{ "best_metric": 0.04723460599780083, "best_model_checkpoint": "email-spam-detection-distilbert/checkpoint-25035", "epoch": 3.0, "eval_steps": 1, "global_step": 25035, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 6.000000000000001e-08, "loss": 0.6443, "step": 1 }, { "epoch": 0.0, "learning_rate": 1.2000000000000002e-07, "loss": 0.7478, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.8e-07, "loss": 0.7152, "step": 3 }, { "epoch": 0.0, "learning_rate": 2.4000000000000003e-07, "loss": 0.6659, "step": 4 }, { "epoch": 0.0, "learning_rate": 3.0000000000000004e-07, "loss": 0.6099, "step": 5 }, { "epoch": 0.0, "learning_rate": 3.6e-07, "loss": 0.6864, "step": 6 }, { "epoch": 0.0, "learning_rate": 4.2000000000000006e-07, "loss": 0.7346, "step": 7 }, { "epoch": 0.0, "learning_rate": 4.800000000000001e-07, "loss": 0.6942, "step": 8 }, { "epoch": 0.0, "learning_rate": 5.4e-07, "loss": 0.6638, "step": 9 }, { "epoch": 0.0, "learning_rate": 6.000000000000001e-07, "loss": 0.6748, "step": 10 }, { "epoch": 0.0, "learning_rate": 6.6e-07, "loss": 0.6946, "step": 11 }, { "epoch": 0.0, "learning_rate": 7.2e-07, "loss": 0.7562, "step": 12 }, { "epoch": 0.0, "learning_rate": 7.8e-07, "loss": 0.8251, "step": 13 }, { "epoch": 0.0, "learning_rate": 8.400000000000001e-07, "loss": 0.6736, "step": 14 }, { "epoch": 0.0, "learning_rate": 9e-07, "loss": 0.7176, "step": 15 }, { "epoch": 0.0, "learning_rate": 9.600000000000001e-07, "loss": 0.7866, "step": 16 }, { "epoch": 0.0, "learning_rate": 1.0200000000000002e-06, "loss": 0.6449, "step": 17 }, { "epoch": 0.0, "learning_rate": 1.08e-06, "loss": 0.6987, "step": 18 }, { "epoch": 0.0, "learning_rate": 1.14e-06, "loss": 0.6969, "step": 19 }, { "epoch": 0.0, "learning_rate": 1.2000000000000002e-06, "loss": 0.7843, "step": 20 }, { "epoch": 0.0, "learning_rate": 1.26e-06, "loss": 0.7418, "step": 21 }, { "epoch": 0.0, "learning_rate": 1.32e-06, "loss": 0.6686, "step": 22 }, { "epoch": 0.0, "learning_rate": 1.3800000000000001e-06, "loss": 0.7123, "step": 23 }, { "epoch": 0.0, "learning_rate": 1.44e-06, "loss": 0.6735, "step": 24 }, { "epoch": 0.0, "learning_rate": 1.5e-06, "loss": 0.6911, "step": 25 }, { "epoch": 0.0, "learning_rate": 1.56e-06, "loss": 0.6768, "step": 26 }, { "epoch": 0.0, "learning_rate": 1.6200000000000002e-06, "loss": 0.7297, "step": 27 }, { "epoch": 0.0, "learning_rate": 1.6800000000000002e-06, "loss": 0.6315, "step": 28 }, { "epoch": 0.0, "learning_rate": 1.7399999999999999e-06, "loss": 0.7526, "step": 29 }, { "epoch": 0.0, "learning_rate": 1.8e-06, "loss": 0.6918, "step": 30 }, { "epoch": 0.0, "learning_rate": 1.86e-06, "loss": 0.7225, "step": 31 }, { "epoch": 0.0, "learning_rate": 1.9200000000000003e-06, "loss": 0.7203, "step": 32 }, { "epoch": 0.0, "learning_rate": 1.98e-06, "loss": 0.741, "step": 33 }, { "epoch": 0.0, "learning_rate": 2.0400000000000004e-06, "loss": 0.6852, "step": 34 }, { "epoch": 0.0, "learning_rate": 2.1e-06, "loss": 0.6203, "step": 35 }, { "epoch": 0.0, "learning_rate": 2.16e-06, "loss": 0.6847, "step": 36 }, { "epoch": 0.0, "learning_rate": 2.22e-06, "loss": 0.7187, "step": 37 }, { "epoch": 0.0, "learning_rate": 2.28e-06, "loss": 0.6868, "step": 38 }, { "epoch": 0.0, "learning_rate": 2.34e-06, "loss": 0.7118, "step": 39 }, { "epoch": 0.0, "learning_rate": 2.4000000000000003e-06, "loss": 0.7014, "step": 40 }, { "epoch": 0.0, "learning_rate": 2.4599999999999997e-06, "loss": 0.6241, "step": 41 }, { "epoch": 0.01, "learning_rate": 2.52e-06, "loss": 0.6202, "step": 42 }, { "epoch": 0.01, "learning_rate": 2.58e-06, "loss": 0.6847, "step": 43 }, { "epoch": 0.01, "learning_rate": 2.64e-06, "loss": 0.7151, "step": 44 }, { "epoch": 0.01, "learning_rate": 2.7e-06, "loss": 0.7158, "step": 45 }, { "epoch": 0.01, "learning_rate": 2.7600000000000003e-06, "loss": 0.7229, "step": 46 }, { "epoch": 0.01, "learning_rate": 2.82e-06, "loss": 0.6889, "step": 47 }, { "epoch": 0.01, "learning_rate": 2.88e-06, "loss": 0.6562, "step": 48 }, { "epoch": 0.01, "learning_rate": 2.9400000000000002e-06, "loss": 0.7117, "step": 49 }, { "epoch": 0.01, "learning_rate": 3e-06, "loss": 0.6785, "step": 50 }, { "epoch": 0.01, "learning_rate": 2.9999280143971204e-06, "loss": 0.68, "step": 51 }, { "epoch": 0.01, "learning_rate": 2.999856028794241e-06, "loss": 0.637, "step": 52 }, { "epoch": 0.01, "learning_rate": 2.9997840431913616e-06, "loss": 0.6637, "step": 53 }, { "epoch": 0.01, "learning_rate": 2.9997120575884823e-06, "loss": 0.707, "step": 54 }, { "epoch": 0.01, "learning_rate": 2.9996400719856027e-06, "loss": 0.7101, "step": 55 }, { "epoch": 0.01, "learning_rate": 2.9995680863827235e-06, "loss": 0.6917, "step": 56 }, { "epoch": 0.01, "learning_rate": 2.999496100779844e-06, "loss": 0.697, "step": 57 }, { "epoch": 0.01, "learning_rate": 2.9994241151769646e-06, "loss": 0.659, "step": 58 }, { "epoch": 0.01, "learning_rate": 2.9993521295740854e-06, "loss": 0.6586, "step": 59 }, { "epoch": 0.01, "learning_rate": 2.999280143971206e-06, "loss": 0.6349, "step": 60 }, { "epoch": 0.01, "learning_rate": 2.9992081583683265e-06, "loss": 0.6546, "step": 61 }, { "epoch": 0.01, "learning_rate": 2.999136172765447e-06, "loss": 0.665, "step": 62 }, { "epoch": 0.01, "learning_rate": 2.9990641871625677e-06, "loss": 0.6952, "step": 63 }, { "epoch": 0.01, "learning_rate": 2.998992201559688e-06, "loss": 0.6586, "step": 64 }, { "epoch": 0.01, "learning_rate": 2.998920215956809e-06, "loss": 0.6481, "step": 65 }, { "epoch": 0.01, "learning_rate": 2.998848230353929e-06, "loss": 0.6872, "step": 66 }, { "epoch": 0.01, "learning_rate": 2.99877624475105e-06, "loss": 0.69, "step": 67 }, { "epoch": 0.01, "learning_rate": 2.9987042591481703e-06, "loss": 0.6817, "step": 68 }, { "epoch": 0.01, "learning_rate": 2.998632273545291e-06, "loss": 0.6628, "step": 69 }, { "epoch": 0.01, "learning_rate": 2.9985602879424114e-06, "loss": 0.6579, "step": 70 }, { "epoch": 0.01, "learning_rate": 2.998488302339532e-06, "loss": 0.6448, "step": 71 }, { "epoch": 0.01, "learning_rate": 2.9984163167366525e-06, "loss": 0.6862, "step": 72 }, { "epoch": 0.01, "learning_rate": 2.9983443311337733e-06, "loss": 0.6315, "step": 73 }, { "epoch": 0.01, "learning_rate": 2.998272345530894e-06, "loss": 0.7003, "step": 74 }, { "epoch": 0.01, "learning_rate": 2.9982003599280145e-06, "loss": 0.6425, "step": 75 }, { "epoch": 0.01, "learning_rate": 2.9981283743251352e-06, "loss": 0.6811, "step": 76 }, { "epoch": 0.01, "learning_rate": 2.9980563887222556e-06, "loss": 0.6739, "step": 77 }, { "epoch": 0.01, "learning_rate": 2.9979844031193764e-06, "loss": 0.6894, "step": 78 }, { "epoch": 0.01, "learning_rate": 2.9979124175164967e-06, "loss": 0.6898, "step": 79 }, { "epoch": 0.01, "learning_rate": 2.9978404319136175e-06, "loss": 0.6598, "step": 80 }, { "epoch": 0.01, "learning_rate": 2.997768446310738e-06, "loss": 0.7027, "step": 81 }, { "epoch": 0.01, "learning_rate": 2.9976964607078586e-06, "loss": 0.6275, "step": 82 }, { "epoch": 0.01, "learning_rate": 2.997624475104979e-06, "loss": 0.644, "step": 83 }, { "epoch": 0.01, "learning_rate": 2.9975524895020998e-06, "loss": 0.6812, "step": 84 }, { "epoch": 0.01, "learning_rate": 2.99748050389922e-06, "loss": 0.7064, "step": 85 }, { "epoch": 0.01, "learning_rate": 2.9974085182963405e-06, "loss": 0.6534, "step": 86 }, { "epoch": 0.01, "learning_rate": 2.9973365326934613e-06, "loss": 0.6535, "step": 87 }, { "epoch": 0.01, "learning_rate": 2.997264547090582e-06, "loss": 0.6643, "step": 88 }, { "epoch": 0.01, "learning_rate": 2.9971925614877024e-06, "loss": 0.6725, "step": 89 }, { "epoch": 0.01, "learning_rate": 2.997120575884823e-06, "loss": 0.6323, "step": 90 }, { "epoch": 0.01, "learning_rate": 2.997048590281944e-06, "loss": 0.6546, "step": 91 }, { "epoch": 0.01, "learning_rate": 2.9969766046790643e-06, "loss": 0.6666, "step": 92 }, { "epoch": 0.01, "learning_rate": 2.996904619076185e-06, "loss": 0.6773, "step": 93 }, { "epoch": 0.01, "learning_rate": 2.9968326334733054e-06, "loss": 0.664, "step": 94 }, { "epoch": 0.01, "learning_rate": 2.9967606478704262e-06, "loss": 0.6812, "step": 95 }, { "epoch": 0.01, "learning_rate": 2.9966886622675466e-06, "loss": 0.6356, "step": 96 }, { "epoch": 0.01, "learning_rate": 2.996616676664667e-06, "loss": 0.6471, "step": 97 }, { "epoch": 0.01, "learning_rate": 2.9965446910617877e-06, "loss": 0.6363, "step": 98 }, { "epoch": 0.01, "learning_rate": 2.996472705458908e-06, "loss": 0.638, "step": 99 }, { "epoch": 0.01, "learning_rate": 2.996400719856029e-06, "loss": 0.6293, "step": 100 }, { "epoch": 0.01, "learning_rate": 2.996328734253149e-06, "loss": 0.6686, "step": 101 }, { "epoch": 0.01, "learning_rate": 2.99625674865027e-06, "loss": 0.651, "step": 102 }, { "epoch": 0.01, "learning_rate": 2.9961847630473903e-06, "loss": 0.6789, "step": 103 }, { "epoch": 0.01, "learning_rate": 2.996112777444511e-06, "loss": 0.643, "step": 104 }, { "epoch": 0.01, "learning_rate": 2.996040791841632e-06, "loss": 0.6613, "step": 105 }, { "epoch": 0.01, "learning_rate": 2.9959688062387527e-06, "loss": 0.6366, "step": 106 }, { "epoch": 0.01, "learning_rate": 2.995896820635873e-06, "loss": 0.638, "step": 107 }, { "epoch": 0.01, "learning_rate": 2.9958248350329934e-06, "loss": 0.6422, "step": 108 }, { "epoch": 0.01, "learning_rate": 2.995752849430114e-06, "loss": 0.6451, "step": 109 }, { "epoch": 0.01, "learning_rate": 2.9956808638272345e-06, "loss": 0.6702, "step": 110 }, { "epoch": 0.01, "learning_rate": 2.9956088782243553e-06, "loss": 0.582, "step": 111 }, { "epoch": 0.01, "learning_rate": 2.9955368926214756e-06, "loss": 0.6227, "step": 112 }, { "epoch": 0.01, "learning_rate": 2.9954649070185964e-06, "loss": 0.5813, "step": 113 }, { "epoch": 0.01, "learning_rate": 2.9953929214157168e-06, "loss": 0.6168, "step": 114 }, { "epoch": 0.01, "learning_rate": 2.9953209358128376e-06, "loss": 0.5951, "step": 115 }, { "epoch": 0.01, "learning_rate": 2.995248950209958e-06, "loss": 0.6516, "step": 116 }, { "epoch": 0.01, "learning_rate": 2.9951769646070787e-06, "loss": 0.5763, "step": 117 }, { "epoch": 0.01, "learning_rate": 2.995104979004199e-06, "loss": 0.6386, "step": 118 }, { "epoch": 0.01, "learning_rate": 2.99503299340132e-06, "loss": 0.562, "step": 119 }, { "epoch": 0.01, "learning_rate": 2.9949610077984406e-06, "loss": 0.5636, "step": 120 }, { "epoch": 0.01, "learning_rate": 2.994889022195561e-06, "loss": 0.5845, "step": 121 }, { "epoch": 0.01, "learning_rate": 2.9948170365926817e-06, "loss": 0.5923, "step": 122 }, { "epoch": 0.01, "learning_rate": 2.994745050989802e-06, "loss": 0.6091, "step": 123 }, { "epoch": 0.01, "learning_rate": 2.994673065386923e-06, "loss": 0.5646, "step": 124 }, { "epoch": 0.01, "learning_rate": 2.9946010797840432e-06, "loss": 0.6205, "step": 125 }, { "epoch": 0.02, "learning_rate": 2.994529094181164e-06, "loss": 0.5659, "step": 126 }, { "epoch": 0.02, "learning_rate": 2.9944571085782844e-06, "loss": 0.5494, "step": 127 }, { "epoch": 0.02, "learning_rate": 2.994385122975405e-06, "loss": 0.6188, "step": 128 }, { "epoch": 0.02, "learning_rate": 2.9943131373725255e-06, "loss": 0.5581, "step": 129 }, { "epoch": 0.02, "learning_rate": 2.9942411517696463e-06, "loss": 0.6175, "step": 130 }, { "epoch": 0.02, "learning_rate": 2.9941691661667666e-06, "loss": 0.526, "step": 131 }, { "epoch": 0.02, "learning_rate": 2.994097180563887e-06, "loss": 0.5344, "step": 132 }, { "epoch": 0.02, "learning_rate": 2.9940251949610078e-06, "loss": 0.5837, "step": 133 }, { "epoch": 0.02, "learning_rate": 2.9939532093581285e-06, "loss": 0.6449, "step": 134 }, { "epoch": 0.02, "learning_rate": 2.993881223755249e-06, "loss": 0.5717, "step": 135 }, { "epoch": 0.02, "learning_rate": 2.9938092381523697e-06, "loss": 0.6344, "step": 136 }, { "epoch": 0.02, "learning_rate": 2.9937372525494904e-06, "loss": 0.4951, "step": 137 }, { "epoch": 0.02, "learning_rate": 2.993665266946611e-06, "loss": 0.5986, "step": 138 }, { "epoch": 0.02, "learning_rate": 2.9935932813437316e-06, "loss": 0.4899, "step": 139 }, { "epoch": 0.02, "learning_rate": 2.993521295740852e-06, "loss": 0.5303, "step": 140 }, { "epoch": 0.02, "learning_rate": 2.9934493101379727e-06, "loss": 0.4675, "step": 141 }, { "epoch": 0.02, "learning_rate": 2.993377324535093e-06, "loss": 0.507, "step": 142 }, { "epoch": 0.02, "learning_rate": 2.9933053389322134e-06, "loss": 0.5314, "step": 143 }, { "epoch": 0.02, "learning_rate": 2.993233353329334e-06, "loss": 0.5612, "step": 144 }, { "epoch": 0.02, "learning_rate": 2.9931613677264546e-06, "loss": 0.6848, "step": 145 }, { "epoch": 0.02, "learning_rate": 2.9930893821235753e-06, "loss": 0.6311, "step": 146 }, { "epoch": 0.02, "learning_rate": 2.9930173965206957e-06, "loss": 0.594, "step": 147 }, { "epoch": 0.02, "learning_rate": 2.9929454109178165e-06, "loss": 0.4442, "step": 148 }, { "epoch": 0.02, "learning_rate": 2.992873425314937e-06, "loss": 0.5896, "step": 149 }, { "epoch": 0.02, "learning_rate": 2.9928014397120576e-06, "loss": 0.5003, "step": 150 }, { "epoch": 0.02, "learning_rate": 2.9927294541091784e-06, "loss": 0.6424, "step": 151 }, { "epoch": 0.02, "learning_rate": 2.992657468506299e-06, "loss": 0.5194, "step": 152 }, { "epoch": 0.02, "learning_rate": 2.9925854829034195e-06, "loss": 0.357, "step": 153 }, { "epoch": 0.02, "learning_rate": 2.99251349730054e-06, "loss": 0.5783, "step": 154 }, { "epoch": 0.02, "learning_rate": 2.9924415116976606e-06, "loss": 0.5791, "step": 155 }, { "epoch": 0.02, "learning_rate": 2.992369526094781e-06, "loss": 0.4497, "step": 156 }, { "epoch": 0.02, "learning_rate": 2.9922975404919018e-06, "loss": 0.5615, "step": 157 }, { "epoch": 0.02, "learning_rate": 2.992225554889022e-06, "loss": 0.4484, "step": 158 }, { "epoch": 0.02, "learning_rate": 2.992153569286143e-06, "loss": 0.5185, "step": 159 }, { "epoch": 0.02, "learning_rate": 2.9920815836832633e-06, "loss": 0.3688, "step": 160 }, { "epoch": 0.02, "learning_rate": 2.992009598080384e-06, "loss": 0.5007, "step": 161 }, { "epoch": 0.02, "learning_rate": 2.9919376124775044e-06, "loss": 0.3821, "step": 162 }, { "epoch": 0.02, "learning_rate": 2.991865626874625e-06, "loss": 0.4235, "step": 163 }, { "epoch": 0.02, "learning_rate": 2.9917936412717455e-06, "loss": 0.5124, "step": 164 }, { "epoch": 0.02, "learning_rate": 2.9917216556688663e-06, "loss": 0.4509, "step": 165 }, { "epoch": 0.02, "learning_rate": 2.991649670065987e-06, "loss": 0.3715, "step": 166 }, { "epoch": 0.02, "learning_rate": 2.9915776844631074e-06, "loss": 0.4183, "step": 167 }, { "epoch": 0.02, "learning_rate": 2.9915056988602282e-06, "loss": 0.4456, "step": 168 }, { "epoch": 0.02, "learning_rate": 2.9914337132573486e-06, "loss": 0.4074, "step": 169 }, { "epoch": 0.02, "learning_rate": 2.9913617276544694e-06, "loss": 0.4632, "step": 170 }, { "epoch": 0.02, "learning_rate": 2.9912897420515897e-06, "loss": 0.3877, "step": 171 }, { "epoch": 0.02, "learning_rate": 2.9912177564487105e-06, "loss": 0.5141, "step": 172 }, { "epoch": 0.02, "learning_rate": 2.991145770845831e-06, "loss": 0.4813, "step": 173 }, { "epoch": 0.02, "learning_rate": 2.9910737852429516e-06, "loss": 0.4361, "step": 174 }, { "epoch": 0.02, "learning_rate": 2.991001799640072e-06, "loss": 0.3732, "step": 175 }, { "epoch": 0.02, "learning_rate": 2.9909298140371928e-06, "loss": 0.5574, "step": 176 }, { "epoch": 0.02, "learning_rate": 2.990857828434313e-06, "loss": 0.4707, "step": 177 }, { "epoch": 0.02, "learning_rate": 2.9907858428314335e-06, "loss": 0.6145, "step": 178 }, { "epoch": 0.02, "learning_rate": 2.9907138572285543e-06, "loss": 0.3325, "step": 179 }, { "epoch": 0.02, "learning_rate": 2.9906418716256746e-06, "loss": 0.4664, "step": 180 }, { "epoch": 0.02, "learning_rate": 2.9905698860227954e-06, "loss": 0.3629, "step": 181 }, { "epoch": 0.02, "learning_rate": 2.990497900419916e-06, "loss": 0.4273, "step": 182 }, { "epoch": 0.02, "learning_rate": 2.990425914817037e-06, "loss": 0.2908, "step": 183 }, { "epoch": 0.02, "learning_rate": 2.9903539292141573e-06, "loss": 0.5378, "step": 184 }, { "epoch": 0.02, "learning_rate": 2.990281943611278e-06, "loss": 0.4726, "step": 185 }, { "epoch": 0.02, "learning_rate": 2.9902099580083984e-06, "loss": 0.4236, "step": 186 }, { "epoch": 0.02, "learning_rate": 2.990137972405519e-06, "loss": 0.326, "step": 187 }, { "epoch": 0.02, "learning_rate": 2.9900659868026396e-06, "loss": 0.4189, "step": 188 }, { "epoch": 0.02, "learning_rate": 2.98999400119976e-06, "loss": 0.517, "step": 189 }, { "epoch": 0.02, "learning_rate": 2.9899220155968807e-06, "loss": 0.6908, "step": 190 }, { "epoch": 0.02, "learning_rate": 2.989850029994001e-06, "loss": 0.3762, "step": 191 }, { "epoch": 0.02, "learning_rate": 2.989778044391122e-06, "loss": 0.2868, "step": 192 }, { "epoch": 0.02, "learning_rate": 2.989706058788242e-06, "loss": 0.4088, "step": 193 }, { "epoch": 0.02, "learning_rate": 2.989634073185363e-06, "loss": 0.3175, "step": 194 }, { "epoch": 0.02, "learning_rate": 2.9895620875824833e-06, "loss": 0.3311, "step": 195 }, { "epoch": 0.02, "learning_rate": 2.989490101979604e-06, "loss": 0.4216, "step": 196 }, { "epoch": 0.02, "learning_rate": 2.989418116376725e-06, "loss": 0.4506, "step": 197 }, { "epoch": 0.02, "learning_rate": 2.9893461307738452e-06, "loss": 0.5093, "step": 198 }, { "epoch": 0.02, "learning_rate": 2.989274145170966e-06, "loss": 0.4923, "step": 199 }, { "epoch": 0.02, "learning_rate": 2.989202159568087e-06, "loss": 0.5234, "step": 200 }, { "epoch": 0.02, "learning_rate": 2.989130173965207e-06, "loss": 0.4887, "step": 201 }, { "epoch": 0.02, "learning_rate": 2.9890581883623275e-06, "loss": 0.4449, "step": 202 }, { "epoch": 0.02, "learning_rate": 2.9889862027594483e-06, "loss": 0.3455, "step": 203 }, { "epoch": 0.02, "learning_rate": 2.9889142171565686e-06, "loss": 0.7352, "step": 204 }, { "epoch": 0.02, "learning_rate": 2.9888422315536894e-06, "loss": 0.3337, "step": 205 }, { "epoch": 0.02, "learning_rate": 2.9887702459508098e-06, "loss": 0.3727, "step": 206 }, { "epoch": 0.02, "learning_rate": 2.9886982603479305e-06, "loss": 0.3784, "step": 207 }, { "epoch": 0.02, "learning_rate": 2.988626274745051e-06, "loss": 0.2429, "step": 208 }, { "epoch": 0.03, "learning_rate": 2.9885542891421717e-06, "loss": 0.4806, "step": 209 }, { "epoch": 0.03, "learning_rate": 2.988482303539292e-06, "loss": 0.3787, "step": 210 }, { "epoch": 0.03, "learning_rate": 2.988410317936413e-06, "loss": 0.4954, "step": 211 }, { "epoch": 0.03, "learning_rate": 2.988338332333533e-06, "loss": 0.351, "step": 212 }, { "epoch": 0.03, "learning_rate": 2.988266346730654e-06, "loss": 0.3838, "step": 213 }, { "epoch": 0.03, "learning_rate": 2.9881943611277747e-06, "loss": 0.245, "step": 214 }, { "epoch": 0.03, "learning_rate": 2.988122375524895e-06, "loss": 0.4977, "step": 215 }, { "epoch": 0.03, "learning_rate": 2.988050389922016e-06, "loss": 0.6061, "step": 216 }, { "epoch": 0.03, "learning_rate": 2.9879784043191362e-06, "loss": 0.3471, "step": 217 }, { "epoch": 0.03, "learning_rate": 2.987906418716257e-06, "loss": 0.283, "step": 218 }, { "epoch": 0.03, "learning_rate": 2.9878344331133773e-06, "loss": 0.1691, "step": 219 }, { "epoch": 0.03, "learning_rate": 2.987762447510498e-06, "loss": 0.725, "step": 220 }, { "epoch": 0.03, "learning_rate": 2.9876904619076185e-06, "loss": 0.5104, "step": 221 }, { "epoch": 0.03, "learning_rate": 2.9876184763047393e-06, "loss": 0.3587, "step": 222 }, { "epoch": 0.03, "learning_rate": 2.9875464907018596e-06, "loss": 0.5577, "step": 223 }, { "epoch": 0.03, "learning_rate": 2.98747450509898e-06, "loss": 0.5938, "step": 224 }, { "epoch": 0.03, "learning_rate": 2.9874025194961007e-06, "loss": 0.227, "step": 225 }, { "epoch": 0.03, "learning_rate": 2.987330533893221e-06, "loss": 0.4635, "step": 226 }, { "epoch": 0.03, "learning_rate": 2.987258548290342e-06, "loss": 0.3724, "step": 227 }, { "epoch": 0.03, "learning_rate": 2.9871865626874627e-06, "loss": 0.4051, "step": 228 }, { "epoch": 0.03, "learning_rate": 2.9871145770845834e-06, "loss": 0.2834, "step": 229 }, { "epoch": 0.03, "learning_rate": 2.987042591481704e-06, "loss": 0.5185, "step": 230 }, { "epoch": 0.03, "learning_rate": 2.9869706058788246e-06, "loss": 0.3881, "step": 231 }, { "epoch": 0.03, "learning_rate": 2.986898620275945e-06, "loss": 0.3498, "step": 232 }, { "epoch": 0.03, "learning_rate": 2.9868266346730657e-06, "loss": 0.5892, "step": 233 }, { "epoch": 0.03, "learning_rate": 2.986754649070186e-06, "loss": 0.3497, "step": 234 }, { "epoch": 0.03, "learning_rate": 2.9866826634673064e-06, "loss": 0.4868, "step": 235 }, { "epoch": 0.03, "learning_rate": 2.986610677864427e-06, "loss": 0.5813, "step": 236 }, { "epoch": 0.03, "learning_rate": 2.9865386922615475e-06, "loss": 0.5307, "step": 237 }, { "epoch": 0.03, "learning_rate": 2.9864667066586683e-06, "loss": 0.2627, "step": 238 }, { "epoch": 0.03, "learning_rate": 2.9863947210557887e-06, "loss": 0.2727, "step": 239 }, { "epoch": 0.03, "learning_rate": 2.9863227354529095e-06, "loss": 0.3205, "step": 240 }, { "epoch": 0.03, "learning_rate": 2.98625074985003e-06, "loss": 0.1898, "step": 241 }, { "epoch": 0.03, "learning_rate": 2.9861787642471506e-06, "loss": 0.2787, "step": 242 }, { "epoch": 0.03, "learning_rate": 2.9861067786442714e-06, "loss": 0.4048, "step": 243 }, { "epoch": 0.03, "learning_rate": 2.9860347930413917e-06, "loss": 0.6191, "step": 244 }, { "epoch": 0.03, "learning_rate": 2.9859628074385125e-06, "loss": 0.2396, "step": 245 }, { "epoch": 0.03, "learning_rate": 2.9858908218356333e-06, "loss": 0.1613, "step": 246 }, { "epoch": 0.03, "learning_rate": 2.9858188362327536e-06, "loss": 0.2928, "step": 247 }, { "epoch": 0.03, "learning_rate": 2.985746850629874e-06, "loss": 0.307, "step": 248 }, { "epoch": 0.03, "learning_rate": 2.9856748650269948e-06, "loss": 0.5204, "step": 249 }, { "epoch": 0.03, "learning_rate": 2.985602879424115e-06, "loss": 0.1896, "step": 250 }, { "epoch": 0.03, "learning_rate": 2.985530893821236e-06, "loss": 0.4803, "step": 251 }, { "epoch": 0.03, "learning_rate": 2.9854589082183563e-06, "loss": 0.274, "step": 252 }, { "epoch": 0.03, "learning_rate": 2.985386922615477e-06, "loss": 0.1767, "step": 253 }, { "epoch": 0.03, "learning_rate": 2.9853149370125974e-06, "loss": 0.2321, "step": 254 }, { "epoch": 0.03, "learning_rate": 2.985242951409718e-06, "loss": 0.3735, "step": 255 }, { "epoch": 0.03, "learning_rate": 2.9851709658068385e-06, "loss": 0.3029, "step": 256 }, { "epoch": 0.03, "learning_rate": 2.9850989802039593e-06, "loss": 0.2808, "step": 257 }, { "epoch": 0.03, "learning_rate": 2.9850269946010797e-06, "loss": 0.4043, "step": 258 }, { "epoch": 0.03, "learning_rate": 2.9849550089982004e-06, "loss": 0.4669, "step": 259 }, { "epoch": 0.03, "learning_rate": 2.9848830233953212e-06, "loss": 0.2095, "step": 260 }, { "epoch": 0.03, "learning_rate": 2.9848110377924416e-06, "loss": 0.2929, "step": 261 }, { "epoch": 0.03, "learning_rate": 2.9847390521895624e-06, "loss": 0.6471, "step": 262 }, { "epoch": 0.03, "learning_rate": 2.9846670665866827e-06, "loss": 0.569, "step": 263 }, { "epoch": 0.03, "learning_rate": 2.9845950809838035e-06, "loss": 0.2169, "step": 264 }, { "epoch": 0.03, "learning_rate": 2.984523095380924e-06, "loss": 0.1841, "step": 265 }, { "epoch": 0.03, "learning_rate": 2.9844511097780446e-06, "loss": 0.2299, "step": 266 }, { "epoch": 0.03, "learning_rate": 2.984379124175165e-06, "loss": 0.2362, "step": 267 }, { "epoch": 0.03, "learning_rate": 2.9843071385722858e-06, "loss": 0.4815, "step": 268 }, { "epoch": 0.03, "learning_rate": 2.984235152969406e-06, "loss": 0.2483, "step": 269 }, { "epoch": 0.03, "learning_rate": 2.9841631673665265e-06, "loss": 0.3003, "step": 270 }, { "epoch": 0.03, "learning_rate": 2.9840911817636472e-06, "loss": 0.8531, "step": 271 }, { "epoch": 0.03, "learning_rate": 2.9840191961607676e-06, "loss": 0.2572, "step": 272 }, { "epoch": 0.03, "learning_rate": 2.9839472105578884e-06, "loss": 0.1711, "step": 273 }, { "epoch": 0.03, "learning_rate": 2.983875224955009e-06, "loss": 0.3667, "step": 274 }, { "epoch": 0.03, "learning_rate": 2.98380323935213e-06, "loss": 0.1474, "step": 275 }, { "epoch": 0.03, "learning_rate": 2.9837312537492503e-06, "loss": 0.4127, "step": 276 }, { "epoch": 0.03, "learning_rate": 2.983659268146371e-06, "loss": 0.4241, "step": 277 }, { "epoch": 0.03, "learning_rate": 2.9835872825434914e-06, "loss": 0.1358, "step": 278 }, { "epoch": 0.03, "learning_rate": 2.983515296940612e-06, "loss": 0.2662, "step": 279 }, { "epoch": 0.03, "learning_rate": 2.9834433113377326e-06, "loss": 0.2744, "step": 280 }, { "epoch": 0.03, "learning_rate": 2.983371325734853e-06, "loss": 0.4064, "step": 281 }, { "epoch": 0.03, "learning_rate": 2.9832993401319737e-06, "loss": 0.161, "step": 282 }, { "epoch": 0.03, "learning_rate": 2.983227354529094e-06, "loss": 0.2638, "step": 283 }, { "epoch": 0.03, "learning_rate": 2.983155368926215e-06, "loss": 0.172, "step": 284 }, { "epoch": 0.03, "learning_rate": 2.983083383323335e-06, "loss": 0.1118, "step": 285 }, { "epoch": 0.03, "learning_rate": 2.983011397720456e-06, "loss": 0.3429, "step": 286 }, { "epoch": 0.03, "learning_rate": 2.9829394121175763e-06, "loss": 0.2672, "step": 287 }, { "epoch": 0.03, "learning_rate": 2.982867426514697e-06, "loss": 0.0923, "step": 288 }, { "epoch": 0.03, "learning_rate": 2.982795440911818e-06, "loss": 0.2008, "step": 289 }, { "epoch": 0.03, "learning_rate": 2.9827234553089382e-06, "loss": 0.2189, "step": 290 }, { "epoch": 0.03, "learning_rate": 2.982651469706059e-06, "loss": 0.5997, "step": 291 }, { "epoch": 0.03, "learning_rate": 2.9825794841031798e-06, "loss": 0.4427, "step": 292 }, { "epoch": 0.04, "learning_rate": 2.9825074985003e-06, "loss": 0.4828, "step": 293 }, { "epoch": 0.04, "learning_rate": 2.9824355128974205e-06, "loss": 0.2958, "step": 294 }, { "epoch": 0.04, "learning_rate": 2.9823635272945413e-06, "loss": 0.225, "step": 295 }, { "epoch": 0.04, "learning_rate": 2.9822915416916616e-06, "loss": 0.099, "step": 296 }, { "epoch": 0.04, "learning_rate": 2.9822195560887824e-06, "loss": 0.3458, "step": 297 }, { "epoch": 0.04, "learning_rate": 2.9821475704859028e-06, "loss": 0.0884, "step": 298 }, { "epoch": 0.04, "learning_rate": 2.9820755848830235e-06, "loss": 0.2096, "step": 299 }, { "epoch": 0.04, "learning_rate": 2.982003599280144e-06, "loss": 0.1305, "step": 300 }, { "epoch": 0.04, "learning_rate": 2.9819316136772647e-06, "loss": 0.3927, "step": 301 }, { "epoch": 0.04, "learning_rate": 2.981859628074385e-06, "loss": 0.168, "step": 302 }, { "epoch": 0.04, "learning_rate": 2.981787642471506e-06, "loss": 0.265, "step": 303 }, { "epoch": 0.04, "learning_rate": 2.981715656868626e-06, "loss": 0.5666, "step": 304 }, { "epoch": 0.04, "learning_rate": 2.981643671265747e-06, "loss": 0.1012, "step": 305 }, { "epoch": 0.04, "learning_rate": 2.9815716856628677e-06, "loss": 0.2275, "step": 306 }, { "epoch": 0.04, "learning_rate": 2.981499700059988e-06, "loss": 0.3214, "step": 307 }, { "epoch": 0.04, "learning_rate": 2.981427714457109e-06, "loss": 0.1115, "step": 308 }, { "epoch": 0.04, "learning_rate": 2.981355728854229e-06, "loss": 0.3473, "step": 309 }, { "epoch": 0.04, "learning_rate": 2.98128374325135e-06, "loss": 0.2609, "step": 310 }, { "epoch": 0.04, "learning_rate": 2.9812117576484703e-06, "loss": 0.6249, "step": 311 }, { "epoch": 0.04, "learning_rate": 2.981139772045591e-06, "loss": 0.2169, "step": 312 }, { "epoch": 0.04, "learning_rate": 2.9810677864427115e-06, "loss": 0.2487, "step": 313 }, { "epoch": 0.04, "learning_rate": 2.9809958008398323e-06, "loss": 0.1626, "step": 314 }, { "epoch": 0.04, "learning_rate": 2.9809238152369526e-06, "loss": 0.3269, "step": 315 }, { "epoch": 0.04, "learning_rate": 2.980851829634073e-06, "loss": 0.4175, "step": 316 }, { "epoch": 0.04, "learning_rate": 2.9807798440311937e-06, "loss": 0.3258, "step": 317 }, { "epoch": 0.04, "learning_rate": 2.980707858428314e-06, "loss": 0.2831, "step": 318 }, { "epoch": 0.04, "learning_rate": 2.980635872825435e-06, "loss": 0.1444, "step": 319 }, { "epoch": 0.04, "learning_rate": 2.9805638872225557e-06, "loss": 0.5743, "step": 320 }, { "epoch": 0.04, "learning_rate": 2.9804919016196764e-06, "loss": 0.1471, "step": 321 }, { "epoch": 0.04, "learning_rate": 2.9804199160167968e-06, "loss": 0.1907, "step": 322 }, { "epoch": 0.04, "learning_rate": 2.9803479304139176e-06, "loss": 0.1057, "step": 323 }, { "epoch": 0.04, "learning_rate": 2.980275944811038e-06, "loss": 0.1393, "step": 324 }, { "epoch": 0.04, "learning_rate": 2.9802039592081587e-06, "loss": 0.175, "step": 325 }, { "epoch": 0.04, "learning_rate": 2.980131973605279e-06, "loss": 0.0992, "step": 326 }, { "epoch": 0.04, "learning_rate": 2.9800599880024e-06, "loss": 0.0498, "step": 327 }, { "epoch": 0.04, "learning_rate": 2.97998800239952e-06, "loss": 0.1064, "step": 328 }, { "epoch": 0.04, "learning_rate": 2.9799160167966405e-06, "loss": 0.1296, "step": 329 }, { "epoch": 0.04, "learning_rate": 2.9798440311937613e-06, "loss": 0.3928, "step": 330 }, { "epoch": 0.04, "learning_rate": 2.9797720455908817e-06, "loss": 0.1588, "step": 331 }, { "epoch": 0.04, "learning_rate": 2.9797000599880025e-06, "loss": 0.1646, "step": 332 }, { "epoch": 0.04, "learning_rate": 2.979628074385123e-06, "loss": 0.3486, "step": 333 }, { "epoch": 0.04, "learning_rate": 2.9795560887822436e-06, "loss": 0.3889, "step": 334 }, { "epoch": 0.04, "learning_rate": 2.979484103179364e-06, "loss": 0.1577, "step": 335 }, { "epoch": 0.04, "learning_rate": 2.9794121175764847e-06, "loss": 0.2712, "step": 336 }, { "epoch": 0.04, "learning_rate": 2.9793401319736055e-06, "loss": 0.5399, "step": 337 }, { "epoch": 0.04, "learning_rate": 2.9792681463707263e-06, "loss": 0.1963, "step": 338 }, { "epoch": 0.04, "learning_rate": 2.9791961607678466e-06, "loss": 0.3652, "step": 339 }, { "epoch": 0.04, "learning_rate": 2.979124175164967e-06, "loss": 0.204, "step": 340 }, { "epoch": 0.04, "learning_rate": 2.9790521895620878e-06, "loss": 0.2837, "step": 341 }, { "epoch": 0.04, "learning_rate": 2.978980203959208e-06, "loss": 0.1287, "step": 342 }, { "epoch": 0.04, "learning_rate": 2.978908218356329e-06, "loss": 0.0584, "step": 343 }, { "epoch": 0.04, "learning_rate": 2.9788362327534493e-06, "loss": 0.3584, "step": 344 }, { "epoch": 0.04, "learning_rate": 2.97876424715057e-06, "loss": 0.7483, "step": 345 }, { "epoch": 0.04, "learning_rate": 2.9786922615476904e-06, "loss": 0.2066, "step": 346 }, { "epoch": 0.04, "learning_rate": 2.978620275944811e-06, "loss": 0.4354, "step": 347 }, { "epoch": 0.04, "learning_rate": 2.9785482903419315e-06, "loss": 0.7611, "step": 348 }, { "epoch": 0.04, "learning_rate": 2.9784763047390523e-06, "loss": 0.3588, "step": 349 }, { "epoch": 0.04, "learning_rate": 2.9784043191361727e-06, "loss": 0.121, "step": 350 }, { "epoch": 0.04, "learning_rate": 2.9783323335332934e-06, "loss": 0.4053, "step": 351 }, { "epoch": 0.04, "learning_rate": 2.9782603479304142e-06, "loss": 0.0752, "step": 352 }, { "epoch": 0.04, "learning_rate": 2.9781883623275346e-06, "loss": 0.2269, "step": 353 }, { "epoch": 0.04, "learning_rate": 2.9781163767246553e-06, "loss": 0.2669, "step": 354 }, { "epoch": 0.04, "learning_rate": 2.9780443911217757e-06, "loss": 0.5194, "step": 355 }, { "epoch": 0.04, "learning_rate": 2.9779724055188965e-06, "loss": 0.2706, "step": 356 }, { "epoch": 0.04, "learning_rate": 2.977900419916017e-06, "loss": 0.0521, "step": 357 }, { "epoch": 0.04, "learning_rate": 2.9778284343131376e-06, "loss": 0.2501, "step": 358 }, { "epoch": 0.04, "learning_rate": 2.977756448710258e-06, "loss": 0.2602, "step": 359 }, { "epoch": 0.04, "learning_rate": 2.9776844631073787e-06, "loss": 0.2869, "step": 360 }, { "epoch": 0.04, "learning_rate": 2.977612477504499e-06, "loss": 0.2187, "step": 361 }, { "epoch": 0.04, "learning_rate": 2.9775404919016195e-06, "loss": 0.0719, "step": 362 }, { "epoch": 0.04, "learning_rate": 2.9774685062987402e-06, "loss": 0.2253, "step": 363 }, { "epoch": 0.04, "learning_rate": 2.9773965206958606e-06, "loss": 0.6581, "step": 364 }, { "epoch": 0.04, "learning_rate": 2.9773245350929814e-06, "loss": 0.0766, "step": 365 }, { "epoch": 0.04, "learning_rate": 2.977252549490102e-06, "loss": 0.4327, "step": 366 }, { "epoch": 0.04, "learning_rate": 2.9771805638872225e-06, "loss": 0.2297, "step": 367 }, { "epoch": 0.04, "learning_rate": 2.9771085782843433e-06, "loss": 0.0831, "step": 368 }, { "epoch": 0.04, "learning_rate": 2.977036592681464e-06, "loss": 0.5078, "step": 369 }, { "epoch": 0.04, "learning_rate": 2.9769646070785844e-06, "loss": 0.4176, "step": 370 }, { "epoch": 0.04, "learning_rate": 2.976892621475705e-06, "loss": 0.1349, "step": 371 }, { "epoch": 0.04, "learning_rate": 2.9768206358728255e-06, "loss": 0.5379, "step": 372 }, { "epoch": 0.04, "learning_rate": 2.9767486502699463e-06, "loss": 0.1064, "step": 373 }, { "epoch": 0.04, "learning_rate": 2.9766766646670667e-06, "loss": 0.0826, "step": 374 }, { "epoch": 0.04, "learning_rate": 2.976604679064187e-06, "loss": 0.2125, "step": 375 }, { "epoch": 0.05, "learning_rate": 2.976532693461308e-06, "loss": 0.1695, "step": 376 }, { "epoch": 0.05, "learning_rate": 2.976460707858428e-06, "loss": 0.1223, "step": 377 }, { "epoch": 0.05, "learning_rate": 2.976388722255549e-06, "loss": 0.128, "step": 378 }, { "epoch": 0.05, "learning_rate": 2.9763167366526693e-06, "loss": 0.2338, "step": 379 }, { "epoch": 0.05, "learning_rate": 2.97624475104979e-06, "loss": 0.0667, "step": 380 }, { "epoch": 0.05, "learning_rate": 2.9761727654469104e-06, "loss": 0.2225, "step": 381 }, { "epoch": 0.05, "learning_rate": 2.9761007798440312e-06, "loss": 0.3662, "step": 382 }, { "epoch": 0.05, "learning_rate": 2.976028794241152e-06, "loss": 0.2212, "step": 383 }, { "epoch": 0.05, "learning_rate": 2.9759568086382728e-06, "loss": 0.222, "step": 384 }, { "epoch": 0.05, "learning_rate": 2.975884823035393e-06, "loss": 0.2122, "step": 385 }, { "epoch": 0.05, "learning_rate": 2.9758128374325135e-06, "loss": 0.2133, "step": 386 }, { "epoch": 0.05, "learning_rate": 2.9757408518296343e-06, "loss": 0.0622, "step": 387 }, { "epoch": 0.05, "learning_rate": 2.9756688662267546e-06, "loss": 0.0578, "step": 388 }, { "epoch": 0.05, "learning_rate": 2.9755968806238754e-06, "loss": 0.4012, "step": 389 }, { "epoch": 0.05, "learning_rate": 2.9755248950209958e-06, "loss": 0.0792, "step": 390 }, { "epoch": 0.05, "learning_rate": 2.9754529094181165e-06, "loss": 0.4505, "step": 391 }, { "epoch": 0.05, "learning_rate": 2.975380923815237e-06, "loss": 0.0976, "step": 392 }, { "epoch": 0.05, "learning_rate": 2.9753089382123577e-06, "loss": 0.1397, "step": 393 }, { "epoch": 0.05, "learning_rate": 2.975236952609478e-06, "loss": 0.351, "step": 394 }, { "epoch": 0.05, "learning_rate": 2.975164967006599e-06, "loss": 0.1489, "step": 395 }, { "epoch": 0.05, "learning_rate": 2.975092981403719e-06, "loss": 0.0483, "step": 396 }, { "epoch": 0.05, "learning_rate": 2.97502099580084e-06, "loss": 0.2285, "step": 397 }, { "epoch": 0.05, "learning_rate": 2.9749490101979607e-06, "loss": 0.2727, "step": 398 }, { "epoch": 0.05, "learning_rate": 2.974877024595081e-06, "loss": 0.2785, "step": 399 }, { "epoch": 0.05, "learning_rate": 2.974805038992202e-06, "loss": 0.5291, "step": 400 }, { "epoch": 0.05, "learning_rate": 2.974733053389322e-06, "loss": 0.4213, "step": 401 }, { "epoch": 0.05, "learning_rate": 2.974661067786443e-06, "loss": 0.0624, "step": 402 }, { "epoch": 0.05, "learning_rate": 2.9745890821835633e-06, "loss": 0.165, "step": 403 }, { "epoch": 0.05, "learning_rate": 2.974517096580684e-06, "loss": 0.1602, "step": 404 }, { "epoch": 0.05, "learning_rate": 2.9744451109778045e-06, "loss": 0.1537, "step": 405 }, { "epoch": 0.05, "learning_rate": 2.9743731253749252e-06, "loss": 0.2374, "step": 406 }, { "epoch": 0.05, "learning_rate": 2.9743011397720456e-06, "loss": 0.4063, "step": 407 }, { "epoch": 0.05, "learning_rate": 2.974229154169166e-06, "loss": 0.2, "step": 408 }, { "epoch": 0.05, "learning_rate": 2.9741571685662867e-06, "loss": 0.2248, "step": 409 }, { "epoch": 0.05, "learning_rate": 2.974085182963407e-06, "loss": 0.7955, "step": 410 }, { "epoch": 0.05, "learning_rate": 2.974013197360528e-06, "loss": 0.4349, "step": 411 }, { "epoch": 0.05, "learning_rate": 2.9739412117576486e-06, "loss": 0.124, "step": 412 }, { "epoch": 0.05, "learning_rate": 2.973869226154769e-06, "loss": 0.0787, "step": 413 }, { "epoch": 0.05, "learning_rate": 2.9737972405518898e-06, "loss": 0.0889, "step": 414 }, { "epoch": 0.05, "learning_rate": 2.9737252549490106e-06, "loss": 0.1303, "step": 415 }, { "epoch": 0.05, "learning_rate": 2.973653269346131e-06, "loss": 0.4741, "step": 416 }, { "epoch": 0.05, "learning_rate": 2.9735812837432517e-06, "loss": 0.1288, "step": 417 }, { "epoch": 0.05, "learning_rate": 2.973509298140372e-06, "loss": 0.1725, "step": 418 }, { "epoch": 0.05, "learning_rate": 2.973437312537493e-06, "loss": 0.1939, "step": 419 }, { "epoch": 0.05, "learning_rate": 2.973365326934613e-06, "loss": 0.0498, "step": 420 }, { "epoch": 0.05, "learning_rate": 2.9732933413317335e-06, "loss": 0.0542, "step": 421 }, { "epoch": 0.05, "learning_rate": 2.9732213557288543e-06, "loss": 0.0569, "step": 422 }, { "epoch": 0.05, "learning_rate": 2.9731493701259747e-06, "loss": 0.4231, "step": 423 }, { "epoch": 0.05, "learning_rate": 2.9730773845230954e-06, "loss": 0.1495, "step": 424 }, { "epoch": 0.05, "learning_rate": 2.973005398920216e-06, "loss": 0.0467, "step": 425 }, { "epoch": 0.05, "learning_rate": 2.9729334133173366e-06, "loss": 0.2503, "step": 426 }, { "epoch": 0.05, "learning_rate": 2.972861427714457e-06, "loss": 0.1004, "step": 427 }, { "epoch": 0.05, "learning_rate": 2.9727894421115777e-06, "loss": 0.0955, "step": 428 }, { "epoch": 0.05, "learning_rate": 2.9727174565086985e-06, "loss": 0.0613, "step": 429 }, { "epoch": 0.05, "learning_rate": 2.9726454709058193e-06, "loss": 0.0701, "step": 430 }, { "epoch": 0.05, "learning_rate": 2.9725734853029396e-06, "loss": 0.0906, "step": 431 }, { "epoch": 0.05, "learning_rate": 2.97250149970006e-06, "loss": 0.0781, "step": 432 }, { "epoch": 0.05, "learning_rate": 2.9724295140971808e-06, "loss": 0.0543, "step": 433 }, { "epoch": 0.05, "learning_rate": 2.972357528494301e-06, "loss": 0.1401, "step": 434 }, { "epoch": 0.05, "learning_rate": 2.972285542891422e-06, "loss": 0.7412, "step": 435 }, { "epoch": 0.05, "learning_rate": 2.9722135572885422e-06, "loss": 0.4488, "step": 436 }, { "epoch": 0.05, "learning_rate": 2.972141571685663e-06, "loss": 0.6403, "step": 437 }, { "epoch": 0.05, "learning_rate": 2.9720695860827834e-06, "loss": 0.2496, "step": 438 }, { "epoch": 0.05, "learning_rate": 2.971997600479904e-06, "loss": 0.0559, "step": 439 }, { "epoch": 0.05, "learning_rate": 2.9719256148770245e-06, "loss": 0.2011, "step": 440 }, { "epoch": 0.05, "learning_rate": 2.9718536292741453e-06, "loss": 0.342, "step": 441 }, { "epoch": 0.05, "learning_rate": 2.9717816436712656e-06, "loss": 0.2924, "step": 442 }, { "epoch": 0.05, "learning_rate": 2.9717096580683864e-06, "loss": 0.0477, "step": 443 }, { "epoch": 0.05, "learning_rate": 2.971637672465507e-06, "loss": 0.0831, "step": 444 }, { "epoch": 0.05, "learning_rate": 2.9715656868626276e-06, "loss": 0.7163, "step": 445 }, { "epoch": 0.05, "learning_rate": 2.9714937012597483e-06, "loss": 0.1016, "step": 446 }, { "epoch": 0.05, "learning_rate": 2.9714217156568687e-06, "loss": 0.03, "step": 447 }, { "epoch": 0.05, "learning_rate": 2.9713497300539895e-06, "loss": 0.046, "step": 448 }, { "epoch": 0.05, "learning_rate": 2.97127774445111e-06, "loss": 0.1045, "step": 449 }, { "epoch": 0.05, "learning_rate": 2.9712057588482306e-06, "loss": 0.3782, "step": 450 }, { "epoch": 0.05, "learning_rate": 2.971133773245351e-06, "loss": 0.1354, "step": 451 }, { "epoch": 0.05, "learning_rate": 2.9710617876424717e-06, "loss": 0.3478, "step": 452 }, { "epoch": 0.05, "learning_rate": 2.970989802039592e-06, "loss": 0.1911, "step": 453 }, { "epoch": 0.05, "learning_rate": 2.970917816436713e-06, "loss": 0.2747, "step": 454 }, { "epoch": 0.05, "learning_rate": 2.9708458308338332e-06, "loss": 0.2077, "step": 455 }, { "epoch": 0.05, "learning_rate": 2.9707738452309536e-06, "loss": 0.3388, "step": 456 }, { "epoch": 0.05, "learning_rate": 2.9707018596280744e-06, "loss": 0.0785, "step": 457 }, { "epoch": 0.05, "learning_rate": 2.9706298740251947e-06, "loss": 0.0709, "step": 458 }, { "epoch": 0.06, "learning_rate": 2.9705578884223155e-06, "loss": 0.3352, "step": 459 }, { "epoch": 0.06, "learning_rate": 2.9704859028194363e-06, "loss": 0.1458, "step": 460 }, { "epoch": 0.06, "learning_rate": 2.970413917216557e-06, "loss": 0.0593, "step": 461 }, { "epoch": 0.06, "learning_rate": 2.9703419316136774e-06, "loss": 0.1629, "step": 462 }, { "epoch": 0.06, "learning_rate": 2.970269946010798e-06, "loss": 0.0567, "step": 463 }, { "epoch": 0.06, "learning_rate": 2.9701979604079185e-06, "loss": 0.227, "step": 464 }, { "epoch": 0.06, "learning_rate": 2.9701259748050393e-06, "loss": 0.0258, "step": 465 }, { "epoch": 0.06, "learning_rate": 2.9700539892021597e-06, "loss": 0.0422, "step": 466 }, { "epoch": 0.06, "learning_rate": 2.96998200359928e-06, "loss": 0.245, "step": 467 }, { "epoch": 0.06, "learning_rate": 2.969910017996401e-06, "loss": 0.4043, "step": 468 }, { "epoch": 0.06, "learning_rate": 2.969838032393521e-06, "loss": 0.0947, "step": 469 }, { "epoch": 0.06, "learning_rate": 2.969766046790642e-06, "loss": 0.0491, "step": 470 }, { "epoch": 0.06, "learning_rate": 2.9696940611877623e-06, "loss": 0.5133, "step": 471 }, { "epoch": 0.06, "learning_rate": 2.969622075584883e-06, "loss": 0.0467, "step": 472 }, { "epoch": 0.06, "learning_rate": 2.9695500899820034e-06, "loss": 0.0382, "step": 473 }, { "epoch": 0.06, "learning_rate": 2.969478104379124e-06, "loss": 0.1283, "step": 474 }, { "epoch": 0.06, "learning_rate": 2.969406118776245e-06, "loss": 0.1862, "step": 475 }, { "epoch": 0.06, "learning_rate": 2.9693341331733653e-06, "loss": 0.1167, "step": 476 }, { "epoch": 0.06, "learning_rate": 2.969262147570486e-06, "loss": 0.3902, "step": 477 }, { "epoch": 0.06, "learning_rate": 2.9691901619676065e-06, "loss": 0.1771, "step": 478 }, { "epoch": 0.06, "learning_rate": 2.9691181763647273e-06, "loss": 0.0247, "step": 479 }, { "epoch": 0.06, "learning_rate": 2.9690461907618476e-06, "loss": 0.0361, "step": 480 }, { "epoch": 0.06, "learning_rate": 2.9689742051589684e-06, "loss": 0.1153, "step": 481 }, { "epoch": 0.06, "learning_rate": 2.9689022195560887e-06, "loss": 0.9338, "step": 482 }, { "epoch": 0.06, "learning_rate": 2.9688302339532095e-06, "loss": 0.3624, "step": 483 }, { "epoch": 0.06, "learning_rate": 2.96875824835033e-06, "loss": 0.1594, "step": 484 }, { "epoch": 0.06, "learning_rate": 2.9686862627474507e-06, "loss": 0.3077, "step": 485 }, { "epoch": 0.06, "learning_rate": 2.968614277144571e-06, "loss": 0.0335, "step": 486 }, { "epoch": 0.06, "learning_rate": 2.9685422915416918e-06, "loss": 0.2005, "step": 487 }, { "epoch": 0.06, "learning_rate": 2.968470305938812e-06, "loss": 0.1476, "step": 488 }, { "epoch": 0.06, "learning_rate": 2.968398320335933e-06, "loss": 0.0667, "step": 489 }, { "epoch": 0.06, "learning_rate": 2.9683263347330533e-06, "loss": 0.4534, "step": 490 }, { "epoch": 0.06, "learning_rate": 2.968254349130174e-06, "loss": 0.1251, "step": 491 }, { "epoch": 0.06, "learning_rate": 2.968182363527295e-06, "loss": 0.0461, "step": 492 }, { "epoch": 0.06, "learning_rate": 2.968110377924415e-06, "loss": 0.0255, "step": 493 }, { "epoch": 0.06, "learning_rate": 2.968038392321536e-06, "loss": 0.0492, "step": 494 }, { "epoch": 0.06, "learning_rate": 2.9679664067186563e-06, "loss": 0.1044, "step": 495 }, { "epoch": 0.06, "learning_rate": 2.967894421115777e-06, "loss": 0.1041, "step": 496 }, { "epoch": 0.06, "learning_rate": 2.9678224355128975e-06, "loss": 0.0373, "step": 497 }, { "epoch": 0.06, "learning_rate": 2.9677504499100182e-06, "loss": 0.0248, "step": 498 }, { "epoch": 0.06, "learning_rate": 2.9676784643071386e-06, "loss": 0.0453, "step": 499 }, { "epoch": 0.06, "learning_rate": 2.9676064787042594e-06, "loss": 0.1927, "step": 500 }, { "epoch": 0.06, "learning_rate": 2.9675344931013797e-06, "loss": 0.5172, "step": 501 }, { "epoch": 0.06, "learning_rate": 2.9674625074985e-06, "loss": 0.0546, "step": 502 }, { "epoch": 0.06, "learning_rate": 2.967390521895621e-06, "loss": 0.0224, "step": 503 }, { "epoch": 0.06, "learning_rate": 2.967318536292741e-06, "loss": 0.1585, "step": 504 }, { "epoch": 0.06, "learning_rate": 2.967246550689862e-06, "loss": 0.1464, "step": 505 }, { "epoch": 0.06, "learning_rate": 2.9671745650869828e-06, "loss": 0.0471, "step": 506 }, { "epoch": 0.06, "learning_rate": 2.9671025794841035e-06, "loss": 0.1485, "step": 507 }, { "epoch": 0.06, "learning_rate": 2.967030593881224e-06, "loss": 0.0457, "step": 508 }, { "epoch": 0.06, "learning_rate": 2.9669586082783447e-06, "loss": 0.0468, "step": 509 }, { "epoch": 0.06, "learning_rate": 2.966886622675465e-06, "loss": 0.0488, "step": 510 }, { "epoch": 0.06, "learning_rate": 2.966814637072586e-06, "loss": 0.0847, "step": 511 }, { "epoch": 0.06, "learning_rate": 2.966742651469706e-06, "loss": 0.0476, "step": 512 }, { "epoch": 0.06, "learning_rate": 2.9666706658668265e-06, "loss": 0.0368, "step": 513 }, { "epoch": 0.06, "learning_rate": 2.9665986802639473e-06, "loss": 0.1264, "step": 514 }, { "epoch": 0.06, "learning_rate": 2.9665266946610677e-06, "loss": 0.0292, "step": 515 }, { "epoch": 0.06, "learning_rate": 2.9664547090581884e-06, "loss": 0.1513, "step": 516 }, { "epoch": 0.06, "learning_rate": 2.966382723455309e-06, "loss": 0.5006, "step": 517 }, { "epoch": 0.06, "learning_rate": 2.9663107378524296e-06, "loss": 0.0371, "step": 518 }, { "epoch": 0.06, "learning_rate": 2.96623875224955e-06, "loss": 0.1535, "step": 519 }, { "epoch": 0.06, "learning_rate": 2.9661667666466707e-06, "loss": 0.0338, "step": 520 }, { "epoch": 0.06, "learning_rate": 2.9660947810437915e-06, "loss": 0.0242, "step": 521 }, { "epoch": 0.06, "learning_rate": 2.966022795440912e-06, "loss": 0.0294, "step": 522 }, { "epoch": 0.06, "learning_rate": 2.9659508098380326e-06, "loss": 0.0245, "step": 523 }, { "epoch": 0.06, "learning_rate": 2.965878824235153e-06, "loss": 0.0431, "step": 524 }, { "epoch": 0.06, "learning_rate": 2.9658068386322737e-06, "loss": 0.4372, "step": 525 }, { "epoch": 0.06, "learning_rate": 2.965734853029394e-06, "loss": 0.0532, "step": 526 }, { "epoch": 0.06, "learning_rate": 2.965662867426515e-06, "loss": 0.207, "step": 527 }, { "epoch": 0.06, "learning_rate": 2.9655908818236352e-06, "loss": 0.0876, "step": 528 }, { "epoch": 0.06, "learning_rate": 2.965518896220756e-06, "loss": 0.0397, "step": 529 }, { "epoch": 0.06, "learning_rate": 2.9654469106178764e-06, "loss": 0.0411, "step": 530 }, { "epoch": 0.06, "learning_rate": 2.965374925014997e-06, "loss": 0.0978, "step": 531 }, { "epoch": 0.06, "learning_rate": 2.9653029394121175e-06, "loss": 0.0662, "step": 532 }, { "epoch": 0.06, "learning_rate": 2.9652309538092383e-06, "loss": 0.0495, "step": 533 }, { "epoch": 0.06, "learning_rate": 2.9651589682063586e-06, "loss": 0.0145, "step": 534 }, { "epoch": 0.06, "learning_rate": 2.9650869826034794e-06, "loss": 0.4768, "step": 535 }, { "epoch": 0.06, "learning_rate": 2.9650149970005998e-06, "loss": 0.1635, "step": 536 }, { "epoch": 0.06, "learning_rate": 2.9649430113977206e-06, "loss": 0.0581, "step": 537 }, { "epoch": 0.06, "learning_rate": 2.9648710257948413e-06, "loss": 0.5406, "step": 538 }, { "epoch": 0.06, "learning_rate": 2.9647990401919617e-06, "loss": 0.0566, "step": 539 }, { "epoch": 0.06, "learning_rate": 2.9647270545890825e-06, "loss": 0.4038, "step": 540 }, { "epoch": 0.06, "learning_rate": 2.964655068986203e-06, "loss": 0.6815, "step": 541 }, { "epoch": 0.06, "learning_rate": 2.9645830833833236e-06, "loss": 0.0272, "step": 542 }, { "epoch": 0.07, "learning_rate": 2.964511097780444e-06, "loss": 0.5433, "step": 543 }, { "epoch": 0.07, "learning_rate": 2.9644391121775647e-06, "loss": 0.072, "step": 544 }, { "epoch": 0.07, "learning_rate": 2.964367126574685e-06, "loss": 0.0997, "step": 545 }, { "epoch": 0.07, "learning_rate": 2.964295140971806e-06, "loss": 0.0245, "step": 546 }, { "epoch": 0.07, "learning_rate": 2.9642231553689262e-06, "loss": 0.0238, "step": 547 }, { "epoch": 0.07, "learning_rate": 2.9641511697660466e-06, "loss": 0.026, "step": 548 }, { "epoch": 0.07, "learning_rate": 2.9640791841631674e-06, "loss": 0.0347, "step": 549 }, { "epoch": 0.07, "learning_rate": 2.9640071985602877e-06, "loss": 0.2414, "step": 550 }, { "epoch": 0.07, "learning_rate": 2.9639352129574085e-06, "loss": 0.177, "step": 551 }, { "epoch": 0.07, "learning_rate": 2.9638632273545293e-06, "loss": 0.1023, "step": 552 }, { "epoch": 0.07, "learning_rate": 2.96379124175165e-06, "loss": 0.4882, "step": 553 }, { "epoch": 0.07, "learning_rate": 2.9637192561487704e-06, "loss": 0.0542, "step": 554 }, { "epoch": 0.07, "learning_rate": 2.963647270545891e-06, "loss": 0.016, "step": 555 }, { "epoch": 0.07, "learning_rate": 2.9635752849430115e-06, "loss": 0.0188, "step": 556 }, { "epoch": 0.07, "learning_rate": 2.9635032993401323e-06, "loss": 0.0682, "step": 557 }, { "epoch": 0.07, "learning_rate": 2.9634313137372527e-06, "loss": 0.0446, "step": 558 }, { "epoch": 0.07, "learning_rate": 2.963359328134373e-06, "loss": 0.507, "step": 559 }, { "epoch": 0.07, "learning_rate": 2.963287342531494e-06, "loss": 0.0402, "step": 560 }, { "epoch": 0.07, "learning_rate": 2.963215356928614e-06, "loss": 0.1083, "step": 561 }, { "epoch": 0.07, "learning_rate": 2.963143371325735e-06, "loss": 0.2035, "step": 562 }, { "epoch": 0.07, "learning_rate": 2.9630713857228553e-06, "loss": 0.4693, "step": 563 }, { "epoch": 0.07, "learning_rate": 2.962999400119976e-06, "loss": 0.0214, "step": 564 }, { "epoch": 0.07, "learning_rate": 2.9629274145170964e-06, "loss": 0.0172, "step": 565 }, { "epoch": 0.07, "learning_rate": 2.962855428914217e-06, "loss": 0.0313, "step": 566 }, { "epoch": 0.07, "learning_rate": 2.962783443311338e-06, "loss": 0.0273, "step": 567 }, { "epoch": 0.07, "learning_rate": 2.9627114577084583e-06, "loss": 0.5779, "step": 568 }, { "epoch": 0.07, "learning_rate": 2.962639472105579e-06, "loss": 0.3639, "step": 569 }, { "epoch": 0.07, "learning_rate": 2.9625674865026995e-06, "loss": 0.0223, "step": 570 }, { "epoch": 0.07, "learning_rate": 2.9624955008998202e-06, "loss": 0.0965, "step": 571 }, { "epoch": 0.07, "learning_rate": 2.9624235152969406e-06, "loss": 0.0453, "step": 572 }, { "epoch": 0.07, "learning_rate": 2.9623515296940614e-06, "loss": 0.0164, "step": 573 }, { "epoch": 0.07, "learning_rate": 2.9622795440911817e-06, "loss": 0.039, "step": 574 }, { "epoch": 0.07, "learning_rate": 2.9622075584883025e-06, "loss": 0.1934, "step": 575 }, { "epoch": 0.07, "learning_rate": 2.962135572885423e-06, "loss": 0.1061, "step": 576 }, { "epoch": 0.07, "learning_rate": 2.9620635872825436e-06, "loss": 0.1374, "step": 577 }, { "epoch": 0.07, "learning_rate": 2.961991601679664e-06, "loss": 0.1398, "step": 578 }, { "epoch": 0.07, "learning_rate": 2.9619196160767848e-06, "loss": 0.0214, "step": 579 }, { "epoch": 0.07, "learning_rate": 2.961847630473905e-06, "loss": 0.0255, "step": 580 }, { "epoch": 0.07, "learning_rate": 2.961775644871026e-06, "loss": 0.1336, "step": 581 }, { "epoch": 0.07, "learning_rate": 2.9617036592681463e-06, "loss": 0.0205, "step": 582 }, { "epoch": 0.07, "learning_rate": 2.961631673665267e-06, "loss": 0.0834, "step": 583 }, { "epoch": 0.07, "learning_rate": 2.961559688062388e-06, "loss": 0.4266, "step": 584 }, { "epoch": 0.07, "learning_rate": 2.961487702459508e-06, "loss": 0.0991, "step": 585 }, { "epoch": 0.07, "learning_rate": 2.961415716856629e-06, "loss": 0.0755, "step": 586 }, { "epoch": 0.07, "learning_rate": 2.9613437312537493e-06, "loss": 0.0856, "step": 587 }, { "epoch": 0.07, "learning_rate": 2.96127174565087e-06, "loss": 0.0173, "step": 588 }, { "epoch": 0.07, "learning_rate": 2.9611997600479904e-06, "loss": 0.0397, "step": 589 }, { "epoch": 0.07, "learning_rate": 2.9611277744451112e-06, "loss": 0.0644, "step": 590 }, { "epoch": 0.07, "learning_rate": 2.9610557888422316e-06, "loss": 0.2015, "step": 591 }, { "epoch": 0.07, "learning_rate": 2.9609838032393524e-06, "loss": 0.7372, "step": 592 }, { "epoch": 0.07, "learning_rate": 2.9609118176364727e-06, "loss": 0.0259, "step": 593 }, { "epoch": 0.07, "learning_rate": 2.960839832033593e-06, "loss": 0.0319, "step": 594 }, { "epoch": 0.07, "learning_rate": 2.960767846430714e-06, "loss": 0.0123, "step": 595 }, { "epoch": 0.07, "learning_rate": 2.960695860827834e-06, "loss": 0.2927, "step": 596 }, { "epoch": 0.07, "learning_rate": 2.960623875224955e-06, "loss": 0.5401, "step": 597 }, { "epoch": 0.07, "learning_rate": 2.9605518896220758e-06, "loss": 0.0225, "step": 598 }, { "epoch": 0.07, "learning_rate": 2.960479904019196e-06, "loss": 0.0496, "step": 599 }, { "epoch": 0.07, "learning_rate": 2.960407918416317e-06, "loss": 0.0211, "step": 600 }, { "epoch": 0.07, "learning_rate": 2.9603359328134377e-06, "loss": 0.1112, "step": 601 }, { "epoch": 0.07, "learning_rate": 2.960263947210558e-06, "loss": 0.0142, "step": 602 }, { "epoch": 0.07, "learning_rate": 2.960191961607679e-06, "loss": 0.4208, "step": 603 }, { "epoch": 0.07, "learning_rate": 2.960119976004799e-06, "loss": 0.6123, "step": 604 }, { "epoch": 0.07, "learning_rate": 2.9600479904019195e-06, "loss": 0.0248, "step": 605 }, { "epoch": 0.07, "learning_rate": 2.9599760047990403e-06, "loss": 0.0128, "step": 606 }, { "epoch": 0.07, "learning_rate": 2.9599040191961607e-06, "loss": 0.2291, "step": 607 }, { "epoch": 0.07, "learning_rate": 2.9598320335932814e-06, "loss": 0.4336, "step": 608 }, { "epoch": 0.07, "learning_rate": 2.9597600479904018e-06, "loss": 0.1119, "step": 609 }, { "epoch": 0.07, "learning_rate": 2.9596880623875226e-06, "loss": 0.1202, "step": 610 }, { "epoch": 0.07, "learning_rate": 2.959616076784643e-06, "loss": 0.4529, "step": 611 }, { "epoch": 0.07, "learning_rate": 2.9595440911817637e-06, "loss": 0.0171, "step": 612 }, { "epoch": 0.07, "learning_rate": 2.959472105578884e-06, "loss": 0.0156, "step": 613 }, { "epoch": 0.07, "learning_rate": 2.959400119976005e-06, "loss": 0.0202, "step": 614 }, { "epoch": 0.07, "learning_rate": 2.9593281343731256e-06, "loss": 0.0688, "step": 615 }, { "epoch": 0.07, "learning_rate": 2.959256148770246e-06, "loss": 0.0303, "step": 616 }, { "epoch": 0.07, "learning_rate": 2.9591841631673667e-06, "loss": 0.3074, "step": 617 }, { "epoch": 0.07, "learning_rate": 2.959112177564487e-06, "loss": 0.0429, "step": 618 }, { "epoch": 0.07, "learning_rate": 2.959040191961608e-06, "loss": 0.0252, "step": 619 }, { "epoch": 0.07, "learning_rate": 2.9589682063587282e-06, "loss": 0.2037, "step": 620 }, { "epoch": 0.07, "learning_rate": 2.958896220755849e-06, "loss": 0.0138, "step": 621 }, { "epoch": 0.07, "learning_rate": 2.9588242351529694e-06, "loss": 0.0154, "step": 622 }, { "epoch": 0.07, "learning_rate": 2.95875224955009e-06, "loss": 0.0233, "step": 623 }, { "epoch": 0.07, "learning_rate": 2.9586802639472105e-06, "loss": 0.1088, "step": 624 }, { "epoch": 0.07, "learning_rate": 2.9586082783443313e-06, "loss": 0.0508, "step": 625 }, { "epoch": 0.08, "learning_rate": 2.9585362927414516e-06, "loss": 0.0352, "step": 626 }, { "epoch": 0.08, "learning_rate": 2.9584643071385724e-06, "loss": 0.1063, "step": 627 }, { "epoch": 0.08, "learning_rate": 2.9583923215356928e-06, "loss": 0.0115, "step": 628 }, { "epoch": 0.08, "learning_rate": 2.9583203359328135e-06, "loss": 0.2322, "step": 629 }, { "epoch": 0.08, "learning_rate": 2.9582483503299343e-06, "loss": 0.2505, "step": 630 }, { "epoch": 0.08, "learning_rate": 2.9581763647270547e-06, "loss": 0.1762, "step": 631 }, { "epoch": 0.08, "learning_rate": 2.9581043791241755e-06, "loss": 0.2081, "step": 632 }, { "epoch": 0.08, "learning_rate": 2.958032393521296e-06, "loss": 0.2649, "step": 633 }, { "epoch": 0.08, "learning_rate": 2.9579604079184166e-06, "loss": 0.019, "step": 634 }, { "epoch": 0.08, "learning_rate": 2.957888422315537e-06, "loss": 0.0182, "step": 635 }, { "epoch": 0.08, "learning_rate": 2.9578164367126577e-06, "loss": 0.1205, "step": 636 }, { "epoch": 0.08, "learning_rate": 2.957744451109778e-06, "loss": 0.0139, "step": 637 }, { "epoch": 0.08, "learning_rate": 2.957672465506899e-06, "loss": 0.0079, "step": 638 }, { "epoch": 0.08, "learning_rate": 2.957600479904019e-06, "loss": 0.0159, "step": 639 }, { "epoch": 0.08, "learning_rate": 2.9575284943011396e-06, "loss": 0.0105, "step": 640 }, { "epoch": 0.08, "learning_rate": 2.9574565086982603e-06, "loss": 0.021, "step": 641 }, { "epoch": 0.08, "learning_rate": 2.9573845230953807e-06, "loss": 0.4002, "step": 642 }, { "epoch": 0.08, "learning_rate": 2.9573125374925015e-06, "loss": 0.0179, "step": 643 }, { "epoch": 0.08, "learning_rate": 2.9572405518896223e-06, "loss": 0.1368, "step": 644 }, { "epoch": 0.08, "learning_rate": 2.9571685662867426e-06, "loss": 1.1773, "step": 645 }, { "epoch": 0.08, "learning_rate": 2.9570965806838634e-06, "loss": 0.0292, "step": 646 }, { "epoch": 0.08, "learning_rate": 2.957024595080984e-06, "loss": 0.1339, "step": 647 }, { "epoch": 0.08, "learning_rate": 2.9569526094781045e-06, "loss": 0.6097, "step": 648 }, { "epoch": 0.08, "learning_rate": 2.9568806238752253e-06, "loss": 0.0583, "step": 649 }, { "epoch": 0.08, "learning_rate": 2.9568086382723457e-06, "loss": 0.3201, "step": 650 }, { "epoch": 0.08, "learning_rate": 2.956736652669466e-06, "loss": 0.0125, "step": 651 }, { "epoch": 0.08, "learning_rate": 2.956664667066587e-06, "loss": 0.2661, "step": 652 }, { "epoch": 0.08, "learning_rate": 2.956592681463707e-06, "loss": 0.6312, "step": 653 }, { "epoch": 0.08, "learning_rate": 2.956520695860828e-06, "loss": 0.0105, "step": 654 }, { "epoch": 0.08, "learning_rate": 2.9564487102579483e-06, "loss": 0.1882, "step": 655 }, { "epoch": 0.08, "learning_rate": 2.956376724655069e-06, "loss": 0.0459, "step": 656 }, { "epoch": 0.08, "learning_rate": 2.9563047390521894e-06, "loss": 0.0103, "step": 657 }, { "epoch": 0.08, "learning_rate": 2.95623275344931e-06, "loss": 0.0551, "step": 658 }, { "epoch": 0.08, "learning_rate": 2.9561607678464305e-06, "loss": 0.0316, "step": 659 }, { "epoch": 0.08, "learning_rate": 2.9560887822435513e-06, "loss": 0.0594, "step": 660 }, { "epoch": 0.08, "learning_rate": 2.956016796640672e-06, "loss": 0.4792, "step": 661 }, { "epoch": 0.08, "learning_rate": 2.955944811037793e-06, "loss": 0.0104, "step": 662 }, { "epoch": 0.08, "learning_rate": 2.9558728254349132e-06, "loss": 0.2586, "step": 663 }, { "epoch": 0.08, "learning_rate": 2.9558008398320336e-06, "loss": 0.4397, "step": 664 }, { "epoch": 0.08, "learning_rate": 2.9557288542291544e-06, "loss": 0.1137, "step": 665 }, { "epoch": 0.08, "learning_rate": 2.9556568686262747e-06, "loss": 0.0103, "step": 666 }, { "epoch": 0.08, "learning_rate": 2.9555848830233955e-06, "loss": 0.0136, "step": 667 }, { "epoch": 0.08, "learning_rate": 2.955512897420516e-06, "loss": 0.0124, "step": 668 }, { "epoch": 0.08, "learning_rate": 2.9554409118176366e-06, "loss": 0.0086, "step": 669 }, { "epoch": 0.08, "learning_rate": 2.955368926214757e-06, "loss": 0.0588, "step": 670 }, { "epoch": 0.08, "learning_rate": 2.9552969406118778e-06, "loss": 0.3912, "step": 671 }, { "epoch": 0.08, "learning_rate": 2.955224955008998e-06, "loss": 0.0176, "step": 672 }, { "epoch": 0.08, "learning_rate": 2.955152969406119e-06, "loss": 0.0271, "step": 673 }, { "epoch": 0.08, "learning_rate": 2.9550809838032393e-06, "loss": 0.2149, "step": 674 }, { "epoch": 0.08, "learning_rate": 2.95500899820036e-06, "loss": 0.0138, "step": 675 }, { "epoch": 0.08, "learning_rate": 2.954937012597481e-06, "loss": 0.0123, "step": 676 }, { "epoch": 0.08, "learning_rate": 2.954865026994601e-06, "loss": 0.0086, "step": 677 }, { "epoch": 0.08, "learning_rate": 2.954793041391722e-06, "loss": 0.0193, "step": 678 }, { "epoch": 0.08, "learning_rate": 2.9547210557888423e-06, "loss": 0.1436, "step": 679 }, { "epoch": 0.08, "learning_rate": 2.954649070185963e-06, "loss": 0.0304, "step": 680 }, { "epoch": 0.08, "learning_rate": 2.9545770845830834e-06, "loss": 0.6252, "step": 681 }, { "epoch": 0.08, "learning_rate": 2.9545050989802042e-06, "loss": 0.2068, "step": 682 }, { "epoch": 0.08, "learning_rate": 2.9544331133773246e-06, "loss": 0.0138, "step": 683 }, { "epoch": 0.08, "learning_rate": 2.9543611277744454e-06, "loss": 0.1119, "step": 684 }, { "epoch": 0.08, "learning_rate": 2.9542891421715657e-06, "loss": 0.4581, "step": 685 }, { "epoch": 0.08, "learning_rate": 2.954217156568686e-06, "loss": 0.0494, "step": 686 }, { "epoch": 0.08, "learning_rate": 2.954145170965807e-06, "loss": 0.0153, "step": 687 }, { "epoch": 0.08, "learning_rate": 2.954073185362927e-06, "loss": 0.0165, "step": 688 }, { "epoch": 0.08, "learning_rate": 2.954001199760048e-06, "loss": 0.0469, "step": 689 }, { "epoch": 0.08, "learning_rate": 2.9539292141571688e-06, "loss": 0.1013, "step": 690 }, { "epoch": 0.08, "learning_rate": 2.953857228554289e-06, "loss": 0.0433, "step": 691 }, { "epoch": 0.08, "learning_rate": 2.95378524295141e-06, "loss": 0.049, "step": 692 }, { "epoch": 0.08, "learning_rate": 2.9537132573485307e-06, "loss": 0.0108, "step": 693 }, { "epoch": 0.08, "learning_rate": 2.953641271745651e-06, "loss": 0.4539, "step": 694 }, { "epoch": 0.08, "learning_rate": 2.953569286142772e-06, "loss": 0.0429, "step": 695 }, { "epoch": 0.08, "learning_rate": 2.953497300539892e-06, "loss": 0.0096, "step": 696 }, { "epoch": 0.08, "learning_rate": 2.9534253149370125e-06, "loss": 0.0786, "step": 697 }, { "epoch": 0.08, "learning_rate": 2.9533533293341333e-06, "loss": 0.0829, "step": 698 }, { "epoch": 0.08, "learning_rate": 2.9532813437312536e-06, "loss": 0.5523, "step": 699 }, { "epoch": 0.08, "learning_rate": 2.9532093581283744e-06, "loss": 0.7013, "step": 700 }, { "epoch": 0.08, "learning_rate": 2.9531373725254948e-06, "loss": 0.0718, "step": 701 }, { "epoch": 0.08, "learning_rate": 2.9530653869226156e-06, "loss": 0.0085, "step": 702 }, { "epoch": 0.08, "learning_rate": 2.952993401319736e-06, "loss": 0.0095, "step": 703 }, { "epoch": 0.08, "learning_rate": 2.9529214157168567e-06, "loss": 0.014, "step": 704 }, { "epoch": 0.08, "learning_rate": 2.952849430113977e-06, "loss": 0.0076, "step": 705 }, { "epoch": 0.08, "learning_rate": 2.952777444511098e-06, "loss": 0.5079, "step": 706 }, { "epoch": 0.08, "learning_rate": 2.9527054589082186e-06, "loss": 0.0405, "step": 707 }, { "epoch": 0.08, "learning_rate": 2.9526334733053394e-06, "loss": 0.0117, "step": 708 }, { "epoch": 0.08, "learning_rate": 2.9525614877024597e-06, "loss": 0.0547, "step": 709 }, { "epoch": 0.09, "learning_rate": 2.95248950209958e-06, "loss": 0.0119, "step": 710 }, { "epoch": 0.09, "learning_rate": 2.952417516496701e-06, "loss": 0.074, "step": 711 }, { "epoch": 0.09, "learning_rate": 2.9523455308938212e-06, "loss": 0.0283, "step": 712 }, { "epoch": 0.09, "learning_rate": 2.952273545290942e-06, "loss": 0.3319, "step": 713 }, { "epoch": 0.09, "learning_rate": 2.9522015596880624e-06, "loss": 0.2662, "step": 714 }, { "epoch": 0.09, "learning_rate": 2.952129574085183e-06, "loss": 0.214, "step": 715 }, { "epoch": 0.09, "learning_rate": 2.9520575884823035e-06, "loss": 0.1363, "step": 716 }, { "epoch": 0.09, "learning_rate": 2.9519856028794243e-06, "loss": 0.5484, "step": 717 }, { "epoch": 0.09, "learning_rate": 2.9519136172765446e-06, "loss": 0.1755, "step": 718 }, { "epoch": 0.09, "learning_rate": 2.9518416316736654e-06, "loss": 0.0291, "step": 719 }, { "epoch": 0.09, "learning_rate": 2.9517696460707858e-06, "loss": 0.616, "step": 720 }, { "epoch": 0.09, "learning_rate": 2.9516976604679065e-06, "loss": 0.2248, "step": 721 }, { "epoch": 0.09, "learning_rate": 2.9516256748650273e-06, "loss": 0.0098, "step": 722 }, { "epoch": 0.09, "learning_rate": 2.9515536892621477e-06, "loss": 0.1568, "step": 723 }, { "epoch": 0.09, "learning_rate": 2.9514817036592684e-06, "loss": 0.0111, "step": 724 }, { "epoch": 0.09, "learning_rate": 2.951409718056389e-06, "loss": 0.0093, "step": 725 }, { "epoch": 0.09, "learning_rate": 2.9513377324535096e-06, "loss": 0.2537, "step": 726 }, { "epoch": 0.09, "learning_rate": 2.95126574685063e-06, "loss": 0.0273, "step": 727 }, { "epoch": 0.09, "learning_rate": 2.9511937612477507e-06, "loss": 0.4718, "step": 728 }, { "epoch": 0.09, "learning_rate": 2.951121775644871e-06, "loss": 0.0355, "step": 729 }, { "epoch": 0.09, "learning_rate": 2.951049790041992e-06, "loss": 0.0068, "step": 730 }, { "epoch": 0.09, "learning_rate": 2.950977804439112e-06, "loss": 0.2235, "step": 731 }, { "epoch": 0.09, "learning_rate": 2.9509058188362326e-06, "loss": 0.1346, "step": 732 }, { "epoch": 0.09, "learning_rate": 2.9508338332333533e-06, "loss": 0.0083, "step": 733 }, { "epoch": 0.09, "learning_rate": 2.9507618476304737e-06, "loss": 0.5519, "step": 734 }, { "epoch": 0.09, "learning_rate": 2.9506898620275945e-06, "loss": 0.0384, "step": 735 }, { "epoch": 0.09, "learning_rate": 2.950617876424715e-06, "loss": 0.0074, "step": 736 }, { "epoch": 0.09, "learning_rate": 2.9505458908218356e-06, "loss": 0.267, "step": 737 }, { "epoch": 0.09, "learning_rate": 2.9504739052189564e-06, "loss": 0.2328, "step": 738 }, { "epoch": 0.09, "learning_rate": 2.950401919616077e-06, "loss": 0.0154, "step": 739 }, { "epoch": 0.09, "learning_rate": 2.9503299340131975e-06, "loss": 0.2525, "step": 740 }, { "epoch": 0.09, "learning_rate": 2.9502579484103183e-06, "loss": 0.6324, "step": 741 }, { "epoch": 0.09, "learning_rate": 2.9501859628074386e-06, "loss": 0.031, "step": 742 }, { "epoch": 0.09, "learning_rate": 2.950113977204559e-06, "loss": 0.1184, "step": 743 }, { "epoch": 0.09, "learning_rate": 2.9500419916016798e-06, "loss": 0.0124, "step": 744 }, { "epoch": 0.09, "learning_rate": 2.9499700059988e-06, "loss": 0.7175, "step": 745 }, { "epoch": 0.09, "learning_rate": 2.949898020395921e-06, "loss": 0.0624, "step": 746 }, { "epoch": 0.09, "learning_rate": 2.9498260347930413e-06, "loss": 0.0639, "step": 747 }, { "epoch": 0.09, "learning_rate": 2.949754049190162e-06, "loss": 0.2815, "step": 748 }, { "epoch": 0.09, "learning_rate": 2.9496820635872824e-06, "loss": 0.6537, "step": 749 }, { "epoch": 0.09, "learning_rate": 2.949610077984403e-06, "loss": 0.0089, "step": 750 }, { "epoch": 0.09, "learning_rate": 2.9495380923815235e-06, "loss": 0.0098, "step": 751 }, { "epoch": 0.09, "learning_rate": 2.9494661067786443e-06, "loss": 0.013, "step": 752 }, { "epoch": 0.09, "learning_rate": 2.949394121175765e-06, "loss": 0.1348, "step": 753 }, { "epoch": 0.09, "learning_rate": 2.9493221355728855e-06, "loss": 0.1828, "step": 754 }, { "epoch": 0.09, "learning_rate": 2.9492501499700062e-06, "loss": 0.0151, "step": 755 }, { "epoch": 0.09, "learning_rate": 2.9491781643671266e-06, "loss": 0.0405, "step": 756 }, { "epoch": 0.09, "learning_rate": 2.9491061787642474e-06, "loss": 0.0281, "step": 757 }, { "epoch": 0.09, "learning_rate": 2.9490341931613677e-06, "loss": 0.0847, "step": 758 }, { "epoch": 0.09, "learning_rate": 2.9489622075584885e-06, "loss": 0.011, "step": 759 }, { "epoch": 0.09, "learning_rate": 2.948890221955609e-06, "loss": 0.3642, "step": 760 }, { "epoch": 0.09, "learning_rate": 2.9488182363527296e-06, "loss": 0.0101, "step": 761 }, { "epoch": 0.09, "learning_rate": 2.94874625074985e-06, "loss": 0.0088, "step": 762 }, { "epoch": 0.09, "learning_rate": 2.9486742651469708e-06, "loss": 0.0089, "step": 763 }, { "epoch": 0.09, "learning_rate": 2.948602279544091e-06, "loss": 0.0181, "step": 764 }, { "epoch": 0.09, "learning_rate": 2.948530293941212e-06, "loss": 0.0226, "step": 765 }, { "epoch": 0.09, "learning_rate": 2.9484583083383323e-06, "loss": 0.4153, "step": 766 }, { "epoch": 0.09, "learning_rate": 2.948386322735453e-06, "loss": 0.022, "step": 767 }, { "epoch": 0.09, "learning_rate": 2.9483143371325734e-06, "loss": 0.3288, "step": 768 }, { "epoch": 0.09, "learning_rate": 2.948242351529694e-06, "loss": 0.0104, "step": 769 }, { "epoch": 0.09, "learning_rate": 2.948170365926815e-06, "loss": 0.171, "step": 770 }, { "epoch": 0.09, "learning_rate": 2.9480983803239353e-06, "loss": 0.0072, "step": 771 }, { "epoch": 0.09, "learning_rate": 2.948026394721056e-06, "loss": 0.0116, "step": 772 }, { "epoch": 0.09, "learning_rate": 2.9479544091181764e-06, "loss": 0.0578, "step": 773 }, { "epoch": 0.09, "learning_rate": 2.947882423515297e-06, "loss": 0.7039, "step": 774 }, { "epoch": 0.09, "learning_rate": 2.9478104379124176e-06, "loss": 0.0387, "step": 775 }, { "epoch": 0.09, "learning_rate": 2.9477384523095383e-06, "loss": 0.0142, "step": 776 }, { "epoch": 0.09, "learning_rate": 2.9476664667066587e-06, "loss": 0.1318, "step": 777 }, { "epoch": 0.09, "learning_rate": 2.947594481103779e-06, "loss": 0.0569, "step": 778 }, { "epoch": 0.09, "learning_rate": 2.9475224955009e-06, "loss": 0.0911, "step": 779 }, { "epoch": 0.09, "learning_rate": 2.94745050989802e-06, "loss": 0.6561, "step": 780 }, { "epoch": 0.09, "learning_rate": 2.947378524295141e-06, "loss": 0.3069, "step": 781 }, { "epoch": 0.09, "learning_rate": 2.9473065386922613e-06, "loss": 0.615, "step": 782 }, { "epoch": 0.09, "learning_rate": 2.947234553089382e-06, "loss": 0.0155, "step": 783 }, { "epoch": 0.09, "learning_rate": 2.947162567486503e-06, "loss": 0.0066, "step": 784 }, { "epoch": 0.09, "learning_rate": 2.9470905818836237e-06, "loss": 0.0228, "step": 785 }, { "epoch": 0.09, "learning_rate": 2.947018596280744e-06, "loss": 0.0086, "step": 786 }, { "epoch": 0.09, "learning_rate": 2.946946610677865e-06, "loss": 0.0227, "step": 787 }, { "epoch": 0.09, "learning_rate": 2.946874625074985e-06, "loss": 0.1147, "step": 788 }, { "epoch": 0.09, "learning_rate": 2.946802639472106e-06, "loss": 0.0194, "step": 789 }, { "epoch": 0.09, "learning_rate": 2.9467306538692263e-06, "loss": 0.0549, "step": 790 }, { "epoch": 0.09, "learning_rate": 2.9466586682663466e-06, "loss": 0.127, "step": 791 }, { "epoch": 0.09, "learning_rate": 2.9465866826634674e-06, "loss": 0.008, "step": 792 }, { "epoch": 0.1, "learning_rate": 2.9465146970605878e-06, "loss": 0.0313, "step": 793 }, { "epoch": 0.1, "learning_rate": 2.9464427114577085e-06, "loss": 0.0055, "step": 794 }, { "epoch": 0.1, "learning_rate": 2.946370725854829e-06, "loss": 0.3775, "step": 795 }, { "epoch": 0.1, "learning_rate": 2.9462987402519497e-06, "loss": 0.4489, "step": 796 }, { "epoch": 0.1, "learning_rate": 2.94622675464907e-06, "loss": 0.7177, "step": 797 }, { "epoch": 0.1, "learning_rate": 2.946154769046191e-06, "loss": 0.5371, "step": 798 }, { "epoch": 0.1, "learning_rate": 2.9460827834433116e-06, "loss": 0.0081, "step": 799 }, { "epoch": 0.1, "learning_rate": 2.946010797840432e-06, "loss": 0.0256, "step": 800 }, { "epoch": 0.1, "learning_rate": 2.9459388122375527e-06, "loss": 0.6656, "step": 801 }, { "epoch": 0.1, "learning_rate": 2.945866826634673e-06, "loss": 0.014, "step": 802 }, { "epoch": 0.1, "learning_rate": 2.945794841031794e-06, "loss": 0.2495, "step": 803 }, { "epoch": 0.1, "learning_rate": 2.9457228554289142e-06, "loss": 0.0134, "step": 804 }, { "epoch": 0.1, "learning_rate": 2.945650869826035e-06, "loss": 0.6982, "step": 805 }, { "epoch": 0.1, "learning_rate": 2.9455788842231553e-06, "loss": 0.0085, "step": 806 }, { "epoch": 0.1, "learning_rate": 2.945506898620276e-06, "loss": 0.0085, "step": 807 }, { "epoch": 0.1, "learning_rate": 2.9454349130173965e-06, "loss": 0.3387, "step": 808 }, { "epoch": 0.1, "learning_rate": 2.9453629274145173e-06, "loss": 0.1715, "step": 809 }, { "epoch": 0.1, "learning_rate": 2.9452909418116376e-06, "loss": 0.0598, "step": 810 }, { "epoch": 0.1, "learning_rate": 2.9452189562087584e-06, "loss": 0.0095, "step": 811 }, { "epoch": 0.1, "learning_rate": 2.9451469706058787e-06, "loss": 0.0497, "step": 812 }, { "epoch": 0.1, "learning_rate": 2.9450749850029995e-06, "loss": 0.2021, "step": 813 }, { "epoch": 0.1, "learning_rate": 2.94500299940012e-06, "loss": 0.0071, "step": 814 }, { "epoch": 0.1, "learning_rate": 2.9449310137972407e-06, "loss": 0.5565, "step": 815 }, { "epoch": 0.1, "learning_rate": 2.9448590281943614e-06, "loss": 0.0951, "step": 816 }, { "epoch": 0.1, "learning_rate": 2.944787042591482e-06, "loss": 0.0489, "step": 817 }, { "epoch": 0.1, "learning_rate": 2.9447150569886026e-06, "loss": 0.5777, "step": 818 }, { "epoch": 0.1, "learning_rate": 2.944643071385723e-06, "loss": 0.0088, "step": 819 }, { "epoch": 0.1, "learning_rate": 2.9445710857828437e-06, "loss": 0.3695, "step": 820 }, { "epoch": 0.1, "learning_rate": 2.944499100179964e-06, "loss": 0.0066, "step": 821 }, { "epoch": 0.1, "learning_rate": 2.944427114577085e-06, "loss": 0.0165, "step": 822 }, { "epoch": 0.1, "learning_rate": 2.944355128974205e-06, "loss": 0.0271, "step": 823 }, { "epoch": 0.1, "learning_rate": 2.9442831433713255e-06, "loss": 0.009, "step": 824 }, { "epoch": 0.1, "learning_rate": 2.9442111577684463e-06, "loss": 0.0073, "step": 825 }, { "epoch": 0.1, "learning_rate": 2.9441391721655667e-06, "loss": 0.0076, "step": 826 }, { "epoch": 0.1, "learning_rate": 2.9440671865626875e-06, "loss": 0.2882, "step": 827 }, { "epoch": 0.1, "learning_rate": 2.943995200959808e-06, "loss": 0.0272, "step": 828 }, { "epoch": 0.1, "learning_rate": 2.9439232153569286e-06, "loss": 0.0518, "step": 829 }, { "epoch": 0.1, "learning_rate": 2.9438512297540494e-06, "loss": 0.0063, "step": 830 }, { "epoch": 0.1, "learning_rate": 2.94377924415117e-06, "loss": 0.0078, "step": 831 }, { "epoch": 0.1, "learning_rate": 2.9437072585482905e-06, "loss": 0.0126, "step": 832 }, { "epoch": 0.1, "learning_rate": 2.9436352729454113e-06, "loss": 0.1116, "step": 833 }, { "epoch": 0.1, "learning_rate": 2.9435632873425316e-06, "loss": 0.0066, "step": 834 }, { "epoch": 0.1, "learning_rate": 2.9434913017396524e-06, "loss": 0.0109, "step": 835 }, { "epoch": 0.1, "learning_rate": 2.9434193161367728e-06, "loss": 0.0112, "step": 836 }, { "epoch": 0.1, "learning_rate": 2.943347330533893e-06, "loss": 0.4592, "step": 837 }, { "epoch": 0.1, "learning_rate": 2.943275344931014e-06, "loss": 1.1572, "step": 838 }, { "epoch": 0.1, "learning_rate": 2.9432033593281343e-06, "loss": 0.3896, "step": 839 }, { "epoch": 0.1, "learning_rate": 2.943131373725255e-06, "loss": 0.0071, "step": 840 }, { "epoch": 0.1, "learning_rate": 2.9430593881223754e-06, "loss": 0.0059, "step": 841 }, { "epoch": 0.1, "learning_rate": 2.942987402519496e-06, "loss": 0.0047, "step": 842 }, { "epoch": 0.1, "learning_rate": 2.9429154169166165e-06, "loss": 0.0108, "step": 843 }, { "epoch": 0.1, "learning_rate": 2.9428434313137373e-06, "loss": 0.0715, "step": 844 }, { "epoch": 0.1, "learning_rate": 2.942771445710858e-06, "loss": 0.8014, "step": 845 }, { "epoch": 0.1, "learning_rate": 2.9426994601079784e-06, "loss": 0.2266, "step": 846 }, { "epoch": 0.1, "learning_rate": 2.9426274745050992e-06, "loss": 0.383, "step": 847 }, { "epoch": 0.1, "learning_rate": 2.9425554889022196e-06, "loss": 0.0522, "step": 848 }, { "epoch": 0.1, "learning_rate": 2.9424835032993404e-06, "loss": 0.0874, "step": 849 }, { "epoch": 0.1, "learning_rate": 2.9424115176964607e-06, "loss": 0.0069, "step": 850 }, { "epoch": 0.1, "learning_rate": 2.9423395320935815e-06, "loss": 0.0075, "step": 851 }, { "epoch": 0.1, "learning_rate": 2.942267546490702e-06, "loss": 0.2684, "step": 852 }, { "epoch": 0.1, "learning_rate": 2.9421955608878226e-06, "loss": 0.0086, "step": 853 }, { "epoch": 0.1, "learning_rate": 2.942123575284943e-06, "loss": 0.0212, "step": 854 }, { "epoch": 0.1, "learning_rate": 2.9420515896820638e-06, "loss": 0.0097, "step": 855 }, { "epoch": 0.1, "learning_rate": 2.941979604079184e-06, "loss": 0.2457, "step": 856 }, { "epoch": 0.1, "learning_rate": 2.941907618476305e-06, "loss": 0.0054, "step": 857 }, { "epoch": 0.1, "learning_rate": 2.9418356328734252e-06, "loss": 1.202, "step": 858 }, { "epoch": 0.1, "learning_rate": 2.9417636472705456e-06, "loss": 0.0141, "step": 859 }, { "epoch": 0.1, "learning_rate": 2.9416916616676664e-06, "loss": 0.3258, "step": 860 }, { "epoch": 0.1, "learning_rate": 2.941619676064787e-06, "loss": 0.0082, "step": 861 }, { "epoch": 0.1, "learning_rate": 2.941547690461908e-06, "loss": 0.0667, "step": 862 }, { "epoch": 0.1, "learning_rate": 2.9414757048590283e-06, "loss": 0.0062, "step": 863 }, { "epoch": 0.1, "learning_rate": 2.941403719256149e-06, "loss": 0.0319, "step": 864 }, { "epoch": 0.1, "learning_rate": 2.9413317336532694e-06, "loss": 0.0224, "step": 865 }, { "epoch": 0.1, "learning_rate": 2.94125974805039e-06, "loss": 0.0099, "step": 866 }, { "epoch": 0.1, "learning_rate": 2.9411877624475106e-06, "loss": 0.2746, "step": 867 }, { "epoch": 0.1, "learning_rate": 2.9411157768446313e-06, "loss": 0.442, "step": 868 }, { "epoch": 0.1, "learning_rate": 2.9410437912417517e-06, "loss": 0.285, "step": 869 }, { "epoch": 0.1, "learning_rate": 2.940971805638872e-06, "loss": 0.0561, "step": 870 }, { "epoch": 0.1, "learning_rate": 2.940899820035993e-06, "loss": 0.0082, "step": 871 }, { "epoch": 0.1, "learning_rate": 2.940827834433113e-06, "loss": 0.018, "step": 872 }, { "epoch": 0.1, "learning_rate": 2.940755848830234e-06, "loss": 0.0054, "step": 873 }, { "epoch": 0.1, "learning_rate": 2.9406838632273543e-06, "loss": 0.0135, "step": 874 }, { "epoch": 0.1, "learning_rate": 2.940611877624475e-06, "loss": 0.5306, "step": 875 }, { "epoch": 0.1, "learning_rate": 2.940539892021596e-06, "loss": 0.0077, "step": 876 }, { "epoch": 0.11, "learning_rate": 2.9404679064187162e-06, "loss": 0.0078, "step": 877 }, { "epoch": 0.11, "learning_rate": 2.940395920815837e-06, "loss": 0.0113, "step": 878 }, { "epoch": 0.11, "learning_rate": 2.9403239352129578e-06, "loss": 0.057, "step": 879 }, { "epoch": 0.11, "learning_rate": 2.940251949610078e-06, "loss": 0.2122, "step": 880 }, { "epoch": 0.11, "learning_rate": 2.940179964007199e-06, "loss": 0.0085, "step": 881 }, { "epoch": 0.11, "learning_rate": 2.9401079784043193e-06, "loss": 0.0043, "step": 882 }, { "epoch": 0.11, "learning_rate": 2.9400359928014396e-06, "loss": 0.4066, "step": 883 }, { "epoch": 0.11, "learning_rate": 2.9399640071985604e-06, "loss": 0.0054, "step": 884 }, { "epoch": 0.11, "learning_rate": 2.9398920215956808e-06, "loss": 0.5227, "step": 885 }, { "epoch": 0.11, "learning_rate": 2.9398200359928015e-06, "loss": 0.5164, "step": 886 }, { "epoch": 0.11, "learning_rate": 2.939748050389922e-06, "loss": 0.0292, "step": 887 }, { "epoch": 0.11, "learning_rate": 2.9396760647870427e-06, "loss": 0.2606, "step": 888 }, { "epoch": 0.11, "learning_rate": 2.939604079184163e-06, "loss": 0.2841, "step": 889 }, { "epoch": 0.11, "learning_rate": 2.939532093581284e-06, "loss": 0.0058, "step": 890 }, { "epoch": 0.11, "learning_rate": 2.939460107978404e-06, "loss": 0.0061, "step": 891 }, { "epoch": 0.11, "learning_rate": 2.939388122375525e-06, "loss": 0.1645, "step": 892 }, { "epoch": 0.11, "learning_rate": 2.9393161367726457e-06, "loss": 0.411, "step": 893 }, { "epoch": 0.11, "learning_rate": 2.939244151169766e-06, "loss": 0.0062, "step": 894 }, { "epoch": 0.11, "learning_rate": 2.939172165566887e-06, "loss": 0.0148, "step": 895 }, { "epoch": 0.11, "learning_rate": 2.939100179964007e-06, "loss": 0.0363, "step": 896 }, { "epoch": 0.11, "learning_rate": 2.939028194361128e-06, "loss": 0.4786, "step": 897 }, { "epoch": 0.11, "learning_rate": 2.9389562087582483e-06, "loss": 0.6193, "step": 898 }, { "epoch": 0.11, "learning_rate": 2.938884223155369e-06, "loss": 0.9876, "step": 899 }, { "epoch": 0.11, "learning_rate": 2.9388122375524895e-06, "loss": 0.0792, "step": 900 }, { "epoch": 0.11, "learning_rate": 2.9387402519496103e-06, "loss": 0.1937, "step": 901 }, { "epoch": 0.11, "learning_rate": 2.9386682663467306e-06, "loss": 0.0084, "step": 902 }, { "epoch": 0.11, "learning_rate": 2.9385962807438514e-06, "loss": 0.0214, "step": 903 }, { "epoch": 0.11, "learning_rate": 2.9385242951409717e-06, "loss": 0.0069, "step": 904 }, { "epoch": 0.11, "learning_rate": 2.938452309538092e-06, "loss": 0.0074, "step": 905 }, { "epoch": 0.11, "learning_rate": 2.938380323935213e-06, "loss": 0.0157, "step": 906 }, { "epoch": 0.11, "learning_rate": 2.9383083383323337e-06, "loss": 0.0068, "step": 907 }, { "epoch": 0.11, "learning_rate": 2.9382363527294544e-06, "loss": 0.0065, "step": 908 }, { "epoch": 0.11, "learning_rate": 2.9381643671265748e-06, "loss": 0.0058, "step": 909 }, { "epoch": 0.11, "learning_rate": 2.9380923815236956e-06, "loss": 0.0075, "step": 910 }, { "epoch": 0.11, "learning_rate": 2.938020395920816e-06, "loss": 0.0048, "step": 911 }, { "epoch": 0.11, "learning_rate": 2.9379484103179367e-06, "loss": 0.0069, "step": 912 }, { "epoch": 0.11, "learning_rate": 2.937876424715057e-06, "loss": 0.0072, "step": 913 }, { "epoch": 0.11, "learning_rate": 2.937804439112178e-06, "loss": 0.5336, "step": 914 }, { "epoch": 0.11, "learning_rate": 2.937732453509298e-06, "loss": 0.0086, "step": 915 }, { "epoch": 0.11, "learning_rate": 2.937660467906419e-06, "loss": 0.0042, "step": 916 }, { "epoch": 0.11, "learning_rate": 2.9375884823035393e-06, "loss": 0.0128, "step": 917 }, { "epoch": 0.11, "learning_rate": 2.9375164967006597e-06, "loss": 0.0118, "step": 918 }, { "epoch": 0.11, "learning_rate": 2.9374445110977805e-06, "loss": 0.9113, "step": 919 }, { "epoch": 0.11, "learning_rate": 2.937372525494901e-06, "loss": 0.0055, "step": 920 }, { "epoch": 0.11, "learning_rate": 2.9373005398920216e-06, "loss": 0.0228, "step": 921 }, { "epoch": 0.11, "learning_rate": 2.9372285542891424e-06, "loss": 0.0124, "step": 922 }, { "epoch": 0.11, "learning_rate": 2.9371565686862627e-06, "loss": 0.0223, "step": 923 }, { "epoch": 0.11, "learning_rate": 2.9370845830833835e-06, "loss": 0.0048, "step": 924 }, { "epoch": 0.11, "learning_rate": 2.9370125974805043e-06, "loss": 0.4606, "step": 925 }, { "epoch": 0.11, "learning_rate": 2.9369406118776246e-06, "loss": 0.0086, "step": 926 }, { "epoch": 0.11, "learning_rate": 2.9368686262747454e-06, "loss": 0.005, "step": 927 }, { "epoch": 0.11, "learning_rate": 2.9367966406718658e-06, "loss": 0.0056, "step": 928 }, { "epoch": 0.11, "learning_rate": 2.936724655068986e-06, "loss": 0.3557, "step": 929 }, { "epoch": 0.11, "learning_rate": 2.936652669466107e-06, "loss": 0.0159, "step": 930 }, { "epoch": 0.11, "learning_rate": 2.9365806838632273e-06, "loss": 0.6391, "step": 931 }, { "epoch": 0.11, "learning_rate": 2.936508698260348e-06, "loss": 0.2628, "step": 932 }, { "epoch": 0.11, "learning_rate": 2.9364367126574684e-06, "loss": 0.0059, "step": 933 }, { "epoch": 0.11, "learning_rate": 2.936364727054589e-06, "loss": 0.6658, "step": 934 }, { "epoch": 0.11, "learning_rate": 2.9362927414517095e-06, "loss": 0.0137, "step": 935 }, { "epoch": 0.11, "learning_rate": 2.9362207558488303e-06, "loss": 0.0084, "step": 936 }, { "epoch": 0.11, "learning_rate": 2.9361487702459507e-06, "loss": 0.0122, "step": 937 }, { "epoch": 0.11, "learning_rate": 2.9360767846430714e-06, "loss": 0.0065, "step": 938 }, { "epoch": 0.11, "learning_rate": 2.9360047990401922e-06, "loss": 0.0044, "step": 939 }, { "epoch": 0.11, "learning_rate": 2.9359328134373126e-06, "loss": 0.2077, "step": 940 }, { "epoch": 0.11, "learning_rate": 2.9358608278344333e-06, "loss": 0.0056, "step": 941 }, { "epoch": 0.11, "learning_rate": 2.9357888422315537e-06, "loss": 0.0047, "step": 942 }, { "epoch": 0.11, "learning_rate": 2.9357168566286745e-06, "loss": 0.0055, "step": 943 }, { "epoch": 0.11, "learning_rate": 2.935644871025795e-06, "loss": 0.012, "step": 944 }, { "epoch": 0.11, "learning_rate": 2.9355728854229156e-06, "loss": 0.0055, "step": 945 }, { "epoch": 0.11, "learning_rate": 2.935500899820036e-06, "loss": 0.0058, "step": 946 }, { "epoch": 0.11, "learning_rate": 2.9354289142171567e-06, "loss": 0.0217, "step": 947 }, { "epoch": 0.11, "learning_rate": 2.935356928614277e-06, "loss": 0.0073, "step": 948 }, { "epoch": 0.11, "learning_rate": 2.935284943011398e-06, "loss": 0.0096, "step": 949 }, { "epoch": 0.11, "learning_rate": 2.9352129574085182e-06, "loss": 0.0063, "step": 950 }, { "epoch": 0.11, "learning_rate": 2.9351409718056386e-06, "loss": 0.0079, "step": 951 }, { "epoch": 0.11, "learning_rate": 2.9350689862027594e-06, "loss": 0.006, "step": 952 }, { "epoch": 0.11, "learning_rate": 2.93499700059988e-06, "loss": 0.0083, "step": 953 }, { "epoch": 0.11, "learning_rate": 2.934925014997001e-06, "loss": 0.009, "step": 954 }, { "epoch": 0.11, "learning_rate": 2.9348530293941213e-06, "loss": 0.0192, "step": 955 }, { "epoch": 0.11, "learning_rate": 2.934781043791242e-06, "loss": 0.4174, "step": 956 }, { "epoch": 0.11, "learning_rate": 2.9347090581883624e-06, "loss": 0.1165, "step": 957 }, { "epoch": 0.11, "learning_rate": 2.934637072585483e-06, "loss": 0.0048, "step": 958 }, { "epoch": 0.11, "learning_rate": 2.9345650869826035e-06, "loss": 0.0056, "step": 959 }, { "epoch": 0.12, "learning_rate": 2.9344931013797243e-06, "loss": 0.0047, "step": 960 }, { "epoch": 0.12, "learning_rate": 2.9344211157768447e-06, "loss": 0.0056, "step": 961 }, { "epoch": 0.12, "learning_rate": 2.9343491301739655e-06, "loss": 0.4182, "step": 962 }, { "epoch": 0.12, "learning_rate": 2.934277144571086e-06, "loss": 0.004, "step": 963 }, { "epoch": 0.12, "learning_rate": 2.934205158968206e-06, "loss": 0.0055, "step": 964 }, { "epoch": 0.12, "learning_rate": 2.934133173365327e-06, "loss": 0.1557, "step": 965 }, { "epoch": 0.12, "learning_rate": 2.9340611877624473e-06, "loss": 0.004, "step": 966 }, { "epoch": 0.12, "learning_rate": 2.933989202159568e-06, "loss": 0.0055, "step": 967 }, { "epoch": 0.12, "learning_rate": 2.933917216556689e-06, "loss": 0.0056, "step": 968 }, { "epoch": 0.12, "learning_rate": 2.9338452309538092e-06, "loss": 0.0078, "step": 969 }, { "epoch": 0.12, "learning_rate": 2.93377324535093e-06, "loss": 0.0196, "step": 970 }, { "epoch": 0.12, "learning_rate": 2.9337012597480508e-06, "loss": 0.0056, "step": 971 }, { "epoch": 0.12, "learning_rate": 2.933629274145171e-06, "loss": 0.5429, "step": 972 }, { "epoch": 0.12, "learning_rate": 2.933557288542292e-06, "loss": 0.0167, "step": 973 }, { "epoch": 0.12, "learning_rate": 2.9334853029394123e-06, "loss": 0.2455, "step": 974 }, { "epoch": 0.12, "learning_rate": 2.9334133173365326e-06, "loss": 0.4971, "step": 975 }, { "epoch": 0.12, "learning_rate": 2.9333413317336534e-06, "loss": 0.0066, "step": 976 }, { "epoch": 0.12, "learning_rate": 2.9332693461307738e-06, "loss": 0.0039, "step": 977 }, { "epoch": 0.12, "learning_rate": 2.9331973605278945e-06, "loss": 0.0052, "step": 978 }, { "epoch": 0.12, "learning_rate": 2.933125374925015e-06, "loss": 0.005, "step": 979 }, { "epoch": 0.12, "learning_rate": 2.9330533893221357e-06, "loss": 0.5348, "step": 980 }, { "epoch": 0.12, "learning_rate": 2.932981403719256e-06, "loss": 0.005, "step": 981 }, { "epoch": 0.12, "learning_rate": 2.932909418116377e-06, "loss": 0.004, "step": 982 }, { "epoch": 0.12, "learning_rate": 2.932837432513497e-06, "loss": 0.0048, "step": 983 }, { "epoch": 0.12, "learning_rate": 2.932765446910618e-06, "loss": 0.0041, "step": 984 }, { "epoch": 0.12, "learning_rate": 2.9326934613077387e-06, "loss": 0.0068, "step": 985 }, { "epoch": 0.12, "learning_rate": 2.932621475704859e-06, "loss": 0.0041, "step": 986 }, { "epoch": 0.12, "learning_rate": 2.93254949010198e-06, "loss": 0.4329, "step": 987 }, { "epoch": 0.12, "learning_rate": 2.9324775044991e-06, "loss": 0.0045, "step": 988 }, { "epoch": 0.12, "learning_rate": 2.932405518896221e-06, "loss": 0.0049, "step": 989 }, { "epoch": 0.12, "learning_rate": 2.9323335332933413e-06, "loss": 0.0136, "step": 990 }, { "epoch": 0.12, "learning_rate": 2.932261547690462e-06, "loss": 0.0045, "step": 991 }, { "epoch": 0.12, "learning_rate": 2.9321895620875825e-06, "loss": 0.0295, "step": 992 }, { "epoch": 0.12, "learning_rate": 2.9321175764847032e-06, "loss": 0.0042, "step": 993 }, { "epoch": 0.12, "learning_rate": 2.9320455908818236e-06, "loss": 0.4762, "step": 994 }, { "epoch": 0.12, "learning_rate": 2.9319736052789444e-06, "loss": 0.0712, "step": 995 }, { "epoch": 0.12, "learning_rate": 2.9319016196760647e-06, "loss": 0.5133, "step": 996 }, { "epoch": 0.12, "learning_rate": 2.931829634073185e-06, "loss": 0.004, "step": 997 }, { "epoch": 0.12, "learning_rate": 2.931757648470306e-06, "loss": 0.0033, "step": 998 }, { "epoch": 0.12, "learning_rate": 2.9316856628674266e-06, "loss": 0.0436, "step": 999 }, { "epoch": 0.12, "learning_rate": 2.931613677264547e-06, "loss": 0.5746, "step": 1000 }, { "epoch": 0.12, "learning_rate": 2.9315416916616678e-06, "loss": 0.0445, "step": 1001 }, { "epoch": 0.12, "learning_rate": 2.9314697060587886e-06, "loss": 0.0381, "step": 1002 }, { "epoch": 0.12, "learning_rate": 2.931397720455909e-06, "loss": 0.023, "step": 1003 }, { "epoch": 0.12, "learning_rate": 2.9313257348530297e-06, "loss": 0.0033, "step": 1004 }, { "epoch": 0.12, "learning_rate": 2.93125374925015e-06, "loss": 0.0122, "step": 1005 }, { "epoch": 0.12, "learning_rate": 2.931181763647271e-06, "loss": 0.0038, "step": 1006 }, { "epoch": 0.12, "learning_rate": 2.931109778044391e-06, "loss": 0.0465, "step": 1007 }, { "epoch": 0.12, "learning_rate": 2.931037792441512e-06, "loss": 0.2604, "step": 1008 }, { "epoch": 0.12, "learning_rate": 2.9309658068386323e-06, "loss": 0.0359, "step": 1009 }, { "epoch": 0.12, "learning_rate": 2.9308938212357527e-06, "loss": 0.0044, "step": 1010 }, { "epoch": 0.12, "learning_rate": 2.9308218356328734e-06, "loss": 0.0082, "step": 1011 }, { "epoch": 0.12, "learning_rate": 2.930749850029994e-06, "loss": 0.0041, "step": 1012 }, { "epoch": 0.12, "learning_rate": 2.9306778644271146e-06, "loss": 0.7637, "step": 1013 }, { "epoch": 0.12, "learning_rate": 2.930605878824235e-06, "loss": 0.0041, "step": 1014 }, { "epoch": 0.12, "learning_rate": 2.9305338932213557e-06, "loss": 0.3917, "step": 1015 }, { "epoch": 0.12, "learning_rate": 2.9304619076184765e-06, "loss": 0.0037, "step": 1016 }, { "epoch": 0.12, "learning_rate": 2.9303899220155973e-06, "loss": 0.0219, "step": 1017 }, { "epoch": 0.12, "learning_rate": 2.9303179364127176e-06, "loss": 0.0078, "step": 1018 }, { "epoch": 0.12, "learning_rate": 2.9302459508098384e-06, "loss": 0.2135, "step": 1019 }, { "epoch": 0.12, "learning_rate": 2.9301739652069588e-06, "loss": 0.0046, "step": 1020 }, { "epoch": 0.12, "learning_rate": 2.930101979604079e-06, "loss": 0.0026, "step": 1021 }, { "epoch": 0.12, "learning_rate": 2.9300299940012e-06, "loss": 0.0033, "step": 1022 }, { "epoch": 0.12, "learning_rate": 2.9299580083983202e-06, "loss": 0.0134, "step": 1023 }, { "epoch": 0.12, "learning_rate": 2.929886022795441e-06, "loss": 0.0036, "step": 1024 }, { "epoch": 0.12, "learning_rate": 2.9298140371925614e-06, "loss": 0.0048, "step": 1025 }, { "epoch": 0.12, "learning_rate": 2.929742051589682e-06, "loss": 0.0152, "step": 1026 }, { "epoch": 0.12, "learning_rate": 2.9296700659868025e-06, "loss": 0.0058, "step": 1027 }, { "epoch": 0.12, "learning_rate": 2.9295980803839233e-06, "loss": 0.5598, "step": 1028 }, { "epoch": 0.12, "learning_rate": 2.9295260947810436e-06, "loss": 0.0041, "step": 1029 }, { "epoch": 0.12, "learning_rate": 2.9294541091781644e-06, "loss": 0.6253, "step": 1030 }, { "epoch": 0.12, "learning_rate": 2.929382123575285e-06, "loss": 0.0035, "step": 1031 }, { "epoch": 0.12, "learning_rate": 2.9293101379724056e-06, "loss": 0.0047, "step": 1032 }, { "epoch": 0.12, "learning_rate": 2.9292381523695263e-06, "loss": 0.8183, "step": 1033 }, { "epoch": 0.12, "learning_rate": 2.9291661667666467e-06, "loss": 0.0247, "step": 1034 }, { "epoch": 0.12, "learning_rate": 2.9290941811637675e-06, "loss": 0.0142, "step": 1035 }, { "epoch": 0.12, "learning_rate": 2.929022195560888e-06, "loss": 0.0044, "step": 1036 }, { "epoch": 0.12, "learning_rate": 2.9289502099580086e-06, "loss": 0.0103, "step": 1037 }, { "epoch": 0.12, "learning_rate": 2.928878224355129e-06, "loss": 0.385, "step": 1038 }, { "epoch": 0.12, "learning_rate": 2.9288062387522497e-06, "loss": 0.0094, "step": 1039 }, { "epoch": 0.12, "learning_rate": 2.92873425314937e-06, "loss": 0.0069, "step": 1040 }, { "epoch": 0.12, "learning_rate": 2.928662267546491e-06, "loss": 0.0047, "step": 1041 }, { "epoch": 0.12, "learning_rate": 2.9285902819436112e-06, "loss": 0.0042, "step": 1042 }, { "epoch": 0.12, "learning_rate": 2.928518296340732e-06, "loss": 0.0038, "step": 1043 }, { "epoch": 0.13, "learning_rate": 2.9284463107378524e-06, "loss": 0.0047, "step": 1044 }, { "epoch": 0.13, "learning_rate": 2.928374325134973e-06, "loss": 0.0091, "step": 1045 }, { "epoch": 0.13, "learning_rate": 2.9283023395320935e-06, "loss": 0.0049, "step": 1046 }, { "epoch": 0.13, "learning_rate": 2.9282303539292143e-06, "loss": 0.004, "step": 1047 }, { "epoch": 0.13, "learning_rate": 2.928158368326335e-06, "loss": 0.0077, "step": 1048 }, { "epoch": 0.13, "learning_rate": 2.9280863827234554e-06, "loss": 0.0052, "step": 1049 }, { "epoch": 0.13, "learning_rate": 2.928014397120576e-06, "loss": 0.1371, "step": 1050 }, { "epoch": 0.13, "learning_rate": 2.9279424115176965e-06, "loss": 0.0051, "step": 1051 }, { "epoch": 0.13, "learning_rate": 2.9278704259148173e-06, "loss": 0.6721, "step": 1052 }, { "epoch": 0.13, "learning_rate": 2.9277984403119377e-06, "loss": 0.0057, "step": 1053 }, { "epoch": 0.13, "learning_rate": 2.9277264547090585e-06, "loss": 0.583, "step": 1054 }, { "epoch": 0.13, "learning_rate": 2.927654469106179e-06, "loss": 0.0032, "step": 1055 }, { "epoch": 0.13, "learning_rate": 2.927582483503299e-06, "loss": 0.003, "step": 1056 }, { "epoch": 0.13, "learning_rate": 2.92751049790042e-06, "loss": 0.0037, "step": 1057 }, { "epoch": 0.13, "learning_rate": 2.9274385122975403e-06, "loss": 0.0079, "step": 1058 }, { "epoch": 0.13, "learning_rate": 2.927366526694661e-06, "loss": 0.006, "step": 1059 }, { "epoch": 0.13, "learning_rate": 2.9272945410917814e-06, "loss": 0.0028, "step": 1060 }, { "epoch": 0.13, "learning_rate": 2.927222555488902e-06, "loss": 0.4049, "step": 1061 }, { "epoch": 0.13, "learning_rate": 2.927150569886023e-06, "loss": 0.0191, "step": 1062 }, { "epoch": 0.13, "learning_rate": 2.9270785842831438e-06, "loss": 0.0843, "step": 1063 }, { "epoch": 0.13, "learning_rate": 2.927006598680264e-06, "loss": 0.657, "step": 1064 }, { "epoch": 0.13, "learning_rate": 2.926934613077385e-06, "loss": 0.1867, "step": 1065 }, { "epoch": 0.13, "learning_rate": 2.9268626274745053e-06, "loss": 0.0037, "step": 1066 }, { "epoch": 0.13, "learning_rate": 2.9267906418716256e-06, "loss": 0.2356, "step": 1067 }, { "epoch": 0.13, "learning_rate": 2.9267186562687464e-06, "loss": 0.3433, "step": 1068 }, { "epoch": 0.13, "learning_rate": 2.9266466706658667e-06, "loss": 0.0077, "step": 1069 }, { "epoch": 0.13, "learning_rate": 2.9265746850629875e-06, "loss": 0.9216, "step": 1070 }, { "epoch": 0.13, "learning_rate": 2.926502699460108e-06, "loss": 0.7167, "step": 1071 }, { "epoch": 0.13, "learning_rate": 2.9264307138572287e-06, "loss": 0.0075, "step": 1072 }, { "epoch": 0.13, "learning_rate": 2.926358728254349e-06, "loss": 0.0729, "step": 1073 }, { "epoch": 0.13, "learning_rate": 2.92628674265147e-06, "loss": 0.0869, "step": 1074 }, { "epoch": 0.13, "learning_rate": 2.92621475704859e-06, "loss": 0.4116, "step": 1075 }, { "epoch": 0.13, "learning_rate": 2.926142771445711e-06, "loss": 0.6867, "step": 1076 }, { "epoch": 0.13, "learning_rate": 2.9260707858428317e-06, "loss": 0.0761, "step": 1077 }, { "epoch": 0.13, "learning_rate": 2.925998800239952e-06, "loss": 0.0042, "step": 1078 }, { "epoch": 0.13, "learning_rate": 2.925926814637073e-06, "loss": 0.0057, "step": 1079 }, { "epoch": 0.13, "learning_rate": 2.925854829034193e-06, "loss": 0.003, "step": 1080 }, { "epoch": 0.13, "learning_rate": 2.925782843431314e-06, "loss": 0.01, "step": 1081 }, { "epoch": 0.13, "learning_rate": 2.9257108578284343e-06, "loss": 0.0037, "step": 1082 }, { "epoch": 0.13, "learning_rate": 2.925638872225555e-06, "loss": 0.7341, "step": 1083 }, { "epoch": 0.13, "learning_rate": 2.9255668866226755e-06, "loss": 0.0222, "step": 1084 }, { "epoch": 0.13, "learning_rate": 2.9254949010197962e-06, "loss": 0.0027, "step": 1085 }, { "epoch": 0.13, "learning_rate": 2.9254229154169166e-06, "loss": 1.4908, "step": 1086 }, { "epoch": 0.13, "learning_rate": 2.9253509298140374e-06, "loss": 0.0586, "step": 1087 }, { "epoch": 0.13, "learning_rate": 2.9252789442111577e-06, "loss": 0.0691, "step": 1088 }, { "epoch": 0.13, "learning_rate": 2.9252069586082785e-06, "loss": 0.0192, "step": 1089 }, { "epoch": 0.13, "learning_rate": 2.925134973005399e-06, "loss": 0.0044, "step": 1090 }, { "epoch": 0.13, "learning_rate": 2.9250629874025196e-06, "loss": 0.0065, "step": 1091 }, { "epoch": 0.13, "learning_rate": 2.92499100179964e-06, "loss": 0.0039, "step": 1092 }, { "epoch": 0.13, "learning_rate": 2.9249190161967608e-06, "loss": 0.0253, "step": 1093 }, { "epoch": 0.13, "learning_rate": 2.9248470305938815e-06, "loss": 0.0053, "step": 1094 }, { "epoch": 0.13, "learning_rate": 2.924775044991002e-06, "loss": 0.0691, "step": 1095 }, { "epoch": 0.13, "learning_rate": 2.9247030593881227e-06, "loss": 0.0042, "step": 1096 }, { "epoch": 0.13, "learning_rate": 2.924631073785243e-06, "loss": 0.0301, "step": 1097 }, { "epoch": 0.13, "learning_rate": 2.924559088182364e-06, "loss": 0.5895, "step": 1098 }, { "epoch": 0.13, "learning_rate": 2.924487102579484e-06, "loss": 0.0039, "step": 1099 }, { "epoch": 0.13, "learning_rate": 2.924415116976605e-06, "loss": 0.0033, "step": 1100 }, { "epoch": 0.13, "learning_rate": 2.9243431313737253e-06, "loss": 0.012, "step": 1101 }, { "epoch": 0.13, "learning_rate": 2.9242711457708457e-06, "loss": 0.7306, "step": 1102 }, { "epoch": 0.13, "learning_rate": 2.9241991601679664e-06, "loss": 0.0048, "step": 1103 }, { "epoch": 0.13, "learning_rate": 2.924127174565087e-06, "loss": 0.0035, "step": 1104 }, { "epoch": 0.13, "learning_rate": 2.9240551889622076e-06, "loss": 0.0138, "step": 1105 }, { "epoch": 0.13, "learning_rate": 2.923983203359328e-06, "loss": 0.4516, "step": 1106 }, { "epoch": 0.13, "learning_rate": 2.9239112177564487e-06, "loss": 0.0205, "step": 1107 }, { "epoch": 0.13, "learning_rate": 2.9238392321535695e-06, "loss": 0.014, "step": 1108 }, { "epoch": 0.13, "learning_rate": 2.9237672465506903e-06, "loss": 0.0812, "step": 1109 }, { "epoch": 0.13, "learning_rate": 2.9236952609478106e-06, "loss": 0.0034, "step": 1110 }, { "epoch": 0.13, "learning_rate": 2.9236232753449314e-06, "loss": 0.0048, "step": 1111 }, { "epoch": 0.13, "learning_rate": 2.9235512897420518e-06, "loss": 0.0033, "step": 1112 }, { "epoch": 0.13, "learning_rate": 2.923479304139172e-06, "loss": 0.69, "step": 1113 }, { "epoch": 0.13, "learning_rate": 2.923407318536293e-06, "loss": 0.1559, "step": 1114 }, { "epoch": 0.13, "learning_rate": 2.9233353329334132e-06, "loss": 0.0046, "step": 1115 }, { "epoch": 0.13, "learning_rate": 2.923263347330534e-06, "loss": 0.0031, "step": 1116 }, { "epoch": 0.13, "learning_rate": 2.9231913617276544e-06, "loss": 0.5517, "step": 1117 }, { "epoch": 0.13, "learning_rate": 2.923119376124775e-06, "loss": 0.0045, "step": 1118 }, { "epoch": 0.13, "learning_rate": 2.9230473905218955e-06, "loss": 0.0069, "step": 1119 }, { "epoch": 0.13, "learning_rate": 2.9229754049190163e-06, "loss": 0.003, "step": 1120 }, { "epoch": 0.13, "learning_rate": 2.9229034193161366e-06, "loss": 0.4735, "step": 1121 }, { "epoch": 0.13, "learning_rate": 2.9228314337132574e-06, "loss": 0.0042, "step": 1122 }, { "epoch": 0.13, "learning_rate": 2.9227594481103778e-06, "loss": 0.0108, "step": 1123 }, { "epoch": 0.13, "learning_rate": 2.9226874625074986e-06, "loss": 0.0032, "step": 1124 }, { "epoch": 0.13, "learning_rate": 2.9226154769046193e-06, "loss": 0.0185, "step": 1125 }, { "epoch": 0.13, "learning_rate": 2.9225434913017397e-06, "loss": 0.3465, "step": 1126 }, { "epoch": 0.14, "learning_rate": 2.9224715056988605e-06, "loss": 0.0134, "step": 1127 }, { "epoch": 0.14, "learning_rate": 2.922399520095981e-06, "loss": 0.0046, "step": 1128 }, { "epoch": 0.14, "learning_rate": 2.9223275344931016e-06, "loss": 0.0205, "step": 1129 }, { "epoch": 0.14, "learning_rate": 2.922255548890222e-06, "loss": 0.0044, "step": 1130 }, { "epoch": 0.14, "learning_rate": 2.9221835632873427e-06, "loss": 0.0061, "step": 1131 }, { "epoch": 0.14, "learning_rate": 2.922111577684463e-06, "loss": 0.003, "step": 1132 }, { "epoch": 0.14, "learning_rate": 2.922039592081584e-06, "loss": 0.0134, "step": 1133 }, { "epoch": 0.14, "learning_rate": 2.9219676064787042e-06, "loss": 0.0032, "step": 1134 }, { "epoch": 0.14, "learning_rate": 2.921895620875825e-06, "loss": 0.0031, "step": 1135 }, { "epoch": 0.14, "learning_rate": 2.9218236352729454e-06, "loss": 0.0983, "step": 1136 }, { "epoch": 0.14, "learning_rate": 2.9217516496700657e-06, "loss": 0.0253, "step": 1137 }, { "epoch": 0.14, "learning_rate": 2.9216796640671865e-06, "loss": 0.0033, "step": 1138 }, { "epoch": 0.14, "learning_rate": 2.9216076784643073e-06, "loss": 0.0061, "step": 1139 }, { "epoch": 0.14, "learning_rate": 2.921535692861428e-06, "loss": 0.1121, "step": 1140 }, { "epoch": 0.14, "learning_rate": 2.9214637072585484e-06, "loss": 0.0065, "step": 1141 }, { "epoch": 0.14, "learning_rate": 2.921391721655669e-06, "loss": 1.1041, "step": 1142 }, { "epoch": 0.14, "learning_rate": 2.9213197360527895e-06, "loss": 0.0038, "step": 1143 }, { "epoch": 0.14, "learning_rate": 2.9212477504499103e-06, "loss": 0.652, "step": 1144 }, { "epoch": 0.14, "learning_rate": 2.9211757648470307e-06, "loss": 0.0063, "step": 1145 }, { "epoch": 0.14, "learning_rate": 2.9211037792441514e-06, "loss": 0.0031, "step": 1146 }, { "epoch": 0.14, "learning_rate": 2.921031793641272e-06, "loss": 0.5618, "step": 1147 }, { "epoch": 0.14, "learning_rate": 2.920959808038392e-06, "loss": 0.0042, "step": 1148 }, { "epoch": 0.14, "learning_rate": 2.920887822435513e-06, "loss": 0.2587, "step": 1149 }, { "epoch": 0.14, "learning_rate": 2.9208158368326333e-06, "loss": 0.008, "step": 1150 }, { "epoch": 0.14, "learning_rate": 2.920743851229754e-06, "loss": 0.4423, "step": 1151 }, { "epoch": 0.14, "learning_rate": 2.9206718656268744e-06, "loss": 0.6452, "step": 1152 }, { "epoch": 0.14, "learning_rate": 2.920599880023995e-06, "loss": 0.0034, "step": 1153 }, { "epoch": 0.14, "learning_rate": 2.920527894421116e-06, "loss": 0.1071, "step": 1154 }, { "epoch": 0.14, "learning_rate": 2.9204559088182363e-06, "loss": 0.0065, "step": 1155 }, { "epoch": 0.14, "learning_rate": 2.920383923215357e-06, "loss": 0.0039, "step": 1156 }, { "epoch": 0.14, "learning_rate": 2.920311937612478e-06, "loss": 0.0036, "step": 1157 }, { "epoch": 0.14, "learning_rate": 2.9202399520095982e-06, "loss": 0.003, "step": 1158 }, { "epoch": 0.14, "learning_rate": 2.9201679664067186e-06, "loss": 0.0873, "step": 1159 }, { "epoch": 0.14, "learning_rate": 2.9200959808038394e-06, "loss": 0.0026, "step": 1160 }, { "epoch": 0.14, "learning_rate": 2.9200239952009597e-06, "loss": 0.0035, "step": 1161 }, { "epoch": 0.14, "learning_rate": 2.9199520095980805e-06, "loss": 0.9593, "step": 1162 }, { "epoch": 0.14, "learning_rate": 2.919880023995201e-06, "loss": 0.0035, "step": 1163 }, { "epoch": 0.14, "learning_rate": 2.9198080383923216e-06, "loss": 0.2857, "step": 1164 }, { "epoch": 0.14, "learning_rate": 2.919736052789442e-06, "loss": 0.0072, "step": 1165 }, { "epoch": 0.14, "learning_rate": 2.9196640671865628e-06, "loss": 0.0907, "step": 1166 }, { "epoch": 0.14, "learning_rate": 2.919592081583683e-06, "loss": 0.0032, "step": 1167 }, { "epoch": 0.14, "learning_rate": 2.919520095980804e-06, "loss": 0.0102, "step": 1168 }, { "epoch": 0.14, "learning_rate": 2.9194481103779243e-06, "loss": 0.0167, "step": 1169 }, { "epoch": 0.14, "learning_rate": 2.919376124775045e-06, "loss": 0.0032, "step": 1170 }, { "epoch": 0.14, "learning_rate": 2.919304139172166e-06, "loss": 0.0047, "step": 1171 }, { "epoch": 0.14, "learning_rate": 2.919232153569286e-06, "loss": 0.0061, "step": 1172 }, { "epoch": 0.14, "learning_rate": 2.919160167966407e-06, "loss": 0.0057, "step": 1173 }, { "epoch": 0.14, "learning_rate": 2.9190881823635273e-06, "loss": 0.0201, "step": 1174 }, { "epoch": 0.14, "learning_rate": 2.919016196760648e-06, "loss": 0.0185, "step": 1175 }, { "epoch": 0.14, "learning_rate": 2.9189442111577684e-06, "loss": 0.0023, "step": 1176 }, { "epoch": 0.14, "learning_rate": 2.9188722255548892e-06, "loss": 0.2926, "step": 1177 }, { "epoch": 0.14, "learning_rate": 2.9188002399520096e-06, "loss": 0.1733, "step": 1178 }, { "epoch": 0.14, "learning_rate": 2.9187282543491304e-06, "loss": 0.0324, "step": 1179 }, { "epoch": 0.14, "learning_rate": 2.9186562687462507e-06, "loss": 0.0029, "step": 1180 }, { "epoch": 0.14, "learning_rate": 2.9185842831433715e-06, "loss": 0.0123, "step": 1181 }, { "epoch": 0.14, "learning_rate": 2.918512297540492e-06, "loss": 0.6556, "step": 1182 }, { "epoch": 0.14, "learning_rate": 2.918440311937612e-06, "loss": 0.0023, "step": 1183 }, { "epoch": 0.14, "learning_rate": 2.918368326334733e-06, "loss": 0.9499, "step": 1184 }, { "epoch": 0.14, "learning_rate": 2.9182963407318538e-06, "loss": 0.9959, "step": 1185 }, { "epoch": 0.14, "learning_rate": 2.9182243551289745e-06, "loss": 0.0317, "step": 1186 }, { "epoch": 0.14, "learning_rate": 2.918152369526095e-06, "loss": 0.0036, "step": 1187 }, { "epoch": 0.14, "learning_rate": 2.9180803839232157e-06, "loss": 0.2973, "step": 1188 }, { "epoch": 0.14, "learning_rate": 2.918008398320336e-06, "loss": 0.0325, "step": 1189 }, { "epoch": 0.14, "learning_rate": 2.917936412717457e-06, "loss": 0.6788, "step": 1190 }, { "epoch": 0.14, "learning_rate": 2.917864427114577e-06, "loss": 0.0045, "step": 1191 }, { "epoch": 0.14, "learning_rate": 2.917792441511698e-06, "loss": 0.0329, "step": 1192 }, { "epoch": 0.14, "learning_rate": 2.9177204559088183e-06, "loss": 0.594, "step": 1193 }, { "epoch": 0.14, "learning_rate": 2.9176484703059387e-06, "loss": 0.0117, "step": 1194 }, { "epoch": 0.14, "learning_rate": 2.9175764847030594e-06, "loss": 0.0034, "step": 1195 }, { "epoch": 0.14, "learning_rate": 2.9175044991001798e-06, "loss": 0.6703, "step": 1196 }, { "epoch": 0.14, "learning_rate": 2.9174325134973006e-06, "loss": 0.0077, "step": 1197 }, { "epoch": 0.14, "learning_rate": 2.917360527894421e-06, "loss": 0.2332, "step": 1198 }, { "epoch": 0.14, "learning_rate": 2.9172885422915417e-06, "loss": 0.025, "step": 1199 }, { "epoch": 0.14, "learning_rate": 2.9172165566886625e-06, "loss": 0.3396, "step": 1200 }, { "epoch": 0.14, "learning_rate": 2.917144571085783e-06, "loss": 0.0025, "step": 1201 }, { "epoch": 0.14, "learning_rate": 2.9170725854829036e-06, "loss": 0.0032, "step": 1202 }, { "epoch": 0.14, "learning_rate": 2.9170005998800244e-06, "loss": 0.0031, "step": 1203 }, { "epoch": 0.14, "learning_rate": 2.9169286142771447e-06, "loss": 0.0035, "step": 1204 }, { "epoch": 0.14, "learning_rate": 2.916856628674265e-06, "loss": 0.0038, "step": 1205 }, { "epoch": 0.14, "learning_rate": 2.916784643071386e-06, "loss": 0.0052, "step": 1206 }, { "epoch": 0.14, "learning_rate": 2.9167126574685062e-06, "loss": 0.013, "step": 1207 }, { "epoch": 0.14, "learning_rate": 2.916640671865627e-06, "loss": 1.2136, "step": 1208 }, { "epoch": 0.14, "learning_rate": 2.9165686862627474e-06, "loss": 0.0032, "step": 1209 }, { "epoch": 0.14, "learning_rate": 2.916496700659868e-06, "loss": 0.2562, "step": 1210 }, { "epoch": 0.15, "learning_rate": 2.9164247150569885e-06, "loss": 0.0186, "step": 1211 }, { "epoch": 0.15, "learning_rate": 2.9163527294541093e-06, "loss": 0.5481, "step": 1212 }, { "epoch": 0.15, "learning_rate": 2.9162807438512296e-06, "loss": 0.0034, "step": 1213 }, { "epoch": 0.15, "learning_rate": 2.9162087582483504e-06, "loss": 0.0927, "step": 1214 }, { "epoch": 0.15, "learning_rate": 2.9161367726454708e-06, "loss": 0.5038, "step": 1215 }, { "epoch": 0.15, "learning_rate": 2.9160647870425915e-06, "loss": 0.0024, "step": 1216 }, { "epoch": 0.15, "learning_rate": 2.9159928014397123e-06, "loss": 0.0136, "step": 1217 }, { "epoch": 0.15, "learning_rate": 2.9159208158368327e-06, "loss": 0.1354, "step": 1218 }, { "epoch": 0.15, "learning_rate": 2.9158488302339535e-06, "loss": 0.6409, "step": 1219 }, { "epoch": 0.15, "learning_rate": 2.915776844631074e-06, "loss": 0.0051, "step": 1220 }, { "epoch": 0.15, "learning_rate": 2.9157048590281946e-06, "loss": 0.2611, "step": 1221 }, { "epoch": 0.15, "learning_rate": 2.915632873425315e-06, "loss": 0.0065, "step": 1222 }, { "epoch": 0.15, "learning_rate": 2.9155608878224357e-06, "loss": 0.0174, "step": 1223 }, { "epoch": 0.15, "learning_rate": 2.915488902219556e-06, "loss": 0.0701, "step": 1224 }, { "epoch": 0.15, "learning_rate": 2.915416916616677e-06, "loss": 0.1376, "step": 1225 }, { "epoch": 0.15, "learning_rate": 2.9153449310137972e-06, "loss": 0.0029, "step": 1226 }, { "epoch": 0.15, "learning_rate": 2.915272945410918e-06, "loss": 0.0028, "step": 1227 }, { "epoch": 0.15, "learning_rate": 2.9152009598080383e-06, "loss": 0.6457, "step": 1228 }, { "epoch": 0.15, "learning_rate": 2.9151289742051587e-06, "loss": 0.3767, "step": 1229 }, { "epoch": 0.15, "learning_rate": 2.9150569886022795e-06, "loss": 0.0035, "step": 1230 }, { "epoch": 0.15, "learning_rate": 2.9149850029994003e-06, "loss": 0.0265, "step": 1231 }, { "epoch": 0.15, "learning_rate": 2.914913017396521e-06, "loss": 0.0042, "step": 1232 }, { "epoch": 0.15, "learning_rate": 2.9148410317936414e-06, "loss": 0.0028, "step": 1233 }, { "epoch": 0.15, "learning_rate": 2.914769046190762e-06, "loss": 0.3003, "step": 1234 }, { "epoch": 0.15, "learning_rate": 2.9146970605878825e-06, "loss": 0.2557, "step": 1235 }, { "epoch": 0.15, "learning_rate": 2.9146250749850033e-06, "loss": 0.0089, "step": 1236 }, { "epoch": 0.15, "learning_rate": 2.9145530893821237e-06, "loss": 0.0717, "step": 1237 }, { "epoch": 0.15, "learning_rate": 2.9144811037792444e-06, "loss": 0.0033, "step": 1238 }, { "epoch": 0.15, "learning_rate": 2.914409118176365e-06, "loss": 0.1794, "step": 1239 }, { "epoch": 0.15, "learning_rate": 2.914337132573485e-06, "loss": 0.0031, "step": 1240 }, { "epoch": 0.15, "learning_rate": 2.914265146970606e-06, "loss": 0.0053, "step": 1241 }, { "epoch": 0.15, "learning_rate": 2.9141931613677263e-06, "loss": 0.0052, "step": 1242 }, { "epoch": 0.15, "learning_rate": 2.914121175764847e-06, "loss": 0.1351, "step": 1243 }, { "epoch": 0.15, "learning_rate": 2.9140491901619674e-06, "loss": 0.0179, "step": 1244 }, { "epoch": 0.15, "learning_rate": 2.913977204559088e-06, "loss": 0.2399, "step": 1245 }, { "epoch": 0.15, "learning_rate": 2.913905218956209e-06, "loss": 0.3472, "step": 1246 }, { "epoch": 0.15, "learning_rate": 2.9138332333533293e-06, "loss": 0.0047, "step": 1247 }, { "epoch": 0.15, "learning_rate": 2.91376124775045e-06, "loss": 0.8134, "step": 1248 }, { "epoch": 0.15, "learning_rate": 2.913689262147571e-06, "loss": 0.5008, "step": 1249 }, { "epoch": 0.15, "learning_rate": 2.9136172765446912e-06, "loss": 0.0032, "step": 1250 }, { "epoch": 0.15, "learning_rate": 2.9135452909418116e-06, "loss": 0.003, "step": 1251 }, { "epoch": 0.15, "learning_rate": 2.9134733053389324e-06, "loss": 0.1924, "step": 1252 }, { "epoch": 0.15, "learning_rate": 2.9134013197360527e-06, "loss": 0.002, "step": 1253 }, { "epoch": 0.15, "learning_rate": 2.9133293341331735e-06, "loss": 0.0681, "step": 1254 }, { "epoch": 0.15, "learning_rate": 2.913257348530294e-06, "loss": 0.0055, "step": 1255 }, { "epoch": 0.15, "learning_rate": 2.9131853629274146e-06, "loss": 0.0032, "step": 1256 }, { "epoch": 0.15, "learning_rate": 2.913113377324535e-06, "loss": 0.0051, "step": 1257 }, { "epoch": 0.15, "learning_rate": 2.9130413917216558e-06, "loss": 0.0035, "step": 1258 }, { "epoch": 0.15, "learning_rate": 2.912969406118776e-06, "loss": 0.496, "step": 1259 }, { "epoch": 0.15, "learning_rate": 2.912897420515897e-06, "loss": 0.604, "step": 1260 }, { "epoch": 0.15, "learning_rate": 2.9128254349130173e-06, "loss": 0.0033, "step": 1261 }, { "epoch": 0.15, "learning_rate": 2.912753449310138e-06, "loss": 0.0027, "step": 1262 }, { "epoch": 0.15, "learning_rate": 2.912681463707259e-06, "loss": 0.7969, "step": 1263 }, { "epoch": 0.15, "learning_rate": 2.912609478104379e-06, "loss": 0.7128, "step": 1264 }, { "epoch": 0.15, "learning_rate": 2.9125374925015e-06, "loss": 0.0027, "step": 1265 }, { "epoch": 0.15, "learning_rate": 2.9124655068986203e-06, "loss": 0.0024, "step": 1266 }, { "epoch": 0.15, "learning_rate": 2.912393521295741e-06, "loss": 0.0071, "step": 1267 }, { "epoch": 0.15, "learning_rate": 2.9123215356928614e-06, "loss": 0.003, "step": 1268 }, { "epoch": 0.15, "learning_rate": 2.9122495500899822e-06, "loss": 1.3555, "step": 1269 }, { "epoch": 0.15, "learning_rate": 2.9121775644871026e-06, "loss": 0.0043, "step": 1270 }, { "epoch": 0.15, "learning_rate": 2.9121055788842234e-06, "loss": 0.4319, "step": 1271 }, { "epoch": 0.15, "learning_rate": 2.9120335932813437e-06, "loss": 0.0074, "step": 1272 }, { "epoch": 0.15, "learning_rate": 2.9119616076784645e-06, "loss": 0.008, "step": 1273 }, { "epoch": 0.15, "learning_rate": 2.911889622075585e-06, "loss": 0.0025, "step": 1274 }, { "epoch": 0.15, "learning_rate": 2.911817636472705e-06, "loss": 0.3501, "step": 1275 }, { "epoch": 0.15, "learning_rate": 2.911745650869826e-06, "loss": 0.0043, "step": 1276 }, { "epoch": 0.15, "learning_rate": 2.9116736652669468e-06, "loss": 0.4682, "step": 1277 }, { "epoch": 0.15, "learning_rate": 2.911601679664067e-06, "loss": 0.0028, "step": 1278 }, { "epoch": 0.15, "learning_rate": 2.911529694061188e-06, "loss": 0.2212, "step": 1279 }, { "epoch": 0.15, "learning_rate": 2.9114577084583087e-06, "loss": 0.0048, "step": 1280 }, { "epoch": 0.15, "learning_rate": 2.911385722855429e-06, "loss": 0.0029, "step": 1281 }, { "epoch": 0.15, "learning_rate": 2.91131373725255e-06, "loss": 0.551, "step": 1282 }, { "epoch": 0.15, "learning_rate": 2.91124175164967e-06, "loss": 0.2995, "step": 1283 }, { "epoch": 0.15, "learning_rate": 2.911169766046791e-06, "loss": 0.68, "step": 1284 }, { "epoch": 0.15, "learning_rate": 2.9110977804439113e-06, "loss": 0.0072, "step": 1285 }, { "epoch": 0.15, "learning_rate": 2.9110257948410316e-06, "loss": 0.1872, "step": 1286 }, { "epoch": 0.15, "learning_rate": 2.9109538092381524e-06, "loss": 0.1118, "step": 1287 }, { "epoch": 0.15, "learning_rate": 2.9108818236352728e-06, "loss": 0.0165, "step": 1288 }, { "epoch": 0.15, "learning_rate": 2.9108098380323936e-06, "loss": 0.0896, "step": 1289 }, { "epoch": 0.15, "learning_rate": 2.910737852429514e-06, "loss": 0.0037, "step": 1290 }, { "epoch": 0.15, "learning_rate": 2.9106658668266347e-06, "loss": 0.0031, "step": 1291 }, { "epoch": 0.15, "learning_rate": 2.910593881223755e-06, "loss": 0.0188, "step": 1292 }, { "epoch": 0.15, "learning_rate": 2.910521895620876e-06, "loss": 0.3373, "step": 1293 }, { "epoch": 0.16, "learning_rate": 2.9104499100179966e-06, "loss": 0.8994, "step": 1294 }, { "epoch": 0.16, "learning_rate": 2.9103779244151174e-06, "loss": 0.0986, "step": 1295 }, { "epoch": 0.16, "learning_rate": 2.9103059388122377e-06, "loss": 0.0107, "step": 1296 }, { "epoch": 0.16, "learning_rate": 2.9102339532093585e-06, "loss": 0.0044, "step": 1297 }, { "epoch": 0.16, "learning_rate": 2.910161967606479e-06, "loss": 0.0042, "step": 1298 }, { "epoch": 0.16, "learning_rate": 2.9100899820035992e-06, "loss": 0.5099, "step": 1299 }, { "epoch": 0.16, "learning_rate": 2.91001799640072e-06, "loss": 0.0093, "step": 1300 }, { "epoch": 0.16, "learning_rate": 2.9099460107978404e-06, "loss": 0.0045, "step": 1301 }, { "epoch": 0.16, "learning_rate": 2.909874025194961e-06, "loss": 0.5024, "step": 1302 }, { "epoch": 0.16, "learning_rate": 2.9098020395920815e-06, "loss": 0.0031, "step": 1303 }, { "epoch": 0.16, "learning_rate": 2.9097300539892023e-06, "loss": 0.7096, "step": 1304 }, { "epoch": 0.16, "learning_rate": 2.9096580683863226e-06, "loss": 0.0044, "step": 1305 }, { "epoch": 0.16, "learning_rate": 2.9095860827834434e-06, "loss": 0.0024, "step": 1306 }, { "epoch": 0.16, "learning_rate": 2.9095140971805638e-06, "loss": 0.0032, "step": 1307 }, { "epoch": 0.16, "learning_rate": 2.9094421115776845e-06, "loss": 0.0039, "step": 1308 }, { "epoch": 0.16, "learning_rate": 2.9093701259748053e-06, "loss": 0.0252, "step": 1309 }, { "epoch": 0.16, "learning_rate": 2.9092981403719257e-06, "loss": 1.3322, "step": 1310 }, { "epoch": 0.16, "learning_rate": 2.9092261547690464e-06, "loss": 0.1536, "step": 1311 }, { "epoch": 0.16, "learning_rate": 2.909154169166167e-06, "loss": 0.0169, "step": 1312 }, { "epoch": 0.16, "learning_rate": 2.9090821835632876e-06, "loss": 0.0502, "step": 1313 }, { "epoch": 0.16, "learning_rate": 2.909010197960408e-06, "loss": 0.05, "step": 1314 }, { "epoch": 0.16, "learning_rate": 2.9089382123575287e-06, "loss": 0.0036, "step": 1315 }, { "epoch": 0.16, "learning_rate": 2.908866226754649e-06, "loss": 0.1698, "step": 1316 }, { "epoch": 0.16, "learning_rate": 2.90879424115177e-06, "loss": 0.0071, "step": 1317 }, { "epoch": 0.16, "learning_rate": 2.90872225554889e-06, "loss": 0.2977, "step": 1318 }, { "epoch": 0.16, "learning_rate": 2.908650269946011e-06, "loss": 0.0191, "step": 1319 }, { "epoch": 0.16, "learning_rate": 2.9085782843431313e-06, "loss": 0.5053, "step": 1320 }, { "epoch": 0.16, "learning_rate": 2.9085062987402517e-06, "loss": 0.5653, "step": 1321 }, { "epoch": 0.16, "learning_rate": 2.9084343131373725e-06, "loss": 0.008, "step": 1322 }, { "epoch": 0.16, "learning_rate": 2.9083623275344932e-06, "loss": 0.0034, "step": 1323 }, { "epoch": 0.16, "learning_rate": 2.9082903419316136e-06, "loss": 0.0056, "step": 1324 }, { "epoch": 0.16, "learning_rate": 2.9082183563287344e-06, "loss": 0.003, "step": 1325 }, { "epoch": 0.16, "learning_rate": 2.908146370725855e-06, "loss": 0.0092, "step": 1326 }, { "epoch": 0.16, "learning_rate": 2.9080743851229755e-06, "loss": 0.0034, "step": 1327 }, { "epoch": 0.16, "learning_rate": 2.9080023995200963e-06, "loss": 0.0042, "step": 1328 }, { "epoch": 0.16, "learning_rate": 2.9079304139172167e-06, "loss": 0.0061, "step": 1329 }, { "epoch": 0.16, "learning_rate": 2.9078584283143374e-06, "loss": 0.0051, "step": 1330 }, { "epoch": 0.16, "learning_rate": 2.9077864427114578e-06, "loss": 0.0056, "step": 1331 }, { "epoch": 0.16, "learning_rate": 2.907714457108578e-06, "loss": 0.0037, "step": 1332 }, { "epoch": 0.16, "learning_rate": 2.907642471505699e-06, "loss": 0.2974, "step": 1333 }, { "epoch": 0.16, "learning_rate": 2.9075704859028193e-06, "loss": 0.0051, "step": 1334 }, { "epoch": 0.16, "learning_rate": 2.90749850029994e-06, "loss": 0.0055, "step": 1335 }, { "epoch": 0.16, "learning_rate": 2.9074265146970604e-06, "loss": 0.0046, "step": 1336 }, { "epoch": 0.16, "learning_rate": 2.907354529094181e-06, "loss": 0.003, "step": 1337 }, { "epoch": 0.16, "learning_rate": 2.9072825434913015e-06, "loss": 0.0058, "step": 1338 }, { "epoch": 0.16, "learning_rate": 2.9072105578884223e-06, "loss": 0.0025, "step": 1339 }, { "epoch": 0.16, "learning_rate": 2.907138572285543e-06, "loss": 0.0039, "step": 1340 }, { "epoch": 0.16, "learning_rate": 2.907066586682664e-06, "loss": 0.7075, "step": 1341 }, { "epoch": 0.16, "learning_rate": 2.9069946010797842e-06, "loss": 0.0063, "step": 1342 }, { "epoch": 0.16, "learning_rate": 2.906922615476905e-06, "loss": 0.1718, "step": 1343 }, { "epoch": 0.16, "learning_rate": 2.9068506298740254e-06, "loss": 0.0959, "step": 1344 }, { "epoch": 0.16, "learning_rate": 2.9067786442711457e-06, "loss": 0.2504, "step": 1345 }, { "epoch": 0.16, "learning_rate": 2.9067066586682665e-06, "loss": 0.4884, "step": 1346 }, { "epoch": 0.16, "learning_rate": 2.906634673065387e-06, "loss": 0.0022, "step": 1347 }, { "epoch": 0.16, "learning_rate": 2.9065626874625076e-06, "loss": 0.0032, "step": 1348 }, { "epoch": 0.16, "learning_rate": 2.906490701859628e-06, "loss": 0.003, "step": 1349 }, { "epoch": 0.16, "learning_rate": 2.9064187162567488e-06, "loss": 0.0023, "step": 1350 }, { "epoch": 0.16, "learning_rate": 2.906346730653869e-06, "loss": 0.0029, "step": 1351 }, { "epoch": 0.16, "learning_rate": 2.90627474505099e-06, "loss": 0.0022, "step": 1352 }, { "epoch": 0.16, "learning_rate": 2.9062027594481103e-06, "loss": 0.1094, "step": 1353 }, { "epoch": 0.16, "learning_rate": 2.906130773845231e-06, "loss": 0.7362, "step": 1354 }, { "epoch": 0.16, "learning_rate": 2.906058788242352e-06, "loss": 0.0026, "step": 1355 }, { "epoch": 0.16, "learning_rate": 2.905986802639472e-06, "loss": 0.6562, "step": 1356 }, { "epoch": 0.16, "learning_rate": 2.905914817036593e-06, "loss": 0.5629, "step": 1357 }, { "epoch": 0.16, "learning_rate": 2.9058428314337133e-06, "loss": 0.0023, "step": 1358 }, { "epoch": 0.16, "learning_rate": 2.905770845830834e-06, "loss": 0.0033, "step": 1359 }, { "epoch": 0.16, "learning_rate": 2.9056988602279544e-06, "loss": 0.0029, "step": 1360 }, { "epoch": 0.16, "learning_rate": 2.905626874625075e-06, "loss": 0.0258, "step": 1361 }, { "epoch": 0.16, "learning_rate": 2.9055548890221956e-06, "loss": 0.0906, "step": 1362 }, { "epoch": 0.16, "learning_rate": 2.9054829034193163e-06, "loss": 0.0025, "step": 1363 }, { "epoch": 0.16, "learning_rate": 2.9054109178164367e-06, "loss": 0.0034, "step": 1364 }, { "epoch": 0.16, "learning_rate": 2.9053389322135575e-06, "loss": 0.1114, "step": 1365 }, { "epoch": 0.16, "learning_rate": 2.905266946610678e-06, "loss": 0.3928, "step": 1366 }, { "epoch": 0.16, "learning_rate": 2.905194961007798e-06, "loss": 0.0066, "step": 1367 }, { "epoch": 0.16, "learning_rate": 2.905122975404919e-06, "loss": 0.0037, "step": 1368 }, { "epoch": 0.16, "learning_rate": 2.9050509898020397e-06, "loss": 0.7083, "step": 1369 }, { "epoch": 0.16, "learning_rate": 2.90497900419916e-06, "loss": 0.0034, "step": 1370 }, { "epoch": 0.16, "learning_rate": 2.904907018596281e-06, "loss": 0.0039, "step": 1371 }, { "epoch": 0.16, "learning_rate": 2.9048350329934017e-06, "loss": 0.1206, "step": 1372 }, { "epoch": 0.16, "learning_rate": 2.904763047390522e-06, "loss": 0.0256, "step": 1373 }, { "epoch": 0.16, "learning_rate": 2.904691061787643e-06, "loss": 0.0029, "step": 1374 }, { "epoch": 0.16, "learning_rate": 2.904619076184763e-06, "loss": 0.0321, "step": 1375 }, { "epoch": 0.16, "learning_rate": 2.904547090581884e-06, "loss": 0.1383, "step": 1376 }, { "epoch": 0.17, "learning_rate": 2.9044751049790043e-06, "loss": 0.0031, "step": 1377 }, { "epoch": 0.17, "learning_rate": 2.9044031193761246e-06, "loss": 0.2497, "step": 1378 }, { "epoch": 0.17, "learning_rate": 2.9043311337732454e-06, "loss": 0.15, "step": 1379 }, { "epoch": 0.17, "learning_rate": 2.9042591481703658e-06, "loss": 0.005, "step": 1380 }, { "epoch": 0.17, "learning_rate": 2.9041871625674865e-06, "loss": 0.0083, "step": 1381 }, { "epoch": 0.17, "learning_rate": 2.904115176964607e-06, "loss": 0.0101, "step": 1382 }, { "epoch": 0.17, "learning_rate": 2.9040431913617277e-06, "loss": 0.0039, "step": 1383 }, { "epoch": 0.17, "learning_rate": 2.903971205758848e-06, "loss": 0.7619, "step": 1384 }, { "epoch": 0.17, "learning_rate": 2.903899220155969e-06, "loss": 0.6521, "step": 1385 }, { "epoch": 0.17, "learning_rate": 2.9038272345530896e-06, "loss": 0.003, "step": 1386 }, { "epoch": 0.17, "learning_rate": 2.9037552489502104e-06, "loss": 0.093, "step": 1387 }, { "epoch": 0.17, "learning_rate": 2.9036832633473307e-06, "loss": 0.0042, "step": 1388 }, { "epoch": 0.17, "learning_rate": 2.9036112777444515e-06, "loss": 0.0029, "step": 1389 }, { "epoch": 0.17, "learning_rate": 2.903539292141572e-06, "loss": 0.2547, "step": 1390 }, { "epoch": 0.17, "learning_rate": 2.9034673065386922e-06, "loss": 0.0058, "step": 1391 }, { "epoch": 0.17, "learning_rate": 2.903395320935813e-06, "loss": 0.004, "step": 1392 }, { "epoch": 0.17, "learning_rate": 2.9033233353329333e-06, "loss": 0.0437, "step": 1393 }, { "epoch": 0.17, "learning_rate": 2.903251349730054e-06, "loss": 0.0031, "step": 1394 }, { "epoch": 0.17, "learning_rate": 2.9031793641271745e-06, "loss": 0.0039, "step": 1395 }, { "epoch": 0.17, "learning_rate": 2.9031073785242953e-06, "loss": 0.0021, "step": 1396 }, { "epoch": 0.17, "learning_rate": 2.9030353929214156e-06, "loss": 0.0027, "step": 1397 }, { "epoch": 0.17, "learning_rate": 2.9029634073185364e-06, "loss": 0.0023, "step": 1398 }, { "epoch": 0.17, "learning_rate": 2.9028914217156567e-06, "loss": 0.3655, "step": 1399 }, { "epoch": 0.17, "learning_rate": 2.9028194361127775e-06, "loss": 0.661, "step": 1400 }, { "epoch": 0.17, "learning_rate": 2.902747450509898e-06, "loss": 0.9471, "step": 1401 }, { "epoch": 0.17, "learning_rate": 2.9026754649070187e-06, "loss": 0.0082, "step": 1402 }, { "epoch": 0.17, "learning_rate": 2.9026034793041394e-06, "loss": 0.0033, "step": 1403 }, { "epoch": 0.17, "learning_rate": 2.90253149370126e-06, "loss": 0.1014, "step": 1404 }, { "epoch": 0.17, "learning_rate": 2.9024595080983806e-06, "loss": 0.0029, "step": 1405 }, { "epoch": 0.17, "learning_rate": 2.902387522495501e-06, "loss": 0.0031, "step": 1406 }, { "epoch": 0.17, "learning_rate": 2.9023155368926217e-06, "loss": 0.1213, "step": 1407 }, { "epoch": 0.17, "learning_rate": 2.902243551289742e-06, "loss": 0.0022, "step": 1408 }, { "epoch": 0.17, "learning_rate": 2.902171565686863e-06, "loss": 0.0543, "step": 1409 }, { "epoch": 0.17, "learning_rate": 2.902099580083983e-06, "loss": 0.5447, "step": 1410 }, { "epoch": 0.17, "learning_rate": 2.902027594481104e-06, "loss": 0.248, "step": 1411 }, { "epoch": 0.17, "learning_rate": 2.9019556088782243e-06, "loss": 0.2546, "step": 1412 }, { "epoch": 0.17, "learning_rate": 2.9018836232753447e-06, "loss": 0.004, "step": 1413 }, { "epoch": 0.17, "learning_rate": 2.9018116376724655e-06, "loss": 0.0026, "step": 1414 }, { "epoch": 0.17, "learning_rate": 2.901739652069586e-06, "loss": 0.444, "step": 1415 }, { "epoch": 0.17, "learning_rate": 2.9016676664667066e-06, "loss": 0.0027, "step": 1416 }, { "epoch": 0.17, "learning_rate": 2.9015956808638274e-06, "loss": 0.0025, "step": 1417 }, { "epoch": 0.17, "learning_rate": 2.901523695260948e-06, "loss": 0.004, "step": 1418 }, { "epoch": 0.17, "learning_rate": 2.9014517096580685e-06, "loss": 0.002, "step": 1419 }, { "epoch": 0.17, "learning_rate": 2.9013797240551893e-06, "loss": 0.0098, "step": 1420 }, { "epoch": 0.17, "learning_rate": 2.9013077384523096e-06, "loss": 0.0795, "step": 1421 }, { "epoch": 0.17, "learning_rate": 2.9012357528494304e-06, "loss": 0.1327, "step": 1422 }, { "epoch": 0.17, "learning_rate": 2.9011637672465508e-06, "loss": 0.0032, "step": 1423 }, { "epoch": 0.17, "learning_rate": 2.9010917816436716e-06, "loss": 0.0216, "step": 1424 }, { "epoch": 0.17, "learning_rate": 2.901019796040792e-06, "loss": 0.0537, "step": 1425 }, { "epoch": 0.17, "learning_rate": 2.9009478104379123e-06, "loss": 0.6904, "step": 1426 }, { "epoch": 0.17, "learning_rate": 2.900875824835033e-06, "loss": 0.0032, "step": 1427 }, { "epoch": 0.17, "learning_rate": 2.9008038392321534e-06, "loss": 0.0024, "step": 1428 }, { "epoch": 0.17, "learning_rate": 2.900731853629274e-06, "loss": 0.6644, "step": 1429 }, { "epoch": 0.17, "learning_rate": 2.9006598680263945e-06, "loss": 0.0843, "step": 1430 }, { "epoch": 0.17, "learning_rate": 2.9005878824235153e-06, "loss": 0.004, "step": 1431 }, { "epoch": 0.17, "learning_rate": 2.900515896820636e-06, "loss": 0.003, "step": 1432 }, { "epoch": 0.17, "learning_rate": 2.9004439112177564e-06, "loss": 0.0024, "step": 1433 }, { "epoch": 0.17, "learning_rate": 2.9003719256148772e-06, "loss": 0.0226, "step": 1434 }, { "epoch": 0.17, "learning_rate": 2.900299940011998e-06, "loss": 0.0053, "step": 1435 }, { "epoch": 0.17, "learning_rate": 2.9002279544091184e-06, "loss": 0.0105, "step": 1436 }, { "epoch": 0.17, "learning_rate": 2.9001559688062387e-06, "loss": 0.043, "step": 1437 }, { "epoch": 0.17, "learning_rate": 2.9000839832033595e-06, "loss": 0.0068, "step": 1438 }, { "epoch": 0.17, "learning_rate": 2.90001199760048e-06, "loss": 0.6101, "step": 1439 }, { "epoch": 0.17, "learning_rate": 2.8999400119976006e-06, "loss": 0.0024, "step": 1440 }, { "epoch": 0.17, "learning_rate": 2.899868026394721e-06, "loss": 0.948, "step": 1441 }, { "epoch": 0.17, "learning_rate": 2.8997960407918418e-06, "loss": 0.0029, "step": 1442 }, { "epoch": 0.17, "learning_rate": 2.899724055188962e-06, "loss": 0.0025, "step": 1443 }, { "epoch": 0.17, "learning_rate": 2.899652069586083e-06, "loss": 0.0019, "step": 1444 }, { "epoch": 0.17, "learning_rate": 2.8995800839832032e-06, "loss": 0.6942, "step": 1445 }, { "epoch": 0.17, "learning_rate": 2.899508098380324e-06, "loss": 0.0029, "step": 1446 }, { "epoch": 0.17, "learning_rate": 2.8994361127774444e-06, "loss": 0.131, "step": 1447 }, { "epoch": 0.17, "learning_rate": 2.899364127174565e-06, "loss": 0.2757, "step": 1448 }, { "epoch": 0.17, "learning_rate": 2.899292141571686e-06, "loss": 0.0035, "step": 1449 }, { "epoch": 0.17, "learning_rate": 2.8992201559688063e-06, "loss": 0.0033, "step": 1450 }, { "epoch": 0.17, "learning_rate": 2.899148170365927e-06, "loss": 0.5942, "step": 1451 }, { "epoch": 0.17, "learning_rate": 2.8990761847630474e-06, "loss": 0.2967, "step": 1452 }, { "epoch": 0.17, "learning_rate": 2.899004199160168e-06, "loss": 0.003, "step": 1453 }, { "epoch": 0.17, "learning_rate": 2.8989322135572886e-06, "loss": 0.4187, "step": 1454 }, { "epoch": 0.17, "learning_rate": 2.8988602279544093e-06, "loss": 0.0025, "step": 1455 }, { "epoch": 0.17, "learning_rate": 2.8987882423515297e-06, "loss": 0.0039, "step": 1456 }, { "epoch": 0.17, "learning_rate": 2.8987162567486505e-06, "loss": 0.0421, "step": 1457 }, { "epoch": 0.17, "learning_rate": 2.898644271145771e-06, "loss": 0.0242, "step": 1458 }, { "epoch": 0.17, "learning_rate": 2.898572285542891e-06, "loss": 0.004, "step": 1459 }, { "epoch": 0.17, "learning_rate": 2.898500299940012e-06, "loss": 0.0035, "step": 1460 }, { "epoch": 0.18, "learning_rate": 2.8984283143371323e-06, "loss": 0.0987, "step": 1461 }, { "epoch": 0.18, "learning_rate": 2.898356328734253e-06, "loss": 0.1171, "step": 1462 }, { "epoch": 0.18, "learning_rate": 2.898284343131374e-06, "loss": 0.003, "step": 1463 }, { "epoch": 0.18, "learning_rate": 2.8982123575284946e-06, "loss": 0.0026, "step": 1464 }, { "epoch": 0.18, "learning_rate": 2.898140371925615e-06, "loss": 0.0036, "step": 1465 }, { "epoch": 0.18, "learning_rate": 2.8980683863227358e-06, "loss": 0.8596, "step": 1466 }, { "epoch": 0.18, "learning_rate": 2.897996400719856e-06, "loss": 0.0029, "step": 1467 }, { "epoch": 0.18, "learning_rate": 2.897924415116977e-06, "loss": 0.0525, "step": 1468 }, { "epoch": 0.18, "learning_rate": 2.8978524295140973e-06, "loss": 0.0029, "step": 1469 }, { "epoch": 0.18, "learning_rate": 2.897780443911218e-06, "loss": 0.1271, "step": 1470 }, { "epoch": 0.18, "learning_rate": 2.8977084583083384e-06, "loss": 0.0023, "step": 1471 }, { "epoch": 0.18, "learning_rate": 2.8976364727054588e-06, "loss": 0.0019, "step": 1472 }, { "epoch": 0.18, "learning_rate": 2.8975644871025795e-06, "loss": 0.0043, "step": 1473 }, { "epoch": 0.18, "learning_rate": 2.8974925014997e-06, "loss": 0.3913, "step": 1474 }, { "epoch": 0.18, "learning_rate": 2.8974205158968207e-06, "loss": 0.0026, "step": 1475 }, { "epoch": 0.18, "learning_rate": 2.897348530293941e-06, "loss": 0.7918, "step": 1476 }, { "epoch": 0.18, "learning_rate": 2.897276544691062e-06, "loss": 0.0025, "step": 1477 }, { "epoch": 0.18, "learning_rate": 2.8972045590881826e-06, "loss": 0.0022, "step": 1478 }, { "epoch": 0.18, "learning_rate": 2.897132573485303e-06, "loss": 0.6634, "step": 1479 }, { "epoch": 0.18, "learning_rate": 2.8970605878824237e-06, "loss": 0.0093, "step": 1480 }, { "epoch": 0.18, "learning_rate": 2.8969886022795445e-06, "loss": 0.0022, "step": 1481 }, { "epoch": 0.18, "learning_rate": 2.896916616676665e-06, "loss": 0.0021, "step": 1482 }, { "epoch": 0.18, "learning_rate": 2.896844631073785e-06, "loss": 0.0033, "step": 1483 }, { "epoch": 0.18, "learning_rate": 2.896772645470906e-06, "loss": 0.0283, "step": 1484 }, { "epoch": 0.18, "learning_rate": 2.8967006598680263e-06, "loss": 0.2584, "step": 1485 }, { "epoch": 0.18, "learning_rate": 2.896628674265147e-06, "loss": 0.0029, "step": 1486 }, { "epoch": 0.18, "learning_rate": 2.8965566886622675e-06, "loss": 0.0028, "step": 1487 }, { "epoch": 0.18, "learning_rate": 2.8964847030593883e-06, "loss": 0.0978, "step": 1488 }, { "epoch": 0.18, "learning_rate": 2.8964127174565086e-06, "loss": 0.3327, "step": 1489 }, { "epoch": 0.18, "learning_rate": 2.8963407318536294e-06, "loss": 0.0025, "step": 1490 }, { "epoch": 0.18, "learning_rate": 2.8962687462507497e-06, "loss": 0.0357, "step": 1491 }, { "epoch": 0.18, "learning_rate": 2.8961967606478705e-06, "loss": 0.0037, "step": 1492 }, { "epoch": 0.18, "learning_rate": 2.896124775044991e-06, "loss": 0.4132, "step": 1493 }, { "epoch": 0.18, "learning_rate": 2.8960527894421117e-06, "loss": 0.0026, "step": 1494 }, { "epoch": 0.18, "learning_rate": 2.8959808038392324e-06, "loss": 0.0026, "step": 1495 }, { "epoch": 0.18, "learning_rate": 2.8959088182363528e-06, "loss": 0.1212, "step": 1496 }, { "epoch": 0.18, "learning_rate": 2.8958368326334736e-06, "loss": 0.5421, "step": 1497 }, { "epoch": 0.18, "learning_rate": 2.895764847030594e-06, "loss": 0.0025, "step": 1498 }, { "epoch": 0.18, "learning_rate": 2.8956928614277147e-06, "loss": 0.0027, "step": 1499 }, { "epoch": 0.18, "learning_rate": 2.895620875824835e-06, "loss": 0.4312, "step": 1500 }, { "epoch": 0.18, "learning_rate": 2.895548890221956e-06, "loss": 0.0199, "step": 1501 }, { "epoch": 0.18, "learning_rate": 2.895476904619076e-06, "loss": 0.0038, "step": 1502 }, { "epoch": 0.18, "learning_rate": 2.895404919016197e-06, "loss": 0.0027, "step": 1503 }, { "epoch": 0.18, "learning_rate": 2.8953329334133173e-06, "loss": 0.003, "step": 1504 }, { "epoch": 0.18, "learning_rate": 2.8952609478104377e-06, "loss": 0.0065, "step": 1505 }, { "epoch": 0.18, "learning_rate": 2.8951889622075585e-06, "loss": 0.0104, "step": 1506 }, { "epoch": 0.18, "learning_rate": 2.895116976604679e-06, "loss": 0.0024, "step": 1507 }, { "epoch": 0.18, "learning_rate": 2.8950449910017996e-06, "loss": 0.0026, "step": 1508 }, { "epoch": 0.18, "learning_rate": 2.8949730053989204e-06, "loss": 0.0025, "step": 1509 }, { "epoch": 0.18, "learning_rate": 2.894901019796041e-06, "loss": 0.5683, "step": 1510 }, { "epoch": 0.18, "learning_rate": 2.8948290341931615e-06, "loss": 0.0021, "step": 1511 }, { "epoch": 0.18, "learning_rate": 2.8947570485902823e-06, "loss": 0.6387, "step": 1512 }, { "epoch": 0.18, "learning_rate": 2.8946850629874026e-06, "loss": 0.5011, "step": 1513 }, { "epoch": 0.18, "learning_rate": 2.8946130773845234e-06, "loss": 0.0029, "step": 1514 }, { "epoch": 0.18, "learning_rate": 2.8945410917816438e-06, "loss": 0.0021, "step": 1515 }, { "epoch": 0.18, "learning_rate": 2.8944691061787645e-06, "loss": 0.0018, "step": 1516 }, { "epoch": 0.18, "learning_rate": 2.894397120575885e-06, "loss": 0.0584, "step": 1517 }, { "epoch": 0.18, "learning_rate": 2.8943251349730053e-06, "loss": 0.0042, "step": 1518 }, { "epoch": 0.18, "learning_rate": 2.894253149370126e-06, "loss": 0.0101, "step": 1519 }, { "epoch": 0.18, "learning_rate": 2.8941811637672464e-06, "loss": 0.8253, "step": 1520 }, { "epoch": 0.18, "learning_rate": 2.894109178164367e-06, "loss": 0.002, "step": 1521 }, { "epoch": 0.18, "learning_rate": 2.8940371925614875e-06, "loss": 0.0021, "step": 1522 }, { "epoch": 0.18, "learning_rate": 2.8939652069586083e-06, "loss": 0.7714, "step": 1523 }, { "epoch": 0.18, "learning_rate": 2.8938932213557287e-06, "loss": 0.0137, "step": 1524 }, { "epoch": 0.18, "learning_rate": 2.8938212357528494e-06, "loss": 0.5076, "step": 1525 }, { "epoch": 0.18, "learning_rate": 2.8937492501499702e-06, "loss": 0.0046, "step": 1526 }, { "epoch": 0.18, "learning_rate": 2.893677264547091e-06, "loss": 0.0138, "step": 1527 }, { "epoch": 0.18, "learning_rate": 2.8936052789442113e-06, "loss": 0.002, "step": 1528 }, { "epoch": 0.18, "learning_rate": 2.8935332933413317e-06, "loss": 0.0031, "step": 1529 }, { "epoch": 0.18, "learning_rate": 2.8934613077384525e-06, "loss": 0.3115, "step": 1530 }, { "epoch": 0.18, "learning_rate": 2.893389322135573e-06, "loss": 0.0178, "step": 1531 }, { "epoch": 0.18, "learning_rate": 2.8933173365326936e-06, "loss": 0.0034, "step": 1532 }, { "epoch": 0.18, "learning_rate": 2.893245350929814e-06, "loss": 0.003, "step": 1533 }, { "epoch": 0.18, "learning_rate": 2.8931733653269347e-06, "loss": 0.0258, "step": 1534 }, { "epoch": 0.18, "learning_rate": 2.893101379724055e-06, "loss": 0.0038, "step": 1535 }, { "epoch": 0.18, "learning_rate": 2.893029394121176e-06, "loss": 0.0032, "step": 1536 }, { "epoch": 0.18, "learning_rate": 2.8929574085182962e-06, "loss": 0.0026, "step": 1537 }, { "epoch": 0.18, "learning_rate": 2.892885422915417e-06, "loss": 0.0034, "step": 1538 }, { "epoch": 0.18, "learning_rate": 2.8928134373125374e-06, "loss": 0.0933, "step": 1539 }, { "epoch": 0.18, "learning_rate": 2.892741451709658e-06, "loss": 0.1394, "step": 1540 }, { "epoch": 0.18, "learning_rate": 2.892669466106779e-06, "loss": 0.004, "step": 1541 }, { "epoch": 0.18, "learning_rate": 2.8925974805038993e-06, "loss": 0.0522, "step": 1542 }, { "epoch": 0.18, "learning_rate": 2.89252549490102e-06, "loss": 0.0036, "step": 1543 }, { "epoch": 0.19, "learning_rate": 2.8924535092981404e-06, "loss": 0.0042, "step": 1544 }, { "epoch": 0.19, "learning_rate": 2.892381523695261e-06, "loss": 0.0657, "step": 1545 }, { "epoch": 0.19, "learning_rate": 2.8923095380923816e-06, "loss": 0.0022, "step": 1546 }, { "epoch": 0.19, "learning_rate": 2.8922375524895023e-06, "loss": 0.0033, "step": 1547 }, { "epoch": 0.19, "learning_rate": 2.8921655668866227e-06, "loss": 0.0083, "step": 1548 }, { "epoch": 0.19, "learning_rate": 2.8920935812837435e-06, "loss": 0.0137, "step": 1549 }, { "epoch": 0.19, "learning_rate": 2.892021595680864e-06, "loss": 0.7252, "step": 1550 }, { "epoch": 0.19, "learning_rate": 2.8919496100779846e-06, "loss": 0.0303, "step": 1551 }, { "epoch": 0.19, "learning_rate": 2.891877624475105e-06, "loss": 0.0028, "step": 1552 }, { "epoch": 0.19, "learning_rate": 2.8918056388722253e-06, "loss": 0.0101, "step": 1553 }, { "epoch": 0.19, "learning_rate": 2.891733653269346e-06, "loss": 0.0023, "step": 1554 }, { "epoch": 0.19, "learning_rate": 2.891661667666467e-06, "loss": 0.0045, "step": 1555 }, { "epoch": 0.19, "learning_rate": 2.8915896820635872e-06, "loss": 0.0028, "step": 1556 }, { "epoch": 0.19, "learning_rate": 2.891517696460708e-06, "loss": 0.5717, "step": 1557 }, { "epoch": 0.19, "learning_rate": 2.8914457108578288e-06, "loss": 0.2645, "step": 1558 }, { "epoch": 0.19, "learning_rate": 2.891373725254949e-06, "loss": 0.002, "step": 1559 }, { "epoch": 0.19, "learning_rate": 2.89130173965207e-06, "loss": 0.0255, "step": 1560 }, { "epoch": 0.19, "learning_rate": 2.8912297540491903e-06, "loss": 0.0023, "step": 1561 }, { "epoch": 0.19, "learning_rate": 2.891157768446311e-06, "loss": 0.0021, "step": 1562 }, { "epoch": 0.19, "learning_rate": 2.8910857828434314e-06, "loss": 0.0019, "step": 1563 }, { "epoch": 0.19, "learning_rate": 2.8910137972405518e-06, "loss": 0.0526, "step": 1564 }, { "epoch": 0.19, "learning_rate": 2.8909418116376725e-06, "loss": 0.2289, "step": 1565 }, { "epoch": 0.19, "learning_rate": 2.890869826034793e-06, "loss": 0.002, "step": 1566 }, { "epoch": 0.19, "learning_rate": 2.8907978404319137e-06, "loss": 0.0034, "step": 1567 }, { "epoch": 0.19, "learning_rate": 2.890725854829034e-06, "loss": 0.0031, "step": 1568 }, { "epoch": 0.19, "learning_rate": 2.890653869226155e-06, "loss": 0.0024, "step": 1569 }, { "epoch": 0.19, "learning_rate": 2.890581883623275e-06, "loss": 0.0031, "step": 1570 }, { "epoch": 0.19, "learning_rate": 2.890509898020396e-06, "loss": 0.2258, "step": 1571 }, { "epoch": 0.19, "learning_rate": 2.8904379124175167e-06, "loss": 0.003, "step": 1572 }, { "epoch": 0.19, "learning_rate": 2.8903659268146375e-06, "loss": 0.0023, "step": 1573 }, { "epoch": 0.19, "learning_rate": 2.890293941211758e-06, "loss": 0.0314, "step": 1574 }, { "epoch": 0.19, "learning_rate": 2.890221955608878e-06, "loss": 0.0025, "step": 1575 }, { "epoch": 0.19, "learning_rate": 2.890149970005999e-06, "loss": 0.009, "step": 1576 }, { "epoch": 0.19, "learning_rate": 2.8900779844031193e-06, "loss": 0.0182, "step": 1577 }, { "epoch": 0.19, "learning_rate": 2.89000599880024e-06, "loss": 0.272, "step": 1578 }, { "epoch": 0.19, "learning_rate": 2.8899340131973605e-06, "loss": 0.3822, "step": 1579 }, { "epoch": 0.19, "learning_rate": 2.8898620275944812e-06, "loss": 0.0022, "step": 1580 }, { "epoch": 0.19, "learning_rate": 2.8897900419916016e-06, "loss": 0.0056, "step": 1581 }, { "epoch": 0.19, "learning_rate": 2.8897180563887224e-06, "loss": 0.0024, "step": 1582 }, { "epoch": 0.19, "learning_rate": 2.8896460707858427e-06, "loss": 0.0025, "step": 1583 }, { "epoch": 0.19, "learning_rate": 2.8895740851829635e-06, "loss": 0.0024, "step": 1584 }, { "epoch": 0.19, "learning_rate": 2.889502099580084e-06, "loss": 0.7081, "step": 1585 }, { "epoch": 0.19, "learning_rate": 2.8894301139772046e-06, "loss": 0.0492, "step": 1586 }, { "epoch": 0.19, "learning_rate": 2.8893581283743254e-06, "loss": 0.0021, "step": 1587 }, { "epoch": 0.19, "learning_rate": 2.8892861427714458e-06, "loss": 0.0733, "step": 1588 }, { "epoch": 0.19, "learning_rate": 2.8892141571685666e-06, "loss": 0.0036, "step": 1589 }, { "epoch": 0.19, "learning_rate": 2.889142171565687e-06, "loss": 0.0057, "step": 1590 }, { "epoch": 0.19, "learning_rate": 2.8890701859628077e-06, "loss": 0.0024, "step": 1591 }, { "epoch": 0.19, "learning_rate": 2.888998200359928e-06, "loss": 0.0246, "step": 1592 }, { "epoch": 0.19, "learning_rate": 2.888926214757049e-06, "loss": 0.0019, "step": 1593 }, { "epoch": 0.19, "learning_rate": 2.888854229154169e-06, "loss": 0.7507, "step": 1594 }, { "epoch": 0.19, "learning_rate": 2.88878224355129e-06, "loss": 0.0204, "step": 1595 }, { "epoch": 0.19, "learning_rate": 2.8887102579484103e-06, "loss": 0.0026, "step": 1596 }, { "epoch": 0.19, "learning_rate": 2.888638272345531e-06, "loss": 0.0019, "step": 1597 }, { "epoch": 0.19, "learning_rate": 2.8885662867426514e-06, "loss": 0.0053, "step": 1598 }, { "epoch": 0.19, "learning_rate": 2.888494301139772e-06, "loss": 0.0024, "step": 1599 }, { "epoch": 0.19, "learning_rate": 2.8884223155368926e-06, "loss": 0.1415, "step": 1600 }, { "epoch": 0.19, "learning_rate": 2.8883503299340134e-06, "loss": 0.0033, "step": 1601 }, { "epoch": 0.19, "learning_rate": 2.8882783443311337e-06, "loss": 0.0027, "step": 1602 }, { "epoch": 0.19, "learning_rate": 2.8882063587282545e-06, "loss": 0.0079, "step": 1603 }, { "epoch": 0.19, "learning_rate": 2.8881343731253753e-06, "loss": 0.0023, "step": 1604 }, { "epoch": 0.19, "learning_rate": 2.8880623875224956e-06, "loss": 0.0051, "step": 1605 }, { "epoch": 0.19, "learning_rate": 2.8879904019196164e-06, "loss": 1.357, "step": 1606 }, { "epoch": 0.19, "learning_rate": 2.8879184163167368e-06, "loss": 0.8783, "step": 1607 }, { "epoch": 0.19, "learning_rate": 2.8878464307138575e-06, "loss": 0.0046, "step": 1608 }, { "epoch": 0.19, "learning_rate": 2.887774445110978e-06, "loss": 0.003, "step": 1609 }, { "epoch": 0.19, "learning_rate": 2.8877024595080982e-06, "loss": 0.2235, "step": 1610 }, { "epoch": 0.19, "learning_rate": 2.887630473905219e-06, "loss": 0.0046, "step": 1611 }, { "epoch": 0.19, "learning_rate": 2.8875584883023394e-06, "loss": 0.0021, "step": 1612 }, { "epoch": 0.19, "learning_rate": 2.88748650269946e-06, "loss": 0.649, "step": 1613 }, { "epoch": 0.19, "learning_rate": 2.8874145170965805e-06, "loss": 0.0028, "step": 1614 }, { "epoch": 0.19, "learning_rate": 2.8873425314937013e-06, "loss": 0.8042, "step": 1615 }, { "epoch": 0.19, "learning_rate": 2.8872705458908216e-06, "loss": 0.0026, "step": 1616 }, { "epoch": 0.19, "learning_rate": 2.8871985602879424e-06, "loss": 0.0047, "step": 1617 }, { "epoch": 0.19, "learning_rate": 2.887126574685063e-06, "loss": 0.005, "step": 1618 }, { "epoch": 0.19, "learning_rate": 2.887054589082184e-06, "loss": 0.0022, "step": 1619 }, { "epoch": 0.19, "learning_rate": 2.8869826034793043e-06, "loss": 0.1544, "step": 1620 }, { "epoch": 0.19, "learning_rate": 2.8869106178764247e-06, "loss": 0.3709, "step": 1621 }, { "epoch": 0.19, "learning_rate": 2.8868386322735455e-06, "loss": 0.002, "step": 1622 }, { "epoch": 0.19, "learning_rate": 2.886766646670666e-06, "loss": 0.002, "step": 1623 }, { "epoch": 0.19, "learning_rate": 2.8866946610677866e-06, "loss": 0.0039, "step": 1624 }, { "epoch": 0.19, "learning_rate": 2.886622675464907e-06, "loss": 0.4013, "step": 1625 }, { "epoch": 0.19, "learning_rate": 2.8865506898620277e-06, "loss": 0.0059, "step": 1626 }, { "epoch": 0.19, "learning_rate": 2.886478704259148e-06, "loss": 0.6944, "step": 1627 }, { "epoch": 0.2, "learning_rate": 2.886406718656269e-06, "loss": 0.0018, "step": 1628 }, { "epoch": 0.2, "learning_rate": 2.8863347330533892e-06, "loss": 0.6378, "step": 1629 }, { "epoch": 0.2, "learning_rate": 2.88626274745051e-06, "loss": 0.413, "step": 1630 }, { "epoch": 0.2, "learning_rate": 2.8861907618476304e-06, "loss": 0.0025, "step": 1631 }, { "epoch": 0.2, "learning_rate": 2.886118776244751e-06, "loss": 0.0021, "step": 1632 }, { "epoch": 0.2, "learning_rate": 2.886046790641872e-06, "loss": 0.0236, "step": 1633 }, { "epoch": 0.2, "learning_rate": 2.8859748050389923e-06, "loss": 0.0035, "step": 1634 }, { "epoch": 0.2, "learning_rate": 2.885902819436113e-06, "loss": 0.0023, "step": 1635 }, { "epoch": 0.2, "learning_rate": 2.8858308338332334e-06, "loss": 0.1747, "step": 1636 }, { "epoch": 0.2, "learning_rate": 2.885758848230354e-06, "loss": 0.0022, "step": 1637 }, { "epoch": 0.2, "learning_rate": 2.8856868626274745e-06, "loss": 0.002, "step": 1638 }, { "epoch": 0.2, "learning_rate": 2.8856148770245953e-06, "loss": 0.0027, "step": 1639 }, { "epoch": 0.2, "learning_rate": 2.8855428914217157e-06, "loss": 0.0019, "step": 1640 }, { "epoch": 0.2, "learning_rate": 2.8854709058188365e-06, "loss": 0.0018, "step": 1641 }, { "epoch": 0.2, "learning_rate": 2.885398920215957e-06, "loss": 0.0026, "step": 1642 }, { "epoch": 0.2, "learning_rate": 2.8853269346130776e-06, "loss": 0.0014, "step": 1643 }, { "epoch": 0.2, "learning_rate": 2.885254949010198e-06, "loss": 0.0091, "step": 1644 }, { "epoch": 0.2, "learning_rate": 2.8851829634073183e-06, "loss": 0.002, "step": 1645 }, { "epoch": 0.2, "learning_rate": 2.885110977804439e-06, "loss": 0.8298, "step": 1646 }, { "epoch": 0.2, "learning_rate": 2.88503899220156e-06, "loss": 0.0064, "step": 1647 }, { "epoch": 0.2, "learning_rate": 2.88496700659868e-06, "loss": 0.0092, "step": 1648 }, { "epoch": 0.2, "learning_rate": 2.884895020995801e-06, "loss": 0.0018, "step": 1649 }, { "epoch": 0.2, "learning_rate": 2.8848230353929218e-06, "loss": 0.0029, "step": 1650 }, { "epoch": 0.2, "learning_rate": 2.884751049790042e-06, "loss": 0.0033, "step": 1651 }, { "epoch": 0.2, "learning_rate": 2.884679064187163e-06, "loss": 0.0023, "step": 1652 }, { "epoch": 0.2, "learning_rate": 2.8846070785842833e-06, "loss": 0.006, "step": 1653 }, { "epoch": 0.2, "learning_rate": 2.884535092981404e-06, "loss": 0.0056, "step": 1654 }, { "epoch": 0.2, "learning_rate": 2.8844631073785244e-06, "loss": 0.0066, "step": 1655 }, { "epoch": 0.2, "learning_rate": 2.8843911217756447e-06, "loss": 0.005, "step": 1656 }, { "epoch": 0.2, "learning_rate": 2.8843191361727655e-06, "loss": 0.0018, "step": 1657 }, { "epoch": 0.2, "learning_rate": 2.884247150569886e-06, "loss": 0.0027, "step": 1658 }, { "epoch": 0.2, "learning_rate": 2.8841751649670067e-06, "loss": 0.0023, "step": 1659 }, { "epoch": 0.2, "learning_rate": 2.884103179364127e-06, "loss": 0.0022, "step": 1660 }, { "epoch": 0.2, "learning_rate": 2.884031193761248e-06, "loss": 0.6286, "step": 1661 }, { "epoch": 0.2, "learning_rate": 2.883959208158368e-06, "loss": 0.0162, "step": 1662 }, { "epoch": 0.2, "learning_rate": 2.883887222555489e-06, "loss": 0.0015, "step": 1663 }, { "epoch": 0.2, "learning_rate": 2.8838152369526097e-06, "loss": 0.002, "step": 1664 }, { "epoch": 0.2, "learning_rate": 2.8837432513497305e-06, "loss": 0.0025, "step": 1665 }, { "epoch": 0.2, "learning_rate": 2.883671265746851e-06, "loss": 1.2616, "step": 1666 }, { "epoch": 0.2, "learning_rate": 2.883599280143971e-06, "loss": 0.1128, "step": 1667 }, { "epoch": 0.2, "learning_rate": 2.883527294541092e-06, "loss": 0.0027, "step": 1668 }, { "epoch": 0.2, "learning_rate": 2.8834553089382123e-06, "loss": 0.0031, "step": 1669 }, { "epoch": 0.2, "learning_rate": 2.883383323335333e-06, "loss": 0.8305, "step": 1670 }, { "epoch": 0.2, "learning_rate": 2.8833113377324535e-06, "loss": 0.7148, "step": 1671 }, { "epoch": 0.2, "learning_rate": 2.8832393521295742e-06, "loss": 0.0041, "step": 1672 }, { "epoch": 0.2, "learning_rate": 2.8831673665266946e-06, "loss": 0.0189, "step": 1673 }, { "epoch": 0.2, "learning_rate": 2.8830953809238154e-06, "loss": 0.0021, "step": 1674 }, { "epoch": 0.2, "learning_rate": 2.8830233953209357e-06, "loss": 0.0026, "step": 1675 }, { "epoch": 0.2, "learning_rate": 2.8829514097180565e-06, "loss": 0.0015, "step": 1676 }, { "epoch": 0.2, "learning_rate": 2.882879424115177e-06, "loss": 0.0329, "step": 1677 }, { "epoch": 0.2, "learning_rate": 2.8828074385122976e-06, "loss": 0.0083, "step": 1678 }, { "epoch": 0.2, "learning_rate": 2.882735452909418e-06, "loss": 0.2682, "step": 1679 }, { "epoch": 0.2, "learning_rate": 2.8826634673065388e-06, "loss": 0.0353, "step": 1680 }, { "epoch": 0.2, "learning_rate": 2.8825914817036595e-06, "loss": 0.6175, "step": 1681 }, { "epoch": 0.2, "learning_rate": 2.88251949610078e-06, "loss": 0.0024, "step": 1682 }, { "epoch": 0.2, "learning_rate": 2.8824475104979007e-06, "loss": 0.426, "step": 1683 }, { "epoch": 0.2, "learning_rate": 2.882375524895021e-06, "loss": 0.0201, "step": 1684 }, { "epoch": 0.2, "learning_rate": 2.882303539292142e-06, "loss": 0.0066, "step": 1685 }, { "epoch": 0.2, "learning_rate": 2.882231553689262e-06, "loss": 1.5158, "step": 1686 }, { "epoch": 0.2, "learning_rate": 2.882159568086383e-06, "loss": 0.0016, "step": 1687 }, { "epoch": 0.2, "learning_rate": 2.8820875824835033e-06, "loss": 0.0018, "step": 1688 }, { "epoch": 0.2, "learning_rate": 2.882015596880624e-06, "loss": 0.0019, "step": 1689 }, { "epoch": 0.2, "learning_rate": 2.8819436112777444e-06, "loss": 0.7636, "step": 1690 }, { "epoch": 0.2, "learning_rate": 2.881871625674865e-06, "loss": 0.0546, "step": 1691 }, { "epoch": 0.2, "learning_rate": 2.8817996400719856e-06, "loss": 0.0354, "step": 1692 }, { "epoch": 0.2, "learning_rate": 2.881727654469106e-06, "loss": 0.002, "step": 1693 }, { "epoch": 0.2, "learning_rate": 2.8816556688662267e-06, "loss": 0.0022, "step": 1694 }, { "epoch": 0.2, "learning_rate": 2.8815836832633475e-06, "loss": 0.0021, "step": 1695 }, { "epoch": 0.2, "learning_rate": 2.8815116976604683e-06, "loss": 0.0017, "step": 1696 }, { "epoch": 0.2, "learning_rate": 2.8814397120575886e-06, "loss": 0.0308, "step": 1697 }, { "epoch": 0.2, "learning_rate": 2.8813677264547094e-06, "loss": 0.0024, "step": 1698 }, { "epoch": 0.2, "learning_rate": 2.8812957408518298e-06, "loss": 0.7655, "step": 1699 }, { "epoch": 0.2, "learning_rate": 2.8812237552489505e-06, "loss": 0.0388, "step": 1700 }, { "epoch": 0.2, "learning_rate": 2.881151769646071e-06, "loss": 0.0022, "step": 1701 }, { "epoch": 0.2, "learning_rate": 2.8810797840431912e-06, "loss": 0.0068, "step": 1702 }, { "epoch": 0.2, "learning_rate": 2.881007798440312e-06, "loss": 0.0019, "step": 1703 }, { "epoch": 0.2, "learning_rate": 2.8809358128374324e-06, "loss": 0.0018, "step": 1704 }, { "epoch": 0.2, "learning_rate": 2.880863827234553e-06, "loss": 0.0051, "step": 1705 }, { "epoch": 0.2, "learning_rate": 2.8807918416316735e-06, "loss": 0.0016, "step": 1706 }, { "epoch": 0.2, "learning_rate": 2.8807198560287943e-06, "loss": 0.0025, "step": 1707 }, { "epoch": 0.2, "learning_rate": 2.8806478704259146e-06, "loss": 0.0118, "step": 1708 }, { "epoch": 0.2, "learning_rate": 2.8805758848230354e-06, "loss": 0.0021, "step": 1709 }, { "epoch": 0.2, "learning_rate": 2.880503899220156e-06, "loss": 0.2014, "step": 1710 }, { "epoch": 0.21, "learning_rate": 2.8804319136172766e-06, "loss": 0.0015, "step": 1711 }, { "epoch": 0.21, "learning_rate": 2.8803599280143973e-06, "loss": 0.6916, "step": 1712 }, { "epoch": 0.21, "learning_rate": 2.8802879424115177e-06, "loss": 0.1024, "step": 1713 }, { "epoch": 0.21, "learning_rate": 2.8802159568086385e-06, "loss": 0.0036, "step": 1714 }, { "epoch": 0.21, "learning_rate": 2.880143971205759e-06, "loss": 0.6724, "step": 1715 }, { "epoch": 0.21, "learning_rate": 2.8800719856028796e-06, "loss": 0.005, "step": 1716 }, { "epoch": 0.21, "learning_rate": 2.88e-06, "loss": 0.0031, "step": 1717 }, { "epoch": 0.21, "learning_rate": 2.8799280143971207e-06, "loss": 0.0027, "step": 1718 }, { "epoch": 0.21, "learning_rate": 2.879856028794241e-06, "loss": 0.0019, "step": 1719 }, { "epoch": 0.21, "learning_rate": 2.879784043191362e-06, "loss": 0.0025, "step": 1720 }, { "epoch": 0.21, "learning_rate": 2.8797120575884822e-06, "loss": 0.0019, "step": 1721 }, { "epoch": 0.21, "learning_rate": 2.879640071985603e-06, "loss": 0.0019, "step": 1722 }, { "epoch": 0.21, "learning_rate": 2.8795680863827234e-06, "loss": 0.0048, "step": 1723 }, { "epoch": 0.21, "learning_rate": 2.879496100779844e-06, "loss": 0.0021, "step": 1724 }, { "epoch": 0.21, "learning_rate": 2.8794241151769645e-06, "loss": 0.0029, "step": 1725 }, { "epoch": 0.21, "learning_rate": 2.8793521295740853e-06, "loss": 0.0027, "step": 1726 }, { "epoch": 0.21, "learning_rate": 2.879280143971206e-06, "loss": 0.0022, "step": 1727 }, { "epoch": 0.21, "learning_rate": 2.8792081583683264e-06, "loss": 0.0013, "step": 1728 }, { "epoch": 0.21, "learning_rate": 2.879136172765447e-06, "loss": 0.0415, "step": 1729 }, { "epoch": 0.21, "learning_rate": 2.8790641871625675e-06, "loss": 0.0481, "step": 1730 }, { "epoch": 0.21, "learning_rate": 2.8789922015596883e-06, "loss": 0.8323, "step": 1731 }, { "epoch": 0.21, "learning_rate": 2.8789202159568087e-06, "loss": 0.7727, "step": 1732 }, { "epoch": 0.21, "learning_rate": 2.8788482303539294e-06, "loss": 0.0021, "step": 1733 }, { "epoch": 0.21, "learning_rate": 2.87877624475105e-06, "loss": 0.0052, "step": 1734 }, { "epoch": 0.21, "learning_rate": 2.8787042591481706e-06, "loss": 0.0017, "step": 1735 }, { "epoch": 0.21, "learning_rate": 2.878632273545291e-06, "loss": 0.0017, "step": 1736 }, { "epoch": 0.21, "learning_rate": 2.8785602879424113e-06, "loss": 0.0019, "step": 1737 }, { "epoch": 0.21, "learning_rate": 2.878488302339532e-06, "loss": 0.0206, "step": 1738 }, { "epoch": 0.21, "learning_rate": 2.8784163167366524e-06, "loss": 0.0021, "step": 1739 }, { "epoch": 0.21, "learning_rate": 2.878344331133773e-06, "loss": 0.0031, "step": 1740 }, { "epoch": 0.21, "learning_rate": 2.878272345530894e-06, "loss": 0.0017, "step": 1741 }, { "epoch": 0.21, "learning_rate": 2.8782003599280148e-06, "loss": 0.0021, "step": 1742 }, { "epoch": 0.21, "learning_rate": 2.878128374325135e-06, "loss": 0.0027, "step": 1743 }, { "epoch": 0.21, "learning_rate": 2.878056388722256e-06, "loss": 0.668, "step": 1744 }, { "epoch": 0.21, "learning_rate": 2.8779844031193762e-06, "loss": 0.002, "step": 1745 }, { "epoch": 0.21, "learning_rate": 2.877912417516497e-06, "loss": 0.0039, "step": 1746 }, { "epoch": 0.21, "learning_rate": 2.8778404319136174e-06, "loss": 0.0016, "step": 1747 }, { "epoch": 0.21, "learning_rate": 2.8777684463107377e-06, "loss": 0.0021, "step": 1748 }, { "epoch": 0.21, "learning_rate": 2.8776964607078585e-06, "loss": 0.0019, "step": 1749 }, { "epoch": 0.21, "learning_rate": 2.877624475104979e-06, "loss": 0.002, "step": 1750 }, { "epoch": 0.21, "learning_rate": 2.8775524895020996e-06, "loss": 0.0678, "step": 1751 }, { "epoch": 0.21, "learning_rate": 2.87748050389922e-06, "loss": 0.0018, "step": 1752 }, { "epoch": 0.21, "learning_rate": 2.8774085182963408e-06, "loss": 0.0017, "step": 1753 }, { "epoch": 0.21, "learning_rate": 2.877336532693461e-06, "loss": 0.0054, "step": 1754 }, { "epoch": 0.21, "learning_rate": 2.877264547090582e-06, "loss": 0.0022, "step": 1755 }, { "epoch": 0.21, "learning_rate": 2.8771925614877027e-06, "loss": 0.0027, "step": 1756 }, { "epoch": 0.21, "learning_rate": 2.877120575884823e-06, "loss": 0.003, "step": 1757 }, { "epoch": 0.21, "learning_rate": 2.877048590281944e-06, "loss": 0.2437, "step": 1758 }, { "epoch": 0.21, "learning_rate": 2.8769766046790646e-06, "loss": 0.0019, "step": 1759 }, { "epoch": 0.21, "learning_rate": 2.876904619076185e-06, "loss": 0.0016, "step": 1760 }, { "epoch": 0.21, "learning_rate": 2.8768326334733053e-06, "loss": 0.0346, "step": 1761 }, { "epoch": 0.21, "learning_rate": 2.876760647870426e-06, "loss": 0.0306, "step": 1762 }, { "epoch": 0.21, "learning_rate": 2.8766886622675465e-06, "loss": 0.0028, "step": 1763 }, { "epoch": 0.21, "learning_rate": 2.8766166766646672e-06, "loss": 0.6781, "step": 1764 }, { "epoch": 0.21, "learning_rate": 2.8765446910617876e-06, "loss": 0.0017, "step": 1765 }, { "epoch": 0.21, "learning_rate": 2.8764727054589084e-06, "loss": 0.0027, "step": 1766 }, { "epoch": 0.21, "learning_rate": 2.8764007198560287e-06, "loss": 0.3089, "step": 1767 }, { "epoch": 0.21, "learning_rate": 2.8763287342531495e-06, "loss": 0.0015, "step": 1768 }, { "epoch": 0.21, "learning_rate": 2.87625674865027e-06, "loss": 0.0015, "step": 1769 }, { "epoch": 0.21, "learning_rate": 2.8761847630473906e-06, "loss": 0.0085, "step": 1770 }, { "epoch": 0.21, "learning_rate": 2.876112777444511e-06, "loss": 0.004, "step": 1771 }, { "epoch": 0.21, "learning_rate": 2.8760407918416318e-06, "loss": 0.0158, "step": 1772 }, { "epoch": 0.21, "learning_rate": 2.8759688062387525e-06, "loss": 0.0043, "step": 1773 }, { "epoch": 0.21, "learning_rate": 2.875896820635873e-06, "loss": 0.0029, "step": 1774 }, { "epoch": 0.21, "learning_rate": 2.8758248350329937e-06, "loss": 0.0033, "step": 1775 }, { "epoch": 0.21, "learning_rate": 2.875752849430114e-06, "loss": 0.0095, "step": 1776 }, { "epoch": 0.21, "learning_rate": 2.875680863827235e-06, "loss": 0.7142, "step": 1777 }, { "epoch": 0.21, "learning_rate": 2.875608878224355e-06, "loss": 0.0019, "step": 1778 }, { "epoch": 0.21, "learning_rate": 2.875536892621476e-06, "loss": 0.0313, "step": 1779 }, { "epoch": 0.21, "learning_rate": 2.8754649070185963e-06, "loss": 0.002, "step": 1780 }, { "epoch": 0.21, "learning_rate": 2.875392921415717e-06, "loss": 0.6876, "step": 1781 }, { "epoch": 0.21, "learning_rate": 2.8753209358128374e-06, "loss": 0.002, "step": 1782 }, { "epoch": 0.21, "learning_rate": 2.8752489502099578e-06, "loss": 0.0027, "step": 1783 }, { "epoch": 0.21, "learning_rate": 2.8751769646070786e-06, "loss": 0.6431, "step": 1784 }, { "epoch": 0.21, "learning_rate": 2.875104979004199e-06, "loss": 0.0021, "step": 1785 }, { "epoch": 0.21, "learning_rate": 2.8750329934013197e-06, "loss": 0.0021, "step": 1786 }, { "epoch": 0.21, "learning_rate": 2.8749610077984405e-06, "loss": 0.0021, "step": 1787 }, { "epoch": 0.21, "learning_rate": 2.8748890221955613e-06, "loss": 0.0927, "step": 1788 }, { "epoch": 0.21, "learning_rate": 2.8748170365926816e-06, "loss": 0.2481, "step": 1789 }, { "epoch": 0.21, "learning_rate": 2.8747450509898024e-06, "loss": 0.0061, "step": 1790 }, { "epoch": 0.21, "learning_rate": 2.8746730653869227e-06, "loss": 0.0015, "step": 1791 }, { "epoch": 0.21, "learning_rate": 2.8746010797840435e-06, "loss": 0.0016, "step": 1792 }, { "epoch": 0.21, "learning_rate": 2.874529094181164e-06, "loss": 0.237, "step": 1793 }, { "epoch": 0.21, "learning_rate": 2.8744571085782842e-06, "loss": 0.0019, "step": 1794 }, { "epoch": 0.22, "learning_rate": 2.874385122975405e-06, "loss": 0.0026, "step": 1795 }, { "epoch": 0.22, "learning_rate": 2.8743131373725254e-06, "loss": 0.002, "step": 1796 }, { "epoch": 0.22, "learning_rate": 2.874241151769646e-06, "loss": 0.7055, "step": 1797 }, { "epoch": 0.22, "learning_rate": 2.8741691661667665e-06, "loss": 0.0023, "step": 1798 }, { "epoch": 0.22, "learning_rate": 2.8740971805638873e-06, "loss": 0.0017, "step": 1799 }, { "epoch": 0.22, "learning_rate": 2.8740251949610076e-06, "loss": 0.0017, "step": 1800 }, { "epoch": 0.22, "learning_rate": 2.8739532093581284e-06, "loss": 0.7599, "step": 1801 }, { "epoch": 0.22, "learning_rate": 2.8738812237552488e-06, "loss": 0.0035, "step": 1802 }, { "epoch": 0.22, "learning_rate": 2.8738092381523695e-06, "loss": 0.0017, "step": 1803 }, { "epoch": 0.22, "learning_rate": 2.8737372525494903e-06, "loss": 0.002, "step": 1804 }, { "epoch": 0.22, "learning_rate": 2.873665266946611e-06, "loss": 0.0019, "step": 1805 }, { "epoch": 0.22, "learning_rate": 2.8735932813437315e-06, "loss": 0.1821, "step": 1806 }, { "epoch": 0.22, "learning_rate": 2.873521295740852e-06, "loss": 0.7564, "step": 1807 }, { "epoch": 0.22, "learning_rate": 2.8734493101379726e-06, "loss": 0.002, "step": 1808 }, { "epoch": 0.22, "learning_rate": 2.873377324535093e-06, "loss": 0.0105, "step": 1809 }, { "epoch": 0.22, "learning_rate": 2.8733053389322137e-06, "loss": 0.002, "step": 1810 }, { "epoch": 0.22, "learning_rate": 2.873233353329334e-06, "loss": 0.0015, "step": 1811 }, { "epoch": 0.22, "learning_rate": 2.873161367726455e-06, "loss": 0.0011, "step": 1812 }, { "epoch": 0.22, "learning_rate": 2.8730893821235752e-06, "loss": 0.806, "step": 1813 }, { "epoch": 0.22, "learning_rate": 2.873017396520696e-06, "loss": 0.0021, "step": 1814 }, { "epoch": 0.22, "learning_rate": 2.8729454109178163e-06, "loss": 0.0045, "step": 1815 }, { "epoch": 0.22, "learning_rate": 2.872873425314937e-06, "loss": 0.0016, "step": 1816 }, { "epoch": 0.22, "learning_rate": 2.8728014397120575e-06, "loss": 0.028, "step": 1817 }, { "epoch": 0.22, "learning_rate": 2.8727294541091783e-06, "loss": 0.0021, "step": 1818 }, { "epoch": 0.22, "learning_rate": 2.872657468506299e-06, "loss": 0.0023, "step": 1819 }, { "epoch": 0.22, "learning_rate": 2.8725854829034194e-06, "loss": 0.6126, "step": 1820 }, { "epoch": 0.22, "learning_rate": 2.87251349730054e-06, "loss": 0.0018, "step": 1821 }, { "epoch": 0.22, "learning_rate": 2.8724415116976605e-06, "loss": 0.0015, "step": 1822 }, { "epoch": 0.22, "learning_rate": 2.8723695260947813e-06, "loss": 0.0021, "step": 1823 }, { "epoch": 0.22, "learning_rate": 2.8722975404919017e-06, "loss": 0.0019, "step": 1824 }, { "epoch": 0.22, "learning_rate": 2.8722255548890224e-06, "loss": 0.0476, "step": 1825 }, { "epoch": 0.22, "learning_rate": 2.872153569286143e-06, "loss": 0.0062, "step": 1826 }, { "epoch": 0.22, "learning_rate": 2.8720815836832636e-06, "loss": 0.0025, "step": 1827 }, { "epoch": 0.22, "learning_rate": 2.872009598080384e-06, "loss": 0.0023, "step": 1828 }, { "epoch": 0.22, "learning_rate": 2.8719376124775043e-06, "loss": 0.0016, "step": 1829 }, { "epoch": 0.22, "learning_rate": 2.871865626874625e-06, "loss": 0.0287, "step": 1830 }, { "epoch": 0.22, "learning_rate": 2.8717936412717454e-06, "loss": 0.0019, "step": 1831 }, { "epoch": 0.22, "learning_rate": 2.871721655668866e-06, "loss": 0.0018, "step": 1832 }, { "epoch": 0.22, "learning_rate": 2.871649670065987e-06, "loss": 0.5052, "step": 1833 }, { "epoch": 0.22, "learning_rate": 2.8715776844631073e-06, "loss": 0.8367, "step": 1834 }, { "epoch": 0.22, "learning_rate": 2.871505698860228e-06, "loss": 0.0019, "step": 1835 }, { "epoch": 0.22, "learning_rate": 2.871433713257349e-06, "loss": 0.0014, "step": 1836 }, { "epoch": 0.22, "learning_rate": 2.8713617276544692e-06, "loss": 0.0047, "step": 1837 }, { "epoch": 0.22, "learning_rate": 2.87128974205159e-06, "loss": 0.0208, "step": 1838 }, { "epoch": 0.22, "learning_rate": 2.8712177564487104e-06, "loss": 0.0019, "step": 1839 }, { "epoch": 0.22, "learning_rate": 2.8711457708458307e-06, "loss": 0.5638, "step": 1840 }, { "epoch": 0.22, "learning_rate": 2.8710737852429515e-06, "loss": 0.0025, "step": 1841 }, { "epoch": 0.22, "learning_rate": 2.871001799640072e-06, "loss": 0.0028, "step": 1842 }, { "epoch": 0.22, "learning_rate": 2.8709298140371926e-06, "loss": 0.0039, "step": 1843 }, { "epoch": 0.22, "learning_rate": 2.870857828434313e-06, "loss": 0.0014, "step": 1844 }, { "epoch": 0.22, "learning_rate": 2.8707858428314338e-06, "loss": 0.2972, "step": 1845 }, { "epoch": 0.22, "learning_rate": 2.870713857228554e-06, "loss": 0.5828, "step": 1846 }, { "epoch": 0.22, "learning_rate": 2.870641871625675e-06, "loss": 0.003, "step": 1847 }, { "epoch": 0.22, "learning_rate": 2.8705698860227953e-06, "loss": 0.0105, "step": 1848 }, { "epoch": 0.22, "learning_rate": 2.870497900419916e-06, "loss": 0.0463, "step": 1849 }, { "epoch": 0.22, "learning_rate": 2.870425914817037e-06, "loss": 0.0018, "step": 1850 }, { "epoch": 0.22, "learning_rate": 2.8703539292141576e-06, "loss": 0.0021, "step": 1851 }, { "epoch": 0.22, "learning_rate": 2.870281943611278e-06, "loss": 0.0057, "step": 1852 }, { "epoch": 0.22, "learning_rate": 2.8702099580083983e-06, "loss": 0.0017, "step": 1853 }, { "epoch": 0.22, "learning_rate": 2.870137972405519e-06, "loss": 0.0015, "step": 1854 }, { "epoch": 0.22, "learning_rate": 2.8700659868026394e-06, "loss": 0.0017, "step": 1855 }, { "epoch": 0.22, "learning_rate": 2.8699940011997602e-06, "loss": 0.795, "step": 1856 }, { "epoch": 0.22, "learning_rate": 2.8699220155968806e-06, "loss": 0.0266, "step": 1857 }, { "epoch": 0.22, "learning_rate": 2.8698500299940014e-06, "loss": 0.002, "step": 1858 }, { "epoch": 0.22, "learning_rate": 2.8697780443911217e-06, "loss": 0.1586, "step": 1859 }, { "epoch": 0.22, "learning_rate": 2.8697060587882425e-06, "loss": 0.5111, "step": 1860 }, { "epoch": 0.22, "learning_rate": 2.869634073185363e-06, "loss": 0.0344, "step": 1861 }, { "epoch": 0.22, "learning_rate": 2.8695620875824836e-06, "loss": 0.0018, "step": 1862 }, { "epoch": 0.22, "learning_rate": 2.869490101979604e-06, "loss": 0.002, "step": 1863 }, { "epoch": 0.22, "learning_rate": 2.8694181163767248e-06, "loss": 0.0016, "step": 1864 }, { "epoch": 0.22, "learning_rate": 2.8693461307738455e-06, "loss": 0.0018, "step": 1865 }, { "epoch": 0.22, "learning_rate": 2.869274145170966e-06, "loss": 0.0015, "step": 1866 }, { "epoch": 0.22, "learning_rate": 2.8692021595680867e-06, "loss": 0.3607, "step": 1867 }, { "epoch": 0.22, "learning_rate": 2.869130173965207e-06, "loss": 0.1547, "step": 1868 }, { "epoch": 0.22, "learning_rate": 2.869058188362328e-06, "loss": 0.1064, "step": 1869 }, { "epoch": 0.22, "learning_rate": 2.868986202759448e-06, "loss": 0.0017, "step": 1870 }, { "epoch": 0.22, "learning_rate": 2.868914217156569e-06, "loss": 0.0017, "step": 1871 }, { "epoch": 0.22, "learning_rate": 2.8688422315536893e-06, "loss": 0.0185, "step": 1872 }, { "epoch": 0.22, "learning_rate": 2.86877024595081e-06, "loss": 0.0024, "step": 1873 }, { "epoch": 0.22, "learning_rate": 2.8686982603479304e-06, "loss": 0.0019, "step": 1874 }, { "epoch": 0.22, "learning_rate": 2.8686262747450508e-06, "loss": 0.0018, "step": 1875 }, { "epoch": 0.22, "learning_rate": 2.8685542891421716e-06, "loss": 0.0024, "step": 1876 }, { "epoch": 0.22, "learning_rate": 2.868482303539292e-06, "loss": 0.1989, "step": 1877 }, { "epoch": 0.23, "learning_rate": 2.8684103179364127e-06, "loss": 0.003, "step": 1878 }, { "epoch": 0.23, "learning_rate": 2.8683383323335335e-06, "loss": 0.2338, "step": 1879 }, { "epoch": 0.23, "learning_rate": 2.868266346730654e-06, "loss": 0.0025, "step": 1880 }, { "epoch": 0.23, "learning_rate": 2.8681943611277746e-06, "loss": 0.8032, "step": 1881 }, { "epoch": 0.23, "learning_rate": 2.8681223755248954e-06, "loss": 0.0017, "step": 1882 }, { "epoch": 0.23, "learning_rate": 2.8680503899220157e-06, "loss": 0.0021, "step": 1883 }, { "epoch": 0.23, "learning_rate": 2.8679784043191365e-06, "loss": 0.0156, "step": 1884 }, { "epoch": 0.23, "learning_rate": 2.867906418716257e-06, "loss": 0.0018, "step": 1885 }, { "epoch": 0.23, "learning_rate": 2.8678344331133776e-06, "loss": 0.0018, "step": 1886 }, { "epoch": 0.23, "learning_rate": 2.867762447510498e-06, "loss": 0.7203, "step": 1887 }, { "epoch": 0.23, "learning_rate": 2.8676904619076184e-06, "loss": 0.3159, "step": 1888 }, { "epoch": 0.23, "learning_rate": 2.867618476304739e-06, "loss": 0.0045, "step": 1889 }, { "epoch": 0.23, "learning_rate": 2.8675464907018595e-06, "loss": 0.0016, "step": 1890 }, { "epoch": 0.23, "learning_rate": 2.8674745050989803e-06, "loss": 0.0037, "step": 1891 }, { "epoch": 0.23, "learning_rate": 2.8674025194961006e-06, "loss": 0.6293, "step": 1892 }, { "epoch": 0.23, "learning_rate": 2.8673305338932214e-06, "loss": 0.0015, "step": 1893 }, { "epoch": 0.23, "learning_rate": 2.8672585482903418e-06, "loss": 0.002, "step": 1894 }, { "epoch": 0.23, "learning_rate": 2.8671865626874625e-06, "loss": 0.0024, "step": 1895 }, { "epoch": 0.23, "learning_rate": 2.8671145770845833e-06, "loss": 0.6097, "step": 1896 }, { "epoch": 0.23, "learning_rate": 2.867042591481704e-06, "loss": 0.0056, "step": 1897 }, { "epoch": 0.23, "learning_rate": 2.8669706058788244e-06, "loss": 0.002, "step": 1898 }, { "epoch": 0.23, "learning_rate": 2.866898620275945e-06, "loss": 0.0013, "step": 1899 }, { "epoch": 0.23, "learning_rate": 2.8668266346730656e-06, "loss": 0.3791, "step": 1900 }, { "epoch": 0.23, "learning_rate": 2.866754649070186e-06, "loss": 0.0015, "step": 1901 }, { "epoch": 0.23, "learning_rate": 2.8666826634673067e-06, "loss": 0.0031, "step": 1902 }, { "epoch": 0.23, "learning_rate": 2.866610677864427e-06, "loss": 0.0461, "step": 1903 }, { "epoch": 0.23, "learning_rate": 2.866538692261548e-06, "loss": 0.0017, "step": 1904 }, { "epoch": 0.23, "learning_rate": 2.866466706658668e-06, "loss": 0.0023, "step": 1905 }, { "epoch": 0.23, "learning_rate": 2.866394721055789e-06, "loss": 0.0386, "step": 1906 }, { "epoch": 0.23, "learning_rate": 2.8663227354529093e-06, "loss": 0.0023, "step": 1907 }, { "epoch": 0.23, "learning_rate": 2.86625074985003e-06, "loss": 0.0014, "step": 1908 }, { "epoch": 0.23, "learning_rate": 2.8661787642471505e-06, "loss": 0.0019, "step": 1909 }, { "epoch": 0.23, "learning_rate": 2.8661067786442713e-06, "loss": 0.0027, "step": 1910 }, { "epoch": 0.23, "learning_rate": 2.866034793041392e-06, "loss": 0.0175, "step": 1911 }, { "epoch": 0.23, "learning_rate": 2.8659628074385124e-06, "loss": 0.0017, "step": 1912 }, { "epoch": 0.23, "learning_rate": 2.865890821835633e-06, "loss": 0.0018, "step": 1913 }, { "epoch": 0.23, "learning_rate": 2.8658188362327535e-06, "loss": 0.0026, "step": 1914 }, { "epoch": 0.23, "learning_rate": 2.8657468506298743e-06, "loss": 0.5942, "step": 1915 }, { "epoch": 0.23, "learning_rate": 2.8656748650269947e-06, "loss": 0.0045, "step": 1916 }, { "epoch": 0.23, "learning_rate": 2.8656028794241154e-06, "loss": 0.0056, "step": 1917 }, { "epoch": 0.23, "learning_rate": 2.8655308938212358e-06, "loss": 0.0024, "step": 1918 }, { "epoch": 0.23, "learning_rate": 2.8654589082183566e-06, "loss": 0.3457, "step": 1919 }, { "epoch": 0.23, "learning_rate": 2.865386922615477e-06, "loss": 0.4032, "step": 1920 }, { "epoch": 0.23, "learning_rate": 2.8653149370125973e-06, "loss": 0.0017, "step": 1921 }, { "epoch": 0.23, "learning_rate": 2.865242951409718e-06, "loss": 0.3887, "step": 1922 }, { "epoch": 0.23, "learning_rate": 2.8651709658068384e-06, "loss": 0.0039, "step": 1923 }, { "epoch": 0.23, "learning_rate": 2.865098980203959e-06, "loss": 0.0019, "step": 1924 }, { "epoch": 0.23, "learning_rate": 2.8650269946010795e-06, "loss": 0.0016, "step": 1925 }, { "epoch": 0.23, "learning_rate": 2.8649550089982003e-06, "loss": 0.0022, "step": 1926 }, { "epoch": 0.23, "learning_rate": 2.864883023395321e-06, "loss": 0.0018, "step": 1927 }, { "epoch": 0.23, "learning_rate": 2.864811037792442e-06, "loss": 0.0014, "step": 1928 }, { "epoch": 0.23, "learning_rate": 2.8647390521895622e-06, "loss": 0.0015, "step": 1929 }, { "epoch": 0.23, "learning_rate": 2.864667066586683e-06, "loss": 0.3628, "step": 1930 }, { "epoch": 0.23, "learning_rate": 2.8645950809838034e-06, "loss": 0.0021, "step": 1931 }, { "epoch": 0.23, "learning_rate": 2.864523095380924e-06, "loss": 0.0013, "step": 1932 }, { "epoch": 0.23, "learning_rate": 2.8644511097780445e-06, "loss": 0.0026, "step": 1933 }, { "epoch": 0.23, "learning_rate": 2.864379124175165e-06, "loss": 0.0019, "step": 1934 }, { "epoch": 0.23, "learning_rate": 2.8643071385722856e-06, "loss": 0.8081, "step": 1935 }, { "epoch": 0.23, "learning_rate": 2.864235152969406e-06, "loss": 0.0103, "step": 1936 }, { "epoch": 0.23, "learning_rate": 2.8641631673665268e-06, "loss": 0.0022, "step": 1937 }, { "epoch": 0.23, "learning_rate": 2.864091181763647e-06, "loss": 0.4023, "step": 1938 }, { "epoch": 0.23, "learning_rate": 2.864019196160768e-06, "loss": 0.0017, "step": 1939 }, { "epoch": 0.23, "learning_rate": 2.8639472105578883e-06, "loss": 0.002, "step": 1940 }, { "epoch": 0.23, "learning_rate": 2.863875224955009e-06, "loss": 0.0021, "step": 1941 }, { "epoch": 0.23, "learning_rate": 2.86380323935213e-06, "loss": 0.0018, "step": 1942 }, { "epoch": 0.23, "learning_rate": 2.86373125374925e-06, "loss": 0.0022, "step": 1943 }, { "epoch": 0.23, "learning_rate": 2.863659268146371e-06, "loss": 0.0019, "step": 1944 }, { "epoch": 0.23, "learning_rate": 2.8635872825434913e-06, "loss": 0.068, "step": 1945 }, { "epoch": 0.23, "learning_rate": 2.863515296940612e-06, "loss": 0.0018, "step": 1946 }, { "epoch": 0.23, "learning_rate": 2.8634433113377324e-06, "loss": 0.0016, "step": 1947 }, { "epoch": 0.23, "learning_rate": 2.8633713257348532e-06, "loss": 0.0037, "step": 1948 }, { "epoch": 0.23, "learning_rate": 2.8632993401319736e-06, "loss": 0.002, "step": 1949 }, { "epoch": 0.23, "learning_rate": 2.8632273545290943e-06, "loss": 0.319, "step": 1950 }, { "epoch": 0.23, "learning_rate": 2.8631553689262147e-06, "loss": 0.4548, "step": 1951 }, { "epoch": 0.23, "learning_rate": 2.8630833833233355e-06, "loss": 0.011, "step": 1952 }, { "epoch": 0.23, "learning_rate": 2.863011397720456e-06, "loss": 0.0029, "step": 1953 }, { "epoch": 0.23, "learning_rate": 2.8629394121175766e-06, "loss": 0.0021, "step": 1954 }, { "epoch": 0.23, "learning_rate": 2.862867426514697e-06, "loss": 0.9554, "step": 1955 }, { "epoch": 0.23, "learning_rate": 2.8627954409118177e-06, "loss": 0.0016, "step": 1956 }, { "epoch": 0.23, "learning_rate": 2.862723455308938e-06, "loss": 0.5756, "step": 1957 }, { "epoch": 0.23, "learning_rate": 2.862651469706059e-06, "loss": 0.0044, "step": 1958 }, { "epoch": 0.23, "learning_rate": 2.8625794841031797e-06, "loss": 0.3608, "step": 1959 }, { "epoch": 0.23, "learning_rate": 2.8625074985003e-06, "loss": 0.0034, "step": 1960 }, { "epoch": 0.23, "learning_rate": 2.862435512897421e-06, "loss": 0.0017, "step": 1961 }, { "epoch": 0.24, "learning_rate": 2.862363527294541e-06, "loss": 0.0018, "step": 1962 }, { "epoch": 0.24, "learning_rate": 2.862291541691662e-06, "loss": 0.0176, "step": 1963 }, { "epoch": 0.24, "learning_rate": 2.8622195560887823e-06, "loss": 0.3458, "step": 1964 }, { "epoch": 0.24, "learning_rate": 2.862147570485903e-06, "loss": 0.0019, "step": 1965 }, { "epoch": 0.24, "learning_rate": 2.8620755848830234e-06, "loss": 0.0015, "step": 1966 }, { "epoch": 0.24, "learning_rate": 2.8620035992801438e-06, "loss": 0.4692, "step": 1967 }, { "epoch": 0.24, "learning_rate": 2.8619316136772645e-06, "loss": 0.003, "step": 1968 }, { "epoch": 0.24, "learning_rate": 2.861859628074385e-06, "loss": 0.9009, "step": 1969 }, { "epoch": 0.24, "learning_rate": 2.8617876424715057e-06, "loss": 0.0016, "step": 1970 }, { "epoch": 0.24, "learning_rate": 2.861715656868626e-06, "loss": 0.0035, "step": 1971 }, { "epoch": 0.24, "learning_rate": 2.861643671265747e-06, "loss": 0.0062, "step": 1972 }, { "epoch": 0.24, "learning_rate": 2.8615716856628676e-06, "loss": 0.085, "step": 1973 }, { "epoch": 0.24, "learning_rate": 2.8614997000599884e-06, "loss": 0.0021, "step": 1974 }, { "epoch": 0.24, "learning_rate": 2.8614277144571087e-06, "loss": 0.0062, "step": 1975 }, { "epoch": 0.24, "learning_rate": 2.8613557288542295e-06, "loss": 0.2957, "step": 1976 }, { "epoch": 0.24, "learning_rate": 2.86128374325135e-06, "loss": 0.0018, "step": 1977 }, { "epoch": 0.24, "learning_rate": 2.8612117576484706e-06, "loss": 0.0852, "step": 1978 }, { "epoch": 0.24, "learning_rate": 2.861139772045591e-06, "loss": 0.0019, "step": 1979 }, { "epoch": 0.24, "learning_rate": 2.8610677864427114e-06, "loss": 0.0751, "step": 1980 }, { "epoch": 0.24, "learning_rate": 2.860995800839832e-06, "loss": 0.0022, "step": 1981 }, { "epoch": 0.24, "learning_rate": 2.8609238152369525e-06, "loss": 0.0243, "step": 1982 }, { "epoch": 0.24, "learning_rate": 2.8608518296340733e-06, "loss": 0.0066, "step": 1983 }, { "epoch": 0.24, "learning_rate": 2.8607798440311936e-06, "loss": 0.0018, "step": 1984 }, { "epoch": 0.24, "learning_rate": 2.8607078584283144e-06, "loss": 0.0016, "step": 1985 }, { "epoch": 0.24, "learning_rate": 2.8606358728254348e-06, "loss": 0.1154, "step": 1986 }, { "epoch": 0.24, "learning_rate": 2.8605638872225555e-06, "loss": 0.7959, "step": 1987 }, { "epoch": 0.24, "learning_rate": 2.8604919016196763e-06, "loss": 0.0014, "step": 1988 }, { "epoch": 0.24, "learning_rate": 2.8604199160167967e-06, "loss": 0.2917, "step": 1989 }, { "epoch": 0.24, "learning_rate": 2.8603479304139174e-06, "loss": 0.0023, "step": 1990 }, { "epoch": 0.24, "learning_rate": 2.860275944811038e-06, "loss": 0.003, "step": 1991 }, { "epoch": 0.24, "learning_rate": 2.8602039592081586e-06, "loss": 0.0019, "step": 1992 }, { "epoch": 0.24, "learning_rate": 2.860131973605279e-06, "loss": 0.0015, "step": 1993 }, { "epoch": 0.24, "learning_rate": 2.8600599880023997e-06, "loss": 0.002, "step": 1994 }, { "epoch": 0.24, "learning_rate": 2.85998800239952e-06, "loss": 0.002, "step": 1995 }, { "epoch": 0.24, "learning_rate": 2.859916016796641e-06, "loss": 0.0018, "step": 1996 }, { "epoch": 0.24, "learning_rate": 2.859844031193761e-06, "loss": 0.0043, "step": 1997 }, { "epoch": 0.24, "learning_rate": 2.859772045590882e-06, "loss": 0.0017, "step": 1998 }, { "epoch": 0.24, "learning_rate": 2.8597000599880023e-06, "loss": 0.002, "step": 1999 }, { "epoch": 0.24, "learning_rate": 2.859628074385123e-06, "loss": 0.8341, "step": 2000 }, { "epoch": 0.24, "learning_rate": 2.8595560887822435e-06, "loss": 0.0017, "step": 2001 }, { "epoch": 0.24, "learning_rate": 2.8594841031793642e-06, "loss": 0.0533, "step": 2002 }, { "epoch": 0.24, "learning_rate": 2.8594121175764846e-06, "loss": 0.0017, "step": 2003 }, { "epoch": 0.24, "learning_rate": 2.8593401319736054e-06, "loss": 0.0668, "step": 2004 }, { "epoch": 0.24, "learning_rate": 2.859268146370726e-06, "loss": 0.0017, "step": 2005 }, { "epoch": 0.24, "learning_rate": 2.8591961607678465e-06, "loss": 0.002, "step": 2006 }, { "epoch": 0.24, "learning_rate": 2.8591241751649673e-06, "loss": 0.015, "step": 2007 }, { "epoch": 0.24, "learning_rate": 2.8590521895620876e-06, "loss": 0.0148, "step": 2008 }, { "epoch": 0.24, "learning_rate": 2.8589802039592084e-06, "loss": 0.0013, "step": 2009 }, { "epoch": 0.24, "learning_rate": 2.8589082183563288e-06, "loss": 0.0024, "step": 2010 }, { "epoch": 0.24, "learning_rate": 2.8588362327534496e-06, "loss": 0.0621, "step": 2011 }, { "epoch": 0.24, "learning_rate": 2.85876424715057e-06, "loss": 0.0047, "step": 2012 }, { "epoch": 0.24, "learning_rate": 2.8586922615476907e-06, "loss": 0.0083, "step": 2013 }, { "epoch": 0.24, "learning_rate": 2.858620275944811e-06, "loss": 0.0017, "step": 2014 }, { "epoch": 0.24, "learning_rate": 2.8585482903419314e-06, "loss": 0.0018, "step": 2015 }, { "epoch": 0.24, "learning_rate": 2.858476304739052e-06, "loss": 0.4767, "step": 2016 }, { "epoch": 0.24, "learning_rate": 2.8584043191361725e-06, "loss": 0.0016, "step": 2017 }, { "epoch": 0.24, "learning_rate": 2.8583323335332933e-06, "loss": 0.1315, "step": 2018 }, { "epoch": 0.24, "learning_rate": 2.858260347930414e-06, "loss": 0.0023, "step": 2019 }, { "epoch": 0.24, "learning_rate": 2.858188362327535e-06, "loss": 0.0018, "step": 2020 }, { "epoch": 0.24, "learning_rate": 2.8581163767246552e-06, "loss": 0.0184, "step": 2021 }, { "epoch": 0.24, "learning_rate": 2.858044391121776e-06, "loss": 0.0018, "step": 2022 }, { "epoch": 0.24, "learning_rate": 2.8579724055188964e-06, "loss": 0.7091, "step": 2023 }, { "epoch": 0.24, "learning_rate": 2.857900419916017e-06, "loss": 0.1766, "step": 2024 }, { "epoch": 0.24, "learning_rate": 2.8578284343131375e-06, "loss": 0.8616, "step": 2025 }, { "epoch": 0.24, "learning_rate": 2.857756448710258e-06, "loss": 0.0017, "step": 2026 }, { "epoch": 0.24, "learning_rate": 2.8576844631073786e-06, "loss": 0.4794, "step": 2027 }, { "epoch": 0.24, "learning_rate": 2.857612477504499e-06, "loss": 0.002, "step": 2028 }, { "epoch": 0.24, "learning_rate": 2.8575404919016198e-06, "loss": 0.0026, "step": 2029 }, { "epoch": 0.24, "learning_rate": 2.85746850629874e-06, "loss": 0.0072, "step": 2030 }, { "epoch": 0.24, "learning_rate": 2.857396520695861e-06, "loss": 0.002, "step": 2031 }, { "epoch": 0.24, "learning_rate": 2.8573245350929812e-06, "loss": 0.0088, "step": 2032 }, { "epoch": 0.24, "learning_rate": 2.857252549490102e-06, "loss": 0.0017, "step": 2033 }, { "epoch": 0.24, "learning_rate": 2.857180563887223e-06, "loss": 0.0021, "step": 2034 }, { "epoch": 0.24, "learning_rate": 2.857108578284343e-06, "loss": 0.002, "step": 2035 }, { "epoch": 0.24, "learning_rate": 2.857036592681464e-06, "loss": 0.0021, "step": 2036 }, { "epoch": 0.24, "learning_rate": 2.8569646070785843e-06, "loss": 0.0293, "step": 2037 }, { "epoch": 0.24, "learning_rate": 2.856892621475705e-06, "loss": 0.0031, "step": 2038 }, { "epoch": 0.24, "learning_rate": 2.8568206358728254e-06, "loss": 0.0025, "step": 2039 }, { "epoch": 0.24, "learning_rate": 2.856748650269946e-06, "loss": 0.0017, "step": 2040 }, { "epoch": 0.24, "learning_rate": 2.8566766646670666e-06, "loss": 0.2858, "step": 2041 }, { "epoch": 0.24, "learning_rate": 2.8566046790641873e-06, "loss": 0.0014, "step": 2042 }, { "epoch": 0.24, "learning_rate": 2.8565326934613077e-06, "loss": 0.0015, "step": 2043 }, { "epoch": 0.24, "learning_rate": 2.8564607078584285e-06, "loss": 0.7568, "step": 2044 }, { "epoch": 0.25, "learning_rate": 2.856388722255549e-06, "loss": 0.0276, "step": 2045 }, { "epoch": 0.25, "learning_rate": 2.8563167366526696e-06, "loss": 0.0027, "step": 2046 }, { "epoch": 0.25, "learning_rate": 2.85624475104979e-06, "loss": 0.9217, "step": 2047 }, { "epoch": 0.25, "learning_rate": 2.8561727654469107e-06, "loss": 0.8071, "step": 2048 }, { "epoch": 0.25, "learning_rate": 2.856100779844031e-06, "loss": 0.0033, "step": 2049 }, { "epoch": 0.25, "learning_rate": 2.856028794241152e-06, "loss": 0.0015, "step": 2050 }, { "epoch": 0.25, "learning_rate": 2.8559568086382727e-06, "loss": 0.043, "step": 2051 }, { "epoch": 0.25, "learning_rate": 2.855884823035393e-06, "loss": 0.7099, "step": 2052 }, { "epoch": 0.25, "learning_rate": 2.8558128374325138e-06, "loss": 0.0024, "step": 2053 }, { "epoch": 0.25, "learning_rate": 2.855740851829634e-06, "loss": 0.0018, "step": 2054 }, { "epoch": 0.25, "learning_rate": 2.855668866226755e-06, "loss": 0.1719, "step": 2055 }, { "epoch": 0.25, "learning_rate": 2.8555968806238753e-06, "loss": 0.0024, "step": 2056 }, { "epoch": 0.25, "learning_rate": 2.855524895020996e-06, "loss": 0.0024, "step": 2057 }, { "epoch": 0.25, "learning_rate": 2.8554529094181164e-06, "loss": 0.0019, "step": 2058 }, { "epoch": 0.25, "learning_rate": 2.855380923815237e-06, "loss": 0.2309, "step": 2059 }, { "epoch": 0.25, "learning_rate": 2.8553089382123575e-06, "loss": 0.0015, "step": 2060 }, { "epoch": 0.25, "learning_rate": 2.855236952609478e-06, "loss": 0.0021, "step": 2061 }, { "epoch": 0.25, "learning_rate": 2.8551649670065987e-06, "loss": 0.0019, "step": 2062 }, { "epoch": 0.25, "learning_rate": 2.855092981403719e-06, "loss": 0.4297, "step": 2063 }, { "epoch": 0.25, "learning_rate": 2.85502099580084e-06, "loss": 0.0017, "step": 2064 }, { "epoch": 0.25, "learning_rate": 2.8549490101979606e-06, "loss": 0.4398, "step": 2065 }, { "epoch": 0.25, "learning_rate": 2.8548770245950814e-06, "loss": 0.0017, "step": 2066 }, { "epoch": 0.25, "learning_rate": 2.8548050389922017e-06, "loss": 0.0016, "step": 2067 }, { "epoch": 0.25, "learning_rate": 2.8547330533893225e-06, "loss": 0.002, "step": 2068 }, { "epoch": 0.25, "learning_rate": 2.854661067786443e-06, "loss": 0.0016, "step": 2069 }, { "epoch": 0.25, "learning_rate": 2.8545890821835636e-06, "loss": 0.002, "step": 2070 }, { "epoch": 0.25, "learning_rate": 2.854517096580684e-06, "loss": 0.8694, "step": 2071 }, { "epoch": 0.25, "learning_rate": 2.8544451109778043e-06, "loss": 0.0027, "step": 2072 }, { "epoch": 0.25, "learning_rate": 2.854373125374925e-06, "loss": 0.0034, "step": 2073 }, { "epoch": 0.25, "learning_rate": 2.8543011397720455e-06, "loss": 0.0433, "step": 2074 }, { "epoch": 0.25, "learning_rate": 2.8542291541691663e-06, "loss": 0.002, "step": 2075 }, { "epoch": 0.25, "learning_rate": 2.8541571685662866e-06, "loss": 0.712, "step": 2076 }, { "epoch": 0.25, "learning_rate": 2.8540851829634074e-06, "loss": 0.0027, "step": 2077 }, { "epoch": 0.25, "learning_rate": 2.8540131973605277e-06, "loss": 0.0191, "step": 2078 }, { "epoch": 0.25, "learning_rate": 2.8539412117576485e-06, "loss": 0.0052, "step": 2079 }, { "epoch": 0.25, "learning_rate": 2.853869226154769e-06, "loss": 0.002, "step": 2080 }, { "epoch": 0.25, "learning_rate": 2.8537972405518897e-06, "loss": 0.005, "step": 2081 }, { "epoch": 0.25, "learning_rate": 2.8537252549490104e-06, "loss": 0.0439, "step": 2082 }, { "epoch": 0.25, "learning_rate": 2.853653269346131e-06, "loss": 0.1645, "step": 2083 }, { "epoch": 0.25, "learning_rate": 2.8535812837432516e-06, "loss": 0.002, "step": 2084 }, { "epoch": 0.25, "learning_rate": 2.853509298140372e-06, "loss": 0.0015, "step": 2085 }, { "epoch": 0.25, "learning_rate": 2.8534373125374927e-06, "loss": 0.6766, "step": 2086 }, { "epoch": 0.25, "learning_rate": 2.853365326934613e-06, "loss": 0.0016, "step": 2087 }, { "epoch": 0.25, "learning_rate": 2.853293341331734e-06, "loss": 0.234, "step": 2088 }, { "epoch": 0.25, "learning_rate": 2.853221355728854e-06, "loss": 0.002, "step": 2089 }, { "epoch": 0.25, "learning_rate": 2.853149370125975e-06, "loss": 0.0015, "step": 2090 }, { "epoch": 0.25, "learning_rate": 2.8530773845230953e-06, "loss": 0.0018, "step": 2091 }, { "epoch": 0.25, "learning_rate": 2.853005398920216e-06, "loss": 0.719, "step": 2092 }, { "epoch": 0.25, "learning_rate": 2.8529334133173365e-06, "loss": 0.0015, "step": 2093 }, { "epoch": 0.25, "learning_rate": 2.852861427714457e-06, "loss": 0.0172, "step": 2094 }, { "epoch": 0.25, "learning_rate": 2.8527894421115776e-06, "loss": 0.6322, "step": 2095 }, { "epoch": 0.25, "learning_rate": 2.8527174565086984e-06, "loss": 0.0162, "step": 2096 }, { "epoch": 0.25, "learning_rate": 2.852645470905819e-06, "loss": 0.0012, "step": 2097 }, { "epoch": 0.25, "learning_rate": 2.8525734853029395e-06, "loss": 0.0015, "step": 2098 }, { "epoch": 0.25, "learning_rate": 2.8525014997000603e-06, "loss": 0.0014, "step": 2099 }, { "epoch": 0.25, "learning_rate": 2.8524295140971806e-06, "loss": 0.0014, "step": 2100 }, { "epoch": 0.25, "learning_rate": 2.8523575284943014e-06, "loss": 0.7831, "step": 2101 }, { "epoch": 0.25, "learning_rate": 2.8522855428914218e-06, "loss": 1.3163, "step": 2102 }, { "epoch": 0.25, "learning_rate": 2.8522135572885425e-06, "loss": 0.0016, "step": 2103 }, { "epoch": 0.25, "learning_rate": 2.852141571685663e-06, "loss": 0.0981, "step": 2104 }, { "epoch": 0.25, "learning_rate": 2.8520695860827837e-06, "loss": 0.0044, "step": 2105 }, { "epoch": 0.25, "learning_rate": 2.851997600479904e-06, "loss": 0.0016, "step": 2106 }, { "epoch": 0.25, "learning_rate": 2.8519256148770244e-06, "loss": 0.0045, "step": 2107 }, { "epoch": 0.25, "learning_rate": 2.851853629274145e-06, "loss": 0.0017, "step": 2108 }, { "epoch": 0.25, "learning_rate": 2.8517816436712655e-06, "loss": 0.0018, "step": 2109 }, { "epoch": 0.25, "learning_rate": 2.8517096580683863e-06, "loss": 0.0016, "step": 2110 }, { "epoch": 0.25, "learning_rate": 2.851637672465507e-06, "loss": 0.0014, "step": 2111 }, { "epoch": 0.25, "learning_rate": 2.8515656868626274e-06, "loss": 0.0015, "step": 2112 }, { "epoch": 0.25, "learning_rate": 2.8514937012597482e-06, "loss": 0.4708, "step": 2113 }, { "epoch": 0.25, "learning_rate": 2.851421715656869e-06, "loss": 0.0021, "step": 2114 }, { "epoch": 0.25, "learning_rate": 2.8513497300539893e-06, "loss": 0.0027, "step": 2115 }, { "epoch": 0.25, "learning_rate": 2.85127774445111e-06, "loss": 0.0251, "step": 2116 }, { "epoch": 0.25, "learning_rate": 2.8512057588482305e-06, "loss": 0.0068, "step": 2117 }, { "epoch": 0.25, "learning_rate": 2.851133773245351e-06, "loss": 0.0015, "step": 2118 }, { "epoch": 0.25, "learning_rate": 2.8510617876424716e-06, "loss": 0.0014, "step": 2119 }, { "epoch": 0.25, "learning_rate": 2.850989802039592e-06, "loss": 0.002, "step": 2120 }, { "epoch": 0.25, "learning_rate": 2.8509178164367128e-06, "loss": 0.0033, "step": 2121 }, { "epoch": 0.25, "learning_rate": 2.850845830833833e-06, "loss": 0.003, "step": 2122 }, { "epoch": 0.25, "learning_rate": 2.850773845230954e-06, "loss": 0.002, "step": 2123 }, { "epoch": 0.25, "learning_rate": 2.8507018596280742e-06, "loss": 0.0036, "step": 2124 }, { "epoch": 0.25, "learning_rate": 2.850629874025195e-06, "loss": 0.0014, "step": 2125 }, { "epoch": 0.25, "learning_rate": 2.8505578884223154e-06, "loss": 0.0021, "step": 2126 }, { "epoch": 0.25, "learning_rate": 2.850485902819436e-06, "loss": 0.0403, "step": 2127 }, { "epoch": 0.26, "learning_rate": 2.850413917216557e-06, "loss": 0.5542, "step": 2128 }, { "epoch": 0.26, "learning_rate": 2.8503419316136773e-06, "loss": 0.0016, "step": 2129 }, { "epoch": 0.26, "learning_rate": 2.850269946010798e-06, "loss": 0.002, "step": 2130 }, { "epoch": 0.26, "learning_rate": 2.8501979604079184e-06, "loss": 0.0013, "step": 2131 }, { "epoch": 0.26, "learning_rate": 2.850125974805039e-06, "loss": 0.0044, "step": 2132 }, { "epoch": 0.26, "learning_rate": 2.8500539892021596e-06, "loss": 0.0014, "step": 2133 }, { "epoch": 0.26, "learning_rate": 2.8499820035992803e-06, "loss": 0.0117, "step": 2134 }, { "epoch": 0.26, "learning_rate": 2.8499100179964007e-06, "loss": 0.1087, "step": 2135 }, { "epoch": 0.26, "learning_rate": 2.8498380323935215e-06, "loss": 0.0016, "step": 2136 }, { "epoch": 0.26, "learning_rate": 2.849766046790642e-06, "loss": 0.8208, "step": 2137 }, { "epoch": 0.26, "learning_rate": 2.8496940611877626e-06, "loss": 0.0014, "step": 2138 }, { "epoch": 0.26, "learning_rate": 2.849622075584883e-06, "loss": 0.0015, "step": 2139 }, { "epoch": 0.26, "learning_rate": 2.8495500899820037e-06, "loss": 0.0016, "step": 2140 }, { "epoch": 0.26, "learning_rate": 2.849478104379124e-06, "loss": 0.0014, "step": 2141 }, { "epoch": 0.26, "learning_rate": 2.849406118776245e-06, "loss": 0.9281, "step": 2142 }, { "epoch": 0.26, "learning_rate": 2.8493341331733656e-06, "loss": 0.0018, "step": 2143 }, { "epoch": 0.26, "learning_rate": 2.849262147570486e-06, "loss": 0.0014, "step": 2144 }, { "epoch": 0.26, "learning_rate": 2.8491901619676068e-06, "loss": 0.0017, "step": 2145 }, { "epoch": 0.26, "learning_rate": 2.849118176364727e-06, "loss": 0.0019, "step": 2146 }, { "epoch": 0.26, "learning_rate": 2.849046190761848e-06, "loss": 0.0226, "step": 2147 }, { "epoch": 0.26, "learning_rate": 2.8489742051589683e-06, "loss": 0.0011, "step": 2148 }, { "epoch": 0.26, "learning_rate": 2.848902219556089e-06, "loss": 0.0014, "step": 2149 }, { "epoch": 0.26, "learning_rate": 2.8488302339532094e-06, "loss": 0.0015, "step": 2150 }, { "epoch": 0.26, "learning_rate": 2.84875824835033e-06, "loss": 0.0013, "step": 2151 }, { "epoch": 0.26, "learning_rate": 2.8486862627474505e-06, "loss": 0.0016, "step": 2152 }, { "epoch": 0.26, "learning_rate": 2.848614277144571e-06, "loss": 0.0013, "step": 2153 }, { "epoch": 0.26, "learning_rate": 2.8485422915416917e-06, "loss": 0.019, "step": 2154 }, { "epoch": 0.26, "learning_rate": 2.848470305938812e-06, "loss": 0.1628, "step": 2155 }, { "epoch": 0.26, "learning_rate": 2.848398320335933e-06, "loss": 0.0028, "step": 2156 }, { "epoch": 0.26, "learning_rate": 2.8483263347330536e-06, "loss": 0.0017, "step": 2157 }, { "epoch": 0.26, "learning_rate": 2.848254349130174e-06, "loss": 0.0014, "step": 2158 }, { "epoch": 0.26, "learning_rate": 2.8481823635272947e-06, "loss": 0.002, "step": 2159 }, { "epoch": 0.26, "learning_rate": 2.8481103779244155e-06, "loss": 0.0013, "step": 2160 }, { "epoch": 0.26, "learning_rate": 2.848038392321536e-06, "loss": 0.3687, "step": 2161 }, { "epoch": 0.26, "learning_rate": 2.8479664067186566e-06, "loss": 0.7528, "step": 2162 }, { "epoch": 0.26, "learning_rate": 2.847894421115777e-06, "loss": 0.0221, "step": 2163 }, { "epoch": 0.26, "learning_rate": 2.8478224355128973e-06, "loss": 0.0016, "step": 2164 }, { "epoch": 0.26, "learning_rate": 2.847750449910018e-06, "loss": 0.0019, "step": 2165 }, { "epoch": 0.26, "learning_rate": 2.8476784643071385e-06, "loss": 0.0011, "step": 2166 }, { "epoch": 0.26, "learning_rate": 2.8476064787042592e-06, "loss": 0.0953, "step": 2167 }, { "epoch": 0.26, "learning_rate": 2.8475344931013796e-06, "loss": 0.0026, "step": 2168 }, { "epoch": 0.26, "learning_rate": 2.8474625074985004e-06, "loss": 0.0019, "step": 2169 }, { "epoch": 0.26, "learning_rate": 2.8473905218956207e-06, "loss": 0.0182, "step": 2170 }, { "epoch": 0.26, "learning_rate": 2.8473185362927415e-06, "loss": 0.3358, "step": 2171 }, { "epoch": 0.26, "learning_rate": 2.847246550689862e-06, "loss": 0.0056, "step": 2172 }, { "epoch": 0.26, "learning_rate": 2.8471745650869826e-06, "loss": 0.1384, "step": 2173 }, { "epoch": 0.26, "learning_rate": 2.8471025794841034e-06, "loss": 0.7105, "step": 2174 }, { "epoch": 0.26, "learning_rate": 2.8470305938812238e-06, "loss": 0.161, "step": 2175 }, { "epoch": 0.26, "learning_rate": 2.8469586082783446e-06, "loss": 0.002, "step": 2176 }, { "epoch": 0.26, "learning_rate": 2.846886622675465e-06, "loss": 0.0016, "step": 2177 }, { "epoch": 0.26, "learning_rate": 2.8468146370725857e-06, "loss": 0.1383, "step": 2178 }, { "epoch": 0.26, "learning_rate": 2.846742651469706e-06, "loss": 0.0016, "step": 2179 }, { "epoch": 0.26, "learning_rate": 2.846670665866827e-06, "loss": 0.3314, "step": 2180 }, { "epoch": 0.26, "learning_rate": 2.846598680263947e-06, "loss": 0.0012, "step": 2181 }, { "epoch": 0.26, "learning_rate": 2.846526694661068e-06, "loss": 0.0019, "step": 2182 }, { "epoch": 0.26, "learning_rate": 2.8464547090581883e-06, "loss": 0.2931, "step": 2183 }, { "epoch": 0.26, "learning_rate": 2.846382723455309e-06, "loss": 0.477, "step": 2184 }, { "epoch": 0.26, "learning_rate": 2.8463107378524294e-06, "loss": 0.253, "step": 2185 }, { "epoch": 0.26, "learning_rate": 2.8462387522495502e-06, "loss": 0.5585, "step": 2186 }, { "epoch": 0.26, "learning_rate": 2.8461667666466706e-06, "loss": 0.0012, "step": 2187 }, { "epoch": 0.26, "learning_rate": 2.8460947810437914e-06, "loss": 0.0014, "step": 2188 }, { "epoch": 0.26, "learning_rate": 2.846022795440912e-06, "loss": 0.0041, "step": 2189 }, { "epoch": 0.26, "learning_rate": 2.8459508098380325e-06, "loss": 0.0101, "step": 2190 }, { "epoch": 0.26, "learning_rate": 2.8458788242351533e-06, "loss": 0.0016, "step": 2191 }, { "epoch": 0.26, "learning_rate": 2.8458068386322736e-06, "loss": 0.0013, "step": 2192 }, { "epoch": 0.26, "learning_rate": 2.8457348530293944e-06, "loss": 0.0018, "step": 2193 }, { "epoch": 0.26, "learning_rate": 2.8456628674265148e-06, "loss": 0.0026, "step": 2194 }, { "epoch": 0.26, "learning_rate": 2.8455908818236355e-06, "loss": 0.0012, "step": 2195 }, { "epoch": 0.26, "learning_rate": 2.845518896220756e-06, "loss": 0.0029, "step": 2196 }, { "epoch": 0.26, "learning_rate": 2.8454469106178767e-06, "loss": 0.6954, "step": 2197 }, { "epoch": 0.26, "learning_rate": 2.845374925014997e-06, "loss": 0.4734, "step": 2198 }, { "epoch": 0.26, "learning_rate": 2.8453029394121174e-06, "loss": 0.6647, "step": 2199 }, { "epoch": 0.26, "learning_rate": 2.845230953809238e-06, "loss": 0.0017, "step": 2200 }, { "epoch": 0.26, "learning_rate": 2.8451589682063585e-06, "loss": 0.0013, "step": 2201 }, { "epoch": 0.26, "learning_rate": 2.8450869826034793e-06, "loss": 0.0012, "step": 2202 }, { "epoch": 0.26, "learning_rate": 2.8450149970005997e-06, "loss": 0.0084, "step": 2203 }, { "epoch": 0.26, "learning_rate": 2.8449430113977204e-06, "loss": 0.2255, "step": 2204 }, { "epoch": 0.26, "learning_rate": 2.844871025794841e-06, "loss": 0.5776, "step": 2205 }, { "epoch": 0.26, "learning_rate": 2.844799040191962e-06, "loss": 0.0012, "step": 2206 }, { "epoch": 0.26, "learning_rate": 2.8447270545890823e-06, "loss": 0.0011, "step": 2207 }, { "epoch": 0.26, "learning_rate": 2.844655068986203e-06, "loss": 0.0022, "step": 2208 }, { "epoch": 0.26, "learning_rate": 2.8445830833833235e-06, "loss": 0.0021, "step": 2209 }, { "epoch": 0.26, "learning_rate": 2.844511097780444e-06, "loss": 0.0099, "step": 2210 }, { "epoch": 0.26, "learning_rate": 2.8444391121775646e-06, "loss": 0.0014, "step": 2211 }, { "epoch": 0.27, "learning_rate": 2.844367126574685e-06, "loss": 0.0705, "step": 2212 }, { "epoch": 0.27, "learning_rate": 2.8442951409718057e-06, "loss": 0.2628, "step": 2213 }, { "epoch": 0.27, "learning_rate": 2.844223155368926e-06, "loss": 0.0011, "step": 2214 }, { "epoch": 0.27, "learning_rate": 2.844151169766047e-06, "loss": 0.0035, "step": 2215 }, { "epoch": 0.27, "learning_rate": 2.8440791841631672e-06, "loss": 0.0014, "step": 2216 }, { "epoch": 0.27, "learning_rate": 2.844007198560288e-06, "loss": 0.0012, "step": 2217 }, { "epoch": 0.27, "learning_rate": 2.8439352129574084e-06, "loss": 0.0011, "step": 2218 }, { "epoch": 0.27, "learning_rate": 2.843863227354529e-06, "loss": 0.0019, "step": 2219 }, { "epoch": 0.27, "learning_rate": 2.84379124175165e-06, "loss": 0.0013, "step": 2220 }, { "epoch": 0.27, "learning_rate": 2.8437192561487703e-06, "loss": 0.2339, "step": 2221 }, { "epoch": 0.27, "learning_rate": 2.843647270545891e-06, "loss": 0.0027, "step": 2222 }, { "epoch": 0.27, "learning_rate": 2.8435752849430114e-06, "loss": 0.0015, "step": 2223 }, { "epoch": 0.27, "learning_rate": 2.843503299340132e-06, "loss": 0.0021, "step": 2224 }, { "epoch": 0.27, "learning_rate": 2.8434313137372525e-06, "loss": 0.0022, "step": 2225 }, { "epoch": 0.27, "learning_rate": 2.8433593281343733e-06, "loss": 0.0037, "step": 2226 }, { "epoch": 0.27, "learning_rate": 2.8432873425314937e-06, "loss": 0.5554, "step": 2227 }, { "epoch": 0.27, "learning_rate": 2.8432153569286145e-06, "loss": 0.0021, "step": 2228 }, { "epoch": 0.27, "learning_rate": 2.843143371325735e-06, "loss": 0.51, "step": 2229 }, { "epoch": 0.27, "learning_rate": 2.8430713857228556e-06, "loss": 0.7301, "step": 2230 }, { "epoch": 0.27, "learning_rate": 2.842999400119976e-06, "loss": 0.2057, "step": 2231 }, { "epoch": 0.27, "learning_rate": 2.8429274145170967e-06, "loss": 0.001, "step": 2232 }, { "epoch": 0.27, "learning_rate": 2.842855428914217e-06, "loss": 0.6768, "step": 2233 }, { "epoch": 0.27, "learning_rate": 2.842783443311338e-06, "loss": 0.0024, "step": 2234 }, { "epoch": 0.27, "learning_rate": 2.842711457708458e-06, "loss": 0.7531, "step": 2235 }, { "epoch": 0.27, "learning_rate": 2.842639472105579e-06, "loss": 0.0025, "step": 2236 }, { "epoch": 0.27, "learning_rate": 2.8425674865026998e-06, "loss": 0.0018, "step": 2237 }, { "epoch": 0.27, "learning_rate": 2.84249550089982e-06, "loss": 0.0013, "step": 2238 }, { "epoch": 0.27, "learning_rate": 2.842423515296941e-06, "loss": 0.0012, "step": 2239 }, { "epoch": 0.27, "learning_rate": 2.8423515296940613e-06, "loss": 0.0014, "step": 2240 }, { "epoch": 0.27, "learning_rate": 2.842279544091182e-06, "loss": 0.0054, "step": 2241 }, { "epoch": 0.27, "learning_rate": 2.8422075584883024e-06, "loss": 0.0013, "step": 2242 }, { "epoch": 0.27, "learning_rate": 2.842135572885423e-06, "loss": 0.0012, "step": 2243 }, { "epoch": 0.27, "learning_rate": 2.8420635872825435e-06, "loss": 0.0022, "step": 2244 }, { "epoch": 0.27, "learning_rate": 2.841991601679664e-06, "loss": 0.518, "step": 2245 }, { "epoch": 0.27, "learning_rate": 2.8419196160767847e-06, "loss": 0.0107, "step": 2246 }, { "epoch": 0.27, "learning_rate": 2.841847630473905e-06, "loss": 0.0015, "step": 2247 }, { "epoch": 0.27, "learning_rate": 2.841775644871026e-06, "loss": 0.0376, "step": 2248 }, { "epoch": 0.27, "learning_rate": 2.841703659268146e-06, "loss": 0.0014, "step": 2249 }, { "epoch": 0.27, "learning_rate": 2.841631673665267e-06, "loss": 0.894, "step": 2250 }, { "epoch": 0.27, "learning_rate": 2.8415596880623877e-06, "loss": 0.0022, "step": 2251 }, { "epoch": 0.27, "learning_rate": 2.8414877024595085e-06, "loss": 0.678, "step": 2252 }, { "epoch": 0.27, "learning_rate": 2.841415716856629e-06, "loss": 0.0017, "step": 2253 }, { "epoch": 0.27, "learning_rate": 2.8413437312537496e-06, "loss": 0.0012, "step": 2254 }, { "epoch": 0.27, "learning_rate": 2.84127174565087e-06, "loss": 0.0012, "step": 2255 }, { "epoch": 0.27, "learning_rate": 2.8411997600479903e-06, "loss": 0.0057, "step": 2256 }, { "epoch": 0.27, "learning_rate": 2.841127774445111e-06, "loss": 0.0011, "step": 2257 }, { "epoch": 0.27, "learning_rate": 2.8410557888422315e-06, "loss": 0.0012, "step": 2258 }, { "epoch": 0.27, "learning_rate": 2.8409838032393522e-06, "loss": 0.036, "step": 2259 }, { "epoch": 0.27, "learning_rate": 2.8409118176364726e-06, "loss": 0.0014, "step": 2260 }, { "epoch": 0.27, "learning_rate": 2.8408398320335934e-06, "loss": 0.7219, "step": 2261 }, { "epoch": 0.27, "learning_rate": 2.8407678464307137e-06, "loss": 1.6755, "step": 2262 }, { "epoch": 0.27, "learning_rate": 2.8406958608278345e-06, "loss": 0.0013, "step": 2263 }, { "epoch": 0.27, "learning_rate": 2.840623875224955e-06, "loss": 0.4458, "step": 2264 }, { "epoch": 0.27, "learning_rate": 2.8405518896220756e-06, "loss": 0.0015, "step": 2265 }, { "epoch": 0.27, "learning_rate": 2.8404799040191964e-06, "loss": 0.0019, "step": 2266 }, { "epoch": 0.27, "learning_rate": 2.8404079184163168e-06, "loss": 0.002, "step": 2267 }, { "epoch": 0.27, "learning_rate": 2.8403359328134376e-06, "loss": 0.0012, "step": 2268 }, { "epoch": 0.27, "learning_rate": 2.840263947210558e-06, "loss": 0.0013, "step": 2269 }, { "epoch": 0.27, "learning_rate": 2.8401919616076787e-06, "loss": 0.0376, "step": 2270 }, { "epoch": 0.27, "learning_rate": 2.840119976004799e-06, "loss": 0.0019, "step": 2271 }, { "epoch": 0.27, "learning_rate": 2.84004799040192e-06, "loss": 0.0029, "step": 2272 }, { "epoch": 0.27, "learning_rate": 2.83997600479904e-06, "loss": 0.002, "step": 2273 }, { "epoch": 0.27, "learning_rate": 2.839904019196161e-06, "loss": 0.0014, "step": 2274 }, { "epoch": 0.27, "learning_rate": 2.8398320335932813e-06, "loss": 0.0015, "step": 2275 }, { "epoch": 0.27, "learning_rate": 2.839760047990402e-06, "loss": 0.6934, "step": 2276 }, { "epoch": 0.27, "learning_rate": 2.8396880623875224e-06, "loss": 0.0532, "step": 2277 }, { "epoch": 0.27, "learning_rate": 2.8396160767846432e-06, "loss": 0.0022, "step": 2278 }, { "epoch": 0.27, "learning_rate": 2.8395440911817636e-06, "loss": 0.0032, "step": 2279 }, { "epoch": 0.27, "learning_rate": 2.8394721055788844e-06, "loss": 0.0011, "step": 2280 }, { "epoch": 0.27, "learning_rate": 2.8394001199760047e-06, "loss": 0.0028, "step": 2281 }, { "epoch": 0.27, "learning_rate": 2.8393281343731255e-06, "loss": 0.7534, "step": 2282 }, { "epoch": 0.27, "learning_rate": 2.8392561487702463e-06, "loss": 0.0024, "step": 2283 }, { "epoch": 0.27, "learning_rate": 2.8391841631673666e-06, "loss": 0.0371, "step": 2284 }, { "epoch": 0.27, "learning_rate": 2.8391121775644874e-06, "loss": 0.0013, "step": 2285 }, { "epoch": 0.27, "learning_rate": 2.8390401919616078e-06, "loss": 0.0015, "step": 2286 }, { "epoch": 0.27, "learning_rate": 2.8389682063587285e-06, "loss": 0.0012, "step": 2287 }, { "epoch": 0.27, "learning_rate": 2.838896220755849e-06, "loss": 0.0369, "step": 2288 }, { "epoch": 0.27, "learning_rate": 2.8388242351529697e-06, "loss": 0.0029, "step": 2289 }, { "epoch": 0.27, "learning_rate": 2.83875224955009e-06, "loss": 0.0172, "step": 2290 }, { "epoch": 0.27, "learning_rate": 2.8386802639472104e-06, "loss": 0.0084, "step": 2291 }, { "epoch": 0.27, "learning_rate": 2.838608278344331e-06, "loss": 0.0033, "step": 2292 }, { "epoch": 0.27, "learning_rate": 2.8385362927414515e-06, "loss": 0.2624, "step": 2293 }, { "epoch": 0.27, "learning_rate": 2.8384643071385723e-06, "loss": 1.4834, "step": 2294 }, { "epoch": 0.28, "learning_rate": 2.8383923215356926e-06, "loss": 0.0014, "step": 2295 }, { "epoch": 0.28, "learning_rate": 2.8383203359328134e-06, "loss": 0.0045, "step": 2296 }, { "epoch": 0.28, "learning_rate": 2.838248350329934e-06, "loss": 0.0011, "step": 2297 }, { "epoch": 0.28, "learning_rate": 2.838176364727055e-06, "loss": 0.0017, "step": 2298 }, { "epoch": 0.28, "learning_rate": 2.8381043791241753e-06, "loss": 1.1064, "step": 2299 }, { "epoch": 0.28, "learning_rate": 2.838032393521296e-06, "loss": 0.0037, "step": 2300 }, { "epoch": 0.28, "learning_rate": 2.8379604079184165e-06, "loss": 0.2273, "step": 2301 }, { "epoch": 0.28, "learning_rate": 2.837888422315537e-06, "loss": 0.1965, "step": 2302 }, { "epoch": 0.28, "learning_rate": 2.8378164367126576e-06, "loss": 0.1307, "step": 2303 }, { "epoch": 0.28, "learning_rate": 2.837744451109778e-06, "loss": 0.002, "step": 2304 }, { "epoch": 0.28, "learning_rate": 2.8376724655068987e-06, "loss": 0.9109, "step": 2305 }, { "epoch": 0.28, "learning_rate": 2.837600479904019e-06, "loss": 1.471, "step": 2306 }, { "epoch": 0.28, "learning_rate": 2.83752849430114e-06, "loss": 0.0019, "step": 2307 }, { "epoch": 0.28, "learning_rate": 2.8374565086982602e-06, "loss": 0.0017, "step": 2308 }, { "epoch": 0.28, "learning_rate": 2.837384523095381e-06, "loss": 0.004, "step": 2309 }, { "epoch": 0.28, "learning_rate": 2.8373125374925014e-06, "loss": 0.0012, "step": 2310 }, { "epoch": 0.28, "learning_rate": 2.837240551889622e-06, "loss": 0.002, "step": 2311 }, { "epoch": 0.28, "learning_rate": 2.837168566286743e-06, "loss": 0.0011, "step": 2312 }, { "epoch": 0.28, "learning_rate": 2.8370965806838633e-06, "loss": 0.0026, "step": 2313 }, { "epoch": 0.28, "learning_rate": 2.837024595080984e-06, "loss": 0.0013, "step": 2314 }, { "epoch": 0.28, "learning_rate": 2.8369526094781044e-06, "loss": 0.6474, "step": 2315 }, { "epoch": 0.28, "learning_rate": 2.836880623875225e-06, "loss": 0.0014, "step": 2316 }, { "epoch": 0.28, "learning_rate": 2.8368086382723455e-06, "loss": 0.0022, "step": 2317 }, { "epoch": 0.28, "learning_rate": 2.8367366526694663e-06, "loss": 0.0013, "step": 2318 }, { "epoch": 0.28, "learning_rate": 2.8366646670665867e-06, "loss": 0.2442, "step": 2319 }, { "epoch": 0.28, "learning_rate": 2.8365926814637074e-06, "loss": 0.0013, "step": 2320 }, { "epoch": 0.28, "learning_rate": 2.836520695860828e-06, "loss": 0.0052, "step": 2321 }, { "epoch": 0.28, "learning_rate": 2.8364487102579486e-06, "loss": 0.0012, "step": 2322 }, { "epoch": 0.28, "learning_rate": 2.836376724655069e-06, "loss": 0.0114, "step": 2323 }, { "epoch": 0.28, "learning_rate": 2.8363047390521897e-06, "loss": 0.0088, "step": 2324 }, { "epoch": 0.28, "learning_rate": 2.83623275344931e-06, "loss": 0.002, "step": 2325 }, { "epoch": 0.28, "learning_rate": 2.8361607678464304e-06, "loss": 0.0244, "step": 2326 }, { "epoch": 0.28, "learning_rate": 2.836088782243551e-06, "loss": 0.6475, "step": 2327 }, { "epoch": 0.28, "learning_rate": 2.836016796640672e-06, "loss": 0.002, "step": 2328 }, { "epoch": 0.28, "learning_rate": 2.8359448110377928e-06, "loss": 0.0021, "step": 2329 }, { "epoch": 0.28, "learning_rate": 2.835872825434913e-06, "loss": 0.1904, "step": 2330 }, { "epoch": 0.28, "learning_rate": 2.835800839832034e-06, "loss": 0.0028, "step": 2331 }, { "epoch": 0.28, "learning_rate": 2.8357288542291542e-06, "loss": 0.0012, "step": 2332 }, { "epoch": 0.28, "learning_rate": 2.835656868626275e-06, "loss": 0.0044, "step": 2333 }, { "epoch": 0.28, "learning_rate": 2.8355848830233954e-06, "loss": 0.0012, "step": 2334 }, { "epoch": 0.28, "learning_rate": 2.835512897420516e-06, "loss": 0.0019, "step": 2335 }, { "epoch": 0.28, "learning_rate": 2.8354409118176365e-06, "loss": 0.0011, "step": 2336 }, { "epoch": 0.28, "learning_rate": 2.835368926214757e-06, "loss": 0.0066, "step": 2337 }, { "epoch": 0.28, "learning_rate": 2.8352969406118777e-06, "loss": 0.2676, "step": 2338 }, { "epoch": 0.28, "learning_rate": 2.835224955008998e-06, "loss": 0.3487, "step": 2339 }, { "epoch": 0.28, "learning_rate": 2.8351529694061188e-06, "loss": 0.0011, "step": 2340 }, { "epoch": 0.28, "learning_rate": 2.835080983803239e-06, "loss": 0.0845, "step": 2341 }, { "epoch": 0.28, "learning_rate": 2.83500899820036e-06, "loss": 0.0016, "step": 2342 }, { "epoch": 0.28, "learning_rate": 2.8349370125974807e-06, "loss": 0.0013, "step": 2343 }, { "epoch": 0.28, "learning_rate": 2.834865026994601e-06, "loss": 0.0012, "step": 2344 }, { "epoch": 0.28, "learning_rate": 2.834793041391722e-06, "loss": 0.0016, "step": 2345 }, { "epoch": 0.28, "learning_rate": 2.8347210557888426e-06, "loss": 0.003, "step": 2346 }, { "epoch": 0.28, "learning_rate": 2.834649070185963e-06, "loss": 0.0016, "step": 2347 }, { "epoch": 0.28, "learning_rate": 2.8345770845830833e-06, "loss": 0.5927, "step": 2348 }, { "epoch": 0.28, "learning_rate": 2.834505098980204e-06, "loss": 1.7297, "step": 2349 }, { "epoch": 0.28, "learning_rate": 2.8344331133773245e-06, "loss": 0.0011, "step": 2350 }, { "epoch": 0.28, "learning_rate": 2.8343611277744452e-06, "loss": 0.001, "step": 2351 }, { "epoch": 0.28, "learning_rate": 2.8342891421715656e-06, "loss": 0.0018, "step": 2352 }, { "epoch": 0.28, "learning_rate": 2.8342171565686864e-06, "loss": 0.3725, "step": 2353 }, { "epoch": 0.28, "learning_rate": 2.8341451709658067e-06, "loss": 0.0029, "step": 2354 }, { "epoch": 0.28, "learning_rate": 2.8340731853629275e-06, "loss": 0.002, "step": 2355 }, { "epoch": 0.28, "learning_rate": 2.834001199760048e-06, "loss": 0.0014, "step": 2356 }, { "epoch": 0.28, "learning_rate": 2.8339292141571686e-06, "loss": 0.0034, "step": 2357 }, { "epoch": 0.28, "learning_rate": 2.833857228554289e-06, "loss": 0.0017, "step": 2358 }, { "epoch": 0.28, "learning_rate": 2.8337852429514098e-06, "loss": 0.007, "step": 2359 }, { "epoch": 0.28, "learning_rate": 2.8337132573485305e-06, "loss": 1.6928, "step": 2360 }, { "epoch": 0.28, "learning_rate": 2.833641271745651e-06, "loss": 0.3464, "step": 2361 }, { "epoch": 0.28, "learning_rate": 2.8335692861427717e-06, "loss": 0.0063, "step": 2362 }, { "epoch": 0.28, "learning_rate": 2.833497300539892e-06, "loss": 0.0015, "step": 2363 }, { "epoch": 0.28, "learning_rate": 2.833425314937013e-06, "loss": 0.0227, "step": 2364 }, { "epoch": 0.28, "learning_rate": 2.833353329334133e-06, "loss": 0.0877, "step": 2365 }, { "epoch": 0.28, "learning_rate": 2.833281343731254e-06, "loss": 0.3206, "step": 2366 }, { "epoch": 0.28, "learning_rate": 2.8332093581283743e-06, "loss": 0.0013, "step": 2367 }, { "epoch": 0.28, "learning_rate": 2.833137372525495e-06, "loss": 0.7797, "step": 2368 }, { "epoch": 0.28, "learning_rate": 2.8330653869226154e-06, "loss": 0.0011, "step": 2369 }, { "epoch": 0.28, "learning_rate": 2.832993401319736e-06, "loss": 0.0012, "step": 2370 }, { "epoch": 0.28, "learning_rate": 2.8329214157168566e-06, "loss": 0.0025, "step": 2371 }, { "epoch": 0.28, "learning_rate": 2.832849430113977e-06, "loss": 0.0013, "step": 2372 }, { "epoch": 0.28, "learning_rate": 2.8327774445110977e-06, "loss": 0.7267, "step": 2373 }, { "epoch": 0.28, "learning_rate": 2.8327054589082185e-06, "loss": 0.4701, "step": 2374 }, { "epoch": 0.28, "learning_rate": 2.8326334733053393e-06, "loss": 0.0013, "step": 2375 }, { "epoch": 0.28, "learning_rate": 2.8325614877024596e-06, "loss": 0.0013, "step": 2376 }, { "epoch": 0.28, "learning_rate": 2.8324895020995804e-06, "loss": 0.0012, "step": 2377 }, { "epoch": 0.28, "learning_rate": 2.8324175164967007e-06, "loss": 0.0022, "step": 2378 }, { "epoch": 0.29, "learning_rate": 2.8323455308938215e-06, "loss": 0.0011, "step": 2379 }, { "epoch": 0.29, "learning_rate": 2.832273545290942e-06, "loss": 0.0014, "step": 2380 }, { "epoch": 0.29, "learning_rate": 2.8322015596880627e-06, "loss": 0.3024, "step": 2381 }, { "epoch": 0.29, "learning_rate": 2.832129574085183e-06, "loss": 0.0108, "step": 2382 }, { "epoch": 0.29, "learning_rate": 2.8320575884823034e-06, "loss": 0.0375, "step": 2383 }, { "epoch": 0.29, "learning_rate": 2.831985602879424e-06, "loss": 2.0022, "step": 2384 }, { "epoch": 0.29, "learning_rate": 2.8319136172765445e-06, "loss": 0.0012, "step": 2385 }, { "epoch": 0.29, "learning_rate": 2.8318416316736653e-06, "loss": 0.0013, "step": 2386 }, { "epoch": 0.29, "learning_rate": 2.8317696460707856e-06, "loss": 0.0013, "step": 2387 }, { "epoch": 0.29, "learning_rate": 2.8316976604679064e-06, "loss": 0.0212, "step": 2388 }, { "epoch": 0.29, "learning_rate": 2.831625674865027e-06, "loss": 0.0054, "step": 2389 }, { "epoch": 0.29, "learning_rate": 2.8315536892621475e-06, "loss": 0.0015, "step": 2390 }, { "epoch": 0.29, "learning_rate": 2.8314817036592683e-06, "loss": 0.0027, "step": 2391 }, { "epoch": 0.29, "learning_rate": 2.831409718056389e-06, "loss": 0.7382, "step": 2392 }, { "epoch": 0.29, "learning_rate": 2.8313377324535095e-06, "loss": 0.002, "step": 2393 }, { "epoch": 0.29, "learning_rate": 2.8312657468506302e-06, "loss": 0.0014, "step": 2394 }, { "epoch": 0.29, "learning_rate": 2.8311937612477506e-06, "loss": 0.0026, "step": 2395 }, { "epoch": 0.29, "learning_rate": 2.831121775644871e-06, "loss": 0.4409, "step": 2396 }, { "epoch": 0.29, "learning_rate": 2.8310497900419917e-06, "loss": 0.0041, "step": 2397 }, { "epoch": 0.29, "learning_rate": 2.830977804439112e-06, "loss": 0.0023, "step": 2398 }, { "epoch": 0.29, "learning_rate": 2.830905818836233e-06, "loss": 0.0014, "step": 2399 }, { "epoch": 0.29, "learning_rate": 2.8308338332333532e-06, "loss": 0.7131, "step": 2400 }, { "epoch": 0.29, "learning_rate": 2.830761847630474e-06, "loss": 0.0012, "step": 2401 }, { "epoch": 0.29, "learning_rate": 2.8306898620275943e-06, "loss": 0.7946, "step": 2402 }, { "epoch": 0.29, "learning_rate": 2.830617876424715e-06, "loss": 0.0016, "step": 2403 }, { "epoch": 0.29, "learning_rate": 2.8305458908218355e-06, "loss": 0.1095, "step": 2404 }, { "epoch": 0.29, "learning_rate": 2.8304739052189563e-06, "loss": 0.0071, "step": 2405 }, { "epoch": 0.29, "learning_rate": 2.830401919616077e-06, "loss": 0.048, "step": 2406 }, { "epoch": 0.29, "learning_rate": 2.8303299340131974e-06, "loss": 0.0176, "step": 2407 }, { "epoch": 0.29, "learning_rate": 2.830257948410318e-06, "loss": 0.0015, "step": 2408 }, { "epoch": 0.29, "learning_rate": 2.8301859628074385e-06, "loss": 0.0073, "step": 2409 }, { "epoch": 0.29, "learning_rate": 2.8301139772045593e-06, "loss": 0.0013, "step": 2410 }, { "epoch": 0.29, "learning_rate": 2.8300419916016797e-06, "loss": 0.8309, "step": 2411 }, { "epoch": 0.29, "learning_rate": 2.8299700059988004e-06, "loss": 0.0076, "step": 2412 }, { "epoch": 0.29, "learning_rate": 2.829898020395921e-06, "loss": 0.0036, "step": 2413 }, { "epoch": 0.29, "learning_rate": 2.8298260347930416e-06, "loss": 0.0047, "step": 2414 }, { "epoch": 0.29, "learning_rate": 2.829754049190162e-06, "loss": 0.0014, "step": 2415 }, { "epoch": 0.29, "learning_rate": 2.8296820635872827e-06, "loss": 0.6404, "step": 2416 }, { "epoch": 0.29, "learning_rate": 2.829610077984403e-06, "loss": 0.0023, "step": 2417 }, { "epoch": 0.29, "learning_rate": 2.8295380923815234e-06, "loss": 0.0013, "step": 2418 }, { "epoch": 0.29, "learning_rate": 2.829466106778644e-06, "loss": 0.0047, "step": 2419 }, { "epoch": 0.29, "learning_rate": 2.829394121175765e-06, "loss": 0.0012, "step": 2420 }, { "epoch": 0.29, "learning_rate": 2.8293221355728858e-06, "loss": 0.0013, "step": 2421 }, { "epoch": 0.29, "learning_rate": 2.829250149970006e-06, "loss": 1.1245, "step": 2422 }, { "epoch": 0.29, "learning_rate": 2.829178164367127e-06, "loss": 0.0816, "step": 2423 }, { "epoch": 0.29, "learning_rate": 2.8291061787642472e-06, "loss": 0.4307, "step": 2424 }, { "epoch": 0.29, "learning_rate": 2.829034193161368e-06, "loss": 0.0021, "step": 2425 }, { "epoch": 0.29, "learning_rate": 2.8289622075584884e-06, "loss": 0.0011, "step": 2426 }, { "epoch": 0.29, "learning_rate": 2.828890221955609e-06, "loss": 0.0028, "step": 2427 }, { "epoch": 0.29, "learning_rate": 2.8288182363527295e-06, "loss": 0.0017, "step": 2428 }, { "epoch": 0.29, "learning_rate": 2.82874625074985e-06, "loss": 0.0017, "step": 2429 }, { "epoch": 0.29, "learning_rate": 2.8286742651469706e-06, "loss": 0.0019, "step": 2430 }, { "epoch": 0.29, "learning_rate": 2.828602279544091e-06, "loss": 0.399, "step": 2431 }, { "epoch": 0.29, "learning_rate": 2.8285302939412118e-06, "loss": 0.0015, "step": 2432 }, { "epoch": 0.29, "learning_rate": 2.828458308338332e-06, "loss": 0.0019, "step": 2433 }, { "epoch": 0.29, "learning_rate": 2.828386322735453e-06, "loss": 0.0022, "step": 2434 }, { "epoch": 0.29, "learning_rate": 2.8283143371325737e-06, "loss": 0.0081, "step": 2435 }, { "epoch": 0.29, "learning_rate": 2.828242351529694e-06, "loss": 0.0019, "step": 2436 }, { "epoch": 0.29, "learning_rate": 2.828170365926815e-06, "loss": 0.0018, "step": 2437 }, { "epoch": 0.29, "learning_rate": 2.8280983803239356e-06, "loss": 0.0227, "step": 2438 }, { "epoch": 0.29, "learning_rate": 2.828026394721056e-06, "loss": 0.0018, "step": 2439 }, { "epoch": 0.29, "learning_rate": 2.8279544091181767e-06, "loss": 0.0038, "step": 2440 }, { "epoch": 0.29, "learning_rate": 2.827882423515297e-06, "loss": 0.0025, "step": 2441 }, { "epoch": 0.29, "learning_rate": 2.8278104379124174e-06, "loss": 0.0017, "step": 2442 }, { "epoch": 0.29, "learning_rate": 2.8277384523095382e-06, "loss": 0.5846, "step": 2443 }, { "epoch": 0.29, "learning_rate": 2.8276664667066586e-06, "loss": 0.1254, "step": 2444 }, { "epoch": 0.29, "learning_rate": 2.8275944811037794e-06, "loss": 0.0013, "step": 2445 }, { "epoch": 0.29, "learning_rate": 2.8275224955008997e-06, "loss": 0.8544, "step": 2446 }, { "epoch": 0.29, "learning_rate": 2.8274505098980205e-06, "loss": 0.0018, "step": 2447 }, { "epoch": 0.29, "learning_rate": 2.827378524295141e-06, "loss": 0.0021, "step": 2448 }, { "epoch": 0.29, "learning_rate": 2.8273065386922616e-06, "loss": 0.0042, "step": 2449 }, { "epoch": 0.29, "learning_rate": 2.827234553089382e-06, "loss": 0.0149, "step": 2450 }, { "epoch": 0.29, "learning_rate": 2.8271625674865028e-06, "loss": 0.4615, "step": 2451 }, { "epoch": 0.29, "learning_rate": 2.8270905818836235e-06, "loss": 0.0015, "step": 2452 }, { "epoch": 0.29, "learning_rate": 2.827018596280744e-06, "loss": 0.0017, "step": 2453 }, { "epoch": 0.29, "learning_rate": 2.8269466106778647e-06, "loss": 0.0055, "step": 2454 }, { "epoch": 0.29, "learning_rate": 2.826874625074985e-06, "loss": 0.0013, "step": 2455 }, { "epoch": 0.29, "learning_rate": 2.826802639472106e-06, "loss": 0.002, "step": 2456 }, { "epoch": 0.29, "learning_rate": 2.826730653869226e-06, "loss": 0.6179, "step": 2457 }, { "epoch": 0.29, "learning_rate": 2.826658668266347e-06, "loss": 0.0014, "step": 2458 }, { "epoch": 0.29, "learning_rate": 2.8265866826634673e-06, "loss": 0.0016, "step": 2459 }, { "epoch": 0.29, "learning_rate": 2.826514697060588e-06, "loss": 0.4613, "step": 2460 }, { "epoch": 0.29, "learning_rate": 2.8264427114577084e-06, "loss": 0.3059, "step": 2461 }, { "epoch": 0.3, "learning_rate": 2.826370725854829e-06, "loss": 0.0015, "step": 2462 }, { "epoch": 0.3, "learning_rate": 2.8262987402519496e-06, "loss": 0.0016, "step": 2463 }, { "epoch": 0.3, "learning_rate": 2.82622675464907e-06, "loss": 0.0014, "step": 2464 }, { "epoch": 0.3, "learning_rate": 2.8261547690461907e-06, "loss": 0.0016, "step": 2465 }, { "epoch": 0.3, "learning_rate": 2.8260827834433115e-06, "loss": 0.0011, "step": 2466 }, { "epoch": 0.3, "learning_rate": 2.826010797840432e-06, "loss": 0.2135, "step": 2467 }, { "epoch": 0.3, "learning_rate": 2.8259388122375526e-06, "loss": 0.0019, "step": 2468 }, { "epoch": 0.3, "learning_rate": 2.8258668266346734e-06, "loss": 0.1702, "step": 2469 }, { "epoch": 0.3, "learning_rate": 2.8257948410317937e-06, "loss": 0.002, "step": 2470 }, { "epoch": 0.3, "learning_rate": 2.8257228554289145e-06, "loss": 0.5548, "step": 2471 }, { "epoch": 0.3, "learning_rate": 2.825650869826035e-06, "loss": 0.0029, "step": 2472 }, { "epoch": 0.3, "learning_rate": 2.8255788842231556e-06, "loss": 0.1678, "step": 2473 }, { "epoch": 0.3, "learning_rate": 2.825506898620276e-06, "loss": 0.0032, "step": 2474 }, { "epoch": 0.3, "learning_rate": 2.8254349130173964e-06, "loss": 0.0014, "step": 2475 }, { "epoch": 0.3, "learning_rate": 2.825362927414517e-06, "loss": 0.0012, "step": 2476 }, { "epoch": 0.3, "learning_rate": 2.8252909418116375e-06, "loss": 0.014, "step": 2477 }, { "epoch": 0.3, "learning_rate": 2.8252189562087583e-06, "loss": 0.0032, "step": 2478 }, { "epoch": 0.3, "learning_rate": 2.8251469706058786e-06, "loss": 0.1035, "step": 2479 }, { "epoch": 0.3, "learning_rate": 2.8250749850029994e-06, "loss": 0.0013, "step": 2480 }, { "epoch": 0.3, "learning_rate": 2.8250029994001198e-06, "loss": 0.0162, "step": 2481 }, { "epoch": 0.3, "learning_rate": 2.8249310137972405e-06, "loss": 0.0014, "step": 2482 }, { "epoch": 0.3, "learning_rate": 2.8248590281943613e-06, "loss": 0.0023, "step": 2483 }, { "epoch": 0.3, "learning_rate": 2.824787042591482e-06, "loss": 0.1786, "step": 2484 }, { "epoch": 0.3, "learning_rate": 2.8247150569886025e-06, "loss": 0.0418, "step": 2485 }, { "epoch": 0.3, "learning_rate": 2.8246430713857232e-06, "loss": 0.3789, "step": 2486 }, { "epoch": 0.3, "learning_rate": 2.8245710857828436e-06, "loss": 0.0017, "step": 2487 }, { "epoch": 0.3, "learning_rate": 2.824499100179964e-06, "loss": 0.0013, "step": 2488 }, { "epoch": 0.3, "learning_rate": 2.8244271145770847e-06, "loss": 0.0015, "step": 2489 }, { "epoch": 0.3, "learning_rate": 2.824355128974205e-06, "loss": 0.6984, "step": 2490 }, { "epoch": 0.3, "learning_rate": 2.824283143371326e-06, "loss": 0.0019, "step": 2491 }, { "epoch": 0.3, "learning_rate": 2.824211157768446e-06, "loss": 0.0495, "step": 2492 }, { "epoch": 0.3, "learning_rate": 2.824139172165567e-06, "loss": 0.009, "step": 2493 }, { "epoch": 0.3, "learning_rate": 2.8240671865626873e-06, "loss": 0.0172, "step": 2494 }, { "epoch": 0.3, "learning_rate": 2.823995200959808e-06, "loss": 0.0172, "step": 2495 }, { "epoch": 0.3, "learning_rate": 2.8239232153569285e-06, "loss": 0.0015, "step": 2496 }, { "epoch": 0.3, "learning_rate": 2.8238512297540493e-06, "loss": 0.0019, "step": 2497 }, { "epoch": 0.3, "learning_rate": 2.82377924415117e-06, "loss": 0.0012, "step": 2498 }, { "epoch": 0.3, "learning_rate": 2.8237072585482904e-06, "loss": 0.0012, "step": 2499 }, { "epoch": 0.3, "learning_rate": 2.823635272945411e-06, "loss": 0.8349, "step": 2500 }, { "epoch": 0.3, "learning_rate": 2.8235632873425315e-06, "loss": 0.0016, "step": 2501 }, { "epoch": 0.3, "learning_rate": 2.8234913017396523e-06, "loss": 0.0014, "step": 2502 }, { "epoch": 0.3, "learning_rate": 2.8234193161367727e-06, "loss": 0.0016, "step": 2503 }, { "epoch": 0.3, "learning_rate": 2.8233473305338934e-06, "loss": 0.0012, "step": 2504 }, { "epoch": 0.3, "learning_rate": 2.8232753449310138e-06, "loss": 0.0023, "step": 2505 }, { "epoch": 0.3, "learning_rate": 2.8232033593281346e-06, "loss": 0.0021, "step": 2506 }, { "epoch": 0.3, "learning_rate": 2.823131373725255e-06, "loss": 0.0014, "step": 2507 }, { "epoch": 0.3, "learning_rate": 2.8230593881223757e-06, "loss": 0.0516, "step": 2508 }, { "epoch": 0.3, "learning_rate": 2.822987402519496e-06, "loss": 0.0015, "step": 2509 }, { "epoch": 0.3, "learning_rate": 2.8229154169166164e-06, "loss": 0.3666, "step": 2510 }, { "epoch": 0.3, "learning_rate": 2.822843431313737e-06, "loss": 0.0045, "step": 2511 }, { "epoch": 0.3, "learning_rate": 2.822771445710858e-06, "loss": 0.0018, "step": 2512 }, { "epoch": 0.3, "learning_rate": 2.8226994601079783e-06, "loss": 0.0025, "step": 2513 }, { "epoch": 0.3, "learning_rate": 2.822627474505099e-06, "loss": 0.0012, "step": 2514 }, { "epoch": 0.3, "learning_rate": 2.82255548890222e-06, "loss": 0.9993, "step": 2515 }, { "epoch": 0.3, "learning_rate": 2.8224835032993402e-06, "loss": 0.006, "step": 2516 }, { "epoch": 0.3, "learning_rate": 2.822411517696461e-06, "loss": 1.6272, "step": 2517 }, { "epoch": 0.3, "learning_rate": 2.8223395320935814e-06, "loss": 0.0013, "step": 2518 }, { "epoch": 0.3, "learning_rate": 2.822267546490702e-06, "loss": 0.0038, "step": 2519 }, { "epoch": 0.3, "learning_rate": 2.8221955608878225e-06, "loss": 0.0025, "step": 2520 }, { "epoch": 0.3, "learning_rate": 2.8221235752849433e-06, "loss": 0.0014, "step": 2521 }, { "epoch": 0.3, "learning_rate": 2.8220515896820636e-06, "loss": 0.0124, "step": 2522 }, { "epoch": 0.3, "learning_rate": 2.821979604079184e-06, "loss": 0.001, "step": 2523 }, { "epoch": 0.3, "learning_rate": 2.8219076184763048e-06, "loss": 0.0036, "step": 2524 }, { "epoch": 0.3, "learning_rate": 2.821835632873425e-06, "loss": 0.0015, "step": 2525 }, { "epoch": 0.3, "learning_rate": 2.821763647270546e-06, "loss": 0.0011, "step": 2526 }, { "epoch": 0.3, "learning_rate": 2.8216916616676663e-06, "loss": 0.0016, "step": 2527 }, { "epoch": 0.3, "learning_rate": 2.821619676064787e-06, "loss": 0.0071, "step": 2528 }, { "epoch": 0.3, "learning_rate": 2.821547690461908e-06, "loss": 0.8237, "step": 2529 }, { "epoch": 0.3, "learning_rate": 2.8214757048590286e-06, "loss": 0.0011, "step": 2530 }, { "epoch": 0.3, "learning_rate": 2.821403719256149e-06, "loss": 0.726, "step": 2531 }, { "epoch": 0.3, "learning_rate": 2.8213317336532697e-06, "loss": 0.0015, "step": 2532 }, { "epoch": 0.3, "learning_rate": 2.82125974805039e-06, "loss": 0.0159, "step": 2533 }, { "epoch": 0.3, "learning_rate": 2.8211877624475104e-06, "loss": 0.001, "step": 2534 }, { "epoch": 0.3, "learning_rate": 2.8211157768446312e-06, "loss": 0.012, "step": 2535 }, { "epoch": 0.3, "learning_rate": 2.8210437912417516e-06, "loss": 0.016, "step": 2536 }, { "epoch": 0.3, "learning_rate": 2.8209718056388723e-06, "loss": 0.2225, "step": 2537 }, { "epoch": 0.3, "learning_rate": 2.8208998200359927e-06, "loss": 0.6373, "step": 2538 }, { "epoch": 0.3, "learning_rate": 2.8208278344331135e-06, "loss": 0.0031, "step": 2539 }, { "epoch": 0.3, "learning_rate": 2.820755848830234e-06, "loss": 0.0012, "step": 2540 }, { "epoch": 0.3, "learning_rate": 2.8206838632273546e-06, "loss": 0.0028, "step": 2541 }, { "epoch": 0.3, "learning_rate": 2.820611877624475e-06, "loss": 0.0011, "step": 2542 }, { "epoch": 0.3, "learning_rate": 2.8205398920215957e-06, "loss": 0.0017, "step": 2543 }, { "epoch": 0.3, "learning_rate": 2.8204679064187165e-06, "loss": 0.0015, "step": 2544 }, { "epoch": 0.3, "learning_rate": 2.820395920815837e-06, "loss": 0.0011, "step": 2545 }, { "epoch": 0.31, "learning_rate": 2.8203239352129577e-06, "loss": 0.4811, "step": 2546 }, { "epoch": 0.31, "learning_rate": 2.820251949610078e-06, "loss": 0.0016, "step": 2547 }, { "epoch": 0.31, "learning_rate": 2.820179964007199e-06, "loss": 0.0029, "step": 2548 }, { "epoch": 0.31, "learning_rate": 2.820107978404319e-06, "loss": 0.0023, "step": 2549 }, { "epoch": 0.31, "learning_rate": 2.82003599280144e-06, "loss": 0.0015, "step": 2550 }, { "epoch": 0.31, "learning_rate": 2.8199640071985603e-06, "loss": 0.0019, "step": 2551 }, { "epoch": 0.31, "learning_rate": 2.819892021595681e-06, "loss": 0.0011, "step": 2552 }, { "epoch": 0.31, "learning_rate": 2.8198200359928014e-06, "loss": 0.002, "step": 2553 }, { "epoch": 0.31, "learning_rate": 2.819748050389922e-06, "loss": 0.0104, "step": 2554 }, { "epoch": 0.31, "learning_rate": 2.8196760647870426e-06, "loss": 0.0149, "step": 2555 }, { "epoch": 0.31, "learning_rate": 2.819604079184163e-06, "loss": 0.0016, "step": 2556 }, { "epoch": 0.31, "learning_rate": 2.8195320935812837e-06, "loss": 0.0018, "step": 2557 }, { "epoch": 0.31, "learning_rate": 2.8194601079784045e-06, "loss": 0.0012, "step": 2558 }, { "epoch": 0.31, "learning_rate": 2.819388122375525e-06, "loss": 0.0026, "step": 2559 }, { "epoch": 0.31, "learning_rate": 2.8193161367726456e-06, "loss": 0.0012, "step": 2560 }, { "epoch": 0.31, "learning_rate": 2.8192441511697664e-06, "loss": 0.0011, "step": 2561 }, { "epoch": 0.31, "learning_rate": 2.8191721655668867e-06, "loss": 1.2925, "step": 2562 }, { "epoch": 0.31, "learning_rate": 2.8191001799640075e-06, "loss": 0.4399, "step": 2563 }, { "epoch": 0.31, "learning_rate": 2.819028194361128e-06, "loss": 0.0014, "step": 2564 }, { "epoch": 0.31, "learning_rate": 2.8189562087582486e-06, "loss": 0.0353, "step": 2565 }, { "epoch": 0.31, "learning_rate": 2.818884223155369e-06, "loss": 0.0014, "step": 2566 }, { "epoch": 0.31, "learning_rate": 2.8188122375524898e-06, "loss": 0.009, "step": 2567 }, { "epoch": 0.31, "learning_rate": 2.81874025194961e-06, "loss": 0.0011, "step": 2568 }, { "epoch": 0.31, "learning_rate": 2.8186682663467305e-06, "loss": 0.002, "step": 2569 }, { "epoch": 0.31, "learning_rate": 2.8185962807438513e-06, "loss": 0.0014, "step": 2570 }, { "epoch": 0.31, "learning_rate": 2.8185242951409716e-06, "loss": 0.0011, "step": 2571 }, { "epoch": 0.31, "learning_rate": 2.8184523095380924e-06, "loss": 0.0019, "step": 2572 }, { "epoch": 0.31, "learning_rate": 2.8183803239352128e-06, "loss": 0.0015, "step": 2573 }, { "epoch": 0.31, "learning_rate": 2.8183083383323335e-06, "loss": 0.0012, "step": 2574 }, { "epoch": 0.31, "learning_rate": 2.8182363527294543e-06, "loss": 1.3515, "step": 2575 }, { "epoch": 0.31, "learning_rate": 2.818164367126575e-06, "loss": 0.8094, "step": 2576 }, { "epoch": 0.31, "learning_rate": 2.8180923815236954e-06, "loss": 0.0009, "step": 2577 }, { "epoch": 0.31, "learning_rate": 2.8180203959208162e-06, "loss": 0.1881, "step": 2578 }, { "epoch": 0.31, "learning_rate": 2.8179484103179366e-06, "loss": 0.0012, "step": 2579 }, { "epoch": 0.31, "learning_rate": 2.817876424715057e-06, "loss": 0.002, "step": 2580 }, { "epoch": 0.31, "learning_rate": 2.8178044391121777e-06, "loss": 0.001, "step": 2581 }, { "epoch": 0.31, "learning_rate": 2.817732453509298e-06, "loss": 0.0346, "step": 2582 }, { "epoch": 0.31, "learning_rate": 2.817660467906419e-06, "loss": 0.0025, "step": 2583 }, { "epoch": 0.31, "learning_rate": 2.817588482303539e-06, "loss": 0.4482, "step": 2584 }, { "epoch": 0.31, "learning_rate": 2.81751649670066e-06, "loss": 0.7428, "step": 2585 }, { "epoch": 0.31, "learning_rate": 2.8174445110977803e-06, "loss": 0.1679, "step": 2586 }, { "epoch": 0.31, "learning_rate": 2.817372525494901e-06, "loss": 0.0958, "step": 2587 }, { "epoch": 0.31, "learning_rate": 2.8173005398920215e-06, "loss": 0.0022, "step": 2588 }, { "epoch": 0.31, "learning_rate": 2.8172285542891422e-06, "loss": 0.002, "step": 2589 }, { "epoch": 0.31, "learning_rate": 2.817156568686263e-06, "loss": 0.1888, "step": 2590 }, { "epoch": 0.31, "learning_rate": 2.8170845830833834e-06, "loss": 0.0015, "step": 2591 }, { "epoch": 0.31, "learning_rate": 2.817012597480504e-06, "loss": 0.0012, "step": 2592 }, { "epoch": 0.31, "learning_rate": 2.8169406118776245e-06, "loss": 0.0012, "step": 2593 }, { "epoch": 0.31, "learning_rate": 2.8168686262747453e-06, "loss": 0.0013, "step": 2594 }, { "epoch": 0.31, "learning_rate": 2.8167966406718656e-06, "loss": 0.0011, "step": 2595 }, { "epoch": 0.31, "learning_rate": 2.8167246550689864e-06, "loss": 0.0012, "step": 2596 }, { "epoch": 0.31, "learning_rate": 2.8166526694661068e-06, "loss": 0.1155, "step": 2597 }, { "epoch": 0.31, "learning_rate": 2.8165806838632276e-06, "loss": 0.2681, "step": 2598 }, { "epoch": 0.31, "learning_rate": 2.816508698260348e-06, "loss": 0.0029, "step": 2599 }, { "epoch": 0.31, "learning_rate": 2.8164367126574687e-06, "loss": 0.006, "step": 2600 }, { "epoch": 0.31, "learning_rate": 2.816364727054589e-06, "loss": 0.8494, "step": 2601 }, { "epoch": 0.31, "learning_rate": 2.81629274145171e-06, "loss": 0.0013, "step": 2602 }, { "epoch": 0.31, "learning_rate": 2.81622075584883e-06, "loss": 0.0024, "step": 2603 }, { "epoch": 0.31, "learning_rate": 2.8161487702459505e-06, "loss": 0.0016, "step": 2604 }, { "epoch": 0.31, "learning_rate": 2.8160767846430713e-06, "loss": 0.2237, "step": 2605 }, { "epoch": 0.31, "learning_rate": 2.816004799040192e-06, "loss": 0.8442, "step": 2606 }, { "epoch": 0.31, "learning_rate": 2.815932813437313e-06, "loss": 0.0012, "step": 2607 }, { "epoch": 0.31, "learning_rate": 2.8158608278344332e-06, "loss": 0.0012, "step": 2608 }, { "epoch": 0.31, "learning_rate": 2.815788842231554e-06, "loss": 0.0014, "step": 2609 }, { "epoch": 0.31, "learning_rate": 2.8157168566286744e-06, "loss": 0.001, "step": 2610 }, { "epoch": 0.31, "learning_rate": 2.815644871025795e-06, "loss": 0.5018, "step": 2611 }, { "epoch": 0.31, "learning_rate": 2.8155728854229155e-06, "loss": 0.3839, "step": 2612 }, { "epoch": 0.31, "learning_rate": 2.8155008998200363e-06, "loss": 0.0014, "step": 2613 }, { "epoch": 0.31, "learning_rate": 2.8154289142171566e-06, "loss": 0.0036, "step": 2614 }, { "epoch": 0.31, "learning_rate": 2.815356928614277e-06, "loss": 0.0012, "step": 2615 }, { "epoch": 0.31, "learning_rate": 2.8152849430113978e-06, "loss": 0.7521, "step": 2616 }, { "epoch": 0.31, "learning_rate": 2.815212957408518e-06, "loss": 0.002, "step": 2617 }, { "epoch": 0.31, "learning_rate": 2.815140971805639e-06, "loss": 0.0012, "step": 2618 }, { "epoch": 0.31, "learning_rate": 2.8150689862027592e-06, "loss": 0.0011, "step": 2619 }, { "epoch": 0.31, "learning_rate": 2.81499700059988e-06, "loss": 0.0014, "step": 2620 }, { "epoch": 0.31, "learning_rate": 2.814925014997001e-06, "loss": 0.0014, "step": 2621 }, { "epoch": 0.31, "learning_rate": 2.814853029394121e-06, "loss": 0.0011, "step": 2622 }, { "epoch": 0.31, "learning_rate": 2.814781043791242e-06, "loss": 0.001, "step": 2623 }, { "epoch": 0.31, "learning_rate": 2.8147090581883627e-06, "loss": 0.0014, "step": 2624 }, { "epoch": 0.31, "learning_rate": 2.814637072585483e-06, "loss": 0.0013, "step": 2625 }, { "epoch": 0.31, "learning_rate": 2.8145650869826034e-06, "loss": 0.0011, "step": 2626 }, { "epoch": 0.31, "learning_rate": 2.814493101379724e-06, "loss": 0.001, "step": 2627 }, { "epoch": 0.31, "learning_rate": 2.8144211157768446e-06, "loss": 0.002, "step": 2628 }, { "epoch": 0.32, "learning_rate": 2.8143491301739653e-06, "loss": 0.0713, "step": 2629 }, { "epoch": 0.32, "learning_rate": 2.8142771445710857e-06, "loss": 0.8043, "step": 2630 }, { "epoch": 0.32, "learning_rate": 2.8142051589682065e-06, "loss": 0.0017, "step": 2631 }, { "epoch": 0.32, "learning_rate": 2.814133173365327e-06, "loss": 0.0013, "step": 2632 }, { "epoch": 0.32, "learning_rate": 2.8140611877624476e-06, "loss": 0.0019, "step": 2633 }, { "epoch": 0.32, "learning_rate": 2.813989202159568e-06, "loss": 0.0016, "step": 2634 }, { "epoch": 0.32, "learning_rate": 2.8139172165566887e-06, "loss": 0.0012, "step": 2635 }, { "epoch": 0.32, "learning_rate": 2.813845230953809e-06, "loss": 0.0045, "step": 2636 }, { "epoch": 0.32, "learning_rate": 2.81377324535093e-06, "loss": 0.0017, "step": 2637 }, { "epoch": 0.32, "learning_rate": 2.8137012597480507e-06, "loss": 0.0019, "step": 2638 }, { "epoch": 0.32, "learning_rate": 2.813629274145171e-06, "loss": 0.8176, "step": 2639 }, { "epoch": 0.32, "learning_rate": 2.8135572885422918e-06, "loss": 0.015, "step": 2640 }, { "epoch": 0.32, "learning_rate": 2.813485302939412e-06, "loss": 0.4434, "step": 2641 }, { "epoch": 0.32, "learning_rate": 2.813413317336533e-06, "loss": 0.7601, "step": 2642 }, { "epoch": 0.32, "learning_rate": 2.8133413317336533e-06, "loss": 0.0011, "step": 2643 }, { "epoch": 0.32, "learning_rate": 2.813269346130774e-06, "loss": 0.808, "step": 2644 }, { "epoch": 0.32, "learning_rate": 2.8131973605278944e-06, "loss": 0.0017, "step": 2645 }, { "epoch": 0.32, "learning_rate": 2.813125374925015e-06, "loss": 0.0017, "step": 2646 }, { "epoch": 0.32, "learning_rate": 2.8130533893221355e-06, "loss": 0.002, "step": 2647 }, { "epoch": 0.32, "learning_rate": 2.8129814037192563e-06, "loss": 0.0037, "step": 2648 }, { "epoch": 0.32, "learning_rate": 2.8129094181163767e-06, "loss": 0.002, "step": 2649 }, { "epoch": 0.32, "learning_rate": 2.812837432513497e-06, "loss": 0.0029, "step": 2650 }, { "epoch": 0.32, "learning_rate": 2.812765446910618e-06, "loss": 0.0018, "step": 2651 }, { "epoch": 0.32, "learning_rate": 2.8126934613077386e-06, "loss": 0.7453, "step": 2652 }, { "epoch": 0.32, "learning_rate": 2.8126214757048594e-06, "loss": 0.0011, "step": 2653 }, { "epoch": 0.32, "learning_rate": 2.8125494901019797e-06, "loss": 0.0019, "step": 2654 }, { "epoch": 0.32, "learning_rate": 2.8124775044991005e-06, "loss": 0.0877, "step": 2655 }, { "epoch": 0.32, "learning_rate": 2.812405518896221e-06, "loss": 0.0021, "step": 2656 }, { "epoch": 0.32, "learning_rate": 2.8123335332933416e-06, "loss": 0.0027, "step": 2657 }, { "epoch": 0.32, "learning_rate": 2.812261547690462e-06, "loss": 0.3812, "step": 2658 }, { "epoch": 0.32, "learning_rate": 2.8121895620875828e-06, "loss": 0.0019, "step": 2659 }, { "epoch": 0.32, "learning_rate": 2.812117576484703e-06, "loss": 0.0054, "step": 2660 }, { "epoch": 0.32, "learning_rate": 2.8120455908818235e-06, "loss": 0.3732, "step": 2661 }, { "epoch": 0.32, "learning_rate": 2.8119736052789443e-06, "loss": 0.0011, "step": 2662 }, { "epoch": 0.32, "learning_rate": 2.8119016196760646e-06, "loss": 0.0023, "step": 2663 }, { "epoch": 0.32, "learning_rate": 2.8118296340731854e-06, "loss": 0.4819, "step": 2664 }, { "epoch": 0.32, "learning_rate": 2.8117576484703057e-06, "loss": 0.0027, "step": 2665 }, { "epoch": 0.32, "learning_rate": 2.8116856628674265e-06, "loss": 0.0019, "step": 2666 }, { "epoch": 0.32, "learning_rate": 2.8116136772645473e-06, "loss": 0.0014, "step": 2667 }, { "epoch": 0.32, "learning_rate": 2.8115416916616677e-06, "loss": 0.5527, "step": 2668 }, { "epoch": 0.32, "learning_rate": 2.8114697060587884e-06, "loss": 0.0015, "step": 2669 }, { "epoch": 0.32, "learning_rate": 2.8113977204559092e-06, "loss": 0.0112, "step": 2670 }, { "epoch": 0.32, "learning_rate": 2.8113257348530296e-06, "loss": 0.0028, "step": 2671 }, { "epoch": 0.32, "learning_rate": 2.81125374925015e-06, "loss": 0.0032, "step": 2672 }, { "epoch": 0.32, "learning_rate": 2.8111817636472707e-06, "loss": 0.6232, "step": 2673 }, { "epoch": 0.32, "learning_rate": 2.811109778044391e-06, "loss": 0.1128, "step": 2674 }, { "epoch": 0.32, "learning_rate": 2.811037792441512e-06, "loss": 0.002, "step": 2675 }, { "epoch": 0.32, "learning_rate": 2.810965806838632e-06, "loss": 0.0013, "step": 2676 }, { "epoch": 0.32, "learning_rate": 2.810893821235753e-06, "loss": 0.7306, "step": 2677 }, { "epoch": 0.32, "learning_rate": 2.8108218356328733e-06, "loss": 0.018, "step": 2678 }, { "epoch": 0.32, "learning_rate": 2.810749850029994e-06, "loss": 0.0024, "step": 2679 }, { "epoch": 0.32, "learning_rate": 2.8106778644271145e-06, "loss": 0.0048, "step": 2680 }, { "epoch": 0.32, "learning_rate": 2.8106058788242352e-06, "loss": 0.0015, "step": 2681 }, { "epoch": 0.32, "learning_rate": 2.8105338932213556e-06, "loss": 0.2192, "step": 2682 }, { "epoch": 0.32, "learning_rate": 2.8104619076184764e-06, "loss": 0.0032, "step": 2683 }, { "epoch": 0.32, "learning_rate": 2.810389922015597e-06, "loss": 0.0025, "step": 2684 }, { "epoch": 0.32, "learning_rate": 2.8103179364127175e-06, "loss": 0.6459, "step": 2685 }, { "epoch": 0.32, "learning_rate": 2.8102459508098383e-06, "loss": 0.0163, "step": 2686 }, { "epoch": 0.32, "learning_rate": 2.8101739652069586e-06, "loss": 0.0019, "step": 2687 }, { "epoch": 0.32, "learning_rate": 2.8101019796040794e-06, "loss": 0.006, "step": 2688 }, { "epoch": 0.32, "learning_rate": 2.8100299940011998e-06, "loss": 0.0015, "step": 2689 }, { "epoch": 0.32, "learning_rate": 2.8099580083983205e-06, "loss": 0.0038, "step": 2690 }, { "epoch": 0.32, "learning_rate": 2.809886022795441e-06, "loss": 0.0024, "step": 2691 }, { "epoch": 0.32, "learning_rate": 2.8098140371925617e-06, "loss": 0.0017, "step": 2692 }, { "epoch": 0.32, "learning_rate": 2.809742051589682e-06, "loss": 0.0016, "step": 2693 }, { "epoch": 0.32, "learning_rate": 2.809670065986803e-06, "loss": 0.0023, "step": 2694 }, { "epoch": 0.32, "learning_rate": 2.809598080383923e-06, "loss": 0.3296, "step": 2695 }, { "epoch": 0.32, "learning_rate": 2.8095260947810435e-06, "loss": 0.644, "step": 2696 }, { "epoch": 0.32, "learning_rate": 2.8094541091781643e-06, "loss": 0.0018, "step": 2697 }, { "epoch": 0.32, "learning_rate": 2.809382123575285e-06, "loss": 0.0013, "step": 2698 }, { "epoch": 0.32, "learning_rate": 2.809310137972406e-06, "loss": 0.0013, "step": 2699 }, { "epoch": 0.32, "learning_rate": 2.8092381523695262e-06, "loss": 0.7273, "step": 2700 }, { "epoch": 0.32, "learning_rate": 2.809166166766647e-06, "loss": 0.0013, "step": 2701 }, { "epoch": 0.32, "learning_rate": 2.8090941811637674e-06, "loss": 0.0042, "step": 2702 }, { "epoch": 0.32, "learning_rate": 2.809022195560888e-06, "loss": 0.0064, "step": 2703 }, { "epoch": 0.32, "learning_rate": 2.8089502099580085e-06, "loss": 0.2972, "step": 2704 }, { "epoch": 0.32, "learning_rate": 2.8088782243551293e-06, "loss": 0.6786, "step": 2705 }, { "epoch": 0.32, "learning_rate": 2.8088062387522496e-06, "loss": 0.004, "step": 2706 }, { "epoch": 0.32, "learning_rate": 2.80873425314937e-06, "loss": 0.0017, "step": 2707 }, { "epoch": 0.32, "learning_rate": 2.8086622675464908e-06, "loss": 0.0014, "step": 2708 }, { "epoch": 0.32, "learning_rate": 2.808590281943611e-06, "loss": 0.002, "step": 2709 }, { "epoch": 0.32, "learning_rate": 2.808518296340732e-06, "loss": 0.097, "step": 2710 }, { "epoch": 0.32, "learning_rate": 2.8084463107378522e-06, "loss": 0.0011, "step": 2711 }, { "epoch": 0.32, "learning_rate": 2.808374325134973e-06, "loss": 0.0017, "step": 2712 }, { "epoch": 0.33, "learning_rate": 2.808302339532094e-06, "loss": 0.0015, "step": 2713 }, { "epoch": 0.33, "learning_rate": 2.808230353929214e-06, "loss": 0.002, "step": 2714 }, { "epoch": 0.33, "learning_rate": 2.808158368326335e-06, "loss": 0.0032, "step": 2715 }, { "epoch": 0.33, "learning_rate": 2.8080863827234557e-06, "loss": 0.0025, "step": 2716 }, { "epoch": 0.33, "learning_rate": 2.808014397120576e-06, "loss": 0.0178, "step": 2717 }, { "epoch": 0.33, "learning_rate": 2.8079424115176964e-06, "loss": 0.0054, "step": 2718 }, { "epoch": 0.33, "learning_rate": 2.807870425914817e-06, "loss": 0.004, "step": 2719 }, { "epoch": 0.33, "learning_rate": 2.8077984403119376e-06, "loss": 0.0014, "step": 2720 }, { "epoch": 0.33, "learning_rate": 2.8077264547090583e-06, "loss": 0.0014, "step": 2721 }, { "epoch": 0.33, "learning_rate": 2.8076544691061787e-06, "loss": 0.0011, "step": 2722 }, { "epoch": 0.33, "learning_rate": 2.8075824835032995e-06, "loss": 0.0014, "step": 2723 }, { "epoch": 0.33, "learning_rate": 2.80751049790042e-06, "loss": 0.0013, "step": 2724 }, { "epoch": 0.33, "learning_rate": 2.8074385122975406e-06, "loss": 0.1677, "step": 2725 }, { "epoch": 0.33, "learning_rate": 2.807366526694661e-06, "loss": 0.0016, "step": 2726 }, { "epoch": 0.33, "learning_rate": 2.8072945410917817e-06, "loss": 0.0011, "step": 2727 }, { "epoch": 0.33, "learning_rate": 2.807222555488902e-06, "loss": 0.0053, "step": 2728 }, { "epoch": 0.33, "learning_rate": 2.807150569886023e-06, "loss": 0.0016, "step": 2729 }, { "epoch": 0.33, "learning_rate": 2.8070785842831436e-06, "loss": 0.5468, "step": 2730 }, { "epoch": 0.33, "learning_rate": 2.807006598680264e-06, "loss": 0.0024, "step": 2731 }, { "epoch": 0.33, "learning_rate": 2.8069346130773848e-06, "loss": 0.0031, "step": 2732 }, { "epoch": 0.33, "learning_rate": 2.806862627474505e-06, "loss": 0.0012, "step": 2733 }, { "epoch": 0.33, "learning_rate": 2.806790641871626e-06, "loss": 0.0031, "step": 2734 }, { "epoch": 0.33, "learning_rate": 2.8067186562687463e-06, "loss": 0.0043, "step": 2735 }, { "epoch": 0.33, "learning_rate": 2.806646670665867e-06, "loss": 0.5116, "step": 2736 }, { "epoch": 0.33, "learning_rate": 2.8065746850629874e-06, "loss": 0.0018, "step": 2737 }, { "epoch": 0.33, "learning_rate": 2.806502699460108e-06, "loss": 0.0014, "step": 2738 }, { "epoch": 0.33, "learning_rate": 2.8064307138572285e-06, "loss": 0.0657, "step": 2739 }, { "epoch": 0.33, "learning_rate": 2.8063587282543493e-06, "loss": 0.0015, "step": 2740 }, { "epoch": 0.33, "learning_rate": 2.8062867426514697e-06, "loss": 0.0013, "step": 2741 }, { "epoch": 0.33, "learning_rate": 2.80621475704859e-06, "loss": 0.0013, "step": 2742 }, { "epoch": 0.33, "learning_rate": 2.806142771445711e-06, "loss": 0.6491, "step": 2743 }, { "epoch": 0.33, "learning_rate": 2.8060707858428316e-06, "loss": 0.0098, "step": 2744 }, { "epoch": 0.33, "learning_rate": 2.805998800239952e-06, "loss": 0.0024, "step": 2745 }, { "epoch": 0.33, "learning_rate": 2.8059268146370727e-06, "loss": 0.0018, "step": 2746 }, { "epoch": 0.33, "learning_rate": 2.8058548290341935e-06, "loss": 0.0012, "step": 2747 }, { "epoch": 0.33, "learning_rate": 2.805782843431314e-06, "loss": 0.1973, "step": 2748 }, { "epoch": 0.33, "learning_rate": 2.8057108578284346e-06, "loss": 0.0034, "step": 2749 }, { "epoch": 0.33, "learning_rate": 2.805638872225555e-06, "loss": 0.0027, "step": 2750 }, { "epoch": 0.33, "learning_rate": 2.8055668866226758e-06, "loss": 0.0025, "step": 2751 }, { "epoch": 0.33, "learning_rate": 2.805494901019796e-06, "loss": 0.4528, "step": 2752 }, { "epoch": 0.33, "learning_rate": 2.8054229154169165e-06, "loss": 0.001, "step": 2753 }, { "epoch": 0.33, "learning_rate": 2.8053509298140372e-06, "loss": 0.1867, "step": 2754 }, { "epoch": 0.33, "learning_rate": 2.8052789442111576e-06, "loss": 0.001, "step": 2755 }, { "epoch": 0.33, "learning_rate": 2.8052069586082784e-06, "loss": 0.3565, "step": 2756 }, { "epoch": 0.33, "learning_rate": 2.8051349730053987e-06, "loss": 0.0029, "step": 2757 }, { "epoch": 0.33, "learning_rate": 2.8050629874025195e-06, "loss": 0.906, "step": 2758 }, { "epoch": 0.33, "learning_rate": 2.80499100179964e-06, "loss": 0.0322, "step": 2759 }, { "epoch": 0.33, "learning_rate": 2.8049190161967606e-06, "loss": 0.0015, "step": 2760 }, { "epoch": 0.33, "learning_rate": 2.8048470305938814e-06, "loss": 0.0062, "step": 2761 }, { "epoch": 0.33, "learning_rate": 2.804775044991002e-06, "loss": 0.2427, "step": 2762 }, { "epoch": 0.33, "learning_rate": 2.8047030593881226e-06, "loss": 0.0069, "step": 2763 }, { "epoch": 0.33, "learning_rate": 2.804631073785243e-06, "loss": 0.0009, "step": 2764 }, { "epoch": 0.33, "learning_rate": 2.8045590881823637e-06, "loss": 0.0092, "step": 2765 }, { "epoch": 0.33, "learning_rate": 2.804487102579484e-06, "loss": 0.0008, "step": 2766 }, { "epoch": 0.33, "learning_rate": 2.804415116976605e-06, "loss": 0.0027, "step": 2767 }, { "epoch": 0.33, "learning_rate": 2.804343131373725e-06, "loss": 0.0334, "step": 2768 }, { "epoch": 0.33, "learning_rate": 2.804271145770846e-06, "loss": 0.001, "step": 2769 }, { "epoch": 0.33, "learning_rate": 2.8041991601679663e-06, "loss": 0.0012, "step": 2770 }, { "epoch": 0.33, "learning_rate": 2.804127174565087e-06, "loss": 0.0019, "step": 2771 }, { "epoch": 0.33, "learning_rate": 2.8040551889622074e-06, "loss": 0.0009, "step": 2772 }, { "epoch": 0.33, "learning_rate": 2.8039832033593282e-06, "loss": 0.003, "step": 2773 }, { "epoch": 0.33, "learning_rate": 2.8039112177564486e-06, "loss": 0.0022, "step": 2774 }, { "epoch": 0.33, "learning_rate": 2.8038392321535694e-06, "loss": 0.001, "step": 2775 }, { "epoch": 0.33, "learning_rate": 2.80376724655069e-06, "loss": 0.0016, "step": 2776 }, { "epoch": 0.33, "learning_rate": 2.8036952609478105e-06, "loss": 0.003, "step": 2777 }, { "epoch": 0.33, "learning_rate": 2.8036232753449313e-06, "loss": 0.001, "step": 2778 }, { "epoch": 0.33, "learning_rate": 2.8035512897420516e-06, "loss": 0.0086, "step": 2779 }, { "epoch": 0.33, "learning_rate": 2.8034793041391724e-06, "loss": 0.0764, "step": 2780 }, { "epoch": 0.33, "learning_rate": 2.8034073185362928e-06, "loss": 0.0015, "step": 2781 }, { "epoch": 0.33, "learning_rate": 2.8033353329334135e-06, "loss": 0.0029, "step": 2782 }, { "epoch": 0.33, "learning_rate": 2.803263347330534e-06, "loss": 0.9688, "step": 2783 }, { "epoch": 0.33, "learning_rate": 2.8031913617276547e-06, "loss": 0.0077, "step": 2784 }, { "epoch": 0.33, "learning_rate": 2.803119376124775e-06, "loss": 1.0887, "step": 2785 }, { "epoch": 0.33, "learning_rate": 2.803047390521896e-06, "loss": 1.4541, "step": 2786 }, { "epoch": 0.33, "learning_rate": 2.802975404919016e-06, "loss": 0.001, "step": 2787 }, { "epoch": 0.33, "learning_rate": 2.8029034193161365e-06, "loss": 0.0016, "step": 2788 }, { "epoch": 0.33, "learning_rate": 2.8028314337132573e-06, "loss": 0.0012, "step": 2789 }, { "epoch": 0.33, "learning_rate": 2.802759448110378e-06, "loss": 0.0015, "step": 2790 }, { "epoch": 0.33, "learning_rate": 2.8026874625074984e-06, "loss": 0.1204, "step": 2791 }, { "epoch": 0.33, "learning_rate": 2.802615476904619e-06, "loss": 0.4212, "step": 2792 }, { "epoch": 0.33, "learning_rate": 2.80254349130174e-06, "loss": 0.0014, "step": 2793 }, { "epoch": 0.33, "learning_rate": 2.8024715056988603e-06, "loss": 0.0009, "step": 2794 }, { "epoch": 0.33, "learning_rate": 2.802399520095981e-06, "loss": 0.0012, "step": 2795 }, { "epoch": 0.34, "learning_rate": 2.8023275344931015e-06, "loss": 0.7126, "step": 2796 }, { "epoch": 0.34, "learning_rate": 2.8022555488902223e-06, "loss": 0.0012, "step": 2797 }, { "epoch": 0.34, "learning_rate": 2.8021835632873426e-06, "loss": 0.7829, "step": 2798 }, { "epoch": 0.34, "learning_rate": 2.802111577684463e-06, "loss": 0.001, "step": 2799 }, { "epoch": 0.34, "learning_rate": 2.8020395920815837e-06, "loss": 0.0008, "step": 2800 }, { "epoch": 0.34, "learning_rate": 2.801967606478704e-06, "loss": 0.001, "step": 2801 }, { "epoch": 0.34, "learning_rate": 2.801895620875825e-06, "loss": 0.0011, "step": 2802 }, { "epoch": 0.34, "learning_rate": 2.8018236352729452e-06, "loss": 0.0011, "step": 2803 }, { "epoch": 0.34, "learning_rate": 2.801751649670066e-06, "loss": 0.0011, "step": 2804 }, { "epoch": 0.34, "learning_rate": 2.8016796640671864e-06, "loss": 0.0016, "step": 2805 }, { "epoch": 0.34, "learning_rate": 2.801607678464307e-06, "loss": 0.0013, "step": 2806 }, { "epoch": 0.34, "learning_rate": 2.801535692861428e-06, "loss": 0.0011, "step": 2807 }, { "epoch": 0.34, "learning_rate": 2.8014637072585487e-06, "loss": 0.2373, "step": 2808 }, { "epoch": 0.34, "learning_rate": 2.801391721655669e-06, "loss": 0.0357, "step": 2809 }, { "epoch": 0.34, "learning_rate": 2.8013197360527894e-06, "loss": 0.0015, "step": 2810 }, { "epoch": 0.34, "learning_rate": 2.80124775044991e-06, "loss": 0.0014, "step": 2811 }, { "epoch": 0.34, "learning_rate": 2.8011757648470305e-06, "loss": 0.0024, "step": 2812 }, { "epoch": 0.34, "learning_rate": 2.8011037792441513e-06, "loss": 0.0013, "step": 2813 }, { "epoch": 0.34, "learning_rate": 2.8010317936412717e-06, "loss": 0.0012, "step": 2814 }, { "epoch": 0.34, "learning_rate": 2.8009598080383925e-06, "loss": 0.0198, "step": 2815 }, { "epoch": 0.34, "learning_rate": 2.800887822435513e-06, "loss": 0.0012, "step": 2816 }, { "epoch": 0.34, "learning_rate": 2.8008158368326336e-06, "loss": 0.0011, "step": 2817 }, { "epoch": 0.34, "learning_rate": 2.800743851229754e-06, "loss": 0.0013, "step": 2818 }, { "epoch": 0.34, "learning_rate": 2.8006718656268747e-06, "loss": 0.0665, "step": 2819 }, { "epoch": 0.34, "learning_rate": 2.800599880023995e-06, "loss": 0.0011, "step": 2820 }, { "epoch": 0.34, "learning_rate": 2.800527894421116e-06, "loss": 0.0011, "step": 2821 }, { "epoch": 0.34, "learning_rate": 2.8004559088182366e-06, "loss": 0.0011, "step": 2822 }, { "epoch": 0.34, "learning_rate": 2.800383923215357e-06, "loss": 0.0013, "step": 2823 }, { "epoch": 0.34, "learning_rate": 2.8003119376124778e-06, "loss": 0.001, "step": 2824 }, { "epoch": 0.34, "learning_rate": 2.800239952009598e-06, "loss": 0.9921, "step": 2825 }, { "epoch": 0.34, "learning_rate": 2.800167966406719e-06, "loss": 0.5293, "step": 2826 }, { "epoch": 0.34, "learning_rate": 2.8000959808038393e-06, "loss": 0.0014, "step": 2827 }, { "epoch": 0.34, "learning_rate": 2.80002399520096e-06, "loss": 0.0014, "step": 2828 }, { "epoch": 0.34, "learning_rate": 2.7999520095980804e-06, "loss": 0.0008, "step": 2829 }, { "epoch": 0.34, "learning_rate": 2.799880023995201e-06, "loss": 0.0012, "step": 2830 }, { "epoch": 0.34, "learning_rate": 2.7998080383923215e-06, "loss": 0.0013, "step": 2831 }, { "epoch": 0.34, "learning_rate": 2.7997360527894423e-06, "loss": 0.0012, "step": 2832 }, { "epoch": 0.34, "learning_rate": 2.7996640671865627e-06, "loss": 0.0009, "step": 2833 }, { "epoch": 0.34, "learning_rate": 2.799592081583683e-06, "loss": 0.0039, "step": 2834 }, { "epoch": 0.34, "learning_rate": 2.799520095980804e-06, "loss": 0.0037, "step": 2835 }, { "epoch": 0.34, "learning_rate": 2.7994481103779246e-06, "loss": 0.0031, "step": 2836 }, { "epoch": 0.34, "learning_rate": 2.799376124775045e-06, "loss": 0.001, "step": 2837 }, { "epoch": 0.34, "learning_rate": 2.7993041391721657e-06, "loss": 0.001, "step": 2838 }, { "epoch": 0.34, "learning_rate": 2.7992321535692865e-06, "loss": 0.0009, "step": 2839 }, { "epoch": 0.34, "learning_rate": 2.799160167966407e-06, "loss": 0.0013, "step": 2840 }, { "epoch": 0.34, "learning_rate": 2.7990881823635276e-06, "loss": 0.0026, "step": 2841 }, { "epoch": 0.34, "learning_rate": 2.799016196760648e-06, "loss": 0.4973, "step": 2842 }, { "epoch": 0.34, "learning_rate": 2.7989442111577688e-06, "loss": 0.0199, "step": 2843 }, { "epoch": 0.34, "learning_rate": 2.798872225554889e-06, "loss": 0.0012, "step": 2844 }, { "epoch": 0.34, "learning_rate": 2.7988002399520095e-06, "loss": 0.7046, "step": 2845 }, { "epoch": 0.34, "learning_rate": 2.7987282543491302e-06, "loss": 0.0071, "step": 2846 }, { "epoch": 0.34, "learning_rate": 2.7986562687462506e-06, "loss": 0.0009, "step": 2847 }, { "epoch": 0.34, "learning_rate": 2.7985842831433714e-06, "loss": 0.0013, "step": 2848 }, { "epoch": 0.34, "learning_rate": 2.7985122975404917e-06, "loss": 0.0013, "step": 2849 }, { "epoch": 0.34, "learning_rate": 2.7984403119376125e-06, "loss": 0.0009, "step": 2850 }, { "epoch": 0.34, "learning_rate": 2.798368326334733e-06, "loss": 0.009, "step": 2851 }, { "epoch": 0.34, "learning_rate": 2.7982963407318536e-06, "loss": 0.6104, "step": 2852 }, { "epoch": 0.34, "learning_rate": 2.7982243551289744e-06, "loss": 0.8105, "step": 2853 }, { "epoch": 0.34, "learning_rate": 2.798152369526095e-06, "loss": 0.2619, "step": 2854 }, { "epoch": 0.34, "learning_rate": 2.7980803839232156e-06, "loss": 0.8628, "step": 2855 }, { "epoch": 0.34, "learning_rate": 2.7980083983203363e-06, "loss": 0.0235, "step": 2856 }, { "epoch": 0.34, "learning_rate": 2.7979364127174567e-06, "loss": 0.0024, "step": 2857 }, { "epoch": 0.34, "learning_rate": 2.797864427114577e-06, "loss": 0.3244, "step": 2858 }, { "epoch": 0.34, "learning_rate": 2.797792441511698e-06, "loss": 1.6432, "step": 2859 }, { "epoch": 0.34, "learning_rate": 2.797720455908818e-06, "loss": 0.0059, "step": 2860 }, { "epoch": 0.34, "learning_rate": 2.797648470305939e-06, "loss": 0.0046, "step": 2861 }, { "epoch": 0.34, "learning_rate": 2.7975764847030593e-06, "loss": 0.0041, "step": 2862 }, { "epoch": 0.34, "learning_rate": 2.79750449910018e-06, "loss": 0.0013, "step": 2863 }, { "epoch": 0.34, "learning_rate": 2.7974325134973004e-06, "loss": 0.0014, "step": 2864 }, { "epoch": 0.34, "learning_rate": 2.7973605278944212e-06, "loss": 0.8525, "step": 2865 }, { "epoch": 0.34, "learning_rate": 2.7972885422915416e-06, "loss": 0.002, "step": 2866 }, { "epoch": 0.34, "learning_rate": 2.7972165566886624e-06, "loss": 0.0019, "step": 2867 }, { "epoch": 0.34, "learning_rate": 2.7971445710857827e-06, "loss": 0.001, "step": 2868 }, { "epoch": 0.34, "learning_rate": 2.7970725854829035e-06, "loss": 0.0016, "step": 2869 }, { "epoch": 0.34, "learning_rate": 2.7970005998800243e-06, "loss": 0.046, "step": 2870 }, { "epoch": 0.34, "learning_rate": 2.7969286142771446e-06, "loss": 0.0013, "step": 2871 }, { "epoch": 0.34, "learning_rate": 2.7968566286742654e-06, "loss": 0.008, "step": 2872 }, { "epoch": 0.34, "learning_rate": 2.7967846430713858e-06, "loss": 0.0012, "step": 2873 }, { "epoch": 0.34, "learning_rate": 2.7967126574685065e-06, "loss": 0.0011, "step": 2874 }, { "epoch": 0.34, "learning_rate": 2.796640671865627e-06, "loss": 0.0011, "step": 2875 }, { "epoch": 0.34, "learning_rate": 2.7965686862627477e-06, "loss": 0.0011, "step": 2876 }, { "epoch": 0.34, "learning_rate": 2.796496700659868e-06, "loss": 0.0016, "step": 2877 }, { "epoch": 0.34, "learning_rate": 2.796424715056989e-06, "loss": 0.0032, "step": 2878 }, { "epoch": 0.34, "learning_rate": 2.796352729454109e-06, "loss": 0.3141, "step": 2879 }, { "epoch": 0.35, "learning_rate": 2.7962807438512295e-06, "loss": 0.0026, "step": 2880 }, { "epoch": 0.35, "learning_rate": 2.7962087582483503e-06, "loss": 0.0033, "step": 2881 }, { "epoch": 0.35, "learning_rate": 2.7961367726454706e-06, "loss": 0.0009, "step": 2882 }, { "epoch": 0.35, "learning_rate": 2.7960647870425914e-06, "loss": 0.5223, "step": 2883 }, { "epoch": 0.35, "learning_rate": 2.795992801439712e-06, "loss": 0.0014, "step": 2884 }, { "epoch": 0.35, "learning_rate": 2.795920815836833e-06, "loss": 0.6278, "step": 2885 }, { "epoch": 0.35, "learning_rate": 2.7958488302339533e-06, "loss": 0.0016, "step": 2886 }, { "epoch": 0.35, "learning_rate": 2.795776844631074e-06, "loss": 0.0009, "step": 2887 }, { "epoch": 0.35, "learning_rate": 2.7957048590281945e-06, "loss": 0.4576, "step": 2888 }, { "epoch": 0.35, "learning_rate": 2.7956328734253152e-06, "loss": 0.0009, "step": 2889 }, { "epoch": 0.35, "learning_rate": 2.7955608878224356e-06, "loss": 0.5443, "step": 2890 }, { "epoch": 0.35, "learning_rate": 2.795488902219556e-06, "loss": 0.1541, "step": 2891 }, { "epoch": 0.35, "learning_rate": 2.7954169166166767e-06, "loss": 0.0022, "step": 2892 }, { "epoch": 0.35, "learning_rate": 2.795344931013797e-06, "loss": 0.0076, "step": 2893 }, { "epoch": 0.35, "learning_rate": 2.795272945410918e-06, "loss": 0.0016, "step": 2894 }, { "epoch": 0.35, "learning_rate": 2.7952009598080382e-06, "loss": 0.3468, "step": 2895 }, { "epoch": 0.35, "learning_rate": 2.795128974205159e-06, "loss": 0.0397, "step": 2896 }, { "epoch": 0.35, "learning_rate": 2.7950569886022794e-06, "loss": 0.7538, "step": 2897 }, { "epoch": 0.35, "learning_rate": 2.7949850029994e-06, "loss": 0.6892, "step": 2898 }, { "epoch": 0.35, "learning_rate": 2.794913017396521e-06, "loss": 0.0012, "step": 2899 }, { "epoch": 0.35, "learning_rate": 2.7948410317936413e-06, "loss": 