.8003, "step": 11085 }, { "epoch": 12.43, "learning_rate": 1.580281690140845e-05, "loss": 0.6756, "step": 11086 }, { "epoch": 12.43, "learning_rate": 1.5801408450704225e-05, "loss": 0.8112, "step": 11087 }, { "epoch": 12.43, "learning_rate": 1.5799999999999998e-05, "loss": 0.7719, "step": 11088 }, { "epoch": 12.43, "learning_rate": 1.5798591549295778e-05, "loss": 0.7876, "step": 11089 }, { "epoch": 12.43, "learning_rate": 1.579718309859155e-05, "loss": 0.7948, "step": 11090 }, { "epoch": 12.43, "learning_rate": 1.5795774647887327e-05, "loss": 0.8857, "step": 11091 }, { "epoch": 12.43, "learning_rate": 1.57943661971831e-05, "loss": 0.6662, "step": 11092 }, { "epoch": 12.44, "learning_rate": 1.5792957746478873e-05, "loss": 0.8109, "step": 11093 }, { "epoch": 12.44, "learning_rate": 1.579154929577465e-05, "loss": 0.7531, "step": 11094 }, { "epoch": 12.44, "learning_rate": 1.5790140845070422e-05, "loss": 0.7684, "step": 11095 }, { "epoch": 12.44, "learning_rate": 1.5788732394366198e-05, "loss": 0.82, "step": 11096 }, { "epoch": 12.44, "learning_rate": 1.578732394366197e-05, "loss": 0.6772, "step": 11097 }, { "epoch": 12.44, "learning_rate": 1.5785915492957748e-05, "loss": 0.6138, "step": 11098 }, { "epoch": 12.44, "learning_rate": 1.578450704225352e-05, "loss": 0.7623, "step": 11099 }, { "epoch": 12.44, "learning_rate": 1.5783098591549297e-05, "loss": 0.843, "step": 11100 }, { "epoch": 12.45, "learning_rate": 1.578169014084507e-05, "loss": 0.4624, "step": 11101 }, { "epoch": 12.45, "learning_rate": 1.5780281690140843e-05, "loss": 0.7313, "step": 11102 }, { "epoch": 12.45, "learning_rate": 1.577887323943662e-05, "loss": 0.9192, "step": 11103 }, { "epoch": 12.45, "learning_rate": 1.5777464788732395e-05, "loss": 0.6726, "step": 11104 }, { "epoch": 12.45, "learning_rate": 1.5776056338028172e-05, "loss": 1.2353, "step": 11105 }, { "epoch": 12.45, "learning_rate": 1.5774647887323945e-05, "loss": 0.9969, "step": 11106 }, { "epoch": 12.45, "learning_rate": 1.577323943661972e-05, "loss": 1.1035, "step": 11107 }, { "epoch": 12.45, "learning_rate": 1.5771830985915494e-05, "loss": 0.8796, "step": 11108 }, { "epoch": 12.45, "learning_rate": 1.5770422535211267e-05, "loss": 0.7522, "step": 11109 }, { "epoch": 12.46, "learning_rate": 1.5769014084507043e-05, "loss": 0.7784, "step": 11110 }, { "epoch": 12.46, "learning_rate": 1.5767605633802816e-05, "loss": 0.8673, "step": 11111 }, { "epoch": 12.46, "learning_rate": 1.5766197183098592e-05, "loss": 0.9374, "step": 11112 }, { "epoch": 12.46, "learning_rate": 1.5764788732394365e-05, "loss": 0.92, "step": 11113 }, { "epoch": 12.46, "learning_rate": 1.5763380281690142e-05, "loss": 0.8205, "step": 11114 }, { "epoch": 12.46, "learning_rate": 1.5761971830985915e-05, "loss": 0.8223, "step": 11115 }, { "epoch": 12.46, "learning_rate": 1.576056338028169e-05, "loss": 0.8539, "step": 11116 }, { "epoch": 12.46, "learning_rate": 1.5759154929577464e-05, "loss": 0.7411, "step": 11117 }, { "epoch": 12.46, "learning_rate": 1.5757746478873237e-05, "loss": 1.0092, "step": 11118 }, { "epoch": 12.47, "learning_rate": 1.5756338028169017e-05, "loss": 0.8139, "step": 11119 }, { "epoch": 12.47, "learning_rate": 1.575492957746479e-05, "loss": 0.95, "step": 11120 }, { "epoch": 12.47, "learning_rate": 1.5753521126760566e-05, "loss": 0.7848, "step": 11121 }, { "epoch": 12.47, "learning_rate": 1.575211267605634e-05, "loss": 0.7516, "step": 11122 }, { "epoch": 12.47, "learning_rate": 1.5750704225352115e-05, "loss": 0.6875, "step": 11123 }, { "epoch": 12.47, "learning_rate": 1.5749295774647888e-05, "loss": 0.7018, "step": 11124 }, { "epoch": 12.47, "learning_rate": 1.574788732394366e-05, "loss": 0.8997, "step": 11125 }, { "epoch": 12.47, "learning_rate": 1.5746478873239437e-05, "loss": 1.0138, "step": 11126 }, { "epoch": 12.47, "learning_rate": 1.574507042253521e-05, "loss": 0.7086, "step": 11127 }, { "epoch": 12.48, "learning_rate": 1.5743661971830987e-05, "loss": 0.7987, "step": 11128 }, { "epoch": 12.48, "learning_rate": 1.574225352112676e-05, "loss": 0.8874, "step": 11129 }, { "epoch": 12.48, "learning_rate": 1.5740845070422536e-05, "loss": 0.864, "step": 11130 }, { "epoch": 12.48, "learning_rate": 1.573943661971831e-05, "loss": 0.774, "step": 11131 }, { "epoch": 12.48, "learning_rate": 1.5738028169014085e-05, "loss": 0.8189, "step": 11132 }, { "epoch": 12.48, "learning_rate": 1.5736619718309858e-05, "loss": 0.8834, "step": 11133 }, { "epoch": 12.48, "learning_rate": 1.573521126760563e-05, "loss": 0.7467, "step": 11134 }, { "epoch": 12.48, "learning_rate": 1.573380281690141e-05, "loss": 0.7692, "step": 11135 }, { "epoch": 12.48, "learning_rate": 1.5732394366197184e-05, "loss": 0.7487, "step": 11136 }, { "epoch": 12.49, "learning_rate": 1.573098591549296e-05, "loss": 0.7185, "step": 11137 }, { "epoch": 12.49, "learning_rate": 1.5729577464788733e-05, "loss": 0.807, "step": 11138 }, { "epoch": 12.49, "learning_rate": 1.572816901408451e-05, "loss": 0.7404, "step": 11139 }, { "epoch": 12.49, "learning_rate": 1.5726760563380282e-05, "loss": 0.7974, "step": 11140 }, { "epoch": 12.49, "learning_rate": 1.5725352112676055e-05, "loss": 0.789, "step": 11141 }, { "epoch": 12.49, "learning_rate": 1.572394366197183e-05, "loss": 0.7279, "step": 11142 }, { "epoch": 12.49, "learning_rate": 1.5722535211267605e-05, "loss": 0.701, "step": 11143 }, { "epoch": 12.49, "learning_rate": 1.572112676056338e-05, "loss": 0.6886, "step": 11144 }, { "epoch": 12.49, "learning_rate": 1.5719718309859154e-05, "loss": 0.7101, "step": 11145 }, { "epoch": 12.5, "learning_rate": 1.571830985915493e-05, "loss": 0.8189, "step": 11146 }, { "epoch": 12.5, "learning_rate": 1.5716901408450703e-05, "loss": 0.7516, "step": 11147 }, { "epoch": 12.5, "learning_rate": 1.571549295774648e-05, "loss": 0.7028, "step": 11148 }, { "epoch": 12.5, "learning_rate": 1.5714084507042252e-05, "loss": 0.8755, "step": 11149 }, { "epoch": 12.5, "learning_rate": 1.571267605633803e-05, "loss": 0.6616, "step": 11150 }, { "epoch": 12.5, "learning_rate": 1.5711267605633805e-05, "loss": 0.7906, "step": 11151 }, { "epoch": 12.5, "learning_rate": 1.5709859154929578e-05, "loss": 0.9575, "step": 11152 }, { "epoch": 12.5, "learning_rate": 1.5708450704225354e-05, "loss": 0.6236, "step": 11153 }, { "epoch": 12.5, "learning_rate": 1.5707042253521127e-05, "loss": 0.7092, "step": 11154 }, { "epoch": 12.51, "learning_rate": 1.5705633802816904e-05, "loss": 1.0829, "step": 11155 }, { "epoch": 12.51, "learning_rate": 1.5704225352112677e-05, "loss": 0.9328, "step": 11156 }, { "epoch": 12.51, "learning_rate": 1.570281690140845e-05, "loss": 0.9044, "step": 11157 }, { "epoch": 12.51, "learning_rate": 1.5701408450704226e-05, "loss": 0.8274, "step": 11158 }, { "epoch": 12.51, "learning_rate": 1.57e-05, "loss": 0.7994, "step": 11159 }, { "epoch": 12.51, "learning_rate": 1.5698591549295775e-05, "loss": 0.8727, "step": 11160 }, { "epoch": 12.51, "learning_rate": 1.5697183098591548e-05, "loss": 0.8703, "step": 11161 }, { "epoch": 12.51, "learning_rate": 1.5695774647887324e-05, "loss": 0.9511, "step": 11162 }, { "epoch": 12.51, "learning_rate": 1.5694366197183097e-05, "loss": 0.9173, "step": 11163 }, { "epoch": 12.52, "learning_rate": 1.5692957746478874e-05, "loss": 0.7456, "step": 11164 }, { "epoch": 12.52, "learning_rate": 1.569154929577465e-05, "loss": 0.7744, "step": 11165 }, { "epoch": 12.52, "learning_rate": 1.5690140845070423e-05, "loss": 0.9287, "step": 11166 }, { "epoch": 12.52, "learning_rate": 1.56887323943662e-05, "loss": 0.8194, "step": 11167 }, { "epoch": 12.52, "learning_rate": 1.5687323943661972e-05, "loss": 0.7511, "step": 11168 }, { "epoch": 12.52, "learning_rate": 1.568591549295775e-05, "loss": 0.8079, "step": 11169 }, { "epoch": 12.52, "learning_rate": 1.568450704225352e-05, "loss": 0.856, "step": 11170 }, { "epoch": 12.52, "learning_rate": 1.5683098591549298e-05, "loss": 0.8588, "step": 11171 }, { "epoch": 12.52, "learning_rate": 1.568169014084507e-05, "loss": 0.7885, "step": 11172 }, { "epoch": 12.53, "learning_rate": 1.5680281690140844e-05, "loss": 0.8834, "step": 11173 }, { "epoch": 12.53, "learning_rate": 1.567887323943662e-05, "loss": 0.8956, "step": 11174 }, { "epoch": 12.53, "learning_rate": 1.5677464788732393e-05, "loss": 0.7408, "step": 11175 }, { "epoch": 12.53, "learning_rate": 1.567605633802817e-05, "loss": 0.824, "step": 11176 }, { "epoch": 12.53, "learning_rate": 1.5674647887323942e-05, "loss": 0.7235, "step": 11177 }, { "epoch": 12.53, "learning_rate": 1.567323943661972e-05, "loss": 0.8257, "step": 11178 }, { "epoch": 12.53, "learning_rate": 1.567183098591549e-05, "loss": 0.8248, "step": 11179 }, { "epoch": 12.53, "learning_rate": 1.567042253521127e-05, "loss": 0.7571, "step": 11180 }, { "epoch": 12.53, "learning_rate": 1.5669014084507044e-05, "loss": 0.8277, "step": 11181 }, { "epoch": 12.54, "learning_rate": 1.5667605633802817e-05, "loss": 0.9969, "step": 11182 }, { "epoch": 12.54, "learning_rate": 1.5666197183098593e-05, "loss": 0.7167, "step": 11183 }, { "epoch": 12.54, "learning_rate": 1.5664788732394366e-05, "loss": 0.7828, "step": 11184 }, { "epoch": 12.54, "learning_rate": 1.5663380281690143e-05, "loss": 0.7689, "step": 11185 }, { "epoch": 12.54, "learning_rate": 1.5661971830985916e-05, "loss": 0.7046, "step": 11186 }, { "epoch": 12.54, "learning_rate": 1.5660563380281692e-05, "loss": 0.7807, "step": 11187 }, { "epoch": 12.54, "learning_rate": 1.5659154929577465e-05, "loss": 0.6577, "step": 11188 }, { "epoch": 12.54, "learning_rate": 1.5657746478873238e-05, "loss": 0.9166, "step": 11189 }, { "epoch": 12.54, "learning_rate": 1.5656338028169014e-05, "loss": 0.8072, "step": 11190 }, { "epoch": 12.55, "learning_rate": 1.5654929577464787e-05, "loss": 0.7044, "step": 11191 }, { "epoch": 12.55, "learning_rate": 1.5653521126760563e-05, "loss": 0.7055, "step": 11192 }, { "epoch": 12.55, "learning_rate": 1.5652112676056336e-05, "loss": 0.9147, "step": 11193 }, { "epoch": 12.55, "learning_rate": 1.5650704225352113e-05, "loss": 0.7061, "step": 11194 }, { "epoch": 12.55, "learning_rate": 1.564929577464789e-05, "loss": 0.8487, "step": 11195 }, { "epoch": 12.55, "learning_rate": 1.5647887323943662e-05, "loss": 0.7598, "step": 11196 }, { "epoch": 12.55, "learning_rate": 1.564647887323944e-05, "loss": 0.8171, "step": 11197 }, { "epoch": 12.55, "learning_rate": 1.564507042253521e-05, "loss": 0.7369, "step": 11198 }, { "epoch": 12.55, "learning_rate": 1.5643661971830988e-05, "loss": 0.7191, "step": 11199 }, { "epoch": 12.56, "learning_rate": 1.564225352112676e-05, "loss": 0.7435, "step": 11200 }, { "epoch": 12.56, "learning_rate": 1.5640845070422537e-05, "loss": 0.5431, "step": 11201 }, { "epoch": 12.56, "learning_rate": 1.563943661971831e-05, "loss": 0.4788, "step": 11202 }, { "epoch": 12.56, "learning_rate": 1.5638028169014086e-05, "loss": 0.6575, "step": 11203 }, { "epoch": 12.56, "learning_rate": 1.563661971830986e-05, "loss": 0.6389, "step": 11204 }, { "epoch": 12.56, "learning_rate": 1.5635211267605632e-05, "loss": 1.0487, "step": 11205 }, { "epoch": 12.56, "learning_rate": 1.563380281690141e-05, "loss": 1.0402, "step": 11206 }, { "epoch": 12.56, "learning_rate": 1.563239436619718e-05, "loss": 1.032, "step": 11207 }, { "epoch": 12.57, "learning_rate": 1.5630985915492958e-05, "loss": 0.8164, "step": 11208 }, { "epoch": 12.57, "learning_rate": 1.562957746478873e-05, "loss": 0.8191, "step": 11209 }, { "epoch": 12.57, "learning_rate": 1.562816901408451e-05, "loss": 0.9553, "step": 11210 }, { "epoch": 12.57, "learning_rate": 1.5626760563380283e-05, "loss": 0.8272, "step": 11211 }, { "epoch": 12.57, "learning_rate": 1.5625352112676056e-05, "loss": 0.8838, "step": 11212 }, { "epoch": 12.57, "learning_rate": 1.5623943661971833e-05, "loss": 0.7468, "step": 11213 }, { "epoch": 12.57, "learning_rate": 1.5622535211267606e-05, "loss": 0.7755, "step": 11214 }, { "epoch": 12.57, "learning_rate": 1.5621126760563382e-05, "loss": 0.7987, "step": 11215 }, { "epoch": 12.57, "learning_rate": 1.5619718309859155e-05, "loss": 0.9051, "step": 11216 }, { "epoch": 12.58, "learning_rate": 1.561830985915493e-05, "loss": 0.6419, "step": 11217 }, { "epoch": 12.58, "learning_rate": 1.5616901408450704e-05, "loss": 0.8346, "step": 11218 }, { "epoch": 12.58, "learning_rate": 1.561549295774648e-05, "loss": 0.9171, "step": 11219 }, { "epoch": 12.58, "learning_rate": 1.5614084507042253e-05, "loss": 0.6119, "step": 11220 }, { "epoch": 12.58, "learning_rate": 1.5612676056338026e-05, "loss": 0.8484, "step": 11221 }, { "epoch": 12.58, "learning_rate": 1.5611267605633803e-05, "loss": 0.8342, "step": 11222 }, { "epoch": 12.58, "learning_rate": 1.5609859154929576e-05, "loss": 0.7402, "step": 11223 }, { "epoch": 12.58, "learning_rate": 1.5608450704225352e-05, "loss": 0.8798, "step": 11224 }, { "epoch": 12.58, "learning_rate": 1.5607042253521128e-05, "loss": 0.8455, "step": 11225 }, { "epoch": 12.59, "learning_rate": 1.5605633802816905e-05, "loss": 0.755, "step": 11226 }, { "epoch": 12.59, "learning_rate": 1.5604225352112677e-05, "loss": 0.8717, "step": 11227 }, { "epoch": 12.59, "learning_rate": 1.560281690140845e-05, "loss": 0.9187, "step": 11228 }, { "epoch": 12.59, "learning_rate": 1.5601408450704227e-05, "loss": 0.7276, "step": 11229 }, { "epoch": 12.59, "learning_rate": 1.56e-05, "loss": 0.7521, "step": 11230 }, { "epoch": 12.59, "learning_rate": 1.5598591549295776e-05, "loss": 0.6695, "step": 11231 }, { "epoch": 12.59, "learning_rate": 1.559718309859155e-05, "loss": 0.8771, "step": 11232 }, { "epoch": 12.59, "learning_rate": 1.5595774647887325e-05, "loss": 0.7673, "step": 11233 }, { "epoch": 12.59, "learning_rate": 1.5594366197183098e-05, "loss": 0.7834, "step": 11234 }, { "epoch": 12.6, "learning_rate": 1.5592957746478875e-05, "loss": 0.6979, "step": 11235 }, { "epoch": 12.6, "learning_rate": 1.5591549295774648e-05, "loss": 0.794, "step": 11236 }, { "epoch": 12.6, "learning_rate": 1.559014084507042e-05, "loss": 0.8251, "step": 11237 }, { "epoch": 12.6, "learning_rate": 1.5588732394366197e-05, "loss": 0.8216, "step": 11238 }, { "epoch": 12.6, "learning_rate": 1.558732394366197e-05, "loss": 0.7296, "step": 11239 }, { "epoch": 12.6, "learning_rate": 1.558591549295775e-05, "loss": 0.8446, "step": 11240 }, { "epoch": 12.6, "learning_rate": 1.5584507042253522e-05, "loss": 0.8164, "step": 11241 }, { "epoch": 12.6, "learning_rate": 1.55830985915493e-05, "loss": 0.7704, "step": 11242 }, { "epoch": 12.6, "learning_rate": 1.5581690140845072e-05, "loss": 0.6155, "step": 11243 }, { "epoch": 12.61, "learning_rate": 1.5580281690140845e-05, "loss": 0.7464, "step": 11244 }, { "epoch": 12.61, "learning_rate": 1.557887323943662e-05, "loss": 0.9738, "step": 11245 }, { "epoch": 12.61, "learning_rate": 1.5577464788732394e-05, "loss": 0.6875, "step": 11246 }, { "epoch": 12.61, "learning_rate": 1.557605633802817e-05, "loss": 0.7128, "step": 11247 }, { "epoch": 12.61, "learning_rate": 1.5574647887323943e-05, "loss": 0.7691, "step": 11248 }, { "epoch": 12.61, "learning_rate": 1.557323943661972e-05, "loss": 0.6815, "step": 11249 }, { "epoch": 12.61, "learning_rate": 1.5571830985915492e-05, "loss": 0.5997, "step": 11250 }, { "epoch": 12.61, "learning_rate": 1.557042253521127e-05, "loss": 0.6872, "step": 11251 }, { "epoch": 12.61, "learning_rate": 1.5569014084507042e-05, "loss": 0.7584, "step": 11252 }, { "epoch": 12.62, "learning_rate": 1.5567605633802815e-05, "loss": 0.6487, "step": 11253 }, { "epoch": 12.62, "learning_rate": 1.556619718309859e-05, "loss": 0.6177, "step": 11254 }, { "epoch": 12.62, "learning_rate": 1.5564788732394364e-05, "loss": 1.1173, "step": 11255 }, { "epoch": 12.62, "learning_rate": 1.5563380281690144e-05, "loss": 0.8663, "step": 11256 }, { "epoch": 12.62, "learning_rate": 1.5561971830985917e-05, "loss": 0.6902, "step": 11257 }, { "epoch": 12.62, "learning_rate": 1.5560563380281693e-05, "loss": 0.8231, "step": 11258 }, { "epoch": 12.62, "learning_rate": 1.5559154929577466e-05, "loss": 0.7139, "step": 11259 }, { "epoch": 12.62, "learning_rate": 1.555774647887324e-05, "loss": 0.6704, "step": 11260 }, { "epoch": 12.62, "learning_rate": 1.5556338028169015e-05, "loss": 0.9577, "step": 11261 }, { "epoch": 12.63, "learning_rate": 1.5554929577464788e-05, "loss": 0.811, "step": 11262 }, { "epoch": 12.63, "learning_rate": 1.5553521126760564e-05, "loss": 0.9162, "step": 11263 }, { "epoch": 12.63, "learning_rate": 1.5552112676056337e-05, "loss": 0.7589, "step": 11264 }, { "epoch": 12.63, "learning_rate": 1.5550704225352114e-05, "loss": 0.7127, "step": 11265 }, { "epoch": 12.63, "learning_rate": 1.5549295774647887e-05, "loss": 0.8061, "step": 11266 }, { "epoch": 12.63, "learning_rate": 1.5547887323943663e-05, "loss": 0.746, "step": 11267 }, { "epoch": 12.63, "learning_rate": 1.5546478873239436e-05, "loss": 0.7054, "step": 11268 }, { "epoch": 12.63, "learning_rate": 1.554507042253521e-05, "loss": 0.8675, "step": 11269 }, { "epoch": 12.63, "learning_rate": 1.5543661971830985e-05, "loss": 0.7391, "step": 11270 }, { "epoch": 12.64, "learning_rate": 1.554225352112676e-05, "loss": 0.807, "step": 11271 }, { "epoch": 12.64, "learning_rate": 1.5540845070422538e-05, "loss": 0.7759, "step": 11272 }, { "epoch": 12.64, "learning_rate": 1.553943661971831e-05, "loss": 0.8836, "step": 11273 }, { "epoch": 12.64, "learning_rate": 1.5538028169014087e-05, "loss": 0.9169, "step": 11274 }, { "epoch": 12.64, "learning_rate": 1.553661971830986e-05, "loss": 0.821, "step": 11275 }, { "epoch": 12.64, "learning_rate": 1.5535211267605633e-05, "loss": 0.7734, "step": 11276 }, { "epoch": 12.64, "learning_rate": 1.553380281690141e-05, "loss": 0.7607, "step": 11277 }, { "epoch": 12.64, "learning_rate": 1.5532394366197182e-05, "loss": 0.7571, "step": 11278 }, { "epoch": 12.64, "learning_rate": 1.553098591549296e-05, "loss": 0.8847, "step": 11279 }, { "epoch": 12.65, "learning_rate": 1.552957746478873e-05, "loss": 0.721, "step": 11280 }, { "epoch": 12.65, "learning_rate": 1.5528169014084508e-05, "loss": 0.7368, "step": 11281 }, { "epoch": 12.65, "learning_rate": 1.552676056338028e-05, "loss": 0.7894, "step": 11282 }, { "epoch": 12.65, "learning_rate": 1.5525352112676057e-05, "loss": 0.7466, "step": 11283 }, { "epoch": 12.65, "learning_rate": 1.552394366197183e-05, "loss": 0.8993, "step": 11284 }, { "epoch": 12.65, "learning_rate": 1.5522535211267603e-05, "loss": 0.6842, "step": 11285 }, { "epoch": 12.65, "learning_rate": 1.5521126760563383e-05, "loss": 0.7506, "step": 11286 }, { "epoch": 12.65, "learning_rate": 1.5519718309859156e-05, "loss": 0.8452, "step": 11287 }, { "epoch": 12.65, "learning_rate": 1.5518309859154932e-05, "loss": 0.7272, "step": 11288 }, { "epoch": 12.66, "learning_rate": 1.5516901408450705e-05, "loss": 0.8611, "step": 11289 }, { "epoch": 12.66, "learning_rate": 1.551549295774648e-05, "loss": 0.7981, "step": 11290 }, { "epoch": 12.66, "learning_rate": 1.5514084507042254e-05, "loss": 0.6533, "step": 11291 }, { "epoch": 12.66, "learning_rate": 1.5512676056338027e-05, "loss": 0.6733, "step": 11292 }, { "epoch": 12.66, "learning_rate": 1.5511267605633804e-05, "loss": 0.779, "step": 11293 }, { "epoch": 12.66, "learning_rate": 1.5509859154929577e-05, "loss": 0.7282, "step": 11294 }, { "epoch": 12.66, "learning_rate": 1.5508450704225353e-05, "loss": 0.8145, "step": 11295 }, { "epoch": 12.66, "learning_rate": 1.5507042253521126e-05, "loss": 0.8482, "step": 11296 }, { "epoch": 12.66, "learning_rate": 1.5505633802816902e-05, "loss": 0.6576, "step": 11297 }, { "epoch": 12.67, "learning_rate": 1.5504225352112675e-05, "loss": 0.8082, "step": 11298 }, { "epoch": 12.67, "learning_rate": 1.550281690140845e-05, "loss": 0.7845, "step": 11299 }, { "epoch": 12.67, "learning_rate": 1.5501408450704224e-05, "loss": 0.7868, "step": 11300 }, { "epoch": 12.67, "learning_rate": 1.55e-05, "loss": 0.7896, "step": 11301 }, { "epoch": 12.67, "learning_rate": 1.5498591549295777e-05, "loss": 0.768, "step": 11302 }, { "epoch": 12.67, "learning_rate": 1.549718309859155e-05, "loss": 0.8497, "step": 11303 }, { "epoch": 12.67, "learning_rate": 1.5495774647887326e-05, "loss": 0.58, "step": 11304 }, { "epoch": 12.67, "learning_rate": 1.54943661971831e-05, "loss": 0.8396, "step": 11305 }, { "epoch": 12.67, "learning_rate": 1.5492957746478876e-05, "loss": 0.9923, "step": 11306 }, { "epoch": 12.68, "learning_rate": 1.549154929577465e-05, "loss": 0.8166, "step": 11307 }, { "epoch": 12.68, "learning_rate": 1.549014084507042e-05, "loss": 0.8532, "step": 11308 }, { "epoch": 12.68, "learning_rate": 1.5488732394366198e-05, "loss": 0.8423, "step": 11309 }, { "epoch": 12.68, "learning_rate": 1.548732394366197e-05, "loss": 0.8502, "step": 11310 }, { "epoch": 12.68, "learning_rate": 1.5485915492957747e-05, "loss": 0.9477, "step": 11311 }, { "epoch": 12.68, "learning_rate": 1.548450704225352e-05, "loss": 0.9052, "step": 11312 }, { "epoch": 12.68, "learning_rate": 1.5483098591549296e-05, "loss": 0.8883, "step": 11313 }, { "epoch": 12.68, "learning_rate": 1.548169014084507e-05, "loss": 0.8994, "step": 11314 }, { "epoch": 12.68, "learning_rate": 1.5480281690140846e-05, "loss": 0.8035, "step": 11315 }, { "epoch": 12.69, "learning_rate": 1.5478873239436622e-05, "loss": 0.8233, "step": 11316 }, { "epoch": 12.69, "learning_rate": 1.5477464788732395e-05, "loss": 0.8019, "step": 11317 }, { "epoch": 12.69, "learning_rate": 1.547605633802817e-05, "loss": 0.8165, "step": 11318 }, { "epoch": 12.69, "learning_rate": 1.5474647887323944e-05, "loss": 0.621, "step": 11319 }, { "epoch": 12.69, "learning_rate": 1.547323943661972e-05, "loss": 0.7552, "step": 11320 }, { "epoch": 12.69, "learning_rate": 1.5471830985915493e-05, "loss": 0.9027, "step": 11321 }, { "epoch": 12.69, "learning_rate": 1.547042253521127e-05, "loss": 0.7017, "step": 11322 }, { "epoch": 12.69, "learning_rate": 1.5469014084507043e-05, "loss": 0.6763, "step": 11323 }, { "epoch": 12.7, "learning_rate": 1.5467605633802816e-05, "loss": 0.7039, "step": 11324 }, { "epoch": 12.7, "learning_rate": 1.5466197183098592e-05, "loss": 0.738, "step": 11325 }, { "epoch": 12.7, "learning_rate": 1.5464788732394365e-05, "loss": 0.7808, "step": 11326 }, { "epoch": 12.7, "learning_rate": 1.546338028169014e-05, "loss": 0.718, "step": 11327 }, { "epoch": 12.7, "learning_rate": 1.5461971830985914e-05, "loss": 0.7983, "step": 11328 }, { "epoch": 12.7, "learning_rate": 1.546056338028169e-05, "loss": 0.739, "step": 11329 }, { "epoch": 12.7, "learning_rate": 1.5459154929577463e-05, "loss": 0.8683, "step": 11330 }, { "epoch": 12.7, "learning_rate": 1.545774647887324e-05, "loss": 0.6338, "step": 11331 }, { "epoch": 12.7, "learning_rate": 1.5456338028169016e-05, "loss": 0.8031, "step": 11332 }, { "epoch": 12.71, "learning_rate": 1.545492957746479e-05, "loss": 0.6787, "step": 11333 }, { "epoch": 12.71, "learning_rate": 1.5453521126760565e-05, "loss": 0.7215, "step": 11334 }, { "epoch": 12.71, "learning_rate": 1.545211267605634e-05, "loss": 0.7814, "step": 11335 }, { "epoch": 12.71, "learning_rate": 1.5450704225352115e-05, "loss": 0.734, "step": 11336 }, { "epoch": 12.71, "learning_rate": 1.5449295774647888e-05, "loss": 0.7809, "step": 11337 }, { "epoch": 12.71, "learning_rate": 1.5447887323943664e-05, "loss": 0.8129, "step": 11338 }, { "epoch": 12.71, "learning_rate": 1.5446478873239437e-05, "loss": 0.9209, "step": 11339 }, { "epoch": 12.71, "learning_rate": 1.544507042253521e-05, "loss": 0.6791, "step": 11340 }, { "epoch": 12.71, "learning_rate": 1.5443661971830986e-05, "loss": 0.7592, "step": 11341 }, { "epoch": 12.72, "learning_rate": 1.544225352112676e-05, "loss": 0.7483, "step": 11342 }, { "epoch": 12.72, "learning_rate": 1.5440845070422535e-05, "loss": 0.5559, "step": 11343 }, { "epoch": 12.72, "learning_rate": 1.543943661971831e-05, "loss": 0.8613, "step": 11344 }, { "epoch": 12.72, "learning_rate": 1.5438028169014085e-05, "loss": 0.7193, "step": 11345 }, { "epoch": 12.72, "learning_rate": 1.543661971830986e-05, "loss": 0.7261, "step": 11346 }, { "epoch": 12.72, "learning_rate": 1.5435211267605634e-05, "loss": 0.649, "step": 11347 }, { "epoch": 12.72, "learning_rate": 1.543380281690141e-05, "loss": 0.7529, "step": 11348 }, { "epoch": 12.72, "learning_rate": 1.5432394366197183e-05, "loss": 0.8671, "step": 11349 }, { "epoch": 12.72, "learning_rate": 1.543098591549296e-05, "loss": 0.7636, "step": 11350 }, { "epoch": 12.73, "learning_rate": 1.5429577464788733e-05, "loss": 0.5707, "step": 11351 }, { "epoch": 12.73, "learning_rate": 1.542816901408451e-05, "loss": 0.9021, "step": 11352 }, { "epoch": 12.73, "learning_rate": 1.5426760563380282e-05, "loss": 0.5873, "step": 11353 }, { "epoch": 12.73, "learning_rate": 1.5425352112676058e-05, "loss": 0.67, "step": 11354 }, { "epoch": 12.73, "learning_rate": 1.542394366197183e-05, "loss": 1.0775, "step": 11355 }, { "epoch": 12.73, "learning_rate": 1.5422535211267604e-05, "loss": 0.8823, "step": 11356 }, { "epoch": 12.73, "learning_rate": 1.542112676056338e-05, "loss": 0.8837, "step": 11357 }, { "epoch": 12.73, "learning_rate": 1.5419718309859153e-05, "loss": 0.9875, "step": 11358 }, { "epoch": 12.73, "learning_rate": 1.541830985915493e-05, "loss": 0.8285, "step": 11359 }, { "epoch": 12.74, "learning_rate": 1.5416901408450703e-05, "loss": 0.7755, "step": 11360 }, { "epoch": 12.74, "learning_rate": 1.541549295774648e-05, "loss": 0.7811, "step": 11361 }, { "epoch": 12.74, "learning_rate": 1.5414084507042255e-05, "loss": 0.867, "step": 11362 }, { "epoch": 12.74, "learning_rate": 1.5412676056338028e-05, "loss": 0.781, "step": 11363 }, { "epoch": 12.74, "learning_rate": 1.5411267605633805e-05, "loss": 1.0169, "step": 11364 }, { "epoch": 12.74, "learning_rate": 1.5409859154929577e-05, "loss": 0.6405, "step": 11365 }, { "epoch": 12.74, "learning_rate": 1.5408450704225354e-05, "loss": 0.7457, "step": 11366 }, { "epoch": 12.74, "learning_rate": 1.5407042253521127e-05, "loss": 0.827, "step": 11367 }, { "epoch": 12.74, "learning_rate": 1.5405633802816903e-05, "loss": 0.7032, "step": 11368 }, { "epoch": 12.75, "learning_rate": 1.5404225352112676e-05, "loss": 0.8367, "step": 11369 }, { "epoch": 12.75, "learning_rate": 1.5402816901408452e-05, "loss": 0.8119, "step": 11370 }, { "epoch": 12.75, "learning_rate": 1.5401408450704225e-05, "loss": 0.7338, "step": 11371 }, { "epoch": 12.75, "learning_rate": 1.5399999999999998e-05, "loss": 0.7089, "step": 11372 }, { "epoch": 12.75, "learning_rate": 1.5398591549295775e-05, "loss": 0.7828, "step": 11373 }, { "epoch": 12.75, "learning_rate": 1.5397183098591548e-05, "loss": 0.7909, "step": 11374 }, { "epoch": 12.75, "learning_rate": 1.5395774647887324e-05, "loss": 0.7932, "step": 11375 }, { "epoch": 12.75, "learning_rate": 1.5394366197183097e-05, "loss": 0.8647, "step": 11376 }, { "epoch": 12.75, "learning_rate": 1.5392957746478877e-05, "loss": 0.8073, "step": 11377 }, { "epoch": 12.76, "learning_rate": 1.539154929577465e-05, "loss": 0.7709, "step": 11378 }, { "epoch": 12.76, "learning_rate": 1.5390140845070422e-05, "loss": 0.8942, "step": 11379 }, { "epoch": 12.76, "learning_rate": 1.53887323943662e-05, "loss": 0.7823, "step": 11380 }, { "epoch": 12.76, "learning_rate": 1.538732394366197e-05, "loss": 0.7538, "step": 11381 }, { "epoch": 12.76, "learning_rate": 1.5385915492957748e-05, "loss": 0.7199, "step": 11382 }, { "epoch": 12.76, "learning_rate": 1.538450704225352e-05, "loss": 0.7958, "step": 11383 }, { "epoch": 12.76, "learning_rate": 1.5383098591549297e-05, "loss": 0.7463, "step": 11384 }, { "epoch": 12.76, "learning_rate": 1.538169014084507e-05, "loss": 0.6705, "step": 11385 }, { "epoch": 12.76, "learning_rate": 1.5380281690140847e-05, "loss": 0.7733, "step": 11386 }, { "epoch": 12.77, "learning_rate": 1.537887323943662e-05, "loss": 0.6777, "step": 11387 }, { "epoch": 12.77, "learning_rate": 1.5377464788732392e-05, "loss": 0.6385, "step": 11388 }, { "epoch": 12.77, "learning_rate": 1.537605633802817e-05, "loss": 0.7128, "step": 11389 }, { "epoch": 12.77, "learning_rate": 1.5374647887323942e-05, "loss": 0.7934, "step": 11390 }, { "epoch": 12.77, "learning_rate": 1.5373239436619718e-05, "loss": 0.7585, "step": 11391 }, { "epoch": 12.77, "learning_rate": 1.5371830985915494e-05, "loss": 0.7984, "step": 11392 }, { "epoch": 12.77, "learning_rate": 1.537042253521127e-05, "loss": 0.7693, "step": 11393 }, { "epoch": 12.77, "learning_rate": 1.5369014084507044e-05, "loss": 0.7541, "step": 11394 }, { "epoch": 12.77, "learning_rate": 1.5367605633802817e-05, "loss": 0.7614, "step": 11395 }, { "epoch": 12.78, "learning_rate": 1.5366197183098593e-05, "loss": 0.7064, "step": 11396 }, { "epoch": 12.78, "learning_rate": 1.5364788732394366e-05, "loss": 0.7344, "step": 11397 }, { "epoch": 12.78, "learning_rate": 1.5363380281690142e-05, "loss": 0.7118, "step": 11398 }, { "epoch": 12.78, "learning_rate": 1.5361971830985915e-05, "loss": 0.5029, "step": 11399 }, { "epoch": 12.78, "learning_rate": 1.536056338028169e-05, "loss": 0.6421, "step": 11400 }, { "epoch": 12.78, "learning_rate": 1.5359154929577464e-05, "loss": 0.4986, "step": 11401 }, { "epoch": 12.78, "learning_rate": 1.535774647887324e-05, "loss": 0.6491, "step": 11402 }, { "epoch": 12.78, "learning_rate": 1.5356338028169014e-05, "loss": 0.8146, "step": 11403 }, { "epoch": 12.78, "learning_rate": 1.5354929577464787e-05, "loss": 0.5644, "step": 11404 }, { "epoch": 12.79, "learning_rate": 1.5353521126760563e-05, "loss": 0.9832, "step": 11405 }, { "epoch": 12.79, "learning_rate": 1.5352112676056336e-05, "loss": 1.1507, "step": 11406 }, { "epoch": 12.79, "learning_rate": 1.5350704225352116e-05, "loss": 0.9391, "step": 11407 }, { "epoch": 12.79, "learning_rate": 1.534929577464789e-05, "loss": 0.9217, "step": 11408 }, { "epoch": 12.79, "learning_rate": 1.5347887323943665e-05, "loss": 0.9021, "step": 11409 }, { "epoch": 12.79, "learning_rate": 1.5346478873239438e-05, "loss": 0.8647, "step": 11410 }, { "epoch": 12.79, "learning_rate": 1.534507042253521e-05, "loss": 0.8197, "step": 11411 }, { "epoch": 12.79, "learning_rate": 1.5343661971830987e-05, "loss": 0.9506, "step": 11412 }, { "epoch": 12.79, "learning_rate": 1.534225352112676e-05, "loss": 0.7655, "step": 11413 }, { "epoch": 12.8, "learning_rate": 1.5340845070422536e-05, "loss": 0.7853, "step": 11414 }, { "epoch": 12.8, "learning_rate": 1.533943661971831e-05, "loss": 0.7276, "step": 11415 }, { "epoch": 12.8, "learning_rate": 1.5338028169014086e-05, "loss": 0.8375, "step": 11416 }, { "epoch": 12.8, "learning_rate": 1.533661971830986e-05, "loss": 0.9928, "step": 11417 }, { "epoch": 12.8, "learning_rate": 1.5335211267605635e-05, "loss": 0.8658, "step": 11418 }, { "epoch": 12.8, "learning_rate": 1.5333802816901408e-05, "loss": 0.9629, "step": 11419 }, { "epoch": 12.8, "learning_rate": 1.533239436619718e-05, "loss": 0.863, "step": 11420 }, { "epoch": 12.8, "learning_rate": 1.5330985915492957e-05, "loss": 0.7212, "step": 11421 }, { "epoch": 12.8, "learning_rate": 1.5329577464788734e-05, "loss": 0.8056, "step": 11422 }, { "epoch": 12.81, "learning_rate": 1.532816901408451e-05, "loss": 0.6966, "step": 11423 }, { "epoch": 12.81, "learning_rate": 1.5326760563380283e-05, "loss": 0.9416, "step": 11424 }, { "epoch": 12.81, "learning_rate": 1.532535211267606e-05, "loss": 0.8438, "step": 11425 }, { "epoch": 12.81, "learning_rate": 1.5323943661971832e-05, "loss": 0.7171, "step": 11426 }, { "epoch": 12.81, "learning_rate": 1.5322535211267605e-05, "loss": 0.7398, "step": 11427 }, { "epoch": 12.81, "learning_rate": 1.532112676056338e-05, "loss": 0.6948, "step": 11428 }, { "epoch": 12.81, "learning_rate": 1.5319718309859154e-05, "loss": 0.8232, "step": 11429 }, { "epoch": 12.81, "learning_rate": 1.531830985915493e-05, "loss": 0.8473, "step": 11430 }, { "epoch": 12.82, "learning_rate": 1.5316901408450704e-05, "loss": 0.6628, "step": 11431 }, { "epoch": 12.82, "learning_rate": 1.531549295774648e-05, "loss": 0.8193, "step": 11432 }, { "epoch": 12.82, "learning_rate": 1.5314084507042253e-05, "loss": 0.7843, "step": 11433 }, { "epoch": 12.82, "learning_rate": 1.531267605633803e-05, "loss": 0.749, "step": 11434 }, { "epoch": 12.82, "learning_rate": 1.5311267605633802e-05, "loss": 0.7643, "step": 11435 }, { "epoch": 12.82, "learning_rate": 1.5309859154929575e-05, "loss": 0.7469, "step": 11436 }, { "epoch": 12.82, "learning_rate": 1.5308450704225355e-05, "loss": 0.6159, "step": 11437 }, { "epoch": 12.82, "learning_rate": 1.5307042253521128e-05, "loss": 0.7448, "step": 11438 }, { "epoch": 12.82, "learning_rate": 1.5305633802816904e-05, "loss": 0.7418, "step": 11439 }, { "epoch": 12.83, "learning_rate": 1.5304225352112677e-05, "loss": 0.8257, "step": 11440 }, { "epoch": 12.83, "learning_rate": 1.5302816901408453e-05, "loss": 0.6979, "step": 11441 }, { "epoch": 12.83, "learning_rate": 1.5301408450704226e-05, "loss": 0.6662, "step": 11442 }, { "epoch": 12.83, "learning_rate": 1.53e-05, "loss": 0.8476, "step": 11443 }, { "epoch": 12.83, "learning_rate": 1.5298591549295776e-05, "loss": 0.8863, "step": 11444 }, { "epoch": 12.83, "learning_rate": 1.529718309859155e-05, "loss": 0.5802, "step": 11445 }, { "epoch": 12.83, "learning_rate": 1.5295774647887325e-05, "loss": 0.7061, "step": 11446 }, { "epoch": 12.83, "learning_rate": 1.5294366197183098e-05, "loss": 0.7567, "step": 11447 }, { "epoch": 12.83, "learning_rate": 1.5292957746478874e-05, "loss": 0.6222, "step": 11448 }, { "epoch": 12.84, "learning_rate": 1.5291549295774647e-05, "loss": 0.5727, "step": 11449 }, { "epoch": 12.84, "learning_rate": 1.529014084507042e-05, "loss": 0.7251, "step": 11450 }, { "epoch": 12.84, "learning_rate": 1.5288732394366196e-05, "loss": 0.8303, "step": 11451 }, { "epoch": 12.84, "learning_rate": 1.5287323943661973e-05, "loss": 0.769, "step": 11452 }, { "epoch": 12.84, "learning_rate": 1.528591549295775e-05, "loss": 0.9038, "step": 11453 }, { "epoch": 12.84, "learning_rate": 1.5284507042253522e-05, "loss": 0.9173, "step": 11454 }, { "epoch": 12.84, "learning_rate": 1.5283098591549298e-05, "loss": 0.8942, "step": 11455 }, { "epoch": 12.84, "learning_rate": 1.528169014084507e-05, "loss": 0.7986, "step": 11456 }, { "epoch": 12.84, "learning_rate": 1.5280281690140848e-05, "loss": 0.9268, "step": 11457 }, { "epoch": 12.85, "learning_rate": 1.527887323943662e-05, "loss": 0.7978, "step": 11458 }, { "epoch": 12.85, "learning_rate": 1.5277464788732393e-05, "loss": 0.8918, "step": 11459 }, { "epoch": 12.85, "learning_rate": 1.527605633802817e-05, "loss": 0.7866, "step": 11460 }, { "epoch": 12.85, "learning_rate": 1.5274647887323943e-05, "loss": 0.8543, "step": 11461 }, { "epoch": 12.85, "learning_rate": 1.527323943661972e-05, "loss": 0.818, "step": 11462 }, { "epoch": 12.85, "learning_rate": 1.5271830985915492e-05, "loss": 1.0503, "step": 11463 }, { "epoch": 12.85, "learning_rate": 1.5270422535211268e-05, "loss": 0.7921, "step": 11464 }, { "epoch": 12.85, "learning_rate": 1.526901408450704e-05, "loss": 0.7123, "step": 11465 }, { "epoch": 12.85, "learning_rate": 1.5267605633802814e-05, "loss": 0.7283, "step": 11466 }, { "epoch": 12.86, "learning_rate": 1.5266197183098594e-05, "loss": 0.8316, "step": 11467 }, { "epoch": 12.86, "learning_rate": 1.5264788732394367e-05, "loss": 0.8109, "step": 11468 }, { "epoch": 12.86, "learning_rate": 1.5263380281690143e-05, "loss": 0.7687, "step": 11469 }, { "epoch": 12.86, "learning_rate": 1.5261971830985916e-05, "loss": 0.9314, "step": 11470 }, { "epoch": 12.86, "learning_rate": 1.5260563380281692e-05, "loss": 0.8424, "step": 11471 }, { "epoch": 12.86, "learning_rate": 1.5259154929577465e-05, "loss": 0.6979, "step": 11472 }, { "epoch": 12.86, "learning_rate": 1.525774647887324e-05, "loss": 0.8238, "step": 11473 }, { "epoch": 12.86, "learning_rate": 1.5256338028169015e-05, "loss": 0.6532, "step": 11474 }, { "epoch": 12.86, "learning_rate": 1.525492957746479e-05, "loss": 0.7292, "step": 11475 }, { "epoch": 12.87, "learning_rate": 1.5253521126760564e-05, "loss": 0.7394, "step": 11476 }, { "epoch": 12.87, "learning_rate": 1.5252112676056339e-05, "loss": 0.7885, "step": 11477 }, { "epoch": 12.87, "learning_rate": 1.5250704225352112e-05, "loss": 0.6823, "step": 11478 }, { "epoch": 12.87, "learning_rate": 1.5249295774647886e-05, "loss": 0.6827, "step": 11479 }, { "epoch": 12.87, "learning_rate": 1.524788732394366e-05, "loss": 0.678, "step": 11480 }, { "epoch": 12.87, "learning_rate": 1.5246478873239435e-05, "loss": 0.8831, "step": 11481 }, { "epoch": 12.87, "learning_rate": 1.524507042253521e-05, "loss": 0.7788, "step": 11482 }, { "epoch": 12.87, "learning_rate": 1.5243661971830988e-05, "loss": 0.6793, "step": 11483 }, { "epoch": 12.87, "learning_rate": 1.5242253521126763e-05, "loss": 0.9065, "step": 11484 }, { "epoch": 12.88, "learning_rate": 1.5240845070422537e-05, "loss": 0.64, "step": 11485 }, { "epoch": 12.88, "learning_rate": 1.523943661971831e-05, "loss": 0.7572, "step": 11486 }, { "epoch": 12.88, "learning_rate": 1.5238028169014085e-05, "loss": 0.7108, "step": 11487 }, { "epoch": 12.88, "learning_rate": 1.523661971830986e-05, "loss": 0.6689, "step": 11488 }, { "epoch": 12.88, "learning_rate": 1.5235211267605634e-05, "loss": 0.7364, "step": 11489 }, { "epoch": 12.88, "learning_rate": 1.5233802816901409e-05, "loss": 0.811, "step": 11490 }, { "epoch": 12.88, "learning_rate": 1.5232394366197183e-05, "loss": 0.7448, "step": 11491 }, { "epoch": 12.88, "learning_rate": 1.5230985915492958e-05, "loss": 0.5668, "step": 11492 }, { "epoch": 12.88, "learning_rate": 1.5229577464788733e-05, "loss": 0.8142, "step": 11493 }, { "epoch": 12.89, "learning_rate": 1.5228169014084506e-05, "loss": 0.8722, "step": 11494 }, { "epoch": 12.89, "learning_rate": 1.522676056338028e-05, "loss": 0.6943, "step": 11495 }, { "epoch": 12.89, "learning_rate": 1.5225352112676055e-05, "loss": 0.6868, "step": 11496 }, { "epoch": 12.89, "learning_rate": 1.522394366197183e-05, "loss": 0.7031, "step": 11497 }, { "epoch": 12.89, "learning_rate": 1.5222535211267608e-05, "loss": 0.6146, "step": 11498 }, { "epoch": 12.89, "learning_rate": 1.5221126760563382e-05, "loss": 0.9866, "step": 11499 }, { "epoch": 12.89, "learning_rate": 1.5219718309859157e-05, "loss": 0.5694, "step": 11500 }, { "epoch": 12.89, "learning_rate": 1.5218309859154932e-05, "loss": 0.5508, "step": 11501 }, { "epoch": 12.89, "learning_rate": 1.5216901408450705e-05, "loss": 0.837, "step": 11502 }, { "epoch": 12.9, "learning_rate": 1.5215492957746479e-05, "loss": 0.6995, "step": 11503 }, { "epoch": 12.9, "learning_rate": 1.5214084507042254e-05, "loss": 0.8234, "step": 11504 }, { "epoch": 12.9, "learning_rate": 1.5212676056338028e-05, "loss": 0.8856, "step": 11505 }, { "epoch": 12.9, "learning_rate": 1.5211267605633803e-05, "loss": 0.8995, "step": 11506 }, { "epoch": 12.9, "learning_rate": 1.5209859154929578e-05, "loss": 0.8719, "step": 11507 }, { "epoch": 12.9, "learning_rate": 1.5208450704225352e-05, "loss": 0.937, "step": 11508 }, { "epoch": 12.9, "learning_rate": 1.5207042253521127e-05, "loss": 0.8193, "step": 11509 }, { "epoch": 12.9, "learning_rate": 1.52056338028169e-05, "loss": 0.8066, "step": 11510 }, { "epoch": 12.9, "learning_rate": 1.5204225352112675e-05, "loss": 0.7776, "step": 11511 }, { "epoch": 12.91, "learning_rate": 1.520281690140845e-05, "loss": 0.6951, "step": 11512 }, { "epoch": 12.91, "learning_rate": 1.5201408450704227e-05, "loss": 0.9176, "step": 11513 }, { "epoch": 12.91, "learning_rate": 1.5200000000000002e-05, "loss": 0.8182, "step": 11514 }, { "epoch": 12.91, "learning_rate": 1.5198591549295776e-05, "loss": 0.7381, "step": 11515 }, { "epoch": 12.91, "learning_rate": 1.5197183098591551e-05, "loss": 0.7758, "step": 11516 }, { "epoch": 12.91, "learning_rate": 1.5195774647887326e-05, "loss": 0.978, "step": 11517 }, { "epoch": 12.91, "learning_rate": 1.5194366197183099e-05, "loss": 0.9172, "step": 11518 }, { "epoch": 12.91, "learning_rate": 1.5192957746478873e-05, "loss": 0.79, "step": 11519 }, { "epoch": 12.91, "learning_rate": 1.5191549295774648e-05, "loss": 0.7736, "step": 11520 }, { "epoch": 12.92, "learning_rate": 1.5190140845070423e-05, "loss": 0.8044, "step": 11521 }, { "epoch": 12.92, "learning_rate": 1.5188732394366197e-05, "loss": 0.6916, "step": 11522 }, { "epoch": 12.92, "learning_rate": 1.5187323943661972e-05, "loss": 0.7151, "step": 11523 }, { "epoch": 12.92, "learning_rate": 1.5185915492957747e-05, "loss": 0.8514, "step": 11524 }, { "epoch": 12.92, "learning_rate": 1.5184507042253521e-05, "loss": 0.8142, "step": 11525 }, { "epoch": 12.92, "learning_rate": 1.5183098591549294e-05, "loss": 0.7391, "step": 11526 }, { "epoch": 12.92, "learning_rate": 1.5181690140845069e-05, "loss": 0.86, "step": 11527 }, { "epoch": 12.92, "learning_rate": 1.5180281690140847e-05, "loss": 0.6748, "step": 11528 }, { "epoch": 12.92, "learning_rate": 1.5178873239436621e-05, "loss": 0.6269, "step": 11529 }, { "epoch": 12.93, "learning_rate": 1.5177464788732396e-05, "loss": 0.7091, "step": 11530 }, { "epoch": 12.93, "learning_rate": 1.517605633802817e-05, "loss": 0.8318, "step": 11531 }, { "epoch": 12.93, "learning_rate": 1.5174647887323945e-05, "loss": 0.6847, "step": 11532 }, { "epoch": 12.93, "learning_rate": 1.517323943661972e-05, "loss": 0.7527, "step": 11533 }, { "epoch": 12.93, "learning_rate": 1.5171830985915493e-05, "loss": 0.7435, "step": 11534 }, { "epoch": 12.93, "learning_rate": 1.5170422535211268e-05, "loss": 0.8069, "step": 11535 }, { "epoch": 12.93, "learning_rate": 1.5169014084507042e-05, "loss": 0.689, "step": 11536 }, { "epoch": 12.93, "learning_rate": 1.5167605633802817e-05, "loss": 0.9111, "step": 11537 }, { "epoch": 12.93, "learning_rate": 1.5166197183098591e-05, "loss": 0.8827, "step": 11538 }, { "epoch": 12.94, "learning_rate": 1.5164788732394366e-05, "loss": 0.9456, "step": 11539 }, { "epoch": 12.94, "learning_rate": 1.516338028169014e-05, "loss": 0.7742, "step": 11540 }, { "epoch": 12.94, "learning_rate": 1.5161971830985915e-05, "loss": 0.7138, "step": 11541 }, { "epoch": 12.94, "learning_rate": 1.5160563380281688e-05, "loss": 0.8113, "step": 11542 }, { "epoch": 12.94, "learning_rate": 1.5159154929577466e-05, "loss": 0.6434, "step": 11543 }, { "epoch": 12.94, "learning_rate": 1.5157746478873241e-05, "loss": 0.8084, "step": 11544 }, { "epoch": 12.94, "learning_rate": 1.5156338028169016e-05, "loss": 0.8342, "step": 11545 }, { "epoch": 12.94, "learning_rate": 1.515492957746479e-05, "loss": 0.9256, "step": 11546 }, { "epoch": 12.95, "learning_rate": 1.5153521126760565e-05, "loss": 0.7678, "step": 11547 }, { "epoch": 12.95, "learning_rate": 1.515211267605634e-05, "loss": 0.7402, "step": 11548 }, { "epoch": 12.95, "learning_rate": 1.5150704225352112e-05, "loss": 0.7728, "step": 11549 }, { "epoch": 12.95, "learning_rate": 1.5149295774647887e-05, "loss": 0.6025, "step": 11550 }, { "epoch": 12.95, "learning_rate": 1.5147887323943662e-05, "loss": 0.6518, "step": 11551 }, { "epoch": 12.95, "learning_rate": 1.5146478873239436e-05, "loss": 0.6898, "step": 11552 }, { "epoch": 12.95, "learning_rate": 1.5145070422535211e-05, "loss": 0.6639, "step": 11553 }, { "epoch": 12.95, "learning_rate": 1.5143661971830986e-05, "loss": 0.7039, "step": 11554 }, { "epoch": 12.95, "learning_rate": 1.514225352112676e-05, "loss": 0.918, "step": 11555 }, { "epoch": 12.96, "learning_rate": 1.5140845070422535e-05, "loss": 0.8575, "step": 11556 }, { "epoch": 12.96, "learning_rate": 1.5139436619718308e-05, "loss": 1.0812, "step": 11557 }, { "epoch": 12.96, "learning_rate": 1.5138028169014086e-05, "loss": 0.7495, "step": 11558 }, { "epoch": 12.96, "learning_rate": 1.513661971830986e-05, "loss": 0.9677, "step": 11559 }, { "epoch": 12.96, "learning_rate": 1.5135211267605635e-05, "loss": 0.7599, "step": 11560 }, { "epoch": 12.96, "learning_rate": 1.513380281690141e-05, "loss": 0.7181, "step": 11561 }, { "epoch": 12.96, "learning_rate": 1.5132394366197184e-05, "loss": 0.8445, "step": 11562 }, { "epoch": 12.96, "learning_rate": 1.5130985915492959e-05, "loss": 0.7649, "step": 11563 }, { "epoch": 12.96, "learning_rate": 1.5129577464788734e-05, "loss": 0.8668, "step": 11564 }, { "epoch": 12.97, "learning_rate": 1.5128169014084507e-05, "loss": 0.7193, "step": 11565 }, { "epoch": 12.97, "learning_rate": 1.5126760563380281e-05, "loss": 0.8875, "step": 11566 }, { "epoch": 12.97, "learning_rate": 1.5125352112676056e-05, "loss": 1.1308, "step": 11567 }, { "epoch": 12.97, "learning_rate": 1.512394366197183e-05, "loss": 0.8175, "step": 11568 }, { "epoch": 12.97, "learning_rate": 1.5122535211267605e-05, "loss": 0.6393, "step": 11569 }, { "epoch": 12.97, "learning_rate": 1.512112676056338e-05, "loss": 0.8157, "step": 11570 }, { "epoch": 12.97, "learning_rate": 1.5119718309859154e-05, "loss": 0.7191, "step": 11571 }, { "epoch": 12.97, "learning_rate": 1.5118309859154929e-05, "loss": 0.8584, "step": 11572 }, { "epoch": 12.97, "learning_rate": 1.5116901408450705e-05, "loss": 0.8206, "step": 11573 }, { "epoch": 12.98, "learning_rate": 1.511549295774648e-05, "loss": 0.8765, "step": 11574 }, { "epoch": 12.98, "learning_rate": 1.5114084507042255e-05, "loss": 0.7016, "step": 11575 }, { "epoch": 12.98, "learning_rate": 1.511267605633803e-05, "loss": 0.8532, "step": 11576 }, { "epoch": 12.98, "learning_rate": 1.5111267605633804e-05, "loss": 0.5855, "step": 11577 }, { "epoch": 12.98, "learning_rate": 1.5109859154929579e-05, "loss": 0.9171, "step": 11578 }, { "epoch": 12.98, "learning_rate": 1.5108450704225353e-05, "loss": 0.7876, "step": 11579 }, { "epoch": 12.98, "learning_rate": 1.5107042253521128e-05, "loss": 0.7126, "step": 11580 }, { "epoch": 12.98, "learning_rate": 1.5105633802816901e-05, "loss": 0.6323, "step": 11581 }, { "epoch": 12.98, "learning_rate": 1.5104225352112676e-05, "loss": 0.8613, "step": 11582 }, { "epoch": 12.99, "learning_rate": 1.510281690140845e-05, "loss": 0.803, "step": 11583 }, { "epoch": 12.99, "learning_rate": 1.5101408450704225e-05, "loss": 0.8935, "step": 11584 }, { "epoch": 12.99, "learning_rate": 1.51e-05, "loss": 0.7967, "step": 11585 }, { "epoch": 12.99, "learning_rate": 1.5098591549295774e-05, "loss": 0.7224, "step": 11586 }, { "epoch": 12.99, "learning_rate": 1.5097183098591549e-05, "loss": 0.7887, "step": 11587 }, { "epoch": 12.99, "learning_rate": 1.5095774647887327e-05, "loss": 0.6151, "step": 11588 }, { "epoch": 12.99, "learning_rate": 1.50943661971831e-05, "loss": 0.7297, "step": 11589 }, { "epoch": 12.99, "learning_rate": 1.5092957746478874e-05, "loss": 0.662, "step": 11590 }, { "epoch": 12.99, "learning_rate": 1.5091549295774649e-05, "loss": 0.604, "step": 11591 }, { "epoch": 13.0, "learning_rate": 1.5090140845070424e-05, "loss": 0.5921, "step": 11592 }, { "epoch": 13.0, "learning_rate": 1.5088732394366198e-05, "loss": 0.6253, "step": 11593 }, { "epoch": 13.0, "learning_rate": 1.5087323943661973e-05, "loss": 0.7, "step": 11594 }, { "epoch": 13.0, "learning_rate": 1.5085915492957747e-05, "loss": 0.906, "step": 11595 }, { "epoch": 13.0, "learning_rate": 1.5084507042253522e-05, "loss": 0.6211, "step": 11596 }, { "epoch": 13.0, "learning_rate": 1.5083098591549295e-05, "loss": 0.9965, "step": 11597 }, { "epoch": 13.0, "learning_rate": 1.508169014084507e-05, "loss": 1.0142, "step": 11598 }, { "epoch": 13.0, "learning_rate": 1.5080281690140844e-05, "loss": 0.7286, "step": 11599 }, { "epoch": 13.0, "learning_rate": 1.5078873239436619e-05, "loss": 0.8231, "step": 11600 }, { "epoch": 13.01, "learning_rate": 1.5077464788732394e-05, "loss": 0.7342, "step": 11601 }, { "epoch": 13.01, "learning_rate": 1.5076056338028168e-05, "loss": 0.8064, "step": 11602 }, { "epoch": 13.01, "learning_rate": 1.5074647887323943e-05, "loss": 0.7772, "step": 11603 }, { "epoch": 13.01, "learning_rate": 1.5073239436619721e-05, "loss": 0.6253, "step": 11604 }, { "epoch": 13.01, "learning_rate": 1.5071830985915494e-05, "loss": 0.7697, "step": 11605 }, { "epoch": 13.01, "learning_rate": 1.5070422535211269e-05, "loss": 0.8217, "step": 11606 }, { "epoch": 13.01, "learning_rate": 1.5069014084507043e-05, "loss": 0.8027, "step": 11607 }, { "epoch": 13.01, "learning_rate": 1.5067605633802818e-05, "loss": 0.7514, "step": 11608 }, { "epoch": 13.01, "learning_rate": 1.5066197183098592e-05, "loss": 0.6984, "step": 11609 }, { "epoch": 13.02, "learning_rate": 1.5064788732394367e-05, "loss": 0.6693, "step": 11610 }, { "epoch": 13.02, "learning_rate": 1.5063380281690142e-05, "loss": 0.7479, "step": 11611 }, { "epoch": 13.02, "learning_rate": 1.5061971830985916e-05, "loss": 0.6849, "step": 11612 }, { "epoch": 13.02, "learning_rate": 1.506056338028169e-05, "loss": 0.795, "step": 11613 }, { "epoch": 13.02, "learning_rate": 1.5059154929577464e-05, "loss": 0.7107, "step": 11614 }, { "epoch": 13.02, "learning_rate": 1.5057746478873239e-05, "loss": 0.6784, "step": 11615 }, { "epoch": 13.02, "learning_rate": 1.5056338028169013e-05, "loss": 0.8177, "step": 11616 }, { "epoch": 13.02, "learning_rate": 1.5054929577464788e-05, "loss": 0.78, "step": 11617 }, { "epoch": 13.02, "learning_rate": 1.5053521126760562e-05, "loss": 0.7568, "step": 11618 }, { "epoch": 13.03, "learning_rate": 1.505211267605634e-05, "loss": 0.7607, "step": 11619 }, { "epoch": 13.03, "learning_rate": 1.5050704225352115e-05, "loss": 0.7467, "step": 11620 }, { "epoch": 13.03, "learning_rate": 1.5049295774647888e-05, "loss": 0.7008, "step": 11621 }, { "epoch": 13.03, "learning_rate": 1.5047887323943663e-05, "loss": 0.7116, "step": 11622 }, { "epoch": 13.03, "learning_rate": 1.5046478873239437e-05, "loss": 0.6696, "step": 11623 }, { "epoch": 13.03, "learning_rate": 1.5045070422535212e-05, "loss": 0.7429, "step": 11624 }, { "epoch": 13.03, "learning_rate": 1.5043661971830987e-05, "loss": 0.7818, "step": 11625 }, { "epoch": 13.03, "learning_rate": 1.5042253521126761e-05, "loss": 0.6855, "step": 11626 }, { "epoch": 13.03, "learning_rate": 1.5040845070422536e-05, "loss": 0.7795, "step": 11627 }, { "epoch": 13.04, "learning_rate": 1.503943661971831e-05, "loss": 0.797, "step": 11628 }, { "epoch": 13.04, "learning_rate": 1.5038028169014083e-05, "loss": 0.7562, "step": 11629 }, { "epoch": 13.04, "learning_rate": 1.5036619718309858e-05, "loss": 0.5698, "step": 11630 }, { "epoch": 13.04, "learning_rate": 1.5035211267605633e-05, "loss": 0.6349, "step": 11631 }, { "epoch": 13.04, "learning_rate": 1.5033802816901407e-05, "loss": 0.8299, "step": 11632 }, { "epoch": 13.04, "learning_rate": 1.5032394366197182e-05, "loss": 0.7433, "step": 11633 }, { "epoch": 13.04, "learning_rate": 1.503098591549296e-05, "loss": 0.6015, "step": 11634 }, { "epoch": 13.04, "learning_rate": 1.5029577464788735e-05, "loss": 0.714, "step": 11635 }, { "epoch": 13.04, "learning_rate": 1.502816901408451e-05, "loss": 0.6891, "step": 11636 }, { "epoch": 13.05, "learning_rate": 1.5026760563380282e-05, "loss": 0.5976, "step": 11637 }, { "epoch": 13.05, "learning_rate": 1.5025352112676057e-05, "loss": 0.7706, "step": 11638 }, { "epoch": 13.05, "learning_rate": 1.5023943661971832e-05, "loss": 0.5554, "step": 11639 }, { "epoch": 13.05, "learning_rate": 1.5022535211267606e-05, "loss": 0.6319, "step": 11640 }, { "epoch": 13.05, "learning_rate": 1.502112676056338e-05, "loss": 0.6811, "step": 11641 }, { "epoch": 13.05, "learning_rate": 1.5019718309859155e-05, "loss": 0.6609, "step": 11642 }, { "epoch": 13.05, "learning_rate": 1.501830985915493e-05, "loss": 0.5877, "step": 11643 }, { "epoch": 13.05, "learning_rate": 1.5016901408450705e-05, "loss": 0.7293, "step": 11644 }, { "epoch": 13.05, "learning_rate": 1.5015492957746478e-05, "loss": 0.4352, "step": 11645 }, { "epoch": 13.06, "learning_rate": 1.5014084507042252e-05, "loss": 0.5007, "step": 11646 }, { "epoch": 13.06, "learning_rate": 1.5012676056338027e-05, "loss": 0.974, "step": 11647 }, { "epoch": 13.06, "learning_rate": 1.5011267605633802e-05, "loss": 0.8527, "step": 11648 }, { "epoch": 13.06, "learning_rate": 1.500985915492958e-05, "loss": 0.8464, "step": 11649 }, { "epoch": 13.06, "learning_rate": 1.5008450704225354e-05, "loss": 0.8455, "step": 11650 }, { "epoch": 13.06, "learning_rate": 1.5007042253521129e-05, "loss": 0.7368, "step": 11651 }, { "epoch": 13.06, "learning_rate": 1.5005633802816902e-05, "loss": 0.6735, "step": 11652 }, { "epoch": 13.06, "learning_rate": 1.5004225352112676e-05, "loss": 0.7271, "step": 11653 }, { "epoch": 13.07, "learning_rate": 1.5002816901408451e-05, "loss": 0.7932, "step": 11654 }, { "epoch": 13.07, "learning_rate": 1.5001408450704226e-05, "loss": 0.8466, "step": 11655 }, { "epoch": 13.07, "learning_rate": 1.5e-05, "loss": 0.8577, "step": 11656 }, { "epoch": 13.07, "learning_rate": 1.4998591549295775e-05, "loss": 0.6879, "step": 11657 }, { "epoch": 13.07, "learning_rate": 1.499718309859155e-05, "loss": 0.7335, "step": 11658 }, { "epoch": 13.07, "learning_rate": 1.4995774647887324e-05, "loss": 0.7081, "step": 11659 }, { "epoch": 13.07, "learning_rate": 1.4994366197183099e-05, "loss": 0.7296, "step": 11660 }, { "epoch": 13.07, "learning_rate": 1.4992957746478874e-05, "loss": 0.9511, "step": 11661 }, { "epoch": 13.07, "learning_rate": 1.4991549295774648e-05, "loss": 0.5992, "step": 11662 }, { "epoch": 13.08, "learning_rate": 1.4990140845070423e-05, "loss": 0.6529, "step": 11663 }, { "epoch": 13.08, "learning_rate": 1.4988732394366197e-05, "loss": 0.7581, "step": 11664 }, { "epoch": 13.08, "learning_rate": 1.4987323943661972e-05, "loss": 0.7626, "step": 11665 }, { "epoch": 13.08, "learning_rate": 1.4985915492957747e-05, "loss": 0.6467, "step": 11666 }, { "epoch": 13.08, "learning_rate": 1.4984507042253521e-05, "loss": 0.7624, "step": 11667 }, { "epoch": 13.08, "learning_rate": 1.4983098591549296e-05, "loss": 0.7274, "step": 11668 }, { "epoch": 13.08, "learning_rate": 1.498169014084507e-05, "loss": 0.7594, "step": 11669 }, { "epoch": 13.08, "learning_rate": 1.4980281690140845e-05, "loss": 0.7412, "step": 11670 }, { "epoch": 13.08, "learning_rate": 1.497887323943662e-05, "loss": 0.6639, "step": 11671 }, { "epoch": 13.09, "learning_rate": 1.4977464788732395e-05, "loss": 0.6333, "step": 11672 }, { "epoch": 13.09, "learning_rate": 1.497605633802817e-05, "loss": 0.7122, "step": 11673 }, { "epoch": 13.09, "learning_rate": 1.4974647887323944e-05, "loss": 0.7102, "step": 11674 }, { "epoch": 13.09, "learning_rate": 1.497323943661972e-05, "loss": 0.7243, "step": 11675 }, { "epoch": 13.09, "learning_rate": 1.4971830985915493e-05, "loss": 0.785, "step": 11676 }, { "epoch": 13.09, "learning_rate": 1.4970422535211268e-05, "loss": 0.8457, "step": 11677 }, { "epoch": 13.09, "learning_rate": 1.4969014084507042e-05, "loss": 0.6268, "step": 11678 }, { "epoch": 13.09, "learning_rate": 1.4967605633802817e-05, "loss": 0.7233, "step": 11679 }, { "epoch": 13.09, "learning_rate": 1.4966197183098592e-05, "loss": 0.7556, "step": 11680 }, { "epoch": 13.1, "learning_rate": 1.4964788732394366e-05, "loss": 0.5593, "step": 11681 }, { "epoch": 13.1, "learning_rate": 1.4963380281690141e-05, "loss": 0.5885, "step": 11682 }, { "epoch": 13.1, "learning_rate": 1.4961971830985917e-05, "loss": 0.5753, "step": 11683 }, { "epoch": 13.1, "learning_rate": 1.496056338028169e-05, "loss": 0.7127, "step": 11684 }, { "epoch": 13.1, "learning_rate": 1.4959154929577465e-05, "loss": 0.6102, "step": 11685 }, { "epoch": 13.1, "learning_rate": 1.495774647887324e-05, "loss": 0.6793, "step": 11686 }, { "epoch": 13.1, "learning_rate": 1.4956338028169014e-05, "loss": 0.6251, "step": 11687 }, { "epoch": 13.1, "learning_rate": 1.4954929577464789e-05, "loss": 0.8773, "step": 11688 }, { "epoch": 13.1, "learning_rate": 1.4953521126760563e-05, "loss": 0.6401, "step": 11689 }, { "epoch": 13.11, "learning_rate": 1.4952112676056338e-05, "loss": 0.5667, "step": 11690 }, { "epoch": 13.11, "learning_rate": 1.4950704225352114e-05, "loss": 0.4594, "step": 11691 }, { "epoch": 13.11, "learning_rate": 1.4949295774647887e-05, "loss": 0.6806, "step": 11692 }, { "epoch": 13.11, "learning_rate": 1.4947887323943662e-05, "loss": 0.4469, "step": 11693 }, { "epoch": 13.11, "learning_rate": 1.4946478873239437e-05, "loss": 0.7362, "step": 11694 }, { "epoch": 13.11, "learning_rate": 1.4945070422535211e-05, "loss": 0.42, "step": 11695 }, { "epoch": 13.11, "learning_rate": 1.4943661971830986e-05, "loss": 0.6816, "step": 11696 }, { "epoch": 13.11, "learning_rate": 1.494225352112676e-05, "loss": 0.8037, "step": 11697 }, { "epoch": 13.11, "learning_rate": 1.4940845070422537e-05, "loss": 0.8766, "step": 11698 }, { "epoch": 13.12, "learning_rate": 1.4939436619718311e-05, "loss": 0.7178, "step": 11699 }, { "epoch": 13.12, "learning_rate": 1.4938028169014084e-05, "loss": 0.9261, "step": 11700 }, { "epoch": 13.12, "learning_rate": 1.4936619718309859e-05, "loss": 0.8072, "step": 11701 }, { "epoch": 13.12, "learning_rate": 1.4935211267605634e-05, "loss": 0.9355, "step": 11702 }, { "epoch": 13.12, "learning_rate": 1.4933802816901408e-05, "loss": 0.6971, "step": 11703 }, { "epoch": 13.12, "learning_rate": 1.4932394366197183e-05, "loss": 0.8252, "step": 11704 }, { "epoch": 13.12, "learning_rate": 1.4930985915492958e-05, "loss": 0.75, "step": 11705 }, { "epoch": 13.12, "learning_rate": 1.4929577464788734e-05, "loss": 0.8839, "step": 11706 }, { "epoch": 13.12, "learning_rate": 1.4928169014084509e-05, "loss": 0.7366, "step": 11707 }, { "epoch": 13.13, "learning_rate": 1.4926760563380282e-05, "loss": 0.7399, "step": 11708 }, { "epoch": 13.13, "learning_rate": 1.4925352112676056e-05, "loss": 0.7181, "step": 11709 }, { "epoch": 13.13, "learning_rate": 1.492394366197183e-05, "loss": 0.7882, "step": 11710 }, { "epoch": 13.13, "learning_rate": 1.4922535211267605e-05, "loss": 0.678, "step": 11711 }, { "epoch": 13.13, "learning_rate": 1.492112676056338e-05, "loss": 0.6397, "step": 11712 }, { "epoch": 13.13, "learning_rate": 1.4919718309859156e-05, "loss": 0.808, "step": 11713 }, { "epoch": 13.13, "learning_rate": 1.4918309859154931e-05, "loss": 0.6707, "step": 11714 }, { "epoch": 13.13, "learning_rate": 1.4916901408450706e-05, "loss": 0.8694, "step": 11715 }, { "epoch": 13.13, "learning_rate": 1.4915492957746479e-05, "loss": 0.7945, "step": 11716 }, { "epoch": 13.14, "learning_rate": 1.4914084507042253e-05, "loss": 0.7473, "step": 11717 }, { "epoch": 13.14, "learning_rate": 1.4912676056338028e-05, "loss": 0.6771, "step": 11718 }, { "epoch": 13.14, "learning_rate": 1.4911267605633803e-05, "loss": 0.6943, "step": 11719 }, { "epoch": 13.14, "learning_rate": 1.4909859154929577e-05, "loss": 0.7251, "step": 11720 }, { "epoch": 13.14, "learning_rate": 1.4908450704225354e-05, "loss": 0.6185, "step": 11721 }, { "epoch": 13.14, "learning_rate": 1.4907042253521128e-05, "loss": 0.7711, "step": 11722 }, { "epoch": 13.14, "learning_rate": 1.4905633802816903e-05, "loss": 0.5779, "step": 11723 }, { "epoch": 13.14, "learning_rate": 1.4904225352112676e-05, "loss": 0.6815, "step": 11724 }, { "epoch": 13.14, "learning_rate": 1.490281690140845e-05, "loss": 0.8809, "step": 11725 }, { "epoch": 13.15, "learning_rate": 1.4901408450704225e-05, "loss": 0.7987, "step": 11726 }, { "epoch": 13.15, "learning_rate": 1.49e-05, "loss": 0.7547, "step": 11727 }, { "epoch": 13.15, "learning_rate": 1.4898591549295776e-05, "loss": 0.7147, "step": 11728 }, { "epoch": 13.15, "learning_rate": 1.489718309859155e-05, "loss": 0.5678, "step": 11729 }, { "epoch": 13.15, "learning_rate": 1.4895774647887325e-05, "loss": 0.6124, "step": 11730 }, { "epoch": 13.15, "learning_rate": 1.48943661971831e-05, "loss": 0.7671, "step": 11731 }, { "epoch": 13.15, "learning_rate": 1.4892957746478873e-05, "loss": 0.745, "step": 11732 }, { "epoch": 13.15, "learning_rate": 1.4891549295774647e-05, "loss": 0.6965, "step": 11733 }, { "epoch": 13.15, "learning_rate": 1.4890140845070422e-05, "loss": 0.7294, "step": 11734 }, { "epoch": 13.16, "learning_rate": 1.4888732394366197e-05, "loss": 0.6435, "step": 11735 }, { "epoch": 13.16, "learning_rate": 1.4887323943661973e-05, "loss": 0.6854, "step": 11736 }, { "epoch": 13.16, "learning_rate": 1.4885915492957748e-05, "loss": 0.6573, "step": 11737 }, { "epoch": 13.16, "learning_rate": 1.4884507042253522e-05, "loss": 0.6827, "step": 11738 }, { "epoch": 13.16, "learning_rate": 1.4883098591549297e-05, "loss": 0.6669, "step": 11739 }, { "epoch": 13.16, "learning_rate": 1.488169014084507e-05, "loss": 0.8109, "step": 11740 }, { "epoch": 13.16, "learning_rate": 1.4880281690140845e-05, "loss": 0.5978, "step": 11741 }, { "epoch": 13.16, "learning_rate": 1.487887323943662e-05, "loss": 0.7035, "step": 11742 }, { "epoch": 13.16, "learning_rate": 1.4877464788732396e-05, "loss": 0.7741, "step": 11743 }, { "epoch": 13.17, "learning_rate": 1.487605633802817e-05, "loss": 0.6688, "step": 11744 }, { "epoch": 13.17, "learning_rate": 1.4874647887323945e-05, "loss": 0.4876, "step": 11745 }, { "epoch": 13.17, "learning_rate": 1.487323943661972e-05, "loss": 0.6029, "step": 11746 }, { "epoch": 13.17, "learning_rate": 1.4871830985915494e-05, "loss": 0.8576, "step": 11747 }, { "epoch": 13.17, "learning_rate": 1.4870422535211267e-05, "loss": 0.8351, "step": 11748 }, { "epoch": 13.17, "learning_rate": 1.4869014084507042e-05, "loss": 0.8465, "step": 11749 }, { "epoch": 13.17, "learning_rate": 1.4867605633802816e-05, "loss": 0.7539, "step": 11750 }, { "epoch": 13.17, "learning_rate": 1.4866197183098593e-05, "loss": 0.7536, "step": 11751 }, { "epoch": 13.17, "learning_rate": 1.4864788732394367e-05, "loss": 0.6345, "step": 11752 }, { "epoch": 13.18, "learning_rate": 1.4863380281690142e-05, "loss": 0.7837, "step": 11753 }, { "epoch": 13.18, "learning_rate": 1.4861971830985917e-05, "loss": 0.757, "step": 11754 }, { "epoch": 13.18, "learning_rate": 1.4860563380281691e-05, "loss": 0.74, "step": 11755 }, { "epoch": 13.18, "learning_rate": 1.4859154929577464e-05, "loss": 0.6852, "step": 11756 }, { "epoch": 13.18, "learning_rate": 1.4857746478873239e-05, "loss": 0.8352, "step": 11757 }, { "epoch": 13.18, "learning_rate": 1.4856338028169013e-05, "loss": 0.6733, "step": 11758 }, { "epoch": 13.18, "learning_rate": 1.485492957746479e-05, "loss": 0.8418, "step": 11759 }, { "epoch": 13.18, "learning_rate": 1.4853521126760564e-05, "loss": 0.7295, "step": 11760 }, { "epoch": 13.18, "learning_rate": 1.4852112676056339e-05, "loss": 0.7669, "step": 11761 }, { "epoch": 13.19, "learning_rate": 1.4850704225352114e-05, "loss": 0.8381, "step": 11762 }, { "epoch": 13.19, "learning_rate": 1.4849295774647888e-05, "loss": 0.6887, "step": 11763 }, { "epoch": 13.19, "learning_rate": 1.4847887323943661e-05, "loss": 0.7668, "step": 11764 }, { "epoch": 13.19, "learning_rate": 1.4846478873239436e-05, "loss": 0.8021, "step": 11765 }, { "epoch": 13.19, "learning_rate": 1.4845070422535212e-05, "loss": 0.8253, "step": 11766 }, { "epoch": 13.19, "learning_rate": 1.4843661971830987e-05, "loss": 0.6474, "step": 11767 }, { "epoch": 13.19, "learning_rate": 1.4842253521126761e-05, "loss": 0.6067, "step": 11768 }, { "epoch": 13.19, "learning_rate": 1.4840845070422536e-05, "loss": 0.6087, "step": 11769 }, { "epoch": 13.2, "learning_rate": 1.483943661971831e-05, "loss": 0.653, "step": 11770 }, { "epoch": 13.2, "learning_rate": 1.4838028169014085e-05, "loss": 0.7662, "step": 11771 }, { "epoch": 13.2, "learning_rate": 1.4836619718309858e-05, "loss": 0.6845, "step": 11772 }, { "epoch": 13.2, "learning_rate": 1.4835211267605633e-05, "loss": 0.652, "step": 11773 }, { "epoch": 13.2, "learning_rate": 1.483380281690141e-05, "loss": 0.5908, "step": 11774 }, { "epoch": 13.2, "learning_rate": 1.4832394366197184e-05, "loss": 0.6494, "step": 11775 }, { "epoch": 13.2, "learning_rate": 1.4830985915492959e-05, "loss": 0.7182, "step": 11776 }, { "epoch": 13.2, "learning_rate": 1.4829577464788733e-05, "loss": 0.8362, "step": 11777 }, { "epoch": 13.2, "learning_rate": 1.4828169014084508e-05, "loss": 0.6324, "step": 11778 }, { "epoch": 13.21, "learning_rate": 1.482676056338028e-05, "loss": 0.7218, "step": 11779 }, { "epoch": 13.21, "learning_rate": 1.4825352112676055e-05, "loss": 0.7074, "step": 11780 }, { "epoch": 13.21, "learning_rate": 1.4823943661971832e-05, "loss": 0.672, "step": 11781 }, { "epoch": 13.21, "learning_rate": 1.4822535211267606e-05, "loss": 0.6, "step": 11782 }, { "epoch": 13.21, "learning_rate": 1.4821126760563381e-05, "loss": 0.6351, "step": 11783 }, { "epoch": 13.21, "learning_rate": 1.4819718309859156e-05, "loss": 0.524, "step": 11784 }, { "epoch": 13.21, "learning_rate": 1.481830985915493e-05, "loss": 0.7642, "step": 11785 }, { "epoch": 13.21, "learning_rate": 1.4816901408450705e-05, "loss": 0.8027, "step": 11786 }, { "epoch": 13.21, "learning_rate": 1.4815492957746478e-05, "loss": 0.8756, "step": 11787 }, { "epoch": 13.22, "learning_rate": 1.4814084507042253e-05, "loss": 0.6342, "step": 11788 }, { "epoch": 13.22, "learning_rate": 1.4812676056338029e-05, "loss": 0.835, "step": 11789 }, { "epoch": 13.22, "learning_rate": 1.4811267605633804e-05, "loss": 0.6116, "step": 11790 }, { "epoch": 13.22, "learning_rate": 1.4809859154929578e-05, "loss": 0.6522, "step": 11791 }, { "epoch": 13.22, "learning_rate": 1.4808450704225353e-05, "loss": 0.6776, "step": 11792 }, { "epoch": 13.22, "learning_rate": 1.4807042253521127e-05, "loss": 0.6832, "step": 11793 }, { "epoch": 13.22, "learning_rate": 1.4805633802816902e-05, "loss": 0.7973, "step": 11794 }, { "epoch": 13.22, "learning_rate": 1.4804225352112675e-05, "loss": 0.4917, "step": 11795 }, { "epoch": 13.22, "learning_rate": 1.4802816901408451e-05, "loss": 0.7417, "step": 11796 }, { "epoch": 13.23, "learning_rate": 1.4801408450704226e-05, "loss": 0.9963, "step": 11797 }, { "epoch": 13.23, "learning_rate": 1.48e-05, "loss": 0.8647, "step": 11798 }, { "epoch": 13.23, "learning_rate": 1.4798591549295775e-05, "loss": 0.8061, "step": 11799 }, { "epoch": 13.23, "learning_rate": 1.479718309859155e-05, "loss": 0.6736, "step": 11800 }, { "epoch": 13.23, "learning_rate": 1.4795774647887325e-05, "loss": 0.8716, "step": 11801 }, { "epoch": 13.23, "learning_rate": 1.4794366197183099e-05, "loss": 0.8348, "step": 11802 }, { "epoch": 13.23, "learning_rate": 1.4792957746478872e-05, "loss": 0.8372, "step": 11803 }, { "epoch": 13.23, "learning_rate": 1.4791549295774648e-05, "loss": 0.6528, "step": 11804 }, { "epoch": 13.23, "learning_rate": 1.4790140845070423e-05, "loss": 0.699, "step": 11805 }, { "epoch": 13.24, "learning_rate": 1.4788732394366198e-05, "loss": 0.6729, "step": 11806 }, { "epoch": 13.24, "learning_rate": 1.4787323943661972e-05, "loss": 0.8251, "step": 11807 }, { "epoch": 13.24, "learning_rate": 1.4785915492957747e-05, "loss": 0.6339, "step": 11808 }, { "epoch": 13.24, "learning_rate": 1.4784507042253522e-05, "loss": 0.6153, "step": 11809 }, { "epoch": 13.24, "learning_rate": 1.4783098591549296e-05, "loss": 0.7099, "step": 11810 }, { "epoch": 13.24, "learning_rate": 1.478169014084507e-05, "loss": 0.752, "step": 11811 }, { "epoch": 13.24, "learning_rate": 1.4780281690140846e-05, "loss": 0.7781, "step": 11812 }, { "epoch": 13.24, "learning_rate": 1.477887323943662e-05, "loss": 0.6972, "step": 11813 }, { "epoch": 13.24, "learning_rate": 1.4777464788732395e-05, "loss": 0.6422, "step": 11814 }, { "epoch": 13.25, "learning_rate": 1.477605633802817e-05, "loss": 0.6536, "step": 11815 }, { "epoch": 13.25, "learning_rate": 1.4774647887323944e-05, "loss": 0.6601, "step": 11816 }, { "epoch": 13.25, "learning_rate": 1.4773239436619719e-05, "loss": 0.7016, "step": 11817 }, { "epoch": 13.25, "learning_rate": 1.4771830985915493e-05, "loss": 0.7059, "step": 11818 }, { "epoch": 13.25, "learning_rate": 1.4770422535211268e-05, "loss": 0.5921, "step": 11819 }, { "epoch": 13.25, "learning_rate": 1.4769014084507043e-05, "loss": 0.6277, "step": 11820 }, { "epoch": 13.25, "learning_rate": 1.4767605633802817e-05, "loss": 0.5954, "step": 11821 }, { "epoch": 13.25, "learning_rate": 1.4766197183098592e-05, "loss": 0.6505, "step": 11822 }, { "epoch": 13.25, "learning_rate": 1.4764788732394367e-05, "loss": 0.6109, "step": 11823 }, { "epoch": 13.26, "learning_rate": 1.4763380281690141e-05, "loss": 0.7474, "step": 11824 }, { "epoch": 13.26, "learning_rate": 1.4761971830985916e-05, "loss": 0.7001, "step": 11825 }, { "epoch": 13.26, "learning_rate": 1.476056338028169e-05, "loss": 0.7428, "step": 11826 }, { "epoch": 13.26, "learning_rate": 1.4759154929577465e-05, "loss": 0.7497, "step": 11827 }, { "epoch": 13.26, "learning_rate": 1.475774647887324e-05, "loss": 0.6777, "step": 11828 }, { "epoch": 13.26, "learning_rate": 1.4756338028169014e-05, "loss": 0.5518, "step": 11829 }, { "epoch": 13.26, "learning_rate": 1.4754929577464789e-05, "loss": 0.6675, "step": 11830 }, { "epoch": 13.26, "learning_rate": 1.4753521126760564e-05, "loss": 0.6338, "step": 11831 }, { "epoch": 13.26, "learning_rate": 1.4752112676056338e-05, "loss": 0.6517, "step": 11832 }, { "epoch": 13.27, "learning_rate": 1.4750704225352113e-05, "loss": 0.5937, "step": 11833 }, { "epoch": 13.27, "learning_rate": 1.474929577464789e-05, "loss": 0.5619, "step": 11834 }, { "epoch": 13.27, "learning_rate": 1.4747887323943662e-05, "loss": 0.4883, "step": 11835 }, { "epoch": 13.27, "learning_rate": 1.4746478873239437e-05, "loss": 0.7591, "step": 11836 }, { "epoch": 13.27, "learning_rate": 1.4745070422535211e-05, "loss": 0.6153, "step": 11837 }, { "epoch": 13.27, "learning_rate": 1.4743661971830986e-05, "loss": 0.585, "step": 11838 }, { "epoch": 13.27, "learning_rate": 1.474225352112676e-05, "loss": 0.7403, "step": 11839 }, { "epoch": 13.27, "learning_rate": 1.4740845070422535e-05, "loss": 0.5524, "step": 11840 }, { "epoch": 13.27, "learning_rate": 1.473943661971831e-05, "loss": 0.6215, "step": 11841 }, { "epoch": 13.28, "learning_rate": 1.4738028169014086e-05, "loss": 0.6624, "step": 11842 }, { "epoch": 13.28, "learning_rate": 1.473661971830986e-05, "loss": 0.761, "step": 11843 }, { "epoch": 13.28, "learning_rate": 1.4735211267605634e-05, "loss": 0.7091, "step": 11844 }, { "epoch": 13.28, "learning_rate": 1.4733802816901409e-05, "loss": 0.6307, "step": 11845 }, { "epoch": 13.28, "learning_rate": 1.4732394366197183e-05, "loss": 0.6091, "step": 11846 }, { "epoch": 13.28, "learning_rate": 1.4730985915492958e-05, "loss": 0.9184, "step": 11847 }, { "epoch": 13.28, "learning_rate": 1.4729577464788732e-05, "loss": 0.8075, "step": 11848 }, { "epoch": 13.28, "learning_rate": 1.4728169014084509e-05, "loss": 0.7621, "step": 11849 }, { "epoch": 13.28, "learning_rate": 1.4726760563380283e-05, "loss": 0.7707, "step": 11850 }, { "epoch": 13.29, "learning_rate": 1.4725352112676056e-05, "loss": 0.7412, "step": 11851 }, { "epoch": 13.29, "learning_rate": 1.4723943661971831e-05, "loss": 0.7217, "step": 11852 }, { "epoch": 13.29, "learning_rate": 1.4722535211267606e-05, "loss": 0.7056, "step": 11853 }, { "epoch": 13.29, "learning_rate": 1.472112676056338e-05, "loss": 0.8246, "step": 11854 }, { "epoch": 13.29, "learning_rate": 1.4719718309859155e-05, "loss": 0.8363, "step": 11855 }, { "epoch": 13.29, "learning_rate": 1.471830985915493e-05, "loss": 0.7455, "step": 11856 }, { "epoch": 13.29, "learning_rate": 1.4716901408450706e-05, "loss": 0.6221, "step": 11857 }, { "epoch": 13.29, "learning_rate": 1.471549295774648e-05, "loss": 0.6079, "step": 11858 }, { "epoch": 13.29, "learning_rate": 1.4714084507042253e-05, "loss": 0.7407, "step": 11859 }, { "epoch": 13.3, "learning_rate": 1.4712676056338028e-05, "loss": 0.5813, "step": 11860 }, { "epoch": 13.3, "learning_rate": 1.4711267605633803e-05, "loss": 0.6352, "step": 11861 }, { "epoch": 13.3, "learning_rate": 1.4709859154929577e-05, "loss": 0.6792, "step": 11862 }, { "epoch": 13.3, "learning_rate": 1.4708450704225352e-05, "loss": 0.6183, "step": 11863 }, { "epoch": 13.3, "learning_rate": 1.4707042253521127e-05, "loss": 0.6018, "step": 11864 }, { "epoch": 13.3, "learning_rate": 1.4705633802816903e-05, "loss": 0.8029, "step": 11865 }, { "epoch": 13.3, "learning_rate": 1.4704225352112678e-05, "loss": 0.5957, "step": 11866 }, { "epoch": 13.3, "learning_rate": 1.4704225352112678e-05, "loss": 0.7178, "step": 11867 }, { "epoch": 13.3, "learning_rate": 1.470281690140845e-05, "loss": 0.6782, "step": 11868 }, { "epoch": 13.31, "learning_rate": 1.4701408450704225e-05, "loss": 0.7819, "step": 11869 }, { "epoch": 13.31, "learning_rate": 1.47e-05, "loss": 0.6421, "step": 11870 }, { "epoch": 13.31, "learning_rate": 1.4698591549295775e-05, "loss": 0.5811, "step": 11871 }, { "epoch": 13.31, "learning_rate": 1.4697183098591549e-05, "loss": 0.7384, "step": 11872 }, { "epoch": 13.31, "learning_rate": 1.4695774647887325e-05, "loss": 0.5997, "step": 11873 }, { "epoch": 13.31, "learning_rate": 1.46943661971831e-05, "loss": 0.6832, "step": 11874 }, { "epoch": 13.31, "learning_rate": 1.4692957746478875e-05, "loss": 0.6613, "step": 11875 }, { "epoch": 13.31, "learning_rate": 1.4691549295774648e-05, "loss": 0.6793, "step": 11876 }, { "epoch": 13.32, "learning_rate": 1.4690140845070422e-05, "loss": 0.6924, "step": 11877 }, { "epoch": 13.32, "learning_rate": 1.4688732394366197e-05, "loss": 0.5589, "step": 11878 }, { "epoch": 13.32, "learning_rate": 1.4687323943661972e-05, "loss": 0.7051, "step": 11879 }, { "epoch": 13.32, "learning_rate": 1.4685915492957746e-05, "loss": 0.5691, "step": 11880 }, { "epoch": 13.32, "learning_rate": 1.4684507042253523e-05, "loss": 0.6221, "step": 11881 }, { "epoch": 13.32, "learning_rate": 1.4683098591549297e-05, "loss": 0.7862, "step": 11882 }, { "epoch": 13.32, "learning_rate": 1.468169014084507e-05, "loss": 0.6106, "step": 11883 }, { "epoch": 13.32, "learning_rate": 1.4680281690140845e-05, "loss": 0.6673, "step": 11884 }, { "epoch": 13.32, "learning_rate": 1.467887323943662e-05, "loss": 0.7381, "step": 11885 }, { "epoch": 13.33, "learning_rate": 1.4677464788732394e-05, "loss": 0.7564, "step": 11886 }, { "epoch": 13.33, "learning_rate": 1.4676056338028169e-05, "loss": 0.7167, "step": 11887 }, { "epoch": 13.33, "learning_rate": 1.4674647887323945e-05, "loss": 0.6646, "step": 11888 }, { "epoch": 13.33, "learning_rate": 1.467323943661972e-05, "loss": 0.6457, "step": 11889 }, { "epoch": 13.33, "learning_rate": 1.4671830985915494e-05, "loss": 0.5753, "step": 11890 }, { "epoch": 13.33, "learning_rate": 1.4670422535211267e-05, "loss": 0.5379, "step": 11891 }, { "epoch": 13.33, "learning_rate": 1.4669014084507042e-05, "loss": 0.6363, "step": 11892 }, { "epoch": 13.33, "learning_rate": 1.4667605633802817e-05, "loss": 0.3903, "step": 11893 }, { "epoch": 13.33, "learning_rate": 1.4666197183098591e-05, "loss": 0.6084, "step": 11894 }, { "epoch": 13.34, "learning_rate": 1.4664788732394366e-05, "loss": 0.6718, "step": 11895 }, { "epoch": 13.34, "learning_rate": 1.4663380281690142e-05, "loss": 0.7313, "step": 11896 }, { "epoch": 13.34, "learning_rate": 1.4661971830985917e-05, "loss": 0.9594, "step": 11897 }, { "epoch": 13.34, "learning_rate": 1.4660563380281691e-05, "loss": 0.961, "step": 11898 }, { "epoch": 13.34, "learning_rate": 1.4659154929577464e-05, "loss": 0.7398, "step": 11899 }, { "epoch": 13.34, "learning_rate": 1.4657746478873239e-05, "loss": 0.8816, "step": 11900 }, { "epoch": 13.34, "learning_rate": 1.4656338028169014e-05, "loss": 0.8996, "step": 11901 }, { "epoch": 13.34, "learning_rate": 1.4654929577464788e-05, "loss": 0.6993, "step": 11902 }, { "epoch": 13.34, "learning_rate": 1.4653521126760565e-05, "loss": 0.6589, "step": 11903 }, { "epoch": 13.35, "learning_rate": 1.465211267605634e-05, "loss": 0.8083, "step": 11904 }, { "epoch": 13.35, "learning_rate": 1.4650704225352114e-05, "loss": 0.7387, "step": 11905 }, { "epoch": 13.35, "learning_rate": 1.4649295774647889e-05, "loss": 0.7997, "step": 11906 }, { "epoch": 13.35, "learning_rate": 1.4647887323943661e-05, "loss": 0.6123, "step": 11907 }, { "epoch": 13.35, "learning_rate": 1.4646478873239436e-05, "loss": 0.7469, "step": 11908 }, { "epoch": 13.35, "learning_rate": 1.464507042253521e-05, "loss": 0.6587, "step": 11909 }, { "epoch": 13.35, "learning_rate": 1.4643661971830985e-05, "loss": 0.7874, "step": 11910 }, { "epoch": 13.35, "learning_rate": 1.4642253521126762e-05, "loss": 0.6986, "step": 11911 }, { "epoch": 13.35, "learning_rate": 1.4640845070422536e-05, "loss": 0.6669, "step": 11912 }, { "epoch": 13.36, "learning_rate": 1.4639436619718311e-05, "loss": 0.7244, "step": 11913 }, { "epoch": 13.36, "learning_rate": 1.4638028169014086e-05, "loss": 0.7414, "step": 11914 }, { "epoch": 13.36, "learning_rate": 1.4636619718309859e-05, "loss": 0.6959, "step": 11915 }, { "epoch": 13.36, "learning_rate": 1.4635211267605633e-05, "loss": 0.7736, "step": 11916 }, { "epoch": 13.36, "learning_rate": 1.4633802816901408e-05, "loss": 0.7295, "step": 11917 }, { "epoch": 13.36, "learning_rate": 1.4632394366197184e-05, "loss": 0.7399, "step": 11918 }, { "epoch": 13.36, "learning_rate": 1.4630985915492959e-05, "loss": 0.5782, "step": 11919 }, { "epoch": 13.36, "learning_rate": 1.4629577464788733e-05, "loss": 0.7649, "step": 11920 }, { "epoch": 13.36, "learning_rate": 1.4628169014084508e-05, "loss": 0.5584, "step": 11921 }, { "epoch": 13.37, "learning_rate": 1.4626760563380283e-05, "loss": 0.6518, "step": 11922 }, { "epoch": 13.37, "learning_rate": 1.4625352112676056e-05, "loss": 0.7188, "step": 11923 }, { "epoch": 13.37, "learning_rate": 1.462394366197183e-05, "loss": 0.7833, "step": 11924 }, { "epoch": 13.37, "learning_rate": 1.4622535211267605e-05, "loss": 0.6422, "step": 11925 }, { "epoch": 13.37, "learning_rate": 1.4621126760563381e-05, "loss": 0.667, "step": 11926 }, { "epoch": 13.37, "learning_rate": 1.4619718309859156e-05, "loss": 0.6646, "step": 11927 }, { "epoch": 13.37, "learning_rate": 1.461830985915493e-05, "loss": 0.685, "step": 11928 }, { "epoch": 13.37, "learning_rate": 1.4616901408450705e-05, "loss": 0.6184, "step": 11929 }, { "epoch": 13.37, "learning_rate": 1.461549295774648e-05, "loss": 0.7256, "step": 11930 }, { "epoch": 13.38, "learning_rate": 1.4614084507042253e-05, "loss": 0.8022, "step": 11931 }, { "epoch": 13.38, "learning_rate": 1.4612676056338027e-05, "loss": 0.606, "step": 11932 }, { "epoch": 13.38, "learning_rate": 1.4611267605633802e-05, "loss": 0.7567, "step": 11933 }, { "epoch": 13.38, "learning_rate": 1.4609859154929578e-05, "loss": 0.6754, "step": 11934 }, { "epoch": 13.38, "learning_rate": 1.4608450704225353e-05, "loss": 0.7719, "step": 11935 }, { "epoch": 13.38, "learning_rate": 1.4607042253521128e-05, "loss": 0.5514, "step": 11936 }, { "epoch": 13.38, "learning_rate": 1.4605633802816902e-05, "loss": 0.7588, "step": 11937 }, { "epoch": 13.38, "learning_rate": 1.4604225352112677e-05, "loss": 0.6222, "step": 11938 }, { "epoch": 13.38, "learning_rate": 1.460281690140845e-05, "loss": 0.634, "step": 11939 }, { "epoch": 13.39, "learning_rate": 1.4601408450704224e-05, "loss": 0.6435, "step": 11940 }, { "epoch": 13.39, "learning_rate": 1.46e-05, "loss": 0.5776, "step": 11941 }, { "epoch": 13.39, "learning_rate": 1.4598591549295775e-05, "loss": 0.6302, "step": 11942 }, { "epoch": 13.39, "learning_rate": 1.459718309859155e-05, "loss": 0.7973, "step": 11943 }, { "epoch": 13.39, "learning_rate": 1.4595774647887325e-05, "loss": 0.5446, "step": 11944 }, { "epoch": 13.39, "learning_rate": 1.45943661971831e-05, "loss": 0.6146, "step": 11945 }, { "epoch": 13.39, "learning_rate": 1.4592957746478874e-05, "loss": 0.8341, "step": 11946 }, { "epoch": 13.39, "learning_rate": 1.4591549295774647e-05, "loss": 0.9628, "step": 11947 }, { "epoch": 13.39, "learning_rate": 1.4590140845070422e-05, "loss": 0.896, "step": 11948 }, { "epoch": 13.4, "learning_rate": 1.4588732394366198e-05, "loss": 0.9155, "step": 11949 }, { "epoch": 13.4, "learning_rate": 1.4587323943661973e-05, "loss": 0.7403, "step": 11950 }, { "epoch": 13.4, "learning_rate": 1.4585915492957747e-05, "loss": 0.7638, "step": 11951 }, { "epoch": 13.4, "learning_rate": 1.4584507042253522e-05, "loss": 0.658, "step": 11952 }, { "epoch": 13.4, "learning_rate": 1.4583098591549296e-05, "loss": 0.7396, "step": 11953 }, { "epoch": 13.4, "learning_rate": 1.4581690140845071e-05, "loss": 0.7692, "step": 11954 }, { "epoch": 13.4, "learning_rate": 1.4580281690140844e-05, "loss": 0.7568, "step": 11955 }, { "epoch": 13.4, "learning_rate": 1.457887323943662e-05, "loss": 0.7044, "step": 11956 }, { "epoch": 13.4, "learning_rate": 1.4577464788732395e-05, "loss": 0.7662, "step": 11957 }, { "epoch": 13.41, "learning_rate": 1.457605633802817e-05, "loss": 0.6828, "step": 11958 }, { "epoch": 13.41, "learning_rate": 1.4574647887323944e-05, "loss": 0.6893, "step": 11959 }, { "epoch": 13.41, "learning_rate": 1.4573239436619719e-05, "loss": 0.6377, "step": 11960 }, { "epoch": 13.41, "learning_rate": 1.4571830985915494e-05, "loss": 0.6858, "step": 11961 }, { "epoch": 13.41, "learning_rate": 1.4570422535211268e-05, "loss": 0.6317, "step": 11962 }, { "epoch": 13.41, "learning_rate": 1.4569014084507041e-05, "loss": 0.6325, "step": 11963 }, { "epoch": 13.41, "learning_rate": 1.4567605633802817e-05, "loss": 0.7072, "step": 11964 }, { "epoch": 13.41, "learning_rate": 1.4566197183098592e-05, "loss": 0.7232, "step": 11965 }, { "epoch": 13.41, "learning_rate": 1.4564788732394367e-05, "loss": 0.7114, "step": 11966 }, { "epoch": 13.42, "learning_rate": 1.4563380281690141e-05, "loss": 0.7423, "step": 11967 }, { "epoch": 13.42, "learning_rate": 1.4561971830985916e-05, "loss": 0.6399, "step": 11968 }, { "epoch": 13.42, "learning_rate": 1.456056338028169e-05, "loss": 0.7359, "step": 11969 }, { "epoch": 13.42, "learning_rate": 1.4559154929577465e-05, "loss": 0.691, "step": 11970 }, { "epoch": 13.42, "learning_rate": 1.455774647887324e-05, "loss": 0.6626, "step": 11971 }, { "epoch": 13.42, "learning_rate": 1.4556338028169015e-05, "loss": 0.7447, "step": 11972 }, { "epoch": 13.42, "learning_rate": 1.455492957746479e-05, "loss": 0.6626, "step": 11973 }, { "epoch": 13.42, "learning_rate": 1.4553521126760564e-05, "loss": 0.6365, "step": 11974 }, { "epoch": 13.42, "learning_rate": 1.4552112676056338e-05, "loss": 0.751, "step": 11975 }, { "epoch": 13.43, "learning_rate": 1.4550704225352113e-05, "loss": 0.7671, "step": 11976 }, { "epoch": 13.43, "learning_rate": 1.4549295774647888e-05, "loss": 0.7513, "step": 11977 }, { "epoch": 13.43, "learning_rate": 1.4547887323943662e-05, "loss": 0.5702, "step": 11978 }, { "epoch": 13.43, "learning_rate": 1.4546478873239437e-05, "loss": 0.7172, "step": 11979 }, { "epoch": 13.43, "learning_rate": 1.4545070422535212e-05, "loss": 0.7677, "step": 11980 }, { "epoch": 13.43, "learning_rate": 1.4543661971830986e-05, "loss": 0.8823, "step": 11981 }, { "epoch": 13.43, "learning_rate": 1.4542253521126761e-05, "loss": 0.8951, "step": 11982 }, { "epoch": 13.43, "learning_rate": 1.4540845070422536e-05, "loss": 0.6784, "step": 11983 }, { "epoch": 13.43, "learning_rate": 1.453943661971831e-05, "loss": 0.7584, "step": 11984 }, { "epoch": 13.44, "learning_rate": 1.4538028169014085e-05, "loss": 0.5977, "step": 11985 }, { "epoch": 13.44, "learning_rate": 1.453661971830986e-05, "loss": 0.5884, "step": 11986 }, { "epoch": 13.44, "learning_rate": 1.4535211267605634e-05, "loss": 0.6794, "step": 11987 }, { "epoch": 13.44, "learning_rate": 1.4533802816901409e-05, "loss": 0.5685, "step": 11988 }, { "epoch": 13.44, "learning_rate": 1.4532394366197183e-05, "loss": 0.575, "step": 11989 }, { "epoch": 13.44, "learning_rate": 1.4530985915492958e-05, "loss": 0.6181, "step": 11990 }, { "epoch": 13.44, "learning_rate": 1.4529577464788733e-05, "loss": 0.4944, "step": 11991 }, { "epoch": 13.44, "learning_rate": 1.4528169014084507e-05, "loss": 0.5853, "step": 11992 }, { "epoch": 13.45, "learning_rate": 1.4526760563380282e-05, "loss": 0.6816, "step": 11993 }, { "epoch": 13.45, "learning_rate": 1.4525352112676057e-05, "loss": 0.5794, "step": 11994 }, { "epoch": 13.45, "learning_rate": 1.4523943661971831e-05, "loss": 0.4912, "step": 11995 }, { "epoch": 13.45, "learning_rate": 1.4522535211267606e-05, "loss": 0.6423, "step": 11996 }, { "epoch": 13.45, "learning_rate": 1.452112676056338e-05, "loss": 0.9639, "step": 11997 }, { "epoch": 13.45, "learning_rate": 1.4519718309859155e-05, "loss": 0.8405, "step": 11998 }, { "epoch": 13.45, "learning_rate": 1.451830985915493e-05, "loss": 0.651, "step": 11999 }, { "epoch": 13.45, "learning_rate": 1.4516901408450704e-05, "loss": 0.8868, "step": 12000 }, { "epoch": 13.45, "eval_loss": 1.0145512819290161, "eval_runtime": 826.0099, "eval_samples_per_second": 3.199, "eval_steps_per_second": 0.401, "eval_wer": 0.17021023403411345, "step": 12000 }, { "epoch": 13.45, "learning_rate": 1.4515492957746479e-05, "loss": 0.7019, "step": 12001 }, { "epoch": 13.46, "learning_rate": 1.4514084507042254e-05, "loss": 0.8406, "step": 12002 }, { "epoch": 13.46, "learning_rate": 1.4512676056338028e-05, "loss": 0.6781, "step": 12003 }, { "epoch": 13.46, "learning_rate": 1.4511267605633803e-05, "loss": 0.933, "step": 12004 }, { "epoch": 13.46, "learning_rate": 1.4509859154929578e-05, "loss": 0.6464, "step": 12005 }, { "epoch": 13.46, "learning_rate": 1.4508450704225352e-05, "loss": 0.6523, "step": 12006 }, { "epoch": 13.46, "learning_rate": 1.4507042253521127e-05, "loss": 0.6889, "step": 12007 }, { "epoch": 13.46, "learning_rate": 1.4505633802816902e-05, "loss": 0.7997, "step": 12008 }, { "epoch": 13.46, "learning_rate": 1.4504225352112678e-05, "loss": 0.8298, "step": 12009 }, { "epoch": 13.46, "learning_rate": 1.450281690140845e-05, "loss": 0.8217, "step": 12010 }, { "epoch": 13.47, "learning_rate": 1.4501408450704225e-05, "loss": 0.8152, "step": 12011 }, { "epoch": 13.47, "learning_rate": 1.45e-05, "loss": 0.6685, "step": 12012 }, { "epoch": 13.47, "learning_rate": 1.4498591549295775e-05, "loss": 0.7641, "step": 12013 }, { "epoch": 13.47, "learning_rate": 1.449718309859155e-05, "loss": 0.8005, "step": 12014 }, { "epoch": 13.47, "learning_rate": 1.4495774647887324e-05, "loss": 0.7112, "step": 12015 }, { "epoch": 13.47, "learning_rate": 1.4494366197183099e-05, "loss": 0.6148, "step": 12016 }, { "epoch": 13.47, "learning_rate": 1.4492957746478875e-05, "loss": 0.6922, "step": 12017 }, { "epoch": 13.47, "learning_rate": 1.4491549295774648e-05, "loss": 0.7378, "step": 12018 }, { "epoch": 13.47, "learning_rate": 1.4490140845070423e-05, "loss": 0.6869, "step": 12019 }, { "epoch": 13.48, "learning_rate": 1.4488732394366197e-05, "loss": 0.7366, "step": 12020 }, { "epoch": 13.48, "learning_rate": 1.4487323943661972e-05, "loss": 0.6211, "step": 12021 }, { "epoch": 13.48, "learning_rate": 1.4485915492957746e-05, "loss": 0.6553, "step": 12022 }, { "epoch": 13.48, "learning_rate": 1.4484507042253521e-05, "loss": 0.6648, "step": 12023 }, { "epoch": 13.48, "learning_rate": 1.4483098591549297e-05, "loss": 0.6706, "step": 12024 }, { "epoch": 13.48, "learning_rate": 1.4481690140845072e-05, "loss": 0.5625, "step": 12025 }, { "epoch": 13.48, "learning_rate": 1.4480281690140845e-05, "loss": 0.6687, "step": 12026 }, { "epoch": 13.48, "learning_rate": 1.447887323943662e-05, "loss": 0.6794, "step": 12027 }, { "epoch": 13.48, "learning_rate": 1.4477464788732394e-05, "loss": 0.6008, "step": 12028 }, { "epoch": 13.49, "learning_rate": 1.4476056338028169e-05, "loss": 0.6834, "step": 12029 }, { "epoch": 13.49, "learning_rate": 1.4474647887323944e-05, "loss": 0.5117, "step": 12030 }, { "epoch": 13.49, "learning_rate": 1.4473239436619718e-05, "loss": 0.6316, "step": 12031 }, { "epoch": 13.49, "learning_rate": 1.4471830985915495e-05, "loss": 0.6711, "step": 12032 }, { "epoch": 13.49, "learning_rate": 1.447042253521127e-05, "loss": 0.5636, "step": 12033 }, { "epoch": 13.49, "learning_rate": 1.4469014084507042e-05, "loss": 0.5828, "step": 12034 }, { "epoch": 13.49, "learning_rate": 1.4467605633802817e-05, "loss": 0.6485, "step": 12035 }, { "epoch": 13.49, "learning_rate": 1.4466197183098591e-05, "loss": 0.7383, "step": 12036 }, { "epoch": 13.49, "learning_rate": 1.4464788732394366e-05, "loss": 0.5386, "step": 12037 }, { "epoch": 13.5, "learning_rate": 1.446338028169014e-05, "loss": 0.8553, "step": 12038 }, { "epoch": 13.5, "learning_rate": 1.4461971830985917e-05, "loss": 0.7808, "step": 12039 }, { "epoch": 13.5, "learning_rate": 1.4460563380281692e-05, "loss": 0.7164, "step": 12040 }, { "epoch": 13.5, "learning_rate": 1.4459154929577466e-05, "loss": 0.8085, "step": 12041 }, { "epoch": 13.5, "learning_rate": 1.445774647887324e-05, "loss": 0.4867, "step": 12042 }, { "epoch": 13.5, "learning_rate": 1.4456338028169014e-05, "loss": 0.6285, "step": 12043 }, { "epoch": 13.5, "learning_rate": 1.4454929577464788e-05, "loss": 0.378, "step": 12044 }, { "epoch": 13.5, "learning_rate": 1.4453521126760563e-05, "loss": 0.6348, "step": 12045 }, { "epoch": 13.5, "learning_rate": 1.4452112676056338e-05, "loss": 0.8374, "step": 12046 }, { "epoch": 13.51, "learning_rate": 1.4450704225352114e-05, "loss": 0.8675, "step": 12047 }, { "epoch": 13.51, "learning_rate": 1.4449295774647889e-05, "loss": 0.9097, "step": 12048 }, { "epoch": 13.51, "learning_rate": 1.4447887323943663e-05, "loss": 0.7513, "step": 12049 }, { "epoch": 13.51, "learning_rate": 1.4446478873239436e-05, "loss": 0.7577, "step": 12050 }, { "epoch": 13.51, "learning_rate": 1.4445070422535211e-05, "loss": 0.665, "step": 12051 }, { "epoch": 13.51, "learning_rate": 1.4443661971830986e-05, "loss": 0.7194, "step": 12052 }, { "epoch": 13.51, "learning_rate": 1.444225352112676e-05, "loss": 0.8361, "step": 12053 }, { "epoch": 13.51, "learning_rate": 1.4440845070422535e-05, "loss": 0.6569, "step": 12054 }, { "epoch": 13.51, "learning_rate": 1.4439436619718311e-05, "loss": 0.9261, "step": 12055 }, { "epoch": 13.52, "learning_rate": 1.4438028169014086e-05, "loss": 0.627, "step": 12056 }, { "epoch": 13.52, "learning_rate": 1.443661971830986e-05, "loss": 0.6352, "step": 12057 }, { "epoch": 13.52, "learning_rate": 1.4435211267605633e-05, "loss": 0.6591, "step": 12058 }, { "epoch": 13.52, "learning_rate": 1.4433802816901408e-05, "loss": 0.7531, "step": 12059 }, { "epoch": 13.52, "learning_rate": 1.4432394366197183e-05, "loss": 0.6842, "step": 12060 }, { "epoch": 13.52, "learning_rate": 1.4430985915492957e-05, "loss": 0.7498, "step": 12061 }, { "epoch": 13.52, "learning_rate": 1.4429577464788734e-05, "loss": 0.7378, "step": 12062 }, { "epoch": 13.52, "learning_rate": 1.4428169014084508e-05, "loss": 0.7218, "step": 12063 }, { "epoch": 13.52, "learning_rate": 1.4426760563380283e-05, "loss": 0.6433, "step": 12064 }, { "epoch": 13.53, "learning_rate": 1.4425352112676058e-05, "loss": 0.6298, "step": 12065 }, { "epoch": 13.53, "learning_rate": 1.442394366197183e-05, "loss": 0.7171, "step": 12066 }, { "epoch": 13.53, "learning_rate": 1.4422535211267605e-05, "loss": 0.5058, "step": 12067 }, { "epoch": 13.53, "learning_rate": 1.442112676056338e-05, "loss": 0.7551, "step": 12068 }, { "epoch": 13.53, "learning_rate": 1.4419718309859154e-05, "loss": 0.674, "step": 12069 }, { "epoch": 13.53, "learning_rate": 1.441830985915493e-05, "loss": 0.5714, "step": 12070 }, { "epoch": 13.53, "learning_rate": 1.4416901408450705e-05, "loss": 0.7937, "step": 12071 }, { "epoch": 13.53, "learning_rate": 1.441549295774648e-05, "loss": 0.6399, "step": 12072 }, { "epoch": 13.53, "learning_rate": 1.4414084507042255e-05, "loss": 0.5704, "step": 12073 }, { "epoch": 13.54, "learning_rate": 1.4412676056338028e-05, "loss": 0.7932, "step": 12074 }, { "epoch": 13.54, "learning_rate": 1.4411267605633802e-05, "loss": 0.6284, "step": 12075 }, { "epoch": 13.54, "learning_rate": 1.4409859154929577e-05, "loss": 0.6414, "step": 12076 }, { "epoch": 13.54, "learning_rate": 1.4408450704225353e-05, "loss": 0.8448, "step": 12077 }, { "epoch": 13.54, "learning_rate": 1.4407042253521128e-05, "loss": 0.7238, "step": 12078 }, { "epoch": 13.54, "learning_rate": 1.4405633802816902e-05, "loss": 0.576, "step": 12079 }, { "epoch": 13.54, "learning_rate": 1.4404225352112677e-05, "loss": 0.7724, "step": 12080 }, { "epoch": 13.54, "learning_rate": 1.4402816901408452e-05, "loss": 0.723, "step": 12081 }, { "epoch": 13.54, "learning_rate": 1.4401408450704225e-05, "loss": 0.6179, "step": 12082 }, { "epoch": 13.55, "learning_rate": 1.44e-05, "loss": 0.7986, "step": 12083 }, { "epoch": 13.55, "learning_rate": 1.4398591549295774e-05, "loss": 0.7032, "step": 12084 }, { "epoch": 13.55, "learning_rate": 1.439718309859155e-05, "loss": 0.6182, "step": 12085 }, { "epoch": 13.55, "learning_rate": 1.4395774647887325e-05, "loss": 0.7063, "step": 12086 }, { "epoch": 13.55, "learning_rate": 1.43943661971831e-05, "loss": 0.5643, "step": 12087 }, { "epoch": 13.55, "learning_rate": 1.4392957746478874e-05, "loss": 0.6227, "step": 12088 }, { "epoch": 13.55, "learning_rate": 1.4391549295774649e-05, "loss": 0.7943, "step": 12089 }, { "epoch": 13.55, "learning_rate": 1.4390140845070422e-05, "loss": 0.627, "step": 12090 }, { "epoch": 13.55, "learning_rate": 1.4388732394366196e-05, "loss": 0.7216, "step": 12091 }, { "epoch": 13.56, "learning_rate": 1.4387323943661973e-05, "loss": 0.7418, "step": 12092 }, { "epoch": 13.56, "learning_rate": 1.4385915492957747e-05, "loss": 0.7668, "step": 12093 }, { "epoch": 13.56, "learning_rate": 1.4384507042253522e-05, "loss": 0.5656, "step": 12094 }, { "epoch": 13.56, "learning_rate": 1.4383098591549297e-05, "loss": 0.672, "step": 12095 }, { "epoch": 13.56, "learning_rate": 1.4381690140845071e-05, "loss": 0.4868, "step": 12096 }, { "epoch": 13.56, "learning_rate": 1.4380281690140846e-05, "loss": 0.8572, "step": 12097 }, { "epoch": 13.56, "learning_rate": 1.4378873239436619e-05, "loss": 0.7311, "step": 12098 }, { "epoch": 13.56, "learning_rate": 1.4377464788732394e-05, "loss": 0.8988, "step": 12099 }, { "epoch": 13.57, "learning_rate": 1.437605633802817e-05, "loss": 0.9622, "step": 12100 }, { "epoch": 13.57, "learning_rate": 1.4374647887323945e-05, "loss": 0.8619, "step": 12101 }, { "epoch": 13.57, "learning_rate": 1.437323943661972e-05, "loss": 0.9755, "step": 12102 }, { "epoch": 13.57, "learning_rate": 1.4371830985915494e-05, "loss": 0.6066, "step": 12103 }, { "epoch": 13.57, "learning_rate": 1.4370422535211268e-05, "loss": 0.7329, "step": 12104 }, { "epoch": 13.57, "learning_rate": 1.4369014084507043e-05, "loss": 0.7024, "step": 12105 }, { "epoch": 13.57, "learning_rate": 1.4367605633802816e-05, "loss": 0.6623, "step": 12106 }, { "epoch": 13.57, "learning_rate": 1.436619718309859e-05, "loss": 0.7741, "step": 12107 }, { "epoch": 13.57, "learning_rate": 1.4364788732394367e-05, "loss": 0.9564, "step": 12108 }, { "epoch": 13.58, "learning_rate": 1.4363380281690142e-05, "loss": 0.7825, "step": 12109 }, { "epoch": 13.58, "learning_rate": 1.4361971830985916e-05, "loss": 0.6337, "step": 12110 }, { "epoch": 13.58, "learning_rate": 1.4360563380281691e-05, "loss": 0.6816, "step": 12111 }, { "epoch": 13.58, "learning_rate": 1.4359154929577466e-05, "loss": 0.715, "step": 12112 }, { "epoch": 13.58, "learning_rate": 1.435774647887324e-05, "loss": 0.6384, "step": 12113 }, { "epoch": 13.58, "learning_rate": 1.4356338028169013e-05, "loss": 0.8415, "step": 12114 }, { "epoch": 13.58, "learning_rate": 1.435492957746479e-05, "loss": 0.634, "step": 12115 }, { "epoch": 13.58, "learning_rate": 1.4353521126760564e-05, "loss": 0.6285, "step": 12116 }, { "epoch": 13.58, "learning_rate": 1.4352112676056339e-05, "loss": 0.6098, "step": 12117 }, { "epoch": 13.59, "learning_rate": 1.4350704225352113e-05, "loss": 0.6617, "step": 12118 }, { "epoch": 13.59, "learning_rate": 1.4349295774647888e-05, "loss": 0.6106, "step": 12119 }, { "epoch": 13.59, "learning_rate": 1.4347887323943663e-05, "loss": 0.6433, "step": 12120 }, { "epoch": 13.59, "learning_rate": 1.4346478873239436e-05, "loss": 0.7151, "step": 12121 }, { "epoch": 13.59, "learning_rate": 1.434507042253521e-05, "loss": 0.5997, "step": 12122 }, { "epoch": 13.59, "learning_rate": 1.4343661971830987e-05, "loss": 0.6997, "step": 12123 }, { "epoch": 13.59, "learning_rate": 1.4342253521126761e-05, "loss": 0.5957, "step": 12124 }, { "epoch": 13.59, "learning_rate": 1.4340845070422536e-05, "loss": 0.4919, "step": 12125 }, { "epoch": 13.59, "learning_rate": 1.433943661971831e-05, "loss": 0.5733, "step": 12126 }, { "epoch": 13.6, "learning_rate": 1.4338028169014085e-05, "loss": 0.6545, "step": 12127 }, { "epoch": 13.6, "learning_rate": 1.433661971830986e-05, "loss": 0.5525, "step": 12128 }, { "epoch": 13.6, "learning_rate": 1.4335211267605633e-05, "loss": 0.7629, "step": 12129 }, { "epoch": 13.6, "learning_rate": 1.4333802816901409e-05, "loss": 0.7245, "step": 12130 }, { "epoch": 13.6, "learning_rate": 1.4332394366197184e-05, "loss": 0.7422, "step": 12131 }, { "epoch": 13.6, "learning_rate": 1.4330985915492958e-05, "loss": 0.6119, "step": 12132 }, { "epoch": 13.6, "learning_rate": 1.4329577464788733e-05, "loss": 0.5696, "step": 12133 }, { "epoch": 13.6, "learning_rate": 1.4328169014084508e-05, "loss": 0.6643, "step": 12134 }, { "epoch": 13.6, "learning_rate": 1.4326760563380282e-05, "loss": 0.6259, "step": 12135 }, { "epoch": 13.61, "learning_rate": 1.4325352112676057e-05, "loss": 0.6068, "step": 12136 }, { "epoch": 13.61, "learning_rate": 1.432394366197183e-05, "loss": 0.6476, "step": 12137 }, { "epoch": 13.61, "learning_rate": 1.4322535211267606e-05, "loss": 0.6679, "step": 12138 }, { "epoch": 13.61, "learning_rate": 1.432112676056338e-05, "loss": 0.5905, "step": 12139 }, { "epoch": 13.61, "learning_rate": 1.4319718309859155e-05, "loss": 0.7597, "step": 12140 }, { "epoch": 13.61, "learning_rate": 1.431830985915493e-05, "loss": 0.791, "step": 12141 }, { "epoch": 13.61, "learning_rate": 1.4316901408450705e-05, "loss": 0.7822, "step": 12142 }, { "epoch": 13.61, "learning_rate": 1.431549295774648e-05, "loss": 0.6817, "step": 12143 }, { "epoch": 13.61, "learning_rate": 1.4314084507042254e-05, "loss": 0.7921, "step": 12144 }, { "epoch": 13.62, "learning_rate": 1.4312676056338029e-05, "loss": 0.6854, "step": 12145 }, { "epoch": 13.62, "learning_rate": 1.4311267605633803e-05, "loss": 0.333, "step": 12146 }, { "epoch": 13.62, "learning_rate": 1.4309859154929578e-05, "loss": 0.9569, "step": 12147 }, { "epoch": 13.62, "learning_rate": 1.4308450704225352e-05, "loss": 0.7784, "step": 12148 }, { "epoch": 13.62, "learning_rate": 1.4307042253521127e-05, "loss": 0.8015, "step": 12149 }, { "epoch": 13.62, "learning_rate": 1.4305633802816902e-05, "loss": 0.6914, "step": 12150 }, { "epoch": 13.62, "learning_rate": 1.4304225352112676e-05, "loss": 0.5261, "step": 12151 }, { "epoch": 13.62, "learning_rate": 1.4302816901408451e-05, "loss": 0.6088, "step": 12152 }, { "epoch": 13.62, "learning_rate": 1.4301408450704226e-05, "loss": 0.7553, "step": 12153 }, { "epoch": 13.63, "learning_rate": 1.43e-05, "loss": 0.7268, "step": 12154 }, { "epoch": 13.63, "learning_rate": 1.4298591549295775e-05, "loss": 0.829, "step": 12155 }, { "epoch": 13.63, "learning_rate": 1.429718309859155e-05, "loss": 0.6262, "step": 12156 }, { "epoch": 13.63, "learning_rate": 1.4295774647887324e-05, "loss": 0.6386, "step": 12157 }, { "epoch": 13.63, "learning_rate": 1.4294366197183099e-05, "loss": 0.737, "step": 12158 }, { "epoch": 13.63, "learning_rate": 1.4292957746478873e-05, "loss": 0.6442, "step": 12159 }, { "epoch": 13.63, "learning_rate": 1.4291549295774648e-05, "loss": 0.6396, "step": 12160 }, { "epoch": 13.63, "learning_rate": 1.4290140845070423e-05, "loss": 0.7302, "step": 12161 }, { "epoch": 13.63, "learning_rate": 1.4288732394366197e-05, "loss": 0.6703, "step": 12162 }, { "epoch": 13.64, "learning_rate": 1.4287323943661972e-05, "loss": 0.7592, "step": 12163 }, { "epoch": 13.64, "learning_rate": 1.4285915492957747e-05, "loss": 0.7196, "step": 12164 }, { "epoch": 13.64, "learning_rate": 1.4284507042253521e-05, "loss": 0.5759, "step": 12165 }, { "epoch": 13.64, "learning_rate": 1.4283098591549296e-05, "loss": 0.6631, "step": 12166 }, { "epoch": 13.64, "learning_rate": 1.428169014084507e-05, "loss": 0.6055, "step": 12167 }, { "epoch": 13.64, "learning_rate": 1.4280281690140847e-05, "loss": 0.6052, "step": 12168 }, { "epoch": 13.64, "learning_rate": 1.427887323943662e-05, "loss": 0.7054, "step": 12169 }, { "epoch": 13.64, "learning_rate": 1.4277464788732395e-05, "loss": 0.6488, "step": 12170 }, { "epoch": 13.64, "learning_rate": 1.4276056338028169e-05, "loss": 0.7753, "step": 12171 }, { "epoch": 13.65, "learning_rate": 1.4274647887323944e-05, "loss": 0.7975, "step": 12172 }, { "epoch": 13.65, "learning_rate": 1.4273239436619718e-05, "loss": 0.74, "step": 12173 }, { "epoch": 13.65, "learning_rate": 1.4271830985915493e-05, "loss": 0.6012, "step": 12174 }, { "epoch": 13.65, "learning_rate": 1.4270422535211268e-05, "loss": 0.7859, "step": 12175 }, { "epoch": 13.65, "learning_rate": 1.4269014084507044e-05, "loss": 0.5916, "step": 12176 }, { "epoch": 13.65, "learning_rate": 1.4267605633802817e-05, "loss": 0.6396, "step": 12177 }, { "epoch": 13.65, "learning_rate": 1.4266197183098592e-05, "loss": 0.5726, "step": 12178 }, { "epoch": 13.65, "learning_rate": 1.4264788732394366e-05, "loss": 0.5104, "step": 12179 }, { "epoch": 13.65, "learning_rate": 1.4263380281690141e-05, "loss": 0.7348, "step": 12180 }, { "epoch": 13.66, "learning_rate": 1.4261971830985916e-05, "loss": 0.6298, "step": 12181 }, { "epoch": 13.66, "learning_rate": 1.426056338028169e-05, "loss": 0.5422, "step": 12182 }, { "epoch": 13.66, "learning_rate": 1.4259154929577466e-05, "loss": 0.744, "step": 12183 }, { "epoch": 13.66, "learning_rate": 1.4257746478873241e-05, "loss": 0.5064, "step": 12184 }, { "epoch": 13.66, "learning_rate": 1.4256338028169014e-05, "loss": 0.5912, "step": 12185 }, { "epoch": 13.66, "learning_rate": 1.4254929577464789e-05, "loss": 0.7114, "step": 12186 }, { "epoch": 13.66, "learning_rate": 1.4253521126760563e-05, "loss": 0.5821, "step": 12187 }, { "epoch": 13.66, "learning_rate": 1.4252112676056338e-05, "loss": 0.7199, "step": 12188 }, { "epoch": 13.66, "learning_rate": 1.4250704225352113e-05, "loss": 0.5388, "step": 12189 }, { "epoch": 13.67, "learning_rate": 1.4249295774647887e-05, "loss": 0.6558, "step": 12190 }, { "epoch": 13.67, "learning_rate": 1.4247887323943664e-05, "loss": 0.6531, "step": 12191 }, { "epoch": 13.67, "learning_rate": 1.4246478873239438e-05, "loss": 0.6192, "step": 12192 }, { "epoch": 13.67, "learning_rate": 1.4245070422535211e-05, "loss": 0.5077, "step": 12193 }, { "epoch": 13.67, "learning_rate": 1.4243661971830986e-05, "loss": 0.7721, "step": 12194 }, { "epoch": 13.67, "learning_rate": 1.424225352112676e-05, "loss": 0.5629, "step": 12195 }, { "epoch": 13.67, "learning_rate": 1.4240845070422535e-05, "loss": 0.6726, "step": 12196 }, { "epoch": 13.67, "learning_rate": 1.423943661971831e-05, "loss": 0.8864, "step": 12197 }, { "epoch": 13.67, "learning_rate": 1.4238028169014086e-05, "loss": 0.7061, "step": 12198 }, { "epoch": 13.68, "learning_rate": 1.423661971830986e-05, "loss": 0.9173, "step": 12199 }, { "epoch": 13.68, "learning_rate": 1.4235211267605635e-05, "loss": 0.6201, "step": 12200 }, { "epoch": 13.68, "learning_rate": 1.4233802816901408e-05, "loss": 0.682, "step": 12201 }, { "epoch": 13.68, "learning_rate": 1.4232394366197183e-05, "loss": 0.7305, "step": 12202 }, { "epoch": 13.68, "learning_rate": 1.4230985915492958e-05, "loss": 0.6263, "step": 12203 }, { "epoch": 13.68, "learning_rate": 1.4229577464788732e-05, "loss": 0.6773, "step": 12204 }, { "epoch": 13.68, "learning_rate": 1.4228169014084507e-05, "loss": 0.6675, "step": 12205 }, { "epoch": 13.68, "learning_rate": 1.4226760563380283e-05, "loss": 0.7773, "step": 12206 }, { "epoch": 13.68, "learning_rate": 1.4225352112676058e-05, "loss": 0.5784, "step": 12207 }, { "epoch": 13.69, "learning_rate": 1.4223943661971832e-05, "loss": 0.7641, "step": 12208 }, { "epoch": 13.69, "learning_rate": 1.4222535211267605e-05, "loss": 0.7238, "step": 12209 }, { "epoch": 13.69, "learning_rate": 1.422112676056338e-05, "loss": 0.5508, "step": 12210 }, { "epoch": 13.69, "learning_rate": 1.4219718309859155e-05, "loss": 0.6343, "step": 12211 }, { "epoch": 13.69, "learning_rate": 1.421830985915493e-05, "loss": 0.6143, "step": 12212 }, { "epoch": 13.69, "learning_rate": 1.4216901408450706e-05, "loss": 0.5918, "step": 12213 }, { "epoch": 13.69, "learning_rate": 1.421549295774648e-05, "loss": 0.5987, "step": 12214 }, { "epoch": 13.69, "learning_rate": 1.4214084507042255e-05, "loss": 0.6132, "step": 12215 }, { "epoch": 13.7, "learning_rate": 1.421267605633803e-05, "loss": 0.7537, "step": 12216 }, { "epoch": 13.7, "learning_rate": 1.4211267605633802e-05, "loss": 0.5008, "step": 12217 }, { "epoch": 13.7, "learning_rate": 1.4209859154929577e-05, "loss": 0.77, "step": 12218 }, { "epoch": 13.7, "learning_rate": 1.4208450704225352e-05, "loss": 0.7112, "step": 12219 }, { "epoch": 13.7, "learning_rate": 1.4207042253521126e-05, "loss": 0.8356, "step": 12220 }, { "epoch": 13.7, "learning_rate": 1.4205633802816903e-05, "loss": 0.5416, "step": 12221 }, { "epoch": 13.7, "learning_rate": 1.4204225352112677e-05, "loss": 0.7268, "step": 12222 }, { "epoch": 13.7, "learning_rate": 1.4202816901408452e-05, "loss": 0.7012, "step": 12223 }, { "epoch": 13.7, "learning_rate": 1.4201408450704225e-05, "loss": 0.6953, "step": 12224 }, { "epoch": 13.71, "learning_rate": 1.42e-05, "loss": 0.5226, "step": 12225 }, { "epoch": 13.71, "learning_rate": 1.4198591549295774e-05, "loss": 0.526, "step": 12226 }, { "epoch": 13.71, "learning_rate": 1.4197183098591549e-05, "loss": 0.795, "step": 12227 }, { "epoch": 13.71, "learning_rate": 1.4195774647887323e-05, "loss": 0.5531, "step": 12228 }, { "epoch": 13.71, "learning_rate": 1.41943661971831e-05, "loss": 0.7335, "step": 12229 }, { "epoch": 13.71, "learning_rate": 1.4192957746478874e-05, "loss": 0.6334, "step": 12230 }, { "epoch": 13.71, "learning_rate": 1.4191549295774649e-05, "loss": 0.6392, "step": 12231 }, { "epoch": 13.71, "learning_rate": 1.4190140845070422e-05, "loss": 0.6616, "step": 12232 }, { "epoch": 13.71, "learning_rate": 1.4188732394366197e-05, "loss": 0.7378, "step": 12233 }, { "epoch": 13.72, "learning_rate": 1.4187323943661971e-05, "loss": 0.6209, "step": 12234 }, { "epoch": 13.72, "learning_rate": 1.4185915492957746e-05, "loss": 0.5808, "step": 12235 }, { "epoch": 13.72, "learning_rate": 1.4184507042253522e-05, "loss": 0.7357, "step": 12236 }, { "epoch": 13.72, "learning_rate": 1.4183098591549297e-05, "loss": 0.689, "step": 12237 }, { "epoch": 13.72, "learning_rate": 1.4181690140845072e-05, "loss": 0.7972, "step": 12238 }, { "epoch": 13.72, "learning_rate": 1.4180281690140846e-05, "loss": 0.6814, "step": 12239 }, { "epoch": 13.72, "learning_rate": 1.4178873239436619e-05, "loss": 0.704, "step": 12240 }, { "epoch": 13.72, "learning_rate": 1.4177464788732394e-05, "loss": 0.6543, "step": 12241 }, { "epoch": 13.72, "learning_rate": 1.4176056338028168e-05, "loss": 0.5236, "step": 12242 }, { "epoch": 13.73, "learning_rate": 1.4174647887323943e-05, "loss": 0.3702, "step": 12243 }, { "epoch": 13.73, "learning_rate": 1.417323943661972e-05, "loss": 0.7338, "step": 12244 }, { "epoch": 13.73, "learning_rate": 1.4171830985915494e-05, "loss": 0.6458, "step": 12245 }, { "epoch": 13.73, "learning_rate": 1.4170422535211269e-05, "loss": 0.7927, "step": 12246 }, { "epoch": 13.73, "learning_rate": 1.4169014084507043e-05, "loss": 0.8594, "step": 12247 }, { "epoch": 13.73, "learning_rate": 1.4167605633802816e-05, "loss": 0.8043, "step": 12248 }, { "epoch": 13.73, "learning_rate": 1.4166197183098591e-05, "loss": 0.8903, "step": 12249 }, { "epoch": 13.73, "learning_rate": 1.4164788732394366e-05, "loss": 0.7557, "step": 12250 }, { "epoch": 13.73, "learning_rate": 1.4163380281690142e-05, "loss": 0.7218, "step": 12251 }, { "epoch": 13.74, "learning_rate": 1.4161971830985916e-05, "loss": 0.7133, "step": 12252 }, { "epoch": 13.74, "learning_rate": 1.4160563380281691e-05, "loss": 0.7519, "step": 12253 }, { "epoch": 13.74, "learning_rate": 1.4159154929577466e-05, "loss": 0.7662, "step": 12254 }, { "epoch": 13.74, "learning_rate": 1.415774647887324e-05, "loss": 0.6227, "step": 12255 }, { "epoch": 13.74, "learning_rate": 1.4156338028169013e-05, "loss": 0.6411, "step": 12256 }, { "epoch": 13.74, "learning_rate": 1.4154929577464788e-05, "loss": 0.7551, "step": 12257 }, { "epoch": 13.74, "learning_rate": 1.4153521126760563e-05, "loss": 0.5891, "step": 12258 }, { "epoch": 13.74, "learning_rate": 1.4152112676056339e-05, "loss": 0.5674, "step": 12259 }, { "epoch": 13.74, "learning_rate": 1.4150704225352114e-05, "loss": 0.891, "step": 12260 }, { "epoch": 13.75, "learning_rate": 1.4149295774647888e-05, "loss": 0.5359, "step": 12261 }, { "epoch": 13.75, "learning_rate": 1.4147887323943663e-05, "loss": 0.638, "step": 12262 }, { "epoch": 13.75, "learning_rate": 1.4146478873239437e-05, "loss": 0.7669, "step": 12263 }, { "epoch": 13.75, "learning_rate": 1.414507042253521e-05, "loss": 0.5832, "step": 12264 }, { "epoch": 13.75, "learning_rate": 1.4143661971830985e-05, "loss": 0.6154, "step": 12265 }, { "epoch": 13.75, "learning_rate": 1.4142253521126761e-05, "loss": 0.6438, "step": 12266 }, { "epoch": 13.75, "learning_rate": 1.4140845070422536e-05, "loss": 0.5428, "step": 12267 }, { "epoch": 13.75, "learning_rate": 1.413943661971831e-05, "loss": 0.7332, "step": 12268 }, { "epoch": 13.75, "learning_rate": 1.4138028169014085e-05, "loss": 0.8092, "step": 12269 }, { "epoch": 13.76, "learning_rate": 1.413661971830986e-05, "loss": 0.6616, "step": 12270 }, { "epoch": 13.76, "learning_rate": 1.4135211267605635e-05, "loss": 0.7876, "step": 12271 }, { "epoch": 13.76, "learning_rate": 1.4133802816901408e-05, "loss": 0.668, "step": 12272 }, { "epoch": 13.76, "learning_rate": 1.4132394366197182e-05, "loss": 0.6615, "step": 12273 }, { "epoch": 13.76, "learning_rate": 1.4130985915492959e-05, "loss": 0.7809, "step": 12274 }, { "epoch": 13.76, "learning_rate": 1.4129577464788733e-05, "loss": 0.6333, "step": 12275 }, { "epoch": 13.76, "learning_rate": 1.4128169014084508e-05, "loss": 0.6639, "step": 12276 }, { "epoch": 13.76, "learning_rate": 1.4126760563380282e-05, "loss": 0.7578, "step": 12277 }, { "epoch": 13.76, "learning_rate": 1.4125352112676057e-05, "loss": 0.5658, "step": 12278 }, { "epoch": 13.77, "learning_rate": 1.4123943661971832e-05, "loss": 0.5955, "step": 12279 }, { "epoch": 13.77, "learning_rate": 1.4122535211267605e-05, "loss": 0.839, "step": 12280 }, { "epoch": 13.77, "learning_rate": 1.412112676056338e-05, "loss": 0.6595, "step": 12281 }, { "epoch": 13.77, "learning_rate": 1.4119718309859156e-05, "loss": 0.6889, "step": 12282 }, { "epoch": 13.77, "learning_rate": 1.411830985915493e-05, "loss": 0.7391, "step": 12283 }, { "epoch": 13.77, "learning_rate": 1.4116901408450705e-05, "loss": 0.7621, "step": 12284 }, { "epoch": 13.77, "learning_rate": 1.411549295774648e-05, "loss": 0.5756, "step": 12285 }, { "epoch": 13.77, "learning_rate": 1.4114084507042254e-05, "loss": 0.5295, "step": 12286 }, { "epoch": 13.77, "learning_rate": 1.4112676056338029e-05, "loss": 0.7147, "step": 12287 }, { "epoch": 13.78, "learning_rate": 1.4111267605633802e-05, "loss": 0.625, "step": 12288 }, { "epoch": 13.78, "learning_rate": 1.4109859154929578e-05, "loss": 0.6317, "step": 12289 }, { "epoch": 13.78, "learning_rate": 1.4108450704225353e-05, "loss": 0.5384, "step": 12290 }, { "epoch": 13.78, "learning_rate": 1.4107042253521127e-05, "loss": 0.6282, "step": 12291 }, { "epoch": 13.78, "learning_rate": 1.4105633802816902e-05, "loss": 0.584, "step": 12292 }, { "epoch": 13.78, "learning_rate": 1.4104225352112677e-05, "loss": 0.5583, "step": 12293 }, { "epoch": 13.78, "learning_rate": 1.4102816901408451e-05, "loss": 0.628, "step": 12294 }, { "epoch": 13.78, "learning_rate": 1.4101408450704226e-05, "loss": 0.6459, "step": 12295 }, { "epoch": 13.78, "learning_rate": 1.4099999999999999e-05, "loss": 0.7027, "step": 12296 }, { "epoch": 13.79, "learning_rate": 1.4098591549295775e-05, "loss": 0.7723, "step": 12297 }, { "epoch": 13.79, "learning_rate": 1.409718309859155e-05, "loss": 0.8111, "step": 12298 }, { "epoch": 13.79, "learning_rate": 1.4095774647887324e-05, "loss": 0.7592, "step": 12299 }, { "epoch": 13.79, "learning_rate": 1.4094366197183099e-05, "loss": 0.728, "step": 12300 }, { "epoch": 13.79, "learning_rate": 1.4092957746478874e-05, "loss": 0.6638, "step": 12301 }, { "epoch": 13.79, "learning_rate": 1.4091549295774648e-05, "loss": 0.6567, "step": 12302 }, { "epoch": 13.79, "learning_rate": 1.4090140845070423e-05, "loss": 0.7772, "step": 12303 }, { "epoch": 13.79, "learning_rate": 1.4088732394366198e-05, "loss": 0.7191, "step": 12304 }, { "epoch": 13.79, "learning_rate": 1.4087323943661972e-05, "loss": 0.7706, "step": 12305 }, { "epoch": 13.8, "learning_rate": 1.4085915492957747e-05, "loss": 0.7334, "step": 12306 }, { "epoch": 13.8, "learning_rate": 1.4084507042253522e-05, "loss": 0.6606, "step": 12307 }, { "epoch": 13.8, "learning_rate": 1.4083098591549296e-05, "loss": 0.7125, "step": 12308 }, { "epoch": 13.8, "learning_rate": 1.408169014084507e-05, "loss": 0.709, "step": 12309 }, { "epoch": 13.8, "learning_rate": 1.4080281690140845e-05, "loss": 0.7112, "step": 12310 }, { "epoch": 13.8, "learning_rate": 1.407887323943662e-05, "loss": 0.7113, "step": 12311 }, { "epoch": 13.8, "learning_rate": 1.4077464788732395e-05, "loss": 0.8039, "step": 12312 }, { "epoch": 13.8, "learning_rate": 1.407605633802817e-05, "loss": 0.6355, "step": 12313 }, { "epoch": 13.8, "learning_rate": 1.4074647887323944e-05, "loss": 0.6982, "step": 12314 }, { "epoch": 13.81, "learning_rate": 1.4073239436619719e-05, "loss": 0.6671, "step": 12315 }, { "epoch": 13.81, "learning_rate": 1.4071830985915493e-05, "loss": 0.6348, "step": 12316 }, { "epoch": 13.81, "learning_rate": 1.4070422535211268e-05, "loss": 0.6089, "step": 12317 }, { "epoch": 13.81, "learning_rate": 1.4069014084507043e-05, "loss": 0.7633, "step": 12318 }, { "epoch": 13.81, "learning_rate": 1.4067605633802819e-05, "loss": 0.5035, "step": 12319 }, { "epoch": 13.81, "learning_rate": 1.4066197183098592e-05, "loss": 0.681, "step": 12320 }, { "epoch": 13.81, "learning_rate": 1.4064788732394366e-05, "loss": 0.6062, "step": 12321 }, { "epoch": 13.81, "learning_rate": 1.4063380281690141e-05, "loss": 0.6708, "step": 12322 }, { "epoch": 13.82, "learning_rate": 1.4061971830985916e-05, "loss": 0.7865, "step": 12323 }, { "epoch": 13.82, "learning_rate": 1.406056338028169e-05, "loss": 0.6134, "step": 12324 }, { "epoch": 13.82, "learning_rate": 1.4059154929577465e-05, "loss": 0.6548, "step": 12325 }, { "epoch": 13.82, "learning_rate": 1.405774647887324e-05, "loss": 0.7485, "step": 12326 }, { "epoch": 13.82, "learning_rate": 1.4056338028169016e-05, "loss": 0.6656, "step": 12327 }, { "epoch": 13.82, "learning_rate": 1.4054929577464789e-05, "loss": 0.4625, "step": 12328 }, { "epoch": 13.82, "learning_rate": 1.4053521126760564e-05, "loss": 0.7147, "step": 12329 }, { "epoch": 13.82, "learning_rate": 1.4052112676056338e-05, "loss": 0.5491, "step": 12330 }, { "epoch": 13.82, "learning_rate": 1.4050704225352113e-05, "loss": 0.6564, "step": 12331 }, { "epoch": 13.83, "learning_rate": 1.4049295774647887e-05, "loss": 0.4795, "step": 12332 }, { "epoch": 13.83, "learning_rate": 1.4047887323943662e-05, "loss": 0.6579, "step": 12333 }, { "epoch": 13.83, "learning_rate": 1.4046478873239438e-05, "loss": 0.6775, "step": 12334 }, { "epoch": 13.83, "learning_rate": 1.4045070422535211e-05, "loss": 0.6452, "step": 12335 }, { "epoch": 13.83, "learning_rate": 1.4043661971830986e-05, "loss": 0.7726, "step": 12336 }, { "epoch": 13.83, "learning_rate": 1.404225352112676e-05, "loss": 0.6836, "step": 12337 }, { "epoch": 13.83, "learning_rate": 1.4040845070422535e-05, "loss": 0.6458, "step": 12338 }, { "epoch": 13.83, "learning_rate": 1.403943661971831e-05, "loss": 0.4816, "step": 12339 }, { "epoch": 13.83, "learning_rate": 1.4038028169014085e-05, "loss": 0.615, "step": 12340 }, { "epoch": 13.84, "learning_rate": 1.403661971830986e-05, "loss": 0.721, "step": 12341 }, { "epoch": 13.84, "learning_rate": 1.4035211267605636e-05, "loss": 0.6777, "step": 12342 }, { "epoch": 13.84, "learning_rate": 1.4033802816901408e-05, "loss": 0.6066, "step": 12343 }, { "epoch": 13.84, "learning_rate": 1.4032394366197183e-05, "loss": 0.6469, "step": 12344 }, { "epoch": 13.84, "learning_rate": 1.4030985915492958e-05, "loss": 0.5993, "step": 12345 }, { "epoch": 13.84, "learning_rate": 1.4029577464788732e-05, "loss": 0.6667, "step": 12346 }, { "epoch": 13.84, "learning_rate": 1.4028169014084507e-05, "loss": 1.0632, "step": 12347 }, { "epoch": 13.84, "learning_rate": 1.4026760563380282e-05, "loss": 0.7104, "step": 12348 }, { "epoch": 13.84, "learning_rate": 1.4025352112676056e-05, "loss": 0.6767, "step": 12349 }, { "epoch": 13.85, "learning_rate": 1.4023943661971833e-05, "loss": 0.8374, "step": 12350 }, { "epoch": 13.85, "learning_rate": 1.4022535211267606e-05, "loss": 0.6978, "step": 12351 }, { "epoch": 13.85, "learning_rate": 1.402112676056338e-05, "loss": 0.5759, "step": 12352 }, { "epoch": 13.85, "learning_rate": 1.4019718309859155e-05, "loss": 0.8273, "step": 12353 }, { "epoch": 13.85, "learning_rate": 1.401830985915493e-05, "loss": 0.5848, "step": 12354 }, { "epoch": 13.85, "learning_rate": 1.4016901408450704e-05, "loss": 0.6212, "step": 12355 }, { "epoch": 13.85, "learning_rate": 1.4015492957746479e-05, "loss": 0.7247, "step": 12356 }, { "epoch": 13.85, "learning_rate": 1.4014084507042255e-05, "loss": 0.7069, "step": 12357 }, { "epoch": 13.85, "learning_rate": 1.401267605633803e-05, "loss": 0.7103, "step": 12358 }, { "epoch": 13.86, "learning_rate": 1.4011267605633803e-05, "loss": 0.7496, "step": 12359 }, { "epoch": 13.86, "learning_rate": 1.4009859154929577e-05, "loss": 0.644, "step": 12360 }, { "epoch": 13.86, "learning_rate": 1.4008450704225352e-05, "loss": 0.6177, "step": 12361 }, { "epoch": 13.86, "learning_rate": 1.4007042253521127e-05, "loss": 0.6323, "step": 12362 }, { "epoch": 13.86, "learning_rate": 1.4005633802816901e-05, "loss": 0.581, "step": 12363 }, { "epoch": 13.86, "learning_rate": 1.4004225352112676e-05, "loss": 0.7385, "step": 12364 }, { "epoch": 13.86, "learning_rate": 1.4002816901408452e-05, "loss": 0.6951, "step": 12365 }, { "epoch": 13.86, "learning_rate": 1.4001408450704227e-05, "loss": 0.834, "step": 12366 }, { "epoch": 13.86, "learning_rate": 1.4e-05, "loss": 0.6404, "step": 12367 }, { "epoch": 13.87, "learning_rate": 1.3998591549295774e-05, "loss": 0.6886, "step": 12368 }, { "epoch": 13.87, "learning_rate": 1.3997183098591549e-05, "loss": 0.5833, "step": 12369 }, { "epoch": 13.87, "learning_rate": 1.3995774647887324e-05, "loss": 0.6775, "step": 12370 }, { "epoch": 13.87, "learning_rate": 1.3994366197183098e-05, "loss": 0.6473, "step": 12371 }, { "epoch": 13.87, "learning_rate": 1.3992957746478875e-05, "loss": 0.732, "step": 12372 }, { "epoch": 13.87, "learning_rate": 1.399154929577465e-05, "loss": 0.7406, "step": 12373 }, { "epoch": 13.87, "learning_rate": 1.3990140845070424e-05, "loss": 0.6388, "step": 12374 }, { "epoch": 13.87, "learning_rate": 1.3988732394366197e-05, "loss": 0.6407, "step": 12375 }, { "epoch": 13.87, "learning_rate": 1.3987323943661972e-05, "loss": 0.6712, "step": 12376 }, { "epoch": 13.88, "learning_rate": 1.3985915492957746e-05, "loss": 0.6625, "step": 12377 }, { "epoch": 13.88, "learning_rate": 1.398450704225352e-05, "loss": 0.6539, "step": 12378 }, { "epoch": 13.88, "learning_rate": 1.3983098591549295e-05, "loss": 0.6778, "step": 12379 }, { "epoch": 13.88, "learning_rate": 1.3981690140845072e-05, "loss": 0.6092, "step": 12380 }, { "epoch": 13.88, "learning_rate": 1.3980281690140846e-05, "loss": 0.6609, "step": 12381 }, { "epoch": 13.88, "learning_rate": 1.3978873239436621e-05, "loss": 0.7366, "step": 12382 }, { "epoch": 13.88, "learning_rate": 1.3977464788732394e-05, "loss": 0.655, "step": 12383 }, { "epoch": 13.88, "learning_rate": 1.3976056338028169e-05, "loss": 0.5448, "step": 12384 }, { "epoch": 13.88, "learning_rate": 1.3974647887323943e-05, "loss": 0.5779, "step": 12385 }, { "epoch": 13.89, "learning_rate": 1.3973239436619718e-05, "loss": 0.6585, "step": 12386 }, { "epoch": 13.89, "learning_rate": 1.3971830985915494e-05, "loss": 0.6627, "step": 12387 }, { "epoch": 13.89, "learning_rate": 1.3970422535211269e-05, "loss": 0.7548, "step": 12388 }, { "epoch": 13.89, "learning_rate": 1.3969014084507044e-05, "loss": 0.5399, "step": 12389 }, { "epoch": 13.89, "learning_rate": 1.3967605633802818e-05, "loss": 0.8234, "step": 12390 }, { "epoch": 13.89, "learning_rate": 1.3966197183098591e-05, "loss": 0.5847, "step": 12391 }, { "epoch": 13.89, "learning_rate": 1.3964788732394366e-05, "loss": 0.5487, "step": 12392 }, { "epoch": 13.89, "learning_rate": 1.396338028169014e-05, "loss": 0.4911, "step": 12393 }, { "epoch": 13.89, "learning_rate": 1.3961971830985915e-05, "loss": 0.5869, "step": 12394 }, { "epoch": 13.9, "learning_rate": 1.3960563380281691e-05, "loss": 0.669, "step": 12395 }, { "epoch": 13.9, "learning_rate": 1.3959154929577466e-05, "loss": 0.5544, "step": 12396 }, { "epoch": 13.9, "learning_rate": 1.395774647887324e-05, "loss": 0.8891, "step": 12397 }, { "epoch": 13.9, "learning_rate": 1.3956338028169015e-05, "loss": 0.8173, "step": 12398 }, { "epoch": 13.9, "learning_rate": 1.3954929577464788e-05, "loss": 0.8746, "step": 12399 }, { "epoch": 13.9, "learning_rate": 1.3953521126760563e-05, "loss": 0.7658, "step": 12400 }, { "epoch": 13.9, "learning_rate": 1.3952112676056337e-05, "loss": 0.8526, "step": 12401 }, { "epoch": 13.9, "learning_rate": 1.3950704225352112e-05, "loss": 0.7549, "step": 12402 }, { "epoch": 13.9, "learning_rate": 1.3949295774647888e-05, "loss": 0.7029, "step": 12403 }, { "epoch": 13.91, "learning_rate": 1.3947887323943663e-05, "loss": 0.8068, "step": 12404 }, { "epoch": 13.91, "learning_rate": 1.3946478873239438e-05, "loss": 0.6008, "step": 12405 }, { "epoch": 13.91, "learning_rate": 1.3945070422535212e-05, "loss": 0.6353, "step": 12406 }, { "epoch": 13.91, "learning_rate": 1.3943661971830985e-05, "loss": 0.6719, "step": 12407 }, { "epoch": 13.91, "learning_rate": 1.394225352112676e-05, "loss": 0.7489, "step": 12408 }, { "epoch": 13.91, "learning_rate": 1.3940845070422535e-05, "loss": 0.6483, "step": 12409 }, { "epoch": 13.91, "learning_rate": 1.3939436619718311e-05, "loss": 0.665, "step": 12410 }, { "epoch": 13.91, "learning_rate": 1.3938028169014086e-05, "loss": 0.6342, "step": 12411 }, { "epoch": 13.91, "learning_rate": 1.393661971830986e-05, "loss": 0.6851, "step": 12412 }, { "epoch": 13.92, "learning_rate": 1.3935211267605635e-05, "loss": 0.9334, "step": 12413 }, { "epoch": 13.92, "learning_rate": 1.393380281690141e-05, "loss": 0.739, "step": 12414 }, { "epoch": 13.92, "learning_rate": 1.3932394366197182e-05, "loss": 0.6614, "step": 12415 }, { "epoch": 13.92, "learning_rate": 1.3930985915492957e-05, "loss": 0.6158, "step": 12416 }, { "epoch": 13.92, "learning_rate": 1.3929577464788732e-05, "loss": 0.5879, "step": 12417 }, { "epoch": 13.92, "learning_rate": 1.3928169014084508e-05, "loss": 0.5482, "step": 12418 }, { "epoch": 13.92, "learning_rate": 1.3926760563380283e-05, "loss": 0.5593, "step": 12419 }, { "epoch": 13.92, "learning_rate": 1.3925352112676057e-05, "loss": 0.6359, "step": 12420 }, { "epoch": 13.92, "learning_rate": 1.3923943661971832e-05, "loss": 0.6665, "step": 12421 }, { "epoch": 13.93, "learning_rate": 1.3922535211267607e-05, "loss": 0.7763, "step": 12422 }, { "epoch": 13.93, "learning_rate": 1.392112676056338e-05, "loss": 0.686, "step": 12423 }, { "epoch": 13.93, "learning_rate": 1.3919718309859154e-05, "loss": 0.6275, "step": 12424 }, { "epoch": 13.93, "learning_rate": 1.391830985915493e-05, "loss": 0.5968, "step": 12425 }, { "epoch": 13.93, "learning_rate": 1.3916901408450705e-05, "loss": 0.7796, "step": 12426 }, { "epoch": 13.93, "learning_rate": 1.391549295774648e-05, "loss": 0.5002, "step": 12427 }, { "epoch": 13.93, "learning_rate": 1.3914084507042254e-05, "loss": 0.5154, "step": 12428 }, { "epoch": 13.93, "learning_rate": 1.3912676056338029e-05, "loss": 0.7549, "step": 12429 }, { "epoch": 13.93, "learning_rate": 1.3911267605633804e-05, "loss": 0.6546, "step": 12430 }, { "epoch": 13.94, "learning_rate": 1.3909859154929577e-05, "loss": 0.6262, "step": 12431 }, { "epoch": 13.94, "learning_rate": 1.3908450704225351e-05, "loss": 0.5304, "step": 12432 }, { "epoch": 13.94, "learning_rate": 1.3907042253521128e-05, "loss": 0.7439, "step": 12433 }, { "epoch": 13.94, "learning_rate": 1.3905633802816902e-05, "loss": 0.5983, "step": 12434 }, { "epoch": 13.94, "learning_rate": 1.3904225352112677e-05, "loss": 0.6681, "step": 12435 }, { "epoch": 13.94, "learning_rate": 1.3902816901408451e-05, "loss": 0.6666, "step": 12436 }, { "epoch": 13.94, "learning_rate": 1.3901408450704226e-05, "loss": 0.6475, "step": 12437 }, { "epoch": 13.94, "learning_rate": 1.39e-05, "loss": 0.4792, "step": 12438 }, { "epoch": 13.95, "learning_rate": 1.3898591549295774e-05, "loss": 0.6517, "step": 12439 }, { "epoch": 13.95, "learning_rate": 1.389718309859155e-05, "loss": 0.5353, "step": 12440 }, { "epoch": 13.95, "learning_rate": 1.3895774647887325e-05, "loss": 0.4871, "step": 12441 }, { "epoch": 13.95, "learning_rate": 1.38943661971831e-05, "loss": 0.5392, "step": 12442 }, { "epoch": 13.95, "learning_rate": 1.3892957746478874e-05, "loss": 0.7069, "step": 12443 }, { "epoch": 13.95, "learning_rate": 1.3891549295774649e-05, "loss": 0.5846, "step": 12444 }, { "epoch": 13.95, "learning_rate": 1.3890140845070423e-05, "loss": 0.6141, "step": 12445 }, { "epoch": 13.95, "learning_rate": 1.3888732394366198e-05, "loss": 0.8315, "step": 12446 }, { "epoch": 13.95, "learning_rate": 1.388732394366197e-05, "loss": 0.9185, "step": 12447 }, { "epoch": 13.96, "learning_rate": 1.3885915492957747e-05, "loss": 0.6036, "step": 12448 }, { "epoch": 13.96, "learning_rate": 1.3884507042253522e-05, "loss": 0.7965, "step": 12449 }, { "epoch": 13.96, "learning_rate": 1.3883098591549296e-05, "loss": 0.7397, "step": 12450 }, { "epoch": 13.96, "learning_rate": 1.3881690140845071e-05, "loss": 0.6737, "step": 12451 }, { "epoch": 13.96, "learning_rate": 1.3880281690140846e-05, "loss": 0.7601, "step": 12452 }, { "epoch": 13.96, "learning_rate": 1.387887323943662e-05, "loss": 0.7842, "step": 12453 }, { "epoch": 13.96, "learning_rate": 1.3877464788732395e-05, "loss": 0.6807, "step": 12454 }, { "epoch": 13.96, "learning_rate": 1.387605633802817e-05, "loss": 0.5627, "step": 12455 }, { "epoch": 13.96, "learning_rate": 1.3874647887323944e-05, "loss": 0.8183, "step": 12456 }, { "epoch": 13.97, "learning_rate": 1.3873239436619719e-05, "loss": 0.696, "step": 12457 }, { "epoch": 13.97, "learning_rate": 1.3871830985915493e-05, "loss": 0.8445, "step": 12458 }, { "epoch": 13.97, "learning_rate": 1.3870422535211268e-05, "loss": 0.7404, "step": 12459 }, { "epoch": 13.97, "learning_rate": 1.3869014084507043e-05, "loss": 0.7493, "step": 12460 }, { "epoch": 13.97, "learning_rate": 1.3867605633802817e-05, "loss": 0.5751, "step": 12461 }, { "epoch": 13.97, "learning_rate": 1.386619718309859e-05, "loss": 0.6723, "step": 12462 }, { "epoch": 13.97, "learning_rate": 1.3864788732394367e-05, "loss": 0.8502, "step": 12463 }, { "epoch": 13.97, "learning_rate": 1.3863380281690141e-05, "loss": 0.6435, "step": 12464 }, { "epoch": 13.97, "learning_rate": 1.3861971830985916e-05, "loss": 0.6916, "step": 12465 }, { "epoch": 13.98, "learning_rate": 1.386056338028169e-05, "loss": 0.5573, "step": 12466 }, { "epoch": 13.98, "learning_rate": 1.3859154929577465e-05, "loss": 0.6254, "step": 12467 }, { "epoch": 13.98, "learning_rate": 1.385774647887324e-05, "loss": 0.6661, "step": 12468 }, { "epoch": 13.98, "learning_rate": 1.3856338028169015e-05, "loss": 0.7526, "step": 12469 }, { "epoch": 13.98, "learning_rate": 1.3854929577464787e-05, "loss": 0.7116, "step": 12470 }, { "epoch": 13.98, "learning_rate": 1.3853521126760564e-05, "loss": 0.6627, "step": 12471 }, { "epoch": 13.98, "learning_rate": 1.3852112676056338e-05, "loss": 0.6855, "step": 12472 }, { "epoch": 13.98, "learning_rate": 1.3850704225352113e-05, "loss": 0.6468, "step": 12473 }, { "epoch": 13.98, "learning_rate": 1.3849295774647888e-05, "loss": 0.7376, "step": 12474 }, { "epoch": 13.99, "learning_rate": 1.3847887323943662e-05, "loss": 0.5523, "step": 12475 }, { "epoch": 13.99, "learning_rate": 1.3846478873239437e-05, "loss": 0.7985, "step": 12476 }, { "epoch": 13.99, "learning_rate": 1.3845070422535212e-05, "loss": 0.6828, "step": 12477 }, { "epoch": 13.99, "learning_rate": 1.3843661971830986e-05, "loss": 0.5899, "step": 12478 }, { "epoch": 13.99, "learning_rate": 1.3842253521126761e-05, "loss": 0.5296, "step": 12479 }, { "epoch": 13.99, "learning_rate": 1.3840845070422536e-05, "loss": 0.5904, "step": 12480 }, { "epoch": 13.99, "learning_rate": 1.383943661971831e-05, "loss": 0.6867, "step": 12481 }, { "epoch": 13.99, "learning_rate": 