{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "global_step": 8331, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 2e-08, "loss": 3.6164, "step": 1 }, { "epoch": 0.0, "learning_rate": 4e-08, "loss": 3.4702, "step": 2 }, { "epoch": 0.0, "learning_rate": 6.000000000000001e-08, "loss": 3.4604, "step": 3 }, { "epoch": 0.0, "learning_rate": 8e-08, "loss": 3.5767, "step": 4 }, { "epoch": 0.0, "learning_rate": 1.0000000000000001e-07, "loss": 3.5647, "step": 5 }, { "epoch": 0.0, "learning_rate": 1.2000000000000002e-07, "loss": 3.5424, "step": 6 }, { "epoch": 0.0, "learning_rate": 1.4e-07, "loss": 3.6716, "step": 7 }, { "epoch": 0.0, "learning_rate": 1.6e-07, "loss": 3.5821, "step": 8 }, { "epoch": 0.0, "learning_rate": 1.8e-07, "loss": 3.5822, "step": 9 }, { "epoch": 0.0, "learning_rate": 2.0000000000000002e-07, "loss": 3.6172, "step": 10 }, { "epoch": 0.0, "learning_rate": 2.2e-07, "loss": 3.5909, "step": 11 }, { "epoch": 0.0, "learning_rate": 2.4000000000000003e-07, "loss": 3.5564, "step": 12 }, { "epoch": 0.0, "learning_rate": 2.6e-07, "loss": 3.6067, "step": 13 }, { "epoch": 0.0, "learning_rate": 2.8e-07, "loss": 3.4872, "step": 14 }, { "epoch": 0.0, "learning_rate": 3.0000000000000004e-07, "loss": 3.4353, "step": 15 }, { "epoch": 0.0, "learning_rate": 3.2e-07, "loss": 3.4828, "step": 16 }, { "epoch": 0.0, "learning_rate": 3.4000000000000003e-07, "loss": 3.5568, "step": 17 }, { "epoch": 0.0, "learning_rate": 3.6e-07, "loss": 3.6367, "step": 18 }, { "epoch": 0.0, "learning_rate": 3.8e-07, "loss": 3.5097, "step": 19 }, { "epoch": 0.0, "learning_rate": 4.0000000000000003e-07, "loss": 3.5974, "step": 20 }, { "epoch": 0.0, "learning_rate": 4.2000000000000006e-07, "loss": 3.5344, "step": 21 }, { "epoch": 0.0, "learning_rate": 4.4e-07, "loss": 3.5648, "step": 22 }, { "epoch": 0.0, "learning_rate": 4.6000000000000004e-07, "loss": 3.5091, "step": 23 }, { "epoch": 0.0, "learning_rate": 4.800000000000001e-07, "loss": 3.6095, "step": 24 }, { "epoch": 0.0, "learning_rate": 5.000000000000001e-07, "loss": 3.6139, "step": 25 }, { "epoch": 0.0, "learning_rate": 5.2e-07, "loss": 3.5613, "step": 26 }, { "epoch": 0.0, "learning_rate": 5.4e-07, "loss": 3.6529, "step": 27 }, { "epoch": 0.0, "learning_rate": 5.6e-07, "loss": 3.5447, "step": 28 }, { "epoch": 0.0, "learning_rate": 5.800000000000001e-07, "loss": 3.5633, "step": 29 }, { "epoch": 0.0, "learning_rate": 6.000000000000001e-07, "loss": 3.5116, "step": 30 }, { "epoch": 0.0, "learning_rate": 6.200000000000001e-07, "loss": 3.5893, "step": 31 }, { "epoch": 0.0, "learning_rate": 6.4e-07, "loss": 3.5696, "step": 32 }, { "epoch": 0.0, "learning_rate": 6.6e-07, "loss": 3.5689, "step": 33 }, { "epoch": 0.0, "learning_rate": 6.800000000000001e-07, "loss": 3.5418, "step": 34 }, { "epoch": 0.0, "learning_rate": 7.000000000000001e-07, "loss": 3.6034, "step": 35 }, { "epoch": 0.0, "learning_rate": 7.2e-07, "loss": 3.5592, "step": 36 }, { "epoch": 0.0, "learning_rate": 7.4e-07, "loss": 3.5322, "step": 37 }, { "epoch": 0.0, "learning_rate": 7.6e-07, "loss": 3.5434, "step": 38 }, { "epoch": 0.0, "learning_rate": 7.8e-07, "loss": 3.5934, "step": 39 }, { "epoch": 0.0, "learning_rate": 8.000000000000001e-07, "loss": 3.5001, "step": 40 }, { "epoch": 0.0, "learning_rate": 8.200000000000001e-07, "loss": 3.5205, "step": 41 }, { "epoch": 0.01, "learning_rate": 8.400000000000001e-07, "loss": 3.4779, "step": 42 }, { "epoch": 0.01, "learning_rate": 8.6e-07, "loss": 3.5687, "step": 43 }, { "epoch": 0.01, "learning_rate": 8.8e-07, "loss": 3.5086, "step": 44 }, { "epoch": 0.01, "learning_rate": 9.000000000000001e-07, "loss": 3.447, "step": 45 }, { "epoch": 0.01, "learning_rate": 9.200000000000001e-07, "loss": 3.5478, "step": 46 }, { "epoch": 0.01, "learning_rate": 9.400000000000001e-07, "loss": 3.4941, "step": 47 }, { "epoch": 0.01, "learning_rate": 9.600000000000001e-07, "loss": 3.6038, "step": 48 }, { "epoch": 0.01, "learning_rate": 9.800000000000001e-07, "loss": 3.557, "step": 49 }, { "epoch": 0.01, "learning_rate": 1.0000000000000002e-06, "loss": 3.5297, "step": 50 }, { "epoch": 0.01, "learning_rate": 1.02e-06, "loss": 3.5566, "step": 51 }, { "epoch": 0.01, "learning_rate": 1.04e-06, "loss": 3.6164, "step": 52 }, { "epoch": 0.01, "learning_rate": 1.06e-06, "loss": 3.4671, "step": 53 }, { "epoch": 0.01, "learning_rate": 1.08e-06, "loss": 3.5748, "step": 54 }, { "epoch": 0.01, "learning_rate": 1.1e-06, "loss": 3.3999, "step": 55 }, { "epoch": 0.01, "learning_rate": 1.12e-06, "loss": 3.4698, "step": 56 }, { "epoch": 0.01, "learning_rate": 1.14e-06, "loss": 3.5406, "step": 57 }, { "epoch": 0.01, "learning_rate": 1.1600000000000001e-06, "loss": 3.5285, "step": 58 }, { "epoch": 0.01, "learning_rate": 1.1800000000000001e-06, "loss": 3.5117, "step": 59 }, { "epoch": 0.01, "learning_rate": 1.2000000000000002e-06, "loss": 3.4819, "step": 60 }, { "epoch": 0.01, "learning_rate": 1.2200000000000002e-06, "loss": 3.5661, "step": 61 }, { "epoch": 0.01, "learning_rate": 1.2400000000000002e-06, "loss": 3.3824, "step": 62 }, { "epoch": 0.01, "learning_rate": 1.26e-06, "loss": 3.5119, "step": 63 }, { "epoch": 0.01, "learning_rate": 1.28e-06, "loss": 3.4955, "step": 64 }, { "epoch": 0.01, "learning_rate": 1.3e-06, "loss": 3.5505, "step": 65 }, { "epoch": 0.01, "learning_rate": 1.32e-06, "loss": 3.5437, "step": 66 }, { "epoch": 0.01, "learning_rate": 1.34e-06, "loss": 3.5874, "step": 67 }, { "epoch": 0.01, "learning_rate": 1.3600000000000001e-06, "loss": 3.5087, "step": 68 }, { "epoch": 0.01, "learning_rate": 1.3800000000000001e-06, "loss": 3.486, "step": 69 }, { "epoch": 0.01, "learning_rate": 1.4000000000000001e-06, "loss": 3.5921, "step": 70 }, { "epoch": 0.01, "learning_rate": 1.42e-06, "loss": 3.472, "step": 71 }, { "epoch": 0.01, "learning_rate": 1.44e-06, "loss": 3.457, "step": 72 }, { "epoch": 0.01, "learning_rate": 1.46e-06, "loss": 3.4373, "step": 73 }, { "epoch": 0.01, "learning_rate": 1.48e-06, "loss": 3.4833, "step": 74 }, { "epoch": 0.01, "learning_rate": 1.5e-06, "loss": 3.6109, "step": 75 }, { "epoch": 0.01, "learning_rate": 1.52e-06, "loss": 3.5814, "step": 76 }, { "epoch": 0.01, "learning_rate": 1.54e-06, "loss": 3.5238, "step": 77 }, { "epoch": 0.01, "learning_rate": 1.56e-06, "loss": 3.5157, "step": 78 }, { "epoch": 0.01, "learning_rate": 1.5800000000000001e-06, "loss": 3.4294, "step": 79 }, { "epoch": 0.01, "learning_rate": 1.6000000000000001e-06, "loss": 3.5577, "step": 80 }, { "epoch": 0.01, "learning_rate": 1.6200000000000002e-06, "loss": 3.5463, "step": 81 }, { "epoch": 0.01, "learning_rate": 1.6400000000000002e-06, "loss": 3.3995, "step": 82 }, { "epoch": 0.01, "learning_rate": 1.6600000000000002e-06, "loss": 3.6272, "step": 83 }, { "epoch": 0.01, "learning_rate": 1.6800000000000002e-06, "loss": 3.4859, "step": 84 }, { "epoch": 0.01, "learning_rate": 1.7000000000000002e-06, "loss": 3.517, "step": 85 }, { "epoch": 0.01, "learning_rate": 1.72e-06, "loss": 3.5112, "step": 86 }, { "epoch": 0.01, "learning_rate": 1.74e-06, "loss": 3.4514, "step": 87 }, { "epoch": 0.01, "learning_rate": 1.76e-06, "loss": 3.4751, "step": 88 }, { "epoch": 0.01, "learning_rate": 1.7800000000000001e-06, "loss": 3.5537, "step": 89 }, { "epoch": 0.01, "learning_rate": 1.8000000000000001e-06, "loss": 3.4517, "step": 90 }, { "epoch": 0.01, "learning_rate": 1.8200000000000002e-06, "loss": 3.5745, "step": 91 }, { "epoch": 0.01, "learning_rate": 1.8400000000000002e-06, "loss": 3.4993, "step": 92 }, { "epoch": 0.01, "learning_rate": 1.8600000000000002e-06, "loss": 3.5063, "step": 93 }, { "epoch": 0.01, "learning_rate": 1.8800000000000002e-06, "loss": 3.4392, "step": 94 }, { "epoch": 0.01, "learning_rate": 1.9000000000000002e-06, "loss": 3.5406, "step": 95 }, { "epoch": 0.01, "learning_rate": 1.9200000000000003e-06, "loss": 3.5014, "step": 96 }, { "epoch": 0.01, "learning_rate": 1.94e-06, "loss": 3.4478, "step": 97 }, { "epoch": 0.01, "learning_rate": 1.9600000000000003e-06, "loss": 3.5798, "step": 98 }, { "epoch": 0.01, "learning_rate": 1.98e-06, "loss": 3.5238, "step": 99 }, { "epoch": 0.01, "learning_rate": 2.0000000000000003e-06, "loss": 3.4352, "step": 100 }, { "epoch": 0.01, "learning_rate": 2.02e-06, "loss": 3.4643, "step": 101 }, { "epoch": 0.01, "learning_rate": 2.04e-06, "loss": 3.5393, "step": 102 }, { "epoch": 0.01, "learning_rate": 2.06e-06, "loss": 3.5849, "step": 103 }, { "epoch": 0.01, "learning_rate": 2.08e-06, "loss": 3.4146, "step": 104 }, { "epoch": 0.01, "learning_rate": 2.1000000000000002e-06, "loss": 3.4909, "step": 105 }, { "epoch": 0.01, "learning_rate": 2.12e-06, "loss": 3.5093, "step": 106 }, { "epoch": 0.01, "learning_rate": 2.1400000000000003e-06, "loss": 3.4614, "step": 107 }, { "epoch": 0.01, "learning_rate": 2.16e-06, "loss": 3.5305, "step": 108 }, { "epoch": 0.01, "learning_rate": 2.1800000000000003e-06, "loss": 3.5317, "step": 109 }, { "epoch": 0.01, "learning_rate": 2.2e-06, "loss": 3.4011, "step": 110 }, { "epoch": 0.01, "learning_rate": 2.2200000000000003e-06, "loss": 3.4357, "step": 111 }, { "epoch": 0.01, "learning_rate": 2.24e-06, "loss": 3.5402, "step": 112 }, { "epoch": 0.01, "learning_rate": 2.2600000000000004e-06, "loss": 3.4931, "step": 113 }, { "epoch": 0.01, "learning_rate": 2.28e-06, "loss": 3.5459, "step": 114 }, { "epoch": 0.01, "learning_rate": 2.3000000000000004e-06, "loss": 3.5701, "step": 115 }, { "epoch": 0.01, "learning_rate": 2.3200000000000002e-06, "loss": 3.5272, "step": 116 }, { "epoch": 0.01, "learning_rate": 2.3400000000000005e-06, "loss": 3.5134, "step": 117 }, { "epoch": 0.01, "learning_rate": 2.3600000000000003e-06, "loss": 3.4161, "step": 118 }, { "epoch": 0.01, "learning_rate": 2.38e-06, "loss": 3.5046, "step": 119 }, { "epoch": 0.01, "learning_rate": 2.4000000000000003e-06, "loss": 3.5166, "step": 120 }, { "epoch": 0.01, "learning_rate": 2.42e-06, "loss": 3.4386, "step": 121 }, { "epoch": 0.01, "learning_rate": 2.4400000000000004e-06, "loss": 3.5971, "step": 122 }, { "epoch": 0.01, "learning_rate": 2.46e-06, "loss": 3.4337, "step": 123 }, { "epoch": 0.01, "learning_rate": 2.4800000000000004e-06, "loss": 3.5229, "step": 124 }, { "epoch": 0.02, "learning_rate": 2.5e-06, "loss": 3.5637, "step": 125 }, { "epoch": 0.02, "learning_rate": 2.52e-06, "loss": 3.6026, "step": 126 }, { "epoch": 0.02, "learning_rate": 2.5400000000000002e-06, "loss": 3.5844, "step": 127 }, { "epoch": 0.02, "learning_rate": 2.56e-06, "loss": 3.5149, "step": 128 }, { "epoch": 0.02, "learning_rate": 2.5800000000000003e-06, "loss": 3.4322, "step": 129 }, { "epoch": 0.02, "learning_rate": 2.6e-06, "loss": 3.6132, "step": 130 }, { "epoch": 0.02, "learning_rate": 2.6200000000000003e-06, "loss": 3.4488, "step": 131 }, { "epoch": 0.02, "learning_rate": 2.64e-06, "loss": 3.4212, "step": 132 }, { "epoch": 0.02, "learning_rate": 2.6600000000000004e-06, "loss": 3.4804, "step": 133 }, { "epoch": 0.02, "learning_rate": 2.68e-06, "loss": 3.4269, "step": 134 }, { "epoch": 0.02, "learning_rate": 2.7000000000000004e-06, "loss": 3.4187, "step": 135 }, { "epoch": 0.02, "learning_rate": 2.7200000000000002e-06, "loss": 3.5009, "step": 136 }, { "epoch": 0.02, "learning_rate": 2.7400000000000004e-06, "loss": 3.563, "step": 137 }, { "epoch": 0.02, "learning_rate": 2.7600000000000003e-06, "loss": 3.4627, "step": 138 }, { "epoch": 0.02, "learning_rate": 2.7800000000000005e-06, "loss": 3.3283, "step": 139 }, { "epoch": 0.02, "learning_rate": 2.8000000000000003e-06, "loss": 3.4864, "step": 140 }, { "epoch": 0.02, "learning_rate": 2.82e-06, "loss": 3.4755, "step": 141 }, { "epoch": 0.02, "learning_rate": 2.84e-06, "loss": 3.5335, "step": 142 }, { "epoch": 0.02, "learning_rate": 2.86e-06, "loss": 3.5846, "step": 143 }, { "epoch": 0.02, "learning_rate": 2.88e-06, "loss": 3.5474, "step": 144 }, { "epoch": 0.02, "learning_rate": 2.9e-06, "loss": 3.4547, "step": 145 }, { "epoch": 0.02, "learning_rate": 2.92e-06, "loss": 3.4302, "step": 146 }, { "epoch": 0.02, "learning_rate": 2.9400000000000002e-06, "loss": 3.5217, "step": 147 }, { "epoch": 0.02, "learning_rate": 2.96e-06, "loss": 3.5518, "step": 148 }, { "epoch": 0.02, "learning_rate": 2.9800000000000003e-06, "loss": 3.5234, "step": 149 }, { "epoch": 0.02, "learning_rate": 3e-06, "loss": 3.4657, "step": 150 }, { "epoch": 0.02, "learning_rate": 3.0200000000000003e-06, "loss": 3.4811, "step": 151 }, { "epoch": 0.02, "learning_rate": 3.04e-06, "loss": 3.4306, "step": 152 }, { "epoch": 0.02, "learning_rate": 3.0600000000000003e-06, "loss": 3.5413, "step": 153 }, { "epoch": 0.02, "learning_rate": 3.08e-06, "loss": 3.5566, "step": 154 }, { "epoch": 0.02, "learning_rate": 3.1000000000000004e-06, "loss": 3.4853, "step": 155 }, { "epoch": 0.02, "learning_rate": 3.12e-06, "loss": 3.4404, "step": 156 }, { "epoch": 0.02, "learning_rate": 3.1400000000000004e-06, "loss": 3.4934, "step": 157 }, { "epoch": 0.02, "learning_rate": 3.1600000000000002e-06, "loss": 3.516, "step": 158 }, { "epoch": 0.02, "learning_rate": 3.1800000000000005e-06, "loss": 3.5249, "step": 159 }, { "epoch": 0.02, "learning_rate": 3.2000000000000003e-06, "loss": 3.4669, "step": 160 }, { "epoch": 0.02, "learning_rate": 3.2200000000000005e-06, "loss": 3.5131, "step": 161 }, { "epoch": 0.02, "learning_rate": 3.2400000000000003e-06, "loss": 3.4701, "step": 162 }, { "epoch": 0.02, "learning_rate": 3.2600000000000006e-06, "loss": 3.4417, "step": 163 }, { "epoch": 0.02, "learning_rate": 3.2800000000000004e-06, "loss": 3.5104, "step": 164 }, { "epoch": 0.02, "learning_rate": 3.3000000000000006e-06, "loss": 3.5083, "step": 165 }, { "epoch": 0.02, "learning_rate": 3.3200000000000004e-06, "loss": 3.3667, "step": 166 }, { "epoch": 0.02, "learning_rate": 3.3400000000000006e-06, "loss": 3.4783, "step": 167 }, { "epoch": 0.02, "learning_rate": 3.3600000000000004e-06, "loss": 3.3911, "step": 168 }, { "epoch": 0.02, "learning_rate": 3.3800000000000007e-06, "loss": 3.4214, "step": 169 }, { "epoch": 0.02, "learning_rate": 3.4000000000000005e-06, "loss": 3.3774, "step": 170 }, { "epoch": 0.02, "learning_rate": 3.4200000000000007e-06, "loss": 3.4724, "step": 171 }, { "epoch": 0.02, "learning_rate": 3.44e-06, "loss": 3.5492, "step": 172 }, { "epoch": 0.02, "learning_rate": 3.46e-06, "loss": 3.4824, "step": 173 }, { "epoch": 0.02, "learning_rate": 3.48e-06, "loss": 3.4754, "step": 174 }, { "epoch": 0.02, "learning_rate": 3.5e-06, "loss": 3.5011, "step": 175 }, { "epoch": 0.02, "learning_rate": 3.52e-06, "loss": 3.4478, "step": 176 }, { "epoch": 0.02, "learning_rate": 3.54e-06, "loss": 3.4779, "step": 177 }, { "epoch": 0.02, "learning_rate": 3.5600000000000002e-06, "loss": 3.4549, "step": 178 }, { "epoch": 0.02, "learning_rate": 3.58e-06, "loss": 3.5151, "step": 179 }, { "epoch": 0.02, "learning_rate": 3.6000000000000003e-06, "loss": 3.5118, "step": 180 }, { "epoch": 0.02, "learning_rate": 3.62e-06, "loss": 3.457, "step": 181 }, { "epoch": 0.02, "learning_rate": 3.6400000000000003e-06, "loss": 3.4668, "step": 182 }, { "epoch": 0.02, "learning_rate": 3.66e-06, "loss": 3.5404, "step": 183 }, { "epoch": 0.02, "learning_rate": 3.6800000000000003e-06, "loss": 3.5241, "step": 184 }, { "epoch": 0.02, "learning_rate": 3.7e-06, "loss": 3.3628, "step": 185 }, { "epoch": 0.02, "learning_rate": 3.7200000000000004e-06, "loss": 3.5308, "step": 186 }, { "epoch": 0.02, "learning_rate": 3.74e-06, "loss": 3.5453, "step": 187 }, { "epoch": 0.02, "learning_rate": 3.7600000000000004e-06, "loss": 3.5159, "step": 188 }, { "epoch": 0.02, "learning_rate": 3.7800000000000002e-06, "loss": 3.595, "step": 189 }, { "epoch": 0.02, "learning_rate": 3.8000000000000005e-06, "loss": 3.5548, "step": 190 }, { "epoch": 0.02, "learning_rate": 3.820000000000001e-06, "loss": 3.4849, "step": 191 }, { "epoch": 0.02, "learning_rate": 3.8400000000000005e-06, "loss": 3.4687, "step": 192 }, { "epoch": 0.02, "learning_rate": 3.86e-06, "loss": 3.5025, "step": 193 }, { "epoch": 0.02, "learning_rate": 3.88e-06, "loss": 3.4915, "step": 194 }, { "epoch": 0.02, "learning_rate": 3.900000000000001e-06, "loss": 3.418, "step": 195 }, { "epoch": 0.02, "learning_rate": 3.920000000000001e-06, "loss": 3.4232, "step": 196 }, { "epoch": 0.02, "learning_rate": 3.94e-06, "loss": 3.3747, "step": 197 }, { "epoch": 0.02, "learning_rate": 3.96e-06, "loss": 3.5139, "step": 198 }, { "epoch": 0.02, "learning_rate": 3.980000000000001e-06, "loss": 3.438, "step": 199 }, { "epoch": 0.02, "learning_rate": 4.000000000000001e-06, "loss": 3.5688, "step": 200 }, { "epoch": 0.02, "learning_rate": 4.0200000000000005e-06, "loss": 3.4231, "step": 201 }, { "epoch": 0.02, "learning_rate": 4.04e-06, "loss": 3.4774, "step": 202 }, { "epoch": 0.02, "learning_rate": 4.060000000000001e-06, "loss": 3.5385, "step": 203 }, { "epoch": 0.02, "learning_rate": 4.08e-06, "loss": 3.4729, "step": 204 }, { "epoch": 0.02, "learning_rate": 4.1e-06, "loss": 3.3779, "step": 205 }, { "epoch": 0.02, "learning_rate": 4.12e-06, "loss": 3.48, "step": 206 }, { "epoch": 0.02, "learning_rate": 4.14e-06, "loss": 3.5035, "step": 207 }, { "epoch": 0.02, "learning_rate": 4.16e-06, "loss": 3.423, "step": 208 }, { "epoch": 0.03, "learning_rate": 4.18e-06, "loss": 3.4939, "step": 209 }, { "epoch": 0.03, "learning_rate": 4.2000000000000004e-06, "loss": 3.5306, "step": 210 }, { "epoch": 0.03, "learning_rate": 4.22e-06, "loss": 3.4389, "step": 211 }, { "epoch": 0.03, "learning_rate": 4.24e-06, "loss": 3.6344, "step": 212 }, { "epoch": 0.03, "learning_rate": 4.26e-06, "loss": 3.4444, "step": 213 }, { "epoch": 0.03, "learning_rate": 4.2800000000000005e-06, "loss": 3.531, "step": 214 }, { "epoch": 0.03, "learning_rate": 4.3e-06, "loss": 3.5216, "step": 215 }, { "epoch": 0.03, "learning_rate": 4.32e-06, "loss": 3.5074, "step": 216 }, { "epoch": 0.03, "learning_rate": 4.34e-06, "loss": 3.4796, "step": 217 }, { "epoch": 0.03, "learning_rate": 4.360000000000001e-06, "loss": 3.5106, "step": 218 }, { "epoch": 0.03, "learning_rate": 4.38e-06, "loss": 3.4713, "step": 219 }, { "epoch": 0.03, "learning_rate": 4.4e-06, "loss": 3.5708, "step": 220 }, { "epoch": 0.03, "learning_rate": 4.42e-06, "loss": 3.5622, "step": 221 }, { "epoch": 0.03, "learning_rate": 4.440000000000001e-06, "loss": 3.5348, "step": 222 }, { "epoch": 0.03, "learning_rate": 4.4600000000000005e-06, "loss": 3.4844, "step": 223 }, { "epoch": 0.03, "learning_rate": 4.48e-06, "loss": 3.3848, "step": 224 }, { "epoch": 0.03, "learning_rate": 4.5e-06, "loss": 3.4477, "step": 225 }, { "epoch": 0.03, "learning_rate": 4.520000000000001e-06, "loss": 3.4964, "step": 226 }, { "epoch": 0.03, "learning_rate": 4.540000000000001e-06, "loss": 3.5782, "step": 227 }, { "epoch": 0.03, "learning_rate": 4.56e-06, "loss": 3.5176, "step": 228 }, { "epoch": 0.03, "learning_rate": 4.58e-06, "loss": 3.5451, "step": 229 }, { "epoch": 0.03, "learning_rate": 4.600000000000001e-06, "loss": 3.3885, "step": 230 }, { "epoch": 0.03, "learning_rate": 4.620000000000001e-06, "loss": 3.4008, "step": 231 }, { "epoch": 0.03, "learning_rate": 4.6400000000000005e-06, "loss": 3.4892, "step": 232 }, { "epoch": 0.03, "learning_rate": 4.66e-06, "loss": 3.4706, "step": 233 }, { "epoch": 0.03, "learning_rate": 4.680000000000001e-06, "loss": 3.4611, "step": 234 }, { "epoch": 0.03, "learning_rate": 4.7e-06, "loss": 3.4612, "step": 235 }, { "epoch": 0.03, "learning_rate": 4.7200000000000005e-06, "loss": 3.4997, "step": 236 }, { "epoch": 0.03, "learning_rate": 4.74e-06, "loss": 3.5095, "step": 237 }, { "epoch": 0.03, "learning_rate": 4.76e-06, "loss": 3.6152, "step": 238 }, { "epoch": 0.03, "learning_rate": 4.78e-06, "loss": 3.379, "step": 239 }, { "epoch": 0.03, "learning_rate": 4.800000000000001e-06, "loss": 3.4724, "step": 240 }, { "epoch": 0.03, "learning_rate": 4.8200000000000004e-06, "loss": 3.496, "step": 241 }, { "epoch": 0.03, "learning_rate": 4.84e-06, "loss": 3.471, "step": 242 }, { "epoch": 0.03, "learning_rate": 4.86e-06, "loss": 3.4473, "step": 243 }, { "epoch": 0.03, "learning_rate": 4.880000000000001e-06, "loss": 3.5776, "step": 244 }, { "epoch": 0.03, "learning_rate": 4.9000000000000005e-06, "loss": 3.3948, "step": 245 }, { "epoch": 0.03, "learning_rate": 4.92e-06, "loss": 3.432, "step": 246 }, { "epoch": 0.03, "learning_rate": 4.94e-06, "loss": 3.4749, "step": 247 }, { "epoch": 0.03, "learning_rate": 4.960000000000001e-06, "loss": 3.5184, "step": 248 }, { "epoch": 0.03, "learning_rate": 4.980000000000001e-06, "loss": 3.3524, "step": 249 }, { "epoch": 0.03, "learning_rate": 5e-06, "loss": 3.5028, "step": 250 }, { "epoch": 0.03, "learning_rate": 5.02e-06, "loss": 3.4791, "step": 251 }, { "epoch": 0.03, "learning_rate": 5.04e-06, "loss": 3.4156, "step": 252 }, { "epoch": 0.03, "learning_rate": 5.060000000000001e-06, "loss": 3.4863, "step": 253 }, { "epoch": 0.03, "learning_rate": 5.0800000000000005e-06, "loss": 3.5285, "step": 254 }, { "epoch": 0.03, "learning_rate": 5.1e-06, "loss": 3.4141, "step": 255 }, { "epoch": 0.03, "learning_rate": 5.12e-06, "loss": 3.5133, "step": 256 }, { "epoch": 0.03, "learning_rate": 5.140000000000001e-06, "loss": 3.5152, "step": 257 }, { "epoch": 0.03, "learning_rate": 5.1600000000000006e-06, "loss": 3.5192, "step": 258 }, { "epoch": 0.03, "learning_rate": 5.18e-06, "loss": 3.4712, "step": 259 }, { "epoch": 0.03, "learning_rate": 5.2e-06, "loss": 3.4719, "step": 260 }, { "epoch": 0.03, "learning_rate": 5.220000000000001e-06, "loss": 3.4922, "step": 261 }, { "epoch": 0.03, "learning_rate": 5.240000000000001e-06, "loss": 3.4697, "step": 262 }, { "epoch": 0.03, "learning_rate": 5.2600000000000005e-06, "loss": 3.4879, "step": 263 }, { "epoch": 0.03, "learning_rate": 5.28e-06, "loss": 3.3733, "step": 264 }, { "epoch": 0.03, "learning_rate": 5.300000000000001e-06, "loss": 3.4686, "step": 265 }, { "epoch": 0.03, "learning_rate": 5.320000000000001e-06, "loss": 3.4592, "step": 266 }, { "epoch": 0.03, "learning_rate": 5.3400000000000005e-06, "loss": 3.4686, "step": 267 }, { "epoch": 0.03, "learning_rate": 5.36e-06, "loss": 3.4301, "step": 268 }, { "epoch": 0.03, "learning_rate": 5.380000000000001e-06, "loss": 3.3826, "step": 269 }, { "epoch": 0.03, "learning_rate": 5.400000000000001e-06, "loss": 3.5095, "step": 270 }, { "epoch": 0.03, "learning_rate": 5.420000000000001e-06, "loss": 3.3577, "step": 271 }, { "epoch": 0.03, "learning_rate": 5.4400000000000004e-06, "loss": 3.5313, "step": 272 }, { "epoch": 0.03, "learning_rate": 5.460000000000001e-06, "loss": 3.5679, "step": 273 }, { "epoch": 0.03, "learning_rate": 5.480000000000001e-06, "loss": 3.3919, "step": 274 }, { "epoch": 0.03, "learning_rate": 5.500000000000001e-06, "loss": 3.5052, "step": 275 }, { "epoch": 0.03, "learning_rate": 5.5200000000000005e-06, "loss": 3.5056, "step": 276 }, { "epoch": 0.03, "learning_rate": 5.540000000000001e-06, "loss": 3.4067, "step": 277 }, { "epoch": 0.03, "learning_rate": 5.560000000000001e-06, "loss": 3.4534, "step": 278 }, { "epoch": 0.03, "learning_rate": 5.580000000000001e-06, "loss": 3.5559, "step": 279 }, { "epoch": 0.03, "learning_rate": 5.600000000000001e-06, "loss": 3.4197, "step": 280 }, { "epoch": 0.03, "learning_rate": 5.620000000000001e-06, "loss": 3.3978, "step": 281 }, { "epoch": 0.03, "learning_rate": 5.64e-06, "loss": 3.3737, "step": 282 }, { "epoch": 0.03, "learning_rate": 5.66e-06, "loss": 3.5291, "step": 283 }, { "epoch": 0.03, "learning_rate": 5.68e-06, "loss": 3.6118, "step": 284 }, { "epoch": 0.03, "learning_rate": 5.7e-06, "loss": 3.4452, "step": 285 }, { "epoch": 0.03, "learning_rate": 5.72e-06, "loss": 3.4343, "step": 286 }, { "epoch": 0.03, "learning_rate": 5.74e-06, "loss": 3.5183, "step": 287 }, { "epoch": 0.03, "learning_rate": 5.76e-06, "loss": 3.5366, "step": 288 }, { "epoch": 0.03, "learning_rate": 5.78e-06, "loss": 3.4111, "step": 289 }, { "epoch": 0.03, "learning_rate": 5.8e-06, "loss": 3.4251, "step": 290 }, { "epoch": 0.03, "learning_rate": 5.82e-06, "loss": 3.3772, "step": 291 }, { "epoch": 0.04, "learning_rate": 5.84e-06, "loss": 3.4691, "step": 292 }, { "epoch": 0.04, "learning_rate": 5.86e-06, "loss": 3.5742, "step": 293 }, { "epoch": 0.04, "learning_rate": 5.8800000000000005e-06, "loss": 3.413, "step": 294 }, { "epoch": 0.04, "learning_rate": 5.9e-06, "loss": 3.4888, "step": 295 }, { "epoch": 0.04, "learning_rate": 5.92e-06, "loss": 3.5007, "step": 296 }, { "epoch": 0.04, "learning_rate": 5.94e-06, "loss": 3.5076, "step": 297 }, { "epoch": 0.04, "learning_rate": 5.9600000000000005e-06, "loss": 3.4333, "step": 298 }, { "epoch": 0.04, "learning_rate": 5.98e-06, "loss": 3.4994, "step": 299 }, { "epoch": 0.04, "learning_rate": 6e-06, "loss": 3.403, "step": 300 }, { "epoch": 0.04, "learning_rate": 6.02e-06, "loss": 3.5639, "step": 301 }, { "epoch": 0.04, "learning_rate": 6.040000000000001e-06, "loss": 3.4472, "step": 302 }, { "epoch": 0.04, "learning_rate": 6.0600000000000004e-06, "loss": 3.5485, "step": 303 }, { "epoch": 0.04, "learning_rate": 6.08e-06, "loss": 3.3861, "step": 304 }, { "epoch": 0.04, "learning_rate": 6.1e-06, "loss": 3.3337, "step": 305 }, { "epoch": 0.04, "learning_rate": 6.120000000000001e-06, "loss": 3.4606, "step": 306 }, { "epoch": 0.04, "learning_rate": 6.1400000000000005e-06, "loss": 3.4987, "step": 307 }, { "epoch": 0.04, "learning_rate": 6.16e-06, "loss": 3.4992, "step": 308 }, { "epoch": 0.04, "learning_rate": 6.18e-06, "loss": 3.509, "step": 309 }, { "epoch": 0.04, "learning_rate": 6.200000000000001e-06, "loss": 3.4343, "step": 310 }, { "epoch": 0.04, "learning_rate": 6.220000000000001e-06, "loss": 3.4416, "step": 311 }, { "epoch": 0.04, "learning_rate": 6.24e-06, "loss": 3.5266, "step": 312 }, { "epoch": 0.04, "learning_rate": 6.26e-06, "loss": 3.4589, "step": 313 }, { "epoch": 0.04, "learning_rate": 6.280000000000001e-06, "loss": 3.4018, "step": 314 }, { "epoch": 0.04, "learning_rate": 6.300000000000001e-06, "loss": 3.5158, "step": 315 }, { "epoch": 0.04, "learning_rate": 6.3200000000000005e-06, "loss": 3.4101, "step": 316 }, { "epoch": 0.04, "learning_rate": 6.34e-06, "loss": 3.4509, "step": 317 }, { "epoch": 0.04, "learning_rate": 6.360000000000001e-06, "loss": 3.4038, "step": 318 }, { "epoch": 0.04, "learning_rate": 6.380000000000001e-06, "loss": 3.5194, "step": 319 }, { "epoch": 0.04, "learning_rate": 6.4000000000000006e-06, "loss": 3.4715, "step": 320 }, { "epoch": 0.04, "learning_rate": 6.42e-06, "loss": 3.372, "step": 321 }, { "epoch": 0.04, "learning_rate": 6.440000000000001e-06, "loss": 3.4225, "step": 322 }, { "epoch": 0.04, "learning_rate": 6.460000000000001e-06, "loss": 3.433, "step": 323 }, { "epoch": 0.04, "learning_rate": 6.480000000000001e-06, "loss": 3.3834, "step": 324 }, { "epoch": 0.04, "learning_rate": 6.5000000000000004e-06, "loss": 3.5219, "step": 325 }, { "epoch": 0.04, "learning_rate": 6.520000000000001e-06, "loss": 3.4602, "step": 326 }, { "epoch": 0.04, "learning_rate": 6.540000000000001e-06, "loss": 3.4697, "step": 327 }, { "epoch": 0.04, "learning_rate": 6.560000000000001e-06, "loss": 3.4229, "step": 328 }, { "epoch": 0.04, "learning_rate": 6.5800000000000005e-06, "loss": 3.4922, "step": 329 }, { "epoch": 0.04, "learning_rate": 6.600000000000001e-06, "loss": 3.4047, "step": 330 }, { "epoch": 0.04, "learning_rate": 6.620000000000001e-06, "loss": 3.4781, "step": 331 }, { "epoch": 0.04, "learning_rate": 6.640000000000001e-06, "loss": 3.4487, "step": 332 }, { "epoch": 0.04, "learning_rate": 6.660000000000001e-06, "loss": 3.3814, "step": 333 }, { "epoch": 0.04, "learning_rate": 6.680000000000001e-06, "loss": 3.5575, "step": 334 }, { "epoch": 0.04, "learning_rate": 6.700000000000001e-06, "loss": 3.5155, "step": 335 }, { "epoch": 0.04, "learning_rate": 6.720000000000001e-06, "loss": 3.3543, "step": 336 }, { "epoch": 0.04, "learning_rate": 6.740000000000001e-06, "loss": 3.477, "step": 337 }, { "epoch": 0.04, "learning_rate": 6.760000000000001e-06, "loss": 3.5477, "step": 338 }, { "epoch": 0.04, "learning_rate": 6.780000000000001e-06, "loss": 3.4125, "step": 339 }, { "epoch": 0.04, "learning_rate": 6.800000000000001e-06, "loss": 3.5367, "step": 340 }, { "epoch": 0.04, "learning_rate": 6.820000000000001e-06, "loss": 3.4789, "step": 341 }, { "epoch": 0.04, "learning_rate": 6.8400000000000014e-06, "loss": 3.4742, "step": 342 }, { "epoch": 0.04, "learning_rate": 6.860000000000001e-06, "loss": 3.5918, "step": 343 }, { "epoch": 0.04, "learning_rate": 6.88e-06, "loss": 3.5261, "step": 344 }, { "epoch": 0.04, "learning_rate": 6.9e-06, "loss": 3.3011, "step": 345 }, { "epoch": 0.04, "learning_rate": 6.92e-06, "loss": 3.5385, "step": 346 }, { "epoch": 0.04, "learning_rate": 6.9400000000000005e-06, "loss": 3.4922, "step": 347 }, { "epoch": 0.04, "learning_rate": 6.96e-06, "loss": 3.5355, "step": 348 }, { "epoch": 0.04, "learning_rate": 6.98e-06, "loss": 3.4166, "step": 349 }, { "epoch": 0.04, "learning_rate": 7e-06, "loss": 3.5014, "step": 350 }, { "epoch": 0.04, "learning_rate": 7.0200000000000006e-06, "loss": 3.4994, "step": 351 }, { "epoch": 0.04, "learning_rate": 7.04e-06, "loss": 3.5292, "step": 352 }, { "epoch": 0.04, "learning_rate": 7.06e-06, "loss": 3.4428, "step": 353 }, { "epoch": 0.04, "learning_rate": 7.08e-06, "loss": 3.4447, "step": 354 }, { "epoch": 0.04, "learning_rate": 7.100000000000001e-06, "loss": 3.5367, "step": 355 }, { "epoch": 0.04, "learning_rate": 7.1200000000000004e-06, "loss": 3.4225, "step": 356 }, { "epoch": 0.04, "learning_rate": 7.14e-06, "loss": 3.4368, "step": 357 }, { "epoch": 0.04, "learning_rate": 7.16e-06, "loss": 3.4523, "step": 358 }, { "epoch": 0.04, "learning_rate": 7.180000000000001e-06, "loss": 3.5183, "step": 359 }, { "epoch": 0.04, "learning_rate": 7.2000000000000005e-06, "loss": 3.4799, "step": 360 }, { "epoch": 0.04, "learning_rate": 7.22e-06, "loss": 3.4702, "step": 361 }, { "epoch": 0.04, "learning_rate": 7.24e-06, "loss": 3.3872, "step": 362 }, { "epoch": 0.04, "learning_rate": 7.260000000000001e-06, "loss": 3.5367, "step": 363 }, { "epoch": 0.04, "learning_rate": 7.280000000000001e-06, "loss": 3.5125, "step": 364 }, { "epoch": 0.04, "learning_rate": 7.3e-06, "loss": 3.4316, "step": 365 }, { "epoch": 0.04, "learning_rate": 7.32e-06, "loss": 3.4441, "step": 366 }, { "epoch": 0.04, "learning_rate": 7.340000000000001e-06, "loss": 3.4337, "step": 367 }, { "epoch": 0.04, "learning_rate": 7.360000000000001e-06, "loss": 3.384, "step": 368 }, { "epoch": 0.04, "learning_rate": 7.3800000000000005e-06, "loss": 3.4932, "step": 369 }, { "epoch": 0.04, "learning_rate": 7.4e-06, "loss": 3.5135, "step": 370 }, { "epoch": 0.04, "learning_rate": 7.420000000000001e-06, "loss": 3.5265, "step": 371 }, { "epoch": 0.04, "learning_rate": 7.440000000000001e-06, "loss": 3.4276, "step": 372 }, { "epoch": 0.04, "learning_rate": 7.4600000000000006e-06, "loss": 3.4163, "step": 373 }, { "epoch": 0.04, "learning_rate": 7.48e-06, "loss": 3.3181, "step": 374 }, { "epoch": 0.05, "learning_rate": 7.500000000000001e-06, "loss": 3.4895, "step": 375 }, { "epoch": 0.05, "learning_rate": 7.520000000000001e-06, "loss": 3.4736, "step": 376 }, { "epoch": 0.05, "learning_rate": 7.540000000000001e-06, "loss": 3.5184, "step": 377 }, { "epoch": 0.05, "learning_rate": 7.5600000000000005e-06, "loss": 3.3869, "step": 378 }, { "epoch": 0.05, "learning_rate": 7.58e-06, "loss": 3.4409, "step": 379 }, { "epoch": 0.05, "learning_rate": 7.600000000000001e-06, "loss": 3.4342, "step": 380 }, { "epoch": 0.05, "learning_rate": 7.620000000000001e-06, "loss": 3.3791, "step": 381 }, { "epoch": 0.05, "learning_rate": 7.640000000000001e-06, "loss": 3.5785, "step": 382 }, { "epoch": 0.05, "learning_rate": 7.660000000000001e-06, "loss": 3.3772, "step": 383 }, { "epoch": 0.05, "learning_rate": 7.680000000000001e-06, "loss": 3.4101, "step": 384 }, { "epoch": 0.05, "learning_rate": 7.7e-06, "loss": 3.3965, "step": 385 }, { "epoch": 0.05, "learning_rate": 7.72e-06, "loss": 3.4899, "step": 386 }, { "epoch": 0.05, "learning_rate": 7.74e-06, "loss": 3.4214, "step": 387 }, { "epoch": 0.05, "learning_rate": 7.76e-06, "loss": 3.5041, "step": 388 }, { "epoch": 0.05, "learning_rate": 7.78e-06, "loss": 3.4801, "step": 389 }, { "epoch": 0.05, "learning_rate": 7.800000000000002e-06, "loss": 3.4566, "step": 390 }, { "epoch": 0.05, "learning_rate": 7.820000000000001e-06, "loss": 3.4937, "step": 391 }, { "epoch": 0.05, "learning_rate": 7.840000000000001e-06, "loss": 3.4241, "step": 392 }, { "epoch": 0.05, "learning_rate": 7.860000000000001e-06, "loss": 3.4368, "step": 393 }, { "epoch": 0.05, "learning_rate": 7.88e-06, "loss": 3.4587, "step": 394 }, { "epoch": 0.05, "learning_rate": 7.9e-06, "loss": 3.5883, "step": 395 }, { "epoch": 0.05, "learning_rate": 7.92e-06, "loss": 3.3842, "step": 396 }, { "epoch": 0.05, "learning_rate": 7.94e-06, "loss": 3.5293, "step": 397 }, { "epoch": 0.05, "learning_rate": 7.960000000000002e-06, "loss": 3.5071, "step": 398 }, { "epoch": 0.05, "learning_rate": 7.980000000000002e-06, "loss": 3.4799, "step": 399 }, { "epoch": 0.05, "learning_rate": 8.000000000000001e-06, "loss": 3.4659, "step": 400 }, { "epoch": 0.05, "learning_rate": 8.020000000000001e-06, "loss": 3.4091, "step": 401 }, { "epoch": 0.05, "learning_rate": 8.040000000000001e-06, "loss": 3.5124, "step": 402 }, { "epoch": 0.05, "learning_rate": 8.06e-06, "loss": 3.4844, "step": 403 }, { "epoch": 0.05, "learning_rate": 8.08e-06, "loss": 3.4682, "step": 404 }, { "epoch": 0.05, "learning_rate": 8.1e-06, "loss": 3.4514, "step": 405 }, { "epoch": 0.05, "learning_rate": 8.120000000000002e-06, "loss": 3.4603, "step": 406 }, { "epoch": 0.05, "learning_rate": 8.14e-06, "loss": 3.3393, "step": 407 }, { "epoch": 0.05, "learning_rate": 8.16e-06, "loss": 3.4713, "step": 408 }, { "epoch": 0.05, "learning_rate": 8.18e-06, "loss": 3.4854, "step": 409 }, { "epoch": 0.05, "learning_rate": 8.2e-06, "loss": 3.4348, "step": 410 }, { "epoch": 0.05, "learning_rate": 8.220000000000001e-06, "loss": 3.4176, "step": 411 }, { "epoch": 0.05, "learning_rate": 8.24e-06, "loss": 3.3925, "step": 412 }, { "epoch": 0.05, "learning_rate": 8.26e-06, "loss": 3.4103, "step": 413 }, { "epoch": 0.05, "learning_rate": 8.28e-06, "loss": 3.5636, "step": 414 }, { "epoch": 0.05, "learning_rate": 8.3e-06, "loss": 3.4342, "step": 415 }, { "epoch": 0.05, "learning_rate": 8.32e-06, "loss": 3.4334, "step": 416 }, { "epoch": 0.05, "learning_rate": 8.34e-06, "loss": 3.4358, "step": 417 }, { "epoch": 0.05, "learning_rate": 8.36e-06, "loss": 3.4342, "step": 418 }, { "epoch": 0.05, "learning_rate": 8.380000000000001e-06, "loss": 3.4178, "step": 419 }, { "epoch": 0.05, "learning_rate": 8.400000000000001e-06, "loss": 3.4991, "step": 420 }, { "epoch": 0.05, "learning_rate": 8.42e-06, "loss": 3.5309, "step": 421 }, { "epoch": 0.05, "learning_rate": 8.44e-06, "loss": 3.5378, "step": 422 }, { "epoch": 0.05, "learning_rate": 8.46e-06, "loss": 3.4695, "step": 423 }, { "epoch": 0.05, "learning_rate": 8.48e-06, "loss": 3.4842, "step": 424 }, { "epoch": 0.05, "learning_rate": 8.5e-06, "loss": 3.3807, "step": 425 }, { "epoch": 0.05, "learning_rate": 8.52e-06, "loss": 3.4789, "step": 426 }, { "epoch": 0.05, "learning_rate": 8.540000000000001e-06, "loss": 3.4198, "step": 427 }, { "epoch": 0.05, "learning_rate": 8.560000000000001e-06, "loss": 3.436, "step": 428 }, { "epoch": 0.05, "learning_rate": 8.580000000000001e-06, "loss": 3.409, "step": 429 }, { "epoch": 0.05, "learning_rate": 8.6e-06, "loss": 3.4708, "step": 430 }, { "epoch": 0.05, "learning_rate": 8.62e-06, "loss": 3.4117, "step": 431 }, { "epoch": 0.05, "learning_rate": 8.64e-06, "loss": 3.4232, "step": 432 }, { "epoch": 0.05, "learning_rate": 8.66e-06, "loss": 3.4727, "step": 433 }, { "epoch": 0.05, "learning_rate": 8.68e-06, "loss": 3.4516, "step": 434 }, { "epoch": 0.05, "learning_rate": 8.700000000000001e-06, "loss": 3.3894, "step": 435 }, { "epoch": 0.05, "learning_rate": 8.720000000000001e-06, "loss": 3.462, "step": 436 }, { "epoch": 0.05, "learning_rate": 8.740000000000001e-06, "loss": 3.4965, "step": 437 }, { "epoch": 0.05, "learning_rate": 8.76e-06, "loss": 3.4715, "step": 438 }, { "epoch": 0.05, "learning_rate": 8.78e-06, "loss": 3.458, "step": 439 }, { "epoch": 0.05, "learning_rate": 8.8e-06, "loss": 3.5042, "step": 440 }, { "epoch": 0.05, "learning_rate": 8.82e-06, "loss": 3.4161, "step": 441 }, { "epoch": 0.05, "learning_rate": 8.84e-06, "loss": 3.3241, "step": 442 }, { "epoch": 0.05, "learning_rate": 8.860000000000002e-06, "loss": 3.3387, "step": 443 }, { "epoch": 0.05, "learning_rate": 8.880000000000001e-06, "loss": 3.4501, "step": 444 }, { "epoch": 0.05, "learning_rate": 8.900000000000001e-06, "loss": 3.5328, "step": 445 }, { "epoch": 0.05, "learning_rate": 8.920000000000001e-06, "loss": 3.3749, "step": 446 }, { "epoch": 0.05, "learning_rate": 8.94e-06, "loss": 3.4855, "step": 447 }, { "epoch": 0.05, "learning_rate": 8.96e-06, "loss": 3.5049, "step": 448 }, { "epoch": 0.05, "learning_rate": 8.98e-06, "loss": 3.4508, "step": 449 }, { "epoch": 0.05, "learning_rate": 9e-06, "loss": 3.3561, "step": 450 }, { "epoch": 0.05, "learning_rate": 9.020000000000002e-06, "loss": 3.4835, "step": 451 }, { "epoch": 0.05, "learning_rate": 9.040000000000002e-06, "loss": 3.489, "step": 452 }, { "epoch": 0.05, "learning_rate": 9.060000000000001e-06, "loss": 3.5312, "step": 453 }, { "epoch": 0.05, "learning_rate": 9.080000000000001e-06, "loss": 3.4514, "step": 454 }, { "epoch": 0.05, "learning_rate": 9.100000000000001e-06, "loss": 3.4641, "step": 455 }, { "epoch": 0.05, "learning_rate": 9.12e-06, "loss": 3.3816, "step": 456 }, { "epoch": 0.05, "learning_rate": 9.14e-06, "loss": 3.4733, "step": 457 }, { "epoch": 0.05, "learning_rate": 9.16e-06, "loss": 3.412, "step": 458 }, { "epoch": 0.06, "learning_rate": 9.180000000000002e-06, "loss": 3.4428, "step": 459 }, { "epoch": 0.06, "learning_rate": 9.200000000000002e-06, "loss": 3.477, "step": 460 }, { "epoch": 0.06, "learning_rate": 9.220000000000002e-06, "loss": 3.5042, "step": 461 }, { "epoch": 0.06, "learning_rate": 9.240000000000001e-06, "loss": 3.4824, "step": 462 }, { "epoch": 0.06, "learning_rate": 9.260000000000001e-06, "loss": 3.3663, "step": 463 }, { "epoch": 0.06, "learning_rate": 9.280000000000001e-06, "loss": 3.4138, "step": 464 }, { "epoch": 0.06, "learning_rate": 9.3e-06, "loss": 3.4354, "step": 465 }, { "epoch": 0.06, "learning_rate": 9.32e-06, "loss": 3.5427, "step": 466 }, { "epoch": 0.06, "learning_rate": 9.340000000000002e-06, "loss": 3.3885, "step": 467 }, { "epoch": 0.06, "learning_rate": 9.360000000000002e-06, "loss": 3.4725, "step": 468 }, { "epoch": 0.06, "learning_rate": 9.38e-06, "loss": 3.3745, "step": 469 }, { "epoch": 0.06, "learning_rate": 9.4e-06, "loss": 3.3946, "step": 470 }, { "epoch": 0.06, "learning_rate": 9.42e-06, "loss": 3.5372, "step": 471 }, { "epoch": 0.06, "learning_rate": 9.440000000000001e-06, "loss": 3.4622, "step": 472 }, { "epoch": 0.06, "learning_rate": 9.460000000000001e-06, "loss": 3.4287, "step": 473 }, { "epoch": 0.06, "learning_rate": 9.48e-06, "loss": 3.4092, "step": 474 }, { "epoch": 0.06, "learning_rate": 9.5e-06, "loss": 3.4878, "step": 475 }, { "epoch": 0.06, "learning_rate": 9.52e-06, "loss": 3.4251, "step": 476 }, { "epoch": 0.06, "learning_rate": 9.54e-06, "loss": 3.362, "step": 477 }, { "epoch": 0.06, "learning_rate": 9.56e-06, "loss": 3.4811, "step": 478 }, { "epoch": 0.06, "learning_rate": 9.58e-06, "loss": 3.3981, "step": 479 }, { "epoch": 0.06, "learning_rate": 9.600000000000001e-06, "loss": 3.344, "step": 480 }, { "epoch": 0.06, "learning_rate": 9.620000000000001e-06, "loss": 3.372, "step": 481 }, { "epoch": 0.06, "learning_rate": 9.640000000000001e-06, "loss": 3.4298, "step": 482 }, { "epoch": 0.06, "learning_rate": 9.66e-06, "loss": 3.4184, "step": 483 }, { "epoch": 0.06, "learning_rate": 9.68e-06, "loss": 3.3619, "step": 484 }, { "epoch": 0.06, "learning_rate": 9.7e-06, "loss": 3.3297, "step": 485 }, { "epoch": 0.06, "learning_rate": 9.72e-06, "loss": 3.4464, "step": 486 }, { "epoch": 0.06, "learning_rate": 9.74e-06, "loss": 3.571, "step": 487 }, { "epoch": 0.06, "learning_rate": 9.760000000000001e-06, "loss": 3.549, "step": 488 }, { "epoch": 0.06, "learning_rate": 9.780000000000001e-06, "loss": 3.4555, "step": 489 }, { "epoch": 0.06, "learning_rate": 9.800000000000001e-06, "loss": 3.3967, "step": 490 }, { "epoch": 0.06, "learning_rate": 9.820000000000001e-06, "loss": 3.4414, "step": 491 }, { "epoch": 0.06, "learning_rate": 9.84e-06, "loss": 3.4805, "step": 492 }, { "epoch": 0.06, "learning_rate": 9.86e-06, "loss": 3.4347, "step": 493 }, { "epoch": 0.06, "learning_rate": 9.88e-06, "loss": 3.41, "step": 494 }, { "epoch": 0.06, "learning_rate": 9.9e-06, "loss": 3.3899, "step": 495 }, { "epoch": 0.06, "learning_rate": 9.920000000000002e-06, "loss": 3.4549, "step": 496 }, { "epoch": 0.06, "learning_rate": 9.940000000000001e-06, "loss": 3.3854, "step": 497 }, { "epoch": 0.06, "learning_rate": 9.960000000000001e-06, "loss": 3.3616, "step": 498 }, { "epoch": 0.06, "learning_rate": 9.980000000000001e-06, "loss": 3.4413, "step": 499 }, { "epoch": 0.06, "learning_rate": 1e-05, "loss": 3.4429, "step": 500 }, { "epoch": 0.06, "learning_rate": 9.999999597648802e-06, "loss": 3.4906, "step": 501 }, { "epoch": 0.06, "learning_rate": 9.999998390595268e-06, "loss": 3.5216, "step": 502 }, { "epoch": 0.06, "learning_rate": 9.999996378839595e-06, "loss": 3.4375, "step": 503 }, { "epoch": 0.06, "learning_rate": 9.999993562382106e-06, "loss": 3.4196, "step": 504 }, { "epoch": 0.06, "learning_rate": 9.999989941223254e-06, "loss": 3.4782, "step": 505 }, { "epoch": 0.06, "learning_rate": 9.999985515363622e-06, "loss": 3.3184, "step": 506 }, { "epoch": 0.06, "learning_rate": 9.999980284803922e-06, "loss": 3.4054, "step": 507 }, { "epoch": 0.06, "learning_rate": 9.999974249544996e-06, "loss": 3.4274, "step": 508 }, { "epoch": 0.06, "learning_rate": 9.999967409587817e-06, "loss": 3.3777, "step": 509 }, { "epoch": 0.06, "learning_rate": 9.999959764933483e-06, "loss": 3.5171, "step": 510 }, { "epoch": 0.06, "learning_rate": 9.999951315583226e-06, "loss": 3.4816, "step": 511 }, { "epoch": 0.06, "learning_rate": 9.999942061538403e-06, "loss": 3.4967, "step": 512 }, { "epoch": 0.06, "learning_rate": 9.99993200280051e-06, "loss": 3.539, "step": 513 }, { "epoch": 0.06, "learning_rate": 9.99992113937116e-06, "loss": 3.5277, "step": 514 }, { "epoch": 0.06, "learning_rate": 9.999909471252103e-06, "loss": 3.4376, "step": 515 }, { "epoch": 0.06, "learning_rate": 9.999896998445216e-06, "loss": 3.4447, "step": 516 }, { "epoch": 0.06, "learning_rate": 9.999883720952507e-06, "loss": 3.5765, "step": 517 }, { "epoch": 0.06, "learning_rate": 9.999869638776115e-06, "loss": 3.4718, "step": 518 }, { "epoch": 0.06, "learning_rate": 9.999854751918303e-06, "loss": 3.4798, "step": 519 }, { "epoch": 0.06, "learning_rate": 9.999839060381469e-06, "loss": 3.3298, "step": 520 }, { "epoch": 0.06, "learning_rate": 9.999822564168138e-06, "loss": 3.5549, "step": 521 }, { "epoch": 0.06, "learning_rate": 9.999805263280966e-06, "loss": 3.4561, "step": 522 }, { "epoch": 0.06, "learning_rate": 9.999787157722735e-06, "loss": 3.3638, "step": 523 }, { "epoch": 0.06, "learning_rate": 9.99976824749636e-06, "loss": 3.4093, "step": 524 }, { "epoch": 0.06, "learning_rate": 9.999748532604884e-06, "loss": 3.4774, "step": 525 }, { "epoch": 0.06, "learning_rate": 9.999728013051482e-06, "loss": 3.3955, "step": 526 }, { "epoch": 0.06, "learning_rate": 9.999706688839454e-06, "loss": 3.4866, "step": 527 }, { "epoch": 0.06, "learning_rate": 9.999684559972234e-06, "loss": 3.5428, "step": 528 }, { "epoch": 0.06, "learning_rate": 9.99966162645338e-06, "loss": 3.5869, "step": 529 }, { "epoch": 0.06, "learning_rate": 9.999637888286587e-06, "loss": 3.3764, "step": 530 }, { "epoch": 0.06, "learning_rate": 9.999613345475673e-06, "loss": 3.4635, "step": 531 }, { "epoch": 0.06, "learning_rate": 9.99958799802459e-06, "loss": 3.4189, "step": 532 }, { "epoch": 0.06, "learning_rate": 9.999561845937413e-06, "loss": 3.4488, "step": 533 }, { "epoch": 0.06, "learning_rate": 9.999534889218355e-06, "loss": 3.3789, "step": 534 }, { "epoch": 0.06, "learning_rate": 9.999507127871755e-06, "loss": 3.4508, "step": 535 }, { "epoch": 0.06, "learning_rate": 9.999478561902077e-06, "loss": 3.383, "step": 536 }, { "epoch": 0.06, "learning_rate": 9.999449191313921e-06, "loss": 3.5043, "step": 537 }, { "epoch": 0.06, "learning_rate": 9.999419016112015e-06, "loss": 3.442, "step": 538 }, { "epoch": 0.06, "learning_rate": 9.999388036301213e-06, "loss": 3.4162, "step": 539 }, { "epoch": 0.06, "learning_rate": 9.999356251886503e-06, "loss": 3.363, "step": 540 }, { "epoch": 0.06, "learning_rate": 9.999323662872998e-06, "loss": 3.3994, "step": 541 }, { "epoch": 0.07, "learning_rate": 9.999290269265945e-06, "loss": 3.479, "step": 542 }, { "epoch": 0.07, "learning_rate": 9.999256071070717e-06, "loss": 3.5442, "step": 543 }, { "epoch": 0.07, "learning_rate": 9.999221068292819e-06, "loss": 3.4798, "step": 544 }, { "epoch": 0.07, "learning_rate": 9.999185260937883e-06, "loss": 3.455, "step": 545 }, { "epoch": 0.07, "learning_rate": 9.999148649011672e-06, "loss": 3.3145, "step": 546 }, { "epoch": 0.07, "learning_rate": 9.99911123252008e-06, "loss": 3.5217, "step": 547 }, { "epoch": 0.07, "learning_rate": 9.99907301146913e-06, "loss": 3.4865, "step": 548 }, { "epoch": 0.07, "learning_rate": 9.999033985864968e-06, "loss": 3.5578, "step": 549 }, { "epoch": 0.07, "learning_rate": 9.998994155713879e-06, "loss": 3.5073, "step": 550 }, { "epoch": 0.07, "learning_rate": 9.998953521022272e-06, "loss": 3.4298, "step": 551 }, { "epoch": 0.07, "learning_rate": 9.998912081796687e-06, "loss": 3.4367, "step": 552 }, { "epoch": 0.07, "learning_rate": 9.998869838043793e-06, "loss": 3.3957, "step": 553 }, { "epoch": 0.07, "learning_rate": 9.99882678977039e-06, "loss": 3.4318, "step": 554 }, { "epoch": 0.07, "learning_rate": 9.998782936983405e-06, "loss": 3.3927, "step": 555 }, { "epoch": 0.07, "learning_rate": 9.998738279689897e-06, "loss": 3.5085, "step": 556 }, { "epoch": 0.07, "learning_rate": 9.99869281789705e-06, "loss": 3.4349, "step": 557 }, { "epoch": 0.07, "learning_rate": 9.998646551612184e-06, "loss": 3.3959, "step": 558 }, { "epoch": 0.07, "learning_rate": 9.998599480842742e-06, "loss": 3.46, "step": 559 }, { "epoch": 0.07, "learning_rate": 9.998551605596303e-06, "loss": 3.4866, "step": 560 }, { "epoch": 0.07, "learning_rate": 9.998502925880569e-06, "loss": 3.4413, "step": 561 }, { "epoch": 0.07, "learning_rate": 9.998453441703378e-06, "loss": 3.4223, "step": 562 }, { "epoch": 0.07, "learning_rate": 9.998403153072691e-06, "loss": 3.4185, "step": 563 }, { "epoch": 0.07, "learning_rate": 9.998352059996603e-06, "loss": 3.4465, "step": 564 }, { "epoch": 0.07, "learning_rate": 9.998300162483336e-06, "loss": 3.459, "step": 565 }, { "epoch": 0.07, "learning_rate": 9.998247460541241e-06, "loss": 3.4184, "step": 566 }, { "epoch": 0.07, "learning_rate": 9.998193954178805e-06, "loss": 3.3669, "step": 567 }, { "epoch": 0.07, "learning_rate": 9.998139643404634e-06, "loss": 3.4392, "step": 568 }, { "epoch": 0.07, "learning_rate": 9.998084528227472e-06, "loss": 3.4516, "step": 569 }, { "epoch": 0.07, "learning_rate": 9.998028608656187e-06, "loss": 3.4034, "step": 570 }, { "epoch": 0.07, "learning_rate": 9.997971884699781e-06, "loss": 3.4538, "step": 571 }, { "epoch": 0.07, "learning_rate": 9.997914356367382e-06, "loss": 3.4552, "step": 572 }, { "epoch": 0.07, "learning_rate": 9.997856023668248e-06, "loss": 3.4438, "step": 573 }, { "epoch": 0.07, "learning_rate": 9.997796886611768e-06, "loss": 3.5106, "step": 574 }, { "epoch": 0.07, "learning_rate": 9.997736945207458e-06, "loss": 3.4467, "step": 575 }, { "epoch": 0.07, "learning_rate": 9.997676199464967e-06, "loss": 3.4771, "step": 576 }, { "epoch": 0.07, "learning_rate": 9.997614649394071e-06, "loss": 3.52, "step": 577 }, { "epoch": 0.07, "learning_rate": 9.997552295004677e-06, "loss": 3.4165, "step": 578 }, { "epoch": 0.07, "learning_rate": 9.997489136306817e-06, "loss": 3.4602, "step": 579 }, { "epoch": 0.07, "learning_rate": 9.99742517331066e-06, "loss": 3.3787, "step": 580 }, { "epoch": 0.07, "learning_rate": 9.997360406026496e-06, "loss": 3.3751, "step": 581 }, { "epoch": 0.07, "learning_rate": 9.99729483446475e-06, "loss": 3.4624, "step": 582 }, { "epoch": 0.07, "learning_rate": 9.997228458635979e-06, "loss": 3.491, "step": 583 }, { "epoch": 0.07, "learning_rate": 9.99716127855086e-06, "loss": 3.4861, "step": 584 }, { "epoch": 0.07, "learning_rate": 9.997093294220208e-06, "loss": 3.525, "step": 585 }, { "epoch": 0.07, "learning_rate": 9.997024505654964e-06, "loss": 3.6042, "step": 586 }, { "epoch": 0.07, "learning_rate": 9.996954912866198e-06, "loss": 3.3753, "step": 587 }, { "epoch": 0.07, "learning_rate": 9.99688451586511e-06, "loss": 3.3637, "step": 588 }, { "epoch": 0.07, "learning_rate": 9.996813314663032e-06, "loss": 3.4873, "step": 589 }, { "epoch": 0.07, "learning_rate": 9.996741309271423e-06, "loss": 3.3631, "step": 590 }, { "epoch": 0.07, "learning_rate": 9.996668499701869e-06, "loss": 3.4242, "step": 591 }, { "epoch": 0.07, "learning_rate": 9.99659488596609e-06, "loss": 3.3826, "step": 592 }, { "epoch": 0.07, "learning_rate": 9.996520468075932e-06, "loss": 3.5582, "step": 593 }, { "epoch": 0.07, "learning_rate": 9.996445246043374e-06, "loss": 3.4898, "step": 594 }, { "epoch": 0.07, "learning_rate": 9.99636921988052e-06, "loss": 3.4428, "step": 595 }, { "epoch": 0.07, "learning_rate": 9.996292389599607e-06, "loss": 3.5482, "step": 596 }, { "epoch": 0.07, "learning_rate": 9.996214755213e-06, "loss": 3.4353, "step": 597 }, { "epoch": 0.07, "learning_rate": 9.996136316733192e-06, "loss": 3.4686, "step": 598 }, { "epoch": 0.07, "learning_rate": 9.99605707417281e-06, "loss": 3.4287, "step": 599 }, { "epoch": 0.07, "learning_rate": 9.995977027544604e-06, "loss": 3.5917, "step": 600 }, { "epoch": 0.07, "learning_rate": 9.99589617686146e-06, "loss": 3.5075, "step": 601 }, { "epoch": 0.07, "learning_rate": 9.995814522136386e-06, "loss": 3.5584, "step": 602 }, { "epoch": 0.07, "learning_rate": 9.995732063382528e-06, "loss": 3.5054, "step": 603 }, { "epoch": 0.07, "learning_rate": 9.995648800613154e-06, "loss": 3.4851, "step": 604 }, { "epoch": 0.07, "learning_rate": 9.995564733841666e-06, "loss": 3.5241, "step": 605 }, { "epoch": 0.07, "learning_rate": 9.995479863081593e-06, "loss": 3.4384, "step": 606 }, { "epoch": 0.07, "learning_rate": 9.995394188346595e-06, "loss": 3.4273, "step": 607 }, { "epoch": 0.07, "learning_rate": 9.995307709650458e-06, "loss": 3.5336, "step": 608 }, { "epoch": 0.07, "learning_rate": 9.995220427007102e-06, "loss": 3.4933, "step": 609 }, { "epoch": 0.07, "learning_rate": 9.995132340430574e-06, "loss": 3.4403, "step": 610 }, { "epoch": 0.07, "learning_rate": 9.995043449935051e-06, "loss": 3.4101, "step": 611 }, { "epoch": 0.07, "learning_rate": 9.994953755534837e-06, "loss": 3.435, "step": 612 }, { "epoch": 0.07, "learning_rate": 9.994863257244372e-06, "loss": 3.5992, "step": 613 }, { "epoch": 0.07, "learning_rate": 9.994771955078217e-06, "loss": 3.4418, "step": 614 }, { "epoch": 0.07, "learning_rate": 9.994679849051066e-06, "loss": 3.3828, "step": 615 }, { "epoch": 0.07, "learning_rate": 9.994586939177746e-06, "loss": 3.4438, "step": 616 }, { "epoch": 0.07, "learning_rate": 9.994493225473206e-06, "loss": 3.4881, "step": 617 }, { "epoch": 0.07, "learning_rate": 9.994398707952531e-06, "loss": 3.3663, "step": 618 }, { "epoch": 0.07, "learning_rate": 9.994303386630932e-06, "loss": 3.442, "step": 619 }, { "epoch": 0.07, "learning_rate": 9.99420726152375e-06, "loss": 3.5021, "step": 620 }, { "epoch": 0.07, "learning_rate": 9.994110332646454e-06, "loss": 3.4118, "step": 621 }, { "epoch": 0.07, "learning_rate": 9.994012600014644e-06, "loss": 3.4401, "step": 622 }, { "epoch": 0.07, "learning_rate": 9.993914063644053e-06, "loss": 3.4301, "step": 623 }, { "epoch": 0.07, "learning_rate": 9.993814723550535e-06, "loss": 3.5186, "step": 624 }, { "epoch": 0.08, "learning_rate": 9.99371457975008e-06, "loss": 3.4405, "step": 625 }, { "epoch": 0.08, "learning_rate": 9.993613632258805e-06, "loss": 3.4908, "step": 626 }, { "epoch": 0.08, "learning_rate": 9.993511881092957e-06, "loss": 3.4395, "step": 627 }, { "epoch": 0.08, "learning_rate": 9.99340932626891e-06, "loss": 3.3978, "step": 628 }, { "epoch": 0.08, "learning_rate": 9.993305967803171e-06, "loss": 3.4746, "step": 629 }, { "epoch": 0.08, "learning_rate": 9.993201805712373e-06, "loss": 3.4579, "step": 630 }, { "epoch": 0.08, "learning_rate": 9.99309684001328e-06, "loss": 3.4667, "step": 631 }, { "epoch": 0.08, "learning_rate": 9.992991070722788e-06, "loss": 3.3887, "step": 632 }, { "epoch": 0.08, "learning_rate": 9.992884497857918e-06, "loss": 3.3392, "step": 633 }, { "epoch": 0.08, "learning_rate": 9.992777121435821e-06, "loss": 3.3875, "step": 634 }, { "epoch": 0.08, "learning_rate": 9.992668941473778e-06, "loss": 3.4775, "step": 635 }, { "epoch": 0.08, "learning_rate": 9.9925599579892e-06, "loss": 3.4304, "step": 636 }, { "epoch": 0.08, "learning_rate": 9.992450170999628e-06, "loss": 3.4319, "step": 637 }, { "epoch": 0.08, "learning_rate": 9.99233958052273e-06, "loss": 3.4422, "step": 638 }, { "epoch": 0.08, "learning_rate": 9.992228186576305e-06, "loss": 3.3656, "step": 639 }, { "epoch": 0.08, "learning_rate": 9.992115989178279e-06, "loss": 3.4814, "step": 640 }, { "epoch": 0.08, "learning_rate": 9.992002988346712e-06, "loss": 3.4983, "step": 641 }, { "epoch": 0.08, "learning_rate": 9.99188918409979e-06, "loss": 3.4085, "step": 642 }, { "epoch": 0.08, "learning_rate": 9.991774576455826e-06, "loss": 3.3864, "step": 643 }, { "epoch": 0.08, "learning_rate": 9.991659165433267e-06, "loss": 3.5092, "step": 644 }, { "epoch": 0.08, "learning_rate": 9.991542951050688e-06, "loss": 3.3644, "step": 645 }, { "epoch": 0.08, "learning_rate": 9.991425933326791e-06, "loss": 3.4301, "step": 646 }, { "epoch": 0.08, "learning_rate": 9.991308112280411e-06, "loss": 3.3915, "step": 647 }, { "epoch": 0.08, "learning_rate": 9.991189487930507e-06, "loss": 3.3573, "step": 648 }, { "epoch": 0.08, "learning_rate": 9.991070060296174e-06, "loss": 3.4251, "step": 649 }, { "epoch": 0.08, "learning_rate": 9.990949829396629e-06, "loss": 3.4875, "step": 650 }, { "epoch": 0.08, "learning_rate": 9.990828795251224e-06, "loss": 3.4147, "step": 651 }, { "epoch": 0.08, "learning_rate": 9.990706957879438e-06, "loss": 3.4461, "step": 652 }, { "epoch": 0.08, "learning_rate": 9.990584317300882e-06, "loss": 3.4609, "step": 653 }, { "epoch": 0.08, "learning_rate": 9.990460873535291e-06, "loss": 3.4152, "step": 654 }, { "epoch": 0.08, "learning_rate": 9.990336626602533e-06, "loss": 3.5364, "step": 655 }, { "epoch": 0.08, "learning_rate": 9.990211576522603e-06, "loss": 3.4495, "step": 656 }, { "epoch": 0.08, "learning_rate": 9.990085723315628e-06, "loss": 3.4601, "step": 657 }, { "epoch": 0.08, "learning_rate": 9.989959067001862e-06, "loss": 3.5332, "step": 658 }, { "epoch": 0.08, "learning_rate": 9.98983160760169e-06, "loss": 3.4866, "step": 659 }, { "epoch": 0.08, "learning_rate": 9.989703345135626e-06, "loss": 3.5069, "step": 660 }, { "epoch": 0.08, "learning_rate": 9.989574279624313e-06, "loss": 3.3763, "step": 661 }, { "epoch": 0.08, "learning_rate": 9.98944441108852e-06, "loss": 3.4202, "step": 662 }, { "epoch": 0.08, "learning_rate": 9.989313739549149e-06, "loss": 3.479, "step": 663 }, { "epoch": 0.08, "learning_rate": 9.989182265027232e-06, "loss": 3.3856, "step": 664 }, { "epoch": 0.08, "learning_rate": 9.989049987543927e-06, "loss": 3.3983, "step": 665 }, { "epoch": 0.08, "learning_rate": 9.988916907120525e-06, "loss": 3.4713, "step": 666 }, { "epoch": 0.08, "learning_rate": 9.988783023778441e-06, "loss": 3.5295, "step": 667 }, { "epoch": 0.08, "learning_rate": 9.988648337539224e-06, "loss": 3.4254, "step": 668 }, { "epoch": 0.08, "learning_rate": 9.98851284842455e-06, "loss": 3.4045, "step": 669 }, { "epoch": 0.08, "learning_rate": 9.988376556456225e-06, "loss": 3.4439, "step": 670 }, { "epoch": 0.08, "learning_rate": 9.988239461656184e-06, "loss": 3.435, "step": 671 }, { "epoch": 0.08, "learning_rate": 9.988101564046491e-06, "loss": 3.4802, "step": 672 }, { "epoch": 0.08, "learning_rate": 9.98796286364934e-06, "loss": 3.4666, "step": 673 }, { "epoch": 0.08, "learning_rate": 9.98782336048705e-06, "loss": 3.4836, "step": 674 }, { "epoch": 0.08, "learning_rate": 9.987683054582077e-06, "loss": 3.4769, "step": 675 }, { "epoch": 0.08, "learning_rate": 9.987541945957e-06, "loss": 3.3421, "step": 676 }, { "epoch": 0.08, "learning_rate": 9.987400034634529e-06, "loss": 3.3938, "step": 677 }, { "epoch": 0.08, "learning_rate": 9.987257320637504e-06, "loss": 3.4385, "step": 678 }, { "epoch": 0.08, "learning_rate": 9.987113803988893e-06, "loss": 3.554, "step": 679 }, { "epoch": 0.08, "learning_rate": 9.986969484711795e-06, "loss": 3.3837, "step": 680 }, { "epoch": 0.08, "learning_rate": 9.986824362829434e-06, "loss": 3.4452, "step": 681 }, { "epoch": 0.08, "learning_rate": 9.98667843836517e-06, "loss": 3.428, "step": 682 }, { "epoch": 0.08, "learning_rate": 9.986531711342482e-06, "loss": 3.5427, "step": 683 }, { "epoch": 0.08, "learning_rate": 9.986384181784991e-06, "loss": 3.478, "step": 684 }, { "epoch": 0.08, "learning_rate": 9.986235849716436e-06, "loss": 3.5645, "step": 685 }, { "epoch": 0.08, "learning_rate": 9.98608671516069e-06, "loss": 3.5222, "step": 686 }, { "epoch": 0.08, "learning_rate": 9.985936778141758e-06, "loss": 3.5204, "step": 687 }, { "epoch": 0.08, "learning_rate": 9.98578603868377e-06, "loss": 3.4526, "step": 688 }, { "epoch": 0.08, "learning_rate": 9.985634496810983e-06, "loss": 3.4123, "step": 689 }, { "epoch": 0.08, "learning_rate": 9.985482152547788e-06, "loss": 3.469, "step": 690 }, { "epoch": 0.08, "learning_rate": 9.985329005918702e-06, "loss": 3.3853, "step": 691 }, { "epoch": 0.08, "learning_rate": 9.985175056948377e-06, "loss": 3.4677, "step": 692 }, { "epoch": 0.08, "learning_rate": 9.985020305661586e-06, "loss": 3.4692, "step": 693 }, { "epoch": 0.08, "learning_rate": 9.984864752083234e-06, "loss": 3.4481, "step": 694 }, { "epoch": 0.08, "learning_rate": 9.984708396238358e-06, "loss": 3.5241, "step": 695 }, { "epoch": 0.08, "learning_rate": 9.98455123815212e-06, "loss": 3.4144, "step": 696 }, { "epoch": 0.08, "learning_rate": 9.984393277849817e-06, "loss": 3.3598, "step": 697 }, { "epoch": 0.08, "learning_rate": 9.984234515356868e-06, "loss": 3.4453, "step": 698 }, { "epoch": 0.08, "learning_rate": 9.984074950698823e-06, "loss": 3.39, "step": 699 }, { "epoch": 0.08, "learning_rate": 9.983914583901365e-06, "loss": 3.3988, "step": 700 }, { "epoch": 0.08, "learning_rate": 9.983753414990303e-06, "loss": 3.5772, "step": 701 }, { "epoch": 0.08, "learning_rate": 9.983591443991577e-06, "loss": 3.4938, "step": 702 }, { "epoch": 0.08, "learning_rate": 9.98342867093125e-06, "loss": 3.4249, "step": 703 }, { "epoch": 0.08, "learning_rate": 9.983265095835524e-06, "loss": 3.478, "step": 704 }, { "epoch": 0.08, "learning_rate": 9.98310071873072e-06, "loss": 3.4437, "step": 705 }, { "epoch": 0.08, "learning_rate": 9.9829355396433e-06, "loss": 3.4267, "step": 706 }, { "epoch": 0.08, "learning_rate": 9.98276955859984e-06, "loss": 3.4492, "step": 707 }, { "epoch": 0.08, "learning_rate": 9.982602775627057e-06, "loss": 3.423, "step": 708 }, { "epoch": 0.09, "learning_rate": 9.982435190751793e-06, "loss": 3.5102, "step": 709 }, { "epoch": 0.09, "learning_rate": 9.98226680400102e-06, "loss": 3.4072, "step": 710 }, { "epoch": 0.09, "learning_rate": 9.982097615401838e-06, "loss": 3.5397, "step": 711 }, { "epoch": 0.09, "learning_rate": 9.981927624981474e-06, "loss": 3.5282, "step": 712 }, { "epoch": 0.09, "learning_rate": 9.981756832767289e-06, "loss": 3.4802, "step": 713 }, { "epoch": 0.09, "learning_rate": 9.981585238786767e-06, "loss": 3.5205, "step": 714 }, { "epoch": 0.09, "learning_rate": 9.981412843067529e-06, "loss": 3.5384, "step": 715 }, { "epoch": 0.09, "learning_rate": 9.981239645637317e-06, "loss": 3.4437, "step": 716 }, { "epoch": 0.09, "learning_rate": 9.981065646524007e-06, "loss": 3.4172, "step": 717 }, { "epoch": 0.09, "learning_rate": 9.980890845755602e-06, "loss": 3.4227, "step": 718 }, { "epoch": 0.09, "learning_rate": 9.980715243360235e-06, "loss": 3.4599, "step": 719 }, { "epoch": 0.09, "learning_rate": 9.980538839366167e-06, "loss": 3.3609, "step": 720 }, { "epoch": 0.09, "learning_rate": 9.980361633801788e-06, "loss": 3.3852, "step": 721 }, { "epoch": 0.09, "learning_rate": 9.980183626695618e-06, "loss": 3.4254, "step": 722 }, { "epoch": 0.09, "learning_rate": 9.980004818076308e-06, "loss": 3.4605, "step": 723 }, { "epoch": 0.09, "learning_rate": 9.97982520797263e-06, "loss": 3.4623, "step": 724 }, { "epoch": 0.09, "learning_rate": 9.979644796413496e-06, "loss": 3.5022, "step": 725 }, { "epoch": 0.09, "learning_rate": 9.97946358342794e-06, "loss": 3.4292, "step": 726 }, { "epoch": 0.09, "learning_rate": 9.979281569045126e-06, "loss": 3.4305, "step": 727 }, { "epoch": 0.09, "learning_rate": 9.979098753294346e-06, "loss": 3.4733, "step": 728 }, { "epoch": 0.09, "learning_rate": 9.978915136205025e-06, "loss": 3.5243, "step": 729 }, { "epoch": 0.09, "learning_rate": 9.978730717806713e-06, "loss": 3.2954, "step": 730 }, { "epoch": 0.09, "learning_rate": 9.97854549812909e-06, "loss": 3.441, "step": 731 }, { "epoch": 0.09, "learning_rate": 9.978359477201966e-06, "loss": 3.5624, "step": 732 }, { "epoch": 0.09, "learning_rate": 9.97817265505528e-06, "loss": 3.3732, "step": 733 }, { "epoch": 0.09, "learning_rate": 9.977985031719098e-06, "loss": 3.4079, "step": 734 }, { "epoch": 0.09, "learning_rate": 9.977796607223617e-06, "loss": 3.4818, "step": 735 }, { "epoch": 0.09, "learning_rate": 9.977607381599161e-06, "loss": 3.4067, "step": 736 }, { "epoch": 0.09, "learning_rate": 9.977417354876185e-06, "loss": 3.391, "step": 737 }, { "epoch": 0.09, "learning_rate": 9.977226527085273e-06, "loss": 3.4217, "step": 738 }, { "epoch": 0.09, "learning_rate": 9.977034898257137e-06, "loss": 3.3515, "step": 739 }, { "epoch": 0.09, "learning_rate": 9.976842468422614e-06, "loss": 3.4614, "step": 740 }, { "epoch": 0.09, "learning_rate": 9.976649237612677e-06, "loss": 3.4137, "step": 741 }, { "epoch": 0.09, "learning_rate": 9.976455205858424e-06, "loss": 3.4088, "step": 742 }, { "epoch": 0.09, "learning_rate": 9.976260373191085e-06, "loss": 3.3645, "step": 743 }, { "epoch": 0.09, "learning_rate": 9.976064739642011e-06, "loss": 3.4983, "step": 744 }, { "epoch": 0.09, "learning_rate": 9.975868305242691e-06, "loss": 3.5366, "step": 745 }, { "epoch": 0.09, "learning_rate": 9.975671070024741e-06, "loss": 3.4222, "step": 746 }, { "epoch": 0.09, "learning_rate": 9.9754730340199e-06, "loss": 3.3174, "step": 747 }, { "epoch": 0.09, "learning_rate": 9.975274197260041e-06, "loss": 3.4301, "step": 748 }, { "epoch": 0.09, "learning_rate": 9.975074559777167e-06, "loss": 3.5166, "step": 749 }, { "epoch": 0.09, "learning_rate": 9.974874121603408e-06, "loss": 3.4276, "step": 750 }, { "epoch": 0.09, "learning_rate": 9.974672882771019e-06, "loss": 3.5098, "step": 751 }, { "epoch": 0.09, "learning_rate": 9.97447084331239e-06, "loss": 3.5347, "step": 752 }, { "epoch": 0.09, "learning_rate": 9.974268003260036e-06, "loss": 3.3729, "step": 753 }, { "epoch": 0.09, "learning_rate": 9.974064362646606e-06, "loss": 3.5074, "step": 754 }, { "epoch": 0.09, "learning_rate": 9.973859921504867e-06, "loss": 3.4829, "step": 755 }, { "epoch": 0.09, "learning_rate": 9.973654679867729e-06, "loss": 3.4514, "step": 756 }, { "epoch": 0.09, "learning_rate": 9.97344863776822e-06, "loss": 3.3782, "step": 757 }, { "epoch": 0.09, "learning_rate": 9.973241795239502e-06, "loss": 3.4262, "step": 758 }, { "epoch": 0.09, "learning_rate": 9.97303415231486e-06, "loss": 3.4613, "step": 759 }, { "epoch": 0.09, "learning_rate": 9.972825709027718e-06, "loss": 3.5037, "step": 760 }, { "epoch": 0.09, "learning_rate": 9.972616465411622e-06, "loss": 3.5193, "step": 761 }, { "epoch": 0.09, "learning_rate": 9.972406421500245e-06, "loss": 3.4894, "step": 762 }, { "epoch": 0.09, "learning_rate": 9.972195577327393e-06, "loss": 3.4514, "step": 763 }, { "epoch": 0.09, "learning_rate": 9.971983932926998e-06, "loss": 3.4331, "step": 764 }, { "epoch": 0.09, "learning_rate": 9.971771488333124e-06, "loss": 3.4367, "step": 765 }, { "epoch": 0.09, "learning_rate": 9.971558243579963e-06, "loss": 3.476, "step": 766 }, { "epoch": 0.09, "learning_rate": 9.971344198701831e-06, "loss": 3.4599, "step": 767 }, { "epoch": 0.09, "learning_rate": 9.97112935373318e-06, "loss": 3.5025, "step": 768 }, { "epoch": 0.09, "learning_rate": 9.970913708708586e-06, "loss": 3.4162, "step": 769 }, { "epoch": 0.09, "learning_rate": 9.970697263662753e-06, "loss": 3.4575, "step": 770 }, { "epoch": 0.09, "learning_rate": 9.970480018630521e-06, "loss": 3.4778, "step": 771 }, { "epoch": 0.09, "learning_rate": 9.970261973646848e-06, "loss": 3.4061, "step": 772 }, { "epoch": 0.09, "learning_rate": 9.970043128746828e-06, "loss": 3.4887, "step": 773 }, { "epoch": 0.09, "learning_rate": 9.969823483965683e-06, "loss": 3.3403, "step": 774 }, { "epoch": 0.09, "learning_rate": 9.969603039338761e-06, "loss": 3.4937, "step": 775 }, { "epoch": 0.09, "learning_rate": 9.969381794901543e-06, "loss": 3.4779, "step": 776 }, { "epoch": 0.09, "learning_rate": 9.969159750689635e-06, "loss": 3.3699, "step": 777 }, { "epoch": 0.09, "learning_rate": 9.968936906738774e-06, "loss": 3.4455, "step": 778 }, { "epoch": 0.09, "learning_rate": 9.968713263084819e-06, "loss": 3.5141, "step": 779 }, { "epoch": 0.09, "learning_rate": 9.96848881976377e-06, "loss": 3.5021, "step": 780 }, { "epoch": 0.09, "learning_rate": 9.968263576811747e-06, "loss": 3.4918, "step": 781 }, { "epoch": 0.09, "learning_rate": 9.968037534265001e-06, "loss": 3.4499, "step": 782 }, { "epoch": 0.09, "learning_rate": 9.967810692159909e-06, "loss": 3.5986, "step": 783 }, { "epoch": 0.09, "learning_rate": 9.967583050532982e-06, "loss": 3.4913, "step": 784 }, { "epoch": 0.09, "learning_rate": 9.967354609420856e-06, "loss": 3.4074, "step": 785 }, { "epoch": 0.09, "learning_rate": 9.967125368860295e-06, "loss": 3.4616, "step": 786 }, { "epoch": 0.09, "learning_rate": 9.966895328888195e-06, "loss": 3.3483, "step": 787 }, { "epoch": 0.09, "learning_rate": 9.966664489541576e-06, "loss": 3.4383, "step": 788 }, { "epoch": 0.09, "learning_rate": 9.966432850857593e-06, "loss": 3.4148, "step": 789 }, { "epoch": 0.09, "learning_rate": 9.966200412873523e-06, "loss": 3.54, "step": 790 }, { "epoch": 0.09, "learning_rate": 9.965967175626776e-06, "loss": 3.4446, "step": 791 }, { "epoch": 0.1, "learning_rate": 9.965733139154891e-06, "loss": 3.3735, "step": 792 }, { "epoch": 0.1, "learning_rate": 9.965498303495531e-06, "loss": 3.4771, "step": 793 }, { "epoch": 0.1, "learning_rate": 9.965262668686491e-06, "loss": 3.4271, "step": 794 }, { "epoch": 0.1, "learning_rate": 9.965026234765696e-06, "loss": 3.3488, "step": 795 }, { "epoch": 0.1, "learning_rate": 9.964789001771196e-06, "loss": 3.4353, "step": 796 }, { "epoch": 0.1, "learning_rate": 9.964550969741173e-06, "loss": 3.5156, "step": 797 }, { "epoch": 0.1, "learning_rate": 9.964312138713936e-06, "loss": 3.5993, "step": 798 }, { "epoch": 0.1, "learning_rate": 9.96407250872792e-06, "loss": 3.4657, "step": 799 }, { "epoch": 0.1, "learning_rate": 9.963832079821693e-06, "loss": 3.4349, "step": 800 }, { "epoch": 0.1, "learning_rate": 9.96359085203395e-06, "loss": 3.4237, "step": 801 }, { "epoch": 0.1, "learning_rate": 9.963348825403513e-06, "loss": 3.4044, "step": 802 }, { "epoch": 0.1, "learning_rate": 9.963105999969335e-06, "loss": 3.437, "step": 803 }, { "epoch": 0.1, "learning_rate": 9.962862375770497e-06, "loss": 3.4212, "step": 804 }, { "epoch": 0.1, "learning_rate": 9.962617952846207e-06, "loss": 3.4362, "step": 805 }, { "epoch": 0.1, "learning_rate": 9.962372731235803e-06, "loss": 3.4234, "step": 806 }, { "epoch": 0.1, "learning_rate": 9.962126710978751e-06, "loss": 3.5751, "step": 807 }, { "epoch": 0.1, "learning_rate": 9.961879892114646e-06, "loss": 3.4617, "step": 808 }, { "epoch": 0.1, "learning_rate": 9.961632274683209e-06, "loss": 3.5157, "step": 809 }, { "epoch": 0.1, "learning_rate": 9.961383858724293e-06, "loss": 3.4391, "step": 810 }, { "epoch": 0.1, "learning_rate": 9.961134644277881e-06, "loss": 3.5409, "step": 811 }, { "epoch": 0.1, "learning_rate": 9.960884631384077e-06, "loss": 3.444, "step": 812 }, { "epoch": 0.1, "learning_rate": 9.960633820083121e-06, "loss": 3.4221, "step": 813 }, { "epoch": 0.1, "learning_rate": 9.960382210415377e-06, "loss": 3.5115, "step": 814 }, { "epoch": 0.1, "learning_rate": 9.960129802421344e-06, "loss": 3.4743, "step": 815 }, { "epoch": 0.1, "learning_rate": 9.959876596141638e-06, "loss": 3.3875, "step": 816 }, { "epoch": 0.1, "learning_rate": 9.959622591617013e-06, "loss": 3.4406, "step": 817 }, { "epoch": 0.1, "learning_rate": 9.959367788888348e-06, "loss": 3.4141, "step": 818 }, { "epoch": 0.1, "learning_rate": 9.959112187996654e-06, "loss": 3.4052, "step": 819 }, { "epoch": 0.1, "learning_rate": 9.958855788983065e-06, "loss": 3.4921, "step": 820 }, { "epoch": 0.1, "learning_rate": 9.958598591888844e-06, "loss": 3.4448, "step": 821 }, { "epoch": 0.1, "learning_rate": 9.958340596755388e-06, "loss": 3.4889, "step": 822 }, { "epoch": 0.1, "learning_rate": 9.958081803624216e-06, "loss": 3.5628, "step": 823 }, { "epoch": 0.1, "learning_rate": 9.957822212536984e-06, "loss": 3.4322, "step": 824 }, { "epoch": 0.1, "learning_rate": 9.957561823535461e-06, "loss": 3.4305, "step": 825 }, { "epoch": 0.1, "learning_rate": 9.957300636661563e-06, "loss": 3.4334, "step": 826 }, { "epoch": 0.1, "learning_rate": 9.957038651957321e-06, "loss": 3.422, "step": 827 }, { "epoch": 0.1, "learning_rate": 9.956775869464901e-06, "loss": 3.4525, "step": 828 }, { "epoch": 0.1, "learning_rate": 9.956512289226594e-06, "loss": 3.3199, "step": 829 }, { "epoch": 0.1, "learning_rate": 9.956247911284823e-06, "loss": 3.4254, "step": 830 }, { "epoch": 0.1, "learning_rate": 9.955982735682133e-06, "loss": 3.4837, "step": 831 }, { "epoch": 0.1, "learning_rate": 9.955716762461204e-06, "loss": 3.4588, "step": 832 }, { "epoch": 0.1, "learning_rate": 9.955449991664843e-06, "loss": 3.5319, "step": 833 }, { "epoch": 0.1, "learning_rate": 9.955182423335981e-06, "loss": 3.3925, "step": 834 }, { "epoch": 0.1, "learning_rate": 9.954914057517684e-06, "loss": 3.4798, "step": 835 }, { "epoch": 0.1, "learning_rate": 9.95464489425314e-06, "loss": 3.5225, "step": 836 }, { "epoch": 0.1, "learning_rate": 9.95437493358567e-06, "loss": 3.4596, "step": 837 }, { "epoch": 0.1, "learning_rate": 9.954104175558723e-06, "loss": 3.482, "step": 838 }, { "epoch": 0.1, "learning_rate": 9.953832620215873e-06, "loss": 3.4219, "step": 839 }, { "epoch": 0.1, "learning_rate": 9.953560267600823e-06, "loss": 3.5276, "step": 840 }, { "epoch": 0.1, "learning_rate": 9.953287117757408e-06, "loss": 3.4084, "step": 841 }, { "epoch": 0.1, "learning_rate": 9.953013170729587e-06, "loss": 3.4327, "step": 842 }, { "epoch": 0.1, "learning_rate": 9.95273842656145e-06, "loss": 3.5208, "step": 843 }, { "epoch": 0.1, "learning_rate": 9.952462885297214e-06, "loss": 3.4022, "step": 844 }, { "epoch": 0.1, "learning_rate": 9.952186546981227e-06, "loss": 3.4173, "step": 845 }, { "epoch": 0.1, "learning_rate": 9.95190941165796e-06, "loss": 3.3618, "step": 846 }, { "epoch": 0.1, "learning_rate": 9.951631479372018e-06, "loss": 3.4838, "step": 847 }, { "epoch": 0.1, "learning_rate": 9.951352750168128e-06, "loss": 3.4071, "step": 848 }, { "epoch": 0.1, "learning_rate": 9.951073224091152e-06, "loss": 3.4748, "step": 849 }, { "epoch": 0.1, "learning_rate": 9.950792901186075e-06, "loss": 3.4162, "step": 850 }, { "epoch": 0.1, "learning_rate": 9.950511781498014e-06, "loss": 3.4141, "step": 851 }, { "epoch": 0.1, "learning_rate": 9.950229865072212e-06, "loss": 3.4689, "step": 852 }, { "epoch": 0.1, "learning_rate": 9.949947151954041e-06, "loss": 3.4294, "step": 853 }, { "epoch": 0.1, "learning_rate": 9.949663642189e-06, "loss": 3.4935, "step": 854 }, { "epoch": 0.1, "learning_rate": 9.949379335822718e-06, "loss": 3.5258, "step": 855 }, { "epoch": 0.1, "learning_rate": 9.94909423290095e-06, "loss": 3.4307, "step": 856 }, { "epoch": 0.1, "learning_rate": 9.948808333469582e-06, "loss": 3.366, "step": 857 }, { "epoch": 0.1, "learning_rate": 9.948521637574629e-06, "loss": 3.5176, "step": 858 }, { "epoch": 0.1, "learning_rate": 9.948234145262227e-06, "loss": 3.3777, "step": 859 }, { "epoch": 0.1, "learning_rate": 9.947945856578648e-06, "loss": 3.4782, "step": 860 }, { "epoch": 0.1, "learning_rate": 9.947656771570288e-06, "loss": 3.3909, "step": 861 }, { "epoch": 0.1, "learning_rate": 9.947366890283675e-06, "loss": 3.4431, "step": 862 }, { "epoch": 0.1, "learning_rate": 9.947076212765462e-06, "loss": 3.4192, "step": 863 }, { "epoch": 0.1, "learning_rate": 9.946784739062429e-06, "loss": 3.5476, "step": 864 }, { "epoch": 0.1, "learning_rate": 9.946492469221486e-06, "loss": 3.4498, "step": 865 }, { "epoch": 0.1, "learning_rate": 9.946199403289673e-06, "loss": 3.3686, "step": 866 }, { "epoch": 0.1, "learning_rate": 9.945905541314153e-06, "loss": 3.4932, "step": 867 }, { "epoch": 0.1, "learning_rate": 9.945610883342225e-06, "loss": 3.5045, "step": 868 }, { "epoch": 0.1, "learning_rate": 9.945315429421307e-06, "loss": 3.3892, "step": 869 }, { "epoch": 0.1, "learning_rate": 9.945019179598951e-06, "loss": 3.4815, "step": 870 }, { "epoch": 0.1, "learning_rate": 9.944722133922838e-06, "loss": 3.4265, "step": 871 }, { "epoch": 0.1, "learning_rate": 9.944424292440771e-06, "loss": 3.4695, "step": 872 }, { "epoch": 0.1, "learning_rate": 9.944125655200685e-06, "loss": 3.4305, "step": 873 }, { "epoch": 0.1, "learning_rate": 9.943826222250645e-06, "loss": 3.4663, "step": 874 }, { "epoch": 0.11, "learning_rate": 9.943525993638842e-06, "loss": 3.3873, "step": 875 }, { "epoch": 0.11, "learning_rate": 9.943224969413591e-06, "loss": 3.4435, "step": 876 }, { "epoch": 0.11, "learning_rate": 9.942923149623344e-06, "loss": 3.4128, "step": 877 }, { "epoch": 0.11, "learning_rate": 9.942620534316676e-06, "loss": 3.4427, "step": 878 }, { "epoch": 0.11, "learning_rate": 9.942317123542284e-06, "loss": 3.3928, "step": 879 }, { "epoch": 0.11, "learning_rate": 9.942012917349005e-06, "loss": 3.3585, "step": 880 }, { "epoch": 0.11, "learning_rate": 9.941707915785796e-06, "loss": 3.4509, "step": 881 }, { "epoch": 0.11, "learning_rate": 9.941402118901743e-06, "loss": 3.5779, "step": 882 }, { "epoch": 0.11, "learning_rate": 9.941095526746063e-06, "loss": 3.4985, "step": 883 }, { "epoch": 0.11, "learning_rate": 9.940788139368098e-06, "loss": 3.4742, "step": 884 }, { "epoch": 0.11, "learning_rate": 9.940479956817321e-06, "loss": 3.4746, "step": 885 }, { "epoch": 0.11, "learning_rate": 9.940170979143329e-06, "loss": 3.4864, "step": 886 }, { "epoch": 0.11, "learning_rate": 9.939861206395849e-06, "loss": 3.4257, "step": 887 }, { "epoch": 0.11, "learning_rate": 9.939550638624735e-06, "loss": 3.41, "step": 888 }, { "epoch": 0.11, "learning_rate": 9.939239275879972e-06, "loss": 3.4895, "step": 889 }, { "epoch": 0.11, "learning_rate": 9.938927118211672e-06, "loss": 3.3722, "step": 890 }, { "epoch": 0.11, "learning_rate": 9.938614165670071e-06, "loss": 3.45, "step": 891 }, { "epoch": 0.11, "learning_rate": 9.938300418305536e-06, "loss": 3.4062, "step": 892 }, { "epoch": 0.11, "learning_rate": 9.937985876168563e-06, "loss": 3.4924, "step": 893 }, { "epoch": 0.11, "learning_rate": 9.937670539309775e-06, "loss": 3.4987, "step": 894 }, { "epoch": 0.11, "learning_rate": 9.937354407779921e-06, "loss": 3.4442, "step": 895 }, { "epoch": 0.11, "learning_rate": 9.93703748162988e-06, "loss": 3.3911, "step": 896 }, { "epoch": 0.11, "learning_rate": 9.936719760910658e-06, "loss": 3.4257, "step": 897 }, { "epoch": 0.11, "learning_rate": 9.936401245673388e-06, "loss": 3.462, "step": 898 }, { "epoch": 0.11, "learning_rate": 9.936081935969334e-06, "loss": 3.4754, "step": 899 }, { "epoch": 0.11, "learning_rate": 9.935761831849886e-06, "loss": 3.4306, "step": 900 }, { "epoch": 0.11, "learning_rate": 9.935440933366561e-06, "loss": 3.4302, "step": 901 }, { "epoch": 0.11, "learning_rate": 9.935119240571002e-06, "loss": 3.4034, "step": 902 }, { "epoch": 0.11, "learning_rate": 9.934796753514986e-06, "loss": 3.3441, "step": 903 }, { "epoch": 0.11, "learning_rate": 9.934473472250412e-06, "loss": 3.4484, "step": 904 }, { "epoch": 0.11, "learning_rate": 9.93414939682931e-06, "loss": 3.4961, "step": 905 }, { "epoch": 0.11, "learning_rate": 9.93382452730384e-06, "loss": 3.4001, "step": 906 }, { "epoch": 0.11, "learning_rate": 9.933498863726281e-06, "loss": 3.4281, "step": 907 }, { "epoch": 0.11, "learning_rate": 9.933172406149047e-06, "loss": 3.3639, "step": 908 }, { "epoch": 0.11, "learning_rate": 9.932845154624681e-06, "loss": 3.447, "step": 909 }, { "epoch": 0.11, "learning_rate": 9.932517109205849e-06, "loss": 3.5351, "step": 910 }, { "epoch": 0.11, "learning_rate": 9.932188269945346e-06, "loss": 3.5511, "step": 911 }, { "epoch": 0.11, "learning_rate": 9.931858636896098e-06, "loss": 3.3871, "step": 912 }, { "epoch": 0.11, "learning_rate": 9.931528210111154e-06, "loss": 3.4336, "step": 913 }, { "epoch": 0.11, "learning_rate": 9.931196989643695e-06, "loss": 3.4688, "step": 914 }, { "epoch": 0.11, "learning_rate": 9.930864975547024e-06, "loss": 3.4945, "step": 915 }, { "epoch": 0.11, "learning_rate": 9.93053216787458e-06, "loss": 3.516, "step": 916 }, { "epoch": 0.11, "learning_rate": 9.930198566679922e-06, "loss": 3.3887, "step": 917 }, { "epoch": 0.11, "learning_rate": 9.929864172016743e-06, "loss": 3.4984, "step": 918 }, { "epoch": 0.11, "learning_rate": 9.929528983938859e-06, "loss": 3.4366, "step": 919 }, { "epoch": 0.11, "learning_rate": 9.929193002500215e-06, "loss": 3.4802, "step": 920 }, { "epoch": 0.11, "learning_rate": 9.928856227754885e-06, "loss": 3.3441, "step": 921 }, { "epoch": 0.11, "learning_rate": 9.928518659757067e-06, "loss": 3.4596, "step": 922 }, { "epoch": 0.11, "learning_rate": 9.928180298561094e-06, "loss": 3.4514, "step": 923 }, { "epoch": 0.11, "learning_rate": 9.927841144221417e-06, "loss": 3.3914, "step": 924 }, { "epoch": 0.11, "learning_rate": 9.927501196792622e-06, "loss": 3.5244, "step": 925 }, { "epoch": 0.11, "learning_rate": 9.927160456329421e-06, "loss": 3.4104, "step": 926 }, { "epoch": 0.11, "learning_rate": 9.926818922886654e-06, "loss": 3.3745, "step": 927 }, { "epoch": 0.11, "learning_rate": 9.926476596519284e-06, "loss": 3.4546, "step": 928 }, { "epoch": 0.11, "learning_rate": 9.926133477282407e-06, "loss": 3.504, "step": 929 }, { "epoch": 0.11, "learning_rate": 9.925789565231246e-06, "loss": 3.4029, "step": 930 }, { "epoch": 0.11, "learning_rate": 9.925444860421147e-06, "loss": 3.4876, "step": 931 }, { "epoch": 0.11, "learning_rate": 9.925099362907592e-06, "loss": 3.4437, "step": 932 }, { "epoch": 0.11, "learning_rate": 9.924753072746181e-06, "loss": 3.5277, "step": 933 }, { "epoch": 0.11, "learning_rate": 9.924405989992647e-06, "loss": 3.3498, "step": 934 }, { "epoch": 0.11, "learning_rate": 9.924058114702851e-06, "loss": 3.4051, "step": 935 }, { "epoch": 0.11, "learning_rate": 9.923709446932782e-06, "loss": 3.5114, "step": 936 }, { "epoch": 0.11, "learning_rate": 9.923359986738548e-06, "loss": 3.3674, "step": 937 }, { "epoch": 0.11, "learning_rate": 9.9230097341764e-06, "loss": 3.3954, "step": 938 }, { "epoch": 0.11, "learning_rate": 9.9226586893027e-06, "loss": 3.4838, "step": 939 }, { "epoch": 0.11, "learning_rate": 9.92230685217395e-06, "loss": 3.3248, "step": 940 }, { "epoch": 0.11, "learning_rate": 9.921954222846774e-06, "loss": 3.5396, "step": 941 }, { "epoch": 0.11, "learning_rate": 9.921600801377924e-06, "loss": 3.4847, "step": 942 }, { "epoch": 0.11, "learning_rate": 9.92124658782428e-06, "loss": 3.444, "step": 943 }, { "epoch": 0.11, "learning_rate": 9.920891582242849e-06, "loss": 3.3854, "step": 944 }, { "epoch": 0.11, "learning_rate": 9.920535784690765e-06, "loss": 3.4386, "step": 945 }, { "epoch": 0.11, "learning_rate": 9.920179195225292e-06, "loss": 3.449, "step": 946 }, { "epoch": 0.11, "learning_rate": 9.919821813903819e-06, "loss": 3.5078, "step": 947 }, { "epoch": 0.11, "learning_rate": 9.919463640783861e-06, "loss": 3.4428, "step": 948 }, { "epoch": 0.11, "learning_rate": 9.919104675923068e-06, "loss": 3.5067, "step": 949 }, { "epoch": 0.11, "learning_rate": 9.918744919379206e-06, "loss": 3.3985, "step": 950 }, { "epoch": 0.11, "learning_rate": 9.918384371210178e-06, "loss": 3.4598, "step": 951 }, { "epoch": 0.11, "learning_rate": 9.918023031474008e-06, "loss": 3.4229, "step": 952 }, { "epoch": 0.11, "learning_rate": 9.917660900228853e-06, "loss": 3.4775, "step": 953 }, { "epoch": 0.11, "learning_rate": 9.917297977532992e-06, "loss": 3.4984, "step": 954 }, { "epoch": 0.11, "learning_rate": 9.916934263444837e-06, "loss": 3.4899, "step": 955 }, { "epoch": 0.11, "learning_rate": 9.916569758022923e-06, "loss": 3.5023, "step": 956 }, { "epoch": 0.11, "learning_rate": 9.916204461325911e-06, "loss": 3.4004, "step": 957 }, { "epoch": 0.11, "learning_rate": 9.915838373412596e-06, "loss": 3.4825, "step": 958 }, { "epoch": 0.12, "learning_rate": 9.915471494341895e-06, "loss": 3.4019, "step": 959 }, { "epoch": 0.12, "learning_rate": 9.915103824172854e-06, "loss": 3.4219, "step": 960 }, { "epoch": 0.12, "learning_rate": 9.914735362964644e-06, "loss": 3.4299, "step": 961 }, { "epoch": 0.12, "learning_rate": 9.914366110776568e-06, "loss": 3.518, "step": 962 }, { "epoch": 0.12, "learning_rate": 9.91399606766805e-06, "loss": 3.4381, "step": 963 }, { "epoch": 0.12, "learning_rate": 9.91362523369865e-06, "loss": 3.4519, "step": 964 }, { "epoch": 0.12, "learning_rate": 9.913253608928046e-06, "loss": 3.4805, "step": 965 }, { "epoch": 0.12, "learning_rate": 9.91288119341605e-06, "loss": 3.427, "step": 966 }, { "epoch": 0.12, "learning_rate": 9.912507987222597e-06, "loss": 3.554, "step": 967 }, { "epoch": 0.12, "learning_rate": 9.912133990407751e-06, "loss": 3.4269, "step": 968 }, { "epoch": 0.12, "learning_rate": 9.911759203031704e-06, "loss": 3.4521, "step": 969 }, { "epoch": 0.12, "learning_rate": 9.911383625154776e-06, "loss": 3.5242, "step": 970 }, { "epoch": 0.12, "learning_rate": 9.911007256837411e-06, "loss": 3.3804, "step": 971 }, { "epoch": 0.12, "learning_rate": 9.91063009814018e-06, "loss": 3.423, "step": 972 }, { "epoch": 0.12, "learning_rate": 9.910252149123787e-06, "loss": 3.349, "step": 973 }, { "epoch": 0.12, "learning_rate": 9.909873409849056e-06, "loss": 3.4996, "step": 974 }, { "epoch": 0.12, "learning_rate": 9.909493880376943e-06, "loss": 3.4245, "step": 975 }, { "epoch": 0.12, "learning_rate": 9.909113560768529e-06, "loss": 3.3944, "step": 976 }, { "epoch": 0.12, "learning_rate": 9.908732451085024e-06, "loss": 3.4954, "step": 977 }, { "epoch": 0.12, "learning_rate": 9.908350551387764e-06, "loss": 3.4718, "step": 978 }, { "epoch": 0.12, "learning_rate": 9.90796786173821e-06, "loss": 3.4763, "step": 979 }, { "epoch": 0.12, "learning_rate": 9.907584382197955e-06, "loss": 3.3114, "step": 980 }, { "epoch": 0.12, "learning_rate": 9.907200112828714e-06, "loss": 3.4029, "step": 981 }, { "epoch": 0.12, "learning_rate": 9.906815053692333e-06, "loss": 3.4676, "step": 982 }, { "epoch": 0.12, "learning_rate": 9.906429204850785e-06, "loss": 3.5097, "step": 983 }, { "epoch": 0.12, "learning_rate": 9.906042566366164e-06, "loss": 3.4848, "step": 984 }, { "epoch": 0.12, "learning_rate": 9.9056551383007e-06, "loss": 3.3227, "step": 985 }, { "epoch": 0.12, "learning_rate": 9.905266920716743e-06, "loss": 3.359, "step": 986 }, { "epoch": 0.12, "learning_rate": 9.904877913676776e-06, "loss": 3.4505, "step": 987 }, { "epoch": 0.12, "learning_rate": 9.904488117243406e-06, "loss": 3.4386, "step": 988 }, { "epoch": 0.12, "learning_rate": 9.904097531479362e-06, "loss": 3.4862, "step": 989 }, { "epoch": 0.12, "learning_rate": 9.90370615644751e-06, "loss": 3.526, "step": 990 }, { "epoch": 0.12, "learning_rate": 9.903313992210836e-06, "loss": 3.4112, "step": 991 }, { "epoch": 0.12, "learning_rate": 9.902921038832456e-06, "loss": 3.4301, "step": 992 }, { "epoch": 0.12, "learning_rate": 9.902527296375611e-06, "loss": 3.4274, "step": 993 }, { "epoch": 0.12, "learning_rate": 9.902132764903673e-06, "loss": 3.4448, "step": 994 }, { "epoch": 0.12, "learning_rate": 9.901737444480134e-06, "loss": 3.4578, "step": 995 }, { "epoch": 0.12, "learning_rate": 9.90134133516862e-06, "loss": 3.4276, "step": 996 }, { "epoch": 0.12, "learning_rate": 9.900944437032879e-06, "loss": 3.4956, "step": 997 }, { "epoch": 0.12, "learning_rate": 9.90054675013679e-06, "loss": 3.4182, "step": 998 }, { "epoch": 0.12, "learning_rate": 9.900148274544354e-06, "loss": 3.3935, "step": 999 }, { "epoch": 0.12, "learning_rate": 9.899749010319706e-06, "loss": 3.4642, "step": 1000 }, { "epoch": 0.12, "learning_rate": 9.899348957527103e-06, "loss": 3.5304, "step": 1001 }, { "epoch": 0.12, "learning_rate": 9.898948116230924e-06, "loss": 3.3896, "step": 1002 }, { "epoch": 0.12, "learning_rate": 9.898546486495687e-06, "loss": 3.2808, "step": 1003 }, { "epoch": 0.12, "learning_rate": 9.898144068386028e-06, "loss": 3.4702, "step": 1004 }, { "epoch": 0.12, "learning_rate": 9.897740861966714e-06, "loss": 3.432, "step": 1005 }, { "epoch": 0.12, "learning_rate": 9.897336867302633e-06, "loss": 3.5095, "step": 1006 }, { "epoch": 0.12, "learning_rate": 9.89693208445881e-06, "loss": 3.3948, "step": 1007 }, { "epoch": 0.12, "learning_rate": 9.896526513500386e-06, "loss": 3.4448, "step": 1008 }, { "epoch": 0.12, "learning_rate": 9.896120154492636e-06, "loss": 3.5026, "step": 1009 }, { "epoch": 0.12, "learning_rate": 9.895713007500961e-06, "loss": 3.414, "step": 1010 }, { "epoch": 0.12, "learning_rate": 9.895305072590883e-06, "loss": 3.387, "step": 1011 }, { "epoch": 0.12, "learning_rate": 9.894896349828061e-06, "loss": 3.5151, "step": 1012 }, { "epoch": 0.12, "learning_rate": 9.89448683927827e-06, "loss": 3.4917, "step": 1013 }, { "epoch": 0.12, "learning_rate": 9.894076541007421e-06, "loss": 3.4035, "step": 1014 }, { "epoch": 0.12, "learning_rate": 9.893665455081543e-06, "loss": 3.5048, "step": 1015 }, { "epoch": 0.12, "learning_rate": 9.893253581566801e-06, "loss": 3.4407, "step": 1016 }, { "epoch": 0.12, "learning_rate": 9.892840920529478e-06, "loss": 3.3672, "step": 1017 }, { "epoch": 0.12, "learning_rate": 9.892427472035992e-06, "loss": 3.3761, "step": 1018 }, { "epoch": 0.12, "learning_rate": 9.89201323615288e-06, "loss": 3.3841, "step": 1019 }, { "epoch": 0.12, "learning_rate": 9.89159821294681e-06, "loss": 3.4435, "step": 1020 }, { "epoch": 0.12, "learning_rate": 9.891182402484578e-06, "loss": 3.4771, "step": 1021 }, { "epoch": 0.12, "learning_rate": 9.890765804833103e-06, "loss": 3.5166, "step": 1022 }, { "epoch": 0.12, "learning_rate": 9.890348420059433e-06, "loss": 3.4091, "step": 1023 }, { "epoch": 0.12, "learning_rate": 9.889930248230744e-06, "loss": 3.4554, "step": 1024 }, { "epoch": 0.12, "learning_rate": 9.889511289414332e-06, "loss": 3.4671, "step": 1025 }, { "epoch": 0.12, "learning_rate": 9.889091543677628e-06, "loss": 3.4667, "step": 1026 }, { "epoch": 0.12, "learning_rate": 9.888671011088186e-06, "loss": 3.4029, "step": 1027 }, { "epoch": 0.12, "learning_rate": 9.888249691713685e-06, "loss": 3.497, "step": 1028 }, { "epoch": 0.12, "learning_rate": 9.887827585621935e-06, "loss": 3.4132, "step": 1029 }, { "epoch": 0.12, "learning_rate": 9.887404692880868e-06, "loss": 3.4116, "step": 1030 }, { "epoch": 0.12, "learning_rate": 9.886981013558543e-06, "loss": 3.4337, "step": 1031 }, { "epoch": 0.12, "learning_rate": 9.886556547723151e-06, "loss": 3.3665, "step": 1032 }, { "epoch": 0.12, "learning_rate": 9.886131295443003e-06, "loss": 3.4002, "step": 1033 }, { "epoch": 0.12, "learning_rate": 9.88570525678654e-06, "loss": 3.3068, "step": 1034 }, { "epoch": 0.12, "learning_rate": 9.885278431822331e-06, "loss": 3.4693, "step": 1035 }, { "epoch": 0.12, "learning_rate": 9.884850820619066e-06, "loss": 3.4054, "step": 1036 }, { "epoch": 0.12, "learning_rate": 9.884422423245565e-06, "loss": 3.4531, "step": 1037 }, { "epoch": 0.12, "learning_rate": 9.883993239770777e-06, "loss": 3.4129, "step": 1038 }, { "epoch": 0.12, "learning_rate": 9.883563270263773e-06, "loss": 3.3937, "step": 1039 }, { "epoch": 0.12, "learning_rate": 9.883132514793755e-06, "loss": 3.4262, "step": 1040 }, { "epoch": 0.12, "learning_rate": 9.882700973430046e-06, "loss": 3.3662, "step": 1041 }, { "epoch": 0.13, "learning_rate": 9.882268646242101e-06, "loss": 3.3848, "step": 1042 }, { "epoch": 0.13, "learning_rate": 9.881835533299497e-06, "loss": 3.4081, "step": 1043 }, { "epoch": 0.13, "learning_rate": 9.881401634671939e-06, "loss": 3.4298, "step": 1044 }, { "epoch": 0.13, "learning_rate": 9.880966950429262e-06, "loss": 3.4053, "step": 1045 }, { "epoch": 0.13, "learning_rate": 9.880531480641422e-06, "loss": 3.3676, "step": 1046 }, { "epoch": 0.13, "learning_rate": 9.880095225378503e-06, "loss": 3.5551, "step": 1047 }, { "epoch": 0.13, "learning_rate": 9.879658184710718e-06, "loss": 3.3805, "step": 1048 }, { "epoch": 0.13, "learning_rate": 9.879220358708402e-06, "loss": 3.3695, "step": 1049 }, { "epoch": 0.13, "learning_rate": 9.878781747442023e-06, "loss": 3.4138, "step": 1050 }, { "epoch": 0.13, "learning_rate": 9.878342350982166e-06, "loss": 3.4741, "step": 1051 }, { "epoch": 0.13, "learning_rate": 9.877902169399554e-06, "loss": 3.4082, "step": 1052 }, { "epoch": 0.13, "learning_rate": 9.877461202765024e-06, "loss": 3.408, "step": 1053 }, { "epoch": 0.13, "learning_rate": 9.877019451149549e-06, "loss": 3.4039, "step": 1054 }, { "epoch": 0.13, "learning_rate": 9.876576914624224e-06, "loss": 3.3927, "step": 1055 }, { "epoch": 0.13, "learning_rate": 9.876133593260269e-06, "loss": 3.5445, "step": 1056 }, { "epoch": 0.13, "learning_rate": 9.875689487129037e-06, "loss": 3.4408, "step": 1057 }, { "epoch": 0.13, "learning_rate": 9.875244596301996e-06, "loss": 3.4223, "step": 1058 }, { "epoch": 0.13, "learning_rate": 9.874798920850754e-06, "loss": 3.4397, "step": 1059 }, { "epoch": 0.13, "learning_rate": 9.874352460847034e-06, "loss": 3.419, "step": 1060 }, { "epoch": 0.13, "learning_rate": 9.87390521636269e-06, "loss": 3.412, "step": 1061 }, { "epoch": 0.13, "learning_rate": 9.873457187469701e-06, "loss": 3.3977, "step": 1062 }, { "epoch": 0.13, "learning_rate": 9.873008374240176e-06, "loss": 3.5516, "step": 1063 }, { "epoch": 0.13, "learning_rate": 9.872558776746346e-06, "loss": 3.5111, "step": 1064 }, { "epoch": 0.13, "learning_rate": 9.872108395060566e-06, "loss": 3.4711, "step": 1065 }, { "epoch": 0.13, "learning_rate": 9.871657229255327e-06, "loss": 3.4391, "step": 1066 }, { "epoch": 0.13, "learning_rate": 9.871205279403233e-06, "loss": 3.4935, "step": 1067 }, { "epoch": 0.13, "learning_rate": 9.870752545577027e-06, "loss": 3.4768, "step": 1068 }, { "epoch": 0.13, "learning_rate": 9.870299027849567e-06, "loss": 3.4676, "step": 1069 }, { "epoch": 0.13, "learning_rate": 9.869844726293847e-06, "loss": 3.4992, "step": 1070 }, { "epoch": 0.13, "learning_rate": 9.869389640982978e-06, "loss": 3.4603, "step": 1071 }, { "epoch": 0.13, "learning_rate": 9.868933771990206e-06, "loss": 3.4747, "step": 1072 }, { "epoch": 0.13, "learning_rate": 9.868477119388897e-06, "loss": 3.3881, "step": 1073 }, { "epoch": 0.13, "learning_rate": 9.868019683252543e-06, "loss": 3.3935, "step": 1074 }, { "epoch": 0.13, "learning_rate": 9.867561463654766e-06, "loss": 3.4314, "step": 1075 }, { "epoch": 0.13, "learning_rate": 9.867102460669314e-06, "loss": 3.5816, "step": 1076 }, { "epoch": 0.13, "learning_rate": 9.866642674370054e-06, "loss": 3.4397, "step": 1077 }, { "epoch": 0.13, "learning_rate": 9.866182104830988e-06, "loss": 3.4756, "step": 1078 }, { "epoch": 0.13, "learning_rate": 9.86572075212624e-06, "loss": 3.4987, "step": 1079 }, { "epoch": 0.13, "learning_rate": 9.865258616330058e-06, "loss": 3.5033, "step": 1080 }, { "epoch": 0.13, "learning_rate": 9.864795697516823e-06, "loss": 3.475, "step": 1081 }, { "epoch": 0.13, "learning_rate": 9.864331995761033e-06, "loss": 3.4641, "step": 1082 }, { "epoch": 0.13, "learning_rate": 9.863867511137318e-06, "loss": 3.5414, "step": 1083 }, { "epoch": 0.13, "learning_rate": 9.863402243720432e-06, "loss": 3.4456, "step": 1084 }, { "epoch": 0.13, "learning_rate": 9.862936193585255e-06, "loss": 3.4044, "step": 1085 }, { "epoch": 