0.6035, "step": 2900 }, { "epoch": 0.35, "learning_rate": 2.794769046190762e-06, "loss": 0.8192, "step": 2901 }, { "epoch": 0.35, "learning_rate": 2.794697060587883e-06, "loss": 0.002, "step": 2902 }, { "epoch": 0.35, "learning_rate": 2.794625074985003e-06, "loss": 0.0622, "step": 2903 }, { "epoch": 0.35, "learning_rate": 2.7945530893821235e-06, "loss": 0.0019, "step": 2904 }, { "epoch": 0.35, "learning_rate": 2.7944811037792443e-06, "loss": 0.0019, "step": 2905 }, { "epoch": 0.35, "learning_rate": 2.7944091181763647e-06, "loss": 0.0926, "step": 2906 }, { "epoch": 0.35, "learning_rate": 2.7943371325734854e-06, "loss": 0.0011, "step": 2907 }, { "epoch": 0.35, "learning_rate": 2.794265146970606e-06, "loss": 0.0012, "step": 2908 }, { "epoch": 0.35, "learning_rate": 2.7941931613677266e-06, "loss": 0.0016, "step": 2909 }, { "epoch": 0.35, "learning_rate": 2.794121175764847e-06, "loss": 0.3864, "step": 2910 }, { "epoch": 0.35, "learning_rate": 2.7940491901619677e-06, "loss": 0.0016, "step": 2911 }, { "epoch": 0.35, "learning_rate": 2.793977204559088e-06, "loss": 0.0015, "step": 2912 }, { "epoch": 0.35, "learning_rate": 2.793905218956209e-06, "loss": 0.155, "step": 2913 }, { "epoch": 0.35, "learning_rate": 2.793833233353329e-06, "loss": 0.0013, "step": 2914 }, { "epoch": 0.35, "learning_rate": 2.79376124775045e-06, "loss": 0.0011, "step": 2915 }, { "epoch": 0.35, "learning_rate": 2.7936892621475708e-06, "loss": 0.0033, "step": 2916 }, { "epoch": 0.35, "learning_rate": 2.793617276544691e-06, "loss": 0.0014, "step": 2917 }, { "epoch": 0.35, "learning_rate": 2.793545290941812e-06, "loss": 0.0202, "step": 2918 }, { "epoch": 0.35, "learning_rate": 2.7934733053389323e-06, "loss": 0.0059, "step": 2919 }, { "epoch": 0.35, "learning_rate": 2.793401319736053e-06, "loss": 0.0143, "step": 2920 }, { "epoch": 0.35, "learning_rate": 2.7933293341331734e-06, "loss": 0.0019, "step": 2921 }, { "epoch": 0.35, "learning_rate": 2.793257348530294e-06, "loss": 0.0027, "step": 2922 }, { "epoch": 0.35, "learning_rate": 2.7931853629274145e-06, "loss": 0.0014, "step": 2923 }, { "epoch": 0.35, "learning_rate": 2.7931133773245353e-06, "loss": 0.0012, "step": 2924 }, { "epoch": 0.35, "learning_rate": 2.7930413917216557e-06, "loss": 0.0017, "step": 2925 }, { "epoch": 0.35, "learning_rate": 2.792969406118776e-06, "loss": 0.017, "step": 2926 }, { "epoch": 0.35, "learning_rate": 2.7928974205158968e-06, "loss": 0.0037, "step": 2927 }, { "epoch": 0.35, "learning_rate": 2.792825434913017e-06, "loss": 0.0016, "step": 2928 }, { "epoch": 0.35, "learning_rate": 2.792753449310138e-06, "loss": 0.3989, "step": 2929 }, { "epoch": 0.35, "learning_rate": 2.7926814637072587e-06, "loss": 0.1588, "step": 2930 }, { "epoch": 0.35, "learning_rate": 2.7926094781043795e-06, "loss": 0.0013, "step": 2931 }, { "epoch": 0.35, "learning_rate": 2.7925374925015e-06, "loss": 0.4444, "step": 2932 }, { "epoch": 0.35, "learning_rate": 2.7924655068986206e-06, "loss": 0.0016, "step": 2933 }, { "epoch": 0.35, "learning_rate": 2.792393521295741e-06, "loss": 0.0014, "step": 2934 }, { "epoch": 0.35, "learning_rate": 2.7923215356928617e-06, "loss": 0.0013, "step": 2935 }, { "epoch": 0.35, "learning_rate": 2.792249550089982e-06, "loss": 0.0106, "step": 2936 }, { "epoch": 0.35, "learning_rate": 2.7921775644871025e-06, "loss": 0.1568, "step": 2937 }, { "epoch": 0.35, "learning_rate": 2.7921055788842232e-06, "loss": 0.0287, "step": 2938 }, { "epoch": 0.35, "learning_rate": 2.7920335932813436e-06, "loss": 0.0119, "step": 2939 }, { "epoch": 0.35, "learning_rate": 2.7919616076784644e-06, "loss": 0.0011, "step": 2940 }, { "epoch": 0.35, "learning_rate": 2.7918896220755847e-06, "loss": 0.0115, "step": 2941 }, { "epoch": 0.35, "learning_rate": 2.7918176364727055e-06, "loss": 0.0012, "step": 2942 }, { "epoch": 0.35, "learning_rate": 2.791745650869826e-06, "loss": 0.1616, "step": 2943 }, { "epoch": 0.35, "learning_rate": 2.7916736652669466e-06, "loss": 0.0036, "step": 2944 }, { "epoch": 0.35, "learning_rate": 2.7916016796640674e-06, "loss": 0.0021, "step": 2945 }, { "epoch": 0.35, "learning_rate": 2.7915296940611878e-06, "loss": 0.0018, "step": 2946 }, { "epoch": 0.35, "learning_rate": 2.7914577084583085e-06, "loss": 0.0263, "step": 2947 }, { "epoch": 0.35, "learning_rate": 2.7913857228554293e-06, "loss": 0.0024, "step": 2948 }, { "epoch": 0.35, "learning_rate": 2.7913137372525497e-06, "loss": 0.0014, "step": 2949 }, { "epoch": 0.35, "learning_rate": 2.79124175164967e-06, "loss": 0.001, "step": 2950 }, { "epoch": 0.35, "learning_rate": 2.791169766046791e-06, "loss": 0.1153, "step": 2951 }, { "epoch": 0.35, "learning_rate": 2.791097780443911e-06, "loss": 0.0018, "step": 2952 }, { "epoch": 0.35, "learning_rate": 2.791025794841032e-06, "loss": 0.0013, "step": 2953 }, { "epoch": 0.35, "learning_rate": 2.7909538092381523e-06, "loss": 0.1196, "step": 2954 }, { "epoch": 0.35, "learning_rate": 2.790881823635273e-06, "loss": 0.4827, "step": 2955 }, { "epoch": 0.35, "learning_rate": 2.7908098380323934e-06, "loss": 0.0015, "step": 2956 }, { "epoch": 0.35, "learning_rate": 2.7907378524295142e-06, "loss": 0.0014, "step": 2957 }, { "epoch": 0.35, "learning_rate": 2.7906658668266346e-06, "loss": 0.0011, "step": 2958 }, { "epoch": 0.35, "learning_rate": 2.7905938812237553e-06, "loss": 0.9742, "step": 2959 }, { "epoch": 0.35, "learning_rate": 2.7905218956208757e-06, "loss": 0.0013, "step": 2960 }, { "epoch": 0.35, "learning_rate": 2.7904499100179965e-06, "loss": 0.8925, "step": 2961 }, { "epoch": 0.35, "learning_rate": 2.7903779244151173e-06, "loss": 0.0036, "step": 2962 }, { "epoch": 0.36, "learning_rate": 2.7903059388122376e-06, "loss": 0.0119, "step": 2963 }, { "epoch": 0.36, "learning_rate": 2.7902339532093584e-06, "loss": 0.003, "step": 2964 }, { "epoch": 0.36, "learning_rate": 2.7901619676064787e-06, "loss": 0.6833, "step": 2965 }, { "epoch": 0.36, "learning_rate": 2.7900899820035995e-06, "loss": 0.0017, "step": 2966 }, { "epoch": 0.36, "learning_rate": 2.79001799640072e-06, "loss": 0.001, "step": 2967 }, { "epoch": 0.36, "learning_rate": 2.7899460107978407e-06, "loss": 0.6181, "step": 2968 }, { "epoch": 0.36, "learning_rate": 2.789874025194961e-06, "loss": 0.4873, "step": 2969 }, { "epoch": 0.36, "learning_rate": 2.789802039592082e-06, "loss": 0.0012, "step": 2970 }, { "epoch": 0.36, "learning_rate": 2.789730053989202e-06, "loss": 0.6337, "step": 2971 }, { "epoch": 0.36, "learning_rate": 2.7896580683863225e-06, "loss": 0.0011, "step": 2972 }, { "epoch": 0.36, "learning_rate": 2.7895860827834433e-06, "loss": 0.0036, "step": 2973 }, { "epoch": 0.36, "learning_rate": 2.7895140971805636e-06, "loss": 0.0019, "step": 2974 }, { "epoch": 0.36, "learning_rate": 2.7894421115776844e-06, "loss": 0.0016, "step": 2975 }, { "epoch": 0.36, "learning_rate": 2.789370125974805e-06, "loss": 0.4003, "step": 2976 }, { "epoch": 0.36, "learning_rate": 2.789298140371926e-06, "loss": 0.0013, "step": 2977 }, { "epoch": 0.36, "learning_rate": 2.7892261547690463e-06, "loss": 0.0013, "step": 2978 }, { "epoch": 0.36, "learning_rate": 2.789154169166167e-06, "loss": 0.0049, "step": 2979 }, { "epoch": 0.36, "learning_rate": 2.7890821835632875e-06, "loss": 0.002, "step": 2980 }, { "epoch": 0.36, "learning_rate": 2.7890101979604082e-06, "loss": 0.3395, "step": 2981 }, { "epoch": 0.36, "learning_rate": 2.7889382123575286e-06, "loss": 0.001, "step": 2982 }, { "epoch": 0.36, "learning_rate": 2.7888662267546494e-06, "loss": 0.0018, "step": 2983 }, { "epoch": 0.36, "learning_rate": 2.7887942411517697e-06, "loss": 0.002, "step": 2984 }, { "epoch": 0.36, "learning_rate": 2.78872225554889e-06, "loss": 0.0038, "step": 2985 }, { "epoch": 0.36, "learning_rate": 2.788650269946011e-06, "loss": 0.0016, "step": 2986 }, { "epoch": 0.36, "learning_rate": 2.7885782843431312e-06, "loss": 0.0023, "step": 2987 }, { "epoch": 0.36, "learning_rate": 2.788506298740252e-06, "loss": 0.0017, "step": 2988 }, { "epoch": 0.36, "learning_rate": 2.7884343131373723e-06, "loss": 0.0012, "step": 2989 }, { "epoch": 0.36, "learning_rate": 2.788362327534493e-06, "loss": 0.0015, "step": 2990 }, { "epoch": 0.36, "learning_rate": 2.788290341931614e-06, "loss": 0.0011, "step": 2991 }, { "epoch": 0.36, "learning_rate": 2.7882183563287343e-06, "loss": 0.005, "step": 2992 }, { "epoch": 0.36, "learning_rate": 2.788146370725855e-06, "loss": 0.001, "step": 2993 }, { "epoch": 0.36, "learning_rate": 2.788074385122976e-06, "loss": 0.0016, "step": 2994 }, { "epoch": 0.36, "learning_rate": 2.788002399520096e-06, "loss": 0.0013, "step": 2995 }, { "epoch": 0.36, "learning_rate": 2.7879304139172165e-06, "loss": 0.0011, "step": 2996 }, { "epoch": 0.36, "learning_rate": 2.7878584283143373e-06, "loss": 0.0014, "step": 2997 }, { "epoch": 0.36, "learning_rate": 2.7877864427114577e-06, "loss": 0.0013, "step": 2998 }, { "epoch": 0.36, "learning_rate": 2.7877144571085784e-06, "loss": 0.0011, "step": 2999 }, { "epoch": 0.36, "learning_rate": 2.787642471505699e-06, "loss": 0.0022, "step": 3000 }, { "epoch": 0.36, "learning_rate": 2.7875704859028196e-06, "loss": 0.0012, "step": 3001 }, { "epoch": 0.36, "learning_rate": 2.78749850029994e-06, "loss": 0.0118, "step": 3002 }, { "epoch": 0.36, "learning_rate": 2.7874265146970607e-06, "loss": 0.0011, "step": 3003 }, { "epoch": 0.36, "learning_rate": 2.787354529094181e-06, "loss": 0.148, "step": 3004 }, { "epoch": 0.36, "learning_rate": 2.787282543491302e-06, "loss": 0.0017, "step": 3005 }, { "epoch": 0.36, "learning_rate": 2.787210557888422e-06, "loss": 0.0009, "step": 3006 }, { "epoch": 0.36, "learning_rate": 2.787138572285543e-06, "loss": 0.0432, "step": 3007 }, { "epoch": 0.36, "learning_rate": 2.7870665866826638e-06, "loss": 0.0009, "step": 3008 }, { "epoch": 0.36, "learning_rate": 2.786994601079784e-06, "loss": 0.0008, "step": 3009 }, { "epoch": 0.36, "learning_rate": 2.786922615476905e-06, "loss": 0.9104, "step": 3010 }, { "epoch": 0.36, "learning_rate": 2.7868506298740252e-06, "loss": 0.001, "step": 3011 }, { "epoch": 0.36, "learning_rate": 2.786778644271146e-06, "loss": 0.4387, "step": 3012 }, { "epoch": 0.36, "learning_rate": 2.7867066586682664e-06, "loss": 0.0014, "step": 3013 }, { "epoch": 0.36, "learning_rate": 2.786634673065387e-06, "loss": 0.0013, "step": 3014 }, { "epoch": 0.36, "learning_rate": 2.7865626874625075e-06, "loss": 0.0016, "step": 3015 }, { "epoch": 0.36, "learning_rate": 2.7864907018596283e-06, "loss": 0.2371, "step": 3016 }, { "epoch": 0.36, "learning_rate": 2.7864187162567486e-06, "loss": 0.0016, "step": 3017 }, { "epoch": 0.36, "learning_rate": 2.786346730653869e-06, "loss": 0.0012, "step": 3018 }, { "epoch": 0.36, "learning_rate": 2.7862747450509898e-06, "loss": 0.0011, "step": 3019 }, { "epoch": 0.36, "learning_rate": 2.78620275944811e-06, "loss": 0.0012, "step": 3020 }, { "epoch": 0.36, "learning_rate": 2.786130773845231e-06, "loss": 0.1556, "step": 3021 }, { "epoch": 0.36, "learning_rate": 2.7860587882423517e-06, "loss": 0.0021, "step": 3022 }, { "epoch": 0.36, "learning_rate": 2.785986802639472e-06, "loss": 0.0014, "step": 3023 }, { "epoch": 0.36, "learning_rate": 2.785914817036593e-06, "loss": 0.0011, "step": 3024 }, { "epoch": 0.36, "learning_rate": 2.7858428314337136e-06, "loss": 0.0012, "step": 3025 }, { "epoch": 0.36, "learning_rate": 2.785770845830834e-06, "loss": 0.0011, "step": 3026 }, { "epoch": 0.36, "learning_rate": 2.7856988602279547e-06, "loss": 0.0011, "step": 3027 }, { "epoch": 0.36, "learning_rate": 2.785626874625075e-06, "loss": 0.0022, "step": 3028 }, { "epoch": 0.36, "learning_rate": 2.785554889022196e-06, "loss": 0.0045, "step": 3029 }, { "epoch": 0.36, "learning_rate": 2.7854829034193162e-06, "loss": 0.0014, "step": 3030 }, { "epoch": 0.36, "learning_rate": 2.7854109178164366e-06, "loss": 0.0009, "step": 3031 }, { "epoch": 0.36, "learning_rate": 2.7853389322135574e-06, "loss": 0.0444, "step": 3032 }, { "epoch": 0.36, "learning_rate": 2.7852669466106777e-06, "loss": 0.0016, "step": 3033 }, { "epoch": 0.36, "learning_rate": 2.7851949610077985e-06, "loss": 0.0024, "step": 3034 }, { "epoch": 0.36, "learning_rate": 2.785122975404919e-06, "loss": 0.0025, "step": 3035 }, { "epoch": 0.36, "learning_rate": 2.7850509898020396e-06, "loss": 0.0013, "step": 3036 }, { "epoch": 0.36, "learning_rate": 2.78497900419916e-06, "loss": 0.0013, "step": 3037 }, { "epoch": 0.36, "learning_rate": 2.7849070185962808e-06, "loss": 0.001, "step": 3038 }, { "epoch": 0.36, "learning_rate": 2.7848350329934015e-06, "loss": 0.0064, "step": 3039 }, { "epoch": 0.36, "learning_rate": 2.7847630473905223e-06, "loss": 0.0009, "step": 3040 }, { "epoch": 0.36, "learning_rate": 2.7846910617876427e-06, "loss": 0.0011, "step": 3041 }, { "epoch": 0.36, "learning_rate": 2.784619076184763e-06, "loss": 0.0012, "step": 3042 }, { "epoch": 0.36, "learning_rate": 2.784547090581884e-06, "loss": 0.0032, "step": 3043 }, { "epoch": 0.36, "learning_rate": 2.784475104979004e-06, "loss": 0.2405, "step": 3044 }, { "epoch": 0.36, "learning_rate": 2.784403119376125e-06, "loss": 0.001, "step": 3045 }, { "epoch": 0.37, "learning_rate": 2.7843311337732453e-06, "loss": 0.0019, "step": 3046 }, { "epoch": 0.37, "learning_rate": 2.784259148170366e-06, "loss": 0.0014, "step": 3047 }, { "epoch": 0.37, "learning_rate": 2.7841871625674864e-06, "loss": 0.0007, "step": 3048 }, { "epoch": 0.37, "learning_rate": 2.784115176964607e-06, "loss": 0.0012, "step": 3049 }, { "epoch": 0.37, "learning_rate": 2.7840431913617276e-06, "loss": 0.0008, "step": 3050 }, { "epoch": 0.37, "learning_rate": 2.7839712057588483e-06, "loss": 0.0014, "step": 3051 }, { "epoch": 0.37, "learning_rate": 2.7838992201559687e-06, "loss": 0.353, "step": 3052 }, { "epoch": 0.37, "learning_rate": 2.7838272345530895e-06, "loss": 0.0013, "step": 3053 }, { "epoch": 0.37, "learning_rate": 2.7837552489502103e-06, "loss": 0.0019, "step": 3054 }, { "epoch": 0.37, "learning_rate": 2.7836832633473306e-06, "loss": 0.0009, "step": 3055 }, { "epoch": 0.37, "learning_rate": 2.7836112777444514e-06, "loss": 0.0095, "step": 3056 }, { "epoch": 0.37, "learning_rate": 2.7835392921415717e-06, "loss": 0.0011, "step": 3057 }, { "epoch": 0.37, "learning_rate": 2.7834673065386925e-06, "loss": 0.0013, "step": 3058 }, { "epoch": 0.37, "learning_rate": 2.783395320935813e-06, "loss": 0.0072, "step": 3059 }, { "epoch": 0.37, "learning_rate": 2.7833233353329337e-06, "loss": 0.0014, "step": 3060 }, { "epoch": 0.37, "learning_rate": 2.783251349730054e-06, "loss": 0.0012, "step": 3061 }, { "epoch": 0.37, "learning_rate": 2.7831793641271748e-06, "loss": 0.0137, "step": 3062 }, { "epoch": 0.37, "learning_rate": 2.783107378524295e-06, "loss": 0.2774, "step": 3063 }, { "epoch": 0.37, "learning_rate": 2.7830353929214155e-06, "loss": 0.1223, "step": 3064 }, { "epoch": 0.37, "learning_rate": 2.7829634073185363e-06, "loss": 0.0017, "step": 3065 }, { "epoch": 0.37, "learning_rate": 2.7828914217156566e-06, "loss": 0.0011, "step": 3066 }, { "epoch": 0.37, "learning_rate": 2.7828194361127774e-06, "loss": 0.0014, "step": 3067 }, { "epoch": 0.37, "learning_rate": 2.782747450509898e-06, "loss": 0.0013, "step": 3068 }, { "epoch": 0.37, "learning_rate": 2.7826754649070185e-06, "loss": 0.0013, "step": 3069 }, { "epoch": 0.37, "learning_rate": 2.7826034793041393e-06, "loss": 0.0015, "step": 3070 }, { "epoch": 0.37, "learning_rate": 2.78253149370126e-06, "loss": 0.007, "step": 3071 }, { "epoch": 0.37, "learning_rate": 2.7824595080983805e-06, "loss": 0.0021, "step": 3072 }, { "epoch": 0.37, "learning_rate": 2.7823875224955012e-06, "loss": 0.0017, "step": 3073 }, { "epoch": 0.37, "learning_rate": 2.7823155368926216e-06, "loss": 0.0008, "step": 3074 }, { "epoch": 0.37, "learning_rate": 2.7822435512897424e-06, "loss": 0.0012, "step": 3075 }, { "epoch": 0.37, "learning_rate": 2.7821715656868627e-06, "loss": 0.1073, "step": 3076 }, { "epoch": 0.37, "learning_rate": 2.782099580083983e-06, "loss": 0.0016, "step": 3077 }, { "epoch": 0.37, "learning_rate": 2.782027594481104e-06, "loss": 0.0026, "step": 3078 }, { "epoch": 0.37, "learning_rate": 2.781955608878224e-06, "loss": 0.0009, "step": 3079 }, { "epoch": 0.37, "learning_rate": 2.781883623275345e-06, "loss": 0.527, "step": 3080 }, { "epoch": 0.37, "learning_rate": 2.7818116376724653e-06, "loss": 0.5069, "step": 3081 }, { "epoch": 0.37, "learning_rate": 2.781739652069586e-06, "loss": 0.6595, "step": 3082 }, { "epoch": 0.37, "learning_rate": 2.7816676664667065e-06, "loss": 1.017, "step": 3083 }, { "epoch": 0.37, "learning_rate": 2.7815956808638273e-06, "loss": 0.0009, "step": 3084 }, { "epoch": 0.37, "learning_rate": 2.781523695260948e-06, "loss": 1.3545, "step": 3085 }, { "epoch": 0.37, "learning_rate": 2.781451709658069e-06, "loss": 0.0008, "step": 3086 }, { "epoch": 0.37, "learning_rate": 2.781379724055189e-06, "loss": 0.0035, "step": 3087 }, { "epoch": 0.37, "learning_rate": 2.7813077384523095e-06, "loss": 0.0011, "step": 3088 }, { "epoch": 0.37, "learning_rate": 2.7812357528494303e-06, "loss": 0.0009, "step": 3089 }, { "epoch": 0.37, "learning_rate": 2.7811637672465507e-06, "loss": 0.0037, "step": 3090 }, { "epoch": 0.37, "learning_rate": 2.7810917816436714e-06, "loss": 0.0011, "step": 3091 }, { "epoch": 0.37, "learning_rate": 2.7810197960407918e-06, "loss": 0.0429, "step": 3092 }, { "epoch": 0.37, "learning_rate": 2.7809478104379126e-06, "loss": 0.001, "step": 3093 }, { "epoch": 0.37, "learning_rate": 2.780875824835033e-06, "loss": 0.1861, "step": 3094 }, { "epoch": 0.37, "learning_rate": 2.7808038392321537e-06, "loss": 0.0043, "step": 3095 }, { "epoch": 0.37, "learning_rate": 2.780731853629274e-06, "loss": 0.4139, "step": 3096 }, { "epoch": 0.37, "learning_rate": 2.780659868026395e-06, "loss": 0.0047, "step": 3097 }, { "epoch": 0.37, "learning_rate": 2.780587882423515e-06, "loss": 0.0018, "step": 3098 }, { "epoch": 0.37, "learning_rate": 2.780515896820636e-06, "loss": 0.0008, "step": 3099 }, { "epoch": 0.37, "learning_rate": 2.7804439112177567e-06, "loss": 0.0014, "step": 3100 }, { "epoch": 0.37, "learning_rate": 2.780371925614877e-06, "loss": 0.1379, "step": 3101 }, { "epoch": 0.37, "learning_rate": 2.780299940011998e-06, "loss": 0.0014, "step": 3102 }, { "epoch": 0.37, "learning_rate": 2.7802279544091182e-06, "loss": 0.0012, "step": 3103 }, { "epoch": 0.37, "learning_rate": 2.780155968806239e-06, "loss": 0.7865, "step": 3104 }, { "epoch": 0.37, "learning_rate": 2.7800839832033594e-06, "loss": 0.0009, "step": 3105 }, { "epoch": 0.37, "learning_rate": 2.78001199760048e-06, "loss": 0.0008, "step": 3106 }, { "epoch": 0.37, "learning_rate": 2.7799400119976005e-06, "loss": 0.0014, "step": 3107 }, { "epoch": 0.37, "learning_rate": 2.7798680263947213e-06, "loss": 0.0017, "step": 3108 }, { "epoch": 0.37, "learning_rate": 2.7797960407918416e-06, "loss": 0.0009, "step": 3109 }, { "epoch": 0.37, "learning_rate": 2.7797240551889624e-06, "loss": 0.001, "step": 3110 }, { "epoch": 0.37, "learning_rate": 2.7796520695860828e-06, "loss": 0.001, "step": 3111 }, { "epoch": 0.37, "learning_rate": 2.779580083983203e-06, "loss": 0.002, "step": 3112 }, { "epoch": 0.37, "learning_rate": 2.779508098380324e-06, "loss": 0.001, "step": 3113 }, { "epoch": 0.37, "learning_rate": 2.7794361127774447e-06, "loss": 0.0008, "step": 3114 }, { "epoch": 0.37, "learning_rate": 2.779364127174565e-06, "loss": 0.6909, "step": 3115 }, { "epoch": 0.37, "learning_rate": 2.779292141571686e-06, "loss": 0.0011, "step": 3116 }, { "epoch": 0.37, "learning_rate": 2.7792201559688066e-06, "loss": 0.7612, "step": 3117 }, { "epoch": 0.37, "learning_rate": 2.779148170365927e-06, "loss": 0.001, "step": 3118 }, { "epoch": 0.37, "learning_rate": 2.7790761847630477e-06, "loss": 0.0009, "step": 3119 }, { "epoch": 0.37, "learning_rate": 2.779004199160168e-06, "loss": 1.0093, "step": 3120 }, { "epoch": 0.37, "learning_rate": 2.778932213557289e-06, "loss": 0.0011, "step": 3121 }, { "epoch": 0.37, "learning_rate": 2.7788602279544092e-06, "loss": 0.0016, "step": 3122 }, { "epoch": 0.37, "learning_rate": 2.7787882423515296e-06, "loss": 0.0015, "step": 3123 }, { "epoch": 0.37, "learning_rate": 2.7787162567486503e-06, "loss": 0.0009, "step": 3124 }, { "epoch": 0.37, "learning_rate": 2.7786442711457707e-06, "loss": 0.0016, "step": 3125 }, { "epoch": 0.37, "learning_rate": 2.7785722855428915e-06, "loss": 0.0007, "step": 3126 }, { "epoch": 0.37, "learning_rate": 2.778500299940012e-06, "loss": 0.0011, "step": 3127 }, { "epoch": 0.37, "learning_rate": 2.7784283143371326e-06, "loss": 0.0008, "step": 3128 }, { "epoch": 0.37, "learning_rate": 2.778356328734253e-06, "loss": 0.001, "step": 3129 }, { "epoch": 0.38, "learning_rate": 2.7782843431313737e-06, "loss": 0.036, "step": 3130 }, { "epoch": 0.38, "learning_rate": 2.7782123575284945e-06, "loss": 0.0045, "step": 3131 }, { "epoch": 0.38, "learning_rate": 2.7781403719256153e-06, "loss": 0.0006, "step": 3132 }, { "epoch": 0.38, "learning_rate": 2.7780683863227357e-06, "loss": 0.0014, "step": 3133 }, { "epoch": 0.38, "learning_rate": 2.777996400719856e-06, "loss": 0.0992, "step": 3134 }, { "epoch": 0.38, "learning_rate": 2.777924415116977e-06, "loss": 0.0015, "step": 3135 }, { "epoch": 0.38, "learning_rate": 2.777852429514097e-06, "loss": 0.0012, "step": 3136 }, { "epoch": 0.38, "learning_rate": 2.777780443911218e-06, "loss": 0.0011, "step": 3137 }, { "epoch": 0.38, "learning_rate": 2.7777084583083383e-06, "loss": 0.002, "step": 3138 }, { "epoch": 0.38, "learning_rate": 2.777636472705459e-06, "loss": 0.0025, "step": 3139 }, { "epoch": 0.38, "learning_rate": 2.7775644871025794e-06, "loss": 0.4377, "step": 3140 }, { "epoch": 0.38, "learning_rate": 2.7774925014997e-06, "loss": 0.0007, "step": 3141 }, { "epoch": 0.38, "learning_rate": 2.7774205158968206e-06, "loss": 0.6264, "step": 3142 }, { "epoch": 0.38, "learning_rate": 2.7773485302939413e-06, "loss": 0.001, "step": 3143 }, { "epoch": 0.38, "learning_rate": 2.7772765446910617e-06, "loss": 0.4494, "step": 3144 }, { "epoch": 0.38, "learning_rate": 2.7772045590881825e-06, "loss": 0.0048, "step": 3145 }, { "epoch": 0.38, "learning_rate": 2.777132573485303e-06, "loss": 0.001, "step": 3146 }, { "epoch": 0.38, "learning_rate": 2.7770605878824236e-06, "loss": 0.0014, "step": 3147 }, { "epoch": 0.38, "learning_rate": 2.7769886022795444e-06, "loss": 0.0012, "step": 3148 }, { "epoch": 0.38, "learning_rate": 2.7769166166766647e-06, "loss": 0.0017, "step": 3149 }, { "epoch": 0.38, "learning_rate": 2.7768446310737855e-06, "loss": 0.0018, "step": 3150 }, { "epoch": 0.38, "learning_rate": 2.776772645470906e-06, "loss": 0.001, "step": 3151 }, { "epoch": 0.38, "learning_rate": 2.7767006598680266e-06, "loss": 0.0011, "step": 3152 }, { "epoch": 0.38, "learning_rate": 2.776628674265147e-06, "loss": 0.2397, "step": 3153 }, { "epoch": 0.38, "learning_rate": 2.7765566886622678e-06, "loss": 0.0016, "step": 3154 }, { "epoch": 0.38, "learning_rate": 2.776484703059388e-06, "loss": 0.0011, "step": 3155 }, { "epoch": 0.38, "learning_rate": 2.776412717456509e-06, "loss": 0.001, "step": 3156 }, { "epoch": 0.38, "learning_rate": 2.7763407318536293e-06, "loss": 0.3574, "step": 3157 }, { "epoch": 0.38, "learning_rate": 2.7762687462507496e-06, "loss": 0.7335, "step": 3158 }, { "epoch": 0.38, "learning_rate": 2.7761967606478704e-06, "loss": 0.2241, "step": 3159 }, { "epoch": 0.38, "learning_rate": 2.7761247750449908e-06, "loss": 0.7134, "step": 3160 }, { "epoch": 0.38, "learning_rate": 2.7760527894421115e-06, "loss": 0.001, "step": 3161 }, { "epoch": 0.38, "learning_rate": 2.7759808038392323e-06, "loss": 0.0008, "step": 3162 }, { "epoch": 0.38, "learning_rate": 2.775908818236353e-06, "loss": 0.0009, "step": 3163 }, { "epoch": 0.38, "learning_rate": 2.7758368326334734e-06, "loss": 0.0012, "step": 3164 }, { "epoch": 0.38, "learning_rate": 2.7757648470305942e-06, "loss": 0.0012, "step": 3165 }, { "epoch": 0.38, "learning_rate": 2.7756928614277146e-06, "loss": 0.0461, "step": 3166 }, { "epoch": 0.38, "learning_rate": 2.7756208758248354e-06, "loss": 0.0021, "step": 3167 }, { "epoch": 0.38, "learning_rate": 2.7755488902219557e-06, "loss": 0.002, "step": 3168 }, { "epoch": 0.38, "learning_rate": 2.775476904619076e-06, "loss": 0.8676, "step": 3169 }, { "epoch": 0.38, "learning_rate": 2.775404919016197e-06, "loss": 0.0017, "step": 3170 }, { "epoch": 0.38, "learning_rate": 2.775332933413317e-06, "loss": 0.0009, "step": 3171 }, { "epoch": 0.38, "learning_rate": 2.775260947810438e-06, "loss": 0.0011, "step": 3172 }, { "epoch": 0.38, "learning_rate": 2.7751889622075583e-06, "loss": 0.1095, "step": 3173 }, { "epoch": 0.38, "learning_rate": 2.775116976604679e-06, "loss": 0.0012, "step": 3174 }, { "epoch": 0.38, "learning_rate": 2.7750449910017995e-06, "loss": 0.0012, "step": 3175 }, { "epoch": 0.38, "learning_rate": 2.7749730053989202e-06, "loss": 0.0008, "step": 3176 }, { "epoch": 0.38, "learning_rate": 2.774901019796041e-06, "loss": 0.0008, "step": 3177 }, { "epoch": 0.38, "learning_rate": 2.7748290341931614e-06, "loss": 0.812, "step": 3178 }, { "epoch": 0.38, "learning_rate": 2.774757048590282e-06, "loss": 0.1541, "step": 3179 }, { "epoch": 0.38, "learning_rate": 2.7746850629874025e-06, "loss": 0.0005, "step": 3180 }, { "epoch": 0.38, "learning_rate": 2.7746130773845233e-06, "loss": 0.387, "step": 3181 }, { "epoch": 0.38, "learning_rate": 2.7745410917816436e-06, "loss": 1.3801, "step": 3182 }, { "epoch": 0.38, "learning_rate": 2.7744691061787644e-06, "loss": 0.0009, "step": 3183 }, { "epoch": 0.38, "learning_rate": 2.7743971205758848e-06, "loss": 0.2252, "step": 3184 }, { "epoch": 0.38, "learning_rate": 2.7743251349730056e-06, "loss": 0.0013, "step": 3185 }, { "epoch": 0.38, "learning_rate": 2.774253149370126e-06, "loss": 0.0007, "step": 3186 }, { "epoch": 0.38, "learning_rate": 2.7741811637672467e-06, "loss": 0.6362, "step": 3187 }, { "epoch": 0.38, "learning_rate": 2.774109178164367e-06, "loss": 0.0244, "step": 3188 }, { "epoch": 0.38, "learning_rate": 2.774037192561488e-06, "loss": 0.0011, "step": 3189 }, { "epoch": 0.38, "learning_rate": 2.773965206958608e-06, "loss": 0.001, "step": 3190 }, { "epoch": 0.38, "learning_rate": 2.773893221355729e-06, "loss": 0.0014, "step": 3191 }, { "epoch": 0.38, "learning_rate": 2.7738212357528493e-06, "loss": 0.0009, "step": 3192 }, { "epoch": 0.38, "learning_rate": 2.77374925014997e-06, "loss": 0.0008, "step": 3193 }, { "epoch": 0.38, "learning_rate": 2.773677264547091e-06, "loss": 0.0009, "step": 3194 }, { "epoch": 0.38, "learning_rate": 2.7736052789442112e-06, "loss": 0.0036, "step": 3195 }, { "epoch": 0.38, "learning_rate": 2.773533293341332e-06, "loss": 0.0066, "step": 3196 }, { "epoch": 0.38, "learning_rate": 2.7734613077384524e-06, "loss": 0.4436, "step": 3197 }, { "epoch": 0.38, "learning_rate": 2.773389322135573e-06, "loss": 0.0221, "step": 3198 }, { "epoch": 0.38, "learning_rate": 2.7733173365326935e-06, "loss": 0.0158, "step": 3199 }, { "epoch": 0.38, "learning_rate": 2.7732453509298143e-06, "loss": 0.0008, "step": 3200 }, { "epoch": 0.38, "learning_rate": 2.7731733653269346e-06, "loss": 0.0009, "step": 3201 }, { "epoch": 0.38, "learning_rate": 2.7731013797240554e-06, "loss": 0.0008, "step": 3202 }, { "epoch": 0.38, "learning_rate": 2.7730293941211758e-06, "loss": 0.0012, "step": 3203 }, { "epoch": 0.38, "learning_rate": 2.772957408518296e-06, "loss": 0.0006, "step": 3204 }, { "epoch": 0.38, "learning_rate": 2.772885422915417e-06, "loss": 0.0011, "step": 3205 }, { "epoch": 0.38, "learning_rate": 2.7728134373125372e-06, "loss": 0.0035, "step": 3206 }, { "epoch": 0.38, "learning_rate": 2.772741451709658e-06, "loss": 0.0013, "step": 3207 }, { "epoch": 0.38, "learning_rate": 2.772669466106779e-06, "loss": 0.3752, "step": 3208 }, { "epoch": 0.38, "learning_rate": 2.7725974805038996e-06, "loss": 0.0009, "step": 3209 }, { "epoch": 0.38, "learning_rate": 2.77252549490102e-06, "loss": 0.0008, "step": 3210 }, { "epoch": 0.38, "learning_rate": 2.7724535092981407e-06, "loss": 0.0013, "step": 3211 }, { "epoch": 0.38, "learning_rate": 2.772381523695261e-06, "loss": 0.0009, "step": 3212 }, { "epoch": 0.39, "learning_rate": 2.772309538092382e-06, "loss": 0.0009, "step": 3213 }, { "epoch": 0.39, "learning_rate": 2.772237552489502e-06, "loss": 0.0026, "step": 3214 }, { "epoch": 0.39, "learning_rate": 2.7721655668866226e-06, "loss": 0.0018, "step": 3215 }, { "epoch": 0.39, "learning_rate": 2.7720935812837433e-06, "loss": 0.0184, "step": 3216 }, { "epoch": 0.39, "learning_rate": 2.7720215956808637e-06, "loss": 0.0018, "step": 3217 }, { "epoch": 0.39, "learning_rate": 2.7719496100779845e-06, "loss": 0.7108, "step": 3218 }, { "epoch": 0.39, "learning_rate": 2.771877624475105e-06, "loss": 0.0013, "step": 3219 }, { "epoch": 0.39, "learning_rate": 2.7718056388722256e-06, "loss": 0.0018, "step": 3220 }, { "epoch": 0.39, "learning_rate": 2.771733653269346e-06, "loss": 0.0009, "step": 3221 }, { "epoch": 0.39, "learning_rate": 2.7716616676664667e-06, "loss": 0.3143, "step": 3222 }, { "epoch": 0.39, "learning_rate": 2.7715896820635875e-06, "loss": 0.0007, "step": 3223 }, { "epoch": 0.39, "learning_rate": 2.771517696460708e-06, "loss": 0.0009, "step": 3224 }, { "epoch": 0.39, "learning_rate": 2.7714457108578287e-06, "loss": 0.0012, "step": 3225 }, { "epoch": 0.39, "learning_rate": 2.771373725254949e-06, "loss": 0.0037, "step": 3226 }, { "epoch": 0.39, "learning_rate": 2.7713017396520698e-06, "loss": 0.3794, "step": 3227 }, { "epoch": 0.39, "learning_rate": 2.77122975404919e-06, "loss": 0.0013, "step": 3228 }, { "epoch": 0.39, "learning_rate": 2.771157768446311e-06, "loss": 0.8292, "step": 3229 }, { "epoch": 0.39, "learning_rate": 2.7710857828434313e-06, "loss": 0.0014, "step": 3230 }, { "epoch": 0.39, "learning_rate": 2.771013797240552e-06, "loss": 0.0009, "step": 3231 }, { "epoch": 0.39, "learning_rate": 2.7709418116376724e-06, "loss": 0.001, "step": 3232 }, { "epoch": 0.39, "learning_rate": 2.770869826034793e-06, "loss": 0.4473, "step": 3233 }, { "epoch": 0.39, "learning_rate": 2.7707978404319135e-06, "loss": 0.0009, "step": 3234 }, { "epoch": 0.39, "learning_rate": 2.7707258548290343e-06, "loss": 0.0009, "step": 3235 }, { "epoch": 0.39, "learning_rate": 2.7706538692261547e-06, "loss": 0.001, "step": 3236 }, { "epoch": 0.39, "learning_rate": 2.7705818836232755e-06, "loss": 0.0014, "step": 3237 }, { "epoch": 0.39, "learning_rate": 2.770509898020396e-06, "loss": 0.4251, "step": 3238 }, { "epoch": 0.39, "learning_rate": 2.7704379124175166e-06, "loss": 0.4918, "step": 3239 }, { "epoch": 0.39, "learning_rate": 2.7703659268146374e-06, "loss": 0.458, "step": 3240 }, { "epoch": 0.39, "learning_rate": 2.7702939412117577e-06, "loss": 0.6379, "step": 3241 }, { "epoch": 0.39, "learning_rate": 2.7702219556088785e-06, "loss": 0.0011, "step": 3242 }, { "epoch": 0.39, "learning_rate": 2.770149970005999e-06, "loss": 0.0009, "step": 3243 }, { "epoch": 0.39, "learning_rate": 2.7700779844031196e-06, "loss": 0.0013, "step": 3244 }, { "epoch": 0.39, "learning_rate": 2.77000599880024e-06, "loss": 0.0011, "step": 3245 }, { "epoch": 0.39, "learning_rate": 2.7699340131973608e-06, "loss": 0.0016, "step": 3246 }, { "epoch": 0.39, "learning_rate": 2.769862027594481e-06, "loss": 0.001, "step": 3247 }, { "epoch": 0.39, "learning_rate": 2.769790041991602e-06, "loss": 0.0008, "step": 3248 }, { "epoch": 0.39, "learning_rate": 2.7697180563887223e-06, "loss": 0.0008, "step": 3249 }, { "epoch": 0.39, "learning_rate": 2.7696460707858426e-06, "loss": 0.0011, "step": 3250 }, { "epoch": 0.39, "learning_rate": 2.7695740851829634e-06, "loss": 0.0008, "step": 3251 }, { "epoch": 0.39, "learning_rate": 2.7695020995800837e-06, "loss": 0.6342, "step": 3252 }, { "epoch": 0.39, "learning_rate": 2.7694301139772045e-06, "loss": 0.0007, "step": 3253 }, { "epoch": 0.39, "learning_rate": 2.7693581283743253e-06, "loss": 0.0008, "step": 3254 }, { "epoch": 0.39, "learning_rate": 2.769286142771446e-06, "loss": 0.0007, "step": 3255 }, { "epoch": 0.39, "learning_rate": 2.7692141571685664e-06, "loss": 0.124, "step": 3256 }, { "epoch": 0.39, "learning_rate": 2.7691421715656872e-06, "loss": 0.7254, "step": 3257 }, { "epoch": 0.39, "learning_rate": 2.7690701859628076e-06, "loss": 0.0006, "step": 3258 }, { "epoch": 0.39, "learning_rate": 2.7689982003599283e-06, "loss": 0.0008, "step": 3259 }, { "epoch": 0.39, "learning_rate": 2.7689262147570487e-06, "loss": 0.7412, "step": 3260 }, { "epoch": 0.39, "learning_rate": 2.768854229154169e-06, "loss": 0.0008, "step": 3261 }, { "epoch": 0.39, "learning_rate": 2.76878224355129e-06, "loss": 0.0007, "step": 3262 }, { "epoch": 0.39, "learning_rate": 2.76871025794841e-06, "loss": 0.5628, "step": 3263 }, { "epoch": 0.39, "learning_rate": 2.768638272345531e-06, "loss": 0.0012, "step": 3264 }, { "epoch": 0.39, "learning_rate": 2.7685662867426513e-06, "loss": 0.0027, "step": 3265 }, { "epoch": 0.39, "learning_rate": 2.768494301139772e-06, "loss": 0.0008, "step": 3266 }, { "epoch": 0.39, "learning_rate": 2.7684223155368925e-06, "loss": 0.0008, "step": 3267 }, { "epoch": 0.39, "learning_rate": 2.7683503299340132e-06, "loss": 0.0009, "step": 3268 }, { "epoch": 0.39, "learning_rate": 2.7682783443311336e-06, "loss": 0.053, "step": 3269 }, { "epoch": 0.39, "learning_rate": 2.7682063587282544e-06, "loss": 0.0016, "step": 3270 }, { "epoch": 0.39, "learning_rate": 2.768134373125375e-06, "loss": 0.0008, "step": 3271 }, { "epoch": 0.39, "learning_rate": 2.7680623875224955e-06, "loss": 0.0008, "step": 3272 }, { "epoch": 0.39, "learning_rate": 2.7679904019196163e-06, "loss": 0.0008, "step": 3273 }, { "epoch": 0.39, "learning_rate": 2.7679184163167366e-06, "loss": 0.3812, "step": 3274 }, { "epoch": 0.39, "learning_rate": 2.7678464307138574e-06, "loss": 0.0015, "step": 3275 }, { "epoch": 0.39, "learning_rate": 2.7677744451109778e-06, "loss": 0.0011, "step": 3276 }, { "epoch": 0.39, "learning_rate": 2.7677024595080986e-06, "loss": 0.0215, "step": 3277 }, { "epoch": 0.39, "learning_rate": 2.767630473905219e-06, "loss": 0.7886, "step": 3278 }, { "epoch": 0.39, "learning_rate": 2.7675584883023397e-06, "loss": 0.014, "step": 3279 }, { "epoch": 0.39, "learning_rate": 2.76748650269946e-06, "loss": 0.0012, "step": 3280 }, { "epoch": 0.39, "learning_rate": 2.767414517096581e-06, "loss": 0.0008, "step": 3281 }, { "epoch": 0.39, "learning_rate": 2.767342531493701e-06, "loss": 0.0031, "step": 3282 }, { "epoch": 0.39, "learning_rate": 2.767270545890822e-06, "loss": 0.0015, "step": 3283 }, { "epoch": 0.39, "learning_rate": 2.7671985602879423e-06, "loss": 0.001, "step": 3284 }, { "epoch": 0.39, "learning_rate": 2.767126574685063e-06, "loss": 0.0066, "step": 3285 }, { "epoch": 0.39, "learning_rate": 2.767054589082184e-06, "loss": 0.0007, "step": 3286 }, { "epoch": 0.39, "learning_rate": 2.7669826034793042e-06, "loss": 0.0008, "step": 3287 }, { "epoch": 0.39, "learning_rate": 2.766910617876425e-06, "loss": 0.0008, "step": 3288 }, { "epoch": 0.39, "learning_rate": 2.7668386322735454e-06, "loss": 0.0009, "step": 3289 }, { "epoch": 0.39, "learning_rate": 2.766766646670666e-06, "loss": 0.0881, "step": 3290 }, { "epoch": 0.39, "learning_rate": 2.7666946610677865e-06, "loss": 0.0526, "step": 3291 }, { "epoch": 0.39, "learning_rate": 2.7666226754649073e-06, "loss": 0.0008, "step": 3292 }, { "epoch": 0.39, "learning_rate": 2.7665506898620276e-06, "loss": 0.0007, "step": 3293 }, { "epoch": 0.39, "learning_rate": 2.7664787042591484e-06, "loss": 0.0007, "step": 3294 }, { "epoch": 0.39, "learning_rate": 2.7664067186562688e-06, "loss": 0.0007, "step": 3295 }, { "epoch": 0.39, "learning_rate": 2.766334733053389e-06, "loss": 0.0051, "step": 3296 }, { "epoch": 0.4, "learning_rate": 2.76626274745051e-06, "loss": 0.8261, "step": 3297 }, { "epoch": 0.4, "learning_rate": 2.7661907618476302e-06, "loss": 0.0711, "step": 3298 }, { "epoch": 0.4, "learning_rate": 2.766118776244751e-06, "loss": 0.0011, "step": 3299 }, { "epoch": 0.4, "learning_rate": 2.766046790641872e-06, "loss": 0.0007, "step": 3300 }, { "epoch": 0.4, "learning_rate": 2.765974805038992e-06, "loss": 0.8323, "step": 3301 }, { "epoch": 0.4, "learning_rate": 2.765902819436113e-06, "loss": 0.0026, "step": 3302 }, { "epoch": 0.4, "learning_rate": 2.7658308338332337e-06, "loss": 0.4768, "step": 3303 }, { "epoch": 0.4, "learning_rate": 2.765758848230354e-06, "loss": 0.0011, "step": 3304 }, { "epoch": 0.4, "learning_rate": 2.765686862627475e-06, "loss": 0.0007, "step": 3305 }, { "epoch": 0.4, "learning_rate": 2.765614877024595e-06, "loss": 0.0006, "step": 3306 }, { "epoch": 0.4, "learning_rate": 2.7655428914217156e-06, "loss": 0.0009, "step": 3307 }, { "epoch": 0.4, "learning_rate": 2.7654709058188363e-06, "loss": 0.001, "step": 3308 }, { "epoch": 0.4, "learning_rate": 2.7653989202159567e-06, "loss": 0.0037, "step": 3309 }, { "epoch": 0.4, "learning_rate": 2.7653269346130775e-06, "loss": 0.0501, "step": 3310 }, { "epoch": 0.4, "learning_rate": 2.765254949010198e-06, "loss": 0.0024, "step": 3311 }, { "epoch": 0.4, "learning_rate": 2.7651829634073186e-06, "loss": 0.0008, "step": 3312 }, { "epoch": 0.4, "learning_rate": 2.765110977804439e-06, "loss": 0.0125, "step": 3313 }, { "epoch": 0.4, "learning_rate": 2.7650389922015597e-06, "loss": 0.6021, "step": 3314 }, { "epoch": 0.4, "learning_rate": 2.76496700659868e-06, "loss": 0.0007, "step": 3315 }, { "epoch": 0.4, "learning_rate": 2.764895020995801e-06, "loss": 0.001, "step": 3316 }, { "epoch": 0.4, "learning_rate": 2.7648230353929216e-06, "loss": 0.0016, "step": 3317 }, { "epoch": 0.4, "learning_rate": 2.764751049790042e-06, "loss": 0.0017, "step": 3318 }, { "epoch": 0.4, "learning_rate": 2.7646790641871628e-06, "loss": 0.0006, "step": 3319 }, { "epoch": 0.4, "learning_rate": 2.764607078584283e-06, "loss": 0.001, "step": 3320 }, { "epoch": 0.4, "learning_rate": 2.764535092981404e-06, "loss": 0.0007, "step": 3321 }, { "epoch": 0.4, "learning_rate": 2.7644631073785243e-06, "loss": 0.001, "step": 3322 }, { "epoch": 0.4, "learning_rate": 2.764391121775645e-06, "loss": 0.001, "step": 3323 }, { "epoch": 0.4, "learning_rate": 2.7643191361727654e-06, "loss": 0.0008, "step": 3324 }, { "epoch": 0.4, "learning_rate": 2.764247150569886e-06, "loss": 0.0009, "step": 3325 }, { "epoch": 0.4, "learning_rate": 2.7641751649670065e-06, "loss": 0.0012, "step": 3326 }, { "epoch": 0.4, "learning_rate": 2.7641031793641273e-06, "loss": 0.0014, "step": 3327 }, { "epoch": 0.4, "learning_rate": 2.7640311937612477e-06, "loss": 0.3017, "step": 3328 }, { "epoch": 0.4, "learning_rate": 2.7639592081583684e-06, "loss": 0.0011, "step": 3329 }, { "epoch": 0.4, "learning_rate": 2.763887222555489e-06, "loss": 0.0009, "step": 3330 }, { "epoch": 0.4, "learning_rate": 2.7638152369526096e-06, "loss": 0.0008, "step": 3331 }, { "epoch": 0.4, "learning_rate": 2.7637432513497304e-06, "loss": 0.3767, "step": 3332 }, { "epoch": 0.4, "learning_rate": 2.7636712657468507e-06, "loss": 0.0009, "step": 3333 }, { "epoch": 0.4, "learning_rate": 2.7635992801439715e-06, "loss": 0.0009, "step": 3334 }, { "epoch": 0.4, "learning_rate": 2.763527294541092e-06, "loss": 0.6714, "step": 3335 }, { "epoch": 0.4, "learning_rate": 2.7634553089382126e-06, "loss": 0.001, "step": 3336 }, { "epoch": 0.4, "learning_rate": 2.763383323335333e-06, "loss": 0.0008, "step": 3337 }, { "epoch": 0.4, "learning_rate": 2.7633113377324538e-06, "loss": 0.0009, "step": 3338 }, { "epoch": 0.4, "learning_rate": 2.763239352129574e-06, "loss": 0.3179, "step": 3339 }, { "epoch": 0.4, "learning_rate": 2.763167366526695e-06, "loss": 0.0009, "step": 3340 }, { "epoch": 0.4, "learning_rate": 2.7630953809238152e-06, "loss": 0.0009, "step": 3341 }, { "epoch": 0.4, "learning_rate": 2.7630233953209356e-06, "loss": 0.0007, "step": 3342 }, { "epoch": 0.4, "learning_rate": 2.7629514097180564e-06, "loss": 0.5762, "step": 3343 }, { "epoch": 0.4, "learning_rate": 2.7628794241151767e-06, "loss": 0.1007, "step": 3344 }, { "epoch": 0.4, "learning_rate": 2.7628074385122975e-06, "loss": 0.0009, "step": 3345 }, { "epoch": 0.4, "learning_rate": 2.7627354529094183e-06, "loss": 0.3659, "step": 3346 }, { "epoch": 0.4, "learning_rate": 2.7626634673065386e-06, "loss": 0.001, "step": 3347 }, { "epoch": 0.4, "learning_rate": 2.7625914817036594e-06, "loss": 0.0007, "step": 3348 }, { "epoch": 0.4, "learning_rate": 2.76251949610078e-06, "loss": 0.0007, "step": 3349 }, { "epoch": 0.4, "learning_rate": 2.7624475104979006e-06, "loss": 0.0007, "step": 3350 }, { "epoch": 0.4, "learning_rate": 2.7623755248950213e-06, "loss": 0.0007, "step": 3351 }, { "epoch": 0.4, "learning_rate": 2.7623035392921417e-06, "loss": 0.0006, "step": 3352 }, { "epoch": 0.4, "learning_rate": 2.762231553689262e-06, "loss": 0.0007, "step": 3353 }, { "epoch": 0.4, "learning_rate": 2.762159568086383e-06, "loss": 0.0023, "step": 3354 }, { "epoch": 0.4, "learning_rate": 2.762087582483503e-06, "loss": 0.0008, "step": 3355 }, { "epoch": 0.4, "learning_rate": 2.762015596880624e-06, "loss": 0.3321, "step": 3356 }, { "epoch": 0.4, "learning_rate": 2.7619436112777443e-06, "loss": 0.9172, "step": 3357 }, { "epoch": 0.4, "learning_rate": 2.761871625674865e-06, "loss": 0.0006, "step": 3358 }, { "epoch": 0.4, "learning_rate": 2.7617996400719855e-06, "loss": 0.001, "step": 3359 }, { "epoch": 0.4, "learning_rate": 2.7617276544691062e-06, "loss": 0.0075, "step": 3360 }, { "epoch": 0.4, "learning_rate": 2.7616556688662266e-06, "loss": 0.001, "step": 3361 }, { "epoch": 0.4, "learning_rate": 2.7615836832633474e-06, "loss": 0.3305, "step": 3362 }, { "epoch": 0.4, "learning_rate": 2.761511697660468e-06, "loss": 0.0013, "step": 3363 }, { "epoch": 0.4, "learning_rate": 2.761439712057589e-06, "loss": 0.0007, "step": 3364 }, { "epoch": 0.4, "learning_rate": 2.7613677264547093e-06, "loss": 0.0085, "step": 3365 }, { "epoch": 0.4, "learning_rate": 2.7612957408518296e-06, "loss": 0.8119, "step": 3366 }, { "epoch": 0.4, "learning_rate": 2.7612237552489504e-06, "loss": 0.5132, "step": 3367 }, { "epoch": 0.4, "learning_rate": 2.7611517696460708e-06, "loss": 0.0093, "step": 3368 }, { "epoch": 0.4, "learning_rate": 2.7610797840431915e-06, "loss": 0.0015, "step": 3369 }, { "epoch": 0.4, "learning_rate": 2.761007798440312e-06, "loss": 0.0499, "step": 3370 }, { "epoch": 0.4, "learning_rate": 2.7609358128374327e-06, "loss": 0.0009, "step": 3371 }, { "epoch": 0.4, "learning_rate": 2.760863827234553e-06, "loss": 0.2952, "step": 3372 }, { "epoch": 0.4, "learning_rate": 2.760791841631674e-06, "loss": 0.0034, "step": 3373 }, { "epoch": 0.4, "learning_rate": 2.760719856028794e-06, "loss": 0.0639, "step": 3374 }, { "epoch": 0.4, "learning_rate": 2.760647870425915e-06, "loss": 0.0062, "step": 3375 }, { "epoch": 0.4, "learning_rate": 2.7605758848230353e-06, "loss": 0.0009, "step": 3376 }, { "epoch": 0.4, "learning_rate": 2.760503899220156e-06, "loss": 0.001, "step": 3377 }, { "epoch": 0.4, "learning_rate": 2.760431913617277e-06, "loss": 0.0008, "step": 3378 }, { "epoch": 0.4, "learning_rate": 2.760359928014397e-06, "loss": 0.0007, "step": 3379 }, { "epoch": 0.41, "learning_rate": 2.760287942411518e-06, "loss": 0.0014, "step": 3380 }, { "epoch": 0.41, "learning_rate": 2.7602159568086383e-06, "loss": 0.0006, "step": 3381 }, { "epoch": 0.41, "learning_rate": 2.760143971205759e-06, "loss": 0.0101, "step": 3382 }, { "epoch": 0.41, "learning_rate": 2.7600719856028795e-06, "loss": 0.8342, "step": 3383 }, { "epoch": 0.41, "learning_rate": 2.7600000000000003e-06, "loss": 0.0013, "step": 3384 }, { "epoch": 0.41, "learning_rate": 2.7599280143971206e-06, "loss": 0.001, "step": 3385 }, { "epoch": 0.41, "learning_rate": 2.7598560287942414e-06, "loss": 0.0014, "step": 3386 }, { "epoch": 0.41, "learning_rate": 2.7597840431913617e-06, "loss": 0.0014, "step": 3387 }, { "epoch": 0.41, "learning_rate": 2.759712057588482e-06, "loss": 0.001, "step": 3388 }, { "epoch": 0.41, "learning_rate": 2.759640071985603e-06, "loss": 0.0007, "step": 3389 }, { "epoch": 0.41, "learning_rate": 2.7595680863827232e-06, "loss": 0.0008, "step": 3390 }, { "epoch": 0.41, "learning_rate": 2.759496100779844e-06, "loss": 0.0008, "step": 3391 }, { "epoch": 0.41, "learning_rate": 2.759424115176965e-06, "loss": 0.001, "step": 3392 }, { "epoch": 0.41, "learning_rate": 2.759352129574085e-06, "loss": 0.9083, "step": 3393 }, { "epoch": 0.41, "learning_rate": 2.759280143971206e-06, "loss": 0.0009, "step": 3394 }, { "epoch": 0.41, "learning_rate": 2.7592081583683267e-06, "loss": 0.0007, "step": 3395 }, { "epoch": 0.41, "learning_rate": 2.759136172765447e-06, "loss": 0.1687, "step": 3396 }, { "epoch": 0.41, "learning_rate": 2.759064187162568e-06, "loss": 0.001, "step": 3397 }, { "epoch": 0.41, "learning_rate": 2.758992201559688e-06, "loss": 0.0011, "step": 3398 }, { "epoch": 0.41, "learning_rate": 2.7589202159568085e-06, "loss": 0.0151, "step": 3399 }, { "epoch": 0.41, "learning_rate": 2.7588482303539293e-06, "loss": 1.4232, "step": 3400 }, { "epoch": 0.41, "learning_rate": 2.7587762447510497e-06, "loss": 0.0011, "step": 3401 }, { "epoch": 0.41, "learning_rate": 2.7587042591481705e-06, "loss": 0.001, "step": 3402 }, { "epoch": 0.41, "learning_rate": 2.758632273545291e-06, "loss": 0.5428, "step": 3403 }, { "epoch": 0.41, "learning_rate": 2.7585602879424116e-06, "loss": 0.3433, "step": 3404 }, { "epoch": 0.41, "learning_rate": 2.758488302339532e-06, "loss": 0.0028, "step": 3405 }, { "epoch": 0.41, "learning_rate": 2.7584163167366527e-06, "loss": 0.0008, "step": 3406 }, { "epoch": 0.41, "learning_rate": 2.758344331133773e-06, "loss": 0.0013, "step": 3407 }, { "epoch": 0.41, "learning_rate": 2.758272345530894e-06, "loss": 0.3499, "step": 3408 }, { "epoch": 0.41, "learning_rate": 2.7582003599280146e-06, "loss": 0.0013, "step": 3409 }, { "epoch": 0.41, "learning_rate": 2.7581283743251354e-06, "loss": 0.0019, "step": 3410 }, { "epoch": 0.41, "learning_rate": 2.7580563887222558e-06, "loss": 0.0012, "step": 3411 }, { "epoch": 0.41, "learning_rate": 2.757984403119376e-06, "loss": 0.0011, "step": 3412 }, { "epoch": 0.41, "learning_rate": 2.757912417516497e-06, "loss": 0.0016, "step": 3413 }, { "epoch": 0.41, "learning_rate": 2.7578404319136173e-06, "loss": 0.002, "step": 3414 }, { "epoch": 0.41, "learning_rate": 2.757768446310738e-06, "loss": 0.3086, "step": 3415 }, { "epoch": 0.41, "learning_rate": 2.7576964607078584e-06, "loss": 0.0009, "step": 3416 }, { "epoch": 0.41, "learning_rate": 2.757624475104979e-06, "loss": 0.002, "step": 3417 }, { "epoch": 0.41, "learning_rate": 2.7575524895020995e-06, "loss": 0.0012, "step": 3418 }, { "epoch": 0.41, "learning_rate": 2.7574805038992203e-06, "loss": 0.0768, "step": 3419 }, { "epoch": 0.41, "learning_rate": 2.7574085182963407e-06, "loss": 0.001, "step": 3420 }, { "epoch": 0.41, "learning_rate": 2.7573365326934614e-06, "loss": 0.0013, "step": 3421 }, { "epoch": 0.41, "learning_rate": 2.757264547090582e-06, "loss": 0.0008, "step": 3422 }, { "epoch": 0.41, "learning_rate": 2.7571925614877026e-06, "loss": 0.0008, "step": 3423 }, { "epoch": 0.41, "learning_rate": 2.757120575884823e-06, "loss": 0.0011, "step": 3424 }, { "epoch": 0.41, "learning_rate": 2.7570485902819437e-06, "loss": 0.0068, "step": 3425 }, { "epoch": 0.41, "learning_rate": 2.7569766046790645e-06, "loss": 0.0009, "step": 3426 }, { "epoch": 0.41, "learning_rate": 2.756904619076185e-06, "loss": 0.0008, "step": 3427 }, { "epoch": 0.41, "learning_rate": 2.7568326334733056e-06, "loss": 0.0007, "step": 3428 }, { "epoch": 0.41, "learning_rate": 2.756760647870426e-06, "loss": 0.0015, "step": 3429 }, { "epoch": 0.41, "learning_rate": 2.7566886622675468e-06, "loss": 0.0008, "step": 3430 }, { "epoch": 0.41, "learning_rate": 2.756616676664667e-06, "loss": 0.0008, "step": 3431 }, { "epoch": 0.41, "learning_rate": 2.756544691061788e-06, "loss": 0.0007, "step": 3432 }, { "epoch": 0.41, "learning_rate": 2.7564727054589082e-06, "loss": 0.0008, "step": 3433 }, { "epoch": 0.41, "learning_rate": 2.7564007198560286e-06, "loss": 0.0012, "step": 3434 }, { "epoch": 0.41, "learning_rate": 2.7563287342531494e-06, "loss": 0.1149, "step": 3435 }, { "epoch": 0.41, "learning_rate": 2.7562567486502697e-06, "loss": 0.0104, "step": 3436 }, { "epoch": 0.41, "learning_rate": 2.7561847630473905e-06, "loss": 0.0013, "step": 3437 }, { "epoch": 0.41, "learning_rate": 2.756112777444511e-06, "loss": 0.0011, "step": 3438 }, { "epoch": 0.41, "learning_rate": 2.7560407918416316e-06, "loss": 0.0008, "step": 3439 }, { "epoch": 0.41, "learning_rate": 2.7559688062387524e-06, "loss": 0.0007, "step": 3440 }, { "epoch": 0.41, "learning_rate": 2.755896820635873e-06, "loss": 1.8927, "step": 3441 }, { "epoch": 0.41, "learning_rate": 2.7558248350329936e-06, "loss": 0.1902, "step": 3442 }, { "epoch": 0.41, "learning_rate": 2.7557528494301143e-06, "loss": 0.0007, "step": 3443 }, { "epoch": 0.41, "learning_rate": 2.7556808638272347e-06, "loss": 0.8278, "step": 3444 }, { "epoch": 0.41, "learning_rate": 2.755608878224355e-06, "loss": 0.004, "step": 3445 }, { "epoch": 0.41, "learning_rate": 2.755536892621476e-06, "loss": 0.0011, "step": 3446 }, { "epoch": 0.41, "learning_rate": 2.755464907018596e-06, "loss": 1.8658, "step": 3447 }, { "epoch": 0.41, "learning_rate": 2.755392921415717e-06, "loss": 0.0012, "step": 3448 }, { "epoch": 0.41, "learning_rate": 2.7553209358128373e-06, "loss": 0.0011, "step": 3449 }, { "epoch": 0.41, "learning_rate": 2.755248950209958e-06, "loss": 0.0251, "step": 3450 }, { "epoch": 0.41, "learning_rate": 2.7551769646070784e-06, "loss": 0.0268, "step": 3451 }, { "epoch": 0.41, "learning_rate": 2.7551049790041992e-06, "loss": 0.0007, "step": 3452 }, { "epoch": 0.41, "learning_rate": 2.7550329934013196e-06, "loss": 0.0008, "step": 3453 }, { "epoch": 0.41, "learning_rate": 2.7549610077984404e-06, "loss": 0.5971, "step": 3454 }, { "epoch": 0.41, "learning_rate": 2.754889022195561e-06, "loss": 0.0015, "step": 3455 }, { "epoch": 0.41, "learning_rate": 2.7548170365926815e-06, "loss": 0.0504, "step": 3456 }, { "epoch": 0.41, "learning_rate": 2.7547450509898023e-06, "loss": 0.025, "step": 3457 }, { "epoch": 0.41, "learning_rate": 2.7546730653869226e-06, "loss": 0.0088, "step": 3458 }, { "epoch": 0.41, "learning_rate": 2.7546010797840434e-06, "loss": 0.0014, "step": 3459 }, { "epoch": 0.41, "learning_rate": 2.7545290941811638e-06, "loss": 0.0017, "step": 3460 }, { "epoch": 0.41, "learning_rate": 2.7544571085782845e-06, "loss": 0.0013, "step": 3461 }, { "epoch": 0.41, "learning_rate": 2.754385122975405e-06, "loss": 0.0026, "step": 3462 }, { "epoch": 0.41, "learning_rate": 2.7543131373725257e-06, "loss": 0.0011, "step": 3463 }, { "epoch": 0.42, "learning_rate": 2.754241151769646e-06, "loss": 0.5315, "step": 3464 }, { "epoch": 0.42, "learning_rate": 2.754169166166767e-06, "loss": 0.0009, "step": 3465 }, { "epoch": 0.42, "learning_rate": 2.754097180563887e-06, "loss": 0.0011, "step": 3466 }, { "epoch": 0.42, "learning_rate": 2.754025194961008e-06, "loss": 0.001, "step": 3467 }, { "epoch": 0.42, "learning_rate": 2.7539532093581283e-06, "loss": 0.2503, "step": 3468 }, { "epoch": 0.42, "learning_rate": 2.753881223755249e-06, "loss": 0.0007, "step": 3469 }, { "epoch": 0.42, "learning_rate": 2.7538092381523694e-06, "loss": 0.0572, "step": 3470 }, { "epoch": 0.42, "learning_rate": 2.75373725254949e-06, "loss": 0.0012, "step": 3471 }, { "epoch": 0.42, "learning_rate": 2.753665266946611e-06, "loss": 0.0009, "step": 3472 }, { "epoch": 0.42, "learning_rate": 2.7535932813437313e-06, "loss": 0.0012, "step": 3473 }, { "epoch": 0.42, "learning_rate": 2.753521295740852e-06, "loss": 0.0009, "step": 3474 }, { "epoch": 0.42, "learning_rate": 2.7534493101379725e-06, "loss": 0.0011, "step": 3475 }, { "epoch": 0.42, "learning_rate": 2.7533773245350932e-06, "loss": 0.0041, "step": 3476 }, { "epoch": 0.42, "learning_rate": 2.7533053389322136e-06, "loss": 0.0038, "step": 3477 }, { "epoch": 0.42, "learning_rate": 2.7532333533293344e-06, "loss": 0.0021, "step": 3478 }, { "epoch": 0.42, "learning_rate": 2.7531613677264547e-06, "loss": 0.0008, "step": 3479 }, { "epoch": 0.42, "learning_rate": 2.753089382123575e-06, "loss": 0.0008, "step": 3480 }, { "epoch": 0.42, "learning_rate": 2.753017396520696e-06, "loss": 0.0014, "step": 3481 }, { "epoch": 0.42, "learning_rate": 2.7529454109178162e-06, "loss": 0.0016, "step": 3482 }, { "epoch": 0.42, "learning_rate": 2.752873425314937e-06, "loss": 0.0007, "step": 3483 }, { "epoch": 0.42, "learning_rate": 2.7528014397120574e-06, "loss": 0.0481, "step": 3484 }, { "epoch": 0.42, "learning_rate": 2.752729454109178e-06, "loss": 0.0011, "step": 3485 }, { "epoch": 0.42, "learning_rate": 2.752657468506299e-06, "loss": 0.0028, "step": 3486 }, { "epoch": 0.42, "learning_rate": 2.7525854829034197e-06, "loss": 0.4144, "step": 3487 }, { "epoch": 0.42, "learning_rate": 2.75251349730054e-06, "loss": 0.0059, "step": 3488 }, { "epoch": 0.42, "learning_rate": 2.752441511697661e-06, "loss": 0.001, "step": 3489 }, { "epoch": 0.42, "learning_rate": 2.752369526094781e-06, "loss": 0.0006, "step": 3490 }, { "epoch": 0.42, "learning_rate": 2.752297540491902e-06, "loss": 0.0011, "step": 3491 }, { "epoch": 0.42, "learning_rate": 2.7522255548890223e-06, "loss": 0.0226, "step": 3492 }, { "epoch": 0.42, "learning_rate": 2.7521535692861427e-06, "loss": 0.0007, "step": 3493 }, { "epoch": 0.42, "learning_rate": 2.7520815836832635e-06, "loss": 0.0008, "step": 3494 }, { "epoch": 0.42, "learning_rate": 2.752009598080384e-06, "loss": 0.0007, "step": 3495 }, { "epoch": 0.42, "learning_rate": 2.7519376124775046e-06, "loss": 0.002, "step": 3496 }, { "epoch": 0.42, "learning_rate": 2.751865626874625e-06, "loss": 0.0008, "step": 3497 }, { "epoch": 0.42, "learning_rate": 2.7517936412717457e-06, "loss": 0.0008, "step": 3498 }, { "epoch": 0.42, "learning_rate": 2.751721655668866e-06, "loss": 0.0007, "step": 3499 }, { "epoch": 0.42, "learning_rate": 2.751649670065987e-06, "loss": 0.0014, "step": 3500 }, { "epoch": 0.42, "learning_rate": 2.7515776844631076e-06, "loss": 0.001, "step": 3501 }, { "epoch": 0.42, "learning_rate": 2.751505698860228e-06, "loss": 0.0583, "step": 3502 }, { "epoch": 0.42, "learning_rate": 2.7514337132573488e-06, "loss": 0.001, "step": 3503 }, { "epoch": 0.42, "learning_rate": 2.751361727654469e-06, "loss": 0.001, "step": 3504 }, { "epoch": 0.42, "learning_rate": 2.75128974205159e-06, "loss": 0.001, "step": 3505 }, { "epoch": 0.42, "learning_rate": 2.7512177564487103e-06, "loss": 0.001, "step": 3506 }, { "epoch": 0.42, "learning_rate": 2.751145770845831e-06, "loss": 0.0026, "step": 3507 }, { "epoch": 0.42, "learning_rate": 2.7510737852429514e-06, "loss": 0.0009, "step": 3508 }, { "epoch": 0.42, "learning_rate": 2.751001799640072e-06, "loss": 0.7651, "step": 3509 }, { "epoch": 0.42, "learning_rate": 2.7509298140371925e-06, "loss": 0.0013, "step": 3510 }, { "epoch": 0.42, "learning_rate": 2.7508578284343133e-06, "loss": 0.0012, "step": 3511 }, { "epoch": 0.42, "learning_rate": 2.7507858428314337e-06, "loss": 0.0246, "step": 3512 }, { "epoch": 0.42, "learning_rate": 2.7507138572285544e-06, "loss": 0.0007, "step": 3513 }, { "epoch": 0.42, "learning_rate": 2.7506418716256748e-06, "loss": 0.0016, "step": 3514 }, { "epoch": 0.42, "learning_rate": 2.7505698860227956e-06, "loss": 0.0008, "step": 3515 }, { "epoch": 0.42, "learning_rate": 2.750497900419916e-06, "loss": 0.0011, "step": 3516 }, { "epoch": 0.42, "learning_rate": 2.7504259148170367e-06, "loss": 0.0009, "step": 3517 }, { "epoch": 0.42, "learning_rate": 2.7503539292141575e-06, "loss": 0.001, "step": 3518 }, { "epoch": 0.42, "learning_rate": 2.750281943611278e-06, "loss": 0.0008, "step": 3519 }, { "epoch": 0.42, "learning_rate": 2.7502099580083986e-06, "loss": 0.0007, "step": 3520 }, { "epoch": 0.42, "learning_rate": 2.750137972405519e-06, "loss": 0.0006, "step": 3521 }, { "epoch": 0.42, "learning_rate": 2.7500659868026397e-06, "loss": 0.0083, "step": 3522 }, { "epoch": 0.42, "learning_rate": 2.74999400119976e-06, "loss": 0.0009, "step": 3523 }, { "epoch": 0.42, "learning_rate": 2.749922015596881e-06, "loss": 0.0035, "step": 3524 }, { "epoch": 0.42, "learning_rate": 2.7498500299940012e-06, "loss": 0.058, "step": 3525 }, { "epoch": 0.42, "learning_rate": 2.7497780443911216e-06, "loss": 0.2238, "step": 3526 }, { "epoch": 0.42, "learning_rate": 2.7497060587882424e-06, "loss": 0.0008, "step": 3527 }, { "epoch": 0.42, "learning_rate": 2.7496340731853627e-06, "loss": 0.0017, "step": 3528 }, { "epoch": 0.42, "learning_rate": 2.7495620875824835e-06, "loss": 0.3968, "step": 3529 }, { "epoch": 0.42, "learning_rate": 2.749490101979604e-06, "loss": 0.6939, "step": 3530 }, { "epoch": 0.42, "learning_rate": 2.7494181163767246e-06, "loss": 0.8034, "step": 3531 }, { "epoch": 0.42, "learning_rate": 2.7493461307738454e-06, "loss": 0.1257, "step": 3532 }, { "epoch": 0.42, "learning_rate": 2.749274145170966e-06, "loss": 0.0008, "step": 3533 }, { "epoch": 0.42, "learning_rate": 2.7492021595680865e-06, "loss": 0.001, "step": 3534 }, { "epoch": 0.42, "learning_rate": 2.7491301739652073e-06, "loss": 0.0013, "step": 3535 }, { "epoch": 0.42, "learning_rate": 2.7490581883623277e-06, "loss": 0.0017, "step": 3536 }, { "epoch": 0.42, "learning_rate": 2.7489862027594485e-06, "loss": 0.0007, "step": 3537 }, { "epoch": 0.42, "learning_rate": 2.748914217156569e-06, "loss": 0.0008, "step": 3538 }, { "epoch": 0.42, "learning_rate": 2.748842231553689e-06, "loss": 0.0297, "step": 3539 }, { "epoch": 0.42, "learning_rate": 2.74877024595081e-06, "loss": 0.0005, "step": 3540 }, { "epoch": 0.42, "learning_rate": 2.7486982603479303e-06, "loss": 0.9577, "step": 3541 }, { "epoch": 0.42, "learning_rate": 2.748626274745051e-06, "loss": 0.0043, "step": 3542 }, { "epoch": 0.42, "learning_rate": 2.7485542891421714e-06, "loss": 0.0007, "step": 3543 }, { "epoch": 0.42, "learning_rate": 2.7484823035392922e-06, "loss": 0.4004, "step": 3544 }, { "epoch": 0.42, "learning_rate": 2.7484103179364126e-06, "loss": 0.0009, "step": 3545 }, { "epoch": 0.42, "learning_rate": 2.7483383323335333e-06, "loss": 0.0033, "step": 3546 }, { "epoch": 0.43, "learning_rate": 2.7482663467306537e-06, "loss": 0.0007, "step": 3547 }, { "epoch": 0.43, "learning_rate": 2.7481943611277745e-06, "loss": 0.0014, "step": 3548 }, { "epoch": 0.43, "learning_rate": 2.7481223755248953e-06, "loss": 0.0013, "step": 3549 }, { "epoch": 0.43, "learning_rate": 2.7480503899220156e-06, "loss": 0.0166, "step": 3550 }, { "epoch": 0.43, "learning_rate": 2.7479784043191364e-06, "loss": 0.8432, "step": 3551 }, { "epoch": 0.43, "learning_rate": 2.7479064187162567e-06, "loss": 0.0012, "step": 3552 }, { "epoch": 0.43, "learning_rate": 2.7478344331133775e-06, "loss": 0.2555, "step": 3553 }, { "epoch": 0.43, "learning_rate": 2.747762447510498e-06, "loss": 0.0027, "step": 3554 }, { "epoch": 0.43, "learning_rate": 2.7476904619076187e-06, "loss": 0.0008, "step": 3555 }, { "epoch": 0.43, "learning_rate": 2.747618476304739e-06, "loss": 0.0023, "step": 3556 }, { "epoch": 0.43, "learning_rate": 2.74754649070186e-06, "loss": 0.8836, "step": 3557 }, { "epoch": 0.43, "learning_rate": 2.74747450509898e-06, "loss": 0.0087, "step": 3558 }, { "epoch": 0.43, "learning_rate": 2.747402519496101e-06, "loss": 0.0035, "step": 3559 }, { "epoch": 0.43, "learning_rate": 2.7473305338932213e-06, "loss": 0.001, "step": 3560 }, { "epoch": 0.43, "learning_rate": 2.7472585482903416e-06, "loss": 0.0329, "step": 3561 }, { "epoch": 0.43, "learning_rate": 2.7471865626874624e-06, "loss": 0.0006, "step": 3562 }, { "epoch": 0.43, "learning_rate": 2.747114577084583e-06, "loss": 0.57, "step": 3563 }, { "epoch": 0.43, "learning_rate": 2.747042591481704e-06, "loss": 0.1806, "step": 3564 }, { "epoch": 0.43, "learning_rate": 2.7469706058788243e-06, "loss": 0.001, "step": 3565 }, { "epoch": 0.43, "learning_rate": 2.746898620275945e-06, "loss": 0.0011, "step": 3566 }, { "epoch": 0.43, "learning_rate": 2.7468266346730655e-06, "loss": 0.0008, "step": 3567 }, { "epoch": 0.43, "learning_rate": 2.7467546490701862e-06, "loss": 0.8969, "step": 3568 }, { "epoch": 0.43, "learning_rate": 2.7466826634673066e-06, "loss": 0.0008, "step": 3569 }, { "epoch": 0.43, "learning_rate": 2.7466106778644274e-06, "loss": 0.0013, "step": 3570 }, { "epoch": 0.43, "learning_rate": 2.7465386922615477e-06, "loss": 0.0011, "step": 3571 }, { "epoch": 0.43, "learning_rate": 2.746466706658668e-06, "loss": 0.001, "step": 3572 }, { "epoch": 0.43, "learning_rate": 2.746394721055789e-06, "loss": 0.0012, "step": 3573 }, { "epoch": 0.43, "learning_rate": 2.7463227354529092e-06, "loss": 0.0009, "step": 3574 }, { "epoch": 0.43, "learning_rate": 2.74625074985003e-06, "loss": 0.0122, "step": 3575 }, { "epoch": 0.43, "learning_rate": 2.7461787642471504e-06, "loss": 0.0018, "step": 3576 }, { "epoch": 0.43, "learning_rate": 2.746106778644271e-06, "loss": 0.0012, "step": 3577 }, { "epoch": 0.43, "learning_rate": 2.746034793041392e-06, "loss": 0.001, "step": 3578 }, { "epoch": 0.43, "learning_rate": 2.7459628074385123e-06, "loss": 0.0033, "step": 3579 }, { "epoch": 0.43, "learning_rate": 2.745890821835633e-06, "loss": 0.0008, "step": 3580 }, { "epoch": 0.43, "learning_rate": 2.745818836232754e-06, "loss": 0.067, "step": 3581 }, { "epoch": 0.43, "learning_rate": 2.745746850629874e-06, "loss": 0.7763, "step": 3582 }, { "epoch": 0.43, "learning_rate": 2.745674865026995e-06, "loss": 0.0025, "step": 3583 }, { "epoch": 0.43, "learning_rate": 2.7456028794241153e-06, "loss": 0.1768, "step": 3584 }, { "epoch": 0.43, "learning_rate": 2.7455308938212357e-06, "loss": 0.4299, "step": 3585 }, { "epoch": 0.43, "learning_rate": 2.7454589082183564e-06, "loss": 0.0007, "step": 3586 }, { "epoch": 0.43, "learning_rate": 2.745386922615477e-06, "loss": 0.5649, "step": 3587 }, { "epoch": 0.43, "learning_rate": 2.7453149370125976e-06, "loss": 0.0006, "step": 3588 }, { "epoch": 0.43, "learning_rate": 2.745242951409718e-06, "loss": 0.0011, "step": 3589 }, { "epoch": 0.43, "learning_rate": 2.7451709658068387e-06, "loss": 0.0019, "step": 3590 }, { "epoch": 0.43, "learning_rate": 2.745098980203959e-06, "loss": 0.0047, "step": 3591 }, { "epoch": 0.43, "learning_rate": 2.74502699460108e-06, "loss": 0.0009, "step": 3592 }, { "epoch": 0.43, "learning_rate": 2.7449550089982e-06, "loss": 0.0012, "step": 3593 }, { "epoch": 0.43, "learning_rate": 2.744883023395321e-06, "loss": 0.0008, "step": 3594 }, { "epoch": 0.43, "learning_rate": 2.7448110377924418e-06, "loss": 0.0007, "step": 3595 }, { "epoch": 0.43, "learning_rate": 2.744739052189562e-06, "loss": 0.0019, "step": 3596 }, { "epoch": 0.43, "learning_rate": 2.744667066586683e-06, "loss": 0.0008, "step": 3597 }, { "epoch": 0.43, "learning_rate": 2.7445950809838032e-06, "loss": 0.0008, "step": 3598 }, { "epoch": 0.43, "learning_rate": 2.744523095380924e-06, "loss": 0.0007, "step": 3599 }, { "epoch": 0.43, "learning_rate": 2.7444511097780444e-06, "loss": 0.0016, "step": 3600 }, { "epoch": 0.43, "learning_rate": 2.744379124175165e-06, "loss": 0.0012, "step": 3601 }, { "epoch": 0.43, "learning_rate": 2.7443071385722855e-06, "loss": 0.7691, "step": 3602 }, { "epoch": 0.43, "learning_rate": 2.7442351529694063e-06, "loss": 0.0009, "step": 3603 }, { "epoch": 0.43, "learning_rate": 2.7441631673665266e-06, "loss": 0.0023, "step": 3604 }, { "epoch": 0.43, "learning_rate": 2.7440911817636474e-06, "loss": 0.0134, "step": 3605 }, { "epoch": 0.43, "learning_rate": 2.7440191961607678e-06, "loss": 0.9875, "step": 3606 }, { "epoch": 0.43, "learning_rate": 2.743947210557888e-06, "loss": 0.0074, "step": 3607 }, { "epoch": 0.43, "learning_rate": 2.743875224955009e-06, "loss": 0.0007, "step": 3608 }, { "epoch": 0.43, "learning_rate": 2.7438032393521297e-06, "loss": 0.002, "step": 3609 }, { "epoch": 0.43, "learning_rate": 2.7437312537492505e-06, "loss": 0.0236, "step": 3610 }, { "epoch": 0.43, "learning_rate": 2.743659268146371e-06, "loss": 0.0033, "step": 3611 }, { "epoch": 0.43, "learning_rate": 2.7435872825434916e-06, "loss": 0.0009, "step": 3612 }, { "epoch": 0.43, "learning_rate": 2.743515296940612e-06, "loss": 0.0007, "step": 3613 }, { "epoch": 0.43, "learning_rate": 2.7434433113377327e-06, "loss": 0.0025, "step": 3614 }, { "epoch": 0.43, "learning_rate": 2.743371325734853e-06, "loss": 0.0011, "step": 3615 }, { "epoch": 0.43, "learning_rate": 2.743299340131974e-06, "loss": 0.0009, "step": 3616 }, { "epoch": 0.43, "learning_rate": 2.7432273545290942e-06, "loss": 0.0014, "step": 3617 }, { "epoch": 0.43, "learning_rate": 2.743155368926215e-06, "loss": 0.3457, "step": 3618 }, { "epoch": 0.43, "learning_rate": 2.7430833833233354e-06, "loss": 0.001, "step": 3619 }, { "epoch": 0.43, "learning_rate": 2.7430113977204557e-06, "loss": 0.0025, "step": 3620 }, { "epoch": 0.43, "learning_rate": 2.7429394121175765e-06, "loss": 0.0012, "step": 3621 }, { "epoch": 0.43, "learning_rate": 2.742867426514697e-06, "loss": 0.0012, "step": 3622 }, { "epoch": 0.43, "learning_rate": 2.7427954409118176e-06, "loss": 0.0013, "step": 3623 }, { "epoch": 0.43, "learning_rate": 2.7427234553089384e-06, "loss": 0.1706, "step": 3624 }, { "epoch": 0.43, "learning_rate": 2.7426514697060588e-06, "loss": 0.0011, "step": 3625 }, { "epoch": 0.43, "learning_rate": 2.7425794841031795e-06, "loss": 0.0019, "step": 3626 }, { "epoch": 0.43, "learning_rate": 2.7425074985003003e-06, "loss": 0.0009, "step": 3627 }, { "epoch": 0.43, "learning_rate": 2.7424355128974207e-06, "loss": 0.0048, "step": 3628 }, { "epoch": 0.43, "learning_rate": 2.7423635272945414e-06, "loss": 0.0011, "step": 3629 }, { "epoch": 0.43, "learning_rate": 2.742291541691662e-06, "loss": 0.6249, "step": 3630 }, { "epoch": 0.44, "learning_rate": 2.742219556088782e-06, "loss": 0.0008, "step": 3631 }, { "epoch": 0.44, "learning_rate": 2.742147570485903e-06, "loss": 0.0011, "step": 3632 }, { "epoch": 0.44, "learning_rate": 2.7420755848830233e-06, "loss": 0.0454, "step": 3633 }, { "epoch": 0.44, "learning_rate": 2.742003599280144e-06, "loss": 0.0014, "step": 3634 }, { "epoch": 0.44, "learning_rate": 2.7419316136772644e-06, "loss": 0.0031, "step": 3635 }, { "epoch": 0.44, "learning_rate": 2.741859628074385e-06, "loss": 0.0107, "step": 3636 }, { "epoch": 0.44, "learning_rate": 2.7417876424715056e-06, "loss": 0.0008, "step": 3637 }, { "epoch": 0.44, "learning_rate": 2.7417156568686263e-06, "loss": 0.0007, "step": 3638 }, { "epoch": 0.44, "learning_rate": 2.7416436712657467e-06, "loss": 0.001, "step": 3639 }, { "epoch": 0.44, "learning_rate": 2.7415716856628675e-06, "loss": 0.8414, "step": 3640 }, { "epoch": 0.44, "learning_rate": 2.7414997000599883e-06, "loss": 0.0007, "step": 3641 }, { "epoch": 0.44, "learning_rate": 2.7414277144571086e-06, "loss": 0.0008, "step": 3642 }, { "epoch": 0.44, "learning_rate": 2.7413557288542294e-06, "loss": 0.0009, "step": 3643 }, { "epoch": 0.44, "learning_rate": 2.7412837432513497e-06, "loss": 0.0012, "step": 3644 }, { "epoch": 0.44, "learning_rate": 2.7412117576484705e-06, "loss": 0.0036, "step": 3645 }, { "epoch": 0.44, "learning_rate": 2.741139772045591e-06, "loss": 0.0015, "step": 3646 }, { "epoch": 0.44, "learning_rate": 2.7410677864427117e-06, "loss": 0.0011, "step": 3647 }, { "epoch": 0.44, "learning_rate": 2.740995800839832e-06, "loss": 0.001, "step": 3648 }, { "epoch": 0.44, "learning_rate": 2.7409238152369528e-06, "loss": 0.0208, "step": 3649 }, { "epoch": 0.44, "learning_rate": 2.740851829634073e-06, "loss": 0.0248, "step": 3650 }, { "epoch": 0.44, "learning_rate": 2.740779844031194e-06, "loss": 0.0006, "step": 3651 }, { "epoch": 0.44, "learning_rate": 2.7407078584283143e-06, "loss": 0.0008, "step": 3652 }, { "epoch": 0.44, "learning_rate": 2.7406358728254346e-06, "loss": 0.0014, "step": 3653 }, { "epoch": 0.44, "learning_rate": 2.7405638872225554e-06, "loss": 0.001, "step": 3654 }, { "epoch": 0.44, "learning_rate": 2.740491901619676e-06, "loss": 0.001, "step": 3655 }, { "epoch": 0.44, "learning_rate": 2.740419916016797e-06, "loss": 0.0395, "step": 3656 }, { "epoch": 0.44, "learning_rate": 2.7403479304139173e-06, "loss": 0.0007, "step": 3657 }, { "epoch": 0.44, "learning_rate": 2.740275944811038e-06, "loss": 0.5178, "step": 3658 }, { "epoch": 0.44, "learning_rate": 2.7402039592081585e-06, "loss": 0.0039, "step": 3659 }, { "epoch": 0.44, "learning_rate": 2.7401319736052792e-06, "loss": 0.0008, "step": 3660 }, { "epoch": 0.44, "learning_rate": 2.7400599880023996e-06, "loss": 0.0006, "step": 3661 }, { "epoch": 0.44, "learning_rate": 2.7399880023995204e-06, "loss": 0.0683, "step": 3662 }, { "epoch": 0.44, "learning_rate": 2.7399160167966407e-06, "loss": 0.0394, "step": 3663 }, { "epoch": 0.44, "learning_rate": 2.7398440311937615e-06, "loss": 0.0017, "step": 3664 }, { "epoch": 0.44, "learning_rate": 2.739772045590882e-06, "loss": 0.0018, "step": 3665 }, { "epoch": 0.44, "learning_rate": 2.739700059988002e-06, "loss": 0.0019, "step": 3666 }, { "epoch": 0.44, "learning_rate": 2.739628074385123e-06, "loss": 0.001, "step": 3667 }, { "epoch": 0.44, "learning_rate": 2.7395560887822433e-06, "loss": 0.0013, "step": 3668 }, { "epoch": 0.44, "learning_rate": 2.739484103179364e-06, "loss": 0.001, "step": 3669 }, { "epoch": 0.44, "learning_rate": 2.7394121175764845e-06, "loss": 0.0007, "step": 3670 }, { "epoch": 0.44, "learning_rate": 2.7393401319736053e-06, "loss": 0.0008, "step": 3671 }, { "epoch": 0.44, "learning_rate": 2.739268146370726e-06, "loss": 0.0277, "step": 3672 }, { "epoch": 0.44, "learning_rate": 2.739196160767847e-06, "loss": 0.0009, "step": 3673 }, { "epoch": 0.44, "learning_rate": 2.739124175164967e-06, "loss": 0.0014, "step": 3674 }, { "epoch": 0.44, "learning_rate": 2.739052189562088e-06, "loss": 0.0006, "step": 3675 }, { "epoch": 0.44, "learning_rate": 2.7389802039592083e-06, "loss": 0.0123, "step": 3676 }, { "epoch": 0.44, "learning_rate": 2.7389082183563287e-06, "loss": 0.001, "step": 3677 }, { "epoch": 0.44, "learning_rate": 2.7388362327534494e-06, "loss": 0.001, "step": 3678 }, { "epoch": 0.44, "learning_rate": 2.73876424715057e-06, "loss": 0.0008, "step": 3679 }, { "epoch": 0.44, "learning_rate": 2.7386922615476906e-06, "loss": 0.001, "step": 3680 }, { "epoch": 0.44, "learning_rate": 2.738620275944811e-06, "loss": 0.0009, "step": 3681 }, { "epoch": 0.44, "learning_rate": 2.7385482903419317e-06, "loss": 0.0009, "step": 3682 }, { "epoch": 0.44, "learning_rate": 2.738476304739052e-06, "loss": 0.8559, "step": 3683 }, { "epoch": 0.44, "learning_rate": 2.738404319136173e-06, "loss": 0.001, "step": 3684 }, { "epoch": 0.44, "learning_rate": 2.738332333533293e-06, "loss": 0.0007, "step": 3685 }, { "epoch": 0.44, "learning_rate": 2.738260347930414e-06, "loss": 0.7709, "step": 3686 }, { "epoch": 0.44, "learning_rate": 2.7381883623275347e-06, "loss": 0.0006, "step": 3687 }, { "epoch": 0.44, "learning_rate": 2.738116376724655e-06, "loss": 0.1635, "step": 3688 }, { "epoch": 0.44, "learning_rate": 2.738044391121776e-06, "loss": 0.001, "step": 3689 }, { "epoch": 0.44, "learning_rate": 2.7379724055188962e-06, "loss": 0.0007, "step": 3690 }, { "epoch": 0.44, "learning_rate": 2.737900419916017e-06, "loss": 0.0011, "step": 3691 }, { "epoch": 0.44, "learning_rate": 2.7378284343131374e-06, "loss": 0.0006, "step": 3692 }, { "epoch": 0.44, "learning_rate": 2.737756448710258e-06, "loss": 0.0014, "step": 3693 }, { "epoch": 0.44, "learning_rate": 2.7376844631073785e-06, "loss": 0.0011, "step": 3694 }, { "epoch": 0.44, "learning_rate": 2.7376124775044993e-06, "loss": 0.0008, "step": 3695 }, { "epoch": 0.44, "learning_rate": 2.7375404919016196e-06, "loss": 0.7078, "step": 3696 }, { "epoch": 0.44, "learning_rate": 2.7374685062987404e-06, "loss": 0.8089, "step": 3697 }, { "epoch": 0.44, "learning_rate": 2.7373965206958608e-06, "loss": 0.0451, "step": 3698 }, { "epoch": 0.44, "learning_rate": 2.7373245350929815e-06, "loss": 0.0007, "step": 3699 }, { "epoch": 0.44, "learning_rate": 2.737252549490102e-06, "loss": 0.0007, "step": 3700 }, { "epoch": 0.44, "learning_rate": 2.7371805638872227e-06, "loss": 0.0009, "step": 3701 }, { "epoch": 0.44, "learning_rate": 2.737108578284343e-06, "loss": 0.0348, "step": 3702 }, { "epoch": 0.44, "learning_rate": 2.737036592681464e-06, "loss": 0.0012, "step": 3703 }, { "epoch": 0.44, "learning_rate": 2.7369646070785846e-06, "loss": 0.4434, "step": 3704 }, { "epoch": 0.44, "learning_rate": 2.736892621475705e-06, "loss": 0.0012, "step": 3705 }, { "epoch": 0.44, "learning_rate": 2.7368206358728257e-06, "loss": 0.0008, "step": 3706 }, { "epoch": 0.44, "learning_rate": 2.736748650269946e-06, "loss": 0.0007, "step": 3707 }, { "epoch": 0.44, "learning_rate": 2.736676664667067e-06, "loss": 0.6997, "step": 3708 }, { "epoch": 0.44, "learning_rate": 2.7366046790641872e-06, "loss": 0.0009, "step": 3709 }, { "epoch": 0.44, "learning_rate": 2.736532693461308e-06, "loss": 0.0007, "step": 3710 }, { "epoch": 0.44, "learning_rate": 2.7364607078584284e-06, "loss": 0.0008, "step": 3711 }, { "epoch": 0.44, "learning_rate": 2.7363887222555487e-06, "loss": 0.0008, "step": 3712 }, { "epoch": 0.44, "learning_rate": 2.7363167366526695e-06, "loss": 0.0007, "step": 3713 }, { "epoch": 0.45, "learning_rate": 2.73624475104979e-06, "loss": 0.001, "step": 3714 }, { "epoch": 0.45, "learning_rate": 2.7361727654469106e-06, "loss": 0.0008, "step": 3715 }, { "epoch": 0.45, "learning_rate": 2.736100779844031e-06, "loss": 0.0009, "step": 3716 }, { "epoch": 0.45, "learning_rate": 2.7360287942411518e-06, "loss": 0.7457, "step": 3717 }, { "epoch": 0.45, "learning_rate": 2.7359568086382725e-06, "loss": 0.0011, "step": 3718 }, { "epoch": 0.45, "learning_rate": 2.7358848230353933e-06, "loss": 0.0018, "step": 3719 }, { "epoch": 0.45, "learning_rate": 2.7358128374325137e-06, "loss": 0.0008, "step": 3720 }, { "epoch": 0.45, "learning_rate": 2.7357408518296344e-06, "loss": 0.0008, "step": 3721 }, { "epoch": 0.45, "learning_rate": 2.735668866226755e-06, "loss": 0.0006, "step": 3722 }, { "epoch": 0.45, "learning_rate": 2.735596880623875e-06, "loss": 0.0377, "step": 3723 }, { "epoch": 0.45, "learning_rate": 2.735524895020996e-06, "loss": 0.0009, "step": 3724 }, { "epoch": 0.45, "learning_rate": 2.7354529094181163e-06, "loss": 0.0014, "step": 3725 }, { "epoch": 0.45, "learning_rate": 2.735380923815237e-06, "loss": 0.0031, "step": 3726 }, { "epoch": 0.45, "learning_rate": 2.7353089382123574e-06, "loss": 0.0007, "step": 3727 }, { "epoch": 0.45, "learning_rate": 2.735236952609478e-06, "loss": 0.0166, "step": 3728 }, { "epoch": 0.45, "learning_rate": 2.7351649670065986e-06, "loss": 0.0006, "step": 3729 }, { "epoch": 0.45, "learning_rate": 2.7350929814037193e-06, "loss": 0.8868, "step": 3730 }, { "epoch": 0.45, "learning_rate": 2.7350209958008397e-06, "loss": 0.0006, "step": 3731 }, { "epoch": 0.45, "learning_rate": 2.7349490101979605e-06, "loss": 0.0007, "step": 3732 }, { "epoch": 0.45, "learning_rate": 2.7348770245950812e-06, "loss": 0.0008, "step": 3733 }, { "epoch": 0.45, "learning_rate": 2.7348050389922016e-06, "loss": 0.0007, "step": 3734 }, { "epoch": 0.45, "learning_rate": 2.7347330533893224e-06, "loss": 0.0009, "step": 3735 }, { "epoch": 0.45, "learning_rate": 2.7346610677864427e-06, "loss": 0.0014, "step": 3736 }, { "epoch": 0.45, "learning_rate": 2.7345890821835635e-06, "loss": 0.6936, "step": 3737 }, { "epoch": 0.45, "learning_rate": 2.734517096580684e-06, "loss": 0.0014, "step": 3738 }, { "epoch": 0.45, "learning_rate": 2.7344451109778046e-06, "loss": 0.001, "step": 3739 }, { "epoch": 0.45, "learning_rate": 2.734373125374925e-06, "loss": 0.001, "step": 3740 }, { "epoch": 0.45, "learning_rate": 2.7343011397720458e-06, "loss": 0.0009, "step": 3741 }, { "epoch": 0.45, "learning_rate": 2.734229154169166e-06, "loss": 0.003, "step": 3742 }, { "epoch": 0.45, "learning_rate": 2.734157168566287e-06, "loss": 0.0008, "step": 3743 }, { "epoch": 0.45, "learning_rate": 2.7340851829634073e-06, "loss": 0.001, "step": 3744 }, { "epoch": 0.45, "learning_rate": 2.734013197360528e-06, "loss": 0.0023, "step": 3745 }, { "epoch": 0.45, "learning_rate": 2.7339412117576484e-06, "loss": 0.0007, "step": 3746 }, { "epoch": 0.45, "learning_rate": 2.733869226154769e-06, "loss": 0.001, "step": 3747 }, { "epoch": 0.45, "learning_rate": 2.7337972405518895e-06, "loss": 0.0006, "step": 3748 }, { "epoch": 0.45, "learning_rate": 2.7337252549490103e-06, "loss": 0.0223, "step": 3749 }, { "epoch": 0.45, "learning_rate": 2.733653269346131e-06, "loss": 0.0014, "step": 3750 }, { "epoch": 0.45, "learning_rate": 2.7335812837432514e-06, "loss": 0.0006, "step": 3751 }, { "epoch": 0.45, "learning_rate": 2.7335092981403722e-06, "loss": 0.0007, "step": 3752 }, { "epoch": 0.45, "learning_rate": 2.7334373125374926e-06, "loss": 0.8089, "step": 3753 }, { "epoch": 0.45, "learning_rate": 2.7333653269346134e-06, "loss": 0.001, "step": 3754 }, { "epoch": 0.45, "learning_rate": 2.7332933413317337e-06, "loss": 0.0029, "step": 3755 }, { "epoch": 0.45, "learning_rate": 2.7332213557288545e-06, "loss": 0.0021, "step": 3756 }, { "epoch": 0.45, "learning_rate": 2.733149370125975e-06, "loss": 0.001, "step": 3757 }, { "epoch": 0.45, "learning_rate": 2.733077384523095e-06, "loss": 0.1572, "step": 3758 }, { "epoch": 0.45, "learning_rate": 2.733005398920216e-06, "loss": 0.0009, "step": 3759 }, { "epoch": 0.45, "learning_rate": 2.7329334133173363e-06, "loss": 0.001, "step": 3760 }, { "epoch": 0.45, "learning_rate": 2.732861427714457e-06, "loss": 0.0008, "step": 3761 }, { "epoch": 0.45, "learning_rate": 2.7327894421115775e-06, "loss": 0.0009, "step": 3762 }, { "epoch": 0.45, "learning_rate": 2.7327174565086982e-06, "loss": 0.0008, "step": 3763 }, { "epoch": 0.45, "learning_rate": 2.732645470905819e-06, "loss": 0.003, "step": 3764 }, { "epoch": 0.45, "learning_rate": 2.73257348530294e-06, "loss": 0.0008, "step": 3765 }, { "epoch": 0.45, "learning_rate": 2.73250149970006e-06, "loss": 0.0009, "step": 3766 }, { "epoch": 0.45, "learning_rate": 2.732429514097181e-06, "loss": 0.0613, "step": 3767 }, { "epoch": 0.45, "learning_rate": 2.7323575284943013e-06, "loss": 0.7743, "step": 3768 }, { "epoch": 0.45, "learning_rate": 2.7322855428914216e-06, "loss": 0.0006, "step": 3769 }, { "epoch": 0.45, "learning_rate": 2.7322135572885424e-06, "loss": 0.0237, "step": 3770 }, { "epoch": 0.45, "learning_rate": 2.7321415716856628e-06, "loss": 0.0025, "step": 3771 }, { "epoch": 0.45, "learning_rate": 2.7320695860827836e-06, "loss": 0.0138, "step": 3772 }, { "epoch": 0.45, "learning_rate": 2.731997600479904e-06, "loss": 0.8269, "step": 3773 }, { "epoch": 0.45, "learning_rate": 2.7319256148770247e-06, "loss": 0.0012, "step": 3774 }, { "epoch": 0.45, "learning_rate": 2.731853629274145e-06, "loss": 0.7612, "step": 3775 }, { "epoch": 0.45, "learning_rate": 2.731781643671266e-06, "loss": 0.0132, "step": 3776 }, { "epoch": 0.45, "learning_rate": 2.731709658068386e-06, "loss": 0.001, "step": 3777 }, { "epoch": 0.45, "learning_rate": 2.731637672465507e-06, "loss": 0.001, "step": 3778 }, { "epoch": 0.45, "learning_rate": 2.7315656868626277e-06, "loss": 0.9108, "step": 3779 }, { "epoch": 0.45, "learning_rate": 2.731493701259748e-06, "loss": 0.001, "step": 3780 }, { "epoch": 0.45, "learning_rate": 2.731421715656869e-06, "loss": 0.0009, "step": 3781 }, { "epoch": 0.45, "learning_rate": 2.7313497300539892e-06, "loss": 0.0122, "step": 3782 }, { "epoch": 0.45, "learning_rate": 2.73127774445111e-06, "loss": 0.0008, "step": 3783 }, { "epoch": 0.45, "learning_rate": 2.7312057588482304e-06, "loss": 0.0015, "step": 3784 }, { "epoch": 0.45, "learning_rate": 2.731133773245351e-06, "loss": 0.001, "step": 3785 }, { "epoch": 0.45, "learning_rate": 2.7310617876424715e-06, "loss": 0.0008, "step": 3786 }, { "epoch": 0.45, "learning_rate": 2.7309898020395923e-06, "loss": 0.0008, "step": 3787 }, { "epoch": 0.45, "learning_rate": 2.7309178164367126e-06, "loss": 0.0009, "step": 3788 }, { "epoch": 0.45, "learning_rate": 2.7308458308338334e-06, "loss": 0.0058, "step": 3789 }, { "epoch": 0.45, "learning_rate": 2.7307738452309538e-06, "loss": 0.0025, "step": 3790 }, { "epoch": 0.45, "learning_rate": 2.7307018596280745e-06, "loss": 0.0014, "step": 3791 }, { "epoch": 0.45, "learning_rate": 2.730629874025195e-06, "loss": 0.0035, "step": 3792 }, { "epoch": 0.45, "learning_rate": 2.7305578884223157e-06, "loss": 0.001, "step": 3793 }, { "epoch": 0.45, "learning_rate": 2.730485902819436e-06, "loss": 0.6326, "step": 3794 }, { "epoch": 0.45, "learning_rate": 2.730413917216557e-06, "loss": 0.0012, "step": 3795 }, { "epoch": 0.45, "learning_rate": 2.7303419316136776e-06, "loss": 0.0019, "step": 3796 }, { "epoch": 0.46, "learning_rate": 2.730269946010798e-06, "loss": 0.0007, "step": 3797 }, { "epoch": 0.46, "learning_rate": 2.7301979604079187e-06, "loss": 0.0011, "step": 3798 }, { "epoch": 0.46, "learning_rate": 2.730125974805039e-06, "loss": 0.0009, "step": 3799 }, { "epoch": 0.46, "learning_rate": 2.73005398920216e-06, "loss": 0.0009, "step": 3800 }, { "epoch": 0.46, "learning_rate": 2.72998200359928e-06, "loss": 0.0009, "step": 3801 }, { "epoch": 0.46, "learning_rate": 2.729910017996401e-06, "loss": 0.9349, "step": 3802 }, { "epoch": 0.46, "learning_rate": 2.7298380323935213e-06, "loss": 0.0007, "step": 3803 }, { "epoch": 0.46, "learning_rate": 2.7297660467906417e-06, "loss": 0.0487, "step": 3804 }, { "epoch": 0.46, "learning_rate": 2.7296940611877625e-06, "loss": 0.001, "step": 3805 }, { "epoch": 0.46, "learning_rate": 2.729622075584883e-06, "loss": 0.3874, "step": 3806 }, { "epoch": 0.46, "learning_rate": 2.7295500899820036e-06, "loss": 0.0007, "step": 3807 }, { "epoch": 0.46, "learning_rate": 2.729478104379124e-06, "loss": 0.0008, "step": 3808 }, { "epoch": 0.46, "learning_rate": 2.7294061187762447e-06, "loss": 0.001, "step": 3809 }, { "epoch": 0.46, "learning_rate": 2.7293341331733655e-06, "loss": 1.3313, "step": 3810 }, { "epoch": 0.46, "learning_rate": 2.729262147570486e-06, "loss": 0.0013, "step": 3811 }, { "epoch": 0.46, "learning_rate": 2.7291901619676067e-06, "loss": 0.0006, "step": 3812 }, { "epoch": 0.46, "learning_rate": 2.7291181763647274e-06, "loss": 0.001, "step": 3813 }, { "epoch": 0.46, "learning_rate": 2.729046190761848e-06, "loss": 0.0101, "step": 3814 }, { "epoch": 0.46, "learning_rate": 2.728974205158968e-06, "loss": 0.0227, "step": 3815 }, { "epoch": 0.46, "learning_rate": 2.728902219556089e-06, "loss": 0.0012, "step": 3816 }, { "epoch": 0.46, "learning_rate": 2.7288302339532093e-06, "loss": 0.0098, "step": 3817 }, { "epoch": 0.46, "learning_rate": 2.72875824835033e-06, "loss": 0.0008, "step": 3818 }, { "epoch": 0.46, "learning_rate": 2.7286862627474504e-06, "loss": 0.0012, "step": 3819 }, { "epoch": 0.46, "learning_rate": 2.728614277144571e-06, "loss": 0.0008, "step": 3820 }, { "epoch": 0.46, "learning_rate": 2.7285422915416915e-06, "loss": 0.0008, "step": 3821 }, { "epoch": 0.46, "learning_rate": 2.7284703059388123e-06, "loss": 0.0011, "step": 3822 }, { "epoch": 0.46, "learning_rate": 2.7283983203359327e-06, "loss": 0.0008, "step": 3823 }, { "epoch": 0.46, "learning_rate": 2.7283263347330535e-06, "loss": 0.3289, "step": 3824 }, { "epoch": 0.46, "learning_rate": 2.728254349130174e-06, "loss": 0.0009, "step": 3825 }, { "epoch": 0.46, "learning_rate": 2.7281823635272946e-06, "loss": 0.0008, "step": 3826 }, { "epoch": 0.46, "learning_rate": 2.7281103779244154e-06, "loss": 0.0007, "step": 3827 }, { "epoch": 0.46, "learning_rate": 2.7280383923215357e-06, "loss": 0.0007, "step": 3828 }, { "epoch": 0.46, "learning_rate": 2.7279664067186565e-06, "loss": 0.0009, "step": 3829 }, { "epoch": 0.46, "learning_rate": 2.727894421115777e-06, "loss": 0.0012, "step": 3830 }, { "epoch": 0.46, "learning_rate": 2.7278224355128976e-06, "loss": 0.0008, "step": 3831 }, { "epoch": 0.46, "learning_rate": 2.727750449910018e-06, "loss": 0.0006, "step": 3832 }, { "epoch": 0.46, "learning_rate": 2.7276784643071388e-06, "loss": 0.0007, "step": 3833 }, { "epoch": 0.46, "learning_rate": 2.727606478704259e-06, "loss": 0.0009, "step": 3834 }, { "epoch": 0.46, "learning_rate": 2.72753449310138e-06, "loss": 0.0009, "step": 3835 }, { "epoch": 0.46, "learning_rate": 2.7274625074985003e-06, "loss": 0.0011, "step": 3836 }, { "epoch": 0.46, "learning_rate": 2.727390521895621e-06, "loss": 0.1463, "step": 3837 }, { "epoch": 0.46, "learning_rate": 2.7273185362927414e-06, "loss": 0.007, "step": 3838 }, { "epoch": 0.46, "learning_rate": 2.7272465506898617e-06, "loss": 0.0017, "step": 3839 }, { "epoch": 0.46, "learning_rate": 2.7271745650869825e-06, "loss": 0.0007, "step": 3840 }, { "epoch": 0.46, "learning_rate": 2.7271025794841033e-06, "loss": 0.7799, "step": 3841 }, { "epoch": 0.46, "learning_rate": 2.727030593881224e-06, "loss": 0.0007, "step": 3842 }, { "epoch": 0.46, "learning_rate": 2.7269586082783444e-06, "loss": 0.0006, "step": 3843 }, { "epoch": 0.46, "learning_rate": 2.7268866226754652e-06, "loss": 0.0038, "step": 3844 }, { "epoch": 0.46, "learning_rate": 2.7268146370725856e-06, "loss": 0.0016, "step": 3845 }, { "epoch": 0.46, "learning_rate": 2.7267426514697063e-06, "loss": 0.0011, "step": 3846 }, { "epoch": 0.46, "learning_rate": 2.7266706658668267e-06, "loss": 0.0011, "step": 3847 }, { "epoch": 0.46, "learning_rate": 2.7265986802639475e-06, "loss": 0.0043, "step": 3848 }, { "epoch": 0.46, "learning_rate": 2.726526694661068e-06, "loss": 0.0157, "step": 3849 }, { "epoch": 0.46, "learning_rate": 2.726454709058188e-06, "loss": 0.1106, "step": 3850 }, { "epoch": 0.46, "learning_rate": 2.726382723455309e-06, "loss": 0.0008, "step": 3851 }, { "epoch": 0.46, "learning_rate": 2.7263107378524293e-06, "loss": 0.0007, "step": 3852 }, { "epoch": 0.46, "learning_rate": 2.72623875224955e-06, "loss": 0.001, "step": 3853 }, { "epoch": 0.46, "learning_rate": 2.7261667666466705e-06, "loss": 0.0009, "step": 3854 }, { "epoch": 0.46, "learning_rate": 2.7260947810437912e-06, "loss": 0.0007, "step": 3855 }, { "epoch": 0.46, "learning_rate": 2.726022795440912e-06, "loss": 0.0007, "step": 3856 }, { "epoch": 0.46, "learning_rate": 2.7259508098380324e-06, "loss": 0.0007, "step": 3857 }, { "epoch": 0.46, "learning_rate": 2.725878824235153e-06, "loss": 0.0014, "step": 3858 }, { "epoch": 0.46, "learning_rate": 2.725806838632274e-06, "loss": 0.0006, "step": 3859 }, { "epoch": 0.46, "learning_rate": 2.7257348530293943e-06, "loss": 0.0006, "step": 3860 }, { "epoch": 0.46, "learning_rate": 2.7256628674265146e-06, "loss": 0.0006, "step": 3861 }, { "epoch": 0.46, "learning_rate": 2.7255908818236354e-06, "loss": 0.0009, "step": 3862 }, { "epoch": 0.46, "learning_rate": 2.7255188962207558e-06, "loss": 0.001, "step": 3863 }, { "epoch": 0.46, "learning_rate": 2.7254469106178766e-06, "loss": 0.005, "step": 3864 }, { "epoch": 0.46, "learning_rate": 2.725374925014997e-06, "loss": 0.0012, "step": 3865 }, { "epoch": 0.46, "learning_rate": 2.7253029394121177e-06, "loss": 0.0066, "step": 3866 }, { "epoch": 0.46, "learning_rate": 2.725230953809238e-06, "loss": 0.3557, "step": 3867 }, { "epoch": 0.46, "learning_rate": 2.725158968206359e-06, "loss": 0.5902, "step": 3868 }, { "epoch": 0.46, "learning_rate": 2.725086982603479e-06, "loss": 0.0006, "step": 3869 }, { "epoch": 0.46, "learning_rate": 2.7250149970006e-06, "loss": 0.0009, "step": 3870 }, { "epoch": 0.46, "learning_rate": 2.7249430113977203e-06, "loss": 0.0682, "step": 3871 }, { "epoch": 0.46, "learning_rate": 2.724871025794841e-06, "loss": 0.5062, "step": 3872 }, { "epoch": 0.46, "learning_rate": 2.724799040191962e-06, "loss": 0.0006, "step": 3873 }, { "epoch": 0.46, "learning_rate": 2.7247270545890822e-06, "loss": 0.0007, "step": 3874 }, { "epoch": 0.46, "learning_rate": 2.724655068986203e-06, "loss": 0.0021, "step": 3875 }, { "epoch": 0.46, "learning_rate": 2.7245830833833234e-06, "loss": 0.0007, "step": 3876 }, { "epoch": 0.46, "learning_rate": 2.724511097780444e-06, "loss": 0.7701, "step": 3877 }, { "epoch": 0.46, "learning_rate": 2.7244391121775645e-06, "loss": 0.0909, "step": 3878 }, { "epoch": 0.46, "learning_rate": 2.7243671265746853e-06, "loss": 0.0228, "step": 3879 }, { "epoch": 0.46, "learning_rate": 2.7242951409718056e-06, "loss": 0.0006, "step": 3880 }, { "epoch": 0.47, "learning_rate": 2.7242231553689264e-06, "loss": 0.0006, "step": 3881 }, { "epoch": 0.47, "learning_rate": 2.7241511697660468e-06, "loss": 0.0007, "step": 3882 }, { "epoch": 0.47, "learning_rate": 2.7240791841631675e-06, "loss": 0.0007, "step": 3883 }, { "epoch": 0.47, "learning_rate": 2.724007198560288e-06, "loss": 0.0014, "step": 3884 }, { "epoch": 0.47, "learning_rate": 2.7239352129574082e-06, "loss": 0.0028, "step": 3885 }, { "epoch": 0.47, "learning_rate": 2.723863227354529e-06, "loss": 0.001, "step": 3886 }, { "epoch": 0.47, "learning_rate": 2.72379124175165e-06, "loss": 0.0007, "step": 3887 }, { "epoch": 0.47, "learning_rate": 2.7237192561487706e-06, "loss": 0.2997, "step": 3888 }, { "epoch": 0.47, "learning_rate": 2.723647270545891e-06, "loss": 0.6271, "step": 3889 }, { "epoch": 0.47, "learning_rate": 2.7235752849430117e-06, "loss": 0.3456, "step": 3890 }, { "epoch": 0.47, "learning_rate": 2.723503299340132e-06, "loss": 0.0009, "step": 3891 }, { "epoch": 0.47, "learning_rate": 2.723431313737253e-06, "loss": 0.0008, "step": 3892 }, { "epoch": 0.47, "learning_rate": 2.723359328134373e-06, "loss": 0.0006, "step": 3893 }, { "epoch": 0.47, "learning_rate": 2.723287342531494e-06, "loss": 0.0009, "step": 3894 }, { "epoch": 0.47, "learning_rate": 2.7232153569286143e-06, "loss": 0.0011, "step": 3895 }, { "epoch": 0.47, "learning_rate": 2.7231433713257347e-06, "loss": 0.0006, "step": 3896 }, { "epoch": 0.47, "learning_rate": 2.7230713857228555e-06, "loss": 0.0015, "step": 3897 }, { "epoch": 0.47, "learning_rate": 2.722999400119976e-06, "loss": 0.7015, "step": 3898 }, { "epoch": 0.47, "learning_rate": 2.7229274145170966e-06, "loss": 0.0013, "step": 3899 }, { "epoch": 0.47, "learning_rate": 2.722855428914217e-06, "loss": 0.0007, "step": 3900 }, { "epoch": 0.47, "learning_rate": 2.7227834433113377e-06, "loss": 0.001, "step": 3901 }, { "epoch": 0.47, "learning_rate": 2.7227114577084585e-06, "loss": 0.0277, "step": 3902 }, { "epoch": 0.47, "learning_rate": 2.722639472105579e-06, "loss": 0.0007, "step": 3903 }, { "epoch": 0.47, "learning_rate": 2.7225674865026996e-06, "loss": 0.0007, "step": 3904 }, { "epoch": 0.47, "learning_rate": 2.7224955008998204e-06, "loss": 0.0007, "step": 3905 }, { "epoch": 0.47, "learning_rate": 2.7224235152969408e-06, "loss": 0.0009, "step": 3906 }, { "epoch": 0.47, "learning_rate": 2.722351529694061e-06, "loss": 0.0007, "step": 3907 }, { "epoch": 0.47, "learning_rate": 2.722279544091182e-06, "loss": 0.4247, "step": 3908 }, { "epoch": 0.47, "learning_rate": 2.7222075584883023e-06, "loss": 0.0007, "step": 3909 }, { "epoch": 0.47, "learning_rate": 2.722135572885423e-06, "loss": 0.0014, "step": 3910 }, { "epoch": 0.47, "learning_rate": 2.7220635872825434e-06, "loss": 0.0008, "step": 3911 }, { "epoch": 0.47, "learning_rate": 2.721991601679664e-06, "loss": 0.0009, "step": 3912 }, { "epoch": 0.47, "learning_rate": 2.7219196160767845e-06, "loss": 0.0007, "step": 3913 }, { "epoch": 0.47, "learning_rate": 2.7218476304739053e-06, "loss": 0.003, "step": 3914 }, { "epoch": 0.47, "learning_rate": 2.7217756448710257e-06, "loss": 0.001, "step": 3915 }, { "epoch": 0.47, "learning_rate": 2.7217036592681464e-06, "loss": 0.0008, "step": 3916 }, { "epoch": 0.47, "learning_rate": 2.721631673665267e-06, "loss": 0.7888, "step": 3917 }, { "epoch": 0.47, "learning_rate": 2.7215596880623876e-06, "loss": 0.0012, "step": 3918 }, { "epoch": 0.47, "learning_rate": 2.7214877024595084e-06, "loss": 0.3422, "step": 3919 }, { "epoch": 0.47, "learning_rate": 2.7214157168566287e-06, "loss": 1.145, "step": 3920 }, { "epoch": 0.47, "learning_rate": 2.7213437312537495e-06, "loss": 0.001, "step": 3921 }, { "epoch": 0.47, "learning_rate": 2.72127174565087e-06, "loss": 0.003, "step": 3922 }, { "epoch": 0.47, "learning_rate": 2.7211997600479906e-06, "loss": 0.0009, "step": 3923 }, { "epoch": 0.47, "learning_rate": 2.721127774445111e-06, "loss": 0.001, "step": 3924 }, { "epoch": 0.47, "learning_rate": 2.7210557888422318e-06, "loss": 0.0008, "step": 3925 }, { "epoch": 0.47, "learning_rate": 2.720983803239352e-06, "loss": 0.0013, "step": 3926 }, { "epoch": 0.47, "learning_rate": 2.720911817636473e-06, "loss": 0.0007, "step": 3927 }, { "epoch": 0.47, "learning_rate": 2.7208398320335933e-06, "loss": 0.0019, "step": 3928 }, { "epoch": 0.47, "learning_rate": 2.720767846430714e-06, "loss": 0.8672, "step": 3929 }, { "epoch": 0.47, "learning_rate": 2.7206958608278344e-06, "loss": 0.0007, "step": 3930 }, { "epoch": 0.47, "learning_rate": 2.7206238752249547e-06, "loss": 0.0031, "step": 3931 }, { "epoch": 0.47, "learning_rate": 2.7205518896220755e-06, "loss": 0.0026, "step": 3932 }, { "epoch": 0.47, "learning_rate": 2.7204799040191963e-06, "loss": 0.0224, "step": 3933 }, { "epoch": 0.47, "learning_rate": 2.720407918416317e-06, "loss": 0.0006, "step": 3934 }, { "epoch": 0.47, "learning_rate": 2.7203359328134374e-06, "loss": 0.0008, "step": 3935 }, { "epoch": 0.47, "learning_rate": 2.720263947210558e-06, "loss": 0.0035, "step": 3936 }, { "epoch": 0.47, "learning_rate": 2.7201919616076786e-06, "loss": 0.0006, "step": 3937 }, { "epoch": 0.47, "learning_rate": 2.7201199760047993e-06, "loss": 0.0019, "step": 3938 }, { "epoch": 0.47, "learning_rate": 2.7200479904019197e-06, "loss": 0.0005, "step": 3939 }, { "epoch": 0.47, "learning_rate": 2.7199760047990405e-06, "loss": 0.0072, "step": 3940 }, { "epoch": 0.47, "learning_rate": 2.719904019196161e-06, "loss": 0.0185, "step": 3941 }, { "epoch": 0.47, "learning_rate": 2.719832033593281e-06, "loss": 0.0096, "step": 3942 }, { "epoch": 0.47, "learning_rate": 2.719760047990402e-06, "loss": 0.0012, "step": 3943 }, { "epoch": 0.47, "learning_rate": 2.7196880623875223e-06, "loss": 0.0007, "step": 3944 }, { "epoch": 0.47, "learning_rate": 2.719616076784643e-06, "loss": 0.0026, "step": 3945 }, { "epoch": 0.47, "learning_rate": 2.7195440911817635e-06, "loss": 0.0007, "step": 3946 }, { "epoch": 0.47, "learning_rate": 2.7194721055788842e-06, "loss": 0.2761, "step": 3947 }, { "epoch": 0.47, "learning_rate": 2.7194001199760046e-06, "loss": 0.0363, "step": 3948 }, { "epoch": 0.47, "learning_rate": 2.7193281343731254e-06, "loss": 0.0005, "step": 3949 }, { "epoch": 0.47, "learning_rate": 2.719256148770246e-06, "loss": 0.0006, "step": 3950 }, { "epoch": 0.47, "learning_rate": 2.719184163167367e-06, "loss": 0.4649, "step": 3951 }, { "epoch": 0.47, "learning_rate": 2.7191121775644873e-06, "loss": 0.0195, "step": 3952 }, { "epoch": 0.47, "learning_rate": 2.719040191961608e-06, "loss": 0.0007, "step": 3953 }, { "epoch": 0.47, "learning_rate": 2.7189682063587284e-06, "loss": 0.6817, "step": 3954 }, { "epoch": 0.47, "learning_rate": 2.7188962207558488e-06, "loss": 0.0008, "step": 3955 }, { "epoch": 0.47, "learning_rate": 2.7188242351529695e-06, "loss": 0.002, "step": 3956 }, { "epoch": 0.47, "learning_rate": 2.71875224955009e-06, "loss": 0.0017, "step": 3957 }, { "epoch": 0.47, "learning_rate": 2.7186802639472107e-06, "loss": 0.0005, "step": 3958 }, { "epoch": 0.47, "learning_rate": 2.718608278344331e-06, "loss": 1.6091, "step": 3959 }, { "epoch": 0.47, "learning_rate": 2.718536292741452e-06, "loss": 0.8705, "step": 3960 }, { "epoch": 0.47, "learning_rate": 2.718464307138572e-06, "loss": 0.0005, "step": 3961 }, { "epoch": 0.47, "learning_rate": 2.718392321535693e-06, "loss": 0.0008, "step": 3962 }, { "epoch": 0.47, "learning_rate": 2.7183203359328133e-06, "loss": 0.0007, "step": 3963 }, { "epoch": 0.48, "learning_rate": 2.718248350329934e-06, "loss": 0.0677, "step": 3964 }, { "epoch": 0.48, "learning_rate": 2.718176364727055e-06, "loss": 0.0006, "step": 3965 }, { "epoch": 0.48, "learning_rate": 2.7181043791241752e-06, "loss": 0.0007, "step": 3966 }, { "epoch": 0.48, "learning_rate": 2.718032393521296e-06, "loss": 0.0004, "step": 3967 }, { "epoch": 0.48, "learning_rate": 2.7179604079184163e-06, "loss": 0.0004, "step": 3968 }, { "epoch": 0.48, "learning_rate": 2.717888422315537e-06, "loss": 0.0008, "step": 3969 }, { "epoch": 0.48, "learning_rate": 2.7178164367126575e-06, "loss": 0.0006, "step": 3970 }, { "epoch": 0.48, "learning_rate": 2.7177444511097783e-06, "loss": 0.7984, "step": 3971 }, { "epoch": 0.48, "learning_rate": 2.7176724655068986e-06, "loss": 0.0012, "step": 3972 }, { "epoch": 0.48, "learning_rate": 2.7176004799040194e-06, "loss": 0.0006, "step": 3973 }, { "epoch": 0.48, "learning_rate": 2.7175284943011397e-06, "loss": 0.0005, "step": 3974 }, { "epoch": 0.48, "learning_rate": 2.7174565086982605e-06, "loss": 0.0011, "step": 3975 }, { "epoch": 0.48, "learning_rate": 2.717384523095381e-06, "loss": 0.0005, "step": 3976 }, { "epoch": 0.48, "learning_rate": 2.7173125374925012e-06, "loss": 0.0627, "step": 3977 }, { "epoch": 0.48, "learning_rate": 2.717240551889622e-06, "loss": 0.0008, "step": 3978 }, { "epoch": 0.48, "learning_rate": 2.717168566286743e-06, "loss": 0.0008, "step": 3979 }, { "epoch": 0.48, "learning_rate": 2.717096580683863e-06, "loss": 0.6643, "step": 3980 }, { "epoch": 0.48, "learning_rate": 2.717024595080984e-06, "loss": 0.0039, "step": 3981 }, { "epoch": 0.48, "learning_rate": 2.7169526094781047e-06, "loss": 0.0006, "step": 3982 }, { "epoch": 0.48, "learning_rate": 2.716880623875225e-06, "loss": 0.0998, "step": 3983 }, { "epoch": 0.48, "learning_rate": 2.716808638272346e-06, "loss": 0.6176, "step": 3984 }, { "epoch": 0.48, "learning_rate": 2.716736652669466e-06, "loss": 0.0006, "step": 3985 }, { "epoch": 0.48, "learning_rate": 2.716664667066587e-06, "loss": 0.0007, "step": 3986 }, { "epoch": 0.48, "learning_rate": 2.7165926814637073e-06, "loss": 0.0008, "step": 3987 }, { "epoch": 0.48, "learning_rate": 2.7165206958608277e-06, "loss": 0.2469, "step": 3988 }, { "epoch": 0.48, "learning_rate": 2.7164487102579485e-06, "loss": 0.0007, "step": 3989 }, { "epoch": 0.48, "learning_rate": 2.716376724655069e-06, "loss": 0.0007, "step": 3990 }, { "epoch": 0.48, "learning_rate": 2.7163047390521896e-06, "loss": 0.0006, "step": 3991 }, { "epoch": 0.48, "learning_rate": 2.71623275344931e-06, "loss": 0.0007, "step": 3992 }, { "epoch": 0.48, "learning_rate": 2.7161607678464307e-06, "loss": 0.9039, "step": 3993 }, { "epoch": 0.48, "learning_rate": 2.716088782243551e-06, "loss": 0.0044, "step": 3994 }, { "epoch": 0.48, "learning_rate": 2.716016796640672e-06, "loss": 0.6639, "step": 3995 }, { "epoch": 0.48, "learning_rate": 2.7159448110377926e-06, "loss": 0.0009, "step": 3996 }, { "epoch": 0.48, "learning_rate": 2.7158728254349134e-06, "loss": 0.0007, "step": 3997 }, { "epoch": 0.48, "learning_rate": 2.7158008398320338e-06, "loss": 0.0016, "step": 3998 }, { "epoch": 0.48, "learning_rate": 2.7157288542291546e-06, "loss": 0.0007, "step": 3999 }, { "epoch": 0.48, "learning_rate": 2.715656868626275e-06, "loss": 0.0007, "step": 4000 }, { "epoch": 0.48, "learning_rate": 2.7155848830233953e-06, "loss": 0.0008, "step": 4001 }, { "epoch": 0.48, "learning_rate": 2.715512897420516e-06, "loss": 0.0009, "step": 4002 }, { "epoch": 0.48, "learning_rate": 2.7154409118176364e-06, "loss": 0.0007, "step": 4003 }, { "epoch": 0.48, "learning_rate": 2.715368926214757e-06, "loss": 0.002, "step": 4004 }, { "epoch": 0.48, "learning_rate": 2.7152969406118775e-06, "loss": 0.0013, "step": 4005 }, { "epoch": 0.48, "learning_rate": 2.7152249550089983e-06, "loss": 0.0006, "step": 4006 }, { "epoch": 0.48, "learning_rate": 2.7151529694061187e-06, "loss": 0.8359, "step": 4007 }, { "epoch": 0.48, "learning_rate": 2.7150809838032394e-06, "loss": 0.0007, "step": 4008 }, { "epoch": 0.48, "learning_rate": 2.71500899820036e-06, "loss": 0.0007, "step": 4009 }, { "epoch": 0.48, "learning_rate": 2.7149370125974806e-06, "loss": 0.0008, "step": 4010 }, { "epoch": 0.48, "learning_rate": 2.7148650269946014e-06, "loss": 0.1514, "step": 4011 }, { "epoch": 0.48, "learning_rate": 2.7147930413917217e-06, "loss": 0.0009, "step": 4012 }, { "epoch": 0.48, "learning_rate": 2.7147210557888425e-06, "loss": 0.0007, "step": 4013 }, { "epoch": 0.48, "learning_rate": 2.714649070185963e-06, "loss": 0.2663, "step": 4014 }, { "epoch": 0.48, "learning_rate": 2.7145770845830836e-06, "loss": 0.0009, "step": 4015 }, { "epoch": 0.48, "learning_rate": 2.714505098980204e-06, "loss": 0.7179, "step": 4016 }, { "epoch": 0.48, "learning_rate": 2.7144331133773248e-06, "loss": 0.0007, "step": 4017 }, { "epoch": 0.48, "learning_rate": 2.714361127774445e-06, "loss": 0.4814, "step": 4018 }, { "epoch": 0.48, "learning_rate": 2.714289142171566e-06, "loss": 0.3875, "step": 4019 }, { "epoch": 0.48, "learning_rate": 2.7142171565686862e-06, "loss": 0.0006, "step": 4020 }, { "epoch": 0.48, "learning_rate": 2.714145170965807e-06, "loss": 0.0007, "step": 4021 }, { "epoch": 0.48, "learning_rate": 2.7140731853629274e-06, "loss": 0.4993, "step": 4022 }, { "epoch": 0.48, "learning_rate": 2.7140011997600477e-06, "loss": 0.0006, "step": 4023 }, { "epoch": 0.48, "learning_rate": 2.7139292141571685e-06, "loss": 0.001, "step": 4024 }, { "epoch": 0.48, "learning_rate": 2.7138572285542893e-06, "loss": 0.0532, "step": 4025 }, { "epoch": 0.48, "learning_rate": 2.7137852429514096e-06, "loss": 0.013, "step": 4026 }, { "epoch": 0.48, "learning_rate": 2.7137132573485304e-06, "loss": 0.0023, "step": 4027 }, { "epoch": 0.48, "learning_rate": 2.713641271745651e-06, "loss": 0.0012, "step": 4028 }, { "epoch": 0.48, "learning_rate": 2.7135692861427716e-06, "loss": 0.0007, "step": 4029 }, { "epoch": 0.48, "learning_rate": 2.7134973005398923e-06, "loss": 0.0008, "step": 4030 }, { "epoch": 0.48, "learning_rate": 2.7134253149370127e-06, "loss": 0.0028, "step": 4031 }, { "epoch": 0.48, "learning_rate": 2.7133533293341335e-06, "loss": 0.0006, "step": 4032 }, { "epoch": 0.48, "learning_rate": 2.713281343731254e-06, "loss": 0.0009, "step": 4033 }, { "epoch": 0.48, "learning_rate": 2.713209358128374e-06, "loss": 0.0006, "step": 4034 }, { "epoch": 0.48, "learning_rate": 2.713137372525495e-06, "loss": 0.0013, "step": 4035 }, { "epoch": 0.48, "learning_rate": 2.7130653869226153e-06, "loss": 0.9801, "step": 4036 }, { "epoch": 0.48, "learning_rate": 2.712993401319736e-06, "loss": 0.0007, "step": 4037 }, { "epoch": 0.48, "learning_rate": 2.7129214157168564e-06, "loss": 0.0006, "step": 4038 }, { "epoch": 0.48, "learning_rate": 2.7128494301139772e-06, "loss": 0.0007, "step": 4039 }, { "epoch": 0.48, "learning_rate": 2.7127774445110976e-06, "loss": 0.0007, "step": 4040 }, { "epoch": 0.48, "learning_rate": 2.7127054589082184e-06, "loss": 0.1455, "step": 4041 }, { "epoch": 0.48, "learning_rate": 2.712633473305339e-06, "loss": 0.8108, "step": 4042 }, { "epoch": 0.48, "learning_rate": 2.71256148770246e-06, "loss": 0.023, "step": 4043 }, { "epoch": 0.48, "learning_rate": 2.7124895020995803e-06, "loss": 0.0012, "step": 4044 }, { "epoch": 0.48, "learning_rate": 2.712417516496701e-06, "loss": 0.0007, "step": 4045 }, { "epoch": 0.48, "learning_rate": 2.7123455308938214e-06, "loss": 0.0006, "step": 4046 }, { "epoch": 0.48, "learning_rate": 2.7122735452909418e-06, "loss": 0.0007, "step": 4047 }, { "epoch": 0.49, "learning_rate": 2.7122015596880625e-06, "loss": 0.0006, "step": 4048 }, { "epoch": 0.49, "learning_rate": 2.712129574085183e-06, "loss": 0.0009, "step": 4049 }, { "epoch": 0.49, "learning_rate": 2.7120575884823037e-06, "loss": 0.0043, "step": 4050 }, { "epoch": 0.49, "learning_rate": 2.711985602879424e-06, "loss": 0.0245, "step": 4051 }, { "epoch": 0.49, "learning_rate": 2.711913617276545e-06, "loss": 0.0017, "step": 4052 }, { "epoch": 0.49, "learning_rate": 2.711841631673665e-06, "loss": 0.0017, "step": 4053 }, { "epoch": 0.49, "learning_rate": 2.711769646070786e-06, "loss": 0.0007, "step": 4054 }, { "epoch": 0.49, "learning_rate": 2.7116976604679063e-06, "loss": 0.0009, "step": 4055 }, { "epoch": 0.49, "learning_rate": 2.711625674865027e-06, "loss": 0.0006, "step": 4056 }, { "epoch": 0.49, "learning_rate": 2.711553689262148e-06, "loss": 0.0222, "step": 4057 }, { "epoch": 0.49, "learning_rate": 2.711481703659268e-06, "loss": 0.0006, "step": 4058 }, { "epoch": 0.49, "learning_rate": 2.711409718056389e-06, "loss": 0.0006, "step": 4059 }, { "epoch": 0.49, "learning_rate": 2.7113377324535093e-06, "loss": 0.0037, "step": 4060 }, { "epoch": 0.49, "learning_rate": 2.71126574685063e-06, "loss": 0.0007, "step": 4061 }, { "epoch": 0.49, "learning_rate": 2.7111937612477505e-06, "loss": 0.0006, "step": 4062 }, { "epoch": 0.49, "learning_rate": 2.7111217756448712e-06, "loss": 0.0007, "step": 4063 }, { "epoch": 0.49, "learning_rate": 2.7110497900419916e-06, "loss": 0.0006, "step": 4064 }, { "epoch": 0.49, "learning_rate": 2.7109778044391124e-06, "loss": 0.6725, "step": 4065 }, { "epoch": 0.49, "learning_rate": 2.7109058188362327e-06, "loss": 0.0012, "step": 4066 }, { "epoch": 0.49, "learning_rate": 2.7108338332333535e-06, "loss": 0.0076, "step": 4067 }, { "epoch": 0.49, "learning_rate": 2.710761847630474e-06, "loss": 0.0016, "step": 4068 }, { "epoch": 0.49, "learning_rate": 2.7106898620275942e-06, "loss": 0.527, "step": 4069 }, { "epoch": 0.49, "learning_rate": 2.710617876424715e-06, "loss": 0.0019, "step": 4070 }, { "epoch": 0.49, "learning_rate": 2.7105458908218354e-06, "loss": 0.0122, "step": 4071 }, { "epoch": 0.49, "learning_rate": 2.710473905218956e-06, "loss": 0.8456, "step": 4072 }, { "epoch": 0.49, "learning_rate": 2.710401919616077e-06, "loss": 0.0009, "step": 4073 }, { "epoch": 0.49, "learning_rate": 2.7103299340131977e-06, "loss": 0.189, "step": 4074 }, { "epoch": 0.49, "learning_rate": 2.710257948410318e-06, "loss": 0.1374, "step": 4075 }, { "epoch": 0.49, "learning_rate": 2.710185962807439e-06, "loss": 0.0012, "step": 4076 }, { "epoch": 0.49, "learning_rate": 2.710113977204559e-06, "loss": 0.0006, "step": 4077 }, { "epoch": 0.49, "learning_rate": 2.71004199160168e-06, "loss": 0.0008, "step": 4078 }, { "epoch": 0.49, "learning_rate": 2.7099700059988003e-06, "loss": 0.0007, "step": 4079 }, { "epoch": 0.49, "learning_rate": 2.709898020395921e-06, "loss": 0.0008, "step": 4080 }, { "epoch": 0.49, "learning_rate": 2.7098260347930415e-06, "loss": 0.0006, "step": 4081 }, { "epoch": 0.49, "learning_rate": 2.709754049190162e-06, "loss": 0.0012, "step": 4082 }, { "epoch": 0.49, "learning_rate": 2.7096820635872826e-06, "loss": 0.0012, "step": 4083 }, { "epoch": 0.49, "learning_rate": 2.709610077984403e-06, "loss": 0.0006, "step": 4084 }, { "epoch": 0.49, "learning_rate": 2.7095380923815237e-06, "loss": 0.6089, "step": 4085 }, { "epoch": 0.49, "learning_rate": 2.709466106778644e-06, "loss": 0.0006, "step": 4086 }, { "epoch": 0.49, "learning_rate": 2.709394121175765e-06, "loss": 0.7387, "step": 4087 }, { "epoch": 0.49, "learning_rate": 2.7093221355728856e-06, "loss": 0.0006, "step": 4088 }, { "epoch": 0.49, "learning_rate": 2.709250149970006e-06, "loss": 0.0034, "step": 4089 }, { "epoch": 0.49, "learning_rate": 2.7091781643671268e-06, "loss": 0.0007, "step": 4090 }, { "epoch": 0.49, "learning_rate": 2.7091061787642475e-06, "loss": 0.0013, "step": 4091 }, { "epoch": 0.49, "learning_rate": 2.709034193161368e-06, "loss": 0.8827, "step": 4092 }, { "epoch": 0.49, "learning_rate": 2.7089622075584883e-06, "loss": 0.0007, "step": 4093 }, { "epoch": 0.49, "learning_rate": 2.708890221955609e-06, "loss": 0.0007, "step": 4094 }, { "epoch": 0.49, "learning_rate": 2.7088182363527294e-06, "loss": 0.0078, "step": 4095 }, { "epoch": 0.49, "learning_rate": 2.70874625074985e-06, "loss": 0.0007, "step": 4096 }, { "epoch": 0.49, "learning_rate": 2.7086742651469705e-06, "loss": 0.0009, "step": 4097 }, { "epoch": 0.49, "learning_rate": 2.7086022795440913e-06, "loss": 0.0007, "step": 4098 }, { "epoch": 0.49, "learning_rate": 2.7085302939412117e-06, "loss": 0.6629, "step": 4099 }, { "epoch": 0.49, "learning_rate": 2.7084583083383324e-06, "loss": 0.002, "step": 4100 }, { "epoch": 0.49, "learning_rate": 2.7083863227354528e-06, "loss": 0.0008, "step": 4101 }, { "epoch": 0.49, "learning_rate": 2.7083143371325736e-06, "loss": 0.0007, "step": 4102 }, { "epoch": 0.49, "learning_rate": 2.708242351529694e-06, "loss": 0.01, "step": 4103 }, { "epoch": 0.49, "learning_rate": 2.7081703659268147e-06, "loss": 0.0009, "step": 4104 }, { "epoch": 0.49, "learning_rate": 2.7080983803239355e-06, "loss": 0.0009, "step": 4105 }, { "epoch": 0.49, "learning_rate": 2.708026394721056e-06, "loss": 0.0011, "step": 4106 }, { "epoch": 0.49, "learning_rate": 2.7079544091181766e-06, "loss": 0.0008, "step": 4107 }, { "epoch": 0.49, "learning_rate": 2.707882423515297e-06, "loss": 0.0005, "step": 4108 }, { "epoch": 0.49, "learning_rate": 2.7078104379124177e-06, "loss": 0.0005, "step": 4109 }, { "epoch": 0.49, "learning_rate": 2.707738452309538e-06, "loss": 0.0008, "step": 4110 }, { "epoch": 0.49, "learning_rate": 2.707666466706659e-06, "loss": 0.0007, "step": 4111 }, { "epoch": 0.49, "learning_rate": 2.7075944811037792e-06, "loss": 0.0007, "step": 4112 }, { "epoch": 0.49, "learning_rate": 2.7075224955009e-06, "loss": 0.001, "step": 4113 }, { "epoch": 0.49, "learning_rate": 2.7074505098980204e-06, "loss": 0.0007, "step": 4114 }, { "epoch": 0.49, "learning_rate": 2.7073785242951407e-06, "loss": 0.0006, "step": 4115 }, { "epoch": 0.49, "learning_rate": 2.7073065386922615e-06, "loss": 0.0025, "step": 4116 }, { "epoch": 0.49, "learning_rate": 2.707234553089382e-06, "loss": 0.3824, "step": 4117 }, { "epoch": 0.49, "learning_rate": 2.7071625674865026e-06, "loss": 0.0167, "step": 4118 }, { "epoch": 0.49, "learning_rate": 2.7070905818836234e-06, "loss": 0.0052, "step": 4119 }, { "epoch": 0.49, "learning_rate": 2.707018596280744e-06, "loss": 0.0005, "step": 4120 }, { "epoch": 0.49, "learning_rate": 2.7069466106778645e-06, "loss": 0.0007, "step": 4121 }, { "epoch": 0.49, "learning_rate": 2.7068746250749853e-06, "loss": 0.0014, "step": 4122 }, { "epoch": 0.49, "learning_rate": 2.7068026394721057e-06, "loss": 0.0046, "step": 4123 }, { "epoch": 0.49, "learning_rate": 2.7067306538692265e-06, "loss": 2.0689, "step": 4124 }, { "epoch": 0.49, "learning_rate": 2.706658668266347e-06, "loss": 0.0007, "step": 4125 }, { "epoch": 0.49, "learning_rate": 2.7065866826634676e-06, "loss": 0.7189, "step": 4126 }, { "epoch": 0.49, "learning_rate": 2.706514697060588e-06, "loss": 0.0037, "step": 4127 }, { "epoch": 0.49, "learning_rate": 2.7064427114577083e-06, "loss": 0.0008, "step": 4128 }, { "epoch": 0.49, "learning_rate": 2.706370725854829e-06, "loss": 0.0009, "step": 4129 }, { "epoch": 0.49, "learning_rate": 2.7062987402519494e-06, "loss": 0.6161, "step": 4130 }, { "epoch": 0.5, "learning_rate": 2.7062267546490702e-06, "loss": 0.0011, "step": 4131 }, { "epoch": 0.5, "learning_rate": 2.7061547690461906e-06, "loss": 0.0006, "step": 4132 }, { "epoch": 0.5, "learning_rate": 2.7060827834433113e-06, "loss": 0.0089, "step": 4133 }, { "epoch": 0.5, "learning_rate": 2.706010797840432e-06, "loss": 0.001, "step": 4134 }, { "epoch": 0.5, "learning_rate": 2.7059388122375525e-06, "loss": 0.0033, "step": 4135 }, { "epoch": 0.5, "learning_rate": 2.7058668266346733e-06, "loss": 0.0006, "step": 4136 }, { "epoch": 0.5, "learning_rate": 2.705794841031794e-06, "loss": 0.0039, "step": 4137 }, { "epoch": 0.5, "learning_rate": 2.7057228554289144e-06, "loss": 0.0006, "step": 4138 }, { "epoch": 0.5, "learning_rate": 2.7056508698260347e-06, "loss": 0.0007, "step": 4139 }, { "epoch": 0.5, "learning_rate": 2.7055788842231555e-06, "loss": 0.001, "step": 4140 }, { "epoch": 0.5, "learning_rate": 2.705506898620276e-06, "loss": 0.0005, "step": 4141 }, { "epoch": 0.5, "learning_rate": 2.7054349130173967e-06, "loss": 0.0017, "step": 4142 }, { "epoch": 0.5, "learning_rate": 2.705362927414517e-06, "loss": 0.0007, "step": 4143 }, { "epoch": 0.5, "learning_rate": 2.705290941811638e-06, "loss": 0.0005, "step": 4144 }, { "epoch": 0.5, "learning_rate": 2.705218956208758e-06, "loss": 0.8464, "step": 4145 }, { "epoch": 0.5, "learning_rate": 2.705146970605879e-06, "loss": 0.0007, "step": 4146 }, { "epoch": 0.5, "learning_rate": 2.7050749850029993e-06, "loss": 0.0011, "step": 4147 }, { "epoch": 0.5, "learning_rate": 2.70500299940012e-06, "loss": 0.0009, "step": 4148 }, { "epoch": 0.5, "learning_rate": 2.7049310137972404e-06, "loss": 0.001, "step": 4149 }, { "epoch": 0.5, "learning_rate": 2.704859028194361e-06, "loss": 0.0222, "step": 4150 }, { "epoch": 0.5, "learning_rate": 2.704787042591482e-06, "loss": 0.0007, "step": 4151 }, { "epoch": 0.5, "learning_rate": 2.7047150569886023e-06, "loss": 0.0007, "step": 4152 }, { "epoch": 0.5, "learning_rate": 2.704643071385723e-06, "loss": 0.0009, "step": 4153 }, { "epoch": 0.5, "learning_rate": 2.7045710857828435e-06, "loss": 0.0006, "step": 4154 }, { "epoch": 0.5, "learning_rate": 2.7044991001799642e-06, "loss": 0.0007, "step": 4155 }, { "epoch": 0.5, "learning_rate": 2.7044271145770846e-06, "loss": 0.0005, "step": 4156 }, { "epoch": 0.5, "learning_rate": 2.7043551289742054e-06, "loss": 0.0007, "step": 4157 }, { "epoch": 0.5, "learning_rate": 2.7042831433713257e-06, "loss": 0.5971, "step": 4158 }, { "epoch": 0.5, "learning_rate": 2.7042111577684465e-06, "loss": 0.8166, "step": 4159 }, { "epoch": 0.5, "learning_rate": 2.704139172165567e-06, "loss": 0.0333, "step": 4160 }, { "epoch": 0.5, "learning_rate": 2.7040671865626872e-06, "loss": 0.0007, "step": 4161 }, { "epoch": 0.5, "learning_rate": 2.703995200959808e-06, "loss": 0.9045, "step": 4162 }, { "epoch": 0.5, "learning_rate": 2.7039232153569284e-06, "loss": 0.7781, "step": 4163 }, { "epoch": 0.5, "learning_rate": 2.703851229754049e-06, "loss": 1.0117, "step": 4164 }, { "epoch": 0.5, "learning_rate": 2.70377924415117e-06, "loss": 0.0032, "step": 4165 }, { "epoch": 0.5, "learning_rate": 2.7037072585482907e-06, "loss": 0.846, "step": 4166 }, { "epoch": 0.5, "learning_rate": 2.703635272945411e-06, "loss": 0.0794, "step": 4167 }, { "epoch": 0.5, "learning_rate": 2.703563287342532e-06, "loss": 0.1899, "step": 4168 }, { "epoch": 0.5, "learning_rate": 2.703491301739652e-06, "loss": 0.0009, "step": 4169 }, { "epoch": 0.5, "learning_rate": 2.703419316136773e-06, "loss": 0.0009, "step": 4170 }, { "epoch": 0.5, "learning_rate": 2.7033473305338933e-06, "loss": 0.7803, "step": 4171 }, { "epoch": 0.5, "learning_rate": 2.703275344931014e-06, "loss": 0.0011, "step": 4172 }, { "epoch": 0.5, "learning_rate": 2.7032033593281344e-06, "loss": 0.0008, "step": 4173 }, { "epoch": 0.5, "learning_rate": 2.703131373725255e-06, "loss": 0.1892, "step": 4174 }, { "epoch": 0.5, "learning_rate": 2.7030593881223756e-06, "loss": 0.0007, "step": 4175 }, { "epoch": 0.5, "learning_rate": 2.702987402519496e-06, "loss": 0.0006, "step": 4176 }, { "epoch": 0.5, "learning_rate": 2.7029154169166167e-06, "loss": 0.0009, "step": 4177 }, { "epoch": 0.5, "learning_rate": 2.702843431313737e-06, "loss": 0.0691, "step": 4178 }, { "epoch": 0.5, "learning_rate": 2.702771445710858e-06, "loss": 0.0007, "step": 4179 }, { "epoch": 0.5, "learning_rate": 2.7026994601079786e-06, "loss": 0.0014, "step": 4180 }, { "epoch": 0.5, "learning_rate": 2.702627474505099e-06, "loss": 0.2043, "step": 4181 }, { "epoch": 0.5, "learning_rate": 2.7025554889022198e-06, "loss": 0.001, "step": 4182 }, { "epoch": 0.5, "learning_rate": 2.7024835032993405e-06, "loss": 0.0071, "step": 4183 }, { "epoch": 0.5, "learning_rate": 2.702411517696461e-06, "loss": 0.0976, "step": 4184 }, { "epoch": 0.5, "learning_rate": 2.7023395320935812e-06, "loss": 0.0119, "step": 4185 }, { "epoch": 0.5, "learning_rate": 2.702267546490702e-06, "loss": 0.0009, "step": 4186 }, { "epoch": 0.5, "learning_rate": 2.7021955608878224e-06, "loss": 0.0025, "step": 4187 }, { "epoch": 0.5, "learning_rate": 2.702123575284943e-06, "loss": 0.005, "step": 4188 }, { "epoch": 0.5, "learning_rate": 2.7020515896820635e-06, "loss": 0.0017, "step": 4189 }, { "epoch": 0.5, "learning_rate": 2.7019796040791843e-06, "loss": 0.0008, "step": 4190 }, { "epoch": 0.5, "learning_rate": 2.7019076184763046e-06, "loss": 0.0015, "step": 4191 }, { "epoch": 0.5, "learning_rate": 2.7018356328734254e-06, "loss": 0.0008, "step": 4192 }, { "epoch": 0.5, "learning_rate": 2.7017636472705458e-06, "loss": 0.0007, "step": 4193 }, { "epoch": 0.5, "learning_rate": 2.7016916616676666e-06, "loss": 0.0011, "step": 4194 }, { "epoch": 0.5, "learning_rate": 2.701619676064787e-06, "loss": 0.0009, "step": 4195 }, { "epoch": 0.5, "learning_rate": 2.7015476904619077e-06, "loss": 0.0007, "step": 4196 }, { "epoch": 0.5, "learning_rate": 2.7014757048590285e-06, "loss": 0.0009, "step": 4197 }, { "epoch": 0.5, "learning_rate": 2.701403719256149e-06, "loss": 0.0006, "step": 4198 }, { "epoch": 0.5, "learning_rate": 2.7013317336532696e-06, "loss": 0.0008, "step": 4199 }, { "epoch": 0.5, "learning_rate": 2.70125974805039e-06, "loss": 0.0006, "step": 4200 }, { "epoch": 0.5, "learning_rate": 2.7011877624475107e-06, "loss": 0.0815, "step": 4201 }, { "epoch": 0.5, "learning_rate": 2.701115776844631e-06, "loss": 0.1347, "step": 4202 }, { "epoch": 0.5, "learning_rate": 2.701043791241752e-06, "loss": 0.0129, "step": 4203 }, { "epoch": 0.5, "learning_rate": 2.7009718056388722e-06, "loss": 0.0006, "step": 4204 }, { "epoch": 0.5, "learning_rate": 2.700899820035993e-06, "loss": 0.6137, "step": 4205 }, { "epoch": 0.5, "learning_rate": 2.7008278344331134e-06, "loss": 0.0007, "step": 4206 }, { "epoch": 0.5, "learning_rate": 2.700755848830234e-06, "loss": 0.0009, "step": 4207 }, { "epoch": 0.5, "learning_rate": 2.7006838632273545e-06, "loss": 0.007, "step": 4208 }, { "epoch": 0.5, "learning_rate": 2.700611877624475e-06, "loss": 0.0009, "step": 4209 }, { "epoch": 0.5, "learning_rate": 2.7005398920215956e-06, "loss": 0.0014, "step": 4210 }, { "epoch": 0.5, "learning_rate": 2.7004679064187164e-06, "loss": 0.0006, "step": 4211 }, { "epoch": 0.5, "learning_rate": 2.7003959208158368e-06, "loss": 0.0007, "step": 4212 }, { "epoch": 0.5, "learning_rate": 2.7003239352129575e-06, "loss": 0.0039, "step": 4213 }, { "epoch": 0.5, "learning_rate": 2.7002519496100783e-06, "loss": 0.0024, "step": 4214 }, { "epoch": 0.51, "learning_rate": 2.7001799640071987e-06, "loss": 0.6217, "step": 4215 }, { "epoch": 0.51, "learning_rate": 2.7001079784043195e-06, "loss": 0.0105, "step": 4216 }, { "epoch": 0.51, "learning_rate": 2.70003599280144e-06, "loss": 0.0017, "step": 4217 }, { "epoch": 0.51, "learning_rate": 2.6999640071985606e-06, "loss": 0.001, "step": 4218 }, { "epoch": 0.51, "learning_rate": 2.699892021595681e-06, "loss": 0.0031, "step": 4219 }, { "epoch": 0.51, "learning_rate": 2.6998200359928013e-06, "loss": 0.012, "step": 4220 }, { "epoch": 0.51, "learning_rate": 2.699748050389922e-06, "loss": 0.0009, "step": 4221 }, { "epoch": 0.51, "learning_rate": 2.6996760647870424e-06, "loss": 0.0007, "step": 4222 }, { "epoch": 0.51, "learning_rate": 2.699604079184163e-06, "loss": 0.0011, "step": 4223 }, { "epoch": 0.51, "learning_rate": 2.6995320935812836e-06, "loss": 0.0286, "step": 4224 }, { "epoch": 0.51, "learning_rate": 2.6994601079784043e-06, "loss": 0.9301, "step": 4225 }, { "epoch": 0.51, "learning_rate": 2.6993881223755247e-06, "loss": 0.0006, "step": 4226 }, { "epoch": 0.51, "learning_rate": 2.6993161367726455e-06, "loss": 0.0017, "step": 4227 }, { "epoch": 0.51, "learning_rate": 2.6992441511697663e-06, "loss": 0.031, "step": 4228 }, { "epoch": 0.51, "learning_rate": 2.699172165566887e-06, "loss": 0.0008, "step": 4229 }, { "epoch": 0.51, "learning_rate": 2.6991001799640074e-06, "loss": 0.0013, "step": 4230 }, { "epoch": 0.51, "learning_rate": 2.6990281943611277e-06, "loss": 0.0008, "step": 4231 }, { "epoch": 0.51, "learning_rate": 2.6989562087582485e-06, "loss": 0.0008, "step": 4232 }, { "epoch": 0.51, "learning_rate": 2.698884223155369e-06, "loss": 0.7871, "step": 4233 }, { "epoch": 0.51, "learning_rate": 2.6988122375524897e-06, "loss": 0.0006, "step": 4234 }, { "epoch": 0.51, "learning_rate": 2.69874025194961e-06, "loss": 0.0007, "step": 4235 }, { "epoch": 0.51, "learning_rate": 2.6986682663467308e-06, "loss": 0.001, "step": 4236 }, { "epoch": 0.51, "learning_rate": 2.698596280743851e-06, "loss": 0.0007, "step": 4237 }, { "epoch": 0.51, "learning_rate": 2.698524295140972e-06, "loss": 0.0006, "step": 4238 }, { "epoch": 0.51, "learning_rate": 2.6984523095380923e-06, "loss": 0.6676, "step": 4239 }, { "epoch": 0.51, "learning_rate": 2.698380323935213e-06, "loss": 0.0007, "step": 4240 }, { "epoch": 0.51, "learning_rate": 2.6983083383323334e-06, "loss": 0.1338, "step": 4241 }, { "epoch": 0.51, "learning_rate": 2.698236352729454e-06, "loss": 0.001, "step": 4242 }, { "epoch": 0.51, "learning_rate": 2.698164367126575e-06, "loss": 0.0006, "step": 4243 }, { "epoch": 0.51, "learning_rate": 2.6980923815236953e-06, "loss": 0.0007, "step": 4244 }, { "epoch": 0.51, "learning_rate": 2.698020395920816e-06, "loss": 0.8623, "step": 4245 }, { "epoch": 0.51, "learning_rate": 2.6979484103179365e-06, "loss": 1.0865, "step": 4246 }, { "epoch": 0.51, "learning_rate": 2.6978764247150572e-06, "loss": 0.0011, "step": 4247 }, { "epoch": 0.51, "learning_rate": 2.6978044391121776e-06, "loss": 0.0007, "step": 4248 }, { "epoch": 0.51, "learning_rate": 2.6977324535092984e-06, "loss": 0.0011, "step": 4249 }, { "epoch": 0.51, "learning_rate": 2.6976604679064187e-06, "loss": 0.0007, "step": 4250 }, { "epoch": 0.51, "learning_rate": 2.6975884823035395e-06, "loss": 0.0016, "step": 4251 }, { "epoch": 0.51, "learning_rate": 2.69751649670066e-06, "loss": 0.0011, "step": 4252 }, { "epoch": 0.51, "learning_rate": 2.6974445110977806e-06, "loss": 0.0008, "step": 4253 }, { "epoch": 0.51, "learning_rate": 2.697372525494901e-06, "loss": 0.0007, "step": 4254 }, { "epoch": 0.51, "learning_rate": 2.6973005398920213e-06, "loss": 0.0008, "step": 4255 }, { "epoch": 0.51, "learning_rate": 2.697228554289142e-06, "loss": 0.0013, "step": 4256 }, { "epoch": 0.51, "learning_rate": 2.697156568686263e-06, "loss": 0.0009, "step": 4257 }, { "epoch": 0.51, "learning_rate": 2.6970845830833833e-06, "loss": 0.0009, "step": 4258 }, { "epoch": 0.51, "learning_rate": 2.697012597480504e-06, "loss": 0.4227, "step": 4259 }, { "epoch": 0.51, "learning_rate": 2.696940611877625e-06, "loss": 0.001, "step": 4260 }, { "epoch": 0.51, "learning_rate": 2.696868626274745e-06, "loss": 0.0036, "step": 4261 }, { "epoch": 0.51, "learning_rate": 2.696796640671866e-06, "loss": 0.0032, "step": 4262 }, { "epoch": 0.51, "learning_rate": 2.6967246550689863e-06, "loss": 0.1533, "step": 4263 }, { "epoch": 0.51, "learning_rate": 2.696652669466107e-06, "loss": 0.0009, "step": 4264 }, { "epoch": 0.51, "learning_rate": 2.6965806838632274e-06, "loss": 0.7118, "step": 4265 }, { "epoch": 0.51, "learning_rate": 2.696508698260348e-06, "loss": 0.0012, "step": 4266 }, { "epoch": 0.51, "learning_rate": 2.6964367126574686e-06, "loss": 0.0026, "step": 4267 }, { "epoch": 0.51, "learning_rate": 2.696364727054589e-06, "loss": 0.0008, "step": 4268 }, { "epoch": 0.51, "learning_rate": 2.6962927414517097e-06, "loss": 0.0016, "step": 4269 }, { "epoch": 0.51, "learning_rate": 2.69622075584883e-06, "loss": 0.0008, "step": 4270 }, { "epoch": 0.51, "learning_rate": 2.696148770245951e-06, "loss": 0.0078, "step": 4271 }, { "epoch": 0.51, "learning_rate": 2.696076784643071e-06, "loss": 0.0009, "step": 4272 }, { "epoch": 0.51, "learning_rate": 2.696004799040192e-06, "loss": 0.0004, "step": 4273 }, { "epoch": 0.51, "learning_rate": 2.6959328134373127e-06, "loss": 0.0012, "step": 4274 }, { "epoch": 0.51, "learning_rate": 2.6958608278344335e-06, "loss": 0.0141, "step": 4275 }, { "epoch": 0.51, "learning_rate": 2.695788842231554e-06, "loss": 0.0009, "step": 4276 }, { "epoch": 0.51, "learning_rate": 2.6957168566286742e-06, "loss": 0.0032, "step": 4277 }, { "epoch": 0.51, "learning_rate": 2.695644871025795e-06, "loss": 0.7744, "step": 4278 }, { "epoch": 0.51, "learning_rate": 2.6955728854229154e-06, "loss": 0.0013, "step": 4279 }, { "epoch": 0.51, "learning_rate": 2.695500899820036e-06, "loss": 0.0006, "step": 4280 }, { "epoch": 0.51, "learning_rate": 2.6954289142171565e-06, "loss": 0.0008, "step": 4281 }, { "epoch": 0.51, "learning_rate": 2.6953569286142773e-06, "loss": 0.0298, "step": 4282 }, { "epoch": 0.51, "learning_rate": 2.6952849430113976e-06, "loss": 0.0009, "step": 4283 }, { "epoch": 0.51, "learning_rate": 2.6952129574085184e-06, "loss": 0.0014, "step": 4284 }, { "epoch": 0.51, "learning_rate": 2.6951409718056388e-06, "loss": 0.043, "step": 4285 }, { "epoch": 0.51, "learning_rate": 2.6950689862027596e-06, "loss": 0.0196, "step": 4286 }, { "epoch": 0.51, "learning_rate": 2.69499700059988e-06, "loss": 0.003, "step": 4287 }, { "epoch": 0.51, "learning_rate": 2.6949250149970007e-06, "loss": 0.0007, "step": 4288 }, { "epoch": 0.51, "learning_rate": 2.6948530293941215e-06, "loss": 0.0011, "step": 4289 }, { "epoch": 0.51, "learning_rate": 2.694781043791242e-06, "loss": 0.0014, "step": 4290 }, { "epoch": 0.51, "learning_rate": 2.6947090581883626e-06, "loss": 0.0011, "step": 4291 }, { "epoch": 0.51, "learning_rate": 2.694637072585483e-06, "loss": 0.0008, "step": 4292 }, { "epoch": 0.51, "learning_rate": 2.6945650869826037e-06, "loss": 0.0011, "step": 4293 }, { "epoch": 0.51, "learning_rate": 2.694493101379724e-06, "loss": 0.0011, "step": 4294 }, { "epoch": 0.51, "learning_rate": 2.694421115776845e-06, "loss": 0.0014, "step": 4295 }, { "epoch": 0.51, "learning_rate": 2.6943491301739652e-06, "loss": 0.0011, "step": 4296 }, { "epoch": 0.51, "learning_rate": 2.694277144571086e-06, "loss": 0.0012, "step": 4297 }, { "epoch": 0.52, "learning_rate": 2.6942051589682064e-06, "loss": 0.0007, "step": 4298 }, { "epoch": 0.52, "learning_rate": 2.694133173365327e-06, "loss": 0.0006, "step": 4299 }, { "epoch": 0.52, "learning_rate": 2.6940611877624475e-06, "loss": 0.5287, "step": 4300 }, { "epoch": 0.52, "learning_rate": 2.693989202159568e-06, "loss": 0.0008, "step": 4301 }, { "epoch": 0.52, "learning_rate": 2.6939172165566886e-06, "loss": 0.0006, "step": 4302 }, { "epoch": 0.52, "learning_rate": 2.6938452309538094e-06, "loss": 0.0011, "step": 4303 }, { "epoch": 0.52, "learning_rate": 2.6937732453509298e-06, "loss": 0.1456, "step": 4304 }, { "epoch": 0.52, "learning_rate": 2.6937012597480505e-06, "loss": 0.001, "step": 4305 }, { "epoch": 0.52, "learning_rate": 2.6936292741451713e-06, "loss": 1.5231, "step": 4306 }, { "epoch": 0.52, "learning_rate": 2.6935572885422917e-06, "loss": 0.072, "step": 4307 }, { "epoch": 0.52, "learning_rate": 2.6934853029394124e-06, "loss": 0.0801, "step": 4308 }, { "epoch": 0.52, "learning_rate": 2.693413317336533e-06, "loss": 0.0045, "step": 4309 }, { "epoch": 0.52, "learning_rate": 2.6933413317336536e-06, "loss": 0.7722, "step": 4310 }, { "epoch": 0.52, "learning_rate": 2.693269346130774e-06, "loss": 0.0007, "step": 4311 }, { "epoch": 0.52, "learning_rate": 2.6931973605278943e-06, "loss": 0.0006, "step": 4312 }, { "epoch": 0.52, "learning_rate": 2.693125374925015e-06, "loss": 0.0006, "step": 4313 }, { "epoch": 0.52, "learning_rate": 2.6930533893221354e-06, "loss": 0.0009, "step": 4314 }, { "epoch": 0.52, "learning_rate": 2.692981403719256e-06, "loss": 0.0012, "step": 4315 }, { "epoch": 0.52, "learning_rate": 2.6929094181163766e-06, "loss": 0.0005, "step": 4316 }, { "epoch": 0.52, "learning_rate": 2.6928374325134973e-06, "loss": 0.0005, "step": 4317 }, { "epoch": 0.52, "learning_rate": 2.6927654469106177e-06, "loss": 0.0013, "step": 4318 }, { "epoch": 0.52, "learning_rate": 2.6926934613077385e-06, "loss": 0.001, "step": 4319 }, { "epoch": 0.52, "learning_rate": 2.6926214757048592e-06, "loss": 0.0027, "step": 4320 }, { "epoch": 0.52, "learning_rate": 2.69254949010198e-06, "loss": 0.024, "step": 4321 }, { "epoch": 0.52, "learning_rate": 2.6924775044991004e-06, "loss": 0.0009, "step": 4322 }, { "epoch": 0.52, "learning_rate": 2.6924055188962207e-06, "loss": 0.0006, "step": 4323 }, { "epoch": 0.52, "learning_rate": 2.6923335332933415e-06, "loss": 0.0008, "step": 4324 }, { "epoch": 0.52, "learning_rate": 2.692261547690462e-06, "loss": 0.0007, "step": 4325 }, { "epoch": 0.52, "learning_rate": 2.6921895620875826e-06, "loss": 0.0462, "step": 4326 }, { "epoch": 0.52, "learning_rate": 2.692117576484703e-06, "loss": 0.0013, "step": 4327 }, { "epoch": 0.52, "learning_rate": 2.6920455908818238e-06, "loss": 0.0012, "step": 4328 }, { "epoch": 0.52, "learning_rate": 2.691973605278944e-06, "loss": 0.9372, "step": 4329 }, { "epoch": 0.52, "learning_rate": 2.691901619676065e-06, "loss": 0.0025, "step": 4330 }, { "epoch": 0.52, "learning_rate": 2.6918296340731853e-06, "loss": 0.0009, "step": 4331 }, { "epoch": 0.52, "learning_rate": 2.691757648470306e-06, "loss": 0.0019, "step": 4332 }, { "epoch": 0.52, "learning_rate": 2.6916856628674264e-06, "loss": 0.4325, "step": 4333 }, { "epoch": 0.52, "learning_rate": 2.691613677264547e-06, "loss": 0.1887, "step": 4334 }, { "epoch": 0.52, "learning_rate": 2.691541691661668e-06, "loss": 0.907, "step": 4335 }, { "epoch": 0.52, "learning_rate": 2.6914697060587883e-06, "loss": 0.0014, "step": 4336 }, { "epoch": 0.52, "learning_rate": 2.691397720455909e-06, "loss": 0.0008, "step": 4337 }, { "epoch": 0.52, "learning_rate": 2.6913257348530294e-06, "loss": 0.3251, "step": 4338 }, { "epoch": 0.52, "learning_rate": 2.6912537492501502e-06, "loss": 1.162, "step": 4339 }, { "epoch": 0.52, "learning_rate": 2.6911817636472706e-06, "loss": 0.0015, "step": 4340 }, { "epoch": 0.52, "learning_rate": 2.6911097780443914e-06, "loss": 0.5195, "step": 4341 }, { "epoch": 0.52, "learning_rate": 2.6910377924415117e-06, "loss": 0.0005, "step": 4342 }, { "epoch": 0.52, "learning_rate": 2.6909658068386325e-06, "loss": 0.001, "step": 4343 }, { "epoch": 0.52, "learning_rate": 2.690893821235753e-06, "loss": 0.0012, "step": 4344 }, { "epoch": 0.52, "learning_rate": 2.6908218356328736e-06, "loss": 0.0011, "step": 4345 }, { "epoch": 0.52, "learning_rate": 2.690749850029994e-06, "loss": 0.0009, "step": 4346 }, { "epoch": 0.52, "learning_rate": 2.6906778644271143e-06, "loss": 0.0015, "step": 4347 }, { "epoch": 0.52, "learning_rate": 2.690605878824235e-06, "loss": 0.0012, "step": 4348 }, { "epoch": 0.52, "learning_rate": 2.6905338932213555e-06, "loss": 0.1529, "step": 4349 }, { "epoch": 0.52, "learning_rate": 2.6904619076184762e-06, "loss": 0.0006, "step": 4350 }, { "epoch": 0.52, "learning_rate": 2.690389922015597e-06, "loss": 0.0009, "step": 4351 }, { "epoch": 0.52, "learning_rate": 2.690317936412718e-06, "loss": 0.0017, "step": 4352 }, { "epoch": 0.52, "learning_rate": 2.690245950809838e-06, "loss": 0.0008, "step": 4353 }, { "epoch": 0.52, "learning_rate": 2.690173965206959e-06, "loss": 0.0009, "step": 4354 }, { "epoch": 0.52, "learning_rate": 2.6901019796040793e-06, "loss": 0.0044, "step": 4355 }, { "epoch": 0.52, "learning_rate": 2.6900299940012e-06, "loss": 0.6686, "step": 4356 }, { "epoch": 0.52, "learning_rate": 2.6899580083983204e-06, "loss": 0.0007, "step": 4357 }, { "epoch": 0.52, "learning_rate": 2.6898860227954408e-06, "loss": 0.4625, "step": 4358 }, { "epoch": 0.52, "learning_rate": 2.6898140371925616e-06, "loss": 0.0017, "step": 4359 }, { "epoch": 0.52, "learning_rate": 2.689742051589682e-06, "loss": 0.0008, "step": 4360 }, { "epoch": 0.52, "learning_rate": 2.6896700659868027e-06, "loss": 0.0008, "step": 4361 }, { "epoch": 0.52, "learning_rate": 2.689598080383923e-06, "loss": 0.0049, "step": 4362 }, { "epoch": 0.52, "learning_rate": 2.689526094781044e-06, "loss": 0.0011, "step": 4363 }, { "epoch": 0.52, "learning_rate": 2.689454109178164e-06, "loss": 0.5078, "step": 4364 }, { "epoch": 0.52, "learning_rate": 2.689382123575285e-06, "loss": 0.0014, "step": 4365 }, { "epoch": 0.52, "learning_rate": 2.6893101379724057e-06, "loss": 0.0008, "step": 4366 }, { "epoch": 0.52, "learning_rate": 2.689238152369526e-06, "loss": 0.6973, "step": 4367 }, { "epoch": 0.52, "learning_rate": 2.689166166766647e-06, "loss": 0.7591, "step": 4368 }, { "epoch": 0.52, "learning_rate": 2.6890941811637672e-06, "loss": 0.0007, "step": 4369 }, { "epoch": 0.52, "learning_rate": 2.689022195560888e-06, "loss": 0.0009, "step": 4370 }, { "epoch": 0.52, "learning_rate": 2.6889502099580084e-06, "loss": 0.0008, "step": 4371 }, { "epoch": 0.52, "learning_rate": 2.688878224355129e-06, "loss": 0.0431, "step": 4372 }, { "epoch": 0.52, "learning_rate": 2.6888062387522495e-06, "loss": 0.4288, "step": 4373 }, { "epoch": 0.52, "learning_rate": 2.6887342531493703e-06, "loss": 0.0015, "step": 4374 }, { "epoch": 0.52, "learning_rate": 2.6886622675464906e-06, "loss": 0.0008, "step": 4375 }, { "epoch": 0.52, "learning_rate": 2.6885902819436114e-06, "loss": 0.0011, "step": 4376 }, { "epoch": 0.52, "learning_rate": 2.6885182963407318e-06, "loss": 0.0009, "step": 4377 }, { "epoch": 0.52, "learning_rate": 2.6884463107378525e-06, "loss": 0.0008, "step": 4378 }, { "epoch": 0.52, "learning_rate": 2.688374325134973e-06, "loss": 0.0006, "step": 4379 }, { "epoch": 0.52, "learning_rate": 2.6883023395320937e-06, "loss": 0.0719, "step": 4380 }, { "epoch": 0.52, "learning_rate": 2.688230353929214e-06, "loss": 0.0011, "step": 4381 }, { "epoch": 0.53, "learning_rate": 2.688158368326335e-06, "loss": 0.0007, "step": 4382 }, { "epoch": 0.53, "learning_rate": 2.6880863827234556e-06, "loss": 0.001, "step": 4383 }, { "epoch": 0.53, "learning_rate": 2.688014397120576e-06, "loss": 0.0013, "step": 4384 }, { "epoch": 0.53, "learning_rate": 2.6879424115176967e-06, "loss": 0.0013, "step": 4385 }, { "epoch": 0.53, "learning_rate": 2.687870425914817e-06, "loss": 0.0009, "step": 4386 }, { "epoch": 0.53, "learning_rate": 2.687798440311938e-06, "loss": 0.0009, "step": 4387 }, { "epoch": 0.53, "learning_rate": 2.687726454709058e-06, "loss": 0.0009, "step": 4388 }, { "epoch": 0.53, "learning_rate": 2.687654469106179e-06, "loss": 0.0009, "step": 4389 }, { "epoch": 0.53, "learning_rate": 2.6875824835032993e-06, "loss": 0.0011, "step": 4390 }, { "epoch": 0.53, "learning_rate": 2.68751049790042e-06, "loss": 0.0008, "step": 4391 }, { "epoch": 0.53, "learning_rate": 2.6874385122975405e-06, "loss": 0.0009, "step": 4392 }, { "epoch": 0.53, "learning_rate": 2.687366526694661e-06, "loss": 0.1059, "step": 4393 }, { "epoch": 0.53, "learning_rate": 2.6872945410917816e-06, "loss": 0.6905, "step": 4394 }, { "epoch": 0.53, "learning_rate": 2.687222555488902e-06, "loss": 0.0006, "step": 4395 }, { "epoch": 0.53, "learning_rate": 2.6871505698860227e-06, "loss": 0.0009, "step": 4396 }, { "epoch": 0.53, "learning_rate": 2.6870785842831435e-06, "loss": 0.0023, "step": 4397 }, { "epoch": 0.53, "learning_rate": 2.6870065986802643e-06, "loss": 0.0008, "step": 4398 }, { "epoch": 0.53, "learning_rate": 2.6869346130773847e-06, "loss": 0.0011, "step": 4399 }, { "epoch": 0.53, "learning_rate": 2.6868626274745054e-06, "loss": 0.0047, "step": 4400 }, { "epoch": 0.53, "learning_rate": 2.686790641871626e-06, "loss": 0.7806, "step": 4401 }, { "epoch": 0.53, "learning_rate": 2.6867186562687466e-06, "loss": 0.0007, "step": 4402 }, { "epoch": 0.53, "learning_rate": 2.686646670665867e-06, "loss": 0.4821, "step": 4403 }, { "epoch": 0.53, "learning_rate": 2.6865746850629873e-06, "loss": 0.0011, "step": 4404 }, { "epoch": 0.53, "learning_rate": 2.686502699460108e-06, "loss": 0.1106, "step": 4405 }, { "epoch": 0.53, "learning_rate": 2.6864307138572284e-06, "loss": 0.0006, "step": 4406 }, { "epoch": 0.53, "learning_rate": 2.686358728254349e-06, "loss": 0.0008, "step": 4407 }, { "epoch": 0.53, "learning_rate": 2.6862867426514695e-06, "loss": 0.1261, "step": 4408 }, { "epoch": 0.53, "learning_rate": 2.6862147570485903e-06, "loss": 0.0011, "step": 4409 }, { "epoch": 0.53, "learning_rate": 2.6861427714457107e-06, "loss": 0.0333, "step": 4410 }, { "epoch": 0.53, "learning_rate": 2.6860707858428315e-06, "loss": 0.0005, "step": 4411 }, { "epoch": 0.53, "learning_rate": 2.6859988002399522e-06, "loss": 0.0008, "step": 4412 }, { "epoch": 0.53, "learning_rate": 2.6859268146370726e-06, "loss": 0.0027, "step": 4413 }, { "epoch": 0.53, "learning_rate": 2.6858548290341934e-06, "loss": 0.2294, "step": 4414 }, { "epoch": 0.53, "learning_rate": 2.6857828434313137e-06, "loss": 0.0007, "step": 4415 }, { "epoch": 0.53, "learning_rate": 2.6857108578284345e-06, "loss": 0.0012, "step": 4416 }, { "epoch": 0.53, "learning_rate": 2.685638872225555e-06, "loss": 0.0008, "step": 4417 }, { "epoch": 0.53, "learning_rate": 2.6855668866226756e-06, "loss": 0.0005, "step": 4418 }, { "epoch": 0.53, "learning_rate": 2.685494901019796e-06, "loss": 0.0007, "step": 4419 }, { "epoch": 0.53, "learning_rate": 2.6854229154169168e-06, "loss": 0.8128, "step": 4420 }, { "epoch": 0.53, "learning_rate": 2.685350929814037e-06, "loss": 0.0005, "step": 4421 }, { "epoch": 0.53, "learning_rate": 2.685278944211158e-06, "loss": 0.0006, "step": 4422 }, { "epoch": 0.53, "learning_rate": 2.6852069586082783e-06, "loss": 0.632, "step": 4423 }, { "epoch": 0.53, "learning_rate": 2.685134973005399e-06, "loss": 0.0032, "step": 4424 }, { "epoch": 0.53, "learning_rate": 2.6850629874025194e-06, "loss": 0.0007, "step": 4425 }, { "epoch": 0.53, "learning_rate": 2.68499100179964e-06, "loss": 0.0022, "step": 4426 }, { "epoch": 0.53, "learning_rate": 2.6849190161967605e-06, "loss": 0.001, "step": 4427 }, { "epoch": 0.53, "learning_rate": 2.6848470305938813e-06, "loss": 0.0012, "step": 4428 }, { "epoch": 0.53, "learning_rate": 2.684775044991002e-06, "loss": 0.0008, "step": 4429 }, { "epoch": 0.53, "learning_rate": 2.6847030593881224e-06, "loss": 0.0383, "step": 4430 }, { "epoch": 0.53, "learning_rate": 2.6846310737852432e-06, "loss": 0.6962, "step": 4431 }, { "epoch": 0.53, "learning_rate": 2.6845590881823636e-06, "loss": 0.0032, "step": 4432 }, { "epoch": 0.53, "learning_rate": 2.6844871025794844e-06, "loss": 0.0015, "step": 4433 }, { "epoch": 0.53, "learning_rate": 2.6844151169766047e-06, "loss": 0.0009, "step": 4434 }, { "epoch": 0.53, "learning_rate": 2.6843431313737255e-06, "loss": 0.0012, "step": 4435 }, { "epoch": 0.53, "learning_rate": 2.684271145770846e-06, "loss": 0.6118, "step": 4436 }, { "epoch": 0.53, "learning_rate": 2.6841991601679666e-06, "loss": 0.0034, "step": 4437 }, { "epoch": 0.53, "learning_rate": 2.684127174565087e-06, "loss": 0.0146, "step": 4438 }, { "epoch": 0.53, "learning_rate": 2.6840551889622073e-06, "loss": 0.0012, "step": 4439 }, { "epoch": 0.53, "learning_rate": 2.683983203359328e-06, "loss": 0.0013, "step": 4440 }, { "epoch": 0.53, "learning_rate": 2.6839112177564485e-06, "loss": 0.0008, "step": 4441 }, { "epoch": 0.53, "learning_rate": 2.6838392321535692e-06, "loss": 0.0307, "step": 4442 }, { "epoch": 0.53, "learning_rate": 2.68376724655069e-06, "loss": 0.8816, "step": 4443 }, { "epoch": 0.53, "learning_rate": 2.683695260947811e-06, "loss": 0.0012, "step": 4444 }, { "epoch": 0.53, "learning_rate": 2.683623275344931e-06, "loss": 0.0015, "step": 4445 }, { "epoch": 0.53, "learning_rate": 2.683551289742052e-06, "loss": 0.0009, "step": 4446 }, { "epoch": 0.53, "learning_rate": 2.6834793041391723e-06, "loss": 0.0018, "step": 4447 }, { "epoch": 0.53, "learning_rate": 2.683407318536293e-06, "loss": 0.0013, "step": 4448 }, { "epoch": 0.53, "learning_rate": 2.6833353329334134e-06, "loss": 0.0007, "step": 4449 }, { "epoch": 0.53, "learning_rate": 2.6832633473305338e-06, "loss": 0.0233, "step": 4450 }, { "epoch": 0.53, "learning_rate": 2.6831913617276546e-06, "loss": 0.0014, "step": 4451 }, { "epoch": 0.53, "learning_rate": 2.683119376124775e-06, "loss": 0.0006, "step": 4452 }, { "epoch": 0.53, "learning_rate": 2.6830473905218957e-06, "loss": 0.0007, "step": 4453 }, { "epoch": 0.53, "learning_rate": 2.682975404919016e-06, "loss": 0.1658, "step": 4454 }, { "epoch": 0.53, "learning_rate": 2.682903419316137e-06, "loss": 0.7666, "step": 4455 }, { "epoch": 0.53, "learning_rate": 2.682831433713257e-06, "loss": 0.0801, "step": 4456 }, { "epoch": 0.53, "learning_rate": 2.682759448110378e-06, "loss": 0.0013, "step": 4457 }, { "epoch": 0.53, "learning_rate": 2.6826874625074987e-06, "loss": 0.001, "step": 4458 }, { "epoch": 0.53, "learning_rate": 2.682615476904619e-06, "loss": 0.7535, "step": 4459 }, { "epoch": 0.53, "learning_rate": 2.68254349130174e-06, "loss": 0.0005, "step": 4460 }, { "epoch": 0.53, "learning_rate": 2.6824715056988606e-06, "loss": 0.0017, "step": 4461 }, { "epoch": 0.53, "learning_rate": 2.682399520095981e-06, "loss": 0.0007, "step": 4462 }, { "epoch": 0.53, "learning_rate": 2.6823275344931014e-06, "loss": 0.0009, "step": 4463 }, { "epoch": 0.53, "learning_rate": 2.682255548890222e-06, "loss": 0.0009, "step": 4464 }, { "epoch": 0.54, "learning_rate": 2.6821835632873425e-06, "loss": 0.002, "step": 4465 }, { "epoch": 0.54, "learning_rate": 2.6821115776844633e-06, "loss": 0.0009, "step": 4466 }, { "epoch": 0.54, "learning_rate": 2.6820395920815836e-06, "loss": 0.1363, "step": 4467 }, { "epoch": 0.54, "learning_rate": 2.6819676064787044e-06, "loss": 0.2131, "step": 4468 }, { "epoch": 0.54, "learning_rate": 2.6818956208758248e-06, "loss": 0.0006, "step": 4469 }, { "epoch": 0.54, "learning_rate": 2.6818236352729455e-06, "loss": 0.0007, "step": 4470 }, { "epoch": 0.54, "learning_rate": 2.681751649670066e-06, "loss": 0.0006, "step": 4471 }, { "epoch": 0.54, "learning_rate": 2.6816796640671867e-06, "loss": 0.0007, "step": 4472 }, { "epoch": 0.54, "learning_rate": 2.681607678464307e-06, "loss": 0.0466, "step": 4473 }, { "epoch": 0.54, "learning_rate": 2.681535692861428e-06, "loss": 0.0019, "step": 4474 }, { "epoch": 0.54, "learning_rate": 2.6814637072585486e-06, "loss": 0.0009, "step": 4475 }, { "epoch": 0.54, "learning_rate": 2.681391721655669e-06, "loss": 0.0017, "step": 4476 }, { "epoch": 0.54, "learning_rate": 2.6813197360527897e-06, "loss": 0.0008, "step": 4477 }, { "epoch": 0.54, "learning_rate": 2.68124775044991e-06, "loss": 0.0013, "step": 4478 }, { "epoch": 0.54, "learning_rate": 2.681175764847031e-06, "loss": 0.0009, "step": 4479 }, { "epoch": 0.54, "learning_rate": 2.681103779244151e-06, "loss": 0.7763, "step": 4480 }, { "epoch": 0.54, "learning_rate": 2.681031793641272e-06, "loss": 0.0006, "step": 4481 }, { "epoch": 0.54, "learning_rate": 2.6809598080383923e-06, "loss": 0.0011, "step": 4482 }, { "epoch": 0.54, "learning_rate": 2.680887822435513e-06, "loss": 0.0008, "step": 4483 }, { "epoch": 0.54, "learning_rate": 2.6808158368326335e-06, "loss": 0.0207, "step": 4484 }, { "epoch": 0.54, "learning_rate": 2.680743851229754e-06, "loss": 0.0011, "step": 4485 }, { "epoch": 0.54, "learning_rate": 2.6806718656268746e-06, "loss": 0.0026, "step": 4486 }, { "epoch": 0.54, "learning_rate": 2.680599880023995e-06, "loss": 0.0012, "step": 4487 }, { "epoch": 0.54, "learning_rate": 2.6805278944211157e-06, "loss": 0.0014, "step": 4488 }, { "epoch": 0.54, "learning_rate": 2.6804559088182365e-06, "loss": 0.0007, "step": 4489 }, { "epoch": 0.54, "learning_rate": 2.680383923215357e-06, "loss": 0.0011, "step": 4490 }, { "epoch": 0.54, "learning_rate": 2.6803119376124776e-06, "loss": 0.177, "step": 4491 }, { "epoch": 0.54, "learning_rate": 2.6802399520095984e-06, "loss": 0.001, "step": 4492 }, { "epoch": 0.54, "learning_rate": 2.6801679664067188e-06, "loss": 0.0007, "step": 4493 }, { "epoch": 0.54, "learning_rate": 2.6800959808038396e-06, "loss": 0.0008, "step": 4494 }, { "epoch": 0.54, "learning_rate": 2.68002399520096e-06, "loss": 0.0008, "step": 4495 }, { "epoch": 0.54, "learning_rate": 2.6799520095980803e-06, "loss": 0.0014, "step": 4496 }, { "epoch": 0.54, "learning_rate": 2.679880023995201e-06, "loss": 0.0065, "step": 4497 }, { "epoch": 0.54, "learning_rate": 2.6798080383923214e-06, "loss": 0.0008, "step": 4498 }, { "epoch": 0.54, "learning_rate": 2.679736052789442e-06, "loss": 0.0013, "step": 4499 }, { "epoch": 0.54, "learning_rate": 2.6796640671865625e-06, "loss": 0.001, "step": 4500 }, { "epoch": 0.54, "learning_rate": 2.6795920815836833e-06, "loss": 0.0012, "step": 4501 }, { "epoch": 0.54, "learning_rate": 2.6795200959808037e-06, "loss": 0.001, "step": 4502 }, { "epoch": 0.54, "learning_rate": 2.6794481103779245e-06, "loss": 0.0011, "step": 4503 }, { "epoch": 0.54, "learning_rate": 2.679376124775045e-06, "loss": 0.0007, "step": 4504 }, { "epoch": 0.54, "learning_rate": 2.6793041391721656e-06, "loss": 0.0007, "step": 4505 }, { "epoch": 0.54, "learning_rate": 2.6792321535692864e-06, "loss": 0.001, "step": 4506 }, { "epoch": 0.54, "learning_rate": 2.679160167966407e-06, "loss": 0.0012, "step": 4507 }, { "epoch": 0.54, "learning_rate": 2.6790881823635275e-06, "loss": 0.0008, "step": 4508 }, { "epoch": 0.54, "learning_rate": 2.679016196760648e-06, "loss": 0.0005, "step": 4509 }, { "epoch": 0.54, "learning_rate": 2.6789442111577686e-06, "loss": 0.0009, "step": 4510 }, { "epoch": 0.54, "learning_rate": 2.678872225554889e-06, "loss": 0.0008, "step": 4511 }, { "epoch": 0.54, "learning_rate": 2.6788002399520098e-06, "loss": 0.8233, "step": 4512 }, { "epoch": 0.54, "learning_rate": 2.67872825434913e-06, "loss": 0.7972, "step": 4513 }, { "epoch": 0.54, "learning_rate": 2.678656268746251e-06, "loss": 0.0008, "step": 4514 }, { "epoch": 0.54, "learning_rate": 2.6785842831433713e-06, "loss": 0.0111, "step": 4515 }, { "epoch": 0.54, "learning_rate": 2.678512297540492e-06, "loss": 0.0007, "step": 4516 }, { "epoch": 0.54, "learning_rate": 2.6784403119376124e-06, "loss": 0.7791, "step": 4517 }, { "epoch": 0.54, "learning_rate": 2.678368326334733e-06, "loss": 0.0218, "step": 4518 }, { "epoch": 0.54, "learning_rate": 2.6782963407318535e-06, "loss": 0.0007, "step": 4519 }, { "epoch": 0.54, "learning_rate": 2.6782243551289743e-06, "loss": 0.0008, "step": 4520 }, { "epoch": 0.54, "learning_rate": 2.678152369526095e-06, "loss": 0.0007, "step": 4521 }, { "epoch": 0.54, "learning_rate": 2.6780803839232154e-06, "loss": 0.6227, "step": 4522 }, { "epoch": 0.54, "learning_rate": 2.678008398320336e-06, "loss": 0.2201, "step": 4523 }, { "epoch": 0.54, "learning_rate": 2.6779364127174566e-06, "loss": 0.0008, "step": 4524 }, { "epoch": 0.54, "learning_rate": 2.6778644271145773e-06, "loss": 0.0026, "step": 4525 }, { "epoch": 0.54, "learning_rate": 2.6777924415116977e-06, "loss": 0.0009, "step": 4526 }, { "epoch": 0.54, "learning_rate": 2.6777204559088185e-06, "loss": 0.0006, "step": 4527 }, { "epoch": 0.54, "learning_rate": 2.677648470305939e-06, "loss": 0.0008, "step": 4528 }, { "epoch": 0.54, "learning_rate": 2.6775764847030596e-06, "loss": 0.6679, "step": 4529 }, { "epoch": 0.54, "learning_rate": 2.67750449910018e-06, "loss": 0.0016, "step": 4530 }, { "epoch": 0.54, "learning_rate": 2.6774325134973003e-06, "loss": 0.0011, "step": 4531 }, { "epoch": 0.54, "learning_rate": 2.677360527894421e-06, "loss": 0.9316, "step": 4532 }, { "epoch": 0.54, "learning_rate": 2.6772885422915415e-06, "loss": 0.0011, "step": 4533 }, { "epoch": 0.54, "learning_rate": 2.6772165566886622e-06, "loss": 0.0007, "step": 4534 }, { "epoch": 0.54, "learning_rate": 2.677144571085783e-06, "loss": 0.0031, "step": 4535 }, { "epoch": 0.54, "learning_rate": 2.6770725854829034e-06, "loss": 0.0043, "step": 4536 }, { "epoch": 0.54, "learning_rate": 2.677000599880024e-06, "loss": 0.0024, "step": 4537 }, { "epoch": 0.54, "learning_rate": 2.676928614277145e-06, "loss": 0.2766, "step": 4538 }, { "epoch": 0.54, "learning_rate": 2.6768566286742653e-06, "loss": 0.0035, "step": 4539 }, { "epoch": 0.54, "learning_rate": 2.676784643071386e-06, "loss": 0.853, "step": 4540 }, { "epoch": 0.54, "learning_rate": 2.6767126574685064e-06, "loss": 0.0008, "step": 4541 }, { "epoch": 0.54, "learning_rate": 2.6766406718656268e-06, "loss": 0.5193, "step": 4542 }, { "epoch": 0.54, "learning_rate": 2.6765686862627475e-06, "loss": 0.0006, "step": 4543 }, { "epoch": 0.54, "learning_rate": 2.676496700659868e-06, "loss": 0.0009, "step": 4544 }, { "epoch": 0.54, "learning_rate": 2.6764247150569887e-06, "loss": 0.0007, "step": 4545 }, { "epoch": 0.54, "learning_rate": 2.676352729454109e-06, "loss": 0.0007, "step": 4546 }, { "epoch": 0.54, "learning_rate": 2.67628074385123e-06, "loss": 0.0008, "step": 4547 }, { "epoch": 0.54, "learning_rate": 2.67620875824835e-06, "loss": 0.0011, "step": 4548 }, { "epoch": 0.55, "learning_rate": 2.676136772645471e-06, "loss": 0.0022, "step": 4549 }, { "epoch": 0.55, "learning_rate": 2.6760647870425913e-06, "loss": 0.6488, "step": 4550 }, { "epoch": 0.55, "learning_rate": 2.675992801439712e-06, "loss": 0.0004, "step": 4551 }, { "epoch": 0.55, "learning_rate": 2.675920815836833e-06, "loss": 0.0007, "step": 4552 }, { "epoch": 0.55, "learning_rate": 2.6758488302339536e-06, "loss": 0.7553, "step": 4553 }, { "epoch": 0.55, "learning_rate": 2.675776844631074e-06, "loss": 0.0017, "step": 4554 }, { "epoch": 0.55, "learning_rate": 2.6757048590281943e-06, "loss": 0.0682, "step": 4555 }, { "epoch": 0.55, "learning_rate": 2.675632873425315e-06, "loss": 0.0018, "step": 4556 }, { "epoch": 0.55, "learning_rate": 2.6755608878224355e-06, "loss": 0.0007, "step": 4557 }, { "epoch": 0.55, "learning_rate": 2.6754889022195563e-06, "loss": 0.0007, "step": 4558 }, { "epoch": 0.55, "learning_rate": 2.6754169166166766e-06, "loss": 0.0006, "step": 4559 }, { "epoch": 0.55, "learning_rate": 2.6753449310137974e-06, "loss": 0.0012, "step": 4560 }, { "epoch": 0.55, "learning_rate": 2.6752729454109177e-06, "loss": 0.0005, "step": 4561 }, { "epoch": 0.55, "learning_rate": 2.6752009598080385e-06, "loss": 0.0009, "step": 4562 }, { "epoch": 0.55, "learning_rate": 2.675128974205159e-06, "loss": 0.0008, "step": 4563 }, { "epoch": 0.55, "learning_rate": 2.6750569886022797e-06, "loss": 0.0011, "step": 4564 }, { "epoch": 0.55, "learning_rate": 2.6749850029994e-06, "loss": 0.0006, "step": 4565 }, { "epoch": 0.55, "learning_rate": 2.674913017396521e-06, "loss": 0.0205, "step": 4566 }, { "epoch": 0.55, "learning_rate": 2.6748410317936416e-06, "loss": 0.0013, "step": 4567 }, { "epoch": 0.55, "learning_rate": 2.674769046190762e-06, "loss": 0.0018, "step": 4568 }, { "epoch": 0.55, "learning_rate": 2.6746970605878827e-06, "loss": 0.0007, "step": 4569 }, { "epoch": 0.55, "learning_rate": 2.674625074985003e-06, "loss": 0.0005, "step": 4570 }, { "epoch": 0.55, "learning_rate": 2.674553089382124e-06, "loss": 0.0008, "step": 4571 }, { "epoch": 0.55, "learning_rate": 2.674481103779244e-06, "loss": 0.0007, "step": 4572 }, { "epoch": 0.55, "learning_rate": 2.674409118176365e-06, "loss": 0.0467, "step": 4573 }, { "epoch": 0.55, "learning_rate": 2.6743371325734853e-06, "loss": 0.0012, "step": 4574 }, { "epoch": 0.55, "learning_rate": 2.674265146970606e-06, "loss": 0.6324, "step": 4575 }, { "epoch": 0.55, "learning_rate": 2.6741931613677265e-06, "loss": 0.0011, "step": 4576 }, { "epoch": 0.55, "learning_rate": 2.674121175764847e-06, "loss": 0.0012, "step": 4577 }, { "epoch": 0.55, "learning_rate": 2.6740491901619676e-06, "loss": 0.0015, "step": 4578 }, { "epoch": 0.55, "learning_rate": 2.673977204559088e-06, "loss": 0.0008, "step": 4579 }, { "epoch": 0.55, "learning_rate": 2.6739052189562087e-06, "loss": 0.3389, "step": 4580 }, { "epoch": 0.55, "learning_rate": 2.6738332333533295e-06, "loss": 0.0008, "step": 4581 }, { "epoch": 0.55, "learning_rate": 2.67376124775045e-06, "loss": 0.0011, "step": 4582 }, { "epoch": 0.55, "learning_rate": 2.6736892621475706e-06, "loss": 0.0007, "step": 4583 }, { "epoch": 0.55, "learning_rate": 2.6736172765446914e-06, "loss": 0.0007, "step": 4584 }, { "epoch": 0.55, "learning_rate": 2.6735452909418118e-06, "loss": 0.0014, "step": 4585 }, { "epoch": 0.55, "learning_rate": 2.6734733053389326e-06, "loss": 0.001, "step": 4586 }, { "epoch": 0.55, "learning_rate": 2.673401319736053e-06, "loss": 0.0222, "step": 4587 }, { "epoch": 0.55, "learning_rate": 2.6733293341331737e-06, "loss": 0.0019, "step": 4588 }, { "epoch": 0.55, "learning_rate": 2.673257348530294e-06, "loss": 0.0007, "step": 4589 }, { "epoch": 0.55, "learning_rate": 2.6731853629274144e-06, "loss": 0.0033, "step": 4590 }, { "epoch": 0.55, "learning_rate": 2.673113377324535e-06, "loss": 0.0289, "step": 4591 }, { "epoch": 0.55, "learning_rate": 2.6730413917216555e-06, "loss": 0.0009, "step": 4592 }, { "epoch": 0.55, "learning_rate": 2.6729694061187763e-06, "loss": 0.0008, "step": 4593 }, { "epoch": 0.55, "learning_rate": 2.6728974205158967e-06, "loss": 0.0016, "step": 4594 }, { "epoch": 0.55, "learning_rate": 2.6728254349130174e-06, "loss": 0.7489, "step": 4595 }, { "epoch": 0.55, "learning_rate": 2.672753449310138e-06, "loss": 0.0027, "step": 4596 }, { "epoch": 0.55, "learning_rate": 2.6726814637072586e-06, "loss": 0.0009, "step": 4597 }, { "epoch": 0.55, "learning_rate": 2.6726094781043794e-06, "loss": 0.0006, "step": 4598 }, { "epoch": 0.55, "learning_rate": 2.6725374925015e-06, "loss": 0.0512, "step": 4599 }, { "epoch": 0.55, "learning_rate": 2.6724655068986205e-06, "loss": 0.0013, "step": 4600 }, { "epoch": 0.55, "learning_rate": 2.672393521295741e-06, "loss": 0.0017, "step": 4601 }, { "epoch": 0.55, "learning_rate": 2.6723215356928616e-06, "loss": 0.0006, "step": 4602 }, { "epoch": 0.55, "learning_rate": 2.672249550089982e-06, "loss": 0.0005, "step": 4603 }, { "epoch": 0.55, "learning_rate": 2.6721775644871028e-06, "loss": 0.0008, "step": 4604 }, { "epoch": 0.55, "learning_rate": 2.672105578884223e-06, "loss": 0.001, "step": 4605 }, { "epoch": 0.55, "learning_rate": 2.672033593281344e-06, "loss": 0.0006, "step": 4606 }, { "epoch": 0.55, "learning_rate": 2.6719616076784642e-06, "loss": 1.1686, "step": 4607 }, { "epoch": 0.55, "learning_rate": 2.671889622075585e-06, "loss": 0.0009, "step": 4608 }, { "epoch": 0.55, "learning_rate": 2.6718176364727054e-06, "loss": 0.0008, "step": 4609 }, { "epoch": 0.55, "learning_rate": 2.671745650869826e-06, "loss": 0.0006, "step": 4610 }, { "epoch": 0.55, "learning_rate": 2.6716736652669465e-06, "loss": 0.0006, "step": 4611 }, { "epoch": 0.55, "learning_rate": 2.6716016796640673e-06, "loss": 0.0009, "step": 4612 }, { "epoch": 0.55, "learning_rate": 2.6715296940611876e-06, "loss": 0.001, "step": 4613 }, { "epoch": 0.55, "learning_rate": 2.6714577084583084e-06, "loss": 0.0012, "step": 4614 }, { "epoch": 0.55, "learning_rate": 2.671385722855429e-06, "loss": 0.0006, "step": 4615 }, { "epoch": 0.55, "learning_rate": 2.6713137372525496e-06, "loss": 0.0234, "step": 4616 }, { "epoch": 0.55, "learning_rate": 2.6712417516496703e-06, "loss": 0.0006, "step": 4617 }, { "epoch": 0.55, "learning_rate": 2.6711697660467907e-06, "loss": 0.0009, "step": 4618 }, { "epoch": 0.55, "learning_rate": 2.6710977804439115e-06, "loss": 0.0012, "step": 4619 }, { "epoch": 0.55, "learning_rate": 2.671025794841032e-06, "loss": 0.0007, "step": 4620 }, { "epoch": 0.55, "learning_rate": 2.6709538092381526e-06, "loss": 0.001, "step": 4621 }, { "epoch": 0.55, "learning_rate": 2.670881823635273e-06, "loss": 0.0011, "step": 4622 }, { "epoch": 0.55, "learning_rate": 2.6708098380323933e-06, "loss": 0.0006, "step": 4623 }, { "epoch": 0.55, "learning_rate": 2.670737852429514e-06, "loss": 0.0005, "step": 4624 }, { "epoch": 0.55, "learning_rate": 2.6706658668266344e-06, "loss": 0.004, "step": 4625 }, { "epoch": 0.55, "learning_rate": 2.6705938812237552e-06, "loss": 0.0008, "step": 4626 }, { "epoch": 0.55, "learning_rate": 2.6705218956208756e-06, "loss": 0.0008, "step": 4627 }, { "epoch": 0.55, "learning_rate": 2.6704499100179964e-06, "loss": 0.0008, "step": 4628 }, { "epoch": 0.55, "learning_rate": 2.670377924415117e-06, "loss": 0.0412, "step": 4629 }, { "epoch": 0.55, "learning_rate": 2.670305938812238e-06, "loss": 0.0007, "step": 4630 }, { "epoch": 0.55, "learning_rate": 2.6702339532093583e-06, "loss": 0.0007, "step": 4631 }, { "epoch": 0.56, "learning_rate": 2.670161967606479e-06, "loss": 0.0006, "step": 4632 }, { "epoch": 0.56, "learning_rate": 2.6700899820035994e-06, "loss": 0.0008, "step": 4633 }, { "epoch": 0.56, "learning_rate": 2.67001799640072e-06, "loss": 0.021, "step": 4634 }, { "epoch": 0.56, "learning_rate": 2.6699460107978405e-06, "loss": 0.0007, "step": 4635 }, { "epoch": 0.56, "learning_rate": 2.669874025194961e-06, "loss": 0.0005, "step": 4636 }, { "epoch": 0.56, "learning_rate": 2.6698020395920817e-06, "loss": 0.0008, "step": 4637 }, { "epoch": 0.56, "learning_rate": 2.669730053989202e-06, "loss": 0.0013, "step": 4638 }, { "epoch": 0.56, "learning_rate": 2.669658068386323e-06, "loss": 0.0025, "step": 4639 }, { "epoch": 0.56, "learning_rate": 2.669586082783443e-06, "loss": 0.5693, "step": 4640 }, { "epoch": 0.56, "learning_rate": 2.669514097180564e-06, "loss": 0.5163, "step": 4641 }, { "epoch": 0.56, "learning_rate": 2.6694421115776843e-06, "loss": 0.0007, "step": 4642 }, { "epoch": 0.56, "learning_rate": 2.669370125974805e-06, "loss": 0.0005, "step": 4643 }, { "epoch": 0.56, "learning_rate": 2.669298140371926e-06, "loss": 0.001, "step": 4644 }, { "epoch": 0.56, "learning_rate": 2.669226154769046e-06, "loss": 0.7721, "step": 4645 }, { "epoch": 0.56, "learning_rate": 2.669154169166167e-06, "loss": 0.0006, "step": 4646 }, { "epoch": 0.56, "learning_rate": 2.6690821835632873e-06, "loss": 0.0008, "step": 4647 }, { "epoch": 0.56, "learning_rate": 2.669010197960408e-06, "loss": 0.7938, "step": 4648 }, { "epoch": 0.56, "learning_rate": 2.6689382123575285e-06, "loss": 0.0007, "step": 4649 }, { "epoch": 0.56, "learning_rate": 2.6688662267546493e-06, "loss": 0.5352, "step": 4650 }, { "epoch": 0.56, "learning_rate": 2.6687942411517696e-06, "loss": 0.0005, "step": 4651 }, { "epoch": 0.56, "learning_rate": 2.6687222555488904e-06, "loss": 0.001, "step": 4652 }, { "epoch": 0.56, "learning_rate": 2.6686502699460107e-06, "loss": 0.0021, "step": 4653 }, { "epoch": 0.56, "learning_rate": 2.6685782843431315e-06, "loss": 0.0518, "step": 4654 }, { "epoch": 0.56, "learning_rate": 2.668506298740252e-06, "loss": 0.001, "step": 4655 }, { "epoch": 0.56, "learning_rate": 2.6684343131373727e-06, "loss": 0.0008, "step": 4656 }, { "epoch": 0.56, "learning_rate": 2.668362327534493e-06, "loss": 0.0051, "step": 4657 }, { "epoch": 0.56, "learning_rate": 2.6682903419316138e-06, "loss": 0.0262, "step": 4658 }, { "epoch": 0.56, "learning_rate": 2.668218356328734e-06, "loss": 0.0007, "step": 4659 }, { "epoch": 0.56, "learning_rate": 2.668146370725855e-06, "loss": 0.0006, "step": 4660 }, { "epoch": 0.56, "learning_rate": 2.6680743851229757e-06, "loss": 0.001, "step": 4661 }, { "epoch": 0.56, "learning_rate": 2.668002399520096e-06, "loss": 0.001, "step": 4662 }, { "epoch": 0.56, "learning_rate": 2.667930413917217e-06, "loss": 0.0008, "step": 4663 }, { "epoch": 0.56, "learning_rate": 2.667858428314337e-06, "loss": 0.0006, "step": 4664 }, { "epoch": 0.56, "learning_rate": 2.667786442711458e-06, "loss": 0.0008, "step": 4665 }, { "epoch": 0.56, "learning_rate": 2.6677144571085783e-06, "loss": 0.0092, "step": 4666 }, { "epoch": 0.56, "learning_rate": 2.667642471505699e-06, "loss": 0.0008, "step": 4667 }, { "epoch": 0.56, "learning_rate": 2.6675704859028195e-06, "loss": 0.122, "step": 4668 }, { "epoch": 0.56, "learning_rate": 2.6674985002999402e-06, "loss": 0.2163, "step": 4669 }, { "epoch": 0.56, "learning_rate": 2.6674265146970606e-06, "loss": 0.2981, "step": 4670 }, { "epoch": 0.56, "learning_rate": 2.667354529094181e-06, "loss": 0.0006, "step": 4671 }, { "epoch": 0.56, "learning_rate": 2.6672825434913017e-06, "loss": 0.001, "step": 4672 }, { "epoch": 0.56, "learning_rate": 2.667210557888422e-06, "loss": 0.0035, "step": 4673 }, { "epoch": 0.56, "learning_rate": 2.667138572285543e-06, "loss": 0.3588, "step": 4674 }, { "epoch": 0.56, "learning_rate": 2.6670665866826636e-06, "loss": 0.0026, "step": 4675 }, { "epoch": 0.56, "learning_rate": 2.6669946010797844e-06, "loss": 0.0005, "step": 4676 }, { "epoch": 0.56, "learning_rate": 2.6669226154769048e-06, "loss": 0.0008, "step": 4677 }, { "epoch": 0.56, "learning_rate": 2.6668506298740255e-06, "loss": 0.0007, "step": 4678 }, { "epoch": 0.56, "learning_rate": 2.666778644271146e-06, "loss": 0.0007, "step": 4679 }, { "epoch": 0.56, "learning_rate": 2.6667066586682667e-06, "loss": 0.0021, "step": 4680 }, { "epoch": 0.56, "learning_rate": 2.666634673065387e-06, "loss": 0.0083, "step": 4681 }, { "epoch": 0.56, "learning_rate": 2.6665626874625074e-06, "loss": 0.0056, "step": 4682 }, { "epoch": 0.56, "learning_rate": 2.666490701859628e-06, "loss": 0.0047, "step": 4683 }, { "epoch": 0.56, "learning_rate": 2.6664187162567485e-06, "loss": 0.001, "step": 4684 }, { "epoch": 0.56, "learning_rate": 2.6663467306538693e-06, "loss": 0.0005, "step": 4685 }, { "epoch": 0.56, "learning_rate": 2.6662747450509897e-06, "loss": 0.6135, "step": 4686 }, { "epoch": 0.56, "learning_rate": 2.6662027594481104e-06, "loss": 0.0008, "step": 4687 }, { "epoch": 0.56, "learning_rate": 2.666130773845231e-06, "loss": 0.0019, "step": 4688 }, { "epoch": 0.56, "learning_rate": 2.6660587882423516e-06, "loss": 0.0322, "step": 4689 }, { "epoch": 0.56, "learning_rate": 2.6659868026394723e-06, "loss": 0.0017, "step": 4690 }, { "epoch": 0.56, "learning_rate": 2.6659148170365927e-06, "loss": 0.0302, "step": 4691 }, { "epoch": 0.56, "learning_rate": 2.6658428314337135e-06, "loss": 0.0009, "step": 4692 }, { "epoch": 0.56, "learning_rate": 2.665770845830834e-06, "loss": 0.0102, "step": 4693 }, { "epoch": 0.56, "learning_rate": 2.6656988602279546e-06, "loss": 0.002, "step": 4694 }, { "epoch": 0.56, "learning_rate": 2.665626874625075e-06, "loss": 0.0006, "step": 4695 }, { "epoch": 0.56, "learning_rate": 2.6655548890221957e-06, "loss": 0.0014, "step": 4696 }, { "epoch": 0.56, "learning_rate": 2.665482903419316e-06, "loss": 0.0006, "step": 4697 }, { "epoch": 0.56, "learning_rate": 2.665410917816437e-06, "loss": 0.0009, "step": 4698 }, { "epoch": 0.56, "learning_rate": 2.6653389322135572e-06, "loss": 0.0218, "step": 4699 }, { "epoch": 0.56, "learning_rate": 2.665266946610678e-06, "loss": 0.0007, "step": 4700 }, { "epoch": 0.56, "learning_rate": 2.6651949610077984e-06, "loss": 0.0174, "step": 4701 }, { "epoch": 0.56, "learning_rate": 2.665122975404919e-06, "loss": 0.0005, "step": 4702 }, { "epoch": 0.56, "learning_rate": 2.6650509898020395e-06, "loss": 0.0006, "step": 4703 }, { "epoch": 0.56, "learning_rate": 2.6649790041991603e-06, "loss": 0.0009, "step": 4704 }, { "epoch": 0.56, "learning_rate": 2.6649070185962806e-06, "loss": 0.0067, "step": 4705 }, { "epoch": 0.56, "learning_rate": 2.6648350329934014e-06, "loss": 0.0058, "step": 4706 }, { "epoch": 0.56, "learning_rate": 2.664763047390522e-06, "loss": 0.0007, "step": 4707 }, { "epoch": 0.56, "learning_rate": 2.6646910617876425e-06, "loss": 0.0005, "step": 4708 }, { "epoch": 0.56, "learning_rate": 2.6646190761847633e-06, "loss": 0.9779, "step": 4709 }, { "epoch": 0.56, "learning_rate": 2.6645470905818837e-06, "loss": 0.0012, "step": 4710 }, { "epoch": 0.56, "learning_rate": 2.6644751049790045e-06, "loss": 0.0015, "step": 4711 }, { "epoch": 0.56, "learning_rate": 2.664403119376125e-06, "loss": 0.0005, "step": 4712 }, { "epoch": 0.56, "learning_rate": 2.6643311337732456e-06, "loss": 0.0006, "step": 4713 }, { "epoch": 0.56, "learning_rate": 2.664259148170366e-06, "loss": 0.0005, "step": 4714 }, { "epoch": 0.57, "learning_rate": 2.6641871625674867e-06, "loss": 0.4797, "step": 4715 }, { "epoch": 0.57, "learning_rate": 2.664115176964607e-06, "loss": 0.0006, "step": 4716 }, { "epoch": 0.57, "learning_rate": 2.6640431913617274e-06, "loss": 0.0053, "step": 4717 }, { "epoch": 0.57, "learning_rate": 2.6639712057588482e-06, "loss": 0.0031, "step": 4718 }, { "epoch": 0.57, "learning_rate": 2.6638992201559686e-06, "loss": 0.0005, "step": 4719 }, { "epoch": 0.57, "learning_rate": 2.6638272345530893e-06, "loss": 0.0006, "step": 4720 }, { "epoch": 0.57, "learning_rate": 2.66375524895021e-06, "loss": 0.0005, "step": 4721 }, { "epoch": 0.57, "learning_rate": 2.663683263347331e-06, "loss": 0.0004, "step": 4722 }, { "epoch": 0.57, "learning_rate": 2.6636112777444513e-06, "loss": 0.9354, "step": 4723 }, { "epoch": 0.57, "learning_rate": 2.663539292141572e-06, "loss": 0.0007, "step": 4724 }, { "epoch": 0.57, "learning_rate": 2.6634673065386924e-06, "loss": 0.0006, "step": 4725 }, { "epoch": 0.57, "learning_rate": 2.663395320935813e-06, "loss": 0.0005, "step": 4726 }, { "epoch": 0.57, "learning_rate": 2.6633233353329335e-06, "loss": 0.0008, "step": 4727 }, { "epoch": 0.57, "learning_rate": 2.663251349730054e-06, "loss": 0.0007, "step": 4728 }, { "epoch": 0.57, "learning_rate": 2.6631793641271747e-06, "loss": 0.5904, "step": 4729 }, { "epoch": 0.57, "learning_rate": 2.663107378524295e-06, "loss": 0.0012, "step": 4730 }, { "epoch": 0.57, "learning_rate": 2.663035392921416e-06, "loss": 0.001, "step": 4731 }, { "epoch": 0.57, "learning_rate": 2.662963407318536e-06, "loss": 0.0012, "step": 4732 }, { "epoch": 0.57, "learning_rate": 2.662891421715657e-06, "loss": 0.0011, "step": 4733 }, { "epoch": 0.57, "learning_rate": 2.6628194361127773e-06, "loss": 0.0007, "step": 4734 }, { "epoch": 0.57, "learning_rate": 2.662747450509898e-06, "loss": 0.0011, "step": 4735 }, { "epoch": 0.57, "learning_rate": 2.662675464907019e-06, "loss": 0.0008, "step": 4736 }, { "epoch": 0.57, "learning_rate": 2.662603479304139e-06, "loss": 0.8055, "step": 4737 }, { "epoch": 0.57, "learning_rate": 2.66253149370126e-06, "loss": 0.0011, "step": 4738 }, { "epoch": 0.57, "learning_rate": 2.6624595080983803e-06, "loss": 0.0006, "step": 4739 }, { "epoch": 0.57, "learning_rate": 2.662387522495501e-06, "loss": 0.0009, "step": 4740 }, { "epoch": 0.57, "learning_rate": 2.6623155368926215e-06, "loss": 0.0006, "step": 4741 }, { "epoch": 0.57, "learning_rate": 2.6622435512897422e-06, "loss": 0.0005, "step": 4742 }, { "epoch": 0.57, "learning_rate": 2.6621715656868626e-06, "loss": 0.618, "step": 4743 }, { "epoch": 0.57, "learning_rate": 2.6620995800839834e-06, "loss": 0.0024, "step": 4744 }, { "epoch": 0.57, "learning_rate": 2.6620275944811037e-06, "loss": 0.0012, "step": 4745 }, { "epoch": 0.57, "learning_rate": 2.6619556088782245e-06, "loss": 0.0009, "step": 4746 }, { "epoch": 0.57, "learning_rate": 2.661883623275345e-06, "loss": 0.0006, "step": 4747 }, { "epoch": 0.57, "learning_rate": 2.6618116376724656e-06, "loss": 0.0008, "step": 4748 }, { "epoch": 0.57, "learning_rate": 2.661739652069586e-06, "loss": 0.0008, "step": 4749 }, { "epoch": 0.57, "learning_rate": 2.6616676664667064e-06, "loss": 0.0007, "step": 4750 }, { "epoch": 0.57, "learning_rate": 2.661595680863827e-06, "loss": 0.8268, "step": 4751 }, { "epoch": 0.57, "learning_rate": 2.661523695260948e-06, "loss": 0.0007, "step": 4752 }, { "epoch": 0.57, "learning_rate": 2.6614517096580687e-06, "loss": 0.0014, "step": 4753 }, { "epoch": 0.57, "learning_rate": 2.661379724055189e-06, "loss": 0.6005, "step": 4754 }, { "epoch": 0.57, "learning_rate": 2.66130773845231e-06, "loss": 0.9787, "step": 4755 }, { "epoch": 0.57, "learning_rate": 2.66123575284943e-06, "loss": 0.006, "step": 4756 }, { "epoch": 0.57, "learning_rate": 2.661163767246551e-06, "loss": 0.001, "step": 4757 }, { "epoch": 0.57, "learning_rate": 2.6610917816436713e-06, "loss": 0.0009, "step": 4758 }, { "epoch": 0.57, "learning_rate": 2.661019796040792e-06, "loss": 0.9076, "step": 4759 }, { "epoch": 0.57, "learning_rate": 2.6609478104379124e-06, "loss": 0.0005, "step": 4760 }, { "epoch": 0.57, "learning_rate": 2.6608758248350332e-06, "loss": 0.0008, "step": 4761 }, { "epoch": 0.57, "learning_rate": 2.6608038392321536e-06, "loss": 0.0006, "step": 4762 }, { "epoch": 0.57, "learning_rate": 2.660731853629274e-06, "loss": 0.0006, "step": 4763 }, { "epoch": 0.57, "learning_rate": 2.6606598680263947e-06, "loss": 0.0006, "step": 4764 }, { "epoch": 0.57, "learning_rate": 2.660587882423515e-06, "loss": 0.0008, "step": 4765 }, { "epoch": 0.57, "learning_rate": 2.660515896820636e-06, "loss": 0.0009, "step": 4766 }, { "epoch": 0.57, "learning_rate": 2.6604439112177566e-06, "loss": 0.0014, "step": 4767 }, { "epoch": 0.57, "learning_rate": 2.660371925614877e-06, "loss": 0.0009, "step": 4768 }, { "epoch": 0.57, "learning_rate": 2.6602999400119978e-06, "loss": 0.0011, "step": 4769 }, { "epoch": 0.57, "learning_rate": 2.6602279544091185e-06, "loss": 0.001, "step": 4770 }, { "epoch": 0.57, "learning_rate": 2.660155968806239e-06, "loss": 0.0007, "step": 4771 }, { "epoch": 0.57, "learning_rate": 2.6600839832033597e-06, "loss": 0.0007, "step": 4772 }, { "epoch": 0.57, "learning_rate": 2.66001199760048e-06, "loss": 0.0048, "step": 4773 }, { "epoch": 0.57, "learning_rate": 2.6599400119976004e-06, "loss": 0.0019, "step": 4774 }, { "epoch": 0.57, "learning_rate": 2.659868026394721e-06, "loss": 0.0015, "step": 4775 }, { "epoch": 0.57, "learning_rate": 2.6597960407918415e-06, "loss": 0.8491, "step": 4776 }, { "epoch": 0.57, "learning_rate": 2.6597240551889623e-06, "loss": 0.0007, "step": 4777 }, { "epoch": 0.57, "learning_rate": 2.6596520695860826e-06, "loss": 0.2215, "step": 4778 }, { "epoch": 0.57, "learning_rate": 2.6595800839832034e-06, "loss": 0.0008, "step": 4779 }, { "epoch": 0.57, "learning_rate": 2.6595080983803238e-06, "loss": 0.0007, "step": 4780 }, { "epoch": 0.57, "learning_rate": 2.6594361127774446e-06, "loss": 0.001, "step": 4781 }, { "epoch": 0.57, "learning_rate": 2.659364127174565e-06, "loss": 0.5702, "step": 4782 }, { "epoch": 0.57, "learning_rate": 2.6592921415716857e-06, "loss": 0.001, "step": 4783 }, { "epoch": 0.57, "learning_rate": 2.6592201559688065e-06, "loss": 0.0009, "step": 4784 }, { "epoch": 0.57, "learning_rate": 2.659148170365927e-06, "loss": 0.0008, "step": 4785 }, { "epoch": 0.57, "learning_rate": 2.6590761847630476e-06, "loss": 1.6081, "step": 4786 }, { "epoch": 0.57, "learning_rate": 2.659004199160168e-06, "loss": 0.0009, "step": 4787 }, { "epoch": 0.57, "learning_rate": 2.6589322135572887e-06, "loss": 0.0008, "step": 4788 }, { "epoch": 0.57, "learning_rate": 2.658860227954409e-06, "loss": 0.0014, "step": 4789 }, { "epoch": 0.57, "learning_rate": 2.65878824235153e-06, "loss": 0.0018, "step": 4790 }, { "epoch": 0.57, "learning_rate": 2.6587162567486502e-06, "loss": 0.0012, "step": 4791 }, { "epoch": 0.57, "learning_rate": 2.658644271145771e-06, "loss": 0.0007, "step": 4792 }, { "epoch": 0.57, "learning_rate": 2.6585722855428914e-06, "loss": 0.0006, "step": 4793 }, { "epoch": 0.57, "learning_rate": 2.658500299940012e-06, "loss": 0.001, "step": 4794 }, { "epoch": 0.57, "learning_rate": 2.6584283143371325e-06, "loss": 0.0021, "step": 4795 }, { "epoch": 0.57, "learning_rate": 2.6583563287342533e-06, "loss": 0.0007, "step": 4796 }, { "epoch": 0.57, "learning_rate": 2.6582843431313736e-06, "loss": 1.3042, "step": 4797 }, { "epoch": 0.57, "learning_rate": 2.6582123575284944e-06, "loss": 0.0046, "step": 4798 }, { "epoch": 0.58, "learning_rate": 2.658140371925615e-06, "loss": 0.0007, "step": 4799 }, { "epoch": 0.58, "learning_rate": 2.6580683863227355e-06, "loss": 0.0026, "step": 4800 }, { "epoch": 0.58, "learning_rate": 2.6579964007198563e-06, "loss": 0.0013, "step": 4801 }, { "epoch": 0.58, "learning_rate": 2.6579244151169767e-06, "loss": 0.0021, "step": 4802 }, { "epoch": 0.58, "learning_rate": 2.6578524295140975e-06, "loss": 0.0009, "step": 4803 }, { "epoch": 0.58, "learning_rate": 2.657780443911218e-06, "loss": 0.0008, "step": 4804 }, { "epoch": 0.58, "learning_rate": 2.6577084583083386e-06, "loss": 0.1166, "step": 4805 }, { "epoch": 0.58, "learning_rate": 2.657636472705459e-06, "loss": 0.0007, "step": 4806 }, { "epoch": 0.58, "learning_rate": 2.6575644871025797e-06, "loss": 0.0007, "step": 4807 }, { "epoch": 0.58, "learning_rate": 2.6574925014997e-06, "loss": 0.0008, "step": 4808 }, { "epoch": 0.58, "learning_rate": 2.6574205158968204e-06, "loss": 0.0006, "step": 4809 }, { "epoch": 0.58, "learning_rate": 2.657348530293941e-06, "loss": 0.0012, "step": 4810 }, { "epoch": 0.58, "learning_rate": 2.6572765446910616e-06, "loss": 0.0006, "step": 4811 }, { "epoch": 0.58, "learning_rate": 2.6572045590881823e-06, "loss": 0.0006, "step": 4812 }, { "epoch": 0.58, "learning_rate": 2.657132573485303e-06, "loss": 0.0249, "step": 4813 }, { "epoch": 0.58, "learning_rate": 2.6570605878824235e-06, "loss": 0.0009, "step": 4814 }, { "epoch": 0.58, "learning_rate": 2.6569886022795443e-06, "loss": 0.6022, "step": 4815 }, { "epoch": 0.58, "learning_rate": 2.656916616676665e-06, "loss": 0.7213, "step": 4816 }, { "epoch": 0.58, "learning_rate": 2.6568446310737854e-06, "loss": 0.0006, "step": 4817 }, { "epoch": 0.58, "learning_rate": 2.656772645470906e-06, "loss": 0.0019, "step": 4818 }, { "epoch": 0.58, "learning_rate": 2.6567006598680265e-06, "loss": 0.6161, "step": 4819 }, { "epoch": 0.58, "learning_rate": 2.656628674265147e-06, "loss": 0.0011, "step": 4820 }, { "epoch": 0.58, "learning_rate": 2.6565566886622677e-06, "loss": 0.9484, "step": 4821 }, { "epoch": 0.58, "learning_rate": 2.656484703059388e-06, "loss": 0.0011, "step": 4822 }, { "epoch": 0.58, "learning_rate": 2.656412717456509e-06, "loss": 0.0008, "step": 4823 }, { "epoch": 0.58, "learning_rate": 2.656340731853629e-06, "loss": 0.0011, "step": 4824 }, { "epoch": 0.58, "learning_rate": 2.65626874625075e-06, "loss": 0.0011, "step": 4825 }, { "epoch": 0.58, "learning_rate": 2.6561967606478703e-06, "loss": 0.3039, "step": 4826 }, { "epoch": 0.58, "learning_rate": 2.656124775044991e-06, "loss": 0.0011, "step": 4827 }, { "epoch": 0.58, "learning_rate": 2.6560527894421114e-06, "loss": 0.3842, "step": 4828 }, { "epoch": 0.58, "learning_rate": 2.655980803839232e-06, "loss": 0.0016, "step": 4829 }, { "epoch": 0.58, "learning_rate": 2.655908818236353e-06, "loss": 0.0006, "step": 4830 }, { "epoch": 0.58, "learning_rate": 2.6558368326334733e-06, "loss": 0.0013, "step": 4831 }, { "epoch": 0.58, "learning_rate": 2.655764847030594e-06, "loss": 0.0014, "step": 4832 }, { "epoch": 0.58, "learning_rate": 2.6556928614277145e-06, "loss": 0.0022, "step": 4833 }, { "epoch": 0.58, "learning_rate": 2.6556208758248352e-06, "loss": 0.0015, "step": 4834 }, { "epoch": 0.58, "learning_rate": 2.6555488902219556e-06, "loss": 0.0006, "step": 4835 }, { "epoch": 0.58, "learning_rate": 2.6554769046190764e-06, "loss": 0.005, "step": 4836 }, { "epoch": 0.58, "learning_rate": 2.6554049190161967e-06, "loss": 0.001, "step": 4837 }, { "epoch": 0.58, "learning_rate": 2.6553329334133175e-06, "loss": 0.3332, "step": 4838 }, { "epoch": 0.58, "learning_rate": 2.655260947810438e-06, "loss": 0.4444, "step": 4839 }, { "epoch": 0.58, "learning_rate": 2.6551889622075586e-06, "loss": 0.0013, "step": 4840 }, { "epoch": 0.58, "learning_rate": 2.655116976604679e-06, "loss": 0.0005, "step": 4841 }, { "epoch": 0.58, "learning_rate": 2.6550449910017998e-06, "loss": 0.0229, "step": 4842 }, { "epoch": 0.58, "learning_rate": 2.65497300539892e-06, "loss": 0.0094, "step": 4843 }, { "epoch": 0.58, "learning_rate": 2.654901019796041e-06, "loss": 0.0009, "step": 4844 }, { "epoch": 0.58, "learning_rate": 2.6548290341931617e-06, "loss": 0.088, "step": 4845 }, { "epoch": 0.58, "learning_rate": 2.654757048590282e-06, "loss": 0.0017, "step": 4846 }, { "epoch": 0.58, "learning_rate": 2.654685062987403e-06, "loss": 0.0008, "step": 4847 }, { "epoch": 0.58, "learning_rate": 2.654613077384523e-06, "loss": 0.001, "step": 4848 }, { "epoch": 0.58, "learning_rate": 2.654541091781644e-06, "loss": 0.0005, "step": 4849 }, { "epoch": 0.58, "learning_rate": 2.6544691061787643e-06, "loss": 0.0064, "step": 4850 }, { "epoch": 0.58, "learning_rate": 2.654397120575885e-06, "loss": 0.0012, "step": 4851 }, { "epoch": 0.58, "learning_rate": 2.6543251349730054e-06, "loss": 0.001, "step": 4852 }, { "epoch": 0.58, "learning_rate": 2.6542531493701262e-06, "loss": 0.0008, "step": 4853 }, { "epoch": 0.58, "learning_rate": 2.6541811637672466e-06, "loss": 0.0019, "step": 4854 }, { "epoch": 0.58, "learning_rate": 2.654109178164367e-06, "loss": 0.0013, "step": 4855 }, { "epoch": 0.58, "learning_rate": 2.6540371925614877e-06, "loss": 0.0011, "step": 4856 }, { "epoch": 0.58, "learning_rate": 2.653965206958608e-06, "loss": 0.0007, "step": 4857 }, { "epoch": 0.58, "learning_rate": 2.653893221355729e-06, "loss": 0.0011, "step": 4858 }, { "epoch": 0.58, "learning_rate": 2.6538212357528496e-06, "loss": 0.0011, "step": 4859 }, { "epoch": 0.58, "learning_rate": 2.65374925014997e-06, "loss": 0.0957, "step": 4860 }, { "epoch": 0.58, "learning_rate": 2.6536772645470907e-06, "loss": 0.001, "step": 4861 }, { "epoch": 0.58, "learning_rate": 2.6536052789442115e-06, "loss": 0.0154, "step": 4862 }, { "epoch": 0.58, "learning_rate": 2.653533293341332e-06, "loss": 0.0072, "step": 4863 }, { "epoch": 0.58, "learning_rate": 2.6534613077384527e-06, "loss": 0.0018, "step": 4864 }, { "epoch": 0.58, "learning_rate": 2.653389322135573e-06, "loss": 0.0006, "step": 4865 }, { "epoch": 0.58, "learning_rate": 2.6533173365326934e-06, "loss": 0.001, "step": 4866 }, { "epoch": 0.58, "learning_rate": 2.653245350929814e-06, "loss": 0.0009, "step": 4867 }, { "epoch": 0.58, "learning_rate": 2.6531733653269345e-06, "loss": 0.0007, "step": 4868 }, { "epoch": 0.58, "learning_rate": 2.6531013797240553e-06, "loss": 0.0004, "step": 4869 }, { "epoch": 0.58, "learning_rate": 2.6530293941211756e-06, "loss": 0.0012, "step": 4870 }, { "epoch": 0.58, "learning_rate": 2.6529574085182964e-06, "loss": 0.0006, "step": 4871 }, { "epoch": 0.58, "learning_rate": 2.6528854229154168e-06, "loss": 0.0016, "step": 4872 }, { "epoch": 0.58, "learning_rate": 2.6528134373125376e-06, "loss": 0.0076, "step": 4873 }, { "epoch": 0.58, "learning_rate": 2.652741451709658e-06, "loss": 0.3102, "step": 4874 }, { "epoch": 0.58, "learning_rate": 2.6526694661067787e-06, "loss": 0.2602, "step": 4875 }, { "epoch": 0.58, "learning_rate": 2.6525974805038995e-06, "loss": 0.0007, "step": 4876 }, { "epoch": 0.58, "learning_rate": 2.65252549490102e-06, "loss": 0.0013, "step": 4877 }, { "epoch": 0.58, "learning_rate": 2.6524535092981406e-06, "loss": 0.0008, "step": 4878 }, { "epoch": 0.58, "learning_rate": 2.652381523695261e-06, "loss": 0.0023, "step": 4879 }, { "epoch": 0.58, "learning_rate": 2.6523095380923817e-06, "loss": 0.0011, "step": 4880 }, { "epoch": 0.58, "learning_rate": 2.652237552489502e-06, "loss": 1.4611, "step": 4881 }, { "epoch": 0.59, "learning_rate": 2.652165566886623e-06, "loss": 0.0011, "step": 4882 }, { "epoch": 0.59, "learning_rate": 2.6520935812837432e-06, "loss": 0.0007, "step": 4883 }, { "epoch": 0.59, "learning_rate": 2.652021595680864e-06, "loss": 0.0007, "step": 4884 }, { "epoch": 0.59, "learning_rate": 2.6519496100779844e-06, "loss": 0.1108, "step": 4885 }, { "epoch": 0.59, "learning_rate": 2.651877624475105e-06, "loss": 0.0018, "step": 4886 }, { "epoch": 0.59, "learning_rate": 2.6518056388722255e-06, "loss": 0.0012, "step": 4887 }, { "epoch": 0.59, "learning_rate": 2.6517336532693463e-06, "loss": 0.0006, "step": 4888 }, { "epoch": 0.59, "learning_rate": 2.6516616676664666e-06, "loss": 0.0012, "step": 4889 }, { "epoch": 0.59, "learning_rate": 2.6515896820635874e-06, "loss": 0.001, "step": 4890 }, { "epoch": 0.59, "learning_rate": 2.6515176964607078e-06, "loss": 0.0007, "step": 4891 }, { "epoch": 0.59, "learning_rate": 2.6514457108578285e-06, "loss": 0.0016, "step": 4892 }, { "epoch": 0.59, "learning_rate": 2.6513737252549493e-06, "loss": 0.0006, "step": 4893 }, { "epoch": 0.59, "learning_rate": 2.6513017396520697e-06, "loss": 0.0007, "step": 4894 }, { "epoch": 0.59, "learning_rate": 2.6512297540491904e-06, "loss": 0.6179, "step": 4895 }, { "epoch": 0.59, "learning_rate": 2.651157768446311e-06, "loss": 0.0017, "step": 4896 }, { "epoch": 0.59, "learning_rate": 2.6510857828434316e-06, "loss": 0.001, "step": 4897 }, { "epoch": 0.59, "learning_rate": 2.651013797240552e-06, "loss": 0.0006, "step": 4898 }, { "epoch": 0.59, "learning_rate": 2.6509418116376727e-06, "loss": 0.0008, "step": 4899 }, { "epoch": 0.59, "learning_rate": 2.650869826034793e-06, "loss": 0.0014, "step": 4900 }, { "epoch": 0.59, "learning_rate": 2.6507978404319134e-06, "loss": 0.0009, "step": 4901 }, { "epoch": 0.59, "learning_rate": 2.650725854829034e-06, "loss": 0.0008, "step": 4902 }, { "epoch": 0.59, "learning_rate": 2.6506538692261546e-06, "loss": 0.0227, "step": 4903 }, { "epoch": 0.59, "learning_rate": 2.6505818836232753e-06, "loss": 0.8241, "step": 4904 }, { "epoch": 0.59, "learning_rate": 2.6505098980203957e-06, "loss": 0.0007, "step": 4905 }, { "epoch": 0.59, "learning_rate": 2.6504379124175165e-06, "loss": 0.0008, "step": 4906 }, { "epoch": 0.59, "learning_rate": 2.6503659268146372e-06, "loss": 0.0009, "step": 4907 }, { "epoch": 0.59, "learning_rate": 2.650293941211758e-06, "loss": 0.6803, "step": 4908 }, { "epoch": 0.59, "learning_rate": 2.6502219556088784e-06, "loss": 0.0006, "step": 4909 }, { "epoch": 0.59, "learning_rate": 2.650149970005999e-06, "loss": 0.0175, "step": 4910 }, { "epoch": 0.59, "learning_rate": 2.6500779844031195e-06, "loss": 0.0005, "step": 4911 }, { "epoch": 0.59, "learning_rate": 2.65000599880024e-06, "loss": 0.0009, "step": 4912 }, { "epoch": 0.59, "learning_rate": 2.6499340131973606e-06, "loss": 0.0008, "step": 4913 }, { "epoch": 0.59, "learning_rate": 2.649862027594481e-06, "loss": 0.0008, "step": 4914 }, { "epoch": 0.59, "learning_rate": 2.6497900419916018e-06, "loss": 0.001, "step": 4915 }, { "epoch": 0.59, "learning_rate": 2.649718056388722e-06, "loss": 0.0018, "step": 4916 }, { "epoch": 0.59, "learning_rate": 2.649646070785843e-06, "loss": 0.0006, "step": 4917 }, { "epoch": 0.59, "learning_rate": 2.6495740851829633e-06, "loss": 0.6909, "step": 4918 }, { "epoch": 0.59, "learning_rate": 2.649502099580084e-06, "loss": 0.0009, "step": 4919 }, { "epoch": 0.59, "learning_rate": 2.6494301139772044e-06, "loss": 0.0012, "step": 4920 }, { "epoch": 0.59, "learning_rate": 2.649358128374325e-06, "loss": 0.2966, "step": 4921 }, { "epoch": 0.59, "learning_rate": 2.649286142771446e-06, "loss": 0.0008, "step": 4922 }, { "epoch": 0.59, "learning_rate": 2.6492141571685663e-06, "loss": 0.0007, "step": 4923 }, { "epoch": 0.59, "learning_rate": 2.649142171565687e-06, "loss": 0.0008, "step": 4924 }, { "epoch": 0.59, "learning_rate": 2.6490701859628074e-06, "loss": 0.0014, "step": 4925 }, { "epoch": 0.59, "learning_rate": 2.6489982003599282e-06, "loss": 0.0011, "step": 4926 }, { "epoch": 0.59, "learning_rate": 2.6489262147570486e-06, "loss": 0.0008, "step": 4927 }, { "epoch": 0.59, "learning_rate": 2.6488542291541694e-06, "loss": 0.0009, "step": 4928 }, { "epoch": 0.59, "learning_rate": 2.6487822435512897e-06, "loss": 0.7197, "step": 4929 }, { "epoch": 0.59, "learning_rate": 2.6487102579484105e-06, "loss": 0.0005, "step": 4930 }, { "epoch": 0.59, "learning_rate": 2.648638272345531e-06, "loss": 0.0005, "step": 4931 }, { "epoch": 0.59, "learning_rate": 2.6485662867426516e-06, "loss": 0.0031, "step": 4932 }, { "epoch": 0.59, "learning_rate": 2.648494301139772e-06, "loss": 0.0005, "step": 4933 }, { "epoch": 0.59, "learning_rate": 2.6484223155368928e-06, "loss": 0.0008, "step": 4934 }, { "epoch": 0.59, "learning_rate": 2.648350329934013e-06, "loss": 0.0289, "step": 4935 }, { "epoch": 0.59, "learning_rate": 2.648278344331134e-06, "loss": 0.9035, "step": 4936 }, { "epoch": 0.59, "learning_rate": 2.6482063587282542e-06, "loss": 0.0005, "step": 4937 }, { "epoch": 0.59, "learning_rate": 2.648134373125375e-06, "loss": 0.0007, "step": 4938 }, { "epoch": 0.59, "learning_rate": 2.648062387522496e-06, "loss": 0.0011, "step": 4939 }, { "epoch": 0.59, "learning_rate": 2.647990401919616e-06, "loss": 0.001, "step": 4940 }, { "epoch": 0.59, "learning_rate": 2.647918416316737e-06, "loss": 0.0007, "step": 4941 }, { "epoch": 0.59, "learning_rate": 2.6478464307138573e-06, "loss": 0.6039, "step": 4942 }, { "epoch": 0.59, "learning_rate": 2.647774445110978e-06, "loss": 0.0016, "step": 4943 }, { "epoch": 0.59, "learning_rate": 2.6477024595080984e-06, "loss": 0.0021, "step": 4944 }, { "epoch": 0.59, "learning_rate": 2.647630473905219e-06, "loss": 0.9256, "step": 4945 }, { "epoch": 0.59, "learning_rate": 2.6475584883023396e-06, "loss": 0.0019, "step": 4946 }, { "epoch": 0.59, "learning_rate": 2.64748650269946e-06, "loss": 0.0005, "step": 4947 }, { "epoch": 0.59, "learning_rate": 2.6474145170965807e-06, "loss": 0.0073, "step": 4948 }, { "epoch": 0.59, "learning_rate": 2.647342531493701e-06, "loss": 0.0009, "step": 4949 }, { "epoch": 0.59, "learning_rate": 2.647270545890822e-06, "loss": 0.0006, "step": 4950 }, { "epoch": 0.59, "learning_rate": 2.647198560287942e-06, "loss": 0.0034, "step": 4951 }, { "epoch": 0.59, "learning_rate": 2.647126574685063e-06, "loss": 0.001, "step": 4952 }, { "epoch": 0.59, "learning_rate": 2.6470545890821837e-06, "loss": 0.0006, "step": 4953 }, { "epoch": 0.59, "learning_rate": 2.6469826034793045e-06, "loss": 0.0009, "step": 4954 }, { "epoch": 0.59, "learning_rate": 2.646910617876425e-06, "loss": 0.001, "step": 4955 }, { "epoch": 0.59, "learning_rate": 2.6468386322735457e-06, "loss": 0.0007, "step": 4956 }, { "epoch": 0.59, "learning_rate": 2.646766646670666e-06, "loss": 0.0007, "step": 4957 }, { "epoch": 0.59, "learning_rate": 2.6466946610677864e-06, "loss": 0.0008, "step": 4958 }, { "epoch": 0.59, "learning_rate": 2.646622675464907e-06, "loss": 0.0008, "step": 4959 }, { "epoch": 0.59, "learning_rate": 2.6465506898620275e-06, "loss": 0.0015, "step": 4960 }, { "epoch": 0.59, "learning_rate": 2.6464787042591483e-06, "loss": 0.0006, "step": 4961 }, { "epoch": 0.59, "learning_rate": 2.6464067186562686e-06, "loss": 0.0007, "step": 4962 }, { "epoch": 0.59, "learning_rate": 2.6463347330533894e-06, "loss": 0.0006, "step": 4963 }, { "epoch": 0.59, "learning_rate": 2.6462627474505098e-06, "loss": 0.001, "step": 4964 }, { "epoch": 0.59, "learning_rate": 2.6461907618476305e-06, "loss": 0.0005, "step": 4965 }, { "epoch": 0.6, "learning_rate": 2.646118776244751e-06, "loss": 0.0006, "step": 4966 }, { "epoch": 0.6, "learning_rate": 2.6460467906418717e-06, "loss": 0.0015, "step": 4967 }, { "epoch": 0.6, "learning_rate": 2.6459748050389925e-06, "loss": 0.0005, "step": 4968 }, { "epoch": 0.6, "learning_rate": 2.645902819436113e-06, "loss": 0.0012, "step": 4969 }, { "epoch": 0.6, "learning_rate": 2.6458308338332336e-06, "loss": 0.0015, "step": 4970 }, { "epoch": 0.6, "learning_rate": 2.645758848230354e-06, "loss": 0.0047, "step": 4971 }, { "epoch": 0.6, "learning_rate": 2.6456868626274747e-06, "loss": 0.0013, "step": 4972 }, { "epoch": 0.6, "learning_rate": 2.645614877024595e-06, "loss": 0.001, "step": 4973 }, { "epoch": 0.6, "learning_rate": 2.645542891421716e-06, "loss": 0.0008, "step": 4974 }, { "epoch": 0.6, "learning_rate": 2.645470905818836e-06, "loss": 0.3298, "step": 4975 }, { "epoch": 0.6, "learning_rate": 2.645398920215957e-06, "loss": 0.0014, "step": 4976 }, { "epoch": 0.6, "learning_rate": 2.6453269346130773e-06, "loss": 0.0011, "step": 4977 }, { "epoch": 0.6, "learning_rate": 2.645254949010198e-06, "loss": 0.0015, "step": 4978 }, { "epoch": 0.6, "learning_rate": 2.6451829634073185e-06, "loss": 0.0005, "step": 4979 }, { "epoch": 0.6, "learning_rate": 2.6451109778044393e-06, "loss": 0.0007, "step": 4980 }, { "epoch": 0.6, "learning_rate": 2.6450389922015596e-06, "loss": 0.001, "step": 4981 }, { "epoch": 0.6, "learning_rate": 2.6449670065986804e-06, "loss": 1.375, "step": 4982 }, { "epoch": 0.6, "learning_rate": 2.6448950209958007e-06, "loss": 0.6003, "step": 4983 }, { "epoch": 0.6, "learning_rate": 2.6448230353929215e-06, "loss": 0.0007, "step": 4984 }, { "epoch": 0.6, "learning_rate": 2.6447510497900423e-06, "loss": 0.0007, "step": 4985 }, { "epoch": 0.6, "learning_rate": 2.6446790641871627e-06, "loss": 0.001, "step": 4986 }, { "epoch": 0.6, "learning_rate": 2.6446070785842834e-06, "loss": 0.0007, "step": 4987 }, { "epoch": 0.6, "learning_rate": 2.644535092981404e-06, "loss": 0.001, "step": 4988 }, { "epoch": 0.6, "learning_rate": 2.6444631073785246e-06, "loss": 0.0007, "step": 4989 }, { "epoch": 0.6, "learning_rate": 2.644391121775645e-06, "loss": 0.0007, "step": 4990 }, { "epoch": 0.6, "learning_rate": 2.6443191361727657e-06, "loss": 0.0006, "step": 4991 }, { "epoch": 0.6, "learning_rate": 2.644247150569886e-06, "loss": 0.3688, "step": 4992 }, { "epoch": 0.6, "learning_rate": 2.6441751649670064e-06, "loss": 0.0006, "step": 4993 }, { "epoch": 0.6, "learning_rate": 2.644103179364127e-06, "loss": 1.5809, "step": 4994 }, { "epoch": 0.6, "learning_rate": 2.6440311937612475e-06, "loss": 0.0008, "step": 4995 }, { "epoch": 0.6, "learning_rate": 2.6439592081583683e-06, "loss": 0.0015, "step": 4996 }, { "epoch": 0.6, "learning_rate": 2.6438872225554887e-06, "loss": 0.3551, "step": 4997 }, { "epoch": 0.6, "learning_rate": 2.6438152369526095e-06, "loss": 0.0006, "step": 4998 }, { "epoch": 0.6, "learning_rate": 2.6437432513497302e-06, "loss": 0.0011, "step": 4999 }, { "epoch": 0.6, "learning_rate": 2.643671265746851e-06, "loss": 0.8125, "step": 5000 }, { "epoch": 0.6, "learning_rate": 2.6435992801439714e-06, "loss": 0.0005, "step": 5001 }, { "epoch": 0.6, "learning_rate": 2.643527294541092e-06, "loss": 0.6772, "step": 5002 }, { "epoch": 0.6, "learning_rate": 2.6434553089382125e-06, "loss": 0.0006, "step": 5003 }, { "epoch": 0.6, "learning_rate": 2.643383323335333e-06, "loss": 0.0007, "step": 5004 }, { "epoch": 0.6, "learning_rate": 2.6433113377324536e-06, "loss": 0.0006, "step": 5005 }, { "epoch": 0.6, "learning_rate": 2.643239352129574e-06, "loss": 0.6694, "step": 5006 }, { "epoch": 0.6, "learning_rate": 2.6431673665266948e-06, "loss": 0.8328, "step": 5007 }, { "epoch": 0.6, "learning_rate": 2.643095380923815e-06, "loss": 0.0014, "step": 5008 }, { "epoch": 0.6, "learning_rate": 2.643023395320936e-06, "loss": 0.0006, "step": 5009 }, { "epoch": 0.6, "learning_rate": 2.6429514097180563e-06, "loss": 0.0006, "step": 5010 }, { "epoch": 0.6, "learning_rate": 2.642879424115177e-06, "loss": 0.0011, "step": 5011 }, { "epoch": 0.6, "learning_rate": 2.6428074385122974e-06, "loss": 0.001, "step": 5012 }, { "epoch": 0.6, "learning_rate": 2.642735452909418e-06, "loss": 0.0014, "step": 5013 }, { "epoch": 0.6, "learning_rate": 2.6426634673065385e-06, "loss": 0.0005, "step": 5014 }, { "epoch": 0.6, "learning_rate": 2.6425914817036593e-06, "loss": 0.8153, "step": 5015 }, { "epoch": 0.6, "learning_rate": 2.64251949610078e-06, "loss": 1.1545, "step": 5016 }, { "epoch": 0.6, "learning_rate": 2.6424475104979004e-06, "loss": 0.0007, "step": 5017 }, { "epoch": 0.6, "learning_rate": 2.6423755248950212e-06, "loss": 0.0031, "step": 5018 }, { "epoch": 0.6, "learning_rate": 2.6423035392921416e-06, "loss": 0.001, "step": 5019 }, { "epoch": 0.6, "learning_rate": 2.6422315536892624e-06, "loss": 0.0022, "step": 5020 }, { "epoch": 0.6, "learning_rate": 2.6421595680863827e-06, "loss": 0.0007, "step": 5021 }, { "epoch": 0.6, "learning_rate": 2.6420875824835035e-06, "loss": 0.7789, "step": 5022 }, { "epoch": 0.6, "learning_rate": 2.642015596880624e-06, "loss": 0.002, "step": 5023 }, { "epoch": 0.6, "learning_rate": 2.6419436112777446e-06, "loss": 0.0007, "step": 5024 }, { "epoch": 0.6, "learning_rate": 2.641871625674865e-06, "loss": 0.0011, "step": 5025 }, { "epoch": 0.6, "learning_rate": 2.6417996400719858e-06, "loss": 0.0009, "step": 5026 }, { "epoch": 0.6, "learning_rate": 2.641727654469106e-06, "loss": 0.0009, "step": 5027 }, { "epoch": 0.6, "learning_rate": 2.6416556688662265e-06, "loss": 0.0384, "step": 5028 }, { "epoch": 0.6, "learning_rate": 2.6415836832633472e-06, "loss": 0.2367, "step": 5029 }, { "epoch": 0.6, "learning_rate": 2.641511697660468e-06, "loss": 0.0524, "step": 5030 }, { "epoch": 0.6, "learning_rate": 2.641439712057589e-06, "loss": 0.0006, "step": 5031 }, { "epoch": 0.6, "learning_rate": 2.641367726454709e-06, "loss": 0.0006, "step": 5032 }, { "epoch": 0.6, "learning_rate": 2.64129574085183e-06, "loss": 0.0007, "step": 5033 }, { "epoch": 0.6, "learning_rate": 2.6412237552489503e-06, "loss": 0.7858, "step": 5034 }, { "epoch": 0.6, "learning_rate": 2.641151769646071e-06, "loss": 0.0006, "step": 5035 }, { "epoch": 0.6, "learning_rate": 2.6410797840431914e-06, "loss": 0.0014, "step": 5036 }, { "epoch": 0.6, "learning_rate": 2.641007798440312e-06, "loss": 0.001, "step": 5037 }, { "epoch": 0.6, "learning_rate": 2.6409358128374326e-06, "loss": 0.0011, "step": 5038 }, { "epoch": 0.6, "learning_rate": 2.640863827234553e-06, "loss": 0.9155, "step": 5039 }, { "epoch": 0.6, "learning_rate": 2.6407918416316737e-06, "loss": 0.5485, "step": 5040 }, { "epoch": 0.6, "learning_rate": 2.640719856028794e-06, "loss": 0.0013, "step": 5041 }, { "epoch": 0.6, "learning_rate": 2.640647870425915e-06, "loss": 0.0006, "step": 5042 }, { "epoch": 0.6, "learning_rate": 2.640575884823035e-06, "loss": 0.9352, "step": 5043 }, { "epoch": 0.6, "learning_rate": 2.640503899220156e-06, "loss": 0.6028, "step": 5044 }, { "epoch": 0.6, "learning_rate": 2.6404319136172767e-06, "loss": 0.0009, "step": 5045 }, { "epoch": 0.6, "learning_rate": 2.640359928014397e-06, "loss": 0.0007, "step": 5046 }, { "epoch": 0.6, "learning_rate": 2.640287942411518e-06, "loss": 0.0135, "step": 5047 }, { "epoch": 0.6, "learning_rate": 2.6402159568086386e-06, "loss": 0.001, "step": 5048 }, { "epoch": 0.61, "learning_rate": 2.640143971205759e-06, "loss": 1.0998, "step": 5049 }, { "epoch": 0.61, "learning_rate": 2.6400719856028798e-06, "loss": 0.0781, "step": 5050 }, { "epoch": 0.61, "learning_rate": 2.64e-06, "loss": 0.0023, "step": 5051 }, { "epoch": 0.61, "learning_rate": 2.6399280143971205e-06, "loss": 0.021, "step": 5052 }, { "epoch": 0.61, "learning_rate": 2.6398560287942413e-06, "loss": 0.0009, "step": 5053 }, { "epoch": 0.61, "learning_rate": 2.6397840431913616e-06, "loss": 0.0011, "step": 5054 }, { "epoch": 0.61, "learning_rate": 2.6397120575884824e-06, "loss": 0.0008, "step": 5055 }, { "epoch": 0.61, "learning_rate": 2.6396400719856028e-06, "loss": 0.0011, "step": 5056 }, { "epoch": 0.61, "learning_rate": 2.6395680863827235e-06, "loss": 0.0011, "step": 5057 }, { "epoch": 0.61, "learning_rate": 2.639496100779844e-06, "loss": 0.0007, "step": 5058 }, { "epoch": 0.61, "learning_rate": 2.6394241151769647e-06, "loss": 0.0008, "step": 5059 }, { "epoch": 0.61, "learning_rate": 2.639352129574085e-06, "loss": 0.001, "step": 5060 }, { "epoch": 0.61, "learning_rate": 2.639280143971206e-06, "loss": 0.3055, "step": 5061 }, { "epoch": 0.61, "learning_rate": 2.6392081583683266e-06, "loss": 0.0006, "step": 5062 }, { "epoch": 0.61, "learning_rate": 2.639136172765447e-06, "loss": 0.0385, "step": 5063 }, { "epoch": 0.61, "learning_rate": 2.6390641871625677e-06, "loss": 0.0008, "step": 5064 }, { "epoch": 0.61, "learning_rate": 2.638992201559688e-06, "loss": 0.0005, "step": 5065 }, { "epoch": 0.61, "learning_rate": 2.638920215956809e-06, "loss": 0.0012, "step": 5066 }, { "epoch": 0.61, "learning_rate": 2.638848230353929e-06, "loss": 0.0006, "step": 5067 }, { "epoch": 0.61, "learning_rate": 2.63877624475105e-06, "loss": 0.0025, "step": 5068 }, { "epoch": 0.61, "learning_rate": 2.6387042591481703e-06, "loss": 0.0014, "step": 5069 }, { "epoch": 0.61, "learning_rate": 2.638632273545291e-06, "loss": 0.0015, "step": 5070 }, { "epoch": 0.61, "learning_rate": 2.6385602879424115e-06, "loss": 0.0004, "step": 5071 }, { "epoch": 0.61, "learning_rate": 2.6384883023395322e-06, "loss": 0.0008, "step": 5072 }, { "epoch": 0.61, "learning_rate": 2.6384163167366526e-06, "loss": 0.0008, "step": 5073 }, { "epoch": 0.61, "learning_rate": 2.638344331133773e-06, "loss": 0.0012, "step": 5074 }, { "epoch": 0.61, "learning_rate": 2.6382723455308937e-06, "loss": 0.0012, "step": 5075 }, { "epoch": 0.61, "learning_rate": 2.6382003599280145e-06, "loss": 0.0006, "step": 5076 }, { "epoch": 0.61, "learning_rate": 2.6381283743251353e-06, "loss": 0.0008, "step": 5077 }, { "epoch": 0.61, "learning_rate": 2.6380563887222556e-06, "loss": 0.6874, "step": 5078 }, { "epoch": 0.61, "learning_rate": 2.6379844031193764e-06, "loss": 0.0009, "step": 5079 }, { "epoch": 0.61, "learning_rate": 2.6379124175164968e-06, "loss": 0.0015, "step": 5080 }, { "epoch": 0.61, "learning_rate": 2.6378404319136176e-06, "loss": 0.0005, "step": 5081 }, { "epoch": 0.61, "learning_rate": 2.637768446310738e-06, "loss": 0.0011, "step": 5082 }, { "epoch": 0.61, "learning_rate": 2.6376964607078587e-06, "loss": 0.0006, "step": 5083 }, { "epoch": 0.61, "learning_rate": 2.637624475104979e-06, "loss": 0.0004, "step": 5084 }, { "epoch": 0.61, "learning_rate": 2.6375524895020994e-06, "loss": 0.0006, "step": 5085 }, { "epoch": 0.61, "learning_rate": 2.63748050389922e-06, "loss": 0.0006, "step": 5086 }, { "epoch": 0.61, "learning_rate": 2.6374085182963405e-06, "loss": 0.0005, "step": 5087 }, { "epoch": 0.61, "learning_rate": 2.6373365326934613e-06, "loss": 0.0006, "step": 5088 }, { "epoch": 0.61, "learning_rate": 2.6372645470905817e-06, "loss": 0.0009, "step": 5089 }, { "epoch": 0.61, "learning_rate": 2.6371925614877025e-06, "loss": 0.001, "step": 5090 }, { "epoch": 0.61, "learning_rate": 2.6371205758848232e-06, "loss": 0.5992, "step": 5091 }, { "epoch": 0.61, "learning_rate": 2.6370485902819436e-06, "loss": 0.0025, "step": 5092 }, { "epoch": 0.61, "learning_rate": 2.6369766046790644e-06, "loss": 0.0008, "step": 5093 }, { "epoch": 0.61, "learning_rate": 2.636904619076185e-06, "loss": 0.0441, "step": 5094 }, { "epoch": 0.61, "learning_rate": 2.6368326334733055e-06, "loss": 0.0007, "step": 5095 }, { "epoch": 0.61, "learning_rate": 2.6367606478704263e-06, "loss": 0.0014, "step": 5096 }, { "epoch": 0.61, "learning_rate": 2.6366886622675466e-06, "loss": 0.0009, "step": 5097 }, { "epoch": 0.61, "learning_rate": 2.636616676664667e-06, "loss": 0.7861, "step": 5098 }, { "epoch": 0.61, "learning_rate": 2.6365446910617878e-06, "loss": 0.0006, "step": 5099 }, { "epoch": 0.61, "learning_rate": 2.636472705458908e-06, "loss": 0.0006, "step": 5100 }, { "epoch": 0.61, "learning_rate": 2.636400719856029e-06, "loss": 0.001, "step": 5101 }, { "epoch": 0.61, "learning_rate": 2.6363287342531493e-06, "loss": 0.0039, "step": 5102 }, { "epoch": 0.61, "learning_rate": 2.63625674865027e-06, "loss": 0.0015, "step": 5103 }, { "epoch": 0.61, "learning_rate": 2.6361847630473904e-06, "loss": 0.8467, "step": 5104 }, { "epoch": 0.61, "learning_rate": 2.636112777444511e-06, "loss": 0.0007, "step": 5105 }, { "epoch": 0.61, "learning_rate": 2.6360407918416315e-06, "loss": 0.0011, "step": 5106 }, { "epoch": 0.61, "learning_rate": 2.6359688062387523e-06, "loss": 0.0009, "step": 5107 }, { "epoch": 0.61, "learning_rate": 2.635896820635873e-06, "loss": 0.001, "step": 5108 }, { "epoch": 0.61, "learning_rate": 2.6358248350329934e-06, "loss": 0.7403, "step": 5109 }, { "epoch": 0.61, "learning_rate": 2.635752849430114e-06, "loss": 0.0005, "step": 5110 }, { "epoch": 0.61, "learning_rate": 2.6356808638272346e-06, "loss": 0.0671, "step": 5111 }, { "epoch": 0.61, "learning_rate": 2.6356088782243553e-06, "loss": 0.4016, "step": 5112 }, { "epoch": 0.61, "learning_rate": 2.6355368926214757e-06, "loss": 0.0007, "step": 5113 }, { "epoch": 0.61, "learning_rate": 2.6354649070185965e-06, "loss": 0.0012, "step": 5114 }, { "epoch": 0.61, "learning_rate": 2.635392921415717e-06, "loss": 0.0028, "step": 5115 }, { "epoch": 0.61, "learning_rate": 2.6353209358128376e-06, "loss": 0.0015, "step": 5116 }, { "epoch": 0.61, "learning_rate": 2.635248950209958e-06, "loss": 0.0012, "step": 5117 }, { "epoch": 0.61, "learning_rate": 2.6351769646070787e-06, "loss": 0.0056, "step": 5118 }, { "epoch": 0.61, "learning_rate": 2.635104979004199e-06, "loss": 0.0005, "step": 5119 }, { "epoch": 0.61, "learning_rate": 2.6350329934013195e-06, "loss": 0.001, "step": 5120 }, { "epoch": 0.61, "learning_rate": 2.6349610077984402e-06, "loss": 0.0009, "step": 5121 }, { "epoch": 0.61, "learning_rate": 2.634889022195561e-06, "loss": 0.423, "step": 5122 }, { "epoch": 0.61, "learning_rate": 2.634817036592682e-06, "loss": 0.0013, "step": 5123 }, { "epoch": 0.61, "learning_rate": 2.634745050989802e-06, "loss": 0.0008, "step": 5124 }, { "epoch": 0.61, "learning_rate": 2.634673065386923e-06, "loss": 0.0013, "step": 5125 }, { "epoch": 0.61, "learning_rate": 2.6346010797840433e-06, "loss": 0.0478, "step": 5126 }, { "epoch": 0.61, "learning_rate": 2.634529094181164e-06, "loss": 0.0012, "step": 5127 }, { "epoch": 0.61, "learning_rate": 2.6344571085782844e-06, "loss": 0.001, "step": 5128 }, { "epoch": 0.61, "learning_rate": 2.634385122975405e-06, "loss": 0.001, "step": 5129 }, { "epoch": 0.61, "learning_rate": 2.6343131373725255e-06, "loss": 0.001, "step": 5130 }, { "epoch": 0.61, "learning_rate": 2.634241151769646e-06, "loss": 0.0011, "step": 5131 }, { "epoch": 0.61, "learning_rate": 2.6341691661667667e-06, "loss": 0.0006, "step": 5132 }, { "epoch": 0.62, "learning_rate": 2.634097180563887e-06, "loss": 0.0007, "step": 5133 }, { "epoch": 0.62, "learning_rate": 2.634025194961008e-06, "loss": 0.0009, "step": 5134 }, { "epoch": 0.62, "learning_rate": 2.633953209358128e-06, "loss": 0.0097, "step": 5135 }, { "epoch": 0.62, "learning_rate": 2.633881223755249e-06, "loss": 0.1879, "step": 5136 }, { "epoch": 0.62, "learning_rate": 2.6338092381523697e-06, "loss": 0.0008, "step": 5137 }, { "epoch": 0.62, "learning_rate": 2.63373725254949e-06, "loss": 0.0013, "step": 5138 }, { "epoch": 0.62, "learning_rate": 2.633665266946611e-06, "loss": 0.0007, "step": 5139 }, { "epoch": 0.62, "learning_rate": 2.6335932813437316e-06, "loss": 0.001, "step": 5140 }, { "epoch": 0.62, "learning_rate": 2.633521295740852e-06, "loss": 0.0011, "step": 5141 }, { "epoch": 0.62, "learning_rate": 2.6334493101379728e-06, "loss": 0.0073, "step": 5142 }, { "epoch": 0.62, "learning_rate": 2.633377324535093e-06, "loss": 0.0034, "step": 5143 }, { "epoch": 0.62, "learning_rate": 2.6333053389322135e-06, "loss": 0.0013, "step": 5144 }, { "epoch": 0.62, "learning_rate": 2.6332333533293343e-06, "loss": 0.0006, "step": 5145 }, { "epoch": 0.62, "learning_rate": 2.6331613677264546e-06, "loss": 0.0011, "step": 5146 }, { "epoch": 0.62, "learning_rate": 2.6330893821235754e-06, "loss": 0.0029, "step": 5147 }, { "epoch": 0.62, "learning_rate": 2.6330173965206957e-06, "loss": 0.1351, "step": 5148 }, { "epoch": 0.62, "learning_rate": 2.6329454109178165e-06, "loss": 0.0016, "step": 5149 }, { "epoch": 0.62, "learning_rate": 2.632873425314937e-06, "loss": 0.8084, "step": 5150 }, { "epoch": 0.62, "learning_rate": 2.6328014397120577e-06, "loss": 0.0009, "step": 5151 }, { "epoch": 0.62, "learning_rate": 2.632729454109178e-06, "loss": 0.0008, "step": 5152 }, { "epoch": 0.62, "learning_rate": 2.632657468506299e-06, "loss": 0.001, "step": 5153 }, { "epoch": 0.62, "learning_rate": 2.6325854829034196e-06, "loss": 0.0012, "step": 5154 }, { "epoch": 0.62, "learning_rate": 2.63251349730054e-06, "loss": 0.0007, "step": 5155 }, { "epoch": 0.62, "learning_rate": 2.6324415116976607e-06, "loss": 0.0008, "step": 5156 }, { "epoch": 0.62, "learning_rate": 2.632369526094781e-06, "loss": 0.0025, "step": 5157 }, { "epoch": 0.62, "learning_rate": 2.632297540491902e-06, "loss": 0.0008, "step": 5158 }, { "epoch": 0.62, "learning_rate": 2.632225554889022e-06, "loss": 0.0007, "step": 5159 }, { "epoch": 0.62, "learning_rate": 2.632153569286143e-06, "loss": 0.0011, "step": 5160 }, { "epoch": 0.62, "learning_rate": 2.6320815836832633e-06, "loss": 0.0017, "step": 5161 }, { "epoch": 0.62, "learning_rate": 2.632009598080384e-06, "loss": 0.0007, "step": 5162 }, { "epoch": 0.62, "learning_rate": 2.6319376124775045e-06, "loss": 0.0008, "step": 5163 }, { "epoch": 0.62, "learning_rate": 2.6318656268746252e-06, "loss": 0.0008, "step": 5164 }, { "epoch": 0.62, "learning_rate": 2.6317936412717456e-06, "loss": 0.0007, "step": 5165 }, { "epoch": 0.62, "learning_rate": 2.631721655668866e-06, "loss": 0.0008, "step": 5166 }, { "epoch": 0.62, "learning_rate": 2.6316496700659867e-06, "loss": 0.0011, "step": 5167 }, { "epoch": 0.62, "learning_rate": 2.6315776844631075e-06, "loss": 0.001, "step": 5168 }, { "epoch": 0.62, "learning_rate": 2.631505698860228e-06, "loss": 0.0007, "step": 5169 }, { "epoch": 0.62, "learning_rate": 2.6314337132573486e-06, "loss": 0.0046, "step": 5170 }, { "epoch": 0.62, "learning_rate": 2.6313617276544694e-06, "loss": 0.0008, "step": 5171 }, { "epoch": 0.62, "learning_rate": 2.6312897420515898e-06, "loss": 0.0008, "step": 5172 }, { "epoch": 0.62, "learning_rate": 2.6312177564487106e-06, "loss": 0.0495, "step": 5173 }, { "epoch": 0.62, "learning_rate": 2.631145770845831e-06, "loss": 0.8532, "step": 5174 }, { "epoch": 0.62, "learning_rate": 2.6310737852429517e-06, "loss": 0.0014, "step": 5175 }, { "epoch": 0.62, "learning_rate": 2.631001799640072e-06, "loss": 0.0005, "step": 5176 }, { "epoch": 0.62, "learning_rate": 2.630929814037193e-06, "loss": 0.0016, "step": 5177 }, { "epoch": 0.62, "learning_rate": 2.630857828434313e-06, "loss": 0.0009, "step": 5178 }, { "epoch": 0.62, "learning_rate": 2.6307858428314335e-06, "loss": 0.0008, "step": 5179 }, { "epoch": 0.62, "learning_rate": 2.6307138572285543e-06, "loss": 0.7253, "step": 5180 }, { "epoch": 0.62, "learning_rate": 2.6306418716256747e-06, "loss": 0.3064, "step": 5181 }, { "epoch": 0.62, "learning_rate": 2.6305698860227954e-06, "loss": 0.0011, "step": 5182 }, { "epoch": 0.62, "learning_rate": 2.630497900419916e-06, "loss": 0.0004, "step": 5183 }, { "epoch": 0.62, "learning_rate": 2.6304259148170366e-06, "loss": 0.0025, "step": 5184 }, { "epoch": 0.62, "learning_rate": 2.6303539292141574e-06, "loss": 0.2088, "step": 5185 }, { "epoch": 0.62, "learning_rate": 2.630281943611278e-06, "loss": 0.0006, "step": 5186 }, { "epoch": 0.62, "learning_rate": 2.6302099580083985e-06, "loss": 0.0011, "step": 5187 }, { "epoch": 0.62, "learning_rate": 2.6301379724055193e-06, "loss": 0.0009, "step": 5188 }, { "epoch": 0.62, "learning_rate": 2.6300659868026396e-06, "loss": 0.0011, "step": 5189 }, { "epoch": 0.62, "learning_rate": 2.62999400119976e-06, "loss": 0.0009, "step": 5190 }, { "epoch": 0.62, "learning_rate": 2.6299220155968808e-06, "loss": 0.001, "step": 5191 }, { "epoch": 0.62, "learning_rate": 2.629850029994001e-06, "loss": 0.0016, "step": 5192 }, { "epoch": 0.62, "learning_rate": 2.629778044391122e-06, "loss": 0.0007, "step": 5193 }, { "epoch": 0.62, "learning_rate": 2.6297060587882422e-06, "loss": 0.0165, "step": 5194 }, { "epoch": 0.62, "learning_rate": 2.629634073185363e-06, "loss": 0.0005, "step": 5195 }, { "epoch": 0.62, "learning_rate": 2.6295620875824834e-06, "loss": 0.0022, "step": 5196 }, { "epoch": 0.62, "learning_rate": 2.629490101979604e-06, "loss": 0.0005, "step": 5197 }, { "epoch": 0.62, "learning_rate": 2.6294181163767245e-06, "loss": 0.0349, "step": 5198 }, { "epoch": 0.62, "learning_rate": 2.6293461307738453e-06, "loss": 0.0006, "step": 5199 }, { "epoch": 0.62, "learning_rate": 2.629274145170966e-06, "loss": 0.0007, "step": 5200 }, { "epoch": 0.62, "learning_rate": 2.6292021595680864e-06, "loss": 0.001, "step": 5201 }, { "epoch": 0.62, "learning_rate": 2.629130173965207e-06, "loss": 0.0006, "step": 5202 }, { "epoch": 0.62, "learning_rate": 2.6290581883623276e-06, "loss": 0.228, "step": 5203 }, { "epoch": 0.62, "learning_rate": 2.6289862027594483e-06, "loss": 0.0009, "step": 5204 }, { "epoch": 0.62, "learning_rate": 2.6289142171565687e-06, "loss": 0.0006, "step": 5205 }, { "epoch": 0.62, "learning_rate": 2.6288422315536895e-06, "loss": 0.2081, "step": 5206 }, { "epoch": 0.62, "learning_rate": 2.62877024595081e-06, "loss": 0.194, "step": 5207 }, { "epoch": 0.62, "learning_rate": 2.6286982603479306e-06, "loss": 0.0017, "step": 5208 }, { "epoch": 0.62, "learning_rate": 2.628626274745051e-06, "loss": 0.0007, "step": 5209 }, { "epoch": 0.62, "learning_rate": 2.6285542891421717e-06, "loss": 0.0005, "step": 5210 }, { "epoch": 0.62, "learning_rate": 2.628482303539292e-06, "loss": 0.0005, "step": 5211 }, { "epoch": 0.62, "learning_rate": 2.6284103179364124e-06, "loss": 0.0006, "step": 5212 }, { "epoch": 0.62, "learning_rate": 2.6283383323335332e-06, "loss": 1.2352, "step": 5213 }, { "epoch": 0.62, "learning_rate": 2.628266346730654e-06, "loss": 0.0007, "step": 5214 }, { "epoch": 0.62, "learning_rate": 2.6281943611277744e-06, "loss": 0.0007, "step": 5215 }, { "epoch": 0.63, "learning_rate": 2.628122375524895e-06, "loss": 0.001, "step": 5216 }, { "epoch": 0.63, "learning_rate": 2.628050389922016e-06, "loss": 0.0169, "step": 5217 }, { "epoch": 0.63, "learning_rate": 2.6279784043191363e-06, "loss": 0.0006, "step": 5218 }, { "epoch": 0.63, "learning_rate": 2.627906418716257e-06, "loss": 0.0005, "step": 5219 }, { "epoch": 0.63, "learning_rate": 2.6278344331133774e-06, "loss": 0.001, "step": 5220 }, { "epoch": 0.63, "learning_rate": 2.627762447510498e-06, "loss": 0.0007, "step": 5221 }, { "epoch": 0.63, "learning_rate": 2.6276904619076185e-06, "loss": 0.0006, "step": 5222 }, { "epoch": 0.63, "learning_rate": 2.6276184763047393e-06, "loss": 0.0166, "step": 5223 }, { "epoch": 0.63, "learning_rate": 2.6275464907018597e-06, "loss": 0.0047, "step": 5224 }, { "epoch": 0.63, "learning_rate": 2.62747450509898e-06, "loss": 0.0005, "step": 5225 }, { "epoch": 0.63, "learning_rate": 2.627402519496101e-06, "loss": 0.0007, "step": 5226 }, { "epoch": 0.63, "learning_rate": 2.627330533893221e-06, "loss": 0.0042, "step": 5227 }, { "epoch": 0.63, "learning_rate": 2.627258548290342e-06, "loss": 0.0007, "step": 5228 }, { "epoch": 0.63, "learning_rate": 2.6271865626874623e-06, "loss": 0.0013, "step": 5229 }, { "epoch": 0.63, "learning_rate": 2.627114577084583e-06, "loss": 0.4901, "step": 5230 }, { "epoch": 0.63, "learning_rate": 2.627042591481704e-06, "loss": 0.0005, "step": 5231 }, { "epoch": 0.63, "learning_rate": 2.6269706058788246e-06, "loss": 0.0008, "step": 5232 }, { "epoch": 0.63, "learning_rate": 2.626898620275945e-06, "loss": 0.0007, "step": 5233 }, { "epoch": 0.63, "learning_rate": 2.6268266346730658e-06, "loss": 0.0005, "step": 5234 }, { "epoch": 0.63, "learning_rate": 2.626754649070186e-06, "loss": 0.2064, "step": 5235 }, { "epoch": 0.63, "learning_rate": 2.6266826634673065e-06, "loss": 0.005, "step": 5236 }, { "epoch": 0.63, "learning_rate": 2.6266106778644273e-06, "loss": 0.0005, "step": 5237 }, { "epoch": 0.63, "learning_rate": 2.6265386922615476e-06, "loss": 0.8681, "step": 5238 }, { "epoch": 0.63, "learning_rate": 2.6264667066586684e-06, "loss": 0.0006, "step": 5239 }, { "epoch": 0.63, "learning_rate": 2.6263947210557887e-06, "loss": 0.0006, "step": 5240 }, { "epoch": 0.63, "learning_rate": 2.6263227354529095e-06, "loss": 0.2271, "step": 5241 }, { "epoch": 0.63, "learning_rate": 2.62625074985003e-06, "loss": 0.0016, "step": 5242 }, { "epoch": 0.63, "learning_rate": 2.6261787642471507e-06, "loss": 0.004, "step": 5243 }, { "epoch": 0.63, "learning_rate": 2.626106778644271e-06, "loss": 0.001, "step": 5244 }, { "epoch": 0.63, "learning_rate": 2.6260347930413918e-06, "loss": 0.0006, "step": 5245 }, { "epoch": 0.63, "learning_rate": 2.6259628074385126e-06, "loss": 0.0005, "step": 5246 }, { "epoch": 0.63, "learning_rate": 2.625890821835633e-06, "loss": 0.0023, "step": 5247 }, { "epoch": 0.63, "learning_rate": 2.6258188362327537e-06, "loss": 0.0009, "step": 5248 }, { "epoch": 0.63, "learning_rate": 2.625746850629874e-06, "loss": 0.0008, "step": 5249 }, { "epoch": 0.63, "learning_rate": 2.625674865026995e-06, "loss": 0.0008, "step": 5250 }, { "epoch": 0.63, "learning_rate": 2.625602879424115e-06, "loss": 0.0009, "step": 5251 }, { "epoch": 0.63, "learning_rate": 2.625530893821236e-06, "loss": 0.3635, "step": 5252 }, { "epoch": 0.63, "learning_rate": 2.6254589082183563e-06, "loss": 0.0011, "step": 5253 }, { "epoch": 0.63, "learning_rate": 2.625386922615477e-06, "loss": 0.0007, "step": 5254 }, { "epoch": 0.63, "learning_rate": 2.6253149370125975e-06, "loss": 0.0007, "step": 5255 }, { "epoch": 0.63, "learning_rate": 2.6252429514097182e-06, "loss": 0.0008, "step": 5256 }, { "epoch": 0.63, "learning_rate": 2.6251709658068386e-06, "loss": 0.502, "step": 5257 }, { "epoch": 0.63, "learning_rate": 2.625098980203959e-06, "loss": 0.0005, "step": 5258 }, { "epoch": 0.63, "learning_rate": 2.6250269946010797e-06, "loss": 0.7417, "step": 5259 }, { "epoch": 0.63, "learning_rate": 2.6249550089982005e-06, "loss": 0.0008, "step": 5260 }, { "epoch": 0.63, "learning_rate": 2.624883023395321e-06, "loss": 0.0005, "step": 5261 }, { "epoch": 0.63, "learning_rate": 2.6248110377924416e-06, "loss": 0.0006, "step": 5262 }, { "epoch": 0.63, "learning_rate": 2.6247390521895624e-06, "loss": 0.0013, "step": 5263 }, { "epoch": 0.63, "learning_rate": 2.6246670665866828e-06, "loss": 0.0009, "step": 5264 }, { "epoch": 0.63, "learning_rate": 2.6245950809838035e-06, "loss": 0.0004, "step": 5265 }, { "epoch": 0.63, "learning_rate": 2.624523095380924e-06, "loss": 0.0006, "step": 5266 }, { "epoch": 0.63, "learning_rate": 2.6244511097780447e-06, "loss": 0.3257, "step": 5267 }, { "epoch": 0.63, "learning_rate": 2.624379124175165e-06, "loss": 0.0009, "step": 5268 }, { "epoch": 0.63, "learning_rate": 2.624307138572286e-06, "loss": 0.2592, "step": 5269 }, { "epoch": 0.63, "learning_rate": 2.624235152969406e-06, "loss": 0.001, "step": 5270 }, { "epoch": 0.63, "learning_rate": 2.6241631673665265e-06, "loss": 0.0006, "step": 5271 }, { "epoch": 0.63, "learning_rate": 2.6240911817636473e-06, "loss": 0.0004, "step": 5272 }, { "epoch": 0.63, "learning_rate": 2.6240191961607677e-06, "loss": 0.0526, "step": 5273 }, { "epoch": 0.63, "learning_rate": 2.6239472105578884e-06, "loss": 0.0006, "step": 5274 }, { "epoch": 0.63, "learning_rate": 2.623875224955009e-06, "loss": 0.0005, "step": 5275 }, { "epoch": 0.63, "learning_rate": 2.6238032393521296e-06, "loss": 0.4418, "step": 5276 }, { "epoch": 0.63, "learning_rate": 2.6237312537492503e-06, "loss": 0.0007, "step": 5277 }, { "epoch": 0.63, "learning_rate": 2.623659268146371e-06, "loss": 0.0007, "step": 5278 }, { "epoch": 0.63, "learning_rate": 2.6235872825434915e-06, "loss": 0.0007, "step": 5279 }, { "epoch": 0.63, "learning_rate": 2.6235152969406123e-06, "loss": 0.0008, "step": 5280 }, { "epoch": 0.63, "learning_rate": 2.6234433113377326e-06, "loss": 0.001, "step": 5281 }, { "epoch": 0.63, "learning_rate": 2.623371325734853e-06, "loss": 0.0007, "step": 5282 }, { "epoch": 0.63, "learning_rate": 2.6232993401319737e-06, "loss": 0.0006, "step": 5283 }, { "epoch": 0.63, "learning_rate": 2.623227354529094e-06, "loss": 0.001, "step": 5284 }, { "epoch": 0.63, "learning_rate": 2.623155368926215e-06, "loss": 0.0008, "step": 5285 }, { "epoch": 0.63, "learning_rate": 2.6230833833233352e-06, "loss": 0.0005, "step": 5286 }, { "epoch": 0.63, "learning_rate": 2.623011397720456e-06, "loss": 0.0009, "step": 5287 }, { "epoch": 0.63, "learning_rate": 2.6229394121175764e-06, "loss": 0.0009, "step": 5288 }, { "epoch": 0.63, "learning_rate": 2.622867426514697e-06, "loss": 0.0005, "step": 5289 }, { "epoch": 0.63, "learning_rate": 2.6227954409118175e-06, "loss": 0.0011, "step": 5290 }, { "epoch": 0.63, "learning_rate": 2.6227234553089383e-06, "loss": 0.7988, "step": 5291 }, { "epoch": 0.63, "learning_rate": 2.6226514697060586e-06, "loss": 0.001, "step": 5292 }, { "epoch": 0.63, "learning_rate": 2.6225794841031794e-06, "loss": 0.0009, "step": 5293 }, { "epoch": 0.63, "learning_rate": 2.6225074985003e-06, "loss": 0.0006, "step": 5294 }, { "epoch": 0.63, "learning_rate": 2.6224355128974205e-06, "loss": 0.0004, "step": 5295 }, { "epoch": 0.63, "learning_rate": 2.6223635272945413e-06, "loss": 0.0011, "step": 5296 }, { "epoch": 0.63, "learning_rate": 2.6222915416916617e-06, "loss": 0.0007, "step": 5297 }, { "epoch": 0.63, "learning_rate": 2.6222195560887825e-06, "loss": 0.0006, "step": 5298 }, { "epoch": 0.63, "learning_rate": 2.622147570485903e-06, "loss": 0.5793, "step": 5299 }, { "epoch": 0.64, "learning_rate": 2.6220755848830236e-06, "loss": 0.0009, "step": 5300 }, { "epoch": 0.64, "learning_rate": 2.622003599280144e-06, "loss": 0.0007, "step": 5301 }, { "epoch": 0.64, "learning_rate": 2.6219316136772647e-06, "loss": 0.0004, "step": 5302 }, { "epoch": 0.64, "learning_rate": 2.621859628074385e-06, "loss": 0.0008, "step": 5303 }, { "epoch": 0.64, "learning_rate": 2.621787642471506e-06, "loss": 0.0006, "step": 5304 }, { "epoch": 0.64, "learning_rate": 2.6217156568686262e-06, "loss": 0.0056, "step": 5305 }, { "epoch": 0.64, "learning_rate": 2.6216436712657466e-06, "loss": 0.0007, "step": 5306 }, { "epoch": 0.64, "learning_rate": 2.6215716856628674e-06, "loss": 0.0037, "step": 5307 }, { "epoch": 0.64, "learning_rate": 2.621499700059988e-06, "loss": 0.0006, "step": 5308 }, { "epoch": 0.64, "learning_rate": 2.621427714457109e-06, "loss": 0.0004, "step": 5309 }, { "epoch": 0.64, "learning_rate": 2.6213557288542293e-06, "loss": 0.0007, "step": 5310 }, { "epoch": 0.64, "learning_rate": 2.62128374325135e-06, "loss": 0.0008, "step": 5311 }, { "epoch": 0.64, "learning_rate": 2.6212117576484704e-06, "loss": 0.0004, "step": 5312 }, { "epoch": 0.64, "learning_rate": 2.621139772045591e-06, "loss": 0.0009, "step": 5313 }, { "epoch": 0.64, "learning_rate": 2.6210677864427115e-06, "loss": 0.0008, "step": 5314 }, { "epoch": 0.64, "learning_rate": 2.6209958008398323e-06, "loss": 0.0009, "step": 5315 }, { "epoch": 0.64, "learning_rate": 2.6209238152369527e-06, "loss": 0.0006, "step": 5316 }, { "epoch": 0.64, "learning_rate": 2.620851829634073e-06, "loss": 0.0005, "step": 5317 }, { "epoch": 0.64, "learning_rate": 2.620779844031194e-06, "loss": 0.0005, "step": 5318 }, { "epoch": 0.64, "learning_rate": 2.620707858428314e-06, "loss": 0.003, "step": 5319 }, { "epoch": 0.64, "learning_rate": 2.620635872825435e-06, "loss": 0.7923, "step": 5320 }, { "epoch": 0.64, "learning_rate": 2.6205638872225553e-06, "loss": 0.0007, "step": 5321 }, { "epoch": 0.64, "learning_rate": 2.620491901619676e-06, "loss": 0.0048, "step": 5322 }, { "epoch": 0.64, "learning_rate": 2.620419916016797e-06, "loss": 0.0005, "step": 5323 }, { "epoch": 0.64, "learning_rate": 2.620347930413917e-06, "loss": 0.6551, "step": 5324 }, { "epoch": 0.64, "learning_rate": 2.620275944811038e-06, "loss": 0.6381, "step": 5325 }, { "epoch": 0.64, "learning_rate": 2.6202039592081588e-06, "loss": 0.0009, "step": 5326 }, { "epoch": 0.64, "learning_rate": 2.620131973605279e-06, "loss": 0.0009, "step": 5327 }, { "epoch": 0.64, "learning_rate": 2.6200599880023995e-06, "loss": 0.2858, "step": 5328 }, { "epoch": 0.64, "learning_rate": 2.6199880023995202e-06, "loss": 0.0019, "step": 5329 }, { "epoch": 0.64, "learning_rate": 2.6199160167966406e-06, "loss": 0.0007, "step": 5330 }, { "epoch": 0.64, "learning_rate": 2.6198440311937614e-06, "loss": 0.0023, "step": 5331 }, { "epoch": 0.64, "learning_rate": 2.6197720455908817e-06, "loss": 0.0006, "step": 5332 }, { "epoch": 0.64, "learning_rate": 2.6197000599880025e-06, "loss": 0.0008, "step": 5333 }, { "epoch": 0.64, "learning_rate": 2.619628074385123e-06, "loss": 0.0007, "step": 5334 }, { "epoch": 0.64, "learning_rate": 2.6195560887822436e-06, "loss": 0.0005, "step": 5335 }, { "epoch": 0.64, "learning_rate": 2.619484103179364e-06, "loss": 0.0006, "step": 5336 }, { "epoch": 0.64, "learning_rate": 2.6194121175764848e-06, "loss": 0.2263, "step": 5337 }, { "epoch": 0.64, "learning_rate": 2.619340131973605e-06, "loss": 0.0005, "step": 5338 }, { "epoch": 0.64, "learning_rate": 2.619268146370726e-06, "loss": 0.0005, "step": 5339 }, { "epoch": 0.64, "learning_rate": 2.6191961607678467e-06, "loss": 0.0004, "step": 5340 }, { "epoch": 0.64, "learning_rate": 2.619124175164967e-06, "loss": 0.0006, "step": 5341 }, { "epoch": 0.64, "learning_rate": 2.619052189562088e-06, "loss": 0.0008, "step": 5342 }, { "epoch": 0.64, "learning_rate": 2.618980203959208e-06, "loss": 0.0005, "step": 5343 }, { "epoch": 0.64, "learning_rate": 2.618908218356329e-06, "loss": 0.0008, "step": 5344 }, { "epoch": 0.64, "learning_rate": 2.6188362327534493e-06, "loss": 0.0007, "step": 5345 }, { "epoch": 0.64, "learning_rate": 2.61876424715057e-06, "loss": 0.4879, "step": 5346 }, { "epoch": 0.64, "learning_rate": 2.6186922615476904e-06, "loss": 0.0009, "step": 5347 }, { "epoch": 0.64, "learning_rate": 2.6186202759448112e-06, "loss": 0.0006, "step": 5348 }, { "epoch": 0.64, "learning_rate": 2.6185482903419316e-06, "loss": 0.0007, "step": 5349 }, { "epoch": 0.64, "learning_rate": 2.6184763047390524e-06, "loss": 0.013, "step": 5350 }, { "epoch": 0.64, "learning_rate": 2.6184043191361727e-06, "loss": 0.0016, "step": 5351 }, { "epoch": 0.64, "learning_rate": 2.618332333533293e-06, "loss": 0.0008, "step": 5352 }, { "epoch": 0.64, "learning_rate": 2.618260347930414e-06, "loss": 0.0093, "step": 5353 }, { "epoch": 0.64, "learning_rate": 2.6181883623275346e-06, "loss": 0.0018, "step": 5354 }, { "epoch": 0.64, "learning_rate": 2.6181163767246554e-06, "loss": 0.0007, "step": 5355 }, { "epoch": 0.64, "learning_rate": 2.6180443911217758e-06, "loss": 0.0005, "step": 5356 }, { "epoch": 0.64, "learning_rate": 2.6179724055188965e-06, "loss": 0.007, "step": 5357 }, { "epoch": 0.64, "learning_rate": 2.617900419916017e-06, "loss": 0.0006, "step": 5358 }, { "epoch": 0.64, "learning_rate": 2.6178284343131377e-06, "loss": 0.0032, "step": 5359 }, { "epoch": 0.64, "learning_rate": 2.617756448710258e-06, "loss": 0.0005, "step": 5360 }, { "epoch": 0.64, "learning_rate": 2.617684463107379e-06, "loss": 0.001, "step": 5361 }, { "epoch": 0.64, "learning_rate": 2.617612477504499e-06, "loss": 0.0005, "step": 5362 }, { "epoch": 0.64, "learning_rate": 2.6175404919016195e-06, "loss": 0.0006, "step": 5363 }, { "epoch": 0.64, "learning_rate": 2.6174685062987403e-06, "loss": 0.0007, "step": 5364 }, { "epoch": 0.64, "learning_rate": 2.6173965206958606e-06, "loss": 0.0006, "step": 5365 }, { "epoch": 0.64, "learning_rate": 2.6173245350929814e-06, "loss": 0.0008, "step": 5366 }, { "epoch": 0.64, "learning_rate": 2.6172525494901018e-06, "loss": 0.0005, "step": 5367 }, { "epoch": 0.64, "learning_rate": 2.6171805638872226e-06, "loss": 0.0707, "step": 5368 }, { "epoch": 0.64, "learning_rate": 2.6171085782843433e-06, "loss": 0.0008, "step": 5369 }, { "epoch": 0.64, "learning_rate": 2.6170365926814637e-06, "loss": 0.0006, "step": 5370 }, { "epoch": 0.64, "learning_rate": 2.6169646070785845e-06, "loss": 0.0011, "step": 5371 }, { "epoch": 0.64, "learning_rate": 2.6168926214757053e-06, "loss": 0.0027, "step": 5372 }, { "epoch": 0.64, "learning_rate": 2.6168206358728256e-06, "loss": 0.0006, "step": 5373 }, { "epoch": 0.64, "learning_rate": 2.616748650269946e-06, "loss": 0.0006, "step": 5374 }, { "epoch": 0.64, "learning_rate": 2.6166766646670667e-06, "loss": 0.3876, "step": 5375 }, { "epoch": 0.64, "learning_rate": 2.616604679064187e-06, "loss": 0.0625, "step": 5376 }, { "epoch": 0.64, "learning_rate": 2.616532693461308e-06, "loss": 0.0021, "step": 5377 }, { "epoch": 0.64, "learning_rate": 2.6164607078584282e-06, "loss": 0.007, "step": 5378 }, { "epoch": 0.64, "learning_rate": 2.616388722255549e-06, "loss": 0.0005, "step": 5379 }, { "epoch": 0.64, "learning_rate": 2.6163167366526694e-06, "loss": 0.3707, "step": 5380 }, { "epoch": 0.64, "learning_rate": 2.61624475104979e-06, "loss": 0.0004, "step": 5381 }, { "epoch": 0.64, "learning_rate": 2.6161727654469105e-06, "loss": 0.0006, "step": 5382 }, { "epoch": 0.65, "learning_rate": 2.6161007798440313e-06, "loss": 0.0006, "step": 5383 }, { "epoch": 0.65, "learning_rate": 2.6160287942411516e-06, "loss": 0.0006, "step": 5384 }, { "epoch": 0.65, "learning_rate": 2.6159568086382724e-06, "loss": 0.0008, "step": 5385 }, { "epoch": 0.65, "learning_rate": 2.615884823035393e-06, "loss": 0.0006, "step": 5386 }, { "epoch": 0.65, "learning_rate": 2.6158128374325135e-06, "loss": 0.0004, "step": 5387 }, { "epoch": 0.65, "learning_rate": 2.6157408518296343e-06, "loss": 0.0067, "step": 5388 }, { "epoch": 0.65, "learning_rate": 2.6156688662267547e-06, "loss": 0.0013, "step": 5389 }, { "epoch": 0.65, "learning_rate": 2.6155968806238755e-06, "loss": 0.0005, "step": 5390 }, { "epoch": 0.65, "learning_rate": 2.615524895020996e-06, "loss": 0.0006, "step": 5391 }, { "epoch": 0.65, "learning_rate": 2.6154529094181166e-06, "loss": 0.0007, "step": 5392 }, { "epoch": 0.65, "learning_rate": 2.615380923815237e-06, "loss": 0.8431, "step": 5393 }, { "epoch": 0.65, "learning_rate": 2.6153089382123577e-06, "loss": 0.0008, "step": 5394 }, { "epoch": 0.65, "learning_rate": 2.615236952609478e-06, "loss": 0.8989, "step": 5395 }, { "epoch": 0.65, "learning_rate": 2.615164967006599e-06, "loss": 0.0005, "step": 5396 }, { "epoch": 0.65, "learning_rate": 2.615092981403719e-06, "loss": 0.0006, "step": 5397 }, { "epoch": 0.65, "learning_rate": 2.6150209958008396e-06, "loss": 0.0005, "step": 5398 }, { "epoch": 0.65, "learning_rate": 2.6149490101979603e-06, "loss": 0.0007, "step": 5399 }, { "epoch": 0.65, "learning_rate": 2.614877024595081e-06, "loss": 0.0006, "step": 5400 }, { "epoch": 0.65, "learning_rate": 2.614805038992202e-06, "loss": 0.0007, "step": 5401 }, { "epoch": 0.65, "learning_rate": 2.6147330533893223e-06, "loss": 0.0006, "step": 5402 }, { "epoch": 0.65, "learning_rate": 2.614661067786443e-06, "loss": 0.0008, "step": 5403 }, { "epoch": 0.65, "learning_rate": 2.6145890821835634e-06, "loss": 0.0006, "step": 5404 }, { "epoch": 0.65, "learning_rate": 2.614517096580684e-06, "loss": 0.0006, "step": 5405 }, { "epoch": 0.65, "learning_rate": 2.6144451109778045e-06, "loss": 0.0005, "step": 5406 }, { "epoch": 0.65, "learning_rate": 2.6143731253749253e-06, "loss": 0.0008, "step": 5407 }, { "epoch": 0.65, "learning_rate": 2.6143011397720457e-06, "loss": 0.1619, "step": 5408 }, { "epoch": 0.65, "learning_rate": 2.614229154169166e-06, "loss": 0.0009, "step": 5409 }, { "epoch": 0.65, "learning_rate": 2.614157168566287e-06, "loss": 0.0009, "step": 5410 }, { "epoch": 0.65, "learning_rate": 2.614085182963407e-06, "loss": 0.3018, "step": 5411 }, { "epoch": 0.65, "learning_rate": 2.614013197360528e-06, "loss": 0.0008, "step": 5412 }, { "epoch": 0.65, "learning_rate": 2.6139412117576483e-06, "loss": 0.0008, "step": 5413 }, { "epoch": 0.65, "learning_rate": 2.613869226154769e-06, "loss": 0.0006, "step": 5414 }, { "epoch": 0.65, "learning_rate": 2.6137972405518894e-06, "loss": 0.0005, "step": 5415 }, { "epoch": 0.65, "learning_rate": 2.61372525494901e-06, "loss": 0.0018, "step": 5416 }, { "epoch": 0.65, "learning_rate": 2.613653269346131e-06, "loss": 0.0011, "step": 5417 }, { "epoch": 0.65, "learning_rate": 2.6135812837432517e-06, "loss": 0.0008, "step": 5418 }, { "epoch": 0.65, "learning_rate": 2.613509298140372e-06, "loss": 0.0007, "step": 5419 }, { "epoch": 0.65, "learning_rate": 2.6134373125374925e-06, "loss": 0.0004, "step": 5420 }, { "epoch": 0.65, "learning_rate": 2.6133653269346132e-06, "loss": 0.0013, "step": 5421 }, { "epoch": 0.65, "learning_rate": 2.6132933413317336e-06, "loss": 0.2814, "step": 5422 }, { "epoch": 0.65, "learning_rate": 2.6132213557288544e-06, "loss": 0.0007, "step": 5423 }, { "epoch": 0.65, "learning_rate": 2.6131493701259747e-06, "loss": 0.2312, "step": 5424 }, { "epoch": 0.65, "learning_rate": 2.6130773845230955e-06, "loss": 0.0008, "step": 5425 }, { "epoch": 0.65, "learning_rate": 2.613005398920216e-06, "loss": 0.0003, "step": 5426 }, { "epoch": 0.65, "learning_rate": 2.6129334133173366e-06, "loss": 0.0004, "step": 5427 }, { "epoch": 0.65, "learning_rate": 2.612861427714457e-06, "loss": 0.0011, "step": 5428 }, { "epoch": 0.65, "learning_rate": 2.6127894421115778e-06, "loss": 0.0006, "step": 5429 }, { "epoch": 0.65, "learning_rate": 2.612717456508698e-06, "loss": 0.001, "step": 5430 }, { "epoch": 0.65, "learning_rate": 2.612645470905819e-06, "loss": 0.0007, "step": 5431 }, { "epoch": 0.65, "learning_rate": 2.6125734853029397e-06, "loss": 0.0006, "step": 5432 }, { "epoch": 0.65, "learning_rate": 2.61250149970006e-06, "loss": 0.0015, "step": 5433 }, { "epoch": 0.65, "learning_rate": 2.612429514097181e-06, "loss": 0.0007, "step": 5434 }, { "epoch": 0.65, "learning_rate": 2.612357528494301e-06, "loss": 0.001, "step": 5435 }, { "epoch": 0.65, "learning_rate": 2.612285542891422e-06, "loss": 0.0008, "step": 5436 }, { "epoch": 0.65, "learning_rate": 2.6122135572885423e-06, "loss": 0.0007, "step": 5437 }, { "epoch": 0.65, "learning_rate": 2.612141571685663e-06, "loss": 0.7963, "step": 5438 }, { "epoch": 0.65, "learning_rate": 2.6120695860827834e-06, "loss": 0.0026, "step": 5439 }, { "epoch": 0.65, "learning_rate": 2.6119976004799042e-06, "loss": 0.0081, "step": 5440 }, { "epoch": 0.65, "learning_rate": 2.6119256148770246e-06, "loss": 0.0009, "step": 5441 }, { "epoch": 0.65, "learning_rate": 2.6118536292741454e-06, "loss": 0.0009, "step": 5442 }, { "epoch": 0.65, "learning_rate": 2.6117816436712657e-06, "loss": 0.0004, "step": 5443 }, { "epoch": 0.65, "learning_rate": 2.611709658068386e-06, "loss": 0.0006, "step": 5444 }, { "epoch": 0.65, "learning_rate": 2.611637672465507e-06, "loss": 0.7782, "step": 5445 }, { "epoch": 0.65, "learning_rate": 2.6115656868626276e-06, "loss": 0.0005, "step": 5446 }, { "epoch": 0.65, "learning_rate": 2.611493701259748e-06, "loss": 0.0004, "step": 5447 }, { "epoch": 0.65, "learning_rate": 2.6114217156568688e-06, "loss": 0.0007, "step": 5448 }, { "epoch": 0.65, "learning_rate": 2.6113497300539895e-06, "loss": 0.001, "step": 5449 }, { "epoch": 0.65, "learning_rate": 2.61127774445111e-06, "loss": 0.0006, "step": 5450 }, { "epoch": 0.65, "learning_rate": 2.6112057588482307e-06, "loss": 0.0006, "step": 5451 }, { "epoch": 0.65, "learning_rate": 2.611133773245351e-06, "loss": 1.4401, "step": 5452 }, { "epoch": 0.65, "learning_rate": 2.611061787642472e-06, "loss": 0.1625, "step": 5453 }, { "epoch": 0.65, "learning_rate": 2.610989802039592e-06, "loss": 0.0005, "step": 5454 }, { "epoch": 0.65, "learning_rate": 2.6109178164367125e-06, "loss": 0.0005, "step": 5455 }, { "epoch": 0.65, "learning_rate": 2.6108458308338333e-06, "loss": 0.4644, "step": 5456 }, { "epoch": 0.65, "learning_rate": 2.6107738452309536e-06, "loss": 0.0004, "step": 5457 }, { "epoch": 0.65, "learning_rate": 2.6107018596280744e-06, "loss": 0.0011, "step": 5458 }, { "epoch": 0.65, "learning_rate": 2.6106298740251948e-06, "loss": 0.0005, "step": 5459 }, { "epoch": 0.65, "learning_rate": 2.6105578884223156e-06, "loss": 0.0007, "step": 5460 }, { "epoch": 0.65, "learning_rate": 2.610485902819436e-06, "loss": 0.0008, "step": 5461 }, { "epoch": 0.65, "learning_rate": 2.6104139172165567e-06, "loss": 0.0008, "step": 5462 }, { "epoch": 0.65, "learning_rate": 2.6103419316136775e-06, "loss": 0.0016, "step": 5463 }, { "epoch": 0.65, "learning_rate": 2.6102699460107982e-06, "loss": 0.001, "step": 5464 }, { "epoch": 0.65, "learning_rate": 2.6101979604079186e-06, "loss": 0.0009, "step": 5465 }, { "epoch": 0.66, "learning_rate": 2.610125974805039e-06, "loss": 0.0009, "step": 5466 }, { "epoch": 0.66, "learning_rate": 2.6100539892021597e-06, "loss": 0.0009, "step": 5467 }, { "epoch": 0.66, "learning_rate": 2.60998200359928e-06, "loss": 0.0005, "step": 5468 }, { "epoch": 0.66, "learning_rate": 2.609910017996401e-06, "loss": 0.0025, "step": 5469 }, { "epoch": 0.66, "learning_rate": 2.6098380323935212e-06, "loss": 0.0005, "step": 5470 }, { "epoch": 0.66, "learning_rate": 2.609766046790642e-06, "loss": 0.0006, "step": 5471 }, { "epoch": 0.66, "learning_rate": 2.6096940611877624e-06, "loss": 0.001, "step": 5472 }, { "epoch": 0.66, "learning_rate": 2.609622075584883e-06, "loss": 0.7756, "step": 5473 }, { "epoch": 0.66, "learning_rate": 2.6095500899820035e-06, "loss": 0.0005, "step": 5474 }, { "epoch": 0.66, "learning_rate": 2.6094781043791243e-06, "loss": 0.0006, "step": 5475 }, { "epoch": 0.66, "learning_rate": 2.6094061187762446e-06, "loss": 0.0004, "step": 5476 }, { "epoch": 0.66, "learning_rate": 2.6093341331733654e-06, "loss": 0.0053, "step": 5477 }, { "epoch": 0.66, "learning_rate": 2.609262147570486e-06, "loss": 0.0035, "step": 5478 }, { "epoch": 0.66, "learning_rate": 2.6091901619676065e-06, "loss": 0.0006, "step": 5479 }, { "epoch": 0.66, "learning_rate": 2.6091181763647273e-06, "loss": 0.0006, "step": 5480 }, { "epoch": 0.66, "learning_rate": 2.6090461907618477e-06, "loss": 0.0007, "step": 5481 }, { "epoch": 0.66, "learning_rate": 2.6089742051589684e-06, "loss": 0.0011, "step": 5482 }, { "epoch": 0.66, "learning_rate": 2.608902219556089e-06, "loss": 0.0004, "step": 5483 }, { "epoch": 0.66, "learning_rate": 2.6088302339532096e-06, "loss": 0.0005, "step": 5484 }, { "epoch": 0.66, "learning_rate": 2.60875824835033e-06, "loss": 0.7628, "step": 5485 }, { "epoch": 0.66, "learning_rate": 2.6086862627474507e-06, "loss": 0.0004, "step": 5486 }, { "epoch": 0.66, "learning_rate": 2.608614277144571e-06, "loss": 0.0005, "step": 5487 }, { "epoch": 0.66, "learning_rate": 2.608542291541692e-06, "loss": 0.0009, "step": 5488 }, { "epoch": 0.66, "learning_rate": 2.608470305938812e-06, "loss": 0.0009, "step": 5489 }, { "epoch": 0.66, "learning_rate": 2.6083983203359326e-06, "loss": 0.003, "step": 5490 }, { "epoch": 0.66, "learning_rate": 2.6083263347330533e-06, "loss": 0.0007, "step": 5491 }, { "epoch": 0.66, "learning_rate": 2.608254349130174e-06, "loss": 0.7462, "step": 5492 }, { "epoch": 0.66, "learning_rate": 2.6081823635272945e-06, "loss": 0.0019, "step": 5493 }, { "epoch": 0.66, "learning_rate": 2.6081103779244152e-06, "loss": 0.0007, "step": 5494 }, { "epoch": 0.66, "learning_rate": 2.608038392321536e-06, "loss": 0.0006, "step": 5495 }, { "epoch": 0.66, "learning_rate": 2.6079664067186564e-06, "loss": 0.0004, "step": 5496 }, { "epoch": 0.66, "learning_rate": 2.607894421115777e-06, "loss": 0.001, "step": 5497 }, { "epoch": 0.66, "learning_rate": 2.6078224355128975e-06, "loss": 0.0004, "step": 5498 }, { "epoch": 0.66, "learning_rate": 2.6077504499100183e-06, "loss": 0.0009, "step": 5499 }, { "epoch": 0.66, "learning_rate": 2.6076784643071386e-06, "loss": 0.5133, "step": 5500 }, { "epoch": 0.66, "learning_rate": 2.607606478704259e-06, "loss": 0.0009, "step": 5501 }, { "epoch": 0.66, "learning_rate": 2.6075344931013798e-06, "loss": 0.0008, "step": 5502 }, { "epoch": 0.66, "learning_rate": 2.6074625074985e-06, "loss": 0.0008, "step": 5503 }, { "epoch": 0.66, "learning_rate": 2.607390521895621e-06, "loss": 0.0007, "step": 5504 }, { "epoch": 0.66, "learning_rate": 2.6073185362927413e-06, "loss": 0.0005, "step": 5505 }, { "epoch": 0.66, "learning_rate": 2.607246550689862e-06, "loss": 0.0008, "step": 5506 }, { "epoch": 0.66, "learning_rate": 2.6071745650869824e-06, "loss": 0.001, "step": 5507 }, { "epoch": 0.66, "learning_rate": 2.607102579484103e-06, "loss": 0.0006, "step": 5508 }, { "epoch": 0.66, "learning_rate": 2.607030593881224e-06, "loss": 0.0006, "step": 5509 }, { "epoch": 0.66, "learning_rate": 2.6069586082783447e-06, "loss": 0.0005, "step": 5510 }, { "epoch": 0.66, "learning_rate": 2.606886622675465e-06, "loss": 0.0012, "step": 5511 }, { "epoch": 0.66, "learning_rate": 2.6068146370725854e-06, "loss": 0.9141, "step": 5512 }, { "epoch": 0.66, "learning_rate": 2.6067426514697062e-06, "loss": 0.0005, "step": 5513 }, { "epoch": 0.66, "learning_rate": 2.6066706658668266e-06, "loss": 0.0005, "step": 5514 }, { "epoch": 0.66, "learning_rate": 2.6065986802639474e-06, "loss": 0.0454, "step": 5515 }, { "epoch": 0.66, "learning_rate": 2.6065266946610677e-06, "loss": 0.0007, "step": 5516 }, { "epoch": 0.66, "learning_rate": 2.6064547090581885e-06, "loss": 0.0004, "step": 5517 }, { "epoch": 0.66, "learning_rate": 2.606382723455309e-06, "loss": 0.0207, "step": 5518 }, { "epoch": 0.66, "learning_rate": 2.6063107378524296e-06, "loss": 0.0005, "step": 5519 }, { "epoch": 0.66, "learning_rate": 2.60623875224955e-06, "loss": 0.0005, "step": 5520 }, { "epoch": 0.66, "learning_rate": 2.6061667666466708e-06, "loss": 0.0007, "step": 5521 }, { "epoch": 0.66, "learning_rate": 2.606094781043791e-06, "loss": 0.3155, "step": 5522 }, { "epoch": 0.66, "learning_rate": 2.606022795440912e-06, "loss": 0.0008, "step": 5523 }, { "epoch": 0.66, "learning_rate": 2.6059508098380327e-06, "loss": 0.0008, "step": 5524 }, { "epoch": 0.66, "learning_rate": 2.605878824235153e-06, "loss": 0.0006, "step": 5525 }, { "epoch": 0.66, "learning_rate": 2.605806838632274e-06, "loss": 0.001, "step": 5526 }, { "epoch": 0.66, "learning_rate": 2.605734853029394e-06, "loss": 0.0012, "step": 5527 }, { "epoch": 0.66, "learning_rate": 2.605662867426515e-06, "loss": 0.0004, "step": 5528 }, { "epoch": 0.66, "learning_rate": 2.6055908818236353e-06, "loss": 0.0007, "step": 5529 }, { "epoch": 0.66, "learning_rate": 2.605518896220756e-06, "loss": 0.0008, "step": 5530 }, { "epoch": 0.66, "learning_rate": 2.6054469106178764e-06, "loss": 0.0014, "step": 5531 }, { "epoch": 0.66, "learning_rate": 2.605374925014997e-06, "loss": 0.0008, "step": 5532 }, { "epoch": 0.66, "learning_rate": 2.6053029394121176e-06, "loss": 0.7558, "step": 5533 }, { "epoch": 0.66, "learning_rate": 2.6052309538092383e-06, "loss": 0.0006, "step": 5534 }, { "epoch": 0.66, "learning_rate": 2.6051589682063587e-06, "loss": 0.0009, "step": 5535 }, { "epoch": 0.66, "learning_rate": 2.605086982603479e-06, "loss": 0.0007, "step": 5536 }, { "epoch": 0.66, "learning_rate": 2.6050149970006e-06, "loss": 0.0028, "step": 5537 }, { "epoch": 0.66, "learning_rate": 2.60494301139772e-06, "loss": 0.0009, "step": 5538 }, { "epoch": 0.66, "learning_rate": 2.604871025794841e-06, "loss": 0.0003, "step": 5539 }, { "epoch": 0.66, "learning_rate": 2.6047990401919617e-06, "loss": 0.0612, "step": 5540 }, { "epoch": 0.66, "learning_rate": 2.6047270545890825e-06, "loss": 0.5177, "step": 5541 }, { "epoch": 0.66, "learning_rate": 2.604655068986203e-06, "loss": 0.0003, "step": 5542 }, { "epoch": 0.66, "learning_rate": 2.6045830833833237e-06, "loss": 0.0065, "step": 5543 }, { "epoch": 0.66, "learning_rate": 2.604511097780444e-06, "loss": 0.0005, "step": 5544 }, { "epoch": 0.66, "learning_rate": 2.604439112177565e-06, "loss": 0.0006, "step": 5545 }, { "epoch": 0.66, "learning_rate": 2.604367126574685e-06, "loss": 0.4828, "step": 5546 }, { "epoch": 0.66, "learning_rate": 2.6042951409718055e-06, "loss": 0.0005, "step": 5547 }, { "epoch": 0.66, "learning_rate": 2.6042231553689263e-06, "loss": 0.7219, "step": 5548 }, { "epoch": 0.66, "learning_rate": 2.6041511697660466e-06, "loss": 0.2624, "step": 5549 }, { "epoch": 0.67, "learning_rate": 2.6040791841631674e-06, "loss": 0.0004, "step": 5550 }, { "epoch": 0.67, "learning_rate": 2.6040071985602878e-06, "loss": 0.0005, "step": 5551 }, { "epoch": 0.67, "learning_rate": 2.6039352129574085e-06, "loss": 0.0004, "step": 5552 }, { "epoch": 0.67, "learning_rate": 2.603863227354529e-06, "loss": 0.1231, "step": 5553 }, { "epoch": 0.67, "learning_rate": 2.6037912417516497e-06, "loss": 0.9208, "step": 5554 }, { "epoch": 0.67, "learning_rate": 2.6037192561487705e-06, "loss": 0.138, "step": 5555 }, { "epoch": 0.67, "learning_rate": 2.603647270545891e-06, "loss": 0.854, "step": 5556 }, { "epoch": 0.67, "learning_rate": 2.6035752849430116e-06, "loss": 0.0057, "step": 5557 }, { "epoch": 0.67, "learning_rate": 2.6035032993401324e-06, "loss": 0.0015, "step": 5558 }, { "epoch": 0.67, "learning_rate": 2.6034313137372527e-06, "loss": 0.001, "step": 5559 }, { "epoch": 0.67, "learning_rate": 2.603359328134373e-06, "loss": 0.0012, "step": 5560 }, { "epoch": 0.67, "learning_rate": 2.603287342531494e-06, "loss": 0.001, "step": 5561 }, { "epoch": 0.67, "learning_rate": 2.6032153569286142e-06, "loss": 0.0021, "step": 5562 }, { "epoch": 0.67, "learning_rate": 2.603143371325735e-06, "loss": 0.0005, "step": 5563 }, { "epoch": 0.67, "learning_rate": 2.6030713857228553e-06, "loss": 0.001, "step": 5564 }, { "epoch": 0.67, "learning_rate": 2.602999400119976e-06, "loss": 0.0005, "step": 5565 }, { "epoch": 0.67, "learning_rate": 2.6029274145170965e-06, "loss": 0.0106, "step": 5566 }, { "epoch": 0.67, "learning_rate": 2.6028554289142173e-06, "loss": 0.0008, "step": 5567 }, { "epoch": 0.67, "learning_rate": 2.6027834433113376e-06, "loss": 0.0033, "step": 5568 }, { "epoch": 0.67, "learning_rate": 2.6027114577084584e-06, "loss": 0.0005, "step": 5569 }, { "epoch": 0.67, "learning_rate": 2.6026394721055787e-06, "loss": 0.0242, "step": 5570 }, { "epoch": 0.67, "learning_rate": 2.6025674865026995e-06, "loss": 0.0007, "step": 5571 }, { "epoch": 0.67, "learning_rate": 2.6024955008998203e-06, "loss": 0.0043, "step": 5572 }, { "epoch": 0.67, "learning_rate": 2.6024235152969407e-06, "loss": 0.0007, "step": 5573 }, { "epoch": 0.67, "learning_rate": 2.6023515296940614e-06, "loss": 0.0009, "step": 5574 }, { "epoch": 0.67, "learning_rate": 2.602279544091182e-06, "loss": 0.0006, "step": 5575 }, { "epoch": 0.67, "learning_rate": 2.6022075584883026e-06, "loss": 0.0006, "step": 5576 }, { "epoch": 0.67, "learning_rate": 2.602135572885423e-06, "loss": 0.0013, "step": 5577 }, { "epoch": 0.67, "learning_rate": 2.6020635872825437e-06, "loss": 0.0007, "step": 5578 }, { "epoch": 0.67, "learning_rate": 2.601991601679664e-06, "loss": 0.0004, "step": 5579 }, { "epoch": 0.67, "learning_rate": 2.601919616076785e-06, "loss": 0.0009, "step": 5580 }, { "epoch": 0.67, "learning_rate": 2.601847630473905e-06, "loss": 0.0007, "step": 5581 }, { "epoch": 0.67, "learning_rate": 2.6017756448710255e-06, "loss": 0.0006, "step": 5582 }, { "epoch": 0.67, "learning_rate": 2.6017036592681463e-06, "loss": 0.0013, "step": 5583 }, { "epoch": 0.67, "learning_rate": 2.6016316736652667e-06, "loss": 0.0008, "step": 5584 }, { "epoch": 0.67, "learning_rate": 2.6015596880623875e-06, "loss": 0.0004, "step": 5585 }, { "epoch": 0.67, "learning_rate": 2.6014877024595082e-06, "loss": 0.0005, "step": 5586 }, { "epoch": 0.67, "learning_rate": 2.601415716856629e-06, "loss": 0.0009, "step": 5587 }, { "epoch": 0.67, "learning_rate": 2.6013437312537494e-06, "loss": 0.0007, "step": 5588 }, { "epoch": 0.67, "learning_rate": 2.60127174565087e-06, "loss": 0.0006, "step": 5589 }, { "epoch": 0.67, "learning_rate": 2.6011997600479905e-06, "loss": 0.0006, "step": 5590 }, { "epoch": 0.67, "learning_rate": 2.6011277744451113e-06, "loss": 0.0006, "step": 5591 }, { "epoch": 0.67, "learning_rate": 2.6010557888422316e-06, "loss": 0.0003, "step": 5592 }, { "epoch": 0.67, "learning_rate": 2.600983803239352e-06, "loss": 0.6895, "step": 5593 }, { "epoch": 0.67, "learning_rate": 2.6009118176364728e-06, "loss": 0.0004, "step": 5594 }, { "epoch": 0.67, "learning_rate": 2.600839832033593e-06, "loss": 0.0011, "step": 5595 }, { "epoch": 0.67, "learning_rate": 2.600767846430714e-06, "loss": 0.0006, "step": 5596 }, { "epoch": 0.67, "learning_rate": 2.6006958608278343e-06, "loss": 0.0008, "step": 5597 }, { "epoch": 0.67, "learning_rate": 2.600623875224955e-06, "loss": 0.0017, "step": 5598 }, { "epoch": 0.67, "learning_rate": 2.6005518896220754e-06, "loss": 0.0007, "step": 5599 }, { "epoch": 0.67, "learning_rate": 2.600479904019196e-06, "loss": 0.0005, "step": 5600 }, { "epoch": 0.67, "learning_rate": 2.600407918416317e-06, "loss": 0.0005, "step": 5601 }, { "epoch": 0.67, "learning_rate": 2.6003359328134373e-06, "loss": 0.0028, "step": 5602 }, { "epoch": 0.67, "learning_rate": 2.600263947210558e-06, "loss": 0.0005, "step": 5603 }, { "epoch": 0.67, "learning_rate": 2.600191961607679e-06, "loss": 0.0005, "step": 5604 }, { "epoch": 0.67, "learning_rate": 2.6001199760047992e-06, "loss": 0.8386, "step": 5605 }, { "epoch": 0.67, "learning_rate": 2.6000479904019196e-06, "loss": 0.8689, "step": 5606 }, { "epoch": 0.67, "learning_rate": 2.5999760047990404e-06, "loss": 0.051, "step": 5607 }, { "epoch": 0.67, "learning_rate": 2.5999040191961607e-06, "loss": 0.0009, "step": 5608 }, { "epoch": 0.67, "learning_rate": 2.5998320335932815e-06, "loss": 0.0007, "step": 5609 }, { "epoch": 0.67, "learning_rate": 2.599760047990402e-06, "loss": 0.0012, "step": 5610 }, { "epoch": 0.67, "learning_rate": 2.5996880623875226e-06, "loss": 0.7507, "step": 5611 }, { "epoch": 0.67, "learning_rate": 2.599616076784643e-06, "loss": 0.0006, "step": 5612 }, { "epoch": 0.67, "learning_rate": 2.5995440911817638e-06, "loss": 0.6892, "step": 5613 }, { "epoch": 0.67, "learning_rate": 2.599472105578884e-06, "loss": 0.0005, "step": 5614 }, { "epoch": 0.67, "learning_rate": 2.599400119976005e-06, "loss": 0.0041, "step": 5615 }, { "epoch": 0.67, "learning_rate": 2.5993281343731252e-06, "loss": 0.0049, "step": 5616 }, { "epoch": 0.67, "learning_rate": 2.599256148770246e-06, "loss": 0.0006, "step": 5617 }, { "epoch": 0.67, "learning_rate": 2.599184163167367e-06, "loss": 0.0009, "step": 5618 }, { "epoch": 0.67, "learning_rate": 2.599112177564487e-06, "loss": 0.0005, "step": 5619 }, { "epoch": 0.67, "learning_rate": 2.599040191961608e-06, "loss": 0.0005, "step": 5620 }, { "epoch": 0.67, "learning_rate": 2.5989682063587283e-06, "loss": 0.0004, "step": 5621 }, { "epoch": 0.67, "learning_rate": 2.598896220755849e-06, "loss": 0.0028, "step": 5622 }, { "epoch": 0.67, "learning_rate": 2.5988242351529694e-06, "loss": 0.112, "step": 5623 }, { "epoch": 0.67, "learning_rate": 2.59875224955009e-06, "loss": 0.0006, "step": 5624 }, { "epoch": 0.67, "learning_rate": 2.5986802639472106e-06, "loss": 0.0005, "step": 5625 }, { "epoch": 0.67, "learning_rate": 2.5986082783443313e-06, "loss": 0.0006, "step": 5626 }, { "epoch": 0.67, "learning_rate": 2.5985362927414517e-06, "loss": 0.0004, "step": 5627 }, { "epoch": 0.67, "learning_rate": 2.598464307138572e-06, "loss": 0.0008, "step": 5628 }, { "epoch": 0.67, "learning_rate": 2.598392321535693e-06, "loss": 0.0005, "step": 5629 }, { "epoch": 0.67, "learning_rate": 2.598320335932813e-06, "loss": 0.0004, "step": 5630 }, { "epoch": 0.67, "learning_rate": 2.598248350329934e-06, "loss": 0.0004, "step": 5631 }, { "epoch": 0.67, "learning_rate": 2.5981763647270547e-06, "loss": 0.0005, "step": 5632 }, { "epoch": 0.68, "learning_rate": 2.5981043791241755e-06, "loss": 0.0006, "step": 5633 }, { "epoch": 0.68, "learning_rate": 2.598032393521296e-06, "loss": 0.0347, "step": 5634 }, { "epoch": 0.68, "learning_rate": 2.5979604079184166e-06, "loss": 0.002, "step": 5635 }, { "epoch": 0.68, "learning_rate": 2.597888422315537e-06, "loss": 0.0133, "step": 5636 }, { "epoch": 0.68, "learning_rate": 2.5978164367126578e-06, "loss": 0.1644, "step": 5637 }, { "epoch": 0.68, "learning_rate": 2.597744451109778e-06, "loss": 0.003, "step": 5638 }, { "epoch": 0.68, "learning_rate": 2.5976724655068985e-06, "loss": 0.0793, "step": 5639 }, { "epoch": 0.68, "learning_rate": 2.5976004799040193e-06, "loss": 0.0007, "step": 5640 }, { "epoch": 0.68, "learning_rate": 2.5975284943011396e-06, "loss": 0.0005, "step": 5641 }, { "epoch": 0.68, "learning_rate": 2.5974565086982604e-06, "loss": 0.0005, "step": 5642 }, { "epoch": 0.68, "learning_rate": 2.5973845230953808e-06, "loss": 0.0005, "step": 5643 }, { "epoch": 0.68, "learning_rate": 2.5973125374925015e-06, "loss": 0.0004, "step": 5644 }, { "epoch": 0.68, "learning_rate": 2.597240551889622e-06, "loss": 0.0014, "step": 5645 }, { "epoch": 0.68, "learning_rate": 2.5971685662867427e-06, "loss": 0.3842, "step": 5646 }, { "epoch": 0.68, "learning_rate": 2.5970965806838634e-06, "loss": 0.0223, "step": 5647 }, { "epoch": 0.68, "learning_rate": 2.597024595080984e-06, "loss": 0.0012, "step": 5648 }, { "epoch": 0.68, "learning_rate": 2.5969526094781046e-06, "loss": 0.0005, "step": 5649 }, { "epoch": 0.68, "learning_rate": 2.5968806238752254e-06, "loss": 0.0008, "step": 5650 }, { "epoch": 0.68, "learning_rate": 2.5968086382723457e-06, "loss": 0.0003, "step": 5651 }, { "epoch": 0.68, "learning_rate": 2.596736652669466e-06, "loss": 0.0003, "step": 5652 }, { "epoch": 0.68, "learning_rate": 2.596664667066587e-06, "loss": 0.5591, "step": 5653 }, { "epoch": 0.68, "learning_rate": 2.596592681463707e-06, "loss": 0.0761, "step": 5654 }, { "epoch": 0.68, "learning_rate": 2.596520695860828e-06, "loss": 0.0004, "step": 5655 }, { "epoch": 0.68, "learning_rate": 2.5964487102579483e-06, "loss": 0.0005, "step": 5656 }, { "epoch": 0.68, "learning_rate": 2.596376724655069e-06, "loss": 0.0006, "step": 5657 }, { "epoch": 0.68, "learning_rate": 2.5963047390521895e-06, "loss": 0.0006, "step": 5658 }, { "epoch": 0.68, "learning_rate": 2.5962327534493103e-06, "loss": 0.3453, "step": 5659 }, { "epoch": 0.68, "learning_rate": 2.5961607678464306e-06, "loss": 0.0005, "step": 5660 }, { "epoch": 0.68, "learning_rate": 2.5960887822435514e-06, "loss": 0.0003, "step": 5661 }, { "epoch": 0.68, "learning_rate": 2.5960167966406717e-06, "loss": 0.0004, "step": 5662 }, { "epoch": 0.68, "learning_rate": 2.5959448110377925e-06, "loss": 0.0005, "step": 5663 }, { "epoch": 0.68, "learning_rate": 2.5958728254349133e-06, "loss": 0.0004, "step": 5664 }, { "epoch": 0.68, "learning_rate": 2.5958008398320337e-06, "loss": 0.0005, "step": 5665 }, { "epoch": 0.68, "learning_rate": 2.5957288542291544e-06, "loss": 0.0008, "step": 5666 }, { "epoch": 0.68, "learning_rate": 2.5956568686262748e-06, "loss": 0.0004, "step": 5667 }, { "epoch": 0.68, "learning_rate": 2.5955848830233956e-06, "loss": 0.0035, "step": 5668 }, { "epoch": 0.68, "learning_rate": 2.595512897420516e-06, "loss": 0.0008, "step": 5669 }, { "epoch": 0.68, "learning_rate": 2.5954409118176367e-06, "loss": 0.0007, "step": 5670 }, { "epoch": 0.68, "learning_rate": 2.595368926214757e-06, "loss": 0.0007, "step": 5671 }, { "epoch": 0.68, "learning_rate": 2.595296940611878e-06, "loss": 0.0005, "step": 5672 }, { "epoch": 0.68, "learning_rate": 2.595224955008998e-06, "loss": 0.0006, "step": 5673 }, { "epoch": 0.68, "learning_rate": 2.5951529694061185e-06, "loss": 0.0007, "step": 5674 }, { "epoch": 0.68, "learning_rate": 2.5950809838032393e-06, "loss": 0.0016, "step": 5675 }, { "epoch": 0.68, "learning_rate": 2.5950089982003597e-06, "loss": 0.0006, "step": 5676 }, { "epoch": 0.68, "learning_rate": 2.5949370125974805e-06, "loss": 0.0012, "step": 5677 }, { "epoch": 0.68, "learning_rate": 2.5948650269946012e-06, "loss": 0.7823, "step": 5678 }, { "epoch": 0.68, "learning_rate": 2.594793041391722e-06, "loss": 0.0003, "step": 5679 }, { "epoch": 0.68, "learning_rate": 2.5947210557888424e-06, "loss": 0.0006, "step": 5680 }, { "epoch": 0.68, "learning_rate": 2.594649070185963e-06, "loss": 0.0013, "step": 5681 }, { "epoch": 0.68, "learning_rate": 2.5945770845830835e-06, "loss": 0.0006, "step": 5682 }, { "epoch": 0.68, "learning_rate": 2.5945050989802043e-06, "loss": 0.0007, "step": 5683 }, { "epoch": 0.68, "learning_rate": 2.5944331133773246e-06, "loss": 0.0008, "step": 5684 }, { "epoch": 0.68, "learning_rate": 2.5943611277744454e-06, "loss": 0.0005, "step": 5685 }, { "epoch": 0.68, "learning_rate": 2.5942891421715658e-06, "loss": 0.0007, "step": 5686 }, { "epoch": 0.68, "learning_rate": 2.594217156568686e-06, "loss": 0.0005, "step": 5687 }, { "epoch": 0.68, "learning_rate": 2.594145170965807e-06, "loss": 0.0004, "step": 5688 }, { "epoch": 0.68, "learning_rate": 2.5940731853629273e-06, "loss": 0.0004, "step": 5689 }, { "epoch": 0.68, "learning_rate": 2.594001199760048e-06, "loss": 0.001, "step": 5690 }, { "epoch": 0.68, "learning_rate": 2.5939292141571684e-06, "loss": 0.003, "step": 5691 }, { "epoch": 0.68, "learning_rate": 2.593857228554289e-06, "loss": 0.0007, "step": 5692 }, { "epoch": 0.68, "learning_rate": 2.5937852429514095e-06, "loss": 0.0007, "step": 5693 }, { "epoch": 0.68, "learning_rate": 2.5937132573485303e-06, "loss": 0.0009, "step": 5694 }, { "epoch": 0.68, "learning_rate": 2.593641271745651e-06, "loss": 0.4591, "step": 5695 }, { "epoch": 0.68, "learning_rate": 2.593569286142772e-06, "loss": 0.0015, "step": 5696 }, { "epoch": 0.68, "learning_rate": 2.5934973005398922e-06, "loss": 0.0006, "step": 5697 }, { "epoch": 0.68, "learning_rate": 2.5934253149370126e-06, "loss": 0.0003, "step": 5698 }, { "epoch": 0.68, "learning_rate": 2.5933533293341333e-06, "loss": 0.0015, "step": 5699 }, { "epoch": 0.68, "learning_rate": 2.5932813437312537e-06, "loss": 0.0003, "step": 5700 }, { "epoch": 0.68, "learning_rate": 2.5932093581283745e-06, "loss": 0.0006, "step": 5701 }, { "epoch": 0.68, "learning_rate": 2.593137372525495e-06, "loss": 0.0005, "step": 5702 }, { "epoch": 0.68, "learning_rate": 2.5930653869226156e-06, "loss": 0.0005, "step": 5703 }, { "epoch": 0.68, "learning_rate": 2.592993401319736e-06, "loss": 0.0007, "step": 5704 }, { "epoch": 0.68, "learning_rate": 2.5929214157168567e-06, "loss": 0.5316, "step": 5705 }, { "epoch": 0.68, "learning_rate": 2.592849430113977e-06, "loss": 0.0005, "step": 5706 }, { "epoch": 0.68, "learning_rate": 2.592777444511098e-06, "loss": 0.0005, "step": 5707 }, { "epoch": 0.68, "learning_rate": 2.5927054589082182e-06, "loss": 0.0008, "step": 5708 }, { "epoch": 0.68, "learning_rate": 2.592633473305339e-06, "loss": 0.0006, "step": 5709 }, { "epoch": 0.68, "learning_rate": 2.59256148770246e-06, "loss": 0.0004, "step": 5710 }, { "epoch": 0.68, "learning_rate": 2.59248950209958e-06, "loss": 0.1496, "step": 5711 }, { "epoch": 0.68, "learning_rate": 2.592417516496701e-06, "loss": 0.0005, "step": 5712 }, { "epoch": 0.68, "learning_rate": 2.5923455308938213e-06, "loss": 0.0015, "step": 5713 }, { "epoch": 0.68, "learning_rate": 2.592273545290942e-06, "loss": 0.0008, "step": 5714 }, { "epoch": 0.68, "learning_rate": 2.5922015596880624e-06, "loss": 0.0006, "step": 5715 }, { "epoch": 0.68, "learning_rate": 2.592129574085183e-06, "loss": 0.0006, "step": 5716 }, { "epoch": 0.69, "learning_rate": 2.5920575884823035e-06, "loss": 0.8517, "step": 5717 }, { "epoch": 0.69, "learning_rate": 2.5919856028794243e-06, "loss": 0.0161, "step": 5718 }, { "epoch": 0.69, "learning_rate": 2.5919136172765447e-06, "loss": 0.0008, "step": 5719 }, { "epoch": 0.69, "learning_rate": 2.591841631673665e-06, "loss": 0.0003, "step": 5720 }, { "epoch": 0.69, "learning_rate": 2.591769646070786e-06, "loss": 0.0005, "step": 5721 }, { "epoch": 0.69, "learning_rate": 2.591697660467906e-06, "loss": 0.0232, "step": 5722 }, { "epoch": 0.69, "learning_rate": 2.591625674865027e-06, "loss": 0.0005, "step": 5723 }, { "epoch": 0.69, "learning_rate": 2.5915536892621477e-06, "loss": 0.0086, "step": 5724 }, { "epoch": 0.69, "learning_rate": 2.591481703659268e-06, "loss": 0.0003, "step": 5725 }, { "epoch": 0.69, "learning_rate": 2.591409718056389e-06, "loss": 0.0033, "step": 5726 }, { "epoch": 0.69, "learning_rate": 2.5913377324535096e-06, "loss": 0.0004, "step": 5727 }, { "epoch": 0.69, "learning_rate": 2.59126574685063e-06, "loss": 0.0005, "step": 5728 }, { "epoch": 0.69, "learning_rate": 2.5911937612477508e-06, "loss": 0.3012, "step": 5729 }, { "epoch": 0.69, "learning_rate": 2.591121775644871e-06, "loss": 0.8402, "step": 5730 }, { "epoch": 0.69, "learning_rate": 2.591049790041992e-06, "loss": 0.4309, "step": 5731 }, { "epoch": 0.69, "learning_rate": 2.5909778044391123e-06, "loss": 0.0006, "step": 5732 }, { "epoch": 0.69, "learning_rate": 2.5909058188362326e-06, "loss": 0.0005, "step": 5733 }, { "epoch": 0.69, "learning_rate": 2.5908338332333534e-06, "loss": 0.0004, "step": 5734 }, { "epoch": 0.69, "learning_rate": 2.5907618476304738e-06, "loss": 0.0014, "step": 5735 }, { "epoch": 0.69, "learning_rate": 2.5906898620275945e-06, "loss": 0.0008, "step": 5736 }, { "epoch": 0.69, "learning_rate": 2.590617876424715e-06, "loss": 0.0003, "step": 5737 }, { "epoch": 0.69, "learning_rate": 2.5905458908218357e-06, "loss": 0.0004, "step": 5738 }, { "epoch": 0.69, "learning_rate": 2.590473905218956e-06, "loss": 0.0008, "step": 5739 }, { "epoch": 0.69, "learning_rate": 2.590401919616077e-06, "loss": 0.0063, "step": 5740 }, { "epoch": 0.69, "learning_rate": 2.5903299340131976e-06, "loss": 0.0005, "step": 5741 }, { "epoch": 0.69, "learning_rate": 2.5902579484103184e-06, "loss": 0.0004, "step": 5742 }, { "epoch": 0.69, "learning_rate": 2.5901859628074387e-06, "loss": 0.0005, "step": 5743 }, { "epoch": 0.69, "learning_rate": 2.590113977204559e-06, "loss": 0.0005, "step": 5744 }, { "epoch": 0.69, "learning_rate": 2.59004199160168e-06, "loss": 0.5214, "step": 5745 }, { "epoch": 0.69, "learning_rate": 2.5899700059988e-06, "loss": 0.0004, "step": 5746 }, { "epoch": 0.69, "learning_rate": 2.589898020395921e-06, "loss": 0.0006, "step": 5747 }, { "epoch": 0.69, "learning_rate": 2.5898260347930413e-06, "loss": 0.5057, "step": 5748 }, { "epoch": 0.69, "learning_rate": 2.589754049190162e-06, "loss": 0.0023, "step": 5749 }, { "epoch": 0.69, "learning_rate": 2.5896820635872825e-06, "loss": 0.0006, "step": 5750 }, { "epoch": 0.69, "learning_rate": 2.5896100779844032e-06, "loss": 0.001, "step": 5751 }, { "epoch": 0.69, "learning_rate": 2.5895380923815236e-06, "loss": 0.6516, "step": 5752 }, { "epoch": 0.69, "learning_rate": 2.5894661067786444e-06, "loss": 0.0367, "step": 5753 }, { "epoch": 0.69, "learning_rate": 2.5893941211757647e-06, "loss": 0.0004, "step": 5754 }, { "epoch": 0.69, "learning_rate": 2.5893221355728855e-06, "loss": 0.7898, "step": 5755 }, { "epoch": 0.69, "learning_rate": 2.5892501499700063e-06, "loss": 0.0007, "step": 5756 }, { "epoch": 0.69, "learning_rate": 2.5891781643671266e-06, "loss": 0.0004, "step": 5757 }, { "epoch": 0.69, "learning_rate": 2.5891061787642474e-06, "loss": 0.9979, "step": 5758 }, { "epoch": 0.69, "learning_rate": 2.5890341931613678e-06, "loss": 0.0005, "step": 5759 }, { "epoch": 0.69, "learning_rate": 2.5889622075584886e-06, "loss": 0.0005, "step": 5760 }, { "epoch": 0.69, "learning_rate": 2.588890221955609e-06, "loss": 0.0004, "step": 5761 }, { "epoch": 0.69, "learning_rate": 2.5888182363527297e-06, "loss": 0.0015, "step": 5762 }, { "epoch": 0.69, "learning_rate": 2.58874625074985e-06, "loss": 0.004, "step": 5763 }, { "epoch": 0.69, "learning_rate": 2.588674265146971e-06, "loss": 0.0004, "step": 5764 }, { "epoch": 0.69, "learning_rate": 2.588602279544091e-06, "loss": 0.8371, "step": 5765 }, { "epoch": 0.69, "learning_rate": 2.588530293941212e-06, "loss": 0.6453, "step": 5766 }, { "epoch": 0.69, "learning_rate": 2.5884583083383323e-06, "loss": 0.0004, "step": 5767 }, { "epoch": 0.69, "learning_rate": 2.5883863227354527e-06, "loss": 0.7676, "step": 5768 }, { "epoch": 0.69, "learning_rate": 2.5883143371325734e-06, "loss": 0.0008, "step": 5769 }, { "epoch": 0.69, "learning_rate": 2.5882423515296942e-06, "loss": 0.0004, "step": 5770 }, { "epoch": 0.69, "learning_rate": 2.5881703659268146e-06, "loss": 0.0042, "step": 5771 }, { "epoch": 0.69, "learning_rate": 2.5880983803239354e-06, "loss": 0.0037, "step": 5772 }, { "epoch": 0.69, "learning_rate": 2.588026394721056e-06, "loss": 0.0005, "step": 5773 }, { "epoch": 0.69, "learning_rate": 2.5879544091181765e-06, "loss": 0.0008, "step": 5774 }, { "epoch": 0.69, "learning_rate": 2.5878824235152973e-06, "loss": 0.0004, "step": 5775 }, { "epoch": 0.69, "learning_rate": 2.5878104379124176e-06, "loss": 0.0006, "step": 5776 }, { "epoch": 0.69, "learning_rate": 2.5877384523095384e-06, "loss": 0.0026, "step": 5777 }, { "epoch": 0.69, "learning_rate": 2.5876664667066588e-06, "loss": 0.0005, "step": 5778 }, { "epoch": 0.69, "learning_rate": 2.587594481103779e-06, "loss": 0.0004, "step": 5779 }, { "epoch": 0.69, "learning_rate": 2.5875224955009e-06, "loss": 0.0006, "step": 5780 }, { "epoch": 0.69, "learning_rate": 2.5874505098980202e-06, "loss": 0.0005, "step": 5781 }, { "epoch": 0.69, "learning_rate": 2.587378524295141e-06, "loss": 0.0013, "step": 5782 }, { "epoch": 0.69, "learning_rate": 2.5873065386922614e-06, "loss": 0.0004, "step": 5783 }, { "epoch": 0.69, "learning_rate": 2.587234553089382e-06, "loss": 