1.3838028169014085e-05, "loss": 0.7728, "step": 12482 }, { "epoch": 13.99, "learning_rate": 1.383661971830986e-05, "loss": 0.6779, "step": 12483 }, { "epoch": 14.0, "learning_rate": 1.3835211267605634e-05, "loss": 0.4791, "step": 12484 }, { "epoch": 14.0, "learning_rate": 1.3833802816901409e-05, "loss": 0.5268, "step": 12485 }, { "epoch": 14.0, "learning_rate": 1.3832394366197183e-05, "loss": 0.6264, "step": 12486 }, { "epoch": 14.0, "learning_rate": 1.3830985915492958e-05, "loss": 0.6049, "step": 12487 }, { "epoch": 14.0, "learning_rate": 1.3829577464788733e-05, "loss": 0.5925, "step": 12488 }, { "epoch": 14.0, "learning_rate": 1.3828169014084507e-05, "loss": 0.7824, "step": 12489 }, { "epoch": 14.0, "learning_rate": 1.3826760563380282e-05, "loss": 0.6574, "step": 12490 }, { "epoch": 14.0, "learning_rate": 1.3825352112676057e-05, "loss": 0.7536, "step": 12491 }, { "epoch": 14.0, "learning_rate": 1.3823943661971831e-05, "loss": 0.6992, "step": 12492 }, { "epoch": 14.01, "learning_rate": 1.3822535211267608e-05, "loss": 0.5933, "step": 12493 }, { "epoch": 14.01, "learning_rate": 1.382112676056338e-05, "loss": 0.7355, "step": 12494 }, { "epoch": 14.01, "learning_rate": 1.3819718309859155e-05, "loss": 0.7496, "step": 12495 }, { "epoch": 14.01, "learning_rate": 1.381830985915493e-05, "loss": 0.5592, "step": 12496 }, { "epoch": 14.01, "learning_rate": 1.3816901408450704e-05, "loss": 0.5716, "step": 12497 }, { "epoch": 14.01, "learning_rate": 1.3815492957746479e-05, "loss": 0.7115, "step": 12498 }, { "epoch": 14.01, "learning_rate": 1.3814084507042254e-05, "loss": 0.549, "step": 12499 }, { "epoch": 14.01, "learning_rate": 1.3812676056338028e-05, "loss": 0.4395, "step": 12500 }, { "epoch": 14.01, "learning_rate": 1.3811267605633805e-05, "loss": 0.6736, "step": 12501 }, { "epoch": 14.02, "learning_rate": 1.3809859154929578e-05, "loss": 0.6393, "step": 12502 }, { "epoch": 14.02, "learning_rate": 1.3808450704225352e-05, "loss": 0.6095, "step": 12503 }, { "epoch": 14.02, "learning_rate": 1.3807042253521127e-05, "loss": 0.6833, "step": 12504 }, { "epoch": 14.02, "learning_rate": 1.3805633802816901e-05, "loss": 0.5839, "step": 12505 }, { "epoch": 14.02, "learning_rate": 1.3804225352112676e-05, "loss": 0.5596, "step": 12506 }, { "epoch": 14.02, "learning_rate": 1.380281690140845e-05, "loss": 0.6884, "step": 12507 }, { "epoch": 14.02, "learning_rate": 1.3801408450704227e-05, "loss": 0.8969, "step": 12508 }, { "epoch": 14.02, "learning_rate": 1.3800000000000002e-05, "loss": 0.5376, "step": 12509 }, { "epoch": 14.02, "learning_rate": 1.3798591549295775e-05, "loss": 0.7072, "step": 12510 }, { "epoch": 14.03, "learning_rate": 1.379718309859155e-05, "loss": 0.6619, "step": 12511 }, { "epoch": 14.03, "learning_rate": 1.3795774647887324e-05, "loss": 0.5831, "step": 12512 }, { "epoch": 14.03, "learning_rate": 1.3794366197183099e-05, "loss": 0.5948, "step": 12513 }, { "epoch": 14.03, "learning_rate": 1.3792957746478873e-05, "loss": 0.5197, "step": 12514 }, { "epoch": 14.03, "learning_rate": 1.3791549295774648e-05, "loss": 0.538, "step": 12515 }, { "epoch": 14.03, "learning_rate": 1.3790140845070424e-05, "loss": 0.631, "step": 12516 }, { "epoch": 14.03, "learning_rate": 1.3788732394366199e-05, "loss": 0.6244, "step": 12517 }, { "epoch": 14.03, "learning_rate": 1.3787323943661972e-05, "loss": 0.4979, "step": 12518 }, { "epoch": 14.03, "learning_rate": 1.3785915492957746e-05, "loss": 0.4989, "step": 12519 }, { "epoch": 14.04, "learning_rate": 1.3784507042253521e-05, "loss": 0.538, "step": 12520 }, { "epoch": 14.04, "learning_rate": 1.3783098591549296e-05, "loss": 0.5532, "step": 12521 }, { "epoch": 14.04, "learning_rate": 1.378169014084507e-05, "loss": 0.5892, "step": 12522 }, { "epoch": 14.04, "learning_rate": 1.3780281690140845e-05, "loss": 0.4696, "step": 12523 }, { "epoch": 14.04, "learning_rate": 1.3778873239436621e-05, "loss": 0.6781, "step": 12524 }, { "epoch": 14.04, "learning_rate": 1.3777464788732396e-05, "loss": 0.626, "step": 12525 }, { "epoch": 14.04, "learning_rate": 1.3776056338028169e-05, "loss": 0.5019, "step": 12526 }, { "epoch": 14.04, "learning_rate": 1.3774647887323943e-05, "loss": 0.4954, "step": 12527 }, { "epoch": 14.04, "learning_rate": 1.3773239436619718e-05, "loss": 0.6295, "step": 12528 }, { "epoch": 14.05, "learning_rate": 1.3771830985915493e-05, "loss": 0.4881, "step": 12529 }, { "epoch": 14.05, "learning_rate": 1.3770422535211267e-05, "loss": 0.621, "step": 12530 }, { "epoch": 14.05, "learning_rate": 1.3769014084507044e-05, "loss": 0.5364, "step": 12531 }, { "epoch": 14.05, "learning_rate": 1.3767605633802818e-05, "loss": 0.6178, "step": 12532 }, { "epoch": 14.05, "learning_rate": 1.3766197183098593e-05, "loss": 0.687, "step": 12533 }, { "epoch": 14.05, "learning_rate": 1.3764788732394366e-05, "loss": 0.5177, "step": 12534 }, { "epoch": 14.05, "learning_rate": 1.376338028169014e-05, "loss": 0.534, "step": 12535 }, { "epoch": 14.05, "learning_rate": 1.3761971830985915e-05, "loss": 0.5337, "step": 12536 }, { "epoch": 14.05, "learning_rate": 1.376056338028169e-05, "loss": 0.5977, "step": 12537 }, { "epoch": 14.06, "learning_rate": 1.3759154929577464e-05, "loss": 0.5504, "step": 12538 }, { "epoch": 14.06, "learning_rate": 1.375774647887324e-05, "loss": 1.0224, "step": 12539 }, { "epoch": 14.06, "learning_rate": 1.3756338028169015e-05, "loss": 0.7158, "step": 12540 }, { "epoch": 14.06, "learning_rate": 1.375492957746479e-05, "loss": 0.6189, "step": 12541 }, { "epoch": 14.06, "learning_rate": 1.3753521126760563e-05, "loss": 0.6633, "step": 12542 }, { "epoch": 14.06, "learning_rate": 1.3752112676056338e-05, "loss": 0.583, "step": 12543 }, { "epoch": 14.06, "learning_rate": 1.3750704225352112e-05, "loss": 0.633, "step": 12544 }, { "epoch": 14.06, "learning_rate": 1.3749295774647887e-05, "loss": 0.5643, "step": 12545 }, { "epoch": 14.07, "learning_rate": 1.3747887323943663e-05, "loss": 0.7412, "step": 12546 }, { "epoch": 14.07, "learning_rate": 1.3746478873239438e-05, "loss": 0.5179, "step": 12547 }, { "epoch": 14.07, "learning_rate": 1.3745070422535213e-05, "loss": 0.6047, "step": 12548 }, { "epoch": 14.07, "learning_rate": 1.3743661971830987e-05, "loss": 0.6468, "step": 12549 }, { "epoch": 14.07, "learning_rate": 1.374225352112676e-05, "loss": 0.6962, "step": 12550 }, { "epoch": 14.07, "learning_rate": 1.3740845070422535e-05, "loss": 0.5997, "step": 12551 }, { "epoch": 14.07, "learning_rate": 1.373943661971831e-05, "loss": 0.633, "step": 12552 }, { "epoch": 14.07, "learning_rate": 1.3738028169014084e-05, "loss": 0.6508, "step": 12553 }, { "epoch": 14.07, "learning_rate": 1.373661971830986e-05, "loss": 0.5487, "step": 12554 }, { "epoch": 14.08, "learning_rate": 1.3735211267605635e-05, "loss": 0.586, "step": 12555 }, { "epoch": 14.08, "learning_rate": 1.373380281690141e-05, "loss": 0.6681, "step": 12556 }, { "epoch": 14.08, "learning_rate": 1.3732394366197184e-05, "loss": 0.6125, "step": 12557 }, { "epoch": 14.08, "learning_rate": 1.3730985915492957e-05, "loss": 0.6006, "step": 12558 }, { "epoch": 14.08, "learning_rate": 1.3729577464788732e-05, "loss": 0.6614, "step": 12559 }, { "epoch": 14.08, "learning_rate": 1.3728169014084507e-05, "loss": 0.4816, "step": 12560 }, { "epoch": 14.08, "learning_rate": 1.3726760563380283e-05, "loss": 0.5485, "step": 12561 }, { "epoch": 14.08, "learning_rate": 1.3725352112676057e-05, "loss": 0.6306, "step": 12562 }, { "epoch": 14.08, "learning_rate": 1.3723943661971832e-05, "loss": 0.6062, "step": 12563 }, { "epoch": 14.09, "learning_rate": 1.3722535211267607e-05, "loss": 0.5301, "step": 12564 }, { "epoch": 14.09, "learning_rate": 1.372112676056338e-05, "loss": 0.6833, "step": 12565 }, { "epoch": 14.09, "learning_rate": 1.3719718309859154e-05, "loss": 0.5486, "step": 12566 }, { "epoch": 14.09, "learning_rate": 1.3718309859154929e-05, "loss": 0.6402, "step": 12567 }, { "epoch": 14.09, "learning_rate": 1.3716901408450704e-05, "loss": 0.5283, "step": 12568 }, { "epoch": 14.09, "learning_rate": 1.371549295774648e-05, "loss": 0.5484, "step": 12569 }, { "epoch": 14.09, "learning_rate": 1.3714084507042255e-05, "loss": 0.566, "step": 12570 }, { "epoch": 14.09, "learning_rate": 1.371267605633803e-05, "loss": 0.5372, "step": 12571 }, { "epoch": 14.09, "learning_rate": 1.3711267605633804e-05, "loss": 0.542, "step": 12572 }, { "epoch": 14.1, "learning_rate": 1.3709859154929577e-05, "loss": 0.6264, "step": 12573 }, { "epoch": 14.1, "learning_rate": 1.3708450704225351e-05, "loss": 0.5872, "step": 12574 }, { "epoch": 14.1, "learning_rate": 1.3707042253521126e-05, "loss": 0.4632, "step": 12575 }, { "epoch": 14.1, "learning_rate": 1.37056338028169e-05, "loss": 0.639, "step": 12576 }, { "epoch": 14.1, "learning_rate": 1.3704225352112677e-05, "loss": 0.6193, "step": 12577 }, { "epoch": 14.1, "learning_rate": 1.3702816901408452e-05, "loss": 0.7118, "step": 12578 }, { "epoch": 14.1, "learning_rate": 1.3701408450704226e-05, "loss": 0.6376, "step": 12579 }, { "epoch": 14.1, "learning_rate": 1.3700000000000001e-05, "loss": 0.7498, "step": 12580 }, { "epoch": 14.1, "learning_rate": 1.3698591549295774e-05, "loss": 0.5718, "step": 12581 }, { "epoch": 14.11, "learning_rate": 1.3697183098591549e-05, "loss": 0.6531, "step": 12582 }, { "epoch": 14.11, "learning_rate": 1.3695774647887323e-05, "loss": 0.5068, "step": 12583 }, { "epoch": 14.11, "learning_rate": 1.36943661971831e-05, "loss": 0.4817, "step": 12584 }, { "epoch": 14.11, "learning_rate": 1.3692957746478874e-05, "loss": 0.5125, "step": 12585 }, { "epoch": 14.11, "learning_rate": 1.3691549295774649e-05, "loss": 0.5497, "step": 12586 }, { "epoch": 14.11, "learning_rate": 1.3690140845070423e-05, "loss": 0.6389, "step": 12587 }, { "epoch": 14.11, "learning_rate": 1.3688732394366198e-05, "loss": 0.6411, "step": 12588 }, { "epoch": 14.11, "learning_rate": 1.3687323943661971e-05, "loss": 0.8109, "step": 12589 }, { "epoch": 14.11, "learning_rate": 1.3685915492957746e-05, "loss": 0.5649, "step": 12590 }, { "epoch": 14.12, "learning_rate": 1.368450704225352e-05, "loss": 0.709, "step": 12591 }, { "epoch": 14.12, "learning_rate": 1.3683098591549297e-05, "loss": 0.6786, "step": 12592 }, { "epoch": 14.12, "learning_rate": 1.3681690140845071e-05, "loss": 0.578, "step": 12593 }, { "epoch": 14.12, "learning_rate": 1.3680281690140846e-05, "loss": 0.5958, "step": 12594 }, { "epoch": 14.12, "learning_rate": 1.367887323943662e-05, "loss": 0.5679, "step": 12595 }, { "epoch": 14.12, "learning_rate": 1.3677464788732395e-05, "loss": 0.6635, "step": 12596 }, { "epoch": 14.12, "learning_rate": 1.3676056338028168e-05, "loss": 0.538, "step": 12597 }, { "epoch": 14.12, "learning_rate": 1.3674647887323943e-05, "loss": 0.6757, "step": 12598 }, { "epoch": 14.12, "learning_rate": 1.3673239436619719e-05, "loss": 0.6476, "step": 12599 }, { "epoch": 14.13, "learning_rate": 1.3671830985915494e-05, "loss": 0.6662, "step": 12600 }, { "epoch": 14.13, "learning_rate": 1.3670422535211268e-05, "loss": 0.5609, "step": 12601 }, { "epoch": 14.13, "learning_rate": 1.3669014084507043e-05, "loss": 0.6059, "step": 12602 }, { "epoch": 14.13, "learning_rate": 1.3667605633802818e-05, "loss": 0.6101, "step": 12603 }, { "epoch": 14.13, "learning_rate": 1.3666197183098592e-05, "loss": 0.568, "step": 12604 }, { "epoch": 14.13, "learning_rate": 1.3664788732394365e-05, "loss": 0.6167, "step": 12605 }, { "epoch": 14.13, "learning_rate": 1.366338028169014e-05, "loss": 0.6028, "step": 12606 }, { "epoch": 14.13, "learning_rate": 1.3661971830985916e-05, "loss": 0.7393, "step": 12607 }, { "epoch": 14.13, "learning_rate": 1.366056338028169e-05, "loss": 0.5984, "step": 12608 }, { "epoch": 14.14, "learning_rate": 1.3659154929577465e-05, "loss": 0.5614, "step": 12609 }, { "epoch": 14.14, "learning_rate": 1.365774647887324e-05, "loss": 0.634, "step": 12610 }, { "epoch": 14.14, "learning_rate": 1.3656338028169015e-05, "loss": 0.512, "step": 12611 }, { "epoch": 14.14, "learning_rate": 1.365492957746479e-05, "loss": 0.6674, "step": 12612 }, { "epoch": 14.14, "learning_rate": 1.3653521126760562e-05, "loss": 0.5922, "step": 12613 }, { "epoch": 14.14, "learning_rate": 1.3652112676056339e-05, "loss": 0.5258, "step": 12614 }, { "epoch": 14.14, "learning_rate": 1.3650704225352113e-05, "loss": 0.65, "step": 12615 }, { "epoch": 14.14, "learning_rate": 1.3649295774647888e-05, "loss": 0.4793, "step": 12616 }, { "epoch": 14.14, "learning_rate": 1.3647887323943663e-05, "loss": 0.6746, "step": 12617 }, { "epoch": 14.15, "learning_rate": 1.3646478873239437e-05, "loss": 0.6632, "step": 12618 }, { "epoch": 14.15, "learning_rate": 1.3645070422535212e-05, "loss": 0.5765, "step": 12619 }, { "epoch": 14.15, "learning_rate": 1.3643661971830986e-05, "loss": 0.525, "step": 12620 }, { "epoch": 14.15, "learning_rate": 1.364225352112676e-05, "loss": 0.6403, "step": 12621 }, { "epoch": 14.15, "learning_rate": 1.3640845070422536e-05, "loss": 0.6106, "step": 12622 }, { "epoch": 14.15, "learning_rate": 1.363943661971831e-05, "loss": 0.6433, "step": 12623 }, { "epoch": 14.15, "learning_rate": 1.3638028169014085e-05, "loss": 0.5242, "step": 12624 }, { "epoch": 14.15, "learning_rate": 1.363661971830986e-05, "loss": 0.6265, "step": 12625 }, { "epoch": 14.15, "learning_rate": 1.3635211267605634e-05, "loss": 0.5069, "step": 12626 }, { "epoch": 14.16, "learning_rate": 1.3633802816901409e-05, "loss": 0.5719, "step": 12627 }, { "epoch": 14.16, "learning_rate": 1.3632394366197184e-05, "loss": 0.6393, "step": 12628 }, { "epoch": 14.16, "learning_rate": 1.3630985915492958e-05, "loss": 0.574, "step": 12629 }, { "epoch": 14.16, "learning_rate": 1.3629577464788733e-05, "loss": 0.4612, "step": 12630 }, { "epoch": 14.16, "learning_rate": 1.3628169014084507e-05, "loss": 0.485, "step": 12631 }, { "epoch": 14.16, "learning_rate": 1.3626760563380282e-05, "loss": 0.3953, "step": 12632 }, { "epoch": 14.16, "learning_rate": 1.3625352112676057e-05, "loss": 0.6914, "step": 12633 }, { "epoch": 14.16, "learning_rate": 1.3623943661971831e-05, "loss": 0.6019, "step": 12634 }, { "epoch": 14.16, "learning_rate": 1.3622535211267606e-05, "loss": 0.3622, "step": 12635 }, { "epoch": 14.17, "learning_rate": 1.362112676056338e-05, "loss": 0.4945, "step": 12636 }, { "epoch": 14.17, "learning_rate": 1.3619718309859155e-05, "loss": 0.4413, "step": 12637 }, { "epoch": 14.17, "learning_rate": 1.361830985915493e-05, "loss": 0.63, "step": 12638 }, { "epoch": 14.17, "learning_rate": 1.3616901408450705e-05, "loss": 0.726, "step": 12639 }, { "epoch": 14.17, "learning_rate": 1.361549295774648e-05, "loss": 0.8805, "step": 12640 }, { "epoch": 14.17, "learning_rate": 1.3614084507042254e-05, "loss": 0.6592, "step": 12641 }, { "epoch": 14.17, "learning_rate": 1.3612676056338028e-05, "loss": 0.789, "step": 12642 }, { "epoch": 14.17, "learning_rate": 1.3611267605633803e-05, "loss": 0.6705, "step": 12643 }, { "epoch": 14.17, "learning_rate": 1.3609859154929578e-05, "loss": 0.5979, "step": 12644 }, { "epoch": 14.18, "learning_rate": 1.3608450704225352e-05, "loss": 0.6031, "step": 12645 }, { "epoch": 14.18, "learning_rate": 1.3607042253521127e-05, "loss": 0.619, "step": 12646 }, { "epoch": 14.18, "learning_rate": 1.3605633802816902e-05, "loss": 0.7213, "step": 12647 }, { "epoch": 14.18, "learning_rate": 1.3604225352112676e-05, "loss": 0.7737, "step": 12648 }, { "epoch": 14.18, "learning_rate": 1.3602816901408451e-05, "loss": 0.6328, "step": 12649 }, { "epoch": 14.18, "learning_rate": 1.3601408450704226e-05, "loss": 0.5994, "step": 12650 }, { "epoch": 14.18, "learning_rate": 1.36e-05, "loss": 0.7081, "step": 12651 }, { "epoch": 14.18, "learning_rate": 1.3598591549295777e-05, "loss": 0.5071, "step": 12652 }, { "epoch": 14.18, "learning_rate": 1.359718309859155e-05, "loss": 0.764, "step": 12653 }, { "epoch": 14.19, "learning_rate": 1.3595774647887324e-05, "loss": 0.807, "step": 12654 }, { "epoch": 14.19, "learning_rate": 1.3594366197183099e-05, "loss": 0.5807, "step": 12655 }, { "epoch": 14.19, "learning_rate": 1.3592957746478873e-05, "loss": 0.6632, "step": 12656 }, { "epoch": 14.19, "learning_rate": 1.3591549295774648e-05, "loss": 0.578, "step": 12657 }, { "epoch": 14.19, "learning_rate": 1.3590140845070423e-05, "loss": 0.5476, "step": 12658 }, { "epoch": 14.19, "learning_rate": 1.3588732394366197e-05, "loss": 0.6644, "step": 12659 }, { "epoch": 14.19, "learning_rate": 1.3587323943661974e-05, "loss": 0.6051, "step": 12660 }, { "epoch": 14.19, "learning_rate": 1.3585915492957747e-05, "loss": 0.5601, "step": 12661 }, { "epoch": 14.2, "learning_rate": 1.3584507042253521e-05, "loss": 0.5976, "step": 12662 }, { "epoch": 14.2, "learning_rate": 1.3583098591549296e-05, "loss": 0.6132, "step": 12663 }, { "epoch": 14.2, "learning_rate": 1.358169014084507e-05, "loss": 0.6312, "step": 12664 }, { "epoch": 14.2, "learning_rate": 1.3580281690140845e-05, "loss": 0.5602, "step": 12665 }, { "epoch": 14.2, "learning_rate": 1.357887323943662e-05, "loss": 0.5859, "step": 12666 }, { "epoch": 14.2, "learning_rate": 1.3577464788732396e-05, "loss": 0.6173, "step": 12667 }, { "epoch": 14.2, "learning_rate": 1.357605633802817e-05, "loss": 0.5628, "step": 12668 }, { "epoch": 14.2, "learning_rate": 1.3574647887323944e-05, "loss": 0.635, "step": 12669 }, { "epoch": 14.2, "learning_rate": 1.3573239436619718e-05, "loss": 0.5767, "step": 12670 }, { "epoch": 14.21, "learning_rate": 1.3571830985915493e-05, "loss": 0.579, "step": 12671 }, { "epoch": 14.21, "learning_rate": 1.3570422535211268e-05, "loss": 0.6327, "step": 12672 }, { "epoch": 14.21, "learning_rate": 1.3569014084507042e-05, "loss": 0.5117, "step": 12673 }, { "epoch": 14.21, "learning_rate": 1.3567605633802817e-05, "loss": 0.699, "step": 12674 }, { "epoch": 14.21, "learning_rate": 1.3566197183098593e-05, "loss": 0.6376, "step": 12675 }, { "epoch": 14.21, "learning_rate": 1.3564788732394366e-05, "loss": 0.5061, "step": 12676 }, { "epoch": 14.21, "learning_rate": 1.356338028169014e-05, "loss": 0.6665, "step": 12677 }, { "epoch": 14.21, "learning_rate": 1.3561971830985915e-05, "loss": 0.6517, "step": 12678 }, { "epoch": 14.21, "learning_rate": 1.356056338028169e-05, "loss": 0.5372, "step": 12679 }, { "epoch": 14.22, "learning_rate": 1.3559154929577465e-05, "loss": 0.8194, "step": 12680 }, { "epoch": 14.22, "learning_rate": 1.355774647887324e-05, "loss": 0.633, "step": 12681 }, { "epoch": 14.22, "learning_rate": 1.3556338028169016e-05, "loss": 0.4683, "step": 12682 }, { "epoch": 14.22, "learning_rate": 1.355492957746479e-05, "loss": 0.5785, "step": 12683 }, { "epoch": 14.22, "learning_rate": 1.3553521126760563e-05, "loss": 0.4634, "step": 12684 }, { "epoch": 14.22, "learning_rate": 1.3552112676056338e-05, "loss": 0.5861, "step": 12685 }, { "epoch": 14.22, "learning_rate": 1.3550704225352113e-05, "loss": 0.7524, "step": 12686 }, { "epoch": 14.22, "learning_rate": 1.3549295774647887e-05, "loss": 0.6106, "step": 12687 }, { "epoch": 14.22, "learning_rate": 1.3547887323943662e-05, "loss": 0.6729, "step": 12688 }, { "epoch": 14.23, "learning_rate": 1.3546478873239436e-05, "loss": 0.8178, "step": 12689 }, { "epoch": 14.23, "learning_rate": 1.3545070422535213e-05, "loss": 0.7333, "step": 12690 }, { "epoch": 14.23, "learning_rate": 1.3543661971830987e-05, "loss": 0.6942, "step": 12691 }, { "epoch": 14.23, "learning_rate": 1.354225352112676e-05, "loss": 0.7348, "step": 12692 }, { "epoch": 14.23, "learning_rate": 1.3540845070422535e-05, "loss": 0.7769, "step": 12693 }, { "epoch": 14.23, "learning_rate": 1.353943661971831e-05, "loss": 0.8942, "step": 12694 }, { "epoch": 14.23, "learning_rate": 1.3538028169014084e-05, "loss": 0.6765, "step": 12695 }, { "epoch": 14.23, "learning_rate": 1.3536619718309859e-05, "loss": 0.6063, "step": 12696 }, { "epoch": 14.23, "learning_rate": 1.3535211267605634e-05, "loss": 0.7477, "step": 12697 }, { "epoch": 14.24, "learning_rate": 1.353380281690141e-05, "loss": 0.6115, "step": 12698 }, { "epoch": 14.24, "learning_rate": 1.3532394366197185e-05, "loss": 0.5183, "step": 12699 }, { "epoch": 14.24, "learning_rate": 1.3530985915492957e-05, "loss": 0.802, "step": 12700 }, { "epoch": 14.24, "learning_rate": 1.3529577464788732e-05, "loss": 0.4985, "step": 12701 }, { "epoch": 14.24, "learning_rate": 1.3528169014084507e-05, "loss": 0.542, "step": 12702 }, { "epoch": 14.24, "learning_rate": 1.3526760563380281e-05, "loss": 0.5931, "step": 12703 }, { "epoch": 14.24, "learning_rate": 1.3525352112676056e-05, "loss": 0.6861, "step": 12704 }, { "epoch": 14.24, "learning_rate": 1.3523943661971832e-05, "loss": 0.5362, "step": 12705 }, { "epoch": 14.24, "learning_rate": 1.3522535211267607e-05, "loss": 0.552, "step": 12706 }, { "epoch": 14.25, "learning_rate": 1.3521126760563382e-05, "loss": 0.5701, "step": 12707 }, { "epoch": 14.25, "learning_rate": 1.3519718309859155e-05, "loss": 0.7831, "step": 12708 }, { "epoch": 14.25, "learning_rate": 1.351830985915493e-05, "loss": 0.6119, "step": 12709 }, { "epoch": 14.25, "learning_rate": 1.3516901408450704e-05, "loss": 0.6377, "step": 12710 }, { "epoch": 14.25, "learning_rate": 1.3515492957746478e-05, "loss": 0.6551, "step": 12711 }, { "epoch": 14.25, "learning_rate": 1.3514084507042253e-05, "loss": 0.6496, "step": 12712 }, { "epoch": 14.25, "learning_rate": 1.351267605633803e-05, "loss": 0.6308, "step": 12713 }, { "epoch": 14.25, "learning_rate": 1.3511267605633804e-05, "loss": 0.693, "step": 12714 }, { "epoch": 14.25, "learning_rate": 1.3509859154929579e-05, "loss": 0.6538, "step": 12715 }, { "epoch": 14.26, "learning_rate": 1.3508450704225352e-05, "loss": 0.547, "step": 12716 }, { "epoch": 14.26, "learning_rate": 1.3507042253521126e-05, "loss": 0.4133, "step": 12717 }, { "epoch": 14.26, "learning_rate": 1.3505633802816901e-05, "loss": 0.7325, "step": 12718 }, { "epoch": 14.26, "learning_rate": 1.3504225352112676e-05, "loss": 0.6355, "step": 12719 }, { "epoch": 14.26, "learning_rate": 1.3502816901408452e-05, "loss": 0.5414, "step": 12720 }, { "epoch": 14.26, "learning_rate": 1.3501408450704227e-05, "loss": 0.5285, "step": 12721 }, { "epoch": 14.26, "learning_rate": 1.3500000000000001e-05, "loss": 0.5789, "step": 12722 }, { "epoch": 14.26, "learning_rate": 1.3498591549295776e-05, "loss": 0.5729, "step": 12723 }, { "epoch": 14.26, "learning_rate": 1.3497183098591549e-05, "loss": 0.5227, "step": 12724 }, { "epoch": 14.27, "learning_rate": 1.3495774647887323e-05, "loss": 0.4667, "step": 12725 }, { "epoch": 14.27, "learning_rate": 1.3494366197183098e-05, "loss": 0.6471, "step": 12726 }, { "epoch": 14.27, "learning_rate": 1.3492957746478873e-05, "loss": 0.5363, "step": 12727 }, { "epoch": 14.27, "learning_rate": 1.3491549295774649e-05, "loss": 0.6066, "step": 12728 }, { "epoch": 14.27, "learning_rate": 1.3490140845070424e-05, "loss": 0.6218, "step": 12729 }, { "epoch": 14.27, "learning_rate": 1.3488732394366198e-05, "loss": 0.689, "step": 12730 }, { "epoch": 14.27, "learning_rate": 1.3487323943661973e-05, "loss": 0.578, "step": 12731 }, { "epoch": 14.27, "learning_rate": 1.3485915492957746e-05, "loss": 0.5652, "step": 12732 }, { "epoch": 14.27, "learning_rate": 1.348450704225352e-05, "loss": 0.6133, "step": 12733 }, { "epoch": 14.28, "learning_rate": 1.3483098591549295e-05, "loss": 0.5704, "step": 12734 }, { "epoch": 14.28, "learning_rate": 1.3481690140845071e-05, "loss": 0.6554, "step": 12735 }, { "epoch": 14.28, "learning_rate": 1.3480281690140846e-05, "loss": 0.6214, "step": 12736 }, { "epoch": 14.28, "learning_rate": 1.347887323943662e-05, "loss": 0.6465, "step": 12737 }, { "epoch": 14.28, "learning_rate": 1.3477464788732395e-05, "loss": 0.3836, "step": 12738 }, { "epoch": 14.28, "learning_rate": 1.347605633802817e-05, "loss": 0.8487, "step": 12739 }, { "epoch": 14.28, "learning_rate": 1.3474647887323943e-05, "loss": 0.6885, "step": 12740 }, { "epoch": 14.28, "learning_rate": 1.3473239436619718e-05, "loss": 0.6392, "step": 12741 }, { "epoch": 14.28, "learning_rate": 1.3471830985915492e-05, "loss": 0.7242, "step": 12742 }, { "epoch": 14.29, "learning_rate": 1.3470422535211269e-05, "loss": 0.7356, "step": 12743 }, { "epoch": 14.29, "learning_rate": 1.3469014084507043e-05, "loss": 0.661, "step": 12744 }, { "epoch": 14.29, "learning_rate": 1.3467605633802818e-05, "loss": 0.6536, "step": 12745 }, { "epoch": 14.29, "learning_rate": 1.3466197183098592e-05, "loss": 0.6371, "step": 12746 }, { "epoch": 14.29, "learning_rate": 1.3464788732394367e-05, "loss": 0.6991, "step": 12747 }, { "epoch": 14.29, "learning_rate": 1.346338028169014e-05, "loss": 0.5599, "step": 12748 }, { "epoch": 14.29, "learning_rate": 1.3461971830985915e-05, "loss": 0.7095, "step": 12749 }, { "epoch": 14.29, "learning_rate": 1.3460563380281691e-05, "loss": 0.7586, "step": 12750 }, { "epoch": 14.29, "learning_rate": 1.3459154929577466e-05, "loss": 0.6041, "step": 12751 }, { "epoch": 14.3, "learning_rate": 1.345774647887324e-05, "loss": 0.6809, "step": 12752 }, { "epoch": 14.3, "learning_rate": 1.3456338028169015e-05, "loss": 0.4826, "step": 12753 }, { "epoch": 14.3, "learning_rate": 1.345492957746479e-05, "loss": 0.4635, "step": 12754 }, { "epoch": 14.3, "learning_rate": 1.3453521126760564e-05, "loss": 0.6078, "step": 12755 }, { "epoch": 14.3, "learning_rate": 1.3452112676056337e-05, "loss": 0.5631, "step": 12756 }, { "epoch": 14.3, "learning_rate": 1.3450704225352112e-05, "loss": 0.5021, "step": 12757 }, { "epoch": 14.3, "learning_rate": 1.3449295774647888e-05, "loss": 0.51, "step": 12758 }, { "epoch": 14.3, "learning_rate": 1.3447887323943663e-05, "loss": 0.6469, "step": 12759 }, { "epoch": 14.3, "learning_rate": 1.3446478873239437e-05, "loss": 0.5892, "step": 12760 }, { "epoch": 14.31, "learning_rate": 1.3445070422535212e-05, "loss": 0.5542, "step": 12761 }, { "epoch": 14.31, "learning_rate": 1.3443661971830987e-05, "loss": 0.5733, "step": 12762 }, { "epoch": 14.31, "learning_rate": 1.3442253521126761e-05, "loss": 0.6704, "step": 12763 }, { "epoch": 14.31, "learning_rate": 1.3440845070422534e-05, "loss": 0.6079, "step": 12764 }, { "epoch": 14.31, "learning_rate": 1.3439436619718309e-05, "loss": 0.6938, "step": 12765 }, { "epoch": 14.31, "learning_rate": 1.3438028169014085e-05, "loss": 0.4885, "step": 12766 }, { "epoch": 14.31, "learning_rate": 1.343661971830986e-05, "loss": 0.6199, "step": 12767 }, { "epoch": 14.31, "learning_rate": 1.3435211267605635e-05, "loss": 0.6363, "step": 12768 }, { "epoch": 14.32, "learning_rate": 1.3433802816901409e-05, "loss": 0.5532, "step": 12769 }, { "epoch": 14.32, "learning_rate": 1.3432394366197184e-05, "loss": 0.5502, "step": 12770 }, { "epoch": 14.32, "learning_rate": 1.3430985915492958e-05, "loss": 0.666, "step": 12771 }, { "epoch": 14.32, "learning_rate": 1.3429577464788731e-05, "loss": 0.631, "step": 12772 }, { "epoch": 14.32, "learning_rate": 1.3428169014084508e-05, "loss": 0.5114, "step": 12773 }, { "epoch": 14.32, "learning_rate": 1.3426760563380282e-05, "loss": 0.7393, "step": 12774 }, { "epoch": 14.32, "learning_rate": 1.3425352112676057e-05, "loss": 0.5627, "step": 12775 }, { "epoch": 14.32, "learning_rate": 1.3423943661971832e-05, "loss": 0.4613, "step": 12776 }, { "epoch": 14.32, "learning_rate": 1.3422535211267606e-05, "loss": 0.4742, "step": 12777 }, { "epoch": 14.33, "learning_rate": 1.3421126760563381e-05, "loss": 0.4821, "step": 12778 }, { "epoch": 14.33, "learning_rate": 1.3419718309859156e-05, "loss": 0.533, "step": 12779 }, { "epoch": 14.33, "learning_rate": 1.3418309859154928e-05, "loss": 0.5965, "step": 12780 }, { "epoch": 14.33, "learning_rate": 1.3416901408450705e-05, "loss": 0.7435, "step": 12781 }, { "epoch": 14.33, "learning_rate": 1.341549295774648e-05, "loss": 0.6031, "step": 12782 }, { "epoch": 14.33, "learning_rate": 1.3414084507042254e-05, "loss": 0.5737, "step": 12783 }, { "epoch": 14.33, "learning_rate": 1.3412676056338029e-05, "loss": 0.6389, "step": 12784 }, { "epoch": 14.33, "learning_rate": 1.3411267605633803e-05, "loss": 0.478, "step": 12785 }, { "epoch": 14.33, "learning_rate": 1.3409859154929578e-05, "loss": 0.5669, "step": 12786 }, { "epoch": 14.34, "learning_rate": 1.3408450704225353e-05, "loss": 0.5536, "step": 12787 }, { "epoch": 14.34, "learning_rate": 1.3407042253521127e-05, "loss": 0.6895, "step": 12788 }, { "epoch": 14.34, "learning_rate": 1.3405633802816902e-05, "loss": 0.8474, "step": 12789 }, { "epoch": 14.34, "learning_rate": 1.3404225352112677e-05, "loss": 0.7162, "step": 12790 }, { "epoch": 14.34, "learning_rate": 1.3402816901408451e-05, "loss": 0.7617, "step": 12791 }, { "epoch": 14.34, "learning_rate": 1.3401408450704226e-05, "loss": 0.7974, "step": 12792 }, { "epoch": 14.34, "learning_rate": 1.34e-05, "loss": 0.763, "step": 12793 }, { "epoch": 14.34, "learning_rate": 1.3398591549295775e-05, "loss": 0.6258, "step": 12794 }, { "epoch": 14.34, "learning_rate": 1.339718309859155e-05, "loss": 0.733, "step": 12795 }, { "epoch": 14.35, "learning_rate": 1.3395774647887324e-05, "loss": 0.6845, "step": 12796 }, { "epoch": 14.35, "learning_rate": 1.3394366197183099e-05, "loss": 0.7198, "step": 12797 }, { "epoch": 14.35, "learning_rate": 1.3392957746478874e-05, "loss": 0.6117, "step": 12798 }, { "epoch": 14.35, "learning_rate": 1.3391549295774648e-05, "loss": 0.6142, "step": 12799 }, { "epoch": 14.35, "learning_rate": 1.3390140845070423e-05, "loss": 0.5939, "step": 12800 }, { "epoch": 14.35, "learning_rate": 1.3388732394366198e-05, "loss": 0.6381, "step": 12801 }, { "epoch": 14.35, "learning_rate": 1.3387323943661972e-05, "loss": 0.7699, "step": 12802 }, { "epoch": 14.35, "learning_rate": 1.3385915492957747e-05, "loss": 0.688, "step": 12803 }, { "epoch": 14.35, "learning_rate": 1.3384507042253521e-05, "loss": 0.6843, "step": 12804 }, { "epoch": 14.36, "learning_rate": 1.3383098591549296e-05, "loss": 0.6114, "step": 12805 }, { "epoch": 14.36, "learning_rate": 1.338169014084507e-05, "loss": 0.6166, "step": 12806 }, { "epoch": 14.36, "learning_rate": 1.3380281690140845e-05, "loss": 0.5765, "step": 12807 }, { "epoch": 14.36, "learning_rate": 1.337887323943662e-05, "loss": 0.6497, "step": 12808 }, { "epoch": 14.36, "learning_rate": 1.3377464788732395e-05, "loss": 0.5543, "step": 12809 }, { "epoch": 14.36, "learning_rate": 1.337605633802817e-05, "loss": 0.5564, "step": 12810 }, { "epoch": 14.36, "learning_rate": 1.3374647887323944e-05, "loss": 0.5329, "step": 12811 }, { "epoch": 14.36, "learning_rate": 1.3373239436619719e-05, "loss": 0.5857, "step": 12812 }, { "epoch": 14.36, "learning_rate": 1.3371830985915493e-05, "loss": 0.5875, "step": 12813 }, { "epoch": 14.37, "learning_rate": 1.3370422535211268e-05, "loss": 0.7063, "step": 12814 }, { "epoch": 14.37, "learning_rate": 1.3369014084507042e-05, "loss": 0.5678, "step": 12815 }, { "epoch": 14.37, "learning_rate": 1.3367605633802817e-05, "loss": 0.5288, "step": 12816 }, { "epoch": 14.37, "learning_rate": 1.3366197183098592e-05, "loss": 0.6834, "step": 12817 }, { "epoch": 14.37, "learning_rate": 1.3364788732394366e-05, "loss": 0.5959, "step": 12818 }, { "epoch": 14.37, "learning_rate": 1.3363380281690141e-05, "loss": 0.593, "step": 12819 }, { "epoch": 14.37, "learning_rate": 1.3361971830985916e-05, "loss": 0.5672, "step": 12820 }, { "epoch": 14.37, "learning_rate": 1.336056338028169e-05, "loss": 0.5135, "step": 12821 }, { "epoch": 14.37, "learning_rate": 1.3359154929577465e-05, "loss": 0.6263, "step": 12822 }, { "epoch": 14.38, "learning_rate": 1.335774647887324e-05, "loss": 0.613, "step": 12823 }, { "epoch": 14.38, "learning_rate": 1.3356338028169014e-05, "loss": 0.7064, "step": 12824 }, { "epoch": 14.38, "learning_rate": 1.3354929577464789e-05, "loss": 0.5633, "step": 12825 }, { "epoch": 14.38, "learning_rate": 1.3353521126760565e-05, "loss": 0.4962, "step": 12826 }, { "epoch": 14.38, "learning_rate": 1.3352112676056338e-05, "loss": 0.5812, "step": 12827 }, { "epoch": 14.38, "learning_rate": 1.3350704225352113e-05, "loss": 0.6234, "step": 12828 }, { "epoch": 14.38, "learning_rate": 1.3349295774647887e-05, "loss": 0.5995, "step": 12829 }, { "epoch": 14.38, "learning_rate": 1.3347887323943662e-05, "loss": 0.6369, "step": 12830 }, { "epoch": 14.38, "learning_rate": 1.3346478873239437e-05, "loss": 0.5861, "step": 12831 }, { "epoch": 14.39, "learning_rate": 1.3345070422535211e-05, "loss": 0.5482, "step": 12832 }, { "epoch": 14.39, "learning_rate": 1.3343661971830986e-05, "loss": 0.573, "step": 12833 }, { "epoch": 14.39, "learning_rate": 1.3342253521126762e-05, "loss": 0.5505, "step": 12834 }, { "epoch": 14.39, "learning_rate": 1.3340845070422535e-05, "loss": 0.4373, "step": 12835 }, { "epoch": 14.39, "learning_rate": 1.333943661971831e-05, "loss": 0.6212, "step": 12836 }, { "epoch": 14.39, "learning_rate": 1.3338028169014085e-05, "loss": 0.6129, "step": 12837 }, { "epoch": 14.39, "learning_rate": 1.3336619718309859e-05, "loss": 0.5912, "step": 12838 }, { "epoch": 14.39, "learning_rate": 1.3335211267605634e-05, "loss": 0.9245, "step": 12839 }, { "epoch": 14.39, "learning_rate": 1.3333802816901408e-05, "loss": 0.8607, "step": 12840 }, { "epoch": 14.4, "learning_rate": 1.3332394366197185e-05, "loss": 0.795, "step": 12841 }, { "epoch": 14.4, "learning_rate": 1.333098591549296e-05, "loss": 0.6619, "step": 12842 }, { "epoch": 14.4, "learning_rate": 1.3329577464788732e-05, "loss": 0.7396, "step": 12843 }, { "epoch": 14.4, "learning_rate": 1.3328169014084507e-05, "loss": 0.6039, "step": 12844 }, { "epoch": 14.4, "learning_rate": 1.3326760563380282e-05, "loss": 0.6015, "step": 12845 }, { "epoch": 14.4, "learning_rate": 1.3325352112676056e-05, "loss": 0.6849, "step": 12846 }, { "epoch": 14.4, "learning_rate": 1.3323943661971831e-05, "loss": 0.6679, "step": 12847 }, { "epoch": 14.4, "learning_rate": 1.3322535211267606e-05, "loss": 0.5849, "step": 12848 }, { "epoch": 14.4, "learning_rate": 1.3321126760563382e-05, "loss": 0.6194, "step": 12849 }, { "epoch": 14.41, "learning_rate": 1.3319718309859156e-05, "loss": 0.5449, "step": 12850 }, { "epoch": 14.41, "learning_rate": 1.331830985915493e-05, "loss": 0.5271, "step": 12851 }, { "epoch": 14.41, "learning_rate": 1.3316901408450704e-05, "loss": 0.4951, "step": 12852 }, { "epoch": 14.41, "learning_rate": 1.3315492957746479e-05, "loss": 0.5138, "step": 12853 }, { "epoch": 14.41, "learning_rate": 1.3314084507042253e-05, "loss": 0.6805, "step": 12854 }, { "epoch": 14.41, "learning_rate": 1.3312676056338028e-05, "loss": 0.5335, "step": 12855 }, { "epoch": 14.41, "learning_rate": 1.3311267605633804e-05, "loss": 0.5664, "step": 12856 }, { "epoch": 14.41, "learning_rate": 1.3309859154929579e-05, "loss": 0.5532, "step": 12857 }, { "epoch": 14.41, "learning_rate": 1.3308450704225354e-05, "loss": 0.6374, "step": 12858 }, { "epoch": 14.42, "learning_rate": 1.3307042253521127e-05, "loss": 0.6347, "step": 12859 }, { "epoch": 14.42, "learning_rate": 1.3305633802816901e-05, "loss": 0.5956, "step": 12860 }, { "epoch": 14.42, "learning_rate": 1.3304225352112676e-05, "loss": 0.5512, "step": 12861 }, { "epoch": 14.42, "learning_rate": 1.330281690140845e-05, "loss": 0.5604, "step": 12862 }, { "epoch": 14.42, "learning_rate": 1.3301408450704225e-05, "loss": 0.542, "step": 12863 }, { "epoch": 14.42, "learning_rate": 1.3300000000000001e-05, "loss": 0.5553, "step": 12864 }, { "epoch": 14.42, "learning_rate": 1.3298591549295776e-05, "loss": 0.625, "step": 12865 }, { "epoch": 14.42, "learning_rate": 1.329718309859155e-05, "loss": 0.5453, "step": 12866 }, { "epoch": 14.42, "learning_rate": 1.3295774647887324e-05, "loss": 0.6615, "step": 12867 }, { "epoch": 14.43, "learning_rate": 1.3294366197183098e-05, "loss": 0.7864, "step": 12868 }, { "epoch": 14.43, "learning_rate": 1.3292957746478873e-05, "loss": 0.5696, "step": 12869 }, { "epoch": 14.43, "learning_rate": 1.3291549295774648e-05, "loss": 0.5124, "step": 12870 }, { "epoch": 14.43, "learning_rate": 1.3290140845070422e-05, "loss": 0.5439, "step": 12871 }, { "epoch": 14.43, "learning_rate": 1.3288732394366199e-05, "loss": 0.6498, "step": 12872 }, { "epoch": 14.43, "learning_rate": 1.3287323943661973e-05, "loss": 0.7031, "step": 12873 }, { "epoch": 14.43, "learning_rate": 1.3285915492957748e-05, "loss": 0.5452, "step": 12874 }, { "epoch": 14.43, "learning_rate": 1.328450704225352e-05, "loss": 0.6526, "step": 12875 }, { "epoch": 14.43, "learning_rate": 1.3283098591549295e-05, "loss": 0.6084, "step": 12876 }, { "epoch": 14.44, "learning_rate": 1.328169014084507e-05, "loss": 0.5639, "step": 12877 }, { "epoch": 14.44, "learning_rate": 1.3280281690140845e-05, "loss": 0.5133, "step": 12878 }, { "epoch": 14.44, "learning_rate": 1.3278873239436621e-05, "loss": 0.6631, "step": 12879 }, { "epoch": 14.44, "learning_rate": 1.3277464788732396e-05, "loss": 0.4799, "step": 12880 }, { "epoch": 14.44, "learning_rate": 1.327605633802817e-05, "loss": 0.7304, "step": 12881 }, { "epoch": 14.44, "learning_rate": 1.3274647887323945e-05, "loss": 0.674, "step": 12882 }, { "epoch": 14.44, "learning_rate": 1.3273239436619718e-05, "loss": 0.5576, "step": 12883 }, { "epoch": 14.44, "learning_rate": 1.3271830985915492e-05, "loss": 0.638, "step": 12884 }, { "epoch": 14.45, "learning_rate": 1.3270422535211267e-05, "loss": 0.4725, "step": 12885 }, { "epoch": 14.45, "learning_rate": 1.3269014084507042e-05, "loss": 0.5332, "step": 12886 }, { "epoch": 14.45, "learning_rate": 1.3267605633802818e-05, "loss": 0.5826, "step": 12887 }, { "epoch": 14.45, "learning_rate": 1.3266197183098593e-05, "loss": 0.4931, "step": 12888 }, { "epoch": 14.45, "learning_rate": 1.3264788732394367e-05, "loss": 0.856, "step": 12889 }, { "epoch": 14.45, "learning_rate": 1.3263380281690142e-05, "loss": 0.6573, "step": 12890 }, { "epoch": 14.45, "learning_rate": 1.3261971830985915e-05, "loss": 0.7117, "step": 12891 }, { "epoch": 14.45, "learning_rate": 1.326056338028169e-05, "loss": 0.6678, "step": 12892 }, { "epoch": 14.45, "learning_rate": 1.3259154929577464e-05, "loss": 0.7449, "step": 12893 }, { "epoch": 14.46, "learning_rate": 1.325774647887324e-05, "loss": 0.6422, "step": 12894 }, { "epoch": 14.46, "learning_rate": 1.3256338028169015e-05, "loss": 0.6652, "step": 12895 }, { "epoch": 14.46, "learning_rate": 1.325492957746479e-05, "loss": 0.6426, "step": 12896 }, { "epoch": 14.46, "learning_rate": 1.3253521126760564e-05, "loss": 0.7607, "step": 12897 }, { "epoch": 14.46, "learning_rate": 1.3252112676056339e-05, "loss": 0.7255, "step": 12898 }, { "epoch": 14.46, "learning_rate": 1.3250704225352112e-05, "loss": 0.6094, "step": 12899 }, { "epoch": 14.46, "learning_rate": 1.3249295774647887e-05, "loss": 0.6928, "step": 12900 }, { "epoch": 14.46, "learning_rate": 1.3247887323943661e-05, "loss": 0.5655, "step": 12901 }, { "epoch": 14.46, "learning_rate": 1.3246478873239438e-05, "loss": 0.6037, "step": 12902 }, { "epoch": 14.47, "learning_rate": 1.3245070422535212e-05, "loss": 0.5852, "step": 12903 }, { "epoch": 14.47, "learning_rate": 1.3243661971830987e-05, "loss": 0.6303, "step": 12904 }, { "epoch": 14.47, "learning_rate": 1.3242253521126762e-05, "loss": 0.509, "step": 12905 }, { "epoch": 14.47, "learning_rate": 1.3240845070422534e-05, "loss": 0.5779, "step": 12906 }, { "epoch": 14.47, "learning_rate": 1.3239436619718309e-05, "loss": 0.7054, "step": 12907 }, { "epoch": 14.47, "learning_rate": 1.3238028169014084e-05, "loss": 0.6865, "step": 12908 }, { "epoch": 14.47, "learning_rate": 1.323661971830986e-05, "loss": 0.7008, "step": 12909 }, { "epoch": 14.47, "learning_rate": 1.3235211267605635e-05, "loss": 0.5612, "step": 12910 }, { "epoch": 14.47, "learning_rate": 1.323380281690141e-05, "loss": 0.4765, "step": 12911 }, { "epoch": 14.48, "learning_rate": 1.3232394366197184e-05, "loss": 0.6313, "step": 12912 }, { "epoch": 14.48, "learning_rate": 1.3230985915492959e-05, "loss": 0.5013, "step": 12913 }, { "epoch": 14.48, "learning_rate": 1.3229577464788732e-05, "loss": 0.5327, "step": 12914 }, { "epoch": 14.48, "learning_rate": 1.3228169014084506e-05, "loss": 0.4678, "step": 12915 }, { "epoch": 14.48, "learning_rate": 1.3226760563380281e-05, "loss": 0.6173, "step": 12916 }, { "epoch": 14.48, "learning_rate": 1.3225352112676057e-05, "loss": 0.6067, "step": 12917 }, { "epoch": 14.48, "learning_rate": 1.3223943661971832e-05, "loss": 0.6533, "step": 12918 }, { "epoch": 14.48, "learning_rate": 1.3222535211267606e-05, "loss": 0.6353, "step": 12919 }, { "epoch": 14.48, "learning_rate": 1.3221126760563381e-05, "loss": 0.5768, "step": 12920 }, { "epoch": 14.49, "learning_rate": 1.3219718309859156e-05, "loss": 0.561, "step": 12921 }, { "epoch": 14.49, "learning_rate": 1.3218309859154929e-05, "loss": 0.5662, "step": 12922 }, { "epoch": 14.49, "learning_rate": 1.3216901408450703e-05, "loss": 0.4891, "step": 12923 }, { "epoch": 14.49, "learning_rate": 1.321549295774648e-05, "loss": 0.6088, "step": 12924 }, { "epoch": 14.49, "learning_rate": 1.3214084507042254e-05, "loss": 0.6797, "step": 12925 }, { "epoch": 14.49, "learning_rate": 1.3212676056338029e-05, "loss": 0.4912, "step": 12926 }, { "epoch": 14.49, "learning_rate": 1.3211267605633804e-05, "loss": 0.5258, "step": 12927 }, { "epoch": 14.49, "learning_rate": 1.3209859154929578e-05, "loss": 0.5865, "step": 12928 }, { "epoch": 14.49, "learning_rate": 1.3208450704225353e-05, "loss": 0.5081, "step": 12929 }, { "epoch": 14.5, "learning_rate": 1.3207042253521126e-05, "loss": 0.4821, "step": 12930 }, { "epoch": 14.5, "learning_rate": 1.32056338028169e-05, "loss": 0.4537, "step": 12931 }, { "epoch": 14.5, "learning_rate": 1.3204225352112677e-05, "loss": 0.675, "step": 12932 }, { "epoch": 14.5, "learning_rate": 1.3202816901408451e-05, "loss": 0.4661, "step": 12933 }, { "epoch": 14.5, "learning_rate": 1.3201408450704226e-05, "loss": 0.5424, "step": 12934 }, { "epoch": 14.5, "learning_rate": 1.32e-05, "loss": 0.6826, "step": 12935 }, { "epoch": 14.5, "learning_rate": 1.3198591549295775e-05, "loss": 0.5826, "step": 12936 }, { "epoch": 14.5, "learning_rate": 1.319718309859155e-05, "loss": 0.525, "step": 12937 }, { "epoch": 14.5, "learning_rate": 1.3195774647887323e-05, "loss": 0.6906, "step": 12938 }, { "epoch": 14.51, "learning_rate": 1.3194366197183098e-05, "loss": 0.7781, "step": 12939 }, { "epoch": 14.51, "learning_rate": 1.3192957746478874e-05, "loss": 0.8174, "step": 12940 }, { "epoch": 14.51, "learning_rate": 1.3191549295774648e-05, "loss": 0.6497, "step": 12941 }, { "epoch": 14.51, "learning_rate": 1.3190140845070423e-05, "loss": 0.5949, "step": 12942 }, { "epoch": 14.51, "learning_rate": 1.3188732394366198e-05, "loss": 0.4889, "step": 12943 }, { "epoch": 14.51, "learning_rate": 1.3187323943661972e-05, "loss": 0.7473, "step": 12944 }, { "epoch": 14.51, "learning_rate": 1.3185915492957747e-05, "loss": 0.6401, "step": 12945 }, { "epoch": 14.51, "learning_rate": 1.318450704225352e-05, "loss": 0.6026, "step": 12946 }, { "epoch": 14.51, "learning_rate": 1.3183098591549296e-05, "loss": 0.7384, "step": 12947 }, { "epoch": 14.52, "learning_rate": 1.3181690140845071e-05, "loss": 0.5896, "step": 12948 }, { "epoch": 14.52, "learning_rate": 1.3180281690140846e-05, "loss": 0.6356, "step": 12949 }, { "epoch": 14.52, "learning_rate": 1.317887323943662e-05, "loss": 0.6773, "step": 12950 }, { "epoch": 14.52, "learning_rate": 1.3177464788732395e-05, "loss": 0.5535, "step": 12951 }, { "epoch": 14.52, "learning_rate": 1.317605633802817e-05, "loss": 0.4968, "step": 12952 }, { "epoch": 14.52, "learning_rate": 1.3174647887323944e-05, "loss": 0.6999, "step": 12953 }, { "epoch": 14.52, "learning_rate": 1.3173239436619717e-05, "loss": 0.6324, "step": 12954 }, { "epoch": 14.52, "learning_rate": 1.3171830985915493e-05, "loss": 0.6137, "step": 12955 }, { "epoch": 14.52, "learning_rate": 1.3170422535211268e-05, "loss": 0.5259, "step": 12956 }, { "epoch": 14.53, "learning_rate": 1.3169014084507043e-05, "loss": 0.71, "step": 12957 }, { "epoch": 14.53, "learning_rate": 1.3167605633802817e-05, "loss": 0.6302, "step": 12958 }, { "epoch": 14.53, "learning_rate": 1.3166197183098592e-05, "loss": 0.6776, "step": 12959 }, { "epoch": 14.53, "learning_rate": 1.3164788732394367e-05, "loss": 0.5954, "step": 12960 }, { "epoch": 14.53, "learning_rate": 1.3163380281690141e-05, "loss": 0.6064, "step": 12961 }, { "epoch": 14.53, "learning_rate": 1.3161971830985916e-05, "loss": 0.5278, "step": 12962 }, { "epoch": 14.53, "learning_rate": 1.316056338028169e-05, "loss": 0.5462, "step": 12963 }, { "epoch": 14.53, "learning_rate": 1.3159154929577465e-05, "loss": 0.6181, "step": 12964 }, { "epoch": 14.53, "learning_rate": 1.315774647887324e-05, "loss": 0.4836, "step": 12965 }, { "epoch": 14.54, "learning_rate": 1.3156338028169014e-05, "loss": 0.517, "step": 12966 }, { "epoch": 14.54, "learning_rate": 1.3154929577464789e-05, "loss": 0.5463, "step": 12967 }, { "epoch": 14.54, "learning_rate": 1.3153521126760564e-05, "loss": 0.4972, "step": 12968 }, { "epoch": 14.54, "learning_rate": 1.3152112676056338e-05, "loss": 0.5039, "step": 12969 }, { "epoch": 14.54, "learning_rate": 1.3150704225352113e-05, "loss": 0.6639, "step": 12970 }, { "epoch": 14.54, "learning_rate": 1.3149295774647888e-05, "loss": 0.5713, "step": 12971 }, { "epoch": 14.54, "learning_rate": 1.3147887323943662e-05, "loss": 0.5903, "step": 12972 }, { "epoch": 14.54, "learning_rate": 1.3146478873239437e-05, "loss": 0.6566, "step": 12973 }, { "epoch": 14.54, "learning_rate": 1.3145070422535212e-05, "loss": 0.6118, "step": 12974 }, { "epoch": 14.55, "learning_rate": 1.3143661971830986e-05, "loss": 0.5349, "step": 12975 }, { "epoch": 14.55, "learning_rate": 1.314225352112676e-05, "loss": 0.5736, "step": 12976 }, { "epoch": 14.55, "learning_rate": 1.3140845070422537e-05, "loss": 0.3924, "step": 12977 }, { "epoch": 14.55, "learning_rate": 1.313943661971831e-05, "loss": 0.4979, "step": 12978 }, { "epoch": 14.55, "learning_rate": 1.3138028169014085e-05, "loss": 0.8018, "step": 12979 }, { "epoch": 14.55, "learning_rate": 1.313661971830986e-05, "loss": 0.6582, "step": 12980 }, { "epoch": 14.55, "learning_rate": 1.3135211267605634e-05, "loss": 0.5806, "step": 12981 }, { "epoch": 14.55, "learning_rate": 1.3133802816901409e-05, "loss": 0.5877, "step": 12982 }, { "epoch": 14.55, "learning_rate": 1.3132394366197183e-05, "loss": 0.656, "step": 12983 }, { "epoch": 14.56, "learning_rate": 1.3130985915492958e-05, "loss": 0.4993, "step": 12984 }, { "epoch": 14.56, "learning_rate": 1.3129577464788734e-05, "loss": 0.5146, "step": 12985 }, { "epoch": 14.56, "learning_rate": 1.3128169014084507e-05, "loss": 0.4866, "step": 12986 }, { "epoch": 14.56, "learning_rate": 1.3126760563380282e-05, "loss": 0.419, "step": 12987 }, { "epoch": 14.56, "learning_rate": 1.3125352112676056e-05, "loss": 0.6055, "step": 12988 }, { "epoch": 14.56, "learning_rate": 1.3123943661971831e-05, "loss": 0.7833, "step": 12989 }, { "epoch": 14.56, "learning_rate": 1.3122535211267606e-05, "loss": 0.5634, "step": 12990 }, { "epoch": 14.56, "learning_rate": 1.312112676056338e-05, "loss": 0.8416, "step": 12991 }, { "epoch": 14.57, "learning_rate": 1.3119718309859155e-05, "loss": 0.65, "step": 12992 }, { "epoch": 14.57, "learning_rate": 1.3118309859154931e-05, "loss": 0.567, "step": 12993 }, { "epoch": 14.57, "learning_rate": 1.3116901408450704e-05, "loss": 0.6181, "step": 12994 }, { "epoch": 14.57, "learning_rate": 1.3115492957746479e-05, "loss": 0.699, "step": 12995 }, { "epoch": 14.57, "learning_rate": 1.3114084507042254e-05, "loss": 0.5937, "step": 12996 }, { "epoch": 14.57, "learning_rate": 1.3112676056338028e-05, "loss": 0.5546, "step": 12997 }, { "epoch": 14.57, "learning_rate": 1.3111267605633803e-05, "loss": 0.6721, "step": 12998 }, { "epoch": 14.57, "learning_rate": 1.3109859154929577e-05, "loss": 0.5458, "step": 12999 }, { "epoch": 14.57, "learning_rate": 1.3108450704225354e-05, "loss": 0.6683, "step": 13000 }, { "epoch": 14.58, "learning_rate": 1.3107042253521128e-05, "loss": 0.705, "step": 13001 }, { "epoch": 14.58, "learning_rate": 1.3105633802816901e-05, "loss": 0.5537, "step": 13002 }, { "epoch": 14.58, "learning_rate": 1.3104225352112676e-05, "loss": 0.6791, "step": 13003 }, { "epoch": 14.58, "learning_rate": 1.310281690140845e-05, "loss": 0.7381, "step": 13004 }, { "epoch": 14.58, "learning_rate": 1.3101408450704225e-05, "loss": 0.562, "step": 13005 }, { "epoch": 14.58, "learning_rate": 1.31e-05, "loss": 0.6095, "step": 13006 }, { "epoch": 14.58, "learning_rate": 1.3098591549295775e-05, "loss": 0.5348, "step": 13007 }, { "epoch": 14.58, "learning_rate": 1.3097183098591551e-05, "loss": 0.6497, "step": 13008 }, { "epoch": 14.58, "learning_rate": 1.3095774647887326e-05, "loss": 0.5535, "step": 13009 }, { "epoch": 14.59, "learning_rate": 1.3094366197183098e-05, "loss": 0.6556, "step": 13010 }, { "epoch": 14.59, "learning_rate": 1.3092957746478873e-05, "loss": 0.5206, "step": 13011 }, { "epoch": 14.59, "learning_rate": 1.3091549295774648e-05, "loss": 0.5296, "step": 13012 }, { "epoch": 14.59, "learning_rate": 1.3090140845070422e-05, "loss": 0.7424, "step": 13013 }, { "epoch": 14.59, "learning_rate": 1.3088732394366197e-05, "loss": 0.6003, "step": 13014 }, { "epoch": 14.59, "learning_rate": 1.3087323943661973e-05, "loss": 0.576, "step": 13015 }, { "epoch": 14.59, "learning_rate": 1.3085915492957748e-05, "loss": 0.5862, "step": 13016 }, { "epoch": 14.59, "learning_rate": 1.3084507042253521e-05, "loss": 0.5558, "step": 13017 }, { "epoch": 14.59, "learning_rate": 1.3083098591549296e-05, "loss": 0.6727, "step": 13018 }, { "epoch": 14.6, "learning_rate": 1.308169014084507e-05, "loss": 0.6554, "step": 13019 }, { "epoch": 14.6, "learning_rate": 1.3080281690140845e-05, "loss": 0.4725, "step": 13020 }, { "epoch": 14.6, "learning_rate": 1.307887323943662e-05, "loss": 0.6516, "step": 13021 }, { "epoch": 14.6, "learning_rate": 1.3077464788732394e-05, "loss": 0.7377, "step": 13022 }, { "epoch": 14.6, "learning_rate": 1.307605633802817e-05, "loss": 0.678, "step": 13023 }, { "epoch": 14.6, "learning_rate": 1.3074647887323945e-05, "loss": 0.4974, "step": 13024 }, { "epoch": 14.6, "learning_rate": 1.3073239436619718e-05, "loss": 0.5156, "step": 13025 }, { "epoch": 14.6, "learning_rate": 1.3071830985915493e-05, "loss": 0.5363, "step": 13026 }, { "epoch": 14.6, "learning_rate": 1.3070422535211267e-05, "loss": 0.5636, "step": 13027 }, { "epoch": 14.61, "learning_rate": 1.3069014084507042e-05, "loss": 0.6596, "step": 13028 }, { "epoch": 14.61, "learning_rate": 1.3067605633802817e-05, "loss": 0.7263, "step": 13029 }, { "epoch": 14.61, "learning_rate": 1.3066197183098593e-05, "loss": 0.5158, "step": 13030 }, { "epoch": 14.61, "learning_rate": 1.3064788732394368e-05, "loss": 0.4698, "step": 13031 }, { "epoch": 14.61, "learning_rate": 1.3063380281690142e-05, "loss": 0.6012, "step": 13032 }, { "epoch": 14.61, "learning_rate": 1.3061971830985915e-05, "loss": 0.5078, "step": 13033 }, { "epoch": 14.61, "learning_rate": 1.306056338028169e-05, "loss": 0.4063, "step": 13034 }, { "epoch": 14.61, "learning_rate": 1.3059154929577464e-05, "loss": 0.4202, "step": 13035 }, { "epoch": 14.61, "learning_rate": 1.3057746478873239e-05, "loss": 0.4634, "step": 13036 }, { "epoch": 14.62, "learning_rate": 1.3056338028169014e-05, "loss": 0.423, "step": 13037 }, { "epoch": 14.62, "learning_rate": 1.305492957746479e-05, "loss": 0.5918, "step": 13038 }, { "epoch": 14.62, "learning_rate": 1.3053521126760565e-05, "loss": 0.6719, "step": 13039 }, { "epoch": 14.62, "learning_rate": 1.305211267605634e-05, "loss": 0.6524, "step": 13040 }, { "epoch": 14.62, "learning_rate": 1.3050704225352112e-05, "loss": 0.7487, "step": 13041 }, { "epoch": 14.62, "learning_rate": 1.3049295774647887e-05, "loss": 0.5986, "step": 13042 }, { "epoch": 14.62, "learning_rate": 1.3047887323943662e-05, "loss": 0.537, "step": 13043 }, { "epoch": 14.62, "learning_rate": 1.3046478873239436e-05, "loss": 0.6178, "step": 13044 }, { "epoch": 14.62, "learning_rate": 1.3045070422535212e-05, "loss": 0.6661, "step": 13045 }, { "epoch": 14.63, "learning_rate": 1.3043661971830987e-05, "loss": 0.6468, "step": 13046 }, { "epoch": 14.63, "learning_rate": 1.3042253521126762e-05, "loss": 0.6113, "step": 13047 }, { "epoch": 14.63, "learning_rate": 1.3040845070422536e-05, "loss": 0.7746, "step": 13048 }, { "epoch": 14.63, "learning_rate": 1.303943661971831e-05, "loss": 0.6396, "step": 13049 }, { "epoch": 14.63, "learning_rate": 1.3038028169014084e-05, "loss": 0.7312, "step": 13050 }, { "epoch": 14.63, "learning_rate": 1.3036619718309859e-05, "loss": 0.643, "step": 13051 }, { "epoch": 14.63, "learning_rate": 1.3035211267605633e-05, "loss": 0.5919, "step": 13052 }, { "epoch": 14.63, "learning_rate": 1.303380281690141e-05, "loss": 0.5616, "step": 13053 }, { "epoch": 14.63, "learning_rate": 1.3032394366197184e-05, "loss": 0.6614, "step": 13054 }, { "epoch": 14.64, "learning_rate": 1.3030985915492959e-05, "loss": 0.6343, "step": 13055 }, { "epoch": 14.64, "learning_rate": 1.3029577464788734e-05, "loss": 0.6177, "step": 13056 }, { "epoch": 14.64, "learning_rate": 1.3028169014084506e-05, "loss": 0.6881, "step": 13057 }, { "epoch": 14.64, "learning_rate": 1.3026760563380281e-05, "loss": 0.6342, "step": 13058 }, { "epoch": 14.64, "learning_rate": 1.3025352112676056e-05, "loss": 0.5326, "step": 13059 }, { "epoch": 14.64, "learning_rate": 1.302394366197183e-05, "loss": 0.6381, "step": 13060 }, { "epoch": 14.64, "learning_rate": 1.3022535211267607e-05, "loss": 0.5858, "step": 13061 }, { "epoch": 14.64, "learning_rate": 1.3021126760563381e-05, "loss": 0.5745, "step": 13062 }, { "epoch": 14.64, "learning_rate": 1.3019718309859156e-05, "loss": 0.5485, "step": 13063 }, { "epoch": 14.65, "learning_rate": 1.301830985915493e-05, "loss": 0.5735, "step": 13064 }, { "epoch": 14.65, "learning_rate": 1.3016901408450704e-05, "loss": 0.5413, "step": 13065 }, { "epoch": 14.65, "learning_rate": 1.3015492957746478e-05, "loss": 0.5739, "step": 13066 }, { "epoch": 14.65, "learning_rate": 1.3014084507042253e-05, "loss": 0.6351, "step": 13067 }, { "epoch": 14.65, "learning_rate": 1.301267605633803e-05, "loss": 0.6081, "step": 13068 }, { "epoch": 14.65, "learning_rate": 1.3011267605633804e-05, "loss": 0.6461, "step": 13069 }, { "epoch": 14.65, "learning_rate": 1.3009859154929578e-05, "loss": 0.6264, "step": 13070 }, { "epoch": 14.65, "learning_rate": 1.3008450704225353e-05, "loss": 0.548, "step": 13071 }, { "epoch": 14.65, "learning_rate": 1.3007042253521128e-05, "loss": 0.4993, "step": 13072 }, { "epoch": 14.66, "learning_rate": 1.30056338028169e-05, "loss": 0.6134, "step": 13073 }, { "epoch": 14.66, "learning_rate": 1.3004225352112675e-05, "loss": 0.5478, "step": 13074 }, { "epoch": 14.66, "learning_rate": 1.300281690140845e-05, "loss": 0.5358, "step": 13075 }, { "epoch": 14.66, "learning_rate": 1.3001408450704226e-05, "loss": 0.667, "step": 13076 }, { "epoch": 14.66, "learning_rate": 1.3000000000000001e-05, "loss": 0.647, "step": 13077 }, { "epoch": 14.66, "learning_rate": 1.2998591549295776e-05, "loss": 0.4676, "step": 13078 }, { "epoch": 14.66, "learning_rate": 1.299718309859155e-05, "loss": 0.569, "step": 13079 }, { "epoch": 14.66, "learning_rate": 1.2995774647887325e-05, "loss": 0.5239, "step": 13080 }, { "epoch": 14.66, "learning_rate": 1.2994366197183098e-05, "loss": 0.6568, "step": 13081 }, { "epoch": 14.67, "learning_rate": 1.2992957746478872e-05, "loss": 0.5141, "step": 13082 }, { "epoch": 14.67, "learning_rate": 1.2991549295774649e-05, "loss": 0.3953, "step": 13083 }, { "epoch": 14.67, "learning_rate": 1.2990140845070423e-05, "loss": 0.5451, "step": 13084 }, { "epoch": 14.67, "learning_rate": 1.2988732394366198e-05, "loss": 0.4317, "step": 13085 }, { "epoch": 14.67, "learning_rate": 1.2987323943661973e-05, "loss": 0.6363, "step": 13086 }, { "epoch": 14.67, "learning_rate": 1.2985915492957747e-05, "loss": 0.6229, "step": 13087 }, { "epoch": 14.67, "learning_rate": 1.2984507042253522e-05, "loss": 0.5665, "step": 13088 }, { "epoch": 14.67, "learning_rate": 1.2983098591549295e-05, "loss": 0.7868, "step": 13089 }, { "epoch": 14.67, "learning_rate": 1.298169014084507e-05, "loss": 0.7102, "step": 13090 }, { "epoch": 14.68, "learning_rate": 1.2980281690140846e-05, "loss": 0.7548, "step": 13091 }, { "epoch": 14.68, "learning_rate": 1.297887323943662e-05, "loss": 0.4613, "step": 13092 }, { "epoch": 14.68, "learning_rate": 1.2977464788732395e-05, "loss": 0.6138, "step": 13093 }, { "epoch": 14.68, "learning_rate": 1.297605633802817e-05, "loss": 0.5079, "step": 13094 }, { "epoch": 14.68, "learning_rate": 1.2974647887323944e-05, "loss": 0.7204, "step": 13095 }, { "epoch": 14.68, "learning_rate": 1.2973239436619719e-05, "loss": 0.5298, "step": 13096 }, { "epoch": 14.68, "learning_rate": 1.2971830985915492e-05, "loss": 0.5006, "step": 13097 }, { "epoch": 14.68, "learning_rate": 1.2970422535211268e-05, "loss": 0.4793, "step": 13098 }, { "epoch": 14.68, "learning_rate": 1.2969014084507043e-05, "loss": 0.6434, "step": 13099 }, { "epoch": 14.69, "learning_rate": 1.2967605633802818e-05, "loss": 0.563, "step": 13100 }, { "epoch": 14.69, "learning_rate": 1.2966197183098592e-05, "loss": 0.7064, "step": 13101 }, { "epoch": 14.69, "learning_rate": 1.2964788732394367e-05, "loss": 0.5857, "step": 13102 }, { "epoch": 14.69, "learning_rate": 1.2963380281690141e-05, "loss": 0.7061, "step": 13103 }, { "epoch": 14.69, "learning_rate": 1.2961971830985916e-05, "loss": 0.6316, "step": 13104 }, { "epoch": 14.69, "learning_rate": 1.2960563380281689e-05, "loss": 0.6197, "step": 13105 }, { "epoch": 14.69, "learning_rate": 1.2959154929577465e-05, "loss": 0.5937, "step": 13106 }, { "epoch": 14.69, "learning_rate": 1.295774647887324e-05, "loss": 0.6882, "step": 13107 }, { "epoch": 14.7, "learning_rate": 1.2956338028169015e-05, "loss": 0.6711, "step": 13108 }, { "epoch": 14.7, "learning_rate": 1.295492957746479e-05, "loss": 0.5444, "step": 13109 }, { "epoch": 14.7, "learning_rate": 1.2953521126760564e-05, "loss": 0.5779, "step": 13110 }, { "epoch": 14.7, "learning_rate": 1.2952112676056339e-05, "loss": 0.5519, "step": 13111 }, { "epoch": 14.7, "learning_rate": 1.2950704225352113e-05, "loss": 0.5452, "step": 13112 }, { "epoch": 14.7, "learning_rate": 1.2949295774647886e-05, "loss": 0.5049, "step": 13113 }, { "epoch": 14.7, "learning_rate": 1.2947887323943662e-05, "loss": 0.6702, "step": 13114 }, { "epoch": 14.7, "learning_rate": 1.2946478873239437e-05, "loss": 0.6046, "step": 13115 }, { "epoch": 14.7, "learning_rate": 1.2945070422535212e-05, "loss": 0.576, "step": 13116 }, { "epoch": 14.71, "learning_rate": 1.2943661971830986e-05, "loss": 0.6218, "step": 13117 }, { "epoch": 14.71, "learning_rate": 1.2942253521126761e-05, "loss": 0.6131, "step": 13118 }, { "epoch": 14.71, "learning_rate": 1.2940845070422536e-05, "loss": 0.6355, "step": 13119 }, { "epoch": 14.71, "learning_rate": 1.293943661971831e-05, "loss": 0.4193, "step": 13120 }, { "epoch": 14.71, "learning_rate": 1.2938028169014085e-05, "loss": 0.5544, "step": 13121 }, { "epoch": 14.71, "learning_rate": 1.293661971830986e-05, "loss": 0.4846, "step": 13122 }, { "epoch": 14.71, "learning_rate": 1.2935211267605634e-05, "loss": 0.7176, "step": 13123 }, { "epoch": 14.71, "learning_rate": 1.2933802816901409e-05, "loss": 0.5346, "step": 13124 }, { "epoch": 14.71, "learning_rate": 1.2932394366197183e-05, "loss": 0.638, "step": 13125 }, { "epoch": 14.72, "learning_rate": 1.2930985915492958e-05, "loss": 0.5931, "step": 13126 }, { "epoch": 14.72, "learning_rate": 1.2929577464788733e-05, "loss": 0.5363, "step": 13127 }, { "epoch": 14.72, "learning_rate": 1.2928169014084507e-05, "loss": 0.7327, "step": 13128 }, { "epoch": 14.72, "learning_rate": 1.2926760563380282e-05, "loss": 0.5471, "step": 13129 }, { "epoch": 14.72, "learning_rate": 1.2925352112676057e-05, "loss": 0.5141, "step": 13130 }, { "epoch": 14.72, "learning_rate": 1.2923943661971831e-05, "loss": 0.5028, "step": 13131 }, { "epoch": 14.72, "learning_rate": 1.2922535211267606e-05, "loss": 0.6213, "step": 13132 }, { "epoch": 14.72, "learning_rate": 1.292112676056338e-05, "loss": 0.4416, "step": 13133 }, { "epoch": 14.72, "learning_rate": 1.2919718309859155e-05, "loss": 0.5816, "step": 13134 }, { "epoch": 14.73, "learning_rate": 1.291830985915493e-05, "loss": 0.4561, "step": 13135 }, { "epoch": 14.73, "learning_rate": 1.2916901408450705e-05, "loss": 0.3827, "step": 13136 }, { "epoch": 14.73, "learning_rate": 1.2915492957746479e-05, "loss": 0.5781, "step": 13137 }, { "epoch": 14.73, "learning_rate": 1.2914084507042254e-05, "loss": 0.4854, "step": 13138 }, { "epoch": 14.73, "learning_rate": 1.2912676056338028e-05, "loss": 0.8031, "step": 13139 }, { "epoch": 14.73, "learning_rate": 1.2911267605633803e-05, "loss": 0.7313, "step": 13140 }, { "epoch": 14.73, "learning_rate": 1.2909859154929578e-05, "loss": 0.8776, "step": 13141 }, { "epoch": 14.73, "learning_rate": 1.2908450704225352e-05, "loss": 0.6782, "step": 13142 }, { "epoch": 14.73, "learning_rate": 1.2907042253521127e-05, "loss": 0.533, "step": 13143 }, { "epoch": 14.74, "learning_rate": 1.2905633802816902e-05, "loss": 0.6755, "step": 13144 }, { "epoch": 14.74, "learning_rate": 1.2904225352112676e-05, "loss": 0.7013, "step": 13145 }, { "epoch": 14.74, "learning_rate": 1.2902816901408451e-05, "loss": 0.7002, "step": 13146 }, { "epoch": 14.74, "learning_rate": 1.2901408450704226e-05, "loss": 0.6361, "step": 13147 }, { "epoch": 14.74, "learning_rate": 1.29e-05, "loss": 0.5917, "step": 13148 }, { "epoch": 14.74, "learning_rate": 1.2898591549295775e-05, "loss": 0.6245, "step": 13149 }, { "epoch": 14.74, "learning_rate": 1.289718309859155e-05, "loss": 0.6223, "step": 13150 }, { "epoch": 14.74, "learning_rate": 1.2895774647887326e-05, "loss": 0.6575, "step": 13151 }, { "epoch": 14.74, "learning_rate": 1.2894366197183099e-05, "loss": 0.6429, "step": 13152 }, { "epoch": 14.75, "learning_rate": 1.2892957746478873e-05, "loss": 0.6409, "step": 13153 }, { "epoch": 14.75, "learning_rate": 1.2891549295774648e-05, "loss": 0.6458, "step": 13154 }, { "epoch": 14.75, "learning_rate": 1.2890140845070423e-05, "loss": 0.6097, "step": 13155 }, { "epoch": 14.75, "learning_rate": 1.2888732394366197e-05, "loss": 0.6909, "step": 13156 }, { "epoch": 14.75, "learning_rate": 1.2887323943661972e-05, "loss": 0.7762, "step": 13157 }, { "epoch": 14.75, "learning_rate": 1.2885915492957747e-05, "loss": 0.5407, "step": 13158 }, { "epoch": 14.75, "learning_rate": 1.2884507042253523e-05, "loss": 0.4633, "step": 13159 }, { "epoch": 14.75, "learning_rate": 1.2883098591549296e-05, "loss": 0.6587, "step": 13160 }, { "epoch": 14.75, "learning_rate": 1.288169014084507e-05, "loss": 0.6388, "step": 13161 }, { "epoch": 14.76, "learning_rate": 1.2880281690140845e-05, "loss": 0.6075, "step": 13162 }, { "epoch": 14.76, "learning_rate": 1.287887323943662e-05, "loss": 0.6126, "step": 13163 }, { "epoch": 14.76, "learning_rate": 1.2877464788732394e-05, "loss": 0.5073, "step": 13164 }, { "epoch": 14.76, "learning_rate": 1.2876056338028169e-05, "loss": 0.5193, "step": 13165 }, { "epoch": 14.76, "learning_rate": 1.2874647887323944e-05, "loss": 0.5194, "step": 13166 }, { "epoch": 14.76, "learning_rate": 1.287323943661972e-05, "loss": 0.6128, "step": 13167 }, { "epoch": 14.76, "learning_rate": 1.2871830985915493e-05, "loss": 0.5838, "step": 13168 }, { "epoch": 14.76, "learning_rate": 1.2870422535211268e-05, "loss": 0.7881, "step": 13169 }, { "epoch": 14.76, "learning_rate": 1.2869014084507042e-05, "loss": 0.4461, "step": 13170 }, { "epoch": 14.77, "learning_rate": 1.2867605633802817e-05, "loss": 0.5177, "step": 13171 }, { "epoch": 14.77, "learning_rate": 1.2866197183098591e-05, "loss": 0.5805, "step": 13172 }, { "epoch": 14.77, "learning_rate": 1.2864788732394366e-05, "loss": 0.483, "step": 13173 }, { "epoch": 14.77, "learning_rate": 1.2863380281690142e-05, "loss": 0.5592, "step": 13174 }, { "epoch": 14.77, "learning_rate": 1.2861971830985917e-05, "loss": 0.42, "step": 13175 }, { "epoch": 14.77, "learning_rate": 1.286056338028169e-05, "loss": 0.5263, "step": 13176 }, { "epoch": 14.77, "learning_rate": 1.2859154929577465e-05, "loss": 0.5582, "step": 13177 }, { "epoch": 14.77, "learning_rate": 1.285774647887324e-05, "loss": 0.5294, "step": 13178 }, { "epoch": 14.77, "learning_rate": 1.2856338028169014e-05, "loss": 0.651, "step": 13179 }, { "epoch": 14.78, "learning_rate": 1.2854929577464789e-05, "loss": 0.5847, "step": 13180 }, { "epoch": 14.78, "learning_rate": 1.2853521126760563e-05, "loss": 0.7205, "step": 13181 }, { "epoch": 14.78, "learning_rate": 1.285211267605634e-05, "loss": 0.5635, "step": 13182 }, { "epoch": 14.78, "learning_rate": 1.2850704225352114e-05, "loss": 0.4717, "step": 13183 }, { "epoch": 14.78, "learning_rate": 1.2849295774647887e-05, "loss": 0.4266, "step": 13184 }, { "epoch": 14.78, "learning_rate": 1.2847887323943662e-05, "loss": 0.4131, "step": 13185 }, { "epoch": 14.78, "learning_rate": 1.2846478873239436e-05, "loss": 0.4597, "step": 13186 }, { "epoch": 14.78, "learning_rate": 1.2845070422535211e-05, "loss": 0.569, "step": 13187 }, { "epoch": 14.78, "learning_rate": 1.2843661971830986e-05, "loss": 0.4254, "step": 13188 }, { "epoch": 14.79, "learning_rate": 1.2842253521126762e-05, "loss": 0.522, "step": 13189 }, { "epoch": 14.79, "learning_rate": 1.2840845070422537e-05, "loss": 0.7128, "step": 13190 }, { "epoch": 14.79, "learning_rate": 1.2839436619718311e-05, "loss": 0.5272, "step": 13191 }, { "epoch": 14.79, "learning_rate": 1.2838028169014084e-05, "loss": 0.6084, "step": 13192 }, { "epoch": 14.79, "learning_rate": 1.2836619718309859e-05, "loss": 0.5603, "step": 13193 }, { "epoch": 14.79, "learning_rate": 1.2835211267605633e-05, "loss": 0.5686, "step": 13194 }, { "epoch": 14.79, "learning_rate": 1.2833802816901408e-05, "loss": 0.6778, "step": 13195 }, { "epoch": 14.79, "learning_rate": 1.2832394366197183e-05, "loss": 0.5857, "step": 13196 }, { "epoch": 14.79, "learning_rate": 1.2830985915492959e-05, "loss": 0.6799, "step": 13197 }, { "epoch": 14.8, "learning_rate": 1.2829577464788734e-05, "loss": 0.6875, "step": 13198 }, { "epoch": 14.8, "learning_rate": 1.2828169014084508e-05, "loss": 0.6097, "step": 13199 }, { "epoch": 14.8, "learning_rate": 1.2826760563380281e-05, "loss": 0.5372, "step": 13200 }, { "epoch": 14.8, "learning_rate": 1.2825352112676056e-05, "loss": 0.5803, "step": 13201 }, { "epoch": 14.8, "learning_rate": 1.282394366197183e-05, "loss": 0.6585, "step": 13202 }, { "epoch": 14.8, "learning_rate": 1.2822535211267605e-05, "loss": 0.6597, "step": 13203 }, { "epoch": 14.8, "learning_rate": 1.2821126760563382e-05, "loss": 0.6449, "step": 13204 }, { "epoch": 14.8, "learning_rate": 1.2819718309859156e-05, "loss": 0.6292, "step": 13205 }, { "epoch": 14.8, "learning_rate": 1.281830985915493e-05, "loss": 0.6616, "step": 13206 }, { "epoch": 14.81, "learning_rate": 1.2816901408450705e-05, "loss": 0.6463, "step": 13207 }, { "epoch": 14.81, "learning_rate": 1.2815492957746478e-05, "loss": 0.5109, "step": 13208 }, { "epoch": 14.81, "learning_rate": 1.2814084507042253e-05, "loss": 0.6277, "step": 13209 }, { "epoch": 14.81, "learning_rate": 1.2812676056338028e-05, "loss": 0.556, "step": 13210 }, { "epoch": 14.81, "learning_rate": 1.2811267605633802e-05, "loss": 0.6309, "step": 13211 }, { "epoch": 14.81, "learning_rate": 1.2809859154929579e-05, "loss": 0.4283, "step": 13212 }, { "epoch": 14.81, "learning_rate": 1.2808450704225353e-05, "loss": 0.6121, "step": 13213 }, { "epoch": 14.81, "learning_rate": 1.2807042253521128e-05, "loss": 0.612, "step": 13214 }, { "epoch": 14.82, "learning_rate": 1.2805633802816903e-05, "loss": 0.5318, "step": 13215 }, { "epoch": 14.82, "learning_rate": 1.2804225352112676e-05, "loss": 0.4979, "step": 13216 }, { "epoch": 14.82, "learning_rate": 1.280281690140845e-05, "loss": 0.5204, "step": 13217 }, { "epoch": 14.82, "learning_rate": 1.2801408450704225e-05, "loss": 0.5093, "step": 13218 }, { "epoch": 14.82, "learning_rate": 1.2800000000000001e-05, "loss": 0.5125, "step": 13219 }, { "epoch": 14.82, "learning_rate": 1.2798591549295776e-05, "loss": 0.6886, "step": 13220 }, { "epoch": 14.82, "learning_rate": 1.279718309859155e-05, "loss": 0.5055, "step": 13221 }, { "epoch": 14.82, "learning_rate": 1.2795774647887325e-05, "loss": 0.6425, "step": 13222 }, { "epoch": 14.82, "learning_rate": 1.27943661971831e-05, "loss": 0.6054, "step": 13223 }, { "epoch": 14.83, "learning_rate": 1.2792957746478873e-05, "loss": 0.5789, "step": 13224 }, { "epoch": 14.83, "learning_rate": 1.2791549295774647e-05, "loss": 0.5632, "step": 13225 }, { "epoch": 14.83, "learning_rate": 1.2790140845070422e-05, "loss": 0.4401, "step": 13226 }, { "epoch": 14.83, "learning_rate": 1.2788732394366198e-05, "loss": 0.566, "step": 13227 }, { "epoch": 14.83, "learning_rate": 1.2787323943661973e-05, "loss": 0.6149, "step": 13228 }, { "epoch": 14.83, "learning_rate": 1.2785915492957747e-05, "loss": 0.5531, "step": 13229 }, { "epoch": 14.83, "learning_rate": 1.2784507042253522e-05, "loss": 0.5308, "step": 13230 }, { "epoch": 14.83, "learning_rate": 1.2783098591549297e-05, "loss": 0.5677, "step": 13231 }, { "epoch": 14.83, "learning_rate": 1.278169014084507e-05, "loss": 0.5833, "step": 13232 }, { "epoch": 14.84, "learning_rate": 1.2780281690140844e-05, "loss": 0.5328, "step": 13233 }, { "epoch": 14.84, "learning_rate": 1.2778873239436619e-05, "loss": 0.4096, "step": 13234 }, { "epoch": 14.84, "learning_rate": 1.2777464788732395e-05, "loss": 0.6245, "step": 13235 }, { "epoch": 14.84, "learning_rate": 1.277605633802817e-05, "loss": 0.627, "step": 13236 }, { "epoch": 14.84, "learning_rate": 1.2774647887323945e-05, "loss": 0.6794, "step": 13237 }, { "epoch": 14.84, "learning_rate": 1.277323943661972e-05, "loss": 0.6048, "step": 13238 }, { "epoch": 14.84, "learning_rate": 1.2771830985915494e-05, "loss": 0.8449, "step": 13239 }, { "epoch": 14.84, "learning_rate": 1.2770422535211267e-05, "loss": 0.6855, "step": 13240 }, { "epoch": 14.84, "learning_rate": 1.2769014084507041e-05, "loss": 0.6893, "step": 13241 }, { "epoch": 14.85, "learning_rate": 1.2767605633802818e-05, "loss": 0.6629, "step": 13242 }, { "epoch": 14.85, "learning_rate": 1.2766197183098592e-05, "loss": 0.7542, "step": 13243 }, { "epoch": 14.85, "learning_rate": 1.2764788732394367e-05, "loss": 0.6115, "step": 13244 }, { "epoch": 14.85, "learning_rate": 1.2763380281690142e-05, "loss": 0.5135, "step": 13245 }, { "epoch": 14.85, "learning_rate": 1.2761971830985916e-05, "loss": 0.6281, "step": 13246 }, { "epoch": 14.85, "learning_rate": 1.276056338028169e-05, "loss": 0.5634, "step": 13247 }, { "epoch": 14.85, "learning_rate": 1.2759154929577464e-05, "loss": 0.8496, "step": 13248 }, { "epoch": 14.85, "learning_rate": 1.2757746478873239e-05, "loss": 0.5444, "step": 13249 }, { "epoch": 14.85, "learning_rate": 1.2756338028169015e-05, "loss": 0.6574, "step": 13250 }, { "epoch": 14.86, "learning_rate": 1.275492957746479e-05, "loss": 0.549, "step": 13251 }, { "epoch": 14.86, "learning_rate": 1.2753521126760564e-05, "loss": 0.6392, "step": 13252 }, { "epoch": 14.86, "learning_rate": 1.2752112676056339e-05, "loss": 0.6564, "step": 13253 }, { "epoch": 14.86, "learning_rate": 1.2750704225352113e-05, "loss": 0.4899, "step": 13254 }, { "epoch": 14.86, "learning_rate": 1.2749295774647886e-05, "loss": 0.602, "step": 13255 }, { "epoch": 14.86, "learning_rate": 1.2747887323943661e-05, "loss": 0.5397, "step": 13256 }, { "epoch": 14.86, "learning_rate": 1.2746478873239437e-05, "loss": 0.4969, "step": 13257 }, { "epoch": 14.86, "learning_rate": 1.2745070422535212e-05, "loss": 0.5863, "step": 13258 }, { "epoch": 14.86, "learning_rate": 1.2743661971830987e-05, "loss": 0.5109, "step": 13259 }, { "epoch": 14.87, "learning_rate": 1.2742253521126761e-05, "loss": 0.6706, "step": 13260 }, { "epoch": 14.87, "learning_rate": 1.2740845070422536e-05, "loss": 0.5782, "step": 13261 }, { "epoch": 14.87, "learning_rate": 1.273943661971831e-05, "loss": 0.566, "step": 13262 }, { "epoch": 14.87, "learning_rate": 1.2738028169014083e-05, "loss": 0.5248, "step": 13263 }, { "epoch": 14.87, "learning_rate": 1.2736619718309858e-05, "loss": 0.7435, "step": 13264 }, { "epoch": 14.87, "learning_rate": 1.2735211267605634e-05, "loss": 0.538, "step": 13265 }, { "epoch": 14.87, "learning_rate": 1.2733802816901409e-05, "loss": 0.6279, "step": 13266 }, { "epoch": 14.87, "learning_rate": 1.2732394366197184e-05, "loss": 0.5709, "step": 13267 }, { "epoch": 14.87, "learning_rate": 1.2730985915492958e-05, "loss": 0.5951, "step": 13268 }, { "epoch": 14.88, "learning_rate": 1.2729577464788733e-05, "loss": 0.6071, "step": 13269 }, { "epoch": 14.88, "learning_rate": 1.2728169014084508e-05, "loss": 0.685, "step": 13270 }, { "epoch": 14.88, "learning_rate": 1.272676056338028e-05, "loss": 0.581, "step": 13271 }, { "epoch": 14.88, "learning_rate": 1.2725352112676057e-05, "loss": 0.5183, "step": 13272 }, { "epoch": 14.88, "learning_rate": 1.2723943661971832e-05, "loss": 0.6973, "step": 13273 }, { "epoch": 14.88, "learning_rate": 1.2722535211267606e-05, "loss": 0.7072, "step": 13274 }, { "epoch": 14.88, "learning_rate": 1.272112676056338e-05, "loss": 0.682, "step": 13275 }, { "epoch": 14.88, "learning_rate": 1.2719718309859155e-05, "loss": 0.5148, "step": 13276 }, { "epoch": 14.88, "learning_rate": 1.271830985915493e-05, "loss": 0.5906, "step": 13277 }, { "epoch": 14.89, "learning_rate": 1.2716901408450705e-05, "loss": 0.5967, "step": 13278 }, { "epoch": 14.89, "learning_rate": 1.2715492957746478e-05, "loss": 0.5537, "step": 13279 }, { "epoch": 14.89, "learning_rate": 1.2714084507042254e-05, "loss": 0.4773, "step": 13280 }, { "epoch": 14.89, "learning_rate": 1.2712676056338029e-05, "loss": 0.5155, "step": 13281 }, { "epoch": 14.89, "learning_rate": 1.2711267605633803e-05, "loss": 0.4995, "step": 13282 }, { "epoch": 14.89, "learning_rate": 1.2709859154929578e-05, "loss": 0.4409, "step": 13283 }, { "epoch": 14.89, "learning_rate": 1.2708450704225353e-05, "loss": 0.6473, "step": 13284 }, { "epoch": 14.89, "learning_rate": 1.2707042253521127e-05, "loss": 0.3419, "step": 13285 }, { "epoch": 14.89, "learning_rate": 1.2705633802816902e-05, "loss": 0.4749, "step": 13286 }, { "epoch": 14.9, "learning_rate": 1.2704225352112675e-05, "loss": 0.577, "step": 13287 }, { "epoch": 14.9, "learning_rate": 1.2702816901408451e-05, "loss": 0.5577, "step": 13288 }, { "epoch": 14.9, "learning_rate": 1.2701408450704226e-05, "loss": 0.8037, "step": 13289 }, { "epoch": 14.9, "learning_rate": 1.27e-05, "loss": 0.5599, "step": 13290 }, { "epoch": 14.9, "learning_rate": 1.2698591549295775e-05, "loss": 0.6576, "step": 13291 }, { "epoch": 14.9, "learning_rate": 1.269718309859155e-05, "loss": 0.6734, "step": 13292 }, { "epoch": 14.9, "learning_rate": 1.2695774647887324e-05, "loss": 0.691, "step": 13293 }, { "epoch": 14.9, "learning_rate": 1.2694366197183099e-05, "loss": 0.6153, "step": 13294 }, { "epoch": 14.9, "learning_rate": 1.2692957746478874e-05, "loss": 0.6369, "step": 13295 }, { "epoch": 14.91, "learning_rate": 1.2691549295774648e-05, "loss": 0.5725, "step": 13296 }, { "epoch": 14.91, "learning_rate": 1.2690140845070423e-05, "loss": 0.6257, "step": 13297 }, { "epoch": 14.91, "learning_rate": 1.2688732394366197e-05, "loss": 0.6523, "step": 13298 }, { "epoch": 14.91, "learning_rate": 1.2687323943661972e-05, "loss": 0.7466, "step": 13299 }, { "epoch": 14.91, "learning_rate": 1.2685915492957747e-05, "loss": 0.6018, "step": 13300 }, { "epoch": 14.91, "learning_rate": 1.2684507042253521e-05, "loss": 0.7763, "step": 13301 }, { "epoch": 14.91, "learning_rate": 1.2683098591549296e-05, "loss": 0.4965, "step": 13302 }, { "epoch": 14.91, "learning_rate": 1.268169014084507e-05, "loss": 0.445, "step": 13303 }, { "epoch": 14.91, "learning_rate": 1.2680281690140845e-05, "loss": 0.5638, "step": 13304 }, { "epoch": 14.92, "learning_rate": 1.267887323943662e-05, "loss": 0.6006, "step": 13305 }, { "epoch": 14.92, "learning_rate": 1.2677464788732395e-05, "loss": 0.7292, "step": 13306 }, { "epoch": 14.92, "learning_rate": 1.267605633802817e-05, "loss": 0.4673, "step": 13307 }, { "epoch": 14.92, "learning_rate": 1.2674647887323944e-05, "loss": 0.5264, "step": 13308 }, { "epoch": 14.92, "learning_rate": 1.2673239436619718e-05, "loss": 0.6051, "step": 13309 }, { "epoch": 14.92, "learning_rate": 1.2671830985915495e-05, "loss": 0.7464, "step": 13310 }, { "epoch": 14.92, "learning_rate": 1.2670422535211268e-05, "loss": 0.5319, "step": 13311 }, { "epoch": 14.92, "learning_rate": 1.2669014084507042e-05, "loss": 0.4915, "step": 13312 }, { "epoch": 14.92, "learning_rate": 1.2667605633802817e-05, "loss": 0.6473, "step": 13313 }, { "epoch": 14.93, "learning_rate": 1.2666197183098592e-05, "loss": 0.6016, "step": 13314 }, { "epoch": 14.93, "learning_rate": 1.2664788732394366e-05, "loss": 0.5077, "step": 13315 }, { "epoch": 14.93, "learning_rate": 1.2663380281690141e-05, "loss": 0.5424, "step": 13316 }, { "epoch": 14.93, "learning_rate": 1.2661971830985916e-05, "loss": 0.5875, "step": 13317 }, { "epoch": 14.93, "learning_rate": 1.2660563380281692e-05, "loss": 0.6774, "step": 13318 }, { "epoch": 14.93, "learning_rate": 1.2659154929577465e-05, "loss": 0.6671, "step": 13319 }, { "epoch": 14.93, "learning_rate": 1.265774647887324e-05, "loss": 0.5386, "step": 13320 }, { "epoch": 14.93, "learning_rate": 1.2656338028169014e-05, "loss": 0.7488, "step": 13321 }, { "epoch": 14.93, "learning_rate": 1.2654929577464789e-05, "loss": 0.5884, "step": 13322 }, { "epoch": 14.94, "learning_rate": 1.2653521126760563e-05, "loss": 0.5914, "step": 13323 }, { "epoch": 14.94, "learning_rate": 1.2652112676056338e-05, "loss": 0.5406, "step": 13324 }, { "epoch": 14.94, "learning_rate": 1.2650704225352114e-05, "loss": 0.5374, "step": 13325 }, { "epoch": 14.94, "learning_rate": 1.2649295774647889e-05, "loss": 0.5675, "step": 13326 }, { "epoch": 14.94, "learning_rate": 1.2647887323943662e-05, "loss": 0.5691, "step": 13327 }, { "epoch": 14.94, "learning_rate": 1.2646478873239437e-05, "loss": 0.4995, "step": 13328 }, { "epoch": 14.94, "learning_rate": 1.2645070422535211e-05, "loss": 0.481, "step": 13329 }, { "epoch": 14.94, "learning_rate": 1.2643661971830986e-05, "loss": 0.7424, "step": 13330 }, { "epoch": 14.95, "learning_rate": 1.264225352112676e-05, "loss": 0.5149, "step": 13331 }, { "epoch": 14.95, "learning_rate": 1.2640845070422535e-05, "loss": 0.4335, "step": 13332 }, { "epoch": 14.95, "learning_rate": 1.2639436619718311e-05, "loss": 0.6255, "step": 13333 }, { "epoch": 14.95, "learning_rate": 1.2638028169014086e-05, "loss": 0.4763, "step": 13334 }, { "epoch": 14.95, "learning_rate": 1.2636619718309859e-05, "loss": 0.5961, "step": 13335 }, { "epoch": 14.95, "learning_rate": 1.2635211267605634e-05, "loss": 0.6166, "step": 13336 }, { "epoch": 14.95, "learning_rate": 1.2633802816901408e-05, "loss": 0.5261, "step": 13337 }, { "epoch": 14.95, "learning_rate": 1.2632394366197183e-05, "loss": 0.5341, "step": 13338 }, { "epoch": 14.95, "learning_rate": 1.2630985915492958e-05, "loss": 0.7028, "step": 13339 }, { "epoch": 14.96, "learning_rate": 1.2629577464788734e-05, "loss": 0.5874, "step": 13340 }, { "epoch": 14.96, "learning_rate": 1.2628169014084509e-05, "loss": 0.8526, "step": 13341 }, { "epoch": 14.96, "learning_rate": 1.2626760563380283e-05, "loss": 0.5755, "step": 13342 }, { "epoch": 14.96, "learning_rate": 1.2625352112676056e-05, "loss": 0.6783, "step": 13343 }, { "epoch": 14.96, "learning_rate": 1.262394366197183e-05, "loss": 0.705, "step": 13344 }, { "epoch": 14.96, "learning_rate": 1.2622535211267605e-05, "loss": 0.6485, "step": 13345 }, { "epoch": 14.96, "learning_rate": 1.262112676056338e-05, "loss": 0.4655, "step": 13346 }, { "epoch": 14.96, "learning_rate": 1.2619718309859155e-05, "loss": 0.5869, "step": 13347 }, { "epoch": 14.96, "learning_rate": 1.2618309859154931e-05, "loss": 0.5906, "step": 13348 }, { "epoch": 14.97, "learning_rate": 1.2616901408450706e-05, "loss": 0.7263, "step": 13349 }, { "epoch": 14.97, "learning_rate": 1.261549295774648e-05, "loss": 0.6151, "step": 13350 }, { "epoch": 14.97, "learning_rate": 1.2614084507042253e-05, "loss": 0.5574, "step": 13351 }, { "epoch": 14.97, "learning_rate": 1.2612676056338028e-05, "loss": 0.673, "step": 13352 }, { "epoch": 14.97, "learning_rate": 1.2611267605633803e-05, "loss": 0.7198, "step": 13353 }, { "epoch": 14.97, "learning_rate": 1.2609859154929577e-05, "loss": 0.6487, "step": 13354 }, { "epoch": 14.97, "learning_rate": 1.2608450704225352e-05, "loss": 0.5544, "step": 13355 }, { "epoch": 14.97, "learning_rate": 1.2607042253521128e-05, "loss": 0.5758, "step": 13356 }, { "epoch": 14.97, "learning_rate": 1.2605633802816903e-05, "loss": 0.5643, "step": 13357 }, { "epoch": 14.98, "learning_rate": 1.2604225352112676e-05, "loss": 0.6636, "step": 13358 }, { "epoch": 14.98, "learning_rate": 1.260281690140845e-05, "loss": 0.5256, "step": 13359 }, { "epoch": 14.98, "learning_rate": 1.2601408450704225e-05, "loss": 0.5797, "step": 13360 }, { "epoch": 14.98, "learning_rate": 1.26e-05, "loss": 0.615, "step": 13361 }, { "epoch": 14.98, "learning_rate": 1.2598591549295774e-05, "loss": 0.6805, "step": 13362 }, { "epoch": 14.98, "learning_rate": 1.259718309859155e-05, "loss": 0.552, "step": 13363 }, { "epoch": 14.98, "learning_rate": 1.2595774647887325e-05, "loss": 0.5741, "step": 13364 }, { "epoch": 14.98, "learning_rate": 1.25943661971831e-05, "loss": 0.6032, "step": 13365 }, { "epoch": 14.98, "learning_rate": 1.2592957746478873e-05, "loss": 0.5841, "step": 13366 }, { "epoch": 14.99, "learning_rate": 1.2591549295774647e-05, "loss": 0.449, "step": 13367 }, { "epoch": 14.99, "learning_rate": 1.2590140845070422e-05, "loss": 0.5807, "step": 13368 }, { "epoch": 14.99, "learning_rate": 1.2588732394366197e-05, "loss": 0.5452, "step": 13369 }, { "epoch": 14.99, "learning_rate": 1.2587323943661971e-05, "loss": 0.5929, "step": 13370 }, { "epoch": 14.99, "learning_rate": 1.2585915492957748e-05, "loss": 0.546, "step": 13371 }, { "epoch": 14.99, "learning_rate": 1.2584507042253522e-05, "loss": 0.7747, "step": 13372 }, { "epoch": 14.99, "learning_rate": 1.2583098591549297e-05, "loss": 0.4997, "step": 13373 }, { "epoch": 14.99, "learning_rate": 1.258169014084507e-05, "loss": 0.4951, "step": 13374 }, { "epoch": 14.99, "learning_rate": 1.2580281690140845e-05, "loss": 0.4781, "step": 13375 }, { "epoch": 15.0, "learning_rate": 1.257887323943662e-05, "loss": 0.6411, "step": 13376 }, { "epoch": 15.0, "learning_rate": 1.2577464788732394e-05, "loss": 0.6664, "step": 13377 }, { "epoch": 15.0, "learning_rate": 1.257605633802817e-05, "loss": 0.5062, "step": 13378 }, { "epoch": 15.0, "learning_rate": 1.2574647887323945e-05, "loss": 0.6189, "step": 13379 }, { "epoch": 15.0, "learning_rate": 1.257323943661972e-05, "loss": 0.4244, "step": 13380 }, { "epoch": 15.0, "learning_rate": 1.2571830985915494e-05, "loss": 0.9306, "step": 13381 }, { "epoch": 15.0, "learning_rate": 1.2570422535211267e-05, "loss": 0.5506, "step": 13382 }, { "epoch": 15.0, "learning_rate": 1.2569014084507042e-05, "loss": 0.8145, "step": 13383 }, { "epoch": 15.0, "learning_rate": 1.2567605633802816e-05, "loss": 0.4821, "step": 13384 }, { "epoch": 15.01, "learning_rate": 1.2566197183098591e-05, "loss": 0.5805, "step": 13385 }, { "epoch": 15.01, "learning_rate": 1.2564788732394367e-05, "loss": 0.5571, "step": 13386 }, { "epoch": 15.01, "learning_rate": 1.2563380281690142e-05, "loss": 0.5677, "step": 13387 }, { "epoch": 15.01, "learning_rate": 1.2561971830985917e-05, "loss": 0.6269, "step": 13388 }, { "epoch": 15.01, "learning_rate": 1.2560563380281691e-05, "loss": 0.5566, "step": 13389 }, { "epoch": 15.01, "learning_rate": 1.2559154929577464e-05, "loss": 0.6395, "step": 13390 }, { "epoch": 15.01, "learning_rate": 1.2557746478873239e-05, "loss": 0.6138, "step": 13391 }, { "epoch": 15.01, "learning_rate": 1.2556338028169013e-05, "loss": 0.6132, "step": 13392 }, { "epoch": 15.01, "learning_rate": 1.255492957746479e-05, "loss": 0.6764, "step": 13393 }, { "epoch": 15.02, "learning_rate": 1.2553521126760564e-05, "loss": 0.5685, "step": 13394 }, { "epoch": 15.02, "learning_rate": 1.2552112676056339e-05, "loss": 0.575, "step": 13395 }, { "epoch": 15.02, "learning_rate": 1.2550704225352114e-05, "loss": 0.512, "step": 13396 }, { "epoch": 15.02, "learning_rate": 1.2549295774647888e-05, "loss": 0.6069, "step": 13397 }, { "epoch": 15.02, "learning_rate": 1.2547887323943661e-05, "loss": 0.5484, "step": 13398 }, { "epoch": 15.02, "learning_rate": 1.2546478873239436e-05, "loss": 0.5391, "step": 13399 }, { "epoch": 15.02, "learning_rate": 1.254507042253521e-05, "loss": 0.5528, "step": 13400 }, { "epoch": 15.02, "learning_rate": 1.2543661971830987e-05, "loss": 0.5395, "step": 13401 }, { "epoch": 15.02, "learning_rate": 1.2542253521126761e-05, "loss": 0.5721, "step": 13402 }, { "epoch": 15.03, "learning_rate": 1.2540845070422536e-05, "loss": 0.4887, "step": 13403 }, { "epoch": 15.03, "learning_rate": 1.253943661971831e-05, "loss": 0.4241, "step": 13404 }, { "epoch": 15.03, "learning_rate": 1.2538028169014085e-05, "loss": 0.5551, "step": 13405 }, { "epoch": 15.03, "learning_rate": 1.2536619718309858e-05, "loss": 0.5518, "step": 13406 }, { "epoch": 15.03, "learning_rate": 1.2535211267605633e-05, "loss": 0.5514, "step": 13407 }, { "epoch": 15.03, "learning_rate": 1.2533802816901408e-05, "loss": 0.5011, "step": 13408 }, { "epoch": 15.03, "learning_rate": 1.2532394366197184e-05, "loss": 0.6309, "step": 13409 }, { "epoch": 15.03, "learning_rate": 1.2530985915492959e-05, "loss": 0.5551, "step": 13410 }, { "epoch": 15.03, "learning_rate": 1.2529577464788733e-05, "loss": 0.6479, "step": 13411 }, { "epoch": 15.04, "learning_rate": 1.2528169014084508e-05, "loss": 0.57, "step": 13412 }, { "epoch": 15.04, "learning_rate": 1.2526760563380282e-05, "loss": 0.4932, "step": 13413 }, { "epoch": 15.04, "learning_rate": 1.2525352112676055e-05, "loss": 0.5286, "step": 13414 }, { "epoch": 15.04, "learning_rate": 1.252394366197183e-05, "loss": 0.4383, "step": 13415 }, { "epoch": 15.04, "learning_rate": 1.2522535211267606e-05, "loss": 0.4509, "step": 13416 }, { "epoch": 15.04, "learning_rate": 1.2521126760563381e-05, "loss": 0.4373, "step": 13417 }, { "epoch": 15.04, "learning_rate": 1.2519718309859156e-05, "loss": 0.5119, "step": 13418 }, { "epoch": 15.04, "learning_rate": 1.251830985915493e-05, "loss": 0.3838, "step": 13419 }, { "epoch": 15.04, "learning_rate": 1.2516901408450705e-05, "loss": 0.4482, "step": 13420 }, { "epoch": 15.05, "learning_rate": 1.251549295774648e-05, "loss": 0.5469, "step": 13421 }, { "epoch": 15.05, "learning_rate": 1.2514084507042253e-05, "loss": 0.6971, "step": 13422 }, { "epoch": 15.05, "learning_rate": 1.2512676056338027e-05, "loss": 0.5513, "step": 13423 }, { "epoch": 15.05, "learning_rate": 1.2511267605633803e-05, "loss": 0.6097, "step": 13424 }, { "epoch": 15.05, "learning_rate": 1.2509859154929578e-05, "loss": 0.5881, "step": 13425 }, { "epoch": 15.05, "learning_rate": 1.2508450704225353e-05, "loss": 0.3488, "step": 13426 }, { "epoch": 15.05, "learning_rate": 1.2507042253521127e-05, "loss": 0.3992, "step": 13427 }, { "epoch": 15.05, "learning_rate": 1.2505633802816902e-05, "loss": 0.4522, "step": 13428 }, { "epoch": 15.05, "learning_rate": 1.2504225352112677e-05, "loss": 0.5217, "step": 13429 }, { "epoch": 15.06, "learning_rate": 1.250281690140845e-05, "loss": 0.5553, "step": 13430 }, { "epoch": 15.06, "learning_rate": 1.2501408450704226e-05, "loss": 0.8149, "step": 13431 }, { "epoch": 15.06, "learning_rate": 1.25e-05, "loss": 0.5369, "step": 13432 }, { "epoch": 15.06, "learning_rate": 1.2498591549295775e-05, "loss": 0.5537, "step": 13433 }, { "epoch": 15.06, "learning_rate": 1.249718309859155e-05, "loss": 0.5527, "step": 13434 }, { "epoch": 15.06, "learning_rate": 1.2495774647887325e-05, "loss": 0.6549, "step": 13435 }, { "epoch": 15.06, "learning_rate": 1.2494366197183099e-05, "loss": 0.5896, "step": 13436 }, { "epoch": 15.06, "learning_rate": 1.2492957746478874e-05, "loss": 0.6292, "step": 13437 }, { "epoch": 15.07, "learning_rate": 1.2491549295774647e-05, "loss": 0.6646, "step": 13438 }, { "epoch": 15.07, "learning_rate": 1.2490140845070423e-05, "loss": 0.6114, "step": 13439 }, { "epoch": 15.07, "learning_rate": 1.2488732394366198e-05, "loss": 0.5543, "step": 13440 }, { "epoch": 15.07, "learning_rate": 1.2487323943661972e-05, "loss": 0.5605, "step": 13441 }, { "epoch": 15.07, "learning_rate": 1.2485915492957747e-05, "loss": 0.6183, "step": 13442 }, { "epoch": 15.07, "learning_rate": 1.2484507042253522e-05, "loss": 0.4724, "step": 13443 }, { "epoch": 15.07, "learning_rate": 1.2483098591549296e-05, "loss": 0.5144, "step": 13444 }, { "epoch": 15.07, "learning_rate": 1.2481690140845071e-05, "loss": 0.6334, "step": 13445 }, { "epoch": 15.07, "learning_rate": 1.2480281690140846e-05, "loss": 0.4881, "step": 13446 }, { "epoch": 15.08, "learning_rate": 1.247887323943662e-05, "loss": 0.5504, "step": 13447 }, { "epoch": 15.08, "learning_rate": 1.2477464788732395e-05, "loss": 0.4756, "step": 13448 }, { "epoch": 15.08, "learning_rate": 1.247605633802817e-05, "loss": 0.4869, "step": 13449 }, { "epoch": 15.08, "learning_rate": 1.2474647887323944e-05, "loss": 0.6615, "step": 13450 }, { "epoch": 15.08, "learning_rate": 1.2473239436619719e-05, "loss": 0.4427, "step": 13451 }, { "epoch": 15.08, "learning_rate": 1.2471830985915493e-05, "loss": 0.4964, "step": 13452 }, { "epoch": 15.08, "learning_rate": 1.2470422535211268e-05, "loss": 0.4993, "step": 13453 }, { "epoch": 15.08, "learning_rate": 1.2469014084507043e-05, "loss": 0.5618, "step": 13454 }, { "epoch": 15.08, "learning_rate": 1.2467605633802817e-05, "loss": 0.4463, "step": 13455 }, { "epoch": 15.09, "learning_rate": 1.2466197183098592e-05, "loss": 0.5831, "step": 13456 }, { "epoch": 15.09, "learning_rate": 1.2464788732394367e-05, "loss": 0.5902, "step": 13457 }, { "epoch": 15.09, "learning_rate": 1.2463380281690141e-05, "loss": 0.5516, "step": 13458 }, { "epoch": 15.09, "learning_rate": 1.2461971830985916e-05, "loss": 0.4377, "step": 13459 }, { "epoch": 15.09, "learning_rate": 1.246056338028169e-05, "loss": 0.5464, "step": 13460 }, { "epoch": 15.09, "learning_rate": 1.2459154929577467e-05, "loss": 0.4141, "step": 13461 }, { "epoch": 15.09, "learning_rate": 1.245774647887324e-05, "loss": 0.5071, "step": 13462 }, { "epoch": 15.09, "learning_rate": 1.2456338028169014e-05, "loss": 0.4718, "step": 13463 }, { "epoch": 15.09, "learning_rate": 1.2454929577464789e-05, "loss": 0.6518, "step": 13464 }, { "epoch": 15.1, "learning_rate": 1.2453521126760564e-05, "loss": 0.5651, "step": 13465 }, { "epoch": 15.1, "learning_rate": 1.2452112676056338e-05, "loss": 0.4745, "step": 13466 }, { "epoch": 15.1, "learning_rate": 1.2450704225352113e-05, "loss": 0.4531, "step": 13467 }, { "epoch": 15.1, "learning_rate": 1.2449295774647888e-05, "loss": 0.4852, "step": 13468 }, { "epoch": 15.1, "learning_rate": 1.2447887323943662e-05, "loss": 0.477, "step": 13469 }, { "epoch": 15.1, "learning_rate": 1.2446478873239437e-05, "loss": 0.4726, "step": 13470 }, { "epoch": 15.1, "learning_rate": 1.2445070422535211e-05, "loss": 0.5612, "step": 13471 }, { "epoch": 15.1, "learning_rate": 1.2443661971830986e-05, "loss": 0.576, "step": 13472 }, { "epoch": 15.1, "learning_rate": 1.244225352112676e-05, "loss": 0.4502, "step": 13473 }, { "epoch": 15.11, "learning_rate": 1.2440845070422535e-05, "loss": 0.5873, "step": 13474 }, { "epoch": 15.11, "learning_rate": 1.243943661971831e-05, "loss": 0.504, "step": 13475 }, { "epoch": 15.11, "learning_rate": 1.2438028169014085e-05, "loss": 0.313, "step": 13476 }, { "epoch": 15.11, "learning_rate": 1.243661971830986e-05, "loss": 0.3902, "step": 13477 }, { "epoch": 15.11, "learning_rate": 1.2435211267605634e-05, "loss": 0.7228, "step": 13478 }, { "epoch": 15.11, "learning_rate": 1.2433802816901409e-05, "loss": 0.3964, "step": 13479 }, { "epoch": 15.11, "learning_rate": 1.2432394366197183e-05, "loss": 0.7015, "step": 13480 }, { "epoch": 15.11, "learning_rate": 1.2430985915492958e-05, "loss": 0.6872, "step": 13481 }, { "epoch": 15.11, "learning_rate": 1.2429577464788732e-05, "loss": 0.6667, "step": 13482 }, { "epoch": 15.12, "learning_rate": 1.2428169014084507e-05, "loss": 0.5425, "step": 13483 }, { "epoch": 15.12, "learning_rate": 1.2426760563380283e-05, "loss": 0.7589, "step": 13484 }, { "epoch": 15.12, "learning_rate": 1.2425352112676056e-05, "loss": 0.4802, "step": 13485 }, { "epoch": 15.12, "learning_rate": 1.2423943661971831e-05, "loss": 0.4785, "step": 13486 }, { "epoch": 15.12, "learning_rate": 1.2422535211267606e-05, "loss": 0.6433, "step": 13487 }, { "epoch": 15.12, "learning_rate": 1.242112676056338e-05, "loss": 0.6447, "step": 13488 }, { "epoch": 15.12, "learning_rate": 1.2419718309859155e-05, "loss": 0.5542, "step": 13489 }, { "epoch": 15.12, "learning_rate": 1.241830985915493e-05, "loss": 0.5595, "step": 13490 }, { "epoch": 15.12, "learning_rate": 1.2416901408450704e-05, "loss": 0.6149, "step": 13491 }, { "epoch": 15.13, "learning_rate": 1.241549295774648e-05, "loss": 0.4418, "step": 13492 }, { "epoch": 15.13, "learning_rate": 1.2414084507042253e-05, "loss": 0.5606, "step": 13493 }, { "epoch": 15.13, "learning_rate": 1.2412676056338028e-05, "loss": 0.5384, "step": 13494 }, { "epoch": 15.13, "learning_rate": 1.2411267605633803e-05, "loss": 0.3739, "step": 13495 }, { "epoch": 15.13, "learning_rate": 1.2409859154929577e-05, "loss": 0.5323, "step": 13496 }, { "epoch": 15.13, "learning_rate": 1.2408450704225352e-05, "loss": 0.6771, "step": 13497 }, { "epoch": 15.13, "learning_rate": 1.2407042253521127e-05, "loss": 0.6175, "step": 13498 }, { "epoch": 15.13, "learning_rate": 1.2405633802816903e-05, "loss": 0.6257, "step": 13499 }, { "epoch": 15.13, "learning_rate": 1.2404225352112678e-05, "loss": 0.5078, "step": 13500 }, { "epoch": 15.13, "eval_loss": 0.882098376750946, "eval_runtime": 826.7086, "eval_samples_per_second": 3.196, "eval_steps_per_second": 0.4, "eval_wer": 0.1547996826656089, "step": 13500 }, { "epoch": 15.14, "learning_rate": 1.240281690140845e-05, "loss": 0.5198, "step": 13501 }, { "epoch": 15.14, "learning_rate": 1.2401408450704225e-05, "loss": 0.4729, "step": 13502 }, { "epoch": 15.14, "learning_rate": 1.24e-05, "loss": 0.5537, "step": 13503 }, { "epoch": 15.14, "learning_rate": 1.2398591549295774e-05, "loss": 0.5586, "step": 13504 }, { "epoch": 15.14, "learning_rate": 1.2397183098591549e-05, "loss": 0.5169, "step": 13505 }, { "epoch": 15.14, "learning_rate": 1.2395774647887324e-05, "loss": 0.4907, "step": 13506 }, { "epoch": 15.14, "learning_rate": 1.23943661971831e-05, "loss": 0.5807, "step": 13507 }, { "epoch": 15.14, "learning_rate": 1.2392957746478875e-05, "loss": 0.5087, "step": 13508 }, { "epoch": 15.14, "learning_rate": 1.2391549295774648e-05, "loss": 0.4439, "step": 13509 }, { "epoch": 15.15, "learning_rate": 1.2390140845070422e-05, "loss": 0.4863, "step": 13510 }, { "epoch": 15.15, "learning_rate": 1.2388732394366197e-05, "loss": 0.6693, "step": 13511 }, { "epoch": 15.15, "learning_rate": 1.2387323943661972e-05, "loss": 0.4574, "step": 13512 }, { "epoch": 15.15, "learning_rate": 1.2385915492957746e-05, "loss": 0.4882, "step": 13513 }, { "epoch": 15.15, "learning_rate": 1.2384507042253523e-05, "loss": 0.3887, "step": 13514 }, { "epoch": 15.15, "learning_rate": 1.2383098591549297e-05, "loss": 0.4751, "step": 13515 }, { "epoch": 15.15, "learning_rate": 1.2381690140845072e-05, "loss": 0.5016, "step": 13516 }, { "epoch": 15.15, "learning_rate": 1.2380281690140845e-05, "loss": 0.5359, "step": 13517 }, { "epoch": 15.15, "learning_rate": 1.237887323943662e-05, "loss": 0.5257, "step": 13518 }, { "epoch": 15.16, "learning_rate": 1.2377464788732394e-05, "loss": 0.4763, "step": 13519 }, { "epoch": 15.16, "learning_rate": 1.2376056338028169e-05, "loss": 0.522, "step": 13520 }, { "epoch": 15.16, "learning_rate": 1.2374647887323943e-05, "loss": 0.5478, "step": 13521 }, { "epoch": 15.16, "learning_rate": 1.237323943661972e-05, "loss": 0.4706, "step": 13522 }, { "epoch": 15.16, "learning_rate": 1.2371830985915494e-05, "loss": 0.7057, "step": 13523 }, { "epoch": 15.16, "learning_rate": 1.2370422535211269e-05, "loss": 0.5221, "step": 13524 }, { "epoch": 15.16, "learning_rate": 1.2369014084507042e-05, "loss": 0.4758, "step": 13525 }, { "epoch": 15.16, "learning_rate": 1.2367605633802817e-05, "loss": 0.5044, "step": 13526 }, { "epoch": 15.16, "learning_rate": 1.2366197183098591e-05, "loss": 0.3779, "step": 13527 }, { "epoch": 15.17, "learning_rate": 1.2364788732394366e-05, "loss": 0.4528, "step": 13528 }, { "epoch": 15.17, "learning_rate": 1.236338028169014e-05, "loss": 0.4974, "step": 13529 }, { "epoch": 15.17, "learning_rate": 1.2361971830985917e-05, "loss": 0.5829, "step": 13530 }, { "epoch": 15.17, "learning_rate": 1.2360563380281691e-05, "loss": 0.644, "step": 13531 }, { "epoch": 15.17, "learning_rate": 1.2359154929577466e-05, "loss": 0.6523, "step": 13532 }, { "epoch": 15.17, "learning_rate": 1.2357746478873239e-05, "loss": 0.6315, "step": 13533 }, { "epoch": 15.17, "learning_rate": 1.2356338028169014e-05, "loss": 0.6657, "step": 13534 }, { "epoch": 15.17, "learning_rate": 1.2354929577464788e-05, "loss": 0.5035, "step": 13535 }, { "epoch": 15.17, "learning_rate": 1.2353521126760563e-05, "loss": 0.6343, "step": 13536 }, { "epoch": 15.18, "learning_rate": 1.235211267605634e-05, "loss": 0.4987, "step": 13537 }, { "epoch": 15.18, "learning_rate": 1.2350704225352114e-05, "loss": 0.5875, "step": 13538 }, { "epoch": 15.18, "learning_rate": 1.2349295774647889e-05, "loss": 0.6374, "step": 13539 }, { "epoch": 15.18, "learning_rate": 1.2347887323943663e-05, "loss": 0.6219, "step": 13540 }, { "epoch": 15.18, "learning_rate": 1.2346478873239436e-05, "loss": 0.5907, "step": 13541 }, { "epoch": 15.18, "learning_rate": 1.234507042253521e-05, "loss": 0.5387, "step": 13542 }, { "epoch": 15.18, "learning_rate": 1.2343661971830985e-05, "loss": 0.6045, "step": 13543 }, { "epoch": 15.18, "learning_rate": 1.234225352112676e-05, "loss": 0.5284, "step": 13544 }, { "epoch": 15.18, "learning_rate": 1.2340845070422536e-05, "loss": 0.6354, "step": 13545 }, { "epoch": 15.19, "learning_rate": 1.2339436619718311e-05, "loss": 0.6003, "step": 13546 }, { "epoch": 15.19, "learning_rate": 1.2338028169014086e-05, "loss": 0.611, "step": 13547 }, { "epoch": 15.19, "learning_rate": 1.233661971830986e-05, "loss": 0.6218, "step": 13548 }, { "epoch": 15.19, "learning_rate": 1.2335211267605633e-05, "loss": 0.536, "step": 13549 }, { "epoch": 15.19, "learning_rate": 1.2333802816901408e-05, "loss": 0.5604, "step": 13550 }, { "epoch": 15.19, "learning_rate": 1.2332394366197182e-05, "loss": 0.6815, "step": 13551 }, { "epoch": 15.19, "learning_rate": 1.2330985915492959e-05, "loss": 0.449, "step": 13552 }, { "epoch": 15.19, "learning_rate": 1.2329577464788733e-05, "loss": 0.5324, "step": 13553 }, { "epoch": 15.2, "learning_rate": 1.2328169014084508e-05, "loss": 0.5695, "step": 13554 }, { "epoch": 15.2, "learning_rate": 1.2326760563380283e-05, "loss": 0.6288, "step": 13555 }, { "epoch": 15.2, "learning_rate": 1.2325352112676057e-05, "loss": 0.578, "step": 13556 }, { "epoch": 15.2, "learning_rate": 1.232394366197183e-05, "loss": 0.5013, "step": 13557 }, { "epoch": 15.2, "learning_rate": 1.2322535211267605e-05, "loss": 0.5241, "step": 13558 }, { "epoch": 15.2, "learning_rate": 1.232112676056338e-05, "loss": 0.5778, "step": 13559 }, { "epoch": 15.2, "learning_rate": 1.2319718309859156e-05, "loss": 0.59, "step": 13560 }, { "epoch": 15.2, "learning_rate": 1.231830985915493e-05, "loss": 0.5207, "step": 13561 }, { "epoch": 15.2, "learning_rate": 1.2316901408450705e-05, "loss": 0.5532, "step": 13562 }, { "epoch": 15.21, "learning_rate": 1.231549295774648e-05, "loss": 0.3914, "step": 13563 }, { "epoch": 15.21, "learning_rate": 1.2314084507042254e-05, "loss": 0.4689, "step": 13564 }, { "epoch": 15.21, "learning_rate": 1.2312676056338027e-05, "loss": 0.6133, "step": 13565 }, { "epoch": 15.21, "learning_rate": 1.2311267605633802e-05, "loss": 0.5257, "step": 13566 }, { "epoch": 15.21, "learning_rate": 1.2309859154929578e-05, "loss": 0.488, "step": 13567 }, { "epoch": 15.21, "learning_rate": 1.2308450704225353e-05, "loss": 0.507, "step": 13568 }, { "epoch": 15.21, "learning_rate": 1.2307042253521128e-05, "loss": 0.4034, "step": 13569 }, { "epoch": 15.21, "learning_rate": 1.2305633802816902e-05, "loss": 0.5743, "step": 13570 }, { "epoch": 15.21, "learning_rate": 1.2304225352112677e-05, "loss": 0.5032, "step": 13571 }, { "epoch": 15.22, "learning_rate": 1.2302816901408452e-05, "loss": 0.5741, "step": 13572 }, { "epoch": 15.22, "learning_rate": 1.2301408450704224e-05, "loss": 0.6354, "step": 13573 }, { "epoch": 15.22, "learning_rate": 1.2299999999999999e-05, "loss": 0.4481, "step": 13574 }, { "epoch": 15.22, "learning_rate": 1.2298591549295775e-05, "loss": 0.361, "step": 13575 }, { "epoch": 15.22, "learning_rate": 1.229718309859155e-05, "loss": 0.4724, "step": 13576 }, { "epoch": 15.22, "learning_rate": 1.2295774647887325e-05, "loss": 0.5814, "step": 13577 }, { "epoch": 15.22, "learning_rate": 1.22943661971831e-05, "loss": 0.4141, "step": 13578 }, { "epoch": 15.22, "learning_rate": 1.2292957746478874e-05, "loss": 0.463, "step": 13579 }, { "epoch": 15.22, "learning_rate": 1.2291549295774649e-05, "loss": 0.5343, "step": 13580 }, { "epoch": 15.23, "learning_rate": 1.2290140845070422e-05, "loss": 0.8173, "step": 13581 }, { "epoch": 15.23, "learning_rate": 1.2288732394366196e-05, "loss": 0.7193, "step": 13582 }, { "epoch": 15.23, "learning_rate": 1.2287323943661973e-05, "loss": 0.5431, "step": 13583 }, { "epoch": 15.23, "learning_rate": 1.2285915492957747e-05, "loss": 0.7995, "step": 13584 }, { "epoch": 15.23, "learning_rate": 1.2284507042253522e-05, "loss": 0.5067, "step": 13585 }, { "epoch": 15.23, "learning_rate": 1.2283098591549296e-05, "loss": 0.6864, "step": 13586 }, { "epoch": 15.23, "learning_rate": 1.2281690140845071e-05, "loss": 0.5113, "step": 13587 }, { "epoch": 15.23, "learning_rate": 1.2280281690140844e-05, "loss": 0.5427, "step": 13588 }, { "epoch": 15.23, "learning_rate": 1.2278873239436619e-05, "loss": 0.5325, "step": 13589 }, { "epoch": 15.24, "learning_rate": 1.2277464788732395e-05, "loss": 0.5909, "step": 13590 }, { "epoch": 15.24, "learning_rate": 1.227605633802817e-05, "loss": 0.5725, "step": 13591 }, { "epoch": 15.24, "learning_rate": 1.2274647887323944e-05, "loss": 0.5702, "step": 13592 }, { "epoch": 15.24, "learning_rate": 1.2273239436619719e-05, "loss": 0.5894, "step": 13593 }, { "epoch": 15.24, "learning_rate": 1.2271830985915494e-05, "loss": 0.3894, "step": 13594 }, { "epoch": 15.24, "learning_rate": 1.2270422535211268e-05, "loss": 0.6354, "step": 13595 }, { "epoch": 15.24, "learning_rate": 1.2269014084507041e-05, "loss": 0.5067, "step": 13596 }, { "epoch": 15.24, "learning_rate": 1.2267605633802816e-05, "loss": 0.5399, "step": 13597 }, { "epoch": 15.24, "learning_rate": 1.2266197183098592e-05, "loss": 0.5305, "step": 13598 }, { "epoch": 15.25, "learning_rate": 1.2264788732394367e-05, "loss": 0.561, "step": 13599 }, { "epoch": 15.25, "learning_rate": 1.2263380281690141e-05, "loss": 0.5566, "step": 13600 }, { "epoch": 15.25, "learning_rate": 1.2261971830985916e-05, "loss": 0.427, "step": 13601 }, { "epoch": 15.25, "learning_rate": 1.226056338028169e-05, "loss": 0.6081, "step": 13602 }, { "epoch": 15.25, "learning_rate": 1.2259154929577465e-05, "loss": 0.5627, "step": 13603 }, { "epoch": 15.25, "learning_rate": 1.2257746478873238e-05, "loss": 0.5766, "step": 13604 }, { "epoch": 15.25, "learning_rate": 1.2256338028169015e-05, "loss": 0.545, "step": 13605 }, { "epoch": 15.25, "learning_rate": 1.225492957746479e-05, "loss": 0.4949, "step": 13606 }, { "epoch": 15.25, "learning_rate": 1.2253521126760564e-05, "loss": 0.5207, "step": 13607 }, { "epoch": 15.26, "learning_rate": 1.2252112676056338e-05, "loss": 0.6342, "step": 13608 }, { "epoch": 15.26, "learning_rate": 1.2250704225352113e-05, "loss": 0.4752, "step": 13609 }, { "epoch": 15.26, "learning_rate": 1.2249295774647888e-05, "loss": 0.5653, "step": 13610 }, { "epoch": 15.26, "learning_rate": 1.2247887323943662e-05, "loss": 0.4976, "step": 13611 }, { "epoch": 15.26, "learning_rate": 1.2246478873239435e-05, "loss": 0.5041, "step": 13612 }, { "epoch": 15.26, "learning_rate": 1.2245070422535212e-05, "loss": 0.4999, "step": 13613 }, { "epoch": 15.26, "learning_rate": 1.2243661971830986e-05, "loss": 0.4753, "step": 13614 }, { "epoch": 15.26, "learning_rate": 1.2242253521126761e-05, "loss": 0.5762, "step": 13615 }, { "epoch": 15.26, "learning_rate": 1.2240845070422536e-05, "loss": 0.5233, "step": 13616 }, { "epoch": 15.27, "learning_rate": 1.223943661971831e-05, "loss": 0.5135, "step": 13617 }, { "epoch": 15.27, "learning_rate": 1.2238028169014085e-05, "loss": 0.4781, "step": 13618 }, { "epoch": 15.27, "learning_rate": 1.223661971830986e-05, "loss": 0.4789, "step": 13619 }, { "epoch": 15.27, "learning_rate": 1.2235211267605634e-05, "loss": 0.5474, "step": 13620 }, { "epoch": 15.27, "learning_rate": 1.2233802816901409e-05, "loss": 0.5625, "step": 13621 }, { "epoch": 15.27, "learning_rate": 1.2232394366197183e-05, "loss": 0.6865, "step": 13622 }, { "epoch": 15.27, "learning_rate": 1.2230985915492958e-05, "loss": 0.4217, "step": 13623 }, { "epoch": 15.27, "learning_rate": 1.2229577464788733e-05, "loss": 0.5923, "step": 13624 }, { "epoch": 15.27, "learning_rate": 1.2228169014084507e-05, "loss": 0.5496, "step": 13625 }, { "epoch": 15.28, "learning_rate": 1.2226760563380282e-05, "loss": 0.5075, "step": 13626 }, { "epoch": 15.28, "learning_rate": 1.2225352112676057e-05, "loss": 0.4372, "step": 13627 }, { "epoch": 15.28, "learning_rate": 1.2223943661971831e-05, "loss": 0.3096, "step": 13628 }, { "epoch": 15.28, "learning_rate": 1.2222535211267606e-05, "loss": 0.5539, "step": 13629 }, { "epoch": 15.28, "learning_rate": 1.222112676056338e-05, "loss": 0.5955, "step": 13630 }, { "epoch": 15.28, "learning_rate": 1.2219718309859155e-05, "loss": 0.7343, "step": 13631 }, { "epoch": 15.28, "learning_rate": 1.221830985915493e-05, "loss": 0.727, "step": 13632 }, { "epoch": 15.28, "learning_rate": 1.2216901408450704e-05, "loss": 0.6293, "step": 13633 }, { "epoch": 15.28, "learning_rate": 1.2215492957746479e-05, "loss": 0.5846, "step": 13634 }, { "epoch": 15.29, "learning_rate": 1.2214084507042255e-05, "loss": 0.648, "step": 13635 }, { "epoch": 15.29, "learning_rate": 1.2212676056338028e-05, "loss": 0.6117, "step": 13636 }, { "epoch": 15.29, "learning_rate": 1.2211267605633803e-05, "loss": 0.5351, "step": 13637 }, { "epoch": 15.29, "learning_rate": 1.2209859154929578e-05, "loss": 0.599, "step": 13638 }, { "epoch": 15.29, "learning_rate": 1.2208450704225352e-05, "loss": 0.5663, "step": 13639 }, { "epoch": 15.29, "learning_rate": 1.2207042253521127e-05, "loss": 0.6062, "step": 13640 }, { "epoch": 15.29, "learning_rate": 1.2205633802816902e-05, "loss": 0.5754, "step": 13641 }, { "epoch": 15.29, "learning_rate": 1.2204225352112676e-05, "loss": 0.6028, "step": 13642 }, { "epoch": 15.29, "learning_rate": 1.2202816901408453e-05, "loss": 0.574, "step": 13643 }, { "epoch": 15.3, "learning_rate": 1.2201408450704225e-05, "loss": 0.6279, "step": 13644 }, { "epoch": 15.3, "learning_rate": 1.22e-05, "loss": 0.5499, "step": 13645 }, { "epoch": 15.3, "learning_rate": 1.2198591549295775e-05, "loss": 0.4695, "step": 13646 }, { "epoch": 15.3, "learning_rate": 1.219718309859155e-05, "loss": 0.5711, "step": 13647 }, { "epoch": 15.3, "learning_rate": 1.2195774647887324e-05, "loss": 0.5067, "step": 13648 }, { "epoch": 15.3, "learning_rate": 1.2194366197183099e-05, "loss": 0.584, "step": 13649 }, { "epoch": 15.3, "learning_rate": 1.2192957746478873e-05, "loss": 0.6062, "step": 13650 }, { "epoch": 15.3, "learning_rate": 1.219154929577465e-05, "loss": 0.6045, "step": 13651 }, { "epoch": 15.3, "learning_rate": 1.2190140845070423e-05, "loss": 0.5917, "step": 13652 }, { "epoch": 15.31, "learning_rate": 1.2188732394366197e-05, "loss": 0.5133, "step": 13653 }, { "epoch": 15.31, "learning_rate": 1.2187323943661972e-05, "loss": 0.5057, "step": 13654 }, { "epoch": 15.31, "learning_rate": 1.2185915492957746e-05, "loss": 0.5959, "step": 13655 }, { "epoch": 15.31, "learning_rate": 1.2184507042253521e-05, "loss": 0.468, "step": 13656 }, { "epoch": 15.31, "learning_rate": 1.2183098591549296e-05, "loss": 0.6013, "step": 13657 }, { "epoch": 15.31, "learning_rate": 1.2181690140845072e-05, "loss": 0.566, "step": 13658 }, { "epoch": 15.31, "learning_rate": 1.2180281690140847e-05, "loss": 0.5596, "step": 13659 }, { "epoch": 15.31, "learning_rate": 1.217887323943662e-05, "loss": 0.4759, "step": 13660 }, { "epoch": 15.32, "learning_rate": 1.2177464788732394e-05, "loss": 0.4796, "step": 13661 }, { "epoch": 15.32, "learning_rate": 1.2176056338028169e-05, "loss": 0.4568, "step": 13662 }, { "epoch": 15.32, "learning_rate": 1.2174647887323944e-05, "loss": 0.5468, "step": 13663 }, { "epoch": 15.32, "learning_rate": 1.2173239436619718e-05, "loss": 0.6122, "step": 13664 }, { "epoch": 15.32, "learning_rate": 1.2171830985915493e-05, "loss": 0.4871, "step": 13665 }, { "epoch": 15.32, "learning_rate": 1.217042253521127e-05, "loss": 0.4726, "step": 13666 }, { "epoch": 15.32, "learning_rate": 1.2169014084507044e-05, "loss": 0.587, "step": 13667 }, { "epoch": 15.32, "learning_rate": 1.2167605633802817e-05, "loss": 0.519, "step": 13668 }, { "epoch": 15.32, "learning_rate": 1.2166197183098591e-05, "loss": 0.4656, "step": 13669 }, { "epoch": 15.33, "learning_rate": 1.2164788732394366e-05, "loss": 0.5043, "step": 13670 }, { "epoch": 15.33, "learning_rate": 1.216338028169014e-05, "loss": 0.4873, "step": 13671 }, { "epoch": 15.33, "learning_rate": 1.2161971830985915e-05, "loss": 0.4308, "step": 13672 }, { "epoch": 15.33, "learning_rate": 1.2160563380281692e-05, "loss": 0.4749, "step": 13673 }, { "epoch": 15.33, "learning_rate": 1.2159154929577466e-05, "loss": 0.4429, "step": 13674 }, { "epoch": 15.33, "learning_rate": 1.2157746478873241e-05, "loss": 0.5654, "step": 13675 }, { "epoch": 15.33, "learning_rate": 1.2156338028169014e-05, "loss": 0.3833, "step": 13676 }, { "epoch": 15.33, "learning_rate": 1.2154929577464788e-05, "loss": 0.4928, "step": 13677 }, { "epoch": 15.33, "learning_rate": 1.2153521126760563e-05, "loss": 0.5334, "step": 13678 }, { "epoch": 15.34, "learning_rate": 1.2152112676056338e-05, "loss": 0.5028, "step": 13679 }, { "epoch": 15.34, "learning_rate": 1.2150704225352112e-05, "loss": 0.5876, "step": 13680 }, { "epoch": 15.34, "learning_rate": 1.2149295774647889e-05, "loss": 0.6222, "step": 13681 }, { "epoch": 15.34, "learning_rate": 1.2147887323943663e-05, "loss": 0.6223, "step": 13682 }, { "epoch": 15.34, "learning_rate": 1.2146478873239438e-05, "loss": 0.7934, "step": 13683 }, { "epoch": 15.34, "learning_rate": 1.2145070422535211e-05, "loss": 0.5447, "step": 13684 }, { "epoch": 15.34, "learning_rate": 1.2143661971830986e-05, "loss": 0.7072, "step": 13685 }, { "epoch": 15.34, "learning_rate": 1.214225352112676e-05, "loss": 0.5881, "step": 13686 }, { "epoch": 15.34, "learning_rate": 1.2140845070422535e-05, "loss": 0.6223, "step": 13687 }, { "epoch": 15.35, "learning_rate": 1.2139436619718311e-05, "loss": 0.732, "step": 13688 }, { "epoch": 15.35, "learning_rate": 1.2138028169014086e-05, "loss": 0.4687, "step": 13689 }, { "epoch": 15.35, "learning_rate": 1.213661971830986e-05, "loss": 0.5826, "step": 13690 }, { "epoch": 15.35, "learning_rate": 1.2135211267605635e-05, "loss": 0.5815, "step": 13691 }, { "epoch": 15.35, "learning_rate": 1.2133802816901408e-05, "loss": 0.6041, "step": 13692 }, { "epoch": 15.35, "learning_rate": 1.2132394366197183e-05, "loss": 0.5321, "step": 13693 }, { "epoch": 15.35, "learning_rate": 1.2130985915492957e-05, "loss": 0.5043, "step": 13694 }, { "epoch": 15.35, "learning_rate": 1.2129577464788732e-05, "loss": 0.665, "step": 13695 }, { "epoch": 15.35, "learning_rate": 1.2128169014084508e-05, "loss": 0.5043, "step": 13696 }, { "epoch": 15.36, "learning_rate": 1.2126760563380283e-05, "loss": 0.4716, "step": 13697 }, { "epoch": 15.36, "learning_rate": 1.2125352112676058e-05, "loss": 0.4333, "step": 13698 }, { "epoch": 15.36, "learning_rate": 1.212394366197183e-05, "loss": 0.521, "step": 13699 }, { "epoch": 15.36, "learning_rate": 1.2122535211267605e-05, "loss": 0.5041, "step": 13700 }, { "epoch": 15.36, "learning_rate": 1.212112676056338e-05, "loss": 0.4844, "step": 13701 }, { "epoch": 15.36, "learning_rate": 1.2119718309859154e-05, "loss": 0.472, "step": 13702 }, { "epoch": 15.36, "learning_rate": 1.2118309859154929e-05, "loss": 0.5988, "step": 13703 }, { "epoch": 15.36, "learning_rate": 1.2116901408450705e-05, "loss": 0.4395, "step": 13704 }, { "epoch": 15.36, "learning_rate": 1.211549295774648e-05, "loss": 0.5349, "step": 13705 }, { "epoch": 15.37, "learning_rate": 1.2114084507042255e-05, "loss": 0.5627, "step": 13706 }, { "epoch": 15.37, "learning_rate": 1.2112676056338028e-05, "loss": 0.6228, "step": 13707 }, { "epoch": 15.37, "learning_rate": 1.2111267605633802e-05, "loss": 0.5425, "step": 13708 }, { "epoch": 15.37, "learning_rate": 1.2109859154929577e-05, "loss": 0.6058, "step": 13709 }, { "epoch": 15.37, "learning_rate": 1.2108450704225352e-05, "loss": 0.6581, "step": 13710 }, { "epoch": 15.37, "learning_rate": 1.2107042253521128e-05, "loss": 0.5784, "step": 13711 }, { "epoch": 15.37, "learning_rate": 1.2105633802816902e-05, "loss": 0.5235, "step": 13712 }, { "epoch": 15.37, "learning_rate": 1.2104225352112677e-05, "loss": 0.5969, "step": 13713 }, { "epoch": 15.37, "learning_rate": 1.2102816901408452e-05, "loss": 0.496, "step": 13714 }, { "epoch": 15.38, "learning_rate": 1.2101408450704225e-05, "loss": 0.5595, "step": 13715 }, { "epoch": 15.38, "learning_rate": 1.21e-05, "loss": 0.6013, "step": 13716 }, { "epoch": 15.38, "learning_rate": 1.2098591549295774e-05, "loss": 0.6053, "step": 13717 }, { "epoch": 15.38, "learning_rate": 1.2097183098591549e-05, "loss": 0.5507, "step": 13718 }, { "epoch": 15.38, "learning_rate": 1.2095774647887325e-05, "loss": 0.5065, "step": 13719 }, { "epoch": 15.38, "learning_rate": 1.20943661971831e-05, "loss": 0.5259, "step": 13720 }, { "epoch": 15.38, "learning_rate": 1.2092957746478874e-05, "loss": 0.4526, "step": 13721 }, { "epoch": 15.38, "learning_rate": 1.2091549295774649e-05, "loss": 0.4627, "step": 13722 }, { "epoch": 15.38, "learning_rate": 1.2090140845070422e-05, "loss": 0.4771, "step": 13723 }, { "epoch": 15.39, "learning_rate": 1.2088732394366196e-05, "loss": 0.4159, "step": 13724 }, { "epoch": 15.39, "learning_rate": 1.2087323943661971e-05, "loss": 0.3445, "step": 13725 }, { "epoch": 15.39, "learning_rate": 1.2085915492957747e-05, "loss": 0.5438, "step": 13726 }, { "epoch": 15.39, "learning_rate": 1.2084507042253522e-05, "loss": 0.5405, "step": 13727 }, { "epoch": 15.39, "learning_rate": 1.2083098591549297e-05, "loss": 0.5918, "step": 13728 }, { "epoch": 15.39, "learning_rate": 1.2081690140845071e-05, "loss": 0.3983, "step": 13729 }, { "epoch": 15.39, "learning_rate": 1.2080281690140846e-05, "loss": 0.4217, "step": 13730 }, { "epoch": 15.39, "learning_rate": 1.2078873239436619e-05, "loss": 0.7208, "step": 13731 }, { "epoch": 15.39, "learning_rate": 1.2077464788732394e-05, "loss": 0.6307, "step": 13732 }, { "epoch": 15.4, "learning_rate": 1.2076056338028168e-05, "loss": 0.5832, "step": 13733 }, { "epoch": 15.4, "learning_rate": 1.2074647887323945e-05, "loss": 0.5196, "step": 13734 }, { "epoch": 15.4, "learning_rate": 1.207323943661972e-05, "loss": 0.5899, "step": 13735 }, { "epoch": 15.4, "learning_rate": 1.2071830985915494e-05, "loss": 0.6088, "step": 13736 }, { "epoch": 15.4, "learning_rate": 1.2070422535211268e-05, "loss": 0.4667, "step": 13737 }, { "epoch": 15.4, "learning_rate": 1.2069014084507043e-05, "loss": 0.5311, "step": 13738 }, { "epoch": 15.4, "learning_rate": 1.2067605633802816e-05, "loss": 0.7135, "step": 13739 }, { "epoch": 15.4, "learning_rate": 1.206619718309859e-05, "loss": 0.5471, "step": 13740 }, { "epoch": 15.4, "learning_rate": 1.2064788732394367e-05, "loss": 0.6188, "step": 13741 }, { "epoch": 15.41, "learning_rate": 1.2063380281690142e-05, "loss": 0.5602, "step": 13742 }, { "epoch": 15.41, "learning_rate": 1.2061971830985916e-05, "loss": 0.5079, "step": 13743 }, { "epoch": 15.41, "learning_rate": 1.2060563380281691e-05, "loss": 0.5263, "step": 13744 }, { "epoch": 15.41, "learning_rate": 1.2059154929577466e-05, "loss": 0.4197, "step": 13745 }, { "epoch": 15.41, "learning_rate": 1.205774647887324e-05, "loss": 0.5354, "step": 13746 }, { "epoch": 15.41, "learning_rate": 1.2056338028169013e-05, "loss": 0.5535, "step": 13747 }, { "epoch": 15.41, "learning_rate": 1.2054929577464788e-05, "loss": 0.525, "step": 13748 }, { "epoch": 15.41, "learning_rate": 1.2053521126760564e-05, "loss": 0.5165, "step": 13749 }, { "epoch": 15.41, "learning_rate": 1.2052112676056339e-05, "loss": 0.5332, "step": 13750 }, { "epoch": 15.42, "learning_rate": 1.2050704225352113e-05, "loss": 0.6332, "step": 13751 }, { "epoch": 15.42, "learning_rate": 1.2049295774647888e-05, "loss": 0.5277, "step": 13752 }, { "epoch": 15.42, "learning_rate": 1.2047887323943663e-05, "loss": 0.5713, "step": 13753 }, { "epoch": 15.42, "learning_rate": 1.2046478873239437e-05, "loss": 0.4349, "step": 13754 }, { "epoch": 15.42, "learning_rate": 1.204507042253521e-05, "loss": 0.5314, "step": 13755 }, { "epoch": 15.42, "learning_rate": 1.2043661971830987e-05, "loss": 0.5036, "step": 13756 }, { "epoch": 15.42, "learning_rate": 1.2042253521126761e-05, "loss": 0.4037, "step": 13757 }, { "epoch": 15.42, "learning_rate": 1.2040845070422536e-05, "loss": 0.5263, "step": 13758 }, { "epoch": 15.42, "learning_rate": 1.203943661971831e-05, "loss": 0.5988, "step": 13759 }, { "epoch": 15.43, "learning_rate": 1.2038028169014085e-05, "loss": 0.6262, "step": 13760 }, { "epoch": 15.43, "learning_rate": 1.203661971830986e-05, "loss": 0.5659, "step": 13761 }, { "epoch": 15.43, "learning_rate": 1.2035211267605634e-05, "loss": 0.5433, "step": 13762 }, { "epoch": 15.43, "learning_rate": 1.2033802816901407e-05, "loss": 0.4465, "step": 13763 }, { "epoch": 15.43, "learning_rate": 1.2032394366197184e-05, "loss": 0.4912, "step": 13764 }, { "epoch": 15.43, "learning_rate": 1.2030985915492958e-05, "loss": 0.5675, "step": 13765 }, { "epoch": 15.43, "learning_rate": 1.2029577464788733e-05, "loss": 0.6039, "step": 13766 }, { "epoch": 15.43, "learning_rate": 1.2028169014084508e-05, "loss": 0.5909, "step": 13767 }, { "epoch": 15.43, "learning_rate": 1.2026760563380282e-05, "loss": 0.3981, "step": 13768 }, { "epoch": 15.44, "learning_rate": 1.2025352112676057e-05, "loss": 0.4713, "step": 13769 }, { "epoch": 15.44, "learning_rate": 1.2023943661971831e-05, "loss": 0.4463, "step": 13770 }, { "epoch": 15.44, "learning_rate": 1.2022535211267604e-05, "loss": 0.5397, "step": 13771 }, { "epoch": 15.44, "learning_rate": 1.202112676056338e-05, "loss": 0.4736, "step": 13772 }, { "epoch": 15.44, "learning_rate": 1.2019718309859155e-05, "loss": 0.4065, "step": 13773 }, { "epoch": 15.44, "learning_rate": 1.201830985915493e-05, "loss": 0.5132, "step": 13774 }, { "epoch": 15.44, "learning_rate": 1.2016901408450705e-05, "loss": 0.3485, "step": 13775 }, { "epoch": 15.44, "learning_rate": 1.201549295774648e-05, "loss": 0.6008, "step": 13776 }, { "epoch": 15.45, "learning_rate": 1.2014084507042254e-05, "loss": 0.3603, "step": 13777 }, { "epoch": 15.45, "learning_rate": 1.2012676056338029e-05, "loss": 0.4925, "step": 13778 }, { "epoch": 15.45, "learning_rate": 1.2011267605633803e-05, "loss": 0.3863, "step": 13779 }, { "epoch": 15.45, "learning_rate": 1.2009859154929578e-05, "loss": 0.4945, "step": 13780 }, { "epoch": 15.45, "learning_rate": 1.2008450704225352e-05, "loss": 0.5107, "step": 13781 }, { "epoch": 15.45, "learning_rate": 1.2007042253521127e-05, "loss": 0.7071, "step": 13782 }, { "epoch": 15.45, "learning_rate": 1.2005633802816902e-05, "loss": 0.6004, "step": 13783 }, { "epoch": 15.45, "learning_rate": 1.2004225352112676e-05, "loss": 0.6516, "step": 13784 }, { "epoch": 15.45, "learning_rate": 1.2002816901408451e-05, "loss": 0.4968, "step": 13785 }, { "epoch": 15.46, "learning_rate": 1.2001408450704226e-05, "loss": 0.5531, "step": 13786 }, { "epoch": 15.46, "learning_rate": 1.2e-05, "loss": 0.5, "step": 13787 }, { "epoch": 15.46, "learning_rate": 1.1998591549295775e-05, "loss": 0.6319, "step": 13788 }, { "epoch": 15.46, "learning_rate": 1.199718309859155e-05, "loss": 0.5798, "step": 13789 }, { "epoch": 15.46, "learning_rate": 1.1995774647887324e-05, "loss": 0.571, "step": 13790 }, { "epoch": 15.46, "learning_rate": 1.1994366197183099e-05, "loss": 0.5602, "step": 13791 }, { "epoch": 15.46, "learning_rate": 1.1992957746478873e-05, "loss": 0.5006, "step": 13792 }, { "epoch": 15.46, "learning_rate": 1.1991549295774648e-05, "loss": 0.5504, "step": 13793 }, { "epoch": 15.46, "learning_rate": 1.1990140845070424e-05, "loss": 0.6225, "step": 13794 }, { "epoch": 15.47, "learning_rate": 1.1988732394366197e-05, "loss": 0.5128, "step": 13795 }, { "epoch": 15.47, "learning_rate": 1.1987323943661972e-05, "loss": 0.4931, "step": 13796 }, { "epoch": 15.47, "learning_rate": 1.1985915492957747e-05, "loss": 0.6373, "step": 13797 }, { "epoch": 15.47, "learning_rate": 1.1984507042253521e-05, "loss": 0.5386, "step": 13798 }, { "epoch": 15.47, "learning_rate": 1.1983098591549296e-05, "loss": 0.4963, "step": 13799 }, { "epoch": 15.47, "learning_rate": 1.198169014084507e-05, "loss": 0.554, "step": 13800 }, { "epoch": 15.47, "learning_rate": 1.1980281690140845e-05, "loss": 0.6781, "step": 13801 }, { "epoch": 15.47, "learning_rate": 1.1978873239436622e-05, "loss": 0.5531, "step": 13802 }, { "epoch": 15.47, "learning_rate": 1.1977464788732395e-05, "loss": 0.5313, "step": 13803 }, { "epoch": 15.48, "learning_rate": 1.1976056338028169e-05, "loss": 0.4044, "step": 13804 }, { "epoch": 15.48, "learning_rate": 1.1974647887323944e-05, "loss": 0.4231, "step": 13805 }, { "epoch": 15.48, "learning_rate": 1.1973239436619718e-05, "loss": 0.4972, "step": 13806 }, { "epoch": 15.48, "learning_rate": 1.1971830985915493e-05, "loss": 0.5678, "step": 13807 }, { "epoch": 15.48, "learning_rate": 1.1970422535211268e-05, "loss": 0.5512, "step": 13808 }, { "epoch": 15.48, "learning_rate": 1.1969014084507044e-05, "loss": 0.6425, "step": 13809 }, { "epoch": 15.48, "learning_rate": 1.1967605633802817e-05, "loss": 0.4951, "step": 13810 }, { "epoch": 15.48, "learning_rate": 1.1966197183098592e-05, "loss": 0.7495, "step": 13811 }, { "epoch": 15.48, "learning_rate": 1.1964788732394366e-05, "loss": 0.4517, "step": 13812 }, { "epoch": 15.49, "learning_rate": 1.1963380281690141e-05, "loss": 0.4884, "step": 13813 }, { "epoch": 15.49, "learning_rate": 1.1961971830985916e-05, "loss": 0.5119, "step": 13814 }, { "epoch": 15.49, "learning_rate": 1.196056338028169e-05, "loss": 0.4671, "step": 13815 }, { "epoch": 15.49, "learning_rate": 1.1959154929577465e-05, "loss": 0.4902, "step": 13816 }, { "epoch": 15.49, "learning_rate": 1.1957746478873241e-05, "loss": 0.5254, "step": 13817 }, { "epoch": 15.49, "learning_rate": 1.1956338028169014e-05, "loss": 0.3808, "step": 13818 }, { "epoch": 15.49, "learning_rate": 1.1954929577464789e-05, "loss": 0.483, "step": 13819 }, { "epoch": 15.49, "learning_rate": 1.1953521126760563e-05, "loss": 0.6735, "step": 13820 }, { "epoch": 15.49, "learning_rate": 1.1952112676056338e-05, "loss": 0.6482, "step": 13821 }, { "epoch": 15.5, "learning_rate": 1.1950704225352113e-05, "loss": 0.4644, "step": 13822 }, { "epoch": 15.5, "learning_rate": 1.1949295774647887e-05, "loss": 0.5827, "step": 13823 }, { "epoch": 15.5, "learning_rate": 1.1947887323943662e-05, "loss": 0.3236, "step": 13824 }, { "epoch": 15.5, "learning_rate": 1.1946478873239438e-05, "loss": 0.4925, "step": 13825 }, { "epoch": 15.5, "learning_rate": 1.1945070422535211e-05, "loss": 0.4011, "step": 13826 }, { "epoch": 15.5, "learning_rate": 1.1943661971830986e-05, "loss": 0.5052, "step": 13827 }, { "epoch": 15.5, "learning_rate": 1.194225352112676e-05, "loss": 0.626, "step": 13828 }, { "epoch": 15.5, "learning_rate": 1.1940845070422535e-05, "loss": 0.4012, "step": 13829 }, { "epoch": 15.5, "learning_rate": 1.193943661971831e-05, "loss": 0.5884, "step": 13830 }, { "epoch": 15.51, "learning_rate": 1.1938028169014084e-05, "loss": 0.5505, "step": 13831 }, { "epoch": 15.51, "learning_rate": 1.193661971830986e-05, "loss": 0.6231, "step": 13832 }, { "epoch": 15.51, "learning_rate": 1.1935211267605635e-05, "loss": 0.6958, "step": 13833 }, { "epoch": 15.51, "learning_rate": 1.1933802816901408e-05, "loss": 0.6141, "step": 13834 }, { "epoch": 15.51, "learning_rate": 1.1932394366197183e-05, "loss": 0.717, "step": 13835 }, { "epoch": 15.51, "learning_rate": 1.1930985915492958e-05, "loss": 0.5862, "step": 13836 }, { "epoch": 15.51, "learning_rate": 1.1929577464788732e-05, "loss": 0.5911, "step": 13837 }, { "epoch": 15.51, "learning_rate": 1.1928169014084507e-05, "loss": 0.6278, "step": 13838 }, { "epoch": 15.51, "learning_rate": 1.1926760563380281e-05, "loss": 0.6215, "step": 13839 }, { "epoch": 15.52, "learning_rate": 1.1925352112676058e-05, "loss": 0.5759, "step": 13840 }, { "epoch": 15.52, "learning_rate": 1.1923943661971832e-05, "loss": 0.4352, "step": 13841 }, { "epoch": 15.52, "learning_rate": 1.1922535211267605e-05, "loss": 0.6319, "step": 13842 }, { "epoch": 15.52, "learning_rate": 1.192112676056338e-05, "loss": 0.5979, "step": 13843 }, { "epoch": 15.52, "learning_rate": 1.1919718309859155e-05, "loss": 0.6439, "step": 13844 }, { "epoch": 15.52, "learning_rate": 1.191830985915493e-05, "loss": 0.5299, "step": 13845 }, { "epoch": 15.52, "learning_rate": 1.1916901408450704e-05, "loss": 0.4855, "step": 13846 }, { "epoch": 15.52, "learning_rate": 1.191549295774648e-05, "loss": 0.5293, "step": 13847 }, { "epoch": 15.52, "learning_rate": 1.1914084507042255e-05, "loss": 0.5348, "step": 13848 }, { "epoch": 15.53, "learning_rate": 1.191267605633803e-05, "loss": 0.6103, "step": 13849 }, { "epoch": 15.53, "learning_rate": 1.1911267605633802e-05, "loss": 0.7268, "step": 13850 }, { "epoch": 15.53, "learning_rate": 1.1909859154929577e-05, "loss": 0.6032, "step": 13851 }, { "epoch": 15.53, "learning_rate": 1.1908450704225352e-05, "loss": 0.4988, "step": 13852 }, { "epoch": 15.53, "learning_rate": 1.1907042253521126e-05, "loss": 0.538, "step": 13853 }, { "epoch": 15.53, "learning_rate": 1.1905633802816901e-05, "loss": 0.5122, "step": 13854 }, { "epoch": 15.53, "learning_rate": 1.1904225352112677e-05, "loss": 0.5166, "step": 13855 }, { "epoch": 15.53, "learning_rate": 1.1902816901408452e-05, "loss": 0.4873, "step": 13856 }, { "epoch": 15.53, "learning_rate": 1.1901408450704227e-05, "loss": 0.501, "step": 13857 }, { "epoch": 15.54, "learning_rate": 1.19e-05, "loss": 0.5864, "step": 13858 }, { "epoch": 15.54, "learning_rate": 1.1898591549295774e-05, "loss": 0.4797, "step": 13859 }, { "epoch": 15.54, "learning_rate": 1.1897183098591549e-05, "loss": 0.5933, "step": 13860 }, { "epoch": 15.54, "learning_rate": 1.1895774647887323e-05, "loss": 0.4695, "step": 13861 }, { "epoch": 15.54, "learning_rate": 1.18943661971831e-05, "loss": 0.4964, "step": 13862 }, { "epoch": 15.54, "learning_rate": 1.1892957746478874e-05, "loss": 0.4901, "step": 13863 }, { "epoch": 15.54, "learning_rate": 1.1891549295774649e-05, "loss": 0.5449, "step": 13864 }, { "epoch": 15.54, "learning_rate": 1.1890140845070424e-05, "loss": 0.521, "step": 13865 }, { "epoch": 15.54, "learning_rate": 1.1888732394366197e-05, "loss": 0.4465, "step": 13866 }, { "epoch": 15.55, "learning_rate": 1.1887323943661971e-05, "loss": 0.5587, "step": 13867 }, { "epoch": 15.55, "learning_rate": 1.1885915492957746e-05, "loss": 0.5184, "step": 13868 }, { "epoch": 15.55, "learning_rate": 1.188450704225352e-05, "loss": 0.5688, "step": 13869 }, { "epoch": 15.55, "learning_rate": 1.1883098591549297e-05, "loss": 0.5271, "step": 13870 }, { "epoch": 15.55, "learning_rate": 1.1881690140845072e-05, "loss": 0.5592, "step": 13871 }, { "epoch": 15.55, "learning_rate": 1.1880281690140846e-05, "loss": 0.5661, "step": 13872 }, { "epoch": 15.55, "learning_rate": 1.187887323943662e-05, "loss": 0.4638, "step": 13873 }, { "epoch": 15.55, "learning_rate": 1.1877464788732394e-05, "loss": 0.3149, "step": 13874 }, { "epoch": 15.55, "learning_rate": 1.1876056338028168e-05, "loss": 0.4698, "step": 13875 }, { "epoch": 15.56, "learning_rate": 1.1874647887323943e-05, "loss": 0.5126, "step": 13876 }, { "epoch": 15.56, "learning_rate": 1.1873239436619718e-05, "loss": 0.3808, "step": 13877 }, { "epoch": 