0.13, "learning_rate": 9.862469360806795e-06, "loss": 3.4108, "step": 1086 }, { "epoch": 0.13, "learning_rate": 9.862001745460184e-06, "loss": 3.4508, "step": 1087 }, { "epoch": 0.13, "learning_rate": 9.861533347620677e-06, "loss": 3.3938, "step": 1088 }, { "epoch": 0.13, "learning_rate": 9.861064167363664e-06, "loss": 3.4158, "step": 1089 }, { "epoch": 0.13, "learning_rate": 9.860594204764649e-06, "loss": 3.3792, "step": 1090 }, { "epoch": 0.13, "learning_rate": 9.860123459899271e-06, "loss": 3.4152, "step": 1091 }, { "epoch": 0.13, "learning_rate": 9.859651932843293e-06, "loss": 3.4366, "step": 1092 }, { "epoch": 0.13, "learning_rate": 9.8591796236726e-06, "loss": 3.4437, "step": 1093 }, { "epoch": 0.13, "learning_rate": 9.858706532463207e-06, "loss": 3.5162, "step": 1094 }, { "epoch": 0.13, "learning_rate": 9.858232659291255e-06, "loss": 3.4529, "step": 1095 }, { "epoch": 0.13, "learning_rate": 9.857758004233007e-06, "loss": 3.4682, "step": 1096 }, { "epoch": 0.13, "learning_rate": 9.857282567364856e-06, "loss": 3.4916, "step": 1097 }, { "epoch": 0.13, "learning_rate": 9.856806348763317e-06, "loss": 3.356, "step": 1098 }, { "epoch": 0.13, "learning_rate": 9.856329348505037e-06, "loss": 3.5061, "step": 1099 }, { "epoch": 0.13, "learning_rate": 9.855851566666779e-06, "loss": 3.4897, "step": 1100 }, { "epoch": 0.13, "learning_rate": 9.85537300332544e-06, "loss": 3.4223, "step": 1101 }, { "epoch": 0.13, "learning_rate": 9.85489365855804e-06, "loss": 3.446, "step": 1102 }, { "epoch": 0.13, "learning_rate": 9.854413532441728e-06, "loss": 3.464, "step": 1103 }, { "epoch": 0.13, "learning_rate": 9.853932625053772e-06, "loss": 3.3746, "step": 1104 }, { "epoch": 0.13, "learning_rate": 9.853450936471571e-06, "loss": 3.4305, "step": 1105 }, { "epoch": 0.13, "learning_rate": 9.852968466772645e-06, "loss": 3.4532, "step": 1106 }, { "epoch": 0.13, "learning_rate": 9.852485216034649e-06, "loss": 3.4046, "step": 1107 }, { "epoch": 0.13, "learning_rate": 9.852001184335353e-06, "loss": 3.467, "step": 1108 }, { "epoch": 0.13, "learning_rate": 9.851516371752658e-06, "loss": 3.4102, "step": 1109 }, { "epoch": 0.13, "learning_rate": 9.851030778364592e-06, "loss": 3.4635, "step": 1110 }, { "epoch": 0.13, "learning_rate": 9.850544404249305e-06, "loss": 3.4038, "step": 1111 }, { "epoch": 0.13, "learning_rate": 9.850057249485074e-06, "loss": 3.4678, "step": 1112 }, { "epoch": 0.13, "learning_rate": 9.849569314150302e-06, "loss": 3.3828, "step": 1113 }, { "epoch": 0.13, "learning_rate": 9.849080598323517e-06, "loss": 3.4293, "step": 1114 }, { "epoch": 0.13, "learning_rate": 9.848591102083375e-06, "loss": 3.4174, "step": 1115 }, { "epoch": 0.13, "learning_rate": 9.848100825508656e-06, "loss": 3.4813, "step": 1116 }, { "epoch": 0.13, "learning_rate": 9.847609768678263e-06, "loss": 3.3908, "step": 1117 }, { "epoch": 0.13, "learning_rate": 9.847117931671228e-06, "loss": 3.5449, "step": 1118 }, { "epoch": 0.13, "learning_rate": 9.846625314566708e-06, "loss": 3.3695, "step": 1119 }, { "epoch": 0.13, "learning_rate": 9.846131917443984e-06, "loss": 3.5116, "step": 1120 }, { "epoch": 0.13, "learning_rate": 9.845637740382465e-06, "loss": 3.3773, "step": 1121 }, { "epoch": 0.13, "learning_rate": 9.845142783461683e-06, "loss": 3.3452, "step": 1122 }, { "epoch": 0.13, "learning_rate": 9.844647046761298e-06, "loss": 3.4534, "step": 1123 }, { "epoch": 0.13, "learning_rate": 9.844150530361092e-06, "loss": 3.3371, "step": 1124 }, { "epoch": 0.14, "learning_rate": 9.843653234340976e-06, "loss": 3.4321, "step": 1125 }, { "epoch": 0.14, "learning_rate": 9.843155158780984e-06, "loss": 3.4084, "step": 1126 }, { "epoch": 0.14, "learning_rate": 9.842656303761276e-06, "loss": 3.4379, "step": 1127 }, { "epoch": 0.14, "learning_rate": 9.842156669362142e-06, "loss": 3.3471, "step": 1128 }, { "epoch": 0.14, "learning_rate": 9.841656255663989e-06, "loss": 3.3603, "step": 1129 }, { "epoch": 0.14, "learning_rate": 9.841155062747354e-06, "loss": 3.3747, "step": 1130 }, { "epoch": 0.14, "learning_rate": 9.840653090692902e-06, "loss": 3.4849, "step": 1131 }, { "epoch": 0.14, "learning_rate": 9.84015033958142e-06, "loss": 3.4973, "step": 1132 }, { "epoch": 0.14, "learning_rate": 9.83964680949382e-06, "loss": 3.5162, "step": 1133 }, { "epoch": 0.14, "learning_rate": 9.83914250051114e-06, "loss": 3.3929, "step": 1134 }, { "epoch": 0.14, "learning_rate": 9.838637412714544e-06, "loss": 3.426, "step": 1135 }, { "epoch": 0.14, "learning_rate": 9.838131546185321e-06, "loss": 3.4174, "step": 1136 }, { "epoch": 0.14, "learning_rate": 9.837624901004887e-06, "loss": 3.3986, "step": 1137 }, { "epoch": 0.14, "learning_rate": 9.83711747725478e-06, "loss": 3.4417, "step": 1138 }, { "epoch": 0.14, "learning_rate": 9.836609275016666e-06, "loss": 3.4555, "step": 1139 }, { "epoch": 0.14, "learning_rate": 9.836100294372334e-06, "loss": 3.5277, "step": 1140 }, { "epoch": 0.14, "learning_rate": 9.8355905354037e-06, "loss": 3.4831, "step": 1141 }, { "epoch": 0.14, "learning_rate": 9.835079998192806e-06, "loss": 3.4233, "step": 1142 }, { "epoch": 0.14, "learning_rate": 9.834568682821819e-06, "loss": 3.3335, "step": 1143 }, { "epoch": 0.14, "learning_rate": 9.834056589373027e-06, "loss": 3.3812, "step": 1144 }, { "epoch": 0.14, "learning_rate": 9.833543717928849e-06, "loss": 3.4117, "step": 1145 }, { "epoch": 0.14, "learning_rate": 9.833030068571824e-06, "loss": 3.5115, "step": 1146 }, { "epoch": 0.14, "learning_rate": 9.832515641384624e-06, "loss": 3.4211, "step": 1147 }, { "epoch": 0.14, "learning_rate": 9.832000436450035e-06, "loss": 3.4069, "step": 1148 }, { "epoch": 0.14, "learning_rate": 9.83148445385098e-06, "loss": 3.3772, "step": 1149 }, { "epoch": 0.14, "learning_rate": 9.8309676936705e-06, "loss": 3.4289, "step": 1150 }, { "epoch": 0.14, "learning_rate": 9.83045015599176e-06, "loss": 3.4304, "step": 1151 }, { "epoch": 0.14, "learning_rate": 9.829931840898052e-06, "loss": 3.2872, "step": 1152 }, { "epoch": 0.14, "learning_rate": 9.829412748472799e-06, "loss": 3.4749, "step": 1153 }, { "epoch": 0.14, "learning_rate": 9.828892878799542e-06, "loss": 3.4136, "step": 1154 }, { "epoch": 0.14, "learning_rate": 9.828372231961947e-06, "loss": 3.3176, "step": 1155 }, { "epoch": 0.14, "learning_rate": 9.82785080804381e-06, "loss": 3.461, "step": 1156 }, { "epoch": 0.14, "learning_rate": 9.827328607129045e-06, "loss": 3.426, "step": 1157 }, { "epoch": 0.14, "learning_rate": 9.826805629301701e-06, "loss": 3.407, "step": 1158 }, { "epoch": 0.14, "learning_rate": 9.826281874645943e-06, "loss": 3.378, "step": 1159 }, { "epoch": 0.14, "learning_rate": 9.825757343246062e-06, "loss": 3.3684, "step": 1160 }, { "epoch": 0.14, "learning_rate": 9.82523203518648e-06, "loss": 3.3772, "step": 1161 }, { "epoch": 0.14, "learning_rate": 9.82470595055174e-06, "loss": 3.5085, "step": 1162 }, { "epoch": 0.14, "learning_rate": 9.824179089426509e-06, "loss": 3.3956, "step": 1163 }, { "epoch": 0.14, "learning_rate": 9.823651451895582e-06, "loss": 3.3978, "step": 1164 }, { "epoch": 0.14, "learning_rate": 9.823123038043875e-06, "loss": 3.3112, "step": 1165 }, { "epoch": 0.14, "learning_rate": 9.822593847956432e-06, "loss": 3.3075, "step": 1166 }, { "epoch": 0.14, "learning_rate": 9.822063881718422e-06, "loss": 3.4348, "step": 1167 }, { "epoch": 0.14, "learning_rate": 9.821533139415138e-06, "loss": 3.4516, "step": 1168 }, { "epoch": 0.14, "learning_rate": 9.821001621131995e-06, "loss": 3.3678, "step": 1169 }, { "epoch": 0.14, "learning_rate": 9.820469326954542e-06, "loss": 3.4341, "step": 1170 }, { "epoch": 0.14, "learning_rate": 9.819936256968442e-06, "loss": 3.4225, "step": 1171 }, { "epoch": 0.14, "learning_rate": 9.819402411259485e-06, "loss": 3.4643, "step": 1172 }, { "epoch": 0.14, "learning_rate": 9.818867789913594e-06, "loss": 3.4441, "step": 1173 }, { "epoch": 0.14, "learning_rate": 9.818332393016809e-06, "loss": 3.5045, "step": 1174 }, { "epoch": 0.14, "learning_rate": 9.817796220655297e-06, "loss": 3.4141, "step": 1175 }, { "epoch": 0.14, "learning_rate": 9.81725927291535e-06, "loss": 3.376, "step": 1176 }, { "epoch": 0.14, "learning_rate": 9.816721549883384e-06, "loss": 3.4126, "step": 1177 }, { "epoch": 0.14, "learning_rate": 9.81618305164594e-06, "loss": 3.4375, "step": 1178 }, { "epoch": 0.14, "learning_rate": 9.815643778289687e-06, "loss": 3.412, "step": 1179 }, { "epoch": 0.14, "learning_rate": 9.815103729901411e-06, "loss": 3.5009, "step": 1180 }, { "epoch": 0.14, "learning_rate": 9.814562906568034e-06, "loss": 3.4476, "step": 1181 }, { "epoch": 0.14, "learning_rate": 9.814021308376592e-06, "loss": 3.4021, "step": 1182 }, { "epoch": 0.14, "learning_rate": 9.813478935414249e-06, "loss": 3.3966, "step": 1183 }, { "epoch": 0.14, "learning_rate": 9.812935787768298e-06, "loss": 3.4298, "step": 1184 }, { "epoch": 0.14, "learning_rate": 9.812391865526155e-06, "loss": 3.4999, "step": 1185 }, { "epoch": 0.14, "learning_rate": 9.811847168775352e-06, "loss": 3.4987, "step": 1186 }, { "epoch": 0.14, "learning_rate": 9.811301697603559e-06, "loss": 3.5166, "step": 1187 }, { "epoch": 0.14, "learning_rate": 9.810755452098563e-06, "loss": 3.4139, "step": 1188 }, { "epoch": 0.14, "learning_rate": 9.810208432348276e-06, "loss": 3.4101, "step": 1189 }, { "epoch": 0.14, "learning_rate": 9.809660638440738e-06, "loss": 3.4526, "step": 1190 }, { "epoch": 0.14, "learning_rate": 9.809112070464107e-06, "loss": 3.6413, "step": 1191 }, { "epoch": 0.14, "learning_rate": 9.808562728506672e-06, "loss": 3.4817, "step": 1192 }, { "epoch": 0.14, "learning_rate": 9.808012612656845e-06, "loss": 3.4262, "step": 1193 }, { "epoch": 0.14, "learning_rate": 9.807461723003162e-06, "loss": 3.3494, "step": 1194 }, { "epoch": 0.14, "learning_rate": 9.806910059634282e-06, "loss": 3.4503, "step": 1195 }, { "epoch": 0.14, "learning_rate": 9.806357622638992e-06, "loss": 3.4753, "step": 1196 }, { "epoch": 0.14, "learning_rate": 9.805804412106197e-06, "loss": 3.3927, "step": 1197 }, { "epoch": 0.14, "learning_rate": 9.805250428124937e-06, "loss": 3.3312, "step": 1198 }, { "epoch": 0.14, "learning_rate": 9.804695670784367e-06, "loss": 3.4647, "step": 1199 }, { "epoch": 0.14, "learning_rate": 9.804140140173772e-06, "loss": 3.5715, "step": 1200 }, { "epoch": 0.14, "learning_rate": 9.803583836382557e-06, "loss": 3.4445, "step": 1201 }, { "epoch": 0.14, "learning_rate": 9.803026759500254e-06, "loss": 3.5246, "step": 1202 }, { "epoch": 0.14, "learning_rate": 9.802468909616521e-06, "loss": 3.3876, "step": 1203 }, { "epoch": 0.14, "learning_rate": 9.801910286821139e-06, "loss": 3.4485, "step": 1204 }, { "epoch": 0.14, "learning_rate": 9.801350891204008e-06, "loss": 3.3985, "step": 1205 }, { "epoch": 0.14, "learning_rate": 9.800790722855165e-06, "loss": 3.5501, "step": 1206 }, { "epoch": 0.14, "learning_rate": 9.800229781864757e-06, "loss": 3.5072, "step": 1207 }, { "epoch": 0.15, "learning_rate": 9.799668068323068e-06, "loss": 3.5006, "step": 1208 }, { "epoch": 0.15, "learning_rate": 9.799105582320494e-06, "loss": 3.4405, "step": 1209 }, { "epoch": 0.15, "learning_rate": 9.798542323947567e-06, "loss": 3.5365, "step": 1210 }, { "epoch": 0.15, "learning_rate": 9.797978293294934e-06, "loss": 3.3804, "step": 1211 }, { "epoch": 0.15, "learning_rate": 9.797413490453375e-06, "loss": 3.5073, "step": 1212 }, { "epoch": 0.15, "learning_rate": 9.796847915513787e-06, "loss": 3.4772, "step": 1213 }, { "epoch": 0.15, "learning_rate": 9.796281568567192e-06, "loss": 3.4516, "step": 1214 }, { "epoch": 0.15, "learning_rate": 9.795714449704744e-06, "loss": 3.4558, "step": 1215 }, { "epoch": 0.15, "learning_rate": 9.795146559017708e-06, "loss": 3.5867, "step": 1216 }, { "epoch": 0.15, "learning_rate": 9.794577896597487e-06, "loss": 3.3828, "step": 1217 }, { "epoch": 0.15, "learning_rate": 9.794008462535595e-06, "loss": 3.5313, "step": 1218 }, { "epoch": 0.15, "learning_rate": 9.793438256923684e-06, "loss": 3.484, "step": 1219 }, { "epoch": 0.15, "learning_rate": 9.792867279853522e-06, "loss": 3.502, "step": 1220 }, { "epoch": 0.15, "learning_rate": 9.792295531416998e-06, "loss": 3.446, "step": 1221 }, { "epoch": 0.15, "learning_rate": 9.791723011706132e-06, "loss": 3.4663, "step": 1222 }, { "epoch": 0.15, "learning_rate": 9.791149720813065e-06, "loss": 3.4449, "step": 1223 }, { "epoch": 0.15, "learning_rate": 9.790575658830066e-06, "loss": 3.4958, "step": 1224 }, { "epoch": 0.15, "learning_rate": 9.79000082584952e-06, "loss": 3.4975, "step": 1225 }, { "epoch": 0.15, "learning_rate": 9.789425221963944e-06, "loss": 3.4699, "step": 1226 }, { "epoch": 0.15, "learning_rate": 9.788848847265974e-06, "loss": 3.5993, "step": 1227 }, { "epoch": 0.15, "learning_rate": 9.788271701848373e-06, "loss": 3.5447, "step": 1228 }, { "epoch": 0.15, "learning_rate": 9.787693785804028e-06, "loss": 3.3684, "step": 1229 }, { "epoch": 0.15, "learning_rate": 9.787115099225948e-06, "loss": 3.4769, "step": 1230 }, { "epoch": 0.15, "learning_rate": 9.786535642207266e-06, "loss": 3.5121, "step": 1231 }, { "epoch": 0.15, "learning_rate": 9.785955414841243e-06, "loss": 3.405, "step": 1232 }, { "epoch": 0.15, "learning_rate": 9.785374417221259e-06, "loss": 3.4654, "step": 1233 }, { "epoch": 0.15, "learning_rate": 9.78479264944082e-06, "loss": 3.469, "step": 1234 }, { "epoch": 0.15, "learning_rate": 9.784210111593556e-06, "loss": 3.4534, "step": 1235 }, { "epoch": 0.15, "learning_rate": 9.783626803773221e-06, "loss": 3.4397, "step": 1236 }, { "epoch": 0.15, "learning_rate": 9.783042726073696e-06, "loss": 3.4337, "step": 1237 }, { "epoch": 0.15, "learning_rate": 9.782457878588977e-06, "loss": 3.4639, "step": 1238 }, { "epoch": 0.15, "learning_rate": 9.781872261413194e-06, "loss": 3.3719, "step": 1239 }, { "epoch": 0.15, "learning_rate": 9.781285874640593e-06, "loss": 3.4094, "step": 1240 }, { "epoch": 0.15, "learning_rate": 9.780698718365551e-06, "loss": 3.4446, "step": 1241 }, { "epoch": 0.15, "learning_rate": 9.780110792682564e-06, "loss": 3.4378, "step": 1242 }, { "epoch": 0.15, "learning_rate": 9.779522097686254e-06, "loss": 3.5047, "step": 1243 }, { "epoch": 0.15, "learning_rate": 9.778932633471363e-06, "loss": 3.461, "step": 1244 }, { "epoch": 0.15, "learning_rate": 9.778342400132761e-06, "loss": 3.4016, "step": 1245 }, { "epoch": 0.15, "learning_rate": 9.77775139776544e-06, "loss": 3.4739, "step": 1246 }, { "epoch": 0.15, "learning_rate": 9.777159626464519e-06, "loss": 3.4871, "step": 1247 }, { "epoch": 0.15, "learning_rate": 9.776567086325234e-06, "loss": 3.5222, "step": 1248 }, { "epoch": 0.15, "learning_rate": 9.77597377744295e-06, "loss": 3.3987, "step": 1249 }, { "epoch": 0.15, "learning_rate": 9.775379699913157e-06, "loss": 3.5085, "step": 1250 }, { "epoch": 0.15, "learning_rate": 9.774784853831462e-06, "loss": 3.5653, "step": 1251 }, { "epoch": 0.15, "learning_rate": 9.774189239293603e-06, "loss": 3.4589, "step": 1252 }, { "epoch": 0.15, "learning_rate": 9.773592856395437e-06, "loss": 3.4223, "step": 1253 }, { "epoch": 0.15, "learning_rate": 9.772995705232946e-06, "loss": 3.4557, "step": 1254 }, { "epoch": 0.15, "learning_rate": 9.772397785902237e-06, "loss": 3.3643, "step": 1255 }, { "epoch": 0.15, "learning_rate": 9.77179909849954e-06, "loss": 3.4104, "step": 1256 }, { "epoch": 0.15, "learning_rate": 9.771199643121205e-06, "loss": 3.4758, "step": 1257 }, { "epoch": 0.15, "learning_rate": 9.770599419863712e-06, "loss": 3.4321, "step": 1258 }, { "epoch": 0.15, "learning_rate": 9.769998428823658e-06, "loss": 3.397, "step": 1259 }, { "epoch": 0.15, "learning_rate": 9.769396670097767e-06, "loss": 3.4754, "step": 1260 }, { "epoch": 0.15, "learning_rate": 9.76879414378289e-06, "loss": 3.4327, "step": 1261 }, { "epoch": 0.15, "learning_rate": 9.768190849975995e-06, "loss": 3.4664, "step": 1262 }, { "epoch": 0.15, "learning_rate": 9.767586788774176e-06, "loss": 3.4553, "step": 1263 }, { "epoch": 0.15, "learning_rate": 9.766981960274653e-06, "loss": 3.3861, "step": 1264 }, { "epoch": 0.15, "learning_rate": 9.766376364574767e-06, "loss": 3.4203, "step": 1265 }, { "epoch": 0.15, "learning_rate": 9.765770001771981e-06, "loss": 3.4611, "step": 1266 }, { "epoch": 0.15, "learning_rate": 9.765162871963885e-06, "loss": 3.4162, "step": 1267 }, { "epoch": 0.15, "learning_rate": 9.764554975248191e-06, "loss": 3.4575, "step": 1268 }, { "epoch": 0.15, "learning_rate": 9.763946311722731e-06, "loss": 3.4199, "step": 1269 }, { "epoch": 0.15, "learning_rate": 9.763336881485467e-06, "loss": 3.3609, "step": 1270 }, { "epoch": 0.15, "learning_rate": 9.762726684634481e-06, "loss": 3.383, "step": 1271 }, { "epoch": 0.15, "learning_rate": 9.762115721267977e-06, "loss": 3.4108, "step": 1272 }, { "epoch": 0.15, "learning_rate": 9.761503991484285e-06, "loss": 3.4947, "step": 1273 }, { "epoch": 0.15, "learning_rate": 9.760891495381854e-06, "loss": 3.5444, "step": 1274 }, { "epoch": 0.15, "learning_rate": 9.760278233059264e-06, "loss": 3.4525, "step": 1275 }, { "epoch": 0.15, "learning_rate": 9.759664204615209e-06, "loss": 3.5154, "step": 1276 }, { "epoch": 0.15, "learning_rate": 9.759049410148515e-06, "loss": 3.3863, "step": 1277 }, { "epoch": 0.15, "learning_rate": 9.758433849758125e-06, "loss": 3.3926, "step": 1278 }, { "epoch": 0.15, "learning_rate": 9.75781752354311e-06, "loss": 3.4082, "step": 1279 }, { "epoch": 0.15, "learning_rate": 9.757200431602657e-06, "loss": 3.4389, "step": 1280 }, { "epoch": 0.15, "learning_rate": 9.756582574036087e-06, "loss": 3.5166, "step": 1281 }, { "epoch": 0.15, "learning_rate": 9.755963950942833e-06, "loss": 3.4286, "step": 1282 }, { "epoch": 0.15, "learning_rate": 9.75534456242246e-06, "loss": 3.3693, "step": 1283 }, { "epoch": 0.15, "learning_rate": 9.754724408574652e-06, "loss": 3.3462, "step": 1284 }, { "epoch": 0.15, "learning_rate": 9.754103489499219e-06, "loss": 3.5652, "step": 1285 }, { "epoch": 0.15, "learning_rate": 9.753481805296085e-06, "loss": 3.3607, "step": 1286 }, { "epoch": 0.15, "learning_rate": 9.752859356065311e-06, "loss": 3.5057, "step": 1287 }, { "epoch": 0.15, "learning_rate": 9.752236141907072e-06, "loss": 3.4328, "step": 1288 }, { "epoch": 0.15, "learning_rate": 9.751612162921669e-06, "loss": 3.3597, "step": 1289 }, { "epoch": 0.15, "learning_rate": 9.750987419209523e-06, "loss": 3.447, "step": 1290 }, { "epoch": 0.15, "learning_rate": 9.750361910871184e-06, "loss": 3.3747, "step": 1291 }, { "epoch": 0.16, "learning_rate": 9.749735638007318e-06, "loss": 3.4408, "step": 1292 }, { "epoch": 0.16, "learning_rate": 9.749108600718722e-06, "loss": 3.4391, "step": 1293 }, { "epoch": 0.16, "learning_rate": 9.748480799106307e-06, "loss": 3.4529, "step": 1294 }, { "epoch": 0.16, "learning_rate": 9.747852233271116e-06, "loss": 3.3451, "step": 1295 }, { "epoch": 0.16, "learning_rate": 9.747222903314308e-06, "loss": 3.3771, "step": 1296 }, { "epoch": 0.16, "learning_rate": 9.74659280933717e-06, "loss": 3.4756, "step": 1297 }, { "epoch": 0.16, "learning_rate": 9.745961951441106e-06, "loss": 3.5366, "step": 1298 }, { "epoch": 0.16, "learning_rate": 9.74533032972765e-06, "loss": 3.3926, "step": 1299 }, { "epoch": 0.16, "learning_rate": 9.744697944298452e-06, "loss": 3.4154, "step": 1300 }, { "epoch": 0.16, "learning_rate": 9.744064795255292e-06, "loss": 3.3858, "step": 1301 }, { "epoch": 0.16, "learning_rate": 9.743430882700067e-06, "loss": 3.3881, "step": 1302 }, { "epoch": 0.16, "learning_rate": 9.742796206734801e-06, "loss": 3.4267, "step": 1303 }, { "epoch": 0.16, "learning_rate": 9.742160767461637e-06, "loss": 3.3292, "step": 1304 }, { "epoch": 0.16, "learning_rate": 9.741524564982845e-06, "loss": 3.469, "step": 1305 }, { "epoch": 0.16, "learning_rate": 9.740887599400815e-06, "loss": 3.3703, "step": 1306 }, { "epoch": 0.16, "learning_rate": 9.740249870818059e-06, "loss": 3.4854, "step": 1307 }, { "epoch": 0.16, "learning_rate": 9.739611379337215e-06, "loss": 3.4555, "step": 1308 }, { "epoch": 0.16, "learning_rate": 9.738972125061043e-06, "loss": 3.4151, "step": 1309 }, { "epoch": 0.16, "learning_rate": 9.738332108092423e-06, "loss": 3.3737, "step": 1310 }, { "epoch": 0.16, "learning_rate": 9.737691328534358e-06, "loss": 3.3771, "step": 1311 }, { "epoch": 0.16, "learning_rate": 9.73704978648998e-06, "loss": 3.4332, "step": 1312 }, { "epoch": 0.16, "learning_rate": 9.736407482062535e-06, "loss": 3.5283, "step": 1313 }, { "epoch": 0.16, "learning_rate": 9.7357644153554e-06, "loss": 3.3193, "step": 1314 }, { "epoch": 0.16, "learning_rate": 9.735120586472066e-06, "loss": 3.4703, "step": 1315 }, { "epoch": 0.16, "learning_rate": 9.734475995516152e-06, "loss": 3.4297, "step": 1316 }, { "epoch": 0.16, "learning_rate": 9.733830642591402e-06, "loss": 3.3808, "step": 1317 }, { "epoch": 0.16, "learning_rate": 9.733184527801677e-06, "loss": 3.438, "step": 1318 }, { "epoch": 0.16, "learning_rate": 9.732537651250961e-06, "loss": 3.46, "step": 1319 }, { "epoch": 0.16, "learning_rate": 9.731890013043367e-06, "loss": 3.4744, "step": 1320 }, { "epoch": 0.16, "learning_rate": 9.731241613283124e-06, "loss": 3.4078, "step": 1321 }, { "epoch": 0.16, "learning_rate": 9.730592452074587e-06, "loss": 3.5284, "step": 1322 }, { "epoch": 0.16, "learning_rate": 9.72994252952223e-06, "loss": 3.3454, "step": 1323 }, { "epoch": 0.16, "learning_rate": 9.729291845730653e-06, "loss": 3.4116, "step": 1324 }, { "epoch": 0.16, "learning_rate": 9.728640400804576e-06, "loss": 3.355, "step": 1325 }, { "epoch": 0.16, "learning_rate": 9.727988194848846e-06, "loss": 3.4318, "step": 1326 }, { "epoch": 0.16, "learning_rate": 9.727335227968428e-06, "loss": 3.4289, "step": 1327 }, { "epoch": 0.16, "learning_rate": 9.726681500268408e-06, "loss": 3.3925, "step": 1328 }, { "epoch": 0.16, "learning_rate": 9.726027011854003e-06, "loss": 3.3871, "step": 1329 }, { "epoch": 0.16, "learning_rate": 9.725371762830541e-06, "loss": 3.5324, "step": 1330 }, { "epoch": 0.16, "learning_rate": 9.724715753303481e-06, "loss": 3.5159, "step": 1331 }, { "epoch": 0.16, "learning_rate": 9.724058983378402e-06, "loss": 3.4128, "step": 1332 }, { "epoch": 0.16, "learning_rate": 9.723401453161e-06, "loss": 3.3703, "step": 1333 }, { "epoch": 0.16, "learning_rate": 9.722743162757107e-06, "loss": 3.4312, "step": 1334 }, { "epoch": 0.16, "learning_rate": 9.722084112272661e-06, "loss": 3.3934, "step": 1335 }, { "epoch": 0.16, "learning_rate": 9.721424301813732e-06, "loss": 3.4738, "step": 1336 }, { "epoch": 0.16, "learning_rate": 9.72076373148651e-06, "loss": 3.4404, "step": 1337 }, { "epoch": 0.16, "learning_rate": 9.720102401397312e-06, "loss": 3.3676, "step": 1338 }, { "epoch": 0.16, "learning_rate": 9.719440311652566e-06, "loss": 3.3891, "step": 1339 }, { "epoch": 0.16, "learning_rate": 9.718777462358833e-06, "loss": 3.3868, "step": 1340 }, { "epoch": 0.16, "learning_rate": 9.71811385362279e-06, "loss": 3.4372, "step": 1341 }, { "epoch": 0.16, "learning_rate": 9.71744948555124e-06, "loss": 3.4755, "step": 1342 }, { "epoch": 0.16, "learning_rate": 9.716784358251107e-06, "loss": 3.4429, "step": 1343 }, { "epoch": 0.16, "learning_rate": 9.716118471829438e-06, "loss": 3.425, "step": 1344 }, { "epoch": 0.16, "learning_rate": 9.715451826393398e-06, "loss": 3.4155, "step": 1345 }, { "epoch": 0.16, "learning_rate": 9.714784422050278e-06, "loss": 3.4321, "step": 1346 }, { "epoch": 0.16, "learning_rate": 9.714116258907494e-06, "loss": 3.4418, "step": 1347 }, { "epoch": 0.16, "learning_rate": 9.713447337072575e-06, "loss": 3.4833, "step": 1348 }, { "epoch": 0.16, "learning_rate": 9.71277765665318e-06, "loss": 3.4185, "step": 1349 }, { "epoch": 0.16, "learning_rate": 9.712107217757089e-06, "loss": 3.4196, "step": 1350 }, { "epoch": 0.16, "learning_rate": 9.711436020492202e-06, "loss": 3.4514, "step": 1351 }, { "epoch": 0.16, "learning_rate": 9.710764064966541e-06, "loss": 3.5187, "step": 1352 }, { "epoch": 0.16, "learning_rate": 9.71009135128825e-06, "loss": 3.335, "step": 1353 }, { "epoch": 0.16, "learning_rate": 9.709417879565601e-06, "loss": 3.5094, "step": 1354 }, { "epoch": 0.16, "learning_rate": 9.708743649906976e-06, "loss": 3.4213, "step": 1355 }, { "epoch": 0.16, "learning_rate": 9.708068662420888e-06, "loss": 3.4125, "step": 1356 }, { "epoch": 0.16, "learning_rate": 9.707392917215972e-06, "loss": 3.4045, "step": 1357 }, { "epoch": 0.16, "learning_rate": 9.706716414400983e-06, "loss": 3.4329, "step": 1358 }, { "epoch": 0.16, "learning_rate": 9.706039154084795e-06, "loss": 3.3383, "step": 1359 }, { "epoch": 0.16, "learning_rate": 9.705361136376408e-06, "loss": 3.5812, "step": 1360 }, { "epoch": 0.16, "learning_rate": 9.704682361384941e-06, "loss": 3.4988, "step": 1361 }, { "epoch": 0.16, "learning_rate": 9.704002829219639e-06, "loss": 3.3619, "step": 1362 }, { "epoch": 0.16, "learning_rate": 9.703322539989863e-06, "loss": 3.4111, "step": 1363 }, { "epoch": 0.16, "learning_rate": 9.702641493805102e-06, "loss": 3.4371, "step": 1364 }, { "epoch": 0.16, "learning_rate": 9.701959690774963e-06, "loss": 3.3209, "step": 1365 }, { "epoch": 0.16, "learning_rate": 9.701277131009177e-06, "loss": 3.367, "step": 1366 }, { "epoch": 0.16, "learning_rate": 9.700593814617592e-06, "loss": 3.4951, "step": 1367 }, { "epoch": 0.16, "learning_rate": 9.699909741710184e-06, "loss": 3.3584, "step": 1368 }, { "epoch": 0.16, "learning_rate": 9.699224912397048e-06, "loss": 3.3737, "step": 1369 }, { "epoch": 0.16, "learning_rate": 9.698539326788398e-06, "loss": 3.4528, "step": 1370 }, { "epoch": 0.16, "learning_rate": 9.697852984994576e-06, "loss": 3.3722, "step": 1371 }, { "epoch": 0.16, "learning_rate": 9.69716588712604e-06, "loss": 3.4461, "step": 1372 }, { "epoch": 0.16, "learning_rate": 9.696478033293373e-06, "loss": 3.495, "step": 1373 }, { "epoch": 0.16, "learning_rate": 9.695789423607278e-06, "loss": 3.4554, "step": 1374 }, { "epoch": 0.17, "learning_rate": 9.695100058178581e-06, "loss": 3.4207, "step": 1375 }, { "epoch": 0.17, "learning_rate": 9.694409937118228e-06, "loss": 3.4549, "step": 1376 }, { "epoch": 0.17, "learning_rate": 9.693719060537286e-06, "loss": 3.5119, "step": 1377 }, { "epoch": 0.17, "learning_rate": 9.693027428546949e-06, "loss": 3.4279, "step": 1378 }, { "epoch": 0.17, "learning_rate": 9.692335041258524e-06, "loss": 3.4363, "step": 1379 }, { "epoch": 0.17, "learning_rate": 9.691641898783446e-06, "loss": 3.4277, "step": 1380 }, { "epoch": 0.17, "learning_rate": 9.690948001233272e-06, "loss": 3.441, "step": 1381 }, { "epoch": 0.17, "learning_rate": 9.690253348719676e-06, "loss": 3.477, "step": 1382 }, { "epoch": 0.17, "learning_rate": 9.689557941354452e-06, "loss": 3.3874, "step": 1383 }, { "epoch": 0.17, "learning_rate": 9.688861779249528e-06, "loss": 3.4367, "step": 1384 }, { "epoch": 0.17, "learning_rate": 9.688164862516938e-06, "loss": 3.3497, "step": 1385 }, { "epoch": 0.17, "learning_rate": 9.687467191268845e-06, "loss": 3.4955, "step": 1386 }, { "epoch": 0.17, "learning_rate": 9.686768765617534e-06, "loss": 3.4051, "step": 1387 }, { "epoch": 0.17, "learning_rate": 9.686069585675408e-06, "loss": 3.4919, "step": 1388 }, { "epoch": 0.17, "learning_rate": 9.685369651554996e-06, "loss": 3.5022, "step": 1389 }, { "epoch": 0.17, "learning_rate": 9.684668963368945e-06, "loss": 3.4766, "step": 1390 }, { "epoch": 0.17, "learning_rate": 9.683967521230022e-06, "loss": 3.4268, "step": 1391 }, { "epoch": 0.17, "learning_rate": 9.68326532525112e-06, "loss": 3.5235, "step": 1392 }, { "epoch": 0.17, "learning_rate": 9.682562375545248e-06, "loss": 3.4756, "step": 1393 }, { "epoch": 0.17, "learning_rate": 9.681858672225543e-06, "loss": 3.4963, "step": 1394 }, { "epoch": 0.17, "learning_rate": 9.681154215405255e-06, "loss": 3.4753, "step": 1395 }, { "epoch": 0.17, "learning_rate": 9.680449005197762e-06, "loss": 3.3425, "step": 1396 }, { "epoch": 0.17, "learning_rate": 9.679743041716563e-06, "loss": 3.4759, "step": 1397 }, { "epoch": 0.17, "learning_rate": 9.679036325075272e-06, "loss": 3.3848, "step": 1398 }, { "epoch": 0.17, "learning_rate": 9.67832885538763e-06, "loss": 3.4278, "step": 1399 }, { "epoch": 0.17, "learning_rate": 9.677620632767496e-06, "loss": 3.4814, "step": 1400 }, { "epoch": 0.17, "learning_rate": 9.676911657328855e-06, "loss": 3.3887, "step": 1401 }, { "epoch": 0.17, "learning_rate": 9.676201929185809e-06, "loss": 3.4165, "step": 1402 }, { "epoch": 0.17, "learning_rate": 9.67549144845258e-06, "loss": 3.4413, "step": 1403 }, { "epoch": 0.17, "learning_rate": 9.674780215243513e-06, "loss": 3.4387, "step": 1404 }, { "epoch": 0.17, "learning_rate": 9.674068229673077e-06, "loss": 3.4404, "step": 1405 }, { "epoch": 0.17, "learning_rate": 9.673355491855857e-06, "loss": 3.3743, "step": 1406 }, { "epoch": 0.17, "learning_rate": 9.672642001906562e-06, "loss": 3.432, "step": 1407 }, { "epoch": 0.17, "learning_rate": 9.671927759940022e-06, "loss": 3.4098, "step": 1408 }, { "epoch": 0.17, "learning_rate": 9.671212766071186e-06, "loss": 3.3532, "step": 1409 }, { "epoch": 0.17, "learning_rate": 9.670497020415127e-06, "loss": 3.4157, "step": 1410 }, { "epoch": 0.17, "learning_rate": 9.669780523087035e-06, "loss": 3.4111, "step": 1411 }, { "epoch": 0.17, "learning_rate": 9.669063274202228e-06, "loss": 3.456, "step": 1412 }, { "epoch": 0.17, "learning_rate": 9.668345273876137e-06, "loss": 3.5778, "step": 1413 }, { "epoch": 0.17, "learning_rate": 9.667626522224316e-06, "loss": 3.3591, "step": 1414 }, { "epoch": 0.17, "learning_rate": 9.666907019362446e-06, "loss": 3.451, "step": 1415 }, { "epoch": 0.17, "learning_rate": 9.66618676540632e-06, "loss": 3.458, "step": 1416 }, { "epoch": 0.17, "learning_rate": 9.665465760471855e-06, "loss": 3.3767, "step": 1417 }, { "epoch": 0.17, "learning_rate": 9.664744004675095e-06, "loss": 3.4152, "step": 1418 }, { "epoch": 0.17, "learning_rate": 9.664021498132197e-06, "loss": 3.4318, "step": 1419 }, { "epoch": 0.17, "learning_rate": 9.66329824095944e-06, "loss": 3.3735, "step": 1420 }, { "epoch": 0.17, "learning_rate": 9.662574233273228e-06, "loss": 3.4397, "step": 1421 }, { "epoch": 0.17, "learning_rate": 9.66184947519008e-06, "loss": 3.4065, "step": 1422 }, { "epoch": 0.17, "learning_rate": 9.661123966826643e-06, "loss": 3.4687, "step": 1423 }, { "epoch": 0.17, "learning_rate": 9.660397708299677e-06, "loss": 3.3984, "step": 1424 }, { "epoch": 0.17, "learning_rate": 9.659670699726069e-06, "loss": 3.4442, "step": 1425 }, { "epoch": 0.17, "learning_rate": 9.658942941222821e-06, "loss": 3.4056, "step": 1426 }, { "epoch": 0.17, "learning_rate": 9.658214432907066e-06, "loss": 3.2759, "step": 1427 }, { "epoch": 0.17, "learning_rate": 9.65748517489604e-06, "loss": 3.4133, "step": 1428 }, { "epoch": 0.17, "learning_rate": 9.656755167307117e-06, "loss": 3.4247, "step": 1429 }, { "epoch": 0.17, "learning_rate": 9.656024410257783e-06, "loss": 3.3618, "step": 1430 }, { "epoch": 0.17, "learning_rate": 9.655292903865648e-06, "loss": 3.459, "step": 1431 }, { "epoch": 0.17, "learning_rate": 9.65456064824844e-06, "loss": 3.3471, "step": 1432 }, { "epoch": 0.17, "learning_rate": 9.653827643524004e-06, "loss": 3.4596, "step": 1433 }, { "epoch": 0.17, "learning_rate": 9.653093889810318e-06, "loss": 3.4352, "step": 1434 }, { "epoch": 0.17, "learning_rate": 9.65235938722547e-06, "loss": 3.3954, "step": 1435 }, { "epoch": 0.17, "learning_rate": 9.651624135887668e-06, "loss": 3.4738, "step": 1436 }, { "epoch": 0.17, "learning_rate": 9.650888135915245e-06, "loss": 3.349, "step": 1437 }, { "epoch": 0.17, "learning_rate": 9.650151387426657e-06, "loss": 3.4443, "step": 1438 }, { "epoch": 0.17, "learning_rate": 9.649413890540471e-06, "loss": 3.4724, "step": 1439 }, { "epoch": 0.17, "learning_rate": 9.648675645375385e-06, "loss": 3.4329, "step": 1440 }, { "epoch": 0.17, "learning_rate": 9.64793665205021e-06, "loss": 3.3652, "step": 1441 }, { "epoch": 0.17, "learning_rate": 9.647196910683879e-06, "loss": 3.5021, "step": 1442 }, { "epoch": 0.17, "learning_rate": 9.646456421395447e-06, "loss": 3.3389, "step": 1443 }, { "epoch": 0.17, "learning_rate": 9.64571518430409e-06, "loss": 3.4505, "step": 1444 }, { "epoch": 0.17, "learning_rate": 9.644973199529103e-06, "loss": 3.472, "step": 1445 }, { "epoch": 0.17, "learning_rate": 9.644230467189901e-06, "loss": 3.46, "step": 1446 }, { "epoch": 0.17, "learning_rate": 9.64348698740602e-06, "loss": 3.4269, "step": 1447 }, { "epoch": 0.17, "learning_rate": 9.642742760297114e-06, "loss": 3.4566, "step": 1448 }, { "epoch": 0.17, "learning_rate": 9.641997785982961e-06, "loss": 3.3932, "step": 1449 }, { "epoch": 0.17, "learning_rate": 9.64125206458346e-06, "loss": 3.4514, "step": 1450 }, { "epoch": 0.17, "learning_rate": 9.640505596218622e-06, "loss": 3.4266, "step": 1451 }, { "epoch": 0.17, "learning_rate": 9.639758381008587e-06, "loss": 3.3689, "step": 1452 }, { "epoch": 0.17, "learning_rate": 9.639010419073612e-06, "loss": 3.3781, "step": 1453 }, { "epoch": 0.17, "learning_rate": 9.638261710534076e-06, "loss": 3.4063, "step": 1454 }, { "epoch": 0.17, "learning_rate": 9.637512255510475e-06, "loss": 3.4765, "step": 1455 }, { "epoch": 0.17, "learning_rate": 9.636762054123425e-06, "loss": 3.4674, "step": 1456 }, { "epoch": 0.17, "learning_rate": 9.636011106493667e-06, "loss": 3.4617, "step": 1457 }, { "epoch": 0.18, "learning_rate": 9.635259412742056e-06, "loss": 3.4603, "step": 1458 }, { "epoch": 0.18, "learning_rate": 9.634506972989573e-06, "loss": 3.5277, "step": 1459 }, { "epoch": 0.18, "learning_rate": 9.633753787357313e-06, "loss": 3.4215, "step": 1460 }, { "epoch": 0.18, "learning_rate": 9.632999855966497e-06, "loss": 3.4935, "step": 1461 }, { "epoch": 0.18, "learning_rate": 9.632245178938458e-06, "loss": 3.3536, "step": 1462 }, { "epoch": 0.18, "learning_rate": 9.631489756394661e-06, "loss": 3.4681, "step": 1463 }, { "epoch": 0.18, "learning_rate": 9.630733588456678e-06, "loss": 3.4204, "step": 1464 }, { "epoch": 0.18, "learning_rate": 9.62997667524621e-06, "loss": 3.4074, "step": 1465 }, { "epoch": 0.18, "learning_rate": 9.629219016885076e-06, "loss": 3.3991, "step": 1466 }, { "epoch": 0.18, "learning_rate": 9.628460613495213e-06, "loss": 3.3918, "step": 1467 }, { "epoch": 0.18, "learning_rate": 9.627701465198677e-06, "loss": 3.4735, "step": 1468 }, { "epoch": 0.18, "learning_rate": 9.626941572117647e-06, "loss": 3.503, "step": 1469 }, { "epoch": 0.18, "learning_rate": 9.62618093437442e-06, "loss": 3.5665, "step": 1470 }, { "epoch": 0.18, "learning_rate": 9.625419552091414e-06, "loss": 3.31, "step": 1471 }, { "epoch": 0.18, "learning_rate": 9.624657425391168e-06, "loss": 3.5379, "step": 1472 }, { "epoch": 0.18, "learning_rate": 9.623894554396337e-06, "loss": 3.4248, "step": 1473 }, { "epoch": 0.18, "learning_rate": 9.623130939229697e-06, "loss": 3.4668, "step": 1474 }, { "epoch": 0.18, "learning_rate": 9.622366580014147e-06, "loss": 3.426, "step": 1475 }, { "epoch": 0.18, "learning_rate": 9.621601476872699e-06, "loss": 3.3912, "step": 1476 }, { "epoch": 0.18, "learning_rate": 9.620835629928493e-06, "loss": 3.4882, "step": 1477 }, { "epoch": 0.18, "learning_rate": 9.620069039304785e-06, "loss": 3.4656, "step": 1478 }, { "epoch": 0.18, "learning_rate": 9.61930170512495e-06, "loss": 3.4178, "step": 1479 }, { "epoch": 0.18, "learning_rate": 9.61853362751248e-06, "loss": 3.4317, "step": 1480 }, { "epoch": 0.18, "learning_rate": 9.617764806590993e-06, "loss": 3.4699, "step": 1481 }, { "epoch": 0.18, "learning_rate": 9.616995242484224e-06, "loss": 3.4795, "step": 1482 }, { "epoch": 0.18, "learning_rate": 9.616224935316024e-06, "loss": 3.3964, "step": 1483 }, { "epoch": 0.18, "learning_rate": 9.615453885210368e-06, "loss": 3.3472, "step": 1484 }, { "epoch": 0.18, "learning_rate": 9.614682092291352e-06, "loss": 3.4467, "step": 1485 }, { "epoch": 0.18, "learning_rate": 9.613909556683183e-06, "loss": 3.4608, "step": 1486 }, { "epoch": 0.18, "learning_rate": 9.613136278510198e-06, "loss": 3.394, "step": 1487 }, { "epoch": 0.18, "learning_rate": 9.612362257896847e-06, "loss": 3.3833, "step": 1488 }, { "epoch": 0.18, "learning_rate": 9.6115874949677e-06, "loss": 3.4486, "step": 1489 }, { "epoch": 0.18, "learning_rate": 9.61081198984745e-06, "loss": 3.3885, "step": 1490 }, { "epoch": 0.18, "learning_rate": 9.610035742660907e-06, "loss": 3.504, "step": 1491 }, { "epoch": 0.18, "learning_rate": 9.609258753533e-06, "loss": 3.3729, "step": 1492 }, { "epoch": 0.18, "learning_rate": 9.608481022588776e-06, "loss": 3.5141, "step": 1493 }, { "epoch": 0.18, "learning_rate": 9.607702549953405e-06, "loss": 3.3721, "step": 1494 }, { "epoch": 0.18, "learning_rate": 9.606923335752179e-06, "loss": 3.3718, "step": 1495 }, { "epoch": 0.18, "learning_rate": 9.606143380110498e-06, "loss": 3.392, "step": 1496 }, { "epoch": 0.18, "learning_rate": 9.605362683153892e-06, "loss": 3.4153, "step": 1497 }, { "epoch": 0.18, "learning_rate": 9.604581245008006e-06, "loss": 3.4949, "step": 1498 }, { "epoch": 0.18, "learning_rate": 9.603799065798605e-06, "loss": 3.4642, "step": 1499 }, { "epoch": 0.18, "learning_rate": 9.603016145651574e-06, "loss": 3.406, "step": 1500 }, { "epoch": 0.18, "learning_rate": 9.602232484692917e-06, "loss": 3.4382, "step": 1501 }, { "epoch": 0.18, "learning_rate": 9.601448083048754e-06, "loss": 3.5427, "step": 1502 }, { "epoch": 0.18, "learning_rate": 9.600662940845331e-06, "loss": 3.4656, "step": 1503 }, { "epoch": 0.18, "learning_rate": 9.599877058209007e-06, "loss": 3.4885, "step": 1504 }, { "epoch": 0.18, "learning_rate": 9.59909043526626e-06, "loss": 3.4854, "step": 1505 }, { "epoch": 0.18, "learning_rate": 9.598303072143694e-06, "loss": 3.4486, "step": 1506 }, { "epoch": 0.18, "learning_rate": 9.597514968968025e-06, "loss": 3.4924, "step": 1507 }, { "epoch": 0.18, "learning_rate": 9.596726125866091e-06, "loss": 3.4592, "step": 1508 }, { "epoch": 0.18, "learning_rate": 9.59593654296485e-06, "loss": 3.4535, "step": 1509 }, { "epoch": 0.18, "learning_rate": 9.595146220391376e-06, "loss": 3.3892, "step": 1510 }, { "epoch": 0.18, "learning_rate": 9.594355158272863e-06, "loss": 3.484, "step": 1511 }, { "epoch": 0.18, "learning_rate": 9.593563356736628e-06, "loss": 3.5407, "step": 1512 }, { "epoch": 0.18, "learning_rate": 9.592770815910104e-06, "loss": 3.4032, "step": 1513 }, { "epoch": 0.18, "learning_rate": 9.59197753592084e-06, "loss": 3.4091, "step": 1514 }, { "epoch": 0.18, "learning_rate": 9.591183516896509e-06, "loss": 3.4534, "step": 1515 }, { "epoch": 0.18, "learning_rate": 9.5903887589649e-06, "loss": 3.4157, "step": 1516 }, { "epoch": 0.18, "learning_rate": 9.58959326225392e-06, "loss": 3.4075, "step": 1517 }, { "epoch": 0.18, "learning_rate": 9.5887970268916e-06, "loss": 3.3596, "step": 1518 }, { "epoch": 0.18, "learning_rate": 9.588000053006084e-06, "loss": 3.5027, "step": 1519 }, { "epoch": 0.18, "learning_rate": 9.58720234072564e-06, "loss": 3.5461, "step": 1520 }, { "epoch": 0.18, "learning_rate": 9.586403890178649e-06, "loss": 3.392, "step": 1521 }, { "epoch": 0.18, "learning_rate": 9.585604701493616e-06, "loss": 3.4115, "step": 1522 }, { "epoch": 0.18, "learning_rate": 9.584804774799163e-06, "loss": 3.4027, "step": 1523 }, { "epoch": 0.18, "learning_rate": 9.58400411022403e-06, "loss": 3.489, "step": 1524 }, { "epoch": 0.18, "learning_rate": 9.583202707897075e-06, "loss": 3.4793, "step": 1525 }, { "epoch": 0.18, "learning_rate": 9.582400567947277e-06, "loss": 3.3496, "step": 1526 }, { "epoch": 0.18, "learning_rate": 9.581597690503734e-06, "loss": 3.4409, "step": 1527 }, { "epoch": 0.18, "learning_rate": 9.58079407569566e-06, "loss": 3.3904, "step": 1528 }, { "epoch": 0.18, "learning_rate": 9.579989723652391e-06, "loss": 3.4115, "step": 1529 }, { "epoch": 0.18, "learning_rate": 9.579184634503378e-06, "loss": 3.3815, "step": 1530 }, { "epoch": 0.18, "learning_rate": 9.578378808378193e-06, "loss": 3.4932, "step": 1531 }, { "epoch": 0.18, "learning_rate": 9.577572245406525e-06, "loss": 3.4761, "step": 1532 }, { "epoch": 0.18, "learning_rate": 9.576764945718187e-06, "loss": 3.4202, "step": 1533 }, { "epoch": 0.18, "learning_rate": 9.5759569094431e-06, "loss": 3.4866, "step": 1534 }, { "epoch": 0.18, "learning_rate": 9.575148136711311e-06, "loss": 3.3564, "step": 1535 }, { "epoch": 0.18, "learning_rate": 9.57433862765299e-06, "loss": 3.4314, "step": 1536 }, { "epoch": 0.18, "learning_rate": 9.573528382398414e-06, "loss": 3.4551, "step": 1537 }, { "epoch": 0.18, "learning_rate": 9.572717401077985e-06, "loss": 3.4193, "step": 1538 }, { "epoch": 0.18, "learning_rate": 9.571905683822225e-06, "loss": 3.4663, "step": 1539 }, { "epoch": 0.18, "learning_rate": 9.57109323076177e-06, "loss": 3.4671, "step": 1540 }, { "epoch": 0.18, "learning_rate": 9.570280042027377e-06, "loss": 3.4067, "step": 1541 }, { "epoch": 0.19, "learning_rate": 9.569466117749922e-06, "loss": 3.4581, "step": 1542 }, { "epoch": 0.19, "learning_rate": 9.568651458060396e-06, "loss": 3.35, "step": 1543 }, { "epoch": 0.19, "learning_rate": 9.567836063089914e-06, "loss": 3.4644, "step": 1544 }, { "epoch": 0.19, "learning_rate": 9.567019932969705e-06, "loss": 3.5095, "step": 1545 }, { "epoch": 0.19, "learning_rate": 9.566203067831117e-06, "loss": 3.4563, "step": 1546 }, { "epoch": 0.19, "learning_rate": 9.565385467805616e-06, "loss": 3.4189, "step": 1547 }, { "epoch": 0.19, "learning_rate": 9.564567133024788e-06, "loss": 3.3808, "step": 1548 }, { "epoch": 0.19, "learning_rate": 9.563748063620334e-06, "loss": 3.4359, "step": 1549 }, { "epoch": 0.19, "learning_rate": 9.562928259724077e-06, "loss": 3.4399, "step": 1550 }, { "epoch": 0.19, "learning_rate": 9.562107721467958e-06, "loss": 3.3818, "step": 1551 }, { "epoch": 0.19, "learning_rate": 9.561286448984035e-06, "loss": 3.4687, "step": 1552 }, { "epoch": 0.19, "learning_rate": 9.560464442404481e-06, "loss": 3.4425, "step": 1553 }, { "epoch": 0.19, "learning_rate": 9.559641701861591e-06, "loss": 3.4422, "step": 1554 }, { "epoch": 0.19, "learning_rate": 9.55881822748778e-06, "loss": 3.4222, "step": 1555 }, { "epoch": 0.19, "learning_rate": 9.557994019415575e-06, "loss": 3.5151, "step": 1556 }, { "epoch": 0.19, "learning_rate": 9.557169077777625e-06, "loss": 3.3644, "step": 1557 }, { "epoch": 0.19, "learning_rate": 9.556343402706698e-06, "loss": 3.3824, "step": 1558 }, { "epoch": 0.19, "learning_rate": 9.555516994335679e-06, "loss": 3.5589, "step": 1559 }, { "epoch": 0.19, "learning_rate": 9.554689852797567e-06, "loss": 3.4799, "step": 1560 }, { "epoch": 0.19, "learning_rate": 9.553861978225488e-06, "loss": 3.4394, "step": 1561 }, { "epoch": 0.19, "learning_rate": 9.553033370752674e-06, "loss": 3.4301, "step": 1562 }, { "epoch": 0.19, "learning_rate": 9.552204030512486e-06, "loss": 3.5592, "step": 1563 }, { "epoch": 0.19, "learning_rate": 9.551373957638399e-06, "loss": 3.4322, "step": 1564 }, { "epoch": 0.19, "learning_rate": 9.550543152264002e-06, "loss": 3.4182, "step": 1565 }, { "epoch": 0.19, "learning_rate": 9.549711614523007e-06, "loss": 3.5068, "step": 1566 }, { "epoch": 0.19, "learning_rate": 9.548879344549242e-06, "loss": 3.388, "step": 1567 }, { "epoch": 0.19, "learning_rate": 9.548046342476654e-06, "loss": 3.4936, "step": 1568 }, { "epoch": 0.19, "learning_rate": 9.547212608439304e-06, "loss": 3.4811, "step": 1569 }, { "epoch": 0.19, "learning_rate": 9.546378142571378e-06, "loss": 3.3881, "step": 1570 }, { "epoch": 0.19, "learning_rate": 9.54554294500717e-06, "loss": 3.5788, "step": 1571 }, { "epoch": 0.19, "learning_rate": 9.5447070158811e-06, "loss": 3.4451, "step": 1572 }, { "epoch": 0.19, "learning_rate": 9.543870355327702e-06, "loss": 3.4423, "step": 1573 }, { "epoch": 0.19, "learning_rate": 9.543032963481632e-06, "loss": 3.4538, "step": 1574 }, { "epoch": 0.19, "learning_rate": 9.542194840477653e-06, "loss": 3.4334, "step": 1575 }, { "epoch": 0.19, "learning_rate": 9.54135598645066e-06, "loss": 3.4514, "step": 1576 }, { "epoch": 0.19, "learning_rate": 9.540516401535655e-06, "loss": 3.4933, "step": 1577 }, { "epoch": 0.19, "learning_rate": 9.539676085867762e-06, "loss": 3.4933, "step": 1578 }, { "epoch": 0.19, "learning_rate": 9.538835039582222e-06, "loss": 3.3279, "step": 1579 }, { "epoch": 0.19, "learning_rate": 9.537993262814391e-06, "loss": 3.4322, "step": 1580 }, { "epoch": 0.19, "learning_rate": 9.53715075569975e-06, "loss": 3.4792, "step": 1581 }, { "epoch": 0.19, "learning_rate": 9.536307518373889e-06, "loss": 3.4766, "step": 1582 }, { "epoch": 0.19, "learning_rate": 9.535463550972519e-06, "loss": 3.5789, "step": 1583 }, { "epoch": 0.19, "learning_rate": 9.534618853631469e-06, "loss": 3.3859, "step": 1584 }, { "epoch": 0.19, "learning_rate": 9.533773426486686e-06, "loss": 3.5408, "step": 1585 }, { "epoch": 0.19, "learning_rate": 9.532927269674232e-06, "loss": 3.5099, "step": 1586 }, { "epoch": 0.19, "learning_rate": 9.532080383330288e-06, "loss": 3.3738, "step": 1587 }, { "epoch": 0.19, "learning_rate": 9.531232767591154e-06, "loss": 3.5878, "step": 1588 }, { "epoch": 0.19, "learning_rate": 9.530384422593242e-06, "loss": 3.3842, "step": 1589 }, { "epoch": 0.19, "learning_rate": 9.529535348473092e-06, "loss": 3.4176, "step": 1590 }, { "epoch": 0.19, "learning_rate": 9.528685545367345e-06, "loss": 3.4834, "step": 1591 }, { "epoch": 0.19, "learning_rate": 9.527835013412776e-06, "loss": 3.474, "step": 1592 }, { "epoch": 0.19, "learning_rate": 9.526983752746268e-06, "loss": 3.3408, "step": 1593 }, { "epoch": 0.19, "learning_rate": 9.526131763504823e-06, "loss": 3.3994, "step": 1594 }, { "epoch": 0.19, "learning_rate": 9.52527904582556e-06, "loss": 3.4832, "step": 1595 }, { "epoch": 0.19, "learning_rate": 9.524425599845714e-06, "loss": 3.5201, "step": 1596 }, { "epoch": 0.19, "learning_rate": 9.523571425702644e-06, "loss": 3.4951, "step": 1597 }, { "epoch": 0.19, "learning_rate": 9.522716523533819e-06, "loss": 3.4819, "step": 1598 }, { "epoch": 0.19, "learning_rate": 9.521860893476826e-06, "loss": 3.4048, "step": 1599 }, { "epoch": 0.19, "learning_rate": 9.521004535669372e-06, "loss": 3.4242, "step": 1600 }, { "epoch": 0.19, "learning_rate": 9.520147450249278e-06, "loss": 3.5183, "step": 1601 }, { "epoch": 0.19, "learning_rate": 9.519289637354483e-06, "loss": 3.4063, "step": 1602 }, { "epoch": 0.19, "learning_rate": 9.51843109712305e-06, "loss": 3.4992, "step": 1603 }, { "epoch": 0.19, "learning_rate": 9.517571829693145e-06, "loss": 3.4125, "step": 1604 }, { "epoch": 0.19, "learning_rate": 9.516711835203063e-06, "loss": 3.4313, "step": 1605 }, { "epoch": 0.19, "learning_rate": 9.515851113791212e-06, "loss": 3.3843, "step": 1606 }, { "epoch": 0.19, "learning_rate": 9.514989665596114e-06, "loss": 3.3601, "step": 1607 }, { "epoch": 0.19, "learning_rate": 9.514127490756417e-06, "loss": 3.4325, "step": 1608 }, { "epoch": 0.19, "learning_rate": 9.513264589410874e-06, "loss": 3.5257, "step": 1609 }, { "epoch": 0.19, "learning_rate": 9.512400961698363e-06, "loss": 3.328, "step": 1610 }, { "epoch": 0.19, "learning_rate": 9.511536607757875e-06, "loss": 3.4398, "step": 1611 }, { "epoch": 0.19, "learning_rate": 9.51067152772852e-06, "loss": 3.416, "step": 1612 }, { "epoch": 0.19, "learning_rate": 9.50980572174953e-06, "loss": 3.4926, "step": 1613 }, { "epoch": 0.19, "learning_rate": 9.508939189960238e-06, "loss": 3.4217, "step": 1614 }, { "epoch": 0.19, "learning_rate": 9.508071932500114e-06, "loss": 3.4025, "step": 1615 }, { "epoch": 0.19, "learning_rate": 9.507203949508727e-06, "loss": 3.4119, "step": 1616 }, { "epoch": 0.19, "learning_rate": 9.506335241125775e-06, "loss": 3.5379, "step": 1617 }, { "epoch": 0.19, "learning_rate": 9.505465807491067e-06, "loss": 3.4391, "step": 1618 }, { "epoch": 0.19, "learning_rate": 9.504595648744532e-06, "loss": 3.4948, "step": 1619 }, { "epoch": 0.19, "learning_rate": 9.503724765026212e-06, "loss": 3.5049, "step": 1620 }, { "epoch": 0.19, "learning_rate": 9.502853156476265e-06, "loss": 3.4099, "step": 1621 }, { "epoch": 0.19, "learning_rate": 9.501980823234974e-06, "loss": 3.4737, "step": 1622 }, { "epoch": 0.19, "learning_rate": 9.501107765442728e-06, "loss": 3.5206, "step": 1623 }, { "epoch": 0.19, "learning_rate": 9.50023398324004e-06, "loss": 3.4142, "step": 1624 }, { "epoch": 0.2, "learning_rate": 9.499359476767533e-06, "loss": 3.366, "step": 1625 }, { "epoch": 0.2, "learning_rate": 9.498484246165954e-06, "loss": 3.3734, "step": 1626 }, { "epoch": 0.2, "learning_rate": 9.497608291576163e-06, "loss": 3.4634, "step": 1627 }, { "epoch": 0.2, "learning_rate": 9.496731613139136e-06, "loss": 3.4052, "step": 1628 }, { "epoch": 0.2, "learning_rate": 9.495854210995966e-06, "loss": 3.4031, "step": 1629 }, { "epoch": 0.2, "learning_rate": 9.49497608528786e-06, "loss": 3.48, "step": 1630 }, { "epoch": 0.2, "learning_rate": 9.494097236156148e-06, "loss": 3.4758, "step": 1631 }, { "epoch": 0.2, "learning_rate": 9.49321766374227e-06, "loss": 3.4149, "step": 1632 }, { "epoch": 0.2, "learning_rate": 9.492337368187787e-06, "loss": 3.4168, "step": 1633 }, { "epoch": 0.2, "learning_rate": 9.491456349634372e-06, "loss": 3.422, "step": 1634 }, { "epoch": 0.2, "learning_rate": 9.490574608223816e-06, "loss": 3.3236, "step": 1635 }, { "epoch": 0.2, "learning_rate": 9.48969214409803e-06, "loss": 3.3599, "step": 1636 }, { "epoch": 0.2, "learning_rate": 9.488808957399036e-06, "loss": 3.4549, "step": 1637 }, { "epoch": 0.2, "learning_rate": 9.487925048268973e-06, "loss": 3.3335, "step": 1638 }, { "epoch": 0.2, "learning_rate": 9.487040416850103e-06, "loss": 3.319, "step": 1639 }, { "epoch": 0.2, "learning_rate": 9.486155063284792e-06, "loss": 3.4662, "step": 1640 }, { "epoch": 0.2, "learning_rate": 9.485268987715534e-06, "loss": 3.5123, "step": 1641 }, { "epoch": 0.2, "learning_rate": 9.484382190284934e-06, "loss": 3.4385, "step": 1642 }, { "epoch": 0.2, "learning_rate": 9.483494671135712e-06, "loss": 3.4078, "step": 1643 }, { "epoch": 0.2, "learning_rate": 9.482606430410708e-06, "loss": 3.4256, "step": 1644 }, { "epoch": 0.2, "learning_rate": 9.481717468252874e-06, "loss": 3.4088, "step": 1645 }, { "epoch": 0.2, "learning_rate": 9.480827784805278e-06, "loss": 3.3292, "step": 1646 }, { "epoch": 0.2, "learning_rate": 9.479937380211111e-06, "loss": 3.4666, "step": 1647 }, { "epoch": 0.2, "learning_rate": 9.479046254613673e-06, "loss": 3.3527, "step": 1648 }, { "epoch": 0.2, "learning_rate": 9.478154408156381e-06, "loss": 3.507, "step": 1649 }, { "epoch": 0.2, "learning_rate": 9.47726184098277e-06, "loss": 3.4043, "step": 1650 }, { "epoch": 0.2, "learning_rate": 9.47636855323649e-06, "loss": 3.4196, "step": 1651 }, { "epoch": 0.2, "learning_rate": 9.475474545061309e-06, "loss": 3.4556, "step": 1652 }, { "epoch": 0.2, "learning_rate": 9.474579816601107e-06, "loss": 3.3754, "step": 1653 }, { "epoch": 0.2, "learning_rate": 9.473684367999881e-06, "loss": 3.542, "step": 1654 }, { "epoch": 0.2, "learning_rate": 9.472788199401748e-06, "loss": 3.4463, "step": 1655 }, { "epoch": 0.2, "learning_rate": 9.471891310950935e-06, "loss": 3.4296, "step": 1656 }, { "epoch": 0.2, "learning_rate": 9.47099370279179e-06, "loss": 3.4393, "step": 1657 }, { "epoch": 0.2, "learning_rate": 9.470095375068773e-06, "loss": 3.4623, "step": 1658 }, { "epoch": 0.2, "learning_rate": 9.469196327926462e-06, "loss": 3.4243, "step": 1659 }, { "epoch": 0.2, "learning_rate": 9.468296561509551e-06, "loss": 3.4568, "step": 1660 }, { "epoch": 0.2, "learning_rate": 9.467396075962846e-06, "loss": 3.5269, "step": 1661 }, { "epoch": 0.2, "learning_rate": 9.466494871431272e-06, "loss": 3.4702, "step": 1662 }, { "epoch": 0.2, "learning_rate": 9.465592948059874e-06, "loss": 3.3759, "step": 1663 }, { "epoch": 0.2, "learning_rate": 9.464690305993802e-06, "loss": 3.4287, "step": 1664 }, { "epoch": 0.2, "learning_rate": 9.463786945378332e-06, "loss": 3.3928, "step": 1665 }, { "epoch": 0.2, "learning_rate": 9.462882866358847e-06, "loss": 3.428, "step": 1666 }, { "epoch": 0.2, "learning_rate": 9.461978069080855e-06, "loss": 3.368, "step": 1667 }, { "epoch": 0.2, "learning_rate": 9.461072553689969e-06, "loss": 3.3669, "step": 1668 }, { "epoch": 0.2, "learning_rate": 9.46016632033193e-06, "loss": 3.4611, "step": 1669 }, { "epoch": 0.2, "learning_rate": 9.45925936915258e-06, "loss": 3.3786, "step": 1670 }, { "epoch": 0.2, "learning_rate": 9.45835170029789e-06, "loss": 3.4772, "step": 1671 }, { "epoch": 0.2, "learning_rate": 9.457443313913937e-06, "loss": 3.3859, "step": 1672 }, { "epoch": 0.2, "learning_rate": 9.456534210146917e-06, "loss": 3.4724, "step": 1673 }, { "epoch": 0.2, "learning_rate": 9.455624389143145e-06, "loss": 3.376, "step": 1674 }, { "epoch": 0.2, "learning_rate": 9.454713851049047e-06, "loss": 3.426, "step": 1675 }, { "epoch": 0.2, "learning_rate": 9.453802596011164e-06, "loss": 3.4729, "step": 1676 }, { "epoch": 0.2, "learning_rate": 9.452890624176153e-06, "loss": 3.3647, "step": 1677 }, { "epoch": 0.2, "learning_rate": 9.45197793569079e-06, "loss": 3.4278, "step": 1678 }, { "epoch": 0.2, "learning_rate": 9.45106453070196e-06, "loss": 3.4492, "step": 1679 }, { "epoch": 0.2, "learning_rate": 9.450150409356673e-06, "loss": 3.3878, "step": 1680 }, { "epoch": 0.2, "learning_rate": 9.44923557180204e-06, "loss": 3.3913, "step": 1681 }, { "epoch": 0.2, "learning_rate": 9.448320018185301e-06, "loss": 3.4491, "step": 1682 }, { "epoch": 0.2, "learning_rate": 9.447403748653804e-06, "loss": 3.4181, "step": 1683 }, { "epoch": 0.2, "learning_rate": 9.446486763355017e-06, "loss": 3.4159, "step": 1684 }, { "epoch": 0.2, "learning_rate": 9.445569062436513e-06, "loss": 3.4226, "step": 1685 }, { "epoch": 0.2, "learning_rate": 9.444650646045994e-06, "loss": 3.4174, "step": 1686 }, { "epoch": 0.2, "learning_rate": 9.443731514331266e-06, "loss": 3.3498, "step": 1687 }, { "epoch": 0.2, "learning_rate": 9.442811667440255e-06, "loss": 3.3348, "step": 1688 }, { "epoch": 0.2, "learning_rate": 9.441891105521005e-06, "loss": 3.5452, "step": 1689 }, { "epoch": 0.2, "learning_rate": 9.440969828721669e-06, "loss": 3.3786, "step": 1690 }, { "epoch": 0.2, "learning_rate": 9.440047837190517e-06, "loss": 3.5017, "step": 1691 }, { "epoch": 0.2, "learning_rate": 9.439125131075937e-06, "loss": 3.5154, "step": 1692 }, { "epoch": 0.2, "learning_rate": 9.438201710526429e-06, "loss": 3.4319, "step": 1693 }, { "epoch": 0.2, "learning_rate": 9.437277575690607e-06, "loss": 3.4862, "step": 1694 }, { "epoch": 0.2, "learning_rate": 9.436352726717205e-06, "loss": 3.5006, "step": 1695 }, { "epoch": 0.2, "learning_rate": 9.435427163755065e-06, "loss": 3.4384, "step": 1696 }, { "epoch": 0.2, "learning_rate": 9.43450088695315e-06, "loss": 3.5187, "step": 1697 }, { "epoch": 0.2, "learning_rate": 9.433573896460535e-06, "loss": 3.3585, "step": 1698 }, { "epoch": 0.2, "learning_rate": 9.432646192426408e-06, "loss": 3.4882, "step": 1699 }, { "epoch": 0.2, "learning_rate": 9.43171777500008e-06, "loss": 3.429, "step": 1700 }, { "epoch": 0.2, "learning_rate": 9.430788644330963e-06, "loss": 3.4077, "step": 1701 }, { "epoch": 0.2, "learning_rate": 9.429858800568599e-06, "loss": 3.4267, "step": 1702 }, { "epoch": 0.2, "learning_rate": 9.428928243862634e-06, "loss": 3.4571, "step": 1703 }, { "epoch": 0.2, "learning_rate": 9.42799697436283e-06, "loss": 3.4328, "step": 1704 }, { "epoch": 0.2, "learning_rate": 9.427064992219071e-06, "loss": 3.357, "step": 1705 }, { "epoch": 0.2, "learning_rate": 9.426132297581347e-06, "loss": 3.4476, "step": 1706 }, { "epoch": 0.2, "learning_rate": 9.425198890599769e-06, "loss": 3.4495, "step": 1707 }, { "epoch": 0.21, "learning_rate": 9.424264771424557e-06, "loss": 3.4854, "step": 1708 }, { "epoch": 0.21, "learning_rate": 9.423329940206052e-06, "loss": 3.4296, "step": 1709 }, { "epoch": 0.21, "learning_rate": 9.422394397094702e-06, "loss": 3.4984, "step": 1710 }, { "epoch": 0.21, "learning_rate": 9.421458142241077e-06, "loss": 3.3472, "step": 1711 }, { "epoch": 0.21, "learning_rate": 9.420521175795857e-06, "loss": 3.4815, "step": 1712 }, { "epoch": 0.21, "learning_rate": 9.419583497909838e-06, "loss": 3.4209, "step": 1713 }, { "epoch": 0.21, "learning_rate": 9.41864510873393e-06, "loss": 3.424, "step": 1714 }, { "epoch": 0.21, "learning_rate": 9.417706008419161e-06, "loss": 3.3175, "step": 1715 }, { "epoch": 0.21, "learning_rate": 9.416766197116664e-06, "loss": 3.403, "step": 1716 }, { "epoch": 0.21, "learning_rate": 9.415825674977698e-06, "loss": 3.3791, "step": 1717 }, { "epoch": 0.21, "learning_rate": 9.41488444215363e-06, "loss": 3.4445, "step": 1718 }, { "epoch": 0.21, "learning_rate": 9.41394249879594e-06, "loss": 3.4286, "step": 1719 }, { "epoch": 0.21, "learning_rate": 9.412999845056228e-06, "loss": 3.3823, "step": 1720 }, { "epoch": 0.21, "learning_rate": 9.412056481086205e-06, "loss": 3.3525, "step": 1721 }, { "epoch": 0.21, "learning_rate": 9.411112407037693e-06, "loss": 3.3421, "step": 1722 }, { "epoch": 0.21, "learning_rate": 9.410167623062633e-06, "loss": 3.4643, "step": 1723 }, { "epoch": 0.21, "learning_rate": 9.409222129313083e-06, "loss": 3.4846, "step": 1724 }, { "epoch": 0.21, "learning_rate": 9.408275925941206e-06, "loss": 3.4781, "step": 1725 }, { "epoch": 0.21, "learning_rate": 9.407329013099286e-06, "loss": 3.3815, "step": 1726 }, { "epoch": 0.21, "learning_rate": 9.406381390939722e-06, "loss": 3.3876, "step": 1727 }, { "epoch": 0.21, "learning_rate": 9.40543305961502e-06, "loss": 3.3684, "step": 1728 }, { "epoch": 0.21, "learning_rate": 9.404484019277812e-06, "loss": 3.5179, "step": 1729 }, { "epoch": 0.21, "learning_rate": 9.40353427008083e-06, "loss": 3.4926, "step": 1730 }, { "epoch": 0.21, "learning_rate": 9.40258381217693e-06, "loss": 3.368, "step": 1731 }, { "epoch": 0.21, "learning_rate": 9.401632645719081e-06, "loss": 3.4065, "step": 1732 }, { "epoch": 0.21, "learning_rate": 9.40068077086036e-06, "loss": 3.4702, "step": 1733 }, { "epoch": 0.21, "learning_rate": 9.399728187753966e-06, "loss": 3.484, "step": 1734 }, { "epoch": 0.21, "learning_rate": 9.398774896553205e-06, "loss": 3.4236, "step": 1735 }, { "epoch": 0.21, "learning_rate": 9.397820897411504e-06, "loss": 3.483, "step": 1736 }, { "epoch": 0.21, "learning_rate": 9.396866190482397e-06, "loss": 3.3362, "step": 1737 }, { "epoch": 0.21, "learning_rate": 9.395910775919534e-06, "loss": 3.3974, "step": 1738 }, { "epoch": 0.21, "learning_rate": 9.394954653876684e-06, "loss": 3.4844, "step": 1739 }, { "epoch": 0.21, "learning_rate": 9.393997824507722e-06, "loss": 3.3774, "step": 1740 }, { "epoch": 0.21, "learning_rate": 9.393040287966642e-06, "loss": 3.4054, "step": 1741 }, { "epoch": 0.21, "learning_rate": 9.39208204440755e-06, "loss": 3.3333, "step": 1742 }, { "epoch": 0.21, "learning_rate": 9.39112309398467e-06, "loss": 3.4571, "step": 1743 }, { "epoch": 0.21, "learning_rate": 9.39016343685233e-06, "loss": 3.5056, "step": 1744 }, { "epoch": 0.21, "learning_rate": 9.389203073164981e-06, "loss": 3.441, "step": 1745 }, { "epoch": 0.21, "learning_rate": 9.388242003077182e-06, "loss": 3.4039, "step": 1746 }, { "epoch": 0.21, "learning_rate": 9.387280226743611e-06, "loss": 3.4494, "step": 1747 }, { "epoch": 0.21, "learning_rate": 9.386317744319056e-06, "loss": 3.4616, "step": 1748 }, { "epoch": 0.21, "learning_rate": 9.385354555958417e-06, "loss": 3.4569, "step": 1749 }, { "epoch": 0.21, "learning_rate": 9.384390661816714e-06, "loss": 3.4096, "step": 1750 }, { "epoch": 0.21, "learning_rate": 9.383426062049072e-06, "loss": 3.5294, "step": 1751 }, { "epoch": 0.21, "learning_rate": 9.38246075681074e-06, "loss": 3.4411, "step": 1752 }, { "epoch": 0.21, "learning_rate": 9.381494746257067e-06, "loss": 3.3914, "step": 1753 }, { "epoch": 0.21, "learning_rate": 9.38052803054353e-06, "loss": 3.4424, "step": 1754 }, { "epoch": 0.21, "learning_rate": 9.379560609825709e-06, "loss": 3.4147, "step": 1755 }, { "epoch": 0.21, "learning_rate": 9.378592484259304e-06, "loss": 3.4344, "step": 1756 }, { "epoch": 0.21, "learning_rate": 9.377623654000121e-06, "loss": 3.4125, "step": 1757 }, { "epoch": 0.21, "learning_rate": 9.376654119204089e-06, "loss": 3.3825, "step": 1758 }, { "epoch": 0.21, "learning_rate": 9.375683880027245e-06, "loss": 3.4298, "step": 1759 }, { "epoch": 0.21, "learning_rate": 9.374712936625734e-06, "loss": 3.469, "step": 1760 }, { "epoch": 0.21, "learning_rate": 9.373741289155829e-06, "loss": 3.4395, "step": 1761 }, { "epoch": 0.21, "learning_rate": 9.372768937773898e-06, "loss": 3.4236, "step": 1762 }, { "epoch": 0.21, "learning_rate": 9.37179588263644e-06, "loss": 3.5094, "step": 1763 }, { "epoch": 0.21, "learning_rate": 9.370822123900051e-06, "loss": 3.4864, "step": 1764 }, { "epoch": 0.21, "learning_rate": 9.369847661721456e-06, "loss": 3.4527, "step": 1765 }, { "epoch": 0.21, "learning_rate": 9.368872496257479e-06, "loss": 3.3789, "step": 1766 }, { "epoch": 0.21, "learning_rate": 9.367896627665068e-06, "loss": 3.3997, "step": 1767 }, { "epoch": 0.21, "learning_rate": 9.366920056101279e-06, "loss": 3.4006, "step": 1768 }, { "epoch": 0.21, "learning_rate": 9.36594278172328e-06, "loss": 3.4735, "step": 1769 }, { "epoch": 0.21, "learning_rate": 9.364964804688355e-06, "loss": 3.4386, "step": 1770 }, { "epoch": 0.21, "learning_rate": 9.3639861251539e-06, "loss": 3.4679, "step": 1771 }, { "epoch": 0.21, "learning_rate": 9.363006743277425e-06, "loss": 3.4095, "step": 1772 }, { "epoch": 0.21, "learning_rate": 9.362026659216551e-06, "loss": 3.4117, "step": 1773 }, { "epoch": 0.21, "learning_rate": 9.361045873129015e-06, "loss": 3.4518, "step": 1774 }, { "epoch": 0.21, "learning_rate": 9.360064385172663e-06, "loss": 3.4192, "step": 1775 }, { "epoch": 0.21, "learning_rate": 9.359082195505456e-06, "loss": 3.4587, "step": 1776 }, { "epoch": 0.21, "learning_rate": 9.35809930428547e-06, "loss": 3.3852, "step": 1777 }, { "epoch": 0.21, "learning_rate": 9.35711571167089e-06, "loss": 3.4736, "step": 1778 }, { "epoch": 0.21, "learning_rate": 9.35613141782002e-06, "loss": 3.4475, "step": 1779 }, { "epoch": 0.21, "learning_rate": 9.355146422891267e-06, "loss": 3.4128, "step": 1780 }, { "epoch": 0.21, "learning_rate": 9.35416072704316e-06, "loss": 3.4085, "step": 1781 }, { "epoch": 0.21, "learning_rate": 9.353174330434337e-06, "loss": 3.4523, "step": 1782 }, { "epoch": 0.21, "learning_rate": 9.352187233223549e-06, "loss": 3.3406, "step": 1783 }, { "epoch": 0.21, "learning_rate": 9.351199435569658e-06, "loss": 3.4247, "step": 1784 }, { "epoch": 0.21, "learning_rate": 9.350210937631644e-06, "loss": 3.4226, "step": 1785 }, { "epoch": 0.21, "learning_rate": 9.349221739568594e-06, "loss": 3.4586, "step": 1786 }, { "epoch": 0.21, "learning_rate": 9.34823184153971e-06, "loss": 3.3583, "step": 1787 }, { "epoch": 0.21, "learning_rate": 9.347241243704308e-06, "loss": 3.4491, "step": 1788 }, { "epoch": 0.21, "learning_rate": 9.346249946221814e-06, "loss": 3.3857, "step": 1789 }, { "epoch": 0.21, "learning_rate": 9.34525794925177e-06, "loss": 3.3986, "step": 1790 }, { "epoch": 0.21, "learning_rate": 9.344265252953825e-06, "loss": 3.4986, "step": 1791 }, { "epoch": 0.22, "learning_rate": 9.343271857487748e-06, "loss": 3.3593, "step": 1792 }, { "epoch": 0.22, "learning_rate": 9.342277763013414e-06, "loss": 3.4408, "step": 1793 }, { "epoch": 0.22, "learning_rate": 9.341282969690813e-06, "loss": 3.4931, "step": 1794 }, { "epoch": 0.22, "learning_rate": 9.340287477680048e-06, "loss": 3.4746, "step": 1795 }, { "epoch": 0.22, "learning_rate": 9.339291287141332e-06, "loss": 3.4506, "step": 1796 }, { "epoch": 0.22, "learning_rate": 9.338294398234997e-06, "loss": 3.5023, "step": 1797 }, { "epoch": 0.22, "learning_rate": 9.33729681112148e-06, "loss": 3.3206, "step": 1798 }, { "epoch": 0.22, "learning_rate": 9.336298525961334e-06, "loss": 3.467, "step": 1799 }, { "epoch": 0.22, "learning_rate": 9.335299542915221e-06, "loss": 3.3804, "step": 1800 }, { "epoch": 0.22, "learning_rate": 9.33429986214392e-06, "loss": 3.4011, "step": 1801 }, { "epoch": 0.22, "learning_rate": 9.333299483808321e-06, "loss": 3.3203, "step": 1802 }, { "epoch": 0.22, "learning_rate": 9.332298408069422e-06, "loss": 3.5329, "step": 1803 }, { "epoch": 0.22, "learning_rate": 9.33129663508834e-06, "loss": 3.411, "step": 1804 }, { "epoch": 0.22, "learning_rate": 9.3302941650263e-06, "loss": 3.4729, "step": 1805 }, { "epoch": 0.22, "learning_rate": 9.32929099804464e-06, "loss": 3.4499, "step": 1806 }, { "epoch": 0.22, "learning_rate": 9.328287134304807e-06, "loss": 3.3265, "step": 1807 }, { "epoch": 0.22, "learning_rate": 9.327282573968369e-06, "loss": 3.4295, "step": 1808 }, { "epoch": 0.22, "learning_rate": 9.326277317196996e-06, "loss": 3.41, "step": 1809 }, { "epoch": 0.22, "learning_rate": 9.325271364152477e-06, "loss": 3.3774, "step": 1810 }, { "epoch": 0.22, "learning_rate": 9.324264714996709e-06, "loss": 3.3781, "step": 1811 }, { "epoch": 0.22, "learning_rate": 9.323257369891702e-06, "loss": 3.3854, "step": 1812 }, { "epoch": 0.22, "learning_rate": 9.322249328999582e-06, "loss": 3.3898, "step": 1813 }, { "epoch": 0.22, "learning_rate": 9.321240592482578e-06, "loss": 3.3326, "step": 1814 }, { "epoch": 0.22, "learning_rate": 9.320231160503044e-06, "loss": 3.4146, "step": 1815 }, { "epoch": 0.22, "learning_rate": 9.31922103322343e-06, "loss": 3.4186, "step": 1816 }, { "epoch": 0.22, "learning_rate": 9.318210210806315e-06, "loss": 3.4738, "step": 1817 }, { "epoch": 0.22, "learning_rate": 9.317198693414375e-06, "loss": 3.4159, "step": 1818 }, { "epoch": 0.22, "learning_rate": 9.316186481210406e-06, "loss": 3.5403, "step": 1819 }, { "epoch": 0.22, "learning_rate": 9.315173574357313e-06, "loss": 3.4085, "step": 1820 }, { "epoch": 0.22, "learning_rate": 9.314159973018115e-06, "loss": 3.42, "step": 1821 }, { "epoch": 0.22, "learning_rate": 9.313145677355943e-06, "loss": 3.4921, "step": 1822 }, { "epoch": 0.22, "learning_rate": 9.312130687534035e-06, "loss": 3.3823, "step": 1823 }, { "epoch": 0.22, "learning_rate": 9.311115003715745e-06, "loss": 3.3936, "step": 1824 }, { "epoch": 0.22, "learning_rate": 9.310098626064539e-06, "loss": 3.4393, "step": 1825 }, { "epoch": 0.22, "learning_rate": 9.309081554743992e-06, "loss": 3.493, "step": 1826 }, { "epoch": 0.22, "learning_rate": 9.308063789917792e-06, "loss": 3.4805, "step": 1827 }, { "epoch": 0.22, "learning_rate": 9.307045331749738e-06, "loss": 3.331, "step": 1828 }, { "epoch": 0.22, "learning_rate": 9.306026180403742e-06, "loss": 3.4691, "step": 1829 }, { "epoch": 0.22, "learning_rate": 9.305006336043828e-06, "loss": 3.3333, "step": 1830 }, { "epoch": 0.22, "learning_rate": 9.303985798834128e-06, "loss": 3.4353, "step": 1831 }, { "epoch": 0.22, "learning_rate": 9.302964568938889e-06, "loss": 3.4915, "step": 1832 }, { "epoch": 0.22, "learning_rate": 9.301942646522467e-06, "loss": 3.5022, "step": 1833 }, { "epoch": 0.22, "learning_rate": 9.300920031749335e-06, "loss": 3.4377, "step": 1834 }, { "epoch": 0.22, "learning_rate": 9.299896724784066e-06, "loss": 3.483, "step": 1835 }, { "epoch": 0.22, "learning_rate": 9.298872725791355e-06, "loss": 3.3672, "step": 1836 }, { "epoch": 0.22, "learning_rate": 9.297848034936007e-06, "loss": 3.3758, "step": 1837 }, { "epoch": 0.22, "learning_rate": 9.296822652382934e-06, "loss": 3.4097, "step": 1838 }, { "epoch": 0.22, "learning_rate": 9.295796578297161e-06, "loss": 3.3821, "step": 1839 }, { "epoch": 0.22, "learning_rate": 9.294769812843826e-06, "loss": 3.2891, "step": 1840 }, { "epoch": 0.22, "learning_rate": 9.293742356188176e-06, "loss": 3.5116, "step": 1841 }, { "epoch": 0.22, "learning_rate": 9.292714208495573e-06, "loss": 3.4946, "step": 1842 }, { "epoch": 0.22, "learning_rate": 9.291685369931484e-06, "loss": 3.4274, "step": 1843 }, { "epoch": 0.22, "learning_rate": 9.290655840661494e-06, "loss": 3.5278, "step": 1844 }, { "epoch": 0.22, "learning_rate": 9.289625620851293e-06, "loss": 3.413, "step": 1845 }, { "epoch": 0.22, "learning_rate": 9.288594710666687e-06, "loss": 3.3515, "step": 1846 }, { "epoch": 0.22, "learning_rate": 9.287563110273593e-06, "loss": 3.4365, "step": 1847 }, { "epoch": 0.22, "learning_rate": 9.286530819838033e-06, "loss": 3.3214, "step": 1848 }, { "epoch": 0.22, "learning_rate": 9.285497839526146e-06, "loss": 3.385, "step": 1849 }, { "epoch": 0.22, "learning_rate": 9.284464169504182e-06, "loss": 3.3579, "step": 1850 }, { "epoch": 0.22, "learning_rate": 9.283429809938497e-06, "loss": 3.4797, "step": 1851 }, { "epoch": 0.22, "learning_rate": 9.282394760995565e-06, "loss": 3.4433, "step": 1852 }, { "epoch": 0.22, "learning_rate": 9.281359022841966e-06, "loss": 3.5008, "step": 1853 }, { "epoch": 0.22, "learning_rate": 9.28032259564439e-06, "loss": 3.4617, "step": 1854 }, { "epoch": 0.22, "learning_rate": 9.279285479569645e-06, "loss": 3.5067, "step": 1855 }, { "epoch": 0.22, "learning_rate": 9.278247674784639e-06, "loss": 3.3276, "step": 1856 }, { "epoch": 0.22, "learning_rate": 9.2772091814564e-06, "loss": 3.4689, "step": 1857 }, { "epoch": 0.22, "learning_rate": 9.276169999752066e-06, "loss": 3.4891, "step": 1858 }, { "epoch": 0.22, "learning_rate": 9.275130129838876e-06, "loss": 3.4542, "step": 1859 }, { "epoch": 0.22, "learning_rate": 9.274089571884194e-06, "loss": 3.4406, "step": 1860 }, { "epoch": 0.22, "learning_rate": 9.273048326055488e-06, "loss": 3.4962, "step": 1861 }, { "epoch": 0.22, "learning_rate": 9.272006392520331e-06, "loss": 3.4212, "step": 1862 }, { "epoch": 0.22, "learning_rate": 9.270963771446418e-06, "loss": 3.2893, "step": 1863 }, { "epoch": 0.22, "learning_rate": 9.269920463001545e-06, "loss": 3.4188, "step": 1864 }, { "epoch": 0.22, "learning_rate": 9.268876467353625e-06, "loss": 3.4997, "step": 1865 }, { "epoch": 0.22, "learning_rate": 9.267831784670678e-06, "loss": 3.4103, "step": 1866 }, { "epoch": 0.22, "learning_rate": 9.266786415120836e-06, "loss": 3.3676, "step": 1867 }, { "epoch": 0.22, "learning_rate": 9.265740358872342e-06, "loss": 3.4368, "step": 1868 }, { "epoch": 0.22, "learning_rate": 9.264693616093548e-06, "loss": 3.4617, "step": 1869 }, { "epoch": 0.22, "learning_rate": 9.263646186952916e-06, "loss": 3.4893, "step": 1870 }, { "epoch": 0.22, "learning_rate": 9.26259807161902e-06, "loss": 3.3344, "step": 1871 }, { "epoch": 0.22, "learning_rate": 9.261549270260548e-06, "loss": 3.4241, "step": 1872 }, { "epoch": 0.22, "learning_rate": 9.260499783046292e-06, "loss": 3.4152, "step": 1873 }, { "epoch": 0.22, "learning_rate": 9.259449610145156e-06, "loss": 3.4773, "step": 1874 }, { "epoch": 0.23, "learning_rate": 9.258398751726156e-06, "loss": 3.4604, "step": 1875 }, { "epoch": 0.23, "learning_rate": 9.257347207958419e-06, "loss": 3.4971, "step": 1876 }, { "epoch": 0.23, "learning_rate": 9.256294979011179e-06, "loss": 3.4673, "step": 1877 }, { "epoch": 0.23, "learning_rate": 9.255242065053783e-06, "loss": 3.4745, "step": 1878 }, { "epoch": 0.23, "learning_rate": 9.254188466255687e-06, "loss": 3.3967, "step": 1879 }, { "epoch": 0.23, "learning_rate": 9.253134182786458e-06, "loss": 3.4697, "step": 1880 }, { "epoch": 0.23, "learning_rate": 9.252079214815775e-06, "loss": 3.3648, "step": 1881 }, { "epoch": 0.23, "learning_rate": 9.251023562513421e-06, "loss": 3.2352, "step": 1882 }, { "epoch": 0.23, "learning_rate": 9.249967226049296e-06, "loss": 3.3641, "step": 1883 }, { "epoch": 0.23, "learning_rate": 9.248910205593407e-06, "loss": 3.4433, "step": 1884 }, { "epoch": 0.23, "learning_rate": 9.24785250131587e-06, "loss": 3.4496, "step": 1885 }, { "epoch": 0.23, "learning_rate": 9.246794113386916e-06, "loss": 3.4564, "step": 1886 }, { "epoch": 0.23, "learning_rate": 9.245735041976877e-06, "loss": 3.4659, "step": 1887 }, { "epoch": 0.23, "learning_rate": 9.244675287256206e-06, "loss": 3.4615, "step": 1888 }, { "epoch": 0.23, "learning_rate": 9.243614849395457e-06, "loss": 3.4029, "step": 1889 }, { "epoch": 0.23, "learning_rate": 9.242553728565298e-06, "loss": 3.4429, "step": 1890 }, { "epoch": 0.23, "learning_rate": 9.241491924936505e-06, "loss": 3.3349, "step": 1891 }, { "epoch": 0.23, "learning_rate": 9.24042943867997e-06, "loss": 3.4183, "step": 1892 }, { "epoch": 0.23, "learning_rate": 9.239366269966683e-06, "loss": 3.3218, "step": 1893 }, { "epoch": 0.23, "learning_rate": 9.238302418967757e-06, "loss": 3.4909, "step": 1894 }, { "epoch": 0.23, "learning_rate": 9.237237885854407e-06, "loss": 3.4601, "step": 1895 }, { "epoch": 0.23, "learning_rate": 9.236172670797956e-06, "loss": 3.4905, "step": 1896 }, { "epoch": 0.23, "learning_rate": 9.235106773969844e-06, "loss": 3.4612, "step": 1897 }, { "epoch": 0.23, "learning_rate": 9.234040195541618e-06, "loss": 3.4325, "step": 1898 }, { "epoch": 0.23, "learning_rate": 9.23297293568493e-06, "loss": 3.4651, "step": 1899 }, { "epoch": 0.23, "learning_rate": 9.231904994571546e-06, "loss": 3.4638, "step": 1900 }, { "epoch": 0.23, "learning_rate": 9.230836372373342e-06, "loss": 3.3637, "step": 1901 }, { "epoch": 0.23, "learning_rate": 9.229767069262305e-06, "loss": 3.3532, "step": 1902 }, { "epoch": 0.23, "learning_rate": 9.228697085410524e-06, "loss": 3.3437, "step": 1903 }, { "epoch": 0.23, "learning_rate": 9.227626420990205e-06, "loss": 3.4052, "step": 1904 }, { "epoch": 0.23, "learning_rate": 9.226555076173664e-06, "loss": 3.3294, "step": 1905 }, { "epoch": 0.23, "learning_rate": 9.22548305113332e-06, "loss": 3.5072, "step": 1906 }, { "epoch": 0.23, "learning_rate": 9.224410346041706e-06, "loss": 3.3592, "step": 1907 }, { "epoch": 0.23, "learning_rate": 9.223336961071463e-06, "loss": 3.4542, "step": 1908 }, { "epoch": 0.23, "learning_rate": 9.222262896395347e-06, "loss": 3.4369, "step": 1909 }, { "epoch": 0.23, "learning_rate": 9.221188152186211e-06, "loss": 3.4777, "step": 1910 }, { "epoch": 0.23, "learning_rate": 9.22011272861703e-06, "loss": 3.3812, "step": 1911 }, { "epoch": 0.23, "learning_rate": 9.219036625860882e-06, "loss": 3.4164, "step": 1912 }, { "epoch": 0.23, "learning_rate": 9.217959844090956e-06, "loss": 3.4123, "step": 1913 }, { "epoch": 0.23, "learning_rate": 9.216882383480547e-06, "loss": 3.4842, "step": 1914 }, { "epoch": 0.23, "learning_rate": 9.215804244203066e-06, "loss": 3.384, "step": 1915 }, { "epoch": 0.23, "learning_rate": 9.214725426432027e-06, "loss": 3.4188, "step": 1916 }, { "epoch": 0.23, "learning_rate": 9.213645930341056e-06, "loss": 3.4201, "step": 1917 }, { "epoch": 0.23, "learning_rate": 9.212565756103887e-06, "loss": 3.3887, "step": 1918 }, { "epoch": 0.23, "learning_rate": 9.211484903894365e-06, "loss": 3.4297, "step": 1919 }, { "epoch": 0.23, "learning_rate": 9.21040337388644e-06, "loss": 3.3563, "step": 1920 }, { "epoch": 0.23, "learning_rate": 9.209321166254178e-06, "loss": 3.4024, "step": 1921 }, { "epoch": 0.23, "learning_rate": 9.208238281171747e-06, "loss": 3.3613, "step": 1922 }, { "epoch": 0.23, "learning_rate": 9.207154718813428e-06, "loss": 3.4039, "step": 1923 }, { "epoch": 0.23, "learning_rate": 9.20607047935361e-06, "loss": 3.4402, "step": 1924 }, { "epoch": 0.23, "learning_rate": 9.204985562966792e-06, "loss": 3.3845, "step": 1925 }, { "epoch": 0.23, "learning_rate": 9.203899969827579e-06, "loss": 3.4604, "step": 1926 }, { "epoch": 0.23, "learning_rate": 9.202813700110688e-06, "loss": 3.3837, "step": 1927 }, { "epoch": 0.23, "learning_rate": 9.201726753990945e-06, "loss": 3.4668, "step": 1928 }, { "epoch": 0.23, "learning_rate": 9.200639131643282e-06, "loss": 3.3492, "step": 1929 }, { "epoch": 0.23, "learning_rate": 9.19955083324274e-06, "loss": 3.3848, "step": 1930 }, { "epoch": 0.23, "learning_rate": 9.198461858964475e-06, "loss": 3.4416, "step": 1931 }, { "epoch": 0.23, "learning_rate": 9.197372208983744e-06, "loss": 3.3783, "step": 1932 }, { "epoch": 0.23, "learning_rate": 9.196281883475913e-06, "loss": 3.3676, "step": 1933 }, { "epoch": 0.23, "learning_rate": 9.195190882616465e-06, "loss": 3.3777, "step": 1934 }, { "epoch": 0.23, "learning_rate": 9.194099206580981e-06, "loss": 3.5047, "step": 1935 }, { "epoch": 0.23, "learning_rate": 9.19300685554516e-06, "loss": 3.5148, "step": 1936 }, { "epoch": 0.23, "learning_rate": 9.191913829684806e-06, "loss": 3.5192, "step": 1937 }, { "epoch": 0.23, "learning_rate": 9.190820129175828e-06, "loss": 3.4595, "step": 1938 }, { "epoch": 0.23, "learning_rate": 9.189725754194248e-06, "loss": 3.4106, "step": 1939 }, { "epoch": 0.23, "learning_rate": 9.188630704916193e-06, "loss": 3.4465, "step": 1940 }, { "epoch": 0.23, "learning_rate": 9.187534981517904e-06, "loss": 3.4063, "step": 1941 }, { "epoch": 0.23, "learning_rate": 9.186438584175726e-06, "loss": 3.3871, "step": 1942 }, { "epoch": 0.23, "learning_rate": 9.185341513066114e-06, "loss": 3.4016, "step": 1943 }, { "epoch": 0.23, "learning_rate": 9.18424376836563e-06, "loss": 3.5322, "step": 1944 }, { "epoch": 0.23, "learning_rate": 9.183145350250948e-06, "loss": 3.4133, "step": 1945 }, { "epoch": 0.23, "learning_rate": 9.182046258898846e-06, "loss": 3.4149, "step": 1946 }, { "epoch": 0.23, "learning_rate": 9.180946494486211e-06, "loss": 3.3719, "step": 1947 }, { "epoch": 0.23, "learning_rate": 9.179846057190044e-06, "loss": 3.3719, "step": 1948 }, { "epoch": 0.23, "learning_rate": 9.178744947187445e-06, "loss": 3.3661, "step": 1949 }, { "epoch": 0.23, "learning_rate": 9.17764316465563e-06, "loss": 3.3653, "step": 1950 }, { "epoch": 0.23, "learning_rate": 9.176540709771921e-06, "loss": 3.5022, "step": 1951 }, { "epoch": 0.23, "learning_rate": 9.175437582713744e-06, "loss": 3.44, "step": 1952 }, { "epoch": 0.23, "learning_rate": 9.174333783658642e-06, "loss": 3.4273, "step": 1953 }, { "epoch": 0.23, "learning_rate": 9.173229312784256e-06, "loss": 3.4172, "step": 1954 }, { "epoch": 0.23, "learning_rate": 9.172124170268344e-06, "loss": 3.492, "step": 1955 }, { "epoch": 0.23, "learning_rate": 9.171018356288765e-06, "loss": 3.4168, "step": 1956 }, { "epoch": 0.23, "learning_rate": 9.169911871023492e-06, "loss": 3.4148, "step": 1957 }, { "epoch": 0.24, "learning_rate": 9.1688047146506e-06, "loss": 3.3602, "step": 1958 }, { "epoch": 0.24, "learning_rate": 9.16769688734828e-06, "loss": 3.3875, "step": 1959 }, { "epoch": 0.24, "learning_rate": 9.16658838929482e-06, "loss": 3.4044, "step": 1960 }, { "epoch": 0.24, "learning_rate": 9.165479220668627e-06, "loss": 3.4202, "step": 1961 }, { "epoch": 0.24, "learning_rate": 9.16436938164821e-06, "loss": 3.4222, "step": 1962 }, { "epoch": 0.24, "learning_rate": 9.163258872412186e-06, "loss": 3.3766, "step": 1963 }, { "epoch": 0.24, "learning_rate": 9.162147693139284e-06, "loss": 3.3475, "step": 1964 }, { "epoch": 0.24, "learning_rate": 9.161035844008333e-06, "loss": 3.3741, "step": 1965 }, { "epoch": 0.24, "learning_rate": 9.159923325198278e-06, "loss": 3.4086, "step": 1966 }, { "epoch": 0.24, "learning_rate": 9.158810136888166e-06, "loss": 3.3994, "step": 1967 }, { "epoch": 0.24, "learning_rate": 9.157696279257159e-06, "loss": 3.3789, "step": 1968 }, { "epoch": 0.24, "learning_rate": 9.156581752484513e-06, "loss": 3.4329, "step": 1969 }, { "epoch": 0.24, "learning_rate": 9.155466556749609e-06, "loss": 3.4004, "step": 1970 }, { "epoch": 0.24, "learning_rate": 9.154350692231923e-06, "loss": 3.3949, "step": 1971 }, { "epoch": 0.24, "learning_rate": 9.153234159111042e-06, "loss": 3.4554, "step": 1972 }, { "epoch": 0.24, "learning_rate": 9.152116957566665e-06, "loss": 3.4151, "step": 1973 }, { "epoch": 0.24, "learning_rate": 9.150999087778592e-06, "loss": 3.3734, "step": 1974 }, { "epoch": 0.24, "learning_rate": 9.149880549926734e-06, "loss": 3.4756, "step": 1975 }, { "epoch": 0.24, "learning_rate": 9.14876134419111e-06, "loss": 3.3627, "step": 1976 }, { "epoch": 0.24, "learning_rate": 9.147641470751843e-06, "loss": 3.4733, "step": 1977 }, { "epoch": 0.24, "learning_rate": 9.14652092978917e-06, "loss": 3.5128, "step": 1978 }, { "epoch": 0.24, "learning_rate": 9.145399721483428e-06, "loss": 3.4871, "step": 1979 }, { "epoch": 0.24, "learning_rate": 9.144277846015066e-06, "loss": 3.4863, "step": 1980 }, { "epoch": 0.24, "learning_rate": 9.14315530356464e-06, "loss": 3.443, "step": 1981 }, { "epoch": 0.24, "learning_rate": 9.142032094312811e-06, "loss": 3.3801, "step": 1982 }, { "epoch": 0.24, "learning_rate": 9.14090821844035e-06, "loss": 3.4871, "step": 1983 }, { "epoch": 0.24, "learning_rate": 9.139783676128133e-06, "loss": 3.4708, "step": 1984 }, { "epoch": 0.24, "learning_rate": 9.138658467557147e-06, "loss": 3.4528, "step": 1985 }, { "epoch": 0.24, "learning_rate": 9.13753259290848e-06, "loss": 3.4141, "step": 1986 }, { "epoch": 0.24, "learning_rate": 9.136406052363332e-06, "loss": 3.3914, "step": 1987 }, { "epoch": 0.24, "learning_rate": 9.13527884610301e-06, "loss": 3.3865, "step": 1988 }, { "epoch": 0.24, "learning_rate": 9.134150974308926e-06, "loss": 3.4613, "step": 1989 }, { "epoch": 0.24, "learning_rate": 9.133022437162602e-06, "loss": 3.4159, "step": 1990 }, { "epoch": 0.24, "learning_rate": 9.131893234845664e-06, "loss": 3.421, "step": 1991 }, { "epoch": 0.24, "learning_rate": 9.130763367539845e-06, "loss": 3.4111, "step": 1992 }, { "epoch": 0.24, "learning_rate": 9.129632835426989e-06, "loss": 3.3641, "step": 1993 }, { "epoch": 0.24, "learning_rate": 9.128501638689043e-06, "loss": 3.3189, "step": 1994 }, { "epoch": 0.24, "learning_rate": 9.127369777508063e-06, "loss": 3.3976, "step": 1995 }, { "epoch": 0.24, "learning_rate": 9.126237252066211e-06, "loss": 3.3936, "step": 1996 }, { "epoch": 0.24, "learning_rate": 9.125104062545754e-06, "loss": 3.4596, "step": 1997 }, { "epoch": 0.24, "learning_rate": 9.123970209129071e-06, "loss": 3.3927, "step": 1998 }, { "epoch": 0.24, "learning_rate": 9.122835691998646e-06, "loss": 3.3934, "step": 1999 }, { "epoch": 0.24, "learning_rate": 9.121700511337064e-06, "loss": 3.4095, "step": 2000 }, { "epoch": 0.24, "learning_rate": 9.120564667327027e-06, "loss": 3.4592, "step": 2001 }, { "epoch": 0.24, "learning_rate": 9.119428160151334e-06, "loss": 3.3815, "step": 2002 }, { "epoch": 0.24, "learning_rate": 9.118290989992894e-06, "loss": 3.4262, "step": 2003 }, { "epoch": 0.24, "learning_rate": 9.11715315703473e-06, "loss": 3.5336, "step": 2004 }, { "epoch": 0.24, "learning_rate": 9.11601466145996e-06, "loss": 3.3819, "step": 2005 }, { "epoch": 0.24, "learning_rate": 9.114875503451816e-06, "loss": 3.4943, "step": 2006 }, { "epoch": 0.24, "learning_rate": 9.113735683193636e-06, "loss": 3.4836, "step": 2007 }, { "epoch": 0.24, "learning_rate": 9.112595200868859e-06, "loss": 3.382, "step": 2008 }, { "epoch": 0.24, "learning_rate": 9.111454056661039e-06, "loss": 3.3408, "step": 2009 }, { "epoch": 0.24, "learning_rate": 9.11031225075383e-06, "loss": 3.4502, "step": 2010 }, { "epoch": 0.24, "learning_rate": 9.109169783330995e-06, "loss": 3.3171, "step": 2011 }, { "epoch": 0.24, "learning_rate": 9.108026654576404e-06, "loss": 3.3834, "step": 2012 }, { "epoch": 0.24, "learning_rate": 9.106882864674031e-06, "loss": 3.3834, "step": 2013 }, { "epoch": 0.24, "learning_rate": 9.10573841380796e-06, "loss": 3.4736, "step": 2014 }, { "epoch": 0.24, "learning_rate": 9.104593302162382e-06, "loss": 3.4456, "step": 2015 }, { "epoch": 0.24, "learning_rate": 9.103447529921584e-06, "loss": 3.4444, "step": 2016 }, { "epoch": 0.24, "learning_rate": 9.102301097269974e-06, "loss": 3.3826, "step": 2017 }, { "epoch": 0.24, "learning_rate": 9.101154004392058e-06, "loss": 3.3956, "step": 2018 }, { "epoch": 0.24, "learning_rate": 9.100006251472448e-06, "loss": 3.4504, "step": 2019 }, { "epoch": 0.24, "learning_rate": 9.098857838695866e-06, "loss": 3.4981, "step": 2020 }, { "epoch": 0.24, "learning_rate": 9.097708766247135e-06, "loss": 3.4717, "step": 2021 }, { "epoch": 0.24, "learning_rate": 9.096559034311192e-06, "loss": 3.4572, "step": 2022 }, { "epoch": 0.24, "learning_rate": 9.09540864307307e-06, "loss": 3.4504, "step": 2023 }, { "epoch": 0.24, "learning_rate": 9.094257592717916e-06, "loss": 3.402, "step": 2024 }, { "epoch": 0.24, "learning_rate": 9.093105883430983e-06, "loss": 3.4468, "step": 2025 }, { "epoch": 0.24, "learning_rate": 9.091953515397624e-06, "loss": 3.3814, "step": 2026 }, { "epoch": 0.24, "learning_rate": 9.090800488803303e-06, "loss": 3.4261, "step": 2027 }, { "epoch": 0.24, "learning_rate": 9.089646803833589e-06, "loss": 3.4992, "step": 2028 }, { "epoch": 0.24, "learning_rate": 9.088492460674157e-06, "loss": 3.3849, "step": 2029 }, { "epoch": 0.24, "learning_rate": 9.087337459510786e-06, "loss": 3.4777, "step": 2030 }, { "epoch": 0.24, "learning_rate": 9.086181800529364e-06, "loss": 3.3506, "step": 2031 }, { "epoch": 0.24, "learning_rate": 9.085025483915881e-06, "loss": 3.4561, "step": 2032 }, { "epoch": 0.24, "learning_rate": 9.083868509856438e-06, "loss": 3.4881, "step": 2033 }, { "epoch": 0.24, "learning_rate": 9.082710878537238e-06, "loss": 3.3723, "step": 2034 }, { "epoch": 0.24, "learning_rate": 9.081552590144588e-06, "loss": 3.4242, "step": 2035 }, { "epoch": 0.24, "learning_rate": 9.080393644864908e-06, "loss": 3.3611, "step": 2036 }, { "epoch": 0.24, "learning_rate": 9.079234042884718e-06, "loss": 3.3972, "step": 2037 }, { "epoch": 0.24, "learning_rate": 9.078073784390644e-06, "loss": 3.4073, "step": 2038 }, { "epoch": 0.24, "learning_rate": 9.076912869569418e-06, "loss": 3.3951, "step": 2039 }, { "epoch": 0.24, "learning_rate": 9.075751298607877e-06, "loss": 3.4792, "step": 2040 }, { "epoch": 0.24, "learning_rate": 9.07458907169297e-06, "loss": 3.4242, "step": 2041 }, { "epoch": 0.25, "learning_rate": 9.073426189011741e-06, "loss": 3.4347, "step": 2042 }, { "epoch": 0.25, "learning_rate": 9.072262650751346e-06, "loss": 3.4612, "step": 2043 }, { "epoch": 0.25, "learning_rate": 9.071098457099048e-06, "loss": 3.4082, "step": 2044 }, { "epoch": 0.25, "learning_rate": 9.06993360824221e-06, "loss": 3.458, "step": 2045 }, { "epoch": 0.25, "learning_rate": 9.068768104368306e-06, "loss": 3.4257, "step": 2046 }, { "epoch": 0.25, "learning_rate": 9.06760194566491e-06, "loss": 3.4335, "step": 2047 }, { "epoch": 0.25, "learning_rate": 9.066435132319703e-06, "loss": 3.3533, "step": 2048 }, { "epoch": 0.25, "learning_rate": 9.065267664520479e-06, "loss": 3.4792, "step": 2049 }, { "epoch": 0.25, "learning_rate": 9.064099542455125e-06, "loss": 3.4366, "step": 2050 }, { "epoch": 0.25, "learning_rate": 9.062930766311642e-06, "loss": 3.3524, "step": 2051 }, { "epoch": 0.25, "learning_rate": 9.06176133627813e-06, "loss": 3.5139, "step": 2052 }, { "epoch": 0.25, "learning_rate": 9.0605912525428e-06, "loss": 3.5036, "step": 2053 }, { "epoch": 0.25, "learning_rate": 9.059420515293969e-06, "loss": 3.358, "step": 2054 }, { "epoch": 0.25, "learning_rate": 9.058249124720048e-06, "loss": 3.4422, "step": 2055 }, { "epoch": 0.25, "learning_rate": 9.05707708100957e-06, "loss": 3.3914, "step": 2056 }, { "epoch": 0.25, "learning_rate": 9.055904384351158e-06, "loss": 3.4365, "step": 2057 }, { "epoch": 0.25, "learning_rate": 9.05473103493355e-06, "loss": 3.4227, "step": 2058 }, { "epoch": 0.25, "learning_rate": 9.053557032945582e-06, "loss": 3.4184, "step": 2059 }, { "epoch": 0.25, "learning_rate": 9.052382378576201e-06, "loss": 3.3461, "step": 2060 }, { "epoch": 0.25, "learning_rate": 9.051207072014456e-06, "loss": 3.4102, "step": 2061 }, { "epoch": 0.25, "learning_rate": 9.0500311134495e-06, "loss": 3.4611, "step": 2062 }, { "epoch": 0.25, "learning_rate": 9.048854503070596e-06, "loss": 3.4026, "step": 2063 }, { "epoch": 0.25, "learning_rate": 9.047677241067103e-06, "loss": 3.5082, "step": 2064 }, { "epoch": 0.25, "learning_rate": 9.046499327628495e-06, "loss": 3.479, "step": 2065 }, { "epoch": 0.25, "learning_rate": 9.045320762944344e-06, "loss": 3.3963, "step": 2066 }, { "epoch": 0.25, "learning_rate": 9.044141547204328e-06, "loss": 3.4444, "step": 2067 }, { "epoch": 0.25, "learning_rate": 9.04296168059823e-06, "loss": 3.3887, "step": 2068 }, { "epoch": 0.25, "learning_rate": 9.041781163315942e-06, "loss": 3.432, "step": 2069 }, { "epoch": 0.25, "learning_rate": 9.040599995547452e-06, "loss": 3.414, "step": 2070 }, { "epoch": 0.25, "learning_rate": 9.03941817748286e-06, "loss": 3.4317, "step": 2071 }, { "epoch": 0.25, "learning_rate": 9.038235709312371e-06, "loss": 3.4724, "step": 2072 }, { "epoch": 0.25, "learning_rate": 9.037052591226287e-06, "loss": 3.4183, "step": 2073 }, { "epoch": 0.25, "learning_rate": 9.035868823415024e-06, "loss": 3.4214, "step": 2074 }, { "epoch": 0.25, "learning_rate": 9.034684406069095e-06, "loss": 3.3732, "step": 2075 }, { "epoch": 0.25, "learning_rate": 9.033499339379122e-06, "loss": 3.3143, "step": 2076 }, { "epoch": 0.25, "learning_rate": 9.03231362353583e-06, "loss": 3.4604, "step": 2077 }, { "epoch": 0.25, "learning_rate": 9.03112725873005e-06, "loss": 3.4397, "step": 2078 }, { "epoch": 0.25, "learning_rate": 9.029940245152715e-06, "loss": 3.3995, "step": 2079 }, { "epoch": 0.25, "learning_rate": 9.028752582994864e-06, "loss": 3.4364, "step": 2080 }, { "epoch": 0.25, "learning_rate": 9.027564272447639e-06, "loss": 3.4496, "step": 2081 }, { "epoch": 0.25, "learning_rate": 9.026375313702286e-06, "loss": 3.4841, "step": 2082 }, { "epoch": 0.25, "learning_rate": 9.025185706950161e-06, "loss": 3.3582, "step": 2083 }, { "epoch": 0.25, "learning_rate": 9.023995452382714e-06, "loss": 3.471, "step": 2084 }, { "epoch": 0.25, "learning_rate": 9.02280455019151e-06, "loss": 3.472, "step": 2085 }, { "epoch": 0.25, "learning_rate": 9.021613000568212e-06, "loss": 3.2722, "step": 2086 }, { "epoch": 0.25, "learning_rate": 9.020420803704589e-06, "loss": 3.4187, "step": 2087 }, { "epoch": 0.25, "learning_rate": 9.019227959792511e-06, "loss": 3.4414, "step": 2088 }, { "epoch": 0.25, "learning_rate": 9.018034469023958e-06, "loss": 3.3297, "step": 2089 }, { "epoch": 0.25, "learning_rate": 9.016840331591011e-06, "loss": 3.3551, "step": 2090 }, { "epoch": 0.25, "learning_rate": 9.015645547685853e-06, "loss": 3.5488, "step": 2091 }, { "epoch": 0.25, "learning_rate": 9.014450117500773e-06, "loss": 3.4802, "step": 2092 }, { "epoch": 0.25, "learning_rate": 9.013254041228166e-06, "loss": 3.4336, "step": 2093 }, { "epoch": 0.25, "learning_rate": 9.012057319060527e-06, "loss": 3.4379, "step": 2094 }, { "epoch": 0.25, "learning_rate": 9.01085995119046e-06, "loss": 3.4774, "step": 2095 }, { "epoch": 0.25, "learning_rate": 9.009661937810667e-06, "loss": 3.4702, "step": 2096 }, { "epoch": 0.25, "learning_rate": 9.008463279113959e-06, "loss": 3.4429, "step": 2097 }, { "epoch": 0.25, "learning_rate": 9.007263975293246e-06, "loss": 3.4132, "step": 2098 }, { "epoch": 0.25, "learning_rate": 9.006064026541549e-06, "loss": 3.4909, "step": 2099 }, { "epoch": 0.25, "learning_rate": 9.004863433051982e-06, "loss": 3.4224, "step": 2100 }, { "epoch": 0.25, "learning_rate": 9.003662195017775e-06, "loss": 3.3267, "step": 2101 }, { "epoch": 0.25, "learning_rate": 9.002460312632252e-06, "loss": 3.4919, "step": 2102 }, { "epoch": 0.25, "learning_rate": 9.00125778608885e-06, "loss": 3.3936, "step": 2103 }, { "epoch": 0.25, "learning_rate": 9.000054615581096e-06, "loss": 3.4757, "step": 2104 }, { "epoch": 0.25, "learning_rate": 8.998850801302634e-06, "loss": 3.391, "step": 2105 }, { "epoch": 0.25, "learning_rate": 8.997646343447206e-06, "loss": 3.4275, "step": 2106 }, { "epoch": 0.25, "learning_rate": 8.996441242208658e-06, "loss": 3.3827, "step": 2107 }, { "epoch": 0.25, "learning_rate": 8.995235497780938e-06, "loss": 3.3389, "step": 2108 }, { "epoch": 0.25, "learning_rate": 8.994029110358102e-06, "loss": 3.5672, "step": 2109 }, { "epoch": 0.25, "learning_rate": 8.992822080134302e-06, "loss": 3.3607, "step": 2110 }, { "epoch": 0.25, "learning_rate": 8.991614407303804e-06, "loss": 3.3651, "step": 2111 }, { "epoch": 0.25, "learning_rate": 8.990406092060966e-06, "loss": 3.4249, "step": 2112 }, { "epoch": 0.25, "learning_rate": 8.989197134600258e-06, "loss": 3.5305, "step": 2113 }, { "epoch": 0.25, "learning_rate": 8.98798753511625e-06, "loss": 3.4992, "step": 2114 }, { "epoch": 0.25, "learning_rate": 8.986777293803614e-06, "loss": 3.342, "step": 2115 }, { "epoch": 0.25, "learning_rate": 8.985566410857127e-06, "loss": 3.4586, "step": 2116 }, { "epoch": 0.25, "learning_rate": 8.98435488647167e-06, "loss": 3.4102, "step": 2117 }, { "epoch": 0.25, "learning_rate": 8.983142720842226e-06, "loss": 3.3864, "step": 2118 }, { "epoch": 0.25, "learning_rate": 8.981929914163883e-06, "loss": 3.4343, "step": 2119 }, { "epoch": 0.25, "learning_rate": 8.980716466631827e-06, "loss": 3.5028, "step": 2120 }, { "epoch": 0.25, "learning_rate": 8.979502378441354e-06, "loss": 3.4115, "step": 2121 }, { "epoch": 0.25, "learning_rate": 8.97828764978786e-06, "loss": 3.4983, "step": 2122 }, { "epoch": 0.25, "learning_rate": 8.977072280866841e-06, "loss": 3.4247, "step": 2123 }, { "epoch": 0.25, "learning_rate": 8.975856271873903e-06, "loss": 3.3861, "step": 2124 }, { "epoch": 0.26, "learning_rate": 8.97463962300475e-06, "loss": 