0.0255, "step": 5784 }, { "epoch": 0.69, "learning_rate": 2.5871625674865025e-06, "loss": 0.0004, "step": 5785 }, { "epoch": 0.69, "learning_rate": 2.5870905818836233e-06, "loss": 0.0006, "step": 5786 }, { "epoch": 0.69, "learning_rate": 2.587018596280744e-06, "loss": 0.0439, "step": 5787 }, { "epoch": 0.69, "learning_rate": 2.586946610677865e-06, "loss": 0.0008, "step": 5788 }, { "epoch": 0.69, "learning_rate": 2.586874625074985e-06, "loss": 0.0005, "step": 5789 }, { "epoch": 0.69, "learning_rate": 2.5868026394721056e-06, "loss": 0.0009, "step": 5790 }, { "epoch": 0.69, "learning_rate": 2.5867306538692263e-06, "loss": 0.0006, "step": 5791 }, { "epoch": 0.69, "learning_rate": 2.5866586682663467e-06, "loss": 0.0005, "step": 5792 }, { "epoch": 0.69, "learning_rate": 2.5865866826634675e-06, "loss": 0.0007, "step": 5793 }, { "epoch": 0.69, "learning_rate": 2.586514697060588e-06, "loss": 0.0005, "step": 5794 }, { "epoch": 0.69, "learning_rate": 2.5864427114577086e-06, "loss": 0.002, "step": 5795 }, { "epoch": 0.69, "learning_rate": 2.586370725854829e-06, "loss": 0.0007, "step": 5796 }, { "epoch": 0.69, "learning_rate": 2.5862987402519497e-06, "loss": 0.0004, "step": 5797 }, { "epoch": 0.69, "learning_rate": 2.58622675464907e-06, "loss": 0.0004, "step": 5798 }, { "epoch": 0.69, "learning_rate": 2.586154769046191e-06, "loss": 0.0007, "step": 5799 }, { "epoch": 0.7, "learning_rate": 2.5860827834433112e-06, "loss": 0.0166, "step": 5800 }, { "epoch": 0.7, "learning_rate": 2.586010797840432e-06, "loss": 0.0007, "step": 5801 }, { "epoch": 0.7, "learning_rate": 2.5859388122375528e-06, "loss": 0.0006, "step": 5802 }, { "epoch": 0.7, "learning_rate": 2.585866826634673e-06, "loss": 0.0004, "step": 5803 }, { "epoch": 0.7, "learning_rate": 2.585794841031794e-06, "loss": 0.0005, "step": 5804 }, { "epoch": 0.7, "learning_rate": 2.5857228554289143e-06, "loss": 0.0005, "step": 5805 }, { "epoch": 0.7, "learning_rate": 2.585650869826035e-06, "loss": 0.0004, "step": 5806 }, { "epoch": 0.7, "learning_rate": 2.5855788842231554e-06, "loss": 0.0004, "step": 5807 }, { "epoch": 0.7, "learning_rate": 2.585506898620276e-06, "loss": 0.8086, "step": 5808 }, { "epoch": 0.7, "learning_rate": 2.5854349130173965e-06, "loss": 0.0325, "step": 5809 }, { "epoch": 0.7, "learning_rate": 2.5853629274145173e-06, "loss": 0.0006, "step": 5810 }, { "epoch": 0.7, "learning_rate": 2.5852909418116377e-06, "loss": 0.0007, "step": 5811 }, { "epoch": 0.7, "learning_rate": 2.5852189562087585e-06, "loss": 0.0006, "step": 5812 }, { "epoch": 0.7, "learning_rate": 2.585146970605879e-06, "loss": 0.0005, "step": 5813 }, { "epoch": 0.7, "learning_rate": 2.585074985002999e-06, "loss": 0.0003, "step": 5814 }, { "epoch": 0.7, "learning_rate": 2.58500299940012e-06, "loss": 0.0003, "step": 5815 }, { "epoch": 0.7, "learning_rate": 2.5849310137972403e-06, "loss": 0.0008, "step": 5816 }, { "epoch": 0.7, "learning_rate": 2.584859028194361e-06, "loss": 0.0011, "step": 5817 }, { "epoch": 0.7, "learning_rate": 2.584787042591482e-06, "loss": 1.0841, "step": 5818 }, { "epoch": 0.7, "learning_rate": 2.5847150569886026e-06, "loss": 0.0007, "step": 5819 }, { "epoch": 0.7, "learning_rate": 2.584643071385723e-06, "loss": 0.0007, "step": 5820 }, { "epoch": 0.7, "learning_rate": 2.5845710857828438e-06, "loss": 0.1586, "step": 5821 }, { "epoch": 0.7, "learning_rate": 2.584499100179964e-06, "loss": 0.0003, "step": 5822 }, { "epoch": 0.7, "learning_rate": 2.584427114577085e-06, "loss": 0.0009, "step": 5823 }, { "epoch": 0.7, "learning_rate": 2.5843551289742053e-06, "loss": 0.2032, "step": 5824 }, { "epoch": 0.7, "learning_rate": 2.5842831433713256e-06, "loss": 0.0004, "step": 5825 }, { "epoch": 0.7, "learning_rate": 2.5842111577684464e-06, "loss": 0.0023, "step": 5826 }, { "epoch": 0.7, "learning_rate": 2.5841391721655667e-06, "loss": 0.0005, "step": 5827 }, { "epoch": 0.7, "learning_rate": 2.5840671865626875e-06, "loss": 0.0005, "step": 5828 }, { "epoch": 0.7, "learning_rate": 2.583995200959808e-06, "loss": 0.0004, "step": 5829 }, { "epoch": 0.7, "learning_rate": 2.5839232153569287e-06, "loss": 0.0003, "step": 5830 }, { "epoch": 0.7, "learning_rate": 2.583851229754049e-06, "loss": 0.0005, "step": 5831 }, { "epoch": 0.7, "learning_rate": 2.5837792441511698e-06, "loss": 0.0006, "step": 5832 }, { "epoch": 0.7, "learning_rate": 2.5837072585482906e-06, "loss": 0.0007, "step": 5833 }, { "epoch": 0.7, "learning_rate": 2.583635272945411e-06, "loss": 0.0014, "step": 5834 }, { "epoch": 0.7, "learning_rate": 2.5835632873425317e-06, "loss": 0.0015, "step": 5835 }, { "epoch": 0.7, "learning_rate": 2.583491301739652e-06, "loss": 0.0008, "step": 5836 }, { "epoch": 0.7, "learning_rate": 2.583419316136773e-06, "loss": 0.0008, "step": 5837 }, { "epoch": 0.7, "learning_rate": 2.583347330533893e-06, "loss": 0.0003, "step": 5838 }, { "epoch": 0.7, "learning_rate": 2.583275344931014e-06, "loss": 0.0009, "step": 5839 }, { "epoch": 0.7, "learning_rate": 2.5832033593281343e-06, "loss": 0.0012, "step": 5840 }, { "epoch": 0.7, "learning_rate": 2.583131373725255e-06, "loss": 0.0004, "step": 5841 }, { "epoch": 0.7, "learning_rate": 2.5830593881223755e-06, "loss": 0.0009, "step": 5842 }, { "epoch": 0.7, "learning_rate": 2.5829874025194962e-06, "loss": 0.0004, "step": 5843 }, { "epoch": 0.7, "learning_rate": 2.5829154169166166e-06, "loss": 0.0003, "step": 5844 }, { "epoch": 0.7, "learning_rate": 2.5828434313137374e-06, "loss": 0.0083, "step": 5845 }, { "epoch": 0.7, "learning_rate": 2.5827714457108577e-06, "loss": 0.0012, "step": 5846 }, { "epoch": 0.7, "learning_rate": 2.5826994601079785e-06, "loss": 0.0005, "step": 5847 }, { "epoch": 0.7, "learning_rate": 2.582627474505099e-06, "loss": 0.002, "step": 5848 }, { "epoch": 0.7, "learning_rate": 2.5825554889022196e-06, "loss": 0.0006, "step": 5849 }, { "epoch": 0.7, "learning_rate": 2.5824835032993404e-06, "loss": 0.6113, "step": 5850 }, { "epoch": 0.7, "learning_rate": 2.5824115176964608e-06, "loss": 0.0006, "step": 5851 }, { "epoch": 0.7, "learning_rate": 2.5823395320935815e-06, "loss": 0.0006, "step": 5852 }, { "epoch": 0.7, "learning_rate": 2.582267546490702e-06, "loss": 0.0004, "step": 5853 }, { "epoch": 0.7, "learning_rate": 2.5821955608878227e-06, "loss": 0.0815, "step": 5854 }, { "epoch": 0.7, "learning_rate": 2.582123575284943e-06, "loss": 0.0005, "step": 5855 }, { "epoch": 0.7, "learning_rate": 2.582051589682064e-06, "loss": 0.0007, "step": 5856 }, { "epoch": 0.7, "learning_rate": 2.581979604079184e-06, "loss": 0.0004, "step": 5857 }, { "epoch": 0.7, "learning_rate": 2.581907618476305e-06, "loss": 0.0006, "step": 5858 }, { "epoch": 0.7, "learning_rate": 2.5818356328734253e-06, "loss": 0.565, "step": 5859 }, { "epoch": 0.7, "learning_rate": 2.5817636472705457e-06, "loss": 0.0012, "step": 5860 }, { "epoch": 0.7, "learning_rate": 2.5816916616676664e-06, "loss": 0.0005, "step": 5861 }, { "epoch": 0.7, "learning_rate": 2.581619676064787e-06, "loss": 0.0006, "step": 5862 }, { "epoch": 0.7, "learning_rate": 2.5815476904619076e-06, "loss": 0.0068, "step": 5863 }, { "epoch": 0.7, "learning_rate": 2.5814757048590283e-06, "loss": 0.0035, "step": 5864 }, { "epoch": 0.7, "learning_rate": 2.581403719256149e-06, "loss": 0.0005, "step": 5865 }, { "epoch": 0.7, "learning_rate": 2.5813317336532695e-06, "loss": 0.0007, "step": 5866 }, { "epoch": 0.7, "learning_rate": 2.5812597480503903e-06, "loss": 0.0008, "step": 5867 }, { "epoch": 0.7, "learning_rate": 2.5811877624475106e-06, "loss": 0.0009, "step": 5868 }, { "epoch": 0.7, "learning_rate": 2.5811157768446314e-06, "loss": 0.2796, "step": 5869 }, { "epoch": 0.7, "learning_rate": 2.5810437912417517e-06, "loss": 1.485, "step": 5870 }, { "epoch": 0.7, "learning_rate": 2.580971805638872e-06, "loss": 0.0005, "step": 5871 }, { "epoch": 0.7, "learning_rate": 2.580899820035993e-06, "loss": 0.855, "step": 5872 }, { "epoch": 0.7, "learning_rate": 2.5808278344331132e-06, "loss": 0.0019, "step": 5873 }, { "epoch": 0.7, "learning_rate": 2.580755848830234e-06, "loss": 0.0004, "step": 5874 }, { "epoch": 0.7, "learning_rate": 2.5806838632273544e-06, "loss": 0.0003, "step": 5875 }, { "epoch": 0.7, "learning_rate": 2.580611877624475e-06, "loss": 0.0345, "step": 5876 }, { "epoch": 0.7, "learning_rate": 2.5805398920215955e-06, "loss": 0.0003, "step": 5877 }, { "epoch": 0.7, "learning_rate": 2.5804679064187163e-06, "loss": 0.0004, "step": 5878 }, { "epoch": 0.7, "learning_rate": 2.580395920815837e-06, "loss": 0.0795, "step": 5879 }, { "epoch": 0.7, "learning_rate": 2.5803239352129574e-06, "loss": 0.0007, "step": 5880 }, { "epoch": 0.7, "learning_rate": 2.580251949610078e-06, "loss": 0.0007, "step": 5881 }, { "epoch": 0.7, "learning_rate": 2.5801799640071986e-06, "loss": 0.0005, "step": 5882 }, { "epoch": 0.7, "learning_rate": 2.5801079784043193e-06, "loss": 0.0003, "step": 5883 }, { "epoch": 0.71, "learning_rate": 2.5800359928014397e-06, "loss": 0.0007, "step": 5884 }, { "epoch": 0.71, "learning_rate": 2.5799640071985605e-06, "loss": 0.0011, "step": 5885 }, { "epoch": 0.71, "learning_rate": 2.579892021595681e-06, "loss": 0.0666, "step": 5886 }, { "epoch": 0.71, "learning_rate": 2.5798200359928016e-06, "loss": 0.0006, "step": 5887 }, { "epoch": 0.71, "learning_rate": 2.579748050389922e-06, "loss": 0.0005, "step": 5888 }, { "epoch": 0.71, "learning_rate": 2.5796760647870427e-06, "loss": 0.5434, "step": 5889 }, { "epoch": 0.71, "learning_rate": 2.579604079184163e-06, "loss": 0.0018, "step": 5890 }, { "epoch": 0.71, "learning_rate": 2.579532093581284e-06, "loss": 0.0007, "step": 5891 }, { "epoch": 0.71, "learning_rate": 2.5794601079784042e-06, "loss": 0.0003, "step": 5892 }, { "epoch": 0.71, "learning_rate": 2.579388122375525e-06, "loss": 0.0004, "step": 5893 }, { "epoch": 0.71, "learning_rate": 2.5793161367726454e-06, "loss": 0.0008, "step": 5894 }, { "epoch": 0.71, "learning_rate": 2.579244151169766e-06, "loss": 0.612, "step": 5895 }, { "epoch": 0.71, "learning_rate": 2.579172165566887e-06, "loss": 0.0002, "step": 5896 }, { "epoch": 0.71, "learning_rate": 2.5791001799640073e-06, "loss": 0.0004, "step": 5897 }, { "epoch": 0.71, "learning_rate": 2.579028194361128e-06, "loss": 0.0004, "step": 5898 }, { "epoch": 0.71, "learning_rate": 2.5789562087582484e-06, "loss": 0.0004, "step": 5899 }, { "epoch": 0.71, "learning_rate": 2.578884223155369e-06, "loss": 0.0006, "step": 5900 }, { "epoch": 0.71, "learning_rate": 2.5788122375524895e-06, "loss": 0.0005, "step": 5901 }, { "epoch": 0.71, "learning_rate": 2.5787402519496103e-06, "loss": 0.0009, "step": 5902 }, { "epoch": 0.71, "learning_rate": 2.5786682663467307e-06, "loss": 0.0004, "step": 5903 }, { "epoch": 0.71, "learning_rate": 2.5785962807438514e-06, "loss": 0.0005, "step": 5904 }, { "epoch": 0.71, "learning_rate": 2.578524295140972e-06, "loss": 0.84, "step": 5905 }, { "epoch": 0.71, "learning_rate": 2.578452309538092e-06, "loss": 0.0005, "step": 5906 }, { "epoch": 0.71, "learning_rate": 2.578380323935213e-06, "loss": 0.6687, "step": 5907 }, { "epoch": 0.71, "learning_rate": 2.5783083383323333e-06, "loss": 0.7942, "step": 5908 }, { "epoch": 0.71, "learning_rate": 2.578236352729454e-06, "loss": 0.0005, "step": 5909 }, { "epoch": 0.71, "learning_rate": 2.578164367126575e-06, "loss": 0.3712, "step": 5910 }, { "epoch": 0.71, "learning_rate": 2.5780923815236956e-06, "loss": 0.0011, "step": 5911 }, { "epoch": 0.71, "learning_rate": 2.578020395920816e-06, "loss": 0.8091, "step": 5912 }, { "epoch": 0.71, "learning_rate": 2.5779484103179368e-06, "loss": 0.0004, "step": 5913 }, { "epoch": 0.71, "learning_rate": 2.577876424715057e-06, "loss": 0.0006, "step": 5914 }, { "epoch": 0.71, "learning_rate": 2.577804439112178e-06, "loss": 0.7874, "step": 5915 }, { "epoch": 0.71, "learning_rate": 2.5777324535092982e-06, "loss": 0.0007, "step": 5916 }, { "epoch": 0.71, "learning_rate": 2.5776604679064186e-06, "loss": 0.0005, "step": 5917 }, { "epoch": 0.71, "learning_rate": 2.5775884823035394e-06, "loss": 0.8925, "step": 5918 }, { "epoch": 0.71, "learning_rate": 2.5775164967006597e-06, "loss": 0.2279, "step": 5919 }, { "epoch": 0.71, "learning_rate": 2.5774445110977805e-06, "loss": 0.0004, "step": 5920 }, { "epoch": 0.71, "learning_rate": 2.577372525494901e-06, "loss": 1.5179, "step": 5921 }, { "epoch": 0.71, "learning_rate": 2.5773005398920216e-06, "loss": 0.0003, "step": 5922 }, { "epoch": 0.71, "learning_rate": 2.577228554289142e-06, "loss": 0.0004, "step": 5923 }, { "epoch": 0.71, "learning_rate": 2.5771565686862628e-06, "loss": 0.0181, "step": 5924 }, { "epoch": 0.71, "learning_rate": 2.5770845830833836e-06, "loss": 0.0011, "step": 5925 }, { "epoch": 0.71, "learning_rate": 2.577012597480504e-06, "loss": 0.0006, "step": 5926 }, { "epoch": 0.71, "learning_rate": 2.5769406118776247e-06, "loss": 0.8316, "step": 5927 }, { "epoch": 0.71, "learning_rate": 2.576868626274745e-06, "loss": 0.0007, "step": 5928 }, { "epoch": 0.71, "learning_rate": 2.576796640671866e-06, "loss": 0.0007, "step": 5929 }, { "epoch": 0.71, "learning_rate": 2.576724655068986e-06, "loss": 0.0005, "step": 5930 }, { "epoch": 0.71, "learning_rate": 2.576652669466107e-06, "loss": 0.0009, "step": 5931 }, { "epoch": 0.71, "learning_rate": 2.5765806838632273e-06, "loss": 0.0007, "step": 5932 }, { "epoch": 0.71, "learning_rate": 2.576508698260348e-06, "loss": 0.0021, "step": 5933 }, { "epoch": 0.71, "learning_rate": 2.5764367126574684e-06, "loss": 0.0026, "step": 5934 }, { "epoch": 0.71, "learning_rate": 2.5763647270545892e-06, "loss": 0.0016, "step": 5935 }, { "epoch": 0.71, "learning_rate": 2.5762927414517096e-06, "loss": 0.4878, "step": 5936 }, { "epoch": 0.71, "learning_rate": 2.5762207558488304e-06, "loss": 0.001, "step": 5937 }, { "epoch": 0.71, "learning_rate": 2.5761487702459507e-06, "loss": 0.0015, "step": 5938 }, { "epoch": 0.71, "learning_rate": 2.5760767846430715e-06, "loss": 0.3098, "step": 5939 }, { "epoch": 0.71, "learning_rate": 2.576004799040192e-06, "loss": 0.2752, "step": 5940 }, { "epoch": 0.71, "learning_rate": 2.5759328134373126e-06, "loss": 0.0008, "step": 5941 }, { "epoch": 0.71, "learning_rate": 2.5758608278344334e-06, "loss": 0.9625, "step": 5942 }, { "epoch": 0.71, "learning_rate": 2.5757888422315538e-06, "loss": 0.0021, "step": 5943 }, { "epoch": 0.71, "learning_rate": 2.5757168566286745e-06, "loss": 0.0005, "step": 5944 }, { "epoch": 0.71, "learning_rate": 2.575644871025795e-06, "loss": 0.099, "step": 5945 }, { "epoch": 0.71, "learning_rate": 2.5755728854229157e-06, "loss": 0.0008, "step": 5946 }, { "epoch": 0.71, "learning_rate": 2.575500899820036e-06, "loss": 0.0007, "step": 5947 }, { "epoch": 0.71, "learning_rate": 2.575428914217157e-06, "loss": 0.3283, "step": 5948 }, { "epoch": 0.71, "learning_rate": 2.575356928614277e-06, "loss": 0.0008, "step": 5949 }, { "epoch": 0.71, "learning_rate": 2.575284943011398e-06, "loss": 0.0006, "step": 5950 }, { "epoch": 0.71, "learning_rate": 2.5752129574085183e-06, "loss": 0.0004, "step": 5951 }, { "epoch": 0.71, "learning_rate": 2.5751409718056386e-06, "loss": 0.0005, "step": 5952 }, { "epoch": 0.71, "learning_rate": 2.5750689862027594e-06, "loss": 0.0026, "step": 5953 }, { "epoch": 0.71, "learning_rate": 2.5749970005998798e-06, "loss": 0.0016, "step": 5954 }, { "epoch": 0.71, "learning_rate": 2.5749250149970006e-06, "loss": 0.002, "step": 5955 }, { "epoch": 0.71, "learning_rate": 2.5748530293941213e-06, "loss": 0.0006, "step": 5956 }, { "epoch": 0.71, "learning_rate": 2.5747810437912417e-06, "loss": 0.0011, "step": 5957 }, { "epoch": 0.71, "learning_rate": 2.5747090581883625e-06, "loss": 0.0009, "step": 5958 }, { "epoch": 0.71, "learning_rate": 2.5746370725854833e-06, "loss": 0.4118, "step": 5959 }, { "epoch": 0.71, "learning_rate": 2.5745650869826036e-06, "loss": 0.2191, "step": 5960 }, { "epoch": 0.71, "learning_rate": 2.5744931013797244e-06, "loss": 0.0006, "step": 5961 }, { "epoch": 0.71, "learning_rate": 2.5744211157768447e-06, "loss": 0.0007, "step": 5962 }, { "epoch": 0.71, "learning_rate": 2.574349130173965e-06, "loss": 0.001, "step": 5963 }, { "epoch": 0.71, "learning_rate": 2.574277144571086e-06, "loss": 0.0009, "step": 5964 }, { "epoch": 0.71, "learning_rate": 2.5742051589682062e-06, "loss": 0.0008, "step": 5965 }, { "epoch": 0.71, "learning_rate": 2.574133173365327e-06, "loss": 0.8342, "step": 5966 }, { "epoch": 0.72, "learning_rate": 2.5740611877624474e-06, "loss": 0.0015, "step": 5967 }, { "epoch": 0.72, "learning_rate": 2.573989202159568e-06, "loss": 0.0006, "step": 5968 }, { "epoch": 0.72, "learning_rate": 2.5739172165566885e-06, "loss": 0.002, "step": 5969 }, { "epoch": 0.72, "learning_rate": 2.5738452309538093e-06, "loss": 0.0009, "step": 5970 }, { "epoch": 0.72, "learning_rate": 2.5737732453509296e-06, "loss": 0.0051, "step": 5971 }, { "epoch": 0.72, "learning_rate": 2.5737012597480504e-06, "loss": 0.0008, "step": 5972 }, { "epoch": 0.72, "learning_rate": 2.573629274145171e-06, "loss": 0.0018, "step": 5973 }, { "epoch": 0.72, "learning_rate": 2.5735572885422915e-06, "loss": 0.0006, "step": 5974 }, { "epoch": 0.72, "learning_rate": 2.5734853029394123e-06, "loss": 0.0005, "step": 5975 }, { "epoch": 0.72, "learning_rate": 2.5734133173365327e-06, "loss": 0.0006, "step": 5976 }, { "epoch": 0.72, "learning_rate": 2.5733413317336535e-06, "loss": 0.0005, "step": 5977 }, { "epoch": 0.72, "learning_rate": 2.573269346130774e-06, "loss": 0.0005, "step": 5978 }, { "epoch": 0.72, "learning_rate": 2.5731973605278946e-06, "loss": 0.0005, "step": 5979 }, { "epoch": 0.72, "learning_rate": 2.573125374925015e-06, "loss": 0.0004, "step": 5980 }, { "epoch": 0.72, "learning_rate": 2.5730533893221357e-06, "loss": 0.0095, "step": 5981 }, { "epoch": 0.72, "learning_rate": 2.572981403719256e-06, "loss": 0.0009, "step": 5982 }, { "epoch": 0.72, "learning_rate": 2.572909418116377e-06, "loss": 0.0026, "step": 5983 }, { "epoch": 0.72, "learning_rate": 2.572837432513497e-06, "loss": 0.8379, "step": 5984 }, { "epoch": 0.72, "learning_rate": 2.572765446910618e-06, "loss": 0.0005, "step": 5985 }, { "epoch": 0.72, "learning_rate": 2.5726934613077383e-06, "loss": 0.0007, "step": 5986 }, { "epoch": 0.72, "learning_rate": 2.572621475704859e-06, "loss": 0.001, "step": 5987 }, { "epoch": 0.72, "learning_rate": 2.57254949010198e-06, "loss": 0.0008, "step": 5988 }, { "epoch": 0.72, "learning_rate": 2.5724775044991003e-06, "loss": 0.0232, "step": 5989 }, { "epoch": 0.72, "learning_rate": 2.572405518896221e-06, "loss": 0.0021, "step": 5990 }, { "epoch": 0.72, "learning_rate": 2.5723335332933414e-06, "loss": 0.3922, "step": 5991 }, { "epoch": 0.72, "learning_rate": 2.572261547690462e-06, "loss": 0.0004, "step": 5992 }, { "epoch": 0.72, "learning_rate": 2.5721895620875825e-06, "loss": 0.0008, "step": 5993 }, { "epoch": 0.72, "learning_rate": 2.5721175764847033e-06, "loss": 0.0015, "step": 5994 }, { "epoch": 0.72, "learning_rate": 2.5720455908818237e-06, "loss": 0.0005, "step": 5995 }, { "epoch": 0.72, "learning_rate": 2.5719736052789444e-06, "loss": 0.0017, "step": 5996 }, { "epoch": 0.72, "learning_rate": 2.571901619676065e-06, "loss": 0.0161, "step": 5997 }, { "epoch": 0.72, "learning_rate": 2.571829634073185e-06, "loss": 0.001, "step": 5998 }, { "epoch": 0.72, "learning_rate": 2.571757648470306e-06, "loss": 0.0035, "step": 5999 }, { "epoch": 0.72, "learning_rate": 2.5716856628674263e-06, "loss": 0.0005, "step": 6000 }, { "epoch": 0.72, "learning_rate": 2.571613677264547e-06, "loss": 0.9807, "step": 6001 }, { "epoch": 0.72, "learning_rate": 2.571541691661668e-06, "loss": 0.7229, "step": 6002 }, { "epoch": 0.72, "learning_rate": 2.571469706058788e-06, "loss": 0.0011, "step": 6003 }, { "epoch": 0.72, "learning_rate": 2.571397720455909e-06, "loss": 0.0006, "step": 6004 }, { "epoch": 0.72, "learning_rate": 2.5713257348530297e-06, "loss": 0.0005, "step": 6005 }, { "epoch": 0.72, "learning_rate": 2.57125374925015e-06, "loss": 0.0005, "step": 6006 }, { "epoch": 0.72, "learning_rate": 2.571181763647271e-06, "loss": 0.0012, "step": 6007 }, { "epoch": 0.72, "learning_rate": 2.5711097780443912e-06, "loss": 0.0005, "step": 6008 }, { "epoch": 0.72, "learning_rate": 2.5710377924415116e-06, "loss": 0.0071, "step": 6009 }, { "epoch": 0.72, "learning_rate": 2.5709658068386324e-06, "loss": 0.0005, "step": 6010 }, { "epoch": 0.72, "learning_rate": 2.5708938212357527e-06, "loss": 0.0007, "step": 6011 }, { "epoch": 0.72, "learning_rate": 2.5708218356328735e-06, "loss": 0.0007, "step": 6012 }, { "epoch": 0.72, "learning_rate": 2.570749850029994e-06, "loss": 0.0004, "step": 6013 }, { "epoch": 0.72, "learning_rate": 2.5706778644271146e-06, "loss": 0.0043, "step": 6014 }, { "epoch": 0.72, "learning_rate": 2.570605878824235e-06, "loss": 0.0006, "step": 6015 }, { "epoch": 0.72, "learning_rate": 2.5705338932213558e-06, "loss": 0.0005, "step": 6016 }, { "epoch": 0.72, "learning_rate": 2.570461907618476e-06, "loss": 0.0004, "step": 6017 }, { "epoch": 0.72, "learning_rate": 2.570389922015597e-06, "loss": 0.0008, "step": 6018 }, { "epoch": 0.72, "learning_rate": 2.5703179364127177e-06, "loss": 0.0006, "step": 6019 }, { "epoch": 0.72, "learning_rate": 2.5702459508098385e-06, "loss": 0.001, "step": 6020 }, { "epoch": 0.72, "learning_rate": 2.570173965206959e-06, "loss": 0.0004, "step": 6021 }, { "epoch": 0.72, "learning_rate": 2.570101979604079e-06, "loss": 0.138, "step": 6022 }, { "epoch": 0.72, "learning_rate": 2.5700299940012e-06, "loss": 0.001, "step": 6023 }, { "epoch": 0.72, "learning_rate": 2.5699580083983203e-06, "loss": 0.8775, "step": 6024 }, { "epoch": 0.72, "learning_rate": 2.569886022795441e-06, "loss": 0.0007, "step": 6025 }, { "epoch": 0.72, "learning_rate": 2.5698140371925614e-06, "loss": 0.0012, "step": 6026 }, { "epoch": 0.72, "learning_rate": 2.5697420515896822e-06, "loss": 0.0016, "step": 6027 }, { "epoch": 0.72, "learning_rate": 2.5696700659868026e-06, "loss": 0.0005, "step": 6028 }, { "epoch": 0.72, "learning_rate": 2.5695980803839234e-06, "loss": 0.0004, "step": 6029 }, { "epoch": 0.72, "learning_rate": 2.5695260947810437e-06, "loss": 0.0004, "step": 6030 }, { "epoch": 0.72, "learning_rate": 2.5694541091781645e-06, "loss": 0.0017, "step": 6031 }, { "epoch": 0.72, "learning_rate": 2.569382123575285e-06, "loss": 0.0006, "step": 6032 }, { "epoch": 0.72, "learning_rate": 2.5693101379724056e-06, "loss": 0.0018, "step": 6033 }, { "epoch": 0.72, "learning_rate": 2.5692381523695264e-06, "loss": 0.0007, "step": 6034 }, { "epoch": 0.72, "learning_rate": 2.5691661667666468e-06, "loss": 0.9468, "step": 6035 }, { "epoch": 0.72, "learning_rate": 2.5690941811637675e-06, "loss": 0.0008, "step": 6036 }, { "epoch": 0.72, "learning_rate": 2.569022195560888e-06, "loss": 0.7615, "step": 6037 }, { "epoch": 0.72, "learning_rate": 2.5689502099580087e-06, "loss": 0.0006, "step": 6038 }, { "epoch": 0.72, "learning_rate": 2.568878224355129e-06, "loss": 0.0007, "step": 6039 }, { "epoch": 0.72, "learning_rate": 2.56880623875225e-06, "loss": 0.0005, "step": 6040 }, { "epoch": 0.72, "learning_rate": 2.56873425314937e-06, "loss": 0.0005, "step": 6041 }, { "epoch": 0.72, "learning_rate": 2.568662267546491e-06, "loss": 0.0006, "step": 6042 }, { "epoch": 0.72, "learning_rate": 2.5685902819436113e-06, "loss": 0.0008, "step": 6043 }, { "epoch": 0.72, "learning_rate": 2.5685182963407316e-06, "loss": 0.5446, "step": 6044 }, { "epoch": 0.72, "learning_rate": 2.5684463107378524e-06, "loss": 0.001, "step": 6045 }, { "epoch": 0.72, "learning_rate": 2.5683743251349728e-06, "loss": 0.0085, "step": 6046 }, { "epoch": 0.72, "learning_rate": 2.5683023395320936e-06, "loss": 0.001, "step": 6047 }, { "epoch": 0.72, "learning_rate": 2.5682303539292143e-06, "loss": 0.0008, "step": 6048 }, { "epoch": 0.72, "learning_rate": 2.5681583683263347e-06, "loss": 0.0008, "step": 6049 }, { "epoch": 0.72, "learning_rate": 2.5680863827234555e-06, "loss": 0.0017, "step": 6050 }, { "epoch": 0.73, "learning_rate": 2.5680143971205762e-06, "loss": 0.0009, "step": 6051 }, { "epoch": 0.73, "learning_rate": 2.5679424115176966e-06, "loss": 0.9206, "step": 6052 }, { "epoch": 0.73, "learning_rate": 2.5678704259148174e-06, "loss": 0.0009, "step": 6053 }, { "epoch": 0.73, "learning_rate": 2.5677984403119377e-06, "loss": 0.0873, "step": 6054 }, { "epoch": 0.73, "learning_rate": 2.567726454709058e-06, "loss": 0.0018, "step": 6055 }, { "epoch": 0.73, "learning_rate": 2.567654469106179e-06, "loss": 0.0015, "step": 6056 }, { "epoch": 0.73, "learning_rate": 2.5675824835032992e-06, "loss": 0.0009, "step": 6057 }, { "epoch": 0.73, "learning_rate": 2.56751049790042e-06, "loss": 0.0007, "step": 6058 }, { "epoch": 0.73, "learning_rate": 2.5674385122975404e-06, "loss": 0.0013, "step": 6059 }, { "epoch": 0.73, "learning_rate": 2.567366526694661e-06, "loss": 0.0009, "step": 6060 }, { "epoch": 0.73, "learning_rate": 2.5672945410917815e-06, "loss": 0.1768, "step": 6061 }, { "epoch": 0.73, "learning_rate": 2.5672225554889023e-06, "loss": 0.0007, "step": 6062 }, { "epoch": 0.73, "learning_rate": 2.5671505698860226e-06, "loss": 0.0007, "step": 6063 }, { "epoch": 0.73, "learning_rate": 2.5670785842831434e-06, "loss": 0.3416, "step": 6064 }, { "epoch": 0.73, "learning_rate": 2.567006598680264e-06, "loss": 0.0006, "step": 6065 }, { "epoch": 0.73, "learning_rate": 2.566934613077385e-06, "loss": 0.0009, "step": 6066 }, { "epoch": 0.73, "learning_rate": 2.5668626274745053e-06, "loss": 0.0009, "step": 6067 }, { "epoch": 0.73, "learning_rate": 2.5667906418716257e-06, "loss": 0.002, "step": 6068 }, { "epoch": 0.73, "learning_rate": 2.5667186562687464e-06, "loss": 0.001, "step": 6069 }, { "epoch": 0.73, "learning_rate": 2.566646670665867e-06, "loss": 0.0011, "step": 6070 }, { "epoch": 0.73, "learning_rate": 2.5665746850629876e-06, "loss": 0.0005, "step": 6071 }, { "epoch": 0.73, "learning_rate": 2.566502699460108e-06, "loss": 0.0006, "step": 6072 }, { "epoch": 0.73, "learning_rate": 2.5664307138572287e-06, "loss": 0.0027, "step": 6073 }, { "epoch": 0.73, "learning_rate": 2.566358728254349e-06, "loss": 0.7847, "step": 6074 }, { "epoch": 0.73, "learning_rate": 2.56628674265147e-06, "loss": 0.0012, "step": 6075 }, { "epoch": 0.73, "learning_rate": 2.56621475704859e-06, "loss": 0.0005, "step": 6076 }, { "epoch": 0.73, "learning_rate": 2.566142771445711e-06, "loss": 0.0006, "step": 6077 }, { "epoch": 0.73, "learning_rate": 2.5660707858428313e-06, "loss": 0.0009, "step": 6078 }, { "epoch": 0.73, "learning_rate": 2.565998800239952e-06, "loss": 0.001, "step": 6079 }, { "epoch": 0.73, "learning_rate": 2.565926814637073e-06, "loss": 0.002, "step": 6080 }, { "epoch": 0.73, "learning_rate": 2.5658548290341932e-06, "loss": 0.0011, "step": 6081 }, { "epoch": 0.73, "learning_rate": 2.565782843431314e-06, "loss": 0.004, "step": 6082 }, { "epoch": 0.73, "learning_rate": 2.5657108578284344e-06, "loss": 0.0312, "step": 6083 }, { "epoch": 0.73, "learning_rate": 2.565638872225555e-06, "loss": 0.0004, "step": 6084 }, { "epoch": 0.73, "learning_rate": 2.5655668866226755e-06, "loss": 0.001, "step": 6085 }, { "epoch": 0.73, "learning_rate": 2.5654949010197963e-06, "loss": 0.138, "step": 6086 }, { "epoch": 0.73, "learning_rate": 2.5654229154169166e-06, "loss": 0.0009, "step": 6087 }, { "epoch": 0.73, "learning_rate": 2.5653509298140374e-06, "loss": 0.9206, "step": 6088 }, { "epoch": 0.73, "learning_rate": 2.5652789442111578e-06, "loss": 0.0012, "step": 6089 }, { "epoch": 0.73, "learning_rate": 2.565206958608278e-06, "loss": 0.0005, "step": 6090 }, { "epoch": 0.73, "learning_rate": 2.565134973005399e-06, "loss": 0.002, "step": 6091 }, { "epoch": 0.73, "learning_rate": 2.5650629874025193e-06, "loss": 0.0008, "step": 6092 }, { "epoch": 0.73, "learning_rate": 2.56499100179964e-06, "loss": 0.0007, "step": 6093 }, { "epoch": 0.73, "learning_rate": 2.5649190161967604e-06, "loss": 0.0012, "step": 6094 }, { "epoch": 0.73, "learning_rate": 2.564847030593881e-06, "loss": 0.0025, "step": 6095 }, { "epoch": 0.73, "learning_rate": 2.564775044991002e-06, "loss": 0.0005, "step": 6096 }, { "epoch": 0.73, "learning_rate": 2.5647030593881227e-06, "loss": 0.0033, "step": 6097 }, { "epoch": 0.73, "learning_rate": 2.564631073785243e-06, "loss": 0.0004, "step": 6098 }, { "epoch": 0.73, "learning_rate": 2.564559088182364e-06, "loss": 0.0014, "step": 6099 }, { "epoch": 0.73, "learning_rate": 2.5644871025794842e-06, "loss": 0.002, "step": 6100 }, { "epoch": 0.73, "learning_rate": 2.5644151169766046e-06, "loss": 0.037, "step": 6101 }, { "epoch": 0.73, "learning_rate": 2.5643431313737254e-06, "loss": 0.1337, "step": 6102 }, { "epoch": 0.73, "learning_rate": 2.5642711457708457e-06, "loss": 0.0006, "step": 6103 }, { "epoch": 0.73, "learning_rate": 2.5641991601679665e-06, "loss": 0.0005, "step": 6104 }, { "epoch": 0.73, "learning_rate": 2.564127174565087e-06, "loss": 0.0005, "step": 6105 }, { "epoch": 0.73, "learning_rate": 2.5640551889622076e-06, "loss": 0.0009, "step": 6106 }, { "epoch": 0.73, "learning_rate": 2.563983203359328e-06, "loss": 0.0057, "step": 6107 }, { "epoch": 0.73, "learning_rate": 2.5639112177564488e-06, "loss": 0.7213, "step": 6108 }, { "epoch": 0.73, "learning_rate": 2.563839232153569e-06, "loss": 0.0012, "step": 6109 }, { "epoch": 0.73, "learning_rate": 2.56376724655069e-06, "loss": 0.0342, "step": 6110 }, { "epoch": 0.73, "learning_rate": 2.5636952609478107e-06, "loss": 0.0004, "step": 6111 }, { "epoch": 0.73, "learning_rate": 2.563623275344931e-06, "loss": 0.0006, "step": 6112 }, { "epoch": 0.73, "learning_rate": 2.563551289742052e-06, "loss": 0.0014, "step": 6113 }, { "epoch": 0.73, "learning_rate": 2.563479304139172e-06, "loss": 0.0007, "step": 6114 }, { "epoch": 0.73, "learning_rate": 2.563407318536293e-06, "loss": 0.0006, "step": 6115 }, { "epoch": 0.73, "learning_rate": 2.5633353329334133e-06, "loss": 0.0019, "step": 6116 }, { "epoch": 0.73, "learning_rate": 2.563263347330534e-06, "loss": 0.0008, "step": 6117 }, { "epoch": 0.73, "learning_rate": 2.5631913617276544e-06, "loss": 0.0007, "step": 6118 }, { "epoch": 0.73, "learning_rate": 2.563119376124775e-06, "loss": 0.0009, "step": 6119 }, { "epoch": 0.73, "learning_rate": 2.5630473905218956e-06, "loss": 0.004, "step": 6120 }, { "epoch": 0.73, "learning_rate": 2.5629754049190163e-06, "loss": 0.0018, "step": 6121 }, { "epoch": 0.73, "learning_rate": 2.5629034193161367e-06, "loss": 1.0302, "step": 6122 }, { "epoch": 0.73, "learning_rate": 2.5628314337132575e-06, "loss": 0.9351, "step": 6123 }, { "epoch": 0.73, "learning_rate": 2.562759448110378e-06, "loss": 0.0013, "step": 6124 }, { "epoch": 0.73, "learning_rate": 2.5626874625074986e-06, "loss": 0.001, "step": 6125 }, { "epoch": 0.73, "learning_rate": 2.562615476904619e-06, "loss": 0.0106, "step": 6126 }, { "epoch": 0.73, "learning_rate": 2.5625434913017397e-06, "loss": 0.0005, "step": 6127 }, { "epoch": 0.73, "learning_rate": 2.5624715056988605e-06, "loss": 0.0148, "step": 6128 }, { "epoch": 0.73, "learning_rate": 2.562399520095981e-06, "loss": 0.1816, "step": 6129 }, { "epoch": 0.73, "learning_rate": 2.5623275344931017e-06, "loss": 0.0005, "step": 6130 }, { "epoch": 0.73, "learning_rate": 2.562255548890222e-06, "loss": 0.0012, "step": 6131 }, { "epoch": 0.73, "learning_rate": 2.562183563287343e-06, "loss": 0.0016, "step": 6132 }, { "epoch": 0.73, "learning_rate": 2.562111577684463e-06, "loss": 0.0011, "step": 6133 }, { "epoch": 0.74, "learning_rate": 2.562039592081584e-06, "loss": 0.0031, "step": 6134 }, { "epoch": 0.74, "learning_rate": 2.5619676064787043e-06, "loss": 0.0012, "step": 6135 }, { "epoch": 0.74, "learning_rate": 2.5618956208758246e-06, "loss": 0.0008, "step": 6136 }, { "epoch": 0.74, "learning_rate": 2.5618236352729454e-06, "loss": 0.0007, "step": 6137 }, { "epoch": 0.74, "learning_rate": 2.5617516496700658e-06, "loss": 0.0006, "step": 6138 }, { "epoch": 0.74, "learning_rate": 2.5616796640671865e-06, "loss": 0.0008, "step": 6139 }, { "epoch": 0.74, "learning_rate": 2.561607678464307e-06, "loss": 0.0009, "step": 6140 }, { "epoch": 0.74, "learning_rate": 2.5615356928614277e-06, "loss": 0.0011, "step": 6141 }, { "epoch": 0.74, "learning_rate": 2.5614637072585485e-06, "loss": 0.0005, "step": 6142 }, { "epoch": 0.74, "learning_rate": 2.5613917216556692e-06, "loss": 0.0009, "step": 6143 }, { "epoch": 0.74, "learning_rate": 2.5613197360527896e-06, "loss": 0.0014, "step": 6144 }, { "epoch": 0.74, "learning_rate": 2.5612477504499104e-06, "loss": 0.0005, "step": 6145 }, { "epoch": 0.74, "learning_rate": 2.5611757648470307e-06, "loss": 0.0008, "step": 6146 }, { "epoch": 0.74, "learning_rate": 2.5611037792441515e-06, "loss": 0.0005, "step": 6147 }, { "epoch": 0.74, "learning_rate": 2.561031793641272e-06, "loss": 0.071, "step": 6148 }, { "epoch": 0.74, "learning_rate": 2.5609598080383922e-06, "loss": 0.0014, "step": 6149 }, { "epoch": 0.74, "learning_rate": 2.560887822435513e-06, "loss": 0.0139, "step": 6150 }, { "epoch": 0.74, "learning_rate": 2.5608158368326333e-06, "loss": 0.0006, "step": 6151 }, { "epoch": 0.74, "learning_rate": 2.560743851229754e-06, "loss": 0.0009, "step": 6152 }, { "epoch": 0.74, "learning_rate": 2.5606718656268745e-06, "loss": 0.0028, "step": 6153 }, { "epoch": 0.74, "learning_rate": 2.5605998800239953e-06, "loss": 0.0006, "step": 6154 }, { "epoch": 0.74, "learning_rate": 2.5605278944211156e-06, "loss": 0.8832, "step": 6155 }, { "epoch": 0.74, "learning_rate": 2.5604559088182364e-06, "loss": 0.0005, "step": 6156 }, { "epoch": 0.74, "learning_rate": 2.560383923215357e-06, "loss": 0.0137, "step": 6157 }, { "epoch": 0.74, "learning_rate": 2.5603119376124775e-06, "loss": 0.0006, "step": 6158 }, { "epoch": 0.74, "learning_rate": 2.5602399520095983e-06, "loss": 0.04, "step": 6159 }, { "epoch": 0.74, "learning_rate": 2.5601679664067187e-06, "loss": 0.0008, "step": 6160 }, { "epoch": 0.74, "learning_rate": 2.5600959808038394e-06, "loss": 0.0005, "step": 6161 }, { "epoch": 0.74, "learning_rate": 2.56002399520096e-06, "loss": 0.0012, "step": 6162 }, { "epoch": 0.74, "learning_rate": 2.5599520095980806e-06, "loss": 0.0009, "step": 6163 }, { "epoch": 0.74, "learning_rate": 2.559880023995201e-06, "loss": 0.0005, "step": 6164 }, { "epoch": 0.74, "learning_rate": 2.5598080383923217e-06, "loss": 0.0007, "step": 6165 }, { "epoch": 0.74, "learning_rate": 2.559736052789442e-06, "loss": 0.0008, "step": 6166 }, { "epoch": 0.74, "learning_rate": 2.559664067186563e-06, "loss": 0.0026, "step": 6167 }, { "epoch": 0.74, "learning_rate": 2.559592081583683e-06, "loss": 0.0009, "step": 6168 }, { "epoch": 0.74, "learning_rate": 2.559520095980804e-06, "loss": 0.0024, "step": 6169 }, { "epoch": 0.74, "learning_rate": 2.5594481103779243e-06, "loss": 0.0007, "step": 6170 }, { "epoch": 0.74, "learning_rate": 2.559376124775045e-06, "loss": 0.0007, "step": 6171 }, { "epoch": 0.74, "learning_rate": 2.5593041391721655e-06, "loss": 0.0011, "step": 6172 }, { "epoch": 0.74, "learning_rate": 2.5592321535692862e-06, "loss": 0.0008, "step": 6173 }, { "epoch": 0.74, "learning_rate": 2.559160167966407e-06, "loss": 0.0005, "step": 6174 }, { "epoch": 0.74, "learning_rate": 2.5590881823635274e-06, "loss": 0.0102, "step": 6175 }, { "epoch": 0.74, "learning_rate": 2.559016196760648e-06, "loss": 0.0008, "step": 6176 }, { "epoch": 0.74, "learning_rate": 2.5589442111577685e-06, "loss": 0.0007, "step": 6177 }, { "epoch": 0.74, "learning_rate": 2.5588722255548893e-06, "loss": 0.0018, "step": 6178 }, { "epoch": 0.74, "learning_rate": 2.5588002399520096e-06, "loss": 0.1133, "step": 6179 }, { "epoch": 0.74, "learning_rate": 2.5587282543491304e-06, "loss": 0.003, "step": 6180 }, { "epoch": 0.74, "learning_rate": 2.5586562687462508e-06, "loss": 0.0005, "step": 6181 }, { "epoch": 0.74, "learning_rate": 2.558584283143371e-06, "loss": 0.9475, "step": 6182 }, { "epoch": 0.74, "learning_rate": 2.558512297540492e-06, "loss": 0.0011, "step": 6183 }, { "epoch": 0.74, "learning_rate": 2.5584403119376123e-06, "loss": 0.8388, "step": 6184 }, { "epoch": 0.74, "learning_rate": 2.558368326334733e-06, "loss": 0.0006, "step": 6185 }, { "epoch": 0.74, "learning_rate": 2.5582963407318534e-06, "loss": 0.0006, "step": 6186 }, { "epoch": 0.74, "learning_rate": 2.558224355128974e-06, "loss": 0.0007, "step": 6187 }, { "epoch": 0.74, "learning_rate": 2.558152369526095e-06, "loss": 0.0009, "step": 6188 }, { "epoch": 0.74, "learning_rate": 2.5580803839232157e-06, "loss": 0.0006, "step": 6189 }, { "epoch": 0.74, "learning_rate": 2.558008398320336e-06, "loss": 0.0007, "step": 6190 }, { "epoch": 0.74, "learning_rate": 2.557936412717457e-06, "loss": 0.0018, "step": 6191 }, { "epoch": 0.74, "learning_rate": 2.5578644271145772e-06, "loss": 0.0004, "step": 6192 }, { "epoch": 0.74, "learning_rate": 2.557792441511698e-06, "loss": 0.0015, "step": 6193 }, { "epoch": 0.74, "learning_rate": 2.5577204559088184e-06, "loss": 0.0009, "step": 6194 }, { "epoch": 0.74, "learning_rate": 2.5576484703059387e-06, "loss": 0.0006, "step": 6195 }, { "epoch": 0.74, "learning_rate": 2.5575764847030595e-06, "loss": 0.0006, "step": 6196 }, { "epoch": 0.74, "learning_rate": 2.55750449910018e-06, "loss": 0.0008, "step": 6197 }, { "epoch": 0.74, "learning_rate": 2.5574325134973006e-06, "loss": 0.0009, "step": 6198 }, { "epoch": 0.74, "learning_rate": 2.557360527894421e-06, "loss": 0.0006, "step": 6199 }, { "epoch": 0.74, "learning_rate": 2.5572885422915418e-06, "loss": 0.0008, "step": 6200 }, { "epoch": 0.74, "learning_rate": 2.557216556688662e-06, "loss": 0.0457, "step": 6201 }, { "epoch": 0.74, "learning_rate": 2.557144571085783e-06, "loss": 0.0011, "step": 6202 }, { "epoch": 0.74, "learning_rate": 2.5570725854829037e-06, "loss": 0.0005, "step": 6203 }, { "epoch": 0.74, "learning_rate": 2.557000599880024e-06, "loss": 0.0008, "step": 6204 }, { "epoch": 0.74, "learning_rate": 2.556928614277145e-06, "loss": 0.0007, "step": 6205 }, { "epoch": 0.74, "learning_rate": 2.556856628674265e-06, "loss": 0.0008, "step": 6206 }, { "epoch": 0.74, "learning_rate": 2.556784643071386e-06, "loss": 0.0008, "step": 6207 }, { "epoch": 0.74, "learning_rate": 2.5567126574685063e-06, "loss": 0.0006, "step": 6208 }, { "epoch": 0.74, "learning_rate": 2.556640671865627e-06, "loss": 0.0013, "step": 6209 }, { "epoch": 0.74, "learning_rate": 2.5565686862627474e-06, "loss": 0.0005, "step": 6210 }, { "epoch": 0.74, "learning_rate": 2.556496700659868e-06, "loss": 0.0006, "step": 6211 }, { "epoch": 0.74, "learning_rate": 2.5564247150569886e-06, "loss": 0.0006, "step": 6212 }, { "epoch": 0.74, "learning_rate": 2.5563527294541093e-06, "loss": 0.0291, "step": 6213 }, { "epoch": 0.74, "learning_rate": 2.5562807438512297e-06, "loss": 0.0006, "step": 6214 }, { "epoch": 0.74, "learning_rate": 2.5562087582483505e-06, "loss": 0.0006, "step": 6215 }, { "epoch": 0.74, "learning_rate": 2.556136772645471e-06, "loss": 1.7599, "step": 6216 }, { "epoch": 0.74, "learning_rate": 2.556064787042591e-06, "loss": 0.001, "step": 6217 }, { "epoch": 0.75, "learning_rate": 2.555992801439712e-06, "loss": 0.0007, "step": 6218 }, { "epoch": 0.75, "learning_rate": 2.5559208158368327e-06, "loss": 0.0007, "step": 6219 }, { "epoch": 0.75, "learning_rate": 2.5558488302339535e-06, "loss": 0.0004, "step": 6220 }, { "epoch": 0.75, "learning_rate": 2.555776844631074e-06, "loss": 0.0008, "step": 6221 }, { "epoch": 0.75, "learning_rate": 2.5557048590281946e-06, "loss": 0.891, "step": 6222 }, { "epoch": 0.75, "learning_rate": 2.555632873425315e-06, "loss": 0.0008, "step": 6223 }, { "epoch": 0.75, "learning_rate": 2.5555608878224358e-06, "loss": 0.0027, "step": 6224 }, { "epoch": 0.75, "learning_rate": 2.555488902219556e-06, "loss": 0.0005, "step": 6225 }, { "epoch": 0.75, "learning_rate": 2.555416916616677e-06, "loss": 0.0008, "step": 6226 }, { "epoch": 0.75, "learning_rate": 2.5553449310137973e-06, "loss": 0.0008, "step": 6227 }, { "epoch": 0.75, "learning_rate": 2.5552729454109176e-06, "loss": 0.0006, "step": 6228 }, { "epoch": 0.75, "learning_rate": 2.5552009598080384e-06, "loss": 0.0005, "step": 6229 }, { "epoch": 0.75, "learning_rate": 2.5551289742051588e-06, "loss": 0.0014, "step": 6230 }, { "epoch": 0.75, "learning_rate": 2.5550569886022795e-06, "loss": 0.0006, "step": 6231 }, { "epoch": 0.75, "learning_rate": 2.5549850029994e-06, "loss": 0.0007, "step": 6232 }, { "epoch": 0.75, "learning_rate": 2.5549130173965207e-06, "loss": 0.0012, "step": 6233 }, { "epoch": 0.75, "learning_rate": 2.5548410317936415e-06, "loss": 0.0176, "step": 6234 }, { "epoch": 0.75, "learning_rate": 2.554769046190762e-06, "loss": 0.0005, "step": 6235 }, { "epoch": 0.75, "learning_rate": 2.5546970605878826e-06, "loss": 0.5509, "step": 6236 }, { "epoch": 0.75, "learning_rate": 2.5546250749850034e-06, "loss": 0.0006, "step": 6237 }, { "epoch": 0.75, "learning_rate": 2.5545530893821237e-06, "loss": 0.0009, "step": 6238 }, { "epoch": 0.75, "learning_rate": 2.5544811037792445e-06, "loss": 0.9246, "step": 6239 }, { "epoch": 0.75, "learning_rate": 2.554409118176365e-06, "loss": 0.0009, "step": 6240 }, { "epoch": 0.75, "learning_rate": 2.554337132573485e-06, "loss": 0.0007, "step": 6241 }, { "epoch": 0.75, "learning_rate": 2.554265146970606e-06, "loss": 0.0005, "step": 6242 }, { "epoch": 0.75, "learning_rate": 2.5541931613677263e-06, "loss": 0.0009, "step": 6243 }, { "epoch": 0.75, "learning_rate": 2.554121175764847e-06, "loss": 0.0005, "step": 6244 }, { "epoch": 0.75, "learning_rate": 2.5540491901619675e-06, "loss": 0.002, "step": 6245 }, { "epoch": 0.75, "learning_rate": 2.5539772045590883e-06, "loss": 0.0005, "step": 6246 }, { "epoch": 0.75, "learning_rate": 2.5539052189562086e-06, "loss": 0.0006, "step": 6247 }, { "epoch": 0.75, "learning_rate": 2.5538332333533294e-06, "loss": 0.0012, "step": 6248 }, { "epoch": 0.75, "learning_rate": 2.5537612477504497e-06, "loss": 0.0061, "step": 6249 }, { "epoch": 0.75, "learning_rate": 2.5536892621475705e-06, "loss": 0.0096, "step": 6250 }, { "epoch": 0.75, "learning_rate": 2.5536172765446913e-06, "loss": 0.0007, "step": 6251 }, { "epoch": 0.75, "learning_rate": 2.5535452909418117e-06, "loss": 0.2042, "step": 6252 }, { "epoch": 0.75, "learning_rate": 2.5534733053389324e-06, "loss": 0.0005, "step": 6253 }, { "epoch": 0.75, "learning_rate": 2.5534013197360528e-06, "loss": 0.0027, "step": 6254 }, { "epoch": 0.75, "learning_rate": 2.5533293341331736e-06, "loss": 0.7634, "step": 6255 }, { "epoch": 0.75, "learning_rate": 2.553257348530294e-06, "loss": 0.0006, "step": 6256 }, { "epoch": 0.75, "learning_rate": 2.5531853629274147e-06, "loss": 0.0064, "step": 6257 }, { "epoch": 0.75, "learning_rate": 2.553113377324535e-06, "loss": 0.9297, "step": 6258 }, { "epoch": 0.75, "learning_rate": 2.553041391721656e-06, "loss": 0.0005, "step": 6259 }, { "epoch": 0.75, "learning_rate": 2.552969406118776e-06, "loss": 0.0007, "step": 6260 }, { "epoch": 0.75, "learning_rate": 2.552897420515897e-06, "loss": 0.0005, "step": 6261 }, { "epoch": 0.75, "learning_rate": 2.5528254349130173e-06, "loss": 0.0004, "step": 6262 }, { "epoch": 0.75, "learning_rate": 2.5527534493101377e-06, "loss": 0.0005, "step": 6263 }, { "epoch": 0.75, "learning_rate": 2.5526814637072585e-06, "loss": 0.1018, "step": 6264 }, { "epoch": 0.75, "learning_rate": 2.5526094781043792e-06, "loss": 0.0006, "step": 6265 }, { "epoch": 0.75, "learning_rate": 2.5525374925015e-06, "loss": 0.0005, "step": 6266 }, { "epoch": 0.75, "learning_rate": 2.5524655068986204e-06, "loss": 0.0005, "step": 6267 }, { "epoch": 0.75, "learning_rate": 2.552393521295741e-06, "loss": 0.0006, "step": 6268 }, { "epoch": 0.75, "learning_rate": 2.5523215356928615e-06, "loss": 0.5685, "step": 6269 }, { "epoch": 0.75, "learning_rate": 2.5522495500899823e-06, "loss": 0.0007, "step": 6270 }, { "epoch": 0.75, "learning_rate": 2.5521775644871026e-06, "loss": 0.0005, "step": 6271 }, { "epoch": 0.75, "learning_rate": 2.5521055788842234e-06, "loss": 0.0008, "step": 6272 }, { "epoch": 0.75, "learning_rate": 2.5520335932813438e-06, "loss": 0.0052, "step": 6273 }, { "epoch": 0.75, "learning_rate": 2.5519616076784645e-06, "loss": 0.0006, "step": 6274 }, { "epoch": 0.75, "learning_rate": 2.551889622075585e-06, "loss": 0.7386, "step": 6275 }, { "epoch": 0.75, "learning_rate": 2.5518176364727053e-06, "loss": 0.0022, "step": 6276 }, { "epoch": 0.75, "learning_rate": 2.551745650869826e-06, "loss": 0.5248, "step": 6277 }, { "epoch": 0.75, "learning_rate": 2.5516736652669464e-06, "loss": 0.0005, "step": 6278 }, { "epoch": 0.75, "learning_rate": 2.551601679664067e-06, "loss": 0.0009, "step": 6279 }, { "epoch": 0.75, "learning_rate": 2.551529694061188e-06, "loss": 0.0009, "step": 6280 }, { "epoch": 0.75, "learning_rate": 2.5514577084583083e-06, "loss": 0.0004, "step": 6281 }, { "epoch": 0.75, "learning_rate": 2.551385722855429e-06, "loss": 0.0006, "step": 6282 }, { "epoch": 0.75, "learning_rate": 2.55131373725255e-06, "loss": 0.3065, "step": 6283 }, { "epoch": 0.75, "learning_rate": 2.5512417516496702e-06, "loss": 0.0004, "step": 6284 }, { "epoch": 0.75, "learning_rate": 2.551169766046791e-06, "loss": 0.0006, "step": 6285 }, { "epoch": 0.75, "learning_rate": 2.5510977804439113e-06, "loss": 0.0011, "step": 6286 }, { "epoch": 0.75, "learning_rate": 2.5510257948410317e-06, "loss": 0.0054, "step": 6287 }, { "epoch": 0.75, "learning_rate": 2.5509538092381525e-06, "loss": 0.0006, "step": 6288 }, { "epoch": 0.75, "learning_rate": 2.550881823635273e-06, "loss": 0.0007, "step": 6289 }, { "epoch": 0.75, "learning_rate": 2.5508098380323936e-06, "loss": 0.0009, "step": 6290 }, { "epoch": 0.75, "learning_rate": 2.550737852429514e-06, "loss": 0.001, "step": 6291 }, { "epoch": 0.75, "learning_rate": 2.5506658668266347e-06, "loss": 0.0113, "step": 6292 }, { "epoch": 0.75, "learning_rate": 2.550593881223755e-06, "loss": 0.0133, "step": 6293 }, { "epoch": 0.75, "learning_rate": 2.550521895620876e-06, "loss": 0.0005, "step": 6294 }, { "epoch": 0.75, "learning_rate": 2.5504499100179962e-06, "loss": 0.0008, "step": 6295 }, { "epoch": 0.75, "learning_rate": 2.550377924415117e-06, "loss": 0.0006, "step": 6296 }, { "epoch": 0.75, "learning_rate": 2.550305938812238e-06, "loss": 0.0007, "step": 6297 }, { "epoch": 0.75, "learning_rate": 2.550233953209358e-06, "loss": 0.5945, "step": 6298 }, { "epoch": 0.75, "learning_rate": 2.550161967606479e-06, "loss": 0.0013, "step": 6299 }, { "epoch": 0.75, "learning_rate": 2.5500899820035993e-06, "loss": 0.0027, "step": 6300 }, { "epoch": 0.76, "learning_rate": 2.55001799640072e-06, "loss": 0.001, "step": 6301 }, { "epoch": 0.76, "learning_rate": 2.5499460107978404e-06, "loss": 0.6751, "step": 6302 }, { "epoch": 0.76, "learning_rate": 2.549874025194961e-06, "loss": 0.0008, "step": 6303 }, { "epoch": 0.76, "learning_rate": 2.5498020395920815e-06, "loss": 0.0007, "step": 6304 }, { "epoch": 0.76, "learning_rate": 2.5497300539892023e-06, "loss": 0.0007, "step": 6305 }, { "epoch": 0.76, "learning_rate": 2.5496580683863227e-06, "loss": 0.5256, "step": 6306 }, { "epoch": 0.76, "learning_rate": 2.5495860827834435e-06, "loss": 0.0084, "step": 6307 }, { "epoch": 0.76, "learning_rate": 2.549514097180564e-06, "loss": 0.0005, "step": 6308 }, { "epoch": 0.76, "learning_rate": 2.549442111577684e-06, "loss": 0.0005, "step": 6309 }, { "epoch": 0.76, "learning_rate": 2.549370125974805e-06, "loss": 0.0006, "step": 6310 }, { "epoch": 0.76, "learning_rate": 2.5492981403719257e-06, "loss": 0.862, "step": 6311 }, { "epoch": 0.76, "learning_rate": 2.5492261547690465e-06, "loss": 0.0005, "step": 6312 }, { "epoch": 0.76, "learning_rate": 2.549154169166167e-06, "loss": 0.0005, "step": 6313 }, { "epoch": 0.76, "learning_rate": 2.5490821835632876e-06, "loss": 0.7556, "step": 6314 }, { "epoch": 0.76, "learning_rate": 2.549010197960408e-06, "loss": 1.1578, "step": 6315 }, { "epoch": 0.76, "learning_rate": 2.5489382123575288e-06, "loss": 0.002, "step": 6316 }, { "epoch": 0.76, "learning_rate": 2.548866226754649e-06, "loss": 0.0007, "step": 6317 }, { "epoch": 0.76, "learning_rate": 2.54879424115177e-06, "loss": 0.0005, "step": 6318 }, { "epoch": 0.76, "learning_rate": 2.5487222555488903e-06, "loss": 0.8966, "step": 6319 }, { "epoch": 0.76, "learning_rate": 2.548650269946011e-06, "loss": 0.0011, "step": 6320 }, { "epoch": 0.76, "learning_rate": 2.5485782843431314e-06, "loss": 0.0005, "step": 6321 }, { "epoch": 0.76, "learning_rate": 2.5485062987402518e-06, "loss": 0.0009, "step": 6322 }, { "epoch": 0.76, "learning_rate": 2.5484343131373725e-06, "loss": 0.0006, "step": 6323 }, { "epoch": 0.76, "learning_rate": 2.548362327534493e-06, "loss": 0.0164, "step": 6324 }, { "epoch": 0.76, "learning_rate": 2.5482903419316137e-06, "loss": 0.0008, "step": 6325 }, { "epoch": 0.76, "learning_rate": 2.5482183563287344e-06, "loss": 0.8068, "step": 6326 }, { "epoch": 0.76, "learning_rate": 2.548146370725855e-06, "loss": 0.3484, "step": 6327 }, { "epoch": 0.76, "learning_rate": 2.5480743851229756e-06, "loss": 0.0006, "step": 6328 }, { "epoch": 0.76, "learning_rate": 2.5480023995200964e-06, "loss": 0.071, "step": 6329 }, { "epoch": 0.76, "learning_rate": 2.5479304139172167e-06, "loss": 0.0008, "step": 6330 }, { "epoch": 0.76, "learning_rate": 2.5478584283143375e-06, "loss": 0.0006, "step": 6331 }, { "epoch": 0.76, "learning_rate": 2.547786442711458e-06, "loss": 0.0008, "step": 6332 }, { "epoch": 0.76, "learning_rate": 2.547714457108578e-06, "loss": 0.0009, "step": 6333 }, { "epoch": 0.76, "learning_rate": 2.547642471505699e-06, "loss": 0.0008, "step": 6334 }, { "epoch": 0.76, "learning_rate": 2.5475704859028193e-06, "loss": 0.0137, "step": 6335 }, { "epoch": 0.76, "learning_rate": 2.54749850029994e-06, "loss": 0.6139, "step": 6336 }, { "epoch": 0.76, "learning_rate": 2.5474265146970605e-06, "loss": 0.0007, "step": 6337 }, { "epoch": 0.76, "learning_rate": 2.5473545290941812e-06, "loss": 0.0005, "step": 6338 }, { "epoch": 0.76, "learning_rate": 2.5472825434913016e-06, "loss": 0.0007, "step": 6339 }, { "epoch": 0.76, "learning_rate": 2.5472105578884224e-06, "loss": 0.0018, "step": 6340 }, { "epoch": 0.76, "learning_rate": 2.5471385722855427e-06, "loss": 0.0006, "step": 6341 }, { "epoch": 0.76, "learning_rate": 2.5470665866826635e-06, "loss": 0.0037, "step": 6342 }, { "epoch": 0.76, "learning_rate": 2.5469946010797843e-06, "loss": 0.002, "step": 6343 }, { "epoch": 0.76, "learning_rate": 2.5469226154769046e-06, "loss": 0.0036, "step": 6344 }, { "epoch": 0.76, "learning_rate": 2.5468506298740254e-06, "loss": 0.0008, "step": 6345 }, { "epoch": 0.76, "learning_rate": 2.5467786442711458e-06, "loss": 0.0006, "step": 6346 }, { "epoch": 0.76, "learning_rate": 2.5467066586682666e-06, "loss": 0.6602, "step": 6347 }, { "epoch": 0.76, "learning_rate": 2.546634673065387e-06, "loss": 0.0017, "step": 6348 }, { "epoch": 0.76, "learning_rate": 2.5465626874625077e-06, "loss": 0.0008, "step": 6349 }, { "epoch": 0.76, "learning_rate": 2.546490701859628e-06, "loss": 0.0007, "step": 6350 }, { "epoch": 0.76, "learning_rate": 2.546418716256749e-06, "loss": 0.6429, "step": 6351 }, { "epoch": 0.76, "learning_rate": 2.546346730653869e-06, "loss": 0.0004, "step": 6352 }, { "epoch": 0.76, "learning_rate": 2.54627474505099e-06, "loss": 0.0007, "step": 6353 }, { "epoch": 0.76, "learning_rate": 2.5462027594481103e-06, "loss": 0.2383, "step": 6354 }, { "epoch": 0.76, "learning_rate": 2.5461307738452307e-06, "loss": 0.0008, "step": 6355 }, { "epoch": 0.76, "learning_rate": 2.5460587882423514e-06, "loss": 0.0021, "step": 6356 }, { "epoch": 0.76, "learning_rate": 2.5459868026394722e-06, "loss": 0.0836, "step": 6357 }, { "epoch": 0.76, "learning_rate": 2.5459148170365926e-06, "loss": 0.0005, "step": 6358 }, { "epoch": 0.76, "learning_rate": 2.5458428314337134e-06, "loss": 0.0005, "step": 6359 }, { "epoch": 0.76, "learning_rate": 2.545770845830834e-06, "loss": 0.0033, "step": 6360 }, { "epoch": 0.76, "learning_rate": 2.5456988602279545e-06, "loss": 0.0009, "step": 6361 }, { "epoch": 0.76, "learning_rate": 2.5456268746250753e-06, "loss": 0.0008, "step": 6362 }, { "epoch": 0.76, "learning_rate": 2.5455548890221956e-06, "loss": 0.0005, "step": 6363 }, { "epoch": 0.76, "learning_rate": 2.5454829034193164e-06, "loss": 0.0038, "step": 6364 }, { "epoch": 0.76, "learning_rate": 2.5454109178164368e-06, "loss": 0.0009, "step": 6365 }, { "epoch": 0.76, "learning_rate": 2.5453389322135575e-06, "loss": 0.0004, "step": 6366 }, { "epoch": 0.76, "learning_rate": 2.545266946610678e-06, "loss": 0.0008, "step": 6367 }, { "epoch": 0.76, "learning_rate": 2.5451949610077982e-06, "loss": 0.0004, "step": 6368 }, { "epoch": 0.76, "learning_rate": 2.545122975404919e-06, "loss": 0.0005, "step": 6369 }, { "epoch": 0.76, "learning_rate": 2.5450509898020394e-06, "loss": 0.0145, "step": 6370 }, { "epoch": 0.76, "learning_rate": 2.54497900419916e-06, "loss": 0.0004, "step": 6371 }, { "epoch": 0.76, "learning_rate": 2.5449070185962805e-06, "loss": 0.0004, "step": 6372 }, { "epoch": 0.76, "learning_rate": 2.5448350329934013e-06, "loss": 0.0005, "step": 6373 }, { "epoch": 0.76, "learning_rate": 2.544763047390522e-06, "loss": 0.0039, "step": 6374 }, { "epoch": 0.76, "learning_rate": 2.544691061787643e-06, "loss": 0.5085, "step": 6375 }, { "epoch": 0.76, "learning_rate": 2.544619076184763e-06, "loss": 0.0012, "step": 6376 }, { "epoch": 0.76, "learning_rate": 2.544547090581884e-06, "loss": 0.0004, "step": 6377 }, { "epoch": 0.76, "learning_rate": 2.5444751049790043e-06, "loss": 0.0019, "step": 6378 }, { "epoch": 0.76, "learning_rate": 2.5444031193761247e-06, "loss": 0.0012, "step": 6379 }, { "epoch": 0.76, "learning_rate": 2.5443311337732455e-06, "loss": 0.0003, "step": 6380 }, { "epoch": 0.76, "learning_rate": 2.544259148170366e-06, "loss": 0.0014, "step": 6381 }, { "epoch": 0.76, "learning_rate": 2.5441871625674866e-06, "loss": 0.3693, "step": 6382 }, { "epoch": 0.76, "learning_rate": 2.544115176964607e-06, "loss": 0.0006, "step": 6383 }, { "epoch": 0.77, "learning_rate": 2.5440431913617277e-06, "loss": 1.0038, "step": 6384 }, { "epoch": 0.77, "learning_rate": 2.543971205758848e-06, "loss": 0.0006, "step": 6385 }, { "epoch": 0.77, "learning_rate": 2.543899220155969e-06, "loss": 0.0005, "step": 6386 }, { "epoch": 0.77, "learning_rate": 2.5438272345530892e-06, "loss": 0.003, "step": 6387 }, { "epoch": 0.77, "learning_rate": 2.54375524895021e-06, "loss": 0.8167, "step": 6388 }, { "epoch": 0.77, "learning_rate": 2.5436832633473308e-06, "loss": 0.0005, "step": 6389 }, { "epoch": 0.77, "learning_rate": 2.543611277744451e-06, "loss": 0.0008, "step": 6390 }, { "epoch": 0.77, "learning_rate": 2.543539292141572e-06, "loss": 0.0006, "step": 6391 }, { "epoch": 0.77, "learning_rate": 2.5434673065386923e-06, "loss": 0.0006, "step": 6392 }, { "epoch": 0.77, "learning_rate": 2.543395320935813e-06, "loss": 0.0006, "step": 6393 }, { "epoch": 0.77, "learning_rate": 2.5433233353329334e-06, "loss": 0.0004, "step": 6394 }, { "epoch": 0.77, "learning_rate": 2.543251349730054e-06, "loss": 0.0012, "step": 6395 }, { "epoch": 0.77, "learning_rate": 2.5431793641271745e-06, "loss": 0.0008, "step": 6396 }, { "epoch": 0.77, "learning_rate": 2.5431073785242953e-06, "loss": 0.0005, "step": 6397 }, { "epoch": 0.77, "learning_rate": 2.5430353929214157e-06, "loss": 0.0006, "step": 6398 }, { "epoch": 0.77, "learning_rate": 2.5429634073185365e-06, "loss": 0.0013, "step": 6399 }, { "epoch": 0.77, "learning_rate": 2.542891421715657e-06, "loss": 0.9587, "step": 6400 }, { "epoch": 0.77, "learning_rate": 2.5428194361127776e-06, "loss": 0.0007, "step": 6401 }, { "epoch": 0.77, "learning_rate": 2.542747450509898e-06, "loss": 0.4785, "step": 6402 }, { "epoch": 0.77, "learning_rate": 2.5426754649070187e-06, "loss": 0.0006, "step": 6403 }, { "epoch": 0.77, "learning_rate": 2.542603479304139e-06, "loss": 0.0011, "step": 6404 }, { "epoch": 0.77, "learning_rate": 2.54253149370126e-06, "loss": 0.3954, "step": 6405 }, { "epoch": 0.77, "learning_rate": 2.5424595080983806e-06, "loss": 0.8613, "step": 6406 }, { "epoch": 0.77, "learning_rate": 2.542387522495501e-06, "loss": 0.0004, "step": 6407 }, { "epoch": 0.77, "learning_rate": 2.5423155368926218e-06, "loss": 0.0009, "step": 6408 }, { "epoch": 0.77, "learning_rate": 2.542243551289742e-06, "loss": 0.0008, "step": 6409 }, { "epoch": 0.77, "learning_rate": 2.542171565686863e-06, "loss": 0.0005, "step": 6410 }, { "epoch": 0.77, "learning_rate": 2.5420995800839833e-06, "loss": 0.0006, "step": 6411 }, { "epoch": 0.77, "learning_rate": 2.542027594481104e-06, "loss": 0.0034, "step": 6412 }, { "epoch": 0.77, "learning_rate": 2.5419556088782244e-06, "loss": 0.0005, "step": 6413 }, { "epoch": 0.77, "learning_rate": 2.5418836232753447e-06, "loss": 0.0065, "step": 6414 }, { "epoch": 0.77, "learning_rate": 2.5418116376724655e-06, "loss": 0.001, "step": 6415 }, { "epoch": 0.77, "learning_rate": 2.541739652069586e-06, "loss": 0.0006, "step": 6416 }, { "epoch": 0.77, "learning_rate": 2.5416676664667067e-06, "loss": 0.0005, "step": 6417 }, { "epoch": 0.77, "learning_rate": 2.541595680863827e-06, "loss": 0.0006, "step": 6418 }, { "epoch": 0.77, "learning_rate": 2.541523695260948e-06, "loss": 0.0005, "step": 6419 }, { "epoch": 0.77, "learning_rate": 2.5414517096580686e-06, "loss": 0.5703, "step": 6420 }, { "epoch": 0.77, "learning_rate": 2.5413797240551893e-06, "loss": 0.5552, "step": 6421 }, { "epoch": 0.77, "learning_rate": 2.5413077384523097e-06, "loss": 0.0005, "step": 6422 }, { "epoch": 0.77, "learning_rate": 2.5412357528494305e-06, "loss": 0.0009, "step": 6423 }, { "epoch": 0.77, "learning_rate": 2.541163767246551e-06, "loss": 0.0009, "step": 6424 }, { "epoch": 0.77, "learning_rate": 2.541091781643671e-06, "loss": 0.001, "step": 6425 }, { "epoch": 0.77, "learning_rate": 2.541019796040792e-06, "loss": 0.0012, "step": 6426 }, { "epoch": 0.77, "learning_rate": 2.5409478104379123e-06, "loss": 0.0007, "step": 6427 }, { "epoch": 0.77, "learning_rate": 2.540875824835033e-06, "loss": 0.0009, "step": 6428 }, { "epoch": 0.77, "learning_rate": 2.5408038392321535e-06, "loss": 0.0008, "step": 6429 }, { "epoch": 0.77, "learning_rate": 2.5407318536292742e-06, "loss": 0.6125, "step": 6430 }, { "epoch": 0.77, "learning_rate": 2.5406598680263946e-06, "loss": 0.0009, "step": 6431 }, { "epoch": 0.77, "learning_rate": 2.5405878824235154e-06, "loss": 0.376, "step": 6432 }, { "epoch": 0.77, "learning_rate": 2.5405158968206357e-06, "loss": 0.0012, "step": 6433 }, { "epoch": 0.77, "learning_rate": 2.5404439112177565e-06, "loss": 0.0054, "step": 6434 }, { "epoch": 0.77, "learning_rate": 2.5403719256148773e-06, "loss": 0.0026, "step": 6435 }, { "epoch": 0.77, "learning_rate": 2.5402999400119976e-06, "loss": 0.8409, "step": 6436 }, { "epoch": 0.77, "learning_rate": 2.5402279544091184e-06, "loss": 0.0008, "step": 6437 }, { "epoch": 0.77, "learning_rate": 2.5401559688062388e-06, "loss": 0.0005, "step": 6438 }, { "epoch": 0.77, "learning_rate": 2.5400839832033595e-06, "loss": 0.0006, "step": 6439 }, { "epoch": 0.77, "learning_rate": 2.54001199760048e-06, "loss": 0.0005, "step": 6440 }, { "epoch": 0.77, "learning_rate": 2.5399400119976007e-06, "loss": 0.0005, "step": 6441 }, { "epoch": 0.77, "learning_rate": 2.539868026394721e-06, "loss": 0.9737, "step": 6442 }, { "epoch": 0.77, "learning_rate": 2.539796040791842e-06, "loss": 0.0007, "step": 6443 }, { "epoch": 0.77, "learning_rate": 2.539724055188962e-06, "loss": 0.8614, "step": 6444 }, { "epoch": 0.77, "learning_rate": 2.539652069586083e-06, "loss": 0.0009, "step": 6445 }, { "epoch": 0.77, "learning_rate": 2.5395800839832033e-06, "loss": 0.0009, "step": 6446 }, { "epoch": 0.77, "learning_rate": 2.539508098380324e-06, "loss": 0.2374, "step": 6447 }, { "epoch": 0.77, "learning_rate": 2.5394361127774444e-06, "loss": 0.0014, "step": 6448 }, { "epoch": 0.77, "learning_rate": 2.5393641271745652e-06, "loss": 0.0003, "step": 6449 }, { "epoch": 0.77, "learning_rate": 2.5392921415716856e-06, "loss": 0.0017, "step": 6450 }, { "epoch": 0.77, "learning_rate": 2.5392201559688064e-06, "loss": 0.0011, "step": 6451 }, { "epoch": 0.77, "learning_rate": 2.539148170365927e-06, "loss": 0.0011, "step": 6452 }, { "epoch": 0.77, "learning_rate": 2.5390761847630475e-06, "loss": 0.4972, "step": 6453 }, { "epoch": 0.77, "learning_rate": 2.5390041991601683e-06, "loss": 0.0018, "step": 6454 }, { "epoch": 0.77, "learning_rate": 2.5389322135572886e-06, "loss": 0.0435, "step": 6455 }, { "epoch": 0.77, "learning_rate": 2.5388602279544094e-06, "loss": 0.0011, "step": 6456 }, { "epoch": 0.77, "learning_rate": 2.5387882423515298e-06, "loss": 0.0008, "step": 6457 }, { "epoch": 0.77, "learning_rate": 2.5387162567486505e-06, "loss": 0.001, "step": 6458 }, { "epoch": 0.77, "learning_rate": 2.538644271145771e-06, "loss": 0.0014, "step": 6459 }, { "epoch": 0.77, "learning_rate": 2.5385722855428912e-06, "loss": 0.0008, "step": 6460 }, { "epoch": 0.77, "learning_rate": 2.538500299940012e-06, "loss": 0.0012, "step": 6461 }, { "epoch": 0.77, "learning_rate": 2.5384283143371324e-06, "loss": 0.0025, "step": 6462 }, { "epoch": 0.77, "learning_rate": 2.538356328734253e-06, "loss": 0.0006, "step": 6463 }, { "epoch": 0.77, "learning_rate": 2.5382843431313735e-06, "loss": 0.0005, "step": 6464 }, { "epoch": 0.77, "learning_rate": 2.5382123575284943e-06, "loss": 0.002, "step": 6465 }, { "epoch": 0.77, "learning_rate": 2.538140371925615e-06, "loss": 0.001, "step": 6466 }, { "epoch": 0.77, "learning_rate": 2.538068386322736e-06, "loss": 0.0287, "step": 6467 }, { "epoch": 0.78, "learning_rate": 2.537996400719856e-06, "loss": 0.002, "step": 6468 }, { "epoch": 0.78, "learning_rate": 2.537924415116977e-06, "loss": 0.0013, "step": 6469 }, { "epoch": 0.78, "learning_rate": 2.5378524295140973e-06, "loss": 0.0015, "step": 6470 }, { "epoch": 0.78, "learning_rate": 2.5377804439112177e-06, "loss": 0.0017, "step": 6471 }, { "epoch": 0.78, "learning_rate": 2.5377084583083385e-06, "loss": 0.0011, "step": 6472 }, { "epoch": 0.78, "learning_rate": 2.537636472705459e-06, "loss": 0.0131, "step": 6473 }, { "epoch": 0.78, "learning_rate": 2.5375644871025796e-06, "loss": 0.0015, "step": 6474 }, { "epoch": 0.78, "learning_rate": 2.5374925014997e-06, "loss": 0.0011, "step": 6475 }, { "epoch": 0.78, "learning_rate": 2.5374205158968207e-06, "loss": 0.0008, "step": 6476 }, { "epoch": 0.78, "learning_rate": 2.537348530293941e-06, "loss": 0.0017, "step": 6477 }, { "epoch": 0.78, "learning_rate": 2.537276544691062e-06, "loss": 0.7008, "step": 6478 }, { "epoch": 0.78, "learning_rate": 2.5372045590881822e-06, "loss": 0.0017, "step": 6479 }, { "epoch": 0.78, "learning_rate": 2.537132573485303e-06, "loss": 0.0011, "step": 6480 }, { "epoch": 0.78, "learning_rate": 2.5370605878824238e-06, "loss": 0.0013, "step": 6481 }, { "epoch": 0.78, "learning_rate": 2.536988602279544e-06, "loss": 0.0007, "step": 6482 }, { "epoch": 0.78, "learning_rate": 2.536916616676665e-06, "loss": 0.7517, "step": 6483 }, { "epoch": 0.78, "learning_rate": 2.5368446310737853e-06, "loss": 0.0015, "step": 6484 }, { "epoch": 0.78, "learning_rate": 2.536772645470906e-06, "loss": 0.0013, "step": 6485 }, { "epoch": 0.78, "learning_rate": 2.5367006598680264e-06, "loss": 0.001, "step": 6486 }, { "epoch": 0.78, "learning_rate": 2.536628674265147e-06, "loss": 0.0033, "step": 6487 }, { "epoch": 0.78, "learning_rate": 2.5365566886622675e-06, "loss": 0.0009, "step": 6488 }, { "epoch": 0.78, "learning_rate": 2.5364847030593883e-06, "loss": 0.0019, "step": 6489 }, { "epoch": 0.78, "learning_rate": 2.5364127174565087e-06, "loss": 0.6724, "step": 6490 }, { "epoch": 0.78, "learning_rate": 2.5363407318536294e-06, "loss": 0.0007, "step": 6491 }, { "epoch": 0.78, "learning_rate": 2.53626874625075e-06, "loss": 0.0009, "step": 6492 }, { "epoch": 0.78, "learning_rate": 2.5361967606478706e-06, "loss": 0.0014, "step": 6493 }, { "epoch": 0.78, "learning_rate": 2.536124775044991e-06, "loss": 0.0007, "step": 6494 }, { "epoch": 0.78, "learning_rate": 2.5360527894421113e-06, "loss": 0.0009, "step": 6495 }, { "epoch": 0.78, "learning_rate": 2.535980803839232e-06, "loss": 0.0008, "step": 6496 }, { "epoch": 0.78, "learning_rate": 2.535908818236353e-06, "loss": 0.2041, "step": 6497 }, { "epoch": 0.78, "learning_rate": 2.5358368326334736e-06, "loss": 0.0021, "step": 6498 }, { "epoch": 0.78, "learning_rate": 2.535764847030594e-06, "loss": 0.001, "step": 6499 }, { "epoch": 0.78, "learning_rate": 2.5356928614277148e-06, "loss": 0.0008, "step": 6500 }, { "epoch": 0.78, "learning_rate": 2.535620875824835e-06, "loss": 0.0007, "step": 6501 }, { "epoch": 0.78, "learning_rate": 2.535548890221956e-06, "loss": 0.4924, "step": 6502 }, { "epoch": 0.78, "learning_rate": 2.5354769046190762e-06, "loss": 0.0016, "step": 6503 }, { "epoch": 0.78, "learning_rate": 2.535404919016197e-06, "loss": 0.0157, "step": 6504 }, { "epoch": 0.78, "learning_rate": 2.5353329334133174e-06, "loss": 0.0049, "step": 6505 }, { "epoch": 0.78, "learning_rate": 2.5352609478104377e-06, "loss": 0.0006, "step": 6506 }, { "epoch": 0.78, "learning_rate": 2.5351889622075585e-06, "loss": 0.0022, "step": 6507 }, { "epoch": 0.78, "learning_rate": 2.535116976604679e-06, "loss": 0.0012, "step": 6508 }, { "epoch": 0.78, "learning_rate": 2.5350449910017996e-06, "loss": 0.0274, "step": 6509 }, { "epoch": 0.78, "learning_rate": 2.53497300539892e-06, "loss": 0.001, "step": 6510 }, { "epoch": 0.78, "learning_rate": 2.5349010197960408e-06, "loss": 0.0007, "step": 6511 }, { "epoch": 0.78, "learning_rate": 2.5348290341931616e-06, "loss": 0.001, "step": 6512 }, { "epoch": 0.78, "learning_rate": 2.534757048590282e-06, "loss": 0.0008, "step": 6513 }, { "epoch": 0.78, "learning_rate": 2.5346850629874027e-06, "loss": 0.5839, "step": 6514 }, { "epoch": 0.78, "learning_rate": 2.5346130773845235e-06, "loss": 0.0014, "step": 6515 }, { "epoch": 0.78, "learning_rate": 2.534541091781644e-06, "loss": 0.003, "step": 6516 }, { "epoch": 0.78, "learning_rate": 2.534469106178764e-06, "loss": 0.0008, "step": 6517 }, { "epoch": 0.78, "learning_rate": 2.534397120575885e-06, "loss": 0.0011, "step": 6518 }, { "epoch": 0.78, "learning_rate": 2.5343251349730053e-06, "loss": 0.0014, "step": 6519 }, { "epoch": 0.78, "learning_rate": 2.534253149370126e-06, "loss": 0.0005, "step": 6520 }, { "epoch": 0.78, "learning_rate": 2.5341811637672464e-06, "loss": 0.0042, "step": 6521 }, { "epoch": 0.78, "learning_rate": 2.5341091781643672e-06, "loss": 0.0007, "step": 6522 }, { "epoch": 0.78, "learning_rate": 2.5340371925614876e-06, "loss": 0.001, "step": 6523 }, { "epoch": 0.78, "learning_rate": 2.5339652069586084e-06, "loss": 0.0004, "step": 6524 }, { "epoch": 0.78, "learning_rate": 2.5338932213557287e-06, "loss": 0.0008, "step": 6525 }, { "epoch": 0.78, "learning_rate": 2.5338212357528495e-06, "loss": 0.0973, "step": 6526 }, { "epoch": 0.78, "learning_rate": 2.53374925014997e-06, "loss": 0.0008, "step": 6527 }, { "epoch": 0.78, "learning_rate": 2.5336772645470906e-06, "loss": 0.0009, "step": 6528 }, { "epoch": 0.78, "learning_rate": 2.5336052789442114e-06, "loss": 0.0012, "step": 6529 }, { "epoch": 0.78, "learning_rate": 2.5335332933413318e-06, "loss": 0.7707, "step": 6530 }, { "epoch": 0.78, "learning_rate": 2.5334613077384525e-06, "loss": 0.0017, "step": 6531 }, { "epoch": 0.78, "learning_rate": 2.533389322135573e-06, "loss": 0.0015, "step": 6532 }, { "epoch": 0.78, "learning_rate": 2.5333173365326937e-06, "loss": 0.0007, "step": 6533 }, { "epoch": 0.78, "learning_rate": 2.533245350929814e-06, "loss": 0.0009, "step": 6534 }, { "epoch": 0.78, "learning_rate": 2.533173365326935e-06, "loss": 0.0017, "step": 6535 }, { "epoch": 0.78, "learning_rate": 2.533101379724055e-06, "loss": 0.0011, "step": 6536 }, { "epoch": 0.78, "learning_rate": 2.533029394121176e-06, "loss": 0.429, "step": 6537 }, { "epoch": 0.78, "learning_rate": 2.5329574085182963e-06, "loss": 0.0016, "step": 6538 }, { "epoch": 0.78, "learning_rate": 2.532885422915417e-06, "loss": 0.3474, "step": 6539 }, { "epoch": 0.78, "learning_rate": 2.5328134373125374e-06, "loss": 0.6044, "step": 6540 }, { "epoch": 0.78, "learning_rate": 2.5327414517096578e-06, "loss": 0.0006, "step": 6541 }, { "epoch": 0.78, "learning_rate": 2.5326694661067786e-06, "loss": 0.0006, "step": 6542 }, { "epoch": 0.78, "learning_rate": 2.5325974805038993e-06, "loss": 0.0007, "step": 6543 }, { "epoch": 0.78, "learning_rate": 2.53252549490102e-06, "loss": 0.0011, "step": 6544 }, { "epoch": 0.78, "learning_rate": 2.5324535092981405e-06, "loss": 0.0013, "step": 6545 }, { "epoch": 0.78, "learning_rate": 2.5323815236952613e-06, "loss": 0.0013, "step": 6546 }, { "epoch": 0.78, "learning_rate": 2.5323095380923816e-06, "loss": 0.0014, "step": 6547 }, { "epoch": 0.78, "learning_rate": 2.5322375524895024e-06, "loss": 0.0009, "step": 6548 }, { "epoch": 0.78, "learning_rate": 2.5321655668866227e-06, "loss": 0.0006, "step": 6549 }, { "epoch": 0.78, "learning_rate": 2.5320935812837435e-06, "loss": 0.0023, "step": 6550 }, { "epoch": 0.79, "learning_rate": 2.532021595680864e-06, "loss": 0.0004, "step": 6551 }, { "epoch": 0.79, "learning_rate": 2.5319496100779842e-06, "loss": 0.0006, "step": 6552 }, { "epoch": 0.79, "learning_rate": 2.531877624475105e-06, "loss": 0.0006, "step": 6553 }, { "epoch": 0.79, "learning_rate": 2.5318056388722254e-06, "loss": 0.0005, "step": 6554 }, { "epoch": 0.79, "learning_rate": 2.531733653269346e-06, "loss": 0.0011, "step": 6555 }, { "epoch": 0.79, "learning_rate": 2.5316616676664665e-06, "loss": 0.0009, "step": 6556 }, { "epoch": 0.79, "learning_rate": 2.5315896820635873e-06, "loss": 0.0005, "step": 6557 }, { "epoch": 0.79, "learning_rate": 2.531517696460708e-06, "loss": 0.0015, "step": 6558 }, { "epoch": 0.79, "learning_rate": 2.5314457108578284e-06, "loss": 0.0016, "step": 6559 }, { "epoch": 0.79, "learning_rate": 2.531373725254949e-06, "loss": 0.0116, "step": 6560 }, { "epoch": 0.79, "learning_rate": 2.53130173965207e-06, "loss": 0.0008, "step": 6561 }, { "epoch": 0.79, "learning_rate": 2.5312297540491903e-06, "loss": 0.0013, "step": 6562 }, { "epoch": 0.79, "learning_rate": 2.5311577684463107e-06, "loss": 0.4789, "step": 6563 }, { "epoch": 0.79, "learning_rate": 2.5310857828434315e-06, "loss": 0.0007, "step": 6564 }, { "epoch": 0.79, "learning_rate": 2.531013797240552e-06, "loss": 0.0009, "step": 6565 }, { "epoch": 0.79, "learning_rate": 2.5309418116376726e-06, "loss": 0.0006, "step": 6566 }, { "epoch": 0.79, "learning_rate": 2.530869826034793e-06, "loss": 0.0008, "step": 6567 }, { "epoch": 0.79, "learning_rate": 2.5307978404319137e-06, "loss": 0.0005, "step": 6568 }, { "epoch": 0.79, "learning_rate": 2.530725854829034e-06, "loss": 0.0003, "step": 6569 }, { "epoch": 0.79, "learning_rate": 2.530653869226155e-06, "loss": 0.0008, "step": 6570 }, { "epoch": 0.79, "learning_rate": 2.5305818836232752e-06, "loss": 0.0005, "step": 6571 }, { "epoch": 0.79, "learning_rate": 2.530509898020396e-06, "loss": 0.0009, "step": 6572 }, { "epoch": 0.79, "learning_rate": 2.5304379124175163e-06, "loss": 0.0006, "step": 6573 }, { "epoch": 0.79, "learning_rate": 2.530365926814637e-06, "loss": 0.0013, "step": 6574 }, { "epoch": 0.79, "learning_rate": 2.530293941211758e-06, "loss": 0.0014, "step": 6575 }, { "epoch": 0.79, "learning_rate": 2.5302219556088783e-06, "loss": 0.0005, "step": 6576 }, { "epoch": 0.79, "learning_rate": 2.530149970005999e-06, "loss": 0.0007, "step": 6577 }, { "epoch": 0.79, "learning_rate": 2.5300779844031194e-06, "loss": 0.0004, "step": 6578 }, { "epoch": 0.79, "learning_rate": 2.53000599880024e-06, "loss": 0.0005, "step": 6579 }, { "epoch": 0.79, "learning_rate": 2.5299340131973605e-06, "loss": 0.9672, "step": 6580 }, { "epoch": 0.79, "learning_rate": 2.5298620275944813e-06, "loss": 0.0006, "step": 6581 }, { "epoch": 0.79, "learning_rate": 2.5297900419916017e-06, "loss": 0.0008, "step": 6582 }, { "epoch": 0.79, "learning_rate": 2.5297180563887224e-06, "loss": 0.0007, "step": 6583 }, { "epoch": 0.79, "learning_rate": 2.529646070785843e-06, "loss": 0.0006, "step": 6584 }, { "epoch": 0.79, "learning_rate": 2.5295740851829636e-06, "loss": 0.0005, "step": 6585 }, { "epoch": 0.79, "learning_rate": 2.529502099580084e-06, "loss": 0.0017, "step": 6586 }, { "epoch": 0.79, "learning_rate": 2.5294301139772043e-06, "loss": 0.0007, "step": 6587 }, { "epoch": 0.79, "learning_rate": 2.529358128374325e-06, "loss": 0.0004, "step": 6588 }, { "epoch": 0.79, "learning_rate": 2.529286142771446e-06, "loss": 0.0004, "step": 6589 }, { "epoch": 0.79, "learning_rate": 2.5292141571685666e-06, "loss": 0.0005, "step": 6590 }, { "epoch": 0.79, "learning_rate": 2.529142171565687e-06, "loss": 0.0431, "step": 6591 }, { "epoch": 0.79, "learning_rate": 2.5290701859628078e-06, "loss": 0.0009, "step": 6592 }, { "epoch": 0.79, "learning_rate": 2.528998200359928e-06, "loss": 0.0055, "step": 6593 }, { "epoch": 0.79, "learning_rate": 2.528926214757049e-06, "loss": 0.0007, "step": 6594 }, { "epoch": 0.79, "learning_rate": 2.5288542291541692e-06, "loss": 0.0006, "step": 6595 }, { "epoch": 0.79, "learning_rate": 2.52878224355129e-06, "loss": 0.0018, "step": 6596 }, { "epoch": 0.79, "learning_rate": 2.5287102579484104e-06, "loss": 0.0007, "step": 6597 }, { "epoch": 0.79, "learning_rate": 2.5286382723455307e-06, "loss": 0.0004, "step": 6598 }, { "epoch": 0.79, "learning_rate": 2.5285662867426515e-06, "loss": 0.0004, "step": 6599 }, { "epoch": 0.79, "learning_rate": 2.528494301139772e-06, "loss": 0.0009, "step": 6600 }, { "epoch": 0.79, "learning_rate": 2.5284223155368926e-06, "loss": 0.001, "step": 6601 }, { "epoch": 0.79, "learning_rate": 2.528350329934013e-06, "loss": 0.0004, "step": 6602 }, { "epoch": 0.79, "learning_rate": 2.5282783443311338e-06, "loss": 0.0007, "step": 6603 }, { "epoch": 0.79, "learning_rate": 2.5282063587282546e-06, "loss": 0.0014, "step": 6604 }, { "epoch": 0.79, "learning_rate": 2.528134373125375e-06, "loss": 0.0006, "step": 6605 }, { "epoch": 0.79, "learning_rate": 2.5280623875224957e-06, "loss": 0.0006, "step": 6606 }, { "epoch": 0.79, "learning_rate": 2.5279904019196165e-06, "loss": 0.9948, "step": 6607 }, { "epoch": 0.79, "learning_rate": 2.527918416316737e-06, "loss": 0.0005, "step": 6608 }, { "epoch": 0.79, "learning_rate": 2.527846430713857e-06, "loss": 0.5783, "step": 6609 }, { "epoch": 0.79, "learning_rate": 2.527774445110978e-06, "loss": 0.0017, "step": 6610 }, { "epoch": 0.79, "learning_rate": 2.5277024595080983e-06, "loss": 0.0008, "step": 6611 }, { "epoch": 0.79, "learning_rate": 2.527630473905219e-06, "loss": 0.6594, "step": 6612 }, { "epoch": 0.79, "learning_rate": 2.5275584883023394e-06, "loss": 1.4711, "step": 6613 }, { "epoch": 0.79, "learning_rate": 2.5274865026994602e-06, "loss": 0.0006, "step": 6614 }, { "epoch": 0.79, "learning_rate": 2.5274145170965806e-06, "loss": 0.0021, "step": 6615 }, { "epoch": 0.79, "learning_rate": 2.5273425314937014e-06, "loss": 0.0007, "step": 6616 }, { "epoch": 0.79, "learning_rate": 2.5272705458908217e-06, "loss": 0.0008, "step": 6617 }, { "epoch": 0.79, "learning_rate": 2.5271985602879425e-06, "loss": 0.1926, "step": 6618 }, { "epoch": 0.79, "learning_rate": 2.527126574685063e-06, "loss": 0.001, "step": 6619 }, { "epoch": 0.79, "learning_rate": 2.5270545890821836e-06, "loss": 0.0007, "step": 6620 }, { "epoch": 0.79, "learning_rate": 2.5269826034793044e-06, "loss": 0.0029, "step": 6621 }, { "epoch": 0.79, "learning_rate": 2.5269106178764248e-06, "loss": 0.0008, "step": 6622 }, { "epoch": 0.79, "learning_rate": 2.5268386322735455e-06, "loss": 0.4903, "step": 6623 }, { "epoch": 0.79, "learning_rate": 2.526766646670666e-06, "loss": 0.0014, "step": 6624 }, { "epoch": 0.79, "learning_rate": 2.5266946610677867e-06, "loss": 0.0009, "step": 6625 }, { "epoch": 0.79, "learning_rate": 2.526622675464907e-06, "loss": 0.0014, "step": 6626 }, { "epoch": 0.79, "learning_rate": 2.526550689862028e-06, "loss": 0.0004, "step": 6627 }, { "epoch": 0.79, "learning_rate": 2.526478704259148e-06, "loss": 0.0011, "step": 6628 }, { "epoch": 0.79, "learning_rate": 2.526406718656269e-06, "loss": 0.0008, "step": 6629 }, { "epoch": 0.79, "learning_rate": 2.5263347330533893e-06, "loss": 0.0007, "step": 6630 }, { "epoch": 0.79, "learning_rate": 2.52626274745051e-06, "loss": 0.0005, "step": 6631 }, { "epoch": 0.79, "learning_rate": 2.5261907618476304e-06, "loss": 0.0008, "step": 6632 }, { "epoch": 0.79, "learning_rate": 2.5261187762447508e-06, "loss": 0.0009, "step": 6633 }, { "epoch": 0.79, "learning_rate": 2.5260467906418716e-06, "loss": 0.0007, "step": 6634 }, { "epoch": 0.8, "learning_rate": 2.5259748050389923e-06, "loss": 0.0008, "step": 6635 }, { "epoch": 0.8, "learning_rate": 2.5259028194361127e-06, "loss": 0.0011, "step": 6636 }, { "epoch": 0.8, "learning_rate": 2.5258308338332335e-06, "loss": 0.8148, "step": 6637 }, { "epoch": 0.8, "learning_rate": 2.5257588482303542e-06, "loss": 0.0017, "step": 6638 }, { "epoch": 0.8, "learning_rate": 2.5256868626274746e-06, "loss": 0.0004, "step": 6639 }, { "epoch": 0.8, "learning_rate": 2.5256148770245954e-06, "loss": 0.0014, "step": 6640 }, { "epoch": 0.8, "learning_rate": 2.5255428914217157e-06, "loss": 0.0012, "step": 6641 }, { "epoch": 0.8, "learning_rate": 2.5254709058188365e-06, "loss": 0.0012, "step": 6642 }, { "epoch": 0.8, "learning_rate": 2.525398920215957e-06, "loss": 0.0007, "step": 6643 }, { "epoch": 0.8, "learning_rate": 2.5253269346130772e-06, "loss": 0.0028, "step": 6644 }, { "epoch": 0.8, "learning_rate": 2.525254949010198e-06, "loss": 0.0007, "step": 6645 }, { "epoch": 0.8, "learning_rate": 2.5251829634073184e-06, "loss": 0.0007, "step": 6646 }, { "epoch": 0.8, "learning_rate": 2.525110977804439e-06, "loss": 0.0008, "step": 6647 }, { "epoch": 0.8, "learning_rate": 2.5250389922015595e-06, "loss": 0.0008, "step": 6648 }, { "epoch": 0.8, "learning_rate": 2.5249670065986803e-06, "loss": 0.0006, "step": 6649 }, { "epoch": 0.8, "learning_rate": 2.5248950209958006e-06, "loss": 0.0013, "step": 6650 }, { "epoch": 0.8, "learning_rate": 2.5248230353929214e-06, "loss": 0.0007, "step": 6651 }, { "epoch": 0.8, "learning_rate": 2.524751049790042e-06, "loss": 0.0029, "step": 6652 }, { "epoch": 0.8, "learning_rate": 2.524679064187163e-06, "loss": 0.0008, "step": 6653 }, { "epoch": 0.8, "learning_rate": 2.5246070785842833e-06, "loss": 0.0012, "step": 6654 }, { "epoch": 0.8, "learning_rate": 2.524535092981404e-06, "loss": 0.0004, "step": 6655 }, { "epoch": 0.8, "learning_rate": 2.5244631073785244e-06, "loss": 0.4973, "step": 6656 }, { "epoch": 0.8, "learning_rate": 2.524391121775645e-06, "loss": 0.001, "step": 6657 }, { "epoch": 0.8, "learning_rate": 2.5243191361727656e-06, "loss": 0.0009, "step": 6658 }, { "epoch": 0.8, "learning_rate": 2.524247150569886e-06, "loss": 0.1036, "step": 6659 }, { "epoch": 0.8, "learning_rate": 2.5241751649670067e-06, "loss": 0.6722, "step": 6660 }, { "epoch": 0.8, "learning_rate": 2.524103179364127e-06, "loss": 0.0007, "step": 6661 }, { "epoch": 0.8, "learning_rate": 2.524031193761248e-06, "loss": 0.059, "step": 6662 }, { "epoch": 0.8, "learning_rate": 2.523959208158368e-06, "loss": 0.0025, "step": 6663 }, { "epoch": 0.8, "learning_rate": 2.523887222555489e-06, "loss": 0.0106, "step": 6664 }, { "epoch": 0.8, "learning_rate": 2.5238152369526093e-06, "loss": 0.0012, "step": 6665 }, { "epoch": 0.8, "learning_rate": 2.52374325134973e-06, "loss": 0.0003, "step": 6666 }, { "epoch": 0.8, "learning_rate": 2.523671265746851e-06, "loss": 0.0006, "step": 6667 }, { "epoch": 0.8, "learning_rate": 2.5235992801439712e-06, "loss": 0.0206, "step": 6668 }, { "epoch": 0.8, "learning_rate": 2.523527294541092e-06, "loss": 0.0005, "step": 6669 }, { "epoch": 0.8, "learning_rate": 2.5234553089382124e-06, "loss": 0.0004, "step": 6670 }, { "epoch": 0.8, "learning_rate": 2.523383323335333e-06, "loss": 0.0005, "step": 6671 }, { "epoch": 0.8, "learning_rate": 2.5233113377324535e-06, "loss": 0.0011, "step": 6672 }, { "epoch": 0.8, "learning_rate": 2.5232393521295743e-06, "loss": 0.0009, "step": 6673 }, { "epoch": 0.8, "learning_rate": 2.5231673665266947e-06, "loss": 0.0006, "step": 6674 }, { "epoch": 0.8, "learning_rate": 2.5230953809238154e-06, "loss": 0.001, "step": 6675 }, { "epoch": 0.8, "learning_rate": 2.5230233953209358e-06, "loss": 0.0014, "step": 6676 }, { "epoch": 0.8, "learning_rate": 2.5229514097180566e-06, "loss": 0.0006, "step": 6677 }, { "epoch": 0.8, "learning_rate": 2.522879424115177e-06, "loss": 0.0007, "step": 6678 }, { "epoch": 0.8, "learning_rate": 2.5228074385122973e-06, "loss": 0.0004, "step": 6679 }, { "epoch": 0.8, "learning_rate": 2.522735452909418e-06, "loss": 0.0004, "step": 6680 }, { "epoch": 0.8, "learning_rate": 2.522663467306539e-06, "loss": 0.0078, "step": 6681 }, { "epoch": 0.8, "learning_rate": 2.522591481703659e-06, "loss": 0.0006, "step": 6682 }, { "epoch": 0.8, "learning_rate": 2.52251949610078e-06, "loss": 1.7281, "step": 6683 }, { "epoch": 0.8, "learning_rate": 2.5224475104979007e-06, "loss": 0.0003, "step": 6684 }, { "epoch": 0.8, "learning_rate": 2.522375524895021e-06, "loss": 0.0006, "step": 6685 }, { "epoch": 0.8, "learning_rate": 2.522303539292142e-06, "loss": 0.9929, "step": 6686 }, { "epoch": 0.8, "learning_rate": 2.5222315536892622e-06, "loss": 0.0007, "step": 6687 }, { "epoch": 0.8, "learning_rate": 2.522159568086383e-06, "loss": 0.0005, "step": 6688 }, { "epoch": 0.8, "learning_rate": 2.5220875824835034e-06, "loss": 0.0006, "step": 6689 }, { "epoch": 0.8, "learning_rate": 2.5220155968806237e-06, "loss": 0.0012, "step": 6690 }, { "epoch": 0.8, "learning_rate": 2.5219436112777445e-06, "loss": 0.0007, "step": 6691 }, { "epoch": 0.8, "learning_rate": 2.521871625674865e-06, "loss": 0.0017, "step": 6692 }, { "epoch": 0.8, "learning_rate": 2.5217996400719856e-06, "loss": 0.0005, "step": 6693 }, { "epoch": 0.8, "learning_rate": 2.521727654469106e-06, "loss": 0.8725, "step": 6694 }, { "epoch": 0.8, "learning_rate": 2.5216556688662268e-06, "loss": 0.6701, "step": 6695 }, { "epoch": 0.8, "learning_rate": 2.521583683263347e-06, "loss": 0.0006, "step": 6696 }, { "epoch": 0.8, "learning_rate": 2.521511697660468e-06, "loss": 0.0007, "step": 6697 }, { "epoch": 0.8, "learning_rate": 2.5214397120575887e-06, "loss": 0.0007, "step": 6698 }, { "epoch": 0.8, "learning_rate": 2.5213677264547095e-06, "loss": 0.0021, "step": 6699 }, { "epoch": 0.8, "learning_rate": 2.52129574085183e-06, "loss": 0.0015, "step": 6700 }, { "epoch": 0.8, "learning_rate": 2.5212237552489506e-06, "loss": 0.0031, "step": 6701 }, { "epoch": 0.8, "learning_rate": 2.521151769646071e-06, "loss": 0.0006, "step": 6702 }, { "epoch": 0.8, "learning_rate": 2.5210797840431913e-06, "loss": 0.7994, "step": 6703 }, { "epoch": 0.8, "learning_rate": 2.521007798440312e-06, "loss": 0.1807, "step": 6704 }, { "epoch": 0.8, "learning_rate": 2.5209358128374324e-06, "loss": 0.0023, "step": 6705 }, { "epoch": 0.8, "learning_rate": 2.520863827234553e-06, "loss": 0.1311, "step": 6706 }, { "epoch": 0.8, "learning_rate": 2.5207918416316736e-06, "loss": 0.0444, "step": 6707 }, { "epoch": 0.8, "learning_rate": 2.5207198560287943e-06, "loss": 0.0021, "step": 6708 }, { "epoch": 0.8, "learning_rate": 2.5206478704259147e-06, "loss": 0.0401, "step": 6709 }, { "epoch": 0.8, "learning_rate": 2.5205758848230355e-06, "loss": 0.0005, "step": 6710 }, { "epoch": 0.8, "learning_rate": 2.520503899220156e-06, "loss": 0.0004, "step": 6711 }, { "epoch": 0.8, "learning_rate": 2.5204319136172766e-06, "loss": 0.0008, "step": 6712 }, { "epoch": 0.8, "learning_rate": 2.5203599280143974e-06, "loss": 0.8874, "step": 6713 }, { "epoch": 0.8, "learning_rate": 2.5202879424115177e-06, "loss": 0.0007, "step": 6714 }, { "epoch": 0.8, "learning_rate": 2.5202159568086385e-06, "loss": 0.0005, "step": 6715 }, { "epoch": 0.8, "learning_rate": 2.520143971205759e-06, "loss": 0.0009, "step": 6716 }, { "epoch": 0.8, "learning_rate": 2.5200719856028797e-06, "loss": 0.0012, "step": 6717 }, { "epoch": 0.81, "learning_rate": 2.52e-06, "loss": 0.0009, "step": 6718 }, { "epoch": 0.81, "learning_rate": 2.519928014397121e-06, "loss": 0.0006, "step": 6719 }, { "epoch": 0.81, "learning_rate": 2.519856028794241e-06, "loss": 0.0006, "step": 6720 }, { "epoch": 0.81, "learning_rate": 2.519784043191362e-06, "loss": 0.001, "step": 6721 }, { "epoch": 0.81, "learning_rate": 2.5197120575884823e-06, "loss": 0.0008, "step": 6722 }, { "epoch": 0.81, "learning_rate": 2.519640071985603e-06, "loss": 0.0007, "step": 6723 }, { "epoch": 0.81, "learning_rate": 2.5195680863827234e-06, "loss": 0.0019, "step": 6724 }, { "epoch": 0.81, "learning_rate": 2.5194961007798438e-06, "loss": 0.0028, "step": 6725 }, { "epoch": 0.81, "learning_rate": 2.5194241151769645e-06, "loss": 0.0008, "step": 6726 }, { "epoch": 0.81, "learning_rate": 2.5193521295740853e-06, "loss": 0.0008, "step": 6727 }, { "epoch": 0.81, "learning_rate": 2.5192801439712057e-06, "loss": 0.0146, "step": 6728 }, { "epoch": 0.81, "learning_rate": 2.5192081583683265e-06, "loss": 0.0008, "step": 6729 }, { "epoch": 0.81, "learning_rate": 2.5191361727654472e-06, "loss": 0.6729, "step": 6730 }, { "epoch": 0.81, "learning_rate": 2.5190641871625676e-06, "loss": 0.0006, "step": 6731 }, { "epoch": 0.81, "learning_rate": 2.5189922015596884e-06, "loss": 0.6486, "step": 6732 }, { "epoch": 0.81, "learning_rate": 2.5189202159568087e-06, "loss": 0.0009, "step": 6733 }, { "epoch": 0.81, "learning_rate": 2.5188482303539295e-06, "loss": 0.0009, "step": 6734 }, { "epoch": 0.81, "learning_rate": 2.51877624475105e-06, "loss": 0.0007, "step": 6735 }, { "epoch": 0.81, "learning_rate": 2.5187042591481706e-06, "loss": 0.0013, "step": 6736 }, { "epoch": 0.81, "learning_rate": 2.518632273545291e-06, "loss": 0.0004, "step": 6737 }, { "epoch": 0.81, "learning_rate": 2.5185602879424113e-06, "loss": 0.0006, "step": 6738 }, { "epoch": 0.81, "learning_rate": 2.518488302339532e-06, "loss": 0.0008, "step": 6739 }, { "epoch": 0.81, "learning_rate": 2.5184163167366525e-06, "loss": 0.001, "step": 6740 }, { "epoch": 0.81, "learning_rate": 2.5183443311337733e-06, "loss": 0.0146, "step": 6741 }, { "epoch": 0.81, "learning_rate": 2.5182723455308936e-06, "loss": 0.0011, "step": 6742 }, { "epoch": 0.81, "learning_rate": 2.5182003599280144e-06, "loss": 0.001, "step": 6743 }, { "epoch": 0.81, "learning_rate": 2.518128374325135e-06, "loss": 0.0037, "step": 6744 }, { "epoch": 0.81, "learning_rate": 2.518056388722256e-06, "loss": 0.0009, "step": 6745 }, { "epoch": 0.81, "learning_rate": 2.5179844031193763e-06, "loss": 0.0009, "step": 6746 }, { "epoch": 0.81, "learning_rate": 2.517912417516497e-06, "loss": 0.0006, "step": 6747 }, { "epoch": 0.81, "learning_rate": 2.5178404319136174e-06, "loss": 0.3288, "step": 6748 }, { "epoch": 0.81, "learning_rate": 2.517768446310738e-06, "loss": 0.0011, "step": 6749 }, { "epoch": 0.81, "learning_rate": 2.5176964607078586e-06, "loss": 0.0014, "step": 6750 }, { "epoch": 0.81, "learning_rate": 2.517624475104979e-06, "loss": 0.0467, "step": 6751 }, { "epoch": 0.81, "learning_rate": 2.5175524895020997e-06, "loss": 0.001, "step": 6752 }, { "epoch": 0.81, "learning_rate": 2.51748050389922e-06, "loss": 0.0017, "step": 6753 }, { "epoch": 0.81, "learning_rate": 2.517408518296341e-06, "loss": 0.0006, "step": 6754 }, { "epoch": 0.81, "learning_rate": 2.517336532693461e-06, "loss": 0.0008, "step": 6755 }, { "epoch": 0.81, "learning_rate": 2.517264547090582e-06, "loss": 0.0015, "step": 6756 }, { "epoch": 0.81, "learning_rate": 2.5171925614877023e-06, "loss": 0.0005, "step": 6757 }, { "epoch": 0.81, "learning_rate": 2.517120575884823e-06, "loss": 0.0015, "step": 6758 }, { "epoch": 0.81, "learning_rate": 2.5170485902819435e-06, "loss": 0.0017, "step": 6759 }, { "epoch": 0.81, "learning_rate": 2.5169766046790642e-06, "loss": 0.0008, "step": 6760 }, { "epoch": 0.81, "learning_rate": 2.516904619076185e-06, "loss": 0.0006, "step": 6761 }, { "epoch": 0.81, "learning_rate": 2.5168326334733054e-06, "loss": 0.001, "step": 6762 }, { "epoch": 0.81, "learning_rate": 2.516760647870426e-06, "loss": 0.0005, "step": 6763 }, { "epoch": 0.81, "learning_rate": 2.5166886622675465e-06, "loss": 0.003, "step": 6764 }, { "epoch": 0.81, "learning_rate": 2.5166166766646673e-06, "loss": 0.0007, "step": 6765 }, { "epoch": 0.81, "learning_rate": 2.5165446910617876e-06, "loss": 0.0004, "step": 6766 }, { "epoch": 0.81, "learning_rate": 2.5164727054589084e-06, "loss": 0.0004, "step": 6767 }, { "epoch": 0.81, "learning_rate": 2.5164007198560288e-06, "loss": 0.0005, "step": 6768 }, { "epoch": 0.81, "learning_rate": 2.5163287342531496e-06, "loss": 1.1653, "step": 6769 }, { "epoch": 0.81, "learning_rate": 2.51625674865027e-06, "loss": 0.0009, "step": 6770 }, { "epoch": 0.81, "learning_rate": 2.5161847630473903e-06, "loss": 0.0006, "step": 6771 }, { "epoch": 0.81, "learning_rate": 2.516112777444511e-06, "loss": 0.7224, "step": 6772 }, { "epoch": 0.81, "learning_rate": 2.5160407918416314e-06, "loss": 0.0013, "step": 6773 }, { "epoch": 0.81, "learning_rate": 2.515968806238752e-06, "loss": 0.0008, "step": 6774 }, { "epoch": 0.81, "learning_rate": 2.515896820635873e-06, "loss": 0.0006, "step": 6775 }, { "epoch": 0.81, "learning_rate": 2.5158248350329937e-06, "loss": 0.0068, "step": 6776 }, { "epoch": 0.81, "learning_rate": 2.515752849430114e-06, "loss": 0.0009, "step": 6777 }, { "epoch": 0.81, "learning_rate": 2.515680863827235e-06, "loss": 0.0007, "step": 6778 }, { "epoch": 0.81, "learning_rate": 2.5156088782243552e-06, "loss": 0.0006, "step": 6779 }, { "epoch": 0.81, "learning_rate": 2.515536892621476e-06, "loss": 0.0004, "step": 6780 }, { "epoch": 0.81, "learning_rate": 2.5154649070185964e-06, "loss": 0.0007, "step": 6781 }, { "epoch": 0.81, "learning_rate": 2.515392921415717e-06, "loss": 0.0065, "step": 6782 }, { "epoch": 0.81, "learning_rate": 2.5153209358128375e-06, "loss": 0.0007, "step": 6783 }, { "epoch": 0.81, "learning_rate": 2.515248950209958e-06, "loss": 0.0009, "step": 6784 }, { "epoch": 0.81, "learning_rate": 2.5151769646070786e-06, "loss": 0.0006, "step": 6785 }, { "epoch": 0.81, "learning_rate": 2.515104979004199e-06, "loss": 0.0007, "step": 6786 }, { "epoch": 0.81, "learning_rate": 2.5150329934013198e-06, "loss": 0.0011, "step": 6787 }, { "epoch": 0.81, "learning_rate": 2.51496100779844e-06, "loss": 0.0027, "step": 6788 }, { "epoch": 0.81, "learning_rate": 2.514889022195561e-06, "loss": 0.0132, "step": 6789 }, { "epoch": 0.81, "learning_rate": 2.5148170365926817e-06, "loss": 0.0007, "step": 6790 }, { "epoch": 0.81, "learning_rate": 2.514745050989802e-06, "loss": 0.4415, "step": 6791 }, { "epoch": 0.81, "learning_rate": 2.514673065386923e-06, "loss": 0.0006, "step": 6792 }, { "epoch": 0.81, "learning_rate": 2.5146010797840436e-06, "loss": 0.0005, "step": 6793 }, { "epoch": 0.81, "learning_rate": 2.514529094181164e-06, "loss": 0.0007, "step": 6794 }, { "epoch": 0.81, "learning_rate": 2.5144571085782843e-06, "loss": 0.0008, "step": 6795 }, { "epoch": 0.81, "learning_rate": 2.514385122975405e-06, "loss": 0.0004, "step": 6796 }, { "epoch": 0.81, "learning_rate": 2.5143131373725254e-06, "loss": 0.0011, "step": 6797 }, { "epoch": 0.81, "learning_rate": 2.514241151769646e-06, "loss": 0.0008, "step": 6798 }, { "epoch": 0.81, "learning_rate": 2.5141691661667666e-06, "loss": 0.0006, "step": 6799 }, { "epoch": 0.81, "learning_rate": 2.5140971805638873e-06, "loss": 0.0008, "step": 6800 }, { "epoch": 0.81, "learning_rate": 2.5140251949610077e-06, "loss": 0.0006, "step": 6801 }, { "epoch": 0.82, "learning_rate": 2.5139532093581285e-06, "loss": 0.0154, "step": 6802 }, { "epoch": 0.82, "learning_rate": 2.513881223755249e-06, "loss": 0.0006, "step": 6803 }, { "epoch": 0.82, "learning_rate": 2.5138092381523696e-06, "loss": 0.0006, "step": 6804 }, { "epoch": 0.82, "learning_rate": 2.51373725254949e-06, "loss": 0.0044, "step": 6805 }, { "epoch": 0.82, "learning_rate": 2.5136652669466107e-06, "loss": 0.0009, "step": 6806 }, { "epoch": 0.82, "learning_rate": 2.5135932813437315e-06, "loss": 0.0008, "step": 6807 }, { "epoch": 0.82, "learning_rate": 2.513521295740852e-06, "loss": 0.0005, "step": 6808 }, { "epoch": 0.82, "learning_rate": 2.5134493101379726e-06, "loss": 0.0011, "step": 6809 }, { "epoch": 0.82, "learning_rate": 2.513377324535093e-06, "loss": 0.0004, "step": 6810 }, { "epoch": 0.82, "learning_rate": 2.5133053389322138e-06, "loss": 0.0005, "step": 6811 }, { "epoch": 0.82, "learning_rate": 2.513233353329334e-06, "loss": 0.1509, "step": 6812 }, { "epoch": 0.82, "learning_rate": 2.513161367726455e-06, "loss": 0.0005, "step": 6813 }, { "epoch": 0.82, "learning_rate": 2.5130893821235753e-06, "loss": 0.0006, "step": 6814 }, { "epoch": 0.82, "learning_rate": 2.513017396520696e-06, "loss": 0.0006, "step": 6815 }, { "epoch": 0.82, "learning_rate": 2.5129454109178164e-06, "loss": 0.0019, "step": 6816 }, { "epoch": 0.82, "learning_rate": 2.5128734253149368e-06, "loss": 0.0017, "step": 6817 }, { "epoch": 0.82, "learning_rate": 2.5128014397120575e-06, "loss": 0.0007, "step": 6818 }, { "epoch": 0.82, "learning_rate": 2.512729454109178e-06, "loss": 0.0006, "step": 6819 }, { "epoch": 0.82, "learning_rate": 2.5126574685062987e-06, "loss": 0.0231, "step": 6820 }, { "epoch": 0.82, "learning_rate": 2.5125854829034195e-06, "loss": 0.0005, "step": 6821 }, { "epoch": 0.82, "learning_rate": 2.5125134973005402e-06, "loss": 0.0007, "step": 6822 }, { "epoch": 0.82, "learning_rate": 2.5124415116976606e-06, "loss": 0.0011, "step": 6823 }, { "epoch": 0.82, "learning_rate": 2.5123695260947814e-06, "loss": 0.0007, "step": 6824 }, { "epoch": 0.82, "learning_rate": 2.5122975404919017e-06, "loss": 0.0007, "step": 6825 }, { "epoch": 0.82, "learning_rate": 2.5122255548890225e-06, "loss": 0.0008, "step": 6826 }, { "epoch": 0.82, "learning_rate": 2.512153569286143e-06, "loss": 0.0007, "step": 6827 }, { "epoch": 0.82, "learning_rate": 2.5120815836832636e-06, "loss": 0.0013, "step": 6828 }, { "epoch": 0.82, "learning_rate": 2.512009598080384e-06, "loss": 0.5665, "step": 6829 }, { "epoch": 0.82, "learning_rate": 2.5119376124775043e-06, "loss": 0.0007, "step": 6830 }, { "epoch": 0.82, "learning_rate": 2.511865626874625e-06, "loss": 0.0013, "step": 6831 }, { "epoch": 0.82, "learning_rate": 2.5117936412717455e-06, "loss": 0.0006, "step": 6832 }, { "epoch": 0.82, "learning_rate": 2.5117216556688663e-06, "loss": 0.0007, "step": 6833 }, { "epoch": 0.82, "learning_rate": 2.5116496700659866e-06, "loss": 0.6197, "step": 6834 }, { "epoch": 0.82, "learning_rate": 2.5115776844631074e-06, "loss": 0.0013, "step": 6835 }, { "epoch": 0.82, "learning_rate": 2.511505698860228e-06, "loss": 0.0005, "step": 6836 }, { "epoch": 0.82, "learning_rate": 2.5114337132573485e-06, "loss": 0.0003, "step": 6837 }, { "epoch": 0.82, "learning_rate": 2.5113617276544693e-06, "loss": 0.0005, "step": 6838 }, { "epoch": 0.82, "learning_rate": 2.51128974205159e-06, "loss": 0.0005, "step": 6839 }, { "epoch": 0.82, "learning_rate": 2.5112177564487104e-06, "loss": 0.0008, "step": 6840 }, { "epoch": 0.82, "learning_rate": 2.5111457708458308e-06, "loss": 0.1583, "step": 6841 }, { "epoch": 0.82, "learning_rate": 2.5110737852429516e-06, "loss": 0.0015, "step": 6842 }, { "epoch": 0.82, "learning_rate": 2.511001799640072e-06, "loss": 0.0009, "step": 6843 }, { "epoch": 0.82, "learning_rate": 2.5109298140371927e-06, "loss": 0.0007, "step": 6844 }, { "epoch": 0.82, "learning_rate": 2.510857828434313e-06, "loss": 0.0005, "step": 6845 }, { "epoch": 0.82, "learning_rate": 2.510785842831434e-06, "loss": 0.6321, "step": 6846 }, { "epoch": 0.82, "learning_rate": 2.510713857228554e-06, "loss": 0.0007, "step": 6847 }, { "epoch": 0.82, "learning_rate": 2.510641871625675e-06, "loss": 0.0004, "step": 6848 }, { "epoch": 0.82, "learning_rate": 2.5105698860227953e-06, "loss": 0.0006, "step": 6849 }, { "epoch": 0.82, "learning_rate": 2.510497900419916e-06, "loss": 0.0009, "step": 6850 }, { "epoch": 0.82, "learning_rate": 2.5104259148170365e-06, "loss": 0.0007, "step": 6851 }, { "epoch": 0.82, "learning_rate": 2.5103539292141572e-06, "loss": 0.0003, "step": 6852 }, { "epoch": 0.82, "learning_rate": 2.510281943611278e-06, "loss": 0.0005, "step": 6853 }, { "epoch": 0.82, "learning_rate": 2.5102099580083984e-06, "loss": 0.0004, "step": 6854 }, { "epoch": 0.82, "learning_rate": 2.510137972405519e-06, "loss": 0.0005, "step": 6855 }, { "epoch": 0.82, "learning_rate": 2.5100659868026395e-06, "loss": 0.0006, "step": 6856 }, { "epoch": 0.82, "learning_rate": 2.5099940011997603e-06, "loss": 0.003, "step": 6857 }, { "epoch": 0.82, "learning_rate": 2.5099220155968806e-06, "loss": 0.0004, "step": 6858 }, { "epoch": 0.82, "learning_rate": 2.5098500299940014e-06, "loss": 0.0184, "step": 6859 }, { "epoch": 0.82, "learning_rate": 2.5097780443911218e-06, "loss": 0.0006, "step": 6860 }, { "epoch": 0.82, "learning_rate": 2.5097060587882425e-06, "loss": 0.0004, "step": 6861 }, { "epoch": 0.82, "learning_rate": 2.509634073185363e-06, "loss": 0.0004, "step": 6862 }, { "epoch": 0.82, "learning_rate": 2.5095620875824837e-06, "loss": 0.0005, "step": 6863 }, { "epoch": 0.82, "learning_rate": 2.509490101979604e-06, "loss": 0.0003, "step": 6864 }, { "epoch": 0.82, "learning_rate": 2.5094181163767244e-06, "loss": 0.0006, "step": 6865 }, { "epoch": 0.82, "learning_rate": 2.509346130773845e-06, "loss": 0.0005, "step": 6866 }, { "epoch": 0.82, "learning_rate": 2.509274145170966e-06, "loss": 0.0005, "step": 6867 }, { "epoch": 0.82, "learning_rate": 2.5092021595680867e-06, "loss": 0.001, "step": 6868 }, { "epoch": 0.82, "learning_rate": 2.509130173965207e-06, "loss": 0.8285, "step": 6869 }, { "epoch": 0.82, "learning_rate": 2.509058188362328e-06, "loss": 0.0005, "step": 6870 }, { "epoch": 0.82, "learning_rate": 2.5089862027594482e-06, "loss": 0.0009, "step": 6871 }, { "epoch": 0.82, "learning_rate": 2.508914217156569e-06, "loss": 0.0003, "step": 6872 }, { "epoch": 0.82, "learning_rate": 2.5088422315536893e-06, "loss": 0.0004, "step": 6873 }, { "epoch": 0.82, "learning_rate": 2.50877024595081e-06, "loss": 0.0018, "step": 6874 }, { "epoch": 0.82, "learning_rate": 2.5086982603479305e-06, "loss": 0.0004, "step": 6875 }, { "epoch": 0.82, "learning_rate": 2.508626274745051e-06, "loss": 0.0006, "step": 6876 }, { "epoch": 0.82, "learning_rate": 2.5085542891421716e-06, "loss": 0.0005, "step": 6877 }, { "epoch": 0.82, "learning_rate": 2.508482303539292e-06, "loss": 0.0004, "step": 6878 }, { "epoch": 0.82, "learning_rate": 2.5084103179364127e-06, "loss": 0.0008, "step": 6879 }, { "epoch": 0.82, "learning_rate": 2.508338332333533e-06, "loss": 0.0004, "step": 6880 }, { "epoch": 0.82, "learning_rate": 2.508266346730654e-06, "loss": 0.0005, "step": 6881 }, { "epoch": 0.82, "learning_rate": 2.5081943611277742e-06, "loss": 0.7999, "step": 6882 }, { "epoch": 0.82, "learning_rate": 2.508122375524895e-06, "loss": 0.0005, "step": 6883 }, { "epoch": 0.82, "learning_rate": 2.508050389922016e-06, "loss": 0.0004, "step": 6884 }, { "epoch": 0.83, "learning_rate": 2.5079784043191366e-06, "loss": 0.0003, "step": 6885 }, { "epoch": 0.83, "learning_rate": 2.507906418716257e-06, "loss": 0.001, "step": 6886 }, { "epoch": 0.83, "learning_rate": 2.5078344331133773e-06, "loss": 0.0004, "step": 6887 }, { "epoch": 0.83, "learning_rate": 2.507762447510498e-06, "loss": 0.0004, "step": 6888 }, { "epoch": 0.83, "learning_rate": 2.5076904619076184e-06, "loss": 0.0005, "step": 6889 }, { "epoch": 0.83, "learning_rate": 2.507618476304739e-06, "loss": 0.0004, "step": 6890 }, { "epoch": 0.83, "learning_rate": 2.5075464907018596e-06, "loss": 0.0016, "step": 6891 }, { "epoch": 0.83, "learning_rate": 2.5074745050989803e-06, "loss": 0.0003, "step": 6892 }, { "epoch": 0.83, "learning_rate": 2.5074025194961007e-06, "loss": 0.0004, "step": 6893 }, { "epoch": 0.83, "learning_rate": 2.5073305338932215e-06, "loss": 0.0003, "step": 6894 }, { "epoch": 0.83, "learning_rate": 2.507258548290342e-06, "loss": 0.144, "step": 6895 }, { "epoch": 0.83, "learning_rate": 2.5071865626874626e-06, "loss": 0.7356, "step": 6896 }, { "epoch": 0.83, "learning_rate": 2.507114577084583e-06, "loss": 0.0003, "step": 6897 }, { "epoch": 0.83, "learning_rate": 2.5070425914817037e-06, "loss": 0.0004, "step": 6898 }, { "epoch": 0.83, "learning_rate": 2.5069706058788245e-06, "loss": 0.0003, "step": 6899 }, { "epoch": 0.83, "learning_rate": 2.506898620275945e-06, "loss": 0.0013, "step": 6900 }, { "epoch": 0.83, "learning_rate": 2.5068266346730656e-06, "loss": 0.0028, "step": 6901 }, { "epoch": 0.83, "learning_rate": 2.506754649070186e-06, "loss": 0.0005, "step": 6902 }, { "epoch": 0.83, "learning_rate": 2.5066826634673068e-06, "loss": 0.0004, "step": 6903 }, { "epoch": 0.83, "learning_rate": 2.506610677864427e-06, "loss": 0.0007, "step": 6904 }, { "epoch": 0.83, "learning_rate": 2.506538692261548e-06, "loss": 0.9697, "step": 6905 }, { "epoch": 0.83, "learning_rate": 2.5064667066586683e-06, "loss": 0.0004, "step": 6906 }, { "epoch": 0.83, "learning_rate": 2.506394721055789e-06, "loss": 0.0005, "step": 6907 }, { "epoch": 0.83, "learning_rate": 2.5063227354529094e-06, "loss": 0.0009, "step": 6908 }, { "epoch": 0.83, "learning_rate": 2.50625074985003e-06, "loss": 0.0022, "step": 6909 }, { "epoch": 0.83, "learning_rate": 2.5061787642471505e-06, "loss": 0.0108, "step": 6910 }, { "epoch": 0.83, "learning_rate": 2.506106778644271e-06, "loss": 0.7853, "step": 6911 }, { "epoch": 0.83, "learning_rate": 2.5060347930413917e-06, "loss": 0.0008, "step": 6912 }, { "epoch": 0.83, "learning_rate": 2.5059628074385124e-06, "loss": 0.0037, "step": 6913 }, { "epoch": 0.83, "learning_rate": 2.505890821835633e-06, "loss": 0.0005, "step": 6914 }, { "epoch": 0.83, "learning_rate": 2.5058188362327536e-06, "loss": 0.0007, "step": 6915 }, { "epoch": 0.83, "learning_rate": 2.5057468506298744e-06, "loss": 0.001, "step": 6916 }, { "epoch": 0.83, "learning_rate": 2.5056748650269947e-06, "loss": 0.0026, "step": 6917 }, { "epoch": 0.83, "learning_rate": 2.5056028794241155e-06, "loss": 0.0005, "step": 6918 }, { "epoch": 0.83, "learning_rate": 2.505530893821236e-06, "loss": 0.0005, "step": 6919 }, { "epoch": 0.83, "learning_rate": 2.5054589082183566e-06, "loss": 0.0079, "step": 6920 }, { "epoch": 0.83, "learning_rate": 2.505386922615477e-06, "loss": 0.0009, "step": 6921 }, { "epoch": 0.83, "learning_rate": 2.5053149370125973e-06, "loss": 0.0003, "step": 6922 }, { "epoch": 0.83, "learning_rate": 2.505242951409718e-06, "loss": 0.0006, "step": 6923 }, { "epoch": 0.83, "learning_rate": 2.5051709658068385e-06, "loss": 0.0006, "step": 6924 }, { "epoch": 0.83, "learning_rate": 2.5050989802039592e-06, "loss": 0.0005, "step": 6925 }, { "epoch": 0.83, "learning_rate": 2.5050269946010796e-06, "loss": 0.0004, "step": 6926 }, { "epoch": 0.83, "learning_rate": 2.5049550089982004e-06, "loss": 0.8115, "step": 6927 }, { "epoch": 0.83, "learning_rate": 2.5048830233953207e-06, "loss": 0.0005, "step": 6928 }, { "epoch": 0.83, "learning_rate": 2.5048110377924415e-06, "loss": 0.8535, "step": 6929 }, { "epoch": 0.83, "learning_rate": 2.5047390521895623e-06, "loss": 0.0003, "step": 6930 }, { "epoch": 0.83, "learning_rate": 2.504667066586683e-06, "loss": 0.0021, "step": 6931 }, { "epoch": 0.83, "learning_rate": 2.5045950809838034e-06, "loss": 0.0004, "step": 6932 }, { "epoch": 0.83, "learning_rate": 2.5045230953809238e-06, "loss": 0.0009, "step": 6933 }, { "epoch": 0.83, "learning_rate": 2.5044511097780446e-06, "loss": 0.0004, "step": 6934 }, { "epoch": 0.83, "learning_rate": 2.504379124175165e-06, "loss": 0.0005, "step": 6935 }, { "epoch": 0.83, "learning_rate": 2.5043071385722857e-06, "loss": 0.0006, "step": 6936 }, { "epoch": 0.83, "learning_rate": 2.504235152969406e-06, "loss": 0.0212, "step": 6937 }, { "epoch": 0.83, "learning_rate": 2.504163167366527e-06, "loss": 0.0008, "step": 6938 }, { "epoch": 0.83, "learning_rate": 2.504091181763647e-06, "loss": 0.3858, "step": 6939 }, { "epoch": 0.83, "learning_rate": 2.504019196160768e-06, "loss": 0.0005, "step": 6940 }, { "epoch": 0.83, "learning_rate": 2.5039472105578883e-06, "loss": 0.0004, "step": 6941 }, { "epoch": 0.83, "learning_rate": 2.503875224955009e-06, "loss": 0.0004, "step": 6942 }, { "epoch": 0.83, "learning_rate": 2.5038032393521294e-06, "loss": 0.0006, "step": 6943 }, { "epoch": 0.83, "learning_rate": 2.5037312537492502e-06, "loss": 0.0009, "step": 6944 }, { "epoch": 0.83, "learning_rate": 2.503659268146371e-06, "loss": 0.0004, "step": 6945 }, { "epoch": 0.83, "learning_rate": 2.5035872825434914e-06, "loss": 0.016, "step": 6946 }, { "epoch": 0.83, "learning_rate": 2.503515296940612e-06, "loss": 0.0007, "step": 6947 }, { "epoch": 0.83, "learning_rate": 2.5034433113377325e-06, "loss": 0.0002, "step": 6948 }, { "epoch": 0.83, "learning_rate": 2.5033713257348533e-06, "loss": 0.0006, "step": 6949 }, { "epoch": 0.83, "learning_rate": 2.5032993401319736e-06, "loss": 0.0006, "step": 6950 }, { "epoch": 0.83, "learning_rate": 2.5032273545290944e-06, "loss": 0.0009, "step": 6951 }, { "epoch": 0.83, "learning_rate": 2.5031553689262148e-06, "loss": 0.0009, "step": 6952 }, { "epoch": 0.83, "learning_rate": 2.5030833833233355e-06, "loss": 0.0004, "step": 6953 }, { "epoch": 0.83, "learning_rate": 2.503011397720456e-06, "loss": 0.001, "step": 6954 }, { "epoch": 0.83, "learning_rate": 2.5029394121175767e-06, "loss": 0.0007, "step": 6955 }, { "epoch": 0.83, "learning_rate": 2.502867426514697e-06, "loss": 0.0007, "step": 6956 }, { "epoch": 0.83, "learning_rate": 2.5027954409118174e-06, "loss": 0.0006, "step": 6957 }, { "epoch": 0.83, "learning_rate": 2.502723455308938e-06, "loss": 0.0007, "step": 6958 }, { "epoch": 0.83, "learning_rate": 2.502651469706059e-06, "loss": 0.0026, "step": 6959 }, { "epoch": 0.83, "learning_rate": 2.5025794841031793e-06, "loss": 0.0003, "step": 6960 }, { "epoch": 0.83, "learning_rate": 2.5025074985003e-06, "loss": 0.0007, "step": 6961 }, { "epoch": 0.83, "learning_rate": 2.502435512897421e-06, "loss": 0.0006, "step": 6962 }, { "epoch": 0.83, "learning_rate": 2.502363527294541e-06, "loss": 0.0004, "step": 6963 }, { "epoch": 0.83, "learning_rate": 2.502291541691662e-06, "loss": 0.0008, "step": 6964 }, { "epoch": 0.83, "learning_rate": 2.5022195560887823e-06, "loss": 0.0004, "step": 6965 }, { "epoch": 0.83, "learning_rate": 2.502147570485903e-06, "loss": 0.0007, "step": 6966 }, { "epoch": 0.83, "learning_rate": 2.5020755848830235e-06, "loss": 0.0004, "step": 6967 }, { "epoch": 0.83, "learning_rate": 2.502003599280144e-06, "loss": 0.5771, "step": 6968 }, { "epoch": 0.84, "learning_rate": 2.5019316136772646e-06, "loss": 0.0008, "step": 6969 }, { "epoch": 0.84, "learning_rate": 2.501859628074385e-06, "loss": 0.0011, "step": 6970 }, { "epoch": 0.84, "learning_rate": 2.5017876424715057e-06, "loss": 0.0004, "step": 6971 }, { "epoch": 0.84, "learning_rate": 2.501715656868626e-06, "loss": 0.0005, "step": 6972 }, { "epoch": 0.84, "learning_rate": 2.501643671265747e-06, "loss": 0.0005, "step": 6973 }, { "epoch": 0.84, "learning_rate": 2.5015716856628672e-06, "loss": 0.0008, "step": 6974 }, { "epoch": 0.84, "learning_rate": 2.501499700059988e-06, "loss": 0.7352, "step": 6975 }, { "epoch": 0.84, "learning_rate": 2.5014277144571088e-06, "loss": 0.0006, "step": 6976 }, { "epoch": 0.84, "learning_rate": 2.5013557288542296e-06, "loss": 0.001, "step": 6977 }, { "epoch": 0.84, "learning_rate": 2.50128374325135e-06, "loss": 0.0006, "step": 6978 }, { "epoch": 0.84, "learning_rate": 2.5012117576484703e-06, "loss": 0.0132, "step": 6979 }, { "epoch": 0.84, "learning_rate": 2.501139772045591e-06, "loss": 0.0004, "step": 6980 }, { "epoch": 0.84, "learning_rate": 2.5010677864427114e-06, "loss": 0.0005, "step": 6981 }, { "epoch": 0.84, "learning_rate": 2.500995800839832e-06, "loss": 0.0003, "step": 6982 }, { "epoch": 0.84, "learning_rate": 2.5009238152369525e-06, "loss": 0.0007, "step": 6983 }, { "epoch": 0.84, "learning_rate": 2.5008518296340733e-06, "loss": 0.0005, "step": 6984 }, { "epoch": 0.84, "learning_rate": 2.5007798440311937e-06, "loss": 0.0019, "step": 6985 }, { "epoch": 0.84, "learning_rate": 2.5007078584283145e-06, "loss": 0.0006, "step": 6986 }, { "epoch": 0.84, "learning_rate": 2.500635872825435e-06, "loss": 0.0007, "step": 6987 }, { "epoch": 0.84, "learning_rate": 2.5005638872225556e-06, "loss": 0.0004, "step": 6988 }, { "epoch": 0.84, "learning_rate": 2.500491901619676e-06, "loss": 0.0007, "step": 6989 }, { "epoch": 0.84, "learning_rate": 2.5004199160167967e-06, "loss": 0.0006, "step": 6990 }, { "epoch": 0.84, "learning_rate": 2.5003479304139175e-06, "loss": 0.0004, "step": 6991 }, { "epoch": 0.84, "learning_rate": 2.500275944811038e-06, "loss": 0.0009, "step": 6992 }, { "epoch": 0.84, "learning_rate": 2.5002039592081586e-06, "loss": 0.0004, "step": 6993 }, { "epoch": 0.84, "learning_rate": 2.500131973605279e-06, "loss": 0.0003, "step": 6994 }, { "epoch": 0.84, "learning_rate": 2.5000599880023998e-06, "loss": 0.0008, "step": 6995 }, { "epoch": 0.84, "learning_rate": 2.49998800239952e-06, "loss": 0.0013, "step": 6996 }, { "epoch": 0.84, "learning_rate": 2.499916016796641e-06, "loss": 0.0006, "step": 6997 }, { "epoch": 0.84, "learning_rate": 2.4998440311937613e-06, "loss": 0.0006, "step": 6998 }, { "epoch": 0.84, "learning_rate": 2.499772045590882e-06, "loss": 0.0004, "step": 6999 }, { "epoch": 0.84, "learning_rate": 2.4997000599880024e-06, "loss": 0.0013, "step": 7000 }, { "epoch": 0.84, "learning_rate": 2.499628074385123e-06, "loss": 0.0003, "step": 7001 }, { "epoch": 0.84, "learning_rate": 2.4995560887822435e-06, "loss": 1.1725, "step": 7002 }, { "epoch": 0.84, "learning_rate": 2.499484103179364e-06, "loss": 0.0006, "step": 7003 }, { "epoch": 0.84, "learning_rate": 2.4994121175764847e-06, "loss": 0.0007, "step": 7004 }, { "epoch": 0.84, "learning_rate": 2.4993401319736054e-06, "loss": 0.0005, "step": 7005 }, { "epoch": 0.84, "learning_rate": 2.499268146370726e-06, "loss": 0.0004, "step": 7006 }, { "epoch": 0.84, "learning_rate": 2.4991961607678466e-06, "loss": 0.0003, "step": 7007 }, { "epoch": 0.84, "learning_rate": 2.4991241751649673e-06, "loss": 0.0005, "step": 7008 }, { "epoch": 0.84, "learning_rate": 2.4990521895620877e-06, "loss": 0.0008, "step": 7009 }, { "epoch": 0.84, "learning_rate": 2.4989802039592085e-06, "loss": 0.0076, "step": 7010 }, { "epoch": 0.84, "learning_rate": 2.498908218356329e-06, "loss": 0.0021, "step": 7011 }, { "epoch": 0.84, "learning_rate": 2.4988362327534496e-06, "loss": 0.0005, "step": 7012 }, { "epoch": 0.84, "learning_rate": 2.49876424715057e-06, "loss": 0.0004, "step": 7013 }, { "epoch": 0.84, "learning_rate": 2.4986922615476903e-06, "loss": 0.0004, "step": 7014 }, { "epoch": 0.84, "learning_rate": 2.498620275944811e-06, "loss": 0.8872, "step": 7015 }, { "epoch": 0.84, "learning_rate": 2.4985482903419315e-06, "loss": 0.0007, "step": 7016 }, { "epoch": 0.84, "learning_rate": 2.4984763047390522e-06, "loss": 0.0007, "step": 7017 }, { "epoch": 0.84, "learning_rate": 2.4984043191361726e-06, "loss": 0.0004, "step": 7018 }, { "epoch": 0.84, "learning_rate": 2.4983323335332934e-06, "loss": 0.004, "step": 7019 }, { "epoch": 0.84, "learning_rate": 2.4982603479304137e-06, "loss": 0.0005, "step": 7020 }, { "epoch": 0.84, "learning_rate": 2.4981883623275345e-06, "loss": 0.0003, "step": 7021 }, { "epoch": 0.84, "learning_rate": 2.4981163767246553e-06, "loss": 0.0004, "step": 7022 }, { "epoch": 0.84, "learning_rate": 2.498044391121776e-06, "loss": 0.0006, "step": 7023 }, { "epoch": 0.84, "learning_rate": 2.4979724055188964e-06, "loss": 0.8096, "step": 7024 }, { "epoch": 0.84, "learning_rate": 2.4979004199160168e-06, "loss": 0.0006, "step": 7025 }, { "epoch": 0.84, "learning_rate": 2.4978284343131375e-06, "loss": 0.0007, "step": 7026 }, { "epoch": 0.84, "learning_rate": 2.497756448710258e-06, "loss": 0.4723, "step": 7027 }, { "epoch": 0.84, "learning_rate": 2.4976844631073787e-06, "loss": 0.0002, "step": 7028 }, { "epoch": 0.84, "learning_rate": 2.497612477504499e-06, "loss": 0.0008, "step": 7029 }, { "epoch": 0.84, "learning_rate": 2.49754049190162e-06, "loss": 0.0009, "step": 7030 }, { "epoch": 0.84, "learning_rate": 2.49746850629874e-06, "loss": 0.0005, "step": 7031 }, { "epoch": 0.84, "learning_rate": 2.497396520695861e-06, "loss": 0.0005, "step": 7032 }, { "epoch": 0.84, "learning_rate": 2.4973245350929813e-06, "loss": 0.0142, "step": 7033 }, { "epoch": 0.84, "learning_rate": 2.497252549490102e-06, "loss": 0.0003, "step": 7034 }, { "epoch": 0.84, "learning_rate": 2.4971805638872224e-06, "loss": 0.001, "step": 7035 }, { "epoch": 0.84, "learning_rate": 2.4971085782843432e-06, "loss": 0.0017, "step": 7036 }, { "epoch": 0.84, "learning_rate": 2.4970365926814636e-06, "loss": 0.9014, "step": 7037 }, { "epoch": 0.84, "learning_rate": 2.4969646070785844e-06, "loss": 0.0005, "step": 7038 }, { "epoch": 0.84, "learning_rate": 2.496892621475705e-06, "loss": 0.0008, "step": 7039 }, { "epoch": 0.84, "learning_rate": 2.4968206358728255e-06, "loss": 0.0007, "step": 7040 }, { "epoch": 0.84, "learning_rate": 2.4967486502699463e-06, "loss": 0.0004, "step": 7041 }, { "epoch": 0.84, "learning_rate": 2.4966766646670666e-06, "loss": 0.0005, "step": 7042 }, { "epoch": 0.84, "learning_rate": 2.4966046790641874e-06, "loss": 0.0006, "step": 7043 }, { "epoch": 0.84, "learning_rate": 2.4965326934613078e-06, "loss": 0.0004, "step": 7044 }, { "epoch": 0.84, "learning_rate": 2.4964607078584285e-06, "loss": 0.0005, "step": 7045 }, { "epoch": 0.84, "learning_rate": 2.496388722255549e-06, "loss": 0.0006, "step": 7046 }, { "epoch": 0.84, "learning_rate": 2.4963167366526697e-06, "loss": 0.0005, "step": 7047 }, { "epoch": 0.84, "learning_rate": 2.49624475104979e-06, "loss": 0.0003, "step": 7048 }, { "epoch": 0.84, "learning_rate": 2.4961727654469104e-06, "loss": 0.4037, "step": 7049 }, { "epoch": 0.84, "learning_rate": 2.496100779844031e-06, "loss": 0.3502, "step": 7050 }, { "epoch": 0.84, "learning_rate": 2.4960287942411515e-06, "loss": 0.0004, "step": 7051 }, { "epoch": 0.85, "learning_rate": 2.4959568086382723e-06, "loss": 0.0007, "step": 7052 }, { "epoch": 0.85, "learning_rate": 2.495884823035393e-06, "loss": 0.0006, "step": 7053 }, { "epoch": 0.85, "learning_rate": 2.495812837432514e-06, "loss": 0.0006, "step": 7054 }, { "epoch": 0.85, "learning_rate": 2.495740851829634e-06, "loss": 0.0006, "step": 7055 }, { "epoch": 0.85, "learning_rate": 2.495668866226755e-06, "loss": 0.0006, "step": 7056 }, { "epoch": 0.85, "learning_rate": 2.4955968806238753e-06, "loss": 0.0003, "step": 7057 }, { "epoch": 0.85, "learning_rate": 2.495524895020996e-06, "loss": 0.0005, "step": 7058 }, { "epoch": 0.85, "learning_rate": 2.4954529094181165e-06, "loss": 0.0003, "step": 7059 }, { "epoch": 0.85, "learning_rate": 2.495380923815237e-06, "loss": 0.0015, "step": 7060 }, { "epoch": 0.85, "learning_rate": 2.4953089382123576e-06, "loss": 0.0007, "step": 7061 }, { "epoch": 0.85, "learning_rate": 2.495236952609478e-06, "loss": 0.0004, "step": 7062 }, { "epoch": 0.85, "learning_rate": 2.4951649670065987e-06, "loss": 0.0005, "step": 7063 }, { "epoch": 0.85, "learning_rate": 2.495092981403719e-06, "loss": 0.331, "step": 7064 }, { "epoch": 0.85, "learning_rate": 2.49502099580084e-06, "loss": 0.0006, "step": 7065 }, { "epoch": 0.85, "learning_rate": 2.4949490101979602e-06, "loss": 0.0003, "step": 7066 }, { "epoch": 0.85, "learning_rate": 2.494877024595081e-06, "loss": 0.0008, "step": 7067 }, { "epoch": 0.85, "learning_rate": 2.4948050389922018e-06, "loss": 0.0005, "step": 7068 }, { "epoch": 0.85, "learning_rate": 2.494733053389322e-06, "loss": 0.8666, "step": 7069 }, { "epoch": 0.85, "learning_rate": 2.494661067786443e-06, "loss": 0.0004, "step": 7070 }, { "epoch": 0.85, "learning_rate": 2.4945890821835633e-06, "loss": 0.0009, "step": 7071 }, { "epoch": 0.85, "learning_rate": 2.494517096580684e-06, "loss": 0.5136, "step": 7072 }, { "epoch": 0.85, "learning_rate": 2.4944451109778044e-06, "loss": 0.0044, "step": 7073 }, { "epoch": 0.85, "learning_rate": 2.494373125374925e-06, "loss": 1.5075, "step": 7074 }, { "epoch": 0.85, "learning_rate": 2.4943011397720455e-06, "loss": 0.0019, "step": 7075 }, { "epoch": 0.85, "learning_rate": 2.4942291541691663e-06, "loss": 0.0004, "step": 7076 }, { "epoch": 0.85, "learning_rate": 2.4941571685662867e-06, "loss": 0.0005, "step": 7077 }, { "epoch": 0.85, "learning_rate": 2.4940851829634074e-06, "loss": 0.0005, "step": 7078 }, { "epoch": 0.85, "learning_rate": 2.494013197360528e-06, "loss": 0.0004, "step": 7079 }, { "epoch": 0.85, "learning_rate": 2.4939412117576486e-06, "loss": 0.0007, "step": 7080 }, { "epoch": 0.85, "learning_rate": 2.493869226154769e-06, "loss": 0.0004, "step": 7081 }, { "epoch": 0.85, "learning_rate": 2.4937972405518897e-06, "loss": 0.029, "step": 7082 }, { "epoch": 0.85, "learning_rate": 2.49372525494901e-06, "loss": 0.0003, "step": 7083 }, { "epoch": 0.85, "learning_rate": 2.493653269346131e-06, "loss": 0.1662, "step": 7084 }, { "epoch": 0.85, "learning_rate": 2.4935812837432516e-06, "loss": 0.0004, "step": 7085 }, { "epoch": 0.85, "learning_rate": 2.493509298140372e-06, "loss": 0.0006, "step": 7086 }, { "epoch": 0.85, "learning_rate": 2.4934373125374928e-06, "loss": 0.0006, "step": 7087 }, { "epoch": 0.85, "learning_rate": 2.493365326934613e-06, "loss": 0.8772, "step": 7088 }, { "epoch": 0.85, "learning_rate": 2.493293341331734e-06, "loss": 0.0007, "step": 7089 }, { "epoch": 0.85, "learning_rate": 2.4932213557288542e-06, "loss": 0.0005, "step": 7090 }, { "epoch": 0.85, "learning_rate": 2.493149370125975e-06, "loss": 0.0006, "step": 7091 }, { "epoch": 0.85, "learning_rate": 2.4930773845230954e-06, "loss": 0.1271, "step": 7092 }, { "epoch": 0.85, "learning_rate": 2.493005398920216e-06, "loss": 0.0005, "step": 7093 }, { "epoch": 0.85, "learning_rate": 2.4929334133173365e-06, "loss": 0.0005, "step": 7094 }, { "epoch": 0.85, "learning_rate": 2.492861427714457e-06, "loss": 0.0006, "step": 7095 }, { "epoch": 0.85, "learning_rate": 2.4927894421115776e-06, "loss": 0.0005, "step": 7096 }, { "epoch": 0.85, "learning_rate": 2.492717456508698e-06, "loss": 0.0006, "step": 7097 }, { "epoch": 0.85, "learning_rate": 2.4926454709058188e-06, "loss": 0.0008, "step": 7098 }, { "epoch": 0.85, "learning_rate": 2.4925734853029396e-06, "loss": 0.0004, "step": 7099 }, { "epoch": 0.85, "learning_rate": 2.4925014997000603e-06, "loss": 0.0006, "step": 7100 }, { "epoch": 0.85, "learning_rate": 2.4924295140971807e-06, "loss": 0.0005, "step": 7101 }, { "epoch": 0.85, "learning_rate": 2.4923575284943015e-06, "loss": 0.0094, "step": 7102 }, { "epoch": 0.85, "learning_rate": 2.492285542891422e-06, "loss": 0.0007, "step": 7103 }, { "epoch": 0.85, "learning_rate": 2.4922135572885426e-06, "loss": 0.0009, "step": 7104 }, { "epoch": 0.85, "learning_rate": 2.492141571685663e-06, "loss": 0.0004, "step": 7105 }, { "epoch": 0.85, "learning_rate": 2.4920695860827833e-06, "loss": 0.0004, "step": 7106 }, { "epoch": 0.85, "learning_rate": 2.491997600479904e-06, "loss": 0.0004, "step": 7107 }, { "epoch": 0.85, "learning_rate": 2.4919256148770245e-06, "loss": 0.0126, "step": 7108 }, { "epoch": 0.85, "learning_rate": 2.4918536292741452e-06, "loss": 0.0006, "step": 7109 }, { "epoch": 0.85, "learning_rate": 2.4917816436712656e-06, "loss": 0.0008, "step": 7110 }, { "epoch": 0.85, "learning_rate": 2.4917096580683864e-06, "loss": 0.3732, "step": 7111 }, { "epoch": 0.85, "learning_rate": 2.4916376724655067e-06, "loss": 0.0007, "step": 7112 }, { "epoch": 0.85, "learning_rate": 2.4915656868626275e-06, "loss": 0.0035, "step": 7113 }, { "epoch": 0.85, "learning_rate": 2.4914937012597483e-06, "loss": 0.0006, "step": 7114 }, { "epoch": 0.85, "learning_rate": 2.4914217156568686e-06, "loss": 0.0004, "step": 7115 }, { "epoch": 0.85, "learning_rate": 2.4913497300539894e-06, "loss": 0.001, "step": 7116 }, { "epoch": 0.85, "learning_rate": 2.49127774445111e-06, "loss": 0.6048, "step": 7117 }, { "epoch": 0.85, "learning_rate": 2.4912057588482305e-06, "loss": 0.0005, "step": 7118 }, { "epoch": 0.85, "learning_rate": 2.491133773245351e-06, "loss": 0.0361, "step": 7119 }, { "epoch": 0.85, "learning_rate": 2.4910617876424717e-06, "loss": 0.0058, "step": 7120 }, { "epoch": 0.85, "learning_rate": 2.490989802039592e-06, "loss": 0.0006, "step": 7121 }, { "epoch": 0.85, "learning_rate": 2.490917816436713e-06, "loss": 0.0006, "step": 7122 }, { "epoch": 0.85, "learning_rate": 2.490845830833833e-06, "loss": 0.0007, "step": 7123 }, { "epoch": 0.85, "learning_rate": 2.490773845230954e-06, "loss": 0.0007, "step": 7124 }, { "epoch": 0.85, "learning_rate": 2.4907018596280743e-06, "loss": 0.1029, "step": 7125 }, { "epoch": 0.85, "learning_rate": 2.490629874025195e-06, "loss": 0.0005, "step": 7126 }, { "epoch": 0.85, "learning_rate": 2.4905578884223154e-06, "loss": 0.0006, "step": 7127 }, { "epoch": 0.85, "learning_rate": 2.490485902819436e-06, "loss": 0.7317, "step": 7128 }, { "epoch": 0.85, "learning_rate": 2.4904139172165566e-06, "loss": 0.2004, "step": 7129 }, { "epoch": 0.85, "learning_rate": 2.4903419316136773e-06, "loss": 0.0004, "step": 7130 }, { "epoch": 0.85, "learning_rate": 2.490269946010798e-06, "loss": 0.0011, "step": 7131 }, { "epoch": 0.85, "learning_rate": 2.4901979604079185e-06, "loss": 0.0004, "step": 7132 }, { "epoch": 0.85, "learning_rate": 2.4901259748050393e-06, "loss": 0.0015, "step": 7133 }, { "epoch": 0.85, "learning_rate": 2.4900539892021596e-06, "loss": 0.0005, "step": 7134 }, { "epoch": 0.86, "learning_rate": 2.4899820035992804e-06, "loss": 0.4762, "step": 7135 }, { "epoch": 0.86, "learning_rate": 2.4899100179964007e-06, "loss": 1.0091, "step": 7136 }, { "epoch": 0.86, "learning_rate": 2.4898380323935215e-06, "loss": 0.0004, "step": 7137 }, { "epoch": 0.86, "learning_rate": 2.489766046790642e-06, "loss": 0.1181, "step": 7138 }, { "epoch": 0.86, "learning_rate": 2.4896940611877627e-06, "loss": 0.0004, "step": 7139 }, { "epoch": 0.86, "learning_rate": 2.489622075584883e-06, "loss": 0.0004, "step": 7140 }, { "epoch": 0.86, "learning_rate": 2.4895500899820034e-06, "loss": 0.0008, "step": 7141 }, { "epoch": 0.86, "learning_rate": 2.489478104379124e-06, "loss": 0.3192, "step": 7142 }, { "epoch": 0.86, "learning_rate": 2.4894061187762445e-06, "loss": 0.0005, "step": 7143 }, { "epoch": 0.86, "learning_rate": 2.4893341331733653e-06, "loss": 0.7193, "step": 7144 }, { "epoch": 0.86, "learning_rate": 2.489262147570486e-06, "loss": 0.4526, "step": 7145 }, { "epoch": 0.86, "learning_rate": 2.489190161967607e-06, "loss": 0.0005, "step": 7146 }, { "epoch": 0.86, "learning_rate": 2.489118176364727e-06, "loss": 0.2606, "step": 7147 }, { "epoch": 0.86, "learning_rate": 2.489046190761848e-06, "loss": 0.0004, "step": 7148 }, { "epoch": 0.86, "learning_rate": 2.4889742051589683e-06, "loss": 0.0011, "step": 7149 }, { "epoch": 0.86, "learning_rate": 2.488902219556089e-06, "loss": 0.001, "step": 7150 }, { "epoch": 0.86, "learning_rate": 2.4888302339532095e-06, "loss": 0.819, "step": 7151 }, { "epoch": 0.86, "learning_rate": 2.48875824835033e-06, "loss": 0.0008, "step": 7152 }, { "epoch": 0.86, "learning_rate": 2.4886862627474506e-06, "loss": 0.0006, "step": 7153 }, { "epoch": 0.86, "learning_rate": 2.488614277144571e-06, "loss": 0.0007, "step": 7154 }, { "epoch": 0.86, "learning_rate": 2.4885422915416917e-06, "loss": 0.0006, "step": 7155 }, { "epoch": 0.86, "learning_rate": 2.488470305938812e-06, "loss": 0.7571, "step": 7156 }, { "epoch": 0.86, "learning_rate": 2.488398320335933e-06, "loss": 0.0006, "step": 7157 }, { "epoch": 0.86, "learning_rate": 2.4883263347330532e-06, "loss": 0.001, "step": 7158 }, { "epoch": 0.86, "learning_rate": 2.488254349130174e-06, "loss": 0.001, "step": 7159 }, { "epoch": 0.86, "learning_rate": 2.4881823635272943e-06, "loss": 0.0004, "step": 7160 }, { "epoch": 0.86, "learning_rate": 2.488110377924415e-06, "loss": 0.0006, "step": 7161 }, { "epoch": 0.86, "learning_rate": 2.488038392321536e-06, "loss": 0.0007, "step": 7162 }, { "epoch": 0.86, "learning_rate": 2.4879664067186567e-06, "loss": 0.0037, "step": 7163 }, { "epoch": 0.86, "learning_rate": 2.487894421115777e-06, "loss": 0.6441, "step": 7164 }, { "epoch": 0.86, "learning_rate": 2.4878224355128974e-06, "loss": 0.0007, "step": 7165 }, { "epoch": 0.86, "learning_rate": 2.487750449910018e-06, "loss": 0.0227, "step": 7166 }, { "epoch": 0.86, "learning_rate": 2.4876784643071385e-06, "loss": 0.5108, "step": 7167 }, { "epoch": 0.86, "learning_rate": 2.4876064787042593e-06, "loss": 0.0005, "step": 7168 }, { "epoch": 0.86, "learning_rate": 2.4875344931013797e-06, "loss": 0.7251, "step": 7169 }, { "epoch": 0.86, "learning_rate": 2.4874625074985004e-06, "loss": 0.0005, "step": 7170 }, { "epoch": 0.86, "learning_rate": 2.487390521895621e-06, "loss": 0.0005, "step": 7171 }, { "epoch": 0.86, "learning_rate": 2.4873185362927416e-06, "loss": 0.0015, "step": 7172 }, { "epoch": 0.86, "learning_rate": 2.487246550689862e-06, "loss": 0.0005, "step": 7173 }, { "epoch": 0.86, "learning_rate": 2.4871745650869827e-06, "loss": 0.001, "step": 7174 }, { "epoch": 0.86, "learning_rate": 2.487102579484103e-06, "loss": 0.8689, "step": 7175 }, { "epoch": 0.86, "learning_rate": 2.487030593881224e-06, "loss": 0.0005, "step": 7176 }, { "epoch": 0.86, "learning_rate": 2.4869586082783446e-06, "loss": 0.0004, "step": 7177 }, { "epoch": 0.86, "learning_rate": 2.486886622675465e-06, "loss": 0.0004, "step": 7178 }, { "epoch": 0.86, "learning_rate": 2.4868146370725858e-06, "loss": 0.0005, "step": 7179 }, { "epoch": 0.86, "learning_rate": 2.486742651469706e-06, "loss": 0.0005, "step": 7180 }, { "epoch": 0.86, "learning_rate": 2.486670665866827e-06, "loss": 0.0007, "step": 7181 }, { "epoch": 0.86, "learning_rate": 2.4865986802639472e-06, "loss": 0.0004, "step": 7182 }, { "epoch": 0.86, "learning_rate": 2.486526694661068e-06, "loss": 0.0012, "step": 7183 }, { "epoch": 0.86, "learning_rate": 2.4864547090581884e-06, "loss": 0.0007, "step": 7184 }, { "epoch": 0.86, "learning_rate": 2.486382723455309e-06, "loss": 0.0008, "step": 7185 }, { "epoch": 0.86, "learning_rate": 2.4863107378524295e-06, "loss": 0.0007, "step": 7186 }, { "epoch": 0.86, "learning_rate": 2.48623875224955e-06, "loss": 0.4172, "step": 7187 }, { "epoch": 0.86, "learning_rate": 2.4861667666466706e-06, "loss": 0.0022, "step": 7188 }, { "epoch": 0.86, "learning_rate": 2.486094781043791e-06, "loss": 0.0005, "step": 7189 }, { "epoch": 0.86, "learning_rate": 2.4860227954409118e-06, "loss": 1.6016, "step": 7190 }, { "epoch": 0.86, "learning_rate": 2.4859508098380326e-06, "loss": 0.0004, "step": 7191 }, { "epoch": 0.86, "learning_rate": 2.485878824235153e-06, "loss": 0.0009, "step": 7192 }, { "epoch": 0.86, "learning_rate": 2.4858068386322737e-06, "loss": 0.0014, "step": 7193 }, { "epoch": 0.86, "learning_rate": 2.4857348530293945e-06, "loss": 0.0005, "step": 7194 }, { "epoch": 0.86, "learning_rate": 2.485662867426515e-06, "loss": 0.9758, "step": 7195 }, { "epoch": 0.86, "learning_rate": 2.4855908818236356e-06, "loss": 0.0007, "step": 7196 }, { "epoch": 0.86, "learning_rate": 2.485518896220756e-06, "loss": 0.0004, "step": 7197 }, { "epoch": 0.86, "learning_rate": 2.4854469106178763e-06, "loss": 0.0007, "step": 7198 }, { "epoch": 0.86, "learning_rate": 2.485374925014997e-06, "loss": 0.8647, "step": 7199 }, { "epoch": 0.86, "learning_rate": 2.4853029394121174e-06, "loss": 0.0008, "step": 7200 }, { "epoch": 0.86, "learning_rate": 2.4852309538092382e-06, "loss": 0.9599, "step": 7201 }, { "epoch": 0.86, "learning_rate": 2.4851589682063586e-06, "loss": 0.0007, "step": 7202 }, { "epoch": 0.86, "learning_rate": 2.4850869826034794e-06, "loss": 0.0005, "step": 7203 }, { "epoch": 0.86, "learning_rate": 2.4850149970005997e-06, "loss": 0.0019, "step": 7204 }, { "epoch": 0.86, "learning_rate": 2.4849430113977205e-06, "loss": 0.0005, "step": 7205 }, { "epoch": 0.86, "learning_rate": 2.484871025794841e-06, "loss": 0.0994, "step": 7206 }, { "epoch": 0.86, "learning_rate": 2.4847990401919616e-06, "loss": 0.0023, "step": 7207 }, { "epoch": 0.86, "learning_rate": 2.4847270545890824e-06, "loss": 0.0006, "step": 7208 }, { "epoch": 0.86, "learning_rate": 2.484655068986203e-06, "loss": 0.0007, "step": 7209 }, { "epoch": 0.86, "learning_rate": 2.4845830833833235e-06, "loss": 0.0018, "step": 7210 }, { "epoch": 0.86, "learning_rate": 2.484511097780444e-06, "loss": 0.0007, "step": 7211 }, { "epoch": 0.86, "learning_rate": 2.4844391121775647e-06, "loss": 0.001, "step": 7212 }, { "epoch": 0.86, "learning_rate": 2.484367126574685e-06, "loss": 0.0007, "step": 7213 }, { "epoch": 0.86, "learning_rate": 2.484295140971806e-06, "loss": 0.0007, "step": 7214 }, { "epoch": 0.86, "learning_rate": 2.484223155368926e-06, "loss": 0.0005, "step": 7215 }, { "epoch": 0.86, "learning_rate": 2.484151169766047e-06, "loss": 0.001, "step": 7216 }, { "epoch": 0.86, "learning_rate": 2.4840791841631673e-06, "loss": 0.0008, "step": 7217 }, { "epoch": 0.86, "learning_rate": 2.484007198560288e-06, "loss": 0.0038, "step": 7218 }, { "epoch": 0.87, "learning_rate": 2.4839352129574084e-06, "loss": 0.0017, "step": 7219 }, { "epoch": 0.87, "learning_rate": 2.483863227354529e-06, "loss": 0.0014, "step": 7220 }, { "epoch": 0.87, "learning_rate": 2.4837912417516496e-06, "loss": 0.0007, "step": 7221 }, { "epoch": 0.87, "learning_rate": 2.4837192561487703e-06, "loss": 0.0005, "step": 7222 }, { "epoch": 0.87, "learning_rate": 2.483647270545891e-06, "loss": 0.0006, "step": 7223 }, { "epoch": 0.87, "learning_rate": 2.4835752849430115e-06, "loss": 0.0004, "step": 7224 }, { "epoch": 0.87, "learning_rate": 2.4835032993401322e-06, "loss": 0.0006, "step": 7225 }, { "epoch": 0.87, "learning_rate": 2.4834313137372526e-06, "loss": 0.0008, "step": 7226 }, { "epoch": 0.87, "learning_rate": 2.4833593281343734e-06, "loss": 0.0007, "step": 7227 }, { "epoch": 0.87, "learning_rate": 2.4832873425314937e-06, "loss": 0.0009, "step": 7228 }, { "epoch": 0.87, "learning_rate": 2.4832153569286145e-06, "loss": 0.0344, "step": 7229 }, { "epoch": 0.87, "learning_rate": 2.483143371325735e-06, "loss": 0.0005, "step": 7230 }, { "epoch": 0.87, "learning_rate": 2.4830713857228556e-06, "loss": 0.0061, "step": 7231 }, { "epoch": 0.87, "learning_rate": 2.482999400119976e-06, "loss": 0.0007, "step": 7232 }, { "epoch": 0.87, "learning_rate": 2.4829274145170964e-06, "loss": 0.002, "step": 7233 }, { "epoch": 0.87, "learning_rate": 2.482855428914217e-06, "loss": 0.0006, "step": 7234 }, { "epoch": 0.87, "learning_rate": 2.4827834433113375e-06, "loss": 0.0008, "step": 7235 }, { "epoch": 0.87, "learning_rate": 2.4827114577084583e-06, "loss": 0.0009, "step": 7236 }, { "epoch": 0.87, "learning_rate": 2.482639472105579e-06, "loss": 0.0007, "step": 7237 }, { "epoch": 0.87, "learning_rate": 2.4825674865026994e-06, "loss": 0.0007, "step": 7238 }, { "epoch": 0.87, "learning_rate": 2.48249550089982e-06, "loss": 0.0009, "step": 7239 }, { "epoch": 0.87, "learning_rate": 2.482423515296941e-06, "loss": 0.0011, "step": 7240 }, { "epoch": 0.87, "learning_rate": 2.4823515296940613e-06, "loss": 0.0022, "step": 7241 }, { "epoch": 0.87, "learning_rate": 2.482279544091182e-06, "loss": 0.0004, "step": 7242 }, { "epoch": 0.87, "learning_rate": 2.4822075584883024e-06, "loss": 0.0006, "step": 7243 }, { "epoch": 0.87, "learning_rate": 2.4821355728854232e-06, "loss": 0.0006, "step": 7244 }, { "epoch": 0.87, "learning_rate": 2.4820635872825436e-06, "loss": 0.046, "step": 7245 }, { "epoch": 0.87, "learning_rate": 2.481991601679664e-06, "loss": 0.0006, "step": 7246 }, { "epoch": 0.87, "learning_rate": 2.4819196160767847e-06, "loss": 0.0051, "step": 7247 }, { "epoch": 0.87, "learning_rate": 2.481847630473905e-06, "loss": 0.0007, "step": 7248 }, { "epoch": 0.87, "learning_rate": 2.481775644871026e-06, "loss": 0.0008, "step": 7249 }, { "epoch": 0.87, "learning_rate": 2.481703659268146e-06, "loss": 0.001, "step": 7250 }, { "epoch": 0.87, "learning_rate": 2.481631673665267e-06, "loss": 0.5177, "step": 7251 }, { "epoch": 0.87, "learning_rate": 2.4815596880623873e-06, "loss": 0.0005, "step": 7252 }, { "epoch": 0.87, "learning_rate": 2.481487702459508e-06, "loss": 0.0007, "step": 7253 }, { "epoch": 0.87, "learning_rate": 2.481415716856629e-06, "loss": 0.0008, "step": 7254 }, { "epoch": 0.87, "learning_rate": 2.4813437312537497e-06, "loss": 0.0004, "step": 7255 }, { "epoch": 0.87, "learning_rate": 2.48127174565087e-06, "loss": 0.0007, "step": 7256 }, { "epoch": 0.87, "learning_rate": 2.4811997600479904e-06, "loss": 0.0006, "step": 7257 }, { "epoch": 0.87, "learning_rate": 2.481127774445111e-06, "loss": 0.0007, "step": 7258 }, { "epoch": 0.87, "learning_rate": 2.4810557888422315e-06, "loss": 0.0005, "step": 7259 }, { "epoch": 0.87, "learning_rate": 2.4809838032393523e-06, "loss": 0.001, "step": 7260 }, { "epoch": 0.87, "learning_rate": 2.4809118176364727e-06, "loss": 0.0003, "step": 7261 }, { "epoch": 0.87, "learning_rate": 2.4808398320335934e-06, "loss": 0.0006, "step": 7262 }, { "epoch": 0.87, "learning_rate": 2.4807678464307138e-06, "loss": 0.0004, "step": 7263 }, { "epoch": 0.87, "learning_rate": 2.4806958608278346e-06, "loss": 0.0005, "step": 7264 }, { "epoch": 0.87, "learning_rate": 2.480623875224955e-06, "loss": 0.0029, "step": 7265 }, { "epoch": 0.87, "learning_rate": 2.4805518896220757e-06, "loss": 0.5046, "step": 7266 }, { "epoch": 0.87, "learning_rate": 2.480479904019196e-06, "loss": 0.0006, "step": 7267 }, { "epoch": 0.87, "learning_rate": 2.480407918416317e-06, "loss": 0.0006, "step": 7268 }, { "epoch": 0.87, "learning_rate": 2.4803359328134376e-06, "loss": 0.0011, "step": 7269 }, { "epoch": 0.87, "learning_rate": 2.480263947210558e-06, "loss": 0.0004, "step": 7270 }, { "epoch": 0.87, "learning_rate": 2.4801919616076787e-06, "loss": 0.0011, "step": 7271 }, { "epoch": 0.87, "learning_rate": 2.480119976004799e-06, "loss": 0.0189, "step": 7272 }, { "epoch": 0.87, "learning_rate": 2.48004799040192e-06, "loss": 0.1034, "step": 7273 }, { "epoch": 0.87, "learning_rate": 2.4799760047990402e-06, "loss": 0.7402, "step": 7274 }, { "epoch": 0.87, "learning_rate": 2.479904019196161e-06, "loss": 0.0007, "step": 7275 }, { "epoch": 0.87, "learning_rate": 2.4798320335932814e-06, "loss": 0.0025, "step": 7276 }, { "epoch": 0.87, "learning_rate": 2.479760047990402e-06, "loss": 0.0005, "step": 7277 }, { "epoch": 0.87, "learning_rate": 2.4796880623875225e-06, "loss": 0.0004, "step": 7278 }, { "epoch": 0.87, "learning_rate": 2.479616076784643e-06, "loss": 0.0016, "step": 7279 }, { "epoch": 0.87, "learning_rate": 2.4795440911817636e-06, "loss": 0.0005, "step": 7280 }, { "epoch": 0.87, "learning_rate": 2.479472105578884e-06, "loss": 0.0004, "step": 7281 }, { "epoch": 0.87, "learning_rate": 2.4794001199760048e-06, "loss": 0.4327, "step": 7282 }, { "epoch": 0.87, "learning_rate": 2.479328134373125e-06, "loss": 0.0012, "step": 7283 }, { "epoch": 0.87, "learning_rate": 2.479256148770246e-06, "loss": 0.0008, "step": 7284 }, { "epoch": 0.87, "learning_rate": 2.4791841631673667e-06, "loss": 0.2983, "step": 7285 }, { "epoch": 0.87, "learning_rate": 2.4791121775644875e-06, "loss": 0.0004, "step": 7286 }, { "epoch": 0.87, "learning_rate": 2.479040191961608e-06, "loss": 0.0005, "step": 7287 }, { "epoch": 0.87, "learning_rate": 2.4789682063587286e-06, "loss": 0.0004, "step": 7288 }, { "epoch": 0.87, "learning_rate": 2.478896220755849e-06, "loss": 0.001, "step": 7289 }, { "epoch": 0.87, "learning_rate": 2.4788242351529697e-06, "loss": 0.0007, "step": 7290 }, { "epoch": 0.87, "learning_rate": 2.47875224955009e-06, "loss": 0.0004, "step": 7291 }, { "epoch": 0.87, "learning_rate": 2.4786802639472104e-06, "loss": 0.004, "step": 7292 }, { "epoch": 0.87, "learning_rate": 2.4786082783443312e-06, "loss": 0.0006, "step": 7293 }, { "epoch": 0.87, "learning_rate": 2.4785362927414516e-06, "loss": 0.0014, "step": 7294 }, { "epoch": 0.87, "learning_rate": 2.4784643071385723e-06, "loss": 0.0192, "step": 7295 }, { "epoch": 0.87, "learning_rate": 2.4783923215356927e-06, "loss": 0.0004, "step": 7296 }, { "epoch": 0.87, "learning_rate": 2.4783203359328135e-06, "loss": 0.0005, "step": 7297 }, { "epoch": 0.87, "learning_rate": 2.478248350329934e-06, "loss": 0.0006, "step": 7298 }, { "epoch": 0.87, "learning_rate": 2.4781763647270546e-06, "loss": 0.0005, "step": 7299 }, { "epoch": 0.87, "learning_rate": 2.4781043791241754e-06, "loss": 0.0004, "step": 7300 }, { "epoch": 0.87, "learning_rate": 2.4780323935212957e-06, "loss": 0.0003, "step": 7301 }, { "epoch": 0.88, "learning_rate": 2.4779604079184165e-06, "loss": 0.0005, "step": 7302 }, { "epoch": 0.88, "learning_rate": 2.477888422315537e-06, "loss": 0.039, "step": 7303 }, { "epoch": 0.88, "learning_rate": 2.4778164367126577e-06, "loss": 0.0046, "step": 7304 }, { "epoch": 0.88, "learning_rate": 2.477744451109778e-06, "loss": 0.0008, "step": 7305 }, { "epoch": 0.88, "learning_rate": 2.477672465506899e-06, "loss": 0.0005, "step": 7306 }, { "epoch": 0.88, "learning_rate": 2.477600479904019e-06, "loss": 0.0016, "step": 7307 }, { "epoch": 0.88, "learning_rate": 2.47752849430114e-06, "loss": 0.0006, "step": 7308 }, { "epoch": 0.88, "learning_rate": 2.4774565086982603e-06, "loss": 0.0009, "step": 7309 }, { "epoch": 0.88, "learning_rate": 2.477384523095381e-06, "loss": 0.0009, "step": 7310 }, { "epoch": 0.88, "learning_rate": 2.4773125374925014e-06, "loss": 0.0007, "step": 7311 }, { "epoch": 0.88, "learning_rate": 2.477240551889622e-06, "loss": 0.0004, "step": 7312 }, { "epoch": 0.88, "learning_rate": 2.4771685662867425e-06, "loss": 0.0004, "step": 7313 }, { "epoch": 0.88, "learning_rate": 2.4770965806838633e-06, "loss": 0.0005, "step": 7314 }, { "epoch": 0.88, "learning_rate": 2.4770245950809837e-06, "loss": 0.0004, "step": 7315 }, { "epoch": 0.88, "learning_rate": 2.4769526094781045e-06, "loss": 0.5795, "step": 7316 }, { "epoch": 0.88, "learning_rate": 2.4768806238752252e-06, "loss": 0.0006, "step": 7317 }, { "epoch": 0.88, "learning_rate": 2.4768086382723456e-06, "loss": 0.0007, "step": 7318 }, { "epoch": 0.88, "learning_rate": 2.4767366526694664e-06, "loss": 0.0003, "step": 7319 }, { "epoch": 0.88, "learning_rate": 2.4766646670665867e-06, "loss": 0.5773, "step": 7320 }, { "epoch": 0.88, "learning_rate": 2.4765926814637075e-06, "loss": 0.0067, "step": 7321 }, { "epoch": 0.88, "learning_rate": 2.476520695860828e-06, "loss": 0.0006, "step": 7322 }, { "epoch": 0.88, "learning_rate": 2.4764487102579486e-06, "loss": 0.0448, "step": 7323 }, { "epoch": 0.88, "learning_rate": 2.476376724655069e-06, "loss": 0.0004, "step": 7324 }, { "epoch": 0.88, "learning_rate": 2.4763047390521894e-06, "loss": 0.8156, "step": 7325 }, { "epoch": 0.88, "learning_rate": 2.47623275344931e-06, "loss": 0.624, "step": 7326 }, { "epoch": 0.88, "learning_rate": 2.4761607678464305e-06, "loss": 0.346, "step": 7327 }, { "epoch": 0.88, "learning_rate": 2.4760887822435513e-06, "loss": 0.0024, "step": 7328 }, { "epoch": 0.88, "learning_rate": 2.4760167966406716e-06, "loss": 0.0287, "step": 7329 }, { "epoch": 0.88, "learning_rate": 2.4759448110377924e-06, "loss": 0.0005, "step": 7330 }, { "epoch": 0.88, "learning_rate": 2.475872825434913e-06, "loss": 0.001, "step": 7331 }, { "epoch": 0.88, "learning_rate": 2.475800839832034e-06, "loss": 0.001, "step": 7332 }, { "epoch": 0.88, "learning_rate": 2.4757288542291543e-06, "loss": 0.0013, "step": 7333 }, { "epoch": 0.88, "learning_rate": 2.475656868626275e-06, "loss": 0.0006, "step": 7334 }, { "epoch": 0.88, "learning_rate": 2.4755848830233954e-06, "loss": 0.001, "step": 7335 }, { "epoch": 0.88, "learning_rate": 2.4755128974205162e-06, "loss": 0.0027, "step": 7336 }, { "epoch": 0.88, "learning_rate": 2.4754409118176366e-06, "loss": 0.4209, "step": 7337 }, { "epoch": 0.88, "learning_rate": 2.475368926214757e-06, "loss": 0.0005, "step": 7338 }, { "epoch": 0.88, "learning_rate": 2.4752969406118777e-06, "loss": 0.0005, "step": 7339 }, { "epoch": 0.88, "learning_rate": 2.475224955008998e-06, "loss": 0.0004, "step": 7340 }, { "epoch": 0.88, "learning_rate": 2.475152969406119e-06, "loss": 0.0005, "step": 7341 }, { "epoch": 0.88, "learning_rate": 2.475080983803239e-06, "loss": 0.0004, "step": 7342 }, { "epoch": 0.88, "learning_rate": 2.47500899820036e-06, "loss": 0.4258, "step": 7343 }, { "epoch": 0.88, "learning_rate": 2.4749370125974803e-06, "loss": 0.0012, "step": 7344 }, { "epoch": 0.88, "learning_rate": 2.474865026994601e-06, "loss": 0.1134, "step": 7345 }, { "epoch": 0.88, "learning_rate": 2.474793041391722e-06, "loss": 0.0118, "step": 7346 }, { "epoch": 0.88, "learning_rate": 2.4747210557888422e-06, "loss": 0.0005, "step": 7347 }, { "epoch": 0.88, "learning_rate": 2.474649070185963e-06, "loss": 0.0007, "step": 7348 }, { "epoch": 0.88, "learning_rate": 2.4745770845830834e-06, "loss": 0.0003, "step": 7349 }, { "epoch": 0.88, "learning_rate": 2.474505098980204e-06, "loss": 0.0016, "step": 7350 }, { "epoch": 0.88, "learning_rate": 2.4744331133773245e-06, "loss": 0.0005, "step": 7351 }, { "epoch": 0.88, "learning_rate": 2.4743611277744453e-06, "loss": 0.0003, "step": 7352 }, { "epoch": 0.88, "learning_rate": 2.4742891421715656e-06, "loss": 0.0008, "step": 7353 }, { "epoch": 0.88, "learning_rate": 2.4742171565686864e-06, "loss": 0.0012, "step": 7354 }, { "epoch": 0.88, "learning_rate": 2.4741451709658068e-06, "loss": 0.0005, "step": 7355 }, { "epoch": 0.88, "learning_rate": 2.4740731853629276e-06, "loss": 0.0025, "step": 7356 }, { "epoch": 0.88, "learning_rate": 2.474001199760048e-06, "loss": 0.7894, "step": 7357 }, { "epoch": 0.88, "learning_rate": 2.4739292141571687e-06, "loss": 0.0031, "step": 7358 }, { "epoch": 0.88, "learning_rate": 2.473857228554289e-06, "loss": 0.0006, "step": 7359 }, { "epoch": 0.88, "learning_rate": 2.47378524295141e-06, "loss": 0.0005, "step": 7360 }, { "epoch": 0.88, "learning_rate": 2.47371325734853e-06, "loss": 0.0005, "step": 7361 }, { "epoch": 0.88, "learning_rate": 2.473641271745651e-06, "loss": 0.0004, "step": 7362 }, { "epoch": 0.88, "learning_rate": 2.4735692861427717e-06, "loss": 0.0005, "step": 7363 }, { "epoch": 0.88, "learning_rate": 2.473497300539892e-06, "loss": 0.0003, "step": 7364 }, { "epoch": 0.88, "learning_rate": 2.473425314937013e-06, "loss": 0.0004, "step": 7365 }, { "epoch": 0.88, "learning_rate": 2.4733533293341332e-06, "loss": 0.002, "step": 7366 }, { "epoch": 0.88, "learning_rate": 2.473281343731254e-06, "loss": 0.0006, "step": 7367 }, { "epoch": 0.88, "learning_rate": 2.4732093581283744e-06, "loss": 0.0004, "step": 7368 }, { "epoch": 0.88, "learning_rate": 2.473137372525495e-06, "loss": 0.0006, "step": 7369 }, { "epoch": 0.88, "learning_rate": 2.4730653869226155e-06, "loss": 0.001, "step": 7370 }, { "epoch": 0.88, "learning_rate": 2.4729934013197363e-06, "loss": 0.0004, "step": 7371 }, { "epoch": 0.88, "learning_rate": 2.4729214157168566e-06, "loss": 0.0005, "step": 7372 }, { "epoch": 0.88, "learning_rate": 2.472849430113977e-06, "loss": 0.0006, "step": 7373 }, { "epoch": 0.88, "learning_rate": 2.4727774445110978e-06, "loss": 0.0006, "step": 7374 }, { "epoch": 0.88, "learning_rate": 2.472705458908218e-06, "loss": 0.0007, "step": 7375 }, { "epoch": 0.88, "learning_rate": 2.472633473305339e-06, "loss": 0.0007, "step": 7376 }, { "epoch": 0.88, "learning_rate": 2.4725614877024597e-06, "loss": 0.0005, "step": 7377 }, { "epoch": 0.88, "learning_rate": 2.4724895020995804e-06, "loss": 0.0005, "step": 7378 }, { "epoch": 0.88, "learning_rate": 2.472417516496701e-06, "loss": 0.0009, "step": 7379 }, { "epoch": 0.88, "learning_rate": 2.4723455308938216e-06, "loss": 0.0003, "step": 7380 }, { "epoch": 0.88, "learning_rate": 2.472273545290942e-06, "loss": 0.0003, "step": 7381 }, { "epoch": 0.88, "learning_rate": 2.4722015596880627e-06, "loss": 0.0006, "step": 7382 }, { "epoch": 0.88, "learning_rate": 2.472129574085183e-06, "loss": 0.0007, "step": 7383 }, { "epoch": 0.88, "learning_rate": 2.4720575884823034e-06, "loss": 0.0003, "step": 7384 }, { "epoch": 0.88, "learning_rate": 2.471985602879424e-06, "loss": 0.4336, "step": 7385 }, { "epoch": 0.89, "learning_rate": 2.4719136172765446e-06, "loss": 0.0005, "step": 7386 }, { "epoch": 0.89, "learning_rate": 2.4718416316736653e-06, "loss": 0.0007, "step": 7387 }, { "epoch": 0.89, "learning_rate": 2.4717696460707857e-06, "loss": 0.0004, "step": 7388 }, { "epoch": 0.89, "learning_rate": 2.4716976604679065e-06, "loss": 0.0005, "step": 7389 }, { "epoch": 0.89, "learning_rate": 2.471625674865027e-06, "loss": 0.0005, "step": 7390 }, { "epoch": 0.89, "learning_rate": 2.4715536892621476e-06, "loss": 0.0009, "step": 7391 }, { "epoch": 0.89, "learning_rate": 2.4714817036592684e-06, "loss": 0.0005, "step": 7392 }, { "epoch": 0.89, "learning_rate": 2.4714097180563887e-06, "loss": 0.0004, "step": 7393 }, { "epoch": 0.89, "learning_rate": 2.4713377324535095e-06, "loss": 0.3432, "step": 7394 }, { "epoch": 0.89, "learning_rate": 2.47126574685063e-06, "loss": 0.0006, "step": 7395 }, { "epoch": 0.89, "learning_rate": 2.4711937612477507e-06, "loss": 0.0005, "step": 7396 }, { "epoch": 0.89, "learning_rate": 2.471121775644871e-06, "loss": 0.0003, "step": 7397 }, { "epoch": 0.89, "learning_rate": 2.4710497900419918e-06, "loss": 0.0007, "step": 7398 }, { "epoch": 0.89, "learning_rate": 2.470977804439112e-06, "loss": 0.0067, "step": 7399 }, { "epoch": 0.89, "learning_rate": 2.470905818836233e-06, "loss": 0.0003, "step": 7400 }, { "epoch": 0.89, "learning_rate": 2.4708338332333533e-06, "loss": 0.0005, "step": 7401 }, { "epoch": 0.89, "learning_rate": 2.470761847630474e-06, "loss": 0.0004, "step": 7402 }, { "epoch": 0.89, "learning_rate": 2.4706898620275944e-06, "loss": 0.0009, "step": 7403 }, { "epoch": 0.89, "learning_rate": 2.470617876424715e-06, "loss": 1.6518, "step": 7404 }, { "epoch": 0.89, "learning_rate": 2.4705458908218355e-06, "loss": 0.8813, "step": 7405 }, { "epoch": 0.89, "learning_rate": 2.4704739052189563e-06, "loss": 0.0005, "step": 7406 }, { "epoch": 0.89, "learning_rate": 2.4704019196160767e-06, "loss": 0.0003, "step": 7407 }, { "epoch": 0.89, "learning_rate": 2.4703299340131975e-06, "loss": 0.0004, "step": 7408 }, { "epoch": 0.89, "learning_rate": 2.4702579484103182e-06, "loss": 0.0012, "step": 7409 }, { "epoch": 0.89, "learning_rate": 2.4701859628074386e-06, "loss": 0.0046, "step": 7410 }, { "epoch": 0.89, "learning_rate": 2.4701139772045594e-06, "loss": 0.8104, "step": 7411 }, { "epoch": 0.89, "learning_rate": 2.4700419916016797e-06, "loss": 0.0006, "step": 7412 }, { "epoch": 0.89, "learning_rate": 2.4699700059988005e-06, "loss": 0.0003, "step": 7413 }, { "epoch": 0.89, "learning_rate": 2.469898020395921e-06, "loss": 0.0004, "step": 7414 }, { "epoch": 0.89, "learning_rate": 2.4698260347930416e-06, "loss": 0.0008, "step": 7415 }, { "epoch": 0.89, "learning_rate": 2.469754049190162e-06, "loss": 0.0011, "step": 7416 }, { "epoch": 0.89, "learning_rate": 2.4696820635872828e-06, "loss": 0.0009, "step": 7417 }, { "epoch": 0.89, "learning_rate": 2.469610077984403e-06, "loss": 0.2553, "step": 7418 }, { "epoch": 0.89, "learning_rate": 2.4695380923815235e-06, "loss": 0.0038, "step": 7419 }, { "epoch": 0.89, "learning_rate": 2.4694661067786443e-06, "loss": 0.0004, "step": 7420 }, { "epoch": 0.89, "learning_rate": 2.4693941211757646e-06, "loss": 0.0031, "step": 7421 }, { "epoch": 0.89, "learning_rate": 2.4693221355728854e-06, "loss": 0.0003, "step": 7422 }, { "epoch": 0.89, "learning_rate": 2.469250149970006e-06, "loss": 0.0005, "step": 7423 }, { "epoch": 0.89, "learning_rate": 2.469178164367127e-06, "loss": 0.6621, "step": 7424 }, { "epoch": 0.89, "learning_rate": 2.4691061787642473e-06, "loss": 0.9541, "step": 7425 }, { "epoch": 0.89, "learning_rate": 2.469034193161368e-06, "loss": 0.0004, "step": 7426 }, { "epoch": 0.89, "learning_rate": 2.4689622075584884e-06, "loss": 0.0006, "step": 7427 }, { "epoch": 0.89, "learning_rate": 2.4688902219556092e-06, "loss": 0.7863, "step": 7428 }, { "epoch": 0.89, "learning_rate": 2.4688182363527296e-06, "loss": 0.0004, "step": 7429 }, { "epoch": 0.89, "learning_rate": 2.46874625074985e-06, "loss": 0.0004, "step": 7430 }, { "epoch": 0.89, "learning_rate": 2.4686742651469707e-06, "loss": 0.001, "step": 7431 }, { "epoch": 0.89, "learning_rate": 2.468602279544091e-06, "loss": 0.0061, "step": 7432 }, { "epoch": 0.89, "learning_rate": 2.468530293941212e-06, "loss": 0.0004, "step": 7433 }, { "epoch": 0.89, "learning_rate": 2.468458308338332e-06, "loss": 0.0004, "step": 7434 }, { "epoch": 0.89, "learning_rate": 2.468386322735453e-06, "loss": 0.0007, "step": 7435 }, { "epoch": 0.89, "learning_rate": 2.4683143371325733e-06, "loss": 0.0006, "step": 7436 }, { "epoch": 0.89, "learning_rate": 2.468242351529694e-06, "loss": 0.0006, "step": 7437 }, { "epoch": 0.89, "learning_rate": 2.4681703659268145e-06, "loss": 0.0005, "step": 7438 }, { "epoch": 0.89, "learning_rate": 2.4680983803239352e-06, "loss": 0.0004, "step": 7439 }, { "epoch": 0.89, "learning_rate": 2.468026394721056e-06, "loss": 0.0005, "step": 7440 }, { "epoch": 0.89, "learning_rate": 2.4679544091181764e-06, "loss": 0.0007, "step": 7441 }, { "epoch": 0.89, "learning_rate": 2.467882423515297e-06, "loss": 0.0004, "step": 7442 }, { "epoch": 0.89, "learning_rate": 2.4678104379124175e-06, "loss": 0.1964, "step": 7443 }, { "epoch": 0.89, "learning_rate": 2.4677384523095383e-06, "loss": 0.0006, "step": 7444 }, { "epoch": 0.89, "learning_rate": 2.4676664667066586e-06, "loss": 0.0005, "step": 7445 }, { "epoch": 0.89, "learning_rate": 2.4675944811037794e-06, "loss": 0.0006, "step": 7446 }, { "epoch": 0.89, "learning_rate": 2.4675224955008998e-06, "loss": 0.0005, "step": 7447 }, { "epoch": 0.89, "learning_rate": 2.4674505098980205e-06, "loss": 0.7617, "step": 7448 }, { "epoch": 0.89, "learning_rate": 2.467378524295141e-06, "loss": 0.0026, "step": 7449 }, { "epoch": 0.89, "learning_rate": 2.4673065386922617e-06, "loss": 0.0004, "step": 7450 }, { "epoch": 0.89, "learning_rate": 2.467234553089382e-06, "loss": 0.0003, "step": 7451 }, { "epoch": 0.89, "learning_rate": 2.4671625674865024e-06, "loss": 0.0006, "step": 7452 }, { "epoch": 0.89, "learning_rate": 2.467090581883623e-06, "loss": 0.7369, "step": 7453 }, { "epoch": 0.89, "learning_rate": 2.467018596280744e-06, "loss": 0.0007, "step": 7454 }, { "epoch": 0.89, "learning_rate": 2.4669466106778647e-06, "loss": 0.0004, "step": 7455 }, { "epoch": 0.89, "learning_rate": 2.466874625074985e-06, "loss": 0.0004, "step": 7456 }, { "epoch": 0.89, "learning_rate": 2.466802639472106e-06, "loss": 0.0004, "step": 7457 }, { "epoch": 0.89, "learning_rate": 2.4667306538692262e-06, "loss": 0.0004, "step": 7458 }, { "epoch": 0.89, "learning_rate": 2.466658668266347e-06, "loss": 0.0006, "step": 7459 }, { "epoch": 0.89, "learning_rate": 2.4665866826634673e-06, "loss": 0.0004, "step": 7460 }, { "epoch": 0.89, "learning_rate": 2.466514697060588e-06, "loss": 0.0006, "step": 7461 }, { "epoch": 0.89, "learning_rate": 2.4664427114577085e-06, "loss": 0.0003, "step": 7462 }, { "epoch": 0.89, "learning_rate": 2.4663707258548293e-06, "loss": 0.0005, "step": 7463 }, { "epoch": 0.89, "learning_rate": 2.4662987402519496e-06, "loss": 0.0003, "step": 7464 }, { "epoch": 0.89, "learning_rate": 2.46622675464907e-06, "loss": 0.0009, "step": 7465 }, { "epoch": 0.89, "learning_rate": 2.4661547690461908e-06, "loss": 0.0009, "step": 7466 }, { "epoch": 0.89, "learning_rate": 2.466082783443311e-06, "loss": 0.0009, "step": 7467 }, { "epoch": 0.89, "learning_rate": 2.466010797840432e-06, "loss": 0.0004, "step": 7468 }, { "epoch": 0.9, "learning_rate": 2.4659388122375527e-06, "loss": 0.0005, "step": 7469 }, { "epoch": 0.9, "learning_rate": 2.465866826634673e-06, "loss": 0.0004, "step": 7470 }, { "epoch": 0.9, "learning_rate": 2.465794841031794e-06, "loss": 0.0007, "step": 7471 }, { "epoch": 0.9, "learning_rate": 2.4657228554289146e-06, "loss": 0.0009, "step": 7472 }, { "epoch": 0.9, "learning_rate": 2.465650869826035e-06, "loss": 0.0008, "step": 7473 }, { "epoch": 0.9, "learning_rate": 2.4655788842231557e-06, "loss": 0.0019, "step": 7474 }, { "epoch": 0.9, "learning_rate": 2.465506898620276e-06, "loss": 0.0004, "step": 7475 }, { "epoch": 0.9, "learning_rate": 2.4654349130173964e-06, "loss": 0.0006, "step": 7476 }, { "epoch": 0.9, "learning_rate": 2.465362927414517e-06, "loss": 0.0008, "step": 7477 }, { "epoch": 0.9, "learning_rate": 2.4652909418116376e-06, "loss": 0.0007, "step": 7478 }, { "epoch": 0.9, "learning_rate": 2.4652189562087583e-06, "loss": 0.0005, "step": 7479 }, { "epoch": 0.9, "learning_rate": 2.4651469706058787e-06, "loss": 0.0004, "step": 7480 }, { "epoch": 0.9, "learning_rate": 2.4650749850029995e-06, "loss": 0.0005, "step": 7481 }, { "epoch": 0.9, "learning_rate": 2.46500299940012e-06, "loss": 0.0007, "step": 7482 }, { "epoch": 0.9, "learning_rate": 2.4649310137972406e-06, "loss": 0.0005, "step": 7483 }, { "epoch": 0.9, "learning_rate": 2.464859028194361e-06, "loss": 0.0017, "step": 7484 }, { "epoch": 0.9, "learning_rate": 2.4647870425914817e-06, "loss": 0.0004, "step": 7485 }, { "epoch": 0.9, "learning_rate": 2.4647150569886025e-06, "loss": 0.934, "step": 7486 }, { "epoch": 0.9, "learning_rate": 2.464643071385723e-06, "loss": 0.0006, "step": 7487 }, { "epoch": 0.9, "learning_rate": 2.4645710857828436e-06, "loss": 0.0005, "step": 7488 }, { "epoch": 0.9, "learning_rate": 2.464499100179964e-06, "loss": 0.0009, "step": 7489 }, { "epoch": 0.9, "learning_rate": 2.4644271145770848e-06, "loss": 0.3909, "step": 7490 }, { "epoch": 0.9, "learning_rate": 2.464355128974205e-06, "loss": 0.0004, "step": 7491 }, { "epoch": 0.9, "learning_rate": 2.464283143371326e-06, "loss": 0.0005, "step": 7492 }, { "epoch": 0.9, "learning_rate": 2.4642111577684463e-06, "loss": 0.0035, "step": 7493 }, { "epoch": 0.9, "learning_rate": 2.464139172165567e-06, "loss": 0.0004, "step": 7494 }, { "epoch": 0.9, "learning_rate": 2.4640671865626874e-06, "loss": 0.0005, "step": 7495 }, { "epoch": 0.9, "learning_rate": 2.463995200959808e-06, "loss": 0.0006, "step": 7496 }, { "epoch": 0.9, "learning_rate": 2.4639232153569285e-06, "loss": 0.0143, "step": 7497 }, { "epoch": 0.9, "learning_rate": 2.4638512297540493e-06, "loss": 0.0006, "step": 7498 }, { "epoch": 0.9, "learning_rate": 2.4637792441511697e-06, "loss": 0.0006, "step": 7499 }, { "epoch": 0.9, "learning_rate": 2.4637072585482904e-06, "loss": 0.0009, "step": 7500 }, { "epoch": 0.9, "learning_rate": 2.4636352729454112e-06, "loss": 0.0005, "step": 7501 }, { "epoch": 0.9, "learning_rate": 2.4635632873425316e-06, "loss": 0.0005, "step": 7502 }, { "epoch": 0.9, "learning_rate": 2.4634913017396524e-06, "loss": 0.0004, "step": 7503 }, { "epoch": 0.9, "learning_rate": 2.4634193161367727e-06, "loss": 0.0009, "step": 7504 }, { "epoch": 0.9, "learning_rate": 2.4633473305338935e-06, "loss": 0.0007, "step": 7505 }, { "epoch": 0.9, "learning_rate": 2.463275344931014e-06, "loss": 0.0007, "step": 7506 }, { "epoch": 0.9, "learning_rate": 2.4632033593281346e-06, "loss": 0.0015, "step": 7507 }, { "epoch": 0.9, "learning_rate": 2.463131373725255e-06, "loss": 0.0014, "step": 7508 }, { "epoch": 0.9, "learning_rate": 2.4630593881223758e-06, "loss": 0.0019, "step": 7509 }, { "epoch": 0.9, "learning_rate": 2.462987402519496e-06, "loss": 0.0007, "step": 7510 }, { "epoch": 0.9, "learning_rate": 2.4629154169166165e-06, "loss": 0.0042, "step": 7511 }, { "epoch": 0.9, "learning_rate": 2.4628434313137372e-06, "loss": 0.0003, "step": 7512 }, { "epoch": 0.9, "learning_rate": 2.4627714457108576e-06, "loss": 0.725, "step": 7513 }, { "epoch": 0.9, "learning_rate": 2.4626994601079784e-06, "loss": 0.0104, "step": 7514 }, { "epoch": 0.9, "learning_rate": 2.462627474505099e-06, "loss": 0.0027, "step": 7515 }, { "epoch": 0.9, "learning_rate": 2.4625554889022195e-06, "loss": 0.0007, "step": 7516 }, { "epoch": 0.9, "learning_rate": 2.4624835032993403e-06, "loss": 0.7682, "step": 7517 }, { "epoch": 0.9, "learning_rate": 2.462411517696461e-06, "loss": 0.0067, "step": 7518 }, { "epoch": 0.9, "learning_rate": 2.4623395320935814e-06, "loss": 0.0007, "step": 7519 }, { "epoch": 0.9, "learning_rate": 2.462267546490702e-06, "loss": 0.3271, "step": 7520 }, { "epoch": 0.9, "learning_rate": 2.4621955608878226e-06, "loss": 0.7992, "step": 7521 }, { "epoch": 0.9, "learning_rate": 2.462123575284943e-06, "loss": 0.0008, "step": 7522 }, { "epoch": 0.9, "learning_rate": 2.4620515896820637e-06, "loss": 0.0007, "step": 7523 }, { "epoch": 0.9, "learning_rate": 2.461979604079184e-06, "loss": 0.0004, "step": 7524 }, { "epoch": 0.9, "learning_rate": 2.461907618476305e-06, "loss": 0.0003, "step": 7525 }, { "epoch": 0.9, "learning_rate": 2.461835632873425e-06, "loss": 0.0004, "step": 7526 }, { "epoch": 0.9, "learning_rate": 2.461763647270546e-06, "loss": 0.0003, "step": 7527 }, { "epoch": 0.9, "learning_rate": 2.4616916616676663e-06, "loss": 0.0004, "step": 7528 }, { "epoch": 0.9, "learning_rate": 2.461619676064787e-06, "loss": 0.0005, "step": 7529 }, { "epoch": 0.9, "learning_rate": 2.4615476904619074e-06, "loss": 0.0006, "step": 7530 }, { "epoch": 0.9, "learning_rate": 2.4614757048590282e-06, "loss": 0.0003, "step": 7531 }, { "epoch": 0.9, "learning_rate": 2.461403719256149e-06, "loss": 0.0003, "step": 7532 }, { "epoch": 0.9, "learning_rate": 2.4613317336532694e-06, "loss": 0.0004, "step": 7533 }, { "epoch": 0.9, "learning_rate": 2.46125974805039e-06, "loss": 0.0005, "step": 7534 }, { "epoch": 0.9, "learning_rate": 2.4611877624475105e-06, "loss": 0.4345, "step": 7535 }, { "epoch": 0.9, "learning_rate": 2.4611157768446313e-06, "loss": 0.0006, "step": 7536 }, { "epoch": 0.9, "learning_rate": 2.4610437912417516e-06, "loss": 0.003, "step": 7537 }, { "epoch": 0.9, "learning_rate": 2.4609718056388724e-06, "loss": 0.5972, "step": 7538 }, { "epoch": 0.9, "learning_rate": 2.4608998200359928e-06, "loss": 0.0004, "step": 7539 }, { "epoch": 0.9, "learning_rate": 2.4608278344331135e-06, "loss": 0.0005, "step": 7540 }, { "epoch": 0.9, "learning_rate": 2.460755848830234e-06, "loss": 0.0003, "step": 7541 }, { "epoch": 0.9, "learning_rate": 2.4606838632273547e-06, "loss": 0.0017, "step": 7542 }, { "epoch": 0.9, "learning_rate": 2.460611877624475e-06, "loss": 0.0005, "step": 7543 }, { "epoch": 0.9, "learning_rate": 2.460539892021596e-06, "loss": 0.0005, "step": 7544 }, { "epoch": 0.9, "learning_rate": 2.460467906418716e-06, "loss": 0.0004, "step": 7545 }, { "epoch": 0.9, "learning_rate": 2.460395920815837e-06, "loss": 0.0005, "step": 7546 }, { "epoch": 0.9, "learning_rate": 2.4603239352129577e-06, "loss": 0.0007, "step": 7547 }, { "epoch": 0.9, "learning_rate": 2.460251949610078e-06, "loss": 0.0004, "step": 7548 }, { "epoch": 0.9, "learning_rate": 2.460179964007199e-06, "loss": 0.0005, "step": 7549 }, { "epoch": 0.9, "learning_rate": 2.460107978404319e-06, "loss": 0.9255, "step": 7550 }, { "epoch": 0.9, "learning_rate": 2.46003599280144e-06, "loss": 0.0004, "step": 7551 }, { "epoch": 0.9, "learning_rate": 2.4599640071985603e-06, "loss": 0.0004, "step": 7552 }, { "epoch": 0.91, "learning_rate": 2.459892021595681e-06, "loss": 0.0003, "step": 7553 }, { "epoch": 0.91, "learning_rate": 2.4598200359928015e-06, "loss": 0.0005, "step": 7554 }, { "epoch": 0.91, "learning_rate": 2.4597480503899223e-06, "loss": 0.0005, "step": 7555 }, { "epoch": 0.91, "learning_rate": 2.4596760647870426e-06, "loss": 0.0009, "step": 7556 }, { "epoch": 0.91, "learning_rate": 2.459604079184163e-06, "loss": 0.8835, "step": 7557 }, { "epoch": 0.91, "learning_rate": 2.4595320935812837e-06, "loss": 0.0003, "step": 7558 }, { "epoch": 0.91, "learning_rate": 2.459460107978404e-06, "loss": 0.0004, "step": 7559 }, { "epoch": 0.91, "learning_rate": 2.459388122375525e-06, "loss": 0.003, "step": 7560 }, { "epoch": 0.91, "learning_rate": 2.4593161367726452e-06, "loss": 0.0004, "step": 7561 }, { "epoch": 0.91, "learning_rate": 2.459244151169766e-06, "loss": 0.8944, "step": 7562 }, { "epoch": 0.91, "learning_rate": 2.4591721655668868e-06, "loss": 0.0005, "step": 7563 }, { "epoch": 0.91, "learning_rate": 2.4591001799640076e-06, "loss": 0.0009, "step": 7564 }, { "epoch": 0.91, "learning_rate": 2.459028194361128e-06, "loss": 0.0006, "step": 7565 }, { "epoch": 0.91, "learning_rate": 2.4589562087582487e-06, "loss": 0.0007, "step": 7566 }, { "epoch": 0.91, "learning_rate": 2.458884223155369e-06, "loss": 0.0006, "step": 7567 }, { "epoch": 0.91, "learning_rate": 2.4588122375524894e-06, "loss": 0.0638, "step": 7568 }, { "epoch": 0.91, "learning_rate": 2.45874025194961e-06, "loss": 0.0004, "step": 7569 }, { "epoch": 0.91, "learning_rate": 2.4586682663467305e-06, "loss": 0.0005, "step": 7570 }, { "epoch": 0.91, "learning_rate": 2.4585962807438513e-06, "loss": 0.0005, "step": 7571 }, { "epoch": 0.91, "learning_rate": 2.4585242951409717e-06, "loss": 0.0004, "step": 7572 }, { "epoch": 0.91, "learning_rate": 2.4584523095380925e-06, "loss": 0.0007, "step": 7573 }, { "epoch": 0.91, "learning_rate": 2.458380323935213e-06, "loss": 0.0021, "step": 7574 }, { "epoch": 0.91, "learning_rate": 2.4583083383323336e-06, "loss": 0.0004, "step": 7575 }, { "epoch": 0.91, "learning_rate": 2.458236352729454e-06, "loss": 0.0005, "step": 7576 }, { "epoch": 0.91, "learning_rate": 2.4581643671265747e-06, "loss": 0.0007, "step": 7577 }, { "epoch": 0.91, "learning_rate": 2.4580923815236955e-06, "loss": 0.045, "step": 7578 }, { "epoch": 0.91, "learning_rate": 2.458020395920816e-06, "loss": 0.0006, "step": 7579 }, { "epoch": 0.91, "learning_rate": 2.4579484103179366e-06, "loss": 0.0005, "step": 7580 }, { "epoch": 0.91, "learning_rate": 2.457876424715057e-06, "loss": 0.0005, "step": 7581 }, { "epoch": 0.91, "learning_rate": 2.4578044391121778e-06, "loss": 0.0005, "step": 7582 }, { "epoch": 0.91, "learning_rate": 2.457732453509298e-06, "loss": 0.0007, "step": 7583 }, { "epoch": 0.91, "learning_rate": 2.457660467906419e-06, "loss": 0.0005, "step": 7584 }, { "epoch": 0.91, "learning_rate": 2.4575884823035393e-06, "loss": 0.0009, "step": 7585 }, { "epoch": 0.91, "learning_rate": 2.45751649670066e-06, "loss": 0.0008, "step": 7586 }, { "epoch": 0.91, "learning_rate": 2.4574445110977804e-06, "loss": 0.0004, "step": 7587 }, { "epoch": 0.91, "learning_rate": 2.457372525494901e-06, "loss": 0.0004, "step": 7588 }, { "epoch": 0.91, "learning_rate": 2.4573005398920215e-06, "loss": 0.0005, "step": 7589 }, { "epoch": 0.91, "learning_rate": 2.4572285542891423e-06, "loss": 0.0003, "step": 7590 }, { "epoch": 0.91, "learning_rate": 2.4571565686862627e-06, "loss": 0.0005, "step": 7591 }, { "epoch": 0.91, "learning_rate": 2.4570845830833834e-06, "loss": 0.0005, "step": 7592 }, { "epoch": 0.91, "learning_rate": 2.457012597480504e-06, "loss": 0.0008, "step": 7593 }, { "epoch": 0.91, "learning_rate": 2.4569406118776246e-06, "loss": 0.6664, "step": 7594 }, { "epoch": 0.91, "learning_rate": 2.4568686262747453e-06, "loss": 0.0004, "step": 7595 }, { "epoch": 0.91, "learning_rate": 2.4567966406718657e-06, "loss": 0.0037, "step": 7596 }, { "epoch": 0.91, "learning_rate": 2.4567246550689865e-06, "loss": 0.0007, "step": 7597 }, { "epoch": 0.91, "learning_rate": 2.456652669466107e-06, "loss": 0.904, "step": 7598 }, { "epoch": 0.91, "learning_rate": 2.4565806838632276e-06, "loss": 0.0009, "step": 7599 }, { "epoch": 0.91, "learning_rate": 2.456508698260348e-06, "loss": 0.0007, "step": 7600 }, { "epoch": 0.91, "learning_rate": 2.4564367126574687e-06, "loss": 0.0004, "step": 7601 }, { "epoch": 0.91, "learning_rate": 2.456364727054589e-06, "loss": 0.0006, "step": 7602 }, { "epoch": 0.91, "learning_rate": 2.4562927414517095e-06, "loss": 0.0003, "step": 7603 }, { "epoch": 0.91, "learning_rate": 2.4562207558488302e-06, "loss": 0.0446, "step": 7604 }, { "epoch": 0.91, "learning_rate": 2.4561487702459506e-06, "loss": 0.0007, "step": 7605 }, { "epoch": 0.91, "learning_rate": 2.4560767846430714e-06, "loss": 0.0004, "step": 7606 }, { "epoch": 0.91, "learning_rate": 2.4560047990401917e-06, "loss": 0.0004, "step": 7607 }, { "epoch": 0.91, "learning_rate": 2.4559328134373125e-06, "loss": 0.0014, "step": 7608 }, { "epoch": 0.91, "learning_rate": 2.4558608278344333e-06, "loss": 0.0015, "step": 7609 }, { "epoch": 0.91, "learning_rate": 2.455788842231554e-06, "loss": 0.0006, "step": 7610 }, { "epoch": 0.91, "learning_rate": 2.4557168566286744e-06, "loss": 0.0005, "step": 7611 }, { "epoch": 0.91, "learning_rate": 2.455644871025795e-06, "loss": 0.0004, "step": 7612 }, { "epoch": 0.91, "learning_rate": 2.4555728854229156e-06, "loss": 0.0538, "step": 7613 }, { "epoch": 0.91, "learning_rate": 2.455500899820036e-06, "loss": 0.0003, "step": 7614 }, { "epoch": 0.91, "learning_rate": 2.4554289142171567e-06, "loss": 0.0006, "step": 7615 }, { "epoch": 0.91, "learning_rate": 2.455356928614277e-06, "loss": 0.0004, "step": 7616 }, { "epoch": 0.91, "learning_rate": 2.455284943011398e-06, "loss": 0.2319, "step": 7617 }, { "epoch": 0.91, "learning_rate": 2.455212957408518e-06, "loss": 0.001, "step": 7618 }, { "epoch": 0.91, "learning_rate": 2.455140971805639e-06, "loss": 0.0007, "step": 7619 }, { "epoch": 0.91, "learning_rate": 2.4550689862027593e-06, "loss": 0.0004, "step": 7620 }, { "epoch": 0.91, "learning_rate": 2.45499700059988e-06, "loss": 0.0004, "step": 7621 }, { "epoch": 0.91, "learning_rate": 2.4549250149970004e-06, "loss": 0.0006, "step": 7622 }, { "epoch": 0.91, "learning_rate": 2.4548530293941212e-06, "loss": 0.0007, "step": 7623 }, { "epoch": 0.91, "learning_rate": 2.454781043791242e-06, "loss": 0.8393, "step": 7624 }, { "epoch": 0.91, "learning_rate": 2.4547090581883624e-06, "loss": 0.9417, "step": 7625 }, { "epoch": 0.91, "learning_rate": 2.454637072585483e-06, "loss": 0.0004, "step": 7626 }, { "epoch": 0.91, "learning_rate": 2.4545650869826035e-06, "loss": 0.0006, "step": 7627 }, { "epoch": 0.91, "learning_rate": 2.4544931013797243e-06, "loss": 0.0007, "step": 7628 }, { "epoch": 0.91, "learning_rate": 2.4544211157768446e-06, "loss": 0.0009, "step": 7629 }, { "epoch": 0.91, "learning_rate": 2.4543491301739654e-06, "loss": 0.0008, "step": 7630 }, { "epoch": 0.91, "learning_rate": 2.4542771445710858e-06, "loss": 0.0005, "step": 7631 }, { "epoch": 0.91, "learning_rate": 2.4542051589682065e-06, "loss": 0.6655, "step": 7632 }, { "epoch": 0.91, "learning_rate": 2.454133173365327e-06, "loss": 0.0005, "step": 7633 }, { "epoch": 0.91, "learning_rate": 2.4540611877624477e-06, "loss": 0.0007, "step": 7634 }, { "epoch": 0.91, "learning_rate": 2.453989202159568e-06, "loss": 0.559, "step": 7635 }, { "epoch": 0.92, "learning_rate": 2.453917216556689e-06, "loss": 0.0005, "step": 7636 }, { "epoch": 0.92, "learning_rate": 2.453845230953809e-06, "loss": 0.0008, "step": 7637 }, { "epoch": 0.92, "learning_rate": 2.45377324535093e-06, "loss": 0.0005, "step": 7638 }, { "epoch": 0.92, "learning_rate": 2.4537012597480503e-06, "loss": 0.0005, "step": 7639 }, { "epoch": 0.92, "learning_rate": 2.453629274145171e-06, "loss": 0.0007, "step": 7640 }, { "epoch": 0.92, "learning_rate": 2.453557288542292e-06, "loss": 0.6369, "step": 7641 }, { "epoch": 0.92, "learning_rate": 2.453485302939412e-06, "loss": 0.0007, "step": 7642 }, { "epoch": 0.92, "learning_rate": 2.453413317336533e-06, "loss": 0.0004, "step": 7643 }, { "epoch": 0.92, "learning_rate": 2.4533413317336533e-06, "loss": 0.018, "step": 7644 }, { "epoch": 0.92, "learning_rate": 2.453269346130774e-06, "loss": 0.0004, "step": 7645 }, { "epoch": 0.92, "learning_rate": 2.4531973605278945e-06, "loss": 0.0226, "step": 7646 }, { "epoch": 0.92, "learning_rate": 2.4531253749250152e-06, "loss": 0.0006, "step": 7647 }, { "epoch": 0.92, "learning_rate": 2.4530533893221356e-06, "loss": 0.001, "step": 7648 }, { "epoch": 0.92, "learning_rate": 2.452981403719256e-06, "loss": 0.8737, "step": 7649 }, { "epoch": 0.92, "learning_rate": 2.4529094181163767e-06, "loss": 0.0004, "step": 7650 }, { "epoch": 0.92, "learning_rate": 2.452837432513497e-06, "loss": 0.0005, "step": 7651 }, { "epoch": 0.92, "learning_rate": 2.452765446910618e-06, "loss": 0.0004, "step": 7652 }, { "epoch": 0.92, "learning_rate": 2.4526934613077382e-06, "loss": 0.0006, "step": 7653 }, { "epoch": 0.92, "learning_rate": 2.452621475704859e-06, "loss": 0.0004, "step": 7654 }, { "epoch": 0.92, "learning_rate": 2.4525494901019798e-06, "loss": 0.0003, "step": 7655 }, { "epoch": 0.92, "learning_rate": 2.4524775044991006e-06, "loss": 0.0011, "step": 7656 }, { "epoch": 0.92, "learning_rate": 2.452405518896221e-06, "loss": 0.0006, "step": 7657 }, { "epoch": 0.92, "learning_rate": 2.4523335332933417e-06, "loss": 0.0413, "step": 7658 }, { "epoch": 0.92, "learning_rate": 2.452261547690462e-06, "loss": 0.001, "step": 7659 }, { "epoch": 0.92, "learning_rate": 2.4521895620875824e-06, "loss": 0.6318, "step": 7660 }, { "epoch": 0.92, "learning_rate": 2.452117576484703e-06, "loss": 0.0007, "step": 7661 }, { "epoch": 0.92, "learning_rate": 2.4520455908818235e-06, "loss": 0.0034, "step": 7662 }, { "epoch": 0.92, "learning_rate": 2.4519736052789443e-06, "loss": 0.0073, "step": 7663 }, { "epoch": 0.92, "learning_rate": 2.4519016196760647e-06, "loss": 0.0009, "step": 7664 }, { "epoch": 0.92, "learning_rate": 2.4518296340731854e-06, "loss": 0.0006, "step": 7665 }, { "epoch": 0.92, "learning_rate": 2.451757648470306e-06, "loss": 0.0008, "step": 7666 }, { "epoch": 0.92, "learning_rate": 2.4516856628674266e-06, "loss": 0.0006, "step": 7667 }, { "epoch": 0.92, "learning_rate": 2.451613677264547e-06, "loss": 0.4667, "step": 7668 }, { "epoch": 0.92, "learning_rate": 2.4515416916616677e-06, "loss": 0.0007, "step": 7669 }, { "epoch": 0.92, "learning_rate": 2.4514697060587885e-06, "loss": 0.0007, "step": 7670 }, { "epoch": 0.92, "learning_rate": 2.451397720455909e-06, "loss": 0.0004, "step": 7671 }, { "epoch": 0.92, "learning_rate": 2.4513257348530296e-06, "loss": 0.0005, "step": 7672 }, { "epoch": 0.92, "learning_rate": 2.45125374925015e-06, "loss": 0.0044, "step": 7673 }, { "epoch": 0.92, "learning_rate": 2.4511817636472708e-06, "loss": 0.0004, "step": 7674 }, { "epoch": 0.92, "learning_rate": 2.451109778044391e-06, "loss": 0.0005, "step": 7675 }, { "epoch": 0.92, "learning_rate": 2.451037792441512e-06, "loss": 0.0005, "step": 7676 }, { "epoch": 0.92, "learning_rate": 2.4509658068386322e-06, "loss": 0.0011, "step": 7677 }, { "epoch": 0.92, "learning_rate": 2.450893821235753e-06, "loss": 0.0007, "step": 7678 }, { "epoch": 0.92, "learning_rate": 2.4508218356328734e-06, "loss": 0.0007, "step": 7679 }, { "epoch": 0.92, "learning_rate": 2.450749850029994e-06, "loss": 0.0007, "step": 7680 }, { "epoch": 0.92, "learning_rate": 2.4506778644271145e-06, "loss": 0.0006, "step": 7681 }, { "epoch": 0.92, "learning_rate": 2.4506058788242353e-06, "loss": 0.0004, "step": 7682 }, { "epoch": 0.92, "learning_rate": 2.4505338932213557e-06, "loss": 0.0006, "step": 7683 }, { "epoch": 0.92, "learning_rate": 2.450461907618476e-06, "loss": 0.0024, "step": 7684 }, { "epoch": 0.92, "learning_rate": 2.4503899220155968e-06, "loss": 0.0006, "step": 7685 }, { "epoch": 0.92, "learning_rate": 2.4503179364127176e-06, "loss": 0.0005, "step": 7686 }, { "epoch": 0.92, "learning_rate": 2.4502459508098383e-06, "loss": 0.047, "step": 7687 }, { "epoch": 0.92, "learning_rate": 2.4501739652069587e-06, "loss": 0.0006, "step": 7688 }, { "epoch": 0.92, "learning_rate": 2.4501019796040795e-06, "loss": 0.0048, "step": 7689 }, { "epoch": 0.92, "learning_rate": 2.4500299940012e-06, "loss": 0.0004, "step": 7690 }, { "epoch": 0.92, "learning_rate": 2.4499580083983206e-06, "loss": 0.0004, "step": 7691 }, { "epoch": 0.92, "learning_rate": 2.449886022795441e-06, "loss": 0.0006, "step": 7692 }, { "epoch": 0.92, "learning_rate": 2.4498140371925617e-06, "loss": 0.0009, "step": 7693 }, { "epoch": 0.92, "learning_rate": 2.449742051589682e-06, "loss": 0.0008, "step": 7694 }, { "epoch": 0.92, "learning_rate": 2.4496700659868025e-06, "loss": 0.0007, "step": 7695 }, { "epoch": 0.92, "learning_rate": 2.4495980803839232e-06, "loss": 0.0003, "step": 7696 }, { "epoch": 0.92, "learning_rate": 2.4495260947810436e-06, "loss": 0.0048, "step": 7697 }, { "epoch": 0.92, "learning_rate": 2.4494541091781644e-06, "loss": 0.0005, "step": 7698 }, { "epoch": 0.92, "learning_rate": 2.4493821235752847e-06, "loss": 0.0007, "step": 7699 }, { "epoch": 0.92, "learning_rate": 2.4493101379724055e-06, "loss": 0.0004, "step": 7700 }, { "epoch": 0.92, "learning_rate": 2.4492381523695263e-06, "loss": 0.0005, "step": 7701 }, { "epoch": 0.92, "learning_rate": 2.4491661667666466e-06, "loss": 0.0004, "step": 7702 }, { "epoch": 0.92, "learning_rate": 2.4490941811637674e-06, "loss": 0.0005, "step": 7703 }, { "epoch": 0.92, "learning_rate": 2.449022195560888e-06, "loss": 0.0003, "step": 7704 }, { "epoch": 0.92, "learning_rate": 2.4489502099580085e-06, "loss": 0.0007, "step": 7705 }, { "epoch": 0.92, "learning_rate": 2.4488782243551293e-06, "loss": 0.0004, "step": 7706 }, { "epoch": 0.92, "learning_rate": 2.4488062387522497e-06, "loss": 0.0004, "step": 7707 }, { "epoch": 0.92, "learning_rate": 2.44873425314937e-06, "loss": 0.531, "step": 7708 }, { "epoch": 0.92, "learning_rate": 2.448662267546491e-06, "loss": 0.0007, "step": 7709 }, { "epoch": 0.92, "learning_rate": 2.448590281943611e-06, "loss": 0.0823, "step": 7710 }, { "epoch": 0.92, "learning_rate": 2.448518296340732e-06, "loss": 0.0013, "step": 7711 }, { "epoch": 0.92, "learning_rate": 2.4484463107378523e-06, "loss": 0.0006, "step": 7712 }, { "epoch": 0.92, "learning_rate": 2.448374325134973e-06, "loss": 0.0026, "step": 7713 }, { "epoch": 0.92, "learning_rate": 2.4483023395320934e-06, "loss": 0.0005, "step": 7714 }, { "epoch": 0.92, "learning_rate": 2.448230353929214e-06, "loss": 0.0004, "step": 7715 }, { "epoch": 0.92, "learning_rate": 2.4481583683263346e-06, "loss": 0.0007, "step": 7716 }, { "epoch": 0.92, "learning_rate": 2.4480863827234553e-06, "loss": 0.0007, "step": 7717 }, { "epoch": 0.92, "learning_rate": 2.448014397120576e-06, "loss": 0.0323, "step": 7718 }, { "epoch": 0.92, "learning_rate": 2.4479424115176965e-06, "loss": 0.6902, "step": 7719 }, { "epoch": 0.93, "learning_rate": 2.4478704259148173e-06, "loss": 0.0034, "step": 7720 }, { "epoch": 0.93, "learning_rate": 2.4477984403119376e-06, "loss": 0.0004, "step": 7721 }, { "epoch": 0.93, "learning_rate": 2.4477264547090584e-06, "loss": 0.0004, "step": 7722 }, { "epoch": 0.93, "learning_rate": 2.4476544691061787e-06, "loss": 0.0002, "step": 7723 }, { "epoch": 0.93, "learning_rate": 2.4475824835032995e-06, "loss": 0.0005, "step": 7724 }, { "epoch": 0.93, "learning_rate": 2.44751049790042e-06, "loss": 0.0005, "step": 7725 }, { "epoch": 0.93, "learning_rate": 2.4474385122975407e-06, "loss": 0.0004, "step": 7726 }, { "epoch": 0.93, "learning_rate": 2.447366526694661e-06, "loss": 0.0005, "step": 7727 }, { "epoch": 0.93, "learning_rate": 2.447294541091782e-06, "loss": 0.0035, "step": 7728 }, { "epoch": 0.93, "learning_rate": 2.447222555488902e-06, "loss": 0.0006, "step": 7729 }, { "epoch": 0.93, "learning_rate": 2.4471505698860225e-06, "loss": 0.002, "step": 7730 }, { "epoch": 0.93, "learning_rate": 2.4470785842831433e-06, "loss": 0.0006, "step": 7731 }, { "epoch": 0.93, "learning_rate": 2.447006598680264e-06, "loss": 0.0004, "step": 7732 }, { "epoch": 0.93, "learning_rate": 2.446934613077385e-06, "loss": 0.0004, "step": 7733 }, { "epoch": 0.93, "learning_rate": 2.446862627474505e-06, "loss": 0.0005, "step": 7734 }, { "epoch": 0.93, "learning_rate": 2.446790641871626e-06, "loss": 0.001, "step": 7735 }, { "epoch": 0.93, "learning_rate": 2.4467186562687463e-06, "loss": 0.0005, "step": 7736 }, { "epoch": 0.93, "learning_rate": 2.446646670665867e-06, "loss": 0.0018, "step": 7737 }, { "epoch": 0.93, "learning_rate": 2.4465746850629875e-06, "loss": 0.0004, "step": 7738 }, { "epoch": 0.93, "learning_rate": 2.4465026994601082e-06, "loss": 0.0004, "step": 7739 }, { "epoch": 0.93, "learning_rate": 2.4464307138572286e-06, "loss": 0.0006, "step": 7740 }, { "epoch": 0.93, "learning_rate": 2.446358728254349e-06, "loss": 0.0023, "step": 7741 }, { "epoch": 0.93, "learning_rate": 2.4462867426514697e-06, "loss": 0.0007, "step": 7742 }, { "epoch": 0.93, "learning_rate": 2.44621475704859e-06, "loss": 0.0006, "step": 7743 }, { "epoch": 0.93, "learning_rate": 2.446142771445711e-06, "loss": 0.0004, "step": 7744 }, { "epoch": 0.93, "learning_rate": 2.4460707858428312e-06, "loss": 0.0003, "step": 7745 }, { "epoch": 0.93, "learning_rate": 2.445998800239952e-06, "loss": 0.0004, "step": 7746 }, { "epoch": 0.93, "learning_rate": 2.4459268146370728e-06, "loss": 0.0003, "step": 7747 }, { "epoch": 0.93, "learning_rate": 2.445854829034193e-06, "loss": 0.0346, "step": 7748 }, { "epoch": 0.93, "learning_rate": 2.445782843431314e-06, "loss": 0.0007, "step": 7749 }, { "epoch": 0.93, "learning_rate": 2.4457108578284347e-06, "loss": 0.0004, "step": 7750 }, { "epoch": 0.93, "learning_rate": 2.445638872225555e-06, "loss": 0.0354, "step": 7751 }, { "epoch": 0.93, "learning_rate": 2.445566886622676e-06, "loss": 0.0318, "step": 7752 }, { "epoch": 0.93, "learning_rate": 2.445494901019796e-06, "loss": 0.3404, "step": 7753 }, { "epoch": 0.93, "learning_rate": 2.4454229154169165e-06, "loss": 0.0003, "step": 7754 }, { "epoch": 0.93, "learning_rate": 2.4453509298140373e-06, "loss": 0.0003, "step": 7755 }, { "epoch": 0.93, "learning_rate": 2.4452789442111577e-06, "loss": 0.0005, "step": 7756 }, { "epoch": 0.93, "learning_rate": 2.4452069586082784e-06, "loss": 0.1608, "step": 7757 }, { "epoch": 0.93, "learning_rate": 2.445134973005399e-06, "loss": 0.0004, "step": 7758 }, { "epoch": 0.93, "learning_rate": 2.4450629874025196e-06, "loss": 0.0005, "step": 7759 }, { "epoch": 0.93, "learning_rate": 2.44499100179964e-06, "loss": 0.0004, "step": 7760 }, { "epoch": 0.93, "learning_rate": 2.4449190161967607e-06, "loss": 0.0008, "step": 7761 }, { "epoch": 0.93, "learning_rate": 2.444847030593881e-06, "loss": 0.0004, "step": 7762 }, { "epoch": 0.93, "learning_rate": 2.444775044991002e-06, "loss": 0.0013, "step": 7763 }, { "epoch": 0.93, "learning_rate": 2.4447030593881226e-06, "loss": 0.0004, "step": 7764 }, { "epoch": 0.93, "learning_rate": 2.444631073785243e-06, "loss": 0.0003, "step": 7765 }, { "epoch": 0.93, "learning_rate": 2.4445590881823638e-06, "loss": 0.0005, "step": 7766 }, { "epoch": 0.93, "learning_rate": 2.444487102579484e-06, "loss": 0.0003, "step": 7767 }, { "epoch": 0.93, "learning_rate": 2.444415116976605e-06, "loss": 0.0004, "step": 7768 }, { "epoch": 0.93, "learning_rate": 2.4443431313737252e-06, "loss": 0.0004, "step": 7769 }, { "epoch": 0.93, "learning_rate": 2.444271145770846e-06, "loss": 0.0003, "step": 7770 }, { "epoch": 0.93, "learning_rate": 2.4441991601679664e-06, "loss": 0.0005, "step": 7771 }, { "epoch": 0.93, "learning_rate": 2.444127174565087e-06, "loss": 0.0003, "step": 7772 }, { "epoch": 0.93, "learning_rate": 2.4440551889622075e-06, "loss": 0.0003, "step": 7773 }, { "epoch": 0.93, "learning_rate": 2.4439832033593283e-06, "loss": 0.0004, "step": 7774 }, { "epoch": 0.93, "learning_rate": 2.4439112177564486e-06, "loss": 1.6611, "step": 7775 }, { "epoch": 0.93, "learning_rate": 2.443839232153569e-06, "loss": 0.0003, "step": 7776 }, { "epoch": 0.93, "learning_rate": 2.4437672465506898e-06, "loss": 0.001, "step": 7777 }, { "epoch": 0.93, "learning_rate": 2.4436952609478106e-06, "loss": 0.4277, "step": 7778 }, { "epoch": 0.93, "learning_rate": 2.4436232753449313e-06, "loss": 0.0003, "step": 7779 }, { "epoch": 0.93, "learning_rate": 2.4435512897420517e-06, "loss": 0.4159, "step": 7780 }, { "epoch": 0.93, "learning_rate": 2.4434793041391725e-06, "loss": 0.0006, "step": 7781 }, { "epoch": 0.93, "learning_rate": 2.443407318536293e-06, "loss": 0.0008, "step": 7782 }, { "epoch": 0.93, "learning_rate": 2.4433353329334136e-06, "loss": 0.0004, "step": 7783 }, { "epoch": 0.93, "learning_rate": 2.443263347330534e-06, "loss": 0.0342, "step": 7784 }, { "epoch": 0.93, "learning_rate": 2.4431913617276547e-06, "loss": 0.0006, "step": 7785 }, { "epoch": 0.93, "learning_rate": 2.443119376124775e-06, "loss": 0.0002, "step": 7786 }, { "epoch": 0.93, "learning_rate": 2.4430473905218954e-06, "loss": 0.0005, "step": 7787 }, { "epoch": 0.93, "learning_rate": 2.4429754049190162e-06, "loss": 0.0365, "step": 7788 }, { "epoch": 0.93, "learning_rate": 2.4429034193161366e-06, "loss": 0.0007, "step": 7789 }, { "epoch": 0.93, "learning_rate": 2.4428314337132574e-06, "loss": 0.0005, "step": 7790 }, { "epoch": 0.93, "learning_rate": 2.4427594481103777e-06, "loss": 0.0004, "step": 7791 }, { "epoch": 0.93, "learning_rate": 2.4426874625074985e-06, "loss": 0.3314, "step": 7792 }, { "epoch": 0.93, "learning_rate": 2.4426154769046193e-06, "loss": 0.0003, "step": 7793 }, { "epoch": 0.93, "learning_rate": 2.4425434913017396e-06, "loss": 0.9568, "step": 7794 }, { "epoch": 0.93, "learning_rate": 2.4424715056988604e-06, "loss": 0.0005, "step": 7795 }, { "epoch": 0.93, "learning_rate": 2.442399520095981e-06, "loss": 0.0005, "step": 7796 }, { "epoch": 0.93, "learning_rate": 2.4423275344931015e-06, "loss": 0.0126, "step": 7797 }, { "epoch": 0.93, "learning_rate": 2.4422555488902223e-06, "loss": 0.0007, "step": 7798 }, { "epoch": 0.93, "learning_rate": 2.4421835632873427e-06, "loss": 0.0003, "step": 7799 }, { "epoch": 0.93, "learning_rate": 2.442111577684463e-06, "loss": 0.0021, "step": 7800 }, { "epoch": 0.93, "learning_rate": 2.442039592081584e-06, "loss": 0.0004, "step": 7801 }, { "epoch": 0.93, "learning_rate": 2.441967606478704e-06, "loss": 0.0004, "step": 7802 }, { "epoch": 0.94, "learning_rate": 2.441895620875825e-06, "loss": 0.0006, "step": 7803 }, { "epoch": 0.94, "learning_rate": 2.4418236352729453e-06, "loss": 0.0015, "step": 7804 }, { "epoch": 0.94, "learning_rate": 2.441751649670066e-06, "loss": 0.0003, "step": 7805 }, { "epoch": 0.94, "learning_rate": 2.4416796640671864e-06, "loss": 0.0136, "step": 7806 }, { "epoch": 0.94, "learning_rate": 2.441607678464307e-06, "loss": 0.0009, "step": 7807 }, { "epoch": 0.94, "learning_rate": 2.4415356928614276e-06, "loss": 0.0019, "step": 7808 }, { "epoch": 0.94, "learning_rate": 2.4414637072585483e-06, "loss": 0.0185, "step": 7809 }, { "epoch": 0.94, "learning_rate": 2.441391721655669e-06, "loss": 0.0005, "step": 7810 }, { "epoch": 0.94, "learning_rate": 2.4413197360527895e-06, "loss": 0.0008, "step": 7811 }, { "epoch": 0.94, "learning_rate": 2.4412477504499102e-06, "loss": 0.0137, "step": 7812 }, { "epoch": 0.94, "learning_rate": 2.4411757648470306e-06, "loss": 0.0003, "step": 7813 }, { "epoch": 0.94, "learning_rate": 2.4411037792441514e-06, "loss": 0.8696, "step": 7814 }, { "epoch": 0.94, "learning_rate": 2.4410317936412717e-06, "loss": 0.0004, "step": 7815 }, { "epoch": 0.94, "learning_rate": 2.4409598080383925e-06, "loss": 0.0004, "step": 7816 }, { "epoch": 0.94, "learning_rate": 2.440887822435513e-06, "loss": 0.0005, "step": 7817 }, { "epoch": 0.94, "learning_rate": 2.4408158368326336e-06, "loss": 0.0004, "step": 7818 }, { "epoch": 0.94, "learning_rate": 2.440743851229754e-06, "loss": 0.001, "step": 7819 }, { "epoch": 0.94, "learning_rate": 2.4406718656268748e-06, "loss": 0.0004, "step": 7820 }, { "epoch": 0.94, "learning_rate": 2.440599880023995e-06, "loss": 0.0004, "step": 7821 }, { "epoch": 0.94, "learning_rate": 2.4405278944211155e-06, "loss": 0.0009, "step": 7822 }, { "epoch": 0.94, "learning_rate": 2.4404559088182363e-06, "loss": 0.0023, "step": 7823 }, { "epoch": 0.94, "learning_rate": 2.440383923215357e-06, "loss": 0.0005, "step": 7824 }, { "epoch": 0.94, "learning_rate": 2.440311937612478e-06, "loss": 0.0005, "step": 7825 }, { "epoch": 0.94, "learning_rate": 2.440239952009598e-06, "loss": 0.0004, "step": 7826 }, { "epoch": 0.94, "learning_rate": 2.440167966406719e-06, "loss": 0.0003, "step": 7827 }, { "epoch": 0.94, "learning_rate": 2.4400959808038393e-06, "loss": 0.0004, "step": 7828 }, { "epoch": 0.94, "learning_rate": 2.44002399520096e-06, "loss": 0.0003, "step": 7829 }, { "epoch": 0.94, "learning_rate": 2.4399520095980805e-06, "loss": 0.0005, "step": 7830 }, { "epoch": 0.94, "learning_rate": 2.4398800239952012e-06, "loss": 0.399, "step": 7831 }, { "epoch": 0.94, "learning_rate": 2.4398080383923216e-06, "loss": 0.0513, "step": 7832 }, { "epoch": 0.94, "learning_rate": 2.4397360527894424e-06, "loss": 0.0082, "step": 7833 }, { "epoch": 0.94, "learning_rate": 2.4396640671865627e-06, "loss": 0.0006, "step": 7834 }, { "epoch": 0.94, "learning_rate": 2.439592081583683e-06, "loss": 0.0005, "step": 7835 }, { "epoch": 0.94, "learning_rate": 2.439520095980804e-06, "loss": 0.0003, "step": 7836 }, { "epoch": 0.94, "learning_rate": 2.439448110377924e-06, "loss": 0.0004, "step": 7837 }, { "epoch": 0.94, "learning_rate": 2.439376124775045e-06, "loss": 0.0004, "step": 7838 }, { "epoch": 0.94, "learning_rate": 2.4393041391721653e-06, "loss": 0.0182, "step": 7839 }, { "epoch": 0.94, "learning_rate": 2.439232153569286e-06, "loss": 0.0027, "step": 7840 }, { "epoch": 0.94, "learning_rate": 2.439160167966407e-06, "loss": 0.0057, "step": 7841 }, { "epoch": 0.94, "learning_rate": 2.4390881823635277e-06, "loss": 0.0004, "step": 7842 }, { "epoch": 0.94, "learning_rate": 2.439016196760648e-06, "loss": 1.0403, "step": 7843 }, { "epoch": 0.94, "learning_rate": 2.438944211157769e-06, "loss": 0.0004, "step": 7844 }, { "epoch": 0.94, "learning_rate": 2.438872225554889e-06, "loss": 0.0587, "step": 7845 }, { "epoch": 0.94, "learning_rate": 2.4388002399520095e-06, "loss": 0.0002, "step": 7846 }, { "epoch": 0.94, "learning_rate": 2.4387282543491303e-06, "loss": 0.0004, "step": 7847 }, { "epoch": 0.94, "learning_rate": 2.4386562687462507e-06, "loss": 0.0007, "step": 7848 }, { "epoch": 0.94, "learning_rate": 2.4385842831433714e-06, "loss": 0.0002, "step": 7849 }, { "epoch": 0.94, "learning_rate": 2.4385122975404918e-06, "loss": 0.0004, "step": 7850 }, { "epoch": 0.94, "learning_rate": 2.4384403119376126e-06, "loss": 0.3233, "step": 7851 }, { "epoch": 0.94, "learning_rate": 2.438368326334733e-06, "loss": 0.0006, "step": 7852 }, { "epoch": 0.94, "learning_rate": 2.4382963407318537e-06, "loss": 0.0008, "step": 7853 }, { "epoch": 0.94, "learning_rate": 2.438224355128974e-06, "loss": 0.0008, "step": 7854 }, { "epoch": 0.94, "learning_rate": 2.438152369526095e-06, "loss": 0.0005, "step": 7855 }, { "epoch": 0.94, "learning_rate": 2.4380803839232156e-06, "loss": 0.0004, "step": 7856 }, { "epoch": 0.94, "learning_rate": 2.438008398320336e-06, "loss": 0.0004, "step": 7857 }, { "epoch": 0.94, "learning_rate": 2.4379364127174567e-06, "loss": 1.0068, "step": 7858 }, { "epoch": 0.94, "learning_rate": 2.437864427114577e-06, "loss": 0.0004, "step": 7859 }, { "epoch": 0.94, "learning_rate": 2.437792441511698e-06, "loss": 0.0005, "step": 7860 }, { "epoch": 0.94, "learning_rate": 2.4377204559088182e-06, "loss": 0.0009, "step": 7861 }, { "epoch": 0.94, "learning_rate": 2.437648470305939e-06, "loss": 0.0011, "step": 7862 }, { "epoch": 0.94, "learning_rate": 2.4375764847030594e-06, "loss": 0.0007, "step": 7863 }, { "epoch": 0.94, "learning_rate": 2.43750449910018e-06, "loss": 0.001, "step": 7864 }, { "epoch": 0.94, "learning_rate": 2.4374325134973005e-06, "loss": 0.0011, "step": 7865 }, { "epoch": 0.94, "learning_rate": 2.4373605278944213e-06, "loss": 0.0006, "step": 7866 }, { "epoch": 0.94, "learning_rate": 2.4372885422915416e-06, "loss": 0.0004, "step": 7867 }, { "epoch": 0.94, "learning_rate": 2.437216556688662e-06, "loss": 0.0005, "step": 7868 }, { "epoch": 0.94, "learning_rate": 2.4371445710857828e-06, "loss": 0.0009, "step": 7869 }, { "epoch": 0.94, "learning_rate": 2.4370725854829035e-06, "loss": 0.0005, "step": 7870 }, { "epoch": 0.94, "learning_rate": 2.437000599880024e-06, "loss": 0.0005, "step": 7871 }, { "epoch": 0.94, "learning_rate": 2.4369286142771447e-06, "loss": 0.0264, "step": 7872 }, { "epoch": 0.94, "learning_rate": 2.4368566286742655e-06, "loss": 0.0005, "step": 7873 }, { "epoch": 0.94, "learning_rate": 2.436784643071386e-06, "loss": 0.0017, "step": 7874 }, { "epoch": 0.94, "learning_rate": 2.4367126574685066e-06, "loss": 0.0003, "step": 7875 }, { "epoch": 0.94, "learning_rate": 2.436640671865627e-06, "loss": 0.0008, "step": 7876 }, { "epoch": 0.94, "learning_rate": 2.4365686862627477e-06, "loss": 0.0007, "step": 7877 }, { "epoch": 0.94, "learning_rate": 2.436496700659868e-06, "loss": 0.0003, "step": 7878 }, { "epoch": 0.94, "learning_rate": 2.436424715056989e-06, "loss": 0.0005, "step": 7879 }, { "epoch": 0.94, "learning_rate": 2.4363527294541092e-06, "loss": 0.0054, "step": 7880 }, { "epoch": 0.94, "learning_rate": 2.4362807438512296e-06, "loss": 0.3423, "step": 7881 }, { "epoch": 0.94, "learning_rate": 2.4362087582483503e-06, "loss": 0.0004, "step": 7882 }, { "epoch": 0.94, "learning_rate": 2.4361367726454707e-06, "loss": 0.0006, "step": 7883 }, { "epoch": 0.94, "learning_rate": 2.4360647870425915e-06, "loss": 0.0005, "step": 7884 }, { "epoch": 0.94, "learning_rate": 2.435992801439712e-06, "loss": 0.0012, "step": 7885 }, { "epoch": 0.94, "learning_rate": 2.4359208158368326e-06, "loss": 0.0004, "step": 7886 }, { "epoch": 0.95, "learning_rate": 2.4358488302339534e-06, "loss": 0.6143, "step": 7887 }, { "epoch": 0.95, "learning_rate": 2.435776844631074e-06, "loss": 0.0005, "step": 7888 }, { "epoch": 0.95, "learning_rate": 2.4357048590281945e-06, "loss": 0.0006, "step": 7889 }, { "epoch": 0.95, "learning_rate": 2.4356328734253153e-06, "loss": 0.0004, "step": 7890 }, { "epoch": 0.95, "learning_rate": 2.4355608878224357e-06, "loss": 0.6206, "step": 7891 }, { "epoch": 0.95, "learning_rate": 2.435488902219556e-06, "loss": 0.0004, "step": 7892 }, { "epoch": 0.95, "learning_rate": 2.435416916616677e-06, "loss": 0.0005, "step": 7893 }, { "epoch": 0.95, "learning_rate": 2.435344931013797e-06, "loss": 0.0003, "step": 7894 }, { "epoch": 0.95, "learning_rate": 2.435272945410918e-06, "loss": 0.1123, "step": 7895 }, { "epoch": 0.95, "learning_rate": 2.4352009598080383e-06, "loss": 0.001, "step": 7896 }, { "epoch": 0.95, "learning_rate": 2.435128974205159e-06, "loss": 0.0006, "step": 7897 }, { "epoch": 0.95, "learning_rate": 2.4350569886022794e-06, "loss": 0.0143, "step": 7898 }, { "epoch": 0.95, "learning_rate": 2.4349850029994e-06, "loss": 0.8945, "step": 7899 }, { "epoch": 0.95, "learning_rate": 2.4349130173965205e-06, "loss": 0.0004, "step": 7900 }, { "epoch": 0.95, "learning_rate": 2.4348410317936413e-06, "loss": 0.0003, "step": 7901 }, { "epoch": 0.95, "learning_rate": 2.434769046190762e-06, "loss": 0.0005, "step": 7902 }, { "epoch": 0.95, "learning_rate": 2.4346970605878825e-06, "loss": 0.0004, "step": 7903 }, { "epoch": 0.95, "learning_rate": 2.4346250749850032e-06, "loss": 0.0004, "step": 7904 }, { "epoch": 0.95, "learning_rate": 2.4345530893821236e-06, "loss": 0.0007, "step": 7905 }, { "epoch": 0.95, "learning_rate": 2.4344811037792444e-06, "loss": 0.7685, "step": 7906 }, { "epoch": 0.95, "learning_rate": 2.4344091181763647e-06, "loss": 0.0003, "step": 7907 }, { "epoch": 0.95, "learning_rate": 2.4343371325734855e-06, "loss": 0.5698, "step": 7908 }, { "epoch": 0.95, "learning_rate": 2.434265146970606e-06, "loss": 0.0005, "step": 7909 }, { "epoch": 0.95, "learning_rate": 2.4341931613677266e-06, "loss": 0.0007, "step": 7910 }, { "epoch": 0.95, "learning_rate": 2.434121175764847e-06, "loss": 0.0005, "step": 7911 }, { "epoch": 0.95, "learning_rate": 2.4340491901619678e-06, "loss": 0.0004, "step": 7912 }, { "epoch": 0.95, "learning_rate": 2.433977204559088e-06, "loss": 0.0168, "step": 7913 }, { "epoch": 0.95, "learning_rate": 2.4339052189562085e-06, "loss": 0.0005, "step": 7914 }, { "epoch": 0.95, "learning_rate": 2.4338332333533293e-06, "loss": 0.0039, "step": 7915 }, { "epoch": 0.95, "learning_rate": 2.43376124775045e-06, "loss": 0.0022, "step": 7916 }, { "epoch": 0.95, "learning_rate": 2.4336892621475704e-06, "loss": 0.0004, "step": 7917 }, { "epoch": 0.95, "learning_rate": 2.433617276544691e-06, "loss": 0.7114, "step": 7918 }, { "epoch": 0.95, "learning_rate": 2.433545290941812e-06, "loss": 0.2912, "step": 7919 }, { "epoch": 0.95, "learning_rate": 2.4334733053389323e-06, "loss": 0.0009, "step": 7920 }, { "epoch": 0.95, "learning_rate": 2.433401319736053e-06, "loss": 0.0015, "step": 7921 }, { "epoch": 0.95, "learning_rate": 2.4333293341331734e-06, "loss": 0.0008, "step": 7922 }, { "epoch": 0.95, "learning_rate": 2.4332573485302942e-06, "loss": 0.0014, "step": 7923 }, { "epoch": 0.95, "learning_rate": 2.4331853629274146e-06, "loss": 0.0006, "step": 7924 }, { "epoch": 0.95, "learning_rate": 2.4331133773245354e-06, "loss": 0.0017, "step": 7925 }, { "epoch": 0.95, "learning_rate": 2.4330413917216557e-06, "loss": 0.0005, "step": 7926 }, { "epoch": 0.95, "learning_rate": 2.432969406118776e-06, "loss": 0.0026, "step": 7927 }, { "epoch": 0.95, "learning_rate": 2.432897420515897e-06, "loss": 0.6396, "step": 7928 }, { "epoch": 0.95, "learning_rate": 2.432825434913017e-06, "loss": 0.4997, "step": 7929 }, { "epoch": 0.95, "learning_rate": 2.432753449310138e-06, "loss": 0.0004, "step": 7930 }, { "epoch": 0.95, "learning_rate": 2.4326814637072583e-06, "loss": 1.1001, "step": 7931 }, { "epoch": 0.95, "learning_rate": 2.432609478104379e-06, "loss": 0.0009, "step": 7932 }, { "epoch": 0.95, "learning_rate": 2.4325374925015e-06, "loss": 0.0004, "step": 7933 }, { "epoch": 0.95, "learning_rate": 2.4324655068986207e-06, "loss": 0.0004, "step": 7934 }, { "epoch": 0.95, "learning_rate": 2.432393521295741e-06, "loss": 0.0009, "step": 7935 }, { "epoch": 0.95, "learning_rate": 2.432321535692862e-06, "loss": 0.0006, "step": 7936 }, { "epoch": 0.95, "learning_rate": 2.432249550089982e-06, "loss": 0.0004, "step": 7937 }, { "epoch": 0.95, "learning_rate": 2.4321775644871025e-06, "loss": 0.0008, "step": 7938 }, { "epoch": 0.95, "learning_rate": 2.4321055788842233e-06, "loss": 0.0003, "step": 7939 }, { "epoch": 0.95, "learning_rate": 2.4320335932813436e-06, "loss": 0.052, "step": 7940 }, { "epoch": 0.95, "learning_rate": 2.4319616076784644e-06, "loss": 0.0003, "step": 7941 }, { "epoch": 0.95, "learning_rate": 2.4318896220755848e-06, "loss": 0.0004, "step": 7942 }, { "epoch": 0.95, "learning_rate": 2.4318176364727056e-06, "loss": 0.0005, "step": 7943 }, { "epoch": 0.95, "learning_rate": 2.431745650869826e-06, "loss": 0.2588, "step": 7944 }, { "epoch": 0.95, "learning_rate": 2.4316736652669467e-06, "loss": 0.0008, "step": 7945 }, { "epoch": 0.95, "learning_rate": 2.431601679664067e-06, "loss": 0.0134, "step": 7946 }, { "epoch": 0.95, "learning_rate": 2.431529694061188e-06, "loss": 0.0008, "step": 7947 }, { "epoch": 0.95, "learning_rate": 2.4314577084583086e-06, "loss": 0.002, "step": 7948 }, { "epoch": 0.95, "learning_rate": 2.431385722855429e-06, "loss": 0.0004, "step": 7949 }, { "epoch": 0.95, "learning_rate": 2.4313137372525497e-06, "loss": 0.0006, "step": 7950 }, { "epoch": 0.95, "learning_rate": 2.43124175164967e-06, "loss": 0.0003, "step": 7951 }, { "epoch": 0.95, "learning_rate": 2.431169766046791e-06, "loss": 0.0411, "step": 7952 }, { "epoch": 0.95, "learning_rate": 2.4310977804439112e-06, "loss": 0.0018, "step": 7953 }, { "epoch": 0.95, "learning_rate": 2.431025794841032e-06, "loss": 0.0003, "step": 7954 }, { "epoch": 0.95, "learning_rate": 2.4309538092381524e-06, "loss": 0.6117, "step": 7955 }, { "epoch": 0.95, "learning_rate": 2.430881823635273e-06, "loss": 0.0005, "step": 7956 }, { "epoch": 0.95, "learning_rate": 2.4308098380323935e-06, "loss": 0.0004, "step": 7957 }, { "epoch": 0.95, "learning_rate": 2.4307378524295143e-06, "loss": 0.0036, "step": 7958 }, { "epoch": 0.95, "learning_rate": 2.4306658668266346e-06, "loss": 0.0004, "step": 7959 }, { "epoch": 0.95, "learning_rate": 2.4305938812237554e-06, "loss": 0.0004, "step": 7960 }, { "epoch": 0.95, "learning_rate": 2.4305218956208758e-06, "loss": 0.0004, "step": 7961 }, { "epoch": 0.95, "learning_rate": 2.430449910017996e-06, "loss": 0.0005, "step": 7962 }, { "epoch": 0.95, "learning_rate": 2.430377924415117e-06, "loss": 0.0006, "step": 7963 }, { "epoch": 0.95, "learning_rate": 2.4303059388122377e-06, "loss": 0.0007, "step": 7964 }, { "epoch": 0.95, "learning_rate": 2.4302339532093585e-06, "loss": 0.0005, "step": 7965 }, { "epoch": 0.95, "learning_rate": 2.430161967606479e-06, "loss": 0.9683, "step": 7966 }, { "epoch": 0.95, "learning_rate": 2.4300899820035996e-06, "loss": 0.0005, "step": 7967 }, { "epoch": 0.95, "learning_rate": 2.43001799640072e-06, "loss": 0.0004, "step": 7968 }, { "epoch": 0.95, "learning_rate": 2.4299460107978407e-06, "loss": 0.3352, "step": 7969 }, { "epoch": 0.96, "learning_rate": 2.429874025194961e-06, "loss": 0.0007, "step": 7970 }, { "epoch": 0.96, "learning_rate": 2.429802039592082e-06, "loss": 0.0015, "step": 7971 }, { "epoch": 0.96, "learning_rate": 2.429730053989202e-06, "loss": 0.0264, "step": 7972 }, { "epoch": 0.96, "learning_rate": 2.4296580683863226e-06, "loss": 0.0006, "step": 7973 }, { "epoch": 0.96, "learning_rate": 2.4295860827834433e-06, "loss": 0.0004, "step": 7974 }, { "epoch": 0.96, "learning_rate": 2.4295140971805637e-06, "loss": 0.0004, "step": 7975 }, { "epoch": 0.96, "learning_rate": 2.4294421115776845e-06, "loss": 0.0003, "step": 7976 }, { "epoch": 0.96, "learning_rate": 2.429370125974805e-06, "loss": 0.0003, "step": 7977 }, { "epoch": 0.96, "learning_rate": 2.4292981403719256e-06, "loss": 0.0004, "step": 7978 }, { "epoch": 0.96, "learning_rate": 2.4292261547690464e-06, "loss": 0.0006, "step": 7979 }, { "epoch": 0.96, "learning_rate": 2.4291541691661667e-06, "loss": 0.0004, "step": 7980 }, { "epoch": 0.96, "learning_rate": 2.4290821835632875e-06, "loss": 0.6819, "step": 7981 }, { "epoch": 0.96, "learning_rate": 2.4290101979604083e-06, "loss": 0.0007, "step": 7982 }, { "epoch": 0.96, "learning_rate": 2.4289382123575287e-06, "loss": 0.0005, "step": 7983 }, { "epoch": 0.96, "learning_rate": 2.428866226754649e-06, "loss": 0.1889, "step": 7984 }, { "epoch": 0.96, "learning_rate": 2.4287942411517698e-06, "loss": 0.0007, "step": 7985 }, { "epoch": 0.96, "learning_rate": 2.42872225554889e-06, "loss": 0.0004, "step": 7986 }, { "epoch": 0.96, "learning_rate": 2.428650269946011e-06, "loss": 0.0004, "step": 7987 }, { "epoch": 0.96, "learning_rate": 2.4285782843431313e-06, "loss": 0.0004, "step": 7988 }, { "epoch": 0.96, "learning_rate": 2.428506298740252e-06, "loss": 0.0003, "step": 7989 }, { "epoch": 0.96, "learning_rate": 2.4284343131373724e-06, "loss": 0.0004, "step": 7990 }, { "epoch": 0.96, "learning_rate": 2.428362327534493e-06, "loss": 0.0005, "step": 7991 }, { "epoch": 0.96, "learning_rate": 2.4282903419316135e-06, "loss": 0.0007, "step": 7992 }, { "epoch": 0.96, "learning_rate": 2.4282183563287343e-06, "loss": 0.5684, "step": 7993 }, { "epoch": 0.96, "learning_rate": 2.4281463707258547e-06, "loss": 0.0003, "step": 7994 }, { "epoch": 0.96, "learning_rate": 2.4280743851229755e-06, "loss": 0.0005, "step": 7995 }, { "epoch": 0.96, "learning_rate": 2.4280023995200962e-06, "loss": 0.0004, "step": 7996 }, { "epoch": 0.96, "learning_rate": 2.4279304139172166e-06, "loss": 0.0007, "step": 7997 }, { "epoch": 0.96, "learning_rate": 2.4278584283143374e-06, "loss": 0.0004, "step": 7998 }, { "epoch": 0.96, "learning_rate": 2.4277864427114577e-06, "loss": 0.0005, "step": 7999 }, { "epoch": 0.96, "learning_rate": 2.4277144571085785e-06, "loss": 0.0004, "step": 8000 }, { "epoch": 0.96, "learning_rate": 2.427642471505699e-06, "loss": 0.905, "step": 8001 }, { "epoch": 0.96, "learning_rate": 2.4275704859028196e-06, "loss": 0.0003, "step": 8002 }, { "epoch": 0.96, "learning_rate": 2.42749850029994e-06, "loss": 0.0003, "step": 8003 }, { "epoch": 0.96, "learning_rate": 2.4274265146970608e-06, "loss": 0.0006, "step": 8004 }, { "epoch": 0.96, "learning_rate": 2.427354529094181e-06, "loss": 0.0005, "step": 8005 }, { "epoch": 0.96, "learning_rate": 2.427282543491302e-06, "loss": 0.0003, "step": 8006 }, { "epoch": 0.96, "learning_rate": 2.4272105578884223e-06, "loss": 0.0007, "step": 8007 }, { "epoch": 0.96, "learning_rate": 2.4271385722855426e-06, "loss": 0.0003, "step": 8008 }, { "epoch": 0.96, "learning_rate": 2.4270665866826634e-06, "loss": 0.0006, "step": 8009 }, { "epoch": 0.96, "learning_rate": 2.426994601079784e-06, "loss": 0.0004, "step": 8010 }, { "epoch": 0.96, "learning_rate": 2.426922615476905e-06, "loss": 0.0004, "step": 8011 }, { "epoch": 0.96, "learning_rate": 2.4268506298740253e-06, "loss": 0.0004, "step": 8012 }, { "epoch": 0.96, "learning_rate": 2.426778644271146e-06, "loss": 0.0006, "step": 8013 }, { "epoch": 0.96, "learning_rate": 2.4267066586682664e-06, "loss": 0.0003, "step": 8014 }, { "epoch": 0.96, "learning_rate": 2.4266346730653872e-06, "loss": 0.0011, "step": 8015 }, { "epoch": 0.96, "learning_rate": 2.4265626874625076e-06, "loss": 0.0005, "step": 8016 }, { "epoch": 0.96, "learning_rate": 2.4264907018596283e-06, "loss": 0.0008, "step": 8017 }, { "epoch": 0.96, "learning_rate": 2.4264187162567487e-06, "loss": 0.0003, "step": 8018 }, { "epoch": 0.96, "learning_rate": 2.426346730653869e-06, "loss": 0.0012, "step": 8019 }, { "epoch": 0.96, "learning_rate": 2.42627474505099e-06, "loss": 0.0003, "step": 8020 }, { "epoch": 0.96, "learning_rate": 2.42620275944811e-06, "loss": 0.0004, "step": 8021 }, { "epoch": 0.96, "learning_rate": 2.426130773845231e-06, "loss": 0.0244, "step": 8022 }, { "epoch": 0.96, "learning_rate": 2.4260587882423513e-06, "loss": 0.0004, "step": 8023 }, { "epoch": 0.96, "learning_rate": 2.425986802639472e-06, "loss": 0.0004, "step": 8024 }, { "epoch": 0.96, "learning_rate": 2.425914817036593e-06, "loss": 0.003, "step": 8025 }, { "epoch": 0.96, "learning_rate": 2.4258428314337132e-06, "loss": 0.0004, "step": 8026 }, { "epoch": 0.96, "learning_rate": 2.425770845830834e-06, "loss": 0.0003, "step": 8027 }, { "epoch": 0.96, "learning_rate": 2.425698860227955e-06, "loss": 1.0469, "step": 8028 }, { "epoch": 0.96, "learning_rate": 2.425626874625075e-06, "loss": 0.0005, "step": 8029 }, { "epoch": 0.96, "learning_rate": 2.4255548890221955e-06, "loss": 0.0005, "step": 8030 }, { "epoch": 0.96, "learning_rate": 2.4254829034193163e-06, "loss": 0.0005, "step": 8031 }, { "epoch": 0.96, "learning_rate": 2.4254109178164366e-06, "loss": 0.0005, "step": 8032 }, { "epoch": 0.96, "learning_rate": 2.4253389322135574e-06, "loss": 0.0003, "step": 8033 }, { "epoch": 0.96, "learning_rate": 2.4252669466106778e-06, "loss": 0.0007, "step": 8034 }, { "epoch": 0.96, "learning_rate": 2.4251949610077985e-06, "loss": 0.0004, "step": 8035 }, { "epoch": 0.96, "learning_rate": 2.425122975404919e-06, "loss": 0.0007, "step": 8036 }, { "epoch": 0.96, "learning_rate": 2.4250509898020397e-06, "loss": 0.0008, "step": 8037 }, { "epoch": 0.96, "learning_rate": 2.42497900419916e-06, "loss": 0.5384, "step": 8038 }, { "epoch": 0.96, "learning_rate": 2.424907018596281e-06, "loss": 0.0004, "step": 8039 }, { "epoch": 0.96, "learning_rate": 2.424835032993401e-06, "loss": 0.0003, "step": 8040 }, { "epoch": 0.96, "learning_rate": 2.424763047390522e-06, "loss": 0.0005, "step": 8041 }, { "epoch": 0.96, "learning_rate": 2.4246910617876427e-06, "loss": 0.0006, "step": 8042 }, { "epoch": 0.96, "learning_rate": 2.424619076184763e-06, "loss": 0.0005, "step": 8043 }, { "epoch": 0.96, "learning_rate": 2.424547090581884e-06, "loss": 0.0003, "step": 8044 }, { "epoch": 0.96, "learning_rate": 2.4244751049790042e-06, "loss": 0.0004, "step": 8045 }, { "epoch": 0.96, "learning_rate": 2.424403119376125e-06, "loss": 0.7025, "step": 8046 }, { "epoch": 0.96, "learning_rate": 2.4243311337732454e-06, "loss": 0.0005, "step": 8047 }, { "epoch": 0.96, "learning_rate": 2.424259148170366e-06, "loss": 0.0005, "step": 8048 }, { "epoch": 0.96, "learning_rate": 2.4241871625674865e-06, "loss": 0.0008, "step": 8049 }, { "epoch": 0.96, "learning_rate": 2.4241151769646073e-06, "loss": 0.0004, "step": 8050 }, { "epoch": 0.96, "learning_rate": 2.4240431913617276e-06, "loss": 0.0019, "step": 8051 }, { "epoch": 0.96, "learning_rate": 2.4239712057588484e-06, "loss": 0.5051, "step": 8052 }, { "epoch": 0.97, "learning_rate": 2.4238992201559688e-06, "loss": 0.0009, "step": 8053 }, { "epoch": 0.97, "learning_rate": 2.423827234553089e-06, "loss": 0.0325, "step": 8054 }, { "epoch": 0.97, "learning_rate": 2.42375524895021e-06, "loss": 0.0007, "step": 8055 }, { "epoch": 0.97, "learning_rate": 2.4236832633473307e-06, "loss": 0.0027, "step": 8056 }, { "epoch": 0.97, "learning_rate": 2.4236112777444514e-06, "loss": 0.0003, "step": 8057 }, { "epoch": 0.97, "learning_rate": 2.423539292141572e-06, "loss": 0.0009, "step": 8058 }, { "epoch": 0.97, "learning_rate": 2.4234673065386926e-06, "loss": 0.0005, "step": 8059 }, { "epoch": 0.97, "learning_rate": 2.423395320935813e-06, "loss": 0.7125, "step": 8060 }, { "epoch": 0.97, "learning_rate": 2.4233233353329337e-06, "loss": 0.1971, "step": 8061 }, { "epoch": 0.97, "learning_rate": 2.423251349730054e-06, "loss": 0.0007, "step": 8062 }, { "epoch": 0.97, "learning_rate": 2.423179364127175e-06, "loss": 0.0004, "step": 8063 }, { "epoch": 0.97, "learning_rate": 2.423107378524295e-06, "loss": 0.0004, "step": 8064 }, { "epoch": 0.97, "learning_rate": 2.4230353929214156e-06, "loss": 0.002, "step": 8065 }, { "epoch": 0.97, "learning_rate": 2.4229634073185363e-06, "loss": 0.0004, "step": 8066 }, { "epoch": 0.97, "learning_rate": 2.4228914217156567e-06, "loss": 0.001, "step": 8067 }, { "epoch": 0.97, "learning_rate": 2.4228194361127775e-06, "loss": 0.0003, "step": 8068 }, { "epoch": 0.97, "learning_rate": 2.422747450509898e-06, "loss": 0.0004, "step": 8069 }, { "epoch": 0.97, "learning_rate": 2.4226754649070186e-06, "loss": 0.0006, "step": 8070 }, { "epoch": 0.97, "learning_rate": 2.4226034793041394e-06, "loss": 0.0013, "step": 8071 }, { "epoch": 0.97, "learning_rate": 2.4225314937012597e-06, "loss": 0.0005, "step": 8072 }, { "epoch": 0.97, "learning_rate": 2.4224595080983805e-06, "loss": 0.0006, "step": 8073 }, { "epoch": 0.97, "learning_rate": 2.4223875224955013e-06, "loss": 0.0005, "step": 8074 }, { "epoch": 0.97, "learning_rate": 2.4223155368926216e-06, "loss": 0.0006, "step": 8075 }, { "epoch": 0.97, "learning_rate": 2.422243551289742e-06, "loss": 0.0003, "step": 8076 }, { "epoch": 0.97, "learning_rate": 2.4221715656868628e-06, "loss": 0.0124, "step": 8077 }, { "epoch": 0.97, "learning_rate": 2.422099580083983e-06, "loss": 0.0005, "step": 8078 }, { "epoch": 0.97, "learning_rate": 2.422027594481104e-06, "loss": 0.001, "step": 8079 }, { "epoch": 0.97, "learning_rate": 2.4219556088782243e-06, "loss": 0.0004, "step": 8080 }, { "epoch": 0.97, "learning_rate": 2.421883623275345e-06, "loss": 0.0003, "step": 8081 }, { "epoch": 0.97, "learning_rate": 2.4218116376724654e-06, "loss": 0.0005, "step": 8082 }, { "epoch": 0.97, "learning_rate": 2.421739652069586e-06, "loss": 0.0004, "step": 8083 }, { "epoch": 0.97, "learning_rate": 2.4216676664667065e-06, "loss": 0.9091, "step": 8084 }, { "epoch": 0.97, "learning_rate": 2.4215956808638273e-06, "loss": 0.0003, "step": 8085 }, { "epoch": 0.97, "learning_rate": 2.4215236952609477e-06, "loss": 0.0007, "step": 8086 }, { "epoch": 0.97, "learning_rate": 2.4214517096580684e-06, "loss": 0.0003, "step": 8087 }, { "epoch": 0.97, "learning_rate": 2.4213797240551892e-06, "loss": 0.0005, "step": 8088 }, { "epoch": 0.97, "learning_rate": 2.4213077384523096e-06, "loss": 0.0007, "step": 8089 }, { "epoch": 0.97, "learning_rate": 2.4212357528494304e-06, "loss": 0.0009, "step": 8090 }, { "epoch": 0.97, "learning_rate": 2.4211637672465507e-06, "loss": 0.0004, "step": 8091 }, { "epoch": 0.97, "learning_rate": 2.4210917816436715e-06, "loss": 0.0005, "step": 8092 }, { "epoch": 0.97, "learning_rate": 2.421019796040792e-06, "loss": 0.0005, "step": 8093 }, { "epoch": 0.97, "learning_rate": 2.4209478104379126e-06, "loss": 0.0004, "step": 8094 }, { "epoch": 0.97, "learning_rate": 2.420875824835033e-06, "loss": 0.0152, "step": 8095 }, { "epoch": 0.97, "learning_rate": 2.4208038392321538e-06, "loss": 0.0005, "step": 8096 }, { "epoch": 0.97, "learning_rate": 2.420731853629274e-06, "loss": 0.0003, "step": 8097 }, { "epoch": 0.97, "learning_rate": 2.420659868026395e-06, "loss": 0.0006, "step": 8098 }, { "epoch": 0.97, "learning_rate": 2.4205878824235152e-06, "loss": 0.0005, "step": 8099 }, { "epoch": 0.97, "learning_rate": 2.4205158968206356e-06, "loss": 0.0008, "step": 8100 }, { "epoch": 0.97, "learning_rate": 2.4204439112177564e-06, "loss": 0.0043, "step": 8101 }, { "epoch": 0.97, "learning_rate": 2.420371925614877e-06, "loss": 0.0048, "step": 8102 }, { "epoch": 0.97, "learning_rate": 2.4202999400119975e-06, "loss": 0.0004, "step": 8103 }, { "epoch": 0.97, "learning_rate": 2.4202279544091183e-06, "loss": 0.0005, "step": 8104 }, { "epoch": 0.97, "learning_rate": 2.420155968806239e-06, "loss": 0.0006, "step": 8105 }, { "epoch": 0.97, "learning_rate": 2.4200839832033594e-06, "loss": 0.8776, "step": 8106 }, { "epoch": 0.97, "learning_rate": 2.42001199760048e-06, "loss": 0.0003, "step": 8107 }, { "epoch": 0.97, "learning_rate": 2.4199400119976006e-06, "loss": 0.7737, "step": 8108 }, { "epoch": 0.97, "learning_rate": 2.4198680263947213e-06, "loss": 0.0004, "step": 8109 }, { "epoch": 0.97, "learning_rate": 2.4197960407918417e-06, "loss": 0.0007, "step": 8110 }, { "epoch": 0.97, "learning_rate": 2.419724055188962e-06, "loss": 0.0006, "step": 8111 }, { "epoch": 0.97, "learning_rate": 2.419652069586083e-06, "loss": 0.0011, "step": 8112 }, { "epoch": 0.97, "learning_rate": 2.419580083983203e-06, "loss": 0.0004, "step": 8113 }, { "epoch": 0.97, "learning_rate": 2.419508098380324e-06, "loss": 0.004, "step": 8114 }, { "epoch": 0.97, "learning_rate": 2.4194361127774443e-06, "loss": 0.0015, "step": 8115 }, { "epoch": 0.97, "learning_rate": 2.419364127174565e-06, "loss": 0.0034, "step": 8116 }, { "epoch": 0.97, "learning_rate": 2.4192921415716854e-06, "loss": 0.0005, "step": 8117 }, { "epoch": 0.97, "learning_rate": 2.4192201559688062e-06, "loss": 0.0005, "step": 8118 }, { "epoch": 0.97, "learning_rate": 2.419148170365927e-06, "loss": 0.0004, "step": 8119 }, { "epoch": 0.97, "learning_rate": 2.4190761847630478e-06, "loss": 0.0006, "step": 8120 }, { "epoch": 0.97, "learning_rate": 2.419004199160168e-06, "loss": 0.0007, "step": 8121 }, { "epoch": 0.97, "learning_rate": 2.4189322135572885e-06, "loss": 0.0004, "step": 8122 }, { "epoch": 0.97, "learning_rate": 2.4188602279544093e-06, "loss": 0.4344, "step": 8123 }, { "epoch": 0.97, "learning_rate": 2.4187882423515296e-06, "loss": 0.0003, "step": 8124 }, { "epoch": 0.97, "learning_rate": 2.4187162567486504e-06, "loss": 0.0006, "step": 8125 }, { "epoch": 0.97, "learning_rate": 2.4186442711457708e-06, "loss": 0.0004, "step": 8126 }, { "epoch": 0.97, "learning_rate": 2.4185722855428915e-06, "loss": 0.0005, "step": 8127 }, { "epoch": 0.97, "learning_rate": 2.418500299940012e-06, "loss": 0.0004, "step": 8128 }, { "epoch": 0.97, "learning_rate": 2.4184283143371327e-06, "loss": 0.0004, "step": 8129 }, { "epoch": 0.97, "learning_rate": 2.418356328734253e-06, "loss": 0.0007, "step": 8130 }, { "epoch": 0.97, "learning_rate": 2.418284343131374e-06, "loss": 0.0006, "step": 8131 }, { "epoch": 0.97, "learning_rate": 2.418212357528494e-06, "loss": 0.0003, "step": 8132 }, { "epoch": 0.97, "learning_rate": 2.418140371925615e-06, "loss": 0.0003, "step": 8133 }, { "epoch": 0.97, "learning_rate": 2.4180683863227357e-06, "loss": 0.0005, "step": 8134 }, { "epoch": 0.97, "learning_rate": 2.417996400719856e-06, "loss": 0.4538, "step": 8135 }, { "epoch": 0.97, "learning_rate": 2.417924415116977e-06, "loss": 0.0003, "step": 8136 }, { "epoch": 0.98, "learning_rate": 2.417852429514097e-06, "loss": 0.0007, "step": 8137 }, { "epoch": 0.98, "learning_rate": 2.417780443911218e-06, "loss": 0.0063, "step": 8138 }, { "epoch": 0.98, "learning_rate": 2.4177084583083383e-06, "loss": 0.0004, "step": 8139 }, { "epoch": 0.98, "learning_rate": 2.417636472705459e-06, "loss": 0.7749, "step": 8140 }, { "epoch": 0.98, "learning_rate": 2.4175644871025795e-06, "loss": 0.0007, "step": 8141 }, { "epoch": 0.98, "learning_rate": 2.4174925014997003e-06, "loss": 0.2134, "step": 8142 }, { "epoch": 0.98, "learning_rate": 2.4174205158968206e-06, "loss": 0.0004, "step": 8143 }, { "epoch": 0.98, "learning_rate": 2.4173485302939414e-06, "loss": 0.0009, "step": 8144 }, { "epoch": 0.98, "learning_rate": 2.4172765446910617e-06, "loss": 0.0039, "step": 8145 }, { "epoch": 0.98, "learning_rate": 2.417204559088182e-06, "loss": 0.0004, "step": 8146 }, { "epoch": 0.98, "learning_rate": 2.417132573485303e-06, "loss": 0.0005, "step": 8147 }, { "epoch": 0.98, "learning_rate": 2.4170605878824237e-06, "loss": 0.1956, "step": 8148 }, { "epoch": 0.98, "learning_rate": 2.416988602279544e-06, "loss": 0.0017, "step": 8149 }, { "epoch": 0.98, "learning_rate": 2.416916616676665e-06, "loss": 1.1083, "step": 8150 }, { "epoch": 0.98, "learning_rate": 2.4168446310737856e-06, "loss": 0.0005, "step": 8151 }, { "epoch": 0.98, "learning_rate": 2.416772645470906e-06, "loss": 0.0004, "step": 8152 }, { "epoch": 0.98, "learning_rate": 2.4167006598680267e-06, "loss": 0.834, "step": 8153 }, { "epoch": 0.98, "learning_rate": 2.416628674265147e-06, "loss": 0.0005, "step": 8154 }, { "epoch": 0.98, "learning_rate": 2.416556688662268e-06, "loss": 0.0004, "step": 8155 }, { "epoch": 0.98, "learning_rate": 2.416484703059388e-06, "loss": 0.0052, "step": 8156 }, { "epoch": 0.98, "learning_rate": 2.4164127174565085e-06, "loss": 0.0007, "step": 8157 }, { "epoch": 0.98, "learning_rate": 2.4163407318536293e-06, "loss": 0.2387, "step": 8158 }, { "epoch": 0.98, "learning_rate": 2.4162687462507497e-06, "loss": 0.0011, "step": 8159 }, { "epoch": 0.98, "learning_rate": 2.4161967606478705e-06, "loss": 0.0004, "step": 8160 }, { "epoch": 0.98, "learning_rate": 2.416124775044991e-06, "loss": 0.0003, "step": 8161 }, { "epoch": 0.98, "learning_rate": 2.4160527894421116e-06, "loss": 0.0006, "step": 8162 }, { "epoch": 0.98, "learning_rate": 2.415980803839232e-06, "loss": 0.0005, "step": 8163 }, { "epoch": 0.98, "learning_rate": 2.4159088182363527e-06, "loss": 0.0654, "step": 8164 }, { "epoch": 0.98, "learning_rate": 2.4158368326334735e-06, "loss": 0.0003, "step": 8165 }, { "epoch": 0.98, "learning_rate": 2.4157648470305943e-06, "loss": 0.0017, "step": 8166 }, { "epoch": 0.98, "learning_rate": 2.4156928614277146e-06, "loss": 0.0005, "step": 8167 }, { "epoch": 0.98, "learning_rate": 2.415620875824835e-06, "loss": 0.0016, "step": 8168 }, { "epoch": 0.98, "learning_rate": 2.4155488902219558e-06, "loss": 0.0004, "step": 8169 }, { "epoch": 0.98, "learning_rate": 2.415476904619076e-06, "loss": 0.0006, "step": 8170 }, { "epoch": 0.98, "learning_rate": 2.415404919016197e-06, "loss": 0.0004, "step": 8171 }, { "epoch": 0.98, "learning_rate": 2.4153329334133173e-06, "loss": 0.0007, "step": 8172 }, { "epoch": 0.98, "learning_rate": 2.415260947810438e-06, "loss": 0.2933, "step": 8173 }, { "epoch": 0.98, "learning_rate": 2.4151889622075584e-06, "loss": 0.0005, "step": 8174 }, { "epoch": 0.98, "learning_rate": 2.415116976604679e-06, "loss": 0.0009, "step": 8175 }, { "epoch": 0.98, "learning_rate": 2.4150449910017995e-06, "loss": 0.0003, "step": 8176 }, { "epoch": 0.98, "learning_rate": 2.4149730053989203e-06, "loss": 0.6274, "step": 8177 }, { "epoch": 0.98, "learning_rate": 2.4149010197960407e-06, "loss": 0.0003, "step": 8178 }, { "epoch": 0.98, "learning_rate": 2.4148290341931614e-06, "loss": 0.0004, "step": 8179 }, { "epoch": 0.98, "learning_rate": 2.4147570485902822e-06, "loss": 1.6603, "step": 8180 }, { "epoch": 0.98, "learning_rate": 2.4146850629874026e-06, "loss": 0.5227, "step": 8181 }, { "epoch": 0.98, "learning_rate": 2.4146130773845234e-06, "loss": 0.0004, "step": 8182 }, { "epoch": 0.98, "learning_rate": 2.4145410917816437e-06, "loss": 0.0004, "step": 8183 }, { "epoch": 0.98, "learning_rate": 2.4144691061787645e-06, "loss": 0.0006, "step": 8184 }, { "epoch": 0.98, "learning_rate": 2.414397120575885e-06, "loss": 0.0006, "step": 8185 }, { "epoch": 0.98, "learning_rate": 2.4143251349730056e-06, "loss": 0.0004, "step": 8186 }, { "epoch": 0.98, "learning_rate": 2.414253149370126e-06, "loss": 0.0004, "step": 8187 }, { "epoch": 0.98, "learning_rate": 2.4141811637672468e-06, "loss": 0.0006, "step": 8188 }, { "epoch": 0.98, "learning_rate": 2.414109178164367e-06, "loss": 0.0005, "step": 8189 }, { "epoch": 0.98, "learning_rate": 2.414037192561488e-06, "loss": 0.0004, "step": 8190 }, { "epoch": 0.98, "learning_rate": 2.4139652069586082e-06, "loss": 0.796, "step": 8191 }, { "epoch": 0.98, "learning_rate": 2.4138932213557286e-06, "loss": 0.0004, "step": 8192 }, { "epoch": 0.98, "learning_rate": 2.4138212357528494e-06, "loss": 0.0006, "step": 8193 }, { "epoch": 0.98, "learning_rate": 2.41374925014997e-06, "loss": 0.0046, "step": 8194 }, { "epoch": 0.98, "learning_rate": 2.4136772645470905e-06, "loss": 0.5799, "step": 8195 }, { "epoch": 0.98, "learning_rate": 2.4136052789442113e-06, "loss": 0.001, "step": 8196 }, { "epoch": 0.98, "learning_rate": 2.413533293341332e-06, "loss": 0.092, "step": 8197 }, { "epoch": 0.98, "learning_rate": 2.4134613077384524e-06, "loss": 0.0003, "step": 8198 }, { "epoch": 0.98, "learning_rate": 2.413389322135573e-06, "loss": 0.0005, "step": 8199 }, { "epoch": 0.98, "learning_rate": 2.4133173365326936e-06, "loss": 0.0005, "step": 8200 }, { "epoch": 0.98, "learning_rate": 2.4132453509298143e-06, "loss": 0.0004, "step": 8201 }, { "epoch": 0.98, "learning_rate": 2.4131733653269347e-06, "loss": 0.0004, "step": 8202 }, { "epoch": 0.98, "learning_rate": 2.413101379724055e-06, "loss": 0.0005, "step": 8203 }, { "epoch": 0.98, "learning_rate": 2.413029394121176e-06, "loss": 0.8408, "step": 8204 }, { "epoch": 0.98, "learning_rate": 2.412957408518296e-06, "loss": 0.0365, "step": 8205 }, { "epoch": 0.98, "learning_rate": 2.412885422915417e-06, "loss": 0.846, "step": 8206 }, { "epoch": 0.98, "learning_rate": 2.4128134373125373e-06, "loss": 0.0004, "step": 8207 }, { "epoch": 0.98, "learning_rate": 2.412741451709658e-06, "loss": 0.0005, "step": 8208 }, { "epoch": 0.98, "learning_rate": 2.4126694661067784e-06, "loss": 0.0105, "step": 8209 }, { "epoch": 0.98, "learning_rate": 2.4125974805038992e-06, "loss": 0.0009, "step": 8210 }, { "epoch": 0.98, "learning_rate": 2.41252549490102e-06, "loss": 0.0004, "step": 8211 }, { "epoch": 0.98, "learning_rate": 2.4124535092981408e-06, "loss": 0.0003, "step": 8212 }, { "epoch": 0.98, "learning_rate": 2.412381523695261e-06, "loss": 0.0011, "step": 8213 }, { "epoch": 0.98, "learning_rate": 2.412309538092382e-06, "loss": 0.0018, "step": 8214 }, { "epoch": 0.98, "learning_rate": 2.4122375524895023e-06, "loss": 0.0003, "step": 8215 }, { "epoch": 0.98, "learning_rate": 2.4121655668866226e-06, "loss": 0.0006, "step": 8216 }, { "epoch": 0.98, "learning_rate": 2.4120935812837434e-06, "loss": 0.0003, "step": 8217 }, { "epoch": 0.98, "learning_rate": 2.4120215956808638e-06, "loss": 0.0004, "step": 8218 }, { "epoch": 0.98, "learning_rate": 2.4119496100779845e-06, "loss": 0.0006, "step": 8219 }, { "epoch": 0.99, "learning_rate": 2.411877624475105e-06, "loss": 0.0005, "step": 8220 }, { "epoch": 0.99, "learning_rate": 2.4118056388722257e-06, "loss": 0.0012, "step": 8221 }, { "epoch": 0.99, "learning_rate": 2.411733653269346e-06, "loss": 0.0003, "step": 8222 }, { "epoch": 0.99, "learning_rate": 2.411661667666467e-06, "loss": 0.0009, "step": 8223 }, { "epoch": 0.99, "learning_rate": 2.411589682063587e-06, "loss": 0.0004, "step": 8224 }, { "epoch": 0.99, "learning_rate": 2.411517696460708e-06, "loss": 0.0006, "step": 8225 }, { "epoch": 0.99, "learning_rate": 2.4114457108578283e-06, "loss": 0.0003, "step": 8226 }, { "epoch": 0.99, "learning_rate": 2.411373725254949e-06, "loss": 0.0006, "step": 8227 }, { "epoch": 0.99, "learning_rate": 2.41130173965207e-06, "loss": 0.0004, "step": 8228 }, { "epoch": 0.99, "learning_rate": 2.41122975404919e-06, "loss": 0.0005, "step": 8229 }, { "epoch": 0.99, "learning_rate": 2.411157768446311e-06, "loss": 0.0013, "step": 8230 }, { "epoch": 0.99, "learning_rate": 2.4110857828434313e-06, "loss": 0.0006, "step": 8231 }, { "epoch": 0.99, "learning_rate": 2.411013797240552e-06, "loss": 0.0285, "step": 8232 }, { "epoch": 0.99, "learning_rate": 2.4109418116376725e-06, "loss": 0.0005, "step": 8233 }, { "epoch": 0.99, "learning_rate": 2.4108698260347932e-06, "loss": 0.0035, "step": 8234 }, { "epoch": 0.99, "learning_rate": 2.4107978404319136e-06, "loss": 0.0008, "step": 8235 }, { "epoch": 0.99, "learning_rate": 2.4107258548290344e-06, "loss": 0.0003, "step": 8236 }, { "epoch": 0.99, "learning_rate": 2.4106538692261547e-06, "loss": 0.0031, "step": 8237 }, { "epoch": 0.99, "learning_rate": 2.410581883623275e-06, "loss": 0.0005, "step": 8238 }, { "epoch": 0.99, "learning_rate": 2.410509898020396e-06, "loss": 0.7488, "step": 8239 }, { "epoch": 0.99, "learning_rate": 2.4104379124175162e-06, "loss": 0.0008, "step": 8240 }, { "epoch": 0.99, "learning_rate": 2.410365926814637e-06, "loss": 0.0004, "step": 8241 }, { "epoch": 0.99, "learning_rate": 2.4102939412117578e-06, "loss": 0.0007, "step": 8242 }, { "epoch": 0.99, "learning_rate": 2.4102219556088786e-06, "loss": 0.0007, "step": 8243 }, { "epoch": 0.99, "learning_rate": 2.410149970005999e-06, "loss": 0.0177, "step": 8244 }, { "epoch": 0.99, "learning_rate": 2.4100779844031197e-06, "loss": 0.0005, "step": 8245 }, { "epoch": 0.99, "learning_rate": 2.41000599880024e-06, "loss": 0.0004, "step": 8246 }, { "epoch": 0.99, "learning_rate": 2.409934013197361e-06, "loss": 0.0003, "step": 8247 }, { "epoch": 0.99, "learning_rate": 2.409862027594481e-06, "loss": 0.0003, "step": 8248 }, { "epoch": 0.99, "learning_rate": 2.4097900419916015e-06, "loss": 0.0006, "step": 8249 }, { "epoch": 0.99, "learning_rate": 2.4097180563887223e-06, "loss": 0.0011, "step": 8250 }, { "epoch": 0.99, "learning_rate": 2.4096460707858427e-06, "loss": 0.0005, "step": 8251 }, { "epoch": 0.99, "learning_rate": 2.4095740851829634e-06, "loss": 0.0002, "step": 8252 }, { "epoch": 0.99, "learning_rate": 2.409502099580084e-06, "loss": 0.5003, "step": 8253 }, { "epoch": 0.99, "learning_rate": 2.4094301139772046e-06, "loss": 0.0009, "step": 8254 }, { "epoch": 0.99, "learning_rate": 2.409358128374325e-06, "loss": 0.0005, "step": 8255 }, { "epoch": 0.99, "learning_rate": 2.4092861427714457e-06, "loss": 0.0006, "step": 8256 }, { "epoch": 0.99, "learning_rate": 2.4092141571685665e-06, "loss": 0.0003, "step": 8257 }, { "epoch": 0.99, "learning_rate": 2.409142171565687e-06, "loss": 0.0003, "step": 8258 }, { "epoch": 0.99, "learning_rate": 2.4090701859628076e-06, "loss": 0.0005, "step": 8259 }, { "epoch": 0.99, "learning_rate": 2.4089982003599284e-06, "loss": 0.0004, "step": 8260 }, { "epoch": 0.99, "learning_rate": 2.4089262147570488e-06, "loss": 0.0007, "step": 8261 }, { "epoch": 0.99, "learning_rate": 2.408854229154169e-06, "loss": 0.0003, "step": 8262 }, { "epoch": 0.99, "learning_rate": 2.40878224355129e-06, "loss": 0.0004, "step": 8263 }, { "epoch": 0.99, "learning_rate": 2.4087102579484103e-06, "loss": 0.7616, "step": 8264 }, { "epoch": 0.99, "learning_rate": 2.408638272345531e-06, "loss": 0.0003, "step": 8265 }, { "epoch": 0.99, "learning_rate": 2.4085662867426514e-06, "loss": 0.0004, "step": 8266 }, { "epoch": 0.99, "learning_rate": 2.408494301139772e-06, "loss": 0.0007, "step": 8267 }, { "epoch": 0.99, "learning_rate": 2.4084223155368925e-06, "loss": 0.0005, "step": 8268 }, { "epoch": 0.99, "learning_rate": 2.4083503299340133e-06, "loss": 0.0029, "step": 8269 }, { "epoch": 0.99, "learning_rate": 2.4082783443311337e-06, "loss": 0.0004, "step": 8270 }, { "epoch": 0.99, "learning_rate": 2.4082063587282544e-06, "loss": 0.0004, "step": 8271 }, { "epoch": 0.99, "learning_rate": 2.4081343731253748e-06, "loss": 0.0003, "step": 8272 }, { "epoch": 0.99, "learning_rate": 2.4080623875224956e-06, "loss": 0.0006, "step": 8273 }, { "epoch": 0.99, "learning_rate": 2.4079904019196163e-06, "loss": 0.0011, "step": 8274 }, { "epoch": 0.99, "learning_rate": 2.4079184163167367e-06, "loss": 0.0003, "step": 8275 }, { "epoch": 0.99, "learning_rate": 2.4078464307138575e-06, "loss": 0.0013, "step": 8276 }, { "epoch": 0.99, "learning_rate": 2.407774445110978e-06, "loss": 0.0012, "step": 8277 }, { "epoch": 0.99, "learning_rate": 2.4077024595080986e-06, "loss": 0.0003, "step": 8278 }, { "epoch": 0.99, "learning_rate": 2.407630473905219e-06, "loss": 0.0004, "step": 8279 }, { "epoch": 0.99, "learning_rate": 2.4075584883023397e-06, "loss": 0.6099, "step": 8280 }, { "epoch": 0.99, "learning_rate": 2.40748650269946e-06, "loss": 0.0004, "step": 8281 }, { "epoch": 0.99, "learning_rate": 2.407414517096581e-06, "loss": 0.0005, "step": 8282 }, { "epoch": 0.99, "learning_rate": 2.4073425314937012e-06, "loss": 0.0008, "step": 8283 }, { "epoch": 0.99, "learning_rate": 2.4072705458908216e-06, "loss": 0.0004, "step": 8284 }, { "epoch": 0.99, "learning_rate": 2.4071985602879424e-06, "loss": 0.0044, "step": 8285 }, { "epoch": 0.99, "learning_rate": 2.4071265746850627e-06, "loss": 0.0013, "step": 8286 }, { "epoch": 0.99, "learning_rate": 2.4070545890821835e-06, "loss": 0.0243, "step": 8287 }, { "epoch": 0.99, "learning_rate": 2.4069826034793043e-06, "loss": 0.001, "step": 8288 }, { "epoch": 0.99, "learning_rate": 2.406910617876425e-06, "loss": 0.0003, "step": 8289 }, { "epoch": 0.99, "learning_rate": 2.4068386322735454e-06, "loss": 0.0005, "step": 8290 }, { "epoch": 0.99, "learning_rate": 2.406766646670666e-06, "loss": 0.7589, "step": 8291 }, { "epoch": 0.99, "learning_rate": 2.4066946610677865e-06, "loss": 0.0006, "step": 8292 }, { "epoch": 0.99, "learning_rate": 2.4066226754649073e-06, "loss": 0.0004, "step": 8293 }, { "epoch": 0.99, "learning_rate": 2.4065506898620277e-06, "loss": 0.0005, "step": 8294 }, { "epoch": 0.99, "learning_rate": 2.406478704259148e-06, "loss": 0.0005, "step": 8295 }, { "epoch": 0.99, "learning_rate": 2.406406718656269e-06, "loss": 0.0004, "step": 8296 }, { "epoch": 0.99, "learning_rate": 2.406334733053389e-06, "loss": 0.0004, "step": 8297 }, { "epoch": 0.99, "learning_rate": 2.40626274745051e-06, "loss": 0.3885, "step": 8298 }, { "epoch": 0.99, "learning_rate": 2.4061907618476303e-06, "loss": 0.0007, "step": 8299 }, { "epoch": 0.99, "learning_rate": 2.406118776244751e-06, "loss": 0.0003, "step": 8300 }, { "epoch": 0.99, "learning_rate": 2.4060467906418714e-06, "loss": 0.0005, "step": 8301 }, { "epoch": 0.99, "learning_rate": 2.4059748050389922e-06, "loss": 0.0005, "step": 8302 }, { "epoch": 0.99, "learning_rate": 2.405902819436113e-06, "loss": 0.0032, "step": 8303 }, { "epoch": 1.0, "learning_rate": 2.4058308338332333e-06, "loss": 0.0003, "step": 8304 }, { "epoch": 1.0, "learning_rate": 2.405758848230354e-06, "loss": 0.0003, "step": 8305 }, { "epoch": 1.0, "learning_rate": 2.405686862627475e-06, "loss": 0.0005, "step": 8306 }, { "epoch": 1.0, "learning_rate": 2.4056148770245953e-06, "loss": 0.0004, "step": 8307 }, { "epoch": 1.0, "learning_rate": 2.4055428914217156e-06, "loss": 0.0005, "step": 8308 }, { "epoch": 1.0, "learning_rate": 2.4054709058188364e-06, "loss": 0.0004, "step": 8309 }, { "epoch": 1.0, "learning_rate": 2.4053989202159567e-06, "loss": 0.0013, "step": 8310 }, { "epoch": 1.0, "learning_rate": 2.4053269346130775e-06, "loss": 0.0005, "step": 8311 }, { "epoch": 1.0, "learning_rate": 2.405254949010198e-06, "loss": 0.0006, "step": 8312 }, { "epoch": 1.0, "learning_rate": 2.4051829634073187e-06, "loss": 0.0058, "step": 8313 }, { "epoch": 1.0, "learning_rate": 2.405110977804439e-06, "loss": 0.0004, "step": 8314 }, { "epoch": 1.0, "learning_rate": 2.40503899220156e-06, "loss": 0.0003, "step": 8315 }, { "epoch": 1.0, "learning_rate": 2.40496700659868e-06, "loss": 0.0006, "step": 8316 }, { "epoch": 1.0, "learning_rate": 2.404895020995801e-06, "loss": 0.001, "step": 8317 }, { "epoch": 1.0, "learning_rate": 2.4048230353929213e-06, "loss": 0.0005, "step": 8318 }, { "epoch": 1.0, "learning_rate": 2.404751049790042e-06, "loss": 0.0007, "step": 8319 }, { "epoch": 1.0, "learning_rate": 2.404679064187163e-06, "loss": 0.0003, "step": 8320 }, { "epoch": 1.0, "learning_rate": 2.404607078584283e-06, "loss": 0.0004, "step": 8321 }, { "epoch": 1.0, "learning_rate": 2.404535092981404e-06, "loss": 0.0003, "step": 8322 }, { "epoch": 1.0, "learning_rate": 2.4044631073785243e-06, "loss": 0.0028, "step": 8323 }, { "epoch": 1.0, "learning_rate": 2.404391121775645e-06, "loss": 0.0005, "step": 8324 }, { "epoch": 1.0, "learning_rate": 2.4043191361727655e-06, "loss": 0.0003, "step": 8325 }, { "epoch": 1.0, "learning_rate": 2.4042471505698862e-06, "loss": 0.0009, "step": 8326 }, { "epoch": 1.0, "learning_rate": 2.4041751649670066e-06, "loss": 0.0003, "step": 8327 }, { "epoch": 1.0, "learning_rate": 2.4041031793641274e-06, "loss": 0.0003, "step": 8328 }, { "epoch": 1.0, "learning_rate": 2.4040311937612477e-06, "loss": 0.0003, "step": 8329 }, { "epoch": 1.0, "learning_rate": 2.403959208158368e-06, "loss": 1.0046, "step": 8330 }, { "epoch": 1.0, "learning_rate": 2.403887222555489e-06, "loss": 0.0007, "step": 8331 }, { "epoch": 1.0, "learning_rate": 2.4038152369526092e-06, "loss": 0.0007, "step": 8332 }, { "epoch": 1.0, "learning_rate": 2.40374325134973e-06, "loss": 0.0003, "step": 8333 }, { "epoch": 1.0, "learning_rate": 2.4036712657468508e-06, "loss": 0.0004, "step": 8334 }, { "epoch": 1.0, "learning_rate": 2.4035992801439716e-06, "loss": 0.0004, "step": 8335 }, { "epoch": 1.0, "learning_rate": 2.403527294541092e-06, "loss": 0.0004, "step": 8336 }, { "epoch": 1.0, "learning_rate": 2.4034553089382127e-06, "loss": 0.0004, "step": 8337 }, { "epoch": 1.0, "learning_rate": 2.403383323335333e-06, "loss": 0.0033, "step": 8338 }, { "epoch": 1.0, "learning_rate": 2.403311337732454e-06, "loss": 0.0004, "step": 8339 }, { "epoch": 1.0, "learning_rate": 2.403239352129574e-06, "loss": 0.5264, "step": 8340 }, { "epoch": 1.0, "learning_rate": 2.403167366526695e-06, "loss": 0.0009, "step": 8341 }, { "epoch": 1.0, "learning_rate": 2.4030953809238153e-06, "loss": 0.0005, "step": 8342 }, { "epoch": 1.0, "learning_rate": 2.4030233953209357e-06, "loss": 0.2106, "step": 8343 }, { "epoch": 1.0, "learning_rate": 2.4029514097180564e-06, "loss": 0.0004, "step": 8344 }, { "epoch": 1.0, "learning_rate": 2.402879424115177e-06, "loss": 0.032, "step": 8345 }, { "epoch": 1.0, "eval_accuracy": 0.9889155182744158, "eval_loss": 0.060990966856479645, "eval_runtime": 158.1606, "eval_samples_per_second": 105.526, "eval_steps_per_second": 1.65, "step": 8345 }, { "epoch": 1.0, "learning_rate": 2.4028074385122976e-06, "loss": 0.0004, "step": 8346 }, { "epoch": 1.0, "learning_rate": 2.402735452909418e-06, "loss": 0.0004, "step": 8347 }, { "epoch": 1.0, "learning_rate": 2.4026634673065387e-06, "loss": 0.0004, "step": 8348 }, { "epoch": 1.0, "learning_rate": 2.4025914817036595e-06, "loss": 0.0004, "step": 8349 }, { "epoch": 1.0, "learning_rate": 2.40251949610078e-06, "loss": 0.0019, "step": 8350 }, { "epoch": 1.0, "learning_rate": 2.4024475104979006e-06, "loss": 0.0028, "step": 8351 }, { "epoch": 1.0, "learning_rate": 2.4023755248950214e-06, "loss": 0.0004, "step": 8352 }, { "epoch": 1.0, "learning_rate": 2.4023035392921418e-06, "loss": 0.0005, "step": 8353 }, { "epoch": 1.0, "learning_rate": 2.402231553689262e-06, "loss": 0.9622, "step": 8354 }, { "epoch": 1.0, "learning_rate": 2.402159568086383e-06, "loss": 0.0005, "step": 8355 }, { "epoch": 1.0, "learning_rate": 2.4020875824835032e-06, "loss": 0.0223, "step": 8356 }, { "epoch": 1.0, "learning_rate": 2.402015596880624e-06, "loss": 0.0004, "step": 8357 }, { "epoch": 1.0, "learning_rate": 2.4019436112777444e-06, "loss": 0.0003, "step": 8358 }, { "epoch": 1.0, "learning_rate": 2.401871625674865e-06, "loss": 0.0005, "step": 8359 }, { "epoch": 1.0, "learning_rate": 2.4017996400719855e-06, "loss": 0.0004, "step": 8360 }, { "epoch": 1.0, "learning_rate": 2.4017276544691063e-06, "loss": 0.0003, "step": 8361 }, { "epoch": 1.0, "learning_rate": 2.4016556688662266e-06, "loss": 0.0003, "step": 8362 }, { "epoch": 1.0, "learning_rate": 2.4015836832633474e-06, "loss": 0.0005, "step": 8363 }, { "epoch": 1.0, "learning_rate": 2.4015116976604678e-06, "loss": 0.0006, "step": 8364 }, { "epoch": 1.0, "learning_rate": 2.4014397120575886e-06, "loss": 0.0012, "step": 8365 }, { "epoch": 1.0, "learning_rate": 2.4013677264547093e-06, "loss": 0.0005, "step": 8366 }, { "epoch": 1.0, "learning_rate": 2.4012957408518297e-06, "loss": 0.0004, "step": 8367 }, { "epoch": 1.0, "learning_rate": 2.4012237552489505e-06, "loss": 0.0006, "step": 8368 }, { "epoch": 1.0, "learning_rate": 2.401151769646071e-06, "loss": 0.0005, "step": 8369 }, { "epoch": 1.0, "learning_rate": 2.4010797840431916e-06, "loss": 0.0008, "step": 8370 }, { "epoch": 1.0, "learning_rate": 2.401007798440312e-06, "loss": 0.0003, "step": 8371 }, { "epoch": 1.0, "learning_rate": 2.4009358128374327e-06, "loss": 0.0005, "step": 8372 }, { "epoch": 1.0, "learning_rate": 2.400863827234553e-06, "loss": 0.0004, "step": 8373 }, { "epoch": 1.0, "learning_rate": 2.400791841631674e-06, "loss": 0.0003, "step": 8374 }, { "epoch": 1.0, "learning_rate": 2.4007198560287942e-06, "loss": 0.4796, "step": 8375 }, { "epoch": 1.0, "learning_rate": 2.4006478704259146e-06, "loss": 0.0007, "step": 8376 }, { "epoch": 1.0, "learning_rate": 2.4005758848230354e-06, "loss": 0.0003, "step": 8377 }, { "epoch": 1.0, "learning_rate": 2.4005038992201557e-06, "loss": 0.0008, "step": 8378 }, { "epoch": 1.0, "learning_rate": 2.4004319136172765e-06, "loss": 0.0007, "step": 8379 }, { "epoch": 1.0, "learning_rate": 2.4003599280143973e-06, "loss": 0.0006, "step": 8380 }, { "epoch": 1.0, "learning_rate": 2.4002879424115176e-06, "loss": 0.0002, "step": 8381 }, { "epoch": 1.0, "learning_rate": 2.4002159568086384e-06, "loss": 0.7902, "step": 8382 }, { "epoch": 1.0, "learning_rate": 2.400143971205759e-06, "loss": 0.0005, "step": 8383 }, { "epoch": 1.0, "learning_rate": 2.4000719856028795e-06, "loss": 0.1666, "step": 8384 }, { "epoch": 1.0, "learning_rate": 2.4000000000000003e-06, "loss": 0.0003, "step": 8385 }, { "epoch": 1.0, "learning_rate": 2.3999280143971207e-06, "loss": 0.0004, "step": 8386 }, { "epoch": 1.01, "learning_rate": 2.3998560287942414e-06, "loss": 0.0004, "step": 8387 }, { "epoch": 1.01, "learning_rate": 2.399784043191362e-06, "loss": 0.0004, "step": 8388 }, { "epoch": 1.01, "learning_rate": 2.399712057588482e-06, "loss": 0.0004, "step": 8389 }, { "epoch": 1.01, "learning_rate": 2.399640071985603e-06, "loss": 0.0004, "step": 8390 }, { "epoch": 1.01, "learning_rate": 2.3995680863827233e-06, "loss": 0.0005, "step": 8391 }, { "epoch": 1.01, "learning_rate": 2.399496100779844e-06, "loss": 0.0005, "step": 8392 }, { "epoch": 1.01, "learning_rate": 2.3994241151769644e-06, "loss": 0.0005, "step": 8393 }, { "epoch": 1.01, "learning_rate": 2.399352129574085e-06, "loss": 0.0011, "step": 8394 }, { "epoch": 1.01, "learning_rate": 2.3992801439712056e-06, "loss": 0.0003, "step": 8395 }, { "epoch": 1.01, "learning_rate": 2.3992081583683263e-06, "loss": 0.0009, "step": 8396 }, { "epoch": 1.01, "learning_rate": 2.399136172765447e-06, "loss": 0.0004, "step": 8397 }, { "epoch": 1.01, "learning_rate": 2.399064187162568e-06, "loss": 0.0004, "step": 8398 }, { "epoch": 1.01, "learning_rate": 2.3989922015596883e-06, "loss": 0.0005, "step": 8399 }, { "epoch": 1.01, "learning_rate": 2.3989202159568086e-06, "loss": 0.0007, "step": 8400 }, { "epoch": 1.01, "learning_rate": 2.3988482303539294e-06, "loss": 0.0101, "step": 8401 }, { "epoch": 1.01, "learning_rate": 2.3987762447510497e-06, "loss": 0.0004, "step": 8402 }, { "epoch": 1.01, "learning_rate": 2.3987042591481705e-06, "loss": 0.0004, "step": 8403 }, { "epoch": 1.01, "learning_rate": 2.398632273545291e-06, "loss": 0.0007, "step": 8404 }, { "epoch": 1.01, "learning_rate": 2.3985602879424117e-06, "loss": 0.0007, "step": 8405 }, { "epoch": 1.01, "learning_rate": 2.398488302339532e-06, "loss": 0.0004, "step": 8406 }, { "epoch": 1.01, "learning_rate": 2.3984163167366528e-06, "loss": 0.0005, "step": 8407 }, { "epoch": 1.01, "learning_rate": 2.398344331133773e-06, "loss": 0.0005, "step": 8408 }, { "epoch": 1.01, "learning_rate": 2.398272345530894e-06, "loss": 0.8973, "step": 8409 }, { "epoch": 1.01, "learning_rate": 2.3982003599280143e-06, "loss": 0.0009, "step": 8410 }, { "epoch": 1.01, "learning_rate": 2.398128374325135e-06, "loss": 0.0008, "step": 8411 }, { "epoch": 1.01, "learning_rate": 2.398056388722256e-06, "loss": 0.0004, "step": 8412 }, { "epoch": 1.01, "learning_rate": 2.397984403119376e-06, "loss": 0.0005, "step": 8413 }, { "epoch": 1.01, "learning_rate": 2.397912417516497e-06, "loss": 0.0005, "step": 8414 }, { "epoch": 1.01, "learning_rate": 2.3978404319136173e-06, "loss": 0.0004, "step": 8415 }, { "epoch": 1.01, "learning_rate": 2.397768446310738e-06, "loss": 0.0006, "step": 8416 }, { "epoch": 1.01, "learning_rate": 2.3976964607078585e-06, "loss": 0.0005, "step": 8417 }, { "epoch": 1.01, "learning_rate": 2.3976244751049792e-06, "loss": 0.0003, "step": 8418 }, { "epoch": 1.01, "learning_rate": 2.3975524895020996e-06, "loss": 0.0004, "step": 8419 }, { "epoch": 1.01, "learning_rate": 2.3974805038992204e-06, "loss": 0.0004, "step": 8420 }, { "epoch": 1.01, "learning_rate": 2.3974085182963407e-06, "loss": 0.7299, "step": 8421 }, { "epoch": 1.01, "learning_rate": 2.397336532693461e-06, "loss": 0.0005, "step": 8422 }, { "epoch": 1.01, "learning_rate": 2.397264547090582e-06, "loss": 0.0004, "step": 8423 }, { "epoch": 1.01, "learning_rate": 2.397192561487702e-06, "loss": 0.0005, "step": 8424 }, { "epoch": 1.01, "learning_rate": 2.397120575884823e-06, "loss": 0.0007, "step": 8425 }, { "epoch": 1.01, "learning_rate": 2.3970485902819438e-06, "loss": 0.5576, "step": 8426 }, { "epoch": 1.01, "learning_rate": 2.396976604679064e-06, "loss": 0.0019, "step": 8427 }, { "epoch": 1.01, "learning_rate": 2.396904619076185e-06, "loss": 0.0004, "step": 8428 }, { "epoch": 1.01, "learning_rate": 2.3968326334733057e-06, "loss": 0.0041, "step": 8429 }, { "epoch": 1.01, "learning_rate": 2.396760647870426e-06, "loss": 0.0005, "step": 8430 }, { "epoch": 1.01, "learning_rate": 2.396688662267547e-06, "loss": 0.9124, "step": 8431 }, { "epoch": 1.01, "learning_rate": 2.396616676664667e-06, "loss": 0.0003, "step": 8432 }, { "epoch": 1.01, "learning_rate": 2.396544691061788e-06, "loss": 0.0003, "step": 8433 }, { "epoch": 1.01, "learning_rate": 2.3964727054589083e-06, "loss": 0.0005, "step": 8434 }, { "epoch": 1.01, "learning_rate": 2.3964007198560287e-06, "loss": 0.0004, "step": 8435 }, { "epoch": 1.01, "learning_rate": 2.3963287342531494e-06, "loss": 0.0004, "step": 8436 }, { "epoch": 1.01, "learning_rate": 2.3962567486502698e-06, "loss": 0.0005, "step": 8437 }, { "epoch": 1.01, "learning_rate": 2.3961847630473906e-06, "loss": 0.0002, "step": 8438 }, { "epoch": 1.01, "learning_rate": 2.396112777444511e-06, "loss": 0.0013, "step": 8439 }, { "epoch": 1.01, "learning_rate": 2.3960407918416317e-06, "loss": 0.0092, "step": 8440 }, { "epoch": 1.01, "learning_rate": 2.395968806238752e-06, "loss": 0.8817, "step": 8441 }, { "epoch": 1.01, "learning_rate": 2.395896820635873e-06, "loss": 0.0006, "step": 8442 }, { "epoch": 1.01, "learning_rate": 2.3958248350329936e-06, "loss": 0.0005, "step": 8443 }, { "epoch": 1.01, "learning_rate": 2.3957528494301144e-06, "loss": 0.0008, "step": 8444 }, { "epoch": 1.01, "learning_rate": 2.3956808638272347e-06, "loss": 0.0005, "step": 8445 }, { "epoch": 1.01, "learning_rate": 2.395608878224355e-06, "loss": 0.2037, "step": 8446 }, { "epoch": 1.01, "learning_rate": 2.395536892621476e-06, "loss": 0.0003, "step": 8447 }, { "epoch": 1.01, "learning_rate": 2.3954649070185962e-06, "loss": 0.0002, "step": 8448 }, { "epoch": 1.01, "learning_rate": 2.395392921415717e-06, "loss": 0.0003, "step": 8449 }, { "epoch": 1.01, "learning_rate": 2.3953209358128374e-06, "loss": 0.0005, "step": 8450 }, { "epoch": 1.01, "learning_rate": 2.395248950209958e-06, "loss": 0.0007, "step": 8451 }, { "epoch": 1.01, "learning_rate": 2.3951769646070785e-06, "loss": 0.0004, "step": 8452 }, { "epoch": 1.01, "learning_rate": 2.3951049790041993e-06, "loss": 0.0005, "step": 8453 }, { "epoch": 1.01, "learning_rate": 2.3950329934013196e-06, "loss": 0.9677, "step": 8454 }, { "epoch": 1.01, "learning_rate": 2.3949610077984404e-06, "loss": 0.0005, "step": 8455 }, { "epoch": 1.01, "learning_rate": 2.3948890221955608e-06, "loss": 0.0003, "step": 8456 }, { "epoch": 1.01, "learning_rate": 2.3948170365926815e-06, "loss": 0.0005, "step": 8457 }, { "epoch": 1.01, "learning_rate": 2.3947450509898023e-06, "loss": 0.0005, "step": 8458 }, { "epoch": 1.01, "learning_rate": 2.3946730653869227e-06, "loss": 0.0008, "step": 8459 }, { "epoch": 1.01, "learning_rate": 2.3946010797840435e-06, "loss": 0.0018, "step": 8460 }, { "epoch": 1.01, "learning_rate": 2.394529094181164e-06, "loss": 0.0006, "step": 8461 }, { "epoch": 1.01, "learning_rate": 2.3944571085782846e-06, "loss": 0.0006, "step": 8462 }, { "epoch": 1.01, "learning_rate": 2.394385122975405e-06, "loss": 0.0007, "step": 8463 }, { "epoch": 1.01, "learning_rate": 2.3943131373725257e-06, "loss": 0.4844, "step": 8464 }, { "epoch": 1.01, "learning_rate": 2.394241151769646e-06, "loss": 0.0095, "step": 8465 }, { "epoch": 1.01, "learning_rate": 2.394169166166767e-06, "loss": 0.8429, "step": 8466 }, { "epoch": 1.01, "learning_rate": 2.3940971805638872e-06, "loss": 0.0003, "step": 8467 }, { "epoch": 1.01, "learning_rate": 2.394025194961008e-06, "loss": 0.0006, "step": 8468 }, { "epoch": 1.01, "learning_rate": 2.3939532093581283e-06, "loss": 0.0006, "step": 8469 }, { "epoch": 1.01, "learning_rate": 2.3938812237552487e-06, "loss": 0.0008, "step": 8470 }, { "epoch": 1.02, "learning_rate": 2.3938092381523695e-06, "loss": 0.0005, "step": 8471 }, { "epoch": 1.02, "learning_rate": 2.3937372525494903e-06, "loss": 0.7689, "step": 8472 }, { "epoch": 1.02, "learning_rate": 2.3936652669466106e-06, "loss": 0.0006, "step": 8473 }, { "epoch": 1.02, "learning_rate": 2.3935932813437314e-06, "loss": 0.0006, "step": 8474 }, { "epoch": 1.02, "learning_rate": 2.393521295740852e-06, "loss": 0.0006, "step": 8475 }, { "epoch": 1.02, "learning_rate": 2.3934493101379725e-06, "loss": 0.0047, "step": 8476 }, { "epoch": 1.02, "learning_rate": 2.3933773245350933e-06, "loss": 0.0033, "step": 8477 }, { "epoch": 1.02, "learning_rate": 2.3933053389322137e-06, "loss": 0.0007, "step": 8478 }, { "epoch": 1.02, "learning_rate": 2.3932333533293344e-06, "loss": 0.5745, "step": 8479 }, { "epoch": 1.02, "learning_rate": 2.393161367726455e-06, "loss": 0.0005, "step": 8480 }, { "epoch": 1.02, "learning_rate": 2.393089382123575e-06, "loss": 0.0003, "step": 8481 }, { "epoch": 1.02, "learning_rate": 2.393017396520696e-06, "loss": 0.0005, "step": 8482 }, { "epoch": 1.02, "learning_rate": 2.3929454109178163e-06, "loss": 0.0194, "step": 8483 }, { "epoch": 1.02, "learning_rate": 2.392873425314937e-06, "loss": 0.1563, "step": 8484 }, { "epoch": 1.02, "learning_rate": 2.3928014397120574e-06, "loss": 0.0005, "step": 8485 }, { "epoch": 1.02, "learning_rate": 2.392729454109178e-06, "loss": 0.0007, "step": 8486 }, { "epoch": 1.02, "learning_rate": 2.3926574685062986e-06, "loss": 0.0007, "step": 8487 }, { "epoch": 1.02, "learning_rate": 2.3925854829034193e-06, "loss": 0.0007, "step": 8488 }, { "epoch": 1.02, "learning_rate": 2.39251349730054e-06, "loss": 0.0008, "step": 8489 }, { "epoch": 1.02, "learning_rate": 2.392441511697661e-06, "loss": 0.0004, "step": 8490 }, { "epoch": 1.02, "learning_rate": 2.3923695260947812e-06, "loss": 0.0007, "step": 8491 }, { "epoch": 1.02, "learning_rate": 2.3922975404919016e-06, "loss": 0.001, "step": 8492 }, { "epoch": 1.02, "learning_rate": 2.3922255548890224e-06, "loss": 0.1814, "step": 8493 }, { "epoch": 1.02, "learning_rate": 2.3921535692861427e-06, "loss": 0.006, "step": 8494 }, { "epoch": 1.02, "learning_rate": 2.3920815836832635e-06, "loss": 0.0006, "step": 8495 }, { "epoch": 1.02, "learning_rate": 2.392009598080384e-06, "loss": 0.0071, "step": 8496 }, { "epoch": 1.02, "learning_rate": 2.3919376124775046e-06, "loss": 0.0004, "step": 8497 }, { "epoch": 1.02, "learning_rate": 2.391865626874625e-06, "loss": 0.0003, "step": 8498 }, { "epoch": 1.02, "learning_rate": 2.3917936412717458e-06, "loss": 0.34, "step": 8499 }, { "epoch": 1.02, "learning_rate": 2.391721655668866e-06, "loss": 0.0321, "step": 8500 }, { "epoch": 1.02, "learning_rate": 2.391649670065987e-06, "loss": 0.0004, "step": 8501 }, { "epoch": 1.02, "learning_rate": 2.3915776844631073e-06, "loss": 0.0018, "step": 8502 }, { "epoch": 1.02, "learning_rate": 2.391505698860228e-06, "loss": 0.0007, "step": 8503 }, { "epoch": 1.02, "learning_rate": 2.3914337132573484e-06, "loss": 0.0004, "step": 8504 }, { "epoch": 1.02, "learning_rate": 2.391361727654469e-06, "loss": 0.0006, "step": 8505 }, { "epoch": 1.02, "learning_rate": 2.39128974205159e-06, "loss": 0.0004, "step": 8506 }, { "epoch": 1.02, "learning_rate": 2.3912177564487103e-06, "loss": 0.0005, "step": 8507 }, { "epoch": 1.02, "learning_rate": 2.391145770845831e-06, "loss": 0.0004, "step": 8508 }, { "epoch": 1.02, "learning_rate": 2.3910737852429514e-06, "loss": 0.0005, "step": 8509 }, { "epoch": 1.02, "learning_rate": 2.3910017996400722e-06, "loss": 1.1274, "step": 8510 }, { "epoch": 1.02, "learning_rate": 2.3909298140371926e-06, "loss": 0.0006, "step": 8511 }, { "epoch": 1.02, "learning_rate": 2.3908578284343134e-06, "loss": 0.0239, "step": 8512 }, { "epoch": 1.02, "learning_rate": 2.3907858428314337e-06, "loss": 0.0008, "step": 8513 }, { "epoch": 1.02, "learning_rate": 2.3907138572285545e-06, "loss": 0.0075, "step": 8514 }, { "epoch": 1.02, "learning_rate": 2.390641871625675e-06, "loss": 0.0003, "step": 8515 }, { "epoch": 1.02, "learning_rate": 2.390569886022795e-06, "loss": 0.0005, "step": 8516 }, { "epoch": 1.02, "learning_rate": 2.390497900419916e-06, "loss": 0.0004, "step": 8517 }, { "epoch": 1.02, "learning_rate": 2.3904259148170363e-06, "loss": 0.0311, "step": 8518 }, { "epoch": 1.02, "learning_rate": 2.390353929214157e-06, "loss": 0.0007, "step": 8519 }, { "epoch": 1.02, "learning_rate": 2.390281943611278e-06, "loss": 0.7159, "step": 8520 }, { "epoch": 1.02, "learning_rate": 2.3902099580083987e-06, "loss": 0.0007, "step": 8521 }, { "epoch": 1.02, "learning_rate": 2.390137972405519e-06, "loss": 0.0004, "step": 8522 }, { "epoch": 1.02, "learning_rate": 2.39006598680264e-06, "loss": 0.0052, "step": 8523 }, { "epoch": 1.02, "learning_rate": 2.38999400119976e-06, "loss": 0.0004, "step": 8524 }, { "epoch": 1.02, "learning_rate": 2.389922015596881e-06, "loss": 0.0003, "step": 8525 }, { "epoch": 1.02, "learning_rate": 2.3898500299940013e-06, "loss": 0.0005, "step": 8526 }, { "epoch": 1.02, "learning_rate": 2.3897780443911216e-06, "loss": 0.0005, "step": 8527 }, { "epoch": 1.02, "learning_rate": 2.3897060587882424e-06, "loss": 0.0005, "step": 8528 }, { "epoch": 1.02, "learning_rate": 2.3896340731853628e-06, "loss": 0.0004, "step": 8529 }, { "epoch": 1.02, "learning_rate": 2.3895620875824836e-06, "loss": 0.0032, "step": 8530 }, { "epoch": 1.02, "learning_rate": 2.389490101979604e-06, "loss": 0.0005, "step": 8531 }, { "epoch": 1.02, "learning_rate": 2.3894181163767247e-06, "loss": 0.0004, "step": 8532 }, { "epoch": 1.02, "learning_rate": 2.389346130773845e-06, "loss": 0.0006, "step": 8533 }, { "epoch": 1.02, "learning_rate": 2.389274145170966e-06, "loss": 0.0006, "step": 8534 }, { "epoch": 1.02, "learning_rate": 2.3892021595680866e-06, "loss": 0.0007, "step": 8535 }, { "epoch": 1.02, "learning_rate": 2.389130173965207e-06, "loss": 0.0005, "step": 8536 }, { "epoch": 1.02, "learning_rate": 2.3890581883623277e-06, "loss": 0.0007, "step": 8537 }, { "epoch": 1.02, "learning_rate": 2.388986202759448e-06, "loss": 0.0005, "step": 8538 }, { "epoch": 1.02, "learning_rate": 2.388914217156569e-06, "loss": 0.0014, "step": 8539 }, { "epoch": 1.02, "learning_rate": 2.3888422315536892e-06, "loss": 0.0006, "step": 8540 }, { "epoch": 1.02, "learning_rate": 2.38877024595081e-06, "loss": 0.0005, "step": 8541 }, { "epoch": 1.02, "learning_rate": 2.3886982603479304e-06, "loss": 0.0006, "step": 8542 }, { "epoch": 1.02, "learning_rate": 2.388626274745051e-06, "loss": 0.0006, "step": 8543 }, { "epoch": 1.02, "learning_rate": 2.3885542891421715e-06, "loss": 0.0006, "step": 8544 }, { "epoch": 1.02, "learning_rate": 2.3884823035392923e-06, "loss": 0.0006, "step": 8545 }, { "epoch": 1.02, "learning_rate": 2.3884103179364126e-06, "loss": 0.0006, "step": 8546 }, { "epoch": 1.02, "learning_rate": 2.3883383323335334e-06, "loss": 0.0004, "step": 8547 }, { "epoch": 1.02, "learning_rate": 2.3882663467306538e-06, "loss": 0.3848, "step": 8548 }, { "epoch": 1.02, "learning_rate": 2.3881943611277745e-06, "loss": 0.0005, "step": 8549 }, { "epoch": 1.02, "learning_rate": 2.388122375524895e-06, "loss": 0.0005, "step": 8550 }, { "epoch": 1.02, "learning_rate": 2.3880503899220157e-06, "loss": 0.001, "step": 8551 }, { "epoch": 1.02, "learning_rate": 2.3879784043191365e-06, "loss": 0.0005, "step": 8552 }, { "epoch": 1.02, "learning_rate": 2.387906418716257e-06, "loss": 0.0006, "step": 8553 }, { "epoch": 1.03, "learning_rate": 2.3878344331133776e-06, "loss": 0.0005, "step": 8554 }, { "epoch": 1.03, "learning_rate": 2.387762447510498e-06, "loss": 0.0026, "step": 8555 }, { "epoch": 1.03, "learning_rate": 2.3876904619076187e-06, "loss": 0.0006, "step": 8556 }, { "epoch": 1.03, "learning_rate": 2.387618476304739e-06, "loss": 0.0003, "step": 8557 }, { "epoch": 1.03, "learning_rate": 2.38754649070186e-06, "loss": 0.0003, "step": 8558 }, { "epoch": 1.03, "learning_rate": 2.38747450509898e-06, "loss": 0.0004, "step": 8559 }, { "epoch": 1.03, "learning_rate": 2.387402519496101e-06, "loss": 0.0006, "step": 8560 }, { "epoch": 1.03, "learning_rate": 2.3873305338932213e-06, "loss": 0.0006, "step": 8561 }, { "epoch": 1.03, "learning_rate": 2.3872585482903417e-06, "loss": 0.0003, "step": 8562 }, { "epoch": 1.03, "learning_rate": 2.3871865626874625e-06, "loss": 0.0003, "step": 8563 }, { "epoch": 1.03, "learning_rate": 2.387114577084583e-06, "loss": 0.0005, "step": 8564 }, { "epoch": 1.03, "learning_rate": 2.3870425914817036e-06, "loss": 0.0008, "step": 8565 }, { "epoch": 1.03, "learning_rate": 2.3869706058788244e-06, "loss": 0.0008, "step": 8566 }, { "epoch": 1.03, "learning_rate": 2.386898620275945e-06, "loss": 0.0619, "step": 8567 }, { "epoch": 1.03, "learning_rate": 2.3868266346730655e-06, "loss": 1.0244, "step": 8568 }, { "epoch": 1.03, "learning_rate": 2.3867546490701863e-06, "loss": 0.0004, "step": 8569 }, { "epoch": 1.03, "learning_rate": 2.3866826634673067e-06, "loss": 0.0006, "step": 8570 }, { "epoch": 1.03, "learning_rate": 2.3866106778644274e-06, "loss": 0.0003, "step": 8571 }, { "epoch": 1.03, "learning_rate": 2.3865386922615478e-06, "loss": 0.0272, "step": 8572 }, { "epoch": 1.03, "learning_rate": 2.386466706658668e-06, "loss": 0.0087, "step": 8573 }, { "epoch": 1.03, "learning_rate": 2.386394721055789e-06, "loss": 0.0005, "step": 8574 }, { "epoch": 1.03, "learning_rate": 2.3863227354529093e-06, "loss": 0.0008, "step": 8575 }, { "epoch": 1.03, "learning_rate": 2.38625074985003e-06, "loss": 0.0007, "step": 8576 }, { "epoch": 1.03, "learning_rate": 2.3861787642471504e-06, "loss": 0.0007, "step": 8577 }, { "epoch": 1.03, "learning_rate": 2.386106778644271e-06, "loss": 0.0004, "step": 8578 }, { "epoch": 1.03, "learning_rate": 2.3860347930413915e-06, "loss": 0.0013, "step": 8579 }, { "epoch": 1.03, "learning_rate": 2.3859628074385123e-06, "loss": 0.0002, "step": 8580 }, { "epoch": 1.03, "learning_rate": 2.385890821835633e-06, "loss": 0.0009, "step": 8581 }, { "epoch": 1.03, "learning_rate": 2.3858188362327535e-06, "loss": 0.0009, "step": 8582 }, { "epoch": 1.03, "learning_rate": 2.3857468506298742e-06, "loss": 0.0005, "step": 8583 }, { "epoch": 1.03, "learning_rate": 2.3856748650269946e-06, "loss": 0.0003, "step": 8584 }, { "epoch": 1.03, "learning_rate": 2.3856028794241154e-06, "loss": 0.001, "step": 8585 }, { "epoch": 1.03, "learning_rate": 2.3855308938212357e-06, "loss": 0.0006, "step": 8586 }, { "epoch": 1.03, "learning_rate": 2.3854589082183565e-06, "loss": 0.0003, "step": 8587 }, { "epoch": 1.03, "learning_rate": 2.385386922615477e-06, "loss": 0.0007, "step": 8588 }, { "epoch": 1.03, "learning_rate": 2.3853149370125976e-06, "loss": 0.1004, "step": 8589 }, { "epoch": 1.03, "learning_rate": 2.385242951409718e-06, "loss": 0.0004, "step": 8590 }, { "epoch": 1.03, "learning_rate": 2.3851709658068388e-06, "loss": 0.0087, "step": 8591 }, { "epoch": 1.03, "learning_rate": 2.385098980203959e-06, "loss": 0.0003, "step": 8592 }, { "epoch": 1.03, "learning_rate": 2.38502699460108e-06, "loss": 0.8801, "step": 8593 }, { "epoch": 1.03, "learning_rate": 2.3849550089982003e-06, "loss": 0.0003, "step": 8594 }, { "epoch": 1.03, "learning_rate": 2.384883023395321e-06, "loss": 0.0007, "step": 8595 }, { "epoch": 1.03, "learning_rate": 2.3848110377924414e-06, "loss": 0.0004, "step": 8596 }, { "epoch": 1.03, "learning_rate": 2.384739052189562e-06, "loss": 0.0004, "step": 8597 }, { "epoch": 1.03, "learning_rate": 2.384667066586683e-06, "loss": 0.0005, "step": 8598 }, { "epoch": 1.03, "learning_rate": 2.3845950809838033e-06, "loss": 0.0004, "step": 8599 }, { "epoch": 1.03, "learning_rate": 2.384523095380924e-06, "loss": 0.0011, "step": 8600 }, { "epoch": 1.03, "learning_rate": 2.3844511097780444e-06, "loss": 0.0004, "step": 8601 }, { "epoch": 1.03, "learning_rate": 2.3843791241751652e-06, "loss": 0.0006, "step": 8602 }, { "epoch": 1.03, "learning_rate": 2.3843071385722856e-06, "loss": 0.001, "step": 8603 }, { "epoch": 1.03, "learning_rate": 2.3842351529694063e-06, "loss": 0.0004, "step": 8604 }, { "epoch": 1.03, "learning_rate": 2.3841631673665267e-06, "loss": 0.0004, "step": 8605 }, { "epoch": 1.03, "learning_rate": 2.3840911817636475e-06, "loss": 0.0004, "step": 8606 }, { "epoch": 1.03, "learning_rate": 2.384019196160768e-06, "loss": 0.0004, "step": 8607 }, { "epoch": 1.03, "learning_rate": 2.383947210557888e-06, "loss": 0.0006, "step": 8608 }, { "epoch": 1.03, "learning_rate": 2.383875224955009e-06, "loss": 0.8976, "step": 8609 }, { "epoch": 1.03, "learning_rate": 2.3838032393521293e-06, "loss": 0.0004, "step": 8610 }, { "epoch": 1.03, "learning_rate": 2.38373125374925e-06, "loss": 0.0005, "step": 8611 }, { "epoch": 1.03, "learning_rate": 2.383659268146371e-06, "loss": 0.0005, "step": 8612 }, { "epoch": 1.03, "learning_rate": 2.3835872825434917e-06, "loss": 0.0007, "step": 8613 }, { "epoch": 1.03, "learning_rate": 2.383515296940612e-06, "loss": 0.0127, "step": 8614 }, { "epoch": 1.03, "learning_rate": 2.383443311337733e-06, "loss": 0.0003, "step": 8615 }, { "epoch": 1.03, "learning_rate": 2.383371325734853e-06, "loss": 0.0008, "step": 8616 }, { "epoch": 1.03, "learning_rate": 2.383299340131974e-06, "loss": 0.0003, "step": 8617 }, { "epoch": 1.03, "learning_rate": 2.3832273545290943e-06, "loss": 0.0004, "step": 8618 }, { "epoch": 1.03, "learning_rate": 2.3831553689262146e-06, "loss": 0.0003, "step": 8619 }, { "epoch": 1.03, "learning_rate": 2.3830833833233354e-06, "loss": 0.0005, "step": 8620 }, { "epoch": 1.03, "learning_rate": 2.3830113977204558e-06, "loss": 0.0006, "step": 8621 }, { "epoch": 1.03, "learning_rate": 2.3829394121175766e-06, "loss": 0.0006, "step": 8622 }, { "epoch": 1.03, "learning_rate": 2.382867426514697e-06, "loss": 0.0005, "step": 8623 }, { "epoch": 1.03, "learning_rate": 2.3827954409118177e-06, "loss": 0.0006, "step": 8624 }, { "epoch": 1.03, "learning_rate": 2.382723455308938e-06, "loss": 0.0006, "step": 8625 }, { "epoch": 1.03, "learning_rate": 2.382651469706059e-06, "loss": 0.0006, "step": 8626 }, { "epoch": 1.03, "learning_rate": 2.382579484103179e-06, "loss": 0.0006, "step": 8627 }, { "epoch": 1.03, "learning_rate": 2.3825074985003e-06, "loss": 0.0005, "step": 8628 }, { "epoch": 1.03, "learning_rate": 2.3824355128974207e-06, "loss": 0.0005, "step": 8629 }, { "epoch": 1.03, "learning_rate": 2.382363527294541e-06, "loss": 0.0003, "step": 8630 }, { "epoch": 1.03, "learning_rate": 2.382291541691662e-06, "loss": 0.0006, "step": 8631 }, { "epoch": 1.03, "learning_rate": 2.3822195560887822e-06, "loss": 0.93, "step": 8632 }, { "epoch": 1.03, "learning_rate": 2.382147570485903e-06, "loss": 0.0007, "step": 8633 }, { "epoch": 1.03, "learning_rate": 2.3820755848830234e-06, "loss": 0.0008, "step": 8634 }, { "epoch": 1.03, "learning_rate": 2.382003599280144e-06, "loss": 0.0005, "step": 8635 }, { "epoch": 1.03, "learning_rate": 2.3819316136772645e-06, "loss": 0.0005, "step": 8636 }, { "epoch": 1.03, "learning_rate": 2.3818596280743853e-06, "loss": 0.0005, "step": 8637 }, { "epoch": 1.04, "learning_rate": 2.3817876424715056e-06, "loss": 0.0004, "step": 8638 }, { "epoch": 1.04, "learning_rate": 2.3817156568686264e-06, "loss": 0.0005, "step": 8639 }, { "epoch": 1.04, "learning_rate": 2.3816436712657468e-06, "loss": 0.0004, "step": 8640 }, { "epoch": 1.04, "learning_rate": 2.3815716856628675e-06, "loss": 0.0004, "step": 8641 }, { "epoch": 1.04, "learning_rate": 2.381499700059988e-06, "loss": 0.0023, "step": 8642 }, { "epoch": 1.04, "learning_rate": 2.3814277144571087e-06, "loss": 0.0006, "step": 8643 }, { "epoch": 1.04, "learning_rate": 2.3813557288542294e-06, "loss": 0.0009, "step": 8644 }, { "epoch": 1.04, "learning_rate": 2.38128374325135e-06, "loss": 0.0016, "step": 8645 }, { "epoch": 1.04, "learning_rate": 2.3812117576484706e-06, "loss": 0.0017, "step": 8646 }, { "epoch": 1.04, "learning_rate": 2.381139772045591e-06, "loss": 0.0005, "step": 8647 }, { "epoch": 1.04, "learning_rate": 2.3810677864427117e-06, "loss": 0.0008, "step": 8648 }, { "epoch": 1.04, "learning_rate": 2.380995800839832e-06, "loss": 0.0006, "step": 8649 }, { "epoch": 1.04, "learning_rate": 2.380923815236953e-06, "loss": 0.0005, "step": 8650 }, { "epoch": 1.04, "learning_rate": 2.380851829634073e-06, "loss": 0.0006, "step": 8651 }, { "epoch": 1.04, "learning_rate": 2.380779844031194e-06, "loss": 0.035, "step": 8652 }, { "epoch": 1.04, "learning_rate": 2.3807078584283143e-06, "loss": 0.0003, "step": 8653 }, { "epoch": 1.04, "learning_rate": 2.3806358728254347e-06, "loss": 0.0016, "step": 8654 }, { "epoch": 1.04, "learning_rate": 2.3805638872225555e-06, "loss": 0.0009, "step": 8655 }, { "epoch": 1.04, "learning_rate": 2.380491901619676e-06, "loss": 0.0009, "step": 8656 }, { "epoch": 1.04, "learning_rate": 2.3804199160167966e-06, "loss": 0.0119, "step": 8657 }, { "epoch": 1.04, "learning_rate": 2.3803479304139174e-06, "loss": 0.001, "step": 8658 }, { "epoch": 1.04, "learning_rate": 2.3802759448110377e-06, "loss": 0.0005, "step": 8659 }, { "epoch": 1.04, "learning_rate": 2.3802039592081585e-06, "loss": 0.0004, "step": 8660 }, { "epoch": 1.04, "learning_rate": 2.3801319736052793e-06, "loss": 0.0003, "step": 8661 }, { "epoch": 1.04, "learning_rate": 2.3800599880023996e-06, "loss": 0.0004, "step": 8662 }, { "epoch": 1.04, "learning_rate": 2.3799880023995204e-06, "loss": 0.0004, "step": 8663 }, { "epoch": 1.04, "learning_rate": 2.3799160167966408e-06, "loss": 0.0004, "step": 8664 }, { "epoch": 1.04, "learning_rate": 2.379844031193761e-06, "loss": 0.0006, "step": 8665 }, { "epoch": 1.04, "learning_rate": 2.379772045590882e-06, "loss": 0.0006, "step": 8666 }, { "epoch": 1.04, "learning_rate": 2.3797000599880023e-06, "loss": 0.0005, "step": 8667 }, { "epoch": 1.04, "learning_rate": 2.379628074385123e-06, "loss": 0.0004, "step": 8668 }, { "epoch": 1.04, "learning_rate": 2.3795560887822434e-06, "loss": 0.0002, "step": 8669 }, { "epoch": 1.04, "learning_rate": 2.379484103179364e-06, "loss": 0.0005, "step": 8670 }, { "epoch": 1.04, "learning_rate": 2.3794121175764845e-06, "loss": 0.0004, "step": 8671 }, { "epoch": 1.04, "learning_rate": 2.3793401319736053e-06, "loss": 0.0011, "step": 8672 }, { "epoch": 1.04, "learning_rate": 2.3792681463707257e-06, "loss": 0.0003, "step": 8673 }, { "epoch": 1.04, "learning_rate": 2.3791961607678464e-06, "loss": 0.0006, "step": 8674 }, { "epoch": 1.04, "learning_rate": 2.3791241751649672e-06, "loss": 0.0005, "step": 8675 }, { "epoch": 1.04, "learning_rate": 2.3790521895620876e-06, "loss": 0.0005, "step": 8676 }, { "epoch": 1.04, "learning_rate": 2.3789802039592084e-06, "loss": 0.0004, "step": 8677 }, { "epoch": 1.04, "learning_rate": 2.3789082183563287e-06, "loss": 0.0008, "step": 8678 }, { "epoch": 1.04, "learning_rate": 2.3788362327534495e-06, "loss": 0.0004, "step": 8679 }, { "epoch": 1.04, "learning_rate": 2.37876424715057e-06, "loss": 0.0004, "step": 8680 }, { "epoch": 1.04, "learning_rate": 2.3786922615476906e-06, "loss": 0.0005, "step": 8681 }, { "epoch": 1.04, "learning_rate": 2.378620275944811e-06, "loss": 0.0006, "step": 8682 }, { "epoch": 1.04, "learning_rate": 2.3785482903419318e-06, "loss": 0.0005, "step": 8683 }, { "epoch": 1.04, "learning_rate": 2.378476304739052e-06, "loss": 0.0004, "step": 8684 }, { "epoch": 1.04, "learning_rate": 2.378404319136173e-06, "loss": 0.0009, "step": 8685 }, { "epoch": 1.04, "learning_rate": 2.3783323335332932e-06, "loss": 0.0005, "step": 8686 }, { "epoch": 1.04, "learning_rate": 2.378260347930414e-06, "loss": 0.0596, "step": 8687 }, { "epoch": 1.04, "learning_rate": 2.3781883623275344e-06, "loss": 0.0004, "step": 8688 }, { "epoch": 1.04, "learning_rate": 2.378116376724655e-06, "loss": 0.0005, "step": 8689 }, { "epoch": 1.04, "learning_rate": 2.378044391121776e-06, "loss": 0.0002, "step": 8690 }, { "epoch": 1.04, "learning_rate": 2.3779724055188963e-06, "loss": 0.0003, "step": 8691 }, { "epoch": 1.04, "learning_rate": 2.377900419916017e-06, "loss": 0.0004, "step": 8692 }, { "epoch": 1.04, "learning_rate": 2.3778284343131374e-06, "loss": 0.0006, "step": 8693 }, { "epoch": 1.04, "learning_rate": 2.377756448710258e-06, "loss": 0.0004, "step": 8694 }, { "epoch": 1.04, "learning_rate": 2.3776844631073786e-06, "loss": 0.0005, "step": 8695 }, { "epoch": 1.04, "learning_rate": 2.3776124775044993e-06, "loss": 0.0004, "step": 8696 }, { "epoch": 1.04, "learning_rate": 2.3775404919016197e-06, "loss": 0.0005, "step": 8697 }, { "epoch": 1.04, "learning_rate": 2.3774685062987405e-06, "loss": 0.0004, "step": 8698 }, { "epoch": 1.04, "learning_rate": 2.377396520695861e-06, "loss": 0.0005, "step": 8699 }, { "epoch": 1.04, "learning_rate": 2.377324535092981e-06, "loss": 0.0002, "step": 8700 }, { "epoch": 1.04, "learning_rate": 2.377252549490102e-06, "loss": 0.0004, "step": 8701 }, { "epoch": 1.04, "learning_rate": 2.3771805638872223e-06, "loss": 0.0006, "step": 8702 }, { "epoch": 1.04, "learning_rate": 2.377108578284343e-06, "loss": 0.0005, "step": 8703 }, { "epoch": 1.04, "learning_rate": 2.377036592681464e-06, "loss": 0.0005, "step": 8704 }, { "epoch": 1.04, "learning_rate": 2.3769646070785842e-06, "loss": 0.0004, "step": 8705 }, { "epoch": 1.04, "learning_rate": 2.376892621475705e-06, "loss": 0.0004, "step": 8706 }, { "epoch": 1.04, "learning_rate": 2.3768206358728258e-06, "loss": 0.0004, "step": 8707 }, { "epoch": 1.04, "learning_rate": 2.376748650269946e-06, "loss": 0.0003, "step": 8708 }, { "epoch": 1.04, "learning_rate": 2.376676664667067e-06, "loss": 0.0003, "step": 8709 }, { "epoch": 1.04, "learning_rate": 2.3766046790641873e-06, "loss": 0.001, "step": 8710 }, { "epoch": 1.04, "learning_rate": 2.3765326934613076e-06, "loss": 0.0042, "step": 8711 }, { "epoch": 1.04, "learning_rate": 2.3764607078584284e-06, "loss": 0.0002, "step": 8712 }, { "epoch": 1.04, "learning_rate": 2.3763887222555488e-06, "loss": 0.0005, "step": 8713 }, { "epoch": 1.04, "learning_rate": 2.3763167366526695e-06, "loss": 0.0017, "step": 8714 }, { "epoch": 1.04, "learning_rate": 2.37624475104979e-06, "loss": 0.0011, "step": 8715 }, { "epoch": 1.04, "learning_rate": 2.3761727654469107e-06, "loss": 0.0005, "step": 8716 }, { "epoch": 1.04, "learning_rate": 2.376100779844031e-06, "loss": 0.0007, "step": 8717 }, { "epoch": 1.04, "learning_rate": 2.376028794241152e-06, "loss": 0.0005, "step": 8718 }, { "epoch": 1.04, "learning_rate": 2.375956808638272e-06, "loss": 0.7863, "step": 8719 }, { "epoch": 1.04, "learning_rate": 2.375884823035393e-06, "loss": 0.0003, "step": 8720 }, { "epoch": 1.05, "learning_rate": 2.3758128374325137e-06, "loss": 0.0005, "step": 8721 }, { "epoch": 1.05, "learning_rate": 2.3757408518296345e-06, "loss": 0.0004, "step": 8722 }, { "epoch": 1.05, "learning_rate": 2.375668866226755e-06, "loss": 0.0003, "step": 8723 }, { "epoch": 1.05, "learning_rate": 2.375596880623875e-06, "loss": 0.0003, "step": 8724 }, { "epoch": 1.05, "learning_rate": 2.375524895020996e-06, "loss": 0.001, "step": 8725 }, { "epoch": 1.05, "learning_rate": 2.3754529094181163e-06, "loss": 0.0004, "step": 8726 }, { "epoch": 1.05, "learning_rate": 2.375380923815237e-06, "loss": 0.0006, "step": 8727 }, { "epoch": 1.05, "learning_rate": 2.3753089382123575e-06, "loss": 0.0003, "step": 8728 }, { "epoch": 1.05, "learning_rate": 2.3752369526094783e-06, "loss": 0.0004, "step": 8729 }, { "epoch": 1.05, "learning_rate": 2.3751649670065986e-06, "loss": 0.0003, "step": 8730 }, { "epoch": 1.05, "learning_rate": 2.3750929814037194e-06, "loss": 0.0003, "step": 8731 }, { "epoch": 1.05, "learning_rate": 2.3750209958008397e-06, "loss": 0.0107, "step": 8732 }, { "epoch": 1.05, "learning_rate": 2.3749490101979605e-06, "loss": 0.0004, "step": 8733 }, { "epoch": 1.05, "learning_rate": 2.374877024595081e-06, "loss": 0.024, "step": 8734 }, { "epoch": 1.05, "learning_rate": 2.3748050389922017e-06, "loss": 0.0002, "step": 8735 }, { "epoch": 1.05, "learning_rate": 2.3747330533893224e-06, "loss": 0.9812, "step": 8736 }, { "epoch": 1.05, "learning_rate": 2.374661067786443e-06, "loss": 0.0004, "step": 8737 }, { "epoch": 1.05, "learning_rate": 2.3745890821835636e-06, "loss": 0.0005, "step": 8738 }, { "epoch": 1.05, "learning_rate": 2.374517096580684e-06, "loss": 0.0005, "step": 8739 }, { "epoch": 1.05, "learning_rate": 2.3744451109778047e-06, "loss": 0.0004, "step": 8740 }, { "epoch": 1.05, "learning_rate": 2.374373125374925e-06, "loss": 0.0003, "step": 8741 }, { "epoch": 1.05, "learning_rate": 2.374301139772046e-06, "loss": 0.0007, "step": 8742 }, { "epoch": 1.05, "learning_rate": 2.374229154169166e-06, "loss": 0.0003, "step": 8743 }, { "epoch": 1.05, "learning_rate": 2.374157168566287e-06, "loss": 0.0002, "step": 8744 }, { "epoch": 1.05, "learning_rate": 2.3740851829634073e-06, "loss": 0.0004, "step": 8745 }, { "epoch": 1.05, "learning_rate": 2.3740131973605277e-06, "loss": 0.0004, "step": 8746 }, { "epoch": 1.05, "learning_rate": 2.3739412117576485e-06, "loss": 0.001, "step": 8747 }, { "epoch": 1.05, "learning_rate": 2.373869226154769e-06, "loss": 0.0004, "step": 8748 }, { "epoch": 1.05, "learning_rate": 2.3737972405518896e-06, "loss": 0.0005, "step": 8749 }, { "epoch": 1.05, "learning_rate": 2.3737252549490104e-06, "loss": 0.0004, "step": 8750 }, { "epoch": 1.05, "learning_rate": 2.3736532693461307e-06, "loss": 0.0006, "step": 8751 }, { "epoch": 1.05, "learning_rate": 2.3735812837432515e-06, "loss": 0.0005, "step": 8752 }, { "epoch": 1.05, "learning_rate": 2.3735092981403723e-06, "loss": 0.0011, "step": 8753 }, { "epoch": 1.05, "learning_rate": 2.3734373125374926e-06, "loss": 0.0004, "step": 8754 }, { "epoch": 1.05, "learning_rate": 2.3733653269346134e-06, "loss": 0.0003, "step": 8755 }, { "epoch": 1.05, "learning_rate": 2.3732933413317338e-06, "loss": 0.0004, "step": 8756 }, { "epoch": 1.05, "learning_rate": 2.373221355728854e-06, "loss": 0.0005, "step": 8757 }, { "epoch": 1.05, "learning_rate": 2.373149370125975e-06, "loss": 0.0004, "step": 8758 }, { "epoch": 1.05, "learning_rate": 2.3730773845230953e-06, "loss": 0.0003, "step": 8759 }, { "epoch": 1.05, "learning_rate": 2.373005398920216e-06, "loss": 0.0004, "step": 8760 }, { "epoch": 1.05, "learning_rate": 2.3729334133173364e-06, "loss": 0.0006, "step": 8761 }, { "epoch": 1.05, "learning_rate": 2.372861427714457e-06, "loss": 0.0003, "step": 8762 }, { "epoch": 1.05, "learning_rate": 2.3727894421115775e-06, "loss": 0.0009, "step": 8763 }, { "epoch": 1.05, "learning_rate": 2.3727174565086983e-06, "loss": 0.0005, "step": 8764 }, { "epoch": 1.05, "learning_rate": 2.3726454709058187e-06, "loss": 0.0005, "step": 8765 }, { "epoch": 1.05, "learning_rate": 2.3725734853029394e-06, "loss": 0.0004, "step": 8766 }, { "epoch": 1.05, "learning_rate": 2.3725014997000602e-06, "loss": 0.0005, "step": 8767 }, { "epoch": 1.05, "learning_rate": 2.372429514097181e-06, "loss": 0.0005, "step": 8768 }, { "epoch": 1.05, "learning_rate": 2.3723575284943014e-06, "loss": 0.0005, "step": 8769 }, { "epoch": 1.05, "learning_rate": 2.3722855428914217e-06, "loss": 0.0004, "step": 8770 }, { "epoch": 1.05, "learning_rate": 2.3722135572885425e-06, "loss": 0.0004, "step": 8771 }, { "epoch": 1.05, "learning_rate": 2.372141571685663e-06, "loss": 0.0003, "step": 8772 }, { "epoch": 1.05, "learning_rate": 2.3720695860827836e-06, "loss": 0.3503, "step": 8773 }, { "epoch": 1.05, "learning_rate": 2.371997600479904e-06, "loss": 0.001, "step": 8774 }, { "epoch": 1.05, "learning_rate": 2.3719256148770248e-06, "loss": 0.923, "step": 8775 }, { "epoch": 1.05, "learning_rate": 2.371853629274145e-06, "loss": 0.0005, "step": 8776 }, { "epoch": 1.05, "learning_rate": 2.371781643671266e-06, "loss": 0.0004, "step": 8777 }, { "epoch": 1.05, "learning_rate": 2.3717096580683862e-06, "loss": 0.0004, "step": 8778 }, { "epoch": 1.05, "learning_rate": 2.371637672465507e-06, "loss": 0.0003, "step": 8779 }, { "epoch": 1.05, "learning_rate": 2.3715656868626274e-06, "loss": 0.0004, "step": 8780 }, { "epoch": 1.05, "learning_rate": 2.371493701259748e-06, "loss": 0.0004, "step": 8781 }, { "epoch": 1.05, "learning_rate": 2.3714217156568685e-06, "loss": 0.0003, "step": 8782 }, { "epoch": 1.05, "learning_rate": 2.3713497300539893e-06, "loss": 0.0006, "step": 8783 }, { "epoch": 1.05, "learning_rate": 2.37127774445111e-06, "loss": 0.0007, "step": 8784 }, { "epoch": 1.05, "learning_rate": 2.3712057588482304e-06, "loss": 0.0041, "step": 8785 }, { "epoch": 1.05, "learning_rate": 2.371133773245351e-06, "loss": 0.0006, "step": 8786 }, { "epoch": 1.05, "learning_rate": 2.3710617876424716e-06, "loss": 0.0004, "step": 8787 }, { "epoch": 1.05, "learning_rate": 2.3709898020395923e-06, "loss": 0.0036, "step": 8788 }, { "epoch": 1.05, "learning_rate": 2.3709178164367127e-06, "loss": 0.0003, "step": 8789 }, { "epoch": 1.05, "learning_rate": 2.3708458308338335e-06, "loss": 0.0003, "step": 8790 }, { "epoch": 1.05, "learning_rate": 2.370773845230954e-06, "loss": 0.0005, "step": 8791 }, { "epoch": 1.05, "learning_rate": 2.370701859628074e-06, "loss": 0.0009, "step": 8792 }, { "epoch": 1.05, "learning_rate": 2.370629874025195e-06, "loss": 0.0004, "step": 8793 }, { "epoch": 1.05, "learning_rate": 2.3705578884223153e-06, "loss": 0.0003, "step": 8794 }, { "epoch": 1.05, "learning_rate": 2.370485902819436e-06, "loss": 0.0004, "step": 8795 }, { "epoch": 1.05, "learning_rate": 2.3704139172165564e-06, "loss": 0.0002, "step": 8796 }, { "epoch": 1.05, "learning_rate": 2.3703419316136772e-06, "loss": 0.0009, "step": 8797 }, { "epoch": 1.05, "learning_rate": 2.370269946010798e-06, "loss": 0.001, "step": 8798 }, { "epoch": 1.05, "learning_rate": 2.3701979604079188e-06, "loss": 0.0003, "step": 8799 }, { "epoch": 1.05, "learning_rate": 2.370125974805039e-06, "loss": 0.0003, "step": 8800 }, { "epoch": 1.05, "learning_rate": 2.37005398920216e-06, "loss": 0.1117, "step": 8801 }, { "epoch": 1.05, "learning_rate": 2.3699820035992803e-06, "loss": 0.0003, "step": 8802 }, { "epoch": 1.05, "learning_rate": 2.369910017996401e-06, "loss": 0.0004, "step": 8803 }, { "epoch": 1.06, "learning_rate": 2.3698380323935214e-06, "loss": 0.0005, "step": 8804 }, { "epoch": 1.06, "learning_rate": 2.3697660467906418e-06, "loss": 0.0002, "step": 8805 }, { "epoch": 1.06, "learning_rate": 2.3696940611877625e-06, "loss": 0.0005, "step": 8806 }, { "epoch": 1.06, "learning_rate": 2.369622075584883e-06, "loss": 0.0003, "step": 8807 }, { "epoch": 1.06, "learning_rate": 2.3695500899820037e-06, "loss": 0.0011, "step": 8808 }, { "epoch": 1.06, "learning_rate": 2.369478104379124e-06, "loss": 0.1112, "step": 8809 }, { "epoch": 1.06, "learning_rate": 2.369406118776245e-06, "loss": 0.0006, "step": 8810 }, { "epoch": 1.06, "learning_rate": 2.369334133173365e-06, "loss": 0.0007, "step": 8811 }, { "epoch": 1.06, "learning_rate": 2.369262147570486e-06, "loss": 0.0004, "step": 8812 }, { "epoch": 1.06, "learning_rate": 2.3691901619676067e-06, "loss": 0.0008, "step": 8813 }, { "epoch": 1.06, "learning_rate": 2.369118176364727e-06, "loss": 0.0005, "step": 8814 }, { "epoch": 1.06, "learning_rate": 2.369046190761848e-06, "loss": 0.0003, "step": 8815 }, { "epoch": 1.06, "learning_rate": 2.368974205158968e-06, "loss": 0.0008, "step": 8816 }, { "epoch": 1.06, "learning_rate": 2.368902219556089e-06, "loss": 0.0004, "step": 8817 }, { "epoch": 1.06, "learning_rate": 2.3688302339532093e-06, "loss": 0.0004, "step": 8818 }, { "epoch": 1.06, "learning_rate": 2.36875824835033e-06, "loss": 0.0003, "step": 8819 }, { "epoch": 1.06, "learning_rate": 2.3686862627474505e-06, "loss": 0.0004, "step": 8820 }, { "epoch": 1.06, "learning_rate": 2.3686142771445712e-06, "loss": 0.0002, "step": 8821 }, { "epoch": 1.06, "learning_rate": 2.3685422915416916e-06, "loss": 0.0005, "step": 8822 }, { "epoch": 1.06, "learning_rate": 2.3684703059388124e-06, "loss": 0.0004, "step": 8823 }, { "epoch": 1.06, "learning_rate": 2.3683983203359327e-06, "loss": 0.0002, "step": 8824 }, { "epoch": 1.06, "learning_rate": 2.3683263347330535e-06, "loss": 0.0003, "step": 8825 }, { "epoch": 1.06, "learning_rate": 2.368254349130174e-06, "loss": 0.0005, "step": 8826 }, { "epoch": 1.06, "learning_rate": 2.3681823635272946e-06, "loss": 0.0004, "step": 8827 }, { "epoch": 1.06, "learning_rate": 2.368110377924415e-06, "loss": 0.0004, "step": 8828 }, { "epoch": 1.06, "learning_rate": 2.3680383923215358e-06, "loss": 0.0004, "step": 8829 }, { "epoch": 1.06, "learning_rate": 2.3679664067186566e-06, "loss": 0.0003, "step": 8830 }, { "epoch": 1.06, "learning_rate": 2.367894421115777e-06, "loss": 0.0022, "step": 8831 }, { "epoch": 1.06, "learning_rate": 2.3678224355128977e-06, "loss": 0.0004, "step": 8832 }, { "epoch": 1.06, "learning_rate": 2.367750449910018e-06, "loss": 0.0003, "step": 8833 }, { "epoch": 1.06, "learning_rate": 2.367678464307139e-06, "loss": 0.0005, "step": 8834 }, { "epoch": 1.06, "learning_rate": 2.367606478704259e-06, "loss": 0.0004, "step": 8835 }, { "epoch": 1.06, "learning_rate": 2.36753449310138e-06, "loss": 0.0004, "step": 8836 }, { "epoch": 1.06, "learning_rate": 2.3674625074985003e-06, "loss": 0.0003, "step": 8837 }, { "epoch": 1.06, "learning_rate": 2.3673905218956207e-06, "loss": 0.0005, "step": 8838 }, { "epoch": 1.06, "learning_rate": 2.3673185362927415e-06, "loss": 0.0004, "step": 8839 }, { "epoch": 1.06, "learning_rate": 2.367246550689862e-06, "loss": 0.0002, "step": 8840 }, { "epoch": 1.06, "learning_rate": 2.3671745650869826e-06, "loss": 0.0006, "step": 8841 }, { "epoch": 1.06, "learning_rate": 2.367102579484103e-06, "loss": 0.0004, "step": 8842 }, { "epoch": 1.06, "learning_rate": 2.3670305938812237e-06, "loss": 0.0003, "step": 8843 }, { "epoch": 1.06, "learning_rate": 2.3669586082783445e-06, "loss": 0.0004, "step": 8844 }, { "epoch": 1.06, "learning_rate": 2.3668866226754653e-06, "loss": 0.0004, "step": 8845 }, { "epoch": 1.06, "learning_rate": 2.3668146370725856e-06, "loss": 0.0003, "step": 8846 }, { "epoch": 1.06, "learning_rate": 2.3667426514697064e-06, "loss": 0.0005, "step": 8847 }, { "epoch": 1.06, "learning_rate": 2.3666706658668268e-06, "loss": 0.0006, "step": 8848 }, { "epoch": 1.06, "learning_rate": 2.3665986802639475e-06, "loss": 0.0004, "step": 8849 }, { "epoch": 1.06, "learning_rate": 2.366526694661068e-06, "loss": 0.0004, "step": 8850 }, { "epoch": 1.06, "learning_rate": 2.3664547090581883e-06, "loss": 0.0002, "step": 8851 }, { "epoch": 1.06, "learning_rate": 2.366382723455309e-06, "loss": 0.0004, "step": 8852 }, { "epoch": 1.06, "learning_rate": 2.3663107378524294e-06, "loss": 0.0003, "step": 8853 }, { "epoch": 1.06, "learning_rate": 2.36623875224955e-06, "loss": 0.0003, "step": 8854 }, { "epoch": 1.06, "learning_rate": 2.3661667666466705e-06, "loss": 0.0003, "step": 8855 }, { "epoch": 1.06, "learning_rate": 2.3660947810437913e-06, "loss": 0.0005, "step": 8856 }, { "epoch": 1.06, "learning_rate": 2.3660227954409117e-06, "loss": 0.0004, "step": 8857 }, { "epoch": 1.06, "learning_rate": 2.3659508098380324e-06, "loss": 0.0005, "step": 8858 }, { "epoch": 1.06, "learning_rate": 2.365878824235153e-06, "loss": 0.0005, "step": 8859 }, { "epoch": 1.06, "learning_rate": 2.3658068386322736e-06, "loss": 0.0024, "step": 8860 }, { "epoch": 1.06, "learning_rate": 2.3657348530293943e-06, "loss": 0.0004, "step": 8861 }, { "epoch": 1.06, "learning_rate": 2.3656628674265147e-06, "loss": 0.0003, "step": 8862 }, { "epoch": 1.06, "learning_rate": 2.3655908818236355e-06, "loss": 0.0004, "step": 8863 }, { "epoch": 1.06, "learning_rate": 2.365518896220756e-06, "loss": 0.0004, "step": 8864 }, { "epoch": 1.06, "learning_rate": 2.3654469106178766e-06, "loss": 0.0003, "step": 8865 }, { "epoch": 1.06, "learning_rate": 2.365374925014997e-06, "loss": 0.0003, "step": 8866 }, { "epoch": 1.06, "learning_rate": 2.3653029394121177e-06, "loss": 0.0004, "step": 8867 }, { "epoch": 1.06, "learning_rate": 2.365230953809238e-06, "loss": 0.0004, "step": 8868 }, { "epoch": 1.06, "learning_rate": 2.365158968206359e-06, "loss": 0.0003, "step": 8869 }, { "epoch": 1.06, "learning_rate": 2.3650869826034792e-06, "loss": 0.0005, "step": 8870 }, { "epoch": 1.06, "learning_rate": 2.3650149970006e-06, "loss": 0.0011, "step": 8871 }, { "epoch": 1.06, "learning_rate": 2.3649430113977204e-06, "loss": 0.0002, "step": 8872 }, { "epoch": 1.06, "learning_rate": 2.364871025794841e-06, "loss": 0.0006, "step": 8873 }, { "epoch": 1.06, "learning_rate": 2.3647990401919615e-06, "loss": 0.0005, "step": 8874 }, { "epoch": 1.06, "learning_rate": 2.3647270545890823e-06, "loss": 0.0004, "step": 8875 }, { "epoch": 1.06, "learning_rate": 2.364655068986203e-06, "loss": 0.0003, "step": 8876 }, { "epoch": 1.06, "learning_rate": 2.3645830833833234e-06, "loss": 0.0003, "step": 8877 }, { "epoch": 1.06, "learning_rate": 2.364511097780444e-06, "loss": 0.806, "step": 8878 }, { "epoch": 1.06, "learning_rate": 2.3644391121775645e-06, "loss": 0.0004, "step": 8879 }, { "epoch": 1.06, "learning_rate": 2.3643671265746853e-06, "loss": 0.0187, "step": 8880 }, { "epoch": 1.06, "learning_rate": 2.3642951409718057e-06, "loss": 0.0016, "step": 8881 }, { "epoch": 1.06, "learning_rate": 2.3642231553689265e-06, "loss": 0.0006, "step": 8882 }, { "epoch": 1.06, "learning_rate": 2.364151169766047e-06, "loss": 0.0005, "step": 8883 }, { "epoch": 1.06, "learning_rate": 2.364079184163167e-06, "loss": 0.0006, "step": 8884 }, { "epoch": 1.06, "learning_rate": 2.364007198560288e-06, "loss": 0.0006, "step": 8885 }, { "epoch": 1.06, "learning_rate": 2.3639352129574083e-06, "loss": 0.0008, "step": 8886 }, { "epoch": 1.06, "learning_rate": 2.363863227354529e-06, "loss": 0.0727, "step": 8887 }, { "epoch": 1.07, "learning_rate": 2.3637912417516494e-06, "loss": 0.0002, "step": 8888 }, { "epoch": 1.07, "learning_rate": 2.3637192561487702e-06, "loss": 0.0005, "step": 8889 }, { "epoch": 1.07, "learning_rate": 2.363647270545891e-06, "loss": 0.0002, "step": 8890 }, { "epoch": 1.07, "learning_rate": 2.3635752849430118e-06, "loss": 0.0003, "step": 8891 }, { "epoch": 1.07, "learning_rate": 2.363503299340132e-06, "loss": 0.0004, "step": 8892 }, { "epoch": 1.07, "learning_rate": 2.363431313737253e-06, "loss": 0.0003, "step": 8893 }, { "epoch": 1.07, "learning_rate": 2.3633593281343733e-06, "loss": 0.0003, "step": 8894 }, { "epoch": 1.07, "learning_rate": 2.363287342531494e-06, "loss": 0.0003, "step": 8895 }, { "epoch": 1.07, "learning_rate": 2.3632153569286144e-06, "loss": 0.0004, "step": 8896 }, { "epoch": 1.07, "learning_rate": 2.3631433713257347e-06, "loss": 0.0004, "step": 8897 }, { "epoch": 1.07, "learning_rate": 2.3630713857228555e-06, "loss": 0.0004, "step": 8898 }, { "epoch": 1.07, "learning_rate": 2.362999400119976e-06, "loss": 0.0003, "step": 8899 }, { "epoch": 1.07, "learning_rate": 2.3629274145170967e-06, "loss": 0.0004, "step": 8900 }, { "epoch": 1.07, "learning_rate": 2.362855428914217e-06, "loss": 0.0004, "step": 8901 }, { "epoch": 1.07, "learning_rate": 2.362783443311338e-06, "loss": 0.0033, "step": 8902 }, { "epoch": 1.07, "learning_rate": 2.362711457708458e-06, "loss": 0.0004, "step": 8903 }, { "epoch": 1.07, "learning_rate": 2.362639472105579e-06, "loss": 0.0003, "step": 8904 }, { "epoch": 1.07, "learning_rate": 2.3625674865026993e-06, "loss": 0.0002, "step": 8905 }, { "epoch": 1.07, "learning_rate": 2.36249550089982e-06, "loss": 0.0004, "step": 8906 }, { "epoch": 1.07, "learning_rate": 2.362423515296941e-06, "loss": 0.0011, "step": 8907 }, { "epoch": 1.07, "learning_rate": 2.362351529694061e-06, "loss": 0.0004, "step": 8908 }, { "epoch": 1.07, "learning_rate": 2.362279544091182e-06, "loss": 0.0003, "step": 8909 }, { "epoch": 1.07, "learning_rate": 2.3622075584883023e-06, "loss": 0.0002, "step": 8910 }, { "epoch": 1.07, "learning_rate": 2.362135572885423e-06, "loss": 0.0004, "step": 8911 }, { "epoch": 1.07, "learning_rate": 2.3620635872825435e-06, "loss": 0.0007, "step": 8912 }, { "epoch": 1.07, "learning_rate": 2.3619916016796642e-06, "loss": 0.0003, "step": 8913 }, { "epoch": 1.07, "learning_rate": 2.3619196160767846e-06, "loss": 0.0009, "step": 8914 }, { "epoch": 1.07, "learning_rate": 2.3618476304739054e-06, "loss": 0.0005, "step": 8915 }, { "epoch": 1.07, "learning_rate": 2.3617756448710257e-06, "loss": 0.0002, "step": 8916 }, { "epoch": 1.07, "learning_rate": 2.3617036592681465e-06, "loss": 0.0033, "step": 8917 }, { "epoch": 1.07, "learning_rate": 2.361631673665267e-06, "loss": 0.0004, "step": 8918 }, { "epoch": 1.07, "learning_rate": 2.3615596880623872e-06, "loss": 0.0008, "step": 8919 }, { "epoch": 1.07, "learning_rate": 2.361487702459508e-06, "loss": 0.0002, "step": 8920 }, { "epoch": 1.07, "learning_rate": 2.3614157168566288e-06, "loss": 0.0004, "step": 8921 }, { "epoch": 1.07, "learning_rate": 2.3613437312537496e-06, "loss": 0.0003, "step": 8922 }, { "epoch": 1.07, "learning_rate": 2.36127174565087e-06, "loss": 0.0004, "step": 8923 }, { "epoch": 1.07, "learning_rate": 2.3611997600479907e-06, "loss": 0.0003, "step": 8924 }, { "epoch": 1.07, "learning_rate": 2.361127774445111e-06, "loss": 0.0824, "step": 8925 }, { "epoch": 1.07, "learning_rate": 2.361055788842232e-06, "loss": 0.0003, "step": 8926 }, { "epoch": 1.07, "learning_rate": 2.360983803239352e-06, "loss": 0.0002, "step": 8927 }, { "epoch": 1.07, "learning_rate": 2.360911817636473e-06, "loss": 0.0009, "step": 8928 }, { "epoch": 1.07, "learning_rate": 2.3608398320335933e-06, "loss": 0.0003, "step": 8929 }, { "epoch": 1.07, "learning_rate": 2.360767846430714e-06, "loss": 0.0729, "step": 8930 }, { "epoch": 1.07, "learning_rate": 2.3606958608278344e-06, "loss": 0.0002, "step": 8931 }, { "epoch": 1.07, "learning_rate": 2.360623875224955e-06, "loss": 0.0004, "step": 8932 }, { "epoch": 1.07, "learning_rate": 2.3605518896220756e-06, "loss": 0.0004, "step": 8933 }, { "epoch": 1.07, "learning_rate": 2.360479904019196e-06, "loss": 0.0004, "step": 8934 }, { "epoch": 1.07, "learning_rate": 2.3604079184163167e-06, "loss": 0.0003, "step": 8935 }, { "epoch": 1.07, "learning_rate": 2.3603359328134375e-06, "loss": 0.0006, "step": 8936 }, { "epoch": 1.07, "learning_rate": 2.360263947210558e-06, "loss": 0.0005, "step": 8937 }, { "epoch": 1.07, "learning_rate": 2.3601919616076786e-06, "loss": 0.0006, "step": 8938 }, { "epoch": 1.07, "learning_rate": 2.3601199760047994e-06, "loss": 0.0005, "step": 8939 }, { "epoch": 1.07, "learning_rate": 2.3600479904019198e-06, "loss": 0.0007, "step": 8940 }, { "epoch": 1.07, "learning_rate": 2.3599760047990405e-06, "loss": 0.0003, "step": 8941 }, { "epoch": 1.07, "learning_rate": 2.359904019196161e-06, "loss": 0.0002, "step": 8942 }, { "epoch": 1.07, "learning_rate": 2.3598320335932812e-06, "loss": 0.0004, "step": 8943 }, { "epoch": 1.07, "learning_rate": 2.359760047990402e-06, "loss": 0.0003, "step": 8944 }, { "epoch": 1.07, "learning_rate": 2.3596880623875224e-06, "loss": 0.0004, "step": 8945 }, { "epoch": 1.07, "learning_rate": 2.359616076784643e-06, "loss": 0.0004, "step": 8946 }, { "epoch": 1.07, "learning_rate": 2.3595440911817635e-06, "loss": 0.0002, "step": 8947 }, { "epoch": 1.07, "learning_rate": 2.3594721055788843e-06, "loss": 0.0004, "step": 8948 }, { "epoch": 1.07, "learning_rate": 2.3594001199760046e-06, "loss": 0.0002, "step": 8949 }, { "epoch": 1.07, "learning_rate": 2.3593281343731254e-06, "loss": 0.0005, "step": 8950 }, { "epoch": 1.07, "learning_rate": 2.3592561487702458e-06, "loss": 0.0001, "step": 8951 }, { "epoch": 1.07, "learning_rate": 2.3591841631673666e-06, "loss": 0.0003, "step": 8952 }, { "epoch": 1.07, "learning_rate": 2.3591121775644873e-06, "loss": 0.0002, "step": 8953 }, { "epoch": 1.07, "learning_rate": 2.3590401919616077e-06, "loss": 0.0005, "step": 8954 }, { "epoch": 1.07, "learning_rate": 2.3589682063587285e-06, "loss": 0.0005, "step": 8955 }, { "epoch": 1.07, "learning_rate": 2.358896220755849e-06, "loss": 0.0005, "step": 8956 }, { "epoch": 1.07, "learning_rate": 2.3588242351529696e-06, "loss": 0.0004, "step": 8957 }, { "epoch": 1.07, "learning_rate": 2.35875224955009e-06, "loss": 0.0003, "step": 8958 }, { "epoch": 1.07, "learning_rate": 2.3586802639472107e-06, "loss": 0.0003, "step": 8959 }, { "epoch": 1.07, "learning_rate": 2.358608278344331e-06, "loss": 0.0003, "step": 8960 }, { "epoch": 1.07, "learning_rate": 2.358536292741452e-06, "loss": 0.8288, "step": 8961 }, { "epoch": 1.07, "learning_rate": 2.3584643071385722e-06, "loss": 0.0004, "step": 8962 }, { "epoch": 1.07, "learning_rate": 2.358392321535693e-06, "loss": 0.0005, "step": 8963 }, { "epoch": 1.07, "learning_rate": 2.3583203359328134e-06, "loss": 0.0002, "step": 8964 }, { "epoch": 1.07, "learning_rate": 2.3582483503299337e-06, "loss": 0.0004, "step": 8965 }, { "epoch": 1.07, "learning_rate": 2.3581763647270545e-06, "loss": 0.0004, "step": 8966 }, { "epoch": 1.07, "learning_rate": 2.3581043791241753e-06, "loss": 0.0016, "step": 8967 }, { "epoch": 1.07, "learning_rate": 2.358032393521296e-06, "loss": 0.0004, "step": 8968 }, { "epoch": 1.07, "learning_rate": 2.3579604079184164e-06, "loss": 0.0003, "step": 8969 }, { "epoch": 1.07, "learning_rate": 2.357888422315537e-06, "loss": 0.0003, "step": 8970 }, { "epoch": 1.08, "learning_rate": 2.3578164367126575e-06, "loss": 0.0007, "step": 8971 }, { "epoch": 1.08, "learning_rate": 2.3577444511097783e-06, "loss": 0.0003, "step": 8972 }, { "epoch": 1.08, "learning_rate": 2.3576724655068987e-06, "loss": 0.0002, "step": 8973 }, { "epoch": 1.08, "learning_rate": 2.3576004799040194e-06, "loss": 0.0002, "step": 8974 }, { "epoch": 1.08, "learning_rate": 2.35752849430114e-06, "loss": 0.0004, "step": 8975 }, { "epoch": 1.08, "learning_rate": 2.3574565086982606e-06, "loss": 0.0004, "step": 8976 }, { "epoch": 1.08, "learning_rate": 2.357384523095381e-06, "loss": 0.0011, "step": 8977 }, { "epoch": 1.08, "learning_rate": 2.3573125374925013e-06, "loss": 0.0003, "step": 8978 }, { "epoch": 1.08, "learning_rate": 2.357240551889622e-06, "loss": 0.0003, "step": 8979 }, { "epoch": 1.08, "learning_rate": 2.3571685662867424e-06, "loss": 0.0006, "step": 8980 }, { "epoch": 1.08, "learning_rate": 2.357096580683863e-06, "loss": 0.0004, "step": 8981 }, { "epoch": 1.08, "learning_rate": 2.357024595080984e-06, "loss": 0.0003, "step": 8982 }, { "epoch": 1.08, "learning_rate": 2.3569526094781043e-06, "loss": 0.0003, "step": 8983 }, { "epoch": 1.08, "learning_rate": 2.356880623875225e-06, "loss": 0.0002, "step": 8984 }, { "epoch": 1.08, "learning_rate": 2.356808638272346e-06, "loss": 0.1391, "step": 8985 }, { "epoch": 1.08, "learning_rate": 2.3567366526694663e-06, "loss": 0.0003, "step": 8986 }, { "epoch": 1.08, "learning_rate": 2.356664667066587e-06, "loss": 0.0002, "step": 8987 }, { "epoch": 1.08, "learning_rate": 2.3565926814637074e-06, "loss": 0.0003, "step": 8988 }, { "epoch": 1.08, "learning_rate": 2.3565206958608277e-06, "loss": 0.0004, "step": 8989 }, { "epoch": 1.08, "learning_rate": 2.3564487102579485e-06, "loss": 0.0002, "step": 8990 }, { "epoch": 1.08, "learning_rate": 2.356376724655069e-06, "loss": 0.0002, "step": 8991 }, { "epoch": 1.08, "learning_rate": 2.3563047390521897e-06, "loss": 0.0005, "step": 8992 }, { "epoch": 1.08, "learning_rate": 2.35623275344931e-06, "loss": 0.0005, "step": 8993 }, { "epoch": 1.08, "learning_rate": 2.3561607678464308e-06, "loss": 0.0004, "step": 8994 }, { "epoch": 1.08, "learning_rate": 2.356088782243551e-06, "loss": 0.0004, "step": 8995 }, { "epoch": 1.08, "learning_rate": 2.356016796640672e-06, "loss": 0.0003, "step": 8996 }, { "epoch": 1.08, "learning_rate": 2.3559448110377923e-06, "loss": 0.0003, "step": 8997 }, { "epoch": 1.08, "learning_rate": 2.355872825434913e-06, "loss": 0.0005, "step": 8998 }, { "epoch": 1.08, "learning_rate": 2.355800839832034e-06, "loss": 0.0007, "step": 8999 }, { "epoch": 1.08, "learning_rate": 2.355728854229154e-06, "loss": 0.0003, "step": 9000 }, { "epoch": 1.08, "learning_rate": 2.355656868626275e-06, "loss": 0.0005, "step": 9001 }, { "epoch": 1.08, "learning_rate": 2.3555848830233953e-06, "loss": 0.0021, "step": 9002 }, { "epoch": 1.08, "learning_rate": 2.355512897420516e-06, "loss": 0.0004, "step": 9003 }, { "epoch": 1.08, "learning_rate": 2.3554409118176365e-06, "loss": 0.0004, "step": 9004 }, { "epoch": 1.08, "learning_rate": 2.3553689262147572e-06, "loss": 0.0004, "step": 9005 }, { "epoch": 1.08, "learning_rate": 2.3552969406118776e-06, "loss": 0.0003, "step": 9006 }, { "epoch": 1.08, "learning_rate": 2.3552249550089984e-06, "loss": 0.0005, "step": 9007 }, { "epoch": 1.08, "learning_rate": 2.3551529694061187e-06, "loss": 0.0008, "step": 9008 }, { "epoch": 1.08, "learning_rate": 2.3550809838032395e-06, "loss": 0.0002, "step": 9009 }, { "epoch": 1.08, "learning_rate": 2.35500899820036e-06, "loss": 0.0003, "step": 9010 }, { "epoch": 1.08, "learning_rate": 2.35493701259748e-06, "loss": 0.0002, "step": 9011 }, { "epoch": 1.08, "learning_rate": 2.354865026994601e-06, "loss": 0.0002, "step": 9012 }, { "epoch": 1.08, "learning_rate": 2.3547930413917218e-06, "loss": 0.0005, "step": 9013 }, { "epoch": 1.08, "learning_rate": 2.3547210557888425e-06, "loss": 0.0108, "step": 9014 }, { "epoch": 1.08, "learning_rate": 2.354649070185963e-06, "loss": 0.0003, "step": 9015 }, { "epoch": 1.08, "learning_rate": 2.3545770845830837e-06, "loss": 0.0014, "step": 9016 }, { "epoch": 1.08, "learning_rate": 2.354505098980204e-06, "loss": 0.3523, "step": 9017 }, { "epoch": 1.08, "learning_rate": 2.354433113377325e-06, "loss": 0.0002, "step": 9018 }, { "epoch": 1.08, "learning_rate": 2.354361127774445e-06, "loss": 0.0003, "step": 9019 }, { "epoch": 1.08, "learning_rate": 2.354289142171566e-06, "loss": 0.0002, "step": 9020 }, { "epoch": 1.08, "learning_rate": 2.3542171565686863e-06, "loss": 0.0002, "step": 9021 }, { "epoch": 1.08, "learning_rate": 2.354145170965807e-06, "loss": 0.0006, "step": 9022 }, { "epoch": 1.08, "learning_rate": 2.3540731853629274e-06, "loss": 0.0002, "step": 9023 }, { "epoch": 1.08, "learning_rate": 2.354001199760048e-06, "loss": 0.0002, "step": 9024 }, { "epoch": 1.08, "learning_rate": 2.3539292141571686e-06, "loss": 0.0005, "step": 9025 }, { "epoch": 1.08, "learning_rate": 2.353857228554289e-06, "loss": 0.0108, "step": 9026 }, { "epoch": 1.08, "learning_rate": 2.3537852429514097e-06, "loss": 0.0036, "step": 9027 }, { "epoch": 1.08, "learning_rate": 2.35371325734853e-06, "loss": 0.0003, "step": 9028 }, { "epoch": 1.08, "learning_rate": 2.353641271745651e-06, "loss": 0.0003, "step": 9029 }, { "epoch": 1.08, "learning_rate": 2.3535692861427716e-06, "loss": 0.0003, "step": 9030 }, { "epoch": 1.08, "learning_rate": 2.3534973005398924e-06, "loss": 0.0004, "step": 9031 }, { "epoch": 1.08, "learning_rate": 2.3534253149370127e-06, "loss": 0.0002, "step": 9032 }, { "epoch": 1.08, "learning_rate": 2.3533533293341335e-06, "loss": 0.0003, "step": 9033 }, { "epoch": 1.08, "learning_rate": 2.353281343731254e-06, "loss": 0.0004, "step": 9034 }, { "epoch": 1.08, "learning_rate": 2.3532093581283742e-06, "loss": 0.0004, "step": 9035 }, { "epoch": 1.08, "learning_rate": 2.353137372525495e-06, "loss": 0.0003, "step": 9036 }, { "epoch": 1.08, "learning_rate": 2.3530653869226154e-06, "loss": 0.0002, "step": 9037 }, { "epoch": 1.08, "learning_rate": 2.352993401319736e-06, "loss": 0.0004, "step": 9038 }, { "epoch": 1.08, "learning_rate": 2.3529214157168565e-06, "loss": 0.0002, "step": 9039 }, { "epoch": 1.08, "learning_rate": 2.3528494301139773e-06, "loss": 0.0003, "step": 9040 }, { "epoch": 1.08, "learning_rate": 2.3527774445110976e-06, "loss": 0.0002, "step": 9041 }, { "epoch": 1.08, "learning_rate": 2.3527054589082184e-06, "loss": 0.1813, "step": 9042 }, { "epoch": 1.08, "learning_rate": 2.3526334733053388e-06, "loss": 0.0002, "step": 9043 }, { "epoch": 1.08, "learning_rate": 2.3525614877024595e-06, "loss": 0.0002, "step": 9044 }, { "epoch": 1.08, "learning_rate": 2.3524895020995803e-06, "loss": 0.0003, "step": 9045 }, { "epoch": 1.08, "learning_rate": 2.3524175164967007e-06, "loss": 0.9132, "step": 9046 }, { "epoch": 1.08, "learning_rate": 2.3523455308938215e-06, "loss": 0.0004, "step": 9047 }, { "epoch": 1.08, "learning_rate": 2.352273545290942e-06, "loss": 0.0007, "step": 9048 }, { "epoch": 1.08, "learning_rate": 2.3522015596880626e-06, "loss": 0.0008, "step": 9049 }, { "epoch": 1.08, "learning_rate": 2.352129574085183e-06, "loss": 0.0017, "step": 9050 }, { "epoch": 1.08, "learning_rate": 2.3520575884823037e-06, "loss": 0.0002, "step": 9051 }, { "epoch": 1.08, "learning_rate": 2.351985602879424e-06, "loss": 0.847, "step": 9052 }, { "epoch": 1.08, "learning_rate": 2.351913617276545e-06, "loss": 0.0002, "step": 9053 }, { "epoch": 1.08, "learning_rate": 2.3518416316736652e-06, "loss": 0.0003, "step": 9054 }, { "epoch": 1.09, "learning_rate": 2.351769646070786e-06, "loss": 0.0003, "step": 9055 }, { "epoch": 1.09, "learning_rate": 2.3516976604679064e-06, "loss": 0.0021, "step": 9056 }, { "epoch": 1.09, "learning_rate": 2.351625674865027e-06, "loss": 0.0004, "step": 9057 }, { "epoch": 1.09, "learning_rate": 2.3515536892621475e-06, "loss": 0.0007, "step": 9058 }, { "epoch": 1.09, "learning_rate": 2.3514817036592683e-06, "loss": 0.0003, "step": 9059 }, { "epoch": 1.09, "learning_rate": 2.3514097180563886e-06, "loss": 0.0002, "step": 9060 }, { "epoch": 1.09, "learning_rate": 2.3513377324535094e-06, "loss": 0.0003, "step": 9061 }, { "epoch": 1.09, "learning_rate": 2.35126574685063e-06, "loss": 0.0002, "step": 9062 }, { "epoch": 1.09, "learning_rate": 2.3511937612477505e-06, "loss": 0.0003, "step": 9063 }, { "epoch": 1.09, "learning_rate": 2.3511217756448713e-06, "loss": 0.0003, "step": 9064 }, { "epoch": 1.09, "learning_rate": 2.3510497900419917e-06, "loss": 0.0002, "step": 9065 }, { "epoch": 1.09, "learning_rate": 2.3509778044391124e-06, "loss": 0.0003, "step": 9066 }, { "epoch": 1.09, "learning_rate": 2.350905818836233e-06, "loss": 0.0002, "step": 9067 }, { "epoch": 1.09, "learning_rate": 2.3508338332333536e-06, "loss": 1.0818, "step": 9068 }, { "epoch": 1.09, "learning_rate": 2.350761847630474e-06, "loss": 0.0002, "step": 9069 }, { "epoch": 1.09, "learning_rate": 2.3506898620275943e-06, "loss": 0.0004, "step": 9070 }, { "epoch": 1.09, "learning_rate": 2.350617876424715e-06, "loss": 0.0003, "step": 9071 }, { "epoch": 1.09, "learning_rate": 2.3505458908218354e-06, "loss": 0.0003, "step": 9072 }, { "epoch": 1.09, "learning_rate": 2.350473905218956e-06, "loss": 0.0002, "step": 9073 }, { "epoch": 1.09, "learning_rate": 2.3504019196160766e-06, "loss": 0.2752, "step": 9074 }, { "epoch": 1.09, "learning_rate": 2.3503299340131973e-06, "loss": 0.0002, "step": 9075 }, { "epoch": 1.09, "learning_rate": 2.350257948410318e-06, "loss": 0.0003, "step": 9076 }, { "epoch": 1.09, "learning_rate": 2.350185962807439e-06, "loss": 0.0009, "step": 9077 }, { "epoch": 1.09, "learning_rate": 2.3501139772045592e-06, "loss": 0.0004, "step": 9078 }, { "epoch": 1.09, "learning_rate": 2.35004199160168e-06, "loss": 0.0004, "step": 9079 }, { "epoch": 1.09, "learning_rate": 2.3499700059988004e-06, "loss": 0.0004, "step": 9080 }, { "epoch": 1.09, "learning_rate": 2.3498980203959207e-06, "loss": 0.0004, "step": 9081 }, { "epoch": 1.09, "learning_rate": 2.3498260347930415e-06, "loss": 0.0002, "step": 9082 }, { "epoch": 1.09, "learning_rate": 2.349754049190162e-06, "loss": 0.0043, "step": 9083 }, { "epoch": 1.09, "learning_rate": 2.3496820635872826e-06, "loss": 0.0004, "step": 9084 }, { "epoch": 1.09, "learning_rate": 2.349610077984403e-06, "loss": 0.6931, "step": 9085 }, { "epoch": 1.09, "learning_rate": 2.3495380923815238e-06, "loss": 0.0012, "step": 9086 }, { "epoch": 1.09, "learning_rate": 2.349466106778644e-06, "loss": 0.0003, "step": 9087 }, { "epoch": 1.09, "learning_rate": 2.349394121175765e-06, "loss": 0.0002, "step": 9088 }, { "epoch": 1.09, "learning_rate": 2.3493221355728853e-06, "loss": 0.0003, "step": 9089 }, { "epoch": 1.09, "learning_rate": 2.349250149970006e-06, "loss": 0.0002, "step": 9090 }, { "epoch": 1.09, "learning_rate": 2.349178164367127e-06, "loss": 0.0003, "step": 9091 }, { "epoch": 1.09, "learning_rate": 2.349106178764247e-06, "loss": 0.0002, "step": 9092 }, { "epoch": 1.09, "learning_rate": 2.349034193161368e-06, "loss": 0.0003, "step": 9093 }, { "epoch": 1.09, "learning_rate": 2.3489622075584883e-06, "loss": 0.0003, "step": 9094 }, { "epoch": 1.09, "learning_rate": 2.348890221955609e-06, "loss": 0.0003, "step": 9095 }, { "epoch": 1.09, "learning_rate": 2.3488182363527294e-06, "loss": 0.0003, "step": 9096 }, { "epoch": 1.09, "learning_rate": 2.3487462507498502e-06, "loss": 0.0003, "step": 9097 }, { "epoch": 1.09, "learning_rate": 2.3486742651469706e-06, "loss": 0.0003, "step": 9098 }, { "epoch": 1.09, "learning_rate": 2.3486022795440914e-06, "loss": 0.0003, "step": 9099 }, { "epoch": 1.09, "learning_rate": 2.3485302939412117e-06, "loss": 0.0022, "step": 9100 }, { "epoch": 1.09, "learning_rate": 2.3484583083383325e-06, "loss": 0.9602, "step": 9101 }, { "epoch": 1.09, "learning_rate": 2.348386322735453e-06, "loss": 0.0004, "step": 9102 }, { "epoch": 1.09, "learning_rate": 2.3483143371325736e-06, "loss": 0.0003, "step": 9103 }, { "epoch": 1.09, "learning_rate": 2.348242351529694e-06, "loss": 0.0002, "step": 9104 }, { "epoch": 1.09, "learning_rate": 2.3481703659268148e-06, "loss": 0.0002, "step": 9105 }, { "epoch": 1.09, "learning_rate": 2.348098380323935e-06, "loss": 0.0003, "step": 9106 }, { "epoch": 1.09, "learning_rate": 2.348026394721056e-06, "loss": 0.0026, "step": 9107 }, { "epoch": 1.09, "learning_rate": 2.3479544091181767e-06, "loss": 0.002, "step": 9108 }, { "epoch": 1.09, "learning_rate": 2.347882423515297e-06, "loss": 0.0003, "step": 9109 }, { "epoch": 1.09, "learning_rate": 2.347810437912418e-06, "loss": 0.0004, "step": 9110 }, { "epoch": 1.09, "learning_rate": 2.347738452309538e-06, "loss": 0.0054, "step": 9111 }, { "epoch": 1.09, "learning_rate": 2.347666466706659e-06, "loss": 0.0003, "step": 9112 }, { "epoch": 1.09, "learning_rate": 2.3475944811037793e-06, "loss": 0.9727, "step": 9113 }, { "epoch": 1.09, "learning_rate": 2.3475224955009e-06, "loss": 0.0005, "step": 9114 }, { "epoch": 1.09, "learning_rate": 2.3474505098980204e-06, "loss": 0.0003, "step": 9115 }, { "epoch": 1.09, "learning_rate": 2.3473785242951408e-06, "loss": 0.0002, "step": 9116 }, { "epoch": 1.09, "learning_rate": 2.3473065386922616e-06, "loss": 0.0003, "step": 9117 }, { "epoch": 1.09, "learning_rate": 2.347234553089382e-06, "loss": 0.0002, "step": 9118 }, { "epoch": 1.09, "learning_rate": 2.3471625674865027e-06, "loss": 0.0003, "step": 9119 }, { "epoch": 1.09, "learning_rate": 2.347090581883623e-06, "loss": 0.0004, "step": 9120 }, { "epoch": 1.09, "learning_rate": 2.347018596280744e-06, "loss": 0.0003, "step": 9121 }, { "epoch": 1.09, "learning_rate": 2.3469466106778646e-06, "loss": 0.0004, "step": 9122 }, { "epoch": 1.09, "learning_rate": 2.3468746250749854e-06, "loss": 0.0004, "step": 9123 }, { "epoch": 1.09, "learning_rate": 2.3468026394721057e-06, "loss": 0.0003, "step": 9124 }, { "epoch": 1.09, "learning_rate": 2.3467306538692265e-06, "loss": 0.0004, "step": 9125 }, { "epoch": 1.09, "learning_rate": 2.346658668266347e-06, "loss": 0.0005, "step": 9126 }, { "epoch": 1.09, "learning_rate": 2.3465866826634672e-06, "loss": 0.0003, "step": 9127 }, { "epoch": 1.09, "learning_rate": 2.346514697060588e-06, "loss": 0.0004, "step": 9128 }, { "epoch": 1.09, "learning_rate": 2.3464427114577084e-06, "loss": 0.0002, "step": 9129 }, { "epoch": 1.09, "learning_rate": 2.346370725854829e-06, "loss": 0.0003, "step": 9130 }, { "epoch": 1.09, "learning_rate": 2.3462987402519495e-06, "loss": 0.0002, "step": 9131 }, { "epoch": 1.09, "learning_rate": 2.3462267546490703e-06, "loss": 0.0003, "step": 9132 }, { "epoch": 1.09, "learning_rate": 2.3461547690461906e-06, "loss": 0.0004, "step": 9133 }, { "epoch": 1.09, "learning_rate": 2.3460827834433114e-06, "loss": 0.0023, "step": 9134 }, { "epoch": 1.09, "learning_rate": 2.3460107978404318e-06, "loss": 0.0003, "step": 9135 }, { "epoch": 1.09, "learning_rate": 2.3459388122375525e-06, "loss": 0.0003, "step": 9136 }, { "epoch": 1.09, "learning_rate": 2.3458668266346733e-06, "loss": 0.1003, "step": 9137 }, { "epoch": 1.1, "learning_rate": 2.3457948410317937e-06, "loss": 0.0003, "step": 9138 }, { "epoch": 1.1, "learning_rate": 2.3457228554289145e-06, "loss": 0.0004, "step": 9139 }, { "epoch": 1.1, "learning_rate": 2.345650869826035e-06, "loss": 0.0006, "step": 9140 }, { "epoch": 1.1, "learning_rate": 2.3455788842231556e-06, "loss": 0.0004, "step": 9141 }, { "epoch": 1.1, "learning_rate": 2.345506898620276e-06, "loss": 0.0004, "step": 9142 }, { "epoch": 1.1, "learning_rate": 2.3454349130173967e-06, "loss": 0.0004, "step": 9143 }, { "epoch": 1.1, "learning_rate": 2.345362927414517e-06, "loss": 0.0007, "step": 9144 }, { "epoch": 1.1, "learning_rate": 2.345290941811638e-06, "loss": 0.0005, "step": 9145 }, { "epoch": 1.1, "learning_rate": 2.345218956208758e-06, "loss": 0.0607, "step": 9146 }, { "epoch": 1.1, "learning_rate": 2.345146970605879e-06, "loss": 0.0004, "step": 9147 }, { "epoch": 1.1, "learning_rate": 2.3450749850029993e-06, "loss": 0.0002, "step": 9148 }, { "epoch": 1.1, "learning_rate": 2.34500299940012e-06, "loss": 0.0003, "step": 9149 }, { "epoch": 1.1, "learning_rate": 2.3449310137972405e-06, "loss": 0.8074, "step": 9150 }, { "epoch": 1.1, "learning_rate": 2.3448590281943613e-06, "loss": 0.0003, "step": 9151 }, { "epoch": 1.1, "learning_rate": 2.3447870425914816e-06, "loss": 0.0005, "step": 9152 }, { "epoch": 1.1, "learning_rate": 2.3447150569886024e-06, "loss": 0.0003, "step": 9153 }, { "epoch": 1.1, "learning_rate": 2.344643071385723e-06, "loss": 0.0004, "step": 9154 }, { "epoch": 1.1, "learning_rate": 2.3445710857828435e-06, "loss": 0.0004, "step": 9155 }, { "epoch": 1.1, "learning_rate": 2.3444991001799643e-06, "loss": 0.0004, "step": 9156 }, { "epoch": 1.1, "learning_rate": 2.3444271145770847e-06, "loss": 0.0003, "step": 9157 }, { "epoch": 1.1, "learning_rate": 2.3443551289742054e-06, "loss": 0.0002, "step": 9158 }, { "epoch": 1.1, "learning_rate": 2.344283143371326e-06, "loss": 0.0192, "step": 9159 }, { "epoch": 1.1, "learning_rate": 2.3442111577684466e-06, "loss": 0.0004, "step": 9160 }, { "epoch": 1.1, "learning_rate": 2.344139172165567e-06, "loss": 0.0003, "step": 9161 }, { "epoch": 1.1, "learning_rate": 2.3440671865626873e-06, "loss": 0.0021, "step": 9162 }, { "epoch": 1.1, "learning_rate": 2.343995200959808e-06, "loss": 0.0015, "step": 9163 }, { "epoch": 1.1, "learning_rate": 2.3439232153569284e-06, "loss": 0.0004, "step": 9164 }, { "epoch": 1.1, "learning_rate": 2.343851229754049e-06, "loss": 0.0003, "step": 9165 }, { "epoch": 1.1, "learning_rate": 2.3437792441511695e-06, "loss": 0.0037, "step": 9166 }, { "epoch": 1.1, "learning_rate": 2.3437072585482903e-06, "loss": 0.0003, "step": 9167 }, { "epoch": 1.1, "learning_rate": 2.343635272945411e-06, "loss": 0.0165, "step": 9168 }, { "epoch": 1.1, "learning_rate": 2.343563287342532e-06, "loss": 0.0004, "step": 9169 }, { "epoch": 1.1, "learning_rate": 2.3434913017396522e-06, "loss": 0.0003, "step": 9170 }, { "epoch": 1.1, "learning_rate": 2.343419316136773e-06, "loss": 0.0003, "step": 9171 }, { "epoch": 1.1, "learning_rate": 2.3433473305338934e-06, "loss": 0.0003, "step": 9172 }, { "epoch": 1.1, "learning_rate": 2.3432753449310137e-06, "loss": 0.0003, "step": 9173 }, { "epoch": 1.1, "learning_rate": 2.3432033593281345e-06, "loss": 0.0003, "step": 9174 }, { "epoch": 1.1, "learning_rate": 2.343131373725255e-06, "loss": 0.0003, "step": 9175 }, { "epoch": 1.1, "learning_rate": 2.3430593881223756e-06, "loss": 0.0005, "step": 9176 }, { "epoch": 1.1, "learning_rate": 2.342987402519496e-06, "loss": 0.0854, "step": 9177 }, { "epoch": 1.1, "learning_rate": 2.3429154169166168e-06, "loss": 0.0004, "step": 9178 }, { "epoch": 1.1, "learning_rate": 2.342843431313737e-06, "loss": 0.0004, "step": 9179 }, { "epoch": 1.1, "learning_rate": 2.342771445710858e-06, "loss": 0.0003, "step": 9180 }, { "epoch": 1.1, "learning_rate": 2.3426994601079783e-06, "loss": 0.0003, "step": 9181 }, { "epoch": 1.1, "learning_rate": 2.342627474505099e-06, "loss": 0.0004, "step": 9182 }, { "epoch": 1.1, "learning_rate": 2.3425554889022194e-06, "loss": 0.0878, "step": 9183 }, { "epoch": 1.1, "learning_rate": 2.34248350329934e-06, "loss": 0.0003, "step": 9184 }, { "epoch": 1.1, "learning_rate": 2.342411517696461e-06, "loss": 0.0003, "step": 9185 }, { "epoch": 1.1, "learning_rate": 2.3423395320935813e-06, "loss": 0.8277, "step": 9186 }, { "epoch": 1.1, "learning_rate": 2.342267546490702e-06, "loss": 0.0004, "step": 9187 }, { "epoch": 1.1, "learning_rate": 2.3421955608878224e-06, "loss": 0.0003, "step": 9188 }, { "epoch": 1.1, "learning_rate": 2.3421235752849432e-06, "loss": 0.0006, "step": 9189 }, { "epoch": 1.1, "learning_rate": 2.3420515896820636e-06, "loss": 0.2098, "step": 9190 }, { "epoch": 1.1, "learning_rate": 2.3419796040791843e-06, "loss": 0.0004, "step": 9191 }, { "epoch": 1.1, "learning_rate": 2.3419076184763047e-06, "loss": 0.0003, "step": 9192 }, { "epoch": 1.1, "learning_rate": 2.3418356328734255e-06, "loss": 0.004, "step": 9193 }, { "epoch": 1.1, "learning_rate": 2.341763647270546e-06, "loss": 0.0002, "step": 9194 }, { "epoch": 1.1, "learning_rate": 2.3416916616676666e-06, "loss": 0.0003, "step": 9195 }, { "epoch": 1.1, "learning_rate": 2.341619676064787e-06, "loss": 0.0005, "step": 9196 }, { "epoch": 1.1, "learning_rate": 2.3415476904619073e-06, "loss": 0.0004, "step": 9197 }, { "epoch": 1.1, "learning_rate": 2.341475704859028e-06, "loss": 0.0005, "step": 9198 }, { "epoch": 1.1, "learning_rate": 2.341403719256149e-06, "loss": 0.0002, "step": 9199 }, { "epoch": 1.1, "learning_rate": 2.3413317336532697e-06, "loss": 0.0003, "step": 9200 }, { "epoch": 1.1, "learning_rate": 2.34125974805039e-06, "loss": 0.0002, "step": 9201 }, { "epoch": 1.1, "learning_rate": 2.341187762447511e-06, "loss": 0.0004, "step": 9202 }, { "epoch": 1.1, "learning_rate": 2.341115776844631e-06, "loss": 0.0002, "step": 9203 }, { "epoch": 1.1, "learning_rate": 2.341043791241752e-06, "loss": 0.0003, "step": 9204 }, { "epoch": 1.1, "learning_rate": 2.3409718056388723e-06, "loss": 0.0004, "step": 9205 }, { "epoch": 1.1, "learning_rate": 2.340899820035993e-06, "loss": 0.0009, "step": 9206 }, { "epoch": 1.1, "learning_rate": 2.3408278344331134e-06, "loss": 0.0086, "step": 9207 }, { "epoch": 1.1, "learning_rate": 2.3407558488302338e-06, "loss": 0.0003, "step": 9208 }, { "epoch": 1.1, "learning_rate": 2.3406838632273546e-06, "loss": 0.0003, "step": 9209 }, { "epoch": 1.1, "learning_rate": 2.340611877624475e-06, "loss": 0.4056, "step": 9210 }, { "epoch": 1.1, "learning_rate": 2.3405398920215957e-06, "loss": 0.0002, "step": 9211 }, { "epoch": 1.1, "learning_rate": 2.340467906418716e-06, "loss": 0.0004, "step": 9212 }, { "epoch": 1.1, "learning_rate": 2.340395920815837e-06, "loss": 0.0002, "step": 9213 }, { "epoch": 1.1, "learning_rate": 2.3403239352129576e-06, "loss": 0.0003, "step": 9214 }, { "epoch": 1.1, "learning_rate": 2.340251949610078e-06, "loss": 0.0012, "step": 9215 }, { "epoch": 1.1, "learning_rate": 2.3401799640071987e-06, "loss": 0.0003, "step": 9216 }, { "epoch": 1.1, "learning_rate": 2.3401079784043195e-06, "loss": 0.0002, "step": 9217 }, { "epoch": 1.1, "learning_rate": 2.34003599280144e-06, "loss": 0.0003, "step": 9218 }, { "epoch": 1.1, "learning_rate": 2.3399640071985602e-06, "loss": 0.0002, "step": 9219 }, { "epoch": 1.1, "learning_rate": 2.339892021595681e-06, "loss": 0.0003, "step": 9220 }, { "epoch": 1.1, "learning_rate": 2.3398200359928014e-06, "loss": 0.0002, "step": 9221 }, { "epoch": 1.11, "learning_rate": 2.339748050389922e-06, "loss": 0.0004, "step": 9222 }, { "epoch": 1.11, "learning_rate": 2.3396760647870425e-06, "loss": 0.0004, "step": 9223 }, { "epoch": 1.11, "learning_rate": 2.3396040791841633e-06, "loss": 0.0003, "step": 9224 }, { "epoch": 1.11, "learning_rate": 2.3395320935812836e-06, "loss": 0.0003, "step": 9225 }, { "epoch": 1.11, "learning_rate": 2.3394601079784044e-06, "loss": 0.0003, "step": 9226 }, { "epoch": 1.11, "learning_rate": 2.3393881223755248e-06, "loss": 0.0003, "step": 9227 }, { "epoch": 1.11, "learning_rate": 2.3393161367726455e-06, "loss": 0.0003, "step": 9228 }, { "epoch": 1.11, "learning_rate": 2.339244151169766e-06, "loss": 0.0004, "step": 9229 }, { "epoch": 1.11, "learning_rate": 2.3391721655668867e-06, "loss": 0.0253, "step": 9230 }, { "epoch": 1.11, "learning_rate": 2.3391001799640074e-06, "loss": 0.0004, "step": 9231 }, { "epoch": 1.11, "learning_rate": 2.339028194361128e-06, "loss": 0.0223, "step": 9232 }, { "epoch": 1.11, "learning_rate": 2.3389562087582486e-06, "loss": 0.0002, "step": 9233 }, { "epoch": 1.11, "learning_rate": 2.338884223155369e-06, "loss": 0.0003, "step": 9234 }, { "epoch": 1.11, "learning_rate": 2.3388122375524897e-06, "loss": 0.0003, "step": 9235 }, { "epoch": 1.11, "learning_rate": 2.33874025194961e-06, "loss": 0.0003, "step": 9236 }, { "epoch": 1.11, "learning_rate": 2.338668266346731e-06, "loss": 0.0781, "step": 9237 }, { "epoch": 1.11, "learning_rate": 2.338596280743851e-06, "loss": 0.0002, "step": 9238 }, { "epoch": 1.11, "learning_rate": 2.338524295140972e-06, "loss": 0.0002, "step": 9239 }, { "epoch": 1.11, "learning_rate": 2.3384523095380923e-06, "loss": 0.0004, "step": 9240 }, { "epoch": 1.11, "learning_rate": 2.338380323935213e-06, "loss": 0.0002, "step": 9241 }, { "epoch": 1.11, "learning_rate": 2.3383083383323335e-06, "loss": 0.0002, "step": 9242 }, { "epoch": 1.11, "learning_rate": 2.338236352729454e-06, "loss": 0.8592, "step": 9243 }, { "epoch": 1.11, "learning_rate": 2.3381643671265746e-06, "loss": 0.0003, "step": 9244 }, { "epoch": 1.11, "learning_rate": 2.3380923815236954e-06, "loss": 0.6176, "step": 9245 }, { "epoch": 1.11, "learning_rate": 2.338020395920816e-06, "loss": 0.0002, "step": 9246 }, { "epoch": 1.11, "learning_rate": 2.3379484103179365e-06, "loss": 0.0002, "step": 9247 }, { "epoch": 1.11, "learning_rate": 2.3378764247150573e-06, "loss": 0.0003, "step": 9248 }, { "epoch": 1.11, "learning_rate": 2.3378044391121776e-06, "loss": 0.0003, "step": 9249 }, { "epoch": 1.11, "learning_rate": 2.3377324535092984e-06, "loss": 0.0002, "step": 9250 }, { "epoch": 1.11, "learning_rate": 2.3376604679064188e-06, "loss": 0.0002, "step": 9251 }, { "epoch": 1.11, "learning_rate": 2.3375884823035396e-06, "loss": 0.0003, "step": 9252 }, { "epoch": 1.11, "learning_rate": 2.33751649670066e-06, "loss": 0.0003, "step": 9253 }, { "epoch": 1.11, "learning_rate": 2.3374445110977803e-06, "loss": 0.0003, "step": 9254 }, { "epoch": 1.11, "learning_rate": 2.337372525494901e-06, "loss": 0.0002, "step": 9255 }, { "epoch": 1.11, "learning_rate": 2.3373005398920214e-06, "loss": 0.2682, "step": 9256 }, { "epoch": 1.11, "learning_rate": 2.337228554289142e-06, "loss": 0.0002, "step": 9257 }, { "epoch": 1.11, "learning_rate": 2.3371565686862625e-06, "loss": 0.0003, "step": 9258 }, { "epoch": 1.11, "learning_rate": 2.3370845830833833e-06, "loss": 0.206, "step": 9259 }, { "epoch": 1.11, "learning_rate": 2.337012597480504e-06, "loss": 0.0004, "step": 9260 }, { "epoch": 1.11, "learning_rate": 2.3369406118776244e-06, "loss": 0.0007, "step": 9261 }, { "epoch": 1.11, "learning_rate": 2.3368686262747452e-06, "loss": 0.0003, "step": 9262 }, { "epoch": 1.11, "learning_rate": 2.336796640671866e-06, "loss": 0.0003, "step": 9263 }, { "epoch": 1.11, "learning_rate": 2.3367246550689864e-06, "loss": 0.0003, "step": 9264 }, { "epoch": 1.11, "learning_rate": 2.3366526694661067e-06, "loss": 0.0002, "step": 9265 }, { "epoch": 1.11, "learning_rate": 2.3365806838632275e-06, "loss": 0.0005, "step": 9266 }, { "epoch": 1.11, "learning_rate": 2.336508698260348e-06, "loss": 0.0003, "step": 9267 }, { "epoch": 1.11, "learning_rate": 2.3364367126574686e-06, "loss": 0.0003, "step": 9268 }, { "epoch": 1.11, "learning_rate": 2.336364727054589e-06, "loss": 0.0003, "step": 9269 }, { "epoch": 1.11, "learning_rate": 2.3362927414517098e-06, "loss": 0.0005, "step": 9270 }, { "epoch": 1.11, "learning_rate": 2.33622075584883e-06, "loss": 0.0004, "step": 9271 }, { "epoch": 1.11, "learning_rate": 2.336148770245951e-06, "loss": 0.0004, "step": 9272 }, { "epoch": 1.11, "learning_rate": 2.3360767846430713e-06, "loss": 0.0002, "step": 9273 }, { "epoch": 1.11, "learning_rate": 2.336004799040192e-06, "loss": 0.0003, "step": 9274 }, { "epoch": 1.11, "learning_rate": 2.3359328134373124e-06, "loss": 0.0004, "step": 9275 }, { "epoch": 1.11, "learning_rate": 2.335860827834433e-06, "loss": 0.0003, "step": 9276 }, { "epoch": 1.11, "learning_rate": 2.335788842231554e-06, "loss": 0.0006, "step": 9277 }, { "epoch": 1.11, "learning_rate": 2.3357168566286743e-06, "loss": 0.0001, "step": 9278 }, { "epoch": 1.11, "learning_rate": 2.335644871025795e-06, "loss": 0.0004, "step": 9279 }, { "epoch": 1.11, "learning_rate": 2.3355728854229154e-06, "loss": 0.0002, "step": 9280 }, { "epoch": 1.11, "learning_rate": 2.335500899820036e-06, "loss": 0.0013, "step": 9281 }, { "epoch": 1.11, "learning_rate": 2.3354289142171566e-06, "loss": 0.0002, "step": 9282 }, { "epoch": 1.11, "learning_rate": 2.3353569286142773e-06, "loss": 0.0004, "step": 9283 }, { "epoch": 1.11, "learning_rate": 2.3352849430113977e-06, "loss": 0.0002, "step": 9284 }, { "epoch": 1.11, "learning_rate": 2.3352129574085185e-06, "loss": 0.0004, "step": 9285 }, { "epoch": 1.11, "learning_rate": 2.335140971805639e-06, "loss": 1.1051, "step": 9286 }, { "epoch": 1.11, "learning_rate": 2.3350689862027596e-06, "loss": 0.0003, "step": 9287 }, { "epoch": 1.11, "learning_rate": 2.33499700059988e-06, "loss": 0.0002, "step": 9288 }, { "epoch": 1.11, "learning_rate": 2.3349250149970003e-06, "loss": 0.0003, "step": 9289 }, { "epoch": 1.11, "learning_rate": 2.334853029394121e-06, "loss": 0.0004, "step": 9290 }, { "epoch": 1.11, "learning_rate": 2.334781043791242e-06, "loss": 0.0002, "step": 9291 }, { "epoch": 1.11, "learning_rate": 2.3347090581883627e-06, "loss": 0.0013, "step": 9292 }, { "epoch": 1.11, "learning_rate": 2.334637072585483e-06, "loss": 0.0003, "step": 9293 }, { "epoch": 1.11, "learning_rate": 2.3345650869826038e-06, "loss": 0.971, "step": 9294 }, { "epoch": 1.11, "learning_rate": 2.334493101379724e-06, "loss": 0.0005, "step": 9295 }, { "epoch": 1.11, "learning_rate": 2.334421115776845e-06, "loss": 0.0003, "step": 9296 }, { "epoch": 1.11, "learning_rate": 2.3343491301739653e-06, "loss": 0.0004, "step": 9297 }, { "epoch": 1.11, "learning_rate": 2.334277144571086e-06, "loss": 0.0002, "step": 9298 }, { "epoch": 1.11, "learning_rate": 2.3342051589682064e-06, "loss": 0.0003, "step": 9299 }, { "epoch": 1.11, "learning_rate": 2.3341331733653268e-06, "loss": 0.0003, "step": 9300 }, { "epoch": 1.11, "learning_rate": 2.3340611877624475e-06, "loss": 0.0003, "step": 9301 }, { "epoch": 1.11, "learning_rate": 2.333989202159568e-06, "loss": 0.0002, "step": 9302 }, { "epoch": 1.11, "learning_rate": 2.3339172165566887e-06, "loss": 0.0008, "step": 9303 }, { "epoch": 1.11, "learning_rate": 2.333845230953809e-06, "loss": 0.0002, "step": 9304 }, { "epoch": 1.12, "learning_rate": 2.33377324535093e-06, "loss": 0.0003, "step": 9305 }, { "epoch": 1.12, "learning_rate": 2.33370125974805e-06, "loss": 0.0004, "step": 9306 }, { "epoch": 1.12, "learning_rate": 2.333629274145171e-06, "loss": 0.0007, "step": 9307 }, { "epoch": 1.12, "learning_rate": 2.3335572885422917e-06, "loss": 0.0005, "step": 9308 }, { "epoch": 1.12, "learning_rate": 2.3334853029394125e-06, "loss": 0.0004, "step": 9309 }, { "epoch": 1.12, "learning_rate": 2.333413317336533e-06, "loss": 0.0007, "step": 9310 }, { "epoch": 1.12, "learning_rate": 2.3333413317336536e-06, "loss": 0.0006, "step": 9311 }, { "epoch": 1.12, "learning_rate": 2.333269346130774e-06, "loss": 0.0002, "step": 9312 }, { "epoch": 1.12, "learning_rate": 2.3331973605278943e-06, "loss": 0.0003, "step": 9313 }, { "epoch": 1.12, "learning_rate": 2.333125374925015e-06, "loss": 0.0003, "step": 9314 }, { "epoch": 1.12, "learning_rate": 2.3330533893221355e-06, "loss": 0.0095, "step": 9315 }, { "epoch": 1.12, "learning_rate": 2.3329814037192563e-06, "loss": 0.0002, "step": 9316 }, { "epoch": 1.12, "learning_rate": 2.3329094181163766e-06, "loss": 0.0007, "step": 9317 }, { "epoch": 1.12, "learning_rate": 2.3328374325134974e-06, "loss": 0.5039, "step": 9318 }, { "epoch": 1.12, "learning_rate": 2.3327654469106177e-06, "loss": 0.0003, "step": 9319 }, { "epoch": 1.12, "learning_rate": 2.3326934613077385e-06, "loss": 0.0003, "step": 9320 }, { "epoch": 1.12, "learning_rate": 2.332621475704859e-06, "loss": 0.0003, "step": 9321 }, { "epoch": 1.12, "learning_rate": 2.3325494901019797e-06, "loss": 0.0005, "step": 9322 }, { "epoch": 1.12, "learning_rate": 2.3324775044991004e-06, "loss": 0.0004, "step": 9323 }, { "epoch": 1.12, "learning_rate": 2.332405518896221e-06, "loss": 0.0003, "step": 9324 }, { "epoch": 1.12, "learning_rate": 2.3323335332933416e-06, "loss": 0.0002, "step": 9325 }, { "epoch": 1.12, "learning_rate": 2.332261547690462e-06, "loss": 0.0002, "step": 9326 }, { "epoch": 1.12, "learning_rate": 2.3321895620875827e-06, "loss": 0.0005, "step": 9327 }, { "epoch": 1.12, "learning_rate": 2.332117576484703e-06, "loss": 0.0004, "step": 9328 }, { "epoch": 1.12, "learning_rate": 2.332045590881824e-06, "loss": 0.0008, "step": 9329 }, { "epoch": 1.12, "learning_rate": 2.331973605278944e-06, "loss": 0.0003, "step": 9330 }, { "epoch": 1.12, "learning_rate": 2.331901619676065e-06, "loss": 0.0002, "step": 9331 }, { "epoch": 1.12, "learning_rate": 2.3318296340731853e-06, "loss": 0.0003, "step": 9332 }, { "epoch": 1.12, "learning_rate": 2.331757648470306e-06, "loss": 0.0002, "step": 9333 }, { "epoch": 1.12, "learning_rate": 2.3316856628674265e-06, "loss": 0.4136, "step": 9334 }, { "epoch": 1.12, "learning_rate": 2.331613677264547e-06, "loss": 0.0011, "step": 9335 }, { "epoch": 1.12, "learning_rate": 2.3315416916616676e-06, "loss": 0.0002, "step": 9336 }, { "epoch": 1.12, "learning_rate": 2.3314697060587884e-06, "loss": 0.0004, "step": 9337 }, { "epoch": 1.12, "learning_rate": 2.3313977204559087e-06, "loss": 0.0095, "step": 9338 }, { "epoch": 1.12, "learning_rate": 2.3313257348530295e-06, "loss": 0.0003, "step": 9339 }, { "epoch": 1.12, "learning_rate": 2.3312537492501503e-06, "loss": 0.0006, "step": 9340 }, { "epoch": 1.12, "learning_rate": 2.3311817636472706e-06, "loss": 0.2282, "step": 9341 }, { "epoch": 1.12, "learning_rate": 2.3311097780443914e-06, "loss": 0.0003, "step": 9342 }, { "epoch": 1.12, "learning_rate": 2.3310377924415118e-06, "loss": 0.0003, "step": 9343 }, { "epoch": 1.12, "learning_rate": 2.3309658068386326e-06, "loss": 0.0003, "step": 9344 }, { "epoch": 1.12, "learning_rate": 2.330893821235753e-06, "loss": 0.0003, "step": 9345 }, { "epoch": 1.12, "learning_rate": 2.3308218356328733e-06, "loss": 0.0002, "step": 9346 }, { "epoch": 1.12, "learning_rate": 2.330749850029994e-06, "loss": 0.9937, "step": 9347 }, { "epoch": 1.12, "learning_rate": 2.3306778644271144e-06, "loss": 0.0003, "step": 9348 }, { "epoch": 1.12, "learning_rate": 2.330605878824235e-06, "loss": 0.0003, "step": 9349 }, { "epoch": 1.12, "learning_rate": 2.3305338932213555e-06, "loss": 0.0004, "step": 9350 }, { "epoch": 1.12, "learning_rate": 2.3304619076184763e-06, "loss": 0.0004, "step": 9351 }, { "epoch": 1.12, "learning_rate": 2.3303899220155967e-06, "loss": 0.0013, "step": 9352 }, { "epoch": 1.12, "learning_rate": 2.3303179364127174e-06, "loss": 0.0004, "step": 9353 }, { "epoch": 1.12, "learning_rate": 2.3302459508098382e-06, "loss": 0.3294, "step": 9354 }, { "epoch": 1.12, "learning_rate": 2.330173965206959e-06, "loss": 0.0003, "step": 9355 }, { "epoch": 1.12, "learning_rate": 2.3301019796040794e-06, "loss": 0.0003, "step": 9356 }, { "epoch": 1.12, "learning_rate": 2.3300299940012e-06, "loss": 0.0004, "step": 9357 }, { "epoch": 1.12, "learning_rate": 2.3299580083983205e-06, "loss": 0.0003, "step": 9358 }, { "epoch": 1.12, "learning_rate": 2.329886022795441e-06, "loss": 0.0004, "step": 9359 }, { "epoch": 1.12, "learning_rate": 2.3298140371925616e-06, "loss": 0.5625, "step": 9360 }, { "epoch": 1.12, "learning_rate": 2.329742051589682e-06, "loss": 0.0003, "step": 9361 }, { "epoch": 1.12, "learning_rate": 2.3296700659868028e-06, "loss": 0.0003, "step": 9362 }, { "epoch": 1.12, "learning_rate": 2.329598080383923e-06, "loss": 0.0003, "step": 9363 }, { "epoch": 1.12, "learning_rate": 2.329526094781044e-06, "loss": 0.0003, "step": 9364 }, { "epoch": 1.12, "learning_rate": 2.3294541091781642e-06, "loss": 0.0003, "step": 9365 }, { "epoch": 1.12, "learning_rate": 2.329382123575285e-06, "loss": 0.0004, "step": 9366 }, { "epoch": 1.12, "learning_rate": 2.3293101379724054e-06, "loss": 0.0003, "step": 9367 }, { "epoch": 1.12, "learning_rate": 2.329238152369526e-06, "loss": 0.3207, "step": 9368 }, { "epoch": 1.12, "learning_rate": 2.329166166766647e-06, "loss": 0.0002, "step": 9369 }, { "epoch": 1.12, "learning_rate": 2.3290941811637673e-06, "loss": 0.0003, "step": 9370 }, { "epoch": 1.12, "learning_rate": 2.329022195560888e-06, "loss": 0.0006, "step": 9371 }, { "epoch": 1.12, "learning_rate": 2.3289502099580084e-06, "loss": 0.0003, "step": 9372 }, { "epoch": 1.12, "learning_rate": 2.328878224355129e-06, "loss": 0.0005, "step": 9373 }, { "epoch": 1.12, "learning_rate": 2.3288062387522496e-06, "loss": 0.0002, "step": 9374 }, { "epoch": 1.12, "learning_rate": 2.3287342531493703e-06, "loss": 0.0003, "step": 9375 }, { "epoch": 1.12, "learning_rate": 2.3286622675464907e-06, "loss": 0.6286, "step": 9376 }, { "epoch": 1.12, "learning_rate": 2.3285902819436115e-06, "loss": 0.0002, "step": 9377 }, { "epoch": 1.12, "learning_rate": 2.328518296340732e-06, "loss": 0.0004, "step": 9378 }, { "epoch": 1.12, "learning_rate": 2.3284463107378526e-06, "loss": 0.0003, "step": 9379 }, { "epoch": 1.12, "learning_rate": 2.328374325134973e-06, "loss": 0.0006, "step": 9380 }, { "epoch": 1.12, "learning_rate": 2.3283023395320933e-06, "loss": 0.0004, "step": 9381 }, { "epoch": 1.12, "learning_rate": 2.328230353929214e-06, "loss": 0.0004, "step": 9382 }, { "epoch": 1.12, "learning_rate": 2.328158368326335e-06, "loss": 0.0003, "step": 9383 }, { "epoch": 1.12, "learning_rate": 2.3280863827234552e-06, "loss": 0.9677, "step": 9384 }, { "epoch": 1.12, "learning_rate": 2.328014397120576e-06, "loss": 0.0003, "step": 9385 }, { "epoch": 1.12, "learning_rate": 2.3279424115176968e-06, "loss": 0.0002, "step": 9386 }, { "epoch": 1.12, "learning_rate": 2.327870425914817e-06, "loss": 0.0002, "step": 9387 }, { "epoch": 1.12, "learning_rate": 2.327798440311938e-06, "loss": 0.0002, "step": 9388 }, { "epoch": 1.13, "learning_rate": 2.3277264547090583e-06, "loss": 0.0003, "step": 9389 }, { "epoch": 1.13, "learning_rate": 2.327654469106179e-06, "loss": 0.0003, "step": 9390 }, { "epoch": 1.13, "learning_rate": 2.3275824835032994e-06, "loss": 0.0409, "step": 9391 }, { "epoch": 1.13, "learning_rate": 2.3275104979004198e-06, "loss": 0.0007, "step": 9392 }, { "epoch": 1.13, "learning_rate": 2.3274385122975405e-06, "loss": 0.0002, "step": 9393 }, { "epoch": 1.13, "learning_rate": 2.327366526694661e-06, "loss": 0.0985, "step": 9394 }, { "epoch": 1.13, "learning_rate": 2.3272945410917817e-06, "loss": 0.0004, "step": 9395 }, { "epoch": 1.13, "learning_rate": 2.327222555488902e-06, "loss": 0.0003, "step": 9396 }, { "epoch": 1.13, "learning_rate": 2.327150569886023e-06, "loss": 0.0003, "step": 9397 }, { "epoch": 1.13, "learning_rate": 2.327078584283143e-06, "loss": 0.0002, "step": 9398 }, { "epoch": 1.13, "learning_rate": 2.327006598680264e-06, "loss": 0.0003, "step": 9399 }, { "epoch": 1.13, "learning_rate": 2.3269346130773847e-06, "loss": 0.0005, "step": 9400 }, { "epoch": 1.13, "learning_rate": 2.3268626274745055e-06, "loss": 0.7684, "step": 9401 }, { "epoch": 1.13, "learning_rate": 2.326790641871626e-06, "loss": 0.0003, "step": 9402 }, { "epoch": 1.13, "learning_rate": 2.3267186562687466e-06, "loss": 0.0004, "step": 9403 }, { "epoch": 1.13, "learning_rate": 2.326646670665867e-06, "loss": 0.0009, "step": 9404 }, { "epoch": 1.13, "learning_rate": 2.3265746850629873e-06, "loss": 0.0003, "step": 9405 }, { "epoch": 1.13, "learning_rate": 2.326502699460108e-06, "loss": 0.0004, "step": 9406 }, { "epoch": 1.13, "learning_rate": 2.3264307138572285e-06, "loss": 0.0004, "step": 9407 }, { "epoch": 1.13, "learning_rate": 2.3263587282543492e-06, "loss": 0.0004, "step": 9408 }, { "epoch": 1.13, "learning_rate": 2.3262867426514696e-06, "loss": 0.0002, "step": 9409 }, { "epoch": 1.13, "learning_rate": 2.3262147570485904e-06, "loss": 0.0003, "step": 9410 }, { "epoch": 1.13, "learning_rate": 2.3261427714457107e-06, "loss": 0.0002, "step": 9411 }, { "epoch": 1.13, "learning_rate": 2.3260707858428315e-06, "loss": 0.0007, "step": 9412 }, { "epoch": 1.13, "learning_rate": 2.325998800239952e-06, "loss": 0.0003, "step": 9413 }, { "epoch": 1.13, "learning_rate": 2.3259268146370727e-06, "loss": 0.0002, "step": 9414 }, { "epoch": 1.13, "learning_rate": 2.3258548290341934e-06, "loss": 0.0004, "step": 9415 }, { "epoch": 1.13, "learning_rate": 2.3257828434313138e-06, "loss": 0.011, "step": 9416 }, { "epoch": 1.13, "learning_rate": 2.3257108578284346e-06, "loss": 0.0006, "step": 9417 }, { "epoch": 1.13, "learning_rate": 2.325638872225555e-06, "loss": 0.0003, "step": 9418 }, { "epoch": 1.13, "learning_rate": 2.3255668866226757e-06, "loss": 0.0003, "step": 9419 }, { "epoch": 1.13, "learning_rate": 2.325494901019796e-06, "loss": 0.0003, "step": 9420 }, { "epoch": 1.13, "learning_rate": 2.325422915416917e-06, "loss": 0.0004, "step": 9421 }, { "epoch": 1.13, "learning_rate": 2.325350929814037e-06, "loss": 0.375, "step": 9422 }, { "epoch": 1.13, "learning_rate": 2.325278944211158e-06, "loss": 0.0002, "step": 9423 }, { "epoch": 1.13, "learning_rate": 2.3252069586082783e-06, "loss": 0.0002, "step": 9424 }, { "epoch": 1.13, "learning_rate": 2.325134973005399e-06, "loss": 0.0004, "step": 9425 }, { "epoch": 1.13, "learning_rate": 2.3250629874025195e-06, "loss": 0.0002, "step": 9426 }, { "epoch": 1.13, "learning_rate": 2.32499100179964e-06, "loss": 0.0097, "step": 9427 }, { "epoch": 1.13, "learning_rate": 2.3249190161967606e-06, "loss": 0.0059, "step": 9428 }, { "epoch": 1.13, "learning_rate": 2.324847030593881e-06, "loss": 0.0005, "step": 9429 }, { "epoch": 1.13, "learning_rate": 2.3247750449910017e-06, "loss": 0.0003, "step": 9430 }, { "epoch": 1.13, "learning_rate": 2.3247030593881225e-06, "loss": 0.0005, "step": 9431 }, { "epoch": 1.13, "learning_rate": 2.3246310737852433e-06, "loss": 0.0003, "step": 9432 }, { "epoch": 1.13, "learning_rate": 2.3245590881823636e-06, "loss": 0.4964, "step": 9433 }, { "epoch": 1.13, "learning_rate": 2.3244871025794844e-06, "loss": 0.001, "step": 9434 }, { "epoch": 1.13, "learning_rate": 2.3244151169766048e-06, "loss": 0.0002, "step": 9435 }, { "epoch": 1.13, "learning_rate": 2.3243431313737255e-06, "loss": 0.7825, "step": 9436 }, { "epoch": 1.13, "learning_rate": 2.324271145770846e-06, "loss": 0.0005, "step": 9437 }, { "epoch": 1.13, "learning_rate": 2.3241991601679667e-06, "loss": 0.0003, "step": 9438 }, { "epoch": 1.13, "learning_rate": 2.324127174565087e-06, "loss": 0.0006, "step": 9439 }, { "epoch": 1.13, "learning_rate": 2.3240551889622074e-06, "loss": 0.6489, "step": 9440 }, { "epoch": 1.13, "learning_rate": 2.323983203359328e-06, "loss": 0.0003, "step": 9441 }, { "epoch": 1.13, "learning_rate": 2.3239112177564485e-06, "loss": 0.0006, "step": 9442 }, { "epoch": 1.13, "learning_rate": 2.3238392321535693e-06, "loss": 0.0003, "step": 9443 }, { "epoch": 1.13, "learning_rate": 2.3237672465506897e-06, "loss": 0.0002, "step": 9444 }, { "epoch": 1.13, "learning_rate": 2.3236952609478104e-06, "loss": 0.435, "step": 9445 }, { "epoch": 1.13, "learning_rate": 2.323623275344931e-06, "loss": 0.0002, "step": 9446 }, { "epoch": 1.13, "learning_rate": 2.3235512897420516e-06, "loss": 0.0032, "step": 9447 }, { "epoch": 1.13, "learning_rate": 2.3234793041391723e-06, "loss": 0.0005, "step": 9448 }, { "epoch": 1.13, "learning_rate": 2.323407318536293e-06, "loss": 0.0006, "step": 9449 }, { "epoch": 1.13, "learning_rate": 2.3233353329334135e-06, "loss": 0.0004, "step": 9450 }, { "epoch": 1.13, "learning_rate": 2.323263347330534e-06, "loss": 0.0004, "step": 9451 }, { "epoch": 1.13, "learning_rate": 2.3231913617276546e-06, "loss": 0.0003, "step": 9452 }, { "epoch": 1.13, "learning_rate": 2.323119376124775e-06, "loss": 0.0003, "step": 9453 }, { "epoch": 1.13, "learning_rate": 2.3230473905218957e-06, "loss": 0.0008, "step": 9454 }, { "epoch": 1.13, "learning_rate": 2.322975404919016e-06, "loss": 0.0003, "step": 9455 }, { "epoch": 1.13, "learning_rate": 2.322903419316137e-06, "loss": 0.0006, "step": 9456 }, { "epoch": 1.13, "learning_rate": 2.3228314337132572e-06, "loss": 0.0002, "step": 9457 }, { "epoch": 1.13, "learning_rate": 2.322759448110378e-06, "loss": 0.0002, "step": 9458 }, { "epoch": 1.13, "learning_rate": 2.3226874625074984e-06, "loss": 0.0004, "step": 9459 }, { "epoch": 1.13, "learning_rate": 2.322615476904619e-06, "loss": 0.0497, "step": 9460 }, { "epoch": 1.13, "learning_rate": 2.3225434913017395e-06, "loss": 0.0004, "step": 9461 }, { "epoch": 1.13, "learning_rate": 2.3224715056988603e-06, "loss": 0.0002, "step": 9462 }, { "epoch": 1.13, "learning_rate": 2.322399520095981e-06, "loss": 0.0003, "step": 9463 }, { "epoch": 1.13, "learning_rate": 2.3223275344931014e-06, "loss": 0.5702, "step": 9464 }, { "epoch": 1.13, "learning_rate": 2.322255548890222e-06, "loss": 0.0004, "step": 9465 }, { "epoch": 1.13, "learning_rate": 2.3221835632873425e-06, "loss": 0.0003, "step": 9466 }, { "epoch": 1.13, "learning_rate": 2.3221115776844633e-06, "loss": 0.0003, "step": 9467 }, { "epoch": 1.13, "learning_rate": 2.3220395920815837e-06, "loss": 0.0005, "step": 9468 }, { "epoch": 1.13, "learning_rate": 2.3219676064787045e-06, "loss": 0.0007, "step": 9469 }, { "epoch": 1.13, "learning_rate": 2.321895620875825e-06, "loss": 0.0005, "step": 9470 }, { "epoch": 1.13, "learning_rate": 2.3218236352729456e-06, "loss": 0.0002, "step": 9471 }, { "epoch": 1.14, "learning_rate": 2.321751649670066e-06, "loss": 0.0002, "step": 9472 }, { "epoch": 1.14, "learning_rate": 2.3216796640671863e-06, "loss": 0.0009, "step": 9473 }, { "epoch": 1.14, "learning_rate": 2.321607678464307e-06, "loss": 0.0003, "step": 9474 }, { "epoch": 1.14, "learning_rate": 2.3215356928614274e-06, "loss": 0.0002, "step": 9475 }, { "epoch": 1.14, "learning_rate": 2.3214637072585482e-06, "loss": 0.0002, "step": 9476 }, { "epoch": 1.14, "learning_rate": 2.321391721655669e-06, "loss": 0.0004, "step": 9477 }, { "epoch": 1.14, "learning_rate": 2.3213197360527898e-06, "loss": 1.0284, "step": 9478 }, { "epoch": 1.14, "learning_rate": 2.32124775044991e-06, "loss": 0.0003, "step": 9479 }, { "epoch": 1.14, "learning_rate": 2.321175764847031e-06, "loss": 0.1366, "step": 9480 }, { "epoch": 1.14, "learning_rate": 2.3211037792441513e-06, "loss": 0.0006, "step": 9481 }, { "epoch": 1.14, "learning_rate": 2.321031793641272e-06, "loss": 0.0004, "step": 9482 }, { "epoch": 1.14, "learning_rate": 2.3209598080383924e-06, "loss": 0.0004, "step": 9483 }, { "epoch": 1.14, "learning_rate": 2.320887822435513e-06, "loss": 0.0003, "step": 9484 }, { "epoch": 1.14, "learning_rate": 2.3208158368326335e-06, "loss": 0.0005, "step": 9485 }, { "epoch": 1.14, "learning_rate": 2.320743851229754e-06, "loss": 0.0003, "step": 9486 }, { "epoch": 1.14, "learning_rate": 2.3206718656268747e-06, "loss": 0.0002, "step": 9487 }, { "epoch": 1.14, "learning_rate": 2.320599880023995e-06, "loss": 0.0004, "step": 9488 }, { "epoch": 1.14, "learning_rate": 2.320527894421116e-06, "loss": 0.0003, "step": 9489 }, { "epoch": 1.14, "learning_rate": 2.320455908818236e-06, "loss": 0.0004, "step": 9490 }, { "epoch": 1.14, "learning_rate": 2.320383923215357e-06, "loss": 0.0003, "step": 9491 }, { "epoch": 1.14, "learning_rate": 2.3203119376124777e-06, "loss": 0.0004, "step": 9492 }, { "epoch": 1.14, "learning_rate": 2.320239952009598e-06, "loss": 0.0004, "step": 9493 }, { "epoch": 1.14, "learning_rate": 2.320167966406719e-06, "loss": 0.0039, "step": 9494 }, { "epoch": 1.14, "learning_rate": 2.3200959808038396e-06, "loss": 0.0007, "step": 9495 }, { "epoch": 1.14, "learning_rate": 2.32002399520096e-06, "loss": 0.0004, "step": 9496 }, { "epoch": 1.14, "learning_rate": 2.3199520095980803e-06, "loss": 0.0003, "step": 9497 }, { "epoch": 1.14, "learning_rate": 2.319880023995201e-06, "loss": 0.0014, "step": 9498 }, { "epoch": 1.14, "learning_rate": 2.3198080383923215e-06, "loss": 0.0006, "step": 9499 }, { "epoch": 1.14, "learning_rate": 2.3197360527894422e-06, "loss": 0.0003, "step": 9500 }, { "epoch": 1.14, "learning_rate": 2.3196640671865626e-06, "loss": 0.5082, "step": 9501 }, { "epoch": 1.14, "learning_rate": 2.3195920815836834e-06, "loss": 0.0015, "step": 9502 }, { "epoch": 1.14, "learning_rate": 2.3195200959808037e-06, "loss": 0.0096, "step": 9503 }, { "epoch": 1.14, "learning_rate": 2.3194481103779245e-06, "loss": 0.0007, "step": 9504 }, { "epoch": 1.14, "learning_rate": 2.319376124775045e-06, "loss": 0.0005, "step": 9505 }, { "epoch": 1.14, "learning_rate": 2.3193041391721656e-06, "loss": 0.0004, "step": 9506 }, { "epoch": 1.14, "learning_rate": 2.319232153569286e-06, "loss": 0.0003, "step": 9507 }, { "epoch": 1.14, "learning_rate": 2.3191601679664068e-06, "loss": 0.0002, "step": 9508 }, { "epoch": 1.14, "learning_rate": 2.3190881823635276e-06, "loss": 0.0004, "step": 9509 }, { "epoch": 1.14, "learning_rate": 2.319016196760648e-06, "loss": 0.0003, "step": 9510 }, { "epoch": 1.14, "learning_rate": 2.3189442111577687e-06, "loss": 0.0003, "step": 9511 }, { "epoch": 1.14, "learning_rate": 2.318872225554889e-06, "loss": 0.3208, "step": 9512 }, { "epoch": 1.14, "learning_rate": 2.31880023995201e-06, "loss": 0.0003, "step": 9513 }, { "epoch": 1.14, "learning_rate": 2.31872825434913e-06, "loss": 0.0006, "step": 9514 }, { "epoch": 1.14, "learning_rate": 2.318656268746251e-06, "loss": 0.0003, "step": 9515 }, { "epoch": 1.14, "learning_rate": 2.3185842831433713e-06, "loss": 0.1517, "step": 9516 }, { "epoch": 1.14, "learning_rate": 2.318512297540492e-06, "loss": 0.0003, "step": 9517 }, { "epoch": 1.14, "learning_rate": 2.3184403119376124e-06, "loss": 0.0004, "step": 9518 }, { "epoch": 1.14, "learning_rate": 2.318368326334733e-06, "loss": 0.0004, "step": 9519 }, { "epoch": 1.14, "learning_rate": 2.3182963407318536e-06, "loss": 0.0002, "step": 9520 }, { "epoch": 1.14, "learning_rate": 2.318224355128974e-06, "loss": 0.0005, "step": 9521 }, { "epoch": 1.14, "learning_rate": 2.3181523695260947e-06, "loss": 0.599, "step": 9522 }, { "epoch": 1.14, "learning_rate": 2.3180803839232155e-06, "loss": 0.0005, "step": 9523 }, { "epoch": 1.14, "learning_rate": 2.3180083983203363e-06, "loss": 0.0002, "step": 9524 }, { "epoch": 1.14, "learning_rate": 2.3179364127174566e-06, "loss": 0.0002, "step": 9525 }, { "epoch": 1.14, "learning_rate": 2.3178644271145774e-06, "loss": 0.0003, "step": 9526 }, { "epoch": 1.14, "learning_rate": 2.3177924415116978e-06, "loss": 0.0003, "step": 9527 }, { "epoch": 1.14, "learning_rate": 2.3177204559088185e-06, "loss": 0.0003, "step": 9528 }, { "epoch": 1.14, "learning_rate": 2.317648470305939e-06, "loss": 0.0003, "step": 9529 }, { "epoch": 1.14, "learning_rate": 2.3175764847030597e-06, "loss": 0.0002, "step": 9530 }, { "epoch": 1.14, "learning_rate": 2.31750449910018e-06, "loss": 0.1939, "step": 9531 }, { "epoch": 1.14, "learning_rate": 2.3174325134973004e-06, "loss": 0.0011, "step": 9532 }, { "epoch": 1.14, "learning_rate": 2.317360527894421e-06, "loss": 0.0034, "step": 9533 }, { "epoch": 1.14, "learning_rate": 2.3172885422915415e-06, "loss": 0.0003, "step": 9534 }, { "epoch": 1.14, "learning_rate": 2.3172165566886623e-06, "loss": 0.0002, "step": 9535 }, { "epoch": 1.14, "learning_rate": 2.3171445710857826e-06, "loss": 0.0004, "step": 9536 }, { "epoch": 1.14, "learning_rate": 2.3170725854829034e-06, "loss": 0.0004, "step": 9537 }, { "epoch": 1.14, "learning_rate": 2.317000599880024e-06, "loss": 0.1007, "step": 9538 }, { "epoch": 1.14, "learning_rate": 2.3169286142771446e-06, "loss": 0.0002, "step": 9539 }, { "epoch": 1.14, "learning_rate": 2.3168566286742653e-06, "loss": 0.0004, "step": 9540 }, { "epoch": 1.14, "learning_rate": 2.316784643071386e-06, "loss": 0.0002, "step": 9541 }, { "epoch": 1.14, "learning_rate": 2.3167126574685065e-06, "loss": 0.0002, "step": 9542 }, { "epoch": 1.14, "learning_rate": 2.316640671865627e-06, "loss": 0.0004, "step": 9543 }, { "epoch": 1.14, "learning_rate": 2.3165686862627476e-06, "loss": 0.0003, "step": 9544 }, { "epoch": 1.14, "learning_rate": 2.316496700659868e-06, "loss": 0.0004, "step": 9545 }, { "epoch": 1.14, "learning_rate": 2.3164247150569887e-06, "loss": 0.0003, "step": 9546 }, { "epoch": 1.14, "learning_rate": 2.316352729454109e-06, "loss": 0.0005, "step": 9547 }, { "epoch": 1.14, "learning_rate": 2.31628074385123e-06, "loss": 0.0017, "step": 9548 }, { "epoch": 1.14, "learning_rate": 2.3162087582483502e-06, "loss": 0.0003, "step": 9549 }, { "epoch": 1.14, "learning_rate": 2.316136772645471e-06, "loss": 0.0004, "step": 9550 }, { "epoch": 1.14, "learning_rate": 2.3160647870425914e-06, "loss": 0.0003, "step": 9551 }, { "epoch": 1.14, "learning_rate": 2.315992801439712e-06, "loss": 0.0003, "step": 9552 }, { "epoch": 1.14, "learning_rate": 2.3159208158368325e-06, "loss": 0.0003, "step": 9553 }, { "epoch": 1.14, "learning_rate": 2.3158488302339533e-06, "loss": 0.0005, "step": 9554 }, { "epoch": 1.14, "learning_rate": 2.315776844631074e-06, "loss": 0.0003, "step": 9555 }, { "epoch": 1.15, "learning_rate": 2.3157048590281944e-06, "loss": 0.0002, "step": 9556 }, { "epoch": 1.15, "learning_rate": 2.315632873425315e-06, "loss": 0.0027, "step": 9557 }, { "epoch": 1.15, "learning_rate": 2.3155608878224355e-06, "loss": 0.8596, "step": 9558 }, { "epoch": 1.15, "learning_rate": 2.3154889022195563e-06, "loss": 0.0003, "step": 9559 }, { "epoch": 1.15, "learning_rate": 2.3154169166166767e-06, "loss": 0.0004, "step": 9560 }, { "epoch": 1.15, "learning_rate": 2.3153449310137975e-06, "loss": 0.0004, "step": 9561 }, { "epoch": 1.15, "learning_rate": 2.315272945410918e-06, "loss": 0.0003, "step": 9562 }, { "epoch": 1.15, "learning_rate": 2.3152009598080386e-06, "loss": 0.8825, "step": 9563 }, { "epoch": 1.15, "learning_rate": 2.315128974205159e-06, "loss": 0.0003, "step": 9564 }, { "epoch": 1.15, "learning_rate": 2.3150569886022797e-06, "loss": 0.0014, "step": 9565 }, { "epoch": 1.15, "learning_rate": 2.3149850029994e-06, "loss": 0.0026, "step": 9566 }, { "epoch": 1.15, "learning_rate": 2.3149130173965204e-06, "loss": 0.0004, "step": 9567 }, { "epoch": 1.15, "learning_rate": 2.314841031793641e-06, "loss": 0.0003, "step": 9568 }, { "epoch": 1.15, "learning_rate": 2.314769046190762e-06, "loss": 0.2159, "step": 9569 }, { "epoch": 1.15, "learning_rate": 2.3146970605878823e-06, "loss": 0.0005, "step": 9570 }, { "epoch": 1.15, "learning_rate": 2.314625074985003e-06, "loss": 0.0003, "step": 9571 }, { "epoch": 1.15, "learning_rate": 2.314553089382124e-06, "loss": 0.0003, "step": 9572 }, { "epoch": 1.15, "learning_rate": 2.3144811037792443e-06, "loss": 0.0004, "step": 9573 }, { "epoch": 1.15, "learning_rate": 2.314409118176365e-06, "loss": 0.0004, "step": 9574 }, { "epoch": 1.15, "learning_rate": 2.3143371325734854e-06, "loss": 0.0011, "step": 9575 }, { "epoch": 1.15, "learning_rate": 2.314265146970606e-06, "loss": 0.0003, "step": 9576 }, { "epoch": 1.15, "learning_rate": 2.3141931613677265e-06, "loss": 0.0004, "step": 9577 }, { "epoch": 1.15, "learning_rate": 2.314121175764847e-06, "loss": 0.0003, "step": 9578 }, { "epoch": 1.15, "learning_rate": 2.3140491901619677e-06, "loss": 0.2901, "step": 9579 }, { "epoch": 1.15, "learning_rate": 2.313977204559088e-06, "loss": 0.0004, "step": 9580 }, { "epoch": 1.15, "learning_rate": 2.3139052189562088e-06, "loss": 0.0005, "step": 9581 }, { "epoch": 1.15, "learning_rate": 2.313833233353329e-06, "loss": 0.0004, "step": 9582 }, { "epoch": 1.15, "learning_rate": 2.31376124775045e-06, "loss": 0.0004, "step": 9583 }, { "epoch": 1.15, "learning_rate": 2.3136892621475703e-06, "loss": 0.0002, "step": 9584 }, { "epoch": 1.15, "learning_rate": 2.313617276544691e-06, "loss": 0.0005, "step": 9585 }, { "epoch": 1.15, "learning_rate": 2.313545290941812e-06, "loss": 0.0003, "step": 9586 }, { "epoch": 1.15, "learning_rate": 2.3134733053389326e-06, "loss": 0.0005, "step": 9587 }, { "epoch": 1.15, "learning_rate": 2.313401319736053e-06, "loss": 0.001, "step": 9588 }, { "epoch": 1.15, "learning_rate": 2.3133293341331733e-06, "loss": 0.0003, "step": 9589 }, { "epoch": 1.15, "learning_rate": 2.313257348530294e-06, "loss": 0.0003, "step": 9590 }, { "epoch": 1.15, "learning_rate": 2.3131853629274145e-06, "loss": 0.0005, "step": 9591 }, { "epoch": 1.15, "learning_rate": 2.3131133773245352e-06, "loss": 0.3883, "step": 9592 }, { "epoch": 1.15, "learning_rate": 2.3130413917216556e-06, "loss": 0.0004, "step": 9593 }, { "epoch": 1.15, "learning_rate": 2.3129694061187764e-06, "loss": 0.0003, "step": 9594 }, { "epoch": 1.15, "learning_rate": 2.3128974205158967e-06, "loss": 0.0004, "step": 9595 }, { "epoch": 1.15, "learning_rate": 2.3128254349130175e-06, "loss": 0.0003, "step": 9596 }, { "epoch": 1.15, "learning_rate": 2.312753449310138e-06, "loss": 0.0005, "step": 9597 }, { "epoch": 1.15, "learning_rate": 2.3126814637072586e-06, "loss": 0.0003, "step": 9598 }, { "epoch": 1.15, "learning_rate": 2.312609478104379e-06, "loss": 0.0002, "step": 9599 }, { "epoch": 1.15, "learning_rate": 2.3125374925014998e-06, "loss": 0.0004, "step": 9600 }, { "epoch": 1.15, "learning_rate": 2.3124655068986205e-06, "loss": 0.0004, "step": 9601 }, { "epoch": 1.15, "learning_rate": 2.312393521295741e-06, "loss": 0.0003, "step": 9602 }, { "epoch": 1.15, "learning_rate": 2.3123215356928617e-06, "loss": 0.0004, "step": 9603 }, { "epoch": 1.15, "learning_rate": 2.312249550089982e-06, "loss": 0.0003, "step": 9604 }, { "epoch": 1.15, "learning_rate": 2.312177564487103e-06, "loss": 0.8862, "step": 9605 }, { "epoch": 1.15, "learning_rate": 2.312105578884223e-06, "loss": 0.0003, "step": 9606 }, { "epoch": 1.15, "learning_rate": 2.312033593281344e-06, "loss": 0.0003, "step": 9607 }, { "epoch": 1.15, "learning_rate": 2.3119616076784643e-06, "loss": 0.0008, "step": 9608 }, { "epoch": 1.15, "learning_rate": 2.311889622075585e-06, "loss": 0.0003, "step": 9609 }, { "epoch": 1.15, "learning_rate": 2.3118176364727054e-06, "loss": 0.0003, "step": 9610 }, { "epoch": 1.15, "learning_rate": 2.3117456508698262e-06, "loss": 0.0003, "step": 9611 }, { "epoch": 1.15, "learning_rate": 2.3116736652669466e-06, "loss": 0.0004, "step": 9612 }, { "epoch": 1.15, "learning_rate": 2.311601679664067e-06, "loss": 0.0052, "step": 9613 }, { "epoch": 1.15, "learning_rate": 2.3115296940611877e-06, "loss": 0.0005, "step": 9614 }, { "epoch": 1.15, "learning_rate": 2.3114577084583085e-06, "loss": 0.0004, "step": 9615 }, { "epoch": 1.15, "learning_rate": 2.311385722855429e-06, "loss": 0.0003, "step": 9616 }, { "epoch": 1.15, "learning_rate": 2.3113137372525496e-06, "loss": 0.0038, "step": 9617 }, { "epoch": 1.15, "learning_rate": 2.3112417516496704e-06, "loss": 0.0005, "step": 9618 }, { "epoch": 1.15, "learning_rate": 2.3111697660467907e-06, "loss": 0.0014, "step": 9619 }, { "epoch": 1.15, "learning_rate": 2.3110977804439115e-06, "loss": 0.0005, "step": 9620 }, { "epoch": 1.15, "learning_rate": 2.311025794841032e-06, "loss": 0.0003, "step": 9621 }, { "epoch": 1.15, "learning_rate": 2.3109538092381527e-06, "loss": 0.0653, "step": 9622 }, { "epoch": 1.15, "learning_rate": 2.310881823635273e-06, "loss": 0.0057, "step": 9623 }, { "epoch": 1.15, "learning_rate": 2.3108098380323934e-06, "loss": 0.0003, "step": 9624 }, { "epoch": 1.15, "learning_rate": 2.310737852429514e-06, "loss": 0.5025, "step": 9625 }, { "epoch": 1.15, "learning_rate": 2.3106658668266345e-06, "loss": 0.0004, "step": 9626 }, { "epoch": 1.15, "learning_rate": 2.3105938812237553e-06, "loss": 0.0003, "step": 9627 }, { "epoch": 1.15, "learning_rate": 2.3105218956208756e-06, "loss": 0.0004, "step": 9628 }, { "epoch": 1.15, "learning_rate": 2.3104499100179964e-06, "loss": 0.0004, "step": 9629 }, { "epoch": 1.15, "learning_rate": 2.3103779244151168e-06, "loss": 0.0005, "step": 9630 }, { "epoch": 1.15, "learning_rate": 2.3103059388122376e-06, "loss": 0.0003, "step": 9631 }, { "epoch": 1.15, "learning_rate": 2.3102339532093583e-06, "loss": 0.0005, "step": 9632 }, { "epoch": 1.15, "learning_rate": 2.310161967606479e-06, "loss": 0.0005, "step": 9633 }, { "epoch": 1.15, "learning_rate": 2.3100899820035995e-06, "loss": 0.0003, "step": 9634 }, { "epoch": 1.15, "learning_rate": 2.31001799640072e-06, "loss": 0.0003, "step": 9635 }, { "epoch": 1.15, "learning_rate": 2.3099460107978406e-06, "loss": 0.0005, "step": 9636 }, { "epoch": 1.15, "learning_rate": 2.309874025194961e-06, "loss": 0.0004, "step": 9637 }, { "epoch": 1.15, "learning_rate": 2.3098020395920817e-06, "loss": 0.8597, "step": 9638 }, { "epoch": 1.16, "learning_rate": 2.309730053989202e-06, "loss": 0.0007, "step": 9639 }, { "epoch": 1.16, "learning_rate": 2.309658068386323e-06, "loss": 0.0067, "step": 9640 }, { "epoch": 1.16, "learning_rate": 2.3095860827834432e-06, "loss": 0.0003, "step": 9641 }, { "epoch": 1.16, "learning_rate": 2.309514097180564e-06, "loss": 0.0003, "step": 9642 }, { "epoch": 1.16, "learning_rate": 2.3094421115776844e-06, "loss": 0.0004, "step": 9643 }, { "epoch": 1.16, "learning_rate": 2.309370125974805e-06, "loss": 0.0005, "step": 9644 }, { "epoch": 1.16, "learning_rate": 2.3092981403719255e-06, "loss": 0.0002, "step": 9645 }, { "epoch": 1.16, "learning_rate": 2.3092261547690463e-06, "loss": 0.0002, "step": 9646 }, { "epoch": 1.16, "learning_rate": 2.309154169166167e-06, "loss": 0.0002, "step": 9647 }, { "epoch": 1.16, "learning_rate": 2.3090821835632874e-06, "loss": 0.0004, "step": 9648 }, { "epoch": 1.16, "learning_rate": 2.309010197960408e-06, "loss": 0.0005, "step": 9649 }, { "epoch": 1.16, "learning_rate": 2.3089382123575285e-06, "loss": 0.0003, "step": 9650 }, { "epoch": 1.16, "learning_rate": 2.3088662267546493e-06, "loss": 0.0004, "step": 9651 }, { "epoch": 1.16, "learning_rate": 2.3087942411517697e-06, "loss": 0.0003, "step": 9652 }, { "epoch": 1.16, "learning_rate": 2.3087222555488904e-06, "loss": 0.0006, "step": 9653 }, { "epoch": 1.16, "learning_rate": 2.308650269946011e-06, "loss": 0.0003, "step": 9654 }, { "epoch": 1.16, "learning_rate": 2.3085782843431316e-06, "loss": 0.037, "step": 9655 }, { "epoch": 1.16, "learning_rate": 2.308506298740252e-06, "loss": 0.0003, "step": 9656 }, { "epoch": 1.16, "learning_rate": 2.3084343131373727e-06, "loss": 0.0002, "step": 9657 }, { "epoch": 1.16, "learning_rate": 2.308362327534493e-06, "loss": 0.0002, "step": 9658 }, { "epoch": 1.16, "learning_rate": 2.3082903419316134e-06, "loss": 0.0003, "step": 9659 }, { "epoch": 1.16, "learning_rate": 2.308218356328734e-06, "loss": 0.0002, "step": 9660 }, { "epoch": 1.16, "learning_rate": 2.308146370725855e-06, "loss": 0.8246, "step": 9661 }, { "epoch": 1.16, "learning_rate": 2.3080743851229753e-06, "loss": 0.0003, "step": 9662 }, { "epoch": 1.16, "learning_rate": 2.308002399520096e-06, "loss": 0.0005, "step": 9663 }, { "epoch": 1.16, "learning_rate": 2.307930413917217e-06, "loss": 0.0003, "step": 9664 }, { "epoch": 1.16, "learning_rate": 2.3078584283143372e-06, "loss": 0.0002, "step": 9665 }, { "epoch": 1.16, "learning_rate": 2.307786442711458e-06, "loss": 0.0003, "step": 9666 }, { "epoch": 1.16, "learning_rate": 2.3077144571085784e-06, "loss": 0.0002, "step": 9667 }, { "epoch": 1.16, "learning_rate": 2.307642471505699e-06, "loss": 0.0002, "step": 9668 }, { "epoch": 1.16, "learning_rate": 2.3075704859028195e-06, "loss": 0.0003, "step": 9669 }, { "epoch": 1.16, "learning_rate": 2.30749850029994e-06, "loss": 0.0003, "step": 9670 }, { "epoch": 1.16, "learning_rate": 2.3074265146970606e-06, "loss": 0.0003, "step": 9671 }, { "epoch": 1.16, "learning_rate": 2.307354529094181e-06, "loss": 0.0002, "step": 9672 }, { "epoch": 1.16, "learning_rate": 2.3072825434913018e-06, "loss": 0.0003, "step": 9673 }, { "epoch": 1.16, "learning_rate": 2.307210557888422e-06, "loss": 0.924, "step": 9674 }, { "epoch": 1.16, "learning_rate": 2.307138572285543e-06, "loss": 0.0003, "step": 9675 }, { "epoch": 1.16, "learning_rate": 2.3070665866826633e-06, "loss": 0.0003, "step": 9676 }, { "epoch": 1.16, "learning_rate": 2.306994601079784e-06, "loss": 0.001, "step": 9677 }, { "epoch": 1.16, "learning_rate": 2.306922615476905e-06, "loss": 0.0006, "step": 9678 }, { "epoch": 1.16, "learning_rate": 2.3068506298740256e-06, "loss": 0.0003, "step": 9679 }, { "epoch": 1.16, "learning_rate": 2.306778644271146e-06, "loss": 0.0003, "step": 9680 }, { "epoch": 1.16, "learning_rate": 2.3067066586682663e-06, "loss": 0.0003, "step": 9681 }, { "epoch": 1.16, "learning_rate": 2.306634673065387e-06, "loss": 0.0003, "step": 9682 }, { "epoch": 1.16, "learning_rate": 2.3065626874625074e-06, "loss": 0.0003, "step": 9683 }, { "epoch": 1.16, "learning_rate": 2.3064907018596282e-06, "loss": 0.0003, "step": 9684 }, { "epoch": 1.16, "learning_rate": 2.3064187162567486e-06, "loss": 0.0003, "step": 9685 }, { "epoch": 1.16, "learning_rate": 2.3063467306538694e-06, "loss": 0.0005, "step": 9686 }, { "epoch": 1.16, "learning_rate": 2.3062747450509897e-06, "loss": 0.0005, "step": 9687 }, { "epoch": 1.16, "learning_rate": 2.3062027594481105e-06, "loss": 0.0005, "step": 9688 }, { "epoch": 1.16, "learning_rate": 2.306130773845231e-06, "loss": 0.0004, "step": 9689 }, { "epoch": 1.16, "learning_rate": 2.3060587882423516e-06, "loss": 0.0004, "step": 9690 }, { "epoch": 1.16, "learning_rate": 2.305986802639472e-06, "loss": 0.5659, "step": 9691 }, { "epoch": 1.16, "learning_rate": 2.3059148170365928e-06, "loss": 0.2366, "step": 9692 }, { "epoch": 1.16, "learning_rate": 2.3058428314337135e-06, "loss": 0.0005, "step": 9693 }, { "epoch": 1.16, "learning_rate": 2.305770845830834e-06, "loss": 0.0009, "step": 9694 }, { "epoch": 1.16, "learning_rate": 2.3056988602279547e-06, "loss": 0.7981, "step": 9695 }, { "epoch": 1.16, "learning_rate": 2.305626874625075e-06, "loss": 0.0007, "step": 9696 }, { "epoch": 1.16, "learning_rate": 2.305554889022196e-06, "loss": 0.0003, "step": 9697 }, { "epoch": 1.16, "learning_rate": 2.305482903419316e-06, "loss": 0.001, "step": 9698 }, { "epoch": 1.16, "learning_rate": 2.305410917816437e-06, "loss": 0.0003, "step": 9699 }, { "epoch": 1.16, "learning_rate": 2.3053389322135573e-06, "loss": 0.0003, "step": 9700 }, { "epoch": 1.16, "learning_rate": 2.305266946610678e-06, "loss": 0.0003, "step": 9701 }, { "epoch": 1.16, "learning_rate": 2.3051949610077984e-06, "loss": 0.0004, "step": 9702 }, { "epoch": 1.16, "learning_rate": 2.305122975404919e-06, "loss": 0.0003, "step": 9703 }, { "epoch": 1.16, "learning_rate": 2.3050509898020396e-06, "loss": 0.8538, "step": 9704 }, { "epoch": 1.16, "learning_rate": 2.30497900419916e-06, "loss": 0.1967, "step": 9705 }, { "epoch": 1.16, "learning_rate": 2.3049070185962807e-06, "loss": 0.0004, "step": 9706 }, { "epoch": 1.16, "learning_rate": 2.304835032993401e-06, "loss": 0.0005, "step": 9707 }, { "epoch": 1.16, "learning_rate": 2.304763047390522e-06, "loss": 0.0004, "step": 9708 }, { "epoch": 1.16, "learning_rate": 2.3046910617876426e-06, "loss": 0.2755, "step": 9709 }, { "epoch": 1.16, "learning_rate": 2.3046190761847634e-06, "loss": 0.0002, "step": 9710 }, { "epoch": 1.16, "learning_rate": 2.3045470905818837e-06, "loss": 0.3377, "step": 9711 }, { "epoch": 1.16, "learning_rate": 2.3044751049790045e-06, "loss": 0.0004, "step": 9712 }, { "epoch": 1.16, "learning_rate": 2.304403119376125e-06, "loss": 0.0026, "step": 9713 }, { "epoch": 1.16, "learning_rate": 2.3043311337732457e-06, "loss": 0.0004, "step": 9714 }, { "epoch": 1.16, "learning_rate": 2.304259148170366e-06, "loss": 0.0003, "step": 9715 }, { "epoch": 1.16, "learning_rate": 2.3041871625674864e-06, "loss": 0.0002, "step": 9716 }, { "epoch": 1.16, "learning_rate": 2.304115176964607e-06, "loss": 0.2183, "step": 9717 }, { "epoch": 1.16, "learning_rate": 2.3040431913617275e-06, "loss": 0.0004, "step": 9718 }, { "epoch": 1.16, "learning_rate": 2.3039712057588483e-06, "loss": 0.0003, "step": 9719 }, { "epoch": 1.16, "learning_rate": 2.3038992201559686e-06, "loss": 0.0003, "step": 9720 }, { "epoch": 1.16, "learning_rate": 2.3038272345530894e-06, "loss": 0.0003, "step": 9721 }, { "epoch": 1.17, "learning_rate": 2.3037552489502098e-06, "loss": 0.0006, "step": 9722 }, { "epoch": 1.17, "learning_rate": 2.3036832633473305e-06, "loss": 0.0003, "step": 9723 }, { "epoch": 1.17, "learning_rate": 2.3036112777444513e-06, "loss": 0.0005, "step": 9724 }, { "epoch": 1.17, "learning_rate": 2.3035392921415717e-06, "loss": 0.0006, "step": 9725 }, { "epoch": 1.17, "learning_rate": 2.3034673065386925e-06, "loss": 0.0004, "step": 9726 }, { "epoch": 1.17, "learning_rate": 2.303395320935813e-06, "loss": 0.0002, "step": 9727 }, { "epoch": 1.17, "learning_rate": 2.3033233353329336e-06, "loss": 0.0003, "step": 9728 }, { "epoch": 1.17, "learning_rate": 2.303251349730054e-06, "loss": 0.0008, "step": 9729 }, { "epoch": 1.17, "learning_rate": 2.3031793641271747e-06, "loss": 0.0008, "step": 9730 }, { "epoch": 1.17, "learning_rate": 2.303107378524295e-06, "loss": 0.0002, "step": 9731 }, { "epoch": 1.17, "learning_rate": 2.303035392921416e-06, "loss": 0.0004, "step": 9732 }, { "epoch": 1.17, "learning_rate": 2.302963407318536e-06, "loss": 0.0007, "step": 9733 }, { "epoch": 1.17, "learning_rate": 2.302891421715657e-06, "loss": 0.0006, "step": 9734 }, { "epoch": 1.17, "learning_rate": 2.3028194361127773e-06, "loss": 0.0005, "step": 9735 }, { "epoch": 1.17, "learning_rate": 2.302747450509898e-06, "loss": 0.0002, "step": 9736 }, { "epoch": 1.17, "learning_rate": 2.3026754649070185e-06, "loss": 0.0006, "step": 9737 }, { "epoch": 1.17, "learning_rate": 2.3026034793041393e-06, "loss": 0.0008, "step": 9738 }, { "epoch": 1.17, "learning_rate": 2.3025314937012596e-06, "loss": 0.0004, "step": 9739 }, { "epoch": 1.17, "learning_rate": 2.3024595080983804e-06, "loss": 0.7181, "step": 9740 }, { "epoch": 1.17, "learning_rate": 2.302387522495501e-06, "loss": 0.0004, "step": 9741 }, { "epoch": 1.17, "learning_rate": 2.3023155368926215e-06, "loss": 0.72, "step": 9742 }, { "epoch": 1.17, "learning_rate": 2.3022435512897423e-06, "loss": 0.0004, "step": 9743 }, { "epoch": 1.17, "learning_rate": 2.3021715656868627e-06, "loss": 0.0005, "step": 9744 }, { "epoch": 1.17, "learning_rate": 2.3020995800839834e-06, "loss": 0.0004, "step": 9745 }, { "epoch": 1.17, "learning_rate": 2.302027594481104e-06, "loss": 0.0004, "step": 9746 }, { "epoch": 1.17, "learning_rate": 2.3019556088782246e-06, "loss": 0.0002, "step": 9747 }, { "epoch": 1.17, "learning_rate": 2.301883623275345e-06, "loss": 0.0004, "step": 9748 }, { "epoch": 1.17, "learning_rate": 2.3018116376724657e-06, "loss": 0.0006, "step": 9749 }, { "epoch": 1.17, "learning_rate": 2.301739652069586e-06, "loss": 0.0004, "step": 9750 }, { "epoch": 1.17, "learning_rate": 2.3016676664667064e-06, "loss": 0.0006, "step": 9751 }, { "epoch": 1.17, "learning_rate": 2.301595680863827e-06, "loss": 0.0005, "step": 9752 }, { "epoch": 1.17, "learning_rate": 2.3015236952609475e-06, "loss": 0.0032, "step": 9753 }, { "epoch": 1.17, "learning_rate": 2.3014517096580683e-06, "loss": 0.0002, "step": 9754 }, { "epoch": 1.17, "learning_rate": 2.301379724055189e-06, "loss": 0.0013, "step": 9755 }, { "epoch": 1.17, "learning_rate": 2.30130773845231e-06, "loss": 0.0003, "step": 9756 }, { "epoch": 1.17, "learning_rate": 2.3012357528494302e-06, "loss": 0.0003, "step": 9757 }, { "epoch": 1.17, "learning_rate": 2.301163767246551e-06, "loss": 0.0003, "step": 9758 }, { "epoch": 1.17, "learning_rate": 2.3010917816436714e-06, "loss": 0.0005, "step": 9759 }, { "epoch": 1.17, "learning_rate": 2.301019796040792e-06, "loss": 0.0006, "step": 9760 }, { "epoch": 1.17, "learning_rate": 2.3009478104379125e-06, "loss": 0.6677, "step": 9761 }, { "epoch": 1.17, "learning_rate": 2.300875824835033e-06, "loss": 0.0002, "step": 9762 }, { "epoch": 1.17, "learning_rate": 2.3008038392321536e-06, "loss": 0.0006, "step": 9763 }, { "epoch": 1.17, "learning_rate": 2.300731853629274e-06, "loss": 0.0004, "step": 9764 }, { "epoch": 1.17, "learning_rate": 2.3006598680263948e-06, "loss": 0.0003, "step": 9765 }, { "epoch": 1.17, "learning_rate": 2.300587882423515e-06, "loss": 0.0004, "step": 9766 }, { "epoch": 1.17, "learning_rate": 2.300515896820636e-06, "loss": 0.0002, "step": 9767 }, { "epoch": 1.17, "learning_rate": 2.3004439112177563e-06, "loss": 0.0006, "step": 9768 }, { "epoch": 1.17, "learning_rate": 2.300371925614877e-06, "loss": 0.0017, "step": 9769 }, { "epoch": 1.17, "learning_rate": 2.300299940011998e-06, "loss": 0.0005, "step": 9770 }, { "epoch": 1.17, "learning_rate": 2.300227954409118e-06, "loss": 0.0003, "step": 9771 }, { "epoch": 1.17, "learning_rate": 2.300155968806239e-06, "loss": 0.0003, "step": 9772 }, { "epoch": 1.17, "learning_rate": 2.3000839832033597e-06, "loss": 0.0005, "step": 9773 }, { "epoch": 1.17, "learning_rate": 2.30001199760048e-06, "loss": 0.0004, "step": 9774 }, { "epoch": 1.17, "learning_rate": 2.2999400119976004e-06, "loss": 0.0003, "step": 9775 }, { "epoch": 1.17, "learning_rate": 2.2998680263947212e-06, "loss": 0.0003, "step": 9776 }, { "epoch": 1.17, "learning_rate": 2.2997960407918416e-06, "loss": 0.0004, "step": 9777 }, { "epoch": 1.17, "learning_rate": 2.2997240551889624e-06, "loss": 0.0072, "step": 9778 }, { "epoch": 1.17, "learning_rate": 2.2996520695860827e-06, "loss": 0.0003, "step": 9779 }, { "epoch": 1.17, "learning_rate": 2.2995800839832035e-06, "loss": 0.0002, "step": 9780 }, { "epoch": 1.17, "learning_rate": 2.299508098380324e-06, "loss": 0.0004, "step": 9781 }, { "epoch": 1.17, "learning_rate": 2.2994361127774446e-06, "loss": 0.0004, "step": 9782 }, { "epoch": 1.17, "learning_rate": 2.299364127174565e-06, "loss": 0.0006, "step": 9783 }, { "epoch": 1.17, "learning_rate": 2.2992921415716858e-06, "loss": 0.0004, "step": 9784 }, { "epoch": 1.17, "learning_rate": 2.299220155968806e-06, "loss": 0.0008, "step": 9785 }, { "epoch": 1.17, "learning_rate": 2.299148170365927e-06, "loss": 0.0005, "step": 9786 }, { "epoch": 1.17, "learning_rate": 2.2990761847630477e-06, "loss": 0.0004, "step": 9787 }, { "epoch": 1.17, "learning_rate": 2.299004199160168e-06, "loss": 0.0003, "step": 9788 }, { "epoch": 1.17, "learning_rate": 2.298932213557289e-06, "loss": 0.0004, "step": 9789 }, { "epoch": 1.17, "learning_rate": 2.298860227954409e-06, "loss": 0.0054, "step": 9790 }, { "epoch": 1.17, "learning_rate": 2.29878824235153e-06, "loss": 0.0003, "step": 9791 }, { "epoch": 1.17, "learning_rate": 2.2987162567486503e-06, "loss": 0.0004, "step": 9792 }, { "epoch": 1.17, "learning_rate": 2.298644271145771e-06, "loss": 0.0011, "step": 9793 }, { "epoch": 1.17, "learning_rate": 2.2985722855428914e-06, "loss": 0.0002, "step": 9794 }, { "epoch": 1.17, "learning_rate": 2.298500299940012e-06, "loss": 0.0005, "step": 9795 }, { "epoch": 1.17, "learning_rate": 2.2984283143371326e-06, "loss": 0.0005, "step": 9796 }, { "epoch": 1.17, "learning_rate": 2.298356328734253e-06, "loss": 0.0003, "step": 9797 }, { "epoch": 1.17, "learning_rate": 2.2982843431313737e-06, "loss": 0.0003, "step": 9798 }, { "epoch": 1.17, "learning_rate": 2.298212357528494e-06, "loss": 0.7626, "step": 9799 }, { "epoch": 1.17, "learning_rate": 2.298140371925615e-06, "loss": 0.0004, "step": 9800 }, { "epoch": 1.17, "learning_rate": 2.2980683863227356e-06, "loss": 0.0003, "step": 9801 }, { "epoch": 1.17, "learning_rate": 2.2979964007198564e-06, "loss": 0.0003, "step": 9802 }, { "epoch": 1.17, "learning_rate": 2.2979244151169767e-06, "loss": 0.0004, "step": 9803 }, { "epoch": 1.17, "learning_rate": 2.2978524295140975e-06, "loss": 0.0003, "step": 9804 }, { "epoch": 1.17, "learning_rate": 2.297780443911218e-06, "loss": 0.0002, "step": 9805 }, { "epoch": 1.18, "learning_rate": 2.2977084583083386e-06, "loss": 0.0002, "step": 9806 }, { "epoch": 1.18, "learning_rate": 2.297636472705459e-06, "loss": 0.0004, "step": 9807 }, { "epoch": 1.18, "learning_rate": 2.2975644871025794e-06, "loss": 0.0542, "step": 9808 }, { "epoch": 1.18, "learning_rate": 2.2974925014997e-06, "loss": 0.1692, "step": 9809 }, { "epoch": 1.18, "learning_rate": 2.2974205158968205e-06, "loss": 0.0005, "step": 9810 }, { "epoch": 1.18, "learning_rate": 2.2973485302939413e-06, "loss": 0.0006, "step": 9811 }, { "epoch": 1.18, "learning_rate": 2.2972765446910616e-06, "loss": 0.0004, "step": 9812 }, { "epoch": 1.18, "learning_rate": 2.2972045590881824e-06, "loss": 0.0002, "step": 9813 }, { "epoch": 1.18, "learning_rate": 2.2971325734853028e-06, "loss": 0.5642, "step": 9814 }, { "epoch": 1.18, "learning_rate": 2.2970605878824235e-06, "loss": 0.0003, "step": 9815 }, { "epoch": 1.18, "learning_rate": 2.2969886022795443e-06, "loss": 0.0004, "step": 9816 }, { "epoch": 1.18, "learning_rate": 2.2969166166766647e-06, "loss": 0.9794, "step": 9817 }, { "epoch": 1.18, "learning_rate": 2.2968446310737854e-06, "loss": 0.0008, "step": 9818 }, { "epoch": 1.18, "learning_rate": 2.2967726454709062e-06, "loss": 0.0113, "step": 9819 }, { "epoch": 1.18, "learning_rate": 2.2967006598680266e-06, "loss": 0.0009, "step": 9820 }, { "epoch": 1.18, "learning_rate": 2.296628674265147e-06, "loss": 0.0004, "step": 9821 }, { "epoch": 1.18, "learning_rate": 2.2965566886622677e-06, "loss": 0.0003, "step": 9822 }, { "epoch": 1.18, "learning_rate": 2.296484703059388e-06, "loss": 0.0003, "step": 9823 }, { "epoch": 1.18, "learning_rate": 2.296412717456509e-06, "loss": 0.0003, "step": 9824 }, { "epoch": 1.18, "learning_rate": 2.296340731853629e-06, "loss": 0.0002, "step": 9825 }, { "epoch": 1.18, "learning_rate": 2.29626874625075e-06, "loss": 0.0005, "step": 9826 }, { "epoch": 1.18, "learning_rate": 2.2961967606478703e-06, "loss": 0.0006, "step": 9827 }, { "epoch": 1.18, "learning_rate": 2.296124775044991e-06, "loss": 0.0004, "step": 9828 }, { "epoch": 1.18, "learning_rate": 2.2960527894421115e-06, "loss": 0.0018, "step": 9829 }, { "epoch": 1.18, "learning_rate": 2.2959808038392322e-06, "loss": 0.0003, "step": 9830 }, { "epoch": 1.18, "learning_rate": 2.2959088182363526e-06, "loss": 0.0004, "step": 9831 }, { "epoch": 1.18, "learning_rate": 2.2958368326334734e-06, "loss": 0.7923, "step": 9832 }, { "epoch": 1.18, "learning_rate": 2.295764847030594e-06, "loss": 0.0011, "step": 9833 }, { "epoch": 1.18, "learning_rate": 2.2956928614277145e-06, "loss": 0.0005, "step": 9834 }, { "epoch": 1.18, "learning_rate": 2.2956208758248353e-06, "loss": 0.0002, "step": 9835 }, { "epoch": 1.18, "learning_rate": 2.2955488902219556e-06, "loss": 0.0004, "step": 9836 }, { "epoch": 1.18, "learning_rate": 2.2954769046190764e-06, "loss": 0.0004, "step": 9837 }, { "epoch": 1.18, "learning_rate": 2.2954049190161968e-06, "loss": 0.0003, "step": 9838 }, { "epoch": 1.18, "learning_rate": 2.2953329334133176e-06, "loss": 0.0006, "step": 9839 }, { "epoch": 1.18, "learning_rate": 2.295260947810438e-06, "loss": 0.4081, "step": 9840 }, { "epoch": 1.18, "learning_rate": 2.2951889622075587e-06, "loss": 0.0003, "step": 9841 }, { "epoch": 1.18, "learning_rate": 2.295116976604679e-06, "loss": 0.0004, "step": 9842 }, { "epoch": 1.18, "learning_rate": 2.2950449910017994e-06, "loss": 0.0005, "step": 9843 }, { "epoch": 1.18, "learning_rate": 2.29497300539892e-06, "loss": 0.0002, "step": 9844 }, { "epoch": 1.18, "learning_rate": 2.2949010197960405e-06, "loss": 0.0005, "step": 9845 }, { "epoch": 1.18, "learning_rate": 2.2948290341931613e-06, "loss": 0.0005, "step": 9846 }, { "epoch": 1.18, "learning_rate": 2.294757048590282e-06, "loss": 0.0005, "step": 9847 }, { "epoch": 1.18, "learning_rate": 2.2946850629874024e-06, "loss": 0.0006, "step": 9848 }, { "epoch": 1.18, "learning_rate": 2.2946130773845232e-06, "loss": 0.0002, "step": 9849 }, { "epoch": 1.18, "learning_rate": 2.294541091781644e-06, "loss": 0.0003, "step": 9850 }, { "epoch": 1.18, "learning_rate": 2.2944691061787644e-06, "loss": 0.0013, "step": 9851 }, { "epoch": 1.18, "learning_rate": 2.294397120575885e-06, "loss": 0.0006, "step": 9852 }, { "epoch": 1.18, "learning_rate": 2.2943251349730055e-06, "loss": 0.0003, "step": 9853 }, { "epoch": 1.18, "learning_rate": 2.294253149370126e-06, "loss": 0.001, "step": 9854 }, { "epoch": 1.18, "learning_rate": 2.2941811637672466e-06, "loss": 0.0004, "step": 9855 }, { "epoch": 1.18, "learning_rate": 2.294109178164367e-06, "loss": 0.0006, "step": 9856 }, { "epoch": 1.18, "learning_rate": 2.2940371925614878e-06, "loss": 1.0207, "step": 9857 }, { "epoch": 1.18, "learning_rate": 2.293965206958608e-06, "loss": 0.0004, "step": 9858 }, { "epoch": 1.18, "learning_rate": 2.293893221355729e-06, "loss": 0.0002, "step": 9859 }, { "epoch": 1.18, "learning_rate": 2.2938212357528493e-06, "loss": 0.0007, "step": 9860 }, { "epoch": 1.18, "learning_rate": 2.29374925014997e-06, "loss": 0.0004, "step": 9861 }, { "epoch": 1.18, "learning_rate": 2.2936772645470904e-06, "loss": 0.0003, "step": 9862 }, { "epoch": 1.18, "learning_rate": 2.293605278944211e-06, "loss": 0.0004, "step": 9863 }, { "epoch": 1.18, "learning_rate": 2.293533293341332e-06, "loss": 0.0003, "step": 9864 }, { "epoch": 1.18, "learning_rate": 2.2934613077384527e-06, "loss": 0.0002, "step": 9865 }, { "epoch": 1.18, "learning_rate": 2.293389322135573e-06, "loss": 0.0004, "step": 9866 }, { "epoch": 1.18, "learning_rate": 2.2933173365326934e-06, "loss": 0.0005, "step": 9867 }, { "epoch": 1.18, "learning_rate": 2.293245350929814e-06, "loss": 0.0006, "step": 9868 }, { "epoch": 1.18, "learning_rate": 2.2931733653269346e-06, "loss": 0.0006, "step": 9869 }, { "epoch": 1.18, "learning_rate": 2.2931013797240553e-06, "loss": 0.7445, "step": 9870 }, { "epoch": 1.18, "learning_rate": 2.2930293941211757e-06, "loss": 0.0003, "step": 9871 }, { "epoch": 1.18, "learning_rate": 2.2929574085182965e-06, "loss": 0.0004, "step": 9872 }, { "epoch": 1.18, "learning_rate": 2.292885422915417e-06, "loss": 0.0003, "step": 9873 }, { "epoch": 1.18, "learning_rate": 2.2928134373125376e-06, "loss": 0.0008, "step": 9874 }, { "epoch": 1.18, "learning_rate": 2.292741451709658e-06, "loss": 0.0002, "step": 9875 }, { "epoch": 1.18, "learning_rate": 2.2926694661067787e-06, "loss": 0.5574, "step": 9876 }, { "epoch": 1.18, "learning_rate": 2.292597480503899e-06, "loss": 0.8007, "step": 9877 }, { "epoch": 1.18, "learning_rate": 2.29252549490102e-06, "loss": 0.0003, "step": 9878 }, { "epoch": 1.18, "learning_rate": 2.2924535092981407e-06, "loss": 0.2967, "step": 9879 }, { "epoch": 1.18, "learning_rate": 2.292381523695261e-06, "loss": 0.0004, "step": 9880 }, { "epoch": 1.18, "learning_rate": 2.292309538092382e-06, "loss": 0.0007, "step": 9881 }, { "epoch": 1.18, "learning_rate": 2.292237552489502e-06, "loss": 0.0007, "step": 9882 }, { "epoch": 1.18, "learning_rate": 2.292165566886623e-06, "loss": 0.0004, "step": 9883 }, { "epoch": 1.18, "learning_rate": 2.2920935812837433e-06, "loss": 0.0172, "step": 9884 }, { "epoch": 1.18, "learning_rate": 2.292021595680864e-06, "loss": 0.0004, "step": 9885 }, { "epoch": 1.18, "learning_rate": 2.2919496100779844e-06, "loss": 0.0005, "step": 9886 }, { "epoch": 1.18, "learning_rate": 2.291877624475105e-06, "loss": 0.0003, "step": 9887 }, { "epoch": 1.18, "learning_rate": 2.2918056388722255e-06, "loss": 0.0002, "step": 9888 }, { "epoch": 1.19, "learning_rate": 2.291733653269346e-06, "loss": 0.0003, "step": 9889 }, { "epoch": 1.19, "learning_rate": 2.2916616676664667e-06, "loss": 0.0005, "step": 9890 }, { "epoch": 1.19, "learning_rate": 2.291589682063587e-06, "loss": 0.0003, "step": 9891 }, { "epoch": 1.19, "learning_rate": 2.291517696460708e-06, "loss": 0.0002, "step": 9892 }, { "epoch": 1.19, "learning_rate": 2.2914457108578286e-06, "loss": 0.0003, "step": 9893 }, { "epoch": 1.19, "learning_rate": 2.291373725254949e-06, "loss": 0.0003, "step": 9894 }, { "epoch": 1.19, "learning_rate": 2.2913017396520697e-06, "loss": 0.0007, "step": 9895 }, { "epoch": 1.19, "learning_rate": 2.2912297540491905e-06, "loss": 0.0003, "step": 9896 }, { "epoch": 1.19, "learning_rate": 2.291157768446311e-06, "loss": 0.0002, "step": 9897 }, { "epoch": 1.19, "learning_rate": 2.2910857828434316e-06, "loss": 0.0005, "step": 9898 }, { "epoch": 1.19, "learning_rate": 2.291013797240552e-06, "loss": 0.0003, "step": 9899 }, { "epoch": 1.19, "learning_rate": 2.2909418116376728e-06, "loss": 0.0004, "step": 9900 }, { "epoch": 1.19, "learning_rate": 2.290869826034793e-06, "loss": 0.0007, "step": 9901 }, { "epoch": 1.19, "learning_rate": 2.2907978404319135e-06, "loss": 0.0011, "step": 9902 }, { "epoch": 1.19, "learning_rate": 2.2907258548290343e-06, "loss": 0.0004, "step": 9903 }, { "epoch": 1.19, "learning_rate": 2.2906538692261546e-06, "loss": 0.0005, "step": 9904 }, { "epoch": 1.19, "learning_rate": 2.2905818836232754e-06, "loss": 0.0004, "step": 9905 }, { "epoch": 1.19, "learning_rate": 2.2905098980203957e-06, "loss": 0.0004, "step": 9906 }, { "epoch": 1.19, "learning_rate": 2.2904379124175165e-06, "loss": 0.2888, "step": 9907 }, { "epoch": 1.19, "learning_rate": 2.290365926814637e-06, "loss": 0.0018, "step": 9908 }, { "epoch": 1.19, "learning_rate": 2.2902939412117577e-06, "loss": 0.0003, "step": 9909 }, { "epoch": 1.19, "learning_rate": 2.2902219556088784e-06, "loss": 0.0003, "step": 9910 }, { "epoch": 1.19, "learning_rate": 2.2901499700059992e-06, "loss": 0.0002, "step": 9911 }, { "epoch": 1.19, "learning_rate": 2.2900779844031196e-06, "loss": 0.0003, "step": 9912 }, { "epoch": 1.19, "learning_rate": 2.29000599880024e-06, "loss": 0.0003, "step": 9913 }, { "epoch": 1.19, "learning_rate": 2.2899340131973607e-06, "loss": 0.0003, "step": 9914 }, { "epoch": 1.19, "learning_rate": 2.289862027594481e-06, "loss": 0.0004, "step": 9915 }, { "epoch": 1.19, "learning_rate": 2.289790041991602e-06, "loss": 0.0004, "step": 9916 }, { "epoch": 1.19, "learning_rate": 2.289718056388722e-06, "loss": 0.0006, "step": 9917 }, { "epoch": 1.19, "learning_rate": 2.289646070785843e-06, "loss": 0.0003, "step": 9918 }, { "epoch": 1.19, "learning_rate": 2.2895740851829633e-06, "loss": 0.0002, "step": 9919 }, { "epoch": 1.19, "learning_rate": 2.289502099580084e-06, "loss": 0.0012, "step": 9920 }, { "epoch": 1.19, "learning_rate": 2.2894301139772045e-06, "loss": 0.0002, "step": 9921 }, { "epoch": 1.19, "learning_rate": 2.2893581283743252e-06, "loss": 0.0013, "step": 9922 }, { "epoch": 1.19, "learning_rate": 2.2892861427714456e-06, "loss": 1.0431, "step": 9923 }, { "epoch": 1.19, "learning_rate": 2.2892141571685664e-06, "loss": 0.0002, "step": 9924 }, { "epoch": 1.19, "learning_rate": 2.289142171565687e-06, "loss": 0.7599, "step": 9925 }, { "epoch": 1.19, "learning_rate": 2.2890701859628075e-06, "loss": 0.0003, "step": 9926 }, { "epoch": 1.19, "learning_rate": 2.2889982003599283e-06, "loss": 0.0003, "step": 9927 }, { "epoch": 1.19, "learning_rate": 2.2889262147570486e-06, "loss": 0.0005, "step": 9928 }, { "epoch": 1.19, "learning_rate": 2.2888542291541694e-06, "loss": 0.9568, "step": 9929 }, { "epoch": 1.19, "learning_rate": 2.2887822435512898e-06, "loss": 1.0724, "step": 9930 }, { "epoch": 1.19, "learning_rate": 2.2887102579484106e-06, "loss": 0.0006, "step": 9931 }, { "epoch": 1.19, "learning_rate": 2.288638272345531e-06, "loss": 0.7628, "step": 9932 }, { "epoch": 1.19, "learning_rate": 2.2885662867426517e-06, "loss": 0.0003, "step": 9933 }, { "epoch": 1.19, "learning_rate": 2.288494301139772e-06, "loss": 0.0019, "step": 9934 }, { "epoch": 1.19, "learning_rate": 2.2884223155368924e-06, "loss": 0.0005, "step": 9935 }, { "epoch": 1.19, "learning_rate": 2.288350329934013e-06, "loss": 0.0005, "step": 9936 }, { "epoch": 1.19, "learning_rate": 2.2882783443311335e-06, "loss": 0.0008, "step": 9937 }, { "epoch": 1.19, "learning_rate": 2.2882063587282543e-06, "loss": 0.003, "step": 9938 }, { "epoch": 1.19, "learning_rate": 2.288134373125375e-06, "loss": 0.0299, "step": 9939 }, { "epoch": 1.19, "learning_rate": 2.2880623875224954e-06, "loss": 0.0006, "step": 9940 }, { "epoch": 1.19, "learning_rate": 2.2879904019196162e-06, "loss": 0.0006, "step": 9941 }, { "epoch": 1.19, "learning_rate": 2.287918416316737e-06, "loss": 0.0004, "step": 9942 }, { "epoch": 1.19, "learning_rate": 2.2878464307138574e-06, "loss": 0.0005, "step": 9943 }, { "epoch": 1.19, "learning_rate": 2.287774445110978e-06, "loss": 0.0003, "step": 9944 }, { "epoch": 1.19, "learning_rate": 2.2877024595080985e-06, "loss": 0.004, "step": 9945 }, { "epoch": 1.19, "learning_rate": 2.2876304739052193e-06, "loss": 0.0003, "step": 9946 }, { "epoch": 1.19, "learning_rate": 2.2875584883023396e-06, "loss": 0.0003, "step": 9947 }, { "epoch": 1.19, "learning_rate": 2.28748650269946e-06, "loss": 0.0008, "step": 9948 }, { "epoch": 1.19, "learning_rate": 2.2874145170965808e-06, "loss": 0.001, "step": 9949 }, { "epoch": 1.19, "learning_rate": 2.287342531493701e-06, "loss": 0.0097, "step": 9950 }, { "epoch": 1.19, "learning_rate": 2.287270545890822e-06, "loss": 0.0004, "step": 9951 }, { "epoch": 1.19, "learning_rate": 2.2871985602879422e-06, "loss": 0.7388, "step": 9952 }, { "epoch": 1.19, "learning_rate": 2.287126574685063e-06, "loss": 0.0736, "step": 9953 }, { "epoch": 1.19, "learning_rate": 2.2870545890821834e-06, "loss": 0.0022, "step": 9954 }, { "epoch": 1.19, "learning_rate": 2.286982603479304e-06, "loss": 0.0036, "step": 9955 }, { "epoch": 1.19, "learning_rate": 2.286910617876425e-06, "loss": 0.0008, "step": 9956 }, { "epoch": 1.19, "learning_rate": 2.2868386322735457e-06, "loss": 0.0009, "step": 9957 }, { "epoch": 1.19, "learning_rate": 2.286766646670666e-06, "loss": 0.6463, "step": 9958 }, { "epoch": 1.19, "learning_rate": 2.2866946610677864e-06, "loss": 0.0005, "step": 9959 }, { "epoch": 1.19, "learning_rate": 2.286622675464907e-06, "loss": 0.4146, "step": 9960 }, { "epoch": 1.19, "learning_rate": 2.2865506898620276e-06, "loss": 0.0004, "step": 9961 }, { "epoch": 1.19, "learning_rate": 2.2864787042591483e-06, "loss": 0.0003, "step": 9962 }, { "epoch": 1.19, "learning_rate": 2.2864067186562687e-06, "loss": 0.0004, "step": 9963 }, { "epoch": 1.19, "learning_rate": 2.2863347330533895e-06, "loss": 0.0006, "step": 9964 }, { "epoch": 1.19, "learning_rate": 2.28626274745051e-06, "loss": 0.0004, "step": 9965 }, { "epoch": 1.19, "learning_rate": 2.2861907618476306e-06, "loss": 0.0004, "step": 9966 }, { "epoch": 1.19, "learning_rate": 2.286118776244751e-06, "loss": 0.0006, "step": 9967 }, { "epoch": 1.19, "learning_rate": 2.2860467906418717e-06, "loss": 0.0008, "step": 9968 }, { "epoch": 1.19, "learning_rate": 2.285974805038992e-06, "loss": 0.0003, "step": 9969 }, { "epoch": 1.19, "learning_rate": 2.285902819436113e-06, "loss": 0.0004, "step": 9970 }, { "epoch": 1.19, "learning_rate": 2.2858308338332332e-06, "loss": 0.1691, "step": 9971 }, { "epoch": 1.19, "learning_rate": 2.285758848230354e-06, "loss": 0.0004, "step": 9972 }, { "epoch": 1.2, "learning_rate": 2.2856868626274748e-06, "loss": 0.0009, "step": 9973 }, { "epoch": 1.2, "learning_rate": 2.285614877024595e-06, "loss": 0.0032, "step": 9974 }, { "epoch": 1.2, "learning_rate": 2.285542891421716e-06, "loss": 0.0008, "step": 9975 }, { "epoch": 1.2, "learning_rate": 2.2854709058188363e-06, "loss": 0.0004, "step": 9976 }, { "epoch": 1.2, "learning_rate": 2.285398920215957e-06, "loss": 0.0007, "step": 9977 }, { "epoch": 1.2, "learning_rate": 2.2853269346130774e-06, "loss": 0.0358, "step": 9978 }, { "epoch": 1.2, "learning_rate": 2.285254949010198e-06, "loss": 0.0006, "step": 9979 }, { "epoch": 1.2, "learning_rate": 2.2851829634073185e-06, "loss": 0.0005, "step": 9980 }, { "epoch": 1.2, "learning_rate": 2.285110977804439e-06, "loss": 0.0003, "step": 9981 }, { "epoch": 1.2, "learning_rate": 2.2850389922015597e-06, "loss": 0.0011, "step": 9982 }, { "epoch": 1.2, "learning_rate": 2.28496700659868e-06, "loss": 0.0005, "step": 9983 }, { "epoch": 1.2, "learning_rate": 2.284895020995801e-06, "loss": 0.0009, "step": 9984 }, { "epoch": 1.2, "learning_rate": 2.284823035392921e-06, "loss": 0.0003, "step": 9985 }, { "epoch": 1.2, "learning_rate": 2.284751049790042e-06, "loss": 0.0005, "step": 9986 }, { "epoch": 1.2, "learning_rate": 2.2846790641871627e-06, "loss": 0.0004, "step": 9987 }, { "epoch": 1.2, "learning_rate": 2.2846070785842835e-06, "loss": 0.0006, "step": 9988 }, { "epoch": 1.2, "learning_rate": 2.284535092981404e-06, "loss": 0.0006, "step": 9989 }, { "epoch": 1.2, "learning_rate": 2.2844631073785246e-06, "loss": 0.6412, "step": 9990 }, { "epoch": 1.2, "learning_rate": 2.284391121775645e-06, "loss": 0.0005, "step": 9991 }, { "epoch": 1.2, "learning_rate": 2.2843191361727658e-06, "loss": 0.002, "step": 9992 }, { "epoch": 1.2, "learning_rate": 2.284247150569886e-06, "loss": 0.0007, "step": 9993 }, { "epoch": 1.2, "learning_rate": 2.2841751649670065e-06, "loss": 0.0006, "step": 9994 }, { "epoch": 1.2, "learning_rate": 2.2841031793641273e-06, "loss": 0.0008, "step": 9995 }, { "epoch": 1.2, "learning_rate": 2.2840311937612476e-06, "loss": 0.1694, "step": 9996 }, { "epoch": 1.2, "learning_rate": 2.2839592081583684e-06, "loss": 0.0003, "step": 9997 }, { "epoch": 1.2, "learning_rate": 2.2838872225554887e-06, "loss": 0.0006, "step": 9998 }, { "epoch": 1.2, "learning_rate": 2.2838152369526095e-06, "loss": 0.0006, "step": 9999 }, { "epoch": 1.2, "learning_rate": 2.28374325134973e-06, "loss": 0.0005, "step": 10000 }, { "epoch": 1.2, "learning_rate": 2.2836712657468507e-06, "loss": 0.0009, "step": 10001 }, { "epoch": 1.2, "learning_rate": 2.2835992801439714e-06, "loss": 0.0004, "step": 10002 }, { "epoch": 1.2, "learning_rate": 2.2835272945410918e-06, "loss": 0.0006, "step": 10003 }, { "epoch": 1.2, "learning_rate": 2.2834553089382126e-06, "loss": 0.0005, "step": 10004 }, { "epoch": 1.2, "learning_rate": 2.283383323335333e-06, "loss": 0.001, "step": 10005 }, { "epoch": 1.2, "learning_rate": 2.2833113377324537e-06, "loss": 0.0057, "step": 10006 }, { "epoch": 1.2, "learning_rate": 2.283239352129574e-06, "loss": 0.0002, "step": 10007 }, { "epoch": 1.2, "learning_rate": 2.283167366526695e-06, "loss": 0.0005, "step": 10008 }, { "epoch": 1.2, "learning_rate": 2.283095380923815e-06, "loss": 0.0005, "step": 10009 }, { "epoch": 1.2, "learning_rate": 2.283023395320936e-06, "loss": 0.3138, "step": 10010 }, { "epoch": 1.2, "learning_rate": 2.2829514097180563e-06, "loss": 0.0007, "step": 10011 }, { "epoch": 1.2, "learning_rate": 2.282879424115177e-06, "loss": 0.0004, "step": 10012 }, { "epoch": 1.2, "learning_rate": 2.2828074385122975e-06, "loss": 0.7688, "step": 10013 }, { "epoch": 1.2, "learning_rate": 2.2827354529094182e-06, "loss": 0.0004, "step": 10014 }, { "epoch": 1.2, "learning_rate": 2.2826634673065386e-06, "loss": 0.8034, "step": 10015 }, { "epoch": 1.2, "learning_rate": 2.2825914817036594e-06, "loss": 0.0008, "step": 10016 }, { "epoch": 1.2, "learning_rate": 2.2825194961007797e-06, "loss": 0.0009, "step": 10017 }, { "epoch": 1.2, "learning_rate": 2.2824475104979005e-06, "loss": 0.1154, "step": 10018 }, { "epoch": 1.2, "learning_rate": 2.2823755248950213e-06, "loss": 0.0002, "step": 10019 }, { "epoch": 1.2, "learning_rate": 2.2823035392921416e-06, "loss": 0.0005, "step": 10020 }, { "epoch": 1.2, "learning_rate": 2.2822315536892624e-06, "loss": 0.0039, "step": 10021 }, { "epoch": 1.2, "learning_rate": 2.2821595680863828e-06, "loss": 0.0003, "step": 10022 }, { "epoch": 1.2, "learning_rate": 2.2820875824835035e-06, "loss": 0.0054, "step": 10023 }, { "epoch": 1.2, "learning_rate": 2.282015596880624e-06, "loss": 0.0002, "step": 10024 }, { "epoch": 1.2, "learning_rate": 2.2819436112777447e-06, "loss": 0.0005, "step": 10025 }, { "epoch": 1.2, "learning_rate": 2.281871625674865e-06, "loss": 0.0006, "step": 10026 }, { "epoch": 1.2, "learning_rate": 2.281799640071986e-06, "loss": 0.0003, "step": 10027 }, { "epoch": 1.2, "learning_rate": 2.281727654469106e-06, "loss": 0.0003, "step": 10028 }, { "epoch": 1.2, "learning_rate": 2.2816556688662265e-06, "loss": 0.0003, "step": 10029 }, { "epoch": 1.2, "learning_rate": 2.2815836832633473e-06, "loss": 0.0004, "step": 10030 }, { "epoch": 1.2, "learning_rate": 2.2815116976604677e-06, "loss": 0.0005, "step": 10031 }, { "epoch": 1.2, "learning_rate": 2.2814397120575884e-06, "loss": 0.0004, "step": 10032 }, { "epoch": 1.2, "learning_rate": 2.2813677264547092e-06, "loss": 0.0003, "step": 10033 }, { "epoch": 1.2, "learning_rate": 2.28129574085183e-06, "loss": 0.0006, "step": 10034 }, { "epoch": 1.2, "learning_rate": 2.2812237552489503e-06, "loss": 0.0021, "step": 10035 }, { "epoch": 1.2, "learning_rate": 2.281151769646071e-06, "loss": 0.0007, "step": 10036 }, { "epoch": 1.2, "learning_rate": 2.2810797840431915e-06, "loss": 0.0003, "step": 10037 }, { "epoch": 1.2, "learning_rate": 2.2810077984403123e-06, "loss": 0.0004, "step": 10038 }, { "epoch": 1.2, "learning_rate": 2.2809358128374326e-06, "loss": 0.0062, "step": 10039 }, { "epoch": 1.2, "learning_rate": 2.280863827234553e-06, "loss": 0.0019, "step": 10040 }, { "epoch": 1.2, "learning_rate": 2.2807918416316737e-06, "loss": 0.0004, "step": 10041 }, { "epoch": 1.2, "learning_rate": 2.280719856028794e-06, "loss": 0.0005, "step": 10042 }, { "epoch": 1.2, "learning_rate": 2.280647870425915e-06, "loss": 0.0003, "step": 10043 }, { "epoch": 1.2, "learning_rate": 2.2805758848230352e-06, "loss": 0.0007, "step": 10044 }, { "epoch": 1.2, "learning_rate": 2.280503899220156e-06, "loss": 0.0004, "step": 10045 }, { "epoch": 1.2, "learning_rate": 2.2804319136172764e-06, "loss": 0.0005, "step": 10046 }, { "epoch": 1.2, "learning_rate": 2.280359928014397e-06, "loss": 0.0004, "step": 10047 }, { "epoch": 1.2, "learning_rate": 2.280287942411518e-06, "loss": 0.0005, "step": 10048 }, { "epoch": 1.2, "learning_rate": 2.2802159568086383e-06, "loss": 0.0002, "step": 10049 }, { "epoch": 1.2, "learning_rate": 2.280143971205759e-06, "loss": 0.0004, "step": 10050 }, { "epoch": 1.2, "learning_rate": 2.2800719856028794e-06, "loss": 0.0006, "step": 10051 }, { "epoch": 1.2, "learning_rate": 2.28e-06, "loss": 0.0004, "step": 10052 }, { "epoch": 1.2, "learning_rate": 2.2799280143971205e-06, "loss": 0.0003, "step": 10053 }, { "epoch": 1.2, "learning_rate": 2.2798560287942413e-06, "loss": 0.0004, "step": 10054 }, { "epoch": 1.2, "learning_rate": 2.2797840431913617e-06, "loss": 0.0004, "step": 10055 }, { "epoch": 1.21, "learning_rate": 2.2797120575884825e-06, "loss": 0.0003, "step": 10056 }, { "epoch": 1.21, "learning_rate": 2.279640071985603e-06, "loss": 0.0006, "step": 10057 }, { "epoch": 1.21, "learning_rate": 2.2795680863827236e-06, "loss": 0.0004, "step": 10058 }, { "epoch": 1.21, "learning_rate": 2.279496100779844e-06, "loss": 0.0004, "step": 10059 }, { "epoch": 1.21, "learning_rate": 2.2794241151769647e-06, "loss": 0.0005, "step": 10060 }, { "epoch": 1.21, "learning_rate": 2.279352129574085e-06, "loss": 0.0003, "step": 10061 }, { "epoch": 1.21, "learning_rate": 2.279280143971206e-06, "loss": 0.0005, "step": 10062 }, { "epoch": 1.21, "learning_rate": 2.2792081583683262e-06, "loss": 0.0004, "step": 10063 }, { "epoch": 1.21, "learning_rate": 2.279136172765447e-06, "loss": 0.0003, "step": 10064 }, { "epoch": 1.21, "learning_rate": 2.2790641871625678e-06, "loss": 0.0003, "step": 10065 }, { "epoch": 1.21, "learning_rate": 2.278992201559688e-06, "loss": 0.0004, "step": 10066 }, { "epoch": 1.21, "learning_rate": 2.278920215956809e-06, "loss": 0.0003, "step": 10067 }, { "epoch": 1.21, "learning_rate": 2.2788482303539293e-06, "loss": 0.0003, "step": 10068 }, { "epoch": 1.21, "learning_rate": 2.27877624475105e-06, "loss": 0.0004, "step": 10069 }, { "epoch": 1.21, "learning_rate": 2.2787042591481704e-06, "loss": 0.0004, "step": 10070 }, { "epoch": 1.21, "learning_rate": 2.278632273545291e-06, "loss": 0.0006, "step": 10071 }, { "epoch": 1.21, "learning_rate": 2.2785602879424115e-06, "loss": 0.0002, "step": 10072 }, { "epoch": 1.21, "learning_rate": 2.2784883023395323e-06, "loss": 0.0143, "step": 10073 }, { "epoch": 1.21, "learning_rate": 2.2784163167366527e-06, "loss": 0.0002, "step": 10074 }, { "epoch": 1.21, "learning_rate": 2.278344331133773e-06, "loss": 0.0005, "step": 10075 }, { "epoch": 1.21, "learning_rate": 2.278272345530894e-06, "loss": 0.0005, "step": 10076 }, { "epoch": 1.21, "learning_rate": 2.278200359928014e-06, "loss": 0.0002, "step": 10077 }, { "epoch": 1.21, "learning_rate": 2.278128374325135e-06, "loss": 0.0003, "step": 10078 }, { "epoch": 1.21, "learning_rate": 2.2780563887222557e-06, "loss": 0.0002, "step": 10079 }, { "epoch": 1.21, "learning_rate": 2.2779844031193765e-06, "loss": 0.0006, "step": 10080 }, { "epoch": 1.21, "learning_rate": 2.277912417516497e-06, "loss": 0.0003, "step": 10081 }, { "epoch": 1.21, "learning_rate": 2.2778404319136176e-06, "loss": 0.0005, "step": 10082 }, { "epoch": 1.21, "learning_rate": 2.277768446310738e-06, "loss": 0.0003, "step": 10083 }, { "epoch": 1.21, "learning_rate": 2.2776964607078588e-06, "loss": 0.0007, "step": 10084 }, { "epoch": 1.21, "learning_rate": 2.277624475104979e-06, "loss": 0.0002, "step": 10085 }, { "epoch": 1.21, "learning_rate": 2.2775524895020995e-06, "loss": 0.0004, "step": 10086 }, { "epoch": 1.21, "learning_rate": 2.2774805038992202e-06, "loss": 0.0008, "step": 10087 }, { "epoch": 1.21, "learning_rate": 2.2774085182963406e-06, "loss": 0.0002, "step": 10088 }, { "epoch": 1.21, "learning_rate": 2.2773365326934614e-06, "loss": 0.0005, "step": 10089 }, { "epoch": 1.21, "learning_rate": 2.2772645470905817e-06, "loss": 0.0007, "step": 10090 }, { "epoch": 1.21, "learning_rate": 2.2771925614877025e-06, "loss": 0.0006, "step": 10091 }, { "epoch": 1.21, "learning_rate": 2.277120575884823e-06, "loss": 0.0003, "step": 10092 }, { "epoch": 1.21, "learning_rate": 2.2770485902819436e-06, "loss": 0.0006, "step": 10093 }, { "epoch": 1.21, "learning_rate": 2.2769766046790644e-06, "loss": 0.8137, "step": 10094 }, { "epoch": 1.21, "learning_rate": 2.2769046190761848e-06, "loss": 0.0002, "step": 10095 }, { "epoch": 1.21, "learning_rate": 2.2768326334733056e-06, "loss": 0.0005, "step": 10096 }, { "epoch": 1.21, "learning_rate": 2.276760647870426e-06, "loss": 0.0004, "step": 10097 }, { "epoch": 1.21, "learning_rate": 2.2766886622675467e-06, "loss": 0.0006, "step": 10098 }, { "epoch": 1.21, "learning_rate": 2.276616676664667e-06, "loss": 0.0004, "step": 10099 }, { "epoch": 1.21, "learning_rate": 2.276544691061788e-06, "loss": 0.0008, "step": 10100 }, { "epoch": 1.21, "learning_rate": 2.276472705458908e-06, "loss": 0.0003, "step": 10101 }, { "epoch": 1.21, "learning_rate": 2.276400719856029e-06, "loss": 0.0003, "step": 10102 }, { "epoch": 1.21, "learning_rate": 2.2763287342531493e-06, "loss": 0.0011, "step": 10103 }, { "epoch": 1.21, "learning_rate": 2.27625674865027e-06, "loss": 0.0004, "step": 10104 }, { "epoch": 1.21, "learning_rate": 2.2761847630473904e-06, "loss": 0.0005, "step": 10105 }, { "epoch": 1.21, "learning_rate": 2.2761127774445112e-06, "loss": 0.0003, "step": 10106 }, { "epoch": 1.21, "learning_rate": 2.2760407918416316e-06, "loss": 0.0003, "step": 10107 }, { "epoch": 1.21, "learning_rate": 2.275968806238752e-06, "loss": 0.0004, "step": 10108 }, { "epoch": 1.21, "learning_rate": 2.2758968206358727e-06, "loss": 0.0002, "step": 10109 }, { "epoch": 1.21, "learning_rate": 2.2758248350329935e-06, "loss": 0.0004, "step": 10110 }, { "epoch": 1.21, "learning_rate": 2.2757528494301143e-06, "loss": 0.0003, "step": 10111 }, { "epoch": 1.21, "learning_rate": 2.2756808638272346e-06, "loss": 0.0004, "step": 10112 }, { "epoch": 1.21, "learning_rate": 2.2756088782243554e-06, "loss": 0.0005, "step": 10113 }, { "epoch": 1.21, "learning_rate": 2.2755368926214758e-06, "loss": 0.0003, "step": 10114 }, { "epoch": 1.21, "learning_rate": 2.2754649070185965e-06, "loss": 0.0003, "step": 10115 }, { "epoch": 1.21, "learning_rate": 2.275392921415717e-06, "loss": 0.0004, "step": 10116 }, { "epoch": 1.21, "learning_rate": 2.2753209358128377e-06, "loss": 0.0003, "step": 10117 }, { "epoch": 1.21, "learning_rate": 2.275248950209958e-06, "loss": 0.0006, "step": 10118 }, { "epoch": 1.21, "learning_rate": 2.275176964607079e-06, "loss": 0.0012, "step": 10119 }, { "epoch": 1.21, "learning_rate": 2.275104979004199e-06, "loss": 1.0336, "step": 10120 }, { "epoch": 1.21, "learning_rate": 2.2750329934013195e-06, "loss": 0.0003, "step": 10121 }, { "epoch": 1.21, "learning_rate": 2.2749610077984403e-06, "loss": 0.0004, "step": 10122 }, { "epoch": 1.21, "learning_rate": 2.2748890221955606e-06, "loss": 0.0002, "step": 10123 }, { "epoch": 1.21, "learning_rate": 2.2748170365926814e-06, "loss": 0.0003, "step": 10124 }, { "epoch": 1.21, "learning_rate": 2.274745050989802e-06, "loss": 0.0015, "step": 10125 }, { "epoch": 1.21, "learning_rate": 2.2746730653869226e-06, "loss": 0.0003, "step": 10126 }, { "epoch": 1.21, "learning_rate": 2.2746010797840433e-06, "loss": 0.0155, "step": 10127 }, { "epoch": 1.21, "learning_rate": 2.274529094181164e-06, "loss": 0.0003, "step": 10128 }, { "epoch": 1.21, "learning_rate": 2.2744571085782845e-06, "loss": 0.0002, "step": 10129 }, { "epoch": 1.21, "learning_rate": 2.2743851229754053e-06, "loss": 0.0006, "step": 10130 }, { "epoch": 1.21, "learning_rate": 2.2743131373725256e-06, "loss": 0.0004, "step": 10131 }, { "epoch": 1.21, "learning_rate": 2.274241151769646e-06, "loss": 0.0004, "step": 10132 }, { "epoch": 1.21, "learning_rate": 2.2741691661667667e-06, "loss": 0.0005, "step": 10133 }, { "epoch": 1.21, "learning_rate": 2.274097180563887e-06, "loss": 0.0005, "step": 10134 }, { "epoch": 1.21, "learning_rate": 2.274025194961008e-06, "loss": 0.0004, "step": 10135 }, { "epoch": 1.21, "learning_rate": 2.2739532093581282e-06, "loss": 0.0004, "step": 10136 }, { "epoch": 1.21, "learning_rate": 2.273881223755249e-06, "loss": 0.0022, "step": 10137 }, { "epoch": 1.21, "learning_rate": 2.2738092381523694e-06, "loss": 0.6327, "step": 10138 }, { "epoch": 1.21, "learning_rate": 2.27373725254949e-06, "loss": 0.0786, "step": 10139 }, { "epoch": 1.22, "learning_rate": 2.2736652669466105e-06, "loss": 0.0003, "step": 10140 }, { "epoch": 1.22, "learning_rate": 2.2735932813437313e-06, "loss": 0.0003, "step": 10141 }, { "epoch": 1.22, "learning_rate": 2.273521295740852e-06, "loss": 0.0005, "step": 10142 }, { "epoch": 1.22, "learning_rate": 2.2734493101379724e-06, "loss": 0.0002, "step": 10143 }, { "epoch": 1.22, "learning_rate": 2.273377324535093e-06, "loss": 0.0003, "step": 10144 }, { "epoch": 1.22, "learning_rate": 2.2733053389322135e-06, "loss": 0.8978, "step": 10145 }, { "epoch": 1.22, "learning_rate": 2.2732333533293343e-06, "loss": 0.0005, "step": 10146 }, { "epoch": 1.22, "learning_rate": 2.2731613677264547e-06, "loss": 0.0006, "step": 10147 }, { "epoch": 1.22, "learning_rate": 2.2730893821235755e-06, "loss": 0.8565, "step": 10148 }, { "epoch": 1.22, "learning_rate": 2.273017396520696e-06, "loss": 0.0023, "step": 10149 }, { "epoch": 1.22, "learning_rate": 2.2729454109178166e-06, "loss": 0.0007, "step": 10150 }, { "epoch": 1.22, "learning_rate": 2.272873425314937e-06, "loss": 0.0003, "step": 10151 }, { "epoch": 1.22, "learning_rate": 2.2728014397120577e-06, "loss": 0.0005, "step": 10152 }, { "epoch": 1.22, "learning_rate": 2.272729454109178e-06, "loss": 1.0554, "step": 10153 }, { "epoch": 1.22, "learning_rate": 2.272657468506299e-06, "loss": 0.0004, "step": 10154 }, { "epoch": 1.22, "learning_rate": 2.272585482903419e-06, "loss": 0.0002, "step": 10155 }, { "epoch": 1.22, "learning_rate": 2.27251349730054e-06, "loss": 0.0003, "step": 10156 }, { "epoch": 1.22, "learning_rate": 2.2724415116976608e-06, "loss": 0.0008, "step": 10157 }, { "epoch": 1.22, "learning_rate": 2.272369526094781e-06, "loss": 0.0007, "step": 10158 }, { "epoch": 1.22, "learning_rate": 2.272297540491902e-06, "loss": 0.001, "step": 10159 }, { "epoch": 1.22, "learning_rate": 2.2722255548890223e-06, "loss": 0.0004, "step": 10160 }, { "epoch": 1.22, "learning_rate": 2.272153569286143e-06, "loss": 0.0004, "step": 10161 }, { "epoch": 1.22, "learning_rate": 2.2720815836832634e-06, "loss": 0.0002, "step": 10162 }, { "epoch": 1.22, "learning_rate": 2.272009598080384e-06, "loss": 0.0008, "step": 10163 }, { "epoch": 1.22, "learning_rate": 2.2719376124775045e-06, "loss": 0.0002, "step": 10164 }, { "epoch": 1.22, "learning_rate": 2.2718656268746253e-06, "loss": 0.0004, "step": 10165 }, { "epoch": 1.22, "learning_rate": 2.2717936412717457e-06, "loss": 0.8076, "step": 10166 }, { "epoch": 1.22, "learning_rate": 2.271721655668866e-06, "loss": 0.0002, "step": 10167 }, { "epoch": 1.22, "learning_rate": 2.271649670065987e-06, "loss": 0.0039, "step": 10168 }, { "epoch": 1.22, "learning_rate": 2.271577684463107e-06, "loss": 0.7249, "step": 10169 }, { "epoch": 1.22, "learning_rate": 2.271505698860228e-06, "loss": 0.0006, "step": 10170 }, { "epoch": 1.22, "learning_rate": 2.2714337132573487e-06, "loss": 0.0004, "step": 10171 }, { "epoch": 1.22, "learning_rate": 2.271361727654469e-06, "loss": 0.0029, "step": 10172 }, { "epoch": 1.22, "learning_rate": 2.27128974205159e-06, "loss": 0.0055, "step": 10173 }, { "epoch": 1.22, "learning_rate": 2.2712177564487106e-06, "loss": 0.0005, "step": 10174 }, { "epoch": 1.22, "learning_rate": 2.271145770845831e-06, "loss": 0.0004, "step": 10175 }, { "epoch": 1.22, "learning_rate": 2.2710737852429517e-06, "loss": 0.0005, "step": 10176 }, { "epoch": 1.22, "learning_rate": 2.271001799640072e-06, "loss": 0.0003, "step": 10177 }, { "epoch": 1.22, "learning_rate": 2.2709298140371925e-06, "loss": 0.0004, "step": 10178 }, { "epoch": 1.22, "learning_rate": 2.2708578284343132e-06, "loss": 0.0007, "step": 10179 }, { "epoch": 1.22, "learning_rate": 2.2707858428314336e-06, "loss": 0.0004, "step": 10180 }, { "epoch": 1.22, "learning_rate": 2.2707138572285544e-06, "loss": 0.0004, "step": 10181 }, { "epoch": 1.22, "learning_rate": 2.2706418716256747e-06, "loss": 0.0005, "step": 10182 }, { "epoch": 1.22, "learning_rate": 2.2705698860227955e-06, "loss": 0.0004, "step": 10183 }, { "epoch": 1.22, "learning_rate": 2.270497900419916e-06, "loss": 0.0004, "step": 10184 }, { "epoch": 1.22, "learning_rate": 2.2704259148170366e-06, "loss": 0.001, "step": 10185 }, { "epoch": 1.22, "learning_rate": 2.270353929214157e-06, "loss": 0.0033, "step": 10186 }, { "epoch": 1.22, "learning_rate": 2.2702819436112778e-06, "loss": 0.0008, "step": 10187 }, { "epoch": 1.22, "learning_rate": 2.2702099580083985e-06, "loss": 0.0004, "step": 10188 }, { "epoch": 1.22, "learning_rate": 2.270137972405519e-06, "loss": 0.0004, "step": 10189 }, { "epoch": 1.22, "learning_rate": 2.2700659868026397e-06, "loss": 0.0005, "step": 10190 }, { "epoch": 1.22, "learning_rate": 2.26999400119976e-06, "loss": 0.3087, "step": 10191 }, { "epoch": 1.22, "learning_rate": 2.269922015596881e-06, "loss": 0.0005, "step": 10192 }, { "epoch": 1.22, "learning_rate": 2.269850029994001e-06, "loss": 0.0004, "step": 10193 }, { "epoch": 1.22, "learning_rate": 2.269778044391122e-06, "loss": 0.0003, "step": 10194 }, { "epoch": 1.22, "learning_rate": 2.2697060587882423e-06, "loss": 0.0168, "step": 10195 }, { "epoch": 1.22, "learning_rate": 2.269634073185363e-06, "loss": 0.0004, "step": 10196 }, { "epoch": 1.22, "learning_rate": 2.2695620875824834e-06, "loss": 0.0006, "step": 10197 }, { "epoch": 1.22, "learning_rate": 2.2694901019796042e-06, "loss": 0.0003, "step": 10198 }, { "epoch": 1.22, "learning_rate": 2.2694181163767246e-06, "loss": 0.0008, "step": 10199 }, { "epoch": 1.22, "learning_rate": 2.2693461307738453e-06, "loss": 0.0003, "step": 10200 }, { "epoch": 1.22, "learning_rate": 2.2692741451709657e-06, "loss": 0.0053, "step": 10201 }, { "epoch": 1.22, "learning_rate": 2.2692021595680865e-06, "loss": 0.0003, "step": 10202 }, { "epoch": 1.22, "learning_rate": 2.2691301739652073e-06, "loss": 0.0004, "step": 10203 }, { "epoch": 1.22, "learning_rate": 2.2690581883623276e-06, "loss": 0.0005, "step": 10204 }, { "epoch": 1.22, "learning_rate": 2.2689862027594484e-06, "loss": 0.0004, "step": 10205 }, { "epoch": 1.22, "learning_rate": 2.2689142171565687e-06, "loss": 0.0004, "step": 10206 }, { "epoch": 1.22, "learning_rate": 2.2688422315536895e-06, "loss": 0.0003, "step": 10207 }, { "epoch": 1.22, "learning_rate": 2.26877024595081e-06, "loss": 0.0006, "step": 10208 }, { "epoch": 1.22, "learning_rate": 2.2686982603479307e-06, "loss": 0.0032, "step": 10209 }, { "epoch": 1.22, "learning_rate": 2.268626274745051e-06, "loss": 0.0007, "step": 10210 }, { "epoch": 1.22, "learning_rate": 2.268554289142172e-06, "loss": 0.0003, "step": 10211 }, { "epoch": 1.22, "learning_rate": 2.268482303539292e-06, "loss": 0.0006, "step": 10212 }, { "epoch": 1.22, "learning_rate": 2.2684103179364125e-06, "loss": 0.0014, "step": 10213 }, { "epoch": 1.22, "learning_rate": 2.2683383323335333e-06, "loss": 0.0011, "step": 10214 }, { "epoch": 1.22, "learning_rate": 2.2682663467306536e-06, "loss": 0.0007, "step": 10215 }, { "epoch": 1.22, "learning_rate": 2.2681943611277744e-06, "loss": 0.0003, "step": 10216 }, { "epoch": 1.22, "learning_rate": 2.268122375524895e-06, "loss": 0.0004, "step": 10217 }, { "epoch": 1.22, "learning_rate": 2.2680503899220156e-06, "loss": 0.0006, "step": 10218 }, { "epoch": 1.22, "learning_rate": 2.2679784043191363e-06, "loss": 0.0004, "step": 10219 }, { "epoch": 1.22, "learning_rate": 2.267906418716257e-06, "loss": 0.0003, "step": 10220 }, { "epoch": 1.22, "learning_rate": 2.2678344331133775e-06, "loss": 0.0006, "step": 10221 }, { "epoch": 1.22, "learning_rate": 2.2677624475104982e-06, "loss": 0.0748, "step": 10222 }, { "epoch": 1.23, "learning_rate": 2.2676904619076186e-06, "loss": 0.8587, "step": 10223 }, { "epoch": 1.23, "learning_rate": 2.267618476304739e-06, "loss": 0.0004, "step": 10224 }, { "epoch": 1.23, "learning_rate": 2.2675464907018597e-06, "loss": 0.0003, "step": 10225 }, { "epoch": 1.23, "learning_rate": 2.26747450509898e-06, "loss": 0.0003, "step": 10226 }, { "epoch": 1.23, "learning_rate": 2.267402519496101e-06, "loss": 0.0004, "step": 10227 }, { "epoch": 1.23, "learning_rate": 2.2673305338932212e-06, "loss": 0.0002, "step": 10228 }, { "epoch": 1.23, "learning_rate": 2.267258548290342e-06, "loss": 0.0005, "step": 10229 }, { "epoch": 1.23, "learning_rate": 2.2671865626874624e-06, "loss": 0.0003, "step": 10230 }, { "epoch": 1.23, "learning_rate": 2.267114577084583e-06, "loss": 0.0005, "step": 10231 }, { "epoch": 1.23, "learning_rate": 2.2670425914817035e-06, "loss": 0.1704, "step": 10232 }, { "epoch": 1.23, "learning_rate": 2.2669706058788243e-06, "loss": 0.8883, "step": 10233 }, { "epoch": 1.23, "learning_rate": 2.266898620275945e-06, "loss": 0.0004, "step": 10234 }, { "epoch": 1.23, "learning_rate": 2.2668266346730654e-06, "loss": 0.0004, "step": 10235 }, { "epoch": 1.23, "learning_rate": 2.266754649070186e-06, "loss": 0.0004, "step": 10236 }, { "epoch": 1.23, "learning_rate": 2.2666826634673065e-06, "loss": 0.0004, "step": 10237 }, { "epoch": 1.23, "learning_rate": 2.2666106778644273e-06, "loss": 0.0003, "step": 10238 }, { "epoch": 1.23, "learning_rate": 2.2665386922615477e-06, "loss": 0.0003, "step": 10239 }, { "epoch": 1.23, "learning_rate": 2.2664667066586684e-06, "loss": 0.0004, "step": 10240 }, { "epoch": 1.23, "learning_rate": 2.266394721055789e-06, "loss": 0.0004, "step": 10241 }, { "epoch": 1.23, "learning_rate": 2.2663227354529096e-06, "loss": 0.0006, "step": 10242 }, { "epoch": 1.23, "learning_rate": 2.26625074985003e-06, "loss": 0.0003, "step": 10243 }, { "epoch": 1.23, "learning_rate": 2.2661787642471507e-06, "loss": 0.0003, "step": 10244 }, { "epoch": 1.23, "learning_rate": 2.266106778644271e-06, "loss": 0.0007, "step": 10245 }, { "epoch": 1.23, "learning_rate": 2.266034793041392e-06, "loss": 0.0005, "step": 10246 }, { "epoch": 1.23, "learning_rate": 2.265962807438512e-06, "loss": 0.0003, "step": 10247 }, { "epoch": 1.23, "learning_rate": 2.265890821835633e-06, "loss": 0.0003, "step": 10248 }, { "epoch": 1.23, "learning_rate": 2.2658188362327533e-06, "loss": 0.0006, "step": 10249 }, { "epoch": 1.23, "learning_rate": 2.265746850629874e-06, "loss": 0.0008, "step": 10250 }, { "epoch": 1.23, "learning_rate": 2.265674865026995e-06, "loss": 0.0006, "step": 10251 }, { "epoch": 1.23, "learning_rate": 2.2656028794241152e-06, "loss": 0.0003, "step": 10252 }, { "epoch": 1.23, "learning_rate": 2.265530893821236e-06, "loss": 0.0004, "step": 10253 }, { "epoch": 1.23, "learning_rate": 2.2654589082183564e-06, "loss": 0.0005, "step": 10254 }, { "epoch": 1.23, "learning_rate": 2.265386922615477e-06, "loss": 0.5025, "step": 10255 }, { "epoch": 1.23, "learning_rate": 2.2653149370125975e-06, "loss": 0.0005, "step": 10256 }, { "epoch": 1.23, "learning_rate": 2.2652429514097183e-06, "loss": 0.0003, "step": 10257 }, { "epoch": 1.23, "learning_rate": 2.2651709658068386e-06, "loss": 0.0003, "step": 10258 }, { "epoch": 1.23, "learning_rate": 2.265098980203959e-06, "loss": 0.0003, "step": 10259 }, { "epoch": 1.23, "learning_rate": 2.2650269946010798e-06, "loss": 0.0002, "step": 10260 }, { "epoch": 1.23, "learning_rate": 2.2649550089982e-06, "loss": 0.0006, "step": 10261 }, { "epoch": 1.23, "learning_rate": 2.264883023395321e-06, "loss": 0.0007, "step": 10262 }, { "epoch": 1.23, "learning_rate": 2.2648110377924413e-06, "loss": 0.0006, "step": 10263 }, { "epoch": 1.23, "learning_rate": 2.264739052189562e-06, "loss": 0.0005, "step": 10264 }, { "epoch": 1.23, "learning_rate": 2.264667066586683e-06, "loss": 0.0005, "step": 10265 }, { "epoch": 1.23, "learning_rate": 2.2645950809838036e-06, "loss": 0.0004, "step": 10266 }, { "epoch": 1.23, "learning_rate": 2.264523095380924e-06, "loss": 0.0005, "step": 10267 }, { "epoch": 1.23, "learning_rate": 2.2644511097780447e-06, "loss": 0.0027, "step": 10268 }, { "epoch": 1.23, "learning_rate": 2.264379124175165e-06, "loss": 0.6929, "step": 10269 }, { "epoch": 1.23, "learning_rate": 2.2643071385722854e-06, "loss": 0.0005, "step": 10270 }, { "epoch": 1.23, "learning_rate": 2.2642351529694062e-06, "loss": 0.0004, "step": 10271 }, { "epoch": 1.23, "learning_rate": 2.2641631673665266e-06, "loss": 0.0007, "step": 10272 }, { "epoch": 1.23, "learning_rate": 2.2640911817636474e-06, "loss": 0.0005, "step": 10273 }, { "epoch": 1.23, "learning_rate": 2.2640191961607677e-06, "loss": 0.0004, "step": 10274 }, { "epoch": 1.23, "learning_rate": 2.2639472105578885e-06, "loss": 0.0006, "step": 10275 }, { "epoch": 1.23, "learning_rate": 2.263875224955009e-06, "loss": 0.0003, "step": 10276 }, { "epoch": 1.23, "learning_rate": 2.2638032393521296e-06, "loss": 0.0003, "step": 10277 }, { "epoch": 1.23, "learning_rate": 2.26373125374925e-06, "loss": 0.0005, "step": 10278 }, { "epoch": 1.23, "learning_rate": 2.2636592681463708e-06, "loss": 0.0004, "step": 10279 }, { "epoch": 1.23, "learning_rate": 2.2635872825434915e-06, "loss": 0.0005, "step": 10280 }, { "epoch": 1.23, "learning_rate": 2.263515296940612e-06, "loss": 0.0003, "step": 10281 }, { "epoch": 1.23, "learning_rate": 2.2634433113377327e-06, "loss": 0.758, "step": 10282 }, { "epoch": 1.23, "learning_rate": 2.263371325734853e-06, "loss": 0.0003, "step": 10283 }, { "epoch": 1.23, "learning_rate": 2.263299340131974e-06, "loss": 0.0006, "step": 10284 }, { "epoch": 1.23, "learning_rate": 2.263227354529094e-06, "loss": 0.0006, "step": 10285 }, { "epoch": 1.23, "learning_rate": 2.263155368926215e-06, "loss": 0.0003, "step": 10286 }, { "epoch": 1.23, "learning_rate": 2.2630833833233353e-06, "loss": 0.8048, "step": 10287 }, { "epoch": 1.23, "learning_rate": 2.263011397720456e-06, "loss": 0.0003, "step": 10288 }, { "epoch": 1.23, "learning_rate": 2.2629394121175764e-06, "loss": 0.0003, "step": 10289 }, { "epoch": 1.23, "learning_rate": 2.262867426514697e-06, "loss": 0.0007, "step": 10290 }, { "epoch": 1.23, "learning_rate": 2.2627954409118176e-06, "loss": 0.0002, "step": 10291 }, { "epoch": 1.23, "learning_rate": 2.2627234553089383e-06, "loss": 0.0003, "step": 10292 }, { "epoch": 1.23, "learning_rate": 2.2626514697060587e-06, "loss": 0.0003, "step": 10293 }, { "epoch": 1.23, "learning_rate": 2.2625794841031795e-06, "loss": 0.0004, "step": 10294 }, { "epoch": 1.23, "learning_rate": 2.2625074985003e-06, "loss": 0.0004, "step": 10295 }, { "epoch": 1.23, "learning_rate": 2.2624355128974206e-06, "loss": 0.0005, "step": 10296 }, { "epoch": 1.23, "learning_rate": 2.2623635272945414e-06, "loss": 0.6724, "step": 10297 }, { "epoch": 1.23, "learning_rate": 2.2622915416916617e-06, "loss": 0.0003, "step": 10298 }, { "epoch": 1.23, "learning_rate": 2.2622195560887825e-06, "loss": 0.0009, "step": 10299 }, { "epoch": 1.23, "learning_rate": 2.262147570485903e-06, "loss": 0.7234, "step": 10300 }, { "epoch": 1.23, "learning_rate": 2.2620755848830237e-06, "loss": 0.0003, "step": 10301 }, { "epoch": 1.23, "learning_rate": 2.262003599280144e-06, "loss": 0.0003, "step": 10302 }, { "epoch": 1.23, "learning_rate": 2.2619316136772648e-06, "loss": 0.0003, "step": 10303 }, { "epoch": 1.23, "learning_rate": 2.261859628074385e-06, "loss": 0.0004, "step": 10304 }, { "epoch": 1.23, "learning_rate": 2.2617876424715055e-06, "loss": 0.0005, "step": 10305 }, { "epoch": 1.23, "learning_rate": 2.2617156568686263e-06, "loss": 0.0004, "step": 10306 }, { "epoch": 1.24, "learning_rate": 2.2616436712657466e-06, "loss": 0.0003, "step": 10307 }, { "epoch": 1.24, "learning_rate": 2.2615716856628674e-06, "loss": 0.0016, "step": 10308 }, { "epoch": 1.24, "learning_rate": 2.2614997000599878e-06, "loss": 0.0004, "step": 10309 }, { "epoch": 1.24, "learning_rate": 2.2614277144571085e-06, "loss": 0.0015, "step": 10310 }, { "epoch": 1.24, "learning_rate": 2.2613557288542293e-06, "loss": 0.0004, "step": 10311 }, { "epoch": 1.24, "learning_rate": 2.26128374325135e-06, "loss": 0.2084, "step": 10312 }, { "epoch": 1.24, "learning_rate": 2.2612117576484705e-06, "loss": 0.0006, "step": 10313 }, { "epoch": 1.24, "learning_rate": 2.2611397720455912e-06, "loss": 0.0003, "step": 10314 }, { "epoch": 1.24, "learning_rate": 2.2610677864427116e-06, "loss": 0.0003, "step": 10315 }, { "epoch": 1.24, "learning_rate": 2.260995800839832e-06, "loss": 0.0004, "step": 10316 }, { "epoch": 1.24, "learning_rate": 2.2609238152369527e-06, "loss": 0.0002, "step": 10317 }, { "epoch": 1.24, "learning_rate": 2.260851829634073e-06, "loss": 0.2121, "step": 10318 }, { "epoch": 1.24, "learning_rate": 2.260779844031194e-06, "loss": 0.0016, "step": 10319 }, { "epoch": 1.24, "learning_rate": 2.260707858428314e-06, "loss": 0.0005, "step": 10320 }, { "epoch": 1.24, "learning_rate": 2.260635872825435e-06, "loss": 0.0004, "step": 10321 }, { "epoch": 1.24, "learning_rate": 2.2605638872225553e-06, "loss": 0.0003, "step": 10322 }, { "epoch": 1.24, "learning_rate": 2.260491901619676e-06, "loss": 0.0003, "step": 10323 }, { "epoch": 1.24, "learning_rate": 2.2604199160167965e-06, "loss": 0.0003, "step": 10324 }, { "epoch": 1.24, "learning_rate": 2.2603479304139173e-06, "loss": 0.0004, "step": 10325 }, { "epoch": 1.24, "learning_rate": 2.260275944811038e-06, "loss": 0.0003, "step": 10326 }, { "epoch": 1.24, "learning_rate": 2.2602039592081584e-06, "loss": 0.0003, "step": 10327 }, { "epoch": 1.24, "learning_rate": 2.260131973605279e-06, "loss": 0.0002, "step": 10328 }, { "epoch": 1.24, "learning_rate": 2.2600599880023995e-06, "loss": 0.0004, "step": 10329 }, { "epoch": 1.24, "learning_rate": 2.2599880023995203e-06, "loss": 0.0004, "step": 10330 }, { "epoch": 1.24, "learning_rate": 2.2599160167966407e-06, "loss": 1.1529, "step": 10331 }, { "epoch": 1.24, "learning_rate": 2.2598440311937614e-06, "loss": 0.0004, "step": 10332 }, { "epoch": 1.24, "learning_rate": 2.259772045590882e-06, "loss": 0.0003, "step": 10333 }, { "epoch": 1.24, "learning_rate": 2.2597000599880026e-06, "loss": 0.1013, "step": 10334 }, { "epoch": 1.24, "learning_rate": 2.259628074385123e-06, "loss": 0.0003, "step": 10335 }, { "epoch": 1.24, "learning_rate": 2.2595560887822437e-06, "loss": 0.0004, "step": 10336 }, { "epoch": 1.24, "learning_rate": 2.259484103179364e-06, "loss": 0.0004, "step": 10337 }, { "epoch": 1.24, "learning_rate": 2.259412117576485e-06, "loss": 0.0003, "step": 10338 }, { "epoch": 1.24, "learning_rate": 2.259340131973605e-06, "loss": 0.7934, "step": 10339 }, { "epoch": 1.24, "learning_rate": 2.259268146370726e-06, "loss": 0.0003, "step": 10340 }, { "epoch": 1.24, "learning_rate": 2.2591961607678463e-06, "loss": 0.0003, "step": 10341 }, { "epoch": 1.24, "learning_rate": 2.259124175164967e-06, "loss": 0.0006, "step": 10342 }, { "epoch": 1.24, "learning_rate": 2.259052189562088e-06, "loss": 0.0006, "step": 10343 }, { "epoch": 1.24, "learning_rate": 2.2589802039592082e-06, "loss": 0.0004, "step": 10344 }, { "epoch": 1.24, "learning_rate": 2.258908218356329e-06, "loss": 0.0004, "step": 10345 }, { "epoch": 1.24, "learning_rate": 2.2588362327534494e-06, "loss": 0.0007, "step": 10346 }, { "epoch": 1.24, "learning_rate": 2.25876424715057e-06, "loss": 0.0004, "step": 10347 }, { "epoch": 1.24, "learning_rate": 2.2586922615476905e-06, "loss": 0.0003, "step": 10348 }, { "epoch": 1.24, "learning_rate": 2.2586202759448113e-06, "loss": 0.9238, "step": 10349 }, { "epoch": 1.24, "learning_rate": 2.2585482903419316e-06, "loss": 0.0003, "step": 10350 }, { "epoch": 1.24, "learning_rate": 2.258476304739052e-06, "loss": 0.0003, "step": 10351 }, { "epoch": 1.24, "learning_rate": 2.2584043191361728e-06, "loss": 0.0299, "step": 10352 }, { "epoch": 1.24, "learning_rate": 2.258332333533293e-06, "loss": 0.9044, "step": 10353 }, { "epoch": 1.24, "learning_rate": 2.258260347930414e-06, "loss": 0.0004, "step": 10354 }, { "epoch": 1.24, "learning_rate": 2.2581883623275343e-06, "loss": 0.0004, "step": 10355 }, { "epoch": 1.24, "learning_rate": 2.258116376724655e-06, "loss": 0.0005, "step": 10356 }, { "epoch": 1.24, "learning_rate": 2.258044391121776e-06, "loss": 0.0004, "step": 10357 }, { "epoch": 1.24, "learning_rate": 2.2579724055188966e-06, "loss": 0.0003, "step": 10358 }, { "epoch": 1.24, "learning_rate": 2.257900419916017e-06, "loss": 0.0003, "step": 10359 }, { "epoch": 1.24, "learning_rate": 2.2578284343131377e-06, "loss": 0.0004, "step": 10360 }, { "epoch": 1.24, "learning_rate": 2.257756448710258e-06, "loss": 0.0028, "step": 10361 }, { "epoch": 1.24, "learning_rate": 2.2576844631073784e-06, "loss": 0.0007, "step": 10362 }, { "epoch": 1.24, "learning_rate": 2.2576124775044992e-06, "loss": 0.0006, "step": 10363 }, { "epoch": 1.24, "learning_rate": 2.2575404919016196e-06, "loss": 0.0039, "step": 10364 }, { "epoch": 1.24, "learning_rate": 2.2574685062987404e-06, "loss": 0.0012, "step": 10365 }, { "epoch": 1.24, "learning_rate": 2.2573965206958607e-06, "loss": 0.0003, "step": 10366 }, { "epoch": 1.24, "learning_rate": 2.2573245350929815e-06, "loss": 0.0014, "step": 10367 }, { "epoch": 1.24, "learning_rate": 2.257252549490102e-06, "loss": 0.9877, "step": 10368 }, { "epoch": 1.24, "learning_rate": 2.2571805638872226e-06, "loss": 0.0011, "step": 10369 }, { "epoch": 1.24, "learning_rate": 2.257108578284343e-06, "loss": 0.0004, "step": 10370 }, { "epoch": 1.24, "learning_rate": 2.2570365926814638e-06, "loss": 0.0004, "step": 10371 }, { "epoch": 1.24, "learning_rate": 2.256964607078584e-06, "loss": 0.0018, "step": 10372 }, { "epoch": 1.24, "learning_rate": 2.256892621475705e-06, "loss": 0.0006, "step": 10373 }, { "epoch": 1.24, "learning_rate": 2.2568206358728257e-06, "loss": 0.001, "step": 10374 }, { "epoch": 1.24, "learning_rate": 2.256748650269946e-06, "loss": 0.0007, "step": 10375 }, { "epoch": 1.24, "learning_rate": 2.256676664667067e-06, "loss": 0.0006, "step": 10376 }, { "epoch": 1.24, "learning_rate": 2.256604679064187e-06, "loss": 0.0004, "step": 10377 }, { "epoch": 1.24, "learning_rate": 2.256532693461308e-06, "loss": 0.0006, "step": 10378 }, { "epoch": 1.24, "learning_rate": 2.2564607078584283e-06, "loss": 0.0004, "step": 10379 }, { "epoch": 1.24, "learning_rate": 2.256388722255549e-06, "loss": 0.0004, "step": 10380 }, { "epoch": 1.24, "learning_rate": 2.2563167366526694e-06, "loss": 0.0006, "step": 10381 }, { "epoch": 1.24, "learning_rate": 2.25624475104979e-06, "loss": 0.6948, "step": 10382 }, { "epoch": 1.24, "learning_rate": 2.2561727654469106e-06, "loss": 0.0008, "step": 10383 }, { "epoch": 1.24, "learning_rate": 2.2561007798440313e-06, "loss": 0.0037, "step": 10384 }, { "epoch": 1.24, "learning_rate": 2.2560287942411517e-06, "loss": 0.0006, "step": 10385 }, { "epoch": 1.24, "learning_rate": 2.255956808638272e-06, "loss": 0.0006, "step": 10386 }, { "epoch": 1.24, "learning_rate": 2.255884823035393e-06, "loss": 0.0006, "step": 10387 }, { "epoch": 1.24, "learning_rate": 2.2558128374325136e-06, "loss": 0.0008, "step": 10388 }, { "epoch": 1.24, "learning_rate": 2.2557408518296344e-06, "loss": 0.0008, "step": 10389 }, { "epoch": 1.25, "learning_rate": 2.2556688662267547e-06, "loss": 0.0003, "step": 10390 }, { "epoch": 1.25, "learning_rate": 2.2555968806238755e-06, "loss": 0.0004, "step": 10391 }, { "epoch": 1.25, "learning_rate": 2.255524895020996e-06, "loss": 0.5553, "step": 10392 }, { "epoch": 1.25, "learning_rate": 2.2554529094181166e-06, "loss": 0.7108, "step": 10393 }, { "epoch": 1.25, "learning_rate": 2.255380923815237e-06, "loss": 0.0004, "step": 10394 }, { "epoch": 1.25, "learning_rate": 2.2553089382123578e-06, "loss": 0.0007, "step": 10395 }, { "epoch": 1.25, "learning_rate": 2.255236952609478e-06, "loss": 0.0006, "step": 10396 }, { "epoch": 1.25, "learning_rate": 2.2551649670065985e-06, "loss": 0.0004, "step": 10397 }, { "epoch": 1.25, "learning_rate": 2.2550929814037193e-06, "loss": 0.0006, "step": 10398 }, { "epoch": 1.25, "learning_rate": 2.2550209958008396e-06, "loss": 0.0005, "step": 10399 }, { "epoch": 1.25, "learning_rate": 2.2549490101979604e-06, "loss": 0.0004, "step": 10400 }, { "epoch": 1.25, "learning_rate": 2.2548770245950808e-06, "loss": 0.0007, "step": 10401 }, { "epoch": 1.25, "learning_rate": 2.2548050389922015e-06, "loss": 0.3044, "step": 10402 }, { "epoch": 1.25, "learning_rate": 2.2547330533893223e-06, "loss": 0.0006, "step": 10403 }, { "epoch": 1.25, "learning_rate": 2.2546610677864427e-06, "loss": 0.0533, "step": 10404 }, { "epoch": 1.25, "learning_rate": 2.2545890821835634e-06, "loss": 0.0003, "step": 10405 }, { "epoch": 1.25, "learning_rate": 2.2545170965806842e-06, "loss": 0.0006, "step": 10406 }, { "epoch": 1.25, "learning_rate": 2.2544451109778046e-06, "loss": 0.0006, "step": 10407 }, { "epoch": 1.25, "learning_rate": 2.2543731253749254e-06, "loss": 0.0004, "step": 10408 }, { "epoch": 1.25, "learning_rate": 2.2543011397720457e-06, "loss": 0.0005, "step": 10409 }, { "epoch": 1.25, "learning_rate": 2.254229154169166e-06, "loss": 0.9818, "step": 10410 }, { "epoch": 1.25, "learning_rate": 2.254157168566287e-06, "loss": 0.0009, "step": 10411 }, { "epoch": 1.25, "learning_rate": 2.254085182963407e-06, "loss": 0.0005, "step": 10412 }, { "epoch": 1.25, "learning_rate": 2.254013197360528e-06, "loss": 0.0004, "step": 10413 }, { "epoch": 1.25, "learning_rate": 2.2539412117576483e-06, "loss": 0.001, "step": 10414 }, { "epoch": 1.25, "learning_rate": 2.253869226154769e-06, "loss": 0.0006, "step": 10415 }, { "epoch": 1.25, "learning_rate": 2.2537972405518895e-06, "loss": 0.0012, "step": 10416 }, { "epoch": 1.25, "learning_rate": 2.2537252549490102e-06, "loss": 0.0008, "step": 10417 }, { "epoch": 1.25, "learning_rate": 2.2536532693461306e-06, "loss": 0.0004, "step": 10418 }, { "epoch": 1.25, "learning_rate": 2.2535812837432514e-06, "loss": 0.0005, "step": 10419 }, { "epoch": 1.25, "learning_rate": 2.253509298140372e-06, "loss": 0.0012, "step": 10420 }, { "epoch": 1.25, "learning_rate": 2.2534373125374925e-06, "loss": 0.0007, "step": 10421 }, { "epoch": 1.25, "learning_rate": 2.2533653269346133e-06, "loss": 0.0009, "step": 10422 }, { "epoch": 1.25, "learning_rate": 2.2532933413317336e-06, "loss": 0.2728, "step": 10423 }, { "epoch": 1.25, "learning_rate": 2.2532213557288544e-06, "loss": 0.7104, "step": 10424 }, { "epoch": 1.25, "learning_rate": 2.2531493701259748e-06, "loss": 0.9029, "step": 10425 }, { "epoch": 1.25, "learning_rate": 2.2530773845230956e-06, "loss": 0.0032, "step": 10426 }, { "epoch": 1.25, "learning_rate": 2.253005398920216e-06, "loss": 0.94, "step": 10427 }, { "epoch": 1.25, "learning_rate": 2.2529334133173367e-06, "loss": 0.9292, "step": 10428 }, { "epoch": 1.25, "learning_rate": 2.252861427714457e-06, "loss": 0.0007, "step": 10429 }, { "epoch": 1.25, "learning_rate": 2.252789442111578e-06, "loss": 0.001, "step": 10430 }, { "epoch": 1.25, "learning_rate": 2.252717456508698e-06, "loss": 0.0099, "step": 10431 }, { "epoch": 1.25, "learning_rate": 2.2526454709058185e-06, "loss": 0.0249, "step": 10432 }, { "epoch": 1.25, "learning_rate": 2.2525734853029393e-06, "loss": 0.0012, "step": 10433 }, { "epoch": 1.25, "learning_rate": 2.25250149970006e-06, "loss": 0.001, "step": 10434 }, { "epoch": 1.25, "learning_rate": 2.252429514097181e-06, "loss": 0.0007, "step": 10435 }, { "epoch": 1.25, "learning_rate": 2.2523575284943012e-06, "loss": 0.0002, "step": 10436 }, { "epoch": 1.25, "learning_rate": 2.252285542891422e-06, "loss": 0.0136, "step": 10437 }, { "epoch": 1.25, "learning_rate": 2.2522135572885424e-06, "loss": 0.0006, "step": 10438 }, { "epoch": 1.25, "learning_rate": 2.252141571685663e-06, "loss": 0.1568, "step": 10439 }, { "epoch": 1.25, "learning_rate": 2.2520695860827835e-06, "loss": 0.0012, "step": 10440 }, { "epoch": 1.25, "learning_rate": 2.2519976004799043e-06, "loss": 0.7855, "step": 10441 }, { "epoch": 1.25, "learning_rate": 2.2519256148770246e-06, "loss": 0.0009, "step": 10442 }, { "epoch": 1.25, "learning_rate": 2.251853629274145e-06, "loss": 0.001, "step": 10443 }, { "epoch": 1.25, "learning_rate": 2.2517816436712658e-06, "loss": 0.1314, "step": 10444 }, { "epoch": 1.25, "learning_rate": 2.251709658068386e-06, "loss": 0.1055, "step": 10445 }, { "epoch": 1.25, "learning_rate": 2.251637672465507e-06, "loss": 0.0009, "step": 10446 }, { "epoch": 1.25, "learning_rate": 2.2515656868626273e-06, "loss": 0.0008, "step": 10447 }, { "epoch": 1.25, "learning_rate": 2.251493701259748e-06, "loss": 0.0011, "step": 10448 }, { "epoch": 1.25, "learning_rate": 2.251421715656869e-06, "loss": 0.0011, "step": 10449 }, { "epoch": 1.25, "learning_rate": 2.251349730053989e-06, "loss": 0.0015, "step": 10450 }, { "epoch": 1.25, "learning_rate": 2.25127774445111e-06, "loss": 0.0015, "step": 10451 }, { "epoch": 1.25, "learning_rate": 2.2512057588482307e-06, "loss": 0.0014, "step": 10452 }, { "epoch": 1.25, "learning_rate": 2.251133773245351e-06, "loss": 0.0006, "step": 10453 }, { "epoch": 1.25, "learning_rate": 2.251061787642472e-06, "loss": 0.0014, "step": 10454 }, { "epoch": 1.25, "learning_rate": 2.250989802039592e-06, "loss": 0.0026, "step": 10455 }, { "epoch": 1.25, "learning_rate": 2.2509178164367126e-06, "loss": 0.2549, "step": 10456 }, { "epoch": 1.25, "learning_rate": 2.2508458308338333e-06, "loss": 0.7093, "step": 10457 }, { "epoch": 1.25, "learning_rate": 2.2507738452309537e-06, "loss": 0.0012, "step": 10458 }, { "epoch": 1.25, "learning_rate": 2.2507018596280745e-06, "loss": 0.0081, "step": 10459 }, { "epoch": 1.25, "learning_rate": 2.250629874025195e-06, "loss": 0.0014, "step": 10460 }, { "epoch": 1.25, "learning_rate": 2.2505578884223156e-06, "loss": 0.0009, "step": 10461 }, { "epoch": 1.25, "learning_rate": 2.250485902819436e-06, "loss": 0.001, "step": 10462 }, { "epoch": 1.25, "learning_rate": 2.2504139172165567e-06, "loss": 0.0029, "step": 10463 }, { "epoch": 1.25, "learning_rate": 2.250341931613677e-06, "loss": 0.0012, "step": 10464 }, { "epoch": 1.25, "learning_rate": 2.250269946010798e-06, "loss": 0.0014, "step": 10465 }, { "epoch": 1.25, "learning_rate": 2.2501979604079187e-06, "loss": 0.0012, "step": 10466 }, { "epoch": 1.25, "learning_rate": 2.250125974805039e-06, "loss": 0.0007, "step": 10467 }, { "epoch": 1.25, "learning_rate": 2.25005398920216e-06, "loss": 0.001, "step": 10468 }, { "epoch": 1.25, "learning_rate": 2.24998200359928e-06, "loss": 0.0007, "step": 10469 }, { "epoch": 1.25, "learning_rate": 2.249910017996401e-06, "loss": 0.0004, "step": 10470 }, { "epoch": 1.25, "learning_rate": 2.2498380323935213e-06, "loss": 0.0173, "step": 10471 }, { "epoch": 1.25, "learning_rate": 2.249766046790642e-06, "loss": 0.0006, "step": 10472 }, { "epoch": 1.26, "learning_rate": 2.2496940611877624e-06, "loss": 0.4316, "step": 10473 }, { "epoch": 1.26, "learning_rate": 2.249622075584883e-06, "loss": 0.0007, "step": 10474 }, { "epoch": 1.26, "learning_rate": 2.2495500899820035e-06, "loss": 0.0014, "step": 10475 }, { "epoch": 1.26, "learning_rate": 2.2494781043791243e-06, "loss": 0.0006, "step": 10476 }, { "epoch": 1.26, "learning_rate": 2.2494061187762447e-06, "loss": 0.0006, "step": 10477 }, { "epoch": 1.26, "learning_rate": 2.249334133173365e-06, "loss": 0.0011, "step": 10478 }, { "epoch": 1.26, "learning_rate": 2.249262147570486e-06, "loss": 0.0008, "step": 10479 }, { "epoch": 1.26, "learning_rate": 2.2491901619676066e-06, "loss": 0.0013, "step": 10480 }, { "epoch": 1.26, "learning_rate": 2.2491181763647274e-06, "loss": 0.0006, "step": 10481 }, { "epoch": 1.26, "learning_rate": 2.2490461907618477e-06, "loss": 0.0008, "step": 10482 }, { "epoch": 1.26, "learning_rate": 2.2489742051589685e-06, "loss": 0.0011, "step": 10483 }, { "epoch": 1.26, "learning_rate": 2.248902219556089e-06, "loss": 0.0008, "step": 10484 }, { "epoch": 1.26, "learning_rate": 2.2488302339532096e-06, "loss": 0.0083, "step": 10485 }, { "epoch": 1.26, "learning_rate": 2.24875824835033e-06, "loss": 0.0005, "step": 10486 }, { "epoch": 1.26, "learning_rate": 2.2486862627474508e-06, "loss": 0.0014, "step": 10487 }, { "epoch": 1.26, "learning_rate": 2.248614277144571e-06, "loss": 0.0005, "step": 10488 }, { "epoch": 1.26, "learning_rate": 2.2485422915416915e-06, "loss": 0.0008, "step": 10489 }, { "epoch": 1.26, "learning_rate": 2.2484703059388123e-06, "loss": 0.0007, "step": 10490 }, { "epoch": 1.26, "learning_rate": 2.2483983203359326e-06, "loss": 0.0019, "step": 10491 }, { "epoch": 1.26, "learning_rate": 2.2483263347330534e-06, "loss": 0.0004, "step": 10492 }, { "epoch": 1.26, "learning_rate": 2.2482543491301737e-06, "loss": 0.0012, "step": 10493 }, { "epoch": 1.26, "learning_rate": 2.2481823635272945e-06, "loss": 0.6823, "step": 10494 }, { "epoch": 1.26, "learning_rate": 2.2481103779244153e-06, "loss": 0.0009, "step": 10495 }, { "epoch": 1.26, "learning_rate": 2.2480383923215357e-06, "loss": 0.0013, "step": 10496 }, { "epoch": 1.26, "learning_rate": 2.2479664067186564e-06, "loss": 0.8596, "step": 10497 }, { "epoch": 1.26, "learning_rate": 2.2478944211157772e-06, "loss": 0.0006, "step": 10498 }, { "epoch": 1.26, "learning_rate": 2.2478224355128976e-06, "loss": 0.002, "step": 10499 }, { "epoch": 1.26, "learning_rate": 2.2477504499100184e-06, "loss": 0.0008, "step": 10500 }, { "epoch": 1.26, "learning_rate": 2.2476784643071387e-06, "loss": 0.0007, "step": 10501 }, { "epoch": 1.26, "learning_rate": 2.247606478704259e-06, "loss": 0.0015, "step": 10502 }, { "epoch": 1.26, "learning_rate": 2.24753449310138e-06, "loss": 0.0007, "step": 10503 }, { "epoch": 1.26, "learning_rate": 2.2474625074985e-06, "loss": 0.0005, "step": 10504 }, { "epoch": 1.26, "learning_rate": 2.247390521895621e-06, "loss": 0.0011, "step": 10505 }, { "epoch": 1.26, "learning_rate": 2.2473185362927413e-06, "loss": 0.6739, "step": 10506 }, { "epoch": 1.26, "learning_rate": 2.247246550689862e-06, "loss": 0.0006, "step": 10507 }, { "epoch": 1.26, "learning_rate": 2.2471745650869825e-06, "loss": 0.0009, "step": 10508 }, { "epoch": 1.26, "learning_rate": 2.2471025794841032e-06, "loss": 0.118, "step": 10509 }, { "epoch": 1.26, "learning_rate": 2.2470305938812236e-06, "loss": 0.0004, "step": 10510 }, { "epoch": 1.26, "learning_rate": 2.2469586082783444e-06, "loss": 0.0007, "step": 10511 }, { "epoch": 1.26, "learning_rate": 2.246886622675465e-06, "loss": 0.0007, "step": 10512 }, { "epoch": 1.26, "learning_rate": 2.2468146370725855e-06, "loss": 0.0004, "step": 10513 }, { "epoch": 1.26, "learning_rate": 2.2467426514697063e-06, "loss": 0.3772, "step": 10514 }, { "epoch": 1.26, "learning_rate": 2.2466706658668266e-06, "loss": 0.0009, "step": 10515 }, { "epoch": 1.26, "learning_rate": 2.2465986802639474e-06, "loss": 0.0011, "step": 10516 }, { "epoch": 1.26, "learning_rate": 2.2465266946610678e-06, "loss": 0.0007, "step": 10517 }, { "epoch": 1.26, "learning_rate": 2.2464547090581886e-06, "loss": 0.0008, "step": 10518 }, { "epoch": 1.26, "learning_rate": 2.246382723455309e-06, "loss": 0.0003, "step": 10519 }, { "epoch": 1.26, "learning_rate": 2.2463107378524297e-06, "loss": 0.0007, "step": 10520 }, { "epoch": 1.26, "learning_rate": 2.24623875224955e-06, "loss": 0.0005, "step": 10521 }, { "epoch": 1.26, "learning_rate": 2.246166766646671e-06, "loss": 0.0009, "step": 10522 }, { "epoch": 1.26, "learning_rate": 2.246094781043791e-06, "loss": 0.0015, "step": 10523 }, { "epoch": 1.26, "learning_rate": 2.2460227954409115e-06, "loss": 0.0009, "step": 10524 }, { "epoch": 1.26, "learning_rate": 2.2459508098380323e-06, "loss": 0.0013, "step": 10525 }, { "epoch": 1.26, "learning_rate": 2.245878824235153e-06, "loss": 0.0005, "step": 10526 }, { "epoch": 1.26, "learning_rate": 2.2458068386322734e-06, "loss": 0.0007, "step": 10527 }, { "epoch": 1.26, "learning_rate": 2.2457348530293942e-06, "loss": 0.0005, "step": 10528 }, { "epoch": 1.26, "learning_rate": 2.245662867426515e-06, "loss": 0.001, "step": 10529 }, { "epoch": 1.26, "learning_rate": 2.2455908818236354e-06, "loss": 0.0006, "step": 10530 }, { "epoch": 1.26, "learning_rate": 2.245518896220756e-06, "loss": 0.0006, "step": 10531 }, { "epoch": 1.26, "learning_rate": 2.2454469106178765e-06, "loss": 0.0019, "step": 10532 }, { "epoch": 1.26, "learning_rate": 2.2453749250149973e-06, "loss": 0.0018, "step": 10533 }, { "epoch": 1.26, "learning_rate": 2.2453029394121176e-06, "loss": 0.3994, "step": 10534 }, { "epoch": 1.26, "learning_rate": 2.2452309538092384e-06, "loss": 0.0014, "step": 10535 }, { "epoch": 1.26, "learning_rate": 2.2451589682063588e-06, "loss": 0.0016, "step": 10536 }, { "epoch": 1.26, "learning_rate": 2.245086982603479e-06, "loss": 0.0017, "step": 10537 }, { "epoch": 1.26, "learning_rate": 2.2450149970006e-06, "loss": 0.0007, "step": 10538 }, { "epoch": 1.26, "learning_rate": 2.2449430113977202e-06, "loss": 0.0003, "step": 10539 }, { "epoch": 1.26, "learning_rate": 2.244871025794841e-06, "loss": 0.0008, "step": 10540 }, { "epoch": 1.26, "learning_rate": 2.2447990401919614e-06, "loss": 0.0021, "step": 10541 }, { "epoch": 1.26, "learning_rate": 2.244727054589082e-06, "loss": 0.0008, "step": 10542 }, { "epoch": 1.26, "learning_rate": 2.244655068986203e-06, "loss": 0.0019, "step": 10543 }, { "epoch": 1.26, "learning_rate": 2.2445830833833237e-06, "loss": 0.0004, "step": 10544 }, { "epoch": 1.26, "learning_rate": 2.244511097780444e-06, "loss": 0.0026, "step": 10545 }, { "epoch": 1.26, "learning_rate": 2.244439112177565e-06, "loss": 0.0008, "step": 10546 }, { "epoch": 1.26, "learning_rate": 2.244367126574685e-06, "loss": 0.0006, "step": 10547 }, { "epoch": 1.26, "learning_rate": 2.2442951409718056e-06, "loss": 0.0011, "step": 10548 }, { "epoch": 1.26, "learning_rate": 2.2442231553689263e-06, "loss": 0.0015, "step": 10549 }, { "epoch": 1.26, "learning_rate": 2.2441511697660467e-06, "loss": 0.0006, "step": 10550 }, { "epoch": 1.26, "learning_rate": 2.2440791841631675e-06, "loss": 0.0006, "step": 10551 }, { "epoch": 1.26, "learning_rate": 2.244007198560288e-06, "loss": 0.001, "step": 10552 }, { "epoch": 1.26, "learning_rate": 2.2439352129574086e-06, "loss": 0.0007, "step": 10553 }, { "epoch": 1.26, "learning_rate": 2.243863227354529e-06, "loss": 0.0011, "step": 10554 }, { "epoch": 1.26, "learning_rate": 2.2437912417516497e-06, "loss": 0.0006, "step": 10555 }, { "epoch": 1.26, "learning_rate": 2.24371925614877e-06, "loss": 0.0007, "step": 10556 }, { "epoch": 1.27, "learning_rate": 2.243647270545891e-06, "loss": 0.0004, "step": 10557 }, { "epoch": 1.27, "learning_rate": 2.2435752849430116e-06, "loss": 0.0006, "step": 10558 }, { "epoch": 1.27, "learning_rate": 2.243503299340132e-06, "loss": 0.0007, "step": 10559 }, { "epoch": 1.27, "learning_rate": 2.2434313137372528e-06, "loss": 0.0013, "step": 10560 }, { "epoch": 1.27, "learning_rate": 2.243359328134373e-06, "loss": 0.0009, "step": 10561 }, { "epoch": 1.27, "learning_rate": 2.243287342531494e-06, "loss": 0.0006, "step": 10562 }, { "epoch": 1.27, "learning_rate": 2.2432153569286143e-06, "loss": 0.0013, "step": 10563 }, { "epoch": 1.27, "learning_rate": 2.243143371325735e-06, "loss": 0.0009, "step": 10564 }, { "epoch": 1.27, "learning_rate": 2.2430713857228554e-06, "loss": 0.0004, "step": 10565 }, { "epoch": 1.27, "learning_rate": 2.242999400119976e-06, "loss": 0.0004, "step": 10566 }, { "epoch": 1.27, "learning_rate": 2.2429274145170965e-06, "loss": 0.6792, "step": 10567 }, { "epoch": 1.27, "learning_rate": 2.2428554289142173e-06, "loss": 0.0006, "step": 10568 }, { "epoch": 1.27, "learning_rate": 2.2427834433113377e-06, "loss": 0.0032, "step": 10569 }, { "epoch": 1.27, "learning_rate": 2.242711457708458e-06, "loss": 0.0004, "step": 10570 }, { "epoch": 1.27, "learning_rate": 2.242639472105579e-06, "loss": 0.0031, "step": 10571 }, { "epoch": 1.27, "learning_rate": 2.2425674865026996e-06, "loss": 0.0004, "step": 10572 }, { "epoch": 1.27, "learning_rate": 2.24249550089982e-06, "loss": 0.0166, "step": 10573 }, { "epoch": 1.27, "learning_rate": 2.2424235152969407e-06, "loss": 0.4326, "step": 10574 }, { "epoch": 1.27, "learning_rate": 2.2423515296940615e-06, "loss": 0.0004, "step": 10575 }, { "epoch": 1.27, "learning_rate": 2.242279544091182e-06, "loss": 0.0008, "step": 10576 }, { "epoch": 1.27, "learning_rate": 2.2422075584883026e-06, "loss": 0.0074, "step": 10577 }, { "epoch": 1.27, "learning_rate": 2.242135572885423e-06, "loss": 0.0005, "step": 10578 }, { "epoch": 1.27, "learning_rate": 2.2420635872825438e-06, "loss": 0.0004, "step": 10579 }, { "epoch": 1.27, "learning_rate": 2.241991601679664e-06, "loss": 0.0006, "step": 10580 }, { "epoch": 1.27, "learning_rate": 2.241919616076785e-06, "loss": 0.0005, "step": 10581 }, { "epoch": 1.27, "learning_rate": 2.2418476304739053e-06, "loss": 0.0008, "step": 10582 }, { "epoch": 1.27, "learning_rate": 2.2417756448710256e-06, "loss": 0.0004, "step": 10583 }, { "epoch": 1.27, "learning_rate": 2.2417036592681464e-06, "loss": 0.0008, "step": 10584 }, { "epoch": 1.27, "learning_rate": 2.2416316736652667e-06, "loss": 0.0004, "step": 10585 }, { "epoch": 1.27, "learning_rate": 2.2415596880623875e-06, "loss": 0.0003, "step": 10586 }, { "epoch": 1.27, "learning_rate": 2.241487702459508e-06, "loss": 0.0017, "step": 10587 }, { "epoch": 1.27, "learning_rate": 2.2414157168566287e-06, "loss": 0.0006, "step": 10588 }, { "epoch": 1.27, "learning_rate": 2.2413437312537494e-06, "loss": 0.0219, "step": 10589 }, { "epoch": 1.27, "learning_rate": 2.24127174565087e-06, "loss": 0.0011, "step": 10590 }, { "epoch": 1.27, "learning_rate": 2.2411997600479906e-06, "loss": 0.0004, "step": 10591 }, { "epoch": 1.27, "learning_rate": 2.2411277744451113e-06, "loss": 0.0003, "step": 10592 }, { "epoch": 1.27, "learning_rate": 2.2410557888422317e-06, "loss": 0.0006, "step": 10593 }, { "epoch": 1.27, "learning_rate": 2.240983803239352e-06, "loss": 0.0002, "step": 10594 }, { "epoch": 1.27, "learning_rate": 2.240911817636473e-06, "loss": 0.0002, "step": 10595 }, { "epoch": 1.27, "learning_rate": 2.240839832033593e-06, "loss": 0.0004, "step": 10596 }, { "epoch": 1.27, "learning_rate": 2.240767846430714e-06, "loss": 0.0004, "step": 10597 }, { "epoch": 1.27, "learning_rate": 2.2406958608278343e-06, "loss": 0.0008, "step": 10598 }, { "epoch": 1.27, "learning_rate": 2.240623875224955e-06, "loss": 0.0004, "step": 10599 }, { "epoch": 1.27, "learning_rate": 2.2405518896220755e-06, "loss": 0.0004, "step": 10600 }, { "epoch": 1.27, "learning_rate": 2.2404799040191962e-06, "loss": 0.0004, "step": 10601 }, { "epoch": 1.27, "learning_rate": 2.2404079184163166e-06, "loss": 0.0004, "step": 10602 }, { "epoch": 1.27, "learning_rate": 2.2403359328134374e-06, "loss": 0.0002, "step": 10603 }, { "epoch": 1.27, "learning_rate": 2.240263947210558e-06, "loss": 0.7638, "step": 10604 }, { "epoch": 1.27, "learning_rate": 2.2401919616076785e-06, "loss": 0.0008, "step": 10605 }, { "epoch": 1.27, "learning_rate": 2.2401199760047993e-06, "loss": 0.0003, "step": 10606 }, { "epoch": 1.27, "learning_rate": 2.2400479904019196e-06, "loss": 0.0003, "step": 10607 }, { "epoch": 1.27, "learning_rate": 2.2399760047990404e-06, "loss": 0.0003, "step": 10608 }, { "epoch": 1.27, "learning_rate": 2.2399040191961608e-06, "loss": 0.5373, "step": 10609 }, { "epoch": 1.27, "learning_rate": 2.2398320335932815e-06, "loss": 0.0006, "step": 10610 }, { "epoch": 1.27, "learning_rate": 2.239760047990402e-06, "loss": 0.0003, "step": 10611 }, { "epoch": 1.27, "learning_rate": 2.2396880623875227e-06, "loss": 0.0005, "step": 10612 }, { "epoch": 1.27, "learning_rate": 2.239616076784643e-06, "loss": 0.0003, "step": 10613 }, { "epoch": 1.27, "learning_rate": 2.239544091181764e-06, "loss": 0.7969, "step": 10614 }, { "epoch": 1.27, "learning_rate": 2.239472105578884e-06, "loss": 0.0004, "step": 10615 }, { "epoch": 1.27, "learning_rate": 2.2394001199760045e-06, "loss": 0.0031, "step": 10616 }, { "epoch": 1.27, "learning_rate": 2.2393281343731253e-06, "loss": 0.001, "step": 10617 }, { "epoch": 1.27, "learning_rate": 2.239256148770246e-06, "loss": 0.0006, "step": 10618 }, { "epoch": 1.27, "learning_rate": 2.2391841631673664e-06, "loss": 0.0037, "step": 10619 }, { "epoch": 1.27, "learning_rate": 2.2391121775644872e-06, "loss": 0.0009, "step": 10620 }, { "epoch": 1.27, "learning_rate": 2.239040191961608e-06, "loss": 0.0004, "step": 10621 }, { "epoch": 1.27, "learning_rate": 2.2389682063587283e-06, "loss": 0.0004, "step": 10622 }, { "epoch": 1.27, "learning_rate": 2.238896220755849e-06, "loss": 0.0005, "step": 10623 }, { "epoch": 1.27, "learning_rate": 2.2388242351529695e-06, "loss": 0.0005, "step": 10624 }, { "epoch": 1.27, "learning_rate": 2.2387522495500903e-06, "loss": 0.0002, "step": 10625 }, { "epoch": 1.27, "learning_rate": 2.2386802639472106e-06, "loss": 0.0007, "step": 10626 }, { "epoch": 1.27, "learning_rate": 2.2386082783443314e-06, "loss": 0.0003, "step": 10627 }, { "epoch": 1.27, "learning_rate": 2.2385362927414517e-06, "loss": 0.1224, "step": 10628 }, { "epoch": 1.27, "learning_rate": 2.238464307138572e-06, "loss": 0.0006, "step": 10629 }, { "epoch": 1.27, "learning_rate": 2.238392321535693e-06, "loss": 0.7084, "step": 10630 }, { "epoch": 1.27, "learning_rate": 2.2383203359328132e-06, "loss": 0.0009, "step": 10631 }, { "epoch": 1.27, "learning_rate": 2.238248350329934e-06, "loss": 0.0003, "step": 10632 }, { "epoch": 1.27, "learning_rate": 2.2381763647270544e-06, "loss": 0.0004, "step": 10633 }, { "epoch": 1.27, "learning_rate": 2.238104379124175e-06, "loss": 0.0004, "step": 10634 }, { "epoch": 1.27, "learning_rate": 2.238032393521296e-06, "loss": 0.0002, "step": 10635 }, { "epoch": 1.27, "learning_rate": 2.2379604079184167e-06, "loss": 0.0002, "step": 10636 }, { "epoch": 1.27, "learning_rate": 2.237888422315537e-06, "loss": 0.0004, "step": 10637 }, { "epoch": 1.27, "learning_rate": 2.237816436712658e-06, "loss": 0.0007, "step": 10638 }, { "epoch": 1.27, "learning_rate": 2.237744451109778e-06, "loss": 0.0006, "step": 10639 }, { "epoch": 1.28, "learning_rate": 2.2376724655068985e-06, "loss": 0.3807, "step": 10640 }, { "epoch": 1.28, "learning_rate": 2.2376004799040193e-06, "loss": 0.0007, "step": 10641 }, { "epoch": 1.28, "learning_rate": 2.2375284943011397e-06, "loss": 0.0003, "step": 10642 }, { "epoch": 1.28, "learning_rate": 2.2374565086982605e-06, "loss": 0.0002, "step": 10643 }, { "epoch": 1.28, "learning_rate": 2.237384523095381e-06, "loss": 0.0009, "step": 10644 }, { "epoch": 1.28, "learning_rate": 2.2373125374925016e-06, "loss": 0.0004, "step": 10645 }, { "epoch": 1.28, "learning_rate": 2.237240551889622e-06, "loss": 0.009, "step": 10646 }, { "epoch": 1.28, "learning_rate": 2.2371685662867427e-06, "loss": 0.0006, "step": 10647 }, { "epoch": 1.28, "learning_rate": 2.237096580683863e-06, "loss": 0.0008, "step": 10648 }, { "epoch": 1.28, "learning_rate": 2.237024595080984e-06, "loss": 0.0136, "step": 10649 }, { "epoch": 1.28, "learning_rate": 2.2369526094781042e-06, "loss": 0.0003, "step": 10650 }, { "epoch": 1.28, "learning_rate": 2.236880623875225e-06, "loss": 0.0004, "step": 10651 }, { "epoch": 1.28, "learning_rate": 2.2368086382723458e-06, "loss": 0.0003, "step": 10652 }, { "epoch": 1.28, "learning_rate": 2.236736652669466e-06, "loss": 0.0003, "step": 10653 }, { "epoch": 1.28, "learning_rate": 2.236664667066587e-06, "loss": 0.0009, "step": 10654 }, { "epoch": 1.28, "learning_rate": 2.2365926814637073e-06, "loss": 0.0003, "step": 10655 }, { "epoch": 1.28, "learning_rate": 2.236520695860828e-06, "loss": 0.0005, "step": 10656 }, { "epoch": 1.28, "learning_rate": 2.2364487102579484e-06, "loss": 0.0003, "step": 10657 }, { "epoch": 1.28, "learning_rate": 2.236376724655069e-06, "loss": 0.0006, "step": 10658 }, { "epoch": 1.28, "learning_rate": 2.2363047390521895e-06, "loss": 0.0005, "step": 10659 }, { "epoch": 1.28, "learning_rate": 2.2362327534493103e-06, "loss": 0.0003, "step": 10660 }, { "epoch": 1.28, "learning_rate": 2.2361607678464307e-06, "loss": 0.0006, "step": 10661 }, { "epoch": 1.28, "learning_rate": 2.2360887822435514e-06, "loss": 0.7162, "step": 10662 }, { "epoch": 1.28, "learning_rate": 2.236016796640672e-06, "loss": 0.0005, "step": 10663 }, { "epoch": 1.28, "learning_rate": 2.235944811037792e-06, "loss": 0.0006, "step": 10664 }, { "epoch": 1.28, "learning_rate": 2.235872825434913e-06, "loss": 0.0005, "step": 10665 }, { "epoch": 1.28, "learning_rate": 2.2358008398320337e-06, "loss": 0.0003, "step": 10666 }, { "epoch": 1.28, "learning_rate": 2.2357288542291545e-06, "loss": 0.0002, "step": 10667 }, { "epoch": 1.28, "learning_rate": 2.235656868626275e-06, "loss": 0.0003, "step": 10668 }, { "epoch": 1.28, "learning_rate": 2.2355848830233956e-06, "loss": 0.001, "step": 10669 }, { "epoch": 1.28, "learning_rate": 2.235512897420516e-06, "loss": 0.0005, "step": 10670 }, { "epoch": 1.28, "learning_rate": 2.2354409118176368e-06, "loss": 0.7961, "step": 10671 }, { "epoch": 1.28, "learning_rate": 2.235368926214757e-06, "loss": 0.0004, "step": 10672 }, { "epoch": 1.28, "learning_rate": 2.235296940611878e-06, "loss": 0.0039, "step": 10673 }, { "epoch": 1.28, "learning_rate": 2.2352249550089982e-06, "loss": 0.0005, "step": 10674 }, { "epoch": 1.28, "learning_rate": 2.2351529694061186e-06, "loss": 0.0008, "step": 10675 }, { "epoch": 1.28, "learning_rate": 2.2350809838032394e-06, "loss": 0.0006, "step": 10676 }, { "epoch": 1.28, "learning_rate": 2.2350089982003597e-06, "loss": 0.0005, "step": 10677 }, { "epoch": 1.28, "learning_rate": 2.2349370125974805e-06, "loss": 0.0004, "step": 10678 }, { "epoch": 1.28, "learning_rate": 2.234865026994601e-06, "loss": 0.0004, "step": 10679 }, { "epoch": 1.28, "learning_rate": 2.2347930413917216e-06, "loss": 0.0006, "step": 10680 }, { "epoch": 1.28, "learning_rate": 2.2347210557888424e-06, "loss": 0.0004, "step": 10681 }, { "epoch": 1.28, "learning_rate": 2.2346490701859628e-06, "loss": 0.8671, "step": 10682 }, { "epoch": 1.28, "learning_rate": 2.2345770845830836e-06, "loss": 0.0003, "step": 10683 }, { "epoch": 1.28, "learning_rate": 2.2345050989802043e-06, "loss": 0.365, "step": 10684 }, { "epoch": 1.28, "learning_rate": 2.2344331133773247e-06, "loss": 0.0003, "step": 10685 }, { "epoch": 1.28, "learning_rate": 2.234361127774445e-06, "loss": 0.0003, "step": 10686 }, { "epoch": 1.28, "learning_rate": 2.234289142171566e-06, "loss": 1.1025, "step": 10687 }, { "epoch": 1.28, "learning_rate": 2.234217156568686e-06, "loss": 0.0003, "step": 10688 }, { "epoch": 1.28, "learning_rate": 2.234145170965807e-06, "loss": 0.0004, "step": 10689 }, { "epoch": 1.28, "learning_rate": 2.2340731853629273e-06, "loss": 0.0008, "step": 10690 }, { "epoch": 1.28, "learning_rate": 2.234001199760048e-06, "loss": 0.0004, "step": 10691 }, { "epoch": 1.28, "learning_rate": 2.2339292141571684e-06, "loss": 0.0005, "step": 10692 }, { "epoch": 1.28, "learning_rate": 2.2338572285542892e-06, "loss": 0.0005, "step": 10693 }, { "epoch": 1.28, "learning_rate": 2.2337852429514096e-06, "loss": 0.0006, "step": 10694 }, { "epoch": 1.28, "learning_rate": 2.2337132573485304e-06, "loss": 0.0006, "step": 10695 }, { "epoch": 1.28, "learning_rate": 2.2336412717456507e-06, "loss": 0.0009, "step": 10696 }, { "epoch": 1.28, "learning_rate": 2.2335692861427715e-06, "loss": 0.0003, "step": 10697 }, { "epoch": 1.28, "learning_rate": 2.2334973005398923e-06, "loss": 0.0004, "step": 10698 }, { "epoch": 1.28, "learning_rate": 2.2334253149370126e-06, "loss": 0.0004, "step": 10699 }, { "epoch": 1.28, "learning_rate": 2.2333533293341334e-06, "loss": 0.0004, "step": 10700 }, { "epoch": 1.28, "learning_rate": 2.2332813437312538e-06, "loss": 0.0008, "step": 10701 }, { "epoch": 1.28, "learning_rate": 2.2332093581283745e-06, "loss": 0.0004, "step": 10702 }, { "epoch": 1.28, "learning_rate": 2.233137372525495e-06, "loss": 0.0005, "step": 10703 }, { "epoch": 1.28, "learning_rate": 2.2330653869226157e-06, "loss": 0.0005, "step": 10704 }, { "epoch": 1.28, "learning_rate": 2.232993401319736e-06, "loss": 0.0014, "step": 10705 }, { "epoch": 1.28, "learning_rate": 2.232921415716857e-06, "loss": 0.0257, "step": 10706 }, { "epoch": 1.28, "learning_rate": 2.232849430113977e-06, "loss": 0.0007, "step": 10707 }, { "epoch": 1.28, "learning_rate": 2.232777444511098e-06, "loss": 0.0004, "step": 10708 }, { "epoch": 1.28, "learning_rate": 2.2327054589082183e-06, "loss": 0.0007, "step": 10709 }, { "epoch": 1.28, "learning_rate": 2.2326334733053386e-06, "loss": 0.0007, "step": 10710 }, { "epoch": 1.28, "learning_rate": 2.2325614877024594e-06, "loss": 0.0183, "step": 10711 }, { "epoch": 1.28, "learning_rate": 2.23248950209958e-06, "loss": 0.0004, "step": 10712 }, { "epoch": 1.28, "learning_rate": 2.232417516496701e-06, "loss": 0.0003, "step": 10713 }, { "epoch": 1.28, "learning_rate": 2.2323455308938213e-06, "loss": 0.0005, "step": 10714 }, { "epoch": 1.28, "learning_rate": 2.232273545290942e-06, "loss": 0.0004, "step": 10715 }, { "epoch": 1.28, "learning_rate": 2.2322015596880625e-06, "loss": 0.0009, "step": 10716 }, { "epoch": 1.28, "learning_rate": 2.2321295740851833e-06, "loss": 0.0008, "step": 10717 }, { "epoch": 1.28, "learning_rate": 2.2320575884823036e-06, "loss": 0.0006, "step": 10718 }, { "epoch": 1.28, "learning_rate": 2.2319856028794244e-06, "loss": 0.0003, "step": 10719 }, { "epoch": 1.28, "learning_rate": 2.2319136172765447e-06, "loss": 0.9068, "step": 10720 }, { "epoch": 1.28, "learning_rate": 2.231841631673665e-06, "loss": 0.0006, "step": 10721 }, { "epoch": 1.28, "learning_rate": 2.231769646070786e-06, "loss": 0.8689, "step": 10722 }, { "epoch": 1.28, "learning_rate": 2.2316976604679062e-06, "loss": 0.0013, "step": 10723 }, { "epoch": 1.29, "learning_rate": 2.231625674865027e-06, "loss": 0.0002, "step": 10724 }, { "epoch": 1.29, "learning_rate": 2.2315536892621474e-06, "loss": 0.0005, "step": 10725 }, { "epoch": 1.29, "learning_rate": 2.231481703659268e-06, "loss": 0.0006, "step": 10726 }, { "epoch": 1.29, "learning_rate": 2.231409718056389e-06, "loss": 0.0006, "step": 10727 }, { "epoch": 1.29, "learning_rate": 2.2313377324535093e-06, "loss": 0.0005, "step": 10728 }, { "epoch": 1.29, "learning_rate": 2.23126574685063e-06, "loss": 0.0004, "step": 10729 }, { "epoch": 1.29, "learning_rate": 2.231193761247751e-06, "loss": 0.0003, "step": 10730 }, { "epoch": 1.29, "learning_rate": 2.231121775644871e-06, "loss": 0.0004, "step": 10731 }, { "epoch": 1.29, "learning_rate": 2.2310497900419915e-06, "loss": 0.0012, "step": 10732 }, { "epoch": 1.29, "learning_rate": 2.2309778044391123e-06, "loss": 0.0011, "step": 10733 }, { "epoch": 1.29, "learning_rate": 2.2309058188362327e-06, "loss": 0.0005, "step": 10734 }, { "epoch": 1.29, "learning_rate": 2.2308338332333535e-06, "loss": 0.6544, "step": 10735 }, { "epoch": 1.29, "learning_rate": 2.230761847630474e-06, "loss": 0.0005, "step": 10736 }, { "epoch": 1.29, "learning_rate": 2.2306898620275946e-06, "loss": 0.0464, "step": 10737 }, { "epoch": 1.29, "learning_rate": 2.230617876424715e-06, "loss": 0.0008, "step": 10738 }, { "epoch": 1.29, "learning_rate": 2.2305458908218357e-06, "loss": 0.0002, "step": 10739 }, { "epoch": 1.29, "learning_rate": 2.230473905218956e-06, "loss": 0.098, "step": 10740 }, { "epoch": 1.29, "learning_rate": 2.230401919616077e-06, "loss": 0.0004, "step": 10741 }, { "epoch": 1.29, "learning_rate": 2.230329934013197e-06, "loss": 0.0021, "step": 10742 }, { "epoch": 1.29, "learning_rate": 2.230257948410318e-06, "loss": 0.0004, "step": 10743 }, { "epoch": 1.29, "learning_rate": 2.2301859628074388e-06, "loss": 0.0004, "step": 10744 }, { "epoch": 1.29, "learning_rate": 2.230113977204559e-06, "loss": 0.0004, "step": 10745 }, { "epoch": 1.29, "learning_rate": 2.23004199160168e-06, "loss": 0.0027, "step": 10746 }, { "epoch": 1.29, "learning_rate": 2.2299700059988003e-06, "loss": 0.0004, "step": 10747 }, { "epoch": 1.29, "learning_rate": 2.229898020395921e-06, "loss": 0.0008, "step": 10748 }, { "epoch": 1.29, "learning_rate": 2.2298260347930414e-06, "loss": 0.0113, "step": 10749 }, { "epoch": 1.29, "learning_rate": 2.229754049190162e-06, "loss": 0.0003, "step": 10750 }, { "epoch": 1.29, "learning_rate": 2.2296820635872825e-06, "loss": 0.0003, "step": 10751 }, { "epoch": 1.29, "learning_rate": 2.2296100779844033e-06, "loss": 0.0004, "step": 10752 }, { "epoch": 1.29, "learning_rate": 2.2295380923815237e-06, "loss": 0.0004, "step": 10753 }, { "epoch": 1.29, "learning_rate": 2.2294661067786444e-06, "loss": 0.0005, "step": 10754 }, { "epoch": 1.29, "learning_rate": 2.229394121175765e-06, "loss": 0.0005, "step": 10755 }, { "epoch": 1.29, "learning_rate": 2.229322135572885e-06, "loss": 0.2466, "step": 10756 }, { "epoch": 1.29, "learning_rate": 2.229250149970006e-06, "loss": 0.0014, "step": 10757 }, { "epoch": 1.29, "learning_rate": 2.2291781643671267e-06, "loss": 0.0007, "step": 10758 }, { "epoch": 1.29, "learning_rate": 2.2291061787642475e-06, "loss": 0.0005, "step": 10759 }, { "epoch": 1.29, "learning_rate": 2.229034193161368e-06, "loss": 0.9508, "step": 10760 }, { "epoch": 1.29, "learning_rate": 2.2289622075584886e-06, "loss": 0.0005, "step": 10761 }, { "epoch": 1.29, "learning_rate": 2.228890221955609e-06, "loss": 0.0005, "step": 10762 }, { "epoch": 1.29, "learning_rate": 2.2288182363527297e-06, "loss": 0.0004, "step": 10763 }, { "epoch": 1.29, "learning_rate": 2.22874625074985e-06, "loss": 0.0008, "step": 10764 }, { "epoch": 1.29, "learning_rate": 2.228674265146971e-06, "loss": 0.0004, "step": 10765 }, { "epoch": 1.29, "learning_rate": 2.2286022795440912e-06, "loss": 0.0005, "step": 10766 }, { "epoch": 1.29, "learning_rate": 2.2285302939412116e-06, "loss": 0.0004, "step": 10767 }, { "epoch": 1.29, "learning_rate": 2.2284583083383324e-06, "loss": 0.0005, "step": 10768 }, { "epoch": 1.29, "learning_rate": 2.2283863227354527e-06, "loss": 0.0005, "step": 10769 }, { "epoch": 1.29, "learning_rate": 2.2283143371325735e-06, "loss": 0.0007, "step": 10770 }, { "epoch": 1.29, "learning_rate": 2.228242351529694e-06, "loss": 0.0003, "step": 10771 }, { "epoch": 1.29, "learning_rate": 2.2281703659268146e-06, "loss": 0.0008, "step": 10772 }, { "epoch": 1.29, "learning_rate": 2.228098380323935e-06, "loss": 0.0006, "step": 10773 }, { "epoch": 1.29, "learning_rate": 2.2280263947210558e-06, "loss": 0.0005, "step": 10774 }, { "epoch": 1.29, "learning_rate": 2.2279544091181765e-06, "loss": 0.0003, "step": 10775 }, { "epoch": 1.29, "learning_rate": 2.2278824235152973e-06, "loss": 0.0005, "step": 10776 }, { "epoch": 1.29, "learning_rate": 2.2278104379124177e-06, "loss": 0.0003, "step": 10777 }, { "epoch": 1.29, "learning_rate": 2.227738452309538e-06, "loss": 0.0006, "step": 10778 }, { "epoch": 1.29, "learning_rate": 2.227666466706659e-06, "loss": 0.0003, "step": 10779 }, { "epoch": 1.29, "learning_rate": 2.227594481103779e-06, "loss": 0.0007, "step": 10780 }, { "epoch": 1.29, "learning_rate": 2.2275224955009e-06, "loss": 0.0004, "step": 10781 }, { "epoch": 1.29, "learning_rate": 2.2274505098980203e-06, "loss": 0.0004, "step": 10782 }, { "epoch": 1.29, "learning_rate": 2.227378524295141e-06, "loss": 0.0008, "step": 10783 }, { "epoch": 1.29, "learning_rate": 2.2273065386922614e-06, "loss": 0.717, "step": 10784 }, { "epoch": 1.29, "learning_rate": 2.2272345530893822e-06, "loss": 0.1531, "step": 10785 }, { "epoch": 1.29, "learning_rate": 2.2271625674865026e-06, "loss": 0.0005, "step": 10786 }, { "epoch": 1.29, "learning_rate": 2.2270905818836234e-06, "loss": 0.6923, "step": 10787 }, { "epoch": 1.29, "learning_rate": 2.2270185962807437e-06, "loss": 0.0007, "step": 10788 }, { "epoch": 1.29, "learning_rate": 2.2269466106778645e-06, "loss": 0.0005, "step": 10789 }, { "epoch": 1.29, "learning_rate": 2.2268746250749853e-06, "loss": 0.0002, "step": 10790 }, { "epoch": 1.29, "learning_rate": 2.2268026394721056e-06, "loss": 0.0003, "step": 10791 }, { "epoch": 1.29, "learning_rate": 2.2267306538692264e-06, "loss": 0.0004, "step": 10792 }, { "epoch": 1.29, "learning_rate": 2.2266586682663468e-06, "loss": 0.0006, "step": 10793 }, { "epoch": 1.29, "learning_rate": 2.2265866826634675e-06, "loss": 0.0006, "step": 10794 }, { "epoch": 1.29, "learning_rate": 2.226514697060588e-06, "loss": 0.0006, "step": 10795 }, { "epoch": 1.29, "learning_rate": 2.2264427114577087e-06, "loss": 0.0008, "step": 10796 }, { "epoch": 1.29, "learning_rate": 2.226370725854829e-06, "loss": 0.0004, "step": 10797 }, { "epoch": 1.29, "learning_rate": 2.22629874025195e-06, "loss": 0.0279, "step": 10798 }, { "epoch": 1.29, "learning_rate": 2.22622675464907e-06, "loss": 0.0003, "step": 10799 }, { "epoch": 1.29, "learning_rate": 2.226154769046191e-06, "loss": 0.4107, "step": 10800 }, { "epoch": 1.29, "learning_rate": 2.2260827834433113e-06, "loss": 0.0008, "step": 10801 }, { "epoch": 1.29, "learning_rate": 2.2260107978404316e-06, "loss": 0.0019, "step": 10802 }, { "epoch": 1.29, "learning_rate": 2.2259388122375524e-06, "loss": 0.0003, "step": 10803 }, { "epoch": 1.29, "learning_rate": 2.225866826634673e-06, "loss": 0.0009, "step": 10804 }, { "epoch": 1.29, "learning_rate": 2.2257948410317936e-06, "loss": 0.0005, "step": 10805 }, { "epoch": 1.29, "learning_rate": 2.2257228554289143e-06, "loss": 0.0008, "step": 10806 }, { "epoch": 1.3, "learning_rate": 2.225650869826035e-06, "loss": 0.0007, "step": 10807 }, { "epoch": 1.3, "learning_rate": 2.2255788842231555e-06, "loss": 0.0012, "step": 10808 }, { "epoch": 1.3, "learning_rate": 2.2255068986202762e-06, "loss": 0.0004, "step": 10809 }, { "epoch": 1.3, "learning_rate": 2.2254349130173966e-06, "loss": 0.0004, "step": 10810 }, { "epoch": 1.3, "learning_rate": 2.2253629274145174e-06, "loss": 0.0005, "step": 10811 }, { "epoch": 1.3, "learning_rate": 2.2252909418116377e-06, "loss": 0.001, "step": 10812 }, { "epoch": 1.3, "learning_rate": 2.225218956208758e-06, "loss": 0.0007, "step": 10813 }, { "epoch": 1.3, "learning_rate": 2.225146970605879e-06, "loss": 0.0006, "step": 10814 }, { "epoch": 1.3, "learning_rate": 2.2250749850029992e-06, "loss": 0.0527, "step": 10815 }, { "epoch": 1.3, "learning_rate": 2.22500299940012e-06, "loss": 0.0002, "step": 10816 }, { "epoch": 1.3, "learning_rate": 2.2249310137972404e-06, "loss": 0.8827, "step": 10817 }, { "epoch": 1.3, "learning_rate": 2.224859028194361e-06, "loss": 0.0002, "step": 10818 }, { "epoch": 1.3, "learning_rate": 2.2247870425914815e-06, "loss": 0.0008, "step": 10819 }, { "epoch": 1.3, "learning_rate": 2.2247150569886023e-06, "loss": 0.0013, "step": 10820 }, { "epoch": 1.3, "learning_rate": 2.224643071385723e-06, "loss": 0.0003, "step": 10821 }, { "epoch": 1.3, "learning_rate": 2.224571085782844e-06, "loss": 0.0007, "step": 10822 }, { "epoch": 1.3, "learning_rate": 2.224499100179964e-06, "loss": 0.0004, "step": 10823 }, { "epoch": 1.3, "learning_rate": 2.2244271145770845e-06, "loss": 0.0008, "step": 10824 }, { "epoch": 1.3, "learning_rate": 2.2243551289742053e-06, "loss": 0.0004, "step": 10825 }, { "epoch": 1.3, "learning_rate": 2.2242831433713257e-06, "loss": 0.9711, "step": 10826 }, { "epoch": 1.3, "learning_rate": 2.2242111577684464e-06, "loss": 0.0006, "step": 10827 }, { "epoch": 1.3, "learning_rate": 2.224139172165567e-06, "loss": 0.0069, "step": 10828 }, { "epoch": 1.3, "learning_rate": 2.2240671865626876e-06, "loss": 0.0016, "step": 10829 }, { "epoch": 1.3, "learning_rate": 2.223995200959808e-06, "loss": 0.0002, "step": 10830 }, { "epoch": 1.3, "learning_rate": 2.2239232153569287e-06, "loss": 0.0476, "step": 10831 }, { "epoch": 1.3, "learning_rate": 2.223851229754049e-06, "loss": 0.0005, "step": 10832 }, { "epoch": 1.3, "learning_rate": 2.22377924415117e-06, "loss": 0.0003, "step": 10833 }, { "epoch": 1.3, "learning_rate": 2.22370725854829e-06, "loss": 0.6362, "step": 10834 }, { "epoch": 1.3, "learning_rate": 2.223635272945411e-06, "loss": 0.0004, "step": 10835 }, { "epoch": 1.3, "learning_rate": 2.2235632873425318e-06, "loss": 0.0009, "step": 10836 }, { "epoch": 1.3, "learning_rate": 2.223491301739652e-06, "loss": 0.0004, "step": 10837 }, { "epoch": 1.3, "learning_rate": 2.223419316136773e-06, "loss": 0.0004, "step": 10838 }, { "epoch": 1.3, "learning_rate": 2.2233473305338932e-06, "loss": 0.0008, "step": 10839 }, { "epoch": 1.3, "learning_rate": 2.223275344931014e-06, "loss": 0.0005, "step": 10840 }, { "epoch": 1.3, "learning_rate": 2.2232033593281344e-06, "loss": 0.0004, "step": 10841 }, { "epoch": 1.3, "learning_rate": 2.223131373725255e-06, "loss": 0.0005, "step": 10842 }, { "epoch": 1.3, "learning_rate": 2.2230593881223755e-06, "loss": 0.0013, "step": 10843 }, { "epoch": 1.3, "learning_rate": 2.2229874025194963e-06, "loss": 0.0015, "step": 10844 }, { "epoch": 1.3, "learning_rate": 2.2229154169166166e-06, "loss": 0.0004, "step": 10845 }, { "epoch": 1.3, "learning_rate": 2.2228434313137374e-06, "loss": 0.0009, "step": 10846 }, { "epoch": 1.3, "learning_rate": 2.2227714457108578e-06, "loss": 0.0007, "step": 10847 }, { "epoch": 1.3, "learning_rate": 2.222699460107978e-06, "loss": 0.0052, "step": 10848 }, { "epoch": 1.3, "learning_rate": 2.222627474505099e-06, "loss": 0.0003, "step": 10849 }, { "epoch": 1.3, "learning_rate": 2.2225554889022197e-06, "loss": 0.0005, "step": 10850 }, { "epoch": 1.3, "learning_rate": 2.22248350329934e-06, "loss": 0.0011, "step": 10851 }, { "epoch": 1.3, "learning_rate": 2.222411517696461e-06, "loss": 0.001, "step": 10852 }, { "epoch": 1.3, "learning_rate": 2.2223395320935816e-06, "loss": 0.77, "step": 10853 }, { "epoch": 1.3, "learning_rate": 2.222267546490702e-06, "loss": 0.0004, "step": 10854 }, { "epoch": 1.3, "learning_rate": 2.2221955608878227e-06, "loss": 0.0005, "step": 10855 }, { "epoch": 1.3, "learning_rate": 2.222123575284943e-06, "loss": 0.7931, "step": 10856 }, { "epoch": 1.3, "learning_rate": 2.222051589682064e-06, "loss": 0.0007, "step": 10857 }, { "epoch": 1.3, "learning_rate": 2.2219796040791842e-06, "loss": 0.0006, "step": 10858 }, { "epoch": 1.3, "learning_rate": 2.2219076184763046e-06, "loss": 0.0031, "step": 10859 }, { "epoch": 1.3, "learning_rate": 2.2218356328734254e-06, "loss": 0.0003, "step": 10860 }, { "epoch": 1.3, "learning_rate": 2.2217636472705457e-06, "loss": 0.0008, "step": 10861 }, { "epoch": 1.3, "learning_rate": 2.2216916616676665e-06, "loss": 0.0006, "step": 10862 }, { "epoch": 1.3, "learning_rate": 2.221619676064787e-06, "loss": 0.001, "step": 10863 }, { "epoch": 1.3, "learning_rate": 2.2215476904619076e-06, "loss": 0.0006, "step": 10864 }, { "epoch": 1.3, "learning_rate": 2.221475704859028e-06, "loss": 0.0009, "step": 10865 }, { "epoch": 1.3, "learning_rate": 2.2214037192561488e-06, "loss": 0.0002, "step": 10866 }, { "epoch": 1.3, "learning_rate": 2.2213317336532695e-06, "loss": 0.0019, "step": 10867 }, { "epoch": 1.3, "learning_rate": 2.2212597480503903e-06, "loss": 0.004, "step": 10868 }, { "epoch": 1.3, "learning_rate": 2.2211877624475107e-06, "loss": 0.0008, "step": 10869 }, { "epoch": 1.3, "learning_rate": 2.2211157768446315e-06, "loss": 0.0005, "step": 10870 }, { "epoch": 1.3, "learning_rate": 2.221043791241752e-06, "loss": 0.0006, "step": 10871 }, { "epoch": 1.3, "learning_rate": 2.220971805638872e-06, "loss": 0.0009, "step": 10872 }, { "epoch": 1.3, "learning_rate": 2.220899820035993e-06, "loss": 0.0011, "step": 10873 }, { "epoch": 1.3, "learning_rate": 2.2208278344331133e-06, "loss": 0.0004, "step": 10874 }, { "epoch": 1.3, "learning_rate": 2.220755848830234e-06, "loss": 0.0005, "step": 10875 }, { "epoch": 1.3, "learning_rate": 2.2206838632273544e-06, "loss": 0.0002, "step": 10876 }, { "epoch": 1.3, "learning_rate": 2.220611877624475e-06, "loss": 0.0005, "step": 10877 }, { "epoch": 1.3, "learning_rate": 2.2205398920215956e-06, "loss": 0.0004, "step": 10878 }, { "epoch": 1.3, "learning_rate": 2.2204679064187163e-06, "loss": 0.0008, "step": 10879 }, { "epoch": 1.3, "learning_rate": 2.2203959208158367e-06, "loss": 0.0018, "step": 10880 }, { "epoch": 1.3, "learning_rate": 2.2203239352129575e-06, "loss": 0.4312, "step": 10881 }, { "epoch": 1.3, "learning_rate": 2.2202519496100783e-06, "loss": 0.0013, "step": 10882 }, { "epoch": 1.3, "learning_rate": 2.2201799640071986e-06, "loss": 0.0004, "step": 10883 }, { "epoch": 1.3, "learning_rate": 2.2201079784043194e-06, "loss": 0.0005, "step": 10884 }, { "epoch": 1.3, "learning_rate": 2.2200359928014397e-06, "loss": 0.0005, "step": 10885 }, { "epoch": 1.3, "learning_rate": 2.2199640071985605e-06, "loss": 0.0003, "step": 10886 }, { "epoch": 1.3, "learning_rate": 2.219892021595681e-06, "loss": 0.0005, "step": 10887 }, { "epoch": 1.3, "learning_rate": 2.2198200359928017e-06, "loss": 0.7969, "step": 10888 }, { "epoch": 1.3, "learning_rate": 2.219748050389922e-06, "loss": 0.0003, "step": 10889 }, { "epoch": 1.3, "learning_rate": 2.219676064787043e-06, "loss": 0.0003, "step": 10890 }, { "epoch": 1.31, "learning_rate": 2.219604079184163e-06, "loss": 0.0006, "step": 10891 }, { "epoch": 1.31, "learning_rate": 2.219532093581284e-06, "loss": 0.0004, "step": 10892 }, { "epoch": 1.31, "learning_rate": 2.2194601079784043e-06, "loss": 0.0003, "step": 10893 }, { "epoch": 1.31, "learning_rate": 2.2193881223755246e-06, "loss": 0.0006, "step": 10894 }, { "epoch": 1.31, "learning_rate": 2.2193161367726454e-06, "loss": 0.0005, "step": 10895 }, { "epoch": 1.31, "learning_rate": 2.219244151169766e-06, "loss": 0.792, "step": 10896 }, { "epoch": 1.31, "learning_rate": 2.2191721655668865e-06, "loss": 0.0003, "step": 10897 }, { "epoch": 1.31, "learning_rate": 2.2191001799640073e-06, "loss": 0.0754, "step": 10898 }, { "epoch": 1.31, "learning_rate": 2.219028194361128e-06, "loss": 0.0061, "step": 10899 }, { "epoch": 1.31, "learning_rate": 2.2189562087582485e-06, "loss": 0.0006, "step": 10900 }, { "epoch": 1.31, "learning_rate": 2.2188842231553692e-06, "loss": 0.0003, "step": 10901 }, { "epoch": 1.31, "learning_rate": 2.2188122375524896e-06, "loss": 0.0006, "step": 10902 }, { "epoch": 1.31, "learning_rate": 2.2187402519496104e-06, "loss": 0.0005, "step": 10903 }, { "epoch": 1.31, "learning_rate": 2.2186682663467307e-06, "loss": 0.0004, "step": 10904 }, { "epoch": 1.31, "learning_rate": 2.218596280743851e-06, "loss": 0.0007, "step": 10905 }, { "epoch": 1.31, "learning_rate": 2.218524295140972e-06, "loss": 0.0005, "step": 10906 }, { "epoch": 1.31, "learning_rate": 2.2184523095380922e-06, "loss": 0.0004, "step": 10907 }, { "epoch": 1.31, "learning_rate": 2.218380323935213e-06, "loss": 0.0003, "step": 10908 }, { "epoch": 1.31, "learning_rate": 2.2183083383323333e-06, "loss": 0.0004, "step": 10909 }, { "epoch": 1.31, "learning_rate": 2.218236352729454e-06, "loss": 0.0087, "step": 10910 }, { "epoch": 1.31, "learning_rate": 2.2181643671265745e-06, "loss": 0.0003, "step": 10911 }, { "epoch": 1.31, "learning_rate": 2.2180923815236953e-06, "loss": 0.0002, "step": 10912 }, { "epoch": 1.31, "learning_rate": 2.218020395920816e-06, "loss": 0.0002, "step": 10913 }, { "epoch": 1.31, "learning_rate": 2.2179484103179364e-06, "loss": 0.0003, "step": 10914 }, { "epoch": 1.31, "learning_rate": 2.217876424715057e-06, "loss": 0.0004, "step": 10915 }, { "epoch": 1.31, "learning_rate": 2.217804439112178e-06, "loss": 0.0006, "step": 10916 }, { "epoch": 1.31, "learning_rate": 2.2177324535092983e-06, "loss": 0.0005, "step": 10917 }, { "epoch": 1.31, "learning_rate": 2.2176604679064187e-06, "loss": 0.0003, "step": 10918 }, { "epoch": 1.31, "learning_rate": 2.2175884823035394e-06, "loss": 0.0004, "step": 10919 }, { "epoch": 1.31, "learning_rate": 2.21751649670066e-06, "loss": 0.0004, "step": 10920 }, { "epoch": 1.31, "learning_rate": 2.2174445110977806e-06, "loss": 0.0015, "step": 10921 }, { "epoch": 1.31, "learning_rate": 2.217372525494901e-06, "loss": 0.6992, "step": 10922 }, { "epoch": 1.31, "learning_rate": 2.2173005398920217e-06, "loss": 0.0006, "step": 10923 }, { "epoch": 1.31, "learning_rate": 2.217228554289142e-06, "loss": 0.0969, "step": 10924 }, { "epoch": 1.31, "learning_rate": 2.217156568686263e-06, "loss": 0.0003, "step": 10925 }, { "epoch": 1.31, "learning_rate": 2.217084583083383e-06, "loss": 0.0004, "step": 10926 }, { "epoch": 1.31, "learning_rate": 2.217012597480504e-06, "loss": 0.0005, "step": 10927 }, { "epoch": 1.31, "learning_rate": 2.2169406118776243e-06, "loss": 0.0103, "step": 10928 }, { "epoch": 1.31, "learning_rate": 2.216868626274745e-06, "loss": 0.0003, "step": 10929 }, { "epoch": 1.31, "learning_rate": 2.216796640671866e-06, "loss": 0.4521, "step": 10930 }, { "epoch": 1.31, "learning_rate": 2.2167246550689862e-06, "loss": 0.0004, "step": 10931 }, { "epoch": 1.31, "learning_rate": 2.216652669466107e-06, "loss": 0.0004, "step": 10932 }, { "epoch": 1.31, "learning_rate": 2.2165806838632274e-06, "loss": 0.0008, "step": 10933 }, { "epoch": 1.31, "learning_rate": 2.216508698260348e-06, "loss": 0.0006, "step": 10934 }, { "epoch": 1.31, "learning_rate": 2.2164367126574685e-06, "loss": 0.0246, "step": 10935 }, { "epoch": 1.31, "learning_rate": 2.2163647270545893e-06, "loss": 0.0004, "step": 10936 }, { "epoch": 1.31, "learning_rate": 2.2162927414517096e-06, "loss": 0.0004, "step": 10937 }, { "epoch": 1.31, "learning_rate": 2.2162207558488304e-06, "loss": 0.0003, "step": 10938 }, { "epoch": 1.31, "learning_rate": 2.2161487702459508e-06, "loss": 0.0004, "step": 10939 }, { "epoch": 1.31, "learning_rate": 2.216076784643071e-06, "loss": 0.0006, "step": 10940 }, { "epoch": 1.31, "learning_rate": 2.216004799040192e-06, "loss": 0.0007, "step": 10941 }, { "epoch": 1.31, "learning_rate": 2.2159328134373123e-06, "loss": 0.0003, "step": 10942 }, { "epoch": 1.31, "learning_rate": 2.215860827834433e-06, "loss": 0.0006, "step": 10943 }, { "epoch": 1.31, "learning_rate": 2.215788842231554e-06, "loss": 0.0003, "step": 10944 }, { "epoch": 1.31, "learning_rate": 2.2157168566286746e-06, "loss": 0.0006, "step": 10945 }, { "epoch": 1.31, "learning_rate": 2.215644871025795e-06, "loss": 0.0002, "step": 10946 }, { "epoch": 1.31, "learning_rate": 2.2155728854229157e-06, "loss": 0.0003, "step": 10947 }, { "epoch": 1.31, "learning_rate": 2.215500899820036e-06, "loss": 0.7542, "step": 10948 }, { "epoch": 1.31, "learning_rate": 2.215428914217157e-06, "loss": 0.0003, "step": 10949 }, { "epoch": 1.31, "learning_rate": 2.2153569286142772e-06, "loss": 0.0003, "step": 10950 }, { "epoch": 1.31, "learning_rate": 2.2152849430113976e-06, "loss": 0.0006, "step": 10951 }, { "epoch": 1.31, "learning_rate": 2.2152129574085184e-06, "loss": 0.0004, "step": 10952 }, { "epoch": 1.31, "learning_rate": 2.2151409718056387e-06, "loss": 0.0003, "step": 10953 }, { "epoch": 1.31, "learning_rate": 2.2150689862027595e-06, "loss": 0.0009, "step": 10954 }, { "epoch": 1.31, "learning_rate": 2.21499700059988e-06, "loss": 0.0006, "step": 10955 }, { "epoch": 1.31, "learning_rate": 2.2149250149970006e-06, "loss": 0.1957, "step": 10956 }, { "epoch": 1.31, "learning_rate": 2.214853029394121e-06, "loss": 0.0004, "step": 10957 }, { "epoch": 1.31, "learning_rate": 2.2147810437912418e-06, "loss": 0.0004, "step": 10958 }, { "epoch": 1.31, "learning_rate": 2.2147090581883625e-06, "loss": 0.0005, "step": 10959 }, { "epoch": 1.31, "learning_rate": 2.214637072585483e-06, "loss": 0.0003, "step": 10960 }, { "epoch": 1.31, "learning_rate": 2.2145650869826037e-06, "loss": 0.0004, "step": 10961 }, { "epoch": 1.31, "learning_rate": 2.2144931013797244e-06, "loss": 0.0002, "step": 10962 }, { "epoch": 1.31, "learning_rate": 2.214421115776845e-06, "loss": 0.0005, "step": 10963 }, { "epoch": 1.31, "learning_rate": 2.214349130173965e-06, "loss": 0.0003, "step": 10964 }, { "epoch": 1.31, "learning_rate": 2.214277144571086e-06, "loss": 0.0005, "step": 10965 }, { "epoch": 1.31, "learning_rate": 2.2142051589682063e-06, "loss": 0.0003, "step": 10966 }, { "epoch": 1.31, "learning_rate": 2.214133173365327e-06, "loss": 0.0012, "step": 10967 }, { "epoch": 1.31, "learning_rate": 2.2140611877624474e-06, "loss": 0.0003, "step": 10968 }, { "epoch": 1.31, "learning_rate": 2.213989202159568e-06, "loss": 0.0004, "step": 10969 }, { "epoch": 1.31, "learning_rate": 2.2139172165566886e-06, "loss": 0.0004, "step": 10970 }, { "epoch": 1.31, "learning_rate": 2.2138452309538093e-06, "loss": 0.0012, "step": 10971 }, { "epoch": 1.31, "learning_rate": 2.2137732453509297e-06, "loss": 0.0007, "step": 10972 }, { "epoch": 1.31, "learning_rate": 2.2137012597480505e-06, "loss": 0.0004, "step": 10973 }, { "epoch": 1.32, "learning_rate": 2.213629274145171e-06, "loss": 0.0004, "step": 10974 }, { "epoch": 1.32, "learning_rate": 2.2135572885422916e-06, "loss": 0.0003, "step": 10975 }, { "epoch": 1.32, "learning_rate": 2.2134853029394124e-06, "loss": 0.0005, "step": 10976 }, { "epoch": 1.32, "learning_rate": 2.2134133173365327e-06, "loss": 0.0005, "step": 10977 }, { "epoch": 1.32, "learning_rate": 2.2133413317336535e-06, "loss": 0.0023, "step": 10978 }, { "epoch": 1.32, "learning_rate": 2.213269346130774e-06, "loss": 0.0003, "step": 10979 }, { "epoch": 1.32, "learning_rate": 2.2131973605278946e-06, "loss": 0.0908, "step": 10980 }, { "epoch": 1.32, "learning_rate": 2.213125374925015e-06, "loss": 0.0002, "step": 10981 }, { "epoch": 1.32, "learning_rate": 2.2130533893221358e-06, "loss": 0.0006, "step": 10982 }, { "epoch": 1.32, "learning_rate": 2.212981403719256e-06, "loss": 0.0004, "step": 10983 }, { "epoch": 1.32, "learning_rate": 2.212909418116377e-06, "loss": 0.0002, "step": 10984 }, { "epoch": 1.32, "learning_rate": 2.2128374325134973e-06, "loss": 0.0005, "step": 10985 }, { "epoch": 1.32, "learning_rate": 2.2127654469106176e-06, "loss": 0.0004, "step": 10986 }, { "epoch": 1.32, "learning_rate": 2.2126934613077384e-06, "loss": 0.0045, "step": 10987 }, { "epoch": 1.32, "learning_rate": 2.2126214757048588e-06, "loss": 0.0004, "step": 10988 }, { "epoch": 1.32, "learning_rate": 2.2125494901019795e-06, "loss": 0.0032, "step": 10989 }, { "epoch": 1.32, "learning_rate": 2.2124775044991003e-06, "loss": 0.0004, "step": 10990 }, { "epoch": 1.32, "learning_rate": 2.212405518896221e-06, "loss": 0.0005, "step": 10991 }, { "epoch": 1.32, "learning_rate": 2.2123335332933414e-06, "loss": 0.002, "step": 10992 }, { "epoch": 1.32, "learning_rate": 2.2122615476904622e-06, "loss": 0.0003, "step": 10993 }, { "epoch": 1.32, "learning_rate": 2.2121895620875826e-06, "loss": 0.0003, "step": 10994 }, { "epoch": 1.32, "learning_rate": 2.2121175764847034e-06, "loss": 0.0005, "step": 10995 }, { "epoch": 1.32, "learning_rate": 2.2120455908818237e-06, "loss": 0.0003, "step": 10996 }, { "epoch": 1.32, "learning_rate": 2.2119736052789445e-06, "loss": 0.0003, "step": 10997 }, { "epoch": 1.32, "learning_rate": 2.211901619676065e-06, "loss": 0.8069, "step": 10998 }, { "epoch": 1.32, "learning_rate": 2.211829634073185e-06, "loss": 0.0412, "step": 10999 }, { "epoch": 1.32, "learning_rate": 2.211757648470306e-06, "loss": 0.0007, "step": 11000 }, { "epoch": 1.32, "learning_rate": 2.2116856628674263e-06, "loss": 0.0005, "step": 11001 }, { "epoch": 1.32, "learning_rate": 2.211613677264547e-06, "loss": 0.9226, "step": 11002 }, { "epoch": 1.32, "learning_rate": 2.2115416916616675e-06, "loss": 0.0007, "step": 11003 }, { "epoch": 1.32, "learning_rate": 2.2114697060587883e-06, "loss": 0.0003, "step": 11004 }, { "epoch": 1.32, "learning_rate": 2.211397720455909e-06, "loss": 0.0012, "step": 11005 }, { "epoch": 1.32, "learning_rate": 2.2113257348530294e-06, "loss": 0.9041, "step": 11006 }, { "epoch": 1.32, "learning_rate": 2.21125374925015e-06, "loss": 0.0007, "step": 11007 }, { "epoch": 1.32, "learning_rate": 2.211181763647271e-06, "loss": 0.0003, "step": 11008 }, { "epoch": 1.32, "learning_rate": 2.2111097780443913e-06, "loss": 0.0044, "step": 11009 }, { "epoch": 1.32, "learning_rate": 2.2110377924415117e-06, "loss": 0.0019, "step": 11010 }, { "epoch": 1.32, "learning_rate": 2.2109658068386324e-06, "loss": 0.0003, "step": 11011 }, { "epoch": 1.32, "learning_rate": 2.2108938212357528e-06, "loss": 0.0006, "step": 11012 }, { "epoch": 1.32, "learning_rate": 2.2108218356328736e-06, "loss": 0.0004, "step": 11013 }, { "epoch": 1.32, "learning_rate": 2.210749850029994e-06, "loss": 0.0004, "step": 11014 }, { "epoch": 1.32, "learning_rate": 2.2106778644271147e-06, "loss": 0.0003, "step": 11015 }, { "epoch": 1.32, "learning_rate": 2.210605878824235e-06, "loss": 0.0011, "step": 11016 }, { "epoch": 1.32, "learning_rate": 2.210533893221356e-06, "loss": 0.0003, "step": 11017 }, { "epoch": 1.32, "learning_rate": 2.210461907618476e-06, "loss": 0.0004, "step": 11018 }, { "epoch": 1.32, "learning_rate": 2.210389922015597e-06, "loss": 0.0007, "step": 11019 }, { "epoch": 1.32, "learning_rate": 2.2103179364127173e-06, "loss": 0.0004, "step": 11020 }, { "epoch": 1.32, "learning_rate": 2.210245950809838e-06, "loss": 0.0011, "step": 11021 }, { "epoch": 1.32, "learning_rate": 2.210173965206959e-06, "loss": 0.0004, "step": 11022 }, { "epoch": 1.32, "learning_rate": 2.2101019796040792e-06, "loss": 0.0003, "step": 11023 }, { "epoch": 1.32, "learning_rate": 2.2100299940012e-06, "loss": 0.0018, "step": 11024 }, { "epoch": 1.32, "learning_rate": 2.2099580083983204e-06, "loss": 0.0002, "step": 11025 }, { "epoch": 1.32, "learning_rate": 2.209886022795441e-06, "loss": 0.0003, "step": 11026 }, { "epoch": 1.32, "learning_rate": 2.2098140371925615e-06, "loss": 0.0004, "step": 11027 }, { "epoch": 1.32, "learning_rate": 2.2097420515896823e-06, "loss": 0.0005, "step": 11028 }, { "epoch": 1.32, "learning_rate": 2.2096700659868026e-06, "loss": 0.0004, "step": 11029 }, { "epoch": 1.32, "learning_rate": 2.2095980803839234e-06, "loss": 0.0003, "step": 11030 }, { "epoch": 1.32, "learning_rate": 2.2095260947810438e-06, "loss": 1.5402, "step": 11031 }, { "epoch": 1.32, "learning_rate": 2.209454109178164e-06, "loss": 0.0003, "step": 11032 }, { "epoch": 1.32, "learning_rate": 2.209382123575285e-06, "loss": 0.0003, "step": 11033 }, { "epoch": 1.32, "learning_rate": 2.2093101379724053e-06, "loss": 0.4648, "step": 11034 }, { "epoch": 1.32, "learning_rate": 2.209238152369526e-06, "loss": 0.0002, "step": 11035 }, { "epoch": 1.32, "learning_rate": 2.209166166766647e-06, "loss": 0.002, "step": 11036 }, { "epoch": 1.32, "learning_rate": 2.2090941811637676e-06, "loss": 0.0014, "step": 11037 }, { "epoch": 1.32, "learning_rate": 2.209022195560888e-06, "loss": 0.0006, "step": 11038 }, { "epoch": 1.32, "learning_rate": 2.2089502099580087e-06, "loss": 0.0004, "step": 11039 }, { "epoch": 1.32, "learning_rate": 2.208878224355129e-06, "loss": 0.0009, "step": 11040 }, { "epoch": 1.32, "learning_rate": 2.20880623875225e-06, "loss": 0.0007, "step": 11041 }, { "epoch": 1.32, "learning_rate": 2.2087342531493702e-06, "loss": 0.3923, "step": 11042 }, { "epoch": 1.32, "learning_rate": 2.208662267546491e-06, "loss": 0.0004, "step": 11043 }, { "epoch": 1.32, "learning_rate": 2.2085902819436113e-06, "loss": 0.0007, "step": 11044 }, { "epoch": 1.32, "learning_rate": 2.2085182963407317e-06, "loss": 0.0004, "step": 11045 }, { "epoch": 1.32, "learning_rate": 2.2084463107378525e-06, "loss": 0.0005, "step": 11046 }, { "epoch": 1.32, "learning_rate": 2.208374325134973e-06, "loss": 0.0004, "step": 11047 }, { "epoch": 1.32, "learning_rate": 2.2083023395320936e-06, "loss": 0.0003, "step": 11048 }, { "epoch": 1.32, "learning_rate": 2.208230353929214e-06, "loss": 0.0003, "step": 11049 }, { "epoch": 1.32, "learning_rate": 2.2081583683263347e-06, "loss": 0.0004, "step": 11050 }, { "epoch": 1.32, "learning_rate": 2.208086382723455e-06, "loss": 0.0004, "step": 11051 }, { "epoch": 1.32, "learning_rate": 2.208014397120576e-06, "loss": 0.0003, "step": 11052 }, { "epoch": 1.32, "learning_rate": 2.2079424115176967e-06, "loss": 0.0003, "step": 11053 }, { "epoch": 1.32, "learning_rate": 2.2078704259148174e-06, "loss": 0.0004, "step": 11054 }, { "epoch": 1.32, "learning_rate": 2.207798440311938e-06, "loss": 0.0096, "step": 11055 }, { "epoch": 1.32, "learning_rate": 2.207726454709058e-06, "loss": 0.0003, "step": 11056 }, { "epoch": 1.32, "learning_rate": 2.207654469106179e-06, "loss": 0.005, "step": 11057 }, { "epoch": 1.33, "learning_rate": 2.2075824835032993e-06, "loss": 0.0069, "step": 11058 }, { "epoch": 1.33, "learning_rate": 2.20751049790042e-06, "loss": 0.0006, "step": 11059 }, { "epoch": 1.33, "learning_rate": 2.2074385122975404e-06, "loss": 0.0004, "step": 11060 }, { "epoch": 1.33, "learning_rate": 2.207366526694661e-06, "loss": 0.0004, "step": 11061 }, { "epoch": 1.33, "learning_rate": 2.2072945410917815e-06, "loss": 0.0004, "step": 11062 }, { "epoch": 1.33, "learning_rate": 2.2072225554889023e-06, "loss": 0.0004, "step": 11063 }, { "epoch": 1.33, "learning_rate": 2.2071505698860227e-06, "loss": 0.0006, "step": 11064 }, { "epoch": 1.33, "learning_rate": 2.2070785842831435e-06, "loss": 0.0002, "step": 11065 }, { "epoch": 1.33, "learning_rate": 2.207006598680264e-06, "loss": 0.0007, "step": 11066 }, { "epoch": 1.33, "learning_rate": 2.2069346130773846e-06, "loss": 0.0005, "step": 11067 }, { "epoch": 1.33, "learning_rate": 2.2068626274745054e-06, "loss": 0.0004, "step": 11068 }, { "epoch": 1.33, "learning_rate": 2.2067906418716257e-06, "loss": 0.0003, "step": 11069 }, { "epoch": 1.33, "learning_rate": 2.2067186562687465e-06, "loss": 0.1096, "step": 11070 }, { "epoch": 1.33, "learning_rate": 2.206646670665867e-06, "loss": 0.0005, "step": 11071 }, { "epoch": 1.33, "learning_rate": 2.2065746850629876e-06, "loss": 0.0002, "step": 11072 }, { "epoch": 1.33, "learning_rate": 2.206502699460108e-06, "loss": 0.0005, "step": 11073 }, { "epoch": 1.33, "learning_rate": 2.2064307138572288e-06, "loss": 0.0033, "step": 11074 }, { "epoch": 1.33, "learning_rate": 2.206358728254349e-06, "loss": 0.0006, "step": 11075 }, { "epoch": 1.33, "learning_rate": 2.20628674265147e-06, "loss": 0.0002, "step": 11076 }, { "epoch": 1.33, "learning_rate": 2.2062147570485903e-06, "loss": 0.0004, "step": 11077 }, { "epoch": 1.33, "learning_rate": 2.2061427714457106e-06, "loss": 0.0003, "step": 11078 }, { "epoch": 1.33, "learning_rate": 2.2060707858428314e-06, "loss": 0.0007, "step": 11079 }, { "epoch": 1.33, "learning_rate": 2.2059988002399518e-06, "loss": 0.0004, "step": 11080 }, { "epoch": 1.33, "learning_rate": 2.2059268146370725e-06, "loss": 0.0003, "step": 11081 }, { "epoch": 1.33, "learning_rate": 2.2058548290341933e-06, "loss": 0.001, "step": 11082 }, { "epoch": 1.33, "learning_rate": 2.2057828434313137e-06, "loss": 0.0003, "step": 11083 }, { "epoch": 1.33, "learning_rate": 2.2057108578284344e-06, "loss": 0.8732, "step": 11084 }, { "epoch": 1.33, "learning_rate": 2.2056388722255552e-06, "loss": 0.0003, "step": 11085 }, { "epoch": 1.33, "learning_rate": 2.2055668866226756e-06, "loss": 0.0004, "step": 11086 }, { "epoch": 1.33, "learning_rate": 2.2054949010197964e-06, "loss": 0.0004, "step": 11087 }, { "epoch": 1.33, "learning_rate": 2.2054229154169167e-06, "loss": 0.0005, "step": 11088 }, { "epoch": 1.33, "learning_rate": 2.2053509298140375e-06, "loss": 0.0003, "step": 11089 }, { "epoch": 1.33, "learning_rate": 2.205278944211158e-06, "loss": 0.0003, "step": 11090 }, { "epoch": 1.33, "learning_rate": 2.205206958608278e-06, "loss": 0.0003, "step": 11091 }, { "epoch": 1.33, "learning_rate": 2.205134973005399e-06, "loss": 0.0003, "step": 11092 }, { "epoch": 1.33, "learning_rate": 2.2050629874025193e-06, "loss": 0.0003, "step": 11093 }, { "epoch": 1.33, "learning_rate": 2.20499100179964e-06, "loss": 0.0009, "step": 11094 }, { "epoch": 1.33, "learning_rate": 2.2049190161967605e-06, "loss": 0.0005, "step": 11095 }, { "epoch": 1.33, "learning_rate": 2.2048470305938812e-06, "loss": 0.0002, "step": 11096 }, { "epoch": 1.33, "learning_rate": 2.2047750449910016e-06, "loss": 0.0005, "step": 11097 }, { "epoch": 1.33, "learning_rate": 2.2047030593881224e-06, "loss": 0.0003, "step": 11098 }, { "epoch": 1.33, "learning_rate": 2.204631073785243e-06, "loss": 0.0003, "step": 11099 }, { "epoch": 1.33, "learning_rate": 2.204559088182364e-06, "loss": 0.0004, "step": 11100 }, { "epoch": 1.33, "learning_rate": 2.2044871025794843e-06, "loss": 0.8112, "step": 11101 }, { "epoch": 1.33, "learning_rate": 2.2044151169766046e-06, "loss": 0.5474, "step": 11102 }, { "epoch": 1.33, "learning_rate": 2.2043431313737254e-06, "loss": 0.0004, "step": 11103 }, { "epoch": 1.33, "learning_rate": 2.2042711457708458e-06, "loss": 0.0003, "step": 11104 }, { "epoch": 1.33, "learning_rate": 2.2041991601679666e-06, "loss": 0.0004, "step": 11105 }, { "epoch": 1.33, "learning_rate": 2.204127174565087e-06, "loss": 0.0002, "step": 11106 }, { "epoch": 1.33, "learning_rate": 2.2040551889622077e-06, "loss": 0.0004, "step": 11107 }, { "epoch": 1.33, "learning_rate": 2.203983203359328e-06, "loss": 0.0002, "step": 11108 }, { "epoch": 1.33, "learning_rate": 2.203911217756449e-06, "loss": 0.0006, "step": 11109 }, { "epoch": 1.33, "learning_rate": 2.203839232153569e-06, "loss": 0.0005, "step": 11110 }, { "epoch": 1.33, "learning_rate": 2.20376724655069e-06, "loss": 0.0006, "step": 11111 }, { "epoch": 1.33, "learning_rate": 2.2036952609478103e-06, "loss": 0.0004, "step": 11112 }, { "epoch": 1.33, "learning_rate": 2.203623275344931e-06, "loss": 0.7043, "step": 11113 }, { "epoch": 1.33, "learning_rate": 2.203551289742052e-06, "loss": 0.0003, "step": 11114 }, { "epoch": 1.33, "learning_rate": 2.2034793041391722e-06, "loss": 0.0004, "step": 11115 }, { "epoch": 1.33, "learning_rate": 2.203407318536293e-06, "loss": 0.001, "step": 11116 }, { "epoch": 1.33, "learning_rate": 2.2033353329334134e-06, "loss": 0.0201, "step": 11117 }, { "epoch": 1.33, "learning_rate": 2.203263347330534e-06, "loss": 0.8863, "step": 11118 }, { "epoch": 1.33, "learning_rate": 2.2031913617276545e-06, "loss": 0.1261, "step": 11119 }, { "epoch": 1.33, "learning_rate": 2.2031193761247753e-06, "loss": 0.0004, "step": 11120 }, { "epoch": 1.33, "learning_rate": 2.2030473905218956e-06, "loss": 0.0003, "step": 11121 }, { "epoch": 1.33, "learning_rate": 2.2029754049190164e-06, "loss": 0.0005, "step": 11122 }, { "epoch": 1.33, "learning_rate": 2.2029034193161368e-06, "loss": 0.0004, "step": 11123 }, { "epoch": 1.33, "learning_rate": 2.2028314337132575e-06, "loss": 0.0009, "step": 11124 }, { "epoch": 1.33, "learning_rate": 2.202759448110378e-06, "loss": 0.0005, "step": 11125 }, { "epoch": 1.33, "learning_rate": 2.2026874625074982e-06, "loss": 0.0002, "step": 11126 }, { "epoch": 1.33, "learning_rate": 2.202615476904619e-06, "loss": 0.0235, "step": 11127 }, { "epoch": 1.33, "learning_rate": 2.20254349130174e-06, "loss": 0.0002, "step": 11128 }, { "epoch": 1.33, "learning_rate": 2.20247150569886e-06, "loss": 0.0003, "step": 11129 }, { "epoch": 1.33, "learning_rate": 2.202399520095981e-06, "loss": 0.0003, "step": 11130 }, { "epoch": 1.33, "learning_rate": 2.2023275344931017e-06, "loss": 0.0002, "step": 11131 }, { "epoch": 1.33, "learning_rate": 2.202255548890222e-06, "loss": 0.0006, "step": 11132 }, { "epoch": 1.33, "learning_rate": 2.202183563287343e-06, "loss": 0.0006, "step": 11133 }, { "epoch": 1.33, "learning_rate": 2.202111577684463e-06, "loss": 0.0746, "step": 11134 }, { "epoch": 1.33, "learning_rate": 2.202039592081584e-06, "loss": 0.0003, "step": 11135 }, { "epoch": 1.33, "learning_rate": 2.2019676064787043e-06, "loss": 0.0005, "step": 11136 }, { "epoch": 1.33, "learning_rate": 2.2018956208758247e-06, "loss": 0.0004, "step": 11137 }, { "epoch": 1.33, "learning_rate": 2.2018236352729455e-06, "loss": 0.0005, "step": 11138 }, { "epoch": 1.33, "learning_rate": 2.201751649670066e-06, "loss": 0.0005, "step": 11139 }, { "epoch": 1.33, "learning_rate": 2.2016796640671866e-06, "loss": 0.0006, "step": 11140 }, { "epoch": 1.34, "learning_rate": 2.201607678464307e-06, "loss": 0.0005, "step": 11141 }, { "epoch": 1.34, "learning_rate": 2.2015356928614277e-06, "loss": 0.0002, "step": 11142 }, { "epoch": 1.34, "learning_rate": 2.201463707258548e-06, "loss": 0.0006, "step": 11143 }, { "epoch": 1.34, "learning_rate": 2.201391721655669e-06, "loss": 0.0007, "step": 11144 }, { "epoch": 1.34, "learning_rate": 2.2013197360527897e-06, "loss": 0.0004, "step": 11145 }, { "epoch": 1.34, "learning_rate": 2.2012477504499104e-06, "loss": 0.006, "step": 11146 }, { "epoch": 1.34, "learning_rate": 2.2011757648470308e-06, "loss": 0.0011, "step": 11147 }, { "epoch": 1.34, "learning_rate": 2.201103779244151e-06, "loss": 0.0006, "step": 11148 }, { "epoch": 1.34, "learning_rate": 2.201031793641272e-06, "loss": 0.0005, "step": 11149 }, { "epoch": 1.34, "learning_rate": 2.2009598080383923e-06, "loss": 0.0004, "step": 11150 }, { "epoch": 1.34, "learning_rate": 2.200887822435513e-06, "loss": 0.0006, "step": 11151 }, { "epoch": 1.34, "learning_rate": 2.2008158368326334e-06, "loss": 0.0003, "step": 11152 }, { "epoch": 1.34, "learning_rate": 2.200743851229754e-06, "loss": 0.0003, "step": 11153 }, { "epoch": 1.34, "learning_rate": 2.2006718656268745e-06, "loss": 0.4316, "step": 11154 }, { "epoch": 1.34, "learning_rate": 2.2005998800239953e-06, "loss": 0.0003, "step": 11155 }, { "epoch": 1.34, "learning_rate": 2.2005278944211157e-06, "loss": 0.5459, "step": 11156 }, { "epoch": 1.34, "learning_rate": 2.2004559088182365e-06, "loss": 0.0005, "step": 11157 }, { "epoch": 1.34, "learning_rate": 2.200383923215357e-06, "loss": 0.0003, "step": 11158 }, { "epoch": 1.34, "learning_rate": 2.2003119376124776e-06, "loss": 0.0005, "step": 11159 }, { "epoch": 1.34, "learning_rate": 2.2002399520095984e-06, "loss": 0.0004, "step": 11160 }, { "epoch": 1.34, "learning_rate": 2.2001679664067187e-06, "loss": 0.0004, "step": 11161 }, { "epoch": 1.34, "learning_rate": 2.2000959808038395e-06, "loss": 0.0002, "step": 11162 }, { "epoch": 1.34, "learning_rate": 2.20002399520096e-06, "loss": 0.9905, "step": 11163 }, { "epoch": 1.34, "learning_rate": 2.1999520095980806e-06, "loss": 0.0003, "step": 11164 }, { "epoch": 1.34, "learning_rate": 2.199880023995201e-06, "loss": 0.0007, "step": 11165 }, { "epoch": 1.34, "learning_rate": 2.1998080383923218e-06, "loss": 0.0005, "step": 11166 }, { "epoch": 1.34, "learning_rate": 2.199736052789442e-06, "loss": 0.0008, "step": 11167 }, { "epoch": 1.34, "learning_rate": 2.199664067186563e-06, "loss": 0.0006, "step": 11168 }, { "epoch": 1.34, "learning_rate": 2.1995920815836833e-06, "loss": 0.0003, "step": 11169 }, { "epoch": 1.34, "learning_rate": 2.199520095980804e-06, "loss": 0.0004, "step": 11170 }, { "epoch": 1.34, "learning_rate": 2.1994481103779244e-06, "loss": 0.0012, "step": 11171 }, { "epoch": 1.34, "learning_rate": 2.1993761247750447e-06, "loss": 0.0006, "step": 11172 }, { "epoch": 1.34, "learning_rate": 2.1993041391721655e-06, "loss": 0.0002, "step": 11173 }, { "epoch": 1.34, "learning_rate": 2.199232153569286e-06, "loss": 0.8705, "step": 11174 }, { "epoch": 1.34, "learning_rate": 2.1991601679664067e-06, "loss": 0.809, "step": 11175 }, { "epoch": 1.34, "learning_rate": 2.1990881823635274e-06, "loss": 0.6982, "step": 11176 }, { "epoch": 1.34, "learning_rate": 2.199016196760648e-06, "loss": 0.0003, "step": 11177 }, { "epoch": 1.34, "learning_rate": 2.1989442111577686e-06, "loss": 0.0003, "step": 11178 }, { "epoch": 1.34, "learning_rate": 2.1988722255548893e-06, "loss": 0.0004, "step": 11179 }, { "epoch": 1.34, "learning_rate": 2.1988002399520097e-06, "loss": 0.0005, "step": 11180 }, { "epoch": 1.34, "learning_rate": 2.1987282543491305e-06, "loss": 0.0024, "step": 11181 }, { "epoch": 1.34, "learning_rate": 2.198656268746251e-06, "loss": 0.0026, "step": 11182 }, { "epoch": 1.34, "learning_rate": 2.198584283143371e-06, "loss": 0.0009, "step": 11183 }, { "epoch": 1.34, "learning_rate": 2.198512297540492e-06, "loss": 0.0005, "step": 11184 }, { "epoch": 1.34, "learning_rate": 2.1984403119376123e-06, "loss": 0.0004, "step": 11185 }, { "epoch": 1.34, "learning_rate": 2.198368326334733e-06, "loss": 0.0009, "step": 11186 }, { "epoch": 1.34, "learning_rate": 2.1982963407318535e-06, "loss": 0.0004, "step": 11187 }, { "epoch": 1.34, "learning_rate": 2.1982243551289742e-06, "loss": 0.0005, "step": 11188 }, { "epoch": 1.34, "learning_rate": 2.1981523695260946e-06, "loss": 0.0003, "step": 11189 }, { "epoch": 1.34, "learning_rate": 2.1980803839232154e-06, "loss": 0.0003, "step": 11190 }, { "epoch": 1.34, "learning_rate": 2.198008398320336e-06, "loss": 0.0003, "step": 11191 }, { "epoch": 1.34, "learning_rate": 2.1979364127174565e-06, "loss": 0.3293, "step": 11192 }, { "epoch": 1.34, "learning_rate": 2.1978644271145773e-06, "loss": 0.0004, "step": 11193 }, { "epoch": 1.34, "learning_rate": 2.1977924415116976e-06, "loss": 0.0005, "step": 11194 }, { "epoch": 1.34, "learning_rate": 2.1977204559088184e-06, "loss": 0.0006, "step": 11195 }, { "epoch": 1.34, "learning_rate": 2.1976484703059388e-06, "loss": 0.0004, "step": 11196 }, { "epoch": 1.34, "learning_rate": 2.1975764847030595e-06, "loss": 0.0004, "step": 11197 }, { "epoch": 1.34, "learning_rate": 2.19750449910018e-06, "loss": 0.6906, "step": 11198 }, { "epoch": 1.34, "learning_rate": 2.1974325134973007e-06, "loss": 0.0004, "step": 11199 }, { "epoch": 1.34, "learning_rate": 2.197360527894421e-06, "loss": 0.0005, "step": 11200 }, { "epoch": 1.34, "learning_rate": 2.197288542291542e-06, "loss": 0.0003, "step": 11201 }, { "epoch": 1.34, "learning_rate": 2.197216556688662e-06, "loss": 0.0003, "step": 11202 }, { "epoch": 1.34, "learning_rate": 2.197144571085783e-06, "loss": 0.5753, "step": 11203 }, { "epoch": 1.34, "learning_rate": 2.1970725854829033e-06, "loss": 0.0002, "step": 11204 }, { "epoch": 1.34, "learning_rate": 2.197000599880024e-06, "loss": 0.0006, "step": 11205 }, { "epoch": 1.34, "learning_rate": 2.1969286142771444e-06, "loss": 0.0005, "step": 11206 }, { "epoch": 1.34, "learning_rate": 2.1968566286742652e-06, "loss": 0.0005, "step": 11207 }, { "epoch": 1.34, "learning_rate": 2.196784643071386e-06, "loss": 0.0008, "step": 11208 }, { "epoch": 1.34, "learning_rate": 2.1967126574685063e-06, "loss": 0.0003, "step": 11209 }, { "epoch": 1.34, "learning_rate": 2.196640671865627e-06, "loss": 0.0176, "step": 11210 }, { "epoch": 1.34, "learning_rate": 2.1965686862627475e-06, "loss": 0.0004, "step": 11211 }, { "epoch": 1.34, "learning_rate": 2.1964967006598683e-06, "loss": 0.0004, "step": 11212 }, { "epoch": 1.34, "learning_rate": 2.1964247150569886e-06, "loss": 0.0005, "step": 11213 }, { "epoch": 1.34, "learning_rate": 2.1963527294541094e-06, "loss": 0.0002, "step": 11214 }, { "epoch": 1.34, "learning_rate": 2.1962807438512297e-06, "loss": 0.0004, "step": 11215 }, { "epoch": 1.34, "learning_rate": 2.1962087582483505e-06, "loss": 0.002, "step": 11216 }, { "epoch": 1.34, "learning_rate": 2.196136772645471e-06, "loss": 0.0004, "step": 11217 }, { "epoch": 1.34, "learning_rate": 2.1960647870425912e-06, "loss": 0.0003, "step": 11218 }, { "epoch": 1.34, "learning_rate": 2.195992801439712e-06, "loss": 0.1609, "step": 11219 }, { "epoch": 1.34, "learning_rate": 2.1959208158368324e-06, "loss": 0.0003, "step": 11220 }, { "epoch": 1.34, "learning_rate": 2.195848830233953e-06, "loss": 0.0003, "step": 11221 }, { "epoch": 1.34, "learning_rate": 2.195776844631074e-06, "loss": 0.3508, "step": 11222 }, { "epoch": 1.34, "learning_rate": 2.1957048590281947e-06, "loss": 0.9068, "step": 11223 }, { "epoch": 1.34, "learning_rate": 2.195632873425315e-06, "loss": 0.0001, "step": 11224 }, { "epoch": 1.35, "learning_rate": 2.195560887822436e-06, "loss": 0.0011, "step": 11225 }, { "epoch": 1.35, "learning_rate": 2.195488902219556e-06, "loss": 0.0003, "step": 11226 }, { "epoch": 1.35, "learning_rate": 2.195416916616677e-06, "loss": 0.0006, "step": 11227 }, { "epoch": 1.35, "learning_rate": 2.1953449310137973e-06, "loss": 0.1506, "step": 11228 }, { "epoch": 1.35, "learning_rate": 2.1952729454109177e-06, "loss": 0.0004, "step": 11229 }, { "epoch": 1.35, "learning_rate": 2.1952009598080385e-06, "loss": 0.0004, "step": 11230 }, { "epoch": 1.35, "learning_rate": 2.195128974205159e-06, "loss": 0.0005, "step": 11231 }, { "epoch": 1.35, "learning_rate": 2.1950569886022796e-06, "loss": 0.0007, "step": 11232 }, { "epoch": 1.35, "learning_rate": 2.1949850029994e-06, "loss": 0.0004, "step": 11233 }, { "epoch": 1.35, "learning_rate": 2.1949130173965207e-06, "loss": 0.0005, "step": 11234 }, { "epoch": 1.35, "learning_rate": 2.194841031793641e-06, "loss": 0.0007, "step": 11235 }, { "epoch": 1.35, "learning_rate": 2.194769046190762e-06, "loss": 0.0004, "step": 11236 }, { "epoch": 1.35, "learning_rate": 2.1946970605878826e-06, "loss": 0.0006, "step": 11237 }, { "epoch": 1.35, "learning_rate": 2.194625074985003e-06, "loss": 0.0006, "step": 11238 }, { "epoch": 1.35, "learning_rate": 2.1945530893821238e-06, "loss": 0.0005, "step": 11239 }, { "epoch": 1.35, "learning_rate": 2.194481103779244e-06, "loss": 0.0007, "step": 11240 }, { "epoch": 1.35, "learning_rate": 2.194409118176365e-06, "loss": 0.0005, "step": 11241 }, { "epoch": 1.35, "learning_rate": 2.1943371325734853e-06, "loss": 0.0005, "step": 11242 }, { "epoch": 1.35, "learning_rate": 2.194265146970606e-06, "loss": 0.0028, "step": 11243 }, { "epoch": 1.35, "learning_rate": 2.1941931613677264e-06, "loss": 0.0003, "step": 11244 }, { "epoch": 1.35, "learning_rate": 2.194121175764847e-06, "loss": 0.0005, "step": 11245 }, { "epoch": 1.35, "learning_rate": 2.1940491901619675e-06, "loss": 0.0004, "step": 11246 }, { "epoch": 1.35, "learning_rate": 2.1939772045590883e-06, "loss": 0.0867, "step": 11247 }, { "epoch": 1.35, "learning_rate": 2.1939052189562087e-06, "loss": 0.0005, "step": 11248 }, { "epoch": 1.35, "learning_rate": 2.1938332333533294e-06, "loss": 0.001, "step": 11249 }, { "epoch": 1.35, "learning_rate": 2.19376124775045e-06, "loss": 0.0003, "step": 11250 }, { "epoch": 1.35, "learning_rate": 2.1936892621475706e-06, "loss": 0.0004, "step": 11251 }, { "epoch": 1.35, "learning_rate": 2.193617276544691e-06, "loss": 0.0004, "step": 11252 }, { "epoch": 1.35, "learning_rate": 2.1935452909418117e-06, "loss": 0.0004, "step": 11253 }, { "epoch": 1.35, "learning_rate": 2.1934733053389325e-06, "loss": 0.0005, "step": 11254 }, { "epoch": 1.35, "learning_rate": 2.193401319736053e-06, "loss": 0.9335, "step": 11255 }, { "epoch": 1.35, "learning_rate": 2.1933293341331736e-06, "loss": 0.2066, "step": 11256 }, { "epoch": 1.35, "learning_rate": 2.193257348530294e-06, "loss": 0.0003, "step": 11257 }, { "epoch": 1.35, "learning_rate": 2.1931853629274148e-06, "loss": 0.0002, "step": 11258 }, { "epoch": 1.35, "learning_rate": 2.193113377324535e-06, "loss": 0.0004, "step": 11259 }, { "epoch": 1.35, "learning_rate": 2.193041391721656e-06, "loss": 0.0003, "step": 11260 }, { "epoch": 1.35, "learning_rate": 2.1929694061187762e-06, "loss": 0.0002, "step": 11261 }, { "epoch": 1.35, "learning_rate": 2.192897420515897e-06, "loss": 0.0007, "step": 11262 }, { "epoch": 1.35, "learning_rate": 2.1928254349130174e-06, "loss": 0.0005, "step": 11263 }, { "epoch": 1.35, "learning_rate": 2.1927534493101377e-06, "loss": 0.0009, "step": 11264 }, { "epoch": 1.35, "learning_rate": 2.1926814637072585e-06, "loss": 0.0007, "step": 11265 }, { "epoch": 1.35, "learning_rate": 2.192609478104379e-06, "loss": 0.0006, "step": 11266 }, { "epoch": 1.35, "learning_rate": 2.1925374925014996e-06, "loss": 0.0003, "step": 11267 }, { "epoch": 1.35, "learning_rate": 2.1924655068986204e-06, "loss": 0.0005, "step": 11268 }, { "epoch": 1.35, "learning_rate": 2.192393521295741e-06, "loss": 0.0004, "step": 11269 }, { "epoch": 1.35, "learning_rate": 2.1923215356928616e-06, "loss": 0.0002, "step": 11270 }, { "epoch": 1.35, "learning_rate": 2.1922495500899823e-06, "loss": 0.0073, "step": 11271 }, { "epoch": 1.35, "learning_rate": 2.1921775644871027e-06, "loss": 0.0004, "step": 11272 }, { "epoch": 1.35, "learning_rate": 2.1921055788842235e-06, "loss": 0.8787, "step": 11273 }, { "epoch": 1.35, "learning_rate": 2.192033593281344e-06, "loss": 0.0005, "step": 11274 }, { "epoch": 1.35, "learning_rate": 2.191961607678464e-06, "loss": 0.0003, "step": 11275 }, { "epoch": 1.35, "learning_rate": 2.191889622075585e-06, "loss": 0.0003, "step": 11276 }, { "epoch": 1.35, "learning_rate": 2.1918176364727053e-06, "loss": 0.0012, "step": 11277 }, { "epoch": 1.35, "learning_rate": 2.191745650869826e-06, "loss": 0.0006, "step": 11278 }, { "epoch": 1.35, "learning_rate": 2.1916736652669464e-06, "loss": 0.0007, "step": 11279 }, { "epoch": 1.35, "learning_rate": 2.1916016796640672e-06, "loss": 0.0004, "step": 11280 }, { "epoch": 1.35, "learning_rate": 2.1915296940611876e-06, "loss": 0.0003, "step": 11281 }, { "epoch": 1.35, "learning_rate": 2.1914577084583084e-06, "loss": 0.008, "step": 11282 }, { "epoch": 1.35, "learning_rate": 2.191385722855429e-06, "loss": 0.0002, "step": 11283 }, { "epoch": 1.35, "learning_rate": 2.1913137372525495e-06, "loss": 0.0011, "step": 11284 }, { "epoch": 1.35, "learning_rate": 2.1912417516496703e-06, "loss": 0.0006, "step": 11285 }, { "epoch": 1.35, "learning_rate": 2.1911697660467906e-06, "loss": 0.0004, "step": 11286 }, { "epoch": 1.35, "learning_rate": 2.1910977804439114e-06, "loss": 0.0005, "step": 11287 }, { "epoch": 1.35, "learning_rate": 2.1910257948410318e-06, "loss": 0.0003, "step": 11288 }, { "epoch": 1.35, "learning_rate": 2.1909538092381525e-06, "loss": 0.0005, "step": 11289 }, { "epoch": 1.35, "learning_rate": 2.190881823635273e-06, "loss": 0.0003, "step": 11290 }, { "epoch": 1.35, "learning_rate": 2.1908098380323937e-06, "loss": 0.0004, "step": 11291 }, { "epoch": 1.35, "learning_rate": 2.190737852429514e-06, "loss": 0.0004, "step": 11292 }, { "epoch": 1.35, "learning_rate": 2.190665866826635e-06, "loss": 0.0004, "step": 11293 }, { "epoch": 1.35, "learning_rate": 2.190593881223755e-06, "loss": 0.0004, "step": 11294 }, { "epoch": 1.35, "learning_rate": 2.190521895620876e-06, "loss": 0.0019, "step": 11295 }, { "epoch": 1.35, "learning_rate": 2.1904499100179963e-06, "loss": 0.0004, "step": 11296 }, { "epoch": 1.35, "learning_rate": 2.190377924415117e-06, "loss": 0.0003, "step": 11297 }, { "epoch": 1.35, "learning_rate": 2.1903059388122374e-06, "loss": 0.0006, "step": 11298 }, { "epoch": 1.35, "learning_rate": 2.190233953209358e-06, "loss": 0.0009, "step": 11299 }, { "epoch": 1.35, "learning_rate": 2.190161967606479e-06, "loss": 0.0005, "step": 11300 }, { "epoch": 1.35, "learning_rate": 2.1900899820035993e-06, "loss": 0.0005, "step": 11301 }, { "epoch": 1.35, "learning_rate": 2.19001799640072e-06, "loss": 0.0009, "step": 11302 }, { "epoch": 1.35, "learning_rate": 2.1899460107978405e-06, "loss": 0.0005, "step": 11303 }, { "epoch": 1.35, "learning_rate": 2.1898740251949613e-06, "loss": 0.0004, "step": 11304 }, { "epoch": 1.35, "learning_rate": 2.1898020395920816e-06, "loss": 0.0003, "step": 11305 }, { "epoch": 1.35, "learning_rate": 2.1897300539892024e-06, "loss": 0.0043, "step": 11306 }, { "epoch": 1.35, "learning_rate": 2.1896580683863227e-06, "loss": 0.0006, "step": 11307 }, { "epoch": 1.36, "learning_rate": 2.1895860827834435e-06, "loss": 0.0007, "step": 11308 }, { "epoch": 1.36, "learning_rate": 2.189514097180564e-06, "loss": 0.238, "step": 11309 }, { "epoch": 1.36, "learning_rate": 2.1894421115776842e-06, "loss": 0.0004, "step": 11310 }, { "epoch": 1.36, "learning_rate": 2.189370125974805e-06, "loss": 0.0005, "step": 11311 }, { "epoch": 1.36, "learning_rate": 2.1892981403719254e-06, "loss": 0.0005, "step": 11312 }, { "epoch": 1.36, "learning_rate": 2.189226154769046e-06, "loss": 0.0002, "step": 11313 }, { "epoch": 1.36, "learning_rate": 2.189154169166167e-06, "loss": 0.0003, "step": 11314 }, { "epoch": 1.36, "learning_rate": 2.1890821835632873e-06, "loss": 0.001, "step": 11315 }, { "epoch": 1.36, "learning_rate": 2.189010197960408e-06, "loss": 0.0003, "step": 11316 }, { "epoch": 1.36, "learning_rate": 2.188938212357529e-06, "loss": 0.0005, "step": 11317 }, { "epoch": 1.36, "learning_rate": 2.188866226754649e-06, "loss": 0.0004, "step": 11318 }, { "epoch": 1.36, "learning_rate": 2.18879424115177e-06, "loss": 0.0004, "step": 11319 }, { "epoch": 1.36, "learning_rate": 2.1887222555488903e-06, "loss": 0.0007, "step": 11320 }, { "epoch": 1.36, "learning_rate": 2.1886502699460107e-06, "loss": 0.0008, "step": 11321 }, { "epoch": 1.36, "learning_rate": 2.1885782843431315e-06, "loss": 0.0013, "step": 11322 }, { "epoch": 1.36, "learning_rate": 2.188506298740252e-06, "loss": 0.0006, "step": 11323 }, { "epoch": 1.36, "learning_rate": 2.1884343131373726e-06, "loss": 0.005, "step": 11324 }, { "epoch": 1.36, "learning_rate": 2.188362327534493e-06, "loss": 0.0005, "step": 11325 }, { "epoch": 1.36, "learning_rate": 2.1882903419316137e-06, "loss": 0.9906, "step": 11326 }, { "epoch": 1.36, "learning_rate": 2.188218356328734e-06, "loss": 0.0013, "step": 11327 }, { "epoch": 1.36, "learning_rate": 2.188146370725855e-06, "loss": 0.0006, "step": 11328 }, { "epoch": 1.36, "learning_rate": 2.188074385122975e-06, "loss": 0.0008, "step": 11329 }, { "epoch": 1.36, "learning_rate": 2.188002399520096e-06, "loss": 0.0002, "step": 11330 }, { "epoch": 1.36, "learning_rate": 2.1879304139172168e-06, "loss": 0.0198, "step": 11331 }, { "epoch": 1.36, "learning_rate": 2.187858428314337e-06, "loss": 0.0003, "step": 11332 }, { "epoch": 1.36, "learning_rate": 2.187786442711458e-06, "loss": 0.0009, "step": 11333 }, { "epoch": 1.36, "learning_rate": 2.1877144571085783e-06, "loss": 0.0084, "step": 11334 }, { "epoch": 1.36, "learning_rate": 2.187642471505699e-06, "loss": 0.0004, "step": 11335 }, { "epoch": 1.36, "learning_rate": 2.1875704859028194e-06, "loss": 0.0005, "step": 11336 }, { "epoch": 1.36, "learning_rate": 2.18749850029994e-06, "loss": 0.0007, "step": 11337 }, { "epoch": 1.36, "learning_rate": 2.1874265146970605e-06, "loss": 0.0006, "step": 11338 }, { "epoch": 1.36, "learning_rate": 2.1873545290941813e-06, "loss": 0.0004, "step": 11339 }, { "epoch": 1.36, "learning_rate": 2.1872825434913017e-06, "loss": 0.0004, "step": 11340 }, { "epoch": 1.36, "learning_rate": 2.1872105578884224e-06, "loss": 0.0009, "step": 11341 }, { "epoch": 1.36, "learning_rate": 2.187138572285543e-06, "loss": 0.0009, "step": 11342 }, { "epoch": 1.36, "learning_rate": 2.1870665866826636e-06, "loss": 0.0004, "step": 11343 }, { "epoch": 1.36, "learning_rate": 2.186994601079784e-06, "loss": 0.0013, "step": 11344 }, { "epoch": 1.36, "learning_rate": 2.1869226154769047e-06, "loss": 0.0002, "step": 11345 }, { "epoch": 1.36, "learning_rate": 2.1868506298740255e-06, "loss": 0.0003, "step": 11346 }, { "epoch": 1.36, "learning_rate": 2.186778644271146e-06, "loss": 0.0004, "step": 11347 }, { "epoch": 1.36, "learning_rate": 2.1867066586682666e-06, "loss": 0.5113, "step": 11348 }, { "epoch": 1.36, "learning_rate": 2.186634673065387e-06, "loss": 0.0006, "step": 11349 }, { "epoch": 1.36, "learning_rate": 2.1865626874625077e-06, "loss": 0.0003, "step": 11350 }, { "epoch": 1.36, "learning_rate": 2.186490701859628e-06, "loss": 0.0004, "step": 11351 }, { "epoch": 1.36, "learning_rate": 2.186418716256749e-06, "loss": 0.3734, "step": 11352 }, { "epoch": 1.36, "learning_rate": 2.1863467306538692e-06, "loss": 0.0004, "step": 11353 }, { "epoch": 1.36, "learning_rate": 2.18627474505099e-06, "loss": 0.0318, "step": 11354 }, { "epoch": 1.36, "learning_rate": 2.1862027594481104e-06, "loss": 0.0004, "step": 11355 }, { "epoch": 1.36, "learning_rate": 2.1861307738452307e-06, "loss": 0.0089, "step": 11356 }, { "epoch": 1.36, "learning_rate": 2.1860587882423515e-06, "loss": 0.0003, "step": 11357 }, { "epoch": 1.36, "learning_rate": 2.185986802639472e-06, "loss": 0.0002, "step": 11358 }, { "epoch": 1.36, "learning_rate": 2.1859148170365926e-06, "loss": 0.0008, "step": 11359 }, { "epoch": 1.36, "learning_rate": 2.1858428314337134e-06, "loss": 0.0005, "step": 11360 }, { "epoch": 1.36, "learning_rate": 2.1857708458308338e-06, "loss": 0.0104, "step": 11361 }, { "epoch": 1.36, "learning_rate": 2.1856988602279546e-06, "loss": 0.0005, "step": 11362 }, { "epoch": 1.36, "learning_rate": 2.1856268746250753e-06, "loss": 0.0002, "step": 11363 }, { "epoch": 1.36, "learning_rate": 2.1855548890221957e-06, "loss": 0.0004, "step": 11364 }, { "epoch": 1.36, "learning_rate": 2.1854829034193165e-06, "loss": 0.0766, "step": 11365 }, { "epoch": 1.36, "learning_rate": 2.185410917816437e-06, "loss": 0.0005, "step": 11366 }, { "epoch": 1.36, "learning_rate": 2.185338932213557e-06, "loss": 0.0003, "step": 11367 }, { "epoch": 1.36, "learning_rate": 2.185266946610678e-06, "loss": 0.0005, "step": 11368 }, { "epoch": 1.36, "learning_rate": 2.1851949610077983e-06, "loss": 0.0005, "step": 11369 }, { "epoch": 1.36, "learning_rate": 2.185122975404919e-06, "loss": 0.0005, "step": 11370 }, { "epoch": 1.36, "learning_rate": 2.1850509898020394e-06, "loss": 0.0003, "step": 11371 }, { "epoch": 1.36, "learning_rate": 2.1849790041991602e-06, "loss": 0.9975, "step": 11372 }, { "epoch": 1.36, "learning_rate": 2.1849070185962806e-06, "loss": 0.0004, "step": 11373 }, { "epoch": 1.36, "learning_rate": 2.1848350329934014e-06, "loss": 0.0003, "step": 11374 }, { "epoch": 1.36, "learning_rate": 2.1847630473905217e-06, "loss": 0.0007, "step": 11375 }, { "epoch": 1.36, "learning_rate": 2.1846910617876425e-06, "loss": 0.001, "step": 11376 }, { "epoch": 1.36, "learning_rate": 2.1846190761847633e-06, "loss": 0.0003, "step": 11377 }, { "epoch": 1.36, "learning_rate": 2.184547090581884e-06, "loss": 0.0003, "step": 11378 }, { "epoch": 1.36, "learning_rate": 2.1844751049790044e-06, "loss": 0.0004, "step": 11379 }, { "epoch": 1.36, "learning_rate": 2.1844031193761248e-06, "loss": 0.0004, "step": 11380 }, { "epoch": 1.36, "learning_rate": 2.1843311337732455e-06, "loss": 0.0005, "step": 11381 }, { "epoch": 1.36, "learning_rate": 2.184259148170366e-06, "loss": 0.0003, "step": 11382 }, { "epoch": 1.36, "learning_rate": 2.1841871625674867e-06, "loss": 0.0005, "step": 11383 }, { "epoch": 1.36, "learning_rate": 2.184115176964607e-06, "loss": 0.0001, "step": 11384 }, { "epoch": 1.36, "learning_rate": 2.184043191361728e-06, "loss": 0.0004, "step": 11385 }, { "epoch": 1.36, "learning_rate": 2.183971205758848e-06, "loss": 0.5199, "step": 11386 }, { "epoch": 1.36, "learning_rate": 2.183899220155969e-06, "loss": 0.0002, "step": 11387 }, { "epoch": 1.36, "learning_rate": 2.1838272345530893e-06, "loss": 0.0004, "step": 11388 }, { "epoch": 1.36, "learning_rate": 2.18375524895021e-06, "loss": 0.0003, "step": 11389 }, { "epoch": 1.36, "learning_rate": 2.1836832633473304e-06, "loss": 0.0004, "step": 11390 }, { "epoch": 1.37, "learning_rate": 2.183611277744451e-06, "loss": 0.0004, "step": 11391 }, { "epoch": 1.37, "learning_rate": 2.183539292141572e-06, "loss": 0.0036, "step": 11392 }, { "epoch": 1.37, "learning_rate": 2.1834673065386923e-06, "loss": 0.0004, "step": 11393 }, { "epoch": 1.37, "learning_rate": 2.183395320935813e-06, "loss": 0.0003, "step": 11394 }, { "epoch": 1.37, "learning_rate": 2.1833233353329335e-06, "loss": 0.0003, "step": 11395 }, { "epoch": 1.37, "learning_rate": 2.1832513497300542e-06, "loss": 0.0004, "step": 11396 }, { "epoch": 1.37, "learning_rate": 2.1831793641271746e-06, "loss": 0.0003, "step": 11397 }, { "epoch": 1.37, "learning_rate": 2.1831073785242954e-06, "loss": 0.0003, "step": 11398 }, { "epoch": 1.37, "learning_rate": 2.1830353929214157e-06, "loss": 0.0002, "step": 11399 }, { "epoch": 1.37, "learning_rate": 2.1829634073185365e-06, "loss": 0.0005, "step": 11400 }, { "epoch": 1.37, "learning_rate": 2.182891421715657e-06, "loss": 0.0003, "step": 11401 }, { "epoch": 1.37, "learning_rate": 2.1828194361127772e-06, "loss": 0.0003, "step": 11402 }, { "epoch": 1.37, "learning_rate": 2.182747450509898e-06, "loss": 0.001, "step": 11403 }, { "epoch": 1.37, "learning_rate": 2.1826754649070184e-06, "loss": 0.0005, "step": 11404 }, { "epoch": 1.37, "learning_rate": 2.182603479304139e-06, "loss": 0.0004, "step": 11405 }, { "epoch": 1.37, "learning_rate": 2.18253149370126e-06, "loss": 0.0003, "step": 11406 }, { "epoch": 1.37, "learning_rate": 2.1824595080983803e-06, "loss": 0.1071, "step": 11407 }, { "epoch": 1.37, "learning_rate": 2.182387522495501e-06, "loss": 0.0007, "step": 11408 }, { "epoch": 1.37, "learning_rate": 2.182315536892622e-06, "loss": 0.0006, "step": 11409 }, { "epoch": 1.37, "learning_rate": 2.182243551289742e-06, "loss": 0.0005, "step": 11410 }, { "epoch": 1.37, "learning_rate": 2.182171565686863e-06, "loss": 0.0002, "step": 11411 }, { "epoch": 1.37, "learning_rate": 2.1820995800839833e-06, "loss": 0.0005, "step": 11412 }, { "epoch": 1.37, "learning_rate": 2.1820275944811037e-06, "loss": 0.0003, "step": 11413 }, { "epoch": 1.37, "learning_rate": 2.1819556088782244e-06, "loss": 0.8287, "step": 11414 }, { "epoch": 1.37, "learning_rate": 2.181883623275345e-06, "loss": 0.0014, "step": 11415 }, { "epoch": 1.37, "learning_rate": 2.1818116376724656e-06, "loss": 0.063, "step": 11416 }, { "epoch": 1.37, "learning_rate": 2.181739652069586e-06, "loss": 0.8905, "step": 11417 }, { "epoch": 1.37, "learning_rate": 2.1816676664667067e-06, "loss": 0.0003, "step": 11418 }, { "epoch": 1.37, "learning_rate": 2.181595680863827e-06, "loss": 0.0004, "step": 11419 }, { "epoch": 1.37, "learning_rate": 2.181523695260948e-06, "loss": 0.0002, "step": 11420 }, { "epoch": 1.37, "learning_rate": 2.181451709658068e-06, "loss": 0.0034, "step": 11421 }, { "epoch": 1.37, "learning_rate": 2.181379724055189e-06, "loss": 0.0002, "step": 11422 }, { "epoch": 1.37, "learning_rate": 2.1813077384523098e-06, "loss": 0.0002, "step": 11423 }, { "epoch": 1.37, "learning_rate": 2.1812357528494305e-06, "loss": 0.0002, "step": 11424 }, { "epoch": 1.37, "learning_rate": 2.181163767246551e-06, "loss": 0.0003, "step": 11425 }, { "epoch": 1.37, "learning_rate": 2.1810917816436712e-06, "loss": 0.0003, "step": 11426 }, { "epoch": 1.37, "learning_rate": 2.181019796040792e-06, "loss": 0.0003, "step": 11427 }, { "epoch": 1.37, "learning_rate": 2.1809478104379124e-06, "loss": 0.0003, "step": 11428 }, { "epoch": 1.37, "learning_rate": 2.180875824835033e-06, "loss": 0.0005, "step": 11429 }, { "epoch": 1.37, "learning_rate": 2.1808038392321535e-06, "loss": 0.0006, "step": 11430 }, { "epoch": 1.37, "learning_rate": 2.1807318536292743e-06, "loss": 0.0003, "step": 11431 }, { "epoch": 1.37, "learning_rate": 2.1806598680263946e-06, "loss": 0.0005, "step": 11432 }, { "epoch": 1.37, "learning_rate": 2.1805878824235154e-06, "loss": 0.0003, "step": 11433 }, { "epoch": 1.37, "learning_rate": 2.1805158968206358e-06, "loss": 0.0008, "step": 11434 }, { "epoch": 1.37, "learning_rate": 2.1804439112177566e-06, "loss": 0.0005, "step": 11435 }, { "epoch": 1.37, "learning_rate": 2.180371925614877e-06, "loss": 0.4916, "step": 11436 }, { "epoch": 1.37, "learning_rate": 2.1802999400119977e-06, "loss": 0.0012, "step": 11437 }, { "epoch": 1.37, "learning_rate": 2.1802279544091185e-06, "loss": 0.0004, "step": 11438 }, { "epoch": 1.37, "learning_rate": 2.180155968806239e-06, "loss": 0.0003, "step": 11439 }, { "epoch": 1.37, "learning_rate": 2.1800839832033596e-06, "loss": 0.0006, "step": 11440 }, { "epoch": 1.37, "learning_rate": 2.18001199760048e-06, "loss": 0.0004, "step": 11441 }, { "epoch": 1.37, "learning_rate": 2.1799400119976007e-06, "loss": 0.0003, "step": 11442 }, { "epoch": 1.37, "learning_rate": 2.179868026394721e-06, "loss": 0.0015, "step": 11443 }, { "epoch": 1.37, "learning_rate": 2.179796040791842e-06, "loss": 0.0002, "step": 11444 }, { "epoch": 1.37, "learning_rate": 2.1797240551889622e-06, "loss": 0.0003, "step": 11445 }, { "epoch": 1.37, "learning_rate": 2.179652069586083e-06, "loss": 0.0004, "step": 11446 }, { "epoch": 1.37, "learning_rate": 2.1795800839832034e-06, "loss": 0.0004, "step": 11447 }, { "epoch": 1.37, "learning_rate": 2.1795080983803237e-06, "loss": 0.0008, "step": 11448 }, { "epoch": 1.37, "learning_rate": 2.1794361127774445e-06, "loss": 0.0003, "step": 11449 }, { "epoch": 1.37, "learning_rate": 2.179364127174565e-06, "loss": 0.0004, "step": 11450 }, { "epoch": 1.37, "learning_rate": 2.1792921415716856e-06, "loss": 0.6449, "step": 11451 }, { "epoch": 1.37, "learning_rate": 2.179220155968806e-06, "loss": 0.0004, "step": 11452 }, { "epoch": 1.37, "learning_rate": 2.1791481703659268e-06, "loss": 0.0004, "step": 11453 }, { "epoch": 1.37, "learning_rate": 2.1790761847630475e-06, "loss": 0.0004, "step": 11454 }, { "epoch": 1.37, "learning_rate": 2.1790041991601683e-06, "loss": 0.0005, "step": 11455 }, { "epoch": 1.37, "learning_rate": 2.1789322135572887e-06, "loss": 0.0021, "step": 11456 }, { "epoch": 1.37, "learning_rate": 2.1788602279544095e-06, "loss": 0.0003, "step": 11457 }, { "epoch": 1.37, "learning_rate": 2.17878824235153e-06, "loss": 0.0003, "step": 11458 }, { "epoch": 1.37, "learning_rate": 2.17871625674865e-06, "loss": 0.0004, "step": 11459 }, { "epoch": 1.37, "learning_rate": 2.178644271145771e-06, "loss": 0.0002, "step": 11460 }, { "epoch": 1.37, "learning_rate": 2.1785722855428913e-06, "loss": 0.0004, "step": 11461 }, { "epoch": 1.37, "learning_rate": 2.178500299940012e-06, "loss": 0.0001, "step": 11462 }, { "epoch": 1.37, "learning_rate": 2.1784283143371324e-06, "loss": 0.0366, "step": 11463 }, { "epoch": 1.37, "learning_rate": 2.178356328734253e-06, "loss": 0.2895, "step": 11464 }, { "epoch": 1.37, "learning_rate": 2.1782843431313736e-06, "loss": 0.0002, "step": 11465 }, { "epoch": 1.37, "learning_rate": 2.1782123575284943e-06, "loss": 0.0002, "step": 11466 }, { "epoch": 1.37, "learning_rate": 2.1781403719256147e-06, "loss": 0.6157, "step": 11467 }, { "epoch": 1.37, "learning_rate": 2.1780683863227355e-06, "loss": 0.0039, "step": 11468 }, { "epoch": 1.37, "learning_rate": 2.1779964007198563e-06, "loss": 0.7245, "step": 11469 }, { "epoch": 1.37, "learning_rate": 2.1779244151169766e-06, "loss": 0.0005, "step": 11470 }, { "epoch": 1.37, "learning_rate": 2.1778524295140974e-06, "loss": 0.0003, "step": 11471 }, { "epoch": 1.37, "learning_rate": 2.1777804439112177e-06, "loss": 0.0004, "step": 11472 }, { "epoch": 1.37, "learning_rate": 2.1777084583083385e-06, "loss": 0.0004, "step": 11473 }, { "epoch": 1.37, "learning_rate": 2.177636472705459e-06, "loss": 0.2161, "step": 11474 }, { "epoch": 1.38, "learning_rate": 2.1775644871025797e-06, "loss": 0.0015, "step": 11475 }, { "epoch": 1.38, "learning_rate": 2.1774925014997e-06, "loss": 0.0005, "step": 11476 }, { "epoch": 1.38, "learning_rate": 2.177420515896821e-06, "loss": 0.0005, "step": 11477 }, { "epoch": 1.38, "learning_rate": 2.177348530293941e-06, "loss": 0.0002, "step": 11478 }, { "epoch": 1.38, "learning_rate": 2.177276544691062e-06, "loss": 0.0006, "step": 11479 }, { "epoch": 1.38, "learning_rate": 2.1772045590881823e-06, "loss": 0.0002, "step": 11480 }, { "epoch": 1.38, "learning_rate": 2.177132573485303e-06, "loss": 0.0004, "step": 11481 }, { "epoch": 1.38, "learning_rate": 2.1770605878824234e-06, "loss": 0.2058, "step": 11482 }, { "epoch": 1.38, "learning_rate": 2.176988602279544e-06, "loss": 0.0019, "step": 11483 }, { "epoch": 1.38, "learning_rate": 2.1769166166766645e-06, "loss": 0.0867, "step": 11484 }, { "epoch": 1.38, "learning_rate": 2.1768446310737853e-06, "loss": 0.0005, "step": 11485 }, { "epoch": 1.38, "learning_rate": 2.176772645470906e-06, "loss": 0.0006, "step": 11486 }, { "epoch": 1.38, "learning_rate": 2.1767006598680265e-06, "loss": 0.0003, "step": 11487 }, { "epoch": 1.38, "learning_rate": 2.1766286742651472e-06, "loss": 0.0004, "step": 11488 }, { "epoch": 1.38, "learning_rate": 2.1765566886622676e-06, "loss": 0.0015, "step": 11489 }, { "epoch": 1.38, "learning_rate": 2.1764847030593884e-06, "loss": 0.0002, "step": 11490 }, { "epoch": 1.38, "learning_rate": 2.1764127174565087e-06, "loss": 0.0004, "step": 11491 }, { "epoch": 1.38, "learning_rate": 2.1763407318536295e-06, "loss": 0.0118, "step": 11492 }, { "epoch": 1.38, "learning_rate": 2.17626874625075e-06, "loss": 0.0015, "step": 11493 }, { "epoch": 1.38, "learning_rate": 2.1761967606478702e-06, "loss": 0.0008, "step": 11494 }, { "epoch": 1.38, "learning_rate": 2.176124775044991e-06, "loss": 0.0002, "step": 11495 }, { "epoch": 1.38, "learning_rate": 2.1760527894421113e-06, "loss": 0.0003, "step": 11496 }, { "epoch": 1.38, "learning_rate": 2.175980803839232e-06, "loss": 0.0008, "step": 11497 }, { "epoch": 1.38, "learning_rate": 2.1759088182363525e-06, "loss": 0.0004, "step": 11498 }, { "epoch": 1.38, "learning_rate": 2.1758368326334733e-06, "loss": 0.0004, "step": 11499 }, { "epoch": 1.38, "learning_rate": 2.175764847030594e-06, "loss": 0.0003, "step": 11500 }, { "epoch": 1.38, "learning_rate": 2.175692861427715e-06, "loss": 0.0005, "step": 11501 }, { "epoch": 1.38, "learning_rate": 2.175620875824835e-06, "loss": 0.0002, "step": 11502 }, { "epoch": 1.38, "learning_rate": 2.175548890221956e-06, "loss": 0.0003, "step": 11503 }, { "epoch": 1.38, "learning_rate": 2.1754769046190763e-06, "loss": 0.57, "step": 11504 }, { "epoch": 1.38, "learning_rate": 2.175404919016197e-06, "loss": 0.0002, "step": 11505 }, { "epoch": 1.38, "learning_rate": 2.1753329334133174e-06, "loss": 0.0003, "step": 11506 }, { "epoch": 1.38, "learning_rate": 2.175260947810438e-06, "loss": 0.0002, "step": 11507 }, { "epoch": 1.38, "learning_rate": 2.1751889622075586e-06, "loss": 0.0003, "step": 11508 }, { "epoch": 1.38, "learning_rate": 2.175116976604679e-06, "loss": 0.0011, "step": 11509 }, { "epoch": 1.38, "learning_rate": 2.1750449910017997e-06, "loss": 0.0002, "step": 11510 }, { "epoch": 1.38, "learning_rate": 2.17497300539892e-06, "loss": 0.9609, "step": 11511 }, { "epoch": 1.38, "learning_rate": 2.174901019796041e-06, "loss": 0.0002, "step": 11512 }, { "epoch": 1.38, "learning_rate": 2.174829034193161e-06, "loss": 0.0004, "step": 11513 }, { "epoch": 1.38, "learning_rate": 2.174757048590282e-06, "loss": 0.0002, "step": 11514 }, { "epoch": 1.38, "learning_rate": 2.1746850629874028e-06, "loss": 0.0004, "step": 11515 }, { "epoch": 1.38, "learning_rate": 2.174613077384523e-06, "loss": 0.0033, "step": 11516 }, { "epoch": 1.38, "learning_rate": 2.174541091781644e-06, "loss": 0.0003, "step": 11517 }, { "epoch": 1.38, "learning_rate": 2.1744691061787642e-06, "loss": 0.0159, "step": 11518 }, { "epoch": 1.38, "learning_rate": 2.174397120575885e-06, "loss": 0.002, "step": 11519 }, { "epoch": 1.38, "learning_rate": 2.1743251349730054e-06, "loss": 0.0003, "step": 11520 }, { "epoch": 1.38, "learning_rate": 2.174253149370126e-06, "loss": 0.0003, "step": 11521 }, { "epoch": 1.38, "learning_rate": 2.1741811637672465e-06, "loss": 0.1443, "step": 11522 }, { "epoch": 1.38, "learning_rate": 2.1741091781643673e-06, "loss": 0.0004, "step": 11523 }, { "epoch": 1.38, "learning_rate": 2.1740371925614876e-06, "loss": 0.0042, "step": 11524 }, { "epoch": 1.38, "learning_rate": 2.1739652069586084e-06, "loss": 0.0006, "step": 11525 }, { "epoch": 1.38, "learning_rate": 2.1738932213557288e-06, "loss": 0.0005, "step": 11526 }, { "epoch": 1.38, "learning_rate": 2.1738212357528496e-06, "loss": 0.0009, "step": 11527 }, { "epoch": 1.38, "learning_rate": 2.17374925014997e-06, "loss": 0.0002, "step": 11528 }, { "epoch": 1.38, "learning_rate": 2.1736772645470907e-06, "loss": 0.0003, "step": 11529 }, { "epoch": 1.38, "learning_rate": 2.173605278944211e-06, "loss": 