15.56, "learning_rate": 1.1871830985915494e-05, "loss": 0.4133, "step": 13878 }, { "epoch": 15.56, "learning_rate": 1.1870422535211269e-05, "loss": 0.5353, "step": 13879 }, { "epoch": 15.56, "learning_rate": 1.1869014084507043e-05, "loss": 0.3265, "step": 13880 }, { "epoch": 15.56, "learning_rate": 1.1867605633802818e-05, "loss": 0.743, "step": 13881 }, { "epoch": 15.56, "learning_rate": 1.1866197183098591e-05, "loss": 0.7469, "step": 13882 }, { "epoch": 15.56, "learning_rate": 1.1864788732394366e-05, "loss": 0.5165, "step": 13883 }, { "epoch": 15.57, "learning_rate": 1.186338028169014e-05, "loss": 0.697, "step": 13884 }, { "epoch": 15.57, "learning_rate": 1.1861971830985916e-05, "loss": 0.581, "step": 13885 }, { "epoch": 15.57, "learning_rate": 1.1860563380281691e-05, "loss": 0.5945, "step": 13886 }, { "epoch": 15.57, "learning_rate": 1.1859154929577466e-05, "loss": 0.4822, "step": 13887 }, { "epoch": 15.57, "learning_rate": 1.185774647887324e-05, "loss": 0.5311, "step": 13888 }, { "epoch": 15.57, "learning_rate": 1.1856338028169015e-05, "loss": 0.4918, "step": 13889 }, { "epoch": 15.57, "learning_rate": 1.1854929577464788e-05, "loss": 0.5823, "step": 13890 }, { "epoch": 15.57, "learning_rate": 1.1853521126760563e-05, "loss": 0.5258, "step": 13891 }, { "epoch": 15.57, "learning_rate": 1.1852112676056337e-05, "loss": 0.4773, "step": 13892 }, { "epoch": 15.58, "learning_rate": 1.1850704225352114e-05, "loss": 0.7082, "step": 13893 }, { "epoch": 15.58, "learning_rate": 1.1849295774647888e-05, "loss": 0.5327, "step": 13894 }, { "epoch": 15.58, "learning_rate": 1.1847887323943663e-05, "loss": 0.4819, "step": 13895 }, { "epoch": 15.58, "learning_rate": 1.1846478873239437e-05, "loss": 0.5524, "step": 13896 }, { "epoch": 15.58, "learning_rate": 1.1845070422535212e-05, "loss": 0.5086, "step": 13897 }, { "epoch": 15.58, "learning_rate": 1.1843661971830985e-05, "loss": 0.7164, "step": 13898 }, { "epoch": 15.58, "learning_rate": 1.184225352112676e-05, "loss": 0.5239, "step": 13899 }, { "epoch": 15.58, "learning_rate": 1.1840845070422536e-05, "loss": 0.6155, "step": 13900 }, { "epoch": 15.58, "learning_rate": 1.183943661971831e-05, "loss": 0.5374, "step": 13901 }, { "epoch": 15.59, "learning_rate": 1.1838028169014085e-05, "loss": 0.6087, "step": 13902 }, { "epoch": 15.59, "learning_rate": 1.183661971830986e-05, "loss": 0.4644, "step": 13903 }, { "epoch": 15.59, "learning_rate": 1.1835211267605635e-05, "loss": 0.5094, "step": 13904 }, { "epoch": 15.59, "learning_rate": 1.183380281690141e-05, "loss": 0.4833, "step": 13905 }, { "epoch": 15.59, "learning_rate": 1.1832394366197182e-05, "loss": 0.4425, "step": 13906 }, { "epoch": 15.59, "learning_rate": 1.1830985915492957e-05, "loss": 0.4689, "step": 13907 }, { "epoch": 15.59, "learning_rate": 1.1829577464788733e-05, "loss": 0.4323, "step": 13908 }, { "epoch": 15.59, "learning_rate": 1.1828169014084508e-05, "loss": 0.5391, "step": 13909 }, { "epoch": 15.59, "learning_rate": 1.1826760563380282e-05, "loss": 0.4568, "step": 13910 }, { "epoch": 15.6, "learning_rate": 1.1825352112676057e-05, "loss": 0.5134, "step": 13911 }, { "epoch": 15.6, "learning_rate": 1.1823943661971832e-05, "loss": 0.5688, "step": 13912 }, { "epoch": 15.6, "learning_rate": 1.1822535211267606e-05, "loss": 0.5766, "step": 13913 }, { "epoch": 15.6, "learning_rate": 1.182112676056338e-05, "loss": 0.4951, "step": 13914 }, { "epoch": 15.6, "learning_rate": 1.1819718309859156e-05, "loss": 0.5805, "step": 13915 }, { "epoch": 15.6, "learning_rate": 1.181830985915493e-05, "loss": 0.5518, "step": 13916 }, { "epoch": 15.6, "learning_rate": 1.1816901408450705e-05, "loss": 0.4737, "step": 13917 }, { "epoch": 15.6, "learning_rate": 1.181549295774648e-05, "loss": 0.5297, "step": 13918 }, { "epoch": 15.6, "learning_rate": 1.1814084507042254e-05, "loss": 0.4238, "step": 13919 }, { "epoch": 15.61, "learning_rate": 1.1812676056338029e-05, "loss": 0.494, "step": 13920 }, { "epoch": 15.61, "learning_rate": 1.1811267605633803e-05, "loss": 0.5045, "step": 13921 }, { "epoch": 15.61, "learning_rate": 1.1809859154929576e-05, "loss": 0.5224, "step": 13922 }, { "epoch": 15.61, "learning_rate": 1.1808450704225353e-05, "loss": 0.4203, "step": 13923 }, { "epoch": 15.61, "learning_rate": 1.1807042253521127e-05, "loss": 0.3612, "step": 13924 }, { "epoch": 15.61, "learning_rate": 1.1805633802816902e-05, "loss": 0.4641, "step": 13925 }, { "epoch": 15.61, "learning_rate": 1.1804225352112677e-05, "loss": 0.4578, "step": 13926 }, { "epoch": 15.61, "learning_rate": 1.1802816901408451e-05, "loss": 0.4216, "step": 13927 }, { "epoch": 15.61, "learning_rate": 1.1801408450704226e-05, "loss": 0.4034, "step": 13928 }, { "epoch": 15.62, "learning_rate": 1.1799999999999999e-05, "loss": 0.5347, "step": 13929 }, { "epoch": 15.62, "learning_rate": 1.1798591549295775e-05, "loss": 0.6987, "step": 13930 }, { "epoch": 15.62, "learning_rate": 1.179718309859155e-05, "loss": 0.6575, "step": 13931 }, { "epoch": 15.62, "learning_rate": 1.1795774647887324e-05, "loss": 0.6921, "step": 13932 }, { "epoch": 15.62, "learning_rate": 1.1794366197183099e-05, "loss": 0.6344, "step": 13933 }, { "epoch": 15.62, "learning_rate": 1.1792957746478874e-05, "loss": 0.542, "step": 13934 }, { "epoch": 15.62, "learning_rate": 1.1791549295774648e-05, "loss": 0.5778, "step": 13935 }, { "epoch": 15.62, "learning_rate": 1.1790140845070423e-05, "loss": 0.4947, "step": 13936 }, { "epoch": 15.62, "learning_rate": 1.1788732394366196e-05, "loss": 0.5681, "step": 13937 }, { "epoch": 15.63, "learning_rate": 1.1787323943661972e-05, "loss": 0.6252, "step": 13938 }, { "epoch": 15.63, "learning_rate": 1.1785915492957747e-05, "loss": 0.4938, "step": 13939 }, { "epoch": 15.63, "learning_rate": 1.1784507042253522e-05, "loss": 0.5283, "step": 13940 }, { "epoch": 15.63, "learning_rate": 1.1783098591549296e-05, "loss": 0.4791, "step": 13941 }, { "epoch": 15.63, "learning_rate": 1.178169014084507e-05, "loss": 0.6006, "step": 13942 }, { "epoch": 15.63, "learning_rate": 1.1780281690140845e-05, "loss": 0.5541, "step": 13943 }, { "epoch": 15.63, "learning_rate": 1.177887323943662e-05, "loss": 0.5169, "step": 13944 }, { "epoch": 15.63, "learning_rate": 1.1777464788732393e-05, "loss": 0.5156, "step": 13945 }, { "epoch": 15.63, "learning_rate": 1.177605633802817e-05, "loss": 0.599, "step": 13946 }, { "epoch": 15.64, "learning_rate": 1.1774647887323944e-05, "loss": 0.5058, "step": 13947 }, { "epoch": 15.64, "learning_rate": 1.1773239436619719e-05, "loss": 0.5479, "step": 13948 }, { "epoch": 15.64, "learning_rate": 1.1771830985915493e-05, "loss": 0.6387, "step": 13949 }, { "epoch": 15.64, "learning_rate": 1.1770422535211268e-05, "loss": 0.4828, "step": 13950 }, { "epoch": 15.64, "learning_rate": 1.1769014084507043e-05, "loss": 0.5334, "step": 13951 }, { "epoch": 15.64, "learning_rate": 1.1767605633802817e-05, "loss": 0.4186, "step": 13952 }, { "epoch": 15.64, "learning_rate": 1.1766197183098592e-05, "loss": 0.5064, "step": 13953 }, { "epoch": 15.64, "learning_rate": 1.1764788732394366e-05, "loss": 0.5432, "step": 13954 }, { "epoch": 15.64, "learning_rate": 1.1763380281690141e-05, "loss": 0.4497, "step": 13955 }, { "epoch": 15.65, "learning_rate": 1.1761971830985916e-05, "loss": 0.5871, "step": 13956 }, { "epoch": 15.65, "learning_rate": 1.176056338028169e-05, "loss": 0.4658, "step": 13957 }, { "epoch": 15.65, "learning_rate": 1.1759154929577465e-05, "loss": 0.5369, "step": 13958 }, { "epoch": 15.65, "learning_rate": 1.175774647887324e-05, "loss": 0.5261, "step": 13959 }, { "epoch": 15.65, "learning_rate": 1.1756338028169014e-05, "loss": 0.5668, "step": 13960 }, { "epoch": 15.65, "learning_rate": 1.1754929577464789e-05, "loss": 0.507, "step": 13961 }, { "epoch": 15.65, "learning_rate": 1.1753521126760564e-05, "loss": 0.4445, "step": 13962 }, { "epoch": 15.65, "learning_rate": 1.1752112676056338e-05, "loss": 0.4694, "step": 13963 }, { "epoch": 15.65, "learning_rate": 1.1750704225352113e-05, "loss": 0.6211, "step": 13964 }, { "epoch": 15.66, "learning_rate": 1.1749295774647887e-05, "loss": 0.4253, "step": 13965 }, { "epoch": 15.66, "learning_rate": 1.1747887323943662e-05, "loss": 0.5894, "step": 13966 }, { "epoch": 15.66, "learning_rate": 1.1746478873239437e-05, "loss": 0.6234, "step": 13967 }, { "epoch": 15.66, "learning_rate": 1.1745070422535213e-05, "loss": 0.676, "step": 13968 }, { "epoch": 15.66, "learning_rate": 1.1743661971830986e-05, "loss": 0.5074, "step": 13969 }, { "epoch": 15.66, "learning_rate": 1.174225352112676e-05, "loss": 0.5145, "step": 13970 }, { "epoch": 15.66, "learning_rate": 1.1740845070422535e-05, "loss": 0.5884, "step": 13971 }, { "epoch": 15.66, "learning_rate": 1.173943661971831e-05, "loss": 0.542, "step": 13972 }, { "epoch": 15.66, "learning_rate": 1.1738028169014085e-05, "loss": 0.5275, "step": 13973 }, { "epoch": 15.67, "learning_rate": 1.173661971830986e-05, "loss": 0.5191, "step": 13974 }, { "epoch": 15.67, "learning_rate": 1.1735211267605634e-05, "loss": 0.3728, "step": 13975 }, { "epoch": 15.67, "learning_rate": 1.173380281690141e-05, "loss": 0.5747, "step": 13976 }, { "epoch": 15.67, "learning_rate": 1.1732394366197183e-05, "loss": 0.4187, "step": 13977 }, { "epoch": 15.67, "learning_rate": 1.1730985915492958e-05, "loss": 0.6486, "step": 13978 }, { "epoch": 15.67, "learning_rate": 1.1729577464788732e-05, "loss": 0.3788, "step": 13979 }, { "epoch": 15.67, "learning_rate": 1.1728169014084507e-05, "loss": 0.6784, "step": 13980 }, { "epoch": 15.67, "learning_rate": 1.1726760563380282e-05, "loss": 0.7625, "step": 13981 }, { "epoch": 15.67, "learning_rate": 1.1725352112676056e-05, "loss": 0.5327, "step": 13982 }, { "epoch": 15.68, "learning_rate": 1.1723943661971833e-05, "loss": 0.4546, "step": 13983 }, { "epoch": 15.68, "learning_rate": 1.1722535211267607e-05, "loss": 0.6697, "step": 13984 }, { "epoch": 15.68, "learning_rate": 1.172112676056338e-05, "loss": 0.5577, "step": 13985 }, { "epoch": 15.68, "learning_rate": 1.1719718309859155e-05, "loss": 0.5577, "step": 13986 }, { "epoch": 15.68, "learning_rate": 1.171830985915493e-05, "loss": 0.6212, "step": 13987 }, { "epoch": 15.68, "learning_rate": 1.1716901408450704e-05, "loss": 0.6195, "step": 13988 }, { "epoch": 15.68, "learning_rate": 1.1715492957746479e-05, "loss": 0.4564, "step": 13989 }, { "epoch": 15.68, "learning_rate": 1.1714084507042253e-05, "loss": 0.5106, "step": 13990 }, { "epoch": 15.68, "learning_rate": 1.171267605633803e-05, "loss": 0.5508, "step": 13991 }, { "epoch": 15.69, "learning_rate": 1.1711267605633804e-05, "loss": 0.5272, "step": 13992 }, { "epoch": 15.69, "learning_rate": 1.1709859154929577e-05, "loss": 0.4253, "step": 13993 }, { "epoch": 15.69, "learning_rate": 1.1708450704225352e-05, "loss": 0.578, "step": 13994 }, { "epoch": 15.69, "learning_rate": 1.1707042253521127e-05, "loss": 0.5598, "step": 13995 }, { "epoch": 15.69, "learning_rate": 1.1705633802816901e-05, "loss": 0.4843, "step": 13996 }, { "epoch": 15.69, "learning_rate": 1.1704225352112676e-05, "loss": 0.3894, "step": 13997 }, { "epoch": 15.69, "learning_rate": 1.170281690140845e-05, "loss": 0.4392, "step": 13998 }, { "epoch": 15.69, "learning_rate": 1.1701408450704227e-05, "loss": 0.4165, "step": 13999 }, { "epoch": 15.7, "learning_rate": 1.1700000000000001e-05, "loss": 0.5975, "step": 14000 }, { "epoch": 15.7, "learning_rate": 1.1698591549295774e-05, "loss": 0.4934, "step": 14001 }, { "epoch": 15.7, "learning_rate": 1.1697183098591549e-05, "loss": 0.4494, "step": 14002 }, { "epoch": 15.7, "learning_rate": 1.1695774647887324e-05, "loss": 0.4157, "step": 14003 }, { "epoch": 15.7, "learning_rate": 1.1694366197183098e-05, "loss": 0.5009, "step": 14004 }, { "epoch": 15.7, "learning_rate": 1.1692957746478873e-05, "loss": 0.4921, "step": 14005 }, { "epoch": 15.7, "learning_rate": 1.169154929577465e-05, "loss": 0.4382, "step": 14006 }, { "epoch": 15.7, "learning_rate": 1.1690140845070424e-05, "loss": 0.4499, "step": 14007 }, { "epoch": 15.7, "learning_rate": 1.1688732394366199e-05, "loss": 0.5073, "step": 14008 }, { "epoch": 15.71, "learning_rate": 1.1687323943661972e-05, "loss": 0.5125, "step": 14009 }, { "epoch": 15.71, "learning_rate": 1.1685915492957746e-05, "loss": 0.4912, "step": 14010 }, { "epoch": 15.71, "learning_rate": 1.168450704225352e-05, "loss": 0.5114, "step": 14011 }, { "epoch": 15.71, "learning_rate": 1.1683098591549295e-05, "loss": 0.4317, "step": 14012 }, { "epoch": 15.71, "learning_rate": 1.168169014084507e-05, "loss": 0.4963, "step": 14013 }, { "epoch": 15.71, "learning_rate": 1.1680281690140846e-05, "loss": 0.4046, "step": 14014 }, { "epoch": 15.71, "learning_rate": 1.1678873239436621e-05, "loss": 0.44, "step": 14015 }, { "epoch": 15.71, "learning_rate": 1.1677464788732396e-05, "loss": 0.4669, "step": 14016 }, { "epoch": 15.71, "learning_rate": 1.1676056338028169e-05, "loss": 0.4884, "step": 14017 }, { "epoch": 15.72, "learning_rate": 1.1674647887323943e-05, "loss": 0.3783, "step": 14018 }, { "epoch": 15.72, "learning_rate": 1.1673239436619718e-05, "loss": 0.5766, "step": 14019 }, { "epoch": 15.72, "learning_rate": 1.1671830985915493e-05, "loss": 0.4507, "step": 14020 }, { "epoch": 15.72, "learning_rate": 1.1670422535211269e-05, "loss": 0.5138, "step": 14021 }, { "epoch": 15.72, "learning_rate": 1.1669014084507044e-05, "loss": 0.5463, "step": 14022 }, { "epoch": 15.72, "learning_rate": 1.1667605633802818e-05, "loss": 0.4274, "step": 14023 }, { "epoch": 15.72, "learning_rate": 1.1666197183098593e-05, "loss": 0.5374, "step": 14024 }, { "epoch": 15.72, "learning_rate": 1.1664788732394366e-05, "loss": 0.4166, "step": 14025 }, { "epoch": 15.72, "learning_rate": 1.166338028169014e-05, "loss": 0.4086, "step": 14026 }, { "epoch": 15.73, "learning_rate": 1.1661971830985915e-05, "loss": 0.4559, "step": 14027 }, { "epoch": 15.73, "learning_rate": 1.166056338028169e-05, "loss": 0.4625, "step": 14028 }, { "epoch": 15.73, "learning_rate": 1.1659154929577466e-05, "loss": 0.6378, "step": 14029 }, { "epoch": 15.73, "learning_rate": 1.165774647887324e-05, "loss": 0.4686, "step": 14030 }, { "epoch": 15.73, "learning_rate": 1.1656338028169015e-05, "loss": 0.663, "step": 14031 }, { "epoch": 15.73, "learning_rate": 1.165492957746479e-05, "loss": 0.5419, "step": 14032 }, { "epoch": 15.73, "learning_rate": 1.1653521126760563e-05, "loss": 0.5316, "step": 14033 }, { "epoch": 15.73, "learning_rate": 1.1652112676056337e-05, "loss": 0.6098, "step": 14034 }, { "epoch": 15.73, "learning_rate": 1.1650704225352112e-05, "loss": 0.6304, "step": 14035 }, { "epoch": 15.74, "learning_rate": 1.1649295774647888e-05, "loss": 0.4724, "step": 14036 }, { "epoch": 15.74, "learning_rate": 1.1647887323943663e-05, "loss": 0.6094, "step": 14037 }, { "epoch": 15.74, "learning_rate": 1.1646478873239438e-05, "loss": 0.6201, "step": 14038 }, { "epoch": 15.74, "learning_rate": 1.1645070422535212e-05, "loss": 0.4806, "step": 14039 }, { "epoch": 15.74, "learning_rate": 1.1643661971830985e-05, "loss": 0.4389, "step": 14040 }, { "epoch": 15.74, "learning_rate": 1.164225352112676e-05, "loss": 0.4785, "step": 14041 }, { "epoch": 15.74, "learning_rate": 1.1640845070422535e-05, "loss": 0.6685, "step": 14042 }, { "epoch": 15.74, "learning_rate": 1.163943661971831e-05, "loss": 0.4836, "step": 14043 }, { "epoch": 15.74, "learning_rate": 1.1638028169014086e-05, "loss": 0.5797, "step": 14044 }, { "epoch": 15.75, "learning_rate": 1.163661971830986e-05, "loss": 0.4879, "step": 14045 }, { "epoch": 15.75, "learning_rate": 1.1635211267605635e-05, "loss": 0.536, "step": 14046 }, { "epoch": 15.75, "learning_rate": 1.163380281690141e-05, "loss": 0.4941, "step": 14047 }, { "epoch": 15.75, "learning_rate": 1.1632394366197182e-05, "loss": 0.4817, "step": 14048 }, { "epoch": 15.75, "learning_rate": 1.1630985915492957e-05, "loss": 0.4655, "step": 14049 }, { "epoch": 15.75, "learning_rate": 1.1629577464788732e-05, "loss": 0.5107, "step": 14050 }, { "epoch": 15.75, "learning_rate": 1.1628169014084508e-05, "loss": 0.5437, "step": 14051 }, { "epoch": 15.75, "learning_rate": 1.1626760563380283e-05, "loss": 0.5113, "step": 14052 }, { "epoch": 15.75, "learning_rate": 1.1625352112676057e-05, "loss": 0.5748, "step": 14053 }, { "epoch": 15.76, "learning_rate": 1.1623943661971832e-05, "loss": 0.4305, "step": 14054 }, { "epoch": 15.76, "learning_rate": 1.1622535211267607e-05, "loss": 0.3735, "step": 14055 }, { "epoch": 15.76, "learning_rate": 1.162112676056338e-05, "loss": 0.5113, "step": 14056 }, { "epoch": 15.76, "learning_rate": 1.1619718309859154e-05, "loss": 0.4996, "step": 14057 }, { "epoch": 15.76, "learning_rate": 1.1618309859154929e-05, "loss": 0.4403, "step": 14058 }, { "epoch": 15.76, "learning_rate": 1.1616901408450705e-05, "loss": 0.4105, "step": 14059 }, { "epoch": 15.76, "learning_rate": 1.161549295774648e-05, "loss": 0.6279, "step": 14060 }, { "epoch": 15.76, "learning_rate": 1.1614084507042254e-05, "loss": 0.6353, "step": 14061 }, { "epoch": 15.76, "learning_rate": 1.1612676056338029e-05, "loss": 0.5428, "step": 14062 }, { "epoch": 15.77, "learning_rate": 1.1611267605633804e-05, "loss": 0.4354, "step": 14063 }, { "epoch": 15.77, "learning_rate": 1.1609859154929577e-05, "loss": 0.5511, "step": 14064 }, { "epoch": 15.77, "learning_rate": 1.1608450704225351e-05, "loss": 0.4637, "step": 14065 }, { "epoch": 15.77, "learning_rate": 1.1607042253521126e-05, "loss": 0.4374, "step": 14066 }, { "epoch": 15.77, "learning_rate": 1.1605633802816902e-05, "loss": 0.4969, "step": 14067 }, { "epoch": 15.77, "learning_rate": 1.1604225352112677e-05, "loss": 0.5222, "step": 14068 }, { "epoch": 15.77, "learning_rate": 1.1602816901408451e-05, "loss": 0.6331, "step": 14069 }, { "epoch": 15.77, "learning_rate": 1.1601408450704226e-05, "loss": 0.5183, "step": 14070 }, { "epoch": 15.77, "learning_rate": 1.16e-05, "loss": 0.4029, "step": 14071 }, { "epoch": 15.78, "learning_rate": 1.1598591549295774e-05, "loss": 0.4159, "step": 14072 }, { "epoch": 15.78, "learning_rate": 1.1597183098591548e-05, "loss": 0.6839, "step": 14073 }, { "epoch": 15.78, "learning_rate": 1.1595774647887325e-05, "loss": 0.3574, "step": 14074 }, { "epoch": 15.78, "learning_rate": 1.15943661971831e-05, "loss": 0.5996, "step": 14075 }, { "epoch": 15.78, "learning_rate": 1.1592957746478874e-05, "loss": 0.4669, "step": 14076 }, { "epoch": 15.78, "learning_rate": 1.1591549295774649e-05, "loss": 0.3693, "step": 14077 }, { "epoch": 15.78, "learning_rate": 1.1590140845070423e-05, "loss": 0.5707, "step": 14078 }, { "epoch": 15.78, "learning_rate": 1.1588732394366198e-05, "loss": 0.5961, "step": 14079 }, { "epoch": 15.78, "learning_rate": 1.158732394366197e-05, "loss": 0.5005, "step": 14080 }, { "epoch": 15.79, "learning_rate": 1.1585915492957745e-05, "loss": 0.6771, "step": 14081 }, { "epoch": 15.79, "learning_rate": 1.1584507042253522e-05, "loss": 0.6989, "step": 14082 }, { "epoch": 15.79, "learning_rate": 1.1583098591549296e-05, "loss": 0.6023, "step": 14083 }, { "epoch": 15.79, "learning_rate": 1.1581690140845071e-05, "loss": 0.4872, "step": 14084 }, { "epoch": 15.79, "learning_rate": 1.1580281690140846e-05, "loss": 0.6186, "step": 14085 }, { "epoch": 15.79, "learning_rate": 1.157887323943662e-05, "loss": 0.4604, "step": 14086 }, { "epoch": 15.79, "learning_rate": 1.1577464788732395e-05, "loss": 0.5838, "step": 14087 }, { "epoch": 15.79, "learning_rate": 1.1576056338028168e-05, "loss": 0.585, "step": 14088 }, { "epoch": 15.79, "learning_rate": 1.1574647887323944e-05, "loss": 0.6037, "step": 14089 }, { "epoch": 15.8, "learning_rate": 1.1573239436619719e-05, "loss": 0.5995, "step": 14090 }, { "epoch": 15.8, "learning_rate": 1.1571830985915493e-05, "loss": 0.6467, "step": 14091 }, { "epoch": 15.8, "learning_rate": 1.1570422535211268e-05, "loss": 0.4647, "step": 14092 }, { "epoch": 15.8, "learning_rate": 1.1569014084507043e-05, "loss": 0.5563, "step": 14093 }, { "epoch": 15.8, "learning_rate": 1.1567605633802817e-05, "loss": 0.4985, "step": 14094 }, { "epoch": 15.8, "learning_rate": 1.1566197183098592e-05, "loss": 0.7628, "step": 14095 }, { "epoch": 15.8, "learning_rate": 1.1564788732394365e-05, "loss": 0.4821, "step": 14096 }, { "epoch": 15.8, "learning_rate": 1.1563380281690141e-05, "loss": 0.5734, "step": 14097 }, { "epoch": 15.8, "learning_rate": 1.1561971830985916e-05, "loss": 0.5658, "step": 14098 }, { "epoch": 15.81, "learning_rate": 1.156056338028169e-05, "loss": 0.4648, "step": 14099 }, { "epoch": 15.81, "learning_rate": 1.1559154929577465e-05, "loss": 0.554, "step": 14100 }, { "epoch": 15.81, "learning_rate": 1.155774647887324e-05, "loss": 0.5723, "step": 14101 }, { "epoch": 15.81, "learning_rate": 1.1556338028169015e-05, "loss": 0.4757, "step": 14102 }, { "epoch": 15.81, "learning_rate": 1.1554929577464789e-05, "loss": 0.4233, "step": 14103 }, { "epoch": 15.81, "learning_rate": 1.1553521126760564e-05, "loss": 0.5975, "step": 14104 }, { "epoch": 15.81, "learning_rate": 1.1552112676056338e-05, "loss": 0.5623, "step": 14105 }, { "epoch": 15.81, "learning_rate": 1.1550704225352113e-05, "loss": 0.5458, "step": 14106 }, { "epoch": 15.82, "learning_rate": 1.1549295774647888e-05, "loss": 0.4768, "step": 14107 }, { "epoch": 15.82, "learning_rate": 1.1547887323943662e-05, "loss": 0.4027, "step": 14108 }, { "epoch": 15.82, "learning_rate": 1.1546478873239437e-05, "loss": 0.4351, "step": 14109 }, { "epoch": 15.82, "learning_rate": 1.1545070422535212e-05, "loss": 0.5384, "step": 14110 }, { "epoch": 15.82, "learning_rate": 1.1543661971830986e-05, "loss": 0.5638, "step": 14111 }, { "epoch": 15.82, "learning_rate": 1.1542253521126761e-05, "loss": 0.4484, "step": 14112 }, { "epoch": 15.82, "learning_rate": 1.1540845070422536e-05, "loss": 0.4207, "step": 14113 }, { "epoch": 15.82, "learning_rate": 1.153943661971831e-05, "loss": 0.509, "step": 14114 }, { "epoch": 15.82, "learning_rate": 1.1538028169014085e-05, "loss": 0.6737, "step": 14115 }, { "epoch": 15.83, "learning_rate": 1.153661971830986e-05, "loss": 0.5175, "step": 14116 }, { "epoch": 15.83, "learning_rate": 1.1535211267605634e-05, "loss": 0.5857, "step": 14117 }, { "epoch": 15.83, "learning_rate": 1.1533802816901409e-05, "loss": 0.4927, "step": 14118 }, { "epoch": 15.83, "learning_rate": 1.1532394366197183e-05, "loss": 0.5622, "step": 14119 }, { "epoch": 15.83, "learning_rate": 1.1530985915492958e-05, "loss": 0.4405, "step": 14120 }, { "epoch": 15.83, "learning_rate": 1.1529577464788733e-05, "loss": 0.697, "step": 14121 }, { "epoch": 15.83, "learning_rate": 1.1528169014084507e-05, "loss": 0.445, "step": 14122 }, { "epoch": 15.83, "learning_rate": 1.1526760563380282e-05, "loss": 0.6277, "step": 14123 }, { "epoch": 15.83, "learning_rate": 1.1525352112676057e-05, "loss": 0.4036, "step": 14124 }, { "epoch": 15.84, "learning_rate": 1.1523943661971831e-05, "loss": 0.5647, "step": 14125 }, { "epoch": 15.84, "learning_rate": 1.1522535211267606e-05, "loss": 0.5518, "step": 14126 }, { "epoch": 15.84, "learning_rate": 1.1521126760563382e-05, "loss": 0.6045, "step": 14127 }, { "epoch": 15.84, "learning_rate": 1.1519718309859155e-05, "loss": 0.45, "step": 14128 }, { "epoch": 15.84, "learning_rate": 1.151830985915493e-05, "loss": 0.4168, "step": 14129 }, { "epoch": 15.84, "learning_rate": 1.1516901408450704e-05, "loss": 0.3196, "step": 14130 }, { "epoch": 15.84, "learning_rate": 1.1515492957746479e-05, "loss": 0.7771, "step": 14131 }, { "epoch": 15.84, "learning_rate": 1.1514084507042254e-05, "loss": 0.6328, "step": 14132 }, { "epoch": 15.84, "learning_rate": 1.1512676056338028e-05, "loss": 0.528, "step": 14133 }, { "epoch": 15.85, "learning_rate": 1.1511267605633803e-05, "loss": 0.5593, "step": 14134 }, { "epoch": 15.85, "learning_rate": 1.150985915492958e-05, "loss": 0.5537, "step": 14135 }, { "epoch": 15.85, "learning_rate": 1.1508450704225352e-05, "loss": 0.6033, "step": 14136 }, { "epoch": 15.85, "learning_rate": 1.1507042253521127e-05, "loss": 0.5517, "step": 14137 }, { "epoch": 15.85, "learning_rate": 1.1505633802816901e-05, "loss": 0.5024, "step": 14138 }, { "epoch": 15.85, "learning_rate": 1.1504225352112676e-05, "loss": 0.5161, "step": 14139 }, { "epoch": 15.85, "learning_rate": 1.150281690140845e-05, "loss": 0.5248, "step": 14140 }, { "epoch": 15.85, "learning_rate": 1.1501408450704225e-05, "loss": 0.5301, "step": 14141 }, { "epoch": 15.85, "learning_rate": 1.1500000000000002e-05, "loss": 0.5759, "step": 14142 }, { "epoch": 15.86, "learning_rate": 1.1498591549295776e-05, "loss": 0.4762, "step": 14143 }, { "epoch": 15.86, "learning_rate": 1.149718309859155e-05, "loss": 0.4938, "step": 14144 }, { "epoch": 15.86, "learning_rate": 1.1495774647887324e-05, "loss": 0.549, "step": 14145 }, { "epoch": 15.86, "learning_rate": 1.1494366197183099e-05, "loss": 0.5792, "step": 14146 }, { "epoch": 15.86, "learning_rate": 1.1492957746478873e-05, "loss": 0.4768, "step": 14147 }, { "epoch": 15.86, "learning_rate": 1.1491549295774648e-05, "loss": 0.4633, "step": 14148 }, { "epoch": 15.86, "learning_rate": 1.1490140845070422e-05, "loss": 0.4762, "step": 14149 }, { "epoch": 15.86, "learning_rate": 1.1488732394366199e-05, "loss": 0.5287, "step": 14150 }, { "epoch": 15.86, "learning_rate": 1.1487323943661972e-05, "loss": 0.5236, "step": 14151 }, { "epoch": 15.87, "learning_rate": 1.1485915492957746e-05, "loss": 0.4653, "step": 14152 }, { "epoch": 15.87, "learning_rate": 1.1484507042253521e-05, "loss": 0.5622, "step": 14153 }, { "epoch": 15.87, "learning_rate": 1.1483098591549296e-05, "loss": 0.4786, "step": 14154 }, { "epoch": 15.87, "learning_rate": 1.148169014084507e-05, "loss": 0.517, "step": 14155 }, { "epoch": 15.87, "learning_rate": 1.1480281690140845e-05, "loss": 0.4108, "step": 14156 }, { "epoch": 15.87, "learning_rate": 1.1478873239436621e-05, "loss": 0.5295, "step": 14157 }, { "epoch": 15.87, "learning_rate": 1.1477464788732396e-05, "loss": 0.4825, "step": 14158 }, { "epoch": 15.87, "learning_rate": 1.1476056338028169e-05, "loss": 0.5019, "step": 14159 }, { "epoch": 15.87, "learning_rate": 1.1474647887323943e-05, "loss": 0.6104, "step": 14160 }, { "epoch": 15.88, "learning_rate": 1.1473239436619718e-05, "loss": 0.5179, "step": 14161 }, { "epoch": 15.88, "learning_rate": 1.1471830985915493e-05, "loss": 0.4589, "step": 14162 }, { "epoch": 15.88, "learning_rate": 1.1470422535211267e-05, "loss": 0.4605, "step": 14163 }, { "epoch": 15.88, "learning_rate": 1.1469014084507042e-05, "loss": 0.682, "step": 14164 }, { "epoch": 15.88, "learning_rate": 1.1467605633802818e-05, "loss": 0.646, "step": 14165 }, { "epoch": 15.88, "learning_rate": 1.1466197183098593e-05, "loss": 0.4005, "step": 14166 }, { "epoch": 15.88, "learning_rate": 1.1464788732394366e-05, "loss": 0.4323, "step": 14167 }, { "epoch": 15.88, "learning_rate": 1.146338028169014e-05, "loss": 0.5202, "step": 14168 }, { "epoch": 15.88, "learning_rate": 1.1461971830985915e-05, "loss": 0.478, "step": 14169 }, { "epoch": 15.89, "learning_rate": 1.146056338028169e-05, "loss": 0.5206, "step": 14170 }, { "epoch": 15.89, "learning_rate": 1.1459154929577464e-05, "loss": 0.4382, "step": 14171 }, { "epoch": 15.89, "learning_rate": 1.1457746478873239e-05, "loss": 0.4871, "step": 14172 }, { "epoch": 15.89, "learning_rate": 1.1456338028169015e-05, "loss": 0.4056, "step": 14173 }, { "epoch": 15.89, "learning_rate": 1.145492957746479e-05, "loss": 0.5327, "step": 14174 }, { "epoch": 15.89, "learning_rate": 1.1453521126760563e-05, "loss": 0.4948, "step": 14175 }, { "epoch": 15.89, "learning_rate": 1.1452112676056338e-05, "loss": 0.5246, "step": 14176 }, { "epoch": 15.89, "learning_rate": 1.1450704225352112e-05, "loss": 0.4834, "step": 14177 }, { "epoch": 15.89, "learning_rate": 1.1449295774647887e-05, "loss": 0.5273, "step": 14178 }, { "epoch": 15.9, "learning_rate": 1.1447887323943662e-05, "loss": 0.3776, "step": 14179 }, { "epoch": 15.9, "learning_rate": 1.1446478873239438e-05, "loss": 0.3491, "step": 14180 }, { "epoch": 15.9, "learning_rate": 1.1445070422535213e-05, "loss": 0.6523, "step": 14181 }, { "epoch": 15.9, "learning_rate": 1.1443661971830987e-05, "loss": 0.6396, "step": 14182 }, { "epoch": 15.9, "learning_rate": 1.144225352112676e-05, "loss": 0.5973, "step": 14183 }, { "epoch": 15.9, "learning_rate": 1.1440845070422535e-05, "loss": 0.5425, "step": 14184 }, { "epoch": 15.9, "learning_rate": 1.143943661971831e-05, "loss": 0.4976, "step": 14185 }, { "epoch": 15.9, "learning_rate": 1.1438028169014084e-05, "loss": 0.5078, "step": 14186 }, { "epoch": 15.9, "learning_rate": 1.1436619718309859e-05, "loss": 0.4529, "step": 14187 }, { "epoch": 15.91, "learning_rate": 1.1435211267605635e-05, "loss": 0.5719, "step": 14188 }, { "epoch": 15.91, "learning_rate": 1.143380281690141e-05, "loss": 0.5361, "step": 14189 }, { "epoch": 15.91, "learning_rate": 1.1432394366197184e-05, "loss": 0.526, "step": 14190 }, { "epoch": 15.91, "learning_rate": 1.1430985915492957e-05, "loss": 0.4536, "step": 14191 }, { "epoch": 15.91, "learning_rate": 1.1429577464788732e-05, "loss": 0.5178, "step": 14192 }, { "epoch": 15.91, "learning_rate": 1.1428169014084507e-05, "loss": 0.5544, "step": 14193 }, { "epoch": 15.91, "learning_rate": 1.1426760563380281e-05, "loss": 0.4448, "step": 14194 }, { "epoch": 15.91, "learning_rate": 1.1425352112676057e-05, "loss": 0.5397, "step": 14195 }, { "epoch": 15.91, "learning_rate": 1.1423943661971832e-05, "loss": 0.5587, "step": 14196 }, { "epoch": 15.92, "learning_rate": 1.1422535211267607e-05, "loss": 0.6274, "step": 14197 }, { "epoch": 15.92, "learning_rate": 1.1421126760563381e-05, "loss": 0.57, "step": 14198 }, { "epoch": 15.92, "learning_rate": 1.1419718309859154e-05, "loss": 0.4841, "step": 14199 }, { "epoch": 15.92, "learning_rate": 1.1418309859154929e-05, "loss": 0.5368, "step": 14200 }, { "epoch": 15.92, "learning_rate": 1.1416901408450704e-05, "loss": 0.4141, "step": 14201 }, { "epoch": 15.92, "learning_rate": 1.1415492957746478e-05, "loss": 0.6146, "step": 14202 }, { "epoch": 15.92, "learning_rate": 1.1414084507042255e-05, "loss": 0.4952, "step": 14203 }, { "epoch": 15.92, "learning_rate": 1.141267605633803e-05, "loss": 0.5021, "step": 14204 }, { "epoch": 15.92, "learning_rate": 1.1411267605633804e-05, "loss": 0.5504, "step": 14205 }, { "epoch": 15.93, "learning_rate": 1.1409859154929579e-05, "loss": 0.6032, "step": 14206 }, { "epoch": 15.93, "learning_rate": 1.1408450704225351e-05, "loss": 0.428, "step": 14207 }, { "epoch": 15.93, "learning_rate": 1.1407042253521126e-05, "loss": 0.4822, "step": 14208 }, { "epoch": 15.93, "learning_rate": 1.14056338028169e-05, "loss": 0.466, "step": 14209 }, { "epoch": 15.93, "learning_rate": 1.1404225352112677e-05, "loss": 0.5007, "step": 14210 }, { "epoch": 15.93, "learning_rate": 1.1402816901408452e-05, "loss": 0.4407, "step": 14211 }, { "epoch": 15.93, "learning_rate": 1.1401408450704226e-05, "loss": 0.5075, "step": 14212 }, { "epoch": 15.93, "learning_rate": 1.1400000000000001e-05, "loss": 0.5312, "step": 14213 }, { "epoch": 15.93, "learning_rate": 1.1398591549295776e-05, "loss": 0.6586, "step": 14214 }, { "epoch": 15.94, "learning_rate": 1.1397183098591549e-05, "loss": 0.4528, "step": 14215 }, { "epoch": 15.94, "learning_rate": 1.1395774647887323e-05, "loss": 0.539, "step": 14216 }, { "epoch": 15.94, "learning_rate": 1.1394366197183098e-05, "loss": 0.4818, "step": 14217 }, { "epoch": 15.94, "learning_rate": 1.1392957746478874e-05, "loss": 0.4567, "step": 14218 }, { "epoch": 15.94, "learning_rate": 1.1391549295774649e-05, "loss": 0.3825, "step": 14219 }, { "epoch": 15.94, "learning_rate": 1.1390140845070423e-05, "loss": 0.5124, "step": 14220 }, { "epoch": 15.94, "learning_rate": 1.1388732394366198e-05, "loss": 0.3589, "step": 14221 }, { "epoch": 15.94, "learning_rate": 1.1387323943661973e-05, "loss": 0.5648, "step": 14222 }, { "epoch": 15.95, "learning_rate": 1.1385915492957746e-05, "loss": 0.4989, "step": 14223 }, { "epoch": 15.95, "learning_rate": 1.138450704225352e-05, "loss": 0.4459, "step": 14224 }, { "epoch": 15.95, "learning_rate": 1.1383098591549297e-05, "loss": 0.6117, "step": 14225 }, { "epoch": 15.95, "learning_rate": 1.1381690140845071e-05, "loss": 0.6403, "step": 14226 }, { "epoch": 15.95, "learning_rate": 1.1380281690140846e-05, "loss": 0.3719, "step": 14227 }, { "epoch": 15.95, "learning_rate": 1.137887323943662e-05, "loss": 0.3118, "step": 14228 }, { "epoch": 15.95, "learning_rate": 1.1377464788732395e-05, "loss": 0.4527, "step": 14229 }, { "epoch": 15.95, "learning_rate": 1.137605633802817e-05, "loss": 0.4721, "step": 14230 }, { "epoch": 15.95, "learning_rate": 1.1374647887323943e-05, "loss": 0.7953, "step": 14231 }, { "epoch": 15.96, "learning_rate": 1.1373239436619717e-05, "loss": 0.6505, "step": 14232 }, { "epoch": 15.96, "learning_rate": 1.1371830985915494e-05, "loss": 0.5868, "step": 14233 }, { "epoch": 15.96, "learning_rate": 1.1370422535211268e-05, "loss": 0.6101, "step": 14234 }, { "epoch": 15.96, "learning_rate": 1.1369014084507043e-05, "loss": 0.3941, "step": 14235 }, { "epoch": 15.96, "learning_rate": 1.1367605633802818e-05, "loss": 0.5578, "step": 14236 }, { "epoch": 15.96, "learning_rate": 1.1366197183098592e-05, "loss": 0.5722, "step": 14237 }, { "epoch": 15.96, "learning_rate": 1.1364788732394367e-05, "loss": 0.5953, "step": 14238 }, { "epoch": 15.96, "learning_rate": 1.136338028169014e-05, "loss": 0.497, "step": 14239 }, { "epoch": 15.96, "learning_rate": 1.1361971830985914e-05, "loss": 0.4759, "step": 14240 }, { "epoch": 15.97, "learning_rate": 1.136056338028169e-05, "loss": 0.6247, "step": 14241 }, { "epoch": 15.97, "learning_rate": 1.1359154929577465e-05, "loss": 0.5197, "step": 14242 }, { "epoch": 15.97, "learning_rate": 1.135774647887324e-05, "loss": 0.5628, "step": 14243 }, { "epoch": 15.97, "learning_rate": 1.1356338028169015e-05, "loss": 0.6478, "step": 14244 }, { "epoch": 15.97, "learning_rate": 1.135492957746479e-05, "loss": 0.426, "step": 14245 }, { "epoch": 15.97, "learning_rate": 1.1353521126760564e-05, "loss": 0.5009, "step": 14246 }, { "epoch": 15.97, "learning_rate": 1.1352112676056337e-05, "loss": 0.5267, "step": 14247 }, { "epoch": 15.97, "learning_rate": 1.1350704225352113e-05, "loss": 0.5055, "step": 14248 }, { "epoch": 15.97, "learning_rate": 1.1349295774647888e-05, "loss": 0.4777, "step": 14249 }, { "epoch": 15.98, "learning_rate": 1.1347887323943663e-05, "loss": 0.5645, "step": 14250 }, { "epoch": 15.98, "learning_rate": 1.1346478873239437e-05, "loss": 0.5485, "step": 14251 }, { "epoch": 15.98, "learning_rate": 1.1345070422535212e-05, "loss": 0.4942, "step": 14252 }, { "epoch": 15.98, "learning_rate": 1.1343661971830986e-05, "loss": 0.4685, "step": 14253 }, { "epoch": 15.98, "learning_rate": 1.1342253521126761e-05, "loss": 0.509, "step": 14254 }, { "epoch": 15.98, "learning_rate": 1.1340845070422534e-05, "loss": 0.4534, "step": 14255 }, { "epoch": 15.98, "learning_rate": 1.133943661971831e-05, "loss": 0.4171, "step": 14256 }, { "epoch": 15.98, "learning_rate": 1.1338028169014085e-05, "loss": 0.4467, "step": 14257 }, { "epoch": 15.98, "learning_rate": 1.133661971830986e-05, "loss": 0.6287, "step": 14258 }, { "epoch": 15.99, "learning_rate": 1.1335211267605634e-05, "loss": 0.5652, "step": 14259 }, { "epoch": 15.99, "learning_rate": 1.1333802816901409e-05, "loss": 0.5646, "step": 14260 }, { "epoch": 15.99, "learning_rate": 1.1332394366197184e-05, "loss": 0.4418, "step": 14261 }, { "epoch": 15.99, "learning_rate": 1.1330985915492958e-05, "loss": 0.4967, "step": 14262 }, { "epoch": 15.99, "learning_rate": 1.1329577464788733e-05, "loss": 0.5273, "step": 14263 }, { "epoch": 15.99, "learning_rate": 1.1328169014084507e-05, "loss": 0.4257, "step": 14264 }, { "epoch": 15.99, "learning_rate": 1.1326760563380282e-05, "loss": 0.5795, "step": 14265 }, { "epoch": 15.99, "learning_rate": 1.1325352112676057e-05, "loss": 0.4405, "step": 14266 }, { "epoch": 15.99, "learning_rate": 1.1323943661971831e-05, "loss": 0.4712, "step": 14267 }, { "epoch": 16.0, "learning_rate": 1.1322535211267606e-05, "loss": 0.5034, "step": 14268 }, { "epoch": 16.0, "learning_rate": 1.132112676056338e-05, "loss": 0.5624, "step": 14269 }, { "epoch": 16.0, "learning_rate": 1.1319718309859154e-05, "loss": 0.4295, "step": 14270 }, { "epoch": 16.0, "learning_rate": 1.131830985915493e-05, "loss": 0.4922, "step": 14271 }, { "epoch": 16.0, "learning_rate": 1.1316901408450705e-05, "loss": 0.5036, "step": 14272 }, { "epoch": 16.0, "learning_rate": 1.131549295774648e-05, "loss": 0.6492, "step": 14273 }, { "epoch": 16.0, "learning_rate": 1.1314084507042254e-05, "loss": 0.5322, "step": 14274 }, { "epoch": 16.0, "learning_rate": 1.1312676056338028e-05, "loss": 0.621, "step": 14275 }, { "epoch": 16.0, "learning_rate": 1.1311267605633803e-05, "loss": 0.5274, "step": 14276 }, { "epoch": 16.01, "learning_rate": 1.1309859154929578e-05, "loss": 0.584, "step": 14277 }, { "epoch": 16.01, "learning_rate": 1.1308450704225352e-05, "loss": 0.6133, "step": 14278 }, { "epoch": 16.01, "learning_rate": 1.1307042253521127e-05, "loss": 0.4737, "step": 14279 }, { "epoch": 16.01, "learning_rate": 1.1305633802816902e-05, "loss": 0.5319, "step": 14280 }, { "epoch": 16.01, "learning_rate": 1.1304225352112676e-05, "loss": 0.4909, "step": 14281 }, { "epoch": 16.01, "learning_rate": 1.1302816901408451e-05, "loss": 0.4993, "step": 14282 }, { "epoch": 16.01, "learning_rate": 1.1301408450704226e-05, "loss": 0.5168, "step": 14283 }, { "epoch": 16.01, "learning_rate": 1.13e-05, "loss": 0.5513, "step": 14284 }, { "epoch": 16.01, "learning_rate": 1.1298591549295775e-05, "loss": 0.5137, "step": 14285 }, { "epoch": 16.02, "learning_rate": 1.129718309859155e-05, "loss": 0.5895, "step": 14286 }, { "epoch": 16.02, "learning_rate": 1.1295774647887324e-05, "loss": 0.4891, "step": 14287 }, { "epoch": 16.02, "learning_rate": 1.1294366197183099e-05, "loss": 0.4695, "step": 14288 }, { "epoch": 16.02, "learning_rate": 1.1292957746478873e-05, "loss": 0.4216, "step": 14289 }, { "epoch": 16.02, "learning_rate": 1.1291549295774648e-05, "loss": 0.4796, "step": 14290 }, { "epoch": 16.02, "learning_rate": 1.1290140845070423e-05, "loss": 0.427, "step": 14291 }, { "epoch": 16.02, "learning_rate": 1.1288732394366197e-05, "loss": 0.5515, "step": 14292 }, { "epoch": 16.02, "learning_rate": 1.1287323943661972e-05, "loss": 0.514, "step": 14293 }, { "epoch": 16.02, "learning_rate": 1.1285915492957747e-05, "loss": 0.4105, "step": 14294 }, { "epoch": 16.03, "learning_rate": 1.1284507042253521e-05, "loss": 0.4742, "step": 14295 }, { "epoch": 16.03, "learning_rate": 1.1283098591549296e-05, "loss": 0.543, "step": 14296 }, { "epoch": 16.03, "learning_rate": 1.128169014084507e-05, "loss": 0.4748, "step": 14297 }, { "epoch": 16.03, "learning_rate": 1.1280281690140845e-05, "loss": 0.5085, "step": 14298 }, { "epoch": 16.03, "learning_rate": 1.127887323943662e-05, "loss": 0.4798, "step": 14299 }, { "epoch": 16.03, "learning_rate": 1.1277464788732394e-05, "loss": 0.5239, "step": 14300 }, { "epoch": 16.03, "learning_rate": 1.127605633802817e-05, "loss": 0.3968, "step": 14301 }, { "epoch": 16.03, "learning_rate": 1.1274647887323944e-05, "loss": 0.501, "step": 14302 }, { "epoch": 16.03, "learning_rate": 1.1273239436619718e-05, "loss": 0.5124, "step": 14303 }, { "epoch": 16.04, "learning_rate": 1.1271830985915493e-05, "loss": 0.4688, "step": 14304 }, { "epoch": 16.04, "learning_rate": 1.1270422535211268e-05, "loss": 0.3742, "step": 14305 }, { "epoch": 16.04, "learning_rate": 1.1269014084507042e-05, "loss": 0.5145, "step": 14306 }, { "epoch": 16.04, "learning_rate": 1.1267605633802817e-05, "loss": 0.3711, "step": 14307 }, { "epoch": 16.04, "learning_rate": 1.1266197183098592e-05, "loss": 0.5463, "step": 14308 }, { "epoch": 16.04, "learning_rate": 1.1264788732394368e-05, "loss": 0.4488, "step": 14309 }, { "epoch": 16.04, "learning_rate": 1.126338028169014e-05, "loss": 0.4128, "step": 14310 }, { "epoch": 16.04, "learning_rate": 1.1261971830985915e-05, "loss": 0.3929, "step": 14311 }, { "epoch": 16.04, "learning_rate": 1.126056338028169e-05, "loss": 0.4234, "step": 14312 }, { "epoch": 16.05, "learning_rate": 1.1259154929577465e-05, "loss": 0.6097, "step": 14313 }, { "epoch": 16.05, "learning_rate": 1.125774647887324e-05, "loss": 0.5144, "step": 14314 }, { "epoch": 16.05, "learning_rate": 1.1256338028169014e-05, "loss": 0.4721, "step": 14315 }, { "epoch": 16.05, "learning_rate": 1.125492957746479e-05, "loss": 0.5174, "step": 14316 }, { "epoch": 16.05, "learning_rate": 1.1253521126760565e-05, "loss": 0.3656, "step": 14317 }, { "epoch": 16.05, "learning_rate": 1.1252112676056338e-05, "loss": 0.3426, "step": 14318 }, { "epoch": 16.05, "learning_rate": 1.1250704225352113e-05, "loss": 0.3933, "step": 14319 }, { "epoch": 16.05, "learning_rate": 1.1249295774647887e-05, "loss": 0.3494, "step": 14320 }, { "epoch": 16.05, "learning_rate": 1.1247887323943662e-05, "loss": 0.3707, "step": 14321 }, { "epoch": 16.06, "learning_rate": 1.1246478873239436e-05, "loss": 0.4613, "step": 14322 }, { "epoch": 16.06, "learning_rate": 1.1245070422535211e-05, "loss": 0.5813, "step": 14323 }, { "epoch": 16.06, "learning_rate": 1.1243661971830987e-05, "loss": 0.6081, "step": 14324 }, { "epoch": 16.06, "learning_rate": 1.1242253521126762e-05, "loss": 0.4975, "step": 14325 }, { "epoch": 16.06, "learning_rate": 1.1240845070422535e-05, "loss": 0.7224, "step": 14326 }, { "epoch": 16.06, "learning_rate": 1.123943661971831e-05, "loss": 0.5977, "step": 14327 }, { "epoch": 16.06, "learning_rate": 1.1238028169014084e-05, "loss": 0.4993, "step": 14328 }, { "epoch": 16.06, "learning_rate": 1.1236619718309859e-05, "loss": 0.4653, "step": 14329 }, { "epoch": 16.07, "learning_rate": 1.1235211267605634e-05, "loss": 0.675, "step": 14330 }, { "epoch": 16.07, "learning_rate": 1.123380281690141e-05, "loss": 0.5648, "step": 14331 }, { "epoch": 16.07, "learning_rate": 1.1232394366197185e-05, "loss": 0.3737, "step": 14332 }, { "epoch": 16.07, "learning_rate": 1.123098591549296e-05, "loss": 0.4823, "step": 14333 }, { "epoch": 16.07, "learning_rate": 1.1229577464788732e-05, "loss": 0.6451, "step": 14334 }, { "epoch": 16.07, "learning_rate": 1.1228169014084507e-05, "loss": 0.5477, "step": 14335 }, { "epoch": 16.07, "learning_rate": 1.1226760563380281e-05, "loss": 0.434, "step": 14336 }, { "epoch": 16.07, "learning_rate": 1.1225352112676056e-05, "loss": 0.51, "step": 14337 }, { "epoch": 16.07, "learning_rate": 1.122394366197183e-05, "loss": 0.435, "step": 14338 }, { "epoch": 16.08, "learning_rate": 1.1222535211267607e-05, "loss": 0.4277, "step": 14339 }, { "epoch": 16.08, "learning_rate": 1.1221126760563382e-05, "loss": 0.4763, "step": 14340 }, { "epoch": 16.08, "learning_rate": 1.1219718309859156e-05, "loss": 0.5402, "step": 14341 }, { "epoch": 16.08, "learning_rate": 1.121830985915493e-05, "loss": 0.5147, "step": 14342 }, { "epoch": 16.08, "learning_rate": 1.1216901408450704e-05, "loss": 0.4596, "step": 14343 }, { "epoch": 16.08, "learning_rate": 1.1215492957746478e-05, "loss": 0.4797, "step": 14344 }, { "epoch": 16.08, "learning_rate": 1.1214084507042253e-05, "loss": 0.4153, "step": 14345 }, { "epoch": 16.08, "learning_rate": 1.121267605633803e-05, "loss": 0.5174, "step": 14346 }, { "epoch": 16.08, "learning_rate": 1.1211267605633804e-05, "loss": 0.5296, "step": 14347 }, { "epoch": 16.09, "learning_rate": 1.1209859154929579e-05, "loss": 0.5312, "step": 14348 }, { "epoch": 16.09, "learning_rate": 1.1208450704225353e-05, "loss": 0.4929, "step": 14349 }, { "epoch": 16.09, "learning_rate": 1.1207042253521126e-05, "loss": 0.4568, "step": 14350 }, { "epoch": 16.09, "learning_rate": 1.1205633802816901e-05, "loss": 0.4259, "step": 14351 }, { "epoch": 16.09, "learning_rate": 1.1204225352112676e-05, "loss": 0.4054, "step": 14352 }, { "epoch": 16.09, "learning_rate": 1.120281690140845e-05, "loss": 0.4789, "step": 14353 }, { "epoch": 16.09, "learning_rate": 1.1201408450704227e-05, "loss": 0.4432, "step": 14354 }, { "epoch": 16.09, "learning_rate": 1.1200000000000001e-05, "loss": 0.4751, "step": 14355 }, { "epoch": 16.09, "learning_rate": 1.1198591549295776e-05, "loss": 0.4188, "step": 14356 }, { "epoch": 16.1, "learning_rate": 1.119718309859155e-05, "loss": 0.5178, "step": 14357 }, { "epoch": 16.1, "learning_rate": 1.1195774647887323e-05, "loss": 0.5044, "step": 14358 }, { "epoch": 16.1, "learning_rate": 1.1194366197183098e-05, "loss": 0.4321, "step": 14359 }, { "epoch": 16.1, "learning_rate": 1.1192957746478873e-05, "loss": 0.4876, "step": 14360 }, { "epoch": 16.1, "learning_rate": 1.1191549295774647e-05, "loss": 0.4796, "step": 14361 }, { "epoch": 16.1, "learning_rate": 1.1190140845070424e-05, "loss": 0.4889, "step": 14362 }, { "epoch": 16.1, "learning_rate": 1.1188732394366198e-05, "loss": 0.4754, "step": 14363 }, { "epoch": 16.1, "learning_rate": 1.1187323943661973e-05, "loss": 0.3968, "step": 14364 }, { "epoch": 16.1, "learning_rate": 1.1185915492957748e-05, "loss": 0.5526, "step": 14365 }, { "epoch": 16.11, "learning_rate": 1.118450704225352e-05, "loss": 0.4668, "step": 14366 }, { "epoch": 16.11, "learning_rate": 1.1183098591549295e-05, "loss": 0.4574, "step": 14367 }, { "epoch": 16.11, "learning_rate": 1.118169014084507e-05, "loss": 0.4118, "step": 14368 }, { "epoch": 16.11, "learning_rate": 1.1180281690140846e-05, "loss": 0.3663, "step": 14369 }, { "epoch": 16.11, "learning_rate": 1.117887323943662e-05, "loss": 0.3649, "step": 14370 }, { "epoch": 16.11, "learning_rate": 1.1177464788732395e-05, "loss": 0.5176, "step": 14371 }, { "epoch": 16.11, "learning_rate": 1.117605633802817e-05, "loss": 0.3838, "step": 14372 }, { "epoch": 16.11, "learning_rate": 1.1174647887323945e-05, "loss": 0.7134, "step": 14373 }, { "epoch": 16.11, "learning_rate": 1.1173239436619718e-05, "loss": 0.5606, "step": 14374 }, { "epoch": 16.12, "learning_rate": 1.1171830985915492e-05, "loss": 0.5468, "step": 14375 }, { "epoch": 16.12, "learning_rate": 1.1170422535211267e-05, "loss": 0.5305, "step": 14376 }, { "epoch": 16.12, "learning_rate": 1.1169014084507043e-05, "loss": 0.3623, "step": 14377 }, { "epoch": 16.12, "learning_rate": 1.1167605633802818e-05, "loss": 0.4967, "step": 14378 }, { "epoch": 16.12, "learning_rate": 1.1166197183098592e-05, "loss": 0.5426, "step": 14379 }, { "epoch": 16.12, "learning_rate": 1.1164788732394367e-05, "loss": 0.4577, "step": 14380 }, { "epoch": 16.12, "learning_rate": 1.116338028169014e-05, "loss": 0.5177, "step": 14381 }, { "epoch": 16.12, "learning_rate": 1.1161971830985915e-05, "loss": 0.5339, "step": 14382 }, { "epoch": 16.12, "learning_rate": 1.116056338028169e-05, "loss": 0.4933, "step": 14383 }, { "epoch": 16.13, "learning_rate": 1.1159154929577466e-05, "loss": 0.5352, "step": 14384 }, { "epoch": 16.13, "learning_rate": 1.115774647887324e-05, "loss": 0.4808, "step": 14385 }, { "epoch": 16.13, "learning_rate": 1.1156338028169015e-05, "loss": 0.4803, "step": 14386 }, { "epoch": 16.13, "learning_rate": 1.115492957746479e-05, "loss": 0.5039, "step": 14387 }, { "epoch": 16.13, "learning_rate": 1.1153521126760564e-05, "loss": 0.4435, "step": 14388 }, { "epoch": 16.13, "learning_rate": 1.1152112676056337e-05, "loss": 0.5896, "step": 14389 }, { "epoch": 16.13, "learning_rate": 1.1150704225352112e-05, "loss": 0.5362, "step": 14390 }, { "epoch": 16.13, "learning_rate": 1.1149295774647886e-05, "loss": 0.5396, "step": 14391 }, { "epoch": 16.13, "learning_rate": 1.1147887323943663e-05, "loss": 0.4628, "step": 14392 }, { "epoch": 16.14, "learning_rate": 1.1146478873239437e-05, "loss": 0.4936, "step": 14393 }, { "epoch": 16.14, "learning_rate": 1.1145070422535212e-05, "loss": 0.4448, "step": 14394 }, { "epoch": 16.14, "learning_rate": 1.1143661971830987e-05, "loss": 0.43, "step": 14395 }, { "epoch": 16.14, "learning_rate": 1.1142253521126761e-05, "loss": 0.4517, "step": 14396 }, { "epoch": 16.14, "learning_rate": 1.1140845070422534e-05, "loss": 0.473, "step": 14397 }, { "epoch": 16.14, "learning_rate": 1.1139436619718309e-05, "loss": 0.4315, "step": 14398 }, { "epoch": 16.14, "learning_rate": 1.1138028169014085e-05, "loss": 0.4197, "step": 14399 }, { "epoch": 16.14, "learning_rate": 1.113661971830986e-05, "loss": 0.4169, "step": 14400 }, { "epoch": 16.14, "learning_rate": 1.1135211267605635e-05, "loss": 0.4436, "step": 14401 }, { "epoch": 16.15, "learning_rate": 1.1133802816901409e-05, "loss": 0.5354, "step": 14402 }, { "epoch": 16.15, "learning_rate": 1.1132394366197184e-05, "loss": 0.4251, "step": 14403 }, { "epoch": 16.15, "learning_rate": 1.1130985915492958e-05, "loss": 0.3178, "step": 14404 }, { "epoch": 16.15, "learning_rate": 1.1129577464788731e-05, "loss": 0.4532, "step": 14405 }, { "epoch": 16.15, "learning_rate": 1.1128169014084506e-05, "loss": 0.4554, "step": 14406 }, { "epoch": 16.15, "learning_rate": 1.1126760563380282e-05, "loss": 0.5085, "step": 14407 }, { "epoch": 16.15, "learning_rate": 1.1125352112676057e-05, "loss": 0.43, "step": 14408 }, { "epoch": 16.15, "learning_rate": 1.1123943661971832e-05, "loss": 0.4574, "step": 14409 }, { "epoch": 16.15, "learning_rate": 1.1122535211267606e-05, "loss": 0.3195, "step": 14410 }, { "epoch": 16.16, "learning_rate": 1.1121126760563381e-05, "loss": 0.4763, "step": 14411 }, { "epoch": 16.16, "learning_rate": 1.1119718309859156e-05, "loss": 0.4389, "step": 14412 }, { "epoch": 16.16, "learning_rate": 1.1118309859154928e-05, "loss": 0.4819, "step": 14413 }, { "epoch": 16.16, "learning_rate": 1.1116901408450703e-05, "loss": 0.4266, "step": 14414 }, { "epoch": 16.16, "learning_rate": 1.111549295774648e-05, "loss": 0.4635, "step": 14415 }, { "epoch": 16.16, "learning_rate": 1.1114084507042254e-05, "loss": 0.3959, "step": 14416 }, { "epoch": 16.16, "learning_rate": 1.1112676056338029e-05, "loss": 0.4803, "step": 14417 }, { "epoch": 16.16, "learning_rate": 1.1111267605633803e-05, "loss": 0.3851, "step": 14418 }, { "epoch": 16.16, "learning_rate": 1.1109859154929578e-05, "loss": 0.378, "step": 14419 }, { "epoch": 16.17, "learning_rate": 1.1108450704225353e-05, "loss": 0.4497, "step": 14420 }, { "epoch": 16.17, "learning_rate": 1.1107042253521126e-05, "loss": 0.267, "step": 14421 }, { "epoch": 16.17, "learning_rate": 1.1105633802816902e-05, "loss": 0.5322, "step": 14422 }, { "epoch": 16.17, "learning_rate": 1.1104225352112677e-05, "loss": 0.5439, "step": 14423 }, { "epoch": 16.17, "learning_rate": 1.1102816901408451e-05, "loss": 0.4852, "step": 14424 }, { "epoch": 16.17, "learning_rate": 1.1101408450704226e-05, "loss": 0.5367, "step": 14425 }, { "epoch": 16.17, "learning_rate": 1.11e-05, "loss": 0.6291, "step": 14426 }, { "epoch": 16.17, "learning_rate": 1.1098591549295775e-05, "loss": 0.4567, "step": 14427 }, { "epoch": 16.17, "learning_rate": 1.109718309859155e-05, "loss": 0.4975, "step": 14428 }, { "epoch": 16.18, "learning_rate": 1.1095774647887323e-05, "loss": 0.4977, "step": 14429 }, { "epoch": 16.18, "learning_rate": 1.1094366197183099e-05, "loss": 0.5618, "step": 14430 }, { "epoch": 16.18, "learning_rate": 1.1092957746478874e-05, "loss": 0.5159, "step": 14431 }, { "epoch": 16.18, "learning_rate": 1.1091549295774648e-05, "loss": 0.5779, "step": 14432 }, { "epoch": 16.18, "learning_rate": 1.1090140845070423e-05, "loss": 0.4594, "step": 14433 }, { "epoch": 16.18, "learning_rate": 1.1088732394366198e-05, "loss": 0.484, "step": 14434 }, { "epoch": 16.18, "learning_rate": 1.1087323943661972e-05, "loss": 0.5846, "step": 14435 }, { "epoch": 16.18, "learning_rate": 1.1085915492957747e-05, "loss": 0.4682, "step": 14436 }, { "epoch": 16.18, "learning_rate": 1.1084507042253521e-05, "loss": 0.3987, "step": 14437 }, { "epoch": 16.19, "learning_rate": 1.1083098591549296e-05, "loss": 0.5232, "step": 14438 }, { "epoch": 16.19, "learning_rate": 1.108169014084507e-05, "loss": 0.6079, "step": 14439 }, { "epoch": 16.19, "learning_rate": 1.1080281690140845e-05, "loss": 0.4786, "step": 14440 }, { "epoch": 16.19, "learning_rate": 1.107887323943662e-05, "loss": 0.6005, "step": 14441 }, { "epoch": 16.19, "learning_rate": 1.1077464788732395e-05, "loss": 0.5134, "step": 14442 }, { "epoch": 16.19, "learning_rate": 1.107605633802817e-05, "loss": 0.5026, "step": 14443 }, { "epoch": 16.19, "learning_rate": 1.1074647887323944e-05, "loss": 0.4647, "step": 14444 }, { "epoch": 16.19, "learning_rate": 1.1073239436619719e-05, "loss": 0.4553, "step": 14445 }, { "epoch": 16.2, "learning_rate": 1.1071830985915493e-05, "loss": 0.3856, "step": 14446 }, { "epoch": 16.2, "learning_rate": 1.1070422535211268e-05, "loss": 0.6252, "step": 14447 }, { "epoch": 16.2, "learning_rate": 1.1069014084507042e-05, "loss": 0.3731, "step": 14448 }, { "epoch": 16.2, "learning_rate": 1.1067605633802817e-05, "loss": 0.4227, "step": 14449 }, { "epoch": 16.2, "learning_rate": 1.1066197183098592e-05, "loss": 0.4507, "step": 14450 }, { "epoch": 16.2, "learning_rate": 1.1064788732394366e-05, "loss": 0.4448, "step": 14451 }, { "epoch": 16.2, "learning_rate": 1.1063380281690143e-05, "loss": 0.4681, "step": 14452 }, { "epoch": 16.2, "learning_rate": 1.1061971830985916e-05, "loss": 0.4656, "step": 14453 }, { "epoch": 16.2, "learning_rate": 1.106056338028169e-05, "loss": 0.3815, "step": 14454 }, { "epoch": 16.21, "learning_rate": 1.1059154929577465e-05, "loss": 0.3991, "step": 14455 }, { "epoch": 16.21, "learning_rate": 1.105774647887324e-05, "loss": 0.4455, "step": 14456 }, { "epoch": 16.21, "learning_rate": 1.1056338028169014e-05, "loss": 0.4164, "step": 14457 }, { "epoch": 16.21, "learning_rate": 1.1054929577464789e-05, "loss": 0.3937, "step": 14458 }, { "epoch": 16.21, "learning_rate": 1.1053521126760563e-05, "loss": 0.4684, "step": 14459 }, { "epoch": 16.21, "learning_rate": 1.105211267605634e-05, "loss": 0.3622, "step": 14460 }, { "epoch": 16.21, "learning_rate": 1.1050704225352113e-05, "loss": 0.4766, "step": 14461 }, { "epoch": 16.21, "learning_rate": 1.1049295774647887e-05, "loss": 0.4668, "step": 14462 }, { "epoch": 16.21, "learning_rate": 1.1047887323943662e-05, "loss": 0.4971, "step": 14463 }, { "epoch": 16.22, "learning_rate": 1.1046478873239437e-05, "loss": 0.5336, "step": 14464 }, { "epoch": 16.22, "learning_rate": 1.1045070422535211e-05, "loss": 0.5632, "step": 14465 }, { "epoch": 16.22, "learning_rate": 1.1043661971830986e-05, "loss": 0.398, "step": 14466 }, { "epoch": 16.22, "learning_rate": 1.1042253521126762e-05, "loss": 0.5109, "step": 14467 }, { "epoch": 16.22, "learning_rate": 1.1040845070422537e-05, "loss": 0.5262, "step": 14468 }, { "epoch": 16.22, "learning_rate": 1.103943661971831e-05, "loss": 0.3479, "step": 14469 }, { "epoch": 16.22, "learning_rate": 1.1038028169014085e-05, "loss": 0.4405, "step": 14470 }, { "epoch": 16.22, "learning_rate": 1.1036619718309859e-05, "loss": 0.3958, "step": 14471 }, { "epoch": 16.22, "learning_rate": 1.1035211267605634e-05, "loss": 0.404, "step": 14472 }, { "epoch": 16.23, "learning_rate": 1.1033802816901408e-05, "loss": 0.6821, "step": 14473 }, { "epoch": 16.23, "learning_rate": 1.1032394366197183e-05, "loss": 0.4886, "step": 14474 }, { "epoch": 16.23, "learning_rate": 1.103098591549296e-05, "loss": 0.5105, "step": 14475 }, { "epoch": 16.23, "learning_rate": 1.1029577464788734e-05, "loss": 0.4388, "step": 14476 }, { "epoch": 16.23, "learning_rate": 1.1028169014084507e-05, "loss": 0.5581, "step": 14477 }, { "epoch": 16.23, "learning_rate": 1.1026760563380282e-05, "loss": 0.542, "step": 14478 }, { "epoch": 16.23, "learning_rate": 1.1025352112676056e-05, "loss": 0.5629, "step": 14479 }, { "epoch": 16.23, "learning_rate": 1.1023943661971831e-05, "loss": 0.4133, "step": 14480 }, { "epoch": 16.23, "learning_rate": 1.1022535211267606e-05, "loss": 0.4651, "step": 14481 }, { "epoch": 16.24, "learning_rate": 1.102112676056338e-05, "loss": 0.4461, "step": 14482 }, { "epoch": 16.24, "learning_rate": 1.1019718309859156e-05, "loss": 0.4888, "step": 14483 }, { "epoch": 16.24, "learning_rate": 1.1018309859154931e-05, "loss": 0.5253, "step": 14484 }, { "epoch": 16.24, "learning_rate": 1.1016901408450704e-05, "loss": 0.5264, "step": 14485 }, { "epoch": 16.24, "learning_rate": 1.1015492957746479e-05, "loss": 0.4589, "step": 14486 }, { "epoch": 16.24, "learning_rate": 1.1014084507042253e-05, "loss": 0.6078, "step": 14487 }, { "epoch": 16.24, "learning_rate": 1.1012676056338028e-05, "loss": 0.4534, "step": 14488 }, { "epoch": 16.24, "learning_rate": 1.1011267605633803e-05, "loss": 0.4592, "step": 14489 }, { "epoch": 16.24, "learning_rate": 1.1009859154929579e-05, "loss": 0.4169, "step": 14490 }, { "epoch": 16.25, "learning_rate": 1.1008450704225354e-05, "loss": 0.5531, "step": 14491 }, { "epoch": 16.25, "learning_rate": 1.1007042253521127e-05, "loss": 0.5017, "step": 14492 }, { "epoch": 16.25, "learning_rate": 1.1005633802816901e-05, "loss": 0.5131, "step": 14493 }, { "epoch": 16.25, "learning_rate": 1.1004225352112676e-05, "loss": 0.5346, "step": 14494 }, { "epoch": 16.25, "learning_rate": 1.100281690140845e-05, "loss": 0.628, "step": 14495 }, { "epoch": 16.25, "learning_rate": 1.1001408450704225e-05, "loss": 0.591, "step": 14496 }, { "epoch": 16.25, "learning_rate": 1.1e-05, "loss": 0.451, "step": 14497 }, { "epoch": 16.25, "learning_rate": 1.0998591549295776e-05, "loss": 0.4626, "step": 14498 }, { "epoch": 16.25, "learning_rate": 1.099718309859155e-05, "loss": 0.3863, "step": 14499 }, { "epoch": 16.26, "learning_rate": 1.0995774647887324e-05, "loss": 0.4491, "step": 14500 }, { "epoch": 16.26, "learning_rate": 1.0994366197183098e-05, "loss": 0.4435, "step": 14501 }, { "epoch": 16.26, "learning_rate": 1.0992957746478873e-05, "loss": 0.5236, "step": 14502 }, { "epoch": 16.26, "learning_rate": 1.0991549295774648e-05, "loss": 0.4839, "step": 14503 }, { "epoch": 16.26, "learning_rate": 1.0990140845070422e-05, "loss": 0.4895, "step": 14504 }, { "epoch": 16.26, "learning_rate": 1.0988732394366199e-05, "loss": 0.5202, "step": 14505 }, { "epoch": 16.26, "learning_rate": 1.0987323943661973e-05, "loss": 0.4899, "step": 14506 }, { "epoch": 16.26, "learning_rate": 1.0985915492957748e-05, "loss": 0.4143, "step": 14507 }, { "epoch": 16.26, "learning_rate": 1.098450704225352e-05, "loss": 0.5382, "step": 14508 }, { "epoch": 16.27, "learning_rate": 1.0983098591549295e-05, "loss": 0.5425, "step": 14509 }, { "epoch": 16.27, "learning_rate": 1.098169014084507e-05, "loss": 0.3293, "step": 14510 }, { "epoch": 16.27, "learning_rate": 1.0980281690140845e-05, "loss": 0.47, "step": 14511 }, { "epoch": 16.27, "learning_rate": 1.097887323943662e-05, "loss": 0.566, "step": 14512 }, { "epoch": 16.27, "learning_rate": 1.0977464788732396e-05, "loss": 0.4639, "step": 14513 }, { "epoch": 16.27, "learning_rate": 1.097605633802817e-05, "loss": 0.5418, "step": 14514 }, { "epoch": 16.27, "learning_rate": 1.0974647887323945e-05, "loss": 0.5004, "step": 14515 }, { "epoch": 16.27, "learning_rate": 1.0973239436619718e-05, "loss": 0.5336, "step": 14516 }, { "epoch": 16.27, "learning_rate": 1.0971830985915492e-05, "loss": 0.4978, "step": 14517 }, { "epoch": 16.28, "learning_rate": 1.0970422535211267e-05, "loss": 0.4697, "step": 14518 }, { "epoch": 16.28, "learning_rate": 1.0969014084507042e-05, "loss": 0.3614, "step": 14519 }, { "epoch": 16.28, "learning_rate": 1.0967605633802818e-05, "loss": 0.3472, "step": 14520 }, { "epoch": 16.28, "learning_rate": 1.0966197183098593e-05, "loss": 0.5109, "step": 14521 }, { "epoch": 16.28, "learning_rate": 1.0964788732394367e-05, "loss": 0.365, "step": 14522 }, { "epoch": 16.28, "learning_rate": 1.0963380281690142e-05, "loss": 0.6484, "step": 14523 }, { "epoch": 16.28, "learning_rate": 1.0961971830985915e-05, "loss": 0.6007, "step": 14524 }, { "epoch": 16.28, "learning_rate": 1.096056338028169e-05, "loss": 0.6369, "step": 14525 }, { "epoch": 16.28, "learning_rate": 1.0959154929577464e-05, "loss": 0.5936, "step": 14526 }, { "epoch": 16.29, "learning_rate": 1.0957746478873239e-05, "loss": 0.708, "step": 14527 }, { "epoch": 16.29, "learning_rate": 1.0956338028169015e-05, "loss": 0.4752, "step": 14528 }, { "epoch": 16.29, "learning_rate": 1.095492957746479e-05, "loss": 0.5546, "step": 14529 }, { "epoch": 16.29, "learning_rate": 1.0953521126760564e-05, "loss": 0.5603, "step": 14530 }, { "epoch": 16.29, "learning_rate": 1.0952112676056339e-05, "loss": 0.5237, "step": 14531 }, { "epoch": 16.29, "learning_rate": 1.0950704225352112e-05, "loss": 0.5885, "step": 14532 }, { "epoch": 16.29, "learning_rate": 1.0949295774647887e-05, "loss": 0.5164, "step": 14533 }, { "epoch": 16.29, "learning_rate": 1.0947887323943661e-05, "loss": 0.5301, "step": 14534 }, { "epoch": 16.29, "learning_rate": 1.0946478873239436e-05, "loss": 0.4836, "step": 14535 }, { "epoch": 16.3, "learning_rate": 1.0945070422535212e-05, "loss": 0.4332, "step": 14536 }, { "epoch": 16.3, "learning_rate": 1.0943661971830987e-05, "loss": 0.5091, "step": 14537 }, { "epoch": 16.3, "learning_rate": 1.0942253521126762e-05, "loss": 0.543, "step": 14538 }, { "epoch": 16.3, "learning_rate": 1.0940845070422536e-05, "loss": 0.451, "step": 14539 }, { "epoch": 16.3, "learning_rate": 1.0939436619718309e-05, "loss": 0.4595, "step": 14540 }, { "epoch": 16.3, "learning_rate": 1.0938028169014084e-05, "loss": 0.5608, "step": 14541 }, { "epoch": 16.3, "learning_rate": 1.0936619718309858e-05, "loss": 0.511, "step": 14542 }, { "epoch": 16.3, "learning_rate": 1.0935211267605635e-05, "loss": 0.5028, "step": 14543 }, { "epoch": 16.3, "learning_rate": 1.093380281690141e-05, "loss": 0.5347, "step": 14544 }, { "epoch": 16.31, "learning_rate": 1.0932394366197184e-05, "loss": 0.5006, "step": 14545 }, { "epoch": 16.31, "learning_rate": 1.0930985915492959e-05, "loss": 0.396, "step": 14546 }, { "epoch": 16.31, "learning_rate": 1.0929577464788733e-05, "loss": 0.3888, "step": 14547 }, { "epoch": 16.31, "learning_rate": 1.0928169014084506e-05, "loss": 0.4556, "step": 14548 }, { "epoch": 16.31, "learning_rate": 1.0926760563380281e-05, "loss": 0.452, "step": 14549 }, { "epoch": 16.31, "learning_rate": 1.0925352112676056e-05, "loss": 0.5066, "step": 14550 }, { "epoch": 16.31, "learning_rate": 1.0923943661971832e-05, "loss": 0.5526, "step": 14551 }, { "epoch": 16.31, "learning_rate": 1.0922535211267606e-05, "loss": 0.5681, "step": 14552 }, { "epoch": 16.32, "learning_rate": 1.0921126760563381e-05, "loss": 0.4603, "step": 14553 }, { "epoch": 16.32, "learning_rate": 1.0919718309859156e-05, "loss": 0.5357, "step": 14554 }, { "epoch": 16.32, "learning_rate": 1.091830985915493e-05, "loss": 0.4265, "step": 14555 }, { "epoch": 16.32, "learning_rate": 1.0916901408450703e-05, "loss": 0.6008, "step": 14556 }, { "epoch": 16.32, "learning_rate": 1.0915492957746478e-05, "loss": 0.5195, "step": 14557 }, { "epoch": 16.32, "learning_rate": 1.0914084507042254e-05, "loss": 0.399, "step": 14558 }, { "epoch": 16.32, "learning_rate": 1.0912676056338029e-05, "loss": 0.4274, "step": 14559 }, { "epoch": 16.32, "learning_rate": 1.0911267605633804e-05, "loss": 0.4192, "step": 14560 }, { "epoch": 16.32, "learning_rate": 1.0909859154929578e-05, "loss": 0.2996, "step": 14561 }, { "epoch": 16.33, "learning_rate": 1.0908450704225353e-05, "loss": 0.4393, "step": 14562 }, { "epoch": 16.33, "learning_rate": 1.0907042253521127e-05, "loss": 0.3913, "step": 14563 }, { "epoch": 16.33, "learning_rate": 1.09056338028169e-05, "loss": 0.5485, "step": 14564 }, { "epoch": 16.33, "learning_rate": 1.0904225352112675e-05, "loss": 0.3972, "step": 14565 }, { "epoch": 16.33, "learning_rate": 1.0902816901408451e-05, "loss": 0.4595, "step": 14566 }, { "epoch": 16.33, "learning_rate": 1.0901408450704226e-05, "loss": 0.378, "step": 14567 }, { "epoch": 16.33, "learning_rate": 1.09e-05, "loss": 0.3894, "step": 14568 }, { "epoch": 16.33, "learning_rate": 1.0898591549295775e-05, "loss": 0.6272, "step": 14569 }, { "epoch": 16.33, "learning_rate": 1.089718309859155e-05, "loss": 0.5313, "step": 14570 }, { "epoch": 16.34, "learning_rate": 1.0895774647887325e-05, "loss": 0.488, "step": 14571 }, { "epoch": 16.34, "learning_rate": 1.0894366197183098e-05, "loss": 0.6852, "step": 14572 }, { "epoch": 16.34, "learning_rate": 1.0892957746478874e-05, "loss": 0.7101, "step": 14573 }, { "epoch": 16.34, "learning_rate": 1.0891549295774648e-05, "loss": 0.6091, "step": 14574 }, { "epoch": 16.34, "learning_rate": 1.0890140845070423e-05, "loss": 0.5588, "step": 14575 }, { "epoch": 16.34, "learning_rate": 1.0888732394366198e-05, "loss": 0.5153, "step": 14576 }, { "epoch": 16.34, "learning_rate": 1.0887323943661972e-05, "loss": 0.4111, "step": 14577 }, { "epoch": 16.34, "learning_rate": 1.0885915492957747e-05, "loss": 0.436, "step": 14578 }, { "epoch": 16.34, "learning_rate": 1.0884507042253522e-05, "loss": 0.4706, "step": 14579 }, { "epoch": 16.35, "learning_rate": 1.0883098591549295e-05, "loss": 0.5536, "step": 14580 }, { "epoch": 16.35, "learning_rate": 1.0881690140845071e-05, "loss": 0.5278, "step": 14581 }, { "epoch": 16.35, "learning_rate": 1.0880281690140846e-05, "loss": 0.5198, "step": 14582 }, { "epoch": 16.35, "learning_rate": 1.087887323943662e-05, "loss": 0.3704, "step": 14583 }, { "epoch": 16.35, "learning_rate": 1.0877464788732395e-05, "loss": 0.5202, "step": 14584 }, { "epoch": 16.35, "learning_rate": 1.087605633802817e-05, "loss": 0.409, "step": 14585 }, { "epoch": 16.35, "learning_rate": 1.0874647887323944e-05, "loss": 0.5166, "step": 14586 }, { "epoch": 16.35, "learning_rate": 1.0873239436619719e-05, "loss": 0.4348, "step": 14587 }, { "epoch": 16.35, "learning_rate": 1.0871830985915492e-05, "loss": 0.562, "step": 14588 }, { "epoch": 16.36, "learning_rate": 1.0870422535211268e-05, "loss": 0.5432, "step": 14589 }, { "epoch": 16.36, "learning_rate": 1.0869014084507043e-05, "loss": 0.4657, "step": 14590 }, { "epoch": 16.36, "learning_rate": 1.0867605633802817e-05, "loss": 0.3889, "step": 14591 }, { "epoch": 16.36, "learning_rate": 1.0866197183098592e-05, "loss": 0.4951, "step": 14592 }, { "epoch": 16.36, "learning_rate": 1.0864788732394367e-05, "loss": 0.393, "step": 14593 }, { "epoch": 16.36, "learning_rate": 1.0863380281690141e-05, "loss": 0.4211, "step": 14594 }, { "epoch": 16.36, "learning_rate": 1.0861971830985916e-05, "loss": 0.456, "step": 14595 }, { "epoch": 16.36, "learning_rate": 1.086056338028169e-05, "loss": 0.5162, "step": 14596 }, { "epoch": 16.36, "learning_rate": 1.0859154929577465e-05, "loss": 0.4795, "step": 14597 }, { "epoch": 16.37, "learning_rate": 1.085774647887324e-05, "loss": 0.5085, "step": 14598 }, { "epoch": 16.37, "learning_rate": 1.0856338028169014e-05, "loss": 0.4698, "step": 14599 }, { "epoch": 16.37, "learning_rate": 1.0854929577464789e-05, "loss": 0.3838, "step": 14600 }, { "epoch": 16.37, "learning_rate": 1.0853521126760564e-05, "loss": 0.429, "step": 14601 }, { "epoch": 16.37, "learning_rate": 1.0852112676056338e-05, "loss": 0.4612, "step": 14602 }, { "epoch": 16.37, "learning_rate": 1.0850704225352113e-05, "loss": 0.4193, "step": 14603 }, { "epoch": 16.37, "learning_rate": 1.0849295774647888e-05, "loss": 0.3923, "step": 14604 }, { "epoch": 16.37, "learning_rate": 1.0847887323943662e-05, "loss": 0.5329, "step": 14605 }, { "epoch": 16.37, "learning_rate": 1.0846478873239437e-05, "loss": 0.4378, "step": 14606 }, { "epoch": 16.38, "learning_rate": 1.0845070422535212e-05, "loss": 0.4489, "step": 14607 }, { "epoch": 16.38, "learning_rate": 1.0843661971830986e-05, "loss": 0.5719, "step": 14608 }, { "epoch": 16.38, "learning_rate": 1.084225352112676e-05, "loss": 0.5387, "step": 14609 }, { "epoch": 16.38, "learning_rate": 1.0840845070422535e-05, "loss": 0.5114, "step": 14610 }, { "epoch": 16.38, "learning_rate": 1.083943661971831e-05, "loss": 0.5029, "step": 14611 }, { "epoch": 16.38, "learning_rate": 1.0838028169014085e-05, "loss": 0.4641, "step": 14612 }, { "epoch": 16.38, "learning_rate": 1.083661971830986e-05, "loss": 0.4171, "step": 14613 }, { "epoch": 16.38, "learning_rate": 1.0835211267605634e-05, "loss": 0.5096, "step": 14614 }, { "epoch": 16.38, "learning_rate": 1.0833802816901409e-05, "loss": 0.4493, "step": 14615 }, { "epoch": 16.39, "learning_rate": 1.0832394366197183e-05, "loss": 0.6037, "step": 14616 }, { "epoch": 16.39, "learning_rate": 1.0830985915492958e-05, "loss": 0.4885, "step": 14617 }, { "epoch": 16.39, "learning_rate": 1.0829577464788733e-05, "loss": 0.383, "step": 14618 }, { "epoch": 16.39, "learning_rate": 1.0828169014084507e-05, "loss": 0.3055, "step": 14619 }, { "epoch": 16.39, "learning_rate": 1.0826760563380282e-05, "loss": 0.4708, "step": 14620 }, { "epoch": 16.39, "learning_rate": 1.0825352112676056e-05, "loss": 0.5088, "step": 14621 }, { "epoch": 16.39, "learning_rate": 1.0823943661971831e-05, "loss": 0.5245, "step": 14622 }, { "epoch": 16.39, "learning_rate": 1.0822535211267606e-05, "loss": 0.7076, "step": 14623 }, { "epoch": 16.39, "learning_rate": 1.082112676056338e-05, "loss": 0.5317, "step": 14624 }, { "epoch": 16.4, "learning_rate": 1.0819718309859155e-05, "loss": 0.5563, "step": 14625 }, { "epoch": 16.4, "learning_rate": 1.0818309859154931e-05, "loss": 0.5474, "step": 14626 }, { "epoch": 16.4, "learning_rate": 1.0816901408450704e-05, "loss": 0.5813, "step": 14627 }, { "epoch": 16.4, "learning_rate": 1.0815492957746479e-05, "loss": 0.4919, "step": 14628 }, { "epoch": 16.4, "learning_rate": 1.0814084507042254e-05, "loss": 0.5058, "step": 14629 }, { "epoch": 16.4, "learning_rate": 1.0812676056338028e-05, "loss": 0.6244, "step": 14630 }, { "epoch": 16.4, "learning_rate": 1.0811267605633803e-05, "loss": 0.5564, "step": 14631 }, { "epoch": 16.4, "learning_rate": 1.0809859154929577e-05, "loss": 0.5388, "step": 14632 }, { "epoch": 16.4, "learning_rate": 1.0808450704225352e-05, "loss": 0.4309, "step": 14633 }, { "epoch": 16.41, "learning_rate": 1.0807042253521128e-05, "loss": 0.5883, "step": 14634 }, { "epoch": 16.41, "learning_rate": 1.0805633802816901e-05, "loss": 0.4511, "step": 14635 }, { "epoch": 16.41, "learning_rate": 1.0804225352112676e-05, "loss": 0.6184, "step": 14636 }, { "epoch": 16.41, "learning_rate": 1.080281690140845e-05, "loss": 0.4921, "step": 14637 }, { "epoch": 16.41, "learning_rate": 1.0801408450704225e-05, "loss": 0.4946, "step": 14638 }, { "epoch": 16.41, "learning_rate": 1.08e-05, "loss": 0.4597, "step": 14639 }, { "epoch": 16.41, "learning_rate": 1.0798591549295775e-05, "loss": 0.4836, "step": 14640 }, { "epoch": 16.41, "learning_rate": 1.0797183098591551e-05, "loss": 0.4683, "step": 14641 }, { "epoch": 16.41, "learning_rate": 1.0795774647887326e-05, "loss": 0.5484, "step": 14642 }, { "epoch": 16.42, "learning_rate": 1.0794366197183098e-05, "loss": 0.5807, "step": 14643 }, { "epoch": 16.42, "learning_rate": 1.0792957746478873e-05, "loss": 0.394, "step": 14644 }, { "epoch": 16.42, "learning_rate": 1.0791549295774648e-05, "loss": 0.524, "step": 14645 }, { "epoch": 16.42, "learning_rate": 1.0790140845070422e-05, "loss": 0.3657, "step": 14646 }, { "epoch": 16.42, "learning_rate": 1.0788732394366197e-05, "loss": 0.4098, "step": 14647 }, { "epoch": 16.42, "learning_rate": 1.0787323943661972e-05, "loss": 0.4585, "step": 14648 }, { "epoch": 16.42, "learning_rate": 1.0785915492957748e-05, "loss": 0.3305, "step": 14649 }, { "epoch": 16.42, "learning_rate": 1.0784507042253523e-05, "loss": 0.4278, "step": 14650 }, { "epoch": 16.42, "learning_rate": 1.0783098591549296e-05, "loss": 0.4515, "step": 14651 }, { "epoch": 16.43, "learning_rate": 1.078169014084507e-05, "loss": 0.493, "step": 14652 }, { "epoch": 16.43, "learning_rate": 1.0780281690140845e-05, "loss": 0.4145, "step": 14653 }, { "epoch": 16.43, "learning_rate": 1.077887323943662e-05, "loss": 0.4487, "step": 14654 }, { "epoch": 16.43, "learning_rate": 1.0777464788732394e-05, "loss": 0.5906, "step": 14655 }, { "epoch": 16.43, "learning_rate": 1.0776056338028169e-05, "loss": 0.5001, "step": 14656 }, { "epoch": 16.43, "learning_rate": 1.0774647887323945e-05, "loss": 0.4783, "step": 14657 }, { "epoch": 16.43, "learning_rate": 1.077323943661972e-05, "loss": 0.4716, "step": 14658 }, { "epoch": 16.43, "learning_rate": 1.0771830985915493e-05, "loss": 0.4528, "step": 14659 }, { "epoch": 16.43, "learning_rate": 1.0770422535211267e-05, "loss": 0.4957, "step": 14660 }, { "epoch": 16.44, "learning_rate": 1.0769014084507042e-05, "loss": 0.4384, "step": 14661 }, { "epoch": 16.44, "learning_rate": 1.0767605633802817e-05, "loss": 0.3529, "step": 14662 }, { "epoch": 16.44, "learning_rate": 1.0766197183098591e-05, "loss": 0.4117, "step": 14663 }, { "epoch": 16.44, "learning_rate": 1.0764788732394368e-05, "loss": 0.5332, "step": 14664 }, { "epoch": 16.44, "learning_rate": 1.0763380281690142e-05, "loss": 0.4505, "step": 14665 }, { "epoch": 16.44, "learning_rate": 1.0761971830985917e-05, "loss": 0.4118, "step": 14666 }, { "epoch": 16.44, "learning_rate": 1.076056338028169e-05, "loss": 0.3119, "step": 14667 }, { "epoch": 16.44, "learning_rate": 1.0759154929577464e-05, "loss": 0.5997, "step": 14668 }, { "epoch": 16.45, "learning_rate": 1.0757746478873239e-05, "loss": 0.3941, "step": 14669 }, { "epoch": 16.45, "learning_rate": 1.0756338028169014e-05, "loss": 0.4982, "step": 14670 }, { "epoch": 16.45, "learning_rate": 1.0754929577464788e-05, "loss": 0.3213, "step": 14671 }, { "epoch": 16.45, "learning_rate": 1.0753521126760565e-05, "loss": 0.3309, "step": 14672 }, { "epoch": 16.45, "learning_rate": 1.075211267605634e-05, "loss": 0.6947, "step": 14673 }, { "epoch": 16.45, "learning_rate": 1.0750704225352114e-05, "loss": 0.7008, "step": 14674 }, { "epoch": 16.45, "learning_rate": 1.0749295774647887e-05, "loss": 0.4478, "step": 14675 }, { "epoch": 16.45, "learning_rate": 1.0747887323943662e-05, "loss": 0.6791, "step": 14676 }, { "epoch": 16.45, "learning_rate": 1.0746478873239436e-05, "loss": 0.4381, "step": 14677 }, { "epoch": 16.46, "learning_rate": 1.074507042253521e-05, "loss": 0.4825, "step": 14678 }, { "epoch": 16.46, "learning_rate": 1.0743661971830987e-05, "loss": 0.5164, "step": 14679 }, { "epoch": 16.46, "learning_rate": 1.0742253521126762e-05, "loss": 0.5674, "step": 14680 }, { "epoch": 16.46, "learning_rate": 1.0740845070422536e-05, "loss": 0.5323, "step": 14681 }, { "epoch": 16.46, "learning_rate": 1.0739436619718311e-05, "loss": 0.5717, "step": 14682 }, { "epoch": 16.46, "learning_rate": 1.0738028169014084e-05, "loss": 0.3526, "step": 14683 }, { "epoch": 16.46, "learning_rate": 1.0736619718309859e-05, "loss": 0.4016, "step": 14684 }, { "epoch": 16.46, "learning_rate": 1.0735211267605633e-05, "loss": 0.4338, "step": 14685 }, { "epoch": 16.46, "learning_rate": 1.0733802816901408e-05, "loss": 0.4431, "step": 14686 }, { "epoch": 16.47, "learning_rate": 1.0732394366197184e-05, "loss": 0.5065, "step": 14687 }, { "epoch": 16.47, "learning_rate": 1.0730985915492959e-05, "loss": 0.4252, "step": 14688 }, { "epoch": 16.47, "learning_rate": 1.0729577464788734e-05, "loss": 0.5588, "step": 14689 }, { "epoch": 16.47, "learning_rate": 1.0728169014084508e-05, "loss": 0.4588, "step": 14690 }, { "epoch": 16.47, "learning_rate": 1.0726760563380281e-05, "loss": 0.4875, "step": 14691 }, { "epoch": 16.47, "learning_rate": 1.0725352112676056e-05, "loss": 0.525, "step": 14692 }, { "epoch": 16.47, "learning_rate": 1.072394366197183e-05, "loss": 0.339, "step": 14693 }, { "epoch": 16.47, "learning_rate": 1.0722535211267607e-05, "loss": 0.3598, "step": 14694 }, { "epoch": 16.47, "learning_rate": 1.0721126760563381e-05, "loss": 0.4699, "step": 14695 }, { "epoch": 16.48, "learning_rate": 1.0719718309859156e-05, "loss": 0.4991, "step": 14696 }, { "epoch": 16.48, "learning_rate": 1.071830985915493e-05, "loss": 0.4662, "step": 14697 }, { "epoch": 16.48, "learning_rate": 1.0716901408450705e-05, "loss": 0.3373, "step": 14698 }, { "epoch": 16.48, "learning_rate": 1.0715492957746478e-05, "loss": 0.4982, "step": 14699 }, { "epoch": 16.48, "learning_rate": 1.0714084507042253e-05, "loss": 0.4855, "step": 14700 }, { "epoch": 16.48, "learning_rate": 1.0712676056338027e-05, "loss": 0.4376, "step": 14701 }, { "epoch": 16.48, "learning_rate": 1.0711267605633804e-05, "loss": 0.5659, "step": 14702 }, { "epoch": 16.48, "learning_rate": 1.0709859154929578e-05, "loss": 0.4663, "step": 14703 }, { "epoch": 16.48, "learning_rate": 1.0708450704225353e-05, "loss": 0.4779, "step": 14704 }, { "epoch": 16.49, "learning_rate": 1.0707042253521128e-05, "loss": 0.4832, "step": 14705 }, { "epoch": 16.49, "learning_rate": 1.0705633802816902e-05, "loss": 0.3879, "step": 14706 }, { "epoch": 16.49, "learning_rate": 1.0704225352112675e-05, "loss": 0.3606, "step": 14707 }, { "epoch": 16.49, "learning_rate": 1.070281690140845e-05, "loss": 0.5515, "step": 14708 }, { "epoch": 16.49, "learning_rate": 1.0701408450704225e-05, "loss": 0.562, "step": 14709 }, { "epoch": 16.49, "learning_rate": 1.0700000000000001e-05, "loss": 0.4282, "step": 14710 }, { "epoch": 16.49, "learning_rate": 1.0698591549295776e-05, "loss": 0.4987, "step": 14711 }, { "epoch": 16.49, "learning_rate": 1.069718309859155e-05, "loss": 0.5116, "step": 14712 }, { "epoch": 16.49, "learning_rate": 1.0695774647887325e-05, "loss": 0.4828, "step": 14713 }, { "epoch": 16.5, "learning_rate": 1.06943661971831e-05, "loss": 0.4871, "step": 14714 }, { "epoch": 16.5, "learning_rate": 1.0692957746478872e-05, "loss": 0.4412, "step": 14715 }, { "epoch": 16.5, "learning_rate": 1.0691549295774647e-05, "loss": 0.368, "step": 14716 }, { "epoch": 16.5, "learning_rate": 1.0690140845070423e-05, "loss": 0.4834, "step": 14717 }, { "epoch": 16.5, "learning_rate": 1.0688732394366198e-05, "loss": 0.3042, "step": 14718 }, { "epoch": 16.5, "learning_rate": 1.0687323943661973e-05, "loss": 0.3856, "step": 14719 }, { "epoch": 16.5, "learning_rate": 1.0685915492957747e-05, "loss": 0.4434, "step": 14720 }, { "epoch": 16.5, "learning_rate": 1.0684507042253522e-05, "loss": 0.2825, "step": 14721 }, { "epoch": 16.5, "learning_rate": 1.0683098591549295e-05, "loss": 0.4098, "step": 14722 }, { "epoch": 16.51, "learning_rate": 1.068169014084507e-05, "loss": 0.5832, "step": 14723 }, { "epoch": 16.51, "learning_rate": 1.0680281690140844e-05, "loss": 0.5028, "step": 14724 }, { "epoch": 16.51, "learning_rate": 1.067887323943662e-05, "loss": 0.6255, "step": 14725 }, { "epoch": 16.51, "learning_rate": 1.0677464788732395e-05, "loss": 0.5005, "step": 14726 }, { "epoch": 16.51, "learning_rate": 1.067605633802817e-05, "loss": 0.635, "step": 14727 }, { "epoch": 16.51, "learning_rate": 1.0674647887323944e-05, "loss": 0.5359, "step": 14728 }, { "epoch": 16.51, "learning_rate": 1.0673239436619719e-05, "loss": 0.4633, "step": 14729 }, { "epoch": 16.51, "learning_rate": 1.0671830985915492e-05, "loss": 0.5822, "step": 14730 }, { "epoch": 16.51, "learning_rate": 1.0670422535211267e-05, "loss": 0.5355, "step": 14731 }, { "epoch": 16.52, "learning_rate": 1.0669014084507043e-05, "loss": 0.7158, "step": 14732 }, { "epoch": 16.52, "learning_rate": 1.0667605633802818e-05, "loss": 0.4507, "step": 14733 }, { "epoch": 16.52, "learning_rate": 1.0666197183098592e-05, "loss": 0.4407, "step": 14734 }, { "epoch": 16.52, "learning_rate": 1.0664788732394367e-05, "loss": 0.5161, "step": 14735 }, { "epoch": 16.52, "learning_rate": 1.0663380281690141e-05, "loss": 0.5866, "step": 14736 }, { "epoch": 16.52, "learning_rate": 1.0661971830985916e-05, "loss": 0.5766, "step": 14737 }, { "epoch": 16.52, "learning_rate": 1.0660563380281689e-05, "loss": 0.442, "step": 14738 }, { "epoch": 16.52, "learning_rate": 1.0659154929577464e-05, "loss": 0.3946, "step": 14739 }, { "epoch": 16.52, "learning_rate": 1.065774647887324e-05, "loss": 0.4617, "step": 14740 }, { "epoch": 16.53, "learning_rate": 1.0656338028169015e-05, "loss": 0.4153, "step": 14741 }, { "epoch": 16.53, "learning_rate": 1.065492957746479e-05, "loss": 0.4651, "step": 14742 }, { "epoch": 16.53, "learning_rate": 1.0653521126760564e-05, "loss": 0.4651, "step": 14743 }, { "epoch": 16.53, "learning_rate": 1.0652112676056339e-05, "loss": 0.4532, "step": 14744 }, { "epoch": 16.53, "learning_rate": 1.0650704225352113e-05, "loss": 0.4461, "step": 14745 }, { "epoch": 16.53, "learning_rate": 1.0649295774647886e-05, "loss": 0.5056, "step": 14746 }, { "epoch": 16.53, "learning_rate": 1.0647887323943662e-05, "loss": 0.4362, "step": 14747 }, { "epoch": 16.53, "learning_rate": 1.0646478873239437e-05, "loss": 0.5444, "step": 14748 }, { "epoch": 16.53, "learning_rate": 1.0645070422535212e-05, "loss": 0.4287, "step": 14749 }, { "epoch": 16.54, "learning_rate": 1.0643661971830986e-05, "loss": 0.4231, "step": 14750 }, { "epoch": 16.54, "learning_rate": 1.0642253521126761e-05, "loss": 0.4258, "step": 14751 }, { "epoch": 16.54, "learning_rate": 1.0640845070422536e-05, "loss": 0.4663, "step": 14752 }, { "epoch": 16.54, "learning_rate": 1.063943661971831e-05, "loss": 0.3295, "step": 14753 }, { "epoch": 16.54, "learning_rate": 1.0638028169014083e-05, "loss": 0.5354, "step": 14754 }, { "epoch": 16.54, "learning_rate": 1.063661971830986e-05, "loss": 0.3133, "step": 14755 }, { "epoch": 16.54, "learning_rate": 1.0635211267605634e-05, "loss": 0.4835, "step": 14756 }, { "epoch": 16.54, "learning_rate": 1.0633802816901409e-05, "loss": 0.4866, "step": 14757 }, { "epoch": 16.54, "learning_rate": 1.0632394366197183e-05, "loss": 0.5458, "step": 14758 }, { "epoch": 16.55, "learning_rate": 1.0630985915492958e-05, "loss": 0.4666, "step": 14759 }, { "epoch": 16.55, "learning_rate": 1.0629577464788733e-05, "loss": 0.5074, "step": 14760 }, { "epoch": 16.55, "learning_rate": 1.0628169014084507e-05, "loss": 0.3838, "step": 14761 }, { "epoch": 16.55, "learning_rate": 1.0626760563380282e-05, "loss": 0.5082, "step": 14762 }, { "epoch": 16.55, "learning_rate": 1.0625352112676057e-05, "loss": 0.3608, "step": 14763 }, { "epoch": 16.55, "learning_rate": 1.0623943661971831e-05, "loss": 0.4353, "step": 14764 }, { "epoch": 16.55, "learning_rate": 1.0622535211267606e-05, "loss": 0.3914, "step": 14765 }, { "epoch": 16.55, "learning_rate": 1.062112676056338e-05, "loss": 0.3937, "step": 14766 }, { "epoch": 16.55, "learning_rate": 1.0619718309859155e-05, "loss": 0.4136, "step": 14767 }, { "epoch": 16.56, "learning_rate": 1.061830985915493e-05, "loss": 0.3929, "step": 14768 }, { "epoch": 16.56, "learning_rate": 1.0616901408450705e-05, "loss": 0.3535, "step": 14769 }, { "epoch": 16.56, "learning_rate": 1.0615492957746479e-05, "loss": 0.4367, "step": 14770 }, { "epoch": 16.56, "learning_rate": 1.0614084507042254e-05, "loss": 0.5204, "step": 14771 }, { "epoch": 16.56, "learning_rate": 1.0612676056338028e-05, "loss": 0.403, "step": 14772 }, { "epoch": 16.56, "learning_rate": 1.0611267605633803e-05, "loss": 0.6311, "step": 14773 }, { "epoch": 16.56, "learning_rate": 1.0609859154929578e-05, "loss": 0.5333, "step": 14774 }, { "epoch": 16.56, "learning_rate": 1.0608450704225352e-05, "loss": 0.5448, "step": 14775 }, { "epoch": 16.57, "learning_rate": 1.0607042253521127e-05, "loss": 0.4778, "step": 14776 }, { "epoch": 16.57, "learning_rate": 1.0605633802816902e-05, "loss": 0.463, "step": 14777 }, { "epoch": 16.57, "learning_rate": 1.0604225352112676e-05, "loss": 0.4972, "step": 14778 }, { "epoch": 16.57, "learning_rate": 1.0602816901408451e-05, "loss": 0.4901, "step": 14779 }, { "epoch": 16.57, "learning_rate": 1.0601408450704226e-05, "loss": 0.4254, "step": 14780 }, { "epoch": 16.57, "learning_rate": 1.06e-05, "loss": 0.4666, "step": 14781 }, { "epoch": 16.57, "learning_rate": 1.0598591549295775e-05, "loss": 0.5034, "step": 14782 }, { "epoch": 16.57, "learning_rate": 1.059718309859155e-05, "loss": 0.6132, "step": 14783 }, { "epoch": 16.57, "learning_rate": 1.0595774647887324e-05, "loss": 0.5395, "step": 14784 }, { "epoch": 16.58, "learning_rate": 1.05943661971831e-05, "loss": 0.4162, "step": 14785 }, { "epoch": 16.58, "learning_rate": 1.0592957746478873e-05, "loss": 0.3555, "step": 14786 }, { "epoch": 16.58, "learning_rate": 1.0591549295774648e-05, "loss": 0.546, "step": 14787 }, { "epoch": 16.58, "learning_rate": 1.0590140845070423e-05, "loss": 0.4108, "step": 14788 }, { "epoch": 16.58, "learning_rate": 1.0588732394366197e-05, "loss": 0.4283, "step": 14789 }, { "epoch": 16.58, "learning_rate": 1.0587323943661972e-05, "loss": 0.486, "step": 14790 }, { "epoch": 16.58, "learning_rate": 1.0585915492957747e-05, "loss": 0.5409, "step": 14791 }, { "epoch": 16.58, "learning_rate": 1.0584507042253521e-05, "loss": 0.344, "step": 14792 }, { "epoch": 16.58, "learning_rate": 1.0583098591549297e-05, "loss": 0.5594, "step": 14793 }, { "epoch": 16.59, "learning_rate": 1.058169014084507e-05, "loss": 0.6204, "step": 14794 }, { "epoch": 16.59, "learning_rate": 1.0580281690140845e-05, "loss": 0.5211, "step": 14795 }, { "epoch": 16.59, "learning_rate": 1.057887323943662e-05, "loss": 0.4777, "step": 14796 }, { "epoch": 16.59, "learning_rate": 1.0577464788732394e-05, "loss": 0.5001, "step": 14797 }, { "epoch": 16.59, "learning_rate": 1.0576056338028169e-05, "loss": 0.4516, "step": 14798 }, { "epoch": 16.59, "learning_rate": 1.0574647887323944e-05, "loss": 0.5738, "step": 14799 }, { "epoch": 16.59, "learning_rate": 1.057323943661972e-05, "loss": 0.6555, "step": 14800 }, { "epoch": 16.59, "learning_rate": 1.0571830985915495e-05, "loss": 0.4637, "step": 14801 }, { "epoch": 16.59, "learning_rate": 1.0570422535211268e-05, "loss": 0.5052, "step": 14802 }, { "epoch": 16.6, "learning_rate": 1.0569014084507042e-05, "loss": 0.5333, "step": 14803 }, { "epoch": 16.6, "learning_rate": 1.0567605633802817e-05, "loss": 0.5089, "step": 14804 }, { "epoch": 16.6, "learning_rate": 1.0566197183098591e-05, "loss": 0.4929, "step": 14805 }, { "epoch": 16.6, "learning_rate": 1.0564788732394366e-05, "loss": 0.4921, "step": 14806 }, { "epoch": 16.6, "learning_rate": 1.056338028169014e-05, "loss": 0.4813, "step": 14807 }, { "epoch": 16.6, "learning_rate": 1.0561971830985917e-05, "loss": 0.4453, "step": 14808 }, { "epoch": 16.6, "learning_rate": 1.0560563380281692e-05, "loss": 0.406, "step": 14809 }, { "epoch": 16.6, "learning_rate": 1.0559154929577465e-05, "loss": 0.4465, "step": 14810 }, { "epoch": 16.6, "learning_rate": 1.055774647887324e-05, "loss": 0.498, "step": 14811 }, { "epoch": 16.61, "learning_rate": 1.0556338028169014e-05, "loss": 0.5474, "step": 14812 }, { "epoch": 16.61, "learning_rate": 1.0554929577464789e-05, "loss": 0.4691, "step": 14813 }, { "epoch": 16.61, "learning_rate": 1.0553521126760563e-05, "loss": 0.5254, "step": 14814 }, { "epoch": 16.61, "learning_rate": 1.055211267605634e-05, "loss": 0.5483, "step": 14815 }, { "epoch": 16.61, "learning_rate": 1.0550704225352114e-05, "loss": 0.481, "step": 14816 }, { "epoch": 16.61, "learning_rate": 1.0549295774647889e-05, "loss": 0.3704, "step": 14817 }, { "epoch": 16.61, "learning_rate": 1.0547887323943662e-05, "loss": 0.5485, "step": 14818 }, { "epoch": 16.61, "learning_rate": 1.0546478873239436e-05, "loss": 0.3971, "step": 14819 }, { "epoch": 16.61, "learning_rate": 1.0545070422535211e-05, "loss": 0.4771, "step": 14820 }, { "epoch": 16.62, "learning_rate": 1.0543661971830986e-05, "loss": 0.3002, "step": 14821 }, { "epoch": 16.62, "learning_rate": 1.054225352112676e-05, "loss": 0.5629, "step": 14822 }, { "epoch": 16.62, "learning_rate": 1.0540845070422537e-05, "loss": 0.6413, "step": 14823 }, { "epoch": 16.62, "learning_rate": 1.0539436619718311e-05, "loss": 0.5923, "step": 14824 }, { "epoch": 16.62, "learning_rate": 1.0538028169014086e-05, "loss": 0.5676, "step": 14825 }, { "epoch": 16.62, "learning_rate": 1.0536619718309859e-05, "loss": 0.4633, "step": 14826 }, { "epoch": 16.62, "learning_rate": 1.0535211267605633e-05, "loss": 0.6023, "step": 14827 }, { "epoch": 16.62, "learning_rate": 1.0533802816901408e-05, "loss": 0.5207, "step": 14828 }, { "epoch": 16.62, "learning_rate": 1.0532394366197183e-05, "loss": 0.6045, "step": 14829 }, { "epoch": 16.63, "learning_rate": 1.0530985915492957e-05, "loss": 0.5511, "step": 14830 }, { "epoch": 16.63, "learning_rate": 1.0529577464788734e-05, "loss": 0.4688, "step": 14831 }, { "epoch": 16.63, "learning_rate": 1.0528169014084508e-05, "loss": 0.5361, "step": 14832 }, { "epoch": 16.63, "learning_rate": 1.0526760563380281e-05, "loss": 0.5823, "step": 14833 }, { "epoch": 16.63, "learning_rate": 1.0525352112676056e-05, "loss": 0.4476, "step": 14834 }, { "epoch": 16.63, "learning_rate": 1.052394366197183e-05, "loss": 0.5793, "step": 14835 }, { "epoch": 16.63, "learning_rate": 1.0522535211267605e-05, "loss": 0.4426, "step": 14836 }, { "epoch": 16.63, "learning_rate": 1.052112676056338e-05, "loss": 0.467, "step": 14837 }, { "epoch": 16.63, "learning_rate": 1.0519718309859156e-05, "loss": 0.3898, "step": 14838 }, { "epoch": 16.64, "learning_rate": 1.051830985915493e-05, "loss": 0.5178, "step": 14839 }, { "epoch": 16.64, "learning_rate": 1.0516901408450705e-05, "loss": 0.4955, "step": 14840 }, { "epoch": 16.64, "learning_rate": 1.0515492957746478e-05, "loss": 0.5448, "step": 14841 }, { "epoch": 16.64, "learning_rate": 1.0514084507042253e-05, "loss": 0.4429, "step": 14842 }, { "epoch": 16.64, "learning_rate": 1.0512676056338028e-05, "loss": 0.4351, "step": 14843 }, { "epoch": 16.64, "learning_rate": 1.0511267605633802e-05, "loss": 0.5684, "step": 14844 }, { "epoch": 16.64, "learning_rate": 1.0509859154929577e-05, "loss": 0.4288, "step": 14845 }, { "epoch": 16.64, "learning_rate": 1.0508450704225353e-05, "loss": 0.4707, "step": 14846 }, { "epoch": 16.64, "learning_rate": 1.0507042253521128e-05, "loss": 0.5067, "step": 14847 }, { "epoch": 16.65, "learning_rate": 1.0505633802816903e-05, "loss": 0.517, "step": 14848 }, { "epoch": 16.65, "learning_rate": 1.0504225352112676e-05, "loss": 0.4889, "step": 14849 }, { "epoch": 16.65, "learning_rate": 1.050281690140845e-05, "loss": 0.3646, "step": 14850 }, { "epoch": 16.65, "learning_rate": 1.0501408450704225e-05, "loss": 0.5031, "step": 14851 }, { "epoch": 16.65, "learning_rate": 1.05e-05, "loss": 0.5046, "step": 14852 }, { "epoch": 16.65, "learning_rate": 1.0498591549295776e-05, "loss": 0.576, "step": 14853 }, { "epoch": 16.65, "learning_rate": 1.049718309859155e-05, "loss": 0.4637, "step": 14854 }, { "epoch": 16.65, "learning_rate": 1.0495774647887325e-05, "loss": 0.3998, "step": 14855 }, { "epoch": 16.65, "learning_rate": 1.04943661971831e-05, "loss": 0.4599, "step": 14856 }, { "epoch": 16.66, "learning_rate": 1.0492957746478873e-05, "loss": 0.5621, "step": 14857 }, { "epoch": 16.66, "learning_rate": 1.0491549295774647e-05, "loss": 0.5434, "step": 14858 }, { "epoch": 16.66, "learning_rate": 1.0490140845070422e-05, "loss": 0.4138, "step": 14859 }, { "epoch": 16.66, "learning_rate": 1.0488732394366197e-05, "loss": 0.3005, "step": 14860 }, { "epoch": 16.66, "learning_rate": 1.0487323943661973e-05, "loss": 0.5703, "step": 14861 }, { "epoch": 16.66, "learning_rate": 1.0485915492957747e-05, "loss": 0.3668, "step": 14862 }, { "epoch": 16.66, "learning_rate": 1.0484507042253522e-05, "loss": 0.4664, "step": 14863 }, { "epoch": 16.66, "learning_rate": 1.0483098591549297e-05, "loss": 0.5309, "step": 14864 }, { "epoch": 16.66, "learning_rate": 1.048169014084507e-05, "loss": 0.5709, "step": 14865 }, { "epoch": 16.67, "learning_rate": 1.0480281690140844e-05, "loss": 0.431, "step": 14866 }, { "epoch": 16.67, "learning_rate": 1.0478873239436619e-05, "loss": 0.3955, "step": 14867 }, { "epoch": 16.67, "learning_rate": 1.0477464788732395e-05, "loss": 0.2979, "step": 14868 }, { "epoch": 16.67, "learning_rate": 1.047605633802817e-05, "loss": 0.6293, "step": 14869 }, { "epoch": 16.67, "learning_rate": 1.0474647887323945e-05, "loss": 0.4177, "step": 14870 }, { "epoch": 16.67, "learning_rate": 1.047323943661972e-05, "loss": 0.5299, "step": 14871 }, { "epoch": 16.67, "learning_rate": 1.0471830985915494e-05, "loss": 0.3714, "step": 14872 }, { "epoch": 16.67, "learning_rate": 1.0470422535211267e-05, "loss": 0.5699, "step": 14873 }, { "epoch": 16.67, "learning_rate": 1.0469014084507041e-05, "loss": 0.5281, "step": 14874 }, { "epoch": 16.68, "learning_rate": 1.0467605633802816e-05, "loss": 0.6271, "step": 14875 }, { "epoch": 16.68, "learning_rate": 1.0466197183098592e-05, "loss": 0.5759, "step": 14876 }, { "epoch": 16.68, "learning_rate": 1.0464788732394367e-05, "loss": 0.4361, "step": 14877 }, { "epoch": 16.68, "learning_rate": 1.0463380281690142e-05, "loss": 0.4984, "step": 14878 }, { "epoch": 16.68, "learning_rate": 1.0461971830985916e-05, "loss": 0.5107, "step": 14879 }, { "epoch": 16.68, "learning_rate": 1.0460563380281691e-05, "loss": 0.4959, "step": 14880 }, { "epoch": 16.68, "learning_rate": 1.0459154929577464e-05, "loss": 0.5029, "step": 14881 }, { "epoch": 16.68, "learning_rate": 1.0457746478873239e-05, "loss": 0.521, "step": 14882 }, { "epoch": 16.68, "learning_rate": 1.0456338028169013e-05, "loss": 0.4083, "step": 14883 }, { "epoch": 16.69, "learning_rate": 1.045492957746479e-05, "loss": 0.5034, "step": 14884 }, { "epoch": 16.69, "learning_rate": 1.0453521126760564e-05, "loss": 0.4335, "step": 14885 }, { "epoch": 16.69, "learning_rate": 1.0452112676056339e-05, "loss": 0.4379, "step": 14886 }, { "epoch": 16.69, "learning_rate": 1.0450704225352113e-05, "loss": 0.5887, "step": 14887 }, { "epoch": 16.69, "learning_rate": 1.0449295774647888e-05, "loss": 0.3891, "step": 14888 }, { "epoch": 16.69, "learning_rate": 1.0447887323943661e-05, "loss": 0.4894, "step": 14889 }, { "epoch": 16.69, "learning_rate": 1.0446478873239436e-05, "loss": 0.4545, "step": 14890 }, { "epoch": 16.69, "learning_rate": 1.0445070422535212e-05, "loss": 0.4749, "step": 14891 }, { "epoch": 16.7, "learning_rate": 1.0443661971830987e-05, "loss": 0.4304, "step": 14892 }, { "epoch": 16.7, "learning_rate": 1.0442253521126761e-05, "loss": 0.5779, "step": 14893 }, { "epoch": 16.7, "learning_rate": 1.0440845070422536e-05, "loss": 0.5616, "step": 14894 }, { "epoch": 16.7, "learning_rate": 1.043943661971831e-05, "loss": 0.4661, "step": 14895 }, { "epoch": 16.7, "learning_rate": 1.0438028169014085e-05, "loss": 0.3806, "step": 14896 }, { "epoch": 16.7, "learning_rate": 1.0436619718309858e-05, "loss": 0.4089, "step": 14897 }, { "epoch": 16.7, "learning_rate": 1.0435211267605633e-05, "loss": 0.5329, "step": 14898 }, { "epoch": 16.7, "learning_rate": 1.0433802816901409e-05, "loss": 0.424, "step": 14899 }, { "epoch": 16.7, "learning_rate": 1.0432394366197184e-05, "loss": 0.4843, "step": 14900 }, { "epoch": 16.71, "learning_rate": 1.0430985915492958e-05, "loss": 0.4676, "step": 14901 }, { "epoch": 16.71, "learning_rate": 1.0429577464788733e-05, "loss": 0.3793, "step": 14902 }, { "epoch": 16.71, "learning_rate": 1.0428169014084508e-05, "loss": 0.4931, "step": 14903 }, { "epoch": 16.71, "learning_rate": 1.0426760563380282e-05, "loss": 0.5198, "step": 14904 }, { "epoch": 16.71, "learning_rate": 1.0425352112676055e-05, "loss": 0.4607, "step": 14905 }, { "epoch": 16.71, "learning_rate": 1.0423943661971832e-05, "loss": 0.4494, "step": 14906 }, { "epoch": 16.71, "learning_rate": 1.0422535211267606e-05, "loss": 0.437, "step": 14907 }, { "epoch": 16.71, "learning_rate": 1.042112676056338e-05, "loss": 0.4678, "step": 14908 }, { "epoch": 16.71, "learning_rate": 1.0419718309859155e-05, "loss": 0.4473, "step": 14909 }, { "epoch": 16.72, "learning_rate": 1.041830985915493e-05, "loss": 0.4559, "step": 14910 }, { "epoch": 16.72, "learning_rate": 1.0416901408450705e-05, "loss": 0.4877, "step": 14911 }, { "epoch": 16.72, "learning_rate": 1.041549295774648e-05, "loss": 0.4418, "step": 14912 }, { "epoch": 16.72, "learning_rate": 1.0414084507042252e-05, "loss": 0.377, "step": 14913 }, { "epoch": 16.72, "learning_rate": 1.0412676056338029e-05, "loss": 0.5199, "step": 14914 }, { "epoch": 16.72, "learning_rate": 1.0411267605633803e-05, "loss": 0.4491, "step": 14915 }, { "epoch": 16.72, "learning_rate": 1.0409859154929578e-05, "loss": 0.2593, "step": 14916 }, { "epoch": 16.72, "learning_rate": 1.0408450704225353e-05, "loss": 0.4102, "step": 14917 }, { "epoch": 16.72, "learning_rate": 1.0407042253521127e-05, "loss": 0.3665, "step": 14918 }, { "epoch": 16.73, "learning_rate": 1.0405633802816902e-05, "loss": 0.4487, "step": 14919 }, { "epoch": 16.73, "learning_rate": 1.0404225352112676e-05, "loss": 0.4217, "step": 14920 }, { "epoch": 16.73, "learning_rate": 1.0402816901408451e-05, "loss": 0.4201, "step": 14921 }, { "epoch": 16.73, "learning_rate": 1.0401408450704226e-05, "loss": 0.451, "step": 14922 }, { "epoch": 16.73, "learning_rate": 1.04e-05, "loss": 0.6488, "step": 14923 }, { "epoch": 16.73, "learning_rate": 1.0398591549295775e-05, "loss": 0.4026, "step": 14924 }, { "epoch": 16.73, "learning_rate": 1.039718309859155e-05, "loss": 0.581, "step": 14925 }, { "epoch": 16.73, "learning_rate": 1.0395774647887324e-05, "loss": 0.4682, "step": 14926 }, { "epoch": 16.73, "learning_rate": 1.0394366197183099e-05, "loss": 0.4858, "step": 14927 }, { "epoch": 16.74, "learning_rate": 1.0392957746478874e-05, "loss": 0.5057, "step": 14928 }, { "epoch": 16.74, "learning_rate": 1.0391549295774648e-05, "loss": 0.54, "step": 14929 }, { "epoch": 16.74, "learning_rate": 1.0390140845070423e-05, "loss": 0.473, "step": 14930 }, { "epoch": 16.74, "learning_rate": 1.0388732394366197e-05, "loss": 0.4928, "step": 14931 }, { "epoch": 16.74, "learning_rate": 1.0387323943661972e-05, "loss": 0.5282, "step": 14932 }, { "epoch": 16.74, "learning_rate": 1.0385915492957747e-05, "loss": 0.4181, "step": 14933 }, { "epoch": 16.74, "learning_rate": 1.0384507042253521e-05, "loss": 0.3946, "step": 14934 }, { "epoch": 16.74, "learning_rate": 1.0383098591549296e-05, "loss": 0.5746, "step": 14935 }, { "epoch": 16.74, "learning_rate": 1.038169014084507e-05, "loss": 0.4972, "step": 14936 }, { "epoch": 16.75, "learning_rate": 1.0380281690140845e-05, "loss": 0.5193, "step": 14937 }, { "epoch": 16.75, "learning_rate": 1.037887323943662e-05, "loss": 0.4109, "step": 14938 }, { "epoch": 16.75, "learning_rate": 1.0377464788732395e-05, "loss": 0.3313, "step": 14939 }, { "epoch": 16.75, "learning_rate": 1.037605633802817e-05, "loss": 0.5356, "step": 14940 }, { "epoch": 16.75, "learning_rate": 1.0374647887323944e-05, "loss": 0.5636, "step": 14941 }, { "epoch": 16.75, "learning_rate": 1.0373239436619718e-05, "loss": 0.4798, "step": 14942 }, { "epoch": 16.75, "learning_rate": 1.0371830985915493e-05, "loss": 0.5523, "step": 14943 }, { "epoch": 16.75, "learning_rate": 1.0370422535211268e-05, "loss": 0.4182, "step": 14944 }, { "epoch": 16.75, "learning_rate": 1.0369014084507042e-05, "loss": 0.434, "step": 14945 }, { "epoch": 16.76, "learning_rate": 1.0367605633802817e-05, "loss": 0.473, "step": 14946 }, { "epoch": 16.76, "learning_rate": 1.0366197183098592e-05, "loss": 0.4711, "step": 14947 }, { "epoch": 16.76, "learning_rate": 1.0364788732394366e-05, "loss": 0.3793, "step": 14948 }, { "epoch": 16.76, "learning_rate": 1.0363380281690141e-05, "loss": 0.4052, "step": 14949 }, { "epoch": 16.76, "learning_rate": 1.0361971830985916e-05, "loss": 0.4506, "step": 14950 }, { "epoch": 16.76, "learning_rate": 1.036056338028169e-05, "loss": 0.5275, "step": 14951 }, { "epoch": 16.76, "learning_rate": 1.0359154929577465e-05, "loss": 0.436, "step": 14952 }, { "epoch": 16.76, "learning_rate": 1.035774647887324e-05, "loss": 0.5118, "step": 14953 }, { "epoch": 16.76, "learning_rate": 1.0356338028169014e-05, "loss": 0.2817, "step": 14954 }, { "epoch": 16.77, "learning_rate": 1.0354929577464789e-05, "loss": 0.3826, "step": 14955 }, { "epoch": 16.77, "learning_rate": 1.0353521126760563e-05, "loss": 0.5424, "step": 14956 }, { "epoch": 16.77, "learning_rate": 1.0352112676056338e-05, "loss": 0.5477, "step": 14957 }, { "epoch": 16.77, "learning_rate": 1.0350704225352113e-05, "loss": 0.5032, "step": 14958 }, { "epoch": 16.77, "learning_rate": 1.0349295774647889e-05, "loss": 0.4239, "step": 14959 }, { "epoch": 16.77, "learning_rate": 1.0347887323943662e-05, "loss": 0.3224, "step": 14960 }, { "epoch": 16.77, "learning_rate": 1.0346478873239437e-05, "loss": 0.482, "step": 14961 }, { "epoch": 16.77, "learning_rate": 1.0345070422535211e-05, "loss": 0.66, "step": 14962 }, { "epoch": 16.77, "learning_rate": 1.0343661971830986e-05, "loss": 0.4098, "step": 14963 }, { "epoch": 16.78, "learning_rate": 1.034225352112676e-05, "loss": 0.3956, "step": 14964 }, { "epoch": 16.78, "learning_rate": 1.0340845070422535e-05, "loss": 0.3716, "step": 14965 }, { "epoch": 16.78, "learning_rate": 1.033943661971831e-05, "loss": 0.4133, "step": 14966 }, { "epoch": 16.78, "learning_rate": 1.0338028169014086e-05, "loss": 0.453, "step": 14967 }, { "epoch": 16.78, "learning_rate": 1.0336619718309859e-05, "loss": 0.483, "step": 14968 }, { "epoch": 16.78, "learning_rate": 1.0335211267605634e-05, "loss": 0.4632, "step": 14969 }, { "epoch": 16.78, "learning_rate": 1.0333802816901408e-05, "loss": 0.4777, "step": 14970 }, { "epoch": 16.78, "learning_rate": 1.0332394366197183e-05, "loss": 0.3574, "step": 14971 }, { "epoch": 16.78, "learning_rate": 1.0330985915492958e-05, "loss": 0.3854, "step": 14972 }, { "epoch": 16.79, "learning_rate": 1.0329577464788732e-05, "loss": 0.5678, "step": 14973 }, { "epoch": 16.79, "learning_rate": 1.0328169014084509e-05, "loss": 0.5884, "step": 14974 }, { "epoch": 16.79, "learning_rate": 1.0326760563380283e-05, "loss": 0.47, "step": 14975 }, { "epoch": 16.79, "learning_rate": 1.0325352112676056e-05, "loss": 0.52, "step": 14976 }, { "epoch": 16.79, "learning_rate": 1.032394366197183e-05, "loss": 0.5005, "step": 14977 }, { "epoch": 16.79, "learning_rate": 1.0322535211267605e-05, "loss": 0.4551, "step": 14978 }, { "epoch": 16.79, "learning_rate": 1.032112676056338e-05, "loss": 0.4899, "step": 14979 }, { "epoch": 16.79, "learning_rate": 1.0319718309859155e-05, "loss": 0.5651, "step": 14980 }, { "epoch": 16.79, "learning_rate": 1.031830985915493e-05, "loss": 0.5158, "step": 14981 }, { "epoch": 16.8, "learning_rate": 1.0316901408450706e-05, "loss": 0.5875, "step": 14982 }, { "epoch": 16.8, "learning_rate": 1.031549295774648e-05, "loss": 0.5252, "step": 14983 }, { "epoch": 16.8, "learning_rate": 1.0314084507042253e-05, "loss": 0.3748, "step": 14984 }, { "epoch": 16.8, "learning_rate": 1.0312676056338028e-05, "loss": 0.5017, "step": 14985 }, { "epoch": 16.8, "learning_rate": 1.0311267605633803e-05, "loss": 0.572, "step": 14986 }, { "epoch": 16.8, "learning_rate": 1.0309859154929577e-05, "loss": 0.4697, "step": 14987 }, { "epoch": 16.8, "learning_rate": 1.0308450704225352e-05, "loss": 0.5856, "step": 14988 }, { "epoch": 16.8, "learning_rate": 1.0307042253521128e-05, "loss": 0.532, "step": 14989 }, { "epoch": 16.8, "learning_rate": 1.0305633802816903e-05, "loss": 0.4935, "step": 14990 }, { "epoch": 16.81, "learning_rate": 1.0304225352112677e-05, "loss": 0.449, "step": 14991 }, { "epoch": 16.81, "learning_rate": 1.030281690140845e-05, "loss": 0.3924, "step": 14992 }, { "epoch": 16.81, "learning_rate": 1.0301408450704225e-05, "loss": 0.4175, "step": 14993 }, { "epoch": 16.81, "learning_rate": 1.03e-05, "loss": 0.4183, "step": 14994 }, { "epoch": 16.81, "learning_rate": 1.0298591549295774e-05, "loss": 0.4365, "step": 14995 }, { "epoch": 16.81, "learning_rate": 1.0297183098591549e-05, "loss": 0.3687, "step": 14996 }, { "epoch": 16.81, "learning_rate": 1.0295774647887325e-05, "loss": 0.5177, "step": 14997 }, { "epoch": 16.81, "learning_rate": 1.02943661971831e-05, "loss": 0.473, "step": 14998 }, { "epoch": 16.82, "learning_rate": 1.0292957746478875e-05, "loss": 0.5003, "step": 14999 }, { "epoch": 16.82, "learning_rate": 1.0291549295774647e-05, "loss": 0.4515, "step": 15000 }, { "epoch": 16.82, "eval_loss": 0.8180882334709167, "eval_runtime": 826.5269, "eval_samples_per_second": 3.197, "eval_steps_per_second": 0.4, "eval_wer": 0.14174930583101944, "step": 15000 }, { "epoch": 16.82, "learning_rate": 1.0290140845070422e-05, "loss": 0.3941, "step": 15001 }, { "epoch": 16.82, "learning_rate": 1.0288732394366197e-05, "loss": 0.4436, "step": 15002 }, { "epoch": 16.82, "learning_rate": 1.0287323943661971e-05, "loss": 0.41, "step": 15003 }, { "epoch": 16.82, "learning_rate": 1.0285915492957746e-05, "loss": 0.3894, "step": 15004 }, { "epoch": 16.82, "learning_rate": 1.0284507042253522e-05, "loss": 0.469, "step": 15005 }, { "epoch": 16.82, "learning_rate": 1.0283098591549297e-05, "loss": 0.4183, "step": 15006 }, { "epoch": 16.82, "learning_rate": 1.0281690140845072e-05, "loss": 0.4508, "step": 15007 }, { "epoch": 16.83, "learning_rate": 1.0280281690140845e-05, "loss": 0.4423, "step": 15008 }, { "epoch": 16.83, "learning_rate": 1.027887323943662e-05, "loss": 0.4908, "step": 15009 }, { "epoch": 16.83, "learning_rate": 1.0277464788732394e-05, "loss": 0.3973, "step": 15010 }, { "epoch": 16.83, "learning_rate": 1.0276056338028168e-05, "loss": 0.3802, "step": 15011 }, { "epoch": 16.83, "learning_rate": 1.0274647887323945e-05, "loss": 0.4965, "step": 15012 }, { "epoch": 16.83, "learning_rate": 1.027323943661972e-05, "loss": 0.4105, "step": 15013 }, { "epoch": 16.83, "learning_rate": 1.0271830985915494e-05, "loss": 0.4688, "step": 15014 }, { "epoch": 16.83, "learning_rate": 1.0270422535211269e-05, "loss": 0.3538, "step": 15015 }, { "epoch": 16.83, "learning_rate": 1.0269014084507042e-05, "loss": 0.3889, "step": 15016 }, { "epoch": 16.84, "learning_rate": 1.0267605633802816e-05, "loss": 0.3262, "step": 15017 }, { "epoch": 16.84, "learning_rate": 1.0266197183098591e-05, "loss": 0.3579, "step": 15018 }, { "epoch": 16.84, "learning_rate": 1.0264788732394366e-05, "loss": 0.3361, "step": 15019 }, { "epoch": 16.84, "learning_rate": 1.0263380281690142e-05, "loss": 0.4137, "step": 15020 }, { "epoch": 16.84, "learning_rate": 1.0261971830985917e-05, "loss": 0.487, "step": 15021 }, { "epoch": 16.84, "learning_rate": 1.0260563380281691e-05, "loss": 0.3328, "step": 15022 }, { "epoch": 16.84, "learning_rate": 1.0259154929577466e-05, "loss": 0.7101, "step": 15023 }, { "epoch": 16.84, "learning_rate": 1.0257746478873239e-05, "loss": 0.6303, "step": 15024 }, { "epoch": 16.84, "learning_rate": 1.0256338028169013e-05, "loss": 0.5001, "step": 15025 }, { "epoch": 16.85, "learning_rate": 1.0254929577464788e-05, "loss": 0.5058, "step": 15026 }, { "epoch": 16.85, "learning_rate": 1.0253521126760564e-05, "loss": 0.5233, "step": 15027 }, { "epoch": 16.85, "learning_rate": 1.0252112676056339e-05, "loss": 0.4975, "step": 15028 }, { "epoch": 16.85, "learning_rate": 1.0250704225352114e-05, "loss": 0.5232, "step": 15029 }, { "epoch": 16.85, "learning_rate": 1.0249295774647888e-05, "loss": 0.5774, "step": 15030 }, { "epoch": 16.85, "learning_rate": 1.0247887323943663e-05, "loss": 0.4677, "step": 15031 }, { "epoch": 16.85, "learning_rate": 1.0246478873239436e-05, "loss": 0.4524, "step": 15032 }, { "epoch": 16.85, "learning_rate": 1.024507042253521e-05, "loss": 0.4516, "step": 15033 }, { "epoch": 16.85, "learning_rate": 1.0243661971830985e-05, "loss": 0.5077, "step": 15034 }, { "epoch": 16.86, "learning_rate": 1.0242253521126761e-05, "loss": 0.572, "step": 15035 }, { "epoch": 16.86, "learning_rate": 1.0240845070422536e-05, "loss": 0.5201, "step": 15036 }, { "epoch": 16.86, "learning_rate": 1.023943661971831e-05, "loss": 0.3854, "step": 15037 }, { "epoch": 16.86, "learning_rate": 1.0238028169014085e-05, "loss": 0.5258, "step": 15038 }, { "epoch": 16.86, "learning_rate": 1.023661971830986e-05, "loss": 0.5065, "step": 15039 }, { "epoch": 16.86, "learning_rate": 1.0235211267605633e-05, "loss": 0.4209, "step": 15040 }, { "epoch": 16.86, "learning_rate": 1.0233802816901408e-05, "loss": 0.5481, "step": 15041 }, { "epoch": 16.86, "learning_rate": 1.0232394366197184e-05, "loss": 0.5413, "step": 15042 }, { "epoch": 16.86, "learning_rate": 1.0230985915492959e-05, "loss": 0.4221, "step": 15043 }, { "epoch": 16.87, "learning_rate": 1.0229577464788733e-05, "loss": 0.4726, "step": 15044 }, { "epoch": 16.87, "learning_rate": 1.0228169014084508e-05, "loss": 0.3796, "step": 15045 }, { "epoch": 16.87, "learning_rate": 1.0226760563380282e-05, "loss": 0.479, "step": 15046 }, { "epoch": 16.87, "learning_rate": 1.0225352112676057e-05, "loss": 0.4386, "step": 15047 }, { "epoch": 16.87, "learning_rate": 1.022394366197183e-05, "loss": 0.5806, "step": 15048 }, { "epoch": 16.87, "learning_rate": 1.0222535211267605e-05, "loss": 0.4046, "step": 15049 }, { "epoch": 16.87, "learning_rate": 1.0221126760563381e-05, "loss": 0.4816, "step": 15050 }, { "epoch": 16.87, "learning_rate": 1.0219718309859156e-05, "loss": 0.6261, "step": 15051 }, { "epoch": 16.87, "learning_rate": 1.021830985915493e-05, "loss": 0.4023, "step": 15052 }, { "epoch": 16.88, "learning_rate": 1.0216901408450705e-05, "loss": 0.4665, "step": 15053 }, { "epoch": 16.88, "learning_rate": 1.021549295774648e-05, "loss": 0.4414, "step": 15054 }, { "epoch": 16.88, "learning_rate": 1.0214084507042254e-05, "loss": 0.4045, "step": 15055 }, { "epoch": 16.88, "learning_rate": 1.0212676056338027e-05, "loss": 0.5097, "step": 15056 }, { "epoch": 16.88, "learning_rate": 1.0211267605633803e-05, "loss": 0.4561, "step": 15057 }, { "epoch": 16.88, "learning_rate": 1.0209859154929578e-05, "loss": 0.395, "step": 15058 }, { "epoch": 16.88, "learning_rate": 1.0208450704225353e-05, "loss": 0.4367, "step": 15059 }, { "epoch": 16.88, "learning_rate": 1.0207042253521127e-05, "loss": 0.3266, "step": 15060 }, { "epoch": 16.88, "learning_rate": 1.0205633802816902e-05, "loss": 0.3914, "step": 15061 }, { "epoch": 16.89, "learning_rate": 1.0204225352112677e-05, "loss": 0.4369, "step": 15062 }, { "epoch": 16.89, "learning_rate": 1.020281690140845e-05, "loss": 0.377, "step": 15063 }, { "epoch": 16.89, "learning_rate": 1.0201408450704224e-05, "loss": 0.4541, "step": 15064 }, { "epoch": 16.89, "learning_rate": 1.02e-05, "loss": 0.3324, "step": 15065 }, { "epoch": 16.89, "learning_rate": 1.0198591549295775e-05, "loss": 0.4454, "step": 15066 }, { "epoch": 16.89, "learning_rate": 1.019718309859155e-05, "loss": 0.4256, "step": 15067 }, { "epoch": 16.89, "learning_rate": 1.0195774647887325e-05, "loss": 0.4755, "step": 15068 }, { "epoch": 16.89, "learning_rate": 1.0194366197183099e-05, "loss": 0.4408, "step": 15069 }, { "epoch": 16.89, "learning_rate": 1.0192957746478874e-05, "loss": 0.3606, "step": 15070 }, { "epoch": 16.9, "learning_rate": 1.0191549295774647e-05, "loss": 0.3976, "step": 15071 }, { "epoch": 16.9, "learning_rate": 1.0190140845070421e-05, "loss": 0.3368, "step": 15072 }, { "epoch": 16.9, "learning_rate": 1.0188732394366198e-05, "loss": 0.6706, "step": 15073 }, { "epoch": 16.9, "learning_rate": 1.0187323943661972e-05, "loss": 0.6562, "step": 15074 }, { "epoch": 16.9, "learning_rate": 1.0185915492957747e-05, "loss": 0.6858, "step": 15075 }, { "epoch": 16.9, "learning_rate": 1.0184507042253522e-05, "loss": 0.5616, "step": 15076 }, { "epoch": 16.9, "learning_rate": 1.0183098591549296e-05, "loss": 0.5651, "step": 15077 }, { "epoch": 16.9, "learning_rate": 1.0181690140845071e-05, "loss": 0.5875, "step": 15078 }, { "epoch": 16.9, "learning_rate": 1.0180281690140844e-05, "loss": 0.5114, "step": 15079 }, { "epoch": 16.91, "learning_rate": 1.017887323943662e-05, "loss": 0.4487, "step": 15080 }, { "epoch": 16.91, "learning_rate": 1.0177464788732395e-05, "loss": 0.3344, "step": 15081 }, { "epoch": 16.91, "learning_rate": 1.017605633802817e-05, "loss": 0.4458, "step": 15082 }, { "epoch": 16.91, "learning_rate": 1.0174647887323944e-05, "loss": 0.4385, "step": 15083 }, { "epoch": 16.91, "learning_rate": 1.0173239436619719e-05, "loss": 0.4102, "step": 15084 }, { "epoch": 16.91, "learning_rate": 1.0171830985915493e-05, "loss": 0.4319, "step": 15085 }, { "epoch": 16.91, "learning_rate": 1.0170422535211268e-05, "loss": 0.5248, "step": 15086 }, { "epoch": 16.91, "learning_rate": 1.0169014084507041e-05, "loss": 0.408, "step": 15087 }, { "epoch": 16.91, "learning_rate": 1.0167605633802817e-05, "loss": 0.4894, "step": 15088 }, { "epoch": 16.92, "learning_rate": 1.0166197183098592e-05, "loss": 0.5965, "step": 15089 }, { "epoch": 16.92, "learning_rate": 1.0164788732394367e-05, "loss": 0.436, "step": 15090 }, { "epoch": 16.92, "learning_rate": 1.0163380281690141e-05, "loss": 0.3611, "step": 15091 }, { "epoch": 16.92, "learning_rate": 1.0161971830985916e-05, "loss": 0.5528, "step": 15092 }, { "epoch": 16.92, "learning_rate": 1.016056338028169e-05, "loss": 0.4494, "step": 15093 }, { "epoch": 16.92, "learning_rate": 1.0159154929577465e-05, "loss": 0.5503, "step": 15094 }, { "epoch": 16.92, "learning_rate": 1.015774647887324e-05, "loss": 0.4474, "step": 15095 }, { "epoch": 16.92, "learning_rate": 1.0156338028169014e-05, "loss": 0.3881, "step": 15096 }, { "epoch": 16.92, "learning_rate": 1.0154929577464789e-05, "loss": 0.4836, "step": 15097 }, { "epoch": 16.93, "learning_rate": 1.0153521126760564e-05, "loss": 0.46, "step": 15098 }, { "epoch": 16.93, "learning_rate": 1.0152112676056338e-05, "loss": 0.4009, "step": 15099 }, { "epoch": 16.93, "learning_rate": 1.0150704225352113e-05, "loss": 0.38, "step": 15100 }, { "epoch": 16.93, "learning_rate": 1.0149295774647888e-05, "loss": 0.5379, "step": 15101 }, { "epoch": 16.93, "learning_rate": 1.0147887323943662e-05, "loss": 0.4049, "step": 15102 }, { "epoch": 16.93, "learning_rate": 1.0146478873239437e-05, "loss": 0.4886, "step": 15103 }, { "epoch": 16.93, "learning_rate": 1.0145070422535211e-05, "loss": 0.4112, "step": 15104 }, { "epoch": 16.93, "learning_rate": 1.0143661971830986e-05, "loss": 0.3847, "step": 15105 }, { "epoch": 16.93, "learning_rate": 1.014225352112676e-05, "loss": 0.5703, "step": 15106 }, { "epoch": 16.94, "learning_rate": 1.0140845070422535e-05, "loss": 0.4458, "step": 15107 }, { "epoch": 16.94, "learning_rate": 1.013943661971831e-05, "loss": 0.4174, "step": 15108 }, { "epoch": 16.94, "learning_rate": 1.0138028169014085e-05, "loss": 0.4228, "step": 15109 }, { "epoch": 16.94, "learning_rate": 1.0136619718309861e-05, "loss": 0.4506, "step": 15110 }, { "epoch": 16.94, "learning_rate": 1.0135211267605634e-05, "loss": 0.4807, "step": 15111 }, { "epoch": 16.94, "learning_rate": 1.0133802816901409e-05, "loss": 0.4089, "step": 15112 }, { "epoch": 16.94, "learning_rate": 1.0132394366197183e-05, "loss": 0.3519, "step": 15113 }, { "epoch": 16.94, "learning_rate": 1.0130985915492958e-05, "loss": 0.4241, "step": 15114 }, { "epoch": 16.95, "learning_rate": 1.0129577464788732e-05, "loss": 0.4399, "step": 15115 }, { "epoch": 16.95, "learning_rate": 1.0128169014084507e-05, "loss": 0.4248, "step": 15116 }, { "epoch": 16.95, "learning_rate": 1.0126760563380282e-05, "loss": 0.3494, "step": 15117 }, { "epoch": 16.95, "learning_rate": 1.0125352112676058e-05, "loss": 0.3973, "step": 15118 }, { "epoch": 16.95, "learning_rate": 1.0123943661971831e-05, "loss": 0.3788, "step": 15119 }, { "epoch": 16.95, "learning_rate": 1.0122535211267606e-05, "loss": 0.343, "step": 15120 }, { "epoch": 16.95, "learning_rate": 1.012112676056338e-05, "loss": 0.3896, "step": 15121 }, { "epoch": 16.95, "learning_rate": 1.0119718309859155e-05, "loss": 0.6705, "step": 15122 }, { "epoch": 16.95, "learning_rate": 1.011830985915493e-05, "loss": 0.6014, "step": 15123 }, { "epoch": 16.96, "learning_rate": 1.0116901408450704e-05, "loss": 0.5438, "step": 15124 }, { "epoch": 16.96, "learning_rate": 1.0115492957746479e-05, "loss": 0.5159, "step": 15125 }, { "epoch": 16.96, "learning_rate": 1.0114084507042255e-05, "loss": 0.4872, "step": 15126 }, { "epoch": 16.96, "learning_rate": 1.0112676056338028e-05, "loss": 0.5524, "step": 15127 }, { "epoch": 16.96, "learning_rate": 1.0111267605633803e-05, "loss": 0.4382, "step": 15128 }, { "epoch": 16.96, "learning_rate": 1.0109859154929577e-05, "loss": 0.4978, "step": 15129 }, { "epoch": 16.96, "learning_rate": 1.0108450704225352e-05, "loss": 0.4463, "step": 15130 }, { "epoch": 16.96, "learning_rate": 1.0107042253521127e-05, "loss": 0.4977, "step": 15131 }, { "epoch": 16.96, "learning_rate": 1.0105633802816901e-05, "loss": 0.47, "step": 15132 }, { "epoch": 16.97, "learning_rate": 1.0104225352112678e-05, "loss": 0.553, "step": 15133 }, { "epoch": 16.97, "learning_rate": 1.0102816901408452e-05, "loss": 0.5411, "step": 15134 }, { "epoch": 16.97, "learning_rate": 1.0101408450704225e-05, "loss": 0.4499, "step": 15135 }, { "epoch": 16.97, "learning_rate": 1.01e-05, "loss": 0.4546, "step": 15136 }, { "epoch": 16.97, "learning_rate": 1.0098591549295774e-05, "loss": 0.4511, "step": 15137 }, { "epoch": 16.97, "learning_rate": 1.0097183098591549e-05, "loss": 0.4302, "step": 15138 }, { "epoch": 16.97, "learning_rate": 1.0095774647887324e-05, "loss": 0.4631, "step": 15139 }, { "epoch": 16.97, "learning_rate": 1.0094366197183098e-05, "loss": 0.469, "step": 15140 }, { "epoch": 16.97, "learning_rate": 1.0092957746478875e-05, "loss": 0.5107, "step": 15141 }, { "epoch": 16.98, "learning_rate": 1.009154929577465e-05, "loss": 0.4855, "step": 15142 }, { "epoch": 16.98, "learning_rate": 1.0090140845070422e-05, "loss": 0.5002, "step": 15143 }, { "epoch": 16.98, "learning_rate": 1.0088732394366197e-05, "loss": 0.4501, "step": 15144 }, { "epoch": 16.98, "learning_rate": 1.0087323943661972e-05, "loss": 0.4368, "step": 15145 }, { "epoch": 16.98, "learning_rate": 1.0085915492957746e-05, "loss": 0.437, "step": 15146 }, { "epoch": 16.98, "learning_rate": 1.0084507042253521e-05, "loss": 0.3667, "step": 15147 }, { "epoch": 16.98, "learning_rate": 1.0083098591549297e-05, "loss": 0.4812, "step": 15148 }, { "epoch": 16.98, "learning_rate": 1.0081690140845072e-05, "loss": 0.5332, "step": 15149 }, { "epoch": 16.98, "learning_rate": 1.0080281690140846e-05, "loss": 0.4266, "step": 15150 }, { "epoch": 16.99, "learning_rate": 1.007887323943662e-05, "loss": 0.3819, "step": 15151 }, { "epoch": 16.99, "learning_rate": 1.0077464788732394e-05, "loss": 0.4247, "step": 15152 }, { "epoch": 16.99, "learning_rate": 1.0076056338028169e-05, "loss": 0.5218, "step": 15153 }, { "epoch": 16.99, "learning_rate": 1.0074647887323943e-05, "loss": 0.4141, "step": 15154 }, { "epoch": 16.99, "learning_rate": 1.0073239436619718e-05, "loss": 0.4656, "step": 15155 }, { "epoch": 16.99, "learning_rate": 1.0071830985915494e-05, "loss": 0.4626, "step": 15156 }, { "epoch": 16.99, "learning_rate": 1.0070422535211269e-05, "loss": 0.5607, "step": 15157 }, { "epoch": 16.99, "learning_rate": 1.0069014084507044e-05, "loss": 0.4159, "step": 15158 }, { "epoch": 16.99, "learning_rate": 1.0067605633802817e-05, "loss": 0.4284, "step": 15159 }, { "epoch": 17.0, "learning_rate": 1.0066197183098591e-05, "loss": 0.4073, "step": 15160 }, { "epoch": 17.0, "learning_rate": 1.0064788732394366e-05, "loss": 0.5079, "step": 15161 }, { "epoch": 17.0, "learning_rate": 1.006338028169014e-05, "loss": 0.3639, "step": 15162 }, { "epoch": 17.0, "learning_rate": 1.0061971830985917e-05, "loss": 0.4189, "step": 15163 }, { "epoch": 17.0, "learning_rate": 1.0060563380281691e-05, "loss": 0.4217, "step": 15164 }, { "epoch": 17.0, "learning_rate": 1.0059154929577466e-05, "loss": 0.5873, "step": 15165 }, { "epoch": 17.0, "learning_rate": 1.005774647887324e-05, "loss": 0.5076, "step": 15166 }, { "epoch": 17.0, "learning_rate": 1.0056338028169014e-05, "loss": 0.5523, "step": 15167 }, { "epoch": 17.0, "learning_rate": 1.0054929577464788e-05, "loss": 0.5195, "step": 15168 }, { "epoch": 17.01, "learning_rate": 1.0053521126760563e-05, "loss": 0.5341, "step": 15169 }, { "epoch": 17.01, "learning_rate": 1.0052112676056338e-05, "loss": 0.4846, "step": 15170 }, { "epoch": 17.01, "learning_rate": 1.0050704225352114e-05, "loss": 0.4723, "step": 15171 }, { "epoch": 17.01, "learning_rate": 1.0049295774647889e-05, "loss": 0.5228, "step": 15172 }, { "epoch": 17.01, "learning_rate": 1.0047887323943663e-05, "loss": 0.4687, "step": 15173 }, { "epoch": 17.01, "learning_rate": 1.0046478873239436e-05, "loss": 0.3779, "step": 15174 }, { "epoch": 17.01, "learning_rate": 1.004507042253521e-05, "loss": 0.5008, "step": 15175 }, { "epoch": 17.01, "learning_rate": 1.0043661971830985e-05, "loss": 0.4261, "step": 15176 }, { "epoch": 17.01, "learning_rate": 1.004225352112676e-05, "loss": 0.4204, "step": 15177 }, { "epoch": 17.02, "learning_rate": 1.0040845070422535e-05, "loss": 0.379, "step": 15178 }, { "epoch": 17.02, "learning_rate": 1.0039436619718311e-05, "loss": 0.4711, "step": 15179 }, { "epoch": 17.02, "learning_rate": 1.0038028169014086e-05, "loss": 0.3526, "step": 15180 }, { "epoch": 17.02, "learning_rate": 1.003661971830986e-05, "loss": 0.4213, "step": 15181 }, { "epoch": 17.02, "learning_rate": 1.0035211267605633e-05, "loss": 0.4258, "step": 15182 }, { "epoch": 17.02, "learning_rate": 1.0033802816901408e-05, "loss": 0.3856, "step": 15183 }, { "epoch": 17.02, "learning_rate": 1.0032394366197182e-05, "loss": 0.5359, "step": 15184 }, { "epoch": 17.02, "learning_rate": 1.0030985915492957e-05, "loss": 0.5488, "step": 15185 }, { "epoch": 17.02, "learning_rate": 1.0029577464788733e-05, "loss": 0.3551, "step": 15186 }, { "epoch": 17.03, "learning_rate": 1.0028169014084508e-05, "loss": 0.379, "step": 15187 }, { "epoch": 17.03, "learning_rate": 1.0026760563380283e-05, "loss": 0.4569, "step": 15188 }, { "epoch": 17.03, "learning_rate": 1.0025352112676057e-05, "loss": 0.4519, "step": 15189 }, { "epoch": 17.03, "learning_rate": 1.002394366197183e-05, "loss": 0.5588, "step": 15190 }, { "epoch": 17.03, "learning_rate": 1.0022535211267605e-05, "loss": 0.4857, "step": 15191 }, { "epoch": 17.03, "learning_rate": 1.002112676056338e-05, "loss": 0.3921, "step": 15192 }, { "epoch": 17.03, "learning_rate": 1.0019718309859154e-05, "loss": 0.4735, "step": 15193 }, { "epoch": 17.03, "learning_rate": 1.001830985915493e-05, "loss": 0.5424, "step": 15194 }, { "epoch": 17.03, "learning_rate": 1.0016901408450705e-05, "loss": 0.428, "step": 15195 }, { "epoch": 17.04, "learning_rate": 1.001549295774648e-05, "loss": 0.4051, "step": 15196 }, { "epoch": 17.04, "learning_rate": 1.0014084507042254e-05, "loss": 0.4563, "step": 15197 }, { "epoch": 17.04, "learning_rate": 1.0012676056338027e-05, "loss": 0.4164, "step": 15198 }, { "epoch": 17.04, "learning_rate": 1.0011267605633802e-05, "loss": 0.4005, "step": 15199 }, { "epoch": 17.04, "learning_rate": 1.0009859154929577e-05, "loss": 0.3972, "step": 15200 }, { "epoch": 17.04, "learning_rate": 1.0008450704225353e-05, "loss": 0.3874, "step": 15201 }, { "epoch": 17.04, "learning_rate": 1.0007042253521128e-05, "loss": 0.4431, "step": 15202 }, { "epoch": 17.04, "learning_rate": 1.0005633802816902e-05, "loss": 0.3631, "step": 15203 }, { "epoch": 17.04, "learning_rate": 1.0004225352112677e-05, "loss": 0.324, "step": 15204 }, { "epoch": 17.05, "learning_rate": 1.0002816901408452e-05, "loss": 0.3847, "step": 15205 }, { "epoch": 17.05, "learning_rate": 1.0001408450704224e-05, "loss": 0.4482, "step": 15206 }, { "epoch": 17.05, "learning_rate": 9.999999999999999e-06, "loss": 0.3769, "step": 15207 }, { "epoch": 17.05, "learning_rate": 9.998591549295774e-06, "loss": 0.4362, "step": 15208 }, { "epoch": 17.05, "learning_rate": 9.99718309859155e-06, "loss": 0.4327, "step": 15209 }, { "epoch": 17.05, "learning_rate": 9.995774647887325e-06, "loss": 0.4076, "step": 15210 }, { "epoch": 17.05, "learning_rate": 9.9943661971831e-06, "loss": 0.4153, "step": 15211 }, { "epoch": 17.05, "learning_rate": 9.992957746478874e-06, "loss": 0.3866, "step": 15212 }, { "epoch": 17.05, "learning_rate": 9.991549295774649e-06, "loss": 0.3948, "step": 15213 }, { "epoch": 17.06, "learning_rate": 9.990140845070422e-06, "loss": 0.4258, "step": 15214 }, { "epoch": 17.06, "learning_rate": 9.988732394366196e-06, "loss": 0.5821, "step": 15215 }, { "epoch": 17.06, "learning_rate": 9.987323943661973e-06, "loss": 0.5126, "step": 15216 }, { "epoch": 17.06, "learning_rate": 9.985915492957747e-06, "loss": 0.4381, "step": 15217 }, { "epoch": 17.06, "learning_rate": 9.984507042253522e-06, "loss": 0.5029, "step": 15218 }, { "epoch": 17.06, "learning_rate": 9.983098591549296e-06, "loss": 0.5297, "step": 15219 }, { "epoch": 17.06, "learning_rate": 9.981690140845071e-06, "loss": 0.5064, "step": 15220 }, { "epoch": 17.06, "learning_rate": 9.980281690140846e-06, "loss": 0.4984, "step": 15221 }, { "epoch": 17.07, "learning_rate": 9.978873239436619e-06, "loss": 0.3964, "step": 15222 }, { "epoch": 17.07, "learning_rate": 9.977464788732393e-06, "loss": 0.37, "step": 15223 }, { "epoch": 17.07, "learning_rate": 9.97605633802817e-06, "loss": 0.5287, "step": 15224 }, { "epoch": 17.07, "learning_rate": 9.974647887323944e-06, "loss": 0.4176, "step": 15225 }, { "epoch": 17.07, "learning_rate": 9.973239436619719e-06, "loss": 0.5308, "step": 15226 }, { "epoch": 17.07, "learning_rate": 9.971830985915494e-06, "loss": 0.4447, "step": 15227 }, { "epoch": 17.07, "learning_rate": 9.970422535211268e-06, "loss": 0.3675, "step": 15228 }, { "epoch": 17.07, "learning_rate": 9.969014084507043e-06, "loss": 0.4484, "step": 15229 }, { "epoch": 17.07, "learning_rate": 9.967605633802816e-06, "loss": 0.4565, "step": 15230 }, { "epoch": 17.08, "learning_rate": 9.966197183098592e-06, "loss": 0.4129, "step": 15231 }, { "epoch": 17.08, "learning_rate": 9.964788732394367e-06, "loss": 0.4001, "step": 15232 }, { "epoch": 17.08, "learning_rate": 9.963380281690141e-06, "loss": 0.3936, "step": 15233 }, { "epoch": 17.08, "learning_rate": 9.961971830985916e-06, "loss": 0.4221, "step": 15234 }, { "epoch": 17.08, "learning_rate": 9.96056338028169e-06, "loss": 0.4565, "step": 15235 }, { "epoch": 17.08, "learning_rate": 9.959154929577465e-06, "loss": 0.4127, "step": 15236 }, { "epoch": 17.08, "learning_rate": 9.95774647887324e-06, "loss": 0.4197, "step": 15237 }, { "epoch": 17.08, "learning_rate": 9.956338028169013e-06, "loss": 0.3557, "step": 15238 }, { "epoch": 17.08, "learning_rate": 9.95492957746479e-06, "loss": 0.4361, "step": 15239 }, { "epoch": 17.09, "learning_rate": 9.953521126760564e-06, "loss": 0.4197, "step": 15240 }, { "epoch": 17.09, "learning_rate": 9.952112676056338e-06, "loss": 0.3976, "step": 15241 }, { "epoch": 17.09, "learning_rate": 9.950704225352113e-06, "loss": 0.379, "step": 15242 }, { "epoch": 17.09, "learning_rate": 9.949295774647888e-06, "loss": 0.3477, "step": 15243 }, { "epoch": 17.09, "learning_rate": 9.947887323943662e-06, "loss": 0.4137, "step": 15244 }, { "epoch": 17.09, "learning_rate": 9.946478873239437e-06, "loss": 0.4653, "step": 15245 }, { "epoch": 17.09, "learning_rate": 9.94507042253521e-06, "loss": 0.394, "step": 15246 }, { "epoch": 17.09, "learning_rate": 9.943661971830986e-06, "loss": 0.4165, "step": 15247 }, { "epoch": 17.09, "learning_rate": 9.942253521126761e-06, "loss": 0.3629, "step": 15248 }, { "epoch": 17.1, "learning_rate": 9.940845070422536e-06, "loss": 0.398, "step": 15249 }, { "epoch": 17.1, "learning_rate": 9.93943661971831e-06, "loss": 0.3892, "step": 15250 }, { "epoch": 17.1, "learning_rate": 9.938028169014085e-06, "loss": 0.3942, "step": 15251 }, { "epoch": 17.1, "learning_rate": 9.93661971830986e-06, "loss": 0.3853, "step": 15252 }, { "epoch": 17.1, "learning_rate": 9.935211267605634e-06, "loss": 0.488, "step": 15253 }, { "epoch": 17.1, "learning_rate": 9.933802816901409e-06, "loss": 0.38, "step": 15254 }, { "epoch": 17.1, "learning_rate": 9.932394366197183e-06, "loss": 0.3958, "step": 15255 }, { "epoch": 17.1, "learning_rate": 9.930985915492958e-06, "loss": 0.3731, "step": 15256 }, { "epoch": 17.1, "learning_rate": 9.929577464788733e-06, "loss": 0.384, "step": 15257 }, { "epoch": 17.11, "learning_rate": 9.928169014084507e-06, "loss": 0.3394, "step": 15258 }, { "epoch": 17.11, "learning_rate": 9.926760563380282e-06, "loss": 0.3729, "step": 15259 }, { "epoch": 17.11, "learning_rate": 9.925352112676057e-06, "loss": 0.4106, "step": 15260 }, { "epoch": 17.11, "learning_rate": 9.923943661971831e-06, "loss": 0.3288, "step": 15261 }, { "epoch": 17.11, "learning_rate": 9.922535211267606e-06, "loss": 0.3849, "step": 15262 }, { "epoch": 17.11, "learning_rate": 9.92112676056338e-06, "loss": 0.3554, "step": 15263 }, { "epoch": 17.11, "learning_rate": 9.919718309859155e-06, "loss": 0.4245, "step": 15264 }, { "epoch": 17.11, "learning_rate": 9.91830985915493e-06, "loss": 0.5698, "step": 15265 }, { "epoch": 17.11, "learning_rate": 9.916901408450704e-06, "loss": 0.4259, "step": 15266 }, { "epoch": 17.12, "learning_rate": 9.915492957746479e-06, "loss": 0.4063, "step": 15267 }, { "epoch": 17.12, "learning_rate": 9.914084507042254e-06, "loss": 0.5095, "step": 15268 }, { "epoch": 17.12, "learning_rate": 9.91267605633803e-06, "loss": 0.726, "step": 15269 }, { "epoch": 17.12, "learning_rate": 9.911267605633803e-06, "loss": 0.4206, "step": 15270 }, { "epoch": 17.12, "learning_rate": 9.909859154929578e-06, "loss": 0.5254, "step": 15271 }, { "epoch": 17.12, "learning_rate": 9.908450704225352e-06, "loss": 0.5218, "step": 15272 }, { "epoch": 17.12, "learning_rate": 9.907042253521127e-06, "loss": 0.5333, "step": 15273 }, { "epoch": 17.12, "learning_rate": 