3.4259, "step": 2125 }, { "epoch": 0.26, "learning_rate": 8.973422334455186e-06, "loss": 3.3979, "step": 2126 }, { "epoch": 0.26, "learning_rate": 8.972204406421126e-06, "loss": 3.4092, "step": 2127 }, { "epoch": 0.26, "learning_rate": 8.970985839098585e-06, "loss": 3.4085, "step": 2128 }, { "epoch": 0.26, "learning_rate": 8.969766632683676e-06, "loss": 3.3606, "step": 2129 }, { "epoch": 0.26, "learning_rate": 8.96854678737262e-06, "loss": 3.3869, "step": 2130 }, { "epoch": 0.26, "learning_rate": 8.967326303361742e-06, "loss": 3.4227, "step": 2131 }, { "epoch": 0.26, "learning_rate": 8.966105180847464e-06, "loss": 3.372, "step": 2132 }, { "epoch": 0.26, "learning_rate": 8.964883420026316e-06, "loss": 3.4639, "step": 2133 }, { "epoch": 0.26, "learning_rate": 8.963661021094929e-06, "loss": 3.4646, "step": 2134 }, { "epoch": 0.26, "learning_rate": 8.962437984250034e-06, "loss": 3.3847, "step": 2135 }, { "epoch": 0.26, "learning_rate": 8.961214309688471e-06, "loss": 3.3641, "step": 2136 }, { "epoch": 0.26, "learning_rate": 8.959989997607174e-06, "loss": 3.4676, "step": 2137 }, { "epoch": 0.26, "learning_rate": 8.958765048203189e-06, "loss": 3.4095, "step": 2138 }, { "epoch": 0.26, "learning_rate": 8.957539461673656e-06, "loss": 3.4615, "step": 2139 }, { "epoch": 0.26, "learning_rate": 8.956313238215824e-06, "loss": 3.4297, "step": 2140 }, { "epoch": 0.26, "learning_rate": 8.955086378027042e-06, "loss": 3.4082, "step": 2141 }, { "epoch": 0.26, "learning_rate": 8.95385888130476e-06, "loss": 3.3638, "step": 2142 }, { "epoch": 0.26, "learning_rate": 8.952630748246534e-06, "loss": 3.4067, "step": 2143 }, { "epoch": 0.26, "learning_rate": 8.951401979050018e-06, "loss": 3.4341, "step": 2144 }, { "epoch": 0.26, "learning_rate": 8.950172573912972e-06, "loss": 3.5291, "step": 2145 }, { "epoch": 0.26, "learning_rate": 8.948942533033256e-06, "loss": 3.4629, "step": 2146 }, { "epoch": 0.26, "learning_rate": 8.947711856608837e-06, "loss": 3.3487, "step": 2147 }, { "epoch": 0.26, "learning_rate": 8.946480544837775e-06, "loss": 3.3223, "step": 2148 }, { "epoch": 0.26, "learning_rate": 8.945248597918242e-06, "loss": 3.4205, "step": 2149 }, { "epoch": 0.26, "learning_rate": 8.944016016048504e-06, "loss": 3.3417, "step": 2150 }, { "epoch": 0.26, "learning_rate": 8.94278279942694e-06, "loss": 3.3787, "step": 2151 }, { "epoch": 0.26, "learning_rate": 8.94154894825202e-06, "loss": 3.4848, "step": 2152 }, { "epoch": 0.26, "learning_rate": 8.94031446272232e-06, "loss": 3.3728, "step": 2153 }, { "epoch": 0.26, "learning_rate": 8.93907934303652e-06, "loss": 3.4048, "step": 2154 }, { "epoch": 0.26, "learning_rate": 8.937843589393401e-06, "loss": 3.4499, "step": 2155 }, { "epoch": 0.26, "learning_rate": 8.936607201991846e-06, "loss": 3.447, "step": 2156 }, { "epoch": 0.26, "learning_rate": 8.935370181030838e-06, "loss": 3.3786, "step": 2157 }, { "epoch": 0.26, "learning_rate": 8.934132526709467e-06, "loss": 3.4996, "step": 2158 }, { "epoch": 0.26, "learning_rate": 8.932894239226916e-06, "loss": 3.4723, "step": 2159 }, { "epoch": 0.26, "learning_rate": 8.931655318782484e-06, "loss": 3.4191, "step": 2160 }, { "epoch": 0.26, "learning_rate": 8.930415765575555e-06, "loss": 3.3667, "step": 2161 }, { "epoch": 0.26, "learning_rate": 8.929175579805629e-06, "loss": 3.3928, "step": 2162 }, { "epoch": 0.26, "learning_rate": 8.9279347616723e-06, "loss": 3.421, "step": 2163 }, { "epoch": 0.26, "learning_rate": 8.926693311375266e-06, "loss": 3.4423, "step": 2164 }, { "epoch": 0.26, "learning_rate": 8.925451229114325e-06, "loss": 3.3665, "step": 2165 }, { "epoch": 0.26, "learning_rate": 8.924208515089382e-06, "loss": 3.3725, "step": 2166 }, { "epoch": 0.26, "learning_rate": 8.922965169500437e-06, "loss": 3.4115, "step": 2167 }, { "epoch": 0.26, "learning_rate": 8.921721192547596e-06, "loss": 3.4402, "step": 2168 }, { "epoch": 0.26, "learning_rate": 8.920476584431063e-06, "loss": 3.4383, "step": 2169 }, { "epoch": 0.26, "learning_rate": 8.919231345351149e-06, "loss": 3.3164, "step": 2170 }, { "epoch": 0.26, "learning_rate": 8.917985475508261e-06, "loss": 3.3896, "step": 2171 }, { "epoch": 0.26, "learning_rate": 8.916738975102914e-06, "loss": 3.3536, "step": 2172 }, { "epoch": 0.26, "learning_rate": 8.915491844335715e-06, "loss": 3.4311, "step": 2173 }, { "epoch": 0.26, "learning_rate": 8.914244083407379e-06, "loss": 3.3894, "step": 2174 }, { "epoch": 0.26, "learning_rate": 8.912995692518722e-06, "loss": 3.3924, "step": 2175 }, { "epoch": 0.26, "learning_rate": 8.911746671870663e-06, "loss": 3.4317, "step": 2176 }, { "epoch": 0.26, "learning_rate": 8.910497021664217e-06, "loss": 3.4054, "step": 2177 }, { "epoch": 0.26, "learning_rate": 8.909246742100505e-06, "loss": 3.4501, "step": 2178 }, { "epoch": 0.26, "learning_rate": 8.907995833380746e-06, "loss": 3.4335, "step": 2179 }, { "epoch": 0.26, "learning_rate": 8.906744295706262e-06, "loss": 3.4053, "step": 2180 }, { "epoch": 0.26, "learning_rate": 8.905492129278478e-06, "loss": 3.4658, "step": 2181 }, { "epoch": 0.26, "learning_rate": 8.904239334298917e-06, "loss": 3.3563, "step": 2182 }, { "epoch": 0.26, "learning_rate": 8.902985910969205e-06, "loss": 3.3396, "step": 2183 }, { "epoch": 0.26, "learning_rate": 8.901731859491065e-06, "loss": 3.4884, "step": 2184 }, { "epoch": 0.26, "learning_rate": 8.900477180066329e-06, "loss": 3.4746, "step": 2185 }, { "epoch": 0.26, "learning_rate": 8.899221872896926e-06, "loss": 3.4733, "step": 2186 }, { "epoch": 0.26, "learning_rate": 8.897965938184883e-06, "loss": 3.4388, "step": 2187 }, { "epoch": 0.26, "learning_rate": 8.896709376132331e-06, "loss": 3.4463, "step": 2188 }, { "epoch": 0.26, "learning_rate": 8.895452186941505e-06, "loss": 3.438, "step": 2189 }, { "epoch": 0.26, "learning_rate": 8.894194370814733e-06, "loss": 3.3864, "step": 2190 }, { "epoch": 0.26, "learning_rate": 8.89293592795445e-06, "loss": 3.3873, "step": 2191 }, { "epoch": 0.26, "learning_rate": 8.891676858563193e-06, "loss": 3.3449, "step": 2192 }, { "epoch": 0.26, "learning_rate": 8.890417162843594e-06, "loss": 3.3769, "step": 2193 }, { "epoch": 0.26, "learning_rate": 8.889156840998389e-06, "loss": 3.5206, "step": 2194 }, { "epoch": 0.26, "learning_rate": 8.887895893230419e-06, "loss": 3.356, "step": 2195 }, { "epoch": 0.26, "learning_rate": 8.886634319742617e-06, "loss": 3.3366, "step": 2196 }, { "epoch": 0.26, "learning_rate": 8.885372120738023e-06, "loss": 3.4433, "step": 2197 }, { "epoch": 0.26, "learning_rate": 8.884109296419778e-06, "loss": 3.4963, "step": 2198 }, { "epoch": 0.26, "learning_rate": 8.882845846991116e-06, "loss": 3.5054, "step": 2199 }, { "epoch": 0.26, "learning_rate": 8.881581772655383e-06, "loss": 3.3602, "step": 2200 }, { "epoch": 0.26, "learning_rate": 8.880317073616016e-06, "loss": 3.3982, "step": 2201 }, { "epoch": 0.26, "learning_rate": 8.879051750076557e-06, "loss": 3.4652, "step": 2202 }, { "epoch": 0.26, "learning_rate": 8.877785802240651e-06, "loss": 3.5435, "step": 2203 }, { "epoch": 0.26, "learning_rate": 8.876519230312035e-06, "loss": 3.3851, "step": 2204 }, { "epoch": 0.26, "learning_rate": 8.875252034494554e-06, "loss": 3.3675, "step": 2205 }, { "epoch": 0.26, "learning_rate": 8.873984214992152e-06, "loss": 3.433, "step": 2206 }, { "epoch": 0.26, "learning_rate": 8.872715772008872e-06, "loss": 3.3978, "step": 2207 }, { "epoch": 0.27, "learning_rate": 8.871446705748858e-06, "loss": 3.4298, "step": 2208 }, { "epoch": 0.27, "learning_rate": 8.870177016416355e-06, "loss": 3.3904, "step": 2209 }, { "epoch": 0.27, "learning_rate": 8.868906704215705e-06, "loss": 3.4505, "step": 2210 }, { "epoch": 0.27, "learning_rate": 8.867635769351353e-06, "loss": 3.3799, "step": 2211 }, { "epoch": 0.27, "learning_rate": 8.866364212027846e-06, "loss": 3.4079, "step": 2212 }, { "epoch": 0.27, "learning_rate": 8.865092032449826e-06, "loss": 3.4245, "step": 2213 }, { "epoch": 0.27, "learning_rate": 8.863819230822043e-06, "loss": 3.5093, "step": 2214 }, { "epoch": 0.27, "learning_rate": 8.862545807349337e-06, "loss": 3.4954, "step": 2215 }, { "epoch": 0.27, "learning_rate": 8.861271762236657e-06, "loss": 3.4301, "step": 2216 }, { "epoch": 0.27, "learning_rate": 8.859997095689044e-06, "loss": 3.4428, "step": 2217 }, { "epoch": 0.27, "learning_rate": 8.85872180791165e-06, "loss": 3.4885, "step": 2218 }, { "epoch": 0.27, "learning_rate": 8.857445899109716e-06, "loss": 3.4258, "step": 2219 }, { "epoch": 0.27, "learning_rate": 8.856169369488587e-06, "loss": 3.3946, "step": 2220 }, { "epoch": 0.27, "learning_rate": 8.854892219253711e-06, "loss": 3.4562, "step": 2221 }, { "epoch": 0.27, "learning_rate": 8.85361444861063e-06, "loss": 3.4324, "step": 2222 }, { "epoch": 0.27, "learning_rate": 8.852336057764994e-06, "loss": 3.465, "step": 2223 }, { "epoch": 0.27, "learning_rate": 8.851057046922542e-06, "loss": 3.5284, "step": 2224 }, { "epoch": 0.27, "learning_rate": 8.849777416289123e-06, "loss": 3.3856, "step": 2225 }, { "epoch": 0.27, "learning_rate": 8.848497166070677e-06, "loss": 3.4491, "step": 2226 }, { "epoch": 0.27, "learning_rate": 8.847216296473254e-06, "loss": 3.4131, "step": 2227 }, { "epoch": 0.27, "learning_rate": 8.845934807702992e-06, "loss": 3.4696, "step": 2228 }, { "epoch": 0.27, "learning_rate": 8.84465269996614e-06, "loss": 3.435, "step": 2229 }, { "epoch": 0.27, "learning_rate": 8.843369973469034e-06, "loss": 3.3123, "step": 2230 }, { "epoch": 0.27, "learning_rate": 8.842086628418122e-06, "loss": 3.4333, "step": 2231 }, { "epoch": 0.27, "learning_rate": 8.840802665019945e-06, "loss": 3.5196, "step": 2232 }, { "epoch": 0.27, "learning_rate": 8.839518083481145e-06, "loss": 3.4831, "step": 2233 }, { "epoch": 0.27, "learning_rate": 8.838232884008462e-06, "loss": 3.5265, "step": 2234 }, { "epoch": 0.27, "learning_rate": 8.836947066808738e-06, "loss": 3.3871, "step": 2235 }, { "epoch": 0.27, "learning_rate": 8.83566063208891e-06, "loss": 3.33, "step": 2236 }, { "epoch": 0.27, "learning_rate": 8.834373580056022e-06, "loss": 3.3801, "step": 2237 }, { "epoch": 0.27, "learning_rate": 8.83308591091721e-06, "loss": 3.4085, "step": 2238 }, { "epoch": 0.27, "learning_rate": 8.831797624879713e-06, "loss": 3.3614, "step": 2239 }, { "epoch": 0.27, "learning_rate": 8.830508722150866e-06, "loss": 3.3913, "step": 2240 }, { "epoch": 0.27, "learning_rate": 8.829219202938109e-06, "loss": 3.5178, "step": 2241 }, { "epoch": 0.27, "learning_rate": 8.827929067448974e-06, "loss": 3.5537, "step": 2242 }, { "epoch": 0.27, "learning_rate": 8.826638315891102e-06, "loss": 3.5239, "step": 2243 }, { "epoch": 0.27, "learning_rate": 8.825346948472221e-06, "loss": 3.5591, "step": 2244 }, { "epoch": 0.27, "learning_rate": 8.824054965400169e-06, "loss": 3.476, "step": 2245 }, { "epoch": 0.27, "learning_rate": 8.822762366882874e-06, "loss": 3.3307, "step": 2246 }, { "epoch": 0.27, "learning_rate": 8.82146915312837e-06, "loss": 3.482, "step": 2247 }, { "epoch": 0.27, "learning_rate": 8.820175324344788e-06, "loss": 3.5713, "step": 2248 }, { "epoch": 0.27, "learning_rate": 8.818880880740356e-06, "loss": 3.4123, "step": 2249 }, { "epoch": 0.27, "learning_rate": 8.817585822523404e-06, "loss": 3.5313, "step": 2250 }, { "epoch": 0.27, "learning_rate": 8.816290149902356e-06, "loss": 3.4624, "step": 2251 }, { "epoch": 0.27, "learning_rate": 8.814993863085742e-06, "loss": 3.4808, "step": 2252 }, { "epoch": 0.27, "learning_rate": 8.813696962282185e-06, "loss": 3.5352, "step": 2253 }, { "epoch": 0.27, "learning_rate": 8.812399447700409e-06, "loss": 3.4485, "step": 2254 }, { "epoch": 0.27, "learning_rate": 8.811101319549236e-06, "loss": 3.3797, "step": 2255 }, { "epoch": 0.27, "learning_rate": 8.80980257803759e-06, "loss": 3.4355, "step": 2256 }, { "epoch": 0.27, "learning_rate": 8.808503223374488e-06, "loss": 3.3756, "step": 2257 }, { "epoch": 0.27, "learning_rate": 8.80720325576905e-06, "loss": 3.4224, "step": 2258 }, { "epoch": 0.27, "learning_rate": 8.805902675430495e-06, "loss": 3.4269, "step": 2259 }, { "epoch": 0.27, "learning_rate": 8.804601482568136e-06, "loss": 3.4507, "step": 2260 }, { "epoch": 0.27, "learning_rate": 8.80329967739139e-06, "loss": 3.3977, "step": 2261 }, { "epoch": 0.27, "learning_rate": 8.801997260109769e-06, "loss": 3.3724, "step": 2262 }, { "epoch": 0.27, "learning_rate": 8.800694230932885e-06, "loss": 3.4353, "step": 2263 }, { "epoch": 0.27, "learning_rate": 8.799390590070448e-06, "loss": 3.3642, "step": 2264 }, { "epoch": 0.27, "learning_rate": 8.798086337732266e-06, "loss": 3.4081, "step": 2265 }, { "epoch": 0.27, "learning_rate": 8.796781474128247e-06, "loss": 3.4675, "step": 2266 }, { "epoch": 0.27, "learning_rate": 8.795475999468395e-06, "loss": 3.4893, "step": 2267 }, { "epoch": 0.27, "learning_rate": 8.794169913962816e-06, "loss": 3.392, "step": 2268 }, { "epoch": 0.27, "learning_rate": 8.79286321782171e-06, "loss": 3.4648, "step": 2269 }, { "epoch": 0.27, "learning_rate": 8.791555911255378e-06, "loss": 3.4176, "step": 2270 }, { "epoch": 0.27, "learning_rate": 8.790247994474219e-06, "loss": 3.4188, "step": 2271 }, { "epoch": 0.27, "learning_rate": 8.788939467688728e-06, "loss": 3.4747, "step": 2272 }, { "epoch": 0.27, "learning_rate": 8.787630331109503e-06, "loss": 3.4558, "step": 2273 }, { "epoch": 0.27, "learning_rate": 8.786320584947234e-06, "loss": 3.4949, "step": 2274 }, { "epoch": 0.27, "learning_rate": 8.785010229412716e-06, "loss": 3.3329, "step": 2275 }, { "epoch": 0.27, "learning_rate": 8.783699264716835e-06, "loss": 3.3757, "step": 2276 }, { "epoch": 0.27, "learning_rate": 8.782387691070577e-06, "loss": 3.3924, "step": 2277 }, { "epoch": 0.27, "learning_rate": 8.781075508685031e-06, "loss": 3.3813, "step": 2278 }, { "epoch": 0.27, "learning_rate": 8.779762717771379e-06, "loss": 3.4683, "step": 2279 }, { "epoch": 0.27, "learning_rate": 8.7784493185409e-06, "loss": 3.5146, "step": 2280 }, { "epoch": 0.27, "learning_rate": 8.777135311204977e-06, "loss": 3.4056, "step": 2281 }, { "epoch": 0.27, "learning_rate": 8.775820695975083e-06, "loss": 3.4218, "step": 2282 }, { "epoch": 0.27, "learning_rate": 8.774505473062796e-06, "loss": 3.4487, "step": 2283 }, { "epoch": 0.27, "learning_rate": 8.773189642679785e-06, "loss": 3.3818, "step": 2284 }, { "epoch": 0.27, "learning_rate": 8.771873205037824e-06, "loss": 3.4372, "step": 2285 }, { "epoch": 0.27, "learning_rate": 8.770556160348782e-06, "loss": 3.4405, "step": 2286 }, { "epoch": 0.27, "learning_rate": 8.76923850882462e-06, "loss": 3.5075, "step": 2287 }, { "epoch": 0.27, "learning_rate": 8.767920250677404e-06, "loss": 3.3806, "step": 2288 }, { "epoch": 0.27, "learning_rate": 8.766601386119296e-06, "loss": 3.4415, "step": 2289 }, { "epoch": 0.27, "learning_rate": 8.765281915362553e-06, "loss": 3.326, "step": 2290 }, { "epoch": 0.27, "learning_rate": 8.763961838619534e-06, "loss": 3.3286, "step": 2291 }, { "epoch": 0.28, "learning_rate": 8.762641156102689e-06, "loss": 3.3961, "step": 2292 }, { "epoch": 0.28, "learning_rate": 8.761319868024572e-06, "loss": 3.4422, "step": 2293 }, { "epoch": 0.28, "learning_rate": 8.759997974597831e-06, "loss": 3.4277, "step": 2294 }, { "epoch": 0.28, "learning_rate": 8.758675476035213e-06, "loss": 3.47, "step": 2295 }, { "epoch": 0.28, "learning_rate": 8.75735237254956e-06, "loss": 3.5038, "step": 2296 }, { "epoch": 0.28, "learning_rate": 8.756028664353811e-06, "loss": 3.3849, "step": 2297 }, { "epoch": 0.28, "learning_rate": 8.754704351661009e-06, "loss": 3.3964, "step": 2298 }, { "epoch": 0.28, "learning_rate": 8.753379434684287e-06, "loss": 3.454, "step": 2299 }, { "epoch": 0.28, "learning_rate": 8.752053913636878e-06, "loss": 3.4448, "step": 2300 }, { "epoch": 0.28, "learning_rate": 8.750727788732111e-06, "loss": 3.4109, "step": 2301 }, { "epoch": 0.28, "learning_rate": 8.749401060183416e-06, "loss": 3.3311, "step": 2302 }, { "epoch": 0.28, "learning_rate": 8.748073728204314e-06, "loss": 3.3717, "step": 2303 }, { "epoch": 0.28, "learning_rate": 8.74674579300843e-06, "loss": 3.4067, "step": 2304 }, { "epoch": 0.28, "learning_rate": 8.745417254809478e-06, "loss": 3.3737, "step": 2305 }, { "epoch": 0.28, "learning_rate": 8.74408811382128e-06, "loss": 3.4805, "step": 2306 }, { "epoch": 0.28, "learning_rate": 8.74275837025774e-06, "loss": 3.4644, "step": 2307 }, { "epoch": 0.28, "learning_rate": 8.741428024332874e-06, "loss": 3.4898, "step": 2308 }, { "epoch": 0.28, "learning_rate": 8.740097076260787e-06, "loss": 3.4539, "step": 2309 }, { "epoch": 0.28, "learning_rate": 8.738765526255683e-06, "loss": 3.3922, "step": 2310 }, { "epoch": 0.28, "learning_rate": 8.737433374531861e-06, "loss": 3.5045, "step": 2311 }, { "epoch": 0.28, "learning_rate": 8.736100621303718e-06, "loss": 3.3656, "step": 2312 }, { "epoch": 0.28, "learning_rate": 8.734767266785749e-06, "loss": 3.4477, "step": 2313 }, { "epoch": 0.28, "learning_rate": 8.733433311192544e-06, "loss": 3.4369, "step": 2314 }, { "epoch": 0.28, "learning_rate": 8.73209875473879e-06, "loss": 3.4511, "step": 2315 }, { "epoch": 0.28, "learning_rate": 8.730763597639273e-06, "loss": 3.406, "step": 2316 }, { "epoch": 0.28, "learning_rate": 8.729427840108873e-06, "loss": 3.5139, "step": 2317 }, { "epoch": 0.28, "learning_rate": 8.728091482362565e-06, "loss": 3.4163, "step": 2318 }, { "epoch": 0.28, "learning_rate": 8.726754524615428e-06, "loss": 3.4125, "step": 2319 }, { "epoch": 0.28, "learning_rate": 8.725416967082629e-06, "loss": 3.4015, "step": 2320 }, { "epoch": 0.28, "learning_rate": 8.724078809979435e-06, "loss": 3.3716, "step": 2321 }, { "epoch": 0.28, "learning_rate": 8.722740053521213e-06, "loss": 3.4165, "step": 2322 }, { "epoch": 0.28, "learning_rate": 8.721400697923418e-06, "loss": 3.3797, "step": 2323 }, { "epoch": 0.28, "learning_rate": 8.72006074340161e-06, "loss": 3.3945, "step": 2324 }, { "epoch": 0.28, "learning_rate": 8.718720190171441e-06, "loss": 3.359, "step": 2325 }, { "epoch": 0.28, "learning_rate": 8.717379038448662e-06, "loss": 3.4283, "step": 2326 }, { "epoch": 0.28, "learning_rate": 8.716037288449117e-06, "loss": 3.3694, "step": 2327 }, { "epoch": 0.28, "learning_rate": 8.714694940388746e-06, "loss": 3.3412, "step": 2328 }, { "epoch": 0.28, "learning_rate": 8.71335199448359e-06, "loss": 3.4283, "step": 2329 }, { "epoch": 0.28, "learning_rate": 8.712008450949783e-06, "loss": 3.3912, "step": 2330 }, { "epoch": 0.28, "learning_rate": 8.710664310003553e-06, "loss": 3.4439, "step": 2331 }, { "epoch": 0.28, "learning_rate": 8.709319571861231e-06, "loss": 3.5196, "step": 2332 }, { "epoch": 0.28, "learning_rate": 8.707974236739238e-06, "loss": 3.4238, "step": 2333 }, { "epoch": 0.28, "learning_rate": 8.70662830485409e-06, "loss": 3.4248, "step": 2334 }, { "epoch": 0.28, "learning_rate": 8.705281776422408e-06, "loss": 3.5106, "step": 2335 }, { "epoch": 0.28, "learning_rate": 8.703934651660894e-06, "loss": 3.4012, "step": 2336 }, { "epoch": 0.28, "learning_rate": 8.702586930786364e-06, "loss": 3.3701, "step": 2337 }, { "epoch": 0.28, "learning_rate": 8.701238614015716e-06, "loss": 3.3204, "step": 2338 }, { "epoch": 0.28, "learning_rate": 8.69988970156595e-06, "loss": 3.4269, "step": 2339 }, { "epoch": 0.28, "learning_rate": 8.698540193654159e-06, "loss": 3.4144, "step": 2340 }, { "epoch": 0.28, "learning_rate": 8.697190090497537e-06, "loss": 3.4085, "step": 2341 }, { "epoch": 0.28, "learning_rate": 8.695839392313366e-06, "loss": 3.4714, "step": 2342 }, { "epoch": 0.28, "learning_rate": 8.694488099319031e-06, "loss": 3.3819, "step": 2343 }, { "epoch": 0.28, "learning_rate": 8.69313621173201e-06, "loss": 3.4576, "step": 2344 }, { "epoch": 0.28, "learning_rate": 8.691783729769874e-06, "loss": 3.4246, "step": 2345 }, { "epoch": 0.28, "learning_rate": 8.690430653650292e-06, "loss": 3.4661, "step": 2346 }, { "epoch": 0.28, "learning_rate": 8.689076983591032e-06, "loss": 3.5316, "step": 2347 }, { "epoch": 0.28, "learning_rate": 8.687722719809952e-06, "loss": 3.4836, "step": 2348 }, { "epoch": 0.28, "learning_rate": 8.686367862525009e-06, "loss": 3.3732, "step": 2349 }, { "epoch": 0.28, "learning_rate": 8.68501241195425e-06, "loss": 3.4852, "step": 2350 }, { "epoch": 0.28, "learning_rate": 8.683656368315829e-06, "loss": 3.3806, "step": 2351 }, { "epoch": 0.28, "learning_rate": 8.682299731827985e-06, "loss": 3.3105, "step": 2352 }, { "epoch": 0.28, "learning_rate": 8.680942502709053e-06, "loss": 3.4552, "step": 2353 }, { "epoch": 0.28, "learning_rate": 8.679584681177471e-06, "loss": 3.4478, "step": 2354 }, { "epoch": 0.28, "learning_rate": 8.678226267451763e-06, "loss": 3.4511, "step": 2355 }, { "epoch": 0.28, "learning_rate": 8.676867261750556e-06, "loss": 3.4525, "step": 2356 }, { "epoch": 0.28, "learning_rate": 8.675507664292568e-06, "loss": 3.3404, "step": 2357 }, { "epoch": 0.28, "learning_rate": 8.674147475296613e-06, "loss": 3.3886, "step": 2358 }, { "epoch": 0.28, "learning_rate": 8.672786694981602e-06, "loss": 3.5202, "step": 2359 }, { "epoch": 0.28, "learning_rate": 8.671425323566536e-06, "loss": 3.4532, "step": 2360 }, { "epoch": 0.28, "learning_rate": 8.670063361270518e-06, "loss": 3.4225, "step": 2361 }, { "epoch": 0.28, "learning_rate": 8.66870080831274e-06, "loss": 3.4018, "step": 2362 }, { "epoch": 0.28, "learning_rate": 8.667337664912497e-06, "loss": 3.4632, "step": 2363 }, { "epoch": 0.28, "learning_rate": 8.665973931289169e-06, "loss": 3.3825, "step": 2364 }, { "epoch": 0.28, "learning_rate": 8.664609607662237e-06, "loss": 3.4991, "step": 2365 }, { "epoch": 0.28, "learning_rate": 8.663244694251277e-06, "loss": 3.4213, "step": 2366 }, { "epoch": 0.28, "learning_rate": 8.66187919127596e-06, "loss": 3.3994, "step": 2367 }, { "epoch": 0.28, "learning_rate": 8.660513098956049e-06, "loss": 3.395, "step": 2368 }, { "epoch": 0.28, "learning_rate": 8.659146417511402e-06, "loss": 3.4728, "step": 2369 }, { "epoch": 0.28, "learning_rate": 8.657779147161976e-06, "loss": 3.3947, "step": 2370 }, { "epoch": 0.28, "learning_rate": 8.656411288127819e-06, "loss": 3.3217, "step": 2371 }, { "epoch": 0.28, "learning_rate": 8.655042840629074e-06, "loss": 3.4234, "step": 2372 }, { "epoch": 0.28, "learning_rate": 8.653673804885981e-06, "loss": 3.4305, "step": 2373 }, { "epoch": 0.28, "learning_rate": 8.652304181118874e-06, "loss": 3.4307, "step": 2374 }, { "epoch": 0.29, "learning_rate": 8.65093396954818e-06, "loss": 3.4101, "step": 2375 }, { "epoch": 0.29, "learning_rate": 8.649563170394421e-06, "loss": 3.3601, "step": 2376 }, { "epoch": 0.29, "learning_rate": 8.648191783878213e-06, "loss": 3.495, "step": 2377 }, { "epoch": 0.29, "learning_rate": 8.64681981022027e-06, "loss": 3.444, "step": 2378 }, { "epoch": 0.29, "learning_rate": 8.6454472496414e-06, "loss": 3.4506, "step": 2379 }, { "epoch": 0.29, "learning_rate": 8.644074102362496e-06, "loss": 3.4925, "step": 2380 }, { "epoch": 0.29, "learning_rate": 8.64270036860456e-06, "loss": 3.2675, "step": 2381 }, { "epoch": 0.29, "learning_rate": 8.641326048588679e-06, "loss": 3.474, "step": 2382 }, { "epoch": 0.29, "learning_rate": 8.639951142536036e-06, "loss": 3.4031, "step": 2383 }, { "epoch": 0.29, "learning_rate": 8.63857565066791e-06, "loss": 3.3657, "step": 2384 }, { "epoch": 0.29, "learning_rate": 8.637199573205674e-06, "loss": 3.443, "step": 2385 }, { "epoch": 0.29, "learning_rate": 8.635822910370793e-06, "loss": 3.479, "step": 2386 }, { "epoch": 0.29, "learning_rate": 8.63444566238483e-06, "loss": 3.4332, "step": 2387 }, { "epoch": 0.29, "learning_rate": 8.633067829469435e-06, "loss": 3.4917, "step": 2388 }, { "epoch": 0.29, "learning_rate": 8.631689411846363e-06, "loss": 3.4143, "step": 2389 }, { "epoch": 0.29, "learning_rate": 8.630310409737455e-06, "loss": 3.3752, "step": 2390 }, { "epoch": 0.29, "learning_rate": 8.628930823364644e-06, "loss": 3.4726, "step": 2391 }, { "epoch": 0.29, "learning_rate": 8.627550652949969e-06, "loss": 3.4856, "step": 2392 }, { "epoch": 0.29, "learning_rate": 8.626169898715553e-06, "loss": 3.3608, "step": 2393 }, { "epoch": 0.29, "learning_rate": 8.624788560883611e-06, "loss": 3.4833, "step": 2394 }, { "epoch": 0.29, "learning_rate": 8.623406639676459e-06, "loss": 3.4686, "step": 2395 }, { "epoch": 0.29, "learning_rate": 8.622024135316504e-06, "loss": 3.4144, "step": 2396 }, { "epoch": 0.29, "learning_rate": 8.620641048026249e-06, "loss": 3.3644, "step": 2397 }, { "epoch": 0.29, "learning_rate": 8.619257378028285e-06, "loss": 3.3878, "step": 2398 }, { "epoch": 0.29, "learning_rate": 8.617873125545303e-06, "loss": 3.4097, "step": 2399 }, { "epoch": 0.29, "learning_rate": 8.616488290800084e-06, "loss": 3.3448, "step": 2400 }, { "epoch": 0.29, "learning_rate": 8.615102874015504e-06, "loss": 3.406, "step": 2401 }, { "epoch": 0.29, "learning_rate": 8.613716875414533e-06, "loss": 3.4444, "step": 2402 }, { "epoch": 0.29, "learning_rate": 8.612330295220234e-06, "loss": 3.3941, "step": 2403 }, { "epoch": 0.29, "learning_rate": 8.610943133655767e-06, "loss": 3.4353, "step": 2404 }, { "epoch": 0.29, "learning_rate": 8.609555390944377e-06, "loss": 3.3877, "step": 2405 }, { "epoch": 0.29, "learning_rate": 8.60816706730941e-06, "loss": 3.4138, "step": 2406 }, { "epoch": 0.29, "learning_rate": 8.606778162974306e-06, "loss": 3.4416, "step": 2407 }, { "epoch": 0.29, "learning_rate": 8.605388678162594e-06, "loss": 3.3819, "step": 2408 }, { "epoch": 0.29, "learning_rate": 8.603998613097898e-06, "loss": 3.3049, "step": 2409 }, { "epoch": 0.29, "learning_rate": 8.602607968003935e-06, "loss": 3.4016, "step": 2410 }, { "epoch": 0.29, "learning_rate": 8.601216743104519e-06, "loss": 3.3432, "step": 2411 }, { "epoch": 0.29, "learning_rate": 8.59982493862355e-06, "loss": 3.3855, "step": 2412 }, { "epoch": 0.29, "learning_rate": 8.598432554785032e-06, "loss": 3.4419, "step": 2413 }, { "epoch": 0.29, "learning_rate": 8.59703959181305e-06, "loss": 3.3852, "step": 2414 }, { "epoch": 0.29, "learning_rate": 8.595646049931791e-06, "loss": 3.4146, "step": 2415 }, { "epoch": 0.29, "learning_rate": 8.594251929365532e-06, "loss": 3.412, "step": 2416 }, { "epoch": 0.29, "learning_rate": 8.592857230338642e-06, "loss": 3.4839, "step": 2417 }, { "epoch": 0.29, "learning_rate": 8.591461953075586e-06, "loss": 3.4378, "step": 2418 }, { "epoch": 0.29, "learning_rate": 8.590066097800919e-06, "loss": 3.4571, "step": 2419 }, { "epoch": 0.29, "learning_rate": 8.588669664739295e-06, "loss": 3.378, "step": 2420 }, { "epoch": 0.29, "learning_rate": 8.58727265411545e-06, "loss": 3.4678, "step": 2421 }, { "epoch": 0.29, "learning_rate": 8.585875066154225e-06, "loss": 3.5279, "step": 2422 }, { "epoch": 0.29, "learning_rate": 8.584476901080547e-06, "loss": 3.3962, "step": 2423 }, { "epoch": 0.29, "learning_rate": 8.583078159119436e-06, "loss": 3.3966, "step": 2424 }, { "epoch": 0.29, "learning_rate": 8.581678840496005e-06, "loss": 3.4094, "step": 2425 }, { "epoch": 0.29, "learning_rate": 8.580278945435466e-06, "loss": 3.4234, "step": 2426 }, { "epoch": 0.29, "learning_rate": 8.578878474163115e-06, "loss": 3.4308, "step": 2427 }, { "epoch": 0.29, "learning_rate": 8.577477426904345e-06, "loss": 3.4336, "step": 2428 }, { "epoch": 0.29, "learning_rate": 8.576075803884642e-06, "loss": 3.475, "step": 2429 }, { "epoch": 0.29, "learning_rate": 8.574673605329582e-06, "loss": 3.4138, "step": 2430 }, { "epoch": 0.29, "learning_rate": 8.57327083146484e-06, "loss": 3.4024, "step": 2431 }, { "epoch": 0.29, "learning_rate": 8.571867482516174e-06, "loss": 3.5069, "step": 2432 }, { "epoch": 0.29, "learning_rate": 8.570463558709441e-06, "loss": 3.4862, "step": 2433 }, { "epoch": 0.29, "learning_rate": 8.56905906027059e-06, "loss": 3.4617, "step": 2434 }, { "epoch": 0.29, "learning_rate": 8.567653987425664e-06, "loss": 3.3778, "step": 2435 }, { "epoch": 0.29, "learning_rate": 8.56624834040079e-06, "loss": 3.4961, "step": 2436 }, { "epoch": 0.29, "learning_rate": 8.5648421194222e-06, "loss": 3.3635, "step": 2437 }, { "epoch": 0.29, "learning_rate": 8.563435324716209e-06, "loss": 3.4539, "step": 2438 }, { "epoch": 0.29, "learning_rate": 8.562027956509226e-06, "loss": 3.3687, "step": 2439 }, { "epoch": 0.29, "learning_rate": 8.560620015027754e-06, "loss": 3.5341, "step": 2440 }, { "epoch": 0.29, "learning_rate": 8.55921150049839e-06, "loss": 3.4312, "step": 2441 }, { "epoch": 0.29, "learning_rate": 8.55780241314782e-06, "loss": 3.4406, "step": 2442 }, { "epoch": 0.29, "learning_rate": 8.55639275320282e-06, "loss": 3.4383, "step": 2443 }, { "epoch": 0.29, "learning_rate": 8.554982520890268e-06, "loss": 3.3724, "step": 2444 }, { "epoch": 0.29, "learning_rate": 8.55357171643712e-06, "loss": 3.3913, "step": 2445 }, { "epoch": 0.29, "learning_rate": 8.552160340070436e-06, "loss": 3.3476, "step": 2446 }, { "epoch": 0.29, "learning_rate": 8.550748392017364e-06, "loss": 3.435, "step": 2447 }, { "epoch": 0.29, "learning_rate": 8.549335872505141e-06, "loss": 3.4071, "step": 2448 }, { "epoch": 0.29, "learning_rate": 8.5479227817611e-06, "loss": 3.3536, "step": 2449 }, { "epoch": 0.29, "learning_rate": 8.546509120012664e-06, "loss": 3.3379, "step": 2450 }, { "epoch": 0.29, "learning_rate": 8.54509488748735e-06, "loss": 3.4962, "step": 2451 }, { "epoch": 0.29, "learning_rate": 8.543680084412764e-06, "loss": 3.4235, "step": 2452 }, { "epoch": 0.29, "learning_rate": 8.542264711016604e-06, "loss": 3.4655, "step": 2453 }, { "epoch": 0.29, "learning_rate": 8.540848767526661e-06, "loss": 3.4765, "step": 2454 }, { "epoch": 0.29, "learning_rate": 8.53943225417082e-06, "loss": 3.4778, "step": 2455 }, { "epoch": 0.29, "learning_rate": 8.538015171177054e-06, "loss": 3.4595, "step": 2456 }, { "epoch": 0.29, "learning_rate": 8.53659751877343e-06, "loss": 3.4762, "step": 2457 }, { "epoch": 0.3, "learning_rate": 8.535179297188103e-06, "loss": 3.4479, "step": 2458 }, { "epoch": 0.3, "learning_rate": 8.533760506649325e-06, "loss": 3.3781, "step": 2459 }, { "epoch": 0.3, "learning_rate": 8.532341147385434e-06, "loss": 3.4134, "step": 2460 }, { "epoch": 0.3, "learning_rate": 8.530921219624866e-06, "loss": 3.4001, "step": 2461 }, { "epoch": 0.3, "learning_rate": 8.52950072359614e-06, "loss": 3.469, "step": 2462 }, { "epoch": 0.3, "learning_rate": 8.528079659527877e-06, "loss": 3.4624, "step": 2463 }, { "epoch": 0.3, "learning_rate": 8.52665802764878e-06, "loss": 3.4272, "step": 2464 }, { "epoch": 0.3, "learning_rate": 8.525235828187648e-06, "loss": 3.4137, "step": 2465 }, { "epoch": 0.3, "learning_rate": 8.52381306137337e-06, "loss": 3.5555, "step": 2466 }, { "epoch": 0.3, "learning_rate": 8.522389727434929e-06, "loss": 3.5104, "step": 2467 }, { "epoch": 0.3, "learning_rate": 8.520965826601394e-06, "loss": 3.3534, "step": 2468 }, { "epoch": 0.3, "learning_rate": 8.519541359101931e-06, "loss": 3.387, "step": 2469 }, { "epoch": 0.3, "learning_rate": 8.51811632516579e-06, "loss": 3.3711, "step": 2470 }, { "epoch": 0.3, "learning_rate": 8.516690725022324e-06, "loss": 3.4254, "step": 2471 }, { "epoch": 0.3, "learning_rate": 8.515264558900963e-06, "loss": 3.4696, "step": 2472 }, { "epoch": 0.3, "learning_rate": 8.513837827031236e-06, "loss": 3.4634, "step": 2473 }, { "epoch": 0.3, "learning_rate": 8.512410529642765e-06, "loss": 3.5334, "step": 2474 }, { "epoch": 0.3, "learning_rate": 8.510982666965257e-06, "loss": 3.4177, "step": 2475 }, { "epoch": 0.3, "learning_rate": 8.509554239228516e-06, "loss": 3.3648, "step": 2476 }, { "epoch": 0.3, "learning_rate": 8.508125246662432e-06, "loss": 3.4773, "step": 2477 }, { "epoch": 0.3, "learning_rate": 8.506695689496987e-06, "loss": 3.4971, "step": 2478 }, { "epoch": 0.3, "learning_rate": 8.505265567962255e-06, "loss": 3.3952, "step": 2479 }, { "epoch": 0.3, "learning_rate": 8.503834882288401e-06, "loss": 3.4865, "step": 2480 }, { "epoch": 0.3, "learning_rate": 8.50240363270568e-06, "loss": 3.4445, "step": 2481 }, { "epoch": 0.3, "learning_rate": 8.50097181944444e-06, "loss": 3.4499, "step": 2482 }, { "epoch": 0.3, "learning_rate": 8.499539442735111e-06, "loss": 3.3792, "step": 2483 }, { "epoch": 0.3, "learning_rate": 8.49810650280823e-06, "loss": 3.5099, "step": 2484 }, { "epoch": 0.3, "learning_rate": 8.496672999894408e-06, "loss": 3.4411, "step": 2485 }, { "epoch": 0.3, "learning_rate": 8.495238934224356e-06, "loss": 3.3865, "step": 2486 }, { "epoch": 0.3, "learning_rate": 8.493804306028873e-06, "loss": 3.4396, "step": 2487 }, { "epoch": 0.3, "learning_rate": 8.492369115538849e-06, "loss": 3.4006, "step": 2488 }, { "epoch": 0.3, "learning_rate": 8.490933362985264e-06, "loss": 3.4204, "step": 2489 }, { "epoch": 0.3, "learning_rate": 8.489497048599189e-06, "loss": 3.367, "step": 2490 }, { "epoch": 0.3, "learning_rate": 8.488060172611784e-06, "loss": 3.4788, "step": 2491 }, { "epoch": 0.3, "learning_rate": 8.486622735254302e-06, "loss": 3.4179, "step": 2492 }, { "epoch": 0.3, "learning_rate": 8.485184736758083e-06, "loss": 3.4909, "step": 2493 }, { "epoch": 0.3, "learning_rate": 8.483746177354563e-06, "loss": 3.356, "step": 2494 }, { "epoch": 0.3, "learning_rate": 8.48230705727526e-06, "loss": 3.5045, "step": 2495 }, { "epoch": 0.3, "learning_rate": 8.48086737675179e-06, "loss": 3.3813, "step": 2496 }, { "epoch": 0.3, "learning_rate": 8.479427136015852e-06, "loss": 3.4805, "step": 2497 }, { "epoch": 0.3, "learning_rate": 8.477986335299242e-06, "loss": 3.4437, "step": 2498 }, { "epoch": 0.3, "learning_rate": 8.476544974833844e-06, "loss": 3.5211, "step": 2499 }, { "epoch": 0.3, "learning_rate": 8.475103054851629e-06, "loss": 3.4525, "step": 2500 }, { "epoch": 0.3, "learning_rate": 8.473660575584662e-06, "loss": 3.4172, "step": 2501 }, { "epoch": 0.3, "learning_rate": 8.472217537265092e-06, "loss": 3.4803, "step": 2502 }, { "epoch": 0.3, "learning_rate": 8.47077394012517e-06, "loss": 3.4123, "step": 2503 }, { "epoch": 0.3, "learning_rate": 8.469329784397222e-06, "loss": 3.3553, "step": 2504 }, { "epoch": 0.3, "learning_rate": 8.467885070313677e-06, "loss": 3.478, "step": 2505 }, { "epoch": 0.3, "learning_rate": 8.466439798107043e-06, "loss": 3.4694, "step": 2506 }, { "epoch": 0.3, "learning_rate": 8.464993968009925e-06, "loss": 3.4227, "step": 2507 }, { "epoch": 0.3, "learning_rate": 8.463547580255014e-06, "loss": 3.4373, "step": 2508 }, { "epoch": 0.3, "learning_rate": 8.462100635075097e-06, "loss": 3.3813, "step": 2509 }, { "epoch": 0.3, "learning_rate": 8.46065313270304e-06, "loss": 3.4947, "step": 2510 }, { "epoch": 0.3, "learning_rate": 8.459205073371807e-06, "loss": 3.3805, "step": 2511 }, { "epoch": 0.3, "learning_rate": 8.457756457314454e-06, "loss": 3.4644, "step": 2512 }, { "epoch": 0.3, "learning_rate": 8.456307284764113e-06, "loss": 3.3629, "step": 2513 }, { "epoch": 0.3, "learning_rate": 8.454857555954024e-06, "loss": 3.4534, "step": 2514 }, { "epoch": 0.3, "learning_rate": 8.453407271117498e-06, "loss": 3.529, "step": 2515 }, { "epoch": 0.3, "learning_rate": 8.451956430487952e-06, "loss": 3.4336, "step": 2516 }, { "epoch": 0.3, "learning_rate": 8.450505034298882e-06, "loss": 3.2696, "step": 2517 }, { "epoch": 0.3, "learning_rate": 8.449053082783874e-06, "loss": 3.3923, "step": 2518 }, { "epoch": 0.3, "learning_rate": 8.44760057617661e-06, "loss": 3.381, "step": 2519 }, { "epoch": 0.3, "learning_rate": 8.446147514710853e-06, "loss": 3.3686, "step": 2520 }, { "epoch": 0.3, "learning_rate": 8.444693898620466e-06, "loss": 3.3836, "step": 2521 }, { "epoch": 0.3, "learning_rate": 8.443239728139387e-06, "loss": 3.5005, "step": 2522 }, { "epoch": 0.3, "learning_rate": 8.441785003501655e-06, "loss": 3.4119, "step": 2523 }, { "epoch": 0.3, "learning_rate": 8.440329724941392e-06, "loss": 3.4135, "step": 2524 }, { "epoch": 0.3, "learning_rate": 8.438873892692813e-06, "loss": 3.2846, "step": 2525 }, { "epoch": 0.3, "learning_rate": 8.437417506990222e-06, "loss": 3.4936, "step": 2526 }, { "epoch": 0.3, "learning_rate": 8.435960568068007e-06, "loss": 3.4385, "step": 2527 }, { "epoch": 0.3, "learning_rate": 8.434503076160649e-06, "loss": 3.3775, "step": 2528 }, { "epoch": 0.3, "learning_rate": 8.433045031502717e-06, "loss": 3.401, "step": 2529 }, { "epoch": 0.3, "learning_rate": 8.431586434328874e-06, "loss": 3.4233, "step": 2530 }, { "epoch": 0.3, "learning_rate": 8.430127284873861e-06, "loss": 3.3582, "step": 2531 }, { "epoch": 0.3, "learning_rate": 8.42866758337252e-06, "loss": 3.4774, "step": 2532 }, { "epoch": 0.3, "learning_rate": 8.42720733005977e-06, "loss": 3.4272, "step": 2533 }, { "epoch": 0.3, "learning_rate": 8.425746525170631e-06, "loss": 3.4864, "step": 2534 }, { "epoch": 0.3, "learning_rate": 8.4242851689402e-06, "loss": 3.5599, "step": 2535 }, { "epoch": 0.3, "learning_rate": 8.422823261603672e-06, "loss": 3.4135, "step": 2536 }, { "epoch": 0.3, "learning_rate": 8.421360803396326e-06, "loss": 3.4173, "step": 2537 }, { "epoch": 0.3, "learning_rate": 8.419897794553533e-06, "loss": 3.3217, "step": 2538 }, { "epoch": 0.3, "learning_rate": 8.418434235310745e-06, "loss": 3.4466, "step": 2539 }, { "epoch": 0.3, "learning_rate": 8.416970125903515e-06, "loss": 3.4464, "step": 2540 }, { "epoch": 0.31, "learning_rate": 8.41550546656747e-06, "loss": 3.303, "step": 2541 }, { "epoch": 0.31, "learning_rate": 8.414040257538337e-06, "loss": 3.5058, "step": 2542 }, { "epoch": 0.31, "learning_rate": 8.412574499051928e-06, "loss": 3.3116, "step": 2543 }, { "epoch": 0.31, "learning_rate": 8.411108191344141e-06, "loss": 3.3683, "step": 2544 }, { "epoch": 0.31, "learning_rate": 8.409641334650966e-06, "loss": 3.3462, "step": 2545 }, { "epoch": 0.31, "learning_rate": 8.408173929208479e-06, "loss": 3.3871, "step": 2546 }, { "epoch": 0.31, "learning_rate": 8.406705975252843e-06, "loss": 3.3772, "step": 2547 }, { "epoch": 0.31, "learning_rate": 8.405237473020316e-06, "loss": 3.4475, "step": 2548 }, { "epoch": 0.31, "learning_rate": 8.403768422747234e-06, "loss": 3.4837, "step": 2549 }, { "epoch": 0.31, "learning_rate": 8.40229882467003e-06, "loss": 3.4238, "step": 2550 }, { "epoch": 0.31, "learning_rate": 8.400828679025218e-06, "loss": 3.3361, "step": 2551 }, { "epoch": 0.31, "learning_rate": 8.39935798604941e-06, "loss": 3.3543, "step": 2552 }, { "epoch": 0.31, "learning_rate": 8.397886745979297e-06, "loss": 3.3994, "step": 2553 }, { "epoch": 0.31, "learning_rate": 8.39641495905166e-06, "loss": 3.4285, "step": 2554 }, { "epoch": 0.31, "learning_rate": 8.39494262550337e-06, "loss": 3.4113, "step": 2555 }, { "epoch": 0.31, "learning_rate": 8.393469745571386e-06, "loss": 3.4715, "step": 2556 }, { "epoch": 0.31, "learning_rate": 8.391996319492753e-06, "loss": 3.4275, "step": 2557 }, { "epoch": 0.31, "learning_rate": 8.390522347504603e-06, "loss": 3.3488, "step": 2558 }, { "epoch": 0.31, "learning_rate": 8.389047829844161e-06, "loss": 3.5291, "step": 2559 }, { "epoch": 0.31, "learning_rate": 8.387572766748734e-06, "loss": 3.3548, "step": 2560 }, { "epoch": 0.31, "learning_rate": 8.386097158455722e-06, "loss": 3.4448, "step": 2561 }, { "epoch": 0.31, "learning_rate": 8.384621005202609e-06, "loss": 3.326, "step": 2562 }, { "epoch": 0.31, "learning_rate": 8.383144307226968e-06, "loss": 3.5032, "step": 2563 }, { "epoch": 0.31, "learning_rate": 8.381667064766456e-06, "loss": 3.4703, "step": 2564 }, { "epoch": 0.31, "learning_rate": 8.380189278058827e-06, "loss": 3.3447, "step": 2565 }, { "epoch": 0.31, "learning_rate": 8.378710947341915e-06, "loss": 3.4018, "step": 2566 }, { "epoch": 0.31, "learning_rate": 8.37723207285364e-06, "loss": 3.4489, "step": 2567 }, { "epoch": 0.31, "learning_rate": 8.375752654832016e-06, "loss": 3.4101, "step": 2568 }, { "epoch": 0.31, "learning_rate": 8.37427269351514e-06, "loss": 3.401, "step": 2569 }, { "epoch": 0.31, "learning_rate": 8.372792189141197e-06, "loss": 3.4924, "step": 2570 }, { "epoch": 0.31, "learning_rate": 8.371311141948462e-06, "loss": 3.3753, "step": 2571 }, { "epoch": 0.31, "learning_rate": 8.369829552175293e-06, "loss": 3.5105, "step": 2572 }, { "epoch": 0.31, "learning_rate": 8.368347420060138e-06, "loss": 3.3881, "step": 2573 }, { "epoch": 0.31, "learning_rate": 8.366864745841537e-06, "loss": 3.409, "step": 2574 }, { "epoch": 0.31, "learning_rate": 8.365381529758108e-06, "loss": 3.4301, "step": 2575 }, { "epoch": 0.31, "learning_rate": 8.363897772048557e-06, "loss": 3.4665, "step": 2576 }, { "epoch": 0.31, "learning_rate": 8.362413472951688e-06, "loss": 3.3943, "step": 2577 }, { "epoch": 0.31, "learning_rate": 8.360928632706381e-06, "loss": 3.3535, "step": 2578 }, { "epoch": 0.31, "learning_rate": 8.359443251551607e-06, "loss": 3.3406, "step": 2579 }, { "epoch": 0.31, "learning_rate": 8.357957329726426e-06, "loss": 3.3473, "step": 2580 }, { "epoch": 0.31, "learning_rate": 8.356470867469979e-06, "loss": 3.4088, "step": 2581 }, { "epoch": 0.31, "learning_rate": 8.3549838650215e-06, "loss": 3.315, "step": 2582 }, { "epoch": 0.31, "learning_rate": 8.35349632262031e-06, "loss": 3.4319, "step": 2583 }, { "epoch": 0.31, "learning_rate": 8.352008240505813e-06, "loss": 3.3673, "step": 2584 }, { "epoch": 0.31, "learning_rate": 8.3505196189175e-06, "loss": 3.4249, "step": 2585 }, { "epoch": 0.31, "learning_rate": 8.349030458094953e-06, "loss": 3.4002, "step": 2586 }, { "epoch": 0.31, "learning_rate": 8.347540758277837e-06, "loss": 3.3921, "step": 2587 }, { "epoch": 0.31, "learning_rate": 8.346050519705906e-06, "loss": 3.4089, "step": 2588 }, { "epoch": 0.31, "learning_rate": 8.344559742619e-06, "loss": 3.3758, "step": 2589 }, { "epoch": 0.31, "learning_rate": 8.343068427257041e-06, "loss": 3.4654, "step": 2590 }, { "epoch": 0.31, "learning_rate": 8.341576573860049e-06, "loss": 3.391, "step": 2591 }, { "epoch": 0.31, "learning_rate": 8.340084182668118e-06, "loss": 3.4939, "step": 2592 }, { "epoch": 0.31, "learning_rate": 8.338591253921437e-06, "loss": 3.4874, "step": 2593 }, { "epoch": 0.31, "learning_rate": 8.337097787860278e-06, "loss": 3.4239, "step": 2594 }, { "epoch": 0.31, "learning_rate": 8.335603784724999e-06, "loss": 3.4323, "step": 2595 }, { "epoch": 0.31, "learning_rate": 8.334109244756047e-06, "loss": 3.3518, "step": 2596 }, { "epoch": 0.31, "learning_rate": 8.332614168193954e-06, "loss": 3.4536, "step": 2597 }, { "epoch": 0.31, "learning_rate": 8.331118555279338e-06, "loss": 3.38, "step": 2598 }, { "epoch": 0.31, "learning_rate": 8.329622406252903e-06, "loss": 3.3912, "step": 2599 }, { "epoch": 0.31, "learning_rate": 8.32812572135544e-06, "loss": 3.5027, "step": 2600 }, { "epoch": 0.31, "learning_rate": 8.326628500827826e-06, "loss": 3.4095, "step": 2601 }, { "epoch": 0.31, "learning_rate": 8.325130744911028e-06, "loss": 3.2968, "step": 2602 }, { "epoch": 0.31, "learning_rate": 8.323632453846091e-06, "loss": 3.4838, "step": 2603 }, { "epoch": 0.31, "learning_rate": 8.322133627874153e-06, "loss": 3.3847, "step": 2604 }, { "epoch": 0.31, "learning_rate": 8.320634267236434e-06, "loss": 3.3837, "step": 2605 }, { "epoch": 0.31, "learning_rate": 8.319134372174242e-06, "loss": 3.3462, "step": 2606 }, { "epoch": 0.31, "learning_rate": 8.317633942928976e-06, "loss": 3.4018, "step": 2607 }, { "epoch": 0.31, "learning_rate": 8.316132979742108e-06, "loss": 3.384, "step": 2608 }, { "epoch": 0.31, "learning_rate": 8.314631482855207e-06, "loss": 3.3654, "step": 2609 }, { "epoch": 0.31, "learning_rate": 8.313129452509926e-06, "loss": 3.4708, "step": 2610 }, { "epoch": 0.31, "learning_rate": 8.311626888948001e-06, "loss": 3.4959, "step": 2611 }, { "epoch": 0.31, "learning_rate": 8.310123792411255e-06, "loss": 3.4641, "step": 2612 }, { "epoch": 0.31, "learning_rate": 8.3086201631416e-06, "loss": 3.4381, "step": 2613 }, { "epoch": 0.31, "learning_rate": 8.307116001381025e-06, "loss": 3.3224, "step": 2614 }, { "epoch": 0.31, "learning_rate": 8.305611307371616e-06, "loss": 3.4844, "step": 2615 }, { "epoch": 0.31, "learning_rate": 8.304106081355541e-06, "loss": 3.367, "step": 2616 }, { "epoch": 0.31, "learning_rate": 8.302600323575043e-06, "loss": 3.4158, "step": 2617 }, { "epoch": 0.31, "learning_rate": 8.301094034272467e-06, "loss": 3.395, "step": 2618 }, { "epoch": 0.31, "learning_rate": 8.299587213690233e-06, "loss": 3.4114, "step": 2619 }, { "epoch": 0.31, "learning_rate": 8.298079862070849e-06, "loss": 3.4823, "step": 2620 }, { "epoch": 0.31, "learning_rate": 8.29657197965691e-06, "loss": 3.3645, "step": 2621 }, { "epoch": 0.31, "learning_rate": 8.295063566691095e-06, "loss": 3.4375, "step": 2622 }, { "epoch": 0.31, "learning_rate": 8.293554623416168e-06, "loss": 3.3587, "step": 2623 }, { "epoch": 0.31, "learning_rate": 8.292045150074981e-06, "loss": 3.3425, "step": 2624 }, { "epoch": 0.32, "learning_rate": 8.290535146910467e-06, "loss": 3.3929, "step": 2625 }, { "epoch": 0.32, "learning_rate": 8.289024614165649e-06, "loss": 3.4706, "step": 2626 }, { "epoch": 0.32, "learning_rate": 8.28751355208363e-06, "loss": 3.3512, "step": 2627 }, { "epoch": 0.32, "learning_rate": 8.286001960907605e-06, "loss": 3.5243, "step": 2628 }, { "epoch": 0.32, "learning_rate": 8.284489840880847e-06, "loss": 3.3745, "step": 2629 }, { "epoch": 0.32, "learning_rate": 8.282977192246717e-06, "loss": 3.5137, "step": 2630 }, { "epoch": 0.32, "learning_rate": 8.281464015248663e-06, "loss": 3.4701, "step": 2631 }, { "epoch": 0.32, "learning_rate": 8.279950310130218e-06, "loss": 3.4062, "step": 2632 }, { "epoch": 0.32, "learning_rate": 8.278436077134995e-06, "loss": 3.3752, "step": 2633 }, { "epoch": 0.32, "learning_rate": 8.276921316506696e-06, "loss": 3.423, "step": 2634 }, { "epoch": 0.32, "learning_rate": 8.27540602848911e-06, "loss": 3.3233, "step": 2635 }, { "epoch": 0.32, "learning_rate": 8.273890213326106e-06, "loss": 3.4271, "step": 2636 }, { "epoch": 0.32, "learning_rate": 8.272373871261639e-06, "loss": 3.3595, "step": 2637 }, { "epoch": 0.32, "learning_rate": 8.270857002539752e-06, "loss": 3.4199, "step": 2638 }, { "epoch": 0.32, "learning_rate": 8.26933960740457e-06, "loss": 3.4674, "step": 2639 }, { "epoch": 0.32, "learning_rate": 8.267821686100304e-06, "loss": 3.3034, "step": 2640 }, { "epoch": 0.32, "learning_rate": 8.266303238871245e-06, "loss": 3.2818, "step": 2641 }, { "epoch": 0.32, "learning_rate": 8.264784265961778e-06, "loss": 3.4334, "step": 2642 }, { "epoch": 0.32, "learning_rate": 8.263264767616365e-06, "loss": 3.5662, "step": 2643 }, { "epoch": 0.32, "learning_rate": 8.261744744079555e-06, "loss": 3.3385, "step": 2644 }, { "epoch": 0.32, "learning_rate": 8.260224195595979e-06, "loss": 3.3831, "step": 2645 }, { "epoch": 0.32, "learning_rate": 8.258703122410358e-06, "loss": 3.4146, "step": 2646 }, { "epoch": 0.32, "learning_rate": 8.257181524767493e-06, "loss": 3.3739, "step": 2647 }, { "epoch": 0.32, "learning_rate": 8.255659402912271e-06, "loss": 3.396, "step": 2648 }, { "epoch": 0.32, "learning_rate": 8.254136757089662e-06, "loss": 3.4764, "step": 2649 }, { "epoch": 0.32, "learning_rate": 8.252613587544722e-06, "loss": 3.409, "step": 2650 }, { "epoch": 0.32, "learning_rate": 8.251089894522593e-06, "loss": 3.3702, "step": 2651 }, { "epoch": 0.32, "learning_rate": 8.249565678268494e-06, "loss": 3.391, "step": 2652 }, { "epoch": 0.32, "learning_rate": 8.248040939027738e-06, "loss": 3.4327, "step": 2653 }, { "epoch": 0.32, "learning_rate": 8.246515677045712e-06, "loss": 3.4737, "step": 2654 }, { "epoch": 0.32, "learning_rate": 8.244989892567898e-06, "loss": 3.4345, "step": 2655 }, { "epoch": 0.32, "learning_rate": 8.243463585839852e-06, "loss": 3.4003, "step": 2656 }, { "epoch": 0.32, "learning_rate": 8.241936757107223e-06, "loss": 3.4457, "step": 2657 }, { "epoch": 0.32, "learning_rate": 8.240409406615735e-06, "loss": 3.4233, "step": 2658 }, { "epoch": 0.32, "learning_rate": 8.238881534611202e-06, "loss": 3.3544, "step": 2659 }, { "epoch": 0.32, "learning_rate": 8.23735314133952e-06, "loss": 3.4666, "step": 2660 }, { "epoch": 0.32, "learning_rate": 8.235824227046674e-06, "loss": 3.493, "step": 2661 }, { "epoch": 0.32, "learning_rate": 8.234294791978722e-06, "loss": 3.3783, "step": 2662 }, { "epoch": 0.32, "learning_rate": 8.232764836381815e-06, "loss": 3.5049, "step": 2663 }, { "epoch": 0.32, "learning_rate": 8.231234360502184e-06, "loss": 3.299, "step": 2664 }, { "epoch": 0.32, "learning_rate": 8.229703364586145e-06, "loss": 3.3551, "step": 2665 }, { "epoch": 0.32, "learning_rate": 8.228171848880098e-06, "loss": 3.3798, "step": 2666 }, { "epoch": 0.32, "learning_rate": 8.226639813630523e-06, "loss": 3.5113, "step": 2667 }, { "epoch": 0.32, "learning_rate": 8.225107259083988e-06, "loss": 3.4463, "step": 2668 }, { "epoch": 0.32, "learning_rate": 8.223574185487145e-06, "loss": 3.4006, "step": 2669 }, { "epoch": 0.32, "learning_rate": 8.222040593086726e-06, "loss": 3.4896, "step": 2670 }, { "epoch": 0.32, "learning_rate": 8.220506482129548e-06, "loss": 3.3638, "step": 2671 }, { "epoch": 0.32, "learning_rate": 8.21897185286251e-06, "loss": 3.406, "step": 2672 }, { "epoch": 0.32, "learning_rate": 8.2174367055326e-06, "loss": 3.4346, "step": 2673 }, { "epoch": 0.32, "learning_rate": 8.215901040386881e-06, "loss": 3.4281, "step": 2674 }, { "epoch": 0.32, "learning_rate": 8.214364857672508e-06, "loss": 3.4421, "step": 2675 }, { "epoch": 0.32, "learning_rate": 8.212828157636711e-06, "loss": 3.4147, "step": 2676 }, { "epoch": 0.32, "learning_rate": 8.211290940526809e-06, "loss": 3.4787, "step": 2677 }, { "epoch": 0.32, "learning_rate": 8.209753206590203e-06, "loss": 3.3865, "step": 2678 }, { "epoch": 0.32, "learning_rate": 8.208214956074375e-06, "loss": 3.3646, "step": 2679 }, { "epoch": 0.32, "learning_rate": 8.206676189226894e-06, "loss": 3.4405, "step": 2680 }, { "epoch": 0.32, "learning_rate": 8.205136906295408e-06, "loss": 3.3967, "step": 2681 }, { "epoch": 0.32, "learning_rate": 8.20359710752765e-06, "loss": 3.3986, "step": 2682 }, { "epoch": 0.32, "learning_rate": 8.202056793171439e-06, "loss": 3.4392, "step": 2683 }, { "epoch": 0.32, "learning_rate": 8.20051596347467e-06, "loss": 3.4165, "step": 2684 }, { "epoch": 0.32, "learning_rate": 8.198974618685324e-06, "loss": 3.4011, "step": 2685 }, { "epoch": 0.32, "learning_rate": 8.197432759051472e-06, "loss": 3.403, "step": 2686 }, { "epoch": 0.32, "learning_rate": 8.195890384821254e-06, "loss": 3.3641, "step": 2687 }, { "epoch": 0.32, "learning_rate": 8.194347496242908e-06, "loss": 3.4131, "step": 2688 }, { "epoch": 0.32, "learning_rate": 8.192804093564741e-06, "loss": 3.4692, "step": 2689 }, { "epoch": 0.32, "learning_rate": 8.191260177035152e-06, "loss": 3.412, "step": 2690 }, { "epoch": 0.32, "learning_rate": 8.189715746902618e-06, "loss": 3.4674, "step": 2691 }, { "epoch": 0.32, "learning_rate": 8.188170803415702e-06, "loss": 3.3843, "step": 2692 }, { "epoch": 0.32, "learning_rate": 8.186625346823048e-06, "loss": 3.4805, "step": 2693 }, { "epoch": 0.32, "learning_rate": 8.185079377373381e-06, "loss": 3.4384, "step": 2694 }, { "epoch": 0.32, "learning_rate": 8.18353289531551e-06, "loss": 3.4545, "step": 2695 }, { "epoch": 0.32, "learning_rate": 8.18198590089833e-06, "loss": 3.4854, "step": 2696 }, { "epoch": 0.32, "learning_rate": 8.180438394370813e-06, "loss": 3.3622, "step": 2697 }, { "epoch": 0.32, "learning_rate": 8.178890375982015e-06, "loss": 3.4275, "step": 2698 }, { "epoch": 0.32, "learning_rate": 8.17734184598107e-06, "loss": 3.4168, "step": 2699 }, { "epoch": 0.32, "learning_rate": 8.175792804617209e-06, "loss": 3.494, "step": 2700 }, { "epoch": 0.32, "learning_rate": 8.174243252139728e-06, "loss": 3.3903, "step": 2701 }, { "epoch": 0.32, "learning_rate": 8.172693188798016e-06, "loss": 3.4152, "step": 2702 }, { "epoch": 0.32, "learning_rate": 8.171142614841538e-06, "loss": 3.4764, "step": 2703 }, { "epoch": 0.32, "learning_rate": 8.169591530519848e-06, "loss": 3.3196, "step": 2704 }, { "epoch": 0.32, "learning_rate": 8.168039936082577e-06, "loss": 3.4705, "step": 2705 }, { "epoch": 0.32, "learning_rate": 8.166487831779439e-06, "loss": 3.501, "step": 2706 }, { "epoch": 0.32, "learning_rate": 8.164935217860228e-06, "loss": 3.3883, "step": 2707 }, { "epoch": 0.33, "learning_rate": 8.163382094574825e-06, "loss": 3.475, "step": 2708 }, { "epoch": 0.33, "learning_rate": 8.16182846217319e-06, "loss": 3.4476, "step": 2709 }, { "epoch": 0.33, "learning_rate": 8.160274320905369e-06, "loss": 3.4194, "step": 2710 }, { "epoch": 0.33, "learning_rate": 8.158719671021478e-06, "loss": 3.4499, "step": 2711 }, { "epoch": 0.33, "learning_rate": 8.157164512771729e-06, "loss": 3.3697, "step": 2712 }, { "epoch": 0.33, "learning_rate": 8.15560884640641e-06, "loss": 3.404, "step": 2713 }, { "epoch": 0.33, "learning_rate": 8.154052672175888e-06, "loss": 3.2945, "step": 2714 }, { "epoch": 0.33, "learning_rate": 8.152495990330617e-06, "loss": 3.4, "step": 2715 }, { "epoch": 0.33, "learning_rate": 8.150938801121126e-06, "loss": 3.5085, "step": 2716 }, { "epoch": 0.33, "learning_rate": 8.149381104798035e-06, "loss": 3.4001, "step": 2717 }, { "epoch": 0.33, "learning_rate": 8.147822901612037e-06, "loss": 3.348, "step": 2718 }, { "epoch": 0.33, "learning_rate": 8.146264191813913e-06, "loss": 3.4451, "step": 2719 }, { "epoch": 0.33, "learning_rate": 8.144704975654519e-06, "loss": 3.4596, "step": 2720 }, { "epoch": 0.33, "learning_rate": 8.143145253384798e-06, "loss": 3.3677, "step": 2721 }, { "epoch": 0.33, "learning_rate": 8.141585025255772e-06, "loss": 3.4259, "step": 2722 }, { "epoch": 0.33, "learning_rate": 8.140024291518543e-06, "loss": 3.439, "step": 2723 }, { "epoch": 0.33, "learning_rate": 8.1384630524243e-06, "loss": 3.4287, "step": 2724 }, { "epoch": 0.33, "learning_rate": 8.136901308224308e-06, "loss": 3.522, "step": 2725 }, { "epoch": 0.33, "learning_rate": 8.135339059169912e-06, "loss": 3.3632, "step": 2726 }, { "epoch": 0.33, "learning_rate": 8.133776305512545e-06, "loss": 3.4064, "step": 2727 }, { "epoch": 0.33, "learning_rate": 8.132213047503717e-06, "loss": 3.371, "step": 2728 }, { "epoch": 0.33, "learning_rate": 8.130649285395018e-06, "loss": 3.433, "step": 2729 }, { "epoch": 0.33, "learning_rate": 8.129085019438119e-06, "loss": 3.3995, "step": 2730 }, { "epoch": 0.33, "learning_rate": 8.127520249884777e-06, "loss": 3.3608, "step": 2731 }, { "epoch": 0.33, "learning_rate": 8.125954976986828e-06, "loss": 3.4388, "step": 2732 }, { "epoch": 0.33, "learning_rate": 8.124389200996181e-06, "loss": 3.4769, "step": 2733 }, { "epoch": 0.33, "learning_rate": 8.122822922164839e-06, "loss": 3.4408, "step": 2734 }, { "epoch": 0.33, "learning_rate": 8.121256140744878e-06, "loss": 3.3959, "step": 2735 }, { "epoch": 0.33, "learning_rate": 8.119688856988458e-06, "loss": 3.4729, "step": 2736 }, { "epoch": 0.33, "learning_rate": 8.118121071147813e-06, "loss": 3.447, "step": 2737 }, { "epoch": 0.33, "learning_rate": 8.11655278347527e-06, "loss": 3.415, "step": 2738 }, { "epoch": 0.33, "learning_rate": 8.114983994223224e-06, "loss": 3.4936, "step": 2739 }, { "epoch": 0.33, "learning_rate": 8.11341470364416e-06, "loss": 3.5015, "step": 2740 }, { "epoch": 0.33, "learning_rate": 8.111844911990641e-06, "loss": 3.3132, "step": 2741 }, { "epoch": 0.33, "learning_rate": 8.110274619515308e-06, "loss": 3.2247, "step": 2742 }, { "epoch": 0.33, "learning_rate": 8.108703826470888e-06, "loss": 3.4404, "step": 2743 }, { "epoch": 0.33, "learning_rate": 8.10713253311018e-06, "loss": 3.4581, "step": 2744 }, { "epoch": 0.33, "learning_rate": 8.105560739686071e-06, "loss": 3.4066, "step": 2745 }, { "epoch": 0.33, "learning_rate": 8.103988446451528e-06, "loss": 3.3132, "step": 2746 }, { "epoch": 0.33, "learning_rate": 8.102415653659596e-06, "loss": 3.3647, "step": 2747 }, { "epoch": 0.33, "learning_rate": 8.100842361563398e-06, "loss": 3.472, "step": 2748 }, { "epoch": 0.33, "learning_rate": 8.099268570416145e-06, "loss": 3.4158, "step": 2749 }, { "epoch": 0.33, "learning_rate": 8.09769428047112e-06, "loss": 3.5325, "step": 2750 }, { "epoch": 0.33, "learning_rate": 8.096119491981693e-06, "loss": 3.4335, "step": 2751 }, { "epoch": 0.33, "learning_rate": 8.094544205201308e-06, "loss": 3.4016, "step": 2752 }, { "epoch": 0.33, "learning_rate": 8.092968420383493e-06, "loss": 3.4047, "step": 2753 }, { "epoch": 0.33, "learning_rate": 8.091392137781859e-06, "loss": 3.4352, "step": 2754 }, { "epoch": 0.33, "learning_rate": 8.08981535765009e-06, "loss": 3.3416, "step": 2755 }, { "epoch": 0.33, "learning_rate": 8.088238080241954e-06, "loss": 3.3405, "step": 2756 }, { "epoch": 0.33, "learning_rate": 8.086660305811302e-06, "loss": 3.4066, "step": 2757 }, { "epoch": 0.33, "learning_rate": 8.085082034612057e-06, "loss": 3.4484, "step": 2758 }, { "epoch": 0.33, "learning_rate": 8.08350326689823e-06, "loss": 3.456, "step": 2759 }, { "epoch": 0.33, "learning_rate": 8.081924002923909e-06, "loss": 3.434, "step": 2760 }, { "epoch": 0.33, "learning_rate": 8.08034424294326e-06, "loss": 3.4817, "step": 2761 }, { "epoch": 0.33, "learning_rate": 8.078763987210529e-06, "loss": 3.4551, "step": 2762 }, { "epoch": 0.33, "learning_rate": 8.077183235980047e-06, "loss": 3.4891, "step": 2763 }, { "epoch": 0.33, "learning_rate": 8.075601989506217e-06, "loss": 3.4034, "step": 2764 }, { "epoch": 0.33, "learning_rate": 8.07402024804353e-06, "loss": 3.4598, "step": 2765 }, { "epoch": 0.33, "learning_rate": 8.072438011846547e-06, "loss": 3.3589, "step": 2766 }, { "epoch": 0.33, "learning_rate": 8.070855281169916e-06, "loss": 3.4454, "step": 2767 }, { "epoch": 0.33, "learning_rate": 8.069272056268364e-06, "loss": 3.4771, "step": 2768 }, { "epoch": 0.33, "learning_rate": 8.067688337396695e-06, "loss": 3.4479, "step": 2769 }, { "epoch": 0.33, "learning_rate": 8.066104124809793e-06, "loss": 3.3781, "step": 2770 }, { "epoch": 0.33, "learning_rate": 8.06451941876262e-06, "loss": 3.4674, "step": 2771 }, { "epoch": 0.33, "learning_rate": 8.062934219510223e-06, "loss": 3.4519, "step": 2772 }, { "epoch": 0.33, "learning_rate": 8.061348527307723e-06, "loss": 3.5199, "step": 2773 }, { "epoch": 0.33, "learning_rate": 8.059762342410324e-06, "loss": 3.5127, "step": 2774 }, { "epoch": 0.33, "learning_rate": 8.058175665073304e-06, "loss": 3.4012, "step": 2775 }, { "epoch": 0.33, "learning_rate": 8.056588495552025e-06, "loss": 3.4838, "step": 2776 }, { "epoch": 0.33, "learning_rate": 8.055000834101929e-06, "loss": 3.3889, "step": 2777 }, { "epoch": 0.33, "learning_rate": 8.053412680978532e-06, "loss": 3.3268, "step": 2778 }, { "epoch": 0.33, "learning_rate": 8.051824036437432e-06, "loss": 3.5202, "step": 2779 }, { "epoch": 0.33, "learning_rate": 8.050234900734309e-06, "loss": 3.3763, "step": 2780 }, { "epoch": 0.33, "learning_rate": 8.048645274124916e-06, "loss": 3.4692, "step": 2781 }, { "epoch": 0.33, "learning_rate": 8.04705515686509e-06, "loss": 3.4661, "step": 2782 }, { "epoch": 0.33, "learning_rate": 8.045464549210748e-06, "loss": 3.4741, "step": 2783 }, { "epoch": 0.33, "learning_rate": 8.043873451417878e-06, "loss": 3.3864, "step": 2784 }, { "epoch": 0.33, "learning_rate": 8.042281863742557e-06, "loss": 3.4431, "step": 2785 }, { "epoch": 0.33, "learning_rate": 8.040689786440929e-06, "loss": 3.3833, "step": 2786 }, { "epoch": 0.33, "learning_rate": 8.039097219769231e-06, "loss": 3.4357, "step": 2787 }, { "epoch": 0.33, "learning_rate": 8.037504163983767e-06, "loss": 3.4395, "step": 2788 }, { "epoch": 0.33, "learning_rate": 8.035910619340927e-06, "loss": 3.4821, "step": 2789 }, { "epoch": 0.33, "learning_rate": 8.034316586097174e-06, "loss": 3.3661, "step": 2790 }, { "epoch": 0.34, "learning_rate": 8.032722064509054e-06, "loss": 3.3765, "step": 2791 }, { "epoch": 0.34, "learning_rate": 8.031127054833192e-06, "loss": 3.4735, "step": 2792 }, { "epoch": 0.34, "learning_rate": 8.029531557326285e-06, "loss": 3.3524, "step": 2793 }, { "epoch": 0.34, "learning_rate": 8.027935572245116e-06, "loss": 3.4087, "step": 2794 }, { "epoch": 0.34, "learning_rate": 8.026339099846544e-06, "loss": 3.3691, "step": 2795 }, { "epoch": 0.34, "learning_rate": 8.024742140387506e-06, "loss": 3.4457, "step": 2796 }, { "epoch": 0.34, "learning_rate": 8.023144694125016e-06, "loss": 3.4905, "step": 2797 }, { "epoch": 0.34, "learning_rate": 8.021546761316168e-06, "loss": 3.4109, "step": 2798 }, { "epoch": 0.34, "learning_rate": 8.019948342218135e-06, "loss": 3.3467, "step": 2799 }, { "epoch": 0.34, "learning_rate": 8.018349437088167e-06, "loss": 3.2976, "step": 2800 }, { "epoch": 0.34, "learning_rate": 8.016750046183593e-06, "loss": 3.4093, "step": 2801 }, { "epoch": 0.34, "learning_rate": 8.015150169761819e-06, "loss": 3.4129, "step": 2802 }, { "epoch": 0.34, "learning_rate": 8.013549808080329e-06, "loss": 3.4756, "step": 2803 }, { "epoch": 0.34, "learning_rate": 8.011948961396688e-06, "loss": 3.3801, "step": 2804 }, { "epoch": 0.34, "learning_rate": 8.010347629968537e-06, "loss": 3.4354, "step": 2805 }, { "epoch": 0.34, "learning_rate": 8.008745814053594e-06, "loss": 3.4625, "step": 2806 }, { "epoch": 0.34, "learning_rate": 8.007143513909654e-06, "loss": 3.5197, "step": 2807 }, { "epoch": 0.34, "learning_rate": 8.005540729794596e-06, "loss": 3.3858, "step": 2808 }, { "epoch": 0.34, "learning_rate": 8.00393746196637e-06, "loss": 3.4151, "step": 2809 }, { "epoch": 0.34, "learning_rate": 8.00233371068301e-06, "loss": 3.4257, "step": 2810 }, { "epoch": 0.34, "learning_rate": 8.00072947620262e-06, "loss": 3.4762, "step": 2811 }, { "epoch": 0.34, "learning_rate": 7.999124758783388e-06, "loss": 3.5001, "step": 2812 }, { "epoch": 0.34, "learning_rate": 7.99751955868358e-06, "loss": 3.4352, "step": 2813 }, { "epoch": 0.34, "learning_rate": 7.995913876161536e-06, "loss": 3.4746, "step": 2814 }, { "epoch": 0.34, "learning_rate": 7.994307711475675e-06, "loss": 3.423, "step": 2815 }, { "epoch": 0.34, "learning_rate": 7.992701064884494e-06, "loss": 3.3888, "step": 2816 }, { "epoch": 0.34, "learning_rate": 7.991093936646568e-06, "loss": 3.3899, "step": 2817 }, { "epoch": 0.34, "learning_rate": 7.989486327020548e-06, "loss": 3.4486, "step": 2818 }, { "epoch": 0.34, "learning_rate": 7.987878236265167e-06, "loss": 3.336, "step": 2819 }, { "epoch": 0.34, "learning_rate": 7.986269664639227e-06, "loss": 3.4668, "step": 2820 }, { "epoch": 0.34, "learning_rate": 7.984660612401615e-06, "loss": 3.4145, "step": 2821 }, { "epoch": 0.34, "learning_rate": 7.98305107981129e-06, "loss": 3.3497, "step": 2822 }, { "epoch": 0.34, "learning_rate": 7.981441067127297e-06, "loss": 3.5161, "step": 2823 }, { "epoch": 0.34, "learning_rate": 7.979830574608747e-06, "loss": 3.432, "step": 2824 }, { "epoch": 0.34, "learning_rate": 7.978219602514833e-06, "loss": 3.4176, "step": 2825 }, { "epoch": 0.34, "learning_rate": 7.976608151104828e-06, "loss": 3.4476, "step": 2826 }, { "epoch": 0.34, "learning_rate": 7.97499622063808e-06, "loss": 3.4641, "step": 2827 }, { "epoch": 0.34, "learning_rate": 7.973383811374013e-06, "loss": 3.3442, "step": 2828 }, { "epoch": 0.34, "learning_rate": 7.971770923572127e-06, "loss": 3.4568, "step": 2829 }, { "epoch": 0.34, "learning_rate": 7.970157557492005e-06, "loss": 3.4585, "step": 2830 }, { "epoch": 0.34, "learning_rate": 7.968543713393299e-06, "loss": 3.4059, "step": 2831 }, { "epoch": 0.34, "learning_rate": 7.966929391535744e-06, "loss": 3.443, "step": 2832 }, { "epoch": 0.34, "learning_rate": 7.96531459217915e-06, "loss": 3.3516, "step": 2833 }, { "epoch": 0.34, "learning_rate": 7.963699315583402e-06, "loss": 3.4375, "step": 2834 }, { "epoch": 0.34, "learning_rate": 7.962083562008464e-06, "loss": 3.469, "step": 2835 }, { "epoch": 0.34, "learning_rate": 7.96046733171438e-06, "loss": 3.3544, "step": 2836 }, { "epoch": 0.34, "learning_rate": 7.95885062496126e-06, "loss": 3.4033, "step": 2837 }, { "epoch": 0.34, "learning_rate": 7.957233442009301e-06, "loss": 3.3742, "step": 2838 }, { "epoch": 0.34, "learning_rate": 7.955615783118773e-06, "loss": 3.4291, "step": 2839 }, { "epoch": 0.34, "learning_rate": 7.953997648550023e-06, "loss": 3.496, "step": 2840 }, { "epoch": 0.34, "learning_rate": 7.952379038563475e-06, "loss": 3.3939, "step": 2841 }, { "epoch": 0.34, "learning_rate": 7.950759953419628e-06, "loss": 3.3793, "step": 2842 }, { "epoch": 0.34, "learning_rate": 7.949140393379058e-06, "loss": 3.4017, "step": 2843 }, { "epoch": 0.34, "learning_rate": 7.947520358702418e-06, "loss": 3.2584, "step": 2844 }, { "epoch": 0.34, "learning_rate": 7.945899849650437e-06, "loss": 3.3764, "step": 2845 }, { "epoch": 0.34, "learning_rate": 7.944278866483921e-06, "loss": 3.3196, "step": 2846 }, { "epoch": 0.34, "learning_rate": 7.942657409463753e-06, "loss": 3.5359, "step": 2847 }, { "epoch": 0.34, "learning_rate": 7.94103547885089e-06, "loss": 3.4042, "step": 2848 }, { "epoch": 0.34, "learning_rate": 7.939413074906364e-06, "loss": 3.3822, "step": 2849 }, { "epoch": 0.34, "learning_rate": 7.937790197891289e-06, "loss": 3.4038, "step": 2850 }, { "epoch": 0.34, "learning_rate": 7.936166848066848e-06, "loss": 3.3585, "step": 2851 }, { "epoch": 0.34, "learning_rate": 7.934543025694308e-06, "loss": 3.3649, "step": 2852 }, { "epoch": 0.34, "learning_rate": 7.932918731035005e-06, "loss": 3.4322, "step": 2853 }, { "epoch": 0.34, "learning_rate": 7.931293964350353e-06, "loss": 3.4879, "step": 2854 }, { "epoch": 0.34, "learning_rate": 7.929668725901847e-06, "loss": 3.3815, "step": 2855 }, { "epoch": 0.34, "learning_rate": 7.928043015951049e-06, "loss": 3.3069, "step": 2856 }, { "epoch": 0.34, "learning_rate": 7.926416834759602e-06, "loss": 3.3892, "step": 2857 }, { "epoch": 0.34, "learning_rate": 7.924790182589226e-06, "loss": 3.3272, "step": 2858 }, { "epoch": 0.34, "learning_rate": 7.923163059701717e-06, "loss": 3.5053, "step": 2859 }, { "epoch": 0.34, "learning_rate": 7.92153546635894e-06, "loss": 3.4452, "step": 2860 }, { "epoch": 0.34, "learning_rate": 7.919907402822846e-06, "loss": 3.4288, "step": 2861 }, { "epoch": 0.34, "learning_rate": 7.918278869355452e-06, "loss": 3.5301, "step": 2862 }, { "epoch": 0.34, "learning_rate": 7.916649866218857e-06, "loss": 3.2899, "step": 2863 }, { "epoch": 0.34, "learning_rate": 7.915020393675233e-06, "loss": 3.4164, "step": 2864 }, { "epoch": 0.34, "learning_rate": 7.913390451986829e-06, "loss": 3.4451, "step": 2865 }, { "epoch": 0.34, "learning_rate": 7.911760041415967e-06, "loss": 3.5509, "step": 2866 }, { "epoch": 0.34, "learning_rate": 7.910129162225048e-06, "loss": 3.3869, "step": 2867 }, { "epoch": 0.34, "learning_rate": 7.908497814676545e-06, "loss": 3.4648, "step": 2868 }, { "epoch": 0.34, "learning_rate": 7.90686599903301e-06, "loss": 3.3959, "step": 2869 }, { "epoch": 0.34, "learning_rate": 7.905233715557063e-06, "loss": 3.35, "step": 2870 }, { "epoch": 0.34, "learning_rate": 7.90360096451141e-06, "loss": 3.4566, "step": 2871 }, { "epoch": 0.34, "learning_rate": 7.901967746158824e-06, "loss": 3.4061, "step": 2872 }, { "epoch": 0.34, "learning_rate": 7.900334060762158e-06, "loss": 3.4751, "step": 2873 }, { "epoch": 0.34, "learning_rate": 7.898699908584335e-06, "loss": 3.4343, "step": 2874 }, { "epoch": 0.35, "learning_rate": 7.89706528988836e-06, "loss": 3.5052, "step": 2875 }, { "epoch": 0.35, "learning_rate": 7.895430204937305e-06, "loss": 3.366, "step": 2876 }, { "epoch": 0.35, "learning_rate": 7.893794653994327e-06, "loss": 3.3551, "step": 2877 }, { "epoch": 0.35, "learning_rate": 7.892158637322647e-06, "loss": 3.3694, "step": 2878 }, { "epoch": 0.35, "learning_rate": 7.890522155185567e-06, "loss": 3.399, "step": 2879 }, { "epoch": 0.35, "learning_rate": 7.888885207846466e-06, "loss": 3.4592, "step": 2880 }, { "epoch": 0.35, "learning_rate": 7.887247795568794e-06, "loss": 3.4005, "step": 2881 }, { "epoch": 0.35, "learning_rate": 7.885609918616077e-06, "loss": 3.4523, "step": 2882 }, { "epoch": 0.35, "learning_rate": 7.883971577251913e-06, "loss": 3.3543, "step": 2883 }, { "epoch": 0.35, "learning_rate": 7.882332771739982e-06, "loss": 3.3031, "step": 2884 }, { "epoch": 0.35, "learning_rate": 7.88069350234403e-06, "loss": 3.3713, "step": 2885 }, { "epoch": 0.35, "learning_rate": 7.879053769327886e-06, "loss": 3.4727, "step": 2886 }, { "epoch": 0.35, "learning_rate": 7.877413572955446e-06, "loss": 3.3277, "step": 2887 }, { "epoch": 0.35, "learning_rate": 7.875772913490685e-06, "loss": 3.4181, "step": 2888 }, { "epoch": 0.35, "learning_rate": 7.874131791197651e-06, "loss": 3.3805, "step": 2889 }, { "epoch": 0.35, "learning_rate": 7.872490206340469e-06, "loss": 3.5122, "step": 2890 }, { "epoch": 0.35, "learning_rate": 7.870848159183333e-06, "loss": 3.3399, "step": 2891 }, { "epoch": 0.35, "learning_rate": 7.869205649990518e-06, "loss": 3.4371, "step": 2892 }, { "epoch": 0.35, "learning_rate": 7.867562679026371e-06, "loss": 3.401, "step": 2893 }, { "epoch": 0.35, "learning_rate": 7.865919246555306e-06, "loss": 3.4077, "step": 2894 }, { "epoch": 0.35, "learning_rate": 7.864275352841829e-06, "loss": 3.4462, "step": 2895 }, { "epoch": 0.35, "learning_rate": 7.862630998150498e-06, "loss": 3.3841, "step": 2896 }, { "epoch": 0.35, "learning_rate": 7.860986182745962e-06, "loss": 3.4804, "step": 2897 }, { "epoch": 0.35, "learning_rate": 7.859340906892936e-06, "loss": 3.3523, "step": 2898 }, { "epoch": 0.35, "learning_rate": 7.857695170856214e-06, "loss": 3.2967, "step": 2899 }, { "epoch": 0.35, "learning_rate": 7.85604897490066e-06, "loss": 3.4175, "step": 2900 }, { "epoch": 0.35, "learning_rate": 7.854402319291213e-06, "loss": 3.4713, "step": 2901 }, { "epoch": 0.35, "learning_rate": 7.852755204292887e-06, "loss": 3.457, "step": 2902 }, { "epoch": 0.35, "learning_rate": 7.85110763017077e-06, "loss": 3.4644, "step": 2903 }, { "epoch": 0.35, "learning_rate": 7.849459597190025e-06, "loss": 3.4669, "step": 2904 }, { "epoch": 0.35, "learning_rate": 7.847811105615884e-06, "loss": 3.405, "step": 2905 }, { "epoch": 0.35, "learning_rate": 7.846162155713659e-06, "loss": 3.502, "step": 2906 }, { "epoch": 0.35, "learning_rate": 7.844512747748728e-06, "loss": 3.5113, "step": 2907 }, { "epoch": 0.35, "learning_rate": 7.842862881986553e-06, "loss": 3.4559, "step": 2908 }, { "epoch": 0.35, "learning_rate": 7.84121255869266e-06, "loss": 3.4417, "step": 2909 }, { "epoch": 0.35, "learning_rate": 7.839561778132657e-06, "loss": 3.3743, "step": 2910 }, { "epoch": 0.35, "learning_rate": 7.837910540572218e-06, "loss": 3.3438, "step": 2911 }, { "epoch": 0.35, "learning_rate": 7.836258846277094e-06, "loss": 3.3839, "step": 2912 }, { "epoch": 0.35, "learning_rate": 7.834606695513111e-06, "loss": 3.3929, "step": 2913 }, { "epoch": 0.35, "learning_rate": 7.832954088546168e-06, "loss": 3.4088, "step": 2914 }, { "epoch": 0.35, "learning_rate": 7.831301025642234e-06, "loss": 3.3961, "step": 2915 }, { "epoch": 0.35, "learning_rate": 7.829647507067355e-06, "loss": 3.4329, "step": 2916 }, { "epoch": 0.35, "learning_rate": 7.827993533087647e-06, "loss": 3.3605, "step": 2917 }, { "epoch": 0.35, "learning_rate": 7.826339103969305e-06, "loss": 3.3484, "step": 2918 }, { "epoch": 0.35, "learning_rate": 7.824684219978591e-06, "loss": 3.4059, "step": 2919 }, { "epoch": 0.35, "learning_rate": 7.823028881381842e-06, "loss": 3.3537, "step": 2920 }, { "epoch": 0.35, "learning_rate": 7.821373088445472e-06, "loss": 3.3575, "step": 2921 }, { "epoch": 0.35, "learning_rate": 7.819716841435963e-06, "loss": 3.5212, "step": 2922 }, { "epoch": 0.35, "learning_rate": 7.818060140619874e-06, "loss": 3.3179, "step": 2923 }, { "epoch": 0.35, "learning_rate": 7.816402986263831e-06, "loss": 3.3733, "step": 2924 }, { "epoch": 0.35, "learning_rate": 7.814745378634544e-06, "loss": 3.4453, "step": 2925 }, { "epoch": 0.35, "learning_rate": 7.813087317998782e-06, "loss": 3.3836, "step": 2926 }, { "epoch": 0.35, "learning_rate": 7.8114288046234e-06, "loss": 3.446, "step": 2927 }, { "epoch": 0.35, "learning_rate": 7.809769838775316e-06, "loss": 3.3894, "step": 2928 }, { "epoch": 0.35, "learning_rate": 7.808110420721527e-06, "loss": 3.4475, "step": 2929 }, { "epoch": 0.35, "learning_rate": 7.806450550729098e-06, "loss": 3.4174, "step": 2930 }, { "epoch": 0.35, "learning_rate": 7.804790229065173e-06, "loss": 3.4081, "step": 2931 }, { "epoch": 0.35, "learning_rate": 7.803129455996961e-06, "loss": 3.415, "step": 2932 }, { "epoch": 0.35, "learning_rate": 7.80146823179175e-06, "loss": 3.401, "step": 2933 }, { "epoch": 0.35, "learning_rate": 7.799806556716898e-06, "loss": 3.4416, "step": 2934 }, { "epoch": 0.35, "learning_rate": 7.798144431039835e-06, "loss": 3.3629, "step": 2935 }, { "epoch": 0.35, "learning_rate": 7.796481855028066e-06, "loss": 3.503, "step": 2936 }, { "epoch": 0.35, "learning_rate": 7.794818828949164e-06, "loss": 3.4376, "step": 2937 }, { "epoch": 0.35, "learning_rate": 7.79315535307078e-06, "loss": 3.3658, "step": 2938 }, { "epoch": 0.35, "learning_rate": 7.791491427660633e-06, "loss": 3.4488, "step": 2939 }, { "epoch": 0.35, "learning_rate": 7.789827052986519e-06, "loss": 3.4008, "step": 2940 }, { "epoch": 0.35, "learning_rate": 7.788162229316297e-06, "loss": 3.3752, "step": 2941 }, { "epoch": 0.35, "learning_rate": 7.786496956917908e-06, "loss": 3.4168, "step": 2942 }, { "epoch": 0.35, "learning_rate": 7.784831236059365e-06, "loss": 3.331, "step": 2943 }, { "epoch": 0.35, "learning_rate": 7.783165067008744e-06, "loss": 3.4512, "step": 2944 }, { "epoch": 0.35, "learning_rate": 7.781498450034203e-06, "loss": 3.489, "step": 2945 }, { "epoch": 0.35, "learning_rate": 7.779831385403966e-06, "loss": 3.4239, "step": 2946 }, { "epoch": 0.35, "learning_rate": 7.778163873386334e-06, "loss": 3.4813, "step": 2947 }, { "epoch": 0.35, "learning_rate": 7.776495914249674e-06, "loss": 3.5218, "step": 2948 }, { "epoch": 0.35, "learning_rate": 7.774827508262431e-06, "loss": 3.4336, "step": 2949 }, { "epoch": 0.35, "learning_rate": 7.773158655693116e-06, "loss": 3.3665, "step": 2950 }, { "epoch": 0.35, "learning_rate": 7.771489356810317e-06, "loss": 3.3881, "step": 2951 }, { "epoch": 0.35, "learning_rate": 7.769819611882691e-06, "loss": 3.4386, "step": 2952 }, { "epoch": 0.35, "learning_rate": 7.768149421178968e-06, "loss": 3.4681, "step": 2953 }, { "epoch": 0.35, "learning_rate": 7.766478784967949e-06, "loss": 3.4142, "step": 2954 }, { "epoch": 0.35, "learning_rate": 7.764807703518506e-06, "loss": 3.4842, "step": 2955 }, { "epoch": 0.35, "learning_rate": 7.763136177099585e-06, "loss": 3.4408, "step": 2956 }, { "epoch": 0.35, "learning_rate": 7.761464205980205e-06, "loss": 3.4605, "step": 2957 }, { "epoch": 0.36, "learning_rate": 7.759791790429449e-06, "loss": 3.4079, "step": 2958 }, { "epoch": 0.36, "learning_rate": 7.758118930716477e-06, "loss": 3.4494, "step": 2959 }, { "epoch": 0.36, "learning_rate": 7.756445627110523e-06, "loss": 3.4617, "step": 2960 }, { "epoch": 0.36, "learning_rate": 7.754771879880886e-06, "loss": 3.4634, "step": 2961 }, { "epoch": 0.36, "learning_rate": 7.753097689296941e-06, "loss": 3.4315, "step": 2962 }, { "epoch": 0.36, "learning_rate": 7.751423055628133e-06, "loss": 3.4951, "step": 2963 }, { "epoch": 0.36, "learning_rate": 7.74974797914398e-06, "loss": 3.4782, "step": 2964 }, { "epoch": 0.36, "learning_rate": 7.748072460114065e-06, "loss": 3.3971, "step": 2965 }, { "epoch": 0.36, "learning_rate": 7.746396498808052e-06, "loss": 3.4366, "step": 2966 }, { "epoch": 0.36, "learning_rate": 7.744720095495669e-06, "loss": 3.4667, "step": 2967 }, { "epoch": 0.36, "learning_rate": 7.743043250446715e-06, "loss": 3.3469, "step": 2968 }, { "epoch": 0.36, "learning_rate": 7.741365963931068e-06, "loss": 3.3668, "step": 2969 }, { "epoch": 0.36, "learning_rate": 7.739688236218665e-06, "loss": 3.4445, "step": 2970 }, { "epoch": 0.36, "learning_rate": 7.738010067579523e-06, "loss": 3.4947, "step": 2971 }, { "epoch": 0.36, "learning_rate": 7.736331458283728e-06, "loss": 3.3973, "step": 2972 }, { "epoch": 0.36, "learning_rate": 7.734652408601432e-06, "loss": 3.4676, "step": 2973 }, { "epoch": 0.36, "learning_rate": 7.73297291880287e-06, "loss": 3.3794, "step": 2974 }, { "epoch": 0.36, "learning_rate": 7.731292989158331e-06, "loss": 3.5442, "step": 2975 }, { "epoch": 0.36, "learning_rate": 7.729612619938192e-06, "loss": 3.3203, "step": 2976 }, { "epoch": 0.36, "learning_rate": 7.727931811412885e-06, "loss": 3.4053, "step": 2977 }, { "epoch": 0.36, "learning_rate": 7.726250563852927e-06, "loss": 3.4737, "step": 2978 }, { "epoch": 0.36, "learning_rate": 7.724568877528893e-06, "loss": 3.5051, "step": 2979 }, { "epoch": 0.36, "learning_rate": 7.722886752711436e-06, "loss": 3.4621, "step": 2980 }, { "epoch": 0.36, "learning_rate": 7.72120418967128e-06, "loss": 3.4288, "step": 2981 }, { "epoch": 0.36, "learning_rate": 7.719521188679215e-06, "loss": 3.4499, "step": 2982 }, { "epoch": 0.36, "learning_rate": 7.717837750006106e-06, "loss": 3.4321, "step": 2983 }, { "epoch": 0.36, "learning_rate": 7.716153873922886e-06, "loss": 3.4478, "step": 2984 }, { "epoch": 0.36, "learning_rate": 7.71446956070056e-06, "loss": 3.3672, "step": 2985 }, { "epoch": 0.36, "learning_rate": 7.712784810610196e-06, "loss": 3.4486, "step": 2986 }, { "epoch": 0.36, "learning_rate": 7.711099623922946e-06, "loss": 3.4167, "step": 2987 }, { "epoch": 0.36, "learning_rate": 7.709414000910023e-06, "loss": 3.2589, "step": 2988 }, { "epoch": 0.36, "learning_rate": 7.707727941842708e-06, "loss": 3.4489, "step": 2989 }, { "epoch": 0.36, "learning_rate": 7.70604144699236e-06, "loss": 3.4586, "step": 2990 }, { "epoch": 0.36, "learning_rate": 7.704354516630403e-06, "loss": 3.3616, "step": 2991 }, { "epoch": 0.36, "learning_rate": 7.702667151028334e-06, "loss": 3.4427, "step": 2992 }, { "epoch": 0.36, "learning_rate": 7.700979350457717e-06, "loss": 3.41, "step": 2993 }, { "epoch": 0.36, "learning_rate": 7.699291115190185e-06, "loss": 3.3574, "step": 2994 }, { "epoch": 0.36, "learning_rate": 7.697602445497449e-06, "loss": 3.4049, "step": 2995 }, { "epoch": 0.36, "learning_rate": 7.69591334165128e-06, "loss": 3.3403, "step": 2996 }, { "epoch": 0.36, "learning_rate": 7.694223803923523e-06, "loss": 3.4007, "step": 2997 }, { "epoch": 0.36, "learning_rate": 7.692533832586097e-06, "loss": 3.4138, "step": 2998 }, { "epoch": 0.36, "learning_rate": 7.690843427910984e-06, "loss": 3.3333, "step": 2999 }, { "epoch": 0.36, "learning_rate": 7.689152590170236e-06, "loss": 3.4716, "step": 3000 }, { "epoch": 0.36, "learning_rate": 7.68746131963598e-06, "loss": 3.3773, "step": 3001 }, { "epoch": 0.36, "learning_rate": 7.685769616580412e-06, "loss": 3.3488, "step": 3002 }, { "epoch": 0.36, "learning_rate": 7.684077481275795e-06, "loss": 3.4125, "step": 3003 }, { "epoch": 0.36, "learning_rate": 7.682384913994458e-06, "loss": 3.3043, "step": 3004 }, { "epoch": 0.36, "learning_rate": 7.680691915008806e-06, "loss": 3.3715, "step": 3005 }, { "epoch": 0.36, "learning_rate": 7.678998484591312e-06, "loss": 3.3947, "step": 3006 }, { "epoch": 0.36, "learning_rate": 7.677304623014514e-06, "loss": 3.3714, "step": 3007 }, { "epoch": 0.36, "learning_rate": 7.675610330551029e-06, "loss": 3.3655, "step": 3008 }, { "epoch": 0.36, "learning_rate": 7.673915607473531e-06, "loss": 3.4919, "step": 3009 }, { "epoch": 0.36, "learning_rate": 7.672220454054773e-06, "loss": 3.4313, "step": 3010 }, { "epoch": 0.36, "learning_rate": 7.670524870567574e-06, "loss": 3.4676, "step": 3011 }, { "epoch": 0.36, "learning_rate": 7.668828857284821e-06, "loss": 3.5135, "step": 3012 }, { "epoch": 0.36, "learning_rate": 7.667132414479471e-06, "loss": 3.4188, "step": 3013 }, { "epoch": 0.36, "learning_rate": 7.665435542424551e-06, "loss": 3.3703, "step": 3014 }, { "epoch": 0.36, "learning_rate": 7.663738241393155e-06, "loss": 3.4748, "step": 3015 }, { "epoch": 0.36, "learning_rate": 7.662040511658448e-06, "loss": 3.365, "step": 3016 }, { "epoch": 0.36, "learning_rate": 7.660342353493666e-06, "loss": 3.3085, "step": 3017 }, { "epoch": 0.36, "learning_rate": 7.658643767172108e-06, "loss": 3.5768, "step": 3018 }, { "epoch": 0.36, "learning_rate": 7.656944752967148e-06, "loss": 3.3532, "step": 3019 }, { "epoch": 0.36, "learning_rate": 7.655245311152223e-06, "loss": 3.4086, "step": 3020 }, { "epoch": 0.36, "learning_rate": 7.653545442000845e-06, "loss": 3.477, "step": 3021 }, { "epoch": 0.36, "learning_rate": 7.65184514578659e-06, "loss": 3.3854, "step": 3022 }, { "epoch": 0.36, "learning_rate": 7.650144422783104e-06, "loss": 3.4228, "step": 3023 }, { "epoch": 0.36, "learning_rate": 7.648443273264105e-06, "loss": 3.3829, "step": 3024 }, { "epoch": 0.36, "learning_rate": 7.646741697503373e-06, "loss": 3.4628, "step": 3025 }, { "epoch": 0.36, "learning_rate": 7.645039695774763e-06, "loss": 3.4239, "step": 3026 }, { "epoch": 0.36, "learning_rate": 7.643337268352196e-06, "loss": 3.3628, "step": 3027 }, { "epoch": 0.36, "learning_rate": 7.64163441550966e-06, "loss": 3.309, "step": 3028 }, { "epoch": 0.36, "learning_rate": 7.639931137521214e-06, "loss": 3.4323, "step": 3029 }, { "epoch": 0.36, "learning_rate": 7.638227434660986e-06, "loss": 3.3589, "step": 3030 }, { "epoch": 0.36, "learning_rate": 7.636523307203168e-06, "loss": 3.3842, "step": 3031 }, { "epoch": 0.36, "learning_rate": 7.634818755422022e-06, "loss": 3.4577, "step": 3032 }, { "epoch": 0.36, "learning_rate": 7.633113779591883e-06, "loss": 3.3667, "step": 3033 }, { "epoch": 0.36, "learning_rate": 7.631408379987149e-06, "loss": 3.4017, "step": 3034 }, { "epoch": 0.36, "learning_rate": 7.629702556882289e-06, "loss": 3.4049, "step": 3035 }, { "epoch": 0.36, "learning_rate": 7.627996310551835e-06, "loss": 3.3822, "step": 3036 }, { "epoch": 0.36, "learning_rate": 7.626289641270397e-06, "loss": 3.5008, "step": 3037 }, { "epoch": 0.36, "learning_rate": 7.624582549312642e-06, "loss": 3.4656, "step": 3038 }, { "epoch": 0.36, "learning_rate": 7.622875034953315e-06, "loss": 3.3358, "step": 3039 }, { "epoch": 0.36, "learning_rate": 7.621167098467218e-06, "loss": 3.4804, "step": 3040 }, { "epoch": 0.37, "learning_rate": 7.619458740129232e-06, "loss": 3.4472, "step": 3041 }, { "epoch": 0.37, "learning_rate": 7.6177499602143e-06, "loss": 3.3718, "step": 3042 }, { "epoch": 0.37, "learning_rate": 7.616040758997432e-06, "loss": 3.4445, "step": 3043 }, { "epoch": 0.37, "learning_rate": 7.614331136753711e-06, "loss": 3.3509, "step": 3044 }, { "epoch": 0.37, "learning_rate": 7.6126210937582825e-06, "loss": 3.352, "step": 3045 }, { "epoch": 0.37, "learning_rate": 7.610910630286361e-06, "loss": 3.4175, "step": 3046 }, { "epoch": 0.37, "learning_rate": 7.6091997466132295e-06, "loss": 3.3841, "step": 3047 }, { "epoch": 0.37, "learning_rate": 7.60748844301424e-06, "loss": 3.4212, "step": 3048 }, { "epoch": 0.37, "learning_rate": 7.605776719764811e-06, "loss": 3.4484, "step": 3049 }, { "epoch": 0.37, "learning_rate": 7.604064577140424e-06, "loss": 3.2866, "step": 3050 }, { "epoch": 0.37, "learning_rate": 7.6023520154166355e-06, "loss": 3.3326, "step": 3051 }, { "epoch": 0.37, "learning_rate": 7.600639034869064e-06, "loss": 3.5087, "step": 3052 }, { "epoch": 0.37, "learning_rate": 7.5989256357734005e-06, "loss": 3.3884, "step": 3053 }, { "epoch": 0.37, "learning_rate": 7.597211818405398e-06, "loss": 3.3172, "step": 3054 }, { "epoch": 0.37, "learning_rate": 7.595497583040878e-06, "loss": 3.4159, "step": 3055 }, { "epoch": 0.37, "learning_rate": 7.593782929955733e-06, "loss": 3.3772, "step": 3056 }, { "epoch": 0.37, "learning_rate": 7.592067859425918e-06, "loss": 3.4678, "step": 3057 }, { "epoch": 0.37, "learning_rate": 7.5903523717274585e-06, "loss": 3.3767, "step": 3058 }, { "epoch": 0.37, "learning_rate": 7.588636467136445e-06, "loss": 3.3605, "step": 3059 }, { "epoch": 0.37, "learning_rate": 7.5869201459290375e-06, "loss": 3.3195, "step": 3060 }, { "epoch": 0.37, "learning_rate": 7.585203408381461e-06, "loss": 3.4888, "step": 3061 }, { "epoch": 0.37, "learning_rate": 7.583486254770007e-06, "loss": 3.4479, "step": 3062 }, { "epoch": 0.37, "learning_rate": 7.581768685371036e-06, "loss": 3.3745, "step": 3063 }, { "epoch": 0.37, "learning_rate": 7.580050700460972e-06, "loss": 3.5303, "step": 3064 }, { "epoch": 0.37, "learning_rate": 7.578332300316313e-06, "loss": 3.4503, "step": 3065 }, { "epoch": 0.37, "learning_rate": 7.576613485213615e-06, "loss": 3.4162, "step": 3066 }, { "epoch": 0.37, "learning_rate": 7.574894255429508e-06, "loss": 3.4708, "step": 3067 }, { "epoch": 0.37, "learning_rate": 7.573174611240684e-06, "loss": 3.3896, "step": 3068 }, { "epoch": 0.37, "learning_rate": 7.571454552923903e-06, "loss": 3.3917, "step": 3069 }, { "epoch": 0.37, "learning_rate": 7.569734080755993e-06, "loss": 3.416, "step": 3070 }, { "epoch": 0.37, "learning_rate": 7.568013195013847e-06, "loss": 3.395, "step": 3071 }, { "epoch": 0.37, "learning_rate": 7.566291895974425e-06, "loss": 3.5257, "step": 3072 }, { "epoch": 0.37, "learning_rate": 7.564570183914754e-06, "loss": 3.5134, "step": 3073 }, { "epoch": 0.37, "learning_rate": 7.562848059111927e-06, "loss": 3.4826, "step": 3074 }, { "epoch": 0.37, "learning_rate": 7.5611255218431045e-06, "loss": 3.4049, "step": 3075 }, { "epoch": 0.37, "learning_rate": 7.5594025723855105e-06, "loss": 3.4597, "step": 3076 }, { "epoch": 0.37, "learning_rate": 7.55767921101644e-06, "loss": 3.3732, "step": 3077 }, { "epoch": 0.37, "learning_rate": 7.55595543801325e-06, "loss": 3.4245, "step": 3078 }, { "epoch": 0.37, "learning_rate": 7.554231253653365e-06, "loss": 3.4503, "step": 3079 }, { "epoch": 0.37, "learning_rate": 7.552506658214277e-06, "loss": 3.4516, "step": 3080 }, { "epoch": 0.37, "learning_rate": 7.550781651973542e-06, "loss": 3.4054, "step": 3081 }, { "epoch": 0.37, "learning_rate": 7.549056235208785e-06, "loss": 3.4908, "step": 3082 }, { "epoch": 0.37, "learning_rate": 7.547330408197695e-06, "loss": 3.3607, "step": 3083 }, { "epoch": 0.37, "learning_rate": 7.5456041712180265e-06, "loss": 3.3734, "step": 3084 }, { "epoch": 0.37, "learning_rate": 7.543877524547602e-06, "loss": 3.4322, "step": 3085 }, { "epoch": 0.37, "learning_rate": 7.542150468464307e-06, "loss": 3.3432, "step": 3086 }, { "epoch": 0.37, "learning_rate": 7.540423003246096e-06, "loss": 3.5016, "step": 3087 }, { "epoch": 0.37, "learning_rate": 7.538695129170989e-06, "loss": 3.5289, "step": 3088 }, { "epoch": 0.37, "learning_rate": 7.53696684651707e-06, "loss": 3.4863, "step": 3089 }, { "epoch": 0.37, "learning_rate": 7.535238155562489e-06, "loss": 3.4409, "step": 3090 }, { "epoch": 0.37, "learning_rate": 7.533509056585463e-06, "loss": 3.465, "step": 3091 }, { "epoch": 0.37, "learning_rate": 7.531779549864273e-06, "loss": 3.3312, "step": 3092 }, { "epoch": 0.37, "learning_rate": 7.530049635677268e-06, "loss": 3.4502, "step": 3093 }, { "epoch": 0.37, "learning_rate": 7.5283193143028625e-06, "loss": 3.4324, "step": 3094 }, { "epoch": 0.37, "learning_rate": 7.526588586019532e-06, "loss": 3.4389, "step": 3095 }, { "epoch": 0.37, "learning_rate": 7.524857451105823e-06, "loss": 3.3675, "step": 3096 }, { "epoch": 0.37, "learning_rate": 7.523125909840343e-06, "loss": 3.3413, "step": 3097 }, { "epoch": 0.37, "learning_rate": 7.521393962501771e-06, "loss": 3.4517, "step": 3098 }, { "epoch": 0.37, "learning_rate": 7.519661609368843e-06, "loss": 3.3848, "step": 3099 }, { "epoch": 0.37, "learning_rate": 7.517928850720368e-06, "loss": 3.3244, "step": 3100 }, { "epoch": 0.37, "learning_rate": 7.516195686835215e-06, "loss": 3.3804, "step": 3101 }, { "epoch": 0.37, "learning_rate": 7.514462117992321e-06, "loss": 3.382, "step": 3102 }, { "epoch": 0.37, "learning_rate": 7.512728144470688e-06, "loss": 3.4398, "step": 3103 }, { "epoch": 0.37, "learning_rate": 7.5109937665493815e-06, "loss": 3.4437, "step": 3104 }, { "epoch": 0.37, "learning_rate": 7.5092589845075325e-06, "loss": 3.496, "step": 3105 }, { "epoch": 0.37, "learning_rate": 7.507523798624341e-06, "loss": 3.4453, "step": 3106 }, { "epoch": 0.37, "learning_rate": 7.505788209179064e-06, "loss": 3.392, "step": 3107 }, { "epoch": 0.37, "learning_rate": 7.5040522164510325e-06, "loss": 3.4336, "step": 3108 }, { "epoch": 0.37, "learning_rate": 7.502315820719634e-06, "loss": 3.447, "step": 3109 }, { "epoch": 0.37, "learning_rate": 7.500579022264327e-06, "loss": 3.4341, "step": 3110 }, { "epoch": 0.37, "learning_rate": 7.498841821364633e-06, "loss": 3.3826, "step": 3111 }, { "epoch": 0.37, "learning_rate": 7.497104218300139e-06, "loss": 3.5363, "step": 3112 }, { "epoch": 0.37, "learning_rate": 7.4953662133504925e-06, "loss": 3.4181, "step": 3113 }, { "epoch": 0.37, "learning_rate": 7.49362780679541e-06, "loss": 3.4054, "step": 3114 }, { "epoch": 0.37, "learning_rate": 7.491888998914671e-06, "loss": 3.4481, "step": 3115 }, { "epoch": 0.37, "learning_rate": 7.490149789988124e-06, "loss": 3.2744, "step": 3116 }, { "epoch": 0.37, "learning_rate": 7.488410180295673e-06, "loss": 3.4754, "step": 3117 }, { "epoch": 0.37, "learning_rate": 7.486670170117295e-06, "loss": 3.3975, "step": 3118 }, { "epoch": 0.37, "learning_rate": 7.484929759733026e-06, "loss": 3.3965, "step": 3119 }, { "epoch": 0.37, "learning_rate": 7.4831889494229695e-06, "loss": 3.3533, "step": 3120 }, { "epoch": 0.37, "learning_rate": 7.4814477394672925e-06, "loss": 3.4794, "step": 3121 }, { "epoch": 0.37, "learning_rate": 7.479706130146225e-06, "loss": 3.5085, "step": 3122 }, { "epoch": 0.37, "learning_rate": 7.477964121740064e-06, "loss": 3.3373, "step": 3123 }, { "epoch": 0.37, "learning_rate": 7.476221714529167e-06, "loss": 3.4213, "step": 3124 }, { "epoch": 0.38, "learning_rate": 7.474478908793961e-06, "loss": 3.5044, "step": 3125 }, { "epoch": 0.38, "learning_rate": 7.472735704814931e-06, "loss": 3.4272, "step": 3126 }, { "epoch": 0.38, "learning_rate": 7.470992102872631e-06, "loss": 3.3897, "step": 3127 }, { "epoch": 0.38, "learning_rate": 7.469248103247675e-06, "loss": 3.4268, "step": 3128 }, { "epoch": 0.38, "learning_rate": 7.467503706220747e-06, "loss": 3.4013, "step": 3129 }, { "epoch": 0.38, "learning_rate": 7.465758912072587e-06, "loss": 3.4333, "step": 3130 }, { "epoch": 0.38, "learning_rate": 7.4640137210840055e-06, "loss": 3.3838, "step": 3131 }, { "epoch": 0.38, "learning_rate": 7.462268133535874e-06, "loss": 3.3713, "step": 3132 }, { "epoch": 0.38, "learning_rate": 7.460522149709126e-06, "loss": 3.3703, "step": 3133 }, { "epoch": 0.38, "learning_rate": 7.458775769884765e-06, "loss": 3.3692, "step": 3134 }, { "epoch": 0.38, "learning_rate": 7.457028994343851e-06, "loss": 3.4486, "step": 3135 }, { "epoch": 0.38, "learning_rate": 7.455281823367512e-06, "loss": 3.3894, "step": 3136 }, { "epoch": 0.38, "learning_rate": 7.453534257236937e-06, "loss": 3.4163, "step": 3137 }, { "epoch": 0.38, "learning_rate": 7.451786296233383e-06, "loss": 3.4515, "step": 3138 }, { "epoch": 0.38, "learning_rate": 7.450037940638166e-06, "loss": 3.3977, "step": 3139 }, { "epoch": 0.38, "learning_rate": 7.448289190732667e-06, "loss": 3.4557, "step": 3140 }, { "epoch": 0.38, "learning_rate": 7.446540046798331e-06, "loss": 3.5091, "step": 3141 }, { "epoch": 0.38, "learning_rate": 7.444790509116666e-06, "loss": 3.5295, "step": 3142 }, { "epoch": 0.38, "learning_rate": 7.443040577969245e-06, "loss": 3.363, "step": 3143 }, { "epoch": 0.38, "learning_rate": 7.441290253637699e-06, "loss": 3.3902, "step": 3144 }, { "epoch": 0.38, "learning_rate": 7.4395395364037306e-06, "loss": 3.4749, "step": 3145 }, { "epoch": 0.38, "learning_rate": 7.437788426549097e-06, "loss": 3.4408, "step": 3146 }, { "epoch": 0.38, "learning_rate": 7.436036924355625e-06, "loss": 3.4711, "step": 3147 }, { "epoch": 0.38, "learning_rate": 7.434285030105203e-06, "loss": 3.4907, "step": 3148 }, { "epoch": 0.38, "learning_rate": 7.432532744079779e-06, "loss": 3.3983, "step": 3149 }, { "epoch": 0.38, "learning_rate": 7.430780066561368e-06, "loss": 3.4074, "step": 3150 }, { "epoch": 0.38, "learning_rate": 7.429026997832047e-06, "loss": 3.2792, "step": 3151 }, { "epoch": 0.38, "learning_rate": 7.427273538173957e-06, "loss": 3.429, "step": 3152 }, { "epoch": 0.38, "learning_rate": 7.425519687869298e-06, "loss": 3.4043, "step": 3153 }, { "epoch": 0.38, "learning_rate": 7.4237654472003365e-06, "loss": 3.4187, "step": 3154 }, { "epoch": 0.38, "learning_rate": 7.422010816449401e-06, "loss": 3.4286, "step": 3155 }, { "epoch": 0.38, "learning_rate": 7.420255795898883e-06, "loss": 3.5089, "step": 3156 }, { "epoch": 0.38, "learning_rate": 7.418500385831237e-06, "loss": 3.432, "step": 3157 }, { "epoch": 0.38, "learning_rate": 7.4167445865289765e-06, "loss": 3.4396, "step": 3158 }, { "epoch": 0.38, "learning_rate": 7.414988398274684e-06, "loss": 3.4214, "step": 3159 }, { "epoch": 0.38, "learning_rate": 7.413231821351e-06, "loss": 3.4018, "step": 3160 }, { "epoch": 0.38, "learning_rate": 7.41147485604063e-06, "loss": 3.4742, "step": 3161 }, { "epoch": 0.38, "learning_rate": 7.409717502626339e-06, "loss": 3.3248, "step": 3162 }, { "epoch": 0.38, "learning_rate": 7.407959761390956e-06, "loss": 3.397, "step": 3163 }, { "epoch": 0.38, "learning_rate": 7.406201632617375e-06, "loss": 