0.0003, "step": 11530 }, { "epoch": 1.38, "learning_rate": 2.173533293341332e-06, "loss": 0.0061, "step": 11531 }, { "epoch": 1.38, "learning_rate": 2.1734613077384526e-06, "loss": 0.0002, "step": 11532 }, { "epoch": 1.38, "learning_rate": 2.173389322135573e-06, "loss": 0.0002, "step": 11533 }, { "epoch": 1.38, "learning_rate": 2.1733173365326937e-06, "loss": 0.0019, "step": 11534 }, { "epoch": 1.38, "learning_rate": 2.173245350929814e-06, "loss": 0.6451, "step": 11535 }, { "epoch": 1.38, "learning_rate": 2.173173365326935e-06, "loss": 0.0002, "step": 11536 }, { "epoch": 1.38, "learning_rate": 2.1731013797240552e-06, "loss": 0.0003, "step": 11537 }, { "epoch": 1.38, "learning_rate": 2.173029394121176e-06, "loss": 0.0007, "step": 11538 }, { "epoch": 1.38, "learning_rate": 2.1729574085182964e-06, "loss": 0.0002, "step": 11539 }, { "epoch": 1.38, "learning_rate": 2.1728854229154167e-06, "loss": 0.0004, "step": 11540 }, { "epoch": 1.38, "learning_rate": 2.1728134373125375e-06, "loss": 0.0002, "step": 11541 }, { "epoch": 1.38, "learning_rate": 2.172741451709658e-06, "loss": 1.0747, "step": 11542 }, { "epoch": 1.38, "learning_rate": 2.1726694661067786e-06, "loss": 0.0003, "step": 11543 }, { "epoch": 1.38, "learning_rate": 2.172597480503899e-06, "loss": 0.0003, "step": 11544 }, { "epoch": 1.38, "learning_rate": 2.1725254949010198e-06, "loss": 0.9468, "step": 11545 }, { "epoch": 1.38, "learning_rate": 2.1724535092981405e-06, "loss": 0.0002, "step": 11546 }, { "epoch": 1.38, "learning_rate": 2.1723815236952613e-06, "loss": 0.0003, "step": 11547 }, { "epoch": 1.38, "learning_rate": 2.1723095380923817e-06, "loss": 0.0004, "step": 11548 }, { "epoch": 1.38, "learning_rate": 2.1722375524895024e-06, "loss": 0.0003, "step": 11549 }, { "epoch": 1.38, "learning_rate": 2.172165566886623e-06, "loss": 0.0004, "step": 11550 }, { "epoch": 1.38, "learning_rate": 2.1720935812837436e-06, "loss": 0.0003, "step": 11551 }, { "epoch": 1.38, "learning_rate": 2.172021595680864e-06, "loss": 0.0002, "step": 11552 }, { "epoch": 1.38, "learning_rate": 2.1719496100779843e-06, "loss": 0.0003, "step": 11553 }, { "epoch": 1.38, "learning_rate": 2.171877624475105e-06, "loss": 0.0002, "step": 11554 }, { "epoch": 1.38, "learning_rate": 2.1718056388722254e-06, "loss": 0.0003, "step": 11555 }, { "epoch": 1.38, "learning_rate": 2.171733653269346e-06, "loss": 0.0005, "step": 11556 }, { "epoch": 1.38, "learning_rate": 2.1716616676664666e-06, "loss": 0.0005, "step": 11557 }, { "epoch": 1.39, "learning_rate": 2.1715896820635873e-06, "loss": 0.0004, "step": 11558 }, { "epoch": 1.39, "learning_rate": 2.1715176964607077e-06, "loss": 0.0003, "step": 11559 }, { "epoch": 1.39, "learning_rate": 2.1714457108578285e-06, "loss": 0.0003, "step": 11560 }, { "epoch": 1.39, "learning_rate": 2.1713737252549492e-06, "loss": 0.0003, "step": 11561 }, { "epoch": 1.39, "learning_rate": 2.1713017396520696e-06, "loss": 0.0004, "step": 11562 }, { "epoch": 1.39, "learning_rate": 2.1712297540491904e-06, "loss": 0.0003, "step": 11563 }, { "epoch": 1.39, "learning_rate": 2.1711577684463107e-06, "loss": 0.0007, "step": 11564 }, { "epoch": 1.39, "learning_rate": 2.1710857828434315e-06, "loss": 0.0004, "step": 11565 }, { "epoch": 1.39, "learning_rate": 2.171013797240552e-06, "loss": 0.0009, "step": 11566 }, { "epoch": 1.39, "learning_rate": 2.1709418116376726e-06, "loss": 0.0002, "step": 11567 }, { "epoch": 1.39, "learning_rate": 2.170869826034793e-06, "loss": 0.0004, "step": 11568 }, { "epoch": 1.39, "learning_rate": 2.1707978404319138e-06, "loss": 0.953, "step": 11569 }, { "epoch": 1.39, "learning_rate": 2.170725854829034e-06, "loss": 0.0003, "step": 11570 }, { "epoch": 1.39, "learning_rate": 2.170653869226155e-06, "loss": 0.0919, "step": 11571 }, { "epoch": 1.39, "learning_rate": 2.1705818836232753e-06, "loss": 0.0005, "step": 11572 }, { "epoch": 1.39, "learning_rate": 2.170509898020396e-06, "loss": 0.6983, "step": 11573 }, { "epoch": 1.39, "learning_rate": 2.1704379124175164e-06, "loss": 0.0003, "step": 11574 }, { "epoch": 1.39, "learning_rate": 2.1703659268146368e-06, "loss": 0.4539, "step": 11575 }, { "epoch": 1.39, "learning_rate": 2.1702939412117575e-06, "loss": 0.0003, "step": 11576 }, { "epoch": 1.39, "learning_rate": 2.1702219556088783e-06, "loss": 0.0004, "step": 11577 }, { "epoch": 1.39, "learning_rate": 2.170149970005999e-06, "loss": 0.0005, "step": 11578 }, { "epoch": 1.39, "learning_rate": 2.1700779844031195e-06, "loss": 0.0006, "step": 11579 }, { "epoch": 1.39, "learning_rate": 2.1700059988002402e-06, "loss": 0.0002, "step": 11580 }, { "epoch": 1.39, "learning_rate": 2.1699340131973606e-06, "loss": 0.9389, "step": 11581 }, { "epoch": 1.39, "learning_rate": 2.1698620275944814e-06, "loss": 0.0003, "step": 11582 }, { "epoch": 1.39, "learning_rate": 2.1697900419916017e-06, "loss": 0.0005, "step": 11583 }, { "epoch": 1.39, "learning_rate": 2.1697180563887225e-06, "loss": 0.0004, "step": 11584 }, { "epoch": 1.39, "learning_rate": 2.169646070785843e-06, "loss": 0.0003, "step": 11585 }, { "epoch": 1.39, "learning_rate": 2.169574085182963e-06, "loss": 0.0003, "step": 11586 }, { "epoch": 1.39, "learning_rate": 2.169502099580084e-06, "loss": 0.0004, "step": 11587 }, { "epoch": 1.39, "learning_rate": 2.1694301139772043e-06, "loss": 0.0006, "step": 11588 }, { "epoch": 1.39, "learning_rate": 2.169358128374325e-06, "loss": 0.0003, "step": 11589 }, { "epoch": 1.39, "learning_rate": 2.1692861427714455e-06, "loss": 0.0006, "step": 11590 }, { "epoch": 1.39, "learning_rate": 2.1692141571685663e-06, "loss": 0.0008, "step": 11591 }, { "epoch": 1.39, "learning_rate": 2.169142171565687e-06, "loss": 0.9445, "step": 11592 }, { "epoch": 1.39, "learning_rate": 2.1690701859628074e-06, "loss": 0.002, "step": 11593 }, { "epoch": 1.39, "learning_rate": 2.168998200359928e-06, "loss": 0.0003, "step": 11594 }, { "epoch": 1.39, "learning_rate": 2.168926214757049e-06, "loss": 0.0006, "step": 11595 }, { "epoch": 1.39, "learning_rate": 2.1688542291541693e-06, "loss": 0.0005, "step": 11596 }, { "epoch": 1.39, "learning_rate": 2.16878224355129e-06, "loss": 0.681, "step": 11597 }, { "epoch": 1.39, "learning_rate": 2.1687102579484104e-06, "loss": 0.0007, "step": 11598 }, { "epoch": 1.39, "learning_rate": 2.1686382723455308e-06, "loss": 0.0006, "step": 11599 }, { "epoch": 1.39, "learning_rate": 2.1685662867426516e-06, "loss": 0.0004, "step": 11600 }, { "epoch": 1.39, "learning_rate": 2.168494301139772e-06, "loss": 0.0003, "step": 11601 }, { "epoch": 1.39, "learning_rate": 2.1684223155368927e-06, "loss": 0.0005, "step": 11602 }, { "epoch": 1.39, "learning_rate": 2.168350329934013e-06, "loss": 0.0009, "step": 11603 }, { "epoch": 1.39, "learning_rate": 2.168278344331134e-06, "loss": 0.001, "step": 11604 }, { "epoch": 1.39, "learning_rate": 2.168206358728254e-06, "loss": 0.0004, "step": 11605 }, { "epoch": 1.39, "learning_rate": 2.168134373125375e-06, "loss": 0.0006, "step": 11606 }, { "epoch": 1.39, "learning_rate": 2.1680623875224953e-06, "loss": 0.0004, "step": 11607 }, { "epoch": 1.39, "learning_rate": 2.167990401919616e-06, "loss": 0.0006, "step": 11608 }, { "epoch": 1.39, "learning_rate": 2.167918416316737e-06, "loss": 0.0003, "step": 11609 }, { "epoch": 1.39, "learning_rate": 2.1678464307138572e-06, "loss": 0.001, "step": 11610 }, { "epoch": 1.39, "learning_rate": 2.167774445110978e-06, "loss": 0.0005, "step": 11611 }, { "epoch": 1.39, "learning_rate": 2.1677024595080984e-06, "loss": 0.0005, "step": 11612 }, { "epoch": 1.39, "learning_rate": 2.167630473905219e-06, "loss": 0.0002, "step": 11613 }, { "epoch": 1.39, "learning_rate": 2.1675584883023395e-06, "loss": 0.0005, "step": 11614 }, { "epoch": 1.39, "learning_rate": 2.1674865026994603e-06, "loss": 0.0005, "step": 11615 }, { "epoch": 1.39, "learning_rate": 2.1674145170965806e-06, "loss": 0.0003, "step": 11616 }, { "epoch": 1.39, "learning_rate": 2.1673425314937014e-06, "loss": 0.0002, "step": 11617 }, { "epoch": 1.39, "learning_rate": 2.1672705458908218e-06, "loss": 0.0002, "step": 11618 }, { "epoch": 1.39, "learning_rate": 2.1671985602879425e-06, "loss": 0.0005, "step": 11619 }, { "epoch": 1.39, "learning_rate": 2.167126574685063e-06, "loss": 0.0005, "step": 11620 }, { "epoch": 1.39, "learning_rate": 2.1670545890821833e-06, "loss": 0.0004, "step": 11621 }, { "epoch": 1.39, "learning_rate": 2.166982603479304e-06, "loss": 0.0004, "step": 11622 }, { "epoch": 1.39, "learning_rate": 2.166910617876425e-06, "loss": 0.0003, "step": 11623 }, { "epoch": 1.39, "learning_rate": 2.1668386322735456e-06, "loss": 0.0004, "step": 11624 }, { "epoch": 1.39, "learning_rate": 2.166766646670666e-06, "loss": 0.0004, "step": 11625 }, { "epoch": 1.39, "learning_rate": 2.1666946610677867e-06, "loss": 0.0006, "step": 11626 }, { "epoch": 1.39, "learning_rate": 2.166622675464907e-06, "loss": 0.0003, "step": 11627 }, { "epoch": 1.39, "learning_rate": 2.166550689862028e-06, "loss": 1.0468, "step": 11628 }, { "epoch": 1.39, "learning_rate": 2.1664787042591482e-06, "loss": 0.0007, "step": 11629 }, { "epoch": 1.39, "learning_rate": 2.166406718656269e-06, "loss": 0.0003, "step": 11630 }, { "epoch": 1.39, "learning_rate": 2.1663347330533893e-06, "loss": 0.6458, "step": 11631 }, { "epoch": 1.39, "learning_rate": 2.16626274745051e-06, "loss": 0.0006, "step": 11632 }, { "epoch": 1.39, "learning_rate": 2.1661907618476305e-06, "loss": 0.0005, "step": 11633 }, { "epoch": 1.39, "learning_rate": 2.166118776244751e-06, "loss": 0.0003, "step": 11634 }, { "epoch": 1.39, "learning_rate": 2.1660467906418716e-06, "loss": 0.0003, "step": 11635 }, { "epoch": 1.39, "learning_rate": 2.165974805038992e-06, "loss": 0.0007, "step": 11636 }, { "epoch": 1.39, "learning_rate": 2.1659028194361127e-06, "loss": 0.0004, "step": 11637 }, { "epoch": 1.39, "learning_rate": 2.1658308338332335e-06, "loss": 0.0008, "step": 11638 }, { "epoch": 1.39, "learning_rate": 2.165758848230354e-06, "loss": 0.0003, "step": 11639 }, { "epoch": 1.39, "learning_rate": 2.1656868626274747e-06, "loss": 0.0003, "step": 11640 }, { "epoch": 1.39, "learning_rate": 2.1656148770245954e-06, "loss": 0.0002, "step": 11641 }, { "epoch": 1.4, "learning_rate": 2.165542891421716e-06, "loss": 0.0005, "step": 11642 }, { "epoch": 1.4, "learning_rate": 2.1654709058188366e-06, "loss": 0.0004, "step": 11643 }, { "epoch": 1.4, "learning_rate": 2.165398920215957e-06, "loss": 0.0005, "step": 11644 }, { "epoch": 1.4, "learning_rate": 2.1653269346130773e-06, "loss": 0.0004, "step": 11645 }, { "epoch": 1.4, "learning_rate": 2.165254949010198e-06, "loss": 0.0002, "step": 11646 }, { "epoch": 1.4, "learning_rate": 2.1651829634073184e-06, "loss": 0.0009, "step": 11647 }, { "epoch": 1.4, "learning_rate": 2.165110977804439e-06, "loss": 0.0009, "step": 11648 }, { "epoch": 1.4, "learning_rate": 2.1650389922015595e-06, "loss": 0.0702, "step": 11649 }, { "epoch": 1.4, "learning_rate": 2.1649670065986803e-06, "loss": 0.0011, "step": 11650 }, { "epoch": 1.4, "learning_rate": 2.1648950209958007e-06, "loss": 0.0005, "step": 11651 }, { "epoch": 1.4, "learning_rate": 2.1648230353929215e-06, "loss": 0.0002, "step": 11652 }, { "epoch": 1.4, "learning_rate": 2.164751049790042e-06, "loss": 0.0007, "step": 11653 }, { "epoch": 1.4, "learning_rate": 2.1646790641871626e-06, "loss": 0.0004, "step": 11654 }, { "epoch": 1.4, "learning_rate": 2.1646070785842834e-06, "loss": 0.0006, "step": 11655 }, { "epoch": 1.4, "learning_rate": 2.1645350929814037e-06, "loss": 0.0003, "step": 11656 }, { "epoch": 1.4, "learning_rate": 2.1644631073785245e-06, "loss": 0.0003, "step": 11657 }, { "epoch": 1.4, "learning_rate": 2.164391121775645e-06, "loss": 0.001, "step": 11658 }, { "epoch": 1.4, "learning_rate": 2.1643191361727656e-06, "loss": 0.0018, "step": 11659 }, { "epoch": 1.4, "learning_rate": 2.164247150569886e-06, "loss": 0.0003, "step": 11660 }, { "epoch": 1.4, "learning_rate": 2.1641751649670068e-06, "loss": 0.0004, "step": 11661 }, { "epoch": 1.4, "learning_rate": 2.164103179364127e-06, "loss": 0.0002, "step": 11662 }, { "epoch": 1.4, "learning_rate": 2.164031193761248e-06, "loss": 0.3922, "step": 11663 }, { "epoch": 1.4, "learning_rate": 2.1639592081583683e-06, "loss": 0.0004, "step": 11664 }, { "epoch": 1.4, "learning_rate": 2.163887222555489e-06, "loss": 0.005, "step": 11665 }, { "epoch": 1.4, "learning_rate": 2.1638152369526094e-06, "loss": 0.0003, "step": 11666 }, { "epoch": 1.4, "learning_rate": 2.1637432513497298e-06, "loss": 0.001, "step": 11667 }, { "epoch": 1.4, "learning_rate": 2.1636712657468505e-06, "loss": 0.0004, "step": 11668 }, { "epoch": 1.4, "learning_rate": 2.1635992801439713e-06, "loss": 0.0005, "step": 11669 }, { "epoch": 1.4, "learning_rate": 2.163527294541092e-06, "loss": 0.4223, "step": 11670 }, { "epoch": 1.4, "learning_rate": 2.1634553089382124e-06, "loss": 0.0003, "step": 11671 }, { "epoch": 1.4, "learning_rate": 2.1633833233353332e-06, "loss": 0.0004, "step": 11672 }, { "epoch": 1.4, "learning_rate": 2.1633113377324536e-06, "loss": 0.0009, "step": 11673 }, { "epoch": 1.4, "learning_rate": 2.1632393521295744e-06, "loss": 0.0004, "step": 11674 }, { "epoch": 1.4, "learning_rate": 2.1631673665266947e-06, "loss": 0.2373, "step": 11675 }, { "epoch": 1.4, "learning_rate": 2.1630953809238155e-06, "loss": 0.0003, "step": 11676 }, { "epoch": 1.4, "learning_rate": 2.163023395320936e-06, "loss": 0.0004, "step": 11677 }, { "epoch": 1.4, "learning_rate": 2.1629514097180566e-06, "loss": 0.0004, "step": 11678 }, { "epoch": 1.4, "learning_rate": 2.162879424115177e-06, "loss": 0.0004, "step": 11679 }, { "epoch": 1.4, "learning_rate": 2.1628074385122973e-06, "loss": 0.0005, "step": 11680 }, { "epoch": 1.4, "learning_rate": 2.162735452909418e-06, "loss": 0.0029, "step": 11681 }, { "epoch": 1.4, "learning_rate": 2.1626634673065385e-06, "loss": 0.0004, "step": 11682 }, { "epoch": 1.4, "learning_rate": 2.1625914817036592e-06, "loss": 0.0002, "step": 11683 }, { "epoch": 1.4, "learning_rate": 2.16251949610078e-06, "loss": 0.0001, "step": 11684 }, { "epoch": 1.4, "learning_rate": 2.1624475104979004e-06, "loss": 0.0002, "step": 11685 }, { "epoch": 1.4, "learning_rate": 2.162375524895021e-06, "loss": 0.0005, "step": 11686 }, { "epoch": 1.4, "learning_rate": 2.162303539292142e-06, "loss": 0.0004, "step": 11687 }, { "epoch": 1.4, "learning_rate": 2.1622315536892623e-06, "loss": 0.0004, "step": 11688 }, { "epoch": 1.4, "learning_rate": 2.162159568086383e-06, "loss": 0.0002, "step": 11689 }, { "epoch": 1.4, "learning_rate": 2.1620875824835034e-06, "loss": 0.0003, "step": 11690 }, { "epoch": 1.4, "learning_rate": 2.1620155968806238e-06, "loss": 0.0009, "step": 11691 }, { "epoch": 1.4, "learning_rate": 2.1619436112777446e-06, "loss": 0.0006, "step": 11692 }, { "epoch": 1.4, "learning_rate": 2.161871625674865e-06, "loss": 0.0004, "step": 11693 }, { "epoch": 1.4, "learning_rate": 2.1617996400719857e-06, "loss": 0.0003, "step": 11694 }, { "epoch": 1.4, "learning_rate": 2.161727654469106e-06, "loss": 0.0005, "step": 11695 }, { "epoch": 1.4, "learning_rate": 2.161655668866227e-06, "loss": 0.0002, "step": 11696 }, { "epoch": 1.4, "learning_rate": 2.161583683263347e-06, "loss": 0.0005, "step": 11697 }, { "epoch": 1.4, "learning_rate": 2.161511697660468e-06, "loss": 0.0063, "step": 11698 }, { "epoch": 1.4, "learning_rate": 2.1614397120575883e-06, "loss": 0.0006, "step": 11699 }, { "epoch": 1.4, "learning_rate": 2.161367726454709e-06, "loss": 0.0003, "step": 11700 }, { "epoch": 1.4, "learning_rate": 2.16129574085183e-06, "loss": 0.0003, "step": 11701 }, { "epoch": 1.4, "learning_rate": 2.1612237552489502e-06, "loss": 0.6664, "step": 11702 }, { "epoch": 1.4, "learning_rate": 2.161151769646071e-06, "loss": 0.0006, "step": 11703 }, { "epoch": 1.4, "learning_rate": 2.1610797840431914e-06, "loss": 0.0003, "step": 11704 }, { "epoch": 1.4, "learning_rate": 2.161007798440312e-06, "loss": 0.0008, "step": 11705 }, { "epoch": 1.4, "learning_rate": 2.1609358128374325e-06, "loss": 0.0007, "step": 11706 }, { "epoch": 1.4, "learning_rate": 2.1608638272345533e-06, "loss": 0.0006, "step": 11707 }, { "epoch": 1.4, "learning_rate": 2.1607918416316736e-06, "loss": 0.0003, "step": 11708 }, { "epoch": 1.4, "learning_rate": 2.1607198560287944e-06, "loss": 0.0004, "step": 11709 }, { "epoch": 1.4, "learning_rate": 2.1606478704259148e-06, "loss": 0.0004, "step": 11710 }, { "epoch": 1.4, "learning_rate": 2.1605758848230355e-06, "loss": 0.0004, "step": 11711 }, { "epoch": 1.4, "learning_rate": 2.160503899220156e-06, "loss": 0.0008, "step": 11712 }, { "epoch": 1.4, "learning_rate": 2.1604319136172762e-06, "loss": 0.0002, "step": 11713 }, { "epoch": 1.4, "learning_rate": 2.160359928014397e-06, "loss": 0.0002, "step": 11714 }, { "epoch": 1.4, "learning_rate": 2.160287942411518e-06, "loss": 0.0096, "step": 11715 }, { "epoch": 1.4, "learning_rate": 2.160215956808638e-06, "loss": 0.4872, "step": 11716 }, { "epoch": 1.4, "learning_rate": 2.160143971205759e-06, "loss": 0.0088, "step": 11717 }, { "epoch": 1.4, "learning_rate": 2.1600719856028797e-06, "loss": 0.0171, "step": 11718 }, { "epoch": 1.4, "learning_rate": 2.16e-06, "loss": 0.0002, "step": 11719 }, { "epoch": 1.4, "learning_rate": 2.159928014397121e-06, "loss": 0.0006, "step": 11720 }, { "epoch": 1.4, "learning_rate": 2.159856028794241e-06, "loss": 0.0002, "step": 11721 }, { "epoch": 1.4, "learning_rate": 2.159784043191362e-06, "loss": 0.9993, "step": 11722 }, { "epoch": 1.4, "learning_rate": 2.1597120575884823e-06, "loss": 0.0004, "step": 11723 }, { "epoch": 1.4, "learning_rate": 2.159640071985603e-06, "loss": 0.9061, "step": 11724 }, { "epoch": 1.41, "learning_rate": 2.1595680863827235e-06, "loss": 0.0004, "step": 11725 }, { "epoch": 1.41, "learning_rate": 2.159496100779844e-06, "loss": 0.0005, "step": 11726 }, { "epoch": 1.41, "learning_rate": 2.1594241151769646e-06, "loss": 0.0006, "step": 11727 }, { "epoch": 1.41, "learning_rate": 2.159352129574085e-06, "loss": 0.0004, "step": 11728 }, { "epoch": 1.41, "learning_rate": 2.1592801439712057e-06, "loss": 0.0003, "step": 11729 }, { "epoch": 1.41, "learning_rate": 2.159208158368326e-06, "loss": 0.0053, "step": 11730 }, { "epoch": 1.41, "learning_rate": 2.159136172765447e-06, "loss": 0.0004, "step": 11731 }, { "epoch": 1.41, "learning_rate": 2.1590641871625677e-06, "loss": 0.0007, "step": 11732 }, { "epoch": 1.41, "learning_rate": 2.1589922015596884e-06, "loss": 0.0006, "step": 11733 }, { "epoch": 1.41, "learning_rate": 2.1589202159568088e-06, "loss": 0.0006, "step": 11734 }, { "epoch": 1.41, "learning_rate": 2.1588482303539296e-06, "loss": 0.0004, "step": 11735 }, { "epoch": 1.41, "learning_rate": 2.15877624475105e-06, "loss": 0.0002, "step": 11736 }, { "epoch": 1.41, "learning_rate": 2.1587042591481703e-06, "loss": 0.0005, "step": 11737 }, { "epoch": 1.41, "learning_rate": 2.158632273545291e-06, "loss": 0.0005, "step": 11738 }, { "epoch": 1.41, "learning_rate": 2.1585602879424114e-06, "loss": 0.0006, "step": 11739 }, { "epoch": 1.41, "learning_rate": 2.158488302339532e-06, "loss": 0.0003, "step": 11740 }, { "epoch": 1.41, "learning_rate": 2.1584163167366525e-06, "loss": 0.0003, "step": 11741 }, { "epoch": 1.41, "learning_rate": 2.1583443311337733e-06, "loss": 0.0022, "step": 11742 }, { "epoch": 1.41, "learning_rate": 2.1582723455308937e-06, "loss": 0.0009, "step": 11743 }, { "epoch": 1.41, "learning_rate": 2.1582003599280145e-06, "loss": 0.0004, "step": 11744 }, { "epoch": 1.41, "learning_rate": 2.158128374325135e-06, "loss": 0.0004, "step": 11745 }, { "epoch": 1.41, "learning_rate": 2.1580563887222556e-06, "loss": 0.0004, "step": 11746 }, { "epoch": 1.41, "learning_rate": 2.1579844031193764e-06, "loss": 0.0007, "step": 11747 }, { "epoch": 1.41, "learning_rate": 2.1579124175164967e-06, "loss": 0.0004, "step": 11748 }, { "epoch": 1.41, "learning_rate": 2.1578404319136175e-06, "loss": 0.0007, "step": 11749 }, { "epoch": 1.41, "learning_rate": 2.157768446310738e-06, "loss": 0.0005, "step": 11750 }, { "epoch": 1.41, "learning_rate": 2.1576964607078586e-06, "loss": 0.0002, "step": 11751 }, { "epoch": 1.41, "learning_rate": 2.157624475104979e-06, "loss": 0.0003, "step": 11752 }, { "epoch": 1.41, "learning_rate": 2.1575524895020998e-06, "loss": 0.0004, "step": 11753 }, { "epoch": 1.41, "learning_rate": 2.15748050389922e-06, "loss": 0.0004, "step": 11754 }, { "epoch": 1.41, "learning_rate": 2.157408518296341e-06, "loss": 0.0006, "step": 11755 }, { "epoch": 1.41, "learning_rate": 2.1573365326934613e-06, "loss": 0.0008, "step": 11756 }, { "epoch": 1.41, "learning_rate": 2.157264547090582e-06, "loss": 0.0056, "step": 11757 }, { "epoch": 1.41, "learning_rate": 2.1571925614877024e-06, "loss": 0.0003, "step": 11758 }, { "epoch": 1.41, "learning_rate": 2.157120575884823e-06, "loss": 0.0005, "step": 11759 }, { "epoch": 1.41, "learning_rate": 2.1570485902819435e-06, "loss": 0.0006, "step": 11760 }, { "epoch": 1.41, "learning_rate": 2.1569766046790643e-06, "loss": 0.0005, "step": 11761 }, { "epoch": 1.41, "learning_rate": 2.1569046190761847e-06, "loss": 0.0005, "step": 11762 }, { "epoch": 1.41, "learning_rate": 2.1568326334733054e-06, "loss": 0.0005, "step": 11763 }, { "epoch": 1.41, "learning_rate": 2.1567606478704262e-06, "loss": 0.0003, "step": 11764 }, { "epoch": 1.41, "learning_rate": 2.1566886622675466e-06, "loss": 0.0006, "step": 11765 }, { "epoch": 1.41, "learning_rate": 2.1566166766646673e-06, "loss": 0.8372, "step": 11766 }, { "epoch": 1.41, "learning_rate": 2.1565446910617877e-06, "loss": 0.0015, "step": 11767 }, { "epoch": 1.41, "learning_rate": 2.1564727054589085e-06, "loss": 0.0002, "step": 11768 }, { "epoch": 1.41, "learning_rate": 2.156400719856029e-06, "loss": 0.7691, "step": 11769 }, { "epoch": 1.41, "learning_rate": 2.1563287342531496e-06, "loss": 0.0002, "step": 11770 }, { "epoch": 1.41, "learning_rate": 2.15625674865027e-06, "loss": 0.0007, "step": 11771 }, { "epoch": 1.41, "learning_rate": 2.1561847630473903e-06, "loss": 0.0007, "step": 11772 }, { "epoch": 1.41, "learning_rate": 2.156112777444511e-06, "loss": 0.0003, "step": 11773 }, { "epoch": 1.41, "learning_rate": 2.1560407918416315e-06, "loss": 0.0004, "step": 11774 }, { "epoch": 1.41, "learning_rate": 2.1559688062387522e-06, "loss": 0.001, "step": 11775 }, { "epoch": 1.41, "learning_rate": 2.1558968206358726e-06, "loss": 0.0004, "step": 11776 }, { "epoch": 1.41, "learning_rate": 2.1558248350329934e-06, "loss": 0.0006, "step": 11777 }, { "epoch": 1.41, "learning_rate": 2.155752849430114e-06, "loss": 0.0006, "step": 11778 }, { "epoch": 1.41, "learning_rate": 2.155680863827235e-06, "loss": 0.0003, "step": 11779 }, { "epoch": 1.41, "learning_rate": 2.1556088782243553e-06, "loss": 0.0006, "step": 11780 }, { "epoch": 1.41, "learning_rate": 2.155536892621476e-06, "loss": 0.0005, "step": 11781 }, { "epoch": 1.41, "learning_rate": 2.1554649070185964e-06, "loss": 0.0005, "step": 11782 }, { "epoch": 1.41, "learning_rate": 2.1553929214157168e-06, "loss": 0.0006, "step": 11783 }, { "epoch": 1.41, "learning_rate": 2.1553209358128375e-06, "loss": 0.0004, "step": 11784 }, { "epoch": 1.41, "learning_rate": 2.155248950209958e-06, "loss": 0.0005, "step": 11785 }, { "epoch": 1.41, "learning_rate": 2.1551769646070787e-06, "loss": 0.0033, "step": 11786 }, { "epoch": 1.41, "learning_rate": 2.155104979004199e-06, "loss": 0.0006, "step": 11787 }, { "epoch": 1.41, "learning_rate": 2.15503299340132e-06, "loss": 0.0006, "step": 11788 }, { "epoch": 1.41, "learning_rate": 2.15496100779844e-06, "loss": 0.0005, "step": 11789 }, { "epoch": 1.41, "learning_rate": 2.154889022195561e-06, "loss": 0.0017, "step": 11790 }, { "epoch": 1.41, "learning_rate": 2.1548170365926813e-06, "loss": 0.0005, "step": 11791 }, { "epoch": 1.41, "learning_rate": 2.154745050989802e-06, "loss": 0.0002, "step": 11792 }, { "epoch": 1.41, "learning_rate": 2.154673065386923e-06, "loss": 0.0003, "step": 11793 }, { "epoch": 1.41, "learning_rate": 2.1546010797840432e-06, "loss": 0.502, "step": 11794 }, { "epoch": 1.41, "learning_rate": 2.154529094181164e-06, "loss": 0.0002, "step": 11795 }, { "epoch": 1.41, "learning_rate": 2.1544571085782843e-06, "loss": 0.0014, "step": 11796 }, { "epoch": 1.41, "learning_rate": 2.154385122975405e-06, "loss": 0.0002, "step": 11797 }, { "epoch": 1.41, "learning_rate": 2.1543131373725255e-06, "loss": 0.9552, "step": 11798 }, { "epoch": 1.41, "learning_rate": 2.1542411517696463e-06, "loss": 0.0004, "step": 11799 }, { "epoch": 1.41, "learning_rate": 2.1541691661667666e-06, "loss": 0.0003, "step": 11800 }, { "epoch": 1.41, "learning_rate": 2.1540971805638874e-06, "loss": 0.0004, "step": 11801 }, { "epoch": 1.41, "learning_rate": 2.1540251949610078e-06, "loss": 0.0008, "step": 11802 }, { "epoch": 1.41, "learning_rate": 2.1539532093581285e-06, "loss": 0.0002, "step": 11803 }, { "epoch": 1.41, "learning_rate": 2.153881223755249e-06, "loss": 0.0071, "step": 11804 }, { "epoch": 1.41, "learning_rate": 2.1538092381523697e-06, "loss": 0.0004, "step": 11805 }, { "epoch": 1.41, "learning_rate": 2.15373725254949e-06, "loss": 0.0005, "step": 11806 }, { "epoch": 1.41, "learning_rate": 2.153665266946611e-06, "loss": 0.0002, "step": 11807 }, { "epoch": 1.41, "learning_rate": 2.153593281343731e-06, "loss": 0.0006, "step": 11808 }, { "epoch": 1.42, "learning_rate": 2.153521295740852e-06, "loss": 0.0008, "step": 11809 }, { "epoch": 1.42, "learning_rate": 2.1534493101379727e-06, "loss": 0.001, "step": 11810 }, { "epoch": 1.42, "learning_rate": 2.153377324535093e-06, "loss": 0.678, "step": 11811 }, { "epoch": 1.42, "learning_rate": 2.153305338932214e-06, "loss": 0.0006, "step": 11812 }, { "epoch": 1.42, "learning_rate": 2.153233353329334e-06, "loss": 0.0004, "step": 11813 }, { "epoch": 1.42, "learning_rate": 2.153161367726455e-06, "loss": 0.0006, "step": 11814 }, { "epoch": 1.42, "learning_rate": 2.1530893821235753e-06, "loss": 0.0006, "step": 11815 }, { "epoch": 1.42, "learning_rate": 2.153017396520696e-06, "loss": 0.0003, "step": 11816 }, { "epoch": 1.42, "learning_rate": 2.1529454109178165e-06, "loss": 0.0005, "step": 11817 }, { "epoch": 1.42, "learning_rate": 2.152873425314937e-06, "loss": 0.0003, "step": 11818 }, { "epoch": 1.42, "learning_rate": 2.1528014397120576e-06, "loss": 0.0015, "step": 11819 }, { "epoch": 1.42, "learning_rate": 2.152729454109178e-06, "loss": 0.9741, "step": 11820 }, { "epoch": 1.42, "learning_rate": 2.1526574685062987e-06, "loss": 0.0007, "step": 11821 }, { "epoch": 1.42, "learning_rate": 2.152585482903419e-06, "loss": 0.0006, "step": 11822 }, { "epoch": 1.42, "learning_rate": 2.15251349730054e-06, "loss": 0.7173, "step": 11823 }, { "epoch": 1.42, "learning_rate": 2.1524415116976606e-06, "loss": 0.0007, "step": 11824 }, { "epoch": 1.42, "learning_rate": 2.1523695260947814e-06, "loss": 0.0007, "step": 11825 }, { "epoch": 1.42, "learning_rate": 2.1522975404919018e-06, "loss": 0.0004, "step": 11826 }, { "epoch": 1.42, "learning_rate": 2.1522255548890226e-06, "loss": 0.0005, "step": 11827 }, { "epoch": 1.42, "learning_rate": 2.152153569286143e-06, "loss": 0.0011, "step": 11828 }, { "epoch": 1.42, "learning_rate": 2.1520815836832633e-06, "loss": 0.0017, "step": 11829 }, { "epoch": 1.42, "learning_rate": 2.152009598080384e-06, "loss": 0.0005, "step": 11830 }, { "epoch": 1.42, "learning_rate": 2.1519376124775044e-06, "loss": 0.0005, "step": 11831 }, { "epoch": 1.42, "learning_rate": 2.151865626874625e-06, "loss": 0.0006, "step": 11832 }, { "epoch": 1.42, "learning_rate": 2.1517936412717455e-06, "loss": 0.0008, "step": 11833 }, { "epoch": 1.42, "learning_rate": 2.1517216556688663e-06, "loss": 0.0005, "step": 11834 }, { "epoch": 1.42, "learning_rate": 2.1516496700659867e-06, "loss": 0.0002, "step": 11835 }, { "epoch": 1.42, "learning_rate": 2.1515776844631074e-06, "loss": 0.0004, "step": 11836 }, { "epoch": 1.42, "learning_rate": 2.151505698860228e-06, "loss": 0.2157, "step": 11837 }, { "epoch": 1.42, "learning_rate": 2.1514337132573486e-06, "loss": 0.0002, "step": 11838 }, { "epoch": 1.42, "learning_rate": 2.1513617276544694e-06, "loss": 0.001, "step": 11839 }, { "epoch": 1.42, "learning_rate": 2.1512897420515897e-06, "loss": 0.0007, "step": 11840 }, { "epoch": 1.42, "learning_rate": 2.1512177564487105e-06, "loss": 0.003, "step": 11841 }, { "epoch": 1.42, "learning_rate": 2.151145770845831e-06, "loss": 0.0008, "step": 11842 }, { "epoch": 1.42, "learning_rate": 2.1510737852429516e-06, "loss": 0.0004, "step": 11843 }, { "epoch": 1.42, "learning_rate": 2.151001799640072e-06, "loss": 0.0004, "step": 11844 }, { "epoch": 1.42, "learning_rate": 2.1509298140371928e-06, "loss": 0.0042, "step": 11845 }, { "epoch": 1.42, "learning_rate": 2.150857828434313e-06, "loss": 0.0008, "step": 11846 }, { "epoch": 1.42, "learning_rate": 2.150785842831434e-06, "loss": 0.0003, "step": 11847 }, { "epoch": 1.42, "learning_rate": 2.1507138572285542e-06, "loss": 0.0005, "step": 11848 }, { "epoch": 1.42, "learning_rate": 2.150641871625675e-06, "loss": 0.0007, "step": 11849 }, { "epoch": 1.42, "learning_rate": 2.1505698860227954e-06, "loss": 0.0006, "step": 11850 }, { "epoch": 1.42, "learning_rate": 2.150497900419916e-06, "loss": 0.0007, "step": 11851 }, { "epoch": 1.42, "learning_rate": 2.1504259148170365e-06, "loss": 0.0003, "step": 11852 }, { "epoch": 1.42, "learning_rate": 2.150353929214157e-06, "loss": 0.0005, "step": 11853 }, { "epoch": 1.42, "learning_rate": 2.1502819436112776e-06, "loss": 0.0002, "step": 11854 }, { "epoch": 1.42, "learning_rate": 2.1502099580083984e-06, "loss": 0.0003, "step": 11855 }, { "epoch": 1.42, "learning_rate": 2.150137972405519e-06, "loss": 0.8492, "step": 11856 }, { "epoch": 1.42, "learning_rate": 2.1500659868026396e-06, "loss": 0.0004, "step": 11857 }, { "epoch": 1.42, "learning_rate": 2.1499940011997603e-06, "loss": 0.0004, "step": 11858 }, { "epoch": 1.42, "learning_rate": 2.1499220155968807e-06, "loss": 0.0004, "step": 11859 }, { "epoch": 1.42, "learning_rate": 2.1498500299940015e-06, "loss": 0.0006, "step": 11860 }, { "epoch": 1.42, "learning_rate": 2.149778044391122e-06, "loss": 0.0003, "step": 11861 }, { "epoch": 1.42, "learning_rate": 2.1497060587882426e-06, "loss": 0.0005, "step": 11862 }, { "epoch": 1.42, "learning_rate": 2.149634073185363e-06, "loss": 0.0004, "step": 11863 }, { "epoch": 1.42, "learning_rate": 2.1495620875824833e-06, "loss": 0.0003, "step": 11864 }, { "epoch": 1.42, "learning_rate": 2.149490101979604e-06, "loss": 0.0008, "step": 11865 }, { "epoch": 1.42, "learning_rate": 2.1494181163767244e-06, "loss": 0.0003, "step": 11866 }, { "epoch": 1.42, "learning_rate": 2.1493461307738452e-06, "loss": 0.0004, "step": 11867 }, { "epoch": 1.42, "learning_rate": 2.1492741451709656e-06, "loss": 0.704, "step": 11868 }, { "epoch": 1.42, "learning_rate": 2.1492021595680864e-06, "loss": 0.0005, "step": 11869 }, { "epoch": 1.42, "learning_rate": 2.149130173965207e-06, "loss": 0.4557, "step": 11870 }, { "epoch": 1.42, "learning_rate": 2.1490581883623275e-06, "loss": 0.0002, "step": 11871 }, { "epoch": 1.42, "learning_rate": 2.1489862027594483e-06, "loss": 0.0023, "step": 11872 }, { "epoch": 1.42, "learning_rate": 2.148914217156569e-06, "loss": 0.0004, "step": 11873 }, { "epoch": 1.42, "learning_rate": 2.1488422315536894e-06, "loss": 0.0005, "step": 11874 }, { "epoch": 1.42, "learning_rate": 2.1487702459508098e-06, "loss": 0.0007, "step": 11875 }, { "epoch": 1.42, "learning_rate": 2.1486982603479305e-06, "loss": 0.0007, "step": 11876 }, { "epoch": 1.42, "learning_rate": 2.148626274745051e-06, "loss": 0.0003, "step": 11877 }, { "epoch": 1.42, "learning_rate": 2.1485542891421717e-06, "loss": 0.0007, "step": 11878 }, { "epoch": 1.42, "learning_rate": 2.148482303539292e-06, "loss": 0.0007, "step": 11879 }, { "epoch": 1.42, "learning_rate": 2.148410317936413e-06, "loss": 0.0009, "step": 11880 }, { "epoch": 1.42, "learning_rate": 2.148338332333533e-06, "loss": 0.1164, "step": 11881 }, { "epoch": 1.42, "learning_rate": 2.148266346730654e-06, "loss": 0.0003, "step": 11882 }, { "epoch": 1.42, "learning_rate": 2.1481943611277743e-06, "loss": 0.0006, "step": 11883 }, { "epoch": 1.42, "learning_rate": 2.148122375524895e-06, "loss": 0.0002, "step": 11884 }, { "epoch": 1.42, "learning_rate": 2.1480503899220154e-06, "loss": 0.0007, "step": 11885 }, { "epoch": 1.42, "learning_rate": 2.147978404319136e-06, "loss": 0.0005, "step": 11886 }, { "epoch": 1.42, "learning_rate": 2.147906418716257e-06, "loss": 0.0006, "step": 11887 }, { "epoch": 1.42, "learning_rate": 2.1478344331133773e-06, "loss": 0.0004, "step": 11888 }, { "epoch": 1.42, "learning_rate": 2.147762447510498e-06, "loss": 0.0004, "step": 11889 }, { "epoch": 1.42, "learning_rate": 2.1476904619076185e-06, "loss": 0.0004, "step": 11890 }, { "epoch": 1.42, "learning_rate": 2.1476184763047393e-06, "loss": 0.0004, "step": 11891 }, { "epoch": 1.43, "learning_rate": 2.1475464907018596e-06, "loss": 0.0003, "step": 11892 }, { "epoch": 1.43, "learning_rate": 2.1474745050989804e-06, "loss": 0.0004, "step": 11893 }, { "epoch": 1.43, "learning_rate": 2.1474025194961007e-06, "loss": 0.0004, "step": 11894 }, { "epoch": 1.43, "learning_rate": 2.1473305338932215e-06, "loss": 0.0006, "step": 11895 }, { "epoch": 1.43, "learning_rate": 2.147258548290342e-06, "loss": 0.0004, "step": 11896 }, { "epoch": 1.43, "learning_rate": 2.1471865626874627e-06, "loss": 0.6902, "step": 11897 }, { "epoch": 1.43, "learning_rate": 2.147114577084583e-06, "loss": 0.001, "step": 11898 }, { "epoch": 1.43, "learning_rate": 2.1470425914817034e-06, "loss": 0.0004, "step": 11899 }, { "epoch": 1.43, "learning_rate": 2.146970605878824e-06, "loss": 0.0005, "step": 11900 }, { "epoch": 1.43, "learning_rate": 2.146898620275945e-06, "loss": 0.0004, "step": 11901 }, { "epoch": 1.43, "learning_rate": 2.1468266346730657e-06, "loss": 0.001, "step": 11902 }, { "epoch": 1.43, "learning_rate": 2.146754649070186e-06, "loss": 0.0004, "step": 11903 }, { "epoch": 1.43, "learning_rate": 2.146682663467307e-06, "loss": 0.0004, "step": 11904 }, { "epoch": 1.43, "learning_rate": 2.146610677864427e-06, "loss": 0.0005, "step": 11905 }, { "epoch": 1.43, "learning_rate": 2.146538692261548e-06, "loss": 0.0006, "step": 11906 }, { "epoch": 1.43, "learning_rate": 2.1464667066586683e-06, "loss": 0.0004, "step": 11907 }, { "epoch": 1.43, "learning_rate": 2.146394721055789e-06, "loss": 0.0003, "step": 11908 }, { "epoch": 1.43, "learning_rate": 2.1463227354529095e-06, "loss": 0.0003, "step": 11909 }, { "epoch": 1.43, "learning_rate": 2.14625074985003e-06, "loss": 0.0003, "step": 11910 }, { "epoch": 1.43, "learning_rate": 2.1461787642471506e-06, "loss": 0.0004, "step": 11911 }, { "epoch": 1.43, "learning_rate": 2.146106778644271e-06, "loss": 0.0005, "step": 11912 }, { "epoch": 1.43, "learning_rate": 2.1460347930413917e-06, "loss": 0.0005, "step": 11913 }, { "epoch": 1.43, "learning_rate": 2.145962807438512e-06, "loss": 0.0002, "step": 11914 }, { "epoch": 1.43, "learning_rate": 2.145890821835633e-06, "loss": 0.0004, "step": 11915 }, { "epoch": 1.43, "learning_rate": 2.1458188362327536e-06, "loss": 0.0003, "step": 11916 }, { "epoch": 1.43, "learning_rate": 2.145746850629874e-06, "loss": 0.0003, "step": 11917 }, { "epoch": 1.43, "learning_rate": 2.1456748650269948e-06, "loss": 0.0004, "step": 11918 }, { "epoch": 1.43, "learning_rate": 2.1456028794241155e-06, "loss": 0.0004, "step": 11919 }, { "epoch": 1.43, "learning_rate": 2.145530893821236e-06, "loss": 0.0005, "step": 11920 }, { "epoch": 1.43, "learning_rate": 2.1454589082183563e-06, "loss": 0.0005, "step": 11921 }, { "epoch": 1.43, "learning_rate": 2.145386922615477e-06, "loss": 0.6402, "step": 11922 }, { "epoch": 1.43, "learning_rate": 2.1453149370125974e-06, "loss": 0.0003, "step": 11923 }, { "epoch": 1.43, "learning_rate": 2.145242951409718e-06, "loss": 0.0011, "step": 11924 }, { "epoch": 1.43, "learning_rate": 2.1451709658068385e-06, "loss": 0.0004, "step": 11925 }, { "epoch": 1.43, "learning_rate": 2.1450989802039593e-06, "loss": 0.3399, "step": 11926 }, { "epoch": 1.43, "learning_rate": 2.1450269946010797e-06, "loss": 0.0002, "step": 11927 }, { "epoch": 1.43, "learning_rate": 2.1449550089982004e-06, "loss": 0.0007, "step": 11928 }, { "epoch": 1.43, "learning_rate": 2.144883023395321e-06, "loss": 0.0003, "step": 11929 }, { "epoch": 1.43, "learning_rate": 2.1448110377924416e-06, "loss": 0.0005, "step": 11930 }, { "epoch": 1.43, "learning_rate": 2.144739052189562e-06, "loss": 0.0003, "step": 11931 }, { "epoch": 1.43, "learning_rate": 2.1446670665866827e-06, "loss": 0.0314, "step": 11932 }, { "epoch": 1.43, "learning_rate": 2.1445950809838035e-06, "loss": 0.0004, "step": 11933 }, { "epoch": 1.43, "learning_rate": 2.144523095380924e-06, "loss": 0.0002, "step": 11934 }, { "epoch": 1.43, "learning_rate": 2.1444511097780446e-06, "loss": 0.0004, "step": 11935 }, { "epoch": 1.43, "learning_rate": 2.144379124175165e-06, "loss": 0.0004, "step": 11936 }, { "epoch": 1.43, "learning_rate": 2.1443071385722858e-06, "loss": 0.0142, "step": 11937 }, { "epoch": 1.43, "learning_rate": 2.144235152969406e-06, "loss": 0.0003, "step": 11938 }, { "epoch": 1.43, "learning_rate": 2.144163167366527e-06, "loss": 0.0003, "step": 11939 }, { "epoch": 1.43, "learning_rate": 2.1440911817636472e-06, "loss": 0.0005, "step": 11940 }, { "epoch": 1.43, "learning_rate": 2.144019196160768e-06, "loss": 0.0009, "step": 11941 }, { "epoch": 1.43, "learning_rate": 2.1439472105578884e-06, "loss": 0.8124, "step": 11942 }, { "epoch": 1.43, "learning_rate": 2.143875224955009e-06, "loss": 0.0034, "step": 11943 }, { "epoch": 1.43, "learning_rate": 2.1438032393521295e-06, "loss": 0.0002, "step": 11944 }, { "epoch": 1.43, "learning_rate": 2.14373125374925e-06, "loss": 0.0002, "step": 11945 }, { "epoch": 1.43, "learning_rate": 2.1436592681463706e-06, "loss": 0.0001, "step": 11946 }, { "epoch": 1.43, "learning_rate": 2.1435872825434914e-06, "loss": 0.0003, "step": 11947 }, { "epoch": 1.43, "learning_rate": 2.143515296940612e-06, "loss": 0.0004, "step": 11948 }, { "epoch": 1.43, "learning_rate": 2.1434433113377326e-06, "loss": 0.0004, "step": 11949 }, { "epoch": 1.43, "learning_rate": 2.1433713257348533e-06, "loss": 0.0004, "step": 11950 }, { "epoch": 1.43, "learning_rate": 2.1432993401319737e-06, "loss": 0.0002, "step": 11951 }, { "epoch": 1.43, "learning_rate": 2.1432273545290945e-06, "loss": 0.0004, "step": 11952 }, { "epoch": 1.43, "learning_rate": 2.143155368926215e-06, "loss": 0.0004, "step": 11953 }, { "epoch": 1.43, "learning_rate": 2.1430833833233356e-06, "loss": 0.0012, "step": 11954 }, { "epoch": 1.43, "learning_rate": 2.143011397720456e-06, "loss": 0.0005, "step": 11955 }, { "epoch": 1.43, "learning_rate": 2.1429394121175763e-06, "loss": 0.0005, "step": 11956 }, { "epoch": 1.43, "learning_rate": 2.142867426514697e-06, "loss": 0.0003, "step": 11957 }, { "epoch": 1.43, "learning_rate": 2.1427954409118174e-06, "loss": 0.0001, "step": 11958 }, { "epoch": 1.43, "learning_rate": 2.1427234553089382e-06, "loss": 0.0008, "step": 11959 }, { "epoch": 1.43, "learning_rate": 2.1426514697060586e-06, "loss": 0.001, "step": 11960 }, { "epoch": 1.43, "learning_rate": 2.1425794841031794e-06, "loss": 0.0005, "step": 11961 }, { "epoch": 1.43, "learning_rate": 2.1425074985003e-06, "loss": 0.0003, "step": 11962 }, { "epoch": 1.43, "learning_rate": 2.1424355128974205e-06, "loss": 0.0002, "step": 11963 }, { "epoch": 1.43, "learning_rate": 2.1423635272945413e-06, "loss": 0.0004, "step": 11964 }, { "epoch": 1.43, "learning_rate": 2.142291541691662e-06, "loss": 0.0004, "step": 11965 }, { "epoch": 1.43, "learning_rate": 2.1422195560887824e-06, "loss": 0.0007, "step": 11966 }, { "epoch": 1.43, "learning_rate": 2.142147570485903e-06, "loss": 0.0004, "step": 11967 }, { "epoch": 1.43, "learning_rate": 2.1420755848830235e-06, "loss": 0.0003, "step": 11968 }, { "epoch": 1.43, "learning_rate": 2.142003599280144e-06, "loss": 0.0003, "step": 11969 }, { "epoch": 1.43, "learning_rate": 2.1419316136772647e-06, "loss": 0.0008, "step": 11970 }, { "epoch": 1.43, "learning_rate": 2.141859628074385e-06, "loss": 0.0208, "step": 11971 }, { "epoch": 1.43, "learning_rate": 2.141787642471506e-06, "loss": 0.0008, "step": 11972 }, { "epoch": 1.43, "learning_rate": 2.141715656868626e-06, "loss": 0.0005, "step": 11973 }, { "epoch": 1.43, "learning_rate": 2.141643671265747e-06, "loss": 0.0005, "step": 11974 }, { "epoch": 1.43, "learning_rate": 2.1415716856628673e-06, "loss": 0.0003, "step": 11975 }, { "epoch": 1.44, "learning_rate": 2.141499700059988e-06, "loss": 0.002, "step": 11976 }, { "epoch": 1.44, "learning_rate": 2.1414277144571084e-06, "loss": 0.0004, "step": 11977 }, { "epoch": 1.44, "learning_rate": 2.141355728854229e-06, "loss": 0.0002, "step": 11978 }, { "epoch": 1.44, "learning_rate": 2.14128374325135e-06, "loss": 0.0005, "step": 11979 }, { "epoch": 1.44, "learning_rate": 2.1412117576484703e-06, "loss": 0.002, "step": 11980 }, { "epoch": 1.44, "learning_rate": 2.141139772045591e-06, "loss": 0.0003, "step": 11981 }, { "epoch": 1.44, "learning_rate": 2.1410677864427115e-06, "loss": 0.0003, "step": 11982 }, { "epoch": 1.44, "learning_rate": 2.1409958008398322e-06, "loss": 0.0002, "step": 11983 }, { "epoch": 1.44, "learning_rate": 2.1409238152369526e-06, "loss": 0.0004, "step": 11984 }, { "epoch": 1.44, "learning_rate": 2.1408518296340734e-06, "loss": 0.1779, "step": 11985 }, { "epoch": 1.44, "learning_rate": 2.1407798440311937e-06, "loss": 0.0002, "step": 11986 }, { "epoch": 1.44, "learning_rate": 2.1407078584283145e-06, "loss": 0.0003, "step": 11987 }, { "epoch": 1.44, "learning_rate": 2.140635872825435e-06, "loss": 0.0005, "step": 11988 }, { "epoch": 1.44, "learning_rate": 2.1405638872225556e-06, "loss": 0.0004, "step": 11989 }, { "epoch": 1.44, "learning_rate": 2.140491901619676e-06, "loss": 0.0005, "step": 11990 }, { "epoch": 1.44, "learning_rate": 2.1404199160167964e-06, "loss": 0.0004, "step": 11991 }, { "epoch": 1.44, "learning_rate": 2.140347930413917e-06, "loss": 0.0004, "step": 11992 }, { "epoch": 1.44, "learning_rate": 2.140275944811038e-06, "loss": 0.0003, "step": 11993 }, { "epoch": 1.44, "learning_rate": 2.1402039592081583e-06, "loss": 0.0004, "step": 11994 }, { "epoch": 1.44, "learning_rate": 2.140131973605279e-06, "loss": 0.0002, "step": 11995 }, { "epoch": 1.44, "learning_rate": 2.1400599880024e-06, "loss": 0.0004, "step": 11996 }, { "epoch": 1.44, "learning_rate": 2.13998800239952e-06, "loss": 0.0003, "step": 11997 }, { "epoch": 1.44, "learning_rate": 2.139916016796641e-06, "loss": 0.0004, "step": 11998 }, { "epoch": 1.44, "learning_rate": 2.1398440311937613e-06, "loss": 0.0007, "step": 11999 }, { "epoch": 1.44, "learning_rate": 2.139772045590882e-06, "loss": 0.0005, "step": 12000 }, { "epoch": 1.44, "learning_rate": 2.1397000599880024e-06, "loss": 0.001, "step": 12001 }, { "epoch": 1.44, "learning_rate": 2.139628074385123e-06, "loss": 0.0002, "step": 12002 }, { "epoch": 1.44, "learning_rate": 2.1395560887822436e-06, "loss": 0.0003, "step": 12003 }, { "epoch": 1.44, "learning_rate": 2.139484103179364e-06, "loss": 0.0002, "step": 12004 }, { "epoch": 1.44, "learning_rate": 2.1394121175764847e-06, "loss": 0.0003, "step": 12005 }, { "epoch": 1.44, "learning_rate": 2.139340131973605e-06, "loss": 0.0002, "step": 12006 }, { "epoch": 1.44, "learning_rate": 2.139268146370726e-06, "loss": 0.0005, "step": 12007 }, { "epoch": 1.44, "learning_rate": 2.139196160767846e-06, "loss": 0.0005, "step": 12008 }, { "epoch": 1.44, "learning_rate": 2.139124175164967e-06, "loss": 0.0005, "step": 12009 }, { "epoch": 1.44, "learning_rate": 2.1390521895620878e-06, "loss": 0.0006, "step": 12010 }, { "epoch": 1.44, "learning_rate": 2.1389802039592085e-06, "loss": 0.0003, "step": 12011 }, { "epoch": 1.44, "learning_rate": 2.138908218356329e-06, "loss": 0.0003, "step": 12012 }, { "epoch": 1.44, "learning_rate": 2.1388362327534497e-06, "loss": 0.0003, "step": 12013 }, { "epoch": 1.44, "learning_rate": 2.13876424715057e-06, "loss": 0.0003, "step": 12014 }, { "epoch": 1.44, "learning_rate": 2.1386922615476904e-06, "loss": 0.0003, "step": 12015 }, { "epoch": 1.44, "learning_rate": 2.138620275944811e-06, "loss": 0.0005, "step": 12016 }, { "epoch": 1.44, "learning_rate": 2.1385482903419315e-06, "loss": 0.0003, "step": 12017 }, { "epoch": 1.44, "learning_rate": 2.1384763047390523e-06, "loss": 0.0123, "step": 12018 }, { "epoch": 1.44, "learning_rate": 2.1384043191361727e-06, "loss": 0.0004, "step": 12019 }, { "epoch": 1.44, "learning_rate": 2.1383323335332934e-06, "loss": 0.002, "step": 12020 }, { "epoch": 1.44, "learning_rate": 2.1382603479304138e-06, "loss": 0.0004, "step": 12021 }, { "epoch": 1.44, "learning_rate": 2.1381883623275346e-06, "loss": 0.0003, "step": 12022 }, { "epoch": 1.44, "learning_rate": 2.138116376724655e-06, "loss": 0.0002, "step": 12023 }, { "epoch": 1.44, "learning_rate": 2.1380443911217757e-06, "loss": 0.0002, "step": 12024 }, { "epoch": 1.44, "learning_rate": 2.1379724055188965e-06, "loss": 0.0003, "step": 12025 }, { "epoch": 1.44, "learning_rate": 2.137900419916017e-06, "loss": 0.0003, "step": 12026 }, { "epoch": 1.44, "learning_rate": 2.1378284343131376e-06, "loss": 0.0003, "step": 12027 }, { "epoch": 1.44, "learning_rate": 2.137756448710258e-06, "loss": 0.0003, "step": 12028 }, { "epoch": 1.44, "learning_rate": 2.1376844631073787e-06, "loss": 0.0127, "step": 12029 }, { "epoch": 1.44, "learning_rate": 2.137612477504499e-06, "loss": 0.0003, "step": 12030 }, { "epoch": 1.44, "learning_rate": 2.13754049190162e-06, "loss": 0.0006, "step": 12031 }, { "epoch": 1.44, "learning_rate": 2.1374685062987402e-06, "loss": 0.0003, "step": 12032 }, { "epoch": 1.44, "learning_rate": 2.137396520695861e-06, "loss": 0.0002, "step": 12033 }, { "epoch": 1.44, "learning_rate": 2.1373245350929814e-06, "loss": 0.0003, "step": 12034 }, { "epoch": 1.44, "learning_rate": 2.137252549490102e-06, "loss": 0.0003, "step": 12035 }, { "epoch": 1.44, "learning_rate": 2.1371805638872225e-06, "loss": 0.0004, "step": 12036 }, { "epoch": 1.44, "learning_rate": 2.137108578284343e-06, "loss": 0.0003, "step": 12037 }, { "epoch": 1.44, "learning_rate": 2.1370365926814636e-06, "loss": 0.0002, "step": 12038 }, { "epoch": 1.44, "learning_rate": 2.1369646070785844e-06, "loss": 0.0002, "step": 12039 }, { "epoch": 1.44, "learning_rate": 2.1368926214757048e-06, "loss": 0.0003, "step": 12040 }, { "epoch": 1.44, "learning_rate": 2.1368206358728255e-06, "loss": 0.0005, "step": 12041 }, { "epoch": 1.44, "learning_rate": 2.1367486502699463e-06, "loss": 0.0003, "step": 12042 }, { "epoch": 1.44, "learning_rate": 2.1366766646670667e-06, "loss": 0.0002, "step": 12043 }, { "epoch": 1.44, "learning_rate": 2.1366046790641875e-06, "loss": 0.0002, "step": 12044 }, { "epoch": 1.44, "learning_rate": 2.136532693461308e-06, "loss": 0.0003, "step": 12045 }, { "epoch": 1.44, "learning_rate": 2.1364607078584286e-06, "loss": 0.001, "step": 12046 }, { "epoch": 1.44, "learning_rate": 2.136388722255549e-06, "loss": 0.0003, "step": 12047 }, { "epoch": 1.44, "learning_rate": 2.1363167366526693e-06, "loss": 0.0003, "step": 12048 }, { "epoch": 1.44, "learning_rate": 2.13624475104979e-06, "loss": 0.0003, "step": 12049 }, { "epoch": 1.44, "learning_rate": 2.1361727654469104e-06, "loss": 0.0003, "step": 12050 }, { "epoch": 1.44, "learning_rate": 2.136100779844031e-06, "loss": 0.0004, "step": 12051 }, { "epoch": 1.44, "learning_rate": 2.1360287942411516e-06, "loss": 0.0004, "step": 12052 }, { "epoch": 1.44, "learning_rate": 2.1359568086382723e-06, "loss": 0.0002, "step": 12053 }, { "epoch": 1.44, "learning_rate": 2.1358848230353927e-06, "loss": 0.0003, "step": 12054 }, { "epoch": 1.44, "learning_rate": 2.1358128374325135e-06, "loss": 0.0002, "step": 12055 }, { "epoch": 1.44, "learning_rate": 2.1357408518296343e-06, "loss": 0.0003, "step": 12056 }, { "epoch": 1.44, "learning_rate": 2.135668866226755e-06, "loss": 0.0002, "step": 12057 }, { "epoch": 1.44, "learning_rate": 2.1355968806238754e-06, "loss": 0.0004, "step": 12058 }, { "epoch": 1.45, "learning_rate": 2.135524895020996e-06, "loss": 0.0004, "step": 12059 }, { "epoch": 1.45, "learning_rate": 2.1354529094181165e-06, "loss": 0.0003, "step": 12060 }, { "epoch": 1.45, "learning_rate": 2.135380923815237e-06, "loss": 0.0003, "step": 12061 }, { "epoch": 1.45, "learning_rate": 2.1353089382123577e-06, "loss": 0.0002, "step": 12062 }, { "epoch": 1.45, "learning_rate": 2.135236952609478e-06, "loss": 0.0004, "step": 12063 }, { "epoch": 1.45, "learning_rate": 2.135164967006599e-06, "loss": 0.0002, "step": 12064 }, { "epoch": 1.45, "learning_rate": 2.135092981403719e-06, "loss": 0.0003, "step": 12065 }, { "epoch": 1.45, "learning_rate": 2.13502099580084e-06, "loss": 0.0002, "step": 12066 }, { "epoch": 1.45, "learning_rate": 2.1349490101979603e-06, "loss": 0.0004, "step": 12067 }, { "epoch": 1.45, "learning_rate": 2.134877024595081e-06, "loss": 0.0009, "step": 12068 }, { "epoch": 1.45, "learning_rate": 2.1348050389922014e-06, "loss": 0.0002, "step": 12069 }, { "epoch": 1.45, "learning_rate": 2.134733053389322e-06, "loss": 0.0002, "step": 12070 }, { "epoch": 1.45, "learning_rate": 2.134661067786443e-06, "loss": 0.0003, "step": 12071 }, { "epoch": 1.45, "learning_rate": 2.1345890821835633e-06, "loss": 0.0003, "step": 12072 }, { "epoch": 1.45, "learning_rate": 2.134517096580684e-06, "loss": 0.0003, "step": 12073 }, { "epoch": 1.45, "learning_rate": 2.1344451109778045e-06, "loss": 0.0005, "step": 12074 }, { "epoch": 1.45, "learning_rate": 2.1343731253749252e-06, "loss": 0.0003, "step": 12075 }, { "epoch": 1.45, "learning_rate": 2.1343011397720456e-06, "loss": 0.0003, "step": 12076 }, { "epoch": 1.45, "learning_rate": 2.1342291541691664e-06, "loss": 0.0003, "step": 12077 }, { "epoch": 1.45, "learning_rate": 2.1341571685662867e-06, "loss": 1.0034, "step": 12078 }, { "epoch": 1.45, "learning_rate": 2.1340851829634075e-06, "loss": 0.0004, "step": 12079 }, { "epoch": 1.45, "learning_rate": 2.134013197360528e-06, "loss": 0.0003, "step": 12080 }, { "epoch": 1.45, "learning_rate": 2.1339412117576486e-06, "loss": 0.0002, "step": 12081 }, { "epoch": 1.45, "learning_rate": 2.133869226154769e-06, "loss": 0.0004, "step": 12082 }, { "epoch": 1.45, "learning_rate": 2.1337972405518893e-06, "loss": 0.0004, "step": 12083 }, { "epoch": 1.45, "learning_rate": 2.13372525494901e-06, "loss": 0.0002, "step": 12084 }, { "epoch": 1.45, "learning_rate": 2.133653269346131e-06, "loss": 0.0004, "step": 12085 }, { "epoch": 1.45, "learning_rate": 2.1335812837432513e-06, "loss": 0.0003, "step": 12086 }, { "epoch": 1.45, "learning_rate": 2.133509298140372e-06, "loss": 0.0003, "step": 12087 }, { "epoch": 1.45, "learning_rate": 2.133437312537493e-06, "loss": 0.0003, "step": 12088 }, { "epoch": 1.45, "learning_rate": 2.133365326934613e-06, "loss": 0.0004, "step": 12089 }, { "epoch": 1.45, "learning_rate": 2.133293341331734e-06, "loss": 0.0003, "step": 12090 }, { "epoch": 1.45, "learning_rate": 2.1332213557288543e-06, "loss": 0.0004, "step": 12091 }, { "epoch": 1.45, "learning_rate": 2.133149370125975e-06, "loss": 0.0017, "step": 12092 }, { "epoch": 1.45, "learning_rate": 2.1330773845230954e-06, "loss": 0.0002, "step": 12093 }, { "epoch": 1.45, "learning_rate": 2.1330053989202162e-06, "loss": 0.0004, "step": 12094 }, { "epoch": 1.45, "learning_rate": 2.1329334133173366e-06, "loss": 0.0002, "step": 12095 }, { "epoch": 1.45, "learning_rate": 2.132861427714457e-06, "loss": 0.0002, "step": 12096 }, { "epoch": 1.45, "learning_rate": 2.1327894421115777e-06, "loss": 0.0005, "step": 12097 }, { "epoch": 1.45, "learning_rate": 2.132717456508698e-06, "loss": 0.0003, "step": 12098 }, { "epoch": 1.45, "learning_rate": 2.132645470905819e-06, "loss": 0.0002, "step": 12099 }, { "epoch": 1.45, "learning_rate": 2.132573485302939e-06, "loss": 0.0004, "step": 12100 }, { "epoch": 1.45, "learning_rate": 2.13250149970006e-06, "loss": 0.0004, "step": 12101 }, { "epoch": 1.45, "learning_rate": 2.1324295140971808e-06, "loss": 0.0002, "step": 12102 }, { "epoch": 1.45, "learning_rate": 2.1323575284943015e-06, "loss": 0.0004, "step": 12103 }, { "epoch": 1.45, "learning_rate": 2.132285542891422e-06, "loss": 0.0004, "step": 12104 }, { "epoch": 1.45, "learning_rate": 2.1322135572885427e-06, "loss": 0.0002, "step": 12105 }, { "epoch": 1.45, "learning_rate": 2.132141571685663e-06, "loss": 0.0004, "step": 12106 }, { "epoch": 1.45, "learning_rate": 2.1320695860827834e-06, "loss": 0.0005, "step": 12107 }, { "epoch": 1.45, "learning_rate": 2.131997600479904e-06, "loss": 0.0002, "step": 12108 }, { "epoch": 1.45, "learning_rate": 2.1319256148770245e-06, "loss": 0.0002, "step": 12109 }, { "epoch": 1.45, "learning_rate": 2.1318536292741453e-06, "loss": 0.0002, "step": 12110 }, { "epoch": 1.45, "learning_rate": 2.1317816436712656e-06, "loss": 0.0003, "step": 12111 }, { "epoch": 1.45, "learning_rate": 2.1317096580683864e-06, "loss": 0.0005, "step": 12112 }, { "epoch": 1.45, "learning_rate": 2.1316376724655068e-06, "loss": 0.0001, "step": 12113 }, { "epoch": 1.45, "learning_rate": 2.1315656868626276e-06, "loss": 0.0003, "step": 12114 }, { "epoch": 1.45, "learning_rate": 2.131493701259748e-06, "loss": 0.002, "step": 12115 }, { "epoch": 1.45, "learning_rate": 2.1314217156568687e-06, "loss": 0.0003, "step": 12116 }, { "epoch": 1.45, "learning_rate": 2.131349730053989e-06, "loss": 0.0005, "step": 12117 }, { "epoch": 1.45, "learning_rate": 2.13127774445111e-06, "loss": 0.0001, "step": 12118 }, { "epoch": 1.45, "learning_rate": 2.1312057588482306e-06, "loss": 0.0006, "step": 12119 }, { "epoch": 1.45, "learning_rate": 2.131133773245351e-06, "loss": 0.0004, "step": 12120 }, { "epoch": 1.45, "learning_rate": 2.1310617876424717e-06, "loss": 0.0005, "step": 12121 }, { "epoch": 1.45, "learning_rate": 2.130989802039592e-06, "loss": 0.3477, "step": 12122 }, { "epoch": 1.45, "learning_rate": 2.130917816436713e-06, "loss": 0.0061, "step": 12123 }, { "epoch": 1.45, "learning_rate": 2.1308458308338332e-06, "loss": 0.0002, "step": 12124 }, { "epoch": 1.45, "learning_rate": 2.130773845230954e-06, "loss": 0.0003, "step": 12125 }, { "epoch": 1.45, "learning_rate": 2.1307018596280744e-06, "loss": 0.0005, "step": 12126 }, { "epoch": 1.45, "learning_rate": 2.130629874025195e-06, "loss": 0.0004, "step": 12127 }, { "epoch": 1.45, "learning_rate": 2.1305578884223155e-06, "loss": 0.0002, "step": 12128 }, { "epoch": 1.45, "learning_rate": 2.130485902819436e-06, "loss": 0.0002, "step": 12129 }, { "epoch": 1.45, "learning_rate": 2.1304139172165566e-06, "loss": 0.0003, "step": 12130 }, { "epoch": 1.45, "learning_rate": 2.130341931613677e-06, "loss": 0.4662, "step": 12131 }, { "epoch": 1.45, "learning_rate": 2.1302699460107978e-06, "loss": 0.0002, "step": 12132 }, { "epoch": 1.45, "learning_rate": 2.1301979604079185e-06, "loss": 0.0002, "step": 12133 }, { "epoch": 1.45, "learning_rate": 2.1301259748050393e-06, "loss": 0.0003, "step": 12134 }, { "epoch": 1.45, "learning_rate": 2.1300539892021597e-06, "loss": 0.0004, "step": 12135 }, { "epoch": 1.45, "learning_rate": 2.1299820035992804e-06, "loss": 0.0003, "step": 12136 }, { "epoch": 1.45, "learning_rate": 2.129910017996401e-06, "loss": 0.0005, "step": 12137 }, { "epoch": 1.45, "learning_rate": 2.1298380323935216e-06, "loss": 0.0003, "step": 12138 }, { "epoch": 1.45, "learning_rate": 2.129766046790642e-06, "loss": 0.0004, "step": 12139 }, { "epoch": 1.45, "learning_rate": 2.1296940611877627e-06, "loss": 0.0002, "step": 12140 }, { "epoch": 1.45, "learning_rate": 2.129622075584883e-06, "loss": 0.0004, "step": 12141 }, { "epoch": 1.46, "learning_rate": 2.1295500899820034e-06, "loss": 0.0002, "step": 12142 }, { "epoch": 1.46, "learning_rate": 2.129478104379124e-06, "loss": 0.0007, "step": 12143 }, { "epoch": 1.46, "learning_rate": 2.1294061187762446e-06, "loss": 0.0002, "step": 12144 }, { "epoch": 1.46, "learning_rate": 2.1293341331733653e-06, "loss": 0.0003, "step": 12145 }, { "epoch": 1.46, "learning_rate": 2.1292621475704857e-06, "loss": 0.0003, "step": 12146 }, { "epoch": 1.46, "learning_rate": 2.1291901619676065e-06, "loss": 0.0003, "step": 12147 }, { "epoch": 1.46, "learning_rate": 2.1291181763647272e-06, "loss": 1.0484, "step": 12148 }, { "epoch": 1.46, "learning_rate": 2.1290461907618476e-06, "loss": 0.0002, "step": 12149 }, { "epoch": 1.46, "learning_rate": 2.1289742051589684e-06, "loss": 0.0001, "step": 12150 }, { "epoch": 1.46, "learning_rate": 2.128902219556089e-06, "loss": 0.0135, "step": 12151 }, { "epoch": 1.46, "learning_rate": 2.1288302339532095e-06, "loss": 0.0004, "step": 12152 }, { "epoch": 1.46, "learning_rate": 2.12875824835033e-06, "loss": 0.0001, "step": 12153 }, { "epoch": 1.46, "learning_rate": 2.1286862627474506e-06, "loss": 0.0002, "step": 12154 }, { "epoch": 1.46, "learning_rate": 2.128614277144571e-06, "loss": 0.0003, "step": 12155 }, { "epoch": 1.46, "learning_rate": 2.1285422915416918e-06, "loss": 0.0004, "step": 12156 }, { "epoch": 1.46, "learning_rate": 2.128470305938812e-06, "loss": 0.0003, "step": 12157 }, { "epoch": 1.46, "learning_rate": 2.128398320335933e-06, "loss": 0.0004, "step": 12158 }, { "epoch": 1.46, "learning_rate": 2.1283263347330533e-06, "loss": 0.0002, "step": 12159 }, { "epoch": 1.46, "learning_rate": 2.128254349130174e-06, "loss": 0.0004, "step": 12160 }, { "epoch": 1.46, "learning_rate": 2.1281823635272944e-06, "loss": 0.0009, "step": 12161 }, { "epoch": 1.46, "learning_rate": 2.128110377924415e-06, "loss": 0.0003, "step": 12162 }, { "epoch": 1.46, "learning_rate": 2.1280383923215355e-06, "loss": 0.0004, "step": 12163 }, { "epoch": 1.46, "learning_rate": 2.1279664067186563e-06, "loss": 0.0155, "step": 12164 }, { "epoch": 1.46, "learning_rate": 2.127894421115777e-06, "loss": 0.0003, "step": 12165 }, { "epoch": 1.46, "learning_rate": 2.1278224355128975e-06, "loss": 0.0004, "step": 12166 }, { "epoch": 1.46, "learning_rate": 2.1277504499100182e-06, "loss": 0.0005, "step": 12167 }, { "epoch": 1.46, "learning_rate": 2.1276784643071386e-06, "loss": 0.0003, "step": 12168 }, { "epoch": 1.46, "learning_rate": 2.1276064787042594e-06, "loss": 0.0005, "step": 12169 }, { "epoch": 1.46, "learning_rate": 2.1275344931013797e-06, "loss": 0.0004, "step": 12170 }, { "epoch": 1.46, "learning_rate": 2.1274625074985005e-06, "loss": 0.0003, "step": 12171 }, { "epoch": 1.46, "learning_rate": 2.127390521895621e-06, "loss": 0.0003, "step": 12172 }, { "epoch": 1.46, "learning_rate": 2.1273185362927416e-06, "loss": 0.0191, "step": 12173 }, { "epoch": 1.46, "learning_rate": 2.127246550689862e-06, "loss": 0.0004, "step": 12174 }, { "epoch": 1.46, "learning_rate": 2.1271745650869823e-06, "loss": 0.0004, "step": 12175 }, { "epoch": 1.46, "learning_rate": 2.127102579484103e-06, "loss": 0.0002, "step": 12176 }, { "epoch": 1.46, "learning_rate": 2.1270305938812235e-06, "loss": 0.0003, "step": 12177 }, { "epoch": 1.46, "learning_rate": 2.1269586082783443e-06, "loss": 0.0002, "step": 12178 }, { "epoch": 1.46, "learning_rate": 2.126886622675465e-06, "loss": 0.0002, "step": 12179 }, { "epoch": 1.46, "learning_rate": 2.126814637072586e-06, "loss": 0.0005, "step": 12180 }, { "epoch": 1.46, "learning_rate": 2.126742651469706e-06, "loss": 0.0002, "step": 12181 }, { "epoch": 1.46, "learning_rate": 2.126670665866827e-06, "loss": 0.0003, "step": 12182 }, { "epoch": 1.46, "learning_rate": 2.1265986802639473e-06, "loss": 0.0003, "step": 12183 }, { "epoch": 1.46, "learning_rate": 2.126526694661068e-06, "loss": 0.0005, "step": 12184 }, { "epoch": 1.46, "learning_rate": 2.1264547090581884e-06, "loss": 0.0003, "step": 12185 }, { "epoch": 1.46, "learning_rate": 2.126382723455309e-06, "loss": 0.0003, "step": 12186 }, { "epoch": 1.46, "learning_rate": 2.1263107378524296e-06, "loss": 0.0002, "step": 12187 }, { "epoch": 1.46, "learning_rate": 2.12623875224955e-06, "loss": 0.0005, "step": 12188 }, { "epoch": 1.46, "learning_rate": 2.1261667666466707e-06, "loss": 0.0005, "step": 12189 }, { "epoch": 1.46, "learning_rate": 2.126094781043791e-06, "loss": 0.0002, "step": 12190 }, { "epoch": 1.46, "learning_rate": 2.126022795440912e-06, "loss": 0.0005, "step": 12191 }, { "epoch": 1.46, "learning_rate": 2.125950809838032e-06, "loss": 0.0003, "step": 12192 }, { "epoch": 1.46, "learning_rate": 2.125878824235153e-06, "loss": 0.0002, "step": 12193 }, { "epoch": 1.46, "learning_rate": 2.1258068386322737e-06, "loss": 0.0002, "step": 12194 }, { "epoch": 1.46, "learning_rate": 2.125734853029394e-06, "loss": 0.0004, "step": 12195 }, { "epoch": 1.46, "learning_rate": 2.125662867426515e-06, "loss": 0.0007, "step": 12196 }, { "epoch": 1.46, "learning_rate": 2.1255908818236357e-06, "loss": 0.0002, "step": 12197 }, { "epoch": 1.46, "learning_rate": 2.125518896220756e-06, "loss": 0.0003, "step": 12198 }, { "epoch": 1.46, "learning_rate": 2.1254469106178764e-06, "loss": 0.0003, "step": 12199 }, { "epoch": 1.46, "learning_rate": 2.125374925014997e-06, "loss": 0.0003, "step": 12200 }, { "epoch": 1.46, "learning_rate": 2.1253029394121175e-06, "loss": 0.0002, "step": 12201 }, { "epoch": 1.46, "learning_rate": 2.1252309538092383e-06, "loss": 0.0003, "step": 12202 }, { "epoch": 1.46, "learning_rate": 2.1251589682063586e-06, "loss": 0.6353, "step": 12203 }, { "epoch": 1.46, "learning_rate": 2.1250869826034794e-06, "loss": 0.0008, "step": 12204 }, { "epoch": 1.46, "learning_rate": 2.1250149970005998e-06, "loss": 0.0001, "step": 12205 }, { "epoch": 1.46, "learning_rate": 2.1249430113977205e-06, "loss": 0.0003, "step": 12206 }, { "epoch": 1.46, "learning_rate": 2.124871025794841e-06, "loss": 0.0014, "step": 12207 }, { "epoch": 1.46, "learning_rate": 2.1247990401919617e-06, "loss": 0.0004, "step": 12208 }, { "epoch": 1.46, "learning_rate": 2.124727054589082e-06, "loss": 0.0006, "step": 12209 }, { "epoch": 1.46, "learning_rate": 2.124655068986203e-06, "loss": 0.9821, "step": 12210 }, { "epoch": 1.46, "learning_rate": 2.1245830833833236e-06, "loss": 0.0003, "step": 12211 }, { "epoch": 1.46, "learning_rate": 2.124511097780444e-06, "loss": 0.0005, "step": 12212 }, { "epoch": 1.46, "learning_rate": 2.1244391121775647e-06, "loss": 0.0001, "step": 12213 }, { "epoch": 1.46, "learning_rate": 2.124367126574685e-06, "loss": 0.0005, "step": 12214 }, { "epoch": 1.46, "learning_rate": 2.124295140971806e-06, "loss": 0.0003, "step": 12215 }, { "epoch": 1.46, "learning_rate": 2.1242231553689262e-06, "loss": 0.0003, "step": 12216 }, { "epoch": 1.46, "learning_rate": 2.124151169766047e-06, "loss": 0.0003, "step": 12217 }, { "epoch": 1.46, "learning_rate": 2.1240791841631673e-06, "loss": 0.0002, "step": 12218 }, { "epoch": 1.46, "learning_rate": 2.124007198560288e-06, "loss": 0.0005, "step": 12219 }, { "epoch": 1.46, "learning_rate": 2.1239352129574085e-06, "loss": 0.0003, "step": 12220 }, { "epoch": 1.46, "learning_rate": 2.1238632273545293e-06, "loss": 0.0003, "step": 12221 }, { "epoch": 1.46, "learning_rate": 2.1237912417516496e-06, "loss": 0.0003, "step": 12222 }, { "epoch": 1.46, "learning_rate": 2.12371925614877e-06, "loss": 0.0005, "step": 12223 }, { "epoch": 1.46, "learning_rate": 2.1236472705458907e-06, "loss": 0.0002, "step": 12224 }, { "epoch": 1.46, "learning_rate": 2.1235752849430115e-06, "loss": 0.0003, "step": 12225 }, { "epoch": 1.47, "learning_rate": 2.1235032993401323e-06, "loss": 0.0001, "step": 12226 }, { "epoch": 1.47, "learning_rate": 2.1234313137372527e-06, "loss": 0.0005, "step": 12227 }, { "epoch": 1.47, "learning_rate": 2.1233593281343734e-06, "loss": 0.0011, "step": 12228 }, { "epoch": 1.47, "learning_rate": 2.123287342531494e-06, "loss": 0.0002, "step": 12229 }, { "epoch": 1.47, "learning_rate": 2.1232153569286146e-06, "loss": 0.0003, "step": 12230 }, { "epoch": 1.47, "learning_rate": 2.123143371325735e-06, "loss": 0.0003, "step": 12231 }, { "epoch": 1.47, "learning_rate": 2.1230713857228557e-06, "loss": 0.0003, "step": 12232 }, { "epoch": 1.47, "learning_rate": 2.122999400119976e-06, "loss": 0.0004, "step": 12233 }, { "epoch": 1.47, "learning_rate": 2.1229274145170964e-06, "loss": 0.0005, "step": 12234 }, { "epoch": 1.47, "learning_rate": 2.122855428914217e-06, "loss": 0.0004, "step": 12235 }, { "epoch": 1.47, "learning_rate": 2.1227834433113376e-06, "loss": 0.0004, "step": 12236 }, { "epoch": 1.47, "learning_rate": 2.1227114577084583e-06, "loss": 0.0004, "step": 12237 }, { "epoch": 1.47, "learning_rate": 2.1226394721055787e-06, "loss": 0.0005, "step": 12238 }, { "epoch": 1.47, "learning_rate": 2.1225674865026995e-06, "loss": 0.0003, "step": 12239 }, { "epoch": 1.47, "learning_rate": 2.1224955008998202e-06, "loss": 0.0004, "step": 12240 }, { "epoch": 1.47, "learning_rate": 2.1224235152969406e-06, "loss": 0.0003, "step": 12241 }, { "epoch": 1.47, "learning_rate": 2.1223515296940614e-06, "loss": 0.0006, "step": 12242 }, { "epoch": 1.47, "learning_rate": 2.122279544091182e-06, "loss": 0.0002, "step": 12243 }, { "epoch": 1.47, "learning_rate": 2.1222075584883025e-06, "loss": 0.0003, "step": 12244 }, { "epoch": 1.47, "learning_rate": 2.122135572885423e-06, "loss": 0.0004, "step": 12245 }, { "epoch": 1.47, "learning_rate": 2.1220635872825436e-06, "loss": 0.0002, "step": 12246 }, { "epoch": 1.47, "learning_rate": 2.121991601679664e-06, "loss": 0.0002, "step": 12247 }, { "epoch": 1.47, "learning_rate": 2.1219196160767848e-06, "loss": 0.0004, "step": 12248 }, { "epoch": 1.47, "learning_rate": 2.121847630473905e-06, "loss": 0.0001, "step": 12249 }, { "epoch": 1.47, "learning_rate": 2.121775644871026e-06, "loss": 0.0005, "step": 12250 }, { "epoch": 1.47, "learning_rate": 2.1217036592681463e-06, "loss": 0.0003, "step": 12251 }, { "epoch": 1.47, "learning_rate": 2.121631673665267e-06, "loss": 0.0006, "step": 12252 }, { "epoch": 1.47, "learning_rate": 2.1215596880623874e-06, "loss": 0.0182, "step": 12253 }, { "epoch": 1.47, "learning_rate": 2.121487702459508e-06, "loss": 0.0005, "step": 12254 }, { "epoch": 1.47, "learning_rate": 2.1214157168566285e-06, "loss": 0.0004, "step": 12255 }, { "epoch": 1.47, "learning_rate": 2.1213437312537493e-06, "loss": 0.0002, "step": 12256 }, { "epoch": 1.47, "learning_rate": 2.12127174565087e-06, "loss": 0.0002, "step": 12257 }, { "epoch": 1.47, "learning_rate": 2.1211997600479904e-06, "loss": 0.0002, "step": 12258 }, { "epoch": 1.47, "learning_rate": 2.1211277744451112e-06, "loss": 0.0003, "step": 12259 }, { "epoch": 1.47, "learning_rate": 2.1210557888422316e-06, "loss": 0.0002, "step": 12260 }, { "epoch": 1.47, "learning_rate": 2.1209838032393524e-06, "loss": 0.0004, "step": 12261 }, { "epoch": 1.47, "learning_rate": 2.1209118176364727e-06, "loss": 0.0002, "step": 12262 }, { "epoch": 1.47, "learning_rate": 2.1208398320335935e-06, "loss": 0.8509, "step": 12263 }, { "epoch": 1.47, "learning_rate": 2.120767846430714e-06, "loss": 0.0003, "step": 12264 }, { "epoch": 1.47, "learning_rate": 2.1206958608278346e-06, "loss": 0.0003, "step": 12265 }, { "epoch": 1.47, "learning_rate": 2.120623875224955e-06, "loss": 0.0002, "step": 12266 }, { "epoch": 1.47, "learning_rate": 2.1205518896220758e-06, "loss": 0.0003, "step": 12267 }, { "epoch": 1.47, "learning_rate": 2.120479904019196e-06, "loss": 0.0004, "step": 12268 }, { "epoch": 1.47, "learning_rate": 2.1204079184163165e-06, "loss": 0.0004, "step": 12269 }, { "epoch": 1.47, "learning_rate": 2.1203359328134372e-06, "loss": 0.0002, "step": 12270 }, { "epoch": 1.47, "learning_rate": 2.120263947210558e-06, "loss": 0.0004, "step": 12271 }, { "epoch": 1.47, "learning_rate": 2.1201919616076784e-06, "loss": 0.0006, "step": 12272 }, { "epoch": 1.47, "learning_rate": 2.120119976004799e-06, "loss": 0.076, "step": 12273 }, { "epoch": 1.47, "learning_rate": 2.12004799040192e-06, "loss": 0.0003, "step": 12274 }, { "epoch": 1.47, "learning_rate": 2.1199760047990403e-06, "loss": 0.0003, "step": 12275 }, { "epoch": 1.47, "learning_rate": 2.119904019196161e-06, "loss": 0.0002, "step": 12276 }, { "epoch": 1.47, "learning_rate": 2.1198320335932814e-06, "loss": 0.0006, "step": 12277 }, { "epoch": 1.47, "learning_rate": 2.119760047990402e-06, "loss": 0.0002, "step": 12278 }, { "epoch": 1.47, "learning_rate": 2.1196880623875226e-06, "loss": 0.0005, "step": 12279 }, { "epoch": 1.47, "learning_rate": 2.119616076784643e-06, "loss": 0.0003, "step": 12280 }, { "epoch": 1.47, "learning_rate": 2.1195440911817637e-06, "loss": 0.0149, "step": 12281 }, { "epoch": 1.47, "learning_rate": 2.119472105578884e-06, "loss": 0.0004, "step": 12282 }, { "epoch": 1.47, "learning_rate": 2.119400119976005e-06, "loss": 0.0005, "step": 12283 }, { "epoch": 1.47, "learning_rate": 2.119328134373125e-06, "loss": 0.0003, "step": 12284 }, { "epoch": 1.47, "learning_rate": 2.119256148770246e-06, "loss": 0.0001, "step": 12285 }, { "epoch": 1.47, "learning_rate": 2.1191841631673663e-06, "loss": 0.0002, "step": 12286 }, { "epoch": 1.47, "learning_rate": 2.119112177564487e-06, "loss": 0.0001, "step": 12287 }, { "epoch": 1.47, "learning_rate": 2.119040191961608e-06, "loss": 0.0005, "step": 12288 }, { "epoch": 1.47, "learning_rate": 2.1189682063587286e-06, "loss": 0.8955, "step": 12289 }, { "epoch": 1.47, "learning_rate": 2.118896220755849e-06, "loss": 0.0003, "step": 12290 }, { "epoch": 1.47, "learning_rate": 2.1188242351529694e-06, "loss": 0.0003, "step": 12291 }, { "epoch": 1.47, "learning_rate": 2.11875224955009e-06, "loss": 0.0005, "step": 12292 }, { "epoch": 1.47, "learning_rate": 2.1186802639472105e-06, "loss": 0.0003, "step": 12293 }, { "epoch": 1.47, "learning_rate": 2.1186082783443313e-06, "loss": 0.0004, "step": 12294 }, { "epoch": 1.47, "learning_rate": 2.1185362927414516e-06, "loss": 0.0005, "step": 12295 }, { "epoch": 1.47, "learning_rate": 2.1184643071385724e-06, "loss": 0.0002, "step": 12296 }, { "epoch": 1.47, "learning_rate": 2.1183923215356928e-06, "loss": 0.0005, "step": 12297 }, { "epoch": 1.47, "learning_rate": 2.1183203359328135e-06, "loss": 0.0145, "step": 12298 }, { "epoch": 1.47, "learning_rate": 2.118248350329934e-06, "loss": 0.7481, "step": 12299 }, { "epoch": 1.47, "learning_rate": 2.1181763647270547e-06, "loss": 0.0006, "step": 12300 }, { "epoch": 1.47, "learning_rate": 2.118104379124175e-06, "loss": 0.0121, "step": 12301 }, { "epoch": 1.47, "learning_rate": 2.118032393521296e-06, "loss": 0.0007, "step": 12302 }, { "epoch": 1.47, "learning_rate": 2.1179604079184166e-06, "loss": 0.0009, "step": 12303 }, { "epoch": 1.47, "learning_rate": 2.117888422315537e-06, "loss": 0.0005, "step": 12304 }, { "epoch": 1.47, "learning_rate": 2.1178164367126577e-06, "loss": 0.0002, "step": 12305 }, { "epoch": 1.47, "learning_rate": 2.117744451109778e-06, "loss": 0.0006, "step": 12306 }, { "epoch": 1.47, "learning_rate": 2.117672465506899e-06, "loss": 0.0003, "step": 12307 }, { "epoch": 1.47, "learning_rate": 2.117600479904019e-06, "loss": 0.0002, "step": 12308 }, { "epoch": 1.48, "learning_rate": 2.11752849430114e-06, "loss": 0.0002, "step": 12309 }, { "epoch": 1.48, "learning_rate": 2.1174565086982603e-06, "loss": 0.0003, "step": 12310 }, { "epoch": 1.48, "learning_rate": 2.117384523095381e-06, "loss": 0.0003, "step": 12311 }, { "epoch": 1.48, "learning_rate": 2.1173125374925015e-06, "loss": 0.0009, "step": 12312 }, { "epoch": 1.48, "learning_rate": 2.1172405518896223e-06, "loss": 0.0002, "step": 12313 }, { "epoch": 1.48, "learning_rate": 2.1171685662867426e-06, "loss": 0.0003, "step": 12314 }, { "epoch": 1.48, "learning_rate": 2.117096580683863e-06, "loss": 0.0004, "step": 12315 }, { "epoch": 1.48, "learning_rate": 2.1170245950809837e-06, "loss": 0.0006, "step": 12316 }, { "epoch": 1.48, "learning_rate": 2.1169526094781045e-06, "loss": 0.0013, "step": 12317 }, { "epoch": 1.48, "learning_rate": 2.116880623875225e-06, "loss": 0.0002, "step": 12318 }, { "epoch": 1.48, "learning_rate": 2.1168086382723457e-06, "loss": 0.001, "step": 12319 }, { "epoch": 1.48, "learning_rate": 2.1167366526694664e-06, "loss": 0.0004, "step": 12320 }, { "epoch": 1.48, "learning_rate": 2.1166646670665868e-06, "loss": 0.0003, "step": 12321 }, { "epoch": 1.48, "learning_rate": 2.1165926814637076e-06, "loss": 0.2325, "step": 12322 }, { "epoch": 1.48, "learning_rate": 2.116520695860828e-06, "loss": 0.0004, "step": 12323 }, { "epoch": 1.48, "learning_rate": 2.1164487102579487e-06, "loss": 0.0002, "step": 12324 }, { "epoch": 1.48, "learning_rate": 2.116376724655069e-06, "loss": 0.0003, "step": 12325 }, { "epoch": 1.48, "learning_rate": 2.1163047390521894e-06, "loss": 0.0004, "step": 12326 }, { "epoch": 1.48, "learning_rate": 2.11623275344931e-06, "loss": 0.0004, "step": 12327 }, { "epoch": 1.48, "learning_rate": 2.1161607678464305e-06, "loss": 0.0005, "step": 12328 }, { "epoch": 1.48, "learning_rate": 2.1160887822435513e-06, "loss": 0.0006, "step": 12329 }, { "epoch": 1.48, "learning_rate": 2.1160167966406717e-06, "loss": 0.0003, "step": 12330 }, { "epoch": 1.48, "learning_rate": 2.1159448110377925e-06, "loss": 0.0005, "step": 12331 }, { "epoch": 1.48, "learning_rate": 2.115872825434913e-06, "loss": 0.0005, "step": 12332 }, { "epoch": 1.48, "learning_rate": 2.1158008398320336e-06, "loss": 0.0333, "step": 12333 }, { "epoch": 1.48, "learning_rate": 2.1157288542291544e-06, "loss": 0.001, "step": 12334 }, { "epoch": 1.48, "learning_rate": 2.115656868626275e-06, "loss": 0.0006, "step": 12335 }, { "epoch": 1.48, "learning_rate": 2.1155848830233955e-06, "loss": 0.0002, "step": 12336 }, { "epoch": 1.48, "learning_rate": 2.115512897420516e-06, "loss": 0.0003, "step": 12337 }, { "epoch": 1.48, "learning_rate": 2.1154409118176366e-06, "loss": 0.0002, "step": 12338 }, { "epoch": 1.48, "learning_rate": 2.115368926214757e-06, "loss": 0.0002, "step": 12339 }, { "epoch": 1.48, "learning_rate": 2.1152969406118778e-06, "loss": 0.0006, "step": 12340 }, { "epoch": 1.48, "learning_rate": 2.115224955008998e-06, "loss": 0.0003, "step": 12341 }, { "epoch": 1.48, "learning_rate": 2.115152969406119e-06, "loss": 0.0007, "step": 12342 }, { "epoch": 1.48, "learning_rate": 2.1150809838032393e-06, "loss": 0.0001, "step": 12343 }, { "epoch": 1.48, "learning_rate": 2.11500899820036e-06, "loss": 0.7447, "step": 12344 }, { "epoch": 1.48, "learning_rate": 2.1149370125974804e-06, "loss": 0.0002, "step": 12345 }, { "epoch": 1.48, "learning_rate": 2.114865026994601e-06, "loss": 0.0004, "step": 12346 }, { "epoch": 1.48, "learning_rate": 2.1147930413917215e-06, "loss": 0.0003, "step": 12347 }, { "epoch": 1.48, "learning_rate": 2.1147210557888423e-06, "loss": 0.0003, "step": 12348 }, { "epoch": 1.48, "learning_rate": 2.114649070185963e-06, "loss": 0.0005, "step": 12349 }, { "epoch": 1.48, "learning_rate": 2.1145770845830834e-06, "loss": 0.0003, "step": 12350 }, { "epoch": 1.48, "learning_rate": 2.1145050989802042e-06, "loss": 0.0003, "step": 12351 }, { "epoch": 1.48, "learning_rate": 2.1144331133773246e-06, "loss": 0.0136, "step": 12352 }, { "epoch": 1.48, "learning_rate": 2.1143611277744453e-06, "loss": 0.0004, "step": 12353 }, { "epoch": 1.48, "learning_rate": 2.1142891421715657e-06, "loss": 0.0006, "step": 12354 }, { "epoch": 1.48, "learning_rate": 2.1142171565686865e-06, "loss": 0.0003, "step": 12355 }, { "epoch": 1.48, "learning_rate": 2.114145170965807e-06, "loss": 0.0005, "step": 12356 }, { "epoch": 1.48, "learning_rate": 2.1140731853629276e-06, "loss": 0.0021, "step": 12357 }, { "epoch": 1.48, "learning_rate": 2.114001199760048e-06, "loss": 0.0005, "step": 12358 }, { "epoch": 1.48, "learning_rate": 2.1139292141571687e-06, "loss": 0.0005, "step": 12359 }, { "epoch": 1.48, "learning_rate": 2.113857228554289e-06, "loss": 0.003, "step": 12360 }, { "epoch": 1.48, "learning_rate": 2.1137852429514095e-06, "loss": 0.0002, "step": 12361 }, { "epoch": 1.48, "learning_rate": 2.1137132573485302e-06, "loss": 0.0005, "step": 12362 }, { "epoch": 1.48, "learning_rate": 2.113641271745651e-06, "loss": 0.0005, "step": 12363 }, { "epoch": 1.48, "learning_rate": 2.1135692861427714e-06, "loss": 0.0003, "step": 12364 }, { "epoch": 1.48, "learning_rate": 2.113497300539892e-06, "loss": 0.0005, "step": 12365 }, { "epoch": 1.48, "learning_rate": 2.113425314937013e-06, "loss": 0.0003, "step": 12366 }, { "epoch": 1.48, "learning_rate": 2.1133533293341333e-06, "loss": 0.0005, "step": 12367 }, { "epoch": 1.48, "learning_rate": 2.113281343731254e-06, "loss": 0.0003, "step": 12368 }, { "epoch": 1.48, "learning_rate": 2.1132093581283744e-06, "loss": 0.0008, "step": 12369 }, { "epoch": 1.48, "learning_rate": 2.113137372525495e-06, "loss": 0.0001, "step": 12370 }, { "epoch": 1.48, "learning_rate": 2.1130653869226155e-06, "loss": 0.0003, "step": 12371 }, { "epoch": 1.48, "learning_rate": 2.112993401319736e-06, "loss": 0.0002, "step": 12372 }, { "epoch": 1.48, "learning_rate": 2.1129214157168567e-06, "loss": 0.0003, "step": 12373 }, { "epoch": 1.48, "learning_rate": 2.112849430113977e-06, "loss": 0.0001, "step": 12374 }, { "epoch": 1.48, "learning_rate": 2.112777444511098e-06, "loss": 0.0004, "step": 12375 }, { "epoch": 1.48, "learning_rate": 2.112705458908218e-06, "loss": 0.0003, "step": 12376 }, { "epoch": 1.48, "learning_rate": 2.112633473305339e-06, "loss": 0.0003, "step": 12377 }, { "epoch": 1.48, "learning_rate": 2.1125614877024593e-06, "loss": 0.0003, "step": 12378 }, { "epoch": 1.48, "learning_rate": 2.11248950209958e-06, "loss": 0.196, "step": 12379 }, { "epoch": 1.48, "learning_rate": 2.112417516496701e-06, "loss": 0.0003, "step": 12380 }, { "epoch": 1.48, "learning_rate": 2.1123455308938216e-06, "loss": 0.8996, "step": 12381 }, { "epoch": 1.48, "learning_rate": 2.112273545290942e-06, "loss": 0.0004, "step": 12382 }, { "epoch": 1.48, "learning_rate": 2.1122015596880624e-06, "loss": 0.0003, "step": 12383 }, { "epoch": 1.48, "learning_rate": 2.112129574085183e-06, "loss": 0.0004, "step": 12384 }, { "epoch": 1.48, "learning_rate": 2.1120575884823035e-06, "loss": 0.0004, "step": 12385 }, { "epoch": 1.48, "learning_rate": 2.1119856028794243e-06, "loss": 0.0163, "step": 12386 }, { "epoch": 1.48, "learning_rate": 2.1119136172765446e-06, "loss": 0.0003, "step": 12387 }, { "epoch": 1.48, "learning_rate": 2.1118416316736654e-06, "loss": 0.9914, "step": 12388 }, { "epoch": 1.48, "learning_rate": 2.1117696460707858e-06, "loss": 0.0006, "step": 12389 }, { "epoch": 1.48, "learning_rate": 2.1116976604679065e-06, "loss": 0.0002, "step": 12390 }, { "epoch": 1.48, "learning_rate": 2.111625674865027e-06, "loss": 0.0003, "step": 12391 }, { "epoch": 1.48, "learning_rate": 2.1115536892621477e-06, "loss": 0.0002, "step": 12392 }, { "epoch": 1.49, "learning_rate": 2.111481703659268e-06, "loss": 0.6744, "step": 12393 }, { "epoch": 1.49, "learning_rate": 2.111409718056389e-06, "loss": 0.0002, "step": 12394 }, { "epoch": 1.49, "learning_rate": 2.111337732453509e-06, "loss": 0.0003, "step": 12395 }, { "epoch": 1.49, "learning_rate": 2.11126574685063e-06, "loss": 0.0004, "step": 12396 }, { "epoch": 1.49, "learning_rate": 2.1111937612477507e-06, "loss": 0.0005, "step": 12397 }, { "epoch": 1.49, "learning_rate": 2.111121775644871e-06, "loss": 0.0005, "step": 12398 }, { "epoch": 1.49, "learning_rate": 2.111049790041992e-06, "loss": 0.0005, "step": 12399 }, { "epoch": 1.49, "learning_rate": 2.110977804439112e-06, "loss": 0.001, "step": 12400 }, { "epoch": 1.49, "learning_rate": 2.110905818836233e-06, "loss": 0.0003, "step": 12401 }, { "epoch": 1.49, "learning_rate": 2.1108338332333533e-06, "loss": 0.0002, "step": 12402 }, { "epoch": 1.49, "learning_rate": 2.110761847630474e-06, "loss": 0.536, "step": 12403 }, { "epoch": 1.49, "learning_rate": 2.1106898620275945e-06, "loss": 0.0013, "step": 12404 }, { "epoch": 1.49, "learning_rate": 2.1106178764247152e-06, "loss": 0.0003, "step": 12405 }, { "epoch": 1.49, "learning_rate": 2.1105458908218356e-06, "loss": 0.001, "step": 12406 }, { "epoch": 1.49, "learning_rate": 2.110473905218956e-06, "loss": 0.0261, "step": 12407 }, { "epoch": 1.49, "learning_rate": 2.1104019196160767e-06, "loss": 0.0222, "step": 12408 }, { "epoch": 1.49, "learning_rate": 2.110329934013197e-06, "loss": 0.0003, "step": 12409 }, { "epoch": 1.49, "learning_rate": 2.110257948410318e-06, "loss": 0.0003, "step": 12410 }, { "epoch": 1.49, "learning_rate": 2.1101859628074386e-06, "loss": 0.0002, "step": 12411 }, { "epoch": 1.49, "learning_rate": 2.1101139772045594e-06, "loss": 0.0004, "step": 12412 }, { "epoch": 1.49, "learning_rate": 2.1100419916016798e-06, "loss": 0.0002, "step": 12413 }, { "epoch": 1.49, "learning_rate": 2.1099700059988006e-06, "loss": 0.0003, "step": 12414 }, { "epoch": 1.49, "learning_rate": 2.109898020395921e-06, "loss": 0.0006, "step": 12415 }, { "epoch": 1.49, "learning_rate": 2.1098260347930417e-06, "loss": 0.0002, "step": 12416 }, { "epoch": 1.49, "learning_rate": 2.109754049190162e-06, "loss": 0.0473, "step": 12417 }, { "epoch": 1.49, "learning_rate": 2.1096820635872824e-06, "loss": 0.0004, "step": 12418 }, { "epoch": 1.49, "learning_rate": 2.109610077984403e-06, "loss": 0.0003, "step": 12419 }, { "epoch": 1.49, "learning_rate": 2.1095380923815235e-06, "loss": 0.0003, "step": 12420 }, { "epoch": 1.49, "learning_rate": 2.1094661067786443e-06, "loss": 0.0006, "step": 12421 }, { "epoch": 1.49, "learning_rate": 2.1093941211757647e-06, "loss": 0.0003, "step": 12422 }, { "epoch": 1.49, "learning_rate": 2.1093221355728854e-06, "loss": 0.0059, "step": 12423 }, { "epoch": 1.49, "learning_rate": 2.109250149970006e-06, "loss": 0.0004, "step": 12424 }, { "epoch": 1.49, "learning_rate": 2.1091781643671266e-06, "loss": 0.0002, "step": 12425 }, { "epoch": 1.49, "learning_rate": 2.1091061787642474e-06, "loss": 0.0001, "step": 12426 }, { "epoch": 1.49, "learning_rate": 2.1090341931613677e-06, "loss": 0.006, "step": 12427 }, { "epoch": 1.49, "learning_rate": 2.1089622075584885e-06, "loss": 0.0006, "step": 12428 }, { "epoch": 1.49, "learning_rate": 2.108890221955609e-06, "loss": 0.0003, "step": 12429 }, { "epoch": 1.49, "learning_rate": 2.1088182363527296e-06, "loss": 0.0004, "step": 12430 }, { "epoch": 1.49, "learning_rate": 2.10874625074985e-06, "loss": 0.0003, "step": 12431 }, { "epoch": 1.49, "learning_rate": 2.1086742651469708e-06, "loss": 0.0004, "step": 12432 }, { "epoch": 1.49, "learning_rate": 2.108602279544091e-06, "loss": 0.0159, "step": 12433 }, { "epoch": 1.49, "learning_rate": 2.108530293941212e-06, "loss": 0.6182, "step": 12434 }, { "epoch": 1.49, "learning_rate": 2.1084583083383322e-06, "loss": 0.0596, "step": 12435 }, { "epoch": 1.49, "learning_rate": 2.108386322735453e-06, "loss": 0.0005, "step": 12436 }, { "epoch": 1.49, "learning_rate": 2.1083143371325734e-06, "loss": 0.0004, "step": 12437 }, { "epoch": 1.49, "learning_rate": 2.108242351529694e-06, "loss": 0.0011, "step": 12438 }, { "epoch": 1.49, "learning_rate": 2.1081703659268145e-06, "loss": 0.0002, "step": 12439 }, { "epoch": 1.49, "learning_rate": 2.1080983803239353e-06, "loss": 0.0003, "step": 12440 }, { "epoch": 1.49, "learning_rate": 2.1080263947210556e-06, "loss": 0.0005, "step": 12441 }, { "epoch": 1.49, "learning_rate": 2.1079544091181764e-06, "loss": 0.5485, "step": 12442 }, { "epoch": 1.49, "learning_rate": 2.107882423515297e-06, "loss": 0.0001, "step": 12443 }, { "epoch": 1.49, "learning_rate": 2.1078104379124176e-06, "loss": 0.2312, "step": 12444 }, { "epoch": 1.49, "learning_rate": 2.1077384523095383e-06, "loss": 0.8175, "step": 12445 }, { "epoch": 1.49, "learning_rate": 2.1076664667066587e-06, "loss": 0.0003, "step": 12446 }, { "epoch": 1.49, "learning_rate": 2.1075944811037795e-06, "loss": 0.0004, "step": 12447 }, { "epoch": 1.49, "learning_rate": 2.1075224955009e-06, "loss": 0.0009, "step": 12448 }, { "epoch": 1.49, "learning_rate": 2.1074505098980206e-06, "loss": 0.0002, "step": 12449 }, { "epoch": 1.49, "learning_rate": 2.107378524295141e-06, "loss": 0.0002, "step": 12450 }, { "epoch": 1.49, "learning_rate": 2.1073065386922617e-06, "loss": 0.0004, "step": 12451 }, { "epoch": 1.49, "learning_rate": 2.107234553089382e-06, "loss": 0.0002, "step": 12452 }, { "epoch": 1.49, "learning_rate": 2.1071625674865025e-06, "loss": 0.0006, "step": 12453 }, { "epoch": 1.49, "learning_rate": 2.1070905818836232e-06, "loss": 0.0002, "step": 12454 }, { "epoch": 1.49, "learning_rate": 2.1070185962807436e-06, "loss": 0.0003, "step": 12455 }, { "epoch": 1.49, "learning_rate": 2.1069466106778644e-06, "loss": 0.0003, "step": 12456 }, { "epoch": 1.49, "learning_rate": 2.106874625074985e-06, "loss": 0.0001, "step": 12457 }, { "epoch": 1.49, "learning_rate": 2.106802639472106e-06, "loss": 0.0004, "step": 12458 }, { "epoch": 1.49, "learning_rate": 2.1067306538692263e-06, "loss": 0.0005, "step": 12459 }, { "epoch": 1.49, "learning_rate": 2.106658668266347e-06, "loss": 0.0004, "step": 12460 }, { "epoch": 1.49, "learning_rate": 2.1065866826634674e-06, "loss": 0.0047, "step": 12461 }, { "epoch": 1.49, "learning_rate": 2.106514697060588e-06, "loss": 0.0053, "step": 12462 }, { "epoch": 1.49, "learning_rate": 2.1064427114577085e-06, "loss": 0.0003, "step": 12463 }, { "epoch": 1.49, "learning_rate": 2.106370725854829e-06, "loss": 0.0011, "step": 12464 }, { "epoch": 1.49, "learning_rate": 2.1062987402519497e-06, "loss": 0.0003, "step": 12465 }, { "epoch": 1.49, "learning_rate": 2.10622675464907e-06, "loss": 0.041, "step": 12466 }, { "epoch": 1.49, "learning_rate": 2.106154769046191e-06, "loss": 0.0008, "step": 12467 }, { "epoch": 1.49, "learning_rate": 2.106082783443311e-06, "loss": 0.0003, "step": 12468 }, { "epoch": 1.49, "learning_rate": 2.106010797840432e-06, "loss": 0.0003, "step": 12469 }, { "epoch": 1.49, "learning_rate": 2.1059388122375523e-06, "loss": 0.0003, "step": 12470 }, { "epoch": 1.49, "learning_rate": 2.105866826634673e-06, "loss": 0.0002, "step": 12471 }, { "epoch": 1.49, "learning_rate": 2.105794841031794e-06, "loss": 0.0009, "step": 12472 }, { "epoch": 1.49, "learning_rate": 2.105722855428914e-06, "loss": 0.0004, "step": 12473 }, { "epoch": 1.49, "learning_rate": 2.105650869826035e-06, "loss": 0.0002, "step": 12474 }, { "epoch": 1.49, "learning_rate": 2.1055788842231558e-06, "loss": 0.0002, "step": 12475 }, { "epoch": 1.5, "learning_rate": 2.105506898620276e-06, "loss": 0.0002, "step": 12476 }, { "epoch": 1.5, "learning_rate": 2.1054349130173965e-06, "loss": 0.0003, "step": 12477 }, { "epoch": 1.5, "learning_rate": 2.1053629274145173e-06, "loss": 0.0002, "step": 12478 }, { "epoch": 1.5, "learning_rate": 2.1052909418116376e-06, "loss": 0.0003, "step": 12479 }, { "epoch": 1.5, "learning_rate": 2.1052189562087584e-06, "loss": 0.0004, "step": 12480 }, { "epoch": 1.5, "learning_rate": 2.1051469706058787e-06, "loss": 0.0003, "step": 12481 }, { "epoch": 1.5, "learning_rate": 2.1050749850029995e-06, "loss": 0.0003, "step": 12482 }, { "epoch": 1.5, "learning_rate": 2.10500299940012e-06, "loss": 0.0002, "step": 12483 }, { "epoch": 1.5, "learning_rate": 2.1049310137972407e-06, "loss": 0.0002, "step": 12484 }, { "epoch": 1.5, "learning_rate": 2.104859028194361e-06, "loss": 0.0003, "step": 12485 }, { "epoch": 1.5, "learning_rate": 2.104787042591482e-06, "loss": 0.0002, "step": 12486 }, { "epoch": 1.5, "learning_rate": 2.104715056988602e-06, "loss": 0.0003, "step": 12487 }, { "epoch": 1.5, "learning_rate": 2.104643071385723e-06, "loss": 0.0001, "step": 12488 }, { "epoch": 1.5, "learning_rate": 2.1045710857828437e-06, "loss": 0.0003, "step": 12489 }, { "epoch": 1.5, "learning_rate": 2.104499100179964e-06, "loss": 0.7478, "step": 12490 }, { "epoch": 1.5, "learning_rate": 2.104427114577085e-06, "loss": 0.0004, "step": 12491 }, { "epoch": 1.5, "learning_rate": 2.104355128974205e-06, "loss": 0.0002, "step": 12492 }, { "epoch": 1.5, "learning_rate": 2.104283143371326e-06, "loss": 0.0005, "step": 12493 }, { "epoch": 1.5, "learning_rate": 2.1042111577684463e-06, "loss": 0.0002, "step": 12494 }, { "epoch": 1.5, "learning_rate": 2.104139172165567e-06, "loss": 0.0004, "step": 12495 }, { "epoch": 1.5, "learning_rate": 2.1040671865626875e-06, "loss": 0.0002, "step": 12496 }, { "epoch": 1.5, "learning_rate": 2.1039952009598082e-06, "loss": 0.4244, "step": 12497 }, { "epoch": 1.5, "learning_rate": 2.1039232153569286e-06, "loss": 0.6719, "step": 12498 }, { "epoch": 1.5, "learning_rate": 2.103851229754049e-06, "loss": 0.0002, "step": 12499 }, { "epoch": 1.5, "learning_rate": 2.1037792441511697e-06, "loss": 0.0003, "step": 12500 }, { "epoch": 1.5, "learning_rate": 2.10370725854829e-06, "loss": 0.0004, "step": 12501 }, { "epoch": 1.5, "learning_rate": 2.103635272945411e-06, "loss": 0.0004, "step": 12502 }, { "epoch": 1.5, "learning_rate": 2.1035632873425316e-06, "loss": 0.0004, "step": 12503 }, { "epoch": 1.5, "learning_rate": 2.1034913017396524e-06, "loss": 0.0004, "step": 12504 }, { "epoch": 1.5, "learning_rate": 2.1034193161367728e-06, "loss": 0.0003, "step": 12505 }, { "epoch": 1.5, "learning_rate": 2.1033473305338935e-06, "loss": 0.0101, "step": 12506 }, { "epoch": 1.5, "learning_rate": 2.103275344931014e-06, "loss": 0.0006, "step": 12507 }, { "epoch": 1.5, "learning_rate": 2.1032033593281347e-06, "loss": 0.0003, "step": 12508 }, { "epoch": 1.5, "learning_rate": 2.103131373725255e-06, "loss": 0.0013, "step": 12509 }, { "epoch": 1.5, "learning_rate": 2.1030593881223754e-06, "loss": 0.0004, "step": 12510 }, { "epoch": 1.5, "learning_rate": 2.102987402519496e-06, "loss": 0.0483, "step": 12511 }, { "epoch": 1.5, "learning_rate": 2.1029154169166165e-06, "loss": 0.0003, "step": 12512 }, { "epoch": 1.5, "learning_rate": 2.1028434313137373e-06, "loss": 0.0018, "step": 12513 }, { "epoch": 1.5, "learning_rate": 2.1027714457108577e-06, "loss": 0.0004, "step": 12514 }, { "epoch": 1.5, "learning_rate": 2.1026994601079784e-06, "loss": 0.0002, "step": 12515 }, { "epoch": 1.5, "learning_rate": 2.102627474505099e-06, "loss": 0.0013, "step": 12516 }, { "epoch": 1.5, "learning_rate": 2.1025554889022196e-06, "loss": 0.0003, "step": 12517 }, { "epoch": 1.5, "learning_rate": 2.10248350329934e-06, "loss": 0.0003, "step": 12518 }, { "epoch": 1.5, "learning_rate": 2.1024115176964607e-06, "loss": 0.0001, "step": 12519 }, { "epoch": 1.5, "learning_rate": 2.1023395320935815e-06, "loss": 0.0004, "step": 12520 }, { "epoch": 1.5, "learning_rate": 2.1022675464907023e-06, "loss": 0.0004, "step": 12521 }, { "epoch": 1.5, "learning_rate": 2.1021955608878226e-06, "loss": 0.0003, "step": 12522 }, { "epoch": 1.5, "learning_rate": 2.102123575284943e-06, "loss": 0.0005, "step": 12523 }, { "epoch": 1.5, "learning_rate": 2.1020515896820638e-06, "loss": 0.0002, "step": 12524 }, { "epoch": 1.5, "learning_rate": 2.101979604079184e-06, "loss": 0.0002, "step": 12525 }, { "epoch": 1.5, "learning_rate": 2.101907618476305e-06, "loss": 0.0003, "step": 12526 }, { "epoch": 1.5, "learning_rate": 2.1018356328734252e-06, "loss": 0.0002, "step": 12527 }, { "epoch": 1.5, "learning_rate": 2.101763647270546e-06, "loss": 0.0002, "step": 12528 }, { "epoch": 1.5, "learning_rate": 2.1016916616676664e-06, "loss": 0.0003, "step": 12529 }, { "epoch": 1.5, "learning_rate": 2.101619676064787e-06, "loss": 0.0004, "step": 12530 }, { "epoch": 1.5, "learning_rate": 2.1015476904619075e-06, "loss": 0.0002, "step": 12531 }, { "epoch": 1.5, "learning_rate": 2.1014757048590283e-06, "loss": 0.0002, "step": 12532 }, { "epoch": 1.5, "learning_rate": 2.1014037192561486e-06, "loss": 0.0004, "step": 12533 }, { "epoch": 1.5, "learning_rate": 2.1013317336532694e-06, "loss": 0.0003, "step": 12534 }, { "epoch": 1.5, "learning_rate": 2.10125974805039e-06, "loss": 0.0003, "step": 12535 }, { "epoch": 1.5, "learning_rate": 2.1011877624475106e-06, "loss": 0.6039, "step": 12536 }, { "epoch": 1.5, "learning_rate": 2.1011157768446313e-06, "loss": 0.0005, "step": 12537 }, { "epoch": 1.5, "learning_rate": 2.1010437912417517e-06, "loss": 0.2194, "step": 12538 }, { "epoch": 1.5, "learning_rate": 2.1009718056388725e-06, "loss": 0.0002, "step": 12539 }, { "epoch": 1.5, "learning_rate": 2.100899820035993e-06, "loss": 0.0004, "step": 12540 }, { "epoch": 1.5, "learning_rate": 2.1008278344331136e-06, "loss": 0.003, "step": 12541 }, { "epoch": 1.5, "learning_rate": 2.100755848830234e-06, "loss": 0.0004, "step": 12542 }, { "epoch": 1.5, "learning_rate": 2.1006838632273547e-06, "loss": 0.0006, "step": 12543 }, { "epoch": 1.5, "learning_rate": 2.100611877624475e-06, "loss": 0.0002, "step": 12544 }, { "epoch": 1.5, "learning_rate": 2.1005398920215954e-06, "loss": 0.0002, "step": 12545 }, { "epoch": 1.5, "learning_rate": 2.1004679064187162e-06, "loss": 0.0005, "step": 12546 }, { "epoch": 1.5, "learning_rate": 2.1003959208158366e-06, "loss": 0.0003, "step": 12547 }, { "epoch": 1.5, "learning_rate": 2.1003239352129574e-06, "loss": 0.0003, "step": 12548 }, { "epoch": 1.5, "learning_rate": 2.100251949610078e-06, "loss": 0.0003, "step": 12549 }, { "epoch": 1.5, "learning_rate": 2.1001799640071985e-06, "loss": 0.0002, "step": 12550 }, { "epoch": 1.5, "learning_rate": 2.1001079784043193e-06, "loss": 0.0002, "step": 12551 }, { "epoch": 1.5, "learning_rate": 2.10003599280144e-06, "loss": 0.0002, "step": 12552 }, { "epoch": 1.5, "learning_rate": 2.0999640071985604e-06, "loss": 0.0002, "step": 12553 }, { "epoch": 1.5, "learning_rate": 2.099892021595681e-06, "loss": 0.0002, "step": 12554 }, { "epoch": 1.5, "learning_rate": 2.0998200359928015e-06, "loss": 0.0026, "step": 12555 }, { "epoch": 1.5, "learning_rate": 2.099748050389922e-06, "loss": 0.0003, "step": 12556 }, { "epoch": 1.5, "learning_rate": 2.0996760647870427e-06, "loss": 0.0005, "step": 12557 }, { "epoch": 1.5, "learning_rate": 2.099604079184163e-06, "loss": 0.0003, "step": 12558 }, { "epoch": 1.5, "learning_rate": 2.099532093581284e-06, "loss": 0.0005, "step": 12559 }, { "epoch": 1.51, "learning_rate": 2.099460107978404e-06, "loss": 0.0002, "step": 12560 }, { "epoch": 1.51, "learning_rate": 2.099388122375525e-06, "loss": 0.0003, "step": 12561 }, { "epoch": 1.51, "learning_rate": 2.0993161367726453e-06, "loss": 0.0002, "step": 12562 }, { "epoch": 1.51, "learning_rate": 2.099244151169766e-06, "loss": 0.0006, "step": 12563 }, { "epoch": 1.51, "learning_rate": 2.0991721655668864e-06, "loss": 0.0002, "step": 12564 }, { "epoch": 1.51, "learning_rate": 2.099100179964007e-06, "loss": 0.0002, "step": 12565 }, { "epoch": 1.51, "learning_rate": 2.099028194361128e-06, "loss": 0.0002, "step": 12566 }, { "epoch": 1.51, "learning_rate": 2.0989562087582488e-06, "loss": 0.0001, "step": 12567 }, { "epoch": 1.51, "learning_rate": 2.098884223155369e-06, "loss": 0.0003, "step": 12568 }, { "epoch": 1.51, "learning_rate": 2.0988122375524895e-06, "loss": 0.0002, "step": 12569 }, { "epoch": 1.51, "learning_rate": 2.0987402519496102e-06, "loss": 0.0003, "step": 12570 }, { "epoch": 1.51, "learning_rate": 2.0986682663467306e-06, "loss": 0.3915, "step": 12571 }, { "epoch": 1.51, "learning_rate": 2.0985962807438514e-06, "loss": 0.0004, "step": 12572 }, { "epoch": 1.51, "learning_rate": 2.0985242951409717e-06, "loss": 0.0004, "step": 12573 }, { "epoch": 1.51, "learning_rate": 2.0984523095380925e-06, "loss": 0.0001, "step": 12574 }, { "epoch": 1.51, "learning_rate": 2.098380323935213e-06, "loss": 0.0002, "step": 12575 }, { "epoch": 1.51, "learning_rate": 2.0983083383323336e-06, "loss": 0.0003, "step": 12576 }, { "epoch": 1.51, "learning_rate": 2.098236352729454e-06, "loss": 0.0017, "step": 12577 }, { "epoch": 1.51, "learning_rate": 2.0981643671265748e-06, "loss": 0.0002, "step": 12578 }, { "epoch": 1.51, "learning_rate": 2.098092381523695e-06, "loss": 0.0003, "step": 12579 }, { "epoch": 1.51, "learning_rate": 2.098020395920816e-06, "loss": 0.001, "step": 12580 }, { "epoch": 1.51, "learning_rate": 2.0979484103179367e-06, "loss": 0.2643, "step": 12581 }, { "epoch": 1.51, "learning_rate": 2.097876424715057e-06, "loss": 0.0003, "step": 12582 }, { "epoch": 1.51, "learning_rate": 2.097804439112178e-06, "loss": 0.0002, "step": 12583 }, { "epoch": 1.51, "learning_rate": 2.097732453509298e-06, "loss": 0.0002, "step": 12584 }, { "epoch": 1.51, "learning_rate": 2.097660467906419e-06, "loss": 0.0021, "step": 12585 }, { "epoch": 1.51, "learning_rate": 2.0975884823035393e-06, "loss": 0.0003, "step": 12586 }, { "epoch": 1.51, "learning_rate": 2.09751649670066e-06, "loss": 0.0002, "step": 12587 }, { "epoch": 1.51, "learning_rate": 2.0974445110977804e-06, "loss": 0.0002, "step": 12588 }, { "epoch": 1.51, "learning_rate": 2.0973725254949012e-06, "loss": 0.0004, "step": 12589 }, { "epoch": 1.51, "learning_rate": 2.0973005398920216e-06, "loss": 0.0006, "step": 12590 }, { "epoch": 1.51, "learning_rate": 2.097228554289142e-06, "loss": 0.0003, "step": 12591 }, { "epoch": 1.51, "learning_rate": 2.0971565686862627e-06, "loss": 0.0001, "step": 12592 }, { "epoch": 1.51, "learning_rate": 2.097084583083383e-06, "loss": 0.0002, "step": 12593 }, { "epoch": 1.51, "learning_rate": 2.097012597480504e-06, "loss": 0.0002, "step": 12594 }, { "epoch": 1.51, "learning_rate": 2.0969406118776246e-06, "loss": 0.0005, "step": 12595 }, { "epoch": 1.51, "learning_rate": 2.096868626274745e-06, "loss": 0.0003, "step": 12596 }, { "epoch": 1.51, "learning_rate": 2.0967966406718658e-06, "loss": 0.0002, "step": 12597 }, { "epoch": 1.51, "learning_rate": 2.0967246550689865e-06, "loss": 0.0003, "step": 12598 }, { "epoch": 1.51, "learning_rate": 2.096652669466107e-06, "loss": 0.0001, "step": 12599 }, { "epoch": 1.51, "learning_rate": 2.0965806838632277e-06, "loss": 0.0002, "step": 12600 }, { "epoch": 1.51, "learning_rate": 2.096508698260348e-06, "loss": 0.0003, "step": 12601 }, { "epoch": 1.51, "learning_rate": 2.096436712657469e-06, "loss": 0.0002, "step": 12602 }, { "epoch": 1.51, "learning_rate": 2.096364727054589e-06, "loss": 0.0002, "step": 12603 }, { "epoch": 1.51, "learning_rate": 2.0962927414517095e-06, "loss": 0.0003, "step": 12604 }, { "epoch": 1.51, "learning_rate": 2.0962207558488303e-06, "loss": 0.0003, "step": 12605 }, { "epoch": 1.51, "learning_rate": 2.0961487702459507e-06, "loss": 0.0003, "step": 12606 }, { "epoch": 1.51, "learning_rate": 2.0960767846430714e-06, "loss": 0.0003, "step": 12607 }, { "epoch": 1.51, "learning_rate": 2.0960047990401918e-06, "loss": 0.0002, "step": 12608 }, { "epoch": 1.51, "learning_rate": 2.0959328134373126e-06, "loss": 0.0003, "step": 12609 }, { "epoch": 1.51, "learning_rate": 2.095860827834433e-06, "loss": 0.0003, "step": 12610 }, { "epoch": 1.51, "learning_rate": 2.0957888422315537e-06, "loss": 0.0003, "step": 12611 }, { "epoch": 1.51, "learning_rate": 2.0957168566286745e-06, "loss": 0.0008, "step": 12612 }, { "epoch": 1.51, "learning_rate": 2.0956448710257953e-06, "loss": 0.0001, "step": 12613 }, { "epoch": 1.51, "learning_rate": 2.0955728854229156e-06, "loss": 0.0003, "step": 12614 }, { "epoch": 1.51, "learning_rate": 2.095500899820036e-06, "loss": 0.0002, "step": 12615 }, { "epoch": 1.51, "learning_rate": 2.0954289142171567e-06, "loss": 0.0063, "step": 12616 }, { "epoch": 1.51, "learning_rate": 2.095356928614277e-06, "loss": 0.0003, "step": 12617 }, { "epoch": 1.51, "learning_rate": 2.095284943011398e-06, "loss": 0.0002, "step": 12618 }, { "epoch": 1.51, "learning_rate": 2.0952129574085182e-06, "loss": 0.0003, "step": 12619 }, { "epoch": 1.51, "learning_rate": 2.095140971805639e-06, "loss": 0.0005, "step": 12620 }, { "epoch": 1.51, "learning_rate": 2.0950689862027594e-06, "loss": 0.0003, "step": 12621 }, { "epoch": 1.51, "learning_rate": 2.09499700059988e-06, "loss": 0.0003, "step": 12622 }, { "epoch": 1.51, "learning_rate": 2.0949250149970005e-06, "loss": 0.0002, "step": 12623 }, { "epoch": 1.51, "learning_rate": 2.0948530293941213e-06, "loss": 0.0002, "step": 12624 }, { "epoch": 1.51, "learning_rate": 2.0947810437912416e-06, "loss": 0.0007, "step": 12625 }, { "epoch": 1.51, "learning_rate": 2.0947090581883624e-06, "loss": 0.0003, "step": 12626 }, { "epoch": 1.51, "learning_rate": 2.094637072585483e-06, "loss": 0.0001, "step": 12627 }, { "epoch": 1.51, "learning_rate": 2.0945650869826035e-06, "loss": 0.0032, "step": 12628 }, { "epoch": 1.51, "learning_rate": 2.0944931013797243e-06, "loss": 0.0002, "step": 12629 }, { "epoch": 1.51, "learning_rate": 2.0944211157768447e-06, "loss": 0.0004, "step": 12630 }, { "epoch": 1.51, "learning_rate": 2.0943491301739655e-06, "loss": 0.0083, "step": 12631 }, { "epoch": 1.51, "learning_rate": 2.094277144571086e-06, "loss": 0.0003, "step": 12632 }, { "epoch": 1.51, "learning_rate": 2.0942051589682066e-06, "loss": 1.0388, "step": 12633 }, { "epoch": 1.51, "learning_rate": 2.094133173365327e-06, "loss": 0.0001, "step": 12634 }, { "epoch": 1.51, "learning_rate": 2.0940611877624477e-06, "loss": 0.0002, "step": 12635 }, { "epoch": 1.51, "learning_rate": 2.093989202159568e-06, "loss": 0.0021, "step": 12636 }, { "epoch": 1.51, "learning_rate": 2.0939172165566884e-06, "loss": 0.0002, "step": 12637 }, { "epoch": 1.51, "learning_rate": 2.0938452309538092e-06, "loss": 0.0002, "step": 12638 }, { "epoch": 1.51, "learning_rate": 2.0937732453509296e-06, "loss": 0.0016, "step": 12639 }, { "epoch": 1.51, "learning_rate": 2.0937012597480503e-06, "loss": 0.0003, "step": 12640 }, { "epoch": 1.51, "learning_rate": 2.0936292741451707e-06, "loss": 0.0002, "step": 12641 }, { "epoch": 1.51, "learning_rate": 2.0935572885422915e-06, "loss": 0.0001, "step": 12642 }, { "epoch": 1.52, "learning_rate": 2.0934853029394123e-06, "loss": 0.8724, "step": 12643 }, { "epoch": 1.52, "learning_rate": 2.093413317336533e-06, "loss": 0.0004, "step": 12644 }, { "epoch": 1.52, "learning_rate": 2.0933413317336534e-06, "loss": 0.0002, "step": 12645 }, { "epoch": 1.52, "learning_rate": 2.093269346130774e-06, "loss": 0.0004, "step": 12646 }, { "epoch": 1.52, "learning_rate": 2.0931973605278945e-06, "loss": 0.0002, "step": 12647 }, { "epoch": 1.52, "learning_rate": 2.0931253749250153e-06, "loss": 0.0002, "step": 12648 }, { "epoch": 1.52, "learning_rate": 2.0930533893221357e-06, "loss": 0.0003, "step": 12649 }, { "epoch": 1.52, "learning_rate": 2.092981403719256e-06, "loss": 0.0002, "step": 12650 }, { "epoch": 1.52, "learning_rate": 2.092909418116377e-06, "loss": 0.0002, "step": 12651 }, { "epoch": 1.52, "learning_rate": 2.092837432513497e-06, "loss": 0.0002, "step": 12652 }, { "epoch": 1.52, "learning_rate": 2.092765446910618e-06, "loss": 0.0003, "step": 12653 }, { "epoch": 1.52, "learning_rate": 2.0926934613077383e-06, "loss": 0.0002, "step": 12654 }, { "epoch": 1.52, "learning_rate": 2.092621475704859e-06, "loss": 0.0003, "step": 12655 }, { "epoch": 1.52, "learning_rate": 2.0925494901019794e-06, "loss": 0.0003, "step": 12656 }, { "epoch": 1.52, "learning_rate": 2.0924775044991e-06, "loss": 0.0007, "step": 12657 }, { "epoch": 1.52, "learning_rate": 2.092405518896221e-06, "loss": 0.0002, "step": 12658 }, { "epoch": 1.52, "learning_rate": 2.0923335332933413e-06, "loss": 0.0015, "step": 12659 }, { "epoch": 1.52, "learning_rate": 2.092261547690462e-06, "loss": 0.0002, "step": 12660 }, { "epoch": 1.52, "learning_rate": 2.0921895620875825e-06, "loss": 0.0002, "step": 12661 }, { "epoch": 1.52, "learning_rate": 2.0921175764847032e-06, "loss": 0.0003, "step": 12662 }, { "epoch": 1.52, "learning_rate": 2.0920455908818236e-06, "loss": 0.0003, "step": 12663 }, { "epoch": 1.52, "learning_rate": 2.0919736052789444e-06, "loss": 0.0002, "step": 12664 }, { "epoch": 1.52, "learning_rate": 2.0919016196760647e-06, "loss": 0.0002, "step": 12665 }, { "epoch": 1.52, "learning_rate": 2.0918296340731855e-06, "loss": 0.0003, "step": 12666 }, { "epoch": 1.52, "learning_rate": 2.091757648470306e-06, "loss": 0.0004, "step": 12667 }, { "epoch": 1.52, "learning_rate": 2.0916856628674266e-06, "loss": 0.0002, "step": 12668 }, { "epoch": 1.52, "learning_rate": 2.091613677264547e-06, "loss": 0.0003, "step": 12669 }, { "epoch": 1.52, "learning_rate": 2.0915416916616678e-06, "loss": 0.0002, "step": 12670 }, { "epoch": 1.52, "learning_rate": 2.091469706058788e-06, "loss": 0.3618, "step": 12671 }, { "epoch": 1.52, "learning_rate": 2.091397720455909e-06, "loss": 0.0026, "step": 12672 }, { "epoch": 1.52, "learning_rate": 2.0913257348530293e-06, "loss": 0.0003, "step": 12673 }, { "epoch": 1.52, "learning_rate": 2.09125374925015e-06, "loss": 0.0003, "step": 12674 }, { "epoch": 1.52, "learning_rate": 2.091181763647271e-06, "loss": 0.0003, "step": 12675 }, { "epoch": 1.52, "learning_rate": 2.091109778044391e-06, "loss": 0.0008, "step": 12676 }, { "epoch": 1.52, "learning_rate": 2.091037792441512e-06, "loss": 0.0003, "step": 12677 }, { "epoch": 1.52, "learning_rate": 2.0909658068386323e-06, "loss": 0.0002, "step": 12678 }, { "epoch": 1.52, "learning_rate": 2.090893821235753e-06, "loss": 0.0002, "step": 12679 }, { "epoch": 1.52, "learning_rate": 2.0908218356328734e-06, "loss": 0.0002, "step": 12680 }, { "epoch": 1.52, "learning_rate": 2.0907498500299942e-06, "loss": 0.0007, "step": 12681 }, { "epoch": 1.52, "learning_rate": 2.0906778644271146e-06, "loss": 0.0001, "step": 12682 }, { "epoch": 1.52, "learning_rate": 2.090605878824235e-06, "loss": 0.0003, "step": 12683 }, { "epoch": 1.52, "learning_rate": 2.0905338932213557e-06, "loss": 0.0002, "step": 12684 }, { "epoch": 1.52, "learning_rate": 2.090461907618476e-06, "loss": 0.0001, "step": 12685 }, { "epoch": 1.52, "learning_rate": 2.090389922015597e-06, "loss": 0.0003, "step": 12686 }, { "epoch": 1.52, "learning_rate": 2.090317936412717e-06, "loss": 0.0002, "step": 12687 }, { "epoch": 1.52, "learning_rate": 2.090245950809838e-06, "loss": 0.0002, "step": 12688 }, { "epoch": 1.52, "learning_rate": 2.0901739652069588e-06, "loss": 0.0003, "step": 12689 }, { "epoch": 1.52, "learning_rate": 2.0901019796040795e-06, "loss": 0.0003, "step": 12690 }, { "epoch": 1.52, "learning_rate": 2.0900299940012e-06, "loss": 0.0003, "step": 12691 }, { "epoch": 1.52, "learning_rate": 2.0899580083983207e-06, "loss": 0.0003, "step": 12692 }, { "epoch": 1.52, "learning_rate": 2.089886022795441e-06, "loss": 0.0004, "step": 12693 }, { "epoch": 1.52, "learning_rate": 2.089814037192562e-06, "loss": 0.0002, "step": 12694 }, { "epoch": 1.52, "learning_rate": 2.089742051589682e-06, "loss": 0.0002, "step": 12695 }, { "epoch": 1.52, "learning_rate": 2.0896700659868025e-06, "loss": 0.0001, "step": 12696 }, { "epoch": 1.52, "learning_rate": 2.0895980803839233e-06, "loss": 0.0006, "step": 12697 }, { "epoch": 1.52, "learning_rate": 2.0895260947810436e-06, "loss": 0.0002, "step": 12698 }, { "epoch": 1.52, "learning_rate": 2.0894541091781644e-06, "loss": 0.0002, "step": 12699 }, { "epoch": 1.52, "learning_rate": 2.0893821235752848e-06, "loss": 0.0003, "step": 12700 }, { "epoch": 1.52, "learning_rate": 2.0893101379724056e-06, "loss": 0.0001, "step": 12701 }, { "epoch": 1.52, "learning_rate": 2.089238152369526e-06, "loss": 0.0005, "step": 12702 }, { "epoch": 1.52, "learning_rate": 2.0891661667666467e-06, "loss": 0.0002, "step": 12703 }, { "epoch": 1.52, "learning_rate": 2.0890941811637675e-06, "loss": 0.0002, "step": 12704 }, { "epoch": 1.52, "learning_rate": 2.089022195560888e-06, "loss": 0.0004, "step": 12705 }, { "epoch": 1.52, "learning_rate": 2.0889502099580086e-06, "loss": 0.0002, "step": 12706 }, { "epoch": 1.52, "learning_rate": 2.088878224355129e-06, "loss": 0.0003, "step": 12707 }, { "epoch": 1.52, "learning_rate": 2.0888062387522497e-06, "loss": 0.0102, "step": 12708 }, { "epoch": 1.52, "learning_rate": 2.08873425314937e-06, "loss": 0.0001, "step": 12709 }, { "epoch": 1.52, "learning_rate": 2.088662267546491e-06, "loss": 0.0003, "step": 12710 }, { "epoch": 1.52, "learning_rate": 2.0885902819436112e-06, "loss": 0.0004, "step": 12711 }, { "epoch": 1.52, "learning_rate": 2.088518296340732e-06, "loss": 0.0002, "step": 12712 }, { "epoch": 1.52, "learning_rate": 2.0884463107378524e-06, "loss": 0.0002, "step": 12713 }, { "epoch": 1.52, "learning_rate": 2.088374325134973e-06, "loss": 0.0002, "step": 12714 }, { "epoch": 1.52, "learning_rate": 2.0883023395320935e-06, "loss": 0.8782, "step": 12715 }, { "epoch": 1.52, "learning_rate": 2.0882303539292143e-06, "loss": 0.0001, "step": 12716 }, { "epoch": 1.52, "learning_rate": 2.0881583683263346e-06, "loss": 0.0005, "step": 12717 }, { "epoch": 1.52, "learning_rate": 2.0880863827234554e-06, "loss": 0.0002, "step": 12718 }, { "epoch": 1.52, "learning_rate": 2.0880143971205758e-06, "loss": 0.0003, "step": 12719 }, { "epoch": 1.52, "learning_rate": 2.0879424115176965e-06, "loss": 0.0003, "step": 12720 }, { "epoch": 1.52, "learning_rate": 2.0878704259148173e-06, "loss": 0.9028, "step": 12721 }, { "epoch": 1.52, "learning_rate": 2.0877984403119377e-06, "loss": 0.0005, "step": 12722 }, { "epoch": 1.52, "learning_rate": 2.0877264547090584e-06, "loss": 0.0003, "step": 12723 }, { "epoch": 1.52, "learning_rate": 2.087654469106179e-06, "loss": 0.0001, "step": 12724 }, { "epoch": 1.52, "learning_rate": 2.0875824835032996e-06, "loss": 0.7823, "step": 12725 }, { "epoch": 1.52, "learning_rate": 2.08751049790042e-06, "loss": 0.0004, "step": 12726 }, { "epoch": 1.53, "learning_rate": 2.0874385122975407e-06, "loss": 0.0004, "step": 12727 }, { "epoch": 1.53, "learning_rate": 2.087366526694661e-06, "loss": 0.0003, "step": 12728 }, { "epoch": 1.53, "learning_rate": 2.087294541091782e-06, "loss": 0.001, "step": 12729 }, { "epoch": 1.53, "learning_rate": 2.087222555488902e-06, "loss": 0.0006, "step": 12730 }, { "epoch": 1.53, "learning_rate": 2.0871505698860226e-06, "loss": 0.0005, "step": 12731 }, { "epoch": 1.53, "learning_rate": 2.0870785842831433e-06, "loss": 0.0005, "step": 12732 }, { "epoch": 1.53, "learning_rate": 2.0870065986802637e-06, "loss": 0.0002, "step": 12733 }, { "epoch": 1.53, "learning_rate": 2.0869346130773845e-06, "loss": 0.0003, "step": 12734 }, { "epoch": 1.53, "learning_rate": 2.0868626274745053e-06, "loss": 0.0002, "step": 12735 }, { "epoch": 1.53, "learning_rate": 2.086790641871626e-06, "loss": 0.0002, "step": 12736 }, { "epoch": 1.53, "learning_rate": 2.0867186562687464e-06, "loss": 0.0002, "step": 12737 }, { "epoch": 1.53, "learning_rate": 2.086646670665867e-06, "loss": 0.1565, "step": 12738 }, { "epoch": 1.53, "learning_rate": 2.0865746850629875e-06, "loss": 0.0003, "step": 12739 }, { "epoch": 1.53, "learning_rate": 2.0865026994601083e-06, "loss": 0.0003, "step": 12740 }, { "epoch": 1.53, "learning_rate": 2.0864307138572287e-06, "loss": 0.0008, "step": 12741 }, { "epoch": 1.53, "learning_rate": 2.086358728254349e-06, "loss": 0.0003, "step": 12742 }, { "epoch": 1.53, "learning_rate": 2.0862867426514698e-06, "loss": 0.0007, "step": 12743 }, { "epoch": 1.53, "learning_rate": 2.08621475704859e-06, "loss": 0.0002, "step": 12744 }, { "epoch": 1.53, "learning_rate": 2.086142771445711e-06, "loss": 0.0003, "step": 12745 }, { "epoch": 1.53, "learning_rate": 2.0860707858428313e-06, "loss": 0.0002, "step": 12746 }, { "epoch": 1.53, "learning_rate": 2.085998800239952e-06, "loss": 0.0002, "step": 12747 }, { "epoch": 1.53, "learning_rate": 2.0859268146370724e-06, "loss": 0.0002, "step": 12748 }, { "epoch": 1.53, "learning_rate": 2.085854829034193e-06, "loss": 0.0005, "step": 12749 }, { "epoch": 1.53, "learning_rate": 2.085782843431314e-06, "loss": 0.0004, "step": 12750 }, { "epoch": 1.53, "learning_rate": 2.0857108578284343e-06, "loss": 0.0002, "step": 12751 }, { "epoch": 1.53, "learning_rate": 2.085638872225555e-06, "loss": 0.0002, "step": 12752 }, { "epoch": 1.53, "learning_rate": 2.0855668866226755e-06, "loss": 0.0004, "step": 12753 }, { "epoch": 1.53, "learning_rate": 2.0854949010197962e-06, "loss": 0.0005, "step": 12754 }, { "epoch": 1.53, "learning_rate": 2.0854229154169166e-06, "loss": 0.0004, "step": 12755 }, { "epoch": 1.53, "learning_rate": 2.0853509298140374e-06, "loss": 0.0002, "step": 12756 }, { "epoch": 1.53, "learning_rate": 2.0852789442111577e-06, "loss": 0.0003, "step": 12757 }, { "epoch": 1.53, "learning_rate": 2.0852069586082785e-06, "loss": 0.0004, "step": 12758 }, { "epoch": 1.53, "learning_rate": 2.085134973005399e-06, "loss": 0.0002, "step": 12759 }, { "epoch": 1.53, "learning_rate": 2.0850629874025196e-06, "loss": 0.0002, "step": 12760 }, { "epoch": 1.53, "learning_rate": 2.08499100179964e-06, "loss": 0.0006, "step": 12761 }, { "epoch": 1.53, "learning_rate": 2.0849190161967608e-06, "loss": 0.0022, "step": 12762 }, { "epoch": 1.53, "learning_rate": 2.084847030593881e-06, "loss": 0.0001, "step": 12763 }, { "epoch": 1.53, "learning_rate": 2.084775044991002e-06, "loss": 0.0002, "step": 12764 }, { "epoch": 1.53, "learning_rate": 2.0847030593881223e-06, "loss": 0.0003, "step": 12765 }, { "epoch": 1.53, "learning_rate": 2.084631073785243e-06, "loss": 0.0002, "step": 12766 }, { "epoch": 1.53, "learning_rate": 2.084559088182364e-06, "loss": 0.0007, "step": 12767 }, { "epoch": 1.53, "learning_rate": 2.084487102579484e-06, "loss": 0.0003, "step": 12768 }, { "epoch": 1.53, "learning_rate": 2.084415116976605e-06, "loss": 0.0002, "step": 12769 }, { "epoch": 1.53, "learning_rate": 2.0843431313737253e-06, "loss": 0.01, "step": 12770 }, { "epoch": 1.53, "learning_rate": 2.084271145770846e-06, "loss": 0.0189, "step": 12771 }, { "epoch": 1.53, "learning_rate": 2.0841991601679664e-06, "loss": 0.0003, "step": 12772 }, { "epoch": 1.53, "learning_rate": 2.0841271745650872e-06, "loss": 0.0004, "step": 12773 }, { "epoch": 1.53, "learning_rate": 2.0840551889622076e-06, "loss": 1.0572, "step": 12774 }, { "epoch": 1.53, "learning_rate": 2.0839832033593283e-06, "loss": 0.0006, "step": 12775 }, { "epoch": 1.53, "learning_rate": 2.0839112177564487e-06, "loss": 0.0002, "step": 12776 }, { "epoch": 1.53, "learning_rate": 2.083839232153569e-06, "loss": 0.0002, "step": 12777 }, { "epoch": 1.53, "learning_rate": 2.08376724655069e-06, "loss": 0.7857, "step": 12778 }, { "epoch": 1.53, "learning_rate": 2.08369526094781e-06, "loss": 0.0002, "step": 12779 }, { "epoch": 1.53, "learning_rate": 2.083623275344931e-06, "loss": 1.1877, "step": 12780 }, { "epoch": 1.53, "learning_rate": 2.0835512897420517e-06, "loss": 0.0002, "step": 12781 }, { "epoch": 1.53, "learning_rate": 2.0834793041391725e-06, "loss": 0.0004, "step": 12782 }, { "epoch": 1.53, "learning_rate": 2.083407318536293e-06, "loss": 0.0002, "step": 12783 }, { "epoch": 1.53, "learning_rate": 2.0833353329334137e-06, "loss": 0.0004, "step": 12784 }, { "epoch": 1.53, "learning_rate": 2.083263347330534e-06, "loss": 0.0007, "step": 12785 }, { "epoch": 1.53, "learning_rate": 2.083191361727655e-06, "loss": 0.0001, "step": 12786 }, { "epoch": 1.53, "learning_rate": 2.083119376124775e-06, "loss": 0.0003, "step": 12787 }, { "epoch": 1.53, "learning_rate": 2.0830473905218955e-06, "loss": 0.0002, "step": 12788 }, { "epoch": 1.53, "learning_rate": 2.0829754049190163e-06, "loss": 0.0002, "step": 12789 }, { "epoch": 1.53, "learning_rate": 2.0829034193161366e-06, "loss": 0.0002, "step": 12790 }, { "epoch": 1.53, "learning_rate": 2.0828314337132574e-06, "loss": 0.0003, "step": 12791 }, { "epoch": 1.53, "learning_rate": 2.0827594481103778e-06, "loss": 0.0002, "step": 12792 }, { "epoch": 1.53, "learning_rate": 2.0826874625074985e-06, "loss": 0.0008, "step": 12793 }, { "epoch": 1.53, "learning_rate": 2.082615476904619e-06, "loss": 0.0002, "step": 12794 }, { "epoch": 1.53, "learning_rate": 2.0825434913017397e-06, "loss": 0.0003, "step": 12795 }, { "epoch": 1.53, "learning_rate": 2.08247150569886e-06, "loss": 0.0009, "step": 12796 }, { "epoch": 1.53, "learning_rate": 2.082399520095981e-06, "loss": 0.0002, "step": 12797 }, { "epoch": 1.53, "learning_rate": 2.0823275344931016e-06, "loss": 0.0005, "step": 12798 }, { "epoch": 1.53, "learning_rate": 2.082255548890222e-06, "loss": 0.0003, "step": 12799 }, { "epoch": 1.53, "learning_rate": 2.0821835632873427e-06, "loss": 0.0004, "step": 12800 }, { "epoch": 1.53, "learning_rate": 2.082111577684463e-06, "loss": 0.0003, "step": 12801 }, { "epoch": 1.53, "learning_rate": 2.082039592081584e-06, "loss": 0.0002, "step": 12802 }, { "epoch": 1.53, "learning_rate": 2.0819676064787042e-06, "loss": 0.0009, "step": 12803 }, { "epoch": 1.53, "learning_rate": 2.081895620875825e-06, "loss": 0.0004, "step": 12804 }, { "epoch": 1.53, "learning_rate": 2.0818236352729453e-06, "loss": 0.0003, "step": 12805 }, { "epoch": 1.53, "learning_rate": 2.081751649670066e-06, "loss": 0.0003, "step": 12806 }, { "epoch": 1.53, "learning_rate": 2.0816796640671865e-06, "loss": 0.1225, "step": 12807 }, { "epoch": 1.53, "learning_rate": 2.0816076784643073e-06, "loss": 0.0003, "step": 12808 }, { "epoch": 1.53, "learning_rate": 2.0815356928614276e-06, "loss": 0.0003, "step": 12809 }, { "epoch": 1.54, "learning_rate": 2.0814637072585484e-06, "loss": 0.0004, "step": 12810 }, { "epoch": 1.54, "learning_rate": 2.0813917216556688e-06, "loss": 0.001, "step": 12811 }, { "epoch": 1.54, "learning_rate": 2.0813197360527895e-06, "loss": 0.0001, "step": 12812 }, { "epoch": 1.54, "learning_rate": 2.0812477504499103e-06, "loss": 0.0002, "step": 12813 }, { "epoch": 1.54, "learning_rate": 2.0811757648470307e-06, "loss": 0.0002, "step": 12814 }, { "epoch": 1.54, "learning_rate": 2.0811037792441514e-06, "loss": 0.0001, "step": 12815 }, { "epoch": 1.54, "learning_rate": 2.081031793641272e-06, "loss": 0.0159, "step": 12816 }, { "epoch": 1.54, "learning_rate": 2.0809598080383926e-06, "loss": 0.0002, "step": 12817 }, { "epoch": 1.54, "learning_rate": 2.080887822435513e-06, "loss": 0.0002, "step": 12818 }, { "epoch": 1.54, "learning_rate": 2.0808158368326337e-06, "loss": 0.0043, "step": 12819 }, { "epoch": 1.54, "learning_rate": 2.080743851229754e-06, "loss": 0.0002, "step": 12820 }, { "epoch": 1.54, "learning_rate": 2.080671865626875e-06, "loss": 0.0002, "step": 12821 }, { "epoch": 1.54, "learning_rate": 2.080599880023995e-06, "loss": 0.0004, "step": 12822 }, { "epoch": 1.54, "learning_rate": 2.0805278944211156e-06, "loss": 0.0002, "step": 12823 }, { "epoch": 1.54, "learning_rate": 2.0804559088182363e-06, "loss": 0.0003, "step": 12824 }, { "epoch": 1.54, "learning_rate": 2.0803839232153567e-06, "loss": 0.0003, "step": 12825 }, { "epoch": 1.54, "learning_rate": 2.0803119376124775e-06, "loss": 0.0002, "step": 12826 }, { "epoch": 1.54, "learning_rate": 2.0802399520095982e-06, "loss": 0.0005, "step": 12827 }, { "epoch": 1.54, "learning_rate": 2.0801679664067186e-06, "loss": 0.0002, "step": 12828 }, { "epoch": 1.54, "learning_rate": 2.0800959808038394e-06, "loss": 0.0003, "step": 12829 }, { "epoch": 1.54, "learning_rate": 2.08002399520096e-06, "loss": 0.0007, "step": 12830 }, { "epoch": 1.54, "learning_rate": 2.0799520095980805e-06, "loss": 0.0003, "step": 12831 }, { "epoch": 1.54, "learning_rate": 2.0798800239952013e-06, "loss": 0.0002, "step": 12832 }, { "epoch": 1.54, "learning_rate": 2.0798080383923216e-06, "loss": 0.0002, "step": 12833 }, { "epoch": 1.54, "learning_rate": 2.079736052789442e-06, "loss": 0.0002, "step": 12834 }, { "epoch": 1.54, "learning_rate": 2.0796640671865628e-06, "loss": 0.0002, "step": 12835 }, { "epoch": 1.54, "learning_rate": 2.079592081583683e-06, "loss": 0.0002, "step": 12836 }, { "epoch": 1.54, "learning_rate": 2.079520095980804e-06, "loss": 0.1905, "step": 12837 }, { "epoch": 1.54, "learning_rate": 2.0794481103779243e-06, "loss": 0.1256, "step": 12838 }, { "epoch": 1.54, "learning_rate": 2.079376124775045e-06, "loss": 0.0003, "step": 12839 }, { "epoch": 1.54, "learning_rate": 2.0793041391721654e-06, "loss": 0.0002, "step": 12840 }, { "epoch": 1.54, "learning_rate": 2.079232153569286e-06, "loss": 0.0001, "step": 12841 }, { "epoch": 1.54, "learning_rate": 2.0791601679664065e-06, "loss": 0.0002, "step": 12842 }, { "epoch": 1.54, "learning_rate": 2.0790881823635273e-06, "loss": 0.0002, "step": 12843 }, { "epoch": 1.54, "learning_rate": 2.079016196760648e-06, "loss": 0.0003, "step": 12844 }, { "epoch": 1.54, "learning_rate": 2.0789442111577684e-06, "loss": 0.4997, "step": 12845 }, { "epoch": 1.54, "learning_rate": 2.0788722255548892e-06, "loss": 0.0003, "step": 12846 }, { "epoch": 1.54, "learning_rate": 2.0788002399520096e-06, "loss": 0.0002, "step": 12847 }, { "epoch": 1.54, "learning_rate": 2.0787282543491304e-06, "loss": 0.6863, "step": 12848 }, { "epoch": 1.54, "learning_rate": 2.0786562687462507e-06, "loss": 0.0002, "step": 12849 }, { "epoch": 1.54, "learning_rate": 2.0785842831433715e-06, "loss": 0.0006, "step": 12850 }, { "epoch": 1.54, "learning_rate": 2.078512297540492e-06, "loss": 0.0003, "step": 12851 }, { "epoch": 1.54, "learning_rate": 2.0784403119376126e-06, "loss": 0.0002, "step": 12852 }, { "epoch": 1.54, "learning_rate": 2.078368326334733e-06, "loss": 0.0002, "step": 12853 }, { "epoch": 1.54, "learning_rate": 2.0782963407318538e-06, "loss": 0.0004, "step": 12854 }, { "epoch": 1.54, "learning_rate": 2.078224355128974e-06, "loss": 0.6373, "step": 12855 }, { "epoch": 1.54, "learning_rate": 2.078152369526095e-06, "loss": 0.002, "step": 12856 }, { "epoch": 1.54, "learning_rate": 2.0780803839232152e-06, "loss": 0.0002, "step": 12857 }, { "epoch": 1.54, "learning_rate": 2.078008398320336e-06, "loss": 0.0003, "step": 12858 }, { "epoch": 1.54, "learning_rate": 2.077936412717457e-06, "loss": 0.0004, "step": 12859 }, { "epoch": 1.54, "learning_rate": 2.077864427114577e-06, "loss": 0.0001, "step": 12860 }, { "epoch": 1.54, "learning_rate": 2.077792441511698e-06, "loss": 0.0002, "step": 12861 }, { "epoch": 1.54, "learning_rate": 2.0777204559088183e-06, "loss": 0.0006, "step": 12862 }, { "epoch": 1.54, "learning_rate": 2.077648470305939e-06, "loss": 0.0003, "step": 12863 }, { "epoch": 1.54, "learning_rate": 2.0775764847030594e-06, "loss": 0.0003, "step": 12864 }, { "epoch": 1.54, "learning_rate": 2.07750449910018e-06, "loss": 0.0005, "step": 12865 }, { "epoch": 1.54, "learning_rate": 2.0774325134973006e-06, "loss": 0.0004, "step": 12866 }, { "epoch": 1.54, "learning_rate": 2.0773605278944213e-06, "loss": 0.0003, "step": 12867 }, { "epoch": 1.54, "learning_rate": 2.0772885422915417e-06, "loss": 0.0003, "step": 12868 }, { "epoch": 1.54, "learning_rate": 2.077216556688662e-06, "loss": 0.0002, "step": 12869 }, { "epoch": 1.54, "learning_rate": 2.077144571085783e-06, "loss": 0.023, "step": 12870 }, { "epoch": 1.54, "learning_rate": 2.077072585482903e-06, "loss": 0.0003, "step": 12871 }, { "epoch": 1.54, "learning_rate": 2.077000599880024e-06, "loss": 0.0003, "step": 12872 }, { "epoch": 1.54, "learning_rate": 2.0769286142771447e-06, "loss": 0.0003, "step": 12873 }, { "epoch": 1.54, "learning_rate": 2.076856628674265e-06, "loss": 0.0006, "step": 12874 }, { "epoch": 1.54, "learning_rate": 2.076784643071386e-06, "loss": 0.0004, "step": 12875 }, { "epoch": 1.54, "learning_rate": 2.0767126574685067e-06, "loss": 0.0003, "step": 12876 }, { "epoch": 1.54, "learning_rate": 2.076640671865627e-06, "loss": 0.0003, "step": 12877 }, { "epoch": 1.54, "learning_rate": 2.0765686862627478e-06, "loss": 0.0002, "step": 12878 }, { "epoch": 1.54, "learning_rate": 2.076496700659868e-06, "loss": 0.0003, "step": 12879 }, { "epoch": 1.54, "learning_rate": 2.0764247150569885e-06, "loss": 0.0006, "step": 12880 }, { "epoch": 1.54, "learning_rate": 2.0763527294541093e-06, "loss": 0.2954, "step": 12881 }, { "epoch": 1.54, "learning_rate": 2.0762807438512296e-06, "loss": 0.0003, "step": 12882 }, { "epoch": 1.54, "learning_rate": 2.0762087582483504e-06, "loss": 0.0003, "step": 12883 }, { "epoch": 1.54, "learning_rate": 2.0761367726454708e-06, "loss": 0.0005, "step": 12884 }, { "epoch": 1.54, "learning_rate": 2.0760647870425915e-06, "loss": 0.0004, "step": 12885 }, { "epoch": 1.54, "learning_rate": 2.075992801439712e-06, "loss": 0.0029, "step": 12886 }, { "epoch": 1.54, "learning_rate": 2.0759208158368327e-06, "loss": 0.0002, "step": 12887 }, { "epoch": 1.54, "learning_rate": 2.075848830233953e-06, "loss": 0.0004, "step": 12888 }, { "epoch": 1.54, "learning_rate": 2.075776844631074e-06, "loss": 0.0004, "step": 12889 }, { "epoch": 1.54, "learning_rate": 2.0757048590281946e-06, "loss": 0.0002, "step": 12890 }, { "epoch": 1.54, "learning_rate": 2.075632873425315e-06, "loss": 0.0001, "step": 12891 }, { "epoch": 1.54, "learning_rate": 2.0755608878224357e-06, "loss": 0.0001, "step": 12892 }, { "epoch": 1.54, "learning_rate": 2.075488902219556e-06, "loss": 0.0011, "step": 12893 }, { "epoch": 1.55, "learning_rate": 2.075416916616677e-06, "loss": 0.0003, "step": 12894 }, { "epoch": 1.55, "learning_rate": 2.075344931013797e-06, "loss": 0.0005, "step": 12895 }, { "epoch": 1.55, "learning_rate": 2.075272945410918e-06, "loss": 0.0003, "step": 12896 }, { "epoch": 1.55, "learning_rate": 2.0752009598080383e-06, "loss": 0.0003, "step": 12897 }, { "epoch": 1.55, "learning_rate": 2.075128974205159e-06, "loss": 0.0003, "step": 12898 }, { "epoch": 1.55, "learning_rate": 2.0750569886022795e-06, "loss": 0.0002, "step": 12899 }, { "epoch": 1.55, "learning_rate": 2.0749850029994003e-06, "loss": 0.0002, "step": 12900 }, { "epoch": 1.55, "learning_rate": 2.0749130173965206e-06, "loss": 0.0015, "step": 12901 }, { "epoch": 1.55, "learning_rate": 2.0748410317936414e-06, "loss": 0.0003, "step": 12902 }, { "epoch": 1.55, "learning_rate": 2.0747690461907617e-06, "loss": 0.0004, "step": 12903 }, { "epoch": 1.55, "learning_rate": 2.0746970605878825e-06, "loss": 0.0005, "step": 12904 }, { "epoch": 1.55, "learning_rate": 2.0746250749850033e-06, "loss": 0.0002, "step": 12905 }, { "epoch": 1.55, "learning_rate": 2.0745530893821237e-06, "loss": 0.0004, "step": 12906 }, { "epoch": 1.55, "learning_rate": 2.0744811037792444e-06, "loss": 0.0002, "step": 12907 }, { "epoch": 1.55, "learning_rate": 2.0744091181763648e-06, "loss": 0.0003, "step": 12908 }, { "epoch": 1.55, "learning_rate": 2.0743371325734856e-06, "loss": 0.0018, "step": 12909 }, { "epoch": 1.55, "learning_rate": 2.074265146970606e-06, "loss": 0.0002, "step": 12910 }, { "epoch": 1.55, "learning_rate": 2.0741931613677267e-06, "loss": 0.0002, "step": 12911 }, { "epoch": 1.55, "learning_rate": 2.074121175764847e-06, "loss": 0.0002, "step": 12912 }, { "epoch": 1.55, "learning_rate": 2.074049190161968e-06, "loss": 0.0002, "step": 12913 }, { "epoch": 1.55, "learning_rate": 2.073977204559088e-06, "loss": 0.0007, "step": 12914 }, { "epoch": 1.55, "learning_rate": 2.0739052189562085e-06, "loss": 0.0002, "step": 12915 }, { "epoch": 1.55, "learning_rate": 2.0738332333533293e-06, "loss": 0.0007, "step": 12916 }, { "epoch": 1.55, "learning_rate": 2.0737612477504497e-06, "loss": 0.0002, "step": 12917 }, { "epoch": 1.55, "learning_rate": 2.0736892621475705e-06, "loss": 0.0001, "step": 12918 }, { "epoch": 1.55, "learning_rate": 2.073617276544691e-06, "loss": 0.0002, "step": 12919 }, { "epoch": 1.55, "learning_rate": 2.0735452909418116e-06, "loss": 0.0003, "step": 12920 }, { "epoch": 1.55, "learning_rate": 2.0734733053389324e-06, "loss": 0.0002, "step": 12921 }, { "epoch": 1.55, "learning_rate": 2.073401319736053e-06, "loss": 0.0002, "step": 12922 }, { "epoch": 1.55, "learning_rate": 2.0733293341331735e-06, "loss": 0.0002, "step": 12923 }, { "epoch": 1.55, "learning_rate": 2.0732573485302943e-06, "loss": 0.0002, "step": 12924 }, { "epoch": 1.55, "learning_rate": 2.0731853629274146e-06, "loss": 0.0003, "step": 12925 }, { "epoch": 1.55, "learning_rate": 2.073113377324535e-06, "loss": 0.0002, "step": 12926 }, { "epoch": 1.55, "learning_rate": 2.0730413917216558e-06, "loss": 0.0001, "step": 12927 }, { "epoch": 1.55, "learning_rate": 2.072969406118776e-06, "loss": 0.0002, "step": 12928 }, { "epoch": 1.55, "learning_rate": 2.072897420515897e-06, "loss": 0.0002, "step": 12929 }, { "epoch": 1.55, "learning_rate": 2.0728254349130173e-06, "loss": 0.0003, "step": 12930 }, { "epoch": 1.55, "learning_rate": 2.072753449310138e-06, "loss": 0.0001, "step": 12931 }, { "epoch": 1.55, "learning_rate": 2.0726814637072584e-06, "loss": 0.0003, "step": 12932 }, { "epoch": 1.55, "learning_rate": 2.072609478104379e-06, "loss": 0.0016, "step": 12933 }, { "epoch": 1.55, "learning_rate": 2.0725374925014995e-06, "loss": 0.0002, "step": 12934 }, { "epoch": 1.55, "learning_rate": 2.0724655068986203e-06, "loss": 0.0001, "step": 12935 }, { "epoch": 1.55, "learning_rate": 2.072393521295741e-06, "loss": 0.0002, "step": 12936 }, { "epoch": 1.55, "learning_rate": 2.0723215356928614e-06, "loss": 0.0002, "step": 12937 }, { "epoch": 1.55, "learning_rate": 2.0722495500899822e-06, "loss": 0.0002, "step": 12938 }, { "epoch": 1.55, "learning_rate": 2.0721775644871026e-06, "loss": 0.8746, "step": 12939 }, { "epoch": 1.55, "learning_rate": 2.0721055788842233e-06, "loss": 0.0002, "step": 12940 }, { "epoch": 1.55, "learning_rate": 2.0720335932813437e-06, "loss": 0.0001, "step": 12941 }, { "epoch": 1.55, "learning_rate": 2.0719616076784645e-06, "loss": 1.0456, "step": 12942 }, { "epoch": 1.55, "learning_rate": 2.071889622075585e-06, "loss": 0.0001, "step": 12943 }, { "epoch": 1.55, "learning_rate": 2.0718176364727056e-06, "loss": 0.0004, "step": 12944 }, { "epoch": 1.55, "learning_rate": 2.071745650869826e-06, "loss": 0.0004, "step": 12945 }, { "epoch": 1.55, "learning_rate": 2.0716736652669467e-06, "loss": 0.0003, "step": 12946 }, { "epoch": 1.55, "learning_rate": 2.071601679664067e-06, "loss": 0.9988, "step": 12947 }, { "epoch": 1.55, "learning_rate": 2.071529694061188e-06, "loss": 0.0002, "step": 12948 }, { "epoch": 1.55, "learning_rate": 2.0714577084583082e-06, "loss": 0.0002, "step": 12949 }, { "epoch": 1.55, "learning_rate": 2.071385722855429e-06, "loss": 0.0003, "step": 12950 }, { "epoch": 1.55, "learning_rate": 2.0713137372525494e-06, "loss": 0.0003, "step": 12951 }, { "epoch": 1.55, "learning_rate": 2.07124175164967e-06, "loss": 0.0004, "step": 12952 }, { "epoch": 1.55, "learning_rate": 2.071169766046791e-06, "loss": 0.9308, "step": 12953 }, { "epoch": 1.55, "learning_rate": 2.0710977804439113e-06, "loss": 0.9655, "step": 12954 }, { "epoch": 1.55, "learning_rate": 2.071025794841032e-06, "loss": 0.0003, "step": 12955 }, { "epoch": 1.55, "learning_rate": 2.0709538092381524e-06, "loss": 0.0002, "step": 12956 }, { "epoch": 1.55, "learning_rate": 2.070881823635273e-06, "loss": 0.0002, "step": 12957 }, { "epoch": 1.55, "learning_rate": 2.0708098380323936e-06, "loss": 0.5987, "step": 12958 }, { "epoch": 1.55, "learning_rate": 2.0707378524295143e-06, "loss": 0.0002, "step": 12959 }, { "epoch": 1.55, "learning_rate": 2.0706658668266347e-06, "loss": 0.0002, "step": 12960 }, { "epoch": 1.55, "learning_rate": 2.070593881223755e-06, "loss": 0.0003, "step": 12961 }, { "epoch": 1.55, "learning_rate": 2.070521895620876e-06, "loss": 0.0004, "step": 12962 }, { "epoch": 1.55, "learning_rate": 2.070449910017996e-06, "loss": 0.0003, "step": 12963 }, { "epoch": 1.55, "learning_rate": 2.070377924415117e-06, "loss": 0.0002, "step": 12964 }, { "epoch": 1.55, "learning_rate": 2.0703059388122373e-06, "loss": 0.0008, "step": 12965 }, { "epoch": 1.55, "learning_rate": 2.070233953209358e-06, "loss": 0.0003, "step": 12966 }, { "epoch": 1.55, "learning_rate": 2.070161967606479e-06, "loss": 0.0002, "step": 12967 }, { "epoch": 1.55, "learning_rate": 2.0700899820035996e-06, "loss": 0.0008, "step": 12968 }, { "epoch": 1.55, "learning_rate": 2.07001799640072e-06, "loss": 0.05, "step": 12969 }, { "epoch": 1.55, "learning_rate": 2.0699460107978408e-06, "loss": 0.0002, "step": 12970 }, { "epoch": 1.55, "learning_rate": 2.069874025194961e-06, "loss": 0.0001, "step": 12971 }, { "epoch": 1.55, "learning_rate": 2.0698020395920815e-06, "loss": 0.0007, "step": 12972 }, { "epoch": 1.55, "learning_rate": 2.0697300539892023e-06, "loss": 0.0021, "step": 12973 }, { "epoch": 1.55, "learning_rate": 2.0696580683863226e-06, "loss": 0.0003, "step": 12974 }, { "epoch": 1.55, "learning_rate": 2.0695860827834434e-06, "loss": 0.0002, "step": 12975 }, { "epoch": 1.55, "learning_rate": 2.0695140971805638e-06, "loss": 0.0009, "step": 12976 }, { "epoch": 1.56, "learning_rate": 2.0694421115776845e-06, "loss": 1.0645, "step": 12977 }, { "epoch": 1.56, "learning_rate": 2.069370125974805e-06, "loss": 0.7177, "step": 12978 }, { "epoch": 1.56, "learning_rate": 2.0692981403719257e-06, "loss": 0.0006, "step": 12979 }, { "epoch": 1.56, "learning_rate": 2.069226154769046e-06, "loss": 0.0001, "step": 12980 }, { "epoch": 1.56, "learning_rate": 2.069154169166167e-06, "loss": 0.0002, "step": 12981 }, { "epoch": 1.56, "learning_rate": 2.0690821835632876e-06, "loss": 0.0009, "step": 12982 }, { "epoch": 1.56, "learning_rate": 2.069010197960408e-06, "loss": 0.0005, "step": 12983 }, { "epoch": 1.56, "learning_rate": 2.0689382123575287e-06, "loss": 0.0017, "step": 12984 }, { "epoch": 1.56, "learning_rate": 2.068866226754649e-06, "loss": 0.0003, "step": 12985 }, { "epoch": 1.56, "learning_rate": 2.06879424115177e-06, "loss": 0.0013, "step": 12986 }, { "epoch": 1.56, "learning_rate": 2.06872225554889e-06, "loss": 0.0004, "step": 12987 }, { "epoch": 1.56, "learning_rate": 2.068650269946011e-06, "loss": 0.0004, "step": 12988 }, { "epoch": 1.56, "learning_rate": 2.0685782843431313e-06, "loss": 0.0007, "step": 12989 }, { "epoch": 1.56, "learning_rate": 2.068506298740252e-06, "loss": 0.0003, "step": 12990 }, { "epoch": 1.56, "learning_rate": 2.0684343131373725e-06, "loss": 0.0003, "step": 12991 }, { "epoch": 1.56, "learning_rate": 2.0683623275344932e-06, "loss": 0.0006, "step": 12992 }, { "epoch": 1.56, "learning_rate": 2.0682903419316136e-06, "loss": 0.0005, "step": 12993 }, { "epoch": 1.56, "learning_rate": 2.0682183563287344e-06, "loss": 0.0052, "step": 12994 }, { "epoch": 1.56, "learning_rate": 2.0681463707258547e-06, "loss": 0.0004, "step": 12995 }, { "epoch": 1.56, "learning_rate": 2.0680743851229755e-06, "loss": 0.0002, "step": 12996 }, { "epoch": 1.56, "learning_rate": 2.068002399520096e-06, "loss": 0.0002, "step": 12997 }, { "epoch": 1.56, "learning_rate": 2.0679304139172166e-06, "loss": 0.0002, "step": 12998 }, { "epoch": 1.56, "learning_rate": 2.0678584283143374e-06, "loss": 0.0001, "step": 12999 }, { "epoch": 1.56, "learning_rate": 2.0677864427114578e-06, "loss": 0.0002, "step": 13000 }, { "epoch": 1.56, "learning_rate": 2.0677144571085786e-06, "loss": 0.0004, "step": 13001 }, { "epoch": 1.56, "learning_rate": 2.067642471505699e-06, "loss": 0.0002, "step": 13002 }, { "epoch": 1.56, "learning_rate": 2.0675704859028197e-06, "loss": 0.0001, "step": 13003 }, { "epoch": 1.56, "learning_rate": 2.06749850029994e-06, "loss": 0.0003, "step": 13004 }, { "epoch": 1.56, "learning_rate": 2.067426514697061e-06, "loss": 0.0003, "step": 13005 }, { "epoch": 1.56, "learning_rate": 2.067354529094181e-06, "loss": 0.0003, "step": 13006 }, { "epoch": 1.56, "learning_rate": 2.0672825434913015e-06, "loss": 0.0004, "step": 13007 }, { "epoch": 1.56, "learning_rate": 2.0672105578884223e-06, "loss": 0.0196, "step": 13008 }, { "epoch": 1.56, "learning_rate": 2.0671385722855427e-06, "loss": 0.0002, "step": 13009 }, { "epoch": 1.56, "learning_rate": 2.0670665866826634e-06, "loss": 0.0002, "step": 13010 }, { "epoch": 1.56, "learning_rate": 2.066994601079784e-06, "loss": 0.0005, "step": 13011 }, { "epoch": 1.56, "learning_rate": 2.0669226154769046e-06, "loss": 0.0003, "step": 13012 }, { "epoch": 1.56, "learning_rate": 2.0668506298740254e-06, "loss": 0.0004, "step": 13013 }, { "epoch": 1.56, "learning_rate": 2.066778644271146e-06, "loss": 0.0003, "step": 13014 }, { "epoch": 1.56, "learning_rate": 2.0667066586682665e-06, "loss": 0.0006, "step": 13015 }, { "epoch": 1.56, "learning_rate": 2.0666346730653873e-06, "loss": 0.0004, "step": 13016 }, { "epoch": 1.56, "learning_rate": 2.0665626874625076e-06, "loss": 0.0002, "step": 13017 }, { "epoch": 1.56, "learning_rate": 2.066490701859628e-06, "loss": 0.0004, "step": 13018 }, { "epoch": 1.56, "learning_rate": 2.0664187162567488e-06, "loss": 0.0005, "step": 13019 }, { "epoch": 1.56, "learning_rate": 2.066346730653869e-06, "loss": 0.0001, "step": 13020 }, { "epoch": 1.56, "learning_rate": 2.06627474505099e-06, "loss": 0.0004, "step": 13021 }, { "epoch": 1.56, "learning_rate": 2.0662027594481102e-06, "loss": 0.0002, "step": 13022 }, { "epoch": 1.56, "learning_rate": 2.066130773845231e-06, "loss": 0.0001, "step": 13023 }, { "epoch": 1.56, "learning_rate": 2.0660587882423514e-06, "loss": 0.0004, "step": 13024 }, { "epoch": 1.56, "learning_rate": 2.065986802639472e-06, "loss": 0.0002, "step": 13025 }, { "epoch": 1.56, "learning_rate": 2.0659148170365925e-06, "loss": 0.0002, "step": 13026 }, { "epoch": 1.56, "learning_rate": 2.0658428314337133e-06, "loss": 0.0003, "step": 13027 }, { "epoch": 1.56, "learning_rate": 2.065770845830834e-06, "loss": 0.0002, "step": 13028 }, { "epoch": 1.56, "learning_rate": 2.0656988602279544e-06, "loss": 0.0002, "step": 13029 }, { "epoch": 1.56, "learning_rate": 2.065626874625075e-06, "loss": 0.0003, "step": 13030 }, { "epoch": 1.56, "learning_rate": 2.0655548890221956e-06, "loss": 0.0002, "step": 13031 }, { "epoch": 1.56, "learning_rate": 2.0654829034193163e-06, "loss": 0.0001, "step": 13032 }, { "epoch": 1.56, "learning_rate": 2.0654109178164367e-06, "loss": 0.0003, "step": 13033 }, { "epoch": 1.56, "learning_rate": 2.0653389322135575e-06, "loss": 0.0003, "step": 13034 }, { "epoch": 1.56, "learning_rate": 2.065266946610678e-06, "loss": 0.0002, "step": 13035 }, { "epoch": 1.56, "learning_rate": 2.0651949610077986e-06, "loss": 0.0002, "step": 13036 }, { "epoch": 1.56, "learning_rate": 2.065122975404919e-06, "loss": 0.0005, "step": 13037 }, { "epoch": 1.56, "learning_rate": 2.0650509898020397e-06, "loss": 0.0003, "step": 13038 }, { "epoch": 1.56, "learning_rate": 2.06497900419916e-06, "loss": 0.0001, "step": 13039 }, { "epoch": 1.56, "learning_rate": 2.064907018596281e-06, "loss": 0.0007, "step": 13040 }, { "epoch": 1.56, "learning_rate": 2.0648350329934012e-06, "loss": 0.0002, "step": 13041 }, { "epoch": 1.56, "learning_rate": 2.0647630473905216e-06, "loss": 0.0004, "step": 13042 }, { "epoch": 1.56, "learning_rate": 2.0646910617876424e-06, "loss": 0.0004, "step": 13043 }, { "epoch": 1.56, "learning_rate": 2.064619076184763e-06, "loss": 0.0004, "step": 13044 }, { "epoch": 1.56, "learning_rate": 2.064547090581884e-06, "loss": 0.0022, "step": 13045 }, { "epoch": 1.56, "learning_rate": 2.0644751049790043e-06, "loss": 0.0002, "step": 13046 }, { "epoch": 1.56, "learning_rate": 2.064403119376125e-06, "loss": 0.0001, "step": 13047 }, { "epoch": 1.56, "learning_rate": 2.0643311337732454e-06, "loss": 0.0002, "step": 13048 }, { "epoch": 1.56, "learning_rate": 2.064259148170366e-06, "loss": 0.0003, "step": 13049 }, { "epoch": 1.56, "learning_rate": 2.0641871625674865e-06, "loss": 0.0005, "step": 13050 }, { "epoch": 1.56, "learning_rate": 2.0641151769646073e-06, "loss": 0.0003, "step": 13051 }, { "epoch": 1.56, "learning_rate": 2.0640431913617277e-06, "loss": 0.0002, "step": 13052 }, { "epoch": 1.56, "learning_rate": 2.063971205758848e-06, "loss": 0.0002, "step": 13053 }, { "epoch": 1.56, "learning_rate": 2.063899220155969e-06, "loss": 0.0002, "step": 13054 }, { "epoch": 1.56, "learning_rate": 2.063827234553089e-06, "loss": 0.0004, "step": 13055 }, { "epoch": 1.56, "learning_rate": 2.06375524895021e-06, "loss": 0.0001, "step": 13056 }, { "epoch": 1.56, "learning_rate": 2.0636832633473303e-06, "loss": 0.053, "step": 13057 }, { "epoch": 1.56, "learning_rate": 2.063611277744451e-06, "loss": 0.0003, "step": 13058 }, { "epoch": 1.56, "learning_rate": 2.063539292141572e-06, "loss": 0.0004, "step": 13059 }, { "epoch": 1.57, "learning_rate": 2.063467306538692e-06, "loss": 0.0006, "step": 13060 }, { "epoch": 1.57, "learning_rate": 2.063395320935813e-06, "loss": 0.0002, "step": 13061 }, { "epoch": 1.57, "learning_rate": 2.0633233353329338e-06, "loss": 0.0002, "step": 13062 }, { "epoch": 1.57, "learning_rate": 2.063251349730054e-06, "loss": 0.9375, "step": 13063 }, { "epoch": 1.57, "learning_rate": 2.063179364127175e-06, "loss": 0.0002, "step": 13064 }, { "epoch": 1.57, "learning_rate": 2.0631073785242953e-06, "loss": 0.0002, "step": 13065 }, { "epoch": 1.57, "learning_rate": 2.0630353929214156e-06, "loss": 0.0002, "step": 13066 }, { "epoch": 1.57, "learning_rate": 2.0629634073185364e-06, "loss": 0.0002, "step": 13067 }, { "epoch": 1.57, "learning_rate": 2.0628914217156567e-06, "loss": 0.0002, "step": 13068 }, { "epoch": 1.57, "learning_rate": 2.0628194361127775e-06, "loss": 0.0004, "step": 13069 }, { "epoch": 1.57, "learning_rate": 2.062747450509898e-06, "loss": 0.0002, "step": 13070 }, { "epoch": 1.57, "learning_rate": 2.0626754649070187e-06, "loss": 0.0665, "step": 13071 }, { "epoch": 1.57, "learning_rate": 2.062603479304139e-06, "loss": 0.0004, "step": 13072 }, { "epoch": 1.57, "learning_rate": 2.06253149370126e-06, "loss": 0.0002, "step": 13073 }, { "epoch": 1.57, "learning_rate": 2.06245950809838e-06, "loss": 0.0021, "step": 13074 }, { "epoch": 1.57, "learning_rate": 2.062387522495501e-06, "loss": 1.0657, "step": 13075 }, { "epoch": 1.57, "learning_rate": 2.0623155368926217e-06, "loss": 0.0002, "step": 13076 }, { "epoch": 1.57, "learning_rate": 2.062243551289742e-06, "loss": 0.0001, "step": 13077 }, { "epoch": 1.57, "learning_rate": 2.062171565686863e-06, "loss": 0.0007, "step": 13078 }, { "epoch": 1.57, "learning_rate": 2.062099580083983e-06, "loss": 0.0002, "step": 13079 }, { "epoch": 1.57, "learning_rate": 2.062027594481104e-06, "loss": 0.0003, "step": 13080 }, { "epoch": 1.57, "learning_rate": 2.0619556088782243e-06, "loss": 0.0002, "step": 13081 }, { "epoch": 1.57, "learning_rate": 2.061883623275345e-06, "loss": 0.0006, "step": 13082 }, { "epoch": 1.57, "learning_rate": 2.0618116376724655e-06, "loss": 0.0003, "step": 13083 }, { "epoch": 1.57, "learning_rate": 2.0617396520695862e-06, "loss": 0.0003, "step": 13084 }, { "epoch": 1.57, "learning_rate": 2.0616676664667066e-06, "loss": 0.0003, "step": 13085 }, { "epoch": 1.57, "learning_rate": 2.0615956808638274e-06, "loss": 0.1792, "step": 13086 }, { "epoch": 1.57, "learning_rate": 2.0615236952609477e-06, "loss": 0.0008, "step": 13087 }, { "epoch": 1.57, "learning_rate": 2.061451709658068e-06, "loss": 0.0004, "step": 13088 }, { "epoch": 1.57, "learning_rate": 2.061379724055189e-06, "loss": 0.0004, "step": 13089 }, { "epoch": 1.57, "learning_rate": 2.0613077384523096e-06, "loss": 0.0003, "step": 13090 }, { "epoch": 1.57, "learning_rate": 2.0612357528494304e-06, "loss": 0.0002, "step": 13091 }, { "epoch": 1.57, "learning_rate": 2.0611637672465508e-06, "loss": 0.0002, "step": 13092 }, { "epoch": 1.57, "learning_rate": 2.0610917816436716e-06, "loss": 0.0003, "step": 13093 }, { "epoch": 1.57, "learning_rate": 2.061019796040792e-06, "loss": 0.0015, "step": 13094 }, { "epoch": 1.57, "learning_rate": 2.0609478104379127e-06, "loss": 0.0003, "step": 13095 }, { "epoch": 1.57, "learning_rate": 2.060875824835033e-06, "loss": 0.0006, "step": 13096 }, { "epoch": 1.57, "learning_rate": 2.060803839232154e-06, "loss": 0.0003, "step": 13097 }, { "epoch": 1.57, "learning_rate": 2.060731853629274e-06, "loss": 0.0006, "step": 13098 }, { "epoch": 1.57, "learning_rate": 2.0606598680263945e-06, "loss": 0.0003, "step": 13099 }, { "epoch": 1.57, "learning_rate": 2.0605878824235153e-06, "loss": 0.0004, "step": 13100 }, { "epoch": 1.57, "learning_rate": 2.0605158968206357e-06, "loss": 0.0004, "step": 13101 }, { "epoch": 1.57, "learning_rate": 2.0604439112177564e-06, "loss": 0.0003, "step": 13102 }, { "epoch": 1.57, "learning_rate": 2.060371925614877e-06, "loss": 0.0002, "step": 13103 }, { "epoch": 1.57, "learning_rate": 2.0602999400119976e-06, "loss": 0.0002, "step": 13104 }, { "epoch": 1.57, "learning_rate": 2.0602279544091184e-06, "loss": 0.0001, "step": 13105 }, { "epoch": 1.57, "learning_rate": 2.0601559688062387e-06, "loss": 0.0003, "step": 13106 }, { "epoch": 1.57, "learning_rate": 2.0600839832033595e-06, "loss": 0.0005, "step": 13107 }, { "epoch": 1.57, "learning_rate": 2.0600119976004803e-06, "loss": 1.117, "step": 13108 }, { "epoch": 1.57, "learning_rate": 2.0599400119976006e-06, "loss": 0.0007, "step": 13109 }, { "epoch": 1.57, "learning_rate": 2.0598680263947214e-06, "loss": 0.0003, "step": 13110 }, { "epoch": 1.57, "learning_rate": 2.0597960407918418e-06, "loss": 0.0013, "step": 13111 }, { "epoch": 1.57, "learning_rate": 2.059724055188962e-06, "loss": 0.0004, "step": 13112 }, { "epoch": 1.57, "learning_rate": 2.059652069586083e-06, "loss": 0.0003, "step": 13113 }, { "epoch": 1.57, "learning_rate": 2.0595800839832032e-06, "loss": 0.0005, "step": 13114 }, { "epoch": 1.57, "learning_rate": 2.059508098380324e-06, "loss": 0.0003, "step": 13115 }, { "epoch": 1.57, "learning_rate": 2.0594361127774444e-06, "loss": 0.0128, "step": 13116 }, { "epoch": 1.57, "learning_rate": 2.059364127174565e-06, "loss": 0.0003, "step": 13117 }, { "epoch": 1.57, "learning_rate": 2.0592921415716855e-06, "loss": 0.0004, "step": 13118 }, { "epoch": 1.57, "learning_rate": 2.0592201559688063e-06, "loss": 0.0002, "step": 13119 }, { "epoch": 1.57, "learning_rate": 2.0591481703659266e-06, "loss": 0.0005, "step": 13120 }, { "epoch": 1.57, "learning_rate": 2.0590761847630474e-06, "loss": 0.0002, "step": 13121 }, { "epoch": 1.57, "learning_rate": 2.059004199160168e-06, "loss": 0.0002, "step": 13122 }, { "epoch": 1.57, "learning_rate": 2.0589322135572886e-06, "loss": 0.0003, "step": 13123 }, { "epoch": 1.57, "learning_rate": 2.0588602279544093e-06, "loss": 0.0003, "step": 13124 }, { "epoch": 1.57, "learning_rate": 2.0587882423515297e-06, "loss": 0.0002, "step": 13125 }, { "epoch": 1.57, "learning_rate": 2.0587162567486505e-06, "loss": 0.0002, "step": 13126 }, { "epoch": 1.57, "learning_rate": 2.058644271145771e-06, "loss": 0.0002, "step": 13127 }, { "epoch": 1.57, "learning_rate": 2.0585722855428916e-06, "loss": 0.0003, "step": 13128 }, { "epoch": 1.57, "learning_rate": 2.058500299940012e-06, "loss": 0.0004, "step": 13129 }, { "epoch": 1.57, "learning_rate": 2.0584283143371327e-06, "loss": 0.0006, "step": 13130 }, { "epoch": 1.57, "learning_rate": 2.058356328734253e-06, "loss": 0.0002, "step": 13131 }, { "epoch": 1.57, "learning_rate": 2.058284343131374e-06, "loss": 0.0005, "step": 13132 }, { "epoch": 1.57, "learning_rate": 2.0582123575284942e-06, "loss": 0.0002, "step": 13133 }, { "epoch": 1.57, "learning_rate": 2.0581403719256146e-06, "loss": 0.7911, "step": 13134 }, { "epoch": 1.57, "learning_rate": 2.0580683863227354e-06, "loss": 0.9558, "step": 13135 }, { "epoch": 1.57, "learning_rate": 2.057996400719856e-06, "loss": 0.0002, "step": 13136 }, { "epoch": 1.57, "learning_rate": 2.057924415116977e-06, "loss": 0.0002, "step": 13137 }, { "epoch": 1.57, "learning_rate": 2.0578524295140973e-06, "loss": 0.0002, "step": 13138 }, { "epoch": 1.57, "learning_rate": 2.057780443911218e-06, "loss": 0.0003, "step": 13139 }, { "epoch": 1.57, "learning_rate": 2.0577084583083384e-06, "loss": 0.0003, "step": 13140 }, { "epoch": 1.57, "learning_rate": 2.057636472705459e-06, "loss": 0.0003, "step": 13141 }, { "epoch": 1.57, "learning_rate": 2.0575644871025795e-06, "loss": 0.0004, "step": 13142 }, { "epoch": 1.57, "learning_rate": 2.0574925014997003e-06, "loss": 0.0006, "step": 13143 }, { "epoch": 1.58, "learning_rate": 2.0574205158968207e-06, "loss": 0.0003, "step": 13144 }, { "epoch": 1.58, "learning_rate": 2.057348530293941e-06, "loss": 0.0005, "step": 13145 }, { "epoch": 1.58, "learning_rate": 2.057276544691062e-06, "loss": 0.0001, "step": 13146 }, { "epoch": 1.58, "learning_rate": 2.057204559088182e-06, "loss": 0.0002, "step": 13147 }, { "epoch": 1.58, "learning_rate": 2.057132573485303e-06, "loss": 0.0002, "step": 13148 }, { "epoch": 1.58, "learning_rate": 2.0570605878824233e-06, "loss": 0.0002, "step": 13149 }, { "epoch": 1.58, "learning_rate": 2.056988602279544e-06, "loss": 0.0024, "step": 13150 }, { "epoch": 1.58, "learning_rate": 2.056916616676665e-06, "loss": 0.0006, "step": 13151 }, { "epoch": 1.58, "learning_rate": 2.056844631073785e-06, "loss": 0.0001, "step": 13152 }, { "epoch": 1.58, "learning_rate": 2.056772645470906e-06, "loss": 0.0003, "step": 13153 }, { "epoch": 1.58, "learning_rate": 2.0567006598680268e-06, "loss": 0.0002, "step": 13154 }, { "epoch": 1.58, "learning_rate": 2.056628674265147e-06, "loss": 0.0002, "step": 13155 }, { "epoch": 1.58, "learning_rate": 2.056556688662268e-06, "loss": 0.0222, "step": 13156 }, { "epoch": 1.58, "learning_rate": 2.0564847030593882e-06, "loss": 0.0003, "step": 13157 }, { "epoch": 1.58, "learning_rate": 2.0564127174565086e-06, "loss": 0.6786, "step": 13158 }, { "epoch": 1.58, "learning_rate": 2.0563407318536294e-06, "loss": 0.0002, "step": 13159 }, { "epoch": 1.58, "learning_rate": 2.0562687462507497e-06, "loss": 0.0002, "step": 13160 }, { "epoch": 1.58, "learning_rate": 2.0561967606478705e-06, "loss": 0.0002, "step": 13161 }, { "epoch": 1.58, "learning_rate": 2.056124775044991e-06, "loss": 0.0002, "step": 13162 }, { "epoch": 1.58, "learning_rate": 2.0560527894421116e-06, "loss": 0.6805, "step": 13163 }, { "epoch": 1.58, "learning_rate": 2.055980803839232e-06, "loss": 0.0004, "step": 13164 }, { "epoch": 1.58, "learning_rate": 2.0559088182363528e-06, "loss": 0.0003, "step": 13165 }, { "epoch": 1.58, "learning_rate": 2.055836832633473e-06, "loss": 0.0004, "step": 13166 }, { "epoch": 1.58, "learning_rate": 2.055764847030594e-06, "loss": 0.0004, "step": 13167 }, { "epoch": 1.58, "learning_rate": 2.0556928614277147e-06, "loss": 0.0003, "step": 13168 }, { "epoch": 1.58, "learning_rate": 2.055620875824835e-06, "loss": 0.0003, "step": 13169 }, { "epoch": 1.58, "learning_rate": 2.055548890221956e-06, "loss": 1.0436, "step": 13170 }, { "epoch": 1.58, "learning_rate": 2.055476904619076e-06, "loss": 0.0002, "step": 13171 }, { "epoch": 1.58, "learning_rate": 2.055404919016197e-06, "loss": 0.0004, "step": 13172 }, { "epoch": 1.58, "learning_rate": 2.0553329334133173e-06, "loss": 0.0003, "step": 13173 }, { "epoch": 1.58, "learning_rate": 2.055260947810438e-06, "loss": 0.0004, "step": 13174 }, { "epoch": 1.58, "learning_rate": 2.0551889622075585e-06, "loss": 0.0002, "step": 13175 }, { "epoch": 1.58, "learning_rate": 2.0551169766046792e-06, "loss": 0.0011, "step": 13176 }, { "epoch": 1.58, "learning_rate": 2.0550449910017996e-06, "loss": 0.0004, "step": 13177 }, { "epoch": 1.58, "learning_rate": 2.0549730053989204e-06, "loss": 0.0003, "step": 13178 }, { "epoch": 1.58, "learning_rate": 2.0549010197960407e-06, "loss": 0.0005, "step": 13179 }, { "epoch": 1.58, "learning_rate": 2.054829034193161e-06, "loss": 0.0004, "step": 13180 }, { "epoch": 1.58, "learning_rate": 2.054757048590282e-06, "loss": 0.0003, "step": 13181 }, { "epoch": 1.58, "learning_rate": 2.0546850629874026e-06, "loss": 0.7626, "step": 13182 }, { "epoch": 1.58, "learning_rate": 2.0546130773845234e-06, "loss": 0.0005, "step": 13183 }, { "epoch": 1.58, "learning_rate": 2.0545410917816438e-06, "loss": 0.0002, "step": 13184 }, { "epoch": 1.58, "learning_rate": 2.0544691061787645e-06, "loss": 0.0004, "step": 13185 }, { "epoch": 1.58, "learning_rate": 2.054397120575885e-06, "loss": 0.0005, "step": 13186 }, { "epoch": 1.58, "learning_rate": 2.0543251349730057e-06, "loss": 0.0003, "step": 13187 }, { "epoch": 1.58, "learning_rate": 2.054253149370126e-06, "loss": 0.0004, "step": 13188 }, { "epoch": 1.58, "learning_rate": 2.054181163767247e-06, "loss": 0.0004, "step": 13189 }, { "epoch": 1.58, "learning_rate": 2.054109178164367e-06, "loss": 0.0006, "step": 13190 }, { "epoch": 1.58, "learning_rate": 2.054037192561488e-06, "loss": 0.0004, "step": 13191 }, { "epoch": 1.58, "learning_rate": 2.0539652069586083e-06, "loss": 0.0001, "step": 13192 }, { "epoch": 1.58, "learning_rate": 2.0538932213557287e-06, "loss": 0.0013, "step": 13193 }, { "epoch": 1.58, "learning_rate": 2.0538212357528494e-06, "loss": 0.0003, "step": 13194 }, { "epoch": 1.58, "learning_rate": 2.0537492501499698e-06, "loss": 0.5906, "step": 13195 }, { "epoch": 1.58, "learning_rate": 2.0536772645470906e-06, "loss": 0.0003, "step": 13196 }, { "epoch": 1.58, "learning_rate": 2.053605278944211e-06, "loss": 0.0034, "step": 13197 }, { "epoch": 1.58, "learning_rate": 2.0535332933413317e-06, "loss": 0.0009, "step": 13198 }, { "epoch": 1.58, "learning_rate": 2.0534613077384525e-06, "loss": 0.0004, "step": 13199 }, { "epoch": 1.58, "learning_rate": 2.0533893221355733e-06, "loss": 0.0007, "step": 13200 }, { "epoch": 1.58, "learning_rate": 2.0533173365326936e-06, "loss": 0.0004, "step": 13201 }, { "epoch": 1.58, "learning_rate": 2.0532453509298144e-06, "loss": 0.0007, "step": 13202 }, { "epoch": 1.58, "learning_rate": 2.0531733653269347e-06, "loss": 0.0003, "step": 13203 }, { "epoch": 1.58, "learning_rate": 2.053101379724055e-06, "loss": 0.0073, "step": 13204 }, { "epoch": 1.58, "learning_rate": 2.053029394121176e-06, "loss": 0.0002, "step": 13205 }, { "epoch": 1.58, "learning_rate": 2.0529574085182962e-06, "loss": 0.1533, "step": 13206 }, { "epoch": 1.58, "learning_rate": 2.052885422915417e-06, "loss": 0.0002, "step": 13207 }, { "epoch": 1.58, "learning_rate": 2.0528134373125374e-06, "loss": 0.0003, "step": 13208 }, { "epoch": 1.58, "learning_rate": 2.052741451709658e-06, "loss": 0.0002, "step": 13209 }, { "epoch": 1.58, "learning_rate": 2.0526694661067785e-06, "loss": 0.0002, "step": 13210 }, { "epoch": 1.58, "learning_rate": 2.0525974805038993e-06, "loss": 0.0004, "step": 13211 }, { "epoch": 1.58, "learning_rate": 2.0525254949010196e-06, "loss": 0.0004, "step": 13212 }, { "epoch": 1.58, "learning_rate": 2.0524535092981404e-06, "loss": 0.0003, "step": 13213 }, { "epoch": 1.58, "learning_rate": 2.052381523695261e-06, "loss": 0.0004, "step": 13214 }, { "epoch": 1.58, "learning_rate": 2.0523095380923815e-06, "loss": 0.0004, "step": 13215 }, { "epoch": 1.58, "learning_rate": 2.0522375524895023e-06, "loss": 0.0003, "step": 13216 }, { "epoch": 1.58, "learning_rate": 2.0521655668866227e-06, "loss": 0.9708, "step": 13217 }, { "epoch": 1.58, "learning_rate": 2.0520935812837435e-06, "loss": 0.0003, "step": 13218 }, { "epoch": 1.58, "learning_rate": 2.052021595680864e-06, "loss": 0.0002, "step": 13219 }, { "epoch": 1.58, "learning_rate": 2.0519496100779846e-06, "loss": 0.0005, "step": 13220 }, { "epoch": 1.58, "learning_rate": 2.051877624475105e-06, "loss": 0.0004, "step": 13221 }, { "epoch": 1.58, "learning_rate": 2.0518056388722257e-06, "loss": 0.0002, "step": 13222 }, { "epoch": 1.58, "learning_rate": 2.051733653269346e-06, "loss": 0.0002, "step": 13223 }, { "epoch": 1.58, "learning_rate": 2.051661667666467e-06, "loss": 0.0005, "step": 13224 }, { "epoch": 1.58, "learning_rate": 2.0515896820635872e-06, "loss": 0.0004, "step": 13225 }, { "epoch": 1.58, "learning_rate": 2.0515176964607076e-06, "loss": 0.0004, "step": 13226 }, { "epoch": 1.59, "learning_rate": 2.0514457108578283e-06, "loss": 0.0002, "step": 13227 }, { "epoch": 1.59, "learning_rate": 2.051373725254949e-06, "loss": 0.0002, "step": 13228 }, { "epoch": 1.59, "learning_rate": 2.0513017396520695e-06, "loss": 0.0003, "step": 13229 }, { "epoch": 1.59, "learning_rate": 2.0512297540491903e-06, "loss": 0.0002, "step": 13230 }, { "epoch": 1.59, "learning_rate": 2.051157768446311e-06, "loss": 0.0004, "step": 13231 }, { "epoch": 1.59, "learning_rate": 2.0510857828434314e-06, "loss": 0.0002, "step": 13232 }, { "epoch": 1.59, "learning_rate": 2.051013797240552e-06, "loss": 0.0005, "step": 13233 }, { "epoch": 1.59, "learning_rate": 2.0509418116376725e-06, "loss": 0.0005, "step": 13234 }, { "epoch": 1.59, "learning_rate": 2.0508698260347933e-06, "loss": 0.0001, "step": 13235 }, { "epoch": 1.59, "learning_rate": 2.0507978404319137e-06, "loss": 0.0004, "step": 13236 }, { "epoch": 1.59, "learning_rate": 2.0507258548290344e-06, "loss": 0.0003, "step": 13237 }, { "epoch": 1.59, "learning_rate": 2.050653869226155e-06, "loss": 0.0002, "step": 13238 }, { "epoch": 1.59, "learning_rate": 2.050581883623275e-06, "loss": 0.0003, "step": 13239 }, { "epoch": 1.59, "learning_rate": 2.050509898020396e-06, "loss": 0.0004, "step": 13240 }, { "epoch": 1.59, "learning_rate": 2.0504379124175163e-06, "loss": 0.0009, "step": 13241 }, { "epoch": 1.59, "learning_rate": 2.050365926814637e-06, "loss": 0.0003, "step": 13242 }, { "epoch": 1.59, "learning_rate": 2.0502939412117574e-06, "loss": 0.0004, "step": 13243 }, { "epoch": 1.59, "learning_rate": 2.050221955608878e-06, "loss": 0.0004, "step": 13244 }, { "epoch": 1.59, "learning_rate": 2.050149970005999e-06, "loss": 0.0003, "step": 13245 }, { "epoch": 1.59, "learning_rate": 2.0500779844031198e-06, "loss": 0.0003, "step": 13246 }, { "epoch": 1.59, "learning_rate": 2.05000599880024e-06, "loss": 0.0001, "step": 13247 }, { "epoch": 1.59, "learning_rate": 2.049934013197361e-06, "loss": 0.0003, "step": 13248 }, { "epoch": 1.59, "learning_rate": 2.0498620275944812e-06, "loss": 0.0006, "step": 13249 }, { "epoch": 1.59, "learning_rate": 2.0497900419916016e-06, "loss": 0.0003, "step": 13250 }, { "epoch": 1.59, "learning_rate": 2.0497180563887224e-06, "loss": 0.0005, "step": 13251 }, { "epoch": 1.59, "learning_rate": 2.0496460707858427e-06, "loss": 0.0005, "step": 13252 }, { "epoch": 1.59, "learning_rate": 2.0495740851829635e-06, "loss": 0.0003, "step": 13253 }, { "epoch": 1.59, "learning_rate": 2.049502099580084e-06, "loss": 0.0003, "step": 13254 }, { "epoch": 1.59, "learning_rate": 2.0494301139772046e-06, "loss": 0.0004, "step": 13255 }, { "epoch": 1.59, "learning_rate": 2.049358128374325e-06, "loss": 0.0004, "step": 13256 }, { "epoch": 1.59, "learning_rate": 2.0492861427714458e-06, "loss": 0.0002, "step": 13257 }, { "epoch": 1.59, "learning_rate": 2.049214157168566e-06, "loss": 0.0003, "step": 13258 }, { "epoch": 1.59, "learning_rate": 2.049142171565687e-06, "loss": 0.0003, "step": 13259 }, { "epoch": 1.59, "learning_rate": 2.0490701859628077e-06, "loss": 0.8346, "step": 13260 }, { "epoch": 1.59, "learning_rate": 2.048998200359928e-06, "loss": 0.7225, "step": 13261 }, { "epoch": 1.59, "learning_rate": 2.048926214757049e-06, "loss": 0.0002, "step": 13262 }, { "epoch": 1.59, "learning_rate": 2.048854229154169e-06, "loss": 0.0003, "step": 13263 }, { "epoch": 1.59, "learning_rate": 2.04878224355129e-06, "loss": 0.0003, "step": 13264 }, { "epoch": 1.59, "learning_rate": 2.0487102579484103e-06, "loss": 0.0002, "step": 13265 }, { "epoch": 1.59, "learning_rate": 2.048638272345531e-06, "loss": 0.0006, "step": 13266 }, { "epoch": 1.59, "learning_rate": 2.0485662867426514e-06, "loss": 0.0003, "step": 13267 }, { "epoch": 1.59, "learning_rate": 2.0484943011397722e-06, "loss": 0.0265, "step": 13268 }, { "epoch": 1.59, "learning_rate": 2.0484223155368926e-06, "loss": 0.0004, "step": 13269 }, { "epoch": 1.59, "learning_rate": 2.0483503299340134e-06, "loss": 0.4362, "step": 13270 }, { "epoch": 1.59, "learning_rate": 2.0482783443311337e-06, "loss": 0.0007, "step": 13271 }, { "epoch": 1.59, "learning_rate": 2.048206358728254e-06, "loss": 0.3264, "step": 13272 }, { "epoch": 1.59, "learning_rate": 2.048134373125375e-06, "loss": 0.0004, "step": 13273 }, { "epoch": 1.59, "learning_rate": 2.0480623875224956e-06, "loss": 0.0012, "step": 13274 }, { "epoch": 1.59, "learning_rate": 2.047990401919616e-06, "loss": 0.0003, "step": 13275 }, { "epoch": 1.59, "learning_rate": 2.0479184163167368e-06, "loss": 0.0009, "step": 13276 }, { "epoch": 1.59, "learning_rate": 2.0478464307138575e-06, "loss": 0.0006, "step": 13277 }, { "epoch": 1.59, "learning_rate": 2.047774445110978e-06, "loss": 0.0004, "step": 13278 }, { "epoch": 1.59, "learning_rate": 2.0477024595080987e-06, "loss": 0.0005, "step": 13279 }, { "epoch": 1.59, "learning_rate": 2.047630473905219e-06, "loss": 0.0003, "step": 13280 }, { "epoch": 1.59, "learning_rate": 2.04755848830234e-06, "loss": 0.0006, "step": 13281 }, { "epoch": 1.59, "learning_rate": 2.04748650269946e-06, "loss": 0.001, "step": 13282 }, { "epoch": 1.59, "learning_rate": 2.047414517096581e-06, "loss": 0.0004, "step": 13283 }, { "epoch": 1.59, "learning_rate": 2.0473425314937013e-06, "loss": 0.0003, "step": 13284 }, { "epoch": 1.59, "learning_rate": 2.0472705458908216e-06, "loss": 0.0003, "step": 13285 }, { "epoch": 1.59, "learning_rate": 2.0471985602879424e-06, "loss": 0.0005, "step": 13286 }, { "epoch": 1.59, "learning_rate": 2.0471265746850628e-06, "loss": 0.0003, "step": 13287 }, { "epoch": 1.59, "learning_rate": 2.0470545890821836e-06, "loss": 0.0003, "step": 13288 }, { "epoch": 1.59, "learning_rate": 2.046982603479304e-06, "loss": 0.0002, "step": 13289 }, { "epoch": 1.59, "learning_rate": 2.0469106178764247e-06, "loss": 0.0141, "step": 13290 }, { "epoch": 1.59, "learning_rate": 2.0468386322735455e-06, "loss": 0.0003, "step": 13291 }, { "epoch": 1.59, "learning_rate": 2.0467666466706662e-06, "loss": 0.0003, "step": 13292 }, { "epoch": 1.59, "learning_rate": 2.0466946610677866e-06, "loss": 0.0005, "step": 13293 }, { "epoch": 1.59, "learning_rate": 2.0466226754649074e-06, "loss": 0.0006, "step": 13294 }, { "epoch": 1.59, "learning_rate": 2.0465506898620277e-06, "loss": 0.3725, "step": 13295 }, { "epoch": 1.59, "learning_rate": 2.046478704259148e-06, "loss": 0.0004, "step": 13296 }, { "epoch": 1.59, "learning_rate": 2.046406718656269e-06, "loss": 0.0007, "step": 13297 }, { "epoch": 1.59, "learning_rate": 2.0463347330533892e-06, "loss": 0.0007, "step": 13298 }, { "epoch": 1.59, "learning_rate": 2.04626274745051e-06, "loss": 0.0029, "step": 13299 }, { "epoch": 1.59, "learning_rate": 2.0461907618476304e-06, "loss": 0.0006, "step": 13300 }, { "epoch": 1.59, "learning_rate": 2.046118776244751e-06, "loss": 0.0004, "step": 13301 }, { "epoch": 1.59, "learning_rate": 2.0460467906418715e-06, "loss": 0.0002, "step": 13302 }, { "epoch": 1.59, "learning_rate": 2.0459748050389923e-06, "loss": 0.0005, "step": 13303 }, { "epoch": 1.59, "learning_rate": 2.0459028194361126e-06, "loss": 0.0002, "step": 13304 }, { "epoch": 1.59, "learning_rate": 2.0458308338332334e-06, "loss": 0.0009, "step": 13305 }, { "epoch": 1.59, "learning_rate": 2.045758848230354e-06, "loss": 0.0004, "step": 13306 }, { "epoch": 1.59, "learning_rate": 2.0456868626274745e-06, "loss": 0.0003, "step": 13307 }, { "epoch": 1.59, "learning_rate": 2.0456148770245953e-06, "loss": 0.0003, "step": 13308 }, { "epoch": 1.59, "learning_rate": 2.0455428914217157e-06, "loss": 0.0004, "step": 13309 }, { "epoch": 1.59, "learning_rate": 2.0454709058188365e-06, "loss": 0.0005, "step": 13310 }, { "epoch": 1.6, "learning_rate": 2.045398920215957e-06, "loss": 0.0004, "step": 13311 }, { "epoch": 1.6, "learning_rate": 2.0453269346130776e-06, "loss": 0.0003, "step": 13312 }, { "epoch": 1.6, "learning_rate": 2.045254949010198e-06, "loss": 0.241, "step": 13313 }, { "epoch": 1.6, "learning_rate": 2.0451829634073187e-06, "loss": 0.0003, "step": 13314 }, { "epoch": 1.6, "learning_rate": 2.045110977804439e-06, "loss": 0.0004, "step": 13315 }, { "epoch": 1.6, "learning_rate": 2.04503899220156e-06, "loss": 0.0005, "step": 13316 }, { "epoch": 1.6, "learning_rate": 2.04496700659868e-06, "loss": 0.0005, "step": 13317 }, { "epoch": 1.6, "learning_rate": 2.044895020995801e-06, "loss": 0.0004, "step": 13318 }, { "epoch": 1.6, "learning_rate": 2.0448230353929213e-06, "loss": 0.0006, "step": 13319 }, { "epoch": 1.6, "learning_rate": 2.0447510497900417e-06, "loss": 0.0003, "step": 13320 }, { "epoch": 1.6, "learning_rate": 2.0446790641871625e-06, "loss": 0.0009, "step": 13321 }, { "epoch": 1.6, "learning_rate": 2.0446070785842833e-06, "loss": 0.0004, "step": 13322 }, { "epoch": 1.6, "learning_rate": 2.044535092981404e-06, "loss": 0.626, "step": 13323 }, { "epoch": 1.6, "learning_rate": 2.0444631073785244e-06, "loss": 0.0003, "step": 13324 }, { "epoch": 1.6, "learning_rate": 2.044391121775645e-06, "loss": 0.0004, "step": 13325 }, { "epoch": 1.6, "learning_rate": 2.0443191361727655e-06, "loss": 0.0006, "step": 13326 }, { "epoch": 1.6, "learning_rate": 2.0442471505698863e-06, "loss": 0.0006, "step": 13327 }, { "epoch": 1.6, "learning_rate": 2.0441751649670067e-06, "loss": 0.0003, "step": 13328 }, { "epoch": 1.6, "learning_rate": 2.0441031793641274e-06, "loss": 0.0002, "step": 13329 }, { "epoch": 1.6, "learning_rate": 2.0440311937612478e-06, "loss": 0.0002, "step": 13330 }, { "epoch": 1.6, "learning_rate": 2.043959208158368e-06, "loss": 0.0005, "step": 13331 }, { "epoch": 1.6, "learning_rate": 2.043887222555489e-06, "loss": 0.0002, "step": 13332 }, { "epoch": 1.6, "learning_rate": 2.0438152369526093e-06, "loss": 0.0008, "step": 13333 }, { "epoch": 1.6, "learning_rate": 2.04374325134973e-06, "loss": 0.0031, "step": 13334 }, { "epoch": 1.6, "learning_rate": 2.0436712657468504e-06, "loss": 0.0004, "step": 13335 }, { "epoch": 1.6, "learning_rate": 2.043599280143971e-06, "loss": 0.0003, "step": 13336 }, { "epoch": 1.6, "learning_rate": 2.043527294541092e-06, "loss": 0.0003, "step": 13337 }, { "epoch": 1.6, "learning_rate": 2.0434553089382123e-06, "loss": 0.0002, "step": 13338 }, { "epoch": 1.6, "learning_rate": 2.043383323335333e-06, "loss": 0.0003, "step": 13339 }, { "epoch": 1.6, "learning_rate": 2.043311337732454e-06, "loss": 0.0004, "step": 13340 }, { "epoch": 1.6, "learning_rate": 2.0432393521295742e-06, "loss": 0.0001, "step": 13341 }, { "epoch": 1.6, "learning_rate": 2.0431673665266946e-06, "loss": 0.0004, "step": 13342 }, { "epoch": 1.6, "learning_rate": 2.0430953809238154e-06, "loss": 0.0003, "step": 13343 }, { "epoch": 1.6, "learning_rate": 2.0430233953209357e-06, "loss": 0.0015, "step": 13344 }, { "epoch": 1.6, "learning_rate": 2.0429514097180565e-06, "loss": 0.0005, "step": 13345 }, { "epoch": 1.6, "learning_rate": 2.042879424115177e-06, "loss": 0.0002, "step": 13346 }, { "epoch": 1.6, "learning_rate": 2.0428074385122976e-06, "loss": 0.1526, "step": 13347 }, { "epoch": 1.6, "learning_rate": 2.042735452909418e-06, "loss": 0.0005, "step": 13348 }, { "epoch": 1.6, "learning_rate": 2.0426634673065388e-06, "loss": 0.0003, "step": 13349 }, { "epoch": 1.6, "learning_rate": 2.042591481703659e-06, "loss": 0.6201, "step": 13350 }, { "epoch": 1.6, "learning_rate": 2.04251949610078e-06, "loss": 0.0003, "step": 13351 }, { "epoch": 1.6, "learning_rate": 2.0424475104979003e-06, "loss": 0.0315, "step": 13352 }, { "epoch": 1.6, "learning_rate": 2.042375524895021e-06, "loss": 0.0002, "step": 13353 }, { "epoch": 1.6, "learning_rate": 2.042303539292142e-06, "loss": 0.0002, "step": 13354 }, { "epoch": 1.6, "learning_rate": 2.042231553689262e-06, "loss": 0.0005, "step": 13355 }, { "epoch": 1.6, "learning_rate": 2.042159568086383e-06, "loss": 0.0005, "step": 13356 }, { "epoch": 1.6, "learning_rate": 2.0420875824835033e-06, "loss": 0.0045, "step": 13357 }, { "epoch": 1.6, "learning_rate": 2.042015596880624e-06, "loss": 0.0003, "step": 13358 }, { "epoch": 1.6, "learning_rate": 2.0419436112777444e-06, "loss": 0.8921, "step": 13359 }, { "epoch": 1.6, "learning_rate": 2.0418716256748652e-06, "loss": 0.0004, "step": 13360 }, { "epoch": 1.6, "learning_rate": 2.0417996400719856e-06, "loss": 0.0004, "step": 13361 }, { "epoch": 1.6, "learning_rate": 2.0417276544691063e-06, "loss": 0.0002, "step": 13362 }, { "epoch": 1.6, "learning_rate": 2.0416556688662267e-06, "loss": 0.0003, "step": 13363 }, { "epoch": 1.6, "learning_rate": 2.0415836832633475e-06, "loss": 0.0003, "step": 13364 }, { "epoch": 1.6, "learning_rate": 2.041511697660468e-06, "loss": 0.0004, "step": 13365 }, { "epoch": 1.6, "learning_rate": 2.041439712057588e-06, "loss": 0.0003, "step": 13366 }, { "epoch": 1.6, "learning_rate": 2.041367726454709e-06, "loss": 0.0002, "step": 13367 }, { "epoch": 1.6, "learning_rate": 2.0412957408518297e-06, "loss": 0.0071, "step": 13368 }, { "epoch": 1.6, "learning_rate": 2.0412237552489505e-06, "loss": 0.0006, "step": 13369 }, { "epoch": 1.6, "learning_rate": 2.041151769646071e-06, "loss": 0.0004, "step": 13370 }, { "epoch": 1.6, "learning_rate": 2.0410797840431917e-06, "loss": 0.0003, "step": 13371 }, { "epoch": 1.6, "learning_rate": 2.041007798440312e-06, "loss": 0.0005, "step": 13372 }, { "epoch": 1.6, "learning_rate": 2.040935812837433e-06, "loss": 0.5803, "step": 13373 }, { "epoch": 1.6, "learning_rate": 2.040863827234553e-06, "loss": 0.0003, "step": 13374 }, { "epoch": 1.6, "learning_rate": 2.040791841631674e-06, "loss": 0.0003, "step": 13375 }, { "epoch": 1.6, "learning_rate": 2.0407198560287943e-06, "loss": 0.0004, "step": 13376 }, { "epoch": 1.6, "learning_rate": 2.0406478704259146e-06, "loss": 0.0002, "step": 13377 }, { "epoch": 1.6, "learning_rate": 2.0405758848230354e-06, "loss": 0.0005, "step": 13378 }, { "epoch": 1.6, "learning_rate": 2.0405038992201558e-06, "loss": 0.0003, "step": 13379 }, { "epoch": 1.6, "learning_rate": 2.0404319136172765e-06, "loss": 0.0003, "step": 13380 }, { "epoch": 1.6, "learning_rate": 2.040359928014397e-06, "loss": 0.0004, "step": 13381 }, { "epoch": 1.6, "learning_rate": 2.0402879424115177e-06, "loss": 0.0003, "step": 13382 }, { "epoch": 1.6, "learning_rate": 2.0402159568086385e-06, "loss": 0.0004, "step": 13383 }, { "epoch": 1.6, "learning_rate": 2.040143971205759e-06, "loss": 0.0003, "step": 13384 }, { "epoch": 1.6, "learning_rate": 2.0400719856028796e-06, "loss": 0.0006, "step": 13385 }, { "epoch": 1.6, "learning_rate": 2.0400000000000004e-06, "loss": 0.0004, "step": 13386 }, { "epoch": 1.6, "learning_rate": 2.0399280143971207e-06, "loss": 0.0005, "step": 13387 }, { "epoch": 1.6, "learning_rate": 2.039856028794241e-06, "loss": 0.0003, "step": 13388 }, { "epoch": 1.6, "learning_rate": 2.039784043191362e-06, "loss": 0.0013, "step": 13389 }, { "epoch": 1.6, "learning_rate": 2.0397120575884822e-06, "loss": 0.0002, "step": 13390 }, { "epoch": 1.6, "learning_rate": 2.039640071985603e-06, "loss": 0.0003, "step": 13391 }, { "epoch": 1.6, "learning_rate": 2.0395680863827234e-06, "loss": 0.0003, "step": 13392 }, { "epoch": 1.6, "learning_rate": 2.039496100779844e-06, "loss": 0.0004, "step": 13393 }, { "epoch": 1.61, "learning_rate": 2.0394241151769645e-06, "loss": 0.0009, "step": 13394 }, { "epoch": 1.61, "learning_rate": 2.0393521295740853e-06, "loss": 0.0003, "step": 13395 }, { "epoch": 1.61, "learning_rate": 2.0392801439712056e-06, "loss": 0.0003, "step": 13396 }, { "epoch": 1.61, "learning_rate": 2.0392081583683264e-06, "loss": 0.0002, "step": 13397 }, { "epoch": 1.61, "learning_rate": 2.0391361727654468e-06, "loss": 0.0002, "step": 13398 }, { "epoch": 1.61, "learning_rate": 2.0390641871625675e-06, "loss": 0.0002, "step": 13399 }, { "epoch": 1.61, "learning_rate": 2.0389922015596883e-06, "loss": 0.0031, "step": 13400 }, { "epoch": 1.61, "learning_rate": 2.0389202159568087e-06, "loss": 0.0003, "step": 13401 }, { "epoch": 1.61, "learning_rate": 2.0388482303539294e-06, "loss": 0.0003, "step": 13402 }, { "epoch": 1.61, "learning_rate": 2.03877624475105e-06, "loss": 0.0005, "step": 13403 }, { "epoch": 1.61, "learning_rate": 2.0387042591481706e-06, "loss": 0.0005, "step": 13404 }, { "epoch": 1.61, "learning_rate": 2.038632273545291e-06, "loss": 0.0002, "step": 13405 }, { "epoch": 1.61, "learning_rate": 2.0385602879424117e-06, "loss": 0.0002, "step": 13406 }, { "epoch": 1.61, "learning_rate": 2.038488302339532e-06, "loss": 0.0002, "step": 13407 }, { "epoch": 1.61, "learning_rate": 2.038416316736653e-06, "loss": 0.0111, "step": 13408 }, { "epoch": 1.61, "learning_rate": 2.038344331133773e-06, "loss": 0.0003, "step": 13409 }, { "epoch": 1.61, "learning_rate": 2.038272345530894e-06, "loss": 0.0002, "step": 13410 }, { "epoch": 1.61, "learning_rate": 2.0382003599280143e-06, "loss": 0.0019, "step": 13411 }, { "epoch": 1.61, "learning_rate": 2.0381283743251347e-06, "loss": 0.0003, "step": 13412 }, { "epoch": 1.61, "learning_rate": 2.0380563887222555e-06, "loss": 0.0002, "step": 13413 }, { "epoch": 1.61, "learning_rate": 2.0379844031193762e-06, "loss": 0.0002, "step": 13414 }, { "epoch": 1.61, "learning_rate": 2.037912417516497e-06, "loss": 0.0004, "step": 13415 }, { "epoch": 1.61, "learning_rate": 2.0378404319136174e-06, "loss": 0.0002, "step": 13416 }, { "epoch": 1.61, "learning_rate": 2.037768446310738e-06, "loss": 0.001, "step": 13417 }, { "epoch": 1.61, "learning_rate": 2.0376964607078585e-06, "loss": 0.0003, "step": 13418 }, { "epoch": 1.61, "learning_rate": 2.0376244751049793e-06, "loss": 0.0003, "step": 13419 }, { "epoch": 1.61, "learning_rate": 2.0375524895020996e-06, "loss": 0.0001, "step": 13420 }, { "epoch": 1.61, "learning_rate": 2.0374805038992204e-06, "loss": 0.0003, "step": 13421 }, { "epoch": 1.61, "learning_rate": 2.0374085182963408e-06, "loss": 0.0004, "step": 13422 }, { "epoch": 1.61, "learning_rate": 2.037336532693461e-06, "loss": 0.0408, "step": 13423 }, { "epoch": 1.61, "learning_rate": 2.037264547090582e-06, "loss": 0.0002, "step": 13424 }, { "epoch": 1.61, "learning_rate": 2.0371925614877023e-06, "loss": 0.0007, "step": 13425 }, { "epoch": 1.61, "learning_rate": 2.037120575884823e-06, "loss": 0.0002, "step": 13426 }, { "epoch": 1.61, "learning_rate": 2.0370485902819434e-06, "loss": 0.0002, "step": 13427 }, { "epoch": 1.61, "learning_rate": 2.036976604679064e-06, "loss": 0.0005, "step": 13428 }, { "epoch": 1.61, "learning_rate": 2.036904619076185e-06, "loss": 0.0009, "step": 13429 }, { "epoch": 1.61, "learning_rate": 2.0368326334733053e-06, "loss": 0.0003, "step": 13430 }, { "epoch": 1.61, "learning_rate": 2.036760647870426e-06, "loss": 0.0002, "step": 13431 }, { "epoch": 1.61, "learning_rate": 2.036688662267547e-06, "loss": 0.0002, "step": 13432 }, { "epoch": 1.61, "learning_rate": 2.0366166766646672e-06, "loss": 0.0001, "step": 13433 }, { "epoch": 1.61, "learning_rate": 2.0365446910617876e-06, "loss": 0.0002, "step": 13434 }, { "epoch": 1.61, "learning_rate": 2.0364727054589084e-06, "loss": 0.0011, "step": 13435 }, { "epoch": 1.61, "learning_rate": 2.0364007198560287e-06, "loss": 1.0203, "step": 13436 }, { "epoch": 1.61, "learning_rate": 2.0363287342531495e-06, "loss": 0.0003, "step": 13437 }, { "epoch": 1.61, "learning_rate": 2.03625674865027e-06, "loss": 0.0011, "step": 13438 }, { "epoch": 1.61, "learning_rate": 2.0361847630473906e-06, "loss": 0.0001, "step": 13439 }, { "epoch": 1.61, "learning_rate": 2.036112777444511e-06, "loss": 0.0003, "step": 13440 }, { "epoch": 1.61, "learning_rate": 2.0360407918416318e-06, "loss": 0.0003, "step": 13441 }, { "epoch": 1.61, "learning_rate": 2.035968806238752e-06, "loss": 0.0002, "step": 13442 }, { "epoch": 1.61, "learning_rate": 2.035896820635873e-06, "loss": 0.0002, "step": 13443 }, { "epoch": 1.61, "learning_rate": 2.0358248350329932e-06, "loss": 0.0002, "step": 13444 }, { "epoch": 1.61, "learning_rate": 2.035752849430114e-06, "loss": 0.0002, "step": 13445 }, { "epoch": 1.61, "learning_rate": 2.035680863827235e-06, "loss": 0.0004, "step": 13446 }, { "epoch": 1.61, "learning_rate": 2.035608878224355e-06, "loss": 0.0004, "step": 13447 }, { "epoch": 1.61, "learning_rate": 2.035536892621476e-06, "loss": 0.0002, "step": 13448 }, { "epoch": 1.61, "learning_rate": 2.0354649070185963e-06, "loss": 0.1721, "step": 13449 }, { "epoch": 1.61, "learning_rate": 2.035392921415717e-06, "loss": 0.1003, "step": 13450 }, { "epoch": 1.61, "learning_rate": 2.0353209358128374e-06, "loss": 0.0002, "step": 13451 }, { "epoch": 1.61, "learning_rate": 2.035248950209958e-06, "loss": 0.0003, "step": 13452 }, { "epoch": 1.61, "learning_rate": 2.0351769646070786e-06, "loss": 0.0005, "step": 13453 }, { "epoch": 1.61, "learning_rate": 2.0351049790041993e-06, "loss": 0.0004, "step": 13454 }, { "epoch": 1.61, "learning_rate": 2.0350329934013197e-06, "loss": 0.0061, "step": 13455 }, { "epoch": 1.61, "learning_rate": 2.0349610077984405e-06, "loss": 0.0002, "step": 13456 }, { "epoch": 1.61, "learning_rate": 2.034889022195561e-06, "loss": 0.0028, "step": 13457 }, { "epoch": 1.61, "learning_rate": 2.034817036592681e-06, "loss": 0.0003, "step": 13458 }, { "epoch": 1.61, "learning_rate": 2.034745050989802e-06, "loss": 0.0002, "step": 13459 }, { "epoch": 1.61, "learning_rate": 2.0346730653869227e-06, "loss": 0.0002, "step": 13460 }, { "epoch": 1.61, "learning_rate": 2.034601079784043e-06, "loss": 0.0006, "step": 13461 }, { "epoch": 1.61, "learning_rate": 2.034529094181164e-06, "loss": 0.0005, "step": 13462 }, { "epoch": 1.61, "learning_rate": 2.0344571085782847e-06, "loss": 0.7806, "step": 13463 }, { "epoch": 1.61, "learning_rate": 2.034385122975405e-06, "loss": 0.0003, "step": 13464 }, { "epoch": 1.61, "learning_rate": 2.0343131373725258e-06, "loss": 0.0006, "step": 13465 }, { "epoch": 1.61, "learning_rate": 2.034241151769646e-06, "loss": 0.0002, "step": 13466 }, { "epoch": 1.61, "learning_rate": 2.034169166166767e-06, "loss": 0.0002, "step": 13467 }, { "epoch": 1.61, "learning_rate": 2.0340971805638873e-06, "loss": 0.0002, "step": 13468 }, { "epoch": 1.61, "learning_rate": 2.0340251949610076e-06, "loss": 0.0003, "step": 13469 }, { "epoch": 1.61, "learning_rate": 2.0339532093581284e-06, "loss": 0.0003, "step": 13470 }, { "epoch": 1.61, "learning_rate": 2.0338812237552488e-06, "loss": 0.0002, "step": 13471 }, { "epoch": 1.61, "learning_rate": 2.0338092381523695e-06, "loss": 0.0002, "step": 13472 }, { "epoch": 1.61, "learning_rate": 2.03373725254949e-06, "loss": 0.0002, "step": 13473 }, { "epoch": 1.61, "learning_rate": 2.0336652669466107e-06, "loss": 0.0004, "step": 13474 }, { "epoch": 1.61, "learning_rate": 2.033593281343731e-06, "loss": 0.0003, "step": 13475 }, { "epoch": 1.61, "learning_rate": 2.033521295740852e-06, "loss": 0.0005, "step": 13476 }, { "epoch": 1.61, "learning_rate": 2.0334493101379726e-06, "loss": 0.5887, "step": 13477 }, { "epoch": 1.62, "learning_rate": 2.0333773245350934e-06, "loss": 0.6412, "step": 13478 }, { "epoch": 1.62, "learning_rate": 2.0333053389322137e-06, "loss": 0.0002, "step": 13479 }, { "epoch": 1.62, "learning_rate": 2.033233353329334e-06, "loss": 0.0002, "step": 13480 }, { "epoch": 1.62, "learning_rate": 2.033161367726455e-06, "loss": 0.0002, "step": 13481 }, { "epoch": 1.62, "learning_rate": 2.033089382123575e-06, "loss": 0.0003, "step": 13482 }, { "epoch": 1.62, "learning_rate": 2.033017396520696e-06, "loss": 0.0004, "step": 13483 }, { "epoch": 1.62, "learning_rate": 2.0329454109178163e-06, "loss": 0.0003, "step": 13484 }, { "epoch": 1.62, "learning_rate": 2.032873425314937e-06, "loss": 0.0001, "step": 13485 }, { "epoch": 1.62, "learning_rate": 2.0328014397120575e-06, "loss": 0.0002, "step": 13486 }, { "epoch": 1.62, "learning_rate": 2.0327294541091783e-06, "loss": 0.9406, "step": 13487 }, { "epoch": 1.62, "learning_rate": 2.0326574685062986e-06, "loss": 0.0002, "step": 13488 }, { "epoch": 1.62, "learning_rate": 2.0325854829034194e-06, "loss": 1.055, "step": 13489 }, { "epoch": 1.62, "learning_rate": 2.0325134973005397e-06, "loss": 0.0004, "step": 13490 }, { "epoch": 1.62, "learning_rate": 2.0324415116976605e-06, "loss": 0.0002, "step": 13491 }, { "epoch": 1.62, "learning_rate": 2.0323695260947813e-06, "loss": 0.0003, "step": 13492 }, { "epoch": 1.62, "learning_rate": 2.0322975404919017e-06, "loss": 0.0002, "step": 13493 }, { "epoch": 1.62, "learning_rate": 2.0322255548890224e-06, "loss": 0.0003, "step": 13494 }, { "epoch": 1.62, "learning_rate": 2.032153569286143e-06, "loss": 0.0005, "step": 13495 }, { "epoch": 1.62, "learning_rate": 2.0320815836832636e-06, "loss": 0.0004, "step": 13496 }, { "epoch": 1.62, "learning_rate": 2.032009598080384e-06, "loss": 0.0003, "step": 13497 }, { "epoch": 1.62, "learning_rate": 2.0319376124775047e-06, "loss": 0.0003, "step": 13498 }, { "epoch": 1.62, "learning_rate": 2.031865626874625e-06, "loss": 0.0002, "step": 13499 }, { "epoch": 1.62, "learning_rate": 2.031793641271746e-06, "loss": 0.0005, "step": 13500 }, { "epoch": 1.62, "learning_rate": 2.031721655668866e-06, "loss": 0.0002, "step": 13501 }, { "epoch": 1.62, "learning_rate": 2.031649670065987e-06, "loss": 0.0007, "step": 13502 }, { "epoch": 1.62, "learning_rate": 2.0315776844631073e-06, "loss": 0.0002, "step": 13503 }, { "epoch": 1.62, "learning_rate": 2.0315056988602277e-06, "loss": 0.0003, "step": 13504 }, { "epoch": 1.62, "learning_rate": 2.0314337132573485e-06, "loss": 0.0003, "step": 13505 }, { "epoch": 1.62, "learning_rate": 2.0313617276544692e-06, "loss": 0.0397, "step": 13506 }, { "epoch": 1.62, "learning_rate": 2.0312897420515896e-06, "loss": 0.0004, "step": 13507 }, { "epoch": 1.62, "learning_rate": 2.0312177564487104e-06, "loss": 0.0004, "step": 13508 }, { "epoch": 1.62, "learning_rate": 2.031145770845831e-06, "loss": 0.0004, "step": 13509 }, { "epoch": 1.62, "learning_rate": 2.0310737852429515e-06, "loss": 0.6308, "step": 13510 }, { "epoch": 1.62, "learning_rate": 2.0310017996400723e-06, "loss": 0.0002, "step": 13511 }, { "epoch": 1.62, "learning_rate": 2.0309298140371926e-06, "loss": 0.0013, "step": 13512 }, { "epoch": 1.62, "learning_rate": 2.0308578284343134e-06, "loss": 0.0008, "step": 13513 }, { "epoch": 1.62, "learning_rate": 2.0307858428314338e-06, "loss": 0.0003, "step": 13514 }, { "epoch": 1.62, "learning_rate": 2.030713857228554e-06, "loss": 0.0003, "step": 13515 }, { "epoch": 1.62, "learning_rate": 2.030641871625675e-06, "loss": 0.3325, "step": 13516 }, { "epoch": 1.62, "learning_rate": 2.0305698860227953e-06, "loss": 0.0004, "step": 13517 }, { "epoch": 1.62, "learning_rate": 2.030497900419916e-06, "loss": 0.0002, "step": 13518 }, { "epoch": 1.62, "learning_rate": 2.0304259148170364e-06, "loss": 0.0002, "step": 13519 }, { "epoch": 1.62, "learning_rate": 2.030353929214157e-06, "loss": 0.0004, "step": 13520 }, { "epoch": 1.62, "learning_rate": 2.0302819436112775e-06, "loss": 0.0005, "step": 13521 }, { "epoch": 1.62, "learning_rate": 2.0302099580083983e-06, "loss": 0.0003, "step": 13522 }, { "epoch": 1.62, "learning_rate": 2.030137972405519e-06, "loss": 0.0002, "step": 13523 }, { "epoch": 1.62, "learning_rate": 2.03006598680264e-06, "loss": 0.0004, "step": 13524 }, { "epoch": 1.62, "learning_rate": 2.0299940011997602e-06, "loss": 0.0002, "step": 13525 }, { "epoch": 1.62, "learning_rate": 2.0299220155968806e-06, "loss": 0.0004, "step": 13526 }, { "epoch": 1.62, "learning_rate": 2.0298500299940014e-06, "loss": 0.0002, "step": 13527 }, { "epoch": 1.62, "learning_rate": 2.0297780443911217e-06, "loss": 0.0003, "step": 13528 }, { "epoch": 1.62, "learning_rate": 2.0297060587882425e-06, "loss": 0.0002, "step": 13529 }, { "epoch": 1.62, "learning_rate": 2.029634073185363e-06, "loss": 0.0003, "step": 13530 }, { "epoch": 1.62, "learning_rate": 2.0295620875824836e-06, "loss": 0.0001, "step": 13531 }, { "epoch": 1.62, "learning_rate": 2.029490101979604e-06, "loss": 0.9571, "step": 13532 }, { "epoch": 1.62, "learning_rate": 2.0294181163767248e-06, "loss": 0.0003, "step": 13533 }, { "epoch": 1.62, "learning_rate": 2.029346130773845e-06, "loss": 0.0006, "step": 13534 }, { "epoch": 1.62, "learning_rate": 2.029274145170966e-06, "loss": 0.0019, "step": 13535 }, { "epoch": 1.62, "learning_rate": 2.0292021595680862e-06, "loss": 0.0002, "step": 13536 }, { "epoch": 1.62, "learning_rate": 2.029130173965207e-06, "loss": 0.5287, "step": 13537 }, { "epoch": 1.62, "learning_rate": 2.029058188362328e-06, "loss": 0.0003, "step": 13538 }, { "epoch": 1.62, "learning_rate": 2.028986202759448e-06, "loss": 0.0002, "step": 13539 }, { "epoch": 1.62, "learning_rate": 2.028914217156569e-06, "loss": 0.0003, "step": 13540 }, { "epoch": 1.62, "learning_rate": 2.0288422315536893e-06, "loss": 0.0002, "step": 13541 }, { "epoch": 1.62, "learning_rate": 2.02877024595081e-06, "loss": 0.0003, "step": 13542 }, { "epoch": 1.62, "learning_rate": 2.0286982603479304e-06, "loss": 0.0002, "step": 13543 }, { "epoch": 1.62, "learning_rate": 2.028626274745051e-06, "loss": 0.0004, "step": 13544 }, { "epoch": 1.62, "learning_rate": 2.0285542891421716e-06, "loss": 0.0002, "step": 13545 }, { "epoch": 1.62, "learning_rate": 2.0284823035392923e-06, "loss": 0.0009, "step": 13546 }, { "epoch": 1.62, "learning_rate": 2.0284103179364127e-06, "loss": 0.0003, "step": 13547 }, { "epoch": 1.62, "learning_rate": 2.0283383323335335e-06, "loss": 0.0003, "step": 13548 }, { "epoch": 1.62, "learning_rate": 2.028266346730654e-06, "loss": 0.0002, "step": 13549 }, { "epoch": 1.62, "learning_rate": 2.028194361127774e-06, "loss": 0.0002, "step": 13550 }, { "epoch": 1.62, "learning_rate": 2.028122375524895e-06, "loss": 0.0003, "step": 13551 }, { "epoch": 1.62, "learning_rate": 2.0280503899220157e-06, "loss": 0.0004, "step": 13552 }, { "epoch": 1.62, "learning_rate": 2.027978404319136e-06, "loss": 0.764, "step": 13553 }, { "epoch": 1.62, "learning_rate": 2.027906418716257e-06, "loss": 0.0003, "step": 13554 }, { "epoch": 1.62, "learning_rate": 2.0278344331133776e-06, "loss": 0.0004, "step": 13555 }, { "epoch": 1.62, "learning_rate": 2.027762447510498e-06, "loss": 0.0004, "step": 13556 }, { "epoch": 1.62, "learning_rate": 2.0276904619076188e-06, "loss": 0.7398, "step": 13557 }, { "epoch": 1.62, "learning_rate": 2.027618476304739e-06, "loss": 1.0042, "step": 13558 }, { "epoch": 1.62, "learning_rate": 2.02754649070186e-06, "loss": 0.0006, "step": 13559 }, { "epoch": 1.62, "learning_rate": 2.0274745050989803e-06, "loss": 0.0006, "step": 13560 }, { "epoch": 1.63, "learning_rate": 2.0274025194961006e-06, "loss": 0.0005, "step": 13561 }, { "epoch": 1.63, "learning_rate": 2.0273305338932214e-06, "loss": 0.0003, "step": 13562 }, { "epoch": 1.63, "learning_rate": 2.0272585482903418e-06, "loss": 0.3934, "step": 13563 }, { "epoch": 1.63, "learning_rate": 2.0271865626874625e-06, "loss": 0.0005, "step": 13564 }, { "epoch": 1.63, "learning_rate": 2.027114577084583e-06, "loss": 0.3554, "step": 13565 }, { "epoch": 1.63, "learning_rate": 2.0270425914817037e-06, "loss": 0.0018, "step": 13566 }, { "epoch": 1.63, "learning_rate": 2.026970605878824e-06, "loss": 0.0006, "step": 13567 }, { "epoch": 1.63, "learning_rate": 2.026898620275945e-06, "loss": 0.2541, "step": 13568 }, { "epoch": 1.63, "learning_rate": 2.0268266346730656e-06, "loss": 0.0007, "step": 13569 }, { "epoch": 1.63, "learning_rate": 2.0267546490701864e-06, "loss": 0.0004, "step": 13570 }, { "epoch": 1.63, "learning_rate": 2.0266826634673067e-06, "loss": 0.0021, "step": 13571 }, { "epoch": 1.63, "learning_rate": 2.0266106778644275e-06, "loss": 0.0006, "step": 13572 }, { "epoch": 1.63, "learning_rate": 2.026538692261548e-06, "loss": 0.0004, "step": 13573 }, { "epoch": 1.63, "learning_rate": 2.026466706658668e-06, "loss": 0.0004, "step": 13574 }, { "epoch": 1.63, "learning_rate": 2.026394721055789e-06, "loss": 0.0016, "step": 13575 }, { "epoch": 1.63, "learning_rate": 2.0263227354529093e-06, "loss": 0.0002, "step": 13576 }, { "epoch": 1.63, "learning_rate": 2.02625074985003e-06, "loss": 0.0002, "step": 13577 }, { "epoch": 1.63, "learning_rate": 2.0261787642471505e-06, "loss": 0.0003, "step": 13578 }, { "epoch": 1.63, "learning_rate": 2.0261067786442712e-06, "loss": 0.0002, "step": 13579 }, { "epoch": 1.63, "learning_rate": 2.0260347930413916e-06, "loss": 0.0005, "step": 13580 }, { "epoch": 1.63, "learning_rate": 2.0259628074385124e-06, "loss": 0.0003, "step": 13581 }, { "epoch": 1.63, "learning_rate": 2.0258908218356327e-06, "loss": 0.0005, "step": 13582 }, { "epoch": 1.63, "learning_rate": 2.0258188362327535e-06, "loss": 0.0024, "step": 13583 }, { "epoch": 1.63, "learning_rate": 2.0257468506298743e-06, "loss": 0.0005, "step": 13584 }, { "epoch": 1.63, "learning_rate": 2.0256748650269946e-06, "loss": 0.0002, "step": 13585 }, { "epoch": 1.63, "learning_rate": 2.0256028794241154e-06, "loss": 0.0004, "step": 13586 }, { "epoch": 1.63, "learning_rate": 2.0255308938212358e-06, "loss": 0.0003, "step": 13587 }, { "epoch": 1.63, "learning_rate": 2.0254589082183566e-06, "loss": 0.0003, "step": 13588 }, { "epoch": 1.63, "learning_rate": 2.025386922615477e-06, "loss": 0.0002, "step": 13589 }, { "epoch": 1.63, "learning_rate": 2.0253149370125977e-06, "loss": 0.0005, "step": 13590 }, { "epoch": 1.63, "learning_rate": 2.025242951409718e-06, "loss": 0.0006, "step": 13591 }, { "epoch": 1.63, "learning_rate": 2.025170965806839e-06, "loss": 0.0002, "step": 13592 }, { "epoch": 1.63, "learning_rate": 2.025098980203959e-06, "loss": 0.0002, "step": 13593 }, { "epoch": 1.63, "learning_rate": 2.02502699460108e-06, "loss": 0.0003, "step": 13594 }, { "epoch": 1.63, "learning_rate": 2.0249550089982003e-06, "loss": 0.0004, "step": 13595 }, { "epoch": 1.63, "learning_rate": 2.0248830233953207e-06, "loss": 0.0004, "step": 13596 }, { "epoch": 1.63, "learning_rate": 2.0248110377924414e-06, "loss": 0.0007, "step": 13597 }, { "epoch": 1.63, "learning_rate": 2.024739052189562e-06, "loss": 0.0003, "step": 13598 }, { "epoch": 1.63, "learning_rate": 2.0246670665866826e-06, "loss": 0.0004, "step": 13599 }, { "epoch": 1.63, "learning_rate": 2.0245950809838034e-06, "loss": 0.0005, "step": 13600 }, { "epoch": 1.63, "learning_rate": 2.024523095380924e-06, "loss": 0.0002, "step": 13601 }, { "epoch": 1.63, "learning_rate": 2.0244511097780445e-06, "loss": 0.0087, "step": 13602 }, { "epoch": 1.63, "learning_rate": 2.0243791241751653e-06, "loss": 0.0004, "step": 13603 }, { "epoch": 1.63, "learning_rate": 2.0243071385722856e-06, "loss": 0.0003, "step": 13604 }, { "epoch": 1.63, "learning_rate": 2.0242351529694064e-06, "loss": 0.0003, "step": 13605 }, { "epoch": 1.63, "learning_rate": 2.0241631673665268e-06, "loss": 0.0003, "step": 13606 }, { "epoch": 1.63, "learning_rate": 2.024091181763647e-06, "loss": 0.0003, "step": 13607 }, { "epoch": 1.63, "learning_rate": 2.024019196160768e-06, "loss": 0.0003, "step": 13608 }, { "epoch": 1.63, "learning_rate": 2.0239472105578883e-06, "loss": 0.0001, "step": 13609 }, { "epoch": 1.63, "learning_rate": 2.023875224955009e-06, "loss": 0.0007, "step": 13610 }, { "epoch": 1.63, "learning_rate": 2.0238032393521294e-06, "loss": 0.259, "step": 13611 }, { "epoch": 1.63, "learning_rate": 2.02373125374925e-06, "loss": 0.0003, "step": 13612 }, { "epoch": 1.63, "learning_rate": 2.0236592681463705e-06, "loss": 0.0006, "step": 13613 }, { "epoch": 1.63, "learning_rate": 2.0235872825434913e-06, "loss": 0.0003, "step": 13614 }, { "epoch": 1.63, "learning_rate": 2.023515296940612e-06, "loss": 0.0002, "step": 13615 }, { "epoch": 1.63, "learning_rate": 2.0234433113377324e-06, "loss": 0.0004, "step": 13616 }, { "epoch": 1.63, "learning_rate": 2.023371325734853e-06, "loss": 0.0003, "step": 13617 }, { "epoch": 1.63, "learning_rate": 2.023299340131974e-06, "loss": 0.0003, "step": 13618 }, { "epoch": 1.63, "learning_rate": 2.0232273545290943e-06, "loss": 0.0004, "step": 13619 }, { "epoch": 1.63, "learning_rate": 2.0231553689262147e-06, "loss": 0.0004, "step": 13620 }, { "epoch": 1.63, "learning_rate": 2.0230833833233355e-06, "loss": 0.0002, "step": 13621 }, { "epoch": 1.63, "learning_rate": 2.023011397720456e-06, "loss": 0.0002, "step": 13622 }, { "epoch": 1.63, "learning_rate": 2.0229394121175766e-06, "loss": 0.0003, "step": 13623 }, { "epoch": 1.63, "learning_rate": 2.022867426514697e-06, "loss": 0.0004, "step": 13624 }, { "epoch": 1.63, "learning_rate": 2.0227954409118177e-06, "loss": 0.0003, "step": 13625 }, { "epoch": 1.63, "learning_rate": 2.022723455308938e-06, "loss": 0.0005, "step": 13626 }, { "epoch": 1.63, "learning_rate": 2.022651469706059e-06, "loss": 0.0002, "step": 13627 }, { "epoch": 1.63, "learning_rate": 2.0225794841031792e-06, "loss": 0.0003, "step": 13628 }, { "epoch": 1.63, "learning_rate": 2.0225074985003e-06, "loss": 0.0002, "step": 13629 }, { "epoch": 1.63, "learning_rate": 2.0224355128974204e-06, "loss": 0.0002, "step": 13630 }, { "epoch": 1.63, "learning_rate": 2.022363527294541e-06, "loss": 0.9401, "step": 13631 }, { "epoch": 1.63, "learning_rate": 2.022291541691662e-06, "loss": 0.0003, "step": 13632 }, { "epoch": 1.63, "learning_rate": 2.0222195560887823e-06, "loss": 0.0006, "step": 13633 }, { "epoch": 1.63, "learning_rate": 2.022147570485903e-06, "loss": 0.0005, "step": 13634 }, { "epoch": 1.63, "learning_rate": 2.0220755848830234e-06, "loss": 0.001, "step": 13635 }, { "epoch": 1.63, "learning_rate": 2.022003599280144e-06, "loss": 0.0002, "step": 13636 }, { "epoch": 1.63, "learning_rate": 2.0219316136772645e-06, "loss": 0.0003, "step": 13637 }, { "epoch": 1.63, "learning_rate": 2.0218596280743853e-06, "loss": 0.0004, "step": 13638 }, { "epoch": 1.63, "learning_rate": 2.0217876424715057e-06, "loss": 0.0005, "step": 13639 }, { "epoch": 1.63, "learning_rate": 2.0217156568686265e-06, "loss": 0.0028, "step": 13640 }, { "epoch": 1.63, "learning_rate": 2.021643671265747e-06, "loss": 0.0004, "step": 13641 }, { "epoch": 1.63, "learning_rate": 2.021571685662867e-06, "loss": 0.0007, "step": 13642 }, { "epoch": 1.63, "learning_rate": 2.021499700059988e-06, "loss": 0.9084, "step": 13643 }, { "epoch": 1.63, "learning_rate": 2.0214277144571083e-06, "loss": 0.0001, "step": 13644 }, { "epoch": 1.64, "learning_rate": 2.021355728854229e-06, "loss": 0.0004, "step": 13645 }, { "epoch": 1.64, "learning_rate": 2.02128374325135e-06, "loss": 0.0104, "step": 13646 }, { "epoch": 1.64, "learning_rate": 2.0212117576484706e-06, "loss": 0.0003, "step": 13647 }, { "epoch": 1.64, "learning_rate": 2.021139772045591e-06, "loss": 0.0005, "step": 13648 }, { "epoch": 1.64, "learning_rate": 2.0210677864427118e-06, "loss": 0.0009, "step": 13649 }, { "epoch": 1.64, "learning_rate": 2.020995800839832e-06, "loss": 0.0004, "step": 13650 }, { "epoch": 1.64, "learning_rate": 2.020923815236953e-06, "loss": 0.0004, "step": 13651 }, { "epoch": 1.64, "learning_rate": 2.0208518296340733e-06, "loss": 0.0003, "step": 13652 }, { "epoch": 1.64, "learning_rate": 2.0207798440311936e-06, "loss": 0.0004, "step": 13653 }, { "epoch": 1.64, "learning_rate": 2.0207078584283144e-06, "loss": 0.0003, "step": 13654 }, { "epoch": 1.64, "learning_rate": 2.0206358728254347e-06, "loss": 0.0002, "step": 13655 }, { "epoch": 1.64, "learning_rate": 2.0205638872225555e-06, "loss": 0.0004, "step": 13656 }, { "epoch": 1.64, "learning_rate": 2.020491901619676e-06, "loss": 0.0003, "step": 13657 }, { "epoch": 1.64, "learning_rate": 2.0204199160167967e-06, "loss": 0.0004, "step": 13658 }, { "epoch": 1.64, "learning_rate": 2.020347930413917e-06, "loss": 0.6409, "step": 13659 }, { "epoch": 1.64, "learning_rate": 2.020275944811038e-06, "loss": 0.0004, "step": 13660 }, { "epoch": 1.64, "learning_rate": 2.0202039592081586e-06, "loss": 0.0002, "step": 13661 }, { "epoch": 1.64, "learning_rate": 2.020131973605279e-06, "loss": 0.0002, "step": 13662 }, { "epoch": 1.64, "learning_rate": 2.0200599880023997e-06, "loss": 0.0006, "step": 13663 }, { "epoch": 1.64, "learning_rate": 2.0199880023995205e-06, "loss": 0.0004, "step": 13664 }, { "epoch": 1.64, "learning_rate": 2.019916016796641e-06, "loss": 0.0003, "step": 13665 }, { "epoch": 1.64, "learning_rate": 2.019844031193761e-06, "loss": 0.0003, "step": 13666 }, { "epoch": 1.64, "learning_rate": 2.019772045590882e-06, "loss": 0.0003, "step": 13667 }, { "epoch": 1.64, "learning_rate": 2.0197000599880023e-06, "loss": 0.0003, "step": 13668 }, { "epoch": 1.64, "learning_rate": 2.019628074385123e-06, "loss": 0.0003, "step": 13669 }, { "epoch": 1.64, "learning_rate": 2.0195560887822435e-06, "loss": 0.0009, "step": 13670 }, { "epoch": 1.64, "learning_rate": 2.0194841031793642e-06, "loss": 0.044, "step": 13671 }, { "epoch": 1.64, "learning_rate": 2.0194121175764846e-06, "loss": 0.0002, "step": 13672 }, { "epoch": 1.64, "learning_rate": 2.0193401319736054e-06, "loss": 0.0004, "step": 13673 }, { "epoch": 1.64, "learning_rate": 2.0192681463707257e-06, "loss": 0.0002, "step": 13674 }, { "epoch": 1.64, "learning_rate": 2.0191961607678465e-06, "loss": 0.0005, "step": 13675 }, { "epoch": 1.64, "learning_rate": 2.019124175164967e-06, "loss": 0.0004, "step": 13676 }, { "epoch": 1.64, "learning_rate": 2.0190521895620876e-06, "loss": 0.0004, "step": 13677 }, { "epoch": 1.64, "learning_rate": 2.0189802039592084e-06, "loss": 0.0002, "step": 13678 }, { "epoch": 1.64, "learning_rate": 2.0189082183563288e-06, "loss": 0.0005, "step": 13679 }, { "epoch": 1.64, "learning_rate": 2.0188362327534496e-06, "loss": 0.0003, "step": 13680 }, { "epoch": 1.64, "learning_rate": 2.01876424715057e-06, "loss": 0.0004, "step": 13681 }, { "epoch": 1.64, "learning_rate": 2.0186922615476907e-06, "loss": 0.0002, "step": 13682 }, { "epoch": 1.64, "learning_rate": 2.018620275944811e-06, "loss": 0.0004, "step": 13683 }, { "epoch": 1.64, "learning_rate": 2.018548290341932e-06, "loss": 1.6881, "step": 13684 }, { "epoch": 1.64, "learning_rate": 2.018476304739052e-06, "loss": 0.7402, "step": 13685 }, { "epoch": 1.64, "learning_rate": 2.018404319136173e-06, "loss": 0.001, "step": 13686 }, { "epoch": 1.64, "learning_rate": 2.0183323335332933e-06, "loss": 0.8325, "step": 13687 }, { "epoch": 1.64, "learning_rate": 2.0182603479304137e-06, "loss": 0.0003, "step": 13688 }, { "epoch": 1.64, "learning_rate": 2.0181883623275344e-06, "loss": 0.0003, "step": 13689 }, { "epoch": 1.64, "learning_rate": 2.018116376724655e-06, "loss": 0.0003, "step": 13690 }, { "epoch": 1.64, "learning_rate": 2.0180443911217756e-06, "loss": 0.0004, "step": 13691 }, { "epoch": 1.64, "learning_rate": 2.0179724055188964e-06, "loss": 0.0003, "step": 13692 }, { "epoch": 1.64, "learning_rate": 2.017900419916017e-06, "loss": 0.3739, "step": 13693 }, { "epoch": 1.64, "learning_rate": 2.0178284343131375e-06, "loss": 0.0007, "step": 13694 }, { "epoch": 1.64, "learning_rate": 2.0177564487102583e-06, "loss": 0.0004, "step": 13695 }, { "epoch": 1.64, "learning_rate": 2.0176844631073786e-06, "loss": 0.0003, "step": 13696 }, { "epoch": 1.64, "learning_rate": 2.0176124775044994e-06, "loss": 0.0006, "step": 13697 }, { "epoch": 1.64, "learning_rate": 2.0175404919016198e-06, "loss": 0.0004, "step": 13698 }, { "epoch": 1.64, "learning_rate": 2.0174685062987405e-06, "loss": 0.0001, "step": 13699 }, { "epoch": 1.64, "learning_rate": 2.017396520695861e-06, "loss": 0.0005, "step": 13700 }, { "epoch": 1.64, "learning_rate": 2.0173245350929812e-06, "loss": 0.0009, "step": 13701 }, { "epoch": 1.64, "learning_rate": 2.017252549490102e-06, "loss": 0.0004, "step": 13702 }, { "epoch": 1.64, "learning_rate": 2.0171805638872224e-06, "loss": 0.0004, "step": 13703 }, { "epoch": 1.64, "learning_rate": 2.017108578284343e-06, "loss": 0.0002, "step": 13704 }, { "epoch": 1.64, "learning_rate": 2.0170365926814635e-06, "loss": 0.0003, "step": 13705 }, { "epoch": 1.64, "learning_rate": 2.0169646070785843e-06, "loss": 0.0003, "step": 13706 }, { "epoch": 1.64, "learning_rate": 2.016892621475705e-06, "loss": 0.0002, "step": 13707 }, { "epoch": 1.64, "learning_rate": 2.0168206358728254e-06, "loss": 0.8045, "step": 13708 }, { "epoch": 1.64, "learning_rate": 2.016748650269946e-06, "loss": 0.0003, "step": 13709 }, { "epoch": 1.64, "learning_rate": 2.016676664667067e-06, "loss": 0.8552, "step": 13710 }, { "epoch": 1.64, "learning_rate": 2.0166046790641873e-06, "loss": 0.0003, "step": 13711 }, { "epoch": 1.64, "learning_rate": 2.0165326934613077e-06, "loss": 0.0005, "step": 13712 }, { "epoch": 1.64, "learning_rate": 2.0164607078584285e-06, "loss": 0.0004, "step": 13713 }, { "epoch": 1.64, "learning_rate": 2.016388722255549e-06, "loss": 0.0002, "step": 13714 }, { "epoch": 1.64, "learning_rate": 2.0163167366526696e-06, "loss": 0.0004, "step": 13715 }, { "epoch": 1.64, "learning_rate": 2.01624475104979e-06, "loss": 0.3344, "step": 13716 }, { "epoch": 1.64, "learning_rate": 2.0161727654469107e-06, "loss": 0.0007, "step": 13717 }, { "epoch": 1.64, "learning_rate": 2.016100779844031e-06, "loss": 0.0009, "step": 13718 }, { "epoch": 1.64, "learning_rate": 2.016028794241152e-06, "loss": 0.0009, "step": 13719 }, { "epoch": 1.64, "learning_rate": 2.0159568086382722e-06, "loss": 0.9484, "step": 13720 }, { "epoch": 1.64, "learning_rate": 2.015884823035393e-06, "loss": 0.0002, "step": 13721 }, { "epoch": 1.64, "learning_rate": 2.0158128374325134e-06, "loss": 0.0004, "step": 13722 }, { "epoch": 1.64, "learning_rate": 2.015740851829634e-06, "loss": 0.5564, "step": 13723 }, { "epoch": 1.64, "learning_rate": 2.015668866226755e-06, "loss": 0.0006, "step": 13724 }, { "epoch": 1.64, "learning_rate": 2.0155968806238753e-06, "loss": 0.001, "step": 13725 }, { "epoch": 1.64, "learning_rate": 2.015524895020996e-06, "loss": 0.0009, "step": 13726 }, { "epoch": 1.64, "learning_rate": 2.0154529094181164e-06, "loss": 0.0003, "step": 13727 }, { "epoch": 1.65, "learning_rate": 2.015380923815237e-06, "loss": 0.0004, "step": 13728 }, { "epoch": 1.65, "learning_rate": 2.0153089382123575e-06, "loss": 0.0004, "step": 13729 }, { "epoch": 1.65, "learning_rate": 2.0152369526094783e-06, "loss": 0.0005, "step": 13730 }, { "epoch": 1.65, "learning_rate": 2.0151649670065987e-06, "loss": 0.0003, "step": 13731 }, { "epoch": 1.65, "learning_rate": 2.0150929814037194e-06, "loss": 0.0008, "step": 13732 }, { "epoch": 1.65, "learning_rate": 2.01502099580084e-06, "loss": 0.0003, "step": 13733 }, { "epoch": 1.65, "learning_rate": 2.01494901019796e-06, "loss": 0.028, "step": 13734 }, { "epoch": 1.65, "learning_rate": 2.014877024595081e-06, "loss": 0.0009, "step": 13735 }, { "epoch": 1.65, "learning_rate": 2.0148050389922013e-06, "loss": 0.0004, "step": 13736 }, { "epoch": 1.65, "learning_rate": 2.014733053389322e-06, "loss": 0.0002, "step": 13737 }, { "epoch": 1.65, "learning_rate": 2.014661067786443e-06, "loss": 0.0008, "step": 13738 }, { "epoch": 1.65, "learning_rate": 2.014589082183563e-06, "loss": 0.0005, "step": 13739 }, { "epoch": 1.65, "learning_rate": 2.014517096580684e-06, "loss": 0.001, "step": 13740 }, { "epoch": 1.65, "learning_rate": 2.0144451109778048e-06, "loss": 0.0003, "step": 13741 }, { "epoch": 1.65, "learning_rate": 2.014373125374925e-06, "loss": 0.0007, "step": 13742 }, { "epoch": 1.65, "learning_rate": 2.014301139772046e-06, "loss": 0.0005, "step": 13743 }, { "epoch": 1.65, "learning_rate": 2.0142291541691662e-06, "loss": 0.0005, "step": 13744 }, { "epoch": 1.65, "learning_rate": 2.014157168566287e-06, "loss": 0.0001, "step": 13745 }, { "epoch": 1.65, "learning_rate": 2.0140851829634074e-06, "loss": 0.7579, "step": 13746 }, { "epoch": 1.65, "learning_rate": 2.0140131973605277e-06, "loss": 0.0094, "step": 13747 }, { "epoch": 1.65, "learning_rate": 2.0139412117576485e-06, "loss": 0.0004, "step": 13748 }, { "epoch": 1.65, "learning_rate": 2.013869226154769e-06, "loss": 0.0003, "step": 13749 }, { "epoch": 1.65, "learning_rate": 2.0137972405518897e-06, "loss": 0.0008, "step": 13750 }, { "epoch": 1.65, "learning_rate": 2.01372525494901e-06, "loss": 0.0003, "step": 13751 }, { "epoch": 1.65, "learning_rate": 2.0136532693461308e-06, "loss": 0.0004, "step": 13752 }, { "epoch": 1.65, "learning_rate": 2.013581283743251e-06, "loss": 0.2641, "step": 13753 }, { "epoch": 1.65, "learning_rate": 2.013509298140372e-06, "loss": 0.004, "step": 13754 }, { "epoch": 1.65, "learning_rate": 2.0134373125374927e-06, "loss": 0.0005, "step": 13755 }, { "epoch": 1.65, "learning_rate": 2.0133653269346135e-06, "loss": 0.0006, "step": 13756 }, { "epoch": 1.65, "learning_rate": 2.013293341331734e-06, "loss": 0.0005, "step": 13757 }, { "epoch": 1.65, "learning_rate": 2.013221355728854e-06, "loss": 0.0005, "step": 13758 }, { "epoch": 1.65, "learning_rate": 2.013149370125975e-06, "loss": 0.0018, "step": 13759 }, { "epoch": 1.65, "learning_rate": 2.0130773845230953e-06, "loss": 0.0005, "step": 13760 }, { "epoch": 1.65, "learning_rate": 2.013005398920216e-06, "loss": 0.0004, "step": 13761 }, { "epoch": 1.65, "learning_rate": 2.0129334133173365e-06, "loss": 0.0007, "step": 13762 }, { "epoch": 1.65, "learning_rate": 2.0128614277144572e-06, "loss": 0.0005, "step": 13763 }, { "epoch": 1.65, "learning_rate": 2.0127894421115776e-06, "loss": 0.0003, "step": 13764 }, { "epoch": 1.65, "learning_rate": 2.0127174565086984e-06, "loss": 0.0003, "step": 13765 }, { "epoch": 1.65, "learning_rate": 2.0126454709058187e-06, "loss": 0.0003, "step": 13766 }, { "epoch": 1.65, "learning_rate": 2.0125734853029395e-06, "loss": 0.0004, "step": 13767 }, { "epoch": 1.65, "learning_rate": 2.01250149970006e-06, "loss": 0.0105, "step": 13768 }, { "epoch": 1.65, "learning_rate": 2.0124295140971806e-06, "loss": 0.0003, "step": 13769 }, { "epoch": 1.65, "learning_rate": 2.0123575284943014e-06, "loss": 0.0002, "step": 13770 }, { "epoch": 1.65, "learning_rate": 2.0122855428914218e-06, "loss": 0.0006, "step": 13771 }, { "epoch": 1.65, "learning_rate": 2.0122135572885425e-06, "loss": 0.0004, "step": 13772 }, { "epoch": 1.65, "learning_rate": 2.012141571685663e-06, "loss": 0.0004, "step": 13773 }, { "epoch": 1.65, "learning_rate": 2.0120695860827837e-06, "loss": 0.0006, "step": 13774 }, { "epoch": 1.65, "learning_rate": 2.011997600479904e-06, "loss": 0.0007, "step": 13775 }, { "epoch": 1.65, "learning_rate": 2.011925614877025e-06, "loss": 0.0005, "step": 13776 }, { "epoch": 1.65, "learning_rate": 2.011853629274145e-06, "loss": 0.0003, "step": 13777 }, { "epoch": 1.65, "learning_rate": 2.011781643671266e-06, "loss": 0.0005, "step": 13778 }, { "epoch": 1.65, "learning_rate": 2.0117096580683863e-06, "loss": 0.0005, "step": 13779 }, { "epoch": 1.65, "learning_rate": 2.011637672465507e-06, "loss": 0.0006, "step": 13780 }, { "epoch": 1.65, "learning_rate": 2.0115656868626274e-06, "loss": 0.0005, "step": 13781 }, { "epoch": 1.65, "learning_rate": 2.0114937012597478e-06, "loss": 0.0004, "step": 13782 }, { "epoch": 1.65, "learning_rate": 2.0114217156568686e-06, "loss": 0.0002, "step": 13783 }, { "epoch": 1.65, "learning_rate": 2.0113497300539893e-06, "loss": 0.0004, "step": 13784 }, { "epoch": 1.65, "learning_rate": 2.0112777444511097e-06, "loss": 0.0004, "step": 13785 }, { "epoch": 1.65, "learning_rate": 2.0112057588482305e-06, "loss": 0.0004, "step": 13786 }, { "epoch": 1.65, "learning_rate": 2.0111337732453513e-06, "loss": 0.0003, "step": 13787 }, { "epoch": 1.65, "learning_rate": 2.0110617876424716e-06, "loss": 0.0003, "step": 13788 }, { "epoch": 1.65, "learning_rate": 2.0109898020395924e-06, "loss": 0.0006, "step": 13789 }, { "epoch": 1.65, "learning_rate": 2.0109178164367127e-06, "loss": 0.0002, "step": 13790 }, { "epoch": 1.65, "learning_rate": 2.0108458308338335e-06, "loss": 0.0117, "step": 13791 }, { "epoch": 1.65, "learning_rate": 2.010773845230954e-06, "loss": 0.0006, "step": 13792 }, { "epoch": 1.65, "learning_rate": 2.0107018596280742e-06, "loss": 0.0004, "step": 13793 }, { "epoch": 1.65, "learning_rate": 2.010629874025195e-06, "loss": 0.0005, "step": 13794 }, { "epoch": 1.65, "learning_rate": 2.0105578884223154e-06, "loss": 0.0002, "step": 13795 }, { "epoch": 1.65, "learning_rate": 2.010485902819436e-06, "loss": 0.0006, "step": 13796 }, { "epoch": 1.65, "learning_rate": 2.0104139172165565e-06, "loss": 0.0002, "step": 13797 }, { "epoch": 1.65, "learning_rate": 2.0103419316136773e-06, "loss": 0.0003, "step": 13798 }, { "epoch": 1.65, "learning_rate": 2.0102699460107976e-06, "loss": 0.0002, "step": 13799 }, { "epoch": 1.65, "learning_rate": 2.0101979604079184e-06, "loss": 0.0004, "step": 13800 }, { "epoch": 1.65, "learning_rate": 2.010125974805039e-06, "loss": 0.0006, "step": 13801 }, { "epoch": 1.65, "learning_rate": 2.01005398920216e-06, "loss": 0.0005, "step": 13802 }, { "epoch": 1.65, "learning_rate": 2.0099820035992803e-06, "loss": 0.0002, "step": 13803 }, { "epoch": 1.65, "learning_rate": 2.0099100179964007e-06, "loss": 0.0003, "step": 13804 }, { "epoch": 1.65, "learning_rate": 2.0098380323935215e-06, "loss": 0.0006, "step": 13805 }, { "epoch": 1.65, "learning_rate": 2.009766046790642e-06, "loss": 0.0013, "step": 13806 }, { "epoch": 1.65, "learning_rate": 2.0096940611877626e-06, "loss": 0.0006, "step": 13807 }, { "epoch": 1.65, "learning_rate": 2.009622075584883e-06, "loss": 0.0005, "step": 13808 }, { "epoch": 1.65, "learning_rate": 2.0095500899820037e-06, "loss": 0.0003, "step": 13809 }, { "epoch": 1.65, "learning_rate": 2.009478104379124e-06, "loss": 0.0003, "step": 13810 }, { "epoch": 1.66, "learning_rate": 2.009406118776245e-06, "loss": 0.0006, "step": 13811 }, { "epoch": 1.66, "learning_rate": 2.0093341331733652e-06, "loss": 0.0001, "step": 13812 }, { "epoch": 1.66, "learning_rate": 2.009262147570486e-06, "loss": 0.0007, "step": 13813 }, { "epoch": 1.66, "learning_rate": 2.0091901619676063e-06, "loss": 0.0003, "step": 13814 }, { "epoch": 1.66, "learning_rate": 2.009118176364727e-06, "loss": 0.0003, "step": 13815 }, { "epoch": 1.66, "learning_rate": 2.009046190761848e-06, "loss": 0.0002, "step": 13816 }, { "epoch": 1.66, "learning_rate": 2.0089742051589683e-06, "loss": 0.0014, "step": 13817 }, { "epoch": 1.66, "learning_rate": 2.008902219556089e-06, "loss": 0.0003, "step": 13818 }, { "epoch": 1.66, "learning_rate": 2.0088302339532094e-06, "loss": 0.0004, "step": 13819 }, { "epoch": 1.66, "learning_rate": 2.00875824835033e-06, "loss": 0.0003, "step": 13820 }, { "epoch": 1.66, "learning_rate": 2.0086862627474505e-06, "loss": 0.0002, "step": 13821 }, { "epoch": 1.66, "learning_rate": 2.0086142771445713e-06, "loss": 0.0009, "step": 13822 }, { "epoch": 1.66, "learning_rate": 2.0085422915416917e-06, "loss": 0.0004, "step": 13823 }, { "epoch": 1.66, "learning_rate": 2.0084703059388124e-06, "loss": 0.0005, "step": 13824 }, { "epoch": 1.66, "learning_rate": 2.008398320335933e-06, "loss": 0.0004, "step": 13825 }, { "epoch": 1.66, "learning_rate": 2.0083263347330536e-06, "loss": 0.0002, "step": 13826 }, { "epoch": 1.66, "learning_rate": 2.008254349130174e-06, "loss": 0.0003, "step": 13827 }, { "epoch": 1.66, "learning_rate": 2.0081823635272943e-06, "loss": 0.0003, "step": 13828 }, { "epoch": 1.66, "learning_rate": 2.008110377924415e-06, "loss": 0.0003, "step": 13829 }, { "epoch": 1.66, "learning_rate": 2.008038392321536e-06, "loss": 0.0007, "step": 13830 }, { "epoch": 1.66, "learning_rate": 2.007966406718656e-06, "loss": 0.0003, "step": 13831 }, { "epoch": 1.66, "learning_rate": 2.007894421115777e-06, "loss": 0.0004, "step": 13832 }, { "epoch": 1.66, "learning_rate": 2.0078224355128978e-06, "loss": 0.0003, "step": 13833 }, { "epoch": 1.66, "learning_rate": 2.007750449910018e-06, "loss": 0.0008, "step": 13834 }, { "epoch": 1.66, "learning_rate": 2.007678464307139e-06, "loss": 0.4183, "step": 13835 }, { "epoch": 1.66, "learning_rate": 2.0076064787042592e-06, "loss": 0.5375, "step": 13836 }, { "epoch": 1.66, "learning_rate": 2.00753449310138e-06, "loss": 1.0147, "step": 13837 }, { "epoch": 1.66, "learning_rate": 2.0074625074985004e-06, "loss": 0.0008, "step": 13838 }, { "epoch": 1.66, "learning_rate": 2.0073905218956207e-06, "loss": 0.0004, "step": 13839 }, { "epoch": 1.66, "learning_rate": 2.0073185362927415e-06, "loss": 0.0003, "step": 13840 }, { "epoch": 1.66, "learning_rate": 2.007246550689862e-06, "loss": 0.0003, "step": 13841 }, { "epoch": 1.66, "learning_rate": 2.0071745650869826e-06, "loss": 0.6752, "step": 13842 }, { "epoch": 1.66, "learning_rate": 2.007102579484103e-06, "loss": 0.001, "step": 13843 }, { "epoch": 1.66, "learning_rate": 2.0070305938812238e-06, "loss": 0.0004, "step": 13844 }, { "epoch": 1.66, "learning_rate": 2.006958608278344e-06, "loss": 0.0004, "step": 13845 }, { "epoch": 1.66, "learning_rate": 2.006886622675465e-06, "loss": 0.0002, "step": 13846 }, { "epoch": 1.66, "learning_rate": 2.0068146370725857e-06, "loss": 0.0004, "step": 13847 }, { "epoch": 1.66, "learning_rate": 2.0067426514697065e-06, "loss": 0.0003, "step": 13848 }, { "epoch": 1.66, "learning_rate": 2.006670665866827e-06, "loss": 0.001, "step": 13849 }, { "epoch": 1.66, "learning_rate": 2.006598680263947e-06, "loss": 0.0005, "step": 13850 }, { "epoch": 1.66, "learning_rate": 2.006526694661068e-06, "loss": 0.0012, "step": 13851 }, { "epoch": 1.66, "learning_rate": 2.0064547090581883e-06, "loss": 0.0003, "step": 13852 }, { "epoch": 1.66, "learning_rate": 2.006382723455309e-06, "loss": 0.0005, "step": 13853 }, { "epoch": 1.66, "learning_rate": 2.0063107378524294e-06, "loss": 0.0003, "step": 13854 }, { "epoch": 1.66, "learning_rate": 2.0062387522495502e-06, "loss": 0.0004, "step": 13855 }, { "epoch": 1.66, "learning_rate": 2.0061667666466706e-06, "loss": 0.0055, "step": 13856 }, { "epoch": 1.66, "learning_rate": 2.0060947810437914e-06, "loss": 0.0007, "step": 13857 }, { "epoch": 1.66, "learning_rate": 2.0060227954409117e-06, "loss": 0.0003, "step": 13858 }, { "epoch": 1.66, "learning_rate": 2.0059508098380325e-06, "loss": 0.001, "step": 13859 }, { "epoch": 1.66, "learning_rate": 2.005878824235153e-06, "loss": 0.0002, "step": 13860 }, { "epoch": 1.66, "learning_rate": 2.0058068386322736e-06, "loss": 0.0003, "step": 13861 }, { "epoch": 1.66, "learning_rate": 2.005734853029394e-06, "loss": 0.0008, "step": 13862 }, { "epoch": 1.66, "learning_rate": 2.0056628674265148e-06, "loss": 0.0005, "step": 13863 }, { "epoch": 1.66, "learning_rate": 2.0055908818236355e-06, "loss": 0.0002, "step": 13864 }, { "epoch": 1.66, "learning_rate": 2.005518896220756e-06, "loss": 0.8223, "step": 13865 }, { "epoch": 1.66, "learning_rate": 2.0054469106178767e-06, "loss": 0.0004, "step": 13866 }, { "epoch": 1.66, "learning_rate": 2.005374925014997e-06, "loss": 0.0009, "step": 13867 }, { "epoch": 1.66, "learning_rate": 2.005302939412118e-06, "loss": 0.0005, "step": 13868 }, { "epoch": 1.66, "learning_rate": 2.005230953809238e-06, "loss": 0.0003, "step": 13869 }, { "epoch": 1.66, "learning_rate": 2.005158968206359e-06, "loss": 0.0008, "step": 13870 }, { "epoch": 1.66, "learning_rate": 2.0050869826034793e-06, "loss": 0.0003, "step": 13871 }, { "epoch": 1.66, "learning_rate": 2.0050149970006e-06, "loss": 0.0003, "step": 13872 }, { "epoch": 1.66, "learning_rate": 2.0049430113977204e-06, "loss": 0.0125, "step": 13873 }, { "epoch": 1.66, "learning_rate": 2.0048710257948408e-06, "loss": 0.0003, "step": 13874 }, { "epoch": 1.66, "learning_rate": 2.0047990401919616e-06, "loss": 0.0002, "step": 13875 }, { "epoch": 1.66, "learning_rate": 2.004727054589082e-06, "loss": 0.0001, "step": 13876 }, { "epoch": 1.66, "learning_rate": 2.0046550689862027e-06, "loss": 0.0002, "step": 13877 }, { "epoch": 1.66, "learning_rate": 2.0045830833833235e-06, "loss": 0.0005, "step": 13878 }, { "epoch": 1.66, "learning_rate": 2.0045110977804442e-06, "loss": 0.0004, "step": 13879 }, { "epoch": 1.66, "learning_rate": 2.0044391121775646e-06, "loss": 0.0003, "step": 13880 }, { "epoch": 1.66, "learning_rate": 2.0043671265746854e-06, "loss": 0.4642, "step": 13881 }, { "epoch": 1.66, "learning_rate": 2.0042951409718057e-06, "loss": 0.0014, "step": 13882 }, { "epoch": 1.66, "learning_rate": 2.0042231553689265e-06, "loss": 0.0004, "step": 13883 }, { "epoch": 1.66, "learning_rate": 2.004151169766047e-06, "loss": 0.0002, "step": 13884 }, { "epoch": 1.66, "learning_rate": 2.0040791841631672e-06, "loss": 0.0003, "step": 13885 }, { "epoch": 1.66, "learning_rate": 2.004007198560288e-06, "loss": 0.0003, "step": 13886 }, { "epoch": 1.66, "learning_rate": 2.0039352129574084e-06, "loss": 0.0002, "step": 13887 }, { "epoch": 1.66, "learning_rate": 2.003863227354529e-06, "loss": 0.0009, "step": 13888 }, { "epoch": 1.66, "learning_rate": 2.0037912417516495e-06, "loss": 0.0006, "step": 13889 }, { "epoch": 1.66, "learning_rate": 2.0037192561487703e-06, "loss": 0.0024, "step": 13890 }, { "epoch": 1.66, "learning_rate": 2.0036472705458906e-06, "loss": 0.0104, "step": 13891 }, { "epoch": 1.66, "learning_rate": 2.0035752849430114e-06, "loss": 0.1165, "step": 13892 }, { "epoch": 1.66, "learning_rate": 2.003503299340132e-06, "loss": 0.0003, "step": 13893 }, { "epoch": 1.66, "learning_rate": 2.0034313137372525e-06, "loss": 0.0003, "step": 13894 }, { "epoch": 1.67, "learning_rate": 2.0033593281343733e-06, "loss": 0.0002, "step": 13895 }, { "epoch": 1.67, "learning_rate": 2.0032873425314937e-06, "loss": 0.8168, "step": 13896 }, { "epoch": 1.67, "learning_rate": 2.0032153569286145e-06, "loss": 0.0006, "step": 13897 }, { "epoch": 1.67, "learning_rate": 2.003143371325735e-06, "loss": 0.0002, "step": 13898 }, { "epoch": 1.67, "learning_rate": 2.0030713857228556e-06, "loss": 0.0004, "step": 13899 }, { "epoch": 1.67, "learning_rate": 2.002999400119976e-06, "loss": 0.0004, "step": 13900 }, { "epoch": 1.67, "learning_rate": 2.0029274145170967e-06, "loss": 0.0006, "step": 13901 }, { "epoch": 1.67, "learning_rate": 2.002855428914217e-06, "loss": 0.5822, "step": 13902 }, { "epoch": 1.67, "learning_rate": 2.002783443311338e-06, "loss": 0.0003, "step": 13903 }, { "epoch": 1.67, "learning_rate": 2.002711457708458e-06, "loss": 0.0005, "step": 13904 }, { "epoch": 1.67, "learning_rate": 2.002639472105579e-06, "loss": 0.0003, "step": 13905 }, { "epoch": 1.67, "learning_rate": 2.0025674865026993e-06, "loss": 0.0003, "step": 13906 }, { "epoch": 1.67, "learning_rate": 2.00249550089982e-06, "loss": 0.0003, "step": 13907 }, { "epoch": 1.67, "learning_rate": 2.0024235152969405e-06, "loss": 0.0006, "step": 13908 }, { "epoch": 1.67, "learning_rate": 2.0023515296940613e-06, "loss": 0.0002, "step": 13909 }, { "epoch": 1.67, "learning_rate": 2.002279544091182e-06, "loss": 0.0001, "step": 13910 }, { "epoch": 1.67, "learning_rate": 2.0022075584883024e-06, "loss": 0.0004, "step": 13911 }, { "epoch": 1.67, "learning_rate": 2.002135572885423e-06, "loss": 0.0002, "step": 13912 }, { "epoch": 1.67, "learning_rate": 2.0020635872825435e-06, "loss": 0.0002, "step": 13913 }, { "epoch": 1.67, "learning_rate": 2.0019916016796643e-06, "loss": 0.0003, "step": 13914 }, { "epoch": 1.67, "learning_rate": 2.0019196160767847e-06, "loss": 0.0005, "step": 13915 }, { "epoch": 1.67, "learning_rate": 2.0018476304739054e-06, "loss": 0.0003, "step": 13916 }, { "epoch": 1.67, "learning_rate": 2.0017756448710258e-06, "loss": 0.0002, "step": 13917 }, { "epoch": 1.67, "learning_rate": 2.0017036592681466e-06, "loss": 0.0003, "step": 13918 }, { "epoch": 1.67, "learning_rate": 2.001631673665267e-06, "loss": 0.0004, "step": 13919 }, { "epoch": 1.67, "learning_rate": 2.0015596880623873e-06, "loss": 0.0001, "step": 13920 }, { "epoch": 1.67, "learning_rate": 2.001487702459508e-06, "loss": 0.0004, "step": 13921 }, { "epoch": 1.67, "learning_rate": 2.0014157168566284e-06, "loss": 0.0026, "step": 13922 }, { "epoch": 1.67, "learning_rate": 2.001343731253749e-06, "loss": 0.0003, "step": 13923 }, { "epoch": 1.67, "learning_rate": 2.00127174565087e-06, "loss": 0.0027, "step": 13924 }, { "epoch": 1.67, "learning_rate": 2.0011997600479907e-06, "loss": 0.19, "step": 13925 }, { "epoch": 1.67, "learning_rate": 2.001127774445111e-06, "loss": 0.0002, "step": 13926 }, { "epoch": 1.67, "learning_rate": 2.001055788842232e-06, "loss": 0.0787, "step": 13927 }, { "epoch": 1.67, "learning_rate": 2.0009838032393522e-06, "loss": 0.0005, "step": 13928 }, { "epoch": 1.67, "learning_rate": 2.000911817636473e-06, "loss": 0.0007, "step": 13929 }, { "epoch": 1.67, "learning_rate": 2.0008398320335934e-06, "loss": 0.0004, "step": 13930 }, { "epoch": 1.67, "learning_rate": 2.0007678464307137e-06, "loss": 0.0002, "step": 13931 }, { "epoch": 1.67, "learning_rate": 2.0006958608278345e-06, "loss": 0.0005, "step": 13932 }, { "epoch": 1.67, "learning_rate": 2.000623875224955e-06, "loss": 0.0004, "step": 13933 }, { "epoch": 1.67, "learning_rate": 2.0005518896220756e-06, "loss": 0.0003, "step": 13934 }, { "epoch": 1.67, "learning_rate": 2.000479904019196e-06, "loss": 0.0005, "step": 13935 }, { "epoch": 1.67, "learning_rate": 2.0004079184163168e-06, "loss": 0.0008, "step": 13936 }, { "epoch": 1.67, "learning_rate": 2.000335932813437e-06, "loss": 0.0002, "step": 13937 }, { "epoch": 1.67, "learning_rate": 2.000263947210558e-06, "loss": 0.0003, "step": 13938 }, { "epoch": 1.67, "learning_rate": 2.0001919616076787e-06, "loss": 0.0007, "step": 13939 }, { "epoch": 1.67, "learning_rate": 2.000119976004799e-06, "loss": 0.5781, "step": 13940 }, { "epoch": 1.67, "learning_rate": 2.00004799040192e-06, "loss": 0.0003, "step": 13941 }, { "epoch": 1.67, "learning_rate": 1.99997600479904e-06, "loss": 0.0004, "step": 13942 }, { "epoch": 1.67, "learning_rate": 1.999904019196161e-06, "loss": 0.001, "step": 13943 }, { "epoch": 1.67, "learning_rate": 1.9998320335932813e-06, "loss": 0.0002, "step": 13944 }, { "epoch": 1.67, "learning_rate": 1.999760047990402e-06, "loss": 0.0003, "step": 13945 }, { "epoch": 1.67, "learning_rate": 1.9996880623875224e-06, "loss": 0.0003, "step": 13946 }, { "epoch": 1.67, "learning_rate": 1.9996160767846432e-06, "loss": 0.0005, "step": 13947 }, { "epoch": 1.67, "learning_rate": 1.9995440911817636e-06, "loss": 0.0312, "step": 13948 }, { "epoch": 1.67, "learning_rate": 1.9994721055788843e-06, "loss": 0.0609, "step": 13949 }, { "epoch": 1.67, "learning_rate": 1.9994001199760047e-06, "loss": 0.0003, "step": 13950 }, { "epoch": 1.67, "learning_rate": 1.9993281343731255e-06, "loss": 0.0008, "step": 13951 }, { "epoch": 1.67, "learning_rate": 1.999256148770246e-06, "loss": 0.0002, "step": 13952 }, { "epoch": 1.67, "learning_rate": 1.9991841631673666e-06, "loss": 0.0009, "step": 13953 }, { "epoch": 1.67, "learning_rate": 1.999112177564487e-06, "loss": 0.0006, "step": 13954 }, { "epoch": 1.67, "learning_rate": 1.9990401919616077e-06, "loss": 0.0002, "step": 13955 }, { "epoch": 1.67, "learning_rate": 1.9989682063587285e-06, "loss": 0.0003, "step": 13956 }, { "epoch": 1.67, "learning_rate": 1.998896220755849e-06, "loss": 0.0015, "step": 13957 }, { "epoch": 1.67, "learning_rate": 1.9988242351529697e-06, "loss": 0.0002, "step": 13958 }, { "epoch": 1.67, "learning_rate": 1.99875224955009e-06, "loss": 0.0002, "step": 13959 }, { "epoch": 1.67, "learning_rate": 1.998680263947211e-06, "loss": 0.0009, "step": 13960 }, { "epoch": 1.67, "learning_rate": 1.998608278344331e-06, "loss": 0.0002, "step": 13961 }, { "epoch": 1.67, "learning_rate": 1.998536292741452e-06, "loss": 0.0006, "step": 13962 }, { "epoch": 1.67, "learning_rate": 1.9984643071385723e-06, "loss": 0.0072, "step": 13963 }, { "epoch": 1.67, "learning_rate": 1.998392321535693e-06, "loss": 0.0003, "step": 13964 }, { "epoch": 1.67, "learning_rate": 1.9983203359328134e-06, "loss": 0.0002, "step": 13965 }, { "epoch": 1.67, "learning_rate": 1.9982483503299338e-06, "loss": 0.0003, "step": 13966 }, { "epoch": 1.67, "learning_rate": 1.9981763647270546e-06, "loss": 0.0002, "step": 13967 }, { "epoch": 1.67, "learning_rate": 1.998104379124175e-06, "loss": 0.0002, "step": 13968 }, { "epoch": 1.67, "learning_rate": 1.9980323935212957e-06, "loss": 0.0004, "step": 13969 }, { "epoch": 1.67, "learning_rate": 1.9979604079184165e-06, "loss": 0.0002, "step": 13970 }, { "epoch": 1.67, "learning_rate": 1.9978884223155372e-06, "loss": 0.0003, "step": 13971 }, { "epoch": 1.67, "learning_rate": 1.9978164367126576e-06, "loss": 0.0003, "step": 13972 }, { "epoch": 1.67, "learning_rate": 1.9977444511097784e-06, "loss": 0.0003, "step": 13973 }, { "epoch": 1.67, "learning_rate": 1.9976724655068987e-06, "loss": 0.0002, "step": 13974 }, { "epoch": 1.67, "learning_rate": 1.9976004799040195e-06, "loss": 0.0003, "step": 13975 }, { "epoch": 1.67, "learning_rate": 1.99752849430114e-06, "loss": 0.2242, "step": 13976 }, { "epoch": 1.67, "learning_rate": 1.9974565086982602e-06, "loss": 0.2642, "step": 13977 }, { "epoch": 1.68, "learning_rate": 1.997384523095381e-06, "loss": 0.0003, "step": 13978 }, { "epoch": 1.68, "learning_rate": 1.9973125374925014e-06, "loss": 0.0009, "step": 13979 }, { "epoch": 1.68, "learning_rate": 1.997240551889622e-06, "loss": 0.0003, "step": 13980 }, { "epoch": 1.68, "learning_rate": 1.9971685662867425e-06, "loss": 0.0003, "step": 13981 }, { "epoch": 1.68, "learning_rate": 1.9970965806838633e-06, "loss": 0.0002, "step": 13982 }, { "epoch": 1.68, "learning_rate": 1.9970245950809836e-06, "loss": 0.0003, "step": 13983 }, { "epoch": 1.68, "learning_rate": 1.9969526094781044e-06, "loss": 0.0002, "step": 13984 }, { "epoch": 1.68, "learning_rate": 1.9968806238752248e-06, "loss": 0.0006, "step": 13985 }, { "epoch": 1.68, "learning_rate": 1.9968086382723455e-06, "loss": 0.0004, "step": 13986 }, { "epoch": 1.68, "learning_rate": 1.9967366526694663e-06, "loss": 0.0003, "step": 13987 }, { "epoch": 1.68, "learning_rate": 1.9966646670665867e-06, "loss": 0.0003, "step": 13988 }, { "epoch": 1.68, "learning_rate": 1.9965926814637074e-06, "loss": 0.5506, "step": 13989 }, { "epoch": 1.68, "learning_rate": 1.996520695860828e-06, "loss": 0.0002, "step": 13990 }, { "epoch": 1.68, "learning_rate": 1.9964487102579486e-06, "loss": 0.0002, "step": 13991 }, { "epoch": 1.68, "learning_rate": 1.996376724655069e-06, "loss": 0.0004, "step": 13992 }, { "epoch": 1.68, "learning_rate": 1.9963047390521897e-06, "loss": 0.0004, "step": 13993 }, { "epoch": 1.68, "learning_rate": 1.99623275344931e-06, "loss": 0.0002, "step": 13994 }, { "epoch": 1.68, "learning_rate": 1.996160767846431e-06, "loss": 0.0002, "step": 13995 }, { "epoch": 1.68, "learning_rate": 1.996088782243551e-06, "loss": 0.0003, "step": 13996 }, { "epoch": 1.68, "learning_rate": 1.996016796640672e-06, "loss": 0.3903, "step": 13997 }, { "epoch": 1.68, "learning_rate": 1.9959448110377923e-06, "loss": 0.0003, "step": 13998 }, { "epoch": 1.68, "learning_rate": 1.995872825434913e-06, "loss": 0.0003, "step": 13999 }, { "epoch": 1.68, "learning_rate": 1.9958008398320335e-06, "loss": 0.0003, "step": 14000 }, { "epoch": 1.68, "learning_rate": 1.9957288542291542e-06, "loss": 0.0004, "step": 14001 }, { "epoch": 1.68, "learning_rate": 1.995656868626275e-06, "loss": 0.0003, "step": 14002 }, { "epoch": 1.68, "learning_rate": 1.9955848830233954e-06, "loss": 0.4251, "step": 14003 }, { "epoch": 1.68, "learning_rate": 1.995512897420516e-06, "loss": 0.0003, "step": 14004 }, { "epoch": 1.68, "learning_rate": 1.9954409118176365e-06, "loss": 0.0003, "step": 14005 }, { "epoch": 1.68, "learning_rate": 1.9953689262147573e-06, "loss": 0.0002, "step": 14006 }, { "epoch": 1.68, "learning_rate": 1.9952969406118776e-06, "loss": 0.0003, "step": 14007 }, { "epoch": 1.68, "learning_rate": 1.9952249550089984e-06, "loss": 0.0006, "step": 14008 }, { "epoch": 1.68, "learning_rate": 1.9951529694061188e-06, "loss": 0.0002, "step": 14009 }, { "epoch": 1.68, "learning_rate": 1.9950809838032396e-06, "loss": 0.0001, "step": 14010 }, { "epoch": 1.68, "learning_rate": 1.99500899820036e-06, "loss": 0.0004, "step": 14011 }, { "epoch": 1.68, "learning_rate": 1.9949370125974803e-06, "loss": 0.057, "step": 14012 }, { "epoch": 1.68, "learning_rate": 1.994865026994601e-06, "loss": 0.0002, "step": 14013 }, { "epoch": 1.68, "learning_rate": 1.9947930413917214e-06, "loss": 0.0003, "step": 14014 }, { "epoch": 1.68, "learning_rate": 1.994721055788842e-06, "loss": 0.0003, "step": 14015 }, { "epoch": 1.68, "learning_rate": 1.994649070185963e-06, "loss": 0.0003, "step": 14016 }, { "epoch": 1.68, "learning_rate": 1.9945770845830833e-06, "loss": 0.0002, "step": 14017 }, { "epoch": 1.68, "learning_rate": 1.994505098980204e-06, "loss": 0.0002, "step": 14018 }, { "epoch": 1.68, "learning_rate": 1.994433113377325e-06, "loss": 0.0004, "step": 14019 }, { "epoch": 1.68, "learning_rate": 1.9943611277744452e-06, "loss": 0.2454, "step": 14020 }, { "epoch": 1.68, "learning_rate": 1.994289142171566e-06, "loss": 0.0003, "step": 14021 }, { "epoch": 1.68, "learning_rate": 1.9942171565686864e-06, "loss": 0.0005, "step": 14022 }, { "epoch": 1.68, "learning_rate": 1.9941451709658067e-06, "loss": 0.0003, "step": 14023 }, { "epoch": 1.68, "learning_rate": 1.9940731853629275e-06, "loss": 0.0002, "step": 14024 }, { "epoch": 1.68, "learning_rate": 1.994001199760048e-06, "loss": 0.0003, "step": 14025 }, { "epoch": 1.68, "learning_rate": 1.9939292141571686e-06, "loss": 0.0004, "step": 14026 }, { "epoch": 1.68, "learning_rate": 1.993857228554289e-06, "loss": 0.0003, "step": 14027 }, { "epoch": 1.68, "learning_rate": 1.9937852429514098e-06, "loss": 0.0002, "step": 14028 }, { "epoch": 1.68, "learning_rate": 1.99371325734853e-06, "loss": 0.0002, "step": 14029 }, { "epoch": 1.68, "learning_rate": 1.993641271745651e-06, "loss": 0.0002, "step": 14030 }, { "epoch": 1.68, "learning_rate": 1.9935692861427712e-06, "loss": 0.0002, "step": 14031 }, { "epoch": 1.68, "learning_rate": 1.993497300539892e-06, "loss": 0.0002, "step": 14032 }, { "epoch": 1.68, "learning_rate": 1.993425314937013e-06, "loss": 0.0002, "step": 14033 }, { "epoch": 1.68, "learning_rate": 1.9933533293341336e-06, "loss": 0.0003, "step": 14034 }, { "epoch": 1.68, "learning_rate": 1.993281343731254e-06, "loss": 0.0003, "step": 14035 }, { "epoch": 1.68, "learning_rate": 1.9932093581283743e-06, "loss": 0.0006, "step": 14036 }, { "epoch": 1.68, "learning_rate": 1.993137372525495e-06, "loss": 0.001, "step": 14037 }, { "epoch": 1.68, "learning_rate": 1.9930653869226154e-06, "loss": 0.0002, "step": 14038 }, { "epoch": 1.68, "learning_rate": 1.992993401319736e-06, "loss": 1.0276, "step": 14039 }, { "epoch": 1.68, "learning_rate": 1.9929214157168566e-06, "loss": 0.0003, "step": 14040 }, { "epoch": 1.68, "learning_rate": 1.9928494301139773e-06, "loss": 0.0002, "step": 14041 }, { "epoch": 1.68, "learning_rate": 1.9927774445110977e-06, "loss": 0.0003, "step": 14042 }, { "epoch": 1.68, "learning_rate": 1.9927054589082185e-06, "loss": 0.0003, "step": 14043 }, { "epoch": 1.68, "learning_rate": 1.992633473305339e-06, "loss": 0.0004, "step": 14044 }, { "epoch": 1.68, "learning_rate": 1.9925614877024596e-06, "loss": 0.0003, "step": 14045 }, { "epoch": 1.68, "learning_rate": 1.99248950209958e-06, "loss": 0.814, "step": 14046 }, { "epoch": 1.68, "learning_rate": 1.9924175164967007e-06, "loss": 0.0007, "step": 14047 }, { "epoch": 1.68, "learning_rate": 1.9923455308938215e-06, "loss": 0.0003, "step": 14048 }, { "epoch": 1.68, "learning_rate": 1.992273545290942e-06, "loss": 0.0003, "step": 14049 }, { "epoch": 1.68, "learning_rate": 1.9922015596880627e-06, "loss": 0.0003, "step": 14050 }, { "epoch": 1.68, "learning_rate": 1.992129574085183e-06, "loss": 0.0006, "step": 14051 }, { "epoch": 1.68, "learning_rate": 1.9920575884823038e-06, "loss": 0.0003, "step": 14052 }, { "epoch": 1.68, "learning_rate": 1.991985602879424e-06, "loss": 0.0002, "step": 14053 }, { "epoch": 1.68, "learning_rate": 1.991913617276545e-06, "loss": 0.0004, "step": 14054 }, { "epoch": 1.68, "learning_rate": 1.9918416316736653e-06, "loss": 0.0002, "step": 14055 }, { "epoch": 1.68, "learning_rate": 1.991769646070786e-06, "loss": 0.6762, "step": 14056 }, { "epoch": 1.68, "learning_rate": 1.9916976604679064e-06, "loss": 0.0003, "step": 14057 }, { "epoch": 1.68, "learning_rate": 1.9916256748650268e-06, "loss": 0.0002, "step": 14058 }, { "epoch": 1.68, "learning_rate": 1.9915536892621475e-06, "loss": 0.0003, "step": 14059 }, { "epoch": 1.68, "learning_rate": 1.991481703659268e-06, "loss": 0.0004, "step": 14060 }, { "epoch": 1.68, "learning_rate": 1.9914097180563887e-06, "loss": 0.0002, "step": 14061 }, { "epoch": 1.69, "learning_rate": 1.9913377324535095e-06, "loss": 0.0029, "step": 14062 }, { "epoch": 1.69, "learning_rate": 1.99126574685063e-06, "loss": 0.0003, "step": 14063 }, { "epoch": 1.69, "learning_rate": 1.9911937612477506e-06, "loss": 0.0002, "step": 14064 }, { "epoch": 1.69, "learning_rate": 1.9911217756448714e-06, "loss": 0.0002, "step": 14065 }, { "epoch": 1.69, "learning_rate": 1.9910497900419917e-06, "loss": 0.0003, "step": 14066 }, { "epoch": 1.69, "learning_rate": 1.9909778044391125e-06, "loss": 0.0007, "step": 14067 }, { "epoch": 1.69, "learning_rate": 1.990905818836233e-06, "loss": 0.0002, "step": 14068 }, { "epoch": 1.69, "learning_rate": 1.990833833233353e-06, "loss": 0.0002, "step": 14069 }, { "epoch": 1.69, "learning_rate": 1.990761847630474e-06, "loss": 0.0004, "step": 14070 }, { "epoch": 1.69, "learning_rate": 1.9906898620275943e-06, "loss": 0.0825, "step": 14071 }, { "epoch": 1.69, "learning_rate": 1.990617876424715e-06, "loss": 0.0002, "step": 14072 }, { "epoch": 1.69, "learning_rate": 1.9905458908218355e-06, "loss": 0.0249, "step": 14073 }, { "epoch": 1.69, "learning_rate": 1.9904739052189563e-06, "loss": 0.0002, "step": 14074 }, { "epoch": 1.69, "learning_rate": 1.9904019196160766e-06, "loss": 0.0004, "step": 14075 }, { "epoch": 1.69, "learning_rate": 1.9903299340131974e-06, "loss": 0.0003, "step": 14076 }, { "epoch": 1.69, "learning_rate": 1.9902579484103177e-06, "loss": 0.0002, "step": 14077 }, { "epoch": 1.69, "learning_rate": 1.9901859628074385e-06, "loss": 0.0002, "step": 14078 }, { "epoch": 1.69, "learning_rate": 1.9901139772045593e-06, "loss": 0.0002, "step": 14079 }, { "epoch": 1.69, "learning_rate": 1.99004199160168e-06, "loss": 0.0002, "step": 14080 }, { "epoch": 1.69, "learning_rate": 1.9899700059988004e-06, "loss": 0.0003, "step": 14081 }, { "epoch": 1.69, "learning_rate": 1.989898020395921e-06, "loss": 0.8256, "step": 14082 }, { "epoch": 1.69, "learning_rate": 1.9898260347930416e-06, "loss": 0.759, "step": 14083 }, { "epoch": 1.69, "learning_rate": 1.989754049190162e-06, "loss": 0.0002, "step": 14084 }, { "epoch": 1.69, "learning_rate": 1.9896820635872827e-06, "loss": 0.0002, "step": 14085 }, { "epoch": 1.69, "learning_rate": 1.989610077984403e-06, "loss": 0.0003, "step": 14086 }, { "epoch": 1.69, "learning_rate": 1.989538092381524e-06, "loss": 0.0005, "step": 14087 }, { "epoch": 1.69, "learning_rate": 1.989466106778644e-06, "loss": 0.0002, "step": 14088 }, { "epoch": 1.69, "learning_rate": 1.989394121175765e-06, "loss": 0.0003, "step": 14089 }, { "epoch": 1.69, "learning_rate": 1.9893221355728853e-06, "loss": 0.0004, "step": 14090 }, { "epoch": 1.69, "learning_rate": 1.989250149970006e-06, "loss": 0.0002, "step": 14091 }, { "epoch": 1.69, "learning_rate": 1.9891781643671265e-06, "loss": 0.0002, "step": 14092 }, { "epoch": 1.69, "learning_rate": 1.9891061787642472e-06, "loss": 0.0002, "step": 14093 }, { "epoch": 1.69, "learning_rate": 1.989034193161368e-06, "loss": 1.0508, "step": 14094 }, { "epoch": 1.69, "learning_rate": 1.9889622075584884e-06, "loss": 0.0005, "step": 14095 }, { "epoch": 1.69, "learning_rate": 1.988890221955609e-06, "loss": 0.0002, "step": 14096 }, { "epoch": 1.69, "learning_rate": 1.9888182363527295e-06, "loss": 0.0049, "step": 14097 }, { "epoch": 1.69, "learning_rate": 1.9887462507498503e-06, "loss": 0.0005, "step": 14098 }, { "epoch": 1.69, "learning_rate": 1.9886742651469706e-06, "loss": 0.0003, "step": 14099 }, { "epoch": 1.69, "learning_rate": 1.9886022795440914e-06, "loss": 0.0004, "step": 14100 }, { "epoch": 1.69, "learning_rate": 1.9885302939412118e-06, "loss": 0.0002, "step": 14101 }, { "epoch": 1.69, "learning_rate": 1.9884583083383325e-06, "loss": 0.0006, "step": 14102 }, { "epoch": 1.69, "learning_rate": 1.988386322735453e-06, "loss": 0.0002, "step": 14103 }, { "epoch": 1.69, "learning_rate": 1.9883143371325733e-06, "loss": 0.0003, "step": 14104 }, { "epoch": 1.69, "learning_rate": 1.988242351529694e-06, "loss": 0.0003, "step": 14105 }, { "epoch": 1.69, "learning_rate": 1.9881703659268144e-06, "loss": 0.0003, "step": 14106 }, { "epoch": 1.69, "learning_rate": 1.988098380323935e-06, "loss": 0.1771, "step": 14107 }, { "epoch": 1.69, "learning_rate": 1.988026394721056e-06, "loss": 0.0004, "step": 14108 }, { "epoch": 1.69, "learning_rate": 1.9879544091181763e-06, "loss": 0.0002, "step": 14109 }, { "epoch": 1.69, "learning_rate": 1.987882423515297e-06, "loss": 0.0004, "step": 14110 }, { "epoch": 1.69, "learning_rate": 1.987810437912418e-06, "loss": 0.5478, "step": 14111 }, { "epoch": 1.69, "learning_rate": 1.9877384523095382e-06, "loss": 0.0003, "step": 14112 }, { "epoch": 1.69, "learning_rate": 1.987666466706659e-06, "loss": 0.0003, "step": 14113 }, { "epoch": 1.69, "learning_rate": 1.9875944811037794e-06, "loss": 0.0003, "step": 14114 }, { "epoch": 1.69, "learning_rate": 1.9875224955008997e-06, "loss": 0.0005, "step": 14115 }, { "epoch": 1.69, "learning_rate": 1.9874505098980205e-06, "loss": 0.0002, "step": 14116 }, { "epoch": 1.69, "learning_rate": 1.987378524295141e-06, "loss": 0.0003, "step": 14117 }, { "epoch": 1.69, "learning_rate": 1.9873065386922616e-06, "loss": 0.0011, "step": 14118 }, { "epoch": 1.69, "learning_rate": 1.987234553089382e-06, "loss": 0.0004, "step": 14119 }, { "epoch": 1.69, "learning_rate": 1.9871625674865028e-06, "loss": 0.0009, "step": 14120 }, { "epoch": 1.69, "learning_rate": 1.987090581883623e-06, "loss": 0.0003, "step": 14121 }, { "epoch": 1.69, "learning_rate": 1.987018596280744e-06, "loss": 0.0002, "step": 14122 }, { "epoch": 1.69, "learning_rate": 1.9869466106778642e-06, "loss": 0.0009, "step": 14123 }, { "epoch": 1.69, "learning_rate": 1.986874625074985e-06, "loss": 0.0005, "step": 14124 }, { "epoch": 1.69, "learning_rate": 1.986802639472106e-06, "loss": 0.0003, "step": 14125 }, { "epoch": 1.69, "learning_rate": 1.9867306538692266e-06, "loss": 0.0002, "step": 14126 }, { "epoch": 1.69, "learning_rate": 1.986658668266347e-06, "loss": 0.0008, "step": 14127 }, { "epoch": 1.69, "learning_rate": 1.9865866826634673e-06, "loss": 0.0005, "step": 14128 }, { "epoch": 1.69, "learning_rate": 1.986514697060588e-06, "loss": 0.0002, "step": 14129 }, { "epoch": 1.69, "learning_rate": 1.9864427114577084e-06, "loss": 0.0004, "step": 14130 }, { "epoch": 1.69, "learning_rate": 1.986370725854829e-06, "loss": 0.0003, "step": 14131 }, { "epoch": 1.69, "learning_rate": 1.9862987402519496e-06, "loss": 0.0002, "step": 14132 }, { "epoch": 1.69, "learning_rate": 1.9862267546490703e-06, "loss": 0.0003, "step": 14133 }, { "epoch": 1.69, "learning_rate": 1.9861547690461907e-06, "loss": 0.0006, "step": 14134 }, { "epoch": 1.69, "learning_rate": 1.9860827834433115e-06, "loss": 0.0001, "step": 14135 }, { "epoch": 1.69, "learning_rate": 1.986010797840432e-06, "loss": 0.3538, "step": 14136 }, { "epoch": 1.69, "learning_rate": 1.9859388122375526e-06, "loss": 0.0003, "step": 14137 }, { "epoch": 1.69, "learning_rate": 1.985866826634673e-06, "loss": 0.0001, "step": 14138 }, { "epoch": 1.69, "learning_rate": 1.9857948410317937e-06, "loss": 0.0008, "step": 14139 }, { "epoch": 1.69, "learning_rate": 1.985722855428914e-06, "loss": 0.0003, "step": 14140 }, { "epoch": 1.69, "learning_rate": 1.985650869826035e-06, "loss": 0.0002, "step": 14141 }, { "epoch": 1.69, "learning_rate": 1.9855788842231556e-06, "loss": 0.0003, "step": 14142 }, { "epoch": 1.69, "learning_rate": 1.985506898620276e-06, "loss": 0.0002, "step": 14143 }, { "epoch": 1.69, "learning_rate": 1.9854349130173968e-06, "loss": 0.0002, "step": 14144 }, { "epoch": 1.7, "learning_rate": 1.985362927414517e-06, "loss": 0.0004, "step": 14145 }, { "epoch": 1.7, "learning_rate": 1.985290941811638e-06, "loss": 0.0002, "step": 14146 }, { "epoch": 1.7, "learning_rate": 1.9852189562087583e-06, "loss": 0.0004, "step": 14147 }, { "epoch": 1.7, "learning_rate": 1.985146970605879e-06, "loss": 0.0032, "step": 14148 }, { "epoch": 1.7, "learning_rate": 1.9850749850029994e-06, "loss": 0.0002, "step": 14149 }, { "epoch": 1.7, "learning_rate": 1.9850029994001198e-06, "loss": 0.0003, "step": 14150 }, { "epoch": 1.7, "learning_rate": 1.9849310137972405e-06, "loss": 0.9091, "step": 14151 }, { "epoch": 1.7, "learning_rate": 1.984859028194361e-06, "loss": 0.0003, "step": 14152 }, { "epoch": 1.7, "learning_rate": 1.9847870425914817e-06, "loss": 0.0005, "step": 14153 }, { "epoch": 1.7, "learning_rate": 1.984715056988602e-06, "loss": 0.0003, "step": 14154 }, { "epoch": 1.7, "learning_rate": 1.984643071385723e-06, "loss": 0.0003, "step": 14155 }, { "epoch": 1.7, "learning_rate": 1.9845710857828436e-06, "loss": 0.0002, "step": 14156 }, { "epoch": 1.7, "learning_rate": 1.9844991001799644e-06, "loss": 0.0012, "step": 14157 }, { "epoch": 1.7, "learning_rate": 1.9844271145770847e-06, "loss": 0.0003, "step": 14158 }, { "epoch": 1.7, "learning_rate": 1.9843551289742055e-06, "loss": 0.0003, "step": 14159 }, { "epoch": 1.7, "learning_rate": 1.984283143371326e-06, "loss": 0.0023, "step": 14160 }, { "epoch": 1.7, "learning_rate": 1.9842111577684466e-06, "loss": 0.0002, "step": 14161 }, { "epoch": 1.7, "learning_rate": 1.984139172165567e-06, "loss": 0.0002, "step": 14162 }, { "epoch": 1.7, "learning_rate": 1.9840671865626873e-06, "loss": 0.0002, "step": 14163 }, { "epoch": 1.7, "learning_rate": 1.983995200959808e-06, "loss": 0.0006, "step": 14164 }, { "epoch": 1.7, "learning_rate": 1.9839232153569285e-06, "loss": 0.0003, "step": 14165 }, { "epoch": 1.7, "learning_rate": 1.9838512297540492e-06, "loss": 0.0011, "step": 14166 }, { "epoch": 1.7, "learning_rate": 1.9837792441511696e-06, "loss": 0.0002, "step": 14167 }, { "epoch": 1.7, "learning_rate": 1.9837072585482904e-06, "loss": 0.0002, "step": 14168 }, { "epoch": 1.7, "learning_rate": 1.9836352729454107e-06, "loss": 0.0002, "step": 14169 }, { "epoch": 1.7, "learning_rate": 1.9835632873425315e-06, "loss": 0.0007, "step": 14170 }, { "epoch": 1.7, "learning_rate": 1.9834913017396523e-06, "loss": 0.0002, "step": 14171 }, { "epoch": 1.7, "learning_rate": 1.9834193161367726e-06, "loss": 0.0314, "step": 14172 }, { "epoch": 1.7, "learning_rate": 1.9833473305338934e-06, "loss": 0.0005, "step": 14173 }, { "epoch": 1.7, "learning_rate": 1.9832753449310138e-06, "loss": 0.0002, "step": 14174 }, { "epoch": 1.7, "learning_rate": 1.9832033593281346e-06, "loss": 0.0002, "step": 14175 }, { "epoch": 1.7, "learning_rate": 1.983131373725255e-06, "loss": 0.0043, "step": 14176 }, { "epoch": 1.7, "learning_rate": 1.9830593881223757e-06, "loss": 0.0031, "step": 14177 }, { "epoch": 1.7, "learning_rate": 1.982987402519496e-06, "loss": 0.0003, "step": 14178 }, { "epoch": 1.7, "learning_rate": 1.982915416916617e-06, "loss": 0.0001, "step": 14179 }, { "epoch": 1.7, "learning_rate": 1.982843431313737e-06, "loss": 0.0002, "step": 14180 }, { "epoch": 1.7, "learning_rate": 1.982771445710858e-06, "loss": 0.0003, "step": 14181 }, { "epoch": 1.7, "learning_rate": 1.9826994601079783e-06, "loss": 0.0003, "step": 14182 }, { "epoch": 1.7, "learning_rate": 1.982627474505099e-06, "loss": 0.0003, "step": 14183 }, { "epoch": 1.7, "learning_rate": 1.9825554889022195e-06, "loss": 0.0009, "step": 14184 }, { "epoch": 1.7, "learning_rate": 1.9824835032993402e-06, "loss": 0.0003, "step": 14185 }, { "epoch": 1.7, "learning_rate": 1.9824115176964606e-06, "loss": 0.0003, "step": 14186 }, { "epoch": 1.7, "learning_rate": 1.9823395320935814e-06, "loss": 0.0007, "step": 14187 }, { "epoch": 1.7, "learning_rate": 1.982267546490702e-06, "loss": 0.2912, "step": 14188 }, { "epoch": 1.7, "learning_rate": 1.9821955608878225e-06, "loss": 0.0002, "step": 14189 }, { "epoch": 1.7, "learning_rate": 1.9821235752849433e-06, "loss": 0.0003, "step": 14190 }, { "epoch": 1.7, "learning_rate": 1.9820515896820636e-06, "loss": 0.0002, "step": 14191 }, { "epoch": 1.7, "learning_rate": 1.9819796040791844e-06, "loss": 0.6456, "step": 14192 }, { "epoch": 1.7, "learning_rate": 1.9819076184763048e-06, "loss": 0.0002, "step": 14193 }, { "epoch": 1.7, "learning_rate": 1.9818356328734255e-06, "loss": 0.0002, "step": 14194 }, { "epoch": 1.7, "learning_rate": 1.981763647270546e-06, "loss": 0.0005, "step": 14195 }, { "epoch": 1.7, "learning_rate": 1.9816916616676663e-06, "loss": 0.0004, "step": 14196 }, { "epoch": 1.7, "learning_rate": 1.981619676064787e-06, "loss": 0.0064, "step": 14197 }, { "epoch": 1.7, "learning_rate": 1.9815476904619074e-06, "loss": 0.0015, "step": 14198 }, { "epoch": 1.7, "learning_rate": 1.981475704859028e-06, "loss": 0.0003, "step": 14199 }, { "epoch": 1.7, "learning_rate": 1.9814037192561485e-06, "loss": 0.0004, "step": 14200 }, { "epoch": 1.7, "learning_rate": 1.9813317336532693e-06, "loss": 0.0004, "step": 14201 }, { "epoch": 1.7, "learning_rate": 1.98125974805039e-06, "loss": 0.0003, "step": 14202 }, { "epoch": 1.7, "learning_rate": 1.981187762447511e-06, "loss": 0.0003, "step": 14203 }, { "epoch": 1.7, "learning_rate": 1.981115776844631e-06, "loss": 0.0004, "step": 14204 }, { "epoch": 1.7, "learning_rate": 1.981043791241752e-06, "loss": 0.0003, "step": 14205 }, { "epoch": 1.7, "learning_rate": 1.9809718056388723e-06, "loss": 0.0021, "step": 14206 }, { "epoch": 1.7, "learning_rate": 1.980899820035993e-06, "loss": 0.0003, "step": 14207 }, { "epoch": 1.7, "learning_rate": 1.9808278344331135e-06, "loss": 0.0003, "step": 14208 }, { "epoch": 1.7, "learning_rate": 1.980755848830234e-06, "loss": 0.0008, "step": 14209 }, { "epoch": 1.7, "learning_rate": 1.9806838632273546e-06, "loss": 0.0004, "step": 14210 }, { "epoch": 1.7, "learning_rate": 1.980611877624475e-06, "loss": 0.0005, "step": 14211 }, { "epoch": 1.7, "learning_rate": 1.9805398920215957e-06, "loss": 0.0004, "step": 14212 }, { "epoch": 1.7, "learning_rate": 1.980467906418716e-06, "loss": 0.0002, "step": 14213 }, { "epoch": 1.7, "learning_rate": 1.980395920815837e-06, "loss": 0.0002, "step": 14214 }, { "epoch": 1.7, "learning_rate": 1.9803239352129572e-06, "loss": 0.0003, "step": 14215 }, { "epoch": 1.7, "learning_rate": 1.980251949610078e-06, "loss": 0.0002, "step": 14216 }, { "epoch": 1.7, "learning_rate": 1.980179964007199e-06, "loss": 0.3679, "step": 14217 }, { "epoch": 1.7, "learning_rate": 1.980107978404319e-06, "loss": 0.0005, "step": 14218 }, { "epoch": 1.7, "learning_rate": 1.98003599280144e-06, "loss": 0.0004, "step": 14219 }, { "epoch": 1.7, "learning_rate": 1.9799640071985603e-06, "loss": 0.0002, "step": 14220 }, { "epoch": 1.7, "learning_rate": 1.979892021595681e-06, "loss": 0.0004, "step": 14221 }, { "epoch": 1.7, "learning_rate": 1.9798200359928014e-06, "loss": 0.0002, "step": 14222 }, { "epoch": 1.7, "learning_rate": 1.979748050389922e-06, "loss": 0.0002, "step": 14223 }, { "epoch": 1.7, "learning_rate": 1.9796760647870425e-06, "loss": 0.0004, "step": 14224 }, { "epoch": 1.7, "learning_rate": 1.9796040791841633e-06, "loss": 0.0002, "step": 14225 }, { "epoch": 1.7, "learning_rate": 1.9795320935812837e-06, "loss": 0.0002, "step": 14226 }, { "epoch": 1.7, "learning_rate": 1.9794601079784045e-06, "loss": 0.0002, "step": 14227 }, { "epoch": 1.7, "learning_rate": 1.979388122375525e-06, "loss": 0.0177, "step": 14228 }, { "epoch": 1.71, "learning_rate": 1.9793161367726456e-06, "loss": 0.0006, "step": 14229 }, { "epoch": 1.71, "learning_rate": 1.979244151169766e-06, "loss": 0.0023, "step": 14230 }, { "epoch": 1.71, "learning_rate": 1.9791721655668867e-06, "loss": 0.0002, "step": 14231 }, { "epoch": 1.71, "learning_rate": 1.979100179964007e-06, "loss": 0.0002, "step": 14232 }, { "epoch": 1.71, "learning_rate": 1.979028194361128e-06, "loss": 0.0003, "step": 14233 }, { "epoch": 1.71, "learning_rate": 1.9789562087582486e-06, "loss": 0.0003, "step": 14234 }, { "epoch": 1.71, "learning_rate": 1.978884223155369e-06, "loss": 0.0003, "step": 14235 }, { "epoch": 1.71, "learning_rate": 1.9788122375524898e-06, "loss": 0.0003, "step": 14236 }, { "epoch": 1.71, "learning_rate": 1.97874025194961e-06, "loss": 0.0001, "step": 14237 }, { "epoch": 1.71, "learning_rate": 1.978668266346731e-06, "loss": 0.0004, "step": 14238 }, { "epoch": 1.71, "learning_rate": 1.9785962807438513e-06, "loss": 0.0002, "step": 14239 }, { "epoch": 1.71, "learning_rate": 1.978524295140972e-06, "loss": 0.0003, "step": 14240 }, { "epoch": 1.71, "learning_rate": 1.9784523095380924e-06, "loss": 0.0002, "step": 14241 }, { "epoch": 1.71, "learning_rate": 1.9783803239352127e-06, "loss": 0.0002, "step": 14242 }, { "epoch": 1.71, "learning_rate": 1.9783083383323335e-06, "loss": 0.0003, "step": 14243 }, { "epoch": 1.71, "learning_rate": 1.978236352729454e-06, "loss": 0.0002, "step": 14244 }, { "epoch": 1.71, "learning_rate": 1.9781643671265747e-06, "loss": 0.0002, "step": 14245 }, { "epoch": 1.71, "learning_rate": 1.978092381523695e-06, "loss": 0.0002, "step": 14246 }, { "epoch": 1.71, "learning_rate": 1.978020395920816e-06, "loss": 0.0001, "step": 14247 }, { "epoch": 1.71, "learning_rate": 1.9779484103179366e-06, "loss": 0.0002, "step": 14248 }, { "epoch": 1.71, "learning_rate": 1.9778764247150574e-06, "loss": 0.0002, "step": 14249 }, { "epoch": 1.71, "learning_rate": 1.9778044391121777e-06, "loss": 0.0003, "step": 14250 }, { "epoch": 1.71, "learning_rate": 1.9777324535092985e-06, "loss": 0.0003, "step": 14251 }, { "epoch": 1.71, "learning_rate": 1.977660467906419e-06, "loss": 0.0009, "step": 14252 }, { "epoch": 1.71, "learning_rate": 1.9775884823035396e-06, "loss": 0.0002, "step": 14253 }, { "epoch": 1.71, "learning_rate": 1.97751649670066e-06, "loss": 0.0009, "step": 14254 }, { "epoch": 1.71, "learning_rate": 1.9774445110977803e-06, "loss": 0.0002, "step": 14255 }, { "epoch": 1.71, "learning_rate": 1.977372525494901e-06, "loss": 0.0001, "step": 14256 }, { "epoch": 1.71, "learning_rate": 1.9773005398920215e-06, "loss": 0.0002, "step": 14257 }, { "epoch": 1.71, "learning_rate": 1.9772285542891422e-06, "loss": 0.0003, "step": 14258 }, { "epoch": 1.71, "learning_rate": 1.9771565686862626e-06, "loss": 0.0006, "step": 14259 }, { "epoch": 1.71, "learning_rate": 1.9770845830833834e-06, "loss": 0.0002, "step": 14260 }, { "epoch": 1.71, "learning_rate": 1.9770125974805037e-06, "loss": 0.0002, "step": 14261 }, { "epoch": 1.71, "learning_rate": 1.9769406118776245e-06, "loss": 0.4438, "step": 14262 }, { "epoch": 1.71, "learning_rate": 1.976868626274745e-06, "loss": 0.0003, "step": 14263 }, { "epoch": 1.71, "learning_rate": 1.9767966406718656e-06, "loss": 0.0002, "step": 14264 }, { "epoch": 1.71, "learning_rate": 1.9767246550689864e-06, "loss": 0.0003, "step": 14265 }, { "epoch": 1.71, "learning_rate": 1.9766526694661068e-06, "loss": 0.0006, "step": 14266 }, { "epoch": 1.71, "learning_rate": 1.9765806838632276e-06, "loss": 0.0004, "step": 14267 }, { "epoch": 1.71, "learning_rate": 1.976508698260348e-06, "loss": 0.0002, "step": 14268 }, { "epoch": 1.71, "learning_rate": 1.9764367126574687e-06, "loss": 0.0003, "step": 14269 }, { "epoch": 1.71, "learning_rate": 1.976364727054589e-06, "loss": 0.0003, "step": 14270 }, { "epoch": 1.71, "learning_rate": 1.97629274145171e-06, "loss": 0.0002, "step": 14271 }, { "epoch": 1.71, "learning_rate": 1.97622075584883e-06, "loss": 0.0001, "step": 14272 }, { "epoch": 1.71, "learning_rate": 1.976148770245951e-06, "loss": 0.0002, "step": 14273 }, { "epoch": 1.71, "learning_rate": 1.9760767846430713e-06, "loss": 0.7968, "step": 14274 }, { "epoch": 1.71, "learning_rate": 1.976004799040192e-06, "loss": 0.0003, "step": 14275 }, { "epoch": 1.71, "learning_rate": 1.9759328134373124e-06, "loss": 0.0003, "step": 14276 }, { "epoch": 1.71, "learning_rate": 1.975860827834433e-06, "loss": 0.0002, "step": 14277 }, { "epoch": 1.71, "learning_rate": 1.9757888422315536e-06, "loss": 0.0004, "step": 14278 }, { "epoch": 1.71, "learning_rate": 1.9757168566286744e-06, "loss": 0.0002, "step": 14279 }, { "epoch": 1.71, "learning_rate": 1.975644871025795e-06, "loss": 0.0002, "step": 14280 }, { "epoch": 1.71, "learning_rate": 1.9755728854229155e-06, "loss": 0.0002, "step": 14281 }, { "epoch": 1.71, "learning_rate": 1.9755008998200363e-06, "loss": 0.4517, "step": 14282 }, { "epoch": 1.71, "learning_rate": 1.9754289142171566e-06, "loss": 0.0004, "step": 14283 }, { "epoch": 1.71, "learning_rate": 1.9753569286142774e-06, "loss": 0.0003, "step": 14284 }, { "epoch": 1.71, "learning_rate": 1.9752849430113978e-06, "loss": 0.0365, "step": 14285 }, { "epoch": 1.71, "learning_rate": 1.9752129574085185e-06, "loss": 0.0002, "step": 14286 }, { "epoch": 1.71, "learning_rate": 1.975140971805639e-06, "loss": 0.0003, "step": 14287 }, { "epoch": 1.71, "learning_rate": 1.9750689862027597e-06, "loss": 0.0003, "step": 14288 }, { "epoch": 1.71, "learning_rate": 1.97499700059988e-06, "loss": 0.0003, "step": 14289 }, { "epoch": 1.71, "learning_rate": 1.9749250149970004e-06, "loss": 0.0002, "step": 14290 }, { "epoch": 1.71, "learning_rate": 1.974853029394121e-06, "loss": 0.0001, "step": 14291 }, { "epoch": 1.71, "learning_rate": 1.9747810437912415e-06, "loss": 0.0003, "step": 14292 }, { "epoch": 1.71, "learning_rate": 1.9747090581883623e-06, "loss": 0.0002, "step": 14293 }, { "epoch": 1.71, "learning_rate": 1.974637072585483e-06, "loss": 0.0002, "step": 14294 }, { "epoch": 1.71, "learning_rate": 1.9745650869826034e-06, "loss": 0.0002, "step": 14295 }, { "epoch": 1.71, "learning_rate": 1.974493101379724e-06, "loss": 0.0019, "step": 14296 }, { "epoch": 1.71, "learning_rate": 1.974421115776845e-06, "loss": 0.0002, "step": 14297 }, { "epoch": 1.71, "learning_rate": 1.9743491301739653e-06, "loss": 0.0002, "step": 14298 }, { "epoch": 1.71, "learning_rate": 1.974277144571086e-06, "loss": 0.0002, "step": 14299 }, { "epoch": 1.71, "learning_rate": 1.9742051589682065e-06, "loss": 0.0002, "step": 14300 }, { "epoch": 1.71, "learning_rate": 1.974133173365327e-06, "loss": 0.4546, "step": 14301 }, { "epoch": 1.71, "learning_rate": 1.9740611877624476e-06, "loss": 0.0002, "step": 14302 }, { "epoch": 1.71, "learning_rate": 1.973989202159568e-06, "loss": 0.0006, "step": 14303 }, { "epoch": 1.71, "learning_rate": 1.9739172165566887e-06, "loss": 0.0011, "step": 14304 }, { "epoch": 1.71, "learning_rate": 1.973845230953809e-06, "loss": 0.0002, "step": 14305 }, { "epoch": 1.71, "learning_rate": 1.97377324535093e-06, "loss": 0.0005, "step": 14306 }, { "epoch": 1.71, "learning_rate": 1.9737012597480502e-06, "loss": 0.0003, "step": 14307 }, { "epoch": 1.71, "learning_rate": 1.973629274145171e-06, "loss": 0.0008, "step": 14308 }, { "epoch": 1.71, "learning_rate": 1.9735572885422914e-06, "loss": 0.0002, "step": 14309 }, { "epoch": 1.71, "learning_rate": 1.973485302939412e-06, "loss": 0.0003, "step": 14310 }, { "epoch": 1.71, "learning_rate": 1.973413317336533e-06, "loss": 0.0002, "step": 14311 }, { "epoch": 1.72, "learning_rate": 1.9733413317336533e-06, "loss": 0.0018, "step": 14312 }, { "epoch": 1.72, "learning_rate": 1.973269346130774e-06, "loss": 0.0001, "step": 14313 }, { "epoch": 1.72, "learning_rate": 1.9731973605278944e-06, "loss": 0.0001, "step": 14314 }, { "epoch": 1.72, "learning_rate": 1.973125374925015e-06, "loss": 0.0002, "step": 14315 }, { "epoch": 1.72, "learning_rate": 1.9730533893221355e-06, "loss": 0.0002, "step": 14316 }, { "epoch": 1.72, "learning_rate": 1.9729814037192563e-06, "loss": 0.0002, "step": 14317 }, { "epoch": 1.72, "learning_rate": 1.9729094181163767e-06, "loss": 0.0003, "step": 14318 }, { "epoch": 1.72, "learning_rate": 1.9728374325134974e-06, "loss": 0.0002, "step": 14319 }, { "epoch": 1.72, "learning_rate": 1.972765446910618e-06, "loss": 0.0001, "step": 14320 }, { "epoch": 1.72, "learning_rate": 1.9726934613077386e-06, "loss": 0.0001, "step": 14321 }, { "epoch": 1.72, "learning_rate": 1.972621475704859e-06, "loss": 0.0004, "step": 14322 }, { "epoch": 1.72, "learning_rate": 1.9725494901019793e-06, "loss": 0.0001, "step": 14323 }, { "epoch": 1.72, "learning_rate": 1.9724775044991e-06, "loss": 0.0002, "step": 14324 }, { "epoch": 1.72, "learning_rate": 1.972405518896221e-06, "loss": 0.0002, "step": 14325 }, { "epoch": 1.72, "learning_rate": 1.9723335332933416e-06, "loss": 0.0006, "step": 14326 }, { "epoch": 1.72, "learning_rate": 1.972261547690462e-06, "loss": 0.0005, "step": 14327 }, { "epoch": 1.72, "learning_rate": 1.9721895620875828e-06, "loss": 0.0002, "step": 14328 }, { "epoch": 1.72, "learning_rate": 1.972117576484703e-06, "loss": 0.0002, "step": 14329 }, { "epoch": 1.72, "learning_rate": 1.972045590881824e-06, "loss": 0.0001, "step": 14330 }, { "epoch": 1.72, "learning_rate": 1.9719736052789443e-06, "loss": 0.0024, "step": 14331 }, { "epoch": 1.72, "learning_rate": 1.971901619676065e-06, "loss": 0.0002, "step": 14332 }, { "epoch": 1.72, "learning_rate": 1.9718296340731854e-06, "loss": 0.0003, "step": 14333 }, { "epoch": 1.72, "learning_rate": 1.971757648470306e-06, "loss": 0.0003, "step": 14334 }, { "epoch": 1.72, "learning_rate": 1.9716856628674265e-06, "loss": 0.0002, "step": 14335 }, { "epoch": 1.72, "learning_rate": 1.971613677264547e-06, "loss": 0.0003, "step": 14336 }, { "epoch": 1.72, "learning_rate": 1.9715416916616677e-06, "loss": 0.0002, "step": 14337 }, { "epoch": 1.72, "learning_rate": 1.971469706058788e-06, "loss": 0.0002, "step": 14338 }, { "epoch": 1.72, "learning_rate": 1.9713977204559088e-06, "loss": 0.0001, "step": 14339 }, { "epoch": 1.72, "learning_rate": 1.9713257348530296e-06, "loss": 0.0002, "step": 14340 }, { "epoch": 1.72, "learning_rate": 1.97125374925015e-06, "loss": 0.8022, "step": 14341 }, { "epoch": 1.72, "learning_rate": 1.9711817636472707e-06, "loss": 0.5346, "step": 14342 }, { "epoch": 1.72, "learning_rate": 1.9711097780443915e-06, "loss": 0.0002, "step": 14343 }, { "epoch": 1.72, "learning_rate": 1.971037792441512e-06, "loss": 0.4824, "step": 14344 }, { "epoch": 1.72, "learning_rate": 1.9709658068386326e-06, "loss": 0.0003, "step": 14345 }, { "epoch": 1.72, "learning_rate": 1.970893821235753e-06, "loss": 0.0002, "step": 14346 }, { "epoch": 1.72, "learning_rate": 1.9708218356328733e-06, "loss": 0.0002, "step": 14347 }, { "epoch": 1.72, "learning_rate": 1.970749850029994e-06, "loss": 0.0002, "step": 14348 }, { "epoch": 1.72, "learning_rate": 1.9706778644271145e-06, "loss": 0.0003, "step": 14349 }, { "epoch": 1.72, "learning_rate": 1.9706058788242352e-06, "loss": 0.7193, "step": 14350 }, { "epoch": 1.72, "learning_rate": 1.9705338932213556e-06, "loss": 0.0004, "step": 14351 }, { "epoch": 1.72, "learning_rate": 1.9704619076184764e-06, "loss": 0.0003, "step": 14352 }, { "epoch": 1.72, "learning_rate": 1.9703899220155967e-06, "loss": 0.0004, "step": 14353 }, { "epoch": 1.72, "learning_rate": 1.9703179364127175e-06, "loss": 0.0003, "step": 14354 }, { "epoch": 1.72, "learning_rate": 1.970245950809838e-06, "loss": 0.0011, "step": 14355 }, { "epoch": 1.72, "learning_rate": 1.9701739652069586e-06, "loss": 0.0003, "step": 14356 }, { "epoch": 1.72, "learning_rate": 1.9701019796040794e-06, "loss": 0.79, "step": 14357 }, { "epoch": 1.72, "learning_rate": 1.9700299940011998e-06, "loss": 0.0003, "step": 14358 }, { "epoch": 1.72, "learning_rate": 1.9699580083983205e-06, "loss": 0.0017, "step": 14359 }, { "epoch": 1.72, "learning_rate": 1.969886022795441e-06, "loss": 0.0001, "step": 14360 }, { "epoch": 1.72, "learning_rate": 1.9698140371925617e-06, "loss": 0.0002, "step": 14361 }, { "epoch": 1.72, "learning_rate": 1.969742051589682e-06, "loss": 0.0001, "step": 14362 }, { "epoch": 1.72, "learning_rate": 1.969670065986803e-06, "loss": 0.0001, "step": 14363 }, { "epoch": 1.72, "learning_rate": 1.969598080383923e-06, "loss": 0.0002, "step": 14364 }, { "epoch": 1.72, "learning_rate": 1.969526094781044e-06, "loss": 0.0002, "step": 14365 }, { "epoch": 1.72, "learning_rate": 1.9694541091781643e-06, "loss": 0.0007, "step": 14366 }, { "epoch": 1.72, "learning_rate": 1.969382123575285e-06, "loss": 0.0002, "step": 14367 }, { "epoch": 1.72, "learning_rate": 1.9693101379724054e-06, "loss": 0.0003, "step": 14368 }, { "epoch": 1.72, "learning_rate": 1.969238152369526e-06, "loss": 0.0002, "step": 14369 }, { "epoch": 1.72, "learning_rate": 1.9691661667666466e-06, "loss": 0.0002, "step": 14370 }, { "epoch": 1.72, "learning_rate": 1.9690941811637673e-06, "loss": 1.0885, "step": 14371 }, { "epoch": 1.72, "learning_rate": 1.969022195560888e-06, "loss": 0.0002, "step": 14372 }, { "epoch": 1.72, "learning_rate": 1.9689502099580085e-06, "loss": 0.0005, "step": 14373 }, { "epoch": 1.72, "learning_rate": 1.9688782243551293e-06, "loss": 0.0002, "step": 14374 }, { "epoch": 1.72, "learning_rate": 1.9688062387522496e-06, "loss": 0.0002, "step": 14375 }, { "epoch": 1.72, "learning_rate": 1.9687342531493704e-06, "loss": 0.0009, "step": 14376 }, { "epoch": 1.72, "learning_rate": 1.9686622675464907e-06, "loss": 0.0003, "step": 14377 }, { "epoch": 1.72, "learning_rate": 1.9685902819436115e-06, "loss": 0.0006, "step": 14378 }, { "epoch": 1.72, "learning_rate": 1.968518296340732e-06, "loss": 0.0002, "step": 14379 }, { "epoch": 1.72, "learning_rate": 1.9684463107378527e-06, "loss": 0.0002, "step": 14380 }, { "epoch": 1.72, "learning_rate": 