9.905633802816902e-06, "loss": 0.447, "step": 15274 }, { "epoch": 17.12, "learning_rate": 9.904225352112676e-06, "loss": 0.3911, "step": 15275 }, { "epoch": 17.13, "learning_rate": 9.90281690140845e-06, "loss": 0.4796, "step": 15276 }, { "epoch": 17.13, "learning_rate": 9.901408450704225e-06, "loss": 0.5298, "step": 15277 }, { "epoch": 17.13, "learning_rate": 9.9e-06, "loss": 0.511, "step": 15278 }, { "epoch": 17.13, "learning_rate": 9.898591549295775e-06, "loss": 0.4784, "step": 15279 }, { "epoch": 17.13, "learning_rate": 9.89718309859155e-06, "loss": 0.4098, "step": 15280 }, { "epoch": 17.13, "learning_rate": 9.895774647887324e-06, "loss": 0.4955, "step": 15281 }, { "epoch": 17.13, "learning_rate": 9.894366197183099e-06, "loss": 0.5736, "step": 15282 }, { "epoch": 17.13, "learning_rate": 9.892957746478873e-06, "loss": 0.3703, "step": 15283 }, { "epoch": 17.13, "learning_rate": 9.89154929577465e-06, "loss": 0.4998, "step": 15284 }, { "epoch": 17.14, "learning_rate": 9.890140845070423e-06, "loss": 0.4634, "step": 15285 }, { "epoch": 17.14, "learning_rate": 9.888732394366197e-06, "loss": 0.4559, "step": 15286 }, { "epoch": 17.14, "learning_rate": 9.887323943661972e-06, "loss": 0.4602, "step": 15287 }, { "epoch": 17.14, "learning_rate": 9.885915492957746e-06, "loss": 0.3554, "step": 15288 }, { "epoch": 17.14, "learning_rate": 9.884507042253521e-06, "loss": 0.4608, "step": 15289 }, { "epoch": 17.14, "learning_rate": 9.883098591549296e-06, "loss": 0.3896, "step": 15290 }, { "epoch": 17.14, "learning_rate": 9.88169014084507e-06, "loss": 0.3553, "step": 15291 }, { "epoch": 17.14, "learning_rate": 9.880281690140847e-06, "loss": 0.3154, "step": 15292 }, { "epoch": 17.14, "learning_rate": 9.87887323943662e-06, "loss": 0.441, "step": 15293 }, { "epoch": 17.15, "learning_rate": 9.877464788732394e-06, "loss": 0.4118, "step": 15294 }, { "epoch": 17.15, "learning_rate": 9.876056338028169e-06, "loss": 0.3183, "step": 15295 }, { "epoch": 17.15, "learning_rate": 9.874647887323944e-06, "loss": 0.4, "step": 15296 }, { "epoch": 17.15, "learning_rate": 9.873239436619718e-06, "loss": 0.3468, "step": 15297 }, { "epoch": 17.15, "learning_rate": 9.871830985915493e-06, "loss": 0.4439, "step": 15298 }, { "epoch": 17.15, "learning_rate": 9.870422535211267e-06, "loss": 0.4538, "step": 15299 }, { "epoch": 17.15, "learning_rate": 9.869014084507044e-06, "loss": 0.3996, "step": 15300 }, { "epoch": 17.15, "learning_rate": 9.867605633802817e-06, "loss": 0.4389, "step": 15301 }, { "epoch": 17.15, "learning_rate": 9.866197183098591e-06, "loss": 0.324, "step": 15302 }, { "epoch": 17.16, "learning_rate": 9.864788732394366e-06, "loss": 0.3387, "step": 15303 }, { "epoch": 17.16, "learning_rate": 9.86338028169014e-06, "loss": 0.338, "step": 15304 }, { "epoch": 17.16, "learning_rate": 9.861971830985915e-06, "loss": 0.4013, "step": 15305 }, { "epoch": 17.16, "learning_rate": 9.86056338028169e-06, "loss": 0.3754, "step": 15306 }, { "epoch": 17.16, "learning_rate": 9.859154929577466e-06, "loss": 0.4803, "step": 15307 }, { "epoch": 17.16, "learning_rate": 9.857746478873241e-06, "loss": 0.3705, "step": 15308 }, { "epoch": 17.16, "learning_rate": 9.856338028169014e-06, "loss": 0.3122, "step": 15309 }, { "epoch": 17.16, "learning_rate": 9.854929577464788e-06, "loss": 0.3985, "step": 15310 }, { "epoch": 17.16, "learning_rate": 9.853521126760563e-06, "loss": 0.3142, "step": 15311 }, { "epoch": 17.17, "learning_rate": 9.852112676056338e-06, "loss": 0.548, "step": 15312 }, { "epoch": 17.17, "learning_rate": 9.850704225352112e-06, "loss": 0.4386, "step": 15313 }, { "epoch": 17.17, "learning_rate": 9.849295774647887e-06, "loss": 0.3903, "step": 15314 }, { "epoch": 17.17, "learning_rate": 9.847887323943663e-06, "loss": 0.5848, "step": 15315 }, { "epoch": 17.17, "learning_rate": 9.846478873239438e-06, "loss": 0.4639, "step": 15316 }, { "epoch": 17.17, "learning_rate": 9.845070422535211e-06, "loss": 0.5519, "step": 15317 }, { "epoch": 17.17, "learning_rate": 9.843661971830986e-06, "loss": 0.4294, "step": 15318 }, { "epoch": 17.17, "learning_rate": 9.84225352112676e-06, "loss": 0.3973, "step": 15319 }, { "epoch": 17.17, "learning_rate": 9.840845070422535e-06, "loss": 0.457, "step": 15320 }, { "epoch": 17.18, "learning_rate": 9.83943661971831e-06, "loss": 0.3841, "step": 15321 }, { "epoch": 17.18, "learning_rate": 9.838028169014086e-06, "loss": 0.572, "step": 15322 }, { "epoch": 17.18, "learning_rate": 9.83661971830986e-06, "loss": 0.5028, "step": 15323 }, { "epoch": 17.18, "learning_rate": 9.835211267605635e-06, "loss": 0.3821, "step": 15324 }, { "epoch": 17.18, "learning_rate": 9.833802816901408e-06, "loss": 0.4544, "step": 15325 }, { "epoch": 17.18, "learning_rate": 9.832394366197183e-06, "loss": 0.4917, "step": 15326 }, { "epoch": 17.18, "learning_rate": 9.830985915492957e-06, "loss": 0.4988, "step": 15327 }, { "epoch": 17.18, "learning_rate": 9.829577464788732e-06, "loss": 0.4258, "step": 15328 }, { "epoch": 17.18, "learning_rate": 9.828169014084507e-06, "loss": 0.4922, "step": 15329 }, { "epoch": 17.19, "learning_rate": 9.826760563380283e-06, "loss": 0.4168, "step": 15330 }, { "epoch": 17.19, "learning_rate": 9.825352112676058e-06, "loss": 0.4008, "step": 15331 }, { "epoch": 17.19, "learning_rate": 9.823943661971832e-06, "loss": 0.4081, "step": 15332 }, { "epoch": 17.19, "learning_rate": 9.822535211267605e-06, "loss": 0.428, "step": 15333 }, { "epoch": 17.19, "learning_rate": 9.82112676056338e-06, "loss": 0.4692, "step": 15334 }, { "epoch": 17.19, "learning_rate": 9.819718309859154e-06, "loss": 0.3118, "step": 15335 }, { "epoch": 17.19, "learning_rate": 9.818309859154929e-06, "loss": 0.466, "step": 15336 }, { "epoch": 17.19, "learning_rate": 9.816901408450705e-06, "loss": 0.431, "step": 15337 }, { "epoch": 17.2, "learning_rate": 9.81549295774648e-06, "loss": 0.3592, "step": 15338 }, { "epoch": 17.2, "learning_rate": 9.814084507042255e-06, "loss": 0.423, "step": 15339 }, { "epoch": 17.2, "learning_rate": 9.81267605633803e-06, "loss": 0.3528, "step": 15340 }, { "epoch": 17.2, "learning_rate": 9.811267605633802e-06, "loss": 0.4147, "step": 15341 }, { "epoch": 17.2, "learning_rate": 9.809859154929577e-06, "loss": 0.3564, "step": 15342 }, { "epoch": 17.2, "learning_rate": 9.808450704225352e-06, "loss": 0.4167, "step": 15343 }, { "epoch": 17.2, "learning_rate": 9.807042253521126e-06, "loss": 0.4144, "step": 15344 }, { "epoch": 17.2, "learning_rate": 9.805633802816902e-06, "loss": 0.5114, "step": 15345 }, { "epoch": 17.2, "learning_rate": 9.804225352112677e-06, "loss": 0.4951, "step": 15346 }, { "epoch": 17.21, "learning_rate": 9.802816901408452e-06, "loss": 0.3467, "step": 15347 }, { "epoch": 17.21, "learning_rate": 9.801408450704226e-06, "loss": 0.3581, "step": 15348 }, { "epoch": 17.21, "learning_rate": 9.8e-06, "loss": 0.3777, "step": 15349 }, { "epoch": 17.21, "learning_rate": 9.798591549295774e-06, "loss": 0.4698, "step": 15350 }, { "epoch": 17.21, "learning_rate": 9.797183098591549e-06, "loss": 0.2913, "step": 15351 }, { "epoch": 17.21, "learning_rate": 9.795774647887325e-06, "loss": 0.3303, "step": 15352 }, { "epoch": 17.21, "learning_rate": 9.7943661971831e-06, "loss": 0.4803, "step": 15353 }, { "epoch": 17.21, "learning_rate": 9.792957746478874e-06, "loss": 0.4963, "step": 15354 }, { "epoch": 17.21, "learning_rate": 9.791549295774649e-06, "loss": 0.3697, "step": 15355 }, { "epoch": 17.22, "learning_rate": 9.790140845070424e-06, "loss": 0.5616, "step": 15356 }, { "epoch": 17.22, "learning_rate": 9.788732394366196e-06, "loss": 0.2851, "step": 15357 }, { "epoch": 17.22, "learning_rate": 9.787323943661971e-06, "loss": 0.3552, "step": 15358 }, { "epoch": 17.22, "learning_rate": 9.785915492957746e-06, "loss": 0.3764, "step": 15359 }, { "epoch": 17.22, "learning_rate": 9.784507042253522e-06, "loss": 0.4413, "step": 15360 }, { "epoch": 17.22, "learning_rate": 9.783098591549297e-06, "loss": 0.3144, "step": 15361 }, { "epoch": 17.22, "learning_rate": 9.781690140845071e-06, "loss": 0.437, "step": 15362 }, { "epoch": 17.22, "learning_rate": 9.780281690140846e-06, "loss": 0.487, "step": 15363 }, { "epoch": 17.22, "learning_rate": 9.77887323943662e-06, "loss": 0.406, "step": 15364 }, { "epoch": 17.23, "learning_rate": 9.777464788732394e-06, "loss": 0.5609, "step": 15365 }, { "epoch": 17.23, "learning_rate": 9.776056338028168e-06, "loss": 0.4859, "step": 15366 }, { "epoch": 17.23, "learning_rate": 9.774647887323943e-06, "loss": 0.4471, "step": 15367 }, { "epoch": 17.23, "learning_rate": 9.773239436619719e-06, "loss": 0.5004, "step": 15368 }, { "epoch": 17.23, "learning_rate": 9.771830985915494e-06, "loss": 0.4364, "step": 15369 }, { "epoch": 17.23, "learning_rate": 9.770422535211268e-06, "loss": 0.5402, "step": 15370 }, { "epoch": 17.23, "learning_rate": 9.769014084507043e-06, "loss": 0.3863, "step": 15371 }, { "epoch": 17.23, "learning_rate": 9.767605633802818e-06, "loss": 0.5726, "step": 15372 }, { "epoch": 17.23, "learning_rate": 9.76619718309859e-06, "loss": 0.4282, "step": 15373 }, { "epoch": 17.24, "learning_rate": 9.764788732394365e-06, "loss": 0.5691, "step": 15374 }, { "epoch": 17.24, "learning_rate": 9.763380281690142e-06, "loss": 0.3773, "step": 15375 }, { "epoch": 17.24, "learning_rate": 9.761971830985916e-06, "loss": 0.4931, "step": 15376 }, { "epoch": 17.24, "learning_rate": 9.760563380281691e-06, "loss": 0.4765, "step": 15377 }, { "epoch": 17.24, "learning_rate": 9.759154929577466e-06, "loss": 0.4507, "step": 15378 }, { "epoch": 17.24, "learning_rate": 9.75774647887324e-06, "loss": 0.4539, "step": 15379 }, { "epoch": 17.24, "learning_rate": 9.756338028169015e-06, "loss": 0.4634, "step": 15380 }, { "epoch": 17.24, "learning_rate": 9.754929577464788e-06, "loss": 0.5164, "step": 15381 }, { "epoch": 17.24, "learning_rate": 9.753521126760562e-06, "loss": 0.3298, "step": 15382 }, { "epoch": 17.25, "learning_rate": 9.752112676056339e-06, "loss": 0.469, "step": 15383 }, { "epoch": 17.25, "learning_rate": 9.750704225352113e-06, "loss": 0.4855, "step": 15384 }, { "epoch": 17.25, "learning_rate": 9.749295774647888e-06, "loss": 0.4198, "step": 15385 }, { "epoch": 17.25, "learning_rate": 9.747887323943663e-06, "loss": 0.4649, "step": 15386 }, { "epoch": 17.25, "learning_rate": 9.746478873239437e-06, "loss": 0.3688, "step": 15387 }, { "epoch": 17.25, "learning_rate": 9.745070422535212e-06, "loss": 0.4075, "step": 15388 }, { "epoch": 17.25, "learning_rate": 9.743661971830985e-06, "loss": 0.417, "step": 15389 }, { "epoch": 17.25, "learning_rate": 9.742253521126761e-06, "loss": 0.3866, "step": 15390 }, { "epoch": 17.25, "learning_rate": 9.740845070422536e-06, "loss": 0.3557, "step": 15391 }, { "epoch": 17.26, "learning_rate": 9.73943661971831e-06, "loss": 0.4054, "step": 15392 }, { "epoch": 17.26, "learning_rate": 9.738028169014085e-06, "loss": 0.4691, "step": 15393 }, { "epoch": 17.26, "learning_rate": 9.73661971830986e-06, "loss": 0.3826, "step": 15394 }, { "epoch": 17.26, "learning_rate": 9.735211267605634e-06, "loss": 0.4074, "step": 15395 }, { "epoch": 17.26, "learning_rate": 9.733802816901409e-06, "loss": 0.4773, "step": 15396 }, { "epoch": 17.26, "learning_rate": 9.732394366197182e-06, "loss": 0.4631, "step": 15397 }, { "epoch": 17.26, "learning_rate": 9.730985915492958e-06, "loss": 0.3839, "step": 15398 }, { "epoch": 17.26, "learning_rate": 9.729577464788733e-06, "loss": 0.4234, "step": 15399 }, { "epoch": 17.26, "learning_rate": 9.728169014084508e-06, "loss": 0.4478, "step": 15400 }, { "epoch": 17.27, "learning_rate": 9.726760563380282e-06, "loss": 0.3709, "step": 15401 }, { "epoch": 17.27, "learning_rate": 9.725352112676057e-06, "loss": 0.3877, "step": 15402 }, { "epoch": 17.27, "learning_rate": 9.723943661971831e-06, "loss": 0.338, "step": 15403 }, { "epoch": 17.27, "learning_rate": 9.722535211267604e-06, "loss": 0.38, "step": 15404 }, { "epoch": 17.27, "learning_rate": 9.72112676056338e-06, "loss": 0.4327, "step": 15405 }, { "epoch": 17.27, "learning_rate": 9.719718309859155e-06, "loss": 0.3824, "step": 15406 }, { "epoch": 17.27, "learning_rate": 9.71830985915493e-06, "loss": 0.4097, "step": 15407 }, { "epoch": 17.27, "learning_rate": 9.716901408450705e-06, "loss": 0.3992, "step": 15408 }, { "epoch": 17.27, "learning_rate": 9.71549295774648e-06, "loss": 0.3302, "step": 15409 }, { "epoch": 17.28, "learning_rate": 9.714084507042254e-06, "loss": 0.2877, "step": 15410 }, { "epoch": 17.28, "learning_rate": 9.712676056338029e-06, "loss": 0.316, "step": 15411 }, { "epoch": 17.28, "learning_rate": 9.711267605633802e-06, "loss": 0.4708, "step": 15412 }, { "epoch": 17.28, "learning_rate": 9.709859154929578e-06, "loss": 0.4154, "step": 15413 }, { "epoch": 17.28, "learning_rate": 9.708450704225352e-06, "loss": 0.3154, "step": 15414 }, { "epoch": 17.28, "learning_rate": 9.707042253521127e-06, "loss": 0.5258, "step": 15415 }, { "epoch": 17.28, "learning_rate": 9.705633802816902e-06, "loss": 0.5677, "step": 15416 }, { "epoch": 17.28, "learning_rate": 9.704225352112676e-06, "loss": 0.5279, "step": 15417 }, { "epoch": 17.28, "learning_rate": 9.702816901408451e-06, "loss": 0.5025, "step": 15418 }, { "epoch": 17.29, "learning_rate": 9.701408450704226e-06, "loss": 0.4149, "step": 15419 }, { "epoch": 17.29, "learning_rate": 9.699999999999999e-06, "loss": 0.4724, "step": 15420 }, { "epoch": 17.29, "learning_rate": 9.698591549295775e-06, "loss": 0.4963, "step": 15421 }, { "epoch": 17.29, "learning_rate": 9.69718309859155e-06, "loss": 0.5047, "step": 15422 }, { "epoch": 17.29, "learning_rate": 9.695774647887324e-06, "loss": 0.3845, "step": 15423 }, { "epoch": 17.29, "learning_rate": 9.694366197183099e-06, "loss": 0.4233, "step": 15424 }, { "epoch": 17.29, "learning_rate": 9.692957746478873e-06, "loss": 0.4666, "step": 15425 }, { "epoch": 17.29, "learning_rate": 9.691549295774648e-06, "loss": 0.4812, "step": 15426 }, { "epoch": 17.29, "learning_rate": 9.690140845070423e-06, "loss": 0.4996, "step": 15427 }, { "epoch": 17.3, "learning_rate": 9.688732394366197e-06, "loss": 0.4165, "step": 15428 }, { "epoch": 17.3, "learning_rate": 9.687323943661972e-06, "loss": 0.3703, "step": 15429 }, { "epoch": 17.3, "learning_rate": 9.685915492957747e-06, "loss": 0.3351, "step": 15430 }, { "epoch": 17.3, "learning_rate": 9.684507042253521e-06, "loss": 0.5535, "step": 15431 }, { "epoch": 17.3, "learning_rate": 9.683098591549296e-06, "loss": 0.4532, "step": 15432 }, { "epoch": 17.3, "learning_rate": 9.68169014084507e-06, "loss": 0.5261, "step": 15433 }, { "epoch": 17.3, "learning_rate": 9.680281690140845e-06, "loss": 0.4949, "step": 15434 }, { "epoch": 17.3, "learning_rate": 9.67887323943662e-06, "loss": 0.4271, "step": 15435 }, { "epoch": 17.3, "learning_rate": 9.677464788732395e-06, "loss": 0.4875, "step": 15436 }, { "epoch": 17.31, "learning_rate": 9.676056338028169e-06, "loss": 0.3782, "step": 15437 }, { "epoch": 17.31, "learning_rate": 9.674647887323944e-06, "loss": 0.4881, "step": 15438 }, { "epoch": 17.31, "learning_rate": 9.673239436619718e-06, "loss": 0.3873, "step": 15439 }, { "epoch": 17.31, "learning_rate": 9.671830985915493e-06, "loss": 0.3826, "step": 15440 }, { "epoch": 17.31, "learning_rate": 9.670422535211268e-06, "loss": 0.5362, "step": 15441 }, { "epoch": 17.31, "learning_rate": 9.669014084507042e-06, "loss": 0.4339, "step": 15442 }, { "epoch": 17.31, "learning_rate": 9.667605633802819e-06, "loss": 0.4399, "step": 15443 }, { "epoch": 17.31, "learning_rate": 9.666197183098592e-06, "loss": 0.4321, "step": 15444 }, { "epoch": 17.32, "learning_rate": 9.664788732394366e-06, "loss": 0.4078, "step": 15445 }, { "epoch": 17.32, "learning_rate": 9.663380281690141e-06, "loss": 0.5011, "step": 15446 }, { "epoch": 17.32, "learning_rate": 9.661971830985916e-06, "loss": 0.4463, "step": 15447 }, { "epoch": 17.32, "learning_rate": 9.66056338028169e-06, "loss": 0.3454, "step": 15448 }, { "epoch": 17.32, "learning_rate": 9.659154929577465e-06, "loss": 0.3786, "step": 15449 }, { "epoch": 17.32, "learning_rate": 9.65774647887324e-06, "loss": 0.4353, "step": 15450 }, { "epoch": 17.32, "learning_rate": 9.656338028169016e-06, "loss": 0.4111, "step": 15451 }, { "epoch": 17.32, "learning_rate": 9.654929577464789e-06, "loss": 0.4515, "step": 15452 }, { "epoch": 17.32, "learning_rate": 9.653521126760563e-06, "loss": 0.3339, "step": 15453 }, { "epoch": 17.33, "learning_rate": 9.652112676056338e-06, "loss": 0.4645, "step": 15454 }, { "epoch": 17.33, "learning_rate": 9.650704225352113e-06, "loss": 0.3678, "step": 15455 }, { "epoch": 17.33, "learning_rate": 9.649295774647887e-06, "loss": 0.4393, "step": 15456 }, { "epoch": 17.33, "learning_rate": 9.647887323943662e-06, "loss": 0.4493, "step": 15457 }, { "epoch": 17.33, "learning_rate": 9.646478873239438e-06, "loss": 0.3598, "step": 15458 }, { "epoch": 17.33, "learning_rate": 9.645070422535213e-06, "loss": 0.3778, "step": 15459 }, { "epoch": 17.33, "learning_rate": 9.643661971830986e-06, "loss": 0.4688, "step": 15460 }, { "epoch": 17.33, "learning_rate": 9.64225352112676e-06, "loss": 0.2884, "step": 15461 }, { "epoch": 17.33, "learning_rate": 9.640845070422535e-06, "loss": 0.4602, "step": 15462 }, { "epoch": 17.34, "learning_rate": 9.63943661971831e-06, "loss": 0.4164, "step": 15463 }, { "epoch": 17.34, "learning_rate": 9.638028169014084e-06, "loss": 0.384, "step": 15464 }, { "epoch": 17.34, "learning_rate": 9.636619718309859e-06, "loss": 0.8142, "step": 15465 }, { "epoch": 17.34, "learning_rate": 9.635211267605635e-06, "loss": 0.5791, "step": 15466 }, { "epoch": 17.34, "learning_rate": 9.63380281690141e-06, "loss": 0.5042, "step": 15467 }, { "epoch": 17.34, "learning_rate": 9.632394366197183e-06, "loss": 0.6571, "step": 15468 }, { "epoch": 17.34, "learning_rate": 9.630985915492958e-06, "loss": 0.5293, "step": 15469 }, { "epoch": 17.34, "learning_rate": 9.629577464788732e-06, "loss": 0.502, "step": 15470 }, { "epoch": 17.34, "learning_rate": 9.628169014084507e-06, "loss": 0.4642, "step": 15471 }, { "epoch": 17.35, "learning_rate": 9.626760563380281e-06, "loss": 0.4503, "step": 15472 }, { "epoch": 17.35, "learning_rate": 9.625352112676058e-06, "loss": 0.4449, "step": 15473 }, { "epoch": 17.35, "learning_rate": 9.623943661971832e-06, "loss": 0.4694, "step": 15474 }, { "epoch": 17.35, "learning_rate": 9.622535211267607e-06, "loss": 0.4922, "step": 15475 }, { "epoch": 17.35, "learning_rate": 9.62112676056338e-06, "loss": 0.4731, "step": 15476 }, { "epoch": 17.35, "learning_rate": 9.619718309859155e-06, "loss": 0.4177, "step": 15477 }, { "epoch": 17.35, "learning_rate": 9.61830985915493e-06, "loss": 0.5581, "step": 15478 }, { "epoch": 17.35, "learning_rate": 9.616901408450704e-06, "loss": 0.4244, "step": 15479 }, { "epoch": 17.35, "learning_rate": 9.615492957746479e-06, "loss": 0.5108, "step": 15480 }, { "epoch": 17.36, "learning_rate": 9.614084507042255e-06, "loss": 0.4673, "step": 15481 }, { "epoch": 17.36, "learning_rate": 9.61267605633803e-06, "loss": 0.4056, "step": 15482 }, { "epoch": 17.36, "learning_rate": 9.611267605633804e-06, "loss": 0.5102, "step": 15483 }, { "epoch": 17.36, "learning_rate": 9.609859154929577e-06, "loss": 0.4703, "step": 15484 }, { "epoch": 17.36, "learning_rate": 9.608450704225352e-06, "loss": 0.433, "step": 15485 }, { "epoch": 17.36, "learning_rate": 9.607042253521126e-06, "loss": 0.4839, "step": 15486 }, { "epoch": 17.36, "learning_rate": 9.605633802816901e-06, "loss": 0.49, "step": 15487 }, { "epoch": 17.36, "learning_rate": 9.604225352112676e-06, "loss": 0.5863, "step": 15488 }, { "epoch": 17.36, "learning_rate": 9.602816901408452e-06, "loss": 0.4489, "step": 15489 }, { "epoch": 17.37, "learning_rate": 9.601408450704227e-06, "loss": 0.4201, "step": 15490 }, { "epoch": 17.37, "learning_rate": 9.600000000000001e-06, "loss": 0.4115, "step": 15491 }, { "epoch": 17.37, "learning_rate": 9.598591549295774e-06, "loss": 0.499, "step": 15492 }, { "epoch": 17.37, "learning_rate": 9.597183098591549e-06, "loss": 0.4305, "step": 15493 }, { "epoch": 17.37, "learning_rate": 9.595774647887323e-06, "loss": 0.4962, "step": 15494 }, { "epoch": 17.37, "learning_rate": 9.594366197183098e-06, "loss": 0.5335, "step": 15495 }, { "epoch": 17.37, "learning_rate": 9.592957746478874e-06, "loss": 0.354, "step": 15496 }, { "epoch": 17.37, "learning_rate": 9.591549295774649e-06, "loss": 0.4666, "step": 15497 }, { "epoch": 17.37, "learning_rate": 9.590140845070424e-06, "loss": 0.3988, "step": 15498 }, { "epoch": 17.38, "learning_rate": 9.588732394366198e-06, "loss": 0.4405, "step": 15499 }, { "epoch": 17.38, "learning_rate": 9.587323943661971e-06, "loss": 0.4644, "step": 15500 }, { "epoch": 17.38, "learning_rate": 9.585915492957746e-06, "loss": 0.4948, "step": 15501 }, { "epoch": 17.38, "learning_rate": 9.58450704225352e-06, "loss": 0.4602, "step": 15502 }, { "epoch": 17.38, "learning_rate": 9.583098591549295e-06, "loss": 0.3986, "step": 15503 }, { "epoch": 17.38, "learning_rate": 9.581690140845072e-06, "loss": 0.5021, "step": 15504 }, { "epoch": 17.38, "learning_rate": 9.580281690140846e-06, "loss": 0.4216, "step": 15505 }, { "epoch": 17.38, "learning_rate": 9.57887323943662e-06, "loss": 0.4751, "step": 15506 }, { "epoch": 17.38, "learning_rate": 9.577464788732395e-06, "loss": 0.3879, "step": 15507 }, { "epoch": 17.39, "learning_rate": 9.576056338028168e-06, "loss": 0.4025, "step": 15508 }, { "epoch": 17.39, "learning_rate": 9.574647887323943e-06, "loss": 0.4953, "step": 15509 }, { "epoch": 17.39, "learning_rate": 9.573239436619718e-06, "loss": 0.3647, "step": 15510 }, { "epoch": 17.39, "learning_rate": 9.571830985915494e-06, "loss": 0.4013, "step": 15511 }, { "epoch": 17.39, "learning_rate": 9.570422535211269e-06, "loss": 0.3083, "step": 15512 }, { "epoch": 17.39, "learning_rate": 9.569014084507043e-06, "loss": 0.4471, "step": 15513 }, { "epoch": 17.39, "learning_rate": 9.567605633802818e-06, "loss": 0.3401, "step": 15514 }, { "epoch": 17.39, "learning_rate": 9.566197183098591e-06, "loss": 0.5278, "step": 15515 }, { "epoch": 17.39, "learning_rate": 9.564788732394366e-06, "loss": 0.4601, "step": 15516 }, { "epoch": 17.4, "learning_rate": 9.56338028169014e-06, "loss": 0.4577, "step": 15517 }, { "epoch": 17.4, "learning_rate": 9.561971830985915e-06, "loss": 0.4469, "step": 15518 }, { "epoch": 17.4, "learning_rate": 9.560563380281691e-06, "loss": 0.4082, "step": 15519 }, { "epoch": 17.4, "learning_rate": 9.559154929577466e-06, "loss": 0.612, "step": 15520 }, { "epoch": 17.4, "learning_rate": 9.55774647887324e-06, "loss": 0.6037, "step": 15521 }, { "epoch": 17.4, "learning_rate": 9.556338028169015e-06, "loss": 0.5025, "step": 15522 }, { "epoch": 17.4, "learning_rate": 9.554929577464788e-06, "loss": 0.4458, "step": 15523 }, { "epoch": 17.4, "learning_rate": 9.553521126760563e-06, "loss": 0.571, "step": 15524 }, { "epoch": 17.4, "learning_rate": 9.552112676056337e-06, "loss": 0.4368, "step": 15525 }, { "epoch": 17.41, "learning_rate": 9.550704225352114e-06, "loss": 0.4422, "step": 15526 }, { "epoch": 17.41, "learning_rate": 9.549295774647888e-06, "loss": 0.485, "step": 15527 }, { "epoch": 17.41, "learning_rate": 9.547887323943663e-06, "loss": 0.3752, "step": 15528 }, { "epoch": 17.41, "learning_rate": 9.546478873239437e-06, "loss": 0.4513, "step": 15529 }, { "epoch": 17.41, "learning_rate": 9.545070422535212e-06, "loss": 0.4053, "step": 15530 }, { "epoch": 17.41, "learning_rate": 9.543661971830985e-06, "loss": 0.4787, "step": 15531 }, { "epoch": 17.41, "learning_rate": 9.54225352112676e-06, "loss": 0.398, "step": 15532 }, { "epoch": 17.41, "learning_rate": 9.540845070422534e-06, "loss": 0.3768, "step": 15533 }, { "epoch": 17.41, "learning_rate": 9.53943661971831e-06, "loss": 0.4203, "step": 15534 }, { "epoch": 17.42, "learning_rate": 9.538028169014085e-06, "loss": 0.4367, "step": 15535 }, { "epoch": 17.42, "learning_rate": 9.53661971830986e-06, "loss": 0.4844, "step": 15536 }, { "epoch": 17.42, "learning_rate": 9.535211267605635e-06, "loss": 0.4431, "step": 15537 }, { "epoch": 17.42, "learning_rate": 9.53380281690141e-06, "loss": 0.5429, "step": 15538 }, { "epoch": 17.42, "learning_rate": 9.532394366197182e-06, "loss": 0.4079, "step": 15539 }, { "epoch": 17.42, "learning_rate": 9.530985915492957e-06, "loss": 0.5288, "step": 15540 }, { "epoch": 17.42, "learning_rate": 9.529577464788731e-06, "loss": 0.447, "step": 15541 }, { "epoch": 17.42, "learning_rate": 9.528169014084508e-06, "loss": 0.3155, "step": 15542 }, { "epoch": 17.42, "learning_rate": 9.526760563380282e-06, "loss": 0.5304, "step": 15543 }, { "epoch": 17.43, "learning_rate": 9.525352112676057e-06, "loss": 0.4265, "step": 15544 }, { "epoch": 17.43, "learning_rate": 9.523943661971832e-06, "loss": 0.4726, "step": 15545 }, { "epoch": 17.43, "learning_rate": 9.522535211267606e-06, "loss": 0.415, "step": 15546 }, { "epoch": 17.43, "learning_rate": 9.52112676056338e-06, "loss": 0.4028, "step": 15547 }, { "epoch": 17.43, "learning_rate": 9.519718309859154e-06, "loss": 0.43, "step": 15548 }, { "epoch": 17.43, "learning_rate": 9.51830985915493e-06, "loss": 0.5099, "step": 15549 }, { "epoch": 17.43, "learning_rate": 9.516901408450705e-06, "loss": 0.4607, "step": 15550 }, { "epoch": 17.43, "learning_rate": 9.51549295774648e-06, "loss": 0.4145, "step": 15551 }, { "epoch": 17.43, "learning_rate": 9.514084507042254e-06, "loss": 0.4507, "step": 15552 }, { "epoch": 17.44, "learning_rate": 9.512676056338029e-06, "loss": 0.4349, "step": 15553 }, { "epoch": 17.44, "learning_rate": 9.511267605633803e-06, "loss": 0.3595, "step": 15554 }, { "epoch": 17.44, "learning_rate": 9.509859154929576e-06, "loss": 0.4217, "step": 15555 }, { "epoch": 17.44, "learning_rate": 9.508450704225351e-06, "loss": 0.5394, "step": 15556 }, { "epoch": 17.44, "learning_rate": 9.507042253521127e-06, "loss": 0.3455, "step": 15557 }, { "epoch": 17.44, "learning_rate": 9.505633802816902e-06, "loss": 0.4022, "step": 15558 }, { "epoch": 17.44, "learning_rate": 9.504225352112677e-06, "loss": 0.3793, "step": 15559 }, { "epoch": 17.44, "learning_rate": 9.502816901408451e-06, "loss": 0.2296, "step": 15560 }, { "epoch": 17.45, "learning_rate": 9.501408450704226e-06, "loss": 0.2913, "step": 15561 }, { "epoch": 17.45, "learning_rate": 9.5e-06, "loss": 0.4594, "step": 15562 }, { "epoch": 17.45, "learning_rate": 9.498591549295773e-06, "loss": 0.3985, "step": 15563 }, { "epoch": 17.45, "learning_rate": 9.49718309859155e-06, "loss": 0.3278, "step": 15564 }, { "epoch": 17.45, "learning_rate": 9.495774647887324e-06, "loss": 0.6928, "step": 15565 }, { "epoch": 17.45, "learning_rate": 9.494366197183099e-06, "loss": 0.4877, "step": 15566 }, { "epoch": 17.45, "learning_rate": 9.492957746478874e-06, "loss": 0.513, "step": 15567 }, { "epoch": 17.45, "learning_rate": 9.491549295774648e-06, "loss": 0.4033, "step": 15568 }, { "epoch": 17.45, "learning_rate": 9.490140845070423e-06, "loss": 0.5542, "step": 15569 }, { "epoch": 17.46, "learning_rate": 9.488732394366198e-06, "loss": 0.4184, "step": 15570 }, { "epoch": 17.46, "learning_rate": 9.48732394366197e-06, "loss": 0.4231, "step": 15571 }, { "epoch": 17.46, "learning_rate": 9.485915492957747e-06, "loss": 0.5028, "step": 15572 }, { "epoch": 17.46, "learning_rate": 9.484507042253522e-06, "loss": 0.4542, "step": 15573 }, { "epoch": 17.46, "learning_rate": 9.483098591549296e-06, "loss": 0.4838, "step": 15574 }, { "epoch": 17.46, "learning_rate": 9.48169014084507e-06, "loss": 0.4675, "step": 15575 }, { "epoch": 17.46, "learning_rate": 9.480281690140845e-06, "loss": 0.4297, "step": 15576 }, { "epoch": 17.46, "learning_rate": 9.47887323943662e-06, "loss": 0.35, "step": 15577 }, { "epoch": 17.46, "learning_rate": 9.477464788732395e-06, "loss": 0.4557, "step": 15578 }, { "epoch": 17.47, "learning_rate": 9.47605633802817e-06, "loss": 0.3573, "step": 15579 }, { "epoch": 17.47, "learning_rate": 9.474647887323944e-06, "loss": 0.4153, "step": 15580 }, { "epoch": 17.47, "learning_rate": 9.473239436619719e-06, "loss": 0.43, "step": 15581 }, { "epoch": 17.47, "learning_rate": 9.471830985915493e-06, "loss": 0.5642, "step": 15582 }, { "epoch": 17.47, "learning_rate": 9.470422535211268e-06, "loss": 0.4433, "step": 15583 }, { "epoch": 17.47, "learning_rate": 9.469014084507043e-06, "loss": 0.422, "step": 15584 }, { "epoch": 17.47, "learning_rate": 9.467605633802817e-06, "loss": 0.4495, "step": 15585 }, { "epoch": 17.47, "learning_rate": 9.466197183098592e-06, "loss": 0.4989, "step": 15586 }, { "epoch": 17.47, "learning_rate": 9.464788732394366e-06, "loss": 0.3882, "step": 15587 }, { "epoch": 17.48, "learning_rate": 9.463380281690141e-06, "loss": 0.4148, "step": 15588 }, { "epoch": 17.48, "learning_rate": 9.461971830985916e-06, "loss": 0.3694, "step": 15589 }, { "epoch": 17.48, "learning_rate": 9.46056338028169e-06, "loss": 0.3608, "step": 15590 }, { "epoch": 17.48, "learning_rate": 9.459154929577465e-06, "loss": 0.4218, "step": 15591 }, { "epoch": 17.48, "learning_rate": 9.45774647887324e-06, "loss": 0.2979, "step": 15592 }, { "epoch": 17.48, "learning_rate": 9.456338028169014e-06, "loss": 0.4235, "step": 15593 }, { "epoch": 17.48, "learning_rate": 9.454929577464789e-06, "loss": 0.4592, "step": 15594 }, { "epoch": 17.48, "learning_rate": 9.453521126760564e-06, "loss": 0.4747, "step": 15595 }, { "epoch": 17.48, "learning_rate": 9.452112676056338e-06, "loss": 0.4084, "step": 15596 }, { "epoch": 17.49, "learning_rate": 9.450704225352113e-06, "loss": 0.3521, "step": 15597 }, { "epoch": 17.49, "learning_rate": 9.449295774647887e-06, "loss": 0.33, "step": 15598 }, { "epoch": 17.49, "learning_rate": 9.447887323943662e-06, "loss": 0.3724, "step": 15599 }, { "epoch": 17.49, "learning_rate": 9.446478873239437e-06, "loss": 0.4369, "step": 15600 }, { "epoch": 17.49, "learning_rate": 9.445070422535211e-06, "loss": 0.3967, "step": 15601 }, { "epoch": 17.49, "learning_rate": 9.443661971830988e-06, "loss": 0.4262, "step": 15602 }, { "epoch": 17.49, "learning_rate": 9.44225352112676e-06, "loss": 0.4436, "step": 15603 }, { "epoch": 17.49, "learning_rate": 9.440845070422535e-06, "loss": 0.4866, "step": 15604 }, { "epoch": 17.49, "learning_rate": 9.43943661971831e-06, "loss": 0.393, "step": 15605 }, { "epoch": 17.5, "learning_rate": 9.438028169014085e-06, "loss": 0.508, "step": 15606 }, { "epoch": 17.5, "learning_rate": 9.43661971830986e-06, "loss": 0.2861, "step": 15607 }, { "epoch": 17.5, "learning_rate": 9.435211267605634e-06, "loss": 0.432, "step": 15608 }, { "epoch": 17.5, "learning_rate": 9.433802816901408e-06, "loss": 0.4375, "step": 15609 }, { "epoch": 17.5, "learning_rate": 9.432394366197185e-06, "loss": 0.3355, "step": 15610 }, { "epoch": 17.5, "learning_rate": 9.430985915492958e-06, "loss": 0.4304, "step": 15611 }, { "epoch": 17.5, "learning_rate": 9.429577464788732e-06, "loss": 0.4013, "step": 15612 }, { "epoch": 17.5, "learning_rate": 9.428169014084507e-06, "loss": 0.3684, "step": 15613 }, { "epoch": 17.5, "learning_rate": 9.426760563380282e-06, "loss": 0.3983, "step": 15614 }, { "epoch": 17.51, "learning_rate": 9.425352112676056e-06, "loss": 0.6231, "step": 15615 }, { "epoch": 17.51, "learning_rate": 9.423943661971831e-06, "loss": 0.5407, "step": 15616 }, { "epoch": 17.51, "learning_rate": 9.422535211267607e-06, "loss": 0.5578, "step": 15617 }, { "epoch": 17.51, "learning_rate": 9.42112676056338e-06, "loss": 0.4734, "step": 15618 }, { "epoch": 17.51, "learning_rate": 9.419718309859155e-06, "loss": 0.3813, "step": 15619 }, { "epoch": 17.51, "learning_rate": 9.41830985915493e-06, "loss": 0.5335, "step": 15620 }, { "epoch": 17.51, "learning_rate": 9.416901408450704e-06, "loss": 0.4434, "step": 15621 }, { "epoch": 17.51, "learning_rate": 9.415492957746479e-06, "loss": 0.5365, "step": 15622 }, { "epoch": 17.51, "learning_rate": 9.414084507042253e-06, "loss": 0.4641, "step": 15623 }, { "epoch": 17.52, "learning_rate": 9.412676056338028e-06, "loss": 0.4238, "step": 15624 }, { "epoch": 17.52, "learning_rate": 9.411267605633804e-06, "loss": 0.3759, "step": 15625 }, { "epoch": 17.52, "learning_rate": 9.409859154929577e-06, "loss": 0.4653, "step": 15626 }, { "epoch": 17.52, "learning_rate": 9.408450704225352e-06, "loss": 0.4256, "step": 15627 }, { "epoch": 17.52, "learning_rate": 9.407042253521127e-06, "loss": 0.4515, "step": 15628 }, { "epoch": 17.52, "learning_rate": 9.405633802816901e-06, "loss": 0.4999, "step": 15629 }, { "epoch": 17.52, "learning_rate": 9.404225352112676e-06, "loss": 0.4766, "step": 15630 }, { "epoch": 17.52, "learning_rate": 9.40281690140845e-06, "loss": 0.4258, "step": 15631 }, { "epoch": 17.52, "learning_rate": 9.401408450704227e-06, "loss": 0.4019, "step": 15632 }, { "epoch": 17.53, "learning_rate": 9.400000000000001e-06, "loss": 0.4589, "step": 15633 }, { "epoch": 17.53, "learning_rate": 9.398591549295774e-06, "loss": 0.4482, "step": 15634 }, { "epoch": 17.53, "learning_rate": 9.397183098591549e-06, "loss": 0.4314, "step": 15635 }, { "epoch": 17.53, "learning_rate": 9.395774647887324e-06, "loss": 0.4573, "step": 15636 }, { "epoch": 17.53, "learning_rate": 9.394366197183098e-06, "loss": 0.3379, "step": 15637 }, { "epoch": 17.53, "learning_rate": 9.392957746478873e-06, "loss": 0.3235, "step": 15638 }, { "epoch": 17.53, "learning_rate": 9.391549295774648e-06, "loss": 0.3599, "step": 15639 }, { "epoch": 17.53, "learning_rate": 9.390140845070424e-06, "loss": 0.403, "step": 15640 }, { "epoch": 17.53, "learning_rate": 9.388732394366199e-06, "loss": 0.3956, "step": 15641 }, { "epoch": 17.54, "learning_rate": 9.387323943661972e-06, "loss": 0.4039, "step": 15642 }, { "epoch": 17.54, "learning_rate": 9.385915492957746e-06, "loss": 0.4665, "step": 15643 }, { "epoch": 17.54, "learning_rate": 9.38450704225352e-06, "loss": 0.3805, "step": 15644 }, { "epoch": 17.54, "learning_rate": 9.383098591549295e-06, "loss": 0.3237, "step": 15645 }, { "epoch": 17.54, "learning_rate": 9.38169014084507e-06, "loss": 0.4313, "step": 15646 }, { "epoch": 17.54, "learning_rate": 9.380281690140846e-06, "loss": 0.4381, "step": 15647 }, { "epoch": 17.54, "learning_rate": 9.378873239436621e-06, "loss": 0.4712, "step": 15648 }, { "epoch": 17.54, "learning_rate": 9.377464788732396e-06, "loss": 0.3616, "step": 15649 }, { "epoch": 17.54, "learning_rate": 9.376056338028169e-06, "loss": 0.5308, "step": 15650 }, { "epoch": 17.55, "learning_rate": 9.374647887323943e-06, "loss": 0.3566, "step": 15651 }, { "epoch": 17.55, "learning_rate": 9.373239436619718e-06, "loss": 0.3703, "step": 15652 }, { "epoch": 17.55, "learning_rate": 9.371830985915493e-06, "loss": 0.3736, "step": 15653 }, { "epoch": 17.55, "learning_rate": 9.370422535211267e-06, "loss": 0.3818, "step": 15654 }, { "epoch": 17.55, "learning_rate": 9.369014084507044e-06, "loss": 0.3809, "step": 15655 }, { "epoch": 17.55, "learning_rate": 9.367605633802818e-06, "loss": 0.361, "step": 15656 }, { "epoch": 17.55, "learning_rate": 9.366197183098593e-06, "loss": 0.3785, "step": 15657 }, { "epoch": 17.55, "learning_rate": 9.364788732394366e-06, "loss": 0.3638, "step": 15658 }, { "epoch": 17.55, "learning_rate": 9.36338028169014e-06, "loss": 0.4011, "step": 15659 }, { "epoch": 17.56, "learning_rate": 9.361971830985915e-06, "loss": 0.3347, "step": 15660 }, { "epoch": 17.56, "learning_rate": 9.36056338028169e-06, "loss": 0.4156, "step": 15661 }, { "epoch": 17.56, "learning_rate": 9.359154929577464e-06, "loss": 0.3581, "step": 15662 }, { "epoch": 17.56, "learning_rate": 9.35774647887324e-06, "loss": 0.3134, "step": 15663 }, { "epoch": 17.56, "learning_rate": 9.356338028169015e-06, "loss": 0.3936, "step": 15664 }, { "epoch": 17.56, "learning_rate": 9.35492957746479e-06, "loss": 0.5759, "step": 15665 }, { "epoch": 17.56, "learning_rate": 9.353521126760563e-06, "loss": 0.5014, "step": 15666 }, { "epoch": 17.56, "learning_rate": 9.352112676056337e-06, "loss": 0.5278, "step": 15667 }, { "epoch": 17.57, "learning_rate": 9.350704225352112e-06, "loss": 0.4083, "step": 15668 }, { "epoch": 17.57, "learning_rate": 9.349295774647887e-06, "loss": 0.4391, "step": 15669 }, { "epoch": 17.57, "learning_rate": 9.347887323943663e-06, "loss": 0.5418, "step": 15670 }, { "epoch": 17.57, "learning_rate": 9.346478873239438e-06, "loss": 0.4996, "step": 15671 }, { "epoch": 17.57, "learning_rate": 9.345070422535212e-06, "loss": 0.4232, "step": 15672 }, { "epoch": 17.57, "learning_rate": 9.343661971830987e-06, "loss": 0.4639, "step": 15673 }, { "epoch": 17.57, "learning_rate": 9.34225352112676e-06, "loss": 0.5223, "step": 15674 }, { "epoch": 17.57, "learning_rate": 9.340845070422535e-06, "loss": 0.3975, "step": 15675 }, { "epoch": 17.57, "learning_rate": 9.33943661971831e-06, "loss": 0.4016, "step": 15676 }, { "epoch": 17.58, "learning_rate": 9.338028169014084e-06, "loss": 0.4391, "step": 15677 }, { "epoch": 17.58, "learning_rate": 9.33661971830986e-06, "loss": 0.4526, "step": 15678 }, { "epoch": 17.58, "learning_rate": 9.335211267605635e-06, "loss": 0.4018, "step": 15679 }, { "epoch": 17.58, "learning_rate": 9.33380281690141e-06, "loss": 0.4358, "step": 15680 }, { "epoch": 17.58, "learning_rate": 9.332394366197184e-06, "loss": 0.3778, "step": 15681 }, { "epoch": 17.58, "learning_rate": 9.330985915492957e-06, "loss": 0.4251, "step": 15682 }, { "epoch": 17.58, "learning_rate": 9.329577464788732e-06, "loss": 0.5053, "step": 15683 }, { "epoch": 17.58, "learning_rate": 9.328169014084506e-06, "loss": 0.404, "step": 15684 }, { "epoch": 17.58, "learning_rate": 9.326760563380283e-06, "loss": 0.4262, "step": 15685 }, { "epoch": 17.59, "learning_rate": 9.325352112676057e-06, "loss": 0.4411, "step": 15686 }, { "epoch": 17.59, "learning_rate": 9.323943661971832e-06, "loss": 0.3886, "step": 15687 }, { "epoch": 17.59, "learning_rate": 9.322535211267607e-06, "loss": 0.4367, "step": 15688 }, { "epoch": 17.59, "learning_rate": 9.321126760563381e-06, "loss": 0.4183, "step": 15689 }, { "epoch": 17.59, "learning_rate": 9.319718309859154e-06, "loss": 0.4855, "step": 15690 }, { "epoch": 17.59, "learning_rate": 9.318309859154929e-06, "loss": 0.4578, "step": 15691 }, { "epoch": 17.59, "learning_rate": 9.316901408450703e-06, "loss": 0.4004, "step": 15692 }, { "epoch": 17.59, "learning_rate": 9.31549295774648e-06, "loss": 0.3966, "step": 15693 }, { "epoch": 17.59, "learning_rate": 9.314084507042254e-06, "loss": 0.4315, "step": 15694 }, { "epoch": 17.6, "learning_rate": 9.312676056338029e-06, "loss": 0.3949, "step": 15695 }, { "epoch": 17.6, "learning_rate": 9.311267605633804e-06, "loss": 0.398, "step": 15696 }, { "epoch": 17.6, "learning_rate": 9.309859154929578e-06, "loss": 0.4001, "step": 15697 }, { "epoch": 17.6, "learning_rate": 9.308450704225351e-06, "loss": 0.4915, "step": 15698 }, { "epoch": 17.6, "learning_rate": 9.307042253521126e-06, "loss": 0.6339, "step": 15699 }, { "epoch": 17.6, "learning_rate": 9.305633802816902e-06, "loss": 0.4189, "step": 15700 }, { "epoch": 17.6, "learning_rate": 9.304225352112677e-06, "loss": 0.3505, "step": 15701 }, { "epoch": 17.6, "learning_rate": 9.302816901408451e-06, "loss": 0.3928, "step": 15702 }, { "epoch": 17.6, "learning_rate": 9.301408450704226e-06, "loss": 0.4925, "step": 15703 }, { "epoch": 17.61, "learning_rate": 9.3e-06, "loss": 0.4095, "step": 15704 }, { "epoch": 17.61, "learning_rate": 9.298591549295775e-06, "loss": 0.4008, "step": 15705 }, { "epoch": 17.61, "learning_rate": 9.297183098591548e-06, "loss": 0.3955, "step": 15706 }, { "epoch": 17.61, "learning_rate": 9.295774647887323e-06, "loss": 0.4286, "step": 15707 }, { "epoch": 17.61, "learning_rate": 9.2943661971831e-06, "loss": 0.35, "step": 15708 }, { "epoch": 17.61, "learning_rate": 9.292957746478874e-06, "loss": 0.3446, "step": 15709 }, { "epoch": 17.61, "learning_rate": 9.291549295774649e-06, "loss": 0.3482, "step": 15710 }, { "epoch": 17.61, "learning_rate": 9.290140845070423e-06, "loss": 0.2891, "step": 15711 }, { "epoch": 17.61, "learning_rate": 9.288732394366198e-06, "loss": 0.2958, "step": 15712 }, { "epoch": 17.62, "learning_rate": 9.287323943661972e-06, "loss": 0.3496, "step": 15713 }, { "epoch": 17.62, "learning_rate": 9.285915492957745e-06, "loss": 0.2926, "step": 15714 }, { "epoch": 17.62, "learning_rate": 9.28450704225352e-06, "loss": 0.5952, "step": 15715 }, { "epoch": 17.62, "learning_rate": 9.283098591549296e-06, "loss": 0.543, "step": 15716 }, { "epoch": 17.62, "learning_rate": 9.281690140845071e-06, "loss": 0.4809, "step": 15717 }, { "epoch": 17.62, "learning_rate": 9.280281690140846e-06, "loss": 0.5351, "step": 15718 }, { "epoch": 17.62, "learning_rate": 9.27887323943662e-06, "loss": 0.4098, "step": 15719 }, { "epoch": 17.62, "learning_rate": 9.277464788732395e-06, "loss": 0.4764, "step": 15720 }, { "epoch": 17.62, "learning_rate": 9.27605633802817e-06, "loss": 0.4865, "step": 15721 }, { "epoch": 17.63, "learning_rate": 9.274647887323943e-06, "loss": 0.4468, "step": 15722 }, { "epoch": 17.63, "learning_rate": 9.273239436619719e-06, "loss": 0.484, "step": 15723 }, { "epoch": 17.63, "learning_rate": 9.271830985915493e-06, "loss": 0.4686, "step": 15724 }, { "epoch": 17.63, "learning_rate": 9.270422535211268e-06, "loss": 0.4527, "step": 15725 }, { "epoch": 17.63, "learning_rate": 9.269014084507043e-06, "loss": 0.5275, "step": 15726 }, { "epoch": 17.63, "learning_rate": 9.267605633802817e-06, "loss": 0.4248, "step": 15727 }, { "epoch": 17.63, "learning_rate": 9.266197183098592e-06, "loss": 0.4888, "step": 15728 }, { "epoch": 17.63, "learning_rate": 9.264788732394367e-06, "loss": 0.4575, "step": 15729 }, { "epoch": 17.63, "learning_rate": 9.26338028169014e-06, "loss": 0.4329, "step": 15730 }, { "epoch": 17.64, "learning_rate": 9.261971830985916e-06, "loss": 0.3828, "step": 15731 }, { "epoch": 17.64, "learning_rate": 9.26056338028169e-06, "loss": 0.4416, "step": 15732 }, { "epoch": 17.64, "learning_rate": 9.259154929577465e-06, "loss": 0.3828, "step": 15733 }, { "epoch": 17.64, "learning_rate": 9.25774647887324e-06, "loss": 0.4599, "step": 15734 }, { "epoch": 17.64, "learning_rate": 9.256338028169015e-06, "loss": 0.4543, "step": 15735 }, { "epoch": 17.64, "learning_rate": 9.254929577464789e-06, "loss": 0.3619, "step": 15736 }, { "epoch": 17.64, "learning_rate": 9.253521126760564e-06, "loss": 0.3382, "step": 15737 }, { "epoch": 17.64, "learning_rate": 9.252112676056338e-06, "loss": 0.3846, "step": 15738 }, { "epoch": 17.64, "learning_rate": 9.250704225352113e-06, "loss": 0.4741, "step": 15739 }, { "epoch": 17.65, "learning_rate": 9.249295774647888e-06, "loss": 0.3638, "step": 15740 }, { "epoch": 17.65, "learning_rate": 9.247887323943662e-06, "loss": 0.393, "step": 15741 }, { "epoch": 17.65, "learning_rate": 9.246478873239437e-06, "loss": 0.5225, "step": 15742 }, { "epoch": 17.65, "learning_rate": 9.245070422535212e-06, "loss": 0.43, "step": 15743 }, { "epoch": 17.65, "learning_rate": 9.243661971830986e-06, "loss": 0.4356, "step": 15744 }, { "epoch": 17.65, "learning_rate": 9.24225352112676e-06, "loss": 0.4427, "step": 15745 }, { "epoch": 17.65, "learning_rate": 9.240845070422536e-06, "loss": 0.3505, "step": 15746 }, { "epoch": 17.65, "learning_rate": 9.23943661971831e-06, "loss": 0.3842, "step": 15747 }, { "epoch": 17.65, "learning_rate": 9.238028169014085e-06, "loss": 0.4779, "step": 15748 }, { "epoch": 17.66, "learning_rate": 9.23661971830986e-06, "loss": 0.4183, "step": 15749 }, { "epoch": 17.66, "learning_rate": 9.235211267605634e-06, "loss": 0.4167, "step": 15750 }, { "epoch": 17.66, "learning_rate": 9.233802816901409e-06, "loss": 0.3909, "step": 15751 }, { "epoch": 17.66, "learning_rate": 9.232394366197183e-06, "loss": 0.4244, "step": 15752 }, { "epoch": 17.66, "learning_rate": 9.230985915492958e-06, "loss": 0.294, "step": 15753 }, { "epoch": 17.66, "learning_rate": 9.229577464788733e-06, "loss": 0.3779, "step": 15754 }, { "epoch": 17.66, "learning_rate": 9.228169014084507e-06, "loss": 0.3823, "step": 15755 }, { "epoch": 17.66, "learning_rate": 9.226760563380282e-06, "loss": 0.3108, "step": 15756 }, { "epoch": 17.66, "learning_rate": 9.225352112676057e-06, "loss": 0.4712, "step": 15757 }, { "epoch": 17.67, "learning_rate": 9.223943661971831e-06, "loss": 0.4097, "step": 15758 }, { "epoch": 17.67, "learning_rate": 9.222535211267606e-06, "loss": 0.3506, "step": 15759 }, { "epoch": 17.67, "learning_rate": 9.22112676056338e-06, "loss": 0.4673, "step": 15760 }, { "epoch": 17.67, "learning_rate": 9.219718309859155e-06, "loss": 0.3885, "step": 15761 }, { "epoch": 17.67, "learning_rate": 9.21830985915493e-06, "loss": 0.4797, "step": 15762 }, { "epoch": 17.67, "learning_rate": 9.216901408450704e-06, "loss": 0.3239, "step": 15763 }, { "epoch": 17.67, "learning_rate": 9.215492957746479e-06, "loss": 0.459, "step": 15764 }, { "epoch": 17.67, "learning_rate": 9.214084507042254e-06, "loss": 0.5263, "step": 15765 }, { "epoch": 17.67, "learning_rate": 9.212676056338028e-06, "loss": 0.5066, "step": 15766 }, { "epoch": 17.68, "learning_rate": 9.211267605633803e-06, "loss": 0.5152, "step": 15767 }, { "epoch": 17.68, "learning_rate": 9.20985915492958e-06, "loss": 0.6492, "step": 15768 }, { "epoch": 17.68, "learning_rate": 9.208450704225352e-06, "loss": 0.3558, "step": 15769 }, { "epoch": 17.68, "learning_rate": 9.207042253521127e-06, "loss": 0.4257, "step": 15770 }, { "epoch": 17.68, "learning_rate": 9.205633802816901e-06, "loss": 0.3931, "step": 15771 }, { "epoch": 17.68, "learning_rate": 9.204225352112676e-06, "loss": 0.442, "step": 15772 }, { "epoch": 17.68, "learning_rate": 9.20281690140845e-06, "loss": 0.4936, "step": 15773 }, { "epoch": 17.68, "learning_rate": 9.201408450704225e-06, "loss": 0.4678, "step": 15774 }, { "epoch": 17.68, "learning_rate": 9.2e-06, "loss": 0.5009, "step": 15775 }, { "epoch": 17.69, "learning_rate": 9.198591549295776e-06, "loss": 0.4021, "step": 15776 }, { "epoch": 17.69, "learning_rate": 9.19718309859155e-06, "loss": 0.4607, "step": 15777 }, { "epoch": 17.69, "learning_rate": 9.195774647887324e-06, "loss": 0.5013, "step": 15778 }, { "epoch": 17.69, "learning_rate": 9.194366197183099e-06, "loss": 0.3953, "step": 15779 }, { "epoch": 17.69, "learning_rate": 9.192957746478873e-06, "loss": 0.3707, "step": 15780 }, { "epoch": 17.69, "learning_rate": 9.191549295774648e-06, "loss": 0.5157, "step": 15781 }, { "epoch": 17.69, "learning_rate": 9.190140845070422e-06, "loss": 0.5401, "step": 15782 }, { "epoch": 17.69, "learning_rate": 9.188732394366197e-06, "loss": 0.42, "step": 15783 }, { "epoch": 17.7, "learning_rate": 9.187323943661973e-06, "loss": 0.424, "step": 15784 }, { "epoch": 17.7, "learning_rate": 9.185915492957746e-06, "loss": 0.4571, "step": 15785 }, { "epoch": 17.7, "learning_rate": 9.184507042253521e-06, "loss": 0.5451, "step": 15786 }, { "epoch": 17.7, "learning_rate": 9.183098591549296e-06, "loss": 0.3689, "step": 15787 }, { "epoch": 17.7, "learning_rate": 9.18169014084507e-06, "loss": 0.4031, "step": 15788 }, { "epoch": 17.7, "learning_rate": 9.180281690140845e-06, "loss": 0.4951, "step": 15789 }, { "epoch": 17.7, "learning_rate": 9.17887323943662e-06, "loss": 0.3011, "step": 15790 }, { "epoch": 17.7, "learning_rate": 9.177464788732396e-06, "loss": 0.4672, "step": 15791 }, { "epoch": 17.7, "learning_rate": 9.17605633802817e-06, "loss": 0.4136, "step": 15792 }, { "epoch": 17.71, "learning_rate": 9.174647887323943e-06, "loss": 0.3884, "step": 15793 }, { "epoch": 17.71, "learning_rate": 9.173239436619718e-06, "loss": 0.4594, "step": 15794 }, { "epoch": 17.71, "learning_rate": 9.171830985915493e-06, "loss": 0.4984, "step": 15795 }, { "epoch": 17.71, "learning_rate": 9.170422535211267e-06, "loss": 0.3909, "step": 15796 }, { "epoch": 17.71, "learning_rate": 9.169014084507042e-06, "loss": 0.3841, "step": 15797 }, { "epoch": 17.71, "learning_rate": 9.167605633802817e-06, "loss": 0.2665, "step": 15798 }, { "epoch": 17.71, "learning_rate": 9.166197183098593e-06, "loss": 0.479, "step": 15799 }, { "epoch": 17.71, "learning_rate": 9.164788732394368e-06, "loss": 0.431, "step": 15800 }, { "epoch": 17.71, "learning_rate": 9.16338028169014e-06, "loss": 0.3731, "step": 15801 }, { "epoch": 17.72, "learning_rate": 9.161971830985915e-06, "loss": 0.4066, "step": 15802 }, { "epoch": 17.72, "learning_rate": 9.16056338028169e-06, "loss": 0.4179, "step": 15803 }, { "epoch": 17.72, "learning_rate": 9.159154929577464e-06, "loss": 0.3657, "step": 15804 }, { "epoch": 17.72, "learning_rate": 9.157746478873239e-06, "loss": 0.3828, "step": 15805 }, { "epoch": 17.72, "learning_rate": 9.156338028169015e-06, "loss": 0.4595, "step": 15806 }, { "epoch": 17.72, "learning_rate": 9.15492957746479e-06, "loss": 0.442, "step": 15807 }, { "epoch": 17.72, "learning_rate": 9.153521126760565e-06, "loss": 0.315, "step": 15808 }, { "epoch": 17.72, "learning_rate": 9.152112676056338e-06, "loss": 0.3864, "step": 15809 }, { "epoch": 17.72, "learning_rate": 9.150704225352112e-06, "loss": 0.4484, "step": 15810 }, { "epoch": 17.73, "learning_rate": 9.149295774647887e-06, "loss": 0.3526, "step": 15811 }, { "epoch": 17.73, "learning_rate": 9.147887323943662e-06, "loss": 0.4513, "step": 15812 }, { "epoch": 17.73, "learning_rate": 9.146478873239436e-06, "loss": 0.3594, "step": 15813 }, { "epoch": 17.73, "learning_rate": 9.145070422535213e-06, "loss": 0.4868, "step": 15814 }, { "epoch": 17.73, "learning_rate": 9.143661971830987e-06, "loss": 0.5213, "step": 15815 }, { "epoch": 17.73, "learning_rate": 9.142253521126762e-06, "loss": 0.5582, "step": 15816 }, { "epoch": 17.73, "learning_rate": 9.140845070422535e-06, "loss": 0.4664, "step": 15817 }, { "epoch": 17.73, "learning_rate": 9.13943661971831e-06, "loss": 0.5327, "step": 15818 }, { "epoch": 17.73, "learning_rate": 9.138028169014084e-06, "loss": 0.4698, "step": 15819 }, { "epoch": 17.74, "learning_rate": 9.136619718309859e-06, "loss": 0.407, "step": 15820 }, { "epoch": 17.74, "learning_rate": 9.135211267605635e-06, "loss": 0.3929, "step": 15821 }, { "epoch": 17.74, "learning_rate": 9.13380281690141e-06, "loss": 0.6086, "step": 15822 }, { "epoch": 17.74, "learning_rate": 9.132394366197184e-06, "loss": 0.4641, "step": 15823 }, { "epoch": 17.74, "learning_rate": 9.130985915492959e-06, "loss": 0.4846, "step": 15824 }, { "epoch": 17.74, "learning_rate": 9.129577464788732e-06, "loss": 0.4653, "step": 15825 }, { "epoch": 17.74, "learning_rate": 9.128169014084507e-06, "loss": 0.5819, "step": 15826 }, { "epoch": 17.74, "learning_rate": 9.126760563380281e-06, "loss": 0.5156, "step": 15827 }, { "epoch": 17.74, "learning_rate": 9.125352112676056e-06, "loss": 0.3949, "step": 15828 }, { "epoch": 17.75, "learning_rate": 9.123943661971832e-06, "loss": 0.4466, "step": 15829 }, { "epoch": 17.75, "learning_rate": 9.122535211267607e-06, "loss": 0.4632, "step": 15830 }, { "epoch": 17.75, "learning_rate": 9.121126760563381e-06, "loss": 0.4432, "step": 15831 }, { "epoch": 17.75, "learning_rate": 9.119718309859156e-06, "loss": 0.3796, "step": 15832 }, { "epoch": 17.75, "learning_rate": 9.118309859154929e-06, "loss": 0.5463, "step": 15833 }, { "epoch": 17.75, "learning_rate": 9.116901408450704e-06, "loss": 0.359, "step": 15834 }, { "epoch": 17.75, "learning_rate": 9.115492957746478e-06, "loss": 0.3569, "step": 15835 }, { "epoch": 17.75, "learning_rate": 9.114084507042253e-06, "loss": 0.4248, "step": 15836 }, { "epoch": 17.75, "learning_rate": 9.11267605633803e-06, "loss": 0.4635, "step": 15837 }, { "epoch": 17.76, "learning_rate": 9.111267605633804e-06, "loss": 0.3633, "step": 15838 }, { "epoch": 17.76, "learning_rate": 9.109859154929579e-06, "loss": 0.3215, "step": 15839 }, { "epoch": 17.76, "learning_rate": 9.108450704225353e-06, "loss": 0.4001, "step": 15840 }, { "epoch": 17.76, "learning_rate": 9.107042253521126e-06, "loss": 0.4742, "step": 15841 }, { "epoch": 17.76, "learning_rate": 9.1056338028169e-06, "loss": 0.3916, "step": 15842 }, { "epoch": 17.76, "learning_rate": 9.104225352112675e-06, "loss": 0.3952, "step": 15843 }, { "epoch": 17.76, "learning_rate": 9.102816901408452e-06, "loss": 0.3653, "step": 15844 }, { "epoch": 17.76, "learning_rate": 9.101408450704226e-06, "loss": 0.3646, "step": 15845 }, { "epoch": 17.76, "learning_rate": 9.100000000000001e-06, "loss": 0.4341, "step": 15846 }, { "epoch": 17.77, "learning_rate": 9.098591549295776e-06, "loss": 0.5164, "step": 15847 }, { "epoch": 17.77, "learning_rate": 9.09718309859155e-06, "loss": 0.4055, "step": 15848 }, { "epoch": 17.77, "learning_rate": 9.095774647887323e-06, "loss": 0.3404, "step": 15849 }, { "epoch": 17.77, "learning_rate": 9.094366197183098e-06, "loss": 0.4963, "step": 15850 }, { "epoch": 17.77, "learning_rate": 9.092957746478872e-06, "loss": 0.4423, "step": 15851 }, { "epoch": 17.77, "learning_rate": 9.091549295774649e-06, "loss": 0.327, "step": 15852 }, { "epoch": 17.77, "learning_rate": 9.090140845070423e-06, "loss": 0.3146, "step": 15853 }, { "epoch": 17.77, "learning_rate": 9.088732394366198e-06, "loss": 0.3484, "step": 15854 }, { "epoch": 17.77, "learning_rate": 9.087323943661973e-06, "loss": 0.4513, "step": 15855 }, { "epoch": 17.78, "learning_rate": 9.085915492957746e-06, "loss": 0.5164, "step": 15856 }, { "epoch": 17.78, "learning_rate": 9.08450704225352e-06, "loss": 0.405, "step": 15857 }, { "epoch": 17.78, "learning_rate": 9.083098591549295e-06, "loss": 0.3585, "step": 15858 }, { "epoch": 17.78, "learning_rate": 9.081690140845071e-06, "loss": 0.382, "step": 15859 }, { "epoch": 17.78, "learning_rate": 9.080281690140846e-06, "loss": 0.5954, "step": 15860 }, { "epoch": 17.78, "learning_rate": 9.07887323943662e-06, "loss": 0.2962, "step": 15861 }, { "epoch": 17.78, "learning_rate": 9.077464788732395e-06, "loss": 0.4016, "step": 15862 }, { "epoch": 17.78, "learning_rate": 9.07605633802817e-06, "loss": 0.3484, "step": 15863 }, { "epoch": 17.78, "learning_rate": 9.074647887323943e-06, "loss": 0.4241, "step": 15864 }, { "epoch": 17.79, "learning_rate": 9.073239436619717e-06, "loss": 0.5634, "step": 15865 }, { "epoch": 17.79, "learning_rate": 9.071830985915492e-06, "loss": 0.4709, "step": 15866 }, { "epoch": 17.79, "learning_rate": 9.070422535211268e-06, "loss": 0.4826, "step": 15867 }, { "epoch": 17.79, "learning_rate": 9.069014084507043e-06, "loss": 0.4496, "step": 15868 }, { "epoch": 17.79, "learning_rate": 9.067605633802818e-06, "loss": 0.4209, "step": 15869 }, { "epoch": 17.79, "learning_rate": 9.066197183098592e-06, "loss": 0.409, "step": 15870 }, { "epoch": 17.79, "learning_rate": 9.064788732394367e-06, "loss": 0.4134, "step": 15871 }, { "epoch": 17.79, "learning_rate": 9.06338028169014e-06, "loss": 0.3583, "step": 15872 }, { "epoch": 17.79, "learning_rate": 9.061971830985914e-06, "loss": 0.3873, "step": 15873 }, { "epoch": 17.8, "learning_rate": 9.06056338028169e-06, "loss": 0.409, "step": 15874 }, { "epoch": 17.8, "learning_rate": 9.059154929577465e-06, "loss": 0.4275, "step": 15875 }, { "epoch": 17.8, "learning_rate": 9.05774647887324e-06, "loss": 0.3932, "step": 15876 }, { "epoch": 17.8, "learning_rate": 9.056338028169015e-06, "loss": 0.388, "step": 15877 }, { "epoch": 17.8, "learning_rate": 9.05492957746479e-06, "loss": 0.3418, "step": 15878 }, { "epoch": 17.8, "learning_rate": 9.053521126760564e-06, "loss": 0.437, "step": 15879 }, { "epoch": 17.8, "learning_rate": 9.052112676056337e-06, "loss": 0.5129, "step": 15880 }, { "epoch": 17.8, "learning_rate": 9.050704225352112e-06, "loss": 0.4387, "step": 15881 }, { "epoch": 17.8, "learning_rate": 9.049295774647888e-06, "loss": 0.4158, "step": 15882 }, { "epoch": 17.81, "learning_rate": 9.047887323943663e-06, "loss": 0.4018, "step": 15883 }, { "epoch": 17.81, "learning_rate": 9.046478873239437e-06, "loss": 0.4213, "step": 15884 }, { "epoch": 17.81, "learning_rate": 9.045070422535212e-06, "loss": 0.4134, "step": 15885 }, { "epoch": 17.81, "learning_rate": 9.043661971830986e-06, "loss": 0.4429, "step": 15886 }, { "epoch": 17.81, "learning_rate": 9.042253521126761e-06, "loss": 0.3075, "step": 15887 }, { "epoch": 17.81, "learning_rate": 9.040845070422534e-06, "loss": 0.3943, "step": 15888 }, { "epoch": 17.81, "learning_rate": 9.039436619718309e-06, "loss": 0.4778, "step": 15889 }, { "epoch": 17.81, "learning_rate": 9.038028169014085e-06, "loss": 0.4193, "step": 15890 }, { "epoch": 17.82, "learning_rate": 9.03661971830986e-06, "loss": 0.4212, "step": 15891 }, { "epoch": 17.82, "learning_rate": 9.035211267605634e-06, "loss": 0.4861, "step": 15892 }, { "epoch": 17.82, "learning_rate": 9.033802816901409e-06, "loss": 0.3808, "step": 15893 }, { "epoch": 17.82, "learning_rate": 9.032394366197184e-06, "loss": 0.4899, "step": 15894 }, { "epoch": 17.82, "learning_rate": 9.030985915492958e-06, "loss": 0.3321, "step": 15895 }, { "epoch": 17.82, "learning_rate": 9.029577464788731e-06, "loss": 0.4122, "step": 15896 }, { "epoch": 17.82, "learning_rate": 9.028169014084507e-06, "loss": 0.3548, "step": 15897 }, { "epoch": 17.82, "learning_rate": 9.026760563380282e-06, "loss": 0.4688, "step": 15898 }, { "epoch": 17.82, "learning_rate": 9.025352112676057e-06, "loss": 0.4586, "step": 15899 }, { "epoch": 17.83, "learning_rate": 9.023943661971831e-06, "loss": 0.4369, "step": 15900 }, { "epoch": 17.83, "learning_rate": 9.022535211267606e-06, "loss": 0.4598, "step": 15901 }, { "epoch": 17.83, "learning_rate": 9.02112676056338e-06, "loss": 0.4036, "step": 15902 }, { "epoch": 17.83, "learning_rate": 9.019718309859155e-06, "loss": 0.3018, "step": 15903 }, { "epoch": 17.83, "learning_rate": 9.018309859154928e-06, "loss": 0.3831, "step": 15904 }, { "epoch": 17.83, "learning_rate": 9.016901408450705e-06, "loss": 0.4846, "step": 15905 }, { "epoch": 17.83, "learning_rate": 9.01549295774648e-06, "loss": 0.5278, "step": 15906 }, { "epoch": 17.83, "learning_rate": 9.014084507042254e-06, "loss": 0.3755, "step": 15907 }, { "epoch": 17.83, "learning_rate": 9.012676056338028e-06, "loss": 0.3414, "step": 15908 }, { "epoch": 17.84, "learning_rate": 9.011267605633803e-06, "loss": 0.3584, "step": 15909 }, { "epoch": 17.84, "learning_rate": 9.009859154929578e-06, "loss": 0.4326, "step": 15910 }, { "epoch": 17.84, "learning_rate": 9.008450704225352e-06, "loss": 0.3386, "step": 15911 }, { "epoch": 17.84, "learning_rate": 9.007042253521127e-06, "loss": 0.3462, "step": 15912 }, { "epoch": 17.84, "learning_rate": 9.005633802816902e-06, "loss": 0.3292, "step": 15913 }, { "epoch": 17.84, "learning_rate": 9.004225352112676e-06, "loss": 0.488, "step": 15914 }, { "epoch": 17.84, "learning_rate": 9.002816901408451e-06, "loss": 0.5607, "step": 15915 }, { "epoch": 17.84, "learning_rate": 9.001408450704226e-06, "loss": 0.522, "step": 15916 }, { "epoch": 17.84, "learning_rate": 9e-06, "loss": 0.492, "step": 15917 }, { "epoch": 17.85, "learning_rate": 8.998591549295775e-06, "loss": 0.487, "step": 15918 }, { "epoch": 17.85, "learning_rate": 8.99718309859155e-06, "loss": 0.4033, "step": 15919 }, { "epoch": 17.85, "learning_rate": 8.995774647887324e-06, "loss": 0.4325, "step": 15920 }, { "epoch": 17.85, "learning_rate": 8.994366197183099e-06, "loss": 0.4694, "step": 15921 }, { "epoch": 17.85, "learning_rate": 8.992957746478873e-06, "loss": 0.466, "step": 15922 }, { "epoch": 17.85, "learning_rate": 8.991549295774648e-06, "loss": 0.3281, "step": 15923 }, { "epoch": 17.85, "learning_rate": 8.990140845070423e-06, "loss": 0.4329, "step": 15924 }, { "epoch": 17.85, "learning_rate": 8.988732394366197e-06, "loss": 0.3992, "step": 15925 }, { "epoch": 17.85, "learning_rate": 8.987323943661972e-06, "loss": 0.372, "step": 15926 }, { "epoch": 17.86, "learning_rate": 8.985915492957748e-06, "loss": 0.4322, "step": 15927 }, { "epoch": 17.86, "learning_rate": 8.984507042253521e-06, "loss": 0.49, "step": 15928 }, { "epoch": 17.86, "learning_rate": 8.983098591549296e-06, "loss": 0.4547, "step": 15929 }, { "epoch": 17.86, "learning_rate": 8.98169014084507e-06, "loss": 0.543, "step": 15930 }, { "epoch": 17.86, "learning_rate": 8.980281690140845e-06, "loss": 0.4613, "step": 15931 }, { "epoch": 17.86, "learning_rate": 8.97887323943662e-06, "loss": 0.3208, "step": 15932 }, { "epoch": 17.86, "learning_rate": 8.977464788732394e-06, "loss": 0.3868, "step": 15933 }, { "epoch": 17.86, "learning_rate": 8.976056338028169e-06, "loss": 0.5847, "step": 15934 }, { "epoch": 17.86, "learning_rate": 8.974647887323945e-06, "loss": 0.4025, "step": 15935 }, { "epoch": 17.87, "learning_rate": 8.973239436619718e-06, "loss": 0.3666, "step": 15936 }, { "epoch": 17.87, "learning_rate": 8.971830985915493e-06, "loss": 0.4876, "step": 15937 }, { "epoch": 17.87, "learning_rate": 8.970422535211268e-06, "loss": 0.4712, "step": 15938 }, { "epoch": 17.87, "learning_rate": 8.969014084507042e-06, "loss": 0.5154, "step": 15939 }, { "epoch": 17.87, "learning_rate": 8.967605633802817e-06, "loss": 0.3966, "step": 15940 }, { "epoch": 17.87, "learning_rate": 8.966197183098592e-06, "loss": 0.3594, "step": 15941 }, { "epoch": 17.87, "learning_rate": 8.964788732394368e-06, "loss": 0.4629, "step": 15942 }, { "epoch": 17.87, "learning_rate": 8.963380281690142e-06, "loss": 0.3642, "step": 15943 }, { "epoch": 17.87, "learning_rate": 8.961971830985915e-06, "loss": 0.4387, "step": 15944 }, { "epoch": 17.88, "learning_rate": 8.96056338028169e-06, "loss": 0.3752, "step": 15945 }, { "epoch": 17.88, "learning_rate": 8.959154929577465e-06, "loss": 0.437, "step": 15946 }, { "epoch": 17.88, "learning_rate": 8.95774647887324e-06, "loss": 0.4685, "step": 15947 }, { "epoch": 17.88, "learning_rate": 8.956338028169014e-06, "loss": 0.398, "step": 15948 }, { "epoch": 17.88, "learning_rate": 8.954929577464789e-06, "loss": 0.3728, "step": 15949 }, { "epoch": 17.88, "learning_rate": 8.953521126760565e-06, "loss": 0.3804, "step": 15950 }, { "epoch": 17.88, "learning_rate": 8.95211267605634e-06, "loss": 0.5153, "step": 15951 }, { "epoch": 17.88, "learning_rate": 8.950704225352113e-06, "loss": 0.4219, "step": 15952 }, { "epoch": 17.88, "learning_rate": 8.949295774647887e-06, "loss": 0.4148, "step": 15953 }, { "epoch": 17.89, "learning_rate": 8.947887323943662e-06, "loss": 0.4378, "step": 15954 }, { "epoch": 17.89, "learning_rate": 8.946478873239436e-06, "loss": 0.5177, "step": 15955 }, { "epoch": 17.89, "learning_rate": 8.945070422535211e-06, "loss": 0.4663, "step": 15956 }, { "epoch": 17.89, "learning_rate": 8.943661971830986e-06, "loss": 0.4117, "step": 15957 }, { "epoch": 17.89, "learning_rate": 8.942253521126762e-06, "loss": 0.4248, "step": 15958 }, { "epoch": 17.89, "learning_rate": 8.940845070422535e-06, "loss": 0.4266, "step": 15959 }, { "epoch": 17.89, "learning_rate": 8.93943661971831e-06, "loss": 0.3505, "step": 15960 }, { "epoch": 17.89, "learning_rate": 8.938028169014084e-06, "loss": 0.3033, "step": 15961 }, { "epoch": 17.89, "learning_rate": 8.936619718309859e-06, "loss": 0.3185, "step": 15962 }, { "epoch": 17.9, "learning_rate": 8.935211267605634e-06, "loss": 0.3573, "step": 15963 }, { "epoch": 17.9, "learning_rate": 8.933802816901408e-06, "loss": 0.3759, "step": 15964 }, { "epoch": 17.9, "learning_rate": 8.932394366197185e-06, "loss": 0.6099, "step": 15965 }, { "epoch": 17.9, "learning_rate": 8.93098591549296e-06, "loss": 0.6084, "step": 15966 }, { "epoch": 17.9, "learning_rate": 8.929577464788732e-06, "loss": 0.5384, "step": 15967 }, { "epoch": 17.9, "learning_rate": 8.928169014084507e-06, "loss": 0.3748, "step": 15968 }, { "epoch": 17.9, "learning_rate": 8.926760563380281e-06, "loss": 0.4477, "step": 15969 }, { "epoch": 17.9, "learning_rate": 8.925352112676056e-06, "loss": 0.4814, "step": 15970 }, { "epoch": 17.9, "learning_rate": 8.92394366197183e-06, "loss": 0.4309, "step": 15971 }, { "epoch": 17.91, "learning_rate": 8.922535211267605e-06, "loss": 0.4495, "step": 15972 }, { "epoch": 17.91, "learning_rate": 8.921126760563382e-06, "loss": 0.4077, "step": 15973 }, { "epoch": 17.91, "learning_rate": 8.919718309859156e-06, "loss": 0.5226, "step": 15974 }, { "epoch": 17.91, "learning_rate": 8.91830985915493e-06, "loss": 0.3588, "step": 15975 }, { "epoch": 17.91, "learning_rate": 8.916901408450704e-06, "loss": 0.4581, "step": 15976 }, { "epoch": 17.91, "learning_rate": 8.915492957746478e-06, "loss": 0.4354, "step": 15977 }, { "epoch": 17.91, "learning_rate": 8.914084507042253e-06, "loss": 0.3867, "step": 15978 }, { "epoch": 17.91, "learning_rate": 8.912676056338028e-06, "loss": 0.5008, "step": 15979 }, { "epoch": 17.91, "learning_rate": 8.911267605633804e-06, "loss": 0.3472, "step": 15980 }, { "epoch": 17.92, "learning_rate": 8.909859154929579e-06, "loss": 0.4775, "step": 15981 }, { "epoch": 17.92, "learning_rate": 8.908450704225353e-06, "loss": 0.3745, "step": 15982 }, { "epoch": 17.92, "learning_rate": 8.907042253521126e-06, "loss": 0.403, "step": 15983 }, { "epoch": 17.92, "learning_rate": 8.905633802816901e-06, "loss": 0.268, "step": 15984 }, { "epoch": 17.92, "learning_rate": 8.904225352112676e-06, "loss": 0.4986, "step": 15985 }, { "epoch": 17.92, "learning_rate": 8.90281690140845e-06, "loss": 0.4107, "step": 15986 }, { "epoch": 17.92, "learning_rate": 8.901408450704225e-06, "loss": 0.4658, "step": 15987 }, { "epoch": 17.92, "learning_rate": 8.900000000000001e-06, "loss": 0.4227, "step": 15988 }, { "epoch": 17.92, "learning_rate": 8.898591549295776e-06, "loss": 0.5053, "step": 15989 }, { "epoch": 17.93, "learning_rate": 8.89718309859155e-06, "loss": 0.3835, "step": 15990 }, { "epoch": 17.93, "learning_rate": 8.895774647887323e-06, "loss": 0.4062, "step": 15991 }, { "epoch": 17.93, "learning_rate": 8.894366197183098e-06, "loss": 0.4547, "step": 15992 }, { "epoch": 17.93, "learning_rate": 8.892957746478873e-06, "loss": 0.4716, "step": 15993 }, { "epoch": 17.93, "learning_rate": 8.891549295774647e-06, "loss": 0.4785, "step": 15994 }, { "epoch": 17.93, "learning_rate": 8.890140845070424e-06, "loss": 0.4863, "step": 15995 }, { "epoch": 17.93, "learning_rate": 8.888732394366198e-06, "loss": 0.3849, "step": 15996 }, { "epoch": 17.93, "learning_rate": 8.887323943661973e-06, "loss": 0.369, "step": 15997 }, { "epoch": 17.93, "learning_rate": 8.885915492957748e-06, "loss": 0.3654, "step": 15998 }, { "epoch": 17.94, "learning_rate": 8.88450704225352e-06, "loss": 0.4632, "step": 15999 }, { "epoch": 17.94, "learning_rate": 8.883098591549295e-06, "loss": 0.3794, "step": 16000 }, { "epoch": 17.94, "learning_rate": 8.88169014084507e-06, "loss": 0.4155, "step": 16001 }, { "epoch": 17.94, "learning_rate": 8.880281690140844e-06, "loss": 0.4113, "step": 16002 }, { "epoch": 17.94, "learning_rate": 8.87887323943662e-06, "loss": 0.3546, "step": 16003 }, { "epoch": 17.94, "learning_rate": 8.877464788732395e-06, "loss": 0.331, "step": 16004 }, { "epoch": 17.94, "learning_rate": 8.87605633802817e-06, "loss": 0.4126, "step": 16005 }, { "epoch": 17.94, "learning_rate": 8.874647887323945e-06, "loss": 0.4136, "step": 16006 }, { "epoch": 17.95, "learning_rate": 8.873239436619718e-06, "loss": 0.4304, "step": 16007 }, { "epoch": 17.95, "learning_rate": 8.871830985915492e-06, "loss": 0.4937, "step": 16008 }, { "epoch": 17.95, "learning_rate": 8.870422535211267e-06, "loss": 0.4987, "step": 16009 }, { "epoch": 17.95, "learning_rate": 8.869014084507042e-06, "loss": 0.3575, "step": 16010 }, { "epoch": 17.95, "learning_rate": 8.867605633802818e-06, "loss": 0.3263, "step": 16011 }, { "epoch": 17.95, "learning_rate": 8.866197183098592e-06, "loss": 0.3238, "step": 16012 }, { "epoch": 17.95, "learning_rate": 8.864788732394367e-06, "loss": 0.4172, "step": 16013 }, { "epoch": 17.95, "learning_rate": 8.863380281690142e-06, "loss": 0.5037, "step": 16014 }, { "epoch": 17.95, "learning_rate": 8.861971830985915e-06, "loss": 0.7704, "step": 16015 }, { "epoch": 17.96, "learning_rate": 8.86056338028169e-06, "loss": 0.4906, "step": 16016 }, { "epoch": 17.96, "learning_rate": 8.859154929577464e-06, "loss": 0.5688, "step": 16017 }, { "epoch": 17.96, "learning_rate": 8.85774647887324e-06, "loss": 0.5066, "step": 16018 }, { "epoch": 17.96, "learning_rate": 8.856338028169015e-06, "loss": 0.4276, "step": 16019 }, { "epoch": 17.96, "learning_rate": 8.85492957746479e-06, "loss": 0.6016, "step": 16020 }, { "epoch": 17.96, "learning_rate": 8.853521126760564e-06, "loss": 0.3595, "step": 16021 }, { "epoch": 17.96, "learning_rate": 8.852112676056339e-06, "loss": 0.4928, "step": 16022 }, { "epoch": 17.96, "learning_rate": 8.850704225352112e-06, "loss": 0.4214, "step": 16023 }, { "epoch": 17.96, "learning_rate": 8.849295774647886e-06, "loss": 0.4905, "step": 16024 }, { "epoch": 17.97, "learning_rate": 8.847887323943661e-06, "loss": 0.43, "step": 16025 }, { "epoch": 17.97, "learning_rate": 8.846478873239437e-06, "loss": 0.403, "step": 16026 }, { "epoch": 17.97, "learning_rate": 8.845070422535212e-06, "loss": 0.4632, "step": 16027 }, { "epoch": 17.97, "learning_rate": 8.843661971830987e-06, "loss": 0.4028, "step": 16028 }, { "epoch": 17.97, "learning_rate": 8.842253521126761e-06, "loss": 0.3859, "step": 16029 }, { "epoch": 17.97, "learning_rate": 8.840845070422536e-06, "loss": 0.4376, "step": 16030 }, { "epoch": 17.97, "learning_rate": 8.839436619718309e-06, "loss": 0.3592, "step": 16031 }, { "epoch": 17.97, "learning_rate": 8.838028169014084e-06, "loss": 0.4906, "step": 16032 }, { "epoch": 17.97, "learning_rate": 8.83661971830986e-06, "loss": 0.5043, "step": 16033 }, { "epoch": 17.98, "learning_rate": 8.835211267605635e-06, "loss": 0.4373, "step": 16034 }, { "epoch": 17.98, "learning_rate": 8.833802816901409e-06, "loss": 0.4695, "step": 16035 }, { "epoch": 17.98, "learning_rate": 8.832394366197184e-06, "loss": 0.3221, "step": 16036 }, { "epoch": 17.98, "learning_rate": 8.830985915492958e-06, "loss": 0.3681, "step": 16037 }, { "epoch": 17.98, "learning_rate": 8.829577464788733e-06, "loss": 0.4953, "step": 16038 }, { "epoch": 17.98, "learning_rate": 8.828169014084506e-06, "loss": 0.4361, "step": 16039 }, { "epoch": 17.98, "learning_rate": 8.82676056338028e-06, "loss": 0.3548, "step": 16040 }, { "epoch": 17.98, "learning_rate": 8.825352112676057e-06, "loss": 0.3761, "step": 16041 }, { "epoch": 17.98, "learning_rate": 8.823943661971832e-06, "loss": 0.4208, "step": 16042 }, { "epoch": 17.99, "learning_rate": 8.822535211267606e-06, "loss": 0.3733, "step": 16043 }, { "epoch": 17.99, "learning_rate": 8.821126760563381e-06, "loss": 0.5143, "step": 16044 }, { "epoch": 17.99, "learning_rate": 8.819718309859156e-06, "loss": 0.3789, "step": 16045 }, { "epoch": 17.99, "learning_rate": 8.81830985915493e-06, "loss": 0.373, "step": 16046 }, { "epoch": 17.99, "learning_rate": 8.816901408450703e-06, "loss": 0.4648, "step": 16047 }, { "epoch": 17.99, "learning_rate": 8.81549295774648e-06, "loss": 0.3879, "step": 16048 }, { "epoch": 17.99, "learning_rate": 8.814084507042254e-06, "loss": 0.4054, "step": 16049 }, { "epoch": 17.99, "learning_rate": 8.812676056338029e-06, "loss": 0.4186, "step": 16050 }, { "epoch": 17.99, "learning_rate": 8.811267605633803e-06, "loss": 0.476, "step": 16051 }, { "epoch": 18.0, "learning_rate": 8.809859154929578e-06, "loss": 0.2349, "step": 16052 }, { "epoch": 18.0, "learning_rate": 8.808450704225353e-06, "loss": 0.3572, "step": 16053 }, { "epoch": 18.0, "learning_rate": 8.807042253521127e-06, "loss": 0.3628, "step": 16054 }, { "epoch": 18.0, "learning_rate": 8.8056338028169e-06, "loss": 0.3185, "step": 16055 }, { "epoch": 18.0, "learning_rate": 8.804225352112677e-06, "loss": 0.4435, "step": 16056 }, { "epoch": 18.0, "learning_rate": 8.802816901408451e-06, "loss": 0.4951, "step": 16057 }, { "epoch": 18.0, "learning_rate": 8.801408450704226e-06, "loss": 0.4311, "step": 16058 }, { "epoch": 18.0, "learning_rate": 8.8e-06, "loss": 0.5205, "step": 16059 }, { "epoch": 18.0, "learning_rate": 8.798591549295775e-06, "loss": 0.5016, "step": 16060 }, { "epoch": 18.01, "learning_rate": 8.79718309859155e-06, "loss": 0.3861, "step": 16061 }, { "epoch": 18.01, "learning_rate": 8.795774647887324e-06, "loss": 0.4898, "step": 16062 }, { "epoch": 18.01, "learning_rate": 8.794366197183099e-06, "loss": 0.4321, "step": 16063 }, { "epoch": 18.01, "learning_rate": 8.792957746478874e-06, "loss": 0.3241, "step": 16064 }, { "epoch": 18.01, "learning_rate": 8.791549295774648e-06, "loss": 0.4512, "step": 16065 }, { "epoch": 18.01, "learning_rate": 8.790140845070423e-06, "loss": 0.4085, "step": 16066 }, { "epoch": 18.01, "learning_rate": 8.788732394366198e-06, "loss": 0.541, "step": 16067 }, { "epoch": 18.01, "learning_rate": 8.787323943661972e-06, "loss": 0.4571, "step": 16068 }, { "epoch": 18.01, "learning_rate": 8.785915492957747e-06, "loss": 0.4464, "step": 16069 }, { "epoch": 18.02, "learning_rate": 8.784507042253521e-06, "loss": 0.3903, "step": 16070 }, { "epoch": 18.02, "learning_rate": 8.783098591549296e-06, "loss": 0.5404, "step": 16071 }, { "epoch": 18.02, "learning_rate": 8.78169014084507e-06, "loss": 0.3619, "step": 16072 }, { "epoch": 18.02, "learning_rate": 8.780281690140845e-06, "loss": 0.371, "step": 16073 }, { "epoch": 18.02, "learning_rate": 8.77887323943662e-06, "loss": 0.3032, "step": 16074 }, { "epoch": 18.02, "learning_rate": 8.777464788732395e-06, "loss": 0.3805, "step": 16075 }, { "epoch": 18.02, "learning_rate": 8.77605633802817e-06, "loss": 0.3475, "step": 16076 }, { "epoch": 18.02, "learning_rate": 8.774647887323944e-06, "loss": 0.3685, "step": 16077 }, { "epoch": 18.02, "learning_rate": 8.773239436619719e-06, "loss": 0.3652, "step": 16078 }, { "epoch": 18.03, "learning_rate": 8.771830985915493e-06, "loss": 0.4081, "step": 16079 }, { "epoch": 18.03, "learning_rate": 8.770422535211268e-06, "loss": 0.3639, "step": 16080 }, { "epoch": 18.03, "learning_rate": 8.769014084507042e-06, "loss": 0.3684, "step": 16081 }, { "epoch": 18.03, "learning_rate": 8.767605633802817e-06, "loss": 0.3923, "step": 16082 }, { "epoch": 18.03, "learning_rate": 8.766197183098592e-06, "loss": 0.4222, "step": 16083 }, { "epoch": 18.03, "learning_rate": 8.764788732394366e-06, "loss": 0.3879, "step": 16084 }, { "epoch": 18.03, "learning_rate": 8.763380281690141e-06, "loss": 0.4048, "step": 16085 }, { "epoch": 18.03, "learning_rate": 8.761971830985916e-06, "loss": 0.3443, "step": 16086 }, { "epoch": 18.03, "learning_rate": 8.76056338028169e-06, "loss": 0.4219, "step": 16087 }, { "epoch": 18.04, "learning_rate": 8.759154929577465e-06, "loss": 0.4961, "step": 16088 }, { "epoch": 18.04, "learning_rate": 8.75774647887324e-06, "loss": 0.4173, "step": 16089 }, { "epoch": 18.04, "learning_rate": 8.756338028169014e-06, "loss": 0.3481, "step": 16090 }, { "epoch": 18.04, "learning_rate": 8.754929577464789e-06, "loss": 0.3939, "step": 16091 }, { "epoch": 18.04, "learning_rate": 8.753521126760563e-06, "loss": 0.3172, "step": 16092 }, { "epoch": 18.04, "learning_rate": 8.752112676056338e-06, "loss": 0.3417, "step": 16093 }, { "epoch": 18.04, "learning_rate": 8.750704225352113e-06, "loss": 0.3499, "step": 16094 }, { "epoch": 18.04, "learning_rate": 8.749295774647887e-06, "loss": 0.3066, "step": 16095 }, { "epoch": 18.04, "learning_rate": 8.747887323943662e-06, "loss": 0.4553, "step": 16096 }, { "epoch": 18.05, "learning_rate": 8.746478873239437e-06, "loss": 0.4784, "step": 16097 }, { "epoch": 18.05, "learning_rate": 8.745070422535211e-06, "loss": 0.4702, "step": 16098 }, { "epoch": 18.05, "learning_rate": 8.743661971830986e-06, "loss": 0.3703, "step": 16099 }, { "epoch": 18.05, "learning_rate": 8.74225352112676e-06, "loss": 0.368, "step": 16100 }, { "epoch": 18.05, "learning_rate": 8.740845070422537e-06, "loss": 0.432, "step": 16101 }, { "epoch": 18.05, "learning_rate": 8.73943661971831e-06, "loss": 0.2547, "step": 16102 }, { "epoch": 18.05, "learning_rate": 8.738028169014084e-06, "loss": 0.3912, "step": 16103 }, { "epoch": 18.05, "learning_rate": 8.736619718309859e-06, "loss": 0.3718, "step": 16104 }, { "epoch": 18.05, "learning_rate": 8.735211267605634e-06, "loss": 0.3611, "step": 16105 }, { "epoch": 18.06, "learning_rate": 8.733802816901408e-06, "loss": 0.3223, "step": 16106 }, { "epoch": 18.06, "learning_rate": 8.732394366197183e-06, "loss": 0.565, "step": 16107 }, { "epoch": 18.06, "learning_rate": 8.730985915492958e-06, "loss": 0.4273, "step": 16108 }, { "epoch": 18.06, "learning_rate": 8.729577464788734e-06, "loss": 0.4271, "step": 16109 }, { "epoch": 18.06, "learning_rate": 8.728169014084507e-06, "loss": 0.4087, "step": 16110 }, { "epoch": 18.06, "learning_rate": 8.726760563380282e-06, "loss": 0.3934, "step": 16111 }, { "epoch": 18.06, "learning_rate": 8.725352112676056e-06, "loss": 0.5171, "step": 16112 }, { "epoch": 18.06, "learning_rate": 8.723943661971831e-06, "loss": 0.4127, "step": 16113 }, { "epoch": 18.07, "learning_rate": 8.722535211267606e-06, "loss": 0.4319, "step": 16114 }, { "epoch": 18.07, "learning_rate": 8.72112676056338e-06, "loss": 0.4048, "step": 16115 }, { "epoch": 18.07, "learning_rate": 8.719718309859156e-06, "loss": 0.4182, "step": 16116 }, { "epoch": 18.07, "learning_rate": 8.718309859154931e-06, "loss": 0.4719, "step": 16117 }, { "epoch": 18.07, "learning_rate": 8.716901408450704e-06, "loss": 0.5243, "step": 16118 }, { "epoch": 18.07, "learning_rate": 8.715492957746479e-06, "loss": 0.3658, "step": 16119 }, { "epoch": 18.07, "learning_rate": 8.714084507042253e-06, "loss": 0.4065, "step": 16120 }, { "epoch": 18.07, "learning_rate": 8.712676056338028e-06, "loss": 0.2771, "step": 16121 }, { "epoch": 18.07, "learning_rate": 8.711267605633803e-06, "loss": 0.414, "step": 16122 }, { "epoch": 18.08, "learning_rate": 8.709859154929577e-06, "loss": 0.5076, "step": 16123 }, { "epoch": 18.08, "learning_rate": 8.708450704225354e-06, "loss": 0.398, "step": 16124 }, { "epoch": 18.08, "learning_rate": 8.707042253521128e-06, "loss": 0.4115, "step": 16125 }, { "epoch": 18.08, "learning_rate": 8.705633802816901e-06, "loss": 0.2981, "step": 16126 }, { "epoch": 18.08, "learning_rate": 8.704225352112676e-06, "loss": 0.4475, "step": 16127 }, { "epoch": 18.08, "learning_rate": 8.70281690140845e-06, "loss": 0.4937, "step": 16128 }, { "epoch": 18.08, "learning_rate": 8.701408450704225e-06, "loss": 0.3699, "step": 16129 }, { "epoch": 18.08, "learning_rate": 8.7e-06, "loss": 0.4101, "step": 16130 }, { "epoch": 18.08, "learning_rate": 8.698591549295774e-06, "loss": 0.3851, "step": 16131 }, { "epoch": 18.09, "learning_rate": 8.69718309859155e-06, "loss": 0.3205, "step": 16132 }, { "epoch": 18.09, "learning_rate": 8.695774647887325e-06, "loss": 0.3851, "step": 16133 }, { "epoch": 18.09, "learning_rate": 8.694366197183098e-06, "loss": 0.3808, "step": 16134 }, { "epoch": 18.09, "learning_rate": 8.692957746478873e-06, "loss": 0.3829, "step": 16135 }, { "epoch": 18.09, "learning_rate": 8.691549295774648e-06, "loss": 0.3715, "step": 16136 }, { "epoch": 18.09, "learning_rate": 8.690140845070422e-06, "loss": 0.465, "step": 16137 }, { "epoch": 18.09, "learning_rate": 8.688732394366197e-06, "loss": 0.3459, "step": 16138 }, { "epoch": 18.09, "learning_rate": 8.687323943661973e-06, "loss": 0.4441, "step": 16139 }, { "epoch": 18.09, "learning_rate": 8.685915492957748e-06, "loss": 0.435, "step": 16140 }, { "epoch": 18.1, "learning_rate": 8.684507042253522e-06, "loss": 0.6056, "step": 16141 }, { "epoch": 18.1, "learning_rate": 8.683098591549295e-06, "loss": 0.4795, "step": 16142 }, { "epoch": 18.1, "learning_rate": 8.68169014084507e-06, "loss": 0.3154, "step": 16143 }, { "epoch": 18.1, "learning_rate": 8.680281690140845e-06, "loss": 0.4093, "step": 16144 }, { "epoch": 18.1, "learning_rate": 8.67887323943662e-06, "loss": 0.3892, "step": 16145 }, { "epoch": 18.1, "learning_rate": 8.677464788732394e-06, "loss": 0.4421, "step": 16146 }, { "epoch": 18.1, "learning_rate": 8.67605633802817e-06, "loss": 0.3524, "step": 16147 }, { "epoch": 18.1, "learning_rate": 8.674647887323945e-06, "loss": 0.3858, "step": 16148 }, { "epoch": 18.1, "learning_rate": 8.67323943661972e-06, "loss": 0.4015, "step": 16149 }, { "epoch": 18.11, "learning_rate": 8.671830985915492e-06, "loss": 0.3945, "step": 16150 }, { "epoch": 18.11, "learning_rate": 8.670422535211267e-06, "loss": 0.3724, "step": 16151 }, { "epoch": 18.11, "learning_rate": 8.669014084507042e-06, "loss": 0.3084, "step": 16152 }, { "epoch": 18.11, "learning_rate": 8.667605633802816e-06, "loss": 0.3426, "step": 16153 }, { "epoch": 18.11, "learning_rate": 8.666197183098593e-06, "loss": 0.2583, "step": 16154 }, { "epoch": 18.11, "learning_rate": 8.664788732394367e-06, "loss": 0.4231, "step": 16155 }, { "epoch": 18.11, "learning_rate": 8.663380281690142e-06, "loss": 0.3476, "step": 16156 }, { "epoch": 18.11, "learning_rate": 8.661971830985917e-06, "loss": 0.445, "step": 16157 }, { "epoch": 18.11, "learning_rate": 8.66056338028169e-06, "loss": 0.5022, "step": 16158 }, { "epoch": 18.12, "learning_rate": 8.659154929577464e-06, "loss": 0.4289, "step": 16159 }, { "epoch": 18.12, "learning_rate": 8.657746478873239e-06, "loss": 0.421, "step": 16160 }, { "epoch": 18.12, "learning_rate": 8.656338028169013e-06, "loss": 0.4834, "step": 16161 }, { "epoch": 18.12, "learning_rate": 8.65492957746479e-06, "loss": 0.4198, "step": 16162 }, { "epoch": 18.12, "learning_rate": 8.653521126760564e-06, "loss": 0.5114, "step": 16163 }, { "epoch": 18.12, "learning_rate": 8.652112676056339e-06, "loss": 0.4164, "step": 16164 }, { "epoch": 18.12, "learning_rate": 8.650704225352114e-06, "loss": 0.3663, "step": 16165 }, { "epoch": 18.12, "learning_rate": 8.649295774647887e-06, "loss": 0.43, "step": 16166 }, { "epoch": 18.12, "learning_rate": 8.647887323943661e-06, "loss": 0.4489, "step": 16167 }, { "epoch": 18.13, "learning_rate": 8.646478873239436e-06, "loss": 0.3938, "step": 16168 }, { "epoch": 18.13, "learning_rate": 8.645070422535212e-06, "loss": 0.2804, "step": 16169 }, { "epoch": 18.13, "learning_rate": 8.643661971830987e-06, "loss": 0.3659, "step": 16170 }, { "epoch": 18.13, "learning_rate": 8.642253521126762e-06, "loss": 0.4104, "step": 16171 }, { "epoch": 18.13, "learning_rate": 8.640845070422536e-06, "loss": 0.344, "step": 16172 }, { "epoch": 18.13, "learning_rate": 8.63943661971831e-06, "loss": 0.4355, "step": 16173 }, { "epoch": 18.13, "learning_rate": 8.638028169014084e-06, "loss": 0.38, "step": 16174 }, { "epoch": 18.13, "learning_rate": 8.636619718309858e-06, "loss": 0.384, "step": 16175 }, { "epoch": 18.13, "learning_rate": 8.635211267605633e-06, "loss": 0.3977, "step": 16176 }, { "epoch": 18.14, "learning_rate": 8.63380281690141e-06, "loss": 0.3669, "step": 16177 }, { "epoch": 18.14, "learning_rate": 8.632394366197184e-06, "loss": 0.374, "step": 16178 }, { "epoch": 18.14, "learning_rate": 8.630985915492959e-06, "loss": 0.4557, "step": 16179 }, { "epoch": 18.14, "learning_rate": 8.629577464788733e-06, "loss": 0.3458, "step": 16180 }, { "epoch": 18.14, "learning_rate": 8.628169014084508e-06, "loss": 0.4277, "step": 16181 }, { "epoch": 18.14, "learning_rate": 8.626760563380281e-06, "loss": 0.5352, "step": 16182 }, { "epoch": 18.14, "learning_rate": 8.625352112676056e-06, "loss": 0.3268, "step": 16183 }, { "epoch": 18.14, "learning_rate": 8.62394366197183e-06, "loss": 0.4325, "step": 16184 }, { "epoch": 18.14, "learning_rate": 8.622535211267606e-06, "loss": 0.3875, "step": 16185 }, { "epoch": 18.15, "learning_rate": 8.621126760563381e-06, "loss": 0.3935, "step": 16186 }, { "epoch": 18.15, "learning_rate": 8.619718309859156e-06, "loss": 0.3602, "step": 16187 }, { "epoch": 18.15, "learning_rate": 8.61830985915493e-06, "loss": 0.4513, "step": 16188 }, { "epoch": 18.15, "learning_rate": 8.616901408450705e-06, "loss": 0.344, "step": 16189 }, { "epoch": 18.15, "learning_rate": 8.615492957746478e-06, "loss": 0.3697, "step": 16190 }, { "epoch": 18.15, "learning_rate": 8.614084507042253e-06, "loss": 0.3105, "step": 16191 }, { "epoch": 18.15, "learning_rate": 8.612676056338029e-06, "loss": 0.3643, "step": 16192 }, { "epoch": 18.15, "learning_rate": 8.611267605633804e-06, "loss": 0.3345, "step": 16193 }, { "epoch": 18.15, "learning_rate": 8.609859154929578e-06, "loss": 0.3878, "step": 16194 }, { "epoch": 18.16, "learning_rate": 8.608450704225353e-06, "loss": 0.3804, "step": 16195 }, { "epoch": 18.16, "learning_rate": 8.607042253521127e-06, "loss": 0.4062, "step": 16196 }, { "epoch": 18.16, "learning_rate": 8.6056338028169e-06, "loss": 0.4043, "step": 16197 }, { "epoch": 18.16, "learning_rate": 8.604225352112675e-06, "loss": 0.4087, "step": 16198 }, { "epoch": 18.16, "learning_rate": 8.60281690140845e-06, "loss": 0.2987, "step": 16199 }, { "epoch": 18.16, "learning_rate": 8.601408450704226e-06, "loss": 0.3981, "step": 16200 }, { "epoch": 18.16, "learning_rate": 8.6e-06, "loss": 0.3877, "step": 16201 }, { "epoch": 18.16, "learning_rate": 8.598591549295775e-06, "loss": 0.3138, "step": 16202 }, { "epoch": 18.16, "learning_rate": 8.59718309859155e-06, "loss": 0.2737, "step": 16203 }, { "epoch": 18.17, "learning_rate": 8.595774647887325e-06, "loss": 0.3529, "step": 16204 }, { "epoch": 18.17, "learning_rate": 8.594366197183098e-06, "loss": 0.4142, "step": 16205 }, { "epoch": 18.17, "learning_rate": 8.592957746478872e-06, "loss": 0.5385, "step": 16206 }, { "epoch": 18.17, "learning_rate": 8.591549295774648e-06, "loss": 0.6345, "step": 16207 }, { "epoch": 18.17, "learning_rate": 8.590140845070423e-06, "loss": 0.4948, "step": 16208 }, { "epoch": 18.17, "learning_rate": 8.588732394366198e-06, "loss": 0.3822, "step": 16209 }, { "epoch": 18.17, "learning_rate": 8.587323943661972e-06, "loss": 0.3721, "step": 16210 }, { "epoch": 18.17, "learning_rate": 8.585915492957747e-06, "loss": 0.47, "step": 16211 }, { "epoch": 18.17, "learning_rate": 8.584507042253522e-06, "loss": 0.3643, "step": 16212 }, { "epoch": 18.18, "learning_rate": 8.583098591549295e-06, "loss": 0.4377, "step": 16213 }, { "epoch": 18.18, "learning_rate": 8.58169014084507e-06, "loss": 0.4431, "step": 16214 }, { "epoch": 18.18, "learning_rate": 8.580281690140846e-06, "loss": 0.5178, "step": 16215 }, { "epoch": 18.18, "learning_rate": 8.57887323943662e-06, "loss": 0.402, "step": 16216 }, { "epoch": 18.18, "learning_rate": 8.577464788732395e-06, "loss": 0.3973, "step": 16217 }, { "epoch": 18.18, "learning_rate": 8.577464788732395e-06, "loss": 0.3337, "step": 16218 }, { "epoch": 18.18, "learning_rate": 8.57605633802817e-06, "loss": 0.3452, "step": 16219 }, { "epoch": 18.18, "learning_rate": 8.574647887323944e-06, "loss": 0.492, "step": 16220 }, { "epoch": 18.18, "learning_rate": 8.573239436619719e-06, "loss": 0.5136, "step": 16221 }, { "epoch": 18.19, "learning_rate": 8.571830985915492e-06, "loss": 0.4067, "step": 16222 }, { "epoch": 18.19, "learning_rate": 8.570422535211268e-06, "loss": 0.4147, "step": 16223 }, { "epoch": 18.19, "learning_rate": 8.569014084507043e-06, "loss": 0.3944, "step": 16224 }, { "epoch": 18.19, "learning_rate": 8.567605633802817e-06, "loss": 0.5443, "step": 16225 }, { "epoch": 18.19, "learning_rate": 8.566197183098592e-06, "loss": 0.3806, "step": 16226 }, { "epoch": 18.19, "learning_rate": 8.564788732394367e-06, "loss": 0.4214, "step": 16227 }, { "epoch": 18.19, "learning_rate": 8.563380281690141e-06, "loss": 0.389, "step": 16228 }, { "epoch": 18.19, "learning_rate": 8.561971830985916e-06, "loss": 0.3514, "step": 16229 }, { "epoch": 18.2, "learning_rate": 8.560563380281689e-06, "loss": 0.4053, "step": 16230 }, { "epoch": 18.2, "learning_rate": 8.559154929577465e-06, "loss": 0.4, "step": 16231 }, { "epoch": 18.2, "learning_rate": 8.55774647887324e-06, "loss": 0.4234, "step": 16232 }, { "epoch": 18.2, "learning_rate": 8.556338028169014e-06, "loss": 0.401, "step": 16233 }, { "epoch": 18.2, "learning_rate": 8.554929577464789e-06, "loss": 0.4062, "step": 16234 }, { "epoch": 18.2, "learning_rate": 8.553521126760564e-06, "loss": 0.4449, "step": 16235 }, { "epoch": 18.2, "learning_rate": 8.552112676056338e-06, "loss": 0.442, "step": 16236 }, { "epoch": 18.2, "learning_rate": 8.550704225352113e-06, "loss": 0.3363, "step": 16237 }, { "epoch": 18.2, "learning_rate": 8.549295774647888e-06, "loss": 0.3405, "step": 16238 }, { "epoch": 18.21, "learning_rate": 8.547887323943662e-06, "loss": 0.4012, "step": 16239 }, { "epoch": 18.21, "learning_rate": 8.546478873239437e-06, "loss": 0.407, "step": 16240 }, { "epoch": 18.21, "learning_rate": 8.545070422535212e-06, "loss": 0.3981, "step": 16241 }, { "epoch": 18.21, "learning_rate": 8.543661971830986e-06, "loss": 0.4388, "step": 16242 }, { "epoch": 18.21, "learning_rate": 8.54225352112676e-06, "loss": 0.4452, "step": 16243 }, { "epoch": 18.21, "learning_rate": 8.540845070422535e-06, "loss": 0.2762, "step": 16244 }, { "epoch": 18.21, "learning_rate": 8.53943661971831e-06, "loss": 0.4582, "step": 16245 }, { "epoch": 18.21, "learning_rate": 8.538028169014085e-06, "loss": 0.4157, "step": 16246 }, { "epoch": 18.21, "learning_rate": 8.53661971830986e-06, "loss": 0.2779, "step": 16247 }, { "epoch": 18.22, "learning_rate": 8.535211267605634e-06, "loss": 0.348, "step": 16248 }, { "epoch": 18.22, "learning_rate": 8.533802816901409e-06, "loss": 0.4094, "step": 16249 }, { "epoch": 18.22, "learning_rate": 8.532394366197183e-06, "loss": 0.345, "step": 16250 }, { "epoch": 18.22, "learning_rate": 8.530985915492958e-06, "loss": 0.2858, "step": 16251 }, { "epoch": 18.22, "learning_rate": 8.529577464788733e-06, "loss": 0.3694, "step": 16252 }, { "epoch": 18.22, "learning_rate": 8.528169014084507e-06, "loss": 0.3486, "step": 16253 }, { "epoch": 18.22, "learning_rate": 8.526760563380282e-06, "loss": 0.3628, "step": 16254 }, { "epoch": 18.22, "learning_rate": 8.525352112676056e-06, "loss": 0.3591, "step": 16255 }, { "epoch": 18.22, "learning_rate": 8.523943661971831e-06, "loss": 0.2845, "step": 16256 }, { "epoch": 18.23, "learning_rate": 8.522535211267606e-06, "loss": 0.5876, "step": 16257 }, { "epoch": 18.23, "learning_rate": 8.52112676056338e-06, "loss": 0.4347, "step": 16258 }, { "epoch": 18.23, "learning_rate": 8.519718309859155e-06, "loss": 0.4258, "step": 16259 }, { "epoch": 18.23, "learning_rate": 8.51830985915493e-06, "loss": 0.5451, "step": 16260 }, { "epoch": 18.23, "learning_rate": 8.516901408450706e-06, "loss": 0.3315, "step": 16261 }, { "epoch": 18.23, "learning_rate": 8.515492957746479e-06, "loss": 0.4457, "step": 16262 }, { "epoch": 18.23, "learning_rate": 8.514084507042254e-06, "loss": 0.5169, "step": 16263 }, { "epoch": 18.23, "learning_rate": 8.512676056338028e-06, "loss": 0.3894, "step": 16264 }, { "epoch": 18.23, "learning_rate": 8.511267605633803e-06, "loss": 0.3683, "step": 16265 }, { "epoch": 18.24, "learning_rate": 8.509859154929577e-06, "loss": 0.4608, "step": 16266 }, { "epoch": 18.24, "learning_rate": 8.508450704225352e-06, "loss": 0.3703, "step": 16267 }, { "epoch": 18.24, "learning_rate": 8.507042253521127e-06, "loss": 0.4775, "step": 16268 }, { "epoch": 18.24, "learning_rate": 8.505633802816903e-06, "loss": 0.4283, "step": 16269 }, { "epoch": 18.24, "learning_rate": 8.504225352112676e-06, "loss": 0.473, "step": 16270 }, { "epoch": 18.24, "learning_rate": 8.50281690140845e-06, "loss": 0.313, "step": 16271 }, { "epoch": 18.24, "learning_rate": 8.501408450704225e-06, "loss": 0.3697, "step": 16272 }, { "epoch": 18.24, "learning_rate": 8.5e-06, "loss": 0.4197, "step": 16273 }, { "epoch": 18.24, "learning_rate": 8.498591549295775e-06, "loss": 0.4227, "step": 16274 }, { "epoch": 18.25, "learning_rate": 8.49718309859155e-06, "loss": 0.3875, "step": 16275 }, { "epoch": 18.25, "learning_rate": 8.495774647887326e-06, "loss": 0.5373, "step": 16276 }, { "epoch": 18.25, "learning_rate": 8.4943661971831e-06, "loss": 0.431, "step": 16277 }, { "epoch": 18.25, "learning_rate": 8.492957746478873e-06, "loss": 0.3128, "step": 16278 }, { "epoch": 18.25, "learning_rate": 8.491549295774648e-06, "loss": 0.3916, "step": 16279 }, { "epoch": 18.25, "learning_rate": 8.490140845070422e-06, "loss": 0.3727, "step": 16280 }, { "epoch": 18.25, "learning_rate": 8.488732394366197e-06, "loss": 0.4015, "step": 16281 }, { "epoch": 18.25, "learning_rate": 8.487323943661972e-06, "loss": 0.3989, "step": 16282 }, { "epoch": 18.25, "learning_rate": 8.485915492957746e-06, "loss": 0.374, "step": 16283 }, { "epoch": 18.26, "learning_rate": 8.484507042253523e-06, "loss": 0.3788, "step": 16284 }, { "epoch": 18.26, "learning_rate": 8.483098591549297e-06, "loss": 0.3111, "step": 16285 }, { "epoch": 18.26, "learning_rate": 8.48169014084507e-06, "loss": 0.3766, "step": 16286 }, { "epoch": 18.26, "learning_rate": 8.480281690140845e-06, "loss": 0.4001, "step": 16287 }, { "epoch": 18.26, "learning_rate": 8.47887323943662e-06, "loss": 0.3955, "step": 16288 }, { "epoch": 18.26, "learning_rate": 8.477464788732394e-06, "loss": 0.3738, "step": 16289 }, { "epoch": 18.26, "learning_rate": 8.476056338028169e-06, "loss": 0.3684, "step": 16290 }, { "epoch": 18.26, "learning_rate": 8.474647887323945e-06, "loss": 0.2583, "step": 16291 }, { "epoch": 18.26, "learning_rate": 8.47323943661972e-06, "loss": 0.3649, "step": 16292 }, { "epoch": 18.27, "learning_rate": 8.471830985915494e-06, "loss": 0.4479, "step": 16293 }, { "epoch": 18.27, "learning_rate": 8.470422535211267e-06, "loss": 0.341, "step": 16294 }, { "epoch": 18.27, "learning_rate": 8.469014084507042e-06, "loss": 0.3953, "step": 16295 }, { "epoch": 18.27, "learning_rate": 8.467605633802817e-06, "loss": 0.3159, "step": 16296 }, { "epoch": 18.27, "learning_rate": 8.466197183098591e-06, "loss": 0.3184, "step": 16297 }, { "epoch": 18.27, "learning_rate": 8.464788732394366e-06, "loss": 0.4067, "step": 16298 }, { "epoch": 18.27, "learning_rate": 8.463380281690142e-06, "loss": 0.4148, "step": 16299 }, { "epoch": 18.27, "learning_rate": 8.461971830985917e-06, "loss": 0.3583, "step": 16300 }, { "epoch": 18.27, "learning_rate": 8.46056338028169e-06, "loss": 0.3587, "step": 16301 }, { "epoch": 18.28, "learning_rate": 8.459154929577464e-06, "loss": 0.2574, "step": 16302 }, { "epoch": 18.28, "learning_rate": 8.457746478873239e-06, "loss": 0.3549, "step": 16303 }, { "epoch": 18.28, "learning_rate": 8.456338028169014e-06, "loss": 0.2554, "step": 16304 }, { "epoch": 18.28, "learning_rate": 8.454929577464788e-06, "loss": 0.4843, "step": 16305 }, { "epoch": 18.28, "learning_rate": 8.453521126760563e-06, "loss": 0.4411, "step": 16306 }, { "epoch": 18.28, "learning_rate": 8.45211267605634e-06, "loss": 0.6041, "step": 16307 }, { "epoch": 18.28, "learning_rate": 8.450704225352114e-06, "loss": 0.4571, "step": 16308 }, { "epoch": 18.28, "learning_rate": 8.449295774647887e-06, "loss": 0.4858, "step": 16309 }, { "epoch": 18.28, "learning_rate": 8.447887323943662e-06, "loss": 0.468, "step": 16310 }, { "epoch": 18.29, "learning_rate": 8.446478873239436e-06, "loss": 0.4867, "step": 16311 }, { "epoch": 18.29, "learning_rate": 8.44507042253521e-06, "loss": 0.3731, "step": 16312 }, { "epoch": 18.29, "learning_rate": 8.443661971830985e-06, "loss": 0.3656, "step": 16313 }, { "epoch": 18.29, "learning_rate": 8.442253521126762e-06, "loss": 0.4256, "step": 16314 }, { "epoch": 18.29, "learning_rate": 8.440845070422536e-06, "loss": 0.5541, "step": 16315 }, { "epoch": 18.29, "learning_rate": 8.439436619718311e-06, "loss": 0.3909, "step": 16316 }, { "epoch": 18.29, "learning_rate": 8.438028169014084e-06, "loss": 0.4009, "step": 16317 }, { "epoch": 18.29, "learning_rate": 8.436619718309859e-06, "loss": 0.4651, "step": 16318 }, { "epoch": 18.29, "learning_rate": 8.435211267605633e-06, "loss": 0.516, "step": 16319 }, { "epoch": 18.3, "learning_rate": 8.433802816901408e-06, "loss": 0.3754, "step": 16320 }, { "epoch": 18.3, "learning_rate": 8.432394366197183e-06, "loss": 0.3608, "step": 16321 }, { "epoch": 18.3, "learning_rate": 8.430985915492959e-06, "loss": 0.4125, "step": 16322 }, { "epoch": 18.3, "learning_rate": 8.429577464788734e-06, "loss": 0.3503, "step": 16323 }, { "epoch": 18.3, "learning_rate": 8.428169014084508e-06, "loss": 0.4239, "step": 16324 }, { "epoch": 18.3, "learning_rate": 8.426760563380281e-06, "loss": 0.3176, "step": 16325 }, { "epoch": 18.3, "learning_rate": 8.425352112676056e-06, "loss": 0.5063, "step": 16326 }, { "epoch": 18.3, "learning_rate": 8.42394366197183e-06, "loss": 0.478, "step": 16327 }, { "epoch": 18.3, "learning_rate": 8.422535211267605e-06, "loss": 0.4579, "step": 16328 }, { "epoch": 18.31, "learning_rate": 8.421126760563381e-06, "loss": 0.376, "step": 16329 }, { "epoch": 18.31, "learning_rate": 8.419718309859156e-06, "loss": 0.3546, "step": 16330 }, { "epoch": 18.31, "learning_rate": 8.41830985915493e-06, "loss": 0.3509, "step": 16331 }, { "epoch": 18.31, "learning_rate": 8.416901408450705e-06, "loss": 0.3734, "step": 16332 }, { "epoch": 18.31, "learning_rate": 8.415492957746478e-06, "loss": 0.3938, "step": 16333 }, { "epoch": 18.31, "learning_rate": 8.414084507042253e-06, "loss": 0.367, "step": 16334 }, { "epoch": 18.31, "learning_rate": 8.412676056338027e-06, "loss": 0.3976, "step": 16335 }, { "epoch": 18.31, "learning_rate": 8.411267605633802e-06, "loss": 0.4305, "step": 16336 }, { "epoch": 18.32, "learning_rate": 8.409859154929578e-06, "loss": 0.3355, "step": 16337 }, { "epoch": 18.32, "learning_rate": 8.408450704225353e-06, "loss": 0.3245, "step": 16338 }, { "epoch": 18.32, "learning_rate": 8.407042253521128e-06, "loss": 0.3213, "step": 16339 }, { "epoch": 18.32, "learning_rate": 8.405633802816902e-06, "loss": 0.3348, "step": 16340 }, { "epoch": 18.32, "learning_rate": 8.404225352112675e-06, "loss": 0.4297, "step": 16341 }, { "epoch": 18.32, "learning_rate": 8.40281690140845e-06, "loss": 0.3852, "step": 16342 }, { "epoch": 18.32, "learning_rate": 8.401408450704225e-06, "loss": 0.3126, "step": 16343 }, { "epoch": 18.32, "learning_rate": 8.400000000000001e-06, "loss": 0.4542, "step": 16344 }, { "epoch": 18.32, "learning_rate": 8.398591549295776e-06, "loss": 0.3634, "step": 16345 }, { "epoch": 18.33, "learning_rate": 8.39718309859155e-06, "loss": 0.3565, "step": 16346 }, { "epoch": 18.33, "learning_rate": 8.395774647887325e-06, "loss": 0.3639, "step": 16347 }, { "epoch": 18.33, "learning_rate": 8.3943661971831e-06, "loss": 0.3572, "step": 16348 }, { "epoch": 18.33, "learning_rate": 8.392957746478872e-06, "loss": 0.2897, "step": 16349 }, { "epoch": 18.33, "learning_rate": 8.391549295774647e-06, "loss": 0.2916, "step": 16350 }, { "epoch": 18.33, "learning_rate": 8.390140845070422e-06, "loss": 0.3664, "step": 16351 }, { "epoch": 18.33, "learning_rate": 8.388732394366198e-06, "loss": 0.3026, "step": 16352 }, { "epoch": 18.33, "learning_rate": 8.387323943661973e-06, "loss": 0.2133, "step": 16353 }, { "epoch": 18.33, "learning_rate": 8.385915492957747e-06, "loss": 0.421, "step": 16354 }, { "epoch": 18.34, "learning_rate": 8.384507042253522e-06, "loss": 0.2424, "step": 16355 }, { "epoch": 18.34, "learning_rate": 8.383098591549297e-06, "loss": 0.4785, "step": 16356 }, { "epoch": 18.34, "learning_rate": 8.38169014084507e-06, "loss": 0.5382, "step": 16357 }, { "epoch": 18.34, "learning_rate": 8.380281690140844e-06, "loss": 0.5809, "step": 16358 }, { "epoch": 18.34, "learning_rate": 8.37887323943662e-06, "loss": 0.5545, "step": 16359 }, { "epoch": 18.34, "learning_rate": 8.377464788732395e-06, "loss": 0.4556, "step": 16360 }, { "epoch": 18.34, "learning_rate": 8.37605633802817e-06, "loss": 0.4481, "step": 16361 }, { "epoch": 18.34, "learning_rate": 8.374647887323944e-06, "loss": 0.4777, "step": 16362 }, { "epoch": 18.34, "learning_rate": 8.373239436619719e-06, "loss": 0.4379, "step": 16363 }, { "epoch": 18.35, "learning_rate": 8.371830985915494e-06, "loss": 0.524, "step": 16364 }, { "epoch": 18.35, "learning_rate": 8.370422535211267e-06, "loss": 0.3333, "step": 16365 }, { "epoch": 18.35, "learning_rate": 8.369014084507041e-06, "loss": 0.3591, "step": 16366 }, { "epoch": 18.35, "learning_rate": 8.367605633802818e-06, "loss": 0.3819, "step": 16367 }, { "epoch": 18.35, "learning_rate": 8.366197183098592e-06, "loss": 0.3938, "step": 16368 }, { "epoch": 18.35, "learning_rate": 8.364788732394367e-06, "loss": 0.4175, "step": 16369 }, { "epoch": 18.35, "learning_rate": 8.363380281690141e-06, "loss": 0.3609, "step": 16370 }, { "epoch": 18.35, "learning_rate": 8.361971830985916e-06, "loss": 0.4721, "step": 16371 }, { "epoch": 18.35, "learning_rate": 8.36056338028169e-06, "loss": 0.4543, "step": 16372 }, { "epoch": 18.36, "learning_rate": 8.359154929577464e-06, "loss": 0.485, "step": 16373 }, { "epoch": 18.36, "learning_rate": 8.357746478873238e-06, "loss": 0.4364, "step": 16374 }, { "epoch": 18.36, "learning_rate": 8.356338028169015e-06, "loss": 0.3724, "step": 16375 }, { "epoch": 18.36, "learning_rate": 8.35492957746479e-06, "loss": 0.472, "step": 16376 }, { "epoch": 18.36, "learning_rate": 8.353521126760564e-06, "loss": 0.3495, "step": 16377 }, { "epoch": 18.36, "learning_rate": 8.352112676056339e-06, "loss": 0.4123, "step": 16378 }, { "epoch": 18.36, "learning_rate": 8.350704225352113e-06, "loss": 0.3136, "step": 16379 }, { "epoch": 18.36, "learning_rate": 8.349295774647888e-06, "loss": 0.3412, "step": 16380 }, { "epoch": 18.36, "learning_rate": 8.34788732394366e-06, "loss": 0.3613, "step": 16381 }, { "epoch": 18.37, "learning_rate": 8.346478873239437e-06, "loss": 0.4124, "step": 16382 }, { "epoch": 18.37, "learning_rate": 8.345070422535212e-06, "loss": 0.3272, "step": 16383 }, { "epoch": 18.37, "learning_rate": 8.343661971830986e-06, "loss": 0.4316, "step": 16384 }, { "epoch": 18.37, "learning_rate": 8.342253521126761e-06, "loss": 0.3797, "step": 16385 }, { "epoch": 18.37, "learning_rate": 8.340845070422536e-06, "loss": 0.3385, "step": 16386 }, { "epoch": 18.37, "learning_rate": 8.33943661971831e-06, "loss": 0.3886, "step": 16387 }, { "epoch": 18.37, "learning_rate": 8.338028169014085e-06, "loss": 0.2627, "step": 16388 }, { "epoch": 18.37, "learning_rate": 8.336619718309858e-06, "loss": 0.3744, "step": 16389 }, { "epoch": 18.37, "learning_rate": 8.335211267605634e-06, "loss": 0.3646, "step": 16390 }, { "epoch": 18.38, "learning_rate": 8.333802816901409e-06, "loss": 0.4162, "step": 16391 }, { "epoch": 18.38, "learning_rate": 8.332394366197183e-06, "loss": 0.4335, "step": 16392 }, { "epoch": 18.38, "learning_rate": 8.330985915492958e-06, "loss": 0.3682, "step": 16393 }, { "epoch": 18.38, "learning_rate": 8.329577464788733e-06, "loss": 0.4107, "step": 16394 }, { "epoch": 18.38, "learning_rate": 8.328169014084507e-06, "loss": 0.3084, "step": 16395 }, { "epoch": 18.38, "learning_rate": 8.326760563380282e-06, "loss": 0.38, "step": 16396 }, { "epoch": 18.38, "learning_rate": 8.325352112676057e-06, "loss": 0.4113, "step": 16397 }, { "epoch": 18.38, "learning_rate": 8.323943661971831e-06, "loss": 0.4112, "step": 16398 }, { "epoch": 18.38, "learning_rate": 8.322535211267606e-06, "loss": 0.3322, "step": 16399 }, { "epoch": 18.39, "learning_rate": 8.32112676056338e-06, "loss": 0.366, "step": 16400 }, { "epoch": 18.39, "learning_rate": 8.319718309859155e-06, "loss": 0.335, "step": 16401 }, { "epoch": 18.39, "learning_rate": 8.31830985915493e-06, "loss": 0.3349, "step": 16402 }, { "epoch": 18.39, "learning_rate": 8.316901408450705e-06, "loss": 0.2819, "step": 16403 }, { "epoch": 18.39, "learning_rate": 8.315492957746479e-06, "loss": 0.3343, "step": 16404 }, { "epoch": 18.39, "learning_rate": 8.314084507042254e-06, "loss": 0.2663, "step": 16405 }, { "epoch": 18.39, "learning_rate": 8.312676056338028e-06, "loss": 0.3189, "step": 16406 }, { "epoch": 18.39, "learning_rate": 8.311267605633803e-06, "loss": 0.5595, "step": 16407 }, { "epoch": 18.39, "learning_rate": 8.309859154929578e-06, "loss": 0.5227, "step": 16408 }, { "epoch": 18.4, "learning_rate": 8.308450704225352e-06, "loss": 0.3903, "step": 16409 }, { "epoch": 18.4, "learning_rate": 8.307042253521127e-06, "loss": 0.4148, "step": 16410 }, { "epoch": 18.4, "learning_rate": 8.305633802816902e-06, "loss": 0.3409, "step": 16411 }, { "epoch": 18.4, "learning_rate": 8.304225352112676e-06, "loss": 0.4259, "step": 16412 }, { "epoch": 18.4, "learning_rate": 8.302816901408451e-06, "loss": 0.4296, "step": 16413 }, { "epoch": 18.4, "learning_rate": 8.301408450704226e-06, "loss": 0.4436, "step": 16414 }, { "epoch": 18.4, "learning_rate": 8.3e-06, "loss": 0.3585, "step": 16415 }, { "epoch": 18.4, "learning_rate": 8.298591549295775e-06, "loss": 0.4817, "step": 16416 }, { "epoch": 18.4, "learning_rate": 8.29718309859155e-06, "loss": 0.4407, "step": 16417 }, { "epoch": 18.41, "learning_rate": 8.295774647887324e-06, "loss": 0.3962, "step": 16418 }, { "epoch": 18.41, "learning_rate": 8.294366197183099e-06, "loss": 0.4364, "step": 16419 }, { "epoch": 18.41, "learning_rate": 8.292957746478873e-06, "loss": 0.45, "step": 16420 }, { "epoch": 18.41, "learning_rate": 8.291549295774648e-06, "loss": 0.4116, "step": 16421 }, { "epoch": 18.41, "learning_rate": 8.290140845070423e-06, "loss": 0.4836, "step": 16422 }, { "epoch": 18.41, "learning_rate": 8.288732394366197e-06, "loss": 0.3902, "step": 16423 }, { "epoch": 18.41, "learning_rate": 8.287323943661972e-06, "loss": 0.433, "step": 16424 }, { "epoch": 18.41, "learning_rate": 8.285915492957747e-06, "loss": 0.293, "step": 16425 }, { "epoch": 18.41, "learning_rate": 8.284507042253521e-06, "loss": 0.4481, "step": 16426 }, { "epoch": 18.42, "learning_rate": 8.283098591549296e-06, "loss": 0.3996, "step": 16427 }, { "epoch": 18.42, "learning_rate": 8.28169014084507e-06, "loss": 0.3464, "step": 16428 }, { "epoch": 18.42, "learning_rate": 8.280281690140845e-06, "loss": 0.432, "step": 16429 }, { "epoch": 18.42, "learning_rate": 8.27887323943662e-06, "loss": 0.3573, "step": 16430 }, { "epoch": 18.42, "learning_rate": 8.277464788732394e-06, "loss": 0.4475, "step": 16431 }, { "epoch": 18.42, "learning_rate": 8.276056338028169e-06, "loss": 0.3727, "step": 16432 }, { "epoch": 18.42, "learning_rate": 8.274647887323944e-06, "loss": 0.3089, "step": 16433 }, { "epoch": 18.42, "learning_rate": 8.273239436619718e-06, "loss": 0.4448, "step": 16434 }, { "epoch": 18.42, "learning_rate": 8.271830985915495e-06, "loss": 0.4267, "step": 16435 }, { "epoch": 18.43, "learning_rate": 8.270422535211268e-06, "loss": 0.3616, "step": 16436 }, { "epoch": 18.43, "learning_rate": 8.269014084507042e-06, "loss": 0.4691, "step": 16437 }, { "epoch": 18.43, "learning_rate": 8.267605633802817e-06, "loss": 0.3695, "step": 16438 }, { "epoch": 18.43, "learning_rate": 8.266197183098591e-06, "loss": 0.4686, "step": 16439 }, { "epoch": 18.43, "learning_rate": 8.264788732394366e-06, "loss": 0.3415, "step": 16440 }, { "epoch": 18.43, "learning_rate": 8.26338028169014e-06, "loss": 0.4249, "step": 16441 }, { "epoch": 18.43, "learning_rate": 8.261971830985915e-06, "loss": 0.3901, "step": 16442 }, { "epoch": 18.43, "learning_rate": 8.260563380281692e-06, "loss": 0.2852, "step": 16443 }, { "epoch": 18.43, "learning_rate": 8.259154929577465e-06, "loss": 0.4007, "step": 16444 }, { "epoch": 18.44, "learning_rate": 8.25774647887324e-06, "loss": 0.4048, "step": 16445 }, { "epoch": 18.44, "learning_rate": 8.256338028169014e-06, "loss": 0.4006, "step": 16446 }, { "epoch": 18.44, "learning_rate": 8.254929577464789e-06, "loss": 0.3188, "step": 16447 }, { "epoch": 18.44, "learning_rate": 8.253521126760563e-06, "loss": 0.4412, "step": 16448 }, { "epoch": 18.44, "learning_rate": 8.252112676056338e-06, "loss": 0.3188, "step": 16449 }, { "epoch": 18.44, "learning_rate": 8.250704225352114e-06, "loss": 0.3872, "step": 16450 }, { "epoch": 18.44, "learning_rate": 8.249295774647889e-06, "loss": 0.3398, "step": 16451 }, { "epoch": 18.44, "learning_rate": 8.247887323943662e-06, "loss": 0.4016, "step": 16452 }, { "epoch": 18.45, "learning_rate": 8.246478873239436e-06, "loss": 0.2759, "step": 16453 }, { "epoch": 18.45, "learning_rate": 8.245070422535211e-06, "loss": 0.37, "step": 16454 }, { "epoch": 18.45, "learning_rate": 8.243661971830986e-06, "loss": 0.3678, "step": 16455 }, { "epoch": 18.45, "learning_rate": 8.24225352112676e-06, "loss": 0.4642, "step": 16456 }, { "epoch": 18.45, "learning_rate": 8.240845070422535e-06, "loss": 0.5374, "step": 16457 }, { "epoch": 18.45, "learning_rate": 8.239436619718311e-06, "loss": 0.4385, "step": 16458 }, { "epoch": 18.45, "learning_rate": 8.238028169014086e-06, "loss": 0.4171, "step": 16459 }, { "epoch": 18.45, "learning_rate": 8.236619718309859e-06, "loss": 0.5175, "step": 16460 }, { "epoch": 18.45, "learning_rate": 8.235211267605633e-06, "loss": 0.5583, "step": 16461 }, { "epoch": 18.46, "learning_rate": 8.233802816901408e-06, "loss": 0.4153, "step": 16462 }, { "epoch": 18.46, "learning_rate": 8.232394366197183e-06, "loss": 0.4377, "step": 16463 }, { "epoch": 18.46, "learning_rate": 8.230985915492957e-06, "loss": 0.4925, "step": 16464 }, { "epoch": 18.46, "learning_rate": 8.229577464788734e-06, "loss": 0.4636, "step": 16465 }, { "epoch": 18.46, "learning_rate": 8.228169014084508e-06, "loss": 0.4017, "step": 16466 }, { "epoch": 18.46, "learning_rate": 8.226760563380283e-06, "loss": 0.4361, "step": 16467 }, { "epoch": 18.46, "learning_rate": 8.225352112676056e-06, "loss": 0.353, "step": 16468 }, { "epoch": 18.46, "learning_rate": 8.22394366197183e-06, "loss": 0.427, "step": 16469 }, { "epoch": 18.46, "learning_rate": 8.222535211267605e-06, "loss": 0.425, "step": 16470 }, { "epoch": 18.47, "learning_rate": 8.22112676056338e-06, "loss": 0.4852, "step": 16471 }, { "epoch": 18.47, "learning_rate": 8.219718309859154e-06, "loss": 0.3178, "step": 16472 }, { "epoch": 18.47, "learning_rate": 8.21830985915493e-06, "loss": 0.4581, "step": 16473 }, { "epoch": 18.47, "learning_rate": 8.216901408450705e-06, "loss": 0.3803, "step": 16474 }, { "epoch": 18.47, "learning_rate": 8.21549295774648e-06, "loss": 0.3647, "step": 16475 }, { "epoch": 18.47, "learning_rate": 8.214084507042253e-06, "loss": 0.3868, "step": 16476 }, { "epoch": 18.47, "learning_rate": 8.212676056338028e-06, "loss": 0.3017, "step": 16477 }, { "epoch": 18.47, "learning_rate": 8.211267605633802e-06, "loss": 0.452, "step": 16478 }, { "epoch": 18.47, "learning_rate": 8.209859154929577e-06, "loss": 0.3264, "step": 16479 }, { "epoch": 18.48, "learning_rate": 8.208450704225353e-06, "loss": 0.3585, "step": 16480 }, { "epoch": 18.48, "learning_rate": 8.207042253521128e-06, "loss": 0.298, "step": 16481 }, { "epoch": 18.48, "learning_rate": 8.205633802816903e-06, "loss": 0.5046, "step": 16482 }, { "epoch": 18.48, "learning_rate": 8.204225352112677e-06, "loss": 0.3122, "step": 16483 }, { "epoch": 18.48, "learning_rate": 8.20281690140845e-06, "loss": 0.3459, "step": 16484 }, { "epoch": 18.48, "learning_rate": 8.201408450704225e-06, "loss": 0.4393, "step": 16485 }, { "epoch": 18.48, "learning_rate": 8.2e-06, "loss": 0.4681, "step": 16486 }, { "epoch": 18.48, "learning_rate": 8.198591549295774e-06, "loss": 0.5594, "step": 16487 }, { "epoch": 18.48, "learning_rate": 8.19718309859155e-06, "loss": 0.4181, "step": 16488 }, { "epoch": 18.49, "learning_rate": 8.195774647887325e-06, "loss": 0.3266, "step": 16489 }, { "epoch": 18.49, "learning_rate": 8.1943661971831e-06, "loss": 0.3371, "step": 16490 }, { "epoch": 18.49, "learning_rate": 8.192957746478874e-06, "loss": 0.3552, "step": 16491 }, { "epoch": 18.49, "learning_rate": 8.191549295774647e-06, "loss": 0.3437, "step": 16492 }, { "epoch": 18.49, "learning_rate": 8.190140845070422e-06, "loss": 0.401, "step": 16493 }, { "epoch": 18.49, "learning_rate": 8.188732394366197e-06, "loss": 0.3871, "step": 16494 }, { "epoch": 18.49, "learning_rate": 8.187323943661971e-06, "loss": 0.2835, "step": 16495 }, { "epoch": 18.49, "learning_rate": 8.185915492957747e-06, "loss": 0.374, "step": 16496 }, { "epoch": 18.49, "learning_rate": 8.184507042253522e-06, "loss": 0.2843, "step": 16497 }, { "epoch": 18.5, "learning_rate": 8.183098591549297e-06, "loss": 0.4754, "step": 16498 }, { "epoch": 18.5, "learning_rate": 8.181690140845071e-06, "loss": 0.299, "step": 16499 }, { "epoch": 18.5, "learning_rate": 8.180281690140844e-06, "loss": 0.3902, "step": 16500 }, { "epoch": 18.5, "eval_loss": 0.7765135765075684, "eval_runtime": 828.6657, "eval_samples_per_second": 3.188, "eval_steps_per_second": 0.399, "eval_wer": 0.13639428798095993, "step": 16500 }, { "epoch": 18.5, "learning_rate": 8.178873239436619e-06, "loss": 0.3373, "step": 16501 }, { "epoch": 18.5, "learning_rate": 8.177464788732394e-06, "loss": 0.4043, "step": 16502 }, { "epoch": 18.5, "learning_rate": 8.17605633802817e-06, "loss": 0.3023, "step": 16503 }, { "epoch": 18.5, "learning_rate": 8.174647887323945e-06, "loss": 0.4282, "step": 16504 }, { "epoch": 18.5, "learning_rate": 8.17323943661972e-06, "loss": 0.3322, "step": 16505 }, { "epoch": 18.5, "learning_rate": 8.171830985915494e-06, "loss": 0.2943, "step": 16506 }, { "epoch": 18.51, "learning_rate": 8.170422535211268e-06, "loss": 0.5555, "step": 16507 }, { "epoch": 18.51, "learning_rate": 8.169014084507041e-06, "loss": 0.4849, "step": 16508 }, { "epoch": 18.51, "learning_rate": 8.167605633802816e-06, "loss": 0.404, "step": 16509 }, { "epoch": 18.51, "learning_rate": 8.16619718309859e-06, "loss": 0.4259, "step": 16510 }, { "epoch": 18.51, "learning_rate": 8.164788732394367e-06, "loss": 0.3939, "step": 16511 }, { "epoch": 18.51, "learning_rate": 8.163380281690142e-06, "loss": 0.4443, "step": 16512 }, { "epoch": 18.51, "learning_rate": 8.161971830985916e-06, "loss": 0.4185, "step": 16513 }, { "epoch": 18.51, "learning_rate": 8.160563380281691e-06, "loss": 0.5047, "step": 16514 }, { "epoch": 18.51, "learning_rate": 8.159154929577466e-06, "loss": 0.4459, "step": 16515 }, { "epoch": 18.52, "learning_rate": 8.157746478873239e-06, "loss": 0.3946, "step": 16516 }, { "epoch": 18.52, "learning_rate": 8.156338028169013e-06, "loss": 0.3767, "step": 16517 }, { "epoch": 18.52, "learning_rate": 8.15492957746479e-06, "loss": 0.4409, "step": 16518 }, { "epoch": 18.52, "learning_rate": 8.153521126760564e-06, "loss": 0.382, "step": 16519 }, { "epoch": 18.52, "learning_rate": 8.152112676056339e-06, "loss": 0.3937, "step": 16520 }, { "epoch": 18.52, "learning_rate": 8.150704225352113e-06, "loss": 0.4731, "step": 16521 }, { "epoch": 18.52, "learning_rate": 8.149295774647888e-06, "loss": 0.3646, "step": 16522 }, { "epoch": 18.52, "learning_rate": 8.147887323943663e-06, "loss": 0.4216, "step": 16523 }, { "epoch": 18.52, "learning_rate": 8.146478873239436e-06, "loss": 0.4216, "step": 16524 }, { "epoch": 18.53, "learning_rate": 8.14507042253521e-06, "loss": 0.4868, "step": 16525 }, { "epoch": 18.53, "learning_rate": 8.143661971830987e-06, "loss": 0.4395, "step": 16526 }, { "epoch": 18.53, "learning_rate": 8.142253521126761e-06, "loss": 0.3852, "step": 16527 }, { "epoch": 18.53, "learning_rate": 8.140845070422536e-06, "loss": 0.3992, "step": 16528 }, { "epoch": 18.53, "learning_rate": 8.13943661971831e-06, "loss": 0.2738, "step": 16529 }, { "epoch": 18.53, "learning_rate": 8.138028169014085e-06, "loss": 0.3554, "step": 16530 }, { "epoch": 18.53, "learning_rate": 8.13661971830986e-06, "loss": 0.4475, "step": 16531 }, { "epoch": 18.53, "learning_rate": 8.135211267605633e-06, "loss": 0.4136, "step": 16532 }, { "epoch": 18.53, "learning_rate": 8.133802816901409e-06, "loss": 0.3347, "step": 16533 }, { "epoch": 18.54, "learning_rate": 8.132394366197184e-06, "loss": 0.3423, "step": 16534 }, { "epoch": 18.54, "learning_rate": 8.130985915492958e-06, "loss": 0.3343, "step": 16535 }, { "epoch": 18.54, "learning_rate": 8.129577464788733e-06, "loss": 0.3945, "step": 16536 }, { "epoch": 18.54, "learning_rate": 8.128169014084508e-06, "loss": 0.4725, "step": 16537 }, { "epoch": 18.54, "learning_rate": 8.126760563380282e-06, "loss": 0.3582, "step": 16538 }, { "epoch": 18.54, "learning_rate": 8.125352112676055e-06, "loss": 0.4568, "step": 16539 }, { "epoch": 18.54, "learning_rate": 8.12394366197183e-06, "loss": 0.3573, "step": 16540 }, { "epoch": 18.54, "learning_rate": 8.122535211267606e-06, "loss": 0.3302, "step": 16541 }, { "epoch": 18.54, "learning_rate": 8.12112676056338e-06, "loss": 0.3454, "step": 16542 }, { "epoch": 18.55, "learning_rate": 8.119718309859155e-06, "loss": 0.3251, "step": 16543 }, { "epoch": 18.55, "learning_rate": 8.11830985915493e-06, "loss": 0.3741, "step": 16544 }, { "epoch": 18.55, "learning_rate": 8.116901408450705e-06, "loss": 0.3093, "step": 16545 }, { "epoch": 18.55, "learning_rate": 8.11549295774648e-06, "loss": 0.3447, "step": 16546 }, { "epoch": 18.55, "learning_rate": 8.114084507042252e-06, "loss": 0.3973, "step": 16547 }, { "epoch": 18.55, "learning_rate": 8.112676056338027e-06, "loss": 0.3396, "step": 16548 }, { "epoch": 18.55, "learning_rate": 8.111267605633803e-06, "loss": 0.479, "step": 16549 }, { "epoch": 18.55, "learning_rate": 8.109859154929578e-06, "loss": 0.3682, "step": 16550 }, { "epoch": 18.55, "learning_rate": 8.108450704225353e-06, "loss": 0.3334, "step": 16551 }, { "epoch": 18.56, "learning_rate": 8.107042253521127e-06, "loss": 0.4957, "step": 16552 }, { "epoch": 18.56, "learning_rate": 8.105633802816902e-06, "loss": 0.3063, "step": 16553 }, { "epoch": 18.56, "learning_rate": 8.104225352112676e-06, "loss": 0.2786, "step": 16554 }, { "epoch": 18.56, "learning_rate": 8.10281690140845e-06, "loss": 0.3608, "step": 16555 }, { "epoch": 18.56, "learning_rate": 8.101408450704226e-06, "loss": 0.4704, "step": 16556 }, { "epoch": 18.56, "learning_rate": 8.1e-06, "loss": 0.5123, "step": 16557 }, { "epoch": 18.56, "learning_rate": 8.098591549295775e-06, "loss": 0.5491, "step": 16558 }, { "epoch": 18.56, "learning_rate": 8.09718309859155e-06, "loss": 0.5239, "step": 16559 }, { "epoch": 18.57, "learning_rate": 8.095774647887324e-06, "loss": 0.4985, "step": 16560 }, { "epoch": 18.57, "learning_rate": 8.094366197183099e-06, "loss": 0.335, "step": 16561 }, { "epoch": 18.57, "learning_rate": 8.092957746478874e-06, "loss": 0.3986, "step": 16562 }, { "epoch": 18.57, "learning_rate": 8.091549295774647e-06, "loss": 0.4243, "step": 16563 }, { "epoch": 18.57, "learning_rate": 8.090140845070423e-06, "loss": 0.4268, "step": 16564 }, { "epoch": 18.57, "learning_rate": 8.088732394366197e-06, "loss": 0.4113, "step": 16565 }, { "epoch": 18.57, "learning_rate": 8.087323943661972e-06, "loss": 0.3866, "step": 16566 }, { "epoch": 18.57, "learning_rate": 8.085915492957747e-06, "loss": 0.3726, "step": 16567 }, { "epoch": 18.57, "learning_rate": 8.084507042253521e-06, "loss": 0.3647, "step": 16568 }, { "epoch": 18.58, "learning_rate": 8.083098591549296e-06, "loss": 0.4187, "step": 16569 }, { "epoch": 18.58, "learning_rate": 8.08169014084507e-06, "loss": 0.3668, "step": 16570 }, { "epoch": 18.58, "learning_rate": 8.080281690140845e-06, "loss": 0.3778, "step": 16571 }, { "epoch": 18.58, "learning_rate": 8.07887323943662e-06, "loss": 0.3399, "step": 16572 }, { "epoch": 18.58, "learning_rate": 8.077464788732395e-06, "loss": 0.4916, "step": 16573 }, { "epoch": 18.58, "learning_rate": 8.07605633802817e-06, "loss": 0.457, "step": 16574 }, { "epoch": 18.58, "learning_rate": 8.074647887323944e-06, "loss": 0.4818, "step": 16575 }, { "epoch": 18.58, "learning_rate": 8.073239436619718e-06, "loss": 0.4585, "step": 16576 }, { "epoch": 18.58, "learning_rate": 8.071830985915493e-06, "loss": 0.4581, "step": 16577 }, { "epoch": 18.59, "learning_rate": 8.070422535211268e-06, "loss": 0.4533, "step": 16578 }, { "epoch": 18.59, "learning_rate": 8.069014084507042e-06, "loss": 0.4164, "step": 16579 }, { "epoch": 18.59, "learning_rate": 8.067605633802817e-06, "loss": 0.2832, "step": 16580 }, { "epoch": 18.59, "learning_rate": 8.066197183098592e-06, "loss": 0.326, "step": 16581 }, { "epoch": 18.59, "learning_rate": 8.064788732394366e-06, "loss": 0.3245, "step": 16582 }, { "epoch": 18.59, "learning_rate": 8.063380281690141e-06, "loss": 0.307, "step": 16583 }, { "epoch": 18.59, "learning_rate": 8.061971830985916e-06, "loss": 0.3507, "step": 16584 }, { "epoch": 18.59, "learning_rate": 8.06056338028169e-06, "loss": 0.3572, "step": 16585 }, { "epoch": 18.59, "learning_rate": 8.059154929577467e-06, "loss": 0.4442, "step": 16586 }, { "epoch": 18.6, "learning_rate": 8.05774647887324e-06, "loss": 0.3493, "step": 16587 }, { "epoch": 18.6, "learning_rate": 8.056338028169014e-06, "loss": 0.3328, "step": 16588 }, { "epoch": 18.6, "learning_rate": 8.054929577464789e-06, "loss": 0.3355, "step": 16589 }, { "epoch": 18.6, "learning_rate": 8.053521126760563e-06, "loss": 0.4187, "step": 16590 }, { "epoch": 18.6, "learning_rate": 8.052112676056338e-06, "loss": 0.4207, "step": 16591 }, { "epoch": 18.6, "learning_rate": 8.050704225352113e-06, "loss": 0.3212, "step": 16592 }, { "epoch": 18.6, "learning_rate": 8.049295774647887e-06, "loss": 0.4602, "step": 16593 }, { "epoch": 18.6, "learning_rate": 8.047887323943664e-06, "loss": 0.2882, "step": 16594 }, { "epoch": 18.6, "learning_rate": 8.046478873239437e-06, "loss": 0.2912, "step": 16595 }, { "epoch": 18.61, "learning_rate": 8.045070422535211e-06, "loss": 0.3807, "step": 16596 }, { "epoch": 18.61, "learning_rate": 8.043661971830986e-06, "loss": 0.319, "step": 16597 }, { "epoch": 18.61, "learning_rate": 8.04225352112676e-06, "loss": 0.4712, "step": 16598 }, { "epoch": 18.61, "learning_rate": 8.040845070422535e-06, "loss": 0.275, "step": 16599 }, { "epoch": 18.61, "learning_rate": 8.03943661971831e-06, "loss": 0.266, "step": 16600 }, { "epoch": 18.61, "learning_rate": 8.038028169014084e-06, "loss": 0.4927, "step": 16601 }, { "epoch": 18.61, "learning_rate": 8.03661971830986e-06, "loss": 0.4158, "step": 16602 }, { "epoch": 18.61, "learning_rate": 8.035211267605634e-06, "loss": 0.3021, "step": 16603 }, { "epoch": 18.61, "learning_rate": 8.033802816901408e-06, "loss": 0.2939, "step": 16604 }, { "epoch": 18.62, "learning_rate": 8.032394366197183e-06, "loss": 0.2245, "step": 16605 }, { "epoch": 18.62, "learning_rate": 8.030985915492958e-06, "loss": 0.2853, "step": 16606 }, { "epoch": 18.62, "learning_rate": 8.029577464788732e-06, "loss": 0.5517, "step": 16607 }, { "epoch": 18.62, "learning_rate": 8.028169014084507e-06, "loss": 0.5871, "step": 16608 }, { "epoch": 18.62, "learning_rate": 8.026760563380283e-06, "loss": 0.4431, "step": 16609 }, { "epoch": 18.62, "learning_rate": 8.025352112676058e-06, "loss": 0.5043, "step": 16610 }, { "epoch": 18.62, "learning_rate": 8.02394366197183e-06, "loss": 0.4173, "step": 16611 }, { "epoch": 18.62, "learning_rate": 8.022535211267605e-06, "loss": 0.3448, "step": 16612 }, { "epoch": 18.62, "learning_rate": 8.02112676056338e-06, "loss": 0.4495, "step": 16613 }, { "epoch": 18.63, "learning_rate": 8.019718309859155e-06, "loss": 0.4471, "step": 16614 }, { "epoch": 18.63, "learning_rate": 8.01830985915493e-06, "loss": 0.5813, "step": 16615 }, { "epoch": 18.63, "learning_rate": 8.016901408450704e-06, "loss": 0.4597, "step": 16616 }, { "epoch": 18.63, "learning_rate": 8.01549295774648e-06, "loss": 0.3792, "step": 16617 }, { "epoch": 18.63, "learning_rate": 8.014084507042255e-06, "loss": 0.4262, "step": 16618 }, { "epoch": 18.63, "learning_rate": 8.012676056338028e-06, "loss": 0.4741, "step": 16619 }, { "epoch": 18.63, "learning_rate": 8.011267605633803e-06, "loss": 0.4701, "step": 16620 }, { "epoch": 18.63, "learning_rate": 8.009859154929577e-06, "loss": 0.4432, "step": 16621 }, { "epoch": 18.63, "learning_rate": 8.008450704225352e-06, "loss": 0.3571, "step": 16622 }, { "epoch": 18.64, "learning_rate": 8.007042253521126e-06, "loss": 0.5585, "step": 16623 }, { "epoch": 18.64, "learning_rate": 8.005633802816903e-06, "loss": 0.4537, "step": 16624 }, { "epoch": 18.64, "learning_rate": 8.004225352112677e-06, "loss": 0.4149, "step": 16625 }, { "epoch": 18.64, "learning_rate": 8.002816901408452e-06, "loss": 0.4454, "step": 16626 }, { "epoch": 18.64, "learning_rate": 8.001408450704225e-06, "loss": 0.4069, "step": 16627 }, { "epoch": 18.64, "learning_rate": 8e-06, "loss": 0.3695, "step": 16628 }, { "epoch": 18.64, "learning_rate": 7.998591549295774e-06, "loss": 0.4966, "step": 16629 }, { "epoch": 18.64, "learning_rate": 7.997183098591549e-06, "loss": 0.4065, "step": 16630 }, { "epoch": 18.64, "learning_rate": 7.995774647887324e-06, "loss": 0.3608, "step": 16631 }, { "epoch": 18.65, "learning_rate": 7.9943661971831e-06, "loss": 0.3258, "step": 16632 }, { "epoch": 18.65, "learning_rate": 7.992957746478875e-06, "loss": 0.3919, "step": 16633 }, { "epoch": 18.65, "learning_rate": 7.99154929577465e-06, "loss": 0.3134, "step": 16634 }, { "epoch": 18.65, "learning_rate": 7.990140845070422e-06, "loss": 0.3642, "step": 16635 }, { "epoch": 18.65, "learning_rate": 7.988732394366197e-06, "loss": 0.4101, "step": 16636 }, { "epoch": 18.65, "learning_rate": 7.987323943661971e-06, "loss": 0.2681, "step": 16637 }, { "epoch": 18.65, "learning_rate": 7.985915492957746e-06, "loss": 0.3891, "step": 16638 }, { "epoch": 18.65, "learning_rate": 7.984507042253522e-06, "loss": 0.3171, "step": 16639 }, { "epoch": 18.65, "learning_rate": 7.983098591549297e-06, "loss": 0.305, "step": 16640 }, { "epoch": 18.66, "learning_rate": 7.981690140845072e-06, "loss": 0.347, "step": 16641 }, { "epoch": 18.66, "learning_rate": 7.980281690140845e-06, "loss": 0.342, "step": 16642 }, { "epoch": 18.66, "learning_rate": 7.97887323943662e-06, "loss": 0.4828, "step": 16643 }, { "epoch": 18.66, "learning_rate": 7.977464788732394e-06, "loss": 0.3747, "step": 16644 }, { "epoch": 18.66, "learning_rate": 7.976056338028168e-06, "loss": 0.3801, "step": 16645 }, { "epoch": 18.66, "learning_rate": 7.974647887323943e-06, "loss": 0.4301, "step": 16646 }, { "epoch": 18.66, "learning_rate": 7.97323943661972e-06, "loss": 0.3537, "step": 16647 }, { "epoch": 18.66, "learning_rate": 7.971830985915494e-06, "loss": 0.3977, "step": 16648 }, { "epoch": 18.66, "learning_rate": 7.970422535211269e-06, "loss": 0.3195, "step": 16649 }, { "epoch": 18.67, "learning_rate": 7.969014084507042e-06, "loss": 0.3816, "step": 16650 }, { "epoch": 18.67, "learning_rate": 7.967605633802816e-06, "loss": 0.3543, "step": 16651 }, { "epoch": 18.67, "learning_rate": 7.966197183098591e-06, "loss": 0.318, "step": 16652 }, { "epoch": 18.67, "learning_rate": 7.964788732394366e-06, "loss": 0.2648, "step": 16653 }, { "epoch": 18.67, "learning_rate": 7.963380281690142e-06, "loss": 0.296, "step": 16654 }, { "epoch": 18.67, "learning_rate": 7.961971830985917e-06, "loss": 0.288, "step": 16655 }, { "epoch": 18.67, "learning_rate": 7.960563380281691e-06, "loss": 0.2788, "step": 16656 }, { "epoch": 18.67, "learning_rate": 7.959154929577466e-06, "loss": 0.5334, "step": 16657 }, { "epoch": 18.67, "learning_rate": 7.957746478873239e-06, "loss": 0.3899, "step": 16658 }, { "epoch": 18.68, "learning_rate": 7.956338028169013e-06, "loss": 0.5784, "step": 16659 }, { "epoch": 18.68, "learning_rate": 7.954929577464788e-06, "loss": 0.5552, "step": 16660 }, { "epoch": 18.68, "learning_rate": 7.953521126760563e-06, "loss": 0.3658, "step": 16661 }, { "epoch": 18.68, "learning_rate": 7.952112676056339e-06, "loss": 0.4674, "step": 16662 }, { "epoch": 18.68, "learning_rate": 7.950704225352114e-06, "loss": 0.4993, "step": 16663 }, { "epoch": 18.68, "learning_rate": 7.949295774647888e-06, "loss": 0.4138, "step": 16664 }, { "epoch": 18.68, "learning_rate": 7.947887323943663e-06, "loss": 0.4293, "step": 16665 }, { "epoch": 18.68, "learning_rate": 7.946478873239436e-06, "loss": 0.3952, "step": 16666 }, { "epoch": 18.68, "learning_rate": 7.94507042253521e-06, "loss": 0.4072, "step": 16667 }, { "epoch": 18.69, "learning_rate": 7.943661971830985e-06, "loss": 0.3864, "step": 16668 }, { "epoch": 18.69, "learning_rate": 7.94225352112676e-06, "loss": 0.4836, "step": 16669 }, { "epoch": 18.69, "learning_rate": 7.940845070422536e-06, "loss": 0.3364, "step": 16670 }, { "epoch": 18.69, "learning_rate": 7.93943661971831e-06, "loss": 0.3736, "step": 16671 }, { "epoch": 18.69, "learning_rate": 7.938028169014085e-06, "loss": 0.4186, "step": 16672 }, { "epoch": 18.69, "learning_rate": 7.93661971830986e-06, "loss": 0.4103, "step": 16673 }, { "epoch": 18.69, "learning_rate": 7.935211267605633e-06, "loss": 0.401, "step": 16674 }, { "epoch": 18.69, "learning_rate": 7.933802816901408e-06, "loss": 0.3651, "step": 16675 }, { "epoch": 18.7, "learning_rate": 7.932394366197182e-06, "loss": 0.3128, "step": 16676 }, { "epoch": 18.7, "learning_rate": 7.930985915492959e-06, "loss": 0.3587, "step": 16677 }, { "epoch": 18.7, "learning_rate": 7.929577464788733e-06, "loss": 0.4287, "step": 16678 }, { "epoch": 18.7, "learning_rate": 7.928169014084508e-06, "loss": 0.4794, "step": 16679 }, { "epoch": 18.7, "learning_rate": 7.926760563380282e-06, "loss": 0.4244, "step": 16680 }, { "epoch": 18.7, "learning_rate": 7.925352112676057e-06, "loss": 0.4191, "step": 16681 }, { "epoch": 18.7, "learning_rate": 7.92394366197183e-06, "loss": 0.3746, "step": 16682 }, { "epoch": 18.7, "learning_rate": 7.922535211267605e-06, "loss": 0.3618, "step": 16683 }, { "epoch": 18.7, "learning_rate": 7.92112676056338e-06, "loss": 0.337, "step": 16684 }, { "epoch": 18.71, "learning_rate": 7.919718309859156e-06, "loss": 0.4065, "step": 16685 }, { "epoch": 