3.405, "step": 3164 }, { "epoch": 0.38, "learning_rate": 7.404443116588548e-06, "loss": 3.4149, "step": 3165 }, { "epoch": 0.38, "learning_rate": 7.402684213587493e-06, "loss": 3.3544, "step": 3166 }, { "epoch": 0.38, "learning_rate": 7.400924923897288e-06, "loss": 3.4276, "step": 3167 }, { "epoch": 0.38, "learning_rate": 7.399165247801072e-06, "loss": 3.4992, "step": 3168 }, { "epoch": 0.38, "learning_rate": 7.3974051855820516e-06, "loss": 3.3675, "step": 3169 }, { "epoch": 0.38, "learning_rate": 7.395644737523492e-06, "loss": 3.4188, "step": 3170 }, { "epoch": 0.38, "learning_rate": 7.393883903908717e-06, "loss": 3.4521, "step": 3171 }, { "epoch": 0.38, "learning_rate": 7.392122685021118e-06, "loss": 3.3669, "step": 3172 }, { "epoch": 0.38, "learning_rate": 7.3903610811441465e-06, "loss": 3.437, "step": 3173 }, { "epoch": 0.38, "learning_rate": 7.388599092561315e-06, "loss": 3.4023, "step": 3174 }, { "epoch": 0.38, "learning_rate": 7.386836719556201e-06, "loss": 3.4079, "step": 3175 }, { "epoch": 0.38, "learning_rate": 7.385073962412439e-06, "loss": 3.3738, "step": 3176 }, { "epoch": 0.38, "learning_rate": 7.383310821413731e-06, "loss": 3.2764, "step": 3177 }, { "epoch": 0.38, "learning_rate": 7.381547296843834e-06, "loss": 3.3943, "step": 3178 }, { "epoch": 0.38, "learning_rate": 7.379783388986574e-06, "loss": 3.4532, "step": 3179 }, { "epoch": 0.38, "learning_rate": 7.378019098125833e-06, "loss": 3.4533, "step": 3180 }, { "epoch": 0.38, "learning_rate": 7.376254424545557e-06, "loss": 3.4281, "step": 3181 }, { "epoch": 0.38, "learning_rate": 7.3744893685297546e-06, "loss": 3.4148, "step": 3182 }, { "epoch": 0.38, "learning_rate": 7.372723930362493e-06, "loss": 3.4371, "step": 3183 }, { "epoch": 0.38, "learning_rate": 7.370958110327906e-06, "loss": 3.4375, "step": 3184 }, { "epoch": 0.38, "learning_rate": 7.369191908710182e-06, "loss": 3.4744, "step": 3185 }, { "epoch": 0.38, "learning_rate": 7.367425325793574e-06, "loss": 3.4137, "step": 3186 }, { "epoch": 0.38, "learning_rate": 7.3656583618624e-06, "loss": 3.3845, "step": 3187 }, { "epoch": 0.38, "learning_rate": 7.363891017201035e-06, "loss": 3.4883, "step": 3188 }, { "epoch": 0.38, "learning_rate": 7.362123292093914e-06, "loss": 3.5088, "step": 3189 }, { "epoch": 0.38, "learning_rate": 7.360355186825537e-06, "loss": 3.4428, "step": 3190 }, { "epoch": 0.38, "learning_rate": 7.358586701680463e-06, "loss": 3.4895, "step": 3191 }, { "epoch": 0.38, "learning_rate": 7.3568178369433155e-06, "loss": 3.4736, "step": 3192 }, { "epoch": 0.38, "learning_rate": 7.355048592898775e-06, "loss": 3.5338, "step": 3193 }, { "epoch": 0.38, "learning_rate": 7.353278969831583e-06, "loss": 3.4552, "step": 3194 }, { "epoch": 0.38, "learning_rate": 7.351508968026544e-06, "loss": 3.4271, "step": 3195 }, { "epoch": 0.38, "learning_rate": 7.3497385877685244e-06, "loss": 3.345, "step": 3196 }, { "epoch": 0.38, "learning_rate": 7.347967829342449e-06, "loss": 3.4985, "step": 3197 }, { "epoch": 0.38, "learning_rate": 7.346196693033305e-06, "loss": 3.4701, "step": 3198 }, { "epoch": 0.38, "learning_rate": 7.3444251791261405e-06, "loss": 3.4483, "step": 3199 }, { "epoch": 0.38, "learning_rate": 7.342653287906061e-06, "loss": 3.35, "step": 3200 }, { "epoch": 0.38, "learning_rate": 7.340881019658237e-06, "loss": 3.4848, "step": 3201 }, { "epoch": 0.38, "learning_rate": 7.3391083746679006e-06, "loss": 3.3762, "step": 3202 }, { "epoch": 0.38, "learning_rate": 7.337335353220339e-06, "loss": 3.4813, "step": 3203 }, { "epoch": 0.38, "learning_rate": 7.335561955600905e-06, "loss": 3.3728, "step": 3204 }, { "epoch": 0.38, "learning_rate": 7.333788182095009e-06, "loss": 3.4698, "step": 3205 }, { "epoch": 0.38, "learning_rate": 7.332014032988123e-06, "loss": 3.3642, "step": 3206 }, { "epoch": 0.38, "learning_rate": 7.330239508565779e-06, "loss": 3.3959, "step": 3207 }, { "epoch": 0.39, "learning_rate": 7.328464609113573e-06, "loss": 3.4482, "step": 3208 }, { "epoch": 0.39, "learning_rate": 7.326689334917153e-06, "loss": 3.4687, "step": 3209 }, { "epoch": 0.39, "learning_rate": 7.324913686262236e-06, "loss": 3.4229, "step": 3210 }, { "epoch": 0.39, "learning_rate": 7.323137663434597e-06, "loss": 3.4213, "step": 3211 }, { "epoch": 0.39, "learning_rate": 7.321361266720065e-06, "loss": 3.3656, "step": 3212 }, { "epoch": 0.39, "learning_rate": 7.3195844964045386e-06, "loss": 3.3698, "step": 3213 }, { "epoch": 0.39, "learning_rate": 7.317807352773969e-06, "loss": 3.4689, "step": 3214 }, { "epoch": 0.39, "learning_rate": 7.316029836114374e-06, "loss": 3.4291, "step": 3215 }, { "epoch": 0.39, "learning_rate": 7.314251946711826e-06, "loss": 3.3498, "step": 3216 }, { "epoch": 0.39, "learning_rate": 7.312473684852457e-06, "loss": 3.4814, "step": 3217 }, { "epoch": 0.39, "learning_rate": 7.310695050822465e-06, "loss": 3.423, "step": 3218 }, { "epoch": 0.39, "learning_rate": 7.308916044908103e-06, "loss": 3.4353, "step": 3219 }, { "epoch": 0.39, "learning_rate": 7.307136667395686e-06, "loss": 3.4041, "step": 3220 }, { "epoch": 0.39, "learning_rate": 7.3053569185715865e-06, "loss": 3.3692, "step": 3221 }, { "epoch": 0.39, "learning_rate": 7.3035767987222385e-06, "loss": 3.3978, "step": 3222 }, { "epoch": 0.39, "learning_rate": 7.3017963081341335e-06, "loss": 3.5039, "step": 3223 }, { "epoch": 0.39, "learning_rate": 7.300015447093828e-06, "loss": 3.4185, "step": 3224 }, { "epoch": 0.39, "learning_rate": 7.298234215887933e-06, "loss": 3.5426, "step": 3225 }, { "epoch": 0.39, "learning_rate": 7.29645261480312e-06, "loss": 3.2663, "step": 3226 }, { "epoch": 0.39, "learning_rate": 7.294670644126122e-06, "loss": 3.3821, "step": 3227 }, { "epoch": 0.39, "learning_rate": 7.292888304143728e-06, "loss": 3.452, "step": 3228 }, { "epoch": 0.39, "learning_rate": 7.291105595142793e-06, "loss": 3.3389, "step": 3229 }, { "epoch": 0.39, "learning_rate": 7.289322517410223e-06, "loss": 3.45, "step": 3230 }, { "epoch": 0.39, "learning_rate": 7.287539071232988e-06, "loss": 3.45, "step": 3231 }, { "epoch": 0.39, "learning_rate": 7.2857552568981184e-06, "loss": 3.3317, "step": 3232 }, { "epoch": 0.39, "learning_rate": 7.283971074692702e-06, "loss": 3.4473, "step": 3233 }, { "epoch": 0.39, "learning_rate": 7.282186524903883e-06, "loss": 3.4345, "step": 3234 }, { "epoch": 0.39, "learning_rate": 7.280401607818871e-06, "loss": 3.4558, "step": 3235 }, { "epoch": 0.39, "learning_rate": 7.27861632372493e-06, "loss": 3.3619, "step": 3236 }, { "epoch": 0.39, "learning_rate": 7.276830672909384e-06, "loss": 3.3727, "step": 3237 }, { "epoch": 0.39, "learning_rate": 7.275044655659619e-06, "loss": 3.5442, "step": 3238 }, { "epoch": 0.39, "learning_rate": 7.273258272263073e-06, "loss": 3.45, "step": 3239 }, { "epoch": 0.39, "learning_rate": 7.2714715230072505e-06, "loss": 3.3944, "step": 3240 }, { "epoch": 0.39, "learning_rate": 7.269684408179713e-06, "loss": 3.4064, "step": 3241 }, { "epoch": 0.39, "learning_rate": 7.267896928068078e-06, "loss": 3.3374, "step": 3242 }, { "epoch": 0.39, "learning_rate": 7.266109082960022e-06, "loss": 3.4902, "step": 3243 }, { "epoch": 0.39, "learning_rate": 7.264320873143283e-06, "loss": 3.5286, "step": 3244 }, { "epoch": 0.39, "learning_rate": 7.2625322989056565e-06, "loss": 3.3952, "step": 3245 }, { "epoch": 0.39, "learning_rate": 7.2607433605349965e-06, "loss": 3.3955, "step": 3246 }, { "epoch": 0.39, "learning_rate": 7.2589540583192165e-06, "loss": 3.3606, "step": 3247 }, { "epoch": 0.39, "learning_rate": 7.257164392546286e-06, "loss": 3.415, "step": 3248 }, { "epoch": 0.39, "learning_rate": 7.255374363504234e-06, "loss": 3.4219, "step": 3249 }, { "epoch": 0.39, "learning_rate": 7.253583971481151e-06, "loss": 3.4416, "step": 3250 }, { "epoch": 0.39, "learning_rate": 7.251793216765183e-06, "loss": 3.4233, "step": 3251 }, { "epoch": 0.39, "learning_rate": 7.250002099644533e-06, "loss": 3.335, "step": 3252 }, { "epoch": 0.39, "learning_rate": 7.248210620407466e-06, "loss": 3.3331, "step": 3253 }, { "epoch": 0.39, "learning_rate": 7.246418779342303e-06, "loss": 3.5127, "step": 3254 }, { "epoch": 0.39, "learning_rate": 7.244626576737424e-06, "loss": 3.3968, "step": 3255 }, { "epoch": 0.39, "learning_rate": 7.242834012881268e-06, "loss": 3.4055, "step": 3256 }, { "epoch": 0.39, "learning_rate": 7.241041088062329e-06, "loss": 3.4855, "step": 3257 }, { "epoch": 0.39, "learning_rate": 7.239247802569162e-06, "loss": 3.4872, "step": 3258 }, { "epoch": 0.39, "learning_rate": 7.2374541566903775e-06, "loss": 3.4612, "step": 3259 }, { "epoch": 0.39, "learning_rate": 7.2356601507146515e-06, "loss": 3.4542, "step": 3260 }, { "epoch": 0.39, "learning_rate": 7.233865784930706e-06, "loss": 3.4048, "step": 3261 }, { "epoch": 0.39, "learning_rate": 7.2320710596273295e-06, "loss": 3.4621, "step": 3262 }, { "epoch": 0.39, "learning_rate": 7.230275975093366e-06, "loss": 3.452, "step": 3263 }, { "epoch": 0.39, "learning_rate": 7.228480531617717e-06, "loss": 3.4571, "step": 3264 }, { "epoch": 0.39, "learning_rate": 7.226684729489342e-06, "loss": 3.4526, "step": 3265 }, { "epoch": 0.39, "learning_rate": 7.224888568997258e-06, "loss": 3.3333, "step": 3266 }, { "epoch": 0.39, "learning_rate": 7.223092050430541e-06, "loss": 3.3703, "step": 3267 }, { "epoch": 0.39, "learning_rate": 7.221295174078322e-06, "loss": 3.4263, "step": 3268 }, { "epoch": 0.39, "learning_rate": 7.219497940229794e-06, "loss": 3.3285, "step": 3269 }, { "epoch": 0.39, "learning_rate": 7.217700349174201e-06, "loss": 3.4211, "step": 3270 }, { "epoch": 0.39, "learning_rate": 7.21590240120085e-06, "loss": 3.3584, "step": 3271 }, { "epoch": 0.39, "learning_rate": 7.214104096599103e-06, "loss": 3.3638, "step": 3272 }, { "epoch": 0.39, "learning_rate": 7.21230543565838e-06, "loss": 3.3239, "step": 3273 }, { "epoch": 0.39, "learning_rate": 7.210506418668159e-06, "loss": 3.4797, "step": 3274 }, { "epoch": 0.39, "learning_rate": 7.208707045917976e-06, "loss": 3.344, "step": 3275 }, { "epoch": 0.39, "learning_rate": 7.20690731769742e-06, "loss": 3.3616, "step": 3276 }, { "epoch": 0.39, "learning_rate": 7.205107234296143e-06, "loss": 3.4262, "step": 3277 }, { "epoch": 0.39, "learning_rate": 7.203306796003849e-06, "loss": 3.4063, "step": 3278 }, { "epoch": 0.39, "learning_rate": 7.201506003110301e-06, "loss": 3.471, "step": 3279 }, { "epoch": 0.39, "learning_rate": 7.199704855905322e-06, "loss": 3.2901, "step": 3280 }, { "epoch": 0.39, "learning_rate": 7.197903354678788e-06, "loss": 3.428, "step": 3281 }, { "epoch": 0.39, "learning_rate": 7.196101499720634e-06, "loss": 3.3962, "step": 3282 }, { "epoch": 0.39, "learning_rate": 7.19429929132085e-06, "loss": 3.3292, "step": 3283 }, { "epoch": 0.39, "learning_rate": 7.192496729769486e-06, "loss": 3.3405, "step": 3284 }, { "epoch": 0.39, "learning_rate": 7.190693815356646e-06, "loss": 3.3949, "step": 3285 }, { "epoch": 0.39, "learning_rate": 7.188890548372493e-06, "loss": 3.4154, "step": 3286 }, { "epoch": 0.39, "learning_rate": 7.187086929107243e-06, "loss": 3.3573, "step": 3287 }, { "epoch": 0.39, "learning_rate": 7.185282957851175e-06, "loss": 3.4963, "step": 3288 }, { "epoch": 0.39, "learning_rate": 7.1834786348946185e-06, "loss": 3.4281, "step": 3289 }, { "epoch": 0.39, "learning_rate": 7.181673960527964e-06, "loss": 3.4175, "step": 3290 }, { "epoch": 0.4, "learning_rate": 7.179868935041655e-06, "loss": 3.4315, "step": 3291 }, { "epoch": 0.4, "learning_rate": 7.1780635587261945e-06, "loss": 3.4583, "step": 3292 }, { "epoch": 0.4, "learning_rate": 7.17625783187214e-06, "loss": 3.4033, "step": 3293 }, { "epoch": 0.4, "learning_rate": 7.174451754770104e-06, "loss": 3.4504, "step": 3294 }, { "epoch": 0.4, "learning_rate": 7.172645327710761e-06, "loss": 3.3486, "step": 3295 }, { "epoch": 0.4, "learning_rate": 7.170838550984838e-06, "loss": 3.4644, "step": 3296 }, { "epoch": 0.4, "learning_rate": 7.169031424883115e-06, "loss": 3.4396, "step": 3297 }, { "epoch": 0.4, "learning_rate": 7.167223949696433e-06, "loss": 3.3027, "step": 3298 }, { "epoch": 0.4, "learning_rate": 7.16541612571569e-06, "loss": 3.4961, "step": 3299 }, { "epoch": 0.4, "learning_rate": 7.163607953231837e-06, "loss": 3.3411, "step": 3300 }, { "epoch": 0.4, "learning_rate": 7.161799432535883e-06, "loss": 3.3595, "step": 3301 }, { "epoch": 0.4, "learning_rate": 7.159990563918889e-06, "loss": 3.473, "step": 3302 }, { "epoch": 0.4, "learning_rate": 7.158181347671979e-06, "loss": 3.4424, "step": 3303 }, { "epoch": 0.4, "learning_rate": 7.156371784086327e-06, "loss": 3.4258, "step": 3304 }, { "epoch": 0.4, "learning_rate": 7.154561873453165e-06, "loss": 3.3123, "step": 3305 }, { "epoch": 0.4, "learning_rate": 7.152751616063781e-06, "loss": 3.3878, "step": 3306 }, { "epoch": 0.4, "learning_rate": 7.15094101220952e-06, "loss": 3.5016, "step": 3307 }, { "epoch": 0.4, "learning_rate": 7.149130062181781e-06, "loss": 3.3191, "step": 3308 }, { "epoch": 0.4, "learning_rate": 7.147318766272018e-06, "loss": 3.4435, "step": 3309 }, { "epoch": 0.4, "learning_rate": 7.145507124771742e-06, "loss": 3.4399, "step": 3310 }, { "epoch": 0.4, "learning_rate": 7.143695137972522e-06, "loss": 3.347, "step": 3311 }, { "epoch": 0.4, "learning_rate": 7.141882806165976e-06, "loss": 3.3262, "step": 3312 }, { "epoch": 0.4, "learning_rate": 7.140070129643783e-06, "loss": 3.4316, "step": 3313 }, { "epoch": 0.4, "learning_rate": 7.13825710869768e-06, "loss": 3.3356, "step": 3314 }, { "epoch": 0.4, "learning_rate": 7.13644374361945e-06, "loss": 3.3701, "step": 3315 }, { "epoch": 0.4, "learning_rate": 7.134630034700938e-06, "loss": 3.3802, "step": 3316 }, { "epoch": 0.4, "learning_rate": 7.132815982234045e-06, "loss": 3.4839, "step": 3317 }, { "epoch": 0.4, "learning_rate": 7.131001586510724e-06, "loss": 3.3113, "step": 3318 }, { "epoch": 0.4, "learning_rate": 7.129186847822986e-06, "loss": 3.4962, "step": 3319 }, { "epoch": 0.4, "learning_rate": 7.127371766462896e-06, "loss": 3.3639, "step": 3320 }, { "epoch": 0.4, "learning_rate": 7.125556342722572e-06, "loss": 3.4295, "step": 3321 }, { "epoch": 0.4, "learning_rate": 7.1237405768941905e-06, "loss": 3.37, "step": 3322 }, { "epoch": 0.4, "learning_rate": 7.121924469269982e-06, "loss": 3.3366, "step": 3323 }, { "epoch": 0.4, "learning_rate": 7.120108020142231e-06, "loss": 3.3992, "step": 3324 }, { "epoch": 0.4, "learning_rate": 7.118291229803278e-06, "loss": 3.4521, "step": 3325 }, { "epoch": 0.4, "learning_rate": 7.116474098545518e-06, "loss": 3.4544, "step": 3326 }, { "epoch": 0.4, "learning_rate": 7.114656626661403e-06, "loss": 3.526, "step": 3327 }, { "epoch": 0.4, "learning_rate": 7.112838814443435e-06, "loss": 3.4203, "step": 3328 }, { "epoch": 0.4, "learning_rate": 7.111020662184174e-06, "loss": 3.5157, "step": 3329 }, { "epoch": 0.4, "learning_rate": 7.109202170176235e-06, "loss": 3.3921, "step": 3330 }, { "epoch": 0.4, "learning_rate": 7.1073833387122855e-06, "loss": 3.4613, "step": 3331 }, { "epoch": 0.4, "learning_rate": 7.105564168085053e-06, "loss": 3.4462, "step": 3332 }, { "epoch": 0.4, "learning_rate": 7.103744658587312e-06, "loss": 3.3924, "step": 3333 }, { "epoch": 0.4, "learning_rate": 7.101924810511896e-06, "loss": 3.4014, "step": 3334 }, { "epoch": 0.4, "learning_rate": 7.1001046241516924e-06, "loss": 3.4065, "step": 3335 }, { "epoch": 0.4, "learning_rate": 7.0982840997996435e-06, "loss": 3.4771, "step": 3336 }, { "epoch": 0.4, "learning_rate": 7.096463237748745e-06, "loss": 3.4143, "step": 3337 }, { "epoch": 0.4, "learning_rate": 7.094642038292047e-06, "loss": 3.3882, "step": 3338 }, { "epoch": 0.4, "learning_rate": 7.092820501722654e-06, "loss": 3.4129, "step": 3339 }, { "epoch": 0.4, "learning_rate": 7.090998628333725e-06, "loss": 3.3788, "step": 3340 }, { "epoch": 0.4, "learning_rate": 7.089176418418475e-06, "loss": 3.4794, "step": 3341 }, { "epoch": 0.4, "learning_rate": 7.087353872270169e-06, "loss": 3.4061, "step": 3342 }, { "epoch": 0.4, "learning_rate": 7.08553099018213e-06, "loss": 3.3132, "step": 3343 }, { "epoch": 0.4, "learning_rate": 7.083707772447732e-06, "loss": 3.4151, "step": 3344 }, { "epoch": 0.4, "learning_rate": 7.081884219360406e-06, "loss": 3.4276, "step": 3345 }, { "epoch": 0.4, "learning_rate": 7.0800603312136365e-06, "loss": 3.4211, "step": 3346 }, { "epoch": 0.4, "learning_rate": 7.078236108300958e-06, "loss": 3.4406, "step": 3347 }, { "epoch": 0.4, "learning_rate": 7.076411550915962e-06, "loss": 3.3847, "step": 3348 }, { "epoch": 0.4, "learning_rate": 7.074586659352296e-06, "loss": 3.4504, "step": 3349 }, { "epoch": 0.4, "learning_rate": 7.0727614339036576e-06, "loss": 3.3935, "step": 3350 }, { "epoch": 0.4, "learning_rate": 7.0709358748638e-06, "loss": 3.3787, "step": 3351 }, { "epoch": 0.4, "learning_rate": 7.069109982526527e-06, "loss": 3.4734, "step": 3352 }, { "epoch": 0.4, "learning_rate": 7.067283757185703e-06, "loss": 3.2762, "step": 3353 }, { "epoch": 0.4, "learning_rate": 7.065457199135238e-06, "loss": 3.4046, "step": 3354 }, { "epoch": 0.4, "learning_rate": 7.063630308669102e-06, "loss": 3.4833, "step": 3355 }, { "epoch": 0.4, "learning_rate": 7.061803086081313e-06, "loss": 3.3527, "step": 3356 }, { "epoch": 0.4, "learning_rate": 7.059975531665945e-06, "loss": 3.3897, "step": 3357 }, { "epoch": 0.4, "learning_rate": 7.058147645717128e-06, "loss": 3.4021, "step": 3358 }, { "epoch": 0.4, "learning_rate": 7.056319428529043e-06, "loss": 3.402, "step": 3359 }, { "epoch": 0.4, "learning_rate": 7.05449088039592e-06, "loss": 3.4924, "step": 3360 }, { "epoch": 0.4, "learning_rate": 7.05266200161205e-06, "loss": 3.3742, "step": 3361 }, { "epoch": 0.4, "learning_rate": 7.050832792471773e-06, "loss": 3.4894, "step": 3362 }, { "epoch": 0.4, "learning_rate": 7.049003253269482e-06, "loss": 3.5166, "step": 3363 }, { "epoch": 0.4, "learning_rate": 7.047173384299624e-06, "loss": 3.4421, "step": 3364 }, { "epoch": 0.4, "learning_rate": 7.045343185856701e-06, "loss": 3.3445, "step": 3365 }, { "epoch": 0.4, "learning_rate": 7.043512658235263e-06, "loss": 3.4476, "step": 3366 }, { "epoch": 0.4, "learning_rate": 7.041681801729917e-06, "loss": 3.4181, "step": 3367 }, { "epoch": 0.4, "learning_rate": 7.039850616635324e-06, "loss": 3.4609, "step": 3368 }, { "epoch": 0.4, "learning_rate": 7.038019103246193e-06, "loss": 3.3862, "step": 3369 }, { "epoch": 0.4, "learning_rate": 7.036187261857289e-06, "loss": 3.4944, "step": 3370 }, { "epoch": 0.4, "learning_rate": 7.034355092763432e-06, "loss": 3.313, "step": 3371 }, { "epoch": 0.4, "learning_rate": 7.032522596259488e-06, "loss": 3.4633, "step": 3372 }, { "epoch": 0.4, "learning_rate": 7.030689772640384e-06, "loss": 3.507, "step": 3373 }, { "epoch": 0.4, "learning_rate": 7.0288566222010925e-06, "loss": 3.287, "step": 3374 }, { "epoch": 0.41, "learning_rate": 7.027023145236644e-06, "loss": 3.4709, "step": 3375 }, { "epoch": 0.41, "learning_rate": 7.025189342042118e-06, "loss": 3.3987, "step": 3376 }, { "epoch": 0.41, "learning_rate": 7.0233552129126485e-06, "loss": 3.4738, "step": 3377 }, { "epoch": 0.41, "learning_rate": 7.0215207581434185e-06, "loss": 3.4635, "step": 3378 }, { "epoch": 0.41, "learning_rate": 7.019685978029668e-06, "loss": 3.4322, "step": 3379 }, { "epoch": 0.41, "learning_rate": 7.017850872866687e-06, "loss": 3.5671, "step": 3380 }, { "epoch": 0.41, "learning_rate": 7.01601544294982e-06, "loss": 3.5214, "step": 3381 }, { "epoch": 0.41, "learning_rate": 7.014179688574459e-06, "loss": 3.391, "step": 3382 }, { "epoch": 0.41, "learning_rate": 7.012343610036054e-06, "loss": 3.3413, "step": 3383 }, { "epoch": 0.41, "learning_rate": 7.0105072076301015e-06, "loss": 3.3443, "step": 3384 }, { "epoch": 0.41, "learning_rate": 7.0086704816521555e-06, "loss": 3.4253, "step": 3385 }, { "epoch": 0.41, "learning_rate": 7.006833432397817e-06, "loss": 3.3947, "step": 3386 }, { "epoch": 0.41, "learning_rate": 7.004996060162744e-06, "loss": 3.5134, "step": 3387 }, { "epoch": 0.41, "learning_rate": 7.003158365242644e-06, "loss": 3.4148, "step": 3388 }, { "epoch": 0.41, "learning_rate": 7.001320347933274e-06, "loss": 3.3681, "step": 3389 }, { "epoch": 0.41, "learning_rate": 6.999482008530447e-06, "loss": 3.419, "step": 3390 }, { "epoch": 0.41, "learning_rate": 6.997643347330029e-06, "loss": 3.3913, "step": 3391 }, { "epoch": 0.41, "learning_rate": 6.995804364627929e-06, "loss": 3.3618, "step": 3392 }, { "epoch": 0.41, "learning_rate": 6.99396506072012e-06, "loss": 3.3946, "step": 3393 }, { "epoch": 0.41, "learning_rate": 6.992125435902615e-06, "loss": 3.4907, "step": 3394 }, { "epoch": 0.41, "learning_rate": 6.990285490471489e-06, "loss": 3.4896, "step": 3395 }, { "epoch": 0.41, "learning_rate": 6.988445224722859e-06, "loss": 3.5047, "step": 3396 }, { "epoch": 0.41, "learning_rate": 6.986604638952902e-06, "loss": 3.3985, "step": 3397 }, { "epoch": 0.41, "learning_rate": 6.98476373345784e-06, "loss": 3.4126, "step": 3398 }, { "epoch": 0.41, "learning_rate": 6.982922508533951e-06, "loss": 3.3869, "step": 3399 }, { "epoch": 0.41, "learning_rate": 6.981080964477563e-06, "loss": 3.4474, "step": 3400 }, { "epoch": 0.41, "learning_rate": 6.979239101585054e-06, "loss": 3.4722, "step": 3401 }, { "epoch": 0.41, "learning_rate": 6.977396920152853e-06, "loss": 3.395, "step": 3402 }, { "epoch": 0.41, "learning_rate": 6.975554420477444e-06, "loss": 3.4029, "step": 3403 }, { "epoch": 0.41, "learning_rate": 6.973711602855359e-06, "loss": 3.4604, "step": 3404 }, { "epoch": 0.41, "learning_rate": 6.971868467583181e-06, "loss": 3.3194, "step": 3405 }, { "epoch": 0.41, "learning_rate": 6.970025014957547e-06, "loss": 3.4754, "step": 3406 }, { "epoch": 0.41, "learning_rate": 6.968181245275139e-06, "loss": 3.393, "step": 3407 }, { "epoch": 0.41, "learning_rate": 6.9663371588327e-06, "loss": 3.5288, "step": 3408 }, { "epoch": 0.41, "learning_rate": 6.964492755927015e-06, "loss": 3.4506, "step": 3409 }, { "epoch": 0.41, "learning_rate": 6.962648036854923e-06, "loss": 3.3964, "step": 3410 }, { "epoch": 0.41, "learning_rate": 6.960803001913315e-06, "loss": 3.3687, "step": 3411 }, { "epoch": 0.41, "learning_rate": 6.95895765139913e-06, "loss": 3.4152, "step": 3412 }, { "epoch": 0.41, "learning_rate": 6.957111985609362e-06, "loss": 3.497, "step": 3413 }, { "epoch": 0.41, "learning_rate": 6.95526600484105e-06, "loss": 3.4225, "step": 3414 }, { "epoch": 0.41, "learning_rate": 6.953419709391291e-06, "loss": 3.3305, "step": 3415 }, { "epoch": 0.41, "learning_rate": 6.951573099557225e-06, "loss": 3.3642, "step": 3416 }, { "epoch": 0.41, "learning_rate": 6.949726175636049e-06, "loss": 3.3855, "step": 3417 }, { "epoch": 0.41, "learning_rate": 6.9478789379250076e-06, "loss": 3.4445, "step": 3418 }, { "epoch": 0.41, "learning_rate": 6.946031386721396e-06, "loss": 3.4954, "step": 3419 }, { "epoch": 0.41, "learning_rate": 6.9441835223225574e-06, "loss": 3.4296, "step": 3420 }, { "epoch": 0.41, "learning_rate": 6.9423353450258915e-06, "loss": 3.4163, "step": 3421 }, { "epoch": 0.41, "learning_rate": 6.9404868551288415e-06, "loss": 3.4177, "step": 3422 }, { "epoch": 0.41, "learning_rate": 6.9386380529289075e-06, "loss": 3.4569, "step": 3423 }, { "epoch": 0.41, "learning_rate": 6.936788938723634e-06, "loss": 3.4812, "step": 3424 }, { "epoch": 0.41, "learning_rate": 6.934939512810621e-06, "loss": 3.3648, "step": 3425 }, { "epoch": 0.41, "learning_rate": 6.933089775487514e-06, "loss": 3.4687, "step": 3426 }, { "epoch": 0.41, "learning_rate": 6.931239727052012e-06, "loss": 3.4123, "step": 3427 }, { "epoch": 0.41, "learning_rate": 6.92938936780186e-06, "loss": 3.345, "step": 3428 }, { "epoch": 0.41, "learning_rate": 6.927538698034858e-06, "loss": 3.3167, "step": 3429 }, { "epoch": 0.41, "learning_rate": 6.925687718048853e-06, "loss": 3.3909, "step": 3430 }, { "epoch": 0.41, "learning_rate": 6.9238364281417434e-06, "loss": 3.3687, "step": 3431 }, { "epoch": 0.41, "learning_rate": 6.921984828611477e-06, "loss": 3.3707, "step": 3432 }, { "epoch": 0.41, "learning_rate": 6.920132919756049e-06, "loss": 3.449, "step": 3433 }, { "epoch": 0.41, "learning_rate": 6.918280701873507e-06, "loss": 3.3735, "step": 3434 }, { "epoch": 0.41, "learning_rate": 6.916428175261949e-06, "loss": 3.354, "step": 3435 }, { "epoch": 0.41, "learning_rate": 6.91457534021952e-06, "loss": 3.3521, "step": 3436 }, { "epoch": 0.41, "learning_rate": 6.912722197044419e-06, "loss": 3.4094, "step": 3437 }, { "epoch": 0.41, "learning_rate": 6.9108687460348885e-06, "loss": 3.4211, "step": 3438 }, { "epoch": 0.41, "learning_rate": 6.909014987489226e-06, "loss": 3.3688, "step": 3439 }, { "epoch": 0.41, "learning_rate": 6.907160921705776e-06, "loss": 3.4913, "step": 3440 }, { "epoch": 0.41, "learning_rate": 6.905306548982933e-06, "loss": 3.5193, "step": 3441 }, { "epoch": 0.41, "learning_rate": 6.903451869619139e-06, "loss": 3.5473, "step": 3442 }, { "epoch": 0.41, "learning_rate": 6.901596883912888e-06, "loss": 3.4403, "step": 3443 }, { "epoch": 0.41, "learning_rate": 6.899741592162721e-06, "loss": 3.443, "step": 3444 }, { "epoch": 0.41, "learning_rate": 6.897885994667233e-06, "loss": 3.3625, "step": 3445 }, { "epoch": 0.41, "learning_rate": 6.896030091725061e-06, "loss": 3.3553, "step": 3446 }, { "epoch": 0.41, "learning_rate": 6.894173883634896e-06, "loss": 3.4935, "step": 3447 }, { "epoch": 0.41, "learning_rate": 6.892317370695479e-06, "loss": 3.4307, "step": 3448 }, { "epoch": 0.41, "learning_rate": 6.890460553205593e-06, "loss": 3.5105, "step": 3449 }, { "epoch": 0.41, "learning_rate": 6.888603431464082e-06, "loss": 3.3918, "step": 3450 }, { "epoch": 0.41, "learning_rate": 6.886746005769829e-06, "loss": 3.3942, "step": 3451 }, { "epoch": 0.41, "learning_rate": 6.884888276421766e-06, "loss": 3.3591, "step": 3452 }, { "epoch": 0.41, "learning_rate": 6.883030243718881e-06, "loss": 3.409, "step": 3453 }, { "epoch": 0.41, "learning_rate": 6.881171907960204e-06, "loss": 3.4488, "step": 3454 }, { "epoch": 0.41, "learning_rate": 6.8793132694448195e-06, "loss": 3.3691, "step": 3455 }, { "epoch": 0.41, "learning_rate": 6.877454328471855e-06, "loss": 3.3811, "step": 3456 }, { "epoch": 0.41, "learning_rate": 6.875595085340489e-06, "loss": 3.4359, "step": 3457 }, { "epoch": 0.42, "learning_rate": 6.873735540349951e-06, "loss": 3.4775, "step": 3458 }, { "epoch": 0.42, "learning_rate": 6.871875693799518e-06, "loss": 3.44, "step": 3459 }, { "epoch": 0.42, "learning_rate": 6.8700155459885106e-06, "loss": 3.4272, "step": 3460 }, { "epoch": 0.42, "learning_rate": 6.868155097216305e-06, "loss": 3.3675, "step": 3461 }, { "epoch": 0.42, "learning_rate": 6.866294347782321e-06, "loss": 3.4002, "step": 3462 }, { "epoch": 0.42, "learning_rate": 6.864433297986031e-06, "loss": 3.3977, "step": 3463 }, { "epoch": 0.42, "learning_rate": 6.8625719481269495e-06, "loss": 3.441, "step": 3464 }, { "epoch": 0.42, "learning_rate": 6.8607102985046445e-06, "loss": 3.4359, "step": 3465 }, { "epoch": 0.42, "learning_rate": 6.858848349418733e-06, "loss": 3.4557, "step": 3466 }, { "epoch": 0.42, "learning_rate": 6.8569861011688755e-06, "loss": 3.3903, "step": 3467 }, { "epoch": 0.42, "learning_rate": 6.855123554054786e-06, "loss": 3.3893, "step": 3468 }, { "epoch": 0.42, "learning_rate": 6.853260708376219e-06, "loss": 3.4859, "step": 3469 }, { "epoch": 0.42, "learning_rate": 6.851397564432986e-06, "loss": 3.458, "step": 3470 }, { "epoch": 0.42, "learning_rate": 6.849534122524939e-06, "loss": 3.3109, "step": 3471 }, { "epoch": 0.42, "learning_rate": 6.8476703829519854e-06, "loss": 3.4005, "step": 3472 }, { "epoch": 0.42, "learning_rate": 6.845806346014072e-06, "loss": 3.4543, "step": 3473 }, { "epoch": 0.42, "learning_rate": 6.8439420120112e-06, "loss": 3.4279, "step": 3474 }, { "epoch": 0.42, "learning_rate": 6.842077381243417e-06, "loss": 3.3717, "step": 3475 }, { "epoch": 0.42, "learning_rate": 6.840212454010815e-06, "loss": 3.4722, "step": 3476 }, { "epoch": 0.42, "learning_rate": 6.8383472306135405e-06, "loss": 3.4041, "step": 3477 }, { "epoch": 0.42, "learning_rate": 6.836481711351778e-06, "loss": 3.3241, "step": 3478 }, { "epoch": 0.42, "learning_rate": 6.834615896525769e-06, "loss": 3.4052, "step": 3479 }, { "epoch": 0.42, "learning_rate": 6.832749786435797e-06, "loss": 3.3849, "step": 3480 }, { "epoch": 0.42, "learning_rate": 6.830883381382196e-06, "loss": 3.3948, "step": 3481 }, { "epoch": 0.42, "learning_rate": 6.8290166816653436e-06, "loss": 3.3826, "step": 3482 }, { "epoch": 0.42, "learning_rate": 6.8271496875856705e-06, "loss": 3.4848, "step": 3483 }, { "epoch": 0.42, "learning_rate": 6.825282399443649e-06, "loss": 3.3664, "step": 3484 }, { "epoch": 0.42, "learning_rate": 6.823414817539802e-06, "loss": 3.3259, "step": 3485 }, { "epoch": 0.42, "learning_rate": 6.821546942174701e-06, "loss": 3.4783, "step": 3486 }, { "epoch": 0.42, "learning_rate": 6.81967877364896e-06, "loss": 3.4135, "step": 3487 }, { "epoch": 0.42, "learning_rate": 6.817810312263245e-06, "loss": 3.4653, "step": 3488 }, { "epoch": 0.42, "learning_rate": 6.815941558318265e-06, "loss": 3.4445, "step": 3489 }, { "epoch": 0.42, "learning_rate": 6.814072512114781e-06, "loss": 3.3845, "step": 3490 }, { "epoch": 0.42, "learning_rate": 6.812203173953596e-06, "loss": 3.4181, "step": 3491 }, { "epoch": 0.42, "learning_rate": 6.810333544135561e-06, "loss": 3.4615, "step": 3492 }, { "epoch": 0.42, "learning_rate": 6.8084636229615786e-06, "loss": 3.4542, "step": 3493 }, { "epoch": 0.42, "learning_rate": 6.8065934107325935e-06, "loss": 3.4256, "step": 3494 }, { "epoch": 0.42, "learning_rate": 6.804722907749598e-06, "loss": 3.4228, "step": 3495 }, { "epoch": 0.42, "learning_rate": 6.802852114313632e-06, "loss": 3.4955, "step": 3496 }, { "epoch": 0.42, "learning_rate": 6.800981030725781e-06, "loss": 3.4032, "step": 3497 }, { "epoch": 0.42, "learning_rate": 6.7991096572871785e-06, "loss": 3.4753, "step": 3498 }, { "epoch": 0.42, "learning_rate": 6.797237994299006e-06, "loss": 3.2947, "step": 3499 }, { "epoch": 0.42, "learning_rate": 6.795366042062488e-06, "loss": 3.4829, "step": 3500 }, { "epoch": 0.42, "learning_rate": 6.7934938008788964e-06, "loss": 3.4284, "step": 3501 }, { "epoch": 0.42, "learning_rate": 6.791621271049553e-06, "loss": 3.4311, "step": 3502 }, { "epoch": 0.42, "learning_rate": 6.789748452875823e-06, "loss": 3.3869, "step": 3503 }, { "epoch": 0.42, "learning_rate": 6.787875346659118e-06, "loss": 3.3359, "step": 3504 }, { "epoch": 0.42, "learning_rate": 6.786001952700895e-06, "loss": 3.3343, "step": 3505 }, { "epoch": 0.42, "learning_rate": 6.784128271302663e-06, "loss": 3.442, "step": 3506 }, { "epoch": 0.42, "learning_rate": 6.782254302765972e-06, "loss": 3.4332, "step": 3507 }, { "epoch": 0.42, "learning_rate": 6.780380047392419e-06, "loss": 3.4355, "step": 3508 }, { "epoch": 0.42, "learning_rate": 6.778505505483644e-06, "loss": 3.5309, "step": 3509 }, { "epoch": 0.42, "learning_rate": 6.776630677341341e-06, "loss": 3.4175, "step": 3510 }, { "epoch": 0.42, "learning_rate": 6.774755563267245e-06, "loss": 3.369, "step": 3511 }, { "epoch": 0.42, "learning_rate": 6.772880163563136e-06, "loss": 3.431, "step": 3512 }, { "epoch": 0.42, "learning_rate": 6.771004478530845e-06, "loss": 3.4277, "step": 3513 }, { "epoch": 0.42, "learning_rate": 6.769128508472241e-06, "loss": 3.3986, "step": 3514 }, { "epoch": 0.42, "learning_rate": 6.767252253689247e-06, "loss": 3.3429, "step": 3515 }, { "epoch": 0.42, "learning_rate": 6.765375714483828e-06, "loss": 3.4222, "step": 3516 }, { "epoch": 0.42, "learning_rate": 6.763498891157995e-06, "loss": 3.4475, "step": 3517 }, { "epoch": 0.42, "learning_rate": 6.761621784013803e-06, "loss": 3.4173, "step": 3518 }, { "epoch": 0.42, "learning_rate": 6.759744393353356e-06, "loss": 3.4304, "step": 3519 }, { "epoch": 0.42, "learning_rate": 6.7578667194788025e-06, "loss": 3.3387, "step": 3520 }, { "epoch": 0.42, "learning_rate": 6.755988762692336e-06, "loss": 3.4055, "step": 3521 }, { "epoch": 0.42, "learning_rate": 6.754110523296196e-06, "loss": 3.2818, "step": 3522 }, { "epoch": 0.42, "learning_rate": 6.752232001592667e-06, "loss": 3.4753, "step": 3523 }, { "epoch": 0.42, "learning_rate": 6.750353197884077e-06, "loss": 3.3625, "step": 3524 }, { "epoch": 0.42, "learning_rate": 6.7484741124728044e-06, "loss": 3.4258, "step": 3525 }, { "epoch": 0.42, "learning_rate": 6.746594745661272e-06, "loss": 3.3948, "step": 3526 }, { "epoch": 0.42, "learning_rate": 6.744715097751941e-06, "loss": 3.4766, "step": 3527 }, { "epoch": 0.42, "learning_rate": 6.742835169047324e-06, "loss": 3.4235, "step": 3528 }, { "epoch": 0.42, "learning_rate": 6.74095495984998e-06, "loss": 3.4079, "step": 3529 }, { "epoch": 0.42, "learning_rate": 6.739074470462509e-06, "loss": 3.3901, "step": 3530 }, { "epoch": 0.42, "learning_rate": 6.73719370118756e-06, "loss": 3.4519, "step": 3531 }, { "epoch": 0.42, "learning_rate": 6.7353126523278215e-06, "loss": 3.4357, "step": 3532 }, { "epoch": 0.42, "learning_rate": 6.7334313241860325e-06, "loss": 3.4056, "step": 3533 }, { "epoch": 0.42, "learning_rate": 6.731549717064975e-06, "loss": 3.4949, "step": 3534 }, { "epoch": 0.42, "learning_rate": 6.729667831267474e-06, "loss": 3.5172, "step": 3535 }, { "epoch": 0.42, "learning_rate": 6.727785667096405e-06, "loss": 3.2925, "step": 3536 }, { "epoch": 0.42, "learning_rate": 6.725903224854679e-06, "loss": 3.3922, "step": 3537 }, { "epoch": 0.42, "learning_rate": 6.724020504845261e-06, "loss": 3.5249, "step": 3538 }, { "epoch": 0.42, "learning_rate": 6.722137507371156e-06, "loss": 3.4328, "step": 3539 }, { "epoch": 0.42, "learning_rate": 6.720254232735416e-06, "loss": 3.4173, "step": 3540 }, { "epoch": 0.43, "learning_rate": 6.718370681241133e-06, "loss": 3.3688, "step": 3541 }, { "epoch": 0.43, "learning_rate": 6.716486853191448e-06, "loss": 3.3453, "step": 3542 }, { "epoch": 0.43, "learning_rate": 6.714602748889545e-06, "loss": 3.3626, "step": 3543 }, { "epoch": 0.43, "learning_rate": 6.712718368638653e-06, "loss": 3.346, "step": 3544 }, { "epoch": 0.43, "learning_rate": 6.710833712742046e-06, "loss": 3.3197, "step": 3545 }, { "epoch": 0.43, "learning_rate": 6.708948781503039e-06, "loss": 3.3505, "step": 3546 }, { "epoch": 0.43, "learning_rate": 6.707063575224997e-06, "loss": 3.3231, "step": 3547 }, { "epoch": 0.43, "learning_rate": 6.705178094211323e-06, "loss": 3.3835, "step": 3548 }, { "epoch": 0.43, "learning_rate": 6.70329233876547e-06, "loss": 3.2615, "step": 3549 }, { "epoch": 0.43, "learning_rate": 6.701406309190929e-06, "loss": 3.4554, "step": 3550 }, { "epoch": 0.43, "learning_rate": 6.69952000579124e-06, "loss": 3.4171, "step": 3551 }, { "epoch": 0.43, "learning_rate": 6.697633428869987e-06, "loss": 3.4477, "step": 3552 }, { "epoch": 0.43, "learning_rate": 6.695746578730795e-06, "loss": 3.428, "step": 3553 }, { "epoch": 0.43, "learning_rate": 6.693859455677336e-06, "loss": 3.3632, "step": 3554 }, { "epoch": 0.43, "learning_rate": 6.691972060013322e-06, "loss": 3.4982, "step": 3555 }, { "epoch": 0.43, "learning_rate": 6.690084392042514e-06, "loss": 3.4431, "step": 3556 }, { "epoch": 0.43, "learning_rate": 6.688196452068712e-06, "loss": 3.3117, "step": 3557 }, { "epoch": 0.43, "learning_rate": 6.686308240395766e-06, "loss": 3.3606, "step": 3558 }, { "epoch": 0.43, "learning_rate": 6.684419757327561e-06, "loss": 3.4809, "step": 3559 }, { "epoch": 0.43, "learning_rate": 6.682531003168031e-06, "loss": 3.3963, "step": 3560 }, { "epoch": 0.43, "learning_rate": 6.680641978221155e-06, "loss": 3.3961, "step": 3561 }, { "epoch": 0.43, "learning_rate": 6.678752682790953e-06, "loss": 3.3268, "step": 3562 }, { "epoch": 0.43, "learning_rate": 6.676863117181489e-06, "loss": 3.4132, "step": 3563 }, { "epoch": 0.43, "learning_rate": 6.67497328169687e-06, "loss": 3.4828, "step": 3564 }, { "epoch": 0.43, "learning_rate": 6.673083176641246e-06, "loss": 3.4876, "step": 3565 }, { "epoch": 0.43, "learning_rate": 6.671192802318815e-06, "loss": 3.4338, "step": 3566 }, { "epoch": 0.43, "learning_rate": 6.6693021590338124e-06, "loss": 3.3524, "step": 3567 }, { "epoch": 0.43, "learning_rate": 6.667411247090519e-06, "loss": 3.4329, "step": 3568 }, { "epoch": 0.43, "learning_rate": 6.665520066793259e-06, "loss": 3.4368, "step": 3569 }, { "epoch": 0.43, "learning_rate": 6.6636286184464e-06, "loss": 3.4322, "step": 3570 }, { "epoch": 0.43, "learning_rate": 6.661736902354354e-06, "loss": 3.3432, "step": 3571 }, { "epoch": 0.43, "learning_rate": 6.659844918821574e-06, "loss": 3.4414, "step": 3572 }, { "epoch": 0.43, "learning_rate": 6.657952668152556e-06, "loss": 3.3852, "step": 3573 }, { "epoch": 0.43, "learning_rate": 6.6560601506518395e-06, "loss": 3.3666, "step": 3574 }, { "epoch": 0.43, "learning_rate": 6.654167366624009e-06, "loss": 3.4539, "step": 3575 }, { "epoch": 0.43, "learning_rate": 6.652274316373689e-06, "loss": 3.4196, "step": 3576 }, { "epoch": 0.43, "learning_rate": 6.6503810002055465e-06, "loss": 3.4328, "step": 3577 }, { "epoch": 0.43, "learning_rate": 6.648487418424296e-06, "loss": 3.4639, "step": 3578 }, { "epoch": 0.43, "learning_rate": 6.646593571334688e-06, "loss": 3.4427, "step": 3579 }, { "epoch": 0.43, "learning_rate": 6.64469945924152e-06, "loss": 3.3056, "step": 3580 }, { "epoch": 0.43, "learning_rate": 6.642805082449634e-06, "loss": 3.521, "step": 3581 }, { "epoch": 0.43, "learning_rate": 6.640910441263908e-06, "loss": 3.3642, "step": 3582 }, { "epoch": 0.43, "learning_rate": 6.639015535989268e-06, "loss": 3.3933, "step": 3583 }, { "epoch": 0.43, "learning_rate": 6.6371203669306825e-06, "loss": 3.4109, "step": 3584 }, { "epoch": 0.43, "learning_rate": 6.635224934393159e-06, "loss": 3.4166, "step": 3585 }, { "epoch": 0.43, "learning_rate": 6.6333292386817515e-06, "loss": 3.4196, "step": 3586 }, { "epoch": 0.43, "learning_rate": 6.631433280101551e-06, "loss": 3.4103, "step": 3587 }, { "epoch": 0.43, "learning_rate": 6.629537058957697e-06, "loss": 3.4407, "step": 3588 }, { "epoch": 0.43, "learning_rate": 6.6276405755553675e-06, "loss": 3.3962, "step": 3589 }, { "epoch": 0.43, "learning_rate": 6.625743830199781e-06, "loss": 3.3556, "step": 3590 }, { "epoch": 0.43, "learning_rate": 6.623846823196203e-06, "loss": 3.4356, "step": 3591 }, { "epoch": 0.43, "learning_rate": 6.621949554849938e-06, "loss": 3.3461, "step": 3592 }, { "epoch": 0.43, "learning_rate": 6.620052025466335e-06, "loss": 3.4404, "step": 3593 }, { "epoch": 0.43, "learning_rate": 6.6181542353507815e-06, "loss": 3.3817, "step": 3594 }, { "epoch": 0.43, "learning_rate": 6.616256184808707e-06, "loss": 3.483, "step": 3595 }, { "epoch": 0.43, "learning_rate": 6.614357874145588e-06, "loss": 3.4767, "step": 3596 }, { "epoch": 0.43, "learning_rate": 6.6124593036669386e-06, "loss": 3.4711, "step": 3597 }, { "epoch": 0.43, "learning_rate": 6.610560473678316e-06, "loss": 3.4791, "step": 3598 }, { "epoch": 0.43, "learning_rate": 6.6086613844853165e-06, "loss": 3.3616, "step": 3599 }, { "epoch": 0.43, "learning_rate": 6.606762036393582e-06, "loss": 3.4802, "step": 3600 }, { "epoch": 0.43, "learning_rate": 6.604862429708795e-06, "loss": 3.3646, "step": 3601 }, { "epoch": 0.43, "learning_rate": 6.602962564736679e-06, "loss": 3.4632, "step": 3602 }, { "epoch": 0.43, "learning_rate": 6.601062441782997e-06, "loss": 3.495, "step": 3603 }, { "epoch": 0.43, "learning_rate": 6.599162061153559e-06, "loss": 3.3911, "step": 3604 }, { "epoch": 0.43, "learning_rate": 6.597261423154211e-06, "loss": 3.3794, "step": 3605 }, { "epoch": 0.43, "learning_rate": 6.595360528090844e-06, "loss": 3.3494, "step": 3606 }, { "epoch": 0.43, "learning_rate": 6.593459376269389e-06, "loss": 3.4103, "step": 3607 }, { "epoch": 0.43, "learning_rate": 6.591557967995816e-06, "loss": 3.492, "step": 3608 }, { "epoch": 0.43, "learning_rate": 6.589656303576141e-06, "loss": 3.4472, "step": 3609 }, { "epoch": 0.43, "learning_rate": 6.587754383316417e-06, "loss": 3.4556, "step": 3610 }, { "epoch": 0.43, "learning_rate": 6.58585220752274e-06, "loss": 3.4828, "step": 3611 }, { "epoch": 0.43, "learning_rate": 6.58394977650125e-06, "loss": 3.4657, "step": 3612 }, { "epoch": 0.43, "learning_rate": 6.582047090558122e-06, "loss": 3.3306, "step": 3613 }, { "epoch": 0.43, "learning_rate": 6.580144149999577e-06, "loss": 3.3988, "step": 3614 }, { "epoch": 0.43, "learning_rate": 6.578240955131874e-06, "loss": 3.3723, "step": 3615 }, { "epoch": 0.43, "learning_rate": 6.576337506261314e-06, "loss": 3.3976, "step": 3616 }, { "epoch": 0.43, "learning_rate": 6.574433803694239e-06, "loss": 3.3781, "step": 3617 }, { "epoch": 0.43, "learning_rate": 6.572529847737033e-06, "loss": 3.3651, "step": 3618 }, { "epoch": 0.43, "learning_rate": 6.570625638696117e-06, "loss": 3.3534, "step": 3619 }, { "epoch": 0.43, "learning_rate": 6.56872117687796e-06, "loss": 3.49, "step": 3620 }, { "epoch": 0.43, "learning_rate": 6.566816462589064e-06, "loss": 3.3395, "step": 3621 }, { "epoch": 0.43, "learning_rate": 6.564911496135974e-06, "loss": 3.5079, "step": 3622 }, { "epoch": 0.43, "learning_rate": 6.563006277825276e-06, "loss": 3.433, "step": 3623 }, { "epoch": 0.44, "learning_rate": 6.5611008079636e-06, "loss": 3.4185, "step": 3624 }, { "epoch": 0.44, "learning_rate": 6.559195086857611e-06, "loss": 3.3197, "step": 3625 }, { "epoch": 0.44, "learning_rate": 6.557289114814014e-06, "loss": 3.4058, "step": 3626 }, { "epoch": 0.44, "learning_rate": 6.5553828921395615e-06, "loss": 3.5226, "step": 3627 }, { "epoch": 0.44, "learning_rate": 6.55347641914104e-06, "loss": 3.4639, "step": 3628 }, { "epoch": 0.44, "learning_rate": 6.551569696125277e-06, "loss": 3.3643, "step": 3629 }, { "epoch": 0.44, "learning_rate": 6.549662723399145e-06, "loss": 3.3998, "step": 3630 }, { "epoch": 0.44, "learning_rate": 6.547755501269549e-06, "loss": 3.3812, "step": 3631 }, { "epoch": 0.44, "learning_rate": 6.54584803004344e-06, "loss": 3.3431, "step": 3632 }, { "epoch": 0.44, "learning_rate": 6.543940310027809e-06, "loss": 3.4653, "step": 3633 }, { "epoch": 0.44, "learning_rate": 6.5420323415296824e-06, "loss": 3.4645, "step": 3634 }, { "epoch": 0.44, "learning_rate": 6.540124124856131e-06, "loss": 3.3887, "step": 3635 }, { "epoch": 0.44, "learning_rate": 6.5382156603142645e-06, "loss": 3.3955, "step": 3636 }, { "epoch": 0.44, "learning_rate": 6.536306948211232e-06, "loss": 3.4104, "step": 3637 }, { "epoch": 0.44, "learning_rate": 6.53439798885422e-06, "loss": 3.5215, "step": 3638 }, { "epoch": 0.44, "learning_rate": 6.532488782550464e-06, "loss": 3.4594, "step": 3639 }, { "epoch": 0.44, "learning_rate": 6.530579329607225e-06, "loss": 3.3975, "step": 3640 }, { "epoch": 0.44, "learning_rate": 6.528669630331814e-06, "loss": 3.3919, "step": 3641 }, { "epoch": 0.44, "learning_rate": 6.526759685031581e-06, "loss": 3.3223, "step": 3642 }, { "epoch": 0.44, "learning_rate": 6.524849494013911e-06, "loss": 3.4586, "step": 3643 }, { "epoch": 0.44, "learning_rate": 6.522939057586232e-06, "loss": 3.4285, "step": 3644 }, { "epoch": 0.44, "learning_rate": 6.521028376056009e-06, "loss": 3.3588, "step": 3645 }, { "epoch": 0.44, "learning_rate": 6.519117449730751e-06, "loss": 3.4792, "step": 3646 }, { "epoch": 0.44, "learning_rate": 6.517206278918001e-06, "loss": 3.476, "step": 3647 }, { "epoch": 0.44, "learning_rate": 6.515294863925345e-06, "loss": 3.3099, "step": 3648 }, { "epoch": 0.44, "learning_rate": 6.513383205060407e-06, "loss": 3.4131, "step": 3649 }, { "epoch": 0.44, "learning_rate": 6.5114713026308495e-06, "loss": 3.4474, "step": 3650 }, { "epoch": 0.44, "learning_rate": 6.5095591569443765e-06, "loss": 3.4142, "step": 3651 }, { "epoch": 0.44, "learning_rate": 6.507646768308726e-06, "loss": 3.5439, "step": 3652 }, { "epoch": 0.44, "learning_rate": 6.505734137031683e-06, "loss": 3.4798, "step": 3653 }, { "epoch": 0.44, "learning_rate": 6.503821263421065e-06, "loss": 3.3144, "step": 3654 }, { "epoch": 0.44, "learning_rate": 6.501908147784732e-06, "loss": 3.4402, "step": 3655 }, { "epoch": 0.44, "learning_rate": 6.49999479043058e-06, "loss": 3.3382, "step": 3656 }, { "epoch": 0.44, "learning_rate": 6.498081191666549e-06, "loss": 3.4586, "step": 3657 }, { "epoch": 0.44, "learning_rate": 6.496167351800611e-06, "loss": 3.3381, "step": 3658 }, { "epoch": 0.44, "learning_rate": 6.494253271140781e-06, "loss": 3.3994, "step": 3659 }, { "epoch": 0.44, "learning_rate": 6.492338949995113e-06, "loss": 3.5039, "step": 3660 }, { "epoch": 0.44, "learning_rate": 6.4904243886716985e-06, "loss": 3.3825, "step": 3661 }, { "epoch": 0.44, "learning_rate": 6.4885095874786684e-06, "loss": 3.4666, "step": 3662 }, { "epoch": 0.44, "learning_rate": 6.4865945467241905e-06, "loss": 3.3887, "step": 3663 }, { "epoch": 0.44, "learning_rate": 6.484679266716474e-06, "loss": 3.3899, "step": 3664 }, { "epoch": 0.44, "learning_rate": 6.482763747763763e-06, "loss": 3.401, "step": 3665 }, { "epoch": 0.44, "learning_rate": 6.480847990174343e-06, "loss": 3.4557, "step": 3666 }, { "epoch": 0.44, "learning_rate": 6.478931994256536e-06, "loss": 3.4176, "step": 3667 }, { "epoch": 0.44, "learning_rate": 6.477015760318706e-06, "loss": 3.3155, "step": 3668 }, { "epoch": 0.44, "learning_rate": 6.475099288669251e-06, "loss": 3.4684, "step": 3669 }, { "epoch": 0.44, "learning_rate": 6.473182579616608e-06, "loss": 3.3411, "step": 3670 }, { "epoch": 0.44, "learning_rate": 6.471265633469255e-06, "loss": 3.4309, "step": 3671 }, { "epoch": 0.44, "learning_rate": 6.469348450535704e-06, "loss": 3.3679, "step": 3672 }, { "epoch": 0.44, "learning_rate": 6.4674310311245085e-06, "loss": 3.4263, "step": 3673 }, { "epoch": 0.44, "learning_rate": 6.465513375544259e-06, "loss": 3.4474, "step": 3674 }, { "epoch": 0.44, "learning_rate": 6.463595484103584e-06, "loss": 3.3352, "step": 3675 }, { "epoch": 0.44, "learning_rate": 6.461677357111149e-06, "loss": 3.3679, "step": 3676 }, { "epoch": 0.44, "learning_rate": 6.459758994875659e-06, "loss": 3.4431, "step": 3677 }, { "epoch": 0.44, "learning_rate": 6.457840397705856e-06, "loss": 3.3633, "step": 3678 }, { "epoch": 0.44, "learning_rate": 6.455921565910521e-06, "loss": 3.3823, "step": 3679 }, { "epoch": 0.44, "learning_rate": 6.454002499798471e-06, "loss": 3.3639, "step": 3680 }, { "epoch": 0.44, "learning_rate": 6.452083199678559e-06, "loss": 3.3555, "step": 3681 }, { "epoch": 0.44, "learning_rate": 6.450163665859681e-06, "loss": 3.3687, "step": 3682 }, { "epoch": 0.44, "learning_rate": 6.448243898650767e-06, "loss": 3.4284, "step": 3683 }, { "epoch": 0.44, "learning_rate": 6.446323898360785e-06, "loss": 3.3975, "step": 3684 }, { "epoch": 0.44, "learning_rate": 6.444403665298741e-06, "loss": 3.4486, "step": 3685 }, { "epoch": 0.44, "learning_rate": 6.4424831997736785e-06, "loss": 3.3988, "step": 3686 }, { "epoch": 0.44, "learning_rate": 6.440562502094676e-06, "loss": 3.3647, "step": 3687 }, { "epoch": 0.44, "learning_rate": 6.438641572570856e-06, "loss": 3.3578, "step": 3688 }, { "epoch": 0.44, "learning_rate": 6.43672041151137e-06, "loss": 3.3115, "step": 3689 }, { "epoch": 0.44, "learning_rate": 6.434799019225412e-06, "loss": 3.3951, "step": 3690 }, { "epoch": 0.44, "learning_rate": 6.4328773960222105e-06, "loss": 3.291, "step": 3691 }, { "epoch": 0.44, "learning_rate": 6.430955542211034e-06, "loss": 3.4984, "step": 3692 }, { "epoch": 0.44, "learning_rate": 6.429033458101186e-06, "loss": 3.3943, "step": 3693 }, { "epoch": 0.44, "learning_rate": 6.427111144002009e-06, "loss": 3.3558, "step": 3694 }, { "epoch": 0.44, "learning_rate": 6.425188600222877e-06, "loss": 3.4075, "step": 3695 }, { "epoch": 0.44, "learning_rate": 6.42326582707321e-06, "loss": 3.383, "step": 3696 }, { "epoch": 0.44, "learning_rate": 6.421342824862458e-06, "loss": 3.3931, "step": 3697 }, { "epoch": 0.44, "learning_rate": 6.419419593900109e-06, "loss": 3.3776, "step": 3698 }, { "epoch": 0.44, "learning_rate": 6.4174961344956885e-06, "loss": 3.4678, "step": 3699 }, { "epoch": 0.44, "learning_rate": 6.415572446958762e-06, "loss": 3.3647, "step": 3700 }, { "epoch": 0.44, "learning_rate": 6.413648531598926e-06, "loss": 3.4252, "step": 3701 }, { "epoch": 0.44, "learning_rate": 6.411724388725817e-06, "loss": 3.4249, "step": 3702 }, { "epoch": 0.44, "learning_rate": 6.4098000186491075e-06, "loss": 3.4193, "step": 3703 }, { "epoch": 0.44, "learning_rate": 6.407875421678507e-06, "loss": 3.4374, "step": 3704 }, { "epoch": 0.44, "learning_rate": 6.40595059812376e-06, "loss": 3.351, "step": 3705 }, { "epoch": 0.44, "learning_rate": 6.404025548294651e-06, "loss": 3.375, "step": 3706 }, { "epoch": 0.44, "learning_rate": 6.402100272500995e-06, "loss": 3.3978, "step": 3707 }, { "epoch": 0.45, "learning_rate": 6.400174771052648e-06, "loss": 3.4574, "step": 3708 }, { "epoch": 0.45, "learning_rate": 6.398249044259503e-06, "loss": 3.4029, "step": 3709 }, { "epoch": 0.45, "learning_rate": 6.396323092431486e-06, "loss": 3.3591, "step": 3710 }, { "epoch": 0.45, "learning_rate": 6.3943969158785604e-06, "loss": 3.4564, "step": 3711 }, { "epoch": 0.45, "learning_rate": 6.392470514910726e-06, "loss": 3.4753, "step": 3712 }, { "epoch": 0.45, "learning_rate": 6.3905438898380175e-06, "loss": 3.392, "step": 3713 }, { "epoch": 0.45, "learning_rate": 6.388617040970509e-06, "loss": 3.3841, "step": 3714 }, { "epoch": 0.45, "learning_rate": 6.386689968618309e-06, "loss": 3.4341, "step": 3715 }, { "epoch": 0.45, "learning_rate": 6.384762673091559e-06, "loss": 3.3052, "step": 3716 }, { "epoch": 0.45, "learning_rate": 6.38283515470044e-06, "loss": 3.4399, "step": 3717 }, { "epoch": 0.45, "learning_rate": 6.380907413755169e-06, "loss": 3.3986, "step": 3718 }, { "epoch": 0.45, "learning_rate": 6.378979450565994e-06, "loss": 3.405, "step": 3719 }, { "epoch": 0.45, "learning_rate": 6.3770512654432075e-06, "loss": 3.3539, "step": 3720 }, { "epoch": 0.45, "learning_rate": 6.375122858697128e-06, "loss": 3.3862, "step": 3721 }, { "epoch": 0.45, "learning_rate": 6.3731942306381155e-06, "loss": 3.2997, "step": 3722 }, { "epoch": 0.45, "learning_rate": 6.371265381576566e-06, "loss": 3.3852, "step": 3723 }, { "epoch": 0.45, "learning_rate": 6.369336311822908e-06, "loss": 3.379, "step": 3724 }, { "epoch": 0.45, "learning_rate": 6.367407021687606e-06, "loss": 3.4352, "step": 3725 }, { "epoch": 0.45, "learning_rate": 6.365477511481163e-06, "loss": 3.3968, "step": 3726 }, { "epoch": 0.45, "learning_rate": 6.363547781514113e-06, "loss": 3.487, "step": 3727 }, { "epoch": 0.45, "learning_rate": 6.361617832097029e-06, "loss": 3.2993, "step": 3728 }, { "epoch": 0.45, "learning_rate": 6.359687663540519e-06, "loss": 3.4282, "step": 3729 }, { "epoch": 0.45, "learning_rate": 6.357757276155223e-06, "loss": 3.5395, "step": 3730 }, { "epoch": 0.45, "learning_rate": 6.35582667025182e-06, "loss": 3.4883, "step": 3731 }, { "epoch": 0.45, "learning_rate": 6.353895846141022e-06, "loss": 3.2947, "step": 3732 }, { "epoch": 0.45, "learning_rate": 6.351964804133577e-06, "loss": 3.5094, "step": 3733 }, { "epoch": 0.45, "learning_rate": 6.350033544540268e-06, "loss": 3.3856, "step": 3734 }, { "epoch": 0.45, "learning_rate": 6.348102067671913e-06, "loss": 3.425, "step": 3735 }, { "epoch": 0.45, "learning_rate": 6.346170373839363e-06, "loss": 3.3578, "step": 3736 }, { "epoch": 0.45, "learning_rate": 6.344238463353508e-06, "loss": 3.4128, "step": 3737 }, { "epoch": 0.45, "learning_rate": 6.342306336525271e-06, "loss": 3.4634, "step": 3738 }, { "epoch": 0.45, "learning_rate": 6.340373993665607e-06, "loss": 3.3656, "step": 3739 }, { "epoch": 0.45, "learning_rate": 6.3384414350855095e-06, "loss": 3.4592, "step": 3740 }, { "epoch": 0.45, "learning_rate": 6.3365086610960055e-06, "loss": 3.3579, "step": 3741 }, { "epoch": 0.45, "learning_rate": 6.334575672008157e-06, "loss": 3.3654, "step": 3742 }, { "epoch": 0.45, "learning_rate": 6.332642468133059e-06, "loss": 3.4086, "step": 3743 }, { "epoch": 0.45, "learning_rate": 6.330709049781842e-06, "loss": 3.4548, "step": 3744 }, { "epoch": 0.45, "learning_rate": 6.328775417265673e-06, "loss": 3.3187, "step": 3745 }, { "epoch": 0.45, "learning_rate": 6.32684157089575e-06, "loss": 3.4826, "step": 3746 }, { "epoch": 0.45, "learning_rate": 6.32490751098331e-06, "loss": 3.458, "step": 3747 }, { "epoch": 0.45, "learning_rate": 6.322973237839618e-06, "loss": 3.367, "step": 3748 }, { "epoch": 0.45, "learning_rate": 6.3210387517759785e-06, "loss": 3.385, "step": 3749 }, { "epoch": 0.45, "learning_rate": 6.319104053103728e-06, "loss": 3.3008, "step": 3750 }, { "epoch": 0.45, "learning_rate": 6.317169142134238e-06, "loss": 3.398, "step": 3751 }, { "epoch": 0.45, "learning_rate": 6.315234019178916e-06, "loss": 3.3399, "step": 3752 }, { "epoch": 0.45, "learning_rate": 6.3132986845491984e-06, "loss": 3.4952, "step": 3753 }, { "epoch": 0.45, "learning_rate": 6.311363138556561e-06, "loss": 3.4491, "step": 3754 }, { "epoch": 0.45, "learning_rate": 6.309427381512512e-06, "loss": 3.4163, "step": 3755 }, { "epoch": 0.45, "learning_rate": 6.307491413728592e-06, "loss": 3.4102, "step": 3756 }, { "epoch": 0.45, "learning_rate": 6.305555235516375e-06, "loss": 3.5002, "step": 3757 }, { "epoch": 0.45, "learning_rate": 6.3036188471874725e-06, "loss": 3.4725, "step": 3758 }, { "epoch": 0.45, "learning_rate": 6.301682249053528e-06, "loss": 3.4097, "step": 3759 }, { "epoch": 0.45, "learning_rate": 6.2997454414262185e-06, "loss": 3.4075, "step": 3760 }, { "epoch": 0.45, "learning_rate": 6.2978084246172525e-06, "loss": 3.4979, "step": 3761 }, { "epoch": 0.45, "learning_rate": 6.295871198938376e-06, "loss": 3.4575, "step": 3762 }, { "epoch": 0.45, "learning_rate": 6.293933764701368e-06, "loss": 3.4775, "step": 3763 }, { "epoch": 0.45, "learning_rate": 6.291996122218038e-06, "loss": 3.3657, "step": 3764 }, { "epoch": 0.45, "learning_rate": 6.290058271800232e-06, "loss": 3.4168, "step": 3765 }, { "epoch": 0.45, "learning_rate": 6.2881202137598296e-06, "loss": 3.4305, "step": 3766 }, { "epoch": 0.45, "learning_rate": 6.286181948408743e-06, "loss": 3.4351, "step": 3767 }, { "epoch": 0.45, "learning_rate": 6.284243476058914e-06, "loss": 3.3544, "step": 3768 }, { "epoch": 0.45, "learning_rate": 6.282304797022327e-06, "loss": 3.524, "step": 3769 }, { "epoch": 0.45, "learning_rate": 6.280365911610989e-06, "loss": 3.4806, "step": 3770 }, { "epoch": 0.45, "learning_rate": 6.278426820136947e-06, "loss": 3.3636, "step": 3771 }, { "epoch": 0.45, "learning_rate": 6.276487522912279e-06, "loss": 3.4055, "step": 3772 }, { "epoch": 0.45, "learning_rate": 6.274548020249097e-06, "loss": 3.4435, "step": 3773 }, { "epoch": 0.45, "learning_rate": 6.272608312459546e-06, "loss": 3.3355, "step": 3774 }, { "epoch": 0.45, "learning_rate": 6.270668399855802e-06, "loss": 3.5297, "step": 3775 }, { "epoch": 0.45, "learning_rate": 6.268728282750076e-06, "loss": 3.5441, "step": 3776 }, { "epoch": 0.45, "learning_rate": 6.26678796145461e-06, "loss": 3.3604, "step": 3777 }, { "epoch": 0.45, "learning_rate": 6.2648474362816845e-06, "loss": 3.4696, "step": 3778 }, { "epoch": 0.45, "learning_rate": 6.262906707543604e-06, "loss": 3.3693, "step": 3779 }, { "epoch": 0.45, "learning_rate": 6.2609657755527135e-06, "loss": 3.4164, "step": 3780 }, { "epoch": 0.45, "learning_rate": 6.2590246406213835e-06, "loss": 3.5418, "step": 3781 }, { "epoch": 0.45, "learning_rate": 6.2570833030620256e-06, "loss": 3.4955, "step": 3782 }, { "epoch": 0.45, "learning_rate": 6.255141763187078e-06, "loss": 3.4732, "step": 3783 }, { "epoch": 0.45, "learning_rate": 6.253200021309012e-06, "loss": 3.4683, "step": 3784 }, { "epoch": 0.45, "learning_rate": 6.251258077740333e-06, "loss": 3.3348, "step": 3785 }, { "epoch": 0.45, "learning_rate": 6.2493159327935795e-06, "loss": 3.457, "step": 3786 }, { "epoch": 0.45, "learning_rate": 6.247373586781321e-06, "loss": 3.4091, "step": 3787 }, { "epoch": 0.45, "learning_rate": 6.245431040016159e-06, "loss": 3.4916, "step": 3788 }, { "epoch": 0.45, "learning_rate": 6.2434882928107265e-06, "loss": 3.4604, "step": 3789 }, { "epoch": 0.45, "learning_rate": 6.241545345477692e-06, "loss": 3.3124, "step": 3790 }, { "epoch": 0.46, "learning_rate": 6.239602198329755e-06, "loss": 3.3072, "step": 3791 }, { "epoch": 0.46, "learning_rate": 6.237658851679645e-06, "loss": 3.4634, "step": 3792 }, { "epoch": 0.46, "learning_rate": 6.235715305840126e-06, "loss": 3.3181, "step": 3793 }, { "epoch": 0.46, "learning_rate": 6.233771561123991e-06, "loss": 3.3921, "step": 3794 }, { "epoch": 0.46, "learning_rate": 6.231827617844071e-06, "loss": 3.4334, "step": 3795 }, { "epoch": 0.46, "learning_rate": 6.2298834763132235e-06, "loss": 3.3885, "step": 3796 }, { "epoch": 0.46, "learning_rate": 6.227939136844337e-06, "loss": 3.3709, "step": 3797 }, { "epoch": 0.46, "learning_rate": 6.225994599750338e-06, "loss": 3.4097, "step": 3798 }, { "epoch": 0.46, "learning_rate": 6.224049865344179e-06, "loss": 3.4221, "step": 3799 }, { "epoch": 0.46, "learning_rate": 6.222104933938847e-06, "loss": 3.4326, "step": 3800 }, { "epoch": 0.46, "learning_rate": 6.220159805847363e-06, "loss": 3.3866, "step": 3801 }, { "epoch": 0.46, "learning_rate": 6.218214481382774e-06, "loss": 3.3962, "step": 3802 }, { "epoch": 0.46, "learning_rate": 6.2162689608581606e-06, "loss": 3.4494, "step": 3803 }, { "epoch": 0.46, "learning_rate": 6.214323244586637e-06, "loss": 3.3476, "step": 3804 }, { "epoch": 0.46, "learning_rate": 6.21237733288135e-06, "loss": 3.4322, "step": 3805 }, { "epoch": 0.46, "learning_rate": 6.210431226055471e-06, "loss": 3.3781, "step": 3806 }, { "epoch": 0.46, "learning_rate": 6.208484924422211e-06, "loss": 3.4272, "step": 3807 }, { "epoch": 0.46, "learning_rate": 6.206538428294807e-06, "loss": 3.306, "step": 3808 }, { "epoch": 0.46, "learning_rate": 6.20459173798653e-06, "loss": 3.4201, "step": 3809 }, { "epoch": 0.46, "learning_rate": 6.202644853810681e-06, "loss": 3.4405, "step": 3810 }, { "epoch": 0.46, "learning_rate": 6.200697776080591e-06, "loss": 3.3854, "step": 3811 }, { "epoch": 0.46, "learning_rate": 6.198750505109625e-06, "loss": 3.3777, "step": 3812 }, { "epoch": 0.46, "learning_rate": 6.196803041211179e-06, "loss": 3.444, "step": 3813 }, { "epoch": 0.46, "learning_rate": 6.194855384698676e-06, "loss": 3.4032, "step": 3814 }, { "epoch": 0.46, "learning_rate": 6.192907535885574e-06, "loss": 3.4374, "step": 3815 }, { "epoch": 0.46, "learning_rate": 6.190959495085361e-06, "loss": 3.3053, "step": 3816 }, { "epoch": 0.46, "learning_rate": 6.189011262611556e-06, "loss": 3.4219, "step": 3817 }, { "epoch": 0.46, "learning_rate": 6.187062838777707e-06, "loss": 3.4653, "step": 3818 }, { "epoch": 0.46, "learning_rate": 6.185114223897398e-06, "loss": 3.4175, "step": 3819 }, { "epoch": 0.46, "learning_rate": 6.183165418284235e-06, "loss": 3.3792, "step": 3820 }, { "epoch": 0.46, "learning_rate": 6.1812164222518626e-06, "loss": 3.4682, "step": 3821 }, { "epoch": 0.46, "learning_rate": 6.179267236113952e-06, "loss": 3.4394, "step": 3822 }, { "epoch": 0.46, "learning_rate": 6.177317860184208e-06, "loss": 3.5032, "step": 3823 }, { "epoch": 0.46, "learning_rate": 6.1753682947763605e-06, "loss": 3.3356, "step": 3824 }, { "epoch": 0.46, "learning_rate": 6.173418540204177e-06, "loss": 3.43, "step": 3825 }, { "epoch": 0.46, "learning_rate": 6.17146859678145e-06, "loss": 3.4191, "step": 3826 }, { "epoch": 0.46, "learning_rate": 6.169518464822005e-06, "loss": 3.4301, "step": 3827 }, { "epoch": 0.46, "learning_rate": 6.167568144639699e-06, "loss": 3.4334, "step": 3828 }, { "epoch": 0.46, "learning_rate": 6.165617636548415e-06, "loss": 3.4317, "step": 3829 }, { "epoch": 0.46, "learning_rate": 6.163666940862068e-06, "loss": 3.4192, "step": 3830 }, { "epoch": 0.46, "learning_rate": 6.161716057894606e-06, "loss": 3.4566, "step": 3831 }, { "epoch": 0.46, "learning_rate": 6.159764987960004e-06, "loss": 3.3942, "step": 3832 }, { "epoch": 0.46, "learning_rate": 6.157813731372269e-06, "loss": 3.4291, "step": 3833 }, { "epoch": 0.46, "learning_rate": 6.155862288445435e-06, "loss": 3.3047, "step": 3834 }, { "epoch": 0.46, "learning_rate": 6.153910659493571e-06, "loss": 3.4818, "step": 3835 }, { "epoch": 0.46, "learning_rate": 6.151958844830772e-06, "loss": 3.4187, "step": 3836 }, { "epoch": 0.46, "learning_rate": 6.150006844771164e-06, "loss": 3.4517, "step": 3837 }, { "epoch": 0.46, "learning_rate": 6.148054659628903e-06, "loss": 3.4812, "step": 3838 }, { "epoch": 0.46, "learning_rate": 6.146102289718172e-06, "loss": 3.3891, "step": 3839 }, { "epoch": 0.46, "learning_rate": 6.1441497353531906e-06, "loss": 3.3672, "step": 3840 }, { "epoch": 0.46, "learning_rate": 6.142196996848202e-06, "loss": 3.3797, "step": 3841 }, { "epoch": 0.46, "learning_rate": 6.140244074517481e-06, "loss": 3.4795, "step": 3842 }, { "epoch": 0.46, "learning_rate": 6.13829096867533e-06, "loss": 3.4775, "step": 3843 }, { "epoch": 0.46, "learning_rate": 6.136337679636085e-06, "loss": 3.4292, "step": 3844 }, { "epoch": 0.46, "learning_rate": 6.134384207714109e-06, "loss": 3.4336, "step": 3845 }, { "epoch": 0.46, "learning_rate": 6.132430553223796e-06, "loss": 3.4444, "step": 3846 }, { "epoch": 0.46, "learning_rate": 6.1304767164795645e-06, "loss": 3.3217, "step": 3847 }, { "epoch": 0.46, "learning_rate": 6.128522697795867e-06, "loss": 3.5032, "step": 3848 }, { "epoch": 0.46, "learning_rate": 6.126568497487186e-06, "loss": 3.4736, "step": 3849 }, { "epoch": 0.46, "learning_rate": 6.12461411586803e-06, "loss": 3.483, "step": 3850 }, { "epoch": 0.46, "learning_rate": 6.122659553252941e-06, "loss": 3.3427, "step": 3851 }, { "epoch": 0.46, "learning_rate": 6.120704809956483e-06, "loss": 3.3928, "step": 3852 }, { "epoch": 0.46, "learning_rate": 6.118749886293255e-06, "loss": 3.466, "step": 3853 }, { "epoch": 0.46, "learning_rate": 6.116794782577884e-06, "loss": 3.4081, "step": 3854 }, { "epoch": 0.46, "learning_rate": 6.114839499125026e-06, "loss": 3.4514, "step": 3855 }, { "epoch": 0.46, "learning_rate": 6.112884036249363e-06, "loss": 3.3163, "step": 3856 }, { "epoch": 0.46, "learning_rate": 6.110928394265608e-06, "loss": 3.3551, "step": 3857 }, { "epoch": 0.46, "learning_rate": 6.1089725734885066e-06, "loss": 3.3106, "step": 3858 }, { "epoch": 0.46, "learning_rate": 6.107016574232826e-06, "loss": 3.397, "step": 3859 }, { "epoch": 0.46, "learning_rate": 6.105060396813368e-06, "loss": 3.38, "step": 3860 }, { "epoch": 0.46, "learning_rate": 6.103104041544958e-06, "loss": 3.3731, "step": 3861 }, { "epoch": 0.46, "learning_rate": 6.101147508742456e-06, "loss": 3.4051, "step": 3862 }, { "epoch": 0.46, "learning_rate": 6.0991907987207436e-06, "loss": 3.4229, "step": 3863 }, { "epoch": 0.46, "learning_rate": 6.097233911794738e-06, "loss": 3.482, "step": 3864 }, { "epoch": 0.46, "learning_rate": 6.095276848279379e-06, "loss": 3.4338, "step": 3865 }, { "epoch": 0.46, "learning_rate": 6.09331960848964e-06, "loss": 3.3451, "step": 3866 }, { "epoch": 0.46, "learning_rate": 6.091362192740518e-06, "loss": 3.4411, "step": 3867 }, { "epoch": 0.46, "learning_rate": 6.08940460134704e-06, "loss": 3.3995, "step": 3868 }, { "epoch": 0.46, "learning_rate": 6.087446834624266e-06, "loss": 3.376, "step": 3869 }, { "epoch": 0.46, "learning_rate": 6.085488892887273e-06, "loss": 3.4563, "step": 3870 }, { "epoch": 0.46, "learning_rate": 6.083530776451179e-06, "loss": 3.3865, "step": 3871 }, { "epoch": 0.46, "learning_rate": 6.081572485631122e-06, "loss": 3.4391, "step": 3872 }, { "epoch": 0.46, "learning_rate": 6.079614020742271e-06, "loss": 3.4447, "step": 3873 }, { "epoch": 0.47, "learning_rate": 6.07765538209982e-06, "loss": 3.3903, "step": 3874 }, { "epoch": 0.47, "learning_rate": 6.075696570018994e-06, "loss": 3.4542, "step": 3875 }, { "epoch": 0.47, "learning_rate": 6.073737584815047e-06, "loss": 3.4031, "step": 3876 }, { "epoch": 0.47, "learning_rate": 6.071778426803259e-06, "loss": 3.487, "step": 3877 }, { "epoch": 0.47, "learning_rate": 6.069819096298935e-06, "loss": 3.3409, "step": 3878 }, { "epoch": 0.47, "learning_rate": 6.067859593617414e-06, "loss": 3.3569, "step": 3879 }, { "epoch": 0.47, "learning_rate": 6.065899919074056e-06, "loss": 3.4919, "step": 3880 }, { "epoch": 0.47, "learning_rate": 6.063940072984256e-06, "loss": 3.3245, "step": 3881 }, { "epoch": 0.47, "learning_rate": 6.0619800556634274e-06, "loss": 3.4043, "step": 3882 }, { "epoch": 0.47, "learning_rate": 6.06001986742702e-06, "loss": 3.4977, "step": 3883 }, { "epoch": 0.47, "learning_rate": 6.058059508590508e-06, "loss": 3.422, "step": 3884 }, { "epoch": 0.47, "learning_rate": 6.056098979469389e-06, "loss": 3.3356, "step": 3885 }, { "epoch": 0.47, "learning_rate": 6.054138280379195e-06, "loss": 3.4584, "step": 3886 }, { "epoch": 0.47, "learning_rate": 6.052177411635481e-06, "loss": 3.3247, "step": 3887 }, { "epoch": 0.47, "learning_rate": 6.050216373553829e-06, "loss": 3.4323, "step": 3888 }, { "epoch": 0.47, "learning_rate": 6.048255166449849e-06, "loss": 3.5292, "step": 3889 }, { "epoch": 0.47, "learning_rate": 6.04629379063918e-06, "loss": 3.4482, "step": 3890 }, { "epoch": 0.47, "learning_rate": 6.044332246437488e-06, "loss": 3.4197, "step": 3891 }, { "epoch": 0.47, "learning_rate": 6.042370534160461e-06, "loss": 3.366, "step": 3892 }, { "epoch": 0.47, "learning_rate": 6.040408654123822e-06, "loss": 3.3855, "step": 3893 }, { "epoch": 0.47, "learning_rate": 6.038446606643313e-06, "loss": 3.2925, "step": 3894 }, { "epoch": 0.47, "learning_rate": 6.0364843920347096e-06, "loss": 3.3879, "step": 3895 }, { "epoch": 0.47, "learning_rate": 6.034522010613812e-06, "loss": 3.3995, "step": 3896 }, { "epoch": 0.47, "learning_rate": 6.032559462696446e-06, "loss": 3.3782, "step": 3897 }, { "epoch": 0.47, "learning_rate": 6.030596748598462e-06, "loss": 3.4134, "step": 3898 }, { "epoch": 0.47, "learning_rate": 6.028633868635744e-06, "loss": 3.3466, "step": 3899 }, { "epoch": 0.47, "learning_rate": 6.026670823124199e-06, "loss": 3.3774, "step": 3900 }, { "epoch": 0.47, "learning_rate": 6.024707612379757e-06, "loss": 3.4489, "step": 3901 }, { "epoch": 0.47, "learning_rate": 6.022744236718382e-06, "loss": 3.2794, "step": 3902 }, { "epoch": 0.47, "learning_rate": 6.020780696456059e-06, "loss": 3.4228, "step": 3903 }, { "epoch": 0.47, "learning_rate": 6.0188169919088005e-06, "loss": 3.4611, "step": 3904 }, { "epoch": 0.47, "learning_rate": 6.016853123392647e-06, "loss": 3.4052, "step": 3905 }, { "epoch": 0.47, "learning_rate": 6.0148890912236626e-06, "loss": 3.4182, "step": 3906 }, { "epoch": 0.47, "learning_rate": 6.0129248957179424e-06, "loss": 3.4346, "step": 3907 }, { "epoch": 0.47, "learning_rate": 6.010960537191602e-06, "loss": 3.3442, "step": 3908 }, { "epoch": 0.47, "learning_rate": 6.0089960159607895e-06, "loss": 3.434, "step": 3909 }, { "epoch": 0.47, "learning_rate": 6.007031332341673e-06, "loss": 3.4897, "step": 3910 }, { "epoch": 0.47, "learning_rate": 6.00506648665045e-06, "loss": 3.3819, "step": 3911 }, { "epoch": 0.47, "learning_rate": 6.003101479203346e-06, "loss": 3.3693, "step": 3912 }, { "epoch": 0.47, "learning_rate": 6.001136310316607e-06, "loss": 3.4666, "step": 3913 }, { "epoch": 0.47, "learning_rate": 5.999170980306511e-06, "loss": 3.4467, "step": 3914 }, { "epoch": 0.47, "learning_rate": 5.997205489489359e-06, "loss": 3.4498, "step": 3915 }, { "epoch": 0.47, "learning_rate": 5.995239838181475e-06, "loss": 3.3601, "step": 3916 }, { "epoch": 0.47, "learning_rate": 5.993274026699215e-06, "loss": 3.3896, "step": 3917 }, { "epoch": 0.47, "learning_rate": 5.991308055358957e-06, "loss": 3.4331, "step": 3918 }, { "epoch": 0.47, "learning_rate": 5.989341924477105e-06, "loss": 3.4801, "step": 3919 }, { "epoch": 0.47, "learning_rate": 5.987375634370088e-06, "loss": 3.4484, "step": 3920 }, { "epoch": 0.47, "learning_rate": 5.985409185354364e-06, "loss": 3.4189, "step": 3921 }, { "epoch": 0.47, "learning_rate": 5.983442577746412e-06, "loss": 3.4482, "step": 3922 }, { "epoch": 0.47, "learning_rate": 5.981475811862741e-06, "loss": 3.4455, "step": 3923 }, { "epoch": 0.47, "learning_rate": 5.97950888801988e-06, "loss": 3.4811, "step": 3924 }, { "epoch": 0.47, "learning_rate": 5.977541806534391e-06, "loss": 3.392, "step": 3925 }, { "epoch": 0.47, "learning_rate": 5.975574567722853e-06, "loss": 3.4366, "step": 3926 }, { "epoch": 0.47, "learning_rate": 5.973607171901876e-06, "loss": 3.4243, "step": 3927 }, { "epoch": 0.47, "learning_rate": 5.971639619388094e-06, "loss": 3.3127, "step": 3928 }, { "epoch": 0.47, "learning_rate": 5.969671910498166e-06, "loss": 3.4863, "step": 3929 }, { "epoch": 0.47, "learning_rate": 5.967704045548775e-06, "loss": 3.3525, "step": 3930 }, { "epoch": 0.47, "learning_rate": 5.9657360248566295e-06, "loss": 3.3777, "step": 3931 }, { "epoch": 0.47, "learning_rate": 5.963767848738467e-06, "loss": 3.4491, "step": 3932 }, { "epoch": 0.47, "learning_rate": 5.961799517511042e-06, "loss": 3.3887, "step": 3933 }, { "epoch": 0.47, "learning_rate": 5.959831031491143e-06, "loss": 3.4413, "step": 3934 }, { "epoch": 0.47, "learning_rate": 5.957862390995577e-06, "loss": 3.4367, "step": 3935 }, { "epoch": 0.47, "learning_rate": 5.955893596341179e-06, "loss": 3.3749, "step": 3936 }, { "epoch": 0.47, "learning_rate": 5.953924647844806e-06, "loss": 3.4518, "step": 3937 }, { "epoch": 0.47, "learning_rate": 5.951955545823342e-06, "loss": 3.4616, "step": 3938 }, { "epoch": 0.47, "learning_rate": 5.949986290593697e-06, "loss": 3.4498, "step": 3939 }, { "epoch": 0.47, "learning_rate": 5.9480168824728025e-06, "loss": 3.4361, "step": 3940 }, { "epoch": 0.47, "learning_rate": 5.946047321777616e-06, "loss": 3.4801, "step": 3941 }, { "epoch": 0.47, "learning_rate": 5.94407760882512e-06, "loss": 3.364, "step": 3942 }, { "epoch": 0.47, "learning_rate": 5.942107743932319e-06, "loss": 3.4512, "step": 3943 }, { "epoch": 0.47, "learning_rate": 5.940137727416247e-06, "loss": 3.4192, "step": 3944 }, { "epoch": 0.47, "learning_rate": 5.938167559593958e-06, "loss": 3.3853, "step": 3945 }, { "epoch": 0.47, "learning_rate": 5.936197240782531e-06, "loss": 3.4141, "step": 3946 }, { "epoch": 0.47, "learning_rate": 5.934226771299072e-06, "loss": 3.4188, "step": 3947 }, { "epoch": 0.47, "learning_rate": 5.932256151460707e-06, "loss": 3.4486, "step": 3948 }, { "epoch": 0.47, "learning_rate": 5.930285381584591e-06, "loss": 3.4089, "step": 3949 }, { "epoch": 0.47, "learning_rate": 5.9283144619879e-06, "loss": 3.3294, "step": 3950 }, { "epoch": 0.47, "learning_rate": 5.926343392987832e-06, "loss": 3.4388, "step": 3951 }, { "epoch": 0.47, "learning_rate": 5.9243721749016145e-06, "loss": 3.4179, "step": 3952 }, { "epoch": 0.47, "learning_rate": 5.922400808046497e-06, "loss": 3.5004, "step": 3953 }, { "epoch": 0.47, "learning_rate": 5.920429292739751e-06, "loss": 3.3499, "step": 3954 }, { "epoch": 0.47, "learning_rate": 5.918457629298672e-06, "loss": 3.4089, "step": 3955 }, { "epoch": 0.47, "learning_rate": 5.9164858180405805e-06, "loss": 3.4468, "step": 3956 }, { "epoch": 0.47, "learning_rate": 5.914513859282823e-06, "loss": 3.4333, "step": 3957 }, { "epoch": 0.48, "learning_rate": 5.912541753342766e-06, "loss": 3.4872, "step": 3958 }, { "epoch": 0.48, "learning_rate": 5.910569500537803e-06, "loss": 3.4869, "step": 3959 }, { "epoch": 0.48, "learning_rate": 5.908597101185346e-06, "loss": 3.4088, "step": 3960 }, { "epoch": 0.48, "learning_rate": 5.906624555602836e-06, "loss": 3.3759, "step": 3961 }, { "epoch": 0.48, "learning_rate": 5.904651864107734e-06, "loss": 3.4508, "step": 3962 }, { "epoch": 0.48, "learning_rate": 5.902679027017528e-06, "loss": 3.4187, "step": 3963 }, { "epoch": 0.48, "learning_rate": 5.9007060446497255e-06, "loss": 3.3936, "step": 3964 }, { "epoch": 0.48, "learning_rate": 5.898732917321861e-06, "loss": 3.4399, "step": 3965 }, { "epoch": 0.48, "learning_rate": 5.896759645351488e-06, "loss": 3.454, "step": 3966 }, { "epoch": 0.48, "learning_rate": 5.894786229056189e-06, "loss": 3.3601, "step": 3967 }, { "epoch": 0.48, "learning_rate": 5.892812668753567e-06, "loss": 3.3768, "step": 3968 }, { "epoch": 0.48, "learning_rate": 5.890838964761243e-06, "loss": 3.4056, "step": 3969 }, { "epoch": 0.48, "learning_rate": 5.888865117396869e-06, "loss": 3.3888, "step": 3970 }, { "epoch": 0.48, "learning_rate": 5.886891126978116e-06, "loss": 3.3349, "step": 3971 }, { "epoch": 0.48, "learning_rate": 5.884916993822681e-06, "loss": 3.3552, "step": 3972 }, { "epoch": 0.48, "learning_rate": 5.88294271824828e-06, "loss": 3.4141, "step": 3973 }, { "epoch": 0.48, "learning_rate": 5.8809683005726535e-06, "loss": 3.4121, "step": 3974 }, { "epoch": 0.48, "learning_rate": 5.878993741113567e-06, "loss": 3.4742, "step": 3975 }, { "epoch": 0.48, "learning_rate": 5.877019040188805e-06, "loss": 3.4651, "step": 3976 }, { "epoch": 0.48, "learning_rate": 5.87504419811618e-06, "loss": 3.5436, "step": 3977 }, { "epoch": 0.48, "learning_rate": 5.8730692152135205e-06, "loss": 3.4094, "step": 3978 }, { "epoch": 0.48, "learning_rate": 5.871094091798683e-06, "loss": 3.4546, "step": 3979 }, { "epoch": 0.48, "learning_rate": 5.869118828189543e-06, "loss": 3.3771, "step": 3980 }, { "epoch": 0.48, "learning_rate": 5.867143424704003e-06, "loss": 3.4283, "step": 3981 }, { "epoch": 0.48, "learning_rate": 5.8651678816599835e-06, "loss": 3.3666, "step": 3982 }, { "epoch": 0.48, "learning_rate": 5.863192199375431e-06, "loss": 3.4039, "step": 3983 }, { "epoch": 0.48, "learning_rate": 5.86121637816831e-06, "loss": 3.4901, "step": 3984 }, { "epoch": 0.48, "learning_rate": 5.859240418356614e-06, "loss": 3.3929, "step": 3985 }, { "epoch": 0.48, "learning_rate": 5.8572643202583535e-06, "loss": 3.5344, "step": 3986 }, { "epoch": 0.48, "learning_rate": 5.855288084191561e-06, "loss": 3.3273, "step": 3987 }, { "epoch": 0.48, "learning_rate": 5.853311710474293e-06, "loss": 3.4233, "step": 3988 }, { "epoch": 0.48, "learning_rate": 5.851335199424629e-06, "loss": 3.4508, "step": 3989 }, { "epoch": 0.48, "learning_rate": 5.849358551360671e-06, "loss": 3.4421, "step": 3990 }, { "epoch": 0.48, "learning_rate": 5.84738176660054e-06, "loss": 3.3706, "step": 3991 }, { "epoch": 0.48, "learning_rate": 5.8454048454623805e-06, "loss": 3.5014, "step": 3992 }, { "epoch": 0.48, "learning_rate": 5.843427788264358e-06, "loss": 3.5147, "step": 3993 }, { "epoch": 0.48, "learning_rate": 5.841450595324665e-06, "loss": 3.3777, "step": 3994 }, { "epoch": 0.48, "learning_rate": 5.8394732669615086e-06, "loss": 3.4626, "step": 3995 }, { "epoch": 0.48, "learning_rate": 5.83749580349312e-06, "loss": 3.4826, "step": 3996 }, { "epoch": 0.48, "learning_rate": 5.835518205237758e-06, "loss": 3.4865, "step": 3997 }, { "epoch": 0.48, "learning_rate": 5.833540472513694e-06, "loss": 3.3853, "step": 3998 }, { "epoch": 0.48, "learning_rate": 5.831562605639226e-06, "loss": 3.518, "step": 3999 }, { "epoch": 0.48, "learning_rate": 5.8295846049326744e-06, "loss": 3.421, "step": 4000 }, { "epoch": 0.48, "learning_rate": 5.827606470712377e-06, "loss": 3.3561, "step": 4001 }, { "epoch": 0.48, "learning_rate": 5.825628203296698e-06, "loss": 3.3801, "step": 4002 }, { "epoch": 0.48, "learning_rate": 5.823649803004019e-06, "loss": 3.3167, "step": 4003 }, { "epoch": 0.48, "learning_rate": 5.821671270152748e-06, "loss": 3.3648, "step": 4004 }, { "epoch": 0.48, "learning_rate": 5.819692605061307e-06, "loss": 3.4198, "step": 4005 }, { "epoch": 0.48, "learning_rate": 5.817713808048144e-06, "loss": 3.4165, "step": 4006 }, { "epoch": 0.48, "learning_rate": 5.815734879431729e-06, "loss": 3.4753, "step": 4007 }, { "epoch": 0.48, "learning_rate": 5.813755819530552e-06, "loss": 3.4281, "step": 4008 }, { "epoch": 0.48, "learning_rate": 5.811776628663123e-06, "loss": 3.4025, "step": 4009 }, { "epoch": 0.48, "learning_rate": 5.809797307147973e-06, "loss": 3.4829, "step": 4010 }, { "epoch": 0.48, "learning_rate": 5.807817855303656e-06, "loss": 3.4977, "step": 4011 }, { "epoch": 0.48, "learning_rate": 5.805838273448745e-06, "loss": 3.3388, "step": 4012 }, { "epoch": 0.48, "learning_rate": 5.803858561901836e-06, "loss": 3.4258, "step": 4013 }, { "epoch": 0.48, "learning_rate": 5.801878720981545e-06, "loss": 3.4591, "step": 4014 }, { "epoch": 0.48, "learning_rate": 5.799898751006509e-06, "loss": 3.3593, "step": 4015 }, { "epoch": 0.48, "learning_rate": 5.797918652295382e-06, "loss": 3.3037, "step": 4016 }, { "epoch": 0.48, "learning_rate": 5.7959384251668475e-06, "loss": 3.4589, "step": 4017 }, { "epoch": 0.48, "learning_rate": 5.793958069939598e-06, "loss": 3.4467, "step": 4018 }, { "epoch": 0.48, "learning_rate": 5.791977586932356e-06, "loss": 3.3113, "step": 4019 }, { "epoch": 0.48, "learning_rate": 5.789996976463862e-06, "loss": 3.3749, "step": 4020 }, { "epoch": 0.48, "learning_rate": 5.788016238852876e-06, "loss": 3.2965, "step": 4021 }, { "epoch": 0.48, "learning_rate": 5.786035374418178e-06, "loss": 3.4204, "step": 4022 }, { "epoch": 0.48, "learning_rate": 5.78405438347857e-06, "loss": 3.4118, "step": 4023 }, { "epoch": 0.48, "learning_rate": 5.7820732663528725e-06, "loss": 3.3477, "step": 4024 }, { "epoch": 0.48, "learning_rate": 5.780092023359929e-06, "loss": 3.3153, "step": 4025 }, { "epoch": 0.48, "learning_rate": 5.778110654818602e-06, "loss": 3.3761, "step": 4026 }, { "epoch": 0.48, "learning_rate": 5.776129161047771e-06, "loss": 3.2435, "step": 4027 }, { "epoch": 0.48, "learning_rate": 5.774147542366342e-06, "loss": 3.3582, "step": 4028 }, { "epoch": 0.48, "learning_rate": 5.772165799093234e-06, "loss": 3.4042, "step": 4029 }, { "epoch": 0.48, "learning_rate": 5.770183931547394e-06, "loss": 3.4665, "step": 4030 }, { "epoch": 0.48, "learning_rate": 5.768201940047782e-06, "loss": 3.434, "step": 4031 }, { "epoch": 0.48, "learning_rate": 5.766219824913381e-06, "loss": 3.3961, "step": 4032 }, { "epoch": 0.48, "learning_rate": 5.764237586463193e-06, "loss": 3.419, "step": 4033 }, { "epoch": 0.48, "learning_rate": 5.762255225016241e-06, "loss": 3.4495, "step": 4034 }, { "epoch": 0.48, "learning_rate": 5.7602727408915695e-06, "loss": 3.4426, "step": 4035 }, { "epoch": 0.48, "learning_rate": 5.758290134408237e-06, "loss": 3.4452, "step": 4036 }, { "epoch": 0.48, "learning_rate": 5.756307405885326e-06, "loss": 3.3726, "step": 4037 }, { "epoch": 0.48, "learning_rate": 5.7543245556419395e-06, "loss": 3.3743, "step": 4038 }, { "epoch": 0.48, "learning_rate": 5.752341583997196e-06, "loss": 3.3184, "step": 4039 }, { "epoch": 0.48, "learning_rate": 5.750358491270238e-06, "loss": 3.4356, "step": 4040 }, { "epoch": 0.49, "learning_rate": 5.748375277780224e-06, "loss": 3.5948, "step": 4041 }, { "epoch": 0.49, "learning_rate": 5.746391943846333e-06, "loss": 3.3813, "step": 4042 }, { "epoch": 0.49, "learning_rate": 5.744408489787765e-06, "loss": 3.3696, "step": 4043 }, { "epoch": 0.49, "learning_rate": 5.742424915923738e-06, "loss": 3.363, "step": 4044 }, { "epoch": 0.49, "learning_rate": 5.740441222573489e-06, "loss": 3.5205, "step": 4045 }, { "epoch": 0.49, "learning_rate": 5.738457410056273e-06, "loss": 3.3214, "step": 4046 }, { "epoch": 0.49, "learning_rate": 5.7364734786913675e-06, "loss": 3.4055, "step": 4047 }, { "epoch": 0.49, "learning_rate": 5.734489428798068e-06, "loss": 3.4112, "step": 4048 }, { "epoch": 0.49, "learning_rate": 5.732505260695687e-06, "loss": 3.4176, "step": 4049 }, { "epoch": 0.49, "learning_rate": 5.730520974703559e-06, "loss": 3.4117, "step": 4050 }, { "epoch": 0.49, "learning_rate": 5.728536571141033e-06, "loss": 3.4893, "step": 4051 }, { "epoch": 0.49, "learning_rate": 5.7265520503274814e-06, "loss": 3.4985, "step": 4052 }, { "epoch": 0.49, "learning_rate": 5.724567412582297e-06, "loss": 3.3675, "step": 4053 }, { "epoch": 0.49, "learning_rate": 5.7225826582248835e-06, "loss": 3.4347, "step": 4054 }, { "epoch": 0.49, "learning_rate": 5.720597787574671e-06, "loss": 3.4496, "step": 4055 }, { "epoch": 0.49, "learning_rate": 5.718612800951104e-06, "loss": 3.3735, "step": 4056 }, { "epoch": 0.49, "learning_rate": 5.716627698673648e-06, "loss": 3.3117, "step": 4057 }, { "epoch": 0.49, "learning_rate": 5.714642481061788e-06, "loss": 3.3978, "step": 4058 }, { "epoch": 0.49, "learning_rate": 5.712657148435024e-06, "loss": 3.4186, "step": 4059 }, { "epoch": 0.49, "learning_rate": 5.7106717011128756e-06, "loss": 3.4331, "step": 4060 }, { "epoch": 0.49, "learning_rate": 5.708686139414882e-06, "loss": 3.4057, "step": 4061 }, { "epoch": 0.49, "learning_rate": 5.7067004636606014e-06, "loss": 3.5063, "step": 4062 }, { "epoch": 0.49, "learning_rate": 5.70471467416961e-06, "loss": 3.4243, "step": 4063 }, { "epoch": 0.49, "learning_rate": 5.702728771261501e-06, "loss": 3.3566, "step": 4064 }, { "epoch": 0.49, "learning_rate": 5.700742755255885e-06, "loss": 3.3504, "step": 4065 }, { "epoch": 0.49, "learning_rate": 5.698756626472395e-06, "loss": 3.4538, "step": 4066 }, { "epoch": 0.49, "learning_rate": 5.696770385230679e-06, "loss": 3.4405, "step": 4067 }, { "epoch": 0.49, "learning_rate": 5.694784031850401e-06, "loss": 3.3998, "step": 4068 }, { "epoch": 0.49, "learning_rate": 5.6927975666512495e-06, "loss": 3.4195, "step": 4069 }, { "epoch": 0.49, "learning_rate": 5.690810989952924e-06, "loss": 3.3904, "step": 4070 }, { "epoch": 0.49, "learning_rate": 5.688824302075147e-06, "loss": 3.3574, "step": 4071 }, { "epoch": 0.49, "learning_rate": 5.686837503337656e-06, "loss": 3.3651, "step": 4072 }, { "epoch": 0.49, "learning_rate": 5.684850594060208e-06, "loss": 3.4906, "step": 4073 }, { "epoch": 0.49, "learning_rate": 5.682863574562576e-06, "loss": 3.4133, "step": 4074 }, { "epoch": 0.49, "learning_rate": 5.680876445164552e-06, "loss": 3.3261, "step": 4075 }, { "epoch": 0.49, "learning_rate": 5.678889206185949e-06, "loss": 3.3402, "step": 4076 }, { "epoch": 0.49, "learning_rate": 5.67690185794659e-06, "loss": 3.4598, "step": 4077 }, { "epoch": 0.49, "learning_rate": 5.67491440076632e-06, "loss": 3.3486, "step": 4078 }, { "epoch": 0.49, "learning_rate": 5.672926834965004e-06, "loss": 3.4342, "step": 4079 }, { "epoch": 0.49, "learning_rate": 5.670939160862521e-06, "loss": 3.2815, "step": 4080 }, { "epoch": 0.49, "learning_rate": 5.668951378778765e-06, "loss": 3.4047, "step": 4081 }, { "epoch": 0.49, "learning_rate": 5.666963489033655e-06, "loss": 3.3698, "step": 4082 }, { "epoch": 0.49, "learning_rate": 5.664975491947121e-06, "loss": 3.3358, "step": 4083 }, { "epoch": 0.49, "learning_rate": 5.662987387839111e-06, "loss": 3.455, "step": 4084 }, { "epoch": 0.49, "learning_rate": 5.660999177029596e-06, "loss": 3.4608, "step": 4085 }, { "epoch": 0.49, "learning_rate": 5.659010859838554e-06, "loss": 3.3881, "step": 4086 }, { "epoch": 0.49, "learning_rate": 5.657022436585987e-06, "loss": 3.356, "step": 4087 }, { "epoch": 0.49, "learning_rate": 5.6550339075919154e-06, "loss": 3.4018, "step": 4088 }, { "epoch": 0.49, "learning_rate": 5.653045273176373e-06, "loss": 3.3977, "step": 4089 }, { "epoch": 0.49, "learning_rate": 5.651056533659411e-06, "loss": 3.4726, "step": 4090 }, { "epoch": 0.49, "learning_rate": 5.649067689361096e-06, "loss": 3.4121, "step": 4091 }, { "epoch": 0.49, "learning_rate": 5.647078740601517e-06, "loss": 3.5135, "step": 4092 }, { "epoch": 0.49, "learning_rate": 5.645089687700774e-06, "loss": 3.2813, "step": 4093 }, { "epoch": 0.49, "learning_rate": 5.643100530978989e-06, "loss": 3.331, "step": 4094 }, { "epoch": 0.49, "learning_rate": 5.641111270756294e-06, "loss": 3.4545, "step": 4095 }, { "epoch": 0.49, "learning_rate": 5.639121907352845e-06, "loss": 3.3728, "step": 4096 }, { "epoch": 0.49, "learning_rate": 5.63713244108881e-06, "loss": 3.4455, "step": 4097 }, { "epoch": 0.49, "learning_rate": 5.635142872284374e-06, "loss": 3.3621, "step": 4098 }, { "epoch": 0.49, "learning_rate": 5.633153201259738e-06, "loss": 3.3393, "step": 4099 }, { "epoch": 0.49, "learning_rate": 5.631163428335124e-06, "loss": 3.37, "step": 4100 }, { "epoch": 0.49, "learning_rate": 5.629173553830763e-06, "loss": 3.4209, "step": 4101 }, { "epoch": 0.49, "learning_rate": 5.627183578066909e-06, "loss": 3.4351, "step": 4102 }, { "epoch": 0.49, "learning_rate": 5.62519350136383e-06, "loss": 3.4165, "step": 4103 }, { "epoch": 0.49, "learning_rate": 5.623203324041808e-06, "loss": 3.3758, "step": 4104 }, { "epoch": 0.49, "learning_rate": 5.6212130464211445e-06, "loss": 3.3879, "step": 4105 }, { "epoch": 0.49, "learning_rate": 5.6192226688221556e-06, "loss": 3.4603, "step": 4106 }, { "epoch": 0.49, "learning_rate": 5.617232191565174e-06, "loss": 3.4124, "step": 4107 }, { "epoch": 0.49, "learning_rate": 5.615241614970546e-06, "loss": 3.4228, "step": 4108 }, { "epoch": 0.49, "learning_rate": 5.613250939358636e-06, "loss": 3.5214, "step": 4109 }, { "epoch": 0.49, "learning_rate": 5.611260165049828e-06, "loss": 3.4412, "step": 4110 }, { "epoch": 0.49, "learning_rate": 5.6092692923645145e-06, "loss": 3.3172, "step": 4111 }, { "epoch": 0.49, "learning_rate": 5.607278321623109e-06, "loss": 3.3176, "step": 4112 }, { "epoch": 0.49, "learning_rate": 5.605287253146039e-06, "loss": 3.4006, "step": 4113 }, { "epoch": 0.49, "learning_rate": 5.603296087253749e-06, "loss": 3.3396, "step": 4114 }, { "epoch": 0.49, "learning_rate": 5.601304824266696e-06, "loss": 3.3874, "step": 4115 }, { "epoch": 0.49, "learning_rate": 5.599313464505357e-06, "loss": 3.394, "step": 4116 }, { "epoch": 0.49, "learning_rate": 5.597322008290221e-06, "loss": 3.3735, "step": 4117 }, { "epoch": 0.49, "learning_rate": 5.595330455941795e-06, "loss": 3.385, "step": 4118 }, { "epoch": 0.49, "learning_rate": 5.5933388077805995e-06, "loss": 3.3792, "step": 4119 }, { "epoch": 0.49, "learning_rate": 5.591347064127172e-06, "loss": 3.4119, "step": 4120 }, { "epoch": 0.49, "learning_rate": 5.5893552253020646e-06, "loss": 3.3513, "step": 4121 }, { "epoch": 0.49, "learning_rate": 5.587363291625842e-06, "loss": 3.412, "step": 4122 }, { "epoch": 0.49, "learning_rate": 5.585371263419091e-06, "loss": 3.4512, "step": 4123 }, { "epoch": 0.5, "learning_rate": 5.583379141002408e-06, "loss": 3.3861, "step": 4124 }, { "epoch": 0.5, "learning_rate": 5.581386924696408e-06, "loss": 3.417, "step": 4125 }, { "epoch": 0.5, "learning_rate": 5.579394614821716e-06, "loss": 3.4801, "step": 4126 }, { "epoch": 0.5, "learning_rate": 5.577402211698976e-06, "loss": 3.3968, "step": 4127 }, { "epoch": 0.5, "learning_rate": 5.575409715648847e-06, "loss": 3.3921, "step": 4128 }, { "epoch": 0.5, "learning_rate": 5.573417126992004e-06, "loss": 3.3732, "step": 4129 }, { "epoch": 0.5, "learning_rate": 5.571424446049132e-06, "loss": 3.3673, "step": 4130 }, { "epoch": 0.5, "learning_rate": 5.569431673140935e-06, "loss": 3.4135, "step": 4131 }, { "epoch": 0.5, "learning_rate": 5.567438808588132e-06, "loss": 3.3975, "step": 4132 }, { "epoch": 0.5, "learning_rate": 5.565445852711455e-06, "loss": 3.3137, "step": 4133 }, { "epoch": 0.5, "learning_rate": 5.563452805831652e-06, "loss": 3.4713, "step": 4134 }, { "epoch": 0.5, "learning_rate": 5.561459668269483e-06, "loss": 3.513, "step": 4135 }, { "epoch": 0.5, "learning_rate": 5.559466440345725e-06, "loss": 3.3097, "step": 4136 }, { "epoch": 0.5, "learning_rate": 5.557473122381172e-06, "loss": 3.4335, "step": 4137 }, { "epoch": 0.5, "learning_rate": 5.555479714696625e-06, "loss": 3.4988, "step": 4138 }, { "epoch": 0.5, "learning_rate": 5.553486217612908e-06, "loss": 3.3968, "step": 4139 }, { "epoch": 0.5, "learning_rate": 5.551492631450852e-06, "loss": 3.4802, "step": 4140 }, { "epoch": 0.5, "learning_rate": 5.549498956531308e-06, "loss": 3.4755, "step": 4141 }, { "epoch": 0.5, "learning_rate": 5.547505193175138e-06, "loss": 3.361, "step": 4142 }, { "epoch": 0.5, "learning_rate": 5.54551134170322e-06, "loss": 3.3516, "step": 4143 }, { "epoch": 0.5, "learning_rate": 5.5435174024364445e-06, "loss": 3.414, "step": 4144 }, { "epoch": 0.5, "learning_rate": 5.541523375695717e-06, "loss": 3.3375, "step": 4145 }, { "epoch": 0.5, "learning_rate": 5.539529261801959e-06, "loss": 3.3384, "step": 4146 }, { "epoch": 0.5, "learning_rate": 5.537535061076102e-06, "loss": 3.4369, "step": 4147 }, { "epoch": 0.5, "learning_rate": 5.535540773839096e-06, "loss": 3.4104, "step": 4148 }, { "epoch": 0.5, "learning_rate": 5.533546400411899e-06, "loss": 3.4636, "step": 4149 }, { "epoch": 0.5, "learning_rate": 5.5315519411154884e-06, "loss": 3.3441, "step": 4150 }, { "epoch": 0.5, "learning_rate": 5.529557396270853e-06, "loss": 3.5102, "step": 4151 }, { "epoch": 0.5, "learning_rate": 5.527562766198999e-06, "loss": 3.3916, "step": 4152 }, { "epoch": 0.5, "learning_rate": 5.525568051220938e-06, "loss": 3.3303, "step": 4153 }, { "epoch": 0.5, "learning_rate": 5.523573251657702e-06, "loss": 3.4474, "step": 4154 }, { "epoch": 0.5, "learning_rate": 5.521578367830336e-06, "loss": 3.3765, "step": 4155 }, { "epoch": 0.5, "learning_rate": 5.519583400059898e-06, "loss": 3.395, "step": 4156 }, { "epoch": 0.5, "learning_rate": 5.5175883486674586e-06, "loss": 3.4019, "step": 4157 }, { "epoch": 0.5, "learning_rate": 5.515593213974102e-06, "loss": 3.4062, "step": 4158 }, { "epoch": 0.5, "learning_rate": 5.513597996300925e-06, "loss": 3.3518, "step": 4159 }, { "epoch": 0.5, "learning_rate": 5.511602695969041e-06, "loss": 3.3866, "step": 4160 }, { "epoch": 0.5, "learning_rate": 5.509607313299573e-06, "loss": 3.4109, "step": 4161 }, { "epoch": 0.5, "learning_rate": 5.507611848613661e-06, "loss": 3.3692, "step": 4162 }, { "epoch": 0.5, "learning_rate": 5.5056163022324524e-06, "loss": 3.4061, "step": 4163 }, { "epoch": 0.5, "learning_rate": 5.5036206744771145e-06, "loss": 3.4127, "step": 4164 }, { "epoch": 0.5, "learning_rate": 5.501624965668824e-06, "loss": 3.3791, "step": 4165 }, { "epoch": 0.5, "learning_rate": 5.499629176128771e-06, "loss": 3.4255, "step": 4166 }, { "epoch": 0.5, "learning_rate": 5.49763330617816e-06, "loss": 3.4461, "step": 4167 }, { "epoch": 0.5, "learning_rate": 5.4956373561382035e-06, "loss": 3.3739, "step": 4168 }, { "epoch": 0.5, "learning_rate": 5.493641326330133e-06, "loss": 3.3649, "step": 4169 }, { "epoch": 0.5, "learning_rate": 5.491645217075193e-06, "loss": 3.3878, "step": 4170 }, { "epoch": 0.5, "learning_rate": 5.489649028694635e-06, "loss": 3.3801, "step": 4171 }, { "epoch": 0.5, "learning_rate": 5.487652761509727e-06, "loss": 3.4835, "step": 4172 }, { "epoch": 0.5, "learning_rate": 5.485656415841748e-06, "loss": 3.3916, "step": 4173 }, { "epoch": 0.5, "learning_rate": 5.483659992011993e-06, "loss": 3.3987, "step": 4174 }, { "epoch": 0.5, "learning_rate": 5.48166349034177e-06, "loss": 3.5269, "step": 4175 }, { "epoch": 0.5, "learning_rate": 5.4796669111523916e-06, "loss": 3.391, "step": 4176 }, { "epoch": 0.5, "learning_rate": 5.477670254765189e-06, "loss": 3.3471, "step": 4177 }, { "epoch": 0.5, "learning_rate": 5.475673521501507e-06, "loss": 3.4052, "step": 4178 }, { "epoch": 0.5, "learning_rate": 5.4736767116827015e-06, "loss": 3.4036, "step": 4179 }, { "epoch": 0.5, "learning_rate": 5.471679825630139e-06, "loss": 3.4506, "step": 4180 }, { "epoch": 0.5, "learning_rate": 5.4696828636651975e-06, "loss": 3.4527, "step": 4181 }, { "epoch": 0.5, "learning_rate": 5.467685826109272e-06, "loss": 3.4846, "step": 4182 }, { "epoch": 0.5, "learning_rate": 5.4656887132837646e-06, "loss": 3.4301, "step": 4183 }, { "epoch": 0.5, "learning_rate": 5.463691525510093e-06, "loss": 3.3816, "step": 4184 }, { "epoch": 0.5, "learning_rate": 5.461694263109684e-06, "loss": 3.3801, "step": 4185 }, { "epoch": 0.5, "learning_rate": 5.459696926403979e-06, "loss": 3.396, "step": 4186 }, { "epoch": 0.5, "learning_rate": 5.457699515714429e-06, "loss": 3.436, "step": 4187 }, { "epoch": 0.5, "learning_rate": 5.455702031362502e-06, "loss": 3.3743, "step": 4188 }, { "epoch": 0.5, "learning_rate": 5.45370447366967e-06, "loss": 3.4362, "step": 4189 }, { "epoch": 0.5, "learning_rate": 5.4517068429574215e-06, "loss": 3.3989, "step": 4190 }, { "epoch": 0.5, "learning_rate": 5.449709139547258e-06, "loss": 3.4469, "step": 4191 }, { "epoch": 0.5, "learning_rate": 5.447711363760688e-06, "loss": 3.3802, "step": 4192 }, { "epoch": 0.5, "learning_rate": 5.4457135159192385e-06, "loss": 3.3972, "step": 4193 }, { "epoch": 0.5, "learning_rate": 5.443715596344441e-06, "loss": 3.4598, "step": 4194 }, { "epoch": 0.5, "learning_rate": 5.441717605357843e-06, "loss": 3.389, "step": 4195 }, { "epoch": 0.5, "learning_rate": 5.4397195432810015e-06, "loss": 3.3519, "step": 4196 }, { "epoch": 0.5, "learning_rate": 5.437721410435486e-06, "loss": 3.3762, "step": 4197 }, { "epoch": 0.5, "learning_rate": 5.435723207142877e-06, "loss": 3.3992, "step": 4198 }, { "epoch": 0.5, "learning_rate": 5.4337249337247645e-06, "loss": 3.3885, "step": 4199 }, { "epoch": 0.5, "learning_rate": 5.431726590502754e-06, "loss": 3.4725, "step": 4200 }, { "epoch": 0.5, "learning_rate": 5.429728177798459e-06, "loss": 3.4469, "step": 4201 }, { "epoch": 0.5, "learning_rate": 