1.968374325134973e-06, "loss": 0.0014, "step": 14381 }, { "epoch": 1.72, "learning_rate": 1.9683023395320934e-06, "loss": 0.0002, "step": 14382 }, { "epoch": 1.72, "learning_rate": 1.968230353929214e-06, "loss": 0.0003, "step": 14383 }, { "epoch": 1.72, "learning_rate": 1.9681583683263345e-06, "loss": 0.5955, "step": 14384 }, { "epoch": 1.72, "learning_rate": 1.9680863827234553e-06, "loss": 0.0003, "step": 14385 }, { "epoch": 1.72, "learning_rate": 1.9680143971205756e-06, "loss": 0.0002, "step": 14386 }, { "epoch": 1.72, "learning_rate": 1.9679424115176964e-06, "loss": 0.0003, "step": 14387 }, { "epoch": 1.72, "learning_rate": 1.967870425914817e-06, "loss": 0.0004, "step": 14388 }, { "epoch": 1.72, "learning_rate": 1.967798440311938e-06, "loss": 0.0002, "step": 14389 }, { "epoch": 1.72, "learning_rate": 1.9677264547090583e-06, "loss": 0.0002, "step": 14390 }, { "epoch": 1.72, "learning_rate": 1.967654469106179e-06, "loss": 0.0003, "step": 14391 }, { "epoch": 1.72, "learning_rate": 1.9675824835032995e-06, "loss": 0.0006, "step": 14392 }, { "epoch": 1.72, "learning_rate": 1.96751049790042e-06, "loss": 0.0001, "step": 14393 }, { "epoch": 1.72, "learning_rate": 1.9674385122975406e-06, "loss": 0.0002, "step": 14394 }, { "epoch": 1.72, "learning_rate": 1.967366526694661e-06, "loss": 0.0003, "step": 14395 }, { "epoch": 1.73, "learning_rate": 1.9672945410917817e-06, "loss": 0.0002, "step": 14396 }, { "epoch": 1.73, "learning_rate": 1.967222555488902e-06, "loss": 0.0005, "step": 14397 }, { "epoch": 1.73, "learning_rate": 1.967150569886023e-06, "loss": 0.0003, "step": 14398 }, { "epoch": 1.73, "learning_rate": 1.9670785842831432e-06, "loss": 0.0004, "step": 14399 }, { "epoch": 1.73, "learning_rate": 1.967006598680264e-06, "loss": 0.0001, "step": 14400 }, { "epoch": 1.73, "learning_rate": 1.9669346130773844e-06, "loss": 0.0002, "step": 14401 }, { "epoch": 1.73, "learning_rate": 1.966862627474505e-06, "loss": 0.0003, "step": 14402 }, { "epoch": 1.73, "learning_rate": 1.966790641871626e-06, "loss": 0.0002, "step": 14403 }, { "epoch": 1.73, "learning_rate": 1.9667186562687463e-06, "loss": 0.0002, "step": 14404 }, { "epoch": 1.73, "learning_rate": 1.966646670665867e-06, "loss": 0.0001, "step": 14405 }, { "epoch": 1.73, "learning_rate": 1.9665746850629874e-06, "loss": 0.0002, "step": 14406 }, { "epoch": 1.73, "learning_rate": 1.966502699460108e-06, "loss": 0.0002, "step": 14407 }, { "epoch": 1.73, "learning_rate": 1.9664307138572285e-06, "loss": 0.0001, "step": 14408 }, { "epoch": 1.73, "learning_rate": 1.9663587282543493e-06, "loss": 0.0003, "step": 14409 }, { "epoch": 1.73, "learning_rate": 1.9662867426514697e-06, "loss": 0.0004, "step": 14410 }, { "epoch": 1.73, "learning_rate": 1.9662147570485904e-06, "loss": 0.0004, "step": 14411 }, { "epoch": 1.73, "learning_rate": 1.966142771445711e-06, "loss": 0.0001, "step": 14412 }, { "epoch": 1.73, "learning_rate": 1.9660707858428316e-06, "loss": 0.0001, "step": 14413 }, { "epoch": 1.73, "learning_rate": 1.965998800239952e-06, "loss": 0.0001, "step": 14414 }, { "epoch": 1.73, "learning_rate": 1.9659268146370727e-06, "loss": 0.0004, "step": 14415 }, { "epoch": 1.73, "learning_rate": 1.965854829034193e-06, "loss": 0.0003, "step": 14416 }, { "epoch": 1.73, "learning_rate": 1.965782843431314e-06, "loss": 0.0005, "step": 14417 }, { "epoch": 1.73, "learning_rate": 1.965710857828434e-06, "loss": 0.0003, "step": 14418 }, { "epoch": 1.73, "learning_rate": 1.965638872225555e-06, "loss": 1.0103, "step": 14419 }, { "epoch": 1.73, "learning_rate": 1.9655668866226758e-06, "loss": 0.9108, "step": 14420 }, { "epoch": 1.73, "learning_rate": 1.965494901019796e-06, "loss": 0.0026, "step": 14421 }, { "epoch": 1.73, "learning_rate": 1.965422915416917e-06, "loss": 0.0003, "step": 14422 }, { "epoch": 1.73, "learning_rate": 1.9653509298140372e-06, "loss": 0.0003, "step": 14423 }, { "epoch": 1.73, "learning_rate": 1.965278944211158e-06, "loss": 0.0003, "step": 14424 }, { "epoch": 1.73, "learning_rate": 1.9652069586082784e-06, "loss": 0.0009, "step": 14425 }, { "epoch": 1.73, "learning_rate": 1.965134973005399e-06, "loss": 0.0914, "step": 14426 }, { "epoch": 1.73, "learning_rate": 1.9650629874025195e-06, "loss": 0.0041, "step": 14427 }, { "epoch": 1.73, "learning_rate": 1.96499100179964e-06, "loss": 0.0003, "step": 14428 }, { "epoch": 1.73, "learning_rate": 1.9649190161967606e-06, "loss": 0.0005, "step": 14429 }, { "epoch": 1.73, "learning_rate": 1.964847030593881e-06, "loss": 0.0006, "step": 14430 }, { "epoch": 1.73, "learning_rate": 1.9647750449910018e-06, "loss": 0.0004, "step": 14431 }, { "epoch": 1.73, "learning_rate": 1.964703059388122e-06, "loss": 0.0012, "step": 14432 }, { "epoch": 1.73, "learning_rate": 1.964631073785243e-06, "loss": 0.7753, "step": 14433 }, { "epoch": 1.73, "learning_rate": 1.9645590881823637e-06, "loss": 0.0002, "step": 14434 }, { "epoch": 1.73, "learning_rate": 1.9644871025794845e-06, "loss": 0.2486, "step": 14435 }, { "epoch": 1.73, "learning_rate": 1.964415116976605e-06, "loss": 0.9567, "step": 14436 }, { "epoch": 1.73, "learning_rate": 1.9643431313737256e-06, "loss": 0.0002, "step": 14437 }, { "epoch": 1.73, "learning_rate": 1.964271145770846e-06, "loss": 0.0002, "step": 14438 }, { "epoch": 1.73, "learning_rate": 1.9641991601679663e-06, "loss": 0.0003, "step": 14439 }, { "epoch": 1.73, "learning_rate": 1.964127174565087e-06, "loss": 0.0003, "step": 14440 }, { "epoch": 1.73, "learning_rate": 1.9640551889622074e-06, "loss": 0.0003, "step": 14441 }, { "epoch": 1.73, "learning_rate": 1.9639832033593282e-06, "loss": 0.001, "step": 14442 }, { "epoch": 1.73, "learning_rate": 1.9639112177564486e-06, "loss": 0.0002, "step": 14443 }, { "epoch": 1.73, "learning_rate": 1.9638392321535694e-06, "loss": 0.0003, "step": 14444 }, { "epoch": 1.73, "learning_rate": 1.9637672465506897e-06, "loss": 0.0004, "step": 14445 }, { "epoch": 1.73, "learning_rate": 1.9636952609478105e-06, "loss": 0.0003, "step": 14446 }, { "epoch": 1.73, "learning_rate": 1.963623275344931e-06, "loss": 0.0004, "step": 14447 }, { "epoch": 1.73, "learning_rate": 1.9635512897420516e-06, "loss": 0.0003, "step": 14448 }, { "epoch": 1.73, "learning_rate": 1.9634793041391724e-06, "loss": 0.0003, "step": 14449 }, { "epoch": 1.73, "learning_rate": 1.9634073185362928e-06, "loss": 0.0003, "step": 14450 }, { "epoch": 1.73, "learning_rate": 1.9633353329334135e-06, "loss": 0.8666, "step": 14451 }, { "epoch": 1.73, "learning_rate": 1.963263347330534e-06, "loss": 0.0002, "step": 14452 }, { "epoch": 1.73, "learning_rate": 1.9631913617276547e-06, "loss": 0.0013, "step": 14453 }, { "epoch": 1.73, "learning_rate": 1.963119376124775e-06, "loss": 0.0004, "step": 14454 }, { "epoch": 1.73, "learning_rate": 1.963047390521896e-06, "loss": 0.0004, "step": 14455 }, { "epoch": 1.73, "learning_rate": 1.962975404919016e-06, "loss": 0.207, "step": 14456 }, { "epoch": 1.73, "learning_rate": 1.962903419316137e-06, "loss": 0.0002, "step": 14457 }, { "epoch": 1.73, "learning_rate": 1.9628314337132573e-06, "loss": 0.0007, "step": 14458 }, { "epoch": 1.73, "learning_rate": 1.962759448110378e-06, "loss": 0.0003, "step": 14459 }, { "epoch": 1.73, "learning_rate": 1.9626874625074984e-06, "loss": 0.0004, "step": 14460 }, { "epoch": 1.73, "learning_rate": 1.962615476904619e-06, "loss": 0.3495, "step": 14461 }, { "epoch": 1.73, "learning_rate": 1.9625434913017396e-06, "loss": 0.002, "step": 14462 }, { "epoch": 1.73, "learning_rate": 1.9624715056988603e-06, "loss": 0.0004, "step": 14463 }, { "epoch": 1.73, "learning_rate": 1.9623995200959807e-06, "loss": 0.0014, "step": 14464 }, { "epoch": 1.73, "learning_rate": 1.9623275344931015e-06, "loss": 0.001, "step": 14465 }, { "epoch": 1.73, "learning_rate": 1.9622555488902223e-06, "loss": 0.0003, "step": 14466 }, { "epoch": 1.73, "learning_rate": 1.9621835632873426e-06, "loss": 0.0002, "step": 14467 }, { "epoch": 1.73, "learning_rate": 1.9621115776844634e-06, "loss": 0.0003, "step": 14468 }, { "epoch": 1.73, "learning_rate": 1.9620395920815837e-06, "loss": 0.0003, "step": 14469 }, { "epoch": 1.73, "learning_rate": 1.9619676064787045e-06, "loss": 0.6294, "step": 14470 }, { "epoch": 1.73, "learning_rate": 1.961895620875825e-06, "loss": 0.0003, "step": 14471 }, { "epoch": 1.73, "learning_rate": 1.9618236352729457e-06, "loss": 0.0002, "step": 14472 }, { "epoch": 1.73, "learning_rate": 1.961751649670066e-06, "loss": 0.0005, "step": 14473 }, { "epoch": 1.73, "learning_rate": 1.9616796640671864e-06, "loss": 0.0006, "step": 14474 }, { "epoch": 1.73, "learning_rate": 1.961607678464307e-06, "loss": 0.0004, "step": 14475 }, { "epoch": 1.73, "learning_rate": 1.9615356928614275e-06, "loss": 0.6816, "step": 14476 }, { "epoch": 1.73, "learning_rate": 1.9614637072585483e-06, "loss": 0.0003, "step": 14477 }, { "epoch": 1.73, "learning_rate": 1.9613917216556686e-06, "loss": 0.0002, "step": 14478 }, { "epoch": 1.74, "learning_rate": 1.9613197360527894e-06, "loss": 0.0003, "step": 14479 }, { "epoch": 1.74, "learning_rate": 1.96124775044991e-06, "loss": 0.0006, "step": 14480 }, { "epoch": 1.74, "learning_rate": 1.961175764847031e-06, "loss": 0.0003, "step": 14481 }, { "epoch": 1.74, "learning_rate": 1.9611037792441513e-06, "loss": 0.0006, "step": 14482 }, { "epoch": 1.74, "learning_rate": 1.961031793641272e-06, "loss": 0.0001, "step": 14483 }, { "epoch": 1.74, "learning_rate": 1.9609598080383925e-06, "loss": 0.0004, "step": 14484 }, { "epoch": 1.74, "learning_rate": 1.960887822435513e-06, "loss": 0.0002, "step": 14485 }, { "epoch": 1.74, "learning_rate": 1.9608158368326336e-06, "loss": 0.0005, "step": 14486 }, { "epoch": 1.74, "learning_rate": 1.960743851229754e-06, "loss": 0.0006, "step": 14487 }, { "epoch": 1.74, "learning_rate": 1.9606718656268747e-06, "loss": 0.0006, "step": 14488 }, { "epoch": 1.74, "learning_rate": 1.960599880023995e-06, "loss": 0.0002, "step": 14489 }, { "epoch": 1.74, "learning_rate": 1.960527894421116e-06, "loss": 0.0006, "step": 14490 }, { "epoch": 1.74, "learning_rate": 1.960455908818236e-06, "loss": 0.0006, "step": 14491 }, { "epoch": 1.74, "learning_rate": 1.960383923215357e-06, "loss": 0.0209, "step": 14492 }, { "epoch": 1.74, "learning_rate": 1.9603119376124773e-06, "loss": 0.0012, "step": 14493 }, { "epoch": 1.74, "learning_rate": 1.960239952009598e-06, "loss": 0.0003, "step": 14494 }, { "epoch": 1.74, "learning_rate": 1.960167966406719e-06, "loss": 0.0006, "step": 14495 }, { "epoch": 1.74, "learning_rate": 1.9600959808038393e-06, "loss": 0.0005, "step": 14496 }, { "epoch": 1.74, "learning_rate": 1.96002399520096e-06, "loss": 0.0004, "step": 14497 }, { "epoch": 1.74, "learning_rate": 1.9599520095980804e-06, "loss": 0.0006, "step": 14498 }, { "epoch": 1.74, "learning_rate": 1.959880023995201e-06, "loss": 0.4419, "step": 14499 }, { "epoch": 1.74, "learning_rate": 1.9598080383923215e-06, "loss": 0.0004, "step": 14500 }, { "epoch": 1.74, "learning_rate": 1.9597360527894423e-06, "loss": 0.0004, "step": 14501 }, { "epoch": 1.74, "learning_rate": 1.9596640671865627e-06, "loss": 0.0005, "step": 14502 }, { "epoch": 1.74, "learning_rate": 1.9595920815836834e-06, "loss": 0.0002, "step": 14503 }, { "epoch": 1.74, "learning_rate": 1.959520095980804e-06, "loss": 0.0133, "step": 14504 }, { "epoch": 1.74, "learning_rate": 1.9594481103779246e-06, "loss": 0.0006, "step": 14505 }, { "epoch": 1.74, "learning_rate": 1.959376124775045e-06, "loss": 0.0005, "step": 14506 }, { "epoch": 1.74, "learning_rate": 1.9593041391721657e-06, "loss": 0.087, "step": 14507 }, { "epoch": 1.74, "learning_rate": 1.959232153569286e-06, "loss": 0.0014, "step": 14508 }, { "epoch": 1.74, "learning_rate": 1.959160167966407e-06, "loss": 0.0028, "step": 14509 }, { "epoch": 1.74, "learning_rate": 1.959088182363527e-06, "loss": 0.0002, "step": 14510 }, { "epoch": 1.74, "learning_rate": 1.959016196760648e-06, "loss": 0.0004, "step": 14511 }, { "epoch": 1.74, "learning_rate": 1.9589442111577687e-06, "loss": 0.0005, "step": 14512 }, { "epoch": 1.74, "learning_rate": 1.958872225554889e-06, "loss": 0.0003, "step": 14513 }, { "epoch": 1.74, "learning_rate": 1.95880023995201e-06, "loss": 0.0003, "step": 14514 }, { "epoch": 1.74, "learning_rate": 1.9587282543491302e-06, "loss": 0.0031, "step": 14515 }, { "epoch": 1.74, "learning_rate": 1.958656268746251e-06, "loss": 0.0004, "step": 14516 }, { "epoch": 1.74, "learning_rate": 1.9585842831433714e-06, "loss": 0.0002, "step": 14517 }, { "epoch": 1.74, "learning_rate": 1.958512297540492e-06, "loss": 0.0007, "step": 14518 }, { "epoch": 1.74, "learning_rate": 1.9584403119376125e-06, "loss": 0.0004, "step": 14519 }, { "epoch": 1.74, "learning_rate": 1.958368326334733e-06, "loss": 0.0002, "step": 14520 }, { "epoch": 1.74, "learning_rate": 1.9582963407318536e-06, "loss": 0.0003, "step": 14521 }, { "epoch": 1.74, "learning_rate": 1.958224355128974e-06, "loss": 0.9625, "step": 14522 }, { "epoch": 1.74, "learning_rate": 1.9581523695260948e-06, "loss": 0.0005, "step": 14523 }, { "epoch": 1.74, "learning_rate": 1.958080383923215e-06, "loss": 0.002, "step": 14524 }, { "epoch": 1.74, "learning_rate": 1.958008398320336e-06, "loss": 0.0004, "step": 14525 }, { "epoch": 1.74, "learning_rate": 1.9579364127174567e-06, "loss": 0.0002, "step": 14526 }, { "epoch": 1.74, "learning_rate": 1.9578644271145775e-06, "loss": 0.0005, "step": 14527 }, { "epoch": 1.74, "learning_rate": 1.957792441511698e-06, "loss": 0.0004, "step": 14528 }, { "epoch": 1.74, "learning_rate": 1.9577204559088186e-06, "loss": 0.0003, "step": 14529 }, { "epoch": 1.74, "learning_rate": 1.957648470305939e-06, "loss": 0.0002, "step": 14530 }, { "epoch": 1.74, "learning_rate": 1.9575764847030593e-06, "loss": 0.0002, "step": 14531 }, { "epoch": 1.74, "learning_rate": 1.95750449910018e-06, "loss": 0.0003, "step": 14532 }, { "epoch": 1.74, "learning_rate": 1.9574325134973004e-06, "loss": 0.0001, "step": 14533 }, { "epoch": 1.74, "learning_rate": 1.9573605278944212e-06, "loss": 0.0003, "step": 14534 }, { "epoch": 1.74, "learning_rate": 1.9572885422915416e-06, "loss": 0.0001, "step": 14535 }, { "epoch": 1.74, "learning_rate": 1.9572165566886623e-06, "loss": 0.0003, "step": 14536 }, { "epoch": 1.74, "learning_rate": 1.9571445710857827e-06, "loss": 0.0002, "step": 14537 }, { "epoch": 1.74, "learning_rate": 1.9570725854829035e-06, "loss": 0.9664, "step": 14538 }, { "epoch": 1.74, "learning_rate": 1.957000599880024e-06, "loss": 0.0001, "step": 14539 }, { "epoch": 1.74, "learning_rate": 1.9569286142771446e-06, "loss": 0.0005, "step": 14540 }, { "epoch": 1.74, "learning_rate": 1.956856628674265e-06, "loss": 0.0002, "step": 14541 }, { "epoch": 1.74, "learning_rate": 1.9567846430713858e-06, "loss": 0.0003, "step": 14542 }, { "epoch": 1.74, "learning_rate": 1.9567126574685065e-06, "loss": 0.0003, "step": 14543 }, { "epoch": 1.74, "learning_rate": 1.956640671865627e-06, "loss": 0.0002, "step": 14544 }, { "epoch": 1.74, "learning_rate": 1.9565686862627477e-06, "loss": 0.0002, "step": 14545 }, { "epoch": 1.74, "learning_rate": 1.956496700659868e-06, "loss": 0.0005, "step": 14546 }, { "epoch": 1.74, "learning_rate": 1.956424715056989e-06, "loss": 0.0002, "step": 14547 }, { "epoch": 1.74, "learning_rate": 1.956352729454109e-06, "loss": 0.0187, "step": 14548 }, { "epoch": 1.74, "learning_rate": 1.95628074385123e-06, "loss": 0.0002, "step": 14549 }, { "epoch": 1.74, "learning_rate": 1.9562087582483503e-06, "loss": 0.0001, "step": 14550 }, { "epoch": 1.74, "learning_rate": 1.956136772645471e-06, "loss": 0.0007, "step": 14551 }, { "epoch": 1.74, "learning_rate": 1.9560647870425914e-06, "loss": 0.0002, "step": 14552 }, { "epoch": 1.74, "learning_rate": 1.955992801439712e-06, "loss": 0.003, "step": 14553 }, { "epoch": 1.74, "learning_rate": 1.9559208158368326e-06, "loss": 0.0002, "step": 14554 }, { "epoch": 1.74, "learning_rate": 1.955848830233953e-06, "loss": 0.0007, "step": 14555 }, { "epoch": 1.74, "learning_rate": 1.9557768446310737e-06, "loss": 0.0002, "step": 14556 }, { "epoch": 1.74, "learning_rate": 1.9557048590281945e-06, "loss": 0.0002, "step": 14557 }, { "epoch": 1.74, "learning_rate": 1.9556328734253152e-06, "loss": 0.0001, "step": 14558 }, { "epoch": 1.74, "learning_rate": 1.9555608878224356e-06, "loss": 0.0005, "step": 14559 }, { "epoch": 1.74, "learning_rate": 1.9554889022195564e-06, "loss": 0.0004, "step": 14560 }, { "epoch": 1.74, "learning_rate": 1.9554169166166767e-06, "loss": 0.0003, "step": 14561 }, { "epoch": 1.74, "learning_rate": 1.9553449310137975e-06, "loss": 0.0003, "step": 14562 }, { "epoch": 1.75, "learning_rate": 1.955272945410918e-06, "loss": 0.0003, "step": 14563 }, { "epoch": 1.75, "learning_rate": 1.9552009598080386e-06, "loss": 0.0002, "step": 14564 }, { "epoch": 1.75, "learning_rate": 1.955128974205159e-06, "loss": 0.0009, "step": 14565 }, { "epoch": 1.75, "learning_rate": 1.9550569886022794e-06, "loss": 0.0002, "step": 14566 }, { "epoch": 1.75, "learning_rate": 1.9549850029994e-06, "loss": 0.0002, "step": 14567 }, { "epoch": 1.75, "learning_rate": 1.9549130173965205e-06, "loss": 0.0003, "step": 14568 }, { "epoch": 1.75, "learning_rate": 1.9548410317936413e-06, "loss": 0.0001, "step": 14569 }, { "epoch": 1.75, "learning_rate": 1.9547690461907616e-06, "loss": 0.0002, "step": 14570 }, { "epoch": 1.75, "learning_rate": 1.9546970605878824e-06, "loss": 0.0002, "step": 14571 }, { "epoch": 1.75, "learning_rate": 1.954625074985003e-06, "loss": 0.0002, "step": 14572 }, { "epoch": 1.75, "learning_rate": 1.9545530893821235e-06, "loss": 0.3234, "step": 14573 }, { "epoch": 1.75, "learning_rate": 1.9544811037792443e-06, "loss": 0.0003, "step": 14574 }, { "epoch": 1.75, "learning_rate": 1.954409118176365e-06, "loss": 0.0003, "step": 14575 }, { "epoch": 1.75, "learning_rate": 1.9543371325734854e-06, "loss": 0.0002, "step": 14576 }, { "epoch": 1.75, "learning_rate": 1.954265146970606e-06, "loss": 0.0003, "step": 14577 }, { "epoch": 1.75, "learning_rate": 1.9541931613677266e-06, "loss": 0.0002, "step": 14578 }, { "epoch": 1.75, "learning_rate": 1.954121175764847e-06, "loss": 0.0003, "step": 14579 }, { "epoch": 1.75, "learning_rate": 1.9540491901619677e-06, "loss": 0.0004, "step": 14580 }, { "epoch": 1.75, "learning_rate": 1.953977204559088e-06, "loss": 0.0003, "step": 14581 }, { "epoch": 1.75, "learning_rate": 1.953905218956209e-06, "loss": 0.0001, "step": 14582 }, { "epoch": 1.75, "learning_rate": 1.953833233353329e-06, "loss": 0.0003, "step": 14583 }, { "epoch": 1.75, "learning_rate": 1.95376124775045e-06, "loss": 0.0003, "step": 14584 }, { "epoch": 1.75, "learning_rate": 1.9536892621475703e-06, "loss": 0.0003, "step": 14585 }, { "epoch": 1.75, "learning_rate": 1.953617276544691e-06, "loss": 0.0098, "step": 14586 }, { "epoch": 1.75, "learning_rate": 1.9535452909418115e-06, "loss": 0.0005, "step": 14587 }, { "epoch": 1.75, "learning_rate": 1.9534733053389322e-06, "loss": 0.0005, "step": 14588 }, { "epoch": 1.75, "learning_rate": 1.953401319736053e-06, "loss": 0.0004, "step": 14589 }, { "epoch": 1.75, "learning_rate": 1.9533293341331734e-06, "loss": 0.0002, "step": 14590 }, { "epoch": 1.75, "learning_rate": 1.953257348530294e-06, "loss": 0.0003, "step": 14591 }, { "epoch": 1.75, "learning_rate": 1.9531853629274145e-06, "loss": 0.0417, "step": 14592 }, { "epoch": 1.75, "learning_rate": 1.9531133773245353e-06, "loss": 0.0006, "step": 14593 }, { "epoch": 1.75, "learning_rate": 1.9530413917216556e-06, "loss": 0.0002, "step": 14594 }, { "epoch": 1.75, "learning_rate": 1.9529694061187764e-06, "loss": 0.0024, "step": 14595 }, { "epoch": 1.75, "learning_rate": 1.9528974205158968e-06, "loss": 0.0004, "step": 14596 }, { "epoch": 1.75, "learning_rate": 1.9528254349130176e-06, "loss": 0.0002, "step": 14597 }, { "epoch": 1.75, "learning_rate": 1.952753449310138e-06, "loss": 0.0002, "step": 14598 }, { "epoch": 1.75, "learning_rate": 1.9526814637072587e-06, "loss": 0.0002, "step": 14599 }, { "epoch": 1.75, "learning_rate": 1.952609478104379e-06, "loss": 0.0004, "step": 14600 }, { "epoch": 1.75, "learning_rate": 1.9525374925014994e-06, "loss": 0.0002, "step": 14601 }, { "epoch": 1.75, "learning_rate": 1.95246550689862e-06, "loss": 0.0003, "step": 14602 }, { "epoch": 1.75, "learning_rate": 1.952393521295741e-06, "loss": 0.003, "step": 14603 }, { "epoch": 1.75, "learning_rate": 1.9523215356928617e-06, "loss": 0.0003, "step": 14604 }, { "epoch": 1.75, "learning_rate": 1.952249550089982e-06, "loss": 0.0002, "step": 14605 }, { "epoch": 1.75, "learning_rate": 1.952177564487103e-06, "loss": 0.0004, "step": 14606 }, { "epoch": 1.75, "learning_rate": 1.9521055788842232e-06, "loss": 0.0001, "step": 14607 }, { "epoch": 1.75, "learning_rate": 1.952033593281344e-06, "loss": 0.0003, "step": 14608 }, { "epoch": 1.75, "learning_rate": 1.9519616076784644e-06, "loss": 0.0003, "step": 14609 }, { "epoch": 1.75, "learning_rate": 1.951889622075585e-06, "loss": 0.0002, "step": 14610 }, { "epoch": 1.75, "learning_rate": 1.9518176364727055e-06, "loss": 0.0033, "step": 14611 }, { "epoch": 1.75, "learning_rate": 1.951745650869826e-06, "loss": 0.0003, "step": 14612 }, { "epoch": 1.75, "learning_rate": 1.9516736652669466e-06, "loss": 0.0002, "step": 14613 }, { "epoch": 1.75, "learning_rate": 1.951601679664067e-06, "loss": 0.0002, "step": 14614 }, { "epoch": 1.75, "learning_rate": 1.9515296940611878e-06, "loss": 0.0002, "step": 14615 }, { "epoch": 1.75, "learning_rate": 1.951457708458308e-06, "loss": 0.0003, "step": 14616 }, { "epoch": 1.75, "learning_rate": 1.951385722855429e-06, "loss": 0.0003, "step": 14617 }, { "epoch": 1.75, "learning_rate": 1.9513137372525497e-06, "loss": 0.0002, "step": 14618 }, { "epoch": 1.75, "learning_rate": 1.95124175164967e-06, "loss": 0.0002, "step": 14619 }, { "epoch": 1.75, "learning_rate": 1.951169766046791e-06, "loss": 0.0005, "step": 14620 }, { "epoch": 1.75, "learning_rate": 1.9510977804439116e-06, "loss": 0.0002, "step": 14621 }, { "epoch": 1.75, "learning_rate": 1.951025794841032e-06, "loss": 0.0004, "step": 14622 }, { "epoch": 1.75, "learning_rate": 1.9509538092381523e-06, "loss": 0.0001, "step": 14623 }, { "epoch": 1.75, "learning_rate": 1.950881823635273e-06, "loss": 0.8134, "step": 14624 }, { "epoch": 1.75, "learning_rate": 1.9508098380323934e-06, "loss": 0.0003, "step": 14625 }, { "epoch": 1.75, "learning_rate": 1.950737852429514e-06, "loss": 0.0002, "step": 14626 }, { "epoch": 1.75, "learning_rate": 1.9506658668266346e-06, "loss": 0.0002, "step": 14627 }, { "epoch": 1.75, "learning_rate": 1.9505938812237553e-06, "loss": 0.0003, "step": 14628 }, { "epoch": 1.75, "learning_rate": 1.9505218956208757e-06, "loss": 0.9175, "step": 14629 }, { "epoch": 1.75, "learning_rate": 1.9504499100179965e-06, "loss": 0.0002, "step": 14630 }, { "epoch": 1.75, "learning_rate": 1.950377924415117e-06, "loss": 0.0025, "step": 14631 }, { "epoch": 1.75, "learning_rate": 1.9503059388122376e-06, "loss": 0.0002, "step": 14632 }, { "epoch": 1.75, "learning_rate": 1.950233953209358e-06, "loss": 0.0003, "step": 14633 }, { "epoch": 1.75, "learning_rate": 1.9501619676064787e-06, "loss": 0.0003, "step": 14634 }, { "epoch": 1.75, "learning_rate": 1.9500899820035995e-06, "loss": 0.0003, "step": 14635 }, { "epoch": 1.75, "learning_rate": 1.95001799640072e-06, "loss": 0.0002, "step": 14636 }, { "epoch": 1.75, "learning_rate": 1.9499460107978407e-06, "loss": 0.0002, "step": 14637 }, { "epoch": 1.75, "learning_rate": 1.949874025194961e-06, "loss": 0.001, "step": 14638 }, { "epoch": 1.75, "learning_rate": 1.9498020395920818e-06, "loss": 0.001, "step": 14639 }, { "epoch": 1.75, "learning_rate": 1.949730053989202e-06, "loss": 0.0002, "step": 14640 }, { "epoch": 1.75, "learning_rate": 1.949658068386323e-06, "loss": 0.0003, "step": 14641 }, { "epoch": 1.75, "learning_rate": 1.9495860827834433e-06, "loss": 0.0002, "step": 14642 }, { "epoch": 1.75, "learning_rate": 1.949514097180564e-06, "loss": 0.0002, "step": 14643 }, { "epoch": 1.75, "learning_rate": 1.9494421115776844e-06, "loss": 0.0002, "step": 14644 }, { "epoch": 1.75, "learning_rate": 1.949370125974805e-06, "loss": 0.0001, "step": 14645 }, { "epoch": 1.76, "learning_rate": 1.9492981403719255e-06, "loss": 0.0002, "step": 14646 }, { "epoch": 1.76, "learning_rate": 1.949226154769046e-06, "loss": 0.0002, "step": 14647 }, { "epoch": 1.76, "learning_rate": 1.9491541691661667e-06, "loss": 0.0002, "step": 14648 }, { "epoch": 1.76, "learning_rate": 1.9490821835632875e-06, "loss": 0.0003, "step": 14649 }, { "epoch": 1.76, "learning_rate": 1.9490101979604082e-06, "loss": 0.0003, "step": 14650 }, { "epoch": 1.76, "learning_rate": 1.9489382123575286e-06, "loss": 0.0003, "step": 14651 }, { "epoch": 1.76, "learning_rate": 1.9488662267546494e-06, "loss": 0.0003, "step": 14652 }, { "epoch": 1.76, "learning_rate": 1.9487942411517697e-06, "loss": 0.0005, "step": 14653 }, { "epoch": 1.76, "learning_rate": 1.9487222555488905e-06, "loss": 0.0003, "step": 14654 }, { "epoch": 1.76, "learning_rate": 1.948650269946011e-06, "loss": 0.0002, "step": 14655 }, { "epoch": 1.76, "learning_rate": 1.9485782843431316e-06, "loss": 0.0002, "step": 14656 }, { "epoch": 1.76, "learning_rate": 1.948506298740252e-06, "loss": 0.0001, "step": 14657 }, { "epoch": 1.76, "learning_rate": 1.9484343131373723e-06, "loss": 0.0003, "step": 14658 }, { "epoch": 1.76, "learning_rate": 1.948362327534493e-06, "loss": 0.0002, "step": 14659 }, { "epoch": 1.76, "learning_rate": 1.9482903419316135e-06, "loss": 0.0003, "step": 14660 }, { "epoch": 1.76, "learning_rate": 1.9482183563287343e-06, "loss": 0.0003, "step": 14661 }, { "epoch": 1.76, "learning_rate": 1.9481463707258546e-06, "loss": 0.0003, "step": 14662 }, { "epoch": 1.76, "learning_rate": 1.9480743851229754e-06, "loss": 0.0009, "step": 14663 }, { "epoch": 1.76, "learning_rate": 1.9480023995200957e-06, "loss": 0.104, "step": 14664 }, { "epoch": 1.76, "learning_rate": 1.9479304139172165e-06, "loss": 0.0003, "step": 14665 }, { "epoch": 1.76, "learning_rate": 1.9478584283143373e-06, "loss": 0.0001, "step": 14666 }, { "epoch": 1.76, "learning_rate": 1.947786442711458e-06, "loss": 0.0004, "step": 14667 }, { "epoch": 1.76, "learning_rate": 1.9477144571085784e-06, "loss": 0.0004, "step": 14668 }, { "epoch": 1.76, "learning_rate": 1.9476424715056992e-06, "loss": 0.0003, "step": 14669 }, { "epoch": 1.76, "learning_rate": 1.9475704859028196e-06, "loss": 0.0008, "step": 14670 }, { "epoch": 1.76, "learning_rate": 1.94749850029994e-06, "loss": 0.0003, "step": 14671 }, { "epoch": 1.76, "learning_rate": 1.9474265146970607e-06, "loss": 0.0002, "step": 14672 }, { "epoch": 1.76, "learning_rate": 1.947354529094181e-06, "loss": 0.0002, "step": 14673 }, { "epoch": 1.76, "learning_rate": 1.947282543491302e-06, "loss": 0.0002, "step": 14674 }, { "epoch": 1.76, "learning_rate": 1.947210557888422e-06, "loss": 0.0002, "step": 14675 }, { "epoch": 1.76, "learning_rate": 1.947138572285543e-06, "loss": 0.0002, "step": 14676 }, { "epoch": 1.76, "learning_rate": 1.9470665866826633e-06, "loss": 0.0002, "step": 14677 }, { "epoch": 1.76, "learning_rate": 1.946994601079784e-06, "loss": 0.7544, "step": 14678 }, { "epoch": 1.76, "learning_rate": 1.9469226154769045e-06, "loss": 0.0001, "step": 14679 }, { "epoch": 1.76, "learning_rate": 1.9468506298740252e-06, "loss": 0.001, "step": 14680 }, { "epoch": 1.76, "learning_rate": 1.946778644271146e-06, "loss": 0.0008, "step": 14681 }, { "epoch": 1.76, "learning_rate": 1.9467066586682664e-06, "loss": 0.8118, "step": 14682 }, { "epoch": 1.76, "learning_rate": 1.946634673065387e-06, "loss": 0.0003, "step": 14683 }, { "epoch": 1.76, "learning_rate": 1.9465626874625075e-06, "loss": 0.0015, "step": 14684 }, { "epoch": 1.76, "learning_rate": 1.9464907018596283e-06, "loss": 0.0004, "step": 14685 }, { "epoch": 1.76, "learning_rate": 1.9464187162567486e-06, "loss": 0.0001, "step": 14686 }, { "epoch": 1.76, "learning_rate": 1.9463467306538694e-06, "loss": 0.0003, "step": 14687 }, { "epoch": 1.76, "learning_rate": 1.9462747450509898e-06, "loss": 0.0003, "step": 14688 }, { "epoch": 1.76, "learning_rate": 1.9462027594481106e-06, "loss": 0.0003, "step": 14689 }, { "epoch": 1.76, "learning_rate": 1.946130773845231e-06, "loss": 0.0002, "step": 14690 }, { "epoch": 1.76, "learning_rate": 1.9460587882423517e-06, "loss": 0.0003, "step": 14691 }, { "epoch": 1.76, "learning_rate": 1.945986802639472e-06, "loss": 0.0002, "step": 14692 }, { "epoch": 1.76, "learning_rate": 1.9459148170365924e-06, "loss": 0.0002, "step": 14693 }, { "epoch": 1.76, "learning_rate": 1.945842831433713e-06, "loss": 0.004, "step": 14694 }, { "epoch": 1.76, "learning_rate": 1.945770845830834e-06, "loss": 0.0003, "step": 14695 }, { "epoch": 1.76, "learning_rate": 1.9456988602279543e-06, "loss": 0.0002, "step": 14696 }, { "epoch": 1.76, "learning_rate": 1.945626874625075e-06, "loss": 0.0003, "step": 14697 }, { "epoch": 1.76, "learning_rate": 1.945554889022196e-06, "loss": 0.0003, "step": 14698 }, { "epoch": 1.76, "learning_rate": 1.9454829034193162e-06, "loss": 0.0011, "step": 14699 }, { "epoch": 1.76, "learning_rate": 1.945410917816437e-06, "loss": 0.0002, "step": 14700 }, { "epoch": 1.76, "learning_rate": 1.9453389322135574e-06, "loss": 0.0002, "step": 14701 }, { "epoch": 1.76, "learning_rate": 1.945266946610678e-06, "loss": 0.0003, "step": 14702 }, { "epoch": 1.76, "learning_rate": 1.9451949610077985e-06, "loss": 0.001, "step": 14703 }, { "epoch": 1.76, "learning_rate": 1.945122975404919e-06, "loss": 0.0001, "step": 14704 }, { "epoch": 1.76, "learning_rate": 1.9450509898020396e-06, "loss": 0.0002, "step": 14705 }, { "epoch": 1.76, "learning_rate": 1.94497900419916e-06, "loss": 0.0005, "step": 14706 }, { "epoch": 1.76, "learning_rate": 1.9449070185962808e-06, "loss": 0.0006, "step": 14707 }, { "epoch": 1.76, "learning_rate": 1.944835032993401e-06, "loss": 0.0015, "step": 14708 }, { "epoch": 1.76, "learning_rate": 1.944763047390522e-06, "loss": 0.9646, "step": 14709 }, { "epoch": 1.76, "learning_rate": 1.9446910617876422e-06, "loss": 0.0001, "step": 14710 }, { "epoch": 1.76, "learning_rate": 1.944619076184763e-06, "loss": 0.0003, "step": 14711 }, { "epoch": 1.76, "learning_rate": 1.944547090581884e-06, "loss": 0.0003, "step": 14712 }, { "epoch": 1.76, "learning_rate": 1.9444751049790046e-06, "loss": 0.8871, "step": 14713 }, { "epoch": 1.76, "learning_rate": 1.944403119376125e-06, "loss": 0.0006, "step": 14714 }, { "epoch": 1.76, "learning_rate": 1.9443311337732457e-06, "loss": 0.0001, "step": 14715 }, { "epoch": 1.76, "learning_rate": 1.944259148170366e-06, "loss": 0.0003, "step": 14716 }, { "epoch": 1.76, "learning_rate": 1.9441871625674864e-06, "loss": 0.0001, "step": 14717 }, { "epoch": 1.76, "learning_rate": 1.944115176964607e-06, "loss": 0.0002, "step": 14718 }, { "epoch": 1.76, "learning_rate": 1.9440431913617276e-06, "loss": 0.0002, "step": 14719 }, { "epoch": 1.76, "learning_rate": 1.9439712057588483e-06, "loss": 0.0002, "step": 14720 }, { "epoch": 1.76, "learning_rate": 1.9438992201559687e-06, "loss": 0.0003, "step": 14721 }, { "epoch": 1.76, "learning_rate": 1.9438272345530895e-06, "loss": 0.0002, "step": 14722 }, { "epoch": 1.76, "learning_rate": 1.94375524895021e-06, "loss": 0.0002, "step": 14723 }, { "epoch": 1.76, "learning_rate": 1.9436832633473306e-06, "loss": 0.0004, "step": 14724 }, { "epoch": 1.76, "learning_rate": 1.943611277744451e-06, "loss": 0.0002, "step": 14725 }, { "epoch": 1.76, "learning_rate": 1.9435392921415717e-06, "loss": 0.4115, "step": 14726 }, { "epoch": 1.76, "learning_rate": 1.9434673065386925e-06, "loss": 0.0003, "step": 14727 }, { "epoch": 1.76, "learning_rate": 1.943395320935813e-06, "loss": 0.0002, "step": 14728 }, { "epoch": 1.77, "learning_rate": 1.9433233353329336e-06, "loss": 0.8998, "step": 14729 }, { "epoch": 1.77, "learning_rate": 1.943251349730054e-06, "loss": 0.0002, "step": 14730 }, { "epoch": 1.77, "learning_rate": 1.9431793641271748e-06, "loss": 0.0002, "step": 14731 }, { "epoch": 1.77, "learning_rate": 1.943107378524295e-06, "loss": 0.0002, "step": 14732 }, { "epoch": 1.77, "learning_rate": 1.943035392921416e-06, "loss": 0.1349, "step": 14733 }, { "epoch": 1.77, "learning_rate": 1.9429634073185363e-06, "loss": 0.0002, "step": 14734 }, { "epoch": 1.77, "learning_rate": 1.942891421715657e-06, "loss": 0.0002, "step": 14735 }, { "epoch": 1.77, "learning_rate": 1.9428194361127774e-06, "loss": 0.0002, "step": 14736 }, { "epoch": 1.77, "learning_rate": 1.942747450509898e-06, "loss": 0.0005, "step": 14737 }, { "epoch": 1.77, "learning_rate": 1.9426754649070185e-06, "loss": 0.0003, "step": 14738 }, { "epoch": 1.77, "learning_rate": 1.942603479304139e-06, "loss": 0.0012, "step": 14739 }, { "epoch": 1.77, "learning_rate": 1.9425314937012597e-06, "loss": 0.0004, "step": 14740 }, { "epoch": 1.77, "learning_rate": 1.9424595080983804e-06, "loss": 0.0003, "step": 14741 }, { "epoch": 1.77, "learning_rate": 1.942387522495501e-06, "loss": 0.0001, "step": 14742 }, { "epoch": 1.77, "learning_rate": 1.9423155368926216e-06, "loss": 0.0002, "step": 14743 }, { "epoch": 1.77, "learning_rate": 1.9422435512897424e-06, "loss": 0.0002, "step": 14744 }, { "epoch": 1.77, "learning_rate": 1.9421715656868627e-06, "loss": 0.0003, "step": 14745 }, { "epoch": 1.77, "learning_rate": 1.9420995800839835e-06, "loss": 0.0005, "step": 14746 }, { "epoch": 1.77, "learning_rate": 1.942027594481104e-06, "loss": 0.0002, "step": 14747 }, { "epoch": 1.77, "learning_rate": 1.9419556088782246e-06, "loss": 0.0003, "step": 14748 }, { "epoch": 1.77, "learning_rate": 1.941883623275345e-06, "loss": 0.0002, "step": 14749 }, { "epoch": 1.77, "learning_rate": 1.9418116376724653e-06, "loss": 0.0002, "step": 14750 }, { "epoch": 1.77, "learning_rate": 1.941739652069586e-06, "loss": 0.0004, "step": 14751 }, { "epoch": 1.77, "learning_rate": 1.9416676664667065e-06, "loss": 0.0002, "step": 14752 }, { "epoch": 1.77, "learning_rate": 1.9415956808638272e-06, "loss": 0.0001, "step": 14753 }, { "epoch": 1.77, "learning_rate": 1.9415236952609476e-06, "loss": 0.0008, "step": 14754 }, { "epoch": 1.77, "learning_rate": 1.9414517096580684e-06, "loss": 0.0002, "step": 14755 }, { "epoch": 1.77, "learning_rate": 1.9413797240551887e-06, "loss": 0.0003, "step": 14756 }, { "epoch": 1.77, "learning_rate": 1.9413077384523095e-06, "loss": 0.0002, "step": 14757 }, { "epoch": 1.77, "learning_rate": 1.9412357528494303e-06, "loss": 0.0003, "step": 14758 }, { "epoch": 1.77, "learning_rate": 1.941163767246551e-06, "loss": 0.0001, "step": 14759 }, { "epoch": 1.77, "learning_rate": 1.9410917816436714e-06, "loss": 0.0002, "step": 14760 }, { "epoch": 1.77, "learning_rate": 1.941019796040792e-06, "loss": 0.0003, "step": 14761 }, { "epoch": 1.77, "learning_rate": 1.9409478104379126e-06, "loss": 0.0002, "step": 14762 }, { "epoch": 1.77, "learning_rate": 1.940875824835033e-06, "loss": 0.0003, "step": 14763 }, { "epoch": 1.77, "learning_rate": 1.9408038392321537e-06, "loss": 0.0002, "step": 14764 }, { "epoch": 1.77, "learning_rate": 1.940731853629274e-06, "loss": 0.0004, "step": 14765 }, { "epoch": 1.77, "learning_rate": 1.940659868026395e-06, "loss": 0.0002, "step": 14766 }, { "epoch": 1.77, "learning_rate": 1.940587882423515e-06, "loss": 0.0009, "step": 14767 }, { "epoch": 1.77, "learning_rate": 1.940515896820636e-06, "loss": 0.0003, "step": 14768 }, { "epoch": 1.77, "learning_rate": 1.9404439112177563e-06, "loss": 0.0002, "step": 14769 }, { "epoch": 1.77, "learning_rate": 1.940371925614877e-06, "loss": 0.0005, "step": 14770 }, { "epoch": 1.77, "learning_rate": 1.9402999400119975e-06, "loss": 0.0002, "step": 14771 }, { "epoch": 1.77, "learning_rate": 1.9402279544091182e-06, "loss": 0.0002, "step": 14772 }, { "epoch": 1.77, "learning_rate": 1.940155968806239e-06, "loss": 0.0004, "step": 14773 }, { "epoch": 1.77, "learning_rate": 1.9400839832033594e-06, "loss": 0.0003, "step": 14774 }, { "epoch": 1.77, "learning_rate": 1.94001199760048e-06, "loss": 0.0002, "step": 14775 }, { "epoch": 1.77, "learning_rate": 1.9399400119976005e-06, "loss": 0.0003, "step": 14776 }, { "epoch": 1.77, "learning_rate": 1.9398680263947213e-06, "loss": 0.0001, "step": 14777 }, { "epoch": 1.77, "learning_rate": 1.9397960407918416e-06, "loss": 0.0003, "step": 14778 }, { "epoch": 1.77, "learning_rate": 1.9397240551889624e-06, "loss": 0.4557, "step": 14779 }, { "epoch": 1.77, "learning_rate": 1.9396520695860828e-06, "loss": 0.0005, "step": 14780 }, { "epoch": 1.77, "learning_rate": 1.9395800839832035e-06, "loss": 0.0004, "step": 14781 }, { "epoch": 1.77, "learning_rate": 1.939508098380324e-06, "loss": 0.0002, "step": 14782 }, { "epoch": 1.77, "learning_rate": 1.9394361127774447e-06, "loss": 0.0003, "step": 14783 }, { "epoch": 1.77, "learning_rate": 1.939364127174565e-06, "loss": 0.0003, "step": 14784 }, { "epoch": 1.77, "learning_rate": 1.9392921415716854e-06, "loss": 0.0003, "step": 14785 }, { "epoch": 1.77, "learning_rate": 1.939220155968806e-06, "loss": 0.0002, "step": 14786 }, { "epoch": 1.77, "learning_rate": 1.9391481703659265e-06, "loss": 0.0003, "step": 14787 }, { "epoch": 1.77, "learning_rate": 1.9390761847630473e-06, "loss": 0.0003, "step": 14788 }, { "epoch": 1.77, "learning_rate": 1.939004199160168e-06, "loss": 0.0004, "step": 14789 }, { "epoch": 1.77, "learning_rate": 1.938932213557289e-06, "loss": 0.0015, "step": 14790 }, { "epoch": 1.77, "learning_rate": 1.938860227954409e-06, "loss": 0.0002, "step": 14791 }, { "epoch": 1.77, "learning_rate": 1.93878824235153e-06, "loss": 0.0003, "step": 14792 }, { "epoch": 1.77, "learning_rate": 1.9387162567486503e-06, "loss": 0.0001, "step": 14793 }, { "epoch": 1.77, "learning_rate": 1.938644271145771e-06, "loss": 0.0002, "step": 14794 }, { "epoch": 1.77, "learning_rate": 1.9385722855428915e-06, "loss": 0.0003, "step": 14795 }, { "epoch": 1.77, "learning_rate": 1.9385002999400123e-06, "loss": 0.0002, "step": 14796 }, { "epoch": 1.77, "learning_rate": 1.9384283143371326e-06, "loss": 0.0001, "step": 14797 }, { "epoch": 1.77, "learning_rate": 1.938356328734253e-06, "loss": 0.0239, "step": 14798 }, { "epoch": 1.77, "learning_rate": 1.9382843431313737e-06, "loss": 0.0003, "step": 14799 }, { "epoch": 1.77, "learning_rate": 1.938212357528494e-06, "loss": 0.0002, "step": 14800 }, { "epoch": 1.77, "learning_rate": 1.938140371925615e-06, "loss": 0.0001, "step": 14801 }, { "epoch": 1.77, "learning_rate": 1.9380683863227352e-06, "loss": 0.0003, "step": 14802 }, { "epoch": 1.77, "learning_rate": 1.937996400719856e-06, "loss": 0.0003, "step": 14803 }, { "epoch": 1.77, "learning_rate": 1.937924415116977e-06, "loss": 0.0001, "step": 14804 }, { "epoch": 1.77, "learning_rate": 1.937852429514097e-06, "loss": 0.0004, "step": 14805 }, { "epoch": 1.77, "learning_rate": 1.937780443911218e-06, "loss": 0.0002, "step": 14806 }, { "epoch": 1.77, "learning_rate": 1.9377084583083387e-06, "loss": 0.0002, "step": 14807 }, { "epoch": 1.77, "learning_rate": 1.937636472705459e-06, "loss": 0.5143, "step": 14808 }, { "epoch": 1.77, "learning_rate": 1.9375644871025794e-06, "loss": 0.0002, "step": 14809 }, { "epoch": 1.77, "learning_rate": 1.9374925014997e-06, "loss": 0.0002, "step": 14810 }, { "epoch": 1.77, "learning_rate": 1.9374205158968205e-06, "loss": 0.0002, "step": 14811 }, { "epoch": 1.77, "learning_rate": 1.9373485302939413e-06, "loss": 0.0002, "step": 14812 }, { "epoch": 1.78, "learning_rate": 1.9372765446910617e-06, "loss": 0.0002, "step": 14813 }, { "epoch": 1.78, "learning_rate": 1.9372045590881825e-06, "loss": 0.0002, "step": 14814 }, { "epoch": 1.78, "learning_rate": 1.937132573485303e-06, "loss": 0.0006, "step": 14815 }, { "epoch": 1.78, "learning_rate": 1.9370605878824236e-06, "loss": 0.0001, "step": 14816 }, { "epoch": 1.78, "learning_rate": 1.936988602279544e-06, "loss": 0.0002, "step": 14817 }, { "epoch": 1.78, "learning_rate": 1.9369166166766647e-06, "loss": 0.159, "step": 14818 }, { "epoch": 1.78, "learning_rate": 1.936844631073785e-06, "loss": 0.0002, "step": 14819 }, { "epoch": 1.78, "learning_rate": 1.936772645470906e-06, "loss": 0.0001, "step": 14820 }, { "epoch": 1.78, "learning_rate": 1.9367006598680266e-06, "loss": 0.0001, "step": 14821 }, { "epoch": 1.78, "learning_rate": 1.936628674265147e-06, "loss": 0.0127, "step": 14822 }, { "epoch": 1.78, "learning_rate": 1.9365566886622678e-06, "loss": 0.0002, "step": 14823 }, { "epoch": 1.78, "learning_rate": 1.936484703059388e-06, "loss": 0.0003, "step": 14824 }, { "epoch": 1.78, "learning_rate": 1.936412717456509e-06, "loss": 0.0004, "step": 14825 }, { "epoch": 1.78, "learning_rate": 1.9363407318536293e-06, "loss": 0.4921, "step": 14826 }, { "epoch": 1.78, "learning_rate": 1.93626874625075e-06, "loss": 0.0003, "step": 14827 }, { "epoch": 1.78, "learning_rate": 1.9361967606478704e-06, "loss": 0.0003, "step": 14828 }, { "epoch": 1.78, "learning_rate": 1.936124775044991e-06, "loss": 0.0002, "step": 14829 }, { "epoch": 1.78, "learning_rate": 1.9360527894421115e-06, "loss": 0.0002, "step": 14830 }, { "epoch": 1.78, "learning_rate": 1.935980803839232e-06, "loss": 1.1113, "step": 14831 }, { "epoch": 1.78, "learning_rate": 1.9359088182363527e-06, "loss": 0.0001, "step": 14832 }, { "epoch": 1.78, "learning_rate": 1.935836832633473e-06, "loss": 0.0001, "step": 14833 }, { "epoch": 1.78, "learning_rate": 1.935764847030594e-06, "loss": 0.0002, "step": 14834 }, { "epoch": 1.78, "learning_rate": 1.9356928614277146e-06, "loss": 0.0001, "step": 14835 }, { "epoch": 1.78, "learning_rate": 1.9356208758248354e-06, "loss": 0.0001, "step": 14836 }, { "epoch": 1.78, "learning_rate": 1.9355488902219557e-06, "loss": 0.0002, "step": 14837 }, { "epoch": 1.78, "learning_rate": 1.9354769046190765e-06, "loss": 0.0014, "step": 14838 }, { "epoch": 1.78, "learning_rate": 1.935404919016197e-06, "loss": 0.0083, "step": 14839 }, { "epoch": 1.78, "learning_rate": 1.9353329334133176e-06, "loss": 0.1101, "step": 14840 }, { "epoch": 1.78, "learning_rate": 1.935260947810438e-06, "loss": 0.0002, "step": 14841 }, { "epoch": 1.78, "learning_rate": 1.9351889622075588e-06, "loss": 0.0004, "step": 14842 }, { "epoch": 1.78, "learning_rate": 1.935116976604679e-06, "loss": 0.0001, "step": 14843 }, { "epoch": 1.78, "learning_rate": 1.9350449910017995e-06, "loss": 0.0002, "step": 14844 }, { "epoch": 1.78, "learning_rate": 1.9349730053989202e-06, "loss": 0.0002, "step": 14845 }, { "epoch": 1.78, "learning_rate": 1.9349010197960406e-06, "loss": 0.0003, "step": 14846 }, { "epoch": 1.78, "learning_rate": 1.9348290341931614e-06, "loss": 0.0002, "step": 14847 }, { "epoch": 1.78, "learning_rate": 1.9347570485902817e-06, "loss": 0.0006, "step": 14848 }, { "epoch": 1.78, "learning_rate": 1.9346850629874025e-06, "loss": 0.0002, "step": 14849 }, { "epoch": 1.78, "learning_rate": 1.9346130773845233e-06, "loss": 1.6694, "step": 14850 }, { "epoch": 1.78, "learning_rate": 1.9345410917816436e-06, "loss": 0.0002, "step": 14851 }, { "epoch": 1.78, "learning_rate": 1.9344691061787644e-06, "loss": 0.0006, "step": 14852 }, { "epoch": 1.78, "learning_rate": 1.934397120575885e-06, "loss": 0.0001, "step": 14853 }, { "epoch": 1.78, "learning_rate": 1.9343251349730056e-06, "loss": 0.0001, "step": 14854 }, { "epoch": 1.78, "learning_rate": 1.934253149370126e-06, "loss": 0.0002, "step": 14855 }, { "epoch": 1.78, "learning_rate": 1.9341811637672467e-06, "loss": 0.0002, "step": 14856 }, { "epoch": 1.78, "learning_rate": 1.934109178164367e-06, "loss": 0.0004, "step": 14857 }, { "epoch": 1.78, "learning_rate": 1.934037192561488e-06, "loss": 0.0002, "step": 14858 }, { "epoch": 1.78, "learning_rate": 1.933965206958608e-06, "loss": 0.0002, "step": 14859 }, { "epoch": 1.78, "learning_rate": 1.933893221355729e-06, "loss": 0.0002, "step": 14860 }, { "epoch": 1.78, "learning_rate": 1.9338212357528493e-06, "loss": 0.0002, "step": 14861 }, { "epoch": 1.78, "learning_rate": 1.93374925014997e-06, "loss": 0.0002, "step": 14862 }, { "epoch": 1.78, "learning_rate": 1.9336772645470904e-06, "loss": 0.0001, "step": 14863 }, { "epoch": 1.78, "learning_rate": 1.9336052789442112e-06, "loss": 0.0001, "step": 14864 }, { "epoch": 1.78, "learning_rate": 1.9335332933413316e-06, "loss": 0.0567, "step": 14865 }, { "epoch": 1.78, "learning_rate": 1.9334613077384524e-06, "loss": 0.0001, "step": 14866 }, { "epoch": 1.78, "learning_rate": 1.933389322135573e-06, "loss": 0.0007, "step": 14867 }, { "epoch": 1.78, "learning_rate": 1.9333173365326935e-06, "loss": 0.0002, "step": 14868 }, { "epoch": 1.78, "learning_rate": 1.9332453509298143e-06, "loss": 0.0001, "step": 14869 }, { "epoch": 1.78, "learning_rate": 1.9331733653269346e-06, "loss": 0.0001, "step": 14870 }, { "epoch": 1.78, "learning_rate": 1.9331013797240554e-06, "loss": 0.0002, "step": 14871 }, { "epoch": 1.78, "learning_rate": 1.9330293941211758e-06, "loss": 0.0002, "step": 14872 }, { "epoch": 1.78, "learning_rate": 1.9329574085182965e-06, "loss": 0.0002, "step": 14873 }, { "epoch": 1.78, "learning_rate": 1.932885422915417e-06, "loss": 0.0002, "step": 14874 }, { "epoch": 1.78, "learning_rate": 1.9328134373125377e-06, "loss": 0.0003, "step": 14875 }, { "epoch": 1.78, "learning_rate": 1.932741451709658e-06, "loss": 0.0001, "step": 14876 }, { "epoch": 1.78, "learning_rate": 1.932669466106779e-06, "loss": 0.0002, "step": 14877 }, { "epoch": 1.78, "learning_rate": 1.932597480503899e-06, "loss": 0.0002, "step": 14878 }, { "epoch": 1.78, "learning_rate": 1.9325254949010195e-06, "loss": 0.0002, "step": 14879 }, { "epoch": 1.78, "learning_rate": 1.9324535092981403e-06, "loss": 0.0002, "step": 14880 }, { "epoch": 1.78, "learning_rate": 1.932381523695261e-06, "loss": 0.3256, "step": 14881 }, { "epoch": 1.78, "learning_rate": 1.932309538092382e-06, "loss": 0.0001, "step": 14882 }, { "epoch": 1.78, "learning_rate": 1.932237552489502e-06, "loss": 0.0002, "step": 14883 }, { "epoch": 1.78, "learning_rate": 1.932165566886623e-06, "loss": 0.0002, "step": 14884 }, { "epoch": 1.78, "learning_rate": 1.9320935812837433e-06, "loss": 0.0003, "step": 14885 }, { "epoch": 1.78, "learning_rate": 1.932021595680864e-06, "loss": 0.0002, "step": 14886 }, { "epoch": 1.78, "learning_rate": 1.9319496100779845e-06, "loss": 0.0002, "step": 14887 }, { "epoch": 1.78, "learning_rate": 1.9318776244751052e-06, "loss": 0.0002, "step": 14888 }, { "epoch": 1.78, "learning_rate": 1.9318056388722256e-06, "loss": 0.7148, "step": 14889 }, { "epoch": 1.78, "learning_rate": 1.931733653269346e-06, "loss": 0.0002, "step": 14890 }, { "epoch": 1.78, "learning_rate": 1.9316616676664667e-06, "loss": 0.0004, "step": 14891 }, { "epoch": 1.78, "learning_rate": 1.931589682063587e-06, "loss": 0.0002, "step": 14892 }, { "epoch": 1.78, "learning_rate": 1.931517696460708e-06, "loss": 0.0001, "step": 14893 }, { "epoch": 1.78, "learning_rate": 1.9314457108578282e-06, "loss": 0.0002, "step": 14894 }, { "epoch": 1.78, "learning_rate": 1.931373725254949e-06, "loss": 0.0002, "step": 14895 }, { "epoch": 1.79, "learning_rate": 1.9313017396520698e-06, "loss": 0.0002, "step": 14896 }, { "epoch": 1.79, "learning_rate": 1.93122975404919e-06, "loss": 0.0002, "step": 14897 }, { "epoch": 1.79, "learning_rate": 1.931157768446311e-06, "loss": 0.7967, "step": 14898 }, { "epoch": 1.79, "learning_rate": 1.9310857828434317e-06, "loss": 0.0006, "step": 14899 }, { "epoch": 1.79, "learning_rate": 1.931013797240552e-06, "loss": 0.0004, "step": 14900 }, { "epoch": 1.79, "learning_rate": 1.9309418116376724e-06, "loss": 0.0001, "step": 14901 }, { "epoch": 1.79, "learning_rate": 1.930869826034793e-06, "loss": 0.0003, "step": 14902 }, { "epoch": 1.79, "learning_rate": 1.9307978404319135e-06, "loss": 0.0001, "step": 14903 }, { "epoch": 1.79, "learning_rate": 1.9307258548290343e-06, "loss": 0.0005, "step": 14904 }, { "epoch": 1.79, "learning_rate": 1.9306538692261547e-06, "loss": 0.0304, "step": 14905 }, { "epoch": 1.79, "learning_rate": 1.9305818836232755e-06, "loss": 0.0001, "step": 14906 }, { "epoch": 1.79, "learning_rate": 1.930509898020396e-06, "loss": 0.0001, "step": 14907 }, { "epoch": 1.79, "learning_rate": 1.9304379124175166e-06, "loss": 0.0002, "step": 14908 }, { "epoch": 1.79, "learning_rate": 1.930365926814637e-06, "loss": 0.0001, "step": 14909 }, { "epoch": 1.79, "learning_rate": 1.9302939412117577e-06, "loss": 0.0003, "step": 14910 }, { "epoch": 1.79, "learning_rate": 1.930221955608878e-06, "loss": 0.0001, "step": 14911 }, { "epoch": 1.79, "learning_rate": 1.930149970005999e-06, "loss": 0.0004, "step": 14912 }, { "epoch": 1.79, "learning_rate": 1.9300779844031196e-06, "loss": 0.0003, "step": 14913 }, { "epoch": 1.79, "learning_rate": 1.93000599880024e-06, "loss": 0.0003, "step": 14914 }, { "epoch": 1.79, "learning_rate": 1.9299340131973608e-06, "loss": 0.0001, "step": 14915 }, { "epoch": 1.79, "learning_rate": 1.929862027594481e-06, "loss": 0.0002, "step": 14916 }, { "epoch": 1.79, "learning_rate": 1.929790041991602e-06, "loss": 0.1242, "step": 14917 }, { "epoch": 1.79, "learning_rate": 1.9297180563887223e-06, "loss": 0.0001, "step": 14918 }, { "epoch": 1.79, "learning_rate": 1.929646070785843e-06, "loss": 0.0002, "step": 14919 }, { "epoch": 1.79, "learning_rate": 1.9295740851829634e-06, "loss": 0.0002, "step": 14920 }, { "epoch": 1.79, "learning_rate": 1.929502099580084e-06, "loss": 0.6117, "step": 14921 }, { "epoch": 1.79, "learning_rate": 1.9294301139772045e-06, "loss": 0.0001, "step": 14922 }, { "epoch": 1.79, "learning_rate": 1.9293581283743253e-06, "loss": 0.0002, "step": 14923 }, { "epoch": 1.79, "learning_rate": 1.9292861427714457e-06, "loss": 0.0002, "step": 14924 }, { "epoch": 1.79, "learning_rate": 1.929214157168566e-06, "loss": 0.0002, "step": 14925 }, { "epoch": 1.79, "learning_rate": 1.9291421715656868e-06, "loss": 0.0002, "step": 14926 }, { "epoch": 1.79, "learning_rate": 1.9290701859628076e-06, "loss": 0.0003, "step": 14927 }, { "epoch": 1.79, "learning_rate": 1.9289982003599283e-06, "loss": 0.0002, "step": 14928 }, { "epoch": 1.79, "learning_rate": 1.9289262147570487e-06, "loss": 0.0003, "step": 14929 }, { "epoch": 1.79, "learning_rate": 1.9288542291541695e-06, "loss": 0.0002, "step": 14930 }, { "epoch": 1.79, "learning_rate": 1.92878224355129e-06, "loss": 0.0003, "step": 14931 }, { "epoch": 1.79, "learning_rate": 1.9287102579484106e-06, "loss": 0.0002, "step": 14932 }, { "epoch": 1.79, "learning_rate": 1.928638272345531e-06, "loss": 0.0001, "step": 14933 }, { "epoch": 1.79, "learning_rate": 1.9285662867426517e-06, "loss": 0.0002, "step": 14934 }, { "epoch": 1.79, "learning_rate": 1.928494301139772e-06, "loss": 0.0001, "step": 14935 }, { "epoch": 1.79, "learning_rate": 1.9284223155368925e-06, "loss": 0.0099, "step": 14936 }, { "epoch": 1.79, "learning_rate": 1.9283503299340132e-06, "loss": 0.0005, "step": 14937 }, { "epoch": 1.79, "learning_rate": 1.9282783443311336e-06, "loss": 0.0088, "step": 14938 }, { "epoch": 1.79, "learning_rate": 1.9282063587282544e-06, "loss": 0.0003, "step": 14939 }, { "epoch": 1.79, "learning_rate": 1.9281343731253747e-06, "loss": 0.0003, "step": 14940 }, { "epoch": 1.79, "learning_rate": 1.9280623875224955e-06, "loss": 0.0007, "step": 14941 }, { "epoch": 1.79, "learning_rate": 1.927990401919616e-06, "loss": 0.0003, "step": 14942 }, { "epoch": 1.79, "learning_rate": 1.9279184163167366e-06, "loss": 0.0002, "step": 14943 }, { "epoch": 1.79, "learning_rate": 1.9278464307138574e-06, "loss": 0.0002, "step": 14944 }, { "epoch": 1.79, "learning_rate": 1.927774445110978e-06, "loss": 0.0001, "step": 14945 }, { "epoch": 1.79, "learning_rate": 1.9277024595080985e-06, "loss": 0.001, "step": 14946 }, { "epoch": 1.79, "learning_rate": 1.927630473905219e-06, "loss": 0.0002, "step": 14947 }, { "epoch": 1.79, "learning_rate": 1.9275584883023397e-06, "loss": 0.0931, "step": 14948 }, { "epoch": 1.79, "learning_rate": 1.92748650269946e-06, "loss": 0.1466, "step": 14949 }, { "epoch": 1.79, "learning_rate": 1.927414517096581e-06, "loss": 0.0013, "step": 14950 }, { "epoch": 1.79, "learning_rate": 1.927342531493701e-06, "loss": 0.0002, "step": 14951 }, { "epoch": 1.79, "learning_rate": 1.927270545890822e-06, "loss": 0.0003, "step": 14952 }, { "epoch": 1.79, "learning_rate": 1.9271985602879423e-06, "loss": 0.0002, "step": 14953 }, { "epoch": 1.79, "learning_rate": 1.927126574685063e-06, "loss": 0.0003, "step": 14954 }, { "epoch": 1.79, "learning_rate": 1.9270545890821834e-06, "loss": 0.0001, "step": 14955 }, { "epoch": 1.79, "learning_rate": 1.9269826034793042e-06, "loss": 0.0003, "step": 14956 }, { "epoch": 1.79, "learning_rate": 1.9269106178764246e-06, "loss": 0.0003, "step": 14957 }, { "epoch": 1.79, "learning_rate": 1.9268386322735453e-06, "loss": 0.0001, "step": 14958 }, { "epoch": 1.79, "learning_rate": 1.926766646670666e-06, "loss": 0.0001, "step": 14959 }, { "epoch": 1.79, "learning_rate": 1.9266946610677865e-06, "loss": 0.0002, "step": 14960 }, { "epoch": 1.79, "learning_rate": 1.9266226754649073e-06, "loss": 0.0002, "step": 14961 }, { "epoch": 1.79, "learning_rate": 1.9265506898620276e-06, "loss": 0.0003, "step": 14962 }, { "epoch": 1.79, "learning_rate": 1.9264787042591484e-06, "loss": 0.0001, "step": 14963 }, { "epoch": 1.79, "learning_rate": 1.9264067186562687e-06, "loss": 0.0054, "step": 14964 }, { "epoch": 1.79, "learning_rate": 1.9263347330533895e-06, "loss": 0.9471, "step": 14965 }, { "epoch": 1.79, "learning_rate": 1.92626274745051e-06, "loss": 0.0002, "step": 14966 }, { "epoch": 1.79, "learning_rate": 1.9261907618476307e-06, "loss": 0.0002, "step": 14967 }, { "epoch": 1.79, "learning_rate": 1.926118776244751e-06, "loss": 0.0002, "step": 14968 }, { "epoch": 1.79, "learning_rate": 1.926046790641872e-06, "loss": 0.0002, "step": 14969 }, { "epoch": 1.79, "learning_rate": 1.925974805038992e-06, "loss": 0.0002, "step": 14970 }, { "epoch": 1.79, "learning_rate": 1.9259028194361125e-06, "loss": 0.0002, "step": 14971 }, { "epoch": 1.79, "learning_rate": 1.9258308338332333e-06, "loss": 0.0002, "step": 14972 }, { "epoch": 1.79, "learning_rate": 1.925758848230354e-06, "loss": 0.0002, "step": 14973 }, { "epoch": 1.79, "learning_rate": 1.9256868626274744e-06, "loss": 0.065, "step": 14974 }, { "epoch": 1.79, "learning_rate": 1.925614877024595e-06, "loss": 0.0002, "step": 14975 }, { "epoch": 1.79, "learning_rate": 1.925542891421716e-06, "loss": 0.0003, "step": 14976 }, { "epoch": 1.79, "learning_rate": 1.9254709058188363e-06, "loss": 0.0044, "step": 14977 }, { "epoch": 1.79, "learning_rate": 1.925398920215957e-06, "loss": 0.0002, "step": 14978 }, { "epoch": 1.79, "learning_rate": 1.9253269346130775e-06, "loss": 0.0002, "step": 14979 }, { "epoch": 1.8, "learning_rate": 1.9252549490101982e-06, "loss": 0.8804, "step": 14980 }, { "epoch": 1.8, "learning_rate": 1.9251829634073186e-06, "loss": 0.0002, "step": 14981 }, { "epoch": 1.8, "learning_rate": 1.925110977804439e-06, "loss": 0.0003, "step": 14982 }, { "epoch": 1.8, "learning_rate": 1.9250389922015597e-06, "loss": 0.0002, "step": 14983 }, { "epoch": 1.8, "learning_rate": 1.92496700659868e-06, "loss": 0.0002, "step": 14984 }, { "epoch": 1.8, "learning_rate": 1.924895020995801e-06, "loss": 0.0003, "step": 14985 }, { "epoch": 1.8, "learning_rate": 1.9248230353929212e-06, "loss": 0.0001, "step": 14986 }, { "epoch": 1.8, "learning_rate": 1.924751049790042e-06, "loss": 0.0002, "step": 14987 }, { "epoch": 1.8, "learning_rate": 1.9246790641871624e-06, "loss": 0.0003, "step": 14988 }, { "epoch": 1.8, "learning_rate": 1.924607078584283e-06, "loss": 0.0002, "step": 14989 }, { "epoch": 1.8, "learning_rate": 1.924535092981404e-06, "loss": 0.0004, "step": 14990 }, { "epoch": 1.8, "learning_rate": 1.9244631073785247e-06, "loss": 0.0002, "step": 14991 }, { "epoch": 1.8, "learning_rate": 1.924391121775645e-06, "loss": 0.0002, "step": 14992 }, { "epoch": 1.8, "learning_rate": 1.9243191361727654e-06, "loss": 0.0002, "step": 14993 }, { "epoch": 1.8, "learning_rate": 1.924247150569886e-06, "loss": 0.0002, "step": 14994 }, { "epoch": 1.8, "learning_rate": 1.9241751649670065e-06, "loss": 0.0002, "step": 14995 }, { "epoch": 1.8, "learning_rate": 1.9241031793641273e-06, "loss": 0.0002, "step": 14996 }, { "epoch": 1.8, "learning_rate": 1.9240311937612477e-06, "loss": 0.0001, "step": 14997 }, { "epoch": 1.8, "learning_rate": 1.9239592081583684e-06, "loss": 0.0003, "step": 14998 }, { "epoch": 1.8, "learning_rate": 1.923887222555489e-06, "loss": 0.0003, "step": 14999 }, { "epoch": 1.8, "learning_rate": 1.9238152369526096e-06, "loss": 0.0001, "step": 15000 }, { "epoch": 1.8, "learning_rate": 1.92374325134973e-06, "loss": 0.8938, "step": 15001 }, { "epoch": 1.8, "learning_rate": 1.9236712657468507e-06, "loss": 0.0001, "step": 15002 }, { "epoch": 1.8, "learning_rate": 1.923599280143971e-06, "loss": 0.0002, "step": 15003 }, { "epoch": 1.8, "learning_rate": 1.923527294541092e-06, "loss": 0.0002, "step": 15004 }, { "epoch": 1.8, "learning_rate": 1.9234553089382126e-06, "loss": 0.0001, "step": 15005 }, { "epoch": 1.8, "learning_rate": 1.923383323335333e-06, "loss": 0.0002, "step": 15006 }, { "epoch": 1.8, "learning_rate": 1.9233113377324538e-06, "loss": 0.0001, "step": 15007 }, { "epoch": 1.8, "learning_rate": 1.923239352129574e-06, "loss": 0.0002, "step": 15008 }, { "epoch": 1.8, "learning_rate": 1.923167366526695e-06, "loss": 0.0004, "step": 15009 }, { "epoch": 1.8, "learning_rate": 1.9230953809238152e-06, "loss": 0.4572, "step": 15010 }, { "epoch": 1.8, "learning_rate": 1.923023395320936e-06, "loss": 0.0002, "step": 15011 }, { "epoch": 1.8, "learning_rate": 1.9229514097180564e-06, "loss": 0.0005, "step": 15012 }, { "epoch": 1.8, "learning_rate": 1.922879424115177e-06, "loss": 0.0002, "step": 15013 }, { "epoch": 1.8, "learning_rate": 1.9228074385122975e-06, "loss": 0.0001, "step": 15014 }, { "epoch": 1.8, "learning_rate": 1.9227354529094183e-06, "loss": 0.0004, "step": 15015 }, { "epoch": 1.8, "learning_rate": 1.9226634673065386e-06, "loss": 0.5285, "step": 15016 }, { "epoch": 1.8, "learning_rate": 1.922591481703659e-06, "loss": 0.0002, "step": 15017 }, { "epoch": 1.8, "learning_rate": 1.9225194961007798e-06, "loss": 0.0003, "step": 15018 }, { "epoch": 1.8, "learning_rate": 1.9224475104979006e-06, "loss": 0.001, "step": 15019 }, { "epoch": 1.8, "learning_rate": 1.922375524895021e-06, "loss": 0.0001, "step": 15020 }, { "epoch": 1.8, "learning_rate": 1.9223035392921417e-06, "loss": 0.0002, "step": 15021 }, { "epoch": 1.8, "learning_rate": 1.9222315536892625e-06, "loss": 0.0003, "step": 15022 }, { "epoch": 1.8, "learning_rate": 1.922159568086383e-06, "loss": 0.0005, "step": 15023 }, { "epoch": 1.8, "learning_rate": 1.9220875824835036e-06, "loss": 0.0001, "step": 15024 }, { "epoch": 1.8, "learning_rate": 1.922015596880624e-06, "loss": 0.0002, "step": 15025 }, { "epoch": 1.8, "learning_rate": 1.9219436112777447e-06, "loss": 0.0002, "step": 15026 }, { "epoch": 1.8, "learning_rate": 1.921871625674865e-06, "loss": 0.0003, "step": 15027 }, { "epoch": 1.8, "learning_rate": 1.9217996400719854e-06, "loss": 0.0002, "step": 15028 }, { "epoch": 1.8, "learning_rate": 1.9217276544691062e-06, "loss": 0.0002, "step": 15029 }, { "epoch": 1.8, "learning_rate": 1.9216556688662266e-06, "loss": 0.0002, "step": 15030 }, { "epoch": 1.8, "learning_rate": 1.9215836832633474e-06, "loss": 0.0001, "step": 15031 }, { "epoch": 1.8, "learning_rate": 1.9215116976604677e-06, "loss": 0.0003, "step": 15032 }, { "epoch": 1.8, "learning_rate": 1.9214397120575885e-06, "loss": 0.0013, "step": 15033 }, { "epoch": 1.8, "learning_rate": 1.921367726454709e-06, "loss": 0.0002, "step": 15034 }, { "epoch": 1.8, "learning_rate": 1.9212957408518296e-06, "loss": 0.0009, "step": 15035 }, { "epoch": 1.8, "learning_rate": 1.9212237552489504e-06, "loss": 0.0002, "step": 15036 }, { "epoch": 1.8, "learning_rate": 1.921151769646071e-06, "loss": 0.0002, "step": 15037 }, { "epoch": 1.8, "learning_rate": 1.9210797840431915e-06, "loss": 0.0002, "step": 15038 }, { "epoch": 1.8, "learning_rate": 1.921007798440312e-06, "loss": 0.0002, "step": 15039 }, { "epoch": 1.8, "learning_rate": 1.9209358128374327e-06, "loss": 0.0014, "step": 15040 }, { "epoch": 1.8, "learning_rate": 1.920863827234553e-06, "loss": 0.0002, "step": 15041 }, { "epoch": 1.8, "learning_rate": 1.920791841631674e-06, "loss": 0.0002, "step": 15042 }, { "epoch": 1.8, "learning_rate": 1.920719856028794e-06, "loss": 0.0052, "step": 15043 }, { "epoch": 1.8, "learning_rate": 1.920647870425915e-06, "loss": 0.0003, "step": 15044 }, { "epoch": 1.8, "learning_rate": 1.9205758848230353e-06, "loss": 0.0002, "step": 15045 }, { "epoch": 1.8, "learning_rate": 1.920503899220156e-06, "loss": 0.0001, "step": 15046 }, { "epoch": 1.8, "learning_rate": 1.9204319136172764e-06, "loss": 0.0003, "step": 15047 }, { "epoch": 1.8, "learning_rate": 1.920359928014397e-06, "loss": 0.0007, "step": 15048 }, { "epoch": 1.8, "learning_rate": 1.9202879424115176e-06, "loss": 0.0003, "step": 15049 }, { "epoch": 1.8, "learning_rate": 1.9202159568086383e-06, "loss": 0.0001, "step": 15050 }, { "epoch": 1.8, "learning_rate": 1.920143971205759e-06, "loss": 0.0925, "step": 15051 }, { "epoch": 1.8, "learning_rate": 1.9200719856028795e-06, "loss": 0.1876, "step": 15052 }, { "epoch": 1.8, "learning_rate": 1.9200000000000003e-06, "loss": 0.0001, "step": 15053 }, { "epoch": 1.8, "learning_rate": 1.9199280143971206e-06, "loss": 0.0002, "step": 15054 }, { "epoch": 1.8, "learning_rate": 1.9198560287942414e-06, "loss": 0.0002, "step": 15055 }, { "epoch": 1.8, "learning_rate": 1.9197840431913617e-06, "loss": 0.0906, "step": 15056 }, { "epoch": 1.8, "learning_rate": 1.9197120575884825e-06, "loss": 0.0001, "step": 15057 }, { "epoch": 1.8, "learning_rate": 1.919640071985603e-06, "loss": 0.0014, "step": 15058 }, { "epoch": 1.8, "learning_rate": 1.9195680863827237e-06, "loss": 0.0001, "step": 15059 }, { "epoch": 1.8, "learning_rate": 1.919496100779844e-06, "loss": 0.0002, "step": 15060 }, { "epoch": 1.8, "learning_rate": 1.9194241151769648e-06, "loss": 0.0002, "step": 15061 }, { "epoch": 1.8, "learning_rate": 1.919352129574085e-06, "loss": 0.0002, "step": 15062 }, { "epoch": 1.81, "learning_rate": 1.9192801439712055e-06, "loss": 0.4411, "step": 15063 }, { "epoch": 1.81, "learning_rate": 1.9192081583683263e-06, "loss": 0.0003, "step": 15064 }, { "epoch": 1.81, "learning_rate": 1.9191361727654466e-06, "loss": 0.0002, "step": 15065 }, { "epoch": 1.81, "learning_rate": 1.9190641871625674e-06, "loss": 0.0005, "step": 15066 }, { "epoch": 1.81, "learning_rate": 1.918992201559688e-06, "loss": 0.0001, "step": 15067 }, { "epoch": 1.81, "learning_rate": 1.918920215956809e-06, "loss": 0.0001, "step": 15068 }, { "epoch": 1.81, "learning_rate": 1.9188482303539293e-06, "loss": 0.0002, "step": 15069 }, { "epoch": 1.81, "learning_rate": 1.91877624475105e-06, "loss": 0.0003, "step": 15070 }, { "epoch": 1.81, "learning_rate": 1.9187042591481705e-06, "loss": 0.0004, "step": 15071 }, { "epoch": 1.81, "learning_rate": 1.9186322735452912e-06, "loss": 0.0002, "step": 15072 }, { "epoch": 1.81, "learning_rate": 1.9185602879424116e-06, "loss": 0.0003, "step": 15073 }, { "epoch": 1.81, "learning_rate": 1.918488302339532e-06, "loss": 0.0003, "step": 15074 }, { "epoch": 1.81, "learning_rate": 1.9184163167366527e-06, "loss": 0.0002, "step": 15075 }, { "epoch": 1.81, "learning_rate": 1.918344331133773e-06, "loss": 0.0003, "step": 15076 }, { "epoch": 1.81, "learning_rate": 1.918272345530894e-06, "loss": 0.0001, "step": 15077 }, { "epoch": 1.81, "learning_rate": 1.918200359928014e-06, "loss": 0.0002, "step": 15078 }, { "epoch": 1.81, "learning_rate": 1.918128374325135e-06, "loss": 0.0002, "step": 15079 }, { "epoch": 1.81, "learning_rate": 1.9180563887222553e-06, "loss": 0.0002, "step": 15080 }, { "epoch": 1.81, "learning_rate": 1.917984403119376e-06, "loss": 0.0001, "step": 15081 }, { "epoch": 1.81, "learning_rate": 1.917912417516497e-06, "loss": 0.0001, "step": 15082 }, { "epoch": 1.81, "learning_rate": 1.9178404319136173e-06, "loss": 0.0002, "step": 15083 }, { "epoch": 1.81, "learning_rate": 1.917768446310738e-06, "loss": 0.0002, "step": 15084 }, { "epoch": 1.81, "learning_rate": 1.9176964607078584e-06, "loss": 0.0002, "step": 15085 }, { "epoch": 1.81, "learning_rate": 1.917624475104979e-06, "loss": 0.0001, "step": 15086 }, { "epoch": 1.81, "learning_rate": 1.9175524895020995e-06, "loss": 0.0001, "step": 15087 }, { "epoch": 1.81, "learning_rate": 1.9174805038992203e-06, "loss": 0.0002, "step": 15088 }, { "epoch": 1.81, "learning_rate": 1.9174085182963407e-06, "loss": 0.0003, "step": 15089 }, { "epoch": 1.81, "learning_rate": 1.9173365326934614e-06, "loss": 0.0001, "step": 15090 }, { "epoch": 1.81, "learning_rate": 1.917264547090582e-06, "loss": 0.0002, "step": 15091 }, { "epoch": 1.81, "learning_rate": 1.9171925614877026e-06, "loss": 0.0003, "step": 15092 }, { "epoch": 1.81, "learning_rate": 1.917120575884823e-06, "loss": 0.0001, "step": 15093 }, { "epoch": 1.81, "learning_rate": 1.9170485902819437e-06, "loss": 0.0002, "step": 15094 }, { "epoch": 1.81, "learning_rate": 1.916976604679064e-06, "loss": 0.0002, "step": 15095 }, { "epoch": 1.81, "learning_rate": 1.916904619076185e-06, "loss": 0.0002, "step": 15096 }, { "epoch": 1.81, "learning_rate": 1.916832633473305e-06, "loss": 0.0003, "step": 15097 }, { "epoch": 1.81, "learning_rate": 1.916760647870426e-06, "loss": 0.0004, "step": 15098 }, { "epoch": 1.81, "learning_rate": 1.9166886622675467e-06, "loss": 0.0002, "step": 15099 }, { "epoch": 1.81, "learning_rate": 1.916616676664667e-06, "loss": 0.0003, "step": 15100 }, { "epoch": 1.81, "learning_rate": 1.916544691061788e-06, "loss": 0.0002, "step": 15101 }, { "epoch": 1.81, "learning_rate": 1.9164727054589082e-06, "loss": 0.0004, "step": 15102 }, { "epoch": 1.81, "learning_rate": 1.916400719856029e-06, "loss": 0.0002, "step": 15103 }, { "epoch": 1.81, "learning_rate": 1.9163287342531494e-06, "loss": 0.0004, "step": 15104 }, { "epoch": 1.81, "learning_rate": 1.91625674865027e-06, "loss": 0.0002, "step": 15105 }, { "epoch": 1.81, "learning_rate": 1.9161847630473905e-06, "loss": 0.0001, "step": 15106 }, { "epoch": 1.81, "learning_rate": 1.9161127774445113e-06, "loss": 0.9549, "step": 15107 }, { "epoch": 1.81, "learning_rate": 1.9160407918416316e-06, "loss": 0.0001, "step": 15108 }, { "epoch": 1.81, "learning_rate": 1.915968806238752e-06, "loss": 0.0003, "step": 15109 }, { "epoch": 1.81, "learning_rate": 1.9158968206358728e-06, "loss": 0.0001, "step": 15110 }, { "epoch": 1.81, "learning_rate": 1.915824835032993e-06, "loss": 0.0002, "step": 15111 }, { "epoch": 1.81, "learning_rate": 1.915752849430114e-06, "loss": 0.0002, "step": 15112 }, { "epoch": 1.81, "learning_rate": 1.9156808638272347e-06, "loss": 0.0002, "step": 15113 }, { "epoch": 1.81, "learning_rate": 1.9156088782243555e-06, "loss": 0.0002, "step": 15114 }, { "epoch": 1.81, "learning_rate": 1.915536892621476e-06, "loss": 0.0003, "step": 15115 }, { "epoch": 1.81, "learning_rate": 1.9154649070185966e-06, "loss": 0.0511, "step": 15116 }, { "epoch": 1.81, "learning_rate": 1.915392921415717e-06, "loss": 0.0001, "step": 15117 }, { "epoch": 1.81, "learning_rate": 1.9153209358128377e-06, "loss": 0.0051, "step": 15118 }, { "epoch": 1.81, "learning_rate": 1.915248950209958e-06, "loss": 0.0001, "step": 15119 }, { "epoch": 1.81, "learning_rate": 1.9151769646070784e-06, "loss": 0.0002, "step": 15120 }, { "epoch": 1.81, "learning_rate": 1.9151049790041992e-06, "loss": 0.0001, "step": 15121 }, { "epoch": 1.81, "learning_rate": 1.9150329934013196e-06, "loss": 0.0001, "step": 15122 }, { "epoch": 1.81, "learning_rate": 1.9149610077984404e-06, "loss": 0.0002, "step": 15123 }, { "epoch": 1.81, "learning_rate": 1.9148890221955607e-06, "loss": 0.0002, "step": 15124 }, { "epoch": 1.81, "learning_rate": 1.9148170365926815e-06, "loss": 0.1513, "step": 15125 }, { "epoch": 1.81, "learning_rate": 1.914745050989802e-06, "loss": 0.0001, "step": 15126 }, { "epoch": 1.81, "learning_rate": 1.9146730653869226e-06, "loss": 0.0001, "step": 15127 }, { "epoch": 1.81, "learning_rate": 1.9146010797840434e-06, "loss": 0.0002, "step": 15128 }, { "epoch": 1.81, "learning_rate": 1.9145290941811638e-06, "loss": 0.0003, "step": 15129 }, { "epoch": 1.81, "learning_rate": 1.9144571085782845e-06, "loss": 0.0003, "step": 15130 }, { "epoch": 1.81, "learning_rate": 1.9143851229754053e-06, "loss": 0.0002, "step": 15131 }, { "epoch": 1.81, "learning_rate": 1.9143131373725257e-06, "loss": 0.004, "step": 15132 }, { "epoch": 1.81, "learning_rate": 1.914241151769646e-06, "loss": 0.0001, "step": 15133 }, { "epoch": 1.81, "learning_rate": 1.914169166166767e-06, "loss": 0.0002, "step": 15134 }, { "epoch": 1.81, "learning_rate": 1.914097180563887e-06, "loss": 0.0004, "step": 15135 }, { "epoch": 1.81, "learning_rate": 1.914025194961008e-06, "loss": 0.0002, "step": 15136 }, { "epoch": 1.81, "learning_rate": 1.9139532093581283e-06, "loss": 0.0002, "step": 15137 }, { "epoch": 1.81, "learning_rate": 1.913881223755249e-06, "loss": 0.0003, "step": 15138 }, { "epoch": 1.81, "learning_rate": 1.9138092381523694e-06, "loss": 0.0002, "step": 15139 }, { "epoch": 1.81, "learning_rate": 1.91373725254949e-06, "loss": 1.1975, "step": 15140 }, { "epoch": 1.81, "learning_rate": 1.9136652669466106e-06, "loss": 0.0095, "step": 15141 }, { "epoch": 1.81, "learning_rate": 1.9135932813437313e-06, "loss": 0.0001, "step": 15142 }, { "epoch": 1.81, "learning_rate": 1.9135212957408517e-06, "loss": 0.0016, "step": 15143 }, { "epoch": 1.81, "learning_rate": 1.9134493101379725e-06, "loss": 0.0003, "step": 15144 }, { "epoch": 1.81, "learning_rate": 1.9133773245350932e-06, "loss": 0.0002, "step": 15145 }, { "epoch": 1.81, "learning_rate": 1.9133053389322136e-06, "loss": 0.0002, "step": 15146 }, { "epoch": 1.82, "learning_rate": 1.9132333533293344e-06, "loss": 0.0002, "step": 15147 }, { "epoch": 1.82, "learning_rate": 1.9131613677264547e-06, "loss": 0.0002, "step": 15148 }, { "epoch": 1.82, "learning_rate": 1.9130893821235755e-06, "loss": 0.0003, "step": 15149 }, { "epoch": 1.82, "learning_rate": 1.913017396520696e-06, "loss": 0.0001, "step": 15150 }, { "epoch": 1.82, "learning_rate": 1.9129454109178166e-06, "loss": 0.0001, "step": 15151 }, { "epoch": 1.82, "learning_rate": 1.912873425314937e-06, "loss": 0.0004, "step": 15152 }, { "epoch": 1.82, "learning_rate": 1.9128014397120578e-06, "loss": 0.0002, "step": 15153 }, { "epoch": 1.82, "learning_rate": 1.912729454109178e-06, "loss": 0.0138, "step": 15154 }, { "epoch": 1.82, "learning_rate": 1.9126574685062985e-06, "loss": 0.5285, "step": 15155 }, { "epoch": 1.82, "learning_rate": 1.9125854829034193e-06, "loss": 0.0002, "step": 15156 }, { "epoch": 1.82, "learning_rate": 1.9125134973005396e-06, "loss": 0.0003, "step": 15157 }, { "epoch": 1.82, "learning_rate": 1.9124415116976604e-06, "loss": 0.0001, "step": 15158 }, { "epoch": 1.82, "learning_rate": 1.912369526094781e-06, "loss": 0.0002, "step": 15159 }, { "epoch": 1.82, "learning_rate": 1.912297540491902e-06, "loss": 0.0002, "step": 15160 }, { "epoch": 1.82, "learning_rate": 1.9122255548890223e-06, "loss": 0.0002, "step": 15161 }, { "epoch": 1.82, "learning_rate": 1.912153569286143e-06, "loss": 0.0003, "step": 15162 }, { "epoch": 1.82, "learning_rate": 1.9120815836832634e-06, "loss": 0.0001, "step": 15163 }, { "epoch": 1.82, "learning_rate": 1.9120095980803842e-06, "loss": 0.0001, "step": 15164 }, { "epoch": 1.82, "learning_rate": 1.9119376124775046e-06, "loss": 0.0002, "step": 15165 }, { "epoch": 1.82, "learning_rate": 1.911865626874625e-06, "loss": 0.0001, "step": 15166 }, { "epoch": 1.82, "learning_rate": 1.9117936412717457e-06, "loss": 0.0002, "step": 15167 }, { "epoch": 1.82, "learning_rate": 1.911721655668866e-06, "loss": 0.0001, "step": 15168 }, { "epoch": 1.82, "learning_rate": 1.911649670065987e-06, "loss": 0.0002, "step": 15169 }, { "epoch": 1.82, "learning_rate": 1.911577684463107e-06, "loss": 0.001, "step": 15170 }, { "epoch": 1.82, "learning_rate": 1.911505698860228e-06, "loss": 0.0002, "step": 15171 }, { "epoch": 1.82, "learning_rate": 1.9114337132573483e-06, "loss": 0.0443, "step": 15172 }, { "epoch": 1.82, "learning_rate": 1.911361727654469e-06, "loss": 0.0003, "step": 15173 }, { "epoch": 1.82, "learning_rate": 1.91128974205159e-06, "loss": 0.0002, "step": 15174 }, { "epoch": 1.82, "learning_rate": 1.9112177564487102e-06, "loss": 0.0002, "step": 15175 }, { "epoch": 1.82, "learning_rate": 1.911145770845831e-06, "loss": 0.0001, "step": 15176 }, { "epoch": 1.82, "learning_rate": 1.911073785242952e-06, "loss": 0.0002, "step": 15177 }, { "epoch": 1.82, "learning_rate": 1.911001799640072e-06, "loss": 0.0002, "step": 15178 }, { "epoch": 1.82, "learning_rate": 1.9109298140371925e-06, "loss": 0.0004, "step": 15179 }, { "epoch": 1.82, "learning_rate": 1.9108578284343133e-06, "loss": 0.7522, "step": 15180 }, { "epoch": 1.82, "learning_rate": 1.9107858428314336e-06, "loss": 0.0002, "step": 15181 }, { "epoch": 1.82, "learning_rate": 1.9107138572285544e-06, "loss": 0.0013, "step": 15182 }, { "epoch": 1.82, "learning_rate": 1.9106418716256748e-06, "loss": 0.0003, "step": 15183 }, { "epoch": 1.82, "learning_rate": 1.9105698860227956e-06, "loss": 0.0001, "step": 15184 }, { "epoch": 1.82, "learning_rate": 1.910497900419916e-06, "loss": 0.0002, "step": 15185 }, { "epoch": 1.82, "learning_rate": 1.9104259148170367e-06, "loss": 0.0195, "step": 15186 }, { "epoch": 1.82, "learning_rate": 1.910353929214157e-06, "loss": 0.0002, "step": 15187 }, { "epoch": 1.82, "learning_rate": 1.910281943611278e-06, "loss": 0.0032, "step": 15188 }, { "epoch": 1.82, "learning_rate": 1.910209958008398e-06, "loss": 0.0014, "step": 15189 }, { "epoch": 1.82, "learning_rate": 1.910137972405519e-06, "loss": 0.0005, "step": 15190 }, { "epoch": 1.82, "learning_rate": 1.9100659868026397e-06, "loss": 0.0002, "step": 15191 }, { "epoch": 1.82, "learning_rate": 1.90999400119976e-06, "loss": 0.002, "step": 15192 }, { "epoch": 1.82, "learning_rate": 1.909922015596881e-06, "loss": 0.0002, "step": 15193 }, { "epoch": 1.82, "learning_rate": 1.9098500299940012e-06, "loss": 0.0001, "step": 15194 }, { "epoch": 1.82, "learning_rate": 1.909778044391122e-06, "loss": 0.0002, "step": 15195 }, { "epoch": 1.82, "learning_rate": 1.9097060587882424e-06, "loss": 0.0002, "step": 15196 }, { "epoch": 1.82, "learning_rate": 1.909634073185363e-06, "loss": 0.0002, "step": 15197 }, { "epoch": 1.82, "learning_rate": 1.9095620875824835e-06, "loss": 0.0001, "step": 15198 }, { "epoch": 1.82, "learning_rate": 1.9094901019796043e-06, "loss": 0.5116, "step": 15199 }, { "epoch": 1.82, "learning_rate": 1.9094181163767246e-06, "loss": 0.0002, "step": 15200 }, { "epoch": 1.82, "learning_rate": 1.909346130773845e-06, "loss": 0.0002, "step": 15201 }, { "epoch": 1.82, "learning_rate": 1.9092741451709658e-06, "loss": 0.0002, "step": 15202 }, { "epoch": 1.82, "learning_rate": 1.909202159568086e-06, "loss": 0.0001, "step": 15203 }, { "epoch": 1.82, "learning_rate": 1.909130173965207e-06, "loss": 0.0002, "step": 15204 }, { "epoch": 1.82, "learning_rate": 1.9090581883623277e-06, "loss": 0.0002, "step": 15205 }, { "epoch": 1.82, "learning_rate": 1.908986202759448e-06, "loss": 0.0001, "step": 15206 }, { "epoch": 1.82, "learning_rate": 1.908914217156569e-06, "loss": 0.0014, "step": 15207 }, { "epoch": 1.82, "learning_rate": 1.9088422315536896e-06, "loss": 0.0002, "step": 15208 }, { "epoch": 1.82, "learning_rate": 1.90877024595081e-06, "loss": 0.0001, "step": 15209 }, { "epoch": 1.82, "learning_rate": 1.9086982603479307e-06, "loss": 0.6011, "step": 15210 }, { "epoch": 1.82, "learning_rate": 1.908626274745051e-06, "loss": 0.0002, "step": 15211 }, { "epoch": 1.82, "learning_rate": 1.9085542891421714e-06, "loss": 0.0001, "step": 15212 }, { "epoch": 1.82, "learning_rate": 1.908482303539292e-06, "loss": 0.0004, "step": 15213 }, { "epoch": 1.82, "learning_rate": 1.9084103179364126e-06, "loss": 0.0009, "step": 15214 }, { "epoch": 1.82, "learning_rate": 1.9083383323335333e-06, "loss": 0.0002, "step": 15215 }, { "epoch": 1.82, "learning_rate": 1.9082663467306537e-06, "loss": 0.0002, "step": 15216 }, { "epoch": 1.82, "learning_rate": 1.9081943611277745e-06, "loss": 0.9247, "step": 15217 }, { "epoch": 1.82, "learning_rate": 1.908122375524895e-06, "loss": 0.0002, "step": 15218 }, { "epoch": 1.82, "learning_rate": 1.9080503899220156e-06, "loss": 0.0001, "step": 15219 }, { "epoch": 1.82, "learning_rate": 1.907978404319136e-06, "loss": 0.0002, "step": 15220 }, { "epoch": 1.82, "learning_rate": 1.9079064187162567e-06, "loss": 0.0001, "step": 15221 }, { "epoch": 1.82, "learning_rate": 1.9078344331133775e-06, "loss": 0.0003, "step": 15222 }, { "epoch": 1.82, "learning_rate": 1.9077624475104983e-06, "loss": 0.0003, "step": 15223 }, { "epoch": 1.82, "learning_rate": 1.9076904619076187e-06, "loss": 0.0003, "step": 15224 }, { "epoch": 1.82, "learning_rate": 1.907618476304739e-06, "loss": 0.0003, "step": 15225 }, { "epoch": 1.82, "learning_rate": 1.90754649070186e-06, "loss": 0.0003, "step": 15226 }, { "epoch": 1.82, "learning_rate": 1.90747450509898e-06, "loss": 0.0003, "step": 15227 }, { "epoch": 1.82, "learning_rate": 1.907402519496101e-06, "loss": 0.0002, "step": 15228 }, { "epoch": 1.82, "learning_rate": 1.9073305338932213e-06, "loss": 0.0003, "step": 15229 }, { "epoch": 1.83, "learning_rate": 1.907258548290342e-06, "loss": 0.0001, "step": 15230 }, { "epoch": 1.83, "learning_rate": 1.9071865626874624e-06, "loss": 0.0003, "step": 15231 }, { "epoch": 1.83, "learning_rate": 1.9071145770845832e-06, "loss": 0.0002, "step": 15232 }, { "epoch": 1.83, "learning_rate": 1.9070425914817038e-06, "loss": 0.0002, "step": 15233 }, { "epoch": 1.83, "learning_rate": 1.9069706058788245e-06, "loss": 0.0003, "step": 15234 }, { "epoch": 1.83, "learning_rate": 1.9068986202759449e-06, "loss": 0.0002, "step": 15235 }, { "epoch": 1.83, "learning_rate": 1.9068266346730652e-06, "loss": 0.0003, "step": 15236 }, { "epoch": 1.83, "learning_rate": 1.906754649070186e-06, "loss": 0.0004, "step": 15237 }, { "epoch": 1.83, "learning_rate": 1.9066826634673064e-06, "loss": 0.0002, "step": 15238 }, { "epoch": 1.83, "learning_rate": 1.9066106778644272e-06, "loss": 0.3347, "step": 15239 }, { "epoch": 1.83, "learning_rate": 1.9065386922615477e-06, "loss": 0.0004, "step": 15240 }, { "epoch": 1.83, "learning_rate": 1.9064667066586685e-06, "loss": 0.001, "step": 15241 }, { "epoch": 1.83, "learning_rate": 1.9063947210557889e-06, "loss": 0.0002, "step": 15242 }, { "epoch": 1.83, "learning_rate": 1.9063227354529096e-06, "loss": 0.0002, "step": 15243 }, { "epoch": 1.83, "learning_rate": 1.90625074985003e-06, "loss": 0.0002, "step": 15244 }, { "epoch": 1.83, "learning_rate": 1.9061787642471508e-06, "loss": 0.0003, "step": 15245 }, { "epoch": 1.83, "learning_rate": 1.9061067786442711e-06, "loss": 0.0002, "step": 15246 }, { "epoch": 1.83, "learning_rate": 1.9060347930413917e-06, "loss": 0.0001, "step": 15247 }, { "epoch": 1.83, "learning_rate": 1.9059628074385125e-06, "loss": 0.0002, "step": 15248 }, { "epoch": 1.83, "learning_rate": 1.9058908218356328e-06, "loss": 0.0003, "step": 15249 }, { "epoch": 1.83, "learning_rate": 1.9058188362327536e-06, "loss": 0.0002, "step": 15250 }, { "epoch": 1.83, "learning_rate": 1.905746850629874e-06, "loss": 0.0003, "step": 15251 }, { "epoch": 1.83, "learning_rate": 1.9056748650269947e-06, "loss": 0.0002, "step": 15252 }, { "epoch": 1.83, "learning_rate": 1.905602879424115e-06, "loss": 0.0004, "step": 15253 }, { "epoch": 1.83, "learning_rate": 1.9055308938212359e-06, "loss": 0.0002, "step": 15254 }, { "epoch": 1.83, "learning_rate": 1.9054589082183564e-06, "loss": 0.0002, "step": 15255 }, { "epoch": 1.83, "learning_rate": 1.905386922615477e-06, "loss": 0.0003, "step": 15256 }, { "epoch": 1.83, "learning_rate": 1.9053149370125976e-06, "loss": 0.0002, "step": 15257 }, { "epoch": 1.83, "learning_rate": 1.9052429514097184e-06, "loss": 0.0001, "step": 15258 }, { "epoch": 1.83, "learning_rate": 1.9051709658068387e-06, "loss": 0.0002, "step": 15259 }, { "epoch": 1.83, "learning_rate": 1.905098980203959e-06, "loss": 0.0002, "step": 15260 }, { "epoch": 1.83, "learning_rate": 1.9050269946010798e-06, "loss": 0.0003, "step": 15261 }, { "epoch": 1.83, "learning_rate": 1.9049550089982002e-06, "loss": 0.0003, "step": 15262 }, { "epoch": 1.83, "learning_rate": 1.904883023395321e-06, "loss": 0.0008, "step": 15263 }, { "epoch": 1.83, "learning_rate": 1.9048110377924415e-06, "loss": 0.0002, "step": 15264 }, { "epoch": 1.83, "learning_rate": 1.9047390521895623e-06, "loss": 0.0002, "step": 15265 }, { "epoch": 1.83, "learning_rate": 1.9046670665866827e-06, "loss": 0.0002, "step": 15266 }, { "epoch": 1.83, "learning_rate": 1.9045950809838035e-06, "loss": 0.0002, "step": 15267 }, { "epoch": 1.83, "learning_rate": 1.9045230953809238e-06, "loss": 0.0002, "step": 15268 }, { "epoch": 1.83, "learning_rate": 1.9044511097780446e-06, "loss": 0.0003, "step": 15269 }, { "epoch": 1.83, "learning_rate": 1.904379124175165e-06, "loss": 0.0001, "step": 15270 }, { "epoch": 1.83, "learning_rate": 1.9043071385722855e-06, "loss": 0.0002, "step": 15271 }, { "epoch": 1.83, "learning_rate": 1.9042351529694063e-06, "loss": 0.0001, "step": 15272 }, { "epoch": 1.83, "learning_rate": 1.9041631673665266e-06, "loss": 0.0001, "step": 15273 }, { "epoch": 1.83, "learning_rate": 1.9040911817636474e-06, "loss": 0.0004, "step": 15274 }, { "epoch": 1.83, "learning_rate": 1.9040191961607678e-06, "loss": 0.0003, "step": 15275 }, { "epoch": 1.83, "learning_rate": 1.9039472105578886e-06, "loss": 0.0001, "step": 15276 }, { "epoch": 1.83, "learning_rate": 1.903875224955009e-06, "loss": 0.0001, "step": 15277 }, { "epoch": 1.83, "learning_rate": 1.9038032393521297e-06, "loss": 0.0001, "step": 15278 }, { "epoch": 1.83, "learning_rate": 1.9037312537492503e-06, "loss": 0.0002, "step": 15279 }, { "epoch": 1.83, "learning_rate": 1.9036592681463708e-06, "loss": 0.0002, "step": 15280 }, { "epoch": 1.83, "learning_rate": 1.9035872825434914e-06, "loss": 0.0001, "step": 15281 }, { "epoch": 1.83, "learning_rate": 1.9035152969406117e-06, "loss": 0.0008, "step": 15282 }, { "epoch": 1.83, "learning_rate": 1.9034433113377325e-06, "loss": 0.0002, "step": 15283 }, { "epoch": 1.83, "learning_rate": 1.9033713257348529e-06, "loss": 0.0001, "step": 15284 }, { "epoch": 1.83, "learning_rate": 1.9032993401319737e-06, "loss": 0.0002, "step": 15285 }, { "epoch": 1.83, "learning_rate": 1.9032273545290942e-06, "loss": 0.0002, "step": 15286 }, { "epoch": 1.83, "learning_rate": 1.9031553689262148e-06, "loss": 0.0002, "step": 15287 }, { "epoch": 1.83, "learning_rate": 1.9030833833233354e-06, "loss": 0.0002, "step": 15288 }, { "epoch": 1.83, "learning_rate": 1.9030113977204561e-06, "loss": 0.001, "step": 15289 }, { "epoch": 1.83, "learning_rate": 1.9029394121175765e-06, "loss": 0.0001, "step": 15290 }, { "epoch": 1.83, "learning_rate": 1.9028674265146973e-06, "loss": 0.0002, "step": 15291 }, { "epoch": 1.83, "learning_rate": 1.9027954409118176e-06, "loss": 0.0001, "step": 15292 }, { "epoch": 1.83, "learning_rate": 1.9027234553089382e-06, "loss": 0.0001, "step": 15293 }, { "epoch": 1.83, "learning_rate": 1.9026514697060588e-06, "loss": 0.0004, "step": 15294 }, { "epoch": 1.83, "learning_rate": 1.9025794841031793e-06, "loss": 0.0038, "step": 15295 }, { "epoch": 1.83, "learning_rate": 1.9025074985003e-06, "loss": 0.0002, "step": 15296 }, { "epoch": 1.83, "learning_rate": 1.9024355128974205e-06, "loss": 0.001, "step": 15297 }, { "epoch": 1.83, "learning_rate": 1.9023635272945412e-06, "loss": 0.0001, "step": 15298 }, { "epoch": 1.83, "learning_rate": 1.9022915416916616e-06, "loss": 0.0002, "step": 15299 }, { "epoch": 1.83, "learning_rate": 1.9022195560887824e-06, "loss": 0.0007, "step": 15300 }, { "epoch": 1.83, "learning_rate": 1.9021475704859027e-06, "loss": 0.0001, "step": 15301 }, { "epoch": 1.83, "learning_rate": 1.9020755848830235e-06, "loss": 0.0002, "step": 15302 }, { "epoch": 1.83, "learning_rate": 1.902003599280144e-06, "loss": 0.0089, "step": 15303 }, { "epoch": 1.83, "learning_rate": 1.9019316136772648e-06, "loss": 0.0004, "step": 15304 }, { "epoch": 1.83, "learning_rate": 1.9018596280743852e-06, "loss": 0.0002, "step": 15305 }, { "epoch": 1.83, "learning_rate": 1.9017876424715056e-06, "loss": 0.0001, "step": 15306 }, { "epoch": 1.83, "learning_rate": 1.9017156568686263e-06, "loss": 0.0001, "step": 15307 }, { "epoch": 1.83, "learning_rate": 1.9016436712657467e-06, "loss": 0.0001, "step": 15308 }, { "epoch": 1.83, "learning_rate": 1.9015716856628675e-06, "loss": 0.0002, "step": 15309 }, { "epoch": 1.83, "learning_rate": 1.901499700059988e-06, "loss": 0.0002, "step": 15310 }, { "epoch": 1.83, "learning_rate": 1.9014277144571088e-06, "loss": 0.0007, "step": 15311 }, { "epoch": 1.83, "learning_rate": 1.9013557288542292e-06, "loss": 0.0001, "step": 15312 }, { "epoch": 1.83, "learning_rate": 1.90128374325135e-06, "loss": 0.0004, "step": 15313 }, { "epoch": 1.84, "learning_rate": 1.9012117576484703e-06, "loss": 0.0002, "step": 15314 }, { "epoch": 1.84, "learning_rate": 1.901139772045591e-06, "loss": 0.0002, "step": 15315 }, { "epoch": 1.84, "learning_rate": 1.9010677864427114e-06, "loss": 0.0003, "step": 15316 }, { "epoch": 1.84, "learning_rate": 1.900995800839832e-06, "loss": 0.0001, "step": 15317 }, { "epoch": 1.84, "learning_rate": 1.9009238152369528e-06, "loss": 0.0003, "step": 15318 }, { "epoch": 1.84, "learning_rate": 1.9008518296340731e-06, "loss": 0.0001, "step": 15319 }, { "epoch": 1.84, "learning_rate": 1.900779844031194e-06, "loss": 0.0001, "step": 15320 }, { "epoch": 1.84, "learning_rate": 1.9007078584283143e-06, "loss": 0.0001, "step": 15321 }, { "epoch": 1.84, "learning_rate": 1.900635872825435e-06, "loss": 0.0001, "step": 15322 }, { "epoch": 1.84, "learning_rate": 1.9005638872225554e-06, "loss": 0.0002, "step": 15323 }, { "epoch": 1.84, "learning_rate": 1.9004919016196762e-06, "loss": 0.0003, "step": 15324 }, { "epoch": 1.84, "learning_rate": 1.9004199160167967e-06, "loss": 0.0005, "step": 15325 }, { "epoch": 1.84, "learning_rate": 1.9003479304139173e-06, "loss": 0.6788, "step": 15326 }, { "epoch": 1.84, "learning_rate": 1.9002759448110379e-06, "loss": 0.0002, "step": 15327 }, { "epoch": 1.84, "learning_rate": 1.9002039592081582e-06, "loss": 0.0002, "step": 15328 }, { "epoch": 1.84, "learning_rate": 1.900131973605279e-06, "loss": 0.0002, "step": 15329 }, { "epoch": 1.84, "learning_rate": 1.9000599880023994e-06, "loss": 0.0002, "step": 15330 }, { "epoch": 1.84, "learning_rate": 1.8999880023995201e-06, "loss": 0.0007, "step": 15331 }, { "epoch": 1.84, "learning_rate": 1.8999160167966407e-06, "loss": 0.0002, "step": 15332 }, { "epoch": 1.84, "learning_rate": 1.8998440311937613e-06, "loss": 0.0003, "step": 15333 }, { "epoch": 1.84, "learning_rate": 1.8997720455908819e-06, "loss": 0.0099, "step": 15334 }, { "epoch": 1.84, "learning_rate": 1.8997000599880026e-06, "loss": 0.0012, "step": 15335 }, { "epoch": 1.84, "learning_rate": 1.899628074385123e-06, "loss": 0.0005, "step": 15336 }, { "epoch": 1.84, "learning_rate": 1.8995560887822438e-06, "loss": 0.0003, "step": 15337 }, { "epoch": 1.84, "learning_rate": 1.8994841031793641e-06, "loss": 0.0003, "step": 15338 }, { "epoch": 1.84, "learning_rate": 1.8994121175764847e-06, "loss": 0.0001, "step": 15339 }, { "epoch": 1.84, "learning_rate": 1.8993401319736053e-06, "loss": 0.0002, "step": 15340 }, { "epoch": 1.84, "learning_rate": 1.8992681463707258e-06, "loss": 0.0001, "step": 15341 }, { "epoch": 1.84, "learning_rate": 1.8991961607678466e-06, "loss": 0.0001, "step": 15342 }, { "epoch": 1.84, "learning_rate": 1.899124175164967e-06, "loss": 0.0008, "step": 15343 }, { "epoch": 1.84, "learning_rate": 1.8990521895620877e-06, "loss": 0.0002, "step": 15344 }, { "epoch": 1.84, "learning_rate": 1.898980203959208e-06, "loss": 0.0001, "step": 15345 }, { "epoch": 1.84, "learning_rate": 1.8989082183563289e-06, "loss": 0.0003, "step": 15346 }, { "epoch": 1.84, "learning_rate": 1.8988362327534492e-06, "loss": 0.8485, "step": 15347 }, { "epoch": 1.84, "learning_rate": 1.89876424715057e-06, "loss": 0.0003, "step": 15348 }, { "epoch": 1.84, "learning_rate": 1.8986922615476906e-06, "loss": 0.0034, "step": 15349 }, { "epoch": 1.84, "learning_rate": 1.8986202759448113e-06, "loss": 0.0002, "step": 15350 }, { "epoch": 1.84, "learning_rate": 1.8985482903419317e-06, "loss": 0.7598, "step": 15351 }, { "epoch": 1.84, "learning_rate": 1.898476304739052e-06, "loss": 0.0002, "step": 15352 }, { "epoch": 1.84, "learning_rate": 1.8984043191361728e-06, "loss": 0.0001, "step": 15353 }, { "epoch": 1.84, "learning_rate": 1.8983323335332932e-06, "loss": 0.0001, "step": 15354 }, { "epoch": 1.84, "learning_rate": 1.898260347930414e-06, "loss": 0.0001, "step": 15355 }, { "epoch": 1.84, "learning_rate": 1.8981883623275345e-06, "loss": 0.0002, "step": 15356 }, { "epoch": 1.84, "learning_rate": 1.8981163767246553e-06, "loss": 0.0002, "step": 15357 }, { "epoch": 1.84, "learning_rate": 1.8980443911217757e-06, "loss": 0.0006, "step": 15358 }, { "epoch": 1.84, "learning_rate": 1.8979724055188964e-06, "loss": 0.0042, "step": 15359 }, { "epoch": 1.84, "learning_rate": 1.8979004199160168e-06, "loss": 0.0002, "step": 15360 }, { "epoch": 1.84, "learning_rate": 1.8978284343131376e-06, "loss": 0.0002, "step": 15361 }, { "epoch": 1.84, "learning_rate": 1.897756448710258e-06, "loss": 0.0001, "step": 15362 }, { "epoch": 1.84, "learning_rate": 1.8976844631073785e-06, "loss": 0.0001, "step": 15363 }, { "epoch": 1.84, "learning_rate": 1.8976124775044993e-06, "loss": 0.0003, "step": 15364 }, { "epoch": 1.84, "learning_rate": 1.8975404919016196e-06, "loss": 0.0002, "step": 15365 }, { "epoch": 1.84, "learning_rate": 1.8974685062987404e-06, "loss": 0.0001, "step": 15366 }, { "epoch": 1.84, "learning_rate": 1.8973965206958608e-06, "loss": 0.0001, "step": 15367 }, { "epoch": 1.84, "learning_rate": 1.8973245350929815e-06, "loss": 0.0002, "step": 15368 }, { "epoch": 1.84, "learning_rate": 1.897252549490102e-06, "loss": 0.0006, "step": 15369 }, { "epoch": 1.84, "learning_rate": 1.8971805638872227e-06, "loss": 0.0002, "step": 15370 }, { "epoch": 1.84, "learning_rate": 1.8971085782843432e-06, "loss": 0.0001, "step": 15371 }, { "epoch": 1.84, "learning_rate": 1.8970365926814638e-06, "loss": 0.0001, "step": 15372 }, { "epoch": 1.84, "learning_rate": 1.8969646070785844e-06, "loss": 0.0001, "step": 15373 }, { "epoch": 1.84, "learning_rate": 1.8968926214757047e-06, "loss": 0.0003, "step": 15374 }, { "epoch": 1.84, "learning_rate": 1.8968206358728255e-06, "loss": 0.0001, "step": 15375 }, { "epoch": 1.84, "learning_rate": 1.8967486502699459e-06, "loss": 0.0003, "step": 15376 }, { "epoch": 1.84, "learning_rate": 1.8966766646670666e-06, "loss": 0.0002, "step": 15377 }, { "epoch": 1.84, "learning_rate": 1.8966046790641872e-06, "loss": 0.0244, "step": 15378 }, { "epoch": 1.84, "learning_rate": 1.8965326934613078e-06, "loss": 0.0001, "step": 15379 }, { "epoch": 1.84, "learning_rate": 1.8964607078584283e-06, "loss": 0.0002, "step": 15380 }, { "epoch": 1.84, "learning_rate": 1.8963887222555491e-06, "loss": 0.0003, "step": 15381 }, { "epoch": 1.84, "learning_rate": 1.8963167366526695e-06, "loss": 0.0062, "step": 15382 }, { "epoch": 1.84, "learning_rate": 1.8962447510497903e-06, "loss": 0.0001, "step": 15383 }, { "epoch": 1.84, "learning_rate": 1.8961727654469106e-06, "loss": 0.0001, "step": 15384 }, { "epoch": 1.84, "learning_rate": 1.8961007798440314e-06, "loss": 0.0001, "step": 15385 }, { "epoch": 1.84, "learning_rate": 1.8960287942411517e-06, "loss": 1.0152, "step": 15386 }, { "epoch": 1.84, "learning_rate": 1.8959568086382723e-06, "loss": 0.0001, "step": 15387 }, { "epoch": 1.84, "learning_rate": 1.895884823035393e-06, "loss": 0.0002, "step": 15388 }, { "epoch": 1.84, "learning_rate": 1.8958128374325134e-06, "loss": 0.0002, "step": 15389 }, { "epoch": 1.84, "learning_rate": 1.8957408518296342e-06, "loss": 0.4271, "step": 15390 }, { "epoch": 1.84, "learning_rate": 1.8956688662267546e-06, "loss": 0.0002, "step": 15391 }, { "epoch": 1.84, "learning_rate": 1.8955968806238754e-06, "loss": 0.0002, "step": 15392 }, { "epoch": 1.84, "learning_rate": 1.8955248950209957e-06, "loss": 0.0036, "step": 15393 }, { "epoch": 1.84, "learning_rate": 1.8954529094181165e-06, "loss": 0.0001, "step": 15394 }, { "epoch": 1.84, "learning_rate": 1.895380923815237e-06, "loss": 0.0003, "step": 15395 }, { "epoch": 1.84, "learning_rate": 1.8953089382123578e-06, "loss": 0.0002, "step": 15396 }, { "epoch": 1.85, "learning_rate": 1.8952369526094782e-06, "loss": 0.0001, "step": 15397 }, { "epoch": 1.85, "learning_rate": 1.8951649670065985e-06, "loss": 0.0002, "step": 15398 }, { "epoch": 1.85, "learning_rate": 1.8950929814037193e-06, "loss": 0.0002, "step": 15399 }, { "epoch": 1.85, "learning_rate": 1.8950209958008397e-06, "loss": 0.0002, "step": 15400 }, { "epoch": 1.85, "learning_rate": 1.8949490101979605e-06, "loss": 0.0001, "step": 15401 }, { "epoch": 1.85, "learning_rate": 1.894877024595081e-06, "loss": 0.0001, "step": 15402 }, { "epoch": 1.85, "learning_rate": 1.8948050389922016e-06, "loss": 0.0002, "step": 15403 }, { "epoch": 1.85, "learning_rate": 1.8947330533893222e-06, "loss": 0.0004, "step": 15404 }, { "epoch": 1.85, "learning_rate": 1.894661067786443e-06, "loss": 0.0054, "step": 15405 }, { "epoch": 1.85, "learning_rate": 1.8945890821835633e-06, "loss": 0.002, "step": 15406 }, { "epoch": 1.85, "learning_rate": 1.894517096580684e-06, "loss": 0.0001, "step": 15407 }, { "epoch": 1.85, "learning_rate": 1.8944451109778044e-06, "loss": 0.0008, "step": 15408 }, { "epoch": 1.85, "learning_rate": 1.894373125374925e-06, "loss": 0.0002, "step": 15409 }, { "epoch": 1.85, "learning_rate": 1.8943011397720456e-06, "loss": 0.0001, "step": 15410 }, { "epoch": 1.85, "learning_rate": 1.8942291541691661e-06, "loss": 0.0002, "step": 15411 }, { "epoch": 1.85, "learning_rate": 1.894157168566287e-06, "loss": 0.0002, "step": 15412 }, { "epoch": 1.85, "learning_rate": 1.8940851829634073e-06, "loss": 0.0001, "step": 15413 }, { "epoch": 1.85, "learning_rate": 1.894013197360528e-06, "loss": 0.0002, "step": 15414 }, { "epoch": 1.85, "learning_rate": 1.8939412117576484e-06, "loss": 0.0002, "step": 15415 }, { "epoch": 1.85, "learning_rate": 1.8938692261547692e-06, "loss": 0.0002, "step": 15416 }, { "epoch": 1.85, "learning_rate": 1.8937972405518895e-06, "loss": 0.0001, "step": 15417 }, { "epoch": 1.85, "learning_rate": 1.8937252549490103e-06, "loss": 0.0004, "step": 15418 }, { "epoch": 1.85, "learning_rate": 1.8936532693461309e-06, "loss": 0.0001, "step": 15419 }, { "epoch": 1.85, "learning_rate": 1.8935812837432512e-06, "loss": 0.0002, "step": 15420 }, { "epoch": 1.85, "learning_rate": 1.893509298140372e-06, "loss": 0.0003, "step": 15421 }, { "epoch": 1.85, "learning_rate": 1.8934373125374924e-06, "loss": 0.0002, "step": 15422 }, { "epoch": 1.85, "learning_rate": 1.8933653269346131e-06, "loss": 0.0003, "step": 15423 }, { "epoch": 1.85, "learning_rate": 1.8932933413317335e-06, "loss": 0.0002, "step": 15424 }, { "epoch": 1.85, "learning_rate": 1.8932213557288543e-06, "loss": 0.0002, "step": 15425 }, { "epoch": 1.85, "learning_rate": 1.8931493701259748e-06, "loss": 0.1153, "step": 15426 }, { "epoch": 1.85, "learning_rate": 1.8930773845230956e-06, "loss": 0.001, "step": 15427 }, { "epoch": 1.85, "learning_rate": 1.893005398920216e-06, "loss": 0.0001, "step": 15428 }, { "epoch": 1.85, "learning_rate": 1.8929334133173368e-06, "loss": 0.0004, "step": 15429 }, { "epoch": 1.85, "learning_rate": 1.8928614277144571e-06, "loss": 0.0002, "step": 15430 }, { "epoch": 1.85, "learning_rate": 1.8927894421115779e-06, "loss": 0.0001, "step": 15431 }, { "epoch": 1.85, "learning_rate": 1.8927174565086982e-06, "loss": 0.0002, "step": 15432 }, { "epoch": 1.85, "learning_rate": 1.8926454709058188e-06, "loss": 0.0002, "step": 15433 }, { "epoch": 1.85, "learning_rate": 1.8925734853029396e-06, "loss": 0.0002, "step": 15434 }, { "epoch": 1.85, "learning_rate": 1.89250149970006e-06, "loss": 0.0002, "step": 15435 }, { "epoch": 1.85, "learning_rate": 1.8924295140971807e-06, "loss": 0.0002, "step": 15436 }, { "epoch": 1.85, "learning_rate": 1.892357528494301e-06, "loss": 0.007, "step": 15437 }, { "epoch": 1.85, "learning_rate": 1.8922855428914219e-06, "loss": 0.0002, "step": 15438 }, { "epoch": 1.85, "learning_rate": 1.8922135572885422e-06, "loss": 0.7935, "step": 15439 }, { "epoch": 1.85, "learning_rate": 1.892141571685663e-06, "loss": 0.0002, "step": 15440 }, { "epoch": 1.85, "learning_rate": 1.8920695860827836e-06, "loss": 0.0003, "step": 15441 }, { "epoch": 1.85, "learning_rate": 1.8919976004799041e-06, "loss": 0.0001, "step": 15442 }, { "epoch": 1.85, "learning_rate": 1.8919256148770247e-06, "loss": 0.0811, "step": 15443 }, { "epoch": 1.85, "learning_rate": 1.891853629274145e-06, "loss": 0.0001, "step": 15444 }, { "epoch": 1.85, "learning_rate": 1.8917816436712658e-06, "loss": 0.0001, "step": 15445 }, { "epoch": 1.85, "learning_rate": 1.8917096580683862e-06, "loss": 0.0002, "step": 15446 }, { "epoch": 1.85, "learning_rate": 1.891637672465507e-06, "loss": 0.0018, "step": 15447 }, { "epoch": 1.85, "learning_rate": 1.8915656868626275e-06, "loss": 0.0002, "step": 15448 }, { "epoch": 1.85, "learning_rate": 1.891493701259748e-06, "loss": 0.0001, "step": 15449 }, { "epoch": 1.85, "learning_rate": 1.8914217156568687e-06, "loss": 0.0001, "step": 15450 }, { "epoch": 1.85, "learning_rate": 1.8913497300539894e-06, "loss": 0.0001, "step": 15451 }, { "epoch": 1.85, "learning_rate": 1.8912777444511098e-06, "loss": 0.0001, "step": 15452 }, { "epoch": 1.85, "learning_rate": 1.8912057588482306e-06, "loss": 0.0001, "step": 15453 }, { "epoch": 1.85, "learning_rate": 1.891133773245351e-06, "loss": 0.0002, "step": 15454 }, { "epoch": 1.85, "learning_rate": 1.8910617876424715e-06, "loss": 0.0003, "step": 15455 }, { "epoch": 1.85, "learning_rate": 1.890989802039592e-06, "loss": 0.0004, "step": 15456 }, { "epoch": 1.85, "learning_rate": 1.8909178164367126e-06, "loss": 0.0002, "step": 15457 }, { "epoch": 1.85, "learning_rate": 1.8908458308338334e-06, "loss": 0.0002, "step": 15458 }, { "epoch": 1.85, "learning_rate": 1.8907738452309538e-06, "loss": 0.0002, "step": 15459 }, { "epoch": 1.85, "learning_rate": 1.8907018596280745e-06, "loss": 0.0002, "step": 15460 }, { "epoch": 1.85, "learning_rate": 1.8906298740251949e-06, "loss": 0.0001, "step": 15461 }, { "epoch": 1.85, "learning_rate": 1.8905578884223157e-06, "loss": 0.0001, "step": 15462 }, { "epoch": 1.85, "learning_rate": 1.890485902819436e-06, "loss": 0.0001, "step": 15463 }, { "epoch": 1.85, "learning_rate": 1.8904139172165568e-06, "loss": 0.0002, "step": 15464 }, { "epoch": 1.85, "learning_rate": 1.8903419316136774e-06, "loss": 0.0002, "step": 15465 }, { "epoch": 1.85, "learning_rate": 1.8902699460107977e-06, "loss": 0.0002, "step": 15466 }, { "epoch": 1.85, "learning_rate": 1.8901979604079185e-06, "loss": 0.0002, "step": 15467 }, { "epoch": 1.85, "learning_rate": 1.8901259748050389e-06, "loss": 0.0001, "step": 15468 }, { "epoch": 1.85, "learning_rate": 1.8900539892021596e-06, "loss": 0.0003, "step": 15469 }, { "epoch": 1.85, "learning_rate": 1.88998200359928e-06, "loss": 0.0002, "step": 15470 }, { "epoch": 1.85, "learning_rate": 1.8899100179964008e-06, "loss": 0.0002, "step": 15471 }, { "epoch": 1.85, "learning_rate": 1.8898380323935213e-06, "loss": 0.2169, "step": 15472 }, { "epoch": 1.85, "learning_rate": 1.8897660467906421e-06, "loss": 0.0001, "step": 15473 }, { "epoch": 1.85, "learning_rate": 1.8896940611877625e-06, "loss": 0.0001, "step": 15474 }, { "epoch": 1.85, "learning_rate": 1.8896220755848833e-06, "loss": 0.0002, "step": 15475 }, { "epoch": 1.85, "learning_rate": 1.8895500899820036e-06, "loss": 0.0001, "step": 15476 }, { "epoch": 1.85, "learning_rate": 1.8894781043791244e-06, "loss": 0.0002, "step": 15477 }, { "epoch": 1.85, "learning_rate": 1.8894061187762447e-06, "loss": 0.0001, "step": 15478 }, { "epoch": 1.85, "learning_rate": 1.8893341331733653e-06, "loss": 0.0001, "step": 15479 }, { "epoch": 1.86, "learning_rate": 1.889262147570486e-06, "loss": 0.0001, "step": 15480 }, { "epoch": 1.86, "learning_rate": 1.8891901619676064e-06, "loss": 0.0001, "step": 15481 }, { "epoch": 1.86, "learning_rate": 1.8891181763647272e-06, "loss": 0.0002, "step": 15482 }, { "epoch": 1.86, "learning_rate": 1.8890461907618476e-06, "loss": 0.0001, "step": 15483 }, { "epoch": 1.86, "learning_rate": 1.8889742051589684e-06, "loss": 0.0003, "step": 15484 }, { "epoch": 1.86, "learning_rate": 1.8889022195560887e-06, "loss": 0.0002, "step": 15485 }, { "epoch": 1.86, "learning_rate": 1.8888302339532095e-06, "loss": 0.0002, "step": 15486 }, { "epoch": 1.86, "learning_rate": 1.88875824835033e-06, "loss": 0.0001, "step": 15487 }, { "epoch": 1.86, "learning_rate": 1.8886862627474506e-06, "loss": 0.0002, "step": 15488 }, { "epoch": 1.86, "learning_rate": 1.8886142771445712e-06, "loss": 0.0001, "step": 15489 }, { "epoch": 1.86, "learning_rate": 1.8885422915416915e-06, "loss": 0.0001, "step": 15490 }, { "epoch": 1.86, "learning_rate": 1.8884703059388123e-06, "loss": 0.0002, "step": 15491 }, { "epoch": 1.86, "learning_rate": 1.8883983203359327e-06, "loss": 0.0001, "step": 15492 }, { "epoch": 1.86, "learning_rate": 1.8883263347330535e-06, "loss": 0.0002, "step": 15493 }, { "epoch": 1.86, "learning_rate": 1.888254349130174e-06, "loss": 0.0002, "step": 15494 }, { "epoch": 1.86, "learning_rate": 1.8881823635272946e-06, "loss": 0.0001, "step": 15495 }, { "epoch": 1.86, "learning_rate": 1.8881103779244152e-06, "loss": 0.0002, "step": 15496 }, { "epoch": 1.86, "learning_rate": 1.888038392321536e-06, "loss": 0.0002, "step": 15497 }, { "epoch": 1.86, "learning_rate": 1.8879664067186563e-06, "loss": 0.0002, "step": 15498 }, { "epoch": 1.86, "learning_rate": 1.887894421115777e-06, "loss": 0.0001, "step": 15499 }, { "epoch": 1.86, "learning_rate": 1.8878224355128974e-06, "loss": 0.0001, "step": 15500 }, { "epoch": 1.86, "learning_rate": 1.887750449910018e-06, "loss": 1.031, "step": 15501 }, { "epoch": 1.86, "learning_rate": 1.8876784643071386e-06, "loss": 0.0004, "step": 15502 }, { "epoch": 1.86, "learning_rate": 1.8876064787042591e-06, "loss": 0.0002, "step": 15503 }, { "epoch": 1.86, "learning_rate": 1.88753449310138e-06, "loss": 0.0001, "step": 15504 }, { "epoch": 1.86, "learning_rate": 1.8874625074985003e-06, "loss": 0.965, "step": 15505 }, { "epoch": 1.86, "learning_rate": 1.887390521895621e-06, "loss": 0.0002, "step": 15506 }, { "epoch": 1.86, "learning_rate": 1.8873185362927414e-06, "loss": 0.0001, "step": 15507 }, { "epoch": 1.86, "learning_rate": 1.8872465506898622e-06, "loss": 0.0003, "step": 15508 }, { "epoch": 1.86, "learning_rate": 1.8871745650869825e-06, "loss": 0.0002, "step": 15509 }, { "epoch": 1.86, "learning_rate": 1.8871025794841033e-06, "loss": 0.6183, "step": 15510 }, { "epoch": 1.86, "learning_rate": 1.8870305938812239e-06, "loss": 0.0003, "step": 15511 }, { "epoch": 1.86, "learning_rate": 1.8869586082783446e-06, "loss": 0.0002, "step": 15512 }, { "epoch": 1.86, "learning_rate": 1.886886622675465e-06, "loss": 0.0002, "step": 15513 }, { "epoch": 1.86, "learning_rate": 1.8868146370725854e-06, "loss": 0.0003, "step": 15514 }, { "epoch": 1.86, "learning_rate": 1.8867426514697061e-06, "loss": 0.0002, "step": 15515 }, { "epoch": 1.86, "learning_rate": 1.8866706658668265e-06, "loss": 0.0002, "step": 15516 }, { "epoch": 1.86, "learning_rate": 1.8865986802639473e-06, "loss": 0.0001, "step": 15517 }, { "epoch": 1.86, "learning_rate": 1.8865266946610678e-06, "loss": 0.0002, "step": 15518 }, { "epoch": 1.86, "learning_rate": 1.8864547090581886e-06, "loss": 0.0003, "step": 15519 }, { "epoch": 1.86, "learning_rate": 1.886382723455309e-06, "loss": 0.0002, "step": 15520 }, { "epoch": 1.86, "learning_rate": 1.8863107378524297e-06, "loss": 0.0003, "step": 15521 }, { "epoch": 1.86, "learning_rate": 1.88623875224955e-06, "loss": 1.0476, "step": 15522 }, { "epoch": 1.86, "learning_rate": 1.8861667666466709e-06, "loss": 0.0002, "step": 15523 }, { "epoch": 1.86, "learning_rate": 1.8860947810437912e-06, "loss": 0.0002, "step": 15524 }, { "epoch": 1.86, "learning_rate": 1.8860227954409118e-06, "loss": 0.0001, "step": 15525 }, { "epoch": 1.86, "learning_rate": 1.8859508098380326e-06, "loss": 0.0001, "step": 15526 }, { "epoch": 1.86, "learning_rate": 1.885878824235153e-06, "loss": 0.0002, "step": 15527 }, { "epoch": 1.86, "learning_rate": 1.8858068386322737e-06, "loss": 0.0003, "step": 15528 }, { "epoch": 1.86, "learning_rate": 1.885734853029394e-06, "loss": 0.0002, "step": 15529 }, { "epoch": 1.86, "learning_rate": 1.8856628674265148e-06, "loss": 0.0001, "step": 15530 }, { "epoch": 1.86, "learning_rate": 1.8855908818236352e-06, "loss": 0.0002, "step": 15531 }, { "epoch": 1.86, "learning_rate": 1.885518896220756e-06, "loss": 0.0003, "step": 15532 }, { "epoch": 1.86, "learning_rate": 1.8854469106178763e-06, "loss": 0.0002, "step": 15533 }, { "epoch": 1.86, "learning_rate": 1.8853749250149971e-06, "loss": 0.0003, "step": 15534 }, { "epoch": 1.86, "learning_rate": 1.8853029394121177e-06, "loss": 0.0012, "step": 15535 }, { "epoch": 1.86, "learning_rate": 1.885230953809238e-06, "loss": 0.9131, "step": 15536 }, { "epoch": 1.86, "learning_rate": 1.8851589682063588e-06, "loss": 0.0004, "step": 15537 }, { "epoch": 1.86, "learning_rate": 1.8850869826034792e-06, "loss": 0.0002, "step": 15538 }, { "epoch": 1.86, "learning_rate": 1.8850149970006e-06, "loss": 0.0002, "step": 15539 }, { "epoch": 1.86, "learning_rate": 1.8849430113977203e-06, "loss": 0.1228, "step": 15540 }, { "epoch": 1.86, "learning_rate": 1.884871025794841e-06, "loss": 0.0002, "step": 15541 }, { "epoch": 1.86, "learning_rate": 1.8847990401919616e-06, "loss": 0.0002, "step": 15542 }, { "epoch": 1.86, "learning_rate": 1.8847270545890824e-06, "loss": 0.0002, "step": 15543 }, { "epoch": 1.86, "learning_rate": 1.8846550689862028e-06, "loss": 0.0008, "step": 15544 }, { "epoch": 1.86, "learning_rate": 1.8845830833833236e-06, "loss": 0.0005, "step": 15545 }, { "epoch": 1.86, "learning_rate": 1.884511097780444e-06, "loss": 0.0004, "step": 15546 }, { "epoch": 1.86, "learning_rate": 1.8844391121775643e-06, "loss": 0.0004, "step": 15547 }, { "epoch": 1.86, "learning_rate": 1.884367126574685e-06, "loss": 0.0001, "step": 15548 }, { "epoch": 1.86, "learning_rate": 1.8842951409718056e-06, "loss": 0.0002, "step": 15549 }, { "epoch": 1.86, "learning_rate": 1.8842231553689264e-06, "loss": 0.0003, "step": 15550 }, { "epoch": 1.86, "learning_rate": 1.8841511697660467e-06, "loss": 0.0004, "step": 15551 }, { "epoch": 1.86, "learning_rate": 1.8840791841631675e-06, "loss": 0.7275, "step": 15552 }, { "epoch": 1.86, "learning_rate": 1.8840071985602879e-06, "loss": 0.0003, "step": 15553 }, { "epoch": 1.86, "learning_rate": 1.8839352129574087e-06, "loss": 0.0001, "step": 15554 }, { "epoch": 1.86, "learning_rate": 1.883863227354529e-06, "loss": 0.0006, "step": 15555 }, { "epoch": 1.86, "learning_rate": 1.8837912417516498e-06, "loss": 0.0003, "step": 15556 }, { "epoch": 1.86, "learning_rate": 1.8837192561487704e-06, "loss": 0.0001, "step": 15557 }, { "epoch": 1.86, "learning_rate": 1.883647270545891e-06, "loss": 0.0002, "step": 15558 }, { "epoch": 1.86, "learning_rate": 1.8835752849430115e-06, "loss": 0.0019, "step": 15559 }, { "epoch": 1.86, "learning_rate": 1.8835032993401319e-06, "loss": 0.0003, "step": 15560 }, { "epoch": 1.86, "learning_rate": 1.8834313137372526e-06, "loss": 0.0002, "step": 15561 }, { "epoch": 1.86, "learning_rate": 1.883359328134373e-06, "loss": 0.0002, "step": 15562 }, { "epoch": 1.86, "learning_rate": 1.8832873425314938e-06, "loss": 0.0003, "step": 15563 }, { "epoch": 1.87, "learning_rate": 1.8832153569286143e-06, "loss": 0.0001, "step": 15564 }, { "epoch": 1.87, "learning_rate": 1.883143371325735e-06, "loss": 0.0002, "step": 15565 }, { "epoch": 1.87, "learning_rate": 1.8830713857228555e-06, "loss": 0.0002, "step": 15566 }, { "epoch": 1.87, "learning_rate": 1.8829994001199762e-06, "loss": 0.0002, "step": 15567 }, { "epoch": 1.87, "learning_rate": 1.8829274145170966e-06, "loss": 0.0003, "step": 15568 }, { "epoch": 1.87, "learning_rate": 1.8828554289142174e-06, "loss": 0.0003, "step": 15569 }, { "epoch": 1.87, "learning_rate": 1.8827834433113377e-06, "loss": 0.4609, "step": 15570 }, { "epoch": 1.87, "learning_rate": 1.8827114577084583e-06, "loss": 0.0002, "step": 15571 }, { "epoch": 1.87, "learning_rate": 1.8826394721055789e-06, "loss": 0.0002, "step": 15572 }, { "epoch": 1.87, "learning_rate": 1.8825674865026994e-06, "loss": 0.0002, "step": 15573 }, { "epoch": 1.87, "learning_rate": 1.8824955008998202e-06, "loss": 0.0002, "step": 15574 }, { "epoch": 1.87, "learning_rate": 1.8824235152969406e-06, "loss": 0.0005, "step": 15575 }, { "epoch": 1.87, "learning_rate": 1.8823515296940613e-06, "loss": 0.0001, "step": 15576 }, { "epoch": 1.87, "learning_rate": 1.8822795440911817e-06, "loss": 0.0002, "step": 15577 }, { "epoch": 1.87, "learning_rate": 1.8822075584883025e-06, "loss": 0.0003, "step": 15578 }, { "epoch": 1.87, "learning_rate": 1.8821355728854228e-06, "loss": 0.0002, "step": 15579 }, { "epoch": 1.87, "learning_rate": 1.8820635872825436e-06, "loss": 0.0002, "step": 15580 }, { "epoch": 1.87, "learning_rate": 1.8819916016796642e-06, "loss": 0.0003, "step": 15581 }, { "epoch": 1.87, "learning_rate": 1.8819196160767845e-06, "loss": 0.0002, "step": 15582 }, { "epoch": 1.87, "learning_rate": 1.8818476304739053e-06, "loss": 0.0001, "step": 15583 }, { "epoch": 1.87, "learning_rate": 1.8817756448710257e-06, "loss": 0.0002, "step": 15584 }, { "epoch": 1.87, "learning_rate": 1.8817036592681464e-06, "loss": 0.0002, "step": 15585 }, { "epoch": 1.87, "learning_rate": 1.8816316736652668e-06, "loss": 0.0002, "step": 15586 }, { "epoch": 1.87, "learning_rate": 1.8815596880623876e-06, "loss": 0.0036, "step": 15587 }, { "epoch": 1.87, "learning_rate": 1.8814877024595081e-06, "loss": 0.0002, "step": 15588 }, { "epoch": 1.87, "learning_rate": 1.881415716856629e-06, "loss": 0.0002, "step": 15589 }, { "epoch": 1.87, "learning_rate": 1.8813437312537493e-06, "loss": 0.0003, "step": 15590 }, { "epoch": 1.87, "learning_rate": 1.88127174565087e-06, "loss": 0.0002, "step": 15591 }, { "epoch": 1.87, "learning_rate": 1.8811997600479904e-06, "loss": 0.0003, "step": 15592 }, { "epoch": 1.87, "learning_rate": 1.8811277744451108e-06, "loss": 0.0002, "step": 15593 }, { "epoch": 1.87, "learning_rate": 1.8810557888422315e-06, "loss": 0.0003, "step": 15594 }, { "epoch": 1.87, "learning_rate": 1.8809838032393521e-06, "loss": 0.0003, "step": 15595 }, { "epoch": 1.87, "learning_rate": 1.8809118176364729e-06, "loss": 0.0003, "step": 15596 }, { "epoch": 1.87, "learning_rate": 1.8808398320335932e-06, "loss": 0.0004, "step": 15597 }, { "epoch": 1.87, "learning_rate": 1.880767846430714e-06, "loss": 0.0001, "step": 15598 }, { "epoch": 1.87, "learning_rate": 1.8806958608278344e-06, "loss": 0.0002, "step": 15599 }, { "epoch": 1.87, "learning_rate": 1.8806238752249552e-06, "loss": 0.8451, "step": 15600 }, { "epoch": 1.87, "learning_rate": 1.8805518896220755e-06, "loss": 0.0001, "step": 15601 }, { "epoch": 1.87, "learning_rate": 1.8804799040191963e-06, "loss": 0.0041, "step": 15602 }, { "epoch": 1.87, "learning_rate": 1.8804079184163169e-06, "loss": 0.0002, "step": 15603 }, { "epoch": 1.87, "learning_rate": 1.8803359328134374e-06, "loss": 0.0002, "step": 15604 }, { "epoch": 1.87, "learning_rate": 1.880263947210558e-06, "loss": 0.0002, "step": 15605 }, { "epoch": 1.87, "learning_rate": 1.8801919616076783e-06, "loss": 0.0138, "step": 15606 }, { "epoch": 1.87, "learning_rate": 1.8801199760047991e-06, "loss": 0.0001, "step": 15607 }, { "epoch": 1.87, "learning_rate": 1.8800479904019195e-06, "loss": 0.0002, "step": 15608 }, { "epoch": 1.87, "learning_rate": 1.8799760047990403e-06, "loss": 0.0001, "step": 15609 }, { "epoch": 1.87, "learning_rate": 1.8799040191961608e-06, "loss": 0.0013, "step": 15610 }, { "epoch": 1.87, "learning_rate": 1.8798320335932814e-06, "loss": 1.102, "step": 15611 }, { "epoch": 1.87, "learning_rate": 1.879760047990402e-06, "loss": 0.0003, "step": 15612 }, { "epoch": 1.87, "learning_rate": 1.8796880623875227e-06, "loss": 0.0002, "step": 15613 }, { "epoch": 1.87, "learning_rate": 1.879616076784643e-06, "loss": 0.0002, "step": 15614 }, { "epoch": 1.87, "learning_rate": 1.8795440911817639e-06, "loss": 0.0092, "step": 15615 }, { "epoch": 1.87, "learning_rate": 1.8794721055788842e-06, "loss": 0.0007, "step": 15616 }, { "epoch": 1.87, "learning_rate": 1.8794001199760048e-06, "loss": 0.0007, "step": 15617 }, { "epoch": 1.87, "learning_rate": 1.8793281343731254e-06, "loss": 0.0017, "step": 15618 }, { "epoch": 1.87, "learning_rate": 1.879256148770246e-06, "loss": 0.0013, "step": 15619 }, { "epoch": 1.87, "learning_rate": 1.8791841631673667e-06, "loss": 0.0045, "step": 15620 }, { "epoch": 1.87, "learning_rate": 1.879112177564487e-06, "loss": 0.0003, "step": 15621 }, { "epoch": 1.87, "learning_rate": 1.8790401919616078e-06, "loss": 0.0001, "step": 15622 }, { "epoch": 1.87, "learning_rate": 1.8789682063587282e-06, "loss": 0.0002, "step": 15623 }, { "epoch": 1.87, "learning_rate": 1.878896220755849e-06, "loss": 0.0003, "step": 15624 }, { "epoch": 1.87, "learning_rate": 1.8788242351529693e-06, "loss": 0.0001, "step": 15625 }, { "epoch": 1.87, "learning_rate": 1.8787522495500901e-06, "loss": 0.0001, "step": 15626 }, { "epoch": 1.87, "learning_rate": 1.8786802639472107e-06, "loss": 0.0011, "step": 15627 }, { "epoch": 1.87, "learning_rate": 1.878608278344331e-06, "loss": 0.0002, "step": 15628 }, { "epoch": 1.87, "learning_rate": 1.8785362927414518e-06, "loss": 0.0003, "step": 15629 }, { "epoch": 1.87, "learning_rate": 1.8784643071385722e-06, "loss": 0.0025, "step": 15630 }, { "epoch": 1.87, "learning_rate": 1.878392321535693e-06, "loss": 0.0002, "step": 15631 }, { "epoch": 1.87, "learning_rate": 1.8783203359328133e-06, "loss": 0.0002, "step": 15632 }, { "epoch": 1.87, "learning_rate": 1.878248350329934e-06, "loss": 0.0002, "step": 15633 }, { "epoch": 1.87, "learning_rate": 1.8781763647270546e-06, "loss": 1.0831, "step": 15634 }, { "epoch": 1.87, "learning_rate": 1.8781043791241754e-06, "loss": 0.0003, "step": 15635 }, { "epoch": 1.87, "learning_rate": 1.8780323935212958e-06, "loss": 0.0004, "step": 15636 }, { "epoch": 1.87, "learning_rate": 1.8779604079184166e-06, "loss": 0.0003, "step": 15637 }, { "epoch": 1.87, "learning_rate": 1.877888422315537e-06, "loss": 0.0004, "step": 15638 }, { "epoch": 1.87, "learning_rate": 1.8778164367126577e-06, "loss": 0.0004, "step": 15639 }, { "epoch": 1.87, "learning_rate": 1.877744451109778e-06, "loss": 0.0002, "step": 15640 }, { "epoch": 1.87, "learning_rate": 1.8776724655068986e-06, "loss": 0.0002, "step": 15641 }, { "epoch": 1.87, "learning_rate": 1.8776004799040194e-06, "loss": 0.0006, "step": 15642 }, { "epoch": 1.87, "learning_rate": 1.8775284943011397e-06, "loss": 0.0002, "step": 15643 }, { "epoch": 1.87, "learning_rate": 1.8774565086982605e-06, "loss": 0.0004, "step": 15644 }, { "epoch": 1.87, "learning_rate": 1.8773845230953809e-06, "loss": 0.0002, "step": 15645 }, { "epoch": 1.87, "learning_rate": 1.8773125374925017e-06, "loss": 0.0003, "step": 15646 }, { "epoch": 1.88, "learning_rate": 1.877240551889622e-06, "loss": 0.0002, "step": 15647 }, { "epoch": 1.88, "learning_rate": 1.8771685662867428e-06, "loss": 0.0002, "step": 15648 }, { "epoch": 1.88, "learning_rate": 1.8770965806838634e-06, "loss": 0.0008, "step": 15649 }, { "epoch": 1.88, "learning_rate": 1.877024595080984e-06, "loss": 0.0003, "step": 15650 }, { "epoch": 1.88, "learning_rate": 1.8769526094781045e-06, "loss": 0.0005, "step": 15651 }, { "epoch": 1.88, "learning_rate": 1.8768806238752248e-06, "loss": 0.0009, "step": 15652 }, { "epoch": 1.88, "learning_rate": 1.8768086382723456e-06, "loss": 0.0001, "step": 15653 }, { "epoch": 1.88, "learning_rate": 1.876736652669466e-06, "loss": 0.0002, "step": 15654 }, { "epoch": 1.88, "learning_rate": 1.8766646670665868e-06, "loss": 0.0002, "step": 15655 }, { "epoch": 1.88, "learning_rate": 1.8765926814637071e-06, "loss": 0.0002, "step": 15656 }, { "epoch": 1.88, "learning_rate": 1.8765206958608279e-06, "loss": 0.0002, "step": 15657 }, { "epoch": 1.88, "learning_rate": 1.8764487102579485e-06, "loss": 0.0003, "step": 15658 }, { "epoch": 1.88, "learning_rate": 1.8763767246550692e-06, "loss": 0.0003, "step": 15659 }, { "epoch": 1.88, "learning_rate": 1.8763047390521896e-06, "loss": 0.0002, "step": 15660 }, { "epoch": 1.88, "learning_rate": 1.8762327534493104e-06, "loss": 0.0005, "step": 15661 }, { "epoch": 1.88, "learning_rate": 1.8761607678464307e-06, "loss": 0.0716, "step": 15662 }, { "epoch": 1.88, "learning_rate": 1.876088782243551e-06, "loss": 0.0002, "step": 15663 }, { "epoch": 1.88, "learning_rate": 1.8760167966406719e-06, "loss": 0.0002, "step": 15664 }, { "epoch": 1.88, "learning_rate": 1.8759448110377924e-06, "loss": 0.0001, "step": 15665 }, { "epoch": 1.88, "learning_rate": 1.8758728254349132e-06, "loss": 0.0005, "step": 15666 }, { "epoch": 1.88, "learning_rate": 1.8758008398320336e-06, "loss": 0.0001, "step": 15667 }, { "epoch": 1.88, "learning_rate": 1.8757288542291543e-06, "loss": 0.0003, "step": 15668 }, { "epoch": 1.88, "learning_rate": 1.8756568686262747e-06, "loss": 0.0002, "step": 15669 }, { "epoch": 1.88, "learning_rate": 1.8755848830233955e-06, "loss": 0.0002, "step": 15670 }, { "epoch": 1.88, "learning_rate": 1.8755128974205158e-06, "loss": 0.0003, "step": 15671 }, { "epoch": 1.88, "learning_rate": 1.8754409118176366e-06, "loss": 0.0002, "step": 15672 }, { "epoch": 1.88, "learning_rate": 1.8753689262147572e-06, "loss": 0.0003, "step": 15673 }, { "epoch": 1.88, "learning_rate": 1.8752969406118775e-06, "loss": 0.0005, "step": 15674 }, { "epoch": 1.88, "learning_rate": 1.8752249550089983e-06, "loss": 0.0008, "step": 15675 }, { "epoch": 1.88, "learning_rate": 1.8751529694061187e-06, "loss": 0.0002, "step": 15676 }, { "epoch": 1.88, "learning_rate": 1.8750809838032394e-06, "loss": 0.0233, "step": 15677 }, { "epoch": 1.88, "learning_rate": 1.8750089982003598e-06, "loss": 0.0002, "step": 15678 }, { "epoch": 1.88, "learning_rate": 1.8749370125974806e-06, "loss": 0.3057, "step": 15679 }, { "epoch": 1.88, "learning_rate": 1.8748650269946011e-06, "loss": 0.0003, "step": 15680 }, { "epoch": 1.88, "learning_rate": 1.8747930413917217e-06, "loss": 0.0004, "step": 15681 }, { "epoch": 1.88, "learning_rate": 1.8747210557888423e-06, "loss": 0.0002, "step": 15682 }, { "epoch": 1.88, "learning_rate": 1.874649070185963e-06, "loss": 0.0002, "step": 15683 }, { "epoch": 1.88, "learning_rate": 1.8745770845830834e-06, "loss": 0.0011, "step": 15684 }, { "epoch": 1.88, "learning_rate": 1.8745050989802042e-06, "loss": 0.0001, "step": 15685 }, { "epoch": 1.88, "learning_rate": 1.8744331133773245e-06, "loss": 0.0002, "step": 15686 }, { "epoch": 1.88, "learning_rate": 1.874361127774445e-06, "loss": 0.0002, "step": 15687 }, { "epoch": 1.88, "learning_rate": 1.8742891421715657e-06, "loss": 0.0001, "step": 15688 }, { "epoch": 1.88, "learning_rate": 1.8742171565686862e-06, "loss": 0.0002, "step": 15689 }, { "epoch": 1.88, "learning_rate": 1.874145170965807e-06, "loss": 0.0003, "step": 15690 }, { "epoch": 1.88, "learning_rate": 1.8740731853629274e-06, "loss": 0.0001, "step": 15691 }, { "epoch": 1.88, "learning_rate": 1.8740011997600481e-06, "loss": 0.0002, "step": 15692 }, { "epoch": 1.88, "learning_rate": 1.8739292141571685e-06, "loss": 0.0004, "step": 15693 }, { "epoch": 1.88, "learning_rate": 1.8738572285542893e-06, "loss": 0.0003, "step": 15694 }, { "epoch": 1.88, "learning_rate": 1.8737852429514096e-06, "loss": 0.0005, "step": 15695 }, { "epoch": 1.88, "learning_rate": 1.8737132573485304e-06, "loss": 0.0001, "step": 15696 }, { "epoch": 1.88, "learning_rate": 1.873641271745651e-06, "loss": 0.8625, "step": 15697 }, { "epoch": 1.88, "learning_rate": 1.8735692861427713e-06, "loss": 0.0002, "step": 15698 }, { "epoch": 1.88, "learning_rate": 1.8734973005398921e-06, "loss": 0.0003, "step": 15699 }, { "epoch": 1.88, "learning_rate": 1.8734253149370125e-06, "loss": 0.0002, "step": 15700 }, { "epoch": 1.88, "learning_rate": 1.8733533293341333e-06, "loss": 0.0003, "step": 15701 }, { "epoch": 1.88, "learning_rate": 1.8732813437312536e-06, "loss": 0.0001, "step": 15702 }, { "epoch": 1.88, "learning_rate": 1.8732093581283744e-06, "loss": 0.6525, "step": 15703 }, { "epoch": 1.88, "learning_rate": 1.873137372525495e-06, "loss": 0.0003, "step": 15704 }, { "epoch": 1.88, "learning_rate": 1.8730653869226157e-06, "loss": 0.0002, "step": 15705 }, { "epoch": 1.88, "learning_rate": 1.872993401319736e-06, "loss": 0.0004, "step": 15706 }, { "epoch": 1.88, "learning_rate": 1.8729214157168569e-06, "loss": 0.0006, "step": 15707 }, { "epoch": 1.88, "learning_rate": 1.8728494301139772e-06, "loss": 0.0002, "step": 15708 }, { "epoch": 1.88, "learning_rate": 1.8727774445110976e-06, "loss": 0.0002, "step": 15709 }, { "epoch": 1.88, "learning_rate": 1.8727054589082184e-06, "loss": 0.0001, "step": 15710 }, { "epoch": 1.88, "learning_rate": 1.872633473305339e-06, "loss": 0.0002, "step": 15711 }, { "epoch": 1.88, "learning_rate": 1.8725614877024597e-06, "loss": 0.0004, "step": 15712 }, { "epoch": 1.88, "learning_rate": 1.87248950209958e-06, "loss": 0.0002, "step": 15713 }, { "epoch": 1.88, "learning_rate": 1.8724175164967008e-06, "loss": 0.0003, "step": 15714 }, { "epoch": 1.88, "learning_rate": 1.8723455308938212e-06, "loss": 0.0003, "step": 15715 }, { "epoch": 1.88, "learning_rate": 1.872273545290942e-06, "loss": 0.0003, "step": 15716 }, { "epoch": 1.88, "learning_rate": 1.8722015596880623e-06, "loss": 0.0002, "step": 15717 }, { "epoch": 1.88, "learning_rate": 1.872129574085183e-06, "loss": 0.0002, "step": 15718 }, { "epoch": 1.88, "learning_rate": 1.8720575884823037e-06, "loss": 0.0002, "step": 15719 }, { "epoch": 1.88, "learning_rate": 1.871985602879424e-06, "loss": 0.0002, "step": 15720 }, { "epoch": 1.88, "learning_rate": 1.8719136172765448e-06, "loss": 0.0002, "step": 15721 }, { "epoch": 1.88, "learning_rate": 1.8718416316736652e-06, "loss": 0.0002, "step": 15722 }, { "epoch": 1.88, "learning_rate": 1.871769646070786e-06, "loss": 0.0001, "step": 15723 }, { "epoch": 1.88, "learning_rate": 1.8716976604679063e-06, "loss": 0.0001, "step": 15724 }, { "epoch": 1.88, "learning_rate": 1.871625674865027e-06, "loss": 0.0003, "step": 15725 }, { "epoch": 1.88, "learning_rate": 1.8715536892621476e-06, "loss": 0.0001, "step": 15726 }, { "epoch": 1.88, "learning_rate": 1.8714817036592682e-06, "loss": 0.0002, "step": 15727 }, { "epoch": 1.88, "learning_rate": 1.8714097180563888e-06, "loss": 0.0001, "step": 15728 }, { "epoch": 1.88, "learning_rate": 1.8713377324535095e-06, "loss": 0.0003, "step": 15729 }, { "epoch": 1.88, "learning_rate": 1.87126574685063e-06, "loss": 0.0003, "step": 15730 }, { "epoch": 1.89, "learning_rate": 1.8711937612477507e-06, "loss": 0.0025, "step": 15731 }, { "epoch": 1.89, "learning_rate": 1.871121775644871e-06, "loss": 0.0002, "step": 15732 }, { "epoch": 1.89, "learning_rate": 1.8710497900419916e-06, "loss": 0.0002, "step": 15733 }, { "epoch": 1.89, "learning_rate": 1.8709778044391122e-06, "loss": 0.0002, "step": 15734 }, { "epoch": 1.89, "learning_rate": 1.8709058188362327e-06, "loss": 0.0002, "step": 15735 }, { "epoch": 1.89, "learning_rate": 1.8708338332333535e-06, "loss": 0.0003, "step": 15736 }, { "epoch": 1.89, "learning_rate": 1.8707618476304739e-06, "loss": 0.0002, "step": 15737 }, { "epoch": 1.89, "learning_rate": 1.8706898620275946e-06, "loss": 0.0002, "step": 15738 }, { "epoch": 1.89, "learning_rate": 1.870617876424715e-06, "loss": 0.0002, "step": 15739 }, { "epoch": 1.89, "learning_rate": 1.8705458908218358e-06, "loss": 0.0155, "step": 15740 }, { "epoch": 1.89, "learning_rate": 1.8704739052189561e-06, "loss": 0.005, "step": 15741 }, { "epoch": 1.89, "learning_rate": 1.870401919616077e-06, "loss": 0.0003, "step": 15742 }, { "epoch": 1.89, "learning_rate": 1.8703299340131975e-06, "loss": 0.0002, "step": 15743 }, { "epoch": 1.89, "learning_rate": 1.8702579484103178e-06, "loss": 0.0003, "step": 15744 }, { "epoch": 1.89, "learning_rate": 1.8701859628074386e-06, "loss": 0.0003, "step": 15745 }, { "epoch": 1.89, "learning_rate": 1.870113977204559e-06, "loss": 0.0002, "step": 15746 }, { "epoch": 1.89, "learning_rate": 1.8700419916016797e-06, "loss": 0.0003, "step": 15747 }, { "epoch": 1.89, "learning_rate": 1.8699700059988e-06, "loss": 0.0002, "step": 15748 }, { "epoch": 1.89, "learning_rate": 1.8698980203959209e-06, "loss": 0.0004, "step": 15749 }, { "epoch": 1.89, "learning_rate": 1.8698260347930414e-06, "loss": 0.0003, "step": 15750 }, { "epoch": 1.89, "learning_rate": 1.8697540491901622e-06, "loss": 0.0003, "step": 15751 }, { "epoch": 1.89, "learning_rate": 1.8696820635872826e-06, "loss": 0.0003, "step": 15752 }, { "epoch": 1.89, "learning_rate": 1.8696100779844034e-06, "loss": 0.0002, "step": 15753 }, { "epoch": 1.89, "learning_rate": 1.8695380923815237e-06, "loss": 0.0003, "step": 15754 }, { "epoch": 1.89, "learning_rate": 1.869466106778644e-06, "loss": 0.0003, "step": 15755 }, { "epoch": 1.89, "learning_rate": 1.8693941211757648e-06, "loss": 0.0063, "step": 15756 }, { "epoch": 1.89, "learning_rate": 1.8693221355728854e-06, "loss": 0.0002, "step": 15757 }, { "epoch": 1.89, "learning_rate": 1.8692501499700062e-06, "loss": 0.0001, "step": 15758 }, { "epoch": 1.89, "learning_rate": 1.8691781643671265e-06, "loss": 0.0005, "step": 15759 }, { "epoch": 1.89, "learning_rate": 1.8691061787642473e-06, "loss": 0.0002, "step": 15760 }, { "epoch": 1.89, "learning_rate": 1.8690341931613677e-06, "loss": 0.0002, "step": 15761 }, { "epoch": 1.89, "learning_rate": 1.8689622075584885e-06, "loss": 0.0002, "step": 15762 }, { "epoch": 1.89, "learning_rate": 1.8688902219556088e-06, "loss": 0.0001, "step": 15763 }, { "epoch": 1.89, "learning_rate": 1.8688182363527296e-06, "loss": 0.0002, "step": 15764 }, { "epoch": 1.89, "learning_rate": 1.8687462507498502e-06, "loss": 0.0002, "step": 15765 }, { "epoch": 1.89, "learning_rate": 1.8686742651469707e-06, "loss": 0.0002, "step": 15766 }, { "epoch": 1.89, "learning_rate": 1.8686022795440913e-06, "loss": 0.9172, "step": 15767 }, { "epoch": 1.89, "learning_rate": 1.8685302939412116e-06, "loss": 0.0003, "step": 15768 }, { "epoch": 1.89, "learning_rate": 1.8684583083383324e-06, "loss": 0.0001, "step": 15769 }, { "epoch": 1.89, "learning_rate": 1.8683863227354528e-06, "loss": 0.0082, "step": 15770 }, { "epoch": 1.89, "learning_rate": 1.8683143371325736e-06, "loss": 0.0002, "step": 15771 }, { "epoch": 1.89, "learning_rate": 1.8682423515296941e-06, "loss": 0.0001, "step": 15772 }, { "epoch": 1.89, "learning_rate": 1.8681703659268147e-06, "loss": 0.2429, "step": 15773 }, { "epoch": 1.89, "learning_rate": 1.8680983803239353e-06, "loss": 0.0002, "step": 15774 }, { "epoch": 1.89, "learning_rate": 1.868026394721056e-06, "loss": 0.0001, "step": 15775 }, { "epoch": 1.89, "learning_rate": 1.8679544091181764e-06, "loss": 0.0003, "step": 15776 }, { "epoch": 1.89, "learning_rate": 1.8678824235152972e-06, "loss": 0.9456, "step": 15777 }, { "epoch": 1.89, "learning_rate": 1.8678104379124175e-06, "loss": 0.0004, "step": 15778 }, { "epoch": 1.89, "learning_rate": 1.867738452309538e-06, "loss": 0.0002, "step": 15779 }, { "epoch": 1.89, "learning_rate": 1.8676664667066587e-06, "loss": 0.0006, "step": 15780 }, { "epoch": 1.89, "learning_rate": 1.8675944811037792e-06, "loss": 0.0002, "step": 15781 }, { "epoch": 1.89, "learning_rate": 1.8675224955009e-06, "loss": 0.0002, "step": 15782 }, { "epoch": 1.89, "learning_rate": 1.8674505098980204e-06, "loss": 0.0002, "step": 15783 }, { "epoch": 1.89, "learning_rate": 1.8673785242951411e-06, "loss": 0.8504, "step": 15784 }, { "epoch": 1.89, "learning_rate": 1.8673065386922615e-06, "loss": 0.0002, "step": 15785 }, { "epoch": 1.89, "learning_rate": 1.8672345530893823e-06, "loss": 0.8483, "step": 15786 }, { "epoch": 1.89, "learning_rate": 1.8671625674865026e-06, "loss": 0.0002, "step": 15787 }, { "epoch": 1.89, "learning_rate": 1.8670905818836234e-06, "loss": 0.0002, "step": 15788 }, { "epoch": 1.89, "learning_rate": 1.867018596280744e-06, "loss": 0.0001, "step": 15789 }, { "epoch": 1.89, "learning_rate": 1.8669466106778643e-06, "loss": 0.0002, "step": 15790 }, { "epoch": 1.89, "learning_rate": 1.8668746250749851e-06, "loss": 0.0003, "step": 15791 }, { "epoch": 1.89, "learning_rate": 1.8668026394721055e-06, "loss": 0.0005, "step": 15792 }, { "epoch": 1.89, "learning_rate": 1.8667306538692262e-06, "loss": 0.0003, "step": 15793 }, { "epoch": 1.89, "learning_rate": 1.8666586682663466e-06, "loss": 0.0018, "step": 15794 }, { "epoch": 1.89, "learning_rate": 1.8665866826634674e-06, "loss": 0.0002, "step": 15795 }, { "epoch": 1.89, "learning_rate": 1.866514697060588e-06, "loss": 0.0011, "step": 15796 }, { "epoch": 1.89, "learning_rate": 1.8664427114577087e-06, "loss": 0.0009, "step": 15797 }, { "epoch": 1.89, "learning_rate": 1.866370725854829e-06, "loss": 0.0003, "step": 15798 }, { "epoch": 1.89, "learning_rate": 1.8662987402519499e-06, "loss": 0.0005, "step": 15799 }, { "epoch": 1.89, "learning_rate": 1.8662267546490702e-06, "loss": 1.1348, "step": 15800 }, { "epoch": 1.89, "learning_rate": 1.8661547690461906e-06, "loss": 0.0016, "step": 15801 }, { "epoch": 1.89, "learning_rate": 1.8660827834433113e-06, "loss": 0.0003, "step": 15802 }, { "epoch": 1.89, "learning_rate": 1.866010797840432e-06, "loss": 0.0002, "step": 15803 }, { "epoch": 1.89, "learning_rate": 1.8659388122375525e-06, "loss": 0.0007, "step": 15804 }, { "epoch": 1.89, "learning_rate": 1.865866826634673e-06, "loss": 0.6787, "step": 15805 }, { "epoch": 1.89, "learning_rate": 1.8657948410317938e-06, "loss": 0.0007, "step": 15806 }, { "epoch": 1.89, "learning_rate": 1.8657228554289142e-06, "loss": 0.0003, "step": 15807 }, { "epoch": 1.89, "learning_rate": 1.865650869826035e-06, "loss": 0.0004, "step": 15808 }, { "epoch": 1.89, "learning_rate": 1.8655788842231553e-06, "loss": 0.0005, "step": 15809 }, { "epoch": 1.89, "learning_rate": 1.865506898620276e-06, "loss": 0.0003, "step": 15810 }, { "epoch": 1.89, "learning_rate": 1.8654349130173964e-06, "loss": 0.0001, "step": 15811 }, { "epoch": 1.89, "learning_rate": 1.8653629274145172e-06, "loss": 0.0001, "step": 15812 }, { "epoch": 1.89, "learning_rate": 1.8652909418116378e-06, "loss": 0.0003, "step": 15813 }, { "epoch": 1.9, "learning_rate": 1.8652189562087581e-06, "loss": 0.823, "step": 15814 }, { "epoch": 1.9, "learning_rate": 1.865146970605879e-06, "loss": 0.0003, "step": 15815 }, { "epoch": 1.9, "learning_rate": 1.8650749850029993e-06, "loss": 0.0003, "step": 15816 }, { "epoch": 1.9, "learning_rate": 1.86500299940012e-06, "loss": 0.0004, "step": 15817 }, { "epoch": 1.9, "learning_rate": 1.8649310137972404e-06, "loss": 0.0003, "step": 15818 }, { "epoch": 1.9, "learning_rate": 1.8648590281943612e-06, "loss": 0.0002, "step": 15819 }, { "epoch": 1.9, "learning_rate": 1.8647870425914818e-06, "loss": 0.0036, "step": 15820 }, { "epoch": 1.9, "learning_rate": 1.8647150569886025e-06, "loss": 0.0004, "step": 15821 }, { "epoch": 1.9, "learning_rate": 1.8646430713857229e-06, "loss": 0.0003, "step": 15822 }, { "epoch": 1.9, "learning_rate": 1.8645710857828437e-06, "loss": 0.0002, "step": 15823 }, { "epoch": 1.9, "learning_rate": 1.864499100179964e-06, "loss": 0.0003, "step": 15824 }, { "epoch": 1.9, "learning_rate": 1.8644271145770844e-06, "loss": 0.0018, "step": 15825 }, { "epoch": 1.9, "learning_rate": 1.8643551289742052e-06, "loss": 1.2057, "step": 15826 }, { "epoch": 1.9, "learning_rate": 1.8642831433713257e-06, "loss": 0.0003, "step": 15827 }, { "epoch": 1.9, "learning_rate": 1.8642111577684465e-06, "loss": 0.0003, "step": 15828 }, { "epoch": 1.9, "learning_rate": 1.8641391721655669e-06, "loss": 0.0002, "step": 15829 }, { "epoch": 1.9, "learning_rate": 1.8640671865626876e-06, "loss": 0.0003, "step": 15830 }, { "epoch": 1.9, "learning_rate": 1.863995200959808e-06, "loss": 0.2098, "step": 15831 }, { "epoch": 1.9, "learning_rate": 1.8639232153569288e-06, "loss": 0.0005, "step": 15832 }, { "epoch": 1.9, "learning_rate": 1.8638512297540491e-06, "loss": 0.0008, "step": 15833 }, { "epoch": 1.9, "learning_rate": 1.86377924415117e-06, "loss": 0.0002, "step": 15834 }, { "epoch": 1.9, "learning_rate": 1.8637072585482905e-06, "loss": 0.0006, "step": 15835 }, { "epoch": 1.9, "learning_rate": 1.8636352729454108e-06, "loss": 0.0003, "step": 15836 }, { "epoch": 1.9, "learning_rate": 1.8635632873425316e-06, "loss": 0.0004, "step": 15837 }, { "epoch": 1.9, "learning_rate": 1.863491301739652e-06, "loss": 0.0005, "step": 15838 }, { "epoch": 1.9, "learning_rate": 1.8634193161367727e-06, "loss": 0.0002, "step": 15839 }, { "epoch": 1.9, "learning_rate": 1.863347330533893e-06, "loss": 0.0002, "step": 15840 }, { "epoch": 1.9, "learning_rate": 1.8632753449310139e-06, "loss": 0.0003, "step": 15841 }, { "epoch": 1.9, "learning_rate": 1.8632033593281344e-06, "loss": 0.0005, "step": 15842 }, { "epoch": 1.9, "learning_rate": 1.863131373725255e-06, "loss": 0.0004, "step": 15843 }, { "epoch": 1.9, "learning_rate": 1.8630593881223756e-06, "loss": 0.0002, "step": 15844 }, { "epoch": 1.9, "learning_rate": 1.8629874025194964e-06, "loss": 0.0001, "step": 15845 }, { "epoch": 1.9, "learning_rate": 1.8629154169166167e-06, "loss": 0.0003, "step": 15846 }, { "epoch": 1.9, "learning_rate": 1.8628434313137375e-06, "loss": 0.0002, "step": 15847 }, { "epoch": 1.9, "learning_rate": 1.8627714457108578e-06, "loss": 0.0002, "step": 15848 }, { "epoch": 1.9, "learning_rate": 1.8626994601079784e-06, "loss": 0.4758, "step": 15849 }, { "epoch": 1.9, "learning_rate": 1.862627474505099e-06, "loss": 0.0002, "step": 15850 }, { "epoch": 1.9, "learning_rate": 1.8625554889022195e-06, "loss": 0.0006, "step": 15851 }, { "epoch": 1.9, "learning_rate": 1.8624835032993403e-06, "loss": 0.0003, "step": 15852 }, { "epoch": 1.9, "learning_rate": 1.8624115176964607e-06, "loss": 0.0002, "step": 15853 }, { "epoch": 1.9, "learning_rate": 1.8623395320935815e-06, "loss": 0.0006, "step": 15854 }, { "epoch": 1.9, "learning_rate": 1.8622675464907018e-06, "loss": 0.0004, "step": 15855 }, { "epoch": 1.9, "learning_rate": 1.8621955608878226e-06, "loss": 0.0001, "step": 15856 }, { "epoch": 1.9, "learning_rate": 1.862123575284943e-06, "loss": 0.0003, "step": 15857 }, { "epoch": 1.9, "learning_rate": 1.8620515896820637e-06, "loss": 0.0007, "step": 15858 }, { "epoch": 1.9, "learning_rate": 1.8619796040791843e-06, "loss": 0.0003, "step": 15859 }, { "epoch": 1.9, "learning_rate": 1.8619076184763046e-06, "loss": 0.1232, "step": 15860 }, { "epoch": 1.9, "learning_rate": 1.8618356328734254e-06, "loss": 0.0002, "step": 15861 }, { "epoch": 1.9, "learning_rate": 1.8617636472705458e-06, "loss": 0.0005, "step": 15862 }, { "epoch": 1.9, "learning_rate": 1.8616916616676666e-06, "loss": 0.0014, "step": 15863 }, { "epoch": 1.9, "learning_rate": 1.861619676064787e-06, "loss": 0.0002, "step": 15864 }, { "epoch": 1.9, "learning_rate": 1.8615476904619077e-06, "loss": 0.0002, "step": 15865 }, { "epoch": 1.9, "learning_rate": 1.8614757048590283e-06, "loss": 0.3905, "step": 15866 }, { "epoch": 1.9, "learning_rate": 1.861403719256149e-06, "loss": 0.0003, "step": 15867 }, { "epoch": 1.9, "learning_rate": 1.8613317336532694e-06, "loss": 0.0003, "step": 15868 }, { "epoch": 1.9, "learning_rate": 1.8612597480503902e-06, "loss": 0.4702, "step": 15869 }, { "epoch": 1.9, "learning_rate": 1.8611877624475105e-06, "loss": 0.0002, "step": 15870 }, { "epoch": 1.9, "learning_rate": 1.8611157768446309e-06, "loss": 0.0002, "step": 15871 }, { "epoch": 1.9, "learning_rate": 1.8610437912417517e-06, "loss": 0.373, "step": 15872 }, { "epoch": 1.9, "learning_rate": 1.8609718056388722e-06, "loss": 0.0002, "step": 15873 }, { "epoch": 1.9, "learning_rate": 1.860899820035993e-06, "loss": 0.0004, "step": 15874 }, { "epoch": 1.9, "learning_rate": 1.8608278344331134e-06, "loss": 0.0002, "step": 15875 }, { "epoch": 1.9, "learning_rate": 1.8607558488302341e-06, "loss": 0.0003, "step": 15876 }, { "epoch": 1.9, "learning_rate": 1.8606838632273545e-06, "loss": 0.0008, "step": 15877 }, { "epoch": 1.9, "learning_rate": 1.8606118776244753e-06, "loss": 0.0003, "step": 15878 }, { "epoch": 1.9, "learning_rate": 1.8605398920215956e-06, "loss": 0.0004, "step": 15879 }, { "epoch": 1.9, "learning_rate": 1.8604679064187164e-06, "loss": 0.0004, "step": 15880 }, { "epoch": 1.9, "learning_rate": 1.860395920815837e-06, "loss": 0.0002, "step": 15881 }, { "epoch": 1.9, "learning_rate": 1.8603239352129573e-06, "loss": 0.0003, "step": 15882 }, { "epoch": 1.9, "learning_rate": 1.860251949610078e-06, "loss": 0.0004, "step": 15883 }, { "epoch": 1.9, "learning_rate": 1.8601799640071985e-06, "loss": 0.0003, "step": 15884 }, { "epoch": 1.9, "learning_rate": 1.8601079784043192e-06, "loss": 0.1634, "step": 15885 }, { "epoch": 1.9, "learning_rate": 1.8600359928014396e-06, "loss": 0.0004, "step": 15886 }, { "epoch": 1.9, "learning_rate": 1.8599640071985604e-06, "loss": 0.0003, "step": 15887 }, { "epoch": 1.9, "learning_rate": 1.859892021595681e-06, "loss": 0.0002, "step": 15888 }, { "epoch": 1.9, "learning_rate": 1.8598200359928015e-06, "loss": 0.0006, "step": 15889 }, { "epoch": 1.9, "learning_rate": 1.859748050389922e-06, "loss": 0.0003, "step": 15890 }, { "epoch": 1.9, "learning_rate": 1.8596760647870428e-06, "loss": 0.0002, "step": 15891 }, { "epoch": 1.9, "learning_rate": 1.8596040791841632e-06, "loss": 0.0002, "step": 15892 }, { "epoch": 1.9, "learning_rate": 1.859532093581284e-06, "loss": 0.0002, "step": 15893 }, { "epoch": 1.9, "learning_rate": 1.8594601079784043e-06, "loss": 0.0002, "step": 15894 }, { "epoch": 1.9, "learning_rate": 1.859388122375525e-06, "loss": 0.0008, "step": 15895 }, { "epoch": 1.9, "learning_rate": 1.8593161367726455e-06, "loss": 0.0002, "step": 15896 }, { "epoch": 1.9, "learning_rate": 1.859244151169766e-06, "loss": 0.0003, "step": 15897 }, { "epoch": 1.91, "learning_rate": 1.8591721655668868e-06, "loss": 0.0002, "step": 15898 }, { "epoch": 1.91, "learning_rate": 1.8591001799640072e-06, "loss": 0.0004, "step": 15899 }, { "epoch": 1.91, "learning_rate": 1.859028194361128e-06, "loss": 0.0003, "step": 15900 }, { "epoch": 1.91, "learning_rate": 1.8589562087582483e-06, "loss": 0.0001, "step": 15901 }, { "epoch": 1.91, "learning_rate": 1.858884223155369e-06, "loss": 0.0208, "step": 15902 }, { "epoch": 1.91, "learning_rate": 1.8588122375524894e-06, "loss": 0.0003, "step": 15903 }, { "epoch": 1.91, "learning_rate": 1.8587402519496102e-06, "loss": 0.0102, "step": 15904 }, { "epoch": 1.91, "learning_rate": 1.8586682663467308e-06, "loss": 0.0002, "step": 15905 }, { "epoch": 1.91, "learning_rate": 1.8585962807438511e-06, "loss": 0.0002, "step": 15906 }, { "epoch": 1.91, "learning_rate": 1.858524295140972e-06, "loss": 0.0001, "step": 15907 }, { "epoch": 1.91, "learning_rate": 1.8584523095380923e-06, "loss": 0.0004, "step": 15908 }, { "epoch": 1.91, "learning_rate": 1.858380323935213e-06, "loss": 0.0002, "step": 15909 }, { "epoch": 1.91, "learning_rate": 1.8583083383323334e-06, "loss": 0.0004, "step": 15910 }, { "epoch": 1.91, "learning_rate": 1.8582363527294542e-06, "loss": 0.0003, "step": 15911 }, { "epoch": 1.91, "learning_rate": 1.8581643671265748e-06, "loss": 0.0003, "step": 15912 }, { "epoch": 1.91, "learning_rate": 1.8580923815236955e-06, "loss": 0.0003, "step": 15913 }, { "epoch": 1.91, "learning_rate": 1.8580203959208159e-06, "loss": 0.0002, "step": 15914 }, { "epoch": 1.91, "learning_rate": 1.8579484103179367e-06, "loss": 0.0003, "step": 15915 }, { "epoch": 1.91, "learning_rate": 1.857876424715057e-06, "loss": 0.0003, "step": 15916 }, { "epoch": 1.91, "learning_rate": 1.8578044391121774e-06, "loss": 0.0003, "step": 15917 }, { "epoch": 1.91, "learning_rate": 1.8577324535092982e-06, "loss": 0.0003, "step": 15918 }, { "epoch": 1.91, "learning_rate": 1.8576604679064187e-06, "loss": 0.0003, "step": 15919 }, { "epoch": 1.91, "learning_rate": 1.8575884823035395e-06, "loss": 0.0002, "step": 15920 }, { "epoch": 1.91, "learning_rate": 1.8575164967006599e-06, "loss": 0.0003, "step": 15921 }, { "epoch": 1.91, "learning_rate": 1.8574445110977806e-06, "loss": 0.0003, "step": 15922 }, { "epoch": 1.91, "learning_rate": 1.857372525494901e-06, "loss": 0.0003, "step": 15923 }, { "epoch": 1.91, "learning_rate": 1.8573005398920218e-06, "loss": 0.0002, "step": 15924 }, { "epoch": 1.91, "learning_rate": 1.8572285542891421e-06, "loss": 0.0102, "step": 15925 }, { "epoch": 1.91, "learning_rate": 1.857156568686263e-06, "loss": 0.0002, "step": 15926 }, { "epoch": 1.91, "learning_rate": 1.8570845830833835e-06, "loss": 0.0003, "step": 15927 }, { "epoch": 1.91, "learning_rate": 1.8570125974805038e-06, "loss": 0.0002, "step": 15928 }, { "epoch": 1.91, "learning_rate": 1.8569406118776246e-06, "loss": 0.0002, "step": 15929 }, { "epoch": 1.91, "learning_rate": 1.856868626274745e-06, "loss": 0.0002, "step": 15930 }, { "epoch": 1.91, "learning_rate": 1.8567966406718657e-06, "loss": 0.7751, "step": 15931 }, { "epoch": 1.91, "learning_rate": 1.856724655068986e-06, "loss": 0.0002, "step": 15932 }, { "epoch": 1.91, "learning_rate": 1.8566526694661069e-06, "loss": 0.0003, "step": 15933 }, { "epoch": 1.91, "learning_rate": 1.8565806838632272e-06, "loss": 0.0002, "step": 15934 }, { "epoch": 1.91, "learning_rate": 1.856508698260348e-06, "loss": 0.0002, "step": 15935 }, { "epoch": 1.91, "learning_rate": 1.8564367126574686e-06, "loss": 0.0002, "step": 15936 }, { "epoch": 1.91, "learning_rate": 1.8563647270545893e-06, "loss": 0.0002, "step": 15937 }, { "epoch": 1.91, "learning_rate": 1.8562927414517097e-06, "loss": 0.0002, "step": 15938 }, { "epoch": 1.91, "learning_rate": 1.8562207558488305e-06, "loss": 0.0008, "step": 15939 }, { "epoch": 1.91, "learning_rate": 1.8561487702459508e-06, "loss": 0.0003, "step": 15940 }, { "epoch": 1.91, "learning_rate": 1.8560767846430712e-06, "loss": 0.0002, "step": 15941 }, { "epoch": 1.91, "learning_rate": 1.856004799040192e-06, "loss": 0.0002, "step": 15942 }, { "epoch": 1.91, "learning_rate": 1.8559328134373125e-06, "loss": 0.7364, "step": 15943 }, { "epoch": 1.91, "learning_rate": 1.8558608278344333e-06, "loss": 0.0002, "step": 15944 }, { "epoch": 1.91, "learning_rate": 1.8557888422315537e-06, "loss": 0.0002, "step": 15945 }, { "epoch": 1.91, "learning_rate": 1.8557168566286744e-06, "loss": 0.2667, "step": 15946 }, { "epoch": 1.91, "learning_rate": 1.8556448710257948e-06, "loss": 0.0002, "step": 15947 }, { "epoch": 1.91, "learning_rate": 1.8555728854229156e-06, "loss": 0.0002, "step": 15948 }, { "epoch": 1.91, "learning_rate": 1.855500899820036e-06, "loss": 0.0004, "step": 15949 }, { "epoch": 1.91, "learning_rate": 1.8554289142171567e-06, "loss": 0.0002, "step": 15950 }, { "epoch": 1.91, "learning_rate": 1.8553569286142773e-06, "loss": 0.0001, "step": 15951 }, { "epoch": 1.91, "learning_rate": 1.8552849430113976e-06, "loss": 0.0002, "step": 15952 }, { "epoch": 1.91, "learning_rate": 1.8552129574085184e-06, "loss": 0.0003, "step": 15953 }, { "epoch": 1.91, "learning_rate": 1.8551409718056388e-06, "loss": 0.0007, "step": 15954 }, { "epoch": 1.91, "learning_rate": 1.8550689862027595e-06, "loss": 0.0003, "step": 15955 }, { "epoch": 1.91, "learning_rate": 1.85499700059988e-06, "loss": 0.0004, "step": 15956 }, { "epoch": 1.91, "learning_rate": 1.8549250149970007e-06, "loss": 0.0004, "step": 15957 }, { "epoch": 1.91, "learning_rate": 1.8548530293941212e-06, "loss": 0.0002, "step": 15958 }, { "epoch": 1.91, "learning_rate": 1.8547810437912418e-06, "loss": 0.0002, "step": 15959 }, { "epoch": 1.91, "learning_rate": 1.8547090581883624e-06, "loss": 0.0002, "step": 15960 }, { "epoch": 1.91, "learning_rate": 1.8546370725854832e-06, "loss": 0.0005, "step": 15961 }, { "epoch": 1.91, "learning_rate": 1.8545650869826035e-06, "loss": 0.0002, "step": 15962 }, { "epoch": 1.91, "learning_rate": 1.8544931013797239e-06, "loss": 0.4314, "step": 15963 }, { "epoch": 1.91, "learning_rate": 1.8544211157768446e-06, "loss": 0.0003, "step": 15964 }, { "epoch": 1.91, "learning_rate": 1.8543491301739652e-06, "loss": 0.0002, "step": 15965 }, { "epoch": 1.91, "learning_rate": 1.8542771445710858e-06, "loss": 0.0002, "step": 15966 }, { "epoch": 1.91, "learning_rate": 1.8542051589682063e-06, "loss": 0.0003, "step": 15967 }, { "epoch": 1.91, "learning_rate": 1.8541331733653271e-06, "loss": 0.9553, "step": 15968 }, { "epoch": 1.91, "learning_rate": 1.8540611877624475e-06, "loss": 0.0002, "step": 15969 }, { "epoch": 1.91, "learning_rate": 1.8539892021595683e-06, "loss": 0.0002, "step": 15970 }, { "epoch": 1.91, "learning_rate": 1.8539172165566886e-06, "loss": 0.0004, "step": 15971 }, { "epoch": 1.91, "learning_rate": 1.8538452309538094e-06, "loss": 0.2706, "step": 15972 }, { "epoch": 1.91, "learning_rate": 1.8537732453509297e-06, "loss": 0.0003, "step": 15973 }, { "epoch": 1.91, "learning_rate": 1.8537012597480505e-06, "loss": 0.0002, "step": 15974 }, { "epoch": 1.91, "learning_rate": 1.853629274145171e-06, "loss": 0.0002, "step": 15975 }, { "epoch": 1.91, "learning_rate": 1.8535572885422914e-06, "loss": 0.0002, "step": 15976 }, { "epoch": 1.91, "learning_rate": 1.8534853029394122e-06, "loss": 0.0005, "step": 15977 }, { "epoch": 1.91, "learning_rate": 1.8534133173365326e-06, "loss": 0.0026, "step": 15978 }, { "epoch": 1.91, "learning_rate": 1.8533413317336534e-06, "loss": 0.0002, "step": 15979 }, { "epoch": 1.91, "learning_rate": 1.8532693461307737e-06, "loss": 0.0003, "step": 15980 }, { "epoch": 1.92, "learning_rate": 1.8531973605278945e-06, "loss": 0.053, "step": 15981 }, { "epoch": 1.92, "learning_rate": 1.853125374925015e-06, "loss": 0.0002, "step": 15982 }, { "epoch": 1.92, "learning_rate": 1.8530533893221358e-06, "loss": 0.0003, "step": 15983 }, { "epoch": 1.92, "learning_rate": 1.8529814037192562e-06, "loss": 0.0003, "step": 15984 }, { "epoch": 1.92, "learning_rate": 1.852909418116377e-06, "loss": 0.0004, "step": 15985 }, { "epoch": 1.92, "learning_rate": 1.8528374325134973e-06, "loss": 0.0002, "step": 15986 }, { "epoch": 1.92, "learning_rate": 1.8527654469106177e-06, "loss": 0.0002, "step": 15987 }, { "epoch": 1.92, "learning_rate": 1.8526934613077385e-06, "loss": 0.072, "step": 15988 }, { "epoch": 1.92, "learning_rate": 1.852621475704859e-06, "loss": 0.0005, "step": 15989 }, { "epoch": 1.92, "learning_rate": 1.8525494901019798e-06, "loss": 0.0002, "step": 15990 }, { "epoch": 1.92, "learning_rate": 1.8524775044991002e-06, "loss": 0.0002, "step": 15991 }, { "epoch": 1.92, "learning_rate": 1.852405518896221e-06, "loss": 0.0003, "step": 15992 }, { "epoch": 1.92, "learning_rate": 1.8523335332933413e-06, "loss": 0.0002, "step": 15993 }, { "epoch": 1.92, "learning_rate": 1.852261547690462e-06, "loss": 0.0002, "step": 15994 }, { "epoch": 1.92, "learning_rate": 1.8521895620875824e-06, "loss": 0.0003, "step": 15995 }, { "epoch": 1.92, "learning_rate": 1.8521175764847032e-06, "loss": 0.0003, "step": 15996 }, { "epoch": 1.92, "learning_rate": 1.8520455908818238e-06, "loss": 0.0004, "step": 15997 }, { "epoch": 1.92, "learning_rate": 1.8519736052789441e-06, "loss": 0.0003, "step": 15998 }, { "epoch": 1.92, "learning_rate": 1.851901619676065e-06, "loss": 0.0004, "step": 15999 }, { "epoch": 1.92, "learning_rate": 1.8518296340731853e-06, "loss": 0.0003, "step": 16000 }, { "epoch": 1.92, "learning_rate": 1.851757648470306e-06, "loss": 0.0002, "step": 16001 }, { "epoch": 1.92, "learning_rate": 1.8516856628674264e-06, "loss": 1.054, "step": 16002 }, { "epoch": 1.92, "learning_rate": 1.8516136772645472e-06, "loss": 0.0006, "step": 16003 }, { "epoch": 1.92, "learning_rate": 1.8515416916616677e-06, "loss": 0.0005, "step": 16004 }, { "epoch": 1.92, "learning_rate": 1.8514697060587883e-06, "loss": 0.0003, "step": 16005 }, { "epoch": 1.92, "learning_rate": 1.8513977204559089e-06, "loss": 0.0029, "step": 16006 }, { "epoch": 1.92, "learning_rate": 1.8513257348530297e-06, "loss": 0.0002, "step": 16007 }, { "epoch": 1.92, "learning_rate": 1.85125374925015e-06, "loss": 0.0003, "step": 16008 }, { "epoch": 1.92, "learning_rate": 1.8511817636472704e-06, "loss": 0.0003, "step": 16009 }, { "epoch": 1.92, "learning_rate": 1.8511097780443911e-06, "loss": 0.0004, "step": 16010 }, { "epoch": 1.92, "learning_rate": 1.8510377924415117e-06, "loss": 0.5157, "step": 16011 }, { "epoch": 1.92, "learning_rate": 1.8509658068386323e-06, "loss": 0.0004, "step": 16012 }, { "epoch": 1.92, "learning_rate": 1.8508938212357528e-06, "loss": 0.0002, "step": 16013 }, { "epoch": 1.92, "learning_rate": 1.8508218356328736e-06, "loss": 0.0005, "step": 16014 }, { "epoch": 1.92, "learning_rate": 1.850749850029994e-06, "loss": 0.0003, "step": 16015 }, { "epoch": 1.92, "learning_rate": 1.8506778644271148e-06, "loss": 0.0009, "step": 16016 }, { "epoch": 1.92, "learning_rate": 1.8506058788242351e-06, "loss": 0.0006, "step": 16017 }, { "epoch": 1.92, "learning_rate": 1.8505338932213559e-06, "loss": 0.0003, "step": 16018 }, { "epoch": 1.92, "learning_rate": 1.8504619076184762e-06, "loss": 0.0004, "step": 16019 }, { "epoch": 1.92, "learning_rate": 1.850389922015597e-06, "loss": 0.0004, "step": 16020 }, { "epoch": 1.92, "learning_rate": 1.8503179364127176e-06, "loss": 0.8555, "step": 16021 }, { "epoch": 1.92, "learning_rate": 1.850245950809838e-06, "loss": 0.0002, "step": 16022 }, { "epoch": 1.92, "learning_rate": 1.8501739652069587e-06, "loss": 0.0003, "step": 16023 }, { "epoch": 1.92, "learning_rate": 1.850101979604079e-06, "loss": 0.0001, "step": 16024 }, { "epoch": 1.92, "learning_rate": 1.8500299940011999e-06, "loss": 0.0002, "step": 16025 }, { "epoch": 1.92, "learning_rate": 1.8499580083983202e-06, "loss": 0.0006, "step": 16026 }, { "epoch": 1.92, "learning_rate": 1.849886022795441e-06, "loss": 0.0006, "step": 16027 }, { "epoch": 1.92, "learning_rate": 1.8498140371925616e-06, "loss": 0.0003, "step": 16028 }, { "epoch": 1.92, "learning_rate": 1.8497420515896823e-06, "loss": 1.0247, "step": 16029 }, { "epoch": 1.92, "learning_rate": 1.8496700659868027e-06, "loss": 0.0002, "step": 16030 }, { "epoch": 1.92, "learning_rate": 1.8495980803839235e-06, "loss": 0.0002, "step": 16031 }, { "epoch": 1.92, "learning_rate": 1.8495260947810438e-06, "loss": 0.0003, "step": 16032 }, { "epoch": 1.92, "learning_rate": 1.8494541091781642e-06, "loss": 0.0016, "step": 16033 }, { "epoch": 1.92, "learning_rate": 1.849382123575285e-06, "loss": 0.0003, "step": 16034 }, { "epoch": 1.92, "learning_rate": 1.8493101379724055e-06, "loss": 0.0005, "step": 16035 }, { "epoch": 1.92, "learning_rate": 1.8492381523695263e-06, "loss": 0.0005, "step": 16036 }, { "epoch": 1.92, "learning_rate": 1.8491661667666467e-06, "loss": 0.0003, "step": 16037 }, { "epoch": 1.92, "learning_rate": 1.8490941811637674e-06, "loss": 0.0003, "step": 16038 }, { "epoch": 1.92, "learning_rate": 1.8490221955608878e-06, "loss": 0.0001, "step": 16039 }, { "epoch": 1.92, "learning_rate": 1.8489502099580086e-06, "loss": 0.0004, "step": 16040 }, { "epoch": 1.92, "learning_rate": 1.848878224355129e-06, "loss": 0.0008, "step": 16041 }, { "epoch": 1.92, "learning_rate": 1.8488062387522497e-06, "loss": 0.0005, "step": 16042 }, { "epoch": 1.92, "learning_rate": 1.8487342531493703e-06, "loss": 0.0005, "step": 16043 }, { "epoch": 1.92, "learning_rate": 1.8486622675464906e-06, "loss": 0.0003, "step": 16044 }, { "epoch": 1.92, "learning_rate": 1.8485902819436114e-06, "loss": 0.0008, "step": 16045 }, { "epoch": 1.92, "learning_rate": 1.8485182963407318e-06, "loss": 0.0006, "step": 16046 }, { "epoch": 1.92, "learning_rate": 1.8484463107378525e-06, "loss": 0.0001, "step": 16047 }, { "epoch": 1.92, "learning_rate": 1.8483743251349729e-06, "loss": 0.9779, "step": 16048 }, { "epoch": 1.92, "learning_rate": 1.8483023395320937e-06, "loss": 0.0003, "step": 16049 }, { "epoch": 1.92, "learning_rate": 1.8482303539292142e-06, "loss": 0.0004, "step": 16050 }, { "epoch": 1.92, "learning_rate": 1.8481583683263348e-06, "loss": 0.0002, "step": 16051 }, { "epoch": 1.92, "learning_rate": 1.8480863827234554e-06, "loss": 0.0003, "step": 16052 }, { "epoch": 1.92, "learning_rate": 1.8480143971205762e-06, "loss": 0.0004, "step": 16053 }, { "epoch": 1.92, "learning_rate": 1.8479424115176965e-06, "loss": 0.0008, "step": 16054 }, { "epoch": 1.92, "learning_rate": 1.8478704259148169e-06, "loss": 0.0004, "step": 16055 }, { "epoch": 1.92, "learning_rate": 1.8477984403119376e-06, "loss": 0.0013, "step": 16056 }, { "epoch": 1.92, "learning_rate": 1.847726454709058e-06, "loss": 0.0004, "step": 16057 }, { "epoch": 1.92, "learning_rate": 1.8476544691061788e-06, "loss": 0.0004, "step": 16058 }, { "epoch": 1.92, "learning_rate": 1.8475824835032993e-06, "loss": 0.0009, "step": 16059 }, { "epoch": 1.92, "learning_rate": 1.8475104979004201e-06, "loss": 0.6707, "step": 16060 }, { "epoch": 1.92, "learning_rate": 1.8474385122975405e-06, "loss": 0.0969, "step": 16061 }, { "epoch": 1.92, "learning_rate": 1.8473665266946613e-06, "loss": 0.0003, "step": 16062 }, { "epoch": 1.92, "learning_rate": 1.8472945410917816e-06, "loss": 0.001, "step": 16063 }, { "epoch": 1.92, "learning_rate": 1.8472225554889024e-06, "loss": 0.0008, "step": 16064 }, { "epoch": 1.93, "learning_rate": 1.8471505698860227e-06, "loss": 0.0004, "step": 16065 }, { "epoch": 1.93, "learning_rate": 1.8470785842831435e-06, "loss": 0.0003, "step": 16066 }, { "epoch": 1.93, "learning_rate": 1.847006598680264e-06, "loss": 0.0002, "step": 16067 }, { "epoch": 1.93, "learning_rate": 1.8469346130773844e-06, "loss": 0.8904, "step": 16068 }, { "epoch": 1.93, "learning_rate": 1.8468626274745052e-06, "loss": 0.01, "step": 16069 }, { "epoch": 1.93, "learning_rate": 1.8467906418716256e-06, "loss": 0.0008, "step": 16070 }, { "epoch": 1.93, "learning_rate": 1.8467186562687464e-06, "loss": 0.0003, "step": 16071 }, { "epoch": 1.93, "learning_rate": 1.8466466706658667e-06, "loss": 0.0011, "step": 16072 }, { "epoch": 1.93, "learning_rate": 1.8465746850629875e-06, "loss": 0.0004, "step": 16073 }, { "epoch": 1.93, "learning_rate": 1.846502699460108e-06, "loss": 0.0004, "step": 16074 }, { "epoch": 1.93, "learning_rate": 1.8464307138572286e-06, "loss": 0.0008, "step": 16075 }, { "epoch": 1.93, "learning_rate": 1.8463587282543492e-06, "loss": 0.0007, "step": 16076 }, { "epoch": 1.93, "learning_rate": 1.84628674265147e-06, "loss": 0.0004, "step": 16077 }, { "epoch": 1.93, "learning_rate": 1.8462147570485903e-06, "loss": 0.0008, "step": 16078 }, { "epoch": 1.93, "learning_rate": 1.8461427714457107e-06, "loss": 0.0009, "step": 16079 }, { "epoch": 1.93, "learning_rate": 1.8460707858428315e-06, "loss": 0.0005, "step": 16080 }, { "epoch": 1.93, "learning_rate": 1.845998800239952e-06, "loss": 0.0008, "step": 16081 }, { "epoch": 1.93, "learning_rate": 1.8459268146370726e-06, "loss": 0.0006, "step": 16082 }, { "epoch": 1.93, "learning_rate": 1.8458548290341932e-06, "loss": 0.0047, "step": 16083 }, { "epoch": 1.93, "learning_rate": 1.845782843431314e-06, "loss": 0.0003, "step": 16084 }, { "epoch": 1.93, "learning_rate": 1.8457108578284343e-06, "loss": 0.0004, "step": 16085 }, { "epoch": 1.93, "learning_rate": 1.845638872225555e-06, "loss": 0.6425, "step": 16086 }, { "epoch": 1.93, "learning_rate": 1.8455668866226754e-06, "loss": 0.0003, "step": 16087 }, { "epoch": 1.93, "learning_rate": 1.8454949010197962e-06, "loss": 0.0015, "step": 16088 }, { "epoch": 1.93, "learning_rate": 1.8454229154169166e-06, "loss": 0.0027, "step": 16089 }, { "epoch": 1.93, "learning_rate": 1.8453509298140371e-06, "loss": 0.0015, "step": 16090 }, { "epoch": 1.93, "learning_rate": 1.845278944211158e-06, "loss": 0.0032, "step": 16091 }, { "epoch": 1.93, "learning_rate": 1.8452069586082783e-06, "loss": 0.0358, "step": 16092 }, { "epoch": 1.93, "learning_rate": 1.845134973005399e-06, "loss": 0.0002, "step": 16093 }, { "epoch": 1.93, "learning_rate": 1.8450629874025194e-06, "loss": 0.0092, "step": 16094 }, { "epoch": 1.93, "learning_rate": 1.8449910017996402e-06, "loss": 0.0003, "step": 16095 }, { "epoch": 1.93, "learning_rate": 1.8449190161967605e-06, "loss": 0.0007, "step": 16096 }, { "epoch": 1.93, "learning_rate": 1.8448470305938813e-06, "loss": 0.0015, "step": 16097 }, { "epoch": 1.93, "learning_rate": 1.8447750449910019e-06, "loss": 0.0002, "step": 16098 }, { "epoch": 1.93, "learning_rate": 1.8447030593881226e-06, "loss": 0.0003, "step": 16099 }, { "epoch": 1.93, "learning_rate": 1.844631073785243e-06, "loss": 0.0007, "step": 16100 }, { "epoch": 1.93, "learning_rate": 1.8445590881823638e-06, "loss": 0.0005, "step": 16101 }, { "epoch": 1.93, "learning_rate": 1.8444871025794841e-06, "loss": 0.5703, "step": 16102 }, { "epoch": 1.93, "learning_rate": 1.8444151169766045e-06, "loss": 0.0002, "step": 16103 }, { "epoch": 1.93, "learning_rate": 1.8443431313737253e-06, "loss": 0.0004, "step": 16104 }, { "epoch": 1.93, "learning_rate": 1.8442711457708458e-06, "loss": 0.0011, "step": 16105 }, { "epoch": 1.93, "learning_rate": 1.8441991601679666e-06, "loss": 0.0008, "step": 16106 }, { "epoch": 1.93, "learning_rate": 1.844127174565087e-06, "loss": 0.0002, "step": 16107 }, { "epoch": 1.93, "learning_rate": 1.8440551889622077e-06, "loss": 0.0009, "step": 16108 }, { "epoch": 1.93, "learning_rate": 1.843983203359328e-06, "loss": 0.0003, "step": 16109 }, { "epoch": 1.93, "learning_rate": 1.8439112177564489e-06, "loss": 0.0003, "step": 16110 }, { "epoch": 1.93, "learning_rate": 1.8438392321535692e-06, "loss": 0.001, "step": 16111 }, { "epoch": 1.93, "learning_rate": 1.84376724655069e-06, "loss": 0.0004, "step": 16112 }, { "epoch": 1.93, "learning_rate": 1.8436952609478106e-06, "loss": 0.0006, "step": 16113 }, { "epoch": 1.93, "learning_rate": 1.843623275344931e-06, "loss": 0.0004, "step": 16114 }, { "epoch": 1.93, "learning_rate": 1.8435512897420517e-06, "loss": 0.0007, "step": 16115 }, { "epoch": 1.93, "learning_rate": 1.843479304139172e-06, "loss": 0.0002, "step": 16116 }, { "epoch": 1.93, "learning_rate": 1.8434073185362928e-06, "loss": 0.0004, "step": 16117 }, { "epoch": 1.93, "learning_rate": 1.8433353329334132e-06, "loss": 0.0007, "step": 16118 }, { "epoch": 1.93, "learning_rate": 1.843263347330534e-06, "loss": 0.0004, "step": 16119 }, { "epoch": 1.93, "learning_rate": 1.8431913617276545e-06, "loss": 0.0004, "step": 16120 }, { "epoch": 1.93, "learning_rate": 1.8431193761247751e-06, "loss": 0.0003, "step": 16121 }, { "epoch": 1.93, "learning_rate": 1.8430473905218957e-06, "loss": 0.0004, "step": 16122 }, { "epoch": 1.93, "learning_rate": 1.8429754049190165e-06, "loss": 0.0002, "step": 16123 }, { "epoch": 1.93, "learning_rate": 1.8429034193161368e-06, "loss": 0.0009, "step": 16124 }, { "epoch": 1.93, "learning_rate": 1.8428314337132572e-06, "loss": 0.0004, "step": 16125 }, { "epoch": 1.93, "learning_rate": 1.842759448110378e-06, "loss": 0.0001, "step": 16126 }, { "epoch": 1.93, "learning_rate": 1.8426874625074985e-06, "loss": 0.0004, "step": 16127 }, { "epoch": 1.93, "learning_rate": 1.842615476904619e-06, "loss": 0.0002, "step": 16128 }, { "epoch": 1.93, "learning_rate": 1.8425434913017397e-06, "loss": 0.0002, "step": 16129 }, { "epoch": 1.93, "learning_rate": 1.8424715056988604e-06, "loss": 0.0002, "step": 16130 }, { "epoch": 1.93, "learning_rate": 1.8423995200959808e-06, "loss": 0.0003, "step": 16131 }, { "epoch": 1.93, "learning_rate": 1.8423275344931016e-06, "loss": 0.0003, "step": 16132 }, { "epoch": 1.93, "learning_rate": 1.842255548890222e-06, "loss": 0.0001, "step": 16133 }, { "epoch": 1.93, "learning_rate": 1.8421835632873427e-06, "loss": 0.0003, "step": 16134 }, { "epoch": 1.93, "learning_rate": 1.842111577684463e-06, "loss": 0.0002, "step": 16135 }, { "epoch": 1.93, "learning_rate": 1.8420395920815836e-06, "loss": 0.0003, "step": 16136 }, { "epoch": 1.93, "learning_rate": 1.8419676064787044e-06, "loss": 0.0003, "step": 16137 }, { "epoch": 1.93, "learning_rate": 1.8418956208758248e-06, "loss": 0.0002, "step": 16138 }, { "epoch": 1.93, "learning_rate": 1.8418236352729455e-06, "loss": 0.0001, "step": 16139 }, { "epoch": 1.93, "learning_rate": 1.8417516496700659e-06, "loss": 0.0003, "step": 16140 }, { "epoch": 1.93, "learning_rate": 1.8416796640671867e-06, "loss": 0.0005, "step": 16141 }, { "epoch": 1.93, "learning_rate": 1.841607678464307e-06, "loss": 0.0004, "step": 16142 }, { "epoch": 1.93, "learning_rate": 1.8415356928614278e-06, "loss": 0.0003, "step": 16143 }, { "epoch": 1.93, "learning_rate": 1.8414637072585484e-06, "loss": 0.0003, "step": 16144 }, { "epoch": 1.93, "learning_rate": 1.8413917216556691e-06, "loss": 0.0002, "step": 16145 }, { "epoch": 1.93, "learning_rate": 1.8413197360527895e-06, "loss": 0.0004, "step": 16146 }, { "epoch": 1.93, "learning_rate": 1.8412477504499103e-06, "loss": 0.0003, "step": 16147 }, { "epoch": 1.94, "learning_rate": 1.8411757648470306e-06, "loss": 0.0022, "step": 16148 }, { "epoch": 1.94, "learning_rate": 1.841103779244151e-06, "loss": 0.0004, "step": 16149 }, { "epoch": 1.94, "learning_rate": 1.8410317936412718e-06, "loss": 0.0016, "step": 16150 }, { "epoch": 1.94, "learning_rate": 1.8409598080383923e-06, "loss": 0.0002, "step": 16151 }, { "epoch": 1.94, "learning_rate": 1.8408878224355131e-06, "loss": 0.0003, "step": 16152 }, { "epoch": 1.94, "learning_rate": 1.8408158368326335e-06, "loss": 0.0003, "step": 16153 }, { "epoch": 1.94, "learning_rate": 1.8407438512297542e-06, "loss": 0.0003, "step": 16154 }, { "epoch": 1.94, "learning_rate": 1.8406718656268746e-06, "loss": 0.0004, "step": 16155 }, { "epoch": 1.94, "learning_rate": 1.8405998800239954e-06, "loss": 0.0009, "step": 16156 }, { "epoch": 1.94, "learning_rate": 1.8405278944211157e-06, "loss": 0.0004, "step": 16157 }, { "epoch": 1.94, "learning_rate": 1.8404559088182365e-06, "loss": 0.0003, "step": 16158 }, { "epoch": 1.94, "learning_rate": 1.840383923215357e-06, "loss": 0.0003, "step": 16159 }, { "epoch": 1.94, "learning_rate": 1.8403119376124774e-06, "loss": 0.0002, "step": 16160 }, { "epoch": 1.94, "learning_rate": 1.8402399520095982e-06, "loss": 0.0007, "step": 16161 }, { "epoch": 1.94, "learning_rate": 1.8401679664067186e-06, "loss": 0.0003, "step": 16162 }, { "epoch": 1.94, "learning_rate": 1.8400959808038393e-06, "loss": 0.0001, "step": 16163 }, { "epoch": 1.94, "learning_rate": 1.8400239952009597e-06, "loss": 0.0004, "step": 16164 }, { "epoch": 1.94, "learning_rate": 1.8399520095980805e-06, "loss": 0.0003, "step": 16165 }, { "epoch": 1.94, "learning_rate": 1.839880023995201e-06, "loss": 0.0005, "step": 16166 }, { "epoch": 1.94, "learning_rate": 1.8398080383923216e-06, "loss": 0.0003, "step": 16167 }, { "epoch": 1.94, "learning_rate": 1.8397360527894422e-06, "loss": 0.0003, "step": 16168 }, { "epoch": 1.94, "learning_rate": 1.839664067186563e-06, "loss": 0.0002, "step": 16169 }, { "epoch": 1.94, "learning_rate": 1.8395920815836833e-06, "loss": 0.0002, "step": 16170 }, { "epoch": 1.94, "learning_rate": 1.8395200959808037e-06, "loss": 0.0005, "step": 16171 }, { "epoch": 1.94, "learning_rate": 1.8394481103779244e-06, "loss": 0.0116, "step": 16172 }, { "epoch": 1.94, "learning_rate": 1.839376124775045e-06, "loss": 0.0006, "step": 16173 }, { "epoch": 1.94, "learning_rate": 1.8393041391721656e-06, "loss": 0.0004, "step": 16174 }, { "epoch": 1.94, "learning_rate": 1.8392321535692861e-06, "loss": 0.0005, "step": 16175 }, { "epoch": 1.94, "learning_rate": 1.839160167966407e-06, "loss": 0.0011, "step": 16176 }, { "epoch": 1.94, "learning_rate": 1.8390881823635273e-06, "loss": 0.0008, "step": 16177 }, { "epoch": 1.94, "learning_rate": 1.839016196760648e-06, "loss": 0.0002, "step": 16178 }, { "epoch": 1.94, "learning_rate": 1.8389442111577684e-06, "loss": 0.0003, "step": 16179 }, { "epoch": 1.94, "learning_rate": 1.8388722255548892e-06, "loss": 0.0005, "step": 16180 }, { "epoch": 1.94, "learning_rate": 1.8388002399520095e-06, "loss": 0.0002, "step": 16181 }, { "epoch": 1.94, "learning_rate": 1.8387282543491301e-06, "loss": 0.0002, "step": 16182 }, { "epoch": 1.94, "learning_rate": 1.8386562687462509e-06, "loss": 0.3668, "step": 16183 }, { "epoch": 1.94, "learning_rate": 1.8385842831433712e-06, "loss": 0.0003, "step": 16184 }, { "epoch": 1.94, "learning_rate": 1.838512297540492e-06, "loss": 0.0004, "step": 16185 }, { "epoch": 1.94, "learning_rate": 1.8384403119376124e-06, "loss": 0.8213, "step": 16186 }, { "epoch": 1.94, "learning_rate": 1.8383683263347332e-06, "loss": 0.0005, "step": 16187 }, { "epoch": 1.94, "learning_rate": 1.8382963407318535e-06, "loss": 1.0251, "step": 16188 }, { "epoch": 1.94, "learning_rate": 1.8382243551289743e-06, "loss": 0.8325, "step": 16189 }, { "epoch": 1.94, "learning_rate": 1.8381523695260949e-06, "loss": 0.0004, "step": 16190 }, { "epoch": 1.94, "learning_rate": 1.8380803839232156e-06, "loss": 0.0003, "step": 16191 }, { "epoch": 1.94, "learning_rate": 1.838008398320336e-06, "loss": 0.0003, "step": 16192 }, { "epoch": 1.94, "learning_rate": 1.8379364127174568e-06, "loss": 0.0002, "step": 16193 }, { "epoch": 1.94, "learning_rate": 1.8378644271145771e-06, "loss": 0.0001, "step": 16194 }, { "epoch": 1.94, "learning_rate": 1.8377924415116975e-06, "loss": 0.0004, "step": 16195 }, { "epoch": 1.94, "learning_rate": 1.8377204559088183e-06, "loss": 0.0009, "step": 16196 }, { "epoch": 1.94, "learning_rate": 1.8376484703059388e-06, "loss": 0.0003, "step": 16197 }, { "epoch": 1.94, "learning_rate": 1.8375764847030596e-06, "loss": 0.0005, "step": 16198 }, { "epoch": 1.94, "learning_rate": 1.83750449910018e-06, "loss": 0.0008, "step": 16199 }, { "epoch": 1.94, "learning_rate": 1.8374325134973007e-06, "loss": 0.6967, "step": 16200 }, { "epoch": 1.94, "learning_rate": 1.837360527894421e-06, "loss": 0.4112, "step": 16201 }, { "epoch": 1.94, "learning_rate": 1.8372885422915419e-06, "loss": 0.0002, "step": 16202 }, { "epoch": 1.94, "learning_rate": 1.8372165566886622e-06, "loss": 0.0006, "step": 16203 }, { "epoch": 1.94, "learning_rate": 1.837144571085783e-06, "loss": 0.0259, "step": 16204 }, { "epoch": 1.94, "learning_rate": 1.8370725854829034e-06, "loss": 0.0003, "step": 16205 }, { "epoch": 1.94, "learning_rate": 1.837000599880024e-06, "loss": 0.0007, "step": 16206 }, { "epoch": 1.94, "learning_rate": 1.8369286142771447e-06, "loss": 0.0002, "step": 16207 }, { "epoch": 1.94, "learning_rate": 1.836856628674265e-06, "loss": 0.0006, "step": 16208 }, { "epoch": 1.94, "learning_rate": 1.8367846430713858e-06, "loss": 0.0081, "step": 16209 }, { "epoch": 1.94, "learning_rate": 1.8367126574685062e-06, "loss": 0.0003, "step": 16210 }, { "epoch": 1.94, "learning_rate": 1.836640671865627e-06, "loss": 0.0002, "step": 16211 }, { "epoch": 1.94, "learning_rate": 1.8365686862627473e-06, "loss": 0.0004, "step": 16212 }, { "epoch": 1.94, "learning_rate": 1.8364967006598681e-06, "loss": 0.0003, "step": 16213 }, { "epoch": 1.94, "learning_rate": 1.8364247150569887e-06, "loss": 0.0002, "step": 16214 }, { "epoch": 1.94, "learning_rate": 1.8363527294541095e-06, "loss": 0.0008, "step": 16215 }, { "epoch": 1.94, "learning_rate": 1.8362807438512298e-06, "loss": 0.0003, "step": 16216 }, { "epoch": 1.94, "learning_rate": 1.8362087582483502e-06, "loss": 0.0002, "step": 16217 }, { "epoch": 1.94, "learning_rate": 1.836136772645471e-06, "loss": 0.0003, "step": 16218 }, { "epoch": 1.94, "learning_rate": 1.8360647870425913e-06, "loss": 0.0003, "step": 16219 }, { "epoch": 1.94, "learning_rate": 1.835992801439712e-06, "loss": 0.0005, "step": 16220 }, { "epoch": 1.94, "learning_rate": 1.8359208158368326e-06, "loss": 0.0006, "step": 16221 }, { "epoch": 1.94, "learning_rate": 1.8358488302339534e-06, "loss": 0.0074, "step": 16222 }, { "epoch": 1.94, "learning_rate": 1.8357768446310738e-06, "loss": 0.0003, "step": 16223 }, { "epoch": 1.94, "learning_rate": 1.8357048590281946e-06, "loss": 0.0002, "step": 16224 }, { "epoch": 1.94, "learning_rate": 1.835632873425315e-06, "loss": 0.0004, "step": 16225 }, { "epoch": 1.94, "learning_rate": 1.8355608878224357e-06, "loss": 0.0002, "step": 16226 }, { "epoch": 1.94, "learning_rate": 1.835488902219556e-06, "loss": 0.0002, "step": 16227 }, { "epoch": 1.94, "learning_rate": 1.8354169166166768e-06, "loss": 0.0004, "step": 16228 }, { "epoch": 1.94, "learning_rate": 1.8353449310137974e-06, "loss": 0.0006, "step": 16229 }, { "epoch": 1.94, "learning_rate": 1.8352729454109177e-06, "loss": 0.0002, "step": 16230 }, { "epoch": 1.94, "learning_rate": 1.8352009598080385e-06, "loss": 0.0004, "step": 16231 }, { "epoch": 1.95, "learning_rate": 1.8351289742051589e-06, "loss": 0.0004, "step": 16232 }, { "epoch": 1.95, "learning_rate": 1.8350569886022797e-06, "loss": 0.0002, "step": 16233 }, { "epoch": 1.95, "learning_rate": 1.8349850029994e-06, "loss": 0.0002, "step": 16234 }, { "epoch": 1.95, "learning_rate": 1.8349130173965208e-06, "loss": 0.0003, "step": 16235 }, { "epoch": 1.95, "learning_rate": 1.8348410317936414e-06, "loss": 0.0003, "step": 16236 }, { "epoch": 1.95, "learning_rate": 1.834769046190762e-06, "loss": 0.0003, "step": 16237 }, { "epoch": 1.95, "learning_rate": 1.8346970605878825e-06, "loss": 0.0005, "step": 16238 }, { "epoch": 1.95, "learning_rate": 1.8346250749850033e-06, "loss": 0.0002, "step": 16239 }, { "epoch": 1.95, "learning_rate": 1.8345530893821236e-06, "loss": 0.0051, "step": 16240 }, { "epoch": 1.95, "learning_rate": 1.834481103779244e-06, "loss": 0.6016, "step": 16241 }, { "epoch": 1.95, "learning_rate": 1.8344091181763648e-06, "loss": 0.0002, "step": 16242 }, { "epoch": 1.95, "learning_rate": 1.8343371325734853e-06, "loss": 0.0004, "step": 16243 }, { "epoch": 1.95, "learning_rate": 1.8342651469706059e-06, "loss": 0.0001, "step": 16244 }, { "epoch": 1.95, "learning_rate": 1.8341931613677265e-06, "loss": 0.697, "step": 16245 }, { "epoch": 1.95, "learning_rate": 1.8341211757648472e-06, "loss": 0.0004, "step": 16246 }, { "epoch": 1.95, "learning_rate": 1.8340491901619676e-06, "loss": 0.6885, "step": 16247 }, { "epoch": 1.95, "learning_rate": 1.8339772045590884e-06, "loss": 0.0004, "step": 16248 }, { "epoch": 1.95, "learning_rate": 1.8339052189562087e-06, "loss": 0.0003, "step": 16249 }, { "epoch": 1.95, "learning_rate": 1.8338332333533295e-06, "loss": 0.0003, "step": 16250 }, { "epoch": 1.95, "learning_rate": 1.8337612477504499e-06, "loss": 0.0003, "step": 16251 }, { "epoch": 1.95, "learning_rate": 1.8336892621475704e-06, "loss": 0.0002, "step": 16252 }, { "epoch": 1.95, "learning_rate": 1.8336172765446912e-06, "loss": 0.0003, "step": 16253 }, { "epoch": 1.95, "learning_rate": 1.8335452909418116e-06, "loss": 0.0005, "step": 16254 }, { "epoch": 1.95, "learning_rate": 1.8334733053389323e-06, "loss": 0.0005, "step": 16255 }, { "epoch": 1.95, "learning_rate": 1.8334013197360527e-06, "loss": 0.0028, "step": 16256 }, { "epoch": 1.95, "learning_rate": 1.8333293341331735e-06, "loss": 0.0001, "step": 16257 }, { "epoch": 1.95, "learning_rate": 1.8332573485302938e-06, "loss": 0.0008, "step": 16258 }, { "epoch": 1.95, "learning_rate": 1.8331853629274146e-06, "loss": 0.0002, "step": 16259 }, { "epoch": 1.95, "learning_rate": 1.8331133773245352e-06, "loss": 0.0002, "step": 16260 }, { "epoch": 1.95, "learning_rate": 1.833041391721656e-06, "loss": 0.0005, "step": 16261 }, { "epoch": 1.95, "learning_rate": 1.8329694061187763e-06, "loss": 0.0004, "step": 16262 }, { "epoch": 1.95, "learning_rate": 1.8328974205158967e-06, "loss": 0.0003, "step": 16263 }, { "epoch": 1.95, "learning_rate": 1.8328254349130174e-06, "loss": 0.0011, "step": 16264 }, { "epoch": 1.95, "learning_rate": 1.8327534493101378e-06, "loss": 0.0002, "step": 16265 }, { "epoch": 1.95, "learning_rate": 1.8326814637072586e-06, "loss": 0.0002, "step": 16266 }, { "epoch": 1.95, "learning_rate": 1.8326094781043791e-06, "loss": 0.0003, "step": 16267 }, { "epoch": 1.95, "learning_rate": 1.8325374925015e-06, "loss": 0.0003, "step": 16268 }, { "epoch": 1.95, "learning_rate": 1.8324655068986203e-06, "loss": 0.0004, "step": 16269 }, { "epoch": 1.95, "learning_rate": 1.832393521295741e-06, "loss": 0.0005, "step": 16270 }, { "epoch": 1.95, "learning_rate": 1.8323215356928614e-06, "loss": 0.0001, "step": 16271 }, { "epoch": 1.95, "learning_rate": 1.8322495500899822e-06, "loss": 0.0002, "step": 16272 }, { "epoch": 1.95, "learning_rate": 1.8321775644871025e-06, "loss": 0.0003, "step": 16273 }, { "epoch": 1.95, "learning_rate": 1.8321055788842233e-06, "loss": 0.0003, "step": 16274 }, { "epoch": 1.95, "learning_rate": 1.8320335932813439e-06, "loss": 0.0003, "step": 16275 }, { "epoch": 1.95, "learning_rate": 1.8319616076784642e-06, "loss": 0.0002, "step": 16276 }, { "epoch": 1.95, "learning_rate": 1.831889622075585e-06, "loss": 0.0004, "step": 16277 }, { "epoch": 1.95, "learning_rate": 1.8318176364727054e-06, "loss": 0.0003, "step": 16278 }, { "epoch": 1.95, "learning_rate": 1.8317456508698262e-06, "loss": 0.0003, "step": 16279 }, { "epoch": 1.95, "learning_rate": 1.8316736652669465e-06, "loss": 0.0002, "step": 16280 }, { "epoch": 1.95, "learning_rate": 1.8316016796640673e-06, "loss": 0.0003, "step": 16281 }, { "epoch": 1.95, "learning_rate": 1.8315296940611879e-06, "loss": 0.0003, "step": 16282 }, { "epoch": 1.95, "learning_rate": 1.8314577084583084e-06, "loss": 0.0002, "step": 16283 }, { "epoch": 1.95, "learning_rate": 1.831385722855429e-06, "loss": 0.0003, "step": 16284 }, { "epoch": 1.95, "learning_rate": 1.8313137372525498e-06, "loss": 0.0004, "step": 16285 }, { "epoch": 1.95, "learning_rate": 1.8312417516496701e-06, "loss": 0.0004, "step": 16286 }, { "epoch": 1.95, "learning_rate": 1.8311697660467905e-06, "loss": 0.0002, "step": 16287 }, { "epoch": 1.95, "learning_rate": 1.8310977804439113e-06, "loss": 1.6698, "step": 16288 }, { "epoch": 1.95, "learning_rate": 1.8310257948410318e-06, "loss": 0.0003, "step": 16289 }, { "epoch": 1.95, "learning_rate": 1.8309538092381524e-06, "loss": 0.0002, "step": 16290 }, { "epoch": 1.95, "learning_rate": 1.830881823635273e-06, "loss": 0.0621, "step": 16291 }, { "epoch": 1.95, "learning_rate": 1.8308098380323937e-06, "loss": 0.5557, "step": 16292 }, { "epoch": 1.95, "learning_rate": 1.830737852429514e-06, "loss": 0.0002, "step": 16293 }, { "epoch": 1.95, "learning_rate": 1.8306658668266349e-06, "loss": 0.0002, "step": 16294 }, { "epoch": 1.95, "learning_rate": 1.8305938812237552e-06, "loss": 0.0004, "step": 16295 }, { "epoch": 1.95, "learning_rate": 1.830521895620876e-06, "loss": 0.0003, "step": 16296 }, { "epoch": 1.95, "learning_rate": 1.8304499100179964e-06, "loss": 0.0004, "step": 16297 }, { "epoch": 1.95, "learning_rate": 1.830377924415117e-06, "loss": 0.0005, "step": 16298 }, { "epoch": 1.95, "learning_rate": 1.8303059388122377e-06, "loss": 0.0004, "step": 16299 }, { "epoch": 1.95, "learning_rate": 1.830233953209358e-06, "loss": 0.0002, "step": 16300 }, { "epoch": 1.95, "learning_rate": 1.8301619676064788e-06, "loss": 0.0008, "step": 16301 }, { "epoch": 1.95, "learning_rate": 1.8300899820035992e-06, "loss": 0.0001, "step": 16302 }, { "epoch": 1.95, "learning_rate": 1.83001799640072e-06, "loss": 0.0006, "step": 16303 }, { "epoch": 1.95, "learning_rate": 1.8299460107978403e-06, "loss": 0.0003, "step": 16304 }, { "epoch": 1.95, "learning_rate": 1.829874025194961e-06, "loss": 0.0003, "step": 16305 }, { "epoch": 1.95, "learning_rate": 1.8298020395920817e-06, "loss": 0.0002, "step": 16306 }, { "epoch": 1.95, "learning_rate": 1.8297300539892024e-06, "loss": 0.0001, "step": 16307 }, { "epoch": 1.95, "learning_rate": 1.8296580683863228e-06, "loss": 0.0003, "step": 16308 }, { "epoch": 1.95, "learning_rate": 1.8295860827834432e-06, "loss": 0.0001, "step": 16309 }, { "epoch": 1.95, "learning_rate": 1.829514097180564e-06, "loss": 0.0002, "step": 16310 }, { "epoch": 1.95, "learning_rate": 1.8294421115776843e-06, "loss": 0.0079, "step": 16311 }, { "epoch": 1.95, "learning_rate": 1.829370125974805e-06, "loss": 0.0026, "step": 16312 }, { "epoch": 1.95, "learning_rate": 1.8292981403719256e-06, "loss": 0.0011, "step": 16313 }, { "epoch": 1.95, "learning_rate": 1.8292261547690464e-06, "loss": 0.0004, "step": 16314 }, { "epoch": 1.96, "learning_rate": 1.8291541691661668e-06, "loss": 0.0002, "step": 16315 }, { "epoch": 1.96, "learning_rate": 1.8290821835632875e-06, "loss": 0.0002, "step": 16316 }, { "epoch": 1.96, "learning_rate": 1.829010197960408e-06, "loss": 0.0004, "step": 16317 }, { "epoch": 1.96, "learning_rate": 1.8289382123575287e-06, "loss": 0.0005, "step": 16318 }, { "epoch": 1.96, "learning_rate": 1.828866226754649e-06, "loss": 0.0001, "step": 16319 }, { "epoch": 1.96, "learning_rate": 1.8287942411517698e-06, "loss": 0.0005, "step": 16320 }, { "epoch": 1.96, "learning_rate": 1.8287222555488904e-06, "loss": 0.0002, "step": 16321 }, { "epoch": 1.96, "learning_rate": 1.8286502699460107e-06, "loss": 0.0003, "step": 16322 }, { "epoch": 1.96, "learning_rate": 1.8285782843431315e-06, "loss": 0.0002, "step": 16323 }, { "epoch": 1.96, "learning_rate": 1.8285062987402519e-06, "loss": 0.0327, "step": 16324 }, { "epoch": 1.96, "learning_rate": 1.8284343131373726e-06, "loss": 0.8956, "step": 16325 }, { "epoch": 1.96, "learning_rate": 1.828362327534493e-06, "loss": 0.0004, "step": 16326 }, { "epoch": 1.96, "learning_rate": 1.8282903419316138e-06, "loss": 0.0003, "step": 16327 }, { "epoch": 1.96, "learning_rate": 1.8282183563287341e-06, "loss": 0.0002, "step": 16328 }, { "epoch": 1.96, "learning_rate": 1.828146370725855e-06, "loss": 0.0003, "step": 16329 }, { "epoch": 1.96, "learning_rate": 1.8280743851229755e-06, "loss": 0.0004, "step": 16330 }, { "epoch": 1.96, "learning_rate": 1.8280023995200963e-06, "loss": 0.0002, "step": 16331 }, { "epoch": 1.96, "learning_rate": 1.8279304139172166e-06, "loss": 0.0002, "step": 16332 }, { "epoch": 1.96, "learning_rate": 1.827858428314337e-06, "loss": 0.0008, "step": 16333 }, { "epoch": 1.96, "learning_rate": 1.8277864427114577e-06, "loss": 0.0002, "step": 16334 }, { "epoch": 1.96, "learning_rate": 1.827714457108578e-06, "loss": 0.0003, "step": 16335 }, { "epoch": 1.96, "learning_rate": 1.8276424715056989e-06, "loss": 0.0002, "step": 16336 }, { "epoch": 1.96, "learning_rate": 1.8275704859028194e-06, "loss": 0.0002, "step": 16337 }, { "epoch": 1.96, "learning_rate": 1.8274985002999402e-06, "loss": 0.0003, "step": 16338 }, { "epoch": 1.96, "learning_rate": 1.8274265146970606e-06, "loss": 0.002, "step": 16339 }, { "epoch": 1.96, "learning_rate": 1.8273545290941814e-06, "loss": 0.0003, "step": 16340 }, { "epoch": 1.96, "learning_rate": 1.8272825434913017e-06, "loss": 0.0001, "step": 16341 }, { "epoch": 1.96, "learning_rate": 1.8272105578884225e-06, "loss": 0.0006, "step": 16342 }, { "epoch": 1.96, "learning_rate": 1.8271385722855428e-06, "loss": 0.0003, "step": 16343 }, { "epoch": 1.96, "learning_rate": 1.8270665866826634e-06, "loss": 0.0002, "step": 16344 }, { "epoch": 1.96, "learning_rate": 1.8269946010797842e-06, "loss": 0.0004, "step": 16345 }, { "epoch": 1.96, "learning_rate": 1.8269226154769045e-06, "loss": 0.0003, "step": 16346 }, { "epoch": 1.96, "learning_rate": 1.8268506298740253e-06, "loss": 0.0091, "step": 16347 }, { "epoch": 1.96, "learning_rate": 1.8267786442711457e-06, "loss": 0.0002, "step": 16348 }, { "epoch": 1.96, "learning_rate": 1.8267066586682665e-06, "loss": 0.0003, "step": 16349 }, { "epoch": 1.96, "learning_rate": 1.8266346730653868e-06, "loss": 0.0002, "step": 16350 }, { "epoch": 1.96, "learning_rate": 1.8265626874625076e-06, "loss": 0.0003, "step": 16351 }, { "epoch": 1.96, "learning_rate": 1.8264907018596282e-06, "loss": 0.7928, "step": 16352 }, { "epoch": 1.96, "learning_rate": 1.8264187162567487e-06, "loss": 0.0003, "step": 16353 }, { "epoch": 1.96, "learning_rate": 1.8263467306538693e-06, "loss": 0.0001, "step": 16354 }, { "epoch": 1.96, "learning_rate": 1.82627474505099e-06, "loss": 0.0005, "step": 16355 }, { "epoch": 1.96, "learning_rate": 1.8262027594481104e-06, "loss": 0.0004, "step": 16356 }, { "epoch": 1.96, "learning_rate": 1.8261307738452308e-06, "loss": 0.733, "step": 16357 }, { "epoch": 1.96, "learning_rate": 1.8260587882423516e-06, "loss": 0.0002, "step": 16358 }, { "epoch": 1.96, "learning_rate": 1.8259868026394721e-06, "loss": 0.0002, "step": 16359 }, { "epoch": 1.96, "learning_rate": 1.8259148170365927e-06, "loss": 0.0004, "step": 16360 }, { "epoch": 1.96, "learning_rate": 1.8258428314337133e-06, "loss": 0.0003, "step": 16361 }, { "epoch": 1.96, "learning_rate": 1.825770845830834e-06, "loss": 0.0002, "step": 16362 }, { "epoch": 1.96, "learning_rate": 1.8256988602279544e-06, "loss": 0.0009, "step": 16363 }, { "epoch": 1.96, "learning_rate": 1.8256268746250752e-06, "loss": 0.0001, "step": 16364 }, { "epoch": 1.96, "learning_rate": 1.8255548890221955e-06, "loss": 0.0002, "step": 16365 }, { "epoch": 1.96, "learning_rate": 1.8254829034193163e-06, "loss": 0.0002, "step": 16366 }, { "epoch": 1.96, "learning_rate": 1.8254109178164367e-06, "loss": 0.0002, "step": 16367 }, { "epoch": 1.96, "learning_rate": 1.8253389322135572e-06, "loss": 0.0003, "step": 16368 }, { "epoch": 1.96, "learning_rate": 1.825266946610678e-06, "loss": 0.0002, "step": 16369 }, { "epoch": 1.96, "learning_rate": 1.8251949610077984e-06, "loss": 0.0011, "step": 16370 }, { "epoch": 1.96, "learning_rate": 1.8251229754049191e-06, "loss": 0.0001, "step": 16371 }, { "epoch": 1.96, "learning_rate": 1.8250509898020395e-06, "loss": 0.0002, "step": 16372 }, { "epoch": 1.96, "learning_rate": 1.8249790041991603e-06, "loss": 0.0008, "step": 16373 }, { "epoch": 1.96, "learning_rate": 1.8249070185962806e-06, "loss": 0.0002, "step": 16374 }, { "epoch": 1.96, "learning_rate": 1.8248350329934014e-06, "loss": 0.0003, "step": 16375 }, { "epoch": 1.96, "learning_rate": 1.824763047390522e-06, "loss": 0.0003, "step": 16376 }, { "epoch": 1.96, "learning_rate": 1.8246910617876428e-06, "loss": 0.0003, "step": 16377 }, { "epoch": 1.96, "learning_rate": 1.8246190761847631e-06, "loss": 0.0003, "step": 16378 }, { "epoch": 1.96, "learning_rate": 1.8245470905818835e-06, "loss": 0.0003, "step": 16379 }, { "epoch": 1.96, "learning_rate": 1.8244751049790042e-06, "loss": 0.0004, "step": 16380 }, { "epoch": 1.96, "learning_rate": 1.8244031193761246e-06, "loss": 0.0002, "step": 16381 }, { "epoch": 1.96, "learning_rate": 1.8243311337732454e-06, "loss": 0.0002, "step": 16382 }, { "epoch": 1.96, "learning_rate": 1.824259148170366e-06, "loss": 0.0002, "step": 16383 }, { "epoch": 1.96, "learning_rate": 1.8241871625674867e-06, "loss": 0.7361, "step": 16384 }, { "epoch": 1.96, "learning_rate": 1.824115176964607e-06, "loss": 0.0002, "step": 16385 }, { "epoch": 1.96, "learning_rate": 1.8240431913617279e-06, "loss": 0.8792, "step": 16386 }, { "epoch": 1.96, "learning_rate": 1.8239712057588482e-06, "loss": 0.0004, "step": 16387 }, { "epoch": 1.96, "learning_rate": 1.823899220155969e-06, "loss": 0.0012, "step": 16388 }, { "epoch": 1.96, "learning_rate": 1.8238272345530893e-06, "loss": 0.0002, "step": 16389 }, { "epoch": 1.96, "learning_rate": 1.82375524895021e-06, "loss": 0.0002, "step": 16390 }, { "epoch": 1.96, "learning_rate": 1.8236832633473307e-06, "loss": 0.0004, "step": 16391 }, { "epoch": 1.96, "learning_rate": 1.823611277744451e-06, "loss": 0.225, "step": 16392 }, { "epoch": 1.96, "learning_rate": 1.8235392921415718e-06, "loss": 0.0002, "step": 16393 }, { "epoch": 1.96, "learning_rate": 1.8234673065386922e-06, "loss": 0.0002, "step": 16394 }, { "epoch": 1.96, "learning_rate": 1.823395320935813e-06, "loss": 0.0002, "step": 16395 }, { "epoch": 1.96, "learning_rate": 1.8233233353329333e-06, "loss": 0.0003, "step": 16396 }, { "epoch": 1.96, "learning_rate": 1.823251349730054e-06, "loss": 0.0002, "step": 16397 }, { "epoch": 1.97, "learning_rate": 1.8231793641271747e-06, "loss": 0.0003, "step": 16398 }, { "epoch": 1.97, "learning_rate": 1.8231073785242952e-06, "loss": 0.0002, "step": 16399 }, { "epoch": 1.97, "learning_rate": 1.8230353929214158e-06, "loss": 0.0003, "step": 16400 }, { "epoch": 1.97, "learning_rate": 1.8229634073185366e-06, "loss": 0.0003, "step": 16401 }, { "epoch": 1.97, "learning_rate": 1.822891421715657e-06, "loss": 0.0003, "step": 16402 }, { "epoch": 1.97, "learning_rate": 1.8228194361127773e-06, "loss": 0.0001, "step": 16403 }, { "epoch": 1.97, "learning_rate": 1.822747450509898e-06, "loss": 0.0002, "step": 16404 }, { "epoch": 1.97, "learning_rate": 1.8226754649070186e-06, "loss": 0.0002, "step": 16405 }, { "epoch": 1.97, "learning_rate": 1.8226034793041392e-06, "loss": 0.0002, "step": 16406 }, { "epoch": 1.97, "learning_rate": 1.8225314937012598e-06, "loss": 0.3368, "step": 16407 }, { "epoch": 1.97, "learning_rate": 1.8224595080983805e-06, "loss": 0.0005, "step": 16408 }, { "epoch": 1.97, "learning_rate": 1.8223875224955009e-06, "loss": 0.6995, "step": 16409 }, { "epoch": 1.97, "learning_rate": 1.8223155368926217e-06, "loss": 0.0003, "step": 16410 }, { "epoch": 1.97, "learning_rate": 1.822243551289742e-06, "loss": 0.001, "step": 16411 }, { "epoch": 1.97, "learning_rate": 1.8221715656868628e-06, "loss": 0.0006, "step": 16412 }, { "epoch": 1.97, "learning_rate": 1.8220995800839832e-06, "loss": 0.0002, "step": 16413 }, { "epoch": 1.97, "learning_rate": 1.8220275944811037e-06, "loss": 0.0002, "step": 16414 }, { "epoch": 1.97, "learning_rate": 1.8219556088782245e-06, "loss": 0.0009, "step": 16415 }, { "epoch": 1.97, "learning_rate": 1.8218836232753449e-06, "loss": 0.0002, "step": 16416 }, { "epoch": 1.97, "learning_rate": 1.8218116376724656e-06, "loss": 0.0007, "step": 16417 }, { "epoch": 1.97, "learning_rate": 1.821739652069586e-06, "loss": 0.0002, "step": 16418 }, { "epoch": 1.97, "learning_rate": 1.8216676664667068e-06, "loss": 0.0002, "step": 16419 }, { "epoch": 1.97, "learning_rate": 1.8215956808638271e-06, "loss": 0.0002, "step": 16420 }, { "epoch": 1.97, "learning_rate": 1.821523695260948e-06, "loss": 0.0006, "step": 16421 }, { "epoch": 1.97, "learning_rate": 1.8214517096580685e-06, "loss": 0.0002, "step": 16422 }, { "epoch": 1.97, "learning_rate": 1.8213797240551893e-06, "loss": 0.0003, "step": 16423 }, { "epoch": 1.97, "learning_rate": 1.8213077384523096e-06, "loss": 0.0002, "step": 16424 }, { "epoch": 1.97, "learning_rate": 1.82123575284943e-06, "loss": 0.0003, "step": 16425 }, { "epoch": 1.97, "learning_rate": 1.8211637672465507e-06, "loss": 0.0002, "step": 16426 }, { "epoch": 1.97, "learning_rate": 1.821091781643671e-06, "loss": 0.0002, "step": 16427 }, { "epoch": 1.97, "learning_rate": 1.8210197960407919e-06, "loss": 0.0002, "step": 16428 }, { "epoch": 1.97, "learning_rate": 1.8209478104379124e-06, "loss": 0.0003, "step": 16429 }, { "epoch": 1.97, "learning_rate": 1.8208758248350332e-06, "loss": 0.0007, "step": 16430 }, { "epoch": 1.97, "learning_rate": 1.8208038392321536e-06, "loss": 0.0007, "step": 16431 }, { "epoch": 1.97, "learning_rate": 1.8207318536292744e-06, "loss": 0.0003, "step": 16432 }, { "epoch": 1.97, "learning_rate": 1.8206598680263947e-06, "loss": 0.0005, "step": 16433 }, { "epoch": 1.97, "learning_rate": 1.8205878824235155e-06, "loss": 0.0003, "step": 16434 }, { "epoch": 1.97, "learning_rate": 1.8205158968206358e-06, "loss": 0.6998, "step": 16435 }, { "epoch": 1.97, "learning_rate": 1.8204439112177564e-06, "loss": 0.0003, "step": 16436 }, { "epoch": 1.97, "learning_rate": 1.8203719256148772e-06, "loss": 0.9329, "step": 16437 }, { "epoch": 1.97, "learning_rate": 1.8202999400119975e-06, "loss": 0.0002, "step": 16438 }, { "epoch": 1.97, "learning_rate": 1.8202279544091183e-06, "loss": 0.0003, "step": 16439 }, { "epoch": 1.97, "learning_rate": 1.8201559688062387e-06, "loss": 0.0003, "step": 16440 }, { "epoch": 1.97, "learning_rate": 1.8200839832033595e-06, "loss": 0.0024, "step": 16441 }, { "epoch": 1.97, "learning_rate": 1.8200119976004798e-06, "loss": 0.0002, "step": 16442 }, { "epoch": 1.97, "learning_rate": 1.8199400119976006e-06, "loss": 0.8458, "step": 16443 }, { "epoch": 1.97, "learning_rate": 1.8198680263947212e-06, "loss": 0.0003, "step": 16444 }, { "epoch": 1.97, "learning_rate": 1.8197960407918417e-06, "loss": 0.0002, "step": 16445 }, { "epoch": 1.97, "learning_rate": 1.8197240551889623e-06, "loss": 0.0002, "step": 16446 }, { "epoch": 1.97, "learning_rate": 1.819652069586083e-06, "loss": 0.0003, "step": 16447 }, { "epoch": 1.97, "learning_rate": 1.8195800839832034e-06, "loss": 0.0018, "step": 16448 }, { "epoch": 1.97, "learning_rate": 1.8195080983803238e-06, "loss": 0.0003, "step": 16449 }, { "epoch": 1.97, "learning_rate": 1.8194361127774446e-06, "loss": 0.0007, "step": 16450 }, { "epoch": 1.97, "learning_rate": 1.8193641271745651e-06, "loss": 0.0002, "step": 16451 }, { "epoch": 1.97, "learning_rate": 1.8192921415716857e-06, "loss": 0.0009, "step": 16452 }, { "epoch": 1.97, "learning_rate": 1.8192201559688063e-06, "loss": 0.0004, "step": 16453 }, { "epoch": 1.97, "learning_rate": 1.819148170365927e-06, "loss": 0.0002, "step": 16454 }, { "epoch": 1.97, "learning_rate": 1.8190761847630474e-06, "loss": 0.0003, "step": 16455 }, { "epoch": 1.97, "learning_rate": 1.8190041991601682e-06, "loss": 0.0003, "step": 16456 }, { "epoch": 1.97, "learning_rate": 1.8189322135572885e-06, "loss": 0.0004, "step": 16457 }, { "epoch": 1.97, "learning_rate": 1.8188602279544093e-06, "loss": 0.0003, "step": 16458 }, { "epoch": 1.97, "learning_rate": 1.8187882423515297e-06, "loss": 0.0003, "step": 16459 }, { "epoch": 1.97, "learning_rate": 1.8187162567486502e-06, "loss": 0.0193, "step": 16460 }, { "epoch": 1.97, "learning_rate": 1.818644271145771e-06, "loss": 0.0001, "step": 16461 }, { "epoch": 1.97, "learning_rate": 1.8185722855428914e-06, "loss": 0.0038, "step": 16462 }, { "epoch": 1.97, "learning_rate": 1.8185002999400121e-06, "loss": 0.0003, "step": 16463 }, { "epoch": 1.97, "learning_rate": 1.8184283143371325e-06, "loss": 0.0003, "step": 16464 }, { "epoch": 1.97, "learning_rate": 1.8183563287342533e-06, "loss": 0.0003, "step": 16465 }, { "epoch": 1.97, "learning_rate": 1.8182843431313736e-06, "loss": 0.0011, "step": 16466 }, { "epoch": 1.97, "learning_rate": 1.8182123575284944e-06, "loss": 0.0007, "step": 16467 }, { "epoch": 1.97, "learning_rate": 1.818140371925615e-06, "loss": 0.0004, "step": 16468 }, { "epoch": 1.97, "learning_rate": 1.8180683863227357e-06, "loss": 0.0009, "step": 16469 }, { "epoch": 1.97, "learning_rate": 1.817996400719856e-06, "loss": 0.0003, "step": 16470 }, { "epoch": 1.97, "learning_rate": 1.8179244151169765e-06, "loss": 0.0015, "step": 16471 }, { "epoch": 1.97, "learning_rate": 1.8178524295140972e-06, "loss": 0.0002, "step": 16472 }, { "epoch": 1.97, "learning_rate": 1.8177804439112176e-06, "loss": 0.0004, "step": 16473 }, { "epoch": 1.97, "learning_rate": 1.8177084583083384e-06, "loss": 0.0005, "step": 16474 }, { "epoch": 1.97, "learning_rate": 1.817636472705459e-06, "loss": 0.0002, "step": 16475 }, { "epoch": 1.97, "learning_rate": 1.8175644871025795e-06, "loss": 0.0002, "step": 16476 }, { "epoch": 1.97, "learning_rate": 1.8174925014997e-06, "loss": 0.0006, "step": 16477 }, { "epoch": 1.97, "learning_rate": 1.8174205158968208e-06, "loss": 0.0002, "step": 16478 }, { "epoch": 1.97, "learning_rate": 1.8173485302939412e-06, "loss": 0.0004, "step": 16479 }, { "epoch": 1.97, "learning_rate": 1.817276544691062e-06, "loss": 0.0005, "step": 16480 }, { "epoch": 1.97, "learning_rate": 1.8172045590881823e-06, "loss": 0.0002, "step": 16481 }, { "epoch": 1.98, "learning_rate": 1.8171325734853031e-06, "loss": 0.0003, "step": 16482 }, { "epoch": 1.98, "learning_rate": 1.8170605878824235e-06, "loss": 0.0093, "step": 16483 }, { "epoch": 1.98, "learning_rate": 1.816988602279544e-06, "loss": 0.0003, "step": 16484 }, { "epoch": 1.98, "learning_rate": 1.8169166166766648e-06, "loss": 0.0003, "step": 16485 }, { "epoch": 1.98, "learning_rate": 1.8168446310737852e-06, "loss": 0.0005, "step": 16486 }, { "epoch": 1.98, "learning_rate": 1.816772645470906e-06, "loss": 0.0004, "step": 16487 }, { "epoch": 1.98, "learning_rate": 1.8167006598680263e-06, "loss": 0.0012, "step": 16488 }, { "epoch": 1.98, "learning_rate": 1.816628674265147e-06, "loss": 0.0003, "step": 16489 }, { "epoch": 1.98, "learning_rate": 1.8165566886622674e-06, "loss": 0.0003, "step": 16490 }, { "epoch": 1.98, "learning_rate": 1.8164847030593882e-06, "loss": 0.0003, "step": 16491 }, { "epoch": 1.98, "learning_rate": 1.8164127174565088e-06, "loss": 0.0003, "step": 16492 }, { "epoch": 1.98, "learning_rate": 1.8163407318536296e-06, "loss": 0.0002, "step": 16493 }, { "epoch": 1.98, "learning_rate": 1.81626874625075e-06, "loss": 0.0366, "step": 16494 }, { "epoch": 1.98, "learning_rate": 1.8161967606478703e-06, "loss": 0.0006, "step": 16495 }, { "epoch": 1.98, "learning_rate": 1.816124775044991e-06, "loss": 0.0002, "step": 16496 }, { "epoch": 1.98, "learning_rate": 1.8160527894421114e-06, "loss": 0.001, "step": 16497 }, { "epoch": 1.98, "learning_rate": 1.8159808038392322e-06, "loss": 0.001, "step": 16498 }, { "epoch": 1.98, "learning_rate": 1.8159088182363528e-06, "loss": 0.0002, "step": 16499 }, { "epoch": 1.98, "learning_rate": 1.8158368326334735e-06, "loss": 0.0002, "step": 16500 }, { "epoch": 1.98, "learning_rate": 1.8157648470305939e-06, "loss": 0.0002, "step": 16501 }, { "epoch": 1.98, "learning_rate": 1.8156928614277147e-06, "loss": 0.0002, "step": 16502 }, { "epoch": 1.98, "learning_rate": 1.815620875824835e-06, "loss": 0.0004, "step": 16503 }, { "epoch": 1.98, "learning_rate": 1.8155488902219558e-06, "loss": 0.0004, "step": 16504 }, { "epoch": 1.98, "learning_rate": 1.8154769046190762e-06, "loss": 0.0002, "step": 16505 }, { "epoch": 1.98, "learning_rate": 1.8154049190161967e-06, "loss": 0.0008, "step": 16506 }, { "epoch": 1.98, "learning_rate": 1.8153329334133175e-06, "loss": 0.0004, "step": 16507 }, { "epoch": 1.98, "learning_rate": 1.8152609478104379e-06, "loss": 0.0005, "step": 16508 }, { "epoch": 1.98, "learning_rate": 1.8151889622075586e-06, "loss": 0.0003, "step": 16509 }, { "epoch": 1.98, "learning_rate": 1.815116976604679e-06, "loss": 0.0001, "step": 16510 }, { "epoch": 1.98, "learning_rate": 1.8150449910017998e-06, "loss": 1.0131, "step": 16511 }, { "epoch": 1.98, "learning_rate": 1.8149730053989201e-06, "loss": 0.0002, "step": 16512 }, { "epoch": 1.98, "learning_rate": 1.814901019796041e-06, "loss": 0.0004, "step": 16513 }, { "epoch": 1.98, "learning_rate": 1.8148290341931615e-06, "loss": 0.0003, "step": 16514 }, { "epoch": 1.98, "learning_rate": 1.814757048590282e-06, "loss": 0.0004, "step": 16515 }, { "epoch": 1.98, "learning_rate": 1.8146850629874026e-06, "loss": 0.0001, "step": 16516 }, { "epoch": 1.98, "learning_rate": 1.814613077384523e-06, "loss": 0.0001, "step": 16517 }, { "epoch": 1.98, "learning_rate": 1.8145410917816437e-06, "loss": 0.0007, "step": 16518 }, { "epoch": 1.98, "learning_rate": 1.814469106178764e-06, "loss": 0.0003, "step": 16519 }, { "epoch": 1.98, "learning_rate": 1.8143971205758849e-06, "loss": 0.0002, "step": 16520 }, { "epoch": 1.98, "learning_rate": 1.8143251349730054e-06, "loss": 0.0005, "step": 16521 }, { "epoch": 1.98, "learning_rate": 1.814253149370126e-06, "loss": 0.0002, "step": 16522 }, { "epoch": 1.98, "learning_rate": 1.8141811637672466e-06, "loss": 0.0002, "step": 16523 }, { "epoch": 1.98, "learning_rate": 1.8141091781643673e-06, "loss": 0.0003, "step": 16524 }, { "epoch": 1.98, "learning_rate": 1.8140371925614877e-06, "loss": 0.0004, "step": 16525 }, { "epoch": 1.98, "learning_rate": 1.8139652069586085e-06, "loss": 0.0003, "step": 16526 }, { "epoch": 1.98, "learning_rate": 1.8138932213557288e-06, "loss": 0.0011, "step": 16527 }, { "epoch": 1.98, "learning_rate": 1.8138212357528496e-06, "loss": 0.0002, "step": 16528 }, { "epoch": 1.98, "learning_rate": 1.81374925014997e-06, "loss": 0.0002, "step": 16529 }, { "epoch": 1.98, "learning_rate": 1.8136772645470905e-06, "loss": 0.0003, "step": 16530 }, { "epoch": 1.98, "learning_rate": 1.8136052789442113e-06, "loss": 0.0006, "step": 16531 }, { "epoch": 1.98, "learning_rate": 1.8135332933413317e-06, "loss": 0.0005, "step": 16532 }, { "epoch": 1.98, "learning_rate": 1.8134613077384524e-06, "loss": 0.0002, "step": 16533 }, { "epoch": 1.98, "learning_rate": 1.8133893221355728e-06, "loss": 0.0002, "step": 16534 }, { "epoch": 1.98, "learning_rate": 1.8133173365326936e-06, "loss": 0.0002, "step": 16535 }, { "epoch": 1.98, "learning_rate": 1.813245350929814e-06, "loss": 0.0007, "step": 16536 }, { "epoch": 1.98, "learning_rate": 1.8131733653269347e-06, "loss": 0.0003, "step": 16537 }, { "epoch": 1.98, "learning_rate": 1.8131013797240553e-06, "loss": 0.0003, "step": 16538 }, { "epoch": 1.98, "learning_rate": 1.813029394121176e-06, "loss": 0.0003, "step": 16539 }, { "epoch": 1.98, "learning_rate": 1.8129574085182964e-06, "loss": 0.0004, "step": 16540 }, { "epoch": 1.98, "learning_rate": 1.8128854229154168e-06, "loss": 0.0002, "step": 16541 }, { "epoch": 1.98, "learning_rate": 1.8128134373125375e-06, "loss": 0.0002, "step": 16542 }, { "epoch": 1.98, "learning_rate": 1.812741451709658e-06, "loss": 0.0003, "step": 16543 }, { "epoch": 1.98, "learning_rate": 1.8126694661067787e-06, "loss": 0.0002, "step": 16544 }, { "epoch": 1.98, "learning_rate": 1.8125974805038992e-06, "loss": 0.0001, "step": 16545 }, { "epoch": 1.98, "learning_rate": 1.81252549490102e-06, "loss": 0.0002, "step": 16546 }, { "epoch": 1.98, "learning_rate": 1.8124535092981404e-06, "loss": 0.0003, "step": 16547 }, { "epoch": 1.98, "learning_rate": 1.8123815236952612e-06, "loss": 0.0004, "step": 16548 }, { "epoch": 1.98, "learning_rate": 1.8123095380923815e-06, "loss": 0.0003, "step": 16549 }, { "epoch": 1.98, "learning_rate": 1.8122375524895023e-06, "loss": 1.009, "step": 16550 }, { "epoch": 1.98, "learning_rate": 1.8121655668866226e-06, "loss": 0.0002, "step": 16551 }, { "epoch": 1.98, "learning_rate": 1.8120935812837432e-06, "loss": 0.0002, "step": 16552 }, { "epoch": 1.98, "learning_rate": 1.812021595680864e-06, "loss": 0.0004, "step": 16553 }, { "epoch": 1.98, "learning_rate": 1.8119496100779843e-06, "loss": 0.0002, "step": 16554 }, { "epoch": 1.98, "learning_rate": 1.8118776244751051e-06, "loss": 0.0002, "step": 16555 }, { "epoch": 1.98, "learning_rate": 1.8118056388722255e-06, "loss": 0.0004, "step": 16556 }, { "epoch": 1.98, "learning_rate": 1.8117336532693463e-06, "loss": 0.0002, "step": 16557 }, { "epoch": 1.98, "learning_rate": 1.8116616676664666e-06, "loss": 0.0003, "step": 16558 }, { "epoch": 1.98, "learning_rate": 1.8115896820635874e-06, "loss": 0.0055, "step": 16559 }, { "epoch": 1.98, "learning_rate": 1.811517696460708e-06, "loss": 0.0002, "step": 16560 }, { "epoch": 1.98, "learning_rate": 1.8114457108578285e-06, "loss": 0.0002, "step": 16561 }, { "epoch": 1.98, "learning_rate": 1.811373725254949e-06, "loss": 0.0003, "step": 16562 }, { "epoch": 1.98, "learning_rate": 1.8113017396520694e-06, "loss": 0.0004, "step": 16563 }, { "epoch": 1.98, "learning_rate": 1.8112297540491902e-06, "loss": 0.001, "step": 16564 }, { "epoch": 1.99, "learning_rate": 1.8111577684463106e-06, "loss": 0.0004, "step": 16565 }, { "epoch": 1.99, "learning_rate": 1.8110857828434314e-06, "loss": 0.0004, "step": 16566 }, { "epoch": 1.99, "learning_rate": 1.811013797240552e-06, "loss": 0.0002, "step": 16567 }, { "epoch": 1.99, "learning_rate": 1.8109418116376725e-06, "loss": 0.0004, "step": 16568 }, { "epoch": 1.99, "learning_rate": 1.810869826034793e-06, "loss": 0.0004, "step": 16569 }, { "epoch": 1.99, "learning_rate": 1.8107978404319138e-06, "loss": 0.0002, "step": 16570 }, { "epoch": 1.99, "learning_rate": 1.8107258548290342e-06, "loss": 0.0006, "step": 16571 }, { "epoch": 1.99, "learning_rate": 1.810653869226155e-06, "loss": 0.0004, "step": 16572 }, { "epoch": 1.99, "learning_rate": 1.8105818836232753e-06, "loss": 0.0003, "step": 16573 }, { "epoch": 1.99, "learning_rate": 1.8105098980203961e-06, "loss": 0.0003, "step": 16574 }, { "epoch": 1.99, "learning_rate": 1.8104379124175165e-06, "loss": 0.001, "step": 16575 }, { "epoch": 1.99, "learning_rate": 1.810365926814637e-06, "loss": 0.004, "step": 16576 }, { "epoch": 1.99, "learning_rate": 1.8102939412117578e-06, "loss": 0.0005, "step": 16577 }, { "epoch": 1.99, "learning_rate": 1.8102219556088782e-06, "loss": 0.0008, "step": 16578 }, { "epoch": 1.99, "learning_rate": 1.810149970005999e-06, "loss": 0.0102, "step": 16579 }, { "epoch": 1.99, "learning_rate": 1.8100779844031193e-06, "loss": 0.0002, "step": 16580 }, { "epoch": 1.99, "learning_rate": 1.81000599880024e-06, "loss": 0.0003, "step": 16581 }, { "epoch": 1.99, "learning_rate": 1.8099340131973604e-06, "loss": 0.0004, "step": 16582 }, { "epoch": 1.99, "learning_rate": 1.8098620275944812e-06, "loss": 0.0009, "step": 16583 }, { "epoch": 1.99, "learning_rate": 1.8097900419916018e-06, "loss": 0.0002, "step": 16584 }, { "epoch": 1.99, "learning_rate": 1.8097180563887226e-06, "loss": 0.0002, "step": 16585 }, { "epoch": 1.99, "learning_rate": 1.809646070785843e-06, "loss": 0.0003, "step": 16586 }, { "epoch": 1.99, "learning_rate": 1.8095740851829633e-06, "loss": 0.0013, "step": 16587 }, { "epoch": 1.99, "learning_rate": 1.809502099580084e-06, "loss": 0.0004, "step": 16588 }, { "epoch": 1.99, "learning_rate": 1.8094301139772044e-06, "loss": 0.0003, "step": 16589 }, { "epoch": 1.99, "learning_rate": 1.8093581283743252e-06, "loss": 0.0002, "step": 16590 }, { "epoch": 1.99, "learning_rate": 1.8092861427714457e-06, "loss": 0.0003, "step": 16591 }, { "epoch": 1.99, "learning_rate": 1.8092141571685665e-06, "loss": 0.0002, "step": 16592 }, { "epoch": 1.99, "learning_rate": 1.8091421715656869e-06, "loss": 0.0003, "step": 16593 }, { "epoch": 1.99, "learning_rate": 1.8090701859628077e-06, "loss": 0.0003, "step": 16594 }, { "epoch": 1.99, "learning_rate": 1.808998200359928e-06, "loss": 0.0002, "step": 16595 }, { "epoch": 1.99, "learning_rate": 1.8089262147570488e-06, "loss": 0.0013, "step": 16596 }, { "epoch": 1.99, "learning_rate": 1.8088542291541691e-06, "loss": 0.0002, "step": 16597 }, { "epoch": 1.99, "learning_rate": 1.8087822435512897e-06, "loss": 0.0002, "step": 16598 }, { "epoch": 1.99, "learning_rate": 1.8087102579484105e-06, "loss": 0.0004, "step": 16599 }, { "epoch": 1.99, "learning_rate": 1.8086382723455308e-06, "loss": 0.0003, "step": 16600 }, { "epoch": 1.99, "learning_rate": 1.8085662867426516e-06, "loss": 0.0002, "step": 16601 }, { "epoch": 1.99, "learning_rate": 1.808494301139772e-06, "loss": 0.0003, "step": 16602 }, { "epoch": 1.99, "learning_rate": 1.8084223155368928e-06, "loss": 0.0004, "step": 16603 }, { "epoch": 1.99, "learning_rate": 1.8083503299340131e-06, "loss": 0.0008, "step": 16604 }, { "epoch": 1.99, "learning_rate": 1.8082783443311339e-06, "loss": 0.0003, "step": 16605 }, { "epoch": 1.99, "learning_rate": 1.8082063587282542e-06, "loss": 0.0001, "step": 16606 }, { "epoch": 1.99, "learning_rate": 1.808134373125375e-06, "loss": 0.0001, "step": 16607 }, { "epoch": 1.99, "learning_rate": 1.8080623875224956e-06, "loss": 0.0001, "step": 16608 }, { "epoch": 1.99, "learning_rate": 1.8079904019196164e-06, "loss": 0.0002, "step": 16609 }, { "epoch": 1.99, "learning_rate": 1.8079184163167367e-06, "loss": 0.0009, "step": 16610 }, { "epoch": 1.99, "learning_rate": 1.807846430713857e-06, "loss": 0.0003, "step": 16611 }, { "epoch": 1.99, "learning_rate": 1.8077744451109779e-06, "loss": 0.0007, "step": 16612 }, { "epoch": 1.99, "learning_rate": 1.8077024595080982e-06, "loss": 0.0002, "step": 16613 }, { "epoch": 1.99, "learning_rate": 1.807630473905219e-06, "loss": 0.0002, "step": 16614 }, { "epoch": 1.99, "learning_rate": 1.8075584883023396e-06, "loss": 0.0006, "step": 16615 }, { "epoch": 1.99, "learning_rate": 1.8074865026994603e-06, "loss": 0.0002, "step": 16616 }, { "epoch": 1.99, "learning_rate": 1.8074145170965807e-06, "loss": 0.0005, "step": 16617 }, { "epoch": 1.99, "learning_rate": 1.8073425314937015e-06, "loss": 0.0001, "step": 16618 }, { "epoch": 1.99, "learning_rate": 1.8072705458908218e-06, "loss": 0.0002, "step": 16619 }, { "epoch": 1.99, "learning_rate": 1.8071985602879426e-06, "loss": 0.0002, "step": 16620 }, { "epoch": 1.99, "learning_rate": 1.807126574685063e-06, "loss": 0.0003, "step": 16621 }, { "epoch": 1.99, "learning_rate": 1.8070545890821835e-06, "loss": 0.0002, "step": 16622 }, { "epoch": 1.99, "learning_rate": 1.8069826034793043e-06, "loss": 0.0003, "step": 16623 }, { "epoch": 1.99, "learning_rate": 1.8069106178764247e-06, "loss": 0.0001, "step": 16624 }, { "epoch": 1.99, "learning_rate": 1.8068386322735454e-06, "loss": 0.0976, "step": 16625 }, { "epoch": 1.99, "learning_rate": 1.8067666466706658e-06, "loss": 0.0002, "step": 16626 }, { "epoch": 1.99, "learning_rate": 1.8066946610677866e-06, "loss": 0.0002, "step": 16627 }, { "epoch": 1.99, "learning_rate": 1.806622675464907e-06, "loss": 0.0006, "step": 16628 }, { "epoch": 1.99, "learning_rate": 1.8065506898620277e-06, "loss": 0.0002, "step": 16629 }, { "epoch": 1.99, "learning_rate": 1.8064787042591483e-06, "loss": 0.0007, "step": 16630 }, { "epoch": 1.99, "learning_rate": 1.8064067186562688e-06, "loss": 0.0003, "step": 16631 }, { "epoch": 1.99, "learning_rate": 1.8063347330533894e-06, "loss": 0.0005, "step": 16632 }, { "epoch": 1.99, "learning_rate": 1.8062627474505098e-06, "loss": 0.0001, "step": 16633 }, { "epoch": 1.99, "learning_rate": 1.8061907618476305e-06, "loss": 0.0001, "step": 16634 }, { "epoch": 1.99, "learning_rate": 1.806118776244751e-06, "loss": 0.0003, "step": 16635 }, { "epoch": 1.99, "learning_rate": 1.8060467906418717e-06, "loss": 0.0002, "step": 16636 }, { "epoch": 1.99, "learning_rate": 1.8059748050389922e-06, "loss": 0.0002, "step": 16637 }, { "epoch": 1.99, "learning_rate": 1.8059028194361128e-06, "loss": 0.0003, "step": 16638 }, { "epoch": 1.99, "learning_rate": 1.8058308338332334e-06, "loss": 0.0002, "step": 16639 }, { "epoch": 1.99, "learning_rate": 1.8057588482303542e-06, "loss": 0.0011, "step": 16640 }, { "epoch": 1.99, "learning_rate": 1.8056868626274745e-06, "loss": 0.0002, "step": 16641 }, { "epoch": 1.99, "learning_rate": 1.8056148770245953e-06, "loss": 0.0002, "step": 16642 }, { "epoch": 1.99, "learning_rate": 1.8055428914217156e-06, "loss": 0.0002, "step": 16643 }, { "epoch": 1.99, "learning_rate": 1.8054709058188362e-06, "loss": 0.0004, "step": 16644 }, { "epoch": 1.99, "learning_rate": 1.8053989202159568e-06, "loss": 0.0003, "step": 16645 }, { "epoch": 1.99, "learning_rate": 1.8053269346130773e-06, "loss": 0.0001, "step": 16646 }, { "epoch": 1.99, "learning_rate": 1.8052549490101981e-06, "loss": 0.0001, "step": 16647 }, { "epoch": 1.99, "learning_rate": 1.8051829634073185e-06, "loss": 0.0003, "step": 16648 }, { "epoch": 2.0, "learning_rate": 1.8051109778044393e-06, "loss": 0.0003, "step": 16649 }, { "epoch": 2.0, "learning_rate": 1.8050389922015596e-06, "loss": 0.0001, "step": 16650 }, { "epoch": 2.0, "learning_rate": 1.8049670065986804e-06, "loss": 0.0002, "step": 16651 }, { "epoch": 2.0, "learning_rate": 1.8048950209958007e-06, "loss": 0.0004, "step": 16652 }, { "epoch": 2.0, "learning_rate": 1.8048230353929215e-06, "loss": 0.0002, "step": 16653 }, { "epoch": 2.0, "learning_rate": 1.804751049790042e-06, "loss": 0.0005, "step": 16654 }, { "epoch": 2.0, "learning_rate": 1.8046790641871629e-06, "loss": 1.1278, "step": 16655 }, { "epoch": 2.0, "learning_rate": 1.8046070785842832e-06, "loss": 0.0002, "step": 16656 }, { "epoch": 2.0, "learning_rate": 1.8045350929814036e-06, "loss": 0.0002, "step": 16657 }, { "epoch": 2.0, "learning_rate": 1.8044631073785244e-06, "loss": 0.0002, "step": 16658 }, { "epoch": 2.0, "learning_rate": 1.8043911217756447e-06, "loss": 0.0002, "step": 16659 }, { "epoch": 2.0, "learning_rate": 1.8043191361727655e-06, "loss": 0.0008, "step": 16660 }, { "epoch": 2.0, "learning_rate": 1.804247150569886e-06, "loss": 0.8129, "step": 16661 }, { "epoch": 2.0, "learning_rate": 1.8041751649670068e-06, "loss": 0.0002, "step": 16662 }, { "epoch": 2.0, "learning_rate": 1.8041031793641272e-06, "loss": 0.0002, "step": 16663 }, { "epoch": 2.0, "learning_rate": 1.804031193761248e-06, "loss": 0.0002, "step": 16664 }, { "epoch": 2.0, "learning_rate": 1.8039592081583683e-06, "loss": 0.0002, "step": 16665 }, { "epoch": 2.0, "learning_rate": 1.803887222555489e-06, "loss": 0.0004, "step": 16666 }, { "epoch": 2.0, "learning_rate": 1.8038152369526095e-06, "loss": 0.0002, "step": 16667 }, { "epoch": 2.0, "learning_rate": 1.80374325134973e-06, "loss": 0.0002, "step": 16668 }, { "epoch": 2.0, "learning_rate": 1.8036712657468508e-06, "loss": 0.0007, "step": 16669 }, { "epoch": 2.0, "learning_rate": 1.8035992801439712e-06, "loss": 0.0002, "step": 16670 }, { "epoch": 2.0, "learning_rate": 1.803527294541092e-06, "loss": 0.0003, "step": 16671 }, { "epoch": 2.0, "learning_rate": 1.8034553089382123e-06, "loss": 0.0006, "step": 16672 }, { "epoch": 2.0, "learning_rate": 1.803383323335333e-06, "loss": 0.0005, "step": 16673 }, { "epoch": 2.0, "learning_rate": 1.8033113377324534e-06, "loss": 0.0002, "step": 16674 }, { "epoch": 2.0, "learning_rate": 1.8032393521295742e-06, "loss": 0.0003, "step": 16675 }, { "epoch": 2.0, "learning_rate": 1.8031673665266948e-06, "loss": 0.0002, "step": 16676 }, { "epoch": 2.0, "learning_rate": 1.8030953809238153e-06, "loss": 0.0002, "step": 16677 }, { "epoch": 2.0, "learning_rate": 1.803023395320936e-06, "loss": 0.0004, "step": 16678 }, { "epoch": 2.0, "learning_rate": 1.8029514097180563e-06, "loss": 0.8776, "step": 16679 }, { "epoch": 2.0, "learning_rate": 1.802879424115177e-06, "loss": 0.0003, "step": 16680 }, { "epoch": 2.0, "learning_rate": 1.8028074385122974e-06, "loss": 0.4058, "step": 16681 }, { "epoch": 2.0, "learning_rate": 1.8027354529094182e-06, "loss": 0.0002, "step": 16682 }, { "epoch": 2.0, "learning_rate": 1.8026634673065387e-06, "loss": 0.0003, "step": 16683 }, { "epoch": 2.0, "learning_rate": 1.8025914817036593e-06, "loss": 0.0001, "step": 16684 }, { "epoch": 2.0, "learning_rate": 1.8025194961007799e-06, "loss": 0.0002, "step": 16685 }, { "epoch": 2.0, "learning_rate": 1.8024475104979006e-06, "loss": 0.0004, "step": 16686 }, { "epoch": 2.0, "learning_rate": 1.802375524895021e-06, "loss": 0.0003, "step": 16687 }, { "epoch": 2.0, "learning_rate": 1.8023035392921418e-06, "loss": 0.0003, "step": 16688 }, { "epoch": 2.0, "learning_rate": 1.8022315536892621e-06, "loss": 0.0002, "step": 16689 }, { "epoch": 2.0, "learning_rate": 1.8021595680863827e-06, "loss": 0.0003, "step": 16690 }, { "epoch": 2.0, "eval_accuracy": 0.9907729179149191, "eval_loss": 0.05078105628490448, "eval_runtime": 158.278, "eval_samples_per_second": 105.447, "eval_steps_per_second": 1.649, "step": 16690 }, { "epoch": 2.0, "learning_rate": 1.8020875824835033e-06, "loss": 0.0002, "step": 16691 }, { "epoch": 2.0, "learning_rate": 1.8020155968806238e-06, "loss": 0.0002, "step": 16692 }, { "epoch": 2.0, "learning_rate": 1.8019436112777446e-06, "loss": 0.0002, "step": 16693 }, { "epoch": 2.0, "learning_rate": 1.801871625674865e-06, "loss": 0.0002, "step": 16694 }, { "epoch": 2.0, "learning_rate": 1.8017996400719857e-06, "loss": 0.0002, "step": 16695 }, { "epoch": 2.0, "learning_rate": 1.801727654469106e-06, "loss": 0.0002, "step": 16696 }, { "epoch": 2.0, "learning_rate": 1.8016556688662269e-06, "loss": 0.0008, "step": 16697 }, { "epoch": 2.0, "learning_rate": 1.8015836832633472e-06, "loss": 0.0002, "step": 16698 }, { "epoch": 2.0, "learning_rate": 1.801511697660468e-06, "loss": 0.0002, "step": 16699 }, { "epoch": 2.0, "learning_rate": 1.8014397120575886e-06, "loss": 0.0002, "step": 16700 }, { "epoch": 2.0, "learning_rate": 1.8013677264547094e-06, "loss": 0.0003, "step": 16701 }, { "epoch": 2.0, "learning_rate": 1.8012957408518297e-06, "loss": 0.0003, "step": 16702 }, { "epoch": 2.0, "learning_rate": 1.80122375524895e-06, "loss": 0.0002, "step": 16703 }, { "epoch": 2.0, "learning_rate": 1.8011517696460708e-06, "loss": 0.0002, "step": 16704 }, { "epoch": 2.0, "learning_rate": 1.8010797840431912e-06, "loss": 0.0002, "step": 16705 }, { "epoch": 2.0, "learning_rate": 1.801007798440312e-06, "loss": 0.0002, "step": 16706 }, { "epoch": 2.0, "learning_rate": 1.8009358128374326e-06, "loss": 0.0004, "step": 16707 }, { "epoch": 2.0, "learning_rate": 1.8008638272345533e-06, "loss": 0.5883, "step": 16708 }, { "epoch": 2.0, "learning_rate": 1.8007918416316737e-06, "loss": 0.0002, "step": 16709 }, { "epoch": 2.0, "learning_rate": 1.8007198560287945e-06, "loss": 0.0003, "step": 16710 }, { "epoch": 2.0, "learning_rate": 1.8006478704259148e-06, "loss": 0.0002, "step": 16711 }, { "epoch": 2.0, "learning_rate": 1.8005758848230356e-06, "loss": 0.0003, "step": 16712 }, { "epoch": 2.0, "learning_rate": 1.800503899220156e-06, "loss": 0.0002, "step": 16713 }, { "epoch": 2.0, "learning_rate": 1.8004319136172765e-06, "loss": 0.0001, "step": 16714 }, { "epoch": 2.0, "learning_rate": 1.8003599280143973e-06, "loss": 1.1732, "step": 16715 }, { "epoch": 2.0, "learning_rate": 1.8002879424115177e-06, "loss": 0.0002, "step": 16716 }, { "epoch": 2.0, "learning_rate": 1.8002159568086384e-06, "loss": 0.0002, "step": 16717 }, { "epoch": 2.0, "learning_rate": 1.8001439712057588e-06, "loss": 0.0003, "step": 16718 }, { "epoch": 2.0, "learning_rate": 1.8000719856028796e-06, "loss": 0.0004, "step": 16719 }, { "epoch": 2.0, "learning_rate": 1.8e-06, "loss": 0.0003, "step": 16720 }, { "epoch": 2.0, "learning_rate": 1.7999280143971207e-06, "loss": 0.0003, "step": 16721 }, { "epoch": 2.0, "learning_rate": 1.7998560287942413e-06, "loss": 0.0003, "step": 16722 }, { "epoch": 2.0, "learning_rate": 1.7997840431913618e-06, "loss": 0.0002, "step": 16723 }, { "epoch": 2.0, "learning_rate": 1.7997120575884824e-06, "loss": 0.8252, "step": 16724 }, { "epoch": 2.0, "learning_rate": 1.7996400719856028e-06, "loss": 0.0005, "step": 16725 }, { "epoch": 2.0, "learning_rate": 1.7995680863827235e-06, "loss": 0.0005, "step": 16726 }, { "epoch": 2.0, "learning_rate": 1.7994961007798439e-06, "loss": 0.0002, "step": 16727 }, { "epoch": 2.0, "learning_rate": 1.7994241151769647e-06, "loss": 0.0002, "step": 16728 }, { "epoch": 2.0, "learning_rate": 1.799352129574085e-06, "loss": 0.0003, "step": 16729 }, { "epoch": 2.0, "learning_rate": 1.7992801439712058e-06, "loss": 0.0003, "step": 16730 }, { "epoch": 2.0, "learning_rate": 1.7992081583683264e-06, "loss": 0.0013, "step": 16731 }, { "epoch": 2.01, "learning_rate": 1.7991361727654471e-06, "loss": 0.0002, "step": 16732 }, { "epoch": 2.01, "learning_rate": 1.7990641871625675e-06, "loss": 0.0003, "step": 16733 }, { "epoch": 2.01, "learning_rate": 1.7989922015596883e-06, "loss": 0.0004, "step": 16734 }, { "epoch": 2.01, "learning_rate": 1.7989202159568086e-06, "loss": 0.0015, "step": 16735 }, { "epoch": 2.01, "learning_rate": 1.7988482303539294e-06, "loss": 0.0003, "step": 16736 }, { "epoch": 2.01, "learning_rate": 1.7987762447510498e-06, "loss": 0.0001, "step": 16737 }, { "epoch": 2.01, "learning_rate": 1.7987042591481703e-06, "loss": 0.0003, "step": 16738 }, { "epoch": 2.01, "learning_rate": 1.7986322735452911e-06, "loss": 0.0003, "step": 16739 }, { "epoch": 2.01, "learning_rate": 1.7985602879424115e-06, "loss": 0.0004, "step": 16740 }, { "epoch": 2.01, "learning_rate": 1.7984883023395322e-06, "loss": 0.0006, "step": 16741 }, { "epoch": 2.01, "learning_rate": 1.7984163167366526e-06, "loss": 0.0007, "step": 16742 }, { "epoch": 2.01, "learning_rate": 1.7983443311337734e-06, "loss": 0.0001, "step": 16743 }, { "epoch": 2.01, "learning_rate": 1.7982723455308937e-06, "loss": 0.0003, "step": 16744 }, { "epoch": 2.01, "learning_rate": 1.7982003599280145e-06, "loss": 0.0003, "step": 16745 }, { "epoch": 2.01, "learning_rate": 1.798128374325135e-06, "loss": 0.0005, "step": 16746 }, { "epoch": 2.01, "learning_rate": 1.7980563887222556e-06, "loss": 0.0002, "step": 16747 }, { "epoch": 2.01, "learning_rate": 1.7979844031193762e-06, "loss": 0.0002, "step": 16748 }, { "epoch": 2.01, "learning_rate": 1.7979124175164966e-06, "loss": 0.0003, "step": 16749 }, { "epoch": 2.01, "learning_rate": 1.7978404319136173e-06, "loss": 0.0001, "step": 16750 }, { "epoch": 2.01, "learning_rate": 1.7977684463107377e-06, "loss": 0.0002, "step": 16751 }, { "epoch": 2.01, "learning_rate": 1.7976964607078585e-06, "loss": 0.8137, "step": 16752 }, { "epoch": 2.01, "learning_rate": 1.797624475104979e-06, "loss": 0.0003, "step": 16753 }, { "epoch": 2.01, "learning_rate": 1.7975524895020996e-06, "loss": 0.459, "step": 16754 }, { "epoch": 2.01, "learning_rate": 1.7974805038992202e-06, "loss": 0.0002, "step": 16755 }, { "epoch": 2.01, "learning_rate": 1.797408518296341e-06, "loss": 0.0002, "step": 16756 }, { "epoch": 2.01, "learning_rate": 1.7973365326934613e-06, "loss": 0.0002, "step": 16757 }, { "epoch": 2.01, "learning_rate": 1.797264547090582e-06, "loss": 0.0001, "step": 16758 }, { "epoch": 2.01, "learning_rate": 1.7971925614877024e-06, "loss": 0.0002, "step": 16759 }, { "epoch": 2.01, "learning_rate": 1.797120575884823e-06, "loss": 0.0002, "step": 16760 }, { "epoch": 2.01, "learning_rate": 1.7970485902819436e-06, "loss": 0.0004, "step": 16761 }, { "epoch": 2.01, "learning_rate": 1.7969766046790641e-06, "loss": 0.0001, "step": 16762 }, { "epoch": 2.01, "learning_rate": 1.796904619076185e-06, "loss": 0.0051, "step": 16763 }, { "epoch": 2.01, "learning_rate": 1.7968326334733053e-06, "loss": 0.0002, "step": 16764 }, { "epoch": 2.01, "learning_rate": 1.796760647870426e-06, "loss": 0.0001, "step": 16765 }, { "epoch": 2.01, "learning_rate": 1.7966886622675464e-06, "loss": 0.0003, "step": 16766 }, { "epoch": 2.01, "learning_rate": 1.7966166766646672e-06, "loss": 0.0019, "step": 16767 }, { "epoch": 2.01, "learning_rate": 1.7965446910617875e-06, "loss": 0.0008, "step": 16768 }, { "epoch": 2.01, "learning_rate": 1.7964727054589083e-06, "loss": 0.0004, "step": 16769 }, { "epoch": 2.01, "learning_rate": 1.7964007198560289e-06, "loss": 0.0003, "step": 16770 }, { "epoch": 2.01, "learning_rate": 1.7963287342531492e-06, "loss": 0.6834, "step": 16771 }, { "epoch": 2.01, "learning_rate": 1.79625674865027e-06, "loss": 0.0003, "step": 16772 }, { "epoch": 2.01, "learning_rate": 1.7961847630473904e-06, "loss": 0.0003, "step": 16773 }, { "epoch": 2.01, "learning_rate": 1.7961127774445112e-06, "loss": 0.0004, "step": 16774 }, { "epoch": 2.01, "learning_rate": 1.7960407918416315e-06, "loss": 0.0002, "step": 16775 }, { "epoch": 2.01, "learning_rate": 1.7959688062387523e-06, "loss": 0.0004, "step": 16776 }, { "epoch": 2.01, "learning_rate": 1.7958968206358729e-06, "loss": 0.0003, "step": 16777 }, { "epoch": 2.01, "learning_rate": 1.7958248350329936e-06, "loss": 0.0003, "step": 16778 }, { "epoch": 2.01, "learning_rate": 1.795752849430114e-06, "loss": 0.0004, "step": 16779 }, { "epoch": 2.01, "learning_rate": 1.7956808638272348e-06, "loss": 0.0001, "step": 16780 }, { "epoch": 2.01, "learning_rate": 1.7956088782243551e-06, "loss": 0.0002, "step": 16781 }, { "epoch": 2.01, "learning_rate": 1.795536892621476e-06, "loss": 0.0003, "step": 16782 }, { "epoch": 2.01, "learning_rate": 1.7954649070185963e-06, "loss": 0.0002, "step": 16783 }, { "epoch": 2.01, "learning_rate": 1.7953929214157168e-06, "loss": 0.0001, "step": 16784 }, { "epoch": 2.01, "learning_rate": 1.7953209358128376e-06, "loss": 0.006, "step": 16785 }, { "epoch": 2.01, "learning_rate": 1.795248950209958e-06, "loss": 0.0003, "step": 16786 }, { "epoch": 2.01, "learning_rate": 1.7951769646070787e-06, "loss": 0.0003, "step": 16787 }, { "epoch": 2.01, "learning_rate": 1.795104979004199e-06, "loss": 0.0002, "step": 16788 }, { "epoch": 2.01, "learning_rate": 1.7950329934013199e-06, "loss": 0.0002, "step": 16789 }, { "epoch": 2.01, "learning_rate": 1.7949610077984402e-06, "loss": 0.0004, "step": 16790 }, { "epoch": 2.01, "learning_rate": 1.794889022195561e-06, "loss": 0.0006, "step": 16791 }, { "epoch": 2.01, "learning_rate": 1.7948170365926816e-06, "loss": 0.0003, "step": 16792 }, { "epoch": 2.01, "learning_rate": 1.7947450509898021e-06, "loss": 0.0003, "step": 16793 }, { "epoch": 2.01, "learning_rate": 1.7946730653869227e-06, "loss": 0.0002, "step": 16794 }, { "epoch": 2.01, "learning_rate": 1.794601079784043e-06, "loss": 0.0007, "step": 16795 }, { "epoch": 2.01, "learning_rate": 1.7945290941811638e-06, "loss": 0.0003, "step": 16796 }, { "epoch": 2.01, "learning_rate": 1.7944571085782842e-06, "loss": 0.0002, "step": 16797 }, { "epoch": 2.01, "learning_rate": 1.794385122975405e-06, "loss": 0.0003, "step": 16798 }, { "epoch": 2.01, "learning_rate": 1.7943131373725255e-06, "loss": 0.0002, "step": 16799 }, { "epoch": 2.01, "learning_rate": 1.7942411517696461e-06, "loss": 0.0002, "step": 16800 }, { "epoch": 2.01, "learning_rate": 1.7941691661667667e-06, "loss": 0.0014, "step": 16801 }, { "epoch": 2.01, "learning_rate": 1.7940971805638875e-06, "loss": 0.0003, "step": 16802 }, { "epoch": 2.01, "learning_rate": 1.7940251949610078e-06, "loss": 0.0009, "step": 16803 }, { "epoch": 2.01, "learning_rate": 1.7939532093581286e-06, "loss": 0.0003, "step": 16804 }, { "epoch": 2.01, "learning_rate": 1.793881223755249e-06, "loss": 0.0002, "step": 16805 }, { "epoch": 2.01, "learning_rate": 1.7938092381523695e-06, "loss": 0.0004, "step": 16806 }, { "epoch": 2.01, "learning_rate": 1.79373725254949e-06, "loss": 0.0002, "step": 16807 }, { "epoch": 2.01, "learning_rate": 1.7936652669466106e-06, "loss": 0.0003, "step": 16808 }, { "epoch": 2.01, "learning_rate": 1.7935932813437314e-06, "loss": 0.0001, "step": 16809 }, { "epoch": 2.01, "learning_rate": 1.7935212957408518e-06, "loss": 0.0004, "step": 16810 }, { "epoch": 2.01, "learning_rate": 1.7934493101379726e-06, "loss": 0.0001, "step": 16811 }, { "epoch": 2.01, "learning_rate": 1.793377324535093e-06, "loss": 0.0003, "step": 16812 }, { "epoch": 2.01, "learning_rate": 1.7933053389322137e-06, "loss": 0.0001, "step": 16813 }, { "epoch": 2.01, "learning_rate": 1.793233353329334e-06, "loss": 0.0005, "step": 16814 }, { "epoch": 2.01, "learning_rate": 1.7931613677264548e-06, "loss": 0.0002, "step": 16815 }, { "epoch": 2.02, "learning_rate": 1.7930893821235754e-06, "loss": 0.0001, "step": 16816 }, { "epoch": 2.02, "learning_rate": 1.7930173965206957e-06, "loss": 0.0002, "step": 16817 }, { "epoch": 2.02, "learning_rate": 1.7929454109178165e-06, "loss": 0.0002, "step": 16818 }, { "epoch": 2.02, "learning_rate": 1.7928734253149369e-06, "loss": 0.0003, "step": 16819 }, { "epoch": 2.02, "learning_rate": 1.7928014397120577e-06, "loss": 0.0003, "step": 16820 }, { "epoch": 2.02, "learning_rate": 1.792729454109178e-06, "loss": 0.0002, "step": 16821 }, { "epoch": 2.02, "learning_rate": 1.7926574685062988e-06, "loss": 0.0003, "step": 16822 }, { "epoch": 2.02, "learning_rate": 1.7925854829034194e-06, "loss": 0.0002, "step": 16823 }, { "epoch": 2.02, "learning_rate": 1.7925134973005401e-06, "loss": 0.0002, "step": 16824 }, { "epoch": 2.02, "learning_rate": 1.7924415116976605e-06, "loss": 0.0001, "step": 16825 }, { "epoch": 2.02, "learning_rate": 1.7923695260947813e-06, "loss": 0.0003, "step": 16826 }, { "epoch": 2.02, "learning_rate": 1.7922975404919016e-06, "loss": 0.0005, "step": 16827 }, { "epoch": 2.02, "learning_rate": 1.7922255548890224e-06, "loss": 0.0001, "step": 16828 }, { "epoch": 2.02, "learning_rate": 1.7921535692861428e-06, "loss": 0.0002, "step": 16829 }, { "epoch": 2.02, "learning_rate": 1.7920815836832633e-06, "loss": 0.0002, "step": 16830 }, { "epoch": 2.02, "learning_rate": 1.792009598080384e-06, "loss": 0.0002, "step": 16831 }, { "epoch": 2.02, "learning_rate": 1.7919376124775045e-06, "loss": 0.0001, "step": 16832 }, { "epoch": 2.02, "learning_rate": 1.7918656268746252e-06, "loss": 0.0002, "step": 16833 }, { "epoch": 2.02, "learning_rate": 1.7917936412717456e-06, "loss": 0.0002, "step": 16834 }, { "epoch": 2.02, "learning_rate": 1.7917216556688664e-06, "loss": 0.0002, "step": 16835 }, { "epoch": 2.02, "learning_rate": 1.7916496700659867e-06, "loss": 0.0004, "step": 16836 }, { "epoch": 2.02, "learning_rate": 1.7915776844631075e-06, "loss": 0.0002, "step": 16837 }, { "epoch": 2.02, "learning_rate": 1.791505698860228e-06, "loss": 0.0003, "step": 16838 }, { "epoch": 2.02, "learning_rate": 1.7914337132573486e-06, "loss": 0.0004, "step": 16839 }, { "epoch": 2.02, "learning_rate": 1.7913617276544692e-06, "loss": 0.0001, "step": 16840 }, { "epoch": 2.02, "learning_rate": 1.7912897420515896e-06, "loss": 0.0002, "step": 16841 }, { "epoch": 2.02, "learning_rate": 1.7912177564487103e-06, "loss": 0.0004, "step": 16842 }, { "epoch": 2.02, "learning_rate": 1.7911457708458307e-06, "loss": 0.0004, "step": 16843 }, { "epoch": 2.02, "learning_rate": 1.7910737852429515e-06, "loss": 0.0002, "step": 16844 }, { "epoch": 2.02, "learning_rate": 1.791001799640072e-06, "loss": 0.0003, "step": 16845 }, { "epoch": 2.02, "learning_rate": 1.7909298140371926e-06, "loss": 0.0002, "step": 16846 }, { "epoch": 2.02, "learning_rate": 1.7908578284343132e-06, "loss": 0.0001, "step": 16847 }, { "epoch": 2.02, "learning_rate": 1.790785842831434e-06, "loss": 0.0003, "step": 16848 }, { "epoch": 2.02, "learning_rate": 1.7907138572285543e-06, "loss": 0.0002, "step": 16849 }, { "epoch": 2.02, "learning_rate": 1.790641871625675e-06, "loss": 0.0007, "step": 16850 }, { "epoch": 2.02, "learning_rate": 1.7905698860227954e-06, "loss": 0.0002, "step": 16851 }, { "epoch": 2.02, "learning_rate": 1.790497900419916e-06, "loss": 0.0003, "step": 16852 }, { "epoch": 2.02, "learning_rate": 1.7904259148170366e-06, "loss": 0.0004, "step": 16853 }, { "epoch": 2.02, "learning_rate": 1.7903539292141571e-06, "loss": 0.0008, "step": 16854 }, { "epoch": 2.02, "learning_rate": 1.790281943611278e-06, "loss": 0.0002, "step": 16855 }, { "epoch": 2.02, "learning_rate": 1.7902099580083983e-06, "loss": 0.0002, "step": 16856 }, { "epoch": 2.02, "learning_rate": 1.790137972405519e-06, "loss": 0.0002, "step": 16857 }, { "epoch": 2.02, "learning_rate": 1.7900659868026394e-06, "loss": 0.0002, "step": 16858 }, { "epoch": 2.02, "learning_rate": 1.7899940011997602e-06, "loss": 0.0006, "step": 16859 }, { "epoch": 2.02, "learning_rate": 1.7899220155968805e-06, "loss": 0.0002, "step": 16860 }, { "epoch": 2.02, "learning_rate": 1.7898500299940013e-06, "loss": 0.0004, "step": 16861 }, { "epoch": 2.02, "learning_rate": 1.7897780443911219e-06, "loss": 0.0003, "step": 16862 }, { "epoch": 2.02, "learning_rate": 1.7897060587882427e-06, "loss": 0.0002, "step": 16863 }, { "epoch": 2.02, "learning_rate": 1.789634073185363e-06, "loss": 0.0001, "step": 16864 }, { "epoch": 2.02, "learning_rate": 1.7895620875824834e-06, "loss": 0.0002, "step": 16865 }, { "epoch": 2.02, "learning_rate": 1.7894901019796042e-06, "loss": 0.0003, "step": 16866 }, { "epoch": 2.02, "learning_rate": 1.7894181163767245e-06, "loss": 0.0001, "step": 16867 }, { "epoch": 2.02, "learning_rate": 1.7893461307738453e-06, "loss": 0.0002, "step": 16868 }, { "epoch": 2.02, "learning_rate": 1.7892741451709659e-06, "loss": 0.0105, "step": 16869 }, { "epoch": 2.02, "learning_rate": 1.7892021595680866e-06, "loss": 0.0002, "step": 16870 }, { "epoch": 2.02, "learning_rate": 1.789130173965207e-06, "loss": 0.0001, "step": 16871 }, { "epoch": 2.02, "learning_rate": 1.7890581883623278e-06, "loss": 0.0001, "step": 16872 }, { "epoch": 2.02, "learning_rate": 1.7889862027594481e-06, "loss": 0.0002, "step": 16873 }, { "epoch": 2.02, "learning_rate": 1.788914217156569e-06, "loss": 0.0002, "step": 16874 }, { "epoch": 2.02, "learning_rate": 1.7888422315536893e-06, "loss": 0.0002, "step": 16875 }, { "epoch": 2.02, "learning_rate": 1.7887702459508098e-06, "loss": 0.0002, "step": 16876 }, { "epoch": 2.02, "learning_rate": 1.7886982603479304e-06, "loss": 0.0001, "step": 16877 }, { "epoch": 2.02, "learning_rate": 1.788626274745051e-06, "loss": 0.0003, "step": 16878 }, { "epoch": 2.02, "learning_rate": 1.7885542891421717e-06, "loss": 0.0003, "step": 16879 }, { "epoch": 2.02, "learning_rate": 1.788482303539292e-06, "loss": 0.0004, "step": 16880 }, { "epoch": 2.02, "learning_rate": 1.7884103179364129e-06, "loss": 0.0001, "step": 16881 }, { "epoch": 2.02, "learning_rate": 1.7883383323335332e-06, "loss": 0.0013, "step": 16882 }, { "epoch": 2.02, "learning_rate": 1.788266346730654e-06, "loss": 0.0002, "step": 16883 }, { "epoch": 2.02, "learning_rate": 1.7881943611277744e-06, "loss": 0.0002, "step": 16884 }, { "epoch": 2.02, "learning_rate": 1.7881223755248951e-06, "loss": 0.0002, "step": 16885 }, { "epoch": 2.02, "learning_rate": 1.7880503899220157e-06, "loss": 0.2268, "step": 16886 }, { "epoch": 2.02, "learning_rate": 1.787978404319136e-06, "loss": 0.0002, "step": 16887 }, { "epoch": 2.02, "learning_rate": 1.7879064187162568e-06, "loss": 0.0001, "step": 16888 }, { "epoch": 2.02, "learning_rate": 1.7878344331133772e-06, "loss": 0.0003, "step": 16889 }, { "epoch": 2.02, "learning_rate": 1.787762447510498e-06, "loss": 0.0002, "step": 16890 }, { "epoch": 2.02, "learning_rate": 1.7876904619076183e-06, "loss": 0.0001, "step": 16891 }, { "epoch": 2.02, "learning_rate": 1.787618476304739e-06, "loss": 0.0001, "step": 16892 }, { "epoch": 2.02, "learning_rate": 1.7875464907018597e-06, "loss": 0.0001, "step": 16893 }, { "epoch": 2.02, "learning_rate": 1.7874745050989804e-06, "loss": 0.0002, "step": 16894 }, { "epoch": 2.02, "learning_rate": 1.7874025194961008e-06, "loss": 0.0001, "step": 16895 }, { "epoch": 2.02, "learning_rate": 1.7873305338932216e-06, "loss": 0.0038, "step": 16896 }, { "epoch": 2.02, "learning_rate": 1.787258548290342e-06, "loss": 0.0001, "step": 16897 }, { "epoch": 2.02, "learning_rate": 1.7871865626874623e-06, "loss": 0.0002, "step": 16898 }, { "epoch": 2.03, "learning_rate": 1.787114577084583e-06, "loss": 0.0002, "step": 16899 }, { "epoch": 2.03, "learning_rate": 1.7870425914817036e-06, "loss": 0.0001, "step": 16900 }, { "epoch": 2.03, "learning_rate": 1.7869706058788244e-06, "loss": 0.0002, "step": 16901 }, { "epoch": 2.03, "learning_rate": 1.7868986202759448e-06, "loss": 0.0003, "step": 16902 }, { "epoch": 2.03, "learning_rate": 1.7868266346730655e-06, "loss": 0.0002, "step": 16903 }, { "epoch": 2.03, "learning_rate": 1.786754649070186e-06, "loss": 0.0001, "step": 16904 }, { "epoch": 2.03, "learning_rate": 1.7866826634673067e-06, "loss": 0.0002, "step": 16905 }, { "epoch": 2.03, "learning_rate": 1.786610677864427e-06, "loss": 0.0003, "step": 16906 }, { "epoch": 2.03, "learning_rate": 1.7865386922615478e-06, "loss": 0.0003, "step": 16907 }, { "epoch": 2.03, "learning_rate": 1.7864667066586684e-06, "loss": 0.0002, "step": 16908 }, { "epoch": 2.03, "learning_rate": 1.786394721055789e-06, "loss": 0.0002, "step": 16909 }, { "epoch": 2.03, "learning_rate": 1.7863227354529095e-06, "loss": 0.0003, "step": 16910 }, { "epoch": 2.03, "learning_rate": 1.7862507498500299e-06, "loss": 0.0001, "step": 16911 }, { "epoch": 2.03, "learning_rate": 1.7861787642471506e-06, "loss": 0.0001, "step": 16912 }, { "epoch": 2.03, "learning_rate": 1.786106778644271e-06, "loss": 0.0002, "step": 16913 }, { "epoch": 2.03, "learning_rate": 1.7860347930413918e-06, "loss": 0.0001, "step": 16914 }, { "epoch": 2.03, "learning_rate": 1.7859628074385123e-06, "loss": 0.0028, "step": 16915 }, { "epoch": 2.03, "learning_rate": 1.785890821835633e-06, "loss": 0.0001, "step": 16916 }, { "epoch": 2.03, "learning_rate": 1.7858188362327535e-06, "loss": 0.0003, "step": 16917 }, { "epoch": 2.03, "learning_rate": 1.7857468506298743e-06, "loss": 0.0002, "step": 16918 }, { "epoch": 2.03, "learning_rate": 1.7856748650269946e-06, "loss": 0.0002, "step": 16919 }, { "epoch": 2.03, "learning_rate": 1.7856028794241154e-06, "loss": 0.0004, "step": 16920 }, { "epoch": 2.03, "learning_rate": 1.7855308938212357e-06, "loss": 0.0002, "step": 16921 }, { "epoch": 2.03, "learning_rate": 1.7854589082183563e-06, "loss": 0.0001, "step": 16922 }, { "epoch": 2.03, "learning_rate": 1.7853869226154769e-06, "loss": 0.0002, "step": 16923 }, { "epoch": 2.03, "learning_rate": 1.7853149370125975e-06, "loss": 0.0002, "step": 16924 }, { "epoch": 2.03, "learning_rate": 1.7852429514097182e-06, "loss": 0.0006, "step": 16925 }, { "epoch": 2.03, "learning_rate": 1.7851709658068386e-06, "loss": 0.0001, "step": 16926 }, { "epoch": 2.03, "learning_rate": 1.7850989802039594e-06, "loss": 0.0001, "step": 16927 }, { "epoch": 2.03, "learning_rate": 1.7850269946010797e-06, "loss": 0.0001, "step": 16928 }, { "epoch": 2.03, "learning_rate": 1.7849550089982005e-06, "loss": 0.0002, "step": 16929 }, { "epoch": 2.03, "learning_rate": 1.7848830233953209e-06, "loss": 0.0002, "step": 16930 }, { "epoch": 2.03, "learning_rate": 1.7848110377924416e-06, "loss": 0.0002, "step": 16931 }, { "epoch": 2.03, "learning_rate": 1.7847390521895622e-06, "loss": 0.0002, "step": 16932 }, { "epoch": 2.03, "learning_rate": 1.7846670665866826e-06, "loss": 0.0001, "step": 16933 }, { "epoch": 2.03, "learning_rate": 1.7845950809838033e-06, "loss": 0.0001, "step": 16934 }, { "epoch": 2.03, "learning_rate": 1.7845230953809237e-06, "loss": 0.0002, "step": 16935 }, { "epoch": 2.03, "learning_rate": 1.7844511097780445e-06, "loss": 0.0002, "step": 16936 }, { "epoch": 2.03, "learning_rate": 1.7843791241751648e-06, "loss": 0.0001, "step": 16937 }, { "epoch": 2.03, "learning_rate": 1.7843071385722856e-06, "loss": 0.1368, "step": 16938 }, { "epoch": 2.03, "learning_rate": 1.7842351529694062e-06, "loss": 0.0001, "step": 16939 }, { "epoch": 2.03, "learning_rate": 1.784163167366527e-06, "loss": 0.0007, "step": 16940 }, { "epoch": 2.03, "learning_rate": 1.7840911817636473e-06, "loss": 1.0061, "step": 16941 }, { "epoch": 2.03, "learning_rate": 1.784019196160768e-06, "loss": 0.0002, "step": 16942 }, { "epoch": 2.03, "learning_rate": 1.7839472105578884e-06, "loss": 0.0004, "step": 16943 }, { "epoch": 2.03, "learning_rate": 1.7838752249550092e-06, "loss": 0.0002, "step": 16944 }, { "epoch": 2.03, "learning_rate": 1.7838032393521296e-06, "loss": 0.0002, "step": 16945 }, { "epoch": 2.03, "learning_rate": 1.7837312537492501e-06, "loss": 0.0002, "step": 16946 }, { "epoch": 2.03, "learning_rate": 1.783659268146371e-06, "loss": 0.0002, "step": 16947 }, { "epoch": 2.03, "learning_rate": 1.7835872825434913e-06, "loss": 0.0003, "step": 16948 }, { "epoch": 2.03, "learning_rate": 1.783515296940612e-06, "loss": 0.1802, "step": 16949 }, { "epoch": 2.03, "learning_rate": 1.7834433113377324e-06, "loss": 0.0002, "step": 16950 }, { "epoch": 2.03, "learning_rate": 1.7833713257348532e-06, "loss": 0.0002, "step": 16951 }, { "epoch": 2.03, "learning_rate": 1.7832993401319735e-06, "loss": 0.0004, "step": 16952 }, { "epoch": 2.03, "learning_rate": 1.7832273545290943e-06, "loss": 0.001, "step": 16953 }, { "epoch": 2.03, "learning_rate": 1.7831553689262149e-06, "loss": 0.0009, "step": 16954 }, { "epoch": 2.03, "learning_rate": 1.7830833833233354e-06, "loss": 0.133, "step": 16955 }, { "epoch": 2.03, "learning_rate": 1.783011397720456e-06, "loss": 0.0008, "step": 16956 }, { "epoch": 2.03, "learning_rate": 1.7829394121175764e-06, "loss": 0.0003, "step": 16957 }, { "epoch": 2.03, "learning_rate": 1.7828674265146971e-06, "loss": 0.0003, "step": 16958 }, { "epoch": 2.03, "learning_rate": 1.7827954409118175e-06, "loss": 0.0003, "step": 16959 }, { "epoch": 2.03, "learning_rate": 1.7827234553089383e-06, "loss": 0.0003, "step": 16960 }, { "epoch": 2.03, "learning_rate": 1.7826514697060588e-06, "loss": 0.0001, "step": 16961 }, { "epoch": 2.03, "learning_rate": 1.7825794841031794e-06, "loss": 0.0002, "step": 16962 }, { "epoch": 2.03, "learning_rate": 1.7825074985003e-06, "loss": 0.0002, "step": 16963 }, { "epoch": 2.03, "learning_rate": 1.7824355128974208e-06, "loss": 0.0004, "step": 16964 }, { "epoch": 2.03, "learning_rate": 1.7823635272945411e-06, "loss": 0.0001, "step": 16965 }, { "epoch": 2.03, "learning_rate": 1.7822915416916619e-06, "loss": 0.0002, "step": 16966 }, { "epoch": 2.03, "learning_rate": 1.7822195560887822e-06, "loss": 0.0002, "step": 16967 }, { "epoch": 2.03, "learning_rate": 1.7821475704859028e-06, "loss": 0.0002, "step": 16968 }, { "epoch": 2.03, "learning_rate": 1.7820755848830234e-06, "loss": 0.0003, "step": 16969 }, { "epoch": 2.03, "learning_rate": 1.782003599280144e-06, "loss": 0.0003, "step": 16970 }, { "epoch": 2.03, "learning_rate": 1.7819316136772647e-06, "loss": 0.0001, "step": 16971 }, { "epoch": 2.03, "learning_rate": 1.781859628074385e-06, "loss": 0.0001, "step": 16972 }, { "epoch": 2.03, "learning_rate": 1.7817876424715059e-06, "loss": 0.0002, "step": 16973 }, { "epoch": 2.03, "learning_rate": 1.7817156568686262e-06, "loss": 0.0002, "step": 16974 }, { "epoch": 2.03, "learning_rate": 1.781643671265747e-06, "loss": 0.0002, "step": 16975 }, { "epoch": 2.03, "learning_rate": 1.7815716856628673e-06, "loss": 0.0001, "step": 16976 }, { "epoch": 2.03, "learning_rate": 1.7814997000599881e-06, "loss": 0.0001, "step": 16977 }, { "epoch": 2.03, "learning_rate": 1.7814277144571087e-06, "loss": 0.0001, "step": 16978 }, { "epoch": 2.03, "learning_rate": 1.781355728854229e-06, "loss": 0.0002, "step": 16979 }, { "epoch": 2.03, "learning_rate": 1.7812837432513498e-06, "loss": 0.0002, "step": 16980 }, { "epoch": 2.03, "learning_rate": 1.7812117576484702e-06, "loss": 0.5253, "step": 16981 }, { "epoch": 2.03, "learning_rate": 1.781139772045591e-06, "loss": 0.0001, "step": 16982 }, { "epoch": 2.04, "learning_rate": 1.7810677864427113e-06, "loss": 0.0001, "step": 16983 }, { "epoch": 2.04, "learning_rate": 1.780995800839832e-06, "loss": 0.0002, "step": 16984 }, { "epoch": 2.04, "learning_rate": 1.7809238152369527e-06, "loss": 0.0002, "step": 16985 }, { "epoch": 2.04, "learning_rate": 1.7808518296340734e-06, "loss": 0.0003, "step": 16986 }, { "epoch": 2.04, "learning_rate": 1.7807798440311938e-06, "loss": 0.0002, "step": 16987 }, { "epoch": 2.04, "learning_rate": 1.7807078584283146e-06, "loss": 0.0001, "step": 16988 }, { "epoch": 2.04, "learning_rate": 1.780635872825435e-06, "loss": 0.0002, "step": 16989 }, { "epoch": 2.04, "learning_rate": 1.7805638872225557e-06, "loss": 0.0001, "step": 16990 }, { "epoch": 2.04, "learning_rate": 1.780491901619676e-06, "loss": 0.0002, "step": 16991 }, { "epoch": 2.04, "learning_rate": 1.7804199160167966e-06, "loss": 0.0002, "step": 16992 }, { "epoch": 2.04, "learning_rate": 1.7803479304139174e-06, "loss": 0.0002, "step": 16993 }, { "epoch": 2.04, "learning_rate": 1.7802759448110378e-06, "loss": 0.0001, "step": 16994 }, { "epoch": 2.04, "learning_rate": 1.7802039592081585e-06, "loss": 0.0001, "step": 16995 }, { "epoch": 2.04, "learning_rate": 1.780131973605279e-06, "loss": 0.0001, "step": 16996 }, { "epoch": 2.04, "learning_rate": 1.7800599880023997e-06, "loss": 0.0002, "step": 16997 }, { "epoch": 2.04, "learning_rate": 1.77998800239952e-06, "loss": 0.0002, "step": 16998 }, { "epoch": 2.04, "learning_rate": 1.7799160167966408e-06, "loss": 0.0001, "step": 16999 }, { "epoch": 2.04, "learning_rate": 1.7798440311937612e-06, "loss": 0.0002, "step": 17000 }, { "epoch": 2.04, "learning_rate": 1.779772045590882e-06, "loss": 0.0002, "step": 17001 }, { "epoch": 2.04, "learning_rate": 1.7797000599880025e-06, "loss": 0.0003, "step": 17002 }, { "epoch": 2.04, "learning_rate": 1.7796280743851229e-06, "loss": 0.0002, "step": 17003 }, { "epoch": 2.04, "learning_rate": 1.7795560887822436e-06, "loss": 0.0002, "step": 17004 }, { "epoch": 2.04, "learning_rate": 1.779484103179364e-06, "loss": 0.0001, "step": 17005 }, { "epoch": 2.04, "learning_rate": 1.7794121175764848e-06, "loss": 0.0001, "step": 17006 }, { "epoch": 2.04, "learning_rate": 1.7793401319736051e-06, "loss": 0.0002, "step": 17007 }, { "epoch": 2.04, "learning_rate": 1.779268146370726e-06, "loss": 0.0001, "step": 17008 }, { "epoch": 2.04, "learning_rate": 1.7791961607678465e-06, "loss": 0.0003, "step": 17009 }, { "epoch": 2.04, "learning_rate": 1.7791241751649673e-06, "loss": 0.0002, "step": 17010 }, { "epoch": 2.04, "learning_rate": 1.7790521895620876e-06, "loss": 0.0002, "step": 17011 }, { "epoch": 2.04, "learning_rate": 1.7789802039592084e-06, "loss": 0.0002, "step": 17012 }, { "epoch": 2.04, "learning_rate": 1.7789082183563287e-06, "loss": 0.0002, "step": 17013 }, { "epoch": 2.04, "learning_rate": 1.778836232753449e-06, "loss": 0.0002, "step": 17014 }, { "epoch": 2.04, "learning_rate": 1.7787642471505699e-06, "loss": 0.0002, "step": 17015 }, { "epoch": 2.04, "learning_rate": 1.7786922615476904e-06, "loss": 0.0001, "step": 17016 }, { "epoch": 2.04, "learning_rate": 1.7786202759448112e-06, "loss": 0.0002, "step": 17017 }, { "epoch": 2.04, "learning_rate": 1.7785482903419316e-06, "loss": 0.0002, "step": 17018 }, { "epoch": 2.04, "learning_rate": 1.7784763047390524e-06, "loss": 0.0001, "step": 17019 }, { "epoch": 2.04, "learning_rate": 1.7784043191361727e-06, "loss": 0.0002, "step": 17020 }, { "epoch": 2.04, "learning_rate": 1.7783323335332935e-06, "loss": 0.0002, "step": 17021 }, { "epoch": 2.04, "learning_rate": 1.7782603479304138e-06, "loss": 0.0001, "step": 17022 }, { "epoch": 2.04, "learning_rate": 1.7781883623275346e-06, "loss": 0.0003, "step": 17023 }, { "epoch": 2.04, "learning_rate": 1.7781163767246552e-06, "loss": 0.0002, "step": 17024 }, { "epoch": 2.04, "learning_rate": 1.7780443911217755e-06, "loss": 0.0001, "step": 17025 }, { "epoch": 2.04, "learning_rate": 1.7779724055188963e-06, "loss": 0.0002, "step": 17026 }, { "epoch": 2.04, "learning_rate": 1.7779004199160167e-06, "loss": 0.0001, "step": 17027 }, { "epoch": 2.04, "learning_rate": 1.7778284343131375e-06, "loss": 0.0001, "step": 17028 }, { "epoch": 2.04, "learning_rate": 1.7777564487102578e-06, "loss": 0.0004, "step": 17029 }, { "epoch": 2.04, "learning_rate": 1.7776844631073786e-06, "loss": 0.0003, "step": 17030 }, { "epoch": 2.04, "learning_rate": 1.7776124775044992e-06, "loss": 0.0003, "step": 17031 }, { "epoch": 2.04, "learning_rate": 1.7775404919016197e-06, "loss": 0.0004, "step": 17032 }, { "epoch": 2.04, "learning_rate": 1.7774685062987403e-06, "loss": 0.0002, "step": 17033 }, { "epoch": 2.04, "learning_rate": 1.777396520695861e-06, "loss": 0.0002, "step": 17034 }, { "epoch": 2.04, "learning_rate": 1.7773245350929814e-06, "loss": 0.0001, "step": 17035 }, { "epoch": 2.04, "learning_rate": 1.7772525494901022e-06, "loss": 0.0002, "step": 17036 }, { "epoch": 2.04, "learning_rate": 1.7771805638872226e-06, "loss": 0.0001, "step": 17037 }, { "epoch": 2.04, "learning_rate": 1.7771085782843431e-06, "loss": 0.0002, "step": 17038 }, { "epoch": 2.04, "learning_rate": 1.7770365926814637e-06, "loss": 0.0002, "step": 17039 }, { "epoch": 2.04, "learning_rate": 1.7769646070785843e-06, "loss": 0.0001, "step": 17040 }, { "epoch": 2.04, "learning_rate": 1.776892621475705e-06, "loss": 0.0001, "step": 17041 }, { "epoch": 2.04, "learning_rate": 1.7768206358728254e-06, "loss": 0.0001, "step": 17042 }, { "epoch": 2.04, "learning_rate": 1.7767486502699462e-06, "loss": 0.0002, "step": 17043 }, { "epoch": 2.04, "learning_rate": 1.7766766646670665e-06, "loss": 0.0002, "step": 17044 }, { "epoch": 2.04, "learning_rate": 1.7766046790641873e-06, "loss": 0.0001, "step": 17045 }, { "epoch": 2.04, "learning_rate": 1.7765326934613077e-06, "loss": 0.0001, "step": 17046 }, { "epoch": 2.04, "learning_rate": 1.7764607078584284e-06, "loss": 0.0001, "step": 17047 }, { "epoch": 2.04, "learning_rate": 1.776388722255549e-06, "loss": 0.0001, "step": 17048 }, { "epoch": 2.04, "learning_rate": 1.7763167366526694e-06, "loss": 0.0002, "step": 17049 }, { "epoch": 2.04, "learning_rate": 1.7762447510497901e-06, "loss": 0.0001, "step": 17050 }, { "epoch": 2.04, "learning_rate": 1.7761727654469105e-06, "loss": 0.0001, "step": 17051 }, { "epoch": 2.04, "learning_rate": 1.7761007798440313e-06, "loss": 0.0005, "step": 17052 }, { "epoch": 2.04, "learning_rate": 1.7760287942411516e-06, "loss": 0.8263, "step": 17053 }, { "epoch": 2.04, "learning_rate": 1.7759568086382724e-06, "loss": 0.0001, "step": 17054 }, { "epoch": 2.04, "learning_rate": 1.775884823035393e-06, "loss": 0.0002, "step": 17055 }, { "epoch": 2.04, "learning_rate": 1.7758128374325137e-06, "loss": 0.0004, "step": 17056 }, { "epoch": 2.04, "learning_rate": 1.775740851829634e-06, "loss": 0.0003, "step": 17057 }, { "epoch": 2.04, "learning_rate": 1.7756688662267549e-06, "loss": 0.8119, "step": 17058 }, { "epoch": 2.04, "learning_rate": 1.7755968806238752e-06, "loss": 0.0002, "step": 17059 }, { "epoch": 2.04, "learning_rate": 1.7755248950209956e-06, "loss": 0.0001, "step": 17060 }, { "epoch": 2.04, "learning_rate": 1.7754529094181164e-06, "loss": 0.0002, "step": 17061 }, { "epoch": 2.04, "learning_rate": 1.775380923815237e-06, "loss": 0.0001, "step": 17062 }, { "epoch": 2.04, "learning_rate": 1.7753089382123577e-06, "loss": 0.0002, "step": 17063 }, { "epoch": 2.04, "learning_rate": 1.775236952609478e-06, "loss": 0.0002, "step": 17064 }, { "epoch": 2.04, "learning_rate": 1.7751649670065989e-06, "loss": 0.0003, "step": 17065 }, { "epoch": 2.05, "learning_rate": 1.7750929814037192e-06, "loss": 0.0002, "step": 17066 }, { "epoch": 2.05, "learning_rate": 1.77502099580084e-06, "loss": 0.0003, "step": 17067 }, { "epoch": 2.05, "learning_rate": 1.7749490101979603e-06, "loss": 0.0002, "step": 17068 }, { "epoch": 2.05, "learning_rate": 1.7748770245950811e-06, "loss": 0.0001, "step": 17069 }, { "epoch": 2.05, "learning_rate": 1.7748050389922017e-06, "loss": 0.0002, "step": 17070 }, { "epoch": 2.05, "learning_rate": 1.7747330533893223e-06, "loss": 0.0001, "step": 17071 }, { "epoch": 2.05, "learning_rate": 1.7746610677864428e-06, "loss": 0.0004, "step": 17072 }, { "epoch": 2.05, "learning_rate": 1.7745890821835632e-06, "loss": 0.0001, "step": 17073 }, { "epoch": 2.05, "learning_rate": 1.774517096580684e-06, "loss": 0.0001, "step": 17074 }, { "epoch": 2.05, "learning_rate": 1.7744451109778043e-06, "loss": 0.0003, "step": 17075 }, { "epoch": 2.05, "learning_rate": 1.774373125374925e-06, "loss": 0.0002, "step": 17076 }, { "epoch": 2.05, "learning_rate": 1.7743011397720457e-06, "loss": 0.0002, "step": 17077 }, { "epoch": 2.05, "learning_rate": 1.7742291541691662e-06, "loss": 0.0006, "step": 17078 }, { "epoch": 2.05, "learning_rate": 1.7741571685662868e-06, "loss": 0.0001, "step": 17079 }, { "epoch": 2.05, "learning_rate": 1.7740851829634076e-06, "loss": 0.0001, "step": 17080 }, { "epoch": 2.05, "learning_rate": 1.774013197360528e-06, "loss": 0.0002, "step": 17081 }, { "epoch": 2.05, "learning_rate": 1.7739412117576487e-06, "loss": 0.0002, "step": 17082 }, { "epoch": 2.05, "learning_rate": 1.773869226154769e-06, "loss": 0.0001, "step": 17083 }, { "epoch": 2.05, "learning_rate": 1.7737972405518896e-06, "loss": 0.0002, "step": 17084 }, { "epoch": 2.05, "learning_rate": 1.7737252549490102e-06, "loss": 0.0001, "step": 17085 }, { "epoch": 2.05, "learning_rate": 1.7736532693461308e-06, "loss": 0.0003, "step": 17086 }, { "epoch": 2.05, "learning_rate": 1.7735812837432515e-06, "loss": 0.0002, "step": 17087 }, { "epoch": 2.05, "learning_rate": 1.7735092981403719e-06, "loss": 0.0001, "step": 17088 }, { "epoch": 2.05, "learning_rate": 1.7734373125374927e-06, "loss": 0.0001, "step": 17089 }, { "epoch": 2.05, "learning_rate": 1.773365326934613e-06, "loss": 0.0001, "step": 17090 }, { "epoch": 2.05, "learning_rate": 1.7732933413317338e-06, "loss": 0.0002, "step": 17091 }, { "epoch": 2.05, "learning_rate": 1.7732213557288542e-06, "loss": 0.0002, "step": 17092 }, { "epoch": 2.05, "learning_rate": 1.773149370125975e-06, "loss": 0.0002, "step": 17093 }, { "epoch": 2.05, "learning_rate": 1.7730773845230955e-06, "loss": 0.0003, "step": 17094 }, { "epoch": 2.05, "learning_rate": 1.7730053989202159e-06, "loss": 0.0001, "step": 17095 }, { "epoch": 2.05, "learning_rate": 1.7729334133173366e-06, "loss": 0.0001, "step": 17096 }, { "epoch": 2.05, "learning_rate": 1.772861427714457e-06, "loss": 0.2721, "step": 17097 }, { "epoch": 2.05, "learning_rate": 1.7727894421115778e-06, "loss": 0.0007, "step": 17098 }, { "epoch": 2.05, "learning_rate": 1.7727174565086981e-06, "loss": 0.0002, "step": 17099 }, { "epoch": 2.05, "learning_rate": 1.772645470905819e-06, "loss": 0.0001, "step": 17100 }, { "epoch": 2.05, "learning_rate": 1.7725734853029395e-06, "loss": 0.0002, "step": 17101 }, { "epoch": 2.05, "learning_rate": 1.7725014997000602e-06, "loss": 0.0006, "step": 17102 }, { "epoch": 2.05, "learning_rate": 1.7724295140971806e-06, "loss": 0.0001, "step": 17103 }, { "epoch": 2.05, "learning_rate": 1.7723575284943014e-06, "loss": 0.0004, "step": 17104 }, { "epoch": 2.05, "learning_rate": 1.7722855428914217e-06, "loss": 0.0003, "step": 17105 }, { "epoch": 2.05, "learning_rate": 1.772213557288542e-06, "loss": 0.0007, "step": 17106 }, { "epoch": 2.05, "learning_rate": 1.7721415716856629e-06, "loss": 0.0002, "step": 17107 }, { "epoch": 2.05, "learning_rate": 1.7720695860827834e-06, "loss": 0.0001, "step": 17108 }, { "epoch": 2.05, "learning_rate": 1.7719976004799042e-06, "loss": 0.0001, "step": 17109 }, { "epoch": 2.05, "learning_rate": 1.7719256148770246e-06, "loss": 0.0002, "step": 17110 }, { "epoch": 2.05, "learning_rate": 1.7718536292741453e-06, "loss": 0.0001, "step": 17111 }, { "epoch": 2.05, "learning_rate": 1.7717816436712657e-06, "loss": 0.0004, "step": 17112 }, { "epoch": 2.05, "learning_rate": 1.7717096580683865e-06, "loss": 0.0001, "step": 17113 }, { "epoch": 2.05, "learning_rate": 1.7716376724655068e-06, "loss": 0.0001, "step": 17114 }, { "epoch": 2.05, "learning_rate": 1.7715656868626276e-06, "loss": 0.0005, "step": 17115 }, { "epoch": 2.05, "learning_rate": 1.7714937012597482e-06, "loss": 0.1278, "step": 17116 }, { "epoch": 2.05, "learning_rate": 1.7714217156568687e-06, "loss": 0.0003, "step": 17117 }, { "epoch": 2.05, "learning_rate": 1.7713497300539893e-06, "loss": 0.0466, "step": 17118 }, { "epoch": 2.05, "learning_rate": 1.7712777444511097e-06, "loss": 0.0003, "step": 17119 }, { "epoch": 2.05, "learning_rate": 1.7712057588482304e-06, "loss": 0.0005, "step": 17120 }, { "epoch": 2.05, "learning_rate": 1.7711337732453508e-06, "loss": 0.0002, "step": 17121 }, { "epoch": 2.05, "learning_rate": 1.7710617876424716e-06, "loss": 0.0002, "step": 17122 }, { "epoch": 2.05, "learning_rate": 1.7709898020395921e-06, "loss": 0.0002, "step": 17123 }, { "epoch": 2.05, "learning_rate": 1.7709178164367127e-06, "loss": 0.0002, "step": 17124 }, { "epoch": 2.05, "learning_rate": 1.7708458308338333e-06, "loss": 0.0002, "step": 17125 }, { "epoch": 2.05, "learning_rate": 1.770773845230954e-06, "loss": 0.0003, "step": 17126 }, { "epoch": 2.05, "learning_rate": 1.7707018596280744e-06, "loss": 0.0002, "step": 17127 }, { "epoch": 2.05, "learning_rate": 1.7706298740251952e-06, "loss": 0.0001, "step": 17128 }, { "epoch": 2.05, "learning_rate": 1.7705578884223155e-06, "loss": 0.0001, "step": 17129 }, { "epoch": 2.05, "learning_rate": 1.770485902819436e-06, "loss": 0.0002, "step": 17130 }, { "epoch": 2.05, "learning_rate": 1.7704139172165567e-06, "loss": 0.0002, "step": 17131 }, { "epoch": 2.05, "learning_rate": 1.7703419316136772e-06, "loss": 0.0007, "step": 17132 }, { "epoch": 2.05, "learning_rate": 1.770269946010798e-06, "loss": 0.0001, "step": 17133 }, { "epoch": 2.05, "learning_rate": 1.7701979604079184e-06, "loss": 0.7695, "step": 17134 }, { "epoch": 2.05, "learning_rate": 1.7701259748050392e-06, "loss": 0.0001, "step": 17135 }, { "epoch": 2.05, "learning_rate": 1.7700539892021595e-06, "loss": 0.0002, "step": 17136 }, { "epoch": 2.05, "learning_rate": 1.7699820035992803e-06, "loss": 0.0005, "step": 17137 }, { "epoch": 2.05, "learning_rate": 1.7699100179964006e-06, "loss": 0.0002, "step": 17138 }, { "epoch": 2.05, "learning_rate": 1.7698380323935214e-06, "loss": 0.0001, "step": 17139 }, { "epoch": 2.05, "learning_rate": 1.769766046790642e-06, "loss": 0.0005, "step": 17140 }, { "epoch": 2.05, "learning_rate": 1.7696940611877623e-06, "loss": 0.0004, "step": 17141 }, { "epoch": 2.05, "learning_rate": 1.7696220755848831e-06, "loss": 0.5157, "step": 17142 }, { "epoch": 2.05, "learning_rate": 1.7695500899820035e-06, "loss": 0.0002, "step": 17143 }, { "epoch": 2.05, "learning_rate": 1.7694781043791243e-06, "loss": 0.0001, "step": 17144 }, { "epoch": 2.05, "learning_rate": 1.7694061187762446e-06, "loss": 0.0001, "step": 17145 }, { "epoch": 2.05, "learning_rate": 1.7693341331733654e-06, "loss": 0.0002, "step": 17146 }, { "epoch": 2.05, "learning_rate": 1.769262147570486e-06, "loss": 0.0002, "step": 17147 }, { "epoch": 2.05, "learning_rate": 1.7691901619676065e-06, "loss": 0.0002, "step": 17148 }, { "epoch": 2.06, "learning_rate": 1.769118176364727e-06, "loss": 0.0001, "step": 17149 }, { "epoch": 2.06, "learning_rate": 1.7690461907618479e-06, "loss": 0.0002, "step": 17150 }, { "epoch": 2.06, "learning_rate": 1.7689742051589682e-06, "loss": 0.0002, "step": 17151 }, { "epoch": 2.06, "learning_rate": 1.7689022195560886e-06, "loss": 0.0002, "step": 17152 }, { "epoch": 2.06, "learning_rate": 1.7688302339532094e-06, "loss": 0.0002, "step": 17153 }, { "epoch": 2.06, "learning_rate": 1.76875824835033e-06, "loss": 0.0002, "step": 17154 }, { "epoch": 2.06, "learning_rate": 1.7686862627474505e-06, "loss": 0.0001, "step": 17155 }, { "epoch": 2.06, "learning_rate": 1.768614277144571e-06, "loss": 0.0001, "step": 17156 }, { "epoch": 2.06, "learning_rate": 1.7685422915416918e-06, "loss": 0.0003, "step": 17157 }, { "epoch": 2.06, "learning_rate": 1.7684703059388122e-06, "loss": 0.0002, "step": 17158 }, { "epoch": 2.06, "learning_rate": 1.768398320335933e-06, "loss": 0.0001, "step": 17159 }, { "epoch": 2.06, "learning_rate": 1.7683263347330533e-06, "loss": 0.0002, "step": 17160 }, { "epoch": 2.06, "learning_rate": 1.7682543491301741e-06, "loss": 0.0015, "step": 17161 }, { "epoch": 2.06, "learning_rate": 1.7681823635272945e-06, "loss": 0.0002, "step": 17162 }, { "epoch": 2.06, "learning_rate": 1.7681103779244152e-06, "loss": 0.0003, "step": 17163 }, { "epoch": 2.06, "learning_rate": 1.7680383923215358e-06, "loss": 0.0001, "step": 17164 }, { "epoch": 2.06, "learning_rate": 1.7679664067186562e-06, "loss": 0.0001, "step": 17165 }, { "epoch": 2.06, "learning_rate": 1.767894421115777e-06, "loss": 0.0002, "step": 17166 }, { "epoch": 2.06, "learning_rate": 1.7678224355128973e-06, "loss": 0.0003, "step": 17167 }, { "epoch": 2.06, "learning_rate": 1.767750449910018e-06, "loss": 0.0003, "step": 17168 }, { "epoch": 2.06, "learning_rate": 1.7676784643071384e-06, "loss": 0.0227, "step": 17169 }, { "epoch": 2.06, "learning_rate": 1.7676064787042592e-06, "loss": 0.0001, "step": 17170 }, { "epoch": 2.06, "learning_rate": 1.7675344931013798e-06, "loss": 0.0002, "step": 17171 }, { "epoch": 2.06, "learning_rate": 1.7674625074985006e-06, "loss": 0.0004, "step": 17172 }, { "epoch": 2.06, "learning_rate": 1.767390521895621e-06, "loss": 0.0002, "step": 17173 }, { "epoch": 2.06, "learning_rate": 1.7673185362927417e-06, "loss": 0.0002, "step": 17174 }, { "epoch": 2.06, "learning_rate": 1.767246550689862e-06, "loss": 0.0002, "step": 17175 }, { "epoch": 2.06, "learning_rate": 1.7671745650869824e-06, "loss": 0.0003, "step": 17176 }, { "epoch": 2.06, "learning_rate": 1.7671025794841032e-06, "loss": 0.0001, "step": 17177 }, { "epoch": 2.06, "learning_rate": 1.7670305938812237e-06, "loss": 0.0002, "step": 17178 }, { "epoch": 2.06, "learning_rate": 1.7669586082783445e-06, "loss": 0.0001, "step": 17179 }, { "epoch": 2.06, "learning_rate": 1.7668866226754649e-06, "loss": 0.0001, "step": 17180 }, { "epoch": 2.06, "learning_rate": 1.7668146370725857e-06, "loss": 0.0001, "step": 17181 }, { "epoch": 2.06, "learning_rate": 1.766742651469706e-06, "loss": 0.0002, "step": 17182 }, { "epoch": 2.06, "learning_rate": 1.7666706658668268e-06, "loss": 0.0001, "step": 17183 }, { "epoch": 2.06, "learning_rate": 1.7665986802639471e-06, "loss": 0.0001, "step": 17184 }, { "epoch": 2.06, "learning_rate": 1.766526694661068e-06, "loss": 0.0001, "step": 17185 }, { "epoch": 2.06, "learning_rate": 1.7664547090581885e-06, "loss": 0.0003, "step": 17186 }, { "epoch": 2.06, "learning_rate": 1.7663827234553088e-06, "loss": 0.0002, "step": 17187 }, { "epoch": 2.06, "learning_rate": 1.7663107378524296e-06, "loss": 0.0001, "step": 17188 }, { "epoch": 2.06, "learning_rate": 1.76623875224955e-06, "loss": 0.0001, "step": 17189 }, { "epoch": 2.06, "learning_rate": 1.7661667666466708e-06, "loss": 0.0001, "step": 17190 }, { "epoch": 2.06, "learning_rate": 1.7660947810437911e-06, "loss": 0.0002, "step": 17191 }, { "epoch": 2.06, "learning_rate": 1.7660227954409119e-06, "loss": 0.0002, "step": 17192 }, { "epoch": 2.06, "learning_rate": 1.7659508098380325e-06, "loss": 0.0002, "step": 17193 }, { "epoch": 2.06, "learning_rate": 1.765878824235153e-06, "loss": 0.0001, "step": 17194 }, { "epoch": 2.06, "learning_rate": 1.7658068386322736e-06, "loss": 0.0002, "step": 17195 }, { "epoch": 2.06, "learning_rate": 1.7657348530293944e-06, "loss": 0.0001, "step": 17196 }, { "epoch": 2.06, "learning_rate": 1.7656628674265147e-06, "loss": 0.0001, "step": 17197 }, { "epoch": 2.06, "learning_rate": 1.7655908818236355e-06, "loss": 0.0002, "step": 17198 }, { "epoch": 2.06, "learning_rate": 1.7655188962207559e-06, "loss": 0.0002, "step": 17199 }, { "epoch": 2.06, "learning_rate": 1.7654469106178764e-06, "loss": 0.0001, "step": 17200 }, { "epoch": 2.06, "learning_rate": 1.765374925014997e-06, "loss": 0.704, "step": 17201 }, { "epoch": 2.06, "learning_rate": 1.7653029394121176e-06, "loss": 0.0003, "step": 17202 }, { "epoch": 2.06, "learning_rate": 1.7652309538092383e-06, "loss": 0.0001, "step": 17203 }, { "epoch": 2.06, "learning_rate": 1.7651589682063587e-06, "loss": 0.0002, "step": 17204 }, { "epoch": 2.06, "learning_rate": 1.7650869826034795e-06, "loss": 0.0001, "step": 17205 }, { "epoch": 2.06, "learning_rate": 1.7650149970005998e-06, "loss": 0.0001, "step": 17206 }, { "epoch": 2.06, "learning_rate": 1.7649430113977206e-06, "loss": 0.0002, "step": 17207 }, { "epoch": 2.06, "learning_rate": 1.764871025794841e-06, "loss": 0.0002, "step": 17208 }, { "epoch": 2.06, "learning_rate": 1.7647990401919617e-06, "loss": 0.0001, "step": 17209 }, { "epoch": 2.06, "learning_rate": 1.7647270545890823e-06, "loss": 0.0001, "step": 17210 }, { "epoch": 2.06, "learning_rate": 1.7646550689862027e-06, "loss": 0.0002, "step": 17211 }, { "epoch": 2.06, "learning_rate": 1.7645830833833234e-06, "loss": 0.0001, "step": 17212 }, { "epoch": 2.06, "learning_rate": 1.7645110977804438e-06, "loss": 0.0002, "step": 17213 }, { "epoch": 2.06, "learning_rate": 1.7644391121775646e-06, "loss": 0.0001, "step": 17214 }, { "epoch": 2.06, "learning_rate": 1.764367126574685e-06, "loss": 0.0002, "step": 17215 }, { "epoch": 2.06, "learning_rate": 1.7642951409718057e-06, "loss": 0.0002, "step": 17216 }, { "epoch": 2.06, "learning_rate": 1.7642231553689263e-06, "loss": 0.0001, "step": 17217 }, { "epoch": 2.06, "learning_rate": 1.764151169766047e-06, "loss": 0.0001, "step": 17218 }, { "epoch": 2.06, "learning_rate": 1.7640791841631674e-06, "loss": 0.0002, "step": 17219 }, { "epoch": 2.06, "learning_rate": 1.7640071985602882e-06, "loss": 0.7454, "step": 17220 }, { "epoch": 2.06, "learning_rate": 1.7639352129574085e-06, "loss": 0.0002, "step": 17221 }, { "epoch": 2.06, "learning_rate": 1.763863227354529e-06, "loss": 0.0001, "step": 17222 }, { "epoch": 2.06, "learning_rate": 1.7637912417516497e-06, "loss": 0.0002, "step": 17223 }, { "epoch": 2.06, "learning_rate": 1.7637192561487702e-06, "loss": 0.0001, "step": 17224 }, { "epoch": 2.06, "learning_rate": 1.763647270545891e-06, "loss": 0.0001, "step": 17225 }, { "epoch": 2.06, "learning_rate": 1.7635752849430114e-06, "loss": 0.0002, "step": 17226 }, { "epoch": 2.06, "learning_rate": 1.7635032993401322e-06, "loss": 0.0002, "step": 17227 }, { "epoch": 2.06, "learning_rate": 1.7634313137372525e-06, "loss": 0.0002, "step": 17228 }, { "epoch": 2.06, "learning_rate": 1.7633593281343733e-06, "loss": 0.0001, "step": 17229 }, { "epoch": 2.06, "learning_rate": 1.7632873425314936e-06, "loss": 0.0001, "step": 17230 }, { "epoch": 2.06, "learning_rate": 1.7632153569286144e-06, "loss": 0.0037, "step": 17231 }, { "epoch": 2.06, "learning_rate": 1.763143371325735e-06, "loss": 0.0004, "step": 17232 }, { "epoch": 2.07, "learning_rate": 1.7630713857228553e-06, "loss": 0.0002, "step": 17233 }, { "epoch": 2.07, "learning_rate": 1.7629994001199761e-06, "loss": 0.0002, "step": 17234 }, { "epoch": 2.07, "learning_rate": 1.7629274145170965e-06, "loss": 0.0002, "step": 17235 }, { "epoch": 2.07, "learning_rate": 1.7628554289142173e-06, "loss": 0.0002, "step": 17236 }, { "epoch": 2.07, "learning_rate": 1.7627834433113376e-06, "loss": 0.0003, "step": 17237 }, { "epoch": 2.07, "learning_rate": 1.7627114577084584e-06, "loss": 0.0001, "step": 17238 }, { "epoch": 2.07, "learning_rate": 1.762639472105579e-06, "loss": 0.0001, "step": 17239 }, { "epoch": 2.07, "learning_rate": 1.7625674865026995e-06, "loss": 0.0001, "step": 17240 }, { "epoch": 2.07, "learning_rate": 1.76249550089982e-06, "loss": 0.0001, "step": 17241 }, { "epoch": 2.07, "learning_rate": 1.7624235152969409e-06, "loss": 0.0001, "step": 17242 }, { "epoch": 2.07, "learning_rate": 1.7623515296940612e-06, "loss": 0.0001, "step": 17243 }, { "epoch": 2.07, "learning_rate": 1.762279544091182e-06, "loss": 0.0002, "step": 17244 }, { "epoch": 2.07, "learning_rate": 1.7622075584883024e-06, "loss": 0.0002, "step": 17245 }, { "epoch": 2.07, "learning_rate": 1.762135572885423e-06, "loss": 0.0001, "step": 17246 }, { "epoch": 2.07, "learning_rate": 1.7620635872825435e-06, "loss": 0.0001, "step": 17247 }, { "epoch": 2.07, "learning_rate": 1.761991601679664e-06, "loss": 0.0001, "step": 17248 }, { "epoch": 2.07, "learning_rate": 1.7619196160767848e-06, "loss": 0.0002, "step": 17249 }, { "epoch": 2.07, "learning_rate": 1.7618476304739052e-06, "loss": 0.0001, "step": 17250 }, { "epoch": 2.07, "learning_rate": 1.761775644871026e-06, "loss": 0.0004, "step": 17251 }, { "epoch": 2.07, "learning_rate": 1.7617036592681463e-06, "loss": 0.0004, "step": 17252 }, { "epoch": 2.07, "learning_rate": 1.761631673665267e-06, "loss": 0.0002, "step": 17253 }, { "epoch": 2.07, "learning_rate": 1.7615596880623875e-06, "loss": 0.0002, "step": 17254 }, { "epoch": 2.07, "learning_rate": 1.7614877024595082e-06, "loss": 0.0001, "step": 17255 }, { "epoch": 2.07, "learning_rate": 1.7614157168566288e-06, "loss": 0.0001, "step": 17256 }, { "epoch": 2.07, "learning_rate": 1.7613437312537492e-06, "loss": 0.0002, "step": 17257 }, { "epoch": 2.07, "learning_rate": 1.76127174565087e-06, "loss": 0.6383, "step": 17258 }, { "epoch": 2.07, "learning_rate": 1.7611997600479903e-06, "loss": 0.0002, "step": 17259 }, { "epoch": 2.07, "learning_rate": 1.761127774445111e-06, "loss": 0.755, "step": 17260 }, { "epoch": 2.07, "learning_rate": 1.7610557888422314e-06, "loss": 0.0002, "step": 17261 }, { "epoch": 2.07, "learning_rate": 1.7609838032393522e-06, "loss": 0.0001, "step": 17262 }, { "epoch": 2.07, "learning_rate": 1.7609118176364728e-06, "loss": 0.0001, "step": 17263 }, { "epoch": 2.07, "learning_rate": 1.7608398320335935e-06, "loss": 0.0002, "step": 17264 }, { "epoch": 2.07, "learning_rate": 1.760767846430714e-06, "loss": 0.0001, "step": 17265 }, { "epoch": 2.07, "learning_rate": 1.7606958608278347e-06, "loss": 0.0001, "step": 17266 }, { "epoch": 2.07, "learning_rate": 1.760623875224955e-06, "loss": 0.0001, "step": 17267 }, { "epoch": 2.07, "learning_rate": 1.7605518896220754e-06, "loss": 0.0001, "step": 17268 }, { "epoch": 2.07, "learning_rate": 1.7604799040191962e-06, "loss": 0.0004, "step": 17269 }, { "epoch": 2.07, "learning_rate": 1.7604079184163167e-06, "loss": 0.0001, "step": 17270 }, { "epoch": 2.07, "learning_rate": 1.7603359328134375e-06, "loss": 0.0007, "step": 17271 }, { "epoch": 2.07, "learning_rate": 1.7602639472105579e-06, "loss": 0.0001, "step": 17272 }, { "epoch": 2.07, "learning_rate": 1.7601919616076786e-06, "loss": 0.0002, "step": 17273 }, { "epoch": 2.07, "learning_rate": 1.760119976004799e-06, "loss": 0.0001, "step": 17274 }, { "epoch": 2.07, "learning_rate": 1.7600479904019198e-06, "loss": 0.0004, "step": 17275 }, { "epoch": 2.07, "learning_rate": 1.7599760047990401e-06, "loss": 0.0001, "step": 17276 }, { "epoch": 2.07, "learning_rate": 1.759904019196161e-06, "loss": 0.0003, "step": 17277 }, { "epoch": 2.07, "learning_rate": 1.7598320335932813e-06, "loss": 0.0001, "step": 17278 }, { "epoch": 2.07, "learning_rate": 1.7597600479904018e-06, "loss": 0.0001, "step": 17279 }, { "epoch": 2.07, "learning_rate": 1.7596880623875226e-06, "loss": 0.0002, "step": 17280 }, { "epoch": 2.07, "learning_rate": 1.759616076784643e-06, "loss": 0.0066, "step": 17281 }, { "epoch": 2.07, "learning_rate": 1.7595440911817638e-06, "loss": 0.0002, "step": 17282 }, { "epoch": 2.07, "learning_rate": 1.759472105578884e-06, "loss": 0.0002, "step": 17283 }, { "epoch": 2.07, "learning_rate": 1.7594001199760049e-06, "loss": 0.0002, "step": 17284 }, { "epoch": 2.07, "learning_rate": 1.7593281343731252e-06, "loss": 0.0001, "step": 17285 }, { "epoch": 2.07, "learning_rate": 1.759256148770246e-06, "loss": 0.1451, "step": 17286 }, { "epoch": 2.07, "learning_rate": 1.7591841631673666e-06, "loss": 0.0001, "step": 17287 }, { "epoch": 2.07, "learning_rate": 1.7591121775644874e-06, "loss": 0.0003, "step": 17288 }, { "epoch": 2.07, "learning_rate": 1.7590401919616077e-06, "loss": 0.0001, "step": 17289 }, { "epoch": 2.07, "learning_rate": 1.7589682063587285e-06, "loss": 0.0001, "step": 17290 }, { "epoch": 2.07, "learning_rate": 1.7588962207558489e-06, "loss": 0.0001, "step": 17291 }, { "epoch": 2.07, "learning_rate": 1.7588242351529692e-06, "loss": 1.1593, "step": 17292 }, { "epoch": 2.07, "learning_rate": 1.75875224955009e-06, "loss": 0.0001, "step": 17293 }, { "epoch": 2.07, "learning_rate": 1.7586802639472106e-06, "loss": 0.0004, "step": 17294 }, { "epoch": 2.07, "learning_rate": 1.7586082783443313e-06, "loss": 0.0004, "step": 17295 }, { "epoch": 2.07, "learning_rate": 1.7585362927414517e-06, "loss": 0.0001, "step": 17296 }, { "epoch": 2.07, "learning_rate": 1.7584643071385725e-06, "loss": 0.0002, "step": 17297 }, { "epoch": 2.07, "learning_rate": 1.7583923215356928e-06, "loss": 0.0001, "step": 17298 }, { "epoch": 2.07, "learning_rate": 1.7583203359328136e-06, "loss": 0.0001, "step": 17299 }, { "epoch": 2.07, "learning_rate": 1.758248350329934e-06, "loss": 0.0002, "step": 17300 }, { "epoch": 2.07, "learning_rate": 1.7581763647270547e-06, "loss": 0.5132, "step": 17301 }, { "epoch": 2.07, "learning_rate": 1.7581043791241753e-06, "loss": 0.0002, "step": 17302 }, { "epoch": 2.07, "learning_rate": 1.7580323935212957e-06, "loss": 0.0003, "step": 17303 }, { "epoch": 2.07, "learning_rate": 1.7579604079184164e-06, "loss": 0.0002, "step": 17304 }, { "epoch": 2.07, "learning_rate": 1.7578884223155368e-06, "loss": 0.3547, "step": 17305 }, { "epoch": 2.07, "learning_rate": 1.7578164367126576e-06, "loss": 0.0002, "step": 17306 }, { "epoch": 2.07, "learning_rate": 1.757744451109778e-06, "loss": 0.0002, "step": 17307 }, { "epoch": 2.07, "learning_rate": 1.7576724655068987e-06, "loss": 0.0001, "step": 17308 }, { "epoch": 2.07, "learning_rate": 1.7576004799040193e-06, "loss": 0.0003, "step": 17309 }, { "epoch": 2.07, "learning_rate": 1.7575284943011398e-06, "loss": 0.0001, "step": 17310 }, { "epoch": 2.07, "learning_rate": 1.7574565086982604e-06, "loss": 0.0002, "step": 17311 }, { "epoch": 2.07, "learning_rate": 1.7573845230953812e-06, "loss": 0.0001, "step": 17312 }, { "epoch": 2.07, "learning_rate": 1.7573125374925015e-06, "loss": 0.0002, "step": 17313 }, { "epoch": 2.07, "learning_rate": 1.7572405518896219e-06, "loss": 0.0001, "step": 17314 }, { "epoch": 2.07, "learning_rate": 1.7571685662867427e-06, "loss": 0.0005, "step": 17315 }, { "epoch": 2.08, "learning_rate": 1.7570965806838632e-06, "loss": 0.0002, "step": 17316 }, { "epoch": 2.08, "learning_rate": 1.7570245950809838e-06, "loss": 0.0002, "step": 17317 }, { "epoch": 2.08, "learning_rate": 1.7569526094781044e-06, "loss": 0.0002, "step": 17318 }, { "epoch": 2.08, "learning_rate": 1.7568806238752251e-06, "loss": 0.0001, "step": 17319 }, { "epoch": 2.08, "learning_rate": 1.7568086382723455e-06, "loss": 0.0009, "step": 17320 }, { "epoch": 2.08, "learning_rate": 1.7567366526694663e-06, "loss": 0.0004, "step": 17321 }, { "epoch": 2.08, "learning_rate": 1.7566646670665866e-06, "loss": 0.0001, "step": 17322 }, { "epoch": 2.08, "learning_rate": 1.7565926814637074e-06, "loss": 0.0002, "step": 17323 }, { "epoch": 2.08, "learning_rate": 1.7565206958608278e-06, "loss": 0.0002, "step": 17324 }, { "epoch": 2.08, "learning_rate": 1.7564487102579485e-06, "loss": 0.0075, "step": 17325 }, { "epoch": 2.08, "learning_rate": 1.7563767246550691e-06, "loss": 0.0002, "step": 17326 }, { "epoch": 2.08, "learning_rate": 1.7563047390521895e-06, "loss": 0.0003, "step": 17327 }, { "epoch": 2.08, "learning_rate": 1.7562327534493102e-06, "loss": 0.0001, "step": 17328 }, { "epoch": 2.08, "learning_rate": 1.7561607678464306e-06, "loss": 0.0002, "step": 17329 }, { "epoch": 2.08, "learning_rate": 1.7560887822435514e-06, "loss": 0.0002, "step": 17330 }, { "epoch": 2.08, "learning_rate": 1.7560167966406717e-06, "loss": 0.0001, "step": 17331 }, { "epoch": 2.08, "learning_rate": 1.7559448110377925e-06, "loss": 0.0003, "step": 17332 }, { "epoch": 2.08, "learning_rate": 1.755872825434913e-06, "loss": 0.0002, "step": 17333 }, { "epoch": 2.08, "learning_rate": 1.7558008398320339e-06, "loss": 0.0013, "step": 17334 }, { "epoch": 2.08, "learning_rate": 1.7557288542291542e-06, "loss": 0.0002, "step": 17335 }, { "epoch": 2.08, "learning_rate": 1.755656868626275e-06, "loss": 0.0003, "step": 17336 }, { "epoch": 2.08, "learning_rate": 1.7555848830233953e-06, "loss": 0.0002, "step": 17337 }, { "epoch": 2.08, "learning_rate": 1.7555128974205157e-06, "loss": 0.0001, "step": 17338 }, { "epoch": 2.08, "learning_rate": 1.7554409118176365e-06, "loss": 0.0002, "step": 17339 }, { "epoch": 2.08, "learning_rate": 1.755368926214757e-06, "loss": 0.0003, "step": 17340 }, { "epoch": 2.08, "learning_rate": 1.7552969406118778e-06, "loss": 0.0002, "step": 17341 }, { "epoch": 2.08, "learning_rate": 1.7552249550089982e-06, "loss": 0.0002, "step": 17342 }, { "epoch": 2.08, "learning_rate": 1.755152969406119e-06, "loss": 0.0001, "step": 17343 }, { "epoch": 2.08, "learning_rate": 1.7550809838032393e-06, "loss": 0.0004, "step": 17344 }, { "epoch": 2.08, "learning_rate": 1.75500899820036e-06, "loss": 0.0001, "step": 17345 }, { "epoch": 2.08, "learning_rate": 1.7549370125974804e-06, "loss": 0.0003, "step": 17346 }, { "epoch": 2.08, "learning_rate": 1.7548650269946012e-06, "loss": 0.0002, "step": 17347 }, { "epoch": 2.08, "learning_rate": 1.7547930413917218e-06, "loss": 0.0005, "step": 17348 }, { "epoch": 2.08, "learning_rate": 1.7547210557888421e-06, "loss": 0.0001, "step": 17349 }, { "epoch": 2.08, "learning_rate": 1.754649070185963e-06, "loss": 0.0002, "step": 17350 }, { "epoch": 2.08, "learning_rate": 1.7545770845830833e-06, "loss": 0.0003, "step": 17351 }, { "epoch": 2.08, "learning_rate": 1.754505098980204e-06, "loss": 0.0001, "step": 17352 }, { "epoch": 2.08, "learning_rate": 1.7544331133773244e-06, "loss": 0.0002, "step": 17353 }, { "epoch": 2.08, "learning_rate": 1.7543611277744452e-06, "loss": 0.0002, "step": 17354 }, { "epoch": 2.08, "learning_rate": 1.7542891421715658e-06, "loss": 0.0002, "step": 17355 }, { "epoch": 2.08, "learning_rate": 1.7542171565686863e-06, "loss": 0.0004, "step": 17356 }, { "epoch": 2.08, "learning_rate": 1.754145170965807e-06, "loss": 0.0002, "step": 17357 }, { "epoch": 2.08, "learning_rate": 1.7540731853629277e-06, "loss": 0.0001, "step": 17358 }, { "epoch": 2.08, "learning_rate": 1.754001199760048e-06, "loss": 0.0001, "step": 17359 }, { "epoch": 2.08, "learning_rate": 1.7539292141571684e-06, "loss": 0.0002, "step": 17360 }, { "epoch": 2.08, "learning_rate": 1.7538572285542892e-06, "loss": 0.0001, "step": 17361 }, { "epoch": 2.08, "learning_rate": 1.7537852429514097e-06, "loss": 0.0002, "step": 17362 }, { "epoch": 2.08, "learning_rate": 1.7537132573485303e-06, "loss": 0.0001, "step": 17363 }, { "epoch": 2.08, "learning_rate": 1.7536412717456509e-06, "loss": 0.0002, "step": 17364 }, { "epoch": 2.08, "learning_rate": 1.7535692861427716e-06, "loss": 0.0001, "step": 17365 }, { "epoch": 2.08, "learning_rate": 1.753497300539892e-06, "loss": 0.001, "step": 17366 }, { "epoch": 2.08, "learning_rate": 1.7534253149370128e-06, "loss": 0.0001, "step": 17367 }, { "epoch": 2.08, "learning_rate": 1.7533533293341331e-06, "loss": 1.1745, "step": 17368 }, { "epoch": 2.08, "learning_rate": 1.753281343731254e-06, "loss": 0.0002, "step": 17369 }, { "epoch": 2.08, "learning_rate": 1.7532093581283743e-06, "loss": 0.0001, "step": 17370 }, { "epoch": 2.08, "learning_rate": 1.753137372525495e-06, "loss": 0.0003, "step": 17371 }, { "epoch": 2.08, "learning_rate": 1.7530653869226156e-06, "loss": 0.0003, "step": 17372 }, { "epoch": 2.08, "learning_rate": 1.752993401319736e-06, "loss": 0.0001, "step": 17373 }, { "epoch": 2.08, "learning_rate": 1.7529214157168567e-06, "loss": 0.0007, "step": 17374 }, { "epoch": 2.08, "learning_rate": 1.752849430113977e-06, "loss": 1.1306, "step": 17375 }, { "epoch": 2.08, "learning_rate": 1.7527774445110979e-06, "loss": 0.0002, "step": 17376 }, { "epoch": 2.08, "learning_rate": 1.7527054589082182e-06, "loss": 0.0001, "step": 17377 }, { "epoch": 2.08, "learning_rate": 1.752633473305339e-06, "loss": 0.0003, "step": 17378 }, { "epoch": 2.08, "learning_rate": 1.7525614877024596e-06, "loss": 0.0001, "step": 17379 }, { "epoch": 2.08, "learning_rate": 1.7524895020995804e-06, "loss": 0.0002, "step": 17380 }, { "epoch": 2.08, "learning_rate": 1.7524175164967007e-06, "loss": 0.0003, "step": 17381 }, { "epoch": 2.08, "learning_rate": 1.7523455308938215e-06, "loss": 0.0002, "step": 17382 }, { "epoch": 2.08, "learning_rate": 1.7522735452909418e-06, "loss": 0.0002, "step": 17383 }, { "epoch": 2.08, "learning_rate": 1.7522015596880622e-06, "loss": 0.0002, "step": 17384 }, { "epoch": 2.08, "learning_rate": 1.752129574085183e-06, "loss": 0.0002, "step": 17385 }, { "epoch": 2.08, "learning_rate": 1.7520575884823035e-06, "loss": 0.0005, "step": 17386 }, { "epoch": 2.08, "learning_rate": 1.7519856028794243e-06, "loss": 0.0002, "step": 17387 }, { "epoch": 2.08, "learning_rate": 1.7519136172765447e-06, "loss": 0.0002, "step": 17388 }, { "epoch": 2.08, "learning_rate": 1.7518416316736655e-06, "loss": 0.0001, "step": 17389 }, { "epoch": 2.08, "learning_rate": 1.7517696460707858e-06, "loss": 0.0002, "step": 17390 }, { "epoch": 2.08, "learning_rate": 1.7516976604679066e-06, "loss": 0.0003, "step": 17391 }, { "epoch": 2.08, "learning_rate": 1.751625674865027e-06, "loss": 0.0002, "step": 17392 }, { "epoch": 2.08, "learning_rate": 1.7515536892621477e-06, "loss": 0.0002, "step": 17393 }, { "epoch": 2.08, "learning_rate": 1.7514817036592683e-06, "loss": 0.0002, "step": 17394 }, { "epoch": 2.08, "learning_rate": 1.7514097180563886e-06, "loss": 0.0002, "step": 17395 }, { "epoch": 2.08, "learning_rate": 1.7513377324535094e-06, "loss": 0.0002, "step": 17396 }, { "epoch": 2.08, "learning_rate": 1.7512657468506298e-06, "loss": 0.0001, "step": 17397 }, { "epoch": 2.08, "learning_rate": 1.7511937612477506e-06, "loss": 0.0002, "step": 17398 }, { "epoch": 2.08, "learning_rate": 1.751121775644871e-06, "loss": 0.0001, "step": 17399 }, { "epoch": 2.09, "learning_rate": 1.7510497900419917e-06, "loss": 0.0001, "step": 17400 }, { "epoch": 2.09, "learning_rate": 1.750977804439112e-06, "loss": 0.0003, "step": 17401 }, { "epoch": 2.09, "learning_rate": 1.7509058188362328e-06, "loss": 0.0001, "step": 17402 }, { "epoch": 2.09, "learning_rate": 1.7508338332333534e-06, "loss": 0.0002, "step": 17403 }, { "epoch": 2.09, "learning_rate": 1.7507618476304742e-06, "loss": 0.0002, "step": 17404 }, { "epoch": 2.09, "learning_rate": 1.7506898620275945e-06, "loss": 0.0002, "step": 17405 }, { "epoch": 2.09, "learning_rate": 1.7506178764247149e-06, "loss": 0.0002, "step": 17406 }, { "epoch": 2.09, "learning_rate": 1.7505458908218357e-06, "loss": 0.0002, "step": 17407 }, { "epoch": 2.09, "learning_rate": 1.750473905218956e-06, "loss": 0.0002, "step": 17408 }, { "epoch": 2.09, "learning_rate": 1.7504019196160768e-06, "loss": 0.0001, "step": 17409 }, { "epoch": 2.09, "learning_rate": 1.7503299340131974e-06, "loss": 0.0002, "step": 17410 }, { "epoch": 2.09, "learning_rate": 1.7502579484103181e-06, "loss": 0.1256, "step": 17411 }, { "epoch": 2.09, "learning_rate": 1.7501859628074385e-06, "loss": 0.0002, "step": 17412 }, { "epoch": 2.09, "learning_rate": 1.7501139772045593e-06, "loss": 0.0003, "step": 17413 }, { "epoch": 2.09, "learning_rate": 1.7500419916016796e-06, "loss": 0.0003, "step": 17414 }, { "epoch": 2.09, "learning_rate": 1.7499700059988004e-06, "loss": 0.0002, "step": 17415 }, { "epoch": 2.09, "learning_rate": 1.7498980203959208e-06, "loss": 0.0003, "step": 17416 }, { "epoch": 2.09, "learning_rate": 1.7498260347930415e-06, "loss": 0.0013, "step": 17417 }, { "epoch": 2.09, "learning_rate": 1.749754049190162e-06, "loss": 0.0001, "step": 17418 }, { "epoch": 2.09, "learning_rate": 1.7496820635872825e-06, "loss": 0.0002, "step": 17419 }, { "epoch": 2.09, "learning_rate": 1.7496100779844032e-06, "loss": 0.0002, "step": 17420 }, { "epoch": 2.09, "learning_rate": 1.7495380923815236e-06, "loss": 0.4247, "step": 17421 }, { "epoch": 2.09, "learning_rate": 1.7494661067786444e-06, "loss": 0.0002, "step": 17422 }, { "epoch": 2.09, "learning_rate": 1.7493941211757647e-06, "loss": 0.0002, "step": 17423 }, { "epoch": 2.09, "learning_rate": 1.7493221355728855e-06, "loss": 0.0002, "step": 17424 }, { "epoch": 2.09, "learning_rate": 1.749250149970006e-06, "loss": 0.0002, "step": 17425 }, { "epoch": 2.09, "learning_rate": 1.7491781643671266e-06, "loss": 0.0002, "step": 17426 }, { "epoch": 2.09, "learning_rate": 1.7491061787642472e-06, "loss": 0.0001, "step": 17427 }, { "epoch": 2.09, "learning_rate": 1.749034193161368e-06, "loss": 0.0003, "step": 17428 }, { "epoch": 2.09, "learning_rate": 1.7489622075584883e-06, "loss": 0.0002, "step": 17429 }, { "epoch": 2.09, "learning_rate": 1.7488902219556087e-06, "loss": 1.02, "step": 17430 }, { "epoch": 2.09, "learning_rate": 1.7488182363527295e-06, "loss": 0.0002, "step": 17431 }, { "epoch": 2.09, "learning_rate": 1.74874625074985e-06, "loss": 0.0002, "step": 17432 }, { "epoch": 2.09, "learning_rate": 1.7486742651469706e-06, "loss": 0.0002, "step": 17433 }, { "epoch": 2.09, "learning_rate": 1.7486022795440912e-06, "loss": 0.0004, "step": 17434 }, { "epoch": 2.09, "learning_rate": 1.748530293941212e-06, "loss": 0.0115, "step": 17435 }, { "epoch": 2.09, "learning_rate": 1.7484583083383323e-06, "loss": 0.001, "step": 17436 }, { "epoch": 2.09, "learning_rate": 1.748386322735453e-06, "loss": 0.0001, "step": 17437 }, { "epoch": 2.09, "learning_rate": 1.7483143371325734e-06, "loss": 0.0002, "step": 17438 }, { "epoch": 2.09, "learning_rate": 1.7482423515296942e-06, "loss": 0.0002, "step": 17439 }, { "epoch": 2.09, "learning_rate": 1.7481703659268146e-06, "loss": 0.0001, "step": 17440 }, { "epoch": 2.09, "learning_rate": 1.7480983803239351e-06, "loss": 0.0002, "step": 17441 }, { "epoch": 2.09, "learning_rate": 1.748026394721056e-06, "loss": 0.0002, "step": 17442 }, { "epoch": 2.09, "learning_rate": 1.7479544091181763e-06, "loss": 0.0001, "step": 17443 }, { "epoch": 2.09, "learning_rate": 1.747882423515297e-06, "loss": 0.0002, "step": 17444 }, { "epoch": 2.09, "learning_rate": 1.7478104379124174e-06, "loss": 0.486, "step": 17445 }, { "epoch": 2.09, "learning_rate": 1.7477384523095382e-06, "loss": 0.0002, "step": 17446 }, { "epoch": 2.09, "learning_rate": 1.7476664667066585e-06, "loss": 0.0001, "step": 17447 }, { "epoch": 2.09, "learning_rate": 1.7475944811037793e-06, "loss": 0.0004, "step": 17448 }, { "epoch": 2.09, "learning_rate": 1.7475224955008999e-06, "loss": 0.0002, "step": 17449 }, { "epoch": 2.09, "learning_rate": 1.7474505098980207e-06, "loss": 0.0002, "step": 17450 }, { "epoch": 2.09, "learning_rate": 1.747378524295141e-06, "loss": 0.0001, "step": 17451 }, { "epoch": 2.09, "learning_rate": 1.7473065386922618e-06, "loss": 0.0003, "step": 17452 }, { "epoch": 2.09, "learning_rate": 1.7472345530893822e-06, "loss": 0.0003, "step": 17453 }, { "epoch": 2.09, "learning_rate": 1.7471625674865025e-06, "loss": 0.0001, "step": 17454 }, { "epoch": 2.09, "learning_rate": 1.7470905818836233e-06, "loss": 0.0002, "step": 17455 }, { "epoch": 2.09, "learning_rate": 1.7470185962807439e-06, "loss": 0.0086, "step": 17456 }, { "epoch": 2.09, "learning_rate": 1.7469466106778646e-06, "loss": 0.0002, "step": 17457 }, { "epoch": 2.09, "learning_rate": 1.746874625074985e-06, "loss": 0.0001, "step": 17458 }, { "epoch": 2.09, "learning_rate": 1.7468026394721058e-06, "loss": 0.0002, "step": 17459 }, { "epoch": 2.09, "learning_rate": 1.7467306538692261e-06, "loss": 0.3494, "step": 17460 }, { "epoch": 2.09, "learning_rate": 1.746658668266347e-06, "loss": 0.0002, "step": 17461 }, { "epoch": 2.09, "learning_rate": 1.7465866826634673e-06, "loss": 0.0004, "step": 17462 }, { "epoch": 2.09, "learning_rate": 1.746514697060588e-06, "loss": 0.0002, "step": 17463 }, { "epoch": 2.09, "learning_rate": 1.7464427114577086e-06, "loss": 0.0002, "step": 17464 }, { "epoch": 2.09, "learning_rate": 1.746370725854829e-06, "loss": 0.0002, "step": 17465 }, { "epoch": 2.09, "learning_rate": 1.7462987402519497e-06, "loss": 0.0084, "step": 17466 }, { "epoch": 2.09, "learning_rate": 1.74622675464907e-06, "loss": 0.0003, "step": 17467 }, { "epoch": 2.09, "learning_rate": 1.7461547690461909e-06, "loss": 0.5765, "step": 17468 }, { "epoch": 2.09, "learning_rate": 1.7460827834433112e-06, "loss": 0.0003, "step": 17469 }, { "epoch": 2.09, "learning_rate": 1.746010797840432e-06, "loss": 0.0002, "step": 17470 }, { "epoch": 2.09, "learning_rate": 1.7459388122375526e-06, "loss": 0.0001, "step": 17471 }, { "epoch": 2.09, "learning_rate": 1.7458668266346731e-06, "loss": 0.0001, "step": 17472 }, { "epoch": 2.09, "learning_rate": 1.7457948410317937e-06, "loss": 0.0003, "step": 17473 }, { "epoch": 2.09, "learning_rate": 1.7457228554289145e-06, "loss": 0.0001, "step": 17474 }, { "epoch": 2.09, "learning_rate": 1.7456508698260348e-06, "loss": 0.0001, "step": 17475 }, { "epoch": 2.09, "learning_rate": 1.7455788842231552e-06, "loss": 0.0001, "step": 17476 }, { "epoch": 2.09, "learning_rate": 1.745506898620276e-06, "loss": 0.0002, "step": 17477 }, { "epoch": 2.09, "learning_rate": 1.7454349130173965e-06, "loss": 0.0003, "step": 17478 }, { "epoch": 2.09, "learning_rate": 1.745362927414517e-06, "loss": 0.0001, "step": 17479 }, { "epoch": 2.09, "learning_rate": 1.7452909418116377e-06, "loss": 0.0002, "step": 17480 }, { "epoch": 2.09, "learning_rate": 1.7452189562087584e-06, "loss": 0.0001, "step": 17481 }, { "epoch": 2.09, "learning_rate": 1.7451469706058788e-06, "loss": 0.0003, "step": 17482 }, { "epoch": 2.1, "learning_rate": 1.7450749850029996e-06, "loss": 0.0004, "step": 17483 }, { "epoch": 2.1, "learning_rate": 1.74500299940012e-06, "loss": 0.0002, "step": 17484 }, { "epoch": 2.1, "learning_rate": 1.7449310137972407e-06, "loss": 0.0001, "step": 17485 }, { "epoch": 2.1, "learning_rate": 1.744859028194361e-06, "loss": 0.0001, "step": 17486 }, { "epoch": 2.1, "learning_rate": 1.7447870425914816e-06, "loss": 0.0001, "step": 17487 }, { "epoch": 2.1, "learning_rate": 1.7447150569886024e-06, "loss": 0.0003, "step": 17488 }, { "epoch": 2.1, "learning_rate": 1.7446430713857228e-06, "loss": 0.0001, "step": 17489 }, { "epoch": 2.1, "learning_rate": 1.7445710857828435e-06, "loss": 0.0003, "step": 17490 }, { "epoch": 2.1, "learning_rate": 1.744499100179964e-06, "loss": 0.0003, "step": 17491 }, { "epoch": 2.1, "learning_rate": 1.7444271145770847e-06, "loss": 0.0001, "step": 17492 }, { "epoch": 2.1, "learning_rate": 1.744355128974205e-06, "loss": 0.0002, "step": 17493 }, { "epoch": 2.1, "learning_rate": 1.7442831433713258e-06, "loss": 0.0001, "step": 17494 }, { "epoch": 2.1, "learning_rate": 1.7442111577684464e-06, "loss": 0.0001, "step": 17495 }, { "epoch": 2.1, "learning_rate": 1.7441391721655672e-06, "loss": 0.0001, "step": 17496 }, { "epoch": 2.1, "learning_rate": 1.7440671865626875e-06, "loss": 0.0001, "step": 17497 }, { "epoch": 2.1, "learning_rate": 1.7439952009598083e-06, "loss": 0.0001, "step": 17498 }, { "epoch": 2.1, "learning_rate": 1.7439232153569286e-06, "loss": 0.0001, "step": 17499 }, { "epoch": 2.1, "learning_rate": 1.743851229754049e-06, "loss": 0.0005, "step": 17500 }, { "epoch": 2.1, "learning_rate": 1.7437792441511698e-06, "loss": 0.0001, "step": 17501 }, { "epoch": 2.1, "learning_rate": 1.7437072585482904e-06, "loss": 0.0002, "step": 17502 }, { "epoch": 2.1, "learning_rate": 1.7436352729454111e-06, "loss": 0.0001, "step": 17503 }, { "epoch": 2.1, "learning_rate": 1.7435632873425315e-06, "loss": 0.0003, "step": 17504 }, { "epoch": 2.1, "learning_rate": 1.7434913017396523e-06, "loss": 0.0005, "step": 17505 }, { "epoch": 2.1, "learning_rate": 1.7434193161367726e-06, "loss": 0.0002, "step": 17506 }, { "epoch": 2.1, "learning_rate": 1.7433473305338934e-06, "loss": 0.0002, "step": 17507 }, { "epoch": 2.1, "learning_rate": 1.7432753449310138e-06, "loss": 0.0009, "step": 17508 }, { "epoch": 2.1, "learning_rate": 1.7432033593281345e-06, "loss": 0.0001, "step": 17509 }, { "epoch": 2.1, "learning_rate": 1.743131373725255e-06, "loss": 0.0002, "step": 17510 }, { "epoch": 2.1, "learning_rate": 1.7430593881223755e-06, "loss": 0.0003, "step": 17511 }, { "epoch": 2.1, "learning_rate": 1.7429874025194962e-06, "loss": 0.0001, "step": 17512 }, { "epoch": 2.1, "learning_rate": 1.7429154169166166e-06, "loss": 0.003, "step": 17513 }, { "epoch": 2.1, "learning_rate": 1.7428434313137374e-06, "loss": 0.0001, "step": 17514 }, { "epoch": 2.1, "learning_rate": 1.7427714457108577e-06, "loss": 0.0002, "step": 17515 }, { "epoch": 2.1, "learning_rate": 1.7426994601079785e-06, "loss": 0.0003, "step": 17516 }, { "epoch": 2.1, "learning_rate": 1.742627474505099e-06, "loss": 0.0003, "step": 17517 }, { "epoch": 2.1, "learning_rate": 1.7425554889022196e-06, "loss": 0.0002, "step": 17518 }, { "epoch": 2.1, "learning_rate": 1.7424835032993402e-06, "loss": 0.0236, "step": 17519 }, { "epoch": 2.1, "learning_rate": 1.742411517696461e-06, "loss": 0.0002, "step": 17520 }, { "epoch": 2.1, "learning_rate": 1.7423395320935813e-06, "loss": 0.8951, "step": 17521 }, { "epoch": 2.1, "learning_rate": 1.7422675464907017e-06, "loss": 0.0001, "step": 17522 }, { "epoch": 2.1, "learning_rate": 1.7421955608878225e-06, "loss": 0.0002, "step": 17523 }, { "epoch": 2.1, "learning_rate": 1.742123575284943e-06, "loss": 0.0001, "step": 17524 }, { "epoch": 2.1, "learning_rate": 1.7420515896820636e-06, "loss": 0.8483, "step": 17525 }, { "epoch": 2.1, "learning_rate": 1.7419796040791842e-06, "loss": 0.0002, "step": 17526 }, { "epoch": 2.1, "learning_rate": 1.741907618476305e-06, "loss": 0.0002, "step": 17527 }, { "epoch": 2.1, "learning_rate": 1.7418356328734253e-06, "loss": 0.0001, "step": 17528 }, { "epoch": 2.1, "learning_rate": 1.741763647270546e-06, "loss": 0.0001, "step": 17529 }, { "epoch": 2.1, "learning_rate": 1.7416916616676664e-06, "loss": 0.0001, "step": 17530 }, { "epoch": 2.1, "learning_rate": 1.7416196760647872e-06, "loss": 0.0002, "step": 17531 }, { "epoch": 2.1, "learning_rate": 1.7415476904619076e-06, "loss": 0.0001, "step": 17532 }, { "epoch": 2.1, "learning_rate": 1.7414757048590281e-06, "loss": 0.0003, "step": 17533 }, { "epoch": 2.1, "learning_rate": 1.741403719256149e-06, "loss": 0.0002, "step": 17534 }, { "epoch": 2.1, "learning_rate": 1.7413317336532693e-06, "loss": 0.0002, "step": 17535 }, { "epoch": 2.1, "learning_rate": 1.74125974805039e-06, "loss": 0.0002, "step": 17536 }, { "epoch": 2.1, "learning_rate": 1.7411877624475104e-06, "loss": 0.0004, "step": 17537 }, { "epoch": 2.1, "learning_rate": 1.7411157768446312e-06, "loss": 0.0001, "step": 17538 }, { "epoch": 2.1, "learning_rate": 1.7410437912417515e-06, "loss": 0.0001, "step": 17539 }, { "epoch": 2.1, "learning_rate": 1.7409718056388723e-06, "loss": 0.0003, "step": 17540 }, { "epoch": 2.1, "learning_rate": 1.7408998200359929e-06, "loss": 0.0002, "step": 17541 }, { "epoch": 2.1, "learning_rate": 1.7408278344331137e-06, "loss": 0.0003, "step": 17542 }, { "epoch": 2.1, "learning_rate": 1.740755848830234e-06, "loss": 0.0004, "step": 17543 }, { "epoch": 2.1, "learning_rate": 1.7406838632273548e-06, "loss": 0.0117, "step": 17544 }, { "epoch": 2.1, "learning_rate": 1.7406118776244751e-06, "loss": 0.0003, "step": 17545 }, { "epoch": 2.1, "learning_rate": 1.7405398920215955e-06, "loss": 0.0362, "step": 17546 }, { "epoch": 2.1, "learning_rate": 1.7404679064187163e-06, "loss": 0.0001, "step": 17547 }, { "epoch": 2.1, "learning_rate": 1.7403959208158368e-06, "loss": 0.0002, "step": 17548 }, { "epoch": 2.1, "learning_rate": 1.7403239352129574e-06, "loss": 0.0002, "step": 17549 }, { "epoch": 2.1, "learning_rate": 1.740251949610078e-06, "loss": 0.0004, "step": 17550 }, { "epoch": 2.1, "learning_rate": 1.7401799640071988e-06, "loss": 0.0003, "step": 17551 }, { "epoch": 2.1, "learning_rate": 1.7401079784043191e-06, "loss": 0.0002, "step": 17552 }, { "epoch": 2.1, "learning_rate": 1.7400359928014399e-06, "loss": 0.0002, "step": 17553 }, { "epoch": 2.1, "learning_rate": 1.7399640071985602e-06, "loss": 0.0002, "step": 17554 }, { "epoch": 2.1, "learning_rate": 1.739892021595681e-06, "loss": 0.0002, "step": 17555 }, { "epoch": 2.1, "learning_rate": 1.7398200359928014e-06, "loss": 0.0002, "step": 17556 }, { "epoch": 2.1, "learning_rate": 1.739748050389922e-06, "loss": 0.0002, "step": 17557 }, { "epoch": 2.1, "learning_rate": 1.7396760647870427e-06, "loss": 0.0002, "step": 17558 }, { "epoch": 2.1, "learning_rate": 1.739604079184163e-06, "loss": 0.0001, "step": 17559 }, { "epoch": 2.1, "learning_rate": 1.7395320935812839e-06, "loss": 0.0002, "step": 17560 }, { "epoch": 2.1, "learning_rate": 1.7394601079784042e-06, "loss": 0.0002, "step": 17561 }, { "epoch": 2.1, "learning_rate": 1.739388122375525e-06, "loss": 0.0003, "step": 17562 }, { "epoch": 2.1, "learning_rate": 1.7393161367726453e-06, "loss": 0.0002, "step": 17563 }, { "epoch": 2.1, "learning_rate": 1.7392441511697661e-06, "loss": 0.0002, "step": 17564 }, { "epoch": 2.1, "learning_rate": 1.7391721655668867e-06, "loss": 0.0002, "step": 17565 }, { "epoch": 2.1, "learning_rate": 1.7391001799640075e-06, "loss": 0.0002, "step": 17566 }, { "epoch": 2.11, "learning_rate": 1.7390281943611278e-06, "loss": 0.0003, "step": 17567 }, { "epoch": 2.11, "learning_rate": 1.7389562087582482e-06, "loss": 0.0003, "step": 17568 }, { "epoch": 2.11, "learning_rate": 1.738884223155369e-06, "loss": 0.0193, "step": 17569 }, { "epoch": 2.11, "learning_rate": 1.7388122375524893e-06, "loss": 0.0002, "step": 17570 }, { "epoch": 2.11, "learning_rate": 1.73874025194961e-06, "loss": 0.2532, "step": 17571 }, { "epoch": 2.11, "learning_rate": 1.7386682663467307e-06, "loss": 0.0002, "step": 17572 }, { "epoch": 2.11, "learning_rate": 1.7385962807438514e-06, "loss": 0.0001, "step": 17573 }, { "epoch": 2.11, "learning_rate": 1.7385242951409718e-06, "loss": 0.0004, "step": 17574 }, { "epoch": 2.11, "learning_rate": 1.7384523095380926e-06, "loss": 0.0001, "step": 17575 }, { "epoch": 2.11, "learning_rate": 1.738380323935213e-06, "loss": 0.0001, "step": 17576 }, { "epoch": 2.11, "learning_rate": 1.7383083383323337e-06, "loss": 0.0002, "step": 17577 }, { "epoch": 2.11, "learning_rate": 1.738236352729454e-06, "loss": 0.0002, "step": 17578 }, { "epoch": 2.11, "learning_rate": 1.7381643671265748e-06, "loss": 0.0002, "step": 17579 }, { "epoch": 2.11, "learning_rate": 1.7380923815236954e-06, "loss": 0.0002, "step": 17580 }, { "epoch": 2.11, "learning_rate": 1.7380203959208158e-06, "loss": 0.0001, "step": 17581 }, { "epoch": 2.11, "learning_rate": 1.7379484103179365e-06, "loss": 0.0002, "step": 17582 }, { "epoch": 2.11, "learning_rate": 1.737876424715057e-06, "loss": 0.0001, "step": 17583 }, { "epoch": 2.11, "learning_rate": 1.7378044391121777e-06, "loss": 0.0003, "step": 17584 }, { "epoch": 2.11, "learning_rate": 1.737732453509298e-06, "loss": 0.0001, "step": 17585 }, { "epoch": 2.11, "learning_rate": 1.7376604679064188e-06, "loss": 0.0001, "step": 17586 }, { "epoch": 2.11, "learning_rate": 1.7375884823035394e-06, "loss": 0.6825, "step": 17587 }, { "epoch": 2.11, "learning_rate": 1.73751649670066e-06, "loss": 0.0001, "step": 17588 }, { "epoch": 2.11, "learning_rate": 1.7374445110977805e-06, "loss": 0.5062, "step": 17589 }, { "epoch": 2.11, "learning_rate": 1.7373725254949013e-06, "loss": 0.0148, "step": 17590 }, { "epoch": 2.11, "learning_rate": 1.7373005398920216e-06, "loss": 0.0003, "step": 17591 }, { "epoch": 2.11, "learning_rate": 1.737228554289142e-06, "loss": 0.0002, "step": 17592 }, { "epoch": 2.11, "learning_rate": 1.7371565686862628e-06, "loss": 0.0001, "step": 17593 }, { "epoch": 2.11, "learning_rate": 1.7370845830833833e-06, "loss": 0.0002, "step": 17594 }, { "epoch": 2.11, "learning_rate": 1.737012597480504e-06, "loss": 0.0001, "step": 17595 }, { "epoch": 2.11, "learning_rate": 1.7369406118776245e-06, "loss": 0.0002, "step": 17596 }, { "epoch": 2.11, "learning_rate": 1.7368686262747453e-06, "loss": 0.0002, "step": 17597 }, { "epoch": 2.11, "learning_rate": 1.7367966406718656e-06, "loss": 0.0003, "step": 17598 }, { "epoch": 2.11, "learning_rate": 1.7367246550689864e-06, "loss": 0.0002, "step": 17599 }, { "epoch": 2.11, "learning_rate": 1.7366526694661067e-06, "loss": 0.0004, "step": 17600 }, { "epoch": 2.11, "learning_rate": 1.7365806838632275e-06, "loss": 0.0005, "step": 17601 }, { "epoch": 2.11, "learning_rate": 1.7365086982603479e-06, "loss": 0.0002, "step": 17602 }, { "epoch": 2.11, "learning_rate": 1.7364367126574684e-06, "loss": 0.0002, "step": 17603 }, { "epoch": 2.11, "learning_rate": 1.7363647270545892e-06, "loss": 0.0002, "step": 17604 }, { "epoch": 2.11, "learning_rate": 1.7362927414517096e-06, "loss": 0.0009, "step": 17605 }, { "epoch": 2.11, "learning_rate": 1.7362207558488304e-06, "loss": 0.0002, "step": 17606 }, { "epoch": 2.11, "learning_rate": 1.7361487702459507e-06, "loss": 0.0001, "step": 17607 }, { "epoch": 2.11, "learning_rate": 1.7360767846430715e-06, "loss": 0.0002, "step": 17608 }, { "epoch": 2.11, "learning_rate": 1.7360047990401918e-06, "loss": 0.0003, "step": 17609 }, { "epoch": 2.11, "learning_rate": 1.7359328134373126e-06, "loss": 0.0004, "step": 17610 }, { "epoch": 2.11, "learning_rate": 1.7358608278344332e-06, "loss": 0.0003, "step": 17611 }, { "epoch": 2.11, "learning_rate": 1.735788842231554e-06, "loss": 0.0005, "step": 17612 }, { "epoch": 2.11, "learning_rate": 1.7357168566286743e-06, "loss": 0.0002, "step": 17613 }, { "epoch": 2.11, "learning_rate": 1.7356448710257947e-06, "loss": 0.0004, "step": 17614 }, { "epoch": 2.11, "learning_rate": 1.7355728854229155e-06, "loss": 0.0001, "step": 17615 }, { "epoch": 2.11, "learning_rate": 1.7355008998200358e-06, "loss": 0.0002, "step": 17616 }, { "epoch": 2.11, "learning_rate": 1.7354289142171566e-06, "loss": 0.0002, "step": 17617 }, { "epoch": 2.11, "learning_rate": 1.7353569286142772e-06, "loss": 0.0018, "step": 17618 }, { "epoch": 2.11, "learning_rate": 1.735284943011398e-06, "loss": 0.0002, "step": 17619 }, { "epoch": 2.11, "learning_rate": 1.7352129574085183e-06, "loss": 0.0001, "step": 17620 }, { "epoch": 2.11, "learning_rate": 1.735140971805639e-06, "loss": 0.0001, "step": 17621 }, { "epoch": 2.11, "learning_rate": 1.7350689862027594e-06, "loss": 0.0001, "step": 17622 }, { "epoch": 2.11, "learning_rate": 1.7349970005998802e-06, "loss": 0.0001, "step": 17623 }, { "epoch": 2.11, "learning_rate": 1.7349250149970006e-06, "loss": 0.0001, "step": 17624 }, { "epoch": 2.11, "learning_rate": 1.7348530293941213e-06, "loss": 0.0001, "step": 17625 }, { "epoch": 2.11, "learning_rate": 1.734781043791242e-06, "loss": 0.0002, "step": 17626 }, { "epoch": 2.11, "learning_rate": 1.7347090581883623e-06, "loss": 0.0003, "step": 17627 }, { "epoch": 2.11, "learning_rate": 1.734637072585483e-06, "loss": 0.0002, "step": 17628 }, { "epoch": 2.11, "learning_rate": 1.7345650869826034e-06, "loss": 0.0001, "step": 17629 }, { "epoch": 2.11, "learning_rate": 1.7344931013797242e-06, "loss": 0.0001, "step": 17630 }, { "epoch": 2.11, "learning_rate": 1.7344211157768445e-06, "loss": 0.0013, "step": 17631 }, { "epoch": 2.11, "learning_rate": 1.7343491301739653e-06, "loss": 0.0001, "step": 17632 }, { "epoch": 2.11, "learning_rate": 1.7342771445710859e-06, "loss": 0.0001, "step": 17633 }, { "epoch": 2.11, "learning_rate": 1.7342051589682064e-06, "loss": 0.0011, "step": 17634 }, { "epoch": 2.11, "learning_rate": 1.734133173365327e-06, "loss": 0.0002, "step": 17635 }, { "epoch": 2.11, "learning_rate": 1.7340611877624478e-06, "loss": 0.0002, "step": 17636 }, { "epoch": 2.11, "learning_rate": 1.7339892021595681e-06, "loss": 0.0002, "step": 17637 }, { "epoch": 2.11, "learning_rate": 1.7339172165566885e-06, "loss": 0.0002, "step": 17638 }, { "epoch": 2.11, "learning_rate": 1.7338452309538093e-06, "loss": 0.0002, "step": 17639 }, { "epoch": 2.11, "learning_rate": 1.7337732453509298e-06, "loss": 0.2649, "step": 17640 }, { "epoch": 2.11, "learning_rate": 1.7337012597480504e-06, "loss": 0.0001, "step": 17641 }, { "epoch": 2.11, "learning_rate": 1.733629274145171e-06, "loss": 0.0002, "step": 17642 }, { "epoch": 2.11, "learning_rate": 1.7335572885422918e-06, "loss": 0.0002, "step": 17643 }, { "epoch": 2.11, "learning_rate": 1.733485302939412e-06, "loss": 0.0002, "step": 17644 }, { "epoch": 2.11, "learning_rate": 1.7334133173365329e-06, "loss": 0.0003, "step": 17645 }, { "epoch": 2.11, "learning_rate": 1.7333413317336532e-06, "loss": 0.0003, "step": 17646 }, { "epoch": 2.11, "learning_rate": 1.733269346130774e-06, "loss": 0.0001, "step": 17647 }, { "epoch": 2.11, "learning_rate": 1.7331973605278944e-06, "loss": 0.0002, "step": 17648 }, { "epoch": 2.11, "learning_rate": 1.733125374925015e-06, "loss": 0.0001, "step": 17649 }, { "epoch": 2.12, "learning_rate": 1.7330533893221357e-06, "loss": 0.0002, "step": 17650 }, { "epoch": 2.12, "learning_rate": 1.732981403719256e-06, "loss": 0.0001, "step": 17651 }, { "epoch": 2.12, "learning_rate": 1.7329094181163769e-06, "loss": 0.0003, "step": 17652 }, { "epoch": 2.12, "learning_rate": 1.7328374325134972e-06, "loss": 0.0957, "step": 17653 }, { "epoch": 2.12, "learning_rate": 1.732765446910618e-06, "loss": 0.0001, "step": 17654 }, { "epoch": 2.12, "learning_rate": 1.7326934613077383e-06, "loss": 0.0002, "step": 17655 }, { "epoch": 2.12, "learning_rate": 1.7326214757048591e-06, "loss": 0.0006, "step": 17656 }, { "epoch": 2.12, "learning_rate": 1.7325494901019797e-06, "loss": 0.0005, "step": 17657 }, { "epoch": 2.12, "learning_rate": 1.7324775044991005e-06, "loss": 0.0001, "step": 17658 }, { "epoch": 2.12, "learning_rate": 1.7324055188962208e-06, "loss": 0.0001, "step": 17659 }, { "epoch": 2.12, "learning_rate": 1.7323335332933412e-06, "loss": 0.0002, "step": 17660 }, { "epoch": 2.12, "learning_rate": 1.732261547690462e-06, "loss": 0.0002, "step": 17661 }, { "epoch": 2.12, "learning_rate": 1.7321895620875823e-06, "loss": 0.0003, "step": 17662 }, { "epoch": 2.12, "learning_rate": 1.732117576484703e-06, "loss": 0.0002, "step": 17663 }, { "epoch": 2.12, "learning_rate": 1.7320455908818237e-06, "loss": 0.0002, "step": 17664 }, { "epoch": 2.12, "learning_rate": 1.7319736052789444e-06, "loss": 0.0007, "step": 17665 }, { "epoch": 2.12, "learning_rate": 1.7319016196760648e-06, "loss": 0.0002, "step": 17666 }, { "epoch": 2.12, "learning_rate": 1.7318296340731856e-06, "loss": 0.0002, "step": 17667 }, { "epoch": 2.12, "learning_rate": 1.731757648470306e-06, "loss": 0.0002, "step": 17668 }, { "epoch": 2.12, "learning_rate": 1.7316856628674267e-06, "loss": 0.0002, "step": 17669 }, { "epoch": 2.12, "learning_rate": 1.731613677264547e-06, "loss": 0.0001, "step": 17670 }, { "epoch": 2.12, "learning_rate": 1.7315416916616678e-06, "loss": 0.0001, "step": 17671 }, { "epoch": 2.12, "learning_rate": 1.7314697060587882e-06, "loss": 0.0001, "step": 17672 }, { "epoch": 2.12, "learning_rate": 1.7313977204559088e-06, "loss": 0.9255, "step": 17673 }, { "epoch": 2.12, "learning_rate": 1.7313257348530295e-06, "loss": 0.0002, "step": 17674 }, { "epoch": 2.12, "learning_rate": 1.7312537492501499e-06, "loss": 0.0003, "step": 17675 }, { "epoch": 2.12, "learning_rate": 1.7311817636472707e-06, "loss": 0.0003, "step": 17676 }, { "epoch": 2.12, "learning_rate": 1.731109778044391e-06, "loss": 0.0002, "step": 17677 }, { "epoch": 2.12, "learning_rate": 1.7310377924415118e-06, "loss": 0.0002, "step": 17678 }, { "epoch": 2.12, "learning_rate": 1.7309658068386322e-06, "loss": 0.0002, "step": 17679 }, { "epoch": 2.12, "learning_rate": 1.730893821235753e-06, "loss": 0.0001, "step": 17680 }, { "epoch": 2.12, "learning_rate": 1.7308218356328735e-06, "loss": 0.0003, "step": 17681 }, { "epoch": 2.12, "learning_rate": 1.7307498500299943e-06, "loss": 0.0001, "step": 17682 }, { "epoch": 2.12, "learning_rate": 1.7306778644271146e-06, "loss": 0.0002, "step": 17683 }, { "epoch": 2.12, "learning_rate": 1.730605878824235e-06, "loss": 0.0002, "step": 17684 }, { "epoch": 2.12, "learning_rate": 1.7305338932213558e-06, "loss": 0.0006, "step": 17685 }, { "epoch": 2.12, "learning_rate": 1.7304619076184761e-06, "loss": 0.0002, "step": 17686 }, { "epoch": 2.12, "learning_rate": 1.730389922015597e-06, "loss": 0.0001, "step": 17687 }, { "epoch": 2.12, "learning_rate": 1.7303179364127175e-06, "loss": 0.0001, "step": 17688 }, { "epoch": 2.12, "learning_rate": 1.7302459508098382e-06, "loss": 0.0004, "step": 17689 }, { "epoch": 2.12, "learning_rate": 1.7301739652069586e-06, "loss": 0.0001, "step": 17690 }, { "epoch": 2.12, "learning_rate": 1.7301019796040794e-06, "loss": 0.0002, "step": 17691 }, { "epoch": 2.12, "learning_rate": 1.7300299940011997e-06, "loss": 0.0008, "step": 17692 }, { "epoch": 2.12, "learning_rate": 1.7299580083983205e-06, "loss": 0.0001, "step": 17693 }, { "epoch": 2.12, "learning_rate": 1.7298860227954409e-06, "loss": 0.0002, "step": 17694 }, { "epoch": 2.12, "learning_rate": 1.7298140371925614e-06, "loss": 0.0003, "step": 17695 }, { "epoch": 2.12, "learning_rate": 1.7297420515896822e-06, "loss": 0.0006, "step": 17696 }, { "epoch": 2.12, "learning_rate": 1.7296700659868026e-06, "loss": 0.0024, "step": 17697 }, { "epoch": 2.12, "learning_rate": 1.7295980803839233e-06, "loss": 0.0001, "step": 17698 }, { "epoch": 2.12, "learning_rate": 1.7295260947810437e-06, "loss": 0.0001, "step": 17699 }, { "epoch": 2.12, "learning_rate": 1.7294541091781645e-06, "loss": 0.0002, "step": 17700 }, { "epoch": 2.12, "learning_rate": 1.7293821235752848e-06, "loss": 0.0001, "step": 17701 }, { "epoch": 2.12, "learning_rate": 1.7293101379724056e-06, "loss": 0.0001, "step": 17702 }, { "epoch": 2.12, "learning_rate": 1.7292381523695262e-06, "loss": 0.0001, "step": 17703 }, { "epoch": 2.12, "learning_rate": 1.7291661667666467e-06, "loss": 0.0002, "step": 17704 }, { "epoch": 2.12, "learning_rate": 1.7290941811637673e-06, "loss": 0.0002, "step": 17705 }, { "epoch": 2.12, "learning_rate": 1.729022195560888e-06, "loss": 0.0002, "step": 17706 }, { "epoch": 2.12, "learning_rate": 1.7289502099580084e-06, "loss": 0.0003, "step": 17707 }, { "epoch": 2.12, "learning_rate": 1.7288782243551288e-06, "loss": 0.0001, "step": 17708 }, { "epoch": 2.12, "learning_rate": 1.7288062387522496e-06, "loss": 0.0001, "step": 17709 }, { "epoch": 2.12, "learning_rate": 1.7287342531493701e-06, "loss": 0.0002, "step": 17710 }, { "epoch": 2.12, "learning_rate": 1.7286622675464907e-06, "loss": 0.0074, "step": 17711 }, { "epoch": 2.12, "learning_rate": 1.7285902819436113e-06, "loss": 0.0095, "step": 17712 }, { "epoch": 2.12, "learning_rate": 1.728518296340732e-06, "loss": 0.0002, "step": 17713 }, { "epoch": 2.12, "learning_rate": 1.7284463107378524e-06, "loss": 0.0002, "step": 17714 }, { "epoch": 2.12, "learning_rate": 1.7283743251349732e-06, "loss": 0.0002, "step": 17715 }, { "epoch": 2.12, "learning_rate": 1.7283023395320935e-06, "loss": 0.0001, "step": 17716 }, { "epoch": 2.12, "learning_rate": 1.7282303539292143e-06, "loss": 0.0003, "step": 17717 }, { "epoch": 2.12, "learning_rate": 1.7281583683263347e-06, "loss": 0.0001, "step": 17718 }, { "epoch": 2.12, "learning_rate": 1.7280863827234553e-06, "loss": 0.0003, "step": 17719 }, { "epoch": 2.12, "learning_rate": 1.728014397120576e-06, "loss": 0.0004, "step": 17720 }, { "epoch": 2.12, "learning_rate": 1.7279424115176964e-06, "loss": 0.0002, "step": 17721 }, { "epoch": 2.12, "learning_rate": 1.7278704259148172e-06, "loss": 0.0001, "step": 17722 }, { "epoch": 2.12, "learning_rate": 1.7277984403119375e-06, "loss": 0.0035, "step": 17723 }, { "epoch": 2.12, "learning_rate": 1.7277264547090583e-06, "loss": 0.0002, "step": 17724 }, { "epoch": 2.12, "learning_rate": 1.7276544691061787e-06, "loss": 0.0004, "step": 17725 }, { "epoch": 2.12, "learning_rate": 1.7275824835032994e-06, "loss": 0.0002, "step": 17726 }, { "epoch": 2.12, "learning_rate": 1.72751049790042e-06, "loss": 0.0001, "step": 17727 }, { "epoch": 2.12, "learning_rate": 1.7274385122975408e-06, "loss": 0.0001, "step": 17728 }, { "epoch": 2.12, "learning_rate": 1.7273665266946611e-06, "loss": 0.0004, "step": 17729 }, { "epoch": 2.12, "learning_rate": 1.7272945410917815e-06, "loss": 0.0002, "step": 17730 }, { "epoch": 2.12, "learning_rate": 1.7272225554889023e-06, "loss": 0.0009, "step": 17731 }, { "epoch": 2.12, "learning_rate": 1.7271505698860226e-06, "loss": 0.0001, "step": 17732 }, { "epoch": 2.12, "learning_rate": 1.7270785842831434e-06, "loss": 0.0001, "step": 17733 }, { "epoch": 2.13, "learning_rate": 1.727006598680264e-06, "loss": 0.0002, "step": 17734 }, { "epoch": 2.13, "learning_rate": 1.7269346130773847e-06, "loss": 0.0003, "step": 17735 }, { "epoch": 2.13, "learning_rate": 1.726862627474505e-06, "loss": 0.0003, "step": 17736 }, { "epoch": 2.13, "learning_rate": 1.7267906418716259e-06, "loss": 0.0001, "step": 17737 }, { "epoch": 2.13, "learning_rate": 1.7267186562687462e-06, "loss": 0.0001, "step": 17738 }, { "epoch": 2.13, "learning_rate": 1.726646670665867e-06, "loss": 0.0003, "step": 17739 }, { "epoch": 2.13, "learning_rate": 1.7265746850629874e-06, "loss": 0.0002, "step": 17740 }, { "epoch": 2.13, "learning_rate": 1.726502699460108e-06, "loss": 0.0001, "step": 17741 }, { "epoch": 2.13, "learning_rate": 1.7264307138572287e-06, "loss": 0.0001, "step": 17742 }, { "epoch": 2.13, "learning_rate": 1.726358728254349e-06, "loss": 0.4303, "step": 17743 }, { "epoch": 2.13, "learning_rate": 1.7262867426514698e-06, "loss": 0.0013, "step": 17744 }, { "epoch": 2.13, "learning_rate": 1.7262147570485902e-06, "loss": 0.0001, "step": 17745 }, { "epoch": 2.13, "learning_rate": 1.726142771445711e-06, "loss": 0.0002, "step": 17746 }, { "epoch": 2.13, "learning_rate": 1.7260707858428313e-06, "loss": 0.0001, "step": 17747 }, { "epoch": 2.13, "learning_rate": 1.7259988002399521e-06, "loss": 0.0002, "step": 17748 }, { "epoch": 2.13, "learning_rate": 1.7259268146370727e-06, "loss": 0.0003, "step": 17749 }, { "epoch": 2.13, "learning_rate": 1.7258548290341932e-06, "loss": 0.0001, "step": 17750 }, { "epoch": 2.13, "learning_rate": 1.7257828434313138e-06, "loss": 0.0002, "step": 17751 }, { "epoch": 2.13, "learning_rate": 1.7257108578284346e-06, "loss": 0.0002, "step": 17752 }, { "epoch": 2.13, "learning_rate": 1.725638872225555e-06, "loss": 0.0001, "step": 17753 }, { "epoch": 2.13, "learning_rate": 1.7255668866226753e-06, "loss": 0.0002, "step": 17754 }, { "epoch": 2.13, "learning_rate": 1.725494901019796e-06, "loss": 0.0001, "step": 17755 }, { "epoch": 2.13, "learning_rate": 1.7254229154169166e-06, "loss": 0.0001, "step": 17756 }, { "epoch": 2.13, "learning_rate": 1.7253509298140372e-06, "loss": 0.0001, "step": 17757 }, { "epoch": 2.13, "learning_rate": 1.7252789442111578e-06, "loss": 0.0002, "step": 17758 }, { "epoch": 2.13, "learning_rate": 1.7252069586082786e-06, "loss": 0.0039, "step": 17759 }, { "epoch": 2.13, "learning_rate": 1.725134973005399e-06, "loss": 0.0002, "step": 17760 }, { "epoch": 2.13, "learning_rate": 1.7250629874025197e-06, "loss": 0.0002, "step": 17761 }, { "epoch": 2.13, "learning_rate": 1.72499100179964e-06, "loss": 0.0001, "step": 17762 }, { "epoch": 2.13, "learning_rate": 1.7249190161967608e-06, "loss": 0.0263, "step": 17763 }, { "epoch": 2.13, "learning_rate": 1.7248470305938812e-06, "loss": 0.0001, "step": 17764 }, { "epoch": 2.13, "learning_rate": 1.7247750449910017e-06, "loss": 0.0012, "step": 17765 }, { "epoch": 2.13, "learning_rate": 1.7247030593881225e-06, "loss": 0.0002, "step": 17766 }, { "epoch": 2.13, "learning_rate": 1.7246310737852429e-06, "loss": 0.0001, "step": 17767 }, { "epoch": 2.13, "learning_rate": 1.7245590881823637e-06, "loss": 0.0001, "step": 17768 }, { "epoch": 2.13, "learning_rate": 1.724487102579484e-06, "loss": 0.0005, "step": 17769 }, { "epoch": 2.13, "learning_rate": 1.7244151169766048e-06, "loss": 0.0003, "step": 17770 }, { "epoch": 2.13, "learning_rate": 1.7243431313737251e-06, "loss": 0.0002, "step": 17771 }, { "epoch": 2.13, "learning_rate": 1.724271145770846e-06, "loss": 0.0002, "step": 17772 }, { "epoch": 2.13, "learning_rate": 1.7241991601679665e-06, "loss": 0.0001, "step": 17773 }, { "epoch": 2.13, "learning_rate": 1.7241271745650873e-06, "loss": 0.0001, "step": 17774 }, { "epoch": 2.13, "learning_rate": 1.7240551889622076e-06, "loss": 0.0002, "step": 17775 }, { "epoch": 2.13, "learning_rate": 1.723983203359328e-06, "loss": 0.0002, "step": 17776 }, { "epoch": 2.13, "learning_rate": 1.7239112177564488e-06, "loss": 0.0001, "step": 17777 }, { "epoch": 2.13, "learning_rate": 1.7238392321535691e-06, "loss": 0.0001, "step": 17778 }, { "epoch": 2.13, "learning_rate": 1.7237672465506899e-06, "loss": 0.0001, "step": 17779 }, { "epoch": 2.13, "learning_rate": 1.7236952609478105e-06, "loss": 0.0002, "step": 17780 }, { "epoch": 2.13, "learning_rate": 1.7236232753449312e-06, "loss": 0.0001, "step": 17781 }, { "epoch": 2.13, "learning_rate": 1.7235512897420516e-06, "loss": 0.0001, "step": 17782 }, { "epoch": 2.13, "learning_rate": 1.7234793041391724e-06, "loss": 0.0002, "step": 17783 }, { "epoch": 2.13, "learning_rate": 1.7234073185362927e-06, "loss": 0.0002, "step": 17784 }, { "epoch": 2.13, "learning_rate": 1.7233353329334135e-06, "loss": 0.0002, "step": 17785 }, { "epoch": 2.13, "learning_rate": 1.7232633473305339e-06, "loss": 0.0001, "step": 17786 }, { "epoch": 2.13, "learning_rate": 1.7231913617276544e-06, "loss": 0.0002, "step": 17787 }, { "epoch": 2.13, "learning_rate": 1.7231193761247752e-06, "loss": 0.0001, "step": 17788 }, { "epoch": 2.13, "learning_rate": 1.7230473905218956e-06, "loss": 0.0001, "step": 17789 }, { "epoch": 2.13, "learning_rate": 1.7229754049190163e-06, "loss": 0.0001, "step": 17790 }, { "epoch": 2.13, "learning_rate": 1.7229034193161367e-06, "loss": 0.0001, "step": 17791 }, { "epoch": 2.13, "learning_rate": 1.7228314337132575e-06, "loss": 0.0001, "step": 17792 }, { "epoch": 2.13, "learning_rate": 1.7227594481103778e-06, "loss": 0.0001, "step": 17793 }, { "epoch": 2.13, "learning_rate": 1.7226874625074986e-06, "loss": 0.0002, "step": 17794 }, { "epoch": 2.13, "learning_rate": 1.7226154769046192e-06, "loss": 0.0003, "step": 17795 }, { "epoch": 2.13, "learning_rate": 1.7225434913017397e-06, "loss": 0.0002, "step": 17796 }, { "epoch": 2.13, "learning_rate": 1.7224715056988603e-06, "loss": 0.0002, "step": 17797 }, { "epoch": 2.13, "learning_rate": 1.722399520095981e-06, "loss": 0.0001, "step": 17798 }, { "epoch": 2.13, "learning_rate": 1.7223275344931014e-06, "loss": 0.0001, "step": 17799 }, { "epoch": 2.13, "learning_rate": 1.7222555488902218e-06, "loss": 0.0001, "step": 17800 }, { "epoch": 2.13, "learning_rate": 1.7221835632873426e-06, "loss": 0.0002, "step": 17801 }, { "epoch": 2.13, "learning_rate": 1.722111577684463e-06, "loss": 0.0001, "step": 17802 }, { "epoch": 2.13, "learning_rate": 1.7220395920815837e-06, "loss": 0.0003, "step": 17803 }, { "epoch": 2.13, "learning_rate": 1.7219676064787043e-06, "loss": 0.0002, "step": 17804 }, { "epoch": 2.13, "learning_rate": 1.721895620875825e-06, "loss": 0.0002, "step": 17805 }, { "epoch": 2.13, "learning_rate": 1.7218236352729454e-06, "loss": 0.0001, "step": 17806 }, { "epoch": 2.13, "learning_rate": 1.7217516496700662e-06, "loss": 0.0001, "step": 17807 }, { "epoch": 2.13, "learning_rate": 1.7216796640671865e-06, "loss": 1.0166, "step": 17808 }, { "epoch": 2.13, "learning_rate": 1.7216076784643073e-06, "loss": 0.0018, "step": 17809 }, { "epoch": 2.13, "learning_rate": 1.7215356928614277e-06, "loss": 0.0001, "step": 17810 }, { "epoch": 2.13, "learning_rate": 1.7214637072585482e-06, "loss": 0.0001, "step": 17811 }, { "epoch": 2.13, "learning_rate": 1.721391721655669e-06, "loss": 0.0004, "step": 17812 }, { "epoch": 2.13, "learning_rate": 1.7213197360527894e-06, "loss": 0.0001, "step": 17813 }, { "epoch": 2.13, "learning_rate": 1.7212477504499102e-06, "loss": 0.0001, "step": 17814 }, { "epoch": 2.13, "learning_rate": 1.7211757648470305e-06, "loss": 0.0001, "step": 17815 }, { "epoch": 2.13, "learning_rate": 1.7211037792441513e-06, "loss": 0.0001, "step": 17816 }, { "epoch": 2.14, "learning_rate": 1.7210317936412716e-06, "loss": 0.0001, "step": 17817 }, { "epoch": 2.14, "learning_rate": 1.7209598080383924e-06, "loss": 0.0003, "step": 17818 }, { "epoch": 2.14, "learning_rate": 1.720887822435513e-06, "loss": 0.0001, "step": 17819 }, { "epoch": 2.14, "learning_rate": 1.7208158368326336e-06, "loss": 0.0001, "step": 17820 }, { "epoch": 2.14, "learning_rate": 1.7207438512297541e-06, "loss": 0.0002, "step": 17821 }, { "epoch": 2.14, "learning_rate": 1.7206718656268745e-06, "loss": 0.0002, "step": 17822 }, { "epoch": 2.14, "learning_rate": 1.7205998800239953e-06, "loss": 0.0002, "step": 17823 }, { "epoch": 2.14, "learning_rate": 1.7205278944211156e-06, "loss": 0.0001, "step": 17824 }, { "epoch": 2.14, "learning_rate": 1.7204559088182364e-06, "loss": 0.0001, "step": 17825 }, { "epoch": 2.14, "learning_rate": 1.720383923215357e-06, "loss": 0.0002, "step": 17826 }, { "epoch": 2.14, "learning_rate": 1.7203119376124775e-06, "loss": 0.0001, "step": 17827 }, { "epoch": 2.14, "learning_rate": 1.720239952009598e-06, "loss": 0.0004, "step": 17828 }, { "epoch": 2.14, "learning_rate": 1.7201679664067189e-06, "loss": 0.0002, "step": 17829 }, { "epoch": 2.14, "learning_rate": 1.7200959808038392e-06, "loss": 0.0001, "step": 17830 }, { "epoch": 2.14, "learning_rate": 1.72002399520096e-06, "loss": 0.0001, "step": 17831 }, { "epoch": 2.14, "learning_rate": 1.7199520095980804e-06, "loss": 0.0002, "step": 17832 }, { "epoch": 2.14, "learning_rate": 1.7198800239952011e-06, "loss": 0.0003, "step": 17833 }, { "epoch": 2.14, "learning_rate": 1.7198080383923215e-06, "loss": 0.0002, "step": 17834 }, { "epoch": 2.14, "learning_rate": 1.719736052789442e-06, "loss": 0.0001, "step": 17835 }, { "epoch": 2.14, "learning_rate": 1.7196640671865628e-06, "loss": 0.0001, "step": 17836 }, { "epoch": 2.14, "learning_rate": 1.7195920815836832e-06, "loss": 0.0004, "step": 17837 }, { "epoch": 2.14, "learning_rate": 1.719520095980804e-06, "loss": 0.0001, "step": 17838 }, { "epoch": 2.14, "learning_rate": 1.7194481103779243e-06, "loss": 0.0005, "step": 17839 }, { "epoch": 2.14, "learning_rate": 1.719376124775045e-06, "loss": 0.0003, "step": 17840 }, { "epoch": 2.14, "learning_rate": 1.7193041391721655e-06, "loss": 0.0003, "step": 17841 }, { "epoch": 2.14, "learning_rate": 1.7192321535692862e-06, "loss": 0.0002, "step": 17842 }, { "epoch": 2.14, "learning_rate": 1.7191601679664068e-06, "loss": 0.1281, "step": 17843 }, { "epoch": 2.14, "learning_rate": 1.7190881823635276e-06, "loss": 0.0004, "step": 17844 }, { "epoch": 2.14, "learning_rate": 1.719016196760648e-06, "loss": 0.0001, "step": 17845 }, { "epoch": 2.14, "learning_rate": 1.7189442111577683e-06, "loss": 0.0001, "step": 17846 }, { "epoch": 2.14, "learning_rate": 1.718872225554889e-06, "loss": 0.0001, "step": 17847 }, { "epoch": 2.14, "learning_rate": 1.7188002399520094e-06, "loss": 0.0002, "step": 17848 }, { "epoch": 2.14, "learning_rate": 1.7187282543491302e-06, "loss": 0.0002, "step": 17849 }, { "epoch": 2.14, "learning_rate": 1.7186562687462508e-06, "loss": 0.2229, "step": 17850 }, { "epoch": 2.14, "learning_rate": 1.7185842831433715e-06, "loss": 0.0001, "step": 17851 }, { "epoch": 2.14, "learning_rate": 1.718512297540492e-06, "loss": 0.0002, "step": 17852 }, { "epoch": 2.14, "learning_rate": 1.7184403119376127e-06, "loss": 0.0001, "step": 17853 }, { "epoch": 2.14, "learning_rate": 1.718368326334733e-06, "loss": 0.0001, "step": 17854 }, { "epoch": 2.14, "learning_rate": 1.7182963407318538e-06, "loss": 0.0002, "step": 17855 }, { "epoch": 2.14, "learning_rate": 1.7182243551289742e-06, "loss": 0.0001, "step": 17856 }, { "epoch": 2.14, "learning_rate": 1.7181523695260947e-06, "loss": 0.0001, "step": 17857 }, { "epoch": 2.14, "learning_rate": 1.7180803839232155e-06, "loss": 0.0001, "step": 17858 }, { "epoch": 2.14, "learning_rate": 1.7180083983203359e-06, "loss": 0.0001, "step": 17859 }, { "epoch": 2.14, "learning_rate": 1.7179364127174567e-06, "loss": 0.0013, "step": 17860 }, { "epoch": 2.14, "learning_rate": 1.717864427114577e-06, "loss": 0.2698, "step": 17861 }, { "epoch": 2.14, "learning_rate": 1.7177924415116978e-06, "loss": 0.0001, "step": 17862 }, { "epoch": 2.14, "learning_rate": 1.7177204559088181e-06, "loss": 0.0002, "step": 17863 }, { "epoch": 2.14, "learning_rate": 1.717648470305939e-06, "loss": 0.0001, "step": 17864 }, { "epoch": 2.14, "learning_rate": 1.7175764847030595e-06, "loss": 0.0002, "step": 17865 }, { "epoch": 2.14, "learning_rate": 1.71750449910018e-06, "loss": 0.0001, "step": 17866 }, { "epoch": 2.14, "learning_rate": 1.7174325134973006e-06, "loss": 0.0002, "step": 17867 }, { "epoch": 2.14, "learning_rate": 1.717360527894421e-06, "loss": 0.9496, "step": 17868 }, { "epoch": 2.14, "learning_rate": 1.7172885422915418e-06, "loss": 0.0003, "step": 17869 }, { "epoch": 2.14, "learning_rate": 1.7172165566886621e-06, "loss": 0.0008, "step": 17870 }, { "epoch": 2.14, "learning_rate": 1.7171445710857829e-06, "loss": 0.0004, "step": 17871 }, { "epoch": 2.14, "learning_rate": 1.7170725854829035e-06, "loss": 0.0003, "step": 17872 }, { "epoch": 2.14, "learning_rate": 1.717000599880024e-06, "loss": 0.0003, "step": 17873 }, { "epoch": 2.14, "learning_rate": 1.7169286142771446e-06, "loss": 0.0002, "step": 17874 }, { "epoch": 2.14, "learning_rate": 1.7168566286742654e-06, "loss": 0.0005, "step": 17875 }, { "epoch": 2.14, "learning_rate": 1.7167846430713857e-06, "loss": 0.0004, "step": 17876 }, { "epoch": 2.14, "learning_rate": 1.7167126574685065e-06, "loss": 0.0016, "step": 17877 }, { "epoch": 2.14, "learning_rate": 1.7166406718656269e-06, "loss": 0.0001, "step": 17878 }, { "epoch": 2.14, "learning_rate": 1.7165686862627476e-06, "loss": 0.0001, "step": 17879 }, { "epoch": 2.14, "learning_rate": 1.716496700659868e-06, "loss": 0.0002, "step": 17880 }, { "epoch": 2.14, "learning_rate": 1.7164247150569886e-06, "loss": 1.0566, "step": 17881 }, { "epoch": 2.14, "learning_rate": 1.7163527294541093e-06, "loss": 0.0002, "step": 17882 }, { "epoch": 2.14, "learning_rate": 1.7162807438512297e-06, "loss": 0.0004, "step": 17883 }, { "epoch": 2.14, "learning_rate": 1.7162087582483505e-06, "loss": 0.0138, "step": 17884 }, { "epoch": 2.14, "learning_rate": 1.7161367726454708e-06, "loss": 0.0001, "step": 17885 }, { "epoch": 2.14, "learning_rate": 1.7160647870425916e-06, "loss": 0.0002, "step": 17886 }, { "epoch": 2.14, "learning_rate": 1.715992801439712e-06, "loss": 0.0002, "step": 17887 }, { "epoch": 2.14, "learning_rate": 1.7159208158368327e-06, "loss": 0.0002, "step": 17888 }, { "epoch": 2.14, "learning_rate": 1.7158488302339533e-06, "loss": 0.0002, "step": 17889 }, { "epoch": 2.14, "learning_rate": 1.715776844631074e-06, "loss": 0.0001, "step": 17890 }, { "epoch": 2.14, "learning_rate": 1.7157048590281944e-06, "loss": 0.0001, "step": 17891 }, { "epoch": 2.14, "learning_rate": 1.7156328734253148e-06, "loss": 0.0001, "step": 17892 }, { "epoch": 2.14, "learning_rate": 1.7155608878224356e-06, "loss": 0.0002, "step": 17893 }, { "epoch": 2.14, "learning_rate": 1.715488902219556e-06, "loss": 0.0001, "step": 17894 }, { "epoch": 2.14, "learning_rate": 1.7154169166166767e-06, "loss": 0.0002, "step": 17895 }, { "epoch": 2.14, "learning_rate": 1.7153449310137973e-06, "loss": 0.0004, "step": 17896 }, { "epoch": 2.14, "learning_rate": 1.715272945410918e-06, "loss": 0.0001, "step": 17897 }, { "epoch": 2.14, "learning_rate": 1.7152009598080384e-06, "loss": 0.0001, "step": 17898 }, { "epoch": 2.14, "learning_rate": 1.7151289742051592e-06, "loss": 0.0002, "step": 17899 }, { "epoch": 2.14, "learning_rate": 1.7150569886022795e-06, "loss": 0.0002, "step": 17900 }, { "epoch": 2.15, "learning_rate": 1.7149850029994003e-06, "loss": 0.0001, "step": 17901 }, { "epoch": 2.15, "learning_rate": 1.7149130173965207e-06, "loss": 0.0001, "step": 17902 }, { "epoch": 2.15, "learning_rate": 1.7148410317936412e-06, "loss": 0.0002, "step": 17903 }, { "epoch": 2.15, "learning_rate": 1.714769046190762e-06, "loss": 0.0003, "step": 17904 }, { "epoch": 2.15, "learning_rate": 1.7146970605878824e-06, "loss": 0.0001, "step": 17905 }, { "epoch": 2.15, "learning_rate": 1.7146250749850031e-06, "loss": 0.0001, "step": 17906 }, { "epoch": 2.15, "learning_rate": 1.7145530893821235e-06, "loss": 0.0002, "step": 17907 }, { "epoch": 2.15, "learning_rate": 1.7144811037792443e-06, "loss": 0.0001, "step": 17908 }, { "epoch": 2.15, "learning_rate": 1.7144091181763646e-06, "loss": 0.0003, "step": 17909 }, { "epoch": 2.15, "learning_rate": 1.7143371325734854e-06, "loss": 0.0001, "step": 17910 }, { "epoch": 2.15, "learning_rate": 1.714265146970606e-06, "loss": 0.0003, "step": 17911 }, { "epoch": 2.15, "learning_rate": 1.7141931613677265e-06, "loss": 0.0001, "step": 17912 }, { "epoch": 2.15, "learning_rate": 1.7141211757648471e-06, "loss": 0.0003, "step": 17913 }, { "epoch": 2.15, "learning_rate": 1.7140491901619679e-06, "loss": 0.0001, "step": 17914 }, { "epoch": 2.15, "learning_rate": 1.7139772045590882e-06, "loss": 0.006, "step": 17915 }, { "epoch": 2.15, "learning_rate": 1.7139052189562086e-06, "loss": 0.0002, "step": 17916 }, { "epoch": 2.15, "learning_rate": 1.7138332333533294e-06, "loss": 0.0002, "step": 17917 }, { "epoch": 2.15, "learning_rate": 1.71376124775045e-06, "loss": 0.0002, "step": 17918 }, { "epoch": 2.15, "learning_rate": 1.7136892621475705e-06, "loss": 0.0052, "step": 17919 }, { "epoch": 2.15, "learning_rate": 1.713617276544691e-06, "loss": 0.0001, "step": 17920 }, { "epoch": 2.15, "learning_rate": 1.7135452909418119e-06, "loss": 0.0002, "step": 17921 }, { "epoch": 2.15, "learning_rate": 1.7134733053389322e-06, "loss": 0.0002, "step": 17922 }, { "epoch": 2.15, "learning_rate": 1.713401319736053e-06, "loss": 0.0002, "step": 17923 }, { "epoch": 2.15, "learning_rate": 1.7133293341331733e-06, "loss": 0.0001, "step": 17924 }, { "epoch": 2.15, "learning_rate": 1.7132573485302941e-06, "loss": 0.9248, "step": 17925 }, { "epoch": 2.15, "learning_rate": 1.7131853629274145e-06, "loss": 0.0001, "step": 17926 }, { "epoch": 2.15, "learning_rate": 1.713113377324535e-06, "loss": 0.0001, "step": 17927 }, { "epoch": 2.15, "learning_rate": 1.7130413917216558e-06, "loss": 0.0003, "step": 17928 }, { "epoch": 2.15, "learning_rate": 1.7129694061187762e-06, "loss": 0.0003, "step": 17929 }, { "epoch": 2.15, "learning_rate": 1.712897420515897e-06, "loss": 0.0001, "step": 17930 }, { "epoch": 2.15, "learning_rate": 1.7128254349130173e-06, "loss": 0.0001, "step": 17931 }, { "epoch": 2.15, "learning_rate": 1.712753449310138e-06, "loss": 0.0002, "step": 17932 }, { "epoch": 2.15, "learning_rate": 1.7126814637072584e-06, "loss": 0.0002, "step": 17933 }, { "epoch": 2.15, "learning_rate": 1.7126094781043792e-06, "loss": 0.0001, "step": 17934 }, { "epoch": 2.15, "learning_rate": 1.7125374925014998e-06, "loss": 0.0002, "step": 17935 }, { "epoch": 2.15, "learning_rate": 1.7124655068986206e-06, "loss": 0.0001, "step": 17936 }, { "epoch": 2.15, "learning_rate": 1.712393521295741e-06, "loss": 0.0003, "step": 17937 }, { "epoch": 2.15, "learning_rate": 1.7123215356928613e-06, "loss": 0.0002, "step": 17938 }, { "epoch": 2.15, "learning_rate": 1.712249550089982e-06, "loss": 0.0001, "step": 17939 }, { "epoch": 2.15, "learning_rate": 1.7121775644871024e-06, "loss": 0.0003, "step": 17940 }, { "epoch": 2.15, "learning_rate": 1.7121055788842232e-06, "loss": 0.0002, "step": 17941 }, { "epoch": 2.15, "learning_rate": 1.7120335932813438e-06, "loss": 0.0001, "step": 17942 }, { "epoch": 2.15, "learning_rate": 1.7119616076784645e-06, "loss": 0.0003, "step": 17943 }, { "epoch": 2.15, "learning_rate": 1.711889622075585e-06, "loss": 0.0001, "step": 17944 }, { "epoch": 2.15, "learning_rate": 1.7118176364727057e-06, "loss": 0.0001, "step": 17945 }, { "epoch": 2.15, "learning_rate": 1.711745650869826e-06, "loss": 0.0002, "step": 17946 }, { "epoch": 2.15, "learning_rate": 1.7116736652669468e-06, "loss": 0.0001, "step": 17947 }, { "epoch": 2.15, "learning_rate": 1.7116016796640672e-06, "loss": 0.0001, "step": 17948 }, { "epoch": 2.15, "learning_rate": 1.7115296940611877e-06, "loss": 0.0002, "step": 17949 }, { "epoch": 2.15, "learning_rate": 1.7114577084583083e-06, "loss": 1.8731, "step": 17950 }, { "epoch": 2.15, "learning_rate": 1.7113857228554289e-06, "loss": 0.0002, "step": 17951 }, { "epoch": 2.15, "learning_rate": 1.7113137372525496e-06, "loss": 0.0002, "step": 17952 }, { "epoch": 2.15, "learning_rate": 1.71124175164967e-06, "loss": 0.0001, "step": 17953 }, { "epoch": 2.15, "learning_rate": 1.7111697660467908e-06, "loss": 0.0005, "step": 17954 }, { "epoch": 2.15, "learning_rate": 1.7110977804439111e-06, "loss": 0.0002, "step": 17955 }, { "epoch": 2.15, "learning_rate": 1.711025794841032e-06, "loss": 0.0001, "step": 17956 }, { "epoch": 2.15, "learning_rate": 1.7109538092381523e-06, "loss": 0.0002, "step": 17957 }, { "epoch": 2.15, "learning_rate": 1.710881823635273e-06, "loss": 0.0002, "step": 17958 }, { "epoch": 2.15, "learning_rate": 1.7108098380323936e-06, "loss": 0.0002, "step": 17959 }, { "epoch": 2.15, "learning_rate": 1.7107378524295144e-06, "loss": 0.0004, "step": 17960 }, { "epoch": 2.15, "learning_rate": 1.7106658668266347e-06, "loss": 0.0001, "step": 17961 }, { "epoch": 2.15, "learning_rate": 1.710593881223755e-06, "loss": 0.0002, "step": 17962 }, { "epoch": 2.15, "learning_rate": 1.7105218956208759e-06, "loss": 0.0001, "step": 17963 }, { "epoch": 2.15, "learning_rate": 1.7104499100179962e-06, "loss": 0.0002, "step": 17964 }, { "epoch": 2.15, "learning_rate": 1.710377924415117e-06, "loss": 0.0001, "step": 17965 }, { "epoch": 2.15, "learning_rate": 1.7103059388122376e-06, "loss": 0.0002, "step": 17966 }, { "epoch": 2.15, "learning_rate": 1.7102339532093584e-06, "loss": 0.0001, "step": 17967 }, { "epoch": 2.15, "learning_rate": 1.7101619676064787e-06, "loss": 0.0002, "step": 17968 }, { "epoch": 2.15, "learning_rate": 1.7100899820035995e-06, "loss": 0.0002, "step": 17969 }, { "epoch": 2.15, "learning_rate": 1.7100179964007198e-06, "loss": 0.0002, "step": 17970 }, { "epoch": 2.15, "learning_rate": 1.7099460107978406e-06, "loss": 0.0002, "step": 17971 }, { "epoch": 2.15, "learning_rate": 1.709874025194961e-06, "loss": 0.0005, "step": 17972 }, { "epoch": 2.15, "learning_rate": 1.7098020395920815e-06, "loss": 0.0002, "step": 17973 }, { "epoch": 2.15, "learning_rate": 1.7097300539892023e-06, "loss": 0.0002, "step": 17974 }, { "epoch": 2.15, "learning_rate": 1.7096580683863227e-06, "loss": 0.0008, "step": 17975 }, { "epoch": 2.15, "learning_rate": 1.7095860827834435e-06, "loss": 0.0001, "step": 17976 }, { "epoch": 2.15, "learning_rate": 1.7095140971805638e-06, "loss": 0.0002, "step": 17977 }, { "epoch": 2.15, "learning_rate": 1.7094421115776846e-06, "loss": 0.0004, "step": 17978 }, { "epoch": 2.15, "learning_rate": 1.709370125974805e-06, "loss": 0.0001, "step": 17979 }, { "epoch": 2.15, "learning_rate": 1.7092981403719257e-06, "loss": 0.0002, "step": 17980 }, { "epoch": 2.15, "learning_rate": 1.7092261547690463e-06, "loss": 0.0002, "step": 17981 }, { "epoch": 2.15, "learning_rate": 1.7091541691661669e-06, "loss": 0.0003, "step": 17982 }, { "epoch": 2.15, "learning_rate": 1.7090821835632874e-06, "loss": 0.0002, "step": 17983 }, { "epoch": 2.16, "learning_rate": 1.7090101979604078e-06, "loss": 0.0001, "step": 17984 }, { "epoch": 2.16, "learning_rate": 1.7089382123575286e-06, "loss": 0.0001, "step": 17985 }, { "epoch": 2.16, "learning_rate": 1.708866226754649e-06, "loss": 0.0001, "step": 17986 }, { "epoch": 2.16, "learning_rate": 1.7087942411517697e-06, "loss": 0.0001, "step": 17987 }, { "epoch": 2.16, "learning_rate": 1.7087222555488903e-06, "loss": 0.0108, "step": 17988 }, { "epoch": 2.16, "learning_rate": 1.7086502699460108e-06, "loss": 0.0001, "step": 17989 }, { "epoch": 2.16, "learning_rate": 1.7085782843431314e-06, "loss": 0.0005, "step": 17990 }, { "epoch": 2.16, "learning_rate": 1.7085062987402522e-06, "loss": 0.0002, "step": 17991 }, { "epoch": 2.16, "learning_rate": 1.7084343131373725e-06, "loss": 0.0002, "step": 17992 }, { "epoch": 2.16, "learning_rate": 1.7083623275344933e-06, "loss": 0.1691, "step": 17993 }, { "epoch": 2.16, "learning_rate": 1.7082903419316137e-06, "loss": 0.0002, "step": 17994 }, { "epoch": 2.16, "learning_rate": 1.7082183563287342e-06, "loss": 0.0002, "step": 17995 }, { "epoch": 2.16, "learning_rate": 1.7081463707258548e-06, "loss": 0.0001, "step": 17996 }, { "epoch": 2.16, "learning_rate": 1.7080743851229754e-06, "loss": 0.0023, "step": 17997 }, { "epoch": 2.16, "learning_rate": 1.7080023995200961e-06, "loss": 0.0002, "step": 17998 }, { "epoch": 2.16, "learning_rate": 1.7079304139172165e-06, "loss": 0.0002, "step": 17999 }, { "epoch": 2.16, "learning_rate": 1.7078584283143373e-06, "loss": 0.0001, "step": 18000 }, { "epoch": 2.16, "learning_rate": 1.7077864427114576e-06, "loss": 0.0002, "step": 18001 }, { "epoch": 2.16, "learning_rate": 1.7077144571085784e-06, "loss": 0.0001, "step": 18002 }, { "epoch": 2.16, "learning_rate": 1.7076424715056988e-06, "loss": 0.0002, "step": 18003 }, { "epoch": 2.16, "learning_rate": 1.7075704859028195e-06, "loss": 0.0001, "step": 18004 }, { "epoch": 2.16, "learning_rate": 1.70749850029994e-06, "loss": 0.0001, "step": 18005 }, { "epoch": 2.16, "learning_rate": 1.7074265146970609e-06, "loss": 0.0002, "step": 18006 }, { "epoch": 2.16, "learning_rate": 1.7073545290941812e-06, "loss": 0.0001, "step": 18007 }, { "epoch": 2.16, "learning_rate": 1.7072825434913016e-06, "loss": 0.0002, "step": 18008 }, { "epoch": 2.16, "learning_rate": 1.7072105578884224e-06, "loss": 0.0002, "step": 18009 }, { "epoch": 2.16, "learning_rate": 1.7071385722855427e-06, "loss": 0.0001, "step": 18010 }, { "epoch": 2.16, "learning_rate": 1.7070665866826635e-06, "loss": 0.0225, "step": 18011 }, { "epoch": 2.16, "learning_rate": 1.706994601079784e-06, "loss": 0.0001, "step": 18012 }, { "epoch": 2.16, "learning_rate": 1.7069226154769049e-06, "loss": 0.0002, "step": 18013 }, { "epoch": 2.16, "learning_rate": 1.7068506298740252e-06, "loss": 0.0001, "step": 18014 }, { "epoch": 2.16, "learning_rate": 1.706778644271146e-06, "loss": 0.0001, "step": 18015 }, { "epoch": 2.16, "learning_rate": 1.7067066586682663e-06, "loss": 0.0002, "step": 18016 }, { "epoch": 2.16, "learning_rate": 1.7066346730653871e-06, "loss": 0.0002, "step": 18017 }, { "epoch": 2.16, "learning_rate": 1.7065626874625075e-06, "loss": 0.0003, "step": 18018 }, { "epoch": 2.16, "learning_rate": 1.706490701859628e-06, "loss": 0.0003, "step": 18019 }, { "epoch": 2.16, "learning_rate": 1.7064187162567488e-06, "loss": 0.0001, "step": 18020 }, { "epoch": 2.16, "learning_rate": 1.7063467306538692e-06, "loss": 0.0002, "step": 18021 }, { "epoch": 2.16, "learning_rate": 1.70627474505099e-06, "loss": 0.0002, "step": 18022 }, { "epoch": 2.16, "learning_rate": 1.7062027594481103e-06, "loss": 0.0001, "step": 18023 }, { "epoch": 2.16, "learning_rate": 1.706130773845231e-06, "loss": 0.0002, "step": 18024 }, { "epoch": 2.16, "learning_rate": 1.7060587882423514e-06, "loss": 0.0004, "step": 18025 }, { "epoch": 2.16, "learning_rate": 1.7059868026394722e-06, "loss": 0.0002, "step": 18026 }, { "epoch": 2.16, "learning_rate": 1.7059148170365928e-06, "loss": 0.0001, "step": 18027 }, { "epoch": 2.16, "learning_rate": 1.7058428314337134e-06, "loss": 0.003, "step": 18028 }, { "epoch": 2.16, "learning_rate": 1.705770845830834e-06, "loss": 0.0002, "step": 18029 }, { "epoch": 2.16, "learning_rate": 1.7056988602279543e-06, "loss": 0.0001, "step": 18030 }, { "epoch": 2.16, "learning_rate": 1.705626874625075e-06, "loss": 0.0003, "step": 18031 }, { "epoch": 2.16, "learning_rate": 1.7055548890221954e-06, "loss": 0.0002, "step": 18032 }, { "epoch": 2.16, "learning_rate": 1.7054829034193162e-06, "loss": 0.0002, "step": 18033 }, { "epoch": 2.16, "learning_rate": 1.7054109178164368e-06, "loss": 0.0001, "step": 18034 }, { "epoch": 2.16, "learning_rate": 1.7053389322135573e-06, "loss": 0.0001, "step": 18035 }, { "epoch": 2.16, "learning_rate": 1.7052669466106779e-06, "loss": 0.0001, "step": 18036 }, { "epoch": 2.16, "learning_rate": 1.7051949610077987e-06, "loss": 0.0001, "step": 18037 }, { "epoch": 2.16, "learning_rate": 1.705122975404919e-06, "loss": 0.0002, "step": 18038 }, { "epoch": 2.16, "learning_rate": 1.7050509898020398e-06, "loss": 0.0003, "step": 18039 }, { "epoch": 2.16, "learning_rate": 1.7049790041991602e-06, "loss": 0.0003, "step": 18040 }, { "epoch": 2.16, "learning_rate": 1.704907018596281e-06, "loss": 0.0001, "step": 18041 }, { "epoch": 2.16, "learning_rate": 1.7048350329934013e-06, "loss": 0.0002, "step": 18042 }, { "epoch": 2.16, "learning_rate": 1.7047630473905219e-06, "loss": 0.0002, "step": 18043 }, { "epoch": 2.16, "learning_rate": 1.7046910617876426e-06, "loss": 0.0002, "step": 18044 }, { "epoch": 2.16, "learning_rate": 1.704619076184763e-06, "loss": 0.0003, "step": 18045 }, { "epoch": 2.16, "learning_rate": 1.7045470905818838e-06, "loss": 0.0002, "step": 18046 }, { "epoch": 2.16, "learning_rate": 1.7044751049790041e-06, "loss": 0.0002, "step": 18047 }, { "epoch": 2.16, "learning_rate": 1.704403119376125e-06, "loss": 0.0001, "step": 18048 }, { "epoch": 2.16, "learning_rate": 1.7043311337732453e-06, "loss": 0.0001, "step": 18049 }, { "epoch": 2.16, "learning_rate": 1.704259148170366e-06, "loss": 0.0003, "step": 18050 }, { "epoch": 2.16, "learning_rate": 1.7041871625674866e-06, "loss": 0.0002, "step": 18051 }, { "epoch": 2.16, "learning_rate": 1.7041151769646074e-06, "loss": 0.0001, "step": 18052 }, { "epoch": 2.16, "learning_rate": 1.7040431913617277e-06, "loss": 0.0001, "step": 18053 }, { "epoch": 2.16, "learning_rate": 1.703971205758848e-06, "loss": 0.0002, "step": 18054 }, { "epoch": 2.16, "learning_rate": 1.7038992201559689e-06, "loss": 0.0001, "step": 18055 }, { "epoch": 2.16, "learning_rate": 1.7038272345530892e-06, "loss": 0.0003, "step": 18056 }, { "epoch": 2.16, "learning_rate": 1.70375524895021e-06, "loss": 0.0002, "step": 18057 }, { "epoch": 2.16, "learning_rate": 1.7036832633473306e-06, "loss": 0.0001, "step": 18058 }, { "epoch": 2.16, "learning_rate": 1.7036112777444513e-06, "loss": 0.7808, "step": 18059 }, { "epoch": 2.16, "learning_rate": 1.7035392921415717e-06, "loss": 0.0002, "step": 18060 }, { "epoch": 2.16, "learning_rate": 1.7034673065386925e-06, "loss": 0.0016, "step": 18061 }, { "epoch": 2.16, "learning_rate": 1.7033953209358128e-06, "loss": 0.0001, "step": 18062 }, { "epoch": 2.16, "learning_rate": 1.7033233353329336e-06, "loss": 0.0002, "step": 18063 }, { "epoch": 2.16, "learning_rate": 1.703251349730054e-06, "loss": 0.0001, "step": 18064 }, { "epoch": 2.16, "learning_rate": 1.7031793641271745e-06, "loss": 0.0002, "step": 18065 }, { "epoch": 2.16, "learning_rate": 1.7031073785242953e-06, "loss": 0.0001, "step": 18066 }, { "epoch": 2.17, "learning_rate": 1.7030353929214157e-06, "loss": 0.0134, "step": 18067 }, { "epoch": 2.17, "learning_rate": 1.7029634073185364e-06, "loss": 0.0002, "step": 18068 }, { "epoch": 2.17, "learning_rate": 1.7028914217156568e-06, "loss": 0.0001, "step": 18069 }, { "epoch": 2.17, "learning_rate": 1.7028194361127776e-06, "loss": 0.0002, "step": 18070 }, { "epoch": 2.17, "learning_rate": 1.702747450509898e-06, "loss": 0.0004, "step": 18071 }, { "epoch": 2.17, "learning_rate": 1.7026754649070187e-06, "loss": 0.0016, "step": 18072 }, { "epoch": 2.17, "learning_rate": 1.702603479304139e-06, "loss": 0.0001, "step": 18073 }, { "epoch": 2.17, "learning_rate": 1.7025314937012598e-06, "loss": 0.0001, "step": 18074 }, { "epoch": 2.17, "learning_rate": 1.7024595080983804e-06, "loss": 0.0047, "step": 18075 }, { "epoch": 2.17, "learning_rate": 1.7023875224955008e-06, "loss": 0.0001, "step": 18076 }, { "epoch": 2.17, "learning_rate": 1.7023155368926216e-06, "loss": 0.0001, "step": 18077 }, { "epoch": 2.17, "learning_rate": 1.702243551289742e-06, "loss": 0.0002, "step": 18078 }, { "epoch": 2.17, "learning_rate": 1.7021715656868627e-06, "loss": 0.0003, "step": 18079 }, { "epoch": 2.17, "learning_rate": 1.702099580083983e-06, "loss": 0.0002, "step": 18080 }, { "epoch": 2.17, "learning_rate": 1.7020275944811038e-06, "loss": 0.0002, "step": 18081 }, { "epoch": 2.17, "learning_rate": 1.7019556088782244e-06, "loss": 0.0001, "step": 18082 }, { "epoch": 2.17, "learning_rate": 1.7018836232753452e-06, "loss": 0.0001, "step": 18083 }, { "epoch": 2.17, "learning_rate": 1.7018116376724655e-06, "loss": 0.0001, "step": 18084 }, { "epoch": 2.17, "learning_rate": 1.7017396520695863e-06, "loss": 0.0001, "step": 18085 }, { "epoch": 2.17, "learning_rate": 1.7016676664667067e-06, "loss": 0.0001, "step": 18086 }, { "epoch": 2.17, "learning_rate": 1.7015956808638274e-06, "loss": 0.0002, "step": 18087 }, { "epoch": 2.17, "learning_rate": 1.7015236952609478e-06, "loss": 0.0002, "step": 18088 }, { "epoch": 2.17, "learning_rate": 1.7014517096580684e-06, "loss": 0.0001, "step": 18089 }, { "epoch": 2.17, "learning_rate": 1.7013797240551891e-06, "loss": 0.0004, "step": 18090 }, { "epoch": 2.17, "learning_rate": 1.7013077384523095e-06, "loss": 0.0002, "step": 18091 }, { "epoch": 2.17, "learning_rate": 1.7012357528494303e-06, "loss": 0.0003, "step": 18092 }, { "epoch": 2.17, "learning_rate": 1.7011637672465506e-06, "loss": 0.0001, "step": 18093 }, { "epoch": 2.17, "learning_rate": 1.7010917816436714e-06, "loss": 0.0001, "step": 18094 }, { "epoch": 2.17, "learning_rate": 1.7010197960407918e-06, "loss": 0.0002, "step": 18095 }, { "epoch": 2.17, "learning_rate": 1.7009478104379125e-06, "loss": 0.8109, "step": 18096 }, { "epoch": 2.17, "learning_rate": 1.700875824835033e-06, "loss": 1.1264, "step": 18097 }, { "epoch": 2.17, "learning_rate": 1.7008038392321537e-06, "loss": 0.0006, "step": 18098 }, { "epoch": 2.17, "learning_rate": 1.7007318536292742e-06, "loss": 0.0002, "step": 18099 }, { "epoch": 2.17, "learning_rate": 1.7006598680263946e-06, "loss": 0.0002, "step": 18100 }, { "epoch": 2.17, "learning_rate": 1.7005878824235154e-06, "loss": 0.0003, "step": 18101 }, { "epoch": 2.17, "learning_rate": 1.7005158968206357e-06, "loss": 0.0009, "step": 18102 }, { "epoch": 2.17, "learning_rate": 1.7004439112177565e-06, "loss": 0.0031, "step": 18103 }, { "epoch": 2.17, "learning_rate": 1.700371925614877e-06, "loss": 0.0002, "step": 18104 }, { "epoch": 2.17, "learning_rate": 1.7002999400119976e-06, "loss": 0.0005, "step": 18105 }, { "epoch": 2.17, "learning_rate": 1.7002279544091182e-06, "loss": 0.0002, "step": 18106 }, { "epoch": 2.17, "learning_rate": 1.700155968806239e-06, "loss": 0.0002, "step": 18107 }, { "epoch": 2.17, "learning_rate": 1.7000839832033593e-06, "loss": 0.0006, "step": 18108 }, { "epoch": 2.17, "learning_rate": 1.7000119976004801e-06, "loss": 0.0001, "step": 18109 }, { "epoch": 2.17, "learning_rate": 1.6999400119976005e-06, "loss": 0.0001, "step": 18110 }, { "epoch": 2.17, "learning_rate": 1.699868026394721e-06, "loss": 0.0002, "step": 18111 }, { "epoch": 2.17, "learning_rate": 1.6997960407918416e-06, "loss": 0.0003, "step": 18112 }, { "epoch": 2.17, "learning_rate": 1.6997240551889622e-06, "loss": 0.0001, "step": 18113 }, { "epoch": 2.17, "learning_rate": 1.699652069586083e-06, "loss": 0.0001, "step": 18114 }, { "epoch": 2.17, "learning_rate": 1.6995800839832033e-06, "loss": 0.0002, "step": 18115 }, { "epoch": 2.17, "learning_rate": 1.699508098380324e-06, "loss": 0.0001, "step": 18116 }, { "epoch": 2.17, "learning_rate": 1.6994361127774444e-06, "loss": 0.0001, "step": 18117 }, { "epoch": 2.17, "learning_rate": 1.6993641271745652e-06, "loss": 0.0002, "step": 18118 }, { "epoch": 2.17, "learning_rate": 1.6992921415716856e-06, "loss": 0.0002, "step": 18119 }, { "epoch": 2.17, "learning_rate": 1.6992201559688063e-06, "loss": 0.0055, "step": 18120 }, { "epoch": 2.17, "learning_rate": 1.699148170365927e-06, "loss": 0.0002, "step": 18121 }, { "epoch": 2.17, "learning_rate": 1.6990761847630473e-06, "loss": 0.0001, "step": 18122 }, { "epoch": 2.17, "learning_rate": 1.699004199160168e-06, "loss": 0.0001, "step": 18123 }, { "epoch": 2.17, "learning_rate": 1.6989322135572884e-06, "loss": 0.0003, "step": 18124 }, { "epoch": 2.17, "learning_rate": 1.6988602279544092e-06, "loss": 0.0, "step": 18125 }, { "epoch": 2.17, "learning_rate": 1.6987882423515295e-06, "loss": 0.0001, "step": 18126 }, { "epoch": 2.17, "learning_rate": 1.6987162567486503e-06, "loss": 0.0002, "step": 18127 }, { "epoch": 2.17, "learning_rate": 1.6986442711457709e-06, "loss": 0.0002, "step": 18128 }, { "epoch": 2.17, "learning_rate": 1.6985722855428917e-06, "loss": 0.0002, "step": 18129 }, { "epoch": 2.17, "learning_rate": 1.698500299940012e-06, "loss": 0.0003, "step": 18130 }, { "epoch": 2.17, "learning_rate": 1.6984283143371328e-06, "loss": 0.0002, "step": 18131 }, { "epoch": 2.17, "learning_rate": 1.6983563287342531e-06, "loss": 0.0002, "step": 18132 }, { "epoch": 2.17, "learning_rate": 1.698284343131374e-06, "loss": 0.0001, "step": 18133 }, { "epoch": 2.17, "learning_rate": 1.6982123575284943e-06, "loss": 0.0001, "step": 18134 }, { "epoch": 2.17, "learning_rate": 1.6981403719256148e-06, "loss": 0.0002, "step": 18135 }, { "epoch": 2.17, "learning_rate": 1.6980683863227356e-06, "loss": 0.0002, "step": 18136 }, { "epoch": 2.17, "learning_rate": 1.697996400719856e-06, "loss": 0.0001, "step": 18137 }, { "epoch": 2.17, "learning_rate": 1.6979244151169768e-06, "loss": 0.0001, "step": 18138 }, { "epoch": 2.17, "learning_rate": 1.6978524295140971e-06, "loss": 0.0002, "step": 18139 }, { "epoch": 2.17, "learning_rate": 1.6977804439112179e-06, "loss": 0.0001, "step": 18140 }, { "epoch": 2.17, "learning_rate": 1.6977084583083382e-06, "loss": 0.0001, "step": 18141 }, { "epoch": 2.17, "learning_rate": 1.697636472705459e-06, "loss": 0.0001, "step": 18142 }, { "epoch": 2.17, "learning_rate": 1.6975644871025796e-06, "loss": 0.8061, "step": 18143 }, { "epoch": 2.17, "learning_rate": 1.6974925014997002e-06, "loss": 0.0001, "step": 18144 }, { "epoch": 2.17, "learning_rate": 1.6974205158968207e-06, "loss": 0.0001, "step": 18145 }, { "epoch": 2.17, "learning_rate": 1.697348530293941e-06, "loss": 0.0002, "step": 18146 }, { "epoch": 2.17, "learning_rate": 1.6972765446910619e-06, "loss": 0.0001, "step": 18147 }, { "epoch": 2.17, "learning_rate": 1.6972045590881822e-06, "loss": 0.0001, "step": 18148 }, { "epoch": 2.17, "learning_rate": 1.697132573485303e-06, "loss": 0.8198, "step": 18149 }, { "epoch": 2.17, "learning_rate": 1.6970605878824236e-06, "loss": 0.0002, "step": 18150 }, { "epoch": 2.18, "learning_rate": 1.6969886022795441e-06, "loss": 0.0001, "step": 18151 }, { "epoch": 2.18, "learning_rate": 1.6969166166766647e-06, "loss": 0.0005, "step": 18152 }, { "epoch": 2.18, "learning_rate": 1.6968446310737855e-06, "loss": 0.0113, "step": 18153 }, { "epoch": 2.18, "learning_rate": 1.6967726454709058e-06, "loss": 1.1575, "step": 18154 }, { "epoch": 2.18, "learning_rate": 1.6967006598680266e-06, "loss": 0.0003, "step": 18155 }, { "epoch": 2.18, "learning_rate": 1.696628674265147e-06, "loss": 0.0001, "step": 18156 }, { "epoch": 2.18, "learning_rate": 1.6965566886622675e-06, "loss": 0.0002, "step": 18157 }, { "epoch": 2.18, "learning_rate": 1.696484703059388e-06, "loss": 0.0098, "step": 18158 }, { "epoch": 2.18, "learning_rate": 1.6964127174565087e-06, "loss": 0.0001, "step": 18159 }, { "epoch": 2.18, "learning_rate": 1.6963407318536294e-06, "loss": 0.0262, "step": 18160 }, { "epoch": 2.18, "learning_rate": 1.6962687462507498e-06, "loss": 0.0001, "step": 18161 }, { "epoch": 2.18, "learning_rate": 1.6961967606478706e-06, "loss": 0.0001, "step": 18162 }, { "epoch": 2.18, "learning_rate": 1.696124775044991e-06, "loss": 0.001, "step": 18163 }, { "epoch": 2.18, "learning_rate": 1.6960527894421117e-06, "loss": 0.0001, "step": 18164 }, { "epoch": 2.18, "learning_rate": 1.695980803839232e-06, "loss": 0.0002, "step": 18165 }, { "epoch": 2.18, "learning_rate": 1.6959088182363528e-06, "loss": 0.0001, "step": 18166 }, { "epoch": 2.18, "learning_rate": 1.6958368326334734e-06, "loss": 0.0001, "step": 18167 }, { "epoch": 2.18, "learning_rate": 1.6957648470305942e-06, "loss": 0.0001, "step": 18168 }, { "epoch": 2.18, "learning_rate": 1.6956928614277145e-06, "loss": 0.0002, "step": 18169 }, { "epoch": 2.18, "learning_rate": 1.695620875824835e-06, "loss": 0.0004, "step": 18170 }, { "epoch": 2.18, "learning_rate": 1.6955488902219557e-06, "loss": 0.0002, "step": 18171 }, { "epoch": 2.18, "learning_rate": 1.695476904619076e-06, "loss": 0.0002, "step": 18172 }, { "epoch": 2.18, "learning_rate": 1.6954049190161968e-06, "loss": 0.0003, "step": 18173 }, { "epoch": 2.18, "learning_rate": 1.6953329334133174e-06, "loss": 0.0002, "step": 18174 }, { "epoch": 2.18, "learning_rate": 1.6952609478104382e-06, "loss": 0.0001, "step": 18175 }, { "epoch": 2.18, "learning_rate": 1.6951889622075585e-06, "loss": 0.0003, "step": 18176 }, { "epoch": 2.18, "learning_rate": 1.6951169766046793e-06, "loss": 0.0003, "step": 18177 }, { "epoch": 2.18, "learning_rate": 1.6950449910017996e-06, "loss": 0.0007, "step": 18178 }, { "epoch": 2.18, "learning_rate": 1.6949730053989204e-06, "loss": 0.0002, "step": 18179 }, { "epoch": 2.18, "learning_rate": 1.6949010197960408e-06, "loss": 0.0001, "step": 18180 }, { "epoch": 2.18, "learning_rate": 1.6948290341931613e-06, "loss": 0.0002, "step": 18181 }, { "epoch": 2.18, "learning_rate": 1.6947570485902821e-06, "loss": 0.0001, "step": 18182 }, { "epoch": 2.18, "learning_rate": 1.6946850629874025e-06, "loss": 0.0736, "step": 18183 }, { "epoch": 2.18, "learning_rate": 1.6946130773845233e-06, "loss": 0.0002, "step": 18184 }, { "epoch": 2.18, "learning_rate": 1.6945410917816436e-06, "loss": 0.0004, "step": 18185 }, { "epoch": 2.18, "learning_rate": 1.6944691061787644e-06, "loss": 0.0027, "step": 18186 }, { "epoch": 2.18, "learning_rate": 1.6943971205758847e-06, "loss": 0.0001, "step": 18187 }, { "epoch": 2.18, "learning_rate": 1.6943251349730055e-06, "loss": 0.0001, "step": 18188 }, { "epoch": 2.18, "learning_rate": 1.694253149370126e-06, "loss": 0.0002, "step": 18189 }, { "epoch": 2.18, "learning_rate": 1.6941811637672467e-06, "loss": 0.0002, "step": 18190 }, { "epoch": 2.18, "learning_rate": 1.6941091781643672e-06, "loss": 0.0003, "step": 18191 }, { "epoch": 2.18, "learning_rate": 1.6940371925614876e-06, "loss": 0.0002, "step": 18192 }, { "epoch": 2.18, "learning_rate": 1.6939652069586084e-06, "loss": 0.0002, "step": 18193 }, { "epoch": 2.18, "learning_rate": 1.6938932213557287e-06, "loss": 0.0002, "step": 18194 }, { "epoch": 2.18, "learning_rate": 1.6938212357528495e-06, "loss": 0.0002, "step": 18195 }, { "epoch": 2.18, "learning_rate": 1.69374925014997e-06, "loss": 0.0002, "step": 18196 }, { "epoch": 2.18, "learning_rate": 1.6936772645470906e-06, "loss": 0.0005, "step": 18197 }, { "epoch": 2.18, "learning_rate": 1.6936052789442112e-06, "loss": 0.0002, "step": 18198 }, { "epoch": 2.18, "learning_rate": 1.693533293341332e-06, "loss": 0.0001, "step": 18199 }, { "epoch": 2.18, "learning_rate": 1.6934613077384523e-06, "loss": 0.0001, "step": 18200 }, { "epoch": 2.18, "learning_rate": 1.693389322135573e-06, "loss": 0.0003, "step": 18201 }, { "epoch": 2.18, "learning_rate": 1.6933173365326935e-06, "loss": 0.0002, "step": 18202 }, { "epoch": 2.18, "learning_rate": 1.6932453509298138e-06, "loss": 0.0001, "step": 18203 }, { "epoch": 2.18, "learning_rate": 1.6931733653269346e-06, "loss": 0.0002, "step": 18204 }, { "epoch": 2.18, "learning_rate": 1.6931013797240552e-06, "loss": 0.0001, "step": 18205 }, { "epoch": 2.18, "learning_rate": 1.693029394121176e-06, "loss": 0.0002, "step": 18206 }, { "epoch": 2.18, "learning_rate": 1.6929574085182963e-06, "loss": 0.0003, "step": 18207 }, { "epoch": 2.18, "learning_rate": 1.692885422915417e-06, "loss": 0.0001, "step": 18208 }, { "epoch": 2.18, "learning_rate": 1.6928134373125374e-06, "loss": 0.0002, "step": 18209 }, { "epoch": 2.18, "learning_rate": 1.6927414517096582e-06, "loss": 0.0002, "step": 18210 }, { "epoch": 2.18, "learning_rate": 1.6926694661067786e-06, "loss": 0.001, "step": 18211 }, { "epoch": 2.18, "learning_rate": 1.6925974805038993e-06, "loss": 0.0001, "step": 18212 }, { "epoch": 2.18, "learning_rate": 1.69252549490102e-06, "loss": 0.0002, "step": 18213 }, { "epoch": 2.18, "learning_rate": 1.6924535092981407e-06, "loss": 0.0001, "step": 18214 }, { "epoch": 2.18, "learning_rate": 1.692381523695261e-06, "loss": 0.0001, "step": 18215 }, { "epoch": 2.18, "learning_rate": 1.6923095380923814e-06, "loss": 0.0002, "step": 18216 }, { "epoch": 2.18, "learning_rate": 1.6922375524895022e-06, "loss": 0.0001, "step": 18217 }, { "epoch": 2.18, "learning_rate": 1.6921655668866225e-06, "loss": 0.0002, "step": 18218 }, { "epoch": 2.18, "learning_rate": 1.6920935812837433e-06, "loss": 0.0001, "step": 18219 }, { "epoch": 2.18, "learning_rate": 1.6920215956808639e-06, "loss": 0.0001, "step": 18220 }, { "epoch": 2.18, "learning_rate": 1.6919496100779844e-06, "loss": 0.0001, "step": 18221 }, { "epoch": 2.18, "learning_rate": 1.691877624475105e-06, "loss": 0.0002, "step": 18222 }, { "epoch": 2.18, "learning_rate": 1.6918056388722258e-06, "loss": 0.0004, "step": 18223 }, { "epoch": 2.18, "learning_rate": 1.6917336532693461e-06, "loss": 0.0003, "step": 18224 }, { "epoch": 2.18, "learning_rate": 1.691661667666467e-06, "loss": 0.0002, "step": 18225 }, { "epoch": 2.18, "learning_rate": 1.6915896820635873e-06, "loss": 0.0002, "step": 18226 }, { "epoch": 2.18, "learning_rate": 1.6915176964607078e-06, "loss": 0.0005, "step": 18227 }, { "epoch": 2.18, "learning_rate": 1.6914457108578284e-06, "loss": 0.0002, "step": 18228 }, { "epoch": 2.18, "learning_rate": 1.691373725254949e-06, "loss": 0.0002, "step": 18229 }, { "epoch": 2.18, "learning_rate": 1.6913017396520698e-06, "loss": 0.0002, "step": 18230 }, { "epoch": 2.18, "learning_rate": 1.6912297540491901e-06, "loss": 0.0001, "step": 18231 }, { "epoch": 2.18, "learning_rate": 1.6911577684463109e-06, "loss": 0.0002, "step": 18232 }, { "epoch": 2.18, "learning_rate": 1.6910857828434312e-06, "loss": 0.0003, "step": 18233 }, { "epoch": 2.19, "learning_rate": 1.691013797240552e-06, "loss": 0.0002, "step": 18234 }, { "epoch": 2.19, "learning_rate": 1.6909418116376724e-06, "loss": 0.0004, "step": 18235 }, { "epoch": 2.19, "learning_rate": 1.6908698260347932e-06, "loss": 0.3372, "step": 18236 }, { "epoch": 2.19, "learning_rate": 1.6907978404319137e-06, "loss": 0.0003, "step": 18237 }, { "epoch": 2.19, "learning_rate": 1.690725854829034e-06, "loss": 0.0002, "step": 18238 }, { "epoch": 2.19, "learning_rate": 1.6906538692261549e-06, "loss": 0.0004, "step": 18239 }, { "epoch": 2.19, "learning_rate": 1.6905818836232752e-06, "loss": 0.0003, "step": 18240 }, { "epoch": 2.19, "learning_rate": 1.690509898020396e-06, "loss": 0.0001, "step": 18241 }, { "epoch": 2.19, "learning_rate": 1.6904379124175163e-06, "loss": 0.0002, "step": 18242 }, { "epoch": 2.19, "learning_rate": 1.6903659268146371e-06, "loss": 0.0001, "step": 18243 }, { "epoch": 2.19, "learning_rate": 1.6902939412117577e-06, "loss": 0.0002, "step": 18244 }, { "epoch": 2.19, "learning_rate": 1.6902219556088785e-06, "loss": 0.0001, "step": 18245 }, { "epoch": 2.19, "learning_rate": 1.6901499700059988e-06, "loss": 0.0002, "step": 18246 }, { "epoch": 2.19, "learning_rate": 1.6900779844031196e-06, "loss": 1.1182, "step": 18247 }, { "epoch": 2.19, "learning_rate": 1.69000599880024e-06, "loss": 0.0013, "step": 18248 }, { "epoch": 2.19, "learning_rate": 1.6899340131973603e-06, "loss": 0.0001, "step": 18249 }, { "epoch": 2.19, "learning_rate": 1.689862027594481e-06, "loss": 0.0004, "step": 18250 }, { "epoch": 2.19, "learning_rate": 1.6897900419916017e-06, "loss": 0.0001, "step": 18251 }, { "epoch": 2.19, "learning_rate": 1.6897180563887224e-06, "loss": 0.0001, "step": 18252 }, { "epoch": 2.19, "learning_rate": 1.6896460707858428e-06, "loss": 0.0006, "step": 18253 }, { "epoch": 2.19, "learning_rate": 1.6895740851829636e-06, "loss": 0.0002, "step": 18254 }, { "epoch": 2.19, "learning_rate": 1.689502099580084e-06, "loss": 0.0002, "step": 18255 }, { "epoch": 2.19, "learning_rate": 1.6894301139772047e-06, "loss": 0.0002, "step": 18256 }, { "epoch": 2.19, "learning_rate": 1.689358128374325e-06, "loss": 0.0009, "step": 18257 }, { "epoch": 2.19, "learning_rate": 1.6892861427714458e-06, "loss": 0.0006, "step": 18258 }, { "epoch": 2.19, "learning_rate": 1.6892141571685664e-06, "loss": 0.0013, "step": 18259 }, { "epoch": 2.19, "learning_rate": 1.689142171565687e-06, "loss": 0.0002, "step": 18260 }, { "epoch": 2.19, "learning_rate": 1.6890701859628075e-06, "loss": 0.0003, "step": 18261 }, { "epoch": 2.19, "learning_rate": 1.6889982003599279e-06, "loss": 0.0002, "step": 18262 }, { "epoch": 2.19, "learning_rate": 1.6889262147570487e-06, "loss": 0.0004, "step": 18263 }, { "epoch": 2.19, "learning_rate": 1.688854229154169e-06, "loss": 0.0002, "step": 18264 }, { "epoch": 2.19, "learning_rate": 1.6887822435512898e-06, "loss": 0.0002, "step": 18265 }, { "epoch": 2.19, "learning_rate": 1.6887102579484104e-06, "loss": 0.0002, "step": 18266 }, { "epoch": 2.19, "learning_rate": 1.688638272345531e-06, "loss": 0.0002, "step": 18267 }, { "epoch": 2.19, "learning_rate": 1.6885662867426515e-06, "loss": 0.0001, "step": 18268 }, { "epoch": 2.19, "learning_rate": 1.6884943011397723e-06, "loss": 0.0006, "step": 18269 }, { "epoch": 2.19, "learning_rate": 1.6884223155368926e-06, "loss": 0.0001, "step": 18270 }, { "epoch": 2.19, "learning_rate": 1.6883503299340134e-06, "loss": 0.0002, "step": 18271 }, { "epoch": 2.19, "learning_rate": 1.6882783443311338e-06, "loss": 0.0004, "step": 18272 }, { "epoch": 2.19, "learning_rate": 1.6882063587282543e-06, "loss": 0.0028, "step": 18273 }, { "epoch": 2.19, "learning_rate": 1.688134373125375e-06, "loss": 0.0002, "step": 18274 }, { "epoch": 2.19, "learning_rate": 1.6880623875224955e-06, "loss": 0.0001, "step": 18275 }, { "epoch": 2.19, "learning_rate": 1.6879904019196162e-06, "loss": 0.0003, "step": 18276 }, { "epoch": 2.19, "learning_rate": 1.6879184163167366e-06, "loss": 0.0002, "step": 18277 }, { "epoch": 2.19, "learning_rate": 1.6878464307138574e-06, "loss": 0.0002, "step": 18278 }, { "epoch": 2.19, "learning_rate": 1.6877744451109777e-06, "loss": 0.0001, "step": 18279 }, { "epoch": 2.19, "learning_rate": 1.6877024595080985e-06, "loss": 0.4725, "step": 18280 }, { "epoch": 2.19, "learning_rate": 1.6876304739052189e-06, "loss": 0.0001, "step": 18281 }, { "epoch": 2.19, "learning_rate": 1.6875584883023396e-06, "loss": 1.3502, "step": 18282 }, { "epoch": 2.19, "learning_rate": 1.6874865026994602e-06, "loss": 0.0003, "step": 18283 }, { "epoch": 2.19, "learning_rate": 1.6874145170965806e-06, "loss": 0.0001, "step": 18284 }, { "epoch": 2.19, "learning_rate": 1.6873425314937013e-06, "loss": 0.0002, "step": 18285 }, { "epoch": 2.19, "learning_rate": 1.6872705458908217e-06, "loss": 0.0002, "step": 18286 }, { "epoch": 2.19, "learning_rate": 1.6871985602879425e-06, "loss": 0.7745, "step": 18287 }, { "epoch": 2.19, "learning_rate": 1.6871265746850628e-06, "loss": 0.0002, "step": 18288 }, { "epoch": 2.19, "learning_rate": 1.6870545890821836e-06, "loss": 0.0001, "step": 18289 }, { "epoch": 2.19, "learning_rate": 1.6869826034793042e-06, "loss": 0.1247, "step": 18290 }, { "epoch": 2.19, "learning_rate": 1.686910617876425e-06, "loss": 0.0001, "step": 18291 }, { "epoch": 2.19, "learning_rate": 1.6868386322735453e-06, "loss": 0.0606, "step": 18292 }, { "epoch": 2.19, "learning_rate": 1.686766646670666e-06, "loss": 0.0003, "step": 18293 }, { "epoch": 2.19, "learning_rate": 1.6866946610677864e-06, "loss": 0.0001, "step": 18294 }, { "epoch": 2.19, "learning_rate": 1.6866226754649072e-06, "loss": 0.0534, "step": 18295 }, { "epoch": 2.19, "learning_rate": 1.6865506898620276e-06, "loss": 0.0003, "step": 18296 }, { "epoch": 2.19, "learning_rate": 1.6864787042591482e-06, "loss": 0.0001, "step": 18297 }, { "epoch": 2.19, "learning_rate": 1.686406718656269e-06, "loss": 0.0002, "step": 18298 }, { "epoch": 2.19, "learning_rate": 1.6863347330533893e-06, "loss": 0.0004, "step": 18299 }, { "epoch": 2.19, "learning_rate": 1.68626274745051e-06, "loss": 0.0003, "step": 18300 }, { "epoch": 2.19, "learning_rate": 1.6861907618476304e-06, "loss": 0.0001, "step": 18301 }, { "epoch": 2.19, "learning_rate": 1.6861187762447512e-06, "loss": 0.0002, "step": 18302 }, { "epoch": 2.19, "learning_rate": 1.6860467906418716e-06, "loss": 0.0002, "step": 18303 }, { "epoch": 2.19, "learning_rate": 1.6859748050389923e-06, "loss": 0.0002, "step": 18304 }, { "epoch": 2.19, "learning_rate": 1.685902819436113e-06, "loss": 0.0002, "step": 18305 }, { "epoch": 2.19, "learning_rate": 1.6858308338332335e-06, "loss": 0.0002, "step": 18306 }, { "epoch": 2.19, "learning_rate": 1.685758848230354e-06, "loss": 0.0002, "step": 18307 }, { "epoch": 2.19, "learning_rate": 1.6856868626274744e-06, "loss": 0.0002, "step": 18308 }, { "epoch": 2.19, "learning_rate": 1.6856148770245952e-06, "loss": 0.0002, "step": 18309 }, { "epoch": 2.19, "learning_rate": 1.6855428914217155e-06, "loss": 0.0002, "step": 18310 }, { "epoch": 2.19, "learning_rate": 1.6854709058188363e-06, "loss": 0.0001, "step": 18311 }, { "epoch": 2.19, "learning_rate": 1.6853989202159569e-06, "loss": 0.0001, "step": 18312 }, { "epoch": 2.19, "learning_rate": 1.6853269346130774e-06, "loss": 0.0002, "step": 18313 }, { "epoch": 2.19, "learning_rate": 1.685254949010198e-06, "loss": 0.0002, "step": 18314 }, { "epoch": 2.19, "learning_rate": 1.6851829634073188e-06, "loss": 0.0002, "step": 18315 }, { "epoch": 2.19, "learning_rate": 1.6851109778044391e-06, "loss": 0.0001, "step": 18316 }, { "epoch": 2.19, "learning_rate": 1.68503899220156e-06, "loss": 0.0358, "step": 18317 }, { "epoch": 2.2, "learning_rate": 1.6849670065986803e-06, "loss": 0.0001, "step": 18318 }, { "epoch": 2.2, "learning_rate": 1.6848950209958008e-06, "loss": 0.0001, "step": 18319 }, { "epoch": 2.2, "learning_rate": 1.6848230353929214e-06, "loss": 0.0002, "step": 18320 }, { "epoch": 2.2, "learning_rate": 1.684751049790042e-06, "loss": 0.0002, "step": 18321 }, { "epoch": 2.2, "learning_rate": 1.6846790641871627e-06, "loss": 0.0003, "step": 18322 }, { "epoch": 2.2, "learning_rate": 1.684607078584283e-06, "loss": 0.0004, "step": 18323 }, { "epoch": 2.2, "learning_rate": 1.6845350929814039e-06, "loss": 0.0004, "step": 18324 }, { "epoch": 2.2, "learning_rate": 1.6844631073785242e-06, "loss": 0.0001, "step": 18325 }, { "epoch": 2.2, "learning_rate": 1.684391121775645e-06, "loss": 0.0006, "step": 18326 }, { "epoch": 2.2, "learning_rate": 1.6843191361727654e-06, "loss": 0.0001, "step": 18327 }, { "epoch": 2.2, "learning_rate": 1.6842471505698861e-06, "loss": 0.0002, "step": 18328 }, { "epoch": 2.2, "learning_rate": 1.6841751649670067e-06, "loss": 0.0001, "step": 18329 }, { "epoch": 2.2, "learning_rate": 1.684103179364127e-06, "loss": 0.0002, "step": 18330 }, { "epoch": 2.2, "learning_rate": 1.6840311937612478e-06, "loss": 0.0002, "step": 18331 }, { "epoch": 2.2, "learning_rate": 1.6839592081583682e-06, "loss": 0.0003, "step": 18332 }, { "epoch": 2.2, "learning_rate": 1.683887222555489e-06, "loss": 0.0003, "step": 18333 }, { "epoch": 2.2, "learning_rate": 1.6838152369526093e-06, "loss": 0.0003, "step": 18334 }, { "epoch": 2.2, "learning_rate": 1.6837432513497301e-06, "loss": 0.0002, "step": 18335 }, { "epoch": 2.2, "learning_rate": 1.6836712657468507e-06, "loss": 0.0001, "step": 18336 }, { "epoch": 2.2, "learning_rate": 1.6835992801439715e-06, "loss": 0.0029, "step": 18337 }, { "epoch": 2.2, "learning_rate": 1.6835272945410918e-06, "loss": 0.0002, "step": 18338 }, { "epoch": 2.2, "learning_rate": 1.6834553089382126e-06, "loss": 0.0002, "step": 18339 }, { "epoch": 2.2, "learning_rate": 1.683383323335333e-06, "loss": 0.0006, "step": 18340 }, { "epoch": 2.2, "learning_rate": 1.6833113377324537e-06, "loss": 0.0002, "step": 18341 }, { "epoch": 2.2, "learning_rate": 1.683239352129574e-06, "loss": 0.0001, "step": 18342 }, { "epoch": 2.2, "learning_rate": 1.6831673665266946e-06, "loss": 0.0004, "step": 18343 }, { "epoch": 2.2, "learning_rate": 1.6830953809238152e-06, "loss": 0.4168, "step": 18344 }, { "epoch": 2.2, "learning_rate": 1.6830233953209358e-06, "loss": 0.0001, "step": 18345 }, { "epoch": 2.2, "learning_rate": 1.6829514097180566e-06, "loss": 0.0001, "step": 18346 }, { "epoch": 2.2, "learning_rate": 1.682879424115177e-06, "loss": 0.0002, "step": 18347 }, { "epoch": 2.2, "learning_rate": 1.6828074385122977e-06, "loss": 0.0001, "step": 18348 }, { "epoch": 2.2, "learning_rate": 1.682735452909418e-06, "loss": 0.0002, "step": 18349 }, { "epoch": 2.2, "learning_rate": 1.6826634673065388e-06, "loss": 0.0003, "step": 18350 }, { "epoch": 2.2, "learning_rate": 1.6825914817036592e-06, "loss": 0.0027, "step": 18351 }, { "epoch": 2.2, "learning_rate": 1.68251949610078e-06, "loss": 0.0002, "step": 18352 }, { "epoch": 2.2, "learning_rate": 1.6824475104979005e-06, "loss": 0.0002, "step": 18353 }, { "epoch": 2.2, "learning_rate": 1.6823755248950209e-06, "loss": 0.0002, "step": 18354 }, { "epoch": 2.2, "learning_rate": 1.6823035392921417e-06, "loss": 0.0002, "step": 18355 }, { "epoch": 2.2, "learning_rate": 1.682231553689262e-06, "loss": 0.001, "step": 18356 }, { "epoch": 2.2, "learning_rate": 1.6821595680863828e-06, "loss": 0.0002, "step": 18357 }, { "epoch": 2.2, "learning_rate": 1.6820875824835031e-06, "loss": 0.0001, "step": 18358 }, { "epoch": 2.2, "learning_rate": 1.682015596880624e-06, "loss": 0.2809, "step": 18359 }, { "epoch": 2.2, "learning_rate": 1.6819436112777445e-06, "loss": 0.0001, "step": 18360 }, { "epoch": 2.2, "learning_rate": 1.6818716256748653e-06, "loss": 0.0002, "step": 18361 }, { "epoch": 2.2, "learning_rate": 1.6817996400719856e-06, "loss": 0.0002, "step": 18362 }, { "epoch": 2.2, "learning_rate": 1.6817276544691064e-06, "loss": 0.0004, "step": 18363 }, { "epoch": 2.2, "learning_rate": 1.6816556688662268e-06, "loss": 0.0003, "step": 18364 }, { "epoch": 2.2, "learning_rate": 1.6815836832633471e-06, "loss": 0.0002, "step": 18365 }, { "epoch": 2.2, "learning_rate": 1.681511697660468e-06, "loss": 0.0002, "step": 18366 }, { "epoch": 2.2, "learning_rate": 1.6814397120575885e-06, "loss": 0.0002, "step": 18367 }, { "epoch": 2.2, "learning_rate": 1.6813677264547092e-06, "loss": 1.1669, "step": 18368 }, { "epoch": 2.2, "learning_rate": 1.6812957408518296e-06, "loss": 0.0002, "step": 18369 }, { "epoch": 2.2, "learning_rate": 1.6812237552489504e-06, "loss": 0.0002, "step": 18370 }, { "epoch": 2.2, "learning_rate": 1.6811517696460707e-06, "loss": 0.0002, "step": 18371 }, { "epoch": 2.2, "learning_rate": 1.6810797840431915e-06, "loss": 0.0002, "step": 18372 }, { "epoch": 2.2, "learning_rate": 1.6810077984403119e-06, "loss": 0.0001, "step": 18373 }, { "epoch": 2.2, "learning_rate": 1.6809358128374326e-06, "loss": 0.0002, "step": 18374 }, { "epoch": 2.2, "learning_rate": 1.6808638272345532e-06, "loss": 0.0071, "step": 18375 }, { "epoch": 2.2, "learning_rate": 1.6807918416316736e-06, "loss": 0.0003, "step": 18376 }, { "epoch": 2.2, "learning_rate": 1.6807198560287943e-06, "loss": 0.0003, "step": 18377 }, { "epoch": 2.2, "learning_rate": 1.6806478704259147e-06, "loss": 0.0003, "step": 18378 }, { "epoch": 2.2, "learning_rate": 1.6805758848230355e-06, "loss": 0.0002, "step": 18379 }, { "epoch": 2.2, "learning_rate": 1.6805038992201558e-06, "loss": 0.0001, "step": 18380 }, { "epoch": 2.2, "learning_rate": 1.6804319136172766e-06, "loss": 0.0005, "step": 18381 }, { "epoch": 2.2, "learning_rate": 1.6803599280143972e-06, "loss": 0.0001, "step": 18382 }, { "epoch": 2.2, "learning_rate": 1.6802879424115177e-06, "loss": 0.0002, "step": 18383 }, { "epoch": 2.2, "learning_rate": 1.6802159568086383e-06, "loss": 0.0002, "step": 18384 }, { "epoch": 2.2, "learning_rate": 1.680143971205759e-06, "loss": 0.0001, "step": 18385 }, { "epoch": 2.2, "learning_rate": 1.6800719856028794e-06, "loss": 0.0304, "step": 18386 }, { "epoch": 2.2, "learning_rate": 1.6800000000000002e-06, "loss": 0.0002, "step": 18387 }, { "epoch": 2.2, "learning_rate": 1.6799280143971206e-06, "loss": 0.0018, "step": 18388 }, { "epoch": 2.2, "learning_rate": 1.6798560287942411e-06, "loss": 0.0003, "step": 18389 }, { "epoch": 2.2, "learning_rate": 1.6797840431913617e-06, "loss": 0.0002, "step": 18390 }, { "epoch": 2.2, "learning_rate": 1.6797120575884823e-06, "loss": 0.0002, "step": 18391 }, { "epoch": 2.2, "learning_rate": 1.679640071985603e-06, "loss": 0.0002, "step": 18392 }, { "epoch": 2.2, "learning_rate": 1.6795680863827234e-06, "loss": 0.0002, "step": 18393 }, { "epoch": 2.2, "learning_rate": 1.6794961007798442e-06, "loss": 0.0003, "step": 18394 }, { "epoch": 2.2, "learning_rate": 1.6794241151769645e-06, "loss": 0.0001, "step": 18395 }, { "epoch": 2.2, "learning_rate": 1.6793521295740853e-06, "loss": 0.0002, "step": 18396 }, { "epoch": 2.2, "learning_rate": 1.6792801439712057e-06, "loss": 0.0004, "step": 18397 }, { "epoch": 2.2, "learning_rate": 1.6792081583683265e-06, "loss": 0.0002, "step": 18398 }, { "epoch": 2.2, "learning_rate": 1.679136172765447e-06, "loss": 0.0002, "step": 18399 }, { "epoch": 2.2, "learning_rate": 1.6790641871625674e-06, "loss": 0.0004, "step": 18400 }, { "epoch": 2.21, "learning_rate": 1.6789922015596882e-06, "loss": 0.0003, "step": 18401 }, { "epoch": 2.21, "learning_rate": 1.6789202159568085e-06, "loss": 0.0002, "step": 18402 }, { "epoch": 2.21, "learning_rate": 1.6788482303539293e-06, "loss": 0.0001, "step": 18403 }, { "epoch": 2.21, "learning_rate": 1.6787762447510496e-06, "loss": 0.0001, "step": 18404 }, { "epoch": 2.21, "learning_rate": 1.6787042591481704e-06, "loss": 0.0006, "step": 18405 }, { "epoch": 2.21, "learning_rate": 1.678632273545291e-06, "loss": 0.0002, "step": 18406 }, { "epoch": 2.21, "learning_rate": 1.6785602879424118e-06, "loss": 0.0002, "step": 18407 }, { "epoch": 2.21, "learning_rate": 1.6784883023395321e-06, "loss": 0.0003, "step": 18408 }, { "epoch": 2.21, "learning_rate": 1.678416316736653e-06, "loss": 0.0001, "step": 18409 }, { "epoch": 2.21, "learning_rate": 1.6783443311337733e-06, "loss": 0.0001, "step": 18410 }, { "epoch": 2.21, "learning_rate": 1.6782723455308936e-06, "loss": 0.0011, "step": 18411 }, { "epoch": 2.21, "learning_rate": 1.6782003599280144e-06, "loss": 0.0031, "step": 18412 }, { "epoch": 2.21, "learning_rate": 1.678128374325135e-06, "loss": 0.0002, "step": 18413 }, { "epoch": 2.21, "learning_rate": 1.6780563887222557e-06, "loss": 0.0002, "step": 18414 }, { "epoch": 2.21, "learning_rate": 1.677984403119376e-06, "loss": 0.0002, "step": 18415 }, { "epoch": 2.21, "learning_rate": 1.6779124175164969e-06, "loss": 0.0003, "step": 18416 }, { "epoch": 2.21, "learning_rate": 1.6778404319136172e-06, "loss": 0.0003, "step": 18417 }, { "epoch": 2.21, "learning_rate": 1.677768446310738e-06, "loss": 0.0001, "step": 18418 }, { "epoch": 2.21, "learning_rate": 1.6776964607078584e-06, "loss": 0.0003, "step": 18419 }, { "epoch": 2.21, "learning_rate": 1.6776244751049791e-06, "loss": 0.0002, "step": 18420 }, { "epoch": 2.21, "learning_rate": 1.6775524895020997e-06, "loss": 0.0004, "step": 18421 }, { "epoch": 2.21, "learning_rate": 1.6774805038992203e-06, "loss": 0.0001, "step": 18422 }, { "epoch": 2.21, "learning_rate": 1.6774085182963408e-06, "loss": 0.0003, "step": 18423 }, { "epoch": 2.21, "learning_rate": 1.6773365326934612e-06, "loss": 0.0001, "step": 18424 }, { "epoch": 2.21, "learning_rate": 1.677264547090582e-06, "loss": 0.0004, "step": 18425 }, { "epoch": 2.21, "learning_rate": 1.6771925614877023e-06, "loss": 0.0001, "step": 18426 }, { "epoch": 2.21, "learning_rate": 1.677120575884823e-06, "loss": 0.0003, "step": 18427 }, { "epoch": 2.21, "learning_rate": 1.6770485902819437e-06, "loss": 0.0002, "step": 18428 }, { "epoch": 2.21, "learning_rate": 1.6769766046790642e-06, "loss": 0.0001, "step": 18429 }, { "epoch": 2.21, "learning_rate": 1.6769046190761848e-06, "loss": 0.0002, "step": 18430 }, { "epoch": 2.21, "learning_rate": 1.6768326334733056e-06, "loss": 0.0002, "step": 18431 }, { "epoch": 2.21, "learning_rate": 1.676760647870426e-06, "loss": 0.0002, "step": 18432 }, { "epoch": 2.21, "learning_rate": 1.6766886622675467e-06, "loss": 0.0002, "step": 18433 }, { "epoch": 2.21, "learning_rate": 1.676616676664667e-06, "loss": 0.0001, "step": 18434 }, { "epoch": 2.21, "learning_rate": 1.6765446910617876e-06, "loss": 0.0002, "step": 18435 }, { "epoch": 2.21, "learning_rate": 1.6764727054589082e-06, "loss": 0.0003, "step": 18436 }, { "epoch": 2.21, "learning_rate": 1.6764007198560288e-06, "loss": 0.0002, "step": 18437 }, { "epoch": 2.21, "learning_rate": 1.6763287342531496e-06, "loss": 0.0002, "step": 18438 }, { "epoch": 2.21, "learning_rate": 1.67625674865027e-06, "loss": 0.0002, "step": 18439 }, { "epoch": 2.21, "learning_rate": 1.6761847630473907e-06, "loss": 0.0003, "step": 18440 }, { "epoch": 2.21, "learning_rate": 1.676112777444511e-06, "loss": 0.0001, "step": 18441 }, { "epoch": 2.21, "learning_rate": 1.6760407918416318e-06, "loss": 0.0002, "step": 18442 }, { "epoch": 2.21, "learning_rate": 1.6759688062387522e-06, "loss": 0.0001, "step": 18443 }, { "epoch": 2.21, "learning_rate": 1.675896820635873e-06, "loss": 0.0003, "step": 18444 }, { "epoch": 2.21, "learning_rate": 1.6758248350329935e-06, "loss": 0.0002, "step": 18445 }, { "epoch": 2.21, "learning_rate": 1.6757528494301139e-06, "loss": 0.0004, "step": 18446 }, { "epoch": 2.21, "learning_rate": 1.6756808638272347e-06, "loss": 0.0001, "step": 18447 }, { "epoch": 2.21, "learning_rate": 1.675608878224355e-06, "loss": 0.0001, "step": 18448 }, { "epoch": 2.21, "learning_rate": 1.6755368926214758e-06, "loss": 0.0004, "step": 18449 }, { "epoch": 2.21, "learning_rate": 1.6754649070185961e-06, "loss": 0.0003, "step": 18450 }, { "epoch": 2.21, "learning_rate": 1.675392921415717e-06, "loss": 0.0002, "step": 18451 }, { "epoch": 2.21, "learning_rate": 1.6753209358128375e-06, "loss": 0.0005, "step": 18452 }, { "epoch": 2.21, "learning_rate": 1.6752489502099583e-06, "loss": 0.1221, "step": 18453 }, { "epoch": 2.21, "learning_rate": 1.6751769646070786e-06, "loss": 0.0002, "step": 18454 }, { "epoch": 2.21, "learning_rate": 1.6751049790041994e-06, "loss": 0.0002, "step": 18455 }, { "epoch": 2.21, "learning_rate": 1.6750329934013198e-06, "loss": 0.0003, "step": 18456 }, { "epoch": 2.21, "learning_rate": 1.6749610077984401e-06, "loss": 0.0003, "step": 18457 }, { "epoch": 2.21, "learning_rate": 1.6748890221955609e-06, "loss": 0.0001, "step": 18458 }, { "epoch": 2.21, "learning_rate": 1.6748170365926815e-06, "loss": 0.0425, "step": 18459 }, { "epoch": 2.21, "learning_rate": 1.6747450509898022e-06, "loss": 0.0001, "step": 18460 }, { "epoch": 2.21, "learning_rate": 1.6746730653869226e-06, "loss": 0.0002, "step": 18461 }, { "epoch": 2.21, "learning_rate": 1.6746010797840434e-06, "loss": 0.0002, "step": 18462 }, { "epoch": 2.21, "learning_rate": 1.6745290941811637e-06, "loss": 0.0002, "step": 18463 }, { "epoch": 2.21, "learning_rate": 1.6744571085782845e-06, "loss": 0.0002, "step": 18464 }, { "epoch": 2.21, "learning_rate": 1.6743851229754049e-06, "loss": 0.0002, "step": 18465 }, { "epoch": 2.21, "learning_rate": 1.6743131373725256e-06, "loss": 0.0001, "step": 18466 }, { "epoch": 2.21, "learning_rate": 1.6742411517696462e-06, "loss": 0.0001, "step": 18467 }, { "epoch": 2.21, "learning_rate": 1.6741691661667668e-06, "loss": 0.0003, "step": 18468 }, { "epoch": 2.21, "learning_rate": 1.6740971805638873e-06, "loss": 0.0226, "step": 18469 }, { "epoch": 2.21, "learning_rate": 1.6740251949610077e-06, "loss": 0.0001, "step": 18470 }, { "epoch": 2.21, "learning_rate": 1.6739532093581285e-06, "loss": 0.9149, "step": 18471 }, { "epoch": 2.21, "learning_rate": 1.6738812237552488e-06, "loss": 0.0001, "step": 18472 }, { "epoch": 2.21, "learning_rate": 1.6738092381523696e-06, "loss": 0.0002, "step": 18473 }, { "epoch": 2.21, "learning_rate": 1.67373725254949e-06, "loss": 0.0003, "step": 18474 }, { "epoch": 2.21, "learning_rate": 1.6736652669466107e-06, "loss": 0.0005, "step": 18475 }, { "epoch": 2.21, "learning_rate": 1.6735932813437313e-06, "loss": 0.0001, "step": 18476 }, { "epoch": 2.21, "learning_rate": 1.673521295740852e-06, "loss": 0.0002, "step": 18477 }, { "epoch": 2.21, "learning_rate": 1.6734493101379724e-06, "loss": 0.0002, "step": 18478 }, { "epoch": 2.21, "learning_rate": 1.6733773245350932e-06, "loss": 0.0001, "step": 18479 }, { "epoch": 2.21, "learning_rate": 1.6733053389322136e-06, "loss": 0.0001, "step": 18480 }, { "epoch": 2.21, "learning_rate": 1.673233353329334e-06, "loss": 0.0002, "step": 18481 }, { "epoch": 2.21, "learning_rate": 1.6731613677264547e-06, "loss": 0.0001, "step": 18482 }, { "epoch": 2.21, "learning_rate": 1.6730893821235753e-06, "loss": 0.0038, "step": 18483 }, { "epoch": 2.21, "learning_rate": 1.673017396520696e-06, "loss": 0.0003, "step": 18484 }, { "epoch": 2.22, "learning_rate": 1.6729454109178164e-06, "loss": 0.0001, "step": 18485 }, { "epoch": 2.22, "learning_rate": 1.6728734253149372e-06, "loss": 0.0003, "step": 18486 }, { "epoch": 2.22, "learning_rate": 1.6728014397120575e-06, "loss": 0.0002, "step": 18487 }, { "epoch": 2.22, "learning_rate": 1.6727294541091783e-06, "loss": 0.0002, "step": 18488 }, { "epoch": 2.22, "learning_rate": 1.6726574685062987e-06, "loss": 0.0003, "step": 18489 }, { "epoch": 2.22, "learning_rate": 1.6725854829034194e-06, "loss": 0.0009, "step": 18490 }, { "epoch": 2.22, "learning_rate": 1.67251349730054e-06, "loss": 0.0002, "step": 18491 }, { "epoch": 2.22, "learning_rate": 1.6724415116976604e-06, "loss": 0.0002, "step": 18492 }, { "epoch": 2.22, "learning_rate": 1.6723695260947811e-06, "loss": 0.0004, "step": 18493 }, { "epoch": 2.22, "learning_rate": 1.6722975404919015e-06, "loss": 0.0001, "step": 18494 }, { "epoch": 2.22, "learning_rate": 1.6722255548890223e-06, "loss": 0.0002, "step": 18495 }, { "epoch": 2.22, "learning_rate": 1.6721535692861426e-06, "loss": 0.0002, "step": 18496 }, { "epoch": 2.22, "learning_rate": 1.6720815836832634e-06, "loss": 0.0004, "step": 18497 }, { "epoch": 2.22, "learning_rate": 1.672009598080384e-06, "loss": 0.0002, "step": 18498 }, { "epoch": 2.22, "learning_rate": 1.6719376124775045e-06, "loss": 0.5862, "step": 18499 }, { "epoch": 2.22, "learning_rate": 1.6718656268746251e-06, "loss": 0.0002, "step": 18500 }, { "epoch": 2.22, "learning_rate": 1.6717936412717459e-06, "loss": 0.0001, "step": 18501 }, { "epoch": 2.22, "learning_rate": 1.6717216556688662e-06, "loss": 0.0001, "step": 18502 }, { "epoch": 2.22, "learning_rate": 1.6716496700659866e-06, "loss": 0.0002, "step": 18503 }, { "epoch": 2.22, "learning_rate": 1.6715776844631074e-06, "loss": 0.0002, "step": 18504 }, { "epoch": 2.22, "learning_rate": 1.671505698860228e-06, "loss": 0.4599, "step": 18505 }, { "epoch": 2.22, "learning_rate": 1.6714337132573485e-06, "loss": 0.0001, "step": 18506 }, { "epoch": 2.22, "learning_rate": 1.671361727654469e-06, "loss": 0.0003, "step": 18507 }, { "epoch": 2.22, "learning_rate": 1.6712897420515899e-06, "loss": 0.0001, "step": 18508 }, { "epoch": 2.22, "learning_rate": 1.6712177564487102e-06, "loss": 0.0002, "step": 18509 }, { "epoch": 2.22, "learning_rate": 1.671145770845831e-06, "loss": 0.0002, "step": 18510 }, { "epoch": 2.22, "learning_rate": 1.6710737852429513e-06, "loss": 0.02, "step": 18511 }, { "epoch": 2.22, "learning_rate": 1.6710017996400721e-06, "loss": 0.0001, "step": 18512 }, { "epoch": 2.22, "learning_rate": 1.6709298140371925e-06, "loss": 0.0001, "step": 18513 }, { "epoch": 2.22, "learning_rate": 1.6708578284343133e-06, "loss": 0.0002, "step": 18514 }, { "epoch": 2.22, "learning_rate": 1.6707858428314338e-06, "loss": 0.0003, "step": 18515 }, { "epoch": 2.22, "learning_rate": 1.6707138572285542e-06, "loss": 0.0003, "step": 18516 }, { "epoch": 2.22, "learning_rate": 1.670641871625675e-06, "loss": 0.0002, "step": 18517 }, { "epoch": 2.22, "learning_rate": 1.6705698860227953e-06, "loss": 0.3319, "step": 18518 }, { "epoch": 2.22, "learning_rate": 1.670497900419916e-06, "loss": 0.0002, "step": 18519 }, { "epoch": 2.22, "learning_rate": 1.6704259148170365e-06, "loss": 0.0004, "step": 18520 }, { "epoch": 2.22, "learning_rate": 1.6703539292141572e-06, "loss": 0.0002, "step": 18521 }, { "epoch": 2.22, "learning_rate": 1.6702819436112778e-06, "loss": 0.0003, "step": 18522 }, { "epoch": 2.22, "learning_rate": 1.6702099580083986e-06, "loss": 0.0002, "step": 18523 }, { "epoch": 2.22, "learning_rate": 1.670137972405519e-06, "loss": 0.0004, "step": 18524 }, { "epoch": 2.22, "learning_rate": 1.6700659868026397e-06, "loss": 0.0002, "step": 18525 }, { "epoch": 2.22, "learning_rate": 1.66999400119976e-06, "loss": 0.5835, "step": 18526 }, { "epoch": 2.22, "learning_rate": 1.6699220155968804e-06, "loss": 0.0002, "step": 18527 }, { "epoch": 2.22, "learning_rate": 1.6698500299940012e-06, "loss": 0.0002, "step": 18528 }, { "epoch": 2.22, "learning_rate": 1.6697780443911218e-06, "loss": 0.0002, "step": 18529 }, { "epoch": 2.22, "learning_rate": 1.6697060587882425e-06, "loss": 0.0001, "step": 18530 }, { "epoch": 2.22, "learning_rate": 1.669634073185363e-06, "loss": 0.0006, "step": 18531 }, { "epoch": 2.22, "learning_rate": 1.6695620875824837e-06, "loss": 0.0002, "step": 18532 }, { "epoch": 2.22, "learning_rate": 1.669490101979604e-06, "loss": 0.0011, "step": 18533 }, { "epoch": 2.22, "learning_rate": 1.6694181163767248e-06, "loss": 0.0002, "step": 18534 }, { "epoch": 2.22, "learning_rate": 1.6693461307738452e-06, "loss": 0.0002, "step": 18535 }, { "epoch": 2.22, "learning_rate": 1.669274145170966e-06, "loss": 0.0002, "step": 18536 }, { "epoch": 2.22, "learning_rate": 1.6692021595680865e-06, "loss": 0.0001, "step": 18537 }, { "epoch": 2.22, "learning_rate": 1.6691301739652069e-06, "loss": 0.0001, "step": 18538 }, { "epoch": 2.22, "learning_rate": 1.6690581883623276e-06, "loss": 0.0002, "step": 18539 }, { "epoch": 2.22, "learning_rate": 1.668986202759448e-06, "loss": 0.0003, "step": 18540 }, { "epoch": 2.22, "learning_rate": 1.6689142171565688e-06, "loss": 0.0003, "step": 18541 }, { "epoch": 2.22, "learning_rate": 1.6688422315536891e-06, "loss": 0.0001, "step": 18542 }, { "epoch": 2.22, "learning_rate": 1.66877024595081e-06, "loss": 0.0009, "step": 18543 }, { "epoch": 2.22, "learning_rate": 1.6686982603479305e-06, "loss": 0.0002, "step": 18544 }, { "epoch": 2.22, "learning_rate": 1.668626274745051e-06, "loss": 0.0004, "step": 18545 }, { "epoch": 2.22, "learning_rate": 1.6685542891421716e-06, "loss": 0.0002, "step": 18546 }, { "epoch": 2.22, "learning_rate": 1.6684823035392924e-06, "loss": 0.0003, "step": 18547 }, { "epoch": 2.22, "learning_rate": 1.6684103179364127e-06, "loss": 0.0003, "step": 18548 }, { "epoch": 2.22, "learning_rate": 1.6683383323335335e-06, "loss": 0.0002, "step": 18549 }, { "epoch": 2.22, "learning_rate": 1.6682663467306539e-06, "loss": 0.0001, "step": 18550 }, { "epoch": 2.22, "learning_rate": 1.6681943611277744e-06, "loss": 0.0001, "step": 18551 }, { "epoch": 2.22, "learning_rate": 1.668122375524895e-06, "loss": 0.0001, "step": 18552 }, { "epoch": 2.22, "learning_rate": 1.6680503899220156e-06, "loss": 0.0002, "step": 18553 }, { "epoch": 2.22, "learning_rate": 1.6679784043191364e-06, "loss": 0.3955, "step": 18554 }, { "epoch": 2.22, "learning_rate": 1.6679064187162567e-06, "loss": 0.0007, "step": 18555 }, { "epoch": 2.22, "learning_rate": 1.6678344331133775e-06, "loss": 0.0002, "step": 18556 }, { "epoch": 2.22, "learning_rate": 1.6677624475104978e-06, "loss": 0.0002, "step": 18557 }, { "epoch": 2.22, "learning_rate": 1.6676904619076186e-06, "loss": 0.0001, "step": 18558 }, { "epoch": 2.22, "learning_rate": 1.667618476304739e-06, "loss": 0.0001, "step": 18559 }, { "epoch": 2.22, "learning_rate": 1.6675464907018598e-06, "loss": 0.0002, "step": 18560 }, { "epoch": 2.22, "learning_rate": 1.6674745050989803e-06, "loss": 0.0001, "step": 18561 }, { "epoch": 2.22, "learning_rate": 1.6674025194961007e-06, "loss": 0.0002, "step": 18562 }, { "epoch": 2.22, "learning_rate": 1.6673305338932215e-06, "loss": 0.0002, "step": 18563 }, { "epoch": 2.22, "learning_rate": 1.6672585482903418e-06, "loss": 0.0002, "step": 18564 }, { "epoch": 2.22, "learning_rate": 1.6671865626874626e-06, "loss": 0.0002, "step": 18565 }, { "epoch": 2.22, "learning_rate": 1.667114577084583e-06, "loss": 0.0003, "step": 18566 }, { "epoch": 2.22, "learning_rate": 1.6670425914817037e-06, "loss": 0.0001, "step": 18567 }, { "epoch": 2.23, "learning_rate": 1.6669706058788243e-06, "loss": 0.0002, "step": 18568 }, { "epoch": 2.23, "learning_rate": 1.666898620275945e-06, "loss": 0.0001, "step": 18569 }, { "epoch": 2.23, "learning_rate": 1.6668266346730654e-06, "loss": 0.0001, "step": 18570 }, { "epoch": 2.23, "learning_rate": 1.6667546490701862e-06, "loss": 0.0003, "step": 18571 }, { "epoch": 2.23, "learning_rate": 1.6666826634673066e-06, "loss": 0.0001, "step": 18572 }, { "epoch": 2.23, "learning_rate": 1.666610677864427e-06, "loss": 0.0002, "step": 18573 }, { "epoch": 2.23, "learning_rate": 1.6665386922615477e-06, "loss": 0.0002, "step": 18574 }, { "epoch": 2.23, "learning_rate": 1.6664667066586683e-06, "loss": 0.0002, "step": 18575 }, { "epoch": 2.23, "learning_rate": 1.666394721055789e-06, "loss": 0.0002, "step": 18576 }, { "epoch": 2.23, "learning_rate": 1.6663227354529094e-06, "loss": 0.0001, "step": 18577 }, { "epoch": 2.23, "learning_rate": 1.6662507498500302e-06, "loss": 0.0001, "step": 18578 }, { "epoch": 2.23, "learning_rate": 1.6661787642471505e-06, "loss": 0.0001, "step": 18579 }, { "epoch": 2.23, "learning_rate": 1.6661067786442713e-06, "loss": 0.0003, "step": 18580 }, { "epoch": 2.23, "learning_rate": 1.6660347930413917e-06, "loss": 0.0003, "step": 18581 }, { "epoch": 2.23, "learning_rate": 1.6659628074385124e-06, "loss": 0.0002, "step": 18582 }, { "epoch": 2.23, "learning_rate": 1.665890821835633e-06, "loss": 0.0002, "step": 18583 }, { "epoch": 2.23, "learning_rate": 1.6658188362327534e-06, "loss": 0.0002, "step": 18584 }, { "epoch": 2.23, "learning_rate": 1.6657468506298741e-06, "loss": 0.0001, "step": 18585 }, { "epoch": 2.23, "learning_rate": 1.6656748650269945e-06, "loss": 0.0002, "step": 18586 }, { "epoch": 2.23, "learning_rate": 1.6656028794241153e-06, "loss": 0.0004, "step": 18587 }, { "epoch": 2.23, "learning_rate": 1.6655308938212356e-06, "loss": 0.0003, "step": 18588 }, { "epoch": 2.23, "learning_rate": 1.6654589082183564e-06, "loss": 0.0001, "step": 18589 }, { "epoch": 2.23, "learning_rate": 1.665386922615477e-06, "loss": 0.0005, "step": 18590 }, { "epoch": 2.23, "learning_rate": 1.6653149370125975e-06, "loss": 0.0109, "step": 18591 }, { "epoch": 2.23, "learning_rate": 1.6652429514097181e-06, "loss": 0.0001, "step": 18592 }, { "epoch": 2.23, "learning_rate": 1.6651709658068389e-06, "loss": 0.0001, "step": 18593 }, { "epoch": 2.23, "learning_rate": 1.6650989802039592e-06, "loss": 0.0001, "step": 18594 }, { "epoch": 2.23, "learning_rate": 1.66502699460108e-06, "loss": 0.0002, "step": 18595 }, { "epoch": 2.23, "learning_rate": 1.6649550089982004e-06, "loss": 0.0022, "step": 18596 }, { "epoch": 2.23, "learning_rate": 1.664883023395321e-06, "loss": 0.0003, "step": 18597 }, { "epoch": 2.23, "learning_rate": 1.6648110377924415e-06, "loss": 0.0002, "step": 18598 }, { "epoch": 2.23, "learning_rate": 1.664739052189562e-06, "loss": 0.0006, "step": 18599 }, { "epoch": 2.23, "learning_rate": 1.6646670665866829e-06, "loss": 0.0004, "step": 18600 }, { "epoch": 2.23, "learning_rate": 1.6645950809838032e-06, "loss": 0.0002, "step": 18601 }, { "epoch": 2.23, "learning_rate": 1.664523095380924e-06, "loss": 0.0004, "step": 18602 }, { "epoch": 2.23, "learning_rate": 1.6644511097780443e-06, "loss": 0.0001, "step": 18603 }, { "epoch": 2.23, "learning_rate": 1.6643791241751651e-06, "loss": 0.0001, "step": 18604 }, { "epoch": 2.23, "learning_rate": 1.6643071385722855e-06, "loss": 0.0002, "step": 18605 }, { "epoch": 2.23, "learning_rate": 1.6642351529694063e-06, "loss": 0.0004, "step": 18606 }, { "epoch": 2.23, "learning_rate": 1.6641631673665268e-06, "loss": 0.0002, "step": 18607 }, { "epoch": 2.23, "learning_rate": 1.6640911817636472e-06, "loss": 0.0001, "step": 18608 }, { "epoch": 2.23, "learning_rate": 1.664019196160768e-06, "loss": 0.0001, "step": 18609 }, { "epoch": 2.23, "learning_rate": 1.6639472105578883e-06, "loss": 0.4142, "step": 18610 }, { "epoch": 2.23, "learning_rate": 1.663875224955009e-06, "loss": 0.0067, "step": 18611 }, { "epoch": 2.23, "learning_rate": 1.6638032393521294e-06, "loss": 0.0001, "step": 18612 }, { "epoch": 2.23, "learning_rate": 1.6637312537492502e-06, "loss": 0.0002, "step": 18613 }, { "epoch": 2.23, "learning_rate": 1.6636592681463708e-06, "loss": 0.0001, "step": 18614 }, { "epoch": 2.23, "learning_rate": 1.6635872825434916e-06, "loss": 0.0002, "step": 18615 }, { "epoch": 2.23, "learning_rate": 1.663515296940612e-06, "loss": 0.0003, "step": 18616 }, { "epoch": 2.23, "learning_rate": 1.6634433113377327e-06, "loss": 0.0002, "step": 18617 }, { "epoch": 2.23, "learning_rate": 1.663371325734853e-06, "loss": 0.0002, "step": 18618 }, { "epoch": 2.23, "learning_rate": 1.6632993401319734e-06, "loss": 0.0001, "step": 18619 }, { "epoch": 2.23, "learning_rate": 1.6632273545290942e-06, "loss": 0.0002, "step": 18620 }, { "epoch": 2.23, "learning_rate": 1.6631553689262148e-06, "loss": 0.0001, "step": 18621 }, { "epoch": 2.23, "learning_rate": 1.6630833833233353e-06, "loss": 0.0001, "step": 18622 }, { "epoch": 2.23, "learning_rate": 1.6630113977204559e-06, "loss": 0.0003, "step": 18623 }, { "epoch": 2.23, "learning_rate": 1.6629394121175767e-06, "loss": 0.0001, "step": 18624 }, { "epoch": 2.23, "learning_rate": 1.662867426514697e-06, "loss": 0.0002, "step": 18625 }, { "epoch": 2.23, "learning_rate": 1.6627954409118178e-06, "loss": 0.0002, "step": 18626 }, { "epoch": 2.23, "learning_rate": 1.6627234553089382e-06, "loss": 0.0002, "step": 18627 }, { "epoch": 2.23, "learning_rate": 1.662651469706059e-06, "loss": 0.0001, "step": 18628 }, { "epoch": 2.23, "learning_rate": 1.6625794841031793e-06, "loss": 0.0001, "step": 18629 }, { "epoch": 2.23, "learning_rate": 1.6625074985002999e-06, "loss": 0.0002, "step": 18630 }, { "epoch": 2.23, "learning_rate": 1.6624355128974206e-06, "loss": 0.0001, "step": 18631 }, { "epoch": 2.23, "learning_rate": 1.662363527294541e-06, "loss": 0.0001, "step": 18632 }, { "epoch": 2.23, "learning_rate": 1.6622915416916618e-06, "loss": 0.0001, "step": 18633 }, { "epoch": 2.23, "learning_rate": 1.6622195560887821e-06, "loss": 0.0002, "step": 18634 }, { "epoch": 2.23, "learning_rate": 1.662147570485903e-06, "loss": 0.0001, "step": 18635 }, { "epoch": 2.23, "learning_rate": 1.6620755848830233e-06, "loss": 0.0002, "step": 18636 }, { "epoch": 2.23, "learning_rate": 1.662003599280144e-06, "loss": 0.0003, "step": 18637 }, { "epoch": 2.23, "learning_rate": 1.6619316136772646e-06, "loss": 0.0002, "step": 18638 }, { "epoch": 2.23, "learning_rate": 1.6618596280743854e-06, "loss": 0.0001, "step": 18639 }, { "epoch": 2.23, "learning_rate": 1.6617876424715057e-06, "loss": 0.0001, "step": 18640 }, { "epoch": 2.23, "learning_rate": 1.6617156568686265e-06, "loss": 0.1454, "step": 18641 }, { "epoch": 2.23, "learning_rate": 1.6616436712657469e-06, "loss": 0.0001, "step": 18642 }, { "epoch": 2.23, "learning_rate": 1.6615716856628672e-06, "loss": 0.0002, "step": 18643 }, { "epoch": 2.23, "learning_rate": 1.661499700059988e-06, "loss": 0.0029, "step": 18644 }, { "epoch": 2.23, "learning_rate": 1.6614277144571086e-06, "loss": 0.0002, "step": 18645 }, { "epoch": 2.23, "learning_rate": 1.6613557288542293e-06, "loss": 0.6022, "step": 18646 }, { "epoch": 2.23, "learning_rate": 1.6612837432513497e-06, "loss": 0.0002, "step": 18647 }, { "epoch": 2.23, "learning_rate": 1.6612117576484705e-06, "loss": 0.0002, "step": 18648 }, { "epoch": 2.23, "learning_rate": 1.6611397720455908e-06, "loss": 0.0001, "step": 18649 }, { "epoch": 2.23, "learning_rate": 1.6610677864427116e-06, "loss": 0.0002, "step": 18650 }, { "epoch": 2.23, "learning_rate": 1.660995800839832e-06, "loss": 0.0002, "step": 18651 }, { "epoch": 2.24, "learning_rate": 1.6609238152369527e-06, "loss": 0.0001, "step": 18652 }, { "epoch": 2.24, "learning_rate": 1.6608518296340733e-06, "loss": 0.0002, "step": 18653 }, { "epoch": 2.24, "learning_rate": 1.6607798440311937e-06, "loss": 0.0001, "step": 18654 }, { "epoch": 2.24, "learning_rate": 1.6607078584283145e-06, "loss": 0.0002, "step": 18655 }, { "epoch": 2.24, "learning_rate": 1.6606358728254348e-06, "loss": 0.0002, "step": 18656 }, { "epoch": 2.24, "learning_rate": 1.6605638872225556e-06, "loss": 0.0693, "step": 18657 }, { "epoch": 2.24, "learning_rate": 1.660491901619676e-06, "loss": 0.0002, "step": 18658 }, { "epoch": 2.24, "learning_rate": 1.6604199160167967e-06, "loss": 0.0002, "step": 18659 }, { "epoch": 2.24, "learning_rate": 1.6603479304139173e-06, "loss": 0.0001, "step": 18660 }, { "epoch": 2.24, "learning_rate": 1.6602759448110379e-06, "loss": 0.0052, "step": 18661 }, { "epoch": 2.24, "learning_rate": 1.6602039592081584e-06, "loss": 0.0005, "step": 18662 }, { "epoch": 2.24, "learning_rate": 1.6601319736052792e-06, "loss": 0.0002, "step": 18663 }, { "epoch": 2.24, "learning_rate": 1.6600599880023996e-06, "loss": 0.0001, "step": 18664 }, { "epoch": 2.24, "learning_rate": 1.65998800239952e-06, "loss": 0.0001, "step": 18665 }, { "epoch": 2.24, "learning_rate": 1.6599160167966407e-06, "loss": 0.0004, "step": 18666 }, { "epoch": 2.24, "learning_rate": 1.6598440311937613e-06, "loss": 0.0001, "step": 18667 }, { "epoch": 2.24, "learning_rate": 1.6597720455908818e-06, "loss": 0.0001, "step": 18668 }, { "epoch": 2.24, "learning_rate": 1.6597000599880024e-06, "loss": 0.0001, "step": 18669 }, { "epoch": 2.24, "learning_rate": 1.6596280743851232e-06, "loss": 0.0001, "step": 18670 }, { "epoch": 2.24, "learning_rate": 1.6595560887822435e-06, "loss": 0.0001, "step": 18671 }, { "epoch": 2.24, "learning_rate": 1.6594841031793643e-06, "loss": 0.0002, "step": 18672 }, { "epoch": 2.24, "learning_rate": 1.6594121175764847e-06, "loss": 0.0002, "step": 18673 }, { "epoch": 2.24, "learning_rate": 1.6593401319736054e-06, "loss": 0.0002, "step": 18674 }, { "epoch": 2.24, "learning_rate": 1.6592681463707258e-06, "loss": 0.0002, "step": 18675 }, { "epoch": 2.24, "learning_rate": 1.6591961607678466e-06, "loss": 0.393, "step": 18676 }, { "epoch": 2.24, "learning_rate": 1.6591241751649671e-06, "loss": 0.0001, "step": 18677 }, { "epoch": 2.24, "learning_rate": 1.6590521895620875e-06, "loss": 0.0002, "step": 18678 }, { "epoch": 2.24, "learning_rate": 1.6589802039592083e-06, "loss": 0.0001, "step": 18679 }, { "epoch": 2.24, "learning_rate": 1.6589082183563286e-06, "loss": 0.0001, "step": 18680 }, { "epoch": 2.24, "learning_rate": 1.6588362327534494e-06, "loss": 0.0002, "step": 18681 }, { "epoch": 2.24, "learning_rate": 1.6587642471505698e-06, "loss": 0.0002, "step": 18682 }, { "epoch": 2.24, "learning_rate": 1.6586922615476905e-06, "loss": 0.0001, "step": 18683 }, { "epoch": 2.24, "learning_rate": 1.658620275944811e-06, "loss": 0.0001, "step": 18684 }, { "epoch": 2.24, "learning_rate": 1.6585482903419319e-06, "loss": 0.0001, "step": 18685 }, { "epoch": 2.24, "learning_rate": 1.6584763047390522e-06, "loss": 0.0002, "step": 18686 }, { "epoch": 2.24, "learning_rate": 1.658404319136173e-06, "loss": 0.3008, "step": 18687 }, { "epoch": 2.24, "learning_rate": 1.6583323335332934e-06, "loss": 0.0001, "step": 18688 }, { "epoch": 2.24, "learning_rate": 1.6582603479304137e-06, "loss": 0.0456, "step": 18689 }, { "epoch": 2.24, "learning_rate": 1.6581883623275345e-06, "loss": 0.0001, "step": 18690 }, { "epoch": 2.24, "learning_rate": 1.658116376724655e-06, "loss": 0.0002, "step": 18691 }, { "epoch": 2.24, "learning_rate": 1.6580443911217758e-06, "loss": 0.0002, "step": 18692 }, { "epoch": 2.24, "learning_rate": 1.6579724055188962e-06, "loss": 0.0001, "step": 18693 }, { "epoch": 2.24, "learning_rate": 1.657900419916017e-06, "loss": 0.0002, "step": 18694 }, { "epoch": 2.24, "learning_rate": 1.6578284343131373e-06, "loss": 0.0002, "step": 18695 }, { "epoch": 2.24, "learning_rate": 1.6577564487102581e-06, "loss": 0.0126, "step": 18696 }, { "epoch": 2.24, "learning_rate": 1.6576844631073785e-06, "loss": 0.0001, "step": 18697 }, { "epoch": 2.24, "learning_rate": 1.6576124775044992e-06, "loss": 0.0001, "step": 18698 }, { "epoch": 2.24, "learning_rate": 1.6575404919016198e-06, "loss": 0.0001, "step": 18699 }, { "epoch": 2.24, "learning_rate": 1.6574685062987402e-06, "loss": 0.0002, "step": 18700 }, { "epoch": 2.24, "learning_rate": 1.657396520695861e-06, "loss": 0.0001, "step": 18701 }, { "epoch": 2.24, "learning_rate": 1.6573245350929813e-06, "loss": 0.0002, "step": 18702 }, { "epoch": 2.24, "learning_rate": 1.657252549490102e-06, "loss": 0.0003, "step": 18703 }, { "epoch": 2.24, "learning_rate": 1.6571805638872224e-06, "loss": 0.0002, "step": 18704 }, { "epoch": 2.24, "learning_rate": 1.6571085782843432e-06, "loss": 0.0001, "step": 18705 }, { "epoch": 2.24, "learning_rate": 1.6570365926814638e-06, "loss": 0.0002, "step": 18706 }, { "epoch": 2.24, "learning_rate": 1.6569646070785843e-06, "loss": 0.0001, "step": 18707 }, { "epoch": 2.24, "learning_rate": 1.656892621475705e-06, "loss": 0.0002, "step": 18708 }, { "epoch": 2.24, "learning_rate": 1.6568206358728257e-06, "loss": 0.0001, "step": 18709 }, { "epoch": 2.24, "learning_rate": 1.656748650269946e-06, "loss": 0.0002, "step": 18710 }, { "epoch": 2.24, "learning_rate": 1.6566766646670664e-06, "loss": 0.0001, "step": 18711 }, { "epoch": 2.24, "learning_rate": 1.6566046790641872e-06, "loss": 0.0001, "step": 18712 }, { "epoch": 2.24, "learning_rate": 1.6565326934613077e-06, "loss": 0.0001, "step": 18713 }, { "epoch": 2.24, "learning_rate": 1.6564607078584283e-06, "loss": 0.0001, "step": 18714 }, { "epoch": 2.24, "learning_rate": 1.6563887222555489e-06, "loss": 0.0002, "step": 18715 }, { "epoch": 2.24, "learning_rate": 1.6563167366526697e-06, "loss": 0.0001, "step": 18716 }, { "epoch": 2.24, "learning_rate": 1.65624475104979e-06, "loss": 0.0002, "step": 18717 }, { "epoch": 2.24, "learning_rate": 1.6561727654469108e-06, "loss": 0.0001, "step": 18718 }, { "epoch": 2.24, "learning_rate": 1.6561007798440311e-06, "loss": 0.0003, "step": 18719 }, { "epoch": 2.24, "learning_rate": 1.656028794241152e-06, "loss": 0.0002, "step": 18720 }, { "epoch": 2.24, "learning_rate": 1.6559568086382723e-06, "loss": 0.0002, "step": 18721 }, { "epoch": 2.24, "learning_rate": 1.655884823035393e-06, "loss": 0.0002, "step": 18722 }, { "epoch": 2.24, "learning_rate": 1.6558128374325136e-06, "loss": 0.0001, "step": 18723 }, { "epoch": 2.24, "learning_rate": 1.655740851829634e-06, "loss": 0.0003, "step": 18724 }, { "epoch": 2.24, "learning_rate": 1.6556688662267548e-06, "loss": 0.8572, "step": 18725 }, { "epoch": 2.24, "learning_rate": 1.6555968806238751e-06, "loss": 0.0001, "step": 18726 }, { "epoch": 2.24, "learning_rate": 1.655524895020996e-06, "loss": 0.0002, "step": 18727 }, { "epoch": 2.24, "learning_rate": 1.6554529094181162e-06, "loss": 0.0003, "step": 18728 }, { "epoch": 2.24, "learning_rate": 1.655380923815237e-06, "loss": 0.0001, "step": 18729 }, { "epoch": 2.24, "learning_rate": 1.6553089382123576e-06, "loss": 0.0002, "step": 18730 }, { "epoch": 2.24, "learning_rate": 1.6552369526094784e-06, "loss": 0.0001, "step": 18731 }, { "epoch": 2.24, "learning_rate": 1.6551649670065987e-06, "loss": 0.0001, "step": 18732 }, { "epoch": 2.24, "learning_rate": 1.6550929814037195e-06, "loss": 0.0002, "step": 18733 }, { "epoch": 2.24, "learning_rate": 1.6550209958008399e-06, "loss": 0.0001, "step": 18734 }, { "epoch": 2.25, "learning_rate": 1.6549490101979602e-06, "loss": 0.0003, "step": 18735 }, { "epoch": 2.25, "learning_rate": 1.654877024595081e-06, "loss": 0.0002, "step": 18736 }, { "epoch": 2.25, "learning_rate": 1.6548050389922016e-06, "loss": 0.0002, "step": 18737 }, { "epoch": 2.25, "learning_rate": 1.6547330533893223e-06, "loss": 0.0001, "step": 18738 }, { "epoch": 2.25, "learning_rate": 1.6546610677864427e-06, "loss": 0.0002, "step": 18739 }, { "epoch": 2.25, "learning_rate": 1.6545890821835635e-06, "loss": 0.0002, "step": 18740 }, { "epoch": 2.25, "learning_rate": 1.6545170965806838e-06, "loss": 0.0005, "step": 18741 }, { "epoch": 2.25, "learning_rate": 1.6544451109778046e-06, "loss": 0.0004, "step": 18742 }, { "epoch": 2.25, "learning_rate": 1.654373125374925e-06, "loss": 0.0001, "step": 18743 }, { "epoch": 2.25, "learning_rate": 1.6543011397720457e-06, "loss": 0.0002, "step": 18744 }, { "epoch": 2.25, "learning_rate": 1.654229154169166e-06, "loss": 0.0002, "step": 18745 }, { "epoch": 2.25, "learning_rate": 1.6541571685662867e-06, "loss": 0.0001, "step": 18746 }, { "epoch": 2.25, "learning_rate": 1.6540851829634074e-06, "loss": 0.0001, "step": 18747 }, { "epoch": 2.25, "learning_rate": 1.6540131973605278e-06, "loss": 0.0002, "step": 18748 }, { "epoch": 2.25, "learning_rate": 1.6539412117576486e-06, "loss": 0.8341, "step": 18749 }, { "epoch": 2.25, "learning_rate": 1.653869226154769e-06, "loss": 0.0001, "step": 18750 }, { "epoch": 2.25, "learning_rate": 1.6537972405518897e-06, "loss": 0.0001, "step": 18751 }, { "epoch": 2.25, "learning_rate": 1.65372525494901e-06, "loss": 0.0002, "step": 18752 }, { "epoch": 2.25, "learning_rate": 1.6536532693461308e-06, "loss": 0.0001, "step": 18753 }, { "epoch": 2.25, "learning_rate": 1.6535812837432514e-06, "loss": 0.0002, "step": 18754 }, { "epoch": 2.25, "learning_rate": 1.6535092981403722e-06, "loss": 0.0001, "step": 18755 }, { "epoch": 2.25, "learning_rate": 1.6534373125374925e-06, "loss": 0.0001, "step": 18756 }, { "epoch": 2.25, "learning_rate": 1.653365326934613e-06, "loss": 0.0001, "step": 18757 }, { "epoch": 2.25, "learning_rate": 1.6532933413317337e-06, "loss": 0.0001, "step": 18758 }, { "epoch": 2.25, "learning_rate": 1.653221355728854e-06, "loss": 0.0002, "step": 18759 }, { "epoch": 2.25, "learning_rate": 1.6531493701259748e-06, "loss": 0.0001, "step": 18760 }, { "epoch": 2.25, "learning_rate": 1.6530773845230954e-06, "loss": 0.0001, "step": 18761 }, { "epoch": 2.25, "learning_rate": 1.6530053989202162e-06, "loss": 0.9209, "step": 18762 }, { "epoch": 2.25, "learning_rate": 1.6529334133173365e-06, "loss": 0.0001, "step": 18763 }, { "epoch": 2.25, "learning_rate": 1.6528614277144573e-06, "loss": 0.0029, "step": 18764 }, { "epoch": 2.25, "learning_rate": 1.6527894421115776e-06, "loss": 0.0001, "step": 18765 }, { "epoch": 2.25, "learning_rate": 1.6527174565086984e-06, "loss": 0.0001, "step": 18766 }, { "epoch": 2.25, "learning_rate": 1.6526454709058188e-06, "loss": 0.0002, "step": 18767 }, { "epoch": 2.25, "learning_rate": 1.6525734853029396e-06, "loss": 0.0001, "step": 18768 }, { "epoch": 2.25, "learning_rate": 1.6525014997000601e-06, "loss": 0.0001, "step": 18769 }, { "epoch": 2.25, "learning_rate": 1.6524295140971805e-06, "loss": 0.0002, "step": 18770 }, { "epoch": 2.25, "learning_rate": 1.6523575284943013e-06, "loss": 0.0001, "step": 18771 }, { "epoch": 2.25, "learning_rate": 1.6522855428914216e-06, "loss": 0.0002, "step": 18772 }, { "epoch": 2.25, "learning_rate": 1.6522135572885424e-06, "loss": 0.0002, "step": 18773 }, { "epoch": 2.25, "learning_rate": 1.6521415716856627e-06, "loss": 0.0002, "step": 18774 }, { "epoch": 2.25, "learning_rate": 1.6520695860827835e-06, "loss": 0.0001, "step": 18775 }, { "epoch": 2.25, "learning_rate": 1.651997600479904e-06, "loss": 0.0002, "step": 18776 }, { "epoch": 2.25, "learning_rate": 1.6519256148770247e-06, "loss": 0.0002, "step": 18777 }, { "epoch": 2.25, "learning_rate": 1.6518536292741452e-06, "loss": 0.0002, "step": 18778 }, { "epoch": 2.25, "learning_rate": 1.651781643671266e-06, "loss": 0.0001, "step": 18779 }, { "epoch": 2.25, "learning_rate": 1.6517096580683864e-06, "loss": 0.0003, "step": 18780 }, { "epoch": 2.25, "learning_rate": 1.6516376724655067e-06, "loss": 0.0001, "step": 18781 }, { "epoch": 2.25, "learning_rate": 1.6515656868626275e-06, "loss": 0.0002, "step": 18782 }, { "epoch": 2.25, "learning_rate": 1.651493701259748e-06, "loss": 0.0001, "step": 18783 }, { "epoch": 2.25, "learning_rate": 1.6514217156568686e-06, "loss": 0.0002, "step": 18784 }, { "epoch": 2.25, "learning_rate": 1.6513497300539892e-06, "loss": 0.0002, "step": 18785 }, { "epoch": 2.25, "learning_rate": 1.65127774445111e-06, "loss": 0.0002, "step": 18786 }, { "epoch": 2.25, "learning_rate": 1.6512057588482303e-06, "loss": 0.0001, "step": 18787 }, { "epoch": 2.25, "learning_rate": 1.651133773245351e-06, "loss": 0.0002, "step": 18788 }, { "epoch": 2.25, "learning_rate": 1.6510617876424715e-06, "loss": 0.0001, "step": 18789 }, { "epoch": 2.25, "learning_rate": 1.6509898020395922e-06, "loss": 0.0001, "step": 18790 }, { "epoch": 2.25, "learning_rate": 1.6509178164367126e-06, "loss": 0.0001, "step": 18791 }, { "epoch": 2.25, "learning_rate": 1.6508458308338332e-06, "loss": 0.0002, "step": 18792 }, { "epoch": 2.25, "learning_rate": 1.650773845230954e-06, "loss": 0.0006, "step": 18793 }, { "epoch": 2.25, "learning_rate": 1.6507018596280743e-06, "loss": 0.001, "step": 18794 }, { "epoch": 2.25, "learning_rate": 1.650629874025195e-06, "loss": 0.0001, "step": 18795 }, { "epoch": 2.25, "learning_rate": 1.6505578884223154e-06, "loss": 0.0001, "step": 18796 }, { "epoch": 2.25, "learning_rate": 1.6504859028194362e-06, "loss": 0.0001, "step": 18797 }, { "epoch": 2.25, "learning_rate": 1.6504139172165566e-06, "loss": 0.0003, "step": 18798 }, { "epoch": 2.25, "learning_rate": 1.6503419316136773e-06, "loss": 0.0002, "step": 18799 }, { "epoch": 2.25, "learning_rate": 1.650269946010798e-06, "loss": 0.0002, "step": 18800 }, { "epoch": 2.25, "learning_rate": 1.6501979604079187e-06, "loss": 0.0002, "step": 18801 }, { "epoch": 2.25, "learning_rate": 1.650125974805039e-06, "loss": 0.0002, "step": 18802 }, { "epoch": 2.25, "learning_rate": 1.6500539892021598e-06, "loss": 0.0001, "step": 18803 }, { "epoch": 2.25, "learning_rate": 1.6499820035992802e-06, "loss": 0.0001, "step": 18804 }, { "epoch": 2.25, "learning_rate": 1.6499100179964005e-06, "loss": 0.0001, "step": 18805 }, { "epoch": 2.25, "learning_rate": 1.6498380323935213e-06, "loss": 0.0004, "step": 18806 }, { "epoch": 2.25, "learning_rate": 1.6497660467906419e-06, "loss": 0.0001, "step": 18807 }, { "epoch": 2.25, "learning_rate": 1.6496940611877627e-06, "loss": 0.0001, "step": 18808 }, { "epoch": 2.25, "learning_rate": 1.649622075584883e-06, "loss": 0.0002, "step": 18809 }, { "epoch": 2.25, "learning_rate": 1.6495500899820038e-06, "loss": 0.0001, "step": 18810 }, { "epoch": 2.25, "learning_rate": 1.6494781043791241e-06, "loss": 0.0002, "step": 18811 }, { "epoch": 2.25, "learning_rate": 1.649406118776245e-06, "loss": 0.0001, "step": 18812 }, { "epoch": 2.25, "learning_rate": 1.6493341331733653e-06, "loss": 0.0002, "step": 18813 }, { "epoch": 2.25, "learning_rate": 1.649262147570486e-06, "loss": 0.0002, "step": 18814 }, { "epoch": 2.25, "learning_rate": 1.6491901619676066e-06, "loss": 0.0001, "step": 18815 }, { "epoch": 2.25, "learning_rate": 1.649118176364727e-06, "loss": 0.0075, "step": 18816 }, { "epoch": 2.25, "learning_rate": 1.6490461907618478e-06, "loss": 0.0001, "step": 18817 }, { "epoch": 2.26, "learning_rate": 1.6489742051589681e-06, "loss": 0.0004, "step": 18818 }, { "epoch": 2.26, "learning_rate": 1.6489022195560889e-06, "loss": 1.876, "step": 18819 }, { "epoch": 2.26, "learning_rate": 1.6488302339532092e-06, "loss": 0.0002, "step": 18820 }, { "epoch": 2.26, "learning_rate": 1.64875824835033e-06, "loss": 0.0001, "step": 18821 }, { "epoch": 2.26, "learning_rate": 1.6486862627474506e-06, "loss": 0.0001, "step": 18822 }, { "epoch": 2.26, "learning_rate": 1.6486142771445712e-06, "loss": 0.0001, "step": 18823 }, { "epoch": 2.26, "learning_rate": 1.6485422915416917e-06, "loss": 0.0001, "step": 18824 }, { "epoch": 2.26, "learning_rate": 1.6484703059388125e-06, "loss": 0.0001, "step": 18825 }, { "epoch": 2.26, "learning_rate": 1.6483983203359329e-06, "loss": 0.0002, "step": 18826 }, { "epoch": 2.26, "learning_rate": 1.6483263347330532e-06, "loss": 0.0001, "step": 18827 }, { "epoch": 2.26, "learning_rate": 1.648254349130174e-06, "loss": 0.0001, "step": 18828 }, { "epoch": 2.26, "learning_rate": 1.6481823635272946e-06, "loss": 0.0001, "step": 18829 }, { "epoch": 2.26, "learning_rate": 1.6481103779244151e-06, "loss": 0.0002, "step": 18830 }, { "epoch": 2.26, "learning_rate": 1.6480383923215357e-06, "loss": 0.0002, "step": 18831 }, { "epoch": 2.26, "learning_rate": 1.6479664067186565e-06, "loss": 0.0001, "step": 18832 }, { "epoch": 2.26, "learning_rate": 1.6478944211157768e-06, "loss": 0.0001, "step": 18833 }, { "epoch": 2.26, "learning_rate": 1.6478224355128976e-06, "loss": 0.0002, "step": 18834 }, { "epoch": 2.26, "learning_rate": 1.647750449910018e-06, "loss": 0.0002, "step": 18835 }, { "epoch": 2.26, "learning_rate": 1.6476784643071387e-06, "loss": 0.0003, "step": 18836 }, { "epoch": 2.26, "learning_rate": 1.647606478704259e-06, "loss": 0.0001, "step": 18837 }, { "epoch": 2.26, "learning_rate": 1.6475344931013797e-06, "loss": 0.0001, "step": 18838 }, { "epoch": 2.26, "learning_rate": 1.6474625074985004e-06, "loss": 0.0002, "step": 18839 }, { "epoch": 2.26, "learning_rate": 1.6473905218956208e-06, "loss": 0.0001, "step": 18840 }, { "epoch": 2.26, "learning_rate": 1.6473185362927416e-06, "loss": 0.0002, "step": 18841 }, { "epoch": 2.26, "learning_rate": 1.647246550689862e-06, "loss": 0.0001, "step": 18842 }, { "epoch": 2.26, "learning_rate": 1.6471745650869827e-06, "loss": 0.0001, "step": 18843 }, { "epoch": 2.26, "learning_rate": 1.647102579484103e-06, "loss": 0.0002, "step": 18844 }, { "epoch": 2.26, "learning_rate": 1.6470305938812238e-06, "loss": 0.0002, "step": 18845 }, { "epoch": 2.26, "learning_rate": 1.6469586082783444e-06, "loss": 0.0002, "step": 18846 }, { "epoch": 2.26, "learning_rate": 1.6468866226754652e-06, "loss": 0.0002, "step": 18847 }, { "epoch": 2.26, "learning_rate": 1.6468146370725855e-06, "loss": 0.1657, "step": 18848 }, { "epoch": 2.26, "learning_rate": 1.6467426514697063e-06, "loss": 0.0001, "step": 18849 }, { "epoch": 2.26, "learning_rate": 1.6466706658668267e-06, "loss": 0.0001, "step": 18850 }, { "epoch": 2.26, "learning_rate": 1.646598680263947e-06, "loss": 0.0001, "step": 18851 }, { "epoch": 2.26, "learning_rate": 1.6465266946610678e-06, "loss": 0.0001, "step": 18852 }, { "epoch": 2.26, "learning_rate": 1.6464547090581884e-06, "loss": 0.0001, "step": 18853 }, { "epoch": 2.26, "learning_rate": 1.6463827234553091e-06, "loss": 0.0003, "step": 18854 }, { "epoch": 2.26, "learning_rate": 1.6463107378524295e-06, "loss": 0.0002, "step": 18855 }, { "epoch": 2.26, "learning_rate": 1.6462387522495503e-06, "loss": 0.0002, "step": 18856 }, { "epoch": 2.26, "learning_rate": 1.6461667666466706e-06, "loss": 0.0001, "step": 18857 }, { "epoch": 2.26, "learning_rate": 1.6460947810437914e-06, "loss": 0.0001, "step": 18858 }, { "epoch": 2.26, "learning_rate": 1.6460227954409118e-06, "loss": 0.0033, "step": 18859 }, { "epoch": 2.26, "learning_rate": 1.6459508098380325e-06, "loss": 0.0002, "step": 18860 }, { "epoch": 2.26, "learning_rate": 1.6458788242351531e-06, "loss": 0.0001, "step": 18861 }, { "epoch": 2.26, "learning_rate": 1.6458068386322735e-06, "loss": 0.0542, "step": 18862 }, { "epoch": 2.26, "learning_rate": 1.6457348530293942e-06, "loss": 0.0001, "step": 18863 }, { "epoch": 2.26, "learning_rate": 1.6456628674265146e-06, "loss": 0.8096, "step": 18864 }, { "epoch": 2.26, "learning_rate": 1.6455908818236354e-06, "loss": 0.0002, "step": 18865 }, { "epoch": 2.26, "learning_rate": 1.6455188962207557e-06, "loss": 0.0001, "step": 18866 }, { "epoch": 2.26, "learning_rate": 1.6454469106178765e-06, "loss": 0.0002, "step": 18867 }, { "epoch": 2.26, "learning_rate": 1.645374925014997e-06, "loss": 0.0001, "step": 18868 }, { "epoch": 2.26, "learning_rate": 1.6453029394121176e-06, "loss": 0.0001, "step": 18869 }, { "epoch": 2.26, "learning_rate": 1.6452309538092382e-06, "loss": 0.0001, "step": 18870 }, { "epoch": 2.26, "learning_rate": 1.645158968206359e-06, "loss": 0.0001, "step": 18871 }, { "epoch": 2.26, "learning_rate": 1.6450869826034794e-06, "loss": 0.0002, "step": 18872 }, { "epoch": 2.26, "learning_rate": 1.6450149970005997e-06, "loss": 0.0001, "step": 18873 }, { "epoch": 2.26, "learning_rate": 1.6449430113977205e-06, "loss": 0.7882, "step": 18874 }, { "epoch": 2.26, "learning_rate": 1.6448710257948408e-06, "loss": 0.0001, "step": 18875 }, { "epoch": 2.26, "learning_rate": 1.6447990401919616e-06, "loss": 0.0001, "step": 18876 }, { "epoch": 2.26, "learning_rate": 1.6447270545890822e-06, "loss": 0.0001, "step": 18877 }, { "epoch": 2.26, "learning_rate": 1.644655068986203e-06, "loss": 0.0001, "step": 18878 }, { "epoch": 2.26, "learning_rate": 1.6445830833833233e-06, "loss": 0.1997, "step": 18879 }, { "epoch": 2.26, "learning_rate": 1.644511097780444e-06, "loss": 0.0001, "step": 18880 }, { "epoch": 2.26, "learning_rate": 1.6444391121775645e-06, "loss": 0.0001, "step": 18881 }, { "epoch": 2.26, "learning_rate": 1.6443671265746852e-06, "loss": 0.0001, "step": 18882 }, { "epoch": 2.26, "learning_rate": 1.6442951409718056e-06, "loss": 0.0004, "step": 18883 }, { "epoch": 2.26, "learning_rate": 1.6442231553689264e-06, "loss": 0.0001, "step": 18884 }, { "epoch": 2.26, "learning_rate": 1.644151169766047e-06, "loss": 0.0002, "step": 18885 }, { "epoch": 2.26, "learning_rate": 1.6440791841631673e-06, "loss": 0.0001, "step": 18886 }, { "epoch": 2.26, "learning_rate": 1.644007198560288e-06, "loss": 0.0001, "step": 18887 }, { "epoch": 2.26, "learning_rate": 1.6439352129574084e-06, "loss": 0.0002, "step": 18888 }, { "epoch": 2.26, "learning_rate": 1.6438632273545292e-06, "loss": 0.0001, "step": 18889 }, { "epoch": 2.26, "learning_rate": 1.6437912417516496e-06, "loss": 0.0001, "step": 18890 }, { "epoch": 2.26, "learning_rate": 1.6437192561487703e-06, "loss": 0.0001, "step": 18891 }, { "epoch": 2.26, "learning_rate": 1.643647270545891e-06, "loss": 0.0002, "step": 18892 }, { "epoch": 2.26, "learning_rate": 1.6435752849430115e-06, "loss": 0.0001, "step": 18893 }, { "epoch": 2.26, "learning_rate": 1.643503299340132e-06, "loss": 0.0001, "step": 18894 }, { "epoch": 2.26, "learning_rate": 1.6434313137372528e-06, "loss": 0.0002, "step": 18895 }, { "epoch": 2.26, "learning_rate": 1.6433593281343732e-06, "loss": 0.0037, "step": 18896 }, { "epoch": 2.26, "learning_rate": 1.6432873425314935e-06, "loss": 1.0742, "step": 18897 }, { "epoch": 2.26, "learning_rate": 1.6432153569286143e-06, "loss": 0.0001, "step": 18898 }, { "epoch": 2.26, "learning_rate": 1.6431433713257349e-06, "loss": 0.0001, "step": 18899 }, { "epoch": 2.26, "learning_rate": 1.6430713857228554e-06, "loss": 0.0001, "step": 18900 }, { "epoch": 2.26, "learning_rate": 1.642999400119976e-06, "loss": 0.0002, "step": 18901 }, { "epoch": 2.27, "learning_rate": 1.6429274145170968e-06, "loss": 0.0001, "step": 18902 }, { "epoch": 2.27, "learning_rate": 1.6428554289142171e-06, "loss": 0.0002, "step": 18903 }, { "epoch": 2.27, "learning_rate": 1.642783443311338e-06, "loss": 0.0002, "step": 18904 }, { "epoch": 2.27, "learning_rate": 1.6427114577084583e-06, "loss": 0.0002, "step": 18905 }, { "epoch": 2.27, "learning_rate": 1.642639472105579e-06, "loss": 0.0001, "step": 18906 }, { "epoch": 2.27, "learning_rate": 1.6425674865026994e-06, "loss": 0.0626, "step": 18907 }, { "epoch": 2.27, "learning_rate": 1.64249550089982e-06, "loss": 0.0001, "step": 18908 }, { "epoch": 2.27, "learning_rate": 1.6424235152969407e-06, "loss": 0.0017, "step": 18909 }, { "epoch": 2.27, "learning_rate": 1.642351529694061e-06, "loss": 0.0002, "step": 18910 }, { "epoch": 2.27, "learning_rate": 1.6422795440911819e-06, "loss": 0.0002, "step": 18911 }, { "epoch": 2.27, "learning_rate": 1.6422075584883022e-06, "loss": 0.0001, "step": 18912 }, { "epoch": 2.27, "learning_rate": 1.642135572885423e-06, "loss": 0.0002, "step": 18913 }, { "epoch": 2.27, "learning_rate": 1.6420635872825434e-06, "loss": 0.0002, "step": 18914 }, { "epoch": 2.27, "learning_rate": 1.6419916016796641e-06, "loss": 0.0006, "step": 18915 }, { "epoch": 2.27, "learning_rate": 1.6419196160767847e-06, "loss": 0.0001, "step": 18916 }, { "epoch": 2.27, "learning_rate": 1.6418476304739055e-06, "loss": 0.0001, "step": 18917 }, { "epoch": 2.27, "learning_rate": 1.6417756448710258e-06, "loss": 0.0001, "step": 18918 }, { "epoch": 2.27, "learning_rate": 1.6417036592681462e-06, "loss": 0.0002, "step": 18919 }, { "epoch": 2.27, "learning_rate": 1.641631673665267e-06, "loss": 0.0002, "step": 18920 }, { "epoch": 2.27, "learning_rate": 1.6415596880623873e-06, "loss": 0.0002, "step": 18921 }, { "epoch": 2.27, "learning_rate": 1.6414877024595081e-06, "loss": 0.0001, "step": 18922 }, { "epoch": 2.27, "learning_rate": 1.6414157168566287e-06, "loss": 0.0001, "step": 18923 }, { "epoch": 2.27, "learning_rate": 1.6413437312537495e-06, "loss": 0.0002, "step": 18924 }, { "epoch": 2.27, "learning_rate": 1.6412717456508698e-06, "loss": 0.0002, "step": 18925 }, { "epoch": 2.27, "learning_rate": 1.6411997600479906e-06, "loss": 0.0002, "step": 18926 }, { "epoch": 2.27, "learning_rate": 1.641127774445111e-06, "loss": 0.0001, "step": 18927 }, { "epoch": 2.27, "learning_rate": 1.6410557888422317e-06, "loss": 0.0001, "step": 18928 }, { "epoch": 2.27, "learning_rate": 1.640983803239352e-06, "loss": 0.0001, "step": 18929 }, { "epoch": 2.27, "learning_rate": 1.6409118176364729e-06, "loss": 0.0001, "step": 18930 }, { "epoch": 2.27, "learning_rate": 1.6408398320335934e-06, "loss": 0.0001, "step": 18931 }, { "epoch": 2.27, "learning_rate": 1.6407678464307138e-06, "loss": 0.0001, "step": 18932 }, { "epoch": 2.27, "learning_rate": 1.6406958608278346e-06, "loss": 0.0028, "step": 18933 }, { "epoch": 2.27, "learning_rate": 1.640623875224955e-06, "loss": 0.0003, "step": 18934 }, { "epoch": 2.27, "learning_rate": 1.6405518896220757e-06, "loss": 0.0002, "step": 18935 }, { "epoch": 2.27, "learning_rate": 1.640479904019196e-06, "loss": 0.0002, "step": 18936 }, { "epoch": 2.27, "learning_rate": 1.6404079184163168e-06, "loss": 0.0001, "step": 18937 }, { "epoch": 2.27, "learning_rate": 1.6403359328134374e-06, "loss": 0.0001, "step": 18938 }, { "epoch": 2.27, "learning_rate": 1.640263947210558e-06, "loss": 0.0001, "step": 18939 }, { "epoch": 2.27, "learning_rate": 1.6401919616076785e-06, "loss": 0.0004, "step": 18940 }, { "epoch": 2.27, "learning_rate": 1.6401199760047993e-06, "loss": 0.0002, "step": 18941 }, { "epoch": 2.27, "learning_rate": 1.6400479904019197e-06, "loss": 0.0002, "step": 18942 }, { "epoch": 2.27, "learning_rate": 1.63997600479904e-06, "loss": 0.0004, "step": 18943 }, { "epoch": 2.27, "learning_rate": 1.6399040191961608e-06, "loss": 0.0002, "step": 18944 }, { "epoch": 2.27, "learning_rate": 1.6398320335932814e-06, "loss": 0.0233, "step": 18945 }, { "epoch": 2.27, "learning_rate": 1.639760047990402e-06, "loss": 0.0001, "step": 18946 }, { "epoch": 2.27, "learning_rate": 1.6396880623875225e-06, "loss": 0.0002, "step": 18947 }, { "epoch": 2.27, "learning_rate": 1.6396160767846433e-06, "loss": 0.0002, "step": 18948 }, { "epoch": 2.27, "learning_rate": 1.6395440911817636e-06, "loss": 0.0001, "step": 18949 }, { "epoch": 2.27, "learning_rate": 1.6394721055788844e-06, "loss": 0.0003, "step": 18950 }, { "epoch": 2.27, "learning_rate": 1.6394001199760048e-06, "loss": 0.0002, "step": 18951 }, { "epoch": 2.27, "learning_rate": 1.6393281343731255e-06, "loss": 0.0001, "step": 18952 }, { "epoch": 2.27, "learning_rate": 1.639256148770246e-06, "loss": 0.0001, "step": 18953 }, { "epoch": 2.27, "learning_rate": 1.6391841631673665e-06, "loss": 0.0002, "step": 18954 }, { "epoch": 2.27, "learning_rate": 1.6391121775644872e-06, "loss": 0.0003, "step": 18955 }, { "epoch": 2.27, "learning_rate": 1.6390401919616076e-06, "loss": 0.0005, "step": 18956 }, { "epoch": 2.27, "learning_rate": 1.6389682063587284e-06, "loss": 0.0002, "step": 18957 }, { "epoch": 2.27, "learning_rate": 1.6388962207558487e-06, "loss": 0.0003, "step": 18958 }, { "epoch": 2.27, "learning_rate": 1.6388242351529695e-06, "loss": 0.0002, "step": 18959 }, { "epoch": 2.27, "learning_rate": 1.6387522495500899e-06, "loss": 0.0001, "step": 18960 }, { "epoch": 2.27, "learning_rate": 1.6386802639472106e-06, "loss": 0.0001, "step": 18961 }, { "epoch": 2.27, "learning_rate": 1.6386082783443312e-06, "loss": 0.0004, "step": 18962 }, { "epoch": 2.27, "learning_rate": 1.638536292741452e-06, "loss": 0.0001, "step": 18963 }, { "epoch": 2.27, "learning_rate": 1.6384643071385723e-06, "loss": 0.0001, "step": 18964 }, { "epoch": 2.27, "learning_rate": 1.6383923215356927e-06, "loss": 0.0001, "step": 18965 }, { "epoch": 2.27, "learning_rate": 1.6383203359328135e-06, "loss": 0.0001, "step": 18966 }, { "epoch": 2.27, "learning_rate": 1.6382483503299338e-06, "loss": 0.0001, "step": 18967 }, { "epoch": 2.27, "learning_rate": 1.6381763647270546e-06, "loss": 0.0001, "step": 18968 }, { "epoch": 2.27, "learning_rate": 1.6381043791241752e-06, "loss": 0.0001, "step": 18969 }, { "epoch": 2.27, "learning_rate": 1.638032393521296e-06, "loss": 0.0002, "step": 18970 }, { "epoch": 2.27, "learning_rate": 1.6379604079184163e-06, "loss": 0.0002, "step": 18971 }, { "epoch": 2.27, "learning_rate": 1.637888422315537e-06, "loss": 0.0001, "step": 18972 }, { "epoch": 2.27, "learning_rate": 1.6378164367126574e-06, "loss": 0.0001, "step": 18973 }, { "epoch": 2.27, "learning_rate": 1.6377444511097782e-06, "loss": 0.0001, "step": 18974 }, { "epoch": 2.27, "learning_rate": 1.6376724655068986e-06, "loss": 0.0001, "step": 18975 }, { "epoch": 2.27, "learning_rate": 1.6376004799040194e-06, "loss": 0.0001, "step": 18976 }, { "epoch": 2.27, "learning_rate": 1.63752849430114e-06, "loss": 0.0001, "step": 18977 }, { "epoch": 2.27, "learning_rate": 1.6374565086982603e-06, "loss": 0.0002, "step": 18978 }, { "epoch": 2.27, "learning_rate": 1.637384523095381e-06, "loss": 0.0002, "step": 18979 }, { "epoch": 2.27, "learning_rate": 1.6373125374925014e-06, "loss": 0.0004, "step": 18980 }, { "epoch": 2.27, "learning_rate": 1.6372405518896222e-06, "loss": 0.0002, "step": 18981 }, { "epoch": 2.27, "learning_rate": 1.6371685662867425e-06, "loss": 0.0003, "step": 18982 }, { "epoch": 2.27, "learning_rate": 1.6370965806838633e-06, "loss": 0.0002, "step": 18983 }, { "epoch": 2.27, "learning_rate": 1.6370245950809839e-06, "loss": 0.0001, "step": 18984 }, { "epoch": 2.28, "learning_rate": 1.6369526094781045e-06, "loss": 0.0002, "step": 18985 }, { "epoch": 2.28, "learning_rate": 1.636880623875225e-06, "loss": 0.0001, "step": 18986 }, { "epoch": 2.28, "learning_rate": 1.6368086382723458e-06, "loss": 0.0002, "step": 18987 }, { "epoch": 2.28, "learning_rate": 1.6367366526694662e-06, "loss": 0.0002, "step": 18988 }, { "epoch": 2.28, "learning_rate": 1.6366646670665865e-06, "loss": 0.0002, "step": 18989 }, { "epoch": 2.28, "learning_rate": 1.6365926814637073e-06, "loss": 0.0001, "step": 18990 }, { "epoch": 2.28, "learning_rate": 1.6365206958608279e-06, "loss": 0.0002, "step": 18991 }, { "epoch": 2.28, "learning_rate": 1.6364487102579484e-06, "loss": 0.0001, "step": 18992 }, { "epoch": 2.28, "learning_rate": 1.636376724655069e-06, "loss": 0.0001, "step": 18993 }, { "epoch": 2.28, "learning_rate": 1.6363047390521898e-06, "loss": 0.0001, "step": 18994 }, { "epoch": 2.28, "learning_rate": 1.6362327534493101e-06, "loss": 0.0002, "step": 18995 }, { "epoch": 2.28, "learning_rate": 1.636160767846431e-06, "loss": 0.0002, "step": 18996 }, { "epoch": 2.28, "learning_rate": 1.6360887822435513e-06, "loss": 0.0006, "step": 18997 }, { "epoch": 2.28, "learning_rate": 1.636016796640672e-06, "loss": 0.0001, "step": 18998 }, { "epoch": 2.28, "learning_rate": 1.6359448110377924e-06, "loss": 0.0002, "step": 18999 }, { "epoch": 2.28, "learning_rate": 1.635872825434913e-06, "loss": 0.0002, "step": 19000 }, { "epoch": 2.28, "learning_rate": 1.6358008398320337e-06, "loss": 0.0002, "step": 19001 }, { "epoch": 2.28, "learning_rate": 1.635728854229154e-06, "loss": 0.0002, "step": 19002 }, { "epoch": 2.28, "learning_rate": 1.6356568686262749e-06, "loss": 0.0002, "step": 19003 }, { "epoch": 2.28, "learning_rate": 1.6355848830233952e-06, "loss": 0.0002, "step": 19004 }, { "epoch": 2.28, "learning_rate": 1.635512897420516e-06, "loss": 0.0001, "step": 19005 }, { "epoch": 2.28, "learning_rate": 1.6354409118176364e-06, "loss": 0.0002, "step": 19006 }, { "epoch": 2.28, "learning_rate": 1.6353689262147571e-06, "loss": 0.0013, "step": 19007 }, { "epoch": 2.28, "learning_rate": 1.6352969406118777e-06, "loss": 0.0001, "step": 19008 }, { "epoch": 2.28, "learning_rate": 1.6352249550089985e-06, "loss": 0.9785, "step": 19009 }, { "epoch": 2.28, "learning_rate": 1.6351529694061188e-06, "loss": 0.0002, "step": 19010 }, { "epoch": 2.28, "learning_rate": 1.6350809838032396e-06, "loss": 0.0001, "step": 19011 }, { "epoch": 2.28, "learning_rate": 1.63500899820036e-06, "loss": 0.0002, "step": 19012 }, { "epoch": 2.28, "learning_rate": 1.6349370125974803e-06, "loss": 0.0002, "step": 19013 }, { "epoch": 2.28, "learning_rate": 1.634865026994601e-06, "loss": 0.0001, "step": 19014 }, { "epoch": 2.28, "learning_rate": 1.6347930413917217e-06, "loss": 0.007, "step": 19015 }, { "epoch": 2.28, "learning_rate": 1.6347210557888422e-06, "loss": 0.0, "step": 19016 }, { "epoch": 2.28, "learning_rate": 1.6346490701859628e-06, "loss": 0.0002, "step": 19017 }, { "epoch": 2.28, "learning_rate": 1.6345770845830836e-06, "loss": 0.0001, "step": 19018 }, { "epoch": 2.28, "learning_rate": 1.634505098980204e-06, "loss": 0.0001, "step": 19019 }, { "epoch": 2.28, "learning_rate": 1.6344331133773247e-06, "loss": 0.0, "step": 19020 }, { "epoch": 2.28, "learning_rate": 1.634361127774445e-06, "loss": 0.0003, "step": 19021 }, { "epoch": 2.28, "learning_rate": 1.6342891421715659e-06, "loss": 0.0006, "step": 19022 }, { "epoch": 2.28, "learning_rate": 1.6342171565686862e-06, "loss": 0.0001, "step": 19023 }, { "epoch": 2.28, "learning_rate": 1.6341451709658068e-06, "loss": 0.0002, "step": 19024 }, { "epoch": 2.28, "learning_rate": 1.6340731853629276e-06, "loss": 0.0266, "step": 19025 }, { "epoch": 2.28, "learning_rate": 1.634001199760048e-06, "loss": 0.0002, "step": 19026 }, { "epoch": 2.28, "learning_rate": 1.6339292141571687e-06, "loss": 0.0002, "step": 19027 }, { "epoch": 2.28, "learning_rate": 1.633857228554289e-06, "loss": 0.0001, "step": 19028 }, { "epoch": 2.28, "learning_rate": 1.6337852429514098e-06, "loss": 0.0001, "step": 19029 }, { "epoch": 2.28, "learning_rate": 1.6337132573485302e-06, "loss": 0.0001, "step": 19030 }, { "epoch": 2.28, "learning_rate": 1.633641271745651e-06, "loss": 0.0001, "step": 19031 }, { "epoch": 2.28, "learning_rate": 1.6335692861427715e-06, "loss": 0.0002, "step": 19032 }, { "epoch": 2.28, "learning_rate": 1.6334973005398923e-06, "loss": 0.0001, "step": 19033 }, { "epoch": 2.28, "learning_rate": 1.6334253149370127e-06, "loss": 1.1013, "step": 19034 }, { "epoch": 2.28, "learning_rate": 1.633353329334133e-06, "loss": 0.0002, "step": 19035 }, { "epoch": 2.28, "learning_rate": 1.6332813437312538e-06, "loss": 0.0003, "step": 19036 }, { "epoch": 2.28, "learning_rate": 1.6332093581283741e-06, "loss": 0.0004, "step": 19037 }, { "epoch": 2.28, "learning_rate": 1.633137372525495e-06, "loss": 0.0002, "step": 19038 }, { "epoch": 2.28, "learning_rate": 1.6330653869226155e-06, "loss": 0.0002, "step": 19039 }, { "epoch": 2.28, "learning_rate": 1.6329934013197363e-06, "loss": 0.0003, "step": 19040 }, { "epoch": 2.28, "learning_rate": 1.6329214157168566e-06, "loss": 0.0002, "step": 19041 }, { "epoch": 2.28, "learning_rate": 1.6328494301139774e-06, "loss": 0.0001, "step": 19042 }, { "epoch": 2.28, "learning_rate": 1.6327774445110978e-06, "loss": 0.0002, "step": 19043 }, { "epoch": 2.28, "learning_rate": 1.6327054589082185e-06, "loss": 0.0003, "step": 19044 }, { "epoch": 2.28, "learning_rate": 1.6326334733053389e-06, "loss": 0.0001, "step": 19045 }, { "epoch": 2.28, "learning_rate": 1.6325614877024595e-06, "loss": 0.0001, "step": 19046 }, { "epoch": 2.28, "learning_rate": 1.6324895020995802e-06, "loss": 0.0003, "step": 19047 }, { "epoch": 2.28, "learning_rate": 1.6324175164967006e-06, "loss": 0.0005, "step": 19048 }, { "epoch": 2.28, "learning_rate": 1.6323455308938214e-06, "loss": 0.0001, "step": 19049 }, { "epoch": 2.28, "learning_rate": 1.6322735452909417e-06, "loss": 0.0002, "step": 19050 }, { "epoch": 2.28, "learning_rate": 1.6322015596880625e-06, "loss": 0.0002, "step": 19051 }, { "epoch": 2.28, "learning_rate": 1.6321295740851829e-06, "loss": 0.0003, "step": 19052 }, { "epoch": 2.28, "learning_rate": 1.6320575884823036e-06, "loss": 0.0002, "step": 19053 }, { "epoch": 2.28, "learning_rate": 1.6319856028794242e-06, "loss": 0.0001, "step": 19054 }, { "epoch": 2.28, "learning_rate": 1.6319136172765448e-06, "loss": 0.0001, "step": 19055 }, { "epoch": 2.28, "learning_rate": 1.6318416316736653e-06, "loss": 0.0002, "step": 19056 }, { "epoch": 2.28, "learning_rate": 1.6317696460707861e-06, "loss": 0.0006, "step": 19057 }, { "epoch": 2.28, "learning_rate": 1.6316976604679065e-06, "loss": 0.0002, "step": 19058 }, { "epoch": 2.28, "learning_rate": 1.6316256748650268e-06, "loss": 0.0001, "step": 19059 }, { "epoch": 2.28, "learning_rate": 1.6315536892621476e-06, "loss": 0.0001, "step": 19060 }, { "epoch": 2.28, "learning_rate": 1.6314817036592682e-06, "loss": 0.0004, "step": 19061 }, { "epoch": 2.28, "learning_rate": 1.6314097180563887e-06, "loss": 0.0001, "step": 19062 }, { "epoch": 2.28, "learning_rate": 1.6313377324535093e-06, "loss": 0.0003, "step": 19063 }, { "epoch": 2.28, "learning_rate": 1.63126574685063e-06, "loss": 0.0002, "step": 19064 }, { "epoch": 2.28, "learning_rate": 1.6311937612477504e-06, "loss": 0.0003, "step": 19065 }, { "epoch": 2.28, "learning_rate": 1.6311217756448712e-06, "loss": 0.0002, "step": 19066 }, { "epoch": 2.28, "learning_rate": 1.6310497900419916e-06, "loss": 0.0003, "step": 19067 }, { "epoch": 2.28, "learning_rate": 1.6309778044391123e-06, "loss": 0.0001, "step": 19068 }, { "epoch": 2.29, "learning_rate": 1.6309058188362327e-06, "loss": 0.0002, "step": 19069 }, { "epoch": 2.29, "learning_rate": 1.6308338332333533e-06, "loss": 0.0842, "step": 19070 }, { "epoch": 2.29, "learning_rate": 1.630761847630474e-06, "loss": 0.0002, "step": 19071 }, { "epoch": 2.29, "learning_rate": 1.6306898620275944e-06, "loss": 0.0004, "step": 19072 }, { "epoch": 2.29, "learning_rate": 1.6306178764247152e-06, "loss": 0.0001, "step": 19073 }, { "epoch": 2.29, "learning_rate": 1.6305458908218355e-06, "loss": 0.0003, "step": 19074 }, { "epoch": 2.29, "learning_rate": 1.6304739052189563e-06, "loss": 0.9569, "step": 19075 }, { "epoch": 2.29, "learning_rate": 1.6304019196160767e-06, "loss": 0.0002, "step": 19076 }, { "epoch": 2.29, "learning_rate": 1.6303299340131974e-06, "loss": 0.0001, "step": 19077 }, { "epoch": 2.29, "learning_rate": 1.630257948410318e-06, "loss": 0.0002, "step": 19078 }, { "epoch": 2.29, "learning_rate": 1.6301859628074388e-06, "loss": 0.0001, "step": 19079 }, { "epoch": 2.29, "learning_rate": 1.6301139772045591e-06, "loss": 0.0002, "step": 19080 }, { "epoch": 2.29, "learning_rate": 1.6300419916016795e-06, "loss": 0.0001, "step": 19081 }, { "epoch": 2.29, "learning_rate": 1.6299700059988003e-06, "loss": 0.61, "step": 19082 }, { "epoch": 2.29, "learning_rate": 1.6298980203959206e-06, "loss": 0.0004, "step": 19083 }, { "epoch": 2.29, "learning_rate": 1.6298260347930414e-06, "loss": 0.0003, "step": 19084 }, { "epoch": 2.29, "learning_rate": 1.629754049190162e-06, "loss": 0.0001, "step": 19085 }, { "epoch": 2.29, "learning_rate": 1.6296820635872828e-06, "loss": 0.0001, "step": 19086 }, { "epoch": 2.29, "learning_rate": 1.6296100779844031e-06, "loss": 0.0002, "step": 19087 }, { "epoch": 2.29, "learning_rate": 1.629538092381524e-06, "loss": 0.0001, "step": 19088 }, { "epoch": 2.29, "learning_rate": 1.6294661067786442e-06, "loss": 0.0003, "step": 19089 }, { "epoch": 2.29, "learning_rate": 1.629394121175765e-06, "loss": 0.0053, "step": 19090 }, { "epoch": 2.29, "learning_rate": 1.6293221355728854e-06, "loss": 0.0001, "step": 19091 }, { "epoch": 2.29, "learning_rate": 1.629250149970006e-06, "loss": 0.0005, "step": 19092 }, { "epoch": 2.29, "learning_rate": 1.6291781643671267e-06, "loss": 0.0001, "step": 19093 }, { "epoch": 2.29, "learning_rate": 1.629106178764247e-06, "loss": 0.0002, "step": 19094 }, { "epoch": 2.29, "learning_rate": 1.6290341931613679e-06, "loss": 0.0001, "step": 19095 }, { "epoch": 2.29, "learning_rate": 1.6289622075584882e-06, "loss": 0.0003, "step": 19096 }, { "epoch": 2.29, "learning_rate": 1.628890221955609e-06, "loss": 0.0002, "step": 19097 }, { "epoch": 2.29, "learning_rate": 1.6288182363527294e-06, "loss": 0.0002, "step": 19098 }, { "epoch": 2.29, "learning_rate": 1.6287462507498501e-06, "loss": 0.0004, "step": 19099 }, { "epoch": 2.29, "learning_rate": 1.6286742651469707e-06, "loss": 0.0002, "step": 19100 }, { "epoch": 2.29, "learning_rate": 1.6286022795440913e-06, "loss": 0.0006, "step": 19101 }, { "epoch": 2.29, "learning_rate": 1.6285302939412118e-06, "loss": 0.0004, "step": 19102 }, { "epoch": 2.29, "learning_rate": 1.6284583083383326e-06, "loss": 0.0002, "step": 19103 }, { "epoch": 2.29, "learning_rate": 1.628386322735453e-06, "loss": 0.0001, "step": 19104 }, { "epoch": 2.29, "learning_rate": 1.6283143371325733e-06, "loss": 0.0004, "step": 19105 }, { "epoch": 2.29, "learning_rate": 1.628242351529694e-06, "loss": 0.0001, "step": 19106 }, { "epoch": 2.29, "learning_rate": 1.6281703659268147e-06, "loss": 0.0001, "step": 19107 }, { "epoch": 2.29, "learning_rate": 1.6280983803239352e-06, "loss": 0.7948, "step": 19108 }, { "epoch": 2.29, "learning_rate": 1.6280263947210558e-06, "loss": 0.0002, "step": 19109 }, { "epoch": 2.29, "learning_rate": 1.6279544091181766e-06, "loss": 0.0002, "step": 19110 }, { "epoch": 2.29, "learning_rate": 1.627882423515297e-06, "loss": 0.0012, "step": 19111 }, { "epoch": 2.29, "learning_rate": 1.6278104379124177e-06, "loss": 0.0002, "step": 19112 }, { "epoch": 2.29, "learning_rate": 1.627738452309538e-06, "loss": 0.0005, "step": 19113 }, { "epoch": 2.29, "learning_rate": 1.6276664667066588e-06, "loss": 0.0002, "step": 19114 }, { "epoch": 2.29, "learning_rate": 1.6275944811037792e-06, "loss": 0.0001, "step": 19115 }, { "epoch": 2.29, "learning_rate": 1.6275224955008998e-06, "loss": 0.0002, "step": 19116 }, { "epoch": 2.29, "learning_rate": 1.6274505098980205e-06, "loss": 0.0003, "step": 19117 }, { "epoch": 2.29, "learning_rate": 1.627378524295141e-06, "loss": 0.0002, "step": 19118 }, { "epoch": 2.29, "learning_rate": 1.6273065386922617e-06, "loss": 0.0003, "step": 19119 }, { "epoch": 2.29, "learning_rate": 1.627234553089382e-06, "loss": 0.0004, "step": 19120 }, { "epoch": 2.29, "learning_rate": 1.6271625674865028e-06, "loss": 0.0004, "step": 19121 }, { "epoch": 2.29, "learning_rate": 1.6270905818836232e-06, "loss": 0.0001, "step": 19122 }, { "epoch": 2.29, "learning_rate": 1.627018596280744e-06, "loss": 0.0004, "step": 19123 }, { "epoch": 2.29, "learning_rate": 1.6269466106778645e-06, "loss": 0.0011, "step": 19124 }, { "epoch": 2.29, "learning_rate": 1.6268746250749853e-06, "loss": 0.0006, "step": 19125 }, { "epoch": 2.29, "learning_rate": 1.6268026394721056e-06, "loss": 0.0011, "step": 19126 }, { "epoch": 2.29, "learning_rate": 1.626730653869226e-06, "loss": 0.0003, "step": 19127 }, { "epoch": 2.29, "learning_rate": 1.6266586682663468e-06, "loss": 0.0002, "step": 19128 }, { "epoch": 2.29, "learning_rate": 1.6265866826634671e-06, "loss": 0.713, "step": 19129 }, { "epoch": 2.29, "learning_rate": 1.626514697060588e-06, "loss": 0.0001, "step": 19130 }, { "epoch": 2.29, "learning_rate": 1.6264427114577085e-06, "loss": 0.0001, "step": 19131 }, { "epoch": 2.29, "learning_rate": 1.6263707258548293e-06, "loss": 0.0002, "step": 19132 }, { "epoch": 2.29, "learning_rate": 1.6262987402519496e-06, "loss": 0.0001, "step": 19133 }, { "epoch": 2.29, "learning_rate": 1.6262267546490704e-06, "loss": 0.0001, "step": 19134 }, { "epoch": 2.29, "learning_rate": 1.6261547690461907e-06, "loss": 0.0001, "step": 19135 }, { "epoch": 2.29, "learning_rate": 1.6260827834433115e-06, "loss": 0.0001, "step": 19136 }, { "epoch": 2.29, "learning_rate": 1.6260107978404319e-06, "loss": 0.0002, "step": 19137 }, { "epoch": 2.29, "learning_rate": 1.6259388122375527e-06, "loss": 0.0002, "step": 19138 }, { "epoch": 2.29, "learning_rate": 1.6258668266346732e-06, "loss": 0.0002, "step": 19139 }, { "epoch": 2.29, "learning_rate": 1.6257948410317936e-06, "loss": 0.0002, "step": 19140 }, { "epoch": 2.29, "learning_rate": 1.6257228554289144e-06, "loss": 0.0008, "step": 19141 }, { "epoch": 2.29, "learning_rate": 1.6256508698260347e-06, "loss": 0.0002, "step": 19142 }, { "epoch": 2.29, "learning_rate": 1.6255788842231555e-06, "loss": 0.0001, "step": 19143 }, { "epoch": 2.29, "learning_rate": 1.6255068986202758e-06, "loss": 0.0002, "step": 19144 }, { "epoch": 2.29, "learning_rate": 1.6254349130173966e-06, "loss": 0.0002, "step": 19145 }, { "epoch": 2.29, "learning_rate": 1.625362927414517e-06, "loss": 0.0001, "step": 19146 }, { "epoch": 2.29, "learning_rate": 1.6252909418116378e-06, "loss": 0.0001, "step": 19147 }, { "epoch": 2.29, "learning_rate": 1.6252189562087583e-06, "loss": 0.0002, "step": 19148 }, { "epoch": 2.29, "learning_rate": 1.625146970605879e-06, "loss": 0.0002, "step": 19149 }, { "epoch": 2.29, "learning_rate": 1.6250749850029995e-06, "loss": 0.0023, "step": 19150 }, { "epoch": 2.29, "learning_rate": 1.6250029994001198e-06, "loss": 0.0003, "step": 19151 }, { "epoch": 2.3, "learning_rate": 1.6249310137972406e-06, "loss": 0.0001, "step": 19152 }, { "epoch": 2.3, "learning_rate": 1.624859028194361e-06, "loss": 0.0001, "step": 19153 }, { "epoch": 2.3, "learning_rate": 1.6247870425914817e-06, "loss": 0.0002, "step": 19154 }, { "epoch": 2.3, "learning_rate": 1.6247150569886023e-06, "loss": 0.0003, "step": 19155 }, { "epoch": 2.3, "learning_rate": 1.624643071385723e-06, "loss": 0.0003, "step": 19156 }, { "epoch": 2.3, "learning_rate": 1.6245710857828434e-06, "loss": 0.0003, "step": 19157 }, { "epoch": 2.3, "learning_rate": 1.6244991001799642e-06, "loss": 0.0002, "step": 19158 }, { "epoch": 2.3, "learning_rate": 1.6244271145770846e-06, "loss": 0.0002, "step": 19159 }, { "epoch": 2.3, "learning_rate": 1.6243551289742053e-06, "loss": 0.0002, "step": 19160 }, { "epoch": 2.3, "learning_rate": 1.6242831433713257e-06, "loss": 0.0001, "step": 19161 }, { "epoch": 2.3, "learning_rate": 1.6242111577684463e-06, "loss": 0.0002, "step": 19162 }, { "epoch": 2.3, "learning_rate": 1.624139172165567e-06, "loss": 0.9338, "step": 19163 }, { "epoch": 2.3, "learning_rate": 1.6240671865626874e-06, "loss": 0.0003, "step": 19164 }, { "epoch": 2.3, "learning_rate": 1.6239952009598082e-06, "loss": 0.0003, "step": 19165 }, { "epoch": 2.3, "learning_rate": 1.6239232153569285e-06, "loss": 0.0002, "step": 19166 }, { "epoch": 2.3, "learning_rate": 1.6238512297540493e-06, "loss": 0.0001, "step": 19167 }, { "epoch": 2.3, "learning_rate": 1.6237792441511697e-06, "loss": 0.0001, "step": 19168 }, { "epoch": 2.3, "learning_rate": 1.6237072585482904e-06, "loss": 0.0004, "step": 19169 }, { "epoch": 2.3, "learning_rate": 1.623635272945411e-06, "loss": 0.0001, "step": 19170 }, { "epoch": 2.3, "learning_rate": 1.6235632873425316e-06, "loss": 0.0001, "step": 19171 }, { "epoch": 2.3, "learning_rate": 1.6234913017396521e-06, "loss": 0.646, "step": 19172 }, { "epoch": 2.3, "learning_rate": 1.6234193161367725e-06, "loss": 0.0002, "step": 19173 }, { "epoch": 2.3, "learning_rate": 1.6233473305338933e-06, "loss": 0.0002, "step": 19174 }, { "epoch": 2.3, "learning_rate": 1.6232753449310136e-06, "loss": 0.0004, "step": 19175 }, { "epoch": 2.3, "learning_rate": 1.6232033593281344e-06, "loss": 0.0003, "step": 19176 }, { "epoch": 2.3, "learning_rate": 1.623131373725255e-06, "loss": 0.0001, "step": 19177 }, { "epoch": 2.3, "learning_rate": 1.6230593881223755e-06, "loss": 0.0002, "step": 19178 }, { "epoch": 2.3, "learning_rate": 1.6229874025194961e-06, "loss": 0.0002, "step": 19179 }, { "epoch": 2.3, "learning_rate": 1.6229154169166169e-06, "loss": 0.0001, "step": 19180 }, { "epoch": 2.3, "learning_rate": 1.6228434313137372e-06, "loss": 0.455, "step": 19181 }, { "epoch": 2.3, "learning_rate": 1.622771445710858e-06, "loss": 0.0001, "step": 19182 }, { "epoch": 2.3, "learning_rate": 1.6226994601079784e-06, "loss": 0.0001, "step": 19183 }, { "epoch": 2.3, "learning_rate": 1.6226274745050992e-06, "loss": 0.0003, "step": 19184 }, { "epoch": 2.3, "learning_rate": 1.6225554889022195e-06, "loss": 0.0003, "step": 19185 }, { "epoch": 2.3, "learning_rate": 1.62248350329934e-06, "loss": 0.0002, "step": 19186 }, { "epoch": 2.3, "learning_rate": 1.6224115176964609e-06, "loss": 0.0001, "step": 19187 }, { "epoch": 2.3, "learning_rate": 1.6223395320935812e-06, "loss": 0.0002, "step": 19188 }, { "epoch": 2.3, "learning_rate": 1.622267546490702e-06, "loss": 0.0003, "step": 19189 }, { "epoch": 2.3, "learning_rate": 1.6221955608878223e-06, "loss": 0.0004, "step": 19190 }, { "epoch": 2.3, "learning_rate": 1.6221235752849431e-06, "loss": 0.0001, "step": 19191 }, { "epoch": 2.3, "learning_rate": 1.6220515896820635e-06, "loss": 0.0002, "step": 19192 }, { "epoch": 2.3, "learning_rate": 1.6219796040791843e-06, "loss": 0.0025, "step": 19193 }, { "epoch": 2.3, "learning_rate": 1.6219076184763048e-06, "loss": 0.0006, "step": 19194 }, { "epoch": 2.3, "learning_rate": 1.6218356328734256e-06, "loss": 0.0003, "step": 19195 }, { "epoch": 2.3, "learning_rate": 1.621763647270546e-06, "loss": 0.0005, "step": 19196 }, { "epoch": 2.3, "learning_rate": 1.6216916616676663e-06, "loss": 0.0002, "step": 19197 }, { "epoch": 2.3, "learning_rate": 1.621619676064787e-06, "loss": 0.0004, "step": 19198 }, { "epoch": 2.3, "learning_rate": 1.6215476904619074e-06, "loss": 0.0002, "step": 19199 }, { "epoch": 2.3, "learning_rate": 1.6214757048590282e-06, "loss": 0.0004, "step": 19200 }, { "epoch": 2.3, "learning_rate": 1.6214037192561488e-06, "loss": 0.0003, "step": 19201 }, { "epoch": 2.3, "learning_rate": 1.6213317336532696e-06, "loss": 0.0002, "step": 19202 }, { "epoch": 2.3, "learning_rate": 1.62125974805039e-06, "loss": 0.0003, "step": 19203 }, { "epoch": 2.3, "learning_rate": 1.6211877624475107e-06, "loss": 0.0001, "step": 19204 }, { "epoch": 2.3, "learning_rate": 1.621115776844631e-06, "loss": 0.0002, "step": 19205 }, { "epoch": 2.3, "learning_rate": 1.6210437912417518e-06, "loss": 0.0002, "step": 19206 }, { "epoch": 2.3, "learning_rate": 1.6209718056388722e-06, "loss": 0.0003, "step": 19207 }, { "epoch": 2.3, "learning_rate": 1.6208998200359928e-06, "loss": 0.4853, "step": 19208 }, { "epoch": 2.3, "learning_rate": 1.6208278344331135e-06, "loss": 0.0002, "step": 19209 }, { "epoch": 2.3, "learning_rate": 1.6207558488302339e-06, "loss": 0.0002, "step": 19210 }, { "epoch": 2.3, "learning_rate": 1.6206838632273547e-06, "loss": 0.0003, "step": 19211 }, { "epoch": 2.3, "learning_rate": 1.620611877624475e-06, "loss": 0.0002, "step": 19212 }, { "epoch": 2.3, "learning_rate": 1.6205398920215958e-06, "loss": 0.0002, "step": 19213 }, { "epoch": 2.3, "learning_rate": 1.6204679064187162e-06, "loss": 0.0001, "step": 19214 }, { "epoch": 2.3, "learning_rate": 1.620395920815837e-06, "loss": 1.516, "step": 19215 }, { "epoch": 2.3, "learning_rate": 1.6203239352129575e-06, "loss": 0.0002, "step": 19216 }, { "epoch": 2.3, "learning_rate": 1.620251949610078e-06, "loss": 0.0002, "step": 19217 }, { "epoch": 2.3, "learning_rate": 1.6201799640071986e-06, "loss": 0.0001, "step": 19218 }, { "epoch": 2.3, "learning_rate": 1.620107978404319e-06, "loss": 0.0002, "step": 19219 }, { "epoch": 2.3, "learning_rate": 1.6200359928014398e-06, "loss": 0.0001, "step": 19220 }, { "epoch": 2.3, "learning_rate": 1.6199640071985601e-06, "loss": 0.0002, "step": 19221 }, { "epoch": 2.3, "learning_rate": 1.619892021595681e-06, "loss": 0.0002, "step": 19222 }, { "epoch": 2.3, "learning_rate": 1.6198200359928015e-06, "loss": 0.0003, "step": 19223 }, { "epoch": 2.3, "learning_rate": 1.619748050389922e-06, "loss": 0.0003, "step": 19224 }, { "epoch": 2.3, "learning_rate": 1.6196760647870426e-06, "loss": 0.0002, "step": 19225 }, { "epoch": 2.3, "learning_rate": 1.6196040791841634e-06, "loss": 0.0001, "step": 19226 }, { "epoch": 2.3, "learning_rate": 1.6195320935812837e-06, "loss": 0.0003, "step": 19227 }, { "epoch": 2.3, "learning_rate": 1.6194601079784045e-06, "loss": 0.0001, "step": 19228 }, { "epoch": 2.3, "learning_rate": 1.6193881223755249e-06, "loss": 0.0002, "step": 19229 }, { "epoch": 2.3, "learning_rate": 1.6193161367726457e-06, "loss": 0.0002, "step": 19230 }, { "epoch": 2.3, "learning_rate": 1.619244151169766e-06, "loss": 0.0001, "step": 19231 }, { "epoch": 2.3, "learning_rate": 1.6191721655668866e-06, "loss": 0.0001, "step": 19232 }, { "epoch": 2.3, "learning_rate": 1.6191001799640074e-06, "loss": 0.0001, "step": 19233 }, { "epoch": 2.3, "learning_rate": 1.6190281943611277e-06, "loss": 0.0002, "step": 19234 }, { "epoch": 2.3, "learning_rate": 1.6189562087582485e-06, "loss": 0.0004, "step": 19235 }, { "epoch": 2.31, "learning_rate": 1.6188842231553688e-06, "loss": 0.0007, "step": 19236 }, { "epoch": 2.31, "learning_rate": 1.6188122375524896e-06, "loss": 0.0001, "step": 19237 }, { "epoch": 2.31, "learning_rate": 1.61874025194961e-06, "loss": 0.0002, "step": 19238 }, { "epoch": 2.31, "learning_rate": 1.6186682663467308e-06, "loss": 0.0001, "step": 19239 }, { "epoch": 2.31, "learning_rate": 1.6185962807438513e-06, "loss": 0.0002, "step": 19240 }, { "epoch": 2.31, "learning_rate": 1.618524295140972e-06, "loss": 0.0002, "step": 19241 }, { "epoch": 2.31, "learning_rate": 1.6184523095380925e-06, "loss": 0.0001, "step": 19242 }, { "epoch": 2.31, "learning_rate": 1.6183803239352128e-06, "loss": 0.9583, "step": 19243 }, { "epoch": 2.31, "learning_rate": 1.6183083383323336e-06, "loss": 0.0001, "step": 19244 }, { "epoch": 2.31, "learning_rate": 1.618236352729454e-06, "loss": 0.0002, "step": 19245 }, { "epoch": 2.31, "learning_rate": 1.6181643671265747e-06, "loss": 0.0002, "step": 19246 }, { "epoch": 2.31, "learning_rate": 1.6180923815236953e-06, "loss": 0.0001, "step": 19247 }, { "epoch": 2.31, "learning_rate": 1.618020395920816e-06, "loss": 0.0001, "step": 19248 }, { "epoch": 2.31, "learning_rate": 1.6179484103179364e-06, "loss": 0.0002, "step": 19249 }, { "epoch": 2.31, "learning_rate": 1.6178764247150572e-06, "loss": 0.0003, "step": 19250 }, { "epoch": 2.31, "learning_rate": 1.6178044391121776e-06, "loss": 0.0001, "step": 19251 }, { "epoch": 2.31, "learning_rate": 1.6177324535092983e-06, "loss": 0.0003, "step": 19252 }, { "epoch": 2.31, "learning_rate": 1.6176604679064187e-06, "loss": 0.0045, "step": 19253 }, { "epoch": 2.31, "learning_rate": 1.6175884823035393e-06, "loss": 0.0004, "step": 19254 }, { "epoch": 2.31, "learning_rate": 1.61751649670066e-06, "loss": 0.0005, "step": 19255 }, { "epoch": 2.31, "learning_rate": 1.6174445110977804e-06, "loss": 0.792, "step": 19256 }, { "epoch": 2.31, "learning_rate": 1.6173725254949012e-06, "loss": 0.0001, "step": 19257 }, { "epoch": 2.31, "learning_rate": 1.6173005398920215e-06, "loss": 0.0004, "step": 19258 }, { "epoch": 2.31, "learning_rate": 1.6172285542891423e-06, "loss": 0.0005, "step": 19259 }, { "epoch": 2.31, "learning_rate": 1.6171565686862627e-06, "loss": 0.0018, "step": 19260 }, { "epoch": 2.31, "learning_rate": 1.6170845830833834e-06, "loss": 0.0009, "step": 19261 }, { "epoch": 2.31, "learning_rate": 1.617012597480504e-06, "loss": 0.0005, "step": 19262 }, { "epoch": 2.31, "learning_rate": 1.6169406118776246e-06, "loss": 0.0006, "step": 19263 }, { "epoch": 2.31, "learning_rate": 1.6168686262747451e-06, "loss": 0.0002, "step": 19264 }, { "epoch": 2.31, "learning_rate": 1.616796640671866e-06, "loss": 0.0003, "step": 19265 }, { "epoch": 2.31, "learning_rate": 1.6167246550689863e-06, "loss": 1.0817, "step": 19266 }, { "epoch": 2.31, "learning_rate": 1.6166526694661066e-06, "loss": 0.0003, "step": 19267 }, { "epoch": 2.31, "learning_rate": 1.6165806838632274e-06, "loss": 0.0006, "step": 19268 }, { "epoch": 2.31, "learning_rate": 1.616508698260348e-06, "loss": 0.0002, "step": 19269 }, { "epoch": 2.31, "learning_rate": 1.6164367126574685e-06, "loss": 0.0003, "step": 19270 }, { "epoch": 2.31, "learning_rate": 1.616364727054589e-06, "loss": 0.0005, "step": 19271 }, { "epoch": 2.31, "learning_rate": 1.6162927414517099e-06, "loss": 0.0002, "step": 19272 }, { "epoch": 2.31, "learning_rate": 1.6162207558488302e-06, "loss": 0.0001, "step": 19273 }, { "epoch": 2.31, "learning_rate": 1.616148770245951e-06, "loss": 0.0002, "step": 19274 }, { "epoch": 2.31, "learning_rate": 1.6160767846430714e-06, "loss": 0.0002, "step": 19275 }, { "epoch": 2.31, "learning_rate": 1.6160047990401921e-06, "loss": 0.0005, "step": 19276 }, { "epoch": 2.31, "learning_rate": 1.6159328134373125e-06, "loss": 0.0002, "step": 19277 }, { "epoch": 2.31, "learning_rate": 1.615860827834433e-06, "loss": 0.0001, "step": 19278 }, { "epoch": 2.31, "learning_rate": 1.6157888422315538e-06, "loss": 0.0004, "step": 19279 }, { "epoch": 2.31, "learning_rate": 1.6157168566286742e-06, "loss": 0.0003, "step": 19280 }, { "epoch": 2.31, "learning_rate": 1.615644871025795e-06, "loss": 0.0018, "step": 19281 }, { "epoch": 2.31, "learning_rate": 1.6155728854229153e-06, "loss": 0.0002, "step": 19282 }, { "epoch": 2.31, "learning_rate": 1.6155008998200361e-06, "loss": 0.002, "step": 19283 }, { "epoch": 2.31, "learning_rate": 1.6154289142171565e-06, "loss": 0.0002, "step": 19284 }, { "epoch": 2.31, "learning_rate": 1.6153569286142772e-06, "loss": 0.0002, "step": 19285 }, { "epoch": 2.31, "learning_rate": 1.6152849430113978e-06, "loss": 0.011, "step": 19286 }, { "epoch": 2.31, "learning_rate": 1.6152129574085186e-06, "loss": 0.002, "step": 19287 }, { "epoch": 2.31, "learning_rate": 1.615140971805639e-06, "loss": 0.1133, "step": 19288 }, { "epoch": 2.31, "learning_rate": 1.6150689862027593e-06, "loss": 0.0001, "step": 19289 }, { "epoch": 2.31, "learning_rate": 1.61499700059988e-06, "loss": 0.0002, "step": 19290 }, { "epoch": 2.31, "learning_rate": 1.6149250149970004e-06, "loss": 0.0002, "step": 19291 }, { "epoch": 2.31, "learning_rate": 1.6148530293941212e-06, "loss": 0.0, "step": 19292 }, { "epoch": 2.31, "learning_rate": 1.6147810437912418e-06, "loss": 0.0003, "step": 19293 }, { "epoch": 2.31, "learning_rate": 1.6147090581883623e-06, "loss": 0.0002, "step": 19294 }, { "epoch": 2.31, "learning_rate": 1.614637072585483e-06, "loss": 0.0002, "step": 19295 }, { "epoch": 2.31, "learning_rate": 1.6145650869826037e-06, "loss": 0.0003, "step": 19296 }, { "epoch": 2.31, "learning_rate": 1.614493101379724e-06, "loss": 0.0002, "step": 19297 }, { "epoch": 2.31, "learning_rate": 1.6144211157768448e-06, "loss": 0.0002, "step": 19298 }, { "epoch": 2.31, "learning_rate": 1.6143491301739652e-06, "loss": 0.5967, "step": 19299 }, { "epoch": 2.31, "learning_rate": 1.6142771445710857e-06, "loss": 0.0002, "step": 19300 }, { "epoch": 2.31, "learning_rate": 1.6142051589682063e-06, "loss": 0.0001, "step": 19301 }, { "epoch": 2.31, "learning_rate": 1.6141331733653269e-06, "loss": 0.0002, "step": 19302 }, { "epoch": 2.31, "learning_rate": 1.6140611877624477e-06, "loss": 0.0001, "step": 19303 }, { "epoch": 2.31, "learning_rate": 1.613989202159568e-06, "loss": 0.0003, "step": 19304 }, { "epoch": 2.31, "learning_rate": 1.6139172165566888e-06, "loss": 0.0002, "step": 19305 }, { "epoch": 2.31, "learning_rate": 1.6138452309538091e-06, "loss": 0.0003, "step": 19306 }, { "epoch": 2.31, "learning_rate": 1.61377324535093e-06, "loss": 0.0003, "step": 19307 }, { "epoch": 2.31, "learning_rate": 1.6137012597480503e-06, "loss": 0.0164, "step": 19308 }, { "epoch": 2.31, "learning_rate": 1.613629274145171e-06, "loss": 0.0002, "step": 19309 }, { "epoch": 2.31, "learning_rate": 1.6135572885422916e-06, "loss": 0.0002, "step": 19310 }, { "epoch": 2.31, "learning_rate": 1.6134853029394124e-06, "loss": 0.0002, "step": 19311 }, { "epoch": 2.31, "learning_rate": 1.6134133173365328e-06, "loss": 0.0002, "step": 19312 }, { "epoch": 2.31, "learning_rate": 1.6133413317336531e-06, "loss": 0.0001, "step": 19313 }, { "epoch": 2.31, "learning_rate": 1.613269346130774e-06, "loss": 0.0002, "step": 19314 }, { "epoch": 2.31, "learning_rate": 1.6131973605278943e-06, "loss": 0.0003, "step": 19315 }, { "epoch": 2.31, "learning_rate": 1.613125374925015e-06, "loss": 0.0002, "step": 19316 }, { "epoch": 2.31, "learning_rate": 1.6130533893221356e-06, "loss": 0.0004, "step": 19317 }, { "epoch": 2.31, "learning_rate": 1.6129814037192564e-06, "loss": 0.0002, "step": 19318 }, { "epoch": 2.32, "learning_rate": 1.6129094181163767e-06, "loss": 0.0002, "step": 19319 }, { "epoch": 2.32, "learning_rate": 1.6128374325134975e-06, "loss": 0.0001, "step": 19320 }, { "epoch": 2.32, "learning_rate": 1.6127654469106179e-06, "loss": 0.0027, "step": 19321 }, { "epoch": 2.32, "learning_rate": 1.6126934613077386e-06, "loss": 0.0098, "step": 19322 }, { "epoch": 2.32, "learning_rate": 1.612621475704859e-06, "loss": 0.0001, "step": 19323 }, { "epoch": 2.32, "learning_rate": 1.6125494901019796e-06, "loss": 0.0002, "step": 19324 }, { "epoch": 2.32, "learning_rate": 1.6124775044991003e-06, "loss": 0.0362, "step": 19325 }, { "epoch": 2.32, "learning_rate": 1.6124055188962207e-06, "loss": 0.0002, "step": 19326 }, { "epoch": 2.32, "learning_rate": 1.6123335332933415e-06, "loss": 0.0002, "step": 19327 }, { "epoch": 2.32, "learning_rate": 1.6122615476904618e-06, "loss": 0.0001, "step": 19328 }, { "epoch": 2.32, "learning_rate": 1.6121895620875826e-06, "loss": 0.0001, "step": 19329 }, { "epoch": 2.32, "learning_rate": 1.612117576484703e-06, "loss": 0.0002, "step": 19330 }, { "epoch": 2.32, "learning_rate": 1.6120455908818237e-06, "loss": 0.0002, "step": 19331 }, { "epoch": 2.32, "learning_rate": 1.6119736052789443e-06, "loss": 0.0003, "step": 19332 }, { "epoch": 2.32, "learning_rate": 1.6119016196760649e-06, "loss": 0.0001, "step": 19333 }, { "epoch": 2.32, "learning_rate": 1.6118296340731854e-06, "loss": 0.0004, "step": 19334 }, { "epoch": 2.32, "learning_rate": 1.6117576484703058e-06, "loss": 0.0006, "step": 19335 }, { "epoch": 2.32, "learning_rate": 1.6116856628674266e-06, "loss": 0.0366, "step": 19336 }, { "epoch": 2.32, "learning_rate": 1.611613677264547e-06, "loss": 0.0267, "step": 19337 }, { "epoch": 2.32, "learning_rate": 1.6115416916616677e-06, "loss": 0.0002, "step": 19338 }, { "epoch": 2.32, "learning_rate": 1.6114697060587883e-06, "loss": 0.0001, "step": 19339 }, { "epoch": 2.32, "learning_rate": 1.6113977204559088e-06, "loss": 0.0001, "step": 19340 }, { "epoch": 2.32, "learning_rate": 1.6113257348530294e-06, "loss": 0.0004, "step": 19341 }, { "epoch": 2.32, "learning_rate": 1.6112537492501502e-06, "loss": 0.0001, "step": 19342 }, { "epoch": 2.32, "learning_rate": 1.6111817636472705e-06, "loss": 0.0001, "step": 19343 }, { "epoch": 2.32, "learning_rate": 1.6111097780443913e-06, "loss": 0.0001, "step": 19344 }, { "epoch": 2.32, "learning_rate": 1.6110377924415117e-06, "loss": 0.0005, "step": 19345 }, { "epoch": 2.32, "learning_rate": 1.6109658068386322e-06, "loss": 0.0003, "step": 19346 }, { "epoch": 2.32, "learning_rate": 1.6108938212357528e-06, "loss": 0.0005, "step": 19347 }, { "epoch": 2.32, "learning_rate": 1.6108218356328734e-06, "loss": 0.0004, "step": 19348 }, { "epoch": 2.32, "learning_rate": 1.6107498500299942e-06, "loss": 0.0002, "step": 19349 }, { "epoch": 2.32, "learning_rate": 1.6106778644271145e-06, "loss": 0.0002, "step": 19350 }, { "epoch": 2.32, "learning_rate": 1.6106058788242353e-06, "loss": 0.0001, "step": 19351 }, { "epoch": 2.32, "learning_rate": 1.6105338932213556e-06, "loss": 0.0001, "step": 19352 }, { "epoch": 2.32, "learning_rate": 1.6104619076184764e-06, "loss": 0.0002, "step": 19353 }, { "epoch": 2.32, "learning_rate": 1.6103899220155968e-06, "loss": 0.0002, "step": 19354 }, { "epoch": 2.32, "learning_rate": 1.6103179364127176e-06, "loss": 0.0002, "step": 19355 }, { "epoch": 2.32, "learning_rate": 1.6102459508098381e-06, "loss": 0.0003, "step": 19356 }, { "epoch": 2.32, "learning_rate": 1.610173965206959e-06, "loss": 0.0004, "step": 19357 }, { "epoch": 2.32, "learning_rate": 1.6101019796040793e-06, "loss": 0.0002, "step": 19358 }, { "epoch": 2.32, "learning_rate": 1.6100299940011996e-06, "loss": 0.0005, "step": 19359 }, { "epoch": 2.32, "learning_rate": 1.6099580083983204e-06, "loss": 0.0004, "step": 19360 }, { "epoch": 2.32, "learning_rate": 1.6098860227954407e-06, "loss": 0.0001, "step": 19361 }, { "epoch": 2.32, "learning_rate": 1.6098140371925615e-06, "loss": 0.0001, "step": 19362 }, { "epoch": 2.32, "learning_rate": 1.609742051589682e-06, "loss": 0.0001, "step": 19363 }, { "epoch": 2.32, "learning_rate": 1.6096700659868029e-06, "loss": 0.0004, "step": 19364 }, { "epoch": 2.32, "learning_rate": 1.6095980803839232e-06, "loss": 0.0002, "step": 19365 }, { "epoch": 2.32, "learning_rate": 1.609526094781044e-06, "loss": 0.0004, "step": 19366 }, { "epoch": 2.32, "learning_rate": 1.6094541091781644e-06, "loss": 0.0003, "step": 19367 }, { "epoch": 2.32, "learning_rate": 1.6093821235752851e-06, "loss": 0.0003, "step": 19368 }, { "epoch": 2.32, "learning_rate": 1.6093101379724055e-06, "loss": 0.0003, "step": 19369 }, { "epoch": 2.32, "learning_rate": 1.609238152369526e-06, "loss": 0.0, "step": 19370 }, { "epoch": 2.32, "learning_rate": 1.6091661667666468e-06, "loss": 0.0001, "step": 19371 }, { "epoch": 2.32, "learning_rate": 1.6090941811637672e-06, "loss": 0.0001, "step": 19372 }, { "epoch": 2.32, "learning_rate": 1.609022195560888e-06, "loss": 0.001, "step": 19373 }, { "epoch": 2.32, "learning_rate": 1.6089502099580083e-06, "loss": 0.0001, "step": 19374 }, { "epoch": 2.32, "learning_rate": 1.608878224355129e-06, "loss": 0.0001, "step": 19375 }, { "epoch": 2.32, "learning_rate": 1.6088062387522495e-06, "loss": 0.0002, "step": 19376 }, { "epoch": 2.32, "learning_rate": 1.6087342531493702e-06, "loss": 0.0001, "step": 19377 }, { "epoch": 2.32, "learning_rate": 1.6086622675464908e-06, "loss": 0.7144, "step": 19378 }, { "epoch": 2.32, "learning_rate": 1.6085902819436114e-06, "loss": 0.0002, "step": 19379 }, { "epoch": 2.32, "learning_rate": 1.608518296340732e-06, "loss": 0.0002, "step": 19380 }, { "epoch": 2.32, "learning_rate": 1.6084463107378523e-06, "loss": 0.0001, "step": 19381 }, { "epoch": 2.32, "learning_rate": 1.608374325134973e-06, "loss": 0.0002, "step": 19382 }, { "epoch": 2.32, "learning_rate": 1.6083023395320934e-06, "loss": 0.0003, "step": 19383 }, { "epoch": 2.32, "learning_rate": 1.6082303539292142e-06, "loss": 0.0002, "step": 19384 }, { "epoch": 2.32, "learning_rate": 1.6081583683263348e-06, "loss": 0.0002, "step": 19385 }, { "epoch": 2.32, "learning_rate": 1.6080863827234553e-06, "loss": 0.0001, "step": 19386 }, { "epoch": 2.32, "learning_rate": 1.608014397120576e-06, "loss": 0.0003, "step": 19387 }, { "epoch": 2.32, "learning_rate": 1.6079424115176967e-06, "loss": 0.0001, "step": 19388 }, { "epoch": 2.32, "learning_rate": 1.607870425914817e-06, "loss": 0.0003, "step": 19389 }, { "epoch": 2.32, "learning_rate": 1.6077984403119378e-06, "loss": 0.0, "step": 19390 }, { "epoch": 2.32, "learning_rate": 1.6077264547090582e-06, "loss": 0.1449, "step": 19391 }, { "epoch": 2.32, "learning_rate": 1.607654469106179e-06, "loss": 0.0002, "step": 19392 }, { "epoch": 2.32, "learning_rate": 1.6075824835032993e-06, "loss": 0.0001, "step": 19393 }, { "epoch": 2.32, "learning_rate": 1.6075104979004199e-06, "loss": 0.0001, "step": 19394 }, { "epoch": 2.32, "learning_rate": 1.6074385122975407e-06, "loss": 0.4744, "step": 19395 }, { "epoch": 2.32, "learning_rate": 1.607366526694661e-06, "loss": 0.0002, "step": 19396 }, { "epoch": 2.32, "learning_rate": 1.6072945410917818e-06, "loss": 0.0001, "step": 19397 }, { "epoch": 2.32, "learning_rate": 1.6072225554889021e-06, "loss": 0.0002, "step": 19398 }, { "epoch": 2.32, "learning_rate": 1.607150569886023e-06, "loss": 0.0004, "step": 19399 }, { "epoch": 2.32, "learning_rate": 1.6070785842831433e-06, "loss": 0.0002, "step": 19400 }, { "epoch": 2.32, "learning_rate": 1.607006598680264e-06, "loss": 1.0697, "step": 19401 }, { "epoch": 2.32, "learning_rate": 1.6069346130773846e-06, "loss": 0.0002, "step": 19402 }, { "epoch": 2.33, "learning_rate": 1.6068626274745054e-06, "loss": 0.0002, "step": 19403 }, { "epoch": 2.33, "learning_rate": 1.6067906418716258e-06, "loss": 0.0001, "step": 19404 }, { "epoch": 2.33, "learning_rate": 1.6067186562687461e-06, "loss": 0.0001, "step": 19405 }, { "epoch": 2.33, "learning_rate": 1.6066466706658669e-06, "loss": 0.0002, "step": 19406 }, { "epoch": 2.33, "learning_rate": 1.6065746850629872e-06, "loss": 0.0001, "step": 19407 }, { "epoch": 2.33, "learning_rate": 1.606502699460108e-06, "loss": 0.0002, "step": 19408 }, { "epoch": 2.33, "learning_rate": 1.6064307138572286e-06, "loss": 0.0002, "step": 19409 }, { "epoch": 2.33, "learning_rate": 1.6063587282543494e-06, "loss": 0.0001, "step": 19410 }, { "epoch": 2.33, "learning_rate": 1.6062867426514697e-06, "loss": 0.4807, "step": 19411 }, { "epoch": 2.33, "learning_rate": 1.6062147570485905e-06, "loss": 0.0001, "step": 19412 }, { "epoch": 2.33, "learning_rate": 1.6061427714457109e-06, "loss": 0.0001, "step": 19413 }, { "epoch": 2.33, "learning_rate": 1.6060707858428316e-06, "loss": 0.1494, "step": 19414 }, { "epoch": 2.33, "learning_rate": 1.605998800239952e-06, "loss": 0.0002, "step": 19415 }, { "epoch": 2.33, "learning_rate": 1.6059268146370726e-06, "loss": 0.0001, "step": 19416 }, { "epoch": 2.33, "learning_rate": 1.6058548290341931e-06, "loss": 0.0003, "step": 19417 }, { "epoch": 2.33, "learning_rate": 1.6057828434313137e-06, "loss": 0.0001, "step": 19418 }, { "epoch": 2.33, "learning_rate": 1.6057108578284345e-06, "loss": 0.0002, "step": 19419 }, { "epoch": 2.33, "learning_rate": 1.6056388722255548e-06, "loss": 0.0003, "step": 19420 }, { "epoch": 2.33, "learning_rate": 1.6055668866226756e-06, "loss": 0.0006, "step": 19421 }, { "epoch": 2.33, "learning_rate": 1.605494901019796e-06, "loss": 0.0003, "step": 19422 }, { "epoch": 2.33, "learning_rate": 1.6054229154169167e-06, "loss": 0.0002, "step": 19423 }, { "epoch": 2.33, "learning_rate": 1.605350929814037e-06, "loss": 0.0003, "step": 19424 }, { "epoch": 2.33, "learning_rate": 1.6052789442111579e-06, "loss": 0.0001, "step": 19425 }, { "epoch": 2.33, "learning_rate": 1.6052069586082784e-06, "loss": 0.0001, "step": 19426 }, { "epoch": 2.33, "learning_rate": 1.6051349730053988e-06, "loss": 0.0002, "step": 19427 }, { "epoch": 2.33, "learning_rate": 1.6050629874025196e-06, "loss": 0.0003, "step": 19428 }, { "epoch": 2.33, "learning_rate": 1.60499100179964e-06, "loss": 0.0002, "step": 19429 }, { "epoch": 2.33, "learning_rate": 1.6049190161967607e-06, "loss": 0.0003, "step": 19430 }, { "epoch": 2.33, "learning_rate": 1.604847030593881e-06, "loss": 0.0002, "step": 19431 }, { "epoch": 2.33, "learning_rate": 1.6047750449910018e-06, "loss": 0.0001, "step": 19432 }, { "epoch": 2.33, "learning_rate": 1.6047030593881224e-06, "loss": 0.0001, "step": 19433 }, { "epoch": 2.33, "learning_rate": 1.6046310737852432e-06, "loss": 0.0002, "step": 19434 }, { "epoch": 2.33, "learning_rate": 1.6045590881823635e-06, "loss": 0.0001, "step": 19435 }, { "epoch": 2.33, "learning_rate": 1.6044871025794843e-06, "loss": 0.0004, "step": 19436 }, { "epoch": 2.33, "learning_rate": 1.6044151169766047e-06, "loss": 0.0004, "step": 19437 }, { "epoch": 2.33, "learning_rate": 1.6043431313737254e-06, "loss": 0.0001, "step": 19438 }, { "epoch": 2.33, "learning_rate": 1.6042711457708458e-06, "loss": 0.0001, "step": 19439 }, { "epoch": 2.33, "learning_rate": 1.6041991601679664e-06, "loss": 0.0001, "step": 19440 }, { "epoch": 2.33, "learning_rate": 1.6041271745650871e-06, "loss": 0.0002, "step": 19441 }, { "epoch": 2.33, "learning_rate": 1.6040551889622075e-06, "loss": 0.0001, "step": 19442 }, { "epoch": 2.33, "learning_rate": 1.6039832033593283e-06, "loss": 0.0002, "step": 19443 }, { "epoch": 2.33, "learning_rate": 1.6039112177564486e-06, "loss": 0.0001, "step": 19444 }, { "epoch": 2.33, "learning_rate": 1.6038392321535694e-06, "loss": 0.0002, "step": 19445 }, { "epoch": 2.33, "learning_rate": 1.6037672465506898e-06, "loss": 0.0006, "step": 19446 }, { "epoch": 2.33, "learning_rate": 1.6036952609478105e-06, "loss": 0.0014, "step": 19447 }, { "epoch": 2.33, "learning_rate": 1.6036232753449311e-06, "loss": 0.0001, "step": 19448 }, { "epoch": 2.33, "learning_rate": 1.6035512897420517e-06, "loss": 0.0002, "step": 19449 }, { "epoch": 2.33, "learning_rate": 1.6034793041391723e-06, "loss": 0.0003, "step": 19450 }, { "epoch": 2.33, "learning_rate": 1.6034073185362926e-06, "loss": 0.0002, "step": 19451 }, { "epoch": 2.33, "learning_rate": 1.6033353329334134e-06, "loss": 0.0002, "step": 19452 }, { "epoch": 2.33, "learning_rate": 1.6032633473305337e-06, "loss": 0.0001, "step": 19453 }, { "epoch": 2.33, "learning_rate": 1.6031913617276545e-06, "loss": 0.0002, "step": 19454 }, { "epoch": 2.33, "learning_rate": 1.603119376124775e-06, "loss": 0.0002, "step": 19455 }, { "epoch": 2.33, "learning_rate": 1.6030473905218957e-06, "loss": 0.0002, "step": 19456 }, { "epoch": 2.33, "learning_rate": 1.6029754049190162e-06, "loss": 0.0001, "step": 19457 }, { "epoch": 2.33, "learning_rate": 1.602903419316137e-06, "loss": 0.0002, "step": 19458 }, { "epoch": 2.33, "learning_rate": 1.6028314337132574e-06, "loss": 0.0001, "step": 19459 }, { "epoch": 2.33, "learning_rate": 1.6027594481103781e-06, "loss": 0.0002, "step": 19460 }, { "epoch": 2.33, "learning_rate": 1.6026874625074985e-06, "loss": 0.0001, "step": 19461 }, { "epoch": 2.33, "learning_rate": 1.602615476904619e-06, "loss": 0.0001, "step": 19462 }, { "epoch": 2.33, "learning_rate": 1.6025434913017396e-06, "loss": 0.0002, "step": 19463 }, { "epoch": 2.33, "learning_rate": 1.6024715056988602e-06, "loss": 0.0001, "step": 19464 }, { "epoch": 2.33, "learning_rate": 1.602399520095981e-06, "loss": 0.0002, "step": 19465 }, { "epoch": 2.33, "learning_rate": 1.6023275344931013e-06, "loss": 0.0001, "step": 19466 }, { "epoch": 2.33, "learning_rate": 1.602255548890222e-06, "loss": 0.0002, "step": 19467 }, { "epoch": 2.33, "learning_rate": 1.6021835632873425e-06, "loss": 0.0001, "step": 19468 }, { "epoch": 2.33, "learning_rate": 1.6021115776844632e-06, "loss": 0.0002, "step": 19469 }, { "epoch": 2.33, "learning_rate": 1.6020395920815836e-06, "loss": 0.0001, "step": 19470 }, { "epoch": 2.33, "learning_rate": 1.6019676064787044e-06, "loss": 0.4991, "step": 19471 }, { "epoch": 2.33, "learning_rate": 1.601895620875825e-06, "loss": 0.0002, "step": 19472 }, { "epoch": 2.33, "learning_rate": 1.6018236352729453e-06, "loss": 0.0003, "step": 19473 }, { "epoch": 2.33, "learning_rate": 1.601751649670066e-06, "loss": 0.0001, "step": 19474 }, { "epoch": 2.33, "learning_rate": 1.6016796640671864e-06, "loss": 0.0003, "step": 19475 }, { "epoch": 2.33, "learning_rate": 1.6016076784643072e-06, "loss": 0.0001, "step": 19476 }, { "epoch": 2.33, "learning_rate": 1.6015356928614276e-06, "loss": 0.0001, "step": 19477 }, { "epoch": 2.33, "learning_rate": 1.6014637072585483e-06, "loss": 0.0002, "step": 19478 }, { "epoch": 2.33, "learning_rate": 1.601391721655669e-06, "loss": 0.0001, "step": 19479 }, { "epoch": 2.33, "learning_rate": 1.6013197360527897e-06, "loss": 0.0001, "step": 19480 }, { "epoch": 2.33, "learning_rate": 1.60124775044991e-06, "loss": 0.0001, "step": 19481 }, { "epoch": 2.33, "learning_rate": 1.6011757648470308e-06, "loss": 0.0003, "step": 19482 }, { "epoch": 2.33, "learning_rate": 1.6011037792441512e-06, "loss": 0.0001, "step": 19483 }, { "epoch": 2.33, "learning_rate": 1.601031793641272e-06, "loss": 0.0001, "step": 19484 }, { "epoch": 2.33, "learning_rate": 1.6009598080383923e-06, "loss": 1.0086, "step": 19485 }, { "epoch": 2.34, "learning_rate": 1.6008878224355129e-06, "loss": 0.0001, "step": 19486 }, { "epoch": 2.34, "learning_rate": 1.6008158368326336e-06, "loss": 0.0001, "step": 19487 }, { "epoch": 2.34, "learning_rate": 1.600743851229754e-06, "loss": 0.0001, "step": 19488 }, { "epoch": 2.34, "learning_rate": 1.6006718656268748e-06, "loss": 0.0001, "step": 19489 }, { "epoch": 2.34, "learning_rate": 1.6005998800239951e-06, "loss": 0.0002, "step": 19490 }, { "epoch": 2.34, "learning_rate": 1.600527894421116e-06, "loss": 0.0002, "step": 19491 }, { "epoch": 2.34, "learning_rate": 1.6004559088182363e-06, "loss": 0.0002, "step": 19492 }, { "epoch": 2.34, "learning_rate": 1.600383923215357e-06, "loss": 0.0001, "step": 19493 }, { "epoch": 2.34, "learning_rate": 1.6003119376124776e-06, "loss": 0.0129, "step": 19494 }, { "epoch": 2.34, "learning_rate": 1.6002399520095982e-06, "loss": 0.0001, "step": 19495 }, { "epoch": 2.34, "learning_rate": 1.6001679664067187e-06, "loss": 0.0002, "step": 19496 }, { "epoch": 2.34, "learning_rate": 1.600095980803839e-06, "loss": 0.0001, "step": 19497 }, { "epoch": 2.34, "learning_rate": 1.6000239952009599e-06, "loss": 0.0001, "step": 19498 }, { "epoch": 2.34, "learning_rate": 1.5999520095980802e-06, "loss": 0.0002, "step": 19499 }, { "epoch": 2.34, "learning_rate": 1.599880023995201e-06, "loss": 0.5741, "step": 19500 }, { "epoch": 2.34, "learning_rate": 1.5998080383923216e-06, "loss": 0.0002, "step": 19501 }, { "epoch": 2.34, "learning_rate": 1.5997360527894421e-06, "loss": 0.0001, "step": 19502 }, { "epoch": 2.34, "learning_rate": 1.5996640671865627e-06, "loss": 0.0002, "step": 19503 }, { "epoch": 2.34, "learning_rate": 1.5995920815836835e-06, "loss": 0.0003, "step": 19504 }, { "epoch": 2.34, "learning_rate": 1.5995200959808038e-06, "loss": 0.0001, "step": 19505 }, { "epoch": 2.34, "learning_rate": 1.5994481103779246e-06, "loss": 0.0003, "step": 19506 }, { "epoch": 2.34, "learning_rate": 1.599376124775045e-06, "loss": 0.0001, "step": 19507 }, { "epoch": 2.34, "learning_rate": 1.5993041391721655e-06, "loss": 0.0003, "step": 19508 }, { "epoch": 2.34, "learning_rate": 1.5992321535692861e-06, "loss": 0.0002, "step": 19509 }, { "epoch": 2.34, "learning_rate": 1.5991601679664067e-06, "loss": 0.0002, "step": 19510 }, { "epoch": 2.34, "learning_rate": 1.5990881823635275e-06, "loss": 0.0002, "step": 19511 }, { "epoch": 2.34, "learning_rate": 1.5990161967606478e-06, "loss": 0.0004, "step": 19512 }, { "epoch": 2.34, "learning_rate": 1.5989442111577686e-06, "loss": 0.0002, "step": 19513 }, { "epoch": 2.34, "learning_rate": 1.598872225554889e-06, "loss": 0.0001, "step": 19514 }, { "epoch": 2.34, "learning_rate": 1.5988002399520097e-06, "loss": 0.0002, "step": 19515 }, { "epoch": 2.34, "learning_rate": 1.59872825434913e-06, "loss": 0.0002, "step": 19516 }, { "epoch": 2.34, "learning_rate": 1.5986562687462509e-06, "loss": 0.0001, "step": 19517 }, { "epoch": 2.34, "learning_rate": 1.5985842831433714e-06, "loss": 0.0001, "step": 19518 }, { "epoch": 2.34, "learning_rate": 1.5985122975404922e-06, "loss": 0.0002, "step": 19519 }, { "epoch": 2.34, "learning_rate": 1.5984403119376126e-06, "loss": 0.0003, "step": 19520 }, { "epoch": 2.34, "learning_rate": 1.598368326334733e-06, "loss": 0.0004, "step": 19521 }, { "epoch": 2.34, "learning_rate": 1.5982963407318537e-06, "loss": 0.0178, "step": 19522 }, { "epoch": 2.34, "learning_rate": 1.598224355128974e-06, "loss": 0.0003, "step": 19523 }, { "epoch": 2.34, "learning_rate": 1.5981523695260948e-06, "loss": 0.0012, "step": 19524 }, { "epoch": 2.34, "learning_rate": 1.5980803839232154e-06, "loss": 0.0002, "step": 19525 }, { "epoch": 2.34, "learning_rate": 1.5980083983203362e-06, "loss": 0.0001, "step": 19526 }, { "epoch": 2.34, "learning_rate": 1.5979364127174565e-06, "loss": 0.0001, "step": 19527 }, { "epoch": 2.34, "learning_rate": 1.5978644271145773e-06, "loss": 0.0002, "step": 19528 }, { "epoch": 2.34, "learning_rate": 1.5977924415116977e-06, "loss": 0.0002, "step": 19529 }, { "epoch": 2.34, "learning_rate": 1.5977204559088184e-06, "loss": 0.0002, "step": 19530 }, { "epoch": 2.34, "learning_rate": 1.5976484703059388e-06, "loss": 0.0002, "step": 19531 }, { "epoch": 2.34, "learning_rate": 1.5975764847030594e-06, "loss": 0.0002, "step": 19532 }, { "epoch": 2.34, "learning_rate": 1.5975044991001801e-06, "loss": 0.0003, "step": 19533 }, { "epoch": 2.34, "learning_rate": 1.5974325134973005e-06, "loss": 0.0002, "step": 19534 }, { "epoch": 2.34, "learning_rate": 1.5973605278944213e-06, "loss": 0.0002, "step": 19535 }, { "epoch": 2.34, "learning_rate": 1.5972885422915416e-06, "loss": 0.0002, "step": 19536 }, { "epoch": 2.34, "learning_rate": 1.5972165566886624e-06, "loss": 0.0002, "step": 19537 }, { "epoch": 2.34, "learning_rate": 1.5971445710857828e-06, "loss": 0.0001, "step": 19538 }, { "epoch": 2.34, "learning_rate": 1.5970725854829035e-06, "loss": 0.0003, "step": 19539 }, { "epoch": 2.34, "learning_rate": 1.5970005998800241e-06, "loss": 0.0002, "step": 19540 }, { "epoch": 2.34, "learning_rate": 1.5969286142771447e-06, "loss": 0.0002, "step": 19541 }, { "epoch": 2.34, "learning_rate": 1.5968566286742652e-06, "loss": 0.015, "step": 19542 }, { "epoch": 2.34, "learning_rate": 1.5967846430713856e-06, "loss": 0.0001, "step": 19543 }, { "epoch": 2.34, "learning_rate": 1.5967126574685064e-06, "loss": 0.0002, "step": 19544 }, { "epoch": 2.34, "learning_rate": 1.5966406718656267e-06, "loss": 0.0002, "step": 19545 }, { "epoch": 2.34, "learning_rate": 1.5965686862627475e-06, "loss": 0.0002, "step": 19546 }, { "epoch": 2.34, "learning_rate": 1.5964967006598679e-06, "loss": 0.0003, "step": 19547 }, { "epoch": 2.34, "learning_rate": 1.5964247150569886e-06, "loss": 0.0002, "step": 19548 }, { "epoch": 2.34, "learning_rate": 1.5963527294541092e-06, "loss": 0.0001, "step": 19549 }, { "epoch": 2.34, "learning_rate": 1.59628074385123e-06, "loss": 0.0001, "step": 19550 }, { "epoch": 2.34, "learning_rate": 1.5962087582483503e-06, "loss": 0.0002, "step": 19551 }, { "epoch": 2.34, "learning_rate": 1.5961367726454711e-06, "loss": 0.0001, "step": 19552 }, { "epoch": 2.34, "learning_rate": 1.5960647870425915e-06, "loss": 0.0002, "step": 19553 }, { "epoch": 2.34, "learning_rate": 1.5959928014397118e-06, "loss": 0.0004, "step": 19554 }, { "epoch": 2.34, "learning_rate": 1.5959208158368326e-06, "loss": 0.0001, "step": 19555 }, { "epoch": 2.34, "learning_rate": 1.5958488302339532e-06, "loss": 0.0001, "step": 19556 }, { "epoch": 2.34, "learning_rate": 1.595776844631074e-06, "loss": 0.0076, "step": 19557 }, { "epoch": 2.34, "learning_rate": 1.5957048590281943e-06, "loss": 0.0001, "step": 19558 }, { "epoch": 2.34, "learning_rate": 1.595632873425315e-06, "loss": 0.0001, "step": 19559 }, { "epoch": 2.34, "learning_rate": 1.5955608878224354e-06, "loss": 0.0002, "step": 19560 }, { "epoch": 2.34, "learning_rate": 1.5954889022195562e-06, "loss": 0.0002, "step": 19561 }, { "epoch": 2.34, "learning_rate": 1.5954169166166766e-06, "loss": 0.0002, "step": 19562 }, { "epoch": 2.34, "learning_rate": 1.5953449310137974e-06, "loss": 0.0012, "step": 19563 }, { "epoch": 2.34, "learning_rate": 1.595272945410918e-06, "loss": 0.0001, "step": 19564 }, { "epoch": 2.34, "learning_rate": 1.5952009598080385e-06, "loss": 0.1089, "step": 19565 }, { "epoch": 2.34, "learning_rate": 1.595128974205159e-06, "loss": 0.0002, "step": 19566 }, { "epoch": 2.34, "learning_rate": 1.5950569886022794e-06, "loss": 0.0002, "step": 19567 }, { "epoch": 2.34, "learning_rate": 1.5949850029994002e-06, "loss": 0.0002, "step": 19568 }, { "epoch": 2.34, "learning_rate": 1.5949130173965205e-06, "loss": 0.0002, "step": 19569 }, { "epoch": 2.35, "learning_rate": 1.5948410317936413e-06, "loss": 0.001, "step": 19570 }, { "epoch": 2.35, "learning_rate": 1.5947690461907619e-06, "loss": 0.0028, "step": 19571 }, { "epoch": 2.35, "learning_rate": 1.5946970605878825e-06, "loss": 0.0003, "step": 19572 }, { "epoch": 2.35, "learning_rate": 1.594625074985003e-06, "loss": 0.0001, "step": 19573 }, { "epoch": 2.35, "learning_rate": 1.5945530893821238e-06, "loss": 0.0001, "step": 19574 }, { "epoch": 2.35, "learning_rate": 1.5944811037792442e-06, "loss": 0.0002, "step": 19575 }, { "epoch": 2.35, "learning_rate": 1.594409118176365e-06, "loss": 0.0001, "step": 19576 }, { "epoch": 2.35, "learning_rate": 1.5943371325734853e-06, "loss": 0.0002, "step": 19577 }, { "epoch": 2.35, "learning_rate": 1.5942651469706059e-06, "loss": 0.0002, "step": 19578 }, { "epoch": 2.35, "learning_rate": 1.5941931613677264e-06, "loss": 0.0001, "step": 19579 }, { "epoch": 2.35, "learning_rate": 1.594121175764847e-06, "loss": 0.0001, "step": 19580 }, { "epoch": 2.35, "learning_rate": 1.5940491901619678e-06, "loss": 0.0001, "step": 19581 }, { "epoch": 2.35, "learning_rate": 1.5939772045590881e-06, "loss": 0.0002, "step": 19582 }, { "epoch": 2.35, "learning_rate": 1.593905218956209e-06, "loss": 0.0001, "step": 19583 }, { "epoch": 2.35, "learning_rate": 1.5938332333533293e-06, "loss": 0.0033, "step": 19584 }, { "epoch": 2.35, "learning_rate": 1.59376124775045e-06, "loss": 0.0002, "step": 19585 }, { "epoch": 2.35, "learning_rate": 1.5936892621475704e-06, "loss": 0.0001, "step": 19586 }, { "epoch": 2.35, "learning_rate": 1.5936172765446912e-06, "loss": 0.0004, "step": 19587 }, { "epoch": 2.35, "learning_rate": 1.5935452909418117e-06, "loss": 0.0001, "step": 19588 }, { "epoch": 2.35, "learning_rate": 1.593473305338932e-06, "loss": 0.0001, "step": 19589 }, { "epoch": 2.35, "learning_rate": 1.5934013197360529e-06, "loss": 0.0002, "step": 19590 }, { "epoch": 2.35, "learning_rate": 1.5933293341331732e-06, "loss": 0.0001, "step": 19591 }, { "epoch": 2.35, "learning_rate": 1.593257348530294e-06, "loss": 0.0002, "step": 19592 }, { "epoch": 2.35, "learning_rate": 1.5931853629274144e-06, "loss": 0.0001, "step": 19593 }, { "epoch": 2.35, "learning_rate": 1.5931133773245351e-06, "loss": 0.0002, "step": 19594 }, { "epoch": 2.35, "learning_rate": 1.5930413917216557e-06, "loss": 0.0001, "step": 19595 }, { "epoch": 2.35, "learning_rate": 1.5929694061187765e-06, "loss": 0.0001, "step": 19596 }, { "epoch": 2.35, "learning_rate": 1.5928974205158968e-06, "loss": 0.0003, "step": 19597 }, { "epoch": 2.35, "learning_rate": 1.5928254349130176e-06, "loss": 0.0002, "step": 19598 }, { "epoch": 2.35, "learning_rate": 1.592753449310138e-06, "loss": 0.0002, "step": 19599 }, { "epoch": 2.35, "learning_rate": 1.5926814637072583e-06, "loss": 0.0001, "step": 19600 }, { "epoch": 2.35, "learning_rate": 1.5926094781043791e-06, "loss": 0.0002, "step": 19601 }, { "epoch": 2.35, "learning_rate": 1.5925374925014997e-06, "loss": 0.0001, "step": 19602 }, { "epoch": 2.35, "learning_rate": 1.5924655068986205e-06, "loss": 0.0002, "step": 19603 }, { "epoch": 2.35, "learning_rate": 1.5923935212957408e-06, "loss": 1.1595, "step": 19604 }, { "epoch": 2.35, "learning_rate": 1.5923215356928616e-06, "loss": 0.0005, "step": 19605 }, { "epoch": 2.35, "learning_rate": 1.592249550089982e-06, "loss": 0.0009, "step": 19606 }, { "epoch": 2.35, "learning_rate": 1.5921775644871027e-06, "loss": 0.0001, "step": 19607 }, { "epoch": 2.35, "learning_rate": 1.592105578884223e-06, "loss": 0.0002, "step": 19608 }, { "epoch": 2.35, "learning_rate": 1.5920335932813439e-06, "loss": 0.0076, "step": 19609 }, { "epoch": 2.35, "learning_rate": 1.5919616076784644e-06, "loss": 0.0001, "step": 19610 }, { "epoch": 2.35, "learning_rate": 1.591889622075585e-06, "loss": 0.0002, "step": 19611 }, { "epoch": 2.35, "learning_rate": 1.5918176364727056e-06, "loss": 0.0002, "step": 19612 }, { "epoch": 2.35, "learning_rate": 1.591745650869826e-06, "loss": 0.0001, "step": 19613 }, { "epoch": 2.35, "learning_rate": 1.5916736652669467e-06, "loss": 0.0003, "step": 19614 }, { "epoch": 2.35, "learning_rate": 1.591601679664067e-06, "loss": 0.0001, "step": 19615 }, { "epoch": 2.35, "learning_rate": 1.5915296940611878e-06, "loss": 0.0001, "step": 19616 }, { "epoch": 2.35, "learning_rate": 1.5914577084583084e-06, "loss": 0.0, "step": 19617 }, { "epoch": 2.35, "learning_rate": 1.591385722855429e-06, "loss": 0.0014, "step": 19618 }, { "epoch": 2.35, "learning_rate": 1.5913137372525495e-06, "loss": 0.0002, "step": 19619 }, { "epoch": 2.35, "learning_rate": 1.5912417516496703e-06, "loss": 0.0002, "step": 19620 }, { "epoch": 2.35, "learning_rate": 1.5911697660467907e-06, "loss": 0.0001, "step": 19621 }, { "epoch": 2.35, "learning_rate": 1.5910977804439114e-06, "loss": 0.0002, "step": 19622 }, { "epoch": 2.35, "learning_rate": 1.5910257948410318e-06, "loss": 0.0002, "step": 19623 }, { "epoch": 2.35, "learning_rate": 1.5909538092381524e-06, "loss": 0.0001, "step": 19624 }, { "epoch": 2.35, "learning_rate": 1.590881823635273e-06, "loss": 0.0004, "step": 19625 }, { "epoch": 2.35, "learning_rate": 1.5908098380323935e-06, "loss": 0.0002, "step": 19626 }, { "epoch": 2.35, "learning_rate": 1.5907378524295143e-06, "loss": 0.0003, "step": 19627 }, { "epoch": 2.35, "learning_rate": 1.5906658668266346e-06, "loss": 0.1261, "step": 19628 }, { "epoch": 2.35, "learning_rate": 1.5905938812237554e-06, "loss": 0.0002, "step": 19629 }, { "epoch": 2.35, "learning_rate": 1.5905218956208758e-06, "loss": 0.0002, "step": 19630 }, { "epoch": 2.35, "learning_rate": 1.5904499100179965e-06, "loss": 0.0002, "step": 19631 }, { "epoch": 2.35, "learning_rate": 1.5903779244151169e-06, "loss": 0.0002, "step": 19632 }, { "epoch": 2.35, "learning_rate": 1.5903059388122377e-06, "loss": 0.0007, "step": 19633 }, { "epoch": 2.35, "learning_rate": 1.5902339532093582e-06, "loss": 0.0005, "step": 19634 }, { "epoch": 2.35, "learning_rate": 1.5901619676064786e-06, "loss": 0.0001, "step": 19635 }, { "epoch": 2.35, "learning_rate": 1.5900899820035994e-06, "loss": 0.7793, "step": 19636 }, { "epoch": 2.35, "learning_rate": 1.5900179964007197e-06, "loss": 0.0001, "step": 19637 }, { "epoch": 2.35, "learning_rate": 1.5899460107978405e-06, "loss": 0.0001, "step": 19638 }, { "epoch": 2.35, "learning_rate": 1.5898740251949609e-06, "loss": 0.0003, "step": 19639 }, { "epoch": 2.35, "learning_rate": 1.5898020395920816e-06, "loss": 0.5204, "step": 19640 }, { "epoch": 2.35, "learning_rate": 1.5897300539892022e-06, "loss": 0.0001, "step": 19641 }, { "epoch": 2.35, "learning_rate": 1.589658068386323e-06, "loss": 0.0002, "step": 19642 }, { "epoch": 2.35, "learning_rate": 1.5895860827834433e-06, "loss": 0.0001, "step": 19643 }, { "epoch": 2.35, "learning_rate": 1.5895140971805641e-06, "loss": 0.2465, "step": 19644 }, { "epoch": 2.35, "learning_rate": 1.5894421115776845e-06, "loss": 0.0001, "step": 19645 }, { "epoch": 2.35, "learning_rate": 1.5893701259748052e-06, "loss": 0.0003, "step": 19646 }, { "epoch": 2.35, "learning_rate": 1.5892981403719256e-06, "loss": 0.0002, "step": 19647 }, { "epoch": 2.35, "learning_rate": 1.5892261547690462e-06, "loss": 0.0002, "step": 19648 }, { "epoch": 2.35, "learning_rate": 1.589154169166167e-06, "loss": 0.0001, "step": 19649 }, { "epoch": 2.35, "learning_rate": 1.5890821835632873e-06, "loss": 0.0002, "step": 19650 }, { "epoch": 2.35, "learning_rate": 1.589010197960408e-06, "loss": 0.0002, "step": 19651 }, { "epoch": 2.35, "learning_rate": 1.5889382123575284e-06, "loss": 0.0003, "step": 19652 }, { "epoch": 2.36, "learning_rate": 1.5888662267546492e-06, "loss": 0.0002, "step": 19653 }, { "epoch": 2.36, "learning_rate": 1.5887942411517696e-06, "loss": 0.0009, "step": 19654 }, { "epoch": 2.36, "learning_rate": 1.5887222555488903e-06, "loss": 0.0002, "step": 19655 }, { "epoch": 2.36, "learning_rate": 1.588650269946011e-06, "loss": 0.0003, "step": 19656 }, { "epoch": 2.36, "learning_rate": 1.5885782843431315e-06, "loss": 0.0001, "step": 19657 }, { "epoch": 2.36, "learning_rate": 1.588506298740252e-06, "loss": 0.0002, "step": 19658 }, { "epoch": 2.36, "learning_rate": 1.5884343131373724e-06, "loss": 0.0001, "step": 19659 }, { "epoch": 2.36, "learning_rate": 1.5883623275344932e-06, "loss": 0.0006, "step": 19660 }, { "epoch": 2.36, "learning_rate": 1.5882903419316135e-06, "loss": 0.0002, "step": 19661 }, { "epoch": 2.36, "learning_rate": 1.5882183563287343e-06, "loss": 0.0002, "step": 19662 }, { "epoch": 2.36, "learning_rate": 1.5881463707258549e-06, "loss": 0.0001, "step": 19663 }, { "epoch": 2.36, "learning_rate": 1.5880743851229754e-06, "loss": 0.0002, "step": 19664 }, { "epoch": 2.36, "learning_rate": 1.588002399520096e-06, "loss": 0.3532, "step": 19665 }, { "epoch": 2.36, "learning_rate": 1.5879304139172168e-06, "loss": 0.0003, "step": 19666 }, { "epoch": 2.36, "learning_rate": 1.5878584283143372e-06, "loss": 0.6838, "step": 19667 }, { "epoch": 2.36, "learning_rate": 1.587786442711458e-06, "loss": 0.0002, "step": 19668 }, { "epoch": 2.36, "learning_rate": 1.5877144571085783e-06, "loss": 0.0001, "step": 19669 }, { "epoch": 2.36, "learning_rate": 1.5876424715056989e-06, "loss": 0.0003, "step": 19670 }, { "epoch": 2.36, "learning_rate": 1.5875704859028194e-06, "loss": 0.0002, "step": 19671 }, { "epoch": 2.36, "learning_rate": 1.58749850029994e-06, "loss": 0.0001, "step": 19672 }, { "epoch": 2.36, "learning_rate": 1.5874265146970608e-06, "loss": 0.0001, "step": 19673 }, { "epoch": 2.36, "learning_rate": 1.5873545290941811e-06, "loss": 0.0001, "step": 19674 }, { "epoch": 2.36, "learning_rate": 1.587282543491302e-06, "loss": 0.0001, "step": 19675 }, { "epoch": 2.36, "learning_rate": 1.5872105578884223e-06, "loss": 0.0002, "step": 19676 }, { "epoch": 2.36, "learning_rate": 1.587138572285543e-06, "loss": 0.0002, "step": 19677 }, { "epoch": 2.36, "learning_rate": 1.5870665866826634e-06, "loss": 0.0001, "step": 19678 }, { "epoch": 2.36, "learning_rate": 1.5869946010797842e-06, "loss": 0.0002, "step": 19679 }, { "epoch": 2.36, "learning_rate": 1.5869226154769047e-06, "loss": 0.0002, "step": 19680 }, { "epoch": 2.36, "learning_rate": 1.586850629874025e-06, "loss": 0.0002, "step": 19681 }, { "epoch": 2.36, "learning_rate": 1.5867786442711459e-06, "loss": 0.0001, "step": 19682 }, { "epoch": 2.36, "learning_rate": 1.5867066586682662e-06, "loss": 0.0002, "step": 19683 }, { "epoch": 2.36, "learning_rate": 1.586634673065387e-06, "loss": 0.0002, "step": 19684 }, { "epoch": 2.36, "learning_rate": 1.5865626874625074e-06, "loss": 0.0002, "step": 19685 }, { "epoch": 2.36, "learning_rate": 1.5864907018596281e-06, "loss": 0.0002, "step": 19686 }, { "epoch": 2.36, "learning_rate": 1.5864187162567487e-06, "loss": 0.0001, "step": 19687 }, { "epoch": 2.36, "learning_rate": 1.5863467306538693e-06, "loss": 0.0002, "step": 19688 }, { "epoch": 2.36, "learning_rate": 1.5862747450509898e-06, "loss": 0.0001, "step": 19689 }, { "epoch": 2.36, "learning_rate": 1.5862027594481106e-06, "loss": 0.0001, "step": 19690 }, { "epoch": 2.36, "learning_rate": 1.586130773845231e-06, "loss": 0.0002, "step": 19691 }, { "epoch": 2.36, "learning_rate": 1.5860587882423517e-06, "loss": 0.0001, "step": 19692 }, { "epoch": 2.36, "learning_rate": 1.585986802639472e-06, "loss": 0.0028, "step": 19693 }, { "epoch": 2.36, "learning_rate": 1.5859148170365927e-06, "loss": 0.0003, "step": 19694 }, { "epoch": 2.36, "learning_rate": 1.5858428314337132e-06, "loss": 0.0238, "step": 19695 }, { "epoch": 2.36, "learning_rate": 1.5857708458308338e-06, "loss": 0.0001, "step": 19696 }, { "epoch": 2.36, "learning_rate": 1.5856988602279546e-06, "loss": 0.0001, "step": 19697 }, { "epoch": 2.36, "learning_rate": 1.585626874625075e-06, "loss": 0.0001, "step": 19698 }, { "epoch": 2.36, "learning_rate": 1.5855548890221957e-06, "loss": 0.0001, "step": 19699 }, { "epoch": 2.36, "learning_rate": 1.585482903419316e-06, "loss": 0.0001, "step": 19700 }, { "epoch": 2.36, "learning_rate": 1.5854109178164368e-06, "loss": 0.0001, "step": 19701 }, { "epoch": 2.36, "learning_rate": 1.5853389322135572e-06, "loss": 0.0002, "step": 19702 }, { "epoch": 2.36, "learning_rate": 1.585266946610678e-06, "loss": 0.0002, "step": 19703 }, { "epoch": 2.36, "learning_rate": 1.5851949610077985e-06, "loss": 1.0077, "step": 19704 }, { "epoch": 2.36, "learning_rate": 1.585122975404919e-06, "loss": 0.0001, "step": 19705 }, { "epoch": 2.36, "learning_rate": 1.5850509898020397e-06, "loss": 0.0002, "step": 19706 }, { "epoch": 2.36, "learning_rate": 1.58497900419916e-06, "loss": 0.0001, "step": 19707 }, { "epoch": 2.36, "learning_rate": 1.5849070185962808e-06, "loss": 0.0002, "step": 19708 }, { "epoch": 2.36, "learning_rate": 1.5848350329934012e-06, "loss": 0.0002, "step": 19709 }, { "epoch": 2.36, "learning_rate": 1.584763047390522e-06, "loss": 0.0002, "step": 19710 }, { "epoch": 2.36, "learning_rate": 1.5846910617876425e-06, "loss": 0.0002, "step": 19711 }, { "epoch": 2.36, "learning_rate": 1.5846190761847633e-06, "loss": 0.0003, "step": 19712 }, { "epoch": 2.36, "learning_rate": 1.5845470905818836e-06, "loss": 0.8038, "step": 19713 }, { "epoch": 2.36, "learning_rate": 1.5844751049790044e-06, "loss": 0.0002, "step": 19714 }, { "epoch": 2.36, "learning_rate": 1.5844031193761248e-06, "loss": 0.0001, "step": 19715 }, { "epoch": 2.36, "learning_rate": 1.5843311337732451e-06, "loss": 0.0002, "step": 19716 }, { "epoch": 2.36, "learning_rate": 1.584259148170366e-06, "loss": 0.0002, "step": 19717 }, { "epoch": 2.36, "learning_rate": 1.5841871625674865e-06, "loss": 0.0002, "step": 19718 }, { "epoch": 2.36, "learning_rate": 1.5841151769646073e-06, "loss": 0.0003, "step": 19719 }, { "epoch": 2.36, "learning_rate": 1.5840431913617276e-06, "loss": 0.0001, "step": 19720 }, { "epoch": 2.36, "learning_rate": 1.5839712057588484e-06, "loss": 0.0005, "step": 19721 }, { "epoch": 2.36, "learning_rate": 1.5838992201559687e-06, "loss": 0.0003, "step": 19722 }, { "epoch": 2.36, "learning_rate": 1.5838272345530895e-06, "loss": 0.0002, "step": 19723 }, { "epoch": 2.36, "learning_rate": 1.5837552489502099e-06, "loss": 0.0003, "step": 19724 }, { "epoch": 2.36, "learning_rate": 1.5836832633473307e-06, "loss": 0.0002, "step": 19725 }, { "epoch": 2.36, "learning_rate": 1.5836112777444512e-06, "loss": 0.0001, "step": 19726 }, { "epoch": 2.36, "learning_rate": 1.5835392921415716e-06, "loss": 0.0001, "step": 19727 }, { "epoch": 2.36, "learning_rate": 1.5834673065386924e-06, "loss": 0.0004, "step": 19728 }, { "epoch": 2.36, "learning_rate": 1.5833953209358127e-06, "loss": 0.0002, "step": 19729 }, { "epoch": 2.36, "learning_rate": 1.5833233353329335e-06, "loss": 0.0004, "step": 19730 }, { "epoch": 2.36, "learning_rate": 1.5832513497300538e-06, "loss": 0.0549, "step": 19731 }, { "epoch": 2.36, "learning_rate": 1.5831793641271746e-06, "loss": 0.0001, "step": 19732 }, { "epoch": 2.36, "learning_rate": 1.5831073785242952e-06, "loss": 0.0005, "step": 19733 }, { "epoch": 2.36, "learning_rate": 1.5830353929214158e-06, "loss": 0.0001, "step": 19734 }, { "epoch": 2.36, "learning_rate": 1.5829634073185363e-06, "loss": 0.0004, "step": 19735 }, { "epoch": 2.37, "learning_rate": 1.582891421715657e-06, "loss": 0.8403, "step": 19736 }, { "epoch": 2.37, "learning_rate": 1.5828194361127775e-06, "loss": 0.0002, "step": 19737 }, { "epoch": 2.37, "learning_rate": 1.5827474505098982e-06, "loss": 0.0001, "step": 19738 }, { "epoch": 2.37, "learning_rate": 1.5826754649070186e-06, "loss": 0.0001, "step": 19739 }, { "epoch": 2.37, "learning_rate": 1.5826034793041392e-06, "loss": 0.1663, "step": 19740 }, { "epoch": 2.37, "learning_rate": 1.5825314937012597e-06, "loss": 0.0001, "step": 19741 }, { "epoch": 2.37, "learning_rate": 1.5824595080983803e-06, "loss": 0.0004, "step": 19742 }, { "epoch": 2.37, "learning_rate": 1.582387522495501e-06, "loss": 0.0004, "step": 19743 }, { "epoch": 2.37, "learning_rate": 1.5823155368926214e-06, "loss": 0.0002, "step": 19744 }, { "epoch": 2.37, "learning_rate": 1.5822435512897422e-06, "loss": 0.0001, "step": 19745 }, { "epoch": 2.37, "learning_rate": 1.5821715656868626e-06, "loss": 0.0001, "step": 19746 }, { "epoch": 2.37, "learning_rate": 1.5820995800839833e-06, "loss": 0.0003, "step": 19747 }, { "epoch": 2.37, "learning_rate": 1.5820275944811037e-06, "loss": 0.0002, "step": 19748 }, { "epoch": 2.37, "learning_rate": 1.5819556088782245e-06, "loss": 0.0001, "step": 19749 }, { "epoch": 2.37, "learning_rate": 1.581883623275345e-06, "loss": 0.0003, "step": 19750 }, { "epoch": 2.37, "learning_rate": 1.5818116376724654e-06, "loss": 0.0001, "step": 19751 }, { "epoch": 2.37, "learning_rate": 1.5817396520695862e-06, "loss": 0.0005, "step": 19752 }, { "epoch": 2.37, "learning_rate": 1.5816676664667065e-06, "loss": 0.0003, "step": 19753 }, { "epoch": 2.37, "learning_rate": 1.5815956808638273e-06, "loss": 0.0002, "step": 19754 }, { "epoch": 2.37, "learning_rate": 1.5815236952609477e-06, "loss": 0.0004, "step": 19755 }, { "epoch": 2.37, "learning_rate": 1.5814517096580684e-06, "loss": 0.0001, "step": 19756 }, { "epoch": 2.37, "learning_rate": 1.581379724055189e-06, "loss": 0.0006, "step": 19757 }, { "epoch": 2.37, "learning_rate": 1.5813077384523098e-06, "loss": 0.0007, "step": 19758 }, { "epoch": 2.37, "learning_rate": 1.5812357528494301e-06, "loss": 0.0001, "step": 19759 }, { "epoch": 2.37, "learning_rate": 1.581163767246551e-06, "loss": 0.0001, "step": 19760 }, { "epoch": 2.37, "learning_rate": 1.5810917816436713e-06, "loss": 0.0003, "step": 19761 }, { "epoch": 2.37, "learning_rate": 1.5810197960407916e-06, "loss": 0.0006, "step": 19762 }, { "epoch": 2.37, "learning_rate": 1.5809478104379124e-06, "loss": 0.0001, "step": 19763 }, { "epoch": 2.37, "learning_rate": 1.580875824835033e-06, "loss": 0.0006, "step": 19764 }, { "epoch": 2.37, "learning_rate": 1.5808038392321538e-06, "loss": 0.0005, "step": 19765 }, { "epoch": 2.37, "learning_rate": 1.5807318536292741e-06, "loss": 0.0001, "step": 19766 }, { "epoch": 2.37, "learning_rate": 1.5806598680263949e-06, "loss": 0.0001, "step": 19767 }, { "epoch": 2.37, "learning_rate": 1.5805878824235152e-06, "loss": 0.0008, "step": 19768 }, { "epoch": 2.37, "learning_rate": 1.580515896820636e-06, "loss": 0.0005, "step": 19769 }, { "epoch": 2.37, "learning_rate": 1.5804439112177564e-06, "loss": 0.0003, "step": 19770 }, { "epoch": 2.37, "learning_rate": 1.5803719256148772e-06, "loss": 0.0002, "step": 19771 }, { "epoch": 2.37, "learning_rate": 1.5802999400119977e-06, "loss": 0.0003, "step": 19772 }, { "epoch": 2.37, "learning_rate": 1.5802279544091183e-06, "loss": 0.0002, "step": 19773 }, { "epoch": 2.37, "learning_rate": 1.5801559688062389e-06, "loss": 0.0004, "step": 19774 }, { "epoch": 2.37, "learning_rate": 1.5800839832033592e-06, "loss": 0.0002, "step": 19775 }, { "epoch": 2.37, "learning_rate": 1.58001199760048e-06, "loss": 0.0002, "step": 19776 }, { "epoch": 2.37, "learning_rate": 1.5799400119976003e-06, "loss": 0.0004, "step": 19777 }, { "epoch": 2.37, "learning_rate": 1.5798680263947211e-06, "loss": 0.0002, "step": 19778 }, { "epoch": 2.37, "learning_rate": 1.5797960407918417e-06, "loss": 0.0001, "step": 19779 }, { "epoch": 2.37, "learning_rate": 1.5797240551889623e-06, "loss": 0.0002, "step": 19780 }, { "epoch": 2.37, "learning_rate": 1.5796520695860828e-06, "loss": 0.0001, "step": 19781 }, { "epoch": 2.37, "learning_rate": 1.5795800839832036e-06, "loss": 0.0001, "step": 19782 }, { "epoch": 2.37, "learning_rate": 1.579508098380324e-06, "loss": 1.0139, "step": 19783 }, { "epoch": 2.37, "learning_rate": 1.5794361127774447e-06, "loss": 0.0005, "step": 19784 }, { "epoch": 2.37, "learning_rate": 1.579364127174565e-06, "loss": 0.0001, "step": 19785 }, { "epoch": 2.37, "learning_rate": 1.5792921415716857e-06, "loss": 0.0002, "step": 19786 }, { "epoch": 2.37, "learning_rate": 1.5792201559688062e-06, "loss": 0.0002, "step": 19787 }, { "epoch": 2.37, "learning_rate": 1.5791481703659268e-06, "loss": 0.0001, "step": 19788 }, { "epoch": 2.37, "learning_rate": 1.5790761847630476e-06, "loss": 0.0144, "step": 19789 }, { "epoch": 2.37, "learning_rate": 1.579004199160168e-06, "loss": 0.0003, "step": 19790 }, { "epoch": 2.37, "learning_rate": 1.5789322135572887e-06, "loss": 0.0002, "step": 19791 }, { "epoch": 2.37, "learning_rate": 1.578860227954409e-06, "loss": 0.0004, "step": 19792 }, { "epoch": 2.37, "learning_rate": 1.5787882423515298e-06, "loss": 0.0003, "step": 19793 }, { "epoch": 2.37, "learning_rate": 1.5787162567486502e-06, "loss": 0.0002, "step": 19794 }, { "epoch": 2.37, "learning_rate": 1.578644271145771e-06, "loss": 0.0003, "step": 19795 }, { "epoch": 2.37, "learning_rate": 1.5785722855428915e-06, "loss": 0.0001, "step": 19796 }, { "epoch": 2.37, "learning_rate": 1.5785002999400119e-06, "loss": 0.0002, "step": 19797 }, { "epoch": 2.37, "learning_rate": 1.5784283143371327e-06, "loss": 0.0002, "step": 19798 }, { "epoch": 2.37, "learning_rate": 1.578356328734253e-06, "loss": 0.0005, "step": 19799 }, { "epoch": 2.37, "learning_rate": 1.5782843431313738e-06, "loss": 0.0002, "step": 19800 }, { "epoch": 2.37, "learning_rate": 1.5782123575284942e-06, "loss": 0.0001, "step": 19801 }, { "epoch": 2.37, "learning_rate": 1.578140371925615e-06, "loss": 0.0001, "step": 19802 }, { "epoch": 2.37, "learning_rate": 1.5780683863227355e-06, "loss": 0.0001, "step": 19803 }, { "epoch": 2.37, "learning_rate": 1.5779964007198563e-06, "loss": 0.0007, "step": 19804 }, { "epoch": 2.37, "learning_rate": 1.5779244151169766e-06, "loss": 0.0004, "step": 19805 }, { "epoch": 2.37, "learning_rate": 1.5778524295140974e-06, "loss": 0.0003, "step": 19806 }, { "epoch": 2.37, "learning_rate": 1.5777804439112178e-06, "loss": 0.0005, "step": 19807 }, { "epoch": 2.37, "learning_rate": 1.5777084583083381e-06, "loss": 0.0001, "step": 19808 }, { "epoch": 2.37, "learning_rate": 1.577636472705459e-06, "loss": 0.0003, "step": 19809 }, { "epoch": 2.37, "learning_rate": 1.5775644871025795e-06, "loss": 0.0002, "step": 19810 }, { "epoch": 2.37, "learning_rate": 1.5774925014997003e-06, "loss": 0.0003, "step": 19811 }, { "epoch": 2.37, "learning_rate": 1.5774205158968206e-06, "loss": 0.0003, "step": 19812 }, { "epoch": 2.37, "learning_rate": 1.5773485302939414e-06, "loss": 0.0002, "step": 19813 }, { "epoch": 2.37, "learning_rate": 1.5772765446910617e-06, "loss": 0.0002, "step": 19814 }, { "epoch": 2.37, "learning_rate": 1.5772045590881825e-06, "loss": 0.0002, "step": 19815 }, { "epoch": 2.37, "learning_rate": 1.5771325734853029e-06, "loss": 0.0002, "step": 19816 }, { "epoch": 2.37, "learning_rate": 1.5770605878824237e-06, "loss": 0.0001, "step": 19817 }, { "epoch": 2.37, "learning_rate": 1.576988602279544e-06, "loss": 0.0001, "step": 19818 }, { "epoch": 2.37, "learning_rate": 1.5769166166766648e-06, "loss": 0.0002, "step": 19819 }, { "epoch": 2.38, "learning_rate": 1.5768446310737854e-06, "loss": 0.0003, "step": 19820 }, { "epoch": 2.38, "learning_rate": 1.5767726454709057e-06, "loss": 0.0002, "step": 19821 }, { "epoch": 2.38, "learning_rate": 1.5767006598680265e-06, "loss": 0.0002, "step": 19822 }, { "epoch": 2.38, "learning_rate": 1.5766286742651468e-06, "loss": 0.0003, "step": 19823 }, { "epoch": 2.38, "learning_rate": 1.5765566886622676e-06, "loss": 0.0011, "step": 19824 }, { "epoch": 2.38, "learning_rate": 1.576484703059388e-06, "loss": 0.0002, "step": 19825 }, { "epoch": 2.38, "learning_rate": 1.5764127174565088e-06, "loss": 0.0001, "step": 19826 }, { "epoch": 2.38, "learning_rate": 1.5763407318536293e-06, "loss": 0.0001, "step": 19827 }, { "epoch": 2.38, "learning_rate": 1.57626874625075e-06, "loss": 0.0002, "step": 19828 }, { "epoch": 2.38, "learning_rate": 1.5761967606478705e-06, "loss": 0.0004, "step": 19829 }, { "epoch": 2.38, "learning_rate": 1.5761247750449912e-06, "loss": 0.0002, "step": 19830 }, { "epoch": 2.38, "learning_rate": 1.5760527894421116e-06, "loss": 0.0002, "step": 19831 }, { "epoch": 2.38, "learning_rate": 1.575980803839232e-06, "loss": 0.0003, "step": 19832 }, { "epoch": 2.38, "learning_rate": 1.5759088182363527e-06, "loss": 0.0002, "step": 19833 }, { "epoch": 2.38, "learning_rate": 1.5758368326334733e-06, "loss": 0.0003, "step": 19834 }, { "epoch": 2.38, "learning_rate": 1.575764847030594e-06, "loss": 0.0002, "step": 19835 }, { "epoch": 2.38, "learning_rate": 1.5756928614277144e-06, "loss": 0.0003, "step": 19836 }, { "epoch": 2.38, "learning_rate": 1.5756208758248352e-06, "loss": 0.6302, "step": 19837 }, { "epoch": 2.38, "learning_rate": 1.5755488902219556e-06, "loss": 0.0001, "step": 19838 }, { "epoch": 2.38, "learning_rate": 1.5754769046190763e-06, "loss": 0.0004, "step": 19839 }, { "epoch": 2.38, "learning_rate": 1.5754049190161967e-06, "loss": 0.0002, "step": 19840 }, { "epoch": 2.38, "learning_rate": 1.5753329334133175e-06, "loss": 0.0001, "step": 19841 }, { "epoch": 2.38, "learning_rate": 1.575260947810438e-06, "loss": 0.0001, "step": 19842 }, { "epoch": 2.38, "learning_rate": 1.5751889622075584e-06, "loss": 0.0001, "step": 19843 }, { "epoch": 2.38, "learning_rate": 1.5751169766046792e-06, "loss": 0.0001, "step": 19844 }, { "epoch": 2.38, "learning_rate": 1.5750449910017995e-06, "loss": 0.0001, "step": 19845 }, { "epoch": 2.38, "learning_rate": 1.5749730053989203e-06, "loss": 0.0002, "step": 19846 }, { "epoch": 2.38, "learning_rate": 1.5749010197960407e-06, "loss": 0.0002, "step": 19847 }, { "epoch": 2.38, "learning_rate": 1.5748290341931614e-06, "loss": 0.7326, "step": 19848 }, { "epoch": 2.38, "learning_rate": 1.574757048590282e-06, "loss": 0.0002, "step": 19849 }, { "epoch": 2.38, "learning_rate": 1.5746850629874026e-06, "loss": 0.0001, "step": 19850 }, { "epoch": 2.38, "learning_rate": 1.5746130773845231e-06, "loss": 0.0008, "step": 19851 }, { "epoch": 2.38, "learning_rate": 1.574541091781644e-06, "loss": 0.0002, "step": 19852 }, { "epoch": 2.38, "learning_rate": 1.5744691061787643e-06, "loss": 0.0002, "step": 19853 }, { "epoch": 2.38, "learning_rate": 1.5743971205758846e-06, "loss": 0.0001, "step": 19854 }, { "epoch": 2.38, "learning_rate": 1.5743251349730054e-06, "loss": 0.0002, "step": 19855 }, { "epoch": 2.38, "learning_rate": 1.574253149370126e-06, "loss": 0.0002, "step": 19856 }, { "epoch": 2.38, "learning_rate": 1.5741811637672465e-06, "loss": 0.0001, "step": 19857 }, { "epoch": 2.38, "learning_rate": 1.574109178164367e-06, "loss": 0.0001, "step": 19858 }, { "epoch": 2.38, "learning_rate": 1.5740371925614879e-06, "loss": 0.0002, "step": 19859 }, { "epoch": 2.38, "learning_rate": 1.5739652069586082e-06, "loss": 0.0001, "step": 19860 }, { "epoch": 2.38, "learning_rate": 1.573893221355729e-06, "loss": 0.0002, "step": 19861 }, { "epoch": 2.38, "learning_rate": 1.5738212357528494e-06, "loss": 0.0002, "step": 19862 }, { "epoch": 2.38, "learning_rate": 1.5737492501499701e-06, "loss": 0.0002, "step": 19863 }, { "epoch": 2.38, "learning_rate": 1.5736772645470905e-06, "loss": 0.0002, "step": 19864 }, { "epoch": 2.38, "learning_rate": 1.5736052789442113e-06, "loss": 0.0003, "step": 19865 }, { "epoch": 2.38, "learning_rate": 1.5735332933413318e-06, "loss": 0.0001, "step": 19866 }, { "epoch": 2.38, "learning_rate": 1.5734613077384522e-06, "loss": 0.0001, "step": 19867 }, { "epoch": 2.38, "learning_rate": 1.573389322135573e-06, "loss": 0.0002, "step": 19868 }, { "epoch": 2.38, "learning_rate": 1.5733173365326933e-06, "loss": 0.0001, "step": 19869 }, { "epoch": 2.38, "learning_rate": 1.5732453509298141e-06, "loss": 0.0003, "step": 19870 }, { "epoch": 2.38, "learning_rate": 1.5731733653269345e-06, "loss": 0.0001, "step": 19871 }, { "epoch": 2.38, "learning_rate": 1.5731013797240552e-06, "loss": 0.0001, "step": 19872 }, { "epoch": 2.38, "learning_rate": 1.5730293941211758e-06, "loss": 0.0004, "step": 19873 }, { "epoch": 2.38, "learning_rate": 1.5729574085182966e-06, "loss": 0.0002, "step": 19874 }, { "epoch": 2.38, "learning_rate": 1.572885422915417e-06, "loss": 0.0001, "step": 19875 }, { "epoch": 2.38, "learning_rate": 1.5728134373125377e-06, "loss": 0.0002, "step": 19876 }, { "epoch": 2.38, "learning_rate": 1.572741451709658e-06, "loss": 0.0006, "step": 19877 }, { "epoch": 2.38, "learning_rate": 1.5726694661067784e-06, "loss": 0.0002, "step": 19878 }, { "epoch": 2.38, "learning_rate": 1.5725974805038992e-06, "loss": 0.0002, "step": 19879 }, { "epoch": 2.38, "learning_rate": 1.5725254949010198e-06, "loss": 0.0002, "step": 19880 }, { "epoch": 2.38, "learning_rate": 1.5724535092981406e-06, "loss": 0.0006, "step": 19881 }, { "epoch": 2.38, "learning_rate": 1.572381523695261e-06, "loss": 0.0001, "step": 19882 }, { "epoch": 2.38, "learning_rate": 1.5723095380923817e-06, "loss": 0.0003, "step": 19883 }, { "epoch": 2.38, "learning_rate": 1.572237552489502e-06, "loss": 0.0002, "step": 19884 }, { "epoch": 2.38, "learning_rate": 1.5721655668866228e-06, "loss": 0.0001, "step": 19885 }, { "epoch": 2.38, "learning_rate": 1.5720935812837432e-06, "loss": 0.0001, "step": 19886 }, { "epoch": 2.38, "learning_rate": 1.572021595680864e-06, "loss": 0.0002, "step": 19887 }, { "epoch": 2.38, "learning_rate": 1.5719496100779845e-06, "loss": 0.0001, "step": 19888 }, { "epoch": 2.38, "learning_rate": 1.5718776244751049e-06, "loss": 0.0001, "step": 19889 }, { "epoch": 2.38, "learning_rate": 1.5718056388722257e-06, "loss": 0.0001, "step": 19890 }, { "epoch": 2.38, "learning_rate": 1.571733653269346e-06, "loss": 0.0002, "step": 19891 }, { "epoch": 2.38, "learning_rate": 1.5716616676664668e-06, "loss": 0.0002, "step": 19892 }, { "epoch": 2.38, "learning_rate": 1.5715896820635872e-06, "loss": 0.5942, "step": 19893 }, { "epoch": 2.38, "learning_rate": 1.571517696460708e-06, "loss": 0.6477, "step": 19894 }, { "epoch": 2.38, "learning_rate": 1.5714457108578285e-06, "loss": 0.0003, "step": 19895 }, { "epoch": 2.38, "learning_rate": 1.571373725254949e-06, "loss": 0.0003, "step": 19896 }, { "epoch": 2.38, "learning_rate": 1.5713017396520696e-06, "loss": 0.0001, "step": 19897 }, { "epoch": 2.38, "learning_rate": 1.5712297540491904e-06, "loss": 0.0001, "step": 19898 }, { "epoch": 2.38, "learning_rate": 1.5711577684463108e-06, "loss": 0.0002, "step": 19899 }, { "epoch": 2.38, "learning_rate": 1.5710857828434315e-06, "loss": 0.0002, "step": 19900 }, { "epoch": 2.38, "learning_rate": 1.571013797240552e-06, "loss": 0.0001, "step": 19901 }, { "epoch": 2.38, "learning_rate": 1.5709418116376725e-06, "loss": 0.0002, "step": 19902 }, { "epoch": 2.39, "learning_rate": 1.570869826034793e-06, "loss": 0.0001, "step": 19903 }, { "epoch": 2.39, "learning_rate": 1.5707978404319136e-06, "loss": 0.0001, "step": 19904 }, { "epoch": 2.39, "learning_rate": 1.5707258548290344e-06, "loss": 0.0001, "step": 19905 }, { "epoch": 2.39, "learning_rate": 1.5706538692261547e-06, "loss": 0.0002, "step": 19906 }, { "epoch": 2.39, "learning_rate": 1.5705818836232755e-06, "loss": 0.0002, "step": 19907 }, { "epoch": 2.39, "learning_rate": 1.5705098980203959e-06, "loss": 0.0002, "step": 19908 }, { "epoch": 2.39, "learning_rate": 1.5704379124175166e-06, "loss": 0.0002, "step": 19909 }, { "epoch": 2.39, "learning_rate": 1.570365926814637e-06, "loss": 0.0002, "step": 19910 }, { "epoch": 2.39, "learning_rate": 1.5702939412117578e-06, "loss": 0.0003, "step": 19911 }, { "epoch": 2.39, "learning_rate": 1.5702219556088783e-06, "loss": 0.0003, "step": 19912 }, { "epoch": 2.39, "learning_rate": 1.5701499700059987e-06, "loss": 0.0001, "step": 19913 }, { "epoch": 2.39, "learning_rate": 1.5700779844031195e-06, "loss": 0.0001, "step": 19914 }, { "epoch": 2.39, "learning_rate": 1.5700059988002398e-06, "loss": 0.0001, "step": 19915 }, { "epoch": 2.39, "learning_rate": 1.5699340131973606e-06, "loss": 0.0001, "step": 19916 }, { "epoch": 2.39, "learning_rate": 1.569862027594481e-06, "loss": 0.0002, "step": 19917 }, { "epoch": 2.39, "learning_rate": 1.5697900419916017e-06, "loss": 0.0002, "step": 19918 }, { "epoch": 2.39, "learning_rate": 1.5697180563887223e-06, "loss": 0.0002, "step": 19919 }, { "epoch": 2.39, "learning_rate": 1.569646070785843e-06, "loss": 0.0002, "step": 19920 }, { "epoch": 2.39, "learning_rate": 1.5695740851829634e-06, "loss": 0.0002, "step": 19921 }, { "epoch": 2.39, "learning_rate": 1.5695020995800842e-06, "loss": 0.0003, "step": 19922 }, { "epoch": 2.39, "learning_rate": 1.5694301139772046e-06, "loss": 0.001, "step": 19923 }, { "epoch": 2.39, "learning_rate": 1.569358128374325e-06, "loss": 0.0004, "step": 19924 }, { "epoch": 2.39, "learning_rate": 1.5692861427714457e-06, "loss": 0.0001, "step": 19925 }, { "epoch": 2.39, "learning_rate": 1.5692141571685663e-06, "loss": 0.0001, "step": 19926 }, { "epoch": 2.39, "learning_rate": 1.569142171565687e-06, "loss": 0.0001, "step": 19927 }, { "epoch": 2.39, "learning_rate": 1.5690701859628074e-06, "loss": 0.0002, "step": 19928 }, { "epoch": 2.39, "learning_rate": 1.5689982003599282e-06, "loss": 0.0003, "step": 19929 }, { "epoch": 2.39, "learning_rate": 1.5689262147570485e-06, "loss": 0.0003, "step": 19930 }, { "epoch": 2.39, "learning_rate": 1.5688542291541693e-06, "loss": 0.0002, "step": 19931 }, { "epoch": 2.39, "learning_rate": 1.5687822435512897e-06, "loss": 0.0002, "step": 19932 }, { "epoch": 2.39, "learning_rate": 1.5687102579484105e-06, "loss": 0.0001, "step": 19933 }, { "epoch": 2.39, "learning_rate": 1.568638272345531e-06, "loss": 0.0003, "step": 19934 }, { "epoch": 2.39, "learning_rate": 1.5685662867426514e-06, "loss": 0.0003, "step": 19935 }, { "epoch": 2.39, "learning_rate": 1.5684943011397722e-06, "loss": 0.0001, "step": 19936 }, { "epoch": 2.39, "learning_rate": 1.5684223155368925e-06, "loss": 0.0002, "step": 19937 }, { "epoch": 2.39, "learning_rate": 1.5683503299340133e-06, "loss": 0.0002, "step": 19938 }, { "epoch": 2.39, "learning_rate": 1.5682783443311336e-06, "loss": 0.0001, "step": 19939 }, { "epoch": 2.39, "learning_rate": 1.5682063587282544e-06, "loss": 0.0002, "step": 19940 }, { "epoch": 2.39, "learning_rate": 1.568134373125375e-06, "loss": 0.0001, "step": 19941 }, { "epoch": 2.39, "learning_rate": 1.5680623875224956e-06, "loss": 0.0004, "step": 19942 }, { "epoch": 2.39, "learning_rate": 1.5679904019196161e-06, "loss": 0.0002, "step": 19943 }, { "epoch": 2.39, "learning_rate": 1.567918416316737e-06, "loss": 0.0002, "step": 19944 }, { "epoch": 2.39, "learning_rate": 1.5678464307138573e-06, "loss": 0.0001, "step": 19945 }, { "epoch": 2.39, "learning_rate": 1.567774445110978e-06, "loss": 0.0001, "step": 19946 }, { "epoch": 2.39, "learning_rate": 1.5677024595080984e-06, "loss": 0.0001, "step": 19947 }, { "epoch": 2.39, "learning_rate": 1.5676304739052187e-06, "loss": 0.0002, "step": 19948 }, { "epoch": 2.39, "learning_rate": 1.5675584883023395e-06, "loss": 0.0001, "step": 19949 }, { "epoch": 2.39, "learning_rate": 1.56748650269946e-06, "loss": 0.0001, "step": 19950 }, { "epoch": 2.39, "learning_rate": 1.5674145170965809e-06, "loss": 0.0002, "step": 19951 }, { "epoch": 2.39, "learning_rate": 1.5673425314937012e-06, "loss": 0.0001, "step": 19952 }, { "epoch": 2.39, "learning_rate": 1.567270545890822e-06, "loss": 0.0002, "step": 19953 }, { "epoch": 2.39, "learning_rate": 1.5671985602879424e-06, "loss": 0.0006, "step": 19954 }, { "epoch": 2.39, "learning_rate": 1.5671265746850631e-06, "loss": 0.0002, "step": 19955 }, { "epoch": 2.39, "learning_rate": 1.5670545890821835e-06, "loss": 0.0003, "step": 19956 }, { "epoch": 2.39, "learning_rate": 1.5669826034793043e-06, "loss": 0.0004, "step": 19957 }, { "epoch": 2.39, "learning_rate": 1.5669106178764248e-06, "loss": 0.0001, "step": 19958 }, { "epoch": 2.39, "learning_rate": 1.5668386322735452e-06, "loss": 0.0001, "step": 19959 }, { "epoch": 2.39, "learning_rate": 1.566766646670666e-06, "loss": 0.0001, "step": 19960 }, { "epoch": 2.39, "learning_rate": 1.5666946610677863e-06, "loss": 0.0001, "step": 19961 }, { "epoch": 2.39, "learning_rate": 1.5666226754649071e-06, "loss": 0.0002, "step": 19962 }, { "epoch": 2.39, "learning_rate": 1.5665506898620275e-06, "loss": 0.0002, "step": 19963 }, { "epoch": 2.39, "learning_rate": 1.5664787042591482e-06, "loss": 0.0002, "step": 19964 }, { "epoch": 2.39, "learning_rate": 1.5664067186562688e-06, "loss": 0.0001, "step": 19965 }, { "epoch": 2.39, "learning_rate": 1.5663347330533894e-06, "loss": 0.0001, "step": 19966 }, { "epoch": 2.39, "learning_rate": 1.56626274745051e-06, "loss": 0.0002, "step": 19967 }, { "epoch": 2.39, "learning_rate": 1.5661907618476307e-06, "loss": 0.0001, "step": 19968 }, { "epoch": 2.39, "learning_rate": 1.566118776244751e-06, "loss": 0.0002, "step": 19969 }, { "epoch": 2.39, "learning_rate": 1.5660467906418714e-06, "loss": 0.0128, "step": 19970 }, { "epoch": 2.39, "learning_rate": 1.5659748050389922e-06, "loss": 0.0003, "step": 19971 }, { "epoch": 2.39, "learning_rate": 1.5659028194361128e-06, "loss": 1.0398, "step": 19972 }, { "epoch": 2.39, "learning_rate": 1.5658308338332333e-06, "loss": 0.0001, "step": 19973 }, { "epoch": 2.39, "learning_rate": 1.565758848230354e-06, "loss": 0.0003, "step": 19974 }, { "epoch": 2.39, "learning_rate": 1.5656868626274747e-06, "loss": 0.0001, "step": 19975 }, { "epoch": 2.39, "learning_rate": 1.565614877024595e-06, "loss": 0.0002, "step": 19976 }, { "epoch": 2.39, "learning_rate": 1.5655428914217158e-06, "loss": 0.0001, "step": 19977 }, { "epoch": 2.39, "learning_rate": 1.5654709058188362e-06, "loss": 0.0001, "step": 19978 }, { "epoch": 2.39, "learning_rate": 1.565398920215957e-06, "loss": 0.0004, "step": 19979 }, { "epoch": 2.39, "learning_rate": 1.5653269346130773e-06, "loss": 0.0005, "step": 19980 }, { "epoch": 2.39, "learning_rate": 1.565254949010198e-06, "loss": 0.0002, "step": 19981 }, { "epoch": 2.39, "learning_rate": 1.5651829634073187e-06, "loss": 0.0002, "step": 19982 }, { "epoch": 2.39, "learning_rate": 1.565110977804439e-06, "loss": 0.0002, "step": 19983 }, { "epoch": 2.39, "learning_rate": 1.5650389922015598e-06, "loss": 0.183, "step": 19984 }, { "epoch": 2.39, "learning_rate": 1.5649670065986801e-06, "loss": 0.0003, "step": 19985 }, { "epoch": 2.39, "learning_rate": 1.564895020995801e-06, "loss": 0.0021, "step": 19986 }, { "epoch": 2.4, "learning_rate": 1.5648230353929213e-06, "loss": 0.9632, "step": 19987 }, { "epoch": 2.4, "learning_rate": 1.564751049790042e-06, "loss": 0.0027, "step": 19988 }, { "epoch": 2.4, "learning_rate": 1.5646790641871626e-06, "loss": 0.0002, "step": 19989 }, { "epoch": 2.4, "learning_rate": 1.5646070785842834e-06, "loss": 0.0003, "step": 19990 }, { "epoch": 2.4, "learning_rate": 1.5645350929814038e-06, "loss": 0.0003, "step": 19991 }, { "epoch": 2.4, "learning_rate": 1.5644631073785245e-06, "loss": 0.0002, "step": 19992 }, { "epoch": 2.4, "learning_rate": 1.5643911217756449e-06, "loss": 0.0001, "step": 19993 }, { "epoch": 2.4, "learning_rate": 1.5643191361727652e-06, "loss": 0.0001, "step": 19994 }, { "epoch": 2.4, "learning_rate": 1.564247150569886e-06, "loss": 0.0003, "step": 19995 }, { "epoch": 2.4, "learning_rate": 1.5641751649670066e-06, "loss": 0.0002, "step": 19996 }, { "epoch": 2.4, "learning_rate": 1.5641031793641274e-06, "loss": 0.0002, "step": 19997 }, { "epoch": 2.4, "learning_rate": 1.5640311937612477e-06, "loss": 0.0001, "step": 19998 }, { "epoch": 2.4, "learning_rate": 1.5639592081583685e-06, "loss": 0.0008, "step": 19999 }, { "epoch": 2.4, "learning_rate": 1.5638872225554889e-06, "loss": 0.0012, "step": 20000 }, { "epoch": 2.4, "learning_rate": 1.5638152369526096e-06, "loss": 0.0004, "step": 20001 }, { "epoch": 2.4, "learning_rate": 1.56374325134973e-06, "loss": 0.0001, "step": 20002 }, { "epoch": 2.4, "learning_rate": 1.5636712657468508e-06, "loss": 0.0001, "step": 20003 }, { "epoch": 2.4, "learning_rate": 1.5635992801439713e-06, "loss": 0.0002, "step": 20004 }, { "epoch": 2.4, "learning_rate": 1.5635272945410917e-06, "loss": 0.0003, "step": 20005 }, { "epoch": 2.4, "learning_rate": 1.5634553089382125e-06, "loss": 0.0001, "step": 20006 }, { "epoch": 2.4, "learning_rate": 1.5633833233353328e-06, "loss": 0.0015, "step": 20007 }, { "epoch": 2.4, "learning_rate": 1.5633113377324536e-06, "loss": 0.0001, "step": 20008 }, { "epoch": 2.4, "learning_rate": 1.563239352129574e-06, "loss": 0.0006, "step": 20009 }, { "epoch": 2.4, "learning_rate": 1.5631673665266947e-06, "loss": 0.0004, "step": 20010 }, { "epoch": 2.4, "learning_rate": 1.5630953809238153e-06, "loss": 0.0002, "step": 20011 }, { "epoch": 2.4, "learning_rate": 1.5630233953209359e-06, "loss": 0.0002, "step": 20012 }, { "epoch": 2.4, "learning_rate": 1.5629514097180564e-06, "loss": 0.0002, "step": 20013 }, { "epoch": 2.4, "learning_rate": 1.5628794241151772e-06, "loss": 0.0853, "step": 20014 }, { "epoch": 2.4, "learning_rate": 1.5628074385122976e-06, "loss": 0.0001, "step": 20015 }, { "epoch": 2.4, "learning_rate": 1.562735452909418e-06, "loss": 0.0004, "step": 20016 }, { "epoch": 2.4, "learning_rate": 1.5626634673065387e-06, "loss": 0.0002, "step": 20017 }, { "epoch": 2.4, "learning_rate": 1.5625914817036593e-06, "loss": 0.0003, "step": 20018 }, { "epoch": 2.4, "learning_rate": 1.5625194961007798e-06, "loss": 0.0001, "step": 20019 }, { "epoch": 2.4, "learning_rate": 1.5624475104979004e-06, "loss": 0.0002, "step": 20020 }, { "epoch": 2.4, "learning_rate": 1.5623755248950212e-06, "loss": 0.0006, "step": 20021 }, { "epoch": 2.4, "learning_rate": 1.5623035392921415e-06, "loss": 0.0362, "step": 20022 }, { "epoch": 2.4, "learning_rate": 1.5622315536892623e-06, "loss": 0.0, "step": 20023 }, { "epoch": 2.4, "learning_rate": 1.5621595680863827e-06, "loss": 0.0001, "step": 20024 }, { "epoch": 2.4, "learning_rate": 1.5620875824835035e-06, "loss": 0.0003, "step": 20025 }, { "epoch": 2.4, "learning_rate": 1.5620155968806238e-06, "loss": 0.0001, "step": 20026 }, { "epoch": 2.4, "learning_rate": 1.5619436112777446e-06, "loss": 0.0001, "step": 20027 }, { "epoch": 2.4, "learning_rate": 1.5618716256748652e-06, "loss": 0.3777, "step": 20028 }, { "epoch": 2.4, "learning_rate": 1.5617996400719855e-06, "loss": 0.0003, "step": 20029 }, { "epoch": 2.4, "learning_rate": 1.5617276544691063e-06, "loss": 0.0033, "step": 20030 }, { "epoch": 2.4, "learning_rate": 1.5616556688662266e-06, "loss": 0.0001, "step": 20031 }, { "epoch": 2.4, "learning_rate": 1.5615836832633474e-06, "loss": 0.0001, "step": 20032 }, { "epoch": 2.4, "learning_rate": 1.5615116976604678e-06, "loss": 0.0006, "step": 20033 }, { "epoch": 2.4, "learning_rate": 1.5614397120575886e-06, "loss": 0.0002, "step": 20034 }, { "epoch": 2.4, "learning_rate": 1.5613677264547091e-06, "loss": 0.0002, "step": 20035 }, { "epoch": 2.4, "learning_rate": 1.56129574085183e-06, "loss": 0.0002, "step": 20036 }, { "epoch": 2.4, "learning_rate": 1.5612237552489503e-06, "loss": 0.0002, "step": 20037 }, { "epoch": 2.4, "learning_rate": 1.561151769646071e-06, "loss": 0.0007, "step": 20038 }, { "epoch": 2.4, "learning_rate": 1.5610797840431914e-06, "loss": 0.0002, "step": 20039 }, { "epoch": 2.4, "learning_rate": 1.5610077984403117e-06, "loss": 0.0003, "step": 20040 }, { "epoch": 2.4, "learning_rate": 1.5609358128374325e-06, "loss": 0.0018, "step": 20041 }, { "epoch": 2.4, "learning_rate": 1.560863827234553e-06, "loss": 0.0007, "step": 20042 }, { "epoch": 2.4, "learning_rate": 1.5607918416316739e-06, "loss": 0.0003, "step": 20043 }, { "epoch": 2.4, "learning_rate": 1.5607198560287942e-06, "loss": 0.0001, "step": 20044 }, { "epoch": 2.4, "learning_rate": 1.560647870425915e-06, "loss": 0.0001, "step": 20045 }, { "epoch": 2.4, "learning_rate": 1.5605758848230354e-06, "loss": 0.0003, "step": 20046 }, { "epoch": 2.4, "learning_rate": 1.5605038992201561e-06, "loss": 0.0004, "step": 20047 }, { "epoch": 2.4, "learning_rate": 1.5604319136172765e-06, "loss": 0.0003, "step": 20048 }, { "epoch": 2.4, "learning_rate": 1.5603599280143973e-06, "loss": 0.0001, "step": 20049 }, { "epoch": 2.4, "learning_rate": 1.5602879424115178e-06, "loss": 0.0002, "step": 20050 }, { "epoch": 2.4, "learning_rate": 1.5602159568086382e-06, "loss": 0.0123, "step": 20051 }, { "epoch": 2.4, "learning_rate": 1.560143971205759e-06, "loss": 0.0001, "step": 20052 }, { "epoch": 2.4, "learning_rate": 1.5600719856028793e-06, "loss": 0.0002, "step": 20053 }, { "epoch": 2.4, "learning_rate": 1.56e-06, "loss": 0.0002, "step": 20054 }, { "epoch": 2.4, "learning_rate": 1.5599280143971205e-06, "loss": 0.0002, "step": 20055 }, { "epoch": 2.4, "learning_rate": 1.5598560287942412e-06, "loss": 0.0049, "step": 20056 }, { "epoch": 2.4, "learning_rate": 1.5597840431913618e-06, "loss": 0.0001, "step": 20057 }, { "epoch": 2.4, "learning_rate": 1.5597120575884824e-06, "loss": 0.042, "step": 20058 }, { "epoch": 2.4, "learning_rate": 1.559640071985603e-06, "loss": 0.0002, "step": 20059 }, { "epoch": 2.4, "learning_rate": 1.5595680863827237e-06, "loss": 0.0003, "step": 20060 }, { "epoch": 2.4, "learning_rate": 1.559496100779844e-06, "loss": 0.0002, "step": 20061 }, { "epoch": 2.4, "learning_rate": 1.5594241151769644e-06, "loss": 0.0001, "step": 20062 }, { "epoch": 2.4, "learning_rate": 1.5593521295740852e-06, "loss": 0.0001, "step": 20063 }, { "epoch": 2.4, "learning_rate": 1.5592801439712058e-06, "loss": 0.0001, "step": 20064 }, { "epoch": 2.4, "learning_rate": 1.5592081583683263e-06, "loss": 0.0002, "step": 20065 }, { "epoch": 2.4, "learning_rate": 1.559136172765447e-06, "loss": 0.0002, "step": 20066 }, { "epoch": 2.4, "learning_rate": 