18.71, "learning_rate": 7.91830985915493e-06, "loss": 0.3177, "step": 16686 }, { "epoch": 18.71, "learning_rate": 7.916901408450705e-06, "loss": 0.3691, "step": 16687 }, { "epoch": 18.71, "learning_rate": 7.91549295774648e-06, "loss": 0.3305, "step": 16688 }, { "epoch": 18.71, "learning_rate": 7.914084507042254e-06, "loss": 0.3563, "step": 16689 }, { "epoch": 18.71, "learning_rate": 7.912676056338027e-06, "loss": 0.4331, "step": 16690 }, { "epoch": 18.71, "learning_rate": 7.911267605633802e-06, "loss": 0.3563, "step": 16691 }, { "epoch": 18.71, "learning_rate": 7.909859154929578e-06, "loss": 0.3671, "step": 16692 }, { "epoch": 18.71, "learning_rate": 7.908450704225353e-06, "loss": 0.3378, "step": 16693 }, { "epoch": 18.72, "learning_rate": 7.907042253521127e-06, "loss": 0.3101, "step": 16694 }, { "epoch": 18.72, "learning_rate": 7.905633802816902e-06, "loss": 0.4535, "step": 16695 }, { "epoch": 18.72, "learning_rate": 7.904225352112677e-06, "loss": 0.3668, "step": 16696 }, { "epoch": 18.72, "learning_rate": 7.902816901408451e-06, "loss": 0.3674, "step": 16697 }, { "epoch": 18.72, "learning_rate": 7.901408450704224e-06, "loss": 0.4333, "step": 16698 }, { "epoch": 18.72, "learning_rate": 7.899999999999999e-06, "loss": 0.2916, "step": 16699 }, { "epoch": 18.72, "learning_rate": 7.898591549295775e-06, "loss": 0.3227, "step": 16700 }, { "epoch": 18.72, "learning_rate": 7.89718309859155e-06, "loss": 0.4717, "step": 16701 }, { "epoch": 18.72, "learning_rate": 7.895774647887325e-06, "loss": 0.4124, "step": 16702 }, { "epoch": 18.73, "learning_rate": 7.894366197183099e-06, "loss": 0.4016, "step": 16703 }, { "epoch": 18.73, "learning_rate": 7.892957746478874e-06, "loss": 0.3243, "step": 16704 }, { "epoch": 18.73, "learning_rate": 7.891549295774648e-06, "loss": 0.3141, "step": 16705 }, { "epoch": 18.73, "learning_rate": 7.890140845070421e-06, "loss": 0.4995, "step": 16706 }, { "epoch": 18.73, "learning_rate": 7.888732394366198e-06, "loss": 0.5276, "step": 16707 }, { "epoch": 18.73, "learning_rate": 7.887323943661972e-06, "loss": 0.3723, "step": 16708 }, { "epoch": 18.73, "learning_rate": 7.885915492957747e-06, "loss": 0.4119, "step": 16709 }, { "epoch": 18.73, "learning_rate": 7.884507042253522e-06, "loss": 0.5341, "step": 16710 }, { "epoch": 18.73, "learning_rate": 7.883098591549296e-06, "loss": 0.3895, "step": 16711 }, { "epoch": 18.74, "learning_rate": 7.881690140845071e-06, "loss": 0.4047, "step": 16712 }, { "epoch": 18.74, "learning_rate": 7.880281690140846e-06, "loss": 0.4435, "step": 16713 }, { "epoch": 18.74, "learning_rate": 7.878873239436618e-06, "loss": 0.3878, "step": 16714 }, { "epoch": 18.74, "learning_rate": 7.877464788732395e-06, "loss": 0.4424, "step": 16715 }, { "epoch": 18.74, "learning_rate": 7.87605633802817e-06, "loss": 0.3572, "step": 16716 }, { "epoch": 18.74, "learning_rate": 7.874647887323944e-06, "loss": 0.375, "step": 16717 }, { "epoch": 18.74, "learning_rate": 7.873239436619719e-06, "loss": 0.4429, "step": 16718 }, { "epoch": 18.74, "learning_rate": 7.871830985915493e-06, "loss": 0.3675, "step": 16719 }, { "epoch": 18.74, "learning_rate": 7.870422535211268e-06, "loss": 0.374, "step": 16720 }, { "epoch": 18.75, "learning_rate": 7.869014084507043e-06, "loss": 0.287, "step": 16721 }, { "epoch": 18.75, "learning_rate": 7.867605633802816e-06, "loss": 0.3631, "step": 16722 }, { "epoch": 18.75, "learning_rate": 7.866197183098592e-06, "loss": 0.4462, "step": 16723 }, { "epoch": 18.75, "learning_rate": 7.864788732394367e-06, "loss": 0.4523, "step": 16724 }, { "epoch": 18.75, "learning_rate": 7.863380281690141e-06, "loss": 0.4322, "step": 16725 }, { "epoch": 18.75, "learning_rate": 7.861971830985916e-06, "loss": 0.4637, "step": 16726 }, { "epoch": 18.75, "learning_rate": 7.86056338028169e-06, "loss": 0.3971, "step": 16727 }, { "epoch": 18.75, "learning_rate": 7.859154929577465e-06, "loss": 0.4324, "step": 16728 }, { "epoch": 18.75, "learning_rate": 7.85774647887324e-06, "loss": 0.4373, "step": 16729 }, { "epoch": 18.76, "learning_rate": 7.856338028169014e-06, "loss": 0.4433, "step": 16730 }, { "epoch": 18.76, "learning_rate": 7.854929577464789e-06, "loss": 0.4377, "step": 16731 }, { "epoch": 18.76, "learning_rate": 7.853521126760564e-06, "loss": 0.3492, "step": 16732 }, { "epoch": 18.76, "learning_rate": 7.852112676056338e-06, "loss": 0.4596, "step": 16733 }, { "epoch": 18.76, "learning_rate": 7.850704225352113e-06, "loss": 0.4599, "step": 16734 }, { "epoch": 18.76, "learning_rate": 7.849295774647888e-06, "loss": 0.4251, "step": 16735 }, { "epoch": 18.76, "learning_rate": 7.847887323943662e-06, "loss": 0.3526, "step": 16736 }, { "epoch": 18.76, "learning_rate": 7.846478873239437e-06, "loss": 0.3227, "step": 16737 }, { "epoch": 18.76, "learning_rate": 7.845070422535211e-06, "loss": 0.4127, "step": 16738 }, { "epoch": 18.77, "learning_rate": 7.843661971830986e-06, "loss": 0.3771, "step": 16739 }, { "epoch": 18.77, "learning_rate": 7.84225352112676e-06, "loss": 0.3742, "step": 16740 }, { "epoch": 18.77, "learning_rate": 7.840845070422535e-06, "loss": 0.4316, "step": 16741 }, { "epoch": 18.77, "learning_rate": 7.83943661971831e-06, "loss": 0.4407, "step": 16742 }, { "epoch": 18.77, "learning_rate": 7.838028169014085e-06, "loss": 0.495, "step": 16743 }, { "epoch": 18.77, "learning_rate": 7.83661971830986e-06, "loss": 0.332, "step": 16744 }, { "epoch": 18.77, "learning_rate": 7.835211267605636e-06, "loss": 0.418, "step": 16745 }, { "epoch": 18.77, "learning_rate": 7.833802816901409e-06, "loss": 0.4376, "step": 16746 }, { "epoch": 18.77, "learning_rate": 7.832394366197183e-06, "loss": 0.4191, "step": 16747 }, { "epoch": 18.78, "learning_rate": 7.830985915492958e-06, "loss": 0.3519, "step": 16748 }, { "epoch": 18.78, "learning_rate": 7.829577464788732e-06, "loss": 0.2747, "step": 16749 }, { "epoch": 18.78, "learning_rate": 7.828169014084507e-06, "loss": 0.2951, "step": 16750 }, { "epoch": 18.78, "learning_rate": 7.826760563380282e-06, "loss": 0.2521, "step": 16751 }, { "epoch": 18.78, "learning_rate": 7.825352112676056e-06, "loss": 0.4356, "step": 16752 }, { "epoch": 18.78, "learning_rate": 7.823943661971831e-06, "loss": 0.3246, "step": 16753 }, { "epoch": 18.78, "learning_rate": 7.822535211267606e-06, "loss": 0.41, "step": 16754 }, { "epoch": 18.78, "learning_rate": 7.82112676056338e-06, "loss": 0.324, "step": 16755 }, { "epoch": 18.78, "learning_rate": 7.819718309859155e-06, "loss": 0.3716, "step": 16756 }, { "epoch": 18.79, "learning_rate": 7.81830985915493e-06, "loss": 0.4956, "step": 16757 }, { "epoch": 18.79, "learning_rate": 7.816901408450704e-06, "loss": 0.554, "step": 16758 }, { "epoch": 18.79, "learning_rate": 7.815492957746479e-06, "loss": 0.451, "step": 16759 }, { "epoch": 18.79, "learning_rate": 7.814084507042255e-06, "loss": 0.4991, "step": 16760 }, { "epoch": 18.79, "learning_rate": 7.812676056338028e-06, "loss": 0.4125, "step": 16761 }, { "epoch": 18.79, "learning_rate": 7.811267605633803e-06, "loss": 0.5175, "step": 16762 }, { "epoch": 18.79, "learning_rate": 7.809859154929577e-06, "loss": 0.3991, "step": 16763 }, { "epoch": 18.79, "learning_rate": 7.808450704225352e-06, "loss": 0.4424, "step": 16764 }, { "epoch": 18.79, "learning_rate": 7.807042253521127e-06, "loss": 0.5022, "step": 16765 }, { "epoch": 18.8, "learning_rate": 7.805633802816901e-06, "loss": 0.391, "step": 16766 }, { "epoch": 18.8, "learning_rate": 7.804225352112676e-06, "loss": 0.4135, "step": 16767 }, { "epoch": 18.8, "learning_rate": 7.802816901408452e-06, "loss": 0.5633, "step": 16768 }, { "epoch": 18.8, "learning_rate": 7.801408450704225e-06, "loss": 0.476, "step": 16769 }, { "epoch": 18.8, "learning_rate": 7.8e-06, "loss": 0.3743, "step": 16770 }, { "epoch": 18.8, "learning_rate": 7.798591549295774e-06, "loss": 0.3595, "step": 16771 }, { "epoch": 18.8, "learning_rate": 7.797183098591549e-06, "loss": 0.3586, "step": 16772 }, { "epoch": 18.8, "learning_rate": 7.795774647887324e-06, "loss": 0.415, "step": 16773 }, { "epoch": 18.8, "learning_rate": 7.794366197183098e-06, "loss": 0.3736, "step": 16774 }, { "epoch": 18.81, "learning_rate": 7.792957746478875e-06, "loss": 0.3122, "step": 16775 }, { "epoch": 18.81, "learning_rate": 7.79154929577465e-06, "loss": 0.4203, "step": 16776 }, { "epoch": 18.81, "learning_rate": 7.790140845070422e-06, "loss": 0.3991, "step": 16777 }, { "epoch": 18.81, "learning_rate": 7.788732394366197e-06, "loss": 0.3877, "step": 16778 }, { "epoch": 18.81, "learning_rate": 7.787323943661972e-06, "loss": 0.3756, "step": 16779 }, { "epoch": 18.81, "learning_rate": 7.785915492957746e-06, "loss": 0.3906, "step": 16780 }, { "epoch": 18.81, "learning_rate": 7.784507042253521e-06, "loss": 0.4352, "step": 16781 }, { "epoch": 18.81, "learning_rate": 7.783098591549296e-06, "loss": 0.3695, "step": 16782 }, { "epoch": 18.82, "learning_rate": 7.781690140845072e-06, "loss": 0.4001, "step": 16783 }, { "epoch": 18.82, "learning_rate": 7.780281690140846e-06, "loss": 0.4004, "step": 16784 }, { "epoch": 18.82, "learning_rate": 7.77887323943662e-06, "loss": 0.372, "step": 16785 }, { "epoch": 18.82, "learning_rate": 7.777464788732394e-06, "loss": 0.3884, "step": 16786 }, { "epoch": 18.82, "learning_rate": 7.776056338028169e-06, "loss": 0.3634, "step": 16787 }, { "epoch": 18.82, "learning_rate": 7.774647887323943e-06, "loss": 0.4315, "step": 16788 }, { "epoch": 18.82, "learning_rate": 7.773239436619718e-06, "loss": 0.4092, "step": 16789 }, { "epoch": 18.82, "learning_rate": 7.771830985915493e-06, "loss": 0.3686, "step": 16790 }, { "epoch": 18.82, "learning_rate": 7.770422535211269e-06, "loss": 0.3672, "step": 16791 }, { "epoch": 18.83, "learning_rate": 7.769014084507044e-06, "loss": 0.3527, "step": 16792 }, { "epoch": 18.83, "learning_rate": 7.767605633802817e-06, "loss": 0.3427, "step": 16793 }, { "epoch": 18.83, "learning_rate": 7.766197183098591e-06, "loss": 0.37, "step": 16794 }, { "epoch": 18.83, "learning_rate": 7.764788732394366e-06, "loss": 0.376, "step": 16795 }, { "epoch": 18.83, "learning_rate": 7.76338028169014e-06, "loss": 0.4244, "step": 16796 }, { "epoch": 18.83, "learning_rate": 7.761971830985915e-06, "loss": 0.4496, "step": 16797 }, { "epoch": 18.83, "learning_rate": 7.760563380281691e-06, "loss": 0.4312, "step": 16798 }, { "epoch": 18.83, "learning_rate": 7.759154929577466e-06, "loss": 0.3787, "step": 16799 }, { "epoch": 18.83, "learning_rate": 7.75774647887324e-06, "loss": 0.3439, "step": 16800 }, { "epoch": 18.84, "learning_rate": 7.756338028169014e-06, "loss": 0.3216, "step": 16801 }, { "epoch": 18.84, "learning_rate": 7.754929577464788e-06, "loss": 0.3383, "step": 16802 }, { "epoch": 18.84, "learning_rate": 7.753521126760563e-06, "loss": 0.2995, "step": 16803 }, { "epoch": 18.84, "learning_rate": 7.752112676056338e-06, "loss": 0.3531, "step": 16804 }, { "epoch": 18.84, "learning_rate": 7.750704225352112e-06, "loss": 0.4255, "step": 16805 }, { "epoch": 18.84, "learning_rate": 7.749295774647889e-06, "loss": 0.377, "step": 16806 }, { "epoch": 18.84, "learning_rate": 7.747887323943663e-06, "loss": 0.5176, "step": 16807 }, { "epoch": 18.84, "learning_rate": 7.746478873239438e-06, "loss": 0.5113, "step": 16808 }, { "epoch": 18.84, "learning_rate": 7.74507042253521e-06, "loss": 0.42, "step": 16809 }, { "epoch": 18.85, "learning_rate": 7.743661971830985e-06, "loss": 0.4252, "step": 16810 }, { "epoch": 18.85, "learning_rate": 7.74225352112676e-06, "loss": 0.4683, "step": 16811 }, { "epoch": 18.85, "learning_rate": 7.740845070422535e-06, "loss": 0.477, "step": 16812 }, { "epoch": 18.85, "learning_rate": 7.739436619718311e-06, "loss": 0.4328, "step": 16813 }, { "epoch": 18.85, "learning_rate": 7.738028169014086e-06, "loss": 0.4348, "step": 16814 }, { "epoch": 18.85, "learning_rate": 7.73661971830986e-06, "loss": 0.4068, "step": 16815 }, { "epoch": 18.85, "learning_rate": 7.735211267605635e-06, "loss": 0.4584, "step": 16816 }, { "epoch": 18.85, "learning_rate": 7.733802816901408e-06, "loss": 0.3312, "step": 16817 }, { "epoch": 18.85, "learning_rate": 7.732394366197182e-06, "loss": 0.3812, "step": 16818 }, { "epoch": 18.86, "learning_rate": 7.730985915492957e-06, "loss": 0.3857, "step": 16819 }, { "epoch": 18.86, "learning_rate": 7.729577464788732e-06, "loss": 0.3657, "step": 16820 }, { "epoch": 18.86, "learning_rate": 7.728169014084508e-06, "loss": 0.3285, "step": 16821 }, { "epoch": 18.86, "learning_rate": 7.726760563380283e-06, "loss": 0.4103, "step": 16822 }, { "epoch": 18.86, "learning_rate": 7.725352112676057e-06, "loss": 0.4306, "step": 16823 }, { "epoch": 18.86, "learning_rate": 7.723943661971832e-06, "loss": 0.423, "step": 16824 }, { "epoch": 18.86, "learning_rate": 7.722535211267605e-06, "loss": 0.2757, "step": 16825 }, { "epoch": 18.86, "learning_rate": 7.72112676056338e-06, "loss": 0.3243, "step": 16826 }, { "epoch": 18.86, "learning_rate": 7.719718309859154e-06, "loss": 0.4288, "step": 16827 }, { "epoch": 18.87, "learning_rate": 7.71830985915493e-06, "loss": 0.4204, "step": 16828 }, { "epoch": 18.87, "learning_rate": 7.716901408450705e-06, "loss": 0.3171, "step": 16829 }, { "epoch": 18.87, "learning_rate": 7.71549295774648e-06, "loss": 0.4151, "step": 16830 }, { "epoch": 18.87, "learning_rate": 7.714084507042254e-06, "loss": 0.3833, "step": 16831 }, { "epoch": 18.87, "learning_rate": 7.712676056338029e-06, "loss": 0.4254, "step": 16832 }, { "epoch": 18.87, "learning_rate": 7.711267605633802e-06, "loss": 0.3365, "step": 16833 }, { "epoch": 18.87, "learning_rate": 7.709859154929577e-06, "loss": 0.3901, "step": 16834 }, { "epoch": 18.87, "learning_rate": 7.708450704225351e-06, "loss": 0.3223, "step": 16835 }, { "epoch": 18.87, "learning_rate": 7.707042253521128e-06, "loss": 0.3585, "step": 16836 }, { "epoch": 18.88, "learning_rate": 7.705633802816902e-06, "loss": 0.4388, "step": 16837 }, { "epoch": 18.88, "learning_rate": 7.704225352112677e-06, "loss": 0.3672, "step": 16838 }, { "epoch": 18.88, "learning_rate": 7.702816901408452e-06, "loss": 0.3695, "step": 16839 }, { "epoch": 18.88, "learning_rate": 7.701408450704226e-06, "loss": 0.4682, "step": 16840 }, { "epoch": 18.88, "learning_rate": 7.699999999999999e-06, "loss": 0.4094, "step": 16841 }, { "epoch": 18.88, "learning_rate": 7.698591549295774e-06, "loss": 0.5132, "step": 16842 }, { "epoch": 18.88, "learning_rate": 7.697183098591548e-06, "loss": 0.4783, "step": 16843 }, { "epoch": 18.88, "learning_rate": 7.695774647887325e-06, "loss": 0.4142, "step": 16844 }, { "epoch": 18.88, "learning_rate": 7.6943661971831e-06, "loss": 0.3814, "step": 16845 }, { "epoch": 18.89, "learning_rate": 7.692957746478874e-06, "loss": 0.3955, "step": 16846 }, { "epoch": 18.89, "learning_rate": 7.691549295774649e-06, "loss": 0.4322, "step": 16847 }, { "epoch": 18.89, "learning_rate": 7.690140845070423e-06, "loss": 0.4316, "step": 16848 }, { "epoch": 18.89, "learning_rate": 7.688732394366196e-06, "loss": 0.4376, "step": 16849 }, { "epoch": 18.89, "learning_rate": 7.687323943661971e-06, "loss": 0.2782, "step": 16850 }, { "epoch": 18.89, "learning_rate": 7.685915492957747e-06, "loss": 0.4957, "step": 16851 }, { "epoch": 18.89, "learning_rate": 7.684507042253522e-06, "loss": 0.3583, "step": 16852 }, { "epoch": 18.89, "learning_rate": 7.683098591549296e-06, "loss": 0.3018, "step": 16853 }, { "epoch": 18.89, "learning_rate": 7.681690140845071e-06, "loss": 0.223, "step": 16854 }, { "epoch": 18.9, "learning_rate": 7.680281690140846e-06, "loss": 0.4161, "step": 16855 }, { "epoch": 18.9, "learning_rate": 7.67887323943662e-06, "loss": 0.3382, "step": 16856 }, { "epoch": 18.9, "learning_rate": 7.677464788732393e-06, "loss": 0.5328, "step": 16857 }, { "epoch": 18.9, "learning_rate": 7.676056338028168e-06, "loss": 0.4969, "step": 16858 }, { "epoch": 18.9, "learning_rate": 7.674647887323944e-06, "loss": 0.3797, "step": 16859 }, { "epoch": 18.9, "learning_rate": 7.673239436619719e-06, "loss": 0.4333, "step": 16860 }, { "epoch": 18.9, "learning_rate": 7.671830985915494e-06, "loss": 0.5071, "step": 16861 }, { "epoch": 18.9, "learning_rate": 7.670422535211268e-06, "loss": 0.5196, "step": 16862 }, { "epoch": 18.9, "learning_rate": 7.669014084507043e-06, "loss": 0.4066, "step": 16863 }, { "epoch": 18.91, "learning_rate": 7.667605633802817e-06, "loss": 0.4435, "step": 16864 }, { "epoch": 18.91, "learning_rate": 7.66619718309859e-06, "loss": 0.3724, "step": 16865 }, { "epoch": 18.91, "learning_rate": 7.664788732394367e-06, "loss": 0.3389, "step": 16866 }, { "epoch": 18.91, "learning_rate": 7.663380281690141e-06, "loss": 0.4039, "step": 16867 }, { "epoch": 18.91, "learning_rate": 7.661971830985916e-06, "loss": 0.3612, "step": 16868 }, { "epoch": 18.91, "learning_rate": 7.66056338028169e-06, "loss": 0.4367, "step": 16869 }, { "epoch": 18.91, "learning_rate": 7.659154929577465e-06, "loss": 0.3513, "step": 16870 }, { "epoch": 18.91, "learning_rate": 7.65774647887324e-06, "loss": 0.4297, "step": 16871 }, { "epoch": 18.91, "learning_rate": 7.656338028169015e-06, "loss": 0.303, "step": 16872 }, { "epoch": 18.92, "learning_rate": 7.654929577464788e-06, "loss": 0.441, "step": 16873 }, { "epoch": 18.92, "learning_rate": 7.653521126760564e-06, "loss": 0.398, "step": 16874 }, { "epoch": 18.92, "learning_rate": 7.652112676056338e-06, "loss": 0.3644, "step": 16875 }, { "epoch": 18.92, "learning_rate": 7.650704225352113e-06, "loss": 0.4278, "step": 16876 }, { "epoch": 18.92, "learning_rate": 7.649295774647888e-06, "loss": 0.4478, "step": 16877 }, { "epoch": 18.92, "learning_rate": 7.647887323943662e-06, "loss": 0.4091, "step": 16878 }, { "epoch": 18.92, "learning_rate": 7.646478873239437e-06, "loss": 0.3439, "step": 16879 }, { "epoch": 18.92, "learning_rate": 7.64507042253521e-06, "loss": 0.4099, "step": 16880 }, { "epoch": 18.92, "learning_rate": 7.643661971830986e-06, "loss": 0.3316, "step": 16881 }, { "epoch": 18.93, "learning_rate": 7.642253521126761e-06, "loss": 0.389, "step": 16882 }, { "epoch": 18.93, "learning_rate": 7.640845070422536e-06, "loss": 0.4103, "step": 16883 }, { "epoch": 18.93, "learning_rate": 7.63943661971831e-06, "loss": 0.3654, "step": 16884 }, { "epoch": 18.93, "learning_rate": 7.638028169014085e-06, "loss": 0.4344, "step": 16885 }, { "epoch": 18.93, "learning_rate": 7.63661971830986e-06, "loss": 0.3474, "step": 16886 }, { "epoch": 18.93, "learning_rate": 7.635211267605634e-06, "loss": 0.4622, "step": 16887 }, { "epoch": 18.93, "learning_rate": 7.633802816901407e-06, "loss": 0.375, "step": 16888 }, { "epoch": 18.93, "learning_rate": 7.632394366197183e-06, "loss": 0.354, "step": 16889 }, { "epoch": 18.93, "learning_rate": 7.630985915492958e-06, "loss": 0.3647, "step": 16890 }, { "epoch": 18.94, "learning_rate": 7.629577464788733e-06, "loss": 0.4209, "step": 16891 }, { "epoch": 18.94, "learning_rate": 7.628169014084507e-06, "loss": 0.4266, "step": 16892 }, { "epoch": 18.94, "learning_rate": 7.626760563380282e-06, "loss": 0.3426, "step": 16893 }, { "epoch": 18.94, "learning_rate": 7.625352112676056e-06, "loss": 0.3525, "step": 16894 }, { "epoch": 18.94, "learning_rate": 7.62394366197183e-06, "loss": 0.3514, "step": 16895 }, { "epoch": 18.94, "learning_rate": 7.622535211267605e-06, "loss": 0.3692, "step": 16896 }, { "epoch": 18.94, "learning_rate": 7.621126760563381e-06, "loss": 0.2957, "step": 16897 }, { "epoch": 18.94, "learning_rate": 7.619718309859155e-06, "loss": 0.299, "step": 16898 }, { "epoch": 18.95, "learning_rate": 7.61830985915493e-06, "loss": 0.4927, "step": 16899 }, { "epoch": 18.95, "learning_rate": 7.6169014084507044e-06, "loss": 0.4516, "step": 16900 }, { "epoch": 18.95, "learning_rate": 7.615492957746479e-06, "loss": 0.3197, "step": 16901 }, { "epoch": 18.95, "learning_rate": 7.614084507042253e-06, "loss": 0.4492, "step": 16902 }, { "epoch": 18.95, "learning_rate": 7.6126760563380275e-06, "loss": 0.375, "step": 16903 }, { "epoch": 18.95, "learning_rate": 7.611267605633804e-06, "loss": 0.3996, "step": 16904 }, { "epoch": 18.95, "learning_rate": 7.6098591549295785e-06, "loss": 0.5216, "step": 16905 }, { "epoch": 18.95, "learning_rate": 7.608450704225352e-06, "loss": 0.2266, "step": 16906 }, { "epoch": 18.95, "learning_rate": 7.607042253521127e-06, "loss": 0.4913, "step": 16907 }, { "epoch": 18.96, "learning_rate": 7.6056338028169015e-06, "loss": 0.4503, "step": 16908 }, { "epoch": 18.96, "learning_rate": 7.604225352112676e-06, "loss": 0.4502, "step": 16909 }, { "epoch": 18.96, "learning_rate": 7.60281690140845e-06, "loss": 0.4022, "step": 16910 }, { "epoch": 18.96, "learning_rate": 7.601408450704225e-06, "loss": 0.3564, "step": 16911 }, { "epoch": 18.96, "learning_rate": 7.600000000000001e-06, "loss": 0.4071, "step": 16912 }, { "epoch": 18.96, "learning_rate": 7.5985915492957756e-06, "loss": 0.452, "step": 16913 }, { "epoch": 18.96, "learning_rate": 7.597183098591549e-06, "loss": 0.3643, "step": 16914 }, { "epoch": 18.96, "learning_rate": 7.595774647887324e-06, "loss": 0.4662, "step": 16915 }, { "epoch": 18.96, "learning_rate": 7.594366197183099e-06, "loss": 0.3936, "step": 16916 }, { "epoch": 18.97, "learning_rate": 7.592957746478873e-06, "loss": 0.3855, "step": 16917 }, { "epoch": 18.97, "learning_rate": 7.591549295774647e-06, "loss": 0.4756, "step": 16918 }, { "epoch": 18.97, "learning_rate": 7.590140845070423e-06, "loss": 0.5045, "step": 16919 }, { "epoch": 18.97, "learning_rate": 7.588732394366198e-06, "loss": 0.3672, "step": 16920 }, { "epoch": 18.97, "learning_rate": 7.587323943661973e-06, "loss": 0.3121, "step": 16921 }, { "epoch": 18.97, "learning_rate": 7.5859154929577465e-06, "loss": 0.321, "step": 16922 }, { "epoch": 18.97, "learning_rate": 7.584507042253521e-06, "loss": 0.3756, "step": 16923 }, { "epoch": 18.97, "learning_rate": 7.583098591549296e-06, "loss": 0.3865, "step": 16924 }, { "epoch": 18.97, "learning_rate": 7.58169014084507e-06, "loss": 0.3652, "step": 16925 }, { "epoch": 18.98, "learning_rate": 7.580281690140844e-06, "loss": 0.3369, "step": 16926 }, { "epoch": 18.98, "learning_rate": 7.5788732394366205e-06, "loss": 0.3805, "step": 16927 }, { "epoch": 18.98, "learning_rate": 7.577464788732395e-06, "loss": 0.3547, "step": 16928 }, { "epoch": 18.98, "learning_rate": 7.57605633802817e-06, "loss": 0.4361, "step": 16929 }, { "epoch": 18.98, "learning_rate": 7.5746478873239436e-06, "loss": 0.3357, "step": 16930 }, { "epoch": 18.98, "learning_rate": 7.573239436619718e-06, "loss": 0.4549, "step": 16931 }, { "epoch": 18.98, "learning_rate": 7.571830985915493e-06, "loss": 0.3298, "step": 16932 }, { "epoch": 18.98, "learning_rate": 7.5704225352112675e-06, "loss": 0.3705, "step": 16933 }, { "epoch": 18.98, "learning_rate": 7.569014084507043e-06, "loss": 0.3158, "step": 16934 }, { "epoch": 18.99, "learning_rate": 7.567605633802818e-06, "loss": 0.3087, "step": 16935 }, { "epoch": 18.99, "learning_rate": 7.566197183098592e-06, "loss": 0.3817, "step": 16936 }, { "epoch": 18.99, "learning_rate": 7.564788732394367e-06, "loss": 0.3996, "step": 16937 }, { "epoch": 18.99, "learning_rate": 7.563380281690141e-06, "loss": 0.3539, "step": 16938 }, { "epoch": 18.99, "learning_rate": 7.561971830985915e-06, "loss": 0.3656, "step": 16939 }, { "epoch": 18.99, "learning_rate": 7.56056338028169e-06, "loss": 0.4473, "step": 16940 }, { "epoch": 18.99, "learning_rate": 7.5591549295774646e-06, "loss": 0.3974, "step": 16941 }, { "epoch": 18.99, "learning_rate": 7.55774647887324e-06, "loss": 0.4205, "step": 16942 }, { "epoch": 18.99, "learning_rate": 7.556338028169015e-06, "loss": 0.3937, "step": 16943 }, { "epoch": 19.0, "learning_rate": 7.554929577464789e-06, "loss": 0.4148, "step": 16944 }, { "epoch": 19.0, "learning_rate": 7.553521126760564e-06, "loss": 0.3846, "step": 16945 }, { "epoch": 19.0, "learning_rate": 7.552112676056338e-06, "loss": 0.4116, "step": 16946 }, { "epoch": 19.0, "learning_rate": 7.550704225352112e-06, "loss": 0.4655, "step": 16947 }, { "epoch": 19.0, "learning_rate": 7.549295774647887e-06, "loss": 0.4324, "step": 16948 }, { "epoch": 19.0, "learning_rate": 7.547887323943663e-06, "loss": 0.5701, "step": 16949 }, { "epoch": 19.0, "learning_rate": 7.546478873239437e-06, "loss": 0.4355, "step": 16950 }, { "epoch": 19.0, "learning_rate": 7.545070422535212e-06, "loss": 0.5055, "step": 16951 }, { "epoch": 19.0, "learning_rate": 7.5436619718309864e-06, "loss": 0.4249, "step": 16952 }, { "epoch": 19.01, "learning_rate": 7.542253521126761e-06, "loss": 0.442, "step": 16953 }, { "epoch": 19.01, "learning_rate": 7.540845070422535e-06, "loss": 0.3345, "step": 16954 }, { "epoch": 19.01, "learning_rate": 7.5394366197183095e-06, "loss": 0.3221, "step": 16955 }, { "epoch": 19.01, "learning_rate": 7.538028169014084e-06, "loss": 0.3984, "step": 16956 }, { "epoch": 19.01, "learning_rate": 7.5366197183098605e-06, "loss": 0.3946, "step": 16957 }, { "epoch": 19.01, "learning_rate": 7.535211267605634e-06, "loss": 0.4585, "step": 16958 }, { "epoch": 19.01, "learning_rate": 7.533802816901409e-06, "loss": 0.4418, "step": 16959 }, { "epoch": 19.01, "learning_rate": 7.5323943661971835e-06, "loss": 0.3877, "step": 16960 }, { "epoch": 19.01, "learning_rate": 7.530985915492958e-06, "loss": 0.4399, "step": 16961 }, { "epoch": 19.02, "learning_rate": 7.529577464788732e-06, "loss": 0.4846, "step": 16962 }, { "epoch": 19.02, "learning_rate": 7.528169014084507e-06, "loss": 0.3217, "step": 16963 }, { "epoch": 19.02, "learning_rate": 7.526760563380281e-06, "loss": 0.3487, "step": 16964 }, { "epoch": 19.02, "learning_rate": 7.5253521126760576e-06, "loss": 0.4144, "step": 16965 }, { "epoch": 19.02, "learning_rate": 7.523943661971831e-06, "loss": 0.2788, "step": 16966 }, { "epoch": 19.02, "learning_rate": 7.522535211267606e-06, "loss": 0.4086, "step": 16967 }, { "epoch": 19.02, "learning_rate": 7.521126760563381e-06, "loss": 0.4083, "step": 16968 }, { "epoch": 19.02, "learning_rate": 7.519718309859155e-06, "loss": 0.3822, "step": 16969 }, { "epoch": 19.02, "learning_rate": 7.518309859154929e-06, "loss": 0.3784, "step": 16970 }, { "epoch": 19.03, "learning_rate": 7.516901408450704e-06, "loss": 0.3002, "step": 16971 }, { "epoch": 19.03, "learning_rate": 7.51549295774648e-06, "loss": 0.3141, "step": 16972 }, { "epoch": 19.03, "learning_rate": 7.514084507042255e-06, "loss": 0.4184, "step": 16973 }, { "epoch": 19.03, "learning_rate": 7.5126760563380285e-06, "loss": 0.3626, "step": 16974 }, { "epoch": 19.03, "learning_rate": 7.511267605633803e-06, "loss": 0.3213, "step": 16975 }, { "epoch": 19.03, "learning_rate": 7.509859154929578e-06, "loss": 0.2766, "step": 16976 }, { "epoch": 19.03, "learning_rate": 7.508450704225352e-06, "loss": 0.4769, "step": 16977 }, { "epoch": 19.03, "learning_rate": 7.507042253521126e-06, "loss": 0.3334, "step": 16978 }, { "epoch": 19.03, "learning_rate": 7.505633802816901e-06, "loss": 0.485, "step": 16979 }, { "epoch": 19.04, "learning_rate": 7.504225352112677e-06, "loss": 0.4566, "step": 16980 }, { "epoch": 19.04, "learning_rate": 7.502816901408451e-06, "loss": 0.338, "step": 16981 }, { "epoch": 19.04, "learning_rate": 7.5014084507042256e-06, "loss": 0.2122, "step": 16982 }, { "epoch": 19.04, "learning_rate": 7.5e-06, "loss": 0.2809, "step": 16983 }, { "epoch": 19.04, "learning_rate": 7.498591549295775e-06, "loss": 0.3884, "step": 16984 }, { "epoch": 19.04, "learning_rate": 7.4971830985915495e-06, "loss": 0.316, "step": 16985 }, { "epoch": 19.04, "learning_rate": 7.495774647887324e-06, "loss": 0.3709, "step": 16986 }, { "epoch": 19.04, "learning_rate": 7.494366197183099e-06, "loss": 0.2767, "step": 16987 }, { "epoch": 19.04, "learning_rate": 7.492957746478873e-06, "loss": 0.413, "step": 16988 }, { "epoch": 19.05, "learning_rate": 7.491549295774648e-06, "loss": 0.4738, "step": 16989 }, { "epoch": 19.05, "learning_rate": 7.490140845070423e-06, "loss": 0.3271, "step": 16990 }, { "epoch": 19.05, "learning_rate": 7.488732394366197e-06, "loss": 0.4619, "step": 16991 }, { "epoch": 19.05, "learning_rate": 7.487323943661972e-06, "loss": 0.2998, "step": 16992 }, { "epoch": 19.05, "learning_rate": 7.4859154929577466e-06, "loss": 0.3188, "step": 16993 }, { "epoch": 19.05, "learning_rate": 7.484507042253521e-06, "loss": 0.3343, "step": 16994 }, { "epoch": 19.05, "learning_rate": 7.483098591549296e-06, "loss": 0.3502, "step": 16995 }, { "epoch": 19.05, "learning_rate": 7.4816901408450705e-06, "loss": 0.3698, "step": 16996 }, { "epoch": 19.05, "learning_rate": 7.480281690140845e-06, "loss": 0.3446, "step": 16997 }, { "epoch": 19.06, "learning_rate": 7.47887323943662e-06, "loss": 0.2404, "step": 16998 }, { "epoch": 19.06, "learning_rate": 7.477464788732394e-06, "loss": 0.4687, "step": 16999 }, { "epoch": 19.06, "learning_rate": 7.476056338028169e-06, "loss": 0.4891, "step": 17000 }, { "epoch": 19.06, "learning_rate": 7.474647887323944e-06, "loss": 0.3832, "step": 17001 }, { "epoch": 19.06, "learning_rate": 7.473239436619718e-06, "loss": 0.4106, "step": 17002 }, { "epoch": 19.06, "learning_rate": 7.471830985915493e-06, "loss": 0.3465, "step": 17003 }, { "epoch": 19.06, "learning_rate": 7.4704225352112684e-06, "loss": 0.399, "step": 17004 }, { "epoch": 19.06, "learning_rate": 7.469014084507042e-06, "loss": 0.3709, "step": 17005 }, { "epoch": 19.07, "learning_rate": 7.467605633802817e-06, "loss": 0.4733, "step": 17006 }, { "epoch": 19.07, "learning_rate": 7.4661971830985915e-06, "loss": 0.3601, "step": 17007 }, { "epoch": 19.07, "learning_rate": 7.464788732394367e-06, "loss": 0.4743, "step": 17008 }, { "epoch": 19.07, "learning_rate": 7.463380281690141e-06, "loss": 0.3587, "step": 17009 }, { "epoch": 19.07, "learning_rate": 7.461971830985915e-06, "loss": 0.3517, "step": 17010 }, { "epoch": 19.07, "learning_rate": 7.46056338028169e-06, "loss": 0.4295, "step": 17011 }, { "epoch": 19.07, "learning_rate": 7.4591549295774655e-06, "loss": 0.336, "step": 17012 }, { "epoch": 19.07, "learning_rate": 7.457746478873239e-06, "loss": 0.4054, "step": 17013 }, { "epoch": 19.07, "learning_rate": 7.456338028169014e-06, "loss": 0.4121, "step": 17014 }, { "epoch": 19.08, "learning_rate": 7.454929577464789e-06, "loss": 0.2864, "step": 17015 }, { "epoch": 19.08, "learning_rate": 7.453521126760564e-06, "loss": 0.328, "step": 17016 }, { "epoch": 19.08, "learning_rate": 7.452112676056338e-06, "loss": 0.3594, "step": 17017 }, { "epoch": 19.08, "learning_rate": 7.4507042253521125e-06, "loss": 0.3675, "step": 17018 }, { "epoch": 19.08, "learning_rate": 7.449295774647888e-06, "loss": 0.344, "step": 17019 }, { "epoch": 19.08, "learning_rate": 7.447887323943663e-06, "loss": 0.4123, "step": 17020 }, { "epoch": 19.08, "learning_rate": 7.446478873239436e-06, "loss": 0.4187, "step": 17021 }, { "epoch": 19.08, "learning_rate": 7.445070422535211e-06, "loss": 0.3219, "step": 17022 }, { "epoch": 19.08, "learning_rate": 7.4436619718309865e-06, "loss": 0.4865, "step": 17023 }, { "epoch": 19.09, "learning_rate": 7.442253521126761e-06, "loss": 0.3905, "step": 17024 }, { "epoch": 19.09, "learning_rate": 7.440845070422535e-06, "loss": 0.39, "step": 17025 }, { "epoch": 19.09, "learning_rate": 7.43943661971831e-06, "loss": 0.3332, "step": 17026 }, { "epoch": 19.09, "learning_rate": 7.438028169014085e-06, "loss": 0.3038, "step": 17027 }, { "epoch": 19.09, "learning_rate": 7.43661971830986e-06, "loss": 0.3368, "step": 17028 }, { "epoch": 19.09, "learning_rate": 7.4352112676056335e-06, "loss": 0.3527, "step": 17029 }, { "epoch": 19.09, "learning_rate": 7.433802816901408e-06, "loss": 0.3309, "step": 17030 }, { "epoch": 19.09, "learning_rate": 7.432394366197184e-06, "loss": 0.3531, "step": 17031 }, { "epoch": 19.09, "learning_rate": 7.430985915492958e-06, "loss": 0.3514, "step": 17032 }, { "epoch": 19.1, "learning_rate": 7.429577464788732e-06, "loss": 0.4183, "step": 17033 }, { "epoch": 19.1, "learning_rate": 7.428169014084507e-06, "loss": 0.3167, "step": 17034 }, { "epoch": 19.1, "learning_rate": 7.426760563380282e-06, "loss": 0.3637, "step": 17035 }, { "epoch": 19.1, "learning_rate": 7.425352112676057e-06, "loss": 0.3543, "step": 17036 }, { "epoch": 19.1, "learning_rate": 7.423943661971831e-06, "loss": 0.2949, "step": 17037 }, { "epoch": 19.1, "learning_rate": 7.422535211267606e-06, "loss": 0.3635, "step": 17038 }, { "epoch": 19.1, "learning_rate": 7.421126760563381e-06, "loss": 0.4213, "step": 17039 }, { "epoch": 19.1, "learning_rate": 7.419718309859155e-06, "loss": 0.3537, "step": 17040 }, { "epoch": 19.1, "learning_rate": 7.418309859154929e-06, "loss": 0.3363, "step": 17041 }, { "epoch": 19.11, "learning_rate": 7.416901408450705e-06, "loss": 0.3448, "step": 17042 }, { "epoch": 19.11, "learning_rate": 7.415492957746479e-06, "loss": 0.3142, "step": 17043 }, { "epoch": 19.11, "learning_rate": 7.414084507042254e-06, "loss": 0.2866, "step": 17044 }, { "epoch": 19.11, "learning_rate": 7.412676056338028e-06, "loss": 0.2818, "step": 17045 }, { "epoch": 19.11, "learning_rate": 7.411267605633803e-06, "loss": 0.4113, "step": 17046 }, { "epoch": 19.11, "learning_rate": 7.409859154929578e-06, "loss": 0.3723, "step": 17047 }, { "epoch": 19.11, "learning_rate": 7.4084507042253525e-06, "loss": 0.2324, "step": 17048 }, { "epoch": 19.11, "learning_rate": 7.407042253521126e-06, "loss": 0.5312, "step": 17049 }, { "epoch": 19.11, "learning_rate": 7.405633802816902e-06, "loss": 0.4062, "step": 17050 }, { "epoch": 19.12, "learning_rate": 7.404225352112676e-06, "loss": 0.4172, "step": 17051 }, { "epoch": 19.12, "learning_rate": 7.402816901408451e-06, "loss": 0.3984, "step": 17052 }, { "epoch": 19.12, "learning_rate": 7.401408450704226e-06, "loss": 0.464, "step": 17053 }, { "epoch": 19.12, "learning_rate": 7.4e-06, "loss": 0.3999, "step": 17054 }, { "epoch": 19.12, "learning_rate": 7.398591549295775e-06, "loss": 0.4732, "step": 17055 }, { "epoch": 19.12, "learning_rate": 7.3971830985915496e-06, "loss": 0.4051, "step": 17056 }, { "epoch": 19.12, "learning_rate": 7.395774647887324e-06, "loss": 0.3907, "step": 17057 }, { "epoch": 19.12, "learning_rate": 7.394366197183099e-06, "loss": 0.4666, "step": 17058 }, { "epoch": 19.12, "learning_rate": 7.3929577464788735e-06, "loss": 0.3325, "step": 17059 }, { "epoch": 19.13, "learning_rate": 7.391549295774648e-06, "loss": 0.4815, "step": 17060 }, { "epoch": 19.13, "learning_rate": 7.390140845070423e-06, "loss": 0.4452, "step": 17061 }, { "epoch": 19.13, "learning_rate": 7.388732394366197e-06, "loss": 0.389, "step": 17062 }, { "epoch": 19.13, "learning_rate": 7.387323943661972e-06, "loss": 0.4246, "step": 17063 }, { "epoch": 19.13, "learning_rate": 7.385915492957747e-06, "loss": 0.3921, "step": 17064 }, { "epoch": 19.13, "learning_rate": 7.384507042253521e-06, "loss": 0.3025, "step": 17065 }, { "epoch": 19.13, "learning_rate": 7.383098591549296e-06, "loss": 0.3714, "step": 17066 }, { "epoch": 19.13, "learning_rate": 7.381690140845071e-06, "loss": 0.2964, "step": 17067 }, { "epoch": 19.13, "learning_rate": 7.380281690140845e-06, "loss": 0.3309, "step": 17068 }, { "epoch": 19.14, "learning_rate": 7.37887323943662e-06, "loss": 0.3772, "step": 17069 }, { "epoch": 19.14, "learning_rate": 7.3774647887323945e-06, "loss": 0.3453, "step": 17070 }, { "epoch": 19.14, "learning_rate": 7.376056338028169e-06, "loss": 0.3306, "step": 17071 }, { "epoch": 19.14, "learning_rate": 7.374647887323945e-06, "loss": 0.3414, "step": 17072 }, { "epoch": 19.14, "learning_rate": 7.373239436619718e-06, "loss": 0.4088, "step": 17073 }, { "epoch": 19.14, "learning_rate": 7.371830985915493e-06, "loss": 0.2856, "step": 17074 }, { "epoch": 19.14, "learning_rate": 7.370422535211268e-06, "loss": 0.3453, "step": 17075 }, { "epoch": 19.14, "learning_rate": 7.369014084507043e-06, "loss": 0.3767, "step": 17076 }, { "epoch": 19.14, "learning_rate": 7.367605633802817e-06, "loss": 0.3345, "step": 17077 }, { "epoch": 19.15, "learning_rate": 7.366197183098592e-06, "loss": 0.3034, "step": 17078 }, { "epoch": 19.15, "learning_rate": 7.364788732394366e-06, "loss": 0.3384, "step": 17079 }, { "epoch": 19.15, "learning_rate": 7.363380281690142e-06, "loss": 0.386, "step": 17080 }, { "epoch": 19.15, "learning_rate": 7.3619718309859155e-06, "loss": 0.2799, "step": 17081 }, { "epoch": 19.15, "learning_rate": 7.36056338028169e-06, "loss": 0.3322, "step": 17082 }, { "epoch": 19.15, "learning_rate": 7.359154929577465e-06, "loss": 0.4616, "step": 17083 }, { "epoch": 19.15, "learning_rate": 7.35774647887324e-06, "loss": 0.4878, "step": 17084 }, { "epoch": 19.15, "learning_rate": 7.356338028169014e-06, "loss": 0.3626, "step": 17085 }, { "epoch": 19.15, "learning_rate": 7.354929577464789e-06, "loss": 0.3046, "step": 17086 }, { "epoch": 19.16, "learning_rate": 7.353521126760563e-06, "loss": 0.3567, "step": 17087 }, { "epoch": 19.16, "learning_rate": 7.352112676056339e-06, "loss": 0.3879, "step": 17088 }, { "epoch": 19.16, "learning_rate": 7.350704225352113e-06, "loss": 0.3128, "step": 17089 }, { "epoch": 19.16, "learning_rate": 7.349295774647887e-06, "loss": 0.3129, "step": 17090 }, { "epoch": 19.16, "learning_rate": 7.347887323943663e-06, "loss": 0.3209, "step": 17091 }, { "epoch": 19.16, "learning_rate": 7.346478873239437e-06, "loss": 0.3249, "step": 17092 }, { "epoch": 19.16, "learning_rate": 7.345070422535211e-06, "loss": 0.4495, "step": 17093 }, { "epoch": 19.16, "learning_rate": 7.343661971830986e-06, "loss": 0.2942, "step": 17094 }, { "epoch": 19.16, "learning_rate": 7.342253521126761e-06, "loss": 0.2246, "step": 17095 }, { "epoch": 19.17, "learning_rate": 7.340845070422535e-06, "loss": 0.4546, "step": 17096 }, { "epoch": 19.17, "learning_rate": 7.33943661971831e-06, "loss": 0.3606, "step": 17097 }, { "epoch": 19.17, "learning_rate": 7.338028169014084e-06, "loss": 0.3786, "step": 17098 }, { "epoch": 19.17, "learning_rate": 7.33661971830986e-06, "loss": 0.529, "step": 17099 }, { "epoch": 19.17, "learning_rate": 7.335211267605634e-06, "loss": 0.4278, "step": 17100 }, { "epoch": 19.17, "learning_rate": 7.333802816901408e-06, "loss": 0.4298, "step": 17101 }, { "epoch": 19.17, "learning_rate": 7.332394366197183e-06, "loss": 0.4014, "step": 17102 }, { "epoch": 19.17, "learning_rate": 7.330985915492958e-06, "loss": 0.4578, "step": 17103 }, { "epoch": 19.17, "learning_rate": 7.329577464788732e-06, "loss": 0.3899, "step": 17104 }, { "epoch": 19.18, "learning_rate": 7.328169014084507e-06, "loss": 0.4202, "step": 17105 }, { "epoch": 19.18, "learning_rate": 7.326760563380282e-06, "loss": 0.4079, "step": 17106 }, { "epoch": 19.18, "learning_rate": 7.325352112676057e-06, "loss": 0.4127, "step": 17107 }, { "epoch": 19.18, "learning_rate": 7.323943661971831e-06, "loss": 0.4184, "step": 17108 }, { "epoch": 19.18, "learning_rate": 7.322535211267605e-06, "loss": 0.3161, "step": 17109 }, { "epoch": 19.18, "learning_rate": 7.321126760563381e-06, "loss": 0.3777, "step": 17110 }, { "epoch": 19.18, "learning_rate": 7.3197183098591555e-06, "loss": 0.3511, "step": 17111 }, { "epoch": 19.18, "learning_rate": 7.318309859154929e-06, "loss": 0.3663, "step": 17112 }, { "epoch": 19.18, "learning_rate": 7.316901408450704e-06, "loss": 0.4651, "step": 17113 }, { "epoch": 19.19, "learning_rate": 7.315492957746479e-06, "loss": 0.4243, "step": 17114 }, { "epoch": 19.19, "learning_rate": 7.314084507042254e-06, "loss": 0.3187, "step": 17115 }, { "epoch": 19.19, "learning_rate": 7.312676056338028e-06, "loss": 0.373, "step": 17116 }, { "epoch": 19.19, "learning_rate": 7.3112676056338025e-06, "loss": 0.326, "step": 17117 }, { "epoch": 19.19, "learning_rate": 7.309859154929578e-06, "loss": 0.3851, "step": 17118 }, { "epoch": 19.19, "learning_rate": 7.308450704225353e-06, "loss": 0.3778, "step": 17119 }, { "epoch": 19.19, "learning_rate": 7.307042253521126e-06, "loss": 0.4223, "step": 17120 }, { "epoch": 19.19, "learning_rate": 7.305633802816901e-06, "loss": 0.4131, "step": 17121 }, { "epoch": 19.2, "learning_rate": 7.3042253521126765e-06, "loss": 0.3651, "step": 17122 }, { "epoch": 19.2, "learning_rate": 7.302816901408451e-06, "loss": 0.3088, "step": 17123 }, { "epoch": 19.2, "learning_rate": 7.301408450704225e-06, "loss": 0.337, "step": 17124 }, { "epoch": 19.2, "learning_rate": 7.3e-06, "loss": 0.3386, "step": 17125 }, { "epoch": 19.2, "learning_rate": 7.298591549295775e-06, "loss": 0.3191, "step": 17126 }, { "epoch": 19.2, "learning_rate": 7.29718309859155e-06, "loss": 0.3587, "step": 17127 }, { "epoch": 19.2, "learning_rate": 7.2957746478873235e-06, "loss": 0.3325, "step": 17128 }, { "epoch": 19.2, "learning_rate": 7.294366197183099e-06, "loss": 0.3056, "step": 17129 }, { "epoch": 19.2, "learning_rate": 7.292957746478874e-06, "loss": 0.3473, "step": 17130 }, { "epoch": 19.21, "learning_rate": 7.291549295774648e-06, "loss": 0.3801, "step": 17131 }, { "epoch": 19.21, "learning_rate": 7.290140845070422e-06, "loss": 0.4114, "step": 17132 }, { "epoch": 19.21, "learning_rate": 7.2887323943661975e-06, "loss": 0.3685, "step": 17133 }, { "epoch": 19.21, "learning_rate": 7.287323943661972e-06, "loss": 0.3205, "step": 17134 }, { "epoch": 19.21, "learning_rate": 7.285915492957747e-06, "loss": 0.429, "step": 17135 }, { "epoch": 19.21, "learning_rate": 7.2845070422535206e-06, "loss": 0.3442, "step": 17136 }, { "epoch": 19.21, "learning_rate": 7.283098591549296e-06, "loss": 0.4569, "step": 17137 }, { "epoch": 19.21, "learning_rate": 7.281690140845071e-06, "loss": 0.3263, "step": 17138 }, { "epoch": 19.21, "learning_rate": 7.280281690140845e-06, "loss": 0.4273, "step": 17139 }, { "epoch": 19.22, "learning_rate": 7.27887323943662e-06, "loss": 0.3594, "step": 17140 }, { "epoch": 19.22, "learning_rate": 7.277464788732395e-06, "loss": 0.3072, "step": 17141 }, { "epoch": 19.22, "learning_rate": 7.276056338028169e-06, "loss": 0.3726, "step": 17142 }, { "epoch": 19.22, "learning_rate": 7.274647887323944e-06, "loss": 0.3063, "step": 17143 }, { "epoch": 19.22, "learning_rate": 7.2732394366197185e-06, "loss": 0.3674, "step": 17144 }, { "epoch": 19.22, "learning_rate": 7.271830985915493e-06, "loss": 0.3252, "step": 17145 }, { "epoch": 19.22, "learning_rate": 7.270422535211268e-06, "loss": 0.2604, "step": 17146 }, { "epoch": 19.22, "learning_rate": 7.2690140845070424e-06, "loss": 0.354, "step": 17147 }, { "epoch": 19.22, "learning_rate": 7.267605633802817e-06, "loss": 0.4536, "step": 17148 }, { "epoch": 19.23, "learning_rate": 7.266197183098592e-06, "loss": 0.5069, "step": 17149 }, { "epoch": 19.23, "learning_rate": 7.264788732394366e-06, "loss": 0.4135, "step": 17150 }, { "epoch": 19.23, "learning_rate": 7.263380281690141e-06, "loss": 0.5005, "step": 17151 }, { "epoch": 19.23, "learning_rate": 7.261971830985916e-06, "loss": 0.4911, "step": 17152 }, { "epoch": 19.23, "learning_rate": 7.26056338028169e-06, "loss": 0.3648, "step": 17153 }, { "epoch": 19.23, "learning_rate": 7.259154929577465e-06, "loss": 0.4363, "step": 17154 }, { "epoch": 19.23, "learning_rate": 7.2577464788732395e-06, "loss": 0.3927, "step": 17155 }, { "epoch": 19.23, "learning_rate": 7.256338028169014e-06, "loss": 0.3972, "step": 17156 }, { "epoch": 19.23, "learning_rate": 7.254929577464789e-06, "loss": 0.3469, "step": 17157 }, { "epoch": 19.24, "learning_rate": 7.2535211267605634e-06, "loss": 0.3444, "step": 17158 }, { "epoch": 19.24, "learning_rate": 7.252112676056339e-06, "loss": 0.342, "step": 17159 }, { "epoch": 19.24, "learning_rate": 7.250704225352113e-06, "loss": 0.4303, "step": 17160 }, { "epoch": 19.24, "learning_rate": 7.249295774647887e-06, "loss": 0.411, "step": 17161 }, { "epoch": 19.24, "learning_rate": 7.247887323943662e-06, "loss": 0.3396, "step": 17162 }, { "epoch": 19.24, "learning_rate": 7.2464788732394375e-06, "loss": 0.3419, "step": 17163 }, { "epoch": 19.24, "learning_rate": 7.245070422535211e-06, "loss": 0.2881, "step": 17164 }, { "epoch": 19.24, "learning_rate": 7.243661971830986e-06, "loss": 0.3342, "step": 17165 }, { "epoch": 19.24, "learning_rate": 7.2422535211267605e-06, "loss": 0.4143, "step": 17166 }, { "epoch": 19.25, "learning_rate": 7.240845070422536e-06, "loss": 0.4356, "step": 17167 }, { "epoch": 19.25, "learning_rate": 7.23943661971831e-06, "loss": 0.3845, "step": 17168 }, { "epoch": 19.25, "learning_rate": 7.2380281690140845e-06, "loss": 0.3476, "step": 17169 }, { "epoch": 19.25, "learning_rate": 7.236619718309859e-06, "loss": 0.4145, "step": 17170 }, { "epoch": 19.25, "learning_rate": 7.235211267605635e-06, "loss": 0.4109, "step": 17171 }, { "epoch": 19.25, "learning_rate": 7.233802816901408e-06, "loss": 0.3805, "step": 17172 }, { "epoch": 19.25, "learning_rate": 7.232394366197183e-06, "loss": 0.3337, "step": 17173 }, { "epoch": 19.25, "learning_rate": 7.2309859154929585e-06, "loss": 0.2593, "step": 17174 }, { "epoch": 19.25, "learning_rate": 7.229577464788733e-06, "loss": 0.3514, "step": 17175 }, { "epoch": 19.26, "learning_rate": 7.228169014084507e-06, "loss": 0.3668, "step": 17176 }, { "epoch": 19.26, "learning_rate": 7.2267605633802816e-06, "loss": 0.3489, "step": 17177 }, { "epoch": 19.26, "learning_rate": 7.225352112676057e-06, "loss": 0.3572, "step": 17178 }, { "epoch": 19.26, "learning_rate": 7.223943661971832e-06, "loss": 0.3972, "step": 17179 }, { "epoch": 19.26, "learning_rate": 7.2225352112676055e-06, "loss": 0.3625, "step": 17180 }, { "epoch": 19.26, "learning_rate": 7.22112676056338e-06, "loss": 0.3745, "step": 17181 }, { "epoch": 19.26, "learning_rate": 7.219718309859156e-06, "loss": 0.3371, "step": 17182 }, { "epoch": 19.26, "learning_rate": 7.21830985915493e-06, "loss": 0.2874, "step": 17183 }, { "epoch": 19.26, "learning_rate": 7.216901408450704e-06, "loss": 0.3912, "step": 17184 }, { "epoch": 19.27, "learning_rate": 7.215492957746479e-06, "loss": 0.3424, "step": 17185 }, { "epoch": 19.27, "learning_rate": 7.214084507042254e-06, "loss": 0.3258, "step": 17186 }, { "epoch": 19.27, "learning_rate": 7.212676056338029e-06, "loss": 0.3003, "step": 17187 }, { "epoch": 19.27, "learning_rate": 7.2112676056338026e-06, "loss": 0.3632, "step": 17188 }, { "epoch": 19.27, "learning_rate": 7.209859154929577e-06, "loss": 0.3871, "step": 17189 }, { "epoch": 19.27, "learning_rate": 7.208450704225353e-06, "loss": 0.316, "step": 17190 }, { "epoch": 19.27, "learning_rate": 7.207042253521127e-06, "loss": 0.3995, "step": 17191 }, { "epoch": 19.27, "learning_rate": 7.205633802816901e-06, "loss": 0.294, "step": 17192 }, { "epoch": 19.27, "learning_rate": 7.204225352112677e-06, "loss": 0.2881, "step": 17193 }, { "epoch": 19.28, "learning_rate": 7.202816901408451e-06, "loss": 0.395, "step": 17194 }, { "epoch": 19.28, "learning_rate": 7.201408450704226e-06, "loss": 0.3033, "step": 17195 }, { "epoch": 19.28, "learning_rate": 7.2e-06, "loss": 0.2895, "step": 17196 }, { "epoch": 19.28, "learning_rate": 7.198591549295775e-06, "loss": 0.2852, "step": 17197 }, { "epoch": 19.28, "learning_rate": 7.19718309859155e-06, "loss": 0.3682, "step": 17198 }, { "epoch": 19.28, "learning_rate": 7.1957746478873244e-06, "loss": 0.451, "step": 17199 }, { "epoch": 19.28, "learning_rate": 7.194366197183098e-06, "loss": 0.5208, "step": 17200 }, { "epoch": 19.28, "learning_rate": 7.192957746478874e-06, "loss": 0.3531, "step": 17201 }, { "epoch": 19.28, "learning_rate": 7.191549295774648e-06, "loss": 0.4388, "step": 17202 }, { "epoch": 19.29, "learning_rate": 7.190140845070423e-06, "loss": 0.4452, "step": 17203 }, { "epoch": 19.29, "learning_rate": 7.188732394366197e-06, "loss": 0.4182, "step": 17204 }, { "epoch": 19.29, "learning_rate": 7.187323943661972e-06, "loss": 0.4273, "step": 17205 }, { "epoch": 19.29, "learning_rate": 7.185915492957747e-06, "loss": 0.4725, "step": 17206 }, { "epoch": 19.29, "learning_rate": 7.1845070422535215e-06, "loss": 0.4498, "step": 17207 }, { "epoch": 19.29, "learning_rate": 7.183098591549295e-06, "loss": 0.3267, "step": 17208 }, { "epoch": 19.29, "learning_rate": 7.181690140845071e-06, "loss": 0.3499, "step": 17209 }, { "epoch": 19.29, "learning_rate": 7.1802816901408454e-06, "loss": 0.4238, "step": 17210 }, { "epoch": 19.29, "learning_rate": 7.17887323943662e-06, "loss": 0.4842, "step": 17211 }, { "epoch": 19.3, "learning_rate": 7.177464788732395e-06, "loss": 0.3644, "step": 17212 }, { "epoch": 19.3, "learning_rate": 7.176056338028169e-06, "loss": 0.3953, "step": 17213 }, { "epoch": 19.3, "learning_rate": 7.174647887323944e-06, "loss": 0.3389, "step": 17214 }, { "epoch": 19.3, "learning_rate": 7.173239436619718e-06, "loss": 0.3897, "step": 17215 }, { "epoch": 19.3, "learning_rate": 7.171830985915493e-06, "loss": 0.2893, "step": 17216 }, { "epoch": 19.3, "learning_rate": 7.170422535211268e-06, "loss": 0.3324, "step": 17217 }, { "epoch": 19.3, "learning_rate": 7.1690140845070425e-06, "loss": 0.3809, "step": 17218 }, { "epoch": 19.3, "learning_rate": 7.167605633802816e-06, "loss": 0.3917, "step": 17219 }, { "epoch": 19.3, "learning_rate": 7.166197183098592e-06, "loss": 0.3052, "step": 17220 }, { "epoch": 19.31, "learning_rate": 7.1647887323943665e-06, "loss": 0.3874, "step": 17221 }, { "epoch": 19.31, "learning_rate": 7.163380281690141e-06, "loss": 0.4231, "step": 17222 }, { "epoch": 19.31, "learning_rate": 7.161971830985915e-06, "loss": 0.4693, "step": 17223 }, { "epoch": 19.31, "learning_rate": 7.16056338028169e-06, "loss": 0.3679, "step": 17224 }, { "epoch": 19.31, "learning_rate": 7.159154929577465e-06, "loss": 0.2661, "step": 17225 }, { "epoch": 19.31, "learning_rate": 7.15774647887324e-06, "loss": 0.3656, "step": 17226 }, { "epoch": 19.31, "learning_rate": 7.156338028169014e-06, "loss": 0.361, "step": 17227 }, { "epoch": 19.31, "learning_rate": 7.154929577464789e-06, "loss": 0.3713, "step": 17228 }, { "epoch": 19.32, "learning_rate": 7.1535211267605636e-06, "loss": 0.3817, "step": 17229 }, { "epoch": 19.32, "learning_rate": 7.152112676056338e-06, "loss": 0.3487, "step": 17230 }, { "epoch": 19.32, "learning_rate": 7.150704225352113e-06, "loss": 0.3334, "step": 17231 }, { "epoch": 19.32, "learning_rate": 7.1492957746478875e-06, "loss": 0.4134, "step": 17232 }, { "epoch": 19.32, "learning_rate": 7.147887323943662e-06, "loss": 0.45, "step": 17233 }, { "epoch": 19.32, "learning_rate": 7.146478873239437e-06, "loss": 0.3163, "step": 17234 }, { "epoch": 19.32, "learning_rate": 7.145070422535211e-06, "loss": 0.3542, "step": 17235 }, { "epoch": 19.32, "learning_rate": 7.143661971830986e-06, "loss": 0.2457, "step": 17236 }, { "epoch": 19.32, "learning_rate": 7.142253521126761e-06, "loss": 0.3231, "step": 17237 }, { "epoch": 19.33, "learning_rate": 7.140845070422535e-06, "loss": 0.299, "step": 17238 }, { "epoch": 19.33, "learning_rate": 7.13943661971831e-06, "loss": 0.3515, "step": 17239 }, { "epoch": 19.33, "learning_rate": 7.1380281690140846e-06, "loss": 0.4224, "step": 17240 }, { "epoch": 19.33, "learning_rate": 7.136619718309859e-06, "loss": 0.3911, "step": 17241 }, { "epoch": 19.33, "learning_rate": 7.135211267605634e-06, "loss": 0.2863, "step": 17242 }, { "epoch": 19.33, "learning_rate": 7.1338028169014085e-06, "loss": 0.22, "step": 17243 }, { "epoch": 19.33, "learning_rate": 7.132394366197183e-06, "loss": 0.3022, "step": 17244 }, { "epoch": 19.33, "learning_rate": 7.130985915492958e-06, "loss": 0.2243, "step": 17245 }, { "epoch": 19.33, "learning_rate": 7.129577464788733e-06, "loss": 0.2329, "step": 17246 }, { "epoch": 19.34, "learning_rate": 7.128169014084507e-06, "loss": 0.4026, "step": 17247 }, { "epoch": 19.34, "learning_rate": 7.126760563380282e-06, "loss": 0.4885, "step": 17248 }, { "epoch": 19.34, "learning_rate": 7.125352112676056e-06, "loss": 0.4668, "step": 17249 }, { "epoch": 19.34, "learning_rate": 7.123943661971832e-06, "loss": 0.3475, "step": 17250 }, { "epoch": 19.34, "learning_rate": 7.122535211267606e-06, "loss": 0.5518, "step": 17251 }, { "epoch": 19.34, "learning_rate": 7.12112676056338e-06, "loss": 0.4137, "step": 17252 }, { "epoch": 19.34, "learning_rate": 7.119718309859155e-06, "loss": 0.4527, "step": 17253 }, { "epoch": 19.34, "learning_rate": 7.11830985915493e-06, "loss": 0.4724, "step": 17254 }, { "epoch": 19.34, "learning_rate": 7.116901408450704e-06, "loss": 0.4135, "step": 17255 }, { "epoch": 19.35, "learning_rate": 7.115492957746479e-06, "loss": 0.3594, "step": 17256 }, { "epoch": 19.35, "learning_rate": 7.114084507042253e-06, "loss": 0.4895, "step": 17257 }, { "epoch": 19.35, "learning_rate": 7.112676056338029e-06, "loss": 0.4468, "step": 17258 }, { "epoch": 19.35, "learning_rate": 7.111267605633803e-06, "loss": 0.3321, "step": 17259 }, { "epoch": 19.35, "learning_rate": 7.109859154929577e-06, "loss": 0.4728, "step": 17260 }, { "epoch": 19.35, "learning_rate": 7.108450704225353e-06, "loss": 0.5025, "step": 17261 }, { "epoch": 19.35, "learning_rate": 7.1070422535211274e-06, "loss": 0.3977, "step": 17262 }, { "epoch": 19.35, "learning_rate": 7.105633802816901e-06, "loss": 0.455, "step": 17263 }, { "epoch": 19.35, "learning_rate": 7.104225352112676e-06, "loss": 0.454, "step": 17264 }, { "epoch": 19.36, "learning_rate": 7.102816901408451e-06, "loss": 0.4557, "step": 17265 }, { "epoch": 19.36, "learning_rate": 7.101408450704226e-06, "loss": 0.4131, "step": 17266 }, { "epoch": 19.36, "learning_rate": 7.1e-06, "loss": 0.3362, "step": 17267 }, { "epoch": 19.36, "learning_rate": 7.098591549295774e-06, "loss": 0.4216, "step": 17268 }, { "epoch": 19.36, "learning_rate": 7.09718309859155e-06, "loss": 0.3527, "step": 17269 }, { "epoch": 19.36, "learning_rate": 7.0957746478873245e-06, "loss": 0.3391, "step": 17270 }, { "epoch": 19.36, "learning_rate": 7.094366197183098e-06, "loss": 0.3421, "step": 17271 }, { "epoch": 19.36, "learning_rate": 7.092957746478873e-06, "loss": 0.3072, "step": 17272 }, { "epoch": 19.36, "learning_rate": 7.0915492957746485e-06, "loss": 0.3642, "step": 17273 }, { "epoch": 19.37, "learning_rate": 7.090140845070423e-06, "loss": 0.3929, "step": 17274 }, { "epoch": 19.37, "learning_rate": 7.088732394366197e-06, "loss": 0.3635, "step": 17275 }, { "epoch": 19.37, "learning_rate": 7.0873239436619715e-06, "loss": 0.3195, "step": 17276 }, { "epoch": 19.37, "learning_rate": 7.085915492957747e-06, "loss": 0.3001, "step": 17277 }, { "epoch": 19.37, "learning_rate": 7.084507042253522e-06, "loss": 0.3972, "step": 17278 }, { "epoch": 19.37, "learning_rate": 7.0830985915492954e-06, "loss": 0.3854, "step": 17279 }, { "epoch": 19.37, "learning_rate": 7.081690140845071e-06, "loss": 0.4031, "step": 17280 }, { "epoch": 19.37, "learning_rate": 7.0802816901408456e-06, "loss": 0.3384, "step": 17281 }, { "epoch": 19.37, "learning_rate": 7.07887323943662e-06, "loss": 0.3024, "step": 17282 }, { "epoch": 19.38, "learning_rate": 7.077464788732394e-06, "loss": 0.3182, "step": 17283 }, { "epoch": 19.38, "learning_rate": 7.0760563380281695e-06, "loss": 0.3264, "step": 17284 }, { "epoch": 19.38, "learning_rate": 7.074647887323944e-06, "loss": 0.2978, "step": 17285 }, { "epoch": 19.38, "learning_rate": 7.073239436619719e-06, "loss": 0.4165, "step": 17286 }, { "epoch": 19.38, "learning_rate": 7.0718309859154925e-06, "loss": 0.3574, "step": 17287 }, { "epoch": 19.38, "learning_rate": 7.070422535211268e-06, "loss": 0.3262, "step": 17288 }, { "epoch": 19.38, "learning_rate": 7.069014084507043e-06, "loss": 0.2379, "step": 17289 }, { "epoch": 19.38, "learning_rate": 7.067605633802817e-06, "loss": 0.2829, "step": 17290 }, { "epoch": 19.38, "learning_rate": 7.066197183098591e-06, "loss": 0.3179, "step": 17291 }, { "epoch": 19.39, "learning_rate": 7.0647887323943666e-06, "loss": 0.3169, "step": 17292 }, { "epoch": 19.39, "learning_rate": 7.063380281690141e-06, "loss": 0.2086, "step": 17293 }, { "epoch": 19.39, "learning_rate": 7.061971830985916e-06, "loss": 0.4738, "step": 17294 }, { "epoch": 19.39, "learning_rate": 7.06056338028169e-06, "loss": 0.2833, "step": 17295 }, { "epoch": 19.39, "learning_rate": 7.059154929577465e-06, "loss": 0.3533, "step": 17296 }, { "epoch": 19.39, "learning_rate": 7.05774647887324e-06, "loss": 0.2308, "step": 17297 }, { "epoch": 19.39, "learning_rate": 7.056338028169014e-06, "loss": 0.2727, "step": 17298 }, { "epoch": 19.39, "learning_rate": 7.054929577464789e-06, "loss": 0.5523, "step": 17299 }, { "epoch": 19.39, "learning_rate": 7.053521126760564e-06, "loss": 0.4367, "step": 17300 }, { "epoch": 19.4, "learning_rate": 7.052112676056338e-06, "loss": 0.3647, "step": 17301 }, { "epoch": 19.4, "learning_rate": 7.050704225352113e-06, "loss": 0.4067, "step": 17302 }, { "epoch": 19.4, "learning_rate": 7.049295774647888e-06, "loss": 0.4041, "step": 17303 }, { "epoch": 19.4, "learning_rate": 7.047887323943662e-06, "loss": 0.4428, "step": 17304 }, { "epoch": 19.4, "learning_rate": 7.046478873239437e-06, "loss": 0.4239, "step": 17305 }, { "epoch": 19.4, "learning_rate": 7.0450704225352115e-06, "loss": 0.383, "step": 17306 }, { "epoch": 19.4, "learning_rate": 7.043661971830986e-06, "loss": 0.3644, "step": 17307 }, { "epoch": 19.4, "learning_rate": 7.042253521126761e-06, "loss": 0.378, "step": 17308 }, { "epoch": 19.4, "learning_rate": 7.040845070422535e-06, "loss": 0.3365, "step": 17309 }, { "epoch": 19.41, "learning_rate": 7.03943661971831e-06, "loss": 0.3753, "step": 17310 }, { "epoch": 19.41, "learning_rate": 7.038028169014085e-06, "loss": 0.3906, "step": 17311 }, { "epoch": 19.41, "learning_rate": 7.036619718309859e-06, "loss": 0.3549, "step": 17312 }, { "epoch": 19.41, "learning_rate": 7.035211267605634e-06, "loss": 0.3331, "step": 17313 }, { "epoch": 19.41, "learning_rate": 7.0338028169014094e-06, "loss": 0.3724, "step": 17314 }, { "epoch": 19.41, "learning_rate": 7.032394366197183e-06, "loss": 0.3881, "step": 17315 }, { "epoch": 19.41, "learning_rate": 7.030985915492958e-06, "loss": 0.3706, "step": 17316 }, { "epoch": 19.41, "learning_rate": 7.0295774647887325e-06, "loss": 0.3257, "step": 17317 }, { "epoch": 19.41, "learning_rate": 7.028169014084508e-06, "loss": 0.3125, "step": 17318 }, { "epoch": 19.42, "learning_rate": 7.026760563380282e-06, "loss": 0.4475, "step": 17319 }, { "epoch": 19.42, "learning_rate": 7.025352112676056e-06, "loss": 0.363, "step": 17320 }, { "epoch": 19.42, "learning_rate": 7.023943661971831e-06, "loss": 0.3969, "step": 17321 }, { "epoch": 19.42, "learning_rate": 7.022535211267606e-06, "loss": 0.339, "step": 17322 }, { "epoch": 19.42, "learning_rate": 7.02112676056338e-06, "loss": 0.3896, "step": 17323 }, { "epoch": 19.42, "learning_rate": 7.019718309859155e-06, "loss": 0.3774, "step": 17324 }, { "epoch": 19.42, "learning_rate": 7.01830985915493e-06, "loss": 0.4121, "step": 17325 }, { "epoch": 19.42, "learning_rate": 7.016901408450704e-06, "loss": 0.3473, "step": 17326 }, { "epoch": 19.42, "learning_rate": 7.015492957746479e-06, "loss": 0.359, "step": 17327 }, { "epoch": 19.43, "learning_rate": 7.0140845070422535e-06, "loss": 0.4083, "step": 17328 }, { "epoch": 19.43, "learning_rate": 7.012676056338028e-06, "loss": 0.3988, "step": 17329 }, { "epoch": 19.43, "learning_rate": 7.011267605633803e-06, "loss": 0.3433, "step": 17330 }, { "epoch": 19.43, "learning_rate": 7.0098591549295774e-06, "loss": 0.2981, "step": 17331 }, { "epoch": 19.43, "learning_rate": 7.008450704225352e-06, "loss": 0.4276, "step": 17332 }, { "epoch": 19.43, "learning_rate": 7.0070422535211276e-06, "loss": 0.3873, "step": 17333 }, { "epoch": 19.43, "learning_rate": 7.005633802816901e-06, "loss": 0.3933, "step": 17334 }, { "epoch": 19.43, "learning_rate": 7.004225352112676e-06, "loss": 0.3043, "step": 17335 }, { "epoch": 19.43, "learning_rate": 7.002816901408451e-06, "loss": 0.3859, "step": 17336 }, { "epoch": 19.44, "learning_rate": 7.001408450704226e-06, "loss": 0.3537, "step": 17337 }, { "epoch": 19.44, "learning_rate": 7e-06, "loss": 0.2616, "step": 17338 }, { "epoch": 19.44, "learning_rate": 6.9985915492957745e-06, "loss": 0.355, "step": 17339 }, { "epoch": 19.44, "learning_rate": 6.997183098591549e-06, "loss": 0.3772, "step": 17340 }, { "epoch": 19.44, "learning_rate": 6.995774647887325e-06, "loss": 0.3134, "step": 17341 }, { "epoch": 19.44, "learning_rate": 6.9943661971830984e-06, "loss": 0.35, "step": 17342 }, { "epoch": 19.44, "learning_rate": 6.992957746478873e-06, "loss": 0.3519, "step": 17343 }, { "epoch": 19.44, "learning_rate": 6.991549295774648e-06, "loss": 0.1663, "step": 17344 }, { "epoch": 19.45, "learning_rate": 6.990140845070423e-06, "loss": 0.198, "step": 17345 }, { "epoch": 19.45, "learning_rate": 6.988732394366197e-06, "loss": 0.2769, "step": 17346 }, { "epoch": 19.45, "learning_rate": 6.987323943661972e-06, "loss": 0.3064, "step": 17347 }, { "epoch": 19.45, "learning_rate": 6.985915492957747e-06, "loss": 0.3855, "step": 17348 }, { "epoch": 19.45, "learning_rate": 6.984507042253522e-06, "loss": 0.5159, "step": 17349 }, { "epoch": 19.45, "learning_rate": 6.9830985915492955e-06, "loss": 0.5044, "step": 17350 }, { "epoch": 19.45, "learning_rate": 6.98169014084507e-06, "loss": 0.4285, "step": 17351 }, { "epoch": 19.45, "learning_rate": 6.980281690140846e-06, "loss": 0.4766, "step": 17352 }, { "epoch": 19.45, "learning_rate": 6.97887323943662e-06, "loss": 0.3447, "step": 17353 }, { "epoch": 19.46, "learning_rate": 6.977464788732394e-06, "loss": 0.3764, "step": 17354 }, { "epoch": 19.46, "learning_rate": 6.976056338028169e-06, "loss": 0.3956, "step": 17355 }, { "epoch": 19.46, "learning_rate": 6.974647887323944e-06, "loss": 0.4397, "step": 17356 }, { "epoch": 19.46, "learning_rate": 6.973239436619719e-06, "loss": 0.3624, "step": 17357 }, { "epoch": 19.46, "learning_rate": 6.971830985915493e-06, "loss": 0.393, "step": 17358 }, { "epoch": 19.46, "learning_rate": 6.970422535211267e-06, "loss": 0.4674, "step": 17359 }, { "epoch": 19.46, "learning_rate": 6.969014084507043e-06, "loss": 0.3453, "step": 17360 }, { "epoch": 19.46, "learning_rate": 6.967605633802817e-06, "loss": 0.4085, "step": 17361 }, { "epoch": 19.46, "learning_rate": 6.966197183098591e-06, "loss": 0.3458, "step": 17362 }, { "epoch": 19.47, "learning_rate": 6.964788732394366e-06, "loss": 0.3513, "step": 17363 }, { "epoch": 19.47, "learning_rate": 6.963380281690141e-06, "loss": 0.3662, "step": 17364 }, { "epoch": 19.47, "learning_rate": 6.961971830985916e-06, "loss": 0.3866, "step": 17365 }, { "epoch": 19.47, "learning_rate": 6.96056338028169e-06, "loss": 0.4206, "step": 17366 }, { "epoch": 19.47, "learning_rate": 6.959154929577465e-06, "loss": 0.3248, "step": 17367 }, { "epoch": 19.47, "learning_rate": 6.95774647887324e-06, "loss": 0.42, "step": 17368 }, { "epoch": 19.47, "learning_rate": 6.9563380281690145e-06, "loss": 0.352, "step": 17369 }, { "epoch": 19.47, "learning_rate": 6.954929577464788e-06, "loss": 0.37, "step": 17370 }, { "epoch": 19.47, "learning_rate": 6.953521126760564e-06, "loss": 0.3185, "step": 17371 }, { "epoch": 19.48, "learning_rate": 6.952112676056338e-06, "loss": 0.3238, "step": 17372 }, { "epoch": 19.48, "learning_rate": 6.950704225352113e-06, "loss": 0.3136, "step": 17373 }, { "epoch": 19.48, "learning_rate": 6.949295774647887e-06, "loss": 0.3511, "step": 17374 }, { "epoch": 19.48, "learning_rate": 6.947887323943662e-06, "loss": 0.296, "step": 17375 }, { "epoch": 19.48, "learning_rate": 6.946478873239437e-06, "loss": 0.2672, "step": 17376 }, { "epoch": 19.48, "learning_rate": 6.945070422535212e-06, "loss": 0.336, "step": 17377 }, { "epoch": 19.48, "learning_rate": 6.943661971830985e-06, "loss": 0.3426, "step": 17378 }, { "epoch": 19.48, "learning_rate": 6.942253521126761e-06, "loss": 0.3462, "step": 17379 }, { "epoch": 19.48, "learning_rate": 6.9408450704225355e-06, "loss": 0.3675, "step": 17380 }, { "epoch": 19.49, "learning_rate": 6.93943661971831e-06, "loss": 0.3145, "step": 17381 }, { "epoch": 19.49, "learning_rate": 6.938028169014085e-06, "loss": 0.3852, "step": 17382 }, { "epoch": 19.49, "learning_rate": 6.9366197183098594e-06, "loss": 0.3248, "step": 17383 }, { "epoch": 19.49, "learning_rate": 6.935211267605634e-06, "loss": 0.3666, "step": 17384 }, { "epoch": 19.49, "learning_rate": 6.933802816901409e-06, "loss": 0.3151, "step": 17385 }, { "epoch": 19.49, "learning_rate": 6.932394366197183e-06, "loss": 0.3481, "step": 17386 }, { "epoch": 19.49, "learning_rate": 6.930985915492958e-06, "loss": 0.329, "step": 17387 }, { "epoch": 19.49, "learning_rate": 6.929577464788733e-06, "loss": 0.3331, "step": 17388 }, { "epoch": 19.49, "learning_rate": 6.928169014084507e-06, "loss": 0.4556, "step": 17389 }, { "epoch": 19.5, "learning_rate": 6.926760563380282e-06, "loss": 0.3751, "step": 17390 }, { "epoch": 19.5, "learning_rate": 6.9253521126760565e-06, "loss": 0.3579, "step": 17391 }, { "epoch": 19.5, "learning_rate": 6.923943661971831e-06, "loss": 0.238, "step": 17392 }, { "epoch": 19.5, "learning_rate": 6.922535211267606e-06, "loss": 0.3144, "step": 17393 }, { "epoch": 19.5, "learning_rate": 6.9211267605633804e-06, "loss": 0.3019, "step": 17394 }, { "epoch": 19.5, "learning_rate": 6.919718309859155e-06, "loss": 0.3987, "step": 17395 }, { "epoch": 19.5, "learning_rate": 6.91830985915493e-06, "loss": 0.3705, "step": 17396 }, { "epoch": 19.5, "learning_rate": 6.916901408450704e-06, "loss": 0.3129, "step": 17397 }, { "epoch": 19.5, "learning_rate": 6.915492957746479e-06, "loss": 0.3391, "step": 17398 }, { "epoch": 19.51, "learning_rate": 6.914084507042254e-06, "loss": 0.38, "step": 17399 }, { "epoch": 19.51, "learning_rate": 6.912676056338028e-06, "loss": 0.5476, "step": 17400 }, { "epoch": 19.51, "learning_rate": 6.911267605633804e-06, "loss": 0.4442, "step": 17401 }, { "epoch": 19.51, "learning_rate": 6.9098591549295775e-06, "loss": 0.4546, "step": 17402 }, { "epoch": 19.51, "learning_rate": 6.908450704225352e-06, "loss": 0.3658, "step": 17403 }, { "epoch": 19.51, "learning_rate": 6.907042253521127e-06, "loss": 0.4476, "step": 17404 }, { "epoch": 19.51, "learning_rate": 6.905633802816902e-06, "loss": 0.487, "step": 17405 }, { "epoch": 19.51, "learning_rate": 6.904225352112676e-06, "loss": 0.3598, "step": 17406 }, { "epoch": 19.51, "learning_rate": 6.902816901408451e-06, "loss": 0.3058, "step": 17407 }, { "epoch": 19.52, "learning_rate": 6.901408450704225e-06, "loss": 0.3439, "step": 17408 }, { "epoch": 19.52, "learning_rate": 6.900000000000001e-06, "loss": 0.3932, "step": 17409 }, { "epoch": 19.52, "learning_rate": 6.898591549295775e-06, "loss": 0.3944, "step": 17410 }, { "epoch": 19.52, "learning_rate": 6.897183098591549e-06, "loss": 0.3806, "step": 17411 }, { "epoch": 19.52, "learning_rate": 6.895774647887324e-06, "loss": 0.3128, "step": 17412 }, { "epoch": 19.52, "learning_rate": 6.894366197183099e-06, "loss": 0.3912, "step": 17413 }, { "epoch": 19.52, "learning_rate": 6.892957746478873e-06, "loss": 0.382, "step": 17414 }, { "epoch": 19.52, "learning_rate": 6.891549295774648e-06, "loss": 0.4157, "step": 17415 }, { "epoch": 19.52, "learning_rate": 6.8901408450704225e-06, "loss": 0.3891, "step": 17416 }, { "epoch": 19.53, "learning_rate": 6.888732394366198e-06, "loss": 0.3312, "step": 17417 }, { "epoch": 19.53, "learning_rate": 6.887323943661972e-06, "loss": 0.4331, "step": 17418 }, { "epoch": 19.53, "learning_rate": 6.885915492957746e-06, "loss": 0.4092, "step": 17419 }, { "epoch": 19.53, "learning_rate": 6.884507042253522e-06, "loss": 0.3369, "step": 17420 }, { "epoch": 19.53, "learning_rate": 6.8830985915492965e-06, "loss": 0.3447, "step": 17421 }, { "epoch": 19.53, "learning_rate": 6.88169014084507e-06, "loss": 0.3179, "step": 17422 }, { "epoch": 19.53, "learning_rate": 6.880281690140845e-06, "loss": 0.2988, "step": 17423 }, { "epoch": 19.53, "learning_rate": 6.87887323943662e-06, "loss": 0.3497, "step": 17424 }, { "epoch": 19.53, "learning_rate": 6.877464788732395e-06, "loss": 0.344, "step": 17425 }, { "epoch": 19.54, "learning_rate": 6.876056338028169e-06, "loss": 0.3545, "step": 17426 }, { "epoch": 19.54, "learning_rate": 6.8746478873239435e-06, "loss": 0.5076, "step": 17427 }, { "epoch": 19.54, "learning_rate": 6.873239436619719e-06, "loss": 0.4333, "step": 17428 }, { "epoch": 19.54, "learning_rate": 6.871830985915494e-06, "loss": 0.373, "step": 17429 }, { "epoch": 19.54, "learning_rate": 6.870422535211267e-06, "loss": 0.3646, "step": 17430 }, { "epoch": 19.54, "learning_rate": 6.869014084507042e-06, "loss": 0.2869, "step": 17431 }, { "epoch": 19.54, "learning_rate": 6.8676056338028175e-06, "loss": 0.3959, "step": 17432 }, { "epoch": 19.54, "learning_rate": 6.866197183098592e-06, "loss": 0.3019, "step": 17433 }, { "epoch": 19.54, "learning_rate": 6.864788732394366e-06, "loss": 0.2969, "step": 17434 }, { "epoch": 19.55, "learning_rate": 6.8633802816901414e-06, "loss": 0.3245, "step": 17435 }, { "epoch": 19.55, "learning_rate": 6.861971830985916e-06, "loss": 0.3198, "step": 17436 }, { "epoch": 19.55, "learning_rate": 6.86056338028169e-06, "loss": 0.2271, "step": 17437 }, { "epoch": 19.55, "learning_rate": 6.8591549295774645e-06, "loss": 0.2911, "step": 17438 }, { "epoch": 19.55, "learning_rate": 6.85774647887324e-06, "loss": 0.4422, "step": 17439 }, { "epoch": 19.55, "learning_rate": 6.856338028169015e-06, "loss": 0.3825, "step": 17440 }, { "epoch": 19.55, "learning_rate": 6.854929577464788e-06, "loss": 0.3789, "step": 17441 }, { "epoch": 19.55, "learning_rate": 6.853521126760563e-06, "loss": 0.2484, "step": 17442 }, { "epoch": 19.55, "learning_rate": 6.8521126760563385e-06, "loss": 0.3244, "step": 17443 }, { "epoch": 19.56, "learning_rate": 6.850704225352113e-06, "loss": 0.2594, "step": 17444 }, { "epoch": 19.56, "learning_rate": 6.849295774647887e-06, "loss": 0.3182, "step": 17445 }, { "epoch": 19.56, "learning_rate": 6.847887323943662e-06, "loss": 0.2996, "step": 17446 }, { "epoch": 19.56, "learning_rate": 6.846478873239437e-06, "loss": 0.2651, "step": 17447 }, { "epoch": 19.56, "learning_rate": 6.845070422535212e-06, "loss": 0.3214, "step": 17448 }, { "epoch": 19.56, "learning_rate": 6.8436619718309855e-06, "loss": 0.4806, "step": 17449 }, { "epoch": 19.56, "learning_rate": 6.84225352112676e-06, "loss": 0.4572, "step": 17450 }, { "epoch": 19.56, "learning_rate": 6.840845070422536e-06, "loss": 0.4531, "step": 17451 }, { "epoch": 19.57, "learning_rate": 6.83943661971831e-06, "loss": 0.4255, "step": 17452 }, { "epoch": 19.57, "learning_rate": 6.838028169014084e-06, "loss": 0.4409, "step": 17453 }, { "epoch": 19.57, "learning_rate": 6.8366197183098595e-06, "loss": 0.4859, "step": 17454 }, { "epoch": 19.57, "learning_rate": 6.835211267605634e-06, "loss": 0.3251, "step": 17455 }, { "epoch": 19.57, "learning_rate": 6.833802816901409e-06, "loss": 0.4429, "step": 17456 }, { "epoch": 19.57, "learning_rate": 6.832394366197183e-06, "loss": 0.3227, "step": 17457 }, { "epoch": 19.57, "learning_rate": 6.830985915492958e-06, "loss": 0.357, "step": 17458 }, { "epoch": 19.57, "learning_rate": 6.829577464788733e-06, "loss": 0.4306, "step": 17459 }, { "epoch": 19.57, "learning_rate": 6.828169014084507e-06, "loss": 0.34, "step": 17460 }, { "epoch": 19.58, "learning_rate": 6.826760563380281e-06, "loss": 0.3398, "step": 17461 }, { "epoch": 19.58, "learning_rate": 6.825352112676057e-06, "loss": 0.3711, "step": 17462 }, { "epoch": 19.58, "learning_rate": 6.823943661971831e-06, "loss": 0.4041, "step": 17463 }, { "epoch": 19.58, "learning_rate": 6.822535211267606e-06, "loss": 0.3423, "step": 17464 }, { "epoch": 19.58, "learning_rate": 6.82112676056338e-06, "loss": 0.3198, "step": 17465 }, { "epoch": 19.58, "learning_rate": 6.819718309859155e-06, "loss": 0.3161, "step": 17466 }, { "epoch": 19.58, "learning_rate": 6.81830985915493e-06, "loss": 0.296, "step": 17467 }, { "epoch": 19.58, "learning_rate": 6.8169014084507045e-06, "loss": 0.4409, "step": 17468 }, { "epoch": 19.58, "learning_rate": 6.815492957746479e-06, "loss": 0.3305, "step": 17469 }, { "epoch": 19.59, "learning_rate": 6.814084507042254e-06, "loss": 0.3913, "step": 17470 }, { "epoch": 19.59, "learning_rate": 6.812676056338028e-06, "loss": 0.3369, "step": 17471 }, { "epoch": 19.59, "learning_rate": 6.811267605633803e-06, "loss": 0.4037, "step": 17472 }, { "epoch": 19.59, "learning_rate": 6.809859154929578e-06, "loss": 0.4434, "step": 17473 }, { "epoch": 19.59, "learning_rate": 6.808450704225352e-06, "loss": 0.4092, "step": 17474 }, { "epoch": 19.59, "learning_rate": 6.807042253521127e-06, "loss": 0.3441, "step": 17475 }, { "epoch": 19.59, "learning_rate": 6.8056338028169016e-06, "loss": 0.3503, "step": 17476 }, { "epoch": 19.59, "learning_rate": 6.804225352112676e-06, "loss": 0.4434, "step": 17477 }, { "epoch": 19.59, "learning_rate": 6.802816901408451e-06, "loss": 0.3397, "step": 17478 }, { "epoch": 19.6, "learning_rate": 6.8014084507042255e-06, "loss": 0.3115, "step": 17479 }, { "epoch": 19.6, "learning_rate": 6.8e-06, "loss": 0.33, "step": 17480 }, { "epoch": 19.6, "learning_rate": 6.798591549295775e-06, "loss": 0.4604, "step": 17481 }, { "epoch": 19.6, "learning_rate": 6.797183098591549e-06, "loss": 0.3435, "step": 17482 }, { "epoch": 19.6, "learning_rate": 6.795774647887324e-06, "loss": 0.4586, "step": 17483 }, { "epoch": 19.6, "learning_rate": 6.794366197183099e-06, "loss": 0.3792, "step": 17484 }, { "epoch": 19.6, "learning_rate": 6.792957746478873e-06, "loss": 0.3754, "step": 17485 }, { "epoch": 19.6, "learning_rate": 6.791549295774648e-06, "loss": 0.3193, "step": 17486 }, { "epoch": 19.6, "learning_rate": 6.790140845070423e-06, "loss": 0.3409, "step": 17487 }, { "epoch": 19.61, "learning_rate": 6.788732394366198e-06, "loss": 0.3515, "step": 17488 }, { "epoch": 19.61, "learning_rate": 6.787323943661972e-06, "loss": 0.1938, "step": 17489 }, { "epoch": 19.61, "learning_rate": 6.7859154929577465e-06, "loss": 0.3723, "step": 17490 }, { "epoch": 19.61, "learning_rate": 6.784507042253521e-06, "loss": 0.4077, "step": 17491 }, { "epoch": 19.61, "learning_rate": 6.783098591549297e-06, "loss": 0.3714, "step": 17492 }, { "epoch": 19.61, "learning_rate": 6.78169014084507e-06, "loss": 0.3333, "step": 17493 }, { "epoch": 19.61, "learning_rate": 6.780281690140845e-06, "loss": 0.4481, "step": 17494 }, { "epoch": 19.61, "learning_rate": 6.77887323943662e-06, "loss": 0.3084, "step": 17495 }, { "epoch": 19.61, "learning_rate": 6.777464788732395e-06, "loss": 0.3268, "step": 17496 }, { "epoch": 19.62, "learning_rate": 6.776056338028169e-06, "loss": 0.24, "step": 17497 }, { "epoch": 19.62, "learning_rate": 6.774647887323944e-06, "loss": 0.2298, "step": 17498 }, { "epoch": 19.62, "learning_rate": 6.773239436619718e-06, "loss": 0.4268, "step": 17499 }, { "epoch": 19.62, "learning_rate": 6.771830985915494e-06, "loss": 0.383, "step": 17500 }, { "epoch": 19.62, "learning_rate": 6.7704225352112675e-06, "loss": 0.4288, "step": 17501 }, { "epoch": 19.62, "learning_rate": 6.769014084507042e-06, "loss": 0.3702, "step": 17502 }, { "epoch": 19.62, "learning_rate": 6.767605633802817e-06, "loss": 0.4311, "step": 17503 }, { "epoch": 19.62, "learning_rate": 6.766197183098592e-06, "loss": 0.3835, "step": 17504 }, { "epoch": 19.62, "learning_rate": 6.764788732394366e-06, "loss": 0.4015, "step": 17505 }, { "epoch": 19.63, "learning_rate": 6.763380281690141e-06, "loss": 0.5104, "step": 17506 }, { "epoch": 19.63, "learning_rate": 6.761971830985916e-06, "loss": 0.3275, "step": 17507 }, { "epoch": 19.63, "learning_rate": 6.760563380281691e-06, "loss": 0.3488, "step": 17508 }, { "epoch": 19.63, "learning_rate": 6.759154929577465e-06, "loss": 0.3974, "step": 17509 }, { "epoch": 19.63, "learning_rate": 6.757746478873239e-06, "loss": 0.393, "step": 17510 }, { "epoch": 19.63, "learning_rate": 6.756338028169015e-06, "loss": 0.3437, "step": 17511 }, { "epoch": 19.63, "learning_rate": 6.754929577464789e-06, "loss": 0.3149, "step": 17512 }, { "epoch": 19.63, "learning_rate": 6.753521126760563e-06, "loss": 0.3516, "step": 17513 }, { "epoch": 19.63, "learning_rate": 6.752112676056338e-06, "loss": 0.3913, "step": 17514 }, { "epoch": 19.64, "learning_rate": 6.750704225352113e-06, "loss": 0.3385, "step": 17515 }, { "epoch": 19.64, "learning_rate": 6.749295774647888e-06, "loss": 0.3788, "step": 17516 }, { "epoch": 19.64, "learning_rate": 6.747887323943662e-06, "loss": 0.3754, "step": 17517 }, { "epoch": 19.64, "learning_rate": 6.746478873239436e-06, "loss": 0.3889, "step": 17518 }, { "epoch": 19.64, "learning_rate": 6.745070422535212e-06, "loss": 0.4079, "step": 17519 }, { "epoch": 19.64, "learning_rate": 6.7436619718309865e-06, "loss": 0.322, "step": 17520 }, { "epoch": 19.64, "learning_rate": 6.74225352112676e-06, "loss": 0.3903, "step": 17521 }, { "epoch": 19.64, "learning_rate": 6.740845070422536e-06, "loss": 0.2959, "step": 17522 }, { "epoch": 19.64, "learning_rate": 6.73943661971831e-06, "loss": 0.371, "step": 17523 }, { "epoch": 19.65, "learning_rate": 6.738028169014085e-06, "loss": 0.3478, "step": 17524 }, { "epoch": 19.65, "learning_rate": 6.736619718309859e-06, "loss": 0.3246, "step": 17525 }, { "epoch": 19.65, "learning_rate": 6.735211267605634e-06, "loss": 0.362, "step": 17526 }, { "epoch": 19.65, "learning_rate": 6.733802816901409e-06, "loss": 0.3004, "step": 17527 }, { "epoch": 19.65, "learning_rate": 6.7323943661971836e-06, "loss": 0.3376, "step": 17528 }, { "epoch": 19.65, "learning_rate": 6.730985915492957e-06, "loss": 0.424, "step": 17529 }, { "epoch": 19.65, "learning_rate": 6.729577464788733e-06, "loss": 0.3584, "step": 17530 }, { "epoch": 19.65, "learning_rate": 6.7281690140845075e-06, "loss": 0.3412, "step": 17531 }, { "epoch": 19.65, "learning_rate": 6.726760563380282e-06, "loss": 0.3766, "step": 17532 }, { "epoch": 19.66, "learning_rate": 6.725352112676056e-06, "loss": 0.3874, "step": 17533 }, { "epoch": 19.66, "learning_rate": 6.723943661971831e-06, "loss": 0.3471, "step": 17534 }, { "epoch": 19.66, "learning_rate": 6.722535211267606e-06, "loss": 0.3635, "step": 17535 }, { "epoch": 19.66, "learning_rate": 6.721126760563381e-06, "loss": 0.3715, "step": 17536 }, { "epoch": 19.66, "learning_rate": 6.7197183098591545e-06, "loss": 0.3345, "step": 17537 }, { "epoch": 19.66, "learning_rate": 6.71830985915493e-06, "loss": 0.369, "step": 17538 }, { "epoch": 19.66, "learning_rate": 6.7169014084507046e-06, "loss": 0.2848, "step": 17539 }, { "epoch": 19.66, "learning_rate": 6.715492957746479e-06, "loss": 0.2626, "step": 17540 }, { "epoch": 19.66, "learning_rate": 6.714084507042254e-06, "loss": 0.4434, "step": 17541 }, { "epoch": 19.67, "learning_rate": 6.7126760563380285e-06, "loss": 0.3299, "step": 17542 }, { "epoch": 19.67, "learning_rate": 6.711267605633803e-06, "loss": 0.3528, "step": 17543 }, { "epoch": 19.67, "learning_rate": 6.709859154929578e-06, "loss": 0.2516, "step": 17544 }, { "epoch": 19.67, "learning_rate": 6.708450704225352e-06, "loss": 0.4111, "step": 17545 }, { "epoch": 19.67, "learning_rate": 6.707042253521127e-06, "loss": 0.2693, "step": 17546 }, { "epoch": 19.67, "learning_rate": 6.705633802816902e-06, "loss": 0.1997, "step": 17547 }, { "epoch": 19.67, "learning_rate": 6.704225352112676e-06, "loss": 0.4825, "step": 17548 }, { "epoch": 19.67, "learning_rate": 6.702816901408451e-06, "loss": 0.5478, "step": 17549 }, { "epoch": 19.67, "learning_rate": 6.701408450704226e-06, "loss": 0.4958, "step": 17550 }, { "epoch": 19.68, "learning_rate": 6.7e-06, "loss": 0.4352, "step": 17551 }, { "epoch": 19.68, "learning_rate": 6.698591549295775e-06, "loss": 0.4824, "step": 17552 }, { "epoch": 19.68, "learning_rate": 6.6971830985915495e-06, "loss": 0.4881, "step": 17553 }, { "epoch": 19.68, "learning_rate": 6.695774647887324e-06, "loss": 0.3969, "step": 17554 }, { "epoch": 19.68, "learning_rate": 6.694366197183099e-06, "loss": 0.3664, "step": 17555 }, { "epoch": 19.68, "learning_rate": 6.692957746478873e-06, "loss": 0.3674, "step": 17556 }, { "epoch": 19.68, "learning_rate": 6.691549295774648e-06, "loss": 0.4239, "step": 17557 }, { "epoch": 19.68, "learning_rate": 6.690140845070423e-06, "loss": 0.4009, "step": 17558 }, { "epoch": 19.68, "learning_rate": 6.688732394366197e-06, "loss": 0.3971, "step": 17559 }, { "epoch": 19.69, "learning_rate": 6.687323943661972e-06, "loss": 0.3769, "step": 17560 }, { "epoch": 19.69, "learning_rate": 6.685915492957747e-06, "loss": 0.375, "step": 17561 }, { "epoch": 19.69, "learning_rate": 6.684507042253521e-06, "loss": 0.3712, "step": 17562 }, { "epoch": 19.69, "learning_rate": 6.683098591549296e-06, "loss": 0.3881, "step": 17563 }, { "epoch": 19.69, "learning_rate": 6.6816901408450705e-06, "loss": 0.3877, "step": 17564 }, { "epoch": 19.69, "learning_rate": 6.680281690140845e-06, "loss": 0.333, "step": 17565 }, { "epoch": 19.69, "learning_rate": 6.67887323943662e-06, "loss": 0.3971, "step": 17566 }, { "epoch": 19.69, "learning_rate": 6.6774647887323944e-06, "loss": 0.3691, "step": 17567 }, { "epoch": 19.7, "learning_rate": 6.676056338028169e-06, "loss": 0.4622, "step": 17568 }, { "epoch": 19.7, "learning_rate": 6.674647887323944e-06, "loss": 0.3302, "step": 17569 }, { "epoch": 19.7, "learning_rate": 6.673239436619718e-06, "loss": 0.3634, "step": 17570 }, { "epoch": 19.7, "learning_rate": 6.671830985915493e-06, "loss": 0.3272, "step": 17571 }, { "epoch": 19.7, "learning_rate": 6.670422535211268e-06, "loss": 0.2593, "step": 17572 }, { "epoch": 19.7, "learning_rate": 6.669014084507042e-06, "loss": 0.3773, "step": 17573 }, { "epoch": 19.7, "learning_rate": 6.667605633802817e-06, "loss": 0.4857, "step": 17574 }, { "epoch": 19.7, "learning_rate": 6.666197183098592e-06, "loss": 0.3718, "step": 17575 }, { "epoch": 19.7, "learning_rate": 6.664788732394366e-06, "loss": 0.4514, "step": 17576 }, { "epoch": 19.71, "learning_rate": 6.663380281690141e-06, "loss": 0.28, "step": 17577 }, { "epoch": 19.71, "learning_rate": 6.6619718309859154e-06, "loss": 0.4234, "step": 17578 }, { "epoch": 19.71, "learning_rate": 6.660563380281691e-06, "loss": 0.3568, "step": 17579 }, { "epoch": 19.71, "learning_rate": 6.659154929577465e-06, "loss": 0.2601, "step": 17580 }, { "epoch": 19.71, "learning_rate": 6.657746478873239e-06, "loss": 0.4665, "step": 17581 }, { "epoch": 19.71, "learning_rate": 6.656338028169014e-06, "loss": 0.3059, "step": 17582 }, { "epoch": 19.71, "learning_rate": 6.6549295774647895e-06, "loss": 0.3887, "step": 17583 }, { "epoch": 19.71, "learning_rate": 6.653521126760563e-06, "loss": 0.354, "step": 17584 }, { "epoch": 19.71, "learning_rate": 6.652112676056338e-06, "loss": 0.3601, "step": 17585 }, { "epoch": 19.72, "learning_rate": 6.6507042253521125e-06, "loss": 0.3455, "step": 17586 }, { "epoch": 19.72, "learning_rate": 6.649295774647888e-06, "loss": 0.3718, "step": 17587 }, { "epoch": 19.72, "learning_rate": 6.647887323943662e-06, "loss": 0.4008, "step": 17588 }, { "epoch": 19.72, "learning_rate": 6.6464788732394365e-06, "loss": 0.3971, "step": 17589 }, { "epoch": 19.72, "learning_rate": 6.645070422535211e-06, "loss": 0.3926, "step": 17590 }, { "epoch": 19.72, "learning_rate": 6.6436619718309866e-06, "loss": 0.2904, "step": 17591 }, { "epoch": 19.72, "learning_rate": 6.64225352112676e-06, "loss": 0.36, "step": 17592 }, { "epoch": 19.72, "learning_rate": 6.640845070422535e-06, "loss": 0.3325, "step": 17593 }, { "epoch": 19.72, "learning_rate": 6.6394366197183105e-06, "loss": 0.1991, "step": 17594 }, { "epoch": 19.73, "learning_rate": 6.638028169014085e-06, "loss": 0.2088, "step": 17595 }, { "epoch": 19.73, "learning_rate": 6.636619718309859e-06, "loss": 0.4295, "step": 17596 }, { "epoch": 19.73, "learning_rate": 6.6352112676056336e-06, "loss": 0.2392, "step": 17597 }, { "epoch": 19.73, "learning_rate": 6.633802816901409e-06, "loss": 0.3646, "step": 17598 }, { "epoch": 19.73, "learning_rate": 6.632394366197184e-06, "loss": 0.5117, "step": 17599 }, { "epoch": 19.73, "learning_rate": 6.6309859154929575e-06, "loss": 0.5685, "step": 17600 }, { "epoch": 19.73, "learning_rate": 6.629577464788732e-06, "loss": 0.4476, "step": 17601 }, { "epoch": 19.73, "learning_rate": 6.628169014084508e-06, "loss": 0.469, "step": 17602 }, { "epoch": 19.73, "learning_rate": 6.626760563380282e-06, "loss": 0.3534, "step": 17603 }, { "epoch": 19.74, "learning_rate": 6.625352112676056e-06, "loss": 0.4821, "step": 17604 }, { "epoch": 19.74, "learning_rate": 6.623943661971831e-06, "loss": 0.3962, "step": 17605 }, { "epoch": 19.74, "learning_rate": 6.622535211267606e-06, "loss": 0.442, "step": 17606 }, { "epoch": 19.74, "learning_rate": 6.621126760563381e-06, "loss": 0.3959, "step": 17607 }, { "epoch": 19.74, "learning_rate": 6.6197183098591546e-06, "loss": 0.3786, "step": 17608 }, { "epoch": 19.74, "learning_rate": 6.61830985915493e-06, "loss": 0.3935, "step": 17609 }, { "epoch": 19.74, "learning_rate": 6.616901408450705e-06, "loss": 0.3756, "step": 17610 }, { "epoch": 19.74, "learning_rate": 6.615492957746479e-06, "loss": 0.3475, "step": 17611 }, { "epoch": 19.74, "learning_rate": 6.614084507042253e-06, "loss": 0.3427, "step": 17612 }, { "epoch": 19.75, "learning_rate": 6.612676056338029e-06, "loss": 0.3281, "step": 17613 }, { "epoch": 19.75, "learning_rate": 6.611267605633803e-06, "loss": 0.388, "step": 17614 }, { "epoch": 19.75, "learning_rate": 6.609859154929578e-06, "loss": 0.4415, "step": 17615 }, { "epoch": 19.75, "learning_rate": 6.608450704225352e-06, "loss": 0.4168, "step": 17616 }, { "epoch": 19.75, "learning_rate": 6.607042253521127e-06, "loss": 0.4332, "step": 17617 }, { "epoch": 19.75, "learning_rate": 6.605633802816902e-06, "loss": 0.3807, "step": 17618 }, { "epoch": 19.75, "learning_rate": 6.6042253521126764e-06, "loss": 0.3155, "step": 17619 }, { "epoch": 19.75, "learning_rate": 6.60281690140845e-06, "loss": 0.4823, "step": 17620 }, { "epoch": 19.75, "learning_rate": 6.601408450704226e-06, "loss": 0.378, "step": 17621 }, { "epoch": 19.76, "learning_rate": 6.6e-06, "loss": 0.2885, "step": 17622 }, { "epoch": 19.76, "learning_rate": 6.598591549295775e-06, "loss": 0.3918, "step": 17623 }, { "epoch": 19.76, "learning_rate": 6.597183098591549e-06, "loss": 0.4108, "step": 17624 }, { "epoch": 19.76, "learning_rate": 6.595774647887324e-06, "loss": 0.4318, "step": 17625 }, { "epoch": 19.76, "learning_rate": 6.594366197183099e-06, "loss": 0.3901, "step": 17626 }, { "epoch": 19.76, "learning_rate": 6.5929577464788735e-06, "loss": 0.3343, "step": 17627 }, { "epoch": 19.76, "learning_rate": 6.591549295774648e-06, "loss": 0.407, "step": 17628 }, { "epoch": 19.76, "learning_rate": 6.590140845070423e-06, "loss": 0.3237, "step": 17629 }, { "epoch": 19.76, "learning_rate": 6.5887323943661974e-06, "loss": 0.3293, "step": 17630 }, { "epoch": 19.77, "learning_rate": 6.587323943661972e-06, "loss": 0.3468, "step": 17631 }, { "epoch": 19.77, "learning_rate": 6.585915492957747e-06, "loss": 0.3734, "step": 17632 }, { "epoch": 19.77, "learning_rate": 6.584507042253521e-06, "loss": 0.4135, "step": 17633 }, { "epoch": 19.77, "learning_rate": 6.583098591549296e-06, "loss": 0.3481, "step": 17634 }, { "epoch": 19.77, "learning_rate": 6.581690140845071e-06, "loss": 0.3906, "step": 17635 }, { "epoch": 19.77, "learning_rate": 6.580281690140845e-06, "loss": 0.322, "step": 17636 }, { "epoch": 19.77, "learning_rate": 6.57887323943662e-06, "loss": 0.3104, "step": 17637 }, { "epoch": 19.77, "learning_rate": 6.5774647887323945e-06, "loss": 0.3713, "step": 17638 }, { "epoch": 19.77, "learning_rate": 6.576056338028169e-06, "loss": 0.2771, "step": 17639 }, { "epoch": 19.78, "learning_rate": 6.574647887323944e-06, "loss": 0.3772, "step": 17640 }, { "epoch": 19.78, "learning_rate": 6.5732394366197184e-06, "loss": 0.2431, "step": 17641 }, { "epoch": 19.78, "learning_rate": 6.571830985915493e-06, "loss": 0.3573, "step": 17642 }, { "epoch": 19.78, "learning_rate": 6.5704225352112686e-06, "loss": 0.3487, "step": 17643 }, { "epoch": 19.78, "learning_rate": 6.569014084507042e-06, "loss": 0.2363, "step": 17644 }, { "epoch": 19.78, "learning_rate": 6.567605633802817e-06, "loss": 0.3748, "step": 17645 }, { "epoch": 19.78, "learning_rate": 6.566197183098592e-06, "loss": 0.352, "step": 17646 }, { "epoch": 19.78, "learning_rate": 6.564788732394367e-06, "loss": 0.3189, "step": 17647 }, { "epoch": 19.78, "learning_rate": 6.563380281690141e-06, "loss": 0.2956, "step": 17648 }, { "epoch": 19.79, "learning_rate": 6.5619718309859155e-06, "loss": 0.4694, "step": 17649 }, { "epoch": 19.79, "learning_rate": 6.56056338028169e-06, "loss": 0.4543, "step": 17650 }, { "epoch": 19.79, "learning_rate": 6.559154929577466e-06, "loss": 0.449, "step": 17651 }, { "epoch": 19.79, "learning_rate": 6.5577464788732395e-06, "loss": 0.405, "step": 17652 }, { "epoch": 19.79, "learning_rate": 6.556338028169014e-06, "loss": 0.4038, "step": 17653 }, { "epoch": 19.79, "learning_rate": 6.554929577464789e-06, "loss": 0.3999, "step": 17654 }, { "epoch": 19.79, "learning_rate": 6.553521126760564e-06, "loss": 0.4105, "step": 17655 }, { "epoch": 19.79, "learning_rate": 6.552112676056338e-06, "loss": 0.3054, "step": 17656 }, { "epoch": 19.79, "learning_rate": 6.550704225352113e-06, "loss": 0.4329, "step": 17657 }, { "epoch": 19.8, "learning_rate": 6.549295774647887e-06, "loss": 0.4268, "step": 17658 }, { "epoch": 19.8, "learning_rate": 6.547887323943663e-06, "loss": 0.3818, "step": 17659 }, { "epoch": 19.8, "learning_rate": 6.5464788732394366e-06, "loss": 0.3763, "step": 17660 }, { "epoch": 19.8, "learning_rate": 6.545070422535211e-06, "loss": 0.3977, "step": 17661 }, { "epoch": 19.8, "learning_rate": 6.543661971830987e-06, "loss": 0.3811, "step": 17662 }, { "epoch": 19.8, "learning_rate": 6.5422535211267605e-06, "loss": 0.3404, "step": 17663 }, { "epoch": 19.8, "learning_rate": 6.540845070422535e-06, "loss": 0.4229, "step": 17664 }, { "epoch": 19.8, "learning_rate": 6.53943661971831e-06, "loss": 0.3963, "step": 17665 }, { "epoch": 19.8, "learning_rate": 6.538028169014085e-06, "loss": 0.4047, "step": 17666 }, { "epoch": 19.81, "learning_rate": 6.536619718309859e-06, "loss": 0.3625, "step": 17667 }, { "epoch": 19.81, "learning_rate": 6.535211267605634e-06, "loss": 0.3242, "step": 17668 }, { "epoch": 19.81, "learning_rate": 6.533802816901408e-06, "loss": 0.2621, "step": 17669 }, { "epoch": 19.81, "learning_rate": 6.532394366197184e-06, "loss": 0.351, "step": 17670 }, { "epoch": 19.81, "learning_rate": 6.5309859154929576e-06, "loss": 0.3544, "step": 17671 }, { "epoch": 19.81, "learning_rate": 6.529577464788732e-06, "loss": 0.3969, "step": 17672 }, { "epoch": 19.81, "learning_rate": 6.528169014084507e-06, "loss": 0.3537, "step": 17673 }, { "epoch": 19.81, "learning_rate": 6.526760563380282e-06, "loss": 0.2777, "step": 17674 }, { "epoch": 19.82, "learning_rate": 6.525352112676056e-06, "loss": 0.2891, "step": 17675 }, { "epoch": 19.82, "learning_rate": 6.523943661971831e-06, "loss": 0.394, "step": 17676 }, { "epoch": 19.82, "learning_rate": 6.522535211267606e-06, "loss": 0.3155, "step": 17677 }, { "epoch": 19.82, "learning_rate": 6.521126760563381e-06, "loss": 0.3883, "step": 17678 }, { "epoch": 19.82, "learning_rate": 6.519718309859155e-06, "loss": 0.3684, "step": 17679 }, { "epoch": 19.82, "learning_rate": 6.518309859154929e-06, "loss": 0.3454, "step": 17680 }, { "epoch": 19.82, "learning_rate": 6.516901408450705e-06, "loss": 0.3517, "step": 17681 }, { "epoch": 19.82, "learning_rate": 6.5154929577464794e-06, "loss": 0.2643, "step": 17682 }, { "epoch": 19.82, "learning_rate": 6.514084507042253e-06, "loss": 0.3764, "step": 17683 }, { "epoch": 19.83, "learning_rate": 6.512676056338028e-06, "loss": 0.3645, "step": 17684 }, { "epoch": 19.83, "learning_rate": 6.511267605633803e-06, "loss": 0.3247, "step": 17685 }, { "epoch": 19.83, "learning_rate": 6.509859154929578e-06, "loss": 0.3059, "step": 17686 }, { "epoch": 19.83, "learning_rate": 6.508450704225352e-06, "loss": 0.4471, "step": 17687 }, { "epoch": 19.83, "learning_rate": 6.507042253521126e-06, "loss": 0.3523, "step": 17688 }, { "epoch": 19.83, "learning_rate": 6.505633802816902e-06, "loss": 0.3575, "step": 17689 }, { "epoch": 19.83, "learning_rate": 6.5042253521126765e-06, "loss": 0.3836, "step": 17690 }, { "epoch": 19.83, "learning_rate": 6.50281690140845e-06, "loss": 0.3483, "step": 17691 }, { "epoch": 19.83, "learning_rate": 6.501408450704225e-06, "loss": 0.2973, "step": 17692 }, { "epoch": 19.84, "learning_rate": 6.5000000000000004e-06, "loss": 0.2832, "step": 17693 }, { "epoch": 19.84, "learning_rate": 6.498591549295775e-06, "loss": 0.3122, "step": 17694 }, { "epoch": 19.84, "learning_rate": 6.497183098591549e-06, "loss": 0.262, "step": 17695 }, { "epoch": 19.84, "learning_rate": 6.495774647887324e-06, "loss": 0.3105, "step": 17696 }, { "epoch": 19.84, "learning_rate": 6.494366197183099e-06, "loss": 0.3748, "step": 17697 }, { "epoch": 19.84, "learning_rate": 6.492957746478874e-06, "loss": 0.3483, "step": 17698 }, { "epoch": 19.84, "learning_rate": 6.4915492957746474e-06, "loss": 0.4753, "step": 17699 }, { "epoch": 19.84, "learning_rate": 6.490140845070423e-06, "loss": 0.4053, "step": 17700 }, { "epoch": 19.84, "learning_rate": 6.4887323943661975e-06, "loss": 0.5018, "step": 17701 }, { "epoch": 19.85, "learning_rate": 6.487323943661972e-06, "loss": 0.5012, "step": 17702 }, { "epoch": 19.85, "learning_rate": 6.485915492957746e-06, "loss": 0.4283, "step": 17703 }, { "epoch": 19.85, "learning_rate": 6.4845070422535215e-06, "loss": 0.345, "step": 17704 }, { "epoch": 19.85, "learning_rate": 6.483098591549296e-06, "loss": 0.4336, "step": 17705 }, { "epoch": 19.85, "learning_rate": 6.481690140845071e-06, "loss": 0.3377, "step": 17706 }, { "epoch": 19.85, "learning_rate": 6.4802816901408445e-06, "loss": 0.4035, "step": 17707 }, { "epoch": 19.85, "learning_rate": 6.47887323943662e-06, "loss": 0.3909, "step": 17708 }, { "epoch": 19.85, "learning_rate": 6.477464788732395e-06, "loss": 0.3834, "step": 17709 }, { "epoch": 19.85, "learning_rate": 6.476056338028169e-06, "loss": 0.2862, "step": 17710 }, { "epoch": 19.86, "learning_rate": 6.474647887323943e-06, "loss": 0.4274, "step": 17711 }, { "epoch": 19.86, "learning_rate": 6.4732394366197186e-06, "loss": 0.3677, "step": 17712 }, { "epoch": 19.86, "learning_rate": 6.471830985915493e-06, "loss": 0.377, "step": 17713 }, { "epoch": 19.86, "learning_rate": 6.470422535211268e-06, "loss": 0.3637, "step": 17714 }, { "epoch": 19.86, "learning_rate": 6.4690140845070425e-06, "loss": 0.3855, "step": 17715 }, { "epoch": 19.86, "learning_rate": 6.467605633802817e-06, "loss": 0.3514, "step": 17716 }, { "epoch": 19.86, "learning_rate": 6.466197183098592e-06, "loss": 0.3466, "step": 17717 }, { "epoch": 19.86, "learning_rate": 6.464788732394366e-06, "loss": 0.4572, "step": 17718 }, { "epoch": 19.86, "learning_rate": 6.463380281690141e-06, "loss": 0.4569, "step": 17719 }, { "epoch": 19.87, "learning_rate": 6.461971830985916e-06, "loss": 0.422, "step": 17720 }, { "epoch": 19.87, "learning_rate": 6.46056338028169e-06, "loss": 0.2242, "step": 17721 }, { "epoch": 19.87, "learning_rate": 6.459154929577465e-06, "loss": 0.3883, "step": 17722 }, { "epoch": 19.87, "learning_rate": 6.4577464788732396e-06, "loss": 0.4289, "step": 17723 }, { "epoch": 19.87, "learning_rate": 6.456338028169014e-06, "loss": 0.3765, "step": 17724 }, { "epoch": 19.87, "learning_rate": 6.454929577464789e-06, "loss": 0.3123, "step": 17725 }, { "epoch": 19.87, "learning_rate": 6.4535211267605635e-06, "loss": 0.3531, "step": 17726 }, { "epoch": 19.87, "learning_rate": 6.452112676056338e-06, "loss": 0.3587, "step": 17727 }, { "epoch": 19.87, "learning_rate": 6.450704225352113e-06, "loss": 0.3286, "step": 17728 }, { "epoch": 19.88, "learning_rate": 6.449295774647887e-06, "loss": 0.3449, "step": 17729 }, { "epoch": 19.88, "learning_rate": 6.447887323943663e-06, "loss": 0.2742, "step": 17730 }, { "epoch": 19.88, "learning_rate": 6.446478873239437e-06, "loss": 0.4097, "step": 17731 }, { "epoch": 19.88, "learning_rate": 6.445070422535211e-06, "loss": 0.3131, "step": 17732 }, { "epoch": 19.88, "learning_rate": 6.443661971830986e-06, "loss": 0.35, "step": 17733 }, { "epoch": 19.88, "learning_rate": 6.4422535211267614e-06, "loss": 0.3016, "step": 17734 }, { "epoch": 19.88, "learning_rate": 6.440845070422535e-06, "loss": 0.3256, "step": 17735 }, { "epoch": 19.88, "learning_rate": 6.43943661971831e-06, "loss": 0.3272, "step": 17736 }, { "epoch": 19.88, "learning_rate": 6.4380281690140845e-06, "loss": 0.3615, "step": 17737 }, { "epoch": 19.89, "learning_rate": 6.43661971830986e-06, "loss": 0.3197, "step": 17738 }, { "epoch": 19.89, "learning_rate": 6.435211267605634e-06, "loss": 0.359, "step": 17739 }, { "epoch": 19.89, "learning_rate": 6.433802816901408e-06, "loss": 0.3778, "step": 17740 }, { "epoch": 19.89, "learning_rate": 6.432394366197183e-06, "loss": 0.2684, "step": 17741 }, { "epoch": 19.89, "learning_rate": 6.4309859154929585e-06, "loss": 0.4471, "step": 17742 }, { "epoch": 19.89, "learning_rate": 6.429577464788732e-06, "loss": 0.3862, "step": 17743 }, { "epoch": 19.89, "learning_rate": 6.428169014084507e-06, "loss": 0.2633, "step": 17744 }, { "epoch": 19.89, "learning_rate": 6.426760563380282e-06, "loss": 0.2961, "step": 17745 }, { "epoch": 19.89, "learning_rate": 6.425352112676057e-06, "loss": 0.3959, "step": 17746 }, { "epoch": 19.9, "learning_rate": 6.423943661971831e-06, "loss": 0.4122, "step": 17747 }, { "epoch": 19.9, "learning_rate": 6.4225352112676055e-06, "loss": 0.2229, "step": 17748 }, { "epoch": 19.9, "learning_rate": 6.421126760563381e-06, "loss": 0.6432, "step": 17749 }, { "epoch": 19.9, "learning_rate": 6.419718309859156e-06, "loss": 0.4489, "step": 17750 }, { "epoch": 19.9, "learning_rate": 6.418309859154929e-06, "loss": 0.3891, "step": 17751 }, { "epoch": 19.9, "learning_rate": 6.416901408450704e-06, "loss": 0.3799, "step": 17752 }, { "epoch": 19.9, "learning_rate": 6.4154929577464795e-06, "loss": 0.4197, "step": 17753 }, { "epoch": 19.9, "learning_rate": 6.414084507042254e-06, "loss": 0.3883, "step": 17754 }, { "epoch": 19.9, "learning_rate": 6.412676056338028e-06, "loss": 0.3703, "step": 17755 }, { "epoch": 19.91, "learning_rate": 6.411267605633803e-06, "loss": 0.4049, "step": 17756 }, { "epoch": 19.91, "learning_rate": 6.409859154929578e-06, "loss": 0.3859, "step": 17757 }, { "epoch": 19.91, "learning_rate": 6.408450704225353e-06, "loss": 0.3958, "step": 17758 }, { "epoch": 19.91, "learning_rate": 6.4070422535211265e-06, "loss": 0.3196, "step": 17759 }, { "epoch": 19.91, "learning_rate": 6.405633802816901e-06, "loss": 0.3638, "step": 17760 }, { "epoch": 19.91, "learning_rate": 6.404225352112677e-06, "loss": 0.4058, "step": 17761 }, { "epoch": 19.91, "learning_rate": 6.402816901408451e-06, "loss": 0.4411, "step": 17762 }, { "epoch": 19.91, "learning_rate": 6.401408450704225e-06, "loss": 0.3863, "step": 17763 }, { "epoch": 19.91, "learning_rate": 6.4000000000000006e-06, "loss": 0.3572, "step": 17764 }, { "epoch": 19.92, "learning_rate": 6.398591549295775e-06, "loss": 0.4279, "step": 17765 }, { "epoch": 19.92, "learning_rate": 6.39718309859155e-06, "loss": 0.4271, "step": 17766 }, { "epoch": 19.92, "learning_rate": 6.395774647887324e-06, "loss": 0.3367, "step": 17767 }, { "epoch": 19.92, "learning_rate": 6.394366197183099e-06, "loss": 0.3907, "step": 17768 }, { "epoch": 19.92, "learning_rate": 6.392957746478874e-06, "loss": 0.3419, "step": 17769 }, { "epoch": 19.92, "learning_rate": 6.391549295774648e-06, "loss": 0.4454, "step": 17770 }, { "epoch": 19.92, "learning_rate": 6.390140845070422e-06, "loss": 0.4124, "step": 17771 }, { "epoch": 19.92, "learning_rate": 6.388732394366198e-06, "loss": 0.341, "step": 17772 }, { "epoch": 19.92, "learning_rate": 6.387323943661972e-06, "loss": 0.3287, "step": 17773 }, { "epoch": 19.93, "learning_rate": 6.385915492957747e-06, "loss": 0.396, "step": 17774 }, { "epoch": 19.93, "learning_rate": 6.384507042253521e-06, "loss": 0.4662, "step": 17775 }, { "epoch": 19.93, "learning_rate": 6.383098591549296e-06, "loss": 0.3244, "step": 17776 }, { "epoch": 19.93, "learning_rate": 6.381690140845071e-06, "loss": 0.374, "step": 17777 }, { "epoch": 19.93, "learning_rate": 6.380281690140845e-06, "loss": 0.426, "step": 17778 }, { "epoch": 19.93, "learning_rate": 6.378873239436619e-06, "loss": 0.3238, "step": 17779 }, { "epoch": 19.93, "learning_rate": 6.377464788732395e-06, "loss": 0.3296, "step": 17780 }, { "epoch": 19.93, "learning_rate": 6.376056338028169e-06, "loss": 0.3663, "step": 17781 }, { "epoch": 19.93, "learning_rate": 6.374647887323943e-06, "loss": 0.3331, "step": 17782 }, { "epoch": 19.94, "learning_rate": 6.373239436619719e-06, "loss": 0.311, "step": 17783 }, { "epoch": 19.94, "learning_rate": 6.371830985915493e-06, "loss": 0.3571, "step": 17784 }, { "epoch": 19.94, "learning_rate": 6.370422535211268e-06, "loss": 0.3161, "step": 17785 }, { "epoch": 19.94, "learning_rate": 6.369014084507042e-06, "loss": 0.3326, "step": 17786 }, { "epoch": 19.94, "learning_rate": 6.367605633802817e-06, "loss": 0.3138, "step": 17787 }, { "epoch": 19.94, "learning_rate": 6.366197183098592e-06, "loss": 0.4167, "step": 17788 }, { "epoch": 19.94, "learning_rate": 6.3647887323943665e-06, "loss": 0.3279, "step": 17789 }, { "epoch": 19.94, "learning_rate": 6.36338028169014e-06, "loss": 0.348, "step": 17790 }, { "epoch": 19.95, "learning_rate": 6.361971830985916e-06, "loss": 0.2583, "step": 17791 }, { "epoch": 19.95, "learning_rate": 6.36056338028169e-06, "loss": 0.2923, "step": 17792 }, { "epoch": 19.95, "learning_rate": 6.359154929577465e-06, "loss": 0.304, "step": 17793 }, { "epoch": 19.95, "learning_rate": 6.357746478873239e-06, "loss": 0.2732, "step": 17794 }, { "epoch": 19.95, "learning_rate": 6.356338028169014e-06, "loss": 0.278, "step": 17795 }, { "epoch": 19.95, "learning_rate": 6.354929577464789e-06, "loss": 0.3052, "step": 17796 }, { "epoch": 19.95, "learning_rate": 6.353521126760564e-06, "loss": 0.3715, "step": 17797 }, { "epoch": 19.95, "learning_rate": 6.352112676056337e-06, "loss": 0.2259, "step": 17798 }, { "epoch": 19.95, "learning_rate": 6.350704225352113e-06, "loss": 0.4768, "step": 17799 }, { "epoch": 19.96, "learning_rate": 6.3492957746478875e-06, "loss": 0.522, "step": 17800 }, { "epoch": 19.96, "learning_rate": 6.347887323943662e-06, "loss": 0.3593, "step": 17801 }, { "epoch": 19.96, "learning_rate": 6.346478873239437e-06, "loss": 0.4078, "step": 17802 }, { "epoch": 19.96, "learning_rate": 6.345070422535211e-06, "loss": 0.3568, "step": 17803 }, { "epoch": 19.96, "learning_rate": 6.343661971830986e-06, "loss": 0.4294, "step": 17804 }, { "epoch": 19.96, "learning_rate": 6.342253521126761e-06, "loss": 0.4962, "step": 17805 }, { "epoch": 19.96, "learning_rate": 6.340845070422535e-06, "loss": 0.4461, "step": 17806 }, { "epoch": 19.96, "learning_rate": 6.33943661971831e-06, "loss": 0.3483, "step": 17807 }, { "epoch": 19.96, "learning_rate": 6.338028169014085e-06, "loss": 0.3712, "step": 17808 }, { "epoch": 19.97, "learning_rate": 6.336619718309859e-06, "loss": 0.3823, "step": 17809 }, { "epoch": 19.97, "learning_rate": 6.335211267605634e-06, "loss": 0.3889, "step": 17810 }, { "epoch": 19.97, "learning_rate": 6.3338028169014085e-06, "loss": 0.3919, "step": 17811 }, { "epoch": 19.97, "learning_rate": 6.332394366197183e-06, "loss": 0.4596, "step": 17812 }, { "epoch": 19.97, "learning_rate": 6.330985915492958e-06, "loss": 0.3332, "step": 17813 }, { "epoch": 19.97, "learning_rate": 6.3295774647887324e-06, "loss": 0.3195, "step": 17814 }, { "epoch": 19.97, "learning_rate": 6.328169014084507e-06, "loss": 0.4484, "step": 17815 }, { "epoch": 19.97, "learning_rate": 6.326760563380282e-06, "loss": 0.3311, "step": 17816 }, { "epoch": 19.97, "learning_rate": 6.325352112676057e-06, "loss": 0.3294, "step": 17817 }, { "epoch": 19.98, "learning_rate": 6.323943661971831e-06, "loss": 0.3494, "step": 17818 }, { "epoch": 19.98, "learning_rate": 6.322535211267606e-06, "loss": 0.2942, "step": 17819 }, { "epoch": 19.98, "learning_rate": 6.32112676056338e-06, "loss": 0.3023, "step": 17820 }, { "epoch": 19.98, "learning_rate": 6.319718309859156e-06, "loss": 0.3421, "step": 17821 }, { "epoch": 19.98, "learning_rate": 6.3183098591549295e-06, "loss": 0.4467, "step": 17822 }, { "epoch": 19.98, "learning_rate": 6.316901408450704e-06, "loss": 0.353, "step": 17823 }, { "epoch": 19.98, "learning_rate": 6.315492957746479e-06, "loss": 0.3296, "step": 17824 }, { "epoch": 19.98, "learning_rate": 6.314084507042254e-06, "loss": 0.427, "step": 17825 }, { "epoch": 19.98, "learning_rate": 6.312676056338028e-06, "loss": 0.4274, "step": 17826 }, { "epoch": 19.99, "learning_rate": 6.311267605633803e-06, "loss": 0.4669, "step": 17827 }, { "epoch": 19.99, "learning_rate": 6.309859154929577e-06, "loss": 0.3843, "step": 17828 }, { "epoch": 19.99, "learning_rate": 6.308450704225353e-06, "loss": 0.3416, "step": 17829 }, { "epoch": 19.99, "learning_rate": 6.307042253521127e-06, "loss": 0.3228, "step": 17830 }, { "epoch": 19.99, "learning_rate": 6.305633802816901e-06, "loss": 0.3153, "step": 17831 }, { "epoch": 19.99, "learning_rate": 6.304225352112676e-06, "loss": 0.3346, "step": 17832 }, { "epoch": 19.99, "learning_rate": 6.302816901408451e-06, "loss": 0.3299, "step": 17833 }, { "epoch": 19.99, "learning_rate": 6.301408450704225e-06, "loss": 0.3511, "step": 17834 }, { "epoch": 19.99, "learning_rate": 6.3e-06, "loss": 0.3806, "step": 17835 }, { "epoch": 20.0, "learning_rate": 6.298591549295775e-06, "loss": 0.2421, "step": 17836 }, { "epoch": 20.0, "learning_rate": 6.29718309859155e-06, "loss": 0.4449, "step": 17837 }, { "epoch": 20.0, "learning_rate": 6.295774647887324e-06, "loss": 0.18, "step": 17838 }, { "epoch": 20.0, "learning_rate": 6.294366197183098e-06, "loss": 0.3116, "step": 17839 }, { "epoch": 20.0, "learning_rate": 6.292957746478874e-06, "loss": 0.4079, "step": 17840 }, { "epoch": 20.0, "learning_rate": 6.2915492957746485e-06, "loss": 0.5361, "step": 17841 }, { "epoch": 20.0, "learning_rate": 6.290140845070422e-06, "loss": 0.4976, "step": 17842 }, { "epoch": 20.0, "learning_rate": 6.288732394366197e-06, "loss": 0.3425, "step": 17843 }, { "epoch": 20.0, "learning_rate": 6.287323943661972e-06, "loss": 0.4092, "step": 17844 }, { "epoch": 20.01, "learning_rate": 6.285915492957747e-06, "loss": 0.4317, "step": 17845 }, { "epoch": 20.01, "learning_rate": 6.284507042253521e-06, "loss": 0.3575, "step": 17846 }, { "epoch": 20.01, "learning_rate": 6.2830985915492955e-06, "loss": 0.4553, "step": 17847 }, { "epoch": 20.01, "learning_rate": 6.281690140845071e-06, "loss": 0.3985, "step": 17848 }, { "epoch": 20.01, "learning_rate": 6.280281690140846e-06, "loss": 0.3905, "step": 17849 }, { "epoch": 20.01, "learning_rate": 6.278873239436619e-06, "loss": 0.3093, "step": 17850 }, { "epoch": 20.01, "learning_rate": 6.277464788732395e-06, "loss": 0.4195, "step": 17851 }, { "epoch": 20.01, "learning_rate": 6.2760563380281695e-06, "loss": 0.3143, "step": 17852 }, { "epoch": 20.01, "learning_rate": 6.274647887323944e-06, "loss": 0.3668, "step": 17853 }, { "epoch": 20.02, "learning_rate": 6.273239436619718e-06, "loss": 0.341, "step": 17854 }, { "epoch": 20.02, "learning_rate": 6.271830985915493e-06, "loss": 0.3782, "step": 17855 }, { "epoch": 20.02, "learning_rate": 6.270422535211268e-06, "loss": 0.3145, "step": 17856 }, { "epoch": 20.02, "learning_rate": 6.269014084507043e-06, "loss": 0.4402, "step": 17857 }, { "epoch": 20.02, "learning_rate": 6.2676056338028165e-06, "loss": 0.3904, "step": 17858 }, { "epoch": 20.02, "learning_rate": 6.266197183098592e-06, "loss": 0.3479, "step": 17859 }, { "epoch": 20.02, "learning_rate": 6.264788732394367e-06, "loss": 0.2724, "step": 17860 }, { "epoch": 20.02, "learning_rate": 6.263380281690141e-06, "loss": 0.2694, "step": 17861 }, { "epoch": 20.02, "learning_rate": 6.261971830985915e-06, "loss": 0.358, "step": 17862 }, { "epoch": 20.03, "learning_rate": 6.2605633802816905e-06, "loss": 0.3012, "step": 17863 }, { "epoch": 20.03, "learning_rate": 6.259154929577465e-06, "loss": 0.3097, "step": 17864 }, { "epoch": 20.03, "learning_rate": 6.25774647887324e-06, "loss": 0.3443, "step": 17865 }, { "epoch": 20.03, "learning_rate": 6.256338028169014e-06, "loss": 0.3305, "step": 17866 }, { "epoch": 20.03, "learning_rate": 6.254929577464789e-06, "loss": 0.3857, "step": 17867 }, { "epoch": 20.03, "learning_rate": 6.253521126760564e-06, "loss": 0.3214, "step": 17868 }, { "epoch": 20.03, "learning_rate": 6.252112676056338e-06, "loss": 0.3436, "step": 17869 }, { "epoch": 20.03, "learning_rate": 6.250704225352113e-06, "loss": 0.3876, "step": 17870 }, { "epoch": 20.03, "learning_rate": 6.249295774647888e-06, "loss": 0.3424, "step": 17871 }, { "epoch": 20.04, "learning_rate": 6.247887323943662e-06, "loss": 0.4355, "step": 17872 }, { "epoch": 20.04, "learning_rate": 6.246478873239437e-06, "loss": 0.3089, "step": 17873 }, { "epoch": 20.04, "learning_rate": 6.2450704225352115e-06, "loss": 0.3136, "step": 17874 }, { "epoch": 20.04, "learning_rate": 6.243661971830986e-06, "loss": 0.3533, "step": 17875 }, { "epoch": 20.04, "learning_rate": 6.242253521126761e-06, "loss": 0.3473, "step": 17876 }, { "epoch": 20.04, "learning_rate": 6.2408450704225354e-06, "loss": 0.2961, "step": 17877 }, { "epoch": 20.04, "learning_rate": 6.23943661971831e-06, "loss": 0.308, "step": 17878 }, { "epoch": 20.04, "learning_rate": 6.238028169014085e-06, "loss": 0.3112, "step": 17879 }, { "epoch": 20.04, "learning_rate": 6.236619718309859e-06, "loss": 0.3143, "step": 17880 }, { "epoch": 20.05, "learning_rate": 6.235211267605634e-06, "loss": 0.3448, "step": 17881 }, { "epoch": 20.05, "learning_rate": 6.233802816901409e-06, "loss": 0.3617, "step": 17882 }, { "epoch": 20.05, "learning_rate": 6.232394366197183e-06, "loss": 0.2964, "step": 17883 }, { "epoch": 20.05, "learning_rate": 6.230985915492958e-06, "loss": 0.2631, "step": 17884 }, { "epoch": 20.05, "learning_rate": 6.229577464788733e-06, "loss": 0.2936, "step": 17885 }, { "epoch": 20.05, "learning_rate": 6.228169014084507e-06, "loss": 0.2552, "step": 17886 }, { "epoch": 20.05, "learning_rate": 6.226760563380282e-06, "loss": 0.2876, "step": 17887 }, { "epoch": 20.05, "learning_rate": 6.2253521126760565e-06, "loss": 0.3045, "step": 17888 }, { "epoch": 20.05, "learning_rate": 6.223943661971831e-06, "loss": 0.2449, "step": 17889 }, { "epoch": 20.06, "learning_rate": 6.222535211267606e-06, "loss": 0.2864, "step": 17890 }, { "epoch": 20.06, "learning_rate": 6.22112676056338e-06, "loss": 0.4294, "step": 17891 }, { "epoch": 20.06, "learning_rate": 6.219718309859155e-06, "loss": 0.3934, "step": 17892 }, { "epoch": 20.06, "learning_rate": 6.21830985915493e-06, "loss": 0.4766, "step": 17893 }, { "epoch": 20.06, "learning_rate": 6.216901408450704e-06, "loss": 0.4339, "step": 17894 }, { "epoch": 20.06, "learning_rate": 6.215492957746479e-06, "loss": 0.4268, "step": 17895 }, { "epoch": 20.06, "learning_rate": 6.2140845070422536e-06, "loss": 0.3685, "step": 17896 }, { "epoch": 20.06, "learning_rate": 6.212676056338028e-06, "loss": 0.3514, "step": 17897 }, { "epoch": 20.07, "learning_rate": 6.211267605633803e-06, "loss": 0.3623, "step": 17898 }, { "epoch": 20.07, "learning_rate": 6.2098591549295775e-06, "loss": 0.3899, "step": 17899 }, { "epoch": 20.07, "learning_rate": 6.208450704225352e-06, "loss": 0.3717, "step": 17900 }, { "epoch": 20.07, "learning_rate": 6.207042253521127e-06, "loss": 0.2607, "step": 17901 }, { "epoch": 20.07, "learning_rate": 6.205633802816901e-06, "loss": 0.3882, "step": 17902 }, { "epoch": 20.07, "learning_rate": 6.204225352112676e-06, "loss": 0.3568, "step": 17903 }, { "epoch": 20.07, "learning_rate": 6.2028169014084515e-06, "loss": 0.3028, "step": 17904 }, { "epoch": 20.07, "learning_rate": 6.201408450704225e-06, "loss": 0.3121, "step": 17905 }, { "epoch": 20.07, "learning_rate": 6.2e-06, "loss": 0.4402, "step": 17906 }, { "epoch": 20.08, "learning_rate": 6.1985915492957746e-06, "loss": 0.3558, "step": 17907 }, { "epoch": 20.08, "learning_rate": 6.19718309859155e-06, "loss": 0.356, "step": 17908 }, { "epoch": 20.08, "learning_rate": 6.195774647887324e-06, "loss": 0.4058, "step": 17909 }, { "epoch": 20.08, "learning_rate": 6.1943661971830985e-06, "loss": 0.2346, "step": 17910 }, { "epoch": 20.08, "learning_rate": 6.192957746478873e-06, "loss": 0.3428, "step": 17911 }, { "epoch": 20.08, "learning_rate": 6.191549295774649e-06, "loss": 0.3338, "step": 17912 }, { "epoch": 20.08, "learning_rate": 6.190140845070422e-06, "loss": 0.4421, "step": 17913 }, { "epoch": 20.08, "learning_rate": 6.188732394366197e-06, "loss": 0.3142, "step": 17914 }, { "epoch": 20.08, "learning_rate": 6.187323943661972e-06, "loss": 0.2983, "step": 17915 }, { "epoch": 20.09, "learning_rate": 6.185915492957747e-06, "loss": 0.3207, "step": 17916 }, { "epoch": 20.09, "learning_rate": 6.184507042253521e-06, "loss": 0.3817, "step": 17917 }, { "epoch": 20.09, "learning_rate": 6.183098591549296e-06, "loss": 0.3843, "step": 17918 }, { "epoch": 20.09, "learning_rate": 6.18169014084507e-06, "loss": 0.2679, "step": 17919 }, { "epoch": 20.09, "learning_rate": 6.180281690140846e-06, "loss": 0.4056, "step": 17920 }, { "epoch": 20.09, "learning_rate": 6.1788732394366195e-06, "loss": 0.327, "step": 17921 }, { "epoch": 20.09, "learning_rate": 6.177464788732394e-06, "loss": 0.3816, "step": 17922 }, { "epoch": 20.09, "learning_rate": 6.17605633802817e-06, "loss": 0.3162, "step": 17923 }, { "epoch": 20.09, "learning_rate": 6.174647887323944e-06, "loss": 0.3193, "step": 17924 }, { "epoch": 20.1, "learning_rate": 6.173239436619718e-06, "loss": 0.3572, "step": 17925 }, { "epoch": 20.1, "learning_rate": 6.171830985915493e-06, "loss": 0.2555, "step": 17926 }, { "epoch": 20.1, "learning_rate": 6.170422535211268e-06, "loss": 0.3139, "step": 17927 }, { "epoch": 20.1, "learning_rate": 6.169014084507043e-06, "loss": 0.3414, "step": 17928 }, { "epoch": 20.1, "learning_rate": 6.167605633802817e-06, "loss": 0.3094, "step": 17929 }, { "epoch": 20.1, "learning_rate": 6.166197183098591e-06, "loss": 0.3533, "step": 17930 }, { "epoch": 20.1, "learning_rate": 6.164788732394367e-06, "loss": 0.3529, "step": 17931 }, { "epoch": 20.1, "learning_rate": 6.163380281690141e-06, "loss": 0.3117, "step": 17932 }, { "epoch": 20.1, "learning_rate": 6.161971830985915e-06, "loss": 0.3331, "step": 17933 }, { "epoch": 20.11, "learning_rate": 6.16056338028169e-06, "loss": 0.376, "step": 17934 }, { "epoch": 20.11, "learning_rate": 6.159154929577465e-06, "loss": 0.2655, "step": 17935 }, { "epoch": 20.11, "learning_rate": 6.15774647887324e-06, "loss": 0.2633, "step": 17936 }, { "epoch": 20.11, "learning_rate": 6.156338028169014e-06, "loss": 0.3485, "step": 17937 }, { "epoch": 20.11, "learning_rate": 6.154929577464789e-06, "loss": 0.4253, "step": 17938 }, { "epoch": 20.11, "learning_rate": 6.153521126760564e-06, "loss": 0.277, "step": 17939 }, { "epoch": 20.11, "learning_rate": 6.1521126760563385e-06, "loss": 0.3206, "step": 17940 }, { "epoch": 20.11, "learning_rate": 6.150704225352112e-06, "loss": 0.5864, "step": 17941 }, { "epoch": 20.11, "learning_rate": 6.149295774647888e-06, "loss": 0.4073, "step": 17942 }, { "epoch": 20.12, "learning_rate": 6.147887323943662e-06, "loss": 0.4646, "step": 17943 }, { "epoch": 20.12, "learning_rate": 6.146478873239437e-06, "loss": 0.4039, "step": 17944 }, { "epoch": 20.12, "learning_rate": 6.145070422535211e-06, "loss": 0.3363, "step": 17945 }, { "epoch": 20.12, "learning_rate": 6.143661971830986e-06, "loss": 0.4237, "step": 17946 }, { "epoch": 20.12, "learning_rate": 6.142253521126761e-06, "loss": 0.4361, "step": 17947 }, { "epoch": 20.12, "learning_rate": 6.1408450704225356e-06, "loss": 0.36, "step": 17948 }, { "epoch": 20.12, "learning_rate": 6.139436619718309e-06, "loss": 0.4004, "step": 17949 }, { "epoch": 20.12, "learning_rate": 6.138028169014085e-06, "loss": 0.4434, "step": 17950 }, { "epoch": 20.12, "learning_rate": 6.1366197183098595e-06, "loss": 0.4341, "step": 17951 }, { "epoch": 20.13, "learning_rate": 6.135211267605634e-06, "loss": 0.3012, "step": 17952 }, { "epoch": 20.13, "learning_rate": 6.133802816901408e-06, "loss": 0.3971, "step": 17953 }, { "epoch": 20.13, "learning_rate": 6.132394366197183e-06, "loss": 0.3152, "step": 17954 }, { "epoch": 20.13, "learning_rate": 6.130985915492958e-06, "loss": 0.36, "step": 17955 }, { "epoch": 20.13, "learning_rate": 6.129577464788733e-06, "loss": 0.3713, "step": 17956 }, { "epoch": 20.13, "learning_rate": 6.128169014084507e-06, "loss": 0.3321, "step": 17957 }, { "epoch": 20.13, "learning_rate": 6.126760563380282e-06, "loss": 0.4398, "step": 17958 }, { "epoch": 20.13, "learning_rate": 6.1253521126760566e-06, "loss": 0.3172, "step": 17959 }, { "epoch": 20.13, "learning_rate": 6.123943661971831e-06, "loss": 0.3225, "step": 17960 }, { "epoch": 20.14, "learning_rate": 6.122535211267606e-06, "loss": 0.3152, "step": 17961 }, { "epoch": 20.14, "learning_rate": 6.1211267605633805e-06, "loss": 0.3285, "step": 17962 }, { "epoch": 20.14, "learning_rate": 6.119718309859155e-06, "loss": 0.3735, "step": 17963 }, { "epoch": 20.14, "learning_rate": 6.11830985915493e-06, "loss": 0.3403, "step": 17964 }, { "epoch": 20.14, "learning_rate": 6.116901408450704e-06, "loss": 0.3293, "step": 17965 }, { "epoch": 20.14, "learning_rate": 6.115492957746479e-06, "loss": 0.2828, "step": 17966 }, { "epoch": 20.14, "learning_rate": 6.114084507042254e-06, "loss": 0.3228, "step": 17967 }, { "epoch": 20.14, "learning_rate": 6.112676056338028e-06, "loss": 0.3536, "step": 17968 }, { "epoch": 20.14, "learning_rate": 6.111267605633803e-06, "loss": 0.3547, "step": 17969 }, { "epoch": 20.15, "learning_rate": 6.109859154929578e-06, "loss": 0.3224, "step": 17970 }, { "epoch": 20.15, "learning_rate": 6.108450704225352e-06, "loss": 0.308, "step": 17971 }, { "epoch": 20.15, "learning_rate": 6.107042253521128e-06, "loss": 0.4598, "step": 17972 }, { "epoch": 20.15, "learning_rate": 6.1056338028169015e-06, "loss": 0.3431, "step": 17973 }, { "epoch": 20.15, "learning_rate": 6.104225352112676e-06, "loss": 0.2886, "step": 17974 }, { "epoch": 20.15, "learning_rate": 6.102816901408451e-06, "loss": 0.3856, "step": 17975 }, { "epoch": 20.15, "learning_rate": 6.101408450704226e-06, "loss": 0.3319, "step": 17976 }, { "epoch": 20.15, "learning_rate": 6.1e-06, "loss": 0.2969, "step": 17977 }, { "epoch": 20.15, "learning_rate": 6.098591549295775e-06, "loss": 0.2997, "step": 17978 }, { "epoch": 20.16, "learning_rate": 6.097183098591549e-06, "loss": 0.3055, "step": 17979 }, { "epoch": 20.16, "learning_rate": 6.095774647887325e-06, "loss": 0.3655, "step": 17980 }, { "epoch": 20.16, "learning_rate": 6.094366197183099e-06, "loss": 0.3215, "step": 17981 }, { "epoch": 20.16, "learning_rate": 6.092957746478873e-06, "loss": 0.3498, "step": 17982 }, { "epoch": 20.16, "learning_rate": 6.091549295774648e-06, "loss": 0.2465, "step": 17983 }, { "epoch": 20.16, "learning_rate": 6.090140845070423e-06, "loss": 0.334, "step": 17984 }, { "epoch": 20.16, "learning_rate": 6.088732394366197e-06, "loss": 0.346, "step": 17985 }, { "epoch": 20.16, "learning_rate": 6.087323943661972e-06, "loss": 0.2265, "step": 17986 }, { "epoch": 20.16, "learning_rate": 6.085915492957746e-06, "loss": 0.2061, "step": 17987 }, { "epoch": 20.17, "learning_rate": 6.084507042253522e-06, "loss": 0.3168, "step": 17988 }, { "epoch": 20.17, "learning_rate": 6.083098591549296e-06, "loss": 0.3652, "step": 17989 }, { "epoch": 20.17, "learning_rate": 6.08169014084507e-06, "loss": 0.2626, "step": 17990 }, { "epoch": 20.17, "learning_rate": 6.080281690140846e-06, "loss": 0.4511, "step": 17991 }, { "epoch": 20.17, "learning_rate": 6.0788732394366205e-06, "loss": 0.3589, "step": 17992 }, { "epoch": 20.17, "learning_rate": 6.077464788732394e-06, "loss": 0.3612, "step": 17993 }, { "epoch": 20.17, "learning_rate": 6.076056338028169e-06, "loss": 0.3991, "step": 17994 }, { "epoch": 20.17, "learning_rate": 6.074647887323944e-06, "loss": 0.316, "step": 17995 }, { "epoch": 20.17, "learning_rate": 6.073239436619719e-06, "loss": 0.3662, "step": 17996 }, { "epoch": 20.18, "learning_rate": 6.071830985915493e-06, "loss": 0.3368, "step": 17997 }, { "epoch": 20.18, "learning_rate": 6.0704225352112674e-06, "loss": 0.3871, "step": 17998 }, { "epoch": 20.18, "learning_rate": 6.069014084507043e-06, "loss": 0.3747, "step": 17999 }, { "epoch": 20.18, "learning_rate": 6.0676056338028176e-06, "loss": 0.3575, "step": 18000 }, { "epoch": 20.18, "eval_loss": 0.7367066740989685, "eval_runtime": 818.2808, "eval_samples_per_second": 3.229, "eval_steps_per_second": 0.405, "eval_wer": 0.1332606108687029, "step": 18000 }, { "epoch": 20.18, "learning_rate": 6.066197183098591e-06, "loss": 0.3431, "step": 18001 }, { "epoch": 20.18, "learning_rate": 6.064788732394366e-06, "loss": 0.3922, "step": 18002 }, { "epoch": 20.18, "learning_rate": 6.0633802816901415e-06, "loss": 0.3953, "step": 18003 }, { "epoch": 20.18, "learning_rate": 6.061971830985915e-06, "loss": 0.3736, "step": 18004 }, { "epoch": 20.18, "learning_rate": 6.06056338028169e-06, "loss": 0.3466, "step": 18005 }, { "epoch": 20.19, "learning_rate": 6.0591549295774645e-06, "loss": 0.435, "step": 18006 }, { "epoch": 20.19, "learning_rate": 6.05774647887324e-06, "loss": 0.275, "step": 18007 }, { "epoch": 20.19, "learning_rate": 6.056338028169014e-06, "loss": 0.4863, "step": 18008 }, { "epoch": 20.19, "learning_rate": 6.0549295774647884e-06, "loss": 0.3668, "step": 18009 }, { "epoch": 20.19, "learning_rate": 6.053521126760564e-06, "loss": 0.3574, "step": 18010 }, { "epoch": 20.19, "learning_rate": 6.0521126760563386e-06, "loss": 0.3005, "step": 18011 }, { "epoch": 20.19, "learning_rate": 6.050704225352112e-06, "loss": 0.4314, "step": 18012 }, { "epoch": 20.19, "learning_rate": 6.049295774647887e-06, "loss": 0.3036, "step": 18013 }, { "epoch": 20.2, "learning_rate": 6.0478873239436625e-06, "loss": 0.3062, "step": 18014 }, { "epoch": 20.2, "learning_rate": 6.046478873239437e-06, "loss": 0.3145, "step": 18015 }, { "epoch": 20.2, "learning_rate": 6.045070422535211e-06, "loss": 0.3429, "step": 18016 }, { "epoch": 20.2, "learning_rate": 6.0436619718309855e-06, "loss": 0.3381, "step": 18017 }, { "epoch": 20.2, "learning_rate": 6.042253521126761e-06, "loss": 0.3472, "step": 18018 }, { "epoch": 20.2, "learning_rate": 6.040845070422536e-06, "loss": 0.2978, "step": 18019 }, { "epoch": 20.2, "learning_rate": 6.0394366197183095e-06, "loss": 0.438, "step": 18020 }, { "epoch": 20.2, "learning_rate": 6.038028169014084e-06, "loss": 0.2925, "step": 18021 }, { "epoch": 20.2, "learning_rate": 6.03661971830986e-06, "loss": 0.3399, "step": 18022 }, { "epoch": 20.21, "learning_rate": 6.035211267605634e-06, "loss": 0.3823, "step": 18023 }, { "epoch": 20.21, "learning_rate": 6.033802816901408e-06, "loss": 0.3479, "step": 18024 }, { "epoch": 20.21, "learning_rate": 6.0323943661971835e-06, "loss": 0.3703, "step": 18025 }, { "epoch": 20.21, "learning_rate": 6.030985915492958e-06, "loss": 0.3337, "step": 18026 }, { "epoch": 20.21, "learning_rate": 6.029577464788733e-06, "loss": 0.3205, "step": 18027 }, { "epoch": 20.21, "learning_rate": 6.0281690140845066e-06, "loss": 0.3381, "step": 18028 }, { "epoch": 20.21, "learning_rate": 6.026760563380282e-06, "loss": 0.3094, "step": 18029 }, { "epoch": 20.21, "learning_rate": 6.025352112676057e-06, "loss": 0.3389, "step": 18030 }, { "epoch": 20.21, "learning_rate": 6.023943661971831e-06, "loss": 0.3613, "step": 18031 }, { "epoch": 20.22, "learning_rate": 6.022535211267605e-06, "loss": 0.301, "step": 18032 }, { "epoch": 20.22, "learning_rate": 6.021126760563381e-06, "loss": 0.5016, "step": 18033 }, { "epoch": 20.22, "learning_rate": 6.019718309859155e-06, "loss": 0.2699, "step": 18034 }, { "epoch": 20.22, "learning_rate": 6.01830985915493e-06, "loss": 0.2075, "step": 18035 }, { "epoch": 20.22, "learning_rate": 6.016901408450704e-06, "loss": 0.2865, "step": 18036 }, { "epoch": 20.22, "learning_rate": 6.015492957746479e-06, "loss": 0.2706, "step": 18037 }, { "epoch": 20.22, "learning_rate": 6.014084507042254e-06, "loss": 0.2378, "step": 18038 }, { "epoch": 20.22, "learning_rate": 6.012676056338028e-06, "loss": 0.2816, "step": 18039 }, { "epoch": 20.22, "learning_rate": 6.011267605633802e-06, "loss": 0.3932, "step": 18040 }, { "epoch": 20.23, "learning_rate": 6.009859154929578e-06, "loss": 0.4696, "step": 18041 }, { "epoch": 20.23, "learning_rate": 6.008450704225352e-06, "loss": 0.5145, "step": 18042 }, { "epoch": 20.23, "learning_rate": 6.007042253521127e-06, "loss": 0.3117, "step": 18043 }, { "epoch": 20.23, "learning_rate": 6.005633802816902e-06, "loss": 0.361, "step": 18044 }, { "epoch": 20.23, "learning_rate": 6.004225352112676e-06, "loss": 0.4047, "step": 18045 }, { "epoch": 20.23, "learning_rate": 6.002816901408451e-06, "loss": 0.3321, "step": 18046 }, { "epoch": 20.23, "learning_rate": 6.0014084507042255e-06, "loss": 0.3647, "step": 18047 }, { "epoch": 20.23, "learning_rate": 6e-06, "loss": 0.3994, "step": 18048 }, { "epoch": 20.23, "learning_rate": 5.998591549295775e-06, "loss": 0.3949, "step": 18049 }, { "epoch": 20.24, "learning_rate": 5.9971830985915494e-06, "loss": 0.4278, "step": 18050 }, { "epoch": 20.24, "learning_rate": 5.995774647887324e-06, "loss": 0.3937, "step": 18051 }, { "epoch": 20.24, "learning_rate": 5.994366197183099e-06, "loss": 0.2807, "step": 18052 }, { "epoch": 20.24, "learning_rate": 5.992957746478873e-06, "loss": 0.359, "step": 18053 }, { "epoch": 20.24, "learning_rate": 5.991549295774648e-06, "loss": 0.3762, "step": 18054 }, { "epoch": 20.24, "learning_rate": 5.990140845070423e-06, "loss": 0.2729, "step": 18055 }, { "epoch": 20.24, "learning_rate": 5.988732394366197e-06, "loss": 0.3282, "step": 18056 }, { "epoch": 20.24, "learning_rate": 5.987323943661972e-06, "loss": 0.3736, "step": 18057 }, { "epoch": 20.24, "learning_rate": 5.9859154929577465e-06, "loss": 0.4402, "step": 18058 }, { "epoch": 20.25, "learning_rate": 5.984507042253522e-06, "loss": 0.3002, "step": 18059 }, { "epoch": 20.25, "learning_rate": 5.983098591549296e-06, "loss": 0.3614, "step": 18060 }, { "epoch": 20.25, "learning_rate": 5.9816901408450704e-06, "loss": 0.3138, "step": 18061 }, { "epoch": 20.25, "learning_rate": 5.980281690140845e-06, "loss": 0.3368, "step": 18062 }, { "epoch": 20.25, "learning_rate": 5.9788732394366206e-06, "loss": 0.4206, "step": 18063 }, { "epoch": 20.25, "learning_rate": 5.977464788732394e-06, "loss": 0.3008, "step": 18064 }, { "epoch": 20.25, "learning_rate": 5.976056338028169e-06, "loss": 0.2617, "step": 18065 }, { "epoch": 20.25, "learning_rate": 5.974647887323944e-06, "loss": 0.3505, "step": 18066 }, { "epoch": 20.25, "learning_rate": 5.973239436619719e-06, "loss": 0.3597, "step": 18067 }, { "epoch": 20.26, "learning_rate": 5.971830985915493e-06, "loss": 0.2978, "step": 18068 }, { "epoch": 20.26, "learning_rate": 5.9704225352112675e-06, "loss": 0.3073, "step": 18069 }, { "epoch": 20.26, "learning_rate": 5.969014084507042e-06, "loss": 0.3045, "step": 18070 }, { "epoch": 20.26, "learning_rate": 5.967605633802818e-06, "loss": 0.3568, "step": 18071 }, { "epoch": 20.26, "learning_rate": 5.9661971830985915e-06, "loss": 0.2456, "step": 18072 }, { "epoch": 20.26, "learning_rate": 5.964788732394366e-06, "loss": 0.3324, "step": 18073 }, { "epoch": 20.26, "learning_rate": 5.963380281690141e-06, "loss": 0.3023, "step": 18074 }, { "epoch": 20.26, "learning_rate": 5.961971830985916e-06, "loss": 0.3836, "step": 18075 }, { "epoch": 20.26, "learning_rate": 5.96056338028169e-06, "loss": 0.3289, "step": 18076 }, { "epoch": 20.27, "learning_rate": 5.959154929577465e-06, "loss": 0.3817, "step": 18077 }, { "epoch": 20.27, "learning_rate": 5.95774647887324e-06, "loss": 0.3639, "step": 18078 }, { "epoch": 20.27, "learning_rate": 5.956338028169015e-06, "loss": 0.3663, "step": 18079 }, { "epoch": 20.27, "learning_rate": 5.9549295774647886e-06, "loss": 0.2982, "step": 18080 }, { "epoch": 20.27, "learning_rate": 5.953521126760563e-06, "loss": 0.318, "step": 18081 }, { "epoch": 20.27, "learning_rate": 5.952112676056339e-06, "loss": 0.3414, "step": 18082 }, { "epoch": 20.27, "learning_rate": 5.950704225352113e-06, "loss": 0.2932, "step": 18083 }, { "epoch": 20.27, "learning_rate": 5.949295774647887e-06, "loss": 0.4228, "step": 18084 }, { "epoch": 20.27, "learning_rate": 5.947887323943662e-06, "loss": 0.4311, "step": 18085 }, { "epoch": 20.28, "learning_rate": 5.946478873239437e-06, "loss": 0.2754, "step": 18086 }, { "epoch": 20.28, "learning_rate": 5.945070422535212e-06, "loss": 0.2868, "step": 18087 }, { "epoch": 20.28, "learning_rate": 5.943661971830986e-06, "loss": 0.2922, "step": 18088 }, { "epoch": 20.28, "learning_rate": 5.94225352112676e-06, "loss": 0.312, "step": 18089 }, { "epoch": 20.28, "learning_rate": 5.940845070422536e-06, "loss": 0.3682, "step": 18090 }, { "epoch": 20.28, "learning_rate": 5.93943661971831e-06, "loss": 0.4573, "step": 18091 }, { "epoch": 20.28, "learning_rate": 5.938028169014084e-06, "loss": 0.4084, "step": 18092 }, { "epoch": 20.28, "learning_rate": 5.936619718309859e-06, "loss": 0.4946, "step": 18093 }, { "epoch": 20.28, "learning_rate": 5.935211267605634e-06, "loss": 0.3175, "step": 18094 }, { "epoch": 20.29, "learning_rate": 5.933802816901409e-06, "loss": 0.4379, "step": 18095 }, { "epoch": 20.29, "learning_rate": 5.932394366197183e-06, "loss": 0.3558, "step": 18096 }, { "epoch": 20.29, "learning_rate": 5.930985915492958e-06, "loss": 0.3454, "step": 18097 }, { "epoch": 20.29, "learning_rate": 5.929577464788733e-06, "loss": 0.454, "step": 18098 }, { "epoch": 20.29, "learning_rate": 5.9281690140845075e-06, "loss": 0.4032, "step": 18099 }, { "epoch": 20.29, "learning_rate": 5.926760563380281e-06, "loss": 0.4249, "step": 18100 }, { "epoch": 20.29, "learning_rate": 5.925352112676057e-06, "loss": 0.3718, "step": 18101 }, { "epoch": 20.29, "learning_rate": 5.9239436619718314e-06, "loss": 0.3949, "step": 18102 }, { "epoch": 20.29, "learning_rate": 5.922535211267606e-06, "loss": 0.3381, "step": 18103 }, { "epoch": 20.3, "learning_rate": 5.92112676056338e-06, "loss": 0.4321, "step": 18104 }, { "epoch": 20.3, "learning_rate": 5.919718309859155e-06, "loss": 0.3717, "step": 18105 }, { "epoch": 20.3, "learning_rate": 5.91830985915493e-06, "loss": 0.3559, "step": 18106 }, { "epoch": 20.3, "learning_rate": 5.916901408450705e-06, "loss": 0.3699, "step": 18107 }, { "epoch": 20.3, "learning_rate": 5.915492957746478e-06, "loss": 0.2903, "step": 18108 }, { "epoch": 20.3, "learning_rate": 5.914084507042254e-06, "loss": 0.3345, "step": 18109 }, { "epoch": 20.3, "learning_rate": 5.9126760563380285e-06, "loss": 0.3699, "step": 18110 }, { "epoch": 20.3, "learning_rate": 5.911267605633803e-06, "loss": 0.3677, "step": 18111 }, { "epoch": 20.3, "learning_rate": 5.909859154929578e-06, "loss": 0.3301, "step": 18112 }, { "epoch": 20.31, "learning_rate": 5.9084507042253524e-06, "loss": 0.3981, "step": 18113 }, { "epoch": 20.31, "learning_rate": 5.907042253521127e-06, "loss": 0.284, "step": 18114 }, { "epoch": 20.31, "learning_rate": 5.905633802816902e-06, "loss": 0.3782, "step": 18115 }, { "epoch": 20.31, "learning_rate": 5.904225352112676e-06, "loss": 0.5132, "step": 18116 }, { "epoch": 20.31, "learning_rate": 5.902816901408451e-06, "loss": 0.3675, "step": 18117 }, { "epoch": 20.31, "learning_rate": 5.901408450704226e-06, "loss": 0.3372, "step": 18118 }, { "epoch": 20.31, "learning_rate": 5.899999999999999e-06, "loss": 0.3966, "step": 18119 }, { "epoch": 20.31, "learning_rate": 5.898591549295775e-06, "loss": 0.3164, "step": 18120 }, { "epoch": 20.32, "learning_rate": 5.8971830985915495e-06, "loss": 0.3612, "step": 18121 }, { "epoch": 20.32, "learning_rate": 5.895774647887324e-06, "loss": 0.2483, "step": 18122 }, { "epoch": 20.32, "learning_rate": 5.894366197183098e-06, "loss": 0.3057, "step": 18123 }, { "epoch": 20.32, "learning_rate": 5.8929577464788734e-06, "loss": 0.3195, "step": 18124 }, { "epoch": 20.32, "learning_rate": 5.891549295774648e-06, "loss": 0.3293, "step": 18125 }, { "epoch": 20.32, "learning_rate": 5.890140845070423e-06, "loss": 0.392, "step": 18126 }, { "epoch": 20.32, "learning_rate": 5.8887323943661965e-06, "loss": 0.4319, "step": 18127 }, { "epoch": 20.32, "learning_rate": 5.887323943661972e-06, "loss": 0.2736, "step": 18128 }, { "epoch": 20.32, "learning_rate": 5.885915492957747e-06, "loss": 0.2654, "step": 18129 }, { "epoch": 20.33, "learning_rate": 5.884507042253521e-06, "loss": 0.3431, "step": 18130 }, { "epoch": 20.33, "learning_rate": 5.883098591549296e-06, "loss": 0.3951, "step": 18131 }, { "epoch": 20.33, "learning_rate": 5.8816901408450706e-06, "loss": 0.2698, "step": 18132 }, { "epoch": 20.33, "learning_rate": 5.880281690140845e-06, "loss": 0.3123, "step": 18133 }, { "epoch": 20.33, "learning_rate": 5.87887323943662e-06, "loss": 0.2105, "step": 18134 }, { "epoch": 20.33, "learning_rate": 5.8774647887323945e-06, "loss": 0.3254, "step": 18135 }, { "epoch": 20.33, "learning_rate": 5.876056338028169e-06, "loss": 0.2702, "step": 18136 }, { "epoch": 20.33, "learning_rate": 5.874647887323944e-06, "loss": 0.4213, "step": 18137 }, { "epoch": 20.33, "learning_rate": 5.873239436619718e-06, "loss": 0.2796, "step": 18138 }, { "epoch": 20.34, "learning_rate": 5.871830985915493e-06, "loss": 0.3076, "step": 18139 }, { "epoch": 20.34, "learning_rate": 5.870422535211268e-06, "loss": 0.2642, "step": 18140 }, { "epoch": 20.34, "learning_rate": 5.869014084507042e-06, "loss": 0.4088, "step": 18141 }, { "epoch": 20.34, "learning_rate": 5.867605633802817e-06, "loss": 0.3724, "step": 18142 }, { "epoch": 20.34, "learning_rate": 5.8661971830985916e-06, "loss": 0.3642, "step": 18143 }, { "epoch": 20.34, "learning_rate": 5.864788732394366e-06, "loss": 0.3886, "step": 18144 }, { "epoch": 20.34, "learning_rate": 5.863380281690141e-06, "loss": 0.3486, "step": 18145 }, { "epoch": 20.34, "learning_rate": 5.861971830985916e-06, "loss": 0.336, "step": 18146 }, { "epoch": 20.34, "learning_rate": 5.86056338028169e-06, "loss": 0.3449, "step": 18147 }, { "epoch": 20.35, "learning_rate": 5.859154929577465e-06, "loss": 0.3598, "step": 18148 }, { "epoch": 20.35, "learning_rate": 5.857746478873239e-06, "loss": 0.2953, "step": 18149 }, { "epoch": 20.35, "learning_rate": 5.856338028169015e-06, "loss": 0.3324, "step": 18150 }, { "epoch": 20.35, "learning_rate": 5.854929577464789e-06, "loss": 0.3604, "step": 18151 }, { "epoch": 20.35, "learning_rate": 5.853521126760563e-06, "loss": 0.3139, "step": 18152 }, { "epoch": 20.35, "learning_rate": 5.852112676056338e-06, "loss": 0.3102, "step": 18153 }, { "epoch": 20.35, "learning_rate": 5.850704225352113e-06, "loss": 0.329, "step": 18154 }, { "epoch": 20.35, "learning_rate": 5.849295774647887e-06, "loss": 0.3429, "step": 18155 }, { "epoch": 20.35, "learning_rate": 5.847887323943662e-06, "loss": 0.3467, "step": 18156 }, { "epoch": 20.36, "learning_rate": 5.8464788732394365e-06, "loss": 0.3628, "step": 18157 }, { "epoch": 20.36, "learning_rate": 5.845070422535212e-06, "loss": 0.2062, "step": 18158 }, { "epoch": 20.36, "learning_rate": 5.843661971830986e-06, "loss": 0.377, "step": 18159 }, { "epoch": 20.36, "learning_rate": 5.84225352112676e-06, "loss": 0.3294, "step": 18160 }, { "epoch": 20.36, "learning_rate": 5.840845070422535e-06, "loss": 0.3826, "step": 18161 }, { "epoch": 20.36, "learning_rate": 5.8394366197183105e-06, "loss": 0.2772, "step": 18162 }, { "epoch": 20.36, "learning_rate": 5.838028169014084e-06, "loss": 0.2974, "step": 18163 }, { "epoch": 20.36, "learning_rate": 5.836619718309859e-06, "loss": 0.3767, "step": 18164 }, { "epoch": 20.36, "learning_rate": 5.8352112676056344e-06, "loss": 0.4185, "step": 18165 }, { "epoch": 20.37, "learning_rate": 5.833802816901409e-06, "loss": 0.3386, "step": 18166 }, { "epoch": 20.37, "learning_rate": 5.832394366197183e-06, "loss": 0.2847, "step": 18167 }, { "epoch": 20.37, "learning_rate": 5.8309859154929575e-06, "loss": 0.2372, "step": 18168 }, { "epoch": 20.37, "learning_rate": 5.829577464788733e-06, "loss": 0.294, "step": 18169 }, { "epoch": 20.37, "learning_rate": 5.828169014084508e-06, "loss": 0.3977, "step": 18170 }, { "epoch": 20.37, "learning_rate": 5.826760563380281e-06, "loss": 0.4097, "step": 18171 }, { "epoch": 20.37, "learning_rate": 5.825352112676056e-06, "loss": 0.2431, "step": 18172 }, { "epoch": 20.37, "learning_rate": 5.8239436619718315e-06, "loss": 0.3726, "step": 18173 }, { "epoch": 20.37, "learning_rate": 5.822535211267606e-06, "loss": 0.2762, "step": 18174 }, { "epoch": 20.38, "learning_rate": 5.82112676056338e-06, "loss": 0.3552, "step": 18175 }, { "epoch": 20.38, "learning_rate": 5.819718309859155e-06, "loss": 0.2532, "step": 18176 }, { "epoch": 20.38, "learning_rate": 5.81830985915493e-06, "loss": 0.3854, "step": 18177 }, { "epoch": 20.38, "learning_rate": 5.816901408450705e-06, "loss": 0.3216, "step": 18178 }, { "epoch": 20.38, "learning_rate": 5.8154929577464785e-06, "loss": 0.307, "step": 18179 }, { "epoch": 20.38, "learning_rate": 5.814084507042254e-06, "loss": 0.2963, "step": 18180 }, { "epoch": 20.38, "learning_rate": 5.812676056338029e-06, "loss": 0.3638, "step": 18181 }, { "epoch": 20.38, "learning_rate": 5.811267605633803e-06, "loss": 0.4207, "step": 18182 }, { "epoch": 20.38, "learning_rate": 5.809859154929577e-06, "loss": 0.3306, "step": 18183 }, { "epoch": 20.39, "learning_rate": 5.8084507042253525e-06, "loss": 0.3553, "step": 18184 }, { "epoch": 20.39, "learning_rate": 5.807042253521127e-06, "loss": 0.3285, "step": 18185 }, { "epoch": 20.39, "learning_rate": 5.805633802816902e-06, "loss": 0.3251, "step": 18186 }, { "epoch": 20.39, "learning_rate": 5.804225352112676e-06, "loss": 0.2646, "step": 18187 }, { "epoch": 20.39, "learning_rate": 5.802816901408451e-06, "loss": 0.2216, "step": 18188 }, { "epoch": 20.39, "learning_rate": 5.801408450704226e-06, "loss": 0.3228, "step": 18189 }, { "epoch": 20.39, "learning_rate": 5.8e-06, "loss": 0.1352, "step": 18190 }, { "epoch": 20.39, "learning_rate": 5.798591549295774e-06, "loss": 0.4745, "step": 18191 }, { "epoch": 20.39, "learning_rate": 5.79718309859155e-06, "loss": 0.4816, "step": 18192 }, { "epoch": 20.4, "learning_rate": 5.795774647887324e-06, "loss": 0.4081, "step": 18193 }, { "epoch": 20.4, "learning_rate": 5.794366197183099e-06, "loss": 0.4118, "step": 18194 }, { "epoch": 20.4, "learning_rate": 5.792957746478873e-06, "loss": 0.3211, "step": 18195 }, { "epoch": 20.4, "learning_rate": 5.791549295774648e-06, "loss": 0.3861, "step": 18196 }, { "epoch": 20.4, "learning_rate": 5.790140845070423e-06, "loss": 0.3378, "step": 18197 }, { "epoch": 20.4, "learning_rate": 5.7887323943661975e-06, "loss": 0.4011, "step": 18198 }, { "epoch": 20.4, "learning_rate": 5.787323943661972e-06, "loss": 0.3831, "step": 18199 }, { "epoch": 20.4, "learning_rate": 5.785915492957747e-06, "loss": 0.4197, "step": 18200 }, { "epoch": 20.4, "learning_rate": 5.784507042253521e-06, "loss": 0.3139, "step": 18201 }, { "epoch": 20.41, "learning_rate": 5.783098591549296e-06, "loss": 0.4131, "step": 18202 }, { "epoch": 20.41, "learning_rate": 5.781690140845071e-06, "loss": 0.4054, "step": 18203 }, { "epoch": 20.41, "learning_rate": 5.780281690140845e-06, "loss": 0.3566, "step": 18204 }, { "epoch": 20.41, "learning_rate": 5.77887323943662e-06, "loss": 0.28, "step": 18205 }, { "epoch": 20.41, "learning_rate": 5.7774647887323946e-06, "loss": 0.375, "step": 18206 }, { "epoch": 20.41, "learning_rate": 5.776056338028169e-06, "loss": 0.4032, "step": 18207 }, { "epoch": 20.41, "learning_rate": 5.774647887323944e-06, "loss": 0.3072, "step": 18208 }, { "epoch": 20.41, "learning_rate": 5.7732394366197185e-06, "loss": 0.3744, "step": 18209 }, { "epoch": 20.41, "learning_rate": 5.771830985915493e-06, "loss": 0.3502, "step": 18210 }, { "epoch": 20.42, "learning_rate": 5.770422535211268e-06, "loss": 0.4216, "step": 18211 }, { "epoch": 20.42, "learning_rate": 5.769014084507042e-06, "loss": 0.3632, "step": 18212 }, { "epoch": 20.42, "learning_rate": 5.767605633802817e-06, "loss": 0.3058, "step": 18213 }, { "epoch": 20.42, "learning_rate": 5.766197183098592e-06, "loss": 0.3275, "step": 18214 }, { "epoch": 20.42, "learning_rate": 5.764788732394366e-06, "loss": 0.3399, "step": 18215 }, { "epoch": 20.42, "learning_rate": 5.763380281690141e-06, "loss": 0.3174, "step": 18216 }, { "epoch": 20.42, "learning_rate": 5.761971830985916e-06, "loss": 0.3602, "step": 18217 }, { "epoch": 20.42, "learning_rate": 5.760563380281691e-06, "loss": 0.2885, "step": 18218 }, { "epoch": 20.42, "learning_rate": 5.759154929577465e-06, "loss": 0.2984, "step": 18219 }, { "epoch": 20.43, "learning_rate": 5.7577464788732395e-06, "loss": 0.3159, "step": 18220 }, { "epoch": 20.43, "learning_rate": 5.756338028169014e-06, "loss": 0.4067, "step": 18221 }, { "epoch": 20.43, "learning_rate": 5.75492957746479e-06, "loss": 0.2543, "step": 18222 }, { "epoch": 20.43, "learning_rate": 5.753521126760563e-06, "loss": 0.3024, "step": 18223 }, { "epoch": 20.43, "learning_rate": 5.752112676056338e-06, "loss": 0.3422, "step": 18224 }, { "epoch": 20.43, "learning_rate": 5.750704225352113e-06, "loss": 0.3256, "step": 18225 }, { "epoch": 20.43, "learning_rate": 5.749295774647888e-06, "loss": 0.3998, "step": 18226 }, { "epoch": 20.43, "learning_rate": 5.747887323943662e-06, "loss": 0.3342, "step": 18227 }, { "epoch": 20.43, "learning_rate": 5.746478873239437e-06, "loss": 0.304, "step": 18228 }, { "epoch": 20.44, "learning_rate": 5.745070422535211e-06, "loss": 0.2597, "step": 18229 }, { "epoch": 20.44, "learning_rate": 5.743661971830986e-06, "loss": 0.3508, "step": 18230 }, { "epoch": 20.44, "learning_rate": 5.7422535211267605e-06, "loss": 0.3957, "step": 18231 }, { "epoch": 20.44, "learning_rate": 5.740845070422535e-06, "loss": 0.4258, "step": 18232 }, { "epoch": 20.44, "learning_rate": 5.739436619718311e-06, "loss": 0.3038, "step": 18233 }, { "epoch": 20.44, "learning_rate": 5.738028169014084e-06, "loss": 0.2323, "step": 18234 }, { "epoch": 20.44, "learning_rate": 5.736619718309859e-06, "loss": 0.2851, "step": 18235 }, { "epoch": 20.44, "learning_rate": 5.735211267605634e-06, "loss": 0.2841, "step": 18236 }, { "epoch": 20.45, "learning_rate": 5.733802816901409e-06, "loss": 0.4034, "step": 18237 }, { "epoch": 20.45, "learning_rate": 5.732394366197183e-06, "loss": 0.2446, "step": 18238 }, { "epoch": 20.45, "learning_rate": 5.730985915492958e-06, "loss": 0.2167, "step": 18239 }, { "epoch": 20.45, "learning_rate": 5.729577464788732e-06, "loss": 0.1468, "step": 18240 }, { "epoch": 20.45, "learning_rate": 5.728169014084508e-06, "loss": 0.4169, "step": 18241 }, { "epoch": 20.45, "learning_rate": 5.7267605633802815e-06, "loss": 0.4868, "step": 18242 }, { "epoch": 20.45, "learning_rate": 5.725352112676056e-06, "loss": 0.4346, "step": 18243 }, { "epoch": 20.45, "learning_rate": 5.723943661971831e-06, "loss": 0.264, "step": 18244 }, { "epoch": 20.45, "learning_rate": 5.722535211267606e-06, "loss": 0.4428, "step": 18245 }, { "epoch": 20.46, "learning_rate": 5.72112676056338e-06, "loss": 0.4322, "step": 18246 }, { "epoch": 20.46, "learning_rate": 5.719718309859155e-06, "loss": 0.5125, "step": 18247 }, { "epoch": 20.46, "learning_rate": 5.718309859154929e-06, "loss": 0.3307, "step": 18248 }, { "epoch": 20.46, "learning_rate": 5.716901408450705e-06, "loss": 0.3331, "step": 18249 }, { "epoch": 20.46, "learning_rate": 5.715492957746479e-06, "loss": 0.3958, "step": 18250 }, { "epoch": 20.46, "learning_rate": 5.714084507042253e-06, "loss": 0.3101, "step": 18251 }, { "epoch": 20.46, "learning_rate": 5.712676056338029e-06, "loss": 0.3671, "step": 18252 }, { "epoch": 20.46, "learning_rate": 5.711267605633803e-06, "loss": 0.3942, "step": 18253 }, { "epoch": 20.46, "learning_rate": 5.709859154929577e-06, "loss": 0.2853, "step": 18254 }, { "epoch": 20.47, "learning_rate": 5.708450704225352e-06, "loss": 0.366, "step": 18255 }, { "epoch": 20.47, "learning_rate": 5.707042253521127e-06, "loss": 0.4536, "step": 18256 }, { "epoch": 20.47, "learning_rate": 5.705633802816902e-06, "loss": 0.3742, "step": 18257 }, { "epoch": 20.47, "learning_rate": 5.704225352112676e-06, "loss": 0.3423, "step": 18258 }, { "epoch": 20.47, "learning_rate": 5.70281690140845e-06, "loss": 0.3787, "step": 18259 }, { "epoch": 20.47, "learning_rate": 5.701408450704226e-06, "loss": 0.3103, "step": 18260 }, { "epoch": 20.47, "learning_rate": 5.7000000000000005e-06, "loss": 0.389, "step": 18261 }, { "epoch": 20.47, "learning_rate": 5.698591549295774e-06, "loss": 0.3637, "step": 18262 }, { "epoch": 20.47, "learning_rate": 5.697183098591549e-06, "loss": 0.3206, "step": 18263 }, { "epoch": 20.48, "learning_rate": 5.695774647887324e-06, "loss": 0.3178, "step": 18264 }, { "epoch": 20.48, "learning_rate": 5.694366197183099e-06, "loss": 0.3673, "step": 18265 }, { "epoch": 20.48, "learning_rate": 5.692957746478873e-06, "loss": 0.3319, "step": 18266 }, { "epoch": 20.48, "learning_rate": 5.691549295774648e-06, "loss": 0.3029, "step": 18267 }, { "epoch": 20.48, "learning_rate": 5.690140845070423e-06, "loss": 0.3689, "step": 18268 }, { "epoch": 20.48, "learning_rate": 5.688732394366198e-06, "loss": 0.3824, "step": 18269 }, { "epoch": 20.48, "learning_rate": 5.687323943661971e-06, "loss": 0.3502, "step": 18270 }, { "epoch": 20.48, "learning_rate": 5.685915492957747e-06, "loss": 0.3244, "step": 18271 }, { "epoch": 20.48, "learning_rate": 5.6845070422535215e-06, "loss": 0.2493, "step": 18272 }, { "epoch": 20.49, "learning_rate": 5.683098591549296e-06, "loss": 0.3765, "step": 18273 }, { "epoch": 20.49, "learning_rate": 5.68169014084507e-06, "loss": 0.2989, "step": 18274 }, { "epoch": 20.49, "learning_rate": 5.680281690140845e-06, "loss": 0.2839, "step": 18275 }, { "epoch": 20.49, "learning_rate": 5.67887323943662e-06, "loss": 0.3891, "step": 18276 }, { "epoch": 20.49, "learning_rate": 5.677464788732395e-06, "loss": 0.2995, "step": 18277 }, { "epoch": 20.49, "learning_rate": 5.6760563380281685e-06, "loss": 0.3233, "step": 18278 }, { "epoch": 20.49, "learning_rate": 5.674647887323944e-06, "loss": 0.2392, "step": 18279 }, { "epoch": 20.49, "learning_rate": 5.673239436619719e-06, "loss": 0.3103, "step": 18280 }, { "epoch": 20.49, "learning_rate": 5.671830985915493e-06, "loss": 0.2708, "step": 18281 }, { "epoch": 20.5, "learning_rate": 5.670422535211267e-06, "loss": 0.3171, "step": 18282 }, { "epoch": 20.5, "learning_rate": 5.6690140845070425e-06, "loss": 0.2964, "step": 18283 }, { "epoch": 20.5, "learning_rate": 5.667605633802817e-06, "loss": 0.2459, "step": 18284 }, { "epoch": 20.5, "learning_rate": 5.666197183098592e-06, "loss": 0.3992, "step": 18285 }, { "epoch": 20.5, "learning_rate": 5.664788732394366e-06, "loss": 0.2649, "step": 18286 }, { "epoch": 20.5, "learning_rate": 5.663380281690141e-06, "loss": 0.314, "step": 18287 }, { "epoch": 20.5, "learning_rate": 5.661971830985916e-06, "loss": 0.3441, "step": 18288 }, { "epoch": 20.5, "learning_rate": 5.66056338028169e-06, "loss": 0.2575, "step": 18289 }, { "epoch": 20.5, "learning_rate": 5.659154929577465e-06, "loss": 0.2694, "step": 18290 }, { "epoch": 20.51, "learning_rate": 5.65774647887324e-06, "loss": 0.4805, "step": 18291 }, { "epoch": 20.51, "learning_rate": 5.656338028169014e-06, "loss": 0.4901, "step": 18292 }, { "epoch": 20.51, "learning_rate": 5.654929577464789e-06, "loss": 0.4288, "step": 18293 }, { "epoch": 20.51, "learning_rate": 5.6535211267605635e-06, "loss": 0.4185, "step": 18294 }, { "epoch": 20.51, "learning_rate": 5.652112676056338e-06, "loss": 0.4438, "step": 18295 }, { "epoch": 20.51, "learning_rate": 5.650704225352113e-06, "loss": 0.3782, "step": 18296 }, { "epoch": 20.51, "learning_rate": 5.6492957746478874e-06, "loss": 0.363, "step": 18297 }, { "epoch": 20.51, "learning_rate": 5.647887323943662e-06, "loss": 0.4429, "step": 18298 }, { "epoch": 20.51, "learning_rate": 5.646478873239437e-06, "loss": 0.357, "step": 18299 }, { "epoch": 20.52, "learning_rate": 5.645070422535211e-06, "loss": 0.3185, "step": 18300 }, { "epoch": 20.52, "learning_rate": 5.643661971830986e-06, "loss": 0.4257, "step": 18301 }, { "epoch": 20.52, "learning_rate": 5.642253521126761e-06, "loss": 0.4849, "step": 18302 }, { "epoch": 20.52, "learning_rate": 5.640845070422535e-06, "loss": 0.3383, "step": 18303 }, { "epoch": 20.52, "learning_rate": 5.63943661971831e-06, "loss": 0.3993, "step": 18304 }, { "epoch": 20.52, "learning_rate": 5.638028169014085e-06, "loss": 0.316, "step": 18305 }, { "epoch": 20.52, "learning_rate": 5.636619718309859e-06, "loss": 0.3206, "step": 18306 }, { "epoch": 20.52, "learning_rate": 5.635211267605634e-06, "loss": 0.3753, "step": 18307 }, { "epoch": 20.52, "learning_rate": 5.6338028169014084e-06, "loss": 0.3481, "step": 18308 }, { "epoch": 20.53, "learning_rate": 5.632394366197184e-06, "loss": 0.3114, "step": 18309 }, { "epoch": 20.53, "learning_rate": 5.630985915492958e-06, "loss": 0.3757, "step": 18310 }, { "epoch": 20.53, "learning_rate": 5.629577464788732e-06, "loss": 0.366, "step": 18311 }, { "epoch": 20.53, "learning_rate": 5.628169014084507e-06, "loss": 0.3064, "step": 18312 }, { "epoch": 20.53, "learning_rate": 5.6267605633802825e-06, "loss": 0.3006, "step": 18313 }, { "epoch": 20.53, "learning_rate": 5.625352112676056e-06, "loss": 0.321, "step": 18314 }, { "epoch": 20.53, "learning_rate": 5.623943661971831e-06, "loss": 0.2937, "step": 18315 }, { "epoch": 20.53, "learning_rate": 5.6225352112676055e-06, "loss": 0.3053, "step": 18316 }, { "epoch": 20.53, "learning_rate": 5.621126760563381e-06, "loss": 0.3268, "step": 18317 }, { "epoch": 20.54, "learning_rate": 5.619718309859155e-06, "loss": 0.3385, "step": 18318 }, { "epoch": 20.54, "learning_rate": 5.6183098591549295e-06, "loss": 0.2717, "step": 18319 }, { "epoch": 20.54, "learning_rate": 5.616901408450705e-06, "loss": 0.2981, "step": 18320 }, { "epoch": 20.54, "learning_rate": 5.61549295774648e-06, "loss": 0.2715, "step": 18321 }, { "epoch": 20.54, "learning_rate": 5.614084507042253e-06, "loss": 0.2731, "step": 18322 }, { "epoch": 20.54, "learning_rate": 5.612676056338028e-06, "loss": 0.3722, "step": 18323 }, { "epoch": 20.54, "learning_rate": 5.6112676056338035e-06, "loss": 0.396, "step": 18324 }, { "epoch": 20.54, "learning_rate": 5.609859154929578e-06, "loss": 0.2928, "step": 18325 }, { "epoch": 20.54, "learning_rate": 5.608450704225352e-06, "loss": 0.4665, "step": 18326 }, { "epoch": 20.55, "learning_rate": 5.6070422535211266e-06, "loss": 0.2712, "step": 18327 }, { "epoch": 20.55, "learning_rate": 5.605633802816902e-06, "loss": 0.2936, "step": 18328 }, { "epoch": 20.55, "learning_rate": 5.604225352112677e-06, "loss": 0.341, "step": 18329 }, { "epoch": 20.55, "learning_rate": 5.6028169014084505e-06, "loss": 0.2603, "step": 18330 }, { "epoch": 20.55, "learning_rate": 5.601408450704225e-06, "loss": 0.2942, "step": 18331 }, { "epoch": 20.55, "learning_rate": 5.600000000000001e-06, "loss": 0.3183, "step": 18332 }, { "epoch": 20.55, "learning_rate": 5.598591549295775e-06, "loss": 0.3762, "step": 18333 }, { "epoch": 20.55, "learning_rate": 5.597183098591549e-06, "loss": 0.3732, "step": 18334 }, { "epoch": 20.55, "learning_rate": 5.595774647887324e-06, "loss": 0.3062, "step": 18335 }, { "epoch": 20.56, "learning_rate": 5.594366197183099e-06, "loss": 0.2534, "step": 18336 }, { "epoch": 20.56, "learning_rate": 5.592957746478874e-06, "loss": 0.2328, "step": 18337 }, { "epoch": 20.56, "learning_rate": 5.5915492957746476e-06, "loss": 0.2347, "step": 18338 }, { "epoch": 20.56, "learning_rate": 5.590140845070423e-06, "loss": 0.3248, "step": 18339 }, { "epoch": 20.56, "learning_rate": 5.588732394366198e-06, "loss": 0.3689, "step": 18340 }, { "epoch": 20.56, "learning_rate": 5.587323943661972e-06, "loss": 0.4761, "step": 18341 }, { "epoch": 20.56, "learning_rate": 5.585915492957746e-06, "loss": 0.4111, "step": 18342 }, { "epoch": 20.56, "learning_rate": 5.584507042253522e-06, "loss": 0.4222, "step": 18343 }, { "epoch": 20.57, "learning_rate": 5.583098591549296e-06, "loss": 0.4408, "step": 18344 }, { "epoch": 20.57, "learning_rate": 5.58169014084507e-06, "loss": 0.4537, "step": 18345 }, { "epoch": 20.57, "learning_rate": 5.580281690140845e-06, "loss": 0.3928, "step": 18346 }, { "epoch": 20.57, "learning_rate": 5.57887323943662e-06, "loss": 0.3867, "step": 18347 }, { "epoch": 20.57, "learning_rate": 5.577464788732395e-06, "loss": 0.4131, "step": 18348 }, { "epoch": 20.57, "learning_rate": 5.576056338028169e-06, "loss": 0.3701, "step": 18349 }, { "epoch": 20.57, "learning_rate": 5.574647887323943e-06, "loss": 0.3697, "step": 18350 }, { "epoch": 20.57, "learning_rate": 5.573239436619719e-06, "loss": 0.277, "step": 18351 }, { "epoch": 20.57, "learning_rate": 5.571830985915493e-06, "loss": 0.3654, "step": 18352 }, { "epoch": 20.58, "learning_rate": 5.570422535211267e-06, "loss": 0.3816, "step": 18353 }, { "epoch": 20.58, "learning_rate": 5.569014084507043e-06, "loss": 0.3178, "step": 18354 }, { "epoch": 20.58, "learning_rate": 5.567605633802817e-06, "loss": 0.4337, "step": 18355 }, { "epoch": 20.58, "learning_rate": 5.566197183098592e-06, "loss": 0.2938, "step": 18356 }, { "epoch": 20.58, "learning_rate": 5.564788732394366e-06, "loss": 0.3477, "step": 18357 }, { "epoch": 20.58, "learning_rate": 5.563380281690141e-06, "loss": 0.3709, "step": 18358 }, { "epoch": 20.58, "learning_rate": 5.561971830985916e-06, "loss": 0.3444, "step": 18359 }, { "epoch": 20.58, "learning_rate": 5.5605633802816904e-06, "loss": 0.3894, "step": 18360 }, { "epoch": 20.58, "learning_rate": 5.559154929577464e-06, "loss": 0.3511, "step": 18361 }, { "epoch": 20.59, "learning_rate": 5.55774647887324e-06, "loss": 0.2793, "step": 18362 }, { "epoch": 20.59, "learning_rate": 5.556338028169014e-06, "loss": 0.3513, "step": 18363 }, { "epoch": 20.59, "learning_rate": 5.554929577464789e-06, "loss": 0.3821, "step": 18364 }, { "epoch": 20.59, "learning_rate": 5.553521126760563e-06, "loss": 0.3623, "step": 18365 }, { "epoch": 20.59, "learning_rate": 5.552112676056338e-06, "loss": 0.2637, "step": 18366 }, { "epoch": 20.59, "learning_rate": 5.550704225352113e-06, "loss": 0.315, "step": 18367 }, { "epoch": 20.59, "learning_rate": 5.5492957746478875e-06, "loss": 0.3756, "step": 18368 }, { "epoch": 20.59, "learning_rate": 5.547887323943661e-06, "loss": 0.292, "step": 18369 }, { "epoch": 20.59, "learning_rate": 5.546478873239437e-06, "loss": 0.3451, "step": 18370 }, { "epoch": 20.6, "learning_rate": 5.5450704225352115e-06, "loss": 0.3773, "step": 18371 }, { "epoch": 20.6, "learning_rate": 5.543661971830986e-06, "loss": 0.2689, "step": 18372 }, { "epoch": 20.6, "learning_rate": 5.542253521126761e-06, "loss": 0.368, "step": 18373 }, { "epoch": 20.6, "learning_rate": 5.540845070422535e-06, "loss": 0.2709, "step": 18374 }, { "epoch": 20.6, "learning_rate": 5.53943661971831e-06, "loss": 0.3557, "step": 18375 }, { "epoch": 20.6, "learning_rate": 5.538028169014085e-06, "loss": 0.5234, "step": 18376 }, { "epoch": 20.6, "learning_rate": 5.536619718309859e-06, "loss": 0.2258, "step": 18377 }, { "epoch": 20.6, "learning_rate": 5.535211267605634e-06, "loss": 0.3469, "step": 18378 }, { "epoch": 20.6, "learning_rate": 5.5338028169014086e-06, "loss": 0.2721, "step": 18379 }, { "epoch": 20.61, "learning_rate": 5.532394366197183e-06, "loss": 0.4644, "step": 18380 }, { "epoch": 20.61, "learning_rate": 5.530985915492958e-06, "loss": 0.2742, "step": 18381 }, { "epoch": 20.61, "learning_rate": 5.5295774647887325e-06, "loss": 0.2848, "step": 18382 }, { "epoch": 20.61, "learning_rate": 5.528169014084507e-06, "loss": 0.2452, "step": 18383 }, { "epoch": 20.61, "learning_rate": 5.526760563380282e-06, "loss": 0.3222, "step": 18384 }, { "epoch": 20.61, "learning_rate": 5.525352112676056e-06, "loss": 0.206, "step": 18385 }, { "epoch": 20.61, "learning_rate": 5.523943661971831e-06, "loss": 0.2858, "step": 18386 }, { "epoch": 20.61, "learning_rate": 5.522535211267606e-06, "loss": 0.2975, "step": 18387 }, { "epoch": 20.61, "learning_rate": 5.521126760563381e-06, "loss": 0.3848, "step": 18388 }, { "epoch": 20.62, "learning_rate": 5.519718309859155e-06, "loss": 0.1867, "step": 18389 }, { "epoch": 20.62, "learning_rate": 5.5183098591549296e-06, "loss": 0.2378, "step": 18390 }, { "epoch": 20.62, "learning_rate": 5.516901408450704e-06, "loss": 0.4746, "step": 18391 }, { "epoch": 20.62, "learning_rate": 5.51549295774648e-06, "loss": 0.5491, "step": 18392 }, { "epoch": 20.62, "learning_rate": 5.5140845070422535e-06, "loss": 0.3281, "step": 18393 }, { "epoch": 20.62, "learning_rate": 5.512676056338028e-06, "loss": 0.5096, "step": 18394 }, { "epoch": 20.62, "learning_rate": 5.511267605633803e-06, "loss": 0.363, "step": 18395 }, { "epoch": 20.62, "learning_rate": 5.509859154929578e-06, "loss": 0.3188, "step": 18396 }, { "epoch": 20.62, "learning_rate": 5.508450704225352e-06, "loss": 0.4462, "step": 18397 }, { "epoch": 20.63, "learning_rate": 5.507042253521127e-06, "loss": 0.3673, "step": 18398 }, { "epoch": 20.63, "learning_rate": 5.505633802816901e-06, "loss": 0.3724, "step": 18399 }, { "epoch": 20.63, "learning_rate": 5.504225352112677e-06, "loss": 0.371, "step": 18400 }, { "epoch": 20.63, "learning_rate": 5.502816901408451e-06, "loss": 0.325, "step": 18401 }, { "epoch": 20.63, "learning_rate": 5.501408450704225e-06, "loss": 0.3366, "step": 18402 }, { "epoch": 20.63, "learning_rate": 5.5e-06, "loss": 0.4023, "step": 18403 }, { "epoch": 20.63, "learning_rate": 5.498591549295775e-06, "loss": 0.3262, "step": 18404 }, { "epoch": 20.63, "learning_rate": 5.497183098591549e-06, "loss": 0.2596, "step": 18405 }, { "epoch": 20.63, "learning_rate": 5.495774647887324e-06, "loss": 0.3682, "step": 18406 }, { "epoch": 20.64, "learning_rate": 5.494366197183099e-06, "loss": 0.3624, "step": 18407 }, { "epoch": 20.64, "learning_rate": 5.492957746478874e-06, "loss": 0.3799, "step": 18408 }, { "epoch": 20.64, "learning_rate": 5.491549295774648e-06, "loss": 0.3809, "step": 18409 }, { "epoch": 20.64, "learning_rate": 5.490140845070422e-06, "loss": 0.4166, "step": 18410 }, { "epoch": 20.64, "learning_rate": 5.488732394366198e-06, "loss": 0.302, "step": 18411 }, { "epoch": 20.64, "learning_rate": 5.4873239436619724e-06, "loss": 0.3346, "step": 18412 }, { "epoch": 20.64, "learning_rate": 5.485915492957746e-06, "loss": 0.2679, "step": 18413 }, { "epoch": 20.64, "learning_rate": 5.484507042253521e-06, "loss": 0.2677, "step": 18414 }, { "epoch": 20.64, "learning_rate": 5.483098591549296e-06, "loss": 0.3499, "step": 18415 }, { "epoch": 20.65, "learning_rate": 5.481690140845071e-06, "loss": 0.327, "step": 18416 }, { "epoch": 20.65, "learning_rate": 5.480281690140845e-06, "loss": 0.2633, "step": 18417 }, { "epoch": 20.65, "learning_rate": 5.478873239436619e-06, "loss": 0.3773, "step": 18418 }, { "epoch": 20.65, "learning_rate": 5.477464788732395e-06, "loss": 0.2656, "step": 18419 }, { "epoch": 20.65, "learning_rate": 5.4760563380281695e-06, "loss": 0.4261, "step": 18420 }, { "epoch": 20.65, "learning_rate": 5.474647887323943e-06, "loss": 0.3297, "step": 18421 }, { "epoch": 20.65, "learning_rate": 5.473239436619718e-06, "loss": 0.3939, "step": 18422 }, { "epoch": 20.65, "learning_rate": 5.4718309859154935e-06, "loss": 0.3401, "step": 18423 }, { "epoch": 20.65, "learning_rate": 5.470422535211268e-06, "loss": 0.3066, "step": 18424 }, { "epoch": 20.66, "learning_rate": 5.469014084507042e-06, "loss": 0.2949, "step": 18425 }, { "epoch": 20.66, "learning_rate": 5.467605633802817e-06, "loss": 0.3007, "step": 18426 }, { "epoch": 20.66, "learning_rate": 5.466197183098592e-06, "loss": 0.2896, "step": 18427 }, { "epoch": 20.66, "learning_rate": 5.464788732394367e-06, "loss": 0.2645, "step": 18428 }, { "epoch": 20.66, "learning_rate": 5.4633802816901404e-06, "loss": 0.3303, "step": 18429 }, { "epoch": 20.66, "learning_rate": 5.461971830985916e-06, "loss": 0.3197, "step": 18430 }, { "epoch": 20.66, "learning_rate": 5.4605633802816906e-06, "loss": 0.2939, "step": 18431 }, { "epoch": 20.66, "learning_rate": 5.459154929577465e-06, "loss": 0.2943, "step": 18432 }, { "epoch": 20.66, "learning_rate": 5.457746478873239e-06, "loss": 0.3089, "step": 18433 }, { "epoch": 20.67, "learning_rate": 5.4563380281690145e-06, "loss": 0.4401, "step": 18434 }, { "epoch": 20.67, "learning_rate": 5.454929577464789e-06, "loss": 0.3194, "step": 18435 }, { "epoch": 20.67, "learning_rate": 5.453521126760564e-06, "loss": 0.3742, "step": 18436 }, { "epoch": 20.67, "learning_rate": 5.4521126760563375e-06, "loss": 0.3154, "step": 18437 }, { "epoch": 20.67, "learning_rate": 5.450704225352113e-06, "loss": 0.2315, "step": 18438 }, { "epoch": 20.67, "learning_rate": 5.449295774647888e-06, "loss": 0.2808, "step": 18439 }, { "epoch": 20.67, "learning_rate": 5.447887323943662e-06, "loss": 0.2759, "step": 18440 }, { "epoch": 20.67, "learning_rate": 5.446478873239437e-06, "loss": 0.4338, "step": 18441 }, { "epoch": 20.67, "learning_rate": 5.4450704225352116e-06, "loss": 0.4275, "step": 18442 }, { "epoch": 20.68, "learning_rate": 5.443661971830986e-06, "loss": 0.3575, "step": 18443 }, { "epoch": 20.68, "learning_rate": 5.442253521126761e-06, "loss": 0.3906, "step": 18444 }, { "epoch": 20.68, "learning_rate": 5.4408450704225355e-06, "loss": 0.3244, "step": 18445 }, { "epoch": 20.68, "learning_rate": 5.43943661971831e-06, "loss": 0.4074, "step": 18446 }, { "epoch": 20.68, "learning_rate": 5.438028169014085e-06, "loss": 0.408, "step": 18447 }, { "epoch": 20.68, "learning_rate": 5.436619718309859e-06, "loss": 0.3665, "step": 18448 }, { "epoch": 20.68, "learning_rate": 5.435211267605634e-06, "loss": 0.4167, "step": 18449 }, { "epoch": 20.68, "learning_rate": 5.433802816901409e-06, "loss": 0.4146, "step": 18450 }, { "epoch": 20.68, "learning_rate": 5.432394366197183e-06, "loss": 0.4038, "step": 18451 }, { "epoch": 20.69, "learning_rate": 5.430985915492958e-06, "loss": 0.4006, "step": 18452 }, { "epoch": 20.69, "learning_rate": 5.429577464788733e-06, "loss": 0.3649, "step": 18453 }, { "epoch": 20.69, "learning_rate": 5.428169014084507e-06, "loss": 0.4308, "step": 18454 }, { "epoch": 20.69, "learning_rate": 5.426760563380282e-06, "loss": 0.3957, "step": 18455 }, { "epoch": 20.69, "learning_rate": 5.4253521126760565e-06, "loss": 0.3418, "step": 18456 }, { "epoch": 20.69, "learning_rate": 5.423943661971831e-06, "loss": 0.356, "step": 18457 }, { "epoch": 20.69, "learning_rate": 5.422535211267606e-06, "loss": 0.3332, "step": 18458 }, { "epoch": 20.69, "learning_rate": 5.42112676056338e-06, "loss": 0.2597, "step": 18459 }, { "epoch": 20.7, "learning_rate": 5.419718309859155e-06, "loss": 0.4026, "step": 18460 }, { "epoch": 20.7, "learning_rate": 5.41830985915493e-06, "loss": 0.3739, "step": 18461 }, { "epoch": 20.7, "learning_rate": 5.416901408450704e-06, "loss": 0.3543, "step": 18462 }, { "epoch": 20.7, "learning_rate": 5.415492957746479e-06, "loss": 0.2395, "step": 18463 }, { "epoch": 20.7, "learning_rate": 5.414084507042254e-06, "loss": 0.3314, "step": 18464 }, { "epoch": 20.7, "learning_rate": 5.412676056338028e-06, "loss": 0.3284, "step": 18465 }, { "epoch": 20.7, "learning_rate": 5.411267605633803e-06, "loss": 0.2772, "step": 18466 }, { "epoch": 20.7, "learning_rate": 5.4098591549295775e-06, "loss": 0.3817, "step": 18467 }, { "epoch": 20.7, "learning_rate": 5.408450704225352e-06, "loss": 0.2964, "step": 18468 }, { "epoch": 20.71, "learning_rate": 5.407042253521127e-06, "loss": 0.3347, "step": 18469 }, { "epoch": 20.71, "learning_rate": 5.405633802816901e-06, "loss": 0.3147, "step": 18470 }, { "epoch": 20.71, "learning_rate": 5.404225352112676e-06, "loss": 0.3858, "step": 18471 }, { "epoch": 20.71, "learning_rate": 5.402816901408451e-06, "loss": 0.3401, "step": 18472 }, { "epoch": 20.71, "learning_rate": 5.401408450704225e-06, "loss": 0.3151, "step": 18473 }, { "epoch": 20.71, "learning_rate": 5.4e-06, "loss": 0.3419, "step": 18474 }, { "epoch": 20.71, "learning_rate": 5.3985915492957755e-06, "loss": 0.3844, "step": 18475 }, { "epoch": 20.71, "learning_rate": 5.397183098591549e-06, "loss": 0.2963, "step": 18476 }, { "epoch": 20.71, "learning_rate": 5.395774647887324e-06, "loss": 0.3388, "step": 18477 }, { "epoch": 20.72, "learning_rate": 5.3943661971830985e-06, "loss": 0.3056, "step": 18478 }, { "epoch": 20.72, "learning_rate": 5.392957746478874e-06, "loss": 0.3181, "step": 18479 }, { "epoch": 20.72, "learning_rate": 5.391549295774648e-06, "loss": 0.2964, "step": 18480 }, { "epoch": 20.72, "learning_rate": 5.3901408450704224e-06, "loss": 0.344, "step": 18481 }, { "epoch": 20.72, "learning_rate": 5.388732394366197e-06, "loss": 0.3132, "step": 18482 }, { "epoch": 20.72, "learning_rate": 5.3873239436619726e-06, "loss": 0.3603, "step": 18483 }, { "epoch": 20.72, "learning_rate": 5.385915492957746e-06, "loss": 0.4234, "step": 18484 }, { "epoch": 20.72, "learning_rate": 5.384507042253521e-06, "loss": 0.2666, "step": 18485 }, { "epoch": 20.72, "learning_rate": 5.383098591549296e-06, "loss": 0.3598, "step": 18486 }, { "epoch": 20.73, "learning_rate": 5.381690140845071e-06, "loss": 0.2739, "step": 18487 }, { "epoch": 20.73, "learning_rate": 5.380281690140845e-06, "loss": 0.2474, "step": 18488 }, { "epoch": 20.73, "learning_rate": 5.3788732394366195e-06, "loss": 0.2291, "step": 18489 }, { "epoch": 20.73, "learning_rate": 5.377464788732394e-06, "loss": 0.3487, "step": 18490 }, { "epoch": 20.73, "learning_rate": 5.37605633802817e-06, "loss": 0.4819, "step": 18491 }, { "epoch": 20.73, "learning_rate": 5.3746478873239434e-06, "loss": 0.3393, "step": 18492 }, { "epoch": 20.73, "learning_rate": 5.373239436619718e-06, "loss": 0.4028, "step": 18493 }, { "epoch": 20.73, "learning_rate": 5.3718309859154936e-06, "loss": 0.4209, "step": 18494 }, { "epoch": 20.73, "learning_rate": 5.370422535211268e-06, "loss": 0.381, "step": 18495 }, { "epoch": 20.74, "learning_rate": 5.369014084507042e-06, "loss": 0.3279, "step": 18496 }, { "epoch": 20.74, "learning_rate": 5.367605633802817e-06, "loss": 0.3845, "step": 18497 }, { "epoch": 20.74, "learning_rate": 5.366197183098592e-06, "loss": 0.316, "step": 18498 }, { "epoch": 20.74, "learning_rate": 5.364788732394367e-06, "loss": 0.3385, "step": 18499 }, { "epoch": 20.74, "learning_rate": 5.3633802816901405e-06, "loss": 0.37, "step": 18500 }, { "epoch": 20.74, "learning_rate": 5.361971830985915e-06, "loss": 0.3273, "step": 18501 }, { "epoch": 20.74, "learning_rate": 5.360563380281691e-06, "loss": 0.3259, "step": 18502 }, { "epoch": 20.74, "learning_rate": 5.359154929577465e-06, "loss": 0.3709, "step": 18503 }, { "epoch": 20.74, "learning_rate": 5.357746478873239e-06, "loss": 0.4374, "step": 18504 }, { "epoch": 20.75, "learning_rate": 5.356338028169014e-06, "loss": 0.3663, "step": 18505 }, { "epoch": 20.75, "learning_rate": 5.354929577464789e-06, "loss": 0.4429, "step": 18506 }, { "epoch": 20.75, "learning_rate": 5.353521126760564e-06, "loss": 0.301, "step": 18507 }, { "epoch": 20.75, "learning_rate": 5.352112676056338e-06, "loss": 0.3684, "step": 18508 }, { "epoch": 20.75, "learning_rate": 5.350704225352112e-06, "loss": 0.4229, "step": 18509 }, { "epoch": 20.75, "learning_rate": 5.349295774647888e-06, "loss": 0.414, "step": 18510 }, { "epoch": 20.75, "learning_rate": 5.347887323943662e-06, "loss": 0.3456, "step": 18511 }, { "epoch": 20.75, "learning_rate": 5.346478873239436e-06, "loss": 0.4, "step": 18512 }, { "epoch": 20.75, "learning_rate": 5.345070422535212e-06, "loss": 0.3883, "step": 18513 }, { "epoch": 20.76, "learning_rate": 5.343661971830986e-06, "loss": 0.3319, "step": 18514 }, { "epoch": 20.76, "learning_rate": 5.342253521126761e-06, "loss": 0.3821, "step": 18515 }, { "epoch": 20.76, "learning_rate": 5.340845070422535e-06, "loss": 0.3791, "step": 18516 }, { "epoch": 20.76, "learning_rate": 5.33943661971831e-06, "loss": 0.2941, "step": 18517 }, { "epoch": 20.76, "learning_rate": 5.338028169014085e-06, "loss": 0.4289, "step": 18518 }, { "epoch": 20.76, "learning_rate": 5.3366197183098595e-06, "loss": 0.2888, "step": 18519 }, { "epoch": 20.76, "learning_rate": 5.335211267605633e-06, "loss": 0.3052, "step": 18520 }, { "epoch": 20.76, "learning_rate": 5.333802816901409e-06, "loss": 0.3495, "step": 18521 }, { "epoch": 20.76, "learning_rate": 5.332394366197183e-06, "loss": 0.3651, "step": 18522 }, { "epoch": 20.77, "learning_rate": 5.330985915492958e-06, "loss": 0.3782, "step": 18523 }, { "epoch": 20.77, "learning_rate": 5.329577464788732e-06, "loss": 0.3042, "step": 18524 }, { "epoch": 20.77, "learning_rate": 5.328169014084507e-06, "loss": 0.3429, "step": 18525 }, { "epoch": 20.77, "learning_rate": 5.326760563380282e-06, "loss": 0.4142, "step": 18526 }, { "epoch": 20.77, "learning_rate": 5.325352112676057e-06, "loss": 0.3494, "step": 18527 }, { "epoch": 20.77, "learning_rate": 5.323943661971831e-06, "loss": 0.2753, "step": 18528 }, { "epoch": 20.77, "learning_rate": 5.322535211267606e-06, "loss": 0.2885, "step": 18529 }, { "epoch": 20.77, "learning_rate": 5.3211267605633805e-06, "loss": 0.3621, "step": 18530 }, { "epoch": 20.77, "learning_rate": 5.319718309859155e-06, "loss": 0.3092, "step": 18531 }, { "epoch": 20.78, "learning_rate": 5.31830985915493e-06, "loss": 0.3968, "step": 18532 }, { "epoch": 20.78, "learning_rate": 5.3169014084507044e-06, "loss": 0.2541, "step": 18533 }, { "epoch": 20.78, "learning_rate": 5.315492957746479e-06, "loss": 0.2887, "step": 18534 }, { "epoch": 20.78, "learning_rate": 5.314084507042254e-06, "loss": 0.2863, "step": 18535 }, { "epoch": 20.78, "learning_rate": 5.312676056338028e-06, "loss": 0.3165, "step": 18536 }, { "epoch": 20.78, "learning_rate": 5.311267605633803e-06, "loss": 0.2115, "step": 18537 }, { "epoch": 20.78, "learning_rate": 5.309859154929578e-06, "loss": 0.3458, "step": 18538 }, { "epoch": 20.78, "learning_rate": 5.308450704225352e-06, "loss": 0.3049, "step": 18539 }, { "epoch": 20.78, "learning_rate": 5.307042253521127e-06, "loss": 0.5347, "step": 18540 }, { "epoch": 20.79, "learning_rate": 5.3056338028169015e-06, "loss": 0.4191, "step": 18541 }, { "epoch": 20.79, "learning_rate": 5.304225352112676e-06, "loss": 0.3938, "step": 18542 }, { "epoch": 20.79, "learning_rate": 5.302816901408451e-06, "loss": 0.4927, "step": 18543 }, { "epoch": 20.79, "learning_rate": 5.3014084507042254e-06, "loss": 0.3897, "step": 18544 }, { "epoch": 20.79, "learning_rate": 5.3e-06, "loss": 0.4268, "step": 18545 }, { "epoch": 20.79, "learning_rate": 5.298591549295775e-06, "loss": 0.4546, "step": 18546 }, { "epoch": 20.79, "learning_rate": 5.29718309859155e-06, "loss": 0.375, "step": 18547 }, { "epoch": 20.79, "learning_rate": 5.295774647887324e-06, "loss": 0.4501, "step": 18548 }, { "epoch": 20.79, "learning_rate": 5.294366197183099e-06, "loss": 0.3671, "step": 18549 }, { "epoch": 20.8, "learning_rate": 5.292957746478873e-06, "loss": 0.3381, "step": 18550 }, { "epoch": 20.8, "learning_rate": 5.291549295774649e-06, "loss": 0.3582, "step": 18551 }, { "epoch": 20.8, "learning_rate": 5.2901408450704225e-06, "loss": 0.3456, "step": 18552 }, { "epoch": 20.8, "learning_rate": 5.288732394366197e-06, "loss": 0.3847, "step": 18553 }, { "epoch": 20.8, "learning_rate": 5.287323943661972e-06, "loss": 0.2999, "step": 18554 }, { "epoch": 20.8, "learning_rate": 5.285915492957747e-06, "loss": 0.3779, "step": 18555 }, { "epoch": 20.8, "learning_rate": 5.284507042253521e-06, "loss": 0.3841, "step": 18556 }, { "epoch": 20.8, "learning_rate": 5.283098591549296e-06, "loss": 0.3562, "step": 18557 }, { "epoch": 20.8, "learning_rate": 5.28169014084507e-06, "loss": 0.3775, "step": 18558 }, { "epoch": 20.81, "learning_rate": 5.280281690140846e-06, "loss": 0.3219, "step": 18559 }, { "epoch": 20.81, "learning_rate": 5.27887323943662e-06, "loss": 0.4062, "step": 18560 }, { "epoch": 20.81, "learning_rate": 5.277464788732394e-06, "loss": 0.3431, "step": 18561 }, { "epoch": 20.81, "learning_rate": 5.27605633802817e-06, "loss": 0.4562, "step": 18562 }, { "epoch": 20.81, "learning_rate": 5.274647887323944e-06, "loss": 0.2757, "step": 18563 }, { "epoch": 20.81, "learning_rate": 5.273239436619718e-06, "loss": 0.3658, "step": 18564 }, { "epoch": 20.81, "learning_rate": 5.271830985915493e-06, "loss": 0.4007, "step": 18565 }, { "epoch": 20.81, "learning_rate": 5.270422535211268e-06, "loss": 0.3677, "step": 18566 }, { "epoch": 20.82, "learning_rate": 5.269014084507043e-06, "loss": 0.3176, "step": 18567 }, { "epoch": 20.82, "learning_rate": 5.267605633802817e-06, "loss": 0.3449, "step": 18568 }, { "epoch": 20.82, "learning_rate": 5.266197183098591e-06, "loss": 0.2798, "step": 18569 }, { "epoch": 20.82, "learning_rate": 5.264788732394367e-06, "loss": 0.3101, "step": 18570 }, { "epoch": 20.82, "learning_rate": 5.263380281690141e-06, "loss": 0.2996, "step": 18571 }, { "epoch": 20.82, "learning_rate": 5.261971830985915e-06, "loss": 0.2858, "step": 18572 }, { "epoch": 20.82, "learning_rate": 5.26056338028169e-06, "loss": 0.3441, "step": 18573 }, { "epoch": 20.82, "learning_rate": 5.259154929577465e-06, "loss": 0.4191, "step": 18574 }, { "epoch": 20.82, "learning_rate": 5.257746478873239e-06, "loss": 0.3029, "step": 18575 }, { "epoch": 20.83, "learning_rate": 5.256338028169014e-06, "loss": 0.3077, "step": 18576 }, { "epoch": 20.83, "learning_rate": 5.2549295774647885e-06, "loss": 0.2918, "step": 18577 }, { "epoch": 20.83, "learning_rate": 5.253521126760564e-06, "loss": 0.3193, "step": 18578 }, { "epoch": 20.83, "learning_rate": 5.252112676056338e-06, "loss": 0.3275, "step": 18579 }, { "epoch": 20.83, "learning_rate": 5.250704225352112e-06, "loss": 0.3753, "step": 18580 }, { "epoch": 20.83, "learning_rate": 5.249295774647888e-06, "loss": 0.3709, "step": 18581 }, { "epoch": 20.83, "learning_rate": 5.2478873239436625e-06, "loss": 0.3085, "step": 18582 }, { "epoch": 20.83, "learning_rate": 5.246478873239436e-06, "loss": 0.5419, "step": 18583 }, { "epoch": 20.83, "learning_rate": 5.245070422535211e-06, "loss": 0.3365, "step": 18584 }, { "epoch": 20.84, "learning_rate": 5.2436619718309864e-06, "loss": 0.3306, "step": 18585 }, { "epoch": 20.84, "learning_rate": 5.242253521126761e-06, "loss": 0.4143, "step": 18586 }, { "epoch": 20.84, "learning_rate": 5.240845070422535e-06, "loss": 0.2226, "step": 18587 }, { "epoch": 20.84, "learning_rate": 5.2394366197183095e-06, "loss": 0.3377, "step": 18588 }, { "epoch": 20.84, "learning_rate": 5.238028169014085e-06, "loss": 0.3354, "step": 18589 }, { "epoch": 20.84, "learning_rate": 5.23661971830986e-06, "loss": 0.3319, "step": 18590 }, { "epoch": 20.84, "learning_rate": 5.235211267605633e-06, "loss": 0.4849, "step": 18591 }, { "epoch": 20.84, "learning_rate": 5.233802816901408e-06, "loss": 0.4536, "step": 18592 }, { "epoch": 20.84, "learning_rate": 5.2323943661971835e-06, "loss": 0.3452, "step": 18593 }, { "epoch": 20.85, "learning_rate": 5.230985915492958e-06, "loss": 0.3809, "step": 18594 }, { "epoch": 20.85, "learning_rate": 5.229577464788732e-06, "loss": 0.3415, "step": 18595 }, { "epoch": 20.85, "learning_rate": 5.228169014084507e-06, "loss": 0.4177, "step": 18596 }, { "epoch": 20.85, "learning_rate": 5.226760563380282e-06, "loss": 0.3487, "step": 18597 }, { "epoch": 20.85, "learning_rate": 5.225352112676057e-06, "loss": 0.3461, "step": 18598 }, { "epoch": 20.85, "learning_rate": 5.2239436619718305e-06, "loss": 0.3155, "step": 18599 }, { "epoch": 20.85, "learning_rate": 5.222535211267606e-06, "loss": 0.4143, "step": 18600 }, { "epoch": 20.85, "learning_rate": 5.221126760563381e-06, "loss": 0.3668, "step": 18601 }, { "epoch": 20.85, "learning_rate": 5.219718309859155e-06, "loss": 0.3721, "step": 18602 }, { "epoch": 20.86, "learning_rate": 5.218309859154929e-06, "loss": 0.3386, "step": 18603 }, { "epoch": 20.86, "learning_rate": 5.2169014084507045e-06, "loss": 0.3421, "step": 18604 }, { "epoch": 20.86, "learning_rate": 5.215492957746479e-06, "loss": 0.3714, "step": 18605 }, { "epoch": 20.86, "learning_rate": 5.214084507042254e-06, "loss": 0.3036, "step": 18606 }, { "epoch": 20.86, "learning_rate": 5.212676056338028e-06, "loss": 0.3587, "step": 18607 }, { "epoch": 20.86, "learning_rate": 5.211267605633803e-06, "loss": 0.3025, "step": 18608 }, { "epoch": 20.86, "learning_rate": 5.209859154929578e-06, "loss": 0.3099, "step": 18609 }, { "epoch": 20.86, "learning_rate": 5.208450704225352e-06, "loss": 0.35, "step": 18610 }, { "epoch": 20.86, "learning_rate": 5.207042253521126e-06, "loss": 0.4116, "step": 18611 }, { "epoch": 20.87, "learning_rate": 5.205633802816902e-06, "loss": 0.3794, "step": 18612 }, { "epoch": 20.87, "learning_rate": 5.204225352112676e-06, "loss": 0.3887, "step": 18613 }, { "epoch": 20.87, "learning_rate": 5.202816901408451e-06, "loss": 0.2725, "step": 18614 }, { "epoch": 20.87, "learning_rate": 5.2014084507042256e-06, "loss": 0.3116, "step": 18615 }, { "epoch": 20.87, "learning_rate": 5.2e-06, "loss": 0.3845, "step": 18616 }, { "epoch": 20.87, "learning_rate": 5.198591549295775e-06, "loss": 0.2602, "step": 18617 }, { "epoch": 20.87, "learning_rate": 5.1971830985915495e-06, "loss": 0.2972, "step": 18618 }, { "epoch": 20.87, "learning_rate": 5.195774647887324e-06, "loss": 0.4556, "step": 18619 }, { "epoch": 20.87, "learning_rate": 5.194366197183099e-06, "loss": 0.3403, "step": 18620 }, { "epoch": 20.88, "learning_rate": 5.192957746478873e-06, "loss": 0.2951, "step": 18621 }, { "epoch": 20.88, "learning_rate": 5.191549295774648e-06, "loss": 0.3201, "step": 18622 }, { "epoch": 20.88, "learning_rate": 5.190140845070423e-06, "loss": 0.3281, "step": 18623 }, { "epoch": 20.88, "learning_rate": 5.188732394366197e-06, "loss": 0.281, "step": 18624 }, { "epoch": 20.88, "learning_rate": 5.187323943661972e-06, "loss": 0.272, "step": 18625 }, { "epoch": 20.88, "learning_rate": 5.1859154929577466e-06, "loss": 0.4038, "step": 18626 }, { "epoch": 20.88, "learning_rate": 5.184507042253521e-06, "loss": 0.31, "step": 18627 }, { "epoch": 20.88, "learning_rate": 5.183098591549296e-06, "loss": 0.2301, "step": 18628 }, { "epoch": 20.88, "learning_rate": 5.1816901408450705e-06, "loss": 0.3389, "step": 18629 }, { "epoch": 20.89, "learning_rate": 5.180281690140845e-06, "loss": 0.3782, "step": 18630 }, { "epoch": 20.89, "learning_rate": 5.17887323943662e-06, "loss": 0.3531, "step": 18631 }, { "epoch": 20.89, "learning_rate": 5.177464788732394e-06, "loss": 0.3335, "step": 18632 }, { "epoch": 20.89, "learning_rate": 5.176056338028169e-06, "loss": 0.3027, "step": 18633 }, { "epoch": 20.89, "learning_rate": 5.1746478873239445e-06, "loss": 0.2745, "step": 18634 }, { "epoch": 20.89, "learning_rate": 5.173239436619718e-06, "loss": 0.3059, "step": 18635 }, { "epoch": 20.89, "learning_rate": 5.171830985915493e-06, "loss": 0.2699, "step": 18636 }, { "epoch": 20.89, "learning_rate": 5.1704225352112676e-06, "loss": 0.2302, "step": 18637 }, { "epoch": 20.89, "learning_rate": 5.169014084507043e-06, "loss": 0.2823, "step": 18638 }, { "epoch": 20.9, "learning_rate": 5.167605633802817e-06, "loss": 0.3, "step": 18639 }, { "epoch": 20.9, "learning_rate": 5.1661971830985915e-06, "loss": 0.2997, "step": 18640 }, { "epoch": 20.9, "learning_rate": 5.164788732394366e-06, "loss": 0.4138, "step": 18641 }, { "epoch": 20.9, "learning_rate": 5.163380281690142e-06, "loss": 0.3924, "step": 18642 }, { "epoch": 20.9, "learning_rate": 5.161971830985915e-06, "loss": 0.4852, "step": 18643 }, { "epoch": 20.9, "learning_rate": 5.16056338028169e-06, "loss": 0.4435, "step": 18644 }, { "epoch": 20.9, "learning_rate": 5.159154929577465e-06, "loss": 0.4669, "step": 18645 }, { "epoch": 20.9, "learning_rate": 5.15774647887324e-06, "loss": 0.3364, "step": 18646 }, { "epoch": 20.9, "learning_rate": 5.156338028169014e-06, "loss": 0.3419, "step": 18647 }, { "epoch": 20.91, "learning_rate": 5.154929577464789e-06, "loss": 0.3462, "step": 18648 }, { "epoch": 20.91, "learning_rate": 5.153521126760564e-06, "loss": 0.3709, "step": 18649 }, { "epoch": 20.91, "learning_rate": 5.152112676056339e-06, "loss": 0.344, "step": 18650 }, { "epoch": 20.91, "learning_rate": 5.1507042253521125e-06, "loss": 0.3253, "step": 18651 }, { "epoch": 20.91, "learning_rate": 5.149295774647887e-06, "loss": 0.4181, "step": 18652 }, { "epoch": 20.91, "learning_rate": 5.147887323943663e-06, "loss": 0.3877, "step": 18653 }, { "epoch": 20.91, "learning_rate": 5.146478873239437e-06, "loss": 0.3568, "step": 18654 }, { "epoch": 20.91, "learning_rate": 5.145070422535211e-06, "loss": 0.3376, "step": 18655 }, { "epoch": 20.91, "learning_rate": 5.143661971830986e-06, "loss": 0.3719, "step": 18656 }, { "epoch": 20.92, "learning_rate": 5.142253521126761e-06, "loss": 0.2833, "step": 18657 }, { "epoch": 20.92, "learning_rate": 5.140845070422536e-06, "loss": 0.3542, "step": 18658 }, { "epoch": 20.92, "learning_rate": 5.13943661971831e-06, "loss": 0.302, "step": 18659 }, { "epoch": 20.92, "learning_rate": 5.138028169014084e-06, "loss": 0.2859, "step": 18660 }, { "epoch": 20.92, "learning_rate": 5.13661971830986e-06, "loss": 0.4049, "step": 18661 }, { "epoch": 20.92, "learning_rate": 5.135211267605634e-06, "loss": 0.3437, "step": 18662 }, { "epoch": 20.92, "learning_rate": 5.133802816901408e-06, "loss": 0.3322, "step": 18663 }, { "epoch": 20.92, "learning_rate": 5.132394366197183e-06, "loss": 0.3734, "step": 18664 }, { "epoch": 20.92, "learning_rate": 5.130985915492958e-06, "loss": 0.2843, "step": 18665 }, { "epoch": 20.93, "learning_rate": 5.129577464788733e-06, "loss": 0.3666, "step": 18666 }, { "epoch": 20.93, "learning_rate": 5.128169014084507e-06, "loss": 0.3436, "step": 18667 }, { "epoch": 20.93, "learning_rate": 5.126760563380282e-06, "loss": 0.2917, "step": 18668 }, { "epoch": 20.93, "learning_rate": 5.125352112676057e-06, "loss": 0.3322, "step": 18669 }, { "epoch": 20.93, "learning_rate": 5.1239436619718315e-06, "loss": 0.3181, "step": 18670 }, { "epoch": 20.93, "learning_rate": 5.122535211267605e-06, "loss": 0.3227, "step": 18671 }, { "epoch": 20.93, "learning_rate": 5.121126760563381e-06, "loss": 0.3014, "step": 18672 }, { "epoch": 20.93, "learning_rate": 5.119718309859155e-06, "loss": 0.2623, "step": 18673 }, { "epoch": 20.93, "learning_rate": 5.11830985915493e-06, "loss": 0.3179, "step": 18674 }, { "epoch": 20.94, "learning_rate": 5.116901408450704e-06, "loss": 0.3093, "step": 18675 }, { "epoch": 20.94, "learning_rate": 5.115492957746479e-06, "loss": 0.2692, "step": 18676 }, { "epoch": 20.94, "learning_rate": 5.114084507042254e-06, "loss": 0.3128, "step": 18677 }, { "epoch": 20.94, "learning_rate": 5.1126760563380286e-06, "loss": 0.2646, "step": 18678 }, { "epoch": 20.94, "learning_rate": 5.111267605633802e-06, "loss": 0.2761, "step": 18679 }, { "epoch": 20.94, "learning_rate": 5.109859154929578e-06, "loss": 0.3319, "step": 18680 }, { "epoch": 20.94, "learning_rate": 5.1084507042253525e-06, "loss": 0.2794, "step": 18681 }, { "epoch": 20.94, "learning_rate": 5.107042253521127e-06, "loss": 0.4185, "step": 18682 }, { "epoch": 20.95, "learning_rate": 5.105633802816902e-06, "loss": 0.2978, "step": 18683 }, { "epoch": 20.95, "learning_rate": 5.104225352112676e-06, "loss": 0.3, "step": 18684 }, { "epoch": 20.95, "learning_rate": 5.102816901408451e-06, "loss": 0.3305, "step": 18685 }, { "epoch": 20.95, "learning_rate": 5.101408450704225e-06, "loss": 0.1937, "step": 18686 }, { "epoch": 20.95, "learning_rate": 5.1e-06, "loss": 0.2673, "step": 18687 }, { "epoch": 20.95, "learning_rate": 5.098591549295775e-06, "loss": 0.3167, "step": 18688 }, { "epoch": 20.95, "learning_rate": 5.0971830985915496e-06, "loss": 0.3453, "step": 18689 }, { "epoch": 20.95, "learning_rate": 5.095774647887323e-06, "loss": 0.2092, "step": 18690 }, { "epoch": 20.95, "learning_rate": 5.094366197183099e-06, "loss": 0.4512, "step": 18691 }, { "epoch": 20.96, "learning_rate": 5.0929577464788735e-06, "loss": 0.3724, "step": 18692 }, { "epoch": 20.96, "learning_rate": 5.091549295774648e-06, "loss": 0.3394, "step": 18693 }, { "epoch": 20.96, "learning_rate": 5.090140845070422e-06, "loss": 0.4025, "step": 18694 }, { "epoch": 20.96, "learning_rate": 5.088732394366197e-06, "loss": 0.3258, "step": 18695 }, { "epoch": 20.96, "learning_rate": 5.087323943661972e-06, "loss": 0.3724, "step": 18696 }, { "epoch": 20.96, "learning_rate": 5.085915492957747e-06, "loss": 0.3579, "step": 18697 }, { "epoch": 20.96, "learning_rate": 5.0845070422535205e-06, "loss": 0.4838, "step": 18698 }, { "epoch": 20.96, "learning_rate": 5.083098591549296e-06, "loss": 0.3445, "step": 18699 }, { "epoch": 20.96, "learning_rate": 5.081690140845071e-06, "loss": 0.3927, "step": 18700 }, { "epoch": 20.97, "learning_rate": 5.080281690140845e-06, "loss": 0.3025, "step": 18701 }, { "epoch": 20.97, "learning_rate": 5.07887323943662e-06, "loss": 0.2809, "step": 18702 }, { "epoch": 20.97, "learning_rate": 5.0774647887323945e-06, "loss": 0.4653, "step": 18703 }, { "epoch": 20.97, "learning_rate": 5.076056338028169e-06, "loss": 0.3847, "step": 18704 }, { "epoch": 20.97, "learning_rate": 5.074647887323944e-06, "loss": 0.2688, "step": 18705 }, { "epoch": 20.97, "learning_rate": 5.073239436619718e-06, "loss": 0.3556, "step": 18706 }, { "epoch": 20.97, "learning_rate": 5.071830985915493e-06, "loss": 0.3501, "step": 18707 }, { "epoch": 20.97, "learning_rate": 5.070422535211268e-06, "loss": 0.3545, "step": 18708 }, { "epoch": 20.97, "learning_rate": 5.069014084507042e-06, "loss": 0.3231, "step": 18709 }, { "epoch": 20.98, "learning_rate": 5.067605633802817e-06, "loss": 0.3197, "step": 18710 }, { "epoch": 20.98, "learning_rate": 5.066197183098592e-06, "loss": 0.3229, "step": 18711 }, { "epoch": 20.98, "learning_rate": 5.064788732394366e-06, "loss": 0.3468, "step": 18712 }, { "epoch": 20.98, "learning_rate": 5.063380281690141e-06, "loss": 0.3619, "step": 18713 }, { "epoch": 20.98, "learning_rate": 5.0619718309859155e-06, "loss": 0.3512, "step": 18714 }, { "epoch": 20.98, "learning_rate": 5.06056338028169e-06, "loss": 0.3709, "step": 18715 }, { "epoch": 20.98, "learning_rate": 5.059154929577465e-06, "loss": 0.3389, "step": 18716 }, { "epoch": 20.98, "learning_rate": 5.0577464788732394e-06, "loss": 0.3367, "step": 18717 }, { "epoch": 20.98, "learning_rate": 5.056338028169014e-06, "loss": 0.3271, "step": 18718 }, { "epoch": 20.99, "learning_rate": 5.054929577464789e-06, "loss": 0.3443, "step": 18719 }, { "epoch": 20.99, "learning_rate": 5.053521126760563e-06, "loss": 0.3354, "step": 18720 }, { "epoch": 20.99, "learning_rate": 5.052112676056339e-06, "loss": 0.3025, "step": 18721 }, { "epoch": 20.99, "learning_rate": 5.050704225352113e-06, "loss": 0.3949, "step": 18722 }, { "epoch": 20.99, "learning_rate": 5.049295774647887e-06, "loss": 0.4131, "step": 18723 }, { "epoch": 20.99, "learning_rate": 5.047887323943662e-06, "loss": 0.355, "step": 18724 }, { "epoch": 20.99, "learning_rate": 5.046478873239437e-06, "loss": 0.3179, "step": 18725 }, { "epoch": 20.99, "learning_rate": 5.045070422535211e-06, "loss": 0.3987, "step": 18726 }, { "epoch": 20.99, "learning_rate": 5.043661971830986e-06, "loss": 0.2864, "step": 18727 }, { "epoch": 21.0, "learning_rate": 5.0422535211267604e-06, "loss": 0.2458, "step": 18728 }, { "epoch": 21.0, "learning_rate": 5.040845070422536e-06, "loss": 0.2408, "step": 18729 }, { "epoch": 21.0, "learning_rate": 5.03943661971831e-06, "loss": 0.2276, "step": 18730 }, { "epoch": 21.0, "learning_rate": 5.038028169014084e-06, "loss": 0.2839, "step": 18731 }, { "epoch": 21.0, "learning_rate": 5.036619718309859e-06, "loss": 0.3947, "step": 18732 }, { "epoch": 21.0, "learning_rate": 5.0352112676056345e-06, "loss": 0.4854, "step": 18733 }, { "epoch": 21.0, "learning_rate": 5.033802816901408e-06, "loss": 0.4022, "step": 18734 }, { "epoch": 21.0, "learning_rate": 5.032394366197183e-06, "loss": 0.3971, "step": 18735 }, { "epoch": 21.0, "learning_rate": 5.030985915492958e-06, "loss": 0.4352, "step": 18736 }, { "epoch": 21.01, "learning_rate": 5.029577464788733e-06, "loss": 0.308, "step": 18737 }, { "epoch": 21.01, "learning_rate": 5.028169014084507e-06, "loss": 0.3548, "step": 18738 }, { "epoch": 21.01, "learning_rate": 5.0267605633802814e-06, "loss": 0.3564, "step": 18739 }, { "epoch": 21.01, "learning_rate": 5.025352112676057e-06, "loss": 0.3491, "step": 18740 }, { "epoch": 21.01, "learning_rate": 5.0239436619718316e-06, "loss": 0.3623, "step": 18741 }, { "epoch": 21.01, "learning_rate": 5.022535211267605e-06, "loss": 0.3187, "step": 18742 }, { "epoch": 21.01, "learning_rate": 5.02112676056338e-06, "loss": 0.2663, "step": 18743 }, { "epoch": 21.01, "learning_rate": 5.0197183098591555e-06, "loss": 0.326, "step": 18744 }, { "epoch": 21.01, "learning_rate": 5.01830985915493e-06, "loss": 0.3339, "step": 18745 }, { "epoch": 21.02, "learning_rate": 5.016901408450704e-06, "loss": 0.344, "step": 18746 }, { "epoch": 21.02, "learning_rate": 5.0154929577464785e-06, "loss": 0.359, "step": 18747 }, { "epoch": 21.02, "learning_rate": 5.014084507042254e-06, "loss": 0.3208, "step": 18748 }, { "epoch": 21.02, "learning_rate": 5.012676056338029e-06, "loss": 0.3109, "step": 18749 }, { "epoch": 21.02, "learning_rate": 5.0112676056338025e-06, "loss": 0.3217, "step": 18750 }, { "epoch": 21.02, "learning_rate": 5.009859154929577e-06, "loss": 0.2639, "step": 18751 }, { "epoch": 21.02, "learning_rate": 5.008450704225353e-06, "loss": 0.3356, "step": 18752 }, { "epoch": 21.02, "learning_rate": 5.007042253521127e-06, "loss": 0.392, "step": 18753 }, { "epoch": 21.02, "learning_rate": 5.005633802816901e-06, "loss": 0.3205, "step": 18754 }, { "epoch": 21.03, "learning_rate": 5.0042253521126765e-06, "loss": 0.3211, "step": 18755 }, { "epoch": 21.03, "learning_rate": 5.002816901408451e-06, "loss": 0.2987, "step": 18756 }, { "epoch": 21.03, "learning_rate": 5.001408450704226e-06, "loss": 0.3015, "step": 18757 }, { "epoch": 21.03, "learning_rate": 4.9999999999999996e-06, "loss": 0.3019, "step": 18758 }, { "epoch": 21.03, "learning_rate": 4.998591549295775e-06, "loss": 0.2827, "step": 18759 }, { "epoch": 21.03, "learning_rate": 4.99718309859155e-06, "loss": 0.2929, "step": 18760 }, { "epoch": 21.03, "learning_rate": 4.995774647887324e-06, "loss": 0.3073, "step": 18761 }, { "epoch": 21.03, "learning_rate": 4.994366197183098e-06, "loss": 0.3086, "step": 18762 }, { "epoch": 21.03, "learning_rate": 4.992957746478874e-06, "loss": 0.3864, "step": 18763 }, { "epoch": 21.04, "learning_rate": 4.991549295774648e-06, "loss": 0.3374, "step": 18764 }, { "epoch": 21.04, "learning_rate": 4.990140845070423e-06, "loss": 0.3719, "step": 18765 }, { "epoch": 21.04, "learning_rate": 4.988732394366197e-06, "loss": 0.3292, "step": 18766 }, { "epoch": 21.04, "learning_rate": 4.987323943661972e-06, "loss": 0.3574, "step": 18767 }, { "epoch": 21.04, "learning_rate": 4.985915492957747e-06, "loss": 0.3782, "step": 18768 }, { "epoch": 21.04, "learning_rate": 4.984507042253521e-06, "loss": 0.3001, "step": 18769 }, { "epoch": 21.04, "learning_rate": 4.983098591549296e-06, "loss": 0.2578, "step": 18770 }, { "epoch": 21.04, "learning_rate": 4.981690140845071e-06, "loss": 0.3126, "step": 18771 }, { "epoch": 21.04, "learning_rate": 4.980281690140845e-06, "loss": 0.2113, "step": 18772 }, { "epoch": 21.05, "learning_rate": 4.97887323943662e-06, "loss": 0.3114, "step": 18773 }, { "epoch": 21.05, "learning_rate": 4.977464788732395e-06, "loss": 0.3094, "step": 18774 }, { "epoch": 21.05, "learning_rate": 4.976056338028169e-06, "loss": 0.2974, "step": 18775 }, { "epoch": 21.05, "learning_rate": 4.974647887323944e-06, "loss": 0.2334, "step": 18776 }, { "epoch": 21.05, "learning_rate": 4.9732394366197185e-06, "loss": 0.2768, "step": 18777 }, { "epoch": 21.05, "learning_rate": 4.971830985915493e-06, "loss": 0.3331, "step": 18778 }, { "epoch": 21.05, "learning_rate": 4.970422535211268e-06, "loss": 0.3676, "step": 18779 }, { "epoch": 21.05, "learning_rate": 4.9690140845070424e-06, "loss": 0.2183, "step": 18780 }, { "epoch": 21.05, "learning_rate": 4.967605633802817e-06, "loss": 0.3519, "step": 18781 }, { "epoch": 21.06, "learning_rate": 4.966197183098592e-06, "loss": 0.2536, "step": 18782 }, { "epoch": 21.06, "learning_rate": 4.964788732394366e-06, "loss": 0.3889, "step": 18783 }, { "epoch": 21.06, "learning_rate": 4.963380281690141e-06, "loss": 0.4322, "step": 18784 }, { "epoch": 21.06, "learning_rate": 4.961971830985916e-06, "loss": 0.3647, "step": 18785 }, { "epoch": 21.06, "learning_rate": 4.96056338028169e-06, "loss": 0.4175, "step": 18786 }, { "epoch": 21.06, "learning_rate": 4.959154929577465e-06, "loss": 0.3698, "step": 18787 }, { "epoch": 21.06, "learning_rate": 4.9577464788732395e-06, "loss": 0.3575, "step": 18788 }, { "epoch": 21.06, "learning_rate": 4.956338028169015e-06, "loss": 0.3636, "step": 18789 }, { "epoch": 21.07, "learning_rate": 4.954929577464789e-06, "loss": 0.3261, "step": 18790 }, { "epoch": 21.07, "learning_rate": 4.9535211267605634e-06, "loss": 0.392, "step": 18791 }, { "epoch": 21.07, "learning_rate": 4.952112676056338e-06, "loss": 0.3581, "step": 18792 }, { "epoch": 21.07, "learning_rate": 4.950704225352113e-06, "loss": 0.3282, "step": 18793 }, { "epoch": 21.07, "learning_rate": 4.949295774647887e-06, "loss": 0.347, "step": 18794 }, { "epoch": 21.07, "learning_rate": 4.947887323943662e-06, "loss": 0.3416, "step": 18795 }, { "epoch": 21.07, "learning_rate": 4.946478873239437e-06, "loss": 0.2932, "step": 18796 }, { "epoch": 21.07, "learning_rate": 4.945070422535211e-06, "loss": 0.323, "step": 18797 }, { "epoch": 21.07, "learning_rate": 4.943661971830986e-06, "loss": 0.3041, "step": 18798 }, { "epoch": 21.08, "learning_rate": 4.9422535211267605e-06, "loss": 0.3978, "step": 18799 }, { "epoch": 21.08, "learning_rate": 4.940845070422535e-06, "loss": 0.3379, "step": 18800 }, { "epoch": 21.08, "learning_rate": 4.93943661971831e-06, "loss": 0.3599, "step": 18801 }, { "epoch": 21.08, "learning_rate": 4.9380281690140845e-06, "loss": 0.3063, "step": 18802 }, { "epoch": 21.08, "learning_rate": 4.936619718309859e-06, "loss": 0.2974, "step": 18803 }, { "epoch": 21.08, "learning_rate": 4.935211267605634e-06, "loss": 0.3624, "step": 18804 }, { "epoch": 21.08, "learning_rate": 4.933802816901408e-06, "loss": 0.3374, "step": 18805 }, { "epoch": 21.08, "learning_rate": 4.932394366197183e-06, "loss": 0.4306, "step": 18806 }, { "epoch": 21.08, "learning_rate": 4.930985915492958e-06, "loss": 0.336, "step": 18807 }, { "epoch": 21.09, "learning_rate": 4.929577464788733e-06, "loss": 0.2515, "step": 18808 }, { "epoch": 21.09, "learning_rate": 4.928169014084507e-06, "loss": 0.3456, "step": 18809 }, { "epoch": 21.09, "learning_rate": 4.9267605633802816e-06, "loss": 0.2343, "step": 18810 }, { "epoch": 21.09, "learning_rate": 4.925352112676056e-06, "loss": 0.3331, "step": 18811 }, { "epoch": 21.09, "learning_rate": 4.923943661971832e-06, "loss": 0.3014, "step": 18812 }, { "epoch": 21.09, "learning_rate": 4.9225352112676055e-06, "loss": 0.2416, "step": 18813 }, { "epoch": 21.09, "learning_rate": 4.92112676056338e-06, "loss": 0.3195, "step": 18814 }, { "epoch": 21.09, "learning_rate": 4.919718309859155e-06, "loss": 0.2577, "step": 18815 }, { "epoch": 21.09, "learning_rate": 4.91830985915493e-06, "loss": 0.3458, "step": 18816 }, { "epoch": 21.1, "learning_rate": 4.916901408450704e-06, "loss": 0.3536, "step": 18817 }, { "epoch": 21.1, "learning_rate": 4.915492957746479e-06, "loss": 0.3816, "step": 18818 }, { "epoch": 21.1, "learning_rate": 4.914084507042253e-06, "loss": 0.2998, "step": 18819 }, { "epoch": 21.1, "learning_rate": 4.912676056338029e-06, "loss": 0.2258, "step": 18820 }, { "epoch": 21.1, "learning_rate": 4.9112676056338026e-06, "loss": 0.3878, "step": 18821 }, { "epoch": 21.1, "learning_rate": 4.909859154929577e-06, "loss": 0.2646, "step": 18822 }, { "epoch": 21.1, "learning_rate": 4.908450704225353e-06, "loss": 0.3552, "step": 18823 }, { "epoch": 21.1, "learning_rate": 4.907042253521127e-06, "loss": 0.3212, "step": 18824 }, { "epoch": 21.1, "learning_rate": 4.905633802816901e-06, "loss": 0.2711, "step": 18825 }, { "epoch": 21.11, "learning_rate": 4.904225352112676e-06, "loss": 0.3077, "step": 18826 }, { "epoch": 21.11, "learning_rate": 4.902816901408451e-06, "loss": 0.3377, "step": 18827 }, { "epoch": 21.11, "learning_rate": 4.901408450704226e-06, "loss": 0.1917, "step": 18828 }, { "epoch": 21.11, "learning_rate": 4.9e-06, "loss": 0.2229, "step": 18829 }, { "epoch": 21.11, "learning_rate": 4.898591549295774e-06, "loss": 0.183, "step": 18830 }, { "epoch": 21.11, "learning_rate": 4.89718309859155e-06, "loss": 0.246, "step": 18831 }, { "epoch": 21.11, "learning_rate": 4.8957746478873244e-06, "loss": 0.3007, "step": 18832 }, { "epoch": 21.11, "learning_rate": 4.894366197183098e-06, "loss": 0.3868, "step": 18833 }, { "epoch": 21.11, "learning_rate": 4.892957746478873e-06, "loss": 0.3949, "step": 18834 }, { "epoch": 21.12, "learning_rate": 4.891549295774648e-06, "loss": 0.3745, "step": 18835 }, { "epoch": 21.12, "learning_rate": 4.890140845070423e-06, "loss": 0.424, "step": 18836 }, { "epoch": 21.12, "learning_rate": 4.888732394366197e-06, "loss": 0.3522, "step": 18837 }, { "epoch": 21.12, "learning_rate": 4.887323943661971e-06, "loss": 0.3432, "step": 18838 }, { "epoch": 21.12, "learning_rate": 4.885915492957747e-06, "loss": 0.3719, "step": 18839 }, { "epoch": 21.12, "learning_rate": 4.8845070422535215e-06, "loss": 0.3411, "step": 18840 }, { "epoch": 21.12, "learning_rate": 4.883098591549295e-06, "loss": 0.3434, "step": 18841 }, { "epoch": 21.12, "learning_rate": 4.881690140845071e-06, "loss": 0.2927, "step": 18842 }, { "epoch": 21.12, "learning_rate": 4.8802816901408454e-06, "loss": 0.285, "step": 18843 }, { "epoch": 21.13, "learning_rate": 4.87887323943662e-06, "loss": 0.4077, "step": 18844 }, { "epoch": 21.13, "learning_rate": 4.877464788732394e-06, "loss": 0.3514, "step": 18845 }, { "epoch": 21.13, "learning_rate": 4.876056338028169e-06, "loss": 0.3241, "step": 18846 }, { "epoch": 21.13, "learning_rate": 4.874647887323944e-06, "loss": 0.3144, "step": 18847 }, { "epoch": 21.13, "learning_rate": 4.873239436619719e-06, "loss": 0.2904, "step": 18848 }, { "epoch": 21.13, "learning_rate": 4.871830985915492e-06, "loss": 0.3743, "step": 18849 }, { "epoch": 21.13, "learning_rate": 4.870422535211268e-06, "loss": 0.2763, "step": 18850 }, { "epoch": 21.13, "learning_rate": 4.8690140845070425e-06, "loss": 0.3219, "step": 18851 }, { "epoch": 21.13, "learning_rate": 4.867605633802817e-06, "loss": 0.3609, "step": 18852 }, { "epoch": 21.14, "learning_rate": 4.866197183098591e-06, "loss": 0.348, "step": 18853 }, { "epoch": 21.14, "learning_rate": 4.8647887323943665e-06, "loss": 0.4101, "step": 18854 }, { "epoch": 21.14, "learning_rate": 4.863380281690141e-06, "loss": 0.3092, "step": 18855 }, { "epoch": 21.14, "learning_rate": 4.861971830985916e-06, "loss": 0.3266, "step": 18856 }, { "epoch": 21.14, "learning_rate": 4.86056338028169e-06, "loss": 0.3106, "step": 18857 }, { "epoch": 21.14, "learning_rate": 4.859154929577465e-06, "loss": 0.2806, "step": 18858 }, { "epoch": 21.14, "learning_rate": 4.85774647887324e-06, "loss": 0.3154, "step": 18859 }, { "epoch": 21.14, "learning_rate": 4.856338028169014e-06, "loss": 0.3642, "step": 18860 }, { "epoch": 21.14, "learning_rate": 4.854929577464789e-06, "loss": 0.3083, "step": 18861 }, { "epoch": 21.15, "learning_rate": 4.8535211267605636e-06, "loss": 0.3255, "step": 18862 }, { "epoch": 21.15, "learning_rate": 4.852112676056338e-06, "loss": 0.2662, "step": 18863 }, { "epoch": 21.15, "learning_rate": 4.850704225352113e-06, "loss": 0.3648, "step": 18864 }, { "epoch": 21.15, "learning_rate": 4.8492957746478875e-06, "loss": 0.2762, "step": 18865 }, { "epoch": 21.15, "learning_rate": 4.847887323943662e-06, "loss": 0.3378, "step": 18866 }, { "epoch": 21.15, "learning_rate": 4.846478873239437e-06, "loss": 0.3385, "step": 18867 }, { "epoch": 21.15, "learning_rate": 4.845070422535211e-06, "loss": 0.2979, "step": 18868 }, { "epoch": 21.15, "learning_rate": 4.843661971830986e-06, "loss": 0.2707, "step": 18869 }, { "epoch": 21.15, "learning_rate": 4.842253521126761e-06, "loss": 0.3512, "step": 18870 }, { "epoch": 21.16, "learning_rate": 4.840845070422535e-06, "loss": 0.295, "step": 18871 }, { "epoch": 21.16, "learning_rate": 4.83943661971831e-06, "loss": 0.2977, "step": 18872 }, { "epoch": 21.16, "learning_rate": 4.8380281690140846e-06, "loss": 0.292, "step": 18873 }, { "epoch": 21.16, "learning_rate": 4.836619718309859e-06, "loss": 0.2908, "step": 18874 }, { "epoch": 21.16, "learning_rate": 4.835211267605634e-06, "loss": 0.2224, "step": 18875 }, { "epoch": 21.16, "learning_rate": 4.833802816901409e-06, "loss": 0.3097, "step": 18876 }, { "epoch": 21.16, "learning_rate": 4.832394366197183e-06, "loss": 0.2297, "step": 18877 }, { "epoch": 21.16, "learning_rate": 4.830985915492958e-06, "loss": 0.2809, "step": 18878 }, { "epoch": 21.16, "learning_rate": 4.829577464788732e-06, "loss": 0.232, "step": 18879 }, { "epoch": 21.17, "learning_rate": 4.828169014084508e-06, "loss": 0.2656, "step": 18880 }, { "epoch": 21.17, "learning_rate": 4.826760563380282e-06, "loss": 0.2462, "step": 18881 }, { "epoch": 21.17, "learning_rate": 4.825352112676056e-06, "loss": 0.3454, "step": 18882 }, { "epoch": 21.17, "learning_rate": 4.823943661971831e-06, "loss": 0.4347, "step": 18883 }, { "epoch": 21.17, "learning_rate": 4.8225352112676064e-06, "loss": 0.5534, "step": 18884 }, { "epoch": 21.17, "learning_rate": 4.82112676056338e-06, "loss": 0.3171, "step": 18885 }, { "epoch": 21.17, "learning_rate": 4.819718309859155e-06, "loss": 0.3902, "step": 18886 }, { "epoch": 21.17, "learning_rate": 4.8183098591549295e-06, "loss": 0.2982, "step": 18887 }, { "epoch": 21.17, "learning_rate": 4.816901408450705e-06, "loss": 0.3795, "step": 18888 }, { "epoch": 21.18, "learning_rate": 4.815492957746479e-06, "loss": 0.4768, "step": 18889 }, { "epoch": 21.18, "learning_rate": 4.814084507042253e-06, "loss": 0.4057, "step": 18890 }, { "epoch": 21.18, "learning_rate": 4.812676056338029e-06, "loss": 0.4384, "step": 18891 }, { "epoch": 21.18, "learning_rate": 4.8112676056338035e-06, "loss": 0.3988, "step": 18892 }, { "epoch": 21.18, "learning_rate": 4.809859154929577e-06, "loss": 0.4003, "step": 18893 }, { "epoch": 21.18, "learning_rate": 4.808450704225352e-06, "loss": 0.2912, "step": 18894 }, { "epoch": 21.18, "learning_rate": 4.8070422535211274e-06, "loss": 0.4202, "step": 18895 }, { "epoch": 21.18, "learning_rate": 4.805633802816902e-06, "loss": 0.2587, "step": 18896 }, { "epoch": 21.18, "learning_rate": 4.804225352112676e-06, "loss": 0.3355, "step": 18897 }, { "epoch": 21.19, "learning_rate": 4.8028169014084505e-06, "loss": 0.3322, "step": 18898 }, { "epoch": 21.19, "learning_rate": 4.801408450704226e-06, "loss": 0.3588, "step": 18899 }, { "epoch": 21.19, "learning_rate": 4.800000000000001e-06, "loss": 0.3753, "step": 18900 }, { "epoch": 21.19, "learning_rate": 4.798591549295774e-06, "loss": 0.2947, "step": 18901 }, { "epoch": 21.19, "learning_rate": 4.797183098591549e-06, "loss": 0.4347, "step": 18902 }, { "epoch": 21.19, "learning_rate": 4.7957746478873245e-06, "loss": 0.2328, "step": 18903 }, { "epoch": 21.19, "learning_rate": 4.794366197183099e-06, "loss": 0.3418, "step": 18904 }, { "epoch": 21.19, "learning_rate": 4.792957746478873e-06, "loss": 0.2787, "step": 18905 }, { "epoch": 21.2, "learning_rate": 4.791549295774648e-06, "loss": 0.361, "step": 18906 }, { "epoch": 21.2, "learning_rate": 4.790140845070423e-06, "loss": 0.4083, "step": 18907 }, { "epoch": 21.2, "learning_rate": 4.788732394366198e-06, "loss": 0.3243, "step": 18908 }, { "epoch": 21.2, "learning_rate": 4.7873239436619715e-06, "loss": 0.3193, "step": 18909 }, { "epoch": 21.2, "learning_rate": 4.785915492957747e-06, "loss": 0.2879, "step": 18910 }, { "epoch": 21.2, "learning_rate": 4.784507042253522e-06, "loss": 0.2607, "step": 18911 }, { "epoch": 21.2, "learning_rate": 4.7830985915492954e-06, "loss": 0.3148, "step": 18912 }, { "epoch": 21.2, "learning_rate": 4.78169014084507e-06, "loss": 0.2017, "step": 18913 }, { "epoch": 21.2, "learning_rate": 4.7802816901408456e-06, "loss": 0.3004, "step": 18914 }, { "epoch": 21.21, "learning_rate": 4.77887323943662e-06, "loss": 0.3395, "step": 18915 }, { "epoch": 21.21, "learning_rate": 4.777464788732394e-06, "loss": 0.2966, "step": 18916 }, { "epoch": 21.21, "learning_rate": 4.776056338028169e-06, "loss": 0.3162, "step": 18917 }, { "epoch": 21.21, "learning_rate": 4.774647887323944e-06, "loss": 0.3275, "step": 18918 }, { "epoch": 21.21, "learning_rate": 4.773239436619719e-06, "loss": 0.3595, "step": 18919 }, { "epoch": 21.21, "learning_rate": 4.7718309859154925e-06, "loss": 0.2193, "step": 18920 }, { "epoch": 21.21, "learning_rate": 4.770422535211267e-06, "loss": 0.3887, "step": 18921 }, { "epoch": 21.21, "learning_rate": 4.769014084507043e-06, "loss": 0.2367, "step": 18922 }, { "epoch": 21.21, "learning_rate": 4.767605633802817e-06, "loss": 0.2605, "step": 18923 }, { "epoch": 21.22, "learning_rate": 4.766197183098591e-06, "loss": 0.3287, "step": 18924 }, { "epoch": 21.22, "learning_rate": 4.764788732394366e-06, "loss": 0.278, "step": 18925 }, { "epoch": 21.22, "learning_rate": 4.763380281690141e-06, "loss": 0.3063, "step": 18926 }, { "epoch": 21.22, "learning_rate": 4.761971830985916e-06, "loss": 0.3142, "step": 18927 }, { "epoch": 21.22, "learning_rate": 4.76056338028169e-06, "loss": 0.4392, "step": 18928 }, { "epoch": 21.22, "learning_rate": 4.759154929577465e-06, "loss": 0.253, "step": 18929 }, { "epoch": 21.22, "learning_rate": 4.75774647887324e-06, "loss": 0.1549, "step": 18930 }, { "epoch": 21.22, "learning_rate": 4.756338028169014e-06, "loss": 0.313, "step": 18931 }, { "epoch": 21.22, "learning_rate": 4.754929577464788e-06, "loss": 0.239, "step": 18932 }, { "epoch": 21.23, "learning_rate": 4.753521126760564e-06, "loss": 0.4481, "step": 18933 }, { "epoch": 21.23, "learning_rate": 4.752112676056338e-06, "loss": 0.3738, "step": 18934 }, { "epoch": 21.23, "learning_rate": 4.750704225352113e-06, "loss": 0.4451, "step": 18935 }, { "epoch": 21.23, "learning_rate": 4.749295774647887e-06, "loss": 0.33, "step": 18936 }, { "epoch": 21.23, "learning_rate": 4.747887323943662e-06, "loss": 0.235, "step": 18937 }, { "epoch": 21.23, "learning_rate": 4.746478873239437e-06, "loss": 0.3058, "step": 18938 }, { "epoch": 21.23, "learning_rate": 4.7450704225352115e-06, "loss": 0.3822, "step": 18939 }, { "epoch": 21.23, "learning_rate": 4.743661971830985e-06, "loss": 0.3297, "step": 18940 }, { "epoch": 21.23, "learning_rate": 4.742253521126761e-06, "loss": 0.3412, "step": 18941 }, { "epoch": 21.24, "learning_rate": 4.740845070422535e-06, "loss": 0.3755, "step": 18942 }, { "epoch": 21.24, "learning_rate": 4.73943661971831e-06, "loss": 0.3646, "step": 18943 }, { "epoch": 21.24, "learning_rate": 4.738028169014085e-06, "loss": 0.3299, "step": 18944 }, { "epoch": 21.24, "learning_rate": 4.736619718309859e-06, "loss": 0.3054, "step": 18945 }, { "epoch": 21.24, "learning_rate": 4.735211267605634e-06, "loss": 0.3195, "step": 18946 }, { "epoch": 21.24, "learning_rate": 4.733802816901409e-06, "loss": 0.2772, "step": 18947 }, { "epoch": 21.24, "learning_rate": 4.732394366197183e-06, "loss": 0.4597, "step": 18948 }, { "epoch": 21.24, "learning_rate": 4.730985915492958e-06, "loss": 0.3294, "step": 18949 }, { "epoch": 21.24, "learning_rate": 4.7295774647887325e-06, "loss": 0.3171, "step": 18950 }, { "epoch": 21.25, "learning_rate": 4.728169014084507e-06, "loss": 0.2924, "step": 18951 }, { "epoch": 21.25, "learning_rate": 4.726760563380282e-06, "loss": 0.433, "step": 18952 }, { "epoch": 21.25, "learning_rate": 4.725352112676056e-06, "loss": 0.2494, "step": 18953 }, { "epoch": 21.25, "learning_rate": 4.723943661971831e-06, "loss": 0.3928, "step": 18954 }, { "epoch": 21.25, "learning_rate": 4.722535211267606e-06, "loss": 0.3243, "step": 18955 }, { "epoch": 21.25, "learning_rate": 4.72112676056338e-06, "loss": 0.3221, "step": 18956 }, { "epoch": 21.25, "learning_rate": 4.719718309859155e-06, "loss": 0.3408, "step": 18957 }, { "epoch": 21.25, "learning_rate": 4.71830985915493e-06, "loss": 0.3364, "step": 18958 }, { "epoch": 21.25, "learning_rate": 4.716901408450704e-06, "loss": 0.3132, "step": 18959 }, { "epoch": 21.26, "learning_rate": 4.715492957746479e-06, "loss": 0.2949, "step": 18960 }, { "epoch": 21.26, "learning_rate": 4.7140845070422535e-06, "loss": 0.2777, "step": 18961 }, { "epoch": 21.26, "learning_rate": 4.712676056338028e-06, "loss": 0.3556, "step": 18962 }, { "epoch": 21.26, "learning_rate": 4.711267605633804e-06, "loss": 0.3209, "step": 18963 }, { "epoch": 21.26, "learning_rate": 4.7098591549295774e-06, "loss": 0.3049, "step": 18964 }, { "epoch": 21.26, "learning_rate": 4.708450704225352e-06, "loss": 0.2777, "step": 18965 }, { "epoch": 21.26, "learning_rate": 4.707042253521127e-06, "loss": 0.3043, "step": 18966 }, { "epoch": 21.26, "learning_rate": 4.705633802816902e-06, "loss": 0.2903, "step": 18967 }, { "epoch": 21.26, "learning_rate": 4.704225352112676e-06, "loss": 0.3981, "step": 18968 }, { "epoch": 21.27, "learning_rate": 4.702816901408451e-06, "loss": 0.2619, "step": 18969 }, { "epoch": 21.27, "learning_rate": 4.701408450704225e-06, "loss": 0.301, "step": 18970 }, { "epoch": 21.27, "learning_rate": 4.700000000000001e-06, "loss": 0.308, "step": 18971 }, { "epoch": 21.27, "learning_rate": 4.6985915492957745e-06, "loss": 0.3063, "step": 18972 }, { "epoch": 21.27, "learning_rate": 4.697183098591549e-06, "loss": 0.2863, "step": 18973 }, { "epoch": 21.27, "learning_rate": 4.695774647887324e-06, "loss": 0.3906, "step": 18974 }, { "epoch": 21.27, "learning_rate": 4.694366197183099e-06, "loss": 0.2595, "step": 18975 }, { "epoch": 21.27, "learning_rate": 4.692957746478873e-06, "loss": 0.2828, "step": 18976 }, { "epoch": 21.27, "learning_rate": 4.691549295774648e-06, "loss": 0.2634, "step": 18977 }, { "epoch": 21.28, "learning_rate": 4.690140845070423e-06, "loss": 0.2974, "step": 18978 }, { "epoch": 21.28, "learning_rate": 4.688732394366198e-06, "loss": 0.2744, "step": 18979 }, { "epoch": 21.28, "learning_rate": 4.687323943661972e-06, "loss": 0.3161, "step": 18980 }, { "epoch": 21.28, "learning_rate": 4.685915492957746e-06, "loss": 0.2952, "step": 18981 }, { "epoch": 21.28, "learning_rate": 4.684507042253522e-06, "loss": 0.3174, "step": 18982 }, { "epoch": 21.28, "learning_rate": 4.683098591549296e-06, "loss": 0.4614, "step": 18983 }, { "epoch": 21.28, "learning_rate": 4.68169014084507e-06, "loss": 0.3963, "step": 18984 }, { "epoch": 21.28, "learning_rate": 4.680281690140845e-06, "loss": 0.3361, "step": 18985 }, { "epoch": 21.28, "learning_rate": 4.67887323943662e-06, "loss": 0.4569, "step": 18986 }, { "epoch": 21.29, "learning_rate": 4.677464788732395e-06, "loss": 0.4024, "step": 18987 }, { "epoch": 21.29, "learning_rate": 4.676056338028169e-06, "loss": 0.3413, "step": 18988 }, { "epoch": 21.29, "learning_rate": 4.674647887323943e-06, "loss": 0.4146, "step": 18989 }, { "epoch": 21.29, "learning_rate": 4.673239436619719e-06, "loss": 0.3193, "step": 18990 }, { "epoch": 21.29, "learning_rate": 4.6718309859154935e-06, "loss": 0.3398, "step": 18991 }, { "epoch": 21.29, "learning_rate": 4.670422535211267e-06, "loss": 0.3602, "step": 18992 }, { "epoch": 21.29, "learning_rate": 4.669014084507042e-06, "loss": 0.3733, "step": 18993 }, { "epoch": 21.29, "learning_rate": 4.667605633802817e-06, "loss": 0.2835, "step": 18994 }, { "epoch": 21.29, "learning_rate": 4.666197183098592e-06, "loss": 0.3958, "step": 18995 }, { "epoch": 21.3, "learning_rate": 4.664788732394366e-06, "loss": 0.3357, "step": 18996 }, { "epoch": 21.3, "learning_rate": 4.663380281690141e-06, "loss": 0.423, "step": 18997 }, { "epoch": 21.3, "learning_rate": 4.661971830985916e-06, "loss": 0.3551, "step": 18998 }, { "epoch": 21.3, "learning_rate": 4.660563380281691e-06, "loss": 0.3567, "step": 18999 }, { "epoch": 21.3, "learning_rate": 4.659154929577464e-06, "loss": 0.3705, "step": 19000 }, { "epoch": 21.3, "learning_rate": 4.65774647887324e-06, "loss": 0.2896, "step": 19001 }, { "epoch": 21.3, "learning_rate": 4.6563380281690145e-06, "loss": 0.373, "step": 19002 }, { "epoch": 21.3, "learning_rate": 4.654929577464789e-06, "loss": 0.2886, "step": 19003 }, { "epoch": 21.3, "learning_rate": 4.653521126760563e-06, "loss": 0.2896, "step": 19004 }, { "epoch": 21.31, "learning_rate": 4.652112676056338e-06, "loss": 0.3235, "step": 19005 }, { "epoch": 21.31, "learning_rate": 4.650704225352113e-06, "loss": 0.4028, "step": 19006 }, { "epoch": 21.31, "learning_rate": 4.649295774647888e-06, "loss": 0.278, "step": 19007 }, { "epoch": 21.31, "learning_rate": 4.6478873239436615e-06, "loss": 0.3314, "step": 19008 }, { "epoch": 21.31, "learning_rate": 4.646478873239437e-06, "loss": 0.3222, "step": 19009 }, { "epoch": 21.31, "learning_rate": 4.645070422535212e-06, "loss": 0.3224, "step": 19010 }, { "epoch": 21.31, "learning_rate": 4.643661971830986e-06, "loss": 0.345, "step": 19011 }, { "epoch": 21.31, "learning_rate": 4.64225352112676e-06, "loss": 0.3195, "step": 19012 }, { "epoch": 21.32, "learning_rate": 4.6408450704225355e-06, "loss": 0.3962, "step": 19013 }, { "epoch": 21.32, "learning_rate": 4.63943661971831e-06, "loss": 0.3621, "step": 19014 }, { "epoch": 21.32, "learning_rate": 4.638028169014085e-06, "loss": 0.2476, "step": 19015 }, { "epoch": 21.32, "learning_rate": 4.6366197183098594e-06, "loss": 0.2789, "step": 19016 }, { "epoch": 21.32, "learning_rate": 4.635211267605634e-06, "loss": 0.3511, "step": 19017 }, { "epoch": 21.32, "learning_rate": 4.633802816901409e-06, "loss": 0.2714, "step": 19018 }, { "epoch": 21.32, "learning_rate": 4.632394366197183e-06, "loss": 0.3016, "step": 19019 }, { "epoch": 21.32, "learning_rate": 4.630985915492958e-06, "loss": 0.3569, "step": 19020 }, { "epoch": 21.32, "learning_rate": 4.629577464788733e-06, "loss": 0.2969, "step": 19021 }, { "epoch": 21.33, "learning_rate": 4.628169014084507e-06, "loss": 0.3213, "step": 19022 }, { "epoch": 21.33, "learning_rate": 4.626760563380282e-06, "loss": 0.3636, "step": 19023 }, { "epoch": 21.33, "learning_rate": 4.6253521126760565e-06, "loss": 0.3309, "step": 19024 }, { "epoch": 21.33, "learning_rate": 4.623943661971831e-06, "loss": 0.2829, "step": 19025 }, { "epoch": 21.33, "learning_rate": 4.622535211267606e-06, "loss": 0.2775, "step": 19026 }, { "epoch": 21.33, "learning_rate": 4.62112676056338e-06, "loss": 0.2834, "step": 19027 }, { "epoch": 21.33, "learning_rate": 4.619718309859155e-06, "loss": 0.2354, "step": 19028 }, { "epoch": 21.33, "learning_rate": 4.61830985915493e-06, "loss": 0.2577, "step": 19029 }, { "epoch": 21.33, "learning_rate": 4.616901408450704e-06, "loss": 0.2879, "step": 19030 }, { "epoch": 21.34, "learning_rate": 4.615492957746479e-06, "loss": 0.3873, "step": 19031 }, { "epoch": 21.34, "learning_rate": 4.614084507042254e-06, "loss": 0.3215, "step": 19032 }, { "epoch": 21.34, "learning_rate": 4.612676056338028e-06, "loss": 0.4904, "step": 19033 }, { "epoch": 21.34, "learning_rate": 4.611267605633803e-06, "loss": 0.3958, "step": 19034 }, { "epoch": 21.34, "learning_rate": 4.6098591549295775e-06, "loss": 0.4204, "step": 19035 }, { "epoch": 21.34, "learning_rate": 4.608450704225352e-06, "loss": 0.3771, "step": 19036 }, { "epoch": 21.34, "learning_rate": 4.607042253521127e-06, "loss": 0.3453, "step": 19037 }, { "epoch": 21.34, "learning_rate": 4.6056338028169015e-06, "loss": 0.3747, "step": 19038 }, { "epoch": 21.34, "learning_rate": 4.604225352112676e-06, "loss": 0.3607, "step": 19039 }, { "epoch": 21.35, "learning_rate": 4.602816901408451e-06, "loss": 0.357, "step": 19040 }, { "epoch": 21.35, "learning_rate": 4.601408450704225e-06, "loss": 0.2686, "step": 19041 }, { "epoch": 21.35, "learning_rate": 4.6e-06, "loss": 0.3791, "step": 19042 }, { "epoch": 21.35, "learning_rate": 4.598591549295775e-06, "loss": 0.4248, "step": 19043 }, { "epoch": 21.35, "learning_rate": 4.597183098591549e-06, "loss": 0.34, "step": 19044 }, { "epoch": 21.35, "learning_rate": 4.595774647887324e-06, "loss": 0.3314, "step": 19045 }, { "epoch": 21.35, "learning_rate": 4.5943661971830986e-06, "loss": 0.3331, "step": 19046 }, { "epoch": 21.35, "learning_rate": 4.592957746478873e-06, "loss": 0.3533, "step": 19047 }, { "epoch": 21.35, "learning_rate": 4.591549295774648e-06, "loss": 0.3228, "step": 19048 }, { "epoch": 21.36, "learning_rate": 4.5901408450704225e-06, "loss": 0.4214, "step": 19049 }, { "epoch": 21.36, "learning_rate": 4.588732394366198e-06, "loss": 0.4375, "step": 19050 }, { "epoch": 21.36, "learning_rate": 4.587323943661972e-06, "loss": 0.3711, "step": 19051 }, { "epoch": 21.36, "learning_rate": 4.585915492957746e-06, "loss": 0.2751, "step": 19052 }, { "epoch": 21.36, "learning_rate": 4.584507042253521e-06, "loss": 0.3784, "step": 19053 }, { "epoch": 21.36, "learning_rate": 4.5830985915492965e-06, "loss": 0.327, "step": 19054 }, { "epoch": 21.36, "learning_rate": 4.58169014084507e-06, "loss": 0.3061, "step": 19055 }, { "epoch": 21.36, "learning_rate": 4.580281690140845e-06, "loss": 0.3445, "step": 19056 }, { "epoch": 21.36, "learning_rate": 4.5788732394366196e-06, "loss": 0.2962, "step": 19057 }, { "epoch": 21.37, "learning_rate": 4.577464788732395e-06, "loss": 0.3205, "step": 19058 }, { "epoch": 21.37, "learning_rate": 4.576056338028169e-06, "loss": 0.3487, "step": 19059 }, { "epoch": 21.37, "learning_rate": 4.5746478873239435e-06, "loss": 0.3279, "step": 19060 }, { "epoch": 21.37, "learning_rate": 4.573239436619718e-06, "loss": 0.2598, "step": 19061 }, { "epoch": 21.37, "learning_rate": 4.571830985915494e-06, "loss": 0.3002, "step": 19062 }, { "epoch": 21.37, "learning_rate": 4.570422535211267e-06, "loss": 0.359, "step": 19063 }, { "epoch": 21.37, "learning_rate": 4.569014084507042e-06, "loss": 0.2969, "step": 19064 }, { "epoch": 21.37, "learning_rate": 4.5676056338028175e-06, "loss": 0.3145, "step": 19065 }, { "epoch": 21.37, "learning_rate": 4.566197183098592e-06, "loss": 0.2646, "step": 19066 }, { "epoch": 21.38, "learning_rate": 4.564788732394366e-06, "loss": 0.3227, "step": 19067 }, { "epoch": 21.38, "learning_rate": 4.563380281690141e-06, "loss": 0.3169, "step": 19068 }, { "epoch": 21.38, "learning_rate": 4.561971830985916e-06, "loss": 0.3018, "step": 19069 }, { "epoch": 21.38, "learning_rate": 4.560563380281691e-06, "loss": 0.2578, "step": 19070 }, { "epoch": 21.38, "learning_rate": 4.5591549295774645e-06, "loss": 0.292, "step": 19071 }, { "epoch": 21.38, "learning_rate": 4.557746478873239e-06, "loss": 0.2839, "step": 19072 }, { "epoch": 21.38, "learning_rate": 4.556338028169015e-06, "loss": 0.3222, "step": 19073 }, { "epoch": 21.38, "learning_rate": 4.554929577464789e-06, "loss": 0.313, "step": 19074 }, { "epoch": 21.38, "learning_rate": 4.553521126760563e-06, "loss": 0.2615, "step": 19075 }, { "epoch": 21.39, "learning_rate": 4.552112676056338e-06, "loss": 0.3824, "step": 19076 }, { "epoch": 21.39, "learning_rate": 4.550704225352113e-06, "loss": 0.3274, "step": 19077 }, { "epoch": 21.39, "learning_rate": 4.549295774647888e-06, "loss": 0.2293, "step": 19078 }, { "epoch": 21.39, "learning_rate": 4.547887323943662e-06, "loss": 0.285, "step": 19079 }, { "epoch": 21.39, "learning_rate": 4.546478873239436e-06, "loss": 0.1678, "step": 19080 }, { "epoch": 21.39, "learning_rate": 4.545070422535212e-06, "loss": 0.2076, "step": 19081 }, { "epoch": 21.39, "learning_rate": 4.543661971830986e-06, "loss": 0.4036, "step": 19082 }, { "epoch": 21.39, "learning_rate": 4.54225352112676e-06, "loss": 0.4188, "step": 19083 }, { "epoch": 21.39, "learning_rate": 4.540845070422536e-06, "loss": 0.4129, "step": 19084 }, { "epoch": 21.4, "learning_rate": 4.53943661971831e-06, "loss": 0.4121, "step": 19085 }, { "epoch": 21.4, "learning_rate": 4.538028169014085e-06, "loss": 0.3477, "step": 19086 }, { "epoch": 21.4, "learning_rate": 4.536619718309859e-06, "loss": 0.3997, "step": 19087 }, { "epoch": 21.4, "learning_rate": 4.535211267605634e-06, "loss": 0.289, "step": 19088 }, { "epoch": 21.4, "learning_rate": 4.533802816901409e-06, "loss": 0.2518, "step": 19089 }, { "epoch": 21.4, "learning_rate": 4.5323943661971835e-06, "loss": 0.3661, "step": 19090 }, { "epoch": 21.4, "learning_rate": 4.530985915492957e-06, "loss": 0.3667, "step": 19091 }, { "epoch": 21.4, "learning_rate": 4.529577464788733e-06, "loss": 0.3275, "step": 19092 }, { "epoch": 21.4, "learning_rate": 4.528169014084507e-06, "loss": 0.3238, "step": 19093 }, { "epoch": 21.41, "learning_rate": 4.526760563380282e-06, "loss": 0.3804, "step": 19094 }, { "epoch": 21.41, "learning_rate": 4.525352112676056e-06, "loss": 0.3319, "step": 19095 }, { "epoch": 21.41, "learning_rate": 4.523943661971831e-06, "loss": 0.2995, "step": 19096 }, { "epoch": 21.41, "learning_rate": 4.522535211267606e-06, "loss": 0.3571, "step": 19097 }, { "epoch": 21.41, "learning_rate": 4.5211267605633806e-06, "loss": 0.2764, "step": 19098 }, { "epoch": 21.41, "learning_rate": 4.519718309859154e-06, "loss": 0.2498, "step": 19099 }, { "epoch": 21.41, "learning_rate": 4.51830985915493e-06, "loss": 0.2824, "step": 19100 }, { "epoch": 21.41, "learning_rate": 4.5169014084507045e-06, "loss": 0.3176, "step": 19101 }, { "epoch": 21.41, "learning_rate": 4.515492957746479e-06, "loss": 0.3057, "step": 19102 }, { "epoch": 21.42, "learning_rate": 4.514084507042254e-06, "loss": 0.2886, "step": 19103 }, { "epoch": 21.42, "learning_rate": 4.512676056338028e-06, "loss": 0.327, "step": 19104 }, { "epoch": 21.42, "learning_rate": 4.511267605633803e-06, "loss": 0.3003, "step": 19105 }, { "epoch": 21.42, "learning_rate": 4.509859154929578e-06, "loss": 0.323, "step": 19106 }, { "epoch": 21.42, "learning_rate": 4.508450704225352e-06, "loss": 0.322, "step": 19107 }, { "epoch": 21.42, "learning_rate": 4.507042253521127e-06, "loss": 0.2686, "step": 19108 }, { "epoch": 21.42, "learning_rate": 4.5056338028169016e-06, "loss": 0.3911, "step": 19109 }, { "epoch": 21.42, "learning_rate": 4.504225352112676e-06, "loss": 0.2836, "step": 19110 }, { "epoch": 21.42, "learning_rate": 4.502816901408451e-06, "loss": 0.237, "step": 19111 }, { "epoch": 21.43, "learning_rate": 4.5014084507042255e-06, "loss": 0.3522, "step": 19112 }, { "epoch": 21.43, "learning_rate": 4.5e-06, "loss": 0.3137, "step": 19113 }, { "epoch": 21.43, "learning_rate": 4.498591549295775e-06, "loss": 0.3223, "step": 19114 }, { "epoch": 21.43, "learning_rate": 4.497183098591549e-06, "loss": 0.3607, "step": 19115 }, { "epoch": 21.43, "learning_rate": 4.495774647887324e-06, "loss": 0.3406, "step": 19116 }, { "epoch": 21.43, "learning_rate": 4.494366197183099e-06, "loss": 0.2942, "step": 19117 }, { "epoch": 21.43, "learning_rate": 4.492957746478874e-06, "loss": 0.3633, "step": 19118 }, { "epoch": 21.43, "learning_rate": 4.491549295774648e-06, "loss": 0.3525, "step": 19119 }, { "epoch": 21.43, "learning_rate": 4.490140845070423e-06, "loss": 0.3186, "step": 19120 }, { "epoch": 21.44, "learning_rate": 4.488732394366197e-06, "loss": 0.2388, "step": 19121 }, { "epoch": 21.44, "learning_rate": 4.487323943661973e-06, "loss": 0.2942, "step": 19122 }, { "epoch": 21.44, "learning_rate": 4.4859154929577465e-06, "loss": 0.2407, "step": 19123 }, { "epoch": 21.44, "learning_rate": 4.484507042253521e-06, "loss": 0.3934, "step": 19124 }, { "epoch": 21.44, "learning_rate": 4.483098591549296e-06, "loss": 0.244, "step": 19125 }, { "epoch": 21.44, "learning_rate": 4.481690140845071e-06, "loss": 0.2429, "step": 19126 }, { "epoch": 21.44, "learning_rate": 4.480281690140845e-06, "loss": 0.3697, "step": 19127 }, { "epoch": 21.44, "learning_rate": 4.47887323943662e-06, "loss": 0.1907, "step": 19128 }, { "epoch": 21.45, "learning_rate": 4.477464788732394e-06, "loss": 0.27, "step": 19129 }, { "epoch": 21.45, "learning_rate": 4.47605633802817e-06, "loss": 0.3129, "step": 19130 }, { "epoch": 21.45, "learning_rate": 4.474647887323944e-06, "loss": 0.2375, "step": 19131 }, { "epoch": 21.45, "learning_rate": 4.473239436619718e-06, "loss": 0.238, "step": 19132 }, { "epoch": 21.45, "learning_rate": 4.471830985915493e-06, "loss": 0.3871, "step": 19133 }, { "epoch": 21.45, "learning_rate": 4.4704225352112675e-06, "loss": 0.3668, "step": 19134 }, { "epoch": 21.45, "learning_rate": 4.469014084507042e-06, "loss": 0.4303, "step": 19135 }, { "epoch": 21.45, "learning_rate": 4.467605633802817e-06, "loss": 0.3345, "step": 19136 }, { "epoch": 21.45, "learning_rate": 4.466197183098592e-06, "loss": 0.3923, "step": 19137 }, { "epoch": 21.46, "learning_rate": 4.464788732394366e-06, "loss": 0.4338, "step": 19138 }, { "epoch": 21.46, "learning_rate": 4.463380281690141e-06, "loss": 0.3351, "step": 19139 }, { "epoch": 21.46, "learning_rate": 4.461971830985915e-06, "loss": 0.3804, "step": 19140 }, { "epoch": 21.46, "learning_rate": 4.460563380281691e-06, "loss": 0.3276, "step": 19141 }, { "epoch": 21.46, "learning_rate": 4.459154929577465e-06, "loss": 0.3657, "step": 19142 }, { "epoch": 21.46, "learning_rate": 4.457746478873239e-06, "loss": 0.3917, "step": 19143 }, { "epoch": 21.46, "learning_rate": 4.456338028169014e-06, "loss": 0.3245, "step": 19144 }, { "epoch": 21.46, "learning_rate": 4.454929577464789e-06, "loss": 0.3354, "step": 19145 }, { "epoch": 21.46, "learning_rate": 4.453521126760563e-06, "loss": 0.3256, "step": 19146 }, { "epoch": 21.47, "learning_rate": 4.452112676056338e-06, "loss": 0.3356, "step": 19147 }, { "epoch": 21.47, "learning_rate": 4.4507042253521124e-06, "loss": 0.4341, "step": 19148 }, { "epoch": 21.47, "learning_rate": 4.449295774647888e-06, "loss": 0.3936, "step": 19149 }, { "epoch": 21.47, "learning_rate": 4.447887323943662e-06, "loss": 0.3011, "step": 19150 }, { "epoch": 21.47, "learning_rate": 4.446478873239436e-06, "loss": 0.2962, "step": 19151 }, { "epoch": 21.47, "learning_rate": 4.445070422535212e-06, "loss": 0.3326, "step": 19152 }, { "epoch": 21.47, "learning_rate": 4.4436619718309865e-06, "loss": 0.2735, "step": 19153 }, { "epoch": 21.47, "learning_rate": 4.44225352112676e-06, "loss": 0.4064, "step": 19154 }, { "epoch": 21.47, "learning_rate": 4.440845070422535e-06, "loss": 0.2801, "step": 19155 }, { "epoch": 21.48, "learning_rate": 4.43943661971831e-06, "loss": 0.4034, "step": 19156 }, { "epoch": 21.48, "learning_rate": 4.438028169014085e-06, "loss": 0.3226, "step": 19157 }, { "epoch": 21.48, "learning_rate": 4.436619718309859e-06, "loss": 0.3215, "step": 19158 }, { "epoch": 21.48, "learning_rate": 4.4352112676056334e-06, "loss": 0.3509, "step": 19159 }, { "epoch": 21.48, "learning_rate": 4.433802816901409e-06, "loss": 0.4082, "step": 19160 }, { "epoch": 21.48, "learning_rate": 4.4323943661971836e-06, "loss": 0.3581, "step": 19161 }, { "epoch": 21.48, "learning_rate": 4.430985915492957e-06, "loss": 0.2941, "step": 19162 }, { "epoch": 21.48, "learning_rate": 4.429577464788732e-06, "loss": 0.3378, "step": 19163 }, { "epoch": 21.48, "learning_rate": 4.4281690140845075e-06, "loss": 0.3268, "step": 19164 }, { "epoch": 21.49, "learning_rate": 4.426760563380282e-06, "loss": 0.2612, "step": 19165 }, { "epoch": 21.49, "learning_rate": 4.425352112676056e-06, "loss": 0.2006, "step": 19166 }, { "epoch": 21.49, "learning_rate": 4.4239436619718305e-06, "loss": 0.3285, "step": 19167 }, { "epoch": 21.49, "learning_rate": 4.422535211267606e-06, "loss": 0.3538, "step": 19168 }, { "epoch": 21.49, "learning_rate": 4.421126760563381e-06, "loss": 0.3331, "step": 19169 }, { "epoch": 21.49, "learning_rate": 4.4197183098591545e-06, "loss": 0.3491, "step": 19170 }, { "epoch": 21.49, "learning_rate": 4.41830985915493e-06, "loss": 0.2746, "step": 19171 }, { "epoch": 21.49, "learning_rate": 4.4169014084507046e-06, "loss": 0.324, "step": 19172 }, { "epoch": 21.49, "learning_rate": 4.415492957746479e-06, "loss": 0.3986, "step": 19173 }, { "epoch": 21.5, "learning_rate": 4.414084507042253e-06, "loss": 0.2774, "step": 19174 }, { "epoch": 21.5, "learning_rate": 4.4126760563380285e-06, "loss": 0.3087, "step": 19175 }, { "epoch": 21.5, "learning_rate": 4.411267605633803e-06, "loss": 0.2164, "step": 19176 }, { "epoch": 21.5, "learning_rate": 4.409859154929578e-06, "loss": 0.2492, "step": 19177 }, { "epoch": 21.5, "learning_rate": 4.4084507042253516e-06, "loss": 0.2878, "step": 19178 }, { "epoch": 21.5, "learning_rate": 4.407042253521127e-06, "loss": 0.2125, "step": 19179 }, { "epoch": 21.5, "learning_rate": 4.405633802816902e-06, "loss": 0.2571, "step": 19180 }, { "epoch": 21.5, "learning_rate": 4.404225352112676e-06, "loss": 0.3483, "step": 19181 }, { "epoch": 21.5, "learning_rate": 4.40281690140845e-06, "loss": 0.2806, "step": 19182 }, { "epoch": 21.51, "learning_rate": 4.401408450704226e-06, "loss": 0.4755, "step": 19183 }, { "epoch": 21.51, "learning_rate": 4.4e-06, "loss": 0.4011, "step": 19184 }, { "epoch": 21.51, "learning_rate": 4.398591549295775e-06, "loss": 0.4982, "step": 19185 }, { "epoch": 21.51, "learning_rate": 4.3971830985915495e-06, "loss": 0.3643, "step": 19186 }, { "epoch": 21.51, "learning_rate": 4.395774647887324e-06, "loss": 0.391, "step": 19187 }, { "epoch": 21.51, "learning_rate": 4.394366197183099e-06, "loss": 0.3921, "step": 19188 }, { "epoch": 21.51, "learning_rate": 4.392957746478873e-06, "loss": 0.4007, "step": 19189 }, { "epoch": 21.51, "learning_rate": 4.391549295774648e-06, "loss": 0.345, "step": 19190 }, { "epoch": 21.51, "learning_rate": 4.390140845070423e-06, "loss": 0.2979, "step": 19191 }, { "epoch": 21.52, "learning_rate": 4.388732394366197e-06, "loss": 0.4083, "step": 19192 }, { "epoch": 21.52, "learning_rate": 4.387323943661972e-06, "loss": 0.3299, "step": 19193 }, { "epoch": 21.52, "learning_rate": 4.385915492957747e-06, "loss": 0.3798, "step": 19194 }, { "epoch": 21.52, "learning_rate": 4.384507042253521e-06, "loss": 0.3241, "step": 19195 }, { "epoch": 21.52, "learning_rate": 4.383098591549296e-06, "loss": 0.2805, "step": 19196 }, { "epoch": 21.52, "learning_rate": 4.3816901408450705e-06, "loss": 0.2853, "step": 19197 }, { "epoch": 21.52, "learning_rate": 4.380281690140845e-06, "loss": 0.277, "step": 19198 }, { "epoch": 21.52, "learning_rate": 4.37887323943662e-06, "loss": 0.3742, "step": 19199 }, { "epoch": 21.52, "learning_rate": 4.3774647887323944e-06, "loss": 0.4261, "step": 19200 }, { "epoch": 21.53, "learning_rate": 4.376056338028169e-06, "loss": 0.3006, "step": 19201 }, { "epoch": 21.53, "learning_rate": 4.374647887323944e-06, "loss": 0.3963, "step": 19202 }, { "epoch": 21.53, "learning_rate": 4.373239436619718e-06, "loss": 0.2454, "step": 19203 }, { "epoch": 21.53, "learning_rate": 4.371830985915493e-06, "loss": 0.346, "step": 19204 }, { "epoch": 21.53, "learning_rate": 4.3704225352112685e-06, "loss": 0.2822, "step": 19205 }, { "epoch": 21.53, "learning_rate": 4.369014084507042e-06, "loss": 0.3146, "step": 19206 }, { "epoch": 21.53, "learning_rate": 4.367605633802817e-06, "loss": 0.4075, "step": 19207 }, { "epoch": 21.53, "learning_rate": 4.3661971830985915e-06, "loss": 0.2864, "step": 19208 }, { "epoch": 21.53, "learning_rate": 4.364788732394367e-06, "loss": 0.27, "step": 19209 }, { "epoch": 21.54, "learning_rate": 4.363380281690141e-06, "loss": 0.3135, "step": 19210 }, { "epoch": 21.54, "learning_rate": 4.3619718309859154e-06, "loss": 0.3784, "step": 19211 }, { "epoch": 21.54, "learning_rate": 4.36056338028169e-06, "loss": 0.3217, "step": 19212 }, { "epoch": 21.54, "learning_rate": 4.3591549295774656e-06, "loss": 0.3293, "step": 19213 }, { "epoch": 21.54, "learning_rate": 4.357746478873239e-06, "loss": 0.2298, "step": 19214 }, { "epoch": 21.54, "learning_rate": 4.356338028169014e-06, "loss": 0.3315, "step": 19215 }, { "epoch": 21.54, "learning_rate": 4.354929577464789e-06, "loss": 0.3485, "step": 19216 }, { "epoch": 21.54, "learning_rate": 4.353521126760564e-06, "loss": 0.349, "step": 19217 }, { "epoch": 21.54, "learning_rate": 4.352112676056338e-06, "loss": 0.3841, "step": 19218 }, { "epoch": 21.55, "learning_rate": 4.3507042253521125e-06, "loss": 0.3158, "step": 19219 }, { "epoch": 21.55, "learning_rate": 4.349295774647887e-06, "loss": 0.2972, "step": 19220 }, { "epoch": 21.55, "learning_rate": 4.347887323943663e-06, "loss": 0.2964, "step": 19221 }, { "epoch": 21.55, "learning_rate": 4.3464788732394364e-06, "loss": 0.3251, "step": 19222 }, { "epoch": 21.55, "learning_rate": 4.345070422535211e-06, "loss": 0.4365, "step": 19223 }, { "epoch": 21.55, "learning_rate": 4.3436619718309866e-06, "loss": 0.2891, "step": 19224 }, { "epoch": 21.55, "learning_rate": 4.342253521126761e-06, "loss": 0.3276, "step": 19225 }, { "epoch": 21.55, "learning_rate": 4.340845070422535e-06, "loss": 0.2954, "step": 19226 }, { "epoch": 21.55, "learning_rate": 4.33943661971831e-06, "loss": 0.2836, "step": 19227 }, { "epoch": 21.56, "learning_rate": 4.338028169014085e-06, "loss": 0.1989, "step": 19228 }, { "epoch": 21.56, "learning_rate": 4.33661971830986e-06, "loss": 0.2556, "step": 19229 }, { "epoch": 21.56, "learning_rate": 4.3352112676056335e-06, "loss": 0.2736, "step": 19230 }, { "epoch": 21.56, "learning_rate": 4.333802816901408e-06, "loss": 0.2355, "step": 19231 }, { "epoch": 21.56, "learning_rate": 4.332394366197184e-06, "loss": 0.3231, "step": 19232 }, { "epoch": 21.56, "learning_rate": 4.330985915492958e-06, "loss": 0.378, "step": 19233 }, { "epoch": 21.56, "learning_rate": 4.329577464788732e-06, "loss": 0.3173, "step": 19234 }, { "epoch": 21.56, "learning_rate": 4.328169014084507e-06, "loss": 0.3659, "step": 19235 }, { "epoch": 21.57, "learning_rate": 4.326760563380282e-06, "loss": 0.3833, "step": 19236 }, { "epoch": 21.57, "learning_rate": 4.325352112676057e-06, "loss": 0.2956, "step": 19237 }, { "epoch": 21.57, "learning_rate": 4.323943661971831e-06, "loss": 0.3475, "step": 19238 }, { "epoch": 21.57, "learning_rate": 4.322535211267606e-06, "loss": 0.3913, "step": 19239 }, { "epoch": 21.57, "learning_rate": 4.321126760563381e-06, "loss": 0.3764, "step": 19240 }, { "epoch": 21.57, "learning_rate": 4.319718309859155e-06, "loss": 0.3378, "step": 19241 }, { "epoch": 21.57, "learning_rate": 4.318309859154929e-06, "loss": 0.3383, "step": 19242 }, { "epoch": 21.57, "learning_rate": 4.316901408450705e-06, "loss": 0.3518, "step": 19243 }, { "epoch": 21.57, "learning_rate": 4.315492957746479e-06, "loss": 0.3753, "step": 19244 }, { "epoch": 21.58, "learning_rate": 4.314084507042254e-06, "loss": 0.319, "step": 19245 }, { "epoch": 21.58, "learning_rate": 4.312676056338028e-06, "loss": 0.3707, "step": 19246 }, { "epoch": 21.58, "learning_rate": 4.311267605633803e-06, "loss": 0.3648, "step": 19247 }, { "epoch": 21.58, "learning_rate": 4.309859154929578e-06, "loss": 0.2774, "step": 19248 }, { "epoch": 21.58, "learning_rate": 4.3084507042253525e-06, "loss": 0.2781, "step": 19249 }, { "epoch": 21.58, "learning_rate": 4.307042253521126e-06, "loss": 0.3573, "step": 19250 }, { "epoch": 21.58, "learning_rate": 4.305633802816902e-06, "loss": 0.379, "step": 19251 }, { "epoch": 21.58, "learning_rate": 4.304225352112676e-06, "loss": 0.2826, "step": 19252 }, { "epoch": 21.58, "learning_rate": 4.30281690140845e-06, "loss": 0.3693, "step": 19253 }, { "epoch": 21.59, "learning_rate": 4.301408450704225e-06, "loss": 0.3272, "step": 19254 }, { "epoch": 21.59, "learning_rate": 4.3e-06, "loss": 0.3596, "step": 19255 }, { "epoch": 21.59, "learning_rate": 4.298591549295775e-06, "loss": 0.3651, "step": 19256 }, { "epoch": 21.59, "learning_rate": 4.297183098591549e-06, "loss": 0.3319, "step": 19257 }, { "epoch": 21.59, "learning_rate": 4.295774647887324e-06, "loss": 0.3496, "step": 19258 }, { "epoch": 21.59, "learning_rate": 4.294366197183099e-06, "loss": 0.3285, "step": 19259 }, { "epoch": 21.59, "learning_rate": 4.2929577464788735e-06, "loss": 0.3112, "step": 19260 }, { "epoch": 21.59, "learning_rate": 4.291549295774647e-06, "loss": 0.3474, "step": 19261 }, { "epoch": 21.59, "learning_rate": 4.290140845070423e-06, "loss": 0.4116, "step": 19262 }, { "epoch": 21.6, "learning_rate": 4.2887323943661974e-06, "loss": 0.3505, "step": 19263 }, { "epoch": 21.6, "learning_rate": 4.287323943661972e-06, "loss": 0.3388, "step": 19264 }, { "epoch": 21.6, "learning_rate": 4.285915492957746e-06, "loss": 0.2698, "step": 19265 }, { "epoch": 21.6, "learning_rate": 4.284507042253521e-06, "loss": 0.2974, "step": 19266 }, { "epoch": 21.6, "learning_rate": 4.283098591549296e-06, "loss": 0.3065, "step": 19267 }, { "epoch": 21.6, "learning_rate": 4.281690140845071e-06, "loss": 0.2523, "step": 19268 }, { "epoch": 21.6, "learning_rate": 4.280281690140844e-06, "loss": 0.2881, "step": 19269 }, { "epoch": 21.6, "learning_rate": 4.27887323943662e-06, "loss": 0.2994, "step": 19270 }, { "epoch": 21.6, "learning_rate": 4.2774647887323945e-06, "loss": 0.2767, "step": 19271 }, { "epoch": 21.61, "learning_rate": 4.276056338028169e-06, "loss": 0.3838, "step": 19272 }, { "epoch": 21.61, "learning_rate": 4.274647887323944e-06, "loss": 0.4482, "step": 19273 }, { "epoch": 21.61, "learning_rate": 4.2732394366197184e-06, "loss": 0.2527, "step": 19274 }, { "epoch": 21.61, "learning_rate": 4.271830985915493e-06, "loss": 0.1793, "step": 19275 }, { "epoch": 21.61, "learning_rate": 4.270422535211268e-06, "loss": 0.3036, "step": 19276 }, { "epoch": 21.61, "learning_rate": 4.269014084507042e-06, "loss": 0.2441, "step": 19277 }, { "epoch": 21.61, "learning_rate": 4.267605633802817e-06, "loss": 0.3077, "step": 19278 }, { "epoch": 21.61, "learning_rate": 4.266197183098592e-06, "loss": 0.2786, "step": 19279 }, { "epoch": 21.61, "learning_rate": 4.264788732394366e-06, "loss": 0.1987, "step": 19280 }, { "epoch": 21.62, "learning_rate": 4.263380281690141e-06, "loss": 0.2979, "step": 19281 }, { "epoch": 21.62, "learning_rate": 4.2619718309859155e-06, "loss": 0.2729, "step": 19282 }, { "epoch": 21.62, "learning_rate": 4.26056338028169e-06, "loss": 0.3868, "step": 19283 }, { "epoch": 21.62, "learning_rate": 4.259154929577465e-06, "loss": 0.4072, "step": 19284 }, { "epoch": 21.62, "learning_rate": 4.2577464788732395e-06, "loss": 0.3875, "step": 19285 }, { "epoch": 21.62, "learning_rate": 4.256338028169014e-06, "loss": 0.2855, "step": 19286 }, { "epoch": 21.62, "learning_rate": 4.254929577464789e-06, "loss": 0.4471, "step": 19287 }, { "epoch": 21.62, "learning_rate": 4.253521126760563e-06, "loss": 0.3498, "step": 19288 }, { "epoch": 21.62, "learning_rate": 4.252112676056338e-06, "loss": 0.2636, "step": 19289 }, { "epoch": 21.63, "learning_rate": 4.250704225352113e-06, "loss": 0.4265, "step": 19290 }, { "epoch": 21.63, "learning_rate": 4.249295774647887e-06, "loss": 0.4246, "step": 19291 }, { "epoch": 21.63, "learning_rate": 4.247887323943663e-06, "loss": 0.3392, "step": 19292 }, { "epoch": 21.63, "learning_rate": 4.2464788732394366e-06, "loss": 0.3232, "step": 19293 }, { "epoch": 21.63, "learning_rate": 4.245070422535211e-06, "loss": 0.3624, "step": 19294 }, { "epoch": 21.63, "learning_rate": 4.243661971830986e-06, "loss": 0.2741, "step": 19295 }, { "epoch": 21.63, "learning_rate": 4.242253521126761e-06, "loss": 0.3425, "step": 19296 }, { "epoch": 21.63, "learning_rate": 4.240845070422535e-06, "loss": 0.3634, "step": 19297 }, { "epoch": 21.63, "learning_rate": 4.23943661971831e-06, "loss": 0.3613, "step": 19298 }, { "epoch": 21.64, "learning_rate": 4.238028169014084e-06, "loss": 0.3785, "step": 19299 }, { "epoch": 21.64, "learning_rate": 4.23661971830986e-06, "loss": 0.3739, "step": 19300 }, { "epoch": 21.64, "learning_rate": 4.235211267605634e-06, "loss": 0.2836, "step": 19301 }, { "epoch": 21.64, "learning_rate": 4.233802816901408e-06, "loss": 0.3037, "step": 19302 }, { "epoch": 21.64, "learning_rate": 4.232394366197183e-06, "loss": 0.3593, "step": 19303 }, { "epoch": 21.64, "learning_rate": 4.230985915492958e-06, "loss": 0.386, "step": 19304 }, { "epoch": 21.64, "learning_rate": 4.229577464788732e-06, "loss": 0.2959, "step": 19305 }, { "epoch": 21.64, "learning_rate": 4.228169014084507e-06, "loss": 0.2995, "step": 19306 }, { "epoch": 21.64, "learning_rate": 4.2267605633802815e-06, "loss": 0.3527, "step": 19307 }, { "epoch": 21.65, "learning_rate": 4.225352112676057e-06, "loss": 0.3113, "step": 19308 }, { "epoch": 21.65, "learning_rate": 4.223943661971831e-06, "loss": 0.3618, "step": 19309 }, { "epoch": 21.65, "learning_rate": 4.222535211267605e-06, "loss": 0.2427, "step": 19310 }, { "epoch": 21.65, "learning_rate": 4.221126760563381e-06, "loss": 0.4039, "step": 19311 }, { "epoch": 21.65, "learning_rate": 4.2197183098591555e-06, "loss": 0.3271, "step": 19312 }, { "epoch": 21.65, "learning_rate": 4.218309859154929e-06, "loss": 0.3284, "step": 19313 }, { "epoch": 21.65, "learning_rate": 4.216901408450704e-06, "loss": 0.3591, "step": 19314 }, { "epoch": 21.65, "learning_rate": 4.2154929577464794e-06, "loss": 0.3713, "step": 19315 }, { "epoch": 21.65, "learning_rate": 4.214084507042254e-06, "loss": 0.2726, "step": 19316 }, { "epoch": 21.66, "learning_rate": 4.212676056338028e-06, "loss": 0.3367, "step": 19317 }, { "epoch": 21.66, "learning_rate": 4.2112676056338025e-06, "loss": 0.3615, "step": 19318 }, { "epoch": 21.66, "learning_rate": 4.209859154929578e-06, "loss": 0.3203, "step": 19319 }, { "epoch": 21.66, "learning_rate": 4.208450704225353e-06, "loss": 0.2585, "step": 19320 }, { "epoch": 21.66, "learning_rate": 4.207042253521126e-06, "loss": 0.3076, "step": 19321 }, { "epoch": 21.66, "learning_rate": 4.205633802816901e-06, "loss": 0.3557, "step": 19322 }, { "epoch": 21.66, "learning_rate": 4.2042253521126765e-06, "loss": 0.3222, "step": 19323 }, { "epoch": 21.66, "learning_rate": 4.202816901408451e-06, "loss": 0.3079, "step": 19324 }, { "epoch": 21.66, "learning_rate": 4.201408450704225e-06, "loss": 0.4152, "step": 19325 }, { "epoch": 21.67, "learning_rate": 4.2000000000000004e-06, "loss": 0.3256, "step": 19326 }, { "epoch": 21.67, "learning_rate": 4.198591549295775e-06, "loss": 0.286, "step": 19327 }, { "epoch": 21.67, "learning_rate": 4.19718309859155e-06, "loss": 0.2672, "step": 19328 }, { "epoch": 21.67, "learning_rate": 4.1957746478873235e-06, "loss": 0.2283, "step": 19329 }, { "epoch": 21.67, "learning_rate": 4.194366197183099e-06, "loss": 0.3509, "step": 19330 }, { "epoch": 21.67, "learning_rate": 4.192957746478874e-06, "loss": 0.278, "step": 19331 }, { "epoch": 21.67, "learning_rate": 4.191549295774648e-06, "loss": 0.198, "step": 19332 }, { "epoch": 21.67, "learning_rate": 4.190140845070422e-06, "loss": 0.6177, "step": 19333 }, { "epoch": 21.67, "learning_rate": 4.1887323943661975e-06, "loss": 0.5155, "step": 19334 }, { "epoch": 21.68, "learning_rate": 4.187323943661972e-06, "loss": 0.3494, "step": 19335 }, { "epoch": 21.68, "learning_rate": 4.185915492957747e-06, "loss": 0.3966, "step": 19336 }, { "epoch": 21.68, "learning_rate": 4.184507042253521e-06, "loss": 0.374, "step": 19337 }, { "epoch": 21.68, "learning_rate": 4.183098591549296e-06, "loss": 0.3769, "step": 19338 }, { "epoch": 21.68, "learning_rate": 4.181690140845071e-06, "loss": 0.2853, "step": 19339 }, { "epoch": 21.68, "learning_rate": 4.180281690140845e-06, "loss": 0.3226, "step": 19340 }, { "epoch": 21.68, "learning_rate": 4.178873239436619e-06, "loss": 0.3096, "step": 19341 }, { "epoch": 21.68, "learning_rate": 4.177464788732395e-06, "loss": 0.3568, "step": 19342 }, { "epoch": 21.68, "learning_rate": 4.176056338028169e-06, "loss": 0.304, "step": 19343 }, { "epoch": 21.69, "learning_rate": 4.174647887323944e-06, "loss": 0.3359, "step": 19344 }, { "epoch": 21.69, "learning_rate": 4.1732394366197186e-06, "loss": 0.3176, "step": 19345 }, { "epoch": 21.69, "learning_rate": 4.171830985915493e-06, "loss": 0.323, "step": 19346 }, { "epoch": 21.69, "learning_rate": 4.170422535211268e-06, "loss": 0.3406, "step": 19347 }, { "epoch": 21.69, "learning_rate": 4.1690140845070425e-06, "loss": 0.3285, "step": 19348 }, { "epoch": 21.69, "learning_rate": 4.167605633802817e-06, "loss": 0.3221, "step": 19349 }, { "epoch": 21.69, "learning_rate": 4.166197183098592e-06, "loss": 0.316, "step": 19350 }, { "epoch": 21.69, "learning_rate": 4.164788732394366e-06, "loss": 0.3985, "step": 19351 }, { "epoch": 21.7, "learning_rate": 4.163380281690141e-06, "loss": 0.4006, "step": 19352 }, { "epoch": 21.7, "learning_rate": 4.161971830985916e-06, "loss": 0.4051, "step": 19353 }, { "epoch": 21.7, "learning_rate": 4.16056338028169e-06, "loss": 0.274, "step": 19354 }, { "epoch": 21.7, "learning_rate": 4.159154929577465e-06, "loss": 0.3173, "step": 19355 }, { "epoch": 21.7, "learning_rate": 4.1577464788732396e-06, "loss": 0.2424, "step": 19356 }, { "epoch": 21.7, "learning_rate": 4.156338028169014e-06, "loss": 0.3151, "step": 19357 }, { "epoch": 21.7, "learning_rate": 4.154929577464789e-06, "loss": 0.2691, "step": 19358 }, { "epoch": 21.7, "learning_rate": 4.1535211267605635e-06, "loss": 0.3119, "step": 19359 }, { "epoch": 21.7, "learning_rate": 4.152112676056338e-06, "loss": 0.2242, "step": 19360 }, { "epoch": 21.71, "learning_rate": 4.150704225352113e-06, "loss": 0.3147, "step": 19361 }, { "epoch": 21.71, "learning_rate": 4.149295774647887e-06, "loss": 0.3748, "step": 19362 }, { "epoch": 21.71, "learning_rate": 4.147887323943662e-06, "loss": 0.3365, "step": 19363 }, { "epoch": 21.71, "learning_rate": 4.146478873239437e-06, "loss": 0.3225, "step": 19364 }, { "epoch": 21.71, "learning_rate": 4.145070422535211e-06, "loss": 0.3652, "step": 19365 }, { "epoch": 21.71, "learning_rate": 4.143661971830986e-06, "loss": 0.3049, "step": 19366 }, { "epoch": 21.71, "learning_rate": 4.142253521126761e-06, "loss": 0.3147, "step": 19367 }, { "epoch": 21.71, "learning_rate": 4.140845070422535e-06, "loss": 0.3188, "step": 19368 }, { "epoch": 21.71, "learning_rate": 4.13943661971831e-06, "loss": 0.3202, "step": 19369 }, { "epoch": 21.72, "learning_rate": 4.1380281690140845e-06, "loss": 0.308, "step": 19370 }, { "epoch": 21.72, "learning_rate": 4.136619718309859e-06, "loss": 0.2889, "step": 19371 }, { "epoch": 21.72, "learning_rate": 4.135211267605634e-06, "loss": 0.3771, "step": 19372 }, { "epoch": 21.72, "learning_rate": 4.133802816901408e-06, "loss": 0.2291, "step": 19373 }, { "epoch": 21.72, "learning_rate": 4.132394366197183e-06, "loss": 0.3252, "step": 19374 }, { "epoch": 21.72, "learning_rate": 4.130985915492958e-06, "loss": 0.3205, "step": 19375 }, { "epoch": 21.72, "learning_rate": 4.129577464788732e-06, "loss": 0.2402, "step": 19376 }, { "epoch": 21.72, "learning_rate": 4.128169014084507e-06, "loss": 0.3064, "step": 19377 }, { "epoch": 21.72, "learning_rate": 4.126760563380282e-06, "loss": 0.2226, "step": 19378 }, { "epoch": 21.73, "learning_rate": 4.125352112676057e-06, "loss": 0.283, "step": 19379 }, { "epoch": 21.73, "learning_rate": 4.123943661971831e-06, "loss": 0.3388, "step": 19380 }, { "epoch": 21.73, "learning_rate": 4.1225352112676055e-06, "loss": 0.329, "step": 19381 }, { "epoch": 21.73, "learning_rate": 4.12112676056338e-06, "loss": 0.196, "step": 19382 }, { "epoch": 21.73, "learning_rate": 4.119718309859156e-06, "loss": 0.4002, "step": 19383 }, { "epoch": 21.73, "learning_rate": 4.118309859154929e-06, "loss": 0.331, "step": 19384 }, { "epoch": 21.73, "learning_rate": 4.116901408450704e-06, "loss": 0.3407, "step": 19385 }, { "epoch": 21.73, "learning_rate": 4.115492957746479e-06, "loss": 0.3577, "step": 19386 }, { "epoch": 21.73, "learning_rate": 4.114084507042254e-06, "loss": 0.4759, "step": 19387 }, { "epoch": 21.74, "learning_rate": 4.112676056338028e-06, "loss": 0.417, "step": 19388 }, { "epoch": 21.74, "learning_rate": 4.111267605633803e-06, "loss": 0.3591, "step": 19389 }, { "epoch": 21.74, "learning_rate": 4.109859154929577e-06, "loss": 0.2986, "step": 19390 }, { "epoch": 21.74, "learning_rate": 4.108450704225353e-06, "loss": 0.3754, "step": 19391 }, { "epoch": 21.74, "learning_rate": 4.1070422535211265e-06, "loss": 0.3334, "step": 19392 }, { "epoch": 21.74, "learning_rate": 4.105633802816901e-06, "loss": 0.3333, "step": 19393 }, { "epoch": 21.74, "learning_rate": 4.104225352112677e-06, "loss": 0.3904, "step": 19394 }, { "epoch": 21.74, "learning_rate": 4.102816901408451e-06, "loss": 0.3543, "step": 19395 }, { "epoch": 21.74, "learning_rate": 4.101408450704225e-06, "loss": 0.3903, "step": 19396 }, { "epoch": 21.75, "learning_rate": 4.1e-06, "loss": 0.3715, "step": 19397 }, { "epoch": 21.75, "learning_rate": 4.098591549295775e-06, "loss": 0.3338, "step": 19398 }, { "epoch": 21.75, "learning_rate": 4.09718309859155e-06, "loss": 0.317, "step": 19399 }, { "epoch": 21.75, "learning_rate": 4.095774647887324e-06, "loss": 0.3259, "step": 19400 }, { "epoch": 21.75, "learning_rate": 4.094366197183098e-06, "loss": 0.2978, "step": 19401 }, { "epoch": 21.75, "learning_rate": 4.092957746478874e-06, "loss": 0.3462, "step": 19402 }, { "epoch": 21.75, "learning_rate": 4.091549295774648e-06, "loss": 0.3777, "step": 19403 }, { "epoch": 21.75, "learning_rate": 4.090140845070422e-06, "loss": 0.2942, "step": 19404 }, { "epoch": 21.75, "learning_rate": 4.088732394366197e-06, "loss": 0.2922, "step": 19405 }, { "epoch": 21.76, "learning_rate": 4.087323943661972e-06, "loss": 0.2566, "step": 19406 }, { "epoch": 21.76, "learning_rate": 4.085915492957747e-06, "loss": 0.2868, "step": 19407 }, { "epoch": 21.76, "learning_rate": 4.084507042253521e-06, "loss": 0.2784, "step": 19408 }, { "epoch": 21.76, "learning_rate": 4.083098591549295e-06, "loss": 0.4295, "step": 19409 }, { "epoch": 21.76, "learning_rate": 4.081690140845071e-06, "loss": 0.3313, "step": 19410 }, { "epoch": 21.76, "learning_rate": 4.0802816901408455e-06, "loss": 0.3727, "step": 19411 }, { "epoch": 21.76, "learning_rate": 4.078873239436619e-06, "loss": 0.3581, "step": 19412 }, { "epoch": 21.76, "learning_rate": 4.077464788732395e-06, "loss": 0.3345, "step": 19413 }, { "epoch": 21.76, "learning_rate": 4.076056338028169e-06, "loss": 0.3159, "step": 19414 }, { "epoch": 21.77, "learning_rate": 4.074647887323944e-06, "loss": 0.2554, "step": 19415 }, { "epoch": 21.77, "learning_rate": 4.073239436619718e-06, "loss": 0.3156, "step": 19416 }, { "epoch": 21.77, "learning_rate": 4.071830985915493e-06, "loss": 0.2811, "step": 19417 }, { "epoch": 21.77, "learning_rate": 4.070422535211268e-06, "loss": 0.3371, "step": 19418 }, { "epoch": 21.77, "learning_rate": 4.069014084507043e-06, "loss": 0.2524, "step": 19419 }, { "epoch": 21.77, "learning_rate": 4.067605633802816e-06, "loss": 0.3098, "step": 19420 }, { "epoch": 21.77, "learning_rate": 4.066197183098592e-06, "loss": 0.3299, "step": 19421 }, { "epoch": 21.77, "learning_rate": 4.0647887323943665e-06, "loss": 0.3036, "step": 19422 }, { "epoch": 21.77, "learning_rate": 4.063380281690141e-06, "loss": 0.3438, "step": 19423 }, { "epoch": 21.78, "learning_rate": 4.061971830985915e-06, "loss": 0.2518, "step": 19424 }, { "epoch": 21.78, "learning_rate": 4.06056338028169e-06, "loss": 0.3405, "step": 19425 }, { "epoch": 21.78, "learning_rate": 4.059154929577465e-06, "loss": 0.3175, "step": 19426 }, { "epoch": 21.78, "learning_rate": 4.05774647887324e-06, "loss": 0.2678, "step": 19427 }, { "epoch": 21.78, "learning_rate": 4.0563380281690135e-06, "loss": 0.3435, "step": 19428 }, { "epoch": 21.78, "learning_rate": 4.054929577464789e-06, "loss": 0.2546, "step": 19429 }, { "epoch": 21.78, "learning_rate": 4.053521126760564e-06, "loss": 0.2358, "step": 19430 }, { "epoch": 21.78, "learning_rate": 4.052112676056338e-06, "loss": 0.2502, "step": 19431 }, { "epoch": 21.78, "learning_rate": 4.050704225352113e-06, "loss": 0.2633, "step": 19432 }, { "epoch": 21.79, "learning_rate": 4.0492957746478875e-06, "loss": 0.449, "step": 19433 }, { "epoch": 21.79, "learning_rate": 4.047887323943662e-06, "loss": 0.3981, "step": 19434 }, { "epoch": 21.79, "learning_rate": 4.046478873239437e-06, "loss": 0.4473, "step": 19435 }, { "epoch": 21.79, "learning_rate": 4.045070422535211e-06, "loss": 0.3274, "step": 19436 }, { "epoch": 21.79, "learning_rate": 4.043661971830986e-06, "loss": 0.3502, "step": 19437 }, { "epoch": 21.79, "learning_rate": 4.042253521126761e-06, "loss": 0.4306, "step": 19438 }, { "epoch": 21.79, "learning_rate": 4.040845070422535e-06, "loss": 0.4246, "step": 19439 }, { "epoch": 21.79, "learning_rate": 4.03943661971831e-06, "loss": 0.3482, "step": 19440 }, { "epoch": 21.79, "learning_rate": 4.038028169014085e-06, "loss": 0.3055, "step": 19441 }, { "epoch": 21.8, "learning_rate": 4.036619718309859e-06, "loss": 0.3172, "step": 19442 }, { "epoch": 21.8, "learning_rate": 4.035211267605634e-06, "loss": 0.2924, "step": 19443 }, { "epoch": 21.8, "learning_rate": 4.0338028169014085e-06, "loss": 0.3068, "step": 19444 }, { "epoch": 21.8, "learning_rate": 4.032394366197183e-06, "loss": 0.3862, "step": 19445 }, { "epoch": 21.8, "learning_rate": 4.030985915492958e-06, "loss": 0.2983, "step": 19446 }, { "epoch": 21.8, "learning_rate": 4.029577464788733e-06, "loss": 0.2852, "step": 19447 }, { "epoch": 21.8, "learning_rate": 4.028169014084507e-06, "loss": 0.3633, "step": 19448 }, { "epoch": 21.8, "learning_rate": 4.026760563380282e-06, "loss": 0.3337, "step": 19449 }, { "epoch": 21.8, "learning_rate": 4.025352112676056e-06, "loss": 0.2707, "step": 19450 }, { "epoch": 21.81, "learning_rate": 4.023943661971832e-06, "loss": 0.3288, "step": 19451 }, { "epoch": 21.81, "learning_rate": 4.022535211267606e-06, "loss": 0.2291, "step": 19452 }, { "epoch": 21.81, "learning_rate": 4.02112676056338e-06, "loss": 0.349, "step": 19453 }, { "epoch": 21.81, "learning_rate": 4.019718309859155e-06, "loss": 0.3408, "step": 19454 }, { "epoch": 21.81, "learning_rate": 4.01830985915493e-06, "loss": 0.282, "step": 19455 }, { "epoch": 21.81, "learning_rate": 4.016901408450704e-06, "loss": 0.2705, "step": 19456 }, { "epoch": 21.81, "learning_rate": 4.015492957746479e-06, "loss": 0.3634, "step": 19457 }, { "epoch": 21.81, "learning_rate": 4.0140845070422534e-06, "loss": 0.329, "step": 19458 }, { "epoch": 21.82, "learning_rate": 4.012676056338029e-06, "loss": 0.2795, "step": 19459 }, { "epoch": 21.82, "learning_rate": 4.011267605633803e-06, "loss": 0.339, "step": 19460 }, { "epoch": 21.82, "learning_rate": 4.009859154929577e-06, "loss": 0.2737, "step": 19461 }, { "epoch": 21.82, "learning_rate": 4.008450704225352e-06, "loss": 0.2527, "step": 19462 }, { "epoch": 21.82, "learning_rate": 4.0070422535211275e-06, "loss": 0.2669, "step": 19463 }, { "epoch": 21.82, "learning_rate": 4.005633802816901e-06, "loss": 0.4024, "step": 19464 }, { "epoch": 21.82, "learning_rate": 4.004225352112676e-06, "loss": 0.2709, "step": 19465 }, { "epoch": 21.82, "learning_rate": 4.002816901408451e-06, "loss": 0.2598, "step": 19466 }, { "epoch": 21.82, "learning_rate": 4.001408450704226e-06, "loss": 0.2812, "step": 19467 }, { "epoch": 21.83, "learning_rate": 4e-06, "loss": 0.3943, "step": 19468 }, { "epoch": 21.83, "learning_rate": 3.9985915492957745e-06, "loss": 0.319, "step": 19469 }, { "epoch": 21.83, "learning_rate": 3.99718309859155e-06, "loss": 0.2945, "step": 19470 }, { "epoch": 21.83, "learning_rate": 3.995774647887325e-06, "loss": 0.3166, "step": 19471 }, { "epoch": 21.83, "learning_rate": 3.994366197183098e-06, "loss": 0.3273, "step": 19472 }, { "epoch": 21.83, "learning_rate": 3.992957746478873e-06, "loss": 0.2554, "step": 19473 }, { "epoch": 21.83, "learning_rate": 3.9915492957746485e-06, "loss": 0.2586, "step": 19474 }, { "epoch": 21.83, "learning_rate": 3.990140845070422e-06, "loss": 0.312, "step": 19475 }, { "epoch": 21.83, "learning_rate": 3.988732394366197e-06, "loss": 0.3558, "step": 19476 }, { "epoch": 21.84, "learning_rate": 3.9873239436619716e-06, "loss": 0.2448, "step": 19477 }, { "epoch": 21.84, "learning_rate": 3.985915492957747e-06, "loss": 0.404, "step": 19478 }, { "epoch": 21.84, "learning_rate": 3.984507042253521e-06, "loss": 0.1636, "step": 19479 }, { "epoch": 21.84, "learning_rate": 3.9830985915492955e-06, "loss": 0.4691, "step": 19480 }, { "epoch": 21.84, "learning_rate": 3.981690140845071e-06, "loss": 0.17, "step": 19481 }, { "epoch": 21.84, "learning_rate": 3.980281690140846e-06, "loss": 0.2434, "step": 19482 }, { "epoch": 21.84, "learning_rate": 3.978873239436619e-06, "loss": 0.4308, "step": 19483 }, { "epoch": 21.84, "learning_rate": 3.977464788732394e-06, "loss": 0.3822, "step": 19484 }, { "epoch": 21.84, "learning_rate": 3.9760563380281695e-06, "loss": 0.461, "step": 19485 }, { "epoch": 21.85, "learning_rate": 3.974647887323944e-06, "loss": 0.4806, "step": 19486 }, { "epoch": 21.85, "learning_rate": 3.973239436619718e-06, "loss": 0.3234, "step": 19487 }, { "epoch": 21.85, "learning_rate": 3.9718309859154926e-06, "loss": 0.3754, "step": 19488 }, { "epoch": 21.85, "learning_rate": 3.970422535211268e-06, "loss": 0.42, "step": 19489 }, { "epoch": 21.85, "learning_rate": 3.969014084507043e-06, "loss": 0.3525, "step": 19490 }, { "epoch": 21.85, "learning_rate": 3.9676056338028165e-06, "loss": 0.3247, "step": 19491 }, { "epoch": 21.85, "learning_rate": 3.966197183098591e-06, "loss": 0.3801, "step": 19492 }, { "epoch": 21.85, "learning_rate": 3.964788732394367e-06, "loss": 0.383, "step": 19493 }, { "epoch": 21.85, "learning_rate": 3.963380281690141e-06, "loss": 0.3752, "step": 19494 }, { "epoch": 21.86, "learning_rate": 3.961971830985915e-06, "loss": 0.3369, "step": 19495 }, { "epoch": 21.86, "learning_rate": 3.96056338028169e-06, "loss": 0.3389, "step": 19496 }, { "epoch": 21.86, "learning_rate": 3.959154929577465e-06, "loss": 0.3834, "step": 19497 }, { "epoch": 21.86, "learning_rate": 3.95774647887324e-06, "loss": 0.3106, "step": 19498 }, { "epoch": 21.86, "learning_rate": 3.956338028169014e-06, "loss": 0.2867, "step": 19499 }, { "epoch": 21.86, "learning_rate": 3.954929577464789e-06, "loss": 0.2903, "step": 19500 }, { "epoch": 21.86, "eval_loss": 0.7210519909858704, "eval_runtime": 818.6541, "eval_samples_per_second": 3.227, "eval_steps_per_second": 0.404, "eval_wer": 0.13006743355811187, "step": 19500 }, { "epoch": 