5.427729695933506e-06, "loss": 3.2438, "step": 4202 }, { "epoch": 0.5, "learning_rate": 5.425731145229528e-06, "loss": 3.3829, "step": 4203 }, { "epoch": 0.5, "learning_rate": 5.423732526008176e-06, "loss": 3.3288, "step": 4204 }, { "epoch": 0.5, "learning_rate": 5.421733838591109e-06, "loss": 3.4247, "step": 4205 }, { "epoch": 0.5, "learning_rate": 5.4197350832999965e-06, "loss": 3.3069, "step": 4206 }, { "epoch": 0.5, "learning_rate": 5.4177362604565165e-06, "loss": 3.4379, "step": 4207 }, { "epoch": 0.51, "learning_rate": 5.415737370382362e-06, "loss": 3.4627, "step": 4208 }, { "epoch": 0.51, "learning_rate": 5.413738413399237e-06, "loss": 3.5206, "step": 4209 }, { "epoch": 0.51, "learning_rate": 5.411739389828853e-06, "loss": 3.3785, "step": 4210 }, { "epoch": 0.51, "learning_rate": 5.4097402999929325e-06, "loss": 3.4089, "step": 4211 }, { "epoch": 0.51, "learning_rate": 5.4077411442132135e-06, "loss": 3.4289, "step": 4212 }, { "epoch": 0.51, "learning_rate": 5.405741922811438e-06, "loss": 3.3521, "step": 4213 }, { "epoch": 0.51, "learning_rate": 5.4037426361093635e-06, "loss": 3.3788, "step": 4214 }, { "epoch": 0.51, "learning_rate": 5.4017432844287556e-06, "loss": 3.388, "step": 4215 }, { "epoch": 0.51, "learning_rate": 5.39974386809139e-06, "loss": 3.4687, "step": 4216 }, { "epoch": 0.51, "learning_rate": 5.397744387419055e-06, "loss": 3.4149, "step": 4217 }, { "epoch": 0.51, "learning_rate": 5.395744842733547e-06, "loss": 3.3828, "step": 4218 }, { "epoch": 0.51, "learning_rate": 5.393745234356675e-06, "loss": 3.3729, "step": 4219 }, { "epoch": 0.51, "learning_rate": 5.391745562610255e-06, "loss": 3.3723, "step": 4220 }, { "epoch": 0.51, "learning_rate": 5.389745827816118e-06, "loss": 3.4395, "step": 4221 }, { "epoch": 0.51, "learning_rate": 5.3877460302960986e-06, "loss": 3.4728, "step": 4222 }, { "epoch": 0.51, "learning_rate": 5.3857461703720486e-06, "loss": 3.4271, "step": 4223 }, { "epoch": 0.51, "learning_rate": 5.383746248365823e-06, "loss": 3.3407, "step": 4224 }, { "epoch": 0.51, "learning_rate": 5.381746264599293e-06, "loss": 3.4514, "step": 4225 }, { "epoch": 0.51, "learning_rate": 5.379746219394337e-06, "loss": 3.4624, "step": 4226 }, { "epoch": 0.51, "learning_rate": 5.377746113072839e-06, "loss": 3.4386, "step": 4227 }, { "epoch": 0.51, "learning_rate": 5.375745945956703e-06, "loss": 3.5256, "step": 4228 }, { "epoch": 0.51, "learning_rate": 5.373745718367836e-06, "loss": 3.4261, "step": 4229 }, { "epoch": 0.51, "learning_rate": 5.37174543062815e-06, "loss": 3.3408, "step": 4230 }, { "epoch": 0.51, "learning_rate": 5.3697450830595775e-06, "loss": 3.3868, "step": 4231 }, { "epoch": 0.51, "learning_rate": 5.367744675984055e-06, "loss": 3.4469, "step": 4232 }, { "epoch": 0.51, "learning_rate": 5.365744209723526e-06, "loss": 3.4567, "step": 4233 }, { "epoch": 0.51, "learning_rate": 5.3637436845999505e-06, "loss": 3.4509, "step": 4234 }, { "epoch": 0.51, "learning_rate": 5.36174310093529e-06, "loss": 3.3544, "step": 4235 }, { "epoch": 0.51, "learning_rate": 5.359742459051522e-06, "loss": 3.3782, "step": 4236 }, { "epoch": 0.51, "learning_rate": 5.35774175927063e-06, "loss": 3.452, "step": 4237 }, { "epoch": 0.51, "learning_rate": 5.3557410019146094e-06, "loss": 3.2925, "step": 4238 }, { "epoch": 0.51, "learning_rate": 5.35374018730546e-06, "loss": 3.2898, "step": 4239 }, { "epoch": 0.51, "learning_rate": 5.351739315765194e-06, "loss": 3.395, "step": 4240 }, { "epoch": 0.51, "learning_rate": 5.349738387615835e-06, "loss": 3.3775, "step": 4241 }, { "epoch": 0.51, "learning_rate": 5.347737403179412e-06, "loss": 3.5445, "step": 4242 }, { "epoch": 0.51, "learning_rate": 5.345736362777963e-06, "loss": 3.4306, "step": 4243 }, { "epoch": 0.51, "learning_rate": 5.343735266733538e-06, "loss": 3.535, "step": 4244 }, { "epoch": 0.51, "learning_rate": 5.341734115368195e-06, "loss": 3.3882, "step": 4245 }, { "epoch": 0.51, "learning_rate": 5.339732909004e-06, "loss": 3.3768, "step": 4246 }, { "epoch": 0.51, "learning_rate": 5.337731647963029e-06, "loss": 3.5528, "step": 4247 }, { "epoch": 0.51, "learning_rate": 5.335730332567364e-06, "loss": 3.3303, "step": 4248 }, { "epoch": 0.51, "learning_rate": 5.333728963139097e-06, "loss": 3.4464, "step": 4249 }, { "epoch": 0.51, "learning_rate": 5.33172754000033e-06, "loss": 3.4621, "step": 4250 }, { "epoch": 0.51, "learning_rate": 5.329726063473174e-06, "loss": 3.3599, "step": 4251 }, { "epoch": 0.51, "learning_rate": 5.327724533879749e-06, "loss": 3.355, "step": 4252 }, { "epoch": 0.51, "learning_rate": 5.325722951542179e-06, "loss": 3.3807, "step": 4253 }, { "epoch": 0.51, "learning_rate": 5.323721316782599e-06, "loss": 3.3583, "step": 4254 }, { "epoch": 0.51, "learning_rate": 5.321719629923158e-06, "loss": 3.4796, "step": 4255 }, { "epoch": 0.51, "learning_rate": 5.319717891286005e-06, "loss": 3.3028, "step": 4256 }, { "epoch": 0.51, "learning_rate": 5.3177161011932995e-06, "loss": 3.3439, "step": 4257 }, { "epoch": 0.51, "learning_rate": 5.315714259967213e-06, "loss": 3.5297, "step": 4258 }, { "epoch": 0.51, "learning_rate": 5.313712367929922e-06, "loss": 3.3745, "step": 4259 }, { "epoch": 0.51, "learning_rate": 5.3117104254036115e-06, "loss": 3.351, "step": 4260 }, { "epoch": 0.51, "learning_rate": 5.309708432710476e-06, "loss": 3.4342, "step": 4261 }, { "epoch": 0.51, "learning_rate": 5.3077063901727165e-06, "loss": 3.5163, "step": 4262 }, { "epoch": 0.51, "learning_rate": 5.305704298112544e-06, "loss": 3.3718, "step": 4263 }, { "epoch": 0.51, "learning_rate": 5.303702156852174e-06, "loss": 3.301, "step": 4264 }, { "epoch": 0.51, "learning_rate": 5.301699966713835e-06, "loss": 3.3354, "step": 4265 }, { "epoch": 0.51, "learning_rate": 5.299697728019756e-06, "loss": 3.3954, "step": 4266 }, { "epoch": 0.51, "learning_rate": 5.297695441092182e-06, "loss": 3.4173, "step": 4267 }, { "epoch": 0.51, "learning_rate": 5.295693106253361e-06, "loss": 3.3843, "step": 4268 }, { "epoch": 0.51, "learning_rate": 5.29369072382555e-06, "loss": 3.4378, "step": 4269 }, { "epoch": 0.51, "learning_rate": 5.291688294131013e-06, "loss": 3.4816, "step": 4270 }, { "epoch": 0.51, "learning_rate": 5.289685817492022e-06, "loss": 3.3356, "step": 4271 }, { "epoch": 0.51, "learning_rate": 5.287683294230855e-06, "loss": 3.4221, "step": 4272 }, { "epoch": 0.51, "learning_rate": 5.2856807246698015e-06, "loss": 3.3212, "step": 4273 }, { "epoch": 0.51, "learning_rate": 5.283678109131157e-06, "loss": 3.4066, "step": 4274 }, { "epoch": 0.51, "learning_rate": 5.28167544793722e-06, "loss": 3.3555, "step": 4275 }, { "epoch": 0.51, "learning_rate": 5.2796727414103e-06, "loss": 3.41, "step": 4276 }, { "epoch": 0.51, "learning_rate": 5.277669989872716e-06, "loss": 3.4295, "step": 4277 }, { "epoch": 0.51, "learning_rate": 5.275667193646789e-06, "loss": 3.3467, "step": 4278 }, { "epoch": 0.51, "learning_rate": 5.2736643530548545e-06, "loss": 3.4559, "step": 4279 }, { "epoch": 0.51, "learning_rate": 5.271661468419247e-06, "loss": 3.3544, "step": 4280 }, { "epoch": 0.51, "learning_rate": 5.269658540062312e-06, "loss": 3.445, "step": 4281 }, { "epoch": 0.51, "learning_rate": 5.267655568306402e-06, "loss": 3.365, "step": 4282 }, { "epoch": 0.51, "learning_rate": 5.265652553473876e-06, "loss": 3.4445, "step": 4283 }, { "epoch": 0.51, "learning_rate": 5.2636494958871005e-06, "loss": 3.3763, "step": 4284 }, { "epoch": 0.51, "learning_rate": 5.26164639586845e-06, "loss": 3.3728, "step": 4285 }, { "epoch": 0.51, "learning_rate": 5.259643253740302e-06, "loss": 3.4262, "step": 4286 }, { "epoch": 0.51, "learning_rate": 5.257640069825044e-06, "loss": 3.4688, "step": 4287 }, { "epoch": 0.51, "learning_rate": 5.255636844445071e-06, "loss": 3.3855, "step": 4288 }, { "epoch": 0.51, "learning_rate": 5.253633577922781e-06, "loss": 3.3811, "step": 4289 }, { "epoch": 0.51, "learning_rate": 5.25163027058058e-06, "loss": 3.2733, "step": 4290 }, { "epoch": 0.52, "learning_rate": 5.249626922740883e-06, "loss": 3.3335, "step": 4291 }, { "epoch": 0.52, "learning_rate": 5.247623534726109e-06, "loss": 3.3604, "step": 4292 }, { "epoch": 0.52, "learning_rate": 5.245620106858686e-06, "loss": 3.3802, "step": 4293 }, { "epoch": 0.52, "learning_rate": 5.243616639461044e-06, "loss": 3.3888, "step": 4294 }, { "epoch": 0.52, "learning_rate": 5.241613132855623e-06, "loss": 3.4368, "step": 4295 }, { "epoch": 0.52, "learning_rate": 5.239609587364869e-06, "loss": 3.3535, "step": 4296 }, { "epoch": 0.52, "learning_rate": 5.237606003311234e-06, "loss": 3.3151, "step": 4297 }, { "epoch": 0.52, "learning_rate": 5.235602381017174e-06, "loss": 3.3993, "step": 4298 }, { "epoch": 0.52, "learning_rate": 5.233598720805153e-06, "loss": 3.3848, "step": 4299 }, { "epoch": 0.52, "learning_rate": 5.231595022997643e-06, "loss": 3.4589, "step": 4300 }, { "epoch": 0.52, "learning_rate": 5.229591287917118e-06, "loss": 3.3567, "step": 4301 }, { "epoch": 0.52, "learning_rate": 5.227587515886062e-06, "loss": 3.4689, "step": 4302 }, { "epoch": 0.52, "learning_rate": 5.225583707226961e-06, "loss": 3.408, "step": 4303 }, { "epoch": 0.52, "learning_rate": 5.22357986226231e-06, "loss": 3.4637, "step": 4304 }, { "epoch": 0.52, "learning_rate": 5.221575981314609e-06, "loss": 3.3415, "step": 4305 }, { "epoch": 0.52, "learning_rate": 5.219572064706364e-06, "loss": 3.3791, "step": 4306 }, { "epoch": 0.52, "learning_rate": 5.217568112760085e-06, "loss": 3.4502, "step": 4307 }, { "epoch": 0.52, "learning_rate": 5.2155641257982895e-06, "loss": 3.3957, "step": 4308 }, { "epoch": 0.52, "learning_rate": 5.213560104143499e-06, "loss": 3.4113, "step": 4309 }, { "epoch": 0.52, "learning_rate": 5.211556048118243e-06, "loss": 3.4563, "step": 4310 }, { "epoch": 0.52, "learning_rate": 5.209551958045057e-06, "loss": 3.3088, "step": 4311 }, { "epoch": 0.52, "learning_rate": 5.207547834246477e-06, "loss": 3.3074, "step": 4312 }, { "epoch": 0.52, "learning_rate": 5.20554367704505e-06, "loss": 3.3731, "step": 4313 }, { "epoch": 0.52, "learning_rate": 5.203539486763324e-06, "loss": 3.4376, "step": 4314 }, { "epoch": 0.52, "learning_rate": 5.2015352637238555e-06, "loss": 3.4029, "step": 4315 }, { "epoch": 0.52, "learning_rate": 5.199531008249205e-06, "loss": 3.3344, "step": 4316 }, { "epoch": 0.52, "learning_rate": 5.197526720661939e-06, "loss": 3.3673, "step": 4317 }, { "epoch": 0.52, "learning_rate": 5.195522401284626e-06, "loss": 3.352, "step": 4318 }, { "epoch": 0.52, "learning_rate": 5.193518050439846e-06, "loss": 3.3921, "step": 4319 }, { "epoch": 0.52, "learning_rate": 5.191513668450178e-06, "loss": 3.3757, "step": 4320 }, { "epoch": 0.52, "learning_rate": 5.189509255638207e-06, "loss": 3.4503, "step": 4321 }, { "epoch": 0.52, "learning_rate": 5.187504812326526e-06, "loss": 3.4404, "step": 4322 }, { "epoch": 0.52, "learning_rate": 5.185500338837732e-06, "loss": 3.422, "step": 4323 }, { "epoch": 0.52, "learning_rate": 5.183495835494422e-06, "loss": 3.4801, "step": 4324 }, { "epoch": 0.52, "learning_rate": 5.181491302619205e-06, "loss": 3.4734, "step": 4325 }, { "epoch": 0.52, "learning_rate": 5.17948674053469e-06, "loss": 3.4316, "step": 4326 }, { "epoch": 0.52, "learning_rate": 5.177482149563494e-06, "loss": 3.3735, "step": 4327 }, { "epoch": 0.52, "learning_rate": 5.175477530028237e-06, "loss": 3.4674, "step": 4328 }, { "epoch": 0.52, "learning_rate": 5.17347288225154e-06, "loss": 3.4012, "step": 4329 }, { "epoch": 0.52, "learning_rate": 5.171468206556035e-06, "loss": 3.528, "step": 4330 }, { "epoch": 0.52, "learning_rate": 5.1694635032643536e-06, "loss": 3.4254, "step": 4331 }, { "epoch": 0.52, "learning_rate": 5.167458772699135e-06, "loss": 3.2969, "step": 4332 }, { "epoch": 0.52, "learning_rate": 5.165454015183023e-06, "loss": 3.4836, "step": 4333 }, { "epoch": 0.52, "learning_rate": 5.163449231038661e-06, "loss": 3.3745, "step": 4334 }, { "epoch": 0.52, "learning_rate": 5.1614444205887015e-06, "loss": 3.4326, "step": 4335 }, { "epoch": 0.52, "learning_rate": 5.159439584155799e-06, "loss": 3.5164, "step": 4336 }, { "epoch": 0.52, "learning_rate": 5.1574347220626156e-06, "loss": 3.453, "step": 4337 }, { "epoch": 0.52, "learning_rate": 5.15542983463181e-06, "loss": 3.4589, "step": 4338 }, { "epoch": 0.52, "learning_rate": 5.1534249221860525e-06, "loss": 3.4645, "step": 4339 }, { "epoch": 0.52, "learning_rate": 5.151419985048014e-06, "loss": 3.3715, "step": 4340 }, { "epoch": 0.52, "learning_rate": 5.149415023540372e-06, "loss": 3.419, "step": 4341 }, { "epoch": 0.52, "learning_rate": 5.147410037985803e-06, "loss": 3.2977, "step": 4342 }, { "epoch": 0.52, "learning_rate": 5.145405028706992e-06, "loss": 3.4655, "step": 4343 }, { "epoch": 0.52, "learning_rate": 5.143399996026627e-06, "loss": 3.4951, "step": 4344 }, { "epoch": 0.52, "learning_rate": 5.141394940267397e-06, "loss": 3.3526, "step": 4345 }, { "epoch": 0.52, "learning_rate": 5.139389861751997e-06, "loss": 3.4331, "step": 4346 }, { "epoch": 0.52, "learning_rate": 5.137384760803128e-06, "loss": 3.4322, "step": 4347 }, { "epoch": 0.52, "learning_rate": 5.135379637743487e-06, "loss": 3.4139, "step": 4348 }, { "epoch": 0.52, "learning_rate": 5.133374492895784e-06, "loss": 3.3994, "step": 4349 }, { "epoch": 0.52, "learning_rate": 5.131369326582724e-06, "loss": 3.4099, "step": 4350 }, { "epoch": 0.52, "learning_rate": 5.1293641391270245e-06, "loss": 3.4135, "step": 4351 }, { "epoch": 0.52, "learning_rate": 5.127358930851397e-06, "loss": 3.3657, "step": 4352 }, { "epoch": 0.52, "learning_rate": 5.1253537020785615e-06, "loss": 3.3575, "step": 4353 }, { "epoch": 0.52, "learning_rate": 5.123348453131242e-06, "loss": 3.4734, "step": 4354 }, { "epoch": 0.52, "learning_rate": 5.121343184332163e-06, "loss": 3.4165, "step": 4355 }, { "epoch": 0.52, "learning_rate": 5.119337896004053e-06, "loss": 3.4134, "step": 4356 }, { "epoch": 0.52, "learning_rate": 5.117332588469646e-06, "loss": 3.2987, "step": 4357 }, { "epoch": 0.52, "learning_rate": 5.115327262051673e-06, "loss": 3.454, "step": 4358 }, { "epoch": 0.52, "learning_rate": 5.113321917072878e-06, "loss": 3.4488, "step": 4359 }, { "epoch": 0.52, "learning_rate": 5.111316553855999e-06, "loss": 3.437, "step": 4360 }, { "epoch": 0.52, "learning_rate": 5.10931117272378e-06, "loss": 3.4273, "step": 4361 }, { "epoch": 0.52, "learning_rate": 5.107305773998968e-06, "loss": 3.348, "step": 4362 }, { "epoch": 0.52, "learning_rate": 5.105300358004313e-06, "loss": 3.4615, "step": 4363 }, { "epoch": 0.52, "learning_rate": 5.1032949250625695e-06, "loss": 3.388, "step": 4364 }, { "epoch": 0.52, "learning_rate": 5.101289475496491e-06, "loss": 3.4113, "step": 4365 }, { "epoch": 0.52, "learning_rate": 5.099284009628833e-06, "loss": 3.4881, "step": 4366 }, { "epoch": 0.52, "learning_rate": 5.097278527782361e-06, "loss": 3.3689, "step": 4367 }, { "epoch": 0.52, "learning_rate": 5.095273030279836e-06, "loss": 3.341, "step": 4368 }, { "epoch": 0.52, "learning_rate": 5.093267517444025e-06, "loss": 3.4262, "step": 4369 }, { "epoch": 0.52, "learning_rate": 5.0912619895976935e-06, "loss": 3.4814, "step": 4370 }, { "epoch": 0.52, "learning_rate": 5.089256447063613e-06, "loss": 3.4558, "step": 4371 }, { "epoch": 0.52, "learning_rate": 5.087250890164558e-06, "loss": 3.324, "step": 4372 }, { "epoch": 0.52, "learning_rate": 5.085245319223302e-06, "loss": 3.4544, "step": 4373 }, { "epoch": 0.53, "learning_rate": 5.083239734562624e-06, "loss": 3.3538, "step": 4374 }, { "epoch": 0.53, "learning_rate": 5.081234136505303e-06, "loss": 3.4137, "step": 4375 }, { "epoch": 0.53, "learning_rate": 5.079228525374121e-06, "loss": 3.4465, "step": 4376 }, { "epoch": 0.53, "learning_rate": 5.077222901491863e-06, "loss": 3.3832, "step": 4377 }, { "epoch": 0.53, "learning_rate": 5.0752172651813146e-06, "loss": 3.4004, "step": 4378 }, { "epoch": 0.53, "learning_rate": 5.073211616765263e-06, "loss": 3.3276, "step": 4379 }, { "epoch": 0.53, "learning_rate": 5.071205956566497e-06, "loss": 3.3516, "step": 4380 }, { "epoch": 0.53, "learning_rate": 5.069200284907811e-06, "loss": 3.324, "step": 4381 }, { "epoch": 0.53, "learning_rate": 5.067194602112e-06, "loss": 3.3526, "step": 4382 }, { "epoch": 0.53, "learning_rate": 5.065188908501855e-06, "loss": 3.4015, "step": 4383 }, { "epoch": 0.53, "learning_rate": 5.063183204400177e-06, "loss": 3.3784, "step": 4384 }, { "epoch": 0.53, "learning_rate": 5.061177490129763e-06, "loss": 3.4517, "step": 4385 }, { "epoch": 0.53, "learning_rate": 5.0591717660134145e-06, "loss": 3.4695, "step": 4386 }, { "epoch": 0.53, "learning_rate": 5.057166032373935e-06, "loss": 3.438, "step": 4387 }, { "epoch": 0.53, "learning_rate": 5.055160289534126e-06, "loss": 3.4392, "step": 4388 }, { "epoch": 0.53, "learning_rate": 5.053154537816792e-06, "loss": 3.4711, "step": 4389 }, { "epoch": 0.53, "learning_rate": 5.051148777544743e-06, "loss": 3.3335, "step": 4390 }, { "epoch": 0.53, "learning_rate": 5.049143009040785e-06, "loss": 3.3834, "step": 4391 }, { "epoch": 0.53, "learning_rate": 5.047137232627727e-06, "loss": 3.3634, "step": 4392 }, { "epoch": 0.53, "learning_rate": 5.04513144862838e-06, "loss": 3.3412, "step": 4393 }, { "epoch": 0.53, "learning_rate": 5.043125657365556e-06, "loss": 3.2963, "step": 4394 }, { "epoch": 0.53, "learning_rate": 5.041119859162068e-06, "loss": 3.3778, "step": 4395 }, { "epoch": 0.53, "learning_rate": 5.039114054340731e-06, "loss": 3.3307, "step": 4396 }, { "epoch": 0.53, "learning_rate": 5.037108243224359e-06, "loss": 3.379, "step": 4397 }, { "epoch": 0.53, "learning_rate": 5.035102426135769e-06, "loss": 3.4793, "step": 4398 }, { "epoch": 0.53, "learning_rate": 5.033096603397776e-06, "loss": 3.4183, "step": 4399 }, { "epoch": 0.53, "learning_rate": 5.031090775333202e-06, "loss": 3.373, "step": 4400 }, { "epoch": 0.53, "learning_rate": 5.029084942264863e-06, "loss": 3.4372, "step": 4401 }, { "epoch": 0.53, "learning_rate": 5.027079104515579e-06, "loss": 3.361, "step": 4402 }, { "epoch": 0.53, "learning_rate": 5.02507326240817e-06, "loss": 3.437, "step": 4403 }, { "epoch": 0.53, "learning_rate": 5.023067416265459e-06, "loss": 3.4432, "step": 4404 }, { "epoch": 0.53, "learning_rate": 5.021061566410268e-06, "loss": 3.3639, "step": 4405 }, { "epoch": 0.53, "learning_rate": 5.019055713165417e-06, "loss": 3.4542, "step": 4406 }, { "epoch": 0.53, "learning_rate": 5.0170498568537304e-06, "loss": 3.5185, "step": 4407 }, { "epoch": 0.53, "learning_rate": 5.015043997798032e-06, "loss": 3.4348, "step": 4408 }, { "epoch": 0.53, "learning_rate": 5.013038136321145e-06, "loss": 3.5311, "step": 4409 }, { "epoch": 0.53, "learning_rate": 5.011032272745895e-06, "loss": 3.3469, "step": 4410 }, { "epoch": 0.53, "learning_rate": 5.009026407395105e-06, "loss": 3.4062, "step": 4411 }, { "epoch": 0.53, "learning_rate": 5.007020540591601e-06, "loss": 3.3014, "step": 4412 }, { "epoch": 0.53, "learning_rate": 5.0050146726582075e-06, "loss": 3.4785, "step": 4413 }, { "epoch": 0.53, "learning_rate": 5.00300880391775e-06, "loss": 3.4595, "step": 4414 }, { "epoch": 0.53, "learning_rate": 5.001002934693054e-06, "loss": 3.3942, "step": 4415 }, { "epoch": 0.53, "learning_rate": 4.9989970653069475e-06, "loss": 3.5443, "step": 4416 }, { "epoch": 0.53, "learning_rate": 4.996991196082252e-06, "loss": 3.407, "step": 4417 }, { "epoch": 0.53, "learning_rate": 4.994985327341795e-06, "loss": 3.4183, "step": 4418 }, { "epoch": 0.53, "learning_rate": 4.992979459408402e-06, "loss": 3.389, "step": 4419 }, { "epoch": 0.53, "learning_rate": 4.990973592604897e-06, "loss": 3.3955, "step": 4420 }, { "epoch": 0.53, "learning_rate": 4.988967727254107e-06, "loss": 3.3448, "step": 4421 }, { "epoch": 0.53, "learning_rate": 4.986961863678856e-06, "loss": 3.3969, "step": 4422 }, { "epoch": 0.53, "learning_rate": 4.984956002201969e-06, "loss": 3.4215, "step": 4423 }, { "epoch": 0.53, "learning_rate": 4.982950143146271e-06, "loss": 3.3755, "step": 4424 }, { "epoch": 0.53, "learning_rate": 4.980944286834586e-06, "loss": 3.4993, "step": 4425 }, { "epoch": 0.53, "learning_rate": 4.978938433589735e-06, "loss": 3.4054, "step": 4426 }, { "epoch": 0.53, "learning_rate": 4.976932583734543e-06, "loss": 3.4335, "step": 4427 }, { "epoch": 0.53, "learning_rate": 4.9749267375918316e-06, "loss": 3.3311, "step": 4428 }, { "epoch": 0.53, "learning_rate": 4.972920895484424e-06, "loss": 3.3703, "step": 4429 }, { "epoch": 0.53, "learning_rate": 4.970915057735139e-06, "loss": 3.4194, "step": 4430 }, { "epoch": 0.53, "learning_rate": 4.9689092246668e-06, "loss": 3.4143, "step": 4431 }, { "epoch": 0.53, "learning_rate": 4.966903396602224e-06, "loss": 3.4146, "step": 4432 }, { "epoch": 0.53, "learning_rate": 4.964897573864232e-06, "loss": 3.4374, "step": 4433 }, { "epoch": 0.53, "learning_rate": 4.962891756775644e-06, "loss": 3.3483, "step": 4434 }, { "epoch": 0.53, "learning_rate": 4.9608859456592715e-06, "loss": 3.3534, "step": 4435 }, { "epoch": 0.53, "learning_rate": 4.958880140837934e-06, "loss": 3.4853, "step": 4436 }, { "epoch": 0.53, "learning_rate": 4.956874342634446e-06, "loss": 3.3471, "step": 4437 }, { "epoch": 0.53, "learning_rate": 4.954868551371622e-06, "loss": 3.3961, "step": 4438 }, { "epoch": 0.53, "learning_rate": 4.952862767372275e-06, "loss": 3.3573, "step": 4439 }, { "epoch": 0.53, "learning_rate": 4.950856990959217e-06, "loss": 3.3346, "step": 4440 }, { "epoch": 0.53, "learning_rate": 4.948851222455259e-06, "loss": 3.3803, "step": 4441 }, { "epoch": 0.53, "learning_rate": 4.946845462183209e-06, "loss": 3.4341, "step": 4442 }, { "epoch": 0.53, "learning_rate": 4.9448397104658765e-06, "loss": 3.4059, "step": 4443 }, { "epoch": 0.53, "learning_rate": 4.942833967626067e-06, "loss": 3.4641, "step": 4444 }, { "epoch": 0.53, "learning_rate": 4.940828233986586e-06, "loss": 3.4118, "step": 4445 }, { "epoch": 0.53, "learning_rate": 4.9388225098702385e-06, "loss": 3.3893, "step": 4446 }, { "epoch": 0.53, "learning_rate": 4.936816795599825e-06, "loss": 3.2766, "step": 4447 }, { "epoch": 0.53, "learning_rate": 4.934811091498146e-06, "loss": 3.4598, "step": 4448 }, { "epoch": 0.53, "learning_rate": 4.932805397888002e-06, "loss": 3.3925, "step": 4449 }, { "epoch": 0.53, "learning_rate": 4.930799715092189e-06, "loss": 3.3715, "step": 4450 }, { "epoch": 0.53, "learning_rate": 4.928794043433503e-06, "loss": 3.4293, "step": 4451 }, { "epoch": 0.53, "learning_rate": 4.926788383234741e-06, "loss": 3.3476, "step": 4452 }, { "epoch": 0.53, "learning_rate": 4.924782734818688e-06, "loss": 3.3818, "step": 4453 }, { "epoch": 0.53, "learning_rate": 4.922777098508139e-06, "loss": 3.3619, "step": 4454 }, { "epoch": 0.53, "learning_rate": 4.92077147462588e-06, "loss": 3.2861, "step": 4455 }, { "epoch": 0.53, "learning_rate": 4.918765863494698e-06, "loss": 3.4466, "step": 4456 }, { "epoch": 0.53, "learning_rate": 4.916760265437378e-06, "loss": 3.528, "step": 4457 }, { "epoch": 0.54, "learning_rate": 4.9147546807766994e-06, "loss": 3.337, "step": 4458 }, { "epoch": 0.54, "learning_rate": 4.912749109835443e-06, "loss": 3.4275, "step": 4459 }, { "epoch": 0.54, "learning_rate": 4.910743552936387e-06, "loss": 3.4408, "step": 4460 }, { "epoch": 0.54, "learning_rate": 4.90873801040231e-06, "loss": 3.4632, "step": 4461 }, { "epoch": 0.54, "learning_rate": 4.906732482555978e-06, "loss": 3.432, "step": 4462 }, { "epoch": 0.54, "learning_rate": 4.904726969720166e-06, "loss": 3.3099, "step": 4463 }, { "epoch": 0.54, "learning_rate": 4.9027214722176405e-06, "loss": 3.4225, "step": 4464 }, { "epoch": 0.54, "learning_rate": 4.9007159903711674e-06, "loss": 3.3474, "step": 4465 }, { "epoch": 0.54, "learning_rate": 4.898710524503512e-06, "loss": 3.3211, "step": 4466 }, { "epoch": 0.54, "learning_rate": 4.896705074937432e-06, "loss": 3.3476, "step": 4467 }, { "epoch": 0.54, "learning_rate": 4.894699641995687e-06, "loss": 3.3213, "step": 4468 }, { "epoch": 0.54, "learning_rate": 4.892694226001033e-06, "loss": 3.4935, "step": 4469 }, { "epoch": 0.54, "learning_rate": 4.890688827276222e-06, "loss": 3.3224, "step": 4470 }, { "epoch": 0.54, "learning_rate": 4.888683446144003e-06, "loss": 3.4499, "step": 4471 }, { "epoch": 0.54, "learning_rate": 4.886678082927123e-06, "loss": 3.3142, "step": 4472 }, { "epoch": 0.54, "learning_rate": 4.884672737948328e-06, "loss": 3.4225, "step": 4473 }, { "epoch": 0.54, "learning_rate": 4.8826674115303565e-06, "loss": 3.359, "step": 4474 }, { "epoch": 0.54, "learning_rate": 4.880662103995948e-06, "loss": 3.3565, "step": 4475 }, { "epoch": 0.54, "learning_rate": 4.878656815667838e-06, "loss": 3.4429, "step": 4476 }, { "epoch": 0.54, "learning_rate": 4.876651546868759e-06, "loss": 3.3526, "step": 4477 }, { "epoch": 0.54, "learning_rate": 4.8746462979214384e-06, "loss": 3.3833, "step": 4478 }, { "epoch": 0.54, "learning_rate": 4.8726410691486055e-06, "loss": 3.3761, "step": 4479 }, { "epoch": 0.54, "learning_rate": 4.870635860872978e-06, "loss": 3.3616, "step": 4480 }, { "epoch": 0.54, "learning_rate": 4.8686306734172764e-06, "loss": 3.3016, "step": 4481 }, { "epoch": 0.54, "learning_rate": 4.866625507104218e-06, "loss": 3.3851, "step": 4482 }, { "epoch": 0.54, "learning_rate": 4.8646203622565145e-06, "loss": 3.3947, "step": 4483 }, { "epoch": 0.54, "learning_rate": 4.862615239196874e-06, "loss": 3.5027, "step": 4484 }, { "epoch": 0.54, "learning_rate": 4.8606101382480034e-06, "loss": 3.3577, "step": 4485 }, { "epoch": 0.54, "learning_rate": 4.858605059732604e-06, "loss": 3.4166, "step": 4486 }, { "epoch": 0.54, "learning_rate": 4.856600003973374e-06, "loss": 3.4697, "step": 4487 }, { "epoch": 0.54, "learning_rate": 4.854594971293009e-06, "loss": 3.4586, "step": 4488 }, { "epoch": 0.54, "learning_rate": 4.852589962014198e-06, "loss": 3.4109, "step": 4489 }, { "epoch": 0.54, "learning_rate": 4.85058497645963e-06, "loss": 3.4094, "step": 4490 }, { "epoch": 0.54, "learning_rate": 4.8485800149519864e-06, "loss": 3.4742, "step": 4491 }, { "epoch": 0.54, "learning_rate": 4.846575077813949e-06, "loss": 3.4356, "step": 4492 }, { "epoch": 0.54, "learning_rate": 4.844570165368191e-06, "loss": 3.3184, "step": 4493 }, { "epoch": 0.54, "learning_rate": 4.842565277937387e-06, "loss": 3.3412, "step": 4494 }, { "epoch": 0.54, "learning_rate": 4.840560415844201e-06, "loss": 3.3593, "step": 4495 }, { "epoch": 0.54, "learning_rate": 4.8385555794112985e-06, "loss": 3.4354, "step": 4496 }, { "epoch": 0.54, "learning_rate": 4.8365507689613414e-06, "loss": 3.3444, "step": 4497 }, { "epoch": 0.54, "learning_rate": 4.834545984816979e-06, "loss": 3.455, "step": 4498 }, { "epoch": 0.54, "learning_rate": 4.832541227300866e-06, "loss": 3.4237, "step": 4499 }, { "epoch": 0.54, "learning_rate": 4.830536496735647e-06, "loss": 3.4144, "step": 4500 }, { "epoch": 0.54, "learning_rate": 4.8285317934439664e-06, "loss": 3.4397, "step": 4501 }, { "epoch": 0.54, "learning_rate": 4.8265271177484615e-06, "loss": 3.4199, "step": 4502 }, { "epoch": 0.54, "learning_rate": 4.824522469971765e-06, "loss": 3.5082, "step": 4503 }, { "epoch": 0.54, "learning_rate": 4.822517850436506e-06, "loss": 3.3709, "step": 4504 }, { "epoch": 0.54, "learning_rate": 4.82051325946531e-06, "loss": 3.4924, "step": 4505 }, { "epoch": 0.54, "learning_rate": 4.818508697380797e-06, "loss": 3.4307, "step": 4506 }, { "epoch": 0.54, "learning_rate": 4.81650416450558e-06, "loss": 3.3997, "step": 4507 }, { "epoch": 0.54, "learning_rate": 4.814499661162271e-06, "loss": 3.3161, "step": 4508 }, { "epoch": 0.54, "learning_rate": 4.812495187673475e-06, "loss": 3.4274, "step": 4509 }, { "epoch": 0.54, "learning_rate": 4.810490744361794e-06, "loss": 3.3469, "step": 4510 }, { "epoch": 0.54, "learning_rate": 4.808486331549824e-06, "loss": 3.4002, "step": 4511 }, { "epoch": 0.54, "learning_rate": 4.806481949560155e-06, "loss": 3.4534, "step": 4512 }, { "epoch": 0.54, "learning_rate": 4.804477598715373e-06, "loss": 3.4252, "step": 4513 }, { "epoch": 0.54, "learning_rate": 4.802473279338062e-06, "loss": 3.3841, "step": 4514 }, { "epoch": 0.54, "learning_rate": 4.800468991750797e-06, "loss": 3.4679, "step": 4515 }, { "epoch": 0.54, "learning_rate": 4.798464736276147e-06, "loss": 3.5065, "step": 4516 }, { "epoch": 0.54, "learning_rate": 4.796460513236678e-06, "loss": 3.3669, "step": 4517 }, { "epoch": 0.54, "learning_rate": 4.794456322954953e-06, "loss": 3.4538, "step": 4518 }, { "epoch": 0.54, "learning_rate": 4.792452165753524e-06, "loss": 3.3716, "step": 4519 }, { "epoch": 0.54, "learning_rate": 4.790448041954945e-06, "loss": 3.3902, "step": 4520 }, { "epoch": 0.54, "learning_rate": 4.7884439518817575e-06, "loss": 3.3313, "step": 4521 }, { "epoch": 0.54, "learning_rate": 4.786439895856502e-06, "loss": 3.4771, "step": 4522 }, { "epoch": 0.54, "learning_rate": 4.784435874201713e-06, "loss": 3.2091, "step": 4523 }, { "epoch": 0.54, "learning_rate": 4.782431887239918e-06, "loss": 3.4085, "step": 4524 }, { "epoch": 0.54, "learning_rate": 4.780427935293639e-06, "loss": 3.3817, "step": 4525 }, { "epoch": 0.54, "learning_rate": 4.778424018685392e-06, "loss": 3.358, "step": 4526 }, { "epoch": 0.54, "learning_rate": 4.776420137737691e-06, "loss": 3.4231, "step": 4527 }, { "epoch": 0.54, "learning_rate": 4.77441629277304e-06, "loss": 3.4903, "step": 4528 }, { "epoch": 0.54, "learning_rate": 4.77241248411394e-06, "loss": 3.4537, "step": 4529 }, { "epoch": 0.54, "learning_rate": 4.7704087120828825e-06, "loss": 3.3789, "step": 4530 }, { "epoch": 0.54, "learning_rate": 4.768404977002358e-06, "loss": 3.3789, "step": 4531 }, { "epoch": 0.54, "learning_rate": 4.766401279194847e-06, "loss": 3.393, "step": 4532 }, { "epoch": 0.54, "learning_rate": 4.7643976189828285e-06, "loss": 3.3909, "step": 4533 }, { "epoch": 0.54, "learning_rate": 4.762393996688768e-06, "loss": 3.4294, "step": 4534 }, { "epoch": 0.54, "learning_rate": 4.760390412635132e-06, "loss": 3.3528, "step": 4535 }, { "epoch": 0.54, "learning_rate": 4.7583868671443776e-06, "loss": 3.4283, "step": 4536 }, { "epoch": 0.54, "learning_rate": 4.756383360538957e-06, "loss": 3.3284, "step": 4537 }, { "epoch": 0.54, "learning_rate": 4.754379893141315e-06, "loss": 3.4863, "step": 4538 }, { "epoch": 0.54, "learning_rate": 4.752376465273892e-06, "loss": 3.3577, "step": 4539 }, { "epoch": 0.54, "learning_rate": 4.750373077259118e-06, "loss": 3.3819, "step": 4540 }, { "epoch": 0.55, "learning_rate": 4.748369729419422e-06, "loss": 3.4287, "step": 4541 }, { "epoch": 0.55, "learning_rate": 4.746366422077222e-06, "loss": 3.4226, "step": 4542 }, { "epoch": 0.55, "learning_rate": 4.744363155554931e-06, "loss": 3.4197, "step": 4543 }, { "epoch": 0.55, "learning_rate": 4.742359930174957e-06, "loss": 3.4051, "step": 4544 }, { "epoch": 0.55, "learning_rate": 4.7403567462596994e-06, "loss": 3.3312, "step": 4545 }, { "epoch": 0.55, "learning_rate": 4.738353604131551e-06, "loss": 3.4828, "step": 4546 }, { "epoch": 0.55, "learning_rate": 4.7363505041129e-06, "loss": 3.4331, "step": 4547 }, { "epoch": 0.55, "learning_rate": 4.734347446526125e-06, "loss": 3.4226, "step": 4548 }, { "epoch": 0.55, "learning_rate": 4.732344431693599e-06, "loss": 3.3644, "step": 4549 }, { "epoch": 0.55, "learning_rate": 4.730341459937689e-06, "loss": 3.4736, "step": 4550 }, { "epoch": 0.55, "learning_rate": 4.7283385315807565e-06, "loss": 3.5098, "step": 4551 }, { "epoch": 0.55, "learning_rate": 4.726335646945148e-06, "loss": 3.4812, "step": 4552 }, { "epoch": 0.55, "learning_rate": 4.7243328063532115e-06, "loss": 3.3456, "step": 4553 }, { "epoch": 0.55, "learning_rate": 4.722330010127286e-06, "loss": 3.3757, "step": 4554 }, { "epoch": 0.55, "learning_rate": 4.720327258589702e-06, "loss": 3.3109, "step": 4555 }, { "epoch": 0.55, "learning_rate": 4.7183245520627816e-06, "loss": 3.4044, "step": 4556 }, { "epoch": 0.55, "learning_rate": 4.7163218908688445e-06, "loss": 3.4336, "step": 4557 }, { "epoch": 0.55, "learning_rate": 4.7143192753301985e-06, "loss": 3.5052, "step": 4558 }, { "epoch": 0.55, "learning_rate": 4.7123167057691446e-06, "loss": 3.3191, "step": 4559 }, { "epoch": 0.55, "learning_rate": 4.710314182507981e-06, "loss": 3.3684, "step": 4560 }, { "epoch": 0.55, "learning_rate": 4.70831170586899e-06, "loss": 3.3783, "step": 4561 }, { "epoch": 0.55, "learning_rate": 4.706309276174452e-06, "loss": 3.5553, "step": 4562 }, { "epoch": 0.55, "learning_rate": 4.70430689374664e-06, "loss": 3.3755, "step": 4563 }, { "epoch": 0.55, "learning_rate": 4.702304558907819e-06, "loss": 3.3635, "step": 4564 }, { "epoch": 0.55, "learning_rate": 4.700302271980245e-06, "loss": 3.3594, "step": 4565 }, { "epoch": 0.55, "learning_rate": 4.6983000332861675e-06, "loss": 3.5015, "step": 4566 }, { "epoch": 0.55, "learning_rate": 4.696297843147827e-06, "loss": 3.3683, "step": 4567 }, { "epoch": 0.55, "learning_rate": 4.694295701887457e-06, "loss": 3.3747, "step": 4568 }, { "epoch": 0.55, "learning_rate": 4.692293609827285e-06, "loss": 3.4209, "step": 4569 }, { "epoch": 0.55, "learning_rate": 4.690291567289525e-06, "loss": 3.3471, "step": 4570 }, { "epoch": 0.55, "learning_rate": 4.688289574596389e-06, "loss": 3.3701, "step": 4571 }, { "epoch": 0.55, "learning_rate": 4.68628763207008e-06, "loss": 3.4757, "step": 4572 }, { "epoch": 0.55, "learning_rate": 4.684285740032788e-06, "loss": 3.2999, "step": 4573 }, { "epoch": 0.55, "learning_rate": 4.682283898806702e-06, "loss": 3.4246, "step": 4574 }, { "epoch": 0.55, "learning_rate": 4.680282108713997e-06, "loss": 3.4472, "step": 4575 }, { "epoch": 0.55, "learning_rate": 4.678280370076843e-06, "loss": 3.3633, "step": 4576 }, { "epoch": 0.55, "learning_rate": 4.676278683217401e-06, "loss": 3.4332, "step": 4577 }, { "epoch": 0.55, "learning_rate": 4.674277048457824e-06, "loss": 3.3826, "step": 4578 }, { "epoch": 0.55, "learning_rate": 4.672275466120254e-06, "loss": 3.3667, "step": 4579 }, { "epoch": 0.55, "learning_rate": 4.670273936526828e-06, "loss": 3.4214, "step": 4580 }, { "epoch": 0.55, "learning_rate": 4.668272459999672e-06, "loss": 3.3772, "step": 4581 }, { "epoch": 0.55, "learning_rate": 4.666271036860906e-06, "loss": 3.3691, "step": 4582 }, { "epoch": 0.55, "learning_rate": 4.664269667432638e-06, "loss": 3.3438, "step": 4583 }, { "epoch": 0.55, "learning_rate": 4.662268352036972e-06, "loss": 3.4416, "step": 4584 }, { "epoch": 0.55, "learning_rate": 4.660267090996e-06, "loss": 3.3875, "step": 4585 }, { "epoch": 0.55, "learning_rate": 4.658265884631804e-06, "loss": 3.4207, "step": 4586 }, { "epoch": 0.55, "learning_rate": 4.656264733266463e-06, "loss": 3.4553, "step": 4587 }, { "epoch": 0.55, "learning_rate": 4.6542636372220384e-06, "loss": 3.3741, "step": 4588 }, { "epoch": 0.55, "learning_rate": 4.65226259682059e-06, "loss": 3.3455, "step": 4589 }, { "epoch": 0.55, "learning_rate": 4.650261612384167e-06, "loss": 3.4177, "step": 4590 }, { "epoch": 0.55, "learning_rate": 4.648260684234807e-06, "loss": 3.3629, "step": 4591 }, { "epoch": 0.55, "learning_rate": 4.646259812694542e-06, "loss": 3.4269, "step": 4592 }, { "epoch": 0.55, "learning_rate": 4.644258998085392e-06, "loss": 3.3484, "step": 4593 }, { "epoch": 0.55, "learning_rate": 4.64225824072937e-06, "loss": 3.4262, "step": 4594 }, { "epoch": 0.55, "learning_rate": 4.640257540948477e-06, "loss": 3.437, "step": 4595 }, { "epoch": 0.55, "learning_rate": 4.638256899064712e-06, "loss": 3.424, "step": 4596 }, { "epoch": 0.55, "learning_rate": 4.636256315400052e-06, "loss": 3.4346, "step": 4597 }, { "epoch": 0.55, "learning_rate": 4.634255790276475e-06, "loss": 3.3697, "step": 4598 }, { "epoch": 0.55, "learning_rate": 4.632255324015948e-06, "loss": 3.296, "step": 4599 }, { "epoch": 0.55, "learning_rate": 4.630254916940424e-06, "loss": 3.3445, "step": 4600 }, { "epoch": 0.55, "learning_rate": 4.628254569371851e-06, "loss": 3.3154, "step": 4601 }, { "epoch": 0.55, "learning_rate": 4.626254281632167e-06, "loss": 3.3886, "step": 4602 }, { "epoch": 0.55, "learning_rate": 4.624254054043297e-06, "loss": 3.4482, "step": 4603 }, { "epoch": 0.55, "learning_rate": 4.62225388692716e-06, "loss": 3.3724, "step": 4604 }, { "epoch": 0.55, "learning_rate": 4.620253780605666e-06, "loss": 3.5265, "step": 4605 }, { "epoch": 0.55, "learning_rate": 4.618253735400709e-06, "loss": 3.4013, "step": 4606 }, { "epoch": 0.55, "learning_rate": 4.6162537516341784e-06, "loss": 3.429, "step": 4607 }, { "epoch": 0.55, "learning_rate": 4.614253829627954e-06, "loss": 3.3645, "step": 4608 }, { "epoch": 0.55, "learning_rate": 4.612253969703902e-06, "loss": 3.3964, "step": 4609 }, { "epoch": 0.55, "learning_rate": 4.610254172183884e-06, "loss": 3.5321, "step": 4610 }, { "epoch": 0.55, "learning_rate": 4.6082544373897455e-06, "loss": 3.4022, "step": 4611 }, { "epoch": 0.55, "learning_rate": 4.606254765643326e-06, "loss": 3.3933, "step": 4612 }, { "epoch": 0.55, "learning_rate": 4.604255157266453e-06, "loss": 3.3789, "step": 4613 }, { "epoch": 0.55, "learning_rate": 4.602255612580947e-06, "loss": 3.3832, "step": 4614 }, { "epoch": 0.55, "learning_rate": 4.600256131908612e-06, "loss": 3.2837, "step": 4615 }, { "epoch": 0.55, "learning_rate": 4.598256715571247e-06, "loss": 3.4294, "step": 4616 }, { "epoch": 0.55, "learning_rate": 4.596257363890638e-06, "loss": 3.4306, "step": 4617 }, { "epoch": 0.55, "learning_rate": 4.594258077188563e-06, "loss": 3.4668, "step": 4618 }, { "epoch": 0.55, "learning_rate": 4.592258855786788e-06, "loss": 3.3492, "step": 4619 }, { "epoch": 0.55, "learning_rate": 4.590259700007068e-06, "loss": 3.3978, "step": 4620 }, { "epoch": 0.55, "learning_rate": 4.588260610171149e-06, "loss": 3.433, "step": 4621 }, { "epoch": 0.55, "learning_rate": 4.586261586600764e-06, "loss": 3.4859, "step": 4622 }, { "epoch": 0.55, "learning_rate": 4.58426262961764e-06, "loss": 3.3503, "step": 4623 }, { "epoch": 0.56, "learning_rate": 4.582263739543486e-06, "loss": 3.4104, "step": 4624 }, { "epoch": 0.56, "learning_rate": 4.580264916700006e-06, "loss": 3.3693, "step": 4625 }, { "epoch": 0.56, "learning_rate": 4.578266161408893e-06, "loss": 3.4286, "step": 4626 }, { "epoch": 0.56, "learning_rate": 4.5762674739918245e-06, "loss": 3.3579, "step": 4627 }, { "epoch": 0.56, "learning_rate": 4.5742688547704725e-06, "loss": 3.431, "step": 4628 }, { "epoch": 0.56, "learning_rate": 4.5722703040664964e-06, "loss": 3.3791, "step": 4629 }, { "epoch": 0.56, "learning_rate": 4.570271822201542e-06, "loss": 3.418, "step": 4630 }, { "epoch": 0.56, "learning_rate": 4.568273409497246e-06, "loss": 3.4563, "step": 4631 }, { "epoch": 0.56, "learning_rate": 4.566275066275238e-06, "loss": 3.3744, "step": 4632 }, { "epoch": 0.56, "learning_rate": 4.564276792857126e-06, "loss": 3.4104, "step": 4633 }, { "epoch": 0.56, "learning_rate": 4.562278589564517e-06, "loss": 3.3898, "step": 4634 }, { "epoch": 0.56, "learning_rate": 4.560280456719001e-06, "loss": 3.4625, "step": 4635 }, { "epoch": 0.56, "learning_rate": 4.558282394642159e-06, "loss": 3.4619, "step": 4636 }, { "epoch": 0.56, "learning_rate": 4.5562844036555605e-06, "loss": 3.385, "step": 4637 }, { "epoch": 0.56, "learning_rate": 4.554286484080763e-06, "loss": 3.4384, "step": 4638 }, { "epoch": 0.56, "learning_rate": 4.5522886362393125e-06, "loss": 3.4316, "step": 4639 }, { "epoch": 0.56, "learning_rate": 4.550290860452744e-06, "loss": 3.372, "step": 4640 }, { "epoch": 0.56, "learning_rate": 4.548293157042581e-06, "loss": 3.357, "step": 4641 }, { "epoch": 0.56, "learning_rate": 4.546295526330334e-06, "loss": 3.4898, "step": 4642 }, { "epoch": 0.56, "learning_rate": 4.544297968637501e-06, "loss": 3.3565, "step": 4643 }, { "epoch": 0.56, "learning_rate": 4.542300484285572e-06, "loss": 3.4592, "step": 4644 }, { "epoch": 0.56, "learning_rate": 4.540303073596023e-06, "loss": 3.2907, "step": 4645 }, { "epoch": 0.56, "learning_rate": 4.538305736890318e-06, "loss": 3.3813, "step": 4646 }, { "epoch": 0.56, "learning_rate": 4.536308474489909e-06, "loss": 3.4373, "step": 4647 }, { "epoch": 0.56, "learning_rate": 4.534311286716236e-06, "loss": 3.3685, "step": 4648 }, { "epoch": 0.56, "learning_rate": 4.532314173890729e-06, "loss": 3.3971, "step": 4649 }, { "epoch": 0.56, "learning_rate": 4.530317136334805e-06, "loss": 3.4086, "step": 4650 }, { "epoch": 0.56, "learning_rate": 4.528320174369864e-06, "loss": 3.3976, "step": 4651 }, { "epoch": 0.56, "learning_rate": 4.5263232883173e-06, "loss": 3.4144, "step": 4652 }, { "epoch": 0.56, "learning_rate": 4.524326478498494e-06, "loss": 3.4041, "step": 4653 }, { "epoch": 0.56, "learning_rate": 4.522329745234812e-06, "loss": 3.5214, "step": 4654 }, { "epoch": 0.56, "learning_rate": 4.52033308884761e-06, "loss": 3.3167, "step": 4655 }, { "epoch": 0.56, "learning_rate": 4.518336509658232e-06, "loss": 3.4756, "step": 4656 }, { "epoch": 0.56, "learning_rate": 4.516340007988006e-06, "loss": 3.481, "step": 4657 }, { "epoch": 0.56, "learning_rate": 4.514343584158252e-06, "loss": 3.4287, "step": 4658 }, { "epoch": 0.56, "learning_rate": 4.512347238490277e-06, "loss": 3.3373, "step": 4659 }, { "epoch": 0.56, "learning_rate": 4.510350971305368e-06, "loss": 3.4777, "step": 4660 }, { "epoch": 0.56, "learning_rate": 4.50835478292481e-06, "loss": 3.3904, "step": 4661 }, { "epoch": 0.56, "learning_rate": 4.506358673669868e-06, "loss": 3.4266, "step": 4662 }, { "epoch": 0.56, "learning_rate": 4.504362643861798e-06, "loss": 3.3782, "step": 4663 }, { "epoch": 0.56, "learning_rate": 4.502366693821843e-06, "loss": 3.461, "step": 4664 }, { "epoch": 0.56, "learning_rate": 4.50037082387123e-06, "loss": 3.3436, "step": 4665 }, { "epoch": 0.56, "learning_rate": 4.498375034331176e-06, "loss": 3.429, "step": 4666 }, { "epoch": 0.56, "learning_rate": 4.496379325522885e-06, "loss": 3.3782, "step": 4667 }, { "epoch": 0.56, "learning_rate": 4.494383697767548e-06, "loss": 3.4115, "step": 4668 }, { "epoch": 0.56, "learning_rate": 4.492388151386341e-06, "loss": 3.3889, "step": 4669 }, { "epoch": 0.56, "learning_rate": 4.490392686700428e-06, "loss": 3.3875, "step": 4670 }, { "epoch": 0.56, "learning_rate": 4.48839730403096e-06, "loss": 3.3617, "step": 4671 }, { "epoch": 0.56, "learning_rate": 4.4864020036990755e-06, "loss": 3.4813, "step": 4672 }, { "epoch": 0.56, "learning_rate": 4.484406786025899e-06, "loss": 3.392, "step": 4673 }, { "epoch": 0.56, "learning_rate": 4.482411651332542e-06, "loss": 3.3126, "step": 4674 }, { "epoch": 0.56, "learning_rate": 4.4804165999401025e-06, "loss": 3.4179, "step": 4675 }, { "epoch": 0.56, "learning_rate": 4.478421632169664e-06, "loss": 3.3977, "step": 4676 }, { "epoch": 0.56, "learning_rate": 4.4764267483423005e-06, "loss": 3.3565, "step": 4677 }, { "epoch": 0.56, "learning_rate": 4.4744319487790655e-06, "loss": 3.4989, "step": 4678 }, { "epoch": 0.56, "learning_rate": 4.472437233801005e-06, "loss": 3.4055, "step": 4679 }, { "epoch": 0.56, "learning_rate": 4.470442603729148e-06, "loss": 3.4861, "step": 4680 }, { "epoch": 0.56, "learning_rate": 4.468448058884513e-06, "loss": 3.4487, "step": 4681 }, { "epoch": 0.56, "learning_rate": 4.466453599588103e-06, "loss": 3.4809, "step": 4682 }, { "epoch": 0.56, "learning_rate": 4.464459226160906e-06, "loss": 3.3312, "step": 4683 }, { "epoch": 0.56, "learning_rate": 4.462464938923898e-06, "loss": 3.4448, "step": 4684 }, { "epoch": 0.56, "learning_rate": 4.460470738198042e-06, "loss": 3.3728, "step": 4685 }, { "epoch": 0.56, "learning_rate": 4.4584766243042835e-06, "loss": 3.3326, "step": 4686 }, { "epoch": 0.56, "learning_rate": 4.456482597563556e-06, "loss": 3.3305, "step": 4687 }, { "epoch": 0.56, "learning_rate": 4.454488658296781e-06, "loss": 3.3906, "step": 4688 }, { "epoch": 0.56, "learning_rate": 4.452494806824863e-06, "loss": 3.2774, "step": 4689 }, { "epoch": 0.56, "learning_rate": 4.4505010434686935e-06, "loss": 3.3238, "step": 4690 }, { "epoch": 0.56, "learning_rate": 4.4485073685491496e-06, "loss": 3.3878, "step": 4691 }, { "epoch": 0.56, "learning_rate": 4.446513782387093e-06, "loss": 3.3373, "step": 4692 }, { "epoch": 0.56, "learning_rate": 4.444520285303375e-06, "loss": 3.3515, "step": 4693 }, { "epoch": 0.56, "learning_rate": 4.442526877618829e-06, "loss": 3.355, "step": 4694 }, { "epoch": 0.56, "learning_rate": 4.440533559654277e-06, "loss": 3.4498, "step": 4695 }, { "epoch": 0.56, "learning_rate": 4.43854033173052e-06, "loss": 3.4356, "step": 4696 }, { "epoch": 0.56, "learning_rate": 4.436547194168351e-06, "loss": 3.4468, "step": 4697 }, { "epoch": 0.56, "learning_rate": 4.4345541472885465e-06, "loss": 3.5134, "step": 4698 }, { "epoch": 0.56, "learning_rate": 4.432561191411869e-06, "loss": 3.3766, "step": 4699 }, { "epoch": 0.56, "learning_rate": 4.430568326859067e-06, "loss": 3.3575, "step": 4700 }, { "epoch": 0.56, "learning_rate": 4.42857555395087e-06, "loss": 3.3834, "step": 4701 }, { "epoch": 0.56, "learning_rate": 4.426582873007999e-06, "loss": 3.3652, "step": 4702 }, { "epoch": 0.56, "learning_rate": 4.424590284351154e-06, "loss": 3.4454, "step": 4703 }, { "epoch": 0.56, "learning_rate": 4.422597788301026e-06, "loss": 3.4153, "step": 4704 }, { "epoch": 0.56, "learning_rate": 4.420605385178287e-06, "loss": 3.4128, "step": 4705 }, { "epoch": 0.56, "learning_rate": 4.418613075303594e-06, "loss": 3.4299, "step": 4706 }, { "epoch": 0.56, "learning_rate": 4.4166208589975925e-06, "loss": 3.4055, "step": 4707 }, { "epoch": 0.57, "learning_rate": 4.41462873658091e-06, "loss": 3.3991, "step": 4708 }, { "epoch": 0.57, "learning_rate": 4.412636708374159e-06, "loss": 3.2884, "step": 4709 }, { "epoch": 0.57, "learning_rate": 4.410644774697937e-06, "loss": 3.4281, "step": 4710 }, { "epoch": 0.57, "learning_rate": 4.408652935872829e-06, "loss": 3.4339, "step": 4711 }, { "epoch": 0.57, "learning_rate": 4.4066611922194005e-06, "loss": 3.3581, "step": 4712 }, { "epoch": 0.57, "learning_rate": 4.404669544058207e-06, "loss": 3.454, "step": 4713 }, { "epoch": 0.57, "learning_rate": 4.402677991709781e-06, "loss": 3.441, "step": 4714 }, { "epoch": 0.57, "learning_rate": 4.400686535494645e-06, "loss": 3.4526, "step": 4715 }, { "epoch": 0.57, "learning_rate": 4.398695175733306e-06, "loss": 3.3269, "step": 4716 }, { "epoch": 0.57, "learning_rate": 4.396703912746253e-06, "loss": 3.3786, "step": 4717 }, { "epoch": 0.57, "learning_rate": 4.3947127468539616e-06, "loss": 3.4152, "step": 4718 }, { "epoch": 0.57, "learning_rate": 4.392721678376892e-06, "loss": 3.3472, "step": 4719 }, { "epoch": 0.57, "learning_rate": 4.390730707635486e-06, "loss": 3.5412, "step": 4720 }, { "epoch": 0.57, "learning_rate": 4.388739834950174e-06, "loss": 3.3634, "step": 4721 }, { "epoch": 0.57, "learning_rate": 4.386749060641365e-06, "loss": 3.4125, "step": 4722 }, { "epoch": 0.57, "learning_rate": 4.384758385029457e-06, "loss": 3.4519, "step": 4723 }, { "epoch": 0.57, "learning_rate": 4.382767808434829e-06, "loss": 3.3186, "step": 4724 }, { "epoch": 0.57, "learning_rate": 4.380777331177846e-06, "loss": 3.3849, "step": 4725 }, { "epoch": 0.57, "learning_rate": 4.378786953578857e-06, "loss": 3.4634, "step": 4726 }, { "epoch": 0.57, "learning_rate": 4.3767966759581925e-06, "loss": 3.3769, "step": 4727 }, { "epoch": 0.57, "learning_rate": 4.374806498636171e-06, "loss": 3.4205, "step": 4728 }, { "epoch": 0.57, "learning_rate": 4.3728164219330915e-06, "loss": 3.4429, "step": 4729 }, { "epoch": 0.57, "learning_rate": 4.370826446169237e-06, "loss": 3.3795, "step": 4730 }, { "epoch": 0.57, "learning_rate": 4.3688365716648796e-06, "loss": 3.3924, "step": 4731 }, { "epoch": 0.57, "learning_rate": 4.3668467987402644e-06, "loss": 3.438, "step": 4732 }, { "epoch": 0.57, "learning_rate": 4.36485712771563e-06, "loss": 3.364, "step": 4733 }, { "epoch": 0.57, "learning_rate": 4.362867558911193e-06, "loss": 3.4463, "step": 4734 }, { "epoch": 0.57, "learning_rate": 4.360878092647156e-06, "loss": 3.4162, "step": 4735 }, { "epoch": 0.57, "learning_rate": 4.358888729243707e-06, "loss": 3.4301, "step": 4736 }, { "epoch": 0.57, "learning_rate": 4.3568994690210134e-06, "loss": 3.3638, "step": 4737 }, { "epoch": 0.57, "learning_rate": 4.354910312299227e-06, "loss": 3.3161, "step": 4738 }, { "epoch": 0.57, "learning_rate": 4.3529212593984845e-06, "loss": 3.4842, "step": 4739 }, { "epoch": 0.57, "learning_rate": 4.350932310638907e-06, "loss": 3.3973, "step": 4740 }, { "epoch": 0.57, "learning_rate": 4.348943466340593e-06, "loss": 3.3284, "step": 4741 }, { "epoch": 0.57, "learning_rate": 4.34695472682363e-06, "loss": 3.4346, "step": 4742 }, { "epoch": 0.57, "learning_rate": 4.344966092408086e-06, "loss": 3.5177, "step": 4743 }, { "epoch": 0.57, "learning_rate": 4.342977563414013e-06, "loss": 3.371, "step": 4744 }, { "epoch": 0.57, "learning_rate": 4.340989140161448e-06, "loss": 3.495, "step": 4745 }, { "epoch": 0.57, "learning_rate": 4.339000822970406e-06, "loss": 3.3623, "step": 4746 }, { "epoch": 0.57, "learning_rate": 4.3370126121608886e-06, "loss": 3.3996, "step": 4747 }, { "epoch": 0.57, "learning_rate": 4.33502450805288e-06, "loss": 3.3687, "step": 4748 }, { "epoch": 0.57, "learning_rate": 4.333036510966347e-06, "loss": 3.39, "step": 4749 }, { "epoch": 0.57, "learning_rate": 4.331048621221236e-06, "loss": 3.3382, "step": 4750 }, { "epoch": 0.57, "learning_rate": 4.329060839137481e-06, "loss": 3.4493, "step": 4751 }, { "epoch": 0.57, "learning_rate": 4.327073165034997e-06, "loss": 3.529, "step": 4752 }, { "epoch": 0.57, "learning_rate": 4.3250855992336806e-06, "loss": 3.3875, "step": 4753 }, { "epoch": 0.57, "learning_rate": 4.323098142053411e-06, "loss": 3.3474, "step": 4754 }, { "epoch": 0.57, "learning_rate": 4.3211107938140525e-06, "loss": 3.3708, "step": 4755 }, { "epoch": 0.57, "learning_rate": 4.319123554835448e-06, "loss": 3.3833, "step": 4756 }, { "epoch": 0.57, "learning_rate": 4.317136425437425e-06, "loss": 3.3649, "step": 4757 }, { "epoch": 0.57, "learning_rate": 4.315149405939795e-06, "loss": 3.444, "step": 4758 }, { "epoch": 0.57, "learning_rate": 4.313162496662347e-06, "loss": 3.4892, "step": 4759 }, { "epoch": 0.57, "learning_rate": 4.3111756979248555e-06, "loss": 3.3689, "step": 4760 }, { "epoch": 0.57, "learning_rate": 4.309189010047078e-06, "loss": 3.3193, "step": 4761 }, { "epoch": 0.57, "learning_rate": 4.307202433348753e-06, "loss": 3.4608, "step": 4762 }, { "epoch": 0.57, "learning_rate": 4.3052159681496e-06, "loss": 3.4269, "step": 4763 }, { "epoch": 0.57, "learning_rate": 4.3032296147693225e-06, "loss": 3.4178, "step": 4764 }, { "epoch": 0.57, "learning_rate": 4.301243373527605e-06, "loss": 3.3709, "step": 4765 }, { "epoch": 0.57, "learning_rate": 4.299257244744115e-06, "loss": 3.3876, "step": 4766 }, { "epoch": 0.57, "learning_rate": 4.297271228738501e-06, "loss": 3.4038, "step": 4767 }, { "epoch": 0.57, "learning_rate": 4.2952853258303915e-06, "loss": 3.4339, "step": 4768 }, { "epoch": 0.57, "learning_rate": 4.293299536339399e-06, "loss": 3.4071, "step": 4769 }, { "epoch": 0.57, "learning_rate": 4.29131386058512e-06, "loss": 3.5508, "step": 4770 }, { "epoch": 0.57, "learning_rate": 4.289328298887127e-06, "loss": 3.3459, "step": 4771 }, { "epoch": 0.57, "learning_rate": 4.287342851564979e-06, "loss": 3.4098, "step": 4772 }, { "epoch": 0.57, "learning_rate": 4.2853575189382134e-06, "loss": 3.4808, "step": 4773 }, { "epoch": 0.57, "learning_rate": 4.283372301326352e-06, "loss": 3.4245, "step": 4774 }, { "epoch": 0.57, "learning_rate": 4.281387199048897e-06, "loss": 3.4492, "step": 4775 }, { "epoch": 0.57, "learning_rate": 4.279402212425332e-06, "loss": 3.3335, "step": 4776 }, { "epoch": 0.57, "learning_rate": 4.277417341775119e-06, "loss": 3.4288, "step": 4777 }, { "epoch": 0.57, "learning_rate": 4.275432587417706e-06, "loss": 3.4147, "step": 4778 }, { "epoch": 0.57, "learning_rate": 4.273447949672519e-06, "loss": 3.4438, "step": 4779 }, { "epoch": 0.57, "learning_rate": 4.271463428858968e-06, "loss": 3.3667, "step": 4780 }, { "epoch": 0.57, "learning_rate": 4.269479025296444e-06, "loss": 3.4449, "step": 4781 }, { "epoch": 0.57, "learning_rate": 4.267494739304314e-06, "loss": 3.4506, "step": 4782 }, { "epoch": 0.57, "learning_rate": 4.265510571201933e-06, "loss": 3.4032, "step": 4783 }, { "epoch": 0.57, "learning_rate": 4.263526521308633e-06, "loss": 3.4011, "step": 4784 }, { "epoch": 0.57, "learning_rate": 4.2615425899437285e-06, "loss": 3.3338, "step": 4785 }, { "epoch": 0.57, "learning_rate": 4.259558777426514e-06, "loss": 3.3362, "step": 4786 }, { "epoch": 0.57, "learning_rate": 4.257575084076264e-06, "loss": 3.3478, "step": 4787 }, { "epoch": 0.57, "learning_rate": 4.255591510212235e-06, "loss": 3.3439, "step": 4788 }, { "epoch": 0.57, "learning_rate": 4.253608056153668e-06, "loss": 3.4405, "step": 4789 }, { "epoch": 0.57, "learning_rate": 4.251624722219778e-06, "loss": 3.3232, "step": 4790 }, { "epoch": 0.58, "learning_rate": 4.249641508729763e-06, "loss": 3.3768, "step": 4791 }, { "epoch": 0.58, "learning_rate": 4.2476584160028045e-06, "loss": 3.3737, "step": 4792 }, { "epoch": 0.58, "learning_rate": 4.245675444358061e-06, "loss": 3.2983, "step": 4793 }, { "epoch": 0.58, "learning_rate": 4.243692594114676e-06, "loss": 3.4182, "step": 4794 }, { "epoch": 0.58, "learning_rate": 4.241709865591765e-06, "loss": 3.4031, "step": 4795 }, { "epoch": 0.58, "learning_rate": 4.239727259108433e-06, "loss": 3.3535, "step": 4796 }, { "epoch": 0.58, "learning_rate": 4.23774477498376e-06, "loss": 3.4039, "step": 4797 }, { "epoch": 0.58, "learning_rate": 4.235762413536809e-06, "loss": 3.3605, "step": 4798 }, { "epoch": 0.58, "learning_rate": 4.233780175086622e-06, "loss": 3.4321, "step": 4799 }, { "epoch": 0.58, "learning_rate": 4.23179805995222e-06, "loss": 3.352, "step": 4800 }, { "epoch": 0.58, "learning_rate": 4.229816068452608e-06, "loss": 3.2877, "step": 4801 }, { "epoch": 0.58, "learning_rate": 4.2278342009067665e-06, "loss": 3.372, "step": 4802 }, { "epoch": 0.58, "learning_rate": 4.225852457633661e-06, "loss": 3.4519, "step": 4803 }, { "epoch": 0.58, "learning_rate": 4.223870838952231e-06, "loss": 3.3611, "step": 4804 }, { "epoch": 0.58, "learning_rate": 4.2218893451814005e-06, "loss": 3.3814, "step": 4805 }, { "epoch": 0.58, "learning_rate": 4.2199079766400725e-06, "loss": 3.4362, "step": 4806 }, { "epoch": 0.58, "learning_rate": 4.217926733647128e-06, "loss": 3.4425, "step": 4807 }, { "epoch": 0.58, "learning_rate": 4.215945616521432e-06, "loss": 3.3989, "step": 4808 }, { "epoch": 0.58, "learning_rate": 4.213964625581823e-06, "loss": 3.4182, "step": 4809 }, { "epoch": 0.58, "learning_rate": 4.211983761147125e-06, "loss": 3.439, "step": 4810 }, { "epoch": 0.58, "learning_rate": 4.210003023536138e-06, "loss": 3.4283, "step": 4811 }, { "epoch": 0.58, "learning_rate": 4.208022413067644e-06, "loss": 3.4556, "step": 4812 }, { "epoch": 0.58, "learning_rate": 4.206041930060405e-06, "loss": 3.3632, "step": 4813 }, { "epoch": 0.58, "learning_rate": 4.204061574833156e-06, "loss": 3.3696, "step": 4814 }, { "epoch": 0.58, "learning_rate": 4.20208134770462e-06, "loss": 3.3067, "step": 4815 }, { "epoch": 0.58, "learning_rate": 4.2001012489934935e-06, "loss": 3.3472, "step": 4816 }, { "epoch": 0.58, "learning_rate": 4.198121279018456e-06, "loss": 3.4005, "step": 4817 }, { "epoch": 0.58, "learning_rate": 4.1961414380981645e-06, "loss": 3.3841, "step": 4818 }, { "epoch": 0.58, "learning_rate": 4.1941617265512565e-06, "loss": 3.4098, "step": 4819 }, { "epoch": 0.58, "learning_rate": 4.1921821446963465e-06, "loss": 3.4098, "step": 4820 }, { "epoch": 0.58, "learning_rate": 4.190202692852029e-06, "loss": 3.3993, "step": 4821 }, { "epoch": 0.58, "learning_rate": 4.18822337133688e-06, "loss": 3.3865, "step": 4822 }, { "epoch": 0.58, "learning_rate": 4.186244180469451e-06, "loss": 3.3354, "step": 4823 }, { "epoch": 0.58, "learning_rate": 4.184265120568272e-06, "loss": 3.3915, "step": 4824 }, { "epoch": 0.58, "learning_rate": 4.182286191951858e-06, "loss": 3.4117, "step": 4825 }, { "epoch": 0.58, "learning_rate": 4.180307394938695e-06, "loss": 3.4498, "step": 4826 }, { "epoch": 0.58, "learning_rate": 4.178328729847254e-06, "loss": 3.389, "step": 4827 }, { "epoch": 0.58, "learning_rate": 4.176350196995981e-06, "loss": 3.3889, "step": 4828 }, { "epoch": 0.58, "learning_rate": 4.1743717967033024e-06, "loss": 3.4867, "step": 4829 }, { "epoch": 0.58, "learning_rate": 4.172393529287623e-06, "loss": 3.454, "step": 4830 }, { "epoch": 0.58, "learning_rate": 4.170415395067329e-06, "loss": 3.4967, "step": 4831 }, { "epoch": 0.58, "learning_rate": 4.1684373943607756e-06, "loss": 3.4571, "step": 4832 }, { "epoch": 0.58, "learning_rate": 4.166459527486309e-06, "loss": 3.4259, "step": 4833 }, { "epoch": 0.58, "learning_rate": 4.164481794762244e-06, "loss": 3.3194, "step": 4834 }, { "epoch": 0.58, "learning_rate": 4.1625041965068805e-06, "loss": 3.4324, "step": 4835 }, { "epoch": 0.58, "learning_rate": 4.160526733038494e-06, "loss": 3.4324, "step": 4836 }, { "epoch": 0.58, "learning_rate": 4.158549404675336e-06, "loss": 3.373, "step": 4837 }, { "epoch": 0.58, "learning_rate": 4.156572211735642e-06, "loss": 3.4454, "step": 4838 }, { "epoch": 0.58, "learning_rate": 4.15459515453762e-06, "loss": 3.3718, "step": 4839 }, { "epoch": 0.58, "learning_rate": 4.152618233399463e-06, "loss": 3.4627, "step": 4840 }, { "epoch": 0.58, "learning_rate": 4.150641448639331e-06, "loss": 3.3161, "step": 4841 }, { "epoch": 0.58, "learning_rate": 4.148664800575372e-06, "loss": 3.3934, "step": 4842 }, { "epoch": 0.58, "learning_rate": 4.146688289525708e-06, "loss": 3.5278, "step": 4843 }, { "epoch": 0.58, "learning_rate": 4.144711915808441e-06, "loss": 3.3566, "step": 4844 }, { "epoch": 0.58, "learning_rate": 4.142735679741648e-06, "loss": 3.3592, "step": 4845 }, { "epoch": 0.58, "learning_rate": 4.140759581643386e-06, "loss": 3.4271, "step": 4846 }, { "epoch": 0.58, "learning_rate": 4.138783621831689e-06, "loss": 3.3153, "step": 4847 }, { "epoch": 0.58, "learning_rate": 4.136807800624569e-06, "loss": 3.404, "step": 4848 }, { "epoch": 0.58, "learning_rate": 4.134832118340017e-06, "loss": 3.3655, "step": 4849 }, { "epoch": 0.58, "learning_rate": 4.132856575295999e-06, "loss": 3.4808, "step": 4850 }, { "epoch": 0.58, "learning_rate": 4.130881171810458e-06, "loss": 3.4172, "step": 4851 }, { "epoch": 0.58, "learning_rate": 4.12890590820132e-06, "loss": 3.3029, "step": 4852 }, { "epoch": 0.58, "learning_rate": 4.126930784786481e-06, "loss": 3.458, "step": 4853 }, { "epoch": 0.58, "learning_rate": 4.124955801883822e-06, "loss": 3.3287, "step": 4854 }, { "epoch": 0.58, "learning_rate": 4.122980959811194e-06, "loss": 3.3465, "step": 4855 }, { "epoch": 0.58, "learning_rate": 4.121006258886433e-06, "loss": 3.4456, "step": 4856 }, { "epoch": 0.58, "learning_rate": 4.1190316994273465e-06, "loss": 3.391, "step": 4857 }, { "epoch": 0.58, "learning_rate": 4.1170572817517226e-06, "loss": 3.3858, "step": 4858 }, { "epoch": 0.58, "learning_rate": 4.1150830061773214e-06, "loss": 3.3867, "step": 4859 }, { "epoch": 0.58, "learning_rate": 4.113108873021885e-06, "loss": 3.3809, "step": 4860 }, { "epoch": 0.58, "learning_rate": 4.111134882603133e-06, "loss": 3.4337, "step": 4861 }, { "epoch": 0.58, "learning_rate": 4.109161035238759e-06, "loss": 3.4323, "step": 4862 }, { "epoch": 0.58, "learning_rate": 4.107187331246435e-06, "loss": 3.4619, "step": 4863 }, { "epoch": 0.58, "learning_rate": 4.1052137709438106e-06, "loss": 3.3876, "step": 4864 }, { "epoch": 0.58, "learning_rate": 4.103240354648511e-06, "loss": 3.4394, "step": 4865 }, { "epoch": 0.58, "learning_rate": 4.1012670826781395e-06, "loss": 3.473, "step": 4866 }, { "epoch": 0.58, "learning_rate": 4.099293955350276e-06, "loss": 3.3398, "step": 4867 }, { "epoch": 0.58, "learning_rate": 4.0973209729824735e-06, "loss": 3.438, "step": 4868 }, { "epoch": 0.58, "learning_rate": 4.095348135892267e-06, "loss": 3.4333, "step": 4869 }, { "epoch": 0.58, "learning_rate": 4.093375444397166e-06, "loss": 3.3638, "step": 4870 }, { "epoch": 0.58, "learning_rate": 4.091402898814656e-06, "loss": 3.3119, "step": 4871 }, { "epoch": 0.58, "learning_rate": 4.0894304994622e-06, "loss": 3.4028, "step": 4872 }, { "epoch": 0.58, "learning_rate": 4.087458246657235e-06, "loss": 3.3776, "step": 4873 }, { "epoch": 0.59, "learning_rate": 4.085486140717177e-06, "loss": 3.424, "step": 4874 }, { "epoch": 0.59, "learning_rate": 4.0835141819594194e-06, "loss": 3.4855, "step": 4875 }, { "epoch": 0.59, "learning_rate": 4.081542370701331e-06, "loss": 3.4458, "step": 4876 }, { "epoch": 0.59, "learning_rate": 4.079570707260252e-06, "loss": 3.3792, "step": 4877 }, { "epoch": 0.59, "learning_rate": 4.077599191953505e-06, "loss": 3.3826, "step": 4878 }, { "epoch": 0.59, "learning_rate": 4.075627825098386e-06, "loss": 3.3111, "step": 4879 }, { "epoch": 0.59, "learning_rate": 4.073656607012169e-06, "loss": 3.4221, "step": 4880 }, { "epoch": 0.59, "learning_rate": 4.071685538012103e-06, "loss": 3.3765, "step": 4881 }, { "epoch": 0.59, "learning_rate": 4.06971461841541e-06, "loss": 3.4243, "step": 4882 }, { "epoch": 0.59, "learning_rate": 4.067743848539293e-06, "loss": 3.496, "step": 4883 }, { "epoch": 0.59, "learning_rate": 4.065773228700929e-06, "loss": 3.3014, "step": 4884 }, { "epoch": 0.59, "learning_rate": 4.063802759217471e-06, "loss": 3.4224, "step": 4885 }, { "epoch": 0.59, "learning_rate": 4.061832440406044e-06, "loss": 3.4158, "step": 4886 }, { "epoch": 0.59, "learning_rate": 4.059862272583755e-06, "loss": 3.3787, "step": 4887 }, { "epoch": 0.59, "learning_rate": 4.057892256067683e-06, "loss": 3.4432, "step": 4888 }, { "epoch": 0.59, "learning_rate": 4.055922391174883e-06, "loss": 3.4162, "step": 4889 }, { "epoch": 0.59, "learning_rate": 4.053952678222385e-06, "loss": 3.3527, "step": 4890 }, { "epoch": 0.59, "learning_rate": 4.051983117527198e-06, "loss": 3.4476, "step": 4891 }, { "epoch": 0.59, "learning_rate": 4.050013709406304e-06, "loss": 3.3905, "step": 4892 }, { "epoch": 0.59, "learning_rate": 4.048044454176658e-06, "loss": 3.4088, "step": 4893 }, { "epoch": 0.59, "learning_rate": 4.046075352155196e-06, "loss": 3.3719, "step": 4894 }, { "epoch": 0.59, "learning_rate": 4.0441064036588235e-06, "loss": 3.3207, "step": 4895 }, { "epoch": 0.59, "learning_rate": 4.042137609004424e-06, "loss": 3.3683, "step": 4896 }, { "epoch": 0.59, "learning_rate": 4.040168968508858e-06, "loss": 3.3697, "step": 4897 }, { "epoch": 0.59, "learning_rate": 4.038200482488959e-06, "loss": 3.3552, "step": 4898 }, { "epoch": 0.59, "learning_rate": 4.036232151261536e-06, "loss": 3.3253, "step": 4899 }, { "epoch": 0.59, "learning_rate": 4.034263975143371e-06, "loss": 3.4345, "step": 4900 }, { "epoch": 0.59, "learning_rate": 4.032295954451227e-06, "loss": 3.3991, "step": 4901 }, { "epoch": 0.59, "learning_rate": 4.030328089501836e-06, "loss": 3.4829, "step": 4902 }, { "epoch": 0.59, "learning_rate": 4.0283603806119086e-06, "loss": 3.4383, "step": 4903 }, { "epoch": 0.59, "learning_rate": 4.026392828098125e-06, "loss": 3.3753, "step": 4904 }, { "epoch": 0.59, "learning_rate": 4.024425432277149e-06, "loss": 3.4204, "step": 4905 }, { "epoch": 0.59, "learning_rate": 4.022458193465612e-06, "loss": 3.3236, "step": 4906 }, { "epoch": 0.59, "learning_rate": 4.020491111980121e-06, "loss": 3.3713, "step": 4907 }, { "epoch": 0.59, "learning_rate": 4.018524188137261e-06, "loss": 3.3088, "step": 4908 }, { "epoch": 0.59, "learning_rate": 4.016557422253589e-06, "loss": 3.3401, "step": 4909 }, { "epoch": 0.59, "learning_rate": 4.014590814645636e-06, "loss": 3.4145, "step": 4910 }, { "epoch": 0.59, "learning_rate": 4.012624365629911e-06, "loss": 3.4555, "step": 4911 }, { "epoch": 0.59, "learning_rate": 4.010658075522897e-06, "loss": 3.4391, "step": 4912 }, { "epoch": 0.59, "learning_rate": 4.0086919446410445e-06, "loss": 3.341, "step": 4913 }, { "epoch": 0.59, "learning_rate": 4.006725973300787e-06, "loss": 3.4129, "step": 4914 }, { "epoch": 0.59, "learning_rate": 4.004760161818526e-06, "loss": 3.4242, "step": 4915 }, { "epoch": 0.59, "learning_rate": 4.002794510510643e-06, "loss": 3.2965, "step": 4916 }, { "epoch": 0.59, "learning_rate": 4.00082901969349e-06, "loss": 3.3058, "step": 4917 }, { "epoch": 0.59, "learning_rate": 3.998863689683394e-06, "loss": 3.46, "step": 4918 }, { "epoch": 0.59, "learning_rate": 3.996898520796655e-06, "loss": 3.4173, "step": 4919 }, { "epoch": 0.59, "learning_rate": 3.9949335133495506e-06, "loss": 3.4088, "step": 4920 }, { "epoch": 0.59, "learning_rate": 3.99296866765833e-06, "loss": 3.3876, "step": 4921 }, { "epoch": 0.59, "learning_rate": 3.991003984039213e-06, "loss": 3.4278, "step": 4922 }, { "epoch": 0.59, "learning_rate": 3.989039462808399e-06, "loss": 3.3753, "step": 4923 }, { "epoch": 0.59, "learning_rate": 3.98707510428206e-06, "loss": 3.4031, "step": 4924 }, { "epoch": 0.59, "learning_rate": 3.985110908776338e-06, "loss": 3.4571, "step": 4925 }, { "epoch": 0.59, "learning_rate": 3.983146876607355e-06, "loss": 3.4442, "step": 4926 }, { "epoch": 0.59, "learning_rate": 3.9811830080912e-06, "loss": 3.3971, "step": 4927 }, { "epoch": 0.59, "learning_rate": 3.979219303543942e-06, "loss": 3.4184, "step": 4928 }, { "epoch": 0.59, "learning_rate": 3.977255763281618e-06, "loss": 3.3722, "step": 4929 }, { "epoch": 0.59, "learning_rate": 3.975292387620244e-06, "loss": 3.3736, "step": 4930 }, { "epoch": 0.59, "learning_rate": 3.973329176875804e-06, "loss": 3.4272, "step": 4931 }, { "epoch": 0.59, "learning_rate": 3.971366131364258e-06, "loss": 3.5148, "step": 4932 }, { "epoch": 0.59, "learning_rate": 3.96940325140154e-06, "loss": 3.2889, "step": 4933 }, { "epoch": 0.59, "learning_rate": 3.967440537303557e-06, "loss": 3.3684, "step": 4934 }, { "epoch": 0.59, "learning_rate": 3.965477989386189e-06, "loss": 3.4845, "step": 4935 }, { "epoch": 0.59, "learning_rate": 3.96351560796529e-06, "loss": 3.4571, "step": 4936 }, { "epoch": 0.59, "learning_rate": 3.961553393356687e-06, "loss": 3.3789, "step": 4937 }, { "epoch": 0.59, "learning_rate": 3.959591345876179e-06, "loss": 3.3274, "step": 4938 }, { "epoch": 0.59, "learning_rate": 3.957629465839541e-06, "loss": 3.3234, "step": 4939 }, { "epoch": 0.59, "learning_rate": 3.955667753562515e-06, "loss": 3.3489, "step": 4940 }, { "epoch": 0.59, "learning_rate": 3.9537062093608215e-06, "loss": 3.3487, "step": 4941 }, { "epoch": 0.59, "learning_rate": 3.951744833550152e-06, "loss": 3.4196, "step": 4942 }, { "epoch": 0.59, "learning_rate": 3.949783626446173e-06, "loss": 3.4703, "step": 4943 }, { "epoch": 0.59, "learning_rate": 3.9478225883645205e-06, "loss": 3.322, "step": 4944 }, { "epoch": 0.59, "learning_rate": 3.945861719620805e-06, "loss": 3.3505, "step": 4945 }, { "epoch": 0.59, "learning_rate": 3.943901020530611e-06, "loss": 3.3408, "step": 4946 }, { "epoch": 0.59, "learning_rate": 3.941940491409493e-06, "loss": 3.3526, "step": 4947 }, { "epoch": 0.59, "learning_rate": 3.939980132572981e-06, "loss": 3.4325, "step": 4948 }, { "epoch": 0.59, "learning_rate": 3.938019944336573e-06, "loss": 3.3821, "step": 4949 }, { "epoch": 0.59, "learning_rate": 3.936059927015747e-06, "loss": 3.2857, "step": 4950 }, { "epoch": 0.59, "learning_rate": 3.934100080925945e-06, "loss": 3.396, "step": 4951 }, { "epoch": 0.59, "learning_rate": 3.932140406382587e-06, "loss": 3.4107, "step": 4952 }, { "epoch": 0.59, "learning_rate": 3.930180903701065e-06, "loss": 3.4751, "step": 4953 }, { "epoch": 0.59, "learning_rate": 3.928221573196742e-06, "loss": 3.4557, "step": 4954 }, { "epoch": 0.59, "learning_rate": 3.926262415184953e-06, "loss": 3.4979, "step": 4955 }, { "epoch": 0.59, "learning_rate": 3.924303429981006e-06, "loss": 3.4074, "step": 4956 }, { "epoch": 0.6, "learning_rate": 3.922344617900183e-06, "loss": 3.4286, "step": 4957 }, { "epoch": 0.6, "learning_rate": 3.920385979257732e-06, "loss": 3.3304, "step": 4958 }, { "epoch": 0.6, "learning_rate": 3.9184275143688796e-06, "loss": 3.3491, "step": 4959 }, { "epoch": 0.6, "learning_rate": 3.916469223548822e-06, "loss": 3.5301, "step": 4960 }, { "epoch": 0.6, "learning_rate": 3.9145111071127275e-06, "loss": 3.376, "step": 4961 }, { "epoch": 0.6, "learning_rate": 3.912553165375737e-06, "loss": 3.3854, "step": 4962 }, { "epoch": 0.6, "learning_rate": 3.91059539865296e-06, "loss": 3.3932, "step": 4963 }, { "epoch": 0.6, "learning_rate": 3.908637807259483e-06, "loss": 3.3965, "step": 4964 }, { "epoch": 0.6, "learning_rate": 3.906680391510361e-06, "loss": 3.4761, "step": 4965 }, { "epoch": 0.6, "learning_rate": 3.9047231517206225e-06, "loss": 3.4154, "step": 4966 }, { "epoch": 0.6, "learning_rate": 3.902766088205264e-06, "loss": 3.4027, "step": 4967 }, { "epoch": 0.6, "learning_rate": 3.900809201279258e-06, "loss": 3.4878, "step": 4968 }, { "epoch": 0.6, "learning_rate": 3.898852491257547e-06, "loss": 3.4177, "step": 4969 }, { "epoch": 0.6, "learning_rate": 3.896895958455043e-06, "loss": 3.4855, "step": 4970 }, { "epoch": 0.6, "learning_rate": 3.894939603186634e-06, "loss": 3.3281, "step": 4971 }, { "epoch": 0.6, "learning_rate": 3.892983425767175e-06, "loss": 3.34, "step": 4972 }, { "epoch": 0.6, "learning_rate": 3.891027426511494e-06, "loss": 3.446, "step": 4973 }, { "epoch": 0.6, "learning_rate": 3.889071605734391e-06, "loss": 3.4988, "step": 4974 }, { "epoch": 0.6, "learning_rate": 3.88711596375064e-06, "loss": 3.3922, "step": 4975 }, { "epoch": 0.6, "learning_rate": 3.885160500874977e-06, "loss": 3.4687, "step": 4976 }, { "epoch": 0.6, "learning_rate": 3.883205217422117e-06, "loss": 3.3514, "step": 4977 }, { "epoch": 0.6, "learning_rate": 3.881250113706747e-06, "loss": 3.3912, "step": 4978 }, { "epoch": 0.6, "learning_rate": 3.879295190043519e-06, "loss": 3.4162, "step": 4979 }, { "epoch": 0.6, "learning_rate": 3.877340446747061e-06, "loss": 3.363, "step": 4980 }, { "epoch": 0.6, "learning_rate": 3.8753858841319705e-06, "loss": 3.3949, "step": 4981 }, { "epoch": 0.6, "learning_rate": 3.873431502512815e-06, "loss": 3.3507, "step": 4982 }, { "epoch": 0.6, "learning_rate": 3.871477302204134e-06, "loss": 3.4097, "step": 4983 }, { "epoch": 0.6, "learning_rate": 3.869523283520438e-06, "loss": 3.5029, "step": 4984 }, { "epoch": 0.6, "learning_rate": 3.867569446776207e-06, "loss": 3.4728, "step": 4985 }, { "epoch": 0.6, "learning_rate": 3.865615792285892e-06, "loss": 3.3449, "step": 4986 }, { "epoch": 0.6, "learning_rate": 3.8636623203639155e-06, "loss": 3.3707, "step": 4987 }, { "epoch": 0.6, "learning_rate": 3.861709031324671e-06, "loss": 3.3277, "step": 4988 }, { "epoch": 0.6, "learning_rate": 3.859755925482521e-06, "loss": 3.4684, "step": 4989 }, { "epoch": 0.6, "learning_rate": 3.857803003151799e-06, "loss": 3.367, "step": 4990 }, { "epoch": 0.6, "learning_rate": 3.855850264646809e-06, "loss": 3.3099, "step": 4991 }, { "epoch": 0.6, "learning_rate": 3.853897710281827e-06, "loss": 3.3275, "step": 4992 }, { "epoch": 0.6, "learning_rate": 3.8519453403711004e-06, "loss": 3.3918, "step": 4993 }, { "epoch": 0.6, "learning_rate": 3.849993155228838e-06, "loss": 3.3734, "step": 4994 }, { "epoch": 0.6, "learning_rate": 3.84804115516923e-06, "loss": 3.3847, "step": 4995 }, { "epoch": 0.6, "learning_rate": 3.84608934050643e-06, "loss": 3.4509, "step": 4996 }, { "epoch": 0.6, "learning_rate": 3.844137711554566e-06, "loss": 3.2987, "step": 4997 }, { "epoch": 0.6, "learning_rate": 3.8421862686277336e-06, "loss": 3.3891, "step": 4998 }, { "epoch": 0.6, "learning_rate": 3.840235012039998e-06, "loss": 3.3857, "step": 4999 }, { "epoch": 0.6, "learning_rate": 3.838283942105396e-06, "loss": 3.4454, "step": 5000 }, { "epoch": 0.6, "learning_rate": 3.836333059137933e-06, "loss": 3.4092, "step": 5001 }, { "epoch": 0.6, "learning_rate": 3.834382363451588e-06, "loss": 3.3686, "step": 5002 }, { "epoch": 0.6, "learning_rate": 3.832431855360303e-06, "loss": 3.4333, "step": 5003 }, { "epoch": 0.6, "learning_rate": 3.830481535177996e-06, "loss": 3.4751, "step": 5004 }, { "epoch": 0.6, "learning_rate": 3.828531403218551e-06, "loss": 3.4994, "step": 5005 }, { "epoch": 0.6, "learning_rate": 3.826581459795824e-06, "loss": 3.4328, "step": 5006 }, { "epoch": 0.6, "learning_rate": 3.82463170522364e-06, "loss": 3.4284, "step": 5007 }, { "epoch": 0.6, "learning_rate": 3.822682139815794e-06, "loss": 3.3532, "step": 5008 }, { "epoch": 0.6, "learning_rate": 3.820732763886049e-06, "loss": 3.4235, "step": 5009 }, { "epoch": 0.6, "learning_rate": 3.818783577748138e-06, "loss": 3.3267, "step": 5010 }, { "epoch": 0.6, "learning_rate": 3.816834581715767e-06, "loss": 3.412, "step": 5011 }, { "epoch": 0.6, "learning_rate": 3.814885776102605e-06, "loss": 3.3556, "step": 5012 }, { "epoch": 0.6, "learning_rate": 3.812937161222293e-06, "loss": 3.4069, "step": 5013 }, { "epoch": 0.6, "learning_rate": 3.8109887373884452e-06, "loss": 3.3608, "step": 5014 }, { "epoch": 0.6, "learning_rate": 3.80904050491464e-06, "loss": 3.375, "step": 5015 }, { "epoch": 0.6, "learning_rate": 3.8070924641144275e-06, "loss": 3.2941, "step": 5016 }, { "epoch": 0.6, "learning_rate": 3.8051446153013256e-06, "loss": 3.3364, "step": 5017 }, { "epoch": 0.6, "learning_rate": 3.8031969587888228e-06, "loss": 3.4732, "step": 5018 }, { "epoch": 0.6, "learning_rate": 3.801249494890376e-06, "loss": 3.5099, "step": 5019 }, { "epoch": 0.6, "learning_rate": 3.799302223919411e-06, "loss": 3.3962, "step": 5020 }, { "epoch": 0.6, "learning_rate": 3.797355146189322e-06, "loss": 3.4393, "step": 5021 }, { "epoch": 0.6, "learning_rate": 3.795408262013472e-06, "loss": 3.3641, "step": 5022 }, { "epoch": 0.6, "learning_rate": 3.793461571705195e-06, "loss": 3.3626, "step": 5023 }, { "epoch": 0.6, "learning_rate": 3.7915150755777906e-06, "loss": 3.2835, "step": 5024 }, { "epoch": 0.6, "learning_rate": 3.7895687739445296e-06, "loss": 3.3657, "step": 5025 }, { "epoch": 0.6, "learning_rate": 3.7876226671186517e-06, "loss": 3.4485, "step": 5026 }, { "epoch": 0.6, "learning_rate": 3.785676755413363e-06, "loss": 3.3873, "step": 5027 }, { "epoch": 0.6, "learning_rate": 3.7837310391418402e-06, "loss": 3.4067, "step": 5028 }, { "epoch": 0.6, "learning_rate": 3.7817855186172292e-06, "loss": 3.4134, "step": 5029 }, { "epoch": 0.6, "learning_rate": 3.7798401941526397e-06, "loss": 3.4335, "step": 5030 }, { "epoch": 0.6, "learning_rate": 3.777895066061154e-06, "loss": 3.4361, "step": 5031 }, { "epoch": 0.6, "learning_rate": 3.775950134655822e-06, "loss": 3.4602, "step": 5032 }, { "epoch": 0.6, "learning_rate": 3.7740054002496633e-06, "loss": 3.4471, "step": 5033 }, { "epoch": 0.6, "learning_rate": 3.772060863155664e-06, "loss": 3.3776, "step": 5034 }, { "epoch": 0.6, "learning_rate": 3.7701165236867786e-06, "loss": 3.4281, "step": 5035 }, { "epoch": 0.6, "learning_rate": 3.76817238215593e-06, "loss": 3.4036, "step": 5036 }, { "epoch": 0.6, "learning_rate": 3.766228438876009e-06, "loss": 3.2817, "step": 5037 }, { "epoch": 0.6, "learning_rate": 3.764284694159877e-06, "loss": 3.4638, "step": 5038 }, { "epoch": 0.6, "learning_rate": 3.7623411483203577e-06, "loss": 3.387, "step": 5039 }, { "epoch": 0.6, "learning_rate": 3.760397801670247e-06, "loss": 3.3969, "step": 5040 }, { "epoch": 0.61, "learning_rate": 3.758454654522309e-06, "loss": 3.4122, "step": 5041 }, { "epoch": 0.61, "learning_rate": 3.7565117071892748e-06, "loss": 3.4003, "step": 5042 }, { "epoch": 0.61, "learning_rate": 3.754568959983843e-06, "loss": 3.3182, "step": 5043 }, { "epoch": 0.61, "learning_rate": 3.7526264132186796e-06, "loss": 3.4551, "step": 5044 }, { "epoch": 0.61, "learning_rate": 3.7506840672064205e-06, "loss": 3.3506, "step": 5045 }, { "epoch": 0.61, "learning_rate": 3.7487419222596664e-06, "loss": 3.5433, "step": 5046 }, { "epoch": 0.61, "learning_rate": 3.746799978690989e-06, "loss": 3.4208, "step": 5047 }, { "epoch": 0.61, "learning_rate": 3.744858236812924e-06, "loss": 3.4722, "step": 5048 }, { "epoch": 0.61, "learning_rate": 3.7429166969379753e-06, "loss": 3.3842, "step": 5049 }, { "epoch": 0.61, "learning_rate": 3.740975359378617e-06, "loss": 3.3662, "step": 5050 }, { "epoch": 0.61, "learning_rate": 3.7390342244472886e-06, "loss": 3.3508, "step": 5051 }, { "epoch": 0.61, "learning_rate": 3.737093292456397e-06, "loss": 3.3772, "step": 5052 }, { "epoch": 0.61, "learning_rate": 3.7351525637183168e-06, "loss": 3.426, "step": 5053 }, { "epoch": 0.61, "learning_rate": 3.7332120385453896e-06, "loss": 3.4999, "step": 5054 }, { "epoch": 0.61, "learning_rate": 3.7312717172499247e-06, "loss": 3.3855, "step": 5055 }, { "epoch": 0.61, "learning_rate": 3.7293316001442005e-06, "loss": 3.3178, "step": 5056 }, { "epoch": 0.61, "learning_rate": 3.727391687540456e-06, "loss": 3.332, "step": 5057 }, { "epoch": 0.61, "learning_rate": 3.725451979750904e-06, "loss": 3.2901, "step": 5058 }, { "epoch": 0.61, "learning_rate": 3.723512477087722e-06, "loss": 3.3759, "step": 5059 }, { "epoch": 0.61, "learning_rate": 3.721573179863054e-06, "loss": 3.4405, "step": 5060 }, { "epoch": 0.61, "learning_rate": 3.7196340883890125e-06, "loss": 3.3469, "step": 5061 }, { "epoch": 0.61, "learning_rate": 3.717695202977675e-06, "loss": 3.4672, "step": 5062 }, { "epoch": 0.61, "learning_rate": 3.7157565239410854e-06, "loss": 3.434, "step": 5063 }, { "epoch": 0.61, "learning_rate": 3.713818051591258e-06, "loss": 3.417, "step": 5064 }, { "epoch": 0.61, "learning_rate": 3.7118797862401712e-06, "loss": 3.3653, "step": 5065 }, { "epoch": 0.61, "learning_rate": 3.709941728199769e-06, "loss": 3.4821, "step": 5066 }, { "epoch": 0.61, "learning_rate": 3.7080038777819638e-06, "loss": 3.4299, "step": 5067 }, { "epoch": 0.61, "learning_rate": 3.706066235298634e-06, "loss": 3.396, "step": 5068 }, { "epoch": 0.61, "learning_rate": 3.7041288010616255e-06, "loss": 3.4228, "step": 5069 }, { "epoch": 0.61, "learning_rate": 3.702191575382749e-06, "loss": 3.3657, "step": 5070 }, { "epoch": 0.61, "learning_rate": 3.7002545585737836e-06, "loss": 3.3864, "step": 5071 }, { "epoch": 0.61, "learning_rate": 3.698317750946473e-06, "loss": 3.2771, "step": 5072 }, { "epoch": 0.61, "learning_rate": 3.6963811528125275e-06, "loss": 3.4195, "step": 5073 }, { "epoch": 0.61, "learning_rate": 3.6944447644836273e-06, "loss": 3.2038, "step": 5074 }, { "epoch": 0.61, "learning_rate": 3.692508586271411e-06, "loss": 3.3918, "step": 5075 }, { "epoch": 0.61, "learning_rate": 3.69057261848749e-06, "loss": 3.366, "step": 5076 }, { "epoch": 0.61, "learning_rate": 3.68863686144344e-06, "loss": 3.4298, "step": 5077 }, { "epoch": 0.61, "learning_rate": 3.6867013154508024e-06, "loss": 3.303, "step": 5078 }, { "epoch": 0.61, "learning_rate": 3.6847659808210855e-06, "loss": 3.394, "step": 5079 }, { "epoch": 0.61, "learning_rate": 3.6828308578657623e-06, "loss": 3.3236, "step": 5080 }, { "epoch": 0.61, "learning_rate": 3.6808959468962736e-06, "loss": 3.3956, "step": 5081 }, { "epoch": 0.61, "learning_rate": 3.678961248224023e-06, "loss": 3.4069, "step": 5082 }, { "epoch": 0.61, "learning_rate": 3.6770267621603844e-06, "loss": 3.436, "step": 5083 }, { "epoch": 0.61, "learning_rate": 3.675092489016693e-06, "loss": 3.3984, "step": 5084 }, { "epoch": 0.61, "learning_rate": 3.673158429104251e-06, "loss": 3.3474, "step": 5085 }, { "epoch": 0.61, "learning_rate": 3.6712245827343286e-06, "loss": 3.3445, "step": 5086 }, { "epoch": 0.61, "learning_rate": 3.669290950218159e-06, "loss": 3.2344, "step": 5087 }, { "epoch": 0.61, "learning_rate": 3.6673575318669425e-06, "loss": 3.3648, "step": 5088 }, { "epoch": 0.61, "learning_rate": 3.6654243279918446e-06, "loss": 3.5219, "step": 5089 }, { "epoch": 0.61, "learning_rate": 3.663491338903995e-06, "loss": 3.3216, "step": 5090 }, { "epoch": 0.61, "learning_rate": 3.66155856491449e-06, "loss": 3.4272, "step": 5091 }, { "epoch": 0.61, "learning_rate": 3.659626006334395e-06, "loss": 3.356, "step": 5092 }, { "epoch": 0.61, "learning_rate": 3.657693663474731e-06, "loss": 3.3993, "step": 5093 }, { "epoch": 0.61, "learning_rate": 3.655761536646493e-06, "loss": 3.4428, "step": 5094 }, { "epoch": 0.61, "learning_rate": 3.6538296261606374e-06, "loss": 3.3033, "step": 5095 }, { "epoch": 0.61, "learning_rate": 3.651897932328089e-06, "loss": 3.4734, "step": 5096 }, { "epoch": 0.61, "learning_rate": 3.6499664554597337e-06, "loss": 3.3483, "step": 5097 }, { "epoch": 0.61, "learning_rate": 3.648035195866424e-06, "loss": 3.428, "step": 5098 }, { "epoch": 0.61, "learning_rate": 3.6461041538589793e-06, "loss": 3.4162, "step": 5099 }, { "epoch": 0.61, "learning_rate": 3.6441733297481807e-06, "loss": 3.3224, "step": 5100 }, { "epoch": 0.61, "learning_rate": 3.6422427238447787e-06, "loss": 3.3887, "step": 5101 }, { "epoch": 0.61, "learning_rate": 3.6403123364594827e-06, "loss": 3.3734, "step": 5102 }, { "epoch": 0.61, "learning_rate": 3.638382167902972e-06, "loss": 3.387, "step": 5103 }, { "epoch": 0.61, "learning_rate": 3.6364522184858885e-06, "loss": 3.3337, "step": 5104 }, { "epoch": 0.61, "learning_rate": 3.634522488518839e-06, "loss": 3.4339, "step": 5105 }, { "epoch": 0.61, "learning_rate": 3.632592978312395e-06, "loss": 3.3742, "step": 5106 }, { "epoch": 0.61, "learning_rate": 3.630663688177093e-06, "loss": 3.3842, "step": 5107 }, { "epoch": 0.61, "learning_rate": 3.6287346184234344e-06, "loss": 3.4519, "step": 5108 }, { "epoch": 0.61, "learning_rate": 3.626805769361884e-06, "loss": 3.4966, "step": 5109 }, { "epoch": 0.61, "learning_rate": 3.624877141302874e-06, "loss": 3.3453, "step": 5110 }, { "epoch": 0.61, "learning_rate": 3.622948734556795e-06, "loss": 3.3955, "step": 5111 }, { "epoch": 0.61, "learning_rate": 3.6210205494340066e-06, "loss": 3.3701, "step": 5112 }, { "epoch": 0.61, "learning_rate": 3.6190925862448334e-06, "loss": 3.4595, "step": 5113 }, { "epoch": 0.61, "learning_rate": 3.617164845299561e-06, "loss": 3.3739, "step": 5114 }, { "epoch": 0.61, "learning_rate": 3.615237326908442e-06, "loss": 3.4118, "step": 5115 }, { "epoch": 0.61, "learning_rate": 3.613310031381693e-06, "loss": 3.3672, "step": 5116 }, { "epoch": 0.61, "learning_rate": 3.611382959029491e-06, "loss": 3.3595, "step": 5117 }, { "epoch": 0.61, "learning_rate": 3.6094561101619817e-06, "loss": 3.3695, "step": 5118 }, { "epoch": 0.61, "learning_rate": 3.607529485089277e-06, "loss": 3.4053, "step": 5119 }, { "epoch": 0.61, "learning_rate": 3.6056030841214417e-06, "loss": 3.3971, "step": 5120 }, { "epoch": 0.61, "learning_rate": 3.6036769075685163e-06, "loss": 3.3341, "step": 5121 }, { "epoch": 0.61, "learning_rate": 3.6017509557404983e-06, "loss": 3.3997, "step": 5122 }, { "epoch": 0.61, "learning_rate": 3.5998252289473524e-06, "loss": 3.5117, "step": 5123 }, { "epoch": 0.62, "learning_rate": 3.5978997274990064e-06, "loss": 3.3272, "step": 5124 }, { "epoch": 0.62, "learning_rate": 3.5959744517053507e-06, "loss": 3.4803, "step": 5125 }, { "epoch": 0.62, "learning_rate": 3.5940494018762405e-06, "loss": 3.3435, "step": 5126 }, { "epoch": 0.62, "learning_rate": 3.5921245783214938e-06, "loss": 3.4538, "step": 5127 }, { "epoch": 0.62, "learning_rate": 3.5901999813508946e-06, "loss": 3.3403, "step": 5128 }, { "epoch": 0.62, "learning_rate": 3.5882756112741847e-06, "loss": 3.353, "step": 5129 }, { "epoch": 0.62, "learning_rate": 3.586351468401076e-06, "loss": 3.4009, "step": 5130 }, { "epoch": 0.62, "learning_rate": 3.5844275530412393e-06, "loss": 3.3546, "step": 5131 }, { "epoch": 0.62, "learning_rate": 3.582503865504312e-06, "loss": 3.4095, "step": 5132 }, { "epoch": 0.62, "learning_rate": 3.5805804060998926e-06, "loss": 3.5034, "step": 5133 }, { "epoch": 0.62, "learning_rate": 3.5786571751375437e-06, "loss": 3.297, "step": 5134 }, { "epoch": 0.62, "learning_rate": 3.5767341729267907e-06, "loss": 3.429, "step": 5135 }, { "epoch": 0.62, "learning_rate": 3.5748113997771227e-06, "loss": 3.4002, "step": 5136 }, { "epoch": 0.62, "learning_rate": 3.5728888559979946e-06, "loss": 3.4214, "step": 5137 }, { "epoch": 0.62, "learning_rate": 3.570966541898816e-06, "loss": 3.4318, "step": 5138 }, { "epoch": 0.62, "learning_rate": 3.569044457788968e-06, "loss": 3.4378, "step": 5139 }, { "epoch": 0.62, "learning_rate": 3.567122603977791e-06, "loss": 3.4204, "step": 5140 }, { "epoch": 0.62, "learning_rate": 3.5652009807745904e-06, "loss": 3.3033, "step": 5141 }, { "epoch": 0.62, "learning_rate": 3.563279588488631e-06, "loss": 3.3251, "step": 5142 }, { "epoch": 0.62, "learning_rate": 3.561358427429145e-06, "loss": 3.3134, "step": 5143 }, { "epoch": 0.62, "learning_rate": 3.5594374979053238e-06, "loss": 3.4483, "step": 5144 }, { "epoch": 0.62, "learning_rate": 3.5575168002263227e-06, "loss": 3.4368, "step": 5145 }, { "epoch": 0.62, "learning_rate": 3.5555963347012606e-06, "loss": 3.3634, "step": 5146 }, { "epoch": 0.62, "learning_rate": 3.5536761016392164e-06, "loss": 3.4407, "step": 5147 }, { "epoch": 0.62, "learning_rate": 3.5517561013492346e-06, "loss": 3.2653, "step": 5148 }, { "epoch": 0.62, "learning_rate": 3.549836334140321e-06, "loss": 3.3543, "step": 5149 }, { "epoch": 0.62, "learning_rate": 3.5479168003214426e-06, "loss": 3.3803, "step": 5150 }, { "epoch": 0.62, "learning_rate": 3.5459975002015314e-06, "loss": 3.4031, "step": 5151 }, { "epoch": 0.62, "learning_rate": 3.5440784340894797e-06, "loss": 3.3987, "step": 5152 }, { "epoch": 0.62, "learning_rate": 3.542159602294144e-06, "loss": 3.3769, "step": 5153 }, { "epoch": 0.62, "learning_rate": 3.5402410051243406e-06, "loss": 3.4273, "step": 5154 }, { "epoch": 0.62, "learning_rate": 3.5383226428888533e-06, "loss": 3.3533, "step": 5155 }, { "epoch": 0.62, "learning_rate": 3.5364045158964185e-06, "loss": 3.3644, "step": 5156 }, { "epoch": 0.62, "learning_rate": 3.534486624455743e-06, "loss": 3.4547, "step": 5157 }, { "epoch": 0.62, "learning_rate": 3.532568968875493e-06, "loss": 3.3276, "step": 5158 }, { "epoch": 0.62, "learning_rate": 3.530651549464298e-06, "loss": 3.3939, "step": 5159 }, { "epoch": 0.62, "learning_rate": 3.528734366530747e-06, "loss": 3.3831, "step": 5160 }, { "epoch": 0.62, "learning_rate": 3.526817420383393e-06, "loss": 3.2966, "step": 5161 }, { "epoch": 0.62, "learning_rate": 3.52490071133075e-06, "loss": 3.288, "step": 5162 }, { "epoch": 0.62, "learning_rate": 3.522984239681294e-06, "loss": 3.4439, "step": 5163 }, { "epoch": 0.62, "learning_rate": 3.521068005743465e-06, "loss": 3.3542, "step": 5164 }, { "epoch": 0.62, "learning_rate": 3.519152009825659e-06, "loss": 3.3564, "step": 5165 }, { "epoch": 0.62, "learning_rate": 3.517236252236239e-06, "loss": 3.3199, "step": 5166 }, { "epoch": 0.62, "learning_rate": 3.5153207332835284e-06, "loss": 3.2672, "step": 5167 }, { "epoch": 0.62, "learning_rate": 3.5134054532758103e-06, "loss": 3.4463, "step": 5168 }, { "epoch": 0.62, "learning_rate": 3.5114904125213324e-06, "loss": 3.4643, "step": 5169 }, { "epoch": 0.62, "learning_rate": 3.509575611328302e-06, "loss": 3.4076, "step": 5170 }, { "epoch": 0.62, "learning_rate": 3.5076610500048876e-06, "loss": 3.4786, "step": 5171 }, { "epoch": 0.62, "learning_rate": 3.5057467288592195e-06, "loss": 3.3479, "step": 5172 }, { "epoch": 0.62, "learning_rate": 3.5038326481993916e-06, "loss": 3.4136, "step": 5173 }, { "epoch": 0.62, "learning_rate": 3.501918808333453e-06, "loss": 3.3025, "step": 5174 }, { "epoch": 0.62, "learning_rate": 3.500005209569421e-06, "loss": 3.4352, "step": 5175 }, { "epoch": 0.62, "learning_rate": 3.4980918522152697e-06, "loss": 3.4455, "step": 5176 }, { "epoch": 0.62, "learning_rate": 3.496178736578936e-06, "loss": 3.4234, "step": 5177 }, { "epoch": 0.62, "learning_rate": 3.4942658629683185e-06, "loss": 3.4441, "step": 5178 }, { "epoch": 0.62, "learning_rate": 3.492353231691275e-06, "loss": 3.4422, "step": 5179 }, { "epoch": 0.62, "learning_rate": 3.4904408430556256e-06, "loss": 3.396, "step": 5180 }, { "epoch": 0.62, "learning_rate": 3.4885286973691513e-06, "loss": 3.4186, "step": 5181 }, { "epoch": 0.62, "learning_rate": 3.4866167949395945e-06, "loss": 3.3531, "step": 5182 }, { "epoch": 0.62, "learning_rate": 3.4847051360746565e-06, "loss": 3.4399, "step": 5183 }, { "epoch": 0.62, "learning_rate": 3.482793721082e-06, "loss": 3.363, "step": 5184 }, { "epoch": 0.62, "learning_rate": 3.48088255026925e-06, "loss": 3.2812, "step": 5185 }, { "epoch": 0.62, "learning_rate": 3.478971623943992e-06, "loss": 3.4462, "step": 5186 }, { "epoch": 0.62, "learning_rate": 3.47706094241377e-06, "loss": 3.4093, "step": 5187 }, { "epoch": 0.62, "learning_rate": 3.47515050598609e-06, "loss": 3.4483, "step": 5188 }, { "epoch": 0.62, "learning_rate": 3.4732403149684204e-06, "loss": 3.4065, "step": 5189 }, { "epoch": 0.62, "learning_rate": 3.4713303696681862e-06, "loss": 3.4279, "step": 5190 }, { "epoch": 0.62, "learning_rate": 3.469420670392778e-06, "loss": 3.4189, "step": 5191 }, { "epoch": 0.62, "learning_rate": 3.467511217449539e-06, "loss": 3.3308, "step": 5192 }, { "epoch": 0.62, "learning_rate": 3.46560201114578e-06, "loss": 3.3727, "step": 5193 }, { "epoch": 0.62, "learning_rate": 3.46369305178877e-06, "loss": 3.4385, "step": 5194 }, { "epoch": 0.62, "learning_rate": 3.4617843396857364e-06, "loss": 3.4048, "step": 5195 }, { "epoch": 0.62, "learning_rate": 3.4598758751438698e-06, "loss": 3.3217, "step": 5196 }, { "epoch": 0.62, "learning_rate": 3.4579676584703192e-06, "loss": 3.4402, "step": 5197 }, { "epoch": 0.62, "learning_rate": 3.4560596899721927e-06, "loss": 3.3914, "step": 5198 }, { "epoch": 0.62, "learning_rate": 3.4541519699565607e-06, "loss": 3.3967, "step": 5199 }, { "epoch": 0.62, "learning_rate": 3.4522444987304533e-06, "loss": 3.3914, "step": 5200 }, { "epoch": 0.62, "learning_rate": 3.4503372766008576e-06, "loss": 3.3769, "step": 5201 }, { "epoch": 0.62, "learning_rate": 3.448430303874724e-06, "loss": 3.3453, "step": 5202 }, { "epoch": 0.62, "learning_rate": 3.4465235808589625e-06, "loss": 3.361, "step": 5203 }, { "epoch": 0.62, "learning_rate": 3.4446171078604406e-06, "loss": 3.5156, "step": 5204 }, { "epoch": 0.62, "learning_rate": 3.442710885185987e-06, "loss": 3.4323, "step": 5205 }, { "epoch": 0.62, "learning_rate": 3.4408049131423914e-06, "loss": 3.419, "step": 5206 }, { "epoch": 0.63, "learning_rate": 3.438899192036401e-06, "loss": 3.4232, "step": 5207 }, { "epoch": 0.63, "learning_rate": 3.436993722174724e-06, "loss": 3.4291, "step": 5208 }, { "epoch": 0.63, "learning_rate": 3.4350885038640285e-06, "loss": 3.4557, "step": 5209 }, { "epoch": 0.63, "learning_rate": 3.4331835374109384e-06, "loss": 3.3867, "step": 5210 }, { "epoch": 0.63, "learning_rate": 3.431278823122042e-06, "loss": 3.3201, "step": 5211 }, { "epoch": 0.63, "learning_rate": 3.4293743613038834e-06, "loss": 3.358, "step": 5212 }, { "epoch": 0.63, "learning_rate": 3.427470152262969e-06, "loss": 3.3229, "step": 5213 }, { "epoch": 0.63, "learning_rate": 3.425566196305763e-06, "loss": 3.3831, "step": 5214 }, { "epoch": 0.63, "learning_rate": 3.4236624937386874e-06, "loss": 3.4996, "step": 5215 }, { "epoch": 0.63, "learning_rate": 3.421759044868127e-06, "loss": 3.3963, "step": 5216 }, { "epoch": 0.63, "learning_rate": 3.4198558500004237e-06, "loss": 3.4103, "step": 5217 }, { "epoch": 0.63, "learning_rate": 3.4179529094418796e-06, "loss": 3.3791, "step": 5218 }, { "epoch": 0.63, "learning_rate": 3.416050223498752e-06, "loss": 3.4525, "step": 5219 }, { "epoch": 0.63, "learning_rate": 3.4141477924772605e-06, "loss": 3.3973, "step": 5220 }, { "epoch": 0.63, "learning_rate": 3.412245616683585e-06, "loss": 3.3988, "step": 5221 }, { "epoch": 0.63, "learning_rate": 3.410343696423861e-06, "loss": 3.4452, "step": 5222 }, { "epoch": 0.63, "learning_rate": 3.4084420320041855e-06, "loss": 3.4361, "step": 5223 }, { "epoch": 0.63, "learning_rate": 3.4065406237306125e-06, "loss": 3.3928, "step": 5224 }, { "epoch": 0.63, "learning_rate": 3.404639471909156e-06, "loss": 3.4528, "step": 5225 }, { "epoch": 0.63, "learning_rate": 3.402738576845789e-06, "loss": 3.3826, "step": 5226 }, { "epoch": 0.63, "learning_rate": 3.400837938846442e-06, "loss": 3.4214, "step": 5227 }, { "epoch": 0.63, "learning_rate": 3.3989375582170038e-06, "loss": 3.3879, "step": 5228 }, { "epoch": 0.63, "learning_rate": 3.397037435263323e-06, "loss": 3.5035, "step": 5229 }, { "epoch": 0.63, "learning_rate": 3.3951375702912066e-06, "loss": 3.3597, "step": 5230 }, { "epoch": 0.63, "learning_rate": 3.3932379636064194e-06, "loss": 3.2934, "step": 5231 }, { "epoch": 0.63, "learning_rate": 3.3913386155146848e-06, "loss": 3.537, "step": 5232 }, { "epoch": 0.63, "learning_rate": 3.3894395263216855e-06, "loss": 3.4497, "step": 5233 }, { "epoch": 0.63, "learning_rate": 3.3875406963330614e-06, "loss": 3.3517, "step": 5234 }, { "epoch": 0.63, "learning_rate": 3.385642125854412e-06, "loss": 3.39, "step": 5235 }, { "epoch": 0.63, "learning_rate": 3.383743815191295e-06, "loss": 3.5113, "step": 5236 }, { "epoch": 0.63, "learning_rate": 3.3818457646492215e-06, "loss": 3.413, "step": 5237 }, { "epoch": 0.63, "learning_rate": 3.3799479745336674e-06, "loss": 3.3336, "step": 5238 }, { "epoch": 0.63, "learning_rate": 3.378050445150063e-06, "loss": 3.4102, "step": 5239 }, { "epoch": 0.63, "learning_rate": 3.376153176803798e-06, "loss": 3.2988, "step": 5240 }, { "epoch": 0.63, "learning_rate": 3.37425616980022e-06, "loss": 3.3966, "step": 5241 }, { "epoch": 0.63, "learning_rate": 3.3723594244446346e-06, "loss": 3.4456, "step": 5242 }, { "epoch": 0.63, "learning_rate": 3.370462941042304e-06, "loss": 3.3674, "step": 5243 }, { "epoch": 0.63, "learning_rate": 3.368566719898449e-06, "loss": 3.3319, "step": 5244 }, { "epoch": 0.63, "learning_rate": 3.36667076131825e-06, "loss": 3.3961, "step": 5245 }, { "epoch": 0.63, "learning_rate": 3.364775065606841e-06, "loss": 3.4695, "step": 5246 }, { "epoch": 0.63, "learning_rate": 3.362879633069318e-06, "loss": 3.3176, "step": 5247 }, { "epoch": 0.63, "learning_rate": 3.3609844640107324e-06, "loss": 3.4575, "step": 5248 }, { "epoch": 0.63, "learning_rate": 3.359089558736094e-06, "loss": 3.3141, "step": 5249 }, { "epoch": 0.63, "learning_rate": 3.357194917550368e-06, "loss": 3.3962, "step": 5250 }, { "epoch": 0.63, "learning_rate": 3.3553005407584805e-06, "loss": 3.3507, "step": 5251 }, { "epoch": 0.63, "learning_rate": 3.3534064286653134e-06, "loss": 3.4379, "step": 5252 }, { "epoch": 0.63, "learning_rate": 3.3515125815757057e-06, "loss": 3.484, "step": 5253 }, { "epoch": 0.63, "learning_rate": 3.3496189997944556e-06, "loss": 3.4099, "step": 5254 }, { "epoch": 0.63, "learning_rate": 3.3477256836263138e-06, "loss": 3.4017, "step": 5255 }, { "epoch": 0.63, "learning_rate": 3.3458326333759927e-06, "loss": 3.3691, "step": 5256 }, { "epoch": 0.63, "learning_rate": 3.3439398493481614e-06, "loss": 3.256, "step": 5257 }, { "epoch": 0.63, "learning_rate": 3.3420473318474457e-06, "loss": 3.3957, "step": 5258 }, { "epoch": 0.63, "learning_rate": 3.3401550811784273e-06, "loss": 3.3798, "step": 5259 }, { "epoch": 0.63, "learning_rate": 3.3382630976456464e-06, "loss": 3.4256, "step": 5260 }, { "epoch": 0.63, "learning_rate": 3.3363713815536e-06, "loss": 3.4196, "step": 5261 }, { "epoch": 0.63, "learning_rate": 3.3344799332067425e-06, "loss": 3.396, "step": 5262 }, { "epoch": 0.63, "learning_rate": 3.3325887529094838e-06, "loss": 3.4361, "step": 5263 }, { "epoch": 0.63, "learning_rate": 3.3306978409661892e-06, "loss": 3.478, "step": 5264 }, { "epoch": 0.63, "learning_rate": 3.3288071976811864e-06, "loss": 3.3221, "step": 5265 }, { "epoch": 0.63, "learning_rate": 3.3269168233587545e-06, "loss": 3.4872, "step": 5266 }, { "epoch": 0.63, "learning_rate": 3.3250267183031316e-06, "loss": 3.3632, "step": 5267 }, { "epoch": 0.63, "learning_rate": 3.3231368828185127e-06, "loss": 3.4232, "step": 5268 }, { "epoch": 0.63, "learning_rate": 3.3212473172090482e-06, "loss": 3.3673, "step": 5269 }, { "epoch": 0.63, "learning_rate": 3.3193580217788455e-06, "loss": 3.4402, "step": 5270 }, { "epoch": 0.63, "learning_rate": 3.3174689968319695e-06, "loss": 3.3374, "step": 5271 }, { "epoch": 0.63, "learning_rate": 3.3155802426724427e-06, "loss": 3.459, "step": 5272 }, { "epoch": 0.63, "learning_rate": 3.313691759604237e-06, "loss": 3.308, "step": 5273 }, { "epoch": 0.63, "learning_rate": 3.3118035479312882e-06, "loss": 3.4632, "step": 5274 }, { "epoch": 0.63, "learning_rate": 3.309915607957487e-06, "loss": 3.3825, "step": 5275 }, { "epoch": 0.63, "learning_rate": 3.3080279399866788e-06, "loss": 3.3241, "step": 5276 }, { "epoch": 0.63, "learning_rate": 3.306140544322666e-06, "loss": 3.4143, "step": 5277 }, { "epoch": 0.63, "learning_rate": 3.3042534212692055e-06, "loss": 3.4183, "step": 5278 }, { "epoch": 0.63, "learning_rate": 3.302366571130014e-06, "loss": 3.5689, "step": 5279 }, { "epoch": 0.63, "learning_rate": 3.3004799942087606e-06, "loss": 3.4145, "step": 5280 }, { "epoch": 0.63, "learning_rate": 3.2985936908090734e-06, "loss": 3.4248, "step": 5281 }, { "epoch": 0.63, "learning_rate": 3.2967076612345327e-06, "loss": 3.4171, "step": 5282 }, { "epoch": 0.63, "learning_rate": 3.2948219057886787e-06, "loss": 3.3864, "step": 5283 }, { "epoch": 0.63, "learning_rate": 3.2929364247750045e-06, "loss": 3.4296, "step": 5284 }, { "epoch": 0.63, "learning_rate": 3.291051218496961e-06, "loss": 3.2449, "step": 5285 }, { "epoch": 0.63, "learning_rate": 3.2891662872579554e-06, "loss": 3.4171, "step": 5286 }, { "epoch": 0.63, "learning_rate": 3.2872816313613475e-06, "loss": 3.3348, "step": 5287 }, { "epoch": 0.63, "learning_rate": 3.285397251110456e-06, "loss": 3.3293, "step": 5288 }, { "epoch": 0.63, "learning_rate": 3.283513146808553e-06, "loss": 3.3457, "step": 5289 }, { "epoch": 0.63, "learning_rate": 3.28162931875887e-06, "loss": 3.3753, "step": 5290 }, { "epoch": 0.64, "learning_rate": 3.2797457672645866e-06, "loss": 3.4517, "step": 5291 }, { "epoch": 0.64, "learning_rate": 3.2778624926288448e-06, "loss": 3.3991, "step": 5292 }, { "epoch": 0.64, "learning_rate": 3.2759794951547397e-06, "loss": 3.3362, "step": 5293 }, { "epoch": 0.64, "learning_rate": 3.2740967751453224e-06, "loss": 3.4439, "step": 5294 }, { "epoch": 0.64, "learning_rate": 3.272214332903597e-06, "loss": 3.3382, "step": 5295 }, { "epoch": 0.64, "learning_rate": 3.2703321687325264e-06, "loss": 3.4625, "step": 5296 }, { "epoch": 0.64, "learning_rate": 3.268450282935026e-06, "loss": 3.4339, "step": 5297 }, { "epoch": 0.64, "learning_rate": 3.2665686758139688e-06, "loss": 3.4369, "step": 5298 }, { "epoch": 0.64, "learning_rate": 3.2646873476721806e-06, "loss": 3.4548, "step": 5299 }, { "epoch": 0.64, "learning_rate": 3.2628062988124428e-06, "loss": 3.3219, "step": 5300 }, { "epoch": 0.64, "learning_rate": 3.260925529537492e-06, "loss": 3.3757, "step": 5301 }, { "epoch": 0.64, "learning_rate": 3.2590450401500218e-06, "loss": 3.2616, "step": 5302 }, { "epoch": 0.64, "learning_rate": 3.257164830952677e-06, "loss": 3.3718, "step": 5303 }, { "epoch": 0.64, "learning_rate": 3.2552849022480613e-06, "loss": 3.3673, "step": 5304 }, { "epoch": 0.64, "learning_rate": 3.25340525433873e-06, "loss": 3.4289, "step": 5305 }, { "epoch": 0.64, "learning_rate": 3.2515258875271947e-06, "loss": 3.3976, "step": 5306 }, { "epoch": 0.64, "learning_rate": 3.249646802115923e-06, "loss": 3.5037, "step": 5307 }, { "epoch": 0.64, "learning_rate": 3.2477679984073363e-06, "loss": 3.3746, "step": 5308 }, { "epoch": 0.64, "learning_rate": 3.245889476703806e-06, "loss": 3.3974, "step": 5309 }, { "epoch": 0.64, "learning_rate": 3.2440112373076654e-06, "loss": 3.4424, "step": 5310 }, { "epoch": 0.64, "learning_rate": 3.242133280521199e-06, "loss": 3.2664, "step": 5311 }, { "epoch": 0.64, "learning_rate": 3.240255606646645e-06, "loss": 3.3728, "step": 5312 }, { "epoch": 0.64, "learning_rate": 3.2383782159861975e-06, "loss": 3.4147, "step": 5313 }, { "epoch": 0.64, "learning_rate": 3.2365011088420063e-06, "loss": 3.4124, "step": 5314 }, { "epoch": 0.64, "learning_rate": 3.2346242855161724e-06, "loss": 3.3915, "step": 5315 }, { "epoch": 0.64, "learning_rate": 3.2327477463107524e-06, "loss": 3.462, "step": 5316 }, { "epoch": 0.64, "learning_rate": 3.2308714915277607e-06, "loss": 3.3757, "step": 5317 }, { "epoch": 0.64, "learning_rate": 3.2289955214691577e-06, "loss": 3.4183, "step": 5318 }, { "epoch": 0.64, "learning_rate": 3.227119836436865e-06, "loss": 3.463, "step": 5319 }, { "epoch": 0.64, "learning_rate": 3.2252444367327567e-06, "loss": 3.3276, "step": 5320 }, { "epoch": 0.64, "learning_rate": 3.2233693226586603e-06, "loss": 3.3555, "step": 5321 }, { "epoch": 0.64, "learning_rate": 3.2214944945163573e-06, "loss": 3.4331, "step": 5322 }, { "epoch": 0.64, "learning_rate": 3.2196199526075835e-06, "loss": 3.4177, "step": 5323 }, { "epoch": 0.64, "learning_rate": 3.217745697234028e-06, "loss": 3.4113, "step": 5324 }, { "epoch": 0.64, "learning_rate": 3.2158717286973363e-06, "loss": 3.4014, "step": 5325 }, { "epoch": 0.64, "learning_rate": 3.213998047299105e-06, "loss": 3.4029, "step": 5326 }, { "epoch": 0.64, "learning_rate": 3.212124653340884e-06, "loss": 3.3505, "step": 5327 }, { "epoch": 0.64, "learning_rate": 3.210251547124179e-06, "loss": 3.3509, "step": 5328 }, { "epoch": 0.64, "learning_rate": 3.208378728950449e-06, "loss": 3.5064, "step": 5329 }, { "epoch": 0.64, "learning_rate": 3.206506199121105e-06, "loss": 3.4456, "step": 5330 }, { "epoch": 0.64, "learning_rate": 3.204633957937514e-06, "loss": 3.4468, "step": 5331 }, { "epoch": 0.64, "learning_rate": 3.2027620057009958e-06, "loss": 3.5008, "step": 5332 }, { "epoch": 0.64, "learning_rate": 3.2008903427128223e-06, "loss": 3.3903, "step": 5333 }, { "epoch": 0.64, "learning_rate": 3.19901896927422e-06, "loss": 3.3584, "step": 5334 }, { "epoch": 0.64, "learning_rate": 3.1971478856863714e-06, "loss": 3.3827, "step": 5335 }, { "epoch": 0.64, "learning_rate": 3.1952770922504046e-06, "loss": 3.3692, "step": 5336 }, { "epoch": 0.64, "learning_rate": 3.193406589267408e-06, "loss": 3.4505, "step": 5337 }, { "epoch": 0.64, "learning_rate": 3.1915363770384223e-06, "loss": 3.4689, "step": 5338 }, { "epoch": 0.64, "learning_rate": 3.1896664558644396e-06, "loss": 3.3364, "step": 5339 }, { "epoch": 0.64, "learning_rate": 3.187796826046406e-06, "loss": 3.432, "step": 5340 }, { "epoch": 0.64, "learning_rate": 3.1859274878852207e-06, "loss": 3.4326, "step": 5341 }, { "epoch": 0.64, "learning_rate": 3.1840584416817354e-06, "loss": 3.4211, "step": 5342 }, { "epoch": 0.64, "learning_rate": 3.182189687736756e-06, "loss": 3.4963, "step": 5343 }, { "epoch": 0.64, "learning_rate": 3.1803212263510418e-06, "loss": 3.4106, "step": 5344 }, { "epoch": 0.64, "learning_rate": 3.178453057825301e-06, "loss": 3.3501, "step": 5345 }, { "epoch": 0.64, "learning_rate": 3.176585182460199e-06, "loss": 3.3251, "step": 5346 }, { "epoch": 0.64, "learning_rate": 3.174717600556353e-06, "loss": 3.3989, "step": 5347 }, { "epoch": 0.64, "learning_rate": 3.1728503124143316e-06, "loss": 3.436, "step": 5348 }, { "epoch": 0.64, "learning_rate": 3.1709833183346573e-06, "loss": 3.4325, "step": 5349 }, { "epoch": 0.64, "learning_rate": 3.1691166186178057e-06, "loss": 3.3361, "step": 5350 }, { "epoch": 0.64, "learning_rate": 3.1672502135642034e-06, "loss": 3.3875, "step": 5351 }, { "epoch": 0.64, "learning_rate": 3.1653841034742316e-06, "loss": 3.3552, "step": 5352 }, { "epoch": 0.64, "learning_rate": 3.1635182886482243e-06, "loss": 3.4406, "step": 5353 }, { "epoch": 0.64, "learning_rate": 3.161652769386463e-06, "loss": 3.335, "step": 5354 }, { "epoch": 0.64, "learning_rate": 3.1597875459891857e-06, "loss": 3.3017, "step": 5355 }, { "epoch": 0.64, "learning_rate": 3.1579226187565846e-06, "loss": 3.4199, "step": 5356 }, { "epoch": 0.64, "learning_rate": 3.1560579879888008e-06, "loss": 3.4643, "step": 5357 }, { "epoch": 0.64, "learning_rate": 3.15419365398593e-06, "loss": 3.4418, "step": 5358 }, { "epoch": 0.64, "learning_rate": 3.1523296170480167e-06, "loss": 3.326, "step": 5359 }, { "epoch": 0.64, "learning_rate": 3.1504658774750618e-06, "loss": 3.3876, "step": 5360 }, { "epoch": 0.64, "learning_rate": 3.148602435567016e-06, "loss": 3.3508, "step": 5361 }, { "epoch": 0.64, "learning_rate": 3.1467392916237832e-06, "loss": 3.4191, "step": 5362 }, { "epoch": 0.64, "learning_rate": 3.144876445945217e-06, "loss": 3.3674, "step": 5363 }, { "epoch": 0.64, "learning_rate": 3.1430138988311253e-06, "loss": 3.3347, "step": 5364 }, { "epoch": 0.64, "learning_rate": 3.1411516505812686e-06, "loss": 3.4564, "step": 5365 }, { "epoch": 0.64, "learning_rate": 3.139289701495356e-06, "loss": 3.442, "step": 5366 }, { "epoch": 0.64, "learning_rate": 3.1374280518730526e-06, "loss": 3.3533, "step": 5367 }, { "epoch": 0.64, "learning_rate": 3.135566702013971e-06, "loss": 3.4427, "step": 5368 }, { "epoch": 0.64, "learning_rate": 3.13370565221768e-06, "loss": 3.3375, "step": 5369 }, { "epoch": 0.64, "learning_rate": 3.1318449027836963e-06, "loss": 3.3746, "step": 5370 }, { "epoch": 0.64, "learning_rate": 3.129984454011492e-06, "loss": 3.3323, "step": 5371 }, { "epoch": 0.64, "learning_rate": 3.128124306200485e-06, "loss": 3.3968, "step": 5372 }, { "epoch": 0.64, "learning_rate": 3.1262644596500503e-06, "loss": 3.3345, "step": 5373 }, { "epoch": 0.65, "learning_rate": 3.1244049146595116e-06, "loss": 3.3201, "step": 5374 }, { "epoch": 0.65, "learning_rate": 3.1225456715281466e-06, "loss": 3.4864, "step": 5375 }, { "epoch": 0.65, "learning_rate": 3.1206867305551818e-06, "loss": 3.3829, "step": 5376 }, { "epoch": 0.65, "learning_rate": 3.118828092039796e-06, "loss": 3.4191, "step": 5377 }, { "epoch": 0.65, "learning_rate": 3.11696975628112e-06, "loss": 3.3816, "step": 5378 }, { "epoch": 0.65, "learning_rate": 3.1151117235782346e-06, "loss": 3.5198, "step": 5379 }, { "epoch": 0.65, "learning_rate": 3.113253994230174e-06, "loss": 3.3795, "step": 5380 }, { "epoch": 0.65, "learning_rate": 3.1113965685359192e-06, "loss": 3.4225, "step": 5381 }, { "epoch": 0.65, "learning_rate": 3.1095394467944074e-06, "loss": 3.309, "step": 5382 }, { "epoch": 0.65, "learning_rate": 3.1076826293045235e-06, "loss": 3.4166, "step": 5383 }, { "epoch": 0.65, "learning_rate": 3.1058261163651054e-06, "loss": 3.3412, "step": 5384 }, { "epoch": 0.65, "learning_rate": 3.103969908274941e-06, "loss": 3.46, "step": 5385 }, { "epoch": 0.65, "learning_rate": 3.1021140053327685e-06, "loss": 3.442, "step": 5386 }, { "epoch": 0.65, "learning_rate": 3.1002584078372787e-06, "loss": 3.4808, "step": 5387 }, { "epoch": 0.65, "learning_rate": 3.0984031160871132e-06, "loss": 3.3234, "step": 5388 }, { "epoch": 0.65, "learning_rate": 3.0965481303808635e-06, "loss": 3.4036, "step": 5389 }, { "epoch": 0.65, "learning_rate": 3.094693451017069e-06, "loss": 3.3315, "step": 5390 }, { "epoch": 0.65, "learning_rate": 3.0928390782942253e-06, "loss": 3.3846, "step": 5391 }, { "epoch": 0.65, "learning_rate": 3.0909850125107744e-06, "loss": 3.4833, "step": 5392 }, { "epoch": 0.65, "learning_rate": 3.089131253965112e-06, "loss": 3.3896, "step": 5393 }, { "epoch": 0.65, "learning_rate": 3.087277802955582e-06, "loss": 3.4252, "step": 5394 }, { "epoch": 0.65, "learning_rate": 3.0854246597804806e-06, "loss": 3.3906, "step": 5395 }, { "epoch": 0.65, "learning_rate": 3.083571824738053e-06, "loss": 3.3472, "step": 5396 }, { "epoch": 0.65, "learning_rate": 3.081719298126495e-06, "loss": 3.3998, "step": 5397 }, { "epoch": 0.65, "learning_rate": 3.0798670802439543e-06, "loss": 3.3608, "step": 5398 }, { "epoch": 0.65, "learning_rate": 3.0780151713885265e-06, "loss": 3.4609, "step": 5399 }, { "epoch": 0.65, "learning_rate": 3.0761635718582582e-06, "loss": 3.3541, "step": 5400 }, { "epoch": 0.65, "learning_rate": 3.0743122819511485e-06, "loss": 3.3139, "step": 5401 }, { "epoch": 0.65, "learning_rate": 3.072461301965144e-06, "loss": 3.4133, "step": 5402 }, { "epoch": 0.65, "learning_rate": 3.0706106321981417e-06, "loss": 3.3776, "step": 5403 }, { "epoch": 0.65, "learning_rate": 3.0687602729479904e-06, "loss": 3.3863, "step": 5404 }, { "epoch": 0.65, "learning_rate": 3.066910224512487e-06, "loss": 3.3763, "step": 5405 }, { "epoch": 0.65, "learning_rate": 3.0650604871893796e-06, "loss": 3.4674, "step": 5406 }, { "epoch": 0.65, "learning_rate": 3.0632110612763677e-06, "loss": 3.3869, "step": 5407 }, { "epoch": 0.65, "learning_rate": 3.0613619470710955e-06, "loss": 3.4472, "step": 5408 }, { "epoch": 0.65, "learning_rate": 3.0595131448711606e-06, "loss": 3.4172, "step": 5409 }, { "epoch": 0.65, "learning_rate": 3.0576646549741106e-06, "loss": 3.4367, "step": 5410 }, { "epoch": 0.65, "learning_rate": 3.0558164776774434e-06, "loss": 3.4245, "step": 5411 }, { "epoch": 0.65, "learning_rate": 3.0539686132786064e-06, "loss": 3.3998, "step": 5412 }, { "epoch": 0.65, "learning_rate": 3.052121062074993e-06, "loss": 3.4904, "step": 5413 }, { "epoch": 0.65, "learning_rate": 3.0502738243639507e-06, "loss": 3.4684, "step": 5414 }, { "epoch": 0.65, "learning_rate": 3.0484269004427747e-06, "loss": 3.416, "step": 5415 }, { "epoch": 0.65, "learning_rate": 3.046580290608712e-06, "loss": 3.5575, "step": 5416 }, { "epoch": 0.65, "learning_rate": 3.0447339951589517e-06, "loss": 3.267, "step": 5417 }, { "epoch": 0.65, "learning_rate": 3.0428880143906415e-06, "loss": 3.4043, "step": 5418 }, { "epoch": 0.65, "learning_rate": 3.041042348600872e-06, "loss": 3.3604, "step": 5419 }, { "epoch": 0.65, "learning_rate": 3.0391969980866874e-06, "loss": 3.4198, "step": 5420 }, { "epoch": 0.65, "learning_rate": 3.0373519631450784e-06, "loss": 3.494, "step": 5421 }, { "epoch": 0.65, "learning_rate": 3.035507244072986e-06, "loss": 3.4154, "step": 5422 }, { "epoch": 0.65, "learning_rate": 3.0336628411673e-06, "loss": 3.3907, "step": 5423 }, { "epoch": 0.65, "learning_rate": 3.0318187547248603e-06, "loss": 3.4607, "step": 5424 }, { "epoch": 0.65, "learning_rate": 3.0299749850424554e-06, "loss": 3.3869, "step": 5425 }, { "epoch": 0.65, "learning_rate": 3.02813153241682e-06, "loss": 3.3133, "step": 5426 }, { "epoch": 0.65, "learning_rate": 3.0262883971446426e-06, "loss": 3.4211, "step": 5427 }, { "epoch": 0.65, "learning_rate": 3.0244455795225567e-06, "loss": 3.4198, "step": 5428 }, { "epoch": 0.65, "learning_rate": 3.0226030798471476e-06, "loss": 3.3392, "step": 5429 }, { "epoch": 0.65, "learning_rate": 3.0207608984149477e-06, "loss": 3.4142, "step": 5430 }, { "epoch": 0.65, "learning_rate": 3.0189190355224378e-06, "loss": 3.4805, "step": 5431 }, { "epoch": 0.65, "learning_rate": 3.0170774914660495e-06, "loss": 3.3961, "step": 5432 }, { "epoch": 0.65, "learning_rate": 3.0152362665421604e-06, "loss": 3.4294, "step": 5433 }, { "epoch": 0.65, "learning_rate": 3.013395361047101e-06, "loss": 3.4388, "step": 5434 }, { "epoch": 0.65, "learning_rate": 3.0115547752771433e-06, "loss": 3.3863, "step": 5435 }, { "epoch": 0.65, "learning_rate": 3.009714509528514e-06, "loss": 3.3124, "step": 5436 }, { "epoch": 0.65, "learning_rate": 3.007874564097386e-06, "loss": 3.4656, "step": 5437 }, { "epoch": 0.65, "learning_rate": 3.006034939279882e-06, "loss": 3.415, "step": 5438 }, { "epoch": 0.65, "learning_rate": 3.0041956353720714e-06, "loss": 3.3275, "step": 5439 }, { "epoch": 0.65, "learning_rate": 3.0023566526699727e-06, "loss": 3.3788, "step": 5440 }, { "epoch": 0.65, "learning_rate": 3.0005179914695528e-06, "loss": 3.3199, "step": 5441 }, { "epoch": 0.65, "learning_rate": 2.998679652066726e-06, "loss": 3.4664, "step": 5442 }, { "epoch": 0.65, "learning_rate": 2.9968416347573583e-06, "loss": 3.3987, "step": 5443 }, { "epoch": 0.65, "learning_rate": 2.995003939837257e-06, "loss": 3.4577, "step": 5444 }, { "epoch": 0.65, "learning_rate": 2.9931665676021836e-06, "loss": 3.5223, "step": 5445 }, { "epoch": 0.65, "learning_rate": 2.9913295183478466e-06, "loss": 3.3414, "step": 5446 }, { "epoch": 0.65, "learning_rate": 2.9894927923698997e-06, "loss": 3.4267, "step": 5447 }, { "epoch": 0.65, "learning_rate": 2.9876563899639475e-06, "loss": 3.328, "step": 5448 }, { "epoch": 0.65, "learning_rate": 2.9858203114255413e-06, "loss": 3.3991, "step": 5449 }, { "epoch": 0.65, "learning_rate": 2.983984557050181e-06, "loss": 3.485, "step": 5450 }, { "epoch": 0.65, "learning_rate": 2.9821491271333126e-06, "loss": 3.4162, "step": 5451 }, { "epoch": 0.65, "learning_rate": 2.9803140219703343e-06, "loss": 3.3615, "step": 5452 }, { "epoch": 0.65, "learning_rate": 2.9784792418565844e-06, "loss": 3.3544, "step": 5453 }, { "epoch": 0.65, "learning_rate": 2.976644787087355e-06, "loss": 3.4685, "step": 5454 }, { "epoch": 0.65, "learning_rate": 2.974810657957884e-06, "loss": 3.4305, "step": 5455 }, { "epoch": 0.65, "learning_rate": 2.9729768547633574e-06, "loss": 3.3434, "step": 5456 }, { "epoch": 0.66, "learning_rate": 2.9711433777989087e-06, "loss": 3.3656, "step": 5457 }, { "epoch": 0.66, "learning_rate": 2.9693102273596176e-06, "loss": 3.4445, "step": 5458 }, { "epoch": 0.66, "learning_rate": 2.967477403740513e-06, "loss": 3.4136, "step": 5459 }, { "epoch": 0.66, "learning_rate": 2.96564490723657e-06, "loss": 3.4494, "step": 5460 }, { "epoch": 0.66, "learning_rate": 2.963812738142713e-06, "loss": 3.4802, "step": 5461 }, { "epoch": 0.66, "learning_rate": 2.96198089675381e-06, "loss": 3.387, "step": 5462 }, { "epoch": 0.66, "learning_rate": 2.960149383364679e-06, "loss": 3.2692, "step": 5463 }, { "epoch": 0.66, "learning_rate": 2.9583181982700844e-06, "loss": 3.3923, "step": 5464 }, { "epoch": 0.66, "learning_rate": 2.9564873417647387e-06, "loss": 3.2677, "step": 5465 }, { "epoch": 0.66, "learning_rate": 2.9546568141433007e-06, "loss": 3.3555, "step": 5466 }, { "epoch": 0.66, "learning_rate": 2.9528266157003764e-06, "loss": 3.4105, "step": 5467 }, { "epoch": 0.66, "learning_rate": 2.950996746730519e-06, "loss": 3.4181, "step": 5468 }, { "epoch": 0.66, "learning_rate": 2.9491672075282278e-06, "loss": 3.3279, "step": 5469 }, { "epoch": 0.66, "learning_rate": 2.9473379983879522e-06, "loss": 3.4274, "step": 5470 }, { "epoch": 0.66, "learning_rate": 2.945509119604082e-06, "loss": 3.4375, "step": 5471 }, { "epoch": 0.66, "learning_rate": 2.9436805714709603e-06, "loss": 3.3972, "step": 5472 }, { "epoch": 0.66, "learning_rate": 2.9418523542828738e-06, "loss": 3.3994, "step": 5473 }, { "epoch": 0.66, "learning_rate": 2.9400244683340563e-06, "loss": 3.4279, "step": 5474 }, { "epoch": 0.66, "learning_rate": 2.9381969139186888e-06, "loss": 3.3216, "step": 5475 }, { "epoch": 0.66, "learning_rate": 2.9363696913309004e-06, "loss": 3.3672, "step": 5476 }, { "epoch": 0.66, "learning_rate": 2.9345428008647625e-06, "loss": 3.5012, "step": 5477 }, { "epoch": 0.66, "learning_rate": 2.9327162428142987e-06, "loss": 3.4645, "step": 5478 }, { "epoch": 0.66, "learning_rate": 2.9308900174734743e-06, "loss": 3.3926, "step": 5479 }, { "epoch": 0.66, "learning_rate": 2.9290641251362027e-06, "loss": 3.3891, "step": 5480 }, { "epoch": 0.66, "learning_rate": 2.927238566096344e-06, "loss": 3.3915, "step": 5481 }, { "epoch": 0.66, "learning_rate": 2.925413340647706e-06, "loss": 3.1913, "step": 5482 }, { "epoch": 0.66, "learning_rate": 2.9235884490840394e-06, "loss": 3.457, "step": 5483 }, { "epoch": 0.66, "learning_rate": 2.921763891699044e-06, "loss": 3.2935, "step": 5484 }, { "epoch": 0.66, "learning_rate": 2.919939668786365e-06, "loss": 3.3309, "step": 5485 }, { "epoch": 0.66, "learning_rate": 2.9181157806395947e-06, "loss": 3.3684, "step": 5486 }, { "epoch": 0.66, "learning_rate": 2.9162922275522687e-06, "loss": 3.3105, "step": 5487 }, { "epoch": 0.66, "learning_rate": 2.914469009817871e-06, "loss": 3.4599, "step": 5488 }, { "epoch": 0.66, "learning_rate": 2.912646127729831e-06, "loss": 3.3021, "step": 5489 }, { "epoch": 0.66, "learning_rate": 2.9108235815815254e-06, "loss": 3.461, "step": 5490 }, { "epoch": 0.66, "learning_rate": 2.9090013716662744e-06, "loss": 3.3524, "step": 5491 }, { "epoch": 0.66, "learning_rate": 2.9071794982773464e-06, "loss": 3.4045, "step": 5492 }, { "epoch": 0.66, "learning_rate": 2.905357961707954e-06, "loss": 3.4049, "step": 5493 }, { "epoch": 0.66, "learning_rate": 2.9035367622512558e-06, "loss": 3.3022, "step": 5494 }, { "epoch": 0.66, "learning_rate": 2.901715900200357e-06, "loss": 3.4308, "step": 5495 }, { "epoch": 0.66, "learning_rate": 2.899895375848308e-06, "loss": 3.3457, "step": 5496 }, { "epoch": 0.66, "learning_rate": 2.8980751894881065e-06, "loss": 3.4317, "step": 5497 }, { "epoch": 0.66, "learning_rate": 2.8962553414126903e-06, "loss": 3.3633, "step": 5498 }, { "epoch": 0.66, "learning_rate": 2.894435831914949e-06, "loss": 3.2746, "step": 5499 }, { "epoch": 0.66, "learning_rate": 2.8926166612877153e-06, "loss": 3.3728, "step": 5500 }, { "epoch": 0.66, "learning_rate": 2.890797829823767e-06, "loss": 3.3879, "step": 5501 }, { "epoch": 0.66, "learning_rate": 2.8889793378158284e-06, "loss": 3.4611, "step": 5502 }, { "epoch": 0.66, "learning_rate": 2.8871611855565673e-06, "loss": 3.3971, "step": 5503 }, { "epoch": 0.66, "learning_rate": 2.8853433733385983e-06, "loss": 3.3616, "step": 5504 }, { "epoch": 0.66, "learning_rate": 2.883525901454482e-06, "loss": 3.3689, "step": 5505 }, { "epoch": 0.66, "learning_rate": 2.8817087701967244e-06, "loss": 3.4951, "step": 5506 }, { "epoch": 0.66, "learning_rate": 2.8798919798577716e-06, "loss": 3.3669, "step": 5507 }, { "epoch": 0.66, "learning_rate": 2.878075530730021e-06, "loss": 3.3123, "step": 5508 }, { "epoch": 0.66, "learning_rate": 2.8762594231058115e-06, "loss": 3.3258, "step": 5509 }, { "epoch": 0.66, "learning_rate": 2.8744436572774302e-06, "loss": 3.391, "step": 5510 }, { "epoch": 0.66, "learning_rate": 2.872628233537107e-06, "loss": 3.412, "step": 5511 }, { "epoch": 0.66, "learning_rate": 2.8708131521770154e-06, "loss": 3.5058, "step": 5512 }, { "epoch": 0.66, "learning_rate": 2.8689984134892768e-06, "loss": 3.4329, "step": 5513 }, { "epoch": 0.66, "learning_rate": 2.867184017765956e-06, "loss": 3.3553, "step": 5514 }, { "epoch": 0.66, "learning_rate": 2.865369965299063e-06, "loss": 3.3857, "step": 5515 }, { "epoch": 0.66, "learning_rate": 2.8635562563805517e-06, "loss": 3.3418, "step": 5516 }, { "epoch": 0.66, "learning_rate": 2.8617428913023216e-06, "loss": 3.5591, "step": 5517 }, { "epoch": 0.66, "learning_rate": 2.859929870356216e-06, "loss": 3.4387, "step": 5518 }, { "epoch": 0.66, "learning_rate": 2.8581171938340244e-06, "loss": 3.3643, "step": 5519 }, { "epoch": 0.66, "learning_rate": 2.856304862027479e-06, "loss": 3.2951, "step": 5520 }, { "epoch": 0.66, "learning_rate": 2.8544928752282576e-06, "loss": 3.3803, "step": 5521 }, { "epoch": 0.66, "learning_rate": 2.852681233727982e-06, "loss": 3.3496, "step": 5522 }, { "epoch": 0.66, "learning_rate": 2.850869937818219e-06, "loss": 3.4615, "step": 5523 }, { "epoch": 0.66, "learning_rate": 2.849058987790481e-06, "loss": 3.3776, "step": 5524 }, { "epoch": 0.66, "learning_rate": 2.84724838393622e-06, "loss": 3.3713, "step": 5525 }, { "epoch": 0.66, "learning_rate": 2.8454381265468368e-06, "loss": 3.3247, "step": 5526 }, { "epoch": 0.66, "learning_rate": 2.843628215913675e-06, "loss": 3.3445, "step": 5527 }, { "epoch": 0.66, "learning_rate": 2.8418186523280223e-06, "loss": 3.4783, "step": 5528 }, { "epoch": 0.66, "learning_rate": 2.8400094360811113e-06, "loss": 3.3444, "step": 5529 }, { "epoch": 0.66, "learning_rate": 2.838200567464119e-06, "loss": 3.2933, "step": 5530 }, { "epoch": 0.66, "learning_rate": 2.8363920467681634e-06, "loss": 3.3696, "step": 5531 }, { "epoch": 0.66, "learning_rate": 2.83458387428431e-06, "loss": 3.4022, "step": 5532 }, { "epoch": 0.66, "learning_rate": 2.8327760503035685e-06, "loss": 3.446, "step": 5533 }, { "epoch": 0.66, "learning_rate": 2.830968575116888e-06, "loss": 3.3011, "step": 5534 }, { "epoch": 0.66, "learning_rate": 2.8291614490151655e-06, "loss": 3.3451, "step": 5535 }, { "epoch": 0.66, "learning_rate": 2.8273546722892403e-06, "loss": 3.3573, "step": 5536 }, { "epoch": 0.66, "learning_rate": 2.8255482452298965e-06, "loss": 3.4452, "step": 5537 }, { "epoch": 0.66, "learning_rate": 2.823742168127862e-06, "loss": 3.2956, "step": 5538 }, { "epoch": 0.66, "learning_rate": 2.8219364412738063e-06, "loss": 3.3995, "step": 5539 }, { "epoch": 0.66, "learning_rate": 2.8201310649583457e-06, "loss": 3.4205, "step": 5540 }, { "epoch": 0.67, "learning_rate": 2.8183260394720365e-06, "loss": 3.3452, "step": 5541 }, { "epoch": 0.67, "learning_rate": 2.8165213651053836e-06, "loss": 3.3187, "step": 5542 }, { "epoch": 0.67, "learning_rate": 2.814717042148827e-06, "loss": 3.3892, "step": 5543 }, { "epoch": 0.67, "learning_rate": 2.8129130708927587e-06, "loss": 3.4232, "step": 5544 }, { "epoch": 0.67, "learning_rate": 2.8111094516275105e-06, "loss": 3.4294, "step": 5545 }, { "epoch": 0.67, "learning_rate": 2.8093061846433566e-06, "loss": 3.4067, "step": 5546 }, { "epoch": 0.67, "learning_rate": 2.8075032702305166e-06, "loss": 3.3522, "step": 5547 }, { "epoch": 0.67, "learning_rate": 2.805700708679152e-06, "loss": 3.3586, "step": 5548 }, { "epoch": 0.67, "learning_rate": 2.8038985002793684e-06, "loss": 3.4326, "step": 5549 }, { "epoch": 0.67, "learning_rate": 2.8020966453212138e-06, "loss": 3.3695, "step": 5550 }, { "epoch": 0.67, "learning_rate": 2.800295144094679e-06, "loss": 3.2854, "step": 5551 }, { "epoch": 0.67, "learning_rate": 2.7984939968896995e-06, "loss": 3.4361, "step": 5552 }, { "epoch": 0.67, "learning_rate": 2.7966932039961525e-06, "loss": 3.3879, "step": 5553 }, { "epoch": 0.67, "learning_rate": 2.794892765703858e-06, "loss": 3.4379, "step": 5554 }, { "epoch": 0.67, "learning_rate": 2.79309268230258e-06, "loss": 3.4113, "step": 5555 }, { "epoch": 0.67, "learning_rate": 2.7912929540820245e-06, "loss": 3.3455, "step": 5556 }, { "epoch": 0.67, "learning_rate": 2.78949358133184e-06, "loss": 3.3879, "step": 5557 }, { "epoch": 0.67, "learning_rate": 2.7876945643416196e-06, "loss": 3.3356, "step": 5558 }, { "epoch": 0.67, "learning_rate": 2.7858959034008977e-06, "loss": 3.43, "step": 5559 }, { "epoch": 0.67, "learning_rate": 2.7840975987991524e-06, "loss": 3.3158, "step": 5560 }, { "epoch": 0.67, "learning_rate": 2.7822996508258015e-06, "loss": 3.4687, "step": 5561 }, { "epoch": 0.67, "learning_rate": 2.7805020597702086e-06, "loss": 3.4727, "step": 5562 }, { "epoch": 0.67, "learning_rate": 2.7787048259216788e-06, "loss": 3.4368, "step": 5563 }, { "epoch": 0.67, "learning_rate": 2.7769079495694606e-06, "loss": 3.4015, "step": 5564 }, { "epoch": 0.67, "learning_rate": 2.7751114310027427e-06, "loss": 3.4059, "step": 5565 }, { "epoch": 0.67, "learning_rate": 2.773315270510659e-06, "loss": 3.4052, "step": 5566 }, { "epoch": 0.67, "learning_rate": 2.7715194683822846e-06, "loss": 3.391, "step": 5567 }, { "epoch": 0.67, "learning_rate": 2.7697240249066347e-06, "loss": 3.4372, "step": 5568 }, { "epoch": 0.67, "learning_rate": 2.7679289403726726e-06, "loss": 3.4282, "step": 5569 }, { "epoch": 0.67, "learning_rate": 2.7661342150692964e-06, "loss": 3.4024, "step": 5570 }, { "epoch": 0.67, "learning_rate": 2.7643398492853514e-06, "loss": 3.4165, "step": 5571 }, { "epoch": 0.67, "learning_rate": 2.762545843309623e-06, "loss": 3.3685, "step": 5572 }, { "epoch": 0.67, "learning_rate": 2.7607521974308404e-06, "loss": 3.4683, "step": 5573 }, { "epoch": 0.67, "learning_rate": 2.758958911937673e-06, "loss": 3.4163, "step": 5574 }, { "epoch": 0.67, "learning_rate": 2.757165987118734e-06, "loss": 3.379, "step": 5575 }, { "epoch": 0.67, "learning_rate": 2.7553734232625774e-06, "loss": 3.4482, "step": 5576 }, { "epoch": 0.67, "learning_rate": 2.753581220657698e-06, "loss": 3.3698, "step": 5577 }, { "epoch": 0.67, "learning_rate": 2.751789379592537e-06, "loss": 3.4333, "step": 5578 }, { "epoch": 0.67, "learning_rate": 2.74999790035547e-06, "loss": 3.3537, "step": 5579 }, { "epoch": 0.67, "learning_rate": 2.7482067832348203e-06, "loss": 3.3257, "step": 5580 }, { "epoch": 0.67, "learning_rate": 2.746416028518851e-06, "loss": 3.3686, "step": 5581 }, { "epoch": 0.67, "learning_rate": 2.744625636495768e-06, "loss": 3.4123, "step": 5582 }, { "epoch": 0.67, "learning_rate": 2.7428356074537165e-06, "loss": 3.4344, "step": 5583 }, { "epoch": 0.67, "learning_rate": 2.7410459416807856e-06, "loss": 3.4965, "step": 5584 }, { "epoch": 0.67, "learning_rate": 2.7392566394650043e-06, "loss": 3.4445, "step": 5585 }, { "epoch": 0.67, "learning_rate": 2.7374677010943447e-06, "loss": 3.4097, "step": 5586 }, { "epoch": 0.67, "learning_rate": 2.7356791268567186e-06, "loss": 3.3751, "step": 5587 }, { "epoch": 0.67, "learning_rate": 2.7338909170399797e-06, "loss": 3.4714, "step": 5588 }, { "epoch": 0.67, "learning_rate": 2.732103071931924e-06, "loss": 3.3017, "step": 5589 }, { "epoch": 0.67, "learning_rate": 2.7303155918202873e-06, "loss": 3.4168, "step": 5590 }, { "epoch": 0.67, "learning_rate": 2.728528476992749e-06, "loss": 3.3949, "step": 5591 }, { "epoch": 0.67, "learning_rate": 2.7267417277369275e-06, "loss": 3.4234, "step": 5592 }, { "epoch": 0.67, "learning_rate": 2.7249553443403827e-06, "loss": 3.4353, "step": 5593 }, { "epoch": 0.67, "learning_rate": 2.7231693270906163e-06, "loss": 3.4776, "step": 5594 }, { "epoch": 0.67, "learning_rate": 2.72138367627507e-06, "loss": 3.4397, "step": 5595 }, { "epoch": 0.67, "learning_rate": 2.7195983921811294e-06, "loss": 3.2888, "step": 5596 }, { "epoch": 0.67, "learning_rate": 2.7178134750961187e-06, "loss": 3.3753, "step": 5597 }, { "epoch": 0.67, "learning_rate": 2.7160289253073003e-06, "loss": 3.3935, "step": 5598 }, { "epoch": 0.67, "learning_rate": 2.7142447431018824e-06, "loss": 3.4115, "step": 5599 }, { "epoch": 0.67, "learning_rate": 2.712460928767013e-06, "loss": 3.32, "step": 5600 }, { "epoch": 0.67, "learning_rate": 2.7106774825897784e-06, "loss": 3.371, "step": 5601 }, { "epoch": 0.67, "learning_rate": 2.7088944048572084e-06, "loss": 3.5224, "step": 5602 }, { "epoch": 0.67, "learning_rate": 2.7071116958562716e-06, "loss": 3.413, "step": 5603 }, { "epoch": 0.67, "learning_rate": 2.7053293558738793e-06, "loss": 3.4405, "step": 5604 }, { "epoch": 0.67, "learning_rate": 2.703547385196881e-06, "loss": 3.3996, "step": 5605 }, { "epoch": 0.67, "learning_rate": 2.7017657841120697e-06, "loss": 3.2366, "step": 5606 }, { "epoch": 0.67, "learning_rate": 2.699984552906174e-06, "loss": 3.4095, "step": 5607 }, { "epoch": 0.67, "learning_rate": 2.698203691865868e-06, "loss": 3.4192, "step": 5608 }, { "epoch": 0.67, "learning_rate": 2.696423201277765e-06, "loss": 3.4143, "step": 5609 }, { "epoch": 0.67, "learning_rate": 2.6946430814284156e-06, "loss": 3.3301, "step": 5610 }, { "epoch": 0.67, "learning_rate": 2.6928633326043157e-06, "loss": 3.3794, "step": 5611 }, { "epoch": 0.67, "learning_rate": 2.691083955091898e-06, "loss": 3.4638, "step": 5612 }, { "epoch": 0.67, "learning_rate": 2.689304949177536e-06, "loss": 3.408, "step": 5613 }, { "epoch": 0.67, "learning_rate": 2.6875263151475438e-06, "loss": 3.4081, "step": 5614 }, { "epoch": 0.67, "learning_rate": 2.6857480532881764e-06, "loss": 3.3959, "step": 5615 }, { "epoch": 0.67, "learning_rate": 2.6839701638856275e-06, "loss": 3.2933, "step": 5616 }, { "epoch": 0.67, "learning_rate": 2.6821926472260308e-06, "loss": 3.3767, "step": 5617 }, { "epoch": 0.67, "learning_rate": 2.6804155035954627e-06, "loss": 3.3233, "step": 5618 }, { "epoch": 0.67, "learning_rate": 2.6786387332799358e-06, "loss": 3.4312, "step": 5619 }, { "epoch": 0.67, "learning_rate": 2.6768623365654046e-06, "loss": 3.3997, "step": 5620 }, { "epoch": 0.67, "learning_rate": 2.675086313737764e-06, "loss": 3.4021, "step": 5621 }, { "epoch": 0.67, "learning_rate": 2.6733106650828465e-06, "loss": 3.3156, "step": 5622 }, { "epoch": 0.67, "learning_rate": 2.671535390886427e-06, "loss": 3.3769, "step": 5623 }, { "epoch": 0.68, "learning_rate": 2.669760491434222e-06, "loss": 3.3704, "step": 5624 }, { "epoch": 0.68, "learning_rate": 2.6679859670118785e-06, "loss": 3.4544, "step": 5625 }, { "epoch": 0.68, "learning_rate": 2.6662118179049924e-06, "loss": 3.3601, "step": 5626 }, { "epoch": 0.68, "learning_rate": 2.6644380443990963e-06, "loss": 3.4298, "step": 5627 }, { "epoch": 0.68, "learning_rate": 2.662664646779662e-06, "loss": 3.4323, "step": 5628 }, { "epoch": 0.68, "learning_rate": 2.6608916253321002e-06, "loss": 3.362, "step": 5629 }, { "epoch": 0.68, "learning_rate": 2.659118980341763e-06, "loss": 3.4435, "step": 5630 }, { "epoch": 0.68, "learning_rate": 2.65734671209394e-06, "loss": 3.5379, "step": 5631 }, { "epoch": 0.68, "learning_rate": 2.6555748208738608e-06, "loss": 3.3869, "step": 5632 }, { "epoch": 0.68, "learning_rate": 2.653803306966697e-06, "loss": 3.3957, "step": 5633 }, { "epoch": 0.68, "learning_rate": 2.6520321706575523e-06, "loss": 3.4721, "step": 5634 }, { "epoch": 0.68, "learning_rate": 2.6502614122314772e-06, "loss": 3.4625, "step": 5635 }, { "epoch": 0.68, "learning_rate": 2.6484910319734567e-06, "loss": 3.3692, "step": 5636 }, { "epoch": 0.68, "learning_rate": 2.646721030168419e-06, "loss": 3.3492, "step": 5637 }, { "epoch": 0.68, "learning_rate": 2.6449514071012265e-06, "loss": 3.3887, "step": 5638 }, { "epoch": 0.68, "learning_rate": 2.643182163056685e-06, "loss": 3.336, "step": 5639 }, { "epoch": 0.68, "learning_rate": 2.6414132983195372e-06, "loss": 3.4728, "step": 5640 }, { "epoch": 0.68, "learning_rate": 2.6396448131744646e-06, "loss": 3.4355, "step": 5641 }, { "epoch": 0.68, "learning_rate": 2.6378767079060894e-06, "loss": 3.4046, "step": 5642 }, { "epoch": 0.68, "learning_rate": 2.6361089827989683e-06, "loss": 3.3411, "step": 5643 }, { "epoch": 0.68, "learning_rate": 2.6343416381376017e-06, "loss": 3.4143, "step": 5644 }, { "epoch": 0.68, "learning_rate": 2.6325746742064273e-06, "loss": 3.3151, "step": 5645 }, { "epoch": 0.68, "learning_rate": 2.6308080912898205e-06, "loss": 3.4351, "step": 5646 }, { "epoch": 0.68, "learning_rate": 2.629041889672096e-06, "loss": 3.3847, "step": 5647 }, { "epoch": 0.68, "learning_rate": 2.6272760696375076e-06, "loss": 3.4468, "step": 5648 }, { "epoch": 0.68, "learning_rate": 2.6255106314702467e-06, "loss": 3.4242, "step": 5649 }, { "epoch": 0.68, "learning_rate": 2.623745575454444e-06, "loss": 3.3899, "step": 5650 }, { "epoch": 0.68, "learning_rate": 2.621980901874168e-06, "loss": 3.4184, "step": 5651 }, { "epoch": 0.68, "learning_rate": 2.6202166110134273e-06, "loss": 3.3454, "step": 5652 }, { "epoch": 0.68, "learning_rate": 2.6184527031561663e-06, "loss": 3.2948, "step": 5653 }, { "epoch": 0.68, "learning_rate": 2.6166891785862703e-06, "loss": 3.2241, "step": 5654 }, { "epoch": 0.68, "learning_rate": 2.6149260375875606e-06, "loss": 3.3783, "step": 5655 }, { "epoch": 0.68, "learning_rate": 2.6131632804437995e-06, "loss": 3.3789, "step": 5656 }, { "epoch": 0.68, "learning_rate": 2.611400907438685e-06, "loss": 3.3303, "step": 5657 }, { "epoch": 0.68, "learning_rate": 2.6096389188558543e-06, "loss": 3.4514, "step": 5658 }, { "epoch": 0.68, "learning_rate": 2.6078773149788827e-06, "loss": 3.4133, "step": 5659 }, { "epoch": 0.68, "learning_rate": 2.606116096091286e-06, "loss": 3.4799, "step": 5660 }, { "epoch": 0.68, "learning_rate": 2.6043552624765112e-06, "loss": 3.4557, "step": 5661 }, { "epoch": 0.68, "learning_rate": 2.6025948144179492e-06, "loss": 3.4317, "step": 5662 }, { "epoch": 0.68, "learning_rate": 2.6008347521989287e-06, "loss": 3.4044, "step": 5663 }, { "epoch": 0.68, "learning_rate": 2.599075076102714e-06, "loss": 3.2898, "step": 5664 }, { "epoch": 0.68, "learning_rate": 2.597315786412508e-06, "loss": 3.4648, "step": 5665 }, { "epoch": 0.68, "learning_rate": 2.5955568834114523e-06, "loss": 3.4191, "step": 5666 }, { "epoch": 0.68, "learning_rate": 2.593798367382626e-06, "loss": 3.3394, "step": 5667 }, { "epoch": 0.68, "learning_rate": 2.5920402386090438e-06, "loss": 3.3797, "step": 5668 }, { "epoch": 0.68, "learning_rate": 2.590282497373664e-06, "loss": 3.3899, "step": 5669 }, { "epoch": 0.68, "learning_rate": 2.588525143959372e-06, "loss": 3.3983, "step": 5670 }, { "epoch": 0.68, "learning_rate": 2.5867681786490012e-06, "loss": 3.4512, "step": 5671 }, { "epoch": 0.68, "learning_rate": 2.5850116017253167e-06, "loss": 3.4866, "step": 5672 }, { "epoch": 0.68, "learning_rate": 2.5832554134710243e-06, "loss": 3.386, "step": 5673 }, { "epoch": 0.68, "learning_rate": 2.5814996141687652e-06, "loss": 3.4129, "step": 5674 }, { "epoch": 0.68, "learning_rate": 2.579744204101118e-06, "loss": 3.4641, "step": 5675 }, { "epoch": 0.68, "learning_rate": 2.5779891835505998e-06, "loss": 3.4076, "step": 5676 }, { "epoch": 0.68, "learning_rate": 2.5762345527996647e-06, "loss": 3.4091, "step": 5677 }, { "epoch": 0.68, "learning_rate": 2.5744803121307055e-06, "loss": 3.463, "step": 5678 }, { "epoch": 0.68, "learning_rate": 2.572726461826046e-06, "loss": 3.4332, "step": 5679 }, { "epoch": 0.68, "learning_rate": 2.5709730021679553e-06, "loss": 3.4435, "step": 5680 }, { "epoch": 0.68, "learning_rate": 2.569219933438634e-06, "loss": 3.373, "step": 5681 }, { "epoch": 0.68, "learning_rate": 2.5674672559202237e-06, "loss": 3.3923, "step": 5682 }, { "epoch": 0.68, "learning_rate": 2.565714969894799e-06, "loss": 3.4137, "step": 5683 }, { "epoch": 0.68, "learning_rate": 2.563963075644376e-06, "loss": 3.3595, "step": 5684 }, { "epoch": 0.68, "learning_rate": 2.5622115734509025e-06, "loss": 3.423, "step": 5685 }, { "epoch": 0.68, "learning_rate": 2.5604604635962694e-06, "loss": 3.4591, "step": 5686 }, { "epoch": 0.68, "learning_rate": 2.5587097463623016e-06, "loss": 3.416, "step": 5687 }, { "epoch": 0.68, "learning_rate": 2.556959422030757e-06, "loss": 3.4904, "step": 5688 }, { "epoch": 0.68, "learning_rate": 2.5552094908833347e-06, "loss": 3.3309, "step": 5689 }, { "epoch": 0.68, "learning_rate": 2.5534599532016697e-06, "loss": 3.3532, "step": 5690 }, { "epoch": 0.68, "learning_rate": 2.5517108092673335e-06, "loss": 3.3301, "step": 5691 }, { "epoch": 0.68, "learning_rate": 2.5499620593618345e-06, "loss": 3.4468, "step": 5692 }, { "epoch": 0.68, "learning_rate": 2.5482137037666176e-06, "loss": 3.2896, "step": 5693 }, { "epoch": 0.68, "learning_rate": 2.546465742763063e-06, "loss": 3.3416, "step": 5694 }, { "epoch": 0.68, "learning_rate": 2.544718176632489e-06, "loss": 3.3528, "step": 5695 }, { "epoch": 0.68, "learning_rate": 2.5429710056561517e-06, "loss": 3.3903, "step": 5696 }, { "epoch": 0.68, "learning_rate": 2.5412242301152366e-06, "loss": 3.391, "step": 5697 }, { "epoch": 0.68, "learning_rate": 2.5394778502908744e-06, "loss": 3.317, "step": 5698 }, { "epoch": 0.68, "learning_rate": 2.5377318664641284e-06, "loss": 3.4196, "step": 5699 }, { "epoch": 0.68, "learning_rate": 2.5359862789159958e-06, "loss": 3.3749, "step": 5700 }, { "epoch": 0.68, "learning_rate": 2.5342410879274137e-06, "loss": 3.4231, "step": 5701 }, { "epoch": 0.68, "learning_rate": 2.532496293779254e-06, "loss": 3.4475, "step": 5702 }, { "epoch": 0.68, "learning_rate": 2.530751896752325e-06, "loss": 3.4108, "step": 5703 }, { "epoch": 0.68, "learning_rate": 2.52900789712737e-06, "loss": 3.3419, "step": 5704 }, { "epoch": 0.68, "learning_rate": 2.5272642951850713e-06, "loss": 3.446, "step": 5705 }, { "epoch": 0.68, "learning_rate": 2.5255210912060418e-06, "loss": 3.3986, "step": 5706 }, { "epoch": 0.69, "learning_rate": 2.523778285470835e-06, "loss": 3.3424, "step": 5707 }, { "epoch": 0.69, "learning_rate": 2.5220358782599384e-06, "loss": 3.4271, "step": 5708 }, { "epoch": 0.69, "learning_rate": 2.520293869853777e-06, "loss": 3.3738, "step": 5709 }, { "epoch": 0.69, "learning_rate": 2.5185522605327096e-06, "loss": 3.4123, "step": 5710 }, { "epoch": 0.69, "learning_rate": 2.516811050577032e-06, "loss": 3.3657, "step": 5711 }, { "epoch": 0.69, "learning_rate": 2.515070240266976e-06, "loss": 3.3769, "step": 5712 }, { "epoch": 0.69, "learning_rate": 2.5133298298827064e-06, "loss": 3.3419, "step": 5713 }, { "epoch": 0.69, "learning_rate": 2.5115898197043276e-06, "loss": 3.3858, "step": 5714 }, { "epoch": 0.69, "learning_rate": 2.509850210011877e-06, "loss": 3.3386, "step": 5715 }, { "epoch": 0.69, "learning_rate": 2.508111001085328e-06, "loss": 3.418, "step": 5716 }, { "epoch": 0.69, "learning_rate": 2.5063721932045903e-06, "loss": 3.419, "step": 5717 }, { "epoch": 0.69, "learning_rate": 2.5046337866495083e-06, "loss": 3.3435, "step": 5718 }, { "epoch": 0.69, "learning_rate": 2.5028957816998623e-06, "loss": 3.4064, "step": 5719 }, { "epoch": 0.69, "learning_rate": 2.5011581786353666e-06, "loss": 3.4056, "step": 5720 }, { "epoch": 0.69, "learning_rate": 2.499420977735672e-06, "loss": 3.3394, "step": 5721 }, { "epoch": 0.69, "learning_rate": 2.497684179280366e-06, "loss": 3.3899, "step": 5722 }, { "epoch": 0.69, "learning_rate": 2.49594778354897e-06, "loss": 3.428, "step": 5723 }, { "epoch": 0.69, "learning_rate": 2.4942117908209375e-06, "loss": 3.424, "step": 5724 }, { "epoch": 0.69, "learning_rate": 2.492476201375661e-06, "loss": 3.359, "step": 5725 }, { "epoch": 0.69, "learning_rate": 2.4907410154924683e-06, "loss": 3.4075, "step": 5726 }, { "epoch": 0.69, "learning_rate": 2.4890062334506197e-06, "loss": 3.4619, "step": 5727 }, { "epoch": 0.69, "learning_rate": 2.4872718555293134e-06, "loss": 3.3642, "step": 5728 }, { "epoch": 0.69, "learning_rate": 2.4855378820076798e-06, "loss": 3.4418, "step": 5729 }, { "epoch": 0.69, "learning_rate": 2.4838043131647853e-06, "loss": 3.4014, "step": 5730 }, { "epoch": 0.69, "learning_rate": 2.482071149279633e-06, "loss": 3.3652, "step": 5731 }, { "epoch": 0.69, "learning_rate": 2.480338390631159e-06, "loss": 3.335, "step": 5732 }, { "epoch": 0.69, "learning_rate": 2.478606037498231e-06, "loss": 3.3399, "step": 5733 }, { "epoch": 0.69, "learning_rate": 2.4768740901596583e-06, "loss": 3.3743, "step": 5734 }, { "epoch": 0.69, "learning_rate": 2.475142548894179e-06, "loss": 3.4542, "step": 5735 }, { "epoch": 0.69, "learning_rate": 2.4734114139804695e-06, "loss": 3.4337, "step": 5736 }, { "epoch": 0.69, "learning_rate": 2.471680685697139e-06, "loss": 3.3712, "step": 5737 }, { "epoch": 0.69, "learning_rate": 2.4699503643227323e-06, "loss": 3.4589, "step": 5738 }, { "epoch": 0.69, "learning_rate": 2.4682204501357275e-06, "loss": 3.4097, "step": 5739 }, { "epoch": 0.69, "learning_rate": 2.4664909434145382e-06, "loss": 3.3204, "step": 5740 }, { "epoch": 0.69, "learning_rate": 2.4647618444375143e-06, "loss": 3.3437, "step": 5741 }, { "epoch": 0.69, "learning_rate": 2.463033153482933e-06, "loss": 3.3432, "step": 5742 }, { "epoch": 0.69, "learning_rate": 2.4613048708290137e-06, "loss": 3.3946, "step": 5743 }, { "epoch": 0.69, "learning_rate": 2.4595769967539065e-06, "loss": 3.3775, "step": 5744 }, { "epoch": 0.69, "learning_rate": 2.457849531535695e-06, "loss": 3.2575, "step": 5745 }, { "epoch": 0.69, "learning_rate": 2.4561224754524003e-06, "loss": 3.3218, "step": 5746 }, { "epoch": 0.69, "learning_rate": 2.454395828781975e-06, "loss": 3.4773, "step": 5747 }, { "epoch": 0.69, "learning_rate": 2.452669591802307e-06, "loss": 3.3299, "step": 5748 }, { "epoch": 0.69, "learning_rate": 2.450943764791216e-06, "loss": 3.3155, "step": 5749 }, { "epoch": 0.69, "learning_rate": 2.4492183480264586e-06, "loss": 3.3529, "step": 5750 }, { "epoch": 0.69, "learning_rate": 2.4474933417857244e-06, "loss": 3.4305, "step": 5751 }, { "epoch": 0.69, "learning_rate": 2.445768746346636e-06, "loss": 3.4127, "step": 5752 }, { "epoch": 0.69, "learning_rate": 2.444044561986751e-06, "loss": 3.4721, "step": 5753 }, { "epoch": 0.69, "learning_rate": 2.4423207889835606e-06, "loss": 3.3861, "step": 5754 }, { "epoch": 0.69, "learning_rate": 2.4405974276144894e-06, "loss": 3.4363, "step": 5755 }, { "epoch": 0.69, "learning_rate": 2.4388744781568964e-06, "loss": 3.4454, "step": 5756 }, { "epoch": 0.69, "learning_rate": 2.4371519408880733e-06, "loss": 3.2814, "step": 5757 }, { "epoch": 0.69, "learning_rate": 2.4354298160852464e-06, "loss": 3.4168, "step": 5758 }, { "epoch": 0.69, "learning_rate": 2.4337081040255773e-06, "loss": 3.3207, "step": 5759 }, { "epoch": 0.69, "learning_rate": 2.4319868049861557e-06, "loss": 3.4287, "step": 5760 }, { "epoch": 0.69, "learning_rate": 2.4302659192440094e-06, "loss": 3.4307, "step": 5761 }, { "epoch": 0.69, "learning_rate": 2.428545447076098e-06, "loss": 3.3205, "step": 5762 }, { "epoch": 0.69, "learning_rate": 2.4268253887593176e-06, "loss": 3.3867, "step": 5763 }, { "epoch": 0.69, "learning_rate": 2.4251057445704927e-06, "loss": 3.5018, "step": 5764 }, { "epoch": 0.69, "learning_rate": 2.423386514786385e-06, "loss": 3.385, "step": 5765 }, { "epoch": 0.69, "learning_rate": 2.4216676996836876e-06, "loss": 3.3935, "step": 5766 }, { "epoch": 0.69, "learning_rate": 2.4199492995390273e-06, "loss": 3.4383, "step": 5767 }, { "epoch": 0.69, "learning_rate": 2.418231314628967e-06, "loss": 3.39, "step": 5768 }, { "epoch": 0.69, "learning_rate": 2.416513745229995e-06, "loss": 3.3949, "step": 5769 }, { "epoch": 0.69, "learning_rate": 2.4147965916185413e-06, "loss": 3.3246, "step": 5770 }, { "epoch": 0.69, "learning_rate": 2.413079854070964e-06, "loss": 3.2518, "step": 5771 }, { "epoch": 0.69, "learning_rate": 2.4113635328635553e-06, "loss": 3.466, "step": 5772 }, { "epoch": 0.69, "learning_rate": 2.4096476282725428e-06, "loss": 3.3548, "step": 5773 }, { "epoch": 0.69, "learning_rate": 2.407932140574083e-06, "loss": 3.406, "step": 5774 }, { "epoch": 0.69, "learning_rate": 2.4062170700442683e-06, "loss": 3.3223, "step": 5775 }, { "epoch": 0.69, "learning_rate": 2.404502416959123e-06, "loss": 3.42, "step": 5776 }, { "epoch": 0.69, "learning_rate": 2.402788181594605e-06, "loss": 3.3642, "step": 5777 }, { "epoch": 0.69, "learning_rate": 2.401074364226602e-06, "loss": 3.4209, "step": 5778 }, { "epoch": 0.69, "learning_rate": 2.3993609651309373e-06, "loss": 3.4659, "step": 5779 }, { "epoch": 0.69, "learning_rate": 2.3976479845833674e-06, "loss": 3.4386, "step": 5780 }, { "epoch": 0.69, "learning_rate": 2.395935422859578e-06, "loss": 3.3539, "step": 5781 }, { "epoch": 0.69, "learning_rate": 2.39422328023519e-06, "loss": 3.4313, "step": 5782 }, { "epoch": 0.69, "learning_rate": 2.392511556985759e-06, "loss": 3.3951, "step": 5783 }, { "epoch": 0.69, "learning_rate": 2.3908002533867697e-06, "loss": 3.4168, "step": 5784 }, { "epoch": 0.69, "learning_rate": 2.3890893697136384e-06, "loss": 3.3127, "step": 5785 }, { "epoch": 0.69, "learning_rate": 2.387378906241719e-06, "loss": 3.2705, "step": 5786 }, { "epoch": 0.69, "learning_rate": 2.3856688632462906e-06, "loss": 3.3868, "step": 5787 }, { "epoch": 0.69, "learning_rate": 2.3839592410025684e-06, "loss": 3.4232, "step": 5788 }, { "epoch": 0.69, "learning_rate": 2.3822500397857016e-06, "loss": 3.3679, "step": 5789 }, { "epoch": 0.69, "learning_rate": 2.3805412598707695e-06, "loss": 3.4207, "step": 5790 }, { "epoch": 0.7, "learning_rate": 2.378832901532783e-06, "loss": 3.3615, "step": 5791 }, { "epoch": 0.7, "learning_rate": 2.3771249650466875e-06, "loss": 3.3858, "step": 5792 }, { "epoch": 0.7, "learning_rate": 2.375417450687359e-06, "loss": 3.4506, "step": 5793 }, { "epoch": 0.7, "learning_rate": 2.3737103587296044e-06, "loss": 3.3661, "step": 5794 }, { "epoch": 0.7, "learning_rate": 2.372003689448167e-06, "loss": 3.3936, "step": 5795 }, { "epoch": 0.7, "learning_rate": 2.370297443117714e-06, "loss": 3.2234, "step": 5796 }, { "epoch": 0.7, "learning_rate": 2.3685916200128526e-06, "loss": 3.3408, "step": 5797 }, { "epoch": 0.7, "learning_rate": 2.3668862204081184e-06, "loss": 3.5149, "step": 5798 }, { "epoch": 0.7, "learning_rate": 2.3651812445779793e-06, "loss": 3.3771, "step": 5799 }, { "epoch": 0.7, "learning_rate": 2.3634766927968344e-06, "loss": 3.3978, "step": 5800 }, { "epoch": 0.7, "learning_rate": 2.3617725653390155e-06, "loss": 3.4353, "step": 5801 }, { "epoch": 0.7, "learning_rate": 2.3600688624787858e-06, "loss": 3.3695, "step": 5802 }, { "epoch": 0.7, "learning_rate": 2.3583655844903404e-06, "loss": 3.3068, "step": 5803 }, { "epoch": 0.7, "learning_rate": 2.3566627316478064e-06, "loss": 3.4929, "step": 5804 }, { "epoch": 0.7, "learning_rate": 2.354960304225239e-06, "loss": 3.4159, "step": 5805 }, { "epoch": 0.7, "learning_rate": 2.3532583024966295e-06, "loss": 3.2987, "step": 5806 }, { "epoch": 0.7, "learning_rate": 2.351556726735898e-06, "loss": 3.4292, "step": 5807 }, { "epoch": 0.7, "learning_rate": 2.349855577216898e-06, "loss": 3.4416, "step": 5808 }, { "epoch": 0.7, "learning_rate": 2.3481548542134126e-06, "loss": 3.4141, "step": 5809 }, { "epoch": 0.7, "learning_rate": 2.3464545579991572e-06, "loss": 3.3766, "step": 5810 }, { "epoch": 0.7, "learning_rate": 2.3447546888477783e-06, "loss": 3.3541, "step": 5811 }, { "epoch": 0.7, "learning_rate": 2.343055247032854e-06, "loss": 3.336, "step": 5812 }, { "epoch": 0.7, "learning_rate": 2.3413562328278923e-06, "loss": 3.3683, "step": 5813 }, { "epoch": 0.7, "learning_rate": 2.3396576465063344e-06, "loss": 3.3331, "step": 5814 }, { "epoch": 0.7, "learning_rate": 2.337959488341551e-06, "loss": 3.4243, "step": 5815 }, { "epoch": 0.7, "learning_rate": 2.3362617586068453e-06, "loss": 3.4071, "step": 5816 }, { "epoch": 0.7, "learning_rate": 2.3345644575754494e-06, "loss": 3.3922, "step": 5817 }, { "epoch": 0.7, "learning_rate": 2.332867585520529e-06, "loss": 3.5095, "step": 5818 }, { "epoch": 0.7, "learning_rate": 2.331171142715179e-06, "loss": 3.382, "step": 5819 }, { "epoch": 0.7, "learning_rate": 2.3294751294324253e-06, "loss": 3.435, "step": 5820 }, { "epoch": 0.7, "learning_rate": 2.327779545945226e-06, "loss": 3.3069, "step": 5821 }, { "epoch": 0.7, "learning_rate": 2.32608439252647e-06, "loss": 3.346, "step": 5822 }, { "epoch": 0.7, "learning_rate": 2.324389669448973e-06, "loss": 3.4214, "step": 5823 }, { "epoch": 0.7, "learning_rate": 2.3226953769854864e-06, "loss": 3.4174, "step": 5824 }, { "epoch": 0.7, "learning_rate": 2.32100151540869e-06, "loss": 3.4558, "step": 5825 }, { "epoch": 0.7, "learning_rate": 2.3193080849911955e-06, "loss": 3.3421, "step": 5826 }, { "epoch": 0.7, "learning_rate": 2.3176150860055433e-06, "loss": 3.3349, "step": 5827 }, { "epoch": 0.7, "learning_rate": 2.3159225187242067e-06, "loss": 3.3695, "step": 5828 }, { "epoch": 0.7, "learning_rate": 2.314230383419587e-06, "loss": 3.3781, "step": 5829 }, { "epoch": 0.7, "learning_rate": 2.3125386803640188e-06, "loss": 3.4152, "step": 5830 }, { "epoch": 0.7, "learning_rate": 2.310847409829766e-06, "loss": 3.4187, "step": 5831 }, { "epoch": 0.7, "learning_rate": 2.309156572089019e-06, "loss": 3.3637, "step": 5832 }, { "epoch": 0.7, "learning_rate": 2.307466167413905e-06, "loss": 3.3461, "step": 5833 }, { "epoch": 0.7, "learning_rate": 2.3057761960764774e-06, "loss": 3.4346, "step": 5834 }, { "epoch": 0.7, "learning_rate": 2.3040866583487215e-06, "loss": 3.3899, "step": 5835 }, { "epoch": 0.7, "learning_rate": 2.3023975545025523e-06, "loss": 3.4461, "step": 5836 }, { "epoch": 0.7, "learning_rate": 2.3007088848098146e-06, "loss": 3.5067, "step": 5837 }, { "epoch": 0.7, "learning_rate": 2.2990206495422847e-06, "loss": 3.4017, "step": 5838 }, { "epoch": 0.7, "learning_rate": 2.297332848971667e-06, "loss": 3.4037, "step": 5839 }, { "epoch": 0.7, "learning_rate": 2.2956454833695986e-06, "loss": 3.3594, "step": 5840 }, { "epoch": 0.7, "learning_rate": 2.293958553007642e-06, "loss": 3.4172, "step": 5841 }, { "epoch": 0.7, "learning_rate": 2.2922720581572937e-06, "loss": 3.3884, "step": 5842 }, { "epoch": 0.7, "learning_rate": 2.29058599908998e-06, "loss": 3.46, "step": 5843 }, { "epoch": 0.7, "learning_rate": 2.2889003760770558e-06, "loss": 3.4748, "step": 5844 }, { "epoch": 0.7, "learning_rate": 2.2872151893898052e-06, "loss": 3.3096, "step": 5845 }, { "epoch": 0.7, "learning_rate": 2.285530439299444e-06, "loss": 3.431, "step": 5846 }, { "epoch": 0.7, "learning_rate": 2.2838461260771154e-06, "loss": 3.4071, "step": 5847 }, { "epoch": 0.7, "learning_rate": 2.282162249993895e-06, "loss": 3.3677, "step": 5848 }, { "epoch": 0.7, "learning_rate": 2.280478811320786e-06, "loss": 3.4669, "step": 5849 }, { "epoch": 0.7, "learning_rate": 2.2787958103287216e-06, "loss": 3.4622, "step": 5850 }, { "epoch": 0.7, "learning_rate": 2.2771132472885647e-06, "loss": 3.3115, "step": 5851 }, { "epoch": 0.7, "learning_rate": 2.275431122471109e-06, "loss": 3.322, "step": 5852 }, { "epoch": 0.7, "learning_rate": 2.273749436147075e-06, "loss": 3.3899, "step": 5853 }, { "epoch": 0.7, "learning_rate": 2.272068188587115e-06, "loss": 3.3854, "step": 5854 }, { "epoch": 0.7, "learning_rate": 2.270387380061809e-06, "loss": 3.2925, "step": 5855 }, { "epoch": 0.7, "learning_rate": 2.268707010841668e-06, "loss": 3.4774, "step": 5856 }, { "epoch": 0.7, "learning_rate": 2.267027081197131e-06, "loss": 3.3466, "step": 5857 }, { "epoch": 0.7, "learning_rate": 2.265347591398569e-06, "loss": 3.339, "step": 5858 }, { "epoch": 0.7, "learning_rate": 2.2636685417162754e-06, "loss": 3.384, "step": 5859 }, { "epoch": 0.7, "learning_rate": 2.261989932420479e-06, "loss": 3.3147, "step": 5860 }, { "epoch": 0.7, "learning_rate": 2.2603117637813373e-06, "loss": 3.4003, "step": 5861 }, { "epoch": 0.7, "learning_rate": 2.2586340360689346e-06, "loss": 3.5565, "step": 5862 }, { "epoch": 0.7, "learning_rate": 2.256956749553285e-06, "loss": 3.3363, "step": 5863 }, { "epoch": 0.7, "learning_rate": 2.255279904504332e-06, "loss": 3.4636, "step": 5864 }, { "epoch": 0.7, "learning_rate": 2.2536035011919485e-06, "loss": 3.3927, "step": 5865 }, { "epoch": 0.7, "learning_rate": 2.2519275398859347e-06, "loss": 3.306, "step": 5866 }, { "epoch": 0.7, "learning_rate": 2.2502520208560237e-06, "loss": 3.2581, "step": 5867 }, { "epoch": 0.7, "learning_rate": 2.2485769443718692e-06, "loss": 3.4092, "step": 5868 }, { "epoch": 0.7, "learning_rate": 2.2469023107030612e-06, "loss": 3.4539, "step": 5869 }, { "epoch": 0.7, "learning_rate": 2.245228120119116e-06, "loss": 3.3438, "step": 5870 }, { "epoch": 0.7, "learning_rate": 2.243554372889479e-06, "loss": 3.4616, "step": 5871 }, { "epoch": 0.7, "learning_rate": 2.2418810692835243e-06, "loss": 3.4072, "step": 5872 }, { "epoch": 0.7, "learning_rate": 2.240208209570553e-06, "loss": 3.4689, "step": 5873 }, { "epoch": 0.71, "learning_rate": 2.2385357940197967e-06, "loss": 3.3867, "step": 5874 }, { "epoch": 0.71, "learning_rate": 2.236863822900414e-06, "loss": 3.4084, "step": 5875 }, { "epoch": 0.71, "learning_rate": 2.235192296481496e-06, "loss": 3.3435, "step": 5876 }, { "epoch": 0.71, "learning_rate": 2.233521215032054e-06, "loss": 3.3917, "step": 5877 }, { "epoch": 0.71, "learning_rate": 2.231850578821035e-06, "loss": 3.3685, "step": 5878 }, { "epoch": 0.71, "learning_rate": 2.230180388117312e-06, "loss": 3.4109, "step": 5879 }, { "epoch": 0.71, "learning_rate": 2.2285106431896842e-06, "loss": 3.4362, "step": 5880 }, { "epoch": 0.71, "learning_rate": 2.2268413443068848e-06, "loss": 3.3041, "step": 5881 }, { "epoch": 0.71, "learning_rate": 2.22517249173757e-06, "loss": 3.4426, "step": 5882 }, { "epoch": 0.71, "learning_rate": 2.2235040857503254e-06, "loss": 3.3934, "step": 5883 }, { "epoch": 0.71, "learning_rate": 2.2218361266136657e-06, "loss": 3.4006, "step": 5884 }, { "epoch": 0.71, "learning_rate": 2.2201686145960344e-06, "loss": 3.3847, "step": 5885 }, { "epoch": 0.71, "learning_rate": 2.2185015499657984e-06, "loss": 3.3785, "step": 5886 }, { "epoch": 0.71, "learning_rate": 2.216834932991257e-06, "loss": 3.3732, "step": 5887 }, { "epoch": 0.71, "learning_rate": 2.215168763940637e-06, "loss": 3.4842, "step": 5888 }, { "epoch": 0.71, "learning_rate": 2.213503043082092e-06, "loss": 3.4433, "step": 5889 }, { "epoch": 0.71, "learning_rate": 2.211837770683704e-06, "loss": 3.3725, "step": 5890 }, { "epoch": 0.71, "learning_rate": 2.210172947013483e-06, "loss": 3.4091, "step": 5891 }, { "epoch": 0.71, "learning_rate": 2.2085085723393666e-06, "loss": 3.4645, "step": 5892 }, { "epoch": 0.71, "learning_rate": 2.20684464692922e-06, "loss": 3.4264, "step": 5893 }, { "epoch": 0.71, "learning_rate": 2.205181171050837e-06, "loss": 3.4605, "step": 5894 }, { "epoch": 0.71, "learning_rate": 2.2035181449719363e-06, "loss": 3.3925, "step": 5895 }, { "epoch": 0.71, "learning_rate": 2.2018555689601663e-06, "loss": 3.4522, "step": 5896 }, { "epoch": 0.71, "learning_rate": 2.2001934432831036e-06, "loss": 3.5023, "step": 5897 }, { "epoch": 0.71, "learning_rate": 2.1985317682082513e-06, "loss": 3.458, "step": 5898 }, { "epoch": 0.71, "learning_rate": 2.196870544003041e-06, "loss": 3.4038, "step": 5899 }, { "epoch": 0.71, "learning_rate": 2.195209770934829e-06, "loss": 3.3554, "step": 5900 }, { "epoch": 0.71, "learning_rate": 2.193549449270903e-06, "loss": 3.365, "step": 5901 }, { "epoch": 0.71, "learning_rate": 2.191889579278475e-06, "loss": 3.4537, "step": 5902 }, { "epoch": 0.71, "learning_rate": 2.1902301612246868e-06, "loss": 3.2987, "step": 5903 }, { "epoch": 0.71, "learning_rate": 2.1885711953766026e-06, "loss": 3.4451, "step": 5904 }, { "epoch": 0.71, "learning_rate": 2.186912682001219e-06, "loss": 3.4707, "step": 5905 }, { "epoch": 0.71, "learning_rate": 2.1852546213654584e-06, "loss": 3.3223, "step": 5906 }, { "epoch": 0.71, "learning_rate": 2.1835970137361697e-06, "loss": 3.3928, "step": 5907 }, { "epoch": 0.71, "learning_rate": 2.181939859380128e-06, "loss": 3.4603, "step": 5908 }, { "epoch": 0.71, "learning_rate": 2.180283158564038e-06, "loss": 3.4811, "step": 5909 }, { "epoch": 0.71, "learning_rate": 2.1786269115545293e-06, "loss": 3.4069, "step": 5910 }, { "epoch": 0.71, "learning_rate": 2.1769711186181584e-06, "loss": 3.4347, "step": 5911 }, { "epoch": 0.71, "learning_rate": 2.1753157800214107e-06, "loss": 3.4759, "step": 5912 }, { "epoch": 0.71, "learning_rate": 2.1736608960306965e-06, "loss": 3.3114, "step": 5913 }, { "epoch": 0.71, "learning_rate": 2.1720064669123532e-06, "loss": 3.3241, "step": 5914 }, { "epoch": 0.71, "learning_rate": 2.1703524929326464e-06, "loss": 3.4659, "step": 5915 }, { "epoch": 0.71, "learning_rate": 2.1686989743577664e-06, "loss": 3.3499, "step": 5916 }, { "epoch": 0.71, "learning_rate": 2.167045911453832e-06, "loss": 3.4015, "step": 5917 }, { "epoch": 0.71, "learning_rate": 2.165393304486888e-06, "loss": 3.3371, "step": 5918 }, { "epoch": 0.71, "learning_rate": 2.163741153722906e-06, "loss": 3.3696, "step": 5919 }, { "epoch": 0.71, "learning_rate": 2.1620894594277826e-06, "loss": 3.3922, "step": 5920 }, { "epoch": 0.71, "learning_rate": 2.1604382218673454e-06, "loss": 3.4603, "step": 5921 }, { "epoch": 0.71, "learning_rate": 2.1587874413073405e-06, "loss": 3.3465, "step": 5922 }, { "epoch": 0.71, "learning_rate": 2.1571371180134486e-06, "loss": 3.4524, "step": 5923 }, { "epoch": 0.71, "learning_rate": 2.155487252251273e-06, "loss": 3.3934, "step": 5924 }, { "epoch": 0.71, "learning_rate": 2.1538378442863434e-06, "loss": 3.3921, "step": 5925 }, { "epoch": 0.71, "learning_rate": 2.152188894384117e-06, "loss": 3.3384, "step": 5926 }, { "epoch": 0.71, "learning_rate": 2.150540402809976e-06, "loss": 3.4449, "step": 5927 }, { "epoch": 0.71, "learning_rate": 2.148892369829229e-06, "loss": 3.4019, "step": 5928 }, { "epoch": 0.71, "learning_rate": 2.1472447957071123e-06, "loss": 3.346, "step": 5929 }, { "epoch": 0.71, "learning_rate": 2.145597680708789e-06, "loss": 3.2964, "step": 5930 }, { "epoch": 0.71, "learning_rate": 2.1439510250993424e-06, "loss": 3.4192, "step": 5931 }, { "epoch": 0.71, "learning_rate": 2.142304829143788e-06, "loss": 3.3259, "step": 5932 }, { "epoch": 0.71, "learning_rate": 2.1406590931070652e-06, "loss": 3.3749, "step": 5933 }, { "epoch": 0.71, "learning_rate": 2.1390138172540402e-06, "loss": 3.3967, "step": 5934 }, { "epoch": 0.71, "learning_rate": 2.1373690018495043e-06, "loss": 3.3618, "step": 5935 }, { "epoch": 0.71, "learning_rate": 2.135724647158174e-06, "loss": 3.3667, "step": 5936 }, { "epoch": 0.71, "learning_rate": 2.134080753444693e-06, "loss": 3.4211, "step": 5937 }, { "epoch": 0.71, "learning_rate": 2.132437320973631e-06, "loss": 3.3894, "step": 5938 }, { "epoch": 0.71, "learning_rate": 2.1307943500094843e-06, "loss": 3.3856, "step": 5939 }, { "epoch": 0.71, "learning_rate": 2.129151840816669e-06, "loss": 3.3085, "step": 5940 }, { "epoch": 0.71, "learning_rate": 2.127509793659534e-06, "loss": 3.3793, "step": 5941 }, { "epoch": 0.71, "learning_rate": 2.1258682088023512e-06, "loss": 3.3272, "step": 5942 }, { "epoch": 0.71, "learning_rate": 2.1242270865093177e-06, "loss": 3.4175, "step": 5943 }, { "epoch": 0.71, "learning_rate": 2.1225864270445563e-06, "loss": 3.3515, "step": 5944 }, { "epoch": 0.71, "learning_rate": 2.120946230672116e-06, "loss": 3.4002, "step": 5945 }, { "epoch": 0.71, "learning_rate": 2.1193064976559706e-06, "loss": 3.3382, "step": 5946 }, { "epoch": 0.71, "learning_rate": 2.11766722826002e-06, "loss": 3.2482, "step": 5947 }, { "epoch": 0.71, "learning_rate": 2.1160284227480878e-06, "loss": 3.4625, "step": 5948 }, { "epoch": 0.71, "learning_rate": 2.1143900813839253e-06, "loss": 3.3357, "step": 5949 }, { "epoch": 0.71, "learning_rate": 2.1127522044312073e-06, "loss": 3.3896, "step": 5950 }, { "epoch": 0.71, "learning_rate": 2.1111147921535348e-06, "loss": 3.3563, "step": 5951 }, { "epoch": 0.71, "learning_rate": 2.1094778448144336e-06, "loss": 3.3684, "step": 5952 }, { "epoch": 0.71, "learning_rate": 2.1078413626773547e-06, "loss": 3.4111, "step": 5953 }, { "epoch": 0.71, "learning_rate": 2.1062053460056747e-06, "loss": 3.4601, "step": 5954 }, { "epoch": 0.71, "learning_rate": 2.1045697950626948e-06, "loss": 3.3119, "step": 5955 }, { "epoch": 0.71, "learning_rate": 2.1029347101116408e-06, "loss": 3.3085, "step": 5956 }, { "epoch": 0.72, "learning_rate": 2.1013000914156663e-06, "loss": 3.4069, "step": 5957 }, { "epoch": 0.72, "learning_rate": 2.0996659392378445e-06, "loss": 3.3559, "step": 5958 }, { "epoch": 0.72, "learning_rate": 2.0980322538411775e-06, "loss": 3.328, "step": 5959 }, { "epoch": 0.72, "learning_rate": 2.0963990354885914e-06, "loss": 3.4059, "step": 5960 }, { "epoch": 0.72, "learning_rate": 2.0947662844429384e-06, "loss": 3.2883, "step": 5961 }, { "epoch": 0.72, "learning_rate": 2.0931340009669927e-06, "loss": 3.4861, "step": 5962 }, { "epoch": 0.72, "learning_rate": 2.0915021853234556e-06, "loss": 3.4453, "step": 5963 }, { "epoch": 0.72, "learning_rate": 2.0898708377749526e-06, "loss": 3.363, "step": 5964 }, { "epoch": 0.72, "learning_rate": 2.088239958584033e-06, "loss": 3.3413, "step": 5965 }, { "epoch": 0.72, "learning_rate": 2.086609548013173e-06, "loss": 3.4648, "step": 5966 }, { "epoch": 0.72, "learning_rate": 2.0849796063247692e-06, "loss": 3.5294, "step": 5967 }, { "epoch": 0.72, "learning_rate": 2.0833501337811453e-06, "loss": 3.3943, "step": 5968 }, { "epoch": 0.72, "learning_rate": 2.0817211306445503e-06, "loss": 3.2999, "step": 5969 }, { "epoch": 0.72, "learning_rate": 2.0800925971771564e-06, "loss": 3.5051, "step": 5970 }, { "epoch": 0.72, "learning_rate": 2.078464533641061e-06, "loss": 3.4986, "step": 5971 }, { "epoch": 0.72, "learning_rate": 2.076836940298285e-06, "loss": 3.4003, "step": 5972 }, { "epoch": 0.72, "learning_rate": 2.075209817410774e-06, "loss": 3.3549, "step": 5973 }, { "epoch": 0.72, "learning_rate": 2.0735831652403987e-06, "loss": 3.3447, "step": 5974 }, { "epoch": 0.72, "learning_rate": 2.0719569840489543e-06, "loss": 3.4529, "step": 5975 }, { "epoch": 0.72, "learning_rate": 2.0703312740981563e-06, "loss": 3.3407, "step": 5976 }, { "epoch": 0.72, "learning_rate": 2.068706035649647e-06, "loss": 3.3934, "step": 5977 }, { "epoch": 0.72, "learning_rate": 2.0670812689649953e-06, "loss": 3.4187, "step": 5978 }, { "epoch": 0.72, "learning_rate": 2.0654569743056924e-06, "loss": 3.3215, "step": 5979 }, { "epoch": 0.72, "learning_rate": 2.0638331519331518e-06, "loss": 3.3824, "step": 5980 }, { "epoch": 0.72, "learning_rate": 2.0622098021087117e-06, "loss": 3.4248, "step": 5981 }, { "epoch": 0.72, "learning_rate": 2.0605869250936362e-06, "loss": 3.34, "step": 5982 }, { "epoch": 0.72, "learning_rate": 2.058964521149111e-06, "loss": 3.3722, "step": 5983 }, { "epoch": 0.72, "learning_rate": 2.057342590536248e-06, "loss": 3.3887, "step": 5984 }, { "epoch": 0.72, "learning_rate": 2.0557211335160794e-06, "loss": 3.3957, "step": 5985 }, { "epoch": 0.72, "learning_rate": 2.054100150349564e-06, "loss": 3.337, "step": 5986 }, { "epoch": 0.72, "learning_rate": 2.0524796412975834e-06, "loss": 3.3744, "step": 5987 }, { "epoch": 0.72, "learning_rate": 2.050859606620943e-06, "loss": 3.4655, "step": 5988 }, { "epoch": 0.72, "learning_rate": 2.049240046580373e-06, "loss": 3.434, "step": 5989 }, { "epoch": 0.72, "learning_rate": 2.0476209614365254e-06, "loss": 3.5263, "step": 5990 }, { "epoch": 0.72, "learning_rate": 2.046002351449977e-06, "loss": 3.3159, "step": 5991 }, { "epoch": 0.72, "learning_rate": 2.0443842168812274e-06, "loss": 3.3929, "step": 5992 }, { "epoch": 0.72, "learning_rate": 2.0427665579907015e-06, "loss": 3.4562, "step": 5993 }, { "epoch": 0.72, "learning_rate": 2.0411493750387423e-06, "loss": 3.3624, "step": 5994 }, { "epoch": 0.72, "learning_rate": 2.039532668285623e-06, "loss": 3.3904, "step": 5995 }, { "epoch": 0.72, "learning_rate": 2.037916437991536e-06, "loss": 3.3755, "step": 5996 }, { "epoch": 0.72, "learning_rate": 2.036300684416599e-06, "loss": 3.3353, "step": 5997 }, { "epoch": 0.72, "learning_rate": 2.034685407820851e-06, "loss": 3.4802, "step": 5998 }, { "epoch": 0.72, "learning_rate": 2.0330706084642564e-06, "loss": 3.4158, "step": 5999 }, { "epoch": 0.72, "learning_rate": 2.0314562866067017e-06, "loss": 3.3664, "step": 6000 }, { "epoch": 0.72, "learning_rate": 2.0298424425079966e-06, "loss": 3.3821, "step": 6001 }, { "epoch": 0.72, "learning_rate": 2.0282290764278752e-06, "loss": 3.4931, "step": 6002 }, { "epoch": 0.72, "learning_rate": 2.02661618862599e-06, "loss": 3.3372, "step": 6003 }, { "epoch": 0.72, "learning_rate": 2.0250037793619227e-06, "loss": 3.3674, "step": 6004 }, { "epoch": 0.72, "learning_rate": 2.0233918488951737e-06, "loss": 3.3218, "step": 6005 }, { "epoch": 0.72, "learning_rate": 2.0217803974851692e-06, "loss": 3.3855, "step": 6006 }, { "epoch": 0.72, "learning_rate": 2.0201694253912556e-06, "loss": 3.3488, "step": 6007 }, { "epoch": 0.72, "learning_rate": 2.018558932872705e-06, "loss": 3.4534, "step": 6008 }, { "epoch": 0.72, "learning_rate": 2.0169489201887097e-06, "loss": 3.3613, "step": 6009 }, { "epoch": 0.72, "learning_rate": 2.0153393875983867e-06, "loss": 3.4316, "step": 6010 }, { "epoch": 0.72, "learning_rate": 2.0137303353607745e-06, "loss": 3.4001, "step": 6011 }, { "epoch": 0.72, "learning_rate": 2.0121217637348346e-06, "loss": 3.2719, "step": 6012 }, { "epoch": 0.72, "learning_rate": 2.010513672979452e-06, "loss": 3.3505, "step": 6013 }, { "epoch": 0.72, "learning_rate": 2.0089060633534326e-06, "loss": 3.4262, "step": 6014 }, { "epoch": 0.72, "learning_rate": 2.0072989351155064e-06, "loss": 3.3946, "step": 6015 }, { "epoch": 0.72, "learning_rate": 2.0056922885243258e-06, "loss": 3.2358, "step": 6016 }, { "epoch": 0.72, "learning_rate": 2.0040861238384645e-06, "loss": 3.3799, "step": 6017 }, { "epoch": 0.72, "learning_rate": 2.00248044131642e-06, "loss": 3.3659, "step": 6018 }, { "epoch": 0.72, "learning_rate": 2.000875241216612e-06, "loss": 3.3713, "step": 6019 }, { "epoch": 0.72, "learning_rate": 1.999270523797382e-06, "loss": 3.3982, "step": 6020 }, { "epoch": 0.72, "learning_rate": 1.9976662893169923e-06, "loss": 3.3185, "step": 6021 }, { "epoch": 0.72, "learning_rate": 1.9960625380336307e-06, "loss": 3.406, "step": 6022 }, { "epoch": 0.72, "learning_rate": 1.994459270205405e-06, "loss": 3.4255, "step": 6023 }, { "epoch": 0.72, "learning_rate": 1.9928564860903466e-06, "loss": 3.4296, "step": 6024 }, { "epoch": 0.72, "learning_rate": 1.991254185946408e-06, "loss": 3.3171, "step": 6025 }, { "epoch": 0.72, "learning_rate": 1.989652370031464e-06, "loss": 3.4803, "step": 6026 }, { "epoch": 0.72, "learning_rate": 1.988051038603312e-06, "loss": 3.3935, "step": 6027 }, { "epoch": 0.72, "learning_rate": 1.986450191919671e-06, "loss": 3.3868, "step": 6028 }, { "epoch": 0.72, "learning_rate": 1.9848498302381837e-06, "loss": 3.2718, "step": 6029 }, { "epoch": 0.72, "learning_rate": 1.983249953816409e-06, "loss": 3.4389, "step": 6030 }, { "epoch": 0.72, "learning_rate": 1.9816505629118344e-06, "loss": 3.3692, "step": 6031 }, { "epoch": 0.72, "learning_rate": 1.9800516577818662e-06, "loss": 3.3806, "step": 6032 }, { "epoch": 0.72, "learning_rate": 1.9784532386838336e-06, "loss": 3.3953, "step": 6033 }, { "epoch": 0.72, "learning_rate": 1.9768553058749855e-06, "loss": 3.4691, "step": 6034 }, { "epoch": 0.72, "learning_rate": 1.9752578596124955e-06, "loss": 3.2952, "step": 6035 }, { "epoch": 0.72, "learning_rate": 1.973660900153457e-06, "loss": 3.4983, "step": 6036 }, { "epoch": 0.72, "learning_rate": 1.9720644277548846e-06, "loss": 3.4004, "step": 6037 }, { "epoch": 0.72, "learning_rate": 1.9704684426737183e-06, "loss": 3.2993, "step": 6038 }, { "epoch": 0.72, "learning_rate": 1.9688729451668116e-06, "loss": 3.3382, "step": 6039 }, { "epoch": 0.73, "learning_rate": 1.967277935490948e-06, "loss": 3.465, "step": 6040 }, { "epoch": 0.73, "learning_rate": 1.9656834139028282e-06, "loss": 3.3861, "step": 6041 }, { "epoch": 0.73, "learning_rate": 1.9640893806590755e-06, "loss": 3.3252, "step": 6042 }, { "epoch": 0.73, "learning_rate": 1.9624958360162342e-06, "loss": 3.379, "step": 6043 }, { "epoch": 0.73, "learning_rate": 1.960902780230771e-06, "loss": 3.3948, "step": 6044 }, { "epoch": 0.73, "learning_rate": 1.959310213559072e-06, "loss": 3.3904, "step": 6045 }, { "epoch": 0.73, "learning_rate": 1.957718136257446e-06, "loss": 3.4686, "step": 6046 }, { "epoch": 0.73, "learning_rate": 1.956126548582123e-06, "loss": 3.3338, "step": 6047 }, { "epoch": 0.73, "learning_rate": 1.954535450789253e-06, "loss": 3.3826, "step": 6048 }, { "epoch": 0.73, "learning_rate": 1.952944843134909e-06, "loss": 3.4146, "step": 6049 }, { "epoch": 0.73, "learning_rate": 1.951354725875084e-06, "loss": 3.3896, "step": 6050 }, { "epoch": 0.73, "learning_rate": 1.949765099265692e-06, "loss": 3.5638, "step": 6051 }, { "epoch": 0.73, "learning_rate": 1.948175963562568e-06, "loss": 3.4886, "step": 6052 }, { "epoch": 0.73, "learning_rate": 1.946587319021469e-06, "loss": 3.3127, "step": 6053 }, { "epoch": 0.73, "learning_rate": 1.9449991658980716e-06, "loss": 3.4373, "step": 6054 }, { "epoch": 0.73, "learning_rate": 1.9434115044479747e-06, "loss": 3.4027, "step": 6055 }, { "epoch": 0.73, "learning_rate": 1.9418243349266974e-06, "loss": 3.294, "step": 6056 }, { "epoch": 0.73, "learning_rate": 1.9402376575896774e-06, "loss": 3.4254, "step": 6057 }, { "epoch": 0.73, "learning_rate": 1.9386514726922774e-06, "loss": 3.4051, "step": 6058 }, { "epoch": 0.73, "learning_rate": 1.9370657804897773e-06, "loss": 3.5131, "step": 6059 }, { "epoch": 0.73, "learning_rate": 1.9354805812373807e-06, "loss": 3.4488, "step": 6060 }, { "epoch": 0.73, "learning_rate": 1.933895875190209e-06, "loss": 3.4288, "step": 6061 }, { "epoch": 0.73, "learning_rate": 1.9323116626033067e-06, "loss": 3.4375, "step": 6062 }, { "epoch": 0.73, "learning_rate": 1.930727943731637e-06, "loss": 3.4545, "step": 6063 }, { "epoch": 0.73, "learning_rate": 1.9291447188300843e-06, "loss": 3.4487, "step": 6064 }, { "epoch": 0.73, "learning_rate": 1.927561988153456e-06, "loss": 3.4162, "step": 6065 }, { "epoch": 0.73, "learning_rate": 1.925979751956473e-06, "loss": 3.414, "step": 6066 }, { "epoch": 0.73, "learning_rate": 1.924398010493784e-06, "loss": 3.3924, "step": 6067 }, { "epoch": 0.73, "learning_rate": 1.9228167640199547e-06, "loss": 3.3659, "step": 6068 }, { "epoch": 0.73, "learning_rate": 1.921236012789472e-06, "loss": 3.4338, "step": 6069 }, { "epoch": 0.73, "learning_rate": 1.919655757056742e-06, "loss": 3.3686, "step": 6070 }, { "epoch": 0.73, "learning_rate": 1.9180759970760927e-06, "loss": 3.462, "step": 6071 }, { "epoch": 0.73, "learning_rate": 1.9164967331017707e-06, "loss": 3.3037, "step": 6072 }, { "epoch": 0.73, "learning_rate": 1.914917965387944e-06, "loss": 3.4586, "step": 6073 }, { "epoch": 0.73, "learning_rate": 1.9133396941887e-06, "loss": 3.4073, "step": 6074 }, { "epoch": 0.73, "learning_rate": 1.9117619197580467e-06, "loss": 3.3893, "step": 6075 }, { "epoch": 0.73, "learning_rate": 1.9101846423499113e-06, "loss": 3.2889, "step": 6076 }, { "epoch": 0.73, "learning_rate": 1.908607862218142e-06, "loss": 3.2774, "step": 6077 }, { "epoch": 0.73, "learning_rate": 1.9070315796165068e-06, "loss": 3.3392, "step": 6078 }, { "epoch": 0.73, "learning_rate": 1.905455794798693e-06, "loss": 3.4203, "step": 6079 }, { "epoch": 0.73, "learning_rate": 1.903880508018308e-06, "loss": 3.32, "step": 6080 }, { "epoch": 0.73, "learning_rate": 1.9023057195288796e-06, "loss": 3.3481, "step": 6081 }, { "epoch": 0.73, "learning_rate": 1.9007314295838548e-06, "loss": 3.3057, "step": 6082 }, { "epoch": 0.73, "learning_rate": 1.8991576384366023e-06, "loss": 3.3265, "step": 6083 }, { "epoch": 0.73, "learning_rate": 1.8975843463404054e-06, "loss": 3.378, "step": 6084 }, { "epoch": 0.73, "learning_rate": 1.8960115535484725e-06, "loss": 3.3978, "step": 6085 }, { "epoch": 0.73, "learning_rate": 1.8944392603139294e-06, "loss": 3.4132, "step": 6086 }, { "epoch": 0.73, "learning_rate": 1.892867466889821e-06, "loss": 3.3692, "step": 6087 }, { "epoch": 0.73, "learning_rate": 1.8912961735291142e-06, "loss": 3.276, "step": 6088 }, { "epoch": 0.73, "learning_rate": 1.8897253804846922e-06, "loss": 3.3514, "step": 6089 }, { "epoch": 0.73, "learning_rate": 1.8881550880093596e-06, "loss": 3.3194, "step": 6090 }, { "epoch": 0.73, "learning_rate": 1.8865852963558401e-06, "loss": 3.4163, "step": 6091 }, { "epoch": 0.73, "learning_rate": 1.8850160057767786e-06, "loss": 3.3826, "step": 6092 }, { "epoch": 0.73, "learning_rate": 1.8834472165247335e-06, "loss": 3.3001, "step": 6093 }, { "epoch": 0.73, "learning_rate": 1.8818789288521888e-06, "loss": 3.4763, "step": 6094 }, { "epoch": 0.73, "learning_rate": 1.880311143011545e-06, "loss": 3.3222, "step": 6095 }, { "epoch": 0.73, "learning_rate": 1.8787438592551232e-06, "loss": 3.3339, "step": 6096 }, { "epoch": 0.73, "learning_rate": 1.8771770778351617e-06, "loss": 3.3177, "step": 6097 }, { "epoch": 0.73, "learning_rate": 1.8756107990038197e-06, "loss": 3.3948, "step": 6098 }, { "epoch": 0.73, "learning_rate": 1.8740450230131747e-06, "loss": 3.3063, "step": 6099 }, { "epoch": 0.73, "learning_rate": 1.872479750115223e-06, "loss": 3.4544, "step": 6100 }, { "epoch": 0.73, "learning_rate": 1.8709149805618826e-06, "loss": 3.375, "step": 6101 }, { "epoch": 0.73, "learning_rate": 1.869350714604985e-06, "loss": 3.3641, "step": 6102 }, { "epoch": 0.73, "learning_rate": 1.8677869524962855e-06, "loss": 3.395, "step": 6103 }, { "epoch": 0.73, "learning_rate": 1.8662236944874562e-06, "loss": 3.3754, "step": 6104 }, { "epoch": 0.73, "learning_rate": 1.8646609408300892e-06, "loss": 3.4806, "step": 6105 }, { "epoch": 0.73, "learning_rate": 1.863098691775695e-06, "loss": 3.4071, "step": 6106 }, { "epoch": 0.73, "learning_rate": 1.8615369475757018e-06, "loss": 3.3825, "step": 6107 }, { "epoch": 0.73, "learning_rate": 1.859975708481458e-06, "loss": 3.3997, "step": 6108 }, { "epoch": 0.73, "learning_rate": 1.8584149747442304e-06, "loss": 3.384, "step": 6109 }, { "epoch": 0.73, "learning_rate": 1.8568547466152038e-06, "loss": 3.4128, "step": 6110 }, { "epoch": 0.73, "learning_rate": 1.855295024345482e-06, "loss": 3.3588, "step": 6111 }, { "epoch": 0.73, "learning_rate": 1.8537358081860879e-06, "loss": 3.4056, "step": 6112 }, { "epoch": 0.73, "learning_rate": 1.8521770983879629e-06, "loss": 3.3738, "step": 6113 }, { "epoch": 0.73, "learning_rate": 1.850618895201965e-06, "loss": 3.3992, "step": 6114 }, { "epoch": 0.73, "learning_rate": 1.8490611988788736e-06, "loss": 3.3149, "step": 6115 }, { "epoch": 0.73, "learning_rate": 1.8475040096693842e-06, "loss": 3.3348, "step": 6116 }, { "epoch": 0.73, "learning_rate": 1.8459473278241125e-06, "loss": 3.3263, "step": 6117 }, { "epoch": 0.73, "learning_rate": 1.8443911535935904e-06, "loss": 3.3556, "step": 6118 }, { "epoch": 0.73, "learning_rate": 1.8428354872282722e-06, "loss": 3.382, "step": 6119 }, { "epoch": 0.73, "learning_rate": 1.8412803289785236e-06, "loss": 3.3826, "step": 6120 }, { "epoch": 0.73, "learning_rate": 1.839725679094634e-06, "loss": 3.3251, "step": 6121 }, { "epoch": 0.73, "learning_rate": 1.83817153782681e-06, "loss": 3.3679, "step": 6122 }, { "epoch": 0.73, "learning_rate": 1.836617905425176e-06, "loss": 3.3839, "step": 6123 }, { "epoch": 0.74, "learning_rate": 1.8350647821397733e-06, "loss": 3.4071, "step": 6124 }, { "epoch": 0.74, "learning_rate": 1.8335121682205631e-06, "loss": 3.4044, "step": 6125 }, { "epoch": 0.74, "learning_rate": 1.831960063917424e-06, "loss": 3.378, "step": 6126 }, { "epoch": 0.74, "learning_rate": 1.830408469480152e-06, "loss": 3.4694, "step": 6127 }, { "epoch": 0.74, "learning_rate": 1.8288573851584634e-06, "loss": 3.325, "step": 6128 }, { "epoch": 0.74, "learning_rate": 1.827306811201987e-06, "loss": 3.4008, "step": 6129 }, { "epoch": 0.74, "learning_rate": 1.8257567478602745e-06, "loss": 3.2295, "step": 6130 }, { "epoch": 0.74, "learning_rate": 1.8242071953827939e-06, "loss": 3.4063, "step": 6131 }, { "epoch": 0.74, "learning_rate": 1.8226581540189303e-06, "loss": 3.4293, "step": 6132 }, { "epoch": 0.74, "learning_rate": 1.8211096240179886e-06, "loss": 3.4258, "step": 6133 }, { "epoch": 0.74, "learning_rate": 1.8195616056291886e-06, "loss": 3.3923, "step": 6134 }, { "epoch": 0.74, "learning_rate": 1.8180140991016698e-06, "loss": 3.3234, "step": 6135 }, { "epoch": 0.74, "learning_rate": 1.816467104684489e-06, "loss": 3.4338, "step": 6136 }, { "epoch": 0.74, "learning_rate": 1.8149206226266208e-06, "loss": 3.4284, "step": 6137 }, { "epoch": 0.74, "learning_rate": 1.8133746531769542e-06, "loss": 3.3983, "step": 6138 }, { "epoch": 0.74, "learning_rate": 1.8118291965842998e-06, "loss": 3.4291, "step": 6139 }, { "epoch": 0.74, "learning_rate": 1.810284253097384e-06, "loss": 3.4365, "step": 6140 }, { "epoch": 0.74, "learning_rate": 1.8087398229648505e-06, "loss": 3.325, "step": 6141 }, { "epoch": 0.74, "learning_rate": 1.8071959064352612e-06, "loss": 3.4566, "step": 6142 }, { "epoch": 0.74, "learning_rate": 1.8056525037570949e-06, "loss": 3.4492, "step": 6143 }, { "epoch": 0.74, "learning_rate": 1.8041096151787469e-06, "loss": 3.2961, "step": 6144 }, { "epoch": 0.74, "learning_rate": 1.8025672409485306e-06, "loss": 3.3179, "step": 6145 }, { "epoch": 0.74, "learning_rate": 1.8010253813146766e-06, "loss": 3.4096, "step": 6146 }, { "epoch": 0.74, "learning_rate": 1.7994840365253324e-06, "loss": 3.4137, "step": 6147 }, { "epoch": 0.74, "learning_rate": 1.797943206828563e-06, "loss": 3.4176, "step": 6148 }, { "epoch": 0.74, "learning_rate": 1.7964028924723504e-06, "loss": 3.4074, "step": 6149 }, { "epoch": 0.74, "learning_rate": 1.7948630937045925e-06, "loss": 3.4468, "step": 6150 }, { "epoch": 0.74, "learning_rate": 1.7933238107731065e-06, "loss": 3.3114, "step": 6151 }, { "epoch": 0.74, "learning_rate": 1.7917850439256251e-06, "loss": 3.4225, "step": 6152 }, { "epoch": 0.74, "learning_rate": 1.7902467934097973e-06, "loss": 3.4479, "step": 6153 }, { "epoch": 0.74, "learning_rate": 1.788709059473191e-06, "loss": 3.3185, "step": 6154 }, { "epoch": 0.74, "learning_rate": 1.787171842363291e-06, "loss": 3.4868, "step": 6155 }, { "epoch": 0.74, "learning_rate": 1.7856351423274942e-06, "loss": 3.4509, "step": 6156 }, { "epoch": 0.74, "learning_rate": 1.7840989596131196e-06, "loss": 3.3954, "step": 6157 }, { "epoch": 0.74, "learning_rate": 1.7825632944674016e-06, "loss": 3.3666, "step": 6158 }, { "epoch": 0.74, "learning_rate": 1.7810281471374907e-06, "loss": 3.4229, "step": 6159 }, { "epoch": 0.74, "learning_rate": 1.7794935178704536e-06, "loss": 3.3863, "step": 6160 }, { "epoch": 0.74, "learning_rate": 1.7779594069132755e-06, "loss": 3.3269, "step": 6161 }, { "epoch": 0.74, "learning_rate": 1.776425814512856e-06, "loss": 3.3681, "step": 6162 }, { "epoch": 0.74, "learning_rate": 1.7748927409160122e-06, "loss": 3.4997, "step": 6163 }, { "epoch": 0.74, "learning_rate": 1.7733601863694798e-06, "loss": 3.4623, "step": 6164 }, { "epoch": 0.74, "learning_rate": 1.7718281511199054e-06, "loss": 3.3684, "step": 6165 }, { "epoch": 0.74, "learning_rate": 1.7702966354138573e-06, "loss": 3.2747, "step": 6166 }, { "epoch": 0.74, "learning_rate": 1.768765639497818e-06, "loss": 3.4763, "step": 6167 }, { "epoch": 0.74, "learning_rate": 1.7672351636181873e-06, "loss": 3.4227, "step": 6168 }, { "epoch": 0.74, "learning_rate": 1.76570520802128e-06, "loss": 3.4258, "step": 6169 }, { "epoch": 0.74, "learning_rate": 1.7641757729533281e-06, "loss": 3.3932, "step": 6170 }, { "epoch": 0.74, "learning_rate": 1.7626468586604795e-06, "loss": 3.3632, "step": 6171 }, { "epoch": 0.74, "learning_rate": 1.761118465388799e-06, "loss": 3.3722, "step": 6172 }, { "epoch": 0.74, "learning_rate": 1.7595905933842666e-06, "loss": 3.4152, "step": 6173 }, { "epoch": 0.74, "learning_rate": 1.758063242892779e-06, "loss": 3.3637, "step": 6174 }, { "epoch": 0.74, "learning_rate": 1.7565364141601483e-06, "loss": 3.3356, "step": 6175 }, { "epoch": 0.74, "learning_rate": 1.755010107432103e-06, "loss": 3.3903, "step": 6176 }, { "epoch": 0.74, "learning_rate": 1.7534843229542876e-06, "loss": 3.3582, "step": 6177 }, { "epoch": 0.74, "learning_rate": 1.751959060972263e-06, "loss": 3.4306, "step": 6178 }, { "epoch": 0.74, "learning_rate": 1.7504343217315056e-06, "loss": 3.4116, "step": 6179 }, { "epoch": 0.74, "learning_rate": 1.7489101054774077e-06, "loss": 3.3388, "step": 6180 }, { "epoch": 0.74, "learning_rate": 1.7473864124552769e-06, "loss": 3.3076, "step": 6181 }, { "epoch": 0.74, "learning_rate": 1.7458632429103389e-06, "loss": 3.3683, "step": 6182 }, { "epoch": 0.74, "learning_rate": 1.7443405970877303e-06, "loss": 3.3613, "step": 6183 }, { "epoch": 0.74, "learning_rate": 1.7428184752325083e-06, "loss": 3.3849, "step": 6184 }, { "epoch": 0.74, "learning_rate": 1.7412968775896433e-06, "loss": 3.2847, "step": 6185 }, { "epoch": 0.74, "learning_rate": 1.739775804404022e-06, "loss": 3.4022, "step": 6186 }, { "epoch": 0.74, "learning_rate": 1.738255255920447e-06, "loss": 3.3443, "step": 6187 }, { "epoch": 0.74, "learning_rate": 1.7367352323836362e-06, "loss": 3.4119, "step": 6188 }, { "epoch": 0.74, "learning_rate": 1.7352157340382226e-06, "loss": 3.3915, "step": 6189 }, { "epoch": 0.74, "learning_rate": 1.7336967611287547e-06, "loss": 3.3074, "step": 6190 }, { "epoch": 0.74, "learning_rate": 1.7321783138996995e-06, "loss": 3.3615, "step": 6191 }, { "epoch": 0.74, "learning_rate": 1.7306603925954319e-06, "loss": 3.4248, "step": 6192 }, { "epoch": 0.74, "learning_rate": 1.7291429974602496e-06, "loss": 3.475, "step": 6193 }, { "epoch": 0.74, "learning_rate": 1.7276261287383628e-06, "loss": 3.445, "step": 6194 }, { "epoch": 0.74, "learning_rate": 1.7261097866738962e-06, "loss": 3.523, "step": 6195 }, { "epoch": 0.74, "learning_rate": 1.7245939715108918e-06, "loss": 3.424, "step": 6196 }, { "epoch": 0.74, "learning_rate": 1.7230786834933045e-06, "loss": 3.3784, "step": 6197 }, { "epoch": 0.74, "learning_rate": 1.7215639228650067e-06, "loss": 3.3939, "step": 6198 }, { "epoch": 0.74, "learning_rate": 1.7200496898697832e-06, "loss": 3.413, "step": 6199 }, { "epoch": 0.74, "learning_rate": 1.7185359847513384e-06, "loss": 3.4597, "step": 6200 }, { "epoch": 0.74, "learning_rate": 1.7170228077532847e-06, "loss": 3.3367, "step": 6201 }, { "epoch": 0.74, "learning_rate": 1.7155101591191553e-06, "loss": 3.2718, "step": 6202 }, { "epoch": 0.74, "learning_rate": 1.7139980390923971e-06, "loss": 3.429, "step": 6203 }, { "epoch": 0.74, "learning_rate": 1.7124864479163706e-06, "loss": 3.3793, "step": 6204 }, { "epoch": 0.74, "learning_rate": 1.7109753858343526e-06, "loss": 3.4208, "step": 6205 }, { "epoch": 0.74, "learning_rate": 1.7094648530895341e-06, "loss": 3.3484, "step": 6206 }, { "epoch": 0.75, "learning_rate": 1.7079548499250203e-06, "loss": 3.3258, "step": 6207 }, { "epoch": 0.75, "learning_rate": 1.7064453765838329e-06, "loss": 3.3115, "step": 6208 }, { "epoch": 0.75, "learning_rate": 1.7049364333089062e-06, "loss": 3.2849, "step": 6209 }, { "epoch": 0.75, "learning_rate": 1.7034280203430914e-06, "loss": 3.3064, "step": 6210 }, { "epoch": 0.75, "learning_rate": 1.7019201379291522e-06, "loss": 3.4753, "step": 6211 }, { "epoch": 0.75, "learning_rate": 1.7004127863097685e-06, "loss": 3.3766, "step": 6212 }, { "epoch": 0.75, "learning_rate": 1.6989059657275343e-06, "loss": 3.4854, "step": 6213 }, { "epoch": 0.75, "learning_rate": 1.6973996764249572e-06, "loss": 3.3616, "step": 6214 }, { "epoch": 0.75, "learning_rate": 1.6958939186444607e-06, "loss": 3.4366, "step": 6215 }, { "epoch": 0.75, "learning_rate": 1.694388692628382e-06, "loss": 3.3435, "step": 6216 }, { "epoch": 0.75, "learning_rate": 1.6928839986189732e-06, "loss": 3.3714, "step": 6217 }, { "epoch": 0.75, "learning_rate": 1.6913798368584022e-06, "loss": 3.3218, "step": 6218 }, { "epoch": 0.75, "learning_rate": 1.6898762075887454e-06, "loss": 3.4604, "step": 6219 }, { "epoch": 0.75, "learning_rate": 1.6883731110520002e-06, "loss": 3.3778, "step": 6220 }, { "epoch": 0.75, "learning_rate": 1.6868705474900754e-06, "loss": 3.3961, "step": 6221 }, { "epoch": 0.75, "learning_rate": 1.6853685171447942e-06, "loss": 3.4263, "step": 6222 }, { "epoch": 0.75, "learning_rate": 1.6838670202578939e-06, "loss": 3.4816, "step": 6223 }, { "epoch": 0.75, "learning_rate": 1.6823660570710265e-06, "loss": 3.3423, "step": 6224 }, { "epoch": 0.75, "learning_rate": 1.6808656278257574e-06, "loss": 3.4325, "step": 6225 }, { "epoch": 0.75, "learning_rate": 1.6793657327635665e-06, "loss": 3.4488, "step": 6226 }, { "epoch": 0.75, "learning_rate": 1.6778663721258499e-06, "loss": 3.3166, "step": 6227 }, { "epoch": 0.75, "learning_rate": 1.676367546153911e-06, "loss": 3.4092, "step": 6228 }, { "epoch": 0.75, "learning_rate": 1.6748692550889738e-06, "loss": 3.3656, "step": 6229 }, { "epoch": 0.75, "learning_rate": 1.6733714991721738e-06, "loss": 3.3449, "step": 6230 }, { "epoch": 0.75, "learning_rate": 1.6718742786445613e-06, "loss": 3.4804, "step": 6231 }, { "epoch": 0.75, "learning_rate": 1.670377593747099e-06, "loss": 3.394, "step": 6232 }, { "epoch": 0.75, "learning_rate": 1.6688814447206642e-06, "loss": 3.35, "step": 6233 }, { "epoch": 0.75, "learning_rate": 1.6673858318060476e-06, "loss": 3.3328, "step": 6234 }, { "epoch": 0.75, "learning_rate": 1.665890755243954e-06, "loss": 3.4298, "step": 6235 }, { "epoch": 0.75, "learning_rate": 1.6643962152750037e-06, "loss": 3.3919, "step": 6236 }, { "epoch": 0.75, "learning_rate": 1.6629022121397248e-06, "loss": 3.4738, "step": 6237 }, { "epoch": 0.75, "learning_rate": 1.6614087460785655e-06, "loss": 3.4802, "step": 6238 }, { "epoch": 0.75, "learning_rate": 1.659915817331884e-06, "loss": 3.3838, "step": 6239 }, { "epoch": 0.75, "learning_rate": 1.6584234261399535e-06, "loss": 3.451, "step": 6240 }, { "epoch": 0.75, "learning_rate": 1.6569315727429596e-06, "loss": 3.4179, "step": 6241 }, { "epoch": 0.75, "learning_rate": 1.655440257381003e-06, "loss": 3.3508, "step": 6242 }, { "epoch": 0.75, "learning_rate": 1.6539494802940952e-06, "loss": 3.4036, "step": 6243 }, { "epoch": 0.75, "learning_rate": 1.6524592417221635e-06, "loss": 3.4723, "step": 6244 }, { "epoch": 0.75, "learning_rate": 1.650969541905048e-06, "loss": 3.3347, "step": 6245 }, { "epoch": 0.75, "learning_rate": 1.6494803810825005e-06, "loss": 3.3619, "step": 6246 }, { "epoch": 0.75, "learning_rate": 1.6479917594941885e-06, "loss": 3.3473, "step": 6247 }, { "epoch": 0.75, "learning_rate": 1.6465036773796905e-06, "loss": 3.2849, "step": 6248 }, { "epoch": 0.75, "learning_rate": 1.6450161349784999e-06, "loss": 3.4306, "step": 6249 }, { "epoch": 0.75, "learning_rate": 1.6435291325300218e-06, "loss": 3.3689, "step": 6250 }, { "epoch": 0.75, "learning_rate": 1.6420426702735752e-06, "loss": 3.4386, "step": 6251 }, { "epoch": 0.75, "learning_rate": 1.640556748448393e-06, "loss": 3.4002, "step": 6252 }, { "epoch": 0.75, "learning_rate": 1.6390713672936192e-06, "loss": 3.3743, "step": 6253 }, { "epoch": 0.75, "learning_rate": 1.6375865270483132e-06, "loss": 3.4382, "step": 6254 }, { "epoch": 0.75, "learning_rate": 1.6361022279514438e-06, "loss": 3.3544, "step": 6255 }, { "epoch": 0.75, "learning_rate": 1.6346184702418955e-06, "loss": 3.3684, "step": 6256 }, { "epoch": 0.75, "learning_rate": 1.633135254158465e-06, "loss": 3.373, "step": 6257 }, { "epoch": 0.75, "learning_rate": 1.6316525799398618e-06, "loss": 3.428, "step": 6258 }, { "epoch": 0.75, "learning_rate": 1.6301704478247089e-06, "loss": 3.4045, "step": 6259 }, { "epoch": 0.75, "learning_rate": 1.6286888580515403e-06, "loss": 3.4747, "step": 6260 }, { "epoch": 0.75, "learning_rate": 1.6272078108588046e-06, "loss": 3.4438, "step": 6261 }, { "epoch": 0.75, "learning_rate": 1.6257273064848617e-06, "loss": 3.4086, "step": 6262 }, { "epoch": 0.75, "learning_rate": 1.624247345167987e-06, "loss": 3.4158, "step": 6263 }, { "epoch": 0.75, "learning_rate": 1.6227679271463626e-06, "loss": 3.3601, "step": 6264 }, { "epoch": 0.75, "learning_rate": 1.6212890526580876e-06, "loss": 3.4357, "step": 6265 }, { "epoch": 0.75, "learning_rate": 1.619810721941174e-06, "loss": 3.404, "step": 6266 }, { "epoch": 0.75, "learning_rate": 1.618332935233544e-06, "loss": 3.3537, "step": 6267 }, { "epoch": 0.75, "learning_rate": 1.6168556927730344e-06, "loss": 