21.86, "learning_rate": 3.953521126760564e-06, "loss": 0.4563, "step": 19501 }, { "epoch": 21.86, "learning_rate": 3.952112676056338e-06, "loss": 0.3543, "step": 19502 }, { "epoch": 21.86, "learning_rate": 3.950704225352112e-06, "loss": 0.3171, "step": 19503 }, { "epoch": 21.87, "learning_rate": 3.949295774647888e-06, "loss": 0.2366, "step": 19504 }, { "epoch": 21.87, "learning_rate": 3.947887323943662e-06, "loss": 0.3632, "step": 19505 }, { "epoch": 21.87, "learning_rate": 3.946478873239437e-06, "loss": 0.3258, "step": 19506 }, { "epoch": 21.87, "learning_rate": 3.945070422535211e-06, "loss": 0.3466, "step": 19507 }, { "epoch": 21.87, "learning_rate": 3.943661971830986e-06, "loss": 0.2746, "step": 19508 }, { "epoch": 21.87, "learning_rate": 3.942253521126761e-06, "loss": 0.3028, "step": 19509 }, { "epoch": 21.87, "learning_rate": 3.9408450704225354e-06, "loss": 0.272, "step": 19510 }, { "epoch": 21.87, "learning_rate": 3.939436619718309e-06, "loss": 0.3179, "step": 19511 }, { "epoch": 21.87, "learning_rate": 3.938028169014085e-06, "loss": 0.3157, "step": 19512 }, { "epoch": 21.88, "learning_rate": 3.936619718309859e-06, "loss": 0.3128, "step": 19513 }, { "epoch": 21.88, "learning_rate": 3.935211267605634e-06, "loss": 0.3216, "step": 19514 }, { "epoch": 21.88, "learning_rate": 3.933802816901408e-06, "loss": 0.3381, "step": 19515 }, { "epoch": 21.88, "learning_rate": 3.932394366197183e-06, "loss": 0.4582, "step": 19516 }, { "epoch": 21.88, "learning_rate": 3.930985915492958e-06, "loss": 0.339, "step": 19517 }, { "epoch": 21.88, "learning_rate": 3.9295774647887325e-06, "loss": 0.3509, "step": 19518 }, { "epoch": 21.88, "learning_rate": 3.928169014084507e-06, "loss": 0.2759, "step": 19519 }, { "epoch": 21.88, "learning_rate": 3.926760563380282e-06, "loss": 0.2929, "step": 19520 }, { "epoch": 21.88, "learning_rate": 3.9253521126760565e-06, "loss": 0.3164, "step": 19521 }, { "epoch": 21.89, "learning_rate": 3.923943661971831e-06, "loss": 0.3163, "step": 19522 }, { "epoch": 21.89, "learning_rate": 3.922535211267606e-06, "loss": 0.3527, "step": 19523 }, { "epoch": 21.89, "learning_rate": 3.92112676056338e-06, "loss": 0.2866, "step": 19524 }, { "epoch": 21.89, "learning_rate": 3.919718309859155e-06, "loss": 0.2822, "step": 19525 }, { "epoch": 21.89, "learning_rate": 3.91830985915493e-06, "loss": 0.2423, "step": 19526 }, { "epoch": 21.89, "learning_rate": 3.916901408450704e-06, "loss": 0.3129, "step": 19527 }, { "epoch": 21.89, "learning_rate": 3.915492957746479e-06, "loss": 0.283, "step": 19528 }, { "epoch": 21.89, "learning_rate": 3.9140845070422536e-06, "loss": 0.3466, "step": 19529 }, { "epoch": 21.89, "learning_rate": 3.912676056338028e-06, "loss": 0.3711, "step": 19530 }, { "epoch": 21.9, "learning_rate": 3.911267605633803e-06, "loss": 0.3508, "step": 19531 }, { "epoch": 21.9, "learning_rate": 3.9098591549295775e-06, "loss": 0.3537, "step": 19532 }, { "epoch": 21.9, "learning_rate": 3.908450704225352e-06, "loss": 0.4408, "step": 19533 }, { "epoch": 21.9, "learning_rate": 3.907042253521128e-06, "loss": 0.3421, "step": 19534 }, { "epoch": 21.9, "learning_rate": 3.905633802816901e-06, "loss": 0.3926, "step": 19535 }, { "epoch": 21.9, "learning_rate": 3.904225352112676e-06, "loss": 0.3669, "step": 19536 }, { "epoch": 21.9, "learning_rate": 3.902816901408451e-06, "loss": 0.3342, "step": 19537 }, { "epoch": 21.9, "learning_rate": 3.901408450704226e-06, "loss": 0.3074, "step": 19538 }, { "epoch": 21.9, "learning_rate": 3.9e-06, "loss": 0.3787, "step": 19539 }, { "epoch": 21.91, "learning_rate": 3.8985915492957746e-06, "loss": 0.3769, "step": 19540 }, { "epoch": 21.91, "learning_rate": 3.897183098591549e-06, "loss": 0.3803, "step": 19541 }, { "epoch": 21.91, "learning_rate": 3.895774647887325e-06, "loss": 0.4282, "step": 19542 }, { "epoch": 21.91, "learning_rate": 3.8943661971830985e-06, "loss": 0.2767, "step": 19543 }, { "epoch": 21.91, "learning_rate": 3.892957746478873e-06, "loss": 0.2794, "step": 19544 }, { "epoch": 21.91, "learning_rate": 3.891549295774648e-06, "loss": 0.3648, "step": 19545 }, { "epoch": 21.91, "learning_rate": 3.890140845070423e-06, "loss": 0.2579, "step": 19546 }, { "epoch": 21.91, "learning_rate": 3.888732394366197e-06, "loss": 0.397, "step": 19547 }, { "epoch": 21.91, "learning_rate": 3.887323943661972e-06, "loss": 0.404, "step": 19548 }, { "epoch": 21.92, "learning_rate": 3.885915492957746e-06, "loss": 0.3193, "step": 19549 }, { "epoch": 21.92, "learning_rate": 3.884507042253522e-06, "loss": 0.3463, "step": 19550 }, { "epoch": 21.92, "learning_rate": 3.883098591549296e-06, "loss": 0.2822, "step": 19551 }, { "epoch": 21.92, "learning_rate": 3.88169014084507e-06, "loss": 0.3832, "step": 19552 }, { "epoch": 21.92, "learning_rate": 3.880281690140846e-06, "loss": 0.3427, "step": 19553 }, { "epoch": 21.92, "learning_rate": 3.87887323943662e-06, "loss": 0.2961, "step": 19554 }, { "epoch": 21.92, "learning_rate": 3.877464788732394e-06, "loss": 0.3602, "step": 19555 }, { "epoch": 21.92, "learning_rate": 3.876056338028169e-06, "loss": 0.3182, "step": 19556 }, { "epoch": 21.92, "learning_rate": 3.874647887323944e-06, "loss": 0.4047, "step": 19557 }, { "epoch": 21.93, "learning_rate": 3.873239436619719e-06, "loss": 0.2629, "step": 19558 }, { "epoch": 21.93, "learning_rate": 3.871830985915493e-06, "loss": 0.2877, "step": 19559 }, { "epoch": 21.93, "learning_rate": 3.870422535211267e-06, "loss": 0.2753, "step": 19560 }, { "epoch": 21.93, "learning_rate": 3.869014084507043e-06, "loss": 0.3536, "step": 19561 }, { "epoch": 21.93, "learning_rate": 3.8676056338028174e-06, "loss": 0.3368, "step": 19562 }, { "epoch": 21.93, "learning_rate": 3.866197183098591e-06, "loss": 0.3706, "step": 19563 }, { "epoch": 21.93, "learning_rate": 3.864788732394366e-06, "loss": 0.3274, "step": 19564 }, { "epoch": 21.93, "learning_rate": 3.863380281690141e-06, "loss": 0.3671, "step": 19565 }, { "epoch": 21.93, "learning_rate": 3.861971830985916e-06, "loss": 0.3456, "step": 19566 }, { "epoch": 21.94, "learning_rate": 3.86056338028169e-06, "loss": 0.3852, "step": 19567 }, { "epoch": 21.94, "learning_rate": 3.859154929577465e-06, "loss": 0.3757, "step": 19568 }, { "epoch": 21.94, "learning_rate": 3.85774647887324e-06, "loss": 0.2958, "step": 19569 }, { "epoch": 21.94, "learning_rate": 3.8563380281690145e-06, "loss": 0.3383, "step": 19570 }, { "epoch": 21.94, "learning_rate": 3.854929577464788e-06, "loss": 0.3639, "step": 19571 }, { "epoch": 21.94, "learning_rate": 3.853521126760564e-06, "loss": 0.2735, "step": 19572 }, { "epoch": 21.94, "learning_rate": 3.8521126760563385e-06, "loss": 0.2558, "step": 19573 }, { "epoch": 21.94, "learning_rate": 3.850704225352113e-06, "loss": 0.3339, "step": 19574 }, { "epoch": 21.95, "learning_rate": 3.849295774647887e-06, "loss": 0.293, "step": 19575 }, { "epoch": 21.95, "learning_rate": 3.847887323943662e-06, "loss": 0.358, "step": 19576 }, { "epoch": 21.95, "learning_rate": 3.846478873239437e-06, "loss": 0.3605, "step": 19577 }, { "epoch": 21.95, "learning_rate": 3.845070422535212e-06, "loss": 0.2658, "step": 19578 }, { "epoch": 21.95, "learning_rate": 3.8436619718309854e-06, "loss": 0.3802, "step": 19579 }, { "epoch": 21.95, "learning_rate": 3.842253521126761e-06, "loss": 0.2888, "step": 19580 }, { "epoch": 21.95, "learning_rate": 3.8408450704225356e-06, "loss": 0.3239, "step": 19581 }, { "epoch": 21.95, "learning_rate": 3.83943661971831e-06, "loss": 0.2746, "step": 19582 }, { "epoch": 21.95, "learning_rate": 3.838028169014084e-06, "loss": 0.5299, "step": 19583 }, { "epoch": 21.96, "learning_rate": 3.8366197183098595e-06, "loss": 0.5126, "step": 19584 }, { "epoch": 21.96, "learning_rate": 3.835211267605634e-06, "loss": 0.3973, "step": 19585 }, { "epoch": 21.96, "learning_rate": 3.833802816901409e-06, "loss": 0.3695, "step": 19586 }, { "epoch": 21.96, "learning_rate": 3.832394366197183e-06, "loss": 0.403, "step": 19587 }, { "epoch": 21.96, "learning_rate": 3.830985915492958e-06, "loss": 0.3509, "step": 19588 }, { "epoch": 21.96, "learning_rate": 3.829577464788733e-06, "loss": 0.3128, "step": 19589 }, { "epoch": 21.96, "learning_rate": 3.828169014084507e-06, "loss": 0.3305, "step": 19590 }, { "epoch": 21.96, "learning_rate": 3.826760563380282e-06, "loss": 0.3771, "step": 19591 }, { "epoch": 21.96, "learning_rate": 3.8253521126760566e-06, "loss": 0.2939, "step": 19592 }, { "epoch": 21.97, "learning_rate": 3.823943661971831e-06, "loss": 0.3526, "step": 19593 }, { "epoch": 21.97, "learning_rate": 3.822535211267605e-06, "loss": 0.3627, "step": 19594 }, { "epoch": 21.97, "learning_rate": 3.8211267605633805e-06, "loss": 0.329, "step": 19595 }, { "epoch": 21.97, "learning_rate": 3.819718309859155e-06, "loss": 0.2854, "step": 19596 }, { "epoch": 21.97, "learning_rate": 3.81830985915493e-06, "loss": 0.3465, "step": 19597 }, { "epoch": 21.97, "learning_rate": 3.8169014084507035e-06, "loss": 0.3356, "step": 19598 }, { "epoch": 21.97, "learning_rate": 3.815492957746479e-06, "loss": 0.3327, "step": 19599 }, { "epoch": 21.97, "learning_rate": 3.8140845070422537e-06, "loss": 0.376, "step": 19600 }, { "epoch": 21.97, "learning_rate": 3.812676056338028e-06, "loss": 0.2707, "step": 19601 }, { "epoch": 21.98, "learning_rate": 3.8112676056338025e-06, "loss": 0.2608, "step": 19602 }, { "epoch": 21.98, "learning_rate": 3.8098591549295776e-06, "loss": 0.3428, "step": 19603 }, { "epoch": 21.98, "learning_rate": 3.8084507042253522e-06, "loss": 0.3494, "step": 19604 }, { "epoch": 21.98, "learning_rate": 3.8070422535211264e-06, "loss": 0.3303, "step": 19605 }, { "epoch": 21.98, "learning_rate": 3.805633802816902e-06, "loss": 0.3415, "step": 19606 }, { "epoch": 21.98, "learning_rate": 3.804225352112676e-06, "loss": 0.2834, "step": 19607 }, { "epoch": 21.98, "learning_rate": 3.8028169014084508e-06, "loss": 0.3615, "step": 19608 }, { "epoch": 21.98, "learning_rate": 3.801408450704225e-06, "loss": 0.2579, "step": 19609 }, { "epoch": 21.98, "learning_rate": 3.8000000000000005e-06, "loss": 0.2354, "step": 19610 }, { "epoch": 21.99, "learning_rate": 3.7985915492957747e-06, "loss": 0.311, "step": 19611 }, { "epoch": 21.99, "learning_rate": 3.7971830985915493e-06, "loss": 0.3054, "step": 19612 }, { "epoch": 21.99, "learning_rate": 3.7957746478873235e-06, "loss": 0.2752, "step": 19613 }, { "epoch": 21.99, "learning_rate": 3.794366197183099e-06, "loss": 0.2819, "step": 19614 }, { "epoch": 21.99, "learning_rate": 3.7929577464788732e-06, "loss": 0.2283, "step": 19615 }, { "epoch": 21.99, "learning_rate": 3.791549295774648e-06, "loss": 0.2344, "step": 19616 }, { "epoch": 21.99, "learning_rate": 3.790140845070422e-06, "loss": 0.3731, "step": 19617 }, { "epoch": 21.99, "learning_rate": 3.7887323943661976e-06, "loss": 0.4467, "step": 19618 }, { "epoch": 21.99, "learning_rate": 3.7873239436619718e-06, "loss": 0.282, "step": 19619 }, { "epoch": 22.0, "learning_rate": 3.7859154929577464e-06, "loss": 0.2283, "step": 19620 }, { "epoch": 22.0, "learning_rate": 3.7845070422535215e-06, "loss": 0.3509, "step": 19621 }, { "epoch": 22.0, "learning_rate": 3.783098591549296e-06, "loss": 0.2533, "step": 19622 }, { "epoch": 22.0, "learning_rate": 3.7816901408450703e-06, "loss": 0.2823, "step": 19623 }, { "epoch": 22.0, "learning_rate": 3.780281690140845e-06, "loss": 0.394, "step": 19624 }, { "epoch": 22.0, "learning_rate": 3.77887323943662e-06, "loss": 0.4085, "step": 19625 }, { "epoch": 22.0, "learning_rate": 3.7774647887323947e-06, "loss": 0.4928, "step": 19626 }, { "epoch": 22.0, "learning_rate": 3.776056338028169e-06, "loss": 0.3344, "step": 19627 }, { "epoch": 22.0, "learning_rate": 3.7746478873239435e-06, "loss": 0.4594, "step": 19628 }, { "epoch": 22.01, "learning_rate": 3.7732394366197186e-06, "loss": 0.3343, "step": 19629 }, { "epoch": 22.01, "learning_rate": 3.7718309859154932e-06, "loss": 0.2545, "step": 19630 }, { "epoch": 22.01, "learning_rate": 3.7704225352112674e-06, "loss": 0.3375, "step": 19631 }, { "epoch": 22.01, "learning_rate": 3.769014084507042e-06, "loss": 0.3425, "step": 19632 }, { "epoch": 22.01, "learning_rate": 3.767605633802817e-06, "loss": 0.2706, "step": 19633 }, { "epoch": 22.01, "learning_rate": 3.7661971830985918e-06, "loss": 0.3496, "step": 19634 }, { "epoch": 22.01, "learning_rate": 3.764788732394366e-06, "loss": 0.3136, "step": 19635 }, { "epoch": 22.01, "learning_rate": 3.7633802816901406e-06, "loss": 0.3602, "step": 19636 }, { "epoch": 22.01, "learning_rate": 3.7619718309859157e-06, "loss": 0.3223, "step": 19637 }, { "epoch": 22.02, "learning_rate": 3.7605633802816903e-06, "loss": 0.4171, "step": 19638 }, { "epoch": 22.02, "learning_rate": 3.7591549295774645e-06, "loss": 0.3417, "step": 19639 }, { "epoch": 22.02, "learning_rate": 3.75774647887324e-06, "loss": 0.2779, "step": 19640 }, { "epoch": 22.02, "learning_rate": 3.7563380281690142e-06, "loss": 0.2524, "step": 19641 }, { "epoch": 22.02, "learning_rate": 3.754929577464789e-06, "loss": 0.2966, "step": 19642 }, { "epoch": 22.02, "learning_rate": 3.753521126760563e-06, "loss": 0.3593, "step": 19643 }, { "epoch": 22.02, "learning_rate": 3.7521126760563386e-06, "loss": 0.3811, "step": 19644 }, { "epoch": 22.02, "learning_rate": 3.7507042253521128e-06, "loss": 0.3214, "step": 19645 }, { "epoch": 22.02, "learning_rate": 3.7492957746478874e-06, "loss": 0.2511, "step": 19646 }, { "epoch": 22.03, "learning_rate": 3.747887323943662e-06, "loss": 0.311, "step": 19647 }, { "epoch": 22.03, "learning_rate": 3.7464788732394367e-06, "loss": 0.3888, "step": 19648 }, { "epoch": 22.03, "learning_rate": 3.7450704225352113e-06, "loss": 0.3082, "step": 19649 }, { "epoch": 22.03, "learning_rate": 3.743661971830986e-06, "loss": 0.3078, "step": 19650 }, { "epoch": 22.03, "learning_rate": 3.7422535211267606e-06, "loss": 0.3214, "step": 19651 }, { "epoch": 22.03, "learning_rate": 3.7408450704225352e-06, "loss": 0.269, "step": 19652 }, { "epoch": 22.03, "learning_rate": 3.73943661971831e-06, "loss": 0.3405, "step": 19653 }, { "epoch": 22.03, "learning_rate": 3.7380281690140845e-06, "loss": 0.295, "step": 19654 }, { "epoch": 22.03, "learning_rate": 3.736619718309859e-06, "loss": 0.3705, "step": 19655 }, { "epoch": 22.04, "learning_rate": 3.7352112676056342e-06, "loss": 0.3219, "step": 19656 }, { "epoch": 22.04, "learning_rate": 3.7338028169014084e-06, "loss": 0.2953, "step": 19657 }, { "epoch": 22.04, "learning_rate": 3.7323943661971835e-06, "loss": 0.2809, "step": 19658 }, { "epoch": 22.04, "learning_rate": 3.7309859154929577e-06, "loss": 0.2797, "step": 19659 }, { "epoch": 22.04, "learning_rate": 3.7295774647887328e-06, "loss": 0.2896, "step": 19660 }, { "epoch": 22.04, "learning_rate": 3.728169014084507e-06, "loss": 0.3052, "step": 19661 }, { "epoch": 22.04, "learning_rate": 3.726760563380282e-06, "loss": 0.223, "step": 19662 }, { "epoch": 22.04, "learning_rate": 3.7253521126760563e-06, "loss": 0.2359, "step": 19663 }, { "epoch": 22.04, "learning_rate": 3.7239436619718313e-06, "loss": 0.2951, "step": 19664 }, { "epoch": 22.05, "learning_rate": 3.7225352112676055e-06, "loss": 0.3071, "step": 19665 }, { "epoch": 22.05, "learning_rate": 3.7211267605633806e-06, "loss": 0.3094, "step": 19666 }, { "epoch": 22.05, "learning_rate": 3.719718309859155e-06, "loss": 0.3301, "step": 19667 }, { "epoch": 22.05, "learning_rate": 3.71830985915493e-06, "loss": 0.3356, "step": 19668 }, { "epoch": 22.05, "learning_rate": 3.716901408450704e-06, "loss": 0.275, "step": 19669 }, { "epoch": 22.05, "learning_rate": 3.715492957746479e-06, "loss": 0.2792, "step": 19670 }, { "epoch": 22.05, "learning_rate": 3.7140845070422534e-06, "loss": 0.1843, "step": 19671 }, { "epoch": 22.05, "learning_rate": 3.7126760563380284e-06, "loss": 0.2616, "step": 19672 }, { "epoch": 22.05, "learning_rate": 3.711267605633803e-06, "loss": 0.2689, "step": 19673 }, { "epoch": 22.06, "learning_rate": 3.7098591549295777e-06, "loss": 0.3828, "step": 19674 }, { "epoch": 22.06, "learning_rate": 3.7084507042253523e-06, "loss": 0.4298, "step": 19675 }, { "epoch": 22.06, "learning_rate": 3.707042253521127e-06, "loss": 0.3409, "step": 19676 }, { "epoch": 22.06, "learning_rate": 3.7056338028169016e-06, "loss": 0.3682, "step": 19677 }, { "epoch": 22.06, "learning_rate": 3.7042253521126762e-06, "loss": 0.3854, "step": 19678 }, { "epoch": 22.06, "learning_rate": 3.702816901408451e-06, "loss": 0.3169, "step": 19679 }, { "epoch": 22.06, "learning_rate": 3.7014084507042255e-06, "loss": 0.2975, "step": 19680 }, { "epoch": 22.06, "learning_rate": 3.7e-06, "loss": 0.3182, "step": 19681 }, { "epoch": 22.07, "learning_rate": 3.6985915492957748e-06, "loss": 0.2448, "step": 19682 }, { "epoch": 22.07, "learning_rate": 3.6971830985915494e-06, "loss": 0.2639, "step": 19683 }, { "epoch": 22.07, "learning_rate": 3.695774647887324e-06, "loss": 0.3477, "step": 19684 }, { "epoch": 22.07, "learning_rate": 3.6943661971830987e-06, "loss": 0.3672, "step": 19685 }, { "epoch": 22.07, "learning_rate": 3.6929577464788733e-06, "loss": 0.2876, "step": 19686 }, { "epoch": 22.07, "learning_rate": 3.691549295774648e-06, "loss": 0.3016, "step": 19687 }, { "epoch": 22.07, "learning_rate": 3.6901408450704226e-06, "loss": 0.2608, "step": 19688 }, { "epoch": 22.07, "learning_rate": 3.6887323943661973e-06, "loss": 0.252, "step": 19689 }, { "epoch": 22.07, "learning_rate": 3.6873239436619723e-06, "loss": 0.3409, "step": 19690 }, { "epoch": 22.08, "learning_rate": 3.6859154929577465e-06, "loss": 0.3643, "step": 19691 }, { "epoch": 22.08, "learning_rate": 3.6845070422535216e-06, "loss": 0.2689, "step": 19692 }, { "epoch": 22.08, "learning_rate": 3.683098591549296e-06, "loss": 0.425, "step": 19693 }, { "epoch": 22.08, "learning_rate": 3.681690140845071e-06, "loss": 0.3508, "step": 19694 }, { "epoch": 22.08, "learning_rate": 3.680281690140845e-06, "loss": 0.3109, "step": 19695 }, { "epoch": 22.08, "learning_rate": 3.67887323943662e-06, "loss": 0.3219, "step": 19696 }, { "epoch": 22.08, "learning_rate": 3.6774647887323944e-06, "loss": 0.3164, "step": 19697 }, { "epoch": 22.08, "learning_rate": 3.6760563380281694e-06, "loss": 0.3604, "step": 19698 }, { "epoch": 22.08, "learning_rate": 3.6746478873239436e-06, "loss": 0.3392, "step": 19699 }, { "epoch": 22.09, "learning_rate": 3.6732394366197187e-06, "loss": 0.3298, "step": 19700 }, { "epoch": 22.09, "learning_rate": 3.671830985915493e-06, "loss": 0.2913, "step": 19701 }, { "epoch": 22.09, "learning_rate": 3.6704225352112675e-06, "loss": 0.3307, "step": 19702 }, { "epoch": 22.09, "learning_rate": 3.669014084507042e-06, "loss": 0.3794, "step": 19703 }, { "epoch": 22.09, "learning_rate": 3.667605633802817e-06, "loss": 0.3029, "step": 19704 }, { "epoch": 22.09, "learning_rate": 3.6661971830985915e-06, "loss": 0.2982, "step": 19705 }, { "epoch": 22.09, "learning_rate": 3.664788732394366e-06, "loss": 0.3251, "step": 19706 }, { "epoch": 22.09, "learning_rate": 3.663380281690141e-06, "loss": 0.2694, "step": 19707 }, { "epoch": 22.09, "learning_rate": 3.6619718309859154e-06, "loss": 0.3261, "step": 19708 }, { "epoch": 22.1, "learning_rate": 3.6605633802816904e-06, "loss": 0.3, "step": 19709 }, { "epoch": 22.1, "learning_rate": 3.6591549295774646e-06, "loss": 0.2857, "step": 19710 }, { "epoch": 22.1, "learning_rate": 3.6577464788732397e-06, "loss": 0.2903, "step": 19711 }, { "epoch": 22.1, "learning_rate": 3.656338028169014e-06, "loss": 0.2577, "step": 19712 }, { "epoch": 22.1, "learning_rate": 3.654929577464789e-06, "loss": 0.317, "step": 19713 }, { "epoch": 22.1, "learning_rate": 3.653521126760563e-06, "loss": 0.2441, "step": 19714 }, { "epoch": 22.1, "learning_rate": 3.6521126760563382e-06, "loss": 0.2614, "step": 19715 }, { "epoch": 22.1, "learning_rate": 3.6507042253521125e-06, "loss": 0.381, "step": 19716 }, { "epoch": 22.1, "learning_rate": 3.6492957746478875e-06, "loss": 0.2039, "step": 19717 }, { "epoch": 22.11, "learning_rate": 3.6478873239436617e-06, "loss": 0.3078, "step": 19718 }, { "epoch": 22.11, "learning_rate": 3.646478873239437e-06, "loss": 0.2319, "step": 19719 }, { "epoch": 22.11, "learning_rate": 3.645070422535211e-06, "loss": 0.2277, "step": 19720 }, { "epoch": 22.11, "learning_rate": 3.643661971830986e-06, "loss": 0.1751, "step": 19721 }, { "epoch": 22.11, "learning_rate": 3.6422535211267603e-06, "loss": 0.2396, "step": 19722 }, { "epoch": 22.11, "learning_rate": 3.6408450704225353e-06, "loss": 0.2648, "step": 19723 }, { "epoch": 22.11, "learning_rate": 3.63943661971831e-06, "loss": 0.3321, "step": 19724 }, { "epoch": 22.11, "learning_rate": 3.6380281690140846e-06, "loss": 0.3767, "step": 19725 }, { "epoch": 22.11, "learning_rate": 3.6366197183098593e-06, "loss": 0.3688, "step": 19726 }, { "epoch": 22.12, "learning_rate": 3.635211267605634e-06, "loss": 0.3441, "step": 19727 }, { "epoch": 22.12, "learning_rate": 3.6338028169014085e-06, "loss": 0.3606, "step": 19728 }, { "epoch": 22.12, "learning_rate": 3.632394366197183e-06, "loss": 0.3436, "step": 19729 }, { "epoch": 22.12, "learning_rate": 3.630985915492958e-06, "loss": 0.3563, "step": 19730 }, { "epoch": 22.12, "learning_rate": 3.6295774647887324e-06, "loss": 0.3982, "step": 19731 }, { "epoch": 22.12, "learning_rate": 3.628169014084507e-06, "loss": 0.3889, "step": 19732 }, { "epoch": 22.12, "learning_rate": 3.6267605633802817e-06, "loss": 0.3971, "step": 19733 }, { "epoch": 22.12, "learning_rate": 3.6253521126760564e-06, "loss": 0.372, "step": 19734 }, { "epoch": 22.12, "learning_rate": 3.623943661971831e-06, "loss": 0.2955, "step": 19735 }, { "epoch": 22.13, "learning_rate": 3.6225352112676056e-06, "loss": 0.3359, "step": 19736 }, { "epoch": 22.13, "learning_rate": 3.6211267605633803e-06, "loss": 0.3487, "step": 19737 }, { "epoch": 22.13, "learning_rate": 3.619718309859155e-06, "loss": 0.357, "step": 19738 }, { "epoch": 22.13, "learning_rate": 3.6183098591549295e-06, "loss": 0.3404, "step": 19739 }, { "epoch": 22.13, "learning_rate": 3.616901408450704e-06, "loss": 0.3216, "step": 19740 }, { "epoch": 22.13, "learning_rate": 3.6154929577464792e-06, "loss": 0.3912, "step": 19741 }, { "epoch": 22.13, "learning_rate": 3.6140845070422535e-06, "loss": 0.3269, "step": 19742 }, { "epoch": 22.13, "learning_rate": 3.6126760563380285e-06, "loss": 0.2978, "step": 19743 }, { "epoch": 22.13, "learning_rate": 3.6112676056338027e-06, "loss": 0.3963, "step": 19744 }, { "epoch": 22.14, "learning_rate": 3.609859154929578e-06, "loss": 0.271, "step": 19745 }, { "epoch": 22.14, "learning_rate": 3.608450704225352e-06, "loss": 0.3391, "step": 19746 }, { "epoch": 22.14, "learning_rate": 3.607042253521127e-06, "loss": 0.2709, "step": 19747 }, { "epoch": 22.14, "learning_rate": 3.6056338028169013e-06, "loss": 0.3246, "step": 19748 }, { "epoch": 22.14, "learning_rate": 3.6042253521126763e-06, "loss": 0.3488, "step": 19749 }, { "epoch": 22.14, "learning_rate": 3.6028169014084506e-06, "loss": 0.2661, "step": 19750 }, { "epoch": 22.14, "learning_rate": 3.6014084507042256e-06, "loss": 0.3124, "step": 19751 }, { "epoch": 22.14, "learning_rate": 3.6e-06, "loss": 0.2871, "step": 19752 }, { "epoch": 22.14, "learning_rate": 3.598591549295775e-06, "loss": 0.2776, "step": 19753 }, { "epoch": 22.15, "learning_rate": 3.597183098591549e-06, "loss": 0.3143, "step": 19754 }, { "epoch": 22.15, "learning_rate": 3.595774647887324e-06, "loss": 0.3432, "step": 19755 }, { "epoch": 22.15, "learning_rate": 3.5943661971830984e-06, "loss": 0.2713, "step": 19756 }, { "epoch": 22.15, "learning_rate": 3.5929577464788734e-06, "loss": 0.2525, "step": 19757 }, { "epoch": 22.15, "learning_rate": 3.5915492957746477e-06, "loss": 0.2541, "step": 19758 }, { "epoch": 22.15, "learning_rate": 3.5901408450704227e-06, "loss": 0.3602, "step": 19759 }, { "epoch": 22.15, "learning_rate": 3.5887323943661974e-06, "loss": 0.312, "step": 19760 }, { "epoch": 22.15, "learning_rate": 3.587323943661972e-06, "loss": 0.2318, "step": 19761 }, { "epoch": 22.15, "learning_rate": 3.5859154929577466e-06, "loss": 0.269, "step": 19762 }, { "epoch": 22.16, "learning_rate": 3.5845070422535213e-06, "loss": 0.2687, "step": 19763 }, { "epoch": 22.16, "learning_rate": 3.583098591549296e-06, "loss": 0.2689, "step": 19764 }, { "epoch": 22.16, "learning_rate": 3.5816901408450705e-06, "loss": 0.3494, "step": 19765 }, { "epoch": 22.16, "learning_rate": 3.580281690140845e-06, "loss": 0.3777, "step": 19766 }, { "epoch": 22.16, "learning_rate": 3.57887323943662e-06, "loss": 0.2586, "step": 19767 }, { "epoch": 22.16, "learning_rate": 3.5774647887323945e-06, "loss": 0.2661, "step": 19768 }, { "epoch": 22.16, "learning_rate": 3.576056338028169e-06, "loss": 0.221, "step": 19769 }, { "epoch": 22.16, "learning_rate": 3.5746478873239437e-06, "loss": 0.1776, "step": 19770 }, { "epoch": 22.16, "learning_rate": 3.5732394366197184e-06, "loss": 0.3602, "step": 19771 }, { "epoch": 22.17, "learning_rate": 3.571830985915493e-06, "loss": 0.2839, "step": 19772 }, { "epoch": 22.17, "learning_rate": 3.5704225352112676e-06, "loss": 0.2193, "step": 19773 }, { "epoch": 22.17, "learning_rate": 3.5690140845070423e-06, "loss": 0.3367, "step": 19774 }, { "epoch": 22.17, "learning_rate": 3.567605633802817e-06, "loss": 0.3555, "step": 19775 }, { "epoch": 22.17, "learning_rate": 3.5661971830985916e-06, "loss": 0.3357, "step": 19776 }, { "epoch": 22.17, "learning_rate": 3.5647887323943666e-06, "loss": 0.3702, "step": 19777 }, { "epoch": 22.17, "learning_rate": 3.563380281690141e-06, "loss": 0.3661, "step": 19778 }, { "epoch": 22.17, "learning_rate": 3.561971830985916e-06, "loss": 0.3932, "step": 19779 }, { "epoch": 22.17, "learning_rate": 3.56056338028169e-06, "loss": 0.3222, "step": 19780 }, { "epoch": 22.18, "learning_rate": 3.559154929577465e-06, "loss": 0.3956, "step": 19781 }, { "epoch": 22.18, "learning_rate": 3.5577464788732394e-06, "loss": 0.2908, "step": 19782 }, { "epoch": 22.18, "learning_rate": 3.5563380281690144e-06, "loss": 0.3658, "step": 19783 }, { "epoch": 22.18, "learning_rate": 3.5549295774647887e-06, "loss": 0.3849, "step": 19784 }, { "epoch": 22.18, "learning_rate": 3.5535211267605637e-06, "loss": 0.4008, "step": 19785 }, { "epoch": 22.18, "learning_rate": 3.552112676056338e-06, "loss": 0.2692, "step": 19786 }, { "epoch": 22.18, "learning_rate": 3.550704225352113e-06, "loss": 0.2686, "step": 19787 }, { "epoch": 22.18, "learning_rate": 3.549295774647887e-06, "loss": 0.3476, "step": 19788 }, { "epoch": 22.18, "learning_rate": 3.5478873239436623e-06, "loss": 0.2374, "step": 19789 }, { "epoch": 22.19, "learning_rate": 3.5464788732394365e-06, "loss": 0.3238, "step": 19790 }, { "epoch": 22.19, "learning_rate": 3.5450704225352115e-06, "loss": 0.3633, "step": 19791 }, { "epoch": 22.19, "learning_rate": 3.5436619718309858e-06, "loss": 0.2968, "step": 19792 }, { "epoch": 22.19, "learning_rate": 3.542253521126761e-06, "loss": 0.2965, "step": 19793 }, { "epoch": 22.19, "learning_rate": 3.5408450704225355e-06, "loss": 0.3232, "step": 19794 }, { "epoch": 22.19, "learning_rate": 3.53943661971831e-06, "loss": 0.3041, "step": 19795 }, { "epoch": 22.19, "learning_rate": 3.5380281690140847e-06, "loss": 0.288, "step": 19796 }, { "epoch": 22.19, "learning_rate": 3.5366197183098594e-06, "loss": 0.2772, "step": 19797 }, { "epoch": 22.2, "learning_rate": 3.535211267605634e-06, "loss": 0.3068, "step": 19798 }, { "epoch": 22.2, "learning_rate": 3.5338028169014086e-06, "loss": 0.2706, "step": 19799 }, { "epoch": 22.2, "learning_rate": 3.5323943661971833e-06, "loss": 0.3077, "step": 19800 }, { "epoch": 22.2, "learning_rate": 3.530985915492958e-06, "loss": 0.3039, "step": 19801 }, { "epoch": 22.2, "learning_rate": 3.5295774647887326e-06, "loss": 0.289, "step": 19802 }, { "epoch": 22.2, "learning_rate": 3.528169014084507e-06, "loss": 0.2123, "step": 19803 }, { "epoch": 22.2, "learning_rate": 3.526760563380282e-06, "loss": 0.3782, "step": 19804 }, { "epoch": 22.2, "learning_rate": 3.5253521126760565e-06, "loss": 0.3061, "step": 19805 }, { "epoch": 22.2, "learning_rate": 3.523943661971831e-06, "loss": 0.3403, "step": 19806 }, { "epoch": 22.21, "learning_rate": 3.5225352112676057e-06, "loss": 0.3006, "step": 19807 }, { "epoch": 22.21, "learning_rate": 3.5211267605633804e-06, "loss": 0.3794, "step": 19808 }, { "epoch": 22.21, "learning_rate": 3.519718309859155e-06, "loss": 0.3847, "step": 19809 }, { "epoch": 22.21, "learning_rate": 3.5183098591549297e-06, "loss": 0.3138, "step": 19810 }, { "epoch": 22.21, "learning_rate": 3.5169014084507047e-06, "loss": 0.2757, "step": 19811 }, { "epoch": 22.21, "learning_rate": 3.515492957746479e-06, "loss": 0.2619, "step": 19812 }, { "epoch": 22.21, "learning_rate": 3.514084507042254e-06, "loss": 0.2852, "step": 19813 }, { "epoch": 22.21, "learning_rate": 3.512676056338028e-06, "loss": 0.2744, "step": 19814 }, { "epoch": 22.21, "learning_rate": 3.511267605633803e-06, "loss": 0.2608, "step": 19815 }, { "epoch": 22.22, "learning_rate": 3.5098591549295775e-06, "loss": 0.2566, "step": 19816 }, { "epoch": 22.22, "learning_rate": 3.508450704225352e-06, "loss": 0.2856, "step": 19817 }, { "epoch": 22.22, "learning_rate": 3.5070422535211268e-06, "loss": 0.2169, "step": 19818 }, { "epoch": 22.22, "learning_rate": 3.5056338028169014e-06, "loss": 0.2901, "step": 19819 }, { "epoch": 22.22, "learning_rate": 3.504225352112676e-06, "loss": 0.28, "step": 19820 }, { "epoch": 22.22, "learning_rate": 3.5028169014084507e-06, "loss": 0.2213, "step": 19821 }, { "epoch": 22.22, "learning_rate": 3.5014084507042253e-06, "loss": 0.3628, "step": 19822 }, { "epoch": 22.22, "learning_rate": 3.5e-06, "loss": 0.3022, "step": 19823 }, { "epoch": 22.22, "learning_rate": 3.4985915492957746e-06, "loss": 0.2819, "step": 19824 }, { "epoch": 22.23, "learning_rate": 3.4971830985915492e-06, "loss": 0.4828, "step": 19825 }, { "epoch": 22.23, "learning_rate": 3.495774647887324e-06, "loss": 0.3182, "step": 19826 }, { "epoch": 22.23, "learning_rate": 3.4943661971830985e-06, "loss": 0.3588, "step": 19827 }, { "epoch": 22.23, "learning_rate": 3.4929577464788736e-06, "loss": 0.4121, "step": 19828 }, { "epoch": 22.23, "learning_rate": 3.4915492957746478e-06, "loss": 0.3812, "step": 19829 }, { "epoch": 22.23, "learning_rate": 3.490140845070423e-06, "loss": 0.3631, "step": 19830 }, { "epoch": 22.23, "learning_rate": 3.488732394366197e-06, "loss": 0.319, "step": 19831 }, { "epoch": 22.23, "learning_rate": 3.487323943661972e-06, "loss": 0.4056, "step": 19832 }, { "epoch": 22.23, "learning_rate": 3.4859154929577463e-06, "loss": 0.3909, "step": 19833 }, { "epoch": 22.24, "learning_rate": 3.4845070422535214e-06, "loss": 0.298, "step": 19834 }, { "epoch": 22.24, "learning_rate": 3.4830985915492956e-06, "loss": 0.305, "step": 19835 }, { "epoch": 22.24, "learning_rate": 3.4816901408450707e-06, "loss": 0.3871, "step": 19836 }, { "epoch": 22.24, "learning_rate": 3.480281690140845e-06, "loss": 0.3457, "step": 19837 }, { "epoch": 22.24, "learning_rate": 3.47887323943662e-06, "loss": 0.3598, "step": 19838 }, { "epoch": 22.24, "learning_rate": 3.477464788732394e-06, "loss": 0.4192, "step": 19839 }, { "epoch": 22.24, "learning_rate": 3.476056338028169e-06, "loss": 0.3761, "step": 19840 }, { "epoch": 22.24, "learning_rate": 3.4746478873239434e-06, "loss": 0.246, "step": 19841 }, { "epoch": 22.24, "learning_rate": 3.4732394366197185e-06, "loss": 0.293, "step": 19842 }, { "epoch": 22.25, "learning_rate": 3.4718309859154927e-06, "loss": 0.253, "step": 19843 }, { "epoch": 22.25, "learning_rate": 3.4704225352112678e-06, "loss": 0.4078, "step": 19844 }, { "epoch": 22.25, "learning_rate": 3.4690140845070424e-06, "loss": 0.3114, "step": 19845 }, { "epoch": 22.25, "learning_rate": 3.467605633802817e-06, "loss": 0.2677, "step": 19846 }, { "epoch": 22.25, "learning_rate": 3.4661971830985917e-06, "loss": 0.2818, "step": 19847 }, { "epoch": 22.25, "learning_rate": 3.4647887323943663e-06, "loss": 0.3055, "step": 19848 }, { "epoch": 22.25, "learning_rate": 3.463380281690141e-06, "loss": 0.2918, "step": 19849 }, { "epoch": 22.25, "learning_rate": 3.4619718309859156e-06, "loss": 0.3677, "step": 19850 }, { "epoch": 22.25, "learning_rate": 3.4605633802816902e-06, "loss": 0.3261, "step": 19851 }, { "epoch": 22.26, "learning_rate": 3.459154929577465e-06, "loss": 0.3266, "step": 19852 }, { "epoch": 22.26, "learning_rate": 3.4577464788732395e-06, "loss": 0.3033, "step": 19853 }, { "epoch": 22.26, "learning_rate": 3.456338028169014e-06, "loss": 0.1971, "step": 19854 }, { "epoch": 22.26, "learning_rate": 3.4549295774647888e-06, "loss": 0.3318, "step": 19855 }, { "epoch": 22.26, "learning_rate": 3.4535211267605634e-06, "loss": 0.3819, "step": 19856 }, { "epoch": 22.26, "learning_rate": 3.452112676056338e-06, "loss": 0.3055, "step": 19857 }, { "epoch": 22.26, "learning_rate": 3.4507042253521127e-06, "loss": 0.2939, "step": 19858 }, { "epoch": 22.26, "learning_rate": 3.4492957746478873e-06, "loss": 0.335, "step": 19859 }, { "epoch": 22.26, "learning_rate": 3.447887323943662e-06, "loss": 0.3041, "step": 19860 }, { "epoch": 22.27, "learning_rate": 3.4464788732394366e-06, "loss": 0.3384, "step": 19861 }, { "epoch": 22.27, "learning_rate": 3.4450704225352112e-06, "loss": 0.3045, "step": 19862 }, { "epoch": 22.27, "learning_rate": 3.443661971830986e-06, "loss": 0.3548, "step": 19863 }, { "epoch": 22.27, "learning_rate": 3.442253521126761e-06, "loss": 0.2757, "step": 19864 }, { "epoch": 22.27, "learning_rate": 3.440845070422535e-06, "loss": 0.2732, "step": 19865 }, { "epoch": 22.27, "learning_rate": 3.43943661971831e-06, "loss": 0.4203, "step": 19866 }, { "epoch": 22.27, "learning_rate": 3.4380281690140844e-06, "loss": 0.2682, "step": 19867 }, { "epoch": 22.27, "learning_rate": 3.4366197183098595e-06, "loss": 0.2689, "step": 19868 }, { "epoch": 22.27, "learning_rate": 3.4352112676056337e-06, "loss": 0.3148, "step": 19869 }, { "epoch": 22.28, "learning_rate": 3.4338028169014088e-06, "loss": 0.2452, "step": 19870 }, { "epoch": 22.28, "learning_rate": 3.432394366197183e-06, "loss": 0.2627, "step": 19871 }, { "epoch": 22.28, "learning_rate": 3.430985915492958e-06, "loss": 0.3435, "step": 19872 }, { "epoch": 22.28, "learning_rate": 3.4295774647887322e-06, "loss": 0.2749, "step": 19873 }, { "epoch": 22.28, "learning_rate": 3.4281690140845073e-06, "loss": 0.1349, "step": 19874 }, { "epoch": 22.28, "learning_rate": 3.4267605633802815e-06, "loss": 0.3669, "step": 19875 }, { "epoch": 22.28, "learning_rate": 3.4253521126760566e-06, "loss": 0.385, "step": 19876 }, { "epoch": 22.28, "learning_rate": 3.423943661971831e-06, "loss": 0.3349, "step": 19877 }, { "epoch": 22.28, "learning_rate": 3.422535211267606e-06, "loss": 0.3713, "step": 19878 }, { "epoch": 22.29, "learning_rate": 3.42112676056338e-06, "loss": 0.3451, "step": 19879 }, { "epoch": 22.29, "learning_rate": 3.419718309859155e-06, "loss": 0.3233, "step": 19880 }, { "epoch": 22.29, "learning_rate": 3.4183098591549298e-06, "loss": 0.3028, "step": 19881 }, { "epoch": 22.29, "learning_rate": 3.4169014084507044e-06, "loss": 0.3925, "step": 19882 }, { "epoch": 22.29, "learning_rate": 3.415492957746479e-06, "loss": 0.3439, "step": 19883 }, { "epoch": 22.29, "learning_rate": 3.4140845070422537e-06, "loss": 0.3286, "step": 19884 }, { "epoch": 22.29, "learning_rate": 3.4126760563380283e-06, "loss": 0.2986, "step": 19885 }, { "epoch": 22.29, "learning_rate": 3.411267605633803e-06, "loss": 0.2683, "step": 19886 }, { "epoch": 22.29, "learning_rate": 3.4098591549295776e-06, "loss": 0.2947, "step": 19887 }, { "epoch": 22.3, "learning_rate": 3.4084507042253522e-06, "loss": 0.4304, "step": 19888 }, { "epoch": 22.3, "learning_rate": 3.407042253521127e-06, "loss": 0.4483, "step": 19889 }, { "epoch": 22.3, "learning_rate": 3.4056338028169015e-06, "loss": 0.331, "step": 19890 }, { "epoch": 22.3, "learning_rate": 3.404225352112676e-06, "loss": 0.3283, "step": 19891 }, { "epoch": 22.3, "learning_rate": 3.4028169014084508e-06, "loss": 0.3677, "step": 19892 }, { "epoch": 22.3, "learning_rate": 3.4014084507042254e-06, "loss": 0.3356, "step": 19893 }, { "epoch": 22.3, "learning_rate": 3.4e-06, "loss": 0.3363, "step": 19894 }, { "epoch": 22.3, "learning_rate": 3.3985915492957747e-06, "loss": 0.3346, "step": 19895 }, { "epoch": 22.3, "learning_rate": 3.3971830985915493e-06, "loss": 0.3317, "step": 19896 }, { "epoch": 22.31, "learning_rate": 3.395774647887324e-06, "loss": 0.2384, "step": 19897 }, { "epoch": 22.31, "learning_rate": 3.394366197183099e-06, "loss": 0.3424, "step": 19898 }, { "epoch": 22.31, "learning_rate": 3.3929577464788732e-06, "loss": 0.3026, "step": 19899 }, { "epoch": 22.31, "learning_rate": 3.3915492957746483e-06, "loss": 0.3044, "step": 19900 }, { "epoch": 22.31, "learning_rate": 3.3901408450704225e-06, "loss": 0.2817, "step": 19901 }, { "epoch": 22.31, "learning_rate": 3.3887323943661976e-06, "loss": 0.3174, "step": 19902 }, { "epoch": 22.31, "learning_rate": 3.387323943661972e-06, "loss": 0.2672, "step": 19903 }, { "epoch": 22.31, "learning_rate": 3.385915492957747e-06, "loss": 0.2883, "step": 19904 }, { "epoch": 22.32, "learning_rate": 3.384507042253521e-06, "loss": 0.3614, "step": 19905 }, { "epoch": 22.32, "learning_rate": 3.383098591549296e-06, "loss": 0.3193, "step": 19906 }, { "epoch": 22.32, "learning_rate": 3.3816901408450703e-06, "loss": 0.4061, "step": 19907 }, { "epoch": 22.32, "learning_rate": 3.3802816901408454e-06, "loss": 0.2416, "step": 19908 }, { "epoch": 22.32, "learning_rate": 3.3788732394366196e-06, "loss": 0.3349, "step": 19909 }, { "epoch": 22.32, "learning_rate": 3.3774647887323947e-06, "loss": 0.2806, "step": 19910 }, { "epoch": 22.32, "learning_rate": 3.376056338028169e-06, "loss": 0.3667, "step": 19911 }, { "epoch": 22.32, "learning_rate": 3.374647887323944e-06, "loss": 0.2556, "step": 19912 }, { "epoch": 22.32, "learning_rate": 3.373239436619718e-06, "loss": 0.3512, "step": 19913 }, { "epoch": 22.33, "learning_rate": 3.3718309859154932e-06, "loss": 0.3089, "step": 19914 }, { "epoch": 22.33, "learning_rate": 3.370422535211268e-06, "loss": 0.3948, "step": 19915 }, { "epoch": 22.33, "learning_rate": 3.3690140845070425e-06, "loss": 0.3875, "step": 19916 }, { "epoch": 22.33, "learning_rate": 3.367605633802817e-06, "loss": 0.4082, "step": 19917 }, { "epoch": 22.33, "learning_rate": 3.3661971830985918e-06, "loss": 0.3057, "step": 19918 }, { "epoch": 22.33, "learning_rate": 3.3647887323943664e-06, "loss": 0.2877, "step": 19919 }, { "epoch": 22.33, "learning_rate": 3.363380281690141e-06, "loss": 0.1841, "step": 19920 }, { "epoch": 22.33, "learning_rate": 3.3619718309859157e-06, "loss": 0.2455, "step": 19921 }, { "epoch": 22.33, "learning_rate": 3.3605633802816903e-06, "loss": 0.2501, "step": 19922 }, { "epoch": 22.34, "learning_rate": 3.359154929577465e-06, "loss": 0.2928, "step": 19923 }, { "epoch": 22.34, "learning_rate": 3.3577464788732396e-06, "loss": 0.3252, "step": 19924 }, { "epoch": 22.34, "learning_rate": 3.3563380281690142e-06, "loss": 0.4136, "step": 19925 }, { "epoch": 22.34, "learning_rate": 3.354929577464789e-06, "loss": 0.4897, "step": 19926 }, { "epoch": 22.34, "learning_rate": 3.3535211267605635e-06, "loss": 0.4485, "step": 19927 }, { "epoch": 22.34, "learning_rate": 3.352112676056338e-06, "loss": 0.3555, "step": 19928 }, { "epoch": 22.34, "learning_rate": 3.350704225352113e-06, "loss": 0.4084, "step": 19929 }, { "epoch": 22.34, "learning_rate": 3.3492957746478874e-06, "loss": 0.325, "step": 19930 }, { "epoch": 22.34, "learning_rate": 3.347887323943662e-06, "loss": 0.2959, "step": 19931 }, { "epoch": 22.35, "learning_rate": 3.3464788732394367e-06, "loss": 0.4155, "step": 19932 }, { "epoch": 22.35, "learning_rate": 3.3450704225352113e-06, "loss": 0.3214, "step": 19933 }, { "epoch": 22.35, "learning_rate": 3.343661971830986e-06, "loss": 0.3607, "step": 19934 }, { "epoch": 22.35, "learning_rate": 3.3422535211267606e-06, "loss": 0.3719, "step": 19935 }, { "epoch": 22.35, "learning_rate": 3.3408450704225353e-06, "loss": 0.3945, "step": 19936 }, { "epoch": 22.35, "learning_rate": 3.33943661971831e-06, "loss": 0.3925, "step": 19937 }, { "epoch": 22.35, "learning_rate": 3.3380281690140845e-06, "loss": 0.4046, "step": 19938 }, { "epoch": 22.35, "learning_rate": 3.336619718309859e-06, "loss": 0.3606, "step": 19939 }, { "epoch": 22.35, "learning_rate": 3.335211267605634e-06, "loss": 0.4281, "step": 19940 }, { "epoch": 22.36, "learning_rate": 3.3338028169014084e-06, "loss": 0.2864, "step": 19941 }, { "epoch": 22.36, "learning_rate": 3.332394366197183e-06, "loss": 0.2968, "step": 19942 }, { "epoch": 22.36, "learning_rate": 3.3309859154929577e-06, "loss": 0.286, "step": 19943 }, { "epoch": 22.36, "learning_rate": 3.3295774647887324e-06, "loss": 0.3271, "step": 19944 }, { "epoch": 22.36, "learning_rate": 3.328169014084507e-06, "loss": 0.3778, "step": 19945 }, { "epoch": 22.36, "learning_rate": 3.3267605633802816e-06, "loss": 0.3064, "step": 19946 }, { "epoch": 22.36, "learning_rate": 3.3253521126760563e-06, "loss": 0.3245, "step": 19947 }, { "epoch": 22.36, "learning_rate": 3.323943661971831e-06, "loss": 0.3547, "step": 19948 }, { "epoch": 22.36, "learning_rate": 3.3225352112676055e-06, "loss": 0.2308, "step": 19949 }, { "epoch": 22.37, "learning_rate": 3.32112676056338e-06, "loss": 0.387, "step": 19950 }, { "epoch": 22.37, "learning_rate": 3.3197183098591552e-06, "loss": 0.4019, "step": 19951 }, { "epoch": 22.37, "learning_rate": 3.3183098591549295e-06, "loss": 0.3263, "step": 19952 }, { "epoch": 22.37, "learning_rate": 3.3169014084507045e-06, "loss": 0.3506, "step": 19953 }, { "epoch": 22.37, "learning_rate": 3.3154929577464787e-06, "loss": 0.303, "step": 19954 }, { "epoch": 22.37, "learning_rate": 3.314084507042254e-06, "loss": 0.3947, "step": 19955 }, { "epoch": 22.37, "learning_rate": 3.312676056338028e-06, "loss": 0.3398, "step": 19956 }, { "epoch": 22.37, "learning_rate": 3.311267605633803e-06, "loss": 0.3612, "step": 19957 }, { "epoch": 22.37, "learning_rate": 3.3098591549295773e-06, "loss": 0.3783, "step": 19958 }, { "epoch": 22.38, "learning_rate": 3.3084507042253523e-06, "loss": 0.3309, "step": 19959 }, { "epoch": 22.38, "learning_rate": 3.3070422535211266e-06, "loss": 0.3597, "step": 19960 }, { "epoch": 22.38, "learning_rate": 3.3056338028169016e-06, "loss": 0.4069, "step": 19961 }, { "epoch": 22.38, "learning_rate": 3.304225352112676e-06, "loss": 0.3958, "step": 19962 }, { "epoch": 22.38, "learning_rate": 3.302816901408451e-06, "loss": 0.3129, "step": 19963 }, { "epoch": 22.38, "learning_rate": 3.301408450704225e-06, "loss": 0.3153, "step": 19964 }, { "epoch": 22.38, "learning_rate": 3.3e-06, "loss": 0.3071, "step": 19965 }, { "epoch": 22.38, "learning_rate": 3.2985915492957744e-06, "loss": 0.3172, "step": 19966 }, { "epoch": 22.38, "learning_rate": 3.2971830985915494e-06, "loss": 0.2603, "step": 19967 }, { "epoch": 22.39, "learning_rate": 3.295774647887324e-06, "loss": 0.3308, "step": 19968 }, { "epoch": 22.39, "learning_rate": 3.2943661971830987e-06, "loss": 0.3059, "step": 19969 }, { "epoch": 22.39, "learning_rate": 3.2929577464788734e-06, "loss": 0.3172, "step": 19970 }, { "epoch": 22.39, "learning_rate": 3.291549295774648e-06, "loss": 0.256, "step": 19971 }, { "epoch": 22.39, "learning_rate": 3.2901408450704226e-06, "loss": 0.2312, "step": 19972 }, { "epoch": 22.39, "learning_rate": 3.2887323943661973e-06, "loss": 0.2035, "step": 19973 }, { "epoch": 22.39, "learning_rate": 3.287323943661972e-06, "loss": 0.2639, "step": 19974 }, { "epoch": 22.39, "learning_rate": 3.2859154929577465e-06, "loss": 0.4014, "step": 19975 }, { "epoch": 22.39, "learning_rate": 3.284507042253521e-06, "loss": 0.3496, "step": 19976 }, { "epoch": 22.4, "learning_rate": 3.283098591549296e-06, "loss": 0.4538, "step": 19977 }, { "epoch": 22.4, "learning_rate": 3.2816901408450705e-06, "loss": 0.3082, "step": 19978 }, { "epoch": 22.4, "learning_rate": 3.280281690140845e-06, "loss": 0.2672, "step": 19979 }, { "epoch": 22.4, "learning_rate": 3.2788732394366197e-06, "loss": 0.4297, "step": 19980 }, { "epoch": 22.4, "learning_rate": 3.2774647887323944e-06, "loss": 0.3453, "step": 19981 }, { "epoch": 22.4, "learning_rate": 3.276056338028169e-06, "loss": 0.3073, "step": 19982 }, { "epoch": 22.4, "learning_rate": 3.2746478873239436e-06, "loss": 0.2953, "step": 19983 }, { "epoch": 22.4, "learning_rate": 3.2732394366197183e-06, "loss": 0.3408, "step": 19984 }, { "epoch": 22.4, "learning_rate": 3.2718309859154933e-06, "loss": 0.3395, "step": 19985 }, { "epoch": 22.41, "learning_rate": 3.2704225352112676e-06, "loss": 0.3077, "step": 19986 }, { "epoch": 22.41, "learning_rate": 3.2690140845070426e-06, "loss": 0.3075, "step": 19987 }, { "epoch": 22.41, "learning_rate": 3.267605633802817e-06, "loss": 0.3365, "step": 19988 }, { "epoch": 22.41, "learning_rate": 3.266197183098592e-06, "loss": 0.2687, "step": 19989 }, { "epoch": 22.41, "learning_rate": 3.264788732394366e-06, "loss": 0.3416, "step": 19990 }, { "epoch": 22.41, "learning_rate": 3.263380281690141e-06, "loss": 0.3185, "step": 19991 }, { "epoch": 22.41, "learning_rate": 3.2619718309859154e-06, "loss": 0.2139, "step": 19992 }, { "epoch": 22.41, "learning_rate": 3.2605633802816904e-06, "loss": 0.3114, "step": 19993 }, { "epoch": 22.41, "learning_rate": 3.2591549295774647e-06, "loss": 0.307, "step": 19994 }, { "epoch": 22.42, "learning_rate": 3.2577464788732397e-06, "loss": 0.3693, "step": 19995 }, { "epoch": 22.42, "learning_rate": 3.256338028169014e-06, "loss": 0.3303, "step": 19996 }, { "epoch": 22.42, "learning_rate": 3.254929577464789e-06, "loss": 0.329, "step": 19997 }, { "epoch": 22.42, "learning_rate": 3.253521126760563e-06, "loss": 0.3493, "step": 19998 }, { "epoch": 22.42, "learning_rate": 3.2521126760563383e-06, "loss": 0.359, "step": 19999 }, { "epoch": 22.42, "learning_rate": 3.2507042253521125e-06, "loss": 0.2644, "step": 20000 }, { "epoch": 22.42, "learning_rate": 3.2492957746478875e-06, "loss": 0.3071, "step": 20001 }, { "epoch": 22.42, "learning_rate": 3.247887323943662e-06, "loss": 0.2776, "step": 20002 }, { "epoch": 22.42, "learning_rate": 3.246478873239437e-06, "loss": 0.3586, "step": 20003 }, { "epoch": 22.43, "learning_rate": 3.2450704225352115e-06, "loss": 0.3297, "step": 20004 }, { "epoch": 22.43, "learning_rate": 3.243661971830986e-06, "loss": 0.3276, "step": 20005 }, { "epoch": 22.43, "learning_rate": 3.2422535211267607e-06, "loss": 0.2659, "step": 20006 }, { "epoch": 22.43, "learning_rate": 3.2408450704225354e-06, "loss": 0.2471, "step": 20007 }, { "epoch": 22.43, "learning_rate": 3.23943661971831e-06, "loss": 0.2855, "step": 20008 }, { "epoch": 22.43, "learning_rate": 3.2380281690140846e-06, "loss": 0.2423, "step": 20009 }, { "epoch": 22.43, "learning_rate": 3.2366197183098593e-06, "loss": 0.299, "step": 20010 }, { "epoch": 22.43, "learning_rate": 3.235211267605634e-06, "loss": 0.2577, "step": 20011 }, { "epoch": 22.43, "learning_rate": 3.2338028169014086e-06, "loss": 0.287, "step": 20012 }, { "epoch": 22.44, "learning_rate": 3.232394366197183e-06, "loss": 0.2664, "step": 20013 }, { "epoch": 22.44, "learning_rate": 3.230985915492958e-06, "loss": 0.3408, "step": 20014 }, { "epoch": 22.44, "learning_rate": 3.2295774647887325e-06, "loss": 0.2798, "step": 20015 }, { "epoch": 22.44, "learning_rate": 3.228169014084507e-06, "loss": 0.2476, "step": 20016 }, { "epoch": 22.44, "learning_rate": 3.2267605633802817e-06, "loss": 0.3649, "step": 20017 }, { "epoch": 22.44, "learning_rate": 3.2253521126760564e-06, "loss": 0.2267, "step": 20018 }, { "epoch": 22.44, "learning_rate": 3.2239436619718314e-06, "loss": 0.2955, "step": 20019 }, { "epoch": 22.44, "learning_rate": 3.2225352112676057e-06, "loss": 0.1969, "step": 20020 }, { "epoch": 22.45, "learning_rate": 3.2211267605633807e-06, "loss": 0.2391, "step": 20021 }, { "epoch": 22.45, "learning_rate": 3.219718309859155e-06, "loss": 0.3215, "step": 20022 }, { "epoch": 22.45, "learning_rate": 3.21830985915493e-06, "loss": 0.1723, "step": 20023 }, { "epoch": 22.45, "learning_rate": 3.216901408450704e-06, "loss": 0.477, "step": 20024 }, { "epoch": 22.45, "learning_rate": 3.2154929577464793e-06, "loss": 0.4026, "step": 20025 }, { "epoch": 22.45, "learning_rate": 3.2140845070422535e-06, "loss": 0.345, "step": 20026 }, { "epoch": 22.45, "learning_rate": 3.2126760563380285e-06, "loss": 0.3499, "step": 20027 }, { "epoch": 22.45, "learning_rate": 3.2112676056338028e-06, "loss": 0.2951, "step": 20028 }, { "epoch": 22.45, "learning_rate": 3.209859154929578e-06, "loss": 0.3837, "step": 20029 }, { "epoch": 22.46, "learning_rate": 3.208450704225352e-06, "loss": 0.2938, "step": 20030 }, { "epoch": 22.46, "learning_rate": 3.207042253521127e-06, "loss": 0.3206, "step": 20031 }, { "epoch": 22.46, "learning_rate": 3.2056338028169013e-06, "loss": 0.3839, "step": 20032 }, { "epoch": 22.46, "learning_rate": 3.2042253521126764e-06, "loss": 0.3242, "step": 20033 }, { "epoch": 22.46, "learning_rate": 3.2028169014084506e-06, "loss": 0.3203, "step": 20034 }, { "epoch": 22.46, "learning_rate": 3.2014084507042256e-06, "loss": 0.3296, "step": 20035 }, { "epoch": 22.46, "learning_rate": 3.2000000000000003e-06, "loss": 0.3153, "step": 20036 }, { "epoch": 22.46, "learning_rate": 3.198591549295775e-06, "loss": 0.3485, "step": 20037 }, { "epoch": 22.46, "learning_rate": 3.1971830985915496e-06, "loss": 0.3589, "step": 20038 }, { "epoch": 22.47, "learning_rate": 3.195774647887324e-06, "loss": 0.3679, "step": 20039 }, { "epoch": 22.47, "learning_rate": 3.194366197183099e-06, "loss": 0.3726, "step": 20040 }, { "epoch": 22.47, "learning_rate": 3.1929577464788735e-06, "loss": 0.3549, "step": 20041 }, { "epoch": 22.47, "learning_rate": 3.191549295774648e-06, "loss": 0.291, "step": 20042 }, { "epoch": 22.47, "learning_rate": 3.1901408450704223e-06, "loss": 0.283, "step": 20043 }, { "epoch": 22.47, "learning_rate": 3.1887323943661974e-06, "loss": 0.3441, "step": 20044 }, { "epoch": 22.47, "learning_rate": 3.1873239436619716e-06, "loss": 0.3824, "step": 20045 }, { "epoch": 22.47, "learning_rate": 3.1859154929577467e-06, "loss": 0.3539, "step": 20046 }, { "epoch": 22.47, "learning_rate": 3.184507042253521e-06, "loss": 0.2917, "step": 20047 }, { "epoch": 22.48, "learning_rate": 3.183098591549296e-06, "loss": 0.268, "step": 20048 }, { "epoch": 22.48, "learning_rate": 3.18169014084507e-06, "loss": 0.3962, "step": 20049 }, { "epoch": 22.48, "learning_rate": 3.180281690140845e-06, "loss": 0.2906, "step": 20050 }, { "epoch": 22.48, "learning_rate": 3.1788732394366194e-06, "loss": 0.2693, "step": 20051 }, { "epoch": 22.48, "learning_rate": 3.1774647887323945e-06, "loss": 0.3339, "step": 20052 }, { "epoch": 22.48, "learning_rate": 3.1760563380281687e-06, "loss": 0.2715, "step": 20053 }, { "epoch": 22.48, "learning_rate": 3.1746478873239438e-06, "loss": 0.2832, "step": 20054 }, { "epoch": 22.48, "learning_rate": 3.1732394366197184e-06, "loss": 0.2616, "step": 20055 }, { "epoch": 22.48, "learning_rate": 3.171830985915493e-06, "loss": 0.2337, "step": 20056 }, { "epoch": 22.49, "learning_rate": 3.1704225352112677e-06, "loss": 0.3161, "step": 20057 }, { "epoch": 22.49, "learning_rate": 3.1690140845070423e-06, "loss": 0.3059, "step": 20058 }, { "epoch": 22.49, "learning_rate": 3.167605633802817e-06, "loss": 0.3003, "step": 20059 }, { "epoch": 22.49, "learning_rate": 3.1661971830985916e-06, "loss": 0.3935, "step": 20060 }, { "epoch": 22.49, "learning_rate": 3.1647887323943662e-06, "loss": 0.3336, "step": 20061 }, { "epoch": 22.49, "learning_rate": 3.163380281690141e-06, "loss": 0.3149, "step": 20062 }, { "epoch": 22.49, "learning_rate": 3.1619718309859155e-06, "loss": 0.2629, "step": 20063 }, { "epoch": 22.49, "learning_rate": 3.16056338028169e-06, "loss": 0.2177, "step": 20064 }, { "epoch": 22.49, "learning_rate": 3.1591549295774648e-06, "loss": 0.2892, "step": 20065 }, { "epoch": 22.5, "learning_rate": 3.1577464788732394e-06, "loss": 0.2947, "step": 20066 }, { "epoch": 22.5, "learning_rate": 3.156338028169014e-06, "loss": 0.26, "step": 20067 }, { "epoch": 22.5, "learning_rate": 3.1549295774647887e-06, "loss": 0.2234, "step": 20068 }, { "epoch": 22.5, "learning_rate": 3.1535211267605633e-06, "loss": 0.312, "step": 20069 }, { "epoch": 22.5, "learning_rate": 3.152112676056338e-06, "loss": 0.3354, "step": 20070 }, { "epoch": 22.5, "learning_rate": 3.1507042253521126e-06, "loss": 0.2868, "step": 20071 }, { "epoch": 22.5, "learning_rate": 3.1492957746478877e-06, "loss": 0.2706, "step": 20072 }, { "epoch": 22.5, "learning_rate": 3.147887323943662e-06, "loss": 0.2097, "step": 20073 }, { "epoch": 22.5, "learning_rate": 3.146478873239437e-06, "loss": 0.2744, "step": 20074 }, { "epoch": 22.51, "learning_rate": 3.145070422535211e-06, "loss": 0.5132, "step": 20075 }, { "epoch": 22.51, "learning_rate": 3.143661971830986e-06, "loss": 0.4058, "step": 20076 }, { "epoch": 22.51, "learning_rate": 3.1422535211267604e-06, "loss": 0.3863, "step": 20077 }, { "epoch": 22.51, "learning_rate": 3.1408450704225355e-06, "loss": 0.3295, "step": 20078 }, { "epoch": 22.51, "learning_rate": 3.1394366197183097e-06, "loss": 0.3295, "step": 20079 }, { "epoch": 22.51, "learning_rate": 3.1380281690140848e-06, "loss": 0.3252, "step": 20080 }, { "epoch": 22.51, "learning_rate": 3.136619718309859e-06, "loss": 0.3425, "step": 20081 }, { "epoch": 22.51, "learning_rate": 3.135211267605634e-06, "loss": 0.3237, "step": 20082 }, { "epoch": 22.51, "learning_rate": 3.1338028169014082e-06, "loss": 0.3009, "step": 20083 }, { "epoch": 22.52, "learning_rate": 3.1323943661971833e-06, "loss": 0.3654, "step": 20084 }, { "epoch": 22.52, "learning_rate": 3.1309859154929575e-06, "loss": 0.2013, "step": 20085 }, { "epoch": 22.52, "learning_rate": 3.1295774647887326e-06, "loss": 0.3876, "step": 20086 }, { "epoch": 22.52, "learning_rate": 3.128169014084507e-06, "loss": 0.3855, "step": 20087 }, { "epoch": 22.52, "learning_rate": 3.126760563380282e-06, "loss": 0.3303, "step": 20088 }, { "epoch": 22.52, "learning_rate": 3.1253521126760565e-06, "loss": 0.2674, "step": 20089 }, { "epoch": 22.52, "learning_rate": 3.123943661971831e-06, "loss": 0.3382, "step": 20090 }, { "epoch": 22.52, "learning_rate": 3.1225352112676058e-06, "loss": 0.2938, "step": 20091 }, { "epoch": 22.52, "learning_rate": 3.1211267605633804e-06, "loss": 0.2812, "step": 20092 }, { "epoch": 22.53, "learning_rate": 3.119718309859155e-06, "loss": 0.3378, "step": 20093 }, { "epoch": 22.53, "learning_rate": 3.1183098591549297e-06, "loss": 0.2552, "step": 20094 }, { "epoch": 22.53, "learning_rate": 3.1169014084507043e-06, "loss": 0.3357, "step": 20095 }, { "epoch": 22.53, "learning_rate": 3.115492957746479e-06, "loss": 0.306, "step": 20096 }, { "epoch": 22.53, "learning_rate": 3.1140845070422536e-06, "loss": 0.2417, "step": 20097 }, { "epoch": 22.53, "learning_rate": 3.1126760563380282e-06, "loss": 0.3403, "step": 20098 }, { "epoch": 22.53, "learning_rate": 3.111267605633803e-06, "loss": 0.2999, "step": 20099 }, { "epoch": 22.53, "learning_rate": 3.1098591549295775e-06, "loss": 0.2936, "step": 20100 }, { "epoch": 22.53, "learning_rate": 3.108450704225352e-06, "loss": 0.3083, "step": 20101 }, { "epoch": 22.54, "learning_rate": 3.1070422535211268e-06, "loss": 0.3032, "step": 20102 }, { "epoch": 22.54, "learning_rate": 3.1056338028169014e-06, "loss": 0.2769, "step": 20103 }, { "epoch": 22.54, "learning_rate": 3.104225352112676e-06, "loss": 0.3396, "step": 20104 }, { "epoch": 22.54, "learning_rate": 3.1028169014084507e-06, "loss": 0.367, "step": 20105 }, { "epoch": 22.54, "learning_rate": 3.1014084507042258e-06, "loss": 0.2625, "step": 20106 }, { "epoch": 22.54, "learning_rate": 3.1e-06, "loss": 0.2888, "step": 20107 }, { "epoch": 22.54, "learning_rate": 3.098591549295775e-06, "loss": 0.3137, "step": 20108 }, { "epoch": 22.54, "learning_rate": 3.0971830985915492e-06, "loss": 0.3242, "step": 20109 }, { "epoch": 22.54, "learning_rate": 3.0957746478873243e-06, "loss": 0.3109, "step": 20110 }, { "epoch": 22.55, "learning_rate": 3.0943661971830985e-06, "loss": 0.3015, "step": 20111 }, { "epoch": 22.55, "learning_rate": 3.0929577464788736e-06, "loss": 0.2652, "step": 20112 }, { "epoch": 22.55, "learning_rate": 3.091549295774648e-06, "loss": 0.3116, "step": 20113 }, { "epoch": 22.55, "learning_rate": 3.090140845070423e-06, "loss": 0.3125, "step": 20114 }, { "epoch": 22.55, "learning_rate": 3.088732394366197e-06, "loss": 0.2695, "step": 20115 }, { "epoch": 22.55, "learning_rate": 3.087323943661972e-06, "loss": 0.2693, "step": 20116 }, { "epoch": 22.55, "learning_rate": 3.0859154929577463e-06, "loss": 0.1876, "step": 20117 }, { "epoch": 22.55, "learning_rate": 3.0845070422535214e-06, "loss": 0.2893, "step": 20118 }, { "epoch": 22.55, "learning_rate": 3.0830985915492956e-06, "loss": 0.3289, "step": 20119 }, { "epoch": 22.56, "learning_rate": 3.0816901408450707e-06, "loss": 0.2082, "step": 20120 }, { "epoch": 22.56, "learning_rate": 3.080281690140845e-06, "loss": 0.2859, "step": 20121 }, { "epoch": 22.56, "learning_rate": 3.07887323943662e-06, "loss": 0.2403, "step": 20122 }, { "epoch": 22.56, "learning_rate": 3.0774647887323946e-06, "loss": 0.2486, "step": 20123 }, { "epoch": 22.56, "learning_rate": 3.0760563380281692e-06, "loss": 0.2093, "step": 20124 }, { "epoch": 22.56, "learning_rate": 3.074647887323944e-06, "loss": 0.4461, "step": 20125 }, { "epoch": 22.56, "learning_rate": 3.0732394366197185e-06, "loss": 0.3774, "step": 20126 }, { "epoch": 22.56, "learning_rate": 3.071830985915493e-06, "loss": 0.4533, "step": 20127 }, { "epoch": 22.57, "learning_rate": 3.0704225352112678e-06, "loss": 0.3368, "step": 20128 }, { "epoch": 22.57, "learning_rate": 3.0690140845070424e-06, "loss": 0.2978, "step": 20129 }, { "epoch": 22.57, "learning_rate": 3.067605633802817e-06, "loss": 0.4185, "step": 20130 }, { "epoch": 22.57, "learning_rate": 3.0661971830985917e-06, "loss": 0.3639, "step": 20131 }, { "epoch": 22.57, "learning_rate": 3.0647887323943663e-06, "loss": 0.4434, "step": 20132 }, { "epoch": 22.57, "learning_rate": 3.063380281690141e-06, "loss": 0.39, "step": 20133 }, { "epoch": 22.57, "learning_rate": 3.0619718309859156e-06, "loss": 0.32, "step": 20134 }, { "epoch": 22.57, "learning_rate": 3.0605633802816902e-06, "loss": 0.2455, "step": 20135 }, { "epoch": 22.57, "learning_rate": 3.059154929577465e-06, "loss": 0.3313, "step": 20136 }, { "epoch": 22.58, "learning_rate": 3.0577464788732395e-06, "loss": 0.2822, "step": 20137 }, { "epoch": 22.58, "learning_rate": 3.056338028169014e-06, "loss": 0.3891, "step": 20138 }, { "epoch": 22.58, "learning_rate": 3.054929577464789e-06, "loss": 0.3636, "step": 20139 }, { "epoch": 22.58, "learning_rate": 3.053521126760564e-06, "loss": 0.3214, "step": 20140 }, { "epoch": 22.58, "learning_rate": 3.052112676056338e-06, "loss": 0.3594, "step": 20141 }, { "epoch": 22.58, "learning_rate": 3.050704225352113e-06, "loss": 0.2436, "step": 20142 }, { "epoch": 22.58, "learning_rate": 3.0492957746478873e-06, "loss": 0.2612, "step": 20143 }, { "epoch": 22.58, "learning_rate": 3.0478873239436624e-06, "loss": 0.3648, "step": 20144 }, { "epoch": 22.58, "learning_rate": 3.0464788732394366e-06, "loss": 0.2843, "step": 20145 }, { "epoch": 22.59, "learning_rate": 3.0450704225352117e-06, "loss": 0.323, "step": 20146 }, { "epoch": 22.59, "learning_rate": 3.043661971830986e-06, "loss": 0.4047, "step": 20147 }, { "epoch": 22.59, "learning_rate": 3.042253521126761e-06, "loss": 0.3007, "step": 20148 }, { "epoch": 22.59, "learning_rate": 3.040845070422535e-06, "loss": 0.3279, "step": 20149 }, { "epoch": 22.59, "learning_rate": 3.0394366197183102e-06, "loss": 0.3003, "step": 20150 }, { "epoch": 22.59, "learning_rate": 3.0380281690140844e-06, "loss": 0.3046, "step": 20151 }, { "epoch": 22.59, "learning_rate": 3.0366197183098595e-06, "loss": 0.3197, "step": 20152 }, { "epoch": 22.59, "learning_rate": 3.0352112676056337e-06, "loss": 0.3795, "step": 20153 }, { "epoch": 22.59, "learning_rate": 3.0338028169014088e-06, "loss": 0.3392, "step": 20154 }, { "epoch": 22.6, "learning_rate": 3.032394366197183e-06, "loss": 0.3249, "step": 20155 }, { "epoch": 22.6, "learning_rate": 3.0309859154929576e-06, "loss": 0.2457, "step": 20156 }, { "epoch": 22.6, "learning_rate": 3.0295774647887323e-06, "loss": 0.2328, "step": 20157 }, { "epoch": 22.6, "learning_rate": 3.028169014084507e-06, "loss": 0.3039, "step": 20158 }, { "epoch": 22.6, "learning_rate": 3.026760563380282e-06, "loss": 0.2764, "step": 20159 }, { "epoch": 22.6, "learning_rate": 3.025352112676056e-06, "loss": 0.3793, "step": 20160 }, { "epoch": 22.6, "learning_rate": 3.0239436619718312e-06, "loss": 0.3248, "step": 20161 }, { "epoch": 22.6, "learning_rate": 3.0225352112676055e-06, "loss": 0.3012, "step": 20162 }, { "epoch": 22.6, "learning_rate": 3.0211267605633805e-06, "loss": 0.2211, "step": 20163 }, { "epoch": 22.61, "learning_rate": 3.0197183098591547e-06, "loss": 0.3108, "step": 20164 }, { "epoch": 22.61, "learning_rate": 3.01830985915493e-06, "loss": 0.2399, "step": 20165 }, { "epoch": 22.61, "learning_rate": 3.016901408450704e-06, "loss": 0.3101, "step": 20166 }, { "epoch": 22.61, "learning_rate": 3.015492957746479e-06, "loss": 0.2562, "step": 20167 }, { "epoch": 22.61, "learning_rate": 3.0140845070422533e-06, "loss": 0.2564, "step": 20168 }, { "epoch": 22.61, "learning_rate": 3.0126760563380283e-06, "loss": 0.3174, "step": 20169 }, { "epoch": 22.61, "learning_rate": 3.0112676056338026e-06, "loss": 0.3195, "step": 20170 }, { "epoch": 22.61, "learning_rate": 3.0098591549295776e-06, "loss": 0.2787, "step": 20171 }, { "epoch": 22.61, "learning_rate": 3.008450704225352e-06, "loss": 0.2259, "step": 20172 }, { "epoch": 22.62, "learning_rate": 3.007042253521127e-06, "loss": 0.2361, "step": 20173 }, { "epoch": 22.62, "learning_rate": 3.005633802816901e-06, "loss": 0.3594, "step": 20174 }, { "epoch": 22.62, "learning_rate": 3.004225352112676e-06, "loss": 0.4233, "step": 20175 }, { "epoch": 22.62, "learning_rate": 3.002816901408451e-06, "loss": 0.4886, "step": 20176 }, { "epoch": 22.62, "learning_rate": 3.0014084507042254e-06, "loss": 0.3162, "step": 20177 }, { "epoch": 22.62, "learning_rate": 3e-06, "loss": 0.3942, "step": 20178 }, { "epoch": 22.62, "learning_rate": 2.9985915492957747e-06, "loss": 0.4813, "step": 20179 }, { "epoch": 22.62, "learning_rate": 2.9971830985915494e-06, "loss": 0.3749, "step": 20180 }, { "epoch": 22.62, "learning_rate": 2.995774647887324e-06, "loss": 0.339, "step": 20181 }, { "epoch": 22.63, "learning_rate": 2.9943661971830986e-06, "loss": 0.3324, "step": 20182 }, { "epoch": 22.63, "learning_rate": 2.9929577464788733e-06, "loss": 0.3336, "step": 20183 }, { "epoch": 22.63, "learning_rate": 2.991549295774648e-06, "loss": 0.3116, "step": 20184 }, { "epoch": 22.63, "learning_rate": 2.9901408450704225e-06, "loss": 0.3551, "step": 20185 }, { "epoch": 22.63, "learning_rate": 2.988732394366197e-06, "loss": 0.2986, "step": 20186 }, { "epoch": 22.63, "learning_rate": 2.987323943661972e-06, "loss": 0.3169, "step": 20187 }, { "epoch": 22.63, "learning_rate": 2.9859154929577465e-06, "loss": 0.2486, "step": 20188 }, { "epoch": 22.63, "learning_rate": 2.984507042253521e-06, "loss": 0.2989, "step": 20189 }, { "epoch": 22.63, "learning_rate": 2.9830985915492957e-06, "loss": 0.2503, "step": 20190 }, { "epoch": 22.64, "learning_rate": 2.9816901408450704e-06, "loss": 0.3694, "step": 20191 }, { "epoch": 22.64, "learning_rate": 2.980281690140845e-06, "loss": 0.3661, "step": 20192 }, { "epoch": 22.64, "learning_rate": 2.97887323943662e-06, "loss": 0.3499, "step": 20193 }, { "epoch": 22.64, "learning_rate": 2.9774647887323943e-06, "loss": 0.2921, "step": 20194 }, { "epoch": 22.64, "learning_rate": 2.9760563380281693e-06, "loss": 0.303, "step": 20195 }, { "epoch": 22.64, "learning_rate": 2.9746478873239436e-06, "loss": 0.2996, "step": 20196 }, { "epoch": 22.64, "learning_rate": 2.9732394366197186e-06, "loss": 0.3185, "step": 20197 }, { "epoch": 22.64, "learning_rate": 2.971830985915493e-06, "loss": 0.2609, "step": 20198 }, { "epoch": 22.64, "learning_rate": 2.970422535211268e-06, "loss": 0.288, "step": 20199 }, { "epoch": 22.65, "learning_rate": 2.969014084507042e-06, "loss": 0.2826, "step": 20200 }, { "epoch": 22.65, "learning_rate": 2.967605633802817e-06, "loss": 0.2932, "step": 20201 }, { "epoch": 22.65, "learning_rate": 2.9661971830985914e-06, "loss": 0.2761, "step": 20202 }, { "epoch": 22.65, "learning_rate": 2.9647887323943664e-06, "loss": 0.343, "step": 20203 }, { "epoch": 22.65, "learning_rate": 2.9633802816901407e-06, "loss": 0.256, "step": 20204 }, { "epoch": 22.65, "learning_rate": 2.9619718309859157e-06, "loss": 0.2961, "step": 20205 }, { "epoch": 22.65, "learning_rate": 2.96056338028169e-06, "loss": 0.2704, "step": 20206 }, { "epoch": 22.65, "learning_rate": 2.959154929577465e-06, "loss": 0.2872, "step": 20207 }, { "epoch": 22.65, "learning_rate": 2.957746478873239e-06, "loss": 0.2716, "step": 20208 }, { "epoch": 22.66, "learning_rate": 2.9563380281690143e-06, "loss": 0.2675, "step": 20209 }, { "epoch": 22.66, "learning_rate": 2.954929577464789e-06, "loss": 0.3152, "step": 20210 }, { "epoch": 22.66, "learning_rate": 2.9535211267605635e-06, "loss": 0.3679, "step": 20211 }, { "epoch": 22.66, "learning_rate": 2.952112676056338e-06, "loss": 0.2257, "step": 20212 }, { "epoch": 22.66, "learning_rate": 2.950704225352113e-06, "loss": 0.392, "step": 20213 }, { "epoch": 22.66, "learning_rate": 2.9492957746478874e-06, "loss": 0.3047, "step": 20214 }, { "epoch": 22.66, "learning_rate": 2.947887323943662e-06, "loss": 0.2999, "step": 20215 }, { "epoch": 22.66, "learning_rate": 2.9464788732394367e-06, "loss": 0.2676, "step": 20216 }, { "epoch": 22.66, "learning_rate": 2.9450704225352114e-06, "loss": 0.3288, "step": 20217 }, { "epoch": 22.67, "learning_rate": 2.943661971830986e-06, "loss": 0.2589, "step": 20218 }, { "epoch": 22.67, "learning_rate": 2.9422535211267606e-06, "loss": 0.1693, "step": 20219 }, { "epoch": 22.67, "learning_rate": 2.9408450704225353e-06, "loss": 0.2603, "step": 20220 }, { "epoch": 22.67, "learning_rate": 2.93943661971831e-06, "loss": 0.2012, "step": 20221 }, { "epoch": 22.67, "learning_rate": 2.9380281690140846e-06, "loss": 0.2276, "step": 20222 }, { "epoch": 22.67, "learning_rate": 2.936619718309859e-06, "loss": 0.2241, "step": 20223 }, { "epoch": 22.67, "learning_rate": 2.935211267605634e-06, "loss": 0.2611, "step": 20224 }, { "epoch": 22.67, "learning_rate": 2.9338028169014085e-06, "loss": 0.4645, "step": 20225 }, { "epoch": 22.67, "learning_rate": 2.932394366197183e-06, "loss": 0.3663, "step": 20226 }, { "epoch": 22.68, "learning_rate": 2.930985915492958e-06, "loss": 0.4362, "step": 20227 }, { "epoch": 22.68, "learning_rate": 2.9295774647887324e-06, "loss": 0.476, "step": 20228 }, { "epoch": 22.68, "learning_rate": 2.9281690140845074e-06, "loss": 0.3252, "step": 20229 }, { "epoch": 22.68, "learning_rate": 2.9267605633802817e-06, "loss": 0.3748, "step": 20230 }, { "epoch": 22.68, "learning_rate": 2.9253521126760567e-06, "loss": 0.442, "step": 20231 }, { "epoch": 22.68, "learning_rate": 2.923943661971831e-06, "loss": 0.3774, "step": 20232 }, { "epoch": 22.68, "learning_rate": 2.922535211267606e-06, "loss": 0.4046, "step": 20233 }, { "epoch": 22.68, "learning_rate": 2.92112676056338e-06, "loss": 0.3342, "step": 20234 }, { "epoch": 22.68, "learning_rate": 2.9197183098591553e-06, "loss": 0.3475, "step": 20235 }, { "epoch": 22.69, "learning_rate": 2.9183098591549295e-06, "loss": 0.3159, "step": 20236 }, { "epoch": 22.69, "learning_rate": 2.9169014084507045e-06, "loss": 0.3018, "step": 20237 }, { "epoch": 22.69, "learning_rate": 2.9154929577464788e-06, "loss": 0.3441, "step": 20238 }, { "epoch": 22.69, "learning_rate": 2.914084507042254e-06, "loss": 0.2905, "step": 20239 }, { "epoch": 22.69, "learning_rate": 2.912676056338028e-06, "loss": 0.2912, "step": 20240 }, { "epoch": 22.69, "learning_rate": 2.911267605633803e-06, "loss": 0.2951, "step": 20241 }, { "epoch": 22.69, "learning_rate": 2.9098591549295773e-06, "loss": 0.2602, "step": 20242 }, { "epoch": 22.69, "learning_rate": 2.9084507042253524e-06, "loss": 0.2728, "step": 20243 }, { "epoch": 22.7, "learning_rate": 2.907042253521127e-06, "loss": 0.3335, "step": 20244 }, { "epoch": 22.7, "learning_rate": 2.9056338028169016e-06, "loss": 0.3863, "step": 20245 }, { "epoch": 22.7, "learning_rate": 2.9042253521126763e-06, "loss": 0.2728, "step": 20246 }, { "epoch": 22.7, "learning_rate": 2.902816901408451e-06, "loss": 0.2817, "step": 20247 }, { "epoch": 22.7, "learning_rate": 2.9014084507042255e-06, "loss": 0.362, "step": 20248 }, { "epoch": 22.7, "learning_rate": 2.9e-06, "loss": 0.3446, "step": 20249 }, { "epoch": 22.7, "learning_rate": 2.898591549295775e-06, "loss": 0.387, "step": 20250 }, { "epoch": 22.7, "learning_rate": 2.8971830985915495e-06, "loss": 0.3336, "step": 20251 }, { "epoch": 22.7, "learning_rate": 2.895774647887324e-06, "loss": 0.2978, "step": 20252 }, { "epoch": 22.71, "learning_rate": 2.8943661971830987e-06, "loss": 0.2389, "step": 20253 }, { "epoch": 22.71, "learning_rate": 2.8929577464788734e-06, "loss": 0.2995, "step": 20254 }, { "epoch": 22.71, "learning_rate": 2.891549295774648e-06, "loss": 0.3122, "step": 20255 }, { "epoch": 22.71, "learning_rate": 2.8901408450704226e-06, "loss": 0.2912, "step": 20256 }, { "epoch": 22.71, "learning_rate": 2.8887323943661973e-06, "loss": 0.3151, "step": 20257 }, { "epoch": 22.71, "learning_rate": 2.887323943661972e-06, "loss": 0.2637, "step": 20258 }, { "epoch": 22.71, "learning_rate": 2.8859154929577466e-06, "loss": 0.3308, "step": 20259 }, { "epoch": 22.71, "learning_rate": 2.884507042253521e-06, "loss": 0.3188, "step": 20260 }, { "epoch": 22.71, "learning_rate": 2.883098591549296e-06, "loss": 0.2848, "step": 20261 }, { "epoch": 22.72, "learning_rate": 2.8816901408450705e-06, "loss": 0.3658, "step": 20262 }, { "epoch": 22.72, "learning_rate": 2.8802816901408455e-06, "loss": 0.2511, "step": 20263 }, { "epoch": 22.72, "learning_rate": 2.8788732394366197e-06, "loss": 0.3002, "step": 20264 }, { "epoch": 22.72, "learning_rate": 2.877464788732395e-06, "loss": 0.243, "step": 20265 }, { "epoch": 22.72, "learning_rate": 2.876056338028169e-06, "loss": 0.2773, "step": 20266 }, { "epoch": 22.72, "learning_rate": 2.874647887323944e-06, "loss": 0.2713, "step": 20267 }, { "epoch": 22.72, "learning_rate": 2.8732394366197183e-06, "loss": 0.2819, "step": 20268 }, { "epoch": 22.72, "learning_rate": 2.871830985915493e-06, "loss": 0.1997, "step": 20269 }, { "epoch": 22.72, "learning_rate": 2.8704225352112676e-06, "loss": 0.2448, "step": 20270 }, { "epoch": 22.73, "learning_rate": 2.869014084507042e-06, "loss": 0.1595, "step": 20271 }, { "epoch": 22.73, "learning_rate": 2.867605633802817e-06, "loss": 0.2808, "step": 20272 }, { "epoch": 22.73, "learning_rate": 2.8661971830985915e-06, "loss": 0.2682, "step": 20273 }, { "epoch": 22.73, "learning_rate": 2.864788732394366e-06, "loss": 0.2042, "step": 20274 }, { "epoch": 22.73, "learning_rate": 2.8633802816901408e-06, "loss": 0.4096, "step": 20275 }, { "epoch": 22.73, "learning_rate": 2.8619718309859154e-06, "loss": 0.4225, "step": 20276 }, { "epoch": 22.73, "learning_rate": 2.86056338028169e-06, "loss": 0.3228, "step": 20277 }, { "epoch": 22.73, "learning_rate": 2.8591549295774647e-06, "loss": 0.3487, "step": 20278 }, { "epoch": 22.73, "learning_rate": 2.8577464788732393e-06, "loss": 0.3562, "step": 20279 }, { "epoch": 22.74, "learning_rate": 2.8563380281690144e-06, "loss": 0.2498, "step": 20280 }, { "epoch": 22.74, "learning_rate": 2.8549295774647886e-06, "loss": 0.2576, "step": 20281 }, { "epoch": 22.74, "learning_rate": 2.8535211267605636e-06, "loss": 0.3725, "step": 20282 }, { "epoch": 22.74, "learning_rate": 2.852112676056338e-06, "loss": 0.288, "step": 20283 }, { "epoch": 22.74, "learning_rate": 2.850704225352113e-06, "loss": 0.3271, "step": 20284 }, { "epoch": 22.74, "learning_rate": 2.849295774647887e-06, "loss": 0.3399, "step": 20285 }, { "epoch": 22.74, "learning_rate": 2.847887323943662e-06, "loss": 0.3246, "step": 20286 }, { "epoch": 22.74, "learning_rate": 2.8464788732394364e-06, "loss": 0.4537, "step": 20287 }, { "epoch": 22.74, "learning_rate": 2.8450704225352115e-06, "loss": 0.4082, "step": 20288 }, { "epoch": 22.75, "learning_rate": 2.8436619718309857e-06, "loss": 0.4295, "step": 20289 }, { "epoch": 22.75, "learning_rate": 2.8422535211267607e-06, "loss": 0.3391, "step": 20290 }, { "epoch": 22.75, "learning_rate": 2.840845070422535e-06, "loss": 0.3538, "step": 20291 }, { "epoch": 22.75, "learning_rate": 2.83943661971831e-06, "loss": 0.3572, "step": 20292 }, { "epoch": 22.75, "learning_rate": 2.8380281690140842e-06, "loss": 0.306, "step": 20293 }, { "epoch": 22.75, "learning_rate": 2.8366197183098593e-06, "loss": 0.3048, "step": 20294 }, { "epoch": 22.75, "learning_rate": 2.8352112676056335e-06, "loss": 0.2495, "step": 20295 }, { "epoch": 22.75, "learning_rate": 2.8338028169014086e-06, "loss": 0.2969, "step": 20296 }, { "epoch": 22.75, "learning_rate": 2.832394366197183e-06, "loss": 0.2676, "step": 20297 }, { "epoch": 22.76, "learning_rate": 2.830985915492958e-06, "loss": 0.3281, "step": 20298 }, { "epoch": 22.76, "learning_rate": 2.8295774647887325e-06, "loss": 0.3116, "step": 20299 }, { "epoch": 22.76, "learning_rate": 2.8295774647887325e-06, "loss": 0.332, "step": 20300 }, { "epoch": 22.76, "learning_rate": 2.828169014084507e-06, "loss": 0.4121, "step": 20301 }, { "epoch": 22.76, "learning_rate": 2.8267605633802818e-06, "loss": 0.2944, "step": 20302 }, { "epoch": 22.76, "learning_rate": 2.8253521126760564e-06, "loss": 0.2186, "step": 20303 }, { "epoch": 22.76, "learning_rate": 2.823943661971831e-06, "loss": 0.2905, "step": 20304 }, { "epoch": 22.76, "learning_rate": 2.8225352112676057e-06, "loss": 0.2666, "step": 20305 }, { "epoch": 22.76, "learning_rate": 2.8211267605633803e-06, "loss": 0.27, "step": 20306 }, { "epoch": 22.77, "learning_rate": 2.819718309859155e-06, "loss": 0.2891, "step": 20307 }, { "epoch": 22.77, "learning_rate": 2.8183098591549296e-06, "loss": 0.4067, "step": 20308 }, { "epoch": 22.77, "learning_rate": 2.8169014084507042e-06, "loss": 0.2324, "step": 20309 }, { "epoch": 22.77, "learning_rate": 2.815492957746479e-06, "loss": 0.3009, "step": 20310 }, { "epoch": 22.77, "learning_rate": 2.8140845070422535e-06, "loss": 0.3264, "step": 20311 }, { "epoch": 22.77, "learning_rate": 2.812676056338028e-06, "loss": 0.2634, "step": 20312 }, { "epoch": 22.77, "learning_rate": 2.8112676056338028e-06, "loss": 0.3091, "step": 20313 }, { "epoch": 22.77, "learning_rate": 2.8098591549295774e-06, "loss": 0.3308, "step": 20314 }, { "epoch": 22.77, "learning_rate": 2.8084507042253525e-06, "loss": 0.2658, "step": 20315 }, { "epoch": 22.78, "learning_rate": 2.8070422535211267e-06, "loss": 0.3031, "step": 20316 }, { "epoch": 22.78, "learning_rate": 2.8056338028169017e-06, "loss": 0.2485, "step": 20317 }, { "epoch": 22.78, "learning_rate": 2.804225352112676e-06, "loss": 0.2694, "step": 20318 }, { "epoch": 22.78, "learning_rate": 2.802816901408451e-06, "loss": 0.2698, "step": 20319 }, { "epoch": 22.78, "learning_rate": 2.8014084507042252e-06, "loss": 0.2258, "step": 20320 }, { "epoch": 22.78, "learning_rate": 2.8000000000000003e-06, "loss": 0.2753, "step": 20321 }, { "epoch": 22.78, "learning_rate": 2.7985915492957745e-06, "loss": 0.2499, "step": 20322 }, { "epoch": 22.78, "learning_rate": 2.7971830985915496e-06, "loss": 0.2959, "step": 20323 }, { "epoch": 22.78, "learning_rate": 2.7957746478873238e-06, "loss": 0.1856, "step": 20324 }, { "epoch": 22.79, "learning_rate": 2.794366197183099e-06, "loss": 0.454, "step": 20325 }, { "epoch": 22.79, "learning_rate": 2.792957746478873e-06, "loss": 0.3185, "step": 20326 }, { "epoch": 22.79, "learning_rate": 2.791549295774648e-06, "loss": 0.4846, "step": 20327 }, { "epoch": 22.79, "learning_rate": 2.7901408450704223e-06, "loss": 0.3805, "step": 20328 }, { "epoch": 22.79, "learning_rate": 2.7887323943661974e-06, "loss": 0.2737, "step": 20329 }, { "epoch": 22.79, "learning_rate": 2.7873239436619716e-06, "loss": 0.3815, "step": 20330 }, { "epoch": 22.79, "learning_rate": 2.7859154929577467e-06, "loss": 0.3139, "step": 20331 }, { "epoch": 22.79, "learning_rate": 2.7845070422535213e-06, "loss": 0.3219, "step": 20332 }, { "epoch": 22.79, "learning_rate": 2.783098591549296e-06, "loss": 0.3337, "step": 20333 }, { "epoch": 22.8, "learning_rate": 2.7816901408450706e-06, "loss": 0.3555, "step": 20334 }, { "epoch": 22.8, "learning_rate": 2.7802816901408452e-06, "loss": 0.3406, "step": 20335 }, { "epoch": 22.8, "learning_rate": 2.77887323943662e-06, "loss": 0.3324, "step": 20336 }, { "epoch": 22.8, "learning_rate": 2.7774647887323945e-06, "loss": 0.2767, "step": 20337 }, { "epoch": 22.8, "learning_rate": 2.776056338028169e-06, "loss": 0.3366, "step": 20338 }, { "epoch": 22.8, "learning_rate": 2.7746478873239438e-06, "loss": 0.2757, "step": 20339 }, { "epoch": 22.8, "learning_rate": 2.7732394366197184e-06, "loss": 0.3603, "step": 20340 }, { "epoch": 22.8, "learning_rate": 2.771830985915493e-06, "loss": 0.2723, "step": 20341 }, { "epoch": 22.8, "learning_rate": 2.7704225352112677e-06, "loss": 0.3034, "step": 20342 }, { "epoch": 22.81, "learning_rate": 2.7690140845070423e-06, "loss": 0.3993, "step": 20343 }, { "epoch": 22.81, "learning_rate": 2.767605633802817e-06, "loss": 0.4111, "step": 20344 }, { "epoch": 22.81, "learning_rate": 2.7661971830985916e-06, "loss": 0.3045, "step": 20345 }, { "epoch": 22.81, "learning_rate": 2.7647887323943662e-06, "loss": 0.2698, "step": 20346 }, { "epoch": 22.81, "learning_rate": 2.763380281690141e-06, "loss": 0.2845, "step": 20347 }, { "epoch": 22.81, "learning_rate": 2.7619718309859155e-06, "loss": 0.325, "step": 20348 }, { "epoch": 22.81, "learning_rate": 2.7605633802816906e-06, "loss": 0.3025, "step": 20349 }, { "epoch": 22.81, "learning_rate": 2.7591549295774648e-06, "loss": 0.3933, "step": 20350 }, { "epoch": 22.82, "learning_rate": 2.75774647887324e-06, "loss": 0.3077, "step": 20351 }, { "epoch": 22.82, "learning_rate": 2.756338028169014e-06, "loss": 0.2986, "step": 20352 }, { "epoch": 22.82, "learning_rate": 2.754929577464789e-06, "loss": 0.294, "step": 20353 }, { "epoch": 22.82, "learning_rate": 2.7535211267605633e-06, "loss": 0.3025, "step": 20354 }, { "epoch": 22.82, "learning_rate": 2.7521126760563384e-06, "loss": 0.2841, "step": 20355 }, { "epoch": 22.82, "learning_rate": 2.7507042253521126e-06, "loss": 0.283, "step": 20356 }, { "epoch": 22.82, "learning_rate": 2.7492957746478877e-06, "loss": 0.2412, "step": 20357 }, { "epoch": 22.82, "learning_rate": 2.747887323943662e-06, "loss": 0.3104, "step": 20358 }, { "epoch": 22.82, "learning_rate": 2.746478873239437e-06, "loss": 0.2948, "step": 20359 }, { "epoch": 22.83, "learning_rate": 2.745070422535211e-06, "loss": 0.2955, "step": 20360 }, { "epoch": 22.83, "learning_rate": 2.7436619718309862e-06, "loss": 0.3316, "step": 20361 }, { "epoch": 22.83, "learning_rate": 2.7422535211267604e-06, "loss": 0.2862, "step": 20362 }, { "epoch": 22.83, "learning_rate": 2.7408450704225355e-06, "loss": 0.2339, "step": 20363 }, { "epoch": 22.83, "learning_rate": 2.7394366197183097e-06, "loss": 0.2783, "step": 20364 }, { "epoch": 22.83, "learning_rate": 2.7380281690140848e-06, "loss": 0.3287, "step": 20365 }, { "epoch": 22.83, "learning_rate": 2.736619718309859e-06, "loss": 0.3303, "step": 20366 }, { "epoch": 22.83, "learning_rate": 2.735211267605634e-06, "loss": 0.2927, "step": 20367 }, { "epoch": 22.83, "learning_rate": 2.7338028169014087e-06, "loss": 0.369, "step": 20368 }, { "epoch": 22.84, "learning_rate": 2.7323943661971833e-06, "loss": 0.2771, "step": 20369 }, { "epoch": 22.84, "learning_rate": 2.730985915492958e-06, "loss": 0.3138, "step": 20370 }, { "epoch": 22.84, "learning_rate": 2.7295774647887326e-06, "loss": 0.2318, "step": 20371 }, { "epoch": 22.84, "learning_rate": 2.7281690140845072e-06, "loss": 0.2599, "step": 20372 }, { "epoch": 22.84, "learning_rate": 2.726760563380282e-06, "loss": 0.218, "step": 20373 }, { "epoch": 22.84, "learning_rate": 2.7253521126760565e-06, "loss": 0.2869, "step": 20374 }, { "epoch": 22.84, "learning_rate": 2.723943661971831e-06, "loss": 0.5213, "step": 20375 }, { "epoch": 22.84, "learning_rate": 2.7225352112676058e-06, "loss": 0.3579, "step": 20376 }, { "epoch": 22.84, "learning_rate": 2.7211267605633804e-06, "loss": 0.2718, "step": 20377 }, { "epoch": 22.85, "learning_rate": 2.719718309859155e-06, "loss": 0.3255, "step": 20378 }, { "epoch": 22.85, "learning_rate": 2.7183098591549297e-06, "loss": 0.3889, "step": 20379 }, { "epoch": 22.85, "learning_rate": 2.7169014084507043e-06, "loss": 0.2432, "step": 20380 }, { "epoch": 22.85, "learning_rate": 2.715492957746479e-06, "loss": 0.3443, "step": 20381 }, { "epoch": 22.85, "learning_rate": 2.7140845070422536e-06, "loss": 0.383, "step": 20382 }, { "epoch": 22.85, "learning_rate": 2.7126760563380282e-06, "loss": 0.3226, "step": 20383 }, { "epoch": 22.85, "learning_rate": 2.711267605633803e-06, "loss": 0.3339, "step": 20384 }, { "epoch": 22.85, "learning_rate": 2.7098591549295775e-06, "loss": 0.2833, "step": 20385 }, { "epoch": 22.85, "learning_rate": 2.708450704225352e-06, "loss": 0.3691, "step": 20386 }, { "epoch": 22.86, "learning_rate": 2.707042253521127e-06, "loss": 0.3519, "step": 20387 }, { "epoch": 22.86, "learning_rate": 2.7056338028169014e-06, "loss": 0.3258, "step": 20388 }, { "epoch": 22.86, "learning_rate": 2.704225352112676e-06, "loss": 0.2993, "step": 20389 }, { "epoch": 22.86, "learning_rate": 2.7028169014084507e-06, "loss": 0.2613, "step": 20390 }, { "epoch": 22.86, "learning_rate": 2.7014084507042253e-06, "loss": 0.3719, "step": 20391 }, { "epoch": 22.86, "learning_rate": 2.7e-06, "loss": 0.2789, "step": 20392 }, { "epoch": 22.86, "learning_rate": 2.6985915492957746e-06, "loss": 0.3563, "step": 20393 }, { "epoch": 22.86, "learning_rate": 2.6971830985915493e-06, "loss": 0.2778, "step": 20394 }, { "epoch": 22.86, "learning_rate": 2.695774647887324e-06, "loss": 0.3257, "step": 20395 }, { "epoch": 22.87, "learning_rate": 2.6943661971830985e-06, "loss": 0.2939, "step": 20396 }, { "epoch": 22.87, "learning_rate": 2.692957746478873e-06, "loss": 0.3016, "step": 20397 }, { "epoch": 22.87, "learning_rate": 2.691549295774648e-06, "loss": 0.2575, "step": 20398 }, { "epoch": 22.87, "learning_rate": 2.6901408450704224e-06, "loss": 0.2435, "step": 20399 }, { "epoch": 22.87, "learning_rate": 2.688732394366197e-06, "loss": 0.2668, "step": 20400 }, { "epoch": 22.87, "learning_rate": 2.6873239436619717e-06, "loss": 0.329, "step": 20401 }, { "epoch": 22.87, "learning_rate": 2.6859154929577468e-06, "loss": 0.2924, "step": 20402 }, { "epoch": 22.87, "learning_rate": 2.684507042253521e-06, "loss": 0.3332, "step": 20403 }, { "epoch": 22.87, "learning_rate": 2.683098591549296e-06, "loss": 0.3622, "step": 20404 }, { "epoch": 22.88, "learning_rate": 2.6816901408450703e-06, "loss": 0.3788, "step": 20405 }, { "epoch": 22.88, "learning_rate": 2.6802816901408453e-06, "loss": 0.2989, "step": 20406 }, { "epoch": 22.88, "learning_rate": 2.6788732394366195e-06, "loss": 0.2884, "step": 20407 }, { "epoch": 22.88, "learning_rate": 2.6774647887323946e-06, "loss": 0.2513, "step": 20408 }, { "epoch": 22.88, "learning_rate": 2.676056338028169e-06, "loss": 0.2558, "step": 20409 }, { "epoch": 22.88, "learning_rate": 2.674647887323944e-06, "loss": 0.2577, "step": 20410 }, { "epoch": 22.88, "learning_rate": 2.673239436619718e-06, "loss": 0.2682, "step": 20411 }, { "epoch": 22.88, "learning_rate": 2.671830985915493e-06, "loss": 0.3659, "step": 20412 }, { "epoch": 22.88, "learning_rate": 2.6704225352112674e-06, "loss": 0.2518, "step": 20413 }, { "epoch": 22.89, "learning_rate": 2.6690140845070424e-06, "loss": 0.3338, "step": 20414 }, { "epoch": 22.89, "learning_rate": 2.6676056338028166e-06, "loss": 0.3374, "step": 20415 }, { "epoch": 22.89, "learning_rate": 2.6661971830985917e-06, "loss": 0.3085, "step": 20416 }, { "epoch": 22.89, "learning_rate": 2.664788732394366e-06, "loss": 0.1968, "step": 20417 }, { "epoch": 22.89, "learning_rate": 2.663380281690141e-06, "loss": 0.259, "step": 20418 }, { "epoch": 22.89, "learning_rate": 2.6619718309859156e-06, "loss": 0.2654, "step": 20419 }, { "epoch": 22.89, "learning_rate": 2.6605633802816903e-06, "loss": 0.2537, "step": 20420 }, { "epoch": 22.89, "learning_rate": 2.659154929577465e-06, "loss": 0.2151, "step": 20421 }, { "epoch": 22.89, "learning_rate": 2.6577464788732395e-06, "loss": 0.4084, "step": 20422 }, { "epoch": 22.9, "learning_rate": 2.656338028169014e-06, "loss": 0.2947, "step": 20423 }, { "epoch": 22.9, "learning_rate": 2.654929577464789e-06, "loss": 0.3243, "step": 20424 }, { "epoch": 22.9, "learning_rate": 2.6535211267605634e-06, "loss": 0.4482, "step": 20425 }, { "epoch": 22.9, "learning_rate": 2.652112676056338e-06, "loss": 0.3449, "step": 20426 }, { "epoch": 22.9, "learning_rate": 2.6507042253521127e-06, "loss": 0.3987, "step": 20427 }, { "epoch": 22.9, "learning_rate": 2.6492957746478874e-06, "loss": 0.3934, "step": 20428 }, { "epoch": 22.9, "learning_rate": 2.647887323943662e-06, "loss": 0.3767, "step": 20429 }, { "epoch": 22.9, "learning_rate": 2.6464788732394366e-06, "loss": 0.3363, "step": 20430 }, { "epoch": 22.9, "learning_rate": 2.6450704225352113e-06, "loss": 0.2907, "step": 20431 }, { "epoch": 22.91, "learning_rate": 2.643661971830986e-06, "loss": 0.3999, "step": 20432 }, { "epoch": 22.91, "learning_rate": 2.6422535211267605e-06, "loss": 0.3329, "step": 20433 }, { "epoch": 22.91, "learning_rate": 2.640845070422535e-06, "loss": 0.3314, "step": 20434 }, { "epoch": 22.91, "learning_rate": 2.63943661971831e-06, "loss": 0.3542, "step": 20435 }, { "epoch": 22.91, "learning_rate": 2.638028169014085e-06, "loss": 0.2919, "step": 20436 }, { "epoch": 22.91, "learning_rate": 2.636619718309859e-06, "loss": 0.2695, "step": 20437 }, { "epoch": 22.91, "learning_rate": 2.635211267605634e-06, "loss": 0.3285, "step": 20438 }, { "epoch": 22.91, "learning_rate": 2.6338028169014084e-06, "loss": 0.2954, "step": 20439 }, { "epoch": 22.91, "learning_rate": 2.6323943661971834e-06, "loss": 0.3847, "step": 20440 }, { "epoch": 22.92, "learning_rate": 2.6309859154929576e-06, "loss": 0.3649, "step": 20441 }, { "epoch": 22.92, "learning_rate": 2.6295774647887327e-06, "loss": 0.2631, "step": 20442 }, { "epoch": 22.92, "learning_rate": 2.628169014084507e-06, "loss": 0.2915, "step": 20443 }, { "epoch": 22.92, "learning_rate": 2.626760563380282e-06, "loss": 0.325, "step": 20444 }, { "epoch": 22.92, "learning_rate": 2.625352112676056e-06, "loss": 0.3503, "step": 20445 }, { "epoch": 22.92, "learning_rate": 2.6239436619718313e-06, "loss": 0.3295, "step": 20446 }, { "epoch": 22.92, "learning_rate": 2.6225352112676055e-06, "loss": 0.2512, "step": 20447 }, { "epoch": 22.92, "learning_rate": 2.6211267605633805e-06, "loss": 0.2493, "step": 20448 }, { "epoch": 22.92, "learning_rate": 2.6197183098591547e-06, "loss": 0.2671, "step": 20449 }, { "epoch": 22.93, "learning_rate": 2.61830985915493e-06, "loss": 0.2587, "step": 20450 }, { "epoch": 22.93, "learning_rate": 2.616901408450704e-06, "loss": 0.3235, "step": 20451 }, { "epoch": 22.93, "learning_rate": 2.615492957746479e-06, "loss": 0.2926, "step": 20452 }, { "epoch": 22.93, "learning_rate": 2.6140845070422533e-06, "loss": 0.3504, "step": 20453 }, { "epoch": 22.93, "learning_rate": 2.6126760563380284e-06, "loss": 0.293, "step": 20454 }, { "epoch": 22.93, "learning_rate": 2.611267605633803e-06, "loss": 0.2992, "step": 20455 }, { "epoch": 22.93, "learning_rate": 2.6098591549295776e-06, "loss": 0.3175, "step": 20456 }, { "epoch": 22.93, "learning_rate": 2.6084507042253523e-06, "loss": 0.2644, "step": 20457 }, { "epoch": 22.93, "learning_rate": 2.607042253521127e-06, "loss": 0.2678, "step": 20458 }, { "epoch": 22.94, "learning_rate": 2.6056338028169015e-06, "loss": 0.2724, "step": 20459 }, { "epoch": 22.94, "learning_rate": 2.604225352112676e-06, "loss": 0.2758, "step": 20460 }, { "epoch": 22.94, "learning_rate": 2.602816901408451e-06, "loss": 0.2632, "step": 20461 }, { "epoch": 22.94, "learning_rate": 2.6014084507042255e-06, "loss": 0.236, "step": 20462 }, { "epoch": 22.94, "learning_rate": 2.6e-06, "loss": 0.3104, "step": 20463 }, { "epoch": 22.94, "learning_rate": 2.5985915492957747e-06, "loss": 0.2746, "step": 20464 }, { "epoch": 22.94, "learning_rate": 2.5971830985915494e-06, "loss": 0.3453, "step": 20465 }, { "epoch": 22.94, "learning_rate": 2.595774647887324e-06, "loss": 0.3053, "step": 20466 }, { "epoch": 22.95, "learning_rate": 2.5943661971830986e-06, "loss": 0.2716, "step": 20467 }, { "epoch": 22.95, "learning_rate": 2.5929577464788733e-06, "loss": 0.2758, "step": 20468 }, { "epoch": 22.95, "learning_rate": 2.591549295774648e-06, "loss": 0.2662, "step": 20469 }, { "epoch": 22.95, "learning_rate": 2.5901408450704226e-06, "loss": 0.2284, "step": 20470 }, { "epoch": 22.95, "learning_rate": 2.588732394366197e-06, "loss": 0.1646, "step": 20471 }, { "epoch": 22.95, "learning_rate": 2.5873239436619723e-06, "loss": 0.3575, "step": 20472 }, { "epoch": 22.95, "learning_rate": 2.5859154929577465e-06, "loss": 0.2663, "step": 20473 }, { "epoch": 22.95, "learning_rate": 2.5845070422535215e-06, "loss": 0.2279, "step": 20474 }, { "epoch": 22.95, "learning_rate": 2.5830985915492957e-06, "loss": 0.4114, "step": 20475 }, { "epoch": 22.96, "learning_rate": 2.581690140845071e-06, "loss": 0.4445, "step": 20476 }, { "epoch": 22.96, "learning_rate": 2.580281690140845e-06, "loss": 0.3795, "step": 20477 }, { "epoch": 22.96, "learning_rate": 2.57887323943662e-06, "loss": 0.3451, "step": 20478 }, { "epoch": 22.96, "learning_rate": 2.5774647887323943e-06, "loss": 0.3742, "step": 20479 }, { "epoch": 22.96, "learning_rate": 2.5760563380281694e-06, "loss": 0.341, "step": 20480 }, { "epoch": 22.96, "learning_rate": 2.5746478873239436e-06, "loss": 0.3304, "step": 20481 }, { "epoch": 22.96, "learning_rate": 2.5732394366197186e-06, "loss": 0.3135, "step": 20482 }, { "epoch": 22.96, "learning_rate": 2.571830985915493e-06, "loss": 0.2926, "step": 20483 }, { "epoch": 22.96, "learning_rate": 2.570422535211268e-06, "loss": 0.3504, "step": 20484 }, { "epoch": 22.97, "learning_rate": 2.569014084507042e-06, "loss": 0.2807, "step": 20485 }, { "epoch": 22.97, "learning_rate": 2.567605633802817e-06, "loss": 0.3225, "step": 20486 }, { "epoch": 22.97, "learning_rate": 2.5661971830985914e-06, "loss": 0.3129, "step": 20487 }, { "epoch": 22.97, "learning_rate": 2.5647887323943665e-06, "loss": 0.3826, "step": 20488 }, { "epoch": 22.97, "learning_rate": 2.563380281690141e-06, "loss": 0.3148, "step": 20489 }, { "epoch": 22.97, "learning_rate": 2.5619718309859157e-06, "loss": 0.2661, "step": 20490 }, { "epoch": 22.97, "learning_rate": 2.5605633802816904e-06, "loss": 0.2875, "step": 20491 }, { "epoch": 22.97, "learning_rate": 2.559154929577465e-06, "loss": 0.2883, "step": 20492 }, { "epoch": 22.97, "learning_rate": 2.5577464788732396e-06, "loss": 0.3865, "step": 20493 }, { "epoch": 22.98, "learning_rate": 2.5563380281690143e-06, "loss": 0.2537, "step": 20494 }, { "epoch": 22.98, "learning_rate": 2.554929577464789e-06, "loss": 0.273, "step": 20495 }, { "epoch": 22.98, "learning_rate": 2.5535211267605636e-06, "loss": 0.3425, "step": 20496 }, { "epoch": 22.98, "learning_rate": 2.552112676056338e-06, "loss": 0.3058, "step": 20497 }, { "epoch": 22.98, "learning_rate": 2.5507042253521124e-06, "loss": 0.3363, "step": 20498 }, { "epoch": 22.98, "learning_rate": 2.5492957746478875e-06, "loss": 0.3005, "step": 20499 }, { "epoch": 22.98, "learning_rate": 2.5478873239436617e-06, "loss": 0.2749, "step": 20500 }, { "epoch": 22.98, "learning_rate": 2.5464788732394367e-06, "loss": 0.3599, "step": 20501 }, { "epoch": 22.98, "learning_rate": 2.545070422535211e-06, "loss": 0.2793, "step": 20502 }, { "epoch": 22.99, "learning_rate": 2.543661971830986e-06, "loss": 0.3533, "step": 20503 }, { "epoch": 22.99, "learning_rate": 2.5422535211267602e-06, "loss": 0.3401, "step": 20504 }, { "epoch": 22.99, "learning_rate": 2.5408450704225353e-06, "loss": 0.2858, "step": 20505 }, { "epoch": 22.99, "learning_rate": 2.53943661971831e-06, "loss": 0.3194, "step": 20506 }, { "epoch": 22.99, "learning_rate": 2.5380281690140846e-06, "loss": 0.2565, "step": 20507 }, { "epoch": 22.99, "learning_rate": 2.536619718309859e-06, "loss": 0.3612, "step": 20508 }, { "epoch": 22.99, "learning_rate": 2.535211267605634e-06, "loss": 0.2251, "step": 20509 }, { "epoch": 22.99, "learning_rate": 2.5338028169014085e-06, "loss": 0.3475, "step": 20510 }, { "epoch": 22.99, "learning_rate": 2.532394366197183e-06, "loss": 0.2026, "step": 20511 }, { "epoch": 23.0, "learning_rate": 2.5309859154929578e-06, "loss": 0.3793, "step": 20512 }, { "epoch": 23.0, "learning_rate": 2.5295774647887324e-06, "loss": 0.2618, "step": 20513 }, { "epoch": 23.0, "learning_rate": 2.528169014084507e-06, "loss": 0.2568, "step": 20514 }, { "epoch": 23.0, "learning_rate": 2.5267605633802817e-06, "loss": 0.3135, "step": 20515 }, { "epoch": 23.0, "learning_rate": 2.5253521126760563e-06, "loss": 0.4653, "step": 20516 }, { "epoch": 23.0, "learning_rate": 2.523943661971831e-06, "loss": 0.4114, "step": 20517 }, { "epoch": 23.0, "learning_rate": 2.5225352112676056e-06, "loss": 0.4412, "step": 20518 }, { "epoch": 23.0, "learning_rate": 2.5211267605633802e-06, "loss": 0.3201, "step": 20519 }, { "epoch": 23.0, "learning_rate": 2.519718309859155e-06, "loss": 0.3909, "step": 20520 }, { "epoch": 23.01, "learning_rate": 2.5183098591549295e-06, "loss": 0.395, "step": 20521 }, { "epoch": 23.01, "learning_rate": 2.516901408450704e-06, "loss": 0.3251, "step": 20522 }, { "epoch": 23.01, "learning_rate": 2.515492957746479e-06, "loss": 0.3287, "step": 20523 }, { "epoch": 23.01, "learning_rate": 2.5140845070422534e-06, "loss": 0.3669, "step": 20524 }, { "epoch": 23.01, "learning_rate": 2.5126760563380285e-06, "loss": 0.3332, "step": 20525 }, { "epoch": 23.01, "learning_rate": 2.5112676056338027e-06, "loss": 0.4213, "step": 20526 }, { "epoch": 23.01, "learning_rate": 2.5098591549295777e-06, "loss": 0.3177, "step": 20527 }, { "epoch": 23.01, "learning_rate": 2.508450704225352e-06, "loss": 0.2626, "step": 20528 }, { "epoch": 23.01, "learning_rate": 2.507042253521127e-06, "loss": 0.3542, "step": 20529 }, { "epoch": 23.02, "learning_rate": 2.5056338028169012e-06, "loss": 0.3218, "step": 20530 }, { "epoch": 23.02, "learning_rate": 2.5042253521126763e-06, "loss": 0.2977, "step": 20531 }, { "epoch": 23.02, "learning_rate": 2.5028169014084505e-06, "loss": 0.3022, "step": 20532 }, { "epoch": 23.02, "learning_rate": 2.5014084507042256e-06, "loss": 0.3097, "step": 20533 }, { "epoch": 23.02, "learning_rate": 2.4999999999999998e-06, "loss": 0.3077, "step": 20534 }, { "epoch": 23.02, "learning_rate": 2.498591549295775e-06, "loss": 0.2773, "step": 20535 }, { "epoch": 23.02, "learning_rate": 2.497183098591549e-06, "loss": 0.3513, "step": 20536 }, { "epoch": 23.02, "learning_rate": 2.495774647887324e-06, "loss": 0.3118, "step": 20537 }, { "epoch": 23.02, "learning_rate": 2.4943661971830983e-06, "loss": 0.2524, "step": 20538 }, { "epoch": 23.03, "learning_rate": 2.4929577464788734e-06, "loss": 0.2456, "step": 20539 }, { "epoch": 23.03, "learning_rate": 2.491549295774648e-06, "loss": 0.3164, "step": 20540 }, { "epoch": 23.03, "learning_rate": 2.4901408450704227e-06, "loss": 0.3607, "step": 20541 }, { "epoch": 23.03, "learning_rate": 2.4887323943661973e-06, "loss": 0.3823, "step": 20542 }, { "epoch": 23.03, "learning_rate": 2.487323943661972e-06, "loss": 0.3042, "step": 20543 }, { "epoch": 23.03, "learning_rate": 2.4859154929577466e-06, "loss": 0.2623, "step": 20544 }, { "epoch": 23.03, "learning_rate": 2.4845070422535212e-06, "loss": 0.2734, "step": 20545 }, { "epoch": 23.03, "learning_rate": 2.483098591549296e-06, "loss": 0.4019, "step": 20546 }, { "epoch": 23.03, "learning_rate": 2.4816901408450705e-06, "loss": 0.3259, "step": 20547 }, { "epoch": 23.04, "learning_rate": 2.480281690140845e-06, "loss": 0.2083, "step": 20548 }, { "epoch": 23.04, "learning_rate": 2.4788732394366198e-06, "loss": 0.2932, "step": 20549 }, { "epoch": 23.04, "learning_rate": 2.4774647887323944e-06, "loss": 0.3379, "step": 20550 }, { "epoch": 23.04, "learning_rate": 2.476056338028169e-06, "loss": 0.2595, "step": 20551 }, { "epoch": 23.04, "learning_rate": 2.4746478873239437e-06, "loss": 0.4104, "step": 20552 }, { "epoch": 23.04, "learning_rate": 2.4732394366197183e-06, "loss": 0.2568, "step": 20553 }, { "epoch": 23.04, "learning_rate": 2.471830985915493e-06, "loss": 0.2434, "step": 20554 }, { "epoch": 23.04, "learning_rate": 2.4704225352112676e-06, "loss": 0.3578, "step": 20555 }, { "epoch": 23.04, "learning_rate": 2.4690140845070422e-06, "loss": 0.3022, "step": 20556 }, { "epoch": 23.05, "learning_rate": 2.467605633802817e-06, "loss": 0.2944, "step": 20557 }, { "epoch": 23.05, "learning_rate": 2.4661971830985915e-06, "loss": 0.2029, "step": 20558 }, { "epoch": 23.05, "learning_rate": 2.4647887323943666e-06, "loss": 0.3532, "step": 20559 }, { "epoch": 23.05, "learning_rate": 2.4633802816901408e-06, "loss": 0.2452, "step": 20560 }, { "epoch": 23.05, "learning_rate": 2.461971830985916e-06, "loss": 0.2271, "step": 20561 }, { "epoch": 23.05, "learning_rate": 2.46056338028169e-06, "loss": 0.2145, "step": 20562 }, { "epoch": 23.05, "learning_rate": 2.459154929577465e-06, "loss": 0.1828, "step": 20563 }, { "epoch": 23.05, "learning_rate": 2.4577464788732393e-06, "loss": 0.2609, "step": 20564 }, { "epoch": 23.05, "learning_rate": 2.4563380281690144e-06, "loss": 0.1972, "step": 20565 }, { "epoch": 23.06, "learning_rate": 2.4549295774647886e-06, "loss": 0.2889, "step": 20566 }, { "epoch": 23.06, "learning_rate": 2.4535211267605637e-06, "loss": 0.4808, "step": 20567 }, { "epoch": 23.06, "learning_rate": 2.452112676056338e-06, "loss": 0.3397, "step": 20568 }, { "epoch": 23.06, "learning_rate": 2.450704225352113e-06, "loss": 0.3568, "step": 20569 }, { "epoch": 23.06, "learning_rate": 2.449295774647887e-06, "loss": 0.3195, "step": 20570 }, { "epoch": 23.06, "learning_rate": 2.4478873239436622e-06, "loss": 0.3634, "step": 20571 }, { "epoch": 23.06, "learning_rate": 2.4464788732394364e-06, "loss": 0.3592, "step": 20572 }, { "epoch": 23.06, "learning_rate": 2.4450704225352115e-06, "loss": 0.3428, "step": 20573 }, { "epoch": 23.07, "learning_rate": 2.4436619718309857e-06, "loss": 0.4365, "step": 20574 }, { "epoch": 23.07, "learning_rate": 2.4422535211267608e-06, "loss": 0.3338, "step": 20575 }, { "epoch": 23.07, "learning_rate": 2.4408450704225354e-06, "loss": 0.2318, "step": 20576 }, { "epoch": 23.07, "learning_rate": 2.43943661971831e-06, "loss": 0.3139, "step": 20577 }, { "epoch": 23.07, "learning_rate": 2.4380281690140847e-06, "loss": 0.2988, "step": 20578 }, { "epoch": 23.07, "learning_rate": 2.4366197183098593e-06, "loss": 0.3539, "step": 20579 }, { "epoch": 23.07, "learning_rate": 2.435211267605634e-06, "loss": 0.3024, "step": 20580 }, { "epoch": 23.07, "learning_rate": 2.4338028169014086e-06, "loss": 0.315, "step": 20581 }, { "epoch": 23.07, "learning_rate": 2.4323943661971832e-06, "loss": 0.2819, "step": 20582 }, { "epoch": 23.08, "learning_rate": 2.430985915492958e-06, "loss": 0.2815, "step": 20583 }, { "epoch": 23.08, "learning_rate": 2.4295774647887325e-06, "loss": 0.2648, "step": 20584 }, { "epoch": 23.08, "learning_rate": 2.428169014084507e-06, "loss": 0.29, "step": 20585 }, { "epoch": 23.08, "learning_rate": 2.4267605633802818e-06, "loss": 0.3322, "step": 20586 }, { "epoch": 23.08, "learning_rate": 2.4253521126760564e-06, "loss": 0.3687, "step": 20587 }, { "epoch": 23.08, "learning_rate": 2.423943661971831e-06, "loss": 0.3258, "step": 20588 }, { "epoch": 23.08, "learning_rate": 2.4225352112676057e-06, "loss": 0.3412, "step": 20589 }, { "epoch": 23.08, "learning_rate": 2.4211267605633803e-06, "loss": 0.2886, "step": 20590 }, { "epoch": 23.08, "learning_rate": 2.419718309859155e-06, "loss": 0.216, "step": 20591 }, { "epoch": 23.09, "learning_rate": 2.4183098591549296e-06, "loss": 0.2304, "step": 20592 }, { "epoch": 23.09, "learning_rate": 2.4169014084507047e-06, "loss": 0.2592, "step": 20593 }, { "epoch": 23.09, "learning_rate": 2.415492957746479e-06, "loss": 0.2528, "step": 20594 }, { "epoch": 23.09, "learning_rate": 2.414084507042254e-06, "loss": 0.2978, "step": 20595 }, { "epoch": 23.09, "learning_rate": 2.412676056338028e-06, "loss": 0.2873, "step": 20596 }, { "epoch": 23.09, "learning_rate": 2.4112676056338032e-06, "loss": 0.295, "step": 20597 }, { "epoch": 23.09, "learning_rate": 2.4098591549295774e-06, "loss": 0.3098, "step": 20598 }, { "epoch": 23.09, "learning_rate": 2.4084507042253525e-06, "loss": 0.2832, "step": 20599 }, { "epoch": 23.09, "learning_rate": 2.4070422535211267e-06, "loss": 0.2842, "step": 20600 }, { "epoch": 23.1, "learning_rate": 2.4056338028169018e-06, "loss": 0.3353, "step": 20601 }, { "epoch": 23.1, "learning_rate": 2.404225352112676e-06, "loss": 0.4527, "step": 20602 }, { "epoch": 23.1, "learning_rate": 2.402816901408451e-06, "loss": 0.2716, "step": 20603 }, { "epoch": 23.1, "learning_rate": 2.4014084507042253e-06, "loss": 0.2951, "step": 20604 }, { "epoch": 23.1, "learning_rate": 2.4000000000000003e-06, "loss": 0.2815, "step": 20605 }, { "epoch": 23.1, "learning_rate": 2.3985915492957745e-06, "loss": 0.3288, "step": 20606 }, { "epoch": 23.1, "learning_rate": 2.3971830985915496e-06, "loss": 0.2594, "step": 20607 }, { "epoch": 23.1, "learning_rate": 2.395774647887324e-06, "loss": 0.2481, "step": 20608 }, { "epoch": 23.1, "learning_rate": 2.394366197183099e-06, "loss": 0.264, "step": 20609 }, { "epoch": 23.11, "learning_rate": 2.3929577464788735e-06, "loss": 0.2508, "step": 20610 }, { "epoch": 23.11, "learning_rate": 2.3915492957746477e-06, "loss": 0.2159, "step": 20611 }, { "epoch": 23.11, "learning_rate": 2.3901408450704228e-06, "loss": 0.217, "step": 20612 }, { "epoch": 23.11, "learning_rate": 2.388732394366197e-06, "loss": 0.2392, "step": 20613 }, { "epoch": 23.11, "learning_rate": 2.387323943661972e-06, "loss": 0.2064, "step": 20614 }, { "epoch": 23.11, "learning_rate": 2.3859154929577463e-06, "loss": 0.3386, "step": 20615 }, { "epoch": 23.11, "learning_rate": 2.3845070422535213e-06, "loss": 0.3572, "step": 20616 }, { "epoch": 23.11, "learning_rate": 2.3830985915492955e-06, "loss": 0.4739, "step": 20617 }, { "epoch": 23.11, "learning_rate": 2.3816901408450706e-06, "loss": 0.3972, "step": 20618 }, { "epoch": 23.12, "learning_rate": 2.380281690140845e-06, "loss": 0.3659, "step": 20619 }, { "epoch": 23.12, "learning_rate": 2.37887323943662e-06, "loss": 0.294, "step": 20620 }, { "epoch": 23.12, "learning_rate": 2.377464788732394e-06, "loss": 0.3683, "step": 20621 }, { "epoch": 23.12, "learning_rate": 2.376056338028169e-06, "loss": 0.2664, "step": 20622 }, { "epoch": 23.12, "learning_rate": 2.3746478873239434e-06, "loss": 0.3285, "step": 20623 }, { "epoch": 23.12, "learning_rate": 2.3732394366197184e-06, "loss": 0.3005, "step": 20624 }, { "epoch": 23.12, "learning_rate": 2.3718309859154926e-06, "loss": 0.2885, "step": 20625 }, { "epoch": 23.12, "learning_rate": 2.3704225352112677e-06, "loss": 0.3701, "step": 20626 }, { "epoch": 23.12, "learning_rate": 2.3690140845070423e-06, "loss": 0.305, "step": 20627 }, { "epoch": 23.13, "learning_rate": 2.367605633802817e-06, "loss": 0.3139, "step": 20628 }, { "epoch": 23.13, "learning_rate": 2.3661971830985916e-06, "loss": 0.3013, "step": 20629 }, { "epoch": 23.13, "learning_rate": 2.3647887323943663e-06, "loss": 0.3083, "step": 20630 }, { "epoch": 23.13, "learning_rate": 2.363380281690141e-06, "loss": 0.3192, "step": 20631 }, { "epoch": 23.13, "learning_rate": 2.3619718309859155e-06, "loss": 0.3403, "step": 20632 }, { "epoch": 23.13, "learning_rate": 2.36056338028169e-06, "loss": 0.3688, "step": 20633 }, { "epoch": 23.13, "learning_rate": 2.359154929577465e-06, "loss": 0.2368, "step": 20634 }, { "epoch": 23.13, "learning_rate": 2.3577464788732394e-06, "loss": 0.3308, "step": 20635 }, { "epoch": 23.13, "learning_rate": 2.356338028169014e-06, "loss": 0.2553, "step": 20636 }, { "epoch": 23.14, "learning_rate": 2.3549295774647887e-06, "loss": 0.2286, "step": 20637 }, { "epoch": 23.14, "learning_rate": 2.3535211267605634e-06, "loss": 0.2403, "step": 20638 }, { "epoch": 23.14, "learning_rate": 2.352112676056338e-06, "loss": 0.2988, "step": 20639 }, { "epoch": 23.14, "learning_rate": 2.3507042253521126e-06, "loss": 0.2766, "step": 20640 }, { "epoch": 23.14, "learning_rate": 2.3492957746478873e-06, "loss": 0.2823, "step": 20641 }, { "epoch": 23.14, "learning_rate": 2.347887323943662e-06, "loss": 0.423, "step": 20642 }, { "epoch": 23.14, "learning_rate": 2.3464788732394365e-06, "loss": 0.4056, "step": 20643 }, { "epoch": 23.14, "learning_rate": 2.3450704225352116e-06, "loss": 0.2633, "step": 20644 }, { "epoch": 23.14, "learning_rate": 2.343661971830986e-06, "loss": 0.2391, "step": 20645 }, { "epoch": 23.15, "learning_rate": 2.342253521126761e-06, "loss": 0.2814, "step": 20646 }, { "epoch": 23.15, "learning_rate": 2.340845070422535e-06, "loss": 0.3207, "step": 20647 }, { "epoch": 23.15, "learning_rate": 2.33943661971831e-06, "loss": 0.3036, "step": 20648 }, { "epoch": 23.15, "learning_rate": 2.3380281690140844e-06, "loss": 0.2845, "step": 20649 }, { "epoch": 23.15, "learning_rate": 2.3366197183098594e-06, "loss": 0.297, "step": 20650 }, { "epoch": 23.15, "learning_rate": 2.3352112676056336e-06, "loss": 0.3371, "step": 20651 }, { "epoch": 23.15, "learning_rate": 2.3338028169014087e-06, "loss": 0.2718, "step": 20652 }, { "epoch": 23.15, "learning_rate": 2.332394366197183e-06, "loss": 0.3117, "step": 20653 }, { "epoch": 23.15, "learning_rate": 2.330985915492958e-06, "loss": 0.248, "step": 20654 }, { "epoch": 23.16, "learning_rate": 2.329577464788732e-06, "loss": 0.2197, "step": 20655 }, { "epoch": 23.16, "learning_rate": 2.3281690140845073e-06, "loss": 0.3185, "step": 20656 }, { "epoch": 23.16, "learning_rate": 2.3267605633802815e-06, "loss": 0.3181, "step": 20657 }, { "epoch": 23.16, "learning_rate": 2.3253521126760565e-06, "loss": 0.2214, "step": 20658 }, { "epoch": 23.16, "learning_rate": 2.3239436619718307e-06, "loss": 0.2778, "step": 20659 }, { "epoch": 23.16, "learning_rate": 2.322535211267606e-06, "loss": 0.2856, "step": 20660 }, { "epoch": 23.16, "learning_rate": 2.32112676056338e-06, "loss": 0.2557, "step": 20661 }, { "epoch": 23.16, "learning_rate": 2.319718309859155e-06, "loss": 0.2126, "step": 20662 }, { "epoch": 23.16, "learning_rate": 2.3183098591549297e-06, "loss": 0.2633, "step": 20663 }, { "epoch": 23.17, "learning_rate": 2.3169014084507044e-06, "loss": 0.2179, "step": 20664 }, { "epoch": 23.17, "learning_rate": 2.315492957746479e-06, "loss": 0.2734, "step": 20665 }, { "epoch": 23.17, "learning_rate": 2.3140845070422536e-06, "loss": 0.1586, "step": 20666 }, { "epoch": 23.17, "learning_rate": 2.3126760563380283e-06, "loss": 0.4464, "step": 20667 }, { "epoch": 23.17, "learning_rate": 2.311267605633803e-06, "loss": 0.4455, "step": 20668 }, { "epoch": 23.17, "learning_rate": 2.3098591549295775e-06, "loss": 0.4078, "step": 20669 }, { "epoch": 23.17, "learning_rate": 2.308450704225352e-06, "loss": 0.3634, "step": 20670 }, { "epoch": 23.17, "learning_rate": 2.307042253521127e-06, "loss": 0.3189, "step": 20671 }, { "epoch": 23.17, "learning_rate": 2.3056338028169015e-06, "loss": 0.3312, "step": 20672 }, { "epoch": 23.18, "learning_rate": 2.304225352112676e-06, "loss": 0.3127, "step": 20673 }, { "epoch": 23.18, "learning_rate": 2.3028169014084507e-06, "loss": 0.3416, "step": 20674 }, { "epoch": 23.18, "learning_rate": 2.3014084507042254e-06, "loss": 0.3028, "step": 20675 }, { "epoch": 23.18, "learning_rate": 2.3e-06, "loss": 0.3443, "step": 20676 }, { "epoch": 23.18, "learning_rate": 2.2985915492957746e-06, "loss": 0.2717, "step": 20677 }, { "epoch": 23.18, "learning_rate": 2.2971830985915493e-06, "loss": 0.402, "step": 20678 }, { "epoch": 23.18, "learning_rate": 2.295774647887324e-06, "loss": 0.3126, "step": 20679 }, { "epoch": 23.18, "learning_rate": 2.294366197183099e-06, "loss": 0.3025, "step": 20680 }, { "epoch": 23.18, "learning_rate": 2.292957746478873e-06, "loss": 0.3222, "step": 20681 }, { "epoch": 23.19, "learning_rate": 2.2915492957746483e-06, "loss": 0.2664, "step": 20682 }, { "epoch": 23.19, "learning_rate": 2.2901408450704225e-06, "loss": 0.3801, "step": 20683 }, { "epoch": 23.19, "learning_rate": 2.2887323943661975e-06, "loss": 0.2441, "step": 20684 }, { "epoch": 23.19, "learning_rate": 2.2873239436619717e-06, "loss": 0.3148, "step": 20685 }, { "epoch": 23.19, "learning_rate": 2.285915492957747e-06, "loss": 0.3986, "step": 20686 }, { "epoch": 23.19, "learning_rate": 2.284507042253521e-06, "loss": 0.2472, "step": 20687 }, { "epoch": 23.19, "learning_rate": 2.283098591549296e-06, "loss": 0.2356, "step": 20688 }, { "epoch": 23.19, "learning_rate": 2.2816901408450703e-06, "loss": 0.2728, "step": 20689 }, { "epoch": 23.2, "learning_rate": 2.2802816901408454e-06, "loss": 0.2858, "step": 20690 }, { "epoch": 23.2, "learning_rate": 2.2788732394366196e-06, "loss": 0.3614, "step": 20691 }, { "epoch": 23.2, "learning_rate": 2.2774647887323946e-06, "loss": 0.2716, "step": 20692 }, { "epoch": 23.2, "learning_rate": 2.276056338028169e-06, "loss": 0.2561, "step": 20693 }, { "epoch": 23.2, "learning_rate": 2.274647887323944e-06, "loss": 0.3188, "step": 20694 }, { "epoch": 23.2, "learning_rate": 2.273239436619718e-06, "loss": 0.2828, "step": 20695 }, { "epoch": 23.2, "learning_rate": 2.271830985915493e-06, "loss": 0.3458, "step": 20696 }, { "epoch": 23.2, "learning_rate": 2.270422535211268e-06, "loss": 0.3565, "step": 20697 }, { "epoch": 23.2, "learning_rate": 2.2690140845070425e-06, "loss": 0.2734, "step": 20698 }, { "epoch": 23.21, "learning_rate": 2.267605633802817e-06, "loss": 0.2673, "step": 20699 }, { "epoch": 23.21, "learning_rate": 2.2661971830985917e-06, "loss": 0.3193, "step": 20700 }, { "epoch": 23.21, "learning_rate": 2.2647887323943664e-06, "loss": 0.2316, "step": 20701 }, { "epoch": 23.21, "learning_rate": 2.263380281690141e-06, "loss": 0.2071, "step": 20702 }, { "epoch": 23.21, "learning_rate": 2.2619718309859156e-06, "loss": 0.2628, "step": 20703 }, { "epoch": 23.21, "learning_rate": 2.2605633802816903e-06, "loss": 0.2687, "step": 20704 }, { "epoch": 23.21, "learning_rate": 2.259154929577465e-06, "loss": 0.2782, "step": 20705 }, { "epoch": 23.21, "learning_rate": 2.2577464788732396e-06, "loss": 0.2517, "step": 20706 }, { "epoch": 23.21, "learning_rate": 2.256338028169014e-06, "loss": 0.2417, "step": 20707 }, { "epoch": 23.22, "learning_rate": 2.254929577464789e-06, "loss": 0.2947, "step": 20708 }, { "epoch": 23.22, "learning_rate": 2.2535211267605635e-06, "loss": 0.269, "step": 20709 }, { "epoch": 23.22, "learning_rate": 2.252112676056338e-06, "loss": 0.3743, "step": 20710 }, { "epoch": 23.22, "learning_rate": 2.2507042253521127e-06, "loss": 0.3431, "step": 20711 }, { "epoch": 23.22, "learning_rate": 2.2492957746478874e-06, "loss": 0.2919, "step": 20712 }, { "epoch": 23.22, "learning_rate": 2.247887323943662e-06, "loss": 0.2497, "step": 20713 }, { "epoch": 23.22, "learning_rate": 2.246478873239437e-06, "loss": 0.3204, "step": 20714 }, { "epoch": 23.22, "learning_rate": 2.2450704225352113e-06, "loss": 0.218, "step": 20715 }, { "epoch": 23.22, "learning_rate": 2.2436619718309863e-06, "loss": 0.3833, "step": 20716 }, { "epoch": 23.23, "learning_rate": 2.2422535211267606e-06, "loss": 0.449, "step": 20717 }, { "epoch": 23.23, "learning_rate": 2.2408450704225356e-06, "loss": 0.3577, "step": 20718 }, { "epoch": 23.23, "learning_rate": 2.23943661971831e-06, "loss": 0.3765, "step": 20719 }, { "epoch": 23.23, "learning_rate": 2.238028169014085e-06, "loss": 0.4023, "step": 20720 }, { "epoch": 23.23, "learning_rate": 2.236619718309859e-06, "loss": 0.2456, "step": 20721 }, { "epoch": 23.23, "learning_rate": 2.2352112676056338e-06, "loss": 0.2688, "step": 20722 }, { "epoch": 23.23, "learning_rate": 2.2338028169014084e-06, "loss": 0.3078, "step": 20723 }, { "epoch": 23.23, "learning_rate": 2.232394366197183e-06, "loss": 0.3419, "step": 20724 }, { "epoch": 23.23, "learning_rate": 2.2309859154929577e-06, "loss": 0.3073, "step": 20725 }, { "epoch": 23.24, "learning_rate": 2.2295774647887323e-06, "loss": 0.3114, "step": 20726 }, { "epoch": 23.24, "learning_rate": 2.228169014084507e-06, "loss": 0.2704, "step": 20727 }, { "epoch": 23.24, "learning_rate": 2.2267605633802816e-06, "loss": 0.279, "step": 20728 }, { "epoch": 23.24, "learning_rate": 2.2253521126760562e-06, "loss": 0.3309, "step": 20729 }, { "epoch": 23.24, "learning_rate": 2.223943661971831e-06, "loss": 0.3278, "step": 20730 }, { "epoch": 23.24, "learning_rate": 2.222535211267606e-06, "loss": 0.3555, "step": 20731 }, { "epoch": 23.24, "learning_rate": 2.22112676056338e-06, "loss": 0.2227, "step": 20732 }, { "epoch": 23.24, "learning_rate": 2.219718309859155e-06, "loss": 0.3042, "step": 20733 }, { "epoch": 23.24, "learning_rate": 2.2183098591549294e-06, "loss": 0.2496, "step": 20734 }, { "epoch": 23.25, "learning_rate": 2.2169014084507045e-06, "loss": 0.2941, "step": 20735 }, { "epoch": 23.25, "learning_rate": 2.2154929577464787e-06, "loss": 0.2489, "step": 20736 }, { "epoch": 23.25, "learning_rate": 2.2140845070422537e-06, "loss": 0.3909, "step": 20737 }, { "epoch": 23.25, "learning_rate": 2.212676056338028e-06, "loss": 0.3792, "step": 20738 }, { "epoch": 23.25, "learning_rate": 2.211267605633803e-06, "loss": 0.3099, "step": 20739 }, { "epoch": 23.25, "learning_rate": 2.2098591549295772e-06, "loss": 0.2526, "step": 20740 }, { "epoch": 23.25, "learning_rate": 2.2084507042253523e-06, "loss": 0.3281, "step": 20741 }, { "epoch": 23.25, "learning_rate": 2.2070422535211265e-06, "loss": 0.25, "step": 20742 }, { "epoch": 23.25, "learning_rate": 2.2056338028169016e-06, "loss": 0.3666, "step": 20743 }, { "epoch": 23.26, "learning_rate": 2.2042253521126758e-06, "loss": 0.2579, "step": 20744 }, { "epoch": 23.26, "learning_rate": 2.202816901408451e-06, "loss": 0.2683, "step": 20745 }, { "epoch": 23.26, "learning_rate": 2.201408450704225e-06, "loss": 0.3186, "step": 20746 }, { "epoch": 23.26, "learning_rate": 2.2e-06, "loss": 0.2585, "step": 20747 }, { "epoch": 23.26, "learning_rate": 2.1985915492957747e-06, "loss": 0.2363, "step": 20748 }, { "epoch": 23.26, "learning_rate": 2.1971830985915494e-06, "loss": 0.2684, "step": 20749 }, { "epoch": 23.26, "learning_rate": 2.195774647887324e-06, "loss": 0.283, "step": 20750 }, { "epoch": 23.26, "learning_rate": 2.1943661971830987e-06, "loss": 0.2761, "step": 20751 }, { "epoch": 23.26, "learning_rate": 2.1929577464788733e-06, "loss": 0.3492, "step": 20752 }, { "epoch": 23.27, "learning_rate": 2.191549295774648e-06, "loss": 0.3078, "step": 20753 }, { "epoch": 23.27, "learning_rate": 2.1901408450704226e-06, "loss": 0.3369, "step": 20754 }, { "epoch": 23.27, "learning_rate": 2.1887323943661972e-06, "loss": 0.2906, "step": 20755 }, { "epoch": 23.27, "learning_rate": 2.187323943661972e-06, "loss": 0.3607, "step": 20756 }, { "epoch": 23.27, "learning_rate": 2.1859154929577465e-06, "loss": 0.3364, "step": 20757 }, { "epoch": 23.27, "learning_rate": 2.184507042253521e-06, "loss": 0.2825, "step": 20758 }, { "epoch": 23.27, "learning_rate": 2.1830985915492958e-06, "loss": 0.271, "step": 20759 }, { "epoch": 23.27, "learning_rate": 2.1816901408450704e-06, "loss": 0.259, "step": 20760 }, { "epoch": 23.27, "learning_rate": 2.180281690140845e-06, "loss": 0.2435, "step": 20761 }, { "epoch": 23.28, "learning_rate": 2.1788732394366197e-06, "loss": 0.2357, "step": 20762 }, { "epoch": 23.28, "learning_rate": 2.1774647887323943e-06, "loss": 0.2217, "step": 20763 }, { "epoch": 23.28, "learning_rate": 2.176056338028169e-06, "loss": 0.3359, "step": 20764 }, { "epoch": 23.28, "learning_rate": 2.1746478873239436e-06, "loss": 0.2503, "step": 20765 }, { "epoch": 23.28, "learning_rate": 2.1732394366197182e-06, "loss": 0.2663, "step": 20766 }, { "epoch": 23.28, "learning_rate": 2.1718309859154933e-06, "loss": 0.3616, "step": 20767 }, { "epoch": 23.28, "learning_rate": 2.1704225352112675e-06, "loss": 0.3479, "step": 20768 }, { "epoch": 23.28, "learning_rate": 2.1690140845070426e-06, "loss": 0.3342, "step": 20769 }, { "epoch": 23.28, "learning_rate": 2.1676056338028168e-06, "loss": 0.36, "step": 20770 }, { "epoch": 23.29, "learning_rate": 2.166197183098592e-06, "loss": 0.5026, "step": 20771 }, { "epoch": 23.29, "learning_rate": 2.164788732394366e-06, "loss": 0.3325, "step": 20772 }, { "epoch": 23.29, "learning_rate": 2.163380281690141e-06, "loss": 0.3109, "step": 20773 }, { "epoch": 23.29, "learning_rate": 2.1619718309859153e-06, "loss": 0.3313, "step": 20774 }, { "epoch": 23.29, "learning_rate": 2.1605633802816904e-06, "loss": 0.3403, "step": 20775 }, { "epoch": 23.29, "learning_rate": 2.1591549295774646e-06, "loss": 0.36, "step": 20776 }, { "epoch": 23.29, "learning_rate": 2.1577464788732397e-06, "loss": 0.2139, "step": 20777 }, { "epoch": 23.29, "learning_rate": 2.156338028169014e-06, "loss": 0.3335, "step": 20778 }, { "epoch": 23.29, "learning_rate": 2.154929577464789e-06, "loss": 0.3218, "step": 20779 }, { "epoch": 23.3, "learning_rate": 2.153521126760563e-06, "loss": 0.2387, "step": 20780 }, { "epoch": 23.3, "learning_rate": 2.152112676056338e-06, "loss": 0.2661, "step": 20781 }, { "epoch": 23.3, "learning_rate": 2.1507042253521124e-06, "loss": 0.3292, "step": 20782 }, { "epoch": 23.3, "learning_rate": 2.1492957746478875e-06, "loss": 0.3376, "step": 20783 }, { "epoch": 23.3, "learning_rate": 2.147887323943662e-06, "loss": 0.3533, "step": 20784 }, { "epoch": 23.3, "learning_rate": 2.1464788732394368e-06, "loss": 0.329, "step": 20785 }, { "epoch": 23.3, "learning_rate": 2.1450704225352114e-06, "loss": 0.2625, "step": 20786 }, { "epoch": 23.3, "learning_rate": 2.143661971830986e-06, "loss": 0.2431, "step": 20787 }, { "epoch": 23.3, "learning_rate": 2.1422535211267607e-06, "loss": 0.3208, "step": 20788 }, { "epoch": 23.31, "learning_rate": 2.1408450704225353e-06, "loss": 0.3175, "step": 20789 }, { "epoch": 23.31, "learning_rate": 2.13943661971831e-06, "loss": 0.2411, "step": 20790 }, { "epoch": 23.31, "learning_rate": 2.1380281690140846e-06, "loss": 0.3474, "step": 20791 }, { "epoch": 23.31, "learning_rate": 2.1366197183098592e-06, "loss": 0.2377, "step": 20792 }, { "epoch": 23.31, "learning_rate": 2.135211267605634e-06, "loss": 0.2484, "step": 20793 }, { "epoch": 23.31, "learning_rate": 2.1338028169014085e-06, "loss": 0.3552, "step": 20794 }, { "epoch": 23.31, "learning_rate": 2.132394366197183e-06, "loss": 0.26, "step": 20795 }, { "epoch": 23.31, "learning_rate": 2.1309859154929578e-06, "loss": 0.3171, "step": 20796 }, { "epoch": 23.32, "learning_rate": 2.1295774647887324e-06, "loss": 0.2129, "step": 20797 }, { "epoch": 23.32, "learning_rate": 2.128169014084507e-06, "loss": 0.2557, "step": 20798 }, { "epoch": 23.32, "learning_rate": 2.1267605633802817e-06, "loss": 0.2269, "step": 20799 }, { "epoch": 23.32, "learning_rate": 2.1253521126760563e-06, "loss": 0.1928, "step": 20800 }, { "epoch": 23.32, "learning_rate": 2.1239436619718314e-06, "loss": 0.3041, "step": 20801 }, { "epoch": 23.32, "learning_rate": 2.1225352112676056e-06, "loss": 0.2817, "step": 20802 }, { "epoch": 23.32, "learning_rate": 2.1211267605633807e-06, "loss": 0.2382, "step": 20803 }, { "epoch": 23.32, "learning_rate": 2.119718309859155e-06, "loss": 0.3003, "step": 20804 }, { "epoch": 23.32, "learning_rate": 2.11830985915493e-06, "loss": 0.285, "step": 20805 }, { "epoch": 23.33, "learning_rate": 2.116901408450704e-06, "loss": 0.204, "step": 20806 }, { "epoch": 23.33, "learning_rate": 2.115492957746479e-06, "loss": 0.2629, "step": 20807 }, { "epoch": 23.33, "learning_rate": 2.1140845070422534e-06, "loss": 0.3021, "step": 20808 }, { "epoch": 23.33, "learning_rate": 2.1126760563380285e-06, "loss": 0.2548, "step": 20809 }, { "epoch": 23.33, "learning_rate": 2.1112676056338027e-06, "loss": 0.2896, "step": 20810 }, { "epoch": 23.33, "learning_rate": 2.1098591549295778e-06, "loss": 0.2265, "step": 20811 }, { "epoch": 23.33, "learning_rate": 2.108450704225352e-06, "loss": 0.2706, "step": 20812 }, { "epoch": 23.33, "learning_rate": 2.107042253521127e-06, "loss": 0.1853, "step": 20813 }, { "epoch": 23.33, "learning_rate": 2.1056338028169012e-06, "loss": 0.345, "step": 20814 }, { "epoch": 23.34, "learning_rate": 2.1042253521126763e-06, "loss": 0.3744, "step": 20815 }, { "epoch": 23.34, "learning_rate": 2.1028169014084505e-06, "loss": 0.3095, "step": 20816 }, { "epoch": 23.34, "learning_rate": 2.1014084507042256e-06, "loss": 0.3915, "step": 20817 }, { "epoch": 23.34, "learning_rate": 2.1000000000000002e-06, "loss": 0.4611, "step": 20818 }, { "epoch": 23.34, "learning_rate": 2.098591549295775e-06, "loss": 0.3919, "step": 20819 }, { "epoch": 23.34, "learning_rate": 2.0971830985915495e-06, "loss": 0.3995, "step": 20820 }, { "epoch": 23.34, "learning_rate": 2.095774647887324e-06, "loss": 0.466, "step": 20821 }, { "epoch": 23.34, "learning_rate": 2.0943661971830988e-06, "loss": 0.2789, "step": 20822 }, { "epoch": 23.34, "learning_rate": 2.0929577464788734e-06, "loss": 0.3888, "step": 20823 }, { "epoch": 23.35, "learning_rate": 2.091549295774648e-06, "loss": 0.3572, "step": 20824 }, { "epoch": 23.35, "learning_rate": 2.0901408450704227e-06, "loss": 0.245, "step": 20825 }, { "epoch": 23.35, "learning_rate": 2.0887323943661973e-06, "loss": 0.3316, "step": 20826 }, { "epoch": 23.35, "learning_rate": 2.087323943661972e-06, "loss": 0.2876, "step": 20827 }, { "epoch": 23.35, "learning_rate": 2.0859154929577466e-06, "loss": 0.392, "step": 20828 }, { "epoch": 23.35, "learning_rate": 2.0845070422535212e-06, "loss": 0.3001, "step": 20829 }, { "epoch": 23.35, "learning_rate": 2.083098591549296e-06, "loss": 0.4071, "step": 20830 }, { "epoch": 23.35, "learning_rate": 2.0816901408450705e-06, "loss": 0.3388, "step": 20831 }, { "epoch": 23.35, "learning_rate": 2.080281690140845e-06, "loss": 0.344, "step": 20832 }, { "epoch": 23.36, "learning_rate": 2.0788732394366198e-06, "loss": 0.392, "step": 20833 }, { "epoch": 23.36, "learning_rate": 2.0774647887323944e-06, "loss": 0.331, "step": 20834 }, { "epoch": 23.36, "learning_rate": 2.076056338028169e-06, "loss": 0.2865, "step": 20835 }, { "epoch": 23.36, "learning_rate": 2.0746478873239437e-06, "loss": 0.3968, "step": 20836 }, { "epoch": 23.36, "learning_rate": 2.0732394366197183e-06, "loss": 0.3692, "step": 20837 }, { "epoch": 23.36, "learning_rate": 2.071830985915493e-06, "loss": 0.2845, "step": 20838 }, { "epoch": 23.36, "learning_rate": 2.0704225352112676e-06, "loss": 0.3026, "step": 20839 }, { "epoch": 23.36, "learning_rate": 2.0690140845070422e-06, "loss": 0.3667, "step": 20840 }, { "epoch": 23.36, "learning_rate": 2.067605633802817e-06, "loss": 0.3107, "step": 20841 }, { "epoch": 23.37, "learning_rate": 2.0661971830985915e-06, "loss": 0.394, "step": 20842 }, { "epoch": 23.37, "learning_rate": 2.064788732394366e-06, "loss": 0.2511, "step": 20843 }, { "epoch": 23.37, "learning_rate": 2.063380281690141e-06, "loss": 0.2378, "step": 20844 }, { "epoch": 23.37, "learning_rate": 2.0619718309859154e-06, "loss": 0.2801, "step": 20845 }, { "epoch": 23.37, "learning_rate": 2.06056338028169e-06, "loss": 0.3202, "step": 20846 }, { "epoch": 23.37, "learning_rate": 2.0591549295774647e-06, "loss": 0.2943, "step": 20847 }, { "epoch": 23.37, "learning_rate": 2.0577464788732393e-06, "loss": 0.3502, "step": 20848 }, { "epoch": 23.37, "learning_rate": 2.056338028169014e-06, "loss": 0.3198, "step": 20849 }, { "epoch": 23.37, "learning_rate": 2.0549295774647886e-06, "loss": 0.3483, "step": 20850 }, { "epoch": 23.38, "learning_rate": 2.0535211267605633e-06, "loss": 0.2238, "step": 20851 }, { "epoch": 23.38, "learning_rate": 2.0521126760563383e-06, "loss": 0.3526, "step": 20852 }, { "epoch": 23.38, "learning_rate": 2.0507042253521125e-06, "loss": 0.3072, "step": 20853 }, { "epoch": 23.38, "learning_rate": 2.0492957746478876e-06, "loss": 0.2934, "step": 20854 }, { "epoch": 23.38, "learning_rate": 2.047887323943662e-06, "loss": 0.2696, "step": 20855 }, { "epoch": 23.38, "learning_rate": 2.046478873239437e-06, "loss": 0.2391, "step": 20856 }, { "epoch": 23.38, "learning_rate": 2.045070422535211e-06, "loss": 0.2615, "step": 20857 }, { "epoch": 23.38, "learning_rate": 2.043661971830986e-06, "loss": 0.2622, "step": 20858 }, { "epoch": 23.38, "learning_rate": 2.0422535211267604e-06, "loss": 0.173, "step": 20859 }, { "epoch": 23.39, "learning_rate": 2.0408450704225354e-06, "loss": 0.2494, "step": 20860 }, { "epoch": 23.39, "learning_rate": 2.0394366197183096e-06, "loss": 0.2176, "step": 20861 }, { "epoch": 23.39, "learning_rate": 2.0380281690140847e-06, "loss": 0.2001, "step": 20862 }, { "epoch": 23.39, "learning_rate": 2.036619718309859e-06, "loss": 0.2149, "step": 20863 }, { "epoch": 23.39, "learning_rate": 2.035211267605634e-06, "loss": 0.2399, "step": 20864 }, { "epoch": 23.39, "learning_rate": 2.033802816901408e-06, "loss": 0.1856, "step": 20865 }, { "epoch": 23.39, "learning_rate": 2.0323943661971832e-06, "loss": 0.2301, "step": 20866 }, { "epoch": 23.39, "learning_rate": 2.0309859154929575e-06, "loss": 0.3808, "step": 20867 }, { "epoch": 23.39, "learning_rate": 2.0295774647887325e-06, "loss": 0.3649, "step": 20868 }, { "epoch": 23.4, "learning_rate": 2.0281690140845067e-06, "loss": 0.4089, "step": 20869 }, { "epoch": 23.4, "learning_rate": 2.026760563380282e-06, "loss": 0.328, "step": 20870 }, { "epoch": 23.4, "learning_rate": 2.0253521126760564e-06, "loss": 0.348, "step": 20871 }, { "epoch": 23.4, "learning_rate": 2.023943661971831e-06, "loss": 0.3682, "step": 20872 }, { "epoch": 23.4, "learning_rate": 2.0225352112676057e-06, "loss": 0.3687, "step": 20873 }, { "epoch": 23.4, "learning_rate": 2.0211267605633803e-06, "loss": 0.3111, "step": 20874 }, { "epoch": 23.4, "learning_rate": 2.019718309859155e-06, "loss": 0.3368, "step": 20875 }, { "epoch": 23.4, "learning_rate": 2.0183098591549296e-06, "loss": 0.3299, "step": 20876 }, { "epoch": 23.4, "learning_rate": 2.0169014084507043e-06, "loss": 0.2777, "step": 20877 }, { "epoch": 23.41, "learning_rate": 2.015492957746479e-06, "loss": 0.368, "step": 20878 }, { "epoch": 23.41, "learning_rate": 2.0140845070422535e-06, "loss": 0.3016, "step": 20879 }, { "epoch": 23.41, "learning_rate": 2.012676056338028e-06, "loss": 0.2099, "step": 20880 }, { "epoch": 23.41, "learning_rate": 2.011267605633803e-06, "loss": 0.3424, "step": 20881 }, { "epoch": 23.41, "learning_rate": 2.0098591549295774e-06, "loss": 0.3106, "step": 20882 }, { "epoch": 23.41, "learning_rate": 2.008450704225352e-06, "loss": 0.3273, "step": 20883 }, { "epoch": 23.41, "learning_rate": 2.0070422535211267e-06, "loss": 0.3175, "step": 20884 }, { "epoch": 23.41, "learning_rate": 2.0056338028169014e-06, "loss": 0.2569, "step": 20885 }, { "epoch": 23.41, "learning_rate": 2.004225352112676e-06, "loss": 0.3173, "step": 20886 }, { "epoch": 23.42, "learning_rate": 2.0028169014084506e-06, "loss": 0.3087, "step": 20887 }, { "epoch": 23.42, "learning_rate": 2.0014084507042257e-06, "loss": 0.2687, "step": 20888 }, { "epoch": 23.42, "learning_rate": 2e-06, "loss": 0.3139, "step": 20889 }, { "epoch": 23.42, "learning_rate": 1.998591549295775e-06, "loss": 0.2494, "step": 20890 }, { "epoch": 23.42, "learning_rate": 1.997183098591549e-06, "loss": 0.354, "step": 20891 }, { "epoch": 23.42, "learning_rate": 1.9957746478873242e-06, "loss": 0.3307, "step": 20892 }, { "epoch": 23.42, "learning_rate": 1.9943661971830985e-06, "loss": 0.2877, "step": 20893 }, { "epoch": 23.42, "learning_rate": 1.9929577464788735e-06, "loss": 0.2248, "step": 20894 }, { "epoch": 23.42, "learning_rate": 1.9915492957746477e-06, "loss": 0.3036, "step": 20895 }, { "epoch": 23.43, "learning_rate": 1.990140845070423e-06, "loss": 0.3236, "step": 20896 }, { "epoch": 23.43, "learning_rate": 1.988732394366197e-06, "loss": 0.3108, "step": 20897 }, { "epoch": 23.43, "learning_rate": 1.987323943661972e-06, "loss": 0.26, "step": 20898 }, { "epoch": 23.43, "learning_rate": 1.9859154929577463e-06, "loss": 0.3528, "step": 20899 }, { "epoch": 23.43, "learning_rate": 1.9845070422535213e-06, "loss": 0.255, "step": 20900 }, { "epoch": 23.43, "learning_rate": 1.9830985915492956e-06, "loss": 0.3044, "step": 20901 }, { "epoch": 23.43, "learning_rate": 1.9816901408450706e-06, "loss": 0.3433, "step": 20902 }, { "epoch": 23.43, "learning_rate": 1.980281690140845e-06, "loss": 0.2995, "step": 