3.366, "step": 6268 }, { "epoch": 0.75, "learning_rate": 1.615378994797393e-06, "loss": 3.434, "step": 6269 }, { "epoch": 0.75, "learning_rate": 1.6139028415442792e-06, "loss": 3.4642, "step": 6270 }, { "epoch": 0.75, "learning_rate": 1.6124272332512668e-06, "loss": 3.4016, "step": 6271 }, { "epoch": 0.75, "learning_rate": 1.6109521701558412e-06, "loss": 3.5386, "step": 6272 }, { "epoch": 0.75, "learning_rate": 1.6094776524953987e-06, "loss": 3.245, "step": 6273 }, { "epoch": 0.75, "learning_rate": 1.6080036805072496e-06, "loss": 3.4245, "step": 6274 }, { "epoch": 0.75, "learning_rate": 1.606530254428616e-06, "loss": 3.4808, "step": 6275 }, { "epoch": 0.75, "learning_rate": 1.6050573744966307e-06, "loss": 3.3669, "step": 6276 }, { "epoch": 0.75, "learning_rate": 1.6035850409483406e-06, "loss": 3.3826, "step": 6277 }, { "epoch": 0.75, "learning_rate": 1.6021132540207035e-06, "loss": 3.3518, "step": 6278 }, { "epoch": 0.75, "learning_rate": 1.6006420139505895e-06, "loss": 3.3717, "step": 6279 }, { "epoch": 0.75, "learning_rate": 1.5991713209747806e-06, "loss": 3.4226, "step": 6280 }, { "epoch": 0.75, "learning_rate": 1.5977011753299726e-06, "loss": 3.4211, "step": 6281 }, { "epoch": 0.75, "learning_rate": 1.5962315772527681e-06, "loss": 3.4237, "step": 6282 }, { "epoch": 0.75, "learning_rate": 1.5947625269796863e-06, "loss": 3.5022, "step": 6283 }, { "epoch": 0.75, "learning_rate": 1.5932940247471574e-06, "loss": 3.4145, "step": 6284 }, { "epoch": 0.75, "learning_rate": 1.5918260707915223e-06, "loss": 3.4118, "step": 6285 }, { "epoch": 0.75, "learning_rate": 1.5903586653490349e-06, "loss": 3.4035, "step": 6286 }, { "epoch": 0.75, "learning_rate": 1.5888918086558596e-06, "loss": 3.4074, "step": 6287 }, { "epoch": 0.75, "learning_rate": 1.5874255009480728e-06, "loss": 3.3202, "step": 6288 }, { "epoch": 0.75, "learning_rate": 1.5859597424616635e-06, "loss": 3.3862, "step": 6289 }, { "epoch": 0.76, "learning_rate": 1.584494533432533e-06, "loss": 3.398, "step": 6290 }, { "epoch": 0.76, "learning_rate": 1.583029874096489e-06, "loss": 3.4206, "step": 6291 }, { "epoch": 0.76, "learning_rate": 1.5815657646892563e-06, "loss": 3.4582, "step": 6292 }, { "epoch": 0.76, "learning_rate": 1.5801022054464693e-06, "loss": 3.344, "step": 6293 }, { "epoch": 0.76, "learning_rate": 1.5786391966036746e-06, "loss": 3.4333, "step": 6294 }, { "epoch": 0.76, "learning_rate": 1.5771767383963289e-06, "loss": 3.3778, "step": 6295 }, { "epoch": 0.76, "learning_rate": 1.5757148310598009e-06, "loss": 3.4456, "step": 6296 }, { "epoch": 0.76, "learning_rate": 1.574253474829371e-06, "loss": 3.3385, "step": 6297 }, { "epoch": 0.76, "learning_rate": 1.57279266994023e-06, "loss": 3.3996, "step": 6298 }, { "epoch": 0.76, "learning_rate": 1.571332416627483e-06, "loss": 3.3281, "step": 6299 }, { "epoch": 0.76, "learning_rate": 1.5698727151261406e-06, "loss": 3.4039, "step": 6300 }, { "epoch": 0.76, "learning_rate": 1.5684135656711286e-06, "loss": 3.366, "step": 6301 }, { "epoch": 0.76, "learning_rate": 1.5669549684972834e-06, "loss": 3.485, "step": 6302 }, { "epoch": 0.76, "learning_rate": 1.565496923839353e-06, "loss": 3.3295, "step": 6303 }, { "epoch": 0.76, "learning_rate": 1.5640394319319952e-06, "loss": 3.4908, "step": 6304 }, { "epoch": 0.76, "learning_rate": 1.5625824930097798e-06, "loss": 3.4012, "step": 6305 }, { "epoch": 0.76, "learning_rate": 1.5611261073071876e-06, "loss": 3.3835, "step": 6306 }, { "epoch": 0.76, "learning_rate": 1.559670275058609e-06, "loss": 3.3659, "step": 6307 }, { "epoch": 0.76, "learning_rate": 1.5582149964983467e-06, "loss": 3.4723, "step": 6308 }, { "epoch": 0.76, "learning_rate": 1.5567602718606145e-06, "loss": 3.3934, "step": 6309 }, { "epoch": 0.76, "learning_rate": 1.5553061013795362e-06, "loss": 3.3393, "step": 6310 }, { "epoch": 0.76, "learning_rate": 1.5538524852891463e-06, "loss": 3.4067, "step": 6311 }, { "epoch": 0.76, "learning_rate": 1.5523994238233909e-06, "loss": 3.4462, "step": 6312 }, { "epoch": 0.76, "learning_rate": 1.5509469172161263e-06, "loss": 3.3453, "step": 6313 }, { "epoch": 0.76, "learning_rate": 1.5494949657011194e-06, "loss": 3.4071, "step": 6314 }, { "epoch": 0.76, "learning_rate": 1.5480435695120482e-06, "loss": 3.4071, "step": 6315 }, { "epoch": 0.76, "learning_rate": 1.5465927288825012e-06, "loss": 3.389, "step": 6316 }, { "epoch": 0.76, "learning_rate": 1.545142444045979e-06, "loss": 3.4288, "step": 6317 }, { "epoch": 0.76, "learning_rate": 1.5436927152358871e-06, "loss": 3.3463, "step": 6318 }, { "epoch": 0.76, "learning_rate": 1.5422435426855487e-06, "loss": 3.3858, "step": 6319 }, { "epoch": 0.76, "learning_rate": 1.540794926628193e-06, "loss": 3.4593, "step": 6320 }, { "epoch": 0.76, "learning_rate": 1.5393468672969614e-06, "loss": 3.3914, "step": 6321 }, { "epoch": 0.76, "learning_rate": 1.5378993649249053e-06, "loss": 3.3885, "step": 6322 }, { "epoch": 0.76, "learning_rate": 1.536452419744986e-06, "loss": 3.3395, "step": 6323 }, { "epoch": 0.76, "learning_rate": 1.5350060319900768e-06, "loss": 3.387, "step": 6324 }, { "epoch": 0.76, "learning_rate": 1.5335602018929586e-06, "loss": 3.3799, "step": 6325 }, { "epoch": 0.76, "learning_rate": 1.5321149296863264e-06, "loss": 3.3427, "step": 6326 }, { "epoch": 0.76, "learning_rate": 1.5306702156027792e-06, "loss": 3.3458, "step": 6327 }, { "epoch": 0.76, "learning_rate": 1.529226059874832e-06, "loss": 3.3979, "step": 6328 }, { "epoch": 0.76, "learning_rate": 1.5277824627349086e-06, "loss": 3.3385, "step": 6329 }, { "epoch": 0.76, "learning_rate": 1.5263394244153407e-06, "loss": 3.3428, "step": 6330 }, { "epoch": 0.76, "learning_rate": 1.524896945148373e-06, "loss": 3.2943, "step": 6331 }, { "epoch": 0.76, "learning_rate": 1.523455025166158e-06, "loss": 3.4177, "step": 6332 }, { "epoch": 0.76, "learning_rate": 1.5220136647007588e-06, "loss": 3.3448, "step": 6333 }, { "epoch": 0.76, "learning_rate": 1.5205728639841494e-06, "loss": 3.4791, "step": 6334 }, { "epoch": 0.76, "learning_rate": 1.519132623248214e-06, "loss": 3.4798, "step": 6335 }, { "epoch": 0.76, "learning_rate": 1.5176929427247427e-06, "loss": 3.3944, "step": 6336 }, { "epoch": 0.76, "learning_rate": 1.51625382264544e-06, "loss": 3.3821, "step": 6337 }, { "epoch": 0.76, "learning_rate": 1.5148152632419182e-06, "loss": 3.395, "step": 6338 }, { "epoch": 0.76, "learning_rate": 1.5133772647457002e-06, "loss": 3.4275, "step": 6339 }, { "epoch": 0.76, "learning_rate": 1.5119398273882179e-06, "loss": 3.336, "step": 6340 }, { "epoch": 0.76, "learning_rate": 1.5105029514008134e-06, "loss": 3.3777, "step": 6341 }, { "epoch": 0.76, "learning_rate": 1.5090666370147378e-06, "loss": 3.2196, "step": 6342 }, { "epoch": 0.76, "learning_rate": 1.5076308844611527e-06, "loss": 3.3918, "step": 6343 }, { "epoch": 0.76, "learning_rate": 1.5061956939711282e-06, "loss": 3.3604, "step": 6344 }, { "epoch": 0.76, "learning_rate": 1.504761065775645e-06, "loss": 3.4006, "step": 6345 }, { "epoch": 0.76, "learning_rate": 1.5033270001055932e-06, "loss": 3.4063, "step": 6346 }, { "epoch": 0.76, "learning_rate": 1.5018934971917714e-06, "loss": 3.4123, "step": 6347 }, { "epoch": 0.76, "learning_rate": 1.5004605572648884e-06, "loss": 3.4535, "step": 6348 }, { "epoch": 0.76, "learning_rate": 1.4990281805555618e-06, "loss": 3.4455, "step": 6349 }, { "epoch": 0.76, "learning_rate": 1.49759636729432e-06, "loss": 3.3877, "step": 6350 }, { "epoch": 0.76, "learning_rate": 1.4961651177115993e-06, "loss": 3.3364, "step": 6351 }, { "epoch": 0.76, "learning_rate": 1.4947344320377449e-06, "loss": 3.3055, "step": 6352 }, { "epoch": 0.76, "learning_rate": 1.4933043105030149e-06, "loss": 3.4415, "step": 6353 }, { "epoch": 0.76, "learning_rate": 1.49187475333757e-06, "loss": 3.4138, "step": 6354 }, { "epoch": 0.76, "learning_rate": 1.4904457607714852e-06, "loss": 3.4026, "step": 6355 }, { "epoch": 0.76, "learning_rate": 1.489017333034743e-06, "loss": 3.3836, "step": 6356 }, { "epoch": 0.76, "learning_rate": 1.4875894703572362e-06, "loss": 3.3943, "step": 6357 }, { "epoch": 0.76, "learning_rate": 1.4861621729687654e-06, "loss": 3.382, "step": 6358 }, { "epoch": 0.76, "learning_rate": 1.4847354410990394e-06, "loss": 3.3405, "step": 6359 }, { "epoch": 0.76, "learning_rate": 1.483309274977679e-06, "loss": 3.3023, "step": 6360 }, { "epoch": 0.76, "learning_rate": 1.4818836748342102e-06, "loss": 3.3044, "step": 6361 }, { "epoch": 0.76, "learning_rate": 1.4804586408980726e-06, "loss": 3.4006, "step": 6362 }, { "epoch": 0.76, "learning_rate": 1.4790341733986085e-06, "loss": 3.3349, "step": 6363 }, { "epoch": 0.76, "learning_rate": 1.4776102725650737e-06, "loss": 3.3899, "step": 6364 }, { "epoch": 0.76, "learning_rate": 1.4761869386266314e-06, "loss": 3.5197, "step": 6365 }, { "epoch": 0.76, "learning_rate": 1.474764171812354e-06, "loss": 3.3681, "step": 6366 }, { "epoch": 0.76, "learning_rate": 1.4733419723512215e-06, "loss": 3.4622, "step": 6367 }, { "epoch": 0.76, "learning_rate": 1.4719203404721244e-06, "loss": 3.4165, "step": 6368 }, { "epoch": 0.76, "learning_rate": 1.4704992764038605e-06, "loss": 3.4237, "step": 6369 }, { "epoch": 0.76, "learning_rate": 1.469078780375136e-06, "loss": 3.3905, "step": 6370 }, { "epoch": 0.76, "learning_rate": 1.4676588526145668e-06, "loss": 3.4497, "step": 6371 }, { "epoch": 0.76, "learning_rate": 1.4662394933506768e-06, "loss": 3.4834, "step": 6372 }, { "epoch": 0.76, "learning_rate": 1.4648207028118977e-06, "loss": 3.3706, "step": 6373 }, { "epoch": 0.77, "learning_rate": 1.4634024812265712e-06, "loss": 3.3004, "step": 6374 }, { "epoch": 0.77, "learning_rate": 1.4619848288229454e-06, "loss": 3.3611, "step": 6375 }, { "epoch": 0.77, "learning_rate": 1.4605677458291794e-06, "loss": 3.4385, "step": 6376 }, { "epoch": 0.77, "learning_rate": 1.4591512324733382e-06, "loss": 3.3396, "step": 6377 }, { "epoch": 0.77, "learning_rate": 1.4577352889833963e-06, "loss": 3.3569, "step": 6378 }, { "epoch": 0.77, "learning_rate": 1.4563199155872366e-06, "loss": 3.3911, "step": 6379 }, { "epoch": 0.77, "learning_rate": 1.4549051125126495e-06, "loss": 3.4798, "step": 6380 }, { "epoch": 0.77, "learning_rate": 1.4534908799873365e-06, "loss": 3.3684, "step": 6381 }, { "epoch": 0.77, "learning_rate": 1.452077218238901e-06, "loss": 3.353, "step": 6382 }, { "epoch": 0.77, "learning_rate": 1.45066412749486e-06, "loss": 3.5552, "step": 6383 }, { "epoch": 0.77, "learning_rate": 1.4492516079826374e-06, "loss": 3.4244, "step": 6384 }, { "epoch": 0.77, "learning_rate": 1.4478396599295641e-06, "loss": 3.4249, "step": 6385 }, { "epoch": 0.77, "learning_rate": 1.4464282835628807e-06, "loss": 3.4587, "step": 6386 }, { "epoch": 0.77, "learning_rate": 1.4450174791097339e-06, "loss": 3.3096, "step": 6387 }, { "epoch": 0.77, "learning_rate": 1.4436072467971796e-06, "loss": 3.3849, "step": 6388 }, { "epoch": 0.77, "learning_rate": 1.4421975868521815e-06, "loss": 3.3394, "step": 6389 }, { "epoch": 0.77, "learning_rate": 1.440788499501612e-06, "loss": 3.3184, "step": 6390 }, { "epoch": 0.77, "learning_rate": 1.4393799849722472e-06, "loss": 3.3531, "step": 6391 }, { "epoch": 0.77, "learning_rate": 1.4379720434907761e-06, "loss": 3.444, "step": 6392 }, { "epoch": 0.77, "learning_rate": 1.4365646752837931e-06, "loss": 3.3337, "step": 6393 }, { "epoch": 0.77, "learning_rate": 1.4351578805778011e-06, "loss": 3.3725, "step": 6394 }, { "epoch": 0.77, "learning_rate": 1.4337516595992102e-06, "loss": 3.3356, "step": 6395 }, { "epoch": 0.77, "learning_rate": 1.4323460125743378e-06, "loss": 3.3494, "step": 6396 }, { "epoch": 0.77, "learning_rate": 1.43094093972941e-06, "loss": 3.3877, "step": 6397 }, { "epoch": 0.77, "learning_rate": 1.4295364412905594e-06, "loss": 3.4679, "step": 6398 }, { "epoch": 0.77, "learning_rate": 1.428132517483829e-06, "loss": 3.3774, "step": 6399 }, { "epoch": 0.77, "learning_rate": 1.426729168535163e-06, "loss": 3.3521, "step": 6400 }, { "epoch": 0.77, "learning_rate": 1.4253263946704187e-06, "loss": 3.4454, "step": 6401 }, { "epoch": 0.77, "learning_rate": 1.4239241961153599e-06, "loss": 3.455, "step": 6402 }, { "epoch": 0.77, "learning_rate": 1.4225225730956565e-06, "loss": 3.3907, "step": 6403 }, { "epoch": 0.77, "learning_rate": 1.4211215258368866e-06, "loss": 3.4739, "step": 6404 }, { "epoch": 0.77, "learning_rate": 1.419721054564535e-06, "loss": 3.4266, "step": 6405 }, { "epoch": 0.77, "learning_rate": 1.4183211595039948e-06, "loss": 3.4379, "step": 6406 }, { "epoch": 0.77, "learning_rate": 1.4169218408805657e-06, "loss": 3.4894, "step": 6407 }, { "epoch": 0.77, "learning_rate": 1.4155230989194547e-06, "loss": 3.3122, "step": 6408 }, { "epoch": 0.77, "learning_rate": 1.4141249338457752e-06, "loss": 3.257, "step": 6409 }, { "epoch": 0.77, "learning_rate": 1.4127273458845498e-06, "loss": 3.3243, "step": 6410 }, { "epoch": 0.77, "learning_rate": 1.4113303352607065e-06, "loss": 3.3964, "step": 6411 }, { "epoch": 0.77, "learning_rate": 1.4099339021990805e-06, "loss": 3.4111, "step": 6412 }, { "epoch": 0.77, "learning_rate": 1.4085380469244148e-06, "loss": 3.3913, "step": 6413 }, { "epoch": 0.77, "learning_rate": 1.4071427696613587e-06, "loss": 3.3777, "step": 6414 }, { "epoch": 0.77, "learning_rate": 1.405748070634469e-06, "loss": 3.426, "step": 6415 }, { "epoch": 0.77, "learning_rate": 1.4043539500682091e-06, "loss": 3.3935, "step": 6416 }, { "epoch": 0.77, "learning_rate": 1.4029604081869513e-06, "loss": 3.3511, "step": 6417 }, { "epoch": 0.77, "learning_rate": 1.4015674452149698e-06, "loss": 3.4328, "step": 6418 }, { "epoch": 0.77, "learning_rate": 1.4001750613764492e-06, "loss": 3.399, "step": 6419 }, { "epoch": 0.77, "learning_rate": 1.3987832568954823e-06, "loss": 3.4431, "step": 6420 }, { "epoch": 0.77, "learning_rate": 1.3973920319960654e-06, "loss": 3.3311, "step": 6421 }, { "epoch": 0.77, "learning_rate": 1.3960013869021032e-06, "loss": 3.4396, "step": 6422 }, { "epoch": 0.77, "learning_rate": 1.3946113218374069e-06, "loss": 3.4615, "step": 6423 }, { "epoch": 0.77, "learning_rate": 1.3932218370256939e-06, "loss": 3.4542, "step": 6424 }, { "epoch": 0.77, "learning_rate": 1.3918329326905893e-06, "loss": 3.4159, "step": 6425 }, { "epoch": 0.77, "learning_rate": 1.390444609055625e-06, "loss": 3.4204, "step": 6426 }, { "epoch": 0.77, "learning_rate": 1.3890568663442354e-06, "loss": 3.4177, "step": 6427 }, { "epoch": 0.77, "learning_rate": 1.3876697047797665e-06, "loss": 3.4085, "step": 6428 }, { "epoch": 0.77, "learning_rate": 1.386283124585468e-06, "loss": 3.3484, "step": 6429 }, { "epoch": 0.77, "learning_rate": 1.3848971259844974e-06, "loss": 3.37, "step": 6430 }, { "epoch": 0.77, "learning_rate": 1.3835117091999178e-06, "loss": 3.3845, "step": 6431 }, { "epoch": 0.77, "learning_rate": 1.3821268744546984e-06, "loss": 3.3642, "step": 6432 }, { "epoch": 0.77, "learning_rate": 1.3807426219717157e-06, "loss": 3.4348, "step": 6433 }, { "epoch": 0.77, "learning_rate": 1.3793589519737522e-06, "loss": 3.4662, "step": 6434 }, { "epoch": 0.77, "learning_rate": 1.3779758646834968e-06, "loss": 3.4567, "step": 6435 }, { "epoch": 0.77, "learning_rate": 1.3765933603235426e-06, "loss": 3.4927, "step": 6436 }, { "epoch": 0.77, "learning_rate": 1.3752114391163913e-06, "loss": 3.4067, "step": 6437 }, { "epoch": 0.77, "learning_rate": 1.3738301012844497e-06, "loss": 3.3583, "step": 6438 }, { "epoch": 0.77, "learning_rate": 1.3724493470500316e-06, "loss": 3.4367, "step": 6439 }, { "epoch": 0.77, "learning_rate": 1.3710691766353557e-06, "loss": 3.3722, "step": 6440 }, { "epoch": 0.77, "learning_rate": 1.3696895902625463e-06, "loss": 3.3717, "step": 6441 }, { "epoch": 0.77, "learning_rate": 1.3683105881536373e-06, "loss": 3.3376, "step": 6442 }, { "epoch": 0.77, "learning_rate": 1.3669321705305643e-06, "loss": 3.3434, "step": 6443 }, { "epoch": 0.77, "learning_rate": 1.365554337615173e-06, "loss": 3.407, "step": 6444 }, { "epoch": 0.77, "learning_rate": 1.3641770896292083e-06, "loss": 3.3978, "step": 6445 }, { "epoch": 0.77, "learning_rate": 1.3628004267943273e-06, "loss": 3.4612, "step": 6446 }, { "epoch": 0.77, "learning_rate": 1.3614243493320906e-06, "loss": 3.4982, "step": 6447 }, { "epoch": 0.77, "learning_rate": 1.3600488574639648e-06, "loss": 3.4175, "step": 6448 }, { "epoch": 0.77, "learning_rate": 1.3586739514113222e-06, "loss": 3.4928, "step": 6449 }, { "epoch": 0.77, "learning_rate": 1.3572996313954407e-06, "loss": 3.405, "step": 6450 }, { "epoch": 0.77, "learning_rate": 1.355925897637504e-06, "loss": 3.4103, "step": 6451 }, { "epoch": 0.77, "learning_rate": 1.354552750358602e-06, "loss": 3.4471, "step": 6452 }, { "epoch": 0.77, "learning_rate": 1.353180189779731e-06, "loss": 3.3547, "step": 6453 }, { "epoch": 0.77, "learning_rate": 1.3518082161217877e-06, "loss": 3.3928, "step": 6454 }, { "epoch": 0.77, "learning_rate": 1.3504368296055808e-06, "loss": 3.392, "step": 6455 }, { "epoch": 0.77, "learning_rate": 1.3490660304518215e-06, "loss": 3.3649, "step": 6456 }, { "epoch": 0.78, "learning_rate": 1.3476958188811268e-06, "loss": 3.5353, "step": 6457 }, { "epoch": 0.78, "learning_rate": 1.3463261951140195e-06, "loss": 3.3842, "step": 6458 }, { "epoch": 0.78, "learning_rate": 1.3449571593709265e-06, "loss": 3.3602, "step": 6459 }, { "epoch": 0.78, "learning_rate": 1.3435887118721824e-06, "loss": 3.4817, "step": 6460 }, { "epoch": 0.78, "learning_rate": 1.3422208528380255e-06, "loss": 3.3644, "step": 6461 }, { "epoch": 0.78, "learning_rate": 1.3408535824886005e-06, "loss": 3.3592, "step": 6462 }, { "epoch": 0.78, "learning_rate": 1.3394869010439537e-06, "loss": 3.4067, "step": 6463 }, { "epoch": 0.78, "learning_rate": 1.3381208087240417e-06, "loss": 3.4009, "step": 6464 }, { "epoch": 0.78, "learning_rate": 1.3367553057487238e-06, "loss": 3.3627, "step": 6465 }, { "epoch": 0.78, "learning_rate": 1.3353903923377637e-06, "loss": 3.4185, "step": 6466 }, { "epoch": 0.78, "learning_rate": 1.3340260687108326e-06, "loss": 3.4099, "step": 6467 }, { "epoch": 0.78, "learning_rate": 1.3326623350875045e-06, "loss": 3.306, "step": 6468 }, { "epoch": 0.78, "learning_rate": 1.3312991916872597e-06, "loss": 3.412, "step": 6469 }, { "epoch": 0.78, "learning_rate": 1.3299366387294837e-06, "loss": 3.3548, "step": 6470 }, { "epoch": 0.78, "learning_rate": 1.328574676433465e-06, "loss": 3.3983, "step": 6471 }, { "epoch": 0.78, "learning_rate": 1.3272133050184e-06, "loss": 3.4428, "step": 6472 }, { "epoch": 0.78, "learning_rate": 1.3258525247033872e-06, "loss": 3.5124, "step": 6473 }, { "epoch": 0.78, "learning_rate": 1.3244923357074318e-06, "loss": 3.3132, "step": 6474 }, { "epoch": 0.78, "learning_rate": 1.3231327382494436e-06, "loss": 3.3209, "step": 6475 }, { "epoch": 0.78, "learning_rate": 1.3217737325482366e-06, "loss": 3.4418, "step": 6476 }, { "epoch": 0.78, "learning_rate": 1.3204153188225293e-06, "loss": 3.3863, "step": 6477 }, { "epoch": 0.78, "learning_rate": 1.3190574972909464e-06, "loss": 3.3593, "step": 6478 }, { "epoch": 0.78, "learning_rate": 1.3177002681720159e-06, "loss": 3.4004, "step": 6479 }, { "epoch": 0.78, "learning_rate": 1.3163436316841722e-06, "loss": 3.4241, "step": 6480 }, { "epoch": 0.78, "learning_rate": 1.3149875880457497e-06, "loss": 3.3125, "step": 6481 }, { "epoch": 0.78, "learning_rate": 1.3136321374749932e-06, "loss": 3.4108, "step": 6482 }, { "epoch": 0.78, "learning_rate": 1.312277280190049e-06, "loss": 3.349, "step": 6483 }, { "epoch": 0.78, "learning_rate": 1.3109230164089688e-06, "loss": 3.3853, "step": 6484 }, { "epoch": 0.78, "learning_rate": 1.3095693463497082e-06, "loss": 3.3768, "step": 6485 }, { "epoch": 0.78, "learning_rate": 1.3082162702301276e-06, "loss": 3.4881, "step": 6486 }, { "epoch": 0.78, "learning_rate": 1.3068637882679913e-06, "loss": 3.5764, "step": 6487 }, { "epoch": 0.78, "learning_rate": 1.3055119006809692e-06, "loss": 3.3597, "step": 6488 }, { "epoch": 0.78, "learning_rate": 1.3041606076866353e-06, "loss": 3.3043, "step": 6489 }, { "epoch": 0.78, "learning_rate": 1.302809909502465e-06, "loss": 3.3994, "step": 6490 }, { "epoch": 0.78, "learning_rate": 1.3014598063458417e-06, "loss": 3.4851, "step": 6491 }, { "epoch": 0.78, "learning_rate": 1.3001102984340514e-06, "loss": 3.3133, "step": 6492 }, { "epoch": 0.78, "learning_rate": 1.298761385984285e-06, "loss": 3.3001, "step": 6493 }, { "epoch": 0.78, "learning_rate": 1.297413069213637e-06, "loss": 3.4815, "step": 6494 }, { "epoch": 0.78, "learning_rate": 1.2960653483391057e-06, "loss": 3.4414, "step": 6495 }, { "epoch": 0.78, "learning_rate": 1.2947182235775945e-06, "loss": 3.3607, "step": 6496 }, { "epoch": 0.78, "learning_rate": 1.2933716951459096e-06, "loss": 3.4074, "step": 6497 }, { "epoch": 0.78, "learning_rate": 1.2920257632607646e-06, "loss": 3.3566, "step": 6498 }, { "epoch": 0.78, "learning_rate": 1.2906804281387702e-06, "loss": 3.2877, "step": 6499 }, { "epoch": 0.78, "learning_rate": 1.2893356899964477e-06, "loss": 3.4757, "step": 6500 }, { "epoch": 0.78, "learning_rate": 1.2879915490502192e-06, "loss": 3.3048, "step": 6501 }, { "epoch": 0.78, "learning_rate": 1.2866480055164115e-06, "loss": 3.4842, "step": 6502 }, { "epoch": 0.78, "learning_rate": 1.2853050596112554e-06, "loss": 3.3481, "step": 6503 }, { "epoch": 0.78, "learning_rate": 1.2839627115508857e-06, "loss": 3.3837, "step": 6504 }, { "epoch": 0.78, "learning_rate": 1.2826209615513396e-06, "loss": 3.3159, "step": 6505 }, { "epoch": 0.78, "learning_rate": 1.2812798098285594e-06, "loss": 3.3677, "step": 6506 }, { "epoch": 0.78, "learning_rate": 1.279939256598391e-06, "loss": 3.3833, "step": 6507 }, { "epoch": 0.78, "learning_rate": 1.278599302076583e-06, "loss": 3.3582, "step": 6508 }, { "epoch": 0.78, "learning_rate": 1.2772599464787888e-06, "loss": 3.4444, "step": 6509 }, { "epoch": 0.78, "learning_rate": 1.2759211900205653e-06, "loss": 3.4085, "step": 6510 }, { "epoch": 0.78, "learning_rate": 1.274583032917372e-06, "loss": 3.3993, "step": 6511 }, { "epoch": 0.78, "learning_rate": 1.2732454753845725e-06, "loss": 3.4233, "step": 6512 }, { "epoch": 0.78, "learning_rate": 1.2719085176374341e-06, "loss": 3.374, "step": 6513 }, { "epoch": 0.78, "learning_rate": 1.2705721598911274e-06, "loss": 3.4404, "step": 6514 }, { "epoch": 0.78, "learning_rate": 1.2692364023607267e-06, "loss": 3.3627, "step": 6515 }, { "epoch": 0.78, "learning_rate": 1.2679012452612111e-06, "loss": 3.2919, "step": 6516 }, { "epoch": 0.78, "learning_rate": 1.2665666888074574e-06, "loss": 3.2932, "step": 6517 }, { "epoch": 0.78, "learning_rate": 1.2652327332142528e-06, "loss": 3.2377, "step": 6518 }, { "epoch": 0.78, "learning_rate": 1.2638993786962833e-06, "loss": 3.4047, "step": 6519 }, { "epoch": 0.78, "learning_rate": 1.2625666254681407e-06, "loss": 3.3388, "step": 6520 }, { "epoch": 0.78, "learning_rate": 1.2612344737443182e-06, "loss": 3.3626, "step": 6521 }, { "epoch": 0.78, "learning_rate": 1.259902923739213e-06, "loss": 3.3891, "step": 6522 }, { "epoch": 0.78, "learning_rate": 1.258571975667126e-06, "loss": 3.3556, "step": 6523 }, { "epoch": 0.78, "learning_rate": 1.2572416297422602e-06, "loss": 3.3755, "step": 6524 }, { "epoch": 0.78, "learning_rate": 1.255911886178724e-06, "loss": 3.3739, "step": 6525 }, { "epoch": 0.78, "learning_rate": 1.254582745190523e-06, "loss": 3.4157, "step": 6526 }, { "epoch": 0.78, "learning_rate": 1.2532542069915722e-06, "loss": 3.4644, "step": 6527 }, { "epoch": 0.78, "learning_rate": 1.2519262717956866e-06, "loss": 3.4002, "step": 6528 }, { "epoch": 0.78, "learning_rate": 1.2505989398165853e-06, "loss": 3.2705, "step": 6529 }, { "epoch": 0.78, "learning_rate": 1.2492722112678896e-06, "loss": 3.3728, "step": 6530 }, { "epoch": 0.78, "learning_rate": 1.2479460863631233e-06, "loss": 3.3517, "step": 6531 }, { "epoch": 0.78, "learning_rate": 1.2466205653157144e-06, "loss": 3.3671, "step": 6532 }, { "epoch": 0.78, "learning_rate": 1.2452956483389922e-06, "loss": 3.3838, "step": 6533 }, { "epoch": 0.78, "learning_rate": 1.2439713356461908e-06, "loss": 3.3463, "step": 6534 }, { "epoch": 0.78, "learning_rate": 1.242647627450444e-06, "loss": 3.3994, "step": 6535 }, { "epoch": 0.78, "learning_rate": 1.2413245239647902e-06, "loss": 3.2853, "step": 6536 }, { "epoch": 0.78, "learning_rate": 1.2400020254021704e-06, "loss": 3.4287, "step": 6537 }, { "epoch": 0.78, "learning_rate": 1.2386801319754276e-06, "loss": 3.4244, "step": 6538 }, { "epoch": 0.78, "learning_rate": 1.2373588438973106e-06, "loss": 3.4558, "step": 6539 }, { "epoch": 0.79, "learning_rate": 1.2360381613804662e-06, "loss": 3.3645, "step": 6540 }, { "epoch": 0.79, "learning_rate": 1.2347180846374456e-06, "loss": 3.394, "step": 6541 }, { "epoch": 0.79, "learning_rate": 1.2333986138807036e-06, "loss": 3.3572, "step": 6542 }, { "epoch": 0.79, "learning_rate": 1.2320797493225967e-06, "loss": 3.3965, "step": 6543 }, { "epoch": 0.79, "learning_rate": 1.2307614911753812e-06, "loss": 3.477, "step": 6544 }, { "epoch": 0.79, "learning_rate": 1.2294438396512198e-06, "loss": 3.4328, "step": 6545 }, { "epoch": 0.79, "learning_rate": 1.2281267949621762e-06, "loss": 3.3658, "step": 6546 }, { "epoch": 0.79, "learning_rate": 1.2268103573202151e-06, "loss": 3.386, "step": 6547 }, { "epoch": 0.79, "learning_rate": 1.225494526937206e-06, "loss": 3.4186, "step": 6548 }, { "epoch": 0.79, "learning_rate": 1.2241793040249184e-06, "loss": 3.4099, "step": 6549 }, { "epoch": 0.79, "learning_rate": 1.2228646887950247e-06, "loss": 3.3849, "step": 6550 }, { "epoch": 0.79, "learning_rate": 1.2215506814591006e-06, "loss": 3.3784, "step": 6551 }, { "epoch": 0.79, "learning_rate": 1.220237282228624e-06, "loss": 3.2998, "step": 6552 }, { "epoch": 0.79, "learning_rate": 1.2189244913149706e-06, "loss": 3.373, "step": 6553 }, { "epoch": 0.79, "learning_rate": 1.2176123089294244e-06, "loss": 3.4548, "step": 6554 }, { "epoch": 0.79, "learning_rate": 1.2163007352831675e-06, "loss": 3.3498, "step": 6555 }, { "epoch": 0.79, "learning_rate": 1.2149897705872854e-06, "loss": 3.3419, "step": 6556 }, { "epoch": 0.79, "learning_rate": 1.2136794150527658e-06, "loss": 3.4346, "step": 6557 }, { "epoch": 0.79, "learning_rate": 1.2123696688904973e-06, "loss": 3.3943, "step": 6558 }, { "epoch": 0.79, "learning_rate": 1.2110605323112718e-06, "loss": 3.4532, "step": 6559 }, { "epoch": 0.79, "learning_rate": 1.2097520055257823e-06, "loss": 3.4109, "step": 6560 }, { "epoch": 0.79, "learning_rate": 1.2084440887446241e-06, "loss": 3.4026, "step": 6561 }, { "epoch": 0.79, "learning_rate": 1.2071367821782925e-06, "loss": 3.2865, "step": 6562 }, { "epoch": 0.79, "learning_rate": 1.205830086037187e-06, "loss": 3.3295, "step": 6563 }, { "epoch": 0.79, "learning_rate": 1.2045240005316073e-06, "loss": 3.4085, "step": 6564 }, { "epoch": 0.79, "learning_rate": 1.203218525871756e-06, "loss": 3.4317, "step": 6565 }, { "epoch": 0.79, "learning_rate": 1.2019136622677368e-06, "loss": 3.4444, "step": 6566 }, { "epoch": 0.79, "learning_rate": 1.2006094099295546e-06, "loss": 3.2631, "step": 6567 }, { "epoch": 0.79, "learning_rate": 1.1993057690671174e-06, "loss": 3.2586, "step": 6568 }, { "epoch": 0.79, "learning_rate": 1.1980027398902328e-06, "loss": 3.4408, "step": 6569 }, { "epoch": 0.79, "learning_rate": 1.1967003226086116e-06, "loss": 3.3471, "step": 6570 }, { "epoch": 0.79, "learning_rate": 1.1953985174318649e-06, "loss": 3.3441, "step": 6571 }, { "epoch": 0.79, "learning_rate": 1.194097324569506e-06, "loss": 3.3836, "step": 6572 }, { "epoch": 0.79, "learning_rate": 1.1927967442309495e-06, "loss": 3.3762, "step": 6573 }, { "epoch": 0.79, "learning_rate": 1.191496776625512e-06, "loss": 3.4279, "step": 6574 }, { "epoch": 0.79, "learning_rate": 1.1901974219624102e-06, "loss": 3.378, "step": 6575 }, { "epoch": 0.79, "learning_rate": 1.1888986804507634e-06, "loss": 3.43, "step": 6576 }, { "epoch": 0.79, "learning_rate": 1.1876005522995915e-06, "loss": 3.4734, "step": 6577 }, { "epoch": 0.79, "learning_rate": 1.1863030377178152e-06, "loss": 3.3382, "step": 6578 }, { "epoch": 0.79, "learning_rate": 1.1850061369142595e-06, "loss": 3.4296, "step": 6579 }, { "epoch": 0.79, "learning_rate": 1.1837098500976453e-06, "loss": 3.3466, "step": 6580 }, { "epoch": 0.79, "learning_rate": 1.1824141774765979e-06, "loss": 3.4827, "step": 6581 }, { "epoch": 0.79, "learning_rate": 1.181119119259645e-06, "loss": 3.4197, "step": 6582 }, { "epoch": 0.79, "learning_rate": 1.1798246756552128e-06, "loss": 3.4343, "step": 6583 }, { "epoch": 0.79, "learning_rate": 1.1785308468716305e-06, "loss": 3.4386, "step": 6584 }, { "epoch": 0.79, "learning_rate": 1.1772376331171264e-06, "loss": 3.4241, "step": 6585 }, { "epoch": 0.79, "learning_rate": 1.1759450345998324e-06, "loss": 3.3187, "step": 6586 }, { "epoch": 0.79, "learning_rate": 1.1746530515277788e-06, "loss": 3.3475, "step": 6587 }, { "epoch": 0.79, "learning_rate": 1.1733616841089003e-06, "loss": 3.3512, "step": 6588 }, { "epoch": 0.79, "learning_rate": 1.172070932551026e-06, "loss": 3.3971, "step": 6589 }, { "epoch": 0.79, "learning_rate": 1.170780797061893e-06, "loss": 3.4238, "step": 6590 }, { "epoch": 0.79, "learning_rate": 1.1694912778491353e-06, "loss": 3.4835, "step": 6591 }, { "epoch": 0.79, "learning_rate": 1.1682023751202892e-06, "loss": 3.4199, "step": 6592 }, { "epoch": 0.79, "learning_rate": 1.1669140890827913e-06, "loss": 3.3493, "step": 6593 }, { "epoch": 0.79, "learning_rate": 1.1656264199439787e-06, "loss": 3.4846, "step": 6594 }, { "epoch": 0.79, "learning_rate": 1.1643393679110898e-06, "loss": 3.4376, "step": 6595 }, { "epoch": 0.79, "learning_rate": 1.1630529331912632e-06, "loss": 3.4806, "step": 6596 }, { "epoch": 0.79, "learning_rate": 1.1617671159915394e-06, "loss": 3.4006, "step": 6597 }, { "epoch": 0.79, "learning_rate": 1.1604819165188568e-06, "loss": 3.3721, "step": 6598 }, { "epoch": 0.79, "learning_rate": 1.1591973349800562e-06, "loss": 3.3877, "step": 6599 }, { "epoch": 0.79, "learning_rate": 1.157913371581879e-06, "loss": 3.3411, "step": 6600 }, { "epoch": 0.79, "learning_rate": 1.1566300265309672e-06, "loss": 3.4355, "step": 6601 }, { "epoch": 0.79, "learning_rate": 1.1553473000338627e-06, "loss": 3.4629, "step": 6602 }, { "epoch": 0.79, "learning_rate": 1.1540651922970085e-06, "loss": 3.3474, "step": 6603 }, { "epoch": 0.79, "learning_rate": 1.1527837035267474e-06, "loss": 3.3579, "step": 6604 }, { "epoch": 0.79, "learning_rate": 1.1515028339293227e-06, "loss": 3.3334, "step": 6605 }, { "epoch": 0.79, "learning_rate": 1.1502225837108788e-06, "loss": 3.4175, "step": 6606 }, { "epoch": 0.79, "learning_rate": 1.1489429530774588e-06, "loss": 3.4036, "step": 6607 }, { "epoch": 0.79, "learning_rate": 1.1476639422350073e-06, "loss": 3.3803, "step": 6608 }, { "epoch": 0.79, "learning_rate": 1.1463855513893695e-06, "loss": 3.4404, "step": 6609 }, { "epoch": 0.79, "learning_rate": 1.1451077807462895e-06, "loss": 3.3404, "step": 6610 }, { "epoch": 0.79, "learning_rate": 1.1438306305114134e-06, "loss": 3.4248, "step": 6611 }, { "epoch": 0.79, "learning_rate": 1.1425541008902852e-06, "loss": 3.2622, "step": 6612 }, { "epoch": 0.79, "learning_rate": 1.141278192088351e-06, "loss": 3.2946, "step": 6613 }, { "epoch": 0.79, "learning_rate": 1.1400029043109552e-06, "loss": 3.4639, "step": 6614 }, { "epoch": 0.79, "learning_rate": 1.1387282377633457e-06, "loss": 3.4214, "step": 6615 }, { "epoch": 0.79, "learning_rate": 1.1374541926506648e-06, "loss": 3.3394, "step": 6616 }, { "epoch": 0.79, "learning_rate": 1.1361807691779591e-06, "loss": 3.4108, "step": 6617 }, { "epoch": 0.79, "learning_rate": 1.1349079675501739e-06, "loss": 3.3339, "step": 6618 }, { "epoch": 0.79, "learning_rate": 1.1336357879721555e-06, "loss": 3.332, "step": 6619 }, { "epoch": 0.79, "learning_rate": 1.1323642306486477e-06, "loss": 3.4132, "step": 6620 }, { "epoch": 0.79, "learning_rate": 1.1310932957842964e-06, "loss": 3.413, "step": 6621 }, { "epoch": 0.79, "learning_rate": 1.1298229835836462e-06, "loss": 3.4325, "step": 6622 }, { "epoch": 0.79, "learning_rate": 1.1285532942511413e-06, "loss": 3.3186, "step": 6623 }, { "epoch": 0.8, "learning_rate": 1.127284227991129e-06, "loss": 3.3592, "step": 6624 }, { "epoch": 0.8, "learning_rate": 1.1260157850078491e-06, "loss": 3.3786, "step": 6625 }, { "epoch": 0.8, "learning_rate": 1.1247479655054473e-06, "loss": 3.4532, "step": 6626 }, { "epoch": 0.8, "learning_rate": 1.1234807696879669e-06, "loss": 3.3792, "step": 6627 }, { "epoch": 0.8, "learning_rate": 1.1222141977593515e-06, "loss": 3.3566, "step": 6628 }, { "epoch": 0.8, "learning_rate": 1.1209482499234437e-06, "loss": 3.4384, "step": 6629 }, { "epoch": 0.8, "learning_rate": 1.1196829263839854e-06, "loss": 3.3202, "step": 6630 }, { "epoch": 0.8, "learning_rate": 1.1184182273446186e-06, "loss": 3.3533, "step": 6631 }, { "epoch": 0.8, "learning_rate": 1.1171541530088848e-06, "loss": 3.3272, "step": 6632 }, { "epoch": 0.8, "learning_rate": 1.1158907035802259e-06, "loss": 3.4481, "step": 6633 }, { "epoch": 0.8, "learning_rate": 1.1146278792619785e-06, "loss": 3.3722, "step": 6634 }, { "epoch": 0.8, "learning_rate": 1.113365680257385e-06, "loss": 3.4546, "step": 6635 }, { "epoch": 0.8, "learning_rate": 1.112104106769582e-06, "loss": 3.4309, "step": 6636 }, { "epoch": 0.8, "learning_rate": 1.1108431590016106e-06, "loss": 3.3944, "step": 6637 }, { "epoch": 0.8, "learning_rate": 1.1095828371564072e-06, "loss": 3.3877, "step": 6638 }, { "epoch": 0.8, "learning_rate": 1.1083231414368084e-06, "loss": 3.3678, "step": 6639 }, { "epoch": 0.8, "learning_rate": 1.1070640720455505e-06, "loss": 3.4287, "step": 6640 }, { "epoch": 0.8, "learning_rate": 1.1058056291852682e-06, "loss": 3.4044, "step": 6641 }, { "epoch": 0.8, "learning_rate": 1.104547813058498e-06, "loss": 3.3558, "step": 6642 }, { "epoch": 0.8, "learning_rate": 1.10329062386767e-06, "loss": 3.3977, "step": 6643 }, { "epoch": 0.8, "learning_rate": 1.1020340618151182e-06, "loss": 3.2537, "step": 6644 }, { "epoch": 0.8, "learning_rate": 1.1007781271030755e-06, "loss": 3.3907, "step": 6645 }, { "epoch": 0.8, "learning_rate": 1.099522819933671e-06, "loss": 3.4238, "step": 6646 }, { "epoch": 0.8, "learning_rate": 1.098268140508935e-06, "loss": 3.3817, "step": 6647 }, { "epoch": 0.8, "learning_rate": 1.0970140890307973e-06, "loss": 3.3521, "step": 6648 }, { "epoch": 0.8, "learning_rate": 1.0957606657010838e-06, "loss": 3.4052, "step": 6649 }, { "epoch": 0.8, "learning_rate": 1.0945078707215224e-06, "loss": 3.3589, "step": 6650 }, { "epoch": 0.8, "learning_rate": 1.0932557042937392e-06, "loss": 3.4162, "step": 6651 }, { "epoch": 0.8, "learning_rate": 1.0920041666192554e-06, "loss": 3.2855, "step": 6652 }, { "epoch": 0.8, "learning_rate": 1.0907532578994966e-06, "loss": 3.2779, "step": 6653 }, { "epoch": 0.8, "learning_rate": 1.089502978335784e-06, "loss": 3.4227, "step": 6654 }, { "epoch": 0.8, "learning_rate": 1.0882533281293378e-06, "loss": 3.4485, "step": 6655 }, { "epoch": 0.8, "learning_rate": 1.087004307481278e-06, "loss": 3.4194, "step": 6656 }, { "epoch": 0.8, "learning_rate": 1.0857559165926224e-06, "loss": 3.3973, "step": 6657 }, { "epoch": 0.8, "learning_rate": 1.0845081556642866e-06, "loss": 3.4349, "step": 6658 }, { "epoch": 0.8, "learning_rate": 1.0832610248970877e-06, "loss": 3.3984, "step": 6659 }, { "epoch": 0.8, "learning_rate": 1.0820145244917396e-06, "loss": 3.3982, "step": 6660 }, { "epoch": 0.8, "learning_rate": 1.0807686546488521e-06, "loss": 3.3956, "step": 6661 }, { "epoch": 0.8, "learning_rate": 1.0795234155689382e-06, "loss": 3.3938, "step": 6662 }, { "epoch": 0.8, "learning_rate": 1.0782788074524064e-06, "loss": 3.4519, "step": 6663 }, { "epoch": 0.8, "learning_rate": 1.0770348304995647e-06, "loss": 3.379, "step": 6664 }, { "epoch": 0.8, "learning_rate": 1.0757914849106195e-06, "loss": 3.4133, "step": 6665 }, { "epoch": 0.8, "learning_rate": 1.0745487708856756e-06, "loss": 3.294, "step": 6666 }, { "epoch": 0.8, "learning_rate": 1.0733066886247356e-06, "loss": 3.3628, "step": 6667 }, { "epoch": 0.8, "learning_rate": 1.0720652383277007e-06, "loss": 3.3343, "step": 6668 }, { "epoch": 0.8, "learning_rate": 1.0708244201943712e-06, "loss": 3.3412, "step": 6669 }, { "epoch": 0.8, "learning_rate": 1.0695842344244451e-06, "loss": 3.3161, "step": 6670 }, { "epoch": 0.8, "learning_rate": 1.0683446812175174e-06, "loss": 3.4465, "step": 6671 }, { "epoch": 0.8, "learning_rate": 1.0671057607730828e-06, "loss": 3.4175, "step": 6672 }, { "epoch": 0.8, "learning_rate": 1.0658674732905345e-06, "loss": 3.3629, "step": 6673 }, { "epoch": 0.8, "learning_rate": 1.0646298189691623e-06, "loss": 3.3591, "step": 6674 }, { "epoch": 0.8, "learning_rate": 1.0633927980081548e-06, "loss": 3.3595, "step": 6675 }, { "epoch": 0.8, "learning_rate": 1.062156410606599e-06, "loss": 3.4712, "step": 6676 }, { "epoch": 0.8, "learning_rate": 1.06092065696348e-06, "loss": 3.3542, "step": 6677 }, { "epoch": 0.8, "learning_rate": 1.0596855372776821e-06, "loss": 3.3296, "step": 6678 }, { "epoch": 0.8, "learning_rate": 1.0584510517479823e-06, "loss": 3.5582, "step": 6679 }, { "epoch": 0.8, "learning_rate": 1.0572172005730614e-06, "loss": 3.3604, "step": 6680 }, { "epoch": 0.8, "learning_rate": 1.0559839839514958e-06, "loss": 3.3547, "step": 6681 }, { "epoch": 0.8, "learning_rate": 1.05475140208176e-06, "loss": 3.3351, "step": 6682 }, { "epoch": 0.8, "learning_rate": 1.0535194551622264e-06, "loss": 3.3248, "step": 6683 }, { "epoch": 0.8, "learning_rate": 1.0522881433911646e-06, "loss": 3.4284, "step": 6684 }, { "epoch": 0.8, "learning_rate": 1.051057466966744e-06, "loss": 3.4004, "step": 6685 }, { "epoch": 0.8, "learning_rate": 1.0498274260870285e-06, "loss": 3.2826, "step": 6686 }, { "epoch": 0.8, "learning_rate": 1.0485980209499835e-06, "loss": 3.4032, "step": 6687 }, { "epoch": 0.8, "learning_rate": 1.0473692517534679e-06, "loss": 3.4477, "step": 6688 }, { "epoch": 0.8, "learning_rate": 1.0461411186952407e-06, "loss": 3.4239, "step": 6689 }, { "epoch": 0.8, "learning_rate": 1.044913621972959e-06, "loss": 3.2928, "step": 6690 }, { "epoch": 0.8, "learning_rate": 1.0436867617841768e-06, "loss": 3.2716, "step": 6691 }, { "epoch": 0.8, "learning_rate": 1.0424605383263448e-06, "loss": 3.4323, "step": 6692 }, { "epoch": 0.8, "learning_rate": 1.041234951796813e-06, "loss": 3.2606, "step": 6693 }, { "epoch": 0.8, "learning_rate": 1.0400100023928266e-06, "loss": 3.3976, "step": 6694 }, { "epoch": 0.8, "learning_rate": 1.0387856903115305e-06, "loss": 3.4047, "step": 6695 }, { "epoch": 0.8, "learning_rate": 1.0375620157499678e-06, "loss": 3.3045, "step": 6696 }, { "epoch": 0.8, "learning_rate": 1.0363389789050732e-06, "loss": 3.3463, "step": 6697 }, { "epoch": 0.8, "learning_rate": 1.0351165799736857e-06, "loss": 3.3406, "step": 6698 }, { "epoch": 0.8, "learning_rate": 1.0338948191525372e-06, "loss": 3.4625, "step": 6699 }, { "epoch": 0.8, "learning_rate": 1.03267369663826e-06, "loss": 3.4724, "step": 6700 }, { "epoch": 0.8, "learning_rate": 1.0314532126273813e-06, "loss": 3.3389, "step": 6701 }, { "epoch": 0.8, "learning_rate": 1.0302333673163263e-06, "loss": 3.36, "step": 6702 }, { "epoch": 0.8, "learning_rate": 1.0290141609014177e-06, "loss": 3.3402, "step": 6703 }, { "epoch": 0.8, "learning_rate": 1.0277955935788752e-06, "loss": 3.4686, "step": 6704 }, { "epoch": 0.8, "learning_rate": 1.0265776655448156e-06, "loss": 3.2384, "step": 6705 }, { "epoch": 0.8, "learning_rate": 1.0253603769952531e-06, "loss": 3.4542, "step": 6706 }, { "epoch": 0.81, "learning_rate": 1.0241437281260985e-06, "loss": 3.3215, "step": 6707 }, { "epoch": 0.81, "learning_rate": 1.0229277191331593e-06, "loss": 3.3595, "step": 6708 }, { "epoch": 0.81, "learning_rate": 1.0217123502121413e-06, "loss": 3.4287, "step": 6709 }, { "epoch": 0.81, "learning_rate": 1.0204976215586466e-06, "loss": 3.4215, "step": 6710 }, { "epoch": 0.81, "learning_rate": 1.0192835333681733e-06, "loss": 3.4849, "step": 6711 }, { "epoch": 0.81, "learning_rate": 1.0180700858361181e-06, "loss": 3.4175, "step": 6712 }, { "epoch": 0.81, "learning_rate": 1.0168572791577742e-06, "loss": 3.4508, "step": 6713 }, { "epoch": 0.81, "learning_rate": 1.0156451135283312e-06, "loss": 3.3466, "step": 6714 }, { "epoch": 0.81, "learning_rate": 1.0144335891428747e-06, "loss": 3.4071, "step": 6715 }, { "epoch": 0.81, "learning_rate": 1.0132227061963879e-06, "loss": 3.442, "step": 6716 }, { "epoch": 0.81, "learning_rate": 1.012012464883752e-06, "loss": 3.4317, "step": 6717 }, { "epoch": 0.81, "learning_rate": 1.010802865399743e-06, "loss": 3.3948, "step": 6718 }, { "epoch": 0.81, "learning_rate": 1.009593907939035e-06, "loss": 3.4032, "step": 6719 }, { "epoch": 0.81, "learning_rate": 1.0083855926961971e-06, "loss": 3.4119, "step": 6720 }, { "epoch": 0.81, "learning_rate": 1.0071779198656979e-06, "loss": 3.3474, "step": 6721 }, { "epoch": 0.81, "learning_rate": 1.0059708896418995e-06, "loss": 3.3158, "step": 6722 }, { "epoch": 0.81, "learning_rate": 1.0047645022190633e-06, "loss": 3.4179, "step": 6723 }, { "epoch": 0.81, "learning_rate": 1.0035587577913441e-06, "loss": 3.3893, "step": 6724 }, { "epoch": 0.81, "learning_rate": 1.0023536565527959e-06, "loss": 3.3748, "step": 6725 }, { "epoch": 0.81, "learning_rate": 1.0011491986973675e-06, "loss": 3.3127, "step": 6726 }, { "epoch": 0.81, "learning_rate": 9.999453844189061e-07, "loss": 3.35, "step": 6727 }, { "epoch": 0.81, "learning_rate": 9.98742213911153e-07, "loss": 3.3301, "step": 6728 }, { "epoch": 0.81, "learning_rate": 9.975396873677485e-07, "loss": 3.2769, "step": 6729 }, { "epoch": 0.81, "learning_rate": 9.963378049822264e-07, "loss": 3.4134, "step": 6730 }, { "epoch": 0.81, "learning_rate": 9.95136566948019e-07, "loss": 3.4232, "step": 6731 }, { "epoch": 0.81, "learning_rate": 9.939359734584552e-07, "loss": 3.3153, "step": 6732 }, { "epoch": 0.81, "learning_rate": 9.927360247067546e-07, "loss": 3.363, "step": 6733 }, { "epoch": 0.81, "learning_rate": 9.915367208860428e-07, "loss": 3.5502, "step": 6734 }, { "epoch": 0.81, "learning_rate": 9.903380621893338e-07, "loss": 3.4168, "step": 6735 }, { "epoch": 0.81, "learning_rate": 9.891400488095416e-07, "loss": 3.368, "step": 6736 }, { "epoch": 0.81, "learning_rate": 9.879426809394732e-07, "loss": 3.4193, "step": 6737 }, { "epoch": 0.81, "learning_rate": 9.867459587718347e-07, "loss": 3.4153, "step": 6738 }, { "epoch": 0.81, "learning_rate": 9.855498824992277e-07, "loss": 3.4662, "step": 6739 }, { "epoch": 0.81, "learning_rate": 9.84354452314148e-07, "loss": 3.4588, "step": 6740 }, { "epoch": 0.81, "learning_rate": 9.831596684089911e-07, "loss": 3.3598, "step": 6741 }, { "epoch": 0.81, "learning_rate": 9.819655309760424e-07, "loss": 3.3685, "step": 6742 }, { "epoch": 0.81, "learning_rate": 9.807720402074894e-07, "loss": 3.4008, "step": 6743 }, { "epoch": 0.81, "learning_rate": 9.79579196295412e-07, "loss": 3.4423, "step": 6744 }, { "epoch": 0.81, "learning_rate": 9.78386999431788e-07, "loss": 3.2848, "step": 6745 }, { "epoch": 0.81, "learning_rate": 9.771954498084896e-07, "loss": 3.3155, "step": 6746 }, { "epoch": 0.81, "learning_rate": 9.760045476172858e-07, "loss": 3.3098, "step": 6747 }, { "epoch": 0.81, "learning_rate": 9.748142930498405e-07, "loss": 3.399, "step": 6748 }, { "epoch": 0.81, "learning_rate": 9.73624686297714e-07, "loss": 3.3959, "step": 6749 }, { "epoch": 0.81, "learning_rate": 9.724357275523632e-07, "loss": 3.3385, "step": 6750 }, { "epoch": 0.81, "learning_rate": 9.712474170051378e-07, "loss": 3.3941, "step": 6751 }, { "epoch": 0.81, "learning_rate": 9.700597548472857e-07, "loss": 3.3036, "step": 6752 }, { "epoch": 0.81, "learning_rate": 9.6887274126995e-07, "loss": 3.316, "step": 6753 }, { "epoch": 0.81, "learning_rate": 9.676863764641698e-07, "loss": 3.4454, "step": 6754 }, { "epoch": 0.81, "learning_rate": 9.665006606208782e-07, "loss": 3.4188, "step": 6755 }, { "epoch": 0.81, "learning_rate": 9.653155939309055e-07, "loss": 3.3859, "step": 6756 }, { "epoch": 0.81, "learning_rate": 9.64131176584977e-07, "loss": 3.3569, "step": 6757 }, { "epoch": 0.81, "learning_rate": 9.629474087737133e-07, "loss": 3.3704, "step": 6758 }, { "epoch": 0.81, "learning_rate": 9.617642906876318e-07, "loss": 3.2776, "step": 6759 }, { "epoch": 0.81, "learning_rate": 9.605818225171414e-07, "loss": 3.3658, "step": 6760 }, { "epoch": 0.81, "learning_rate": 9.594000044525503e-07, "loss": 3.5006, "step": 6761 }, { "epoch": 0.81, "learning_rate": 9.582188366840612e-07, "loss": 3.2891, "step": 6762 }, { "epoch": 0.81, "learning_rate": 9.570383194017713e-07, "loss": 3.4114, "step": 6763 }, { "epoch": 0.81, "learning_rate": 9.558584527956742e-07, "loss": 3.3168, "step": 6764 }, { "epoch": 0.81, "learning_rate": 9.546792370556578e-07, "loss": 3.3245, "step": 6765 }, { "epoch": 0.81, "learning_rate": 9.535006723715056e-07, "loss": 3.4093, "step": 6766 }, { "epoch": 0.81, "learning_rate": 9.52322758932897e-07, "loss": 3.4017, "step": 6767 }, { "epoch": 0.81, "learning_rate": 9.511454969294054e-07, "loss": 3.3778, "step": 6768 }, { "epoch": 0.81, "learning_rate": 9.499688865505003e-07, "loss": 3.3968, "step": 6769 }, { "epoch": 0.81, "learning_rate": 9.487929279855452e-07, "loss": 3.3869, "step": 6770 }, { "epoch": 0.81, "learning_rate": 9.476176214237997e-07, "loss": 3.4109, "step": 6771 }, { "epoch": 0.81, "learning_rate": 9.46442967054419e-07, "loss": 3.3579, "step": 6772 }, { "epoch": 0.81, "learning_rate": 9.452689650664515e-07, "loss": 3.4066, "step": 6773 }, { "epoch": 0.81, "learning_rate": 9.440956156488423e-07, "loss": 3.3079, "step": 6774 }, { "epoch": 0.81, "learning_rate": 9.42922918990431e-07, "loss": 3.4698, "step": 6775 }, { "epoch": 0.81, "learning_rate": 9.417508752799509e-07, "loss": 3.3895, "step": 6776 }, { "epoch": 0.81, "learning_rate": 9.405794847060335e-07, "loss": 3.3657, "step": 6777 }, { "epoch": 0.81, "learning_rate": 9.394087474572e-07, "loss": 3.348, "step": 6778 }, { "epoch": 0.81, "learning_rate": 9.382386637218711e-07, "loss": 3.4685, "step": 6779 }, { "epoch": 0.81, "learning_rate": 9.370692336883596e-07, "loss": 3.4044, "step": 6780 }, { "epoch": 0.81, "learning_rate": 9.359004575448754e-07, "loss": 3.4763, "step": 6781 }, { "epoch": 0.81, "learning_rate": 9.347323354795218e-07, "loss": 3.3771, "step": 6782 }, { "epoch": 0.81, "learning_rate": 9.335648676802961e-07, "loss": 3.4062, "step": 6783 }, { "epoch": 0.81, "learning_rate": 9.323980543350913e-07, "loss": 3.3769, "step": 6784 }, { "epoch": 0.81, "learning_rate": 9.312318956316951e-07, "loss": 3.2786, "step": 6785 }, { "epoch": 0.81, "learning_rate": 9.300663917577918e-07, "loss": 3.3818, "step": 6786 }, { "epoch": 0.81, "learning_rate": 9.28901542900954e-07, "loss": 3.3589, "step": 6787 }, { "epoch": 0.81, "learning_rate": 9.277373492486552e-07, "loss": 3.3828, "step": 6788 }, { "epoch": 0.81, "learning_rate": 9.26573810988261e-07, "loss": 3.3625, "step": 6789 }, { "epoch": 0.82, "learning_rate": 9.254109283070323e-07, "loss": 3.3586, "step": 6790 }, { "epoch": 0.82, "learning_rate": 9.24248701392123e-07, "loss": 3.325, "step": 6791 }, { "epoch": 0.82, "learning_rate": 9.230871304305838e-07, "loss": 3.3271, "step": 6792 }, { "epoch": 0.82, "learning_rate": 9.219262156093572e-07, "loss": 3.4601, "step": 6793 }, { "epoch": 0.82, "learning_rate": 9.207659571152827e-07, "loss": 3.479, "step": 6794 }, { "epoch": 0.82, "learning_rate": 9.196063551350926e-07, "loss": 3.3389, "step": 6795 }, { "epoch": 0.82, "learning_rate": 9.184474098554124e-07, "loss": 3.3462, "step": 6796 }, { "epoch": 0.82, "learning_rate": 9.172891214627644e-07, "loss": 3.3086, "step": 6797 }, { "epoch": 0.82, "learning_rate": 9.161314901435641e-07, "loss": 3.3519, "step": 6798 }, { "epoch": 0.82, "learning_rate": 9.149745160841206e-07, "loss": 3.3, "step": 6799 }, { "epoch": 0.82, "learning_rate": 9.138181994706385e-07, "loss": 3.392, "step": 6800 }, { "epoch": 0.82, "learning_rate": 9.126625404892159e-07, "loss": 3.3127, "step": 6801 }, { "epoch": 0.82, "learning_rate": 9.115075393258449e-07, "loss": 3.4164, "step": 6802 }, { "epoch": 0.82, "learning_rate": 9.10353196166412e-07, "loss": 3.3475, "step": 6803 }, { "epoch": 0.82, "learning_rate": 9.091995111966978e-07, "loss": 3.3213, "step": 6804 }, { "epoch": 0.82, "learning_rate": 9.080464846023773e-07, "loss": 3.5145, "step": 6805 }, { "epoch": 0.82, "learning_rate": 9.068941165690182e-07, "loss": 3.414, "step": 6806 }, { "epoch": 0.82, "learning_rate": 9.057424072820842e-07, "loss": 3.4087, "step": 6807 }, { "epoch": 0.82, "learning_rate": 9.04591356926931e-07, "loss": 3.3258, "step": 6808 }, { "epoch": 0.82, "learning_rate": 9.034409656888099e-07, "loss": 3.3653, "step": 6809 }, { "epoch": 0.82, "learning_rate": 9.02291233752865e-07, "loss": 3.443, "step": 6810 }, { "epoch": 0.82, "learning_rate": 9.011421613041349e-07, "loss": 3.4256, "step": 6811 }, { "epoch": 0.82, "learning_rate": 8.999937485275518e-07, "loss": 3.4614, "step": 6812 }, { "epoch": 0.82, "learning_rate": 8.988459956079431e-07, "loss": 3.3805, "step": 6813 }, { "epoch": 0.82, "learning_rate": 8.976989027300265e-07, "loss": 3.3398, "step": 6814 }, { "epoch": 0.82, "learning_rate": 8.965524700784162e-07, "loss": 3.4451, "step": 6815 }, { "epoch": 0.82, "learning_rate": 8.954066978376203e-07, "loss": 3.3413, "step": 6816 }, { "epoch": 0.82, "learning_rate": 8.942615861920395e-07, "loss": 3.358, "step": 6817 }, { "epoch": 0.82, "learning_rate": 8.931171353259694e-07, "loss": 3.3577, "step": 6818 }, { "epoch": 0.82, "learning_rate": 8.919733454235973e-07, "loss": 3.265, "step": 6819 }, { "epoch": 0.82, "learning_rate": 8.908302166690058e-07, "loss": 3.3154, "step": 6820 }, { "epoch": 0.82, "learning_rate": 8.896877492461708e-07, "loss": 3.3372, "step": 6821 }, { "epoch": 0.82, "learning_rate": 8.885459433389632e-07, "loss": 3.4275, "step": 6822 }, { "epoch": 0.82, "learning_rate": 8.874047991311424e-07, "loss": 3.396, "step": 6823 }, { "epoch": 0.82, "learning_rate": 8.862643168063662e-07, "loss": 3.4321, "step": 6824 }, { "epoch": 0.82, "learning_rate": 8.851244965481842e-07, "loss": 3.4366, "step": 6825 }, { "epoch": 0.82, "learning_rate": 8.839853385400405e-07, "loss": 3.3832, "step": 6826 }, { "epoch": 0.82, "learning_rate": 8.828468429652709e-07, "loss": 3.4439, "step": 6827 }, { "epoch": 0.82, "learning_rate": 8.817090100071057e-07, "loss": 3.4085, "step": 6828 }, { "epoch": 0.82, "learning_rate": 8.805718398486679e-07, "loss": 3.5331, "step": 6829 }, { "epoch": 0.82, "learning_rate": 8.794353326729748e-07, "loss": 3.3925, "step": 6830 }, { "epoch": 0.82, "learning_rate": 8.782994886629364e-07, "loss": 3.3624, "step": 6831 }, { "epoch": 0.82, "learning_rate": 8.771643080013553e-07, "loss": 3.4181, "step": 6832 }, { "epoch": 0.82, "learning_rate": 8.760297908709286e-07, "loss": 3.4255, "step": 6833 }, { "epoch": 0.82, "learning_rate": 8.748959374542459e-07, "loss": 3.4495, "step": 6834 }, { "epoch": 0.82, "learning_rate": 8.737627479337902e-07, "loss": 3.3823, "step": 6835 }, { "epoch": 0.82, "learning_rate": 8.726302224919375e-07, "loss": 3.3409, "step": 6836 }, { "epoch": 0.82, "learning_rate": 8.714983613109567e-07, "loss": 3.4176, "step": 6837 }, { "epoch": 0.82, "learning_rate": 8.703671645730111e-07, "loss": 3.39, "step": 6838 }, { "epoch": 0.82, "learning_rate": 8.692366324601548e-07, "loss": 3.3559, "step": 6839 }, { "epoch": 0.82, "learning_rate": 8.681067651543373e-07, "loss": 3.4115, "step": 6840 }, { "epoch": 0.82, "learning_rate": 8.669775628373988e-07, "loss": 3.4159, "step": 6841 }, { "epoch": 0.82, "learning_rate": 8.65849025691074e-07, "loss": 3.3947, "step": 6842 }, { "epoch": 0.82, "learning_rate": 8.647211538969908e-07, "loss": 3.3072, "step": 6843 }, { "epoch": 0.82, "learning_rate": 8.63593947636669e-07, "loss": 3.4529, "step": 6844 }, { "epoch": 0.82, "learning_rate": 8.624674070915212e-07, "loss": 3.421, "step": 6845 }, { "epoch": 0.82, "learning_rate": 8.613415324428542e-07, "loss": 3.3441, "step": 6846 }, { "epoch": 0.82, "learning_rate": 8.60216323871867e-07, "loss": 3.3451, "step": 6847 }, { "epoch": 0.82, "learning_rate": 8.5909178155965e-07, "loss": 3.4868, "step": 6848 }, { "epoch": 0.82, "learning_rate": 8.579679056871898e-07, "loss": 3.2657, "step": 6849 }, { "epoch": 0.82, "learning_rate": 8.568446964353611e-07, "loss": 3.3663, "step": 6850 }, { "epoch": 0.82, "learning_rate": 8.557221539849347e-07, "loss": 3.5005, "step": 6851 }, { "epoch": 0.82, "learning_rate": 8.54600278516573e-07, "loss": 3.3358, "step": 6852 }, { "epoch": 0.82, "learning_rate": 8.534790702108314e-07, "loss": 3.3201, "step": 6853 }, { "epoch": 0.82, "learning_rate": 8.523585292481574e-07, "loss": 3.3936, "step": 6854 }, { "epoch": 0.82, "learning_rate": 8.512386558088919e-07, "loss": 3.3072, "step": 6855 }, { "epoch": 0.82, "learning_rate": 8.501194500732674e-07, "loss": 3.4046, "step": 6856 }, { "epoch": 0.82, "learning_rate": 8.490009122214093e-07, "loss": 3.3654, "step": 6857 }, { "epoch": 0.82, "learning_rate": 8.478830424333373e-07, "loss": 3.3579, "step": 6858 }, { "epoch": 0.82, "learning_rate": 8.467658408889595e-07, "loss": 3.3675, "step": 6859 }, { "epoch": 0.82, "learning_rate": 8.456493077680795e-07, "loss": 3.2988, "step": 6860 }, { "epoch": 0.82, "learning_rate": 8.445334432503932e-07, "loss": 3.4738, "step": 6861 }, { "epoch": 0.82, "learning_rate": 8.434182475154884e-07, "loss": 3.378, "step": 6862 }, { "epoch": 0.82, "learning_rate": 8.423037207428447e-07, "loss": 3.4463, "step": 6863 }, { "epoch": 0.82, "learning_rate": 8.411898631118348e-07, "loss": 3.3662, "step": 6864 }, { "epoch": 0.82, "learning_rate": 8.400766748017236e-07, "loss": 3.4302, "step": 6865 }, { "epoch": 0.82, "learning_rate": 8.389641559916684e-07, "loss": 3.3194, "step": 6866 }, { "epoch": 0.82, "learning_rate": 8.378523068607181e-07, "loss": 3.3963, "step": 6867 }, { "epoch": 0.82, "learning_rate": 8.367411275878146e-07, "loss": 3.4176, "step": 6868 }, { "epoch": 0.82, "learning_rate": 8.356306183517915e-07, "loss": 3.4266, "step": 6869 }, { "epoch": 0.82, "learning_rate": 8.345207793313742e-07, "loss": 3.4098, "step": 6870 }, { "epoch": 0.82, "learning_rate": 8.334116107051815e-07, "loss": 3.2661, "step": 6871 }, { "epoch": 0.82, "learning_rate": 8.323031126517228e-07, "loss": 3.3636, "step": 6872 }, { "epoch": 0.82, "learning_rate": 8.31195285349401e-07, "loss": 3.3609, "step": 6873 }, { "epoch": 0.83, "learning_rate": 8.300881289765095e-07, "loss": 3.3962, "step": 6874 }, { "epoch": 0.83, "learning_rate": 8.289816437112358e-07, "loss": 3.4118, "step": 6875 }, { "epoch": 0.83, "learning_rate": 8.27875829731658e-07, "loss": 3.4015, "step": 6876 }, { "epoch": 0.83, "learning_rate": 8.267706872157449e-07, "loss": 3.3455, "step": 6877 }, { "epoch": 0.83, "learning_rate": 8.256662163413598e-07, "loss": 3.4584, "step": 6878 }, { "epoch": 0.83, "learning_rate": 8.24562417286256e-07, "loss": 3.3445, "step": 6879 }, { "epoch": 0.83, "learning_rate": 8.234592902280808e-07, "loss": 3.3429, "step": 6880 }, { "epoch": 0.83, "learning_rate": 8.223568353443706e-07, "loss": 3.4396, "step": 6881 }, { "epoch": 0.83, "learning_rate": 8.212550528125557e-07, "loss": 3.3336, "step": 6882 }, { "epoch": 0.83, "learning_rate": 8.201539428099576e-07, "loss": 3.3298, "step": 6883 }, { "epoch": 0.83, "learning_rate": 8.190535055137888e-07, "loss": 3.2559, "step": 6884 }, { "epoch": 0.83, "learning_rate": 8.179537411011562e-07, "loss": 3.2008, "step": 6885 }, { "epoch": 0.83, "learning_rate": 8.168546497490537e-07, "loss": 3.4194, "step": 6886 }, { "epoch": 0.83, "learning_rate": 8.157562316343709e-07, "loss": 3.3551, "step": 6887 }, { "epoch": 0.83, "learning_rate": 8.146584869338875e-07, "loss": 3.2913, "step": 6888 }, { "epoch": 0.83, "learning_rate": 8.135614158242755e-07, "loss": 3.3561, "step": 6889 }, { "epoch": 0.83, "learning_rate": 8.124650184820971e-07, "loss": 3.3987, "step": 6890 }, { "epoch": 0.83, "learning_rate": 8.113692950838082e-07, "loss": 3.3719, "step": 6891 }, { "epoch": 0.83, "learning_rate": 8.102742458057544e-07, "loss": 3.4251, "step": 6892 }, { "epoch": 0.83, "learning_rate": 8.091798708241733e-07, "loss": 3.4545, "step": 6893 }, { "epoch": 0.83, "learning_rate": 8.08086170315196e-07, "loss": 3.3393, "step": 6894 }, { "epoch": 0.83, "learning_rate": 8.069931444548401e-07, "loss": 3.3135, "step": 6895 }, { "epoch": 0.83, "learning_rate": 8.059007934190194e-07, "loss": 3.4205, "step": 6896 }, { "epoch": 0.83, "learning_rate": 8.048091173835371e-07, "loss": 3.3316, "step": 6897 }, { "epoch": 0.83, "learning_rate": 8.037181165240882e-07, "loss": 3.3885, "step": 6898 }, { "epoch": 0.83, "learning_rate": 8.026277910162589e-07, "loss": 3.3509, "step": 6899 }, { "epoch": 0.83, "learning_rate": 8.015381410355261e-07, "loss": 3.4451, "step": 6900 }, { "epoch": 0.83, "learning_rate": 8.0044916675726e-07, "loss": 3.3355, "step": 6901 }, { "epoch": 0.83, "learning_rate": 7.993608683567194e-07, "loss": 3.3413, "step": 6902 }, { "epoch": 0.83, "learning_rate": 7.982732460090558e-07, "loss": 3.4718, "step": 6903 }, { "epoch": 0.83, "learning_rate": 7.971862998893121e-07, "loss": 3.3349, "step": 6904 }, { "epoch": 0.83, "learning_rate": 7.961000301724214e-07, "loss": 3.4315, "step": 6905 }, { "epoch": 0.83, "learning_rate": 7.950144370332092e-07, "loss": 3.3658, "step": 6906 }, { "epoch": 0.83, "learning_rate": 7.939295206463904e-07, "loss": 3.3689, "step": 6907 }, { "epoch": 0.83, "learning_rate": 7.928452811865728e-07, "loss": 3.3482, "step": 6908 }, { "epoch": 0.83, "learning_rate": 7.917617188282539e-07, "loss": 3.4142, "step": 6909 }, { "epoch": 0.83, "learning_rate": 7.906788337458232e-07, "loss": 3.3944, "step": 6910 }, { "epoch": 0.83, "learning_rate": 7.895966261135602e-07, "loss": 3.3527, "step": 6911 }, { "epoch": 0.83, "learning_rate": 7.885150961056376e-07, "loss": 3.3136, "step": 6912 }, { "epoch": 0.83, "learning_rate": 7.874342438961147e-07, "loss": 3.3961, "step": 6913 }, { "epoch": 0.83, "learning_rate": 7.86354069658945e-07, "loss": 3.3299, "step": 6914 }, { "epoch": 0.83, "learning_rate": 7.85274573567974e-07, "loss": 3.3348, "step": 6915 }, { "epoch": 0.83, "learning_rate": 7.841957557969343e-07, "loss": 3.4692, "step": 6916 }, { "epoch": 0.83, "learning_rate": 7.831176165194526e-07, "loss": 3.3498, "step": 6917 }, { "epoch": 0.83, "learning_rate": 7.820401559090451e-07, "loss": 3.4386, "step": 6918 }, { "epoch": 0.83, "learning_rate": 7.809633741391182e-07, "loss": 3.4464, "step": 6919 }, { "epoch": 0.83, "learning_rate": 7.798872713829703e-07, "loss": 3.4225, "step": 6920 }, { "epoch": 0.83, "learning_rate": 7.788118478137901e-07, "loss": 3.3665, "step": 6921 }, { "epoch": 0.83, "learning_rate": 7.777371036046561e-07, "loss": 3.3933, "step": 6922 }, { "epoch": 0.83, "learning_rate": 7.766630389285373e-07, "loss": 3.378, "step": 6923 }, { "epoch": 0.83, "learning_rate": 7.755896539582958e-07, "loss": 3.4796, "step": 6924 }, { "epoch": 0.83, "learning_rate": 7.74516948866682e-07, "loss": 3.4167, "step": 6925 }, { "epoch": 0.83, "learning_rate": 7.734449238263381e-07, "loss": 3.3189, "step": 6926 }, { "epoch": 0.83, "learning_rate": 7.723735790097953e-07, "loss": 3.3662, "step": 6927 }, { "epoch": 0.83, "learning_rate": 7.713029145894773e-07, "loss": 3.4528, "step": 6928 }, { "epoch": 0.83, "learning_rate": 7.702329307376972e-07, "loss": 3.3232, "step": 6929 }, { "epoch": 0.83, "learning_rate": 7.691636276266579e-07, "loss": 3.334, "step": 6930 }, { "epoch": 0.83, "learning_rate": 7.680950054284548e-07, "loss": 3.3873, "step": 6931 }, { "epoch": 0.83, "learning_rate": 7.670270643150717e-07, "loss": 3.3839, "step": 6932 }, { "epoch": 0.83, "learning_rate": 7.659598044583832e-07, "loss": 3.35, "step": 6933 }, { "epoch": 0.83, "learning_rate": 7.648932260301556e-07, "loss": 3.3867, "step": 6934 }, { "epoch": 0.83, "learning_rate": 7.638273292020443e-07, "loss": 3.3857, "step": 6935 }, { "epoch": 0.83, "learning_rate": 7.627621141455943e-07, "loss": 3.3892, "step": 6936 }, { "epoch": 0.83, "learning_rate": 7.61697581032243e-07, "loss": 3.4143, "step": 6937 }, { "epoch": 0.83, "learning_rate": 7.606337300333161e-07, "loss": 3.4344, "step": 6938 }, { "epoch": 0.83, "learning_rate": 7.595705613200322e-07, "loss": 3.4054, "step": 6939 }, { "epoch": 0.83, "learning_rate": 7.585080750634948e-07, "loss": 3.3392, "step": 6940 }, { "epoch": 0.83, "learning_rate": 7.574462714347036e-07, "loss": 3.3343, "step": 6941 }, { "epoch": 0.83, "learning_rate": 7.563851506045445e-07, "loss": 3.3342, "step": 6942 }, { "epoch": 0.83, "learning_rate": 7.553247127437946e-07, "loss": 3.4025, "step": 6943 }, { "epoch": 0.83, "learning_rate": 7.542649580231226e-07, "loss": 3.3334, "step": 6944 }, { "epoch": 0.83, "learning_rate": 7.532058866130848e-07, "loss": 3.4278, "step": 6945 }, { "epoch": 0.83, "learning_rate": 7.521474986841292e-07, "loss": 3.3964, "step": 6946 }, { "epoch": 0.83, "learning_rate": 7.510897944065931e-07, "loss": 3.3753, "step": 6947 }, { "epoch": 0.83, "learning_rate": 7.500327739507051e-07, "loss": 3.2601, "step": 6948 }, { "epoch": 0.83, "learning_rate": 7.489764374865804e-07, "loss": 3.4175, "step": 6949 }, { "epoch": 0.83, "learning_rate": 7.479207851842274e-07, "loss": 3.356, "step": 6950 }, { "epoch": 0.83, "learning_rate": 7.468658172135434e-07, "loss": 3.422, "step": 6951 }, { "epoch": 0.83, "learning_rate": 7.45811533744315e-07, "loss": 3.523, "step": 6952 }, { "epoch": 0.83, "learning_rate": 7.447579349462192e-07, "loss": 3.3645, "step": 6953 }, { "epoch": 0.83, "learning_rate": 7.437050209888231e-07, "loss": 3.4708, "step": 6954 }, { "epoch": 0.83, "learning_rate": 7.426527920415832e-07, "loss": 3.3018, "step": 6955 }, { "epoch": 0.83, "learning_rate": 7.416012482738449e-07, "loss": 3.4037, "step": 6956 }, { "epoch": 0.84, "learning_rate": 7.405503898548461e-07, "loss": 3.4551, "step": 6957 }, { "epoch": 0.84, "learning_rate": 7.395002169537102e-07, "loss": 3.3881, "step": 6958 }, { "epoch": 0.84, "learning_rate": 7.384507297394533e-07, "loss": 3.3414, "step": 6959 }, { "epoch": 0.84, "learning_rate": 7.374019283809802e-07, "loss": 3.4267, "step": 6960 }, { "epoch": 0.84, "learning_rate": 7.363538130470859e-07, "loss": 3.2977, "step": 6961 }, { "epoch": 0.84, "learning_rate": 7.35306383906455e-07, "loss": 3.3394, "step": 6962 }, { "epoch": 0.84, "learning_rate": 7.342596411276604e-07, "loss": 3.2979, "step": 6963 }, { "epoch": 0.84, "learning_rate": 7.332135848791655e-07, "loss": 3.3348, "step": 6964 }, { "epoch": 0.84, "learning_rate": 7.321682153293236e-07, "loss": 3.3856, "step": 6965 }, { "epoch": 0.84, "learning_rate": 7.311235326463761e-07, "loss": 3.494, "step": 6966 }, { "epoch": 0.84, "learning_rate": 7.300795369984559e-07, "loss": 3.3624, "step": 6967 }, { "epoch": 0.84, "learning_rate": 7.290362285535835e-07, "loss": 3.4621, "step": 6968 }, { "epoch": 0.84, "learning_rate": 7.279936074796695e-07, "loss": 3.3886, "step": 6969 }, { "epoch": 0.84, "learning_rate": 7.269516739445137e-07, "loss": 3.3678, "step": 6970 }, { "epoch": 0.84, "learning_rate": 7.259104281158059e-07, "loss": 3.3704, "step": 6971 }, { "epoch": 0.84, "learning_rate": 7.24869870161124e-07, "loss": 3.3876, "step": 6972 }, { "epoch": 0.84, "learning_rate": 7.238300002479364e-07, "loss": 3.375, "step": 6973 }, { "epoch": 0.84, "learning_rate": 7.227908185436e-07, "loss": 3.3445, "step": 6974 }, { "epoch": 0.84, "learning_rate": 7.21752325215363e-07, "loss": 3.365, "step": 6975 }, { "epoch": 0.84, "learning_rate": 7.207145204303573e-07, "loss": 3.4442, "step": 6976 }, { "epoch": 0.84, "learning_rate": 7.196774043556104e-07, "loss": 3.3966, "step": 6977 }, { "epoch": 0.84, "learning_rate": 7.186409771580355e-07, "loss": 3.242, "step": 6978 }, { "epoch": 0.84, "learning_rate": 7.176052390044358e-07, "loss": 3.4522, "step": 6979 }, { "epoch": 0.84, "learning_rate": 7.16570190061503e-07, "loss": 3.3635, "step": 6980 }, { "epoch": 0.84, "learning_rate": 7.155358304958193e-07, "loss": 3.4349, "step": 6981 }, { "epoch": 0.84, "learning_rate": 7.145021604738545e-07, "loss": 3.3992, "step": 6982 }, { "epoch": 0.84, "learning_rate": 7.134691801619681e-07, "loss": 3.3521, "step": 6983 }, { "epoch": 0.84, "learning_rate": 7.124368897264095e-07, "loss": 3.4274, "step": 6984 }, { "epoch": 0.84, "learning_rate": 7.114052893333134e-07, "loss": 3.3323, "step": 6985 }, { "epoch": 0.84, "learning_rate": 7.103743791487078e-07, "loss": 3.4393, "step": 6986 }, { "epoch": 0.84, "learning_rate": 7.093441593385075e-07, "loss": 3.42, "step": 6987 }, { "epoch": 0.84, "learning_rate": 7.083146300685173e-07, "loss": 3.378, "step": 6988 }, { "epoch": 0.84, "learning_rate": 7.07285791504429e-07, "loss": 3.4118, "step": 6989 }, { "epoch": 0.84, "learning_rate": 7.062576438118246e-07, "loss": 3.3844, "step": 6990 }, { "epoch": 0.84, "learning_rate": 7.052301871561757e-07, "loss": 3.3767, "step": 6991 }, { "epoch": 0.84, "learning_rate": 7.042034217028404e-07, "loss": 3.3555, "step": 6992 }, { "epoch": 0.84, "learning_rate": 7.031773476170689e-07, "loss": 3.3839, "step": 6993 }, { "epoch": 0.84, "learning_rate": 7.021519650639952e-07, "loss": 3.3821, "step": 6994 }, { "epoch": 0.84, "learning_rate": 7.011272742086467e-07, "loss": 3.4122, "step": 6995 }, { "epoch": 0.84, "learning_rate": 7.001032752159364e-07, "loss": 3.3547, "step": 6996 }, { "epoch": 0.84, "learning_rate": 6.990799682506683e-07, "loss": 3.3939, "step": 6997 }, { "epoch": 0.84, "learning_rate": 6.980573534775337e-07, "loss": 3.3155, "step": 6998 }, { "epoch": 0.84, "learning_rate": 6.970354310611122e-07, "loss": 3.3653, "step": 6999 }, { "epoch": 0.84, "learning_rate": 6.960142011658727e-07, "loss": 3.4673, "step": 7000 }, { "epoch": 0.84, "learning_rate": 6.949936639561727e-07, "loss": 3.3614, "step": 7001 }, { "epoch": 0.84, "learning_rate": 6.939738195962581e-07, "loss": 3.3945, "step": 7002 }, { "epoch": 0.84, "learning_rate": 6.92954668250263e-07, "loss": 3.4111, "step": 7003 }, { "epoch": 0.84, "learning_rate": 6.919362100822097e-07, "loss": 3.3707, "step": 7004 }, { "epoch": 0.84, "learning_rate": 6.909184452560092e-07, "loss": 3.3907, "step": 7005 }, { "epoch": 0.84, "learning_rate": 6.899013739354621e-07, "loss": 3.4596, "step": 7006 }, { "epoch": 0.84, "learning_rate": 6.888849962842553e-07, "loss": 3.3587, "step": 7007 }, { "epoch": 0.84, "learning_rate": 6.878693124659658e-07, "loss": 3.3636, "step": 7008 }, { "epoch": 0.84, "learning_rate": 6.868543226440577e-07, "loss": 3.39, "step": 7009 }, { "epoch": 0.84, "learning_rate": 6.858400269818843e-07, "loss": 3.3808, "step": 7010 }, { "epoch": 0.84, "learning_rate": 6.848264256426879e-07, "loss": 3.4026, "step": 7011 }, { "epoch": 0.84, "learning_rate": 6.838135187895961e-07, "loss": 3.4633, "step": 7012 }, { "epoch": 0.84, "learning_rate": 6.828013065856265e-07, "loss": 3.4006, "step": 7013 }, { "epoch": 0.84, "learning_rate": 6.817897891936865e-07, "loss": 3.3624, "step": 7014 }, { "epoch": 0.84, "learning_rate": 6.807789667765696e-07, "loss": 3.4517, "step": 7015 }, { "epoch": 0.84, "learning_rate": 6.797688394969576e-07, "loss": 3.3847, "step": 7016 }, { "epoch": 0.84, "learning_rate": 6.787594075174214e-07, "loss": 3.4175, "step": 7017 }, { "epoch": 0.84, "learning_rate": 6.777506710004195e-07, "loss": 3.3134, "step": 7018 }, { "epoch": 0.84, "learning_rate": 6.76742630108298e-07, "loss": 3.4046, "step": 7019 }, { "epoch": 0.84, "learning_rate": 6.757352850032933e-07, "loss": 3.3469, "step": 7020 }, { "epoch": 0.84, "learning_rate": 6.747286358475247e-07, "loss": 3.3276, "step": 7021 }, { "epoch": 0.84, "learning_rate": 6.737226828030052e-07, "loss": 3.331, "step": 7022 }, { "epoch": 0.84, "learning_rate": 6.727174260316322e-07, "loss": 3.3759, "step": 7023 }, { "epoch": 0.84, "learning_rate": 6.717128656951932e-07, "loss": 3.3717, "step": 7024 }, { "epoch": 0.84, "learning_rate": 6.707090019553619e-07, "loss": 3.3754, "step": 7025 }, { "epoch": 0.84, "learning_rate": 6.697058349737007e-07, "loss": 3.4185, "step": 7026 }, { "epoch": 0.84, "learning_rate": 6.687033649116604e-07, "loss": 3.3381, "step": 7027 }, { "epoch": 0.84, "learning_rate": 6.677015919305785e-07, "loss": 3.5004, "step": 7028 }, { "epoch": 0.84, "learning_rate": 6.667005161916807e-07, "loss": 3.4493, "step": 7029 }, { "epoch": 0.84, "learning_rate": 6.657001378560806e-07, "loss": 3.469, "step": 7030 }, { "epoch": 0.84, "learning_rate": 6.647004570847798e-07, "loss": 3.252, "step": 7031 }, { "epoch": 0.84, "learning_rate": 6.637014740386677e-07, "loss": 3.3902, "step": 7032 }, { "epoch": 0.84, "learning_rate": 6.627031888785202e-07, "loss": 3.2792, "step": 7033 }, { "epoch": 0.84, "learning_rate": 6.617056017650031e-07, "loss": 3.3112, "step": 7034 }, { "epoch": 0.84, "learning_rate": 6.607087128586675e-07, "loss": 3.4249, "step": 7035 }, { "epoch": 0.84, "learning_rate": 6.597125223199529e-07, "loss": 3.3343, "step": 7036 }, { "epoch": 0.84, "learning_rate": 6.587170303091883e-07, "loss": 3.401, "step": 7037 }, { "epoch": 0.84, "learning_rate": 6.577222369865882e-07, "loss": 3.4899, "step": 7038 }, { "epoch": 0.84, "learning_rate": 6.567281425122535e-07, "loss": 3.327, "step": 7039 }, { "epoch": 0.85, "learning_rate": 6.55734747046175e-07, "loss": 3.355, "step": 7040 }, { "epoch": 0.85, "learning_rate": 6.54742050748231e-07, "loss": 3.5705, "step": 7041 }, { "epoch": 0.85, "learning_rate": 6.53750053778186e-07, "loss": 3.4609, "step": 7042 }, { "epoch": 0.85, "learning_rate": 6.527587562956922e-07, "loss": 3.4421, "step": 7043 }, { "epoch": 0.85, "learning_rate": 6.517681584602903e-07, "loss": 3.3336, "step": 7044 }, { "epoch": 0.85, "learning_rate": 6.507782604314067e-07, "loss": 3.4326, "step": 7045 }, { "epoch": 0.85, "learning_rate": 6.497890623683567e-07, "loss": 3.4431, "step": 7046 }, { "epoch": 0.85, "learning_rate": 6.48800564430343e-07, "loss": 3.4486, "step": 7047 }, { "epoch": 0.85, "learning_rate": 6.478127667764533e-07, "loss": 3.3596, "step": 7048 }, { "epoch": 0.85, "learning_rate": 6.468256695656644e-07, "loss": 3.5347, "step": 7049 }, { "epoch": 0.85, "learning_rate": 6.458392729568408e-07, "loss": 3.383, "step": 7050 }, { "epoch": 0.85, "learning_rate": 6.448535771087338e-07, "loss": 3.4189, "step": 7051 }, { "epoch": 0.85, "learning_rate": 6.438685821799817e-07, "loss": 3.4072, "step": 7052 }, { "epoch": 0.85, "learning_rate": 6.428842883291098e-07, "loss": 3.2927, "step": 7053 }, { "epoch": 0.85, "learning_rate": 6.419006957145308e-07, "loss": 3.4884, "step": 7054 }, { "epoch": 0.85, "learning_rate": 6.409178044945447e-07, "loss": 3.3877, "step": 7055 }, { "epoch": 0.85, "learning_rate": 6.3993561482734e-07, "loss": 3.3615, "step": 7056 }, { "epoch": 0.85, "learning_rate": 6.389541268709875e-07, "loss": 3.4435, "step": 7057 }, { "epoch": 0.85, "learning_rate": 6.379733407834505e-07, "loss": 3.3406, "step": 7058 }, { "epoch": 0.85, "learning_rate": 6.369932567225767e-07, "loss": 3.4614, "step": 7059 }, { "epoch": 0.85, "learning_rate": 6.360138748461015e-07, "loss": 3.3741, "step": 7060 }, { "epoch": 0.85, "learning_rate": 6.350351953116469e-07, "loss": 3.3908, "step": 7061 }, { "epoch": 0.85, "learning_rate": 6.340572182767219e-07, "loss": 3.3182, "step": 7062 }, { "epoch": 0.85, "learning_rate": 6.330799438987228e-07, "loss": 3.4099, "step": 7063 }, { "epoch": 0.85, "learning_rate": 6.321033723349329e-07, "loss": 3.384, "step": 7064 }, { "epoch": 0.85, "learning_rate": 6.311275037425219e-07, "loss": 3.4241, "step": 7065 }, { "epoch": 0.85, "learning_rate": 6.301523382785458e-07, "loss": 3.3832, "step": 7066 }, { "epoch": 0.85, "learning_rate": 6.291778760999495e-07, "loss": 3.4002, "step": 7067 }, { "epoch": 0.85, "learning_rate": 6.282041173635628e-07, "loss": 3.4118, "step": 7068 }, { "epoch": 0.85, "learning_rate": 6.272310622261024e-07, "loss": 3.3826, "step": 7069 }, { "epoch": 0.85, "learning_rate": 6.262587108441731e-07, "loss": 3.329, "step": 7070 }, { "epoch": 0.85, "learning_rate": 6.252870633742653e-07, "loss": 3.413, "step": 7071 }, { "epoch": 0.85, "learning_rate": 6.243161199727566e-07, "loss": 3.4223, "step": 7072 }, { "epoch": 0.85, "learning_rate": 6.233458807959103e-07, "loss": 3.4972, "step": 7073 }, { "epoch": 0.85, "learning_rate": 6.223763459998794e-07, "loss": 3.3794, "step": 7074 }, { "epoch": 0.85, "learning_rate": 6.21407515740698e-07, "loss": 3.3361, "step": 7075 }, { "epoch": 0.85, "learning_rate": 6.204393901742917e-07, "loss": 3.3321, "step": 7076 }, { "epoch": 0.85, "learning_rate": 6.194719694564716e-07, "loss": 3.3952, "step": 7077 }, { "epoch": 0.85, "learning_rate": 6.185052537429336e-07, "loss": 3.3457, "step": 7078 }, { "epoch": 0.85, "learning_rate": 6.175392431892629e-07, "loss": 3.3452, "step": 7079 }, { "epoch": 0.85, "learning_rate": 6.165739379509284e-07, "loss": 3.43, "step": 7080 }, { "epoch": 0.85, "learning_rate": 6.156093381832878e-07, "loss": 3.4864, "step": 7081 }, { "epoch": 0.85, "learning_rate": 6.146454440415833e-07, "loss": 3.41, "step": 7082 }, { "epoch": 0.85, "learning_rate": 6.136822556809463e-07, "loss": 3.2865, "step": 7083 }, { "epoch": 0.85, "learning_rate": 6.127197732563905e-07, "loss": 3.461, "step": 7084 }, { "epoch": 0.85, "learning_rate": 6.117579969228193e-07, "loss": 3.4071, "step": 7085 }, { "epoch": 0.85, "learning_rate": 6.107969268350211e-07, "loss": 3.373, "step": 7086 }, { "epoch": 0.85, "learning_rate": 6.098365631476716e-07, "loss": 3.3268, "step": 7087 }, { "epoch": 0.85, "learning_rate": 6.088769060153316e-07, "loss": 3.3196, "step": 7088 }, { "epoch": 0.85, "learning_rate": 6.079179555924492e-07, "loss": 3.349, "step": 7089 }, { "epoch": 0.85, "learning_rate": 6.069597120333581e-07, "loss": 3.3506, "step": 7090 }, { "epoch": 0.85, "learning_rate": 6.06002175492279e-07, "loss": 3.3525, "step": 7091 }, { "epoch": 0.85, "learning_rate": 6.050453461233186e-07, "loss": 3.3266, "step": 7092 }, { "epoch": 0.85, "learning_rate": 6.040892240804674e-07, "loss": 3.3524, "step": 7093 }, { "epoch": 0.85, "learning_rate": 6.031338095176059e-07, "loss": 3.3803, "step": 7094 }, { "epoch": 0.85, "learning_rate": 6.021791025884982e-07, "loss": 3.3684, "step": 7095 }, { "epoch": 0.85, "learning_rate": 6.012251034467959e-07, "loss": 3.2396, "step": 7096 }, { "epoch": 0.85, "learning_rate": 6.00271812246036e-07, "loss": 3.4137, "step": 7097 }, { "epoch": 0.85, "learning_rate": 5.99319229139641e-07, "loss": 3.3504, "step": 7098 }, { "epoch": 0.85, "learning_rate": 5.983673542809209e-07, "loss": 3.3562, "step": 7099 }, { "epoch": 0.85, "learning_rate": 5.974161878230694e-07, "loss": 3.411, "step": 7100 }, { "epoch": 0.85, "learning_rate": 5.964657299191712e-07, "loss": 3.3172, "step": 7101 }, { "epoch": 0.85, "learning_rate": 5.955159807221894e-07, "loss": 3.4206, "step": 7102 }, { "epoch": 0.85, "learning_rate": 5.94566940384979e-07, "loss": 3.4635, "step": 7103 }, { "epoch": 0.85, "learning_rate": 5.936186090602792e-07, "loss": 3.3619, "step": 7104 }, { "epoch": 0.85, "learning_rate": 5.926709869007141e-07, "loss": 3.4094, "step": 7105 }, { "epoch": 0.85, "learning_rate": 5.917240740587954e-07, "loss": 3.4903, "step": 7106 }, { "epoch": 0.85, "learning_rate": 5.907778706869183e-07, "loss": 3.4096, "step": 7107 }, { "epoch": 0.85, "learning_rate": 5.898323769373665e-07, "loss": 3.4374, "step": 7108 }, { "epoch": 0.85, "learning_rate": 5.888875929623078e-07, "loss": 3.3505, "step": 7109 }, { "epoch": 0.85, "learning_rate": 5.879435189137978e-07, "loss": 3.3616, "step": 7110 }, { "epoch": 0.85, "learning_rate": 5.870001549437726e-07, "loss": 3.3737, "step": 7111 }, { "epoch": 0.85, "learning_rate": 5.860575012040604e-07, "loss": 3.3524, "step": 7112 }, { "epoch": 0.85, "learning_rate": 5.851155578463713e-07, "loss": 3.4053, "step": 7113 }, { "epoch": 0.85, "learning_rate": 5.841743250223019e-07, "loss": 3.5028, "step": 7114 }, { "epoch": 0.85, "learning_rate": 5.832338028833362e-07, "loss": 3.3796, "step": 7115 }, { "epoch": 0.85, "learning_rate": 5.822939915808407e-07, "loss": 3.3956, "step": 7116 }, { "epoch": 0.85, "learning_rate": 5.813548912660699e-07, "loss": 3.4945, "step": 7117 }, { "epoch": 0.85, "learning_rate": 5.804165020901625e-07, "loss": 3.3666, "step": 7118 }, { "epoch": 0.85, "learning_rate": 5.79478824204145e-07, "loss": 3.5463, "step": 7119 }, { "epoch": 0.85, "learning_rate": 5.785418577589246e-07, "loss": 3.4259, "step": 7120 }, { "epoch": 0.85, "learning_rate": 5.776056029052996e-07, "loss": 3.4336, "step": 7121 }, { "epoch": 0.85, "learning_rate": 5.766700597939506e-07, "loss": 3.4103, "step": 7122 }, { "epoch": 0.85, "learning_rate": 5.757352285754442e-07, "loss": 3.504, "step": 7123 }, { "epoch": 0.86, "learning_rate": 5.748011094002326e-07, "loss": 3.3293, "step": 7124 }, { "epoch": 0.86, "learning_rate": 5.738677024186534e-07, "loss": 3.3864, "step": 7125 }, { "epoch": 0.86, "learning_rate": 5.729350077809298e-07, "loss": 3.3841, "step": 7126 }, { "epoch": 0.86, "learning_rate": 5.7200302563717e-07, "loss": 3.3452, "step": 7127 }, { "epoch": 0.86, "learning_rate": 5.710717561373674e-07, "loss": 3.3164, "step": 7128 }, { "epoch": 0.86, "learning_rate": 5.701411994314015e-07, "loss": 3.3551, "step": 7129 }, { "epoch": 0.86, "learning_rate": 5.692113556690365e-07, "loss": 3.2925, "step": 7130 }, { "epoch": 0.86, "learning_rate": 5.682822249999215e-07, "loss": 3.5181, "step": 7131 }, { "epoch": 0.86, "learning_rate": 5.673538075735912e-07, "loss": 3.4082, "step": 7132 }, { "epoch": 0.86, "learning_rate": 5.664261035394658e-07, "loss": 3.3322, "step": 7133 }, { "epoch": 0.86, "learning_rate": 5.654991130468506e-07, "loss": 3.42, "step": 7134 }, { "epoch": 0.86, "learning_rate": 5.645728362449355e-07, "loss": 3.3518, "step": 7135 }, { "epoch": 0.86, "learning_rate": 5.636472732827958e-07, "loss": 3.3932, "step": 7136 }, { "epoch": 0.86, "learning_rate": 5.627224243093938e-07, "loss": 3.3303, "step": 7137 }, { "epoch": 0.86, "learning_rate": 5.617982894735724e-07, "loss": 3.3328, "step": 7138 }, { "epoch": 0.86, "learning_rate": 5.608748689240634e-07, "loss": 3.3064, "step": 7139 }, { "epoch": 0.86, "learning_rate": 5.599521628094834e-07, "loss": 3.3683, "step": 7140 }, { "epoch": 0.86, "learning_rate": 5.590301712783319e-07, "loss": 3.2807, "step": 7141 }, { "epoch": 0.86, "learning_rate": 5.581088944789953e-07, "loss": 3.486, "step": 7142 }, { "epoch": 0.86, "learning_rate": 5.571883325597443e-07, "loss": 3.3807, "step": 7143 }, { "epoch": 0.86, "learning_rate": 5.562684856687351e-07, "loss": 3.3533, "step": 7144 }, { "epoch": 0.86, "learning_rate": 5.553493539540073e-07, "loss": 3.3882, "step": 7145 }, { "epoch": 0.86, "learning_rate": 5.544309375634882e-07, "loss": 3.4141, "step": 7146 }, { "epoch": 0.86, "learning_rate": 5.535132366449852e-07, "loss": 3.3391, "step": 7147 }, { "epoch": 0.86, "learning_rate": 5.525962513461958e-07, "loss": 3.3943, "step": 7148 }, { "epoch": 0.86, "learning_rate": 5.516799818146989e-07, "loss": 3.4035, "step": 7149 }, { "epoch": 0.86, "learning_rate": 5.507644281979602e-07, "loss": 3.411, "step": 7150 }, { "epoch": 0.86, "learning_rate": 5.498495906433287e-07, "loss": 3.3983, "step": 7151 }, { "epoch": 0.86, "learning_rate": 5.489354692980397e-07, "loss": 3.4339, "step": 7152 }, { "epoch": 0.86, "learning_rate": 5.480220643092116e-07, "loss": 3.3567, "step": 7153 }, { "epoch": 0.86, "learning_rate": 5.471093758238477e-07, "loss": 3.3978, "step": 7154 }, { "epoch": 0.86, "learning_rate": 5.46197403988839e-07, "loss": 3.4129, "step": 7155 }, { "epoch": 0.86, "learning_rate": 5.452861489509554e-07, "loss": 3.4321, "step": 7156 }, { "epoch": 0.86, "learning_rate": 5.443756108568565e-07, "loss": 3.3324, "step": 7157 }, { "epoch": 0.86, "learning_rate": 5.434657898530843e-07, "loss": 3.3682, "step": 7158 }, { "epoch": 0.86, "learning_rate": 5.42556686086066e-07, "loss": 3.4474, "step": 7159 }, { "epoch": 0.86, "learning_rate": 5.416482997021127e-07, "loss": 3.3176, "step": 7160 }, { "epoch": 0.86, "learning_rate": 5.407406308474217e-07, "loss": 3.4082, "step": 7161 }, { "epoch": 0.86, "learning_rate": 5.398336796680726e-07, "loss": 3.4486, "step": 7162 }, { "epoch": 0.86, "learning_rate": 5.389274463100308e-07, "loss": 3.3592, "step": 7163 }, { "epoch": 0.86, "learning_rate": 5.380219309191465e-07, "loss": 3.3323, "step": 7164 }, { "epoch": 0.86, "learning_rate": 5.371171336411529e-07, "loss": 3.4199, "step": 7165 }, { "epoch": 0.86, "learning_rate": 5.362130546216693e-07, "loss": 3.4437, "step": 7166 }, { "epoch": 0.86, "learning_rate": 5.353096940061981e-07, "loss": 3.3868, "step": 7167 }, { "epoch": 0.86, "learning_rate": 5.344070519401267e-07, "loss": 3.3725, "step": 7168 }, { "epoch": 0.86, "learning_rate": 5.335051285687271e-07, "loss": 3.3377, "step": 7169 }, { "epoch": 0.86, "learning_rate": 5.326039240371544e-07, "loss": 3.4468, "step": 7170 }, { "epoch": 0.86, "learning_rate": 5.317034384904502e-07, "loss": 3.3834, "step": 7171 }, { "epoch": 0.86, "learning_rate": 5.308036720735376e-07, "loss": 3.3388, "step": 7172 }, { "epoch": 0.86, "learning_rate": 5.299046249312267e-07, "loss": 3.3233, "step": 7173 }, { "epoch": 0.86, "learning_rate": 5.290062972082111e-07, "loss": 3.4087, "step": 7174 }, { "epoch": 0.86, "learning_rate": 5.281086890490655e-07, "loss": 3.424, "step": 7175 }, { "epoch": 0.86, "learning_rate": 5.272118005982535e-07, "loss": 3.3847, "step": 7176 }, { "epoch": 0.86, "learning_rate": 5.263156320001201e-07, "loss": 3.2485, "step": 7177 }, { "epoch": 0.86, "learning_rate": 5.254201833988947e-07, "loss": 3.3396, "step": 7178 }, { "epoch": 0.86, "learning_rate": 5.245254549386924e-07, "loss": 3.2819, "step": 7179 }, { "epoch": 0.86, "learning_rate": 5.236314467635101e-07, "loss": 3.3475, "step": 7180 }, { "epoch": 0.86, "learning_rate": 5.227381590172309e-07, "loss": 3.3392, "step": 7181 }, { "epoch": 0.86, "learning_rate": 5.2184559184362e-07, "loss": 3.3371, "step": 7182 }, { "epoch": 0.86, "learning_rate": 5.20953745386329e-07, "loss": 3.3586, "step": 7183 }, { "epoch": 0.86, "learning_rate": 5.200626197888903e-07, "loss": 3.3253, "step": 7184 }, { "epoch": 0.86, "learning_rate": 5.191722151947227e-07, "loss": 3.5713, "step": 7185 }, { "epoch": 0.86, "learning_rate": 5.182825317471285e-07, "loss": 3.4185, "step": 7186 }, { "epoch": 0.86, "learning_rate": 5.17393569589294e-07, "loss": 3.4223, "step": 7187 }, { "epoch": 0.86, "learning_rate": 5.16505328864289e-07, "loss": 3.3912, "step": 7188 }, { "epoch": 0.86, "learning_rate": 5.156178097150671e-07, "loss": 3.3252, "step": 7189 }, { "epoch": 0.86, "learning_rate": 5.147310122844668e-07, "loss": 3.3869, "step": 7190 }, { "epoch": 0.86, "learning_rate": 5.138449367152087e-07, "loss": 3.4263, "step": 7191 }, { "epoch": 0.86, "learning_rate": 5.129595831499001e-07, "loss": 3.2964, "step": 7192 }, { "epoch": 0.86, "learning_rate": 5.120749517310281e-07, "loss": 3.4323, "step": 7193 }, { "epoch": 0.86, "learning_rate": 5.111910426009664e-07, "loss": 3.3258, "step": 7194 }, { "epoch": 0.86, "learning_rate": 5.10307855901972e-07, "loss": 3.404, "step": 7195 }, { "epoch": 0.86, "learning_rate": 5.094253917761849e-07, "loss": 3.4903, "step": 7196 }, { "epoch": 0.86, "learning_rate": 5.085436503656288e-07, "loss": 3.3516, "step": 7197 }, { "epoch": 0.86, "learning_rate": 5.076626318122135e-07, "loss": 3.4475, "step": 7198 }, { "epoch": 0.86, "learning_rate": 5.067823362577295e-07, "loss": 3.4321, "step": 7199 }, { "epoch": 0.86, "learning_rate": 5.059027638438519e-07, "loss": 3.4379, "step": 7200 }, { "epoch": 0.86, "learning_rate": 5.050239147121411e-07, "loss": 3.3178, "step": 7201 }, { "epoch": 0.86, "learning_rate": 5.041457890040364e-07, "loss": 3.4651, "step": 7202 }, { "epoch": 0.86, "learning_rate": 5.032683868608656e-07, "loss": 3.4902, "step": 7203 }, { "epoch": 0.86, "learning_rate": 5.023917084238378e-07, "loss": 3.4763, "step": 7204 }, { "epoch": 0.86, "learning_rate": 5.015157538340465e-07, "loss": 3.3453, "step": 7205 }, { "epoch": 0.86, "learning_rate": 5.006405232324679e-07, "loss": 3.4185, "step": 7206 }, { "epoch": 0.87, "learning_rate": 4.997660167599621e-07, "loss": 3.3111, "step": 7207 }, { "epoch": 0.87, "learning_rate": 4.988922345572727e-07, "loss": 3.5107, "step": 7208 }, { "epoch": 0.87, "learning_rate": 4.980191767650266e-07, "loss": 3.3791, "step": 7209 }, { "epoch": 0.87, "learning_rate": 4.971468435237347e-07, "loss": 3.3456, "step": 7210 }, { "epoch": 0.87, "learning_rate": 4.962752349737893e-07, "loss": 3.4337, "step": 7211 }, { "epoch": 0.87, "learning_rate": 4.954043512554685e-07, "loss": 3.4748, "step": 7212 }, { "epoch": 0.87, "learning_rate": 4.945341925089326e-07, "loss": 3.3833, "step": 7213 }, { "epoch": 0.87, "learning_rate": 4.936647588742249e-07, "loss": 3.4545, "step": 7214 }, { "epoch": 0.87, "learning_rate": 4.927960504912732e-07, "loss": 3.3453, "step": 7215 }, { "epoch": 0.87, "learning_rate": 4.919280674998878e-07, "loss": 3.374, "step": 7216 }, { "epoch": 0.87, "learning_rate": 4.910608100397618e-07, "loss": 3.3844, "step": 7217 }, { "epoch": 0.87, "learning_rate": 4.901942782504721e-07, "loss": 3.4238, "step": 7218 }, { "epoch": 0.87, "learning_rate": 4.893284722714803e-07, "loss": 3.3963, "step": 7219 }, { "epoch": 0.87, "learning_rate": 4.884633922421267e-07, "loss": 3.3072, "step": 7220 }, { "epoch": 0.87, "learning_rate": 4.875990383016393e-07, "loss": 3.4332, "step": 7221 }, { "epoch": 0.87, "learning_rate": 4.867354105891281e-07, "loss": 3.4137, "step": 7222 }, { "epoch": 0.87, "learning_rate": 4.858725092435845e-07, "loss": 3.367, "step": 7223 }, { "epoch": 0.87, "learning_rate": 4.850103344038853e-07, "loss": 3.3519, "step": 7224 }, { "epoch": 0.87, "learning_rate": 4.841488862087895e-07, "loss": 3.4447, "step": 7225 }, { "epoch": 0.87, "learning_rate": 4.832881647969379e-07, "loss": 3.3853, "step": 7226 }, { "epoch": 0.87, "learning_rate": 4.824281703068562e-07, "loss": 3.4888, "step": 7227 }, { "epoch": 0.87, "learning_rate": 4.815689028769521e-07, "loss": 3.3327, "step": 7228 }, { "epoch": 0.87, "learning_rate": 4.807103626455167e-07, "loss": 3.3245, "step": 7229 }, { "epoch": 0.87, "learning_rate": 4.798525497507234e-07, "loss": 3.3239, "step": 7230 }, { "epoch": 0.87, "learning_rate": 4.789954643306294e-07, "loss": 3.3817, "step": 7231 }, { "epoch": 0.87, "learning_rate": 4.78139106523175e-07, "loss": 3.3908, "step": 7232 }, { "epoch": 0.87, "learning_rate": 4.772834764661816e-07, "loss": 3.442, "step": 7233 }, { "epoch": 0.87, "learning_rate": 4.764285742973557e-07, "loss": 3.3939, "step": 7234 }, { "epoch": 0.87, "learning_rate": 4.7557440015428503e-07, "loss": 3.4147, "step": 7235 }, { "epoch": 0.87, "learning_rate": 4.747209541744413e-07, "loss": 3.3665, "step": 7236 }, { "epoch": 0.87, "learning_rate": 4.738682364951791e-07, "loss": 3.3769, "step": 7237 }, { "epoch": 0.87, "learning_rate": 4.730162472537331e-07, "loss": 3.3279, "step": 7238 }, { "epoch": 0.87, "learning_rate": 4.7216498658722464e-07, "loss": 3.4418, "step": 7239 }, { "epoch": 0.87, "learning_rate": 4.7131445463265533e-07, "loss": 3.5102, "step": 7240 }, { "epoch": 0.87, "learning_rate": 4.704646515269107e-07, "loss": 3.4651, "step": 7241 }, { "epoch": 0.87, "learning_rate": 4.696155774067579e-07, "loss": 3.3439, "step": 7242 }, { "epoch": 0.87, "learning_rate": 4.687672324088477e-07, "loss": 3.4419, "step": 7243 }, { "epoch": 0.87, "learning_rate": 4.6791961666971254e-07, "loss": 3.3204, "step": 7244 }, { "epoch": 0.87, "learning_rate": 4.6707273032576936e-07, "loss": 3.4221, "step": 7245 }, { "epoch": 0.87, "learning_rate": 4.662265735133159e-07, "loss": 3.4223, "step": 7246 }, { "epoch": 0.87, "learning_rate": 4.653811463685326e-07, "loss": 3.4056, "step": 7247 }, { "epoch": 0.87, "learning_rate": 4.6453644902748287e-07, "loss": 3.3763, "step": 7248 }, { "epoch": 0.87, "learning_rate": 4.6369248162611293e-07, "loss": 3.3743, "step": 7249 }, { "epoch": 0.87, "learning_rate": 4.6284924430025137e-07, "loss": 3.3625, "step": 7250 }, { "epoch": 0.87, "learning_rate": 4.620067371856091e-07, "loss": 3.2602, "step": 7251 }, { "epoch": 0.87, "learning_rate": 4.611649604177798e-07, "loss": 3.3456, "step": 7252 }, { "epoch": 0.87, "learning_rate": 4.603239141322396e-07, "loss": 3.3799, "step": 7253 }, { "epoch": 0.87, "learning_rate": 4.5948359846434644e-07, "loss": 3.3982, "step": 7254 }, { "epoch": 0.87, "learning_rate": 4.586440135493425e-07, "loss": 3.3862, "step": 7255 }, { "epoch": 0.87, "learning_rate": 4.578051595223487e-07, "loss": 3.3525, "step": 7256 }, { "epoch": 0.87, "learning_rate": 4.569670365183715e-07, "loss": 3.3999, "step": 7257 }, { "epoch": 0.87, "learning_rate": 4.56129644672299e-07, "loss": 3.3864, "step": 7258 }, { "epoch": 0.87, "learning_rate": 4.552929841189019e-07, "loss": 3.3748, "step": 7259 }, { "epoch": 0.87, "learning_rate": 4.544570549928318e-07, "loss": 3.383, "step": 7260 }, { "epoch": 0.87, "learning_rate": 4.536218574286244e-07, "loss": 3.3049, "step": 7261 }, { "epoch": 0.87, "learning_rate": 4.527873915606967e-07, "loss": 3.331, "step": 7262 }, { "epoch": 0.87, "learning_rate": 4.519536575233474e-07, "loss": 3.4229, "step": 7263 }, { "epoch": 0.87, "learning_rate": 4.5112065545075856e-07, "loss": 3.3552, "step": 7264 }, { "epoch": 0.87, "learning_rate": 4.502883854769935e-07, "loss": 3.3981, "step": 7265 }, { "epoch": 0.87, "learning_rate": 4.4945684773599895e-07, "loss": 3.3289, "step": 7266 }, { "epoch": 0.87, "learning_rate": 4.4862604236160234e-07, "loss": 3.4021, "step": 7267 }, { "epoch": 0.87, "learning_rate": 4.477959694875139e-07, "loss": 3.4084, "step": 7268 }, { "epoch": 0.87, "learning_rate": 4.4696662924732683e-07, "loss": 3.3588, "step": 7269 }, { "epoch": 0.87, "learning_rate": 4.4613802177451424e-07, "loss": 3.4217, "step": 7270 }, { "epoch": 0.87, "learning_rate": 4.4531014720243295e-07, "loss": 3.3781, "step": 7271 }, { "epoch": 0.87, "learning_rate": 4.4448300566432234e-07, "loss": 3.4667, "step": 7272 }, { "epoch": 0.87, "learning_rate": 4.4365659729330266e-07, "loss": 3.4306, "step": 7273 }, { "epoch": 0.87, "learning_rate": 4.428309222223759e-07, "loss": 3.3217, "step": 7274 }, { "epoch": 0.87, "learning_rate": 4.420059805844268e-07, "loss": 3.3498, "step": 7275 }, { "epoch": 0.87, "learning_rate": 4.4118177251222137e-07, "loss": 3.4478, "step": 7276 }, { "epoch": 0.87, "learning_rate": 4.4035829813840913e-07, "loss": 3.3237, "step": 7277 }, { "epoch": 0.87, "learning_rate": 4.395355575955201e-07, "loss": 3.4755, "step": 7278 }, { "epoch": 0.87, "learning_rate": 4.3871355101596626e-07, "loss": 3.3303, "step": 7279 }, { "epoch": 0.87, "learning_rate": 4.3789227853204163e-07, "loss": 3.3875, "step": 7280 }, { "epoch": 0.87, "learning_rate": 4.3707174027592225e-07, "loss": 3.5319, "step": 7281 }, { "epoch": 0.87, "learning_rate": 4.3625193637966747e-07, "loss": 3.3227, "step": 7282 }, { "epoch": 0.87, "learning_rate": 4.354328669752145e-07, "loss": 3.2239, "step": 7283 }, { "epoch": 0.87, "learning_rate": 4.3461453219438567e-07, "loss": 3.3419, "step": 7284 }, { "epoch": 0.87, "learning_rate": 4.337969321688845e-07, "loss": 3.4603, "step": 7285 }, { "epoch": 0.87, "learning_rate": 4.329800670302958e-07, "loss": 3.3266, "step": 7286 }, { "epoch": 0.87, "learning_rate": 4.321639369100866e-07, "loss": 3.4197, "step": 7287 }, { "epoch": 0.87, "learning_rate": 4.313485419396041e-07, "loss": 3.2105, "step": 7288 }, { "epoch": 0.87, "learning_rate": 4.3053388225007984e-07, "loss": 3.4184, "step": 7289 }, { "epoch": 0.88, "learning_rate": 4.2971995797262466e-07, "loss": 3.3956, "step": 7290 }, { "epoch": 0.88, "learning_rate": 4.289067692382326e-07, "loss": 3.4406, "step": 7291 }, { "epoch": 0.88, "learning_rate": 4.2809431617777787e-07, "loss": 3.4221, "step": 7292 }, { "epoch": 0.88, "learning_rate": 4.2728259892201696e-07, "loss": 3.2942, "step": 7293 }, { "epoch": 0.88, "learning_rate": 4.264716176015882e-07, "loss": 3.4418, "step": 7294 }, { "epoch": 0.88, "learning_rate": 4.25661372347011e-07, "loss": 3.3382, "step": 7295 }, { "epoch": 0.88, "learning_rate": 4.2485186328868776e-07, "loss": 3.3772, "step": 7296 }, { "epoch": 0.88, "learning_rate": 4.2404309055690094e-07, "loss": 3.4145, "step": 7297 }, { "epoch": 0.88, "learning_rate": 4.2323505428181424e-07, "loss": 3.2168, "step": 7298 }, { "epoch": 0.88, "learning_rate": 4.2242775459347407e-07, "loss": 3.4697, "step": 7299 }, { "epoch": 0.88, "learning_rate": 4.216211916218077e-07, "loss": 3.4078, "step": 7300 }, { "epoch": 0.88, "learning_rate": 4.208153654966224e-07, "loss": 3.3059, "step": 7301 }, { "epoch": 0.88, "learning_rate": 4.2001027634760936e-07, "loss": 3.4623, "step": 7302 }, { "epoch": 0.88, "learning_rate": 4.1920592430433946e-07, "loss": 3.404, "step": 7303 }, { "epoch": 0.88, "learning_rate": 4.184023094962664e-07, "loss": 3.3894, "step": 7304 }, { "epoch": 0.88, "learning_rate": 4.1759943205272323e-07, "loss": 3.3078, "step": 7305 }, { "epoch": 0.88, "learning_rate": 4.167972921029262e-07, "loss": 3.4022, "step": 7306 }, { "epoch": 0.88, "learning_rate": 4.1599588977597137e-07, "loss": 3.2626, "step": 7307 }, { "epoch": 0.88, "learning_rate": 4.1519522520083787e-07, "loss": 3.3854, "step": 7308 }, { "epoch": 0.88, "learning_rate": 4.143952985063848e-07, "loss": 3.4019, "step": 7309 }, { "epoch": 0.88, "learning_rate": 4.13596109821352e-07, "loss": 3.392, "step": 7310 }, { "epoch": 0.88, "learning_rate": 4.1279765927436156e-07, "loss": 3.3494, "step": 7311 }, { "epoch": 0.88, "learning_rate": 4.119999469939162e-07, "loss": 3.4001, "step": 7312 }, { "epoch": 0.88, "learning_rate": 4.1120297310840107e-07, "loss": 3.353, "step": 7313 }, { "epoch": 0.88, "learning_rate": 4.1040673774608076e-07, "loss": 3.3694, "step": 7314 }, { "epoch": 0.88, "learning_rate": 4.096112410351022e-07, "loss": 3.3732, "step": 7315 }, { "epoch": 0.88, "learning_rate": 4.088164831034924e-07, "loss": 3.4672, "step": 7316 }, { "epoch": 0.88, "learning_rate": 4.080224640791608e-07, "loss": 3.4652, "step": 7317 }, { "epoch": 0.88, "learning_rate": 4.0722918408989786e-07, "loss": 3.446, "step": 7318 }, { "epoch": 0.88, "learning_rate": 4.0643664326337264e-07, "loss": 3.4347, "step": 7319 }, { "epoch": 0.88, "learning_rate": 4.0564484172713745e-07, "loss": 3.3781, "step": 7320 }, { "epoch": 0.88, "learning_rate": 4.04853779608626e-07, "loss": 3.2672, "step": 7321 }, { "epoch": 0.88, "learning_rate": 4.0406345703515205e-07, "loss": 3.3075, "step": 7322 }, { "epoch": 0.88, "learning_rate": 4.0327387413390983e-07, "loss": 3.4166, "step": 7323 }, { "epoch": 0.88, "learning_rate": 4.024850310319761e-07, "loss": 3.3379, "step": 7324 }, { "epoch": 0.88, "learning_rate": 4.0169692785630656e-07, "loss": 3.3397, "step": 7325 }, { "epoch": 0.88, "learning_rate": 4.009095647337402e-07, "loss": 3.3789, "step": 7326 }, { "epoch": 0.88, "learning_rate": 4.0012294179099466e-07, "loss": 3.2851, "step": 7327 }, { "epoch": 0.88, "learning_rate": 3.993370591546697e-07, "loss": 3.3893, "step": 7328 }, { "epoch": 0.88, "learning_rate": 3.9855191695124584e-07, "loss": 3.3991, "step": 7329 }, { "epoch": 0.88, "learning_rate": 3.977675153070837e-07, "loss": 3.3886, "step": 7330 }, { "epoch": 0.88, "learning_rate": 3.969838543484261e-07, "loss": 3.3812, "step": 7331 }, { "epoch": 0.88, "learning_rate": 3.9620093420139507e-07, "loss": 3.4173, "step": 7332 }, { "epoch": 0.88, "learning_rate": 3.9541875499199465e-07, "loss": 3.435, "step": 7333 }, { "epoch": 0.88, "learning_rate": 3.946373168461087e-07, "loss": 3.2792, "step": 7334 }, { "epoch": 0.88, "learning_rate": 3.9385661988950273e-07, "loss": 3.4066, "step": 7335 }, { "epoch": 0.88, "learning_rate": 3.930766642478229e-07, "loss": 3.4696, "step": 7336 }, { "epoch": 0.88, "learning_rate": 3.922974500465943e-07, "loss": 3.4434, "step": 7337 }, { "epoch": 0.88, "learning_rate": 3.915189774112244e-07, "loss": 3.3917, "step": 7338 }, { "epoch": 0.88, "learning_rate": 3.907412464670013e-07, "loss": 3.5228, "step": 7339 }, { "epoch": 0.88, "learning_rate": 3.8996425733909383e-07, "loss": 3.4112, "step": 7340 }, { "epoch": 0.88, "learning_rate": 3.8918801015255025e-07, "loss": 3.4107, "step": 7341 }, { "epoch": 0.88, "learning_rate": 3.884125050323001e-07, "loss": 3.4023, "step": 7342 }, { "epoch": 0.88, "learning_rate": 3.8763774210315406e-07, "loss": 3.3232, "step": 7343 }, { "epoch": 0.88, "learning_rate": 3.86863721489803e-07, "loss": 3.3958, "step": 7344 }, { "epoch": 0.88, "learning_rate": 3.8609044331681833e-07, "loss": 3.4625, "step": 7345 }, { "epoch": 0.88, "learning_rate": 3.853179077086505e-07, "loss": 3.3847, "step": 7346 }, { "epoch": 0.88, "learning_rate": 3.8454611478963235e-07, "loss": 3.3721, "step": 7347 }, { "epoch": 0.88, "learning_rate": 3.837750646839772e-07, "loss": 3.3669, "step": 7348 }, { "epoch": 0.88, "learning_rate": 3.830047575157775e-07, "loss": 3.405, "step": 7349 }, { "epoch": 0.88, "learning_rate": 3.822351934090074e-07, "loss": 3.4606, "step": 7350 }, { "epoch": 0.88, "learning_rate": 3.814663724875206e-07, "loss": 3.3763, "step": 7351 }, { "epoch": 0.88, "learning_rate": 3.806982948750515e-07, "loss": 3.4599, "step": 7352 }, { "epoch": 0.88, "learning_rate": 3.799309606952156e-07, "loss": 3.386, "step": 7353 }, { "epoch": 0.88, "learning_rate": 3.791643700715075e-07, "loss": 3.3708, "step": 7354 }, { "epoch": 0.88, "learning_rate": 3.783985231273024e-07, "loss": 3.3893, "step": 7355 }, { "epoch": 0.88, "learning_rate": 3.7763341998585613e-07, "loss": 3.3967, "step": 7356 }, { "epoch": 0.88, "learning_rate": 3.768690607703046e-07, "loss": 3.3738, "step": 7357 }, { "epoch": 0.88, "learning_rate": 3.761054456036645e-07, "loss": 3.3679, "step": 7358 }, { "epoch": 0.88, "learning_rate": 3.7534257460883293e-07, "loss": 3.4554, "step": 7359 }, { "epoch": 0.88, "learning_rate": 3.745804479085857e-07, "loss": 3.3815, "step": 7360 }, { "epoch": 0.88, "learning_rate": 3.7381906562558025e-07, "loss": 3.4559, "step": 7361 }, { "epoch": 0.88, "learning_rate": 3.7305842788235347e-07, "loss": 3.3086, "step": 7362 }, { "epoch": 0.88, "learning_rate": 3.722985348013236e-07, "loss": 3.4244, "step": 7363 }, { "epoch": 0.88, "learning_rate": 3.715393865047878e-07, "loss": 3.2749, "step": 7364 }, { "epoch": 0.88, "learning_rate": 3.7078098311492396e-07, "loss": 3.3983, "step": 7365 }, { "epoch": 0.88, "learning_rate": 3.7002332475378935e-07, "loss": 3.3001, "step": 7366 }, { "epoch": 0.88, "learning_rate": 3.69266411543322e-07, "loss": 3.3644, "step": 7367 }, { "epoch": 0.88, "learning_rate": 3.6851024360534003e-07, "loss": 3.4234, "step": 7368 }, { "epoch": 0.88, "learning_rate": 3.677548210615417e-07, "loss": 3.3396, "step": 7369 }, { "epoch": 0.88, "learning_rate": 3.670001440335047e-07, "loss": 3.5083, "step": 7370 }, { "epoch": 0.88, "learning_rate": 3.662462126426874e-07, "loss": 3.4418, "step": 7371 }, { "epoch": 0.88, "learning_rate": 3.65493027010429e-07, "loss": 3.4738, "step": 7372 }, { "epoch": 0.89, "learning_rate": 3.647405872579446e-07, "loss": 3.2672, "step": 7373 }, { "epoch": 0.89, "learning_rate": 3.639888935063346e-07, "loss": 3.4035, "step": 7374 }, { "epoch": 0.89, "learning_rate": 3.632379458765761e-07, "loss": 3.4904, "step": 7375 }, { "epoch": 0.89, "learning_rate": 3.6248774448952695e-07, "loss": 3.4673, "step": 7376 }, { "epoch": 0.89, "learning_rate": 3.617382894659249e-07, "loss": 3.3589, "step": 7377 }, { "epoch": 0.89, "learning_rate": 3.6098958092638857e-07, "loss": 3.4235, "step": 7378 }, { "epoch": 0.89, "learning_rate": 3.602416189914143e-07, "loss": 3.3927, "step": 7379 }, { "epoch": 0.89, "learning_rate": 3.5949440378137967e-07, "loss": 3.3431, "step": 7380 }, { "epoch": 0.89, "learning_rate": 3.5874793541654294e-07, "loss": 3.5045, "step": 7381 }, { "epoch": 0.89, "learning_rate": 3.580022140170397e-07, "loss": 3.3554, "step": 7382 }, { "epoch": 0.89, "learning_rate": 3.5725723970288673e-07, "loss": 3.4465, "step": 7383 }, { "epoch": 0.89, "learning_rate": 3.565130125939814e-07, "loss": 3.4042, "step": 7384 }, { "epoch": 0.89, "learning_rate": 3.557695328100996e-07, "loss": 3.4688, "step": 7385 }, { "epoch": 0.89, "learning_rate": 3.5502680047089734e-07, "loss": 3.3684, "step": 7386 }, { "epoch": 0.89, "learning_rate": 3.5428481569591e-07, "loss": 3.4689, "step": 7387 }, { "epoch": 0.89, "learning_rate": 3.535435786045538e-07, "loss": 3.3323, "step": 7388 }, { "epoch": 0.89, "learning_rate": 3.528030893161227e-07, "loss": 3.3494, "step": 7389 }, { "epoch": 0.89, "learning_rate": 3.5206334794979314e-07, "loss": 3.333, "step": 7390 }, { "epoch": 0.89, "learning_rate": 3.5132435462461757e-07, "loss": 3.3404, "step": 7391 }, { "epoch": 0.89, "learning_rate": 3.505861094595292e-07, "loss": 3.445, "step": 7392 }, { "epoch": 0.89, "learning_rate": 3.4984861257334413e-07, "loss": 3.3918, "step": 7393 }, { "epoch": 0.89, "learning_rate": 3.491118640847546e-07, "loss": 3.2978, "step": 7394 }, { "epoch": 0.89, "learning_rate": 3.4837586411233306e-07, "loss": 3.4669, "step": 7395 }, { "epoch": 0.89, "learning_rate": 3.476406127745313e-07, "loss": 3.339, "step": 7396 }, { "epoch": 0.89, "learning_rate": 3.4690611018968143e-07, "loss": 3.3382, "step": 7397 }, { "epoch": 0.89, "learning_rate": 3.4617235647599446e-07, "loss": 3.3695, "step": 7398 }, { "epoch": 0.89, "learning_rate": 3.4543935175156193e-07, "loss": 3.3352, "step": 7399 }, { "epoch": 0.89, "learning_rate": 3.4470709613435295e-07, "loss": 3.365, "step": 7400 }, { "epoch": 0.89, "learning_rate": 3.43975589742217e-07, "loss": 3.32, "step": 7401 }, { "epoch": 0.89, "learning_rate": 3.4324483269288333e-07, "loss": 3.3509, "step": 7402 }, { "epoch": 0.89, "learning_rate": 3.4251482510396064e-07, "loss": 3.3988, "step": 7403 }, { "epoch": 0.89, "learning_rate": 3.417855670929365e-07, "loss": 3.3428, "step": 7404 }, { "epoch": 0.89, "learning_rate": 3.4105705877717775e-07, "loss": 3.3403, "step": 7405 }, { "epoch": 0.89, "learning_rate": 3.4032930027393165e-07, "loss": 3.3958, "step": 7406 }, { "epoch": 0.89, "learning_rate": 3.3960229170032334e-07, "loss": 3.4272, "step": 7407 }, { "epoch": 0.89, "learning_rate": 3.388760331733587e-07, "loss": 3.3884, "step": 7408 }, { "epoch": 0.89, "learning_rate": 3.3815052480992093e-07, "loss": 3.3702, "step": 7409 }, { "epoch": 0.89, "learning_rate": 3.374257667267739e-07, "loss": 3.3299, "step": 7410 }, { "epoch": 0.89, "learning_rate": 3.3670175904056136e-07, "loss": 3.4284, "step": 7411 }, { "epoch": 0.89, "learning_rate": 3.3597850186780467e-07, "loss": 3.3632, "step": 7412 }, { "epoch": 0.89, "learning_rate": 3.352559953249057e-07, "loss": 3.3964, "step": 7413 }, { "epoch": 0.89, "learning_rate": 3.345342395281448e-07, "loss": 3.4427, "step": 7414 }, { "epoch": 0.89, "learning_rate": 3.3381323459368177e-07, "loss": 3.4392, "step": 7415 }, { "epoch": 0.89, "learning_rate": 3.330929806375555e-07, "loss": 3.3647, "step": 7416 }, { "epoch": 0.89, "learning_rate": 3.3237347777568484e-07, "loss": 3.3236, "step": 7417 }, { "epoch": 0.89, "learning_rate": 3.3165472612386504e-07, "loss": 3.347, "step": 7418 }, { "epoch": 0.89, "learning_rate": 3.30936725797773e-07, "loss": 3.296, "step": 7419 }, { "epoch": 0.89, "learning_rate": 3.302194769129652e-07, "loss": 3.4452, "step": 7420 }, { "epoch": 0.89, "learning_rate": 3.295029795848742e-07, "loss": 3.3418, "step": 7421 }, { "epoch": 0.89, "learning_rate": 3.287872339288151e-07, "loss": 3.38, "step": 7422 }, { "epoch": 0.89, "learning_rate": 3.280722400599795e-07, "loss": 3.3127, "step": 7423 }, { "epoch": 0.89, "learning_rate": 3.2735799809343883e-07, "loss": 3.3639, "step": 7424 }, { "epoch": 0.89, "learning_rate": 3.2664450814414384e-07, "loss": 3.4557, "step": 7425 }, { "epoch": 0.89, "learning_rate": 3.259317703269238e-07, "loss": 3.3872, "step": 7426 }, { "epoch": 0.89, "learning_rate": 3.25219784756487e-07, "loss": 3.407, "step": 7427 }, { "epoch": 0.89, "learning_rate": 3.245085515474211e-07, "loss": 3.3453, "step": 7428 }, { "epoch": 0.89, "learning_rate": 3.237980708141919e-07, "loss": 3.3836, "step": 7429 }, { "epoch": 0.89, "learning_rate": 3.2308834267114453e-07, "loss": 3.3837, "step": 7430 }, { "epoch": 0.89, "learning_rate": 3.2237936723250375e-07, "loss": 3.4531, "step": 7431 }, { "epoch": 0.89, "learning_rate": 3.21671144612371e-07, "loss": 3.3884, "step": 7432 }, { "epoch": 0.89, "learning_rate": 3.2096367492472913e-07, "loss": 3.3907, "step": 7433 }, { "epoch": 0.89, "learning_rate": 3.2025695828343804e-07, "loss": 3.4351, "step": 7434 }, { "epoch": 0.89, "learning_rate": 3.1955099480223796e-07, "loss": 3.3032, "step": 7435 }, { "epoch": 0.89, "learning_rate": 3.188457845947457e-07, "loss": 3.456, "step": 7436 }, { "epoch": 0.89, "learning_rate": 3.181413277744588e-07, "loss": 3.486, "step": 7437 }, { "epoch": 0.89, "learning_rate": 3.174376244547528e-07, "loss": 3.4391, "step": 7438 }, { "epoch": 0.89, "learning_rate": 3.1673467474888196e-07, "loss": 3.4148, "step": 7439 }, { "epoch": 0.89, "learning_rate": 3.160324787699792e-07, "loss": 3.4083, "step": 7440 }, { "epoch": 0.89, "learning_rate": 3.153310366310569e-07, "loss": 3.3564, "step": 7441 }, { "epoch": 0.89, "learning_rate": 3.1463034844500474e-07, "loss": 3.2992, "step": 7442 }, { "epoch": 0.89, "learning_rate": 3.139304143245925e-07, "loss": 3.3817, "step": 7443 }, { "epoch": 0.89, "learning_rate": 3.132312343824684e-07, "loss": 3.3659, "step": 7444 }, { "epoch": 0.89, "learning_rate": 3.1253280873115686e-07, "loss": 3.3432, "step": 7445 }, { "epoch": 0.89, "learning_rate": 3.1183513748306417e-07, "loss": 3.2903, "step": 7446 }, { "epoch": 0.89, "learning_rate": 3.1113822075047375e-07, "loss": 3.4029, "step": 7447 }, { "epoch": 0.89, "learning_rate": 3.1044205864554755e-07, "loss": 3.4378, "step": 7448 }, { "epoch": 0.89, "learning_rate": 3.0974665128032657e-07, "loss": 3.4401, "step": 7449 }, { "epoch": 0.89, "learning_rate": 3.0905199876672954e-07, "loss": 3.3142, "step": 7450 }, { "epoch": 0.89, "learning_rate": 3.083581012165543e-07, "loss": 3.4014, "step": 7451 }, { "epoch": 0.89, "learning_rate": 3.0766495874147705e-07, "loss": 3.4436, "step": 7452 }, { "epoch": 0.89, "learning_rate": 3.069725714530536e-07, "loss": 3.4271, "step": 7453 }, { "epoch": 0.89, "learning_rate": 3.0628093946271475e-07, "loss": 3.3149, "step": 7454 }, { "epoch": 0.89, "learning_rate": 3.055900628817737e-07, "loss": 3.3096, "step": 7455 }, { "epoch": 0.89, "learning_rate": 3.0489994182142045e-07, "loss": 3.2809, "step": 7456 }, { "epoch": 0.9, "learning_rate": 3.0421057639272277e-07, "loss": 3.4356, "step": 7457 }, { "epoch": 0.9, "learning_rate": 3.03521966706628e-07, "loss": 3.4266, "step": 7458 }, { "epoch": 0.9, "learning_rate": 3.028341128739609e-07, "loss": 3.3601, "step": 7459 }, { "epoch": 0.9, "learning_rate": 3.0214701500542563e-07, "loss": 3.4183, "step": 7460 }, { "epoch": 0.9, "learning_rate": 3.0146067321160324e-07, "loss": 3.3543, "step": 7461 }, { "epoch": 0.9, "learning_rate": 3.0077508760295415e-07, "loss": 3.3136, "step": 7462 }, { "epoch": 0.9, "learning_rate": 3.000902582898174e-07, "loss": 3.2961, "step": 7463 }, { "epoch": 0.9, "learning_rate": 2.9940618538240926e-07, "loss": 3.453, "step": 7464 }, { "epoch": 0.9, "learning_rate": 2.9872286899082446e-07, "loss": 3.378, "step": 7465 }, { "epoch": 0.9, "learning_rate": 2.980403092250372e-07, "loss": 3.4201, "step": 7466 }, { "epoch": 0.9, "learning_rate": 2.9735850619489804e-07, "loss": 3.4477, "step": 7467 }, { "epoch": 0.9, "learning_rate": 2.9667746001013684e-07, "loss": 3.4388, "step": 7468 }, { "epoch": 0.9, "learning_rate": 2.9599717078036214e-07, "loss": 3.4503, "step": 7469 }, { "epoch": 0.9, "learning_rate": 2.9531763861505967e-07, "loss": 3.4071, "step": 7470 }, { "epoch": 0.9, "learning_rate": 2.946388636235942e-07, "loss": 3.3727, "step": 7471 }, { "epoch": 0.9, "learning_rate": 2.9396084591520657e-07, "loss": 3.355, "step": 7472 }, { "epoch": 0.9, "learning_rate": 2.932835855990179e-07, "loss": 3.509, "step": 7473 }, { "epoch": 0.9, "learning_rate": 2.926070827840277e-07, "loss": 3.395, "step": 7474 }, { "epoch": 0.9, "learning_rate": 2.9193133757911164e-07, "loss": 3.4142, "step": 7475 }, { "epoch": 0.9, "learning_rate": 2.912563500930254e-07, "loss": 3.4426, "step": 7476 }, { "epoch": 0.9, "learning_rate": 2.905821204344006e-07, "loss": 3.346, "step": 7477 }, { "epoch": 0.9, "learning_rate": 2.899086487117492e-07, "loss": 3.2948, "step": 7478 }, { "epoch": 0.9, "learning_rate": 2.892359350334595e-07, "loss": 3.4515, "step": 7479 }, { "epoch": 0.9, "learning_rate": 2.8856397950779934e-07, "loss": 3.4056, "step": 7480 }, { "epoch": 0.9, "learning_rate": 2.8789278224291174e-07, "loss": 3.3694, "step": 7481 }, { "epoch": 0.9, "learning_rate": 2.872223433468202e-07, "loss": 3.4133, "step": 7482 }, { "epoch": 0.9, "learning_rate": 2.865526629274262e-07, "loss": 3.3782, "step": 7483 }, { "epoch": 0.9, "learning_rate": 2.8588374109250805e-07, "loss": 3.3265, "step": 7484 }, { "epoch": 0.9, "learning_rate": 2.852155779497218e-07, "loss": 3.4206, "step": 7485 }, { "epoch": 0.9, "learning_rate": 2.8454817360660305e-07, "loss": 3.4108, "step": 7486 }, { "epoch": 0.9, "learning_rate": 2.838815281705631e-07, "loss": 3.4117, "step": 7487 }, { "epoch": 0.9, "learning_rate": 2.8321564174889236e-07, "loss": 3.4248, "step": 7488 }, { "epoch": 0.9, "learning_rate": 2.8255051444876e-07, "loss": 3.4477, "step": 7489 }, { "epoch": 0.9, "learning_rate": 2.8188614637721044e-07, "loss": 3.5068, "step": 7490 }, { "epoch": 0.9, "learning_rate": 2.812225376411681e-07, "loss": 3.3453, "step": 7491 }, { "epoch": 0.9, "learning_rate": 2.8055968834743486e-07, "loss": 3.3578, "step": 7492 }, { "epoch": 0.9, "learning_rate": 2.798975986026892e-07, "loss": 3.4275, "step": 7493 }, { "epoch": 0.9, "learning_rate": 2.792362685134892e-07, "loss": 3.4529, "step": 7494 }, { "epoch": 0.9, "learning_rate": 2.7857569818626864e-07, "loss": 3.3299, "step": 7495 }, { "epoch": 0.9, "learning_rate": 2.7791588772734026e-07, "loss": 3.5432, "step": 7496 }, { "epoch": 0.9, "learning_rate": 2.772568372428941e-07, "loss": 3.3423, "step": 7497 }, { "epoch": 0.9, "learning_rate": 2.765985468389998e-07, "loss": 3.3557, "step": 7498 }, { "epoch": 0.9, "learning_rate": 2.759410166216003e-07, "loss": 3.4281, "step": 7499 }, { "epoch": 0.9, "learning_rate": 2.7528424669651997e-07, "loss": 3.2667, "step": 7500 }, { "epoch": 0.9, "learning_rate": 2.7462823716945977e-07, "loss": 3.4065, "step": 7501 }, { "epoch": 0.9, "learning_rate": 2.7397298814599846e-07, "loss": 3.4359, "step": 7502 }, { "epoch": 0.9, "learning_rate": 2.7331849973159186e-07, "loss": 3.4121, "step": 7503 }, { "epoch": 0.9, "learning_rate": 2.7266477203157337e-07, "loss": 3.2979, "step": 7504 }, { "epoch": 0.9, "learning_rate": 2.7201180515115445e-07, "loss": 3.4128, "step": 7505 }, { "epoch": 0.9, "learning_rate": 2.713595991954238e-07, "loss": 3.4409, "step": 7506 }, { "epoch": 0.9, "learning_rate": 2.707081542693485e-07, "loss": 3.4077, "step": 7507 }, { "epoch": 0.9, "learning_rate": 2.7005747047777143e-07, "loss": 3.3558, "step": 7508 }, { "epoch": 0.9, "learning_rate": 2.6940754792541433e-07, "loss": 3.3635, "step": 7509 }, { "epoch": 0.9, "learning_rate": 2.687583867168758e-07, "loss": 3.2929, "step": 7510 }, { "epoch": 0.9, "learning_rate": 2.681099869566328e-07, "loss": 3.499, "step": 7511 }, { "epoch": 0.9, "learning_rate": 2.674623487490385e-07, "loss": 3.3957, "step": 7512 }, { "epoch": 0.9, "learning_rate": 2.66815472198324e-07, "loss": 3.4459, "step": 7513 }, { "epoch": 0.9, "learning_rate": 2.6616935740859885e-07, "loss": 3.4364, "step": 7514 }, { "epoch": 0.9, "learning_rate": 2.6552400448384807e-07, "loss": 3.4115, "step": 7515 }, { "epoch": 0.9, "learning_rate": 2.648794135279359e-07, "loss": 3.3824, "step": 7516 }, { "epoch": 0.9, "learning_rate": 2.642355846446021e-07, "loss": 3.3207, "step": 7517 }, { "epoch": 0.9, "learning_rate": 2.6359251793746596e-07, "loss": 3.4162, "step": 7518 }, { "epoch": 0.9, "learning_rate": 2.629502135100215e-07, "loss": 3.3836, "step": 7519 }, { "epoch": 0.9, "learning_rate": 2.623086714656431e-07, "loss": 3.3891, "step": 7520 }, { "epoch": 0.9, "learning_rate": 2.6166789190757935e-07, "loss": 3.3483, "step": 7521 }, { "epoch": 0.9, "learning_rate": 2.6102787493895886e-07, "loss": 3.3681, "step": 7522 }, { "epoch": 0.9, "learning_rate": 2.603886206627854e-07, "loss": 3.4067, "step": 7523 }, { "epoch": 0.9, "learning_rate": 2.5975012918194154e-07, "loss": 3.3926, "step": 7524 }, { "epoch": 0.9, "learning_rate": 2.591124005991863e-07, "loss": 3.3989, "step": 7525 }, { "epoch": 0.9, "learning_rate": 2.584754350171553e-07, "loss": 3.4602, "step": 7526 }, { "epoch": 0.9, "learning_rate": 2.5783923253836276e-07, "loss": 3.4906, "step": 7527 }, { "epoch": 0.9, "learning_rate": 2.572037932651994e-07, "loss": 3.2634, "step": 7528 }, { "epoch": 0.9, "learning_rate": 2.5656911729993294e-07, "loss": 3.3436, "step": 7529 }, { "epoch": 0.9, "learning_rate": 2.559352047447083e-07, "loss": 3.3104, "step": 7530 }, { "epoch": 0.9, "learning_rate": 2.5530205570154785e-07, "loss": 3.36, "step": 7531 }, { "epoch": 0.9, "learning_rate": 2.546696702723511e-07, "loss": 3.4317, "step": 7532 }, { "epoch": 0.9, "learning_rate": 2.5403804855889444e-07, "loss": 3.464, "step": 7533 }, { "epoch": 0.9, "learning_rate": 2.534071906628316e-07, "loss": 3.4407, "step": 7534 }, { "epoch": 0.9, "learning_rate": 2.527770966856918e-07, "loss": 3.2915, "step": 7535 }, { "epoch": 0.9, "learning_rate": 2.5214776672888407e-07, "loss": 3.4162, "step": 7536 }, { "epoch": 0.9, "learning_rate": 2.515192008936923e-07, "loss": 3.3825, "step": 7537 }, { "epoch": 0.9, "learning_rate": 2.5089139928127893e-07, "loss": 3.4939, "step": 7538 }, { "epoch": 0.9, "learning_rate": 2.50264361992682e-07, "loss": 3.3961, "step": 7539 }, { "epoch": 0.91, "learning_rate": 2.4963808912881747e-07, "loss": 3.3395, "step": 7540 }, { "epoch": 0.91, "learning_rate": 2.4901258079047753e-07, "loss": 3.3807, "step": 7541 }, { "epoch": 0.91, "learning_rate": 2.483878370783327e-07, "loss": 3.3148, "step": 7542 }, { "epoch": 0.91, "learning_rate": 2.4776385809292926e-07, "loss": 3.4171, "step": 7543 }, { "epoch": 0.91, "learning_rate": 2.471406439346902e-07, "loss": 3.4496, "step": 7544 }, { "epoch": 0.91, "learning_rate": 2.465181947039158e-07, "loss": 3.4139, "step": 7545 }, { "epoch": 0.91, "learning_rate": 2.4589651050078376e-07, "loss": 3.2883, "step": 7546 }, { "epoch": 0.91, "learning_rate": 2.4527559142534853e-07, "loss": 3.4893, "step": 7547 }, { "epoch": 0.91, "learning_rate": 2.446554375775401e-07, "loss": 3.3517, "step": 7548 }, { "epoch": 0.91, "learning_rate": 2.440360490571675e-07, "loss": 3.4559, "step": 7549 }, { "epoch": 0.91, "learning_rate": 2.4341742596391494e-07, "loss": 3.3943, "step": 7550 }, { "epoch": 0.91, "learning_rate": 2.4279956839734384e-07, "loss": 3.4037, "step": 7551 }, { "epoch": 0.91, "learning_rate": 2.4218247645689306e-07, "loss": 3.3675, "step": 7552 }, { "epoch": 0.91, "learning_rate": 2.415661502418765e-07, "loss": 3.334, "step": 7553 }, { "epoch": 0.91, "learning_rate": 2.409505898514869e-07, "loss": 3.3529, "step": 7554 }, { "epoch": 0.91, "learning_rate": 2.4033579538479234e-07, "loss": 3.3942, "step": 7555 }, { "epoch": 0.91, "learning_rate": 2.397217669407381e-07, "loss": 3.4792, "step": 7556 }, { "epoch": 0.91, "learning_rate": 2.391085046181468e-07, "loss": 3.321, "step": 7557 }, { "epoch": 0.91, "learning_rate": 2.3849600851571733e-07, "loss": 3.4239, "step": 7558 }, { "epoch": 0.91, "learning_rate": 2.3788427873202402e-07, "loss": 3.4277, "step": 7559 }, { "epoch": 0.91, "learning_rate": 2.372733153655199e-07, "loss": 3.3599, "step": 7560 }, { "epoch": 0.91, "learning_rate": 2.3666311851453348e-07, "loss": 3.2879, "step": 7561 }, { "epoch": 0.91, "learning_rate": 2.360536882772696e-07, "loss": 3.5435, "step": 7562 }, { "epoch": 0.91, "learning_rate": 2.3544502475181086e-07, "loss": 3.3923, "step": 7563 }, { "epoch": 0.91, "learning_rate": 2.3483712803611612e-07, "loss": 3.2845, "step": 7564 }, { "epoch": 0.91, "learning_rate": 2.3422999822801994e-07, "loss": 3.3488, "step": 7565 }, { "epoch": 0.91, "learning_rate": 2.3362363542523415e-07, "loss": 3.2871, "step": 7566 }, { "epoch": 0.91, "learning_rate": 2.330180397253473e-07, "loss": 3.3322, "step": 7567 }, { "epoch": 0.91, "learning_rate": 2.3241321122582428e-07, "loss": 3.3612, "step": 7568 }, { "epoch": 0.91, "learning_rate": 2.3180915002400605e-07, "loss": 3.3381, "step": 7569 }, { "epoch": 0.91, "learning_rate": 2.3120585621711156e-07, "loss": 3.4804, "step": 7570 }, { "epoch": 0.91, "learning_rate": 2.306033299022342e-07, "loss": 3.3593, "step": 7571 }, { "epoch": 0.91, "learning_rate": 2.300015711763448e-07, "loss": 3.3275, "step": 7572 }, { "epoch": 0.91, "learning_rate": 2.2940058013629085e-07, "loss": 3.3481, "step": 7573 }, { "epoch": 0.91, "learning_rate": 2.2880035687879663e-07, "loss": 3.4091, "step": 7574 }, { "epoch": 0.91, "learning_rate": 2.2820090150046159e-07, "loss": 3.4004, "step": 7575 }, { "epoch": 0.91, "learning_rate": 2.276022140977635e-07, "loss": 3.3455, "step": 7576 }, { "epoch": 0.91, "learning_rate": 2.270042947670542e-07, "loss": 3.422, "step": 7577 }, { "epoch": 0.91, "learning_rate": 2.264071436045634e-07, "loss": 3.421, "step": 7578 }, { "epoch": 0.91, "learning_rate": 2.2581076070639807e-07, "loss": 3.2941, "step": 7579 }, { "epoch": 0.91, "learning_rate": 2.2521514616853922e-07, "loss": 3.3612, "step": 7580 }, { "epoch": 0.91, "learning_rate": 2.2462030008684466e-07, "loss": 3.376, "step": 7581 }, { "epoch": 0.91, "learning_rate": 2.2402622255705054e-07, "loss": 3.3589, "step": 7582 }, { "epoch": 0.91, "learning_rate": 2.234329136747676e-07, "loss": 3.4068, "step": 7583 }, { "epoch": 0.91, "learning_rate": 2.2284037353548283e-07, "loss": 3.4581, "step": 7584 }, { "epoch": 0.91, "learning_rate": 2.2224860223456045e-07, "loss": 3.481, "step": 7585 }, { "epoch": 0.91, "learning_rate": 2.2165759986724044e-07, "loss": 3.3153, "step": 7586 }, { "epoch": 0.91, "learning_rate": 2.2106736652863835e-07, "loss": 3.5096, "step": 7587 }, { "epoch": 0.91, "learning_rate": 2.2047790231374766e-07, "loss": 3.4793, "step": 7588 }, { "epoch": 0.91, "learning_rate": 2.198892073174358e-07, "loss": 3.3534, "step": 7589 }, { "epoch": 0.91, "learning_rate": 2.1930128163444864e-07, "loss": 3.3427, "step": 7590 }, { "epoch": 0.91, "learning_rate": 2.1871412535940663e-07, "loss": 3.4419, "step": 7591 }, { "epoch": 0.91, "learning_rate": 2.181277385868069e-07, "loss": 3.4308, "step": 7592 }, { "epoch": 0.91, "learning_rate": 2.175421214110235e-07, "loss": 3.4016, "step": 7593 }, { "epoch": 0.91, "learning_rate": 2.169572739263054e-07, "loss": 3.2962, "step": 7594 }, { "epoch": 0.91, "learning_rate": 2.1637319622677843e-07, "loss": 3.3583, "step": 7595 }, { "epoch": 0.91, "learning_rate": 2.157898884064441e-07, "loss": 3.39, "step": 7596 }, { "epoch": 0.91, "learning_rate": 2.152073505591812e-07, "loss": 3.3821, "step": 7597 }, { "epoch": 0.91, "learning_rate": 2.1462558277874245e-07, "loss": 3.3926, "step": 7598 }, { "epoch": 0.91, "learning_rate": 2.1404458515875802e-07, "loss": 3.4457, "step": 7599 }, { "epoch": 0.91, "learning_rate": 2.1346435779273423e-07, "loss": 3.377, "step": 7600 }, { "epoch": 0.91, "learning_rate": 2.1288490077405298e-07, "loss": 3.3854, "step": 7601 }, { "epoch": 0.91, "learning_rate": 2.123062141959731e-07, "loss": 3.4132, "step": 7602 }, { "epoch": 0.91, "learning_rate": 2.117282981516272e-07, "loss": 3.4047, "step": 7603 }, { "epoch": 0.91, "learning_rate": 2.1115115273402708e-07, "loss": 3.4746, "step": 7604 }, { "epoch": 0.91, "learning_rate": 2.105747780360573e-07, "loss": 3.3719, "step": 7605 }, { "epoch": 0.91, "learning_rate": 2.0999917415048143e-07, "loss": 3.3471, "step": 7606 }, { "epoch": 0.91, "learning_rate": 2.0942434116993593e-07, "loss": 3.436, "step": 7607 }, { "epoch": 0.91, "learning_rate": 2.0885027918693512e-07, "loss": 3.3406, "step": 7608 }, { "epoch": 0.91, "learning_rate": 2.08276988293869e-07, "loss": 3.3941, "step": 7609 }, { "epoch": 0.91, "learning_rate": 2.0770446858300375e-07, "loss": 3.3738, "step": 7610 }, { "epoch": 0.91, "learning_rate": 2.071327201464801e-07, "loss": 3.3651, "step": 7611 }, { "epoch": 0.91, "learning_rate": 2.0656174307631615e-07, "loss": 3.4467, "step": 7612 }, { "epoch": 0.91, "learning_rate": 2.0599153746440447e-07, "loss": 3.4441, "step": 7613 }, { "epoch": 0.91, "learning_rate": 2.0542210340251556e-07, "loss": 3.2741, "step": 7614 }, { "epoch": 0.91, "learning_rate": 2.0485344098229386e-07, "loss": 3.439, "step": 7615 }, { "epoch": 0.91, "learning_rate": 2.0428555029525897e-07, "loss": 3.3539, "step": 7616 }, { "epoch": 0.91, "learning_rate": 2.037184314328089e-07, "loss": 3.4453, "step": 7617 }, { "epoch": 0.91, "learning_rate": 2.0315208448621505e-07, "loss": 3.4222, "step": 7618 }, { "epoch": 0.91, "learning_rate": 2.0258650954662674e-07, "loss": 3.3641, "step": 7619 }, { "epoch": 0.91, "learning_rate": 2.020217067050667e-07, "loss": 3.3412, "step": 7620 }, { "epoch": 0.91, "learning_rate": 2.0145767605243504e-07, "loss": 3.35, "step": 7621 }, { "epoch": 0.91, "learning_rate": 2.0089441767950746e-07, "loss": 3.3759, "step": 7622 }, { "epoch": 0.92, "learning_rate": 2.0033193167693477e-07, "loss": 3.2702, "step": 7623 }, { "epoch": 0.92, "learning_rate": 1.9977021813524344e-07, "loss": 3.3606, "step": 7624 }, { "epoch": 0.92, "learning_rate": 1.9920927714483617e-07, "loss": 3.3277, "step": 7625 }, { "epoch": 0.92, "learning_rate": 1.986491087959913e-07, "loss": 3.4216, "step": 7626 }, { "epoch": 0.92, "learning_rate": 1.9808971317886283e-07, "loss": 3.4016, "step": 7627 }, { "epoch": 0.92, "learning_rate": 1.9753109038347874e-07, "loss": 3.3946, "step": 7628 }, { "epoch": 0.92, "learning_rate": 1.9697324049974597e-07, "loss": 3.3904, "step": 7629 }, { "epoch": 0.92, "learning_rate": 1.964161636174433e-07, "loss": 3.411, "step": 7630 }, { "epoch": 0.92, "learning_rate": 1.9585985982622847e-07, "loss": 3.3161, "step": 7631 }, { "epoch": 0.92, "learning_rate": 1.9530432921563258e-07, "loss": 3.3238, "step": 7632 }, { "epoch": 0.92, "learning_rate": 1.947495718750636e-07, "loss": 3.4561, "step": 7633 }, { "epoch": 0.92, "learning_rate": 1.941955878938029e-07, "loss": 3.4066, "step": 7634 }, { "epoch": 0.92, "learning_rate": 1.9364237736101076e-07, "loss": 3.4168, "step": 7635 }, { "epoch": 0.92, "learning_rate": 1.9308994036571938e-07, "loss": 3.444, "step": 7636 }, { "epoch": 0.92, "learning_rate": 1.925382769968398e-07, "loss": 3.365, "step": 7637 }, { "epoch": 0.92, "learning_rate": 1.919873873431566e-07, "loss": 3.394, "step": 7638 }, { "epoch": 0.92, "learning_rate": 1.9143727149332935e-07, "loss": 3.4543, "step": 7639 }, { "epoch": 0.92, "learning_rate": 1.9088792953589508e-07, "loss": 3.3254, "step": 7640 }, { "epoch": 0.92, "learning_rate": 1.9033936155926468e-07, "loss": 3.4356, "step": 7641 }, { "epoch": 0.92, "learning_rate": 1.8979156765172536e-07, "loss": 3.2678, "step": 7642 }, { "epoch": 0.92, "learning_rate": 1.8924454790143821e-07, "loss": 3.4308, "step": 7643 }, { "epoch": 0.92, "learning_rate": 1.8869830239644172e-07, "loss": 3.5076, "step": 7644 }, { "epoch": 0.92, "learning_rate": 1.8815283122464833e-07, "loss": 3.427, "step": 7645 }, { "epoch": 0.92, "learning_rate": 1.8760813447384728e-07, "loss": 3.3633, "step": 7646 }, { "epoch": 0.92, "learning_rate": 1.870642122317018e-07, "loss": 3.3949, "step": 7647 }, { "epoch": 0.92, "learning_rate": 1.8652106458575126e-07, "loss": 3.3959, "step": 7648 }, { "epoch": 0.92, "learning_rate": 1.8597869162340965e-07, "loss": 3.3541, "step": 7649 }, { "epoch": 0.92, "learning_rate": 1.8543709343196715e-07, "loss": 3.3284, "step": 7650 }, { "epoch": 0.92, "learning_rate": 1.8489627009858901e-07, "loss": 3.3546, "step": 7651 }, { "epoch": 0.92, "learning_rate": 1.8435622171031507e-07, "loss": 3.3777, "step": 7652 }, { "epoch": 0.92, "learning_rate": 1.8381694835406084e-07, "loss": 3.4606, "step": 7653 }, { "epoch": 0.92, "learning_rate": 1.8327845011661793e-07, "loss": 3.4151, "step": 7654 }, { "epoch": 0.92, "learning_rate": 1.8274072708465207e-07, "loss": 3.4171, "step": 7655 }, { "epoch": 0.92, "learning_rate": 1.8220377934470456e-07, "loss": 3.4616, "step": 7656 }, { "epoch": 0.92, "learning_rate": 1.8166760698319242e-07, "loss": 3.4195, "step": 7657 }, { "epoch": 0.92, "learning_rate": 1.8113221008640713e-07, "loss": 3.4115, "step": 7658 }, { "epoch": 0.92, "learning_rate": 1.805975887405159e-07, "loss": 3.2315, "step": 7659 }, { "epoch": 0.92, "learning_rate": 1.8006374303156104e-07, "loss": 3.4039, "step": 7660 }, { "epoch": 0.92, "learning_rate": 1.7953067304545936e-07, "loss": 3.368, "step": 7661 }, { "epoch": 0.92, "learning_rate": 1.7899837886800442e-07, "loss": 3.3576, "step": 7662 }, { "epoch": 0.92, "learning_rate": 1.7846686058486328e-07, "loss": 3.3654, "step": 7663 }, { "epoch": 0.92, "learning_rate": 1.7793611828157865e-07, "loss": 3.4293, "step": 7664 }, { "epoch": 0.92, "learning_rate": 1.7740615204356827e-07, "loss": 3.4426, "step": 7665 }, { "epoch": 0.92, "learning_rate": 1.7687696195612613e-07, "loss": 3.4273, "step": 7666 }, { "epoch": 0.92, "learning_rate": 1.7634854810441914e-07, "loss": 3.3821, "step": 7667 }, { "epoch": 0.92, "learning_rate": 1.7582091057349148e-07, "loss": 3.3368, "step": 7668 }, { "epoch": 0.92, "learning_rate": 1.7529404944826135e-07, "loss": 3.4904, "step": 7669 }, { "epoch": 0.92, "learning_rate": 1.7476796481352088e-07, "loss": 3.4491, "step": 7670 }, { "epoch": 0.92, "learning_rate": 1.7424265675393903e-07, "loss": 3.4152, "step": 7671 }, { "epoch": 0.92, "learning_rate": 1.7371812535405985e-07, "loss": 3.3199, "step": 7672 }, { "epoch": 0.92, "learning_rate": 1.7319437069830026e-07, "loss": 3.3717, "step": 7673 }, { "epoch": 0.92, "learning_rate": 1.7267139287095503e-07, "loss": 3.41, "step": 7674 }, { "epoch": 0.92, "learning_rate": 1.721491919561913e-07, "loss": 3.4581, "step": 7675 }, { "epoch": 0.92, "learning_rate": 1.7162776803805347e-07, "loss": 3.276, "step": 7676 }, { "epoch": 0.92, "learning_rate": 1.711071212004589e-07, "loss": 3.4448, "step": 7677 }, { "epoch": 0.92, "learning_rate": 1.7058725152720112e-07, "loss": 3.3737, "step": 7678 }, { "epoch": 0.92, "learning_rate": 1.7006815910194818e-07, "loss": 3.4041, "step": 7679 }, { "epoch": 0.92, "learning_rate": 1.6954984400824216e-07, "loss": 3.375, "step": 7680 }, { "epoch": 0.92, "learning_rate": 1.690323063295024e-07, "loss": 3.4135, "step": 7681 }, { "epoch": 0.92, "learning_rate": 1.6851554614902067e-07, "loss": 3.324, "step": 7682 }, { "epoch": 0.92, "learning_rate": 1.6799956354996483e-07, "loss": 3.388, "step": 7683 }, { "epoch": 0.92, "learning_rate": 1.6748435861537793e-07, "loss": 3.4239, "step": 7684 }, { "epoch": 0.92, "learning_rate": 1.669699314281764e-07, "loss": 3.549, "step": 7685 }, { "epoch": 0.92, "learning_rate": 1.6645628207115295e-07, "loss": 3.4453, "step": 7686 }, { "epoch": 0.92, "learning_rate": 1.6594341062697416e-07, "loss": 3.377, "step": 7687 }, { "epoch": 0.92, "learning_rate": 1.6543131717818239e-07, "loss": 3.445, "step": 7688 }, { "epoch": 0.92, "learning_rate": 1.6492000180719392e-07, "loss": 3.4703, "step": 7689 }, { "epoch": 0.92, "learning_rate": 1.6440946459630015e-07, "loss": 3.4027, "step": 7690 }, { "epoch": 0.92, "learning_rate": 1.63899705627667e-07, "loss": 3.4298, "step": 7691 }, { "epoch": 0.92, "learning_rate": 1.6339072498333496e-07, "loss": 3.3994, "step": 7692 }, { "epoch": 0.92, "learning_rate": 1.6288252274522075e-07, "loss": 3.4397, "step": 7693 }, { "epoch": 0.92, "learning_rate": 1.6237509899511395e-07, "loss": 3.3586, "step": 7694 }, { "epoch": 0.92, "learning_rate": 1.618684538146792e-07, "loss": 3.3911, "step": 7695 }, { "epoch": 0.92, "learning_rate": 1.613625872854574e-07, "loss": 3.4509, "step": 7696 }, { "epoch": 0.92, "learning_rate": 1.6085749948886176e-07, "loss": 3.2966, "step": 7697 }, { "epoch": 0.92, "learning_rate": 1.603531905061817e-07, "loss": 3.3498, "step": 7698 }, { "epoch": 0.92, "learning_rate": 1.5984966041858119e-07, "loss": 3.4017, "step": 7699 }, { "epoch": 0.92, "learning_rate": 1.5934690930709817e-07, "loss": 3.2896, "step": 7700 }, { "epoch": 0.92, "learning_rate": 1.5884493725264626e-07, "loss": 3.3939, "step": 7701 }, { "epoch": 0.92, "learning_rate": 1.5834374433601252e-07, "loss": 3.3183, "step": 7702 }, { "epoch": 0.92, "learning_rate": 1.5784333063785962e-07, "loss": 3.4003, "step": 7703 }, { "epoch": 0.92, "learning_rate": 1.5734369623872426e-07, "loss": 3.2749, "step": 7704 }, { "epoch": 0.92, "learning_rate": 1.5684484121901766e-07, "loss": 3.3124, "step": 7705 }, { "epoch": 0.92, "learning_rate": 1.563467656590262e-07, "loss": 3.3168, "step": 7706 }, { "epoch": 0.93, "learning_rate": 1.5584946963890956e-07, "loss": 3.3831, "step": 7707 }, { "epoch": 0.93, "learning_rate": 1.553529532387038e-07, "loss": 3.2608, "step": 7708 }, { "epoch": 0.93, "learning_rate": 1.5485721653831775e-07, "loss": 3.3158, "step": 7709 }, { "epoch": 0.93, "learning_rate": 1.5436225961753538e-07, "loss": 3.4719, "step": 7710 }, { "epoch": 0.93, "learning_rate": 1.5386808255601627e-07, "loss": 3.3706, "step": 7711 }, { "epoch": 0.93, "learning_rate": 1.5337468543329293e-07, "loss": 3.4009, "step": 7712 }, { "epoch": 0.93, "learning_rate": 1.5288206832877294e-07, "loss": 3.4507, "step": 7713 }, { "epoch": 0.93, "learning_rate": 1.5239023132173847e-07, "loss": 3.3596, "step": 7714 }, { "epoch": 0.93, "learning_rate": 1.5189917449134563e-07, "loss": 3.4039, "step": 7715 }, { "epoch": 0.93, "learning_rate": 1.514088979166256e-07, "loss": 3.3524, "step": 7716 }, { "epoch": 0.93, "learning_rate": 1.5091940167648366e-07, "loss": 3.3906, "step": 7717 }, { "epoch": 0.93, "learning_rate": 1.504306858497001e-07, "loss": 3.358, "step": 7718 }, { "epoch": 0.93, "learning_rate": 1.4994275051492814e-07, "loss": 3.3431, "step": 7719 }, { "epoch": 0.93, "learning_rate": 1.4945559575069714e-07, "loss": 3.4832, "step": 7720 }, { "epoch": 0.93, "learning_rate": 1.4896922163540939e-07, "loss": 3.4736, "step": 7721 }, { "epoch": 0.93, "learning_rate": 1.484836282473423e-07, "loss": 3.4634, "step": 7722 }, { "epoch": 0.93, "learning_rate": 1.479988156646478e-07, "loss": 3.4185, "step": 7723 }, { "epoch": 0.93, "learning_rate": 1.475147839653518e-07, "loss": 3.3259, "step": 7724 }, { "epoch": 0.93, "learning_rate": 1.4703153322735475e-07, "loss": 3.4392, "step": 7725 }, { "epoch": 0.93, "learning_rate": 1.4654906352843057e-07, "loss": 3.436, "step": 7726 }, { "epoch": 0.93, "learning_rate": 1.460673749462288e-07, "loss": 3.4045, "step": 7727 }, { "epoch": 0.93, "learning_rate": 1.45586467558273e-07, "loss": 3.5171, "step": 7728 }, { "epoch": 0.93, "learning_rate": 1.4510634144195955e-07, "loss": 3.4297, "step": 7729 }, { "epoch": 0.93, "learning_rate": 1.446269966745606e-07, "loss": 3.4422, "step": 7730 }, { "epoch": 0.93, "learning_rate": 1.4414843333322214e-07, "loss": 3.3332, "step": 7731 }, { "epoch": 0.93, "learning_rate": 1.436706514949654e-07, "loss": 3.3475, "step": 7732 }, { "epoch": 0.93, "learning_rate": 1.4319365123668328e-07, "loss": 3.4786, "step": 7733 }, { "epoch": 0.93, "learning_rate": 1.4271743263514548e-07, "loss": 3.4716, "step": 7734 }, { "epoch": 0.93, "learning_rate": 1.422419957669935e-07, "loss": 3.4054, "step": 7735 }, { "epoch": 0.93, "learning_rate": 1.417673407087461e-07, "loss": 3.5327, "step": 7736 }, { "epoch": 0.93, "learning_rate": 1.412934675367933e-07, "loss": 3.4, "step": 7737 }, { "epoch": 0.93, "learning_rate": 1.4082037632740076e-07, "loss": 3.4524, "step": 7738 }, { "epoch": 0.93, "learning_rate": 1.4034806715670868e-07, "loss": 3.3606, "step": 7739 }, { "epoch": 0.93, "learning_rate": 1.398765401007296e-07, "loss": 3.4518, "step": 7740 }, { "epoch": 0.93, "learning_rate": 1.394057952353528e-07, "loss": 3.3432, "step": 7741 }, { "epoch": 0.93, "learning_rate": 1.3893583263633826e-07, "loss": 3.4186, "step": 7742 }, { "epoch": 0.93, "learning_rate": 1.3846665237932322e-07, "loss": 3.4281, "step": 7743 }, { "epoch": 0.93, "learning_rate": 1.3799825453981787e-07, "loss": 3.4252, "step": 7744 }, { "epoch": 0.93, "learning_rate": 1.3753063919320576e-07, "loss": 3.3246, "step": 7745 }, { "epoch": 0.93, "learning_rate": 1.3706380641474558e-07, "loss": 3.4678, "step": 7746 }, { "epoch": 0.93, "learning_rate": 1.3659775627956895e-07, "loss": 3.4446, "step": 7747 }, { "epoch": 0.93, "learning_rate": 1.3613248886268303e-07, "loss": 3.2128, "step": 7748 }, { "epoch": 0.93, "learning_rate": 1.35668004238968e-07, "loss": 3.4081, "step": 7749 }, { "epoch": 0.93, "learning_rate": 1.3520430248317839e-07, "loss": 3.3536, "step": 7750 }, { "epoch": 0.93, "learning_rate": 1.3474138366994238e-07, "loss": 3.4026, "step": 7751 }, { "epoch": 0.93, "learning_rate": 1.3427924787376145e-07, "loss": 3.3748, "step": 7752 }, { "epoch": 0.93, "learning_rate": 1.3381789516901333e-07, "loss": 3.3013, "step": 7753 }, { "epoch": 0.93, "learning_rate": 1.3335732562994752e-07, "loss": 3.5137, "step": 7754 }, { "epoch": 0.93, "learning_rate": 1.328975393306886e-07, "loss": 3.4151, "step": 7755 }, { "epoch": 0.93, "learning_rate": 1.3243853634523464e-07, "loss": 3.2885, "step": 7756 }, { "epoch": 0.93, "learning_rate": 1.3198031674745814e-07, "loss": 3.4251, "step": 7757 }, { "epoch": 0.93, "learning_rate": 1.3152288061110518e-07, "loss": 3.4324, "step": 7758 }, { "epoch": 0.93, "learning_rate": 1.310662280097952e-07, "loss": 3.4642, "step": 7759 }, { "epoch": 0.93, "learning_rate": 1.3061035901702211e-07, "loss": 3.3168, "step": 7760 }, { "epoch": 0.93, "learning_rate": 1.3015527370615456e-07, "loss": 3.22, "step": 7761 }, { "epoch": 0.93, "learning_rate": 1.2970097215043332e-07, "loss": 3.3324, "step": 7762 }, { "epoch": 0.93, "learning_rate": 1.2924745442297437e-07, "loss": 3.4048, "step": 7763 }, { "epoch": 0.93, "learning_rate": 1.287947205967671e-07, "loss": 3.3135, "step": 7764 }, { "epoch": 0.93, "learning_rate": 1.2834277074467438e-07, "loss": 3.4098, "step": 7765 }, { "epoch": 0.93, "learning_rate": 1.2789160493943353e-07, "loss": 3.4071, "step": 7766 }, { "epoch": 0.93, "learning_rate": 1.2744122325365481e-07, "loss": 3.4183, "step": 7767 }, { "epoch": 0.93, "learning_rate": 1.2699162575982415e-07, "loss": 3.3747, "step": 7768 }, { "epoch": 0.93, "learning_rate": 1.2654281253029866e-07, "loss": 3.467, "step": 7769 }, { "epoch": 0.93, "learning_rate": 1.2609478363731108e-07, "loss": 3.3331, "step": 7770 }, { "epoch": 0.93, "learning_rate": 1.2564753915296712e-07, "loss": 3.3831, "step": 7771 }, { "epoch": 0.93, "learning_rate": 1.2520107914924695e-07, "loss": 3.3257, "step": 7772 }, { "epoch": 0.93, "learning_rate": 1.2475540369800365e-07, "loss": 3.5212, "step": 7773 }, { "epoch": 0.93, "learning_rate": 1.2431051287096484e-07, "loss": 3.4516, "step": 7774 }, { "epoch": 0.93, "learning_rate": 1.2386640673973105e-07, "loss": 3.3728, "step": 7775 }, { "epoch": 0.93, "learning_rate": 1.234230853757773e-07, "loss": 3.4693, "step": 7776 }, { "epoch": 0.93, "learning_rate": 1.2298054885045208e-07, "loss": 3.472, "step": 7777 }, { "epoch": 0.93, "learning_rate": 1.225387972349773e-07, "loss": 3.3666, "step": 7778 }, { "epoch": 0.93, "learning_rate": 1.2209783060044776e-07, "loss": 3.3401, "step": 7779 }, { "epoch": 0.93, "learning_rate": 1.2165764901783395e-07, "loss": 3.2989, "step": 7780 }, { "epoch": 0.93, "learning_rate": 1.212182525579786e-07, "loss": 3.3825, "step": 7781 }, { "epoch": 0.93, "learning_rate": 1.2077964129159847e-07, "loss": 3.4113, "step": 7782 }, { "epoch": 0.93, "learning_rate": 1.203418152892838e-07, "loss": 3.3304, "step": 7783 }, { "epoch": 0.93, "learning_rate": 1.1990477462149818e-07, "loss": 3.4209, "step": 7784 }, { "epoch": 0.93, "learning_rate": 1.1946851935857927e-07, "loss": 3.3756, "step": 7785 }, { "epoch": 0.93, "learning_rate": 1.1903304957073868e-07, "loss": 3.4911, "step": 7786 }, { "epoch": 0.93, "learning_rate": 1.185983653280609e-07, "loss": 3.3178, "step": 7787 }, { "epoch": 0.93, "learning_rate": 1.1816446670050385e-07, "loss": 3.4359, "step": 7788 }, { "epoch": 0.93, "learning_rate": 1.1773135375790001e-07, "loss": 3.5134, "step": 7789 }, { "epoch": 0.94, "learning_rate": 1.1729902656995418e-07, "loss": 3.3531, "step": 7790 }, { "epoch": 0.94, "learning_rate": 1.1686748520624513e-07, "loss": 3.3581, "step": 7791 }, { "epoch": 0.94, "learning_rate": 1.1643672973622622e-07, "loss": 3.3981, "step": 7792 }, { "epoch": 0.94, "learning_rate": 1.1600676022922308e-07, "loss": 3.4321, "step": 7793 }, { "epoch": 0.94, "learning_rate": 1.1557757675443537e-07, "loss": 3.2745, "step": 7794 }, { "epoch": 0.94, "learning_rate": 1.1514917938093562e-07, "loss": 3.4383, "step": 7795 }, { "epoch": 0.94, "learning_rate": 1.1472156817767033e-07, "loss": 3.2791, "step": 7796 }, { "epoch": 0.94, "learning_rate": 1.1429474321346002e-07, "loss": 3.3992, "step": 7797 }, { "epoch": 0.94, "learning_rate": 1.138687045569975e-07, "loss": 3.4399, "step": 7798 }, { "epoch": 0.94, "learning_rate": 1.1344345227684961e-07, "loss": 3.2849, "step": 7799 }, { "epoch": 0.94, "learning_rate": 1.1301898644145715e-07, "loss": 3.435, "step": 7800 }, { "epoch": 0.94, "learning_rate": 1.125953071191338e-07, "loss": 3.4682, "step": 7801 }, { "epoch": 0.94, "learning_rate": 1.1217241437806614e-07, "loss": 3.2785, "step": 7802 }, { "epoch": 0.94, "learning_rate": 1.1175030828631528e-07, "loss": 3.312, "step": 7803 }, { "epoch": 0.94, "learning_rate": 1.113289889118152e-07, "loss": 3.3522, "step": 7804 }, { "epoch": 0.94, "learning_rate": 1.1090845632237279e-07, "loss": 3.3688, "step": 7805 }, { "epoch": 0.94, "learning_rate": 1.1048871058566945e-07, "loss": 3.3982, "step": 7806 }, { "epoch": 0.94, "learning_rate": 1.1006975176925838e-07, "loss": 3.3927, "step": 7807 }, { "epoch": 0.94, "learning_rate": 1.096515799405673e-07, "loss": 3.398, "step": 7808 }, { "epoch": 0.94, "learning_rate": 1.0923419516689793e-07, "loss": 3.4011, "step": 7809 }, { "epoch": 0.94, "learning_rate": 1.0881759751542264e-07, "loss": 3.4134, "step": 7810 }, { "epoch": 0.94, "learning_rate": 1.084017870531906e-07, "loss": 3.4524, "step": 7811 }, { "epoch": 0.94, "learning_rate": 1.079867638471216e-07, "loss": 3.3524, "step": 7812 }, { "epoch": 0.94, "learning_rate": 1.0757252796400997e-07, "loss": 3.4184, "step": 7813 }, { "epoch": 0.94, "learning_rate": 1.0715907947052296e-07, "loss": 3.3905, "step": 7814 }, { "epoch": 0.94, "learning_rate": 1.0674641843320066e-07, "loss": 3.437, "step": 7815 }, { "epoch": 0.94, "learning_rate": 1.0633454491845718e-07, "loss": 3.4221, "step": 7816 }, { "epoch": 0.94, "learning_rate": 1.0592345899258061e-07, "loss": 3.4088, "step": 7817 }, { "epoch": 0.94, "learning_rate": 1.0551316072173024e-07, "loss": 3.4037, "step": 7818 }, { "epoch": 0.94, "learning_rate": 1.051036501719399e-07, "loss": 3.3881, "step": 7819 }, { "epoch": 0.94, "learning_rate": 1.0469492740911636e-07, "loss": 3.4062, "step": 7820 }, { "epoch": 0.94, "learning_rate": 1.0428699249903973e-07, "loss": 3.4258, "step": 7821 }, { "epoch": 0.94, "learning_rate": 1.0387984550736362e-07, "loss": 3.4181, "step": 7822 }, { "epoch": 0.94, "learning_rate": 1.0347348649961397e-07, "loss": 3.4103, "step": 7823 }, { "epoch": 0.94, "learning_rate": 1.0306791554119066e-07, "loss": 3.4485, "step": 7824 }, { "epoch": 0.94, "learning_rate": 1.026631326973665e-07, "loss": 3.2627, "step": 7825 }, { "epoch": 0.94, "learning_rate": 1.0225913803328768e-07, "loss": 3.3798, "step": 7826 }, { "epoch": 0.94, "learning_rate": 1.0185593161397223e-07, "loss": 3.3896, "step": 7827 }, { "epoch": 0.94, "learning_rate": 1.0145351350431377e-07, "loss": 3.3939, "step": 7828 }, { "epoch": 0.94, "learning_rate": 1.010518837690766e-07, "loss": 3.3635, "step": 7829 }, { "epoch": 0.94, "learning_rate": 1.0065104247289958e-07, "loss": 3.5176, "step": 7830 }, { "epoch": 0.94, "learning_rate": 1.0025098968029445e-07, "loss": 3.3745, "step": 7831 }, { "epoch": 0.94, "learning_rate": 9.98517254556458e-08, "loss": 3.4242, "step": 7832 }, { "epoch": 0.94, "learning_rate": 9.945324986321114e-08, "loss": 3.3719, "step": 7833 }, { "epoch": 0.94, "learning_rate": 9.905556296712193e-08, "loss": 3.4068, "step": 7834 }, { "epoch": 0.94, "learning_rate": 9.865866483138143e-08, "loss": 3.4104, "step": 7835 }, { "epoch": 0.94, "learning_rate": 9.826255551986685e-08, "loss": 3.3988, "step": 7836 }, { "epoch": 0.94, "learning_rate": 9.786723509632834e-08, "loss": 3.3935, "step": 7837 }, { "epoch": 0.94, "learning_rate": 9.747270362438943e-08, "loss": 3.4145, "step": 7838 }, { "epoch": 0.94, "learning_rate": 9.70789611675449e-08, "loss": 3.4418, "step": 7839 }, { "epoch": 0.94, "learning_rate": 9.668600778916515e-08, "loss": 3.4379, "step": 7840 }, { "epoch": 0.94, "learning_rate": 9.629384355249183e-08, "loss": 3.4305, "step": 7841 }, { "epoch": 0.94, "learning_rate": 9.590246852063945e-08, "loss": 3.4467, "step": 7842 }, { "epoch": 0.94, "learning_rate": 9.551188275659651e-08, "loss": 3.277, "step": 7843 }, { "epoch": 0.94, "learning_rate": 9.512208632322439e-08, "loss": 3.3625, "step": 7844 }, { "epoch": 0.94, "learning_rate": 9.473307928325682e-08, "loss": 3.4354, "step": 7845 }, { "epoch": 0.94, "learning_rate": 9.43448616993009e-08, "loss": 3.4382, "step": 7846 }, { "epoch": 0.94, "learning_rate": 9.39574336338367e-08, "loss": 3.4312, "step": 7847 }, { "epoch": 0.94, "learning_rate": 9.357079514921652e-08, "loss": 3.3889, "step": 7848 }, { "epoch": 0.94, "learning_rate": 9.318494630766727e-08, "loss": 3.4086, "step": 7849 }, { "epoch": 0.94, "learning_rate": 9.279988717128652e-08, "loss": 3.4176, "step": 7850 }, { "epoch": 0.94, "learning_rate": 9.241561780204578e-08, "loss": 3.3947, "step": 7851 }, { "epoch": 0.94, "learning_rate": 9.203213826179002e-08, "loss": 3.3412, "step": 7852 }, { "epoch": 0.94, "learning_rate": 9.164944861223712e-08, "loss": 3.409, "step": 7853 }, { "epoch": 0.94, "learning_rate": 9.126754891497613e-08, "loss": 3.4482, "step": 7854 }, { "epoch": 0.94, "learning_rate": 9.088643923147122e-08, "loss": 3.5351, "step": 7855 }, { "epoch": 0.94, "learning_rate": 9.050611962305778e-08, "loss": 3.3596, "step": 7856 }, { "epoch": 0.94, "learning_rate": 9.012659015094515e-08, "loss": 3.5071, "step": 7857 }, { "epoch": 0.94, "learning_rate": 8.974785087621507e-08, "loss": 3.3894, "step": 7858 }, { "epoch": 0.94, "learning_rate": 8.936990185982097e-08, "loss": 3.3204, "step": 7859 }, { "epoch": 0.94, "learning_rate": 8.899274316259087e-08, "loss": 3.217, "step": 7860 }, { "epoch": 0.94, "learning_rate": 8.861637484522511e-08, "loss": 3.3993, "step": 7861 }, { "epoch": 0.94, "learning_rate": 8.824079696829634e-08, "loss": 3.3529, "step": 7862 }, { "epoch": 0.94, "learning_rate": 8.786600959224956e-08, "loss": 3.3484, "step": 7863 }, { "epoch": 0.94, "learning_rate": 8.749201277740482e-08, "loss": 3.413, "step": 7864 }, { "epoch": 0.94, "learning_rate": 8.711880658395178e-08, "loss": 3.3984, "step": 7865 }, { "epoch": 0.94, "learning_rate": 8.674639107195515e-08, "loss": 3.3361, "step": 7866 }, { "epoch": 0.94, "learning_rate": 8.6374766301352e-08, "loss": 3.4209, "step": 7867 }, { "epoch": 0.94, "learning_rate": 8.60039323319506e-08, "loss": 3.4455, "step": 7868 }, { "epoch": 0.94, "learning_rate": 8.56338892234343e-08, "loss": 3.328, "step": 7869 }, { "epoch": 0.94, "learning_rate": 8.526463703535715e-08, "loss": 3.439, "step": 7870 }, { "epoch": 0.94, "learning_rate": 8.48961758271477e-08, "loss": 3.4317, "step": 7871 }, { "epoch": 0.94, "learning_rate": 8.452850565810578e-08, "loss": 3.345, "step": 7872 }, { "epoch": 0.95, "learning_rate": 8.416162658740457e-08, "loss": 3.4755, "step": 7873 }, { "epoch": 0.95, "learning_rate": 8.379553867408908e-08, "loss": 3.4903, "step": 7874 }, { "epoch": 0.95, "learning_rate": 8.343024197707883e-08, "loss": 3.3765, "step": 7875 }, { "epoch": 0.95, "learning_rate": 8.306573655516403e-08, "loss": 3.3726, "step": 7876 }, { "epoch": 0.95, "learning_rate": 8.270202246700832e-08, "loss": 3.5326, "step": 7877 }, { "epoch": 0.95, "learning_rate": 8.233909977114873e-08, "loss": 3.4599, "step": 7878 }, { "epoch": 0.95, "learning_rate": 8.197696852599302e-08, "loss": 3.3579, "step": 7879 }, { "epoch": 0.95, "learning_rate": 8.161562878982399e-08, "loss": 3.3816, "step": 7880 }, { "epoch": 0.95, "learning_rate": 8.125508062079513e-08, "loss": 3.3484, "step": 7881 }, { "epoch": 0.95, "learning_rate": 8.08953240769339e-08, "loss": 3.3326, "step": 7882 }, { "epoch": 0.95, "learning_rate": 8.053635921613845e-08, "loss": 3.4473, "step": 7883 }, { "epoch": 0.95, "learning_rate": 8.0178186096182e-08, "loss": 3.3801, "step": 7884 }, { "epoch": 0.95, "learning_rate": 7.982080477470899e-08, "loss": 3.3322, "step": 7885 }, { "epoch": 0.95, "learning_rate": 7.946421530923565e-08, "loss": 3.4322, "step": 7886 }, { "epoch": 0.95, "learning_rate": 7.910841775715216e-08, "loss": 3.4239, "step": 7887 }, { "epoch": 0.95, "learning_rate": 7.875341217572108e-08, "loss": 3.4417, "step": 7888 }, { "epoch": 0.95, "learning_rate": 7.839919862207668e-08, "loss": 3.4351, "step": 7889 }, { "epoch": 0.95, "learning_rate": 7.804577715322669e-08, "loss": 3.4689, "step": 7890 }, { "epoch": 0.95, "learning_rate": 7.769314782605064e-08, "loss": 3.3965, "step": 7891 }, { "epoch": 0.95, "learning_rate": 7.734131069730089e-08, "loss": 3.3488, "step": 7892 }, { "epoch": 0.95, "learning_rate": 7.699026582360213e-08, "loss": 3.2916, "step": 7893 }, { "epoch": 0.95, "learning_rate": 7.664001326145255e-08, "loss": 3.332, "step": 7894 }, { "epoch": 0.95, "learning_rate": 7.629055306722033e-08, "loss": 3.3419, "step": 7895 }, { "epoch": 0.95, "learning_rate": 7.594188529714941e-08, "loss": 3.4404, "step": 7896 }, { "epoch": 0.95, "learning_rate": 7.559401000735379e-08, "loss": 3.3514, "step": 7897 }, { "epoch": 0.95, "learning_rate": 7.524692725382032e-08, "loss": 3.4658, "step": 7898 }, { "epoch": 0.95, "learning_rate": 7.490063709240936e-08, "loss": 3.406, "step": 7899 }, { "epoch": 0.95, "learning_rate": 7.455513957885297e-08, "loss": 3.3451, "step": 7900 }, { "epoch": 0.95, "learning_rate": 7.4210434768755e-08, "loss": 3.4905, "step": 7901 }, { "epoch": 0.95, "learning_rate": 7.386652271759331e-08, "loss": 3.3494, "step": 7902 }, { "epoch": 0.95, "learning_rate": 7.352340348071696e-08, "loss": 3.3753, "step": 7903 }, { "epoch": 0.95, "learning_rate": 7.318107711334732e-08, "loss": 3.46, "step": 7904 }, { "epoch": 0.95, "learning_rate": 7.283954367057867e-08, "loss": 3.3471, "step": 7905 }, { "epoch": 0.95, "learning_rate": 7.249880320737812e-08, "loss": 3.3727, "step": 7906 }, { "epoch": 0.95, "learning_rate": 7.215885577858406e-08, "loss": 3.2827, "step": 7907 }, { "epoch": 0.95, "learning_rate": 7.18197014389077e-08, "loss": 3.4053, "step": 7908 }, { "epoch": 0.95, "learning_rate": 7.148134024293318e-08, "loss": 3.4169, "step": 7909 }, { "epoch": 0.95, "learning_rate": 7.114377224511638e-08, "loss": 3.3928, "step": 7910 }, { "epoch": 0.95, "learning_rate": 7.080699749978547e-08, "loss": 3.3711, "step": 7911 }, { "epoch": 0.95, "learning_rate": 7.047101606114159e-08, "loss": 3.4643, "step": 7912 }, { "epoch": 0.95, "learning_rate": 7.013582798325702e-08, "loss": 3.4548, "step": 7913 }, { "epoch": 0.95, "learning_rate": 6.98014333200775e-08, "loss": 3.3538, "step": 7914 }, { "epoch": 0.95, "learning_rate": 6.946783212542108e-08, "loss": 3.4433, "step": 7915 }, { "epoch": 0.95, "learning_rate": 6.913502445297649e-08, "loss": 3.3293, "step": 7916 }, { "epoch": 0.95, "learning_rate": 6.880301035630698e-08, "loss": 3.3169, "step": 7917 }, { "epoch": 0.95, "learning_rate": 6.847178988884707e-08, "loss": 3.3402, "step": 7918 }, { "epoch": 0.95, "learning_rate": 6.814136310390296e-08, "loss": 3.3728, "step": 7919 }, { "epoch": 0.95, "learning_rate": 6.781173005465436e-08, "loss": 3.3913, "step": 7920 }, { "epoch": 0.95, "learning_rate": 6.74828907941516e-08, "loss": 3.3971, "step": 7921 }, { "epoch": 0.95, "learning_rate": 6.715484537531957e-08, "loss": 3.4792, "step": 7922 }, { "epoch": 0.95, "learning_rate": 6.682759385095273e-08, "loss": 3.4496, "step": 7923 }, { "epoch": 0.95, "learning_rate": 6.650113627372002e-08, "loss": 3.3309, "step": 7924 }, { "epoch": 0.95, "learning_rate": 6.617547269616109e-08, "loss": 3.3533, "step": 7925 }, { "epoch": 0.95, "learning_rate": 6.585060317068902e-08, "loss": 3.3497, "step": 7926 }, { "epoch": 0.95, "learning_rate": 6.552652774958813e-08, "loss": 3.5209, "step": 7927 }, { "epoch": 0.95, "learning_rate": 6.520324648501497e-08, "loss": 3.3548, "step": 7928 }, { "epoch": 0.95, "learning_rate": 6.488075942899908e-08, "loss": 3.3779, "step": 7929 }, { "epoch": 0.95, "learning_rate": 6.45590666334417e-08, "loss": 3.3271, "step": 7930 }, { "epoch": 0.95, "learning_rate": 6.42381681501153e-08, "loss": 3.5416, "step": 7931 }, { "epoch": 0.95, "learning_rate": 6.391806403066692e-08, "loss": 3.4351, "step": 7932 }, { "epoch": 0.95, "learning_rate": 6.359875432661255e-08, "loss": 3.3857, "step": 7933 }, { "epoch": 0.95, "learning_rate": 6.328023908934389e-08, "loss": 3.3922, "step": 7934 }, { "epoch": 0.95, "learning_rate": 6.296251837012157e-08, "loss": 3.319, "step": 7935 }, { "epoch": 0.95, "learning_rate": 6.26455922200797e-08, "loss": 3.3685, "step": 7936 }, { "epoch": 0.95, "learning_rate": 6.232946069022583e-08, "loss": 3.3587, "step": 7937 }, { "epoch": 0.95, "learning_rate": 6.201412383143702e-08, "loss": 3.3911, "step": 7938 }, { "epoch": 0.95, "learning_rate": 6.169958169446433e-08, "loss": 3.4339, "step": 7939 }, { "epoch": 0.95, "learning_rate": 6.138583432993006e-08, "loss": 3.3686, "step": 7940 }, { "epoch": 0.95, "learning_rate": 6.107288178832938e-08, "loss": 3.4575, "step": 7941 }, { "epoch": 0.95, "learning_rate": 6.076072412002809e-08, "loss": 3.293, "step": 7942 }, { "epoch": 0.95, "learning_rate": 6.044936137526602e-08, "loss": 3.3739, "step": 7943 }, { "epoch": 0.95, "learning_rate": 6.013879360415309e-08, "loss": 3.5128, "step": 7944 }, { "epoch": 0.95, "learning_rate": 5.982902085667264e-08, "loss": 3.4146, "step": 7945 }, { "epoch": 0.95, "learning_rate": 5.952004318268034e-08, "loss": 3.389, "step": 7946 }, { "epoch": 0.95, "learning_rate": 5.921186063190199e-08, "loss": 3.3229, "step": 7947 }, { "epoch": 0.95, "learning_rate": 5.890447325393733e-08, "loss": 3.4952, "step": 7948 }, { "epoch": 0.95, "learning_rate": 5.8597881098257924e-08, "loss": 3.4029, "step": 7949 }, { "epoch": 0.95, "learning_rate": 5.8292084214205405e-08, "loss": 3.435, "step": 7950 }, { "epoch": 0.95, "learning_rate": 5.7987082650995954e-08, "loss": 3.3634, "step": 7951 }, { "epoch": 0.95, "learning_rate": 5.7682876457716416e-08, "loss": 3.4505, "step": 7952 }, { "epoch": 0.95, "learning_rate": 5.7379465683325976e-08, "loss": 3.4722, "step": 7953 }, { "epoch": 0.95, "learning_rate": 5.7076850376655554e-08, "loss": 3.2677, "step": 7954 }, { "epoch": 0.95, "learning_rate": 5.677503058640788e-08, "loss": 3.3519, "step": 7955 }, { "epoch": 0.95, "learning_rate": 5.6474006361159075e-08, "loss": 3.4944, "step": 7956 }, { "epoch": 0.96, "learning_rate": 5.6173777749354844e-08, "loss": 3.4635, "step": 7957 }, { "epoch": 0.96, "learning_rate": 5.587434479931542e-08, "loss": 3.4153, "step": 7958 }, { "epoch": 0.96, "learning_rate": 5.5575707559230584e-08, "loss": 3.4316, "step": 7959 }, { "epoch": 0.96, "learning_rate": 5.527786607716357e-08, "loss": 3.4503, "step": 7960 }, { "epoch": 0.96, "learning_rate": 5.4980820401048796e-08, "loss": 3.3834, "step": 7961 }, { "epoch": 0.96, "learning_rate": 5.468457057869358e-08, "loss": 3.4117, "step": 7962 }, { "epoch": 0.96, "learning_rate": 5.438911665777591e-08, "loss": 3.3137, "step": 7963 }, { "epoch": 0.96, "learning_rate": 5.4094458685846616e-08, "loss": 3.3246, "step": 7964 }, { "epoch": 0.96, "learning_rate": 5.380059671032833e-08, "loss": 3.4081, "step": 7965 }, { "epoch": 0.96, "learning_rate": 5.3507530778514335e-08, "loss": 3.4682, "step": 7966 }, { "epoch": 0.96, "learning_rate": 5.321526093757245e-08, "loss": 3.3638, "step": 7967 }, { "epoch": 0.96, "learning_rate": 5.292378723453895e-08, "loss": 3.4603, "step": 7968 }, { "epoch": 0.96, "learning_rate": 5.263310971632518e-08, "loss": 3.3919, "step": 7969 }, { "epoch": 0.96, "learning_rate": 5.234322842971207e-08, "loss": 3.3085, "step": 7970 }, { "epoch": 0.96, "learning_rate": 5.2054143421353396e-08, "loss": 3.5007, "step": 7971 }, { "epoch": 0.96, "learning_rate": 5.176585473777473e-08, "loss": 3.467, "step": 7972 }, { "epoch": 0.96, "learning_rate": 5.1478362425372855e-08, "loss": 3.5294, "step": 7973 }, { "epoch": 0.96, "learning_rate": 5.119166653041796e-08, "loss": 3.4542, "step": 7974 }, { "epoch": 0.96, "learning_rate": 5.090576709905093e-08, "loss": 3.4596, "step": 7975 }, { "epoch": 0.96, "learning_rate": 5.062066417728384e-08, "loss": 3.4383, "step": 7976 }, { "epoch": 0.96, "learning_rate": 5.0336357811001654e-08, "loss": 3.4268, "step": 7977 }, { "epoch": 0.96, "learning_rate": 5.005284804596055e-08, "loss": 3.4506, "step": 7978 }, { "epoch": 0.96, "learning_rate": 4.977013492778904e-08, "loss": 3.3973, "step": 7979 }, { "epoch": 0.96, "learning_rate": 4.948821850198682e-08, "loss": 3.4592, "step": 7980 }, { "epoch": 0.96, "learning_rate": 4.920709881392538e-08, "loss": 3.3924, "step": 7981 }, { "epoch": 0.96, "learning_rate": 4.8926775908849646e-08, "loss": 3.3781, "step": 7982 }, { "epoch": 0.96, "learning_rate": 4.8647249831872965e-08, "loss": 3.452, "step": 7983 }, { "epoch": 0.96, "learning_rate": 4.836852062798381e-08, "loss": 3.4108, "step": 7984 }, { "epoch": 0.96, "learning_rate": 4.809058834204072e-08, "loss": 3.3438, "step": 7985 }, { "epoch": 0.96, "learning_rate": 4.781345301877405e-08, "loss": 3.3127, "step": 7986 }, { "epoch": 0.96, "learning_rate": 4.753711470278588e-08, "loss": 3.4409, "step": 7987 }, { "epoch": 0.96, "learning_rate": 4.726157343855064e-08, "loss": 3.3463, "step": 7988 }, { "epoch": 0.96, "learning_rate": 4.698682927041398e-08, "loss": 3.2832, "step": 7989 }, { "epoch": 0.96, "learning_rate": 4.671288224259329e-08, "loss": 3.4438, "step": 7990 }, { "epoch": 0.96, "learning_rate": 4.643973239917776e-08, "loss": 3.4769, "step": 7991 }, { "epoch": 0.96, "learning_rate": 4.616737978412833e-08, "loss": 3.3999, "step": 7992 }, { "epoch": 0.96, "learning_rate": 4.589582444127716e-08, "loss": 3.4934, "step": 7993 }, { "epoch": 0.96, "learning_rate": 4.562506641432929e-08, "loss": 3.4657, "step": 7994 }, { "epoch": 0.96, "learning_rate": 4.535510574685986e-08, "loss": 3.4479, "step": 7995 }, { "epoch": 0.96, "learning_rate": 4.5085942482316904e-08, "loss": 3.3707, "step": 7996 }, { "epoch": 0.96, "learning_rate": 4.481757666401909e-08, "loss": 3.316, "step": 7997 }, { "epoch": 0.96, "learning_rate": 4.4550008335158544e-08, "loss": 3.5205, "step": 7998 }, { "epoch": 0.96, "learning_rate": 4.4283237538796374e-08, "loss": 3.4136, "step": 7999 }, { "epoch": 0.96, "learning_rate": 4.401726431786824e-08, "loss": 3.4202, "step": 8000 }, { "epoch": 0.96, "learning_rate": 4.375208871517878e-08, "loss": 3.3562, "step": 8001 }, { "epoch": 0.96, "learning_rate": 4.3487710773406076e-08, "loss": 3.3917, "step": 8002 }, { "epoch": 0.96, "learning_rate": 4.322413053509944e-08, "loss": 3.3548, "step": 8003 }, { "epoch": 0.96, "learning_rate": 4.296134804267937e-08, "loss": 3.3402, "step": 8004 }, { "epoch": 0.96, "learning_rate": 4.26993633384376e-08, "loss": 3.3098, "step": 8005 }, { "epoch": 0.96, "learning_rate": 4.2438176464539294e-08, "loss": 3.3324, "step": 8006 }, { "epoch": 0.96, "learning_rate": 4.2177787463018593e-08, "loss": 3.3944, "step": 8007 }, { "epoch": 0.96, "learning_rate": 4.19181963757842e-08, "loss": 3.3536, "step": 8008 }, { "epoch": 0.96, "learning_rate": 4.1659403244613264e-08, "loss": 3.415, "step": 8009 }, { "epoch": 0.96, "learning_rate": 4.140140811115745e-08, "loss": 3.3573, "step": 8010 }, { "epoch": 0.96, "learning_rate": 4.114421101693744e-08, "loss": 3.4025, "step": 8011 }, { "epoch": 0.96, "learning_rate": 4.088781200334735e-08, "loss": 3.3365, "step": 8012 }, { "epoch": 0.96, "learning_rate": 4.0632211111652496e-08, "loss": 3.3339, "step": 8013 }, { "epoch": 0.96, "learning_rate": 4.037740838298831e-08, "loss": 3.449, "step": 8014 }, { "epoch": 0.96, "learning_rate": 4.012340385836366e-08, "loss": 3.4583, "step": 8015 }, { "epoch": 0.96, "learning_rate": 3.987019757865807e-08, "loss": 3.4223, "step": 8016 }, { "epoch": 0.96, "learning_rate": 3.961778958462226e-08, "loss": 3.3967, "step": 8017 }, { "epoch": 0.96, "learning_rate": 3.93661799168793e-08, "loss": 3.3405, "step": 8018 }, { "epoch": 0.96, "learning_rate": 3.911536861592346e-08, "loss": 3.4351, "step": 8019 }, { "epoch": 0.96, "learning_rate": 3.8865355722120246e-08, "loss": 3.2648, "step": 8020 }, { "epoch": 0.96, "learning_rate": 3.86161412757069e-08, "loss": 3.416, "step": 8021 }, { "epoch": 0.96, "learning_rate": 3.8367725316791895e-08, "loss": 3.4453, "step": 8022 }, { "epoch": 0.96, "learning_rate": 3.8120107885355475e-08, "loss": 3.3913, "step": 8023 }, { "epoch": 0.96, "learning_rate": 3.787328902124965e-08, "loss": 3.3123, "step": 8024 }, { "epoch": 0.96, "learning_rate": 3.762726876419709e-08, "loss": 3.3167, "step": 8025 }, { "epoch": 0.96, "learning_rate": 3.738204715379279e-08, "loss": 3.3928, "step": 8026 }, { "epoch": 0.96, "learning_rate": 3.713762422950295e-08, "loss": 3.3991, "step": 8027 }, { "epoch": 0.96, "learning_rate": 3.6894000030665014e-08, "loss": 3.31, "step": 8028 }, { "epoch": 0.96, "learning_rate": 3.6651174596487594e-08, "loss": 3.3693, "step": 8029 }, { "epoch": 0.96, "learning_rate": 3.640914796605166e-08, "loss": 3.3705, "step": 8030 }, { "epoch": 0.96, "learning_rate": 3.616792017830828e-08, "loss": 3.4281, "step": 8031 }, { "epoch": 0.96, "learning_rate": 3.592749127208139e-08, "loss": 3.3279, "step": 8032 }, { "epoch": 0.96, "learning_rate": 3.5687861286065605e-08, "loss": 3.3988, "step": 8033 }, { "epoch": 0.96, "learning_rate": 3.544903025882729e-08, "loss": 3.345, "step": 8034 }, { "epoch": 0.96, "learning_rate": 3.521099822880403e-08, "loss": 3.4229, "step": 8035 }, { "epoch": 0.96, "learning_rate": 3.4973765234304627e-08, "loss": 3.3851, "step": 8036 }, { "epoch": 0.96, "learning_rate": 3.4737331313509094e-08, "loss": 3.4351, "step": 8037 }, { "epoch": 0.96, "learning_rate": 3.450169650446977e-08, "loss": 3.457, "step": 8038 }, { "epoch": 0.96, "learning_rate": 3.4266860845110216e-08, "loss": 3.348, "step": 8039 }, { "epoch": 0.97, "learning_rate": 3.4032824373224086e-08, "loss": 3.4365, "step": 8040 }, { "epoch": 0.97, "learning_rate": 3.3799587126477354e-08, "loss": 3.4631, "step": 8041 }, { "epoch": 0.97, "learning_rate": 3.356714914240833e-08, "loss": 3.3744, "step": 8042 }, { "epoch": 0.97, "learning_rate": 3.3335510458424314e-08, "loss": 3.3914, "step": 8043 }, { "epoch": 0.97, "learning_rate": 3.31046711118066e-08, "loss": 3.3648, "step": 8044 }, { "epoch": 0.97, "learning_rate": 3.2874631139706014e-08, "loss": 3.2618, "step": 8045 }, { "epoch": 0.97, "learning_rate": 3.2645390579145176e-08, "loss": 3.2996, "step": 8046 }, { "epoch": 0.97, "learning_rate": 3.2416949467018457e-08, "loss": 3.2466, "step": 8047 }, { "epoch": 0.97, "learning_rate": 3.2189307840091446e-08, "loss": 3.3335, "step": 8048 }, { "epoch": 0.97, "learning_rate": 3.196246573500039e-08, "loss": 3.4039, "step": 8049 }, { "epoch": 0.97, "learning_rate": 3.173642318825332e-08, "loss": 3.3577, "step": 8050 }, { "epoch": 0.97, "learning_rate": 3.1511180236229457e-08, "loss": 3.3701, "step": 8051 }, { "epoch": 0.97, "learning_rate": 3.128673691518091e-08, "loss": 3.3664, "step": 8052 }, { "epoch": 0.97, "learning_rate": 3.106309326122825e-08, "loss": 3.4446, "step": 8053 }, { "epoch": 0.97, "learning_rate": 3.0840249310365445e-08, "loss": 3.3445, "step": 8054 }, { "epoch": 0.97, "learning_rate": 3.0618205098457144e-08, "loss": 3.4385, "step": 8055 }, { "epoch": 0.97, "learning_rate": 3.039696066123865e-08, "loss": 3.2859, "step": 8056 }, { "epoch": 0.97, "learning_rate": 3.017651603431759e-08, "loss": 3.2823, "step": 8057 }, { "epoch": 0.97, "learning_rate": 2.995687125317282e-08, "loss": 3.4627, "step": 8058 }, { "epoch": 0.97, "learning_rate": 2.973802635315326e-08, "loss": 3.3415, "step": 8059 }, { "epoch": 0.97, "learning_rate": 2.9519981369480198e-08, "loss": 3.4297, "step": 8060 }, { "epoch": 0.97, "learning_rate": 2.9302736337246118e-08, "loss": 3.4116, "step": 8061 }, { "epoch": 0.97, "learning_rate": 2.9086291291414714e-08, "loss": 3.3401, "step": 8062 }, { "epoch": 0.97, "learning_rate": 2.8870646266820345e-08, "loss": 3.4281, "step": 8063 }, { "epoch": 0.97, "learning_rate": 2.8655801298169138e-08, "loss": 3.3345, "step": 8064 }, { "epoch": 0.97, "learning_rate": 2.844175642003788e-08, "loss": 3.4169, "step": 8065 }, { "epoch": 0.97, "learning_rate": 2.8228511666876236e-08, "loss": 3.3655, "step": 8066 }, { "epoch": 0.97, "learning_rate": 2.801606707300286e-08, "loss": 3.3419, "step": 8067 }, { "epoch": 0.97, "learning_rate": 2.7804422672608744e-08, "loss": 3.4178, "step": 8068 }, { "epoch": 0.97, "learning_rate": 2.7593578499756635e-08, "loss": 3.3573, "step": 8069 }, { "epoch": 0.97, "learning_rate": 2.7383534588379945e-08, "loss": 3.3505, "step": 8070 }, { "epoch": 0.97, "learning_rate": 2.7174290972282193e-08, "loss": 3.4009, "step": 8071 }, { "epoch": 0.97, "learning_rate": 2.6965847685140323e-08, "loss": 3.3406, "step": 8072 }, { "epoch": 0.97, "learning_rate": 2.6758204760500282e-08, "loss": 3.4784, "step": 8073 }, { "epoch": 0.97, "learning_rate": 2.655136223178145e-08, "loss": 3.3887, "step": 8074 }, { "epoch": 0.97, "learning_rate": 2.63453201322722e-08, "loss": 3.3816, "step": 8075 }, { "epoch": 0.97, "learning_rate": 2.6140078495133225e-08, "loss": 3.3004, "step": 8076 }, { "epoch": 0.97, "learning_rate": 2.5935637353396437e-08, "loss": 3.406, "step": 8077 }, { "epoch": 0.97, "learning_rate": 2.5731996739964403e-08, "loss": 3.467, "step": 8078 }, { "epoch": 0.97, "learning_rate": 2.5529156687611467e-08, "loss": 3.3355, "step": 8079 }, { "epoch": 0.97, "learning_rate": 2.5327117228982066e-08, "loss": 3.3615, "step": 8080 }, { "epoch": 0.97, "learning_rate": 2.512587839659353e-08, "loss": 3.4621, "step": 8081 }, { "epoch": 0.97, "learning_rate": 2.4925440222833273e-08, "loss": 3.3829, "step": 8082 }, { "epoch": 0.97, "learning_rate": 2.4725802739958836e-08, "loss": 3.396, "step": 8083 }, { "epoch": 0.97, "learning_rate": 2.4526965980101182e-08, "loss": 3.4463, "step": 8084 }, { "epoch": 0.97, "learning_rate": 2.4328929975260262e-08, "loss": 3.322, "step": 8085 }, { "epoch": 0.97, "learning_rate": 2.413169475730892e-08, "loss": 3.2861, "step": 8086 }, { "epoch": 0.97, "learning_rate": 2.3935260357989544e-08, "loss": 3.4009, "step": 8087 }, { "epoch": 0.97, "learning_rate": 2.3739626808916838e-08, "loss": 3.5164, "step": 8088 }, { "epoch": 0.97, "learning_rate": 2.3544794141576178e-08, "loss": 3.4437, "step": 8089 }, { "epoch": 0.97, "learning_rate": 2.3350762387323588e-08, "loss": 3.3845, "step": 8090 }, { "epoch": 0.97, "learning_rate": 2.315753157738687e-08, "loss": 3.3764, "step": 8091 }, { "epoch": 0.97, "learning_rate": 2.2965101742865037e-08, "loss": 3.3564, "step": 8092 }, { "epoch": 0.97, "learning_rate": 2.2773472914727757e-08, "loss": 3.4925, "step": 8093 }, { "epoch": 0.97, "learning_rate": 2.2582645123815362e-08, "loss": 3.388, "step": 8094 }, { "epoch": 0.97, "learning_rate": 2.239261840083995e-08, "loss": 3.3813, "step": 8095 }, { "epoch": 0.97, "learning_rate": 2.220339277638428e-08, "loss": 3.412, "step": 8096 }, { "epoch": 0.97, "learning_rate": 2.2014968280903437e-08, "loss": 3.351, "step": 8097 }, { "epoch": 0.97, "learning_rate": 2.1827344944721497e-08, "loss": 3.5161, "step": 8098 }, { "epoch": 0.97, "learning_rate": 2.164052279803486e-08, "loss": 3.346, "step": 8099 }, { "epoch": 0.97, "learning_rate": 2.1454501870911136e-08, "loss": 3.4165, "step": 8100 }, { "epoch": 0.97, "learning_rate": 2.1269282193288608e-08, "loss": 3.3253, "step": 8101 }, { "epoch": 0.97, "learning_rate": 2.1084863794976207e-08, "loss": 3.3304, "step": 8102 }, { "epoch": 0.97, "learning_rate": 2.090124670565463e-08, "loss": 3.294, "step": 8103 }, { "epoch": 0.97, "learning_rate": 2.0718430954875245e-08, "loss": 3.374, "step": 8104 }, { "epoch": 0.97, "learning_rate": 2.0536416572060625e-08, "loss": 3.5442, "step": 8105 }, { "epoch": 0.97, "learning_rate": 2.0355203586503448e-08, "loss": 3.3154, "step": 8106 }, { "epoch": 0.97, "learning_rate": 2.0174792027369273e-08, "loss": 3.2844, "step": 8107 }, { "epoch": 0.97, "learning_rate": 1.9995181923693207e-08, "loss": 3.3255, "step": 8108 }, { "epoch": 0.97, "learning_rate": 1.981637330438213e-08, "loss": 3.4013, "step": 8109 }, { "epoch": 0.97, "learning_rate": 1.9638366198212467e-08, "loss": 3.4501, "step": 8110 }, { "epoch": 0.97, "learning_rate": 1.946116063383463e-08, "loss": 3.2402, "step": 8111 }, { "epoch": 0.97, "learning_rate": 1.9284756639766366e-08, "loss": 3.4105, "step": 8112 }, { "epoch": 0.97, "learning_rate": 1.9109154244398852e-08, "loss": 3.3699, "step": 8113 }, { "epoch": 0.97, "learning_rate": 1.8934353475993926e-08, "loss": 3.3746, "step": 8114 }, { "epoch": 0.97, "learning_rate": 1.8760354362684087e-08, "loss": 3.331, "step": 8115 }, { "epoch": 0.97, "learning_rate": 1.858715693247193e-08, "loss": 3.4157, "step": 8116 }, { "epoch": 0.97, "learning_rate": 1.841476121323349e-08, "loss": 3.348, "step": 8117 }, { "epoch": 0.97, "learning_rate": 1.8243167232712687e-08, "loss": 3.309, "step": 8118 }, { "epoch": 0.97, "learning_rate": 1.807237501852688e-08, "loss": 3.3349, "step": 8119 }, { "epoch": 0.97, "learning_rate": 1.7902384598163514e-08, "loss": 3.3902, "step": 8120 }, { "epoch": 0.97, "learning_rate": 1.7733195998980156e-08, "loss": 3.349, "step": 8121 }, { "epoch": 0.97, "learning_rate": 1.7564809248206694e-08, "loss": 3.4434, "step": 8122 }, { "epoch": 0.98, "learning_rate": 1.7397224372943666e-08, "loss": 3.4487, "step": 8123 }, { "epoch": 0.98, "learning_rate": 1.723044140016117e-08, "loss": 3.3682, "step": 8124 }, { "epoch": 0.98, "learning_rate": 1.706446035670217e-08, "loss": 3.3405, "step": 8125 }, { "epoch": 0.98, "learning_rate": 1.6899281269279756e-08, "loss": 3.3807, "step": 8126 }, { "epoch": 0.98, "learning_rate": 1.673490416447765e-08, "loss": 3.3744, "step": 8127 }, { "epoch": 0.98, "learning_rate": 1.6571329068750808e-08, "loss": 3.5008, "step": 8128 }, { "epoch": 0.98, "learning_rate": 1.6408556008425393e-08, "loss": 3.4127, "step": 8129 }, { "epoch": 0.98, "learning_rate": 1.6246585009697667e-08, "loss": 3.4533, "step": 8130 }, { "epoch": 0.98, "learning_rate": 1.6085416098635675e-08, "loss": 3.3926, "step": 8131 }, { "epoch": 0.98, "learning_rate": 1.5925049301177554e-08, "loss": 3.3157, "step": 8132 }, { "epoch": 0.98, "learning_rate": 1.5765484643133765e-08, "loss": 3.4641, "step": 8133 }, { "epoch": 0.98, "learning_rate": 1.5606722150183772e-08, "loss": 3.4206, "step": 8134 }, { "epoch": 0.98, "learning_rate": 1.5448761847879356e-08, "loss": 3.4647, "step": 8135 }, { "epoch": 0.98, "learning_rate": 1.5291603761642402e-08, "loss": 3.4311, "step": 8136 }, { "epoch": 0.98, "learning_rate": 1.513524791676657e-08, "loss": 3.48, "step": 8137 }, { "epoch": 0.98, "learning_rate": 1.4979694338415617e-08, "loss": 3.3767, "step": 8138 }, { "epoch": 0.98, "learning_rate": 1.4824943051623964e-08, "loss": 3.3967, "step": 8139 }, { "epoch": 0.98, "learning_rate": 1.4670994081297796e-08, "loss": 3.4128, "step": 8140 }, { "epoch": 0.98, "learning_rate": 1.4517847452213408e-08, "loss": 3.4259, "step": 8141 }, { "epoch": 0.98, "learning_rate": 1.436550318901886e-08, "loss": 3.4524, "step": 8142 }, { "epoch": 0.98, "learning_rate": 1.4213961316232316e-08, "loss": 3.369, "step": 8143 }, { "epoch": 0.98, "learning_rate": 1.406322185824205e-08, "loss": 3.4452, "step": 8144 }, { "epoch": 0.98, "learning_rate": 1.391328483930976e-08, "loss": 3.4115, "step": 8145 }, { "epoch": 0.98, "learning_rate": 1.3764150283565036e-08, "loss": 3.3467, "step": 8146 }, { "epoch": 0.98, "learning_rate": 1.3615818215010346e-08, "loss": 3.4041, "step": 8147 }, { "epoch": 0.98, "learning_rate": 1.3468288657518258e-08, "loss": 3.3978, "step": 8148 }, { "epoch": 0.98, "learning_rate": 1.3321561634832003e-08, "loss": 3.3237, "step": 8149 }, { "epoch": 0.98, "learning_rate": 1.3175637170566025e-08, "loss": 3.4753, "step": 8150 }, { "epoch": 0.98, "learning_rate": 1.303051528820598e-08, "loss": 3.2608, "step": 8151 }, { "epoch": 0.98, "learning_rate": 1.2886196011107077e-08, "loss": 3.3913, "step": 8152 }, { "epoch": 0.98, "learning_rate": 1.274267936249629e-08, "loss": 3.3717, "step": 8153 }, { "epoch": 0.98, "learning_rate": 1.2599965365471255e-08, "loss": 3.3926, "step": 8154 }, { "epoch": 0.98, "learning_rate": 1.2458054043000822e-08, "loss": 3.4149, "step": 8155 }, { "epoch": 0.98, "learning_rate": 1.2316945417924497e-08, "loss": 3.2782, "step": 8156 }, { "epoch": 0.98, "learning_rate": 1.2176639512951339e-08, "loss": 3.4139, "step": 8157 }, { "epoch": 0.98, "learning_rate": 1.2037136350662726e-08, "loss": 3.3149, "step": 8158 }, { "epoch": 0.98, "learning_rate": 1.1898435953510701e-08, "loss": 3.338, "step": 8159 }, { "epoch": 0.98, "learning_rate": 1.1760538343817407e-08, "loss": 3.415, "step": 8160 }, { "epoch": 0.98, "learning_rate": 1.1623443543776203e-08, "loss": 3.3725, "step": 8161 }, { "epoch": 0.98, "learning_rate": 1.1487151575451105e-08, "loss": 3.3776, "step": 8162 }, { "epoch": 0.98, "learning_rate": 1.1351662460777347e-08, "loss": 3.4486, "step": 8163 }, { "epoch": 0.98, "learning_rate": 1.121697622156026e-08, "loss": 3.4417, "step": 8164 }, { "epoch": 0.98, "learning_rate": 1.1083092879476953e-08, "loss": 3.3098, "step": 8165 }, { "epoch": 0.98, "learning_rate": 1.0950012456073523e-08, "loss": 3.4616, "step": 8166 }, { "epoch": 0.98, "learning_rate": 1.0817734972768946e-08, "loss": 3.4564, "step": 8167 }, { "epoch": 0.98, "learning_rate": 1.0686260450851748e-08, "loss": 3.391, "step": 8168 }, { "epoch": 0.98, "learning_rate": 1.0555588911481674e-08, "loss": 3.4716, "step": 8169 }, { "epoch": 0.98, "learning_rate": 1.0425720375688563e-08, "loss": 3.409, "step": 8170 }, { "epoch": 0.98, "learning_rate": 1.0296654864374034e-08, "loss": 3.3152, "step": 8171 }, { "epoch": 0.98, "learning_rate": 1.01683923983098e-08, "loss": 3.3566, "step": 8172 }, { "epoch": 0.98, "learning_rate": 1.0040932998138242e-08, "loss": 3.3613, "step": 8173 }, { "epoch": 0.98, "learning_rate": 9.914276684372948e-09, "loss": 3.345, "step": 8174 }, { "epoch": 0.98, "learning_rate": 9.788423477398167e-09, "loss": 3.3989, "step": 8175 }, { "epoch": 0.98, "learning_rate": 9.663373397468811e-09, "loss": 3.3192, "step": 8176 }, { "epoch": 0.98, "learning_rate": 9.539126464709892e-09, "loss": 3.4998, "step": 8177 }, { "epoch": 0.98, "learning_rate": 9.415682699118189e-09, "loss": 3.3988, "step": 8178 }, { "epoch": 0.98, "learning_rate": 9.293042120561147e-09, "loss": 3.4059, "step": 8179 }, { "epoch": 0.98, "learning_rate": 9.171204748776308e-09, "loss": 3.2931, "step": 8180 }, { "epoch": 0.98, "learning_rate": 9.050170603371877e-09, "loss": 3.3767, "step": 8181 }, { "epoch": 0.98, "learning_rate": 8.929939703827828e-09, "loss": 3.3273, "step": 8182 }, { "epoch": 0.98, "learning_rate": 8.810512069493682e-09, "loss": 3.3831, "step": 8183 }, { "epoch": 0.98, "learning_rate": 8.691887719590175e-09, "loss": 3.3902, "step": 8184 }, { "epoch": 0.98, "learning_rate": 8.574066673209259e-09, "loss": 3.3754, "step": 8185 }, { "epoch": 0.98, "learning_rate": 8.45704894931243e-09, "loss": 3.3926, "step": 8186 }, { "epoch": 0.98, "learning_rate": 8.340834566732958e-09, "loss": 3.4778, "step": 8187 }, { "epoch": 0.98, "learning_rate": 8.22542354417477e-09, "loss": 3.2916, "step": 8188 }, { "epoch": 0.98, "learning_rate": 8.110815900211344e-09, "loss": 3.3385, "step": 8189 }, { "epoch": 0.98, "learning_rate": 7.997011653288478e-09, "loss": 3.3545, "step": 8190 }, { "epoch": 0.98, "learning_rate": 7.884010821720967e-09, "loss": 3.3698, "step": 8191 }, { "epoch": 0.98, "learning_rate": 7.771813423696483e-09, "loss": 3.4842, "step": 8192 }, { "epoch": 0.98, "learning_rate": 7.660419477271142e-09, "loss": 3.3594, "step": 8193 }, { "epoch": 0.98, "learning_rate": 7.549829000372822e-09, "loss": 3.3587, "step": 8194 }, { "epoch": 0.98, "learning_rate": 7.440042010800619e-09, "loss": 3.4348, "step": 8195 }, { "epoch": 0.98, "learning_rate": 7.331058526223178e-09, "loss": 3.4433, "step": 8196 }, { "epoch": 0.98, "learning_rate": 7.222878564180358e-09, "loss": 3.51, "step": 8197 }, { "epoch": 0.98, "learning_rate": 7.1155021420826755e-09, "loss": 3.3741, "step": 8198 }, { "epoch": 0.98, "learning_rate": 7.008929277211862e-09, "loss": 3.4457, "step": 8199 }, { "epoch": 0.98, "learning_rate": 6.903159986719199e-09, "loss": 3.3638, "step": 8200 }, { "epoch": 0.98, "learning_rate": 6.798194287627735e-09, "loss": 3.3941, "step": 8201 }, { "epoch": 0.98, "learning_rate": 6.694032196830624e-09, "loss": 3.4513, "step": 8202 }, { "epoch": 0.98, "learning_rate": 6.590673731091124e-09, "loss": 3.3761, "step": 8203 }, { "epoch": 0.98, "learning_rate": 6.488118907044261e-09, "loss": 3.4216, "step": 8204 }, { "epoch": 0.98, "learning_rate": 6.386367741195165e-09, "loss": 3.3682, "step": 8205 }, { "epoch": 0.98, "learning_rate": 6.2854202499201825e-09, "loss": 3.3762, "step": 8206 }, { "epoch": 0.99, "learning_rate": 6.1852764494652055e-09, "loss": 3.3736, "step": 8207 }, { "epoch": 0.99, "learning_rate": 6.085936355947897e-09, "loss": 3.5037, "step": 8208 }, { "epoch": 0.99, "learning_rate": 5.987399985356024e-09, "loss": 3.4064, "step": 8209 }, { "epoch": 0.99, "learning_rate": 5.889667353547457e-09, "loss": 3.3702, "step": 8210 }, { "epoch": 0.99, "learning_rate": 5.792738476252391e-09, "loss": 3.3293, "step": 8211 }, { "epoch": 0.99, "learning_rate": 5.696613369069459e-09, "loss": 3.2764, "step": 8212 }, { "epoch": 0.99, "learning_rate": 5.601292047470175e-09, "loss": 3.3735, "step": 8213 }, { "epoch": 0.99, "learning_rate": 5.5067745267950446e-09, "loss": 3.4316, "step": 8214 }, { "epoch": 0.99, "learning_rate": 5.413060822255234e-09, "loss": 3.3683, "step": 8215 }, { "epoch": 0.99, "learning_rate": 5.320150948934233e-09, "loss": 3.3857, "step": 8216 }, { "epoch": 0.99, "learning_rate": 5.22804492178397e-09, "loss": 3.3419, "step": 8217 }, { "epoch": 0.99, "learning_rate": 5.136742755628698e-09, "loss": 3.3743, "step": 8218 }, { "epoch": 0.99, "learning_rate": 5.046244465162775e-09, "loss": 3.3804, "step": 8219 }, { "epoch": 0.99, "learning_rate": 4.956550064950105e-09, "loss": 3.3766, "step": 8220 }, { "epoch": 0.99, "learning_rate": 4.867659569426919e-09, "loss": 3.4565, "step": 8221 }, { "epoch": 0.99, "learning_rate": 4.779572992898995e-09, "loss": 3.4215, "step": 8222 }, { "epoch": 0.99, "learning_rate": 4.692290349543327e-09, "loss": 3.2623, "step": 8223 }, { "epoch": 0.99, "learning_rate": 4.605811653407011e-09, "loss": 3.3919, "step": 8224 }, { "epoch": 0.99, "learning_rate": 4.520136918407803e-09, "loss": 3.3068, "step": 8225 }, { "epoch": 0.99, "learning_rate": 4.435266158334672e-09, "loss": 3.2998, "step": 8226 }, { "epoch": 0.99, "learning_rate": 4.3511993868461386e-09, "loss": 3.3207, "step": 8227 }, { "epoch": 0.99, "learning_rate": 4.26793661747249e-09, "loss": 3.3898, "step": 8228 }, { "epoch": 0.99, "learning_rate": 4.185477863614118e-09, "loss": 3.3601, "step": 8229 }, { "epoch": 0.99, "learning_rate": 4.103823138541518e-09, "loss": 3.2415, "step": 8230 }, { "epoch": 0.99, "learning_rate": 4.022972455396401e-09, "loss": 3.4764, "step": 8231 }, { "epoch": 0.99, "learning_rate": 3.942925827191135e-09, "loss": 3.3197, "step": 8232 }, { "epoch": 0.99, "learning_rate": 3.863683266808193e-09, "loss": 3.3589, "step": 8233 }, { "epoch": 0.99, "learning_rate": 3.7852447870012636e-09, "loss": 3.3552, "step": 8234 }, { "epoch": 0.99, "learning_rate": 3.707610400394135e-09, "loss": 3.4441, "step": 8235 }, { "epoch": 0.99, "learning_rate": 3.6307801194807037e-09, "loss": 3.3748, "step": 8236 }, { "epoch": 0.99, "learning_rate": 3.5547539566271882e-09, "loss": 3.3752, "step": 8237 }, { "epoch": 0.99, "learning_rate": 3.479531924068802e-09, "loss": 3.3766, "step": 8238 }, { "epoch": 0.99, "learning_rate": 3.405114033910861e-09, "loss": 3.3896, "step": 8239 }, { "epoch": 0.99, "learning_rate": 3.3315002981315624e-09, "loss": 3.3502, "step": 8240 }, { "epoch": 0.99, "learning_rate": 3.2586907285780957e-09, "loss": 3.3379, "step": 8241 }, { "epoch": 0.99, "learning_rate": 3.186685336967754e-09, "loss": 3.3609, "step": 8242 }, { "epoch": 0.99, "learning_rate": 3.1154841348896016e-09, "loss": 3.3782, "step": 8243 }, { "epoch": 0.99, "learning_rate": 3.045087133802804e-09, "loss": 3.3918, "step": 8244 }, { "epoch": 0.99, "learning_rate": 2.9754943450371886e-09, "loss": 3.3708, "step": 8245 }, { "epoch": 0.99, "learning_rate": 2.9067057797932396e-09, "loss": 3.4096, "step": 8246 }, { "epoch": 0.99, "learning_rate": 2.838721449140991e-09, "loss": 3.4434, "step": 8247 }, { "epoch": 0.99, "learning_rate": 2.7715413640222455e-09, "loss": 3.4203, "step": 8248 }, { "epoch": 0.99, "learning_rate": 2.7051655352494654e-09, "loss": 3.4142, "step": 8249 }, { "epoch": 0.99, "learning_rate": 2.6395939735046618e-09, "loss": 3.2863, "step": 8250 }, { "epoch": 0.99, "learning_rate": 2.574826689341614e-09, "loss": 3.4649, "step": 8251 }, { "epoch": 0.99, "learning_rate": 2.5108636931830967e-09, "loss": 3.3748, "step": 8252 }, { "epoch": 0.99, "learning_rate": 2.447704995323652e-09, "loss": 3.3656, "step": 8253 }, { "epoch": 0.99, "learning_rate": 2.3853506059284825e-09, "loss": 3.4346, "step": 8254 }, { "epoch": 0.99, "learning_rate": 2.323800535032894e-09, "loss": 3.4024, "step": 8255 }, { "epoch": 0.99, "learning_rate": 2.263054792542296e-09, "loss": 3.3598, "step": 8256 }, { "epoch": 0.99, "learning_rate": 2.2031133882333133e-09, "loss": 3.3274, "step": 8257 }, { "epoch": 0.99, "learning_rate": 2.143976331753228e-09, "loss": 3.4325, "step": 8258 }, { "epoch": 0.99, "learning_rate": 2.0856436326194276e-09, "loss": 3.382, "step": 8259 }, { "epoch": 0.99, "learning_rate": 2.0281153002199572e-09, "loss": 3.3625, "step": 8260 }, { "epoch": 0.99, "learning_rate": 1.9713913438135226e-09, "loss": 3.3537, "step": 8261 }, { "epoch": 0.99, "learning_rate": 1.915471772528932e-09, "loss": 3.3687, "step": 8262 }, { "epoch": 0.99, "learning_rate": 1.8603565953662084e-09, "loss": 3.4041, "step": 8263 }, { "epoch": 0.99, "learning_rate": 1.806045821196034e-09, "loss": 3.3339, "step": 8264 }, { "epoch": 0.99, "learning_rate": 1.7525394587586398e-09, "loss": 3.4059, "step": 8265 }, { "epoch": 0.99, "learning_rate": 1.69983751666547e-09, "loss": 3.3033, "step": 8266 }, { "epoch": 0.99, "learning_rate": 1.647940003398074e-09, "loss": 3.4102, "step": 8267 }, { "epoch": 0.99, "learning_rate": 1.5968469273097697e-09, "loss": 3.3993, "step": 8268 }, { "epoch": 0.99, "learning_rate": 1.5465582966228686e-09, "loss": 3.4398, "step": 8269 }, { "epoch": 0.99, "learning_rate": 1.4970741194308969e-09, "loss": 3.3934, "step": 8270 }, { "epoch": 0.99, "learning_rate": 1.448394403698039e-09, "loss": 3.3969, "step": 8271 }, { "epoch": 0.99, "learning_rate": 1.400519157258584e-09, "loss": 3.3263, "step": 8272 }, { "epoch": 0.99, "learning_rate": 1.3534483878174798e-09, "loss": 3.3415, "step": 8273 }, { "epoch": 0.99, "learning_rate": 1.307182102950888e-09, "loss": 3.4274, "step": 8274 }, { "epoch": 0.99, "learning_rate": 1.2617203101045194e-09, "loss": 3.443, "step": 8275 }, { "epoch": 0.99, "learning_rate": 1.2170630165952991e-09, "loss": 3.4041, "step": 8276 }, { "epoch": 0.99, "learning_rate": 1.1732102296102554e-09, "loss": 3.4474, "step": 8277 }, { "epoch": 0.99, "learning_rate": 1.1301619562065214e-09, "loss": 3.3857, "step": 8278 }, { "epoch": 0.99, "learning_rate": 1.0879182033135538e-09, "loss": 3.2741, "step": 8279 }, { "epoch": 0.99, "learning_rate": 1.0464789777286932e-09, "loss": 3.4212, "step": 8280 }, { "epoch": 0.99, "learning_rate": 1.0058442861221595e-09, "loss": 3.4523, "step": 8281 }, { "epoch": 0.99, "learning_rate": 9.660141350331664e-10, "loss": 3.3412, "step": 8282 }, { "epoch": 0.99, "learning_rate": 9.269885308721416e-10, "loss": 3.3826, "step": 8283 }, { "epoch": 0.99, "learning_rate": 8.887674799201718e-10, "loss": 3.3728, "step": 8284 }, { "epoch": 0.99, "learning_rate": 8.513509883284476e-10, "loss": 3.4724, "step": 8285 }, { "epoch": 0.99, "learning_rate": 8.147390621182638e-10, "loss": 3.4831, "step": 8286 }, { "epoch": 0.99, "learning_rate": 7.78931707182684e-10, "loss": 3.3732, "step": 8287 }, { "epoch": 0.99, "learning_rate": 7.439289292843211e-10, "loss": 3.2838, "step": 8288 }, { "epoch": 0.99, "learning_rate": 7.097307340564463e-10, "loss": 3.3462, "step": 8289 }, { "epoch": 1.0, "learning_rate": 6.763371270035457e-10, "loss": 3.3885, "step": 8290 }, { "epoch": 1.0, "learning_rate": 6.437481134985435e-10, "loss": 3.3067, "step": 8291 }, { "epoch": 1.0, "learning_rate": 6.119636987877986e-10, "loss": 3.3677, "step": 8292 }, { "epoch": 1.0, "learning_rate": 5.809838879861085e-10, "loss": 3.4079, "step": 8293 }, { "epoch": 1.0, "learning_rate": 5.508086860794848e-10, "loss": 3.4211, "step": 8294 }, { "epoch": 1.0, "learning_rate": 5.214380979240429e-10, "loss": 3.3851, "step": 8295 }, { "epoch": 1.0, "learning_rate": 4.928721282465576e-10, "loss": 3.3709, "step": 8296 }, { "epoch": 1.0, "learning_rate": 4.6511078164557245e-10, "loss": 3.3171, "step": 8297 }, { "epoch": 1.0, "learning_rate": 4.381540625880698e-10, "loss": 3.3572, "step": 8298 }, { "epoch": 1.0, "learning_rate": 4.1200197541224617e-10, "loss": 3.421, "step": 8299 }, { "epoch": 1.0, "learning_rate": 3.866545243280673e-10, "loss": 3.4571, "step": 8300 }, { "epoch": 1.0, "learning_rate": 3.6211171341393736e-10, "loss": 3.4198, "step": 8301 }, { "epoch": 1.0, "learning_rate": 3.3837354662003e-10, "loss": 3.3386, "step": 8302 }, { "epoch": 1.0, "learning_rate": 3.1544002776717765e-10, "loss": 3.4129, "step": 8303 }, { "epoch": 1.0, "learning_rate": 2.933111605463168e-10, "loss": 3.4451, "step": 8304 }, { "epoch": 1.0, "learning_rate": 2.7198694851848785e-10, "loss": 3.4454, "step": 8305 }, { "epoch": 1.0, "learning_rate": 2.514673951153901e-10, "loss": 3.4271, "step": 8306 }, { "epoch": 1.0, "learning_rate": 2.3175250364049218e-10, "loss": 3.3732, "step": 8307 }, { "epoch": 1.0, "learning_rate": 2.1284227726570127e-10, "loss": 3.4123, "step": 8308 }, { "epoch": 1.0, "learning_rate": 1.9473671903469383e-10, "loss": 3.486, "step": 8309 }, { "epoch": 1.0, "learning_rate": 1.7743583186180524e-10, "loss": 3.3873, "step": 8310 }, { "epoch": 1.0, "learning_rate": 1.6093961853091976e-10, "loss": 3.368, "step": 8311 }, { "epoch": 1.0, "learning_rate": 1.452480816971358e-10, "loss": 3.4454, "step": 8312 }, { "epoch": 1.0, "learning_rate": 1.3036122388565553e-10, "loss": 3.3385, "step": 8313 }, { "epoch": 1.0, "learning_rate": 1.1627904749289543e-10, "loss": 3.3052, "step": 8314 }, { "epoch": 1.0, "learning_rate": 1.0300155478537577e-10, "loss": 3.4294, "step": 8315 }, { "epoch": 1.0, "learning_rate": 9.052874789861054e-11, "loss": 3.3243, "step": 8316 }, { "epoch": 1.0, "learning_rate": 7.886062884154833e-11, "loss": 3.4379, "step": 8317 }, { "epoch": 1.0, "learning_rate": 6.799719949157624e-11, "loss": 3.4003, "step": 8318 }, { "epoch": 1.0, "learning_rate": 5.7938461596740436e-11, "loss": 3.495, "step": 8319 }, { "epoch": 1.0, "learning_rate": 4.868441677574609e-11, "loss": 3.3836, "step": 8320 }, { "epoch": 1.0, "learning_rate": 4.02350665185125e-11, "loss": 3.3985, "step": 8321 }, { "epoch": 1.0, "learning_rate": 3.259041218506287e-11, "loss": 3.383, "step": 8322 }, { "epoch": 1.0, "learning_rate": 2.57504550049692e-11, "loss": 3.3982, "step": 8323 }, { "epoch": 1.0, "learning_rate": 1.971519607957273e-11, "loss": 3.3003, "step": 8324 }, { "epoch": 1.0, "learning_rate": 1.4484636379763496e-11, "loss": 3.4248, "step": 8325 }, { "epoch": 1.0, "learning_rate": 1.005877674764566e-11, "loss": 3.3941, "step": 8326 }, { "epoch": 1.0, "learning_rate": 6.437617895427295e-12, "loss": 3.3718, "step": 8327 }, { "epoch": 1.0, "learning_rate": 3.621160405975488e-12, "loss": 3.4042, "step": 8328 }, { "epoch": 1.0, "learning_rate": 1.6094047328163443e-12, "loss": 3.4233, "step": 8329 }, { "epoch": 1.0, "learning_rate": 4.023511995798757e-13, "loss": 3.4596, "step": 8330 }, { "epoch": 1.0, "learning_rate": 0.0, "loss": 3.3813, "step": 8331 }, { "epoch": 1.0, "step": 8331, "total_flos": 1.0740697008470753e+18, "train_loss": 3.412663069653949, "train_runtime": 9270.6529, "train_samples_per_second": 57.507, "train_steps_per_second": 0.899 } ], "max_steps": 8331, "num_train_epochs": 1, "total_flos": 1.0740697008470753e+18, "trial_name": null, "trial_params": null }