20903 }, { "epoch": 23.43, "learning_rate": 1.97887323943662e-06, "loss": 0.2271, "step": 20904 }, { "epoch": 23.44, "learning_rate": 1.9774647887323945e-06, "loss": 0.2401, "step": 20905 }, { "epoch": 23.44, "learning_rate": 1.976056338028169e-06, "loss": 0.2842, "step": 20906 }, { "epoch": 23.44, "learning_rate": 1.974647887323944e-06, "loss": 0.3197, "step": 20907 }, { "epoch": 23.44, "learning_rate": 1.9732394366197184e-06, "loss": 0.2696, "step": 20908 }, { "epoch": 23.44, "learning_rate": 1.971830985915493e-06, "loss": 0.3222, "step": 20909 }, { "epoch": 23.44, "learning_rate": 1.9704225352112677e-06, "loss": 0.2475, "step": 20910 }, { "epoch": 23.44, "learning_rate": 1.9690140845070424e-06, "loss": 0.2416, "step": 20911 }, { "epoch": 23.44, "learning_rate": 1.967605633802817e-06, "loss": 0.148, "step": 20912 }, { "epoch": 23.45, "learning_rate": 1.9661971830985916e-06, "loss": 0.2065, "step": 20913 }, { "epoch": 23.45, "learning_rate": 1.9647887323943663e-06, "loss": 0.246, "step": 20914 }, { "epoch": 23.45, "learning_rate": 1.963380281690141e-06, "loss": 0.2305, "step": 20915 }, { "epoch": 23.45, "learning_rate": 1.9619718309859155e-06, "loss": 0.4363, "step": 20916 }, { "epoch": 23.45, "learning_rate": 1.96056338028169e-06, "loss": 0.4018, "step": 20917 }, { "epoch": 23.45, "learning_rate": 1.959154929577465e-06, "loss": 0.3247, "step": 20918 }, { "epoch": 23.45, "learning_rate": 1.9577464788732395e-06, "loss": 0.3337, "step": 20919 }, { "epoch": 23.45, "learning_rate": 1.956338028169014e-06, "loss": 0.4049, "step": 20920 }, { "epoch": 23.45, "learning_rate": 1.9549295774647887e-06, "loss": 0.3242, "step": 20921 }, { "epoch": 23.46, "learning_rate": 1.953521126760564e-06, "loss": 0.3336, "step": 20922 }, { "epoch": 23.46, "learning_rate": 1.952112676056338e-06, "loss": 0.4027, "step": 20923 }, { "epoch": 23.46, "learning_rate": 1.950704225352113e-06, "loss": 0.3039, "step": 20924 }, { "epoch": 23.46, "learning_rate": 1.9492957746478873e-06, "loss": 0.3449, "step": 20925 }, { "epoch": 23.46, "learning_rate": 1.9478873239436623e-06, "loss": 0.3015, "step": 20926 }, { "epoch": 23.46, "learning_rate": 1.9464788732394366e-06, "loss": 0.3468, "step": 20927 }, { "epoch": 23.46, "learning_rate": 1.9450704225352116e-06, "loss": 0.3853, "step": 20928 }, { "epoch": 23.46, "learning_rate": 1.943661971830986e-06, "loss": 0.3033, "step": 20929 }, { "epoch": 23.46, "learning_rate": 1.942253521126761e-06, "loss": 0.325, "step": 20930 }, { "epoch": 23.47, "learning_rate": 1.940845070422535e-06, "loss": 0.3349, "step": 20931 }, { "epoch": 23.47, "learning_rate": 1.93943661971831e-06, "loss": 0.3428, "step": 20932 }, { "epoch": 23.47, "learning_rate": 1.9380281690140844e-06, "loss": 0.3287, "step": 20933 }, { "epoch": 23.47, "learning_rate": 1.9366197183098594e-06, "loss": 0.3056, "step": 20934 }, { "epoch": 23.47, "learning_rate": 1.9352112676056337e-06, "loss": 0.2675, "step": 20935 }, { "epoch": 23.47, "learning_rate": 1.9338028169014087e-06, "loss": 0.2654, "step": 20936 }, { "epoch": 23.47, "learning_rate": 1.932394366197183e-06, "loss": 0.3201, "step": 20937 }, { "epoch": 23.47, "learning_rate": 1.930985915492958e-06, "loss": 0.3185, "step": 20938 }, { "epoch": 23.47, "learning_rate": 1.9295774647887326e-06, "loss": 0.3221, "step": 20939 }, { "epoch": 23.48, "learning_rate": 1.9281690140845073e-06, "loss": 0.2789, "step": 20940 }, { "epoch": 23.48, "learning_rate": 1.926760563380282e-06, "loss": 0.2898, "step": 20941 }, { "epoch": 23.48, "learning_rate": 1.9253521126760565e-06, "loss": 0.2562, "step": 20942 }, { "epoch": 23.48, "learning_rate": 1.923943661971831e-06, "loss": 0.3465, "step": 20943 }, { "epoch": 23.48, "learning_rate": 1.922535211267606e-06, "loss": 0.2401, "step": 20944 }, { "epoch": 23.48, "learning_rate": 1.9211267605633805e-06, "loss": 0.3181, "step": 20945 }, { "epoch": 23.48, "learning_rate": 1.919718309859155e-06, "loss": 0.2318, "step": 20946 }, { "epoch": 23.48, "learning_rate": 1.9183098591549297e-06, "loss": 0.3611, "step": 20947 }, { "epoch": 23.48, "learning_rate": 1.9169014084507044e-06, "loss": 0.2294, "step": 20948 }, { "epoch": 23.49, "learning_rate": 1.915492957746479e-06, "loss": 0.2969, "step": 20949 }, { "epoch": 23.49, "learning_rate": 1.9140845070422536e-06, "loss": 0.2573, "step": 20950 }, { "epoch": 23.49, "learning_rate": 1.9126760563380283e-06, "loss": 0.4861, "step": 20951 }, { "epoch": 23.49, "learning_rate": 1.9112676056338025e-06, "loss": 0.349, "step": 20952 }, { "epoch": 23.49, "learning_rate": 1.9098591549295776e-06, "loss": 0.2554, "step": 20953 }, { "epoch": 23.49, "learning_rate": 1.9084507042253518e-06, "loss": 0.2441, "step": 20954 }, { "epoch": 23.49, "learning_rate": 1.9070422535211268e-06, "loss": 0.2855, "step": 20955 }, { "epoch": 23.49, "learning_rate": 1.9056338028169013e-06, "loss": 0.334, "step": 20956 }, { "epoch": 23.49, "learning_rate": 1.9042253521126761e-06, "loss": 0.2664, "step": 20957 }, { "epoch": 23.5, "learning_rate": 1.902816901408451e-06, "loss": 0.2268, "step": 20958 }, { "epoch": 23.5, "learning_rate": 1.9014084507042254e-06, "loss": 0.3501, "step": 20959 }, { "epoch": 23.5, "learning_rate": 1.9000000000000002e-06, "loss": 0.2778, "step": 20960 }, { "epoch": 23.5, "learning_rate": 1.8985915492957747e-06, "loss": 0.1931, "step": 20961 }, { "epoch": 23.5, "learning_rate": 1.8971830985915495e-06, "loss": 0.2782, "step": 20962 }, { "epoch": 23.5, "learning_rate": 1.895774647887324e-06, "loss": 0.3137, "step": 20963 }, { "epoch": 23.5, "learning_rate": 1.8943661971830988e-06, "loss": 0.3076, "step": 20964 }, { "epoch": 23.5, "learning_rate": 1.8929577464788732e-06, "loss": 0.2703, "step": 20965 }, { "epoch": 23.5, "learning_rate": 1.891549295774648e-06, "loss": 0.171, "step": 20966 }, { "epoch": 23.51, "learning_rate": 1.8901408450704225e-06, "loss": 0.4033, "step": 20967 }, { "epoch": 23.51, "learning_rate": 1.8887323943661973e-06, "loss": 0.4675, "step": 20968 }, { "epoch": 23.51, "learning_rate": 1.8873239436619718e-06, "loss": 0.3291, "step": 20969 }, { "epoch": 23.51, "learning_rate": 1.8859154929577466e-06, "loss": 0.3229, "step": 20970 }, { "epoch": 23.51, "learning_rate": 1.884507042253521e-06, "loss": 0.3148, "step": 20971 }, { "epoch": 23.51, "learning_rate": 1.8830985915492959e-06, "loss": 0.2712, "step": 20972 }, { "epoch": 23.51, "learning_rate": 1.8816901408450703e-06, "loss": 0.3385, "step": 20973 }, { "epoch": 23.51, "learning_rate": 1.8802816901408452e-06, "loss": 0.374, "step": 20974 }, { "epoch": 23.51, "learning_rate": 1.87887323943662e-06, "loss": 0.3768, "step": 20975 }, { "epoch": 23.52, "learning_rate": 1.8774647887323944e-06, "loss": 0.4163, "step": 20976 }, { "epoch": 23.52, "learning_rate": 1.8760563380281693e-06, "loss": 0.3491, "step": 20977 }, { "epoch": 23.52, "learning_rate": 1.8746478873239437e-06, "loss": 0.3122, "step": 20978 }, { "epoch": 23.52, "learning_rate": 1.8732394366197183e-06, "loss": 0.338, "step": 20979 }, { "epoch": 23.52, "learning_rate": 1.871830985915493e-06, "loss": 0.2499, "step": 20980 }, { "epoch": 23.52, "learning_rate": 1.8704225352112676e-06, "loss": 0.2856, "step": 20981 }, { "epoch": 23.52, "learning_rate": 1.8690140845070423e-06, "loss": 0.2684, "step": 20982 }, { "epoch": 23.52, "learning_rate": 1.8676056338028171e-06, "loss": 0.3083, "step": 20983 }, { "epoch": 23.52, "learning_rate": 1.8661971830985917e-06, "loss": 0.2745, "step": 20984 }, { "epoch": 23.53, "learning_rate": 1.8647887323943664e-06, "loss": 0.2882, "step": 20985 }, { "epoch": 23.53, "learning_rate": 1.863380281690141e-06, "loss": 0.34, "step": 20986 }, { "epoch": 23.53, "learning_rate": 1.8619718309859157e-06, "loss": 0.2754, "step": 20987 }, { "epoch": 23.53, "learning_rate": 1.8605633802816903e-06, "loss": 0.3047, "step": 20988 }, { "epoch": 23.53, "learning_rate": 1.859154929577465e-06, "loss": 0.2692, "step": 20989 }, { "epoch": 23.53, "learning_rate": 1.8577464788732396e-06, "loss": 0.3254, "step": 20990 }, { "epoch": 23.53, "learning_rate": 1.8563380281690142e-06, "loss": 0.3227, "step": 20991 }, { "epoch": 23.53, "learning_rate": 1.8549295774647888e-06, "loss": 0.2504, "step": 20992 }, { "epoch": 23.53, "learning_rate": 1.8535211267605635e-06, "loss": 0.2776, "step": 20993 }, { "epoch": 23.54, "learning_rate": 1.8521126760563381e-06, "loss": 0.24, "step": 20994 }, { "epoch": 23.54, "learning_rate": 1.8507042253521128e-06, "loss": 0.2385, "step": 20995 }, { "epoch": 23.54, "learning_rate": 1.8492957746478874e-06, "loss": 0.294, "step": 20996 }, { "epoch": 23.54, "learning_rate": 1.847887323943662e-06, "loss": 0.2889, "step": 20997 }, { "epoch": 23.54, "learning_rate": 1.8464788732394367e-06, "loss": 0.3311, "step": 20998 }, { "epoch": 23.54, "learning_rate": 1.8450704225352113e-06, "loss": 0.2875, "step": 20999 }, { "epoch": 23.54, "learning_rate": 1.8436619718309862e-06, "loss": 0.2698, "step": 21000 }, { "epoch": 23.54, "eval_loss": 0.717689573764801, "eval_runtime": 818.6465, "eval_samples_per_second": 3.227, "eval_steps_per_second": 0.404, "eval_wer": 0.12834192780642603, "step": 21000 }, { "epoch": 23.54, "learning_rate": 1.8422535211267608e-06, "loss": 0.312, "step": 21001 }, { "epoch": 23.54, "learning_rate": 1.8408450704225354e-06, "loss": 0.2618, "step": 21002 }, { "epoch": 23.55, "learning_rate": 1.83943661971831e-06, "loss": 0.3064, "step": 21003 }, { "epoch": 23.55, "learning_rate": 1.8380281690140847e-06, "loss": 0.2976, "step": 21004 }, { "epoch": 23.55, "learning_rate": 1.8366197183098593e-06, "loss": 0.2979, "step": 21005 }, { "epoch": 23.55, "learning_rate": 1.8352112676056338e-06, "loss": 0.3038, "step": 21006 }, { "epoch": 23.55, "learning_rate": 1.8338028169014084e-06, "loss": 0.3508, "step": 21007 }, { "epoch": 23.55, "learning_rate": 1.832394366197183e-06, "loss": 0.2714, "step": 21008 }, { "epoch": 23.55, "learning_rate": 1.8309859154929577e-06, "loss": 0.3518, "step": 21009 }, { "epoch": 23.55, "learning_rate": 1.8295774647887323e-06, "loss": 0.2493, "step": 21010 }, { "epoch": 23.55, "learning_rate": 1.828169014084507e-06, "loss": 0.3211, "step": 21011 }, { "epoch": 23.56, "learning_rate": 1.8267605633802816e-06, "loss": 0.2354, "step": 21012 }, { "epoch": 23.56, "learning_rate": 1.8253521126760562e-06, "loss": 0.2574, "step": 21013 }, { "epoch": 23.56, "learning_rate": 1.8239436619718309e-06, "loss": 0.2285, "step": 21014 }, { "epoch": 23.56, "learning_rate": 1.8225352112676055e-06, "loss": 0.2064, "step": 21015 }, { "epoch": 23.56, "learning_rate": 1.8211267605633801e-06, "loss": 0.2946, "step": 21016 }, { "epoch": 23.56, "learning_rate": 1.819718309859155e-06, "loss": 0.4741, "step": 21017 }, { "epoch": 23.56, "learning_rate": 1.8183098591549296e-06, "loss": 0.3924, "step": 21018 }, { "epoch": 23.56, "learning_rate": 1.8169014084507043e-06, "loss": 0.4667, "step": 21019 }, { "epoch": 23.57, "learning_rate": 1.815492957746479e-06, "loss": 0.3266, "step": 21020 }, { "epoch": 23.57, "learning_rate": 1.8140845070422535e-06, "loss": 0.3464, "step": 21021 }, { "epoch": 23.57, "learning_rate": 1.8126760563380282e-06, "loss": 0.3753, "step": 21022 }, { "epoch": 23.57, "learning_rate": 1.8112676056338028e-06, "loss": 0.3655, "step": 21023 }, { "epoch": 23.57, "learning_rate": 1.8098591549295775e-06, "loss": 0.3685, "step": 21024 }, { "epoch": 23.57, "learning_rate": 1.808450704225352e-06, "loss": 0.3005, "step": 21025 }, { "epoch": 23.57, "learning_rate": 1.8070422535211267e-06, "loss": 0.4145, "step": 21026 }, { "epoch": 23.57, "learning_rate": 1.8056338028169014e-06, "loss": 0.3096, "step": 21027 }, { "epoch": 23.57, "learning_rate": 1.804225352112676e-06, "loss": 0.2908, "step": 21028 }, { "epoch": 23.58, "learning_rate": 1.8028169014084506e-06, "loss": 0.3125, "step": 21029 }, { "epoch": 23.58, "learning_rate": 1.8014084507042253e-06, "loss": 0.2763, "step": 21030 }, { "epoch": 23.58, "learning_rate": 1.8e-06, "loss": 0.3183, "step": 21031 }, { "epoch": 23.58, "learning_rate": 1.7985915492957746e-06, "loss": 0.3699, "step": 21032 }, { "epoch": 23.58, "learning_rate": 1.7971830985915492e-06, "loss": 0.3414, "step": 21033 }, { "epoch": 23.58, "learning_rate": 1.7957746478873238e-06, "loss": 0.2989, "step": 21034 }, { "epoch": 23.58, "learning_rate": 1.7943661971830987e-06, "loss": 0.4951, "step": 21035 }, { "epoch": 23.58, "learning_rate": 1.7929577464788733e-06, "loss": 0.3206, "step": 21036 }, { "epoch": 23.58, "learning_rate": 1.791549295774648e-06, "loss": 0.3126, "step": 21037 }, { "epoch": 23.59, "learning_rate": 1.7901408450704226e-06, "loss": 0.2516, "step": 21038 }, { "epoch": 23.59, "learning_rate": 1.7887323943661972e-06, "loss": 0.3227, "step": 21039 }, { "epoch": 23.59, "learning_rate": 1.7873239436619719e-06, "loss": 0.3122, "step": 21040 }, { "epoch": 23.59, "learning_rate": 1.7859154929577465e-06, "loss": 0.2761, "step": 21041 }, { "epoch": 23.59, "learning_rate": 1.7845070422535211e-06, "loss": 0.24, "step": 21042 }, { "epoch": 23.59, "learning_rate": 1.7830985915492958e-06, "loss": 0.2893, "step": 21043 }, { "epoch": 23.59, "learning_rate": 1.7816901408450704e-06, "loss": 0.24, "step": 21044 }, { "epoch": 23.59, "learning_rate": 1.780281690140845e-06, "loss": 0.2827, "step": 21045 }, { "epoch": 23.59, "learning_rate": 1.7788732394366197e-06, "loss": 0.2917, "step": 21046 }, { "epoch": 23.6, "learning_rate": 1.7774647887323943e-06, "loss": 0.3936, "step": 21047 }, { "epoch": 23.6, "learning_rate": 1.776056338028169e-06, "loss": 0.38, "step": 21048 }, { "epoch": 23.6, "learning_rate": 1.7746478873239436e-06, "loss": 0.2807, "step": 21049 }, { "epoch": 23.6, "learning_rate": 1.7732394366197182e-06, "loss": 0.2271, "step": 21050 }, { "epoch": 23.6, "learning_rate": 1.7718309859154929e-06, "loss": 0.3081, "step": 21051 }, { "epoch": 23.6, "learning_rate": 1.7704225352112677e-06, "loss": 0.2654, "step": 21052 }, { "epoch": 23.6, "learning_rate": 1.7690140845070424e-06, "loss": 0.2903, "step": 21053 }, { "epoch": 23.6, "learning_rate": 1.767605633802817e-06, "loss": 0.2817, "step": 21054 }, { "epoch": 23.6, "learning_rate": 1.7661971830985916e-06, "loss": 0.3025, "step": 21055 }, { "epoch": 23.61, "learning_rate": 1.7647887323943663e-06, "loss": 0.273, "step": 21056 }, { "epoch": 23.61, "learning_rate": 1.763380281690141e-06, "loss": 0.2585, "step": 21057 }, { "epoch": 23.61, "learning_rate": 1.7619718309859156e-06, "loss": 0.3279, "step": 21058 }, { "epoch": 23.61, "learning_rate": 1.7605633802816902e-06, "loss": 0.3718, "step": 21059 }, { "epoch": 23.61, "learning_rate": 1.7591549295774648e-06, "loss": 0.2898, "step": 21060 }, { "epoch": 23.61, "learning_rate": 1.7577464788732395e-06, "loss": 0.2747, "step": 21061 }, { "epoch": 23.61, "learning_rate": 1.756338028169014e-06, "loss": 0.3086, "step": 21062 }, { "epoch": 23.61, "learning_rate": 1.7549295774647887e-06, "loss": 0.2914, "step": 21063 }, { "epoch": 23.61, "learning_rate": 1.7535211267605634e-06, "loss": 0.1778, "step": 21064 }, { "epoch": 23.62, "learning_rate": 1.752112676056338e-06, "loss": 0.2722, "step": 21065 }, { "epoch": 23.62, "learning_rate": 1.7507042253521127e-06, "loss": 0.1589, "step": 21066 }, { "epoch": 23.62, "learning_rate": 1.7492957746478873e-06, "loss": 0.3642, "step": 21067 }, { "epoch": 23.62, "learning_rate": 1.747887323943662e-06, "loss": 0.4265, "step": 21068 }, { "epoch": 23.62, "learning_rate": 1.7464788732394368e-06, "loss": 0.3959, "step": 21069 }, { "epoch": 23.62, "learning_rate": 1.7450704225352114e-06, "loss": 0.3887, "step": 21070 }, { "epoch": 23.62, "learning_rate": 1.743661971830986e-06, "loss": 0.295, "step": 21071 }, { "epoch": 23.62, "learning_rate": 1.7422535211267607e-06, "loss": 0.3908, "step": 21072 }, { "epoch": 23.62, "learning_rate": 1.7408450704225353e-06, "loss": 0.2654, "step": 21073 }, { "epoch": 23.63, "learning_rate": 1.73943661971831e-06, "loss": 0.4004, "step": 21074 }, { "epoch": 23.63, "learning_rate": 1.7380281690140846e-06, "loss": 0.304, "step": 21075 }, { "epoch": 23.63, "learning_rate": 1.7366197183098592e-06, "loss": 0.3453, "step": 21076 }, { "epoch": 23.63, "learning_rate": 1.7352112676056339e-06, "loss": 0.3586, "step": 21077 }, { "epoch": 23.63, "learning_rate": 1.7338028169014085e-06, "loss": 0.4145, "step": 21078 }, { "epoch": 23.63, "learning_rate": 1.7323943661971832e-06, "loss": 0.3836, "step": 21079 }, { "epoch": 23.63, "learning_rate": 1.7309859154929578e-06, "loss": 0.2582, "step": 21080 }, { "epoch": 23.63, "learning_rate": 1.7295774647887324e-06, "loss": 0.2179, "step": 21081 }, { "epoch": 23.63, "learning_rate": 1.728169014084507e-06, "loss": 0.304, "step": 21082 }, { "epoch": 23.64, "learning_rate": 1.7267605633802817e-06, "loss": 0.2573, "step": 21083 }, { "epoch": 23.64, "learning_rate": 1.7253521126760563e-06, "loss": 0.3122, "step": 21084 }, { "epoch": 23.64, "learning_rate": 1.723943661971831e-06, "loss": 0.2404, "step": 21085 }, { "epoch": 23.64, "learning_rate": 1.7225352112676056e-06, "loss": 0.2489, "step": 21086 }, { "epoch": 23.64, "learning_rate": 1.7211267605633805e-06, "loss": 0.376, "step": 21087 }, { "epoch": 23.64, "learning_rate": 1.719718309859155e-06, "loss": 0.2383, "step": 21088 }, { "epoch": 23.64, "learning_rate": 1.7183098591549297e-06, "loss": 0.2202, "step": 21089 }, { "epoch": 23.64, "learning_rate": 1.7169014084507044e-06, "loss": 0.2941, "step": 21090 }, { "epoch": 23.64, "learning_rate": 1.715492957746479e-06, "loss": 0.3301, "step": 21091 }, { "epoch": 23.65, "learning_rate": 1.7140845070422537e-06, "loss": 0.2985, "step": 21092 }, { "epoch": 23.65, "learning_rate": 1.7126760563380283e-06, "loss": 0.2313, "step": 21093 }, { "epoch": 23.65, "learning_rate": 1.711267605633803e-06, "loss": 0.2803, "step": 21094 }, { "epoch": 23.65, "learning_rate": 1.7098591549295776e-06, "loss": 0.3705, "step": 21095 }, { "epoch": 23.65, "learning_rate": 1.7084507042253522e-06, "loss": 0.2726, "step": 21096 }, { "epoch": 23.65, "learning_rate": 1.7070422535211268e-06, "loss": 0.3663, "step": 21097 }, { "epoch": 23.65, "learning_rate": 1.7056338028169015e-06, "loss": 0.268, "step": 21098 }, { "epoch": 23.65, "learning_rate": 1.7042253521126761e-06, "loss": 0.2709, "step": 21099 }, { "epoch": 23.65, "learning_rate": 1.7028169014084508e-06, "loss": 0.2959, "step": 21100 }, { "epoch": 23.66, "learning_rate": 1.7014084507042254e-06, "loss": 0.3334, "step": 21101 }, { "epoch": 23.66, "learning_rate": 1.7e-06, "loss": 0.3399, "step": 21102 }, { "epoch": 23.66, "learning_rate": 1.6985915492957747e-06, "loss": 0.3235, "step": 21103 }, { "epoch": 23.66, "learning_rate": 1.6971830985915495e-06, "loss": 0.2733, "step": 21104 }, { "epoch": 23.66, "learning_rate": 1.6957746478873242e-06, "loss": 0.2438, "step": 21105 }, { "epoch": 23.66, "learning_rate": 1.6943661971830988e-06, "loss": 0.2836, "step": 21106 }, { "epoch": 23.66, "learning_rate": 1.6929577464788734e-06, "loss": 0.3203, "step": 21107 }, { "epoch": 23.66, "learning_rate": 1.691549295774648e-06, "loss": 0.31, "step": 21108 }, { "epoch": 23.66, "learning_rate": 1.6901408450704227e-06, "loss": 0.2601, "step": 21109 }, { "epoch": 23.67, "learning_rate": 1.6887323943661973e-06, "loss": 0.1854, "step": 21110 }, { "epoch": 23.67, "learning_rate": 1.687323943661972e-06, "loss": 0.3562, "step": 21111 }, { "epoch": 23.67, "learning_rate": 1.6859154929577466e-06, "loss": 0.2823, "step": 21112 }, { "epoch": 23.67, "learning_rate": 1.6845070422535213e-06, "loss": 0.1813, "step": 21113 }, { "epoch": 23.67, "learning_rate": 1.6830985915492959e-06, "loss": 0.2555, "step": 21114 }, { "epoch": 23.67, "learning_rate": 1.6816901408450705e-06, "loss": 0.2533, "step": 21115 }, { "epoch": 23.67, "learning_rate": 1.6802816901408452e-06, "loss": 0.2645, "step": 21116 }, { "epoch": 23.67, "learning_rate": 1.6788732394366198e-06, "loss": 0.5813, "step": 21117 }, { "epoch": 23.67, "learning_rate": 1.6774647887323944e-06, "loss": 0.3486, "step": 21118 }, { "epoch": 23.68, "learning_rate": 1.676056338028169e-06, "loss": 0.2973, "step": 21119 }, { "epoch": 23.68, "learning_rate": 1.6746478873239437e-06, "loss": 0.3225, "step": 21120 }, { "epoch": 23.68, "learning_rate": 1.6732394366197184e-06, "loss": 0.3016, "step": 21121 }, { "epoch": 23.68, "learning_rate": 1.671830985915493e-06, "loss": 0.4478, "step": 21122 }, { "epoch": 23.68, "learning_rate": 1.6704225352112676e-06, "loss": 0.3192, "step": 21123 }, { "epoch": 23.68, "learning_rate": 1.6690140845070423e-06, "loss": 0.36, "step": 21124 }, { "epoch": 23.68, "learning_rate": 1.667605633802817e-06, "loss": 0.3643, "step": 21125 }, { "epoch": 23.68, "learning_rate": 1.6661971830985915e-06, "loss": 0.2852, "step": 21126 }, { "epoch": 23.68, "learning_rate": 1.6647887323943662e-06, "loss": 0.3307, "step": 21127 }, { "epoch": 23.69, "learning_rate": 1.6633802816901408e-06, "loss": 0.3045, "step": 21128 }, { "epoch": 23.69, "learning_rate": 1.6619718309859155e-06, "loss": 0.328, "step": 21129 }, { "epoch": 23.69, "learning_rate": 1.66056338028169e-06, "loss": 0.3263, "step": 21130 }, { "epoch": 23.69, "learning_rate": 1.6591549295774647e-06, "loss": 0.3202, "step": 21131 }, { "epoch": 23.69, "learning_rate": 1.6577464788732394e-06, "loss": 0.2646, "step": 21132 }, { "epoch": 23.69, "learning_rate": 1.656338028169014e-06, "loss": 0.2607, "step": 21133 }, { "epoch": 23.69, "learning_rate": 1.6549295774647886e-06, "loss": 0.2518, "step": 21134 }, { "epoch": 23.69, "learning_rate": 1.6535211267605633e-06, "loss": 0.3289, "step": 21135 }, { "epoch": 23.7, "learning_rate": 1.652112676056338e-06, "loss": 0.2964, "step": 21136 }, { "epoch": 23.7, "learning_rate": 1.6507042253521126e-06, "loss": 0.3394, "step": 21137 }, { "epoch": 23.7, "learning_rate": 1.6492957746478872e-06, "loss": 0.313, "step": 21138 }, { "epoch": 23.7, "learning_rate": 1.647887323943662e-06, "loss": 0.3006, "step": 21139 }, { "epoch": 23.7, "learning_rate": 1.6464788732394367e-06, "loss": 0.3631, "step": 21140 }, { "epoch": 23.7, "learning_rate": 1.6450704225352113e-06, "loss": 0.2988, "step": 21141 }, { "epoch": 23.7, "learning_rate": 1.643661971830986e-06, "loss": 0.2625, "step": 21142 }, { "epoch": 23.7, "learning_rate": 1.6422535211267606e-06, "loss": 0.3535, "step": 21143 }, { "epoch": 23.7, "learning_rate": 1.6408450704225352e-06, "loss": 0.2295, "step": 21144 }, { "epoch": 23.71, "learning_rate": 1.6394366197183099e-06, "loss": 0.3212, "step": 21145 }, { "epoch": 23.71, "learning_rate": 1.6380281690140845e-06, "loss": 0.2879, "step": 21146 }, { "epoch": 23.71, "learning_rate": 1.6366197183098591e-06, "loss": 0.2834, "step": 21147 }, { "epoch": 23.71, "learning_rate": 1.6352112676056338e-06, "loss": 0.2529, "step": 21148 }, { "epoch": 23.71, "learning_rate": 1.6338028169014084e-06, "loss": 0.304, "step": 21149 }, { "epoch": 23.71, "learning_rate": 1.632394366197183e-06, "loss": 0.2994, "step": 21150 }, { "epoch": 23.71, "learning_rate": 1.6309859154929577e-06, "loss": 0.3436, "step": 21151 }, { "epoch": 23.71, "learning_rate": 1.6295774647887323e-06, "loss": 0.2694, "step": 21152 }, { "epoch": 23.71, "learning_rate": 1.628169014084507e-06, "loss": 0.3236, "step": 21153 }, { "epoch": 23.72, "learning_rate": 1.6267605633802816e-06, "loss": 0.2394, "step": 21154 }, { "epoch": 23.72, "learning_rate": 1.6253521126760562e-06, "loss": 0.2687, "step": 21155 }, { "epoch": 23.72, "learning_rate": 1.623943661971831e-06, "loss": 0.3136, "step": 21156 }, { "epoch": 23.72, "learning_rate": 1.6225352112676057e-06, "loss": 0.3547, "step": 21157 }, { "epoch": 23.72, "learning_rate": 1.6211267605633804e-06, "loss": 0.3786, "step": 21158 }, { "epoch": 23.72, "learning_rate": 1.619718309859155e-06, "loss": 0.2729, "step": 21159 }, { "epoch": 23.72, "learning_rate": 1.6183098591549296e-06, "loss": 0.1584, "step": 21160 }, { "epoch": 23.72, "learning_rate": 1.6169014084507043e-06, "loss": 0.3486, "step": 21161 }, { "epoch": 23.72, "learning_rate": 1.615492957746479e-06, "loss": 0.2446, "step": 21162 }, { "epoch": 23.73, "learning_rate": 1.6140845070422536e-06, "loss": 0.2559, "step": 21163 }, { "epoch": 23.73, "learning_rate": 1.6126760563380282e-06, "loss": 0.1913, "step": 21164 }, { "epoch": 23.73, "learning_rate": 1.6112676056338028e-06, "loss": 0.3028, "step": 21165 }, { "epoch": 23.73, "learning_rate": 1.6098591549295775e-06, "loss": 0.2705, "step": 21166 }, { "epoch": 23.73, "learning_rate": 1.608450704225352e-06, "loss": 0.3632, "step": 21167 }, { "epoch": 23.73, "learning_rate": 1.6070422535211267e-06, "loss": 0.286, "step": 21168 }, { "epoch": 23.73, "learning_rate": 1.6056338028169014e-06, "loss": 0.2748, "step": 21169 }, { "epoch": 23.73, "learning_rate": 1.604225352112676e-06, "loss": 0.3446, "step": 21170 }, { "epoch": 23.73, "learning_rate": 1.6028169014084507e-06, "loss": 0.3291, "step": 21171 }, { "epoch": 23.74, "learning_rate": 1.6014084507042253e-06, "loss": 0.329, "step": 21172 }, { "epoch": 23.74, "learning_rate": 1.6000000000000001e-06, "loss": 0.2547, "step": 21173 }, { "epoch": 23.74, "learning_rate": 1.5985915492957748e-06, "loss": 0.3881, "step": 21174 }, { "epoch": 23.74, "learning_rate": 1.5971830985915494e-06, "loss": 0.2557, "step": 21175 }, { "epoch": 23.74, "learning_rate": 1.595774647887324e-06, "loss": 0.3458, "step": 21176 }, { "epoch": 23.74, "learning_rate": 1.5943661971830987e-06, "loss": 0.3398, "step": 21177 }, { "epoch": 23.74, "learning_rate": 1.5929577464788733e-06, "loss": 0.2615, "step": 21178 }, { "epoch": 23.74, "learning_rate": 1.591549295774648e-06, "loss": 0.3298, "step": 21179 }, { "epoch": 23.74, "learning_rate": 1.5901408450704226e-06, "loss": 0.3469, "step": 21180 }, { "epoch": 23.75, "learning_rate": 1.5887323943661972e-06, "loss": 0.3786, "step": 21181 }, { "epoch": 23.75, "learning_rate": 1.5873239436619719e-06, "loss": 0.2718, "step": 21182 }, { "epoch": 23.75, "learning_rate": 1.5859154929577465e-06, "loss": 0.2479, "step": 21183 }, { "epoch": 23.75, "learning_rate": 1.5845070422535212e-06, "loss": 0.3548, "step": 21184 }, { "epoch": 23.75, "learning_rate": 1.5830985915492958e-06, "loss": 0.2465, "step": 21185 }, { "epoch": 23.75, "learning_rate": 1.5816901408450704e-06, "loss": 0.303, "step": 21186 }, { "epoch": 23.75, "learning_rate": 1.580281690140845e-06, "loss": 0.227, "step": 21187 }, { "epoch": 23.75, "learning_rate": 1.5788732394366197e-06, "loss": 0.2894, "step": 21188 }, { "epoch": 23.75, "learning_rate": 1.5774647887323943e-06, "loss": 0.2238, "step": 21189 }, { "epoch": 23.76, "learning_rate": 1.576056338028169e-06, "loss": 0.2181, "step": 21190 }, { "epoch": 23.76, "learning_rate": 1.5746478873239438e-06, "loss": 0.3155, "step": 21191 }, { "epoch": 23.76, "learning_rate": 1.5732394366197185e-06, "loss": 0.277, "step": 21192 }, { "epoch": 23.76, "learning_rate": 1.571830985915493e-06, "loss": 0.3323, "step": 21193 }, { "epoch": 23.76, "learning_rate": 1.5704225352112677e-06, "loss": 0.2785, "step": 21194 }, { "epoch": 23.76, "learning_rate": 1.5690140845070424e-06, "loss": 0.3376, "step": 21195 }, { "epoch": 23.76, "learning_rate": 1.567605633802817e-06, "loss": 0.2685, "step": 21196 }, { "epoch": 23.76, "learning_rate": 1.5661971830985917e-06, "loss": 0.2952, "step": 21197 }, { "epoch": 23.76, "learning_rate": 1.5647887323943663e-06, "loss": 0.3398, "step": 21198 }, { "epoch": 23.77, "learning_rate": 1.563380281690141e-06, "loss": 0.3075, "step": 21199 }, { "epoch": 23.77, "learning_rate": 1.5619718309859156e-06, "loss": 0.2697, "step": 21200 }, { "epoch": 23.77, "learning_rate": 1.5605633802816902e-06, "loss": 0.3305, "step": 21201 }, { "epoch": 23.77, "learning_rate": 1.5591549295774648e-06, "loss": 0.2235, "step": 21202 }, { "epoch": 23.77, "learning_rate": 1.5577464788732395e-06, "loss": 0.2934, "step": 21203 }, { "epoch": 23.77, "learning_rate": 1.5563380281690141e-06, "loss": 0.234, "step": 21204 }, { "epoch": 23.77, "learning_rate": 1.5549295774647888e-06, "loss": 0.286, "step": 21205 }, { "epoch": 23.77, "learning_rate": 1.5535211267605634e-06, "loss": 0.3586, "step": 21206 }, { "epoch": 23.77, "learning_rate": 1.552112676056338e-06, "loss": 0.3004, "step": 21207 }, { "epoch": 23.78, "learning_rate": 1.5507042253521129e-06, "loss": 0.2591, "step": 21208 }, { "epoch": 23.78, "learning_rate": 1.5492957746478875e-06, "loss": 0.3336, "step": 21209 }, { "epoch": 23.78, "learning_rate": 1.5478873239436622e-06, "loss": 0.2062, "step": 21210 }, { "epoch": 23.78, "learning_rate": 1.5464788732394368e-06, "loss": 0.2332, "step": 21211 }, { "epoch": 23.78, "learning_rate": 1.5450704225352114e-06, "loss": 0.2328, "step": 21212 }, { "epoch": 23.78, "learning_rate": 1.543661971830986e-06, "loss": 0.1932, "step": 21213 }, { "epoch": 23.78, "learning_rate": 1.5422535211267607e-06, "loss": 0.3165, "step": 21214 }, { "epoch": 23.78, "learning_rate": 1.5408450704225353e-06, "loss": 0.1575, "step": 21215 }, { "epoch": 23.78, "learning_rate": 1.53943661971831e-06, "loss": 0.2716, "step": 21216 }, { "epoch": 23.79, "learning_rate": 1.5380281690140846e-06, "loss": 0.4907, "step": 21217 }, { "epoch": 23.79, "learning_rate": 1.5366197183098593e-06, "loss": 0.3178, "step": 21218 }, { "epoch": 23.79, "learning_rate": 1.5352112676056339e-06, "loss": 0.3403, "step": 21219 }, { "epoch": 23.79, "learning_rate": 1.5338028169014085e-06, "loss": 0.3479, "step": 21220 }, { "epoch": 23.79, "learning_rate": 1.5323943661971832e-06, "loss": 0.3574, "step": 21221 }, { "epoch": 23.79, "learning_rate": 1.5309859154929578e-06, "loss": 0.3113, "step": 21222 }, { "epoch": 23.79, "learning_rate": 1.5295774647887324e-06, "loss": 0.3171, "step": 21223 }, { "epoch": 23.79, "learning_rate": 1.528169014084507e-06, "loss": 0.3467, "step": 21224 }, { "epoch": 23.79, "learning_rate": 1.526760563380282e-06, "loss": 0.3734, "step": 21225 }, { "epoch": 23.8, "learning_rate": 1.5253521126760566e-06, "loss": 0.3349, "step": 21226 }, { "epoch": 23.8, "learning_rate": 1.5239436619718312e-06, "loss": 0.3033, "step": 21227 }, { "epoch": 23.8, "learning_rate": 1.5225352112676058e-06, "loss": 0.3265, "step": 21228 }, { "epoch": 23.8, "learning_rate": 1.5211267605633805e-06, "loss": 0.3066, "step": 21229 }, { "epoch": 23.8, "learning_rate": 1.5197183098591551e-06, "loss": 0.3834, "step": 21230 }, { "epoch": 23.8, "learning_rate": 1.5183098591549298e-06, "loss": 0.301, "step": 21231 }, { "epoch": 23.8, "learning_rate": 1.5169014084507044e-06, "loss": 0.3123, "step": 21232 }, { "epoch": 23.8, "learning_rate": 1.5154929577464788e-06, "loss": 0.3047, "step": 21233 }, { "epoch": 23.8, "learning_rate": 1.5140845070422535e-06, "loss": 0.3061, "step": 21234 }, { "epoch": 23.81, "learning_rate": 1.512676056338028e-06, "loss": 0.3783, "step": 21235 }, { "epoch": 23.81, "learning_rate": 1.5112676056338027e-06, "loss": 0.3718, "step": 21236 }, { "epoch": 23.81, "learning_rate": 1.5098591549295774e-06, "loss": 0.3151, "step": 21237 }, { "epoch": 23.81, "learning_rate": 1.508450704225352e-06, "loss": 0.355, "step": 21238 }, { "epoch": 23.81, "learning_rate": 1.5070422535211266e-06, "loss": 0.3392, "step": 21239 }, { "epoch": 23.81, "learning_rate": 1.5056338028169013e-06, "loss": 0.2989, "step": 21240 }, { "epoch": 23.81, "learning_rate": 1.504225352112676e-06, "loss": 0.2497, "step": 21241 }, { "epoch": 23.81, "learning_rate": 1.5028169014084506e-06, "loss": 0.2678, "step": 21242 }, { "epoch": 23.82, "learning_rate": 1.5014084507042254e-06, "loss": 0.2737, "step": 21243 }, { "epoch": 23.82, "learning_rate": 1.5e-06, "loss": 0.2973, "step": 21244 }, { "epoch": 23.82, "learning_rate": 1.4985915492957747e-06, "loss": 0.2988, "step": 21245 }, { "epoch": 23.82, "learning_rate": 1.4971830985915493e-06, "loss": 0.2461, "step": 21246 }, { "epoch": 23.82, "learning_rate": 1.495774647887324e-06, "loss": 0.3005, "step": 21247 }, { "epoch": 23.82, "learning_rate": 1.4943661971830986e-06, "loss": 0.284, "step": 21248 }, { "epoch": 23.82, "learning_rate": 1.4929577464788732e-06, "loss": 0.2919, "step": 21249 }, { "epoch": 23.82, "learning_rate": 1.4915492957746479e-06, "loss": 0.3395, "step": 21250 }, { "epoch": 23.82, "learning_rate": 1.4901408450704225e-06, "loss": 0.2681, "step": 21251 }, { "epoch": 23.83, "learning_rate": 1.4887323943661971e-06, "loss": 0.2557, "step": 21252 }, { "epoch": 23.83, "learning_rate": 1.4873239436619718e-06, "loss": 0.2425, "step": 21253 }, { "epoch": 23.83, "learning_rate": 1.4859154929577464e-06, "loss": 0.2841, "step": 21254 }, { "epoch": 23.83, "learning_rate": 1.484507042253521e-06, "loss": 0.2737, "step": 21255 }, { "epoch": 23.83, "learning_rate": 1.4830985915492957e-06, "loss": 0.3592, "step": 21256 }, { "epoch": 23.83, "learning_rate": 1.4816901408450703e-06, "loss": 0.3038, "step": 21257 }, { "epoch": 23.83, "learning_rate": 1.480281690140845e-06, "loss": 0.2389, "step": 21258 }, { "epoch": 23.83, "learning_rate": 1.4788732394366196e-06, "loss": 0.2859, "step": 21259 }, { "epoch": 23.83, "learning_rate": 1.4774647887323944e-06, "loss": 0.3374, "step": 21260 }, { "epoch": 23.84, "learning_rate": 1.476056338028169e-06, "loss": 0.2273, "step": 21261 }, { "epoch": 23.84, "learning_rate": 1.4746478873239437e-06, "loss": 0.2023, "step": 21262 }, { "epoch": 23.84, "learning_rate": 1.4732394366197184e-06, "loss": 0.3033, "step": 21263 }, { "epoch": 23.84, "learning_rate": 1.471830985915493e-06, "loss": 0.1956, "step": 21264 }, { "epoch": 23.84, "learning_rate": 1.4704225352112676e-06, "loss": 0.2185, "step": 21265 }, { "epoch": 23.84, "learning_rate": 1.4690140845070423e-06, "loss": 0.315, "step": 21266 }, { "epoch": 23.84, "learning_rate": 1.467605633802817e-06, "loss": 0.4293, "step": 21267 }, { "epoch": 23.84, "learning_rate": 1.4661971830985916e-06, "loss": 0.3446, "step": 21268 }, { "epoch": 23.84, "learning_rate": 1.4647887323943662e-06, "loss": 0.2924, "step": 21269 }, { "epoch": 23.85, "learning_rate": 1.4633802816901408e-06, "loss": 0.3443, "step": 21270 }, { "epoch": 23.85, "learning_rate": 1.4619718309859155e-06, "loss": 0.2951, "step": 21271 }, { "epoch": 23.85, "learning_rate": 1.46056338028169e-06, "loss": 0.2731, "step": 21272 }, { "epoch": 23.85, "learning_rate": 1.4591549295774647e-06, "loss": 0.2715, "step": 21273 }, { "epoch": 23.85, "learning_rate": 1.4577464788732394e-06, "loss": 0.3702, "step": 21274 }, { "epoch": 23.85, "learning_rate": 1.456338028169014e-06, "loss": 0.3076, "step": 21275 }, { "epoch": 23.85, "learning_rate": 1.4549295774647887e-06, "loss": 0.2285, "step": 21276 }, { "epoch": 23.85, "learning_rate": 1.4535211267605635e-06, "loss": 0.2549, "step": 21277 }, { "epoch": 23.85, "learning_rate": 1.4521126760563381e-06, "loss": 0.3298, "step": 21278 }, { "epoch": 23.86, "learning_rate": 1.4507042253521128e-06, "loss": 0.2542, "step": 21279 }, { "epoch": 23.86, "learning_rate": 1.4492957746478874e-06, "loss": 0.3964, "step": 21280 }, { "epoch": 23.86, "learning_rate": 1.447887323943662e-06, "loss": 0.2499, "step": 21281 }, { "epoch": 23.86, "learning_rate": 1.4464788732394367e-06, "loss": 0.3672, "step": 21282 }, { "epoch": 23.86, "learning_rate": 1.4450704225352113e-06, "loss": 0.3628, "step": 21283 }, { "epoch": 23.86, "learning_rate": 1.443661971830986e-06, "loss": 0.3129, "step": 21284 }, { "epoch": 23.86, "learning_rate": 1.4422535211267606e-06, "loss": 0.292, "step": 21285 }, { "epoch": 23.86, "learning_rate": 1.4408450704225352e-06, "loss": 0.2752, "step": 21286 }, { "epoch": 23.86, "learning_rate": 1.4394366197183099e-06, "loss": 0.3085, "step": 21287 }, { "epoch": 23.87, "learning_rate": 1.4380281690140845e-06, "loss": 0.3838, "step": 21288 }, { "epoch": 23.87, "learning_rate": 1.4366197183098591e-06, "loss": 0.2703, "step": 21289 }, { "epoch": 23.87, "learning_rate": 1.4352112676056338e-06, "loss": 0.3791, "step": 21290 }, { "epoch": 23.87, "learning_rate": 1.4338028169014084e-06, "loss": 0.4298, "step": 21291 }, { "epoch": 23.87, "learning_rate": 1.432394366197183e-06, "loss": 0.2293, "step": 21292 }, { "epoch": 23.87, "learning_rate": 1.4309859154929577e-06, "loss": 0.2902, "step": 21293 }, { "epoch": 23.87, "learning_rate": 1.4295774647887323e-06, "loss": 0.3466, "step": 21294 }, { "epoch": 23.87, "learning_rate": 1.4281690140845072e-06, "loss": 0.2743, "step": 21295 }, { "epoch": 23.87, "learning_rate": 1.4267605633802818e-06, "loss": 0.2774, "step": 21296 }, { "epoch": 23.88, "learning_rate": 1.4253521126760565e-06, "loss": 0.3904, "step": 21297 }, { "epoch": 23.88, "learning_rate": 1.423943661971831e-06, "loss": 0.2794, "step": 21298 }, { "epoch": 23.88, "learning_rate": 1.4225352112676057e-06, "loss": 0.2523, "step": 21299 }, { "epoch": 23.88, "learning_rate": 1.4211267605633804e-06, "loss": 0.2542, "step": 21300 }, { "epoch": 23.88, "learning_rate": 1.419718309859155e-06, "loss": 0.2413, "step": 21301 }, { "epoch": 23.88, "learning_rate": 1.4183098591549296e-06, "loss": 0.3711, "step": 21302 }, { "epoch": 23.88, "learning_rate": 1.4169014084507043e-06, "loss": 0.3292, "step": 21303 }, { "epoch": 23.88, "learning_rate": 1.415492957746479e-06, "loss": 0.2239, "step": 21304 }, { "epoch": 23.88, "learning_rate": 1.4140845070422536e-06, "loss": 0.2756, "step": 21305 }, { "epoch": 23.89, "learning_rate": 1.4126760563380282e-06, "loss": 0.2372, "step": 21306 }, { "epoch": 23.89, "learning_rate": 1.4112676056338028e-06, "loss": 0.2868, "step": 21307 }, { "epoch": 23.89, "learning_rate": 1.4098591549295775e-06, "loss": 0.2958, "step": 21308 }, { "epoch": 23.89, "learning_rate": 1.4084507042253521e-06, "loss": 0.3348, "step": 21309 }, { "epoch": 23.89, "learning_rate": 1.4070422535211267e-06, "loss": 0.2933, "step": 21310 }, { "epoch": 23.89, "learning_rate": 1.4056338028169014e-06, "loss": 0.3912, "step": 21311 }, { "epoch": 23.89, "learning_rate": 1.4042253521126762e-06, "loss": 0.2548, "step": 21312 }, { "epoch": 23.89, "learning_rate": 1.4028169014084509e-06, "loss": 0.1874, "step": 21313 }, { "epoch": 23.89, "learning_rate": 1.4014084507042255e-06, "loss": 0.2597, "step": 21314 }, { "epoch": 23.9, "learning_rate": 1.4000000000000001e-06, "loss": 0.2942, "step": 21315 }, { "epoch": 23.9, "learning_rate": 1.3985915492957748e-06, "loss": 0.2754, "step": 21316 }, { "epoch": 23.9, "learning_rate": 1.3971830985915494e-06, "loss": 0.3855, "step": 21317 }, { "epoch": 23.9, "learning_rate": 1.395774647887324e-06, "loss": 0.3309, "step": 21318 }, { "epoch": 23.9, "learning_rate": 1.3943661971830987e-06, "loss": 0.3347, "step": 21319 }, { "epoch": 23.9, "learning_rate": 1.3929577464788733e-06, "loss": 0.3853, "step": 21320 }, { "epoch": 23.9, "learning_rate": 1.391549295774648e-06, "loss": 0.2677, "step": 21321 }, { "epoch": 23.9, "learning_rate": 1.3901408450704226e-06, "loss": 0.3209, "step": 21322 }, { "epoch": 23.9, "learning_rate": 1.3887323943661972e-06, "loss": 0.3338, "step": 21323 }, { "epoch": 23.91, "learning_rate": 1.3873239436619719e-06, "loss": 0.3271, "step": 21324 }, { "epoch": 23.91, "learning_rate": 1.3859154929577465e-06, "loss": 0.3596, "step": 21325 }, { "epoch": 23.91, "learning_rate": 1.3845070422535212e-06, "loss": 0.2977, "step": 21326 }, { "epoch": 23.91, "learning_rate": 1.3830985915492958e-06, "loss": 0.2613, "step": 21327 }, { "epoch": 23.91, "learning_rate": 1.3816901408450704e-06, "loss": 0.2863, "step": 21328 }, { "epoch": 23.91, "learning_rate": 1.3802816901408453e-06, "loss": 0.2559, "step": 21329 }, { "epoch": 23.91, "learning_rate": 1.37887323943662e-06, "loss": 0.3425, "step": 21330 }, { "epoch": 23.91, "learning_rate": 1.3774647887323946e-06, "loss": 0.3043, "step": 21331 }, { "epoch": 23.91, "learning_rate": 1.3760563380281692e-06, "loss": 0.3132, "step": 21332 }, { "epoch": 23.92, "learning_rate": 1.3746478873239438e-06, "loss": 0.3163, "step": 21333 }, { "epoch": 23.92, "learning_rate": 1.3732394366197185e-06, "loss": 0.3202, "step": 21334 }, { "epoch": 23.92, "learning_rate": 1.3718309859154931e-06, "loss": 0.3554, "step": 21335 }, { "epoch": 23.92, "learning_rate": 1.3704225352112677e-06, "loss": 0.3244, "step": 21336 }, { "epoch": 23.92, "learning_rate": 1.3690140845070424e-06, "loss": 0.3245, "step": 21337 }, { "epoch": 23.92, "learning_rate": 1.367605633802817e-06, "loss": 0.2918, "step": 21338 }, { "epoch": 23.92, "learning_rate": 1.3661971830985917e-06, "loss": 0.4473, "step": 21339 }, { "epoch": 23.92, "learning_rate": 1.3647887323943663e-06, "loss": 0.3034, "step": 21340 }, { "epoch": 23.92, "learning_rate": 1.363380281690141e-06, "loss": 0.3291, "step": 21341 }, { "epoch": 23.93, "learning_rate": 1.3619718309859156e-06, "loss": 0.2971, "step": 21342 }, { "epoch": 23.93, "learning_rate": 1.3605633802816902e-06, "loss": 0.3064, "step": 21343 }, { "epoch": 23.93, "learning_rate": 1.3591549295774648e-06, "loss": 0.3236, "step": 21344 }, { "epoch": 23.93, "learning_rate": 1.3577464788732395e-06, "loss": 0.268, "step": 21345 }, { "epoch": 23.93, "learning_rate": 1.3563380281690141e-06, "loss": 0.3347, "step": 21346 }, { "epoch": 23.93, "learning_rate": 1.3549295774647888e-06, "loss": 0.2911, "step": 21347 }, { "epoch": 23.93, "learning_rate": 1.3535211267605634e-06, "loss": 0.283, "step": 21348 }, { "epoch": 23.93, "learning_rate": 1.352112676056338e-06, "loss": 0.382, "step": 21349 }, { "epoch": 23.93, "learning_rate": 1.3507042253521127e-06, "loss": 0.2713, "step": 21350 }, { "epoch": 23.94, "learning_rate": 1.3492957746478873e-06, "loss": 0.3091, "step": 21351 }, { "epoch": 23.94, "learning_rate": 1.347887323943662e-06, "loss": 0.309, "step": 21352 }, { "epoch": 23.94, "learning_rate": 1.3464788732394366e-06, "loss": 0.3215, "step": 21353 }, { "epoch": 23.94, "learning_rate": 1.3450704225352112e-06, "loss": 0.3279, "step": 21354 }, { "epoch": 23.94, "learning_rate": 1.3436619718309859e-06, "loss": 0.3108, "step": 21355 }, { "epoch": 23.94, "learning_rate": 1.3422535211267605e-06, "loss": 0.2799, "step": 21356 }, { "epoch": 23.94, "learning_rate": 1.3408450704225351e-06, "loss": 0.2861, "step": 21357 }, { "epoch": 23.94, "learning_rate": 1.3394366197183098e-06, "loss": 0.3327, "step": 21358 }, { "epoch": 23.95, "learning_rate": 1.3380281690140844e-06, "loss": 0.2659, "step": 21359 }, { "epoch": 23.95, "learning_rate": 1.336619718309859e-06, "loss": 0.2868, "step": 21360 }, { "epoch": 23.95, "learning_rate": 1.3352112676056337e-06, "loss": 0.3454, "step": 21361 }, { "epoch": 23.95, "learning_rate": 1.3338028169014083e-06, "loss": 0.3163, "step": 21362 }, { "epoch": 23.95, "learning_rate": 1.332394366197183e-06, "loss": 0.3029, "step": 21363 }, { "epoch": 23.95, "learning_rate": 1.3309859154929578e-06, "loss": 0.2788, "step": 21364 }, { "epoch": 23.95, "learning_rate": 1.3295774647887324e-06, "loss": 0.2451, "step": 21365 }, { "epoch": 23.95, "learning_rate": 1.328169014084507e-06, "loss": 0.3453, "step": 21366 }, { "epoch": 23.95, "learning_rate": 1.3267605633802817e-06, "loss": 0.4139, "step": 21367 }, { "epoch": 23.96, "learning_rate": 1.3253521126760564e-06, "loss": 0.3832, "step": 21368 }, { "epoch": 23.96, "learning_rate": 1.323943661971831e-06, "loss": 0.2666, "step": 21369 }, { "epoch": 23.96, "learning_rate": 1.3225352112676056e-06, "loss": 0.3058, "step": 21370 }, { "epoch": 23.96, "learning_rate": 1.3211267605633803e-06, "loss": 0.2318, "step": 21371 }, { "epoch": 23.96, "learning_rate": 1.319718309859155e-06, "loss": 0.3439, "step": 21372 }, { "epoch": 23.96, "learning_rate": 1.3183098591549295e-06, "loss": 0.3623, "step": 21373 }, { "epoch": 23.96, "learning_rate": 1.3169014084507042e-06, "loss": 0.3751, "step": 21374 }, { "epoch": 23.96, "learning_rate": 1.3154929577464788e-06, "loss": 0.2902, "step": 21375 }, { "epoch": 23.96, "learning_rate": 1.3140845070422535e-06, "loss": 0.4081, "step": 21376 }, { "epoch": 23.97, "learning_rate": 1.312676056338028e-06, "loss": 0.2808, "step": 21377 }, { "epoch": 23.97, "learning_rate": 1.3112676056338027e-06, "loss": 0.213, "step": 21378 }, { "epoch": 23.97, "learning_rate": 1.3098591549295774e-06, "loss": 0.3512, "step": 21379 }, { "epoch": 23.97, "learning_rate": 1.308450704225352e-06, "loss": 0.369, "step": 21380 }, { "epoch": 23.97, "learning_rate": 1.3070422535211266e-06, "loss": 0.2595, "step": 21381 }, { "epoch": 23.97, "learning_rate": 1.3056338028169015e-06, "loss": 0.2888, "step": 21382 }, { "epoch": 23.97, "learning_rate": 1.3042253521126761e-06, "loss": 0.2349, "step": 21383 }, { "epoch": 23.97, "learning_rate": 1.3028169014084508e-06, "loss": 0.3275, "step": 21384 }, { "epoch": 23.97, "learning_rate": 1.3014084507042254e-06, "loss": 0.3361, "step": 21385 }, { "epoch": 23.98, "learning_rate": 1.3e-06, "loss": 0.287, "step": 21386 }, { "epoch": 23.98, "learning_rate": 1.2985915492957747e-06, "loss": 0.3321, "step": 21387 }, { "epoch": 23.98, "learning_rate": 1.2971830985915493e-06, "loss": 0.343, "step": 21388 }, { "epoch": 23.98, "learning_rate": 1.295774647887324e-06, "loss": 0.3108, "step": 21389 }, { "epoch": 23.98, "learning_rate": 1.2943661971830986e-06, "loss": 0.2906, "step": 21390 }, { "epoch": 23.98, "learning_rate": 1.2929577464788732e-06, "loss": 0.3056, "step": 21391 }, { "epoch": 23.98, "learning_rate": 1.2915492957746479e-06, "loss": 0.2706, "step": 21392 }, { "epoch": 23.98, "learning_rate": 1.2901408450704225e-06, "loss": 0.26, "step": 21393 }, { "epoch": 23.98, "learning_rate": 1.2887323943661971e-06, "loss": 0.3133, "step": 21394 }, { "epoch": 23.99, "learning_rate": 1.2873239436619718e-06, "loss": 0.3255, "step": 21395 }, { "epoch": 23.99, "learning_rate": 1.2859154929577464e-06, "loss": 0.2663, "step": 21396 }, { "epoch": 23.99, "learning_rate": 1.284507042253521e-06, "loss": 0.3319, "step": 21397 }, { "epoch": 23.99, "learning_rate": 1.2830985915492957e-06, "loss": 0.2283, "step": 21398 }, { "epoch": 23.99, "learning_rate": 1.2816901408450705e-06, "loss": 0.2758, "step": 21399 }, { "epoch": 23.99, "learning_rate": 1.2802816901408452e-06, "loss": 0.2887, "step": 21400 }, { "epoch": 23.99, "learning_rate": 1.2788732394366198e-06, "loss": 0.3888, "step": 21401 }, { "epoch": 23.99, "learning_rate": 1.2774647887323945e-06, "loss": 0.2695, "step": 21402 }, { "epoch": 23.99, "learning_rate": 1.276056338028169e-06, "loss": 0.2624, "step": 21403 }, { "epoch": 24.0, "learning_rate": 1.2746478873239437e-06, "loss": 0.2047, "step": 21404 }, { "epoch": 24.0, "learning_rate": 1.2732394366197184e-06, "loss": 0.2792, "step": 21405 }, { "epoch": 24.0, "learning_rate": 1.271830985915493e-06, "loss": 0.2516, "step": 21406 }, { "epoch": 24.0, "learning_rate": 1.2704225352112676e-06, "loss": 0.181, "step": 21407 }, { "epoch": 24.0, "learning_rate": 1.2690140845070423e-06, "loss": 0.3524, "step": 21408 }, { "epoch": 24.0, "learning_rate": 1.267605633802817e-06, "loss": 0.3964, "step": 21409 }, { "epoch": 24.0, "learning_rate": 1.2661971830985916e-06, "loss": 0.3198, "step": 21410 }, { "epoch": 24.0, "learning_rate": 1.2647887323943662e-06, "loss": 0.2951, "step": 21411 }, { "epoch": 24.0, "learning_rate": 1.2633802816901408e-06, "loss": 0.37, "step": 21412 }, { "epoch": 24.01, "learning_rate": 1.2619718309859155e-06, "loss": 0.303, "step": 21413 }, { "epoch": 24.01, "learning_rate": 1.2605633802816901e-06, "loss": 0.2726, "step": 21414 }, { "epoch": 24.01, "learning_rate": 1.2591549295774647e-06, "loss": 0.2539, "step": 21415 }, { "epoch": 24.01, "learning_rate": 1.2577464788732396e-06, "loss": 0.3648, "step": 21416 }, { "epoch": 24.01, "learning_rate": 1.2563380281690142e-06, "loss": 0.3177, "step": 21417 }, { "epoch": 24.01, "learning_rate": 1.2549295774647889e-06, "loss": 0.3338, "step": 21418 }, { "epoch": 24.01, "learning_rate": 1.2535211267605635e-06, "loss": 0.2917, "step": 21419 }, { "epoch": 24.01, "learning_rate": 1.2521126760563381e-06, "loss": 0.3564, "step": 21420 }, { "epoch": 24.01, "learning_rate": 1.2507042253521128e-06, "loss": 0.2936, "step": 21421 }, { "epoch": 24.02, "learning_rate": 1.2492957746478874e-06, "loss": 0.2556, "step": 21422 }, { "epoch": 24.02, "learning_rate": 1.247887323943662e-06, "loss": 0.3449, "step": 21423 }, { "epoch": 24.02, "learning_rate": 1.2464788732394367e-06, "loss": 0.2613, "step": 21424 }, { "epoch": 24.02, "learning_rate": 1.2450704225352113e-06, "loss": 0.2914, "step": 21425 }, { "epoch": 24.02, "learning_rate": 1.243661971830986e-06, "loss": 0.253, "step": 21426 }, { "epoch": 24.02, "learning_rate": 1.2422535211267606e-06, "loss": 0.225, "step": 21427 }, { "epoch": 24.02, "learning_rate": 1.2408450704225352e-06, "loss": 0.2678, "step": 21428 }, { "epoch": 24.02, "learning_rate": 1.2394366197183099e-06, "loss": 0.3312, "step": 21429 }, { "epoch": 24.02, "learning_rate": 1.2380281690140845e-06, "loss": 0.2996, "step": 21430 }, { "epoch": 24.03, "learning_rate": 1.2366197183098592e-06, "loss": 0.2694, "step": 21431 }, { "epoch": 24.03, "learning_rate": 1.2352112676056338e-06, "loss": 0.3003, "step": 21432 }, { "epoch": 24.03, "learning_rate": 1.2338028169014084e-06, "loss": 0.3544, "step": 21433 }, { "epoch": 24.03, "learning_rate": 1.2323943661971833e-06, "loss": 0.2823, "step": 21434 }, { "epoch": 24.03, "learning_rate": 1.230985915492958e-06, "loss": 0.3182, "step": 21435 }, { "epoch": 24.03, "learning_rate": 1.2295774647887326e-06, "loss": 0.3419, "step": 21436 }, { "epoch": 24.03, "learning_rate": 1.2281690140845072e-06, "loss": 0.2563, "step": 21437 }, { "epoch": 24.03, "learning_rate": 1.2267605633802818e-06, "loss": 0.2408, "step": 21438 }, { "epoch": 24.03, "learning_rate": 1.2253521126760565e-06, "loss": 0.3042, "step": 21439 }, { "epoch": 24.04, "learning_rate": 1.2239436619718311e-06, "loss": 0.3108, "step": 21440 }, { "epoch": 24.04, "learning_rate": 1.2225352112676057e-06, "loss": 0.3122, "step": 21441 }, { "epoch": 24.04, "learning_rate": 1.2211267605633804e-06, "loss": 0.2863, "step": 21442 }, { "epoch": 24.04, "learning_rate": 1.219718309859155e-06, "loss": 0.3005, "step": 21443 }, { "epoch": 24.04, "learning_rate": 1.2183098591549297e-06, "loss": 0.2513, "step": 21444 }, { "epoch": 24.04, "learning_rate": 1.2169014084507043e-06, "loss": 0.2593, "step": 21445 }, { "epoch": 24.04, "learning_rate": 1.215492957746479e-06, "loss": 0.2641, "step": 21446 }, { "epoch": 24.04, "learning_rate": 1.2140845070422536e-06, "loss": 0.2355, "step": 21447 }, { "epoch": 24.04, "learning_rate": 1.2126760563380282e-06, "loss": 0.2923, "step": 21448 }, { "epoch": 24.05, "learning_rate": 1.2112676056338028e-06, "loss": 0.2627, "step": 21449 }, { "epoch": 24.05, "learning_rate": 1.2098591549295775e-06, "loss": 0.2783, "step": 21450 }, { "epoch": 24.05, "learning_rate": 1.2084507042253523e-06, "loss": 0.2588, "step": 21451 }, { "epoch": 24.05, "learning_rate": 1.207042253521127e-06, "loss": 0.2315, "step": 21452 }, { "epoch": 24.05, "learning_rate": 1.2056338028169016e-06, "loss": 0.2669, "step": 21453 }, { "epoch": 24.05, "learning_rate": 1.2042253521126762e-06, "loss": 0.2075, "step": 21454 }, { "epoch": 24.05, "learning_rate": 1.2028169014084509e-06, "loss": 0.2128, "step": 21455 }, { "epoch": 24.05, "learning_rate": 1.2014084507042255e-06, "loss": 0.3737, "step": 21456 }, { "epoch": 24.05, "learning_rate": 1.2000000000000002e-06, "loss": 0.2066, "step": 21457 }, { "epoch": 24.06, "learning_rate": 1.1985915492957748e-06, "loss": 0.213, "step": 21458 }, { "epoch": 24.06, "learning_rate": 1.1971830985915494e-06, "loss": 0.4044, "step": 21459 }, { "epoch": 24.06, "learning_rate": 1.1957746478873239e-06, "loss": 0.3467, "step": 21460 }, { "epoch": 24.06, "learning_rate": 1.1943661971830985e-06, "loss": 0.2966, "step": 21461 }, { "epoch": 24.06, "learning_rate": 1.1929577464788731e-06, "loss": 0.3075, "step": 21462 }, { "epoch": 24.06, "learning_rate": 1.1915492957746478e-06, "loss": 0.324, "step": 21463 }, { "epoch": 24.06, "learning_rate": 1.1901408450704224e-06, "loss": 0.4039, "step": 21464 }, { "epoch": 24.06, "learning_rate": 1.188732394366197e-06, "loss": 0.3218, "step": 21465 }, { "epoch": 24.07, "learning_rate": 1.1873239436619717e-06, "loss": 0.3583, "step": 21466 }, { "epoch": 24.07, "learning_rate": 1.1859154929577463e-06, "loss": 0.3045, "step": 21467 }, { "epoch": 24.07, "learning_rate": 1.1845070422535212e-06, "loss": 0.3486, "step": 21468 }, { "epoch": 24.07, "learning_rate": 1.1830985915492958e-06, "loss": 0.3418, "step": 21469 }, { "epoch": 24.07, "learning_rate": 1.1816901408450704e-06, "loss": 0.306, "step": 21470 }, { "epoch": 24.07, "learning_rate": 1.180281690140845e-06, "loss": 0.3681, "step": 21471 }, { "epoch": 24.07, "learning_rate": 1.1788732394366197e-06, "loss": 0.2454, "step": 21472 }, { "epoch": 24.07, "learning_rate": 1.1774647887323944e-06, "loss": 0.2986, "step": 21473 }, { "epoch": 24.07, "learning_rate": 1.176056338028169e-06, "loss": 0.3477, "step": 21474 }, { "epoch": 24.08, "learning_rate": 1.1746478873239436e-06, "loss": 0.3902, "step": 21475 }, { "epoch": 24.08, "learning_rate": 1.1732394366197183e-06, "loss": 0.2707, "step": 21476 }, { "epoch": 24.08, "learning_rate": 1.171830985915493e-06, "loss": 0.278, "step": 21477 }, { "epoch": 24.08, "learning_rate": 1.1704225352112675e-06, "loss": 0.2357, "step": 21478 }, { "epoch": 24.08, "learning_rate": 1.1690140845070422e-06, "loss": 0.3663, "step": 21479 }, { "epoch": 24.08, "learning_rate": 1.1676056338028168e-06, "loss": 0.2743, "step": 21480 }, { "epoch": 24.08, "learning_rate": 1.1661971830985915e-06, "loss": 0.2877, "step": 21481 }, { "epoch": 24.08, "learning_rate": 1.164788732394366e-06, "loss": 0.1944, "step": 21482 }, { "epoch": 24.08, "learning_rate": 1.1633802816901407e-06, "loss": 0.2796, "step": 21483 }, { "epoch": 24.09, "learning_rate": 1.1619718309859154e-06, "loss": 0.3012, "step": 21484 }, { "epoch": 24.09, "learning_rate": 1.16056338028169e-06, "loss": 0.3239, "step": 21485 }, { "epoch": 24.09, "learning_rate": 1.1591549295774649e-06, "loss": 0.3797, "step": 21486 }, { "epoch": 24.09, "learning_rate": 1.1577464788732395e-06, "loss": 0.3459, "step": 21487 }, { "epoch": 24.09, "learning_rate": 1.1563380281690141e-06, "loss": 0.2757, "step": 21488 }, { "epoch": 24.09, "learning_rate": 1.1549295774647888e-06, "loss": 0.3006, "step": 21489 }, { "epoch": 24.09, "learning_rate": 1.1535211267605634e-06, "loss": 0.3253, "step": 21490 }, { "epoch": 24.09, "learning_rate": 1.152112676056338e-06, "loss": 0.2497, "step": 21491 }, { "epoch": 24.09, "learning_rate": 1.1507042253521127e-06, "loss": 0.2934, "step": 21492 }, { "epoch": 24.1, "learning_rate": 1.1492957746478873e-06, "loss": 0.3181, "step": 21493 }, { "epoch": 24.1, "learning_rate": 1.147887323943662e-06, "loss": 0.2839, "step": 21494 }, { "epoch": 24.1, "learning_rate": 1.1464788732394366e-06, "loss": 0.3164, "step": 21495 }, { "epoch": 24.1, "learning_rate": 1.1450704225352112e-06, "loss": 0.2454, "step": 21496 }, { "epoch": 24.1, "learning_rate": 1.1436619718309859e-06, "loss": 0.2404, "step": 21497 }, { "epoch": 24.1, "learning_rate": 1.1422535211267605e-06, "loss": 0.3111, "step": 21498 }, { "epoch": 24.1, "learning_rate": 1.1408450704225351e-06, "loss": 0.409, "step": 21499 }, { "epoch": 24.1, "learning_rate": 1.1394366197183098e-06, "loss": 0.2637, "step": 21500 }, { "epoch": 24.1, "learning_rate": 1.1380281690140844e-06, "loss": 0.2926, "step": 21501 }, { "epoch": 24.11, "learning_rate": 1.136619718309859e-06, "loss": 0.1928, "step": 21502 }, { "epoch": 24.11, "learning_rate": 1.135211267605634e-06, "loss": 0.1983, "step": 21503 }, { "epoch": 24.11, "learning_rate": 1.1338028169014085e-06, "loss": 0.3063, "step": 21504 }, { "epoch": 24.11, "learning_rate": 1.1323943661971832e-06, "loss": 0.1868, "step": 21505 }, { "epoch": 24.11, "learning_rate": 1.1309859154929578e-06, "loss": 0.1899, "step": 21506 }, { "epoch": 24.11, "learning_rate": 1.1295774647887325e-06, "loss": 0.304, "step": 21507 }, { "epoch": 24.11, "learning_rate": 1.128169014084507e-06, "loss": 0.177, "step": 21508 }, { "epoch": 24.11, "learning_rate": 1.1267605633802817e-06, "loss": 0.4145, "step": 21509 }, { "epoch": 24.11, "learning_rate": 1.1253521126760564e-06, "loss": 0.2966, "step": 21510 }, { "epoch": 24.12, "learning_rate": 1.123943661971831e-06, "loss": 0.3559, "step": 21511 }, { "epoch": 24.12, "learning_rate": 1.1225352112676056e-06, "loss": 0.3379, "step": 21512 }, { "epoch": 24.12, "learning_rate": 1.1211267605633803e-06, "loss": 0.3347, "step": 21513 }, { "epoch": 24.12, "learning_rate": 1.119718309859155e-06, "loss": 0.3365, "step": 21514 }, { "epoch": 24.12, "learning_rate": 1.1183098591549296e-06, "loss": 0.2901, "step": 21515 }, { "epoch": 24.12, "learning_rate": 1.1169014084507042e-06, "loss": 0.3629, "step": 21516 }, { "epoch": 24.12, "learning_rate": 1.1154929577464788e-06, "loss": 0.3054, "step": 21517 }, { "epoch": 24.12, "learning_rate": 1.1140845070422535e-06, "loss": 0.2859, "step": 21518 }, { "epoch": 24.12, "learning_rate": 1.1126760563380281e-06, "loss": 0.2852, "step": 21519 }, { "epoch": 24.13, "learning_rate": 1.111267605633803e-06, "loss": 0.2821, "step": 21520 }, { "epoch": 24.13, "learning_rate": 1.1098591549295776e-06, "loss": 0.3265, "step": 21521 }, { "epoch": 24.13, "learning_rate": 1.1084507042253522e-06, "loss": 0.346, "step": 21522 }, { "epoch": 24.13, "learning_rate": 1.1070422535211269e-06, "loss": 0.2876, "step": 21523 }, { "epoch": 24.13, "learning_rate": 1.1056338028169015e-06, "loss": 0.2959, "step": 21524 }, { "epoch": 24.13, "learning_rate": 1.1042253521126761e-06, "loss": 0.2723, "step": 21525 }, { "epoch": 24.13, "learning_rate": 1.1028169014084508e-06, "loss": 0.2974, "step": 21526 }, { "epoch": 24.13, "learning_rate": 1.1014084507042254e-06, "loss": 0.294, "step": 21527 }, { "epoch": 24.13, "learning_rate": 1.1e-06, "loss": 0.3218, "step": 21528 }, { "epoch": 24.14, "learning_rate": 1.0985915492957747e-06, "loss": 0.373, "step": 21529 }, { "epoch": 24.14, "learning_rate": 1.0971830985915493e-06, "loss": 0.2531, "step": 21530 }, { "epoch": 24.14, "learning_rate": 1.095774647887324e-06, "loss": 0.2734, "step": 21531 }, { "epoch": 24.14, "learning_rate": 1.0943661971830986e-06, "loss": 0.2392, "step": 21532 }, { "epoch": 24.14, "learning_rate": 1.0929577464788732e-06, "loss": 0.2135, "step": 21533 }, { "epoch": 24.14, "learning_rate": 1.0915492957746479e-06, "loss": 0.3062, "step": 21534 }, { "epoch": 24.14, "learning_rate": 1.0901408450704225e-06, "loss": 0.3443, "step": 21535 }, { "epoch": 24.14, "learning_rate": 1.0887323943661972e-06, "loss": 0.322, "step": 21536 }, { "epoch": 24.14, "learning_rate": 1.0873239436619718e-06, "loss": 0.234, "step": 21537 }, { "epoch": 24.15, "learning_rate": 1.0859154929577466e-06, "loss": 0.2681, "step": 21538 }, { "epoch": 24.15, "learning_rate": 1.0845070422535213e-06, "loss": 0.2295, "step": 21539 }, { "epoch": 24.15, "learning_rate": 1.083098591549296e-06, "loss": 0.2902, "step": 21540 }, { "epoch": 24.15, "learning_rate": 1.0816901408450706e-06, "loss": 0.2444, "step": 21541 }, { "epoch": 24.15, "learning_rate": 1.0802816901408452e-06, "loss": 0.2939, "step": 21542 }, { "epoch": 24.15, "learning_rate": 1.0788732394366198e-06, "loss": 0.2812, "step": 21543 }, { "epoch": 24.15, "learning_rate": 1.0774647887323945e-06, "loss": 0.2785, "step": 21544 }, { "epoch": 24.15, "learning_rate": 1.076056338028169e-06, "loss": 0.2265, "step": 21545 }, { "epoch": 24.15, "learning_rate": 1.0746478873239437e-06, "loss": 0.2266, "step": 21546 }, { "epoch": 24.16, "learning_rate": 1.0732394366197184e-06, "loss": 0.2752, "step": 21547 }, { "epoch": 24.16, "learning_rate": 1.071830985915493e-06, "loss": 0.3636, "step": 21548 }, { "epoch": 24.16, "learning_rate": 1.0704225352112677e-06, "loss": 0.2693, "step": 21549 }, { "epoch": 24.16, "learning_rate": 1.0690140845070423e-06, "loss": 0.3414, "step": 21550 }, { "epoch": 24.16, "learning_rate": 1.067605633802817e-06, "loss": 0.2334, "step": 21551 }, { "epoch": 24.16, "learning_rate": 1.0661971830985916e-06, "loss": 0.2725, "step": 21552 }, { "epoch": 24.16, "learning_rate": 1.0647887323943662e-06, "loss": 0.2325, "step": 21553 }, { "epoch": 24.16, "learning_rate": 1.0633802816901408e-06, "loss": 0.218, "step": 21554 }, { "epoch": 24.16, "learning_rate": 1.0619718309859157e-06, "loss": 0.1855, "step": 21555 }, { "epoch": 24.17, "learning_rate": 1.0605633802816903e-06, "loss": 0.1489, "step": 21556 }, { "epoch": 24.17, "learning_rate": 1.059154929577465e-06, "loss": 0.2514, "step": 21557 }, { "epoch": 24.17, "learning_rate": 1.0577464788732396e-06, "loss": 0.2535, "step": 21558 }, { "epoch": 24.17, "learning_rate": 1.0563380281690142e-06, "loss": 0.373, "step": 21559 }, { "epoch": 24.17, "learning_rate": 1.0549295774647889e-06, "loss": 0.4049, "step": 21560 }, { "epoch": 24.17, "learning_rate": 1.0535211267605635e-06, "loss": 0.3266, "step": 21561 }, { "epoch": 24.17, "learning_rate": 1.0521126760563382e-06, "loss": 0.3148, "step": 21562 }, { "epoch": 24.17, "learning_rate": 1.0507042253521128e-06, "loss": 0.3657, "step": 21563 }, { "epoch": 24.17, "learning_rate": 1.0492957746478874e-06, "loss": 0.2925, "step": 21564 }, { "epoch": 24.18, "learning_rate": 1.047887323943662e-06, "loss": 0.427, "step": 21565 }, { "epoch": 24.18, "learning_rate": 1.0464788732394367e-06, "loss": 0.3188, "step": 21566 }, { "epoch": 24.18, "learning_rate": 1.0450704225352113e-06, "loss": 0.2991, "step": 21567 }, { "epoch": 24.18, "learning_rate": 1.043661971830986e-06, "loss": 0.3386, "step": 21568 }, { "epoch": 24.18, "learning_rate": 1.0422535211267606e-06, "loss": 0.3037, "step": 21569 }, { "epoch": 24.18, "learning_rate": 1.0408450704225353e-06, "loss": 0.35, "step": 21570 }, { "epoch": 24.18, "learning_rate": 1.0394366197183099e-06, "loss": 0.2838, "step": 21571 }, { "epoch": 24.18, "learning_rate": 1.0380281690140845e-06, "loss": 0.2883, "step": 21572 }, { "epoch": 24.18, "learning_rate": 1.0366197183098592e-06, "loss": 0.3697, "step": 21573 }, { "epoch": 24.19, "learning_rate": 1.0352112676056338e-06, "loss": 0.2334, "step": 21574 }, { "epoch": 24.19, "learning_rate": 1.0338028169014084e-06, "loss": 0.3374, "step": 21575 }, { "epoch": 24.19, "learning_rate": 1.032394366197183e-06, "loss": 0.3392, "step": 21576 }, { "epoch": 24.19, "learning_rate": 1.0309859154929577e-06, "loss": 0.3703, "step": 21577 }, { "epoch": 24.19, "learning_rate": 1.0295774647887324e-06, "loss": 0.3258, "step": 21578 }, { "epoch": 24.19, "learning_rate": 1.028169014084507e-06, "loss": 0.2872, "step": 21579 }, { "epoch": 24.19, "learning_rate": 1.0267605633802816e-06, "loss": 0.233, "step": 21580 }, { "epoch": 24.19, "learning_rate": 1.0253521126760563e-06, "loss": 0.2855, "step": 21581 }, { "epoch": 24.2, "learning_rate": 1.023943661971831e-06, "loss": 0.3014, "step": 21582 }, { "epoch": 24.2, "learning_rate": 1.0225352112676055e-06, "loss": 0.2798, "step": 21583 }, { "epoch": 24.2, "learning_rate": 1.0211267605633802e-06, "loss": 0.2638, "step": 21584 }, { "epoch": 24.2, "learning_rate": 1.0197183098591548e-06, "loss": 0.2799, "step": 21585 }, { "epoch": 24.2, "learning_rate": 1.0183098591549295e-06, "loss": 0.2691, "step": 21586 }, { "epoch": 24.2, "learning_rate": 1.016901408450704e-06, "loss": 0.2858, "step": 21587 }, { "epoch": 24.2, "learning_rate": 1.0154929577464787e-06, "loss": 0.2737, "step": 21588 }, { "epoch": 24.2, "learning_rate": 1.0140845070422534e-06, "loss": 0.2524, "step": 21589 }, { "epoch": 24.2, "learning_rate": 1.0126760563380282e-06, "loss": 0.2082, "step": 21590 }, { "epoch": 24.21, "learning_rate": 1.0112676056338029e-06, "loss": 0.3073, "step": 21591 }, { "epoch": 24.21, "learning_rate": 1.0098591549295775e-06, "loss": 0.2677, "step": 21592 }, { "epoch": 24.21, "learning_rate": 1.0084507042253521e-06, "loss": 0.349, "step": 21593 }, { "epoch": 24.21, "learning_rate": 1.0070422535211268e-06, "loss": 0.3959, "step": 21594 }, { "epoch": 24.21, "learning_rate": 1.0056338028169014e-06, "loss": 0.3387, "step": 21595 }, { "epoch": 24.21, "learning_rate": 1.004225352112676e-06, "loss": 0.2702, "step": 21596 }, { "epoch": 24.21, "learning_rate": 1.0028169014084507e-06, "loss": 0.2428, "step": 21597 }, { "epoch": 24.21, "learning_rate": 1.0014084507042253e-06, "loss": 0.3475, "step": 21598 }, { "epoch": 24.21, "learning_rate": 1e-06, "loss": 0.2782, "step": 21599 }, { "epoch": 24.22, "learning_rate": 9.985915492957746e-07, "loss": 0.2894, "step": 21600 }, { "epoch": 24.22, "learning_rate": 9.971830985915492e-07, "loss": 0.2868, "step": 21601 }, { "epoch": 24.22, "learning_rate": 9.957746478873239e-07, "loss": 0.2483, "step": 21602 }, { "epoch": 24.22, "learning_rate": 9.943661971830985e-07, "loss": 0.3899, "step": 21603 }, { "epoch": 24.22, "learning_rate": 9.929577464788731e-07, "loss": 0.2778, "step": 21604 }, { "epoch": 24.22, "learning_rate": 9.915492957746478e-07, "loss": 0.3563, "step": 21605 }, { "epoch": 24.22, "learning_rate": 9.901408450704224e-07, "loss": 0.2513, "step": 21606 }, { "epoch": 24.22, "learning_rate": 9.887323943661973e-07, "loss": 0.2344, "step": 21607 }, { "epoch": 24.22, "learning_rate": 9.87323943661972e-07, "loss": 0.2143, "step": 21608 }, { "epoch": 24.23, "learning_rate": 9.859154929577465e-07, "loss": 0.278, "step": 21609 }, { "epoch": 24.23, "learning_rate": 9.845070422535212e-07, "loss": 0.4174, "step": 21610 }, { "epoch": 24.23, "learning_rate": 9.830985915492958e-07, "loss": 0.3427, "step": 21611 }, { "epoch": 24.23, "learning_rate": 9.816901408450705e-07, "loss": 0.4242, "step": 21612 }, { "epoch": 24.23, "learning_rate": 9.80281690140845e-07, "loss": 0.3141, "step": 21613 }, { "epoch": 24.23, "learning_rate": 9.788732394366197e-07, "loss": 0.3533, "step": 21614 }, { "epoch": 24.23, "learning_rate": 9.774647887323944e-07, "loss": 0.373, "step": 21615 }, { "epoch": 24.23, "learning_rate": 9.76056338028169e-07, "loss": 0.3658, "step": 21616 }, { "epoch": 24.23, "learning_rate": 9.746478873239436e-07, "loss": 0.3421, "step": 21617 }, { "epoch": 24.24, "learning_rate": 9.732394366197183e-07, "loss": 0.3446, "step": 21618 }, { "epoch": 24.24, "learning_rate": 9.71830985915493e-07, "loss": 0.2872, "step": 21619 }, { "epoch": 24.24, "learning_rate": 9.704225352112676e-07, "loss": 0.3052, "step": 21620 }, { "epoch": 24.24, "learning_rate": 9.690140845070422e-07, "loss": 0.3078, "step": 21621 }, { "epoch": 24.24, "learning_rate": 9.676056338028168e-07, "loss": 0.356, "step": 21622 }, { "epoch": 24.24, "learning_rate": 9.661971830985915e-07, "loss": 0.3077, "step": 21623 }, { "epoch": 24.24, "learning_rate": 9.647887323943663e-07, "loss": 0.2687, "step": 21624 }, { "epoch": 24.24, "learning_rate": 9.63380281690141e-07, "loss": 0.331, "step": 21625 }, { "epoch": 24.24, "learning_rate": 9.619718309859156e-07, "loss": 0.3053, "step": 21626 }, { "epoch": 24.25, "learning_rate": 9.605633802816902e-07, "loss": 0.3473, "step": 21627 }, { "epoch": 24.25, "learning_rate": 9.591549295774649e-07, "loss": 0.2898, "step": 21628 }, { "epoch": 24.25, "learning_rate": 9.577464788732395e-07, "loss": 0.2687, "step": 21629 }, { "epoch": 24.25, "learning_rate": 9.563380281690141e-07, "loss": 0.3216, "step": 21630 }, { "epoch": 24.25, "learning_rate": 9.549295774647888e-07, "loss": 0.3034, "step": 21631 }, { "epoch": 24.25, "learning_rate": 9.535211267605634e-07, "loss": 0.3023, "step": 21632 }, { "epoch": 24.25, "learning_rate": 9.521126760563381e-07, "loss": 0.2705, "step": 21633 }, { "epoch": 24.25, "learning_rate": 9.507042253521127e-07, "loss": 0.3015, "step": 21634 }, { "epoch": 24.25, "learning_rate": 9.492957746478873e-07, "loss": 0.3129, "step": 21635 }, { "epoch": 24.26, "learning_rate": 9.47887323943662e-07, "loss": 0.3105, "step": 21636 }, { "epoch": 24.26, "learning_rate": 9.464788732394366e-07, "loss": 0.3273, "step": 21637 }, { "epoch": 24.26, "learning_rate": 9.450704225352112e-07, "loss": 0.2577, "step": 21638 }, { "epoch": 24.26, "learning_rate": 9.436619718309859e-07, "loss": 0.2841, "step": 21639 }, { "epoch": 24.26, "learning_rate": 9.422535211267605e-07, "loss": 0.3687, "step": 21640 }, { "epoch": 24.26, "learning_rate": 9.408450704225352e-07, "loss": 0.2434, "step": 21641 }, { "epoch": 24.26, "learning_rate": 9.3943661971831e-07, "loss": 0.2568, "step": 21642 }, { "epoch": 24.26, "learning_rate": 9.380281690140846e-07, "loss": 0.2583, "step": 21643 }, { "epoch": 24.26, "learning_rate": 9.366197183098592e-07, "loss": 0.2722, "step": 21644 }, { "epoch": 24.27, "learning_rate": 9.352112676056338e-07, "loss": 0.292, "step": 21645 }, { "epoch": 24.27, "learning_rate": 9.338028169014086e-07, "loss": 0.3165, "step": 21646 }, { "epoch": 24.27, "learning_rate": 9.323943661971832e-07, "loss": 0.3171, "step": 21647 }, { "epoch": 24.27, "learning_rate": 9.309859154929578e-07, "loss": 0.3017, "step": 21648 }, { "epoch": 24.27, "learning_rate": 9.295774647887325e-07, "loss": 0.214, "step": 21649 }, { "epoch": 24.27, "learning_rate": 9.281690140845071e-07, "loss": 0.3304, "step": 21650 }, { "epoch": 24.27, "learning_rate": 9.267605633802817e-07, "loss": 0.3019, "step": 21651 }, { "epoch": 24.27, "learning_rate": 9.253521126760564e-07, "loss": 0.2209, "step": 21652 }, { "epoch": 24.27, "learning_rate": 9.23943661971831e-07, "loss": 0.2084, "step": 21653 }, { "epoch": 24.28, "learning_rate": 9.225352112676057e-07, "loss": 0.2493, "step": 21654 }, { "epoch": 24.28, "learning_rate": 9.211267605633804e-07, "loss": 0.2832, "step": 21655 }, { "epoch": 24.28, "learning_rate": 9.19718309859155e-07, "loss": 0.2341, "step": 21656 }, { "epoch": 24.28, "learning_rate": 9.183098591549297e-07, "loss": 0.2111, "step": 21657 }, { "epoch": 24.28, "learning_rate": 9.169014084507042e-07, "loss": 0.3907, "step": 21658 }, { "epoch": 24.28, "learning_rate": 9.154929577464788e-07, "loss": 0.4353, "step": 21659 }, { "epoch": 24.28, "learning_rate": 9.140845070422535e-07, "loss": 0.3477, "step": 21660 }, { "epoch": 24.28, "learning_rate": 9.126760563380281e-07, "loss": 0.368, "step": 21661 }, { "epoch": 24.28, "learning_rate": 9.112676056338028e-07, "loss": 0.3212, "step": 21662 }, { "epoch": 24.29, "learning_rate": 9.098591549295775e-07, "loss": 0.4572, "step": 21663 }, { "epoch": 24.29, "learning_rate": 9.084507042253521e-07, "loss": 0.3151, "step": 21664 }, { "epoch": 24.29, "learning_rate": 9.070422535211268e-07, "loss": 0.3328, "step": 21665 }, { "epoch": 24.29, "learning_rate": 9.056338028169014e-07, "loss": 0.3637, "step": 21666 }, { "epoch": 24.29, "learning_rate": 9.04225352112676e-07, "loss": 0.2481, "step": 21667 }, { "epoch": 24.29, "learning_rate": 9.028169014084507e-07, "loss": 0.2474, "step": 21668 }, { "epoch": 24.29, "learning_rate": 9.014084507042253e-07, "loss": 0.2564, "step": 21669 }, { "epoch": 24.29, "learning_rate": 9e-07, "loss": 0.3149, "step": 21670 }, { "epoch": 24.29, "learning_rate": 8.985915492957746e-07, "loss": 0.3066, "step": 21671 }, { "epoch": 24.3, "learning_rate": 8.971830985915493e-07, "loss": 0.3742, "step": 21672 }, { "epoch": 24.3, "learning_rate": 8.95774647887324e-07, "loss": 0.2715, "step": 21673 }, { "epoch": 24.3, "learning_rate": 8.943661971830986e-07, "loss": 0.3355, "step": 21674 }, { "epoch": 24.3, "learning_rate": 8.929577464788733e-07, "loss": 0.2898, "step": 21675 }, { "epoch": 24.3, "learning_rate": 8.915492957746479e-07, "loss": 0.3231, "step": 21676 }, { "epoch": 24.3, "learning_rate": 8.901408450704225e-07, "loss": 0.3065, "step": 21677 }, { "epoch": 24.3, "learning_rate": 8.887323943661972e-07, "loss": 0.2216, "step": 21678 }, { "epoch": 24.3, "learning_rate": 8.873239436619718e-07, "loss": 0.3008, "step": 21679 }, { "epoch": 24.3, "learning_rate": 8.859154929577464e-07, "loss": 0.2558, "step": 21680 }, { "epoch": 24.31, "learning_rate": 8.845070422535212e-07, "loss": 0.342, "step": 21681 }, { "epoch": 24.31, "learning_rate": 8.830985915492958e-07, "loss": 0.2749, "step": 21682 }, { "epoch": 24.31, "learning_rate": 8.816901408450705e-07, "loss": 0.2851, "step": 21683 }, { "epoch": 24.31, "learning_rate": 8.802816901408451e-07, "loss": 0.3217, "step": 21684 }, { "epoch": 24.31, "learning_rate": 8.788732394366197e-07, "loss": 0.2337, "step": 21685 }, { "epoch": 24.31, "learning_rate": 8.774647887323944e-07, "loss": 0.3238, "step": 21686 }, { "epoch": 24.31, "learning_rate": 8.76056338028169e-07, "loss": 0.2431, "step": 21687 }, { "epoch": 24.31, "learning_rate": 8.746478873239436e-07, "loss": 0.2913, "step": 21688 }, { "epoch": 24.32, "learning_rate": 8.732394366197184e-07, "loss": 0.3588, "step": 21689 }, { "epoch": 24.32, "learning_rate": 8.71830985915493e-07, "loss": 0.3056, "step": 21690 }, { "epoch": 24.32, "learning_rate": 8.704225352112677e-07, "loss": 0.2969, "step": 21691 }, { "epoch": 24.32, "learning_rate": 8.690140845070423e-07, "loss": 0.3194, "step": 21692 }, { "epoch": 24.32, "learning_rate": 8.676056338028169e-07, "loss": 0.3441, "step": 21693 }, { "epoch": 24.32, "learning_rate": 8.661971830985916e-07, "loss": 0.3147, "step": 21694 }, { "epoch": 24.32, "learning_rate": 8.647887323943662e-07, "loss": 0.3219, "step": 21695 }, { "epoch": 24.32, "learning_rate": 8.633802816901409e-07, "loss": 0.2664, "step": 21696 }, { "epoch": 24.32, "learning_rate": 8.619718309859155e-07, "loss": 0.2435, "step": 21697 }, { "epoch": 24.33, "learning_rate": 8.605633802816902e-07, "loss": 0.2545, "step": 21698 }, { "epoch": 24.33, "learning_rate": 8.591549295774649e-07, "loss": 0.3279, "step": 21699 }, { "epoch": 24.33, "learning_rate": 8.577464788732395e-07, "loss": 0.2567, "step": 21700 }, { "epoch": 24.33, "learning_rate": 8.563380281690141e-07, "loss": 0.2294, "step": 21701 }, { "epoch": 24.33, "learning_rate": 8.549295774647888e-07, "loss": 0.3522, "step": 21702 }, { "epoch": 24.33, "learning_rate": 8.535211267605634e-07, "loss": 0.2598, "step": 21703 }, { "epoch": 24.33, "learning_rate": 8.521126760563381e-07, "loss": 0.2706, "step": 21704 }, { "epoch": 24.33, "learning_rate": 8.507042253521127e-07, "loss": 0.2597, "step": 21705 }, { "epoch": 24.33, "learning_rate": 8.492957746478873e-07, "loss": 0.2555, "step": 21706 }, { "epoch": 24.34, "learning_rate": 8.478873239436621e-07, "loss": 0.2879, "step": 21707 }, { "epoch": 24.34, "learning_rate": 8.464788732394367e-07, "loss": 0.3613, "step": 21708 }, { "epoch": 24.34, "learning_rate": 8.450704225352114e-07, "loss": 0.3378, "step": 21709 }, { "epoch": 24.34, "learning_rate": 8.43661971830986e-07, "loss": 0.4675, "step": 21710 }, { "epoch": 24.34, "learning_rate": 8.422535211267606e-07, "loss": 0.3648, "step": 21711 }, { "epoch": 24.34, "learning_rate": 8.408450704225353e-07, "loss": 0.3422, "step": 21712 }, { "epoch": 24.34, "learning_rate": 8.394366197183099e-07, "loss": 0.366, "step": 21713 }, { "epoch": 24.34, "learning_rate": 8.380281690140845e-07, "loss": 0.3495, "step": 21714 }, { "epoch": 24.34, "learning_rate": 8.366197183098592e-07, "loss": 0.323, "step": 21715 }, { "epoch": 24.35, "learning_rate": 8.352112676056338e-07, "loss": 0.2935, "step": 21716 }, { "epoch": 24.35, "learning_rate": 8.338028169014085e-07, "loss": 0.3006, "step": 21717 }, { "epoch": 24.35, "learning_rate": 8.323943661971831e-07, "loss": 0.3467, "step": 21718 }, { "epoch": 24.35, "learning_rate": 8.309859154929577e-07, "loss": 0.3046, "step": 21719 }, { "epoch": 24.35, "learning_rate": 8.295774647887324e-07, "loss": 0.3775, "step": 21720 }, { "epoch": 24.35, "learning_rate": 8.28169014084507e-07, "loss": 0.3075, "step": 21721 }, { "epoch": 24.35, "learning_rate": 8.267605633802816e-07, "loss": 0.2567, "step": 21722 }, { "epoch": 24.35, "learning_rate": 8.253521126760563e-07, "loss": 0.2696, "step": 21723 }, { "epoch": 24.35, "learning_rate": 8.23943661971831e-07, "loss": 0.3685, "step": 21724 }, { "epoch": 24.36, "learning_rate": 8.225352112676057e-07, "loss": 0.2905, "step": 21725 }, { "epoch": 24.36, "learning_rate": 8.211267605633803e-07, "loss": 0.3195, "step": 21726 }, { "epoch": 24.36, "learning_rate": 8.197183098591549e-07, "loss": 0.3035, "step": 21727 }, { "epoch": 24.36, "learning_rate": 8.183098591549296e-07, "loss": 0.3304, "step": 21728 }, { "epoch": 24.36, "learning_rate": 8.169014084507042e-07, "loss": 0.2778, "step": 21729 }, { "epoch": 24.36, "learning_rate": 8.154929577464788e-07, "loss": 0.308, "step": 21730 }, { "epoch": 24.36, "learning_rate": 8.140845070422535e-07, "loss": 0.2569, "step": 21731 }, { "epoch": 24.36, "learning_rate": 8.126760563380281e-07, "loss": 0.2697, "step": 21732 }, { "epoch": 24.36, "learning_rate": 8.112676056338029e-07, "loss": 0.3079, "step": 21733 }, { "epoch": 24.37, "learning_rate": 8.098591549295775e-07, "loss": 0.3647, "step": 21734 }, { "epoch": 24.37, "learning_rate": 8.084507042253521e-07, "loss": 0.3499, "step": 21735 }, { "epoch": 24.37, "learning_rate": 8.070422535211268e-07, "loss": 0.2638, "step": 21736 }, { "epoch": 24.37, "learning_rate": 8.056338028169014e-07, "loss": 0.2824, "step": 21737 }, { "epoch": 24.37, "learning_rate": 8.04225352112676e-07, "loss": 0.3913, "step": 21738 }, { "epoch": 24.37, "learning_rate": 8.028169014084507e-07, "loss": 0.3188, "step": 21739 }, { "epoch": 24.37, "learning_rate": 8.014084507042253e-07, "loss": 0.2865, "step": 21740 }, { "epoch": 24.37, "learning_rate": 8.000000000000001e-07, "loss": 0.2636, "step": 21741 }, { "epoch": 24.37, "learning_rate": 7.985915492957747e-07, "loss": 0.233, "step": 21742 }, { "epoch": 24.38, "learning_rate": 7.971830985915493e-07, "loss": 0.2744, "step": 21743 }, { "epoch": 24.38, "learning_rate": 7.95774647887324e-07, "loss": 0.2643, "step": 21744 }, { "epoch": 24.38, "learning_rate": 7.943661971830986e-07, "loss": 0.2859, "step": 21745 }, { "epoch": 24.38, "learning_rate": 7.929577464788733e-07, "loss": 0.2835, "step": 21746 }, { "epoch": 24.38, "learning_rate": 7.915492957746479e-07, "loss": 0.3003, "step": 21747 }, { "epoch": 24.38, "learning_rate": 7.901408450704225e-07, "loss": 0.2883, "step": 21748 }, { "epoch": 24.38, "learning_rate": 7.887323943661972e-07, "loss": 0.3635, "step": 21749 }, { "epoch": 24.38, "learning_rate": 7.873239436619719e-07, "loss": 0.3071, "step": 21750 }, { "epoch": 24.38, "learning_rate": 7.859154929577466e-07, "loss": 0.2207, "step": 21751 }, { "epoch": 24.39, "learning_rate": 7.845070422535212e-07, "loss": 0.2245, "step": 21752 }, { "epoch": 24.39, "learning_rate": 7.830985915492958e-07, "loss": 0.2724, "step": 21753 }, { "epoch": 24.39, "learning_rate": 7.816901408450705e-07, "loss": 0.2609, "step": 21754 }, { "epoch": 24.39, "learning_rate": 7.802816901408451e-07, "loss": 0.263, "step": 21755 }, { "epoch": 24.39, "learning_rate": 7.788732394366197e-07, "loss": 0.3215, "step": 21756 }, { "epoch": 24.39, "learning_rate": 7.774647887323944e-07, "loss": 0.2562, "step": 21757 }, { "epoch": 24.39, "learning_rate": 7.76056338028169e-07, "loss": 0.2726, "step": 21758 }, { "epoch": 24.39, "learning_rate": 7.746478873239438e-07, "loss": 0.4102, "step": 21759 }, { "epoch": 24.39, "learning_rate": 7.732394366197184e-07, "loss": 0.4269, "step": 21760 }, { "epoch": 24.4, "learning_rate": 7.71830985915493e-07, "loss": 0.3055, "step": 21761 }, { "epoch": 24.4, "learning_rate": 7.704225352112677e-07, "loss": 0.3393, "step": 21762 }, { "epoch": 24.4, "learning_rate": 7.690140845070423e-07, "loss": 0.3122, "step": 21763 }, { "epoch": 24.4, "learning_rate": 7.676056338028169e-07, "loss": 0.3371, "step": 21764 }, { "epoch": 24.4, "learning_rate": 7.661971830985916e-07, "loss": 0.3618, "step": 21765 }, { "epoch": 24.4, "learning_rate": 7.647887323943662e-07, "loss": 0.3877, "step": 21766 }, { "epoch": 24.4, "learning_rate": 7.63380281690141e-07, "loss": 0.3593, "step": 21767 }, { "epoch": 24.4, "learning_rate": 7.619718309859156e-07, "loss": 0.4008, "step": 21768 }, { "epoch": 24.4, "learning_rate": 7.605633802816902e-07, "loss": 0.2913, "step": 21769 }, { "epoch": 24.41, "learning_rate": 7.591549295774649e-07, "loss": 0.3007, "step": 21770 }, { "epoch": 24.41, "learning_rate": 7.577464788732394e-07, "loss": 0.2728, "step": 21771 }, { "epoch": 24.41, "learning_rate": 7.56338028169014e-07, "loss": 0.3674, "step": 21772 }, { "epoch": 24.41, "learning_rate": 7.549295774647887e-07, "loss": 0.312, "step": 21773 }, { "epoch": 24.41, "learning_rate": 7.535211267605633e-07, "loss": 0.2446, "step": 21774 }, { "epoch": 24.41, "learning_rate": 7.52112676056338e-07, "loss": 0.2939, "step": 21775 }, { "epoch": 24.41, "learning_rate": 7.507042253521127e-07, "loss": 0.2742, "step": 21776 }, { "epoch": 24.41, "learning_rate": 7.492957746478873e-07, "loss": 0.3784, "step": 21777 }, { "epoch": 24.41, "learning_rate": 7.47887323943662e-07, "loss": 0.3289, "step": 21778 }, { "epoch": 24.42, "learning_rate": 7.464788732394366e-07, "loss": 0.3553, "step": 21779 }, { "epoch": 24.42, "learning_rate": 7.450704225352113e-07, "loss": 0.2778, "step": 21780 }, { "epoch": 24.42, "learning_rate": 7.436619718309859e-07, "loss": 0.3269, "step": 21781 }, { "epoch": 24.42, "learning_rate": 7.422535211267605e-07, "loss": 0.3116, "step": 21782 }, { "epoch": 24.42, "learning_rate": 7.408450704225352e-07, "loss": 0.4061, "step": 21783 }, { "epoch": 24.42, "learning_rate": 7.394366197183098e-07, "loss": 0.2649, "step": 21784 }, { "epoch": 24.42, "learning_rate": 7.380281690140845e-07, "loss": 0.2315, "step": 21785 }, { "epoch": 24.42, "learning_rate": 7.366197183098592e-07, "loss": 0.255, "step": 21786 }, { "epoch": 24.42, "learning_rate": 7.352112676056338e-07, "loss": 0.3133, "step": 21787 }, { "epoch": 24.43, "learning_rate": 7.338028169014085e-07, "loss": 0.3469, "step": 21788 }, { "epoch": 24.43, "learning_rate": 7.323943661971831e-07, "loss": 0.3016, "step": 21789 }, { "epoch": 24.43, "learning_rate": 7.309859154929577e-07, "loss": 0.2749, "step": 21790 }, { "epoch": 24.43, "learning_rate": 7.295774647887324e-07, "loss": 0.3252, "step": 21791 }, { "epoch": 24.43, "learning_rate": 7.28169014084507e-07, "loss": 0.2864, "step": 21792 }, { "epoch": 24.43, "learning_rate": 7.267605633802817e-07, "loss": 0.2465, "step": 21793 }, { "epoch": 24.43, "learning_rate": 7.253521126760564e-07, "loss": 0.2667, "step": 21794 }, { "epoch": 24.43, "learning_rate": 7.23943661971831e-07, "loss": 0.3511, "step": 21795 }, { "epoch": 24.43, "learning_rate": 7.225352112676057e-07, "loss": 0.2999, "step": 21796 }, { "epoch": 24.44, "learning_rate": 7.211267605633803e-07, "loss": 0.2893, "step": 21797 }, { "epoch": 24.44, "learning_rate": 7.197183098591549e-07, "loss": 0.2677, "step": 21798 }, { "epoch": 24.44, "learning_rate": 7.183098591549296e-07, "loss": 0.2862, "step": 21799 }, { "epoch": 24.44, "learning_rate": 7.169014084507042e-07, "loss": 0.2757, "step": 21800 }, { "epoch": 24.44, "learning_rate": 7.154929577464788e-07, "loss": 0.2742, "step": 21801 }, { "epoch": 24.44, "learning_rate": 7.140845070422536e-07, "loss": 0.2759, "step": 21802 }, { "epoch": 24.44, "learning_rate": 7.126760563380282e-07, "loss": 0.2393, "step": 21803 }, { "epoch": 24.44, "learning_rate": 7.112676056338029e-07, "loss": 0.2919, "step": 21804 }, { "epoch": 24.45, "learning_rate": 7.098591549295775e-07, "loss": 0.1986, "step": 21805 }, { "epoch": 24.45, "learning_rate": 7.084507042253521e-07, "loss": 0.2665, "step": 21806 }, { "epoch": 24.45, "learning_rate": 7.070422535211268e-07, "loss": 0.2701, "step": 21807 }, { "epoch": 24.45, "learning_rate": 7.056338028169014e-07, "loss": 0.2915, "step": 21808 }, { "epoch": 24.45, "learning_rate": 7.042253521126761e-07, "loss": 0.345, "step": 21809 }, { "epoch": 24.45, "learning_rate": 7.028169014084507e-07, "loss": 0.3419, "step": 21810 }, { "epoch": 24.45, "learning_rate": 7.014084507042254e-07, "loss": 0.3134, "step": 21811 }, { "epoch": 24.45, "learning_rate": 7.000000000000001e-07, "loss": 0.3438, "step": 21812 }, { "epoch": 24.45, "learning_rate": 6.985915492957747e-07, "loss": 0.263, "step": 21813 }, { "epoch": 24.46, "learning_rate": 6.971830985915493e-07, "loss": 0.3432, "step": 21814 }, { "epoch": 24.46, "learning_rate": 6.95774647887324e-07, "loss": 0.2655, "step": 21815 }, { "epoch": 24.46, "learning_rate": 6.943661971830986e-07, "loss": 0.309, "step": 21816 }, { "epoch": 24.46, "learning_rate": 6.929577464788733e-07, "loss": 0.3108, "step": 21817 }, { "epoch": 24.46, "learning_rate": 6.915492957746479e-07, "loss": 0.2594, "step": 21818 }, { "epoch": 24.46, "learning_rate": 6.901408450704226e-07, "loss": 0.3283, "step": 21819 }, { "epoch": 24.46, "learning_rate": 6.887323943661973e-07, "loss": 0.3325, "step": 21820 }, { "epoch": 24.46, "learning_rate": 6.873239436619719e-07, "loss": 0.2811, "step": 21821 }, { "epoch": 24.46, "learning_rate": 6.859154929577466e-07, "loss": 0.3378, "step": 21822 }, { "epoch": 24.47, "learning_rate": 6.845070422535212e-07, "loss": 0.3604, "step": 21823 }, { "epoch": 24.47, "learning_rate": 6.830985915492958e-07, "loss": 0.3074, "step": 21824 }, { "epoch": 24.47, "learning_rate": 6.816901408450705e-07, "loss": 0.3943, "step": 21825 }, { "epoch": 24.47, "learning_rate": 6.802816901408451e-07, "loss": 0.2531, "step": 21826 }, { "epoch": 24.47, "learning_rate": 6.788732394366197e-07, "loss": 0.3276, "step": 21827 }, { "epoch": 24.47, "learning_rate": 6.774647887323944e-07, "loss": 0.3754, "step": 21828 }, { "epoch": 24.47, "learning_rate": 6.76056338028169e-07, "loss": 0.3149, "step": 21829 }, { "epoch": 24.47, "learning_rate": 6.746478873239437e-07, "loss": 0.2492, "step": 21830 }, { "epoch": 24.47, "learning_rate": 6.732394366197183e-07, "loss": 0.3706, "step": 21831 }, { "epoch": 24.48, "learning_rate": 6.718309859154929e-07, "loss": 0.2291, "step": 21832 }, { "epoch": 24.48, "learning_rate": 6.704225352112676e-07, "loss": 0.2565, "step": 21833 }, { "epoch": 24.48, "learning_rate": 6.690140845070422e-07, "loss": 0.3174, "step": 21834 }, { "epoch": 24.48, "learning_rate": 6.676056338028168e-07, "loss": 0.2295, "step": 21835 }, { "epoch": 24.48, "learning_rate": 6.661971830985915e-07, "loss": 0.2333, "step": 21836 }, { "epoch": 24.48, "learning_rate": 6.647887323943662e-07, "loss": 0.2884, "step": 21837 }, { "epoch": 24.48, "learning_rate": 6.633802816901409e-07, "loss": 0.4161, "step": 21838 }, { "epoch": 24.48, "learning_rate": 6.619718309859155e-07, "loss": 0.2499, "step": 21839 }, { "epoch": 24.48, "learning_rate": 6.605633802816901e-07, "loss": 0.2527, "step": 21840 }, { "epoch": 24.49, "learning_rate": 6.591549295774648e-07, "loss": 0.2744, "step": 21841 }, { "epoch": 24.49, "learning_rate": 6.577464788732394e-07, "loss": 0.3525, "step": 21842 }, { "epoch": 24.49, "learning_rate": 6.56338028169014e-07, "loss": 0.3481, "step": 21843 }, { "epoch": 24.49, "learning_rate": 6.549295774647887e-07, "loss": 0.3683, "step": 21844 }, { "epoch": 24.49, "learning_rate": 6.535211267605633e-07, "loss": 0.2429, "step": 21845 }, { "epoch": 24.49, "learning_rate": 6.521126760563381e-07, "loss": 0.2988, "step": 21846 }, { "epoch": 24.49, "learning_rate": 6.507042253521127e-07, "loss": 0.2763, "step": 21847 }, { "epoch": 24.49, "learning_rate": 6.492957746478873e-07, "loss": 0.2778, "step": 21848 }, { "epoch": 24.49, "learning_rate": 6.47887323943662e-07, "loss": 0.323, "step": 21849 }, { "epoch": 24.5, "learning_rate": 6.464788732394366e-07, "loss": 0.3243, "step": 21850 }, { "epoch": 24.5, "learning_rate": 6.450704225352113e-07, "loss": 0.2642, "step": 21851 }, { "epoch": 24.5, "learning_rate": 6.436619718309859e-07, "loss": 0.2228, "step": 21852 }, { "epoch": 24.5, "learning_rate": 6.422535211267605e-07, "loss": 0.2709, "step": 21853 }, { "epoch": 24.5, "learning_rate": 6.408450704225353e-07, "loss": 0.2714, "step": 21854 }, { "epoch": 24.5, "learning_rate": 6.394366197183099e-07, "loss": 0.1762, "step": 21855 }, { "epoch": 24.5, "learning_rate": 6.380281690140845e-07, "loss": 0.221, "step": 21856 }, { "epoch": 24.5, "learning_rate": 6.366197183098592e-07, "loss": 0.2064, "step": 21857 }, { "epoch": 24.5, "learning_rate": 6.352112676056338e-07, "loss": 0.3273, "step": 21858 }, { "epoch": 24.51, "learning_rate": 6.338028169014085e-07, "loss": 0.4435, "step": 21859 }, { "epoch": 24.51, "learning_rate": 6.323943661971831e-07, "loss": 0.3281, "step": 21860 }, { "epoch": 24.51, "learning_rate": 6.309859154929577e-07, "loss": 0.2977, "step": 21861 }, { "epoch": 24.51, "learning_rate": 6.295774647887324e-07, "loss": 0.3625, "step": 21862 }, { "epoch": 24.51, "learning_rate": 6.281690140845071e-07, "loss": 0.2666, "step": 21863 }, { "epoch": 24.51, "learning_rate": 6.267605633802818e-07, "loss": 0.3525, "step": 21864 }, { "epoch": 24.51, "learning_rate": 6.253521126760564e-07, "loss": 0.3611, "step": 21865 }, { "epoch": 24.51, "learning_rate": 6.23943661971831e-07, "loss": 0.3434, "step": 21866 }, { "epoch": 24.51, "learning_rate": 6.225352112676057e-07, "loss": 0.3877, "step": 21867 }, { "epoch": 24.52, "learning_rate": 6.211267605633803e-07, "loss": 0.3004, "step": 21868 }, { "epoch": 24.52, "learning_rate": 6.197183098591549e-07, "loss": 0.275, "step": 21869 }, { "epoch": 24.52, "learning_rate": 6.183098591549296e-07, "loss": 0.2911, "step": 21870 }, { "epoch": 24.52, "learning_rate": 6.169014084507042e-07, "loss": 0.3006, "step": 21871 }, { "epoch": 24.52, "learning_rate": 6.15492957746479e-07, "loss": 0.294, "step": 21872 }, { "epoch": 24.52, "learning_rate": 6.140845070422536e-07, "loss": 0.3951, "step": 21873 }, { "epoch": 24.52, "learning_rate": 6.126760563380282e-07, "loss": 0.3032, "step": 21874 }, { "epoch": 24.52, "learning_rate": 6.112676056338029e-07, "loss": 0.3805, "step": 21875 }, { "epoch": 24.52, "learning_rate": 6.098591549295775e-07, "loss": 0.2611, "step": 21876 }, { "epoch": 24.53, "learning_rate": 6.084507042253521e-07, "loss": 0.2319, "step": 21877 }, { "epoch": 24.53, "learning_rate": 6.070422535211268e-07, "loss": 0.3674, "step": 21878 }, { "epoch": 24.53, "learning_rate": 6.056338028169014e-07, "loss": 0.3623, "step": 21879 }, { "epoch": 24.53, "learning_rate": 6.042253521126762e-07, "loss": 0.3574, "step": 21880 }, { "epoch": 24.53, "learning_rate": 6.028169014084508e-07, "loss": 0.2893, "step": 21881 }, { "epoch": 24.53, "learning_rate": 6.014084507042254e-07, "loss": 0.2426, "step": 21882 }, { "epoch": 24.53, "learning_rate": 6.000000000000001e-07, "loss": 0.2727, "step": 21883 }, { "epoch": 24.53, "learning_rate": 5.985915492957747e-07, "loss": 0.2551, "step": 21884 }, { "epoch": 24.53, "learning_rate": 5.971830985915492e-07, "loss": 0.3238, "step": 21885 }, { "epoch": 24.54, "learning_rate": 5.957746478873239e-07, "loss": 0.3365, "step": 21886 }, { "epoch": 24.54, "learning_rate": 5.943661971830985e-07, "loss": 0.3006, "step": 21887 }, { "epoch": 24.54, "learning_rate": 5.929577464788732e-07, "loss": 0.263, "step": 21888 }, { "epoch": 24.54, "learning_rate": 5.915492957746479e-07, "loss": 0.3465, "step": 21889 }, { "epoch": 24.54, "learning_rate": 5.901408450704225e-07, "loss": 0.2921, "step": 21890 }, { "epoch": 24.54, "learning_rate": 5.887323943661972e-07, "loss": 0.2397, "step": 21891 }, { "epoch": 24.54, "learning_rate": 5.873239436619718e-07, "loss": 0.2364, "step": 21892 }, { "epoch": 24.54, "learning_rate": 5.859154929577465e-07, "loss": 0.3577, "step": 21893 }, { "epoch": 24.54, "learning_rate": 5.845070422535211e-07, "loss": 0.3565, "step": 21894 }, { "epoch": 24.55, "learning_rate": 5.830985915492957e-07, "loss": 0.2742, "step": 21895 }, { "epoch": 24.55, "learning_rate": 5.816901408450704e-07, "loss": 0.3052, "step": 21896 }, { "epoch": 24.55, "learning_rate": 5.80281690140845e-07, "loss": 0.1928, "step": 21897 }, { "epoch": 24.55, "learning_rate": 5.788732394366197e-07, "loss": 0.2757, "step": 21898 }, { "epoch": 24.55, "learning_rate": 5.774647887323944e-07, "loss": 0.2444, "step": 21899 }, { "epoch": 24.55, "learning_rate": 5.76056338028169e-07, "loss": 0.3034, "step": 21900 }, { "epoch": 24.55, "learning_rate": 5.746478873239437e-07, "loss": 0.3124, "step": 21901 }, { "epoch": 24.55, "learning_rate": 5.732394366197183e-07, "loss": 0.2854, "step": 21902 }, { "epoch": 24.55, "learning_rate": 5.718309859154929e-07, "loss": 0.3191, "step": 21903 }, { "epoch": 24.56, "learning_rate": 5.704225352112676e-07, "loss": 0.1594, "step": 21904 }, { "epoch": 24.56, "learning_rate": 5.690140845070422e-07, "loss": 0.1913, "step": 21905 }, { "epoch": 24.56, "learning_rate": 5.67605633802817e-07, "loss": 0.2007, "step": 21906 }, { "epoch": 24.56, "learning_rate": 5.661971830985916e-07, "loss": 0.3615, "step": 21907 }, { "epoch": 24.56, "learning_rate": 5.647887323943662e-07, "loss": 0.2322, "step": 21908 }, { "epoch": 24.56, "learning_rate": 5.633802816901409e-07, "loss": 0.5545, "step": 21909 }, { "epoch": 24.56, "learning_rate": 5.619718309859155e-07, "loss": 0.3386, "step": 21910 }, { "epoch": 24.56, "learning_rate": 5.605633802816901e-07, "loss": 0.2575, "step": 21911 }, { "epoch": 24.57, "learning_rate": 5.591549295774648e-07, "loss": 0.3033, "step": 21912 }, { "epoch": 24.57, "learning_rate": 5.577464788732394e-07, "loss": 0.4379, "step": 21913 }, { "epoch": 24.57, "learning_rate": 5.563380281690141e-07, "loss": 0.3229, "step": 21914 }, { "epoch": 24.57, "learning_rate": 5.549295774647888e-07, "loss": 0.2972, "step": 21915 }, { "epoch": 24.57, "learning_rate": 5.535211267605634e-07, "loss": 0.3418, "step": 21916 }, { "epoch": 24.57, "learning_rate": 5.521126760563381e-07, "loss": 0.2841, "step": 21917 }, { "epoch": 24.57, "learning_rate": 5.507042253521127e-07, "loss": 0.2933, "step": 21918 }, { "epoch": 24.57, "learning_rate": 5.492957746478873e-07, "loss": 0.3288, "step": 21919 }, { "epoch": 24.57, "learning_rate": 5.47887323943662e-07, "loss": 0.3002, "step": 21920 }, { "epoch": 24.58, "learning_rate": 5.464788732394366e-07, "loss": 0.2868, "step": 21921 }, { "epoch": 24.58, "learning_rate": 5.450704225352113e-07, "loss": 0.3696, "step": 21922 }, { "epoch": 24.58, "learning_rate": 5.436619718309859e-07, "loss": 0.3096, "step": 21923 }, { "epoch": 24.58, "learning_rate": 5.422535211267606e-07, "loss": 0.2915, "step": 21924 }, { "epoch": 24.58, "learning_rate": 5.408450704225353e-07, "loss": 0.2801, "step": 21925 }, { "epoch": 24.58, "learning_rate": 5.394366197183099e-07, "loss": 0.2544, "step": 21926 }, { "epoch": 24.58, "learning_rate": 5.380281690140846e-07, "loss": 0.2718, "step": 21927 }, { "epoch": 24.58, "learning_rate": 5.366197183098592e-07, "loss": 0.2749, "step": 21928 }, { "epoch": 24.58, "learning_rate": 5.352112676056338e-07, "loss": 0.4107, "step": 21929 }, { "epoch": 24.59, "learning_rate": 5.338028169014085e-07, "loss": 0.203, "step": 21930 }, { "epoch": 24.59, "learning_rate": 5.323943661971831e-07, "loss": 0.2812, "step": 21931 }, { "epoch": 24.59, "learning_rate": 5.309859154929578e-07, "loss": 0.2989, "step": 21932 }, { "epoch": 24.59, "learning_rate": 5.295774647887325e-07, "loss": 0.2207, "step": 21933 }, { "epoch": 24.59, "learning_rate": 5.281690140845071e-07, "loss": 0.3185, "step": 21934 }, { "epoch": 24.59, "learning_rate": 5.267605633802818e-07, "loss": 0.2995, "step": 21935 }, { "epoch": 24.59, "learning_rate": 5.253521126760564e-07, "loss": 0.2318, "step": 21936 }, { "epoch": 24.59, "learning_rate": 5.23943661971831e-07, "loss": 0.3657, "step": 21937 }, { "epoch": 24.59, "learning_rate": 5.225352112676057e-07, "loss": 0.3171, "step": 21938 }, { "epoch": 24.6, "learning_rate": 5.211267605633803e-07, "loss": 0.2319, "step": 21939 }, { "epoch": 24.6, "learning_rate": 5.197183098591549e-07, "loss": 0.2235, "step": 21940 }, { "epoch": 24.6, "learning_rate": 5.183098591549296e-07, "loss": 0.3426, "step": 21941 }, { "epoch": 24.6, "learning_rate": 5.169014084507042e-07, "loss": 0.284, "step": 21942 }, { "epoch": 24.6, "learning_rate": 5.154929577464789e-07, "loss": 0.2623, "step": 21943 }, { "epoch": 24.6, "learning_rate": 5.140845070422535e-07, "loss": 0.2986, "step": 21944 }, { "epoch": 24.6, "learning_rate": 5.126760563380281e-07, "loss": 0.2279, "step": 21945 }, { "epoch": 24.6, "learning_rate": 5.112676056338028e-07, "loss": 0.2802, "step": 21946 }, { "epoch": 24.6, "learning_rate": 5.098591549295774e-07, "loss": 0.2177, "step": 21947 }, { "epoch": 24.61, "learning_rate": 5.08450704225352e-07, "loss": 0.2592, "step": 21948 }, { "epoch": 24.61, "learning_rate": 5.070422535211267e-07, "loss": 0.2502, "step": 21949 }, { "epoch": 24.61, "learning_rate": 5.056338028169014e-07, "loss": 0.3092, "step": 21950 }, { "epoch": 24.61, "learning_rate": 5.042253521126761e-07, "loss": 0.3779, "step": 21951 }, { "epoch": 24.61, "learning_rate": 5.028169014084507e-07, "loss": 0.2156, "step": 21952 }, { "epoch": 24.61, "learning_rate": 5.014084507042253e-07, "loss": 0.2931, "step": 21953 }, { "epoch": 24.61, "learning_rate": 5e-07, "loss": 0.218, "step": 21954 }, { "epoch": 24.61, "learning_rate": 4.985915492957746e-07, "loss": 0.2177, "step": 21955 }, { "epoch": 24.61, "learning_rate": 4.971830985915493e-07, "loss": 0.1903, "step": 21956 }, { "epoch": 24.62, "learning_rate": 4.957746478873239e-07, "loss": 0.2009, "step": 21957 }, { "epoch": 24.62, "learning_rate": 4.943661971830986e-07, "loss": 0.2486, "step": 21958 }, { "epoch": 24.62, "learning_rate": 4.929577464788733e-07, "loss": 0.365, "step": 21959 }, { "epoch": 24.62, "learning_rate": 4.915492957746479e-07, "loss": 0.3849, "step": 21960 }, { "epoch": 24.62, "learning_rate": 4.901408450704225e-07, "loss": 0.4798, "step": 21961 }, { "epoch": 24.62, "learning_rate": 4.887323943661972e-07, "loss": 0.4754, "step": 21962 }, { "epoch": 24.62, "learning_rate": 4.873239436619718e-07, "loss": 0.3049, "step": 21963 }, { "epoch": 24.62, "learning_rate": 4.859154929577465e-07, "loss": 0.3112, "step": 21964 }, { "epoch": 24.62, "learning_rate": 4.845070422535211e-07, "loss": 0.3603, "step": 21965 }, { "epoch": 24.63, "learning_rate": 4.830985915492957e-07, "loss": 0.3376, "step": 21966 }, { "epoch": 24.63, "learning_rate": 4.816901408450705e-07, "loss": 0.401, "step": 21967 }, { "epoch": 24.63, "learning_rate": 4.802816901408451e-07, "loss": 0.2582, "step": 21968 }, { "epoch": 24.63, "learning_rate": 4.788732394366198e-07, "loss": 0.3451, "step": 21969 }, { "epoch": 24.63, "learning_rate": 4.774647887323944e-07, "loss": 0.359, "step": 21970 }, { "epoch": 24.63, "learning_rate": 4.7605633802816903e-07, "loss": 0.2953, "step": 21971 }, { "epoch": 24.63, "learning_rate": 4.7464788732394366e-07, "loss": 0.2863, "step": 21972 }, { "epoch": 24.63, "learning_rate": 4.732394366197183e-07, "loss": 0.3118, "step": 21973 }, { "epoch": 24.63, "learning_rate": 4.7183098591549294e-07, "loss": 0.2385, "step": 21974 }, { "epoch": 24.64, "learning_rate": 4.704225352112676e-07, "loss": 0.296, "step": 21975 }, { "epoch": 24.64, "learning_rate": 4.690140845070423e-07, "loss": 0.3813, "step": 21976 }, { "epoch": 24.64, "learning_rate": 4.676056338028169e-07, "loss": 0.2326, "step": 21977 }, { "epoch": 24.64, "learning_rate": 4.661971830985916e-07, "loss": 0.2701, "step": 21978 }, { "epoch": 24.64, "learning_rate": 4.6478873239436623e-07, "loss": 0.3144, "step": 21979 }, { "epoch": 24.64, "learning_rate": 4.6338028169014087e-07, "loss": 0.2827, "step": 21980 }, { "epoch": 24.64, "learning_rate": 4.619718309859155e-07, "loss": 0.3117, "step": 21981 }, { "epoch": 24.64, "learning_rate": 4.605633802816902e-07, "loss": 0.4194, "step": 21982 }, { "epoch": 24.64, "learning_rate": 4.5915492957746484e-07, "loss": 0.2376, "step": 21983 }, { "epoch": 24.65, "learning_rate": 4.577464788732394e-07, "loss": 0.2832, "step": 21984 }, { "epoch": 24.65, "learning_rate": 4.5633802816901406e-07, "loss": 0.3226, "step": 21985 }, { "epoch": 24.65, "learning_rate": 4.5492957746478875e-07, "loss": 0.2558, "step": 21986 }, { "epoch": 24.65, "learning_rate": 4.535211267605634e-07, "loss": 0.2803, "step": 21987 }, { "epoch": 24.65, "learning_rate": 4.52112676056338e-07, "loss": 0.2754, "step": 21988 }, { "epoch": 24.65, "learning_rate": 4.5070422535211266e-07, "loss": 0.2807, "step": 21989 }, { "epoch": 24.65, "learning_rate": 4.492957746478873e-07, "loss": 0.2912, "step": 21990 }, { "epoch": 24.65, "learning_rate": 4.47887323943662e-07, "loss": 0.2553, "step": 21991 }, { "epoch": 24.65, "learning_rate": 4.464788732394366e-07, "loss": 0.2332, "step": 21992 }, { "epoch": 24.66, "learning_rate": 4.4507042253521126e-07, "loss": 0.275, "step": 21993 }, { "epoch": 24.66, "learning_rate": 4.436619718309859e-07, "loss": 0.3005, "step": 21994 }, { "epoch": 24.66, "learning_rate": 4.422535211267606e-07, "loss": 0.2451, "step": 21995 }, { "epoch": 24.66, "learning_rate": 4.4084507042253523e-07, "loss": 0.3006, "step": 21996 }, { "epoch": 24.66, "learning_rate": 4.3943661971830987e-07, "loss": 0.29, "step": 21997 }, { "epoch": 24.66, "learning_rate": 4.380281690140845e-07, "loss": 0.3148, "step": 21998 }, { "epoch": 24.66, "learning_rate": 4.366197183098592e-07, "loss": 0.2797, "step": 21999 }, { "epoch": 24.66, "learning_rate": 4.3521126760563383e-07, "loss": 0.2632, "step": 22000 }, { "epoch": 24.66, "learning_rate": 4.3380281690140847e-07, "loss": 0.2993, "step": 22001 }, { "epoch": 24.67, "learning_rate": 4.323943661971831e-07, "loss": 0.2447, "step": 22002 }, { "epoch": 24.67, "learning_rate": 4.3098591549295774e-07, "loss": 0.1705, "step": 22003 }, { "epoch": 24.67, "learning_rate": 4.2957746478873244e-07, "loss": 0.2214, "step": 22004 }, { "epoch": 24.67, "learning_rate": 4.2816901408450707e-07, "loss": 0.2113, "step": 22005 }, { "epoch": 24.67, "learning_rate": 4.267605633802817e-07, "loss": 0.215, "step": 22006 }, { "epoch": 24.67, "learning_rate": 4.2535211267605635e-07, "loss": 0.2289, "step": 22007 }, { "epoch": 24.67, "learning_rate": 4.2394366197183104e-07, "loss": 0.177, "step": 22008 }, { "epoch": 24.67, "learning_rate": 4.225352112676057e-07, "loss": 0.4065, "step": 22009 }, { "epoch": 24.67, "learning_rate": 4.211267605633803e-07, "loss": 0.3451, "step": 22010 }, { "epoch": 24.68, "learning_rate": 4.1971830985915495e-07, "loss": 0.2927, "step": 22011 }, { "epoch": 24.68, "learning_rate": 4.183098591549296e-07, "loss": 0.3196, "step": 22012 }, { "epoch": 24.68, "learning_rate": 4.169014084507042e-07, "loss": 0.3457, "step": 22013 }, { "epoch": 24.68, "learning_rate": 4.1549295774647886e-07, "loss": 0.3635, "step": 22014 }, { "epoch": 24.68, "learning_rate": 4.140845070422535e-07, "loss": 0.2974, "step": 22015 }, { "epoch": 24.68, "learning_rate": 4.1267605633802814e-07, "loss": 0.3796, "step": 22016 }, { "epoch": 24.68, "learning_rate": 4.1126760563380283e-07, "loss": 0.3275, "step": 22017 }, { "epoch": 24.68, "learning_rate": 4.0985915492957747e-07, "loss": 0.3396, "step": 22018 }, { "epoch": 24.68, "learning_rate": 4.084507042253521e-07, "loss": 0.4044, "step": 22019 }, { "epoch": 24.69, "learning_rate": 4.0704225352112674e-07, "loss": 0.3137, "step": 22020 }, { "epoch": 24.69, "learning_rate": 4.0563380281690143e-07, "loss": 0.3556, "step": 22021 }, { "epoch": 24.69, "learning_rate": 4.0422535211267607e-07, "loss": 0.3241, "step": 22022 }, { "epoch": 24.69, "learning_rate": 4.028169014084507e-07, "loss": 0.3688, "step": 22023 }, { "epoch": 24.69, "learning_rate": 4.0140845070422534e-07, "loss": 0.261, "step": 22024 }, { "epoch": 24.69, "learning_rate": 4.0000000000000003e-07, "loss": 0.3087, "step": 22025 }, { "epoch": 24.69, "learning_rate": 3.9859154929577467e-07, "loss": 0.3755, "step": 22026 }, { "epoch": 24.69, "learning_rate": 3.971830985915493e-07, "loss": 0.303, "step": 22027 }, { "epoch": 24.7, "learning_rate": 3.9577464788732395e-07, "loss": 0.3466, "step": 22028 }, { "epoch": 24.7, "learning_rate": 3.943661971830986e-07, "loss": 0.2366, "step": 22029 }, { "epoch": 24.7, "learning_rate": 3.929577464788733e-07, "loss": 0.2709, "step": 22030 }, { "epoch": 24.7, "learning_rate": 3.915492957746479e-07, "loss": 0.2518, "step": 22031 }, { "epoch": 24.7, "learning_rate": 3.9014084507042255e-07, "loss": 0.2961, "step": 22032 }, { "epoch": 24.7, "learning_rate": 3.887323943661972e-07, "loss": 0.3934, "step": 22033 }, { "epoch": 24.7, "learning_rate": 3.873239436619719e-07, "loss": 0.2535, "step": 22034 }, { "epoch": 24.7, "learning_rate": 3.859154929577465e-07, "loss": 0.293, "step": 22035 }, { "epoch": 24.7, "learning_rate": 3.8450704225352115e-07, "loss": 0.307, "step": 22036 }, { "epoch": 24.71, "learning_rate": 3.830985915492958e-07, "loss": 0.354, "step": 22037 }, { "epoch": 24.71, "learning_rate": 3.816901408450705e-07, "loss": 0.3118, "step": 22038 }, { "epoch": 24.71, "learning_rate": 3.802816901408451e-07, "loss": 0.3104, "step": 22039 }, { "epoch": 24.71, "learning_rate": 3.788732394366197e-07, "loss": 0.3893, "step": 22040 }, { "epoch": 24.71, "learning_rate": 3.7746478873239434e-07, "loss": 0.3002, "step": 22041 }, { "epoch": 24.71, "learning_rate": 3.76056338028169e-07, "loss": 0.3389, "step": 22042 }, { "epoch": 24.71, "learning_rate": 3.7464788732394367e-07, "loss": 0.2814, "step": 22043 }, { "epoch": 24.71, "learning_rate": 3.732394366197183e-07, "loss": 0.253, "step": 22044 }, { "epoch": 24.71, "learning_rate": 3.7183098591549294e-07, "loss": 0.3137, "step": 22045 }, { "epoch": 24.72, "learning_rate": 3.704225352112676e-07, "loss": 0.2101, "step": 22046 }, { "epoch": 24.72, "learning_rate": 3.6901408450704227e-07, "loss": 0.2417, "step": 22047 }, { "epoch": 24.72, "learning_rate": 3.676056338028169e-07, "loss": 0.2397, "step": 22048 }, { "epoch": 24.72, "learning_rate": 3.6619718309859155e-07, "loss": 0.3107, "step": 22049 }, { "epoch": 24.72, "learning_rate": 3.647887323943662e-07, "loss": 0.3028, "step": 22050 }, { "epoch": 24.72, "learning_rate": 3.633802816901409e-07, "loss": 0.2546, "step": 22051 }, { "epoch": 24.72, "learning_rate": 3.619718309859155e-07, "loss": 0.2867, "step": 22052 }, { "epoch": 24.72, "learning_rate": 3.6056338028169015e-07, "loss": 0.3109, "step": 22053 }, { "epoch": 24.72, "learning_rate": 3.591549295774648e-07, "loss": 0.4065, "step": 22054 }, { "epoch": 24.73, "learning_rate": 3.577464788732394e-07, "loss": 0.1582, "step": 22055 }, { "epoch": 24.73, "learning_rate": 3.563380281690141e-07, "loss": 0.2743, "step": 22056 }, { "epoch": 24.73, "learning_rate": 3.5492957746478875e-07, "loss": 0.1937, "step": 22057 }, { "epoch": 24.73, "learning_rate": 3.535211267605634e-07, "loss": 0.2813, "step": 22058 }, { "epoch": 24.73, "learning_rate": 3.5211267605633803e-07, "loss": 0.4653, "step": 22059 }, { "epoch": 24.73, "learning_rate": 3.507042253521127e-07, "loss": 0.3578, "step": 22060 }, { "epoch": 24.73, "learning_rate": 3.4929577464788736e-07, "loss": 0.2958, "step": 22061 }, { "epoch": 24.73, "learning_rate": 3.47887323943662e-07, "loss": 0.3936, "step": 22062 }, { "epoch": 24.73, "learning_rate": 3.4647887323943663e-07, "loss": 0.2604, "step": 22063 }, { "epoch": 24.74, "learning_rate": 3.450704225352113e-07, "loss": 0.308, "step": 22064 }, { "epoch": 24.74, "learning_rate": 3.4366197183098596e-07, "loss": 0.3069, "step": 22065 }, { "epoch": 24.74, "learning_rate": 3.422535211267606e-07, "loss": 0.3533, "step": 22066 }, { "epoch": 24.74, "learning_rate": 3.4084507042253523e-07, "loss": 0.279, "step": 22067 }, { "epoch": 24.74, "learning_rate": 3.3943661971830987e-07, "loss": 0.2729, "step": 22068 }, { "epoch": 24.74, "learning_rate": 3.380281690140845e-07, "loss": 0.2977, "step": 22069 }, { "epoch": 24.74, "learning_rate": 3.3661971830985915e-07, "loss": 0.3212, "step": 22070 }, { "epoch": 24.74, "learning_rate": 3.352112676056338e-07, "loss": 0.2926, "step": 22071 }, { "epoch": 24.74, "learning_rate": 3.338028169014084e-07, "loss": 0.3011, "step": 22072 }, { "epoch": 24.75, "learning_rate": 3.323943661971831e-07, "loss": 0.3039, "step": 22073 }, { "epoch": 24.75, "learning_rate": 3.3098591549295775e-07, "loss": 0.2774, "step": 22074 }, { "epoch": 24.75, "learning_rate": 3.295774647887324e-07, "loss": 0.252, "step": 22075 }, { "epoch": 24.75, "learning_rate": 3.28169014084507e-07, "loss": 0.2872, "step": 22076 }, { "epoch": 24.75, "learning_rate": 3.2676056338028166e-07, "loss": 0.3033, "step": 22077 }, { "epoch": 24.75, "learning_rate": 3.2535211267605635e-07, "loss": 0.3768, "step": 22078 }, { "epoch": 24.75, "learning_rate": 3.23943661971831e-07, "loss": 0.3583, "step": 22079 }, { "epoch": 24.75, "learning_rate": 3.2253521126760563e-07, "loss": 0.2689, "step": 22080 }, { "epoch": 24.75, "learning_rate": 3.2112676056338026e-07, "loss": 0.2058, "step": 22081 }, { "epoch": 24.76, "learning_rate": 3.1971830985915496e-07, "loss": 0.3458, "step": 22082 }, { "epoch": 24.76, "learning_rate": 3.183098591549296e-07, "loss": 0.2974, "step": 22083 }, { "epoch": 24.76, "learning_rate": 3.1690140845070423e-07, "loss": 0.3409, "step": 22084 }, { "epoch": 24.76, "learning_rate": 3.1549295774647887e-07, "loss": 0.2595, "step": 22085 }, { "epoch": 24.76, "learning_rate": 3.1408450704225356e-07, "loss": 0.3146, "step": 22086 }, { "epoch": 24.76, "learning_rate": 3.126760563380282e-07, "loss": 0.3138, "step": 22087 }, { "epoch": 24.76, "learning_rate": 3.1126760563380283e-07, "loss": 0.3408, "step": 22088 }, { "epoch": 24.76, "learning_rate": 3.0985915492957747e-07, "loss": 0.2916, "step": 22089 }, { "epoch": 24.76, "learning_rate": 3.084507042253521e-07, "loss": 0.3263, "step": 22090 }, { "epoch": 24.77, "learning_rate": 3.070422535211268e-07, "loss": 0.3179, "step": 22091 }, { "epoch": 24.77, "learning_rate": 3.0563380281690144e-07, "loss": 0.2789, "step": 22092 }, { "epoch": 24.77, "learning_rate": 3.042253521126761e-07, "loss": 0.2507, "step": 22093 }, { "epoch": 24.77, "learning_rate": 3.028169014084507e-07, "loss": 0.2418, "step": 22094 }, { "epoch": 24.77, "learning_rate": 3.014084507042254e-07, "loss": 0.1909, "step": 22095 }, { "epoch": 24.77, "learning_rate": 3.0000000000000004e-07, "loss": 0.1809, "step": 22096 }, { "epoch": 24.77, "learning_rate": 2.985915492957746e-07, "loss": 0.2276, "step": 22097 }, { "epoch": 24.77, "learning_rate": 2.9718309859154926e-07, "loss": 0.293, "step": 22098 }, { "epoch": 24.77, "learning_rate": 2.9577464788732395e-07, "loss": 0.2158, "step": 22099 }, { "epoch": 24.78, "learning_rate": 2.943661971830986e-07, "loss": 0.2482, "step": 22100 }, { "epoch": 24.78, "learning_rate": 2.9295774647887323e-07, "loss": 0.2268, "step": 22101 }, { "epoch": 24.78, "learning_rate": 2.9154929577464786e-07, "loss": 0.2395, "step": 22102 }, { "epoch": 24.78, "learning_rate": 2.901408450704225e-07, "loss": 0.2061, "step": 22103 }, { "epoch": 24.78, "learning_rate": 2.887323943661972e-07, "loss": 0.2739, "step": 22104 }, { "epoch": 24.78, "learning_rate": 2.8732394366197183e-07, "loss": 0.2299, "step": 22105 }, { "epoch": 24.78, "learning_rate": 2.8591549295774647e-07, "loss": 0.2527, "step": 22106 }, { "epoch": 24.78, "learning_rate": 2.845070422535211e-07, "loss": 0.2627, "step": 22107 }, { "epoch": 24.78, "learning_rate": 2.830985915492958e-07, "loss": 0.2731, "step": 22108 }, { "epoch": 24.79, "learning_rate": 2.8169014084507043e-07, "loss": 0.4323, "step": 22109 }, { "epoch": 24.79, "learning_rate": 2.8028169014084507e-07, "loss": 0.4062, "step": 22110 }, { "epoch": 24.79, "learning_rate": 2.788732394366197e-07, "loss": 0.34, "step": 22111 }, { "epoch": 24.79, "learning_rate": 2.774647887323944e-07, "loss": 0.3588, "step": 22112 }, { "epoch": 24.79, "learning_rate": 2.7605633802816904e-07, "loss": 0.2778, "step": 22113 }, { "epoch": 24.79, "learning_rate": 2.7464788732394367e-07, "loss": 0.2927, "step": 22114 }, { "epoch": 24.79, "learning_rate": 2.732394366197183e-07, "loss": 0.2759, "step": 22115 }, { "epoch": 24.79, "learning_rate": 2.7183098591549295e-07, "loss": 0.3041, "step": 22116 }, { "epoch": 24.79, "learning_rate": 2.7042253521126764e-07, "loss": 0.3363, "step": 22117 }, { "epoch": 24.8, "learning_rate": 2.690140845070423e-07, "loss": 0.3254, "step": 22118 }, { "epoch": 24.8, "learning_rate": 2.676056338028169e-07, "loss": 0.2167, "step": 22119 }, { "epoch": 24.8, "learning_rate": 2.6619718309859155e-07, "loss": 0.3843, "step": 22120 }, { "epoch": 24.8, "learning_rate": 2.6478873239436624e-07, "loss": 0.2509, "step": 22121 }, { "epoch": 24.8, "learning_rate": 2.633802816901409e-07, "loss": 0.3479, "step": 22122 }, { "epoch": 24.8, "learning_rate": 2.619718309859155e-07, "loss": 0.3339, "step": 22123 }, { "epoch": 24.8, "learning_rate": 2.6056338028169015e-07, "loss": 0.307, "step": 22124 }, { "epoch": 24.8, "learning_rate": 2.591549295774648e-07, "loss": 0.294, "step": 22125 }, { "epoch": 24.8, "learning_rate": 2.5774647887323943e-07, "loss": 0.2863, "step": 22126 }, { "epoch": 24.81, "learning_rate": 2.5633802816901407e-07, "loss": 0.2124, "step": 22127 }, { "epoch": 24.81, "learning_rate": 2.549295774647887e-07, "loss": 0.2996, "step": 22128 }, { "epoch": 24.81, "learning_rate": 2.5352112676056334e-07, "loss": 0.3098, "step": 22129 }, { "epoch": 24.81, "learning_rate": 2.5211267605633803e-07, "loss": 0.3561, "step": 22130 }, { "epoch": 24.81, "learning_rate": 2.5070422535211267e-07, "loss": 0.2714, "step": 22131 }, { "epoch": 24.81, "learning_rate": 2.492957746478873e-07, "loss": 0.2978, "step": 22132 }, { "epoch": 24.81, "learning_rate": 2.4788732394366194e-07, "loss": 0.2734, "step": 22133 }, { "epoch": 24.81, "learning_rate": 2.4647887323943664e-07, "loss": 0.246, "step": 22134 }, { "epoch": 24.82, "learning_rate": 2.4507042253521127e-07, "loss": 0.2878, "step": 22135 }, { "epoch": 24.82, "learning_rate": 2.436619718309859e-07, "loss": 0.2274, "step": 22136 }, { "epoch": 24.82, "learning_rate": 2.4225352112676055e-07, "loss": 0.3079, "step": 22137 }, { "epoch": 24.82, "learning_rate": 2.4084507042253524e-07, "loss": 0.2719, "step": 22138 }, { "epoch": 24.82, "learning_rate": 2.394366197183099e-07, "loss": 0.2704, "step": 22139 }, { "epoch": 24.82, "learning_rate": 2.3802816901408451e-07, "loss": 0.2941, "step": 22140 }, { "epoch": 24.82, "learning_rate": 2.3661971830985915e-07, "loss": 0.2417, "step": 22141 }, { "epoch": 24.82, "learning_rate": 2.352112676056338e-07, "loss": 0.2243, "step": 22142 }, { "epoch": 24.82, "learning_rate": 2.3380281690140845e-07, "loss": 0.2974, "step": 22143 }, { "epoch": 24.83, "learning_rate": 2.3239436619718312e-07, "loss": 0.2442, "step": 22144 }, { "epoch": 24.83, "learning_rate": 2.3098591549295775e-07, "loss": 0.3461, "step": 22145 }, { "epoch": 24.83, "learning_rate": 2.2957746478873242e-07, "loss": 0.28, "step": 22146 }, { "epoch": 24.83, "learning_rate": 2.2816901408450703e-07, "loss": 0.2146, "step": 22147 }, { "epoch": 24.83, "learning_rate": 2.267605633802817e-07, "loss": 0.2913, "step": 22148 }, { "epoch": 24.83, "learning_rate": 2.2535211267605633e-07, "loss": 0.337, "step": 22149 }, { "epoch": 24.83, "learning_rate": 2.23943661971831e-07, "loss": 0.2774, "step": 22150 }, { "epoch": 24.83, "learning_rate": 2.2253521126760563e-07, "loss": 0.2602, "step": 22151 }, { "epoch": 24.83, "learning_rate": 2.211267605633803e-07, "loss": 0.3632, "step": 22152 }, { "epoch": 24.84, "learning_rate": 2.1971830985915493e-07, "loss": 0.3306, "step": 22153 }, { "epoch": 24.84, "learning_rate": 2.183098591549296e-07, "loss": 0.2841, "step": 22154 }, { "epoch": 24.84, "learning_rate": 2.1690140845070423e-07, "loss": 0.2015, "step": 22155 }, { "epoch": 24.84, "learning_rate": 2.1549295774647887e-07, "loss": 0.2534, "step": 22156 }, { "epoch": 24.84, "learning_rate": 2.1408450704225354e-07, "loss": 0.2165, "step": 22157 }, { "epoch": 24.84, "learning_rate": 2.1267605633802817e-07, "loss": 0.2978, "step": 22158 }, { "epoch": 24.84, "learning_rate": 2.1126760563380284e-07, "loss": 0.4853, "step": 22159 }, { "epoch": 24.84, "learning_rate": 2.0985915492957748e-07, "loss": 0.3799, "step": 22160 }, { "epoch": 24.84, "learning_rate": 2.084507042253521e-07, "loss": 0.4126, "step": 22161 }, { "epoch": 24.85, "learning_rate": 2.0704225352112675e-07, "loss": 0.3464, "step": 22162 }, { "epoch": 24.85, "learning_rate": 2.0563380281690141e-07, "loss": 0.3628, "step": 22163 }, { "epoch": 24.85, "learning_rate": 2.0422535211267605e-07, "loss": 0.3012, "step": 22164 }, { "epoch": 24.85, "learning_rate": 2.0281690140845072e-07, "loss": 0.3628, "step": 22165 }, { "epoch": 24.85, "learning_rate": 2.0140845070422535e-07, "loss": 0.3243, "step": 22166 }, { "epoch": 24.85, "learning_rate": 2.0000000000000002e-07, "loss": 0.3082, "step": 22167 }, { "epoch": 24.85, "learning_rate": 1.9859154929577465e-07, "loss": 0.3976, "step": 22168 }, { "epoch": 24.85, "learning_rate": 1.971830985915493e-07, "loss": 0.3558, "step": 22169 }, { "epoch": 24.85, "learning_rate": 1.9577464788732396e-07, "loss": 0.3481, "step": 22170 }, { "epoch": 24.86, "learning_rate": 1.943661971830986e-07, "loss": 0.3111, "step": 22171 }, { "epoch": 24.86, "learning_rate": 1.9295774647887326e-07, "loss": 0.3114, "step": 22172 }, { "epoch": 24.86, "learning_rate": 1.915492957746479e-07, "loss": 0.2752, "step": 22173 }, { "epoch": 24.86, "learning_rate": 1.9014084507042256e-07, "loss": 0.28, "step": 22174 }, { "epoch": 24.86, "learning_rate": 1.8873239436619717e-07, "loss": 0.4187, "step": 22175 }, { "epoch": 24.86, "learning_rate": 1.8732394366197183e-07, "loss": 0.3077, "step": 22176 }, { "epoch": 24.86, "learning_rate": 1.8591549295774647e-07, "loss": 0.3681, "step": 22177 }, { "epoch": 24.86, "learning_rate": 1.8450704225352114e-07, "loss": 0.2848, "step": 22178 }, { "epoch": 24.86, "learning_rate": 1.8309859154929577e-07, "loss": 0.2962, "step": 22179 }, { "epoch": 24.87, "learning_rate": 1.8169014084507044e-07, "loss": 0.3217, "step": 22180 }, { "epoch": 24.87, "learning_rate": 1.8028169014084507e-07, "loss": 0.3746, "step": 22181 }, { "epoch": 24.87, "learning_rate": 1.788732394366197e-07, "loss": 0.2901, "step": 22182 }, { "epoch": 24.87, "learning_rate": 1.7746478873239438e-07, "loss": 0.2963, "step": 22183 }, { "epoch": 24.87, "learning_rate": 1.7605633802816901e-07, "loss": 0.2652, "step": 22184 }, { "epoch": 24.87, "learning_rate": 1.7464788732394368e-07, "loss": 0.3115, "step": 22185 }, { "epoch": 24.87, "learning_rate": 1.7323943661971832e-07, "loss": 0.2431, "step": 22186 }, { "epoch": 24.87, "learning_rate": 1.7183098591549298e-07, "loss": 0.2735, "step": 22187 }, { "epoch": 24.87, "learning_rate": 1.7042253521126762e-07, "loss": 0.265, "step": 22188 }, { "epoch": 24.88, "learning_rate": 1.6901408450704225e-07, "loss": 0.3265, "step": 22189 }, { "epoch": 24.88, "learning_rate": 1.676056338028169e-07, "loss": 0.2536, "step": 22190 }, { "epoch": 24.88, "learning_rate": 1.6619718309859156e-07, "loss": 0.2959, "step": 22191 }, { "epoch": 24.88, "learning_rate": 1.647887323943662e-07, "loss": 0.2631, "step": 22192 }, { "epoch": 24.88, "learning_rate": 1.6338028169014083e-07, "loss": 0.4325, "step": 22193 }, { "epoch": 24.88, "learning_rate": 1.619718309859155e-07, "loss": 0.2962, "step": 22194 }, { "epoch": 24.88, "learning_rate": 1.6056338028169013e-07, "loss": 0.2439, "step": 22195 }, { "epoch": 24.88, "learning_rate": 1.591549295774648e-07, "loss": 0.2181, "step": 22196 }, { "epoch": 24.88, "learning_rate": 1.5774647887323943e-07, "loss": 0.3958, "step": 22197 }, { "epoch": 24.89, "learning_rate": 1.563380281690141e-07, "loss": 0.3127, "step": 22198 }, { "epoch": 24.89, "learning_rate": 1.5492957746478874e-07, "loss": 0.2288, "step": 22199 }, { "epoch": 24.89, "learning_rate": 1.535211267605634e-07, "loss": 0.263, "step": 22200 }, { "epoch": 24.89, "learning_rate": 1.5211267605633804e-07, "loss": 0.2387, "step": 22201 }, { "epoch": 24.89, "learning_rate": 1.507042253521127e-07, "loss": 0.2589, "step": 22202 }, { "epoch": 24.89, "learning_rate": 1.492957746478873e-07, "loss": 0.2568, "step": 22203 }, { "epoch": 24.89, "learning_rate": 1.4788732394366198e-07, "loss": 0.2332, "step": 22204 }, { "epoch": 24.89, "learning_rate": 1.4647887323943661e-07, "loss": 0.2369, "step": 22205 }, { "epoch": 24.89, "learning_rate": 1.4507042253521125e-07, "loss": 0.3012, "step": 22206 }, { "epoch": 24.9, "learning_rate": 1.4366197183098591e-07, "loss": 0.3672, "step": 22207 }, { "epoch": 24.9, "learning_rate": 1.4225352112676055e-07, "loss": 0.3932, "step": 22208 }, { "epoch": 24.9, "learning_rate": 1.4084507042253522e-07, "loss": 0.3877, "step": 22209 }, { "epoch": 24.9, "learning_rate": 1.3943661971830985e-07, "loss": 0.4116, "step": 22210 }, { "epoch": 24.9, "learning_rate": 1.3802816901408452e-07, "loss": 0.3788, "step": 22211 }, { "epoch": 24.9, "learning_rate": 1.3661971830985916e-07, "loss": 0.3427, "step": 22212 }, { "epoch": 24.9, "learning_rate": 1.3521126760563382e-07, "loss": 0.4495, "step": 22213 }, { "epoch": 24.9, "learning_rate": 1.3380281690140846e-07, "loss": 0.3398, "step": 22214 }, { "epoch": 24.9, "learning_rate": 1.3239436619718312e-07, "loss": 0.4127, "step": 22215 }, { "epoch": 24.91, "learning_rate": 1.3098591549295776e-07, "loss": 0.4057, "step": 22216 }, { "epoch": 24.91, "learning_rate": 1.295774647887324e-07, "loss": 0.3648, "step": 22217 }, { "epoch": 24.91, "learning_rate": 1.2816901408450703e-07, "loss": 0.3272, "step": 22218 }, { "epoch": 24.91, "learning_rate": 1.2676056338028167e-07, "loss": 0.3418, "step": 22219 }, { "epoch": 24.91, "learning_rate": 1.2535211267605633e-07, "loss": 0.3243, "step": 22220 }, { "epoch": 24.91, "learning_rate": 1.2394366197183097e-07, "loss": 0.332, "step": 22221 }, { "epoch": 24.91, "learning_rate": 1.2253521126760564e-07, "loss": 0.3402, "step": 22222 }, { "epoch": 24.91, "learning_rate": 1.2112676056338027e-07, "loss": 0.2563, "step": 22223 }, { "epoch": 24.91, "learning_rate": 1.1971830985915494e-07, "loss": 0.3078, "step": 22224 }, { "epoch": 24.92, "learning_rate": 1.1830985915492958e-07, "loss": 0.3101, "step": 22225 }, { "epoch": 24.92, "learning_rate": 1.1690140845070423e-07, "loss": 0.2774, "step": 22226 }, { "epoch": 24.92, "learning_rate": 1.1549295774647888e-07, "loss": 0.2896, "step": 22227 }, { "epoch": 24.92, "learning_rate": 1.1408450704225351e-07, "loss": 0.2998, "step": 22228 }, { "epoch": 24.92, "learning_rate": 1.1267605633802817e-07, "loss": 0.3647, "step": 22229 }, { "epoch": 24.92, "learning_rate": 1.1126760563380282e-07, "loss": 0.2648, "step": 22230 }, { "epoch": 24.92, "learning_rate": 1.0985915492957747e-07, "loss": 0.2877, "step": 22231 }, { "epoch": 24.92, "learning_rate": 1.0845070422535212e-07, "loss": 0.2248, "step": 22232 }, { "epoch": 24.92, "learning_rate": 1.0704225352112677e-07, "loss": 0.2906, "step": 22233 }, { "epoch": 24.93, "learning_rate": 1.0563380281690142e-07, "loss": 0.3421, "step": 22234 }, { "epoch": 24.93, "learning_rate": 1.0422535211267606e-07, "loss": 0.2468, "step": 22235 }, { "epoch": 24.93, "learning_rate": 1.0281690140845071e-07, "loss": 0.2905, "step": 22236 }, { "epoch": 24.93, "learning_rate": 1.0140845070422536e-07, "loss": 0.26, "step": 22237 }, { "epoch": 24.93, "learning_rate": 1.0000000000000001e-07, "loss": 0.348, "step": 22238 }, { "epoch": 24.93, "learning_rate": 9.859154929577465e-08, "loss": 0.2747, "step": 22239 }, { "epoch": 24.93, "learning_rate": 9.71830985915493e-08, "loss": 0.3105, "step": 22240 }, { "epoch": 24.93, "learning_rate": 9.577464788732395e-08, "loss": 0.237, "step": 22241 }, { "epoch": 24.93, "learning_rate": 9.436619718309859e-08, "loss": 0.2779, "step": 22242 }, { "epoch": 24.94, "learning_rate": 9.295774647887324e-08, "loss": 0.2678, "step": 22243 }, { "epoch": 24.94, "learning_rate": 9.154929577464789e-08, "loss": 0.326, "step": 22244 }, { "epoch": 24.94, "learning_rate": 9.014084507042254e-08, "loss": 0.2752, "step": 22245 }, { "epoch": 24.94, "learning_rate": 8.873239436619719e-08, "loss": 0.2373, "step": 22246 }, { "epoch": 24.94, "learning_rate": 8.732394366197184e-08, "loss": 0.3072, "step": 22247 }, { "epoch": 24.94, "learning_rate": 8.591549295774649e-08, "loss": 0.2902, "step": 22248 }, { "epoch": 24.94, "learning_rate": 8.450704225352113e-08, "loss": 0.4064, "step": 22249 }, { "epoch": 24.94, "learning_rate": 8.309859154929578e-08, "loss": 0.2926, "step": 22250 }, { "epoch": 24.95, "learning_rate": 8.169014084507042e-08, "loss": 0.2373, "step": 22251 }, { "epoch": 24.95, "learning_rate": 8.028169014084507e-08, "loss": 0.191, "step": 22252 }, { "epoch": 24.95, "learning_rate": 7.887323943661972e-08, "loss": 0.2741, "step": 22253 }, { "epoch": 24.95, "learning_rate": 7.746478873239437e-08, "loss": 0.2047, "step": 22254 }, { "epoch": 24.95, "learning_rate": 7.605633802816902e-08, "loss": 0.2737, "step": 22255 }, { "epoch": 24.95, "learning_rate": 7.464788732394366e-08, "loss": 0.1556, "step": 22256 }, { "epoch": 24.95, "learning_rate": 7.323943661971831e-08, "loss": 0.2619, "step": 22257 }, { "epoch": 24.95, "learning_rate": 7.183098591549296e-08, "loss": 0.2269, "step": 22258 }, { "epoch": 24.95, "learning_rate": 7.042253521126761e-08, "loss": 0.4635, "step": 22259 }, { "epoch": 24.96, "learning_rate": 6.901408450704226e-08, "loss": 0.4206, "step": 22260 }, { "epoch": 24.96, "learning_rate": 6.760563380281691e-08, "loss": 0.321, "step": 22261 }, { "epoch": 24.96, "learning_rate": 6.619718309859156e-08, "loss": 0.3461, "step": 22262 }, { "epoch": 24.96, "learning_rate": 6.47887323943662e-08, "loss": 0.2702, "step": 22263 }, { "epoch": 24.96, "learning_rate": 6.338028169014084e-08, "loss": 0.2709, "step": 22264 }, { "epoch": 24.96, "learning_rate": 6.197183098591549e-08, "loss": 0.3249, "step": 22265 }, { "epoch": 24.96, "learning_rate": 6.056338028169014e-08, "loss": 0.3322, "step": 22266 }, { "epoch": 24.96, "learning_rate": 5.915492957746479e-08, "loss": 0.3062, "step": 22267 }, { "epoch": 24.96, "learning_rate": 5.774647887323944e-08, "loss": 0.3492, "step": 22268 }, { "epoch": 24.97, "learning_rate": 5.633802816901408e-08, "loss": 0.3145, "step": 22269 }, { "epoch": 24.97, "learning_rate": 5.4929577464788733e-08, "loss": 0.255, "step": 22270 }, { "epoch": 24.97, "learning_rate": 5.3521126760563384e-08, "loss": 0.285, "step": 22271 }, { "epoch": 24.97, "learning_rate": 5.211267605633803e-08, "loss": 0.2613, "step": 22272 }, { "epoch": 24.97, "learning_rate": 5.070422535211268e-08, "loss": 0.3005, "step": 22273 }, { "epoch": 24.97, "learning_rate": 4.929577464788732e-08, "loss": 0.3024, "step": 22274 }, { "epoch": 24.97, "learning_rate": 4.7887323943661974e-08, "loss": 0.3238, "step": 22275 }, { "epoch": 24.97, "learning_rate": 4.647887323943662e-08, "loss": 0.345, "step": 22276 }, { "epoch": 24.97, "learning_rate": 4.507042253521127e-08, "loss": 0.2999, "step": 22277 }, { "epoch": 24.98, "learning_rate": 4.366197183098592e-08, "loss": 0.3515, "step": 22278 }, { "epoch": 24.98, "learning_rate": 4.2253521126760564e-08, "loss": 0.2685, "step": 22279 }, { "epoch": 24.98, "learning_rate": 4.084507042253521e-08, "loss": 0.2889, "step": 22280 }, { "epoch": 24.98, "learning_rate": 3.943661971830986e-08, "loss": 0.2118, "step": 22281 }, { "epoch": 24.98, "learning_rate": 3.802816901408451e-08, "loss": 0.2415, "step": 22282 }, { "epoch": 24.98, "learning_rate": 3.6619718309859153e-08, "loss": 0.2951, "step": 22283 }, { "epoch": 24.98, "learning_rate": 3.5211267605633804e-08, "loss": 0.3319, "step": 22284 }, { "epoch": 24.98, "learning_rate": 3.3802816901408455e-08, "loss": 0.3006, "step": 22285 }, { "epoch": 24.98, "learning_rate": 3.23943661971831e-08, "loss": 0.3297, "step": 22286 }, { "epoch": 24.99, "learning_rate": 3.098591549295774e-08, "loss": 0.2937, "step": 22287 }, { "epoch": 24.99, "learning_rate": 2.9577464788732394e-08, "loss": 0.2583, "step": 22288 }, { "epoch": 24.99, "learning_rate": 2.816901408450704e-08, "loss": 0.3569, "step": 22289 }, { "epoch": 24.99, "learning_rate": 2.6760563380281692e-08, "loss": 0.3105, "step": 22290 }, { "epoch": 24.99, "learning_rate": 2.535211267605634e-08, "loss": 0.2591, "step": 22291 }, { "epoch": 24.99, "learning_rate": 2.3943661971830987e-08, "loss": 0.3047, "step": 22292 }, { "epoch": 24.99, "learning_rate": 2.2535211267605634e-08, "loss": 0.1951, "step": 22293 }, { "epoch": 24.99, "learning_rate": 2.1126760563380282e-08, "loss": 0.2899, "step": 22294 }, { "epoch": 24.99, "learning_rate": 1.971830985915493e-08, "loss": 0.3373, "step": 22295 }, { "epoch": 25.0, "learning_rate": 1.8309859154929577e-08, "loss": 0.2235, "step": 22296 }, { "epoch": 25.0, "learning_rate": 1.6901408450704227e-08, "loss": 0.2079, "step": 22297 }, { "epoch": 25.0, "learning_rate": 1.549295774647887e-08, "loss": 0.1841, "step": 22298 }, { "epoch": 25.0, "learning_rate": 1.408450704225352e-08, "loss": 0.2671, "step": 22299 }, { "epoch": 25.0, "learning_rate": 1.267605633802817e-08, "loss": 0.4226, "step": 22300 }, { "epoch": 25.0, "step": 22300, "total_flos": 0.0, "train_loss": 0.08645637407097047, "train_runtime": 35320.0928, "train_samples_per_second": 20.2, "train_steps_per_second": 0.631 } ], "max_steps": 22300, "num_train_epochs": 25, "total_flos": 0.0, "trial_name": null, "trial_params": null }