1.5590641871625677e-06, "loss": 0.0001, "step": 20067 }, { "epoch": 2.4, "learning_rate": 1.558992201559688e-06, "loss": 0.0003, "step": 20068 }, { "epoch": 2.4, "learning_rate": 1.5589202159568088e-06, "loss": 0.0002, "step": 20069 }, { "epoch": 2.41, "learning_rate": 1.5588482303539292e-06, "loss": 0.0004, "step": 20070 }, { "epoch": 2.41, "learning_rate": 1.55877624475105e-06, "loss": 0.0002, "step": 20071 }, { "epoch": 2.41, "learning_rate": 1.5587042591481703e-06, "loss": 0.0002, "step": 20072 }, { "epoch": 2.41, "learning_rate": 1.558632273545291e-06, "loss": 0.0002, "step": 20073 }, { "epoch": 2.41, "learning_rate": 1.5585602879424116e-06, "loss": 0.0002, "step": 20074 }, { "epoch": 2.41, "learning_rate": 1.558488302339532e-06, "loss": 0.0001, "step": 20075 }, { "epoch": 2.41, "learning_rate": 1.5584163167366528e-06, "loss": 0.0002, "step": 20076 }, { "epoch": 2.41, "learning_rate": 1.5583443311337731e-06, "loss": 0.0001, "step": 20077 }, { "epoch": 2.41, "learning_rate": 1.558272345530894e-06, "loss": 0.0002, "step": 20078 }, { "epoch": 2.41, "learning_rate": 1.5582003599280143e-06, "loss": 0.0001, "step": 20079 }, { "epoch": 2.41, "learning_rate": 1.558128374325135e-06, "loss": 0.0001, "step": 20080 }, { "epoch": 2.41, "learning_rate": 1.5580563887222556e-06, "loss": 0.0008, "step": 20081 }, { "epoch": 2.41, "learning_rate": 1.5579844031193764e-06, "loss": 0.0002, "step": 20082 }, { "epoch": 2.41, "learning_rate": 1.5579124175164967e-06, "loss": 0.0001, "step": 20083 }, { "epoch": 2.41, "learning_rate": 1.5578404319136175e-06, "loss": 0.0004, "step": 20084 }, { "epoch": 2.41, "learning_rate": 1.5577684463107379e-06, "loss": 0.0003, "step": 20085 }, { "epoch": 2.41, "learning_rate": 1.5576964607078582e-06, "loss": 0.0001, "step": 20086 }, { "epoch": 2.41, "learning_rate": 1.557624475104979e-06, "loss": 0.0001, "step": 20087 }, { "epoch": 2.41, "learning_rate": 1.5575524895020996e-06, "loss": 0.0002, "step": 20088 }, { "epoch": 2.41, "learning_rate": 1.5574805038992201e-06, "loss": 0.0002, "step": 20089 }, { "epoch": 2.41, "learning_rate": 1.5574085182963407e-06, "loss": 0.0003, "step": 20090 }, { "epoch": 2.41, "learning_rate": 1.5573365326934615e-06, "loss": 0.0001, "step": 20091 }, { "epoch": 2.41, "learning_rate": 1.5572645470905818e-06, "loss": 0.0002, "step": 20092 }, { "epoch": 2.41, "learning_rate": 1.5571925614877026e-06, "loss": 0.0001, "step": 20093 }, { "epoch": 2.41, "learning_rate": 1.557120575884823e-06, "loss": 0.0002, "step": 20094 }, { "epoch": 2.41, "learning_rate": 1.5570485902819438e-06, "loss": 0.0003, "step": 20095 }, { "epoch": 2.41, "learning_rate": 1.5569766046790641e-06, "loss": 0.0002, "step": 20096 }, { "epoch": 2.41, "learning_rate": 1.5569046190761847e-06, "loss": 0.0003, "step": 20097 }, { "epoch": 2.41, "learning_rate": 1.5568326334733055e-06, "loss": 0.0002, "step": 20098 }, { "epoch": 2.41, "learning_rate": 1.5567606478704258e-06, "loss": 0.0001, "step": 20099 }, { "epoch": 2.41, "learning_rate": 1.5566886622675466e-06, "loss": 0.0001, "step": 20100 }, { "epoch": 2.41, "learning_rate": 1.556616676664667e-06, "loss": 0.0003, "step": 20101 }, { "epoch": 2.41, "learning_rate": 1.5565446910617877e-06, "loss": 0.0001, "step": 20102 }, { "epoch": 2.41, "learning_rate": 1.556472705458908e-06, "loss": 0.0001, "step": 20103 }, { "epoch": 2.41, "learning_rate": 1.5564007198560289e-06, "loss": 0.0001, "step": 20104 }, { "epoch": 2.41, "learning_rate": 1.5563287342531494e-06, "loss": 0.0001, "step": 20105 }, { "epoch": 2.41, "learning_rate": 1.5562567486502702e-06, "loss": 0.0001, "step": 20106 }, { "epoch": 2.41, "learning_rate": 1.5561847630473906e-06, "loss": 0.0001, "step": 20107 }, { "epoch": 2.41, "learning_rate": 1.5561127774445113e-06, "loss": 0.0001, "step": 20108 }, { "epoch": 2.41, "learning_rate": 1.5560407918416317e-06, "loss": 0.0004, "step": 20109 }, { "epoch": 2.41, "learning_rate": 1.555968806238752e-06, "loss": 0.0002, "step": 20110 }, { "epoch": 2.41, "learning_rate": 1.5558968206358728e-06, "loss": 0.0002, "step": 20111 }, { "epoch": 2.41, "learning_rate": 1.5558248350329934e-06, "loss": 0.0002, "step": 20112 }, { "epoch": 2.41, "learning_rate": 1.5557528494301142e-06, "loss": 0.0002, "step": 20113 }, { "epoch": 2.41, "learning_rate": 1.5556808638272345e-06, "loss": 0.0001, "step": 20114 }, { "epoch": 2.41, "learning_rate": 1.5556088782243553e-06, "loss": 0.0001, "step": 20115 }, { "epoch": 2.41, "learning_rate": 1.5555368926214757e-06, "loss": 0.0001, "step": 20116 }, { "epoch": 2.41, "learning_rate": 1.5554649070185964e-06, "loss": 0.0003, "step": 20117 }, { "epoch": 2.41, "learning_rate": 1.5553929214157168e-06, "loss": 0.0001, "step": 20118 }, { "epoch": 2.41, "learning_rate": 1.5553209358128376e-06, "loss": 0.0003, "step": 20119 }, { "epoch": 2.41, "learning_rate": 1.5552489502099581e-06, "loss": 0.0002, "step": 20120 }, { "epoch": 2.41, "learning_rate": 1.5551769646070785e-06, "loss": 0.0002, "step": 20121 }, { "epoch": 2.41, "learning_rate": 1.5551049790041993e-06, "loss": 0.0001, "step": 20122 }, { "epoch": 2.41, "learning_rate": 1.5550329934013196e-06, "loss": 0.0001, "step": 20123 }, { "epoch": 2.41, "learning_rate": 1.5549610077984404e-06, "loss": 0.0202, "step": 20124 }, { "epoch": 2.41, "learning_rate": 1.5548890221955608e-06, "loss": 0.0001, "step": 20125 }, { "epoch": 2.41, "learning_rate": 1.5548170365926815e-06, "loss": 0.0008, "step": 20126 }, { "epoch": 2.41, "learning_rate": 1.5547450509898021e-06, "loss": 0.0002, "step": 20127 }, { "epoch": 2.41, "learning_rate": 1.5546730653869227e-06, "loss": 0.0002, "step": 20128 }, { "epoch": 2.41, "learning_rate": 1.5546010797840432e-06, "loss": 0.0001, "step": 20129 }, { "epoch": 2.41, "learning_rate": 1.554529094181164e-06, "loss": 0.0001, "step": 20130 }, { "epoch": 2.41, "learning_rate": 1.5544571085782844e-06, "loss": 0.0001, "step": 20131 }, { "epoch": 2.41, "learning_rate": 1.5543851229754047e-06, "loss": 0.0004, "step": 20132 }, { "epoch": 2.41, "learning_rate": 1.5543131373725255e-06, "loss": 0.0002, "step": 20133 }, { "epoch": 2.41, "learning_rate": 1.554241151769646e-06, "loss": 0.0001, "step": 20134 }, { "epoch": 2.41, "learning_rate": 1.5541691661667666e-06, "loss": 0.0002, "step": 20135 }, { "epoch": 2.41, "learning_rate": 1.5540971805638872e-06, "loss": 0.0002, "step": 20136 }, { "epoch": 2.41, "learning_rate": 1.554025194961008e-06, "loss": 0.0001, "step": 20137 }, { "epoch": 2.41, "learning_rate": 1.5539532093581283e-06, "loss": 0.0001, "step": 20138 }, { "epoch": 2.41, "learning_rate": 1.5538812237552491e-06, "loss": 0.0001, "step": 20139 }, { "epoch": 2.41, "learning_rate": 1.5538092381523695e-06, "loss": 0.0003, "step": 20140 }, { "epoch": 2.41, "learning_rate": 1.5537372525494903e-06, "loss": 0.0002, "step": 20141 }, { "epoch": 2.41, "learning_rate": 1.5536652669466106e-06, "loss": 0.0002, "step": 20142 }, { "epoch": 2.41, "learning_rate": 1.5535932813437312e-06, "loss": 0.0002, "step": 20143 }, { "epoch": 2.41, "learning_rate": 1.553521295740852e-06, "loss": 0.0002, "step": 20144 }, { "epoch": 2.41, "learning_rate": 1.5534493101379723e-06, "loss": 0.0002, "step": 20145 }, { "epoch": 2.41, "learning_rate": 1.553377324535093e-06, "loss": 0.0632, "step": 20146 }, { "epoch": 2.41, "learning_rate": 1.5533053389322134e-06, "loss": 0.0002, "step": 20147 }, { "epoch": 2.41, "learning_rate": 1.5532333533293342e-06, "loss": 0.0001, "step": 20148 }, { "epoch": 2.41, "learning_rate": 1.5531613677264546e-06, "loss": 0.0003, "step": 20149 }, { "epoch": 2.41, "learning_rate": 1.5530893821235754e-06, "loss": 0.0003, "step": 20150 }, { "epoch": 2.41, "learning_rate": 1.553017396520696e-06, "loss": 0.0009, "step": 20151 }, { "epoch": 2.41, "learning_rate": 1.5529454109178167e-06, "loss": 0.0002, "step": 20152 }, { "epoch": 2.41, "learning_rate": 1.552873425314937e-06, "loss": 0.0001, "step": 20153 }, { "epoch": 2.42, "learning_rate": 1.5528014397120578e-06, "loss": 0.0001, "step": 20154 }, { "epoch": 2.42, "learning_rate": 1.5527294541091782e-06, "loss": 1.07, "step": 20155 }, { "epoch": 2.42, "learning_rate": 1.5526574685062985e-06, "loss": 0.0001, "step": 20156 }, { "epoch": 2.42, "learning_rate": 1.5525854829034193e-06, "loss": 0.0002, "step": 20157 }, { "epoch": 2.42, "learning_rate": 1.5525134973005399e-06, "loss": 0.0002, "step": 20158 }, { "epoch": 2.42, "learning_rate": 1.5524415116976607e-06, "loss": 0.0002, "step": 20159 }, { "epoch": 2.42, "learning_rate": 1.552369526094781e-06, "loss": 0.0002, "step": 20160 }, { "epoch": 2.42, "learning_rate": 1.5522975404919018e-06, "loss": 0.0001, "step": 20161 }, { "epoch": 2.42, "learning_rate": 1.5522255548890222e-06, "loss": 0.0003, "step": 20162 }, { "epoch": 2.42, "learning_rate": 1.552153569286143e-06, "loss": 0.0001, "step": 20163 }, { "epoch": 2.42, "learning_rate": 1.5520815836832633e-06, "loss": 0.0001, "step": 20164 }, { "epoch": 2.42, "learning_rate": 1.552009598080384e-06, "loss": 0.0004, "step": 20165 }, { "epoch": 2.42, "learning_rate": 1.5519376124775046e-06, "loss": 0.0002, "step": 20166 }, { "epoch": 2.42, "learning_rate": 1.551865626874625e-06, "loss": 0.9629, "step": 20167 }, { "epoch": 2.42, "learning_rate": 1.5517936412717458e-06, "loss": 0.0002, "step": 20168 }, { "epoch": 2.42, "learning_rate": 1.5517216556688661e-06, "loss": 0.4021, "step": 20169 }, { "epoch": 2.42, "learning_rate": 1.551649670065987e-06, "loss": 0.0002, "step": 20170 }, { "epoch": 2.42, "learning_rate": 1.5515776844631073e-06, "loss": 0.7928, "step": 20171 }, { "epoch": 2.42, "learning_rate": 1.551505698860228e-06, "loss": 0.0002, "step": 20172 }, { "epoch": 2.42, "learning_rate": 1.5514337132573486e-06, "loss": 0.0002, "step": 20173 }, { "epoch": 2.42, "learning_rate": 1.5513617276544692e-06, "loss": 0.0003, "step": 20174 }, { "epoch": 2.42, "learning_rate": 1.5512897420515897e-06, "loss": 0.0011, "step": 20175 }, { "epoch": 2.42, "learning_rate": 1.5512177564487105e-06, "loss": 0.0001, "step": 20176 }, { "epoch": 2.42, "learning_rate": 1.5511457708458309e-06, "loss": 0.0004, "step": 20177 }, { "epoch": 2.42, "learning_rate": 1.5510737852429512e-06, "loss": 0.0001, "step": 20178 }, { "epoch": 2.42, "learning_rate": 1.551001799640072e-06, "loss": 0.0001, "step": 20179 }, { "epoch": 2.42, "learning_rate": 1.5509298140371926e-06, "loss": 0.0003, "step": 20180 }, { "epoch": 2.42, "learning_rate": 1.5508578284343131e-06, "loss": 0.0001, "step": 20181 }, { "epoch": 2.42, "learning_rate": 1.5507858428314337e-06, "loss": 0.0002, "step": 20182 }, { "epoch": 2.42, "learning_rate": 1.5507138572285545e-06, "loss": 0.0002, "step": 20183 }, { "epoch": 2.42, "learning_rate": 1.5506418716256748e-06, "loss": 0.0003, "step": 20184 }, { "epoch": 2.42, "learning_rate": 1.5505698860227956e-06, "loss": 0.0002, "step": 20185 }, { "epoch": 2.42, "learning_rate": 1.550497900419916e-06, "loss": 0.0001, "step": 20186 }, { "epoch": 2.42, "learning_rate": 1.5504259148170368e-06, "loss": 0.0006, "step": 20187 }, { "epoch": 2.42, "learning_rate": 1.5503539292141571e-06, "loss": 0.0002, "step": 20188 }, { "epoch": 2.42, "learning_rate": 1.5502819436112777e-06, "loss": 0.0084, "step": 20189 }, { "epoch": 2.42, "learning_rate": 1.5502099580083985e-06, "loss": 0.0002, "step": 20190 }, { "epoch": 2.42, "learning_rate": 1.5501379724055188e-06, "loss": 0.0003, "step": 20191 }, { "epoch": 2.42, "learning_rate": 1.5500659868026396e-06, "loss": 0.0007, "step": 20192 }, { "epoch": 2.42, "learning_rate": 1.54999400119976e-06, "loss": 0.0002, "step": 20193 }, { "epoch": 2.42, "learning_rate": 1.5499220155968807e-06, "loss": 0.0002, "step": 20194 }, { "epoch": 2.42, "learning_rate": 1.549850029994001e-06, "loss": 0.0002, "step": 20195 }, { "epoch": 2.42, "learning_rate": 1.5497780443911219e-06, "loss": 0.0002, "step": 20196 }, { "epoch": 2.42, "learning_rate": 1.5497060587882424e-06, "loss": 0.0002, "step": 20197 }, { "epoch": 2.42, "learning_rate": 1.5496340731853632e-06, "loss": 0.0001, "step": 20198 }, { "epoch": 2.42, "learning_rate": 1.5495620875824836e-06, "loss": 0.0003, "step": 20199 }, { "epoch": 2.42, "learning_rate": 1.5494901019796043e-06, "loss": 0.0003, "step": 20200 }, { "epoch": 2.42, "learning_rate": 1.5494181163767247e-06, "loss": 0.0002, "step": 20201 }, { "epoch": 2.42, "learning_rate": 1.549346130773845e-06, "loss": 0.0005, "step": 20202 }, { "epoch": 2.42, "learning_rate": 1.5492741451709658e-06, "loss": 0.0002, "step": 20203 }, { "epoch": 2.42, "learning_rate": 1.5492021595680864e-06, "loss": 0.0003, "step": 20204 }, { "epoch": 2.42, "learning_rate": 1.5491301739652072e-06, "loss": 0.0004, "step": 20205 }, { "epoch": 2.42, "learning_rate": 1.5490581883623275e-06, "loss": 0.0001, "step": 20206 }, { "epoch": 2.42, "learning_rate": 1.5489862027594483e-06, "loss": 0.0001, "step": 20207 }, { "epoch": 2.42, "learning_rate": 1.5489142171565687e-06, "loss": 0.0003, "step": 20208 }, { "epoch": 2.42, "learning_rate": 1.5488422315536894e-06, "loss": 0.0003, "step": 20209 }, { "epoch": 2.42, "learning_rate": 1.5487702459508098e-06, "loss": 0.0001, "step": 20210 }, { "epoch": 2.42, "learning_rate": 1.5486982603479306e-06, "loss": 0.0001, "step": 20211 }, { "epoch": 2.42, "learning_rate": 1.5486262747450511e-06, "loss": 0.0002, "step": 20212 }, { "epoch": 2.42, "learning_rate": 1.5485542891421715e-06, "loss": 0.0001, "step": 20213 }, { "epoch": 2.42, "learning_rate": 1.5484823035392923e-06, "loss": 0.0002, "step": 20214 }, { "epoch": 2.42, "learning_rate": 1.5484103179364126e-06, "loss": 0.0002, "step": 20215 }, { "epoch": 2.42, "learning_rate": 1.5483383323335334e-06, "loss": 0.0001, "step": 20216 }, { "epoch": 2.42, "learning_rate": 1.5482663467306538e-06, "loss": 0.0002, "step": 20217 }, { "epoch": 2.42, "learning_rate": 1.5481943611277745e-06, "loss": 0.0003, "step": 20218 }, { "epoch": 2.42, "learning_rate": 1.5481223755248949e-06, "loss": 0.0002, "step": 20219 }, { "epoch": 2.42, "learning_rate": 1.5480503899220157e-06, "loss": 0.0001, "step": 20220 }, { "epoch": 2.42, "learning_rate": 1.5479784043191362e-06, "loss": 0.0042, "step": 20221 }, { "epoch": 2.42, "learning_rate": 1.547906418716257e-06, "loss": 0.0002, "step": 20222 }, { "epoch": 2.42, "learning_rate": 1.5478344331133774e-06, "loss": 0.0002, "step": 20223 }, { "epoch": 2.42, "learning_rate": 1.5477624475104977e-06, "loss": 0.0002, "step": 20224 }, { "epoch": 2.42, "learning_rate": 1.5476904619076185e-06, "loss": 0.0001, "step": 20225 }, { "epoch": 2.42, "learning_rate": 1.5476184763047389e-06, "loss": 0.0002, "step": 20226 }, { "epoch": 2.42, "learning_rate": 1.5475464907018596e-06, "loss": 0.0001, "step": 20227 }, { "epoch": 2.42, "learning_rate": 1.5474745050989802e-06, "loss": 0.7156, "step": 20228 }, { "epoch": 2.42, "learning_rate": 1.547402519496101e-06, "loss": 0.0001, "step": 20229 }, { "epoch": 2.42, "learning_rate": 1.5473305338932213e-06, "loss": 0.0003, "step": 20230 }, { "epoch": 2.42, "learning_rate": 1.5472585482903421e-06, "loss": 0.0002, "step": 20231 }, { "epoch": 2.42, "learning_rate": 1.5471865626874625e-06, "loss": 0.0001, "step": 20232 }, { "epoch": 2.42, "learning_rate": 1.5471145770845832e-06, "loss": 0.0001, "step": 20233 }, { "epoch": 2.42, "learning_rate": 1.5470425914817036e-06, "loss": 0.0001, "step": 20234 }, { "epoch": 2.42, "learning_rate": 1.5469706058788244e-06, "loss": 0.0001, "step": 20235 }, { "epoch": 2.42, "learning_rate": 1.546898620275945e-06, "loss": 0.0446, "step": 20236 }, { "epoch": 2.43, "learning_rate": 1.5468266346730653e-06, "loss": 0.0002, "step": 20237 }, { "epoch": 2.43, "learning_rate": 1.546754649070186e-06, "loss": 0.0, "step": 20238 }, { "epoch": 2.43, "learning_rate": 1.5466826634673064e-06, "loss": 0.0001, "step": 20239 }, { "epoch": 2.43, "learning_rate": 1.5466106778644272e-06, "loss": 0.0002, "step": 20240 }, { "epoch": 2.43, "learning_rate": 1.5465386922615476e-06, "loss": 0.0001, "step": 20241 }, { "epoch": 2.43, "learning_rate": 1.5464667066586683e-06, "loss": 0.0002, "step": 20242 }, { "epoch": 2.43, "learning_rate": 1.546394721055789e-06, "loss": 0.0003, "step": 20243 }, { "epoch": 2.43, "learning_rate": 1.5463227354529095e-06, "loss": 0.0001, "step": 20244 }, { "epoch": 2.43, "learning_rate": 1.54625074985003e-06, "loss": 0.0001, "step": 20245 }, { "epoch": 2.43, "learning_rate": 1.5461787642471508e-06, "loss": 0.0001, "step": 20246 }, { "epoch": 2.43, "learning_rate": 1.5461067786442712e-06, "loss": 0.0002, "step": 20247 }, { "epoch": 2.43, "learning_rate": 1.5460347930413915e-06, "loss": 0.0001, "step": 20248 }, { "epoch": 2.43, "learning_rate": 1.5459628074385123e-06, "loss": 0.0001, "step": 20249 }, { "epoch": 2.43, "learning_rate": 1.5458908218356329e-06, "loss": 0.0001, "step": 20250 }, { "epoch": 2.43, "learning_rate": 1.5458188362327535e-06, "loss": 0.6149, "step": 20251 }, { "epoch": 2.43, "learning_rate": 1.545746850629874e-06, "loss": 0.0002, "step": 20252 }, { "epoch": 2.43, "learning_rate": 1.5456748650269948e-06, "loss": 0.0001, "step": 20253 }, { "epoch": 2.43, "learning_rate": 1.5456028794241152e-06, "loss": 0.0001, "step": 20254 }, { "epoch": 2.43, "learning_rate": 1.545530893821236e-06, "loss": 0.0001, "step": 20255 }, { "epoch": 2.43, "learning_rate": 1.5454589082183563e-06, "loss": 0.0001, "step": 20256 }, { "epoch": 2.43, "learning_rate": 1.545386922615477e-06, "loss": 0.0004, "step": 20257 }, { "epoch": 2.43, "learning_rate": 1.5453149370125974e-06, "loss": 0.0002, "step": 20258 }, { "epoch": 2.43, "learning_rate": 1.545242951409718e-06, "loss": 0.0002, "step": 20259 }, { "epoch": 2.43, "learning_rate": 1.5451709658068388e-06, "loss": 0.0001, "step": 20260 }, { "epoch": 2.43, "learning_rate": 1.5450989802039591e-06, "loss": 0.0002, "step": 20261 }, { "epoch": 2.43, "learning_rate": 1.54502699460108e-06, "loss": 0.0001, "step": 20262 }, { "epoch": 2.43, "learning_rate": 1.5449550089982003e-06, "loss": 0.0001, "step": 20263 }, { "epoch": 2.43, "learning_rate": 1.544883023395321e-06, "loss": 0.0004, "step": 20264 }, { "epoch": 2.43, "learning_rate": 1.5448110377924414e-06, "loss": 0.0296, "step": 20265 }, { "epoch": 2.43, "learning_rate": 1.5447390521895622e-06, "loss": 0.9792, "step": 20266 }, { "epoch": 2.43, "learning_rate": 1.5446670665866827e-06, "loss": 0.0001, "step": 20267 }, { "epoch": 2.43, "learning_rate": 1.5445950809838035e-06, "loss": 0.0001, "step": 20268 }, { "epoch": 2.43, "learning_rate": 1.5445230953809239e-06, "loss": 0.0001, "step": 20269 }, { "epoch": 2.43, "learning_rate": 1.5444511097780442e-06, "loss": 0.0001, "step": 20270 }, { "epoch": 2.43, "learning_rate": 1.544379124175165e-06, "loss": 0.0001, "step": 20271 }, { "epoch": 2.43, "learning_rate": 1.5443071385722854e-06, "loss": 0.0001, "step": 20272 }, { "epoch": 2.43, "learning_rate": 1.5442351529694061e-06, "loss": 0.0005, "step": 20273 }, { "epoch": 2.43, "learning_rate": 1.5441631673665267e-06, "loss": 1.1511, "step": 20274 }, { "epoch": 2.43, "learning_rate": 1.5440911817636475e-06, "loss": 0.0002, "step": 20275 }, { "epoch": 2.43, "learning_rate": 1.5440191961607678e-06, "loss": 0.3476, "step": 20276 }, { "epoch": 2.43, "learning_rate": 1.5439472105578886e-06, "loss": 0.0002, "step": 20277 }, { "epoch": 2.43, "learning_rate": 1.543875224955009e-06, "loss": 0.1156, "step": 20278 }, { "epoch": 2.43, "learning_rate": 1.5438032393521297e-06, "loss": 0.0001, "step": 20279 }, { "epoch": 2.43, "learning_rate": 1.54373125374925e-06, "loss": 0.0001, "step": 20280 }, { "epoch": 2.43, "learning_rate": 1.5436592681463709e-06, "loss": 0.0001, "step": 20281 }, { "epoch": 2.43, "learning_rate": 1.5435872825434914e-06, "loss": 0.0008, "step": 20282 }, { "epoch": 2.43, "learning_rate": 1.5435152969406118e-06, "loss": 0.0005, "step": 20283 }, { "epoch": 2.43, "learning_rate": 1.5434433113377326e-06, "loss": 0.0, "step": 20284 }, { "epoch": 2.43, "learning_rate": 1.543371325734853e-06, "loss": 0.0004, "step": 20285 }, { "epoch": 2.43, "learning_rate": 1.5432993401319737e-06, "loss": 0.0002, "step": 20286 }, { "epoch": 2.43, "learning_rate": 1.543227354529094e-06, "loss": 0.0002, "step": 20287 }, { "epoch": 2.43, "learning_rate": 1.5431553689262148e-06, "loss": 0.0001, "step": 20288 }, { "epoch": 2.43, "learning_rate": 1.5430833833233354e-06, "loss": 0.0005, "step": 20289 }, { "epoch": 2.43, "learning_rate": 1.543011397720456e-06, "loss": 0.0003, "step": 20290 }, { "epoch": 2.43, "learning_rate": 1.5429394121175765e-06, "loss": 0.1543, "step": 20291 }, { "epoch": 2.43, "learning_rate": 1.5428674265146973e-06, "loss": 0.0001, "step": 20292 }, { "epoch": 2.43, "learning_rate": 1.5427954409118177e-06, "loss": 0.0003, "step": 20293 }, { "epoch": 2.43, "learning_rate": 1.542723455308938e-06, "loss": 0.0002, "step": 20294 }, { "epoch": 2.43, "learning_rate": 1.5426514697060588e-06, "loss": 0.0003, "step": 20295 }, { "epoch": 2.43, "learning_rate": 1.5425794841031794e-06, "loss": 0.0002, "step": 20296 }, { "epoch": 2.43, "learning_rate": 1.5425074985003e-06, "loss": 0.0002, "step": 20297 }, { "epoch": 2.43, "learning_rate": 1.5424355128974205e-06, "loss": 0.0002, "step": 20298 }, { "epoch": 2.43, "learning_rate": 1.5423635272945413e-06, "loss": 0.0001, "step": 20299 }, { "epoch": 2.43, "learning_rate": 1.5422915416916616e-06, "loss": 0.0002, "step": 20300 }, { "epoch": 2.43, "learning_rate": 1.5422195560887824e-06, "loss": 0.0465, "step": 20301 }, { "epoch": 2.43, "learning_rate": 1.5421475704859028e-06, "loss": 0.0003, "step": 20302 }, { "epoch": 2.43, "learning_rate": 1.5420755848830236e-06, "loss": 0.0001, "step": 20303 }, { "epoch": 2.43, "learning_rate": 1.542003599280144e-06, "loss": 0.0001, "step": 20304 }, { "epoch": 2.43, "learning_rate": 1.5419316136772645e-06, "loss": 0.0001, "step": 20305 }, { "epoch": 2.43, "learning_rate": 1.5418596280743853e-06, "loss": 0.0002, "step": 20306 }, { "epoch": 2.43, "learning_rate": 1.5417876424715056e-06, "loss": 0.0001, "step": 20307 }, { "epoch": 2.43, "learning_rate": 1.5417156568686264e-06, "loss": 0.0001, "step": 20308 }, { "epoch": 2.43, "learning_rate": 1.5416436712657467e-06, "loss": 0.0001, "step": 20309 }, { "epoch": 2.43, "learning_rate": 1.5415716856628675e-06, "loss": 0.6641, "step": 20310 }, { "epoch": 2.43, "learning_rate": 1.5414997000599879e-06, "loss": 0.0002, "step": 20311 }, { "epoch": 2.43, "learning_rate": 1.5414277144571087e-06, "loss": 0.0001, "step": 20312 }, { "epoch": 2.43, "learning_rate": 1.5413557288542292e-06, "loss": 0.0001, "step": 20313 }, { "epoch": 2.43, "learning_rate": 1.54128374325135e-06, "loss": 0.0001, "step": 20314 }, { "epoch": 2.43, "learning_rate": 1.5412117576484704e-06, "loss": 0.0001, "step": 20315 }, { "epoch": 2.43, "learning_rate": 1.5411397720455907e-06, "loss": 0.0002, "step": 20316 }, { "epoch": 2.43, "learning_rate": 1.5410677864427115e-06, "loss": 0.0001, "step": 20317 }, { "epoch": 2.43, "learning_rate": 1.5409958008398318e-06, "loss": 0.0004, "step": 20318 }, { "epoch": 2.43, "learning_rate": 1.5409238152369526e-06, "loss": 0.0001, "step": 20319 }, { "epoch": 2.43, "learning_rate": 1.5408518296340732e-06, "loss": 0.0003, "step": 20320 }, { "epoch": 2.44, "learning_rate": 1.540779844031194e-06, "loss": 0.0001, "step": 20321 }, { "epoch": 2.44, "learning_rate": 1.5407078584283143e-06, "loss": 0.0001, "step": 20322 }, { "epoch": 2.44, "learning_rate": 1.5406358728254351e-06, "loss": 0.0003, "step": 20323 }, { "epoch": 2.44, "learning_rate": 1.5405638872225555e-06, "loss": 0.0002, "step": 20324 }, { "epoch": 2.44, "learning_rate": 1.5404919016196762e-06, "loss": 0.0024, "step": 20325 }, { "epoch": 2.44, "learning_rate": 1.5404199160167966e-06, "loss": 0.0001, "step": 20326 }, { "epoch": 2.44, "learning_rate": 1.5403479304139174e-06, "loss": 0.0001, "step": 20327 }, { "epoch": 2.44, "learning_rate": 1.540275944811038e-06, "loss": 0.0003, "step": 20328 }, { "epoch": 2.44, "learning_rate": 1.5402039592081583e-06, "loss": 0.0037, "step": 20329 }, { "epoch": 2.44, "learning_rate": 1.540131973605279e-06, "loss": 0.0, "step": 20330 }, { "epoch": 2.44, "learning_rate": 1.5400599880023994e-06, "loss": 0.0002, "step": 20331 }, { "epoch": 2.44, "learning_rate": 1.5399880023995202e-06, "loss": 1.7811, "step": 20332 }, { "epoch": 2.44, "learning_rate": 1.5399160167966406e-06, "loss": 0.0001, "step": 20333 }, { "epoch": 2.44, "learning_rate": 1.5398440311937613e-06, "loss": 0.0002, "step": 20334 }, { "epoch": 2.44, "learning_rate": 1.539772045590882e-06, "loss": 0.0001, "step": 20335 }, { "epoch": 2.44, "learning_rate": 1.5397000599880025e-06, "loss": 0.0002, "step": 20336 }, { "epoch": 2.44, "learning_rate": 1.539628074385123e-06, "loss": 0.0002, "step": 20337 }, { "epoch": 2.44, "learning_rate": 1.5395560887822438e-06, "loss": 0.0003, "step": 20338 }, { "epoch": 2.44, "learning_rate": 1.5394841031793642e-06, "loss": 0.0001, "step": 20339 }, { "epoch": 2.44, "learning_rate": 1.5394121175764845e-06, "loss": 0.0001, "step": 20340 }, { "epoch": 2.44, "learning_rate": 1.5393401319736053e-06, "loss": 0.0001, "step": 20341 }, { "epoch": 2.44, "learning_rate": 1.5392681463707259e-06, "loss": 0.0001, "step": 20342 }, { "epoch": 2.44, "learning_rate": 1.5391961607678464e-06, "loss": 0.0001, "step": 20343 }, { "epoch": 2.44, "learning_rate": 1.539124175164967e-06, "loss": 0.0001, "step": 20344 }, { "epoch": 2.44, "learning_rate": 1.5390521895620878e-06, "loss": 0.0003, "step": 20345 }, { "epoch": 2.44, "learning_rate": 1.5389802039592081e-06, "loss": 0.0002, "step": 20346 }, { "epoch": 2.44, "learning_rate": 1.538908218356329e-06, "loss": 0.0001, "step": 20347 }, { "epoch": 2.44, "learning_rate": 1.5388362327534493e-06, "loss": 0.0005, "step": 20348 }, { "epoch": 2.44, "learning_rate": 1.53876424715057e-06, "loss": 0.0001, "step": 20349 }, { "epoch": 2.44, "learning_rate": 1.5386922615476904e-06, "loss": 0.0003, "step": 20350 }, { "epoch": 2.44, "learning_rate": 1.538620275944811e-06, "loss": 0.0002, "step": 20351 }, { "epoch": 2.44, "learning_rate": 1.5385482903419318e-06, "loss": 0.0001, "step": 20352 }, { "epoch": 2.44, "learning_rate": 1.5384763047390521e-06, "loss": 0.0002, "step": 20353 }, { "epoch": 2.44, "learning_rate": 1.5384043191361729e-06, "loss": 0.0001, "step": 20354 }, { "epoch": 2.44, "learning_rate": 1.5383323335332932e-06, "loss": 0.0004, "step": 20355 }, { "epoch": 2.44, "learning_rate": 1.538260347930414e-06, "loss": 0.0002, "step": 20356 }, { "epoch": 2.44, "learning_rate": 1.5381883623275344e-06, "loss": 0.0002, "step": 20357 }, { "epoch": 2.44, "learning_rate": 1.5381163767246552e-06, "loss": 0.0003, "step": 20358 }, { "epoch": 2.44, "learning_rate": 1.5380443911217757e-06, "loss": 0.0002, "step": 20359 }, { "epoch": 2.44, "learning_rate": 1.5379724055188965e-06, "loss": 0.0001, "step": 20360 }, { "epoch": 2.44, "learning_rate": 1.5379004199160169e-06, "loss": 0.0003, "step": 20361 }, { "epoch": 2.44, "learning_rate": 1.5378284343131376e-06, "loss": 0.0004, "step": 20362 }, { "epoch": 2.44, "learning_rate": 1.537756448710258e-06, "loss": 0.0002, "step": 20363 }, { "epoch": 2.44, "learning_rate": 1.5376844631073783e-06, "loss": 0.0001, "step": 20364 }, { "epoch": 2.44, "learning_rate": 1.5376124775044991e-06, "loss": 0.0004, "step": 20365 }, { "epoch": 2.44, "learning_rate": 1.5375404919016197e-06, "loss": 0.0002, "step": 20366 }, { "epoch": 2.44, "learning_rate": 1.5374685062987403e-06, "loss": 0.0002, "step": 20367 }, { "epoch": 2.44, "learning_rate": 1.5373965206958608e-06, "loss": 0.0002, "step": 20368 }, { "epoch": 2.44, "learning_rate": 1.5373245350929816e-06, "loss": 0.0002, "step": 20369 }, { "epoch": 2.44, "learning_rate": 1.537252549490102e-06, "loss": 0.0, "step": 20370 }, { "epoch": 2.44, "learning_rate": 1.5371805638872227e-06, "loss": 0.0009, "step": 20371 }, { "epoch": 2.44, "learning_rate": 1.537108578284343e-06, "loss": 0.001, "step": 20372 }, { "epoch": 2.44, "learning_rate": 1.5370365926814639e-06, "loss": 0.0001, "step": 20373 }, { "epoch": 2.44, "learning_rate": 1.5369646070785842e-06, "loss": 0.0002, "step": 20374 }, { "epoch": 2.44, "learning_rate": 1.5368926214757048e-06, "loss": 0.0001, "step": 20375 }, { "epoch": 2.44, "learning_rate": 1.5368206358728256e-06, "loss": 0.0002, "step": 20376 }, { "epoch": 2.44, "learning_rate": 1.536748650269946e-06, "loss": 0.0002, "step": 20377 }, { "epoch": 2.44, "learning_rate": 1.5366766646670667e-06, "loss": 0.0002, "step": 20378 }, { "epoch": 2.44, "learning_rate": 1.536604679064187e-06, "loss": 0.0001, "step": 20379 }, { "epoch": 2.44, "learning_rate": 1.5365326934613078e-06, "loss": 0.0003, "step": 20380 }, { "epoch": 2.44, "learning_rate": 1.5364607078584282e-06, "loss": 0.0001, "step": 20381 }, { "epoch": 2.44, "learning_rate": 1.536388722255549e-06, "loss": 0.0001, "step": 20382 }, { "epoch": 2.44, "learning_rate": 1.5363167366526695e-06, "loss": 0.0001, "step": 20383 }, { "epoch": 2.44, "learning_rate": 1.5362447510497903e-06, "loss": 0.0002, "step": 20384 }, { "epoch": 2.44, "learning_rate": 1.5361727654469107e-06, "loss": 0.0001, "step": 20385 }, { "epoch": 2.44, "learning_rate": 1.536100779844031e-06, "loss": 0.0003, "step": 20386 }, { "epoch": 2.44, "learning_rate": 1.5360287942411518e-06, "loss": 0.0002, "step": 20387 }, { "epoch": 2.44, "learning_rate": 1.5359568086382722e-06, "loss": 0.0001, "step": 20388 }, { "epoch": 2.44, "learning_rate": 1.535884823035393e-06, "loss": 0.0002, "step": 20389 }, { "epoch": 2.44, "learning_rate": 1.5358128374325135e-06, "loss": 0.0004, "step": 20390 }, { "epoch": 2.44, "learning_rate": 1.5357408518296343e-06, "loss": 0.0002, "step": 20391 }, { "epoch": 2.44, "learning_rate": 1.5356688662267546e-06, "loss": 0.0017, "step": 20392 }, { "epoch": 2.44, "learning_rate": 1.5355968806238754e-06, "loss": 0.0001, "step": 20393 }, { "epoch": 2.44, "learning_rate": 1.5355248950209958e-06, "loss": 0.8951, "step": 20394 }, { "epoch": 2.44, "learning_rate": 1.5354529094181166e-06, "loss": 0.0002, "step": 20395 }, { "epoch": 2.44, "learning_rate": 1.535380923815237e-06, "loss": 0.0001, "step": 20396 }, { "epoch": 2.44, "learning_rate": 1.5353089382123575e-06, "loss": 0.0004, "step": 20397 }, { "epoch": 2.44, "learning_rate": 1.5352369526094783e-06, "loss": 1.186, "step": 20398 }, { "epoch": 2.44, "learning_rate": 1.5351649670065986e-06, "loss": 0.0001, "step": 20399 }, { "epoch": 2.44, "learning_rate": 1.5350929814037194e-06, "loss": 0.0001, "step": 20400 }, { "epoch": 2.44, "learning_rate": 1.5350209958008397e-06, "loss": 0.9838, "step": 20401 }, { "epoch": 2.44, "learning_rate": 1.5349490101979605e-06, "loss": 0.0001, "step": 20402 }, { "epoch": 2.44, "learning_rate": 1.5348770245950809e-06, "loss": 0.0001, "step": 20403 }, { "epoch": 2.45, "learning_rate": 1.5348050389922017e-06, "loss": 0.0004, "step": 20404 }, { "epoch": 2.45, "learning_rate": 1.5347330533893222e-06, "loss": 0.0002, "step": 20405 }, { "epoch": 2.45, "learning_rate": 1.5346610677864428e-06, "loss": 0.0003, "step": 20406 }, { "epoch": 2.45, "learning_rate": 1.5345890821835634e-06, "loss": 0.0001, "step": 20407 }, { "epoch": 2.45, "learning_rate": 1.5345170965806841e-06, "loss": 0.0001, "step": 20408 }, { "epoch": 2.45, "learning_rate": 1.5344451109778045e-06, "loss": 0.0001, "step": 20409 }, { "epoch": 2.45, "learning_rate": 1.5343731253749248e-06, "loss": 0.0001, "step": 20410 }, { "epoch": 2.45, "learning_rate": 1.5343011397720456e-06, "loss": 0.0001, "step": 20411 }, { "epoch": 2.45, "learning_rate": 1.5342291541691662e-06, "loss": 0.0002, "step": 20412 }, { "epoch": 2.45, "learning_rate": 1.5341571685662868e-06, "loss": 0.0001, "step": 20413 }, { "epoch": 2.45, "learning_rate": 1.5340851829634073e-06, "loss": 0.0007, "step": 20414 }, { "epoch": 2.45, "learning_rate": 1.534013197360528e-06, "loss": 0.0001, "step": 20415 }, { "epoch": 2.45, "learning_rate": 1.5339412117576485e-06, "loss": 0.0001, "step": 20416 }, { "epoch": 2.45, "learning_rate": 1.5338692261547692e-06, "loss": 0.0001, "step": 20417 }, { "epoch": 2.45, "learning_rate": 1.5337972405518896e-06, "loss": 0.0001, "step": 20418 }, { "epoch": 2.45, "learning_rate": 1.5337252549490104e-06, "loss": 0.0001, "step": 20419 }, { "epoch": 2.45, "learning_rate": 1.5336532693461307e-06, "loss": 0.0001, "step": 20420 }, { "epoch": 2.45, "learning_rate": 1.5335812837432513e-06, "loss": 0.0001, "step": 20421 }, { "epoch": 2.45, "learning_rate": 1.533509298140372e-06, "loss": 0.0002, "step": 20422 }, { "epoch": 2.45, "learning_rate": 1.5334373125374924e-06, "loss": 0.0003, "step": 20423 }, { "epoch": 2.45, "learning_rate": 1.5333653269346132e-06, "loss": 0.0001, "step": 20424 }, { "epoch": 2.45, "learning_rate": 1.5332933413317336e-06, "loss": 0.541, "step": 20425 }, { "epoch": 2.45, "learning_rate": 1.5332213557288543e-06, "loss": 0.0002, "step": 20426 }, { "epoch": 2.45, "learning_rate": 1.5331493701259747e-06, "loss": 0.842, "step": 20427 }, { "epoch": 2.45, "learning_rate": 1.5330773845230955e-06, "loss": 0.0001, "step": 20428 }, { "epoch": 2.45, "learning_rate": 1.533005398920216e-06, "loss": 0.0001, "step": 20429 }, { "epoch": 2.45, "learning_rate": 1.5329334133173368e-06, "loss": 0.0001, "step": 20430 }, { "epoch": 2.45, "learning_rate": 1.5328614277144572e-06, "loss": 0.0019, "step": 20431 }, { "epoch": 2.45, "learning_rate": 1.5327894421115775e-06, "loss": 0.0004, "step": 20432 }, { "epoch": 2.45, "learning_rate": 1.5327174565086983e-06, "loss": 0.0002, "step": 20433 }, { "epoch": 2.45, "learning_rate": 1.5326454709058187e-06, "loss": 0.0002, "step": 20434 }, { "epoch": 2.45, "learning_rate": 1.5325734853029394e-06, "loss": 0.0011, "step": 20435 }, { "epoch": 2.45, "learning_rate": 1.53250149970006e-06, "loss": 0.0001, "step": 20436 }, { "epoch": 2.45, "learning_rate": 1.5324295140971808e-06, "loss": 0.0001, "step": 20437 }, { "epoch": 2.45, "learning_rate": 1.5323575284943011e-06, "loss": 0.0022, "step": 20438 }, { "epoch": 2.45, "learning_rate": 1.532285542891422e-06, "loss": 0.0003, "step": 20439 }, { "epoch": 2.45, "learning_rate": 1.5322135572885423e-06, "loss": 0.0005, "step": 20440 }, { "epoch": 2.45, "learning_rate": 1.532141571685663e-06, "loss": 0.0002, "step": 20441 }, { "epoch": 2.45, "learning_rate": 1.5320695860827834e-06, "loss": 0.0006, "step": 20442 }, { "epoch": 2.45, "learning_rate": 1.531997600479904e-06, "loss": 0.0002, "step": 20443 }, { "epoch": 2.45, "learning_rate": 1.5319256148770247e-06, "loss": 0.0001, "step": 20444 }, { "epoch": 2.45, "learning_rate": 1.531853629274145e-06, "loss": 0.0001, "step": 20445 }, { "epoch": 2.45, "learning_rate": 1.5317816436712659e-06, "loss": 0.024, "step": 20446 }, { "epoch": 2.45, "learning_rate": 1.5317096580683862e-06, "loss": 0.0003, "step": 20447 }, { "epoch": 2.45, "learning_rate": 1.531637672465507e-06, "loss": 0.0002, "step": 20448 }, { "epoch": 2.45, "learning_rate": 1.5315656868626274e-06, "loss": 0.0002, "step": 20449 }, { "epoch": 2.45, "learning_rate": 1.5314937012597481e-06, "loss": 0.544, "step": 20450 }, { "epoch": 2.45, "learning_rate": 1.5314217156568687e-06, "loss": 0.0002, "step": 20451 }, { "epoch": 2.45, "learning_rate": 1.5313497300539893e-06, "loss": 0.0004, "step": 20452 }, { "epoch": 2.45, "learning_rate": 1.5312777444511098e-06, "loss": 0.0017, "step": 20453 }, { "epoch": 2.45, "learning_rate": 1.5312057588482306e-06, "loss": 0.0001, "step": 20454 }, { "epoch": 2.45, "learning_rate": 1.531133773245351e-06, "loss": 0.0004, "step": 20455 }, { "epoch": 2.45, "learning_rate": 1.5310617876424713e-06, "loss": 0.0004, "step": 20456 }, { "epoch": 2.45, "learning_rate": 1.5309898020395921e-06, "loss": 0.0009, "step": 20457 }, { "epoch": 2.45, "learning_rate": 1.5309178164367127e-06, "loss": 0.0001, "step": 20458 }, { "epoch": 2.45, "learning_rate": 1.5308458308338332e-06, "loss": 0.0001, "step": 20459 }, { "epoch": 2.45, "learning_rate": 1.5307738452309538e-06, "loss": 0.0005, "step": 20460 }, { "epoch": 2.45, "learning_rate": 1.5307018596280746e-06, "loss": 0.0001, "step": 20461 }, { "epoch": 2.45, "learning_rate": 1.530629874025195e-06, "loss": 0.0001, "step": 20462 }, { "epoch": 2.45, "learning_rate": 1.5305578884223157e-06, "loss": 0.5579, "step": 20463 }, { "epoch": 2.45, "learning_rate": 1.530485902819436e-06, "loss": 0.0001, "step": 20464 }, { "epoch": 2.45, "learning_rate": 1.5304139172165569e-06, "loss": 0.0002, "step": 20465 }, { "epoch": 2.45, "learning_rate": 1.5303419316136772e-06, "loss": 0.0002, "step": 20466 }, { "epoch": 2.45, "learning_rate": 1.5302699460107978e-06, "loss": 0.0002, "step": 20467 }, { "epoch": 2.45, "learning_rate": 1.5301979604079186e-06, "loss": 0.0003, "step": 20468 }, { "epoch": 2.45, "learning_rate": 1.530125974805039e-06, "loss": 0.0001, "step": 20469 }, { "epoch": 2.45, "learning_rate": 1.5300539892021597e-06, "loss": 0.0003, "step": 20470 }, { "epoch": 2.45, "learning_rate": 1.52998200359928e-06, "loss": 0.0003, "step": 20471 }, { "epoch": 2.45, "learning_rate": 1.5299100179964008e-06, "loss": 0.0001, "step": 20472 }, { "epoch": 2.45, "learning_rate": 1.5298380323935212e-06, "loss": 0.0002, "step": 20473 }, { "epoch": 2.45, "learning_rate": 1.529766046790642e-06, "loss": 0.0001, "step": 20474 }, { "epoch": 2.45, "learning_rate": 1.5296940611877625e-06, "loss": 0.0001, "step": 20475 }, { "epoch": 2.45, "learning_rate": 1.5296220755848833e-06, "loss": 0.0002, "step": 20476 }, { "epoch": 2.45, "learning_rate": 1.5295500899820037e-06, "loss": 0.0002, "step": 20477 }, { "epoch": 2.45, "learning_rate": 1.529478104379124e-06, "loss": 0.0002, "step": 20478 }, { "epoch": 2.45, "learning_rate": 1.5294061187762448e-06, "loss": 0.0039, "step": 20479 }, { "epoch": 2.45, "learning_rate": 1.5293341331733652e-06, "loss": 0.0001, "step": 20480 }, { "epoch": 2.45, "learning_rate": 1.529262147570486e-06, "loss": 0.0001, "step": 20481 }, { "epoch": 2.45, "learning_rate": 1.5291901619676065e-06, "loss": 0.0002, "step": 20482 }, { "epoch": 2.45, "learning_rate": 1.5291181763647273e-06, "loss": 0.0001, "step": 20483 }, { "epoch": 2.45, "learning_rate": 1.5290461907618476e-06, "loss": 0.0001, "step": 20484 }, { "epoch": 2.45, "learning_rate": 1.5289742051589684e-06, "loss": 0.0001, "step": 20485 }, { "epoch": 2.45, "learning_rate": 1.5289022195560888e-06, "loss": 0.0001, "step": 20486 }, { "epoch": 2.46, "learning_rate": 1.5288302339532095e-06, "loss": 0.0002, "step": 20487 }, { "epoch": 2.46, "learning_rate": 1.52875824835033e-06, "loss": 0.0002, "step": 20488 }, { "epoch": 2.46, "learning_rate": 1.5286862627474507e-06, "loss": 0.0002, "step": 20489 }, { "epoch": 2.46, "learning_rate": 1.528614277144571e-06, "loss": 0.0001, "step": 20490 }, { "epoch": 2.46, "learning_rate": 1.5285422915416916e-06, "loss": 0.0001, "step": 20491 }, { "epoch": 2.46, "learning_rate": 1.5284703059388124e-06, "loss": 0.0001, "step": 20492 }, { "epoch": 2.46, "learning_rate": 1.5283983203359327e-06, "loss": 0.0002, "step": 20493 }, { "epoch": 2.46, "learning_rate": 1.5283263347330535e-06, "loss": 0.0001, "step": 20494 }, { "epoch": 2.46, "learning_rate": 1.5282543491301739e-06, "loss": 0.0, "step": 20495 }, { "epoch": 2.46, "learning_rate": 1.5281823635272946e-06, "loss": 0.0003, "step": 20496 }, { "epoch": 2.46, "learning_rate": 1.528110377924415e-06, "loss": 0.0001, "step": 20497 }, { "epoch": 2.46, "learning_rate": 1.5280383923215358e-06, "loss": 0.0003, "step": 20498 }, { "epoch": 2.46, "learning_rate": 1.5279664067186563e-06, "loss": 0.0001, "step": 20499 }, { "epoch": 2.46, "learning_rate": 1.5278944211157771e-06, "loss": 0.0001, "step": 20500 }, { "epoch": 2.46, "learning_rate": 1.5278224355128975e-06, "loss": 0.0002, "step": 20501 }, { "epoch": 2.46, "learning_rate": 1.5277504499100178e-06, "loss": 0.0001, "step": 20502 }, { "epoch": 2.46, "learning_rate": 1.5276784643071386e-06, "loss": 0.0002, "step": 20503 }, { "epoch": 2.46, "learning_rate": 1.527606478704259e-06, "loss": 0.0002, "step": 20504 }, { "epoch": 2.46, "learning_rate": 1.5275344931013797e-06, "loss": 0.0002, "step": 20505 }, { "epoch": 2.46, "learning_rate": 1.5274625074985003e-06, "loss": 0.0001, "step": 20506 }, { "epoch": 2.46, "learning_rate": 1.527390521895621e-06, "loss": 0.0002, "step": 20507 }, { "epoch": 2.46, "learning_rate": 1.5273185362927414e-06, "loss": 0.0002, "step": 20508 }, { "epoch": 2.46, "learning_rate": 1.5272465506898622e-06, "loss": 0.0002, "step": 20509 }, { "epoch": 2.46, "learning_rate": 1.5271745650869826e-06, "loss": 0.0002, "step": 20510 }, { "epoch": 2.46, "learning_rate": 1.5271025794841034e-06, "loss": 0.0001, "step": 20511 }, { "epoch": 2.46, "learning_rate": 1.5270305938812237e-06, "loss": 0.0001, "step": 20512 }, { "epoch": 2.46, "learning_rate": 1.5269586082783443e-06, "loss": 0.0001, "step": 20513 }, { "epoch": 2.46, "learning_rate": 1.526886622675465e-06, "loss": 0.0001, "step": 20514 }, { "epoch": 2.46, "learning_rate": 1.5268146370725854e-06, "loss": 0.0001, "step": 20515 }, { "epoch": 2.46, "learning_rate": 1.5267426514697062e-06, "loss": 0.0002, "step": 20516 }, { "epoch": 2.46, "learning_rate": 1.5266706658668265e-06, "loss": 0.0001, "step": 20517 }, { "epoch": 2.46, "learning_rate": 1.5265986802639473e-06, "loss": 0.0003, "step": 20518 }, { "epoch": 2.46, "learning_rate": 1.5265266946610677e-06, "loss": 0.0001, "step": 20519 }, { "epoch": 2.46, "learning_rate": 1.5264547090581885e-06, "loss": 0.0001, "step": 20520 }, { "epoch": 2.46, "learning_rate": 1.526382723455309e-06, "loss": 0.0005, "step": 20521 }, { "epoch": 2.46, "learning_rate": 1.5263107378524296e-06, "loss": 0.0003, "step": 20522 }, { "epoch": 2.46, "learning_rate": 1.5262387522495502e-06, "loss": 0.0001, "step": 20523 }, { "epoch": 2.46, "learning_rate": 1.5261667666466705e-06, "loss": 0.0001, "step": 20524 }, { "epoch": 2.46, "learning_rate": 1.5260947810437913e-06, "loss": 0.0004, "step": 20525 }, { "epoch": 2.46, "learning_rate": 1.5260227954409116e-06, "loss": 0.0002, "step": 20526 }, { "epoch": 2.46, "learning_rate": 1.5259508098380324e-06, "loss": 0.0001, "step": 20527 }, { "epoch": 2.46, "learning_rate": 1.525878824235153e-06, "loss": 0.0001, "step": 20528 }, { "epoch": 2.46, "learning_rate": 1.5258068386322736e-06, "loss": 0.3683, "step": 20529 }, { "epoch": 2.46, "learning_rate": 1.5257348530293941e-06, "loss": 0.0001, "step": 20530 }, { "epoch": 2.46, "learning_rate": 1.525662867426515e-06, "loss": 0.0001, "step": 20531 }, { "epoch": 2.46, "learning_rate": 1.5255908818236353e-06, "loss": 0.0001, "step": 20532 }, { "epoch": 2.46, "learning_rate": 1.525518896220756e-06, "loss": 0.0001, "step": 20533 }, { "epoch": 2.46, "learning_rate": 1.5254469106178764e-06, "loss": 0.0002, "step": 20534 }, { "epoch": 2.46, "learning_rate": 1.5253749250149972e-06, "loss": 0.0002, "step": 20535 }, { "epoch": 2.46, "learning_rate": 1.5253029394121175e-06, "loss": 0.0001, "step": 20536 }, { "epoch": 2.46, "learning_rate": 1.525230953809238e-06, "loss": 0.0001, "step": 20537 }, { "epoch": 2.46, "learning_rate": 1.5251589682063589e-06, "loss": 0.0004, "step": 20538 }, { "epoch": 2.46, "learning_rate": 1.5250869826034792e-06, "loss": 0.0002, "step": 20539 }, { "epoch": 2.46, "learning_rate": 1.5250149970006e-06, "loss": 0.0003, "step": 20540 }, { "epoch": 2.46, "learning_rate": 1.5249430113977204e-06, "loss": 0.0003, "step": 20541 }, { "epoch": 2.46, "learning_rate": 1.5248710257948411e-06, "loss": 0.0004, "step": 20542 }, { "epoch": 2.46, "learning_rate": 1.5247990401919615e-06, "loss": 0.0001, "step": 20543 }, { "epoch": 2.46, "learning_rate": 1.5247270545890823e-06, "loss": 0.0, "step": 20544 }, { "epoch": 2.46, "learning_rate": 1.5246550689862028e-06, "loss": 0.0001, "step": 20545 }, { "epoch": 2.46, "learning_rate": 1.5245830833833236e-06, "loss": 0.0002, "step": 20546 }, { "epoch": 2.46, "learning_rate": 1.524511097780444e-06, "loss": 0.0002, "step": 20547 }, { "epoch": 2.46, "learning_rate": 1.5244391121775643e-06, "loss": 0.0003, "step": 20548 }, { "epoch": 2.46, "learning_rate": 1.5243671265746851e-06, "loss": 0.0005, "step": 20549 }, { "epoch": 2.46, "learning_rate": 1.5242951409718055e-06, "loss": 0.438, "step": 20550 }, { "epoch": 2.46, "learning_rate": 1.5242231553689262e-06, "loss": 0.0001, "step": 20551 }, { "epoch": 2.46, "learning_rate": 1.5241511697660468e-06, "loss": 0.0006, "step": 20552 }, { "epoch": 2.46, "learning_rate": 1.5240791841631676e-06, "loss": 0.0011, "step": 20553 }, { "epoch": 2.46, "learning_rate": 1.524007198560288e-06, "loss": 0.0001, "step": 20554 }, { "epoch": 2.46, "learning_rate": 1.5239352129574087e-06, "loss": 0.0001, "step": 20555 }, { "epoch": 2.46, "learning_rate": 1.523863227354529e-06, "loss": 0.0001, "step": 20556 }, { "epoch": 2.46, "learning_rate": 1.5237912417516499e-06, "loss": 0.0002, "step": 20557 }, { "epoch": 2.46, "learning_rate": 1.5237192561487702e-06, "loss": 0.0002, "step": 20558 }, { "epoch": 2.46, "learning_rate": 1.5236472705458908e-06, "loss": 1.1015, "step": 20559 }, { "epoch": 2.46, "learning_rate": 1.5235752849430116e-06, "loss": 0.0001, "step": 20560 }, { "epoch": 2.46, "learning_rate": 1.523503299340132e-06, "loss": 0.0001, "step": 20561 }, { "epoch": 2.46, "learning_rate": 1.5234313137372527e-06, "loss": 0.0001, "step": 20562 }, { "epoch": 2.46, "learning_rate": 1.523359328134373e-06, "loss": 0.0001, "step": 20563 }, { "epoch": 2.46, "learning_rate": 1.5232873425314938e-06, "loss": 0.0003, "step": 20564 }, { "epoch": 2.46, "learning_rate": 1.5232153569286142e-06, "loss": 0.0156, "step": 20565 }, { "epoch": 2.46, "learning_rate": 1.523143371325735e-06, "loss": 0.0002, "step": 20566 }, { "epoch": 2.46, "learning_rate": 1.5230713857228555e-06, "loss": 0.0002, "step": 20567 }, { "epoch": 2.46, "learning_rate": 1.522999400119976e-06, "loss": 0.0001, "step": 20568 }, { "epoch": 2.46, "learning_rate": 1.5229274145170967e-06, "loss": 0.1537, "step": 20569 }, { "epoch": 2.46, "learning_rate": 1.522855428914217e-06, "loss": 0.0003, "step": 20570 }, { "epoch": 2.47, "learning_rate": 1.5227834433113378e-06, "loss": 0.0001, "step": 20571 }, { "epoch": 2.47, "learning_rate": 1.5227114577084581e-06, "loss": 0.0001, "step": 20572 }, { "epoch": 2.47, "learning_rate": 1.522639472105579e-06, "loss": 0.0031, "step": 20573 }, { "epoch": 2.47, "learning_rate": 1.5225674865026995e-06, "loss": 0.0001, "step": 20574 }, { "epoch": 2.47, "learning_rate": 1.52249550089982e-06, "loss": 0.0001, "step": 20575 }, { "epoch": 2.47, "learning_rate": 1.5224235152969406e-06, "loss": 0.0001, "step": 20576 }, { "epoch": 2.47, "learning_rate": 1.5223515296940614e-06, "loss": 0.0005, "step": 20577 }, { "epoch": 2.47, "learning_rate": 1.5222795440911818e-06, "loss": 0.0009, "step": 20578 }, { "epoch": 2.47, "learning_rate": 1.5222075584883025e-06, "loss": 0.0001, "step": 20579 }, { "epoch": 2.47, "learning_rate": 1.5221355728854229e-06, "loss": 0.0001, "step": 20580 }, { "epoch": 2.47, "learning_rate": 1.5220635872825437e-06, "loss": 0.0001, "step": 20581 }, { "epoch": 2.47, "learning_rate": 1.521991601679664e-06, "loss": 0.0001, "step": 20582 }, { "epoch": 2.47, "learning_rate": 1.5219196160767846e-06, "loss": 0.0001, "step": 20583 }, { "epoch": 2.47, "learning_rate": 1.5218476304739054e-06, "loss": 0.0003, "step": 20584 }, { "epoch": 2.47, "learning_rate": 1.5217756448710257e-06, "loss": 0.7152, "step": 20585 }, { "epoch": 2.47, "learning_rate": 1.5217036592681465e-06, "loss": 0.0001, "step": 20586 }, { "epoch": 2.47, "learning_rate": 1.5216316736652669e-06, "loss": 0.0002, "step": 20587 }, { "epoch": 2.47, "learning_rate": 1.5215596880623876e-06, "loss": 0.0027, "step": 20588 }, { "epoch": 2.47, "learning_rate": 1.521487702459508e-06, "loss": 0.0001, "step": 20589 }, { "epoch": 2.47, "learning_rate": 1.5214157168566288e-06, "loss": 0.0001, "step": 20590 }, { "epoch": 2.47, "learning_rate": 1.5213437312537493e-06, "loss": 0.0002, "step": 20591 }, { "epoch": 2.47, "learning_rate": 1.5212717456508701e-06, "loss": 0.0001, "step": 20592 }, { "epoch": 2.47, "learning_rate": 1.5211997600479905e-06, "loss": 0.0001, "step": 20593 }, { "epoch": 2.47, "learning_rate": 1.5211277744451108e-06, "loss": 0.0002, "step": 20594 }, { "epoch": 2.47, "learning_rate": 1.5210557888422316e-06, "loss": 0.0001, "step": 20595 }, { "epoch": 2.47, "learning_rate": 1.520983803239352e-06, "loss": 0.0003, "step": 20596 }, { "epoch": 2.47, "learning_rate": 1.5209118176364727e-06, "loss": 0.0001, "step": 20597 }, { "epoch": 2.47, "learning_rate": 1.5208398320335933e-06, "loss": 0.0001, "step": 20598 }, { "epoch": 2.47, "learning_rate": 1.520767846430714e-06, "loss": 0.0002, "step": 20599 }, { "epoch": 2.47, "learning_rate": 1.5206958608278344e-06, "loss": 0.0001, "step": 20600 }, { "epoch": 2.47, "learning_rate": 1.5206238752249552e-06, "loss": 0.0001, "step": 20601 }, { "epoch": 2.47, "learning_rate": 1.5205518896220756e-06, "loss": 0.0002, "step": 20602 }, { "epoch": 2.47, "learning_rate": 1.5204799040191964e-06, "loss": 0.0001, "step": 20603 }, { "epoch": 2.47, "learning_rate": 1.5204079184163167e-06, "loss": 0.0012, "step": 20604 }, { "epoch": 2.47, "learning_rate": 1.5203359328134373e-06, "loss": 0.0003, "step": 20605 }, { "epoch": 2.47, "learning_rate": 1.520263947210558e-06, "loss": 0.0002, "step": 20606 }, { "epoch": 2.47, "learning_rate": 1.5201919616076784e-06, "loss": 0.0001, "step": 20607 }, { "epoch": 2.47, "learning_rate": 1.5201199760047992e-06, "loss": 0.0001, "step": 20608 }, { "epoch": 2.47, "learning_rate": 1.5200479904019195e-06, "loss": 0.0001, "step": 20609 }, { "epoch": 2.47, "learning_rate": 1.5199760047990403e-06, "loss": 0.0001, "step": 20610 }, { "epoch": 2.47, "learning_rate": 1.5199040191961607e-06, "loss": 0.0007, "step": 20611 }, { "epoch": 2.47, "learning_rate": 1.5198320335932815e-06, "loss": 0.0002, "step": 20612 }, { "epoch": 2.47, "learning_rate": 1.519760047990402e-06, "loss": 0.0002, "step": 20613 }, { "epoch": 2.47, "learning_rate": 1.5196880623875226e-06, "loss": 0.0001, "step": 20614 }, { "epoch": 2.47, "learning_rate": 1.5196160767846432e-06, "loss": 0.0001, "step": 20615 }, { "epoch": 2.47, "learning_rate": 1.519544091181764e-06, "loss": 0.0002, "step": 20616 }, { "epoch": 2.47, "learning_rate": 1.5194721055788843e-06, "loss": 0.0001, "step": 20617 }, { "epoch": 2.47, "learning_rate": 1.5194001199760046e-06, "loss": 0.0001, "step": 20618 }, { "epoch": 2.47, "learning_rate": 1.5193281343731254e-06, "loss": 0.0001, "step": 20619 }, { "epoch": 2.47, "learning_rate": 1.5192561487702458e-06, "loss": 0.0001, "step": 20620 }, { "epoch": 2.47, "learning_rate": 1.5191841631673666e-06, "loss": 0.0008, "step": 20621 }, { "epoch": 2.47, "learning_rate": 1.5191121775644871e-06, "loss": 0.0001, "step": 20622 }, { "epoch": 2.47, "learning_rate": 1.519040191961608e-06, "loss": 0.1336, "step": 20623 }, { "epoch": 2.47, "learning_rate": 1.5189682063587283e-06, "loss": 0.0001, "step": 20624 }, { "epoch": 2.47, "learning_rate": 1.518896220755849e-06, "loss": 0.0002, "step": 20625 }, { "epoch": 2.47, "learning_rate": 1.5188242351529694e-06, "loss": 0.0001, "step": 20626 }, { "epoch": 2.47, "learning_rate": 1.5187522495500902e-06, "loss": 0.0004, "step": 20627 }, { "epoch": 2.47, "learning_rate": 1.5186802639472105e-06, "loss": 0.0001, "step": 20628 }, { "epoch": 2.47, "learning_rate": 1.518608278344331e-06, "loss": 0.0001, "step": 20629 }, { "epoch": 2.47, "learning_rate": 1.5185362927414519e-06, "loss": 0.0001, "step": 20630 }, { "epoch": 2.47, "learning_rate": 1.5184643071385722e-06, "loss": 0.9477, "step": 20631 }, { "epoch": 2.47, "learning_rate": 1.518392321535693e-06, "loss": 1.1082, "step": 20632 }, { "epoch": 2.47, "learning_rate": 1.5183203359328134e-06, "loss": 0.001, "step": 20633 }, { "epoch": 2.47, "learning_rate": 1.5182483503299341e-06, "loss": 0.0002, "step": 20634 }, { "epoch": 2.47, "learning_rate": 1.5181763647270545e-06, "loss": 0.0002, "step": 20635 }, { "epoch": 2.47, "learning_rate": 1.5181043791241753e-06, "loss": 0.0002, "step": 20636 }, { "epoch": 2.47, "learning_rate": 1.5180323935212958e-06, "loss": 0.0003, "step": 20637 }, { "epoch": 2.47, "learning_rate": 1.5179604079184164e-06, "loss": 0.0002, "step": 20638 }, { "epoch": 2.47, "learning_rate": 1.517888422315537e-06, "loss": 0.0002, "step": 20639 }, { "epoch": 2.47, "learning_rate": 1.5178164367126573e-06, "loss": 0.0001, "step": 20640 }, { "epoch": 2.47, "learning_rate": 1.517744451109778e-06, "loss": 0.0001, "step": 20641 }, { "epoch": 2.47, "learning_rate": 1.5176724655068985e-06, "loss": 0.0001, "step": 20642 }, { "epoch": 2.47, "learning_rate": 1.5176004799040192e-06, "loss": 0.0001, "step": 20643 }, { "epoch": 2.47, "learning_rate": 1.5175284943011398e-06, "loss": 0.0001, "step": 20644 }, { "epoch": 2.47, "learning_rate": 1.5174565086982604e-06, "loss": 0.0002, "step": 20645 }, { "epoch": 2.47, "learning_rate": 1.517384523095381e-06, "loss": 0.0001, "step": 20646 }, { "epoch": 2.47, "learning_rate": 1.5173125374925017e-06, "loss": 0.0002, "step": 20647 }, { "epoch": 2.47, "learning_rate": 1.517240551889622e-06, "loss": 0.0001, "step": 20648 }, { "epoch": 2.47, "learning_rate": 1.5171685662867428e-06, "loss": 0.0002, "step": 20649 }, { "epoch": 2.47, "learning_rate": 1.5170965806838632e-06, "loss": 0.0002, "step": 20650 }, { "epoch": 2.47, "learning_rate": 1.5170245950809838e-06, "loss": 0.0001, "step": 20651 }, { "epoch": 2.47, "learning_rate": 1.5169526094781043e-06, "loss": 0.0002, "step": 20652 }, { "epoch": 2.47, "learning_rate": 1.516880623875225e-06, "loss": 0.0002, "step": 20653 }, { "epoch": 2.48, "learning_rate": 1.5168086382723457e-06, "loss": 0.0003, "step": 20654 }, { "epoch": 2.48, "learning_rate": 1.516736652669466e-06, "loss": 0.0003, "step": 20655 }, { "epoch": 2.48, "learning_rate": 1.5166646670665868e-06, "loss": 0.0001, "step": 20656 }, { "epoch": 2.48, "learning_rate": 1.5165926814637072e-06, "loss": 0.0003, "step": 20657 }, { "epoch": 2.48, "learning_rate": 1.516520695860828e-06, "loss": 0.0002, "step": 20658 }, { "epoch": 2.48, "learning_rate": 1.5164487102579483e-06, "loss": 0.0003, "step": 20659 }, { "epoch": 2.48, "learning_rate": 1.516376724655069e-06, "loss": 0.0002, "step": 20660 }, { "epoch": 2.48, "learning_rate": 1.5163047390521896e-06, "loss": 0.0016, "step": 20661 }, { "epoch": 2.48, "learning_rate": 1.5162327534493104e-06, "loss": 0.0002, "step": 20662 }, { "epoch": 2.48, "learning_rate": 1.5161607678464308e-06, "loss": 0.0001, "step": 20663 }, { "epoch": 2.48, "learning_rate": 1.5160887822435511e-06, "loss": 0.0005, "step": 20664 }, { "epoch": 2.48, "learning_rate": 1.516016796640672e-06, "loss": 0.0002, "step": 20665 }, { "epoch": 2.48, "learning_rate": 1.5159448110377923e-06, "loss": 0.0008, "step": 20666 }, { "epoch": 2.48, "learning_rate": 1.515872825434913e-06, "loss": 0.0003, "step": 20667 }, { "epoch": 2.48, "learning_rate": 1.5158008398320336e-06, "loss": 0.0002, "step": 20668 }, { "epoch": 2.48, "learning_rate": 1.5157288542291544e-06, "loss": 0.0004, "step": 20669 }, { "epoch": 2.48, "learning_rate": 1.5156568686262747e-06, "loss": 0.0002, "step": 20670 }, { "epoch": 2.48, "learning_rate": 1.5155848830233955e-06, "loss": 0.0002, "step": 20671 }, { "epoch": 2.48, "learning_rate": 1.5155128974205159e-06, "loss": 0.0004, "step": 20672 }, { "epoch": 2.48, "learning_rate": 1.5154409118176367e-06, "loss": 0.0002, "step": 20673 }, { "epoch": 2.48, "learning_rate": 1.515368926214757e-06, "loss": 0.0001, "step": 20674 }, { "epoch": 2.48, "learning_rate": 1.5152969406118776e-06, "loss": 0.0002, "step": 20675 }, { "epoch": 2.48, "learning_rate": 1.5152249550089984e-06, "loss": 0.0002, "step": 20676 }, { "epoch": 2.48, "learning_rate": 1.5151529694061187e-06, "loss": 0.0002, "step": 20677 }, { "epoch": 2.48, "learning_rate": 1.5150809838032395e-06, "loss": 0.0001, "step": 20678 }, { "epoch": 2.48, "learning_rate": 1.5150089982003599e-06, "loss": 0.0001, "step": 20679 }, { "epoch": 2.48, "learning_rate": 1.5149370125974806e-06, "loss": 0.0003, "step": 20680 }, { "epoch": 2.48, "learning_rate": 1.514865026994601e-06, "loss": 0.0002, "step": 20681 }, { "epoch": 2.48, "learning_rate": 1.5147930413917218e-06, "loss": 0.4196, "step": 20682 }, { "epoch": 2.48, "learning_rate": 1.5147210557888423e-06, "loss": 0.0002, "step": 20683 }, { "epoch": 2.48, "learning_rate": 1.514649070185963e-06, "loss": 0.0002, "step": 20684 }, { "epoch": 2.48, "learning_rate": 1.5145770845830835e-06, "loss": 0.0003, "step": 20685 }, { "epoch": 2.48, "learning_rate": 1.5145050989802038e-06, "loss": 0.0001, "step": 20686 }, { "epoch": 2.48, "learning_rate": 1.5144331133773246e-06, "loss": 0.0004, "step": 20687 }, { "epoch": 2.48, "learning_rate": 1.514361127774445e-06, "loss": 0.0001, "step": 20688 }, { "epoch": 2.48, "learning_rate": 1.5142891421715657e-06, "loss": 0.0001, "step": 20689 }, { "epoch": 2.48, "learning_rate": 1.5142171565686863e-06, "loss": 0.0005, "step": 20690 }, { "epoch": 2.48, "learning_rate": 1.5141451709658069e-06, "loss": 0.0002, "step": 20691 }, { "epoch": 2.48, "learning_rate": 1.5140731853629274e-06, "loss": 0.0002, "step": 20692 }, { "epoch": 2.48, "learning_rate": 1.5140011997600482e-06, "loss": 0.0002, "step": 20693 }, { "epoch": 2.48, "learning_rate": 1.5139292141571686e-06, "loss": 0.0002, "step": 20694 }, { "epoch": 2.48, "learning_rate": 1.5138572285542893e-06, "loss": 0.0002, "step": 20695 }, { "epoch": 2.48, "learning_rate": 1.5137852429514097e-06, "loss": 0.0002, "step": 20696 }, { "epoch": 2.48, "learning_rate": 1.5137132573485303e-06, "loss": 0.0003, "step": 20697 }, { "epoch": 2.48, "learning_rate": 1.5136412717456508e-06, "loss": 0.0001, "step": 20698 }, { "epoch": 2.48, "learning_rate": 1.5135692861427714e-06, "loss": 0.0001, "step": 20699 }, { "epoch": 2.48, "learning_rate": 1.5134973005398922e-06, "loss": 0.0002, "step": 20700 }, { "epoch": 2.48, "learning_rate": 1.5134253149370125e-06, "loss": 0.0001, "step": 20701 }, { "epoch": 2.48, "learning_rate": 1.5133533293341333e-06, "loss": 0.0002, "step": 20702 }, { "epoch": 2.48, "learning_rate": 1.5132813437312537e-06, "loss": 0.001, "step": 20703 }, { "epoch": 2.48, "learning_rate": 1.5132093581283744e-06, "loss": 0.0002, "step": 20704 }, { "epoch": 2.48, "learning_rate": 1.5131373725254948e-06, "loss": 0.0001, "step": 20705 }, { "epoch": 2.48, "learning_rate": 1.5130653869226156e-06, "loss": 0.0003, "step": 20706 }, { "epoch": 2.48, "learning_rate": 1.5129934013197361e-06, "loss": 0.0009, "step": 20707 }, { "epoch": 2.48, "learning_rate": 1.512921415716857e-06, "loss": 0.0002, "step": 20708 }, { "epoch": 2.48, "learning_rate": 1.5128494301139773e-06, "loss": 0.1098, "step": 20709 }, { "epoch": 2.48, "learning_rate": 1.5127774445110976e-06, "loss": 0.0001, "step": 20710 }, { "epoch": 2.48, "learning_rate": 1.5127054589082184e-06, "loss": 0.609, "step": 20711 }, { "epoch": 2.48, "learning_rate": 1.5126334733053388e-06, "loss": 0.0002, "step": 20712 }, { "epoch": 2.48, "learning_rate": 1.5125614877024595e-06, "loss": 0.0002, "step": 20713 }, { "epoch": 2.48, "learning_rate": 1.5124895020995801e-06, "loss": 0.0002, "step": 20714 }, { "epoch": 2.48, "learning_rate": 1.5124175164967009e-06, "loss": 0.0002, "step": 20715 }, { "epoch": 2.48, "learning_rate": 1.5123455308938212e-06, "loss": 0.0004, "step": 20716 }, { "epoch": 2.48, "learning_rate": 1.512273545290942e-06, "loss": 0.0001, "step": 20717 }, { "epoch": 2.48, "learning_rate": 1.5122015596880624e-06, "loss": 0.0005, "step": 20718 }, { "epoch": 2.48, "learning_rate": 1.5121295740851832e-06, "loss": 0.0003, "step": 20719 }, { "epoch": 2.48, "learning_rate": 1.5120575884823035e-06, "loss": 0.0003, "step": 20720 }, { "epoch": 2.48, "learning_rate": 1.511985602879424e-06, "loss": 0.3418, "step": 20721 }, { "epoch": 2.48, "learning_rate": 1.5119136172765449e-06, "loss": 0.0001, "step": 20722 }, { "epoch": 2.48, "learning_rate": 1.5118416316736652e-06, "loss": 0.0001, "step": 20723 }, { "epoch": 2.48, "learning_rate": 1.511769646070786e-06, "loss": 0.0002, "step": 20724 }, { "epoch": 2.48, "learning_rate": 1.5116976604679063e-06, "loss": 0.0002, "step": 20725 }, { "epoch": 2.48, "learning_rate": 1.5116256748650271e-06, "loss": 0.0001, "step": 20726 }, { "epoch": 2.48, "learning_rate": 1.5115536892621475e-06, "loss": 0.0001, "step": 20727 }, { "epoch": 2.48, "learning_rate": 1.5114817036592683e-06, "loss": 0.0002, "step": 20728 }, { "epoch": 2.48, "learning_rate": 1.5114097180563888e-06, "loss": 0.0002, "step": 20729 }, { "epoch": 2.48, "learning_rate": 1.5113377324535094e-06, "loss": 0.0001, "step": 20730 }, { "epoch": 2.48, "learning_rate": 1.51126574685063e-06, "loss": 0.0219, "step": 20731 }, { "epoch": 2.48, "learning_rate": 1.5111937612477503e-06, "loss": 0.0001, "step": 20732 }, { "epoch": 2.48, "learning_rate": 1.511121775644871e-06, "loss": 0.0002, "step": 20733 }, { "epoch": 2.48, "learning_rate": 1.5110497900419914e-06, "loss": 0.8012, "step": 20734 }, { "epoch": 2.48, "learning_rate": 1.5109778044391122e-06, "loss": 0.0006, "step": 20735 }, { "epoch": 2.48, "learning_rate": 1.5109058188362328e-06, "loss": 0.0004, "step": 20736 }, { "epoch": 2.48, "learning_rate": 1.5108338332333534e-06, "loss": 0.0003, "step": 20737 }, { "epoch": 2.49, "learning_rate": 1.510761847630474e-06, "loss": 0.0002, "step": 20738 }, { "epoch": 2.49, "learning_rate": 1.5106898620275947e-06, "loss": 0.0001, "step": 20739 }, { "epoch": 2.49, "learning_rate": 1.510617876424715e-06, "loss": 0.0001, "step": 20740 }, { "epoch": 2.49, "learning_rate": 1.5105458908218358e-06, "loss": 0.0013, "step": 20741 }, { "epoch": 2.49, "learning_rate": 1.5104739052189562e-06, "loss": 0.0003, "step": 20742 }, { "epoch": 2.49, "learning_rate": 1.510401919616077e-06, "loss": 0.0001, "step": 20743 }, { "epoch": 2.49, "learning_rate": 1.5103299340131973e-06, "loss": 0.0003, "step": 20744 }, { "epoch": 2.49, "learning_rate": 1.5102579484103179e-06, "loss": 0.0002, "step": 20745 }, { "epoch": 2.49, "learning_rate": 1.5101859628074387e-06, "loss": 0.0002, "step": 20746 }, { "epoch": 2.49, "learning_rate": 1.510113977204559e-06, "loss": 0.573, "step": 20747 }, { "epoch": 2.49, "learning_rate": 1.5100419916016798e-06, "loss": 0.0003, "step": 20748 }, { "epoch": 2.49, "learning_rate": 1.5099700059988002e-06, "loss": 0.0003, "step": 20749 }, { "epoch": 2.49, "learning_rate": 1.509898020395921e-06, "loss": 0.0001, "step": 20750 }, { "epoch": 2.49, "learning_rate": 1.5098260347930413e-06, "loss": 0.0001, "step": 20751 }, { "epoch": 2.49, "learning_rate": 1.509754049190162e-06, "loss": 0.0002, "step": 20752 }, { "epoch": 2.49, "learning_rate": 1.5096820635872826e-06, "loss": 0.0002, "step": 20753 }, { "epoch": 2.49, "learning_rate": 1.5096100779844034e-06, "loss": 0.0001, "step": 20754 }, { "epoch": 2.49, "learning_rate": 1.5095380923815238e-06, "loss": 0.0001, "step": 20755 }, { "epoch": 2.49, "learning_rate": 1.5094661067786441e-06, "loss": 0.0005, "step": 20756 }, { "epoch": 2.49, "learning_rate": 1.509394121175765e-06, "loss": 0.0002, "step": 20757 }, { "epoch": 2.49, "learning_rate": 1.5093221355728853e-06, "loss": 0.0001, "step": 20758 }, { "epoch": 2.49, "learning_rate": 1.509250149970006e-06, "loss": 0.0001, "step": 20759 }, { "epoch": 2.49, "learning_rate": 1.5091781643671266e-06, "loss": 0.0001, "step": 20760 }, { "epoch": 2.49, "learning_rate": 1.5091061787642472e-06, "loss": 0.0003, "step": 20761 }, { "epoch": 2.49, "learning_rate": 1.5090341931613677e-06, "loss": 0.0536, "step": 20762 }, { "epoch": 2.49, "learning_rate": 1.5089622075584885e-06, "loss": 0.0001, "step": 20763 }, { "epoch": 2.49, "learning_rate": 1.5088902219556089e-06, "loss": 0.6442, "step": 20764 }, { "epoch": 2.49, "learning_rate": 1.5088182363527297e-06, "loss": 0.0002, "step": 20765 }, { "epoch": 2.49, "learning_rate": 1.50874625074985e-06, "loss": 0.0002, "step": 20766 }, { "epoch": 2.49, "learning_rate": 1.5086742651469706e-06, "loss": 0.3858, "step": 20767 }, { "epoch": 2.49, "learning_rate": 1.5086022795440911e-06, "loss": 0.0002, "step": 20768 }, { "epoch": 2.49, "learning_rate": 1.5085302939412117e-06, "loss": 0.0003, "step": 20769 }, { "epoch": 2.49, "learning_rate": 1.5084583083383325e-06, "loss": 0.0002, "step": 20770 }, { "epoch": 2.49, "learning_rate": 1.5083863227354528e-06, "loss": 0.0003, "step": 20771 }, { "epoch": 2.49, "learning_rate": 1.5083143371325736e-06, "loss": 0.0016, "step": 20772 }, { "epoch": 2.49, "learning_rate": 1.508242351529694e-06, "loss": 0.0001, "step": 20773 }, { "epoch": 2.49, "learning_rate": 1.5081703659268148e-06, "loss": 0.0002, "step": 20774 }, { "epoch": 2.49, "learning_rate": 1.5080983803239351e-06, "loss": 0.0001, "step": 20775 }, { "epoch": 2.49, "learning_rate": 1.5080263947210559e-06, "loss": 0.0002, "step": 20776 }, { "epoch": 2.49, "learning_rate": 1.5079544091181765e-06, "loss": 0.0002, "step": 20777 }, { "epoch": 2.49, "learning_rate": 1.5078824235152968e-06, "loss": 0.0001, "step": 20778 }, { "epoch": 2.49, "learning_rate": 1.5078104379124176e-06, "loss": 0.0002, "step": 20779 }, { "epoch": 2.49, "learning_rate": 1.507738452309538e-06, "loss": 0.0001, "step": 20780 }, { "epoch": 2.49, "learning_rate": 1.5076664667066587e-06, "loss": 0.0002, "step": 20781 }, { "epoch": 2.49, "learning_rate": 1.507594481103779e-06, "loss": 0.0001, "step": 20782 }, { "epoch": 2.49, "learning_rate": 1.5075224955008999e-06, "loss": 0.0001, "step": 20783 }, { "epoch": 2.49, "learning_rate": 1.5074505098980204e-06, "loss": 0.0002, "step": 20784 }, { "epoch": 2.49, "learning_rate": 1.5073785242951412e-06, "loss": 0.0001, "step": 20785 }, { "epoch": 2.49, "learning_rate": 1.5073065386922616e-06, "loss": 0.0003, "step": 20786 }, { "epoch": 2.49, "learning_rate": 1.5072345530893823e-06, "loss": 0.0005, "step": 20787 }, { "epoch": 2.49, "learning_rate": 1.5071625674865027e-06, "loss": 0.0002, "step": 20788 }, { "epoch": 2.49, "learning_rate": 1.5070905818836235e-06, "loss": 0.0001, "step": 20789 }, { "epoch": 2.49, "learning_rate": 1.5070185962807438e-06, "loss": 0.0002, "step": 20790 }, { "epoch": 2.49, "learning_rate": 1.5069466106778644e-06, "loss": 0.0002, "step": 20791 }, { "epoch": 2.49, "learning_rate": 1.5068746250749852e-06, "loss": 0.0003, "step": 20792 }, { "epoch": 2.49, "learning_rate": 1.5068026394721055e-06, "loss": 0.0003, "step": 20793 }, { "epoch": 2.49, "learning_rate": 1.5067306538692263e-06, "loss": 0.0002, "step": 20794 }, { "epoch": 2.49, "learning_rate": 1.5066586682663467e-06, "loss": 0.9555, "step": 20795 }, { "epoch": 2.49, "learning_rate": 1.5065866826634674e-06, "loss": 0.0003, "step": 20796 }, { "epoch": 2.49, "learning_rate": 1.5065146970605878e-06, "loss": 0.7178, "step": 20797 }, { "epoch": 2.49, "learning_rate": 1.5064427114577086e-06, "loss": 0.0002, "step": 20798 }, { "epoch": 2.49, "learning_rate": 1.5063707258548291e-06, "loss": 0.0003, "step": 20799 }, { "epoch": 2.49, "learning_rate": 1.5062987402519497e-06, "loss": 0.0001, "step": 20800 }, { "epoch": 2.49, "learning_rate": 1.5062267546490703e-06, "loss": 0.0005, "step": 20801 }, { "epoch": 2.49, "learning_rate": 1.5061547690461906e-06, "loss": 0.0002, "step": 20802 }, { "epoch": 2.49, "learning_rate": 1.5060827834433114e-06, "loss": 0.0002, "step": 20803 }, { "epoch": 2.49, "learning_rate": 1.5060107978404318e-06, "loss": 0.0003, "step": 20804 }, { "epoch": 2.49, "learning_rate": 1.5059388122375525e-06, "loss": 0.0002, "step": 20805 }, { "epoch": 2.49, "learning_rate": 1.505866826634673e-06, "loss": 0.0002, "step": 20806 }, { "epoch": 2.49, "learning_rate": 1.5057948410317937e-06, "loss": 0.0004, "step": 20807 }, { "epoch": 2.49, "learning_rate": 1.5057228554289142e-06, "loss": 0.0001, "step": 20808 }, { "epoch": 2.49, "learning_rate": 1.505650869826035e-06, "loss": 0.0003, "step": 20809 }, { "epoch": 2.49, "learning_rate": 1.5055788842231554e-06, "loss": 0.0001, "step": 20810 }, { "epoch": 2.49, "learning_rate": 1.5055068986202761e-06, "loss": 0.0001, "step": 20811 }, { "epoch": 2.49, "learning_rate": 1.5054349130173965e-06, "loss": 0.0002, "step": 20812 }, { "epoch": 2.49, "learning_rate": 1.505362927414517e-06, "loss": 0.0004, "step": 20813 }, { "epoch": 2.49, "learning_rate": 1.5052909418116376e-06, "loss": 0.0002, "step": 20814 }, { "epoch": 2.49, "learning_rate": 1.5052189562087582e-06, "loss": 0.0002, "step": 20815 }, { "epoch": 2.49, "learning_rate": 1.505146970605879e-06, "loss": 0.0001, "step": 20816 }, { "epoch": 2.49, "learning_rate": 1.5050749850029993e-06, "loss": 0.0001, "step": 20817 }, { "epoch": 2.49, "learning_rate": 1.5050029994001201e-06, "loss": 0.0003, "step": 20818 }, { "epoch": 2.49, "learning_rate": 1.5049310137972405e-06, "loss": 0.0001, "step": 20819 }, { "epoch": 2.49, "learning_rate": 1.5048590281943613e-06, "loss": 0.0001, "step": 20820 }, { "epoch": 2.5, "learning_rate": 1.5047870425914816e-06, "loss": 0.0002, "step": 20821 }, { "epoch": 2.5, "learning_rate": 1.5047150569886024e-06, "loss": 0.0003, "step": 20822 }, { "epoch": 2.5, "learning_rate": 1.504643071385723e-06, "loss": 0.0002, "step": 20823 }, { "epoch": 2.5, "learning_rate": 1.5045710857828433e-06, "loss": 0.0002, "step": 20824 }, { "epoch": 2.5, "learning_rate": 1.504499100179964e-06, "loss": 0.0003, "step": 20825 }, { "epoch": 2.5, "learning_rate": 1.5044271145770844e-06, "loss": 0.0028, "step": 20826 }, { "epoch": 2.5, "learning_rate": 1.5043551289742052e-06, "loss": 0.0002, "step": 20827 }, { "epoch": 2.5, "learning_rate": 1.5042831433713256e-06, "loss": 0.0001, "step": 20828 }, { "epoch": 2.5, "learning_rate": 1.5042111577684464e-06, "loss": 0.0003, "step": 20829 }, { "epoch": 2.5, "learning_rate": 1.504139172165567e-06, "loss": 0.0001, "step": 20830 }, { "epoch": 2.5, "learning_rate": 1.5040671865626877e-06, "loss": 0.0002, "step": 20831 }, { "epoch": 2.5, "learning_rate": 1.503995200959808e-06, "loss": 0.0003, "step": 20832 }, { "epoch": 2.5, "learning_rate": 1.5039232153569288e-06, "loss": 0.0001, "step": 20833 }, { "epoch": 2.5, "learning_rate": 1.5038512297540492e-06, "loss": 0.0002, "step": 20834 }, { "epoch": 2.5, "learning_rate": 1.50377924415117e-06, "loss": 0.0002, "step": 20835 }, { "epoch": 2.5, "learning_rate": 1.5037072585482903e-06, "loss": 0.0001, "step": 20836 }, { "epoch": 2.5, "learning_rate": 1.5036352729454109e-06, "loss": 0.0001, "step": 20837 }, { "epoch": 2.5, "learning_rate": 1.5035632873425317e-06, "loss": 0.0001, "step": 20838 }, { "epoch": 2.5, "learning_rate": 1.503491301739652e-06, "loss": 0.0001, "step": 20839 }, { "epoch": 2.5, "learning_rate": 1.5034193161367728e-06, "loss": 0.0001, "step": 20840 }, { "epoch": 2.5, "learning_rate": 1.5033473305338932e-06, "loss": 0.0002, "step": 20841 }, { "epoch": 2.5, "learning_rate": 1.503275344931014e-06, "loss": 0.0004, "step": 20842 }, { "epoch": 2.5, "learning_rate": 1.5032033593281343e-06, "loss": 0.0003, "step": 20843 }, { "epoch": 2.5, "learning_rate": 1.503131373725255e-06, "loss": 0.0002, "step": 20844 }, { "epoch": 2.5, "learning_rate": 1.5030593881223756e-06, "loss": 0.0003, "step": 20845 }, { "epoch": 2.5, "learning_rate": 1.5029874025194962e-06, "loss": 0.0002, "step": 20846 }, { "epoch": 2.5, "learning_rate": 1.5029154169166168e-06, "loss": 0.0001, "step": 20847 }, { "epoch": 2.5, "learning_rate": 1.5028434313137371e-06, "loss": 0.0002, "step": 20848 }, { "epoch": 2.5, "learning_rate": 1.502771445710858e-06, "loss": 0.0002, "step": 20849 }, { "epoch": 2.5, "learning_rate": 1.5026994601079783e-06, "loss": 0.0001, "step": 20850 }, { "epoch": 2.5, "learning_rate": 1.502627474505099e-06, "loss": 0.0002, "step": 20851 }, { "epoch": 2.5, "learning_rate": 1.5025554889022196e-06, "loss": 0.0002, "step": 20852 }, { "epoch": 2.5, "learning_rate": 1.5024835032993402e-06, "loss": 0.6679, "step": 20853 }, { "epoch": 2.5, "learning_rate": 1.5024115176964607e-06, "loss": 0.0001, "step": 20854 }, { "epoch": 2.5, "learning_rate": 1.5023395320935815e-06, "loss": 0.0002, "step": 20855 }, { "epoch": 2.5, "learning_rate": 1.5022675464907019e-06, "loss": 0.0003, "step": 20856 }, { "epoch": 2.5, "learning_rate": 1.5021955608878226e-06, "loss": 0.0001, "step": 20857 }, { "epoch": 2.5, "learning_rate": 1.502123575284943e-06, "loss": 0.0001, "step": 20858 }, { "epoch": 2.5, "learning_rate": 1.5020515896820636e-06, "loss": 0.0002, "step": 20859 }, { "epoch": 2.5, "learning_rate": 1.5019796040791841e-06, "loss": 0.0001, "step": 20860 }, { "epoch": 2.5, "learning_rate": 1.5019076184763047e-06, "loss": 0.0003, "step": 20861 }, { "epoch": 2.5, "learning_rate": 1.5018356328734255e-06, "loss": 0.0011, "step": 20862 }, { "epoch": 2.5, "learning_rate": 1.5017636472705458e-06, "loss": 0.0624, "step": 20863 }, { "epoch": 2.5, "learning_rate": 1.5016916616676666e-06, "loss": 0.0001, "step": 20864 }, { "epoch": 2.5, "learning_rate": 1.501619676064787e-06, "loss": 0.0003, "step": 20865 }, { "epoch": 2.5, "learning_rate": 1.5015476904619077e-06, "loss": 0.0005, "step": 20866 }, { "epoch": 2.5, "learning_rate": 1.501475704859028e-06, "loss": 0.0001, "step": 20867 }, { "epoch": 2.5, "learning_rate": 1.5014037192561489e-06, "loss": 0.0001, "step": 20868 }, { "epoch": 2.5, "learning_rate": 1.5013317336532694e-06, "loss": 0.0002, "step": 20869 }, { "epoch": 2.5, "learning_rate": 1.5012597480503902e-06, "loss": 0.0006, "step": 20870 }, { "epoch": 2.5, "learning_rate": 1.5011877624475106e-06, "loss": 0.0001, "step": 20871 }, { "epoch": 2.5, "learning_rate": 1.501115776844631e-06, "loss": 0.0006, "step": 20872 }, { "epoch": 2.5, "learning_rate": 1.5010437912417517e-06, "loss": 0.0001, "step": 20873 }, { "epoch": 2.5, "learning_rate": 1.500971805638872e-06, "loss": 0.0001, "step": 20874 }, { "epoch": 2.5, "learning_rate": 1.5008998200359928e-06, "loss": 0.0003, "step": 20875 }, { "epoch": 2.5, "learning_rate": 1.5008278344331134e-06, "loss": 0.0001, "step": 20876 }, { "epoch": 2.5, "learning_rate": 1.5007558488302342e-06, "loss": 0.2206, "step": 20877 }, { "epoch": 2.5, "learning_rate": 1.5006838632273545e-06, "loss": 0.0002, "step": 20878 }, { "epoch": 2.5, "learning_rate": 1.5006118776244753e-06, "loss": 0.0002, "step": 20879 }, { "epoch": 2.5, "learning_rate": 1.5005398920215957e-06, "loss": 0.0002, "step": 20880 }, { "epoch": 2.5, "learning_rate": 1.5004679064187165e-06, "loss": 0.0005, "step": 20881 }, { "epoch": 2.5, "learning_rate": 1.5003959208158368e-06, "loss": 0.0001, "step": 20882 }, { "epoch": 2.5, "learning_rate": 1.5003239352129574e-06, "loss": 0.0001, "step": 20883 }, { "epoch": 2.5, "learning_rate": 1.5002519496100782e-06, "loss": 0.0001, "step": 20884 }, { "epoch": 2.5, "learning_rate": 1.5001799640071985e-06, "loss": 0.0001, "step": 20885 }, { "epoch": 2.5, "learning_rate": 1.5001079784043193e-06, "loss": 0.0001, "step": 20886 }, { "epoch": 2.5, "learning_rate": 1.5000359928014396e-06, "loss": 0.0001, "step": 20887 }, { "epoch": 2.5, "learning_rate": 1.4999640071985602e-06, "loss": 0.0002, "step": 20888 }, { "epoch": 2.5, "learning_rate": 1.4998920215956808e-06, "loss": 0.0001, "step": 20889 }, { "epoch": 2.5, "learning_rate": 1.4998200359928013e-06, "loss": 0.0002, "step": 20890 }, { "epoch": 2.5, "learning_rate": 1.499748050389922e-06, "loss": 0.0003, "step": 20891 }, { "epoch": 2.5, "learning_rate": 1.4996760647870427e-06, "loss": 0.68, "step": 20892 }, { "epoch": 2.5, "learning_rate": 1.4996040791841633e-06, "loss": 0.0004, "step": 20893 }, { "epoch": 2.5, "learning_rate": 1.4995320935812838e-06, "loss": 0.0001, "step": 20894 }, { "epoch": 2.5, "learning_rate": 1.4994601079784044e-06, "loss": 0.0001, "step": 20895 }, { "epoch": 2.5, "learning_rate": 1.499388122375525e-06, "loss": 0.0004, "step": 20896 }, { "epoch": 2.5, "learning_rate": 1.4993161367726455e-06, "loss": 0.0002, "step": 20897 }, { "epoch": 2.5, "learning_rate": 1.499244151169766e-06, "loss": 0.9992, "step": 20898 }, { "epoch": 2.5, "learning_rate": 1.4991721655668867e-06, "loss": 0.0002, "step": 20899 }, { "epoch": 2.5, "learning_rate": 1.4991001799640072e-06, "loss": 0.6333, "step": 20900 }, { "epoch": 2.5, "learning_rate": 1.4990281943611278e-06, "loss": 0.0001, "step": 20901 }, { "epoch": 2.5, "learning_rate": 1.4989562087582484e-06, "loss": 0.0001, "step": 20902 }, { "epoch": 2.5, "learning_rate": 1.498884223155369e-06, "loss": 0.0002, "step": 20903 }, { "epoch": 2.5, "learning_rate": 1.4988122375524895e-06, "loss": 0.0001, "step": 20904 }, { "epoch": 2.51, "learning_rate": 1.49874025194961e-06, "loss": 0.0002, "step": 20905 }, { "epoch": 2.51, "learning_rate": 1.4986682663467306e-06, "loss": 0.0002, "step": 20906 }, { "epoch": 2.51, "learning_rate": 1.4985962807438512e-06, "loss": 0.0003, "step": 20907 }, { "epoch": 2.51, "learning_rate": 1.498524295140972e-06, "loss": 0.0002, "step": 20908 }, { "epoch": 2.51, "learning_rate": 1.4984523095380925e-06, "loss": 0.0003, "step": 20909 }, { "epoch": 2.51, "learning_rate": 1.4983803239352131e-06, "loss": 0.8324, "step": 20910 }, { "epoch": 2.51, "learning_rate": 1.4983083383323335e-06, "loss": 0.0002, "step": 20911 }, { "epoch": 2.51, "learning_rate": 1.498236352729454e-06, "loss": 0.0001, "step": 20912 }, { "epoch": 2.51, "learning_rate": 1.4981643671265746e-06, "loss": 0.0002, "step": 20913 }, { "epoch": 2.51, "learning_rate": 1.4980923815236952e-06, "loss": 0.0001, "step": 20914 }, { "epoch": 2.51, "learning_rate": 1.498020395920816e-06, "loss": 0.0001, "step": 20915 }, { "epoch": 2.51, "learning_rate": 1.4979484103179365e-06, "loss": 0.0002, "step": 20916 }, { "epoch": 2.51, "learning_rate": 1.497876424715057e-06, "loss": 0.0001, "step": 20917 }, { "epoch": 2.51, "learning_rate": 1.4978044391121776e-06, "loss": 0.3169, "step": 20918 }, { "epoch": 2.51, "learning_rate": 1.4977324535092982e-06, "loss": 0.0002, "step": 20919 }, { "epoch": 2.51, "learning_rate": 1.4976604679064188e-06, "loss": 0.0608, "step": 20920 }, { "epoch": 2.51, "learning_rate": 1.4975884823035393e-06, "loss": 0.0006, "step": 20921 }, { "epoch": 2.51, "learning_rate": 1.49751649670066e-06, "loss": 0.0001, "step": 20922 }, { "epoch": 2.51, "learning_rate": 1.4974445110977805e-06, "loss": 0.0003, "step": 20923 }, { "epoch": 2.51, "learning_rate": 1.497372525494901e-06, "loss": 0.0003, "step": 20924 }, { "epoch": 2.51, "learning_rate": 1.4973005398920216e-06, "loss": 0.8146, "step": 20925 }, { "epoch": 2.51, "learning_rate": 1.4972285542891422e-06, "loss": 0.0002, "step": 20926 }, { "epoch": 2.51, "learning_rate": 1.4971565686862627e-06, "loss": 0.0001, "step": 20927 }, { "epoch": 2.51, "learning_rate": 1.4970845830833833e-06, "loss": 0.0002, "step": 20928 }, { "epoch": 2.51, "learning_rate": 1.4970125974805039e-06, "loss": 0.0001, "step": 20929 }, { "epoch": 2.51, "learning_rate": 1.4969406118776244e-06, "loss": 0.0001, "step": 20930 }, { "epoch": 2.51, "learning_rate": 1.4968686262747452e-06, "loss": 0.0017, "step": 20931 }, { "epoch": 2.51, "learning_rate": 1.4967966406718658e-06, "loss": 0.0004, "step": 20932 }, { "epoch": 2.51, "learning_rate": 1.4967246550689864e-06, "loss": 0.0002, "step": 20933 }, { "epoch": 2.51, "learning_rate": 1.4966526694661067e-06, "loss": 0.0002, "step": 20934 }, { "epoch": 2.51, "learning_rate": 1.4965806838632273e-06, "loss": 0.0002, "step": 20935 }, { "epoch": 2.51, "learning_rate": 1.4965086982603478e-06, "loss": 0.0002, "step": 20936 }, { "epoch": 2.51, "learning_rate": 1.4964367126574684e-06, "loss": 0.0002, "step": 20937 }, { "epoch": 2.51, "learning_rate": 1.4963647270545892e-06, "loss": 0.7403, "step": 20938 }, { "epoch": 2.51, "learning_rate": 1.4962927414517098e-06, "loss": 0.0002, "step": 20939 }, { "epoch": 2.51, "learning_rate": 1.4962207558488303e-06, "loss": 0.0002, "step": 20940 }, { "epoch": 2.51, "learning_rate": 1.4961487702459509e-06, "loss": 0.0002, "step": 20941 }, { "epoch": 2.51, "learning_rate": 1.4960767846430715e-06, "loss": 0.0584, "step": 20942 }, { "epoch": 2.51, "learning_rate": 1.496004799040192e-06, "loss": 0.0001, "step": 20943 }, { "epoch": 2.51, "learning_rate": 1.4959328134373126e-06, "loss": 0.0006, "step": 20944 }, { "epoch": 2.51, "learning_rate": 1.4958608278344332e-06, "loss": 0.0001, "step": 20945 }, { "epoch": 2.51, "learning_rate": 1.4957888422315537e-06, "loss": 0.0002, "step": 20946 }, { "epoch": 2.51, "learning_rate": 1.4957168566286743e-06, "loss": 0.0002, "step": 20947 }, { "epoch": 2.51, "learning_rate": 1.4956448710257949e-06, "loss": 0.0002, "step": 20948 }, { "epoch": 2.51, "learning_rate": 1.4955728854229154e-06, "loss": 0.0002, "step": 20949 }, { "epoch": 2.51, "learning_rate": 1.495500899820036e-06, "loss": 0.1945, "step": 20950 }, { "epoch": 2.51, "learning_rate": 1.4954289142171566e-06, "loss": 0.0002, "step": 20951 }, { "epoch": 2.51, "learning_rate": 1.4953569286142771e-06, "loss": 0.0001, "step": 20952 }, { "epoch": 2.51, "learning_rate": 1.4952849430113977e-06, "loss": 0.0004, "step": 20953 }, { "epoch": 2.51, "learning_rate": 1.4952129574085185e-06, "loss": 0.0002, "step": 20954 }, { "epoch": 2.51, "learning_rate": 1.495140971805639e-06, "loss": 0.0002, "step": 20955 }, { "epoch": 2.51, "learning_rate": 1.4950689862027596e-06, "loss": 0.0191, "step": 20956 }, { "epoch": 2.51, "learning_rate": 1.49499700059988e-06, "loss": 0.6572, "step": 20957 }, { "epoch": 2.51, "learning_rate": 1.4949250149970005e-06, "loss": 0.0004, "step": 20958 }, { "epoch": 2.51, "learning_rate": 1.494853029394121e-06, "loss": 0.0001, "step": 20959 }, { "epoch": 2.51, "learning_rate": 1.4947810437912417e-06, "loss": 0.0001, "step": 20960 }, { "epoch": 2.51, "learning_rate": 1.4947090581883624e-06, "loss": 0.0002, "step": 20961 }, { "epoch": 2.51, "learning_rate": 1.494637072585483e-06, "loss": 0.0002, "step": 20962 }, { "epoch": 2.51, "learning_rate": 1.4945650869826036e-06, "loss": 0.0002, "step": 20963 }, { "epoch": 2.51, "learning_rate": 1.4944931013797241e-06, "loss": 0.0001, "step": 20964 }, { "epoch": 2.51, "learning_rate": 1.4944211157768447e-06, "loss": 0.0001, "step": 20965 }, { "epoch": 2.51, "learning_rate": 1.4943491301739653e-06, "loss": 0.0001, "step": 20966 }, { "epoch": 2.51, "learning_rate": 1.4942771445710858e-06, "loss": 0.0001, "step": 20967 }, { "epoch": 2.51, "learning_rate": 1.4942051589682064e-06, "loss": 0.0003, "step": 20968 }, { "epoch": 2.51, "learning_rate": 1.494133173365327e-06, "loss": 0.0001, "step": 20969 }, { "epoch": 2.51, "learning_rate": 1.4940611877624475e-06, "loss": 0.0001, "step": 20970 }, { "epoch": 2.51, "learning_rate": 1.4939892021595681e-06, "loss": 0.0002, "step": 20971 }, { "epoch": 2.51, "learning_rate": 1.4939172165566887e-06, "loss": 0.0002, "step": 20972 }, { "epoch": 2.51, "learning_rate": 1.4938452309538092e-06, "loss": 0.0003, "step": 20973 }, { "epoch": 2.51, "learning_rate": 1.4937732453509298e-06, "loss": 0.0002, "step": 20974 }, { "epoch": 2.51, "learning_rate": 1.4937012597480504e-06, "loss": 0.0003, "step": 20975 }, { "epoch": 2.51, "learning_rate": 1.493629274145171e-06, "loss": 0.0004, "step": 20976 }, { "epoch": 2.51, "learning_rate": 1.4935572885422917e-06, "loss": 0.0002, "step": 20977 }, { "epoch": 2.51, "learning_rate": 1.4934853029394123e-06, "loss": 0.0001, "step": 20978 }, { "epoch": 2.51, "learning_rate": 1.4934133173365329e-06, "loss": 0.0002, "step": 20979 }, { "epoch": 2.51, "learning_rate": 1.4933413317336532e-06, "loss": 0.0003, "step": 20980 }, { "epoch": 2.51, "learning_rate": 1.4932693461307738e-06, "loss": 0.0001, "step": 20981 }, { "epoch": 2.51, "learning_rate": 1.4931973605278943e-06, "loss": 0.0007, "step": 20982 }, { "epoch": 2.51, "learning_rate": 1.493125374925015e-06, "loss": 0.0001, "step": 20983 }, { "epoch": 2.51, "learning_rate": 1.4930533893221357e-06, "loss": 0.0003, "step": 20984 }, { "epoch": 2.51, "learning_rate": 1.4929814037192563e-06, "loss": 0.0003, "step": 20985 }, { "epoch": 2.51, "learning_rate": 1.4929094181163768e-06, "loss": 0.0002, "step": 20986 }, { "epoch": 2.51, "learning_rate": 1.4928374325134974e-06, "loss": 0.0008, "step": 20987 }, { "epoch": 2.52, "learning_rate": 1.492765446910618e-06, "loss": 0.0002, "step": 20988 }, { "epoch": 2.52, "learning_rate": 1.4926934613077385e-06, "loss": 0.0655, "step": 20989 }, { "epoch": 2.52, "learning_rate": 1.492621475704859e-06, "loss": 0.0001, "step": 20990 }, { "epoch": 2.52, "learning_rate": 1.4925494901019797e-06, "loss": 0.0002, "step": 20991 }, { "epoch": 2.52, "learning_rate": 1.4924775044991002e-06, "loss": 0.0001, "step": 20992 }, { "epoch": 2.52, "learning_rate": 1.4924055188962208e-06, "loss": 0.0001, "step": 20993 }, { "epoch": 2.52, "learning_rate": 1.4923335332933414e-06, "loss": 0.0005, "step": 20994 }, { "epoch": 2.52, "learning_rate": 1.492261547690462e-06, "loss": 0.0002, "step": 20995 }, { "epoch": 2.52, "learning_rate": 1.4921895620875825e-06, "loss": 0.0001, "step": 20996 }, { "epoch": 2.52, "learning_rate": 1.492117576484703e-06, "loss": 0.0001, "step": 20997 }, { "epoch": 2.52, "learning_rate": 1.4920455908818236e-06, "loss": 0.0001, "step": 20998 }, { "epoch": 2.52, "learning_rate": 1.4919736052789442e-06, "loss": 0.0002, "step": 20999 }, { "epoch": 2.52, "learning_rate": 1.491901619676065e-06, "loss": 0.0002, "step": 21000 }, { "epoch": 2.52, "learning_rate": 1.4918296340731855e-06, "loss": 0.0002, "step": 21001 }, { "epoch": 2.52, "learning_rate": 1.491757648470306e-06, "loss": 0.0002, "step": 21002 }, { "epoch": 2.52, "learning_rate": 1.4916856628674265e-06, "loss": 0.0001, "step": 21003 }, { "epoch": 2.52, "learning_rate": 1.491613677264547e-06, "loss": 0.0001, "step": 21004 }, { "epoch": 2.52, "learning_rate": 1.4915416916616676e-06, "loss": 0.0001, "step": 21005 }, { "epoch": 2.52, "learning_rate": 1.4914697060587882e-06, "loss": 0.0003, "step": 21006 }, { "epoch": 2.52, "learning_rate": 1.491397720455909e-06, "loss": 0.0481, "step": 21007 }, { "epoch": 2.52, "learning_rate": 1.4913257348530295e-06, "loss": 0.0001, "step": 21008 }, { "epoch": 2.52, "learning_rate": 1.49125374925015e-06, "loss": 0.0001, "step": 21009 }, { "epoch": 2.52, "learning_rate": 1.4911817636472706e-06, "loss": 0.0002, "step": 21010 }, { "epoch": 2.52, "learning_rate": 1.4911097780443912e-06, "loss": 0.0002, "step": 21011 }, { "epoch": 2.52, "learning_rate": 1.4910377924415118e-06, "loss": 0.0003, "step": 21012 }, { "epoch": 2.52, "learning_rate": 1.4909658068386323e-06, "loss": 0.0004, "step": 21013 }, { "epoch": 2.52, "learning_rate": 1.490893821235753e-06, "loss": 0.0002, "step": 21014 }, { "epoch": 2.52, "learning_rate": 1.4908218356328735e-06, "loss": 0.0001, "step": 21015 }, { "epoch": 2.52, "learning_rate": 1.490749850029994e-06, "loss": 0.0077, "step": 21016 }, { "epoch": 2.52, "learning_rate": 1.4906778644271146e-06, "loss": 0.0001, "step": 21017 }, { "epoch": 2.52, "learning_rate": 1.4906058788242352e-06, "loss": 0.0002, "step": 21018 }, { "epoch": 2.52, "learning_rate": 1.4905338932213557e-06, "loss": 0.4084, "step": 21019 }, { "epoch": 2.52, "learning_rate": 1.4904619076184763e-06, "loss": 0.0001, "step": 21020 }, { "epoch": 2.52, "learning_rate": 1.4903899220155969e-06, "loss": 0.0002, "step": 21021 }, { "epoch": 2.52, "learning_rate": 1.4903179364127174e-06, "loss": 0.0002, "step": 21022 }, { "epoch": 2.52, "learning_rate": 1.4902459508098382e-06, "loss": 0.0002, "step": 21023 }, { "epoch": 2.52, "learning_rate": 1.4901739652069588e-06, "loss": 0.6855, "step": 21024 }, { "epoch": 2.52, "learning_rate": 1.4901019796040793e-06, "loss": 0.4934, "step": 21025 }, { "epoch": 2.52, "learning_rate": 1.4900299940012e-06, "loss": 0.0002, "step": 21026 }, { "epoch": 2.52, "learning_rate": 1.4899580083983203e-06, "loss": 0.0001, "step": 21027 }, { "epoch": 2.52, "learning_rate": 1.4898860227954408e-06, "loss": 0.0002, "step": 21028 }, { "epoch": 2.52, "learning_rate": 1.4898140371925614e-06, "loss": 0.7847, "step": 21029 }, { "epoch": 2.52, "learning_rate": 1.489742051589682e-06, "loss": 0.0004, "step": 21030 }, { "epoch": 2.52, "learning_rate": 1.4896700659868027e-06, "loss": 0.0001, "step": 21031 }, { "epoch": 2.52, "learning_rate": 1.4895980803839233e-06, "loss": 0.0001, "step": 21032 }, { "epoch": 2.52, "learning_rate": 1.4895260947810439e-06, "loss": 0.0047, "step": 21033 }, { "epoch": 2.52, "learning_rate": 1.4894541091781644e-06, "loss": 0.0001, "step": 21034 }, { "epoch": 2.52, "learning_rate": 1.489382123575285e-06, "loss": 0.0003, "step": 21035 }, { "epoch": 2.52, "learning_rate": 1.4893101379724056e-06, "loss": 0.0003, "step": 21036 }, { "epoch": 2.52, "learning_rate": 1.4892381523695261e-06, "loss": 0.0002, "step": 21037 }, { "epoch": 2.52, "learning_rate": 1.4891661667666467e-06, "loss": 0.0009, "step": 21038 }, { "epoch": 2.52, "learning_rate": 1.4890941811637673e-06, "loss": 0.0001, "step": 21039 }, { "epoch": 2.52, "learning_rate": 1.4890221955608879e-06, "loss": 0.0002, "step": 21040 }, { "epoch": 2.52, "learning_rate": 1.4889502099580084e-06, "loss": 0.0002, "step": 21041 }, { "epoch": 2.52, "learning_rate": 1.488878224355129e-06, "loss": 0.0002, "step": 21042 }, { "epoch": 2.52, "learning_rate": 1.4888062387522496e-06, "loss": 0.1025, "step": 21043 }, { "epoch": 2.52, "learning_rate": 1.4887342531493701e-06, "loss": 0.0002, "step": 21044 }, { "epoch": 2.52, "learning_rate": 1.4886622675464907e-06, "loss": 0.0001, "step": 21045 }, { "epoch": 2.52, "learning_rate": 1.4885902819436113e-06, "loss": 0.0006, "step": 21046 }, { "epoch": 2.52, "learning_rate": 1.488518296340732e-06, "loss": 0.4663, "step": 21047 }, { "epoch": 2.52, "learning_rate": 1.4884463107378526e-06, "loss": 0.0002, "step": 21048 }, { "epoch": 2.52, "learning_rate": 1.4883743251349732e-06, "loss": 0.0002, "step": 21049 }, { "epoch": 2.52, "learning_rate": 1.4883023395320935e-06, "loss": 0.0002, "step": 21050 }, { "epoch": 2.52, "learning_rate": 1.488230353929214e-06, "loss": 0.0001, "step": 21051 }, { "epoch": 2.52, "learning_rate": 1.4881583683263347e-06, "loss": 0.0002, "step": 21052 }, { "epoch": 2.52, "learning_rate": 1.4880863827234552e-06, "loss": 0.0001, "step": 21053 }, { "epoch": 2.52, "learning_rate": 1.488014397120576e-06, "loss": 0.0002, "step": 21054 }, { "epoch": 2.52, "learning_rate": 1.4879424115176966e-06, "loss": 0.0002, "step": 21055 }, { "epoch": 2.52, "learning_rate": 1.4878704259148171e-06, "loss": 0.0002, "step": 21056 }, { "epoch": 2.52, "learning_rate": 1.4877984403119377e-06, "loss": 0.0002, "step": 21057 }, { "epoch": 2.52, "learning_rate": 1.4877264547090583e-06, "loss": 0.0004, "step": 21058 }, { "epoch": 2.52, "learning_rate": 1.4876544691061788e-06, "loss": 0.0003, "step": 21059 }, { "epoch": 2.52, "learning_rate": 1.4875824835032994e-06, "loss": 0.0004, "step": 21060 }, { "epoch": 2.52, "learning_rate": 1.48751049790042e-06, "loss": 0.0012, "step": 21061 }, { "epoch": 2.52, "learning_rate": 1.4874385122975405e-06, "loss": 0.002, "step": 21062 }, { "epoch": 2.52, "learning_rate": 1.487366526694661e-06, "loss": 0.0001, "step": 21063 }, { "epoch": 2.52, "learning_rate": 1.4872945410917817e-06, "loss": 0.0002, "step": 21064 }, { "epoch": 2.52, "learning_rate": 1.4872225554889022e-06, "loss": 0.0001, "step": 21065 }, { "epoch": 2.52, "learning_rate": 1.4871505698860228e-06, "loss": 0.0001, "step": 21066 }, { "epoch": 2.52, "learning_rate": 1.4870785842831434e-06, "loss": 0.0002, "step": 21067 }, { "epoch": 2.52, "learning_rate": 1.487006598680264e-06, "loss": 0.0003, "step": 21068 }, { "epoch": 2.52, "learning_rate": 1.4869346130773845e-06, "loss": 0.0002, "step": 21069 }, { "epoch": 2.52, "learning_rate": 1.4868626274745053e-06, "loss": 0.0001, "step": 21070 }, { "epoch": 2.52, "learning_rate": 1.4867906418716258e-06, "loss": 0.0001, "step": 21071 }, { "epoch": 2.53, "learning_rate": 1.4867186562687464e-06, "loss": 0.0003, "step": 21072 }, { "epoch": 2.53, "learning_rate": 1.4866466706658668e-06, "loss": 0.0001, "step": 21073 }, { "epoch": 2.53, "learning_rate": 1.4865746850629873e-06, "loss": 0.0002, "step": 21074 }, { "epoch": 2.53, "learning_rate": 1.486502699460108e-06, "loss": 0.0001, "step": 21075 }, { "epoch": 2.53, "learning_rate": 1.4864307138572285e-06, "loss": 0.0001, "step": 21076 }, { "epoch": 2.53, "learning_rate": 1.4863587282543492e-06, "loss": 0.0002, "step": 21077 }, { "epoch": 2.53, "learning_rate": 1.4862867426514698e-06, "loss": 0.0001, "step": 21078 }, { "epoch": 2.53, "learning_rate": 1.4862147570485904e-06, "loss": 0.0001, "step": 21079 }, { "epoch": 2.53, "learning_rate": 1.486142771445711e-06, "loss": 0.0001, "step": 21080 }, { "epoch": 2.53, "learning_rate": 1.4860707858428315e-06, "loss": 0.0002, "step": 21081 }, { "epoch": 2.53, "learning_rate": 1.485998800239952e-06, "loss": 0.0001, "step": 21082 }, { "epoch": 2.53, "learning_rate": 1.4859268146370726e-06, "loss": 0.0002, "step": 21083 }, { "epoch": 2.53, "learning_rate": 1.4858548290341932e-06, "loss": 0.0002, "step": 21084 }, { "epoch": 2.53, "learning_rate": 1.4857828434313138e-06, "loss": 0.9816, "step": 21085 }, { "epoch": 2.53, "learning_rate": 1.4857108578284343e-06, "loss": 0.0004, "step": 21086 }, { "epoch": 2.53, "learning_rate": 1.485638872225555e-06, "loss": 0.0001, "step": 21087 }, { "epoch": 2.53, "learning_rate": 1.4855668866226755e-06, "loss": 0.0002, "step": 21088 }, { "epoch": 2.53, "learning_rate": 1.485494901019796e-06, "loss": 0.0006, "step": 21089 }, { "epoch": 2.53, "learning_rate": 1.4854229154169166e-06, "loss": 0.0001, "step": 21090 }, { "epoch": 2.53, "learning_rate": 1.4853509298140372e-06, "loss": 0.0006, "step": 21091 }, { "epoch": 2.53, "learning_rate": 1.4852789442111577e-06, "loss": 0.0007, "step": 21092 }, { "epoch": 2.53, "learning_rate": 1.4852069586082785e-06, "loss": 0.0002, "step": 21093 }, { "epoch": 2.53, "learning_rate": 1.485134973005399e-06, "loss": 0.0002, "step": 21094 }, { "epoch": 2.53, "learning_rate": 1.4850629874025197e-06, "loss": 0.0003, "step": 21095 }, { "epoch": 2.53, "learning_rate": 1.48499100179964e-06, "loss": 0.0001, "step": 21096 }, { "epoch": 2.53, "learning_rate": 1.4849190161967606e-06, "loss": 0.0002, "step": 21097 }, { "epoch": 2.53, "learning_rate": 1.4848470305938811e-06, "loss": 0.0005, "step": 21098 }, { "epoch": 2.53, "learning_rate": 1.4847750449910017e-06, "loss": 0.0005, "step": 21099 }, { "epoch": 2.53, "learning_rate": 1.4847030593881225e-06, "loss": 0.0002, "step": 21100 }, { "epoch": 2.53, "learning_rate": 1.484631073785243e-06, "loss": 0.0001, "step": 21101 }, { "epoch": 2.53, "learning_rate": 1.4845590881823636e-06, "loss": 0.0001, "step": 21102 }, { "epoch": 2.53, "learning_rate": 1.4844871025794842e-06, "loss": 0.0002, "step": 21103 }, { "epoch": 2.53, "learning_rate": 1.4844151169766048e-06, "loss": 0.0001, "step": 21104 }, { "epoch": 2.53, "learning_rate": 1.4843431313737253e-06, "loss": 0.0001, "step": 21105 }, { "epoch": 2.53, "learning_rate": 1.4842711457708459e-06, "loss": 0.0003, "step": 21106 }, { "epoch": 2.53, "learning_rate": 1.4841991601679665e-06, "loss": 0.0001, "step": 21107 }, { "epoch": 2.53, "learning_rate": 1.484127174565087e-06, "loss": 0.0008, "step": 21108 }, { "epoch": 2.53, "learning_rate": 1.4840551889622076e-06, "loss": 0.0001, "step": 21109 }, { "epoch": 2.53, "learning_rate": 1.4839832033593282e-06, "loss": 0.0003, "step": 21110 }, { "epoch": 2.53, "learning_rate": 1.4839112177564487e-06, "loss": 0.0005, "step": 21111 }, { "epoch": 2.53, "learning_rate": 1.4838392321535693e-06, "loss": 0.0001, "step": 21112 }, { "epoch": 2.53, "learning_rate": 1.4837672465506899e-06, "loss": 0.0002, "step": 21113 }, { "epoch": 2.53, "learning_rate": 1.4836952609478104e-06, "loss": 0.0002, "step": 21114 }, { "epoch": 2.53, "learning_rate": 1.483623275344931e-06, "loss": 0.0001, "step": 21115 }, { "epoch": 2.53, "learning_rate": 1.4835512897420518e-06, "loss": 0.0001, "step": 21116 }, { "epoch": 2.53, "learning_rate": 1.4834793041391723e-06, "loss": 0.0001, "step": 21117 }, { "epoch": 2.53, "learning_rate": 1.483407318536293e-06, "loss": 0.0002, "step": 21118 }, { "epoch": 2.53, "learning_rate": 1.4833353329334133e-06, "loss": 0.0001, "step": 21119 }, { "epoch": 2.53, "learning_rate": 1.4832633473305338e-06, "loss": 0.0001, "step": 21120 }, { "epoch": 2.53, "learning_rate": 1.4831913617276544e-06, "loss": 0.0003, "step": 21121 }, { "epoch": 2.53, "learning_rate": 1.483119376124775e-06, "loss": 0.0003, "step": 21122 }, { "epoch": 2.53, "learning_rate": 1.4830473905218957e-06, "loss": 0.0001, "step": 21123 }, { "epoch": 2.53, "learning_rate": 1.4829754049190163e-06, "loss": 0.0001, "step": 21124 }, { "epoch": 2.53, "learning_rate": 1.4829034193161369e-06, "loss": 0.0002, "step": 21125 }, { "epoch": 2.53, "learning_rate": 1.4828314337132574e-06, "loss": 0.0002, "step": 21126 }, { "epoch": 2.53, "learning_rate": 1.482759448110378e-06, "loss": 0.0002, "step": 21127 }, { "epoch": 2.53, "learning_rate": 1.4826874625074986e-06, "loss": 0.0002, "step": 21128 }, { "epoch": 2.53, "learning_rate": 1.4826154769046191e-06, "loss": 0.0002, "step": 21129 }, { "epoch": 2.53, "learning_rate": 1.4825434913017397e-06, "loss": 0.0002, "step": 21130 }, { "epoch": 2.53, "learning_rate": 1.4824715056988603e-06, "loss": 0.0001, "step": 21131 }, { "epoch": 2.53, "learning_rate": 1.4823995200959808e-06, "loss": 0.0001, "step": 21132 }, { "epoch": 2.53, "learning_rate": 1.4823275344931014e-06, "loss": 0.0001, "step": 21133 }, { "epoch": 2.53, "learning_rate": 1.482255548890222e-06, "loss": 0.0001, "step": 21134 }, { "epoch": 2.53, "learning_rate": 1.4821835632873425e-06, "loss": 0.0001, "step": 21135 }, { "epoch": 2.53, "learning_rate": 1.4821115776844631e-06, "loss": 0.0006, "step": 21136 }, { "epoch": 2.53, "learning_rate": 1.4820395920815837e-06, "loss": 0.0004, "step": 21137 }, { "epoch": 2.53, "learning_rate": 1.4819676064787042e-06, "loss": 0.0003, "step": 21138 }, { "epoch": 2.53, "learning_rate": 1.481895620875825e-06, "loss": 0.0002, "step": 21139 }, { "epoch": 2.53, "learning_rate": 1.4818236352729456e-06, "loss": 0.0001, "step": 21140 }, { "epoch": 2.53, "learning_rate": 1.4817516496700662e-06, "loss": 0.0002, "step": 21141 }, { "epoch": 2.53, "learning_rate": 1.4816796640671865e-06, "loss": 0.0001, "step": 21142 }, { "epoch": 2.53, "learning_rate": 1.481607678464307e-06, "loss": 0.0002, "step": 21143 }, { "epoch": 2.53, "learning_rate": 1.4815356928614276e-06, "loss": 0.0001, "step": 21144 }, { "epoch": 2.53, "learning_rate": 1.4814637072585482e-06, "loss": 0.0003, "step": 21145 }, { "epoch": 2.53, "learning_rate": 1.481391721655669e-06, "loss": 0.0002, "step": 21146 }, { "epoch": 2.53, "learning_rate": 1.4813197360527896e-06, "loss": 0.0001, "step": 21147 }, { "epoch": 2.53, "learning_rate": 1.4812477504499101e-06, "loss": 0.0001, "step": 21148 }, { "epoch": 2.53, "learning_rate": 1.4811757648470307e-06, "loss": 0.0001, "step": 21149 }, { "epoch": 2.53, "learning_rate": 1.4811037792441513e-06, "loss": 0.0002, "step": 21150 }, { "epoch": 2.53, "learning_rate": 1.4810317936412718e-06, "loss": 0.5256, "step": 21151 }, { "epoch": 2.53, "learning_rate": 1.4809598080383924e-06, "loss": 0.0001, "step": 21152 }, { "epoch": 2.53, "learning_rate": 1.480887822435513e-06, "loss": 0.0001, "step": 21153 }, { "epoch": 2.53, "learning_rate": 1.4808158368326335e-06, "loss": 0.0001, "step": 21154 }, { "epoch": 2.54, "learning_rate": 1.480743851229754e-06, "loss": 0.0002, "step": 21155 }, { "epoch": 2.54, "learning_rate": 1.4806718656268747e-06, "loss": 0.0001, "step": 21156 }, { "epoch": 2.54, "learning_rate": 1.4805998800239952e-06, "loss": 0.0006, "step": 21157 }, { "epoch": 2.54, "learning_rate": 1.4805278944211158e-06, "loss": 0.0003, "step": 21158 }, { "epoch": 2.54, "learning_rate": 1.4804559088182364e-06, "loss": 0.0001, "step": 21159 }, { "epoch": 2.54, "learning_rate": 1.480383923215357e-06, "loss": 1.3104, "step": 21160 }, { "epoch": 2.54, "learning_rate": 1.4803119376124775e-06, "loss": 0.0004, "step": 21161 }, { "epoch": 2.54, "learning_rate": 1.480239952009598e-06, "loss": 0.0002, "step": 21162 }, { "epoch": 2.54, "learning_rate": 1.4801679664067188e-06, "loss": 0.0002, "step": 21163 }, { "epoch": 2.54, "learning_rate": 1.4800959808038394e-06, "loss": 0.0002, "step": 21164 }, { "epoch": 2.54, "learning_rate": 1.4800239952009598e-06, "loss": 0.0002, "step": 21165 }, { "epoch": 2.54, "learning_rate": 1.4799520095980803e-06, "loss": 0.0001, "step": 21166 }, { "epoch": 2.54, "learning_rate": 1.4798800239952009e-06, "loss": 0.0003, "step": 21167 }, { "epoch": 2.54, "learning_rate": 1.4798080383923215e-06, "loss": 0.0002, "step": 21168 }, { "epoch": 2.54, "learning_rate": 1.479736052789442e-06, "loss": 0.0002, "step": 21169 }, { "epoch": 2.54, "learning_rate": 1.4796640671865628e-06, "loss": 0.0002, "step": 21170 }, { "epoch": 2.54, "learning_rate": 1.4795920815836834e-06, "loss": 0.0007, "step": 21171 }, { "epoch": 2.54, "learning_rate": 1.479520095980804e-06, "loss": 0.0001, "step": 21172 }, { "epoch": 2.54, "learning_rate": 1.4794481103779245e-06, "loss": 0.0002, "step": 21173 }, { "epoch": 2.54, "learning_rate": 1.479376124775045e-06, "loss": 0.0002, "step": 21174 }, { "epoch": 2.54, "learning_rate": 1.4793041391721656e-06, "loss": 0.0001, "step": 21175 }, { "epoch": 2.54, "learning_rate": 1.4792321535692862e-06, "loss": 0.0002, "step": 21176 }, { "epoch": 2.54, "learning_rate": 1.4791601679664068e-06, "loss": 0.0001, "step": 21177 }, { "epoch": 2.54, "learning_rate": 1.4790881823635273e-06, "loss": 0.0004, "step": 21178 }, { "epoch": 2.54, "learning_rate": 1.479016196760648e-06, "loss": 0.0002, "step": 21179 }, { "epoch": 2.54, "learning_rate": 1.4789442111577685e-06, "loss": 0.2711, "step": 21180 }, { "epoch": 2.54, "learning_rate": 1.478872225554889e-06, "loss": 0.0004, "step": 21181 }, { "epoch": 2.54, "learning_rate": 1.4788002399520096e-06, "loss": 0.0001, "step": 21182 }, { "epoch": 2.54, "learning_rate": 1.4787282543491302e-06, "loss": 0.0001, "step": 21183 }, { "epoch": 2.54, "learning_rate": 1.4786562687462507e-06, "loss": 0.0005, "step": 21184 }, { "epoch": 2.54, "learning_rate": 1.4785842831433713e-06, "loss": 0.0001, "step": 21185 }, { "epoch": 2.54, "learning_rate": 1.478512297540492e-06, "loss": 0.0001, "step": 21186 }, { "epoch": 2.54, "learning_rate": 1.4784403119376127e-06, "loss": 0.0001, "step": 21187 }, { "epoch": 2.54, "learning_rate": 1.478368326334733e-06, "loss": 0.0002, "step": 21188 }, { "epoch": 2.54, "learning_rate": 1.4782963407318536e-06, "loss": 0.0003, "step": 21189 }, { "epoch": 2.54, "learning_rate": 1.4782243551289741e-06, "loss": 0.0001, "step": 21190 }, { "epoch": 2.54, "learning_rate": 1.4781523695260947e-06, "loss": 0.0001, "step": 21191 }, { "epoch": 2.54, "learning_rate": 1.4780803839232153e-06, "loss": 0.0002, "step": 21192 }, { "epoch": 2.54, "learning_rate": 1.478008398320336e-06, "loss": 0.0001, "step": 21193 }, { "epoch": 2.54, "learning_rate": 1.4779364127174566e-06, "loss": 0.0004, "step": 21194 }, { "epoch": 2.54, "learning_rate": 1.4778644271145772e-06, "loss": 0.0003, "step": 21195 }, { "epoch": 2.54, "learning_rate": 1.4777924415116978e-06, "loss": 0.0004, "step": 21196 }, { "epoch": 2.54, "learning_rate": 1.4777204559088183e-06, "loss": 0.0004, "step": 21197 }, { "epoch": 2.54, "learning_rate": 1.4776484703059389e-06, "loss": 0.0002, "step": 21198 }, { "epoch": 2.54, "learning_rate": 1.4775764847030595e-06, "loss": 0.0003, "step": 21199 }, { "epoch": 2.54, "learning_rate": 1.47750449910018e-06, "loss": 0.0001, "step": 21200 }, { "epoch": 2.54, "learning_rate": 1.4774325134973006e-06, "loss": 0.0003, "step": 21201 }, { "epoch": 2.54, "learning_rate": 1.4773605278944212e-06, "loss": 0.0003, "step": 21202 }, { "epoch": 2.54, "learning_rate": 1.4772885422915417e-06, "loss": 0.0001, "step": 21203 }, { "epoch": 2.54, "learning_rate": 1.4772165566886623e-06, "loss": 0.0002, "step": 21204 }, { "epoch": 2.54, "learning_rate": 1.4771445710857829e-06, "loss": 0.0003, "step": 21205 }, { "epoch": 2.54, "learning_rate": 1.4770725854829034e-06, "loss": 0.0002, "step": 21206 }, { "epoch": 2.54, "learning_rate": 1.477000599880024e-06, "loss": 0.0001, "step": 21207 }, { "epoch": 2.54, "learning_rate": 1.4769286142771446e-06, "loss": 0.0001, "step": 21208 }, { "epoch": 2.54, "learning_rate": 1.4768566286742653e-06, "loss": 0.0002, "step": 21209 }, { "epoch": 2.54, "learning_rate": 1.476784643071386e-06, "loss": 0.0001, "step": 21210 }, { "epoch": 2.54, "learning_rate": 1.4767126574685063e-06, "loss": 0.0004, "step": 21211 }, { "epoch": 2.54, "learning_rate": 1.4766406718656268e-06, "loss": 0.0002, "step": 21212 }, { "epoch": 2.54, "learning_rate": 1.4765686862627474e-06, "loss": 0.0001, "step": 21213 }, { "epoch": 2.54, "learning_rate": 1.476496700659868e-06, "loss": 0.4326, "step": 21214 }, { "epoch": 2.54, "learning_rate": 1.4764247150569885e-06, "loss": 0.0001, "step": 21215 }, { "epoch": 2.54, "learning_rate": 1.4763527294541093e-06, "loss": 0.0003, "step": 21216 }, { "epoch": 2.54, "learning_rate": 1.4762807438512299e-06, "loss": 0.0001, "step": 21217 }, { "epoch": 2.54, "learning_rate": 1.4762087582483504e-06, "loss": 0.006, "step": 21218 }, { "epoch": 2.54, "learning_rate": 1.476136772645471e-06, "loss": 0.0002, "step": 21219 }, { "epoch": 2.54, "learning_rate": 1.4760647870425916e-06, "loss": 0.0001, "step": 21220 }, { "epoch": 2.54, "learning_rate": 1.4759928014397121e-06, "loss": 0.0002, "step": 21221 }, { "epoch": 2.54, "learning_rate": 1.4759208158368327e-06, "loss": 0.5935, "step": 21222 }, { "epoch": 2.54, "learning_rate": 1.4758488302339533e-06, "loss": 0.0001, "step": 21223 }, { "epoch": 2.54, "learning_rate": 1.4757768446310738e-06, "loss": 0.0001, "step": 21224 }, { "epoch": 2.54, "learning_rate": 1.4757048590281944e-06, "loss": 0.0002, "step": 21225 }, { "epoch": 2.54, "learning_rate": 1.475632873425315e-06, "loss": 0.0068, "step": 21226 }, { "epoch": 2.54, "learning_rate": 1.4755608878224355e-06, "loss": 0.0069, "step": 21227 }, { "epoch": 2.54, "learning_rate": 1.475488902219556e-06, "loss": 0.0002, "step": 21228 }, { "epoch": 2.54, "learning_rate": 1.4754169166166767e-06, "loss": 0.0002, "step": 21229 }, { "epoch": 2.54, "learning_rate": 1.4753449310137972e-06, "loss": 0.0002, "step": 21230 }, { "epoch": 2.54, "learning_rate": 1.4752729454109178e-06, "loss": 0.0001, "step": 21231 }, { "epoch": 2.54, "learning_rate": 1.4752009598080386e-06, "loss": 0.0003, "step": 21232 }, { "epoch": 2.54, "learning_rate": 1.4751289742051591e-06, "loss": 0.0001, "step": 21233 }, { "epoch": 2.54, "learning_rate": 1.4750569886022795e-06, "loss": 0.0002, "step": 21234 }, { "epoch": 2.54, "learning_rate": 1.4749850029994e-06, "loss": 0.0002, "step": 21235 }, { "epoch": 2.54, "learning_rate": 1.4749130173965206e-06, "loss": 0.0002, "step": 21236 }, { "epoch": 2.54, "learning_rate": 1.4748410317936412e-06, "loss": 0.0001, "step": 21237 }, { "epoch": 2.54, "learning_rate": 1.4747690461907618e-06, "loss": 0.0001, "step": 21238 }, { "epoch": 2.55, "learning_rate": 1.4746970605878825e-06, "loss": 0.0001, "step": 21239 }, { "epoch": 2.55, "learning_rate": 1.4746250749850031e-06, "loss": 0.0044, "step": 21240 }, { "epoch": 2.55, "learning_rate": 1.4745530893821237e-06, "loss": 0.0001, "step": 21241 }, { "epoch": 2.55, "learning_rate": 1.4744811037792442e-06, "loss": 0.0004, "step": 21242 }, { "epoch": 2.55, "learning_rate": 1.4744091181763648e-06, "loss": 0.0001, "step": 21243 }, { "epoch": 2.55, "learning_rate": 1.4743371325734854e-06, "loss": 0.0002, "step": 21244 }, { "epoch": 2.55, "learning_rate": 1.474265146970606e-06, "loss": 0.0001, "step": 21245 }, { "epoch": 2.55, "learning_rate": 1.4741931613677265e-06, "loss": 0.0002, "step": 21246 }, { "epoch": 2.55, "learning_rate": 1.474121175764847e-06, "loss": 0.0001, "step": 21247 }, { "epoch": 2.55, "learning_rate": 1.4740491901619676e-06, "loss": 0.0002, "step": 21248 }, { "epoch": 2.55, "learning_rate": 1.4739772045590882e-06, "loss": 0.0001, "step": 21249 }, { "epoch": 2.55, "learning_rate": 1.4739052189562088e-06, "loss": 0.0001, "step": 21250 }, { "epoch": 2.55, "learning_rate": 1.4738332333533293e-06, "loss": 0.0002, "step": 21251 }, { "epoch": 2.55, "learning_rate": 1.47376124775045e-06, "loss": 0.0001, "step": 21252 }, { "epoch": 2.55, "learning_rate": 1.4736892621475705e-06, "loss": 0.0001, "step": 21253 }, { "epoch": 2.55, "learning_rate": 1.473617276544691e-06, "loss": 0.0001, "step": 21254 }, { "epoch": 2.55, "learning_rate": 1.4735452909418118e-06, "loss": 0.0001, "step": 21255 }, { "epoch": 2.55, "learning_rate": 1.4734733053389324e-06, "loss": 0.0001, "step": 21256 }, { "epoch": 2.55, "learning_rate": 1.473401319736053e-06, "loss": 0.0001, "step": 21257 }, { "epoch": 2.55, "learning_rate": 1.4733293341331733e-06, "loss": 0.0001, "step": 21258 }, { "epoch": 2.55, "learning_rate": 1.4732573485302939e-06, "loss": 0.0002, "step": 21259 }, { "epoch": 2.55, "learning_rate": 1.4731853629274145e-06, "loss": 0.0001, "step": 21260 }, { "epoch": 2.55, "learning_rate": 1.473113377324535e-06, "loss": 0.0001, "step": 21261 }, { "epoch": 2.55, "learning_rate": 1.4730413917216558e-06, "loss": 0.0001, "step": 21262 }, { "epoch": 2.55, "learning_rate": 1.4729694061187764e-06, "loss": 0.0002, "step": 21263 }, { "epoch": 2.55, "learning_rate": 1.472897420515897e-06, "loss": 0.0002, "step": 21264 }, { "epoch": 2.55, "learning_rate": 1.4728254349130175e-06, "loss": 0.0002, "step": 21265 }, { "epoch": 2.55, "learning_rate": 1.472753449310138e-06, "loss": 0.0002, "step": 21266 }, { "epoch": 2.55, "learning_rate": 1.4726814637072586e-06, "loss": 0.0002, "step": 21267 }, { "epoch": 2.55, "learning_rate": 1.4726094781043792e-06, "loss": 0.0002, "step": 21268 }, { "epoch": 2.55, "learning_rate": 1.4725374925014998e-06, "loss": 0.001, "step": 21269 }, { "epoch": 2.55, "learning_rate": 1.4724655068986203e-06, "loss": 0.0002, "step": 21270 }, { "epoch": 2.55, "learning_rate": 1.472393521295741e-06, "loss": 1.0881, "step": 21271 }, { "epoch": 2.55, "learning_rate": 1.4723215356928615e-06, "loss": 0.0001, "step": 21272 }, { "epoch": 2.55, "learning_rate": 1.472249550089982e-06, "loss": 0.0005, "step": 21273 }, { "epoch": 2.55, "learning_rate": 1.4721775644871026e-06, "loss": 0.0001, "step": 21274 }, { "epoch": 2.55, "learning_rate": 1.4721055788842232e-06, "loss": 0.0001, "step": 21275 }, { "epoch": 2.55, "learning_rate": 1.4720335932813437e-06, "loss": 0.0005, "step": 21276 }, { "epoch": 2.55, "learning_rate": 1.4719616076784643e-06, "loss": 0.0001, "step": 21277 }, { "epoch": 2.55, "learning_rate": 1.471889622075585e-06, "loss": 0.0032, "step": 21278 }, { "epoch": 2.55, "learning_rate": 1.4718176364727056e-06, "loss": 0.0004, "step": 21279 }, { "epoch": 2.55, "learning_rate": 1.4717456508698262e-06, "loss": 0.0001, "step": 21280 }, { "epoch": 2.55, "learning_rate": 1.4716736652669466e-06, "loss": 0.6856, "step": 21281 }, { "epoch": 2.55, "learning_rate": 1.4716016796640671e-06, "loss": 0.0001, "step": 21282 }, { "epoch": 2.55, "learning_rate": 1.4715296940611877e-06, "loss": 0.0001, "step": 21283 }, { "epoch": 2.55, "learning_rate": 1.4714577084583083e-06, "loss": 0.0003, "step": 21284 }, { "epoch": 2.55, "learning_rate": 1.471385722855429e-06, "loss": 0.0002, "step": 21285 }, { "epoch": 2.55, "learning_rate": 1.4713137372525496e-06, "loss": 0.0001, "step": 21286 }, { "epoch": 2.55, "learning_rate": 1.4712417516496702e-06, "loss": 0.0002, "step": 21287 }, { "epoch": 2.55, "learning_rate": 1.4711697660467907e-06, "loss": 0.0001, "step": 21288 }, { "epoch": 2.55, "learning_rate": 1.4710977804439113e-06, "loss": 0.0001, "step": 21289 }, { "epoch": 2.55, "learning_rate": 1.4710257948410319e-06, "loss": 0.0004, "step": 21290 }, { "epoch": 2.55, "learning_rate": 1.4709538092381524e-06, "loss": 0.0003, "step": 21291 }, { "epoch": 2.55, "learning_rate": 1.4708818236352728e-06, "loss": 0.0001, "step": 21292 }, { "epoch": 2.55, "learning_rate": 1.4708098380323936e-06, "loss": 0.0001, "step": 21293 }, { "epoch": 2.55, "learning_rate": 1.4707378524295141e-06, "loss": 0.0001, "step": 21294 }, { "epoch": 2.55, "learning_rate": 1.4706658668266347e-06, "loss": 0.0001, "step": 21295 }, { "epoch": 2.55, "learning_rate": 1.4705938812237553e-06, "loss": 0.0002, "step": 21296 }, { "epoch": 2.55, "learning_rate": 1.4705218956208758e-06, "loss": 0.0002, "step": 21297 }, { "epoch": 2.55, "learning_rate": 1.4704499100179964e-06, "loss": 0.0002, "step": 21298 }, { "epoch": 2.55, "learning_rate": 1.470377924415117e-06, "loss": 0.0001, "step": 21299 }, { "epoch": 2.55, "learning_rate": 1.4703059388122375e-06, "loss": 0.0002, "step": 21300 }, { "epoch": 2.55, "learning_rate": 1.4702339532093581e-06, "loss": 0.0001, "step": 21301 }, { "epoch": 2.55, "learning_rate": 1.4701619676064789e-06, "loss": 0.0001, "step": 21302 }, { "epoch": 2.55, "learning_rate": 1.4700899820035995e-06, "loss": 0.0001, "step": 21303 }, { "epoch": 2.55, "learning_rate": 1.4700179964007198e-06, "loss": 0.0001, "step": 21304 }, { "epoch": 2.55, "learning_rate": 1.4699460107978404e-06, "loss": 0.0003, "step": 21305 }, { "epoch": 2.55, "learning_rate": 1.469874025194961e-06, "loss": 0.0001, "step": 21306 }, { "epoch": 2.55, "learning_rate": 1.4698020395920815e-06, "loss": 0.6781, "step": 21307 }, { "epoch": 2.55, "learning_rate": 1.469730053989202e-06, "loss": 0.0015, "step": 21308 }, { "epoch": 2.55, "learning_rate": 1.4696580683863229e-06, "loss": 0.0002, "step": 21309 }, { "epoch": 2.55, "learning_rate": 1.4695860827834434e-06, "loss": 0.0002, "step": 21310 }, { "epoch": 2.55, "learning_rate": 1.469514097180564e-06, "loss": 0.0001, "step": 21311 }, { "epoch": 2.55, "learning_rate": 1.4694421115776846e-06, "loss": 0.0001, "step": 21312 }, { "epoch": 2.55, "learning_rate": 1.4693701259748051e-06, "loss": 0.0002, "step": 21313 }, { "epoch": 2.55, "learning_rate": 1.4692981403719257e-06, "loss": 0.0002, "step": 21314 }, { "epoch": 2.55, "learning_rate": 1.469226154769046e-06, "loss": 0.0001, "step": 21315 }, { "epoch": 2.55, "learning_rate": 1.4691541691661668e-06, "loss": 0.0004, "step": 21316 }, { "epoch": 2.55, "learning_rate": 1.4690821835632874e-06, "loss": 0.0001, "step": 21317 }, { "epoch": 2.55, "learning_rate": 1.469010197960408e-06, "loss": 0.0002, "step": 21318 }, { "epoch": 2.55, "learning_rate": 1.4689382123575285e-06, "loss": 0.0002, "step": 21319 }, { "epoch": 2.55, "learning_rate": 1.468866226754649e-06, "loss": 0.0002, "step": 21320 }, { "epoch": 2.55, "learning_rate": 1.4687942411517697e-06, "loss": 0.0003, "step": 21321 }, { "epoch": 2.56, "learning_rate": 1.4687222555488902e-06, "loss": 0.0002, "step": 21322 }, { "epoch": 2.56, "learning_rate": 1.4686502699460108e-06, "loss": 0.0002, "step": 21323 }, { "epoch": 2.56, "learning_rate": 1.4685782843431314e-06, "loss": 0.0006, "step": 21324 }, { "epoch": 2.56, "learning_rate": 1.4685062987402521e-06, "loss": 0.0002, "step": 21325 }, { "epoch": 2.56, "learning_rate": 1.4684343131373727e-06, "loss": 0.0007, "step": 21326 }, { "epoch": 2.56, "learning_rate": 1.468362327534493e-06, "loss": 0.0003, "step": 21327 }, { "epoch": 2.56, "learning_rate": 1.4682903419316136e-06, "loss": 0.0001, "step": 21328 }, { "epoch": 2.56, "learning_rate": 1.4682183563287342e-06, "loss": 0.0001, "step": 21329 }, { "epoch": 2.56, "learning_rate": 1.4681463707258548e-06, "loss": 0.0002, "step": 21330 }, { "epoch": 2.56, "learning_rate": 1.4680743851229753e-06, "loss": 0.0002, "step": 21331 }, { "epoch": 2.56, "learning_rate": 1.4680023995200961e-06, "loss": 0.0002, "step": 21332 }, { "epoch": 2.56, "learning_rate": 1.4679304139172167e-06, "loss": 0.0002, "step": 21333 }, { "epoch": 2.56, "learning_rate": 1.4678584283143372e-06, "loss": 0.0001, "step": 21334 }, { "epoch": 2.56, "learning_rate": 1.4677864427114578e-06, "loss": 0.0002, "step": 21335 }, { "epoch": 2.56, "learning_rate": 1.4677144571085784e-06, "loss": 0.0001, "step": 21336 }, { "epoch": 2.56, "learning_rate": 1.467642471505699e-06, "loss": 0.0003, "step": 21337 }, { "epoch": 2.56, "learning_rate": 1.4675704859028193e-06, "loss": 0.0001, "step": 21338 }, { "epoch": 2.56, "learning_rate": 1.46749850029994e-06, "loss": 0.0002, "step": 21339 }, { "epoch": 2.56, "learning_rate": 1.4674265146970606e-06, "loss": 0.0001, "step": 21340 }, { "epoch": 2.56, "learning_rate": 1.4673545290941812e-06, "loss": 0.0005, "step": 21341 }, { "epoch": 2.56, "learning_rate": 1.4672825434913018e-06, "loss": 0.0002, "step": 21342 }, { "epoch": 2.56, "learning_rate": 1.4672105578884223e-06, "loss": 0.0004, "step": 21343 }, { "epoch": 2.56, "learning_rate": 1.467138572285543e-06, "loss": 0.0001, "step": 21344 }, { "epoch": 2.56, "learning_rate": 1.4670665866826635e-06, "loss": 0.0001, "step": 21345 }, { "epoch": 2.56, "learning_rate": 1.466994601079784e-06, "loss": 0.0001, "step": 21346 }, { "epoch": 2.56, "learning_rate": 1.4669226154769046e-06, "loss": 0.0, "step": 21347 }, { "epoch": 2.56, "learning_rate": 1.4668506298740254e-06, "loss": 0.0003, "step": 21348 }, { "epoch": 2.56, "learning_rate": 1.466778644271146e-06, "loss": 0.0001, "step": 21349 }, { "epoch": 2.56, "learning_rate": 1.4667066586682663e-06, "loss": 0.0003, "step": 21350 }, { "epoch": 2.56, "learning_rate": 1.4666346730653869e-06, "loss": 0.0003, "step": 21351 }, { "epoch": 2.56, "learning_rate": 1.4665626874625074e-06, "loss": 0.0001, "step": 21352 }, { "epoch": 2.56, "learning_rate": 1.466490701859628e-06, "loss": 0.0002, "step": 21353 }, { "epoch": 2.56, "learning_rate": 1.4664187162567486e-06, "loss": 0.001, "step": 21354 }, { "epoch": 2.56, "learning_rate": 1.4663467306538694e-06, "loss": 0.0002, "step": 21355 }, { "epoch": 2.56, "learning_rate": 1.46627474505099e-06, "loss": 0.002, "step": 21356 }, { "epoch": 2.56, "learning_rate": 1.4662027594481105e-06, "loss": 0.0002, "step": 21357 }, { "epoch": 2.56, "learning_rate": 1.466130773845231e-06, "loss": 0.0002, "step": 21358 }, { "epoch": 2.56, "learning_rate": 1.4660587882423516e-06, "loss": 0.0001, "step": 21359 }, { "epoch": 2.56, "learning_rate": 1.4659868026394722e-06, "loss": 0.0002, "step": 21360 }, { "epoch": 2.56, "learning_rate": 1.4659148170365925e-06, "loss": 0.0003, "step": 21361 }, { "epoch": 2.56, "learning_rate": 1.4658428314337133e-06, "loss": 0.0003, "step": 21362 }, { "epoch": 2.56, "learning_rate": 1.4657708458308339e-06, "loss": 0.0001, "step": 21363 }, { "epoch": 2.56, "learning_rate": 1.4656988602279545e-06, "loss": 0.0004, "step": 21364 }, { "epoch": 2.56, "learning_rate": 1.465626874625075e-06, "loss": 0.0001, "step": 21365 }, { "epoch": 2.56, "learning_rate": 1.4655548890221956e-06, "loss": 0.0001, "step": 21366 }, { "epoch": 2.56, "learning_rate": 1.4654829034193162e-06, "loss": 0.0003, "step": 21367 }, { "epoch": 2.56, "learning_rate": 1.4654109178164367e-06, "loss": 0.0001, "step": 21368 }, { "epoch": 2.56, "learning_rate": 1.4653389322135573e-06, "loss": 0.0059, "step": 21369 }, { "epoch": 2.56, "learning_rate": 1.4652669466106779e-06, "loss": 1.0318, "step": 21370 }, { "epoch": 2.56, "learning_rate": 1.4651949610077986e-06, "loss": 0.0001, "step": 21371 }, { "epoch": 2.56, "learning_rate": 1.4651229754049192e-06, "loss": 0.0001, "step": 21372 }, { "epoch": 2.56, "learning_rate": 1.4650509898020396e-06, "loss": 0.0001, "step": 21373 }, { "epoch": 2.56, "learning_rate": 1.4649790041991601e-06, "loss": 0.0003, "step": 21374 }, { "epoch": 2.56, "learning_rate": 1.4649070185962807e-06, "loss": 0.0002, "step": 21375 }, { "epoch": 2.56, "learning_rate": 1.4648350329934013e-06, "loss": 0.0031, "step": 21376 }, { "epoch": 2.56, "learning_rate": 1.4647630473905218e-06, "loss": 0.0002, "step": 21377 }, { "epoch": 2.56, "learning_rate": 1.4646910617876426e-06, "loss": 0.0007, "step": 21378 }, { "epoch": 2.56, "learning_rate": 1.4646190761847632e-06, "loss": 0.0006, "step": 21379 }, { "epoch": 2.56, "learning_rate": 1.4645470905818837e-06, "loss": 0.0002, "step": 21380 }, { "epoch": 2.56, "learning_rate": 1.4644751049790043e-06, "loss": 0.0001, "step": 21381 }, { "epoch": 2.56, "learning_rate": 1.4644031193761249e-06, "loss": 0.0003, "step": 21382 }, { "epoch": 2.56, "learning_rate": 1.4643311337732454e-06, "loss": 0.0002, "step": 21383 }, { "epoch": 2.56, "learning_rate": 1.464259148170366e-06, "loss": 0.0008, "step": 21384 }, { "epoch": 2.56, "learning_rate": 1.4641871625674866e-06, "loss": 0.0001, "step": 21385 }, { "epoch": 2.56, "learning_rate": 1.4641151769646071e-06, "loss": 0.0001, "step": 21386 }, { "epoch": 2.56, "learning_rate": 1.4640431913617277e-06, "loss": 0.0001, "step": 21387 }, { "epoch": 2.56, "learning_rate": 1.4639712057588483e-06, "loss": 0.0001, "step": 21388 }, { "epoch": 2.56, "learning_rate": 1.4638992201559688e-06, "loss": 0.0002, "step": 21389 }, { "epoch": 2.56, "learning_rate": 1.4638272345530894e-06, "loss": 0.0001, "step": 21390 }, { "epoch": 2.56, "learning_rate": 1.46375524895021e-06, "loss": 0.0012, "step": 21391 }, { "epoch": 2.56, "learning_rate": 1.4636832633473305e-06, "loss": 0.0002, "step": 21392 }, { "epoch": 2.56, "learning_rate": 1.463611277744451e-06, "loss": 0.0002, "step": 21393 }, { "epoch": 2.56, "learning_rate": 1.4635392921415719e-06, "loss": 0.0001, "step": 21394 }, { "epoch": 2.56, "learning_rate": 1.4634673065386924e-06, "loss": 0.0001, "step": 21395 }, { "epoch": 2.56, "learning_rate": 1.4633953209358128e-06, "loss": 0.0001, "step": 21396 }, { "epoch": 2.56, "learning_rate": 1.4633233353329334e-06, "loss": 0.0002, "step": 21397 }, { "epoch": 2.56, "learning_rate": 1.463251349730054e-06, "loss": 0.0002, "step": 21398 }, { "epoch": 2.56, "learning_rate": 1.4631793641271745e-06, "loss": 0.0004, "step": 21399 }, { "epoch": 2.56, "learning_rate": 1.463107378524295e-06, "loss": 1.118, "step": 21400 }, { "epoch": 2.56, "learning_rate": 1.4630353929214159e-06, "loss": 0.0001, "step": 21401 }, { "epoch": 2.56, "learning_rate": 1.4629634073185364e-06, "loss": 0.0005, "step": 21402 }, { "epoch": 2.56, "learning_rate": 1.462891421715657e-06, "loss": 0.0005, "step": 21403 }, { "epoch": 2.56, "learning_rate": 1.4628194361127776e-06, "loss": 0.0002, "step": 21404 }, { "epoch": 2.57, "learning_rate": 1.4627474505098981e-06, "loss": 0.0037, "step": 21405 }, { "epoch": 2.57, "learning_rate": 1.4626754649070187e-06, "loss": 0.0002, "step": 21406 }, { "epoch": 2.57, "learning_rate": 1.4626034793041393e-06, "loss": 0.0002, "step": 21407 }, { "epoch": 2.57, "learning_rate": 1.4625314937012598e-06, "loss": 0.0002, "step": 21408 }, { "epoch": 2.57, "learning_rate": 1.4624595080983804e-06, "loss": 0.0001, "step": 21409 }, { "epoch": 2.57, "learning_rate": 1.462387522495501e-06, "loss": 0.0001, "step": 21410 }, { "epoch": 2.57, "learning_rate": 1.4623155368926215e-06, "loss": 0.0001, "step": 21411 }, { "epoch": 2.57, "learning_rate": 1.462243551289742e-06, "loss": 0.0002, "step": 21412 }, { "epoch": 2.57, "learning_rate": 1.4621715656868627e-06, "loss": 0.0003, "step": 21413 }, { "epoch": 2.57, "learning_rate": 1.4620995800839832e-06, "loss": 0.0002, "step": 21414 }, { "epoch": 2.57, "learning_rate": 1.4620275944811038e-06, "loss": 0.0001, "step": 21415 }, { "epoch": 2.57, "learning_rate": 1.4619556088782244e-06, "loss": 0.0003, "step": 21416 }, { "epoch": 2.57, "learning_rate": 1.4618836232753451e-06, "loss": 0.0002, "step": 21417 }, { "epoch": 2.57, "learning_rate": 1.4618116376724657e-06, "loss": 0.0002, "step": 21418 }, { "epoch": 2.57, "learning_rate": 1.461739652069586e-06, "loss": 0.0001, "step": 21419 }, { "epoch": 2.57, "learning_rate": 1.4616676664667066e-06, "loss": 0.0001, "step": 21420 }, { "epoch": 2.57, "learning_rate": 1.4615956808638272e-06, "loss": 0.0001, "step": 21421 }, { "epoch": 2.57, "learning_rate": 1.4615236952609478e-06, "loss": 0.8178, "step": 21422 }, { "epoch": 2.57, "learning_rate": 1.4614517096580683e-06, "loss": 0.0001, "step": 21423 }, { "epoch": 2.57, "learning_rate": 1.4613797240551889e-06, "loss": 0.6069, "step": 21424 }, { "epoch": 2.57, "learning_rate": 1.4613077384523097e-06, "loss": 0.0001, "step": 21425 }, { "epoch": 2.57, "learning_rate": 1.4612357528494302e-06, "loss": 0.964, "step": 21426 }, { "epoch": 2.57, "learning_rate": 1.4611637672465508e-06, "loss": 0.0009, "step": 21427 }, { "epoch": 2.57, "learning_rate": 1.4610917816436714e-06, "loss": 0.9167, "step": 21428 }, { "epoch": 2.57, "learning_rate": 1.461019796040792e-06, "loss": 0.0002, "step": 21429 }, { "epoch": 2.57, "learning_rate": 1.4609478104379125e-06, "loss": 0.0005, "step": 21430 }, { "epoch": 2.57, "learning_rate": 1.4608758248350329e-06, "loss": 0.0001, "step": 21431 }, { "epoch": 2.57, "learning_rate": 1.4608038392321536e-06, "loss": 0.0004, "step": 21432 }, { "epoch": 2.57, "learning_rate": 1.4607318536292742e-06, "loss": 0.0001, "step": 21433 }, { "epoch": 2.57, "learning_rate": 1.4606598680263948e-06, "loss": 0.0002, "step": 21434 }, { "epoch": 2.57, "learning_rate": 1.4605878824235153e-06, "loss": 0.0009, "step": 21435 }, { "epoch": 2.57, "learning_rate": 1.460515896820636e-06, "loss": 0.0002, "step": 21436 }, { "epoch": 2.57, "learning_rate": 1.4604439112177565e-06, "loss": 0.0007, "step": 21437 }, { "epoch": 2.57, "learning_rate": 1.460371925614877e-06, "loss": 0.0001, "step": 21438 }, { "epoch": 2.57, "learning_rate": 1.4602999400119976e-06, "loss": 0.0004, "step": 21439 }, { "epoch": 2.57, "learning_rate": 1.4602279544091182e-06, "loss": 0.0003, "step": 21440 }, { "epoch": 2.57, "learning_rate": 1.460155968806239e-06, "loss": 0.0001, "step": 21441 }, { "epoch": 2.57, "learning_rate": 1.4600839832033593e-06, "loss": 0.0055, "step": 21442 }, { "epoch": 2.57, "learning_rate": 1.4600119976004799e-06, "loss": 0.8734, "step": 21443 }, { "epoch": 2.57, "learning_rate": 1.4599400119976004e-06, "loss": 0.0003, "step": 21444 }, { "epoch": 2.57, "learning_rate": 1.459868026394721e-06, "loss": 0.0002, "step": 21445 }, { "epoch": 2.57, "learning_rate": 1.4597960407918416e-06, "loss": 0.0004, "step": 21446 }, { "epoch": 2.57, "learning_rate": 1.4597240551889621e-06, "loss": 0.0057, "step": 21447 }, { "epoch": 2.57, "learning_rate": 1.459652069586083e-06, "loss": 0.7774, "step": 21448 }, { "epoch": 2.57, "learning_rate": 1.4595800839832035e-06, "loss": 0.0003, "step": 21449 }, { "epoch": 2.57, "learning_rate": 1.459508098380324e-06, "loss": 0.0004, "step": 21450 }, { "epoch": 2.57, "learning_rate": 1.4594361127774446e-06, "loss": 0.0003, "step": 21451 }, { "epoch": 2.57, "learning_rate": 1.4593641271745652e-06, "loss": 0.0003, "step": 21452 }, { "epoch": 2.57, "learning_rate": 1.4592921415716857e-06, "loss": 0.6665, "step": 21453 }, { "epoch": 2.57, "learning_rate": 1.459220155968806e-06, "loss": 0.0003, "step": 21454 }, { "epoch": 2.57, "learning_rate": 1.4591481703659269e-06, "loss": 0.0002, "step": 21455 }, { "epoch": 2.57, "learning_rate": 1.4590761847630474e-06, "loss": 0.0001, "step": 21456 }, { "epoch": 2.57, "learning_rate": 1.459004199160168e-06, "loss": 0.0001, "step": 21457 }, { "epoch": 2.57, "learning_rate": 1.4589322135572886e-06, "loss": 0.0001, "step": 21458 }, { "epoch": 2.57, "learning_rate": 1.4588602279544091e-06, "loss": 0.0003, "step": 21459 }, { "epoch": 2.57, "learning_rate": 1.4587882423515297e-06, "loss": 0.0003, "step": 21460 }, { "epoch": 2.57, "learning_rate": 1.4587162567486503e-06, "loss": 0.0002, "step": 21461 }, { "epoch": 2.57, "learning_rate": 1.4586442711457708e-06, "loss": 0.0003, "step": 21462 }, { "epoch": 2.57, "learning_rate": 1.4585722855428914e-06, "loss": 0.0002, "step": 21463 }, { "epoch": 2.57, "learning_rate": 1.4585002999400122e-06, "loss": 0.0004, "step": 21464 }, { "epoch": 2.57, "learning_rate": 1.4584283143371325e-06, "loss": 0.0003, "step": 21465 }, { "epoch": 2.57, "learning_rate": 1.4583563287342531e-06, "loss": 0.0002, "step": 21466 }, { "epoch": 2.57, "learning_rate": 1.4582843431313737e-06, "loss": 0.0009, "step": 21467 }, { "epoch": 2.57, "learning_rate": 1.4582123575284942e-06, "loss": 0.0003, "step": 21468 }, { "epoch": 2.57, "learning_rate": 1.4581403719256148e-06, "loss": 0.0002, "step": 21469 }, { "epoch": 2.57, "learning_rate": 1.4580683863227354e-06, "loss": 0.0002, "step": 21470 }, { "epoch": 2.57, "learning_rate": 1.4579964007198562e-06, "loss": 0.0003, "step": 21471 }, { "epoch": 2.57, "learning_rate": 1.4579244151169767e-06, "loss": 0.0005, "step": 21472 }, { "epoch": 2.57, "learning_rate": 1.4578524295140973e-06, "loss": 0.0004, "step": 21473 }, { "epoch": 2.57, "learning_rate": 1.4577804439112179e-06, "loss": 0.0005, "step": 21474 }, { "epoch": 2.57, "learning_rate": 1.4577084583083384e-06, "loss": 0.0005, "step": 21475 }, { "epoch": 2.57, "learning_rate": 1.457636472705459e-06, "loss": 0.0003, "step": 21476 }, { "epoch": 2.57, "learning_rate": 1.4575644871025794e-06, "loss": 0.0004, "step": 21477 }, { "epoch": 2.57, "learning_rate": 1.4574925014997001e-06, "loss": 0.0006, "step": 21478 }, { "epoch": 2.57, "learning_rate": 1.4574205158968207e-06, "loss": 0.0002, "step": 21479 }, { "epoch": 2.57, "learning_rate": 1.4573485302939413e-06, "loss": 0.0003, "step": 21480 }, { "epoch": 2.57, "learning_rate": 1.4572765446910618e-06, "loss": 0.0003, "step": 21481 }, { "epoch": 2.57, "learning_rate": 1.4572045590881824e-06, "loss": 0.0003, "step": 21482 }, { "epoch": 2.57, "learning_rate": 1.457132573485303e-06, "loss": 0.0002, "step": 21483 }, { "epoch": 2.57, "learning_rate": 1.4570605878824235e-06, "loss": 0.0004, "step": 21484 }, { "epoch": 2.57, "learning_rate": 1.456988602279544e-06, "loss": 0.0002, "step": 21485 }, { "epoch": 2.57, "learning_rate": 1.4569166166766647e-06, "loss": 0.0002, "step": 21486 }, { "epoch": 2.57, "learning_rate": 1.4568446310737854e-06, "loss": 0.0003, "step": 21487 }, { "epoch": 2.57, "learning_rate": 1.4567726454709058e-06, "loss": 0.0875, "step": 21488 }, { "epoch": 2.58, "learning_rate": 1.4567006598680264e-06, "loss": 0.001, "step": 21489 }, { "epoch": 2.58, "learning_rate": 1.456628674265147e-06, "loss": 0.0001, "step": 21490 }, { "epoch": 2.58, "learning_rate": 1.4565566886622675e-06, "loss": 0.0002, "step": 21491 }, { "epoch": 2.58, "learning_rate": 1.456484703059388e-06, "loss": 0.0008, "step": 21492 }, { "epoch": 2.58, "learning_rate": 1.4564127174565086e-06, "loss": 0.0006, "step": 21493 }, { "epoch": 2.58, "learning_rate": 1.4563407318536294e-06, "loss": 0.0002, "step": 21494 }, { "epoch": 2.58, "learning_rate": 1.45626874625075e-06, "loss": 0.0004, "step": 21495 }, { "epoch": 2.58, "learning_rate": 1.4561967606478705e-06, "loss": 0.8985, "step": 21496 }, { "epoch": 2.58, "learning_rate": 1.4561247750449911e-06, "loss": 0.0006, "step": 21497 }, { "epoch": 2.58, "learning_rate": 1.4560527894421117e-06, "loss": 0.0005, "step": 21498 }, { "epoch": 2.58, "learning_rate": 1.4559808038392322e-06, "loss": 0.3444, "step": 21499 }, { "epoch": 2.58, "learning_rate": 1.4559088182363526e-06, "loss": 0.0001, "step": 21500 }, { "epoch": 2.58, "learning_rate": 1.4558368326334734e-06, "loss": 0.5091, "step": 21501 }, { "epoch": 2.58, "learning_rate": 1.455764847030594e-06, "loss": 0.0001, "step": 21502 }, { "epoch": 2.58, "learning_rate": 1.4556928614277145e-06, "loss": 0.0003, "step": 21503 }, { "epoch": 2.58, "learning_rate": 1.455620875824835e-06, "loss": 0.0004, "step": 21504 }, { "epoch": 2.58, "learning_rate": 1.4555488902219556e-06, "loss": 0.0002, "step": 21505 }, { "epoch": 2.58, "learning_rate": 1.4554769046190762e-06, "loss": 0.0008, "step": 21506 }, { "epoch": 2.58, "learning_rate": 1.4554049190161968e-06, "loss": 0.0005, "step": 21507 }, { "epoch": 2.58, "learning_rate": 1.4553329334133173e-06, "loss": 0.0003, "step": 21508 }, { "epoch": 2.58, "learning_rate": 1.455260947810438e-06, "loss": 0.0004, "step": 21509 }, { "epoch": 2.58, "learning_rate": 1.4551889622075587e-06, "loss": 0.0014, "step": 21510 }, { "epoch": 2.58, "learning_rate": 1.4551169766046793e-06, "loss": 0.0002, "step": 21511 }, { "epoch": 2.58, "learning_rate": 1.4550449910017996e-06, "loss": 0.0003, "step": 21512 }, { "epoch": 2.58, "learning_rate": 1.4549730053989202e-06, "loss": 0.0001, "step": 21513 }, { "epoch": 2.58, "learning_rate": 1.4549010197960407e-06, "loss": 0.0005, "step": 21514 }, { "epoch": 2.58, "learning_rate": 1.4548290341931613e-06, "loss": 0.0005, "step": 21515 }, { "epoch": 2.58, "learning_rate": 1.4547570485902819e-06, "loss": 0.0015, "step": 21516 }, { "epoch": 2.58, "learning_rate": 1.4546850629874027e-06, "loss": 0.0005, "step": 21517 }, { "epoch": 2.58, "learning_rate": 1.4546130773845232e-06, "loss": 0.0005, "step": 21518 }, { "epoch": 2.58, "learning_rate": 1.4545410917816438e-06, "loss": 0.0003, "step": 21519 }, { "epoch": 2.58, "learning_rate": 1.4544691061787644e-06, "loss": 0.0003, "step": 21520 }, { "epoch": 2.58, "learning_rate": 1.454397120575885e-06, "loss": 0.0003, "step": 21521 }, { "epoch": 2.58, "learning_rate": 1.4543251349730055e-06, "loss": 0.0001, "step": 21522 }, { "epoch": 2.58, "learning_rate": 1.4542531493701258e-06, "loss": 0.0008, "step": 21523 }, { "epoch": 2.58, "learning_rate": 1.4541811637672466e-06, "loss": 0.0003, "step": 21524 }, { "epoch": 2.58, "learning_rate": 1.4541091781643672e-06, "loss": 0.0001, "step": 21525 }, { "epoch": 2.58, "learning_rate": 1.4540371925614878e-06, "loss": 0.0002, "step": 21526 }, { "epoch": 2.58, "learning_rate": 1.4539652069586083e-06, "loss": 0.0007, "step": 21527 }, { "epoch": 2.58, "learning_rate": 1.4538932213557289e-06, "loss": 0.0003, "step": 21528 }, { "epoch": 2.58, "learning_rate": 1.4538212357528495e-06, "loss": 0.0003, "step": 21529 }, { "epoch": 2.58, "learning_rate": 1.45374925014997e-06, "loss": 0.2264, "step": 21530 }, { "epoch": 2.58, "learning_rate": 1.4536772645470906e-06, "loss": 0.0004, "step": 21531 }, { "epoch": 2.58, "learning_rate": 1.4536052789442112e-06, "loss": 0.0002, "step": 21532 }, { "epoch": 2.58, "learning_rate": 1.453533293341332e-06, "loss": 0.0004, "step": 21533 }, { "epoch": 2.58, "learning_rate": 1.4534613077384525e-06, "loss": 0.0003, "step": 21534 }, { "epoch": 2.58, "learning_rate": 1.4533893221355729e-06, "loss": 0.0002, "step": 21535 }, { "epoch": 2.58, "learning_rate": 1.4533173365326934e-06, "loss": 0.0002, "step": 21536 }, { "epoch": 2.58, "learning_rate": 1.453245350929814e-06, "loss": 0.0003, "step": 21537 }, { "epoch": 2.58, "learning_rate": 1.4531733653269346e-06, "loss": 0.0004, "step": 21538 }, { "epoch": 2.58, "learning_rate": 1.4531013797240551e-06, "loss": 0.001, "step": 21539 }, { "epoch": 2.58, "learning_rate": 1.453029394121176e-06, "loss": 0.0004, "step": 21540 }, { "epoch": 2.58, "learning_rate": 1.4529574085182965e-06, "loss": 0.0003, "step": 21541 }, { "epoch": 2.58, "learning_rate": 1.452885422915417e-06, "loss": 0.0004, "step": 21542 }, { "epoch": 2.58, "learning_rate": 1.4528134373125376e-06, "loss": 0.0002, "step": 21543 }, { "epoch": 2.58, "learning_rate": 1.4527414517096582e-06, "loss": 0.0011, "step": 21544 }, { "epoch": 2.58, "learning_rate": 1.4526694661067787e-06, "loss": 0.0002, "step": 21545 }, { "epoch": 2.58, "learning_rate": 1.452597480503899e-06, "loss": 0.0002, "step": 21546 }, { "epoch": 2.58, "learning_rate": 1.4525254949010199e-06, "loss": 0.0004, "step": 21547 }, { "epoch": 2.58, "learning_rate": 1.4524535092981404e-06, "loss": 0.0005, "step": 21548 }, { "epoch": 2.58, "learning_rate": 1.452381523695261e-06, "loss": 0.0015, "step": 21549 }, { "epoch": 2.58, "learning_rate": 1.4523095380923816e-06, "loss": 0.0006, "step": 21550 }, { "epoch": 2.58, "learning_rate": 1.4522375524895021e-06, "loss": 0.0002, "step": 21551 }, { "epoch": 2.58, "learning_rate": 1.4521655668866227e-06, "loss": 0.0002, "step": 21552 }, { "epoch": 2.58, "learning_rate": 1.4520935812837433e-06, "loss": 0.0002, "step": 21553 }, { "epoch": 2.58, "learning_rate": 1.4520215956808638e-06, "loss": 0.0002, "step": 21554 }, { "epoch": 2.58, "learning_rate": 1.4519496100779844e-06, "loss": 0.0003, "step": 21555 }, { "epoch": 2.58, "learning_rate": 1.4518776244751052e-06, "loss": 0.0001, "step": 21556 }, { "epoch": 2.58, "learning_rate": 1.4518056388722258e-06, "loss": 0.0003, "step": 21557 }, { "epoch": 2.58, "learning_rate": 1.4517336532693461e-06, "loss": 0.0002, "step": 21558 }, { "epoch": 2.58, "learning_rate": 1.4516616676664667e-06, "loss": 0.0003, "step": 21559 }, { "epoch": 2.58, "learning_rate": 1.4515896820635872e-06, "loss": 0.0005, "step": 21560 }, { "epoch": 2.58, "learning_rate": 1.4515176964607078e-06, "loss": 0.0005, "step": 21561 }, { "epoch": 2.58, "learning_rate": 1.4514457108578284e-06, "loss": 0.0001, "step": 21562 }, { "epoch": 2.58, "learning_rate": 1.451373725254949e-06, "loss": 0.0008, "step": 21563 }, { "epoch": 2.58, "learning_rate": 1.4513017396520697e-06, "loss": 0.0002, "step": 21564 }, { "epoch": 2.58, "learning_rate": 1.4512297540491903e-06, "loss": 0.0002, "step": 21565 }, { "epoch": 2.58, "learning_rate": 1.4511577684463109e-06, "loss": 0.0006, "step": 21566 }, { "epoch": 2.58, "learning_rate": 1.4510857828434314e-06, "loss": 0.0002, "step": 21567 }, { "epoch": 2.58, "learning_rate": 1.451013797240552e-06, "loss": 0.0002, "step": 21568 }, { "epoch": 2.58, "learning_rate": 1.4509418116376723e-06, "loss": 0.0002, "step": 21569 }, { "epoch": 2.58, "learning_rate": 1.450869826034793e-06, "loss": 0.0003, "step": 21570 }, { "epoch": 2.58, "learning_rate": 1.4507978404319137e-06, "loss": 0.0005, "step": 21571 }, { "epoch": 2.59, "learning_rate": 1.4507258548290343e-06, "loss": 0.0003, "step": 21572 }, { "epoch": 2.59, "learning_rate": 1.4506538692261548e-06, "loss": 0.0002, "step": 21573 }, { "epoch": 2.59, "learning_rate": 1.4505818836232754e-06, "loss": 0.0003, "step": 21574 }, { "epoch": 2.59, "learning_rate": 1.450509898020396e-06, "loss": 0.0002, "step": 21575 }, { "epoch": 2.59, "learning_rate": 1.4504379124175165e-06, "loss": 0.0003, "step": 21576 }, { "epoch": 2.59, "learning_rate": 1.450365926814637e-06, "loss": 0.0002, "step": 21577 }, { "epoch": 2.59, "learning_rate": 1.4502939412117577e-06, "loss": 0.0001, "step": 21578 }, { "epoch": 2.59, "learning_rate": 1.4502219556088782e-06, "loss": 0.0007, "step": 21579 }, { "epoch": 2.59, "learning_rate": 1.450149970005999e-06, "loss": 0.0002, "step": 21580 }, { "epoch": 2.59, "learning_rate": 1.4500779844031194e-06, "loss": 0.0002, "step": 21581 }, { "epoch": 2.59, "learning_rate": 1.45000599880024e-06, "loss": 0.0002, "step": 21582 }, { "epoch": 2.59, "learning_rate": 1.4499340131973605e-06, "loss": 0.0003, "step": 21583 }, { "epoch": 2.59, "learning_rate": 1.449862027594481e-06, "loss": 0.0001, "step": 21584 }, { "epoch": 2.59, "learning_rate": 1.4497900419916016e-06, "loss": 0.0002, "step": 21585 }, { "epoch": 2.59, "learning_rate": 1.4497180563887222e-06, "loss": 0.0001, "step": 21586 }, { "epoch": 2.59, "learning_rate": 1.449646070785843e-06, "loss": 0.0005, "step": 21587 }, { "epoch": 2.59, "learning_rate": 1.4495740851829635e-06, "loss": 0.0003, "step": 21588 }, { "epoch": 2.59, "learning_rate": 1.449502099580084e-06, "loss": 0.0015, "step": 21589 }, { "epoch": 2.59, "learning_rate": 1.4494301139772047e-06, "loss": 0.0002, "step": 21590 }, { "epoch": 2.59, "learning_rate": 1.4493581283743252e-06, "loss": 0.0001, "step": 21591 }, { "epoch": 2.59, "learning_rate": 1.4492861427714456e-06, "loss": 0.0001, "step": 21592 }, { "epoch": 2.59, "learning_rate": 1.4492141571685662e-06, "loss": 0.0003, "step": 21593 }, { "epoch": 2.59, "learning_rate": 1.449142171565687e-06, "loss": 0.0003, "step": 21594 }, { "epoch": 2.59, "learning_rate": 1.4490701859628075e-06, "loss": 0.0006, "step": 21595 }, { "epoch": 2.59, "learning_rate": 1.448998200359928e-06, "loss": 0.0006, "step": 21596 }, { "epoch": 2.59, "learning_rate": 1.4489262147570486e-06, "loss": 0.0008, "step": 21597 }, { "epoch": 2.59, "learning_rate": 1.4488542291541692e-06, "loss": 0.0003, "step": 21598 }, { "epoch": 2.59, "learning_rate": 1.4487822435512898e-06, "loss": 0.016, "step": 21599 }, { "epoch": 2.59, "learning_rate": 1.4487102579484103e-06, "loss": 0.0004, "step": 21600 }, { "epoch": 2.59, "learning_rate": 1.448638272345531e-06, "loss": 0.0007, "step": 21601 }, { "epoch": 2.59, "learning_rate": 1.4485662867426515e-06, "loss": 0.0001, "step": 21602 }, { "epoch": 2.59, "learning_rate": 1.4484943011397722e-06, "loss": 0.0001, "step": 21603 }, { "epoch": 2.59, "learning_rate": 1.4484223155368926e-06, "loss": 0.0001, "step": 21604 }, { "epoch": 2.59, "learning_rate": 1.4483503299340132e-06, "loss": 0.0002, "step": 21605 }, { "epoch": 2.59, "learning_rate": 1.4482783443311337e-06, "loss": 0.0002, "step": 21606 }, { "epoch": 2.59, "learning_rate": 1.4482063587282543e-06, "loss": 0.0003, "step": 21607 }, { "epoch": 2.59, "learning_rate": 1.4481343731253749e-06, "loss": 0.0002, "step": 21608 }, { "epoch": 2.59, "learning_rate": 1.4480623875224954e-06, "loss": 0.0002, "step": 21609 }, { "epoch": 2.59, "learning_rate": 1.4479904019196162e-06, "loss": 0.0014, "step": 21610 }, { "epoch": 2.59, "learning_rate": 1.4479184163167368e-06, "loss": 0.0002, "step": 21611 }, { "epoch": 2.59, "learning_rate": 1.4478464307138573e-06, "loss": 0.0002, "step": 21612 }, { "epoch": 2.59, "learning_rate": 1.447774445110978e-06, "loss": 0.0004, "step": 21613 }, { "epoch": 2.59, "learning_rate": 1.4477024595080985e-06, "loss": 0.0001, "step": 21614 }, { "epoch": 2.59, "learning_rate": 1.4476304739052188e-06, "loss": 0.0001, "step": 21615 }, { "epoch": 2.59, "learning_rate": 1.4475584883023394e-06, "loss": 0.0002, "step": 21616 }, { "epoch": 2.59, "learning_rate": 1.4474865026994602e-06, "loss": 0.0001, "step": 21617 }, { "epoch": 2.59, "learning_rate": 1.4474145170965808e-06, "loss": 0.0005, "step": 21618 }, { "epoch": 2.59, "learning_rate": 1.4473425314937013e-06, "loss": 0.0001, "step": 21619 }, { "epoch": 2.59, "learning_rate": 1.4472705458908219e-06, "loss": 0.0003, "step": 21620 }, { "epoch": 2.59, "learning_rate": 1.4471985602879425e-06, "loss": 0.0001, "step": 21621 }, { "epoch": 2.59, "learning_rate": 1.447126574685063e-06, "loss": 0.0002, "step": 21622 }, { "epoch": 2.59, "learning_rate": 1.4470545890821836e-06, "loss": 0.0002, "step": 21623 }, { "epoch": 2.59, "learning_rate": 1.4469826034793042e-06, "loss": 0.0003, "step": 21624 }, { "epoch": 2.59, "learning_rate": 1.4469106178764247e-06, "loss": 0.0003, "step": 21625 }, { "epoch": 2.59, "learning_rate": 1.4468386322735455e-06, "loss": 0.0002, "step": 21626 }, { "epoch": 2.59, "learning_rate": 1.4467666466706659e-06, "loss": 0.0006, "step": 21627 }, { "epoch": 2.59, "learning_rate": 1.4466946610677864e-06, "loss": 0.0003, "step": 21628 }, { "epoch": 2.59, "learning_rate": 1.446622675464907e-06, "loss": 0.0002, "step": 21629 }, { "epoch": 2.59, "learning_rate": 1.4465506898620276e-06, "loss": 0.0001, "step": 21630 }, { "epoch": 2.59, "learning_rate": 1.4464787042591481e-06, "loss": 0.0001, "step": 21631 }, { "epoch": 2.59, "learning_rate": 1.4464067186562687e-06, "loss": 0.0003, "step": 21632 }, { "epoch": 2.59, "learning_rate": 1.4463347330533895e-06, "loss": 0.0008, "step": 21633 }, { "epoch": 2.59, "learning_rate": 1.44626274745051e-06, "loss": 0.0002, "step": 21634 }, { "epoch": 2.59, "learning_rate": 1.4461907618476306e-06, "loss": 0.0003, "step": 21635 }, { "epoch": 2.59, "learning_rate": 1.4461187762447512e-06, "loss": 0.0002, "step": 21636 }, { "epoch": 2.59, "learning_rate": 1.4460467906418717e-06, "loss": 0.0001, "step": 21637 }, { "epoch": 2.59, "learning_rate": 1.4459748050389923e-06, "loss": 0.1352, "step": 21638 }, { "epoch": 2.59, "learning_rate": 1.4459028194361127e-06, "loss": 0.0003, "step": 21639 }, { "epoch": 2.59, "learning_rate": 1.4458308338332334e-06, "loss": 0.0005, "step": 21640 }, { "epoch": 2.59, "learning_rate": 1.445758848230354e-06, "loss": 0.0003, "step": 21641 }, { "epoch": 2.59, "learning_rate": 1.4456868626274746e-06, "loss": 0.0002, "step": 21642 }, { "epoch": 2.59, "learning_rate": 1.4456148770245951e-06, "loss": 0.0001, "step": 21643 }, { "epoch": 2.59, "learning_rate": 1.4455428914217157e-06, "loss": 0.0002, "step": 21644 }, { "epoch": 2.59, "learning_rate": 1.4454709058188363e-06, "loss": 0.0001, "step": 21645 }, { "epoch": 2.59, "learning_rate": 1.4453989202159568e-06, "loss": 0.0001, "step": 21646 }, { "epoch": 2.59, "learning_rate": 1.4453269346130774e-06, "loss": 0.0002, "step": 21647 }, { "epoch": 2.59, "learning_rate": 1.445254949010198e-06, "loss": 0.0008, "step": 21648 }, { "epoch": 2.59, "learning_rate": 1.4451829634073187e-06, "loss": 0.0003, "step": 21649 }, { "epoch": 2.59, "learning_rate": 1.445110977804439e-06, "loss": 0.0004, "step": 21650 }, { "epoch": 2.59, "learning_rate": 1.4450389922015597e-06, "loss": 0.0003, "step": 21651 }, { "epoch": 2.59, "learning_rate": 1.4449670065986802e-06, "loss": 0.0001, "step": 21652 }, { "epoch": 2.59, "learning_rate": 1.4448950209958008e-06, "loss": 0.0007, "step": 21653 }, { "epoch": 2.59, "learning_rate": 1.4448230353929214e-06, "loss": 0.0006, "step": 21654 }, { "epoch": 2.59, "learning_rate": 1.444751049790042e-06, "loss": 0.0005, "step": 21655 }, { "epoch": 2.6, "learning_rate": 1.4446790641871627e-06, "loss": 0.0001, "step": 21656 }, { "epoch": 2.6, "learning_rate": 1.4446070785842833e-06, "loss": 0.0001, "step": 21657 }, { "epoch": 2.6, "learning_rate": 1.4445350929814038e-06, "loss": 0.0004, "step": 21658 }, { "epoch": 2.6, "learning_rate": 1.4444631073785244e-06, "loss": 0.0002, "step": 21659 }, { "epoch": 2.6, "learning_rate": 1.444391121775645e-06, "loss": 0.0001, "step": 21660 }, { "epoch": 2.6, "learning_rate": 1.4443191361727655e-06, "loss": 0.0002, "step": 21661 }, { "epoch": 2.6, "learning_rate": 1.444247150569886e-06, "loss": 0.0004, "step": 21662 }, { "epoch": 2.6, "learning_rate": 1.4441751649670067e-06, "loss": 0.0005, "step": 21663 }, { "epoch": 2.6, "learning_rate": 1.4441031793641272e-06, "loss": 0.0008, "step": 21664 }, { "epoch": 2.6, "learning_rate": 1.4440311937612478e-06, "loss": 0.0002, "step": 21665 }, { "epoch": 2.6, "learning_rate": 1.4439592081583684e-06, "loss": 0.0001, "step": 21666 }, { "epoch": 2.6, "learning_rate": 1.443887222555489e-06, "loss": 0.0001, "step": 21667 }, { "epoch": 2.6, "learning_rate": 1.4438152369526095e-06, "loss": 0.0001, "step": 21668 }, { "epoch": 2.6, "learning_rate": 1.44374325134973e-06, "loss": 0.0001, "step": 21669 }, { "epoch": 2.6, "learning_rate": 1.4436712657468506e-06, "loss": 0.0002, "step": 21670 }, { "epoch": 2.6, "learning_rate": 1.4435992801439712e-06, "loss": 0.0009, "step": 21671 }, { "epoch": 2.6, "learning_rate": 1.443527294541092e-06, "loss": 0.0073, "step": 21672 }, { "epoch": 2.6, "learning_rate": 1.4434553089382123e-06, "loss": 0.0003, "step": 21673 }, { "epoch": 2.6, "learning_rate": 1.443383323335333e-06, "loss": 0.0001, "step": 21674 }, { "epoch": 2.6, "learning_rate": 1.4433113377324535e-06, "loss": 0.0001, "step": 21675 }, { "epoch": 2.6, "learning_rate": 1.443239352129574e-06, "loss": 0.0007, "step": 21676 }, { "epoch": 2.6, "learning_rate": 1.4431673665266946e-06, "loss": 0.0001, "step": 21677 }, { "epoch": 2.6, "learning_rate": 1.4430953809238152e-06, "loss": 0.0002, "step": 21678 }, { "epoch": 2.6, "learning_rate": 1.443023395320936e-06, "loss": 0.0002, "step": 21679 }, { "epoch": 2.6, "learning_rate": 1.4429514097180565e-06, "loss": 0.0002, "step": 21680 }, { "epoch": 2.6, "learning_rate": 1.442879424115177e-06, "loss": 0.0002, "step": 21681 }, { "epoch": 2.6, "learning_rate": 1.4428074385122977e-06, "loss": 0.0, "step": 21682 }, { "epoch": 2.6, "learning_rate": 1.4427354529094182e-06, "loss": 0.0002, "step": 21683 }, { "epoch": 2.6, "learning_rate": 1.4426634673065388e-06, "loss": 0.0002, "step": 21684 }, { "epoch": 2.6, "learning_rate": 1.4425914817036591e-06, "loss": 0.0001, "step": 21685 }, { "epoch": 2.6, "learning_rate": 1.44251949610078e-06, "loss": 0.0002, "step": 21686 }, { "epoch": 2.6, "learning_rate": 1.4424475104979005e-06, "loss": 0.0001, "step": 21687 }, { "epoch": 2.6, "learning_rate": 1.442375524895021e-06, "loss": 0.0002, "step": 21688 }, { "epoch": 2.6, "learning_rate": 1.4423035392921416e-06, "loss": 0.0002, "step": 21689 }, { "epoch": 2.6, "learning_rate": 1.4422315536892622e-06, "loss": 0.0001, "step": 21690 }, { "epoch": 2.6, "learning_rate": 1.4421595680863828e-06, "loss": 0.0002, "step": 21691 }, { "epoch": 2.6, "learning_rate": 1.4420875824835033e-06, "loss": 0.0001, "step": 21692 }, { "epoch": 2.6, "learning_rate": 1.442015596880624e-06, "loss": 0.0001, "step": 21693 }, { "epoch": 2.6, "learning_rate": 1.4419436112777445e-06, "loss": 0.0003, "step": 21694 }, { "epoch": 2.6, "learning_rate": 1.4418716256748652e-06, "loss": 0.0004, "step": 21695 }, { "epoch": 2.6, "learning_rate": 1.4417996400719856e-06, "loss": 0.0001, "step": 21696 }, { "epoch": 2.6, "learning_rate": 1.4417276544691062e-06, "loss": 0.0003, "step": 21697 }, { "epoch": 2.6, "learning_rate": 1.4416556688662267e-06, "loss": 0.0001, "step": 21698 }, { "epoch": 2.6, "learning_rate": 1.4415836832633473e-06, "loss": 0.0001, "step": 21699 }, { "epoch": 2.6, "learning_rate": 1.4415116976604679e-06, "loss": 0.0001, "step": 21700 }, { "epoch": 2.6, "learning_rate": 1.4414397120575884e-06, "loss": 0.0002, "step": 21701 }, { "epoch": 2.6, "learning_rate": 1.441367726454709e-06, "loss": 0.0001, "step": 21702 }, { "epoch": 2.6, "learning_rate": 1.4412957408518298e-06, "loss": 0.0001, "step": 21703 }, { "epoch": 2.6, "learning_rate": 1.4412237552489503e-06, "loss": 0.0001, "step": 21704 }, { "epoch": 2.6, "learning_rate": 1.441151769646071e-06, "loss": 0.0006, "step": 21705 }, { "epoch": 2.6, "learning_rate": 1.4410797840431915e-06, "loss": 0.0001, "step": 21706 }, { "epoch": 2.6, "learning_rate": 1.441007798440312e-06, "loss": 0.0002, "step": 21707 }, { "epoch": 2.6, "learning_rate": 1.4409358128374324e-06, "loss": 0.0003, "step": 21708 }, { "epoch": 2.6, "learning_rate": 1.440863827234553e-06, "loss": 0.0001, "step": 21709 }, { "epoch": 2.6, "learning_rate": 1.4407918416316737e-06, "loss": 0.0002, "step": 21710 }, { "epoch": 2.6, "learning_rate": 1.4407198560287943e-06, "loss": 0.0001, "step": 21711 }, { "epoch": 2.6, "learning_rate": 1.4406478704259149e-06, "loss": 0.0001, "step": 21712 }, { "epoch": 2.6, "learning_rate": 1.4405758848230354e-06, "loss": 0.0001, "step": 21713 }, { "epoch": 2.6, "learning_rate": 1.440503899220156e-06, "loss": 0.0002, "step": 21714 }, { "epoch": 2.6, "learning_rate": 1.4404319136172766e-06, "loss": 0.0001, "step": 21715 }, { "epoch": 2.6, "learning_rate": 1.4403599280143971e-06, "loss": 0.0002, "step": 21716 }, { "epoch": 2.6, "learning_rate": 1.4402879424115177e-06, "loss": 0.0002, "step": 21717 }, { "epoch": 2.6, "learning_rate": 1.4402159568086383e-06, "loss": 0.0001, "step": 21718 }, { "epoch": 2.6, "learning_rate": 1.4401439712057588e-06, "loss": 0.0001, "step": 21719 }, { "epoch": 2.6, "learning_rate": 1.4400719856028794e-06, "loss": 0.0001, "step": 21720 }, { "epoch": 2.6, "learning_rate": 1.44e-06, "loss": 0.0003, "step": 21721 }, { "epoch": 2.6, "learning_rate": 1.4399280143971205e-06, "loss": 0.0002, "step": 21722 }, { "epoch": 2.6, "learning_rate": 1.4398560287942411e-06, "loss": 0.0001, "step": 21723 }, { "epoch": 2.6, "learning_rate": 1.4397840431913617e-06, "loss": 0.0003, "step": 21724 }, { "epoch": 2.6, "learning_rate": 1.4397120575884822e-06, "loss": 0.0002, "step": 21725 }, { "epoch": 2.6, "learning_rate": 1.439640071985603e-06, "loss": 0.0001, "step": 21726 }, { "epoch": 2.6, "learning_rate": 1.4395680863827236e-06, "loss": 0.0001, "step": 21727 }, { "epoch": 2.6, "learning_rate": 1.4394961007798442e-06, "loss": 0.0001, "step": 21728 }, { "epoch": 2.6, "learning_rate": 1.4394241151769647e-06, "loss": 0.0001, "step": 21729 }, { "epoch": 2.6, "learning_rate": 1.4393521295740853e-06, "loss": 0.0001, "step": 21730 }, { "epoch": 2.6, "learning_rate": 1.4392801439712056e-06, "loss": 0.0001, "step": 21731 }, { "epoch": 2.6, "learning_rate": 1.4392081583683262e-06, "loss": 0.0002, "step": 21732 }, { "epoch": 2.6, "learning_rate": 1.439136172765447e-06, "loss": 0.0001, "step": 21733 }, { "epoch": 2.6, "learning_rate": 1.4390641871625676e-06, "loss": 0.0001, "step": 21734 }, { "epoch": 2.6, "learning_rate": 1.4389922015596881e-06, "loss": 0.0001, "step": 21735 }, { "epoch": 2.6, "learning_rate": 1.4389202159568087e-06, "loss": 0.0001, "step": 21736 }, { "epoch": 2.6, "learning_rate": 1.4388482303539293e-06, "loss": 0.0002, "step": 21737 }, { "epoch": 2.6, "learning_rate": 1.4387762447510498e-06, "loss": 0.0001, "step": 21738 }, { "epoch": 2.61, "learning_rate": 1.4387042591481704e-06, "loss": 0.0002, "step": 21739 }, { "epoch": 2.61, "learning_rate": 1.438632273545291e-06, "loss": 0.0002, "step": 21740 }, { "epoch": 2.61, "learning_rate": 1.4385602879424115e-06, "loss": 0.0002, "step": 21741 }, { "epoch": 2.61, "learning_rate": 1.4384883023395323e-06, "loss": 0.0522, "step": 21742 }, { "epoch": 2.61, "learning_rate": 1.4384163167366527e-06, "loss": 0.0002, "step": 21743 }, { "epoch": 2.61, "learning_rate": 1.4383443311337732e-06, "loss": 0.0002, "step": 21744 }, { "epoch": 2.61, "learning_rate": 1.4382723455308938e-06, "loss": 0.0001, "step": 21745 }, { "epoch": 2.61, "learning_rate": 1.4382003599280144e-06, "loss": 0.0001, "step": 21746 }, { "epoch": 2.61, "learning_rate": 1.438128374325135e-06, "loss": 0.0001, "step": 21747 }, { "epoch": 2.61, "learning_rate": 1.4380563887222555e-06, "loss": 0.6248, "step": 21748 }, { "epoch": 2.61, "learning_rate": 1.4379844031193763e-06, "loss": 0.0002, "step": 21749 }, { "epoch": 2.61, "learning_rate": 1.4379124175164968e-06, "loss": 0.0001, "step": 21750 }, { "epoch": 2.61, "learning_rate": 1.4378404319136174e-06, "loss": 0.0001, "step": 21751 }, { "epoch": 2.61, "learning_rate": 1.437768446310738e-06, "loss": 0.0003, "step": 21752 }, { "epoch": 2.61, "learning_rate": 1.4376964607078585e-06, "loss": 0.0001, "step": 21753 }, { "epoch": 2.61, "learning_rate": 1.4376244751049789e-06, "loss": 0.0001, "step": 21754 }, { "epoch": 2.61, "learning_rate": 1.4375524895020995e-06, "loss": 0.0002, "step": 21755 }, { "epoch": 2.61, "learning_rate": 1.4374805038992202e-06, "loss": 0.0001, "step": 21756 }, { "epoch": 2.61, "learning_rate": 1.4374085182963408e-06, "loss": 0.0107, "step": 21757 }, { "epoch": 2.61, "learning_rate": 1.4373365326934614e-06, "loss": 0.0004, "step": 21758 }, { "epoch": 2.61, "learning_rate": 1.437264547090582e-06, "loss": 1.1138, "step": 21759 }, { "epoch": 2.61, "learning_rate": 1.4371925614877025e-06, "loss": 0.0001, "step": 21760 }, { "epoch": 2.61, "learning_rate": 1.437120575884823e-06, "loss": 0.0007, "step": 21761 }, { "epoch": 2.61, "learning_rate": 1.4370485902819436e-06, "loss": 0.0002, "step": 21762 }, { "epoch": 2.61, "learning_rate": 1.4369766046790642e-06, "loss": 0.8896, "step": 21763 }, { "epoch": 2.61, "learning_rate": 1.4369046190761848e-06, "loss": 0.0001, "step": 21764 }, { "epoch": 2.61, "learning_rate": 1.4368326334733056e-06, "loss": 0.0002, "step": 21765 }, { "epoch": 2.61, "learning_rate": 1.436760647870426e-06, "loss": 0.0001, "step": 21766 }, { "epoch": 2.61, "learning_rate": 1.4366886622675465e-06, "loss": 0.0001, "step": 21767 }, { "epoch": 2.61, "learning_rate": 1.436616676664667e-06, "loss": 0.0001, "step": 21768 }, { "epoch": 2.61, "learning_rate": 1.4365446910617876e-06, "loss": 0.0009, "step": 21769 }, { "epoch": 2.61, "learning_rate": 1.4364727054589082e-06, "loss": 0.9085, "step": 21770 }, { "epoch": 2.61, "learning_rate": 1.4364007198560287e-06, "loss": 0.0002, "step": 21771 }, { "epoch": 2.61, "learning_rate": 1.4363287342531495e-06, "loss": 0.0002, "step": 21772 }, { "epoch": 2.61, "learning_rate": 1.43625674865027e-06, "loss": 0.0001, "step": 21773 }, { "epoch": 2.61, "learning_rate": 1.4361847630473907e-06, "loss": 0.0001, "step": 21774 }, { "epoch": 2.61, "learning_rate": 1.4361127774445112e-06, "loss": 0.0004, "step": 21775 }, { "epoch": 2.61, "learning_rate": 1.4360407918416318e-06, "loss": 0.0003, "step": 21776 }, { "epoch": 2.61, "learning_rate": 1.4359688062387521e-06, "loss": 0.9309, "step": 21777 }, { "epoch": 2.61, "learning_rate": 1.4358968206358727e-06, "loss": 0.0021, "step": 21778 }, { "epoch": 2.61, "learning_rate": 1.4358248350329935e-06, "loss": 0.0002, "step": 21779 }, { "epoch": 2.61, "learning_rate": 1.435752849430114e-06, "loss": 0.0035, "step": 21780 }, { "epoch": 2.61, "learning_rate": 1.4356808638272346e-06, "loss": 0.0002, "step": 21781 }, { "epoch": 2.61, "learning_rate": 1.4356088782243552e-06, "loss": 0.0002, "step": 21782 }, { "epoch": 2.61, "learning_rate": 1.4355368926214758e-06, "loss": 0.0002, "step": 21783 }, { "epoch": 2.61, "learning_rate": 1.4354649070185963e-06, "loss": 0.0002, "step": 21784 }, { "epoch": 2.61, "learning_rate": 1.4353929214157169e-06, "loss": 0.0001, "step": 21785 }, { "epoch": 2.61, "learning_rate": 1.4353209358128375e-06, "loss": 0.0003, "step": 21786 }, { "epoch": 2.61, "learning_rate": 1.435248950209958e-06, "loss": 0.0002, "step": 21787 }, { "epoch": 2.61, "learning_rate": 1.4351769646070788e-06, "loss": 0.001, "step": 21788 }, { "epoch": 2.61, "learning_rate": 1.4351049790041992e-06, "loss": 0.0001, "step": 21789 }, { "epoch": 2.61, "learning_rate": 1.4350329934013197e-06, "loss": 0.0003, "step": 21790 }, { "epoch": 2.61, "learning_rate": 1.4349610077984403e-06, "loss": 0.0002, "step": 21791 }, { "epoch": 2.61, "learning_rate": 1.4348890221955609e-06, "loss": 0.0005, "step": 21792 }, { "epoch": 2.61, "learning_rate": 1.4348170365926814e-06, "loss": 0.0003, "step": 21793 }, { "epoch": 2.61, "learning_rate": 1.434745050989802e-06, "loss": 0.89, "step": 21794 }, { "epoch": 2.61, "learning_rate": 1.4346730653869228e-06, "loss": 0.0002, "step": 21795 }, { "epoch": 2.61, "learning_rate": 1.4346010797840433e-06, "loss": 0.0002, "step": 21796 }, { "epoch": 2.61, "learning_rate": 1.434529094181164e-06, "loss": 0.0002, "step": 21797 }, { "epoch": 2.61, "learning_rate": 1.4344571085782845e-06, "loss": 0.0002, "step": 21798 }, { "epoch": 2.61, "learning_rate": 1.434385122975405e-06, "loss": 0.0006, "step": 21799 }, { "epoch": 2.61, "learning_rate": 1.4343131373725254e-06, "loss": 0.0006, "step": 21800 }, { "epoch": 2.61, "learning_rate": 1.434241151769646e-06, "loss": 0.0002, "step": 21801 }, { "epoch": 2.61, "learning_rate": 1.4341691661667667e-06, "loss": 0.0001, "step": 21802 }, { "epoch": 2.61, "learning_rate": 1.4340971805638873e-06, "loss": 0.0002, "step": 21803 }, { "epoch": 2.61, "learning_rate": 1.4340251949610079e-06, "loss": 0.0002, "step": 21804 }, { "epoch": 2.61, "learning_rate": 1.4339532093581284e-06, "loss": 0.0001, "step": 21805 }, { "epoch": 2.61, "learning_rate": 1.433881223755249e-06, "loss": 0.0009, "step": 21806 }, { "epoch": 2.61, "learning_rate": 1.4338092381523696e-06, "loss": 0.742, "step": 21807 }, { "epoch": 2.61, "learning_rate": 1.4337372525494901e-06, "loss": 0.0002, "step": 21808 }, { "epoch": 2.61, "learning_rate": 1.4336652669466107e-06, "loss": 0.0001, "step": 21809 }, { "epoch": 2.61, "learning_rate": 1.4335932813437313e-06, "loss": 0.0002, "step": 21810 }, { "epoch": 2.61, "learning_rate": 1.433521295740852e-06, "loss": 0.0003, "step": 21811 }, { "epoch": 2.61, "learning_rate": 1.4334493101379724e-06, "loss": 0.0005, "step": 21812 }, { "epoch": 2.61, "learning_rate": 1.433377324535093e-06, "loss": 0.0001, "step": 21813 }, { "epoch": 2.61, "learning_rate": 1.4333053389322135e-06, "loss": 0.0001, "step": 21814 }, { "epoch": 2.61, "learning_rate": 1.433233353329334e-06, "loss": 0.0009, "step": 21815 }, { "epoch": 2.61, "learning_rate": 1.4331613677264547e-06, "loss": 0.0004, "step": 21816 }, { "epoch": 2.61, "learning_rate": 1.4330893821235752e-06, "loss": 0.0002, "step": 21817 }, { "epoch": 2.61, "learning_rate": 1.433017396520696e-06, "loss": 0.0004, "step": 21818 }, { "epoch": 2.61, "learning_rate": 1.4329454109178166e-06, "loss": 0.0002, "step": 21819 }, { "epoch": 2.61, "learning_rate": 1.4328734253149371e-06, "loss": 0.0004, "step": 21820 }, { "epoch": 2.61, "learning_rate": 1.4328014397120577e-06, "loss": 0.0001, "step": 21821 }, { "epoch": 2.61, "learning_rate": 1.4327294541091783e-06, "loss": 0.0003, "step": 21822 }, { "epoch": 2.62, "learning_rate": 1.4326574685062986e-06, "loss": 0.0002, "step": 21823 }, { "epoch": 2.62, "learning_rate": 1.4325854829034192e-06, "loss": 0.0001, "step": 21824 }, { "epoch": 2.62, "learning_rate": 1.4325134973005398e-06, "loss": 0.0001, "step": 21825 }, { "epoch": 2.62, "learning_rate": 1.4324415116976605e-06, "loss": 0.0002, "step": 21826 }, { "epoch": 2.62, "learning_rate": 1.4323695260947811e-06, "loss": 0.0001, "step": 21827 }, { "epoch": 2.62, "learning_rate": 1.4322975404919017e-06, "loss": 0.0005, "step": 21828 }, { "epoch": 2.62, "learning_rate": 1.4322255548890222e-06, "loss": 0.0003, "step": 21829 }, { "epoch": 2.62, "learning_rate": 1.4321535692861428e-06, "loss": 0.0002, "step": 21830 }, { "epoch": 2.62, "learning_rate": 1.4320815836832634e-06, "loss": 0.0001, "step": 21831 }, { "epoch": 2.62, "learning_rate": 1.432009598080384e-06, "loss": 0.0008, "step": 21832 }, { "epoch": 2.62, "learning_rate": 1.4319376124775045e-06, "loss": 0.0001, "step": 21833 }, { "epoch": 2.62, "learning_rate": 1.431865626874625e-06, "loss": 0.0001, "step": 21834 }, { "epoch": 2.62, "learning_rate": 1.4317936412717457e-06, "loss": 0.4312, "step": 21835 }, { "epoch": 2.62, "learning_rate": 1.4317216556688662e-06, "loss": 0.0002, "step": 21836 }, { "epoch": 2.62, "learning_rate": 1.4316496700659868e-06, "loss": 0.0001, "step": 21837 }, { "epoch": 2.62, "learning_rate": 1.4315776844631074e-06, "loss": 0.0002, "step": 21838 }, { "epoch": 2.62, "learning_rate": 1.431505698860228e-06, "loss": 0.0002, "step": 21839 }, { "epoch": 2.62, "learning_rate": 1.4314337132573485e-06, "loss": 0.0002, "step": 21840 }, { "epoch": 2.62, "learning_rate": 1.431361727654469e-06, "loss": 0.0006, "step": 21841 }, { "epoch": 2.62, "learning_rate": 1.4312897420515898e-06, "loss": 0.0001, "step": 21842 }, { "epoch": 2.62, "learning_rate": 1.4312177564487104e-06, "loss": 0.0001, "step": 21843 }, { "epoch": 2.62, "learning_rate": 1.431145770845831e-06, "loss": 0.0004, "step": 21844 }, { "epoch": 2.62, "learning_rate": 1.4310737852429515e-06, "loss": 0.0003, "step": 21845 }, { "epoch": 2.62, "learning_rate": 1.4310017996400719e-06, "loss": 0.0001, "step": 21846 }, { "epoch": 2.62, "learning_rate": 1.4309298140371925e-06, "loss": 0.0002, "step": 21847 }, { "epoch": 2.62, "learning_rate": 1.430857828434313e-06, "loss": 0.0002, "step": 21848 }, { "epoch": 2.62, "learning_rate": 1.4307858428314338e-06, "loss": 0.0002, "step": 21849 }, { "epoch": 2.62, "learning_rate": 1.4307138572285544e-06, "loss": 0.0003, "step": 21850 }, { "epoch": 2.62, "learning_rate": 1.430641871625675e-06, "loss": 0.0005, "step": 21851 }, { "epoch": 2.62, "learning_rate": 1.4305698860227955e-06, "loss": 0.0002, "step": 21852 }, { "epoch": 2.62, "learning_rate": 1.430497900419916e-06, "loss": 0.0001, "step": 21853 }, { "epoch": 2.62, "learning_rate": 1.4304259148170366e-06, "loss": 0.0002, "step": 21854 }, { "epoch": 2.62, "learning_rate": 1.4303539292141572e-06, "loss": 0.0001, "step": 21855 }, { "epoch": 2.62, "learning_rate": 1.4302819436112778e-06, "loss": 0.0001, "step": 21856 }, { "epoch": 2.62, "learning_rate": 1.4302099580083983e-06, "loss": 0.0001, "step": 21857 }, { "epoch": 2.62, "learning_rate": 1.430137972405519e-06, "loss": 0.0001, "step": 21858 }, { "epoch": 2.62, "learning_rate": 1.4300659868026395e-06, "loss": 0.0002, "step": 21859 }, { "epoch": 2.62, "learning_rate": 1.42999400119976e-06, "loss": 0.0004, "step": 21860 }, { "epoch": 2.62, "learning_rate": 1.4299220155968806e-06, "loss": 0.0002, "step": 21861 }, { "epoch": 2.62, "learning_rate": 1.4298500299940012e-06, "loss": 0.0003, "step": 21862 }, { "epoch": 2.62, "learning_rate": 1.4297780443911217e-06, "loss": 0.0008, "step": 21863 }, { "epoch": 2.62, "learning_rate": 1.4297060587882423e-06, "loss": 0.0001, "step": 21864 }, { "epoch": 2.62, "learning_rate": 1.429634073185363e-06, "loss": 0.0001, "step": 21865 }, { "epoch": 2.62, "learning_rate": 1.4295620875824836e-06, "loss": 0.0001, "step": 21866 }, { "epoch": 2.62, "learning_rate": 1.4294901019796042e-06, "loss": 0.0002, "step": 21867 }, { "epoch": 2.62, "learning_rate": 1.4294181163767248e-06, "loss": 0.8271, "step": 21868 }, { "epoch": 2.62, "learning_rate": 1.4293461307738453e-06, "loss": 0.0001, "step": 21869 }, { "epoch": 2.62, "learning_rate": 1.4292741451709657e-06, "loss": 0.0001, "step": 21870 }, { "epoch": 2.62, "learning_rate": 1.4292021595680863e-06, "loss": 0.0001, "step": 21871 }, { "epoch": 2.62, "learning_rate": 1.429130173965207e-06, "loss": 0.0001, "step": 21872 }, { "epoch": 2.62, "learning_rate": 1.4290581883623276e-06, "loss": 0.0013, "step": 21873 }, { "epoch": 2.62, "learning_rate": 1.4289862027594482e-06, "loss": 0.0002, "step": 21874 }, { "epoch": 2.62, "learning_rate": 1.4289142171565687e-06, "loss": 0.0001, "step": 21875 }, { "epoch": 2.62, "learning_rate": 1.4288422315536893e-06, "loss": 0.0001, "step": 21876 }, { "epoch": 2.62, "learning_rate": 1.4287702459508099e-06, "loss": 0.0002, "step": 21877 }, { "epoch": 2.62, "learning_rate": 1.4286982603479304e-06, "loss": 0.0009, "step": 21878 }, { "epoch": 2.62, "learning_rate": 1.428626274745051e-06, "loss": 1.0334, "step": 21879 }, { "epoch": 2.62, "learning_rate": 1.4285542891421716e-06, "loss": 0.0001, "step": 21880 }, { "epoch": 2.62, "learning_rate": 1.4284823035392921e-06, "loss": 0.0002, "step": 21881 }, { "epoch": 2.62, "learning_rate": 1.4284103179364127e-06, "loss": 0.0001, "step": 21882 }, { "epoch": 2.62, "learning_rate": 1.4283383323335333e-06, "loss": 0.0112, "step": 21883 }, { "epoch": 2.62, "learning_rate": 1.4282663467306538e-06, "loss": 0.0002, "step": 21884 }, { "epoch": 2.62, "learning_rate": 1.4281943611277744e-06, "loss": 0.0002, "step": 21885 }, { "epoch": 2.62, "learning_rate": 1.428122375524895e-06, "loss": 0.0001, "step": 21886 }, { "epoch": 2.62, "learning_rate": 1.4280503899220155e-06, "loss": 0.0004, "step": 21887 }, { "epoch": 2.62, "learning_rate": 1.4279784043191363e-06, "loss": 0.0002, "step": 21888 }, { "epoch": 2.62, "learning_rate": 1.4279064187162569e-06, "loss": 0.0001, "step": 21889 }, { "epoch": 2.62, "learning_rate": 1.4278344331133775e-06, "loss": 0.0001, "step": 21890 }, { "epoch": 2.62, "learning_rate": 1.427762447510498e-06, "loss": 0.0001, "step": 21891 }, { "epoch": 2.62, "learning_rate": 1.4276904619076186e-06, "loss": 0.0003, "step": 21892 }, { "epoch": 2.62, "learning_rate": 1.427618476304739e-06, "loss": 0.0001, "step": 21893 }, { "epoch": 2.62, "learning_rate": 1.4275464907018595e-06, "loss": 0.0001, "step": 21894 }, { "epoch": 2.62, "learning_rate": 1.4274745050989803e-06, "loss": 0.0001, "step": 21895 }, { "epoch": 2.62, "learning_rate": 1.4274025194961009e-06, "loss": 0.0001, "step": 21896 }, { "epoch": 2.62, "learning_rate": 1.4273305338932214e-06, "loss": 0.0003, "step": 21897 }, { "epoch": 2.62, "learning_rate": 1.427258548290342e-06, "loss": 0.0004, "step": 21898 }, { "epoch": 2.62, "learning_rate": 1.4271865626874626e-06, "loss": 0.0002, "step": 21899 }, { "epoch": 2.62, "learning_rate": 1.4271145770845831e-06, "loss": 0.0002, "step": 21900 }, { "epoch": 2.62, "learning_rate": 1.4270425914817037e-06, "loss": 0.0001, "step": 21901 }, { "epoch": 2.62, "learning_rate": 1.4269706058788243e-06, "loss": 0.0001, "step": 21902 }, { "epoch": 2.62, "learning_rate": 1.4268986202759448e-06, "loss": 0.0002, "step": 21903 }, { "epoch": 2.62, "learning_rate": 1.4268266346730654e-06, "loss": 0.0, "step": 21904 }, { "epoch": 2.62, "learning_rate": 1.426754649070186e-06, "loss": 0.0001, "step": 21905 }, { "epoch": 2.63, "learning_rate": 1.4266826634673065e-06, "loss": 0.0001, "step": 21906 }, { "epoch": 2.63, "learning_rate": 1.426610677864427e-06, "loss": 0.253, "step": 21907 }, { "epoch": 2.63, "learning_rate": 1.4265386922615477e-06, "loss": 0.0001, "step": 21908 }, { "epoch": 2.63, "learning_rate": 1.4264667066586682e-06, "loss": 0.0005, "step": 21909 }, { "epoch": 2.63, "learning_rate": 1.4263947210557888e-06, "loss": 0.0001, "step": 21910 }, { "epoch": 2.63, "learning_rate": 1.4263227354529096e-06, "loss": 0.0002, "step": 21911 }, { "epoch": 2.63, "learning_rate": 1.4262507498500301e-06, "loss": 0.0001, "step": 21912 }, { "epoch": 2.63, "learning_rate": 1.4261787642471507e-06, "loss": 0.0002, "step": 21913 }, { "epoch": 2.63, "learning_rate": 1.4261067786442713e-06, "loss": 0.0007, "step": 21914 }, { "epoch": 2.63, "learning_rate": 1.4260347930413918e-06, "loss": 0.0002, "step": 21915 }, { "epoch": 2.63, "learning_rate": 1.4259628074385122e-06, "loss": 0.0005, "step": 21916 }, { "epoch": 2.63, "learning_rate": 1.4258908218356328e-06, "loss": 1.0146, "step": 21917 }, { "epoch": 2.63, "learning_rate": 1.4258188362327535e-06, "loss": 0.0002, "step": 21918 }, { "epoch": 2.63, "learning_rate": 1.4257468506298741e-06, "loss": 0.0002, "step": 21919 }, { "epoch": 2.63, "learning_rate": 1.4256748650269947e-06, "loss": 0.0001, "step": 21920 }, { "epoch": 2.63, "learning_rate": 1.4256028794241152e-06, "loss": 0.0001, "step": 21921 }, { "epoch": 2.63, "learning_rate": 1.4255308938212358e-06, "loss": 0.0029, "step": 21922 }, { "epoch": 2.63, "learning_rate": 1.4254589082183564e-06, "loss": 0.0002, "step": 21923 }, { "epoch": 2.63, "learning_rate": 1.425386922615477e-06, "loss": 0.0001, "step": 21924 }, { "epoch": 2.63, "learning_rate": 1.4253149370125975e-06, "loss": 0.0002, "step": 21925 }, { "epoch": 2.63, "learning_rate": 1.425242951409718e-06, "loss": 0.0001, "step": 21926 }, { "epoch": 2.63, "learning_rate": 1.4251709658068386e-06, "loss": 0.0001, "step": 21927 }, { "epoch": 2.63, "learning_rate": 1.4250989802039592e-06, "loss": 0.0002, "step": 21928 }, { "epoch": 2.63, "learning_rate": 1.4250269946010798e-06, "loss": 0.8288, "step": 21929 }, { "epoch": 2.63, "learning_rate": 1.4249550089982003e-06, "loss": 0.0003, "step": 21930 }, { "epoch": 2.63, "learning_rate": 1.424883023395321e-06, "loss": 0.2094, "step": 21931 }, { "epoch": 2.63, "learning_rate": 1.4248110377924415e-06, "loss": 0.0003, "step": 21932 }, { "epoch": 2.63, "learning_rate": 1.424739052189562e-06, "loss": 0.0002, "step": 21933 }, { "epoch": 2.63, "learning_rate": 1.4246670665866828e-06, "loss": 0.0004, "step": 21934 }, { "epoch": 2.63, "learning_rate": 1.4245950809838034e-06, "loss": 0.0002, "step": 21935 }, { "epoch": 2.63, "learning_rate": 1.424523095380924e-06, "loss": 0.0002, "step": 21936 }, { "epoch": 2.63, "learning_rate": 1.4244511097780445e-06, "loss": 0.0001, "step": 21937 }, { "epoch": 2.63, "learning_rate": 1.424379124175165e-06, "loss": 0.0002, "step": 21938 }, { "epoch": 2.63, "learning_rate": 1.4243071385722854e-06, "loss": 0.0004, "step": 21939 }, { "epoch": 2.63, "learning_rate": 1.424235152969406e-06, "loss": 0.0002, "step": 21940 }, { "epoch": 2.63, "learning_rate": 1.4241631673665268e-06, "loss": 0.0001, "step": 21941 }, { "epoch": 2.63, "learning_rate": 1.4240911817636474e-06, "loss": 0.0001, "step": 21942 }, { "epoch": 2.63, "learning_rate": 1.424019196160768e-06, "loss": 0.0005, "step": 21943 }, { "epoch": 2.63, "learning_rate": 1.4239472105578885e-06, "loss": 0.0002, "step": 21944 }, { "epoch": 2.63, "learning_rate": 1.423875224955009e-06, "loss": 0.0001, "step": 21945 }, { "epoch": 2.63, "learning_rate": 1.4238032393521296e-06, "loss": 0.0001, "step": 21946 }, { "epoch": 2.63, "learning_rate": 1.4237312537492502e-06, "loss": 0.7671, "step": 21947 }, { "epoch": 2.63, "learning_rate": 1.4236592681463708e-06, "loss": 0.0, "step": 21948 }, { "epoch": 2.63, "learning_rate": 1.4235872825434913e-06, "loss": 0.0002, "step": 21949 }, { "epoch": 2.63, "learning_rate": 1.4235152969406119e-06, "loss": 0.0009, "step": 21950 }, { "epoch": 2.63, "learning_rate": 1.4234433113377325e-06, "loss": 0.0015, "step": 21951 }, { "epoch": 2.63, "learning_rate": 1.423371325734853e-06, "loss": 0.0005, "step": 21952 }, { "epoch": 2.63, "learning_rate": 1.4232993401319736e-06, "loss": 0.0074, "step": 21953 }, { "epoch": 2.63, "learning_rate": 1.4232273545290942e-06, "loss": 0.0001, "step": 21954 }, { "epoch": 2.63, "learning_rate": 1.4231553689262147e-06, "loss": 0.0001, "step": 21955 }, { "epoch": 2.63, "learning_rate": 1.4230833833233353e-06, "loss": 0.4724, "step": 21956 }, { "epoch": 2.63, "learning_rate": 1.423011397720456e-06, "loss": 0.0006, "step": 21957 }, { "epoch": 2.63, "learning_rate": 1.4229394121175766e-06, "loss": 0.0004, "step": 21958 }, { "epoch": 2.63, "learning_rate": 1.4228674265146972e-06, "loss": 0.0006, "step": 21959 }, { "epoch": 2.63, "learning_rate": 1.4227954409118178e-06, "loss": 0.0002, "step": 21960 }, { "epoch": 2.63, "learning_rate": 1.4227234553089383e-06, "loss": 0.0003, "step": 21961 }, { "epoch": 2.63, "learning_rate": 1.4226514697060587e-06, "loss": 0.0003, "step": 21962 }, { "epoch": 2.63, "learning_rate": 1.4225794841031793e-06, "loss": 0.0003, "step": 21963 }, { "epoch": 2.63, "learning_rate": 1.4225074985002998e-06, "loss": 0.0003, "step": 21964 }, { "epoch": 2.63, "learning_rate": 1.4224355128974206e-06, "loss": 0.0001, "step": 21965 }, { "epoch": 2.63, "learning_rate": 1.4223635272945412e-06, "loss": 0.0002, "step": 21966 }, { "epoch": 2.63, "learning_rate": 1.4222915416916617e-06, "loss": 0.0004, "step": 21967 }, { "epoch": 2.63, "learning_rate": 1.4222195560887823e-06, "loss": 0.0003, "step": 21968 }, { "epoch": 2.63, "learning_rate": 1.4221475704859029e-06, "loss": 0.0014, "step": 21969 }, { "epoch": 2.63, "learning_rate": 1.4220755848830234e-06, "loss": 0.0002, "step": 21970 }, { "epoch": 2.63, "learning_rate": 1.422003599280144e-06, "loss": 0.0016, "step": 21971 }, { "epoch": 2.63, "learning_rate": 1.4219316136772646e-06, "loss": 0.0003, "step": 21972 }, { "epoch": 2.63, "learning_rate": 1.4218596280743851e-06, "loss": 0.0003, "step": 21973 }, { "epoch": 2.63, "learning_rate": 1.4217876424715057e-06, "loss": 0.0002, "step": 21974 }, { "epoch": 2.63, "learning_rate": 1.4217156568686263e-06, "loss": 0.0003, "step": 21975 }, { "epoch": 2.63, "learning_rate": 1.4216436712657468e-06, "loss": 0.0001, "step": 21976 }, { "epoch": 2.63, "learning_rate": 1.4215716856628674e-06, "loss": 0.0001, "step": 21977 }, { "epoch": 2.63, "learning_rate": 1.421499700059988e-06, "loss": 0.0003, "step": 21978 }, { "epoch": 2.63, "learning_rate": 1.4214277144571085e-06, "loss": 0.0002, "step": 21979 }, { "epoch": 2.63, "learning_rate": 1.421355728854229e-06, "loss": 0.0003, "step": 21980 }, { "epoch": 2.63, "learning_rate": 1.4212837432513499e-06, "loss": 0.0001, "step": 21981 }, { "epoch": 2.63, "learning_rate": 1.4212117576484705e-06, "loss": 0.0006, "step": 21982 }, { "epoch": 2.63, "learning_rate": 1.421139772045591e-06, "loss": 0.0001, "step": 21983 }, { "epoch": 2.63, "learning_rate": 1.4210677864427116e-06, "loss": 0.0001, "step": 21984 }, { "epoch": 2.63, "learning_rate": 1.420995800839832e-06, "loss": 0.0003, "step": 21985 }, { "epoch": 2.63, "learning_rate": 1.4209238152369525e-06, "loss": 0.0001, "step": 21986 }, { "epoch": 2.63, "learning_rate": 1.420851829634073e-06, "loss": 0.0001, "step": 21987 }, { "epoch": 2.63, "learning_rate": 1.4207798440311939e-06, "loss": 0.0006, "step": 21988 }, { "epoch": 2.63, "learning_rate": 1.4207078584283144e-06, "loss": 0.0001, "step": 21989 }, { "epoch": 2.64, "learning_rate": 1.420635872825435e-06, "loss": 0.0003, "step": 21990 }, { "epoch": 2.64, "learning_rate": 1.4205638872225556e-06, "loss": 0.0001, "step": 21991 }, { "epoch": 2.64, "learning_rate": 1.4204919016196761e-06, "loss": 0.0001, "step": 21992 }, { "epoch": 2.64, "learning_rate": 1.4204199160167967e-06, "loss": 0.0007, "step": 21993 }, { "epoch": 2.64, "learning_rate": 1.4203479304139173e-06, "loss": 0.0001, "step": 21994 }, { "epoch": 2.64, "learning_rate": 1.4202759448110378e-06, "loss": 0.0427, "step": 21995 }, { "epoch": 2.64, "learning_rate": 1.4202039592081584e-06, "loss": 0.0003, "step": 21996 }, { "epoch": 2.64, "learning_rate": 1.420131973605279e-06, "loss": 0.0002, "step": 21997 }, { "epoch": 2.64, "learning_rate": 1.4200599880023995e-06, "loss": 0.0004, "step": 21998 }, { "epoch": 2.64, "learning_rate": 1.41998800239952e-06, "loss": 0.0015, "step": 21999 }, { "epoch": 2.64, "learning_rate": 1.4199160167966407e-06, "loss": 0.0001, "step": 22000 }, { "epoch": 2.64, "learning_rate": 1.4198440311937612e-06, "loss": 0.0002, "step": 22001 }, { "epoch": 2.64, "learning_rate": 1.4197720455908818e-06, "loss": 0.0001, "step": 22002 }, { "epoch": 2.64, "learning_rate": 1.4197000599880024e-06, "loss": 0.1492, "step": 22003 }, { "epoch": 2.64, "learning_rate": 1.4196280743851231e-06, "loss": 0.0001, "step": 22004 }, { "epoch": 2.64, "learning_rate": 1.4195560887822437e-06, "loss": 0.3101, "step": 22005 }, { "epoch": 2.64, "learning_rate": 1.4194841031793643e-06, "loss": 0.0001, "step": 22006 }, { "epoch": 2.64, "learning_rate": 1.4194121175764848e-06, "loss": 0.0001, "step": 22007 }, { "epoch": 2.64, "learning_rate": 1.4193401319736052e-06, "loss": 0.0001, "step": 22008 }, { "epoch": 2.64, "learning_rate": 1.4192681463707258e-06, "loss": 0.0002, "step": 22009 }, { "epoch": 2.64, "learning_rate": 1.4191961607678463e-06, "loss": 0.0001, "step": 22010 }, { "epoch": 2.64, "learning_rate": 1.419124175164967e-06, "loss": 0.0001, "step": 22011 }, { "epoch": 2.64, "learning_rate": 1.4190521895620877e-06, "loss": 0.0005, "step": 22012 }, { "epoch": 2.64, "learning_rate": 1.4189802039592082e-06, "loss": 0.0013, "step": 22013 }, { "epoch": 2.64, "learning_rate": 1.4189082183563288e-06, "loss": 0.0003, "step": 22014 }, { "epoch": 2.64, "learning_rate": 1.4188362327534494e-06, "loss": 0.0001, "step": 22015 }, { "epoch": 2.64, "learning_rate": 1.41876424715057e-06, "loss": 0.0003, "step": 22016 }, { "epoch": 2.64, "learning_rate": 1.4186922615476905e-06, "loss": 0.0001, "step": 22017 }, { "epoch": 2.64, "learning_rate": 1.418620275944811e-06, "loss": 0.0002, "step": 22018 }, { "epoch": 2.64, "learning_rate": 1.4185482903419316e-06, "loss": 0.0001, "step": 22019 }, { "epoch": 2.64, "learning_rate": 1.4184763047390522e-06, "loss": 0.0001, "step": 22020 }, { "epoch": 2.64, "learning_rate": 1.4184043191361728e-06, "loss": 0.0001, "step": 22021 }, { "epoch": 2.64, "learning_rate": 1.4183323335332933e-06, "loss": 0.0001, "step": 22022 }, { "epoch": 2.64, "learning_rate": 1.418260347930414e-06, "loss": 0.0002, "step": 22023 }, { "epoch": 2.64, "learning_rate": 1.4181883623275345e-06, "loss": 0.0001, "step": 22024 }, { "epoch": 2.64, "learning_rate": 1.418116376724655e-06, "loss": 0.0001, "step": 22025 }, { "epoch": 2.64, "learning_rate": 1.4180443911217756e-06, "loss": 0.0001, "step": 22026 }, { "epoch": 2.64, "learning_rate": 1.4179724055188964e-06, "loss": 0.0001, "step": 22027 }, { "epoch": 2.64, "learning_rate": 1.417900419916017e-06, "loss": 0.0002, "step": 22028 }, { "epoch": 2.64, "learning_rate": 1.4178284343131375e-06, "loss": 0.0002, "step": 22029 }, { "epoch": 2.64, "learning_rate": 1.417756448710258e-06, "loss": 0.0002, "step": 22030 }, { "epoch": 2.64, "learning_rate": 1.4176844631073784e-06, "loss": 0.0002, "step": 22031 }, { "epoch": 2.64, "learning_rate": 1.417612477504499e-06, "loss": 0.0001, "step": 22032 }, { "epoch": 2.64, "learning_rate": 1.4175404919016196e-06, "loss": 0.0001, "step": 22033 }, { "epoch": 2.64, "learning_rate": 1.4174685062987403e-06, "loss": 0.0001, "step": 22034 }, { "epoch": 2.64, "learning_rate": 1.417396520695861e-06, "loss": 0.0002, "step": 22035 }, { "epoch": 2.64, "learning_rate": 1.4173245350929815e-06, "loss": 0.0002, "step": 22036 }, { "epoch": 2.64, "learning_rate": 1.417252549490102e-06, "loss": 0.0001, "step": 22037 }, { "epoch": 2.64, "learning_rate": 1.4171805638872226e-06, "loss": 0.0001, "step": 22038 }, { "epoch": 2.64, "learning_rate": 1.4171085782843432e-06, "loss": 0.0006, "step": 22039 }, { "epoch": 2.64, "learning_rate": 1.4170365926814637e-06, "loss": 0.0001, "step": 22040 }, { "epoch": 2.64, "learning_rate": 1.4169646070785843e-06, "loss": 0.0002, "step": 22041 }, { "epoch": 2.64, "learning_rate": 1.4168926214757049e-06, "loss": 0.0003, "step": 22042 }, { "epoch": 2.64, "learning_rate": 1.4168206358728254e-06, "loss": 0.0001, "step": 22043 }, { "epoch": 2.64, "learning_rate": 1.416748650269946e-06, "loss": 0.0001, "step": 22044 }, { "epoch": 2.64, "learning_rate": 1.4166766646670666e-06, "loss": 0.0016, "step": 22045 }, { "epoch": 2.64, "learning_rate": 1.4166046790641871e-06, "loss": 0.0001, "step": 22046 }, { "epoch": 2.64, "learning_rate": 1.4165326934613077e-06, "loss": 0.0004, "step": 22047 }, { "epoch": 2.64, "learning_rate": 1.4164607078584283e-06, "loss": 0.0001, "step": 22048 }, { "epoch": 2.64, "learning_rate": 1.4163887222555488e-06, "loss": 0.0003, "step": 22049 }, { "epoch": 2.64, "learning_rate": 1.4163167366526696e-06, "loss": 0.0001, "step": 22050 }, { "epoch": 2.64, "learning_rate": 1.4162447510497902e-06, "loss": 0.0002, "step": 22051 }, { "epoch": 2.64, "learning_rate": 1.4161727654469108e-06, "loss": 0.0001, "step": 22052 }, { "epoch": 2.64, "learning_rate": 1.4161007798440313e-06, "loss": 0.0001, "step": 22053 }, { "epoch": 2.64, "learning_rate": 1.4160287942411517e-06, "loss": 0.0001, "step": 22054 }, { "epoch": 2.64, "learning_rate": 1.4159568086382723e-06, "loss": 0.0001, "step": 22055 }, { "epoch": 2.64, "learning_rate": 1.4158848230353928e-06, "loss": 0.0001, "step": 22056 }, { "epoch": 2.64, "learning_rate": 1.4158128374325136e-06, "loss": 0.0002, "step": 22057 }, { "epoch": 2.64, "learning_rate": 1.4157408518296342e-06, "loss": 0.0001, "step": 22058 }, { "epoch": 2.64, "learning_rate": 1.4156688662267547e-06, "loss": 0.0398, "step": 22059 }, { "epoch": 2.64, "learning_rate": 1.4155968806238753e-06, "loss": 0.6348, "step": 22060 }, { "epoch": 2.64, "learning_rate": 1.4155248950209959e-06, "loss": 0.0001, "step": 22061 }, { "epoch": 2.64, "learning_rate": 1.4154529094181164e-06, "loss": 0.0002, "step": 22062 }, { "epoch": 2.64, "learning_rate": 1.415380923815237e-06, "loss": 0.6036, "step": 22063 }, { "epoch": 2.64, "learning_rate": 1.4153089382123576e-06, "loss": 0.0001, "step": 22064 }, { "epoch": 2.64, "learning_rate": 1.4152369526094781e-06, "loss": 0.0006, "step": 22065 }, { "epoch": 2.64, "learning_rate": 1.4151649670065987e-06, "loss": 0.0001, "step": 22066 }, { "epoch": 2.64, "learning_rate": 1.4150929814037193e-06, "loss": 0.0001, "step": 22067 }, { "epoch": 2.64, "learning_rate": 1.4150209958008398e-06, "loss": 0.0002, "step": 22068 }, { "epoch": 2.64, "learning_rate": 1.4149490101979604e-06, "loss": 0.2941, "step": 22069 }, { "epoch": 2.64, "learning_rate": 1.414877024595081e-06, "loss": 0.0001, "step": 22070 }, { "epoch": 2.64, "learning_rate": 1.4148050389922015e-06, "loss": 0.0002, "step": 22071 }, { "epoch": 2.64, "learning_rate": 1.414733053389322e-06, "loss": 0.0003, "step": 22072 }, { "epoch": 2.65, "learning_rate": 1.4146610677864429e-06, "loss": 0.0003, "step": 22073 }, { "epoch": 2.65, "learning_rate": 1.4145890821835634e-06, "loss": 0.0002, "step": 22074 }, { "epoch": 2.65, "learning_rate": 1.414517096580684e-06, "loss": 0.0001, "step": 22075 }, { "epoch": 2.65, "learning_rate": 1.4144451109778046e-06, "loss": 0.0002, "step": 22076 }, { "epoch": 2.65, "learning_rate": 1.414373125374925e-06, "loss": 0.0001, "step": 22077 }, { "epoch": 2.65, "learning_rate": 1.4143011397720455e-06, "loss": 0.0001, "step": 22078 }, { "epoch": 2.65, "learning_rate": 1.414229154169166e-06, "loss": 0.0001, "step": 22079 }, { "epoch": 2.65, "learning_rate": 1.4141571685662868e-06, "loss": 0.0002, "step": 22080 }, { "epoch": 2.65, "learning_rate": 1.4140851829634074e-06, "loss": 0.0003, "step": 22081 }, { "epoch": 2.65, "learning_rate": 1.414013197360528e-06, "loss": 0.0001, "step": 22082 }, { "epoch": 2.65, "learning_rate": 1.4139412117576485e-06, "loss": 0.0002, "step": 22083 }, { "epoch": 2.65, "learning_rate": 1.4138692261547691e-06, "loss": 0.0002, "step": 22084 }, { "epoch": 2.65, "learning_rate": 1.4137972405518897e-06, "loss": 0.0003, "step": 22085 }, { "epoch": 2.65, "learning_rate": 1.4137252549490102e-06, "loss": 0.0009, "step": 22086 }, { "epoch": 2.65, "learning_rate": 1.4136532693461308e-06, "loss": 0.0002, "step": 22087 }, { "epoch": 2.65, "learning_rate": 1.4135812837432514e-06, "loss": 0.0001, "step": 22088 }, { "epoch": 2.65, "learning_rate": 1.413509298140372e-06, "loss": 0.0001, "step": 22089 }, { "epoch": 2.65, "learning_rate": 1.4134373125374925e-06, "loss": 0.0002, "step": 22090 }, { "epoch": 2.65, "learning_rate": 1.413365326934613e-06, "loss": 0.4541, "step": 22091 }, { "epoch": 2.65, "learning_rate": 1.4132933413317336e-06, "loss": 0.0001, "step": 22092 }, { "epoch": 2.65, "learning_rate": 1.4132213557288542e-06, "loss": 0.0017, "step": 22093 }, { "epoch": 2.65, "learning_rate": 1.4131493701259748e-06, "loss": 0.0002, "step": 22094 }, { "epoch": 2.65, "learning_rate": 1.4130773845230953e-06, "loss": 0.0005, "step": 22095 }, { "epoch": 2.65, "learning_rate": 1.413005398920216e-06, "loss": 0.0003, "step": 22096 }, { "epoch": 2.65, "learning_rate": 1.4129334133173367e-06, "loss": 0.0002, "step": 22097 }, { "epoch": 2.65, "learning_rate": 1.4128614277144573e-06, "loss": 0.0001, "step": 22098 }, { "epoch": 2.65, "learning_rate": 1.4127894421115778e-06, "loss": 0.0004, "step": 22099 }, { "epoch": 2.65, "learning_rate": 1.4127174565086982e-06, "loss": 0.0001, "step": 22100 }, { "epoch": 2.65, "learning_rate": 1.4126454709058187e-06, "loss": 0.0002, "step": 22101 }, { "epoch": 2.65, "learning_rate": 1.4125734853029393e-06, "loss": 0.0004, "step": 22102 }, { "epoch": 2.65, "learning_rate": 1.4125014997000599e-06, "loss": 0.0002, "step": 22103 }, { "epoch": 2.65, "learning_rate": 1.4124295140971807e-06, "loss": 0.0002, "step": 22104 }, { "epoch": 2.65, "learning_rate": 1.4123575284943012e-06, "loss": 1.8615, "step": 22105 }, { "epoch": 2.65, "learning_rate": 1.4122855428914218e-06, "loss": 0.0003, "step": 22106 }, { "epoch": 2.65, "learning_rate": 1.4122135572885424e-06, "loss": 0.0001, "step": 22107 }, { "epoch": 2.65, "learning_rate": 1.412141571685663e-06, "loss": 0.0001, "step": 22108 }, { "epoch": 2.65, "learning_rate": 1.4120695860827835e-06, "loss": 0.0005, "step": 22109 }, { "epoch": 2.65, "learning_rate": 1.411997600479904e-06, "loss": 0.0001, "step": 22110 }, { "epoch": 2.65, "learning_rate": 1.4119256148770246e-06, "loss": 0.0001, "step": 22111 }, { "epoch": 2.65, "learning_rate": 1.4118536292741452e-06, "loss": 0.0001, "step": 22112 }, { "epoch": 2.65, "learning_rate": 1.4117816436712658e-06, "loss": 0.0005, "step": 22113 }, { "epoch": 2.65, "learning_rate": 1.4117096580683863e-06, "loss": 0.0001, "step": 22114 }, { "epoch": 2.65, "learning_rate": 1.4116376724655069e-06, "loss": 0.0001, "step": 22115 }, { "epoch": 2.65, "learning_rate": 1.4115656868626275e-06, "loss": 0.0001, "step": 22116 }, { "epoch": 2.65, "learning_rate": 1.411493701259748e-06, "loss": 0.0003, "step": 22117 }, { "epoch": 2.65, "learning_rate": 1.4114217156568686e-06, "loss": 0.0001, "step": 22118 }, { "epoch": 2.65, "learning_rate": 1.4113497300539892e-06, "loss": 0.0002, "step": 22119 }, { "epoch": 2.65, "learning_rate": 1.41127774445111e-06, "loss": 0.0002, "step": 22120 }, { "epoch": 2.65, "learning_rate": 1.4112057588482305e-06, "loss": 0.0013, "step": 22121 }, { "epoch": 2.65, "learning_rate": 1.411133773245351e-06, "loss": 0.0047, "step": 22122 }, { "epoch": 2.65, "learning_rate": 1.4110617876424716e-06, "loss": 0.0002, "step": 22123 }, { "epoch": 2.65, "learning_rate": 1.410989802039592e-06, "loss": 0.0, "step": 22124 }, { "epoch": 2.65, "learning_rate": 1.4109178164367126e-06, "loss": 0.0001, "step": 22125 }, { "epoch": 2.65, "learning_rate": 1.4108458308338331e-06, "loss": 0.0002, "step": 22126 }, { "epoch": 2.65, "learning_rate": 1.410773845230954e-06, "loss": 0.0001, "step": 22127 }, { "epoch": 2.65, "learning_rate": 1.4107018596280745e-06, "loss": 0.0001, "step": 22128 }, { "epoch": 2.65, "learning_rate": 1.410629874025195e-06, "loss": 0.6079, "step": 22129 }, { "epoch": 2.65, "learning_rate": 1.4105578884223156e-06, "loss": 1.1145, "step": 22130 }, { "epoch": 2.65, "learning_rate": 1.4104859028194362e-06, "loss": 0.0001, "step": 22131 }, { "epoch": 2.65, "learning_rate": 1.4104139172165567e-06, "loss": 0.0001, "step": 22132 }, { "epoch": 2.65, "learning_rate": 1.4103419316136773e-06, "loss": 0.0001, "step": 22133 }, { "epoch": 2.65, "learning_rate": 1.4102699460107979e-06, "loss": 0.0003, "step": 22134 }, { "epoch": 2.65, "learning_rate": 1.4101979604079184e-06, "loss": 0.0001, "step": 22135 }, { "epoch": 2.65, "learning_rate": 1.410125974805039e-06, "loss": 0.0001, "step": 22136 }, { "epoch": 2.65, "learning_rate": 1.4100539892021596e-06, "loss": 0.0001, "step": 22137 }, { "epoch": 2.65, "learning_rate": 1.4099820035992801e-06, "loss": 0.0002, "step": 22138 }, { "epoch": 2.65, "learning_rate": 1.4099100179964007e-06, "loss": 0.0003, "step": 22139 }, { "epoch": 2.65, "learning_rate": 1.4098380323935213e-06, "loss": 0.0001, "step": 22140 }, { "epoch": 2.65, "learning_rate": 1.4097660467906418e-06, "loss": 0.0006, "step": 22141 }, { "epoch": 2.65, "learning_rate": 1.4096940611877624e-06, "loss": 0.0001, "step": 22142 }, { "epoch": 2.65, "learning_rate": 1.4096220755848832e-06, "loss": 0.8049, "step": 22143 }, { "epoch": 2.65, "learning_rate": 1.4095500899820038e-06, "loss": 0.0001, "step": 22144 }, { "epoch": 2.65, "learning_rate": 1.4094781043791243e-06, "loss": 0.0001, "step": 22145 }, { "epoch": 2.65, "learning_rate": 1.4094061187762449e-06, "loss": 0.6001, "step": 22146 }, { "epoch": 2.65, "learning_rate": 1.4093341331733652e-06, "loss": 0.0007, "step": 22147 }, { "epoch": 2.65, "learning_rate": 1.4092621475704858e-06, "loss": 0.0003, "step": 22148 }, { "epoch": 2.65, "learning_rate": 1.4091901619676064e-06, "loss": 0.0002, "step": 22149 }, { "epoch": 2.65, "learning_rate": 1.4091181763647272e-06, "loss": 0.0003, "step": 22150 }, { "epoch": 2.65, "learning_rate": 1.4090461907618477e-06, "loss": 0.0001, "step": 22151 }, { "epoch": 2.65, "learning_rate": 1.4089742051589683e-06, "loss": 0.0003, "step": 22152 }, { "epoch": 2.65, "learning_rate": 1.4089022195560889e-06, "loss": 0.0001, "step": 22153 }, { "epoch": 2.65, "learning_rate": 1.4088302339532094e-06, "loss": 0.0004, "step": 22154 }, { "epoch": 2.65, "learning_rate": 1.40875824835033e-06, "loss": 0.0002, "step": 22155 }, { "epoch": 2.66, "learning_rate": 1.4086862627474506e-06, "loss": 0.0002, "step": 22156 }, { "epoch": 2.66, "learning_rate": 1.4086142771445711e-06, "loss": 0.0001, "step": 22157 }, { "epoch": 2.66, "learning_rate": 1.4085422915416917e-06, "loss": 0.0001, "step": 22158 }, { "epoch": 2.66, "learning_rate": 1.4084703059388123e-06, "loss": 0.0003, "step": 22159 }, { "epoch": 2.66, "learning_rate": 1.4083983203359328e-06, "loss": 0.0002, "step": 22160 }, { "epoch": 2.66, "learning_rate": 1.4083263347330534e-06, "loss": 0.0001, "step": 22161 }, { "epoch": 2.66, "learning_rate": 1.408254349130174e-06, "loss": 0.0001, "step": 22162 }, { "epoch": 2.66, "learning_rate": 1.4081823635272945e-06, "loss": 0.0001, "step": 22163 }, { "epoch": 2.66, "learning_rate": 1.408110377924415e-06, "loss": 0.0006, "step": 22164 }, { "epoch": 2.66, "learning_rate": 1.4080383923215357e-06, "loss": 0.0002, "step": 22165 }, { "epoch": 2.66, "learning_rate": 1.4079664067186564e-06, "loss": 0.0002, "step": 22166 }, { "epoch": 2.66, "learning_rate": 1.407894421115777e-06, "loss": 0.0002, "step": 22167 }, { "epoch": 2.66, "learning_rate": 1.4078224355128976e-06, "loss": 0.0003, "step": 22168 }, { "epoch": 2.66, "learning_rate": 1.4077504499100181e-06, "loss": 0.0004, "step": 22169 }, { "epoch": 2.66, "learning_rate": 1.4076784643071385e-06, "loss": 0.8748, "step": 22170 }, { "epoch": 2.66, "learning_rate": 1.407606478704259e-06, "loss": 0.0001, "step": 22171 }, { "epoch": 2.66, "learning_rate": 1.4075344931013796e-06, "loss": 0.0001, "step": 22172 }, { "epoch": 2.66, "learning_rate": 1.4074625074985004e-06, "loss": 0.0001, "step": 22173 }, { "epoch": 2.66, "learning_rate": 1.407390521895621e-06, "loss": 0.0003, "step": 22174 }, { "epoch": 2.66, "learning_rate": 1.4073185362927415e-06, "loss": 0.0002, "step": 22175 }, { "epoch": 2.66, "learning_rate": 1.407246550689862e-06, "loss": 0.0001, "step": 22176 }, { "epoch": 2.66, "learning_rate": 1.4071745650869827e-06, "loss": 0.0001, "step": 22177 }, { "epoch": 2.66, "learning_rate": 1.4071025794841032e-06, "loss": 0.0001, "step": 22178 }, { "epoch": 2.66, "learning_rate": 1.4070305938812238e-06, "loss": 0.0204, "step": 22179 }, { "epoch": 2.66, "learning_rate": 1.4069586082783444e-06, "loss": 0.0001, "step": 22180 }, { "epoch": 2.66, "learning_rate": 1.406886622675465e-06, "loss": 0.0001, "step": 22181 }, { "epoch": 2.66, "learning_rate": 1.4068146370725855e-06, "loss": 0.0002, "step": 22182 }, { "epoch": 2.66, "learning_rate": 1.406742651469706e-06, "loss": 0.0001, "step": 22183 }, { "epoch": 2.66, "learning_rate": 1.4066706658668266e-06, "loss": 0.0002, "step": 22184 }, { "epoch": 2.66, "learning_rate": 1.4065986802639472e-06, "loss": 0.0005, "step": 22185 }, { "epoch": 2.66, "learning_rate": 1.4065266946610678e-06, "loss": 1.0889, "step": 22186 }, { "epoch": 2.66, "learning_rate": 1.4064547090581883e-06, "loss": 0.0001, "step": 22187 }, { "epoch": 2.66, "learning_rate": 1.406382723455309e-06, "loss": 0.0003, "step": 22188 }, { "epoch": 2.66, "learning_rate": 1.4063107378524297e-06, "loss": 0.0002, "step": 22189 }, { "epoch": 2.66, "learning_rate": 1.4062387522495502e-06, "loss": 0.0002, "step": 22190 }, { "epoch": 2.66, "learning_rate": 1.4061667666466708e-06, "loss": 0.0002, "step": 22191 }, { "epoch": 2.66, "learning_rate": 1.4060947810437914e-06, "loss": 0.0001, "step": 22192 }, { "epoch": 2.66, "learning_rate": 1.4060227954409117e-06, "loss": 0.0003, "step": 22193 }, { "epoch": 2.66, "learning_rate": 1.4059508098380323e-06, "loss": 0.0007, "step": 22194 }, { "epoch": 2.66, "learning_rate": 1.4058788242351529e-06, "loss": 0.0001, "step": 22195 }, { "epoch": 2.66, "learning_rate": 1.4058068386322737e-06, "loss": 0.0002, "step": 22196 }, { "epoch": 2.66, "learning_rate": 1.4057348530293942e-06, "loss": 0.0002, "step": 22197 }, { "epoch": 2.66, "learning_rate": 1.4056628674265148e-06, "loss": 0.0004, "step": 22198 }, { "epoch": 2.66, "learning_rate": 1.4055908818236354e-06, "loss": 0.0004, "step": 22199 }, { "epoch": 2.66, "learning_rate": 1.405518896220756e-06, "loss": 0.0002, "step": 22200 }, { "epoch": 2.66, "learning_rate": 1.4054469106178765e-06, "loss": 0.0002, "step": 22201 }, { "epoch": 2.66, "learning_rate": 1.405374925014997e-06, "loss": 0.0005, "step": 22202 }, { "epoch": 2.66, "learning_rate": 1.4053029394121176e-06, "loss": 0.0003, "step": 22203 }, { "epoch": 2.66, "learning_rate": 1.4052309538092382e-06, "loss": 0.0001, "step": 22204 }, { "epoch": 2.66, "learning_rate": 1.4051589682063588e-06, "loss": 0.0002, "step": 22205 }, { "epoch": 2.66, "learning_rate": 1.4050869826034793e-06, "loss": 0.4722, "step": 22206 }, { "epoch": 2.66, "learning_rate": 1.4050149970005999e-06, "loss": 0.0004, "step": 22207 }, { "epoch": 2.66, "learning_rate": 1.4049430113977205e-06, "loss": 0.0004, "step": 22208 }, { "epoch": 2.66, "learning_rate": 1.404871025794841e-06, "loss": 0.0002, "step": 22209 }, { "epoch": 2.66, "learning_rate": 1.4047990401919616e-06, "loss": 0.0002, "step": 22210 }, { "epoch": 2.66, "learning_rate": 1.4047270545890822e-06, "loss": 0.0002, "step": 22211 }, { "epoch": 2.66, "learning_rate": 1.404655068986203e-06, "loss": 0.0001, "step": 22212 }, { "epoch": 2.66, "learning_rate": 1.4045830833833235e-06, "loss": 0.0001, "step": 22213 }, { "epoch": 2.66, "learning_rate": 1.404511097780444e-06, "loss": 0.0002, "step": 22214 }, { "epoch": 2.66, "learning_rate": 1.4044391121775646e-06, "loss": 0.0001, "step": 22215 }, { "epoch": 2.66, "learning_rate": 1.404367126574685e-06, "loss": 0.0002, "step": 22216 }, { "epoch": 2.66, "learning_rate": 1.4042951409718056e-06, "loss": 0.0004, "step": 22217 }, { "epoch": 2.66, "learning_rate": 1.4042231553689261e-06, "loss": 0.0002, "step": 22218 }, { "epoch": 2.66, "learning_rate": 1.404151169766047e-06, "loss": 0.0002, "step": 22219 }, { "epoch": 2.66, "learning_rate": 1.4040791841631675e-06, "loss": 0.0002, "step": 22220 }, { "epoch": 2.66, "learning_rate": 1.404007198560288e-06, "loss": 0.0001, "step": 22221 }, { "epoch": 2.66, "learning_rate": 1.4039352129574086e-06, "loss": 0.0002, "step": 22222 }, { "epoch": 2.66, "learning_rate": 1.4038632273545292e-06, "loss": 0.0003, "step": 22223 }, { "epoch": 2.66, "learning_rate": 1.4037912417516497e-06, "loss": 0.0001, "step": 22224 }, { "epoch": 2.66, "learning_rate": 1.4037192561487703e-06, "loss": 0.0002, "step": 22225 }, { "epoch": 2.66, "learning_rate": 1.4036472705458909e-06, "loss": 0.0, "step": 22226 }, { "epoch": 2.66, "learning_rate": 1.4035752849430114e-06, "loss": 0.0004, "step": 22227 }, { "epoch": 2.66, "learning_rate": 1.403503299340132e-06, "loss": 0.0004, "step": 22228 }, { "epoch": 2.66, "learning_rate": 1.4034313137372526e-06, "loss": 0.0004, "step": 22229 }, { "epoch": 2.66, "learning_rate": 1.4033593281343731e-06, "loss": 0.0001, "step": 22230 }, { "epoch": 2.66, "learning_rate": 1.4032873425314937e-06, "loss": 0.0001, "step": 22231 }, { "epoch": 2.66, "learning_rate": 1.4032153569286143e-06, "loss": 0.0002, "step": 22232 }, { "epoch": 2.66, "learning_rate": 1.4031433713257348e-06, "loss": 0.0006, "step": 22233 }, { "epoch": 2.66, "learning_rate": 1.4030713857228554e-06, "loss": 0.0002, "step": 22234 }, { "epoch": 2.66, "learning_rate": 1.402999400119976e-06, "loss": 0.0007, "step": 22235 }, { "epoch": 2.66, "learning_rate": 1.4029274145170967e-06, "loss": 0.0004, "step": 22236 }, { "epoch": 2.66, "learning_rate": 1.4028554289142173e-06, "loss": 0.0002, "step": 22237 }, { "epoch": 2.66, "learning_rate": 1.4027834433113379e-06, "loss": 0.0006, "step": 22238 }, { "epoch": 2.66, "learning_rate": 1.4027114577084582e-06, "loss": 0.4087, "step": 22239 }, { "epoch": 2.67, "learning_rate": 1.4026394721055788e-06, "loss": 0.0003, "step": 22240 }, { "epoch": 2.67, "learning_rate": 1.4025674865026994e-06, "loss": 0.0001, "step": 22241 }, { "epoch": 2.67, "learning_rate": 1.40249550089982e-06, "loss": 0.3255, "step": 22242 }, { "epoch": 2.67, "learning_rate": 1.4024235152969407e-06, "loss": 0.0002, "step": 22243 }, { "epoch": 2.67, "learning_rate": 1.4023515296940613e-06, "loss": 0.0002, "step": 22244 }, { "epoch": 2.67, "learning_rate": 1.4022795440911818e-06, "loss": 0.0001, "step": 22245 }, { "epoch": 2.67, "learning_rate": 1.4022075584883024e-06, "loss": 0.0003, "step": 22246 }, { "epoch": 2.67, "learning_rate": 1.402135572885423e-06, "loss": 0.0002, "step": 22247 }, { "epoch": 2.67, "learning_rate": 1.4020635872825435e-06, "loss": 0.0002, "step": 22248 }, { "epoch": 2.67, "learning_rate": 1.4019916016796641e-06, "loss": 0.0003, "step": 22249 }, { "epoch": 2.67, "learning_rate": 1.4019196160767847e-06, "loss": 0.0003, "step": 22250 }, { "epoch": 2.67, "learning_rate": 1.4018476304739052e-06, "loss": 0.0005, "step": 22251 }, { "epoch": 2.67, "learning_rate": 1.4017756448710258e-06, "loss": 0.0003, "step": 22252 }, { "epoch": 2.67, "learning_rate": 1.4017036592681464e-06, "loss": 0.0002, "step": 22253 }, { "epoch": 2.67, "learning_rate": 1.401631673665267e-06, "loss": 0.0004, "step": 22254 }, { "epoch": 2.67, "learning_rate": 1.4015596880623875e-06, "loss": 0.0002, "step": 22255 }, { "epoch": 2.67, "learning_rate": 1.401487702459508e-06, "loss": 0.0001, "step": 22256 }, { "epoch": 2.67, "learning_rate": 1.4014157168566286e-06, "loss": 0.0002, "step": 22257 }, { "epoch": 2.67, "learning_rate": 1.4013437312537492e-06, "loss": 0.0031, "step": 22258 }, { "epoch": 2.67, "learning_rate": 1.40127174565087e-06, "loss": 0.0002, "step": 22259 }, { "epoch": 2.67, "learning_rate": 1.4011997600479906e-06, "loss": 0.0002, "step": 22260 }, { "epoch": 2.67, "learning_rate": 1.4011277744451111e-06, "loss": 0.0004, "step": 22261 }, { "epoch": 2.67, "learning_rate": 1.4010557888422315e-06, "loss": 0.0842, "step": 22262 }, { "epoch": 2.67, "learning_rate": 1.400983803239352e-06, "loss": 0.0004, "step": 22263 }, { "epoch": 2.67, "learning_rate": 1.4009118176364726e-06, "loss": 0.0001, "step": 22264 }, { "epoch": 2.67, "learning_rate": 1.4008398320335932e-06, "loss": 0.0001, "step": 22265 }, { "epoch": 2.67, "learning_rate": 1.400767846430714e-06, "loss": 0.0001, "step": 22266 }, { "epoch": 2.67, "learning_rate": 1.4006958608278345e-06, "loss": 0.0002, "step": 22267 }, { "epoch": 2.67, "learning_rate": 1.400623875224955e-06, "loss": 0.0002, "step": 22268 }, { "epoch": 2.67, "learning_rate": 1.4005518896220757e-06, "loss": 0.0001, "step": 22269 }, { "epoch": 2.67, "learning_rate": 1.4004799040191962e-06, "loss": 0.0002, "step": 22270 }, { "epoch": 2.67, "learning_rate": 1.4004079184163168e-06, "loss": 0.0003, "step": 22271 }, { "epoch": 2.67, "learning_rate": 1.4003359328134374e-06, "loss": 0.0002, "step": 22272 }, { "epoch": 2.67, "learning_rate": 1.400263947210558e-06, "loss": 0.0002, "step": 22273 }, { "epoch": 2.67, "learning_rate": 1.4001919616076785e-06, "loss": 0.0001, "step": 22274 }, { "epoch": 2.67, "learning_rate": 1.400119976004799e-06, "loss": 0.0001, "step": 22275 }, { "epoch": 2.67, "learning_rate": 1.4000479904019196e-06, "loss": 0.0002, "step": 22276 }, { "epoch": 2.67, "learning_rate": 1.3999760047990402e-06, "loss": 0.0003, "step": 22277 }, { "epoch": 2.67, "learning_rate": 1.3999040191961608e-06, "loss": 0.0001, "step": 22278 }, { "epoch": 2.67, "learning_rate": 1.3998320335932813e-06, "loss": 0.0002, "step": 22279 }, { "epoch": 2.67, "learning_rate": 1.399760047990402e-06, "loss": 0.0001, "step": 22280 }, { "epoch": 2.67, "learning_rate": 1.3996880623875225e-06, "loss": 0.0002, "step": 22281 }, { "epoch": 2.67, "learning_rate": 1.3996160767846432e-06, "loss": 0.0002, "step": 22282 }, { "epoch": 2.67, "learning_rate": 1.3995440911817638e-06, "loss": 0.0003, "step": 22283 }, { "epoch": 2.67, "learning_rate": 1.3994721055788844e-06, "loss": 0.0002, "step": 22284 }, { "epoch": 2.67, "learning_rate": 1.3994001199760047e-06, "loss": 0.0005, "step": 22285 }, { "epoch": 2.67, "learning_rate": 1.3993281343731253e-06, "loss": 0.0001, "step": 22286 }, { "epoch": 2.67, "learning_rate": 1.3992561487702459e-06, "loss": 0.0001, "step": 22287 }, { "epoch": 2.67, "learning_rate": 1.3991841631673664e-06, "loss": 0.0002, "step": 22288 }, { "epoch": 2.67, "learning_rate": 1.3991121775644872e-06, "loss": 0.0003, "step": 22289 }, { "epoch": 2.67, "learning_rate": 1.3990401919616078e-06, "loss": 0.0002, "step": 22290 }, { "epoch": 2.67, "learning_rate": 1.3989682063587283e-06, "loss": 0.0002, "step": 22291 }, { "epoch": 2.67, "learning_rate": 1.398896220755849e-06, "loss": 0.0002, "step": 22292 }, { "epoch": 2.67, "learning_rate": 1.3988242351529695e-06, "loss": 0.0002, "step": 22293 }, { "epoch": 2.67, "learning_rate": 1.39875224955009e-06, "loss": 0.0001, "step": 22294 }, { "epoch": 2.67, "learning_rate": 1.3986802639472106e-06, "loss": 0.0006, "step": 22295 }, { "epoch": 2.67, "learning_rate": 1.3986082783443312e-06, "loss": 0.0001, "step": 22296 }, { "epoch": 2.67, "learning_rate": 1.3985362927414517e-06, "loss": 0.0001, "step": 22297 }, { "epoch": 2.67, "learning_rate": 1.3984643071385723e-06, "loss": 0.0764, "step": 22298 }, { "epoch": 2.67, "learning_rate": 1.3983923215356929e-06, "loss": 0.0002, "step": 22299 }, { "epoch": 2.67, "learning_rate": 1.3983203359328134e-06, "loss": 0.0001, "step": 22300 }, { "epoch": 2.67, "learning_rate": 1.398248350329934e-06, "loss": 0.0009, "step": 22301 }, { "epoch": 2.67, "learning_rate": 1.3981763647270546e-06, "loss": 0.0003, "step": 22302 }, { "epoch": 2.67, "learning_rate": 1.3981043791241751e-06, "loss": 0.0001, "step": 22303 }, { "epoch": 2.67, "learning_rate": 1.3980323935212957e-06, "loss": 0.0002, "step": 22304 }, { "epoch": 2.67, "learning_rate": 1.3979604079184165e-06, "loss": 0.0006, "step": 22305 }, { "epoch": 2.67, "learning_rate": 1.397888422315537e-06, "loss": 0.0008, "step": 22306 }, { "epoch": 2.67, "learning_rate": 1.3978164367126576e-06, "loss": 0.0002, "step": 22307 }, { "epoch": 2.67, "learning_rate": 1.397744451109778e-06, "loss": 0.0002, "step": 22308 }, { "epoch": 2.67, "learning_rate": 1.3976724655068985e-06, "loss": 0.0002, "step": 22309 }, { "epoch": 2.67, "learning_rate": 1.3976004799040191e-06, "loss": 0.0002, "step": 22310 }, { "epoch": 2.67, "learning_rate": 1.3975284943011397e-06, "loss": 0.0001, "step": 22311 }, { "epoch": 2.67, "learning_rate": 1.3974565086982605e-06, "loss": 0.0013, "step": 22312 }, { "epoch": 2.67, "learning_rate": 1.397384523095381e-06, "loss": 0.0002, "step": 22313 }, { "epoch": 2.67, "learning_rate": 1.3973125374925016e-06, "loss": 0.0001, "step": 22314 }, { "epoch": 2.67, "learning_rate": 1.3972405518896222e-06, "loss": 0.0001, "step": 22315 }, { "epoch": 2.67, "learning_rate": 1.3971685662867427e-06, "loss": 0.0002, "step": 22316 }, { "epoch": 2.67, "learning_rate": 1.3970965806838633e-06, "loss": 0.0001, "step": 22317 }, { "epoch": 2.67, "learning_rate": 1.3970245950809839e-06, "loss": 0.0003, "step": 22318 }, { "epoch": 2.67, "learning_rate": 1.3969526094781044e-06, "loss": 0.0001, "step": 22319 }, { "epoch": 2.67, "learning_rate": 1.396880623875225e-06, "loss": 0.0002, "step": 22320 }, { "epoch": 2.67, "learning_rate": 1.3968086382723456e-06, "loss": 0.0003, "step": 22321 }, { "epoch": 2.67, "learning_rate": 1.3967366526694661e-06, "loss": 0.0003, "step": 22322 }, { "epoch": 2.68, "learning_rate": 1.3966646670665867e-06, "loss": 0.0004, "step": 22323 }, { "epoch": 2.68, "learning_rate": 1.3965926814637073e-06, "loss": 0.0002, "step": 22324 }, { "epoch": 2.68, "learning_rate": 1.3965206958608278e-06, "loss": 0.0003, "step": 22325 }, { "epoch": 2.68, "learning_rate": 1.3964487102579484e-06, "loss": 0.0008, "step": 22326 }, { "epoch": 2.68, "learning_rate": 1.396376724655069e-06, "loss": 0.0003, "step": 22327 }, { "epoch": 2.68, "learning_rate": 1.3963047390521897e-06, "loss": 0.0002, "step": 22328 }, { "epoch": 2.68, "learning_rate": 1.3962327534493103e-06, "loss": 0.0001, "step": 22329 }, { "epoch": 2.68, "learning_rate": 1.3961607678464309e-06, "loss": 0.0001, "step": 22330 }, { "epoch": 2.68, "learning_rate": 1.3960887822435512e-06, "loss": 0.0001, "step": 22331 }, { "epoch": 2.68, "learning_rate": 1.3960167966406718e-06, "loss": 0.0009, "step": 22332 }, { "epoch": 2.68, "learning_rate": 1.3959448110377924e-06, "loss": 0.0002, "step": 22333 }, { "epoch": 2.68, "learning_rate": 1.395872825434913e-06, "loss": 0.0002, "step": 22334 }, { "epoch": 2.68, "learning_rate": 1.3958008398320337e-06, "loss": 0.0003, "step": 22335 }, { "epoch": 2.68, "learning_rate": 1.3957288542291543e-06, "loss": 0.0002, "step": 22336 }, { "epoch": 2.68, "learning_rate": 1.3956568686262748e-06, "loss": 0.0003, "step": 22337 }, { "epoch": 2.68, "learning_rate": 1.3955848830233954e-06, "loss": 0.0001, "step": 22338 }, { "epoch": 2.68, "learning_rate": 1.395512897420516e-06, "loss": 0.0001, "step": 22339 }, { "epoch": 2.68, "learning_rate": 1.3954409118176365e-06, "loss": 0.0001, "step": 22340 }, { "epoch": 2.68, "learning_rate": 1.3953689262147571e-06, "loss": 0.0003, "step": 22341 }, { "epoch": 2.68, "learning_rate": 1.3952969406118777e-06, "loss": 0.0001, "step": 22342 }, { "epoch": 2.68, "learning_rate": 1.3952249550089982e-06, "loss": 0.0001, "step": 22343 }, { "epoch": 2.68, "learning_rate": 1.3951529694061188e-06, "loss": 0.0002, "step": 22344 }, { "epoch": 2.68, "learning_rate": 1.3950809838032394e-06, "loss": 0.0002, "step": 22345 }, { "epoch": 2.68, "learning_rate": 1.39500899820036e-06, "loss": 0.0002, "step": 22346 }, { "epoch": 2.68, "learning_rate": 1.3949370125974805e-06, "loss": 0.0001, "step": 22347 }, { "epoch": 2.68, "learning_rate": 1.394865026994601e-06, "loss": 0.0001, "step": 22348 }, { "epoch": 2.68, "learning_rate": 1.3947930413917216e-06, "loss": 0.0001, "step": 22349 }, { "epoch": 2.68, "learning_rate": 1.3947210557888422e-06, "loss": 0.0003, "step": 22350 }, { "epoch": 2.68, "learning_rate": 1.394649070185963e-06, "loss": 0.0002, "step": 22351 }, { "epoch": 2.68, "learning_rate": 1.3945770845830836e-06, "loss": 0.0002, "step": 22352 }, { "epoch": 2.68, "learning_rate": 1.3945050989802041e-06, "loss": 0.0001, "step": 22353 }, { "epoch": 2.68, "learning_rate": 1.3944331133773247e-06, "loss": 0.0002, "step": 22354 }, { "epoch": 2.68, "learning_rate": 1.394361127774445e-06, "loss": 0.0001, "step": 22355 }, { "epoch": 2.68, "learning_rate": 1.3942891421715656e-06, "loss": 0.0138, "step": 22356 }, { "epoch": 2.68, "learning_rate": 1.3942171565686862e-06, "loss": 0.0004, "step": 22357 }, { "epoch": 2.68, "learning_rate": 1.394145170965807e-06, "loss": 0.0001, "step": 22358 }, { "epoch": 2.68, "learning_rate": 1.3940731853629275e-06, "loss": 0.0006, "step": 22359 }, { "epoch": 2.68, "learning_rate": 1.394001199760048e-06, "loss": 0.0001, "step": 22360 }, { "epoch": 2.68, "learning_rate": 1.3939292141571687e-06, "loss": 0.0001, "step": 22361 }, { "epoch": 2.68, "learning_rate": 1.3938572285542892e-06, "loss": 0.0002, "step": 22362 }, { "epoch": 2.68, "learning_rate": 1.3937852429514098e-06, "loss": 0.0003, "step": 22363 }, { "epoch": 2.68, "learning_rate": 1.3937132573485304e-06, "loss": 0.0003, "step": 22364 }, { "epoch": 2.68, "learning_rate": 1.393641271745651e-06, "loss": 0.0002, "step": 22365 }, { "epoch": 2.68, "learning_rate": 1.3935692861427715e-06, "loss": 0.0001, "step": 22366 }, { "epoch": 2.68, "learning_rate": 1.393497300539892e-06, "loss": 0.0001, "step": 22367 }, { "epoch": 2.68, "learning_rate": 1.3934253149370126e-06, "loss": 0.0001, "step": 22368 }, { "epoch": 2.68, "learning_rate": 1.3933533293341332e-06, "loss": 0.0002, "step": 22369 }, { "epoch": 2.68, "learning_rate": 1.3932813437312538e-06, "loss": 0.0005, "step": 22370 }, { "epoch": 2.68, "learning_rate": 1.3932093581283743e-06, "loss": 0.0003, "step": 22371 }, { "epoch": 2.68, "learning_rate": 1.3931373725254949e-06, "loss": 0.9939, "step": 22372 }, { "epoch": 2.68, "learning_rate": 1.3930653869226155e-06, "loss": 0.0003, "step": 22373 }, { "epoch": 2.68, "learning_rate": 1.392993401319736e-06, "loss": 0.0005, "step": 22374 }, { "epoch": 2.68, "learning_rate": 1.3929214157168568e-06, "loss": 0.0003, "step": 22375 }, { "epoch": 2.68, "learning_rate": 1.3928494301139774e-06, "loss": 0.0001, "step": 22376 }, { "epoch": 2.68, "learning_rate": 1.392777444511098e-06, "loss": 0.0002, "step": 22377 }, { "epoch": 2.68, "learning_rate": 1.3927054589082183e-06, "loss": 0.0002, "step": 22378 }, { "epoch": 2.68, "learning_rate": 1.3926334733053389e-06, "loss": 0.0002, "step": 22379 }, { "epoch": 2.68, "learning_rate": 1.3925614877024594e-06, "loss": 0.0001, "step": 22380 }, { "epoch": 2.68, "learning_rate": 1.39248950209958e-06, "loss": 0.0002, "step": 22381 }, { "epoch": 2.68, "learning_rate": 1.3924175164967008e-06, "loss": 0.0001, "step": 22382 }, { "epoch": 2.68, "learning_rate": 1.3923455308938213e-06, "loss": 0.0003, "step": 22383 }, { "epoch": 2.68, "learning_rate": 1.392273545290942e-06, "loss": 0.0001, "step": 22384 }, { "epoch": 2.68, "learning_rate": 1.3922015596880625e-06, "loss": 0.0001, "step": 22385 }, { "epoch": 2.68, "learning_rate": 1.392129574085183e-06, "loss": 0.0002, "step": 22386 }, { "epoch": 2.68, "learning_rate": 1.3920575884823036e-06, "loss": 0.0001, "step": 22387 }, { "epoch": 2.68, "learning_rate": 1.3919856028794242e-06, "loss": 0.0002, "step": 22388 }, { "epoch": 2.68, "learning_rate": 1.3919136172765447e-06, "loss": 0.0001, "step": 22389 }, { "epoch": 2.68, "learning_rate": 1.3918416316736653e-06, "loss": 0.0001, "step": 22390 }, { "epoch": 2.68, "learning_rate": 1.3917696460707859e-06, "loss": 0.0001, "step": 22391 }, { "epoch": 2.68, "learning_rate": 1.3916976604679064e-06, "loss": 0.0003, "step": 22392 }, { "epoch": 2.68, "learning_rate": 1.391625674865027e-06, "loss": 0.0007, "step": 22393 }, { "epoch": 2.68, "learning_rate": 1.3915536892621476e-06, "loss": 0.0002, "step": 22394 }, { "epoch": 2.68, "learning_rate": 1.3914817036592681e-06, "loss": 0.0004, "step": 22395 }, { "epoch": 2.68, "learning_rate": 1.3914097180563887e-06, "loss": 0.0001, "step": 22396 }, { "epoch": 2.68, "learning_rate": 1.3913377324535093e-06, "loss": 0.0002, "step": 22397 }, { "epoch": 2.68, "learning_rate": 1.39126574685063e-06, "loss": 0.0003, "step": 22398 }, { "epoch": 2.68, "learning_rate": 1.3911937612477506e-06, "loss": 0.0003, "step": 22399 }, { "epoch": 2.68, "learning_rate": 1.3911217756448712e-06, "loss": 0.0001, "step": 22400 }, { "epoch": 2.68, "learning_rate": 1.3910497900419915e-06, "loss": 0.0002, "step": 22401 }, { "epoch": 2.68, "learning_rate": 1.390977804439112e-06, "loss": 0.0002, "step": 22402 }, { "epoch": 2.68, "learning_rate": 1.3909058188362327e-06, "loss": 0.0003, "step": 22403 }, { "epoch": 2.68, "learning_rate": 1.3908338332333532e-06, "loss": 0.0001, "step": 22404 }, { "epoch": 2.68, "learning_rate": 1.390761847630474e-06, "loss": 0.0002, "step": 22405 }, { "epoch": 2.68, "learning_rate": 1.3906898620275946e-06, "loss": 0.0002, "step": 22406 }, { "epoch": 2.69, "learning_rate": 1.3906178764247151e-06, "loss": 0.0001, "step": 22407 }, { "epoch": 2.69, "learning_rate": 1.3905458908218357e-06, "loss": 0.0002, "step": 22408 }, { "epoch": 2.69, "learning_rate": 1.3904739052189563e-06, "loss": 0.0001, "step": 22409 }, { "epoch": 2.69, "learning_rate": 1.3904019196160769e-06, "loss": 0.0002, "step": 22410 }, { "epoch": 2.69, "learning_rate": 1.3903299340131974e-06, "loss": 0.2947, "step": 22411 }, { "epoch": 2.69, "learning_rate": 1.390257948410318e-06, "loss": 0.0006, "step": 22412 }, { "epoch": 2.69, "learning_rate": 1.3901859628074386e-06, "loss": 0.2255, "step": 22413 }, { "epoch": 2.69, "learning_rate": 1.3901139772045591e-06, "loss": 0.0004, "step": 22414 }, { "epoch": 2.69, "learning_rate": 1.3900419916016797e-06, "loss": 0.0002, "step": 22415 }, { "epoch": 2.69, "learning_rate": 1.3899700059988003e-06, "loss": 0.8464, "step": 22416 }, { "epoch": 2.69, "learning_rate": 1.3898980203959208e-06, "loss": 0.0002, "step": 22417 }, { "epoch": 2.69, "learning_rate": 1.3898260347930414e-06, "loss": 0.0002, "step": 22418 }, { "epoch": 2.69, "learning_rate": 1.389754049190162e-06, "loss": 0.0005, "step": 22419 }, { "epoch": 2.69, "learning_rate": 1.3896820635872825e-06, "loss": 0.0002, "step": 22420 }, { "epoch": 2.69, "learning_rate": 1.3896100779844033e-06, "loss": 0.0001, "step": 22421 }, { "epoch": 2.69, "learning_rate": 1.3895380923815239e-06, "loss": 0.0001, "step": 22422 }, { "epoch": 2.69, "learning_rate": 1.3894661067786444e-06, "loss": 0.0002, "step": 22423 }, { "epoch": 2.69, "learning_rate": 1.3893941211757648e-06, "loss": 0.0001, "step": 22424 }, { "epoch": 2.69, "learning_rate": 1.3893221355728854e-06, "loss": 0.0001, "step": 22425 }, { "epoch": 2.69, "learning_rate": 1.389250149970006e-06, "loss": 0.0005, "step": 22426 }, { "epoch": 2.69, "learning_rate": 1.3891781643671265e-06, "loss": 0.0005, "step": 22427 }, { "epoch": 2.69, "learning_rate": 1.3891061787642473e-06, "loss": 0.0001, "step": 22428 }, { "epoch": 2.69, "learning_rate": 1.3890341931613678e-06, "loss": 0.0002, "step": 22429 }, { "epoch": 2.69, "learning_rate": 1.3889622075584884e-06, "loss": 0.0002, "step": 22430 }, { "epoch": 2.69, "learning_rate": 1.388890221955609e-06, "loss": 0.0002, "step": 22431 }, { "epoch": 2.69, "learning_rate": 1.3888182363527295e-06, "loss": 0.0001, "step": 22432 }, { "epoch": 2.69, "learning_rate": 1.38874625074985e-06, "loss": 0.0002, "step": 22433 }, { "epoch": 2.69, "learning_rate": 1.3886742651469707e-06, "loss": 0.0001, "step": 22434 }, { "epoch": 2.69, "learning_rate": 1.3886022795440912e-06, "loss": 0.0002, "step": 22435 }, { "epoch": 2.69, "learning_rate": 1.3885302939412118e-06, "loss": 0.0001, "step": 22436 }, { "epoch": 2.69, "learning_rate": 1.3884583083383324e-06, "loss": 0.0001, "step": 22437 }, { "epoch": 2.69, "learning_rate": 1.388386322735453e-06, "loss": 0.0002, "step": 22438 }, { "epoch": 2.69, "learning_rate": 1.3883143371325735e-06, "loss": 0.0001, "step": 22439 }, { "epoch": 2.69, "learning_rate": 1.388242351529694e-06, "loss": 0.0006, "step": 22440 }, { "epoch": 2.69, "learning_rate": 1.3881703659268146e-06, "loss": 0.0001, "step": 22441 }, { "epoch": 2.69, "learning_rate": 1.3880983803239352e-06, "loss": 0.0004, "step": 22442 }, { "epoch": 2.69, "learning_rate": 1.3880263947210558e-06, "loss": 0.0002, "step": 22443 }, { "epoch": 2.69, "learning_rate": 1.3879544091181765e-06, "loss": 0.0001, "step": 22444 }, { "epoch": 2.69, "learning_rate": 1.3878824235152971e-06, "loss": 0.0001, "step": 22445 }, { "epoch": 2.69, "learning_rate": 1.3878104379124177e-06, "loss": 0.0001, "step": 22446 }, { "epoch": 2.69, "learning_rate": 1.387738452309538e-06, "loss": 0.0006, "step": 22447 }, { "epoch": 2.69, "learning_rate": 1.3876664667066586e-06, "loss": 0.0001, "step": 22448 }, { "epoch": 2.69, "learning_rate": 1.3875944811037792e-06, "loss": 0.0001, "step": 22449 }, { "epoch": 2.69, "learning_rate": 1.3875224955008997e-06, "loss": 0.0001, "step": 22450 }, { "epoch": 2.69, "learning_rate": 1.3874505098980205e-06, "loss": 0.0003, "step": 22451 }, { "epoch": 2.69, "learning_rate": 1.387378524295141e-06, "loss": 0.0001, "step": 22452 }, { "epoch": 2.69, "learning_rate": 1.3873065386922616e-06, "loss": 0.0001, "step": 22453 }, { "epoch": 2.69, "learning_rate": 1.3872345530893822e-06, "loss": 0.0003, "step": 22454 }, { "epoch": 2.69, "learning_rate": 1.3871625674865028e-06, "loss": 0.0001, "step": 22455 }, { "epoch": 2.69, "learning_rate": 1.3870905818836233e-06, "loss": 0.0002, "step": 22456 }, { "epoch": 2.69, "learning_rate": 1.387018596280744e-06, "loss": 0.0002, "step": 22457 }, { "epoch": 2.69, "learning_rate": 1.3869466106778645e-06, "loss": 0.0001, "step": 22458 }, { "epoch": 2.69, "learning_rate": 1.386874625074985e-06, "loss": 0.0001, "step": 22459 }, { "epoch": 2.69, "learning_rate": 1.3868026394721056e-06, "loss": 0.0001, "step": 22460 }, { "epoch": 2.69, "learning_rate": 1.3867306538692262e-06, "loss": 0.0001, "step": 22461 }, { "epoch": 2.69, "learning_rate": 1.3866586682663467e-06, "loss": 0.0002, "step": 22462 }, { "epoch": 2.69, "learning_rate": 1.3865866826634673e-06, "loss": 0.0001, "step": 22463 }, { "epoch": 2.69, "learning_rate": 1.3865146970605879e-06, "loss": 0.0001, "step": 22464 }, { "epoch": 2.69, "learning_rate": 1.3864427114577084e-06, "loss": 0.0002, "step": 22465 }, { "epoch": 2.69, "learning_rate": 1.386370725854829e-06, "loss": 0.0001, "step": 22466 }, { "epoch": 2.69, "learning_rate": 1.3862987402519498e-06, "loss": 0.0002, "step": 22467 }, { "epoch": 2.69, "learning_rate": 1.3862267546490704e-06, "loss": 0.0002, "step": 22468 }, { "epoch": 2.69, "learning_rate": 1.386154769046191e-06, "loss": 0.0004, "step": 22469 }, { "epoch": 2.69, "learning_rate": 1.3860827834433113e-06, "loss": 0.0002, "step": 22470 }, { "epoch": 2.69, "learning_rate": 1.3860107978404318e-06, "loss": 0.0019, "step": 22471 }, { "epoch": 2.69, "learning_rate": 1.3859388122375524e-06, "loss": 0.0002, "step": 22472 }, { "epoch": 2.69, "learning_rate": 1.385866826634673e-06, "loss": 0.0001, "step": 22473 }, { "epoch": 2.69, "learning_rate": 1.3857948410317938e-06, "loss": 0.0001, "step": 22474 }, { "epoch": 2.69, "learning_rate": 1.3857228554289143e-06, "loss": 0.0003, "step": 22475 }, { "epoch": 2.69, "learning_rate": 1.3856508698260349e-06, "loss": 0.0001, "step": 22476 }, { "epoch": 2.69, "learning_rate": 1.3855788842231555e-06, "loss": 0.0001, "step": 22477 }, { "epoch": 2.69, "learning_rate": 1.385506898620276e-06, "loss": 0.0001, "step": 22478 }, { "epoch": 2.69, "learning_rate": 1.3854349130173966e-06, "loss": 0.0001, "step": 22479 }, { "epoch": 2.69, "learning_rate": 1.3853629274145172e-06, "loss": 0.0044, "step": 22480 }, { "epoch": 2.69, "learning_rate": 1.3852909418116377e-06, "loss": 0.0003, "step": 22481 }, { "epoch": 2.69, "learning_rate": 1.3852189562087583e-06, "loss": 0.0001, "step": 22482 }, { "epoch": 2.69, "learning_rate": 1.3851469706058789e-06, "loss": 0.0001, "step": 22483 }, { "epoch": 2.69, "learning_rate": 1.3850749850029994e-06, "loss": 0.0001, "step": 22484 }, { "epoch": 2.69, "learning_rate": 1.38500299940012e-06, "loss": 0.0005, "step": 22485 }, { "epoch": 2.69, "learning_rate": 1.3849310137972406e-06, "loss": 0.0001, "step": 22486 }, { "epoch": 2.69, "learning_rate": 1.3848590281943611e-06, "loss": 0.0001, "step": 22487 }, { "epoch": 2.69, "learning_rate": 1.3847870425914817e-06, "loss": 0.0002, "step": 22488 }, { "epoch": 2.69, "learning_rate": 1.3847150569886023e-06, "loss": 0.0001, "step": 22489 }, { "epoch": 2.7, "learning_rate": 1.384643071385723e-06, "loss": 0.0001, "step": 22490 }, { "epoch": 2.7, "learning_rate": 1.3845710857828436e-06, "loss": 0.0001, "step": 22491 }, { "epoch": 2.7, "learning_rate": 1.3844991001799642e-06, "loss": 0.0002, "step": 22492 }, { "epoch": 2.7, "learning_rate": 1.3844271145770845e-06, "loss": 0.0002, "step": 22493 }, { "epoch": 2.7, "learning_rate": 1.384355128974205e-06, "loss": 0.0001, "step": 22494 }, { "epoch": 2.7, "learning_rate": 1.3842831433713257e-06, "loss": 0.0001, "step": 22495 }, { "epoch": 2.7, "learning_rate": 1.3842111577684462e-06, "loss": 0.0001, "step": 22496 }, { "epoch": 2.7, "learning_rate": 1.3841391721655668e-06, "loss": 0.0002, "step": 22497 }, { "epoch": 2.7, "learning_rate": 1.3840671865626876e-06, "loss": 0.0, "step": 22498 }, { "epoch": 2.7, "learning_rate": 1.3839952009598081e-06, "loss": 0.0003, "step": 22499 }, { "epoch": 2.7, "learning_rate": 1.3839232153569287e-06, "loss": 0.0001, "step": 22500 }, { "epoch": 2.7, "learning_rate": 1.3838512297540493e-06, "loss": 0.0001, "step": 22501 }, { "epoch": 2.7, "learning_rate": 1.3837792441511698e-06, "loss": 0.0001, "step": 22502 }, { "epoch": 2.7, "learning_rate": 1.3837072585482904e-06, "loss": 0.0022, "step": 22503 }, { "epoch": 2.7, "learning_rate": 1.383635272945411e-06, "loss": 0.0001, "step": 22504 }, { "epoch": 2.7, "learning_rate": 1.3835632873425315e-06, "loss": 0.0003, "step": 22505 }, { "epoch": 2.7, "learning_rate": 1.3834913017396521e-06, "loss": 0.0001, "step": 22506 }, { "epoch": 2.7, "learning_rate": 1.3834193161367727e-06, "loss": 0.0002, "step": 22507 }, { "epoch": 2.7, "learning_rate": 1.3833473305338932e-06, "loss": 0.0001, "step": 22508 }, { "epoch": 2.7, "learning_rate": 1.3832753449310138e-06, "loss": 0.0002, "step": 22509 }, { "epoch": 2.7, "learning_rate": 1.3832033593281344e-06, "loss": 0.0001, "step": 22510 }, { "epoch": 2.7, "learning_rate": 1.383131373725255e-06, "loss": 0.0002, "step": 22511 }, { "epoch": 2.7, "learning_rate": 1.3830593881223755e-06, "loss": 0.0004, "step": 22512 }, { "epoch": 2.7, "learning_rate": 1.382987402519496e-06, "loss": 0.0002, "step": 22513 }, { "epoch": 2.7, "learning_rate": 1.3829154169166169e-06, "loss": 0.0003, "step": 22514 }, { "epoch": 2.7, "learning_rate": 1.3828434313137374e-06, "loss": 0.0001, "step": 22515 }, { "epoch": 2.7, "learning_rate": 1.3827714457108578e-06, "loss": 0.3359, "step": 22516 }, { "epoch": 2.7, "learning_rate": 1.3826994601079783e-06, "loss": 0.0001, "step": 22517 }, { "epoch": 2.7, "learning_rate": 1.382627474505099e-06, "loss": 0.0001, "step": 22518 }, { "epoch": 2.7, "learning_rate": 1.3825554889022195e-06, "loss": 0.0003, "step": 22519 }, { "epoch": 2.7, "learning_rate": 1.38248350329934e-06, "loss": 0.0001, "step": 22520 }, { "epoch": 2.7, "learning_rate": 1.3824115176964608e-06, "loss": 0.0001, "step": 22521 }, { "epoch": 2.7, "learning_rate": 1.3823395320935814e-06, "loss": 0.0001, "step": 22522 }, { "epoch": 2.7, "learning_rate": 1.382267546490702e-06, "loss": 0.0001, "step": 22523 }, { "epoch": 2.7, "learning_rate": 1.3821955608878225e-06, "loss": 0.0001, "step": 22524 }, { "epoch": 2.7, "learning_rate": 1.382123575284943e-06, "loss": 0.0002, "step": 22525 }, { "epoch": 2.7, "learning_rate": 1.3820515896820637e-06, "loss": 0.0001, "step": 22526 }, { "epoch": 2.7, "learning_rate": 1.3819796040791842e-06, "loss": 0.2988, "step": 22527 }, { "epoch": 2.7, "learning_rate": 1.3819076184763048e-06, "loss": 0.0001, "step": 22528 }, { "epoch": 2.7, "learning_rate": 1.3818356328734254e-06, "loss": 0.0001, "step": 22529 }, { "epoch": 2.7, "learning_rate": 1.381763647270546e-06, "loss": 0.0001, "step": 22530 }, { "epoch": 2.7, "learning_rate": 1.3816916616676665e-06, "loss": 0.0002, "step": 22531 }, { "epoch": 2.7, "learning_rate": 1.381619676064787e-06, "loss": 0.0002, "step": 22532 }, { "epoch": 2.7, "learning_rate": 1.3815476904619076e-06, "loss": 0.0001, "step": 22533 }, { "epoch": 2.7, "learning_rate": 1.3814757048590282e-06, "loss": 0.0001, "step": 22534 }, { "epoch": 2.7, "learning_rate": 1.3814037192561488e-06, "loss": 0.0001, "step": 22535 }, { "epoch": 2.7, "learning_rate": 1.3813317336532693e-06, "loss": 0.0002, "step": 22536 }, { "epoch": 2.7, "learning_rate": 1.38125974805039e-06, "loss": 0.0001, "step": 22537 }, { "epoch": 2.7, "learning_rate": 1.3811877624475107e-06, "loss": 0.0001, "step": 22538 }, { "epoch": 2.7, "learning_rate": 1.381115776844631e-06, "loss": 0.0004, "step": 22539 }, { "epoch": 2.7, "learning_rate": 1.3810437912417516e-06, "loss": 0.0001, "step": 22540 }, { "epoch": 2.7, "learning_rate": 1.3809718056388722e-06, "loss": 0.0001, "step": 22541 }, { "epoch": 2.7, "learning_rate": 1.3808998200359927e-06, "loss": 0.0001, "step": 22542 }, { "epoch": 2.7, "learning_rate": 1.3808278344331133e-06, "loss": 0.0001, "step": 22543 }, { "epoch": 2.7, "learning_rate": 1.380755848830234e-06, "loss": 0.0002, "step": 22544 }, { "epoch": 2.7, "learning_rate": 1.3806838632273546e-06, "loss": 0.0001, "step": 22545 }, { "epoch": 2.7, "learning_rate": 1.3806118776244752e-06, "loss": 0.0001, "step": 22546 }, { "epoch": 2.7, "learning_rate": 1.3805398920215958e-06, "loss": 0.0002, "step": 22547 }, { "epoch": 2.7, "learning_rate": 1.3804679064187163e-06, "loss": 0.0001, "step": 22548 }, { "epoch": 2.7, "learning_rate": 1.380395920815837e-06, "loss": 0.0003, "step": 22549 }, { "epoch": 2.7, "learning_rate": 1.3803239352129575e-06, "loss": 0.0109, "step": 22550 }, { "epoch": 2.7, "learning_rate": 1.380251949610078e-06, "loss": 0.0001, "step": 22551 }, { "epoch": 2.7, "learning_rate": 1.3801799640071986e-06, "loss": 0.0002, "step": 22552 }, { "epoch": 2.7, "learning_rate": 1.3801079784043192e-06, "loss": 0.0001, "step": 22553 }, { "epoch": 2.7, "learning_rate": 1.3800359928014397e-06, "loss": 0.0003, "step": 22554 }, { "epoch": 2.7, "learning_rate": 1.3799640071985603e-06, "loss": 0.0002, "step": 22555 }, { "epoch": 2.7, "learning_rate": 1.3798920215956809e-06, "loss": 0.0003, "step": 22556 }, { "epoch": 2.7, "learning_rate": 1.3798200359928014e-06, "loss": 0.0001, "step": 22557 }, { "epoch": 2.7, "learning_rate": 1.379748050389922e-06, "loss": 0.0002, "step": 22558 }, { "epoch": 2.7, "learning_rate": 1.3796760647870426e-06, "loss": 0.0001, "step": 22559 }, { "epoch": 2.7, "learning_rate": 1.3796040791841634e-06, "loss": 0.0005, "step": 22560 }, { "epoch": 2.7, "learning_rate": 1.379532093581284e-06, "loss": 0.0001, "step": 22561 }, { "epoch": 2.7, "learning_rate": 1.3794601079784043e-06, "loss": 0.0001, "step": 22562 }, { "epoch": 2.7, "learning_rate": 1.3793881223755248e-06, "loss": 0.0001, "step": 22563 }, { "epoch": 2.7, "learning_rate": 1.3793161367726454e-06, "loss": 0.0001, "step": 22564 }, { "epoch": 2.7, "learning_rate": 1.379244151169766e-06, "loss": 0.0001, "step": 22565 }, { "epoch": 2.7, "learning_rate": 1.3791721655668865e-06, "loss": 0.0004, "step": 22566 }, { "epoch": 2.7, "learning_rate": 1.3791001799640073e-06, "loss": 0.0001, "step": 22567 }, { "epoch": 2.7, "learning_rate": 1.3790281943611279e-06, "loss": 0.0001, "step": 22568 }, { "epoch": 2.7, "learning_rate": 1.3789562087582485e-06, "loss": 0.0058, "step": 22569 }, { "epoch": 2.7, "learning_rate": 1.378884223155369e-06, "loss": 0.0003, "step": 22570 }, { "epoch": 2.7, "learning_rate": 1.3788122375524896e-06, "loss": 0.0002, "step": 22571 }, { "epoch": 2.7, "learning_rate": 1.3787402519496102e-06, "loss": 0.0001, "step": 22572 }, { "epoch": 2.7, "learning_rate": 1.3786682663467307e-06, "loss": 1.1318, "step": 22573 }, { "epoch": 2.71, "learning_rate": 1.3785962807438513e-06, "loss": 0.0001, "step": 22574 }, { "epoch": 2.71, "learning_rate": 1.3785242951409719e-06, "loss": 0.0001, "step": 22575 }, { "epoch": 2.71, "learning_rate": 1.3784523095380924e-06, "loss": 0.0001, "step": 22576 }, { "epoch": 2.71, "learning_rate": 1.378380323935213e-06, "loss": 0.0001, "step": 22577 }, { "epoch": 2.71, "learning_rate": 1.3783083383323336e-06, "loss": 0.0001, "step": 22578 }, { "epoch": 2.71, "learning_rate": 1.3782363527294541e-06, "loss": 0.0002, "step": 22579 }, { "epoch": 2.71, "learning_rate": 1.3781643671265747e-06, "loss": 0.0001, "step": 22580 }, { "epoch": 2.71, "learning_rate": 1.3780923815236953e-06, "loss": 0.0001, "step": 22581 }, { "epoch": 2.71, "learning_rate": 1.3780203959208158e-06, "loss": 0.7907, "step": 22582 }, { "epoch": 2.71, "learning_rate": 1.3779484103179366e-06, "loss": 0.0001, "step": 22583 }, { "epoch": 2.71, "learning_rate": 1.3778764247150572e-06, "loss": 0.0002, "step": 22584 }, { "epoch": 2.71, "learning_rate": 1.3778044391121775e-06, "loss": 0.0002, "step": 22585 }, { "epoch": 2.71, "learning_rate": 1.377732453509298e-06, "loss": 0.0001, "step": 22586 }, { "epoch": 2.71, "learning_rate": 1.3776604679064187e-06, "loss": 0.0001, "step": 22587 }, { "epoch": 2.71, "learning_rate": 1.3775884823035392e-06, "loss": 0.0014, "step": 22588 }, { "epoch": 2.71, "learning_rate": 1.3775164967006598e-06, "loss": 0.0002, "step": 22589 }, { "epoch": 2.71, "learning_rate": 1.3774445110977806e-06, "loss": 0.0002, "step": 22590 }, { "epoch": 2.71, "learning_rate": 1.3773725254949011e-06, "loss": 0.0001, "step": 22591 }, { "epoch": 2.71, "learning_rate": 1.3773005398920217e-06, "loss": 0.0002, "step": 22592 }, { "epoch": 2.71, "learning_rate": 1.3772285542891423e-06, "loss": 0.0001, "step": 22593 }, { "epoch": 2.71, "learning_rate": 1.3771565686862628e-06, "loss": 0.0001, "step": 22594 }, { "epoch": 2.71, "learning_rate": 1.3770845830833834e-06, "loss": 0.0002, "step": 22595 }, { "epoch": 2.71, "learning_rate": 1.377012597480504e-06, "loss": 0.0001, "step": 22596 }, { "epoch": 2.71, "learning_rate": 1.3769406118776245e-06, "loss": 0.0003, "step": 22597 }, { "epoch": 2.71, "learning_rate": 1.376868626274745e-06, "loss": 0.0001, "step": 22598 }, { "epoch": 2.71, "learning_rate": 1.3767966406718657e-06, "loss": 0.0002, "step": 22599 }, { "epoch": 2.71, "learning_rate": 1.3767246550689862e-06, "loss": 0.0003, "step": 22600 }, { "epoch": 2.71, "learning_rate": 1.3766526694661068e-06, "loss": 0.0002, "step": 22601 }, { "epoch": 2.71, "learning_rate": 1.3765806838632274e-06, "loss": 0.0002, "step": 22602 }, { "epoch": 2.71, "learning_rate": 1.376508698260348e-06, "loss": 0.0003, "step": 22603 }, { "epoch": 2.71, "learning_rate": 1.3764367126574685e-06, "loss": 0.0022, "step": 22604 }, { "epoch": 2.71, "learning_rate": 1.376364727054589e-06, "loss": 0.0002, "step": 22605 }, { "epoch": 2.71, "learning_rate": 1.3762927414517098e-06, "loss": 0.0001, "step": 22606 }, { "epoch": 2.71, "learning_rate": 1.3762207558488304e-06, "loss": 0.0001, "step": 22607 }, { "epoch": 2.71, "learning_rate": 1.376148770245951e-06, "loss": 0.0003, "step": 22608 }, { "epoch": 2.71, "learning_rate": 1.3760767846430713e-06, "loss": 0.0002, "step": 22609 }, { "epoch": 2.71, "learning_rate": 1.376004799040192e-06, "loss": 0.0002, "step": 22610 }, { "epoch": 2.71, "learning_rate": 1.3759328134373125e-06, "loss": 0.0002, "step": 22611 }, { "epoch": 2.71, "learning_rate": 1.375860827834433e-06, "loss": 0.0001, "step": 22612 }, { "epoch": 2.71, "learning_rate": 1.3757888422315538e-06, "loss": 0.0002, "step": 22613 }, { "epoch": 2.71, "learning_rate": 1.3757168566286744e-06, "loss": 0.0001, "step": 22614 }, { "epoch": 2.71, "learning_rate": 1.375644871025795e-06, "loss": 0.0002, "step": 22615 }, { "epoch": 2.71, "learning_rate": 1.3755728854229155e-06, "loss": 0.0001, "step": 22616 }, { "epoch": 2.71, "learning_rate": 1.375500899820036e-06, "loss": 0.0001, "step": 22617 }, { "epoch": 2.71, "learning_rate": 1.3754289142171566e-06, "loss": 0.0002, "step": 22618 }, { "epoch": 2.71, "learning_rate": 1.3753569286142772e-06, "loss": 0.0001, "step": 22619 }, { "epoch": 2.71, "learning_rate": 1.3752849430113978e-06, "loss": 0.0003, "step": 22620 }, { "epoch": 2.71, "learning_rate": 1.3752129574085183e-06, "loss": 0.0001, "step": 22621 }, { "epoch": 2.71, "learning_rate": 1.375140971805639e-06, "loss": 0.0001, "step": 22622 }, { "epoch": 2.71, "learning_rate": 1.3750689862027595e-06, "loss": 0.0001, "step": 22623 }, { "epoch": 2.71, "learning_rate": 1.37499700059988e-06, "loss": 0.0001, "step": 22624 }, { "epoch": 2.71, "learning_rate": 1.3749250149970006e-06, "loss": 0.0001, "step": 22625 }, { "epoch": 2.71, "learning_rate": 1.3748530293941212e-06, "loss": 0.0002, "step": 22626 }, { "epoch": 2.71, "learning_rate": 1.3747810437912418e-06, "loss": 0.0002, "step": 22627 }, { "epoch": 2.71, "learning_rate": 1.3747090581883623e-06, "loss": 0.0002, "step": 22628 }, { "epoch": 2.71, "learning_rate": 1.374637072585483e-06, "loss": 0.0001, "step": 22629 }, { "epoch": 2.71, "learning_rate": 1.3745650869826037e-06, "loss": 0.0003, "step": 22630 }, { "epoch": 2.71, "learning_rate": 1.3744931013797242e-06, "loss": 0.0001, "step": 22631 }, { "epoch": 2.71, "learning_rate": 1.3744211157768446e-06, "loss": 0.0001, "step": 22632 }, { "epoch": 2.71, "learning_rate": 1.3743491301739652e-06, "loss": 0.0001, "step": 22633 }, { "epoch": 2.71, "learning_rate": 1.3742771445710857e-06, "loss": 0.0, "step": 22634 }, { "epoch": 2.71, "learning_rate": 1.3742051589682063e-06, "loss": 0.0001, "step": 22635 }, { "epoch": 2.71, "learning_rate": 1.3741331733653269e-06, "loss": 0.0001, "step": 22636 }, { "epoch": 2.71, "learning_rate": 1.3740611877624476e-06, "loss": 0.0002, "step": 22637 }, { "epoch": 2.71, "learning_rate": 1.3739892021595682e-06, "loss": 0.0001, "step": 22638 }, { "epoch": 2.71, "learning_rate": 1.3739172165566888e-06, "loss": 0.0001, "step": 22639 }, { "epoch": 2.71, "learning_rate": 1.3738452309538093e-06, "loss": 0.0002, "step": 22640 }, { "epoch": 2.71, "learning_rate": 1.37377324535093e-06, "loss": 0.0002, "step": 22641 }, { "epoch": 2.71, "learning_rate": 1.3737012597480505e-06, "loss": 0.0001, "step": 22642 }, { "epoch": 2.71, "learning_rate": 1.3736292741451708e-06, "loss": 0.0001, "step": 22643 }, { "epoch": 2.71, "learning_rate": 1.3735572885422916e-06, "loss": 0.0001, "step": 22644 }, { "epoch": 2.71, "learning_rate": 1.3734853029394122e-06, "loss": 0.0003, "step": 22645 }, { "epoch": 2.71, "learning_rate": 1.3734133173365327e-06, "loss": 0.0002, "step": 22646 }, { "epoch": 2.71, "learning_rate": 1.3733413317336533e-06, "loss": 0.0002, "step": 22647 }, { "epoch": 2.71, "learning_rate": 1.3732693461307739e-06, "loss": 0.0002, "step": 22648 }, { "epoch": 2.71, "learning_rate": 1.3731973605278944e-06, "loss": 0.0001, "step": 22649 }, { "epoch": 2.71, "learning_rate": 1.373125374925015e-06, "loss": 0.0002, "step": 22650 }, { "epoch": 2.71, "learning_rate": 1.3730533893221356e-06, "loss": 0.0002, "step": 22651 }, { "epoch": 2.71, "learning_rate": 1.3729814037192561e-06, "loss": 0.0002, "step": 22652 }, { "epoch": 2.71, "learning_rate": 1.372909418116377e-06, "loss": 0.0001, "step": 22653 }, { "epoch": 2.71, "learning_rate": 1.3728374325134975e-06, "loss": 0.0001, "step": 22654 }, { "epoch": 2.71, "learning_rate": 1.3727654469106178e-06, "loss": 0.0001, "step": 22655 }, { "epoch": 2.71, "learning_rate": 1.3726934613077384e-06, "loss": 0.0029, "step": 22656 }, { "epoch": 2.72, "learning_rate": 1.372621475704859e-06, "loss": 0.0001, "step": 22657 }, { "epoch": 2.72, "learning_rate": 1.3725494901019795e-06, "loss": 0.0001, "step": 22658 }, { "epoch": 2.72, "learning_rate": 1.3724775044991e-06, "loss": 0.0002, "step": 22659 }, { "epoch": 2.72, "learning_rate": 1.3724055188962209e-06, "loss": 0.0001, "step": 22660 }, { "epoch": 2.72, "learning_rate": 1.3723335332933414e-06, "loss": 0.0001, "step": 22661 }, { "epoch": 2.72, "learning_rate": 1.372261547690462e-06, "loss": 0.0002, "step": 22662 }, { "epoch": 2.72, "learning_rate": 1.3721895620875826e-06, "loss": 0.0001, "step": 22663 }, { "epoch": 2.72, "learning_rate": 1.3721175764847031e-06, "loss": 0.0001, "step": 22664 }, { "epoch": 2.72, "learning_rate": 1.3720455908818237e-06, "loss": 0.0003, "step": 22665 }, { "epoch": 2.72, "learning_rate": 1.371973605278944e-06, "loss": 0.0002, "step": 22666 }, { "epoch": 2.72, "learning_rate": 1.3719016196760648e-06, "loss": 0.0005, "step": 22667 }, { "epoch": 2.72, "learning_rate": 1.3718296340731854e-06, "loss": 0.0002, "step": 22668 }, { "epoch": 2.72, "learning_rate": 1.371757648470306e-06, "loss": 0.0002, "step": 22669 }, { "epoch": 2.72, "learning_rate": 1.3716856628674265e-06, "loss": 0.0002, "step": 22670 }, { "epoch": 2.72, "learning_rate": 1.3716136772645471e-06, "loss": 0.0002, "step": 22671 }, { "epoch": 2.72, "learning_rate": 1.3715416916616677e-06, "loss": 0.0001, "step": 22672 }, { "epoch": 2.72, "learning_rate": 1.3714697060587882e-06, "loss": 0.0001, "step": 22673 }, { "epoch": 2.72, "learning_rate": 1.3713977204559088e-06, "loss": 0.0001, "step": 22674 }, { "epoch": 2.72, "learning_rate": 1.3713257348530294e-06, "loss": 0.0001, "step": 22675 }, { "epoch": 2.72, "learning_rate": 1.3712537492501502e-06, "loss": 0.0001, "step": 22676 }, { "epoch": 2.72, "learning_rate": 1.3711817636472707e-06, "loss": 0.0002, "step": 22677 }, { "epoch": 2.72, "learning_rate": 1.371109778044391e-06, "loss": 0.0623, "step": 22678 }, { "epoch": 2.72, "learning_rate": 1.3710377924415116e-06, "loss": 0.0004, "step": 22679 }, { "epoch": 2.72, "learning_rate": 1.3709658068386322e-06, "loss": 0.0001, "step": 22680 }, { "epoch": 2.72, "learning_rate": 1.3708938212357528e-06, "loss": 0.0003, "step": 22681 }, { "epoch": 2.72, "learning_rate": 1.3708218356328733e-06, "loss": 0.0001, "step": 22682 }, { "epoch": 2.72, "learning_rate": 1.3707498500299941e-06, "loss": 0.0001, "step": 22683 }, { "epoch": 2.72, "learning_rate": 1.3706778644271147e-06, "loss": 0.0001, "step": 22684 }, { "epoch": 2.72, "learning_rate": 1.3706058788242353e-06, "loss": 0.0001, "step": 22685 }, { "epoch": 2.72, "learning_rate": 1.3705338932213558e-06, "loss": 0.0002, "step": 22686 }, { "epoch": 2.72, "learning_rate": 1.3704619076184764e-06, "loss": 0.0001, "step": 22687 }, { "epoch": 2.72, "learning_rate": 1.370389922015597e-06, "loss": 0.0002, "step": 22688 }, { "epoch": 2.72, "learning_rate": 1.3703179364127173e-06, "loss": 0.0001, "step": 22689 }, { "epoch": 2.72, "learning_rate": 1.370245950809838e-06, "loss": 0.0002, "step": 22690 }, { "epoch": 2.72, "learning_rate": 1.3701739652069587e-06, "loss": 0.0002, "step": 22691 }, { "epoch": 2.72, "learning_rate": 1.3701019796040792e-06, "loss": 0.0002, "step": 22692 }, { "epoch": 2.72, "learning_rate": 1.3700299940011998e-06, "loss": 0.0002, "step": 22693 }, { "epoch": 2.72, "learning_rate": 1.3699580083983204e-06, "loss": 0.0001, "step": 22694 }, { "epoch": 2.72, "learning_rate": 1.369886022795441e-06, "loss": 0.0001, "step": 22695 }, { "epoch": 2.72, "learning_rate": 1.3698140371925615e-06, "loss": 0.0002, "step": 22696 }, { "epoch": 2.72, "learning_rate": 1.369742051589682e-06, "loss": 0.0001, "step": 22697 }, { "epoch": 2.72, "learning_rate": 1.3696700659868026e-06, "loss": 0.4885, "step": 22698 }, { "epoch": 2.72, "learning_rate": 1.3695980803839234e-06, "loss": 0.0001, "step": 22699 }, { "epoch": 2.72, "learning_rate": 1.369526094781044e-06, "loss": 0.0001, "step": 22700 }, { "epoch": 2.72, "learning_rate": 1.3694541091781643e-06, "loss": 0.5214, "step": 22701 }, { "epoch": 2.72, "learning_rate": 1.369382123575285e-06, "loss": 0.6387, "step": 22702 }, { "epoch": 2.72, "learning_rate": 1.3693101379724055e-06, "loss": 0.0003, "step": 22703 }, { "epoch": 2.72, "learning_rate": 1.369238152369526e-06, "loss": 0.0002, "step": 22704 }, { "epoch": 2.72, "learning_rate": 1.3691661667666466e-06, "loss": 0.0001, "step": 22705 }, { "epoch": 2.72, "learning_rate": 1.3690941811637674e-06, "loss": 0.0002, "step": 22706 }, { "epoch": 2.72, "learning_rate": 1.369022195560888e-06, "loss": 0.0002, "step": 22707 }, { "epoch": 2.72, "learning_rate": 1.3689502099580085e-06, "loss": 0.0015, "step": 22708 }, { "epoch": 2.72, "learning_rate": 1.368878224355129e-06, "loss": 0.0001, "step": 22709 }, { "epoch": 2.72, "learning_rate": 1.3688062387522496e-06, "loss": 0.0006, "step": 22710 }, { "epoch": 2.72, "learning_rate": 1.3687342531493702e-06, "loss": 0.0003, "step": 22711 }, { "epoch": 2.72, "learning_rate": 1.3686622675464908e-06, "loss": 0.0002, "step": 22712 }, { "epoch": 2.72, "learning_rate": 1.3685902819436113e-06, "loss": 0.0003, "step": 22713 }, { "epoch": 2.72, "learning_rate": 1.368518296340732e-06, "loss": 0.0001, "step": 22714 }, { "epoch": 2.72, "learning_rate": 1.3684463107378525e-06, "loss": 0.0003, "step": 22715 }, { "epoch": 2.72, "learning_rate": 1.368374325134973e-06, "loss": 0.0016, "step": 22716 }, { "epoch": 2.72, "learning_rate": 1.3683023395320936e-06, "loss": 0.0001, "step": 22717 }, { "epoch": 2.72, "learning_rate": 1.3682303539292142e-06, "loss": 0.0001, "step": 22718 }, { "epoch": 2.72, "learning_rate": 1.3681583683263347e-06, "loss": 0.0002, "step": 22719 }, { "epoch": 2.72, "learning_rate": 1.3680863827234553e-06, "loss": 0.0001, "step": 22720 }, { "epoch": 2.72, "learning_rate": 1.3680143971205759e-06, "loss": 0.0, "step": 22721 }, { "epoch": 2.72, "learning_rate": 1.3679424115176967e-06, "loss": 0.0002, "step": 22722 }, { "epoch": 2.72, "learning_rate": 1.3678704259148172e-06, "loss": 0.0001, "step": 22723 }, { "epoch": 2.72, "learning_rate": 1.3677984403119376e-06, "loss": 0.0002, "step": 22724 }, { "epoch": 2.72, "learning_rate": 1.3677264547090581e-06, "loss": 0.0002, "step": 22725 }, { "epoch": 2.72, "learning_rate": 1.3676544691061787e-06, "loss": 0.0001, "step": 22726 }, { "epoch": 2.72, "learning_rate": 1.3675824835032993e-06, "loss": 0.0001, "step": 22727 }, { "epoch": 2.72, "learning_rate": 1.3675104979004198e-06, "loss": 0.0001, "step": 22728 }, { "epoch": 2.72, "learning_rate": 1.3674385122975406e-06, "loss": 0.0003, "step": 22729 }, { "epoch": 2.72, "learning_rate": 1.3673665266946612e-06, "loss": 0.0002, "step": 22730 }, { "epoch": 2.72, "learning_rate": 1.3672945410917818e-06, "loss": 0.0004, "step": 22731 }, { "epoch": 2.72, "learning_rate": 1.3672225554889023e-06, "loss": 0.0001, "step": 22732 }, { "epoch": 2.72, "learning_rate": 1.3671505698860229e-06, "loss": 0.0003, "step": 22733 }, { "epoch": 2.72, "learning_rate": 1.3670785842831435e-06, "loss": 0.0002, "step": 22734 }, { "epoch": 2.72, "learning_rate": 1.367006598680264e-06, "loss": 0.0002, "step": 22735 }, { "epoch": 2.72, "learning_rate": 1.3669346130773846e-06, "loss": 0.0001, "step": 22736 }, { "epoch": 2.72, "learning_rate": 1.3668626274745052e-06, "loss": 0.0001, "step": 22737 }, { "epoch": 2.72, "learning_rate": 1.3667906418716257e-06, "loss": 0.0001, "step": 22738 }, { "epoch": 2.72, "learning_rate": 1.3667186562687463e-06, "loss": 0.0001, "step": 22739 }, { "epoch": 2.72, "learning_rate": 1.3666466706658669e-06, "loss": 0.0002, "step": 22740 }, { "epoch": 2.73, "learning_rate": 1.3665746850629874e-06, "loss": 0.0001, "step": 22741 }, { "epoch": 2.73, "learning_rate": 1.366502699460108e-06, "loss": 0.0001, "step": 22742 }, { "epoch": 2.73, "learning_rate": 1.3664307138572286e-06, "loss": 0.0002, "step": 22743 }, { "epoch": 2.73, "learning_rate": 1.3663587282543491e-06, "loss": 0.0001, "step": 22744 }, { "epoch": 2.73, "learning_rate": 1.36628674265147e-06, "loss": 0.0001, "step": 22745 }, { "epoch": 2.73, "learning_rate": 1.3662147570485905e-06, "loss": 0.0002, "step": 22746 }, { "epoch": 2.73, "learning_rate": 1.3661427714457108e-06, "loss": 0.0001, "step": 22747 }, { "epoch": 2.73, "learning_rate": 1.3660707858428314e-06, "loss": 0.0001, "step": 22748 }, { "epoch": 2.73, "learning_rate": 1.365998800239952e-06, "loss": 0.0001, "step": 22749 }, { "epoch": 2.73, "learning_rate": 1.3659268146370725e-06, "loss": 0.0004, "step": 22750 }, { "epoch": 2.73, "learning_rate": 1.365854829034193e-06, "loss": 0.0001, "step": 22751 }, { "epoch": 2.73, "learning_rate": 1.3657828434313139e-06, "loss": 0.0006, "step": 22752 }, { "epoch": 2.73, "learning_rate": 1.3657108578284344e-06, "loss": 0.0001, "step": 22753 }, { "epoch": 2.73, "learning_rate": 1.365638872225555e-06, "loss": 0.0001, "step": 22754 }, { "epoch": 2.73, "learning_rate": 1.3655668866226756e-06, "loss": 0.0001, "step": 22755 }, { "epoch": 2.73, "learning_rate": 1.3654949010197961e-06, "loss": 0.0003, "step": 22756 }, { "epoch": 2.73, "learning_rate": 1.3654229154169167e-06, "loss": 0.0008, "step": 22757 }, { "epoch": 2.73, "learning_rate": 1.3653509298140373e-06, "loss": 0.0002, "step": 22758 }, { "epoch": 2.73, "learning_rate": 1.3652789442111578e-06, "loss": 0.0001, "step": 22759 }, { "epoch": 2.73, "learning_rate": 1.3652069586082784e-06, "loss": 0.0001, "step": 22760 }, { "epoch": 2.73, "learning_rate": 1.365134973005399e-06, "loss": 0.0012, "step": 22761 }, { "epoch": 2.73, "learning_rate": 1.3650629874025195e-06, "loss": 0.0001, "step": 22762 }, { "epoch": 2.73, "learning_rate": 1.36499100179964e-06, "loss": 0.0003, "step": 22763 }, { "epoch": 2.73, "learning_rate": 1.3649190161967607e-06, "loss": 0.0002, "step": 22764 }, { "epoch": 2.73, "learning_rate": 1.3648470305938812e-06, "loss": 0.0001, "step": 22765 }, { "epoch": 2.73, "learning_rate": 1.3647750449910018e-06, "loss": 0.0, "step": 22766 }, { "epoch": 2.73, "learning_rate": 1.3647030593881224e-06, "loss": 0.0002, "step": 22767 }, { "epoch": 2.73, "learning_rate": 1.364631073785243e-06, "loss": 0.0001, "step": 22768 }, { "epoch": 2.73, "learning_rate": 1.3645590881823637e-06, "loss": 0.0001, "step": 22769 }, { "epoch": 2.73, "learning_rate": 1.364487102579484e-06, "loss": 0.8159, "step": 22770 }, { "epoch": 2.73, "learning_rate": 1.3644151169766046e-06, "loss": 1.0802, "step": 22771 }, { "epoch": 2.73, "learning_rate": 1.3643431313737252e-06, "loss": 0.0001, "step": 22772 }, { "epoch": 2.73, "learning_rate": 1.3642711457708458e-06, "loss": 0.0001, "step": 22773 }, { "epoch": 2.73, "learning_rate": 1.3641991601679663e-06, "loss": 0.0001, "step": 22774 }, { "epoch": 2.73, "learning_rate": 1.364127174565087e-06, "loss": 0.0001, "step": 22775 }, { "epoch": 2.73, "learning_rate": 1.3640551889622077e-06, "loss": 0.0001, "step": 22776 }, { "epoch": 2.73, "learning_rate": 1.3639832033593283e-06, "loss": 0.0001, "step": 22777 }, { "epoch": 2.73, "learning_rate": 1.3639112177564488e-06, "loss": 0.0001, "step": 22778 }, { "epoch": 2.73, "learning_rate": 1.3638392321535694e-06, "loss": 0.0005, "step": 22779 }, { "epoch": 2.73, "learning_rate": 1.36376724655069e-06, "loss": 0.0001, "step": 22780 }, { "epoch": 2.73, "learning_rate": 1.3636952609478105e-06, "loss": 0.0037, "step": 22781 }, { "epoch": 2.73, "learning_rate": 1.3636232753449309e-06, "loss": 0.0001, "step": 22782 }, { "epoch": 2.73, "learning_rate": 1.3635512897420517e-06, "loss": 0.0002, "step": 22783 }, { "epoch": 2.73, "learning_rate": 1.3634793041391722e-06, "loss": 0.0002, "step": 22784 }, { "epoch": 2.73, "learning_rate": 1.3634073185362928e-06, "loss": 0.0002, "step": 22785 }, { "epoch": 2.73, "learning_rate": 1.3633353329334134e-06, "loss": 0.0002, "step": 22786 }, { "epoch": 2.73, "learning_rate": 1.363263347330534e-06, "loss": 0.5144, "step": 22787 }, { "epoch": 2.73, "learning_rate": 1.3631913617276545e-06, "loss": 0.0007, "step": 22788 }, { "epoch": 2.73, "learning_rate": 1.363119376124775e-06, "loss": 0.0002, "step": 22789 }, { "epoch": 2.73, "learning_rate": 1.3630473905218956e-06, "loss": 0.0002, "step": 22790 }, { "epoch": 2.73, "learning_rate": 1.3629754049190162e-06, "loss": 0.0001, "step": 22791 }, { "epoch": 2.73, "learning_rate": 1.362903419316137e-06, "loss": 0.0001, "step": 22792 }, { "epoch": 2.73, "learning_rate": 1.3628314337132573e-06, "loss": 0.0001, "step": 22793 }, { "epoch": 2.73, "learning_rate": 1.3627594481103779e-06, "loss": 0.0002, "step": 22794 }, { "epoch": 2.73, "learning_rate": 1.3626874625074985e-06, "loss": 0.0002, "step": 22795 }, { "epoch": 2.73, "learning_rate": 1.362615476904619e-06, "loss": 0.0001, "step": 22796 }, { "epoch": 2.73, "learning_rate": 1.3625434913017396e-06, "loss": 0.0002, "step": 22797 }, { "epoch": 2.73, "learning_rate": 1.3624715056988602e-06, "loss": 0.0001, "step": 22798 }, { "epoch": 2.73, "learning_rate": 1.362399520095981e-06, "loss": 0.0001, "step": 22799 }, { "epoch": 2.73, "learning_rate": 1.3623275344931015e-06, "loss": 0.0002, "step": 22800 }, { "epoch": 2.73, "learning_rate": 1.362255548890222e-06, "loss": 0.0001, "step": 22801 }, { "epoch": 2.73, "learning_rate": 1.3621835632873426e-06, "loss": 0.0001, "step": 22802 }, { "epoch": 2.73, "learning_rate": 1.3621115776844632e-06, "loss": 0.0001, "step": 22803 }, { "epoch": 2.73, "learning_rate": 1.3620395920815838e-06, "loss": 0.0003, "step": 22804 }, { "epoch": 2.73, "learning_rate": 1.3619676064787041e-06, "loss": 0.0003, "step": 22805 }, { "epoch": 2.73, "learning_rate": 1.361895620875825e-06, "loss": 0.0017, "step": 22806 }, { "epoch": 2.73, "learning_rate": 1.3618236352729455e-06, "loss": 0.0001, "step": 22807 }, { "epoch": 2.73, "learning_rate": 1.361751649670066e-06, "loss": 0.0001, "step": 22808 }, { "epoch": 2.73, "learning_rate": 1.3616796640671866e-06, "loss": 0.0002, "step": 22809 }, { "epoch": 2.73, "learning_rate": 1.3616076784643072e-06, "loss": 0.0002, "step": 22810 }, { "epoch": 2.73, "learning_rate": 1.3615356928614277e-06, "loss": 0.0506, "step": 22811 }, { "epoch": 2.73, "learning_rate": 1.3614637072585483e-06, "loss": 0.0002, "step": 22812 }, { "epoch": 2.73, "learning_rate": 1.3613917216556689e-06, "loss": 0.0007, "step": 22813 }, { "epoch": 2.73, "learning_rate": 1.3613197360527894e-06, "loss": 0.0004, "step": 22814 }, { "epoch": 2.73, "learning_rate": 1.3612477504499102e-06, "loss": 0.0001, "step": 22815 }, { "epoch": 2.73, "learning_rate": 1.3611757648470306e-06, "loss": 0.0001, "step": 22816 }, { "epoch": 2.73, "learning_rate": 1.3611037792441511e-06, "loss": 0.0018, "step": 22817 }, { "epoch": 2.73, "learning_rate": 1.3610317936412717e-06, "loss": 0.0002, "step": 22818 }, { "epoch": 2.73, "learning_rate": 1.3609598080383923e-06, "loss": 0.0001, "step": 22819 }, { "epoch": 2.73, "learning_rate": 1.3608878224355128e-06, "loss": 0.0002, "step": 22820 }, { "epoch": 2.73, "learning_rate": 1.3608158368326334e-06, "loss": 0.0002, "step": 22821 }, { "epoch": 2.73, "learning_rate": 1.3607438512297542e-06, "loss": 0.0002, "step": 22822 }, { "epoch": 2.73, "learning_rate": 1.3606718656268747e-06, "loss": 0.0002, "step": 22823 }, { "epoch": 2.74, "learning_rate": 1.3605998800239953e-06, "loss": 0.0001, "step": 22824 }, { "epoch": 2.74, "learning_rate": 1.3605278944211159e-06, "loss": 0.0001, "step": 22825 }, { "epoch": 2.74, "learning_rate": 1.3604559088182364e-06, "loss": 0.0006, "step": 22826 }, { "epoch": 2.74, "learning_rate": 1.360383923215357e-06, "loss": 0.0001, "step": 22827 }, { "epoch": 2.74, "learning_rate": 1.3603119376124774e-06, "loss": 0.0002, "step": 22828 }, { "epoch": 2.74, "learning_rate": 1.3602399520095981e-06, "loss": 0.0002, "step": 22829 }, { "epoch": 2.74, "learning_rate": 1.3601679664067187e-06, "loss": 0.0005, "step": 22830 }, { "epoch": 2.74, "learning_rate": 1.3600959808038393e-06, "loss": 0.0002, "step": 22831 }, { "epoch": 2.74, "learning_rate": 1.3600239952009598e-06, "loss": 0.0001, "step": 22832 }, { "epoch": 2.74, "learning_rate": 1.3599520095980804e-06, "loss": 0.0003, "step": 22833 }, { "epoch": 2.74, "learning_rate": 1.359880023995201e-06, "loss": 0.0065, "step": 22834 }, { "epoch": 2.74, "learning_rate": 1.3598080383923215e-06, "loss": 0.0001, "step": 22835 }, { "epoch": 2.74, "learning_rate": 1.3597360527894421e-06, "loss": 1.0752, "step": 22836 }, { "epoch": 2.74, "learning_rate": 1.3596640671865627e-06, "loss": 0.6753, "step": 22837 }, { "epoch": 2.74, "learning_rate": 1.3595920815836835e-06, "loss": 0.0001, "step": 22838 }, { "epoch": 2.74, "learning_rate": 1.359520095980804e-06, "loss": 0.0001, "step": 22839 }, { "epoch": 2.74, "learning_rate": 1.3594481103779244e-06, "loss": 0.0002, "step": 22840 }, { "epoch": 2.74, "learning_rate": 1.359376124775045e-06, "loss": 0.0004, "step": 22841 }, { "epoch": 2.74, "learning_rate": 1.3593041391721655e-06, "loss": 0.0001, "step": 22842 }, { "epoch": 2.74, "learning_rate": 1.359232153569286e-06, "loss": 0.0002, "step": 22843 }, { "epoch": 2.74, "learning_rate": 1.3591601679664066e-06, "loss": 0.0001, "step": 22844 }, { "epoch": 2.74, "learning_rate": 1.3590881823635274e-06, "loss": 0.0005, "step": 22845 }, { "epoch": 2.74, "learning_rate": 1.359016196760648e-06, "loss": 0.0001, "step": 22846 }, { "epoch": 2.74, "learning_rate": 1.3589442111577686e-06, "loss": 0.0002, "step": 22847 }, { "epoch": 2.74, "learning_rate": 1.3588722255548891e-06, "loss": 0.0009, "step": 22848 }, { "epoch": 2.74, "learning_rate": 1.3588002399520097e-06, "loss": 0.0001, "step": 22849 }, { "epoch": 2.74, "learning_rate": 1.3587282543491303e-06, "loss": 0.0001, "step": 22850 }, { "epoch": 2.74, "learning_rate": 1.3586562687462506e-06, "loss": 0.0001, "step": 22851 }, { "epoch": 2.74, "learning_rate": 1.3585842831433714e-06, "loss": 0.0003, "step": 22852 }, { "epoch": 2.74, "learning_rate": 1.358512297540492e-06, "loss": 0.0002, "step": 22853 }, { "epoch": 2.74, "learning_rate": 1.3584403119376125e-06, "loss": 0.0001, "step": 22854 }, { "epoch": 2.74, "learning_rate": 1.358368326334733e-06, "loss": 0.0002, "step": 22855 }, { "epoch": 2.74, "learning_rate": 1.3582963407318537e-06, "loss": 0.0013, "step": 22856 }, { "epoch": 2.74, "learning_rate": 1.3582243551289742e-06, "loss": 0.0001, "step": 22857 }, { "epoch": 2.74, "learning_rate": 1.3581523695260948e-06, "loss": 0.0001, "step": 22858 }, { "epoch": 2.74, "learning_rate": 1.3580803839232154e-06, "loss": 0.0002, "step": 22859 }, { "epoch": 2.74, "learning_rate": 1.358008398320336e-06, "loss": 0.0002, "step": 22860 }, { "epoch": 2.74, "learning_rate": 1.3579364127174567e-06, "loss": 0.0002, "step": 22861 }, { "epoch": 2.74, "learning_rate": 1.3578644271145773e-06, "loss": 0.0001, "step": 22862 }, { "epoch": 2.74, "learning_rate": 1.3577924415116976e-06, "loss": 0.0002, "step": 22863 }, { "epoch": 2.74, "learning_rate": 1.3577204559088182e-06, "loss": 0.0001, "step": 22864 }, { "epoch": 2.74, "learning_rate": 1.3576484703059388e-06, "loss": 0.0001, "step": 22865 }, { "epoch": 2.74, "learning_rate": 1.3575764847030593e-06, "loss": 0.0002, "step": 22866 }, { "epoch": 2.74, "learning_rate": 1.35750449910018e-06, "loss": 0.0001, "step": 22867 }, { "epoch": 2.74, "learning_rate": 1.3574325134973007e-06, "loss": 0.0002, "step": 22868 }, { "epoch": 2.74, "learning_rate": 1.3573605278944212e-06, "loss": 0.0001, "step": 22869 }, { "epoch": 2.74, "learning_rate": 1.3572885422915418e-06, "loss": 0.0001, "step": 22870 }, { "epoch": 2.74, "learning_rate": 1.3572165566886624e-06, "loss": 0.0001, "step": 22871 }, { "epoch": 2.74, "learning_rate": 1.357144571085783e-06, "loss": 0.0003, "step": 22872 }, { "epoch": 2.74, "learning_rate": 1.3570725854829035e-06, "loss": 0.0001, "step": 22873 }, { "epoch": 2.74, "learning_rate": 1.3570005998800239e-06, "loss": 0.0001, "step": 22874 }, { "epoch": 2.74, "learning_rate": 1.3569286142771446e-06, "loss": 0.0002, "step": 22875 }, { "epoch": 2.74, "learning_rate": 1.3568566286742652e-06, "loss": 0.0, "step": 22876 }, { "epoch": 2.74, "learning_rate": 1.3567846430713858e-06, "loss": 0.0001, "step": 22877 }, { "epoch": 2.74, "learning_rate": 1.3567126574685063e-06, "loss": 0.0002, "step": 22878 }, { "epoch": 2.74, "learning_rate": 1.356640671865627e-06, "loss": 0.0001, "step": 22879 }, { "epoch": 2.74, "learning_rate": 1.3565686862627475e-06, "loss": 0.0002, "step": 22880 }, { "epoch": 2.74, "learning_rate": 1.356496700659868e-06, "loss": 0.0001, "step": 22881 }, { "epoch": 2.74, "learning_rate": 1.3564247150569886e-06, "loss": 0.0001, "step": 22882 }, { "epoch": 2.74, "learning_rate": 1.3563527294541092e-06, "loss": 0.0002, "step": 22883 }, { "epoch": 2.74, "learning_rate": 1.35628074385123e-06, "loss": 0.0002, "step": 22884 }, { "epoch": 2.74, "learning_rate": 1.3562087582483505e-06, "loss": 0.0002, "step": 22885 }, { "epoch": 2.74, "learning_rate": 1.3561367726454709e-06, "loss": 0.6596, "step": 22886 }, { "epoch": 2.74, "learning_rate": 1.3560647870425914e-06, "loss": 0.0004, "step": 22887 }, { "epoch": 2.74, "learning_rate": 1.355992801439712e-06, "loss": 0.0007, "step": 22888 }, { "epoch": 2.74, "learning_rate": 1.3559208158368326e-06, "loss": 0.0001, "step": 22889 }, { "epoch": 2.74, "learning_rate": 1.3558488302339531e-06, "loss": 0.0002, "step": 22890 }, { "epoch": 2.74, "learning_rate": 1.355776844631074e-06, "loss": 0.0001, "step": 22891 }, { "epoch": 2.74, "learning_rate": 1.3557048590281945e-06, "loss": 0.0001, "step": 22892 }, { "epoch": 2.74, "learning_rate": 1.355632873425315e-06, "loss": 0.0002, "step": 22893 }, { "epoch": 2.74, "learning_rate": 1.3555608878224356e-06, "loss": 0.0007, "step": 22894 }, { "epoch": 2.74, "learning_rate": 1.3554889022195562e-06, "loss": 0.0008, "step": 22895 }, { "epoch": 2.74, "learning_rate": 1.3554169166166768e-06, "loss": 0.0002, "step": 22896 }, { "epoch": 2.74, "learning_rate": 1.3553449310137971e-06, "loss": 0.0007, "step": 22897 }, { "epoch": 2.74, "learning_rate": 1.3552729454109177e-06, "loss": 0.0001, "step": 22898 }, { "epoch": 2.74, "learning_rate": 1.3552009598080385e-06, "loss": 0.0002, "step": 22899 }, { "epoch": 2.74, "learning_rate": 1.355128974205159e-06, "loss": 0.9803, "step": 22900 }, { "epoch": 2.74, "learning_rate": 1.3550569886022796e-06, "loss": 0.0003, "step": 22901 }, { "epoch": 2.74, "learning_rate": 1.3549850029994002e-06, "loss": 0.0001, "step": 22902 }, { "epoch": 2.74, "learning_rate": 1.3549130173965207e-06, "loss": 0.0002, "step": 22903 }, { "epoch": 2.74, "learning_rate": 1.3548410317936413e-06, "loss": 0.0025, "step": 22904 }, { "epoch": 2.74, "learning_rate": 1.3547690461907619e-06, "loss": 0.0002, "step": 22905 }, { "epoch": 2.74, "learning_rate": 1.3546970605878824e-06, "loss": 0.0001, "step": 22906 }, { "epoch": 2.74, "learning_rate": 1.354625074985003e-06, "loss": 0.0001, "step": 22907 }, { "epoch": 2.75, "learning_rate": 1.3545530893821238e-06, "loss": 0.0001, "step": 22908 }, { "epoch": 2.75, "learning_rate": 1.3544811037792441e-06, "loss": 0.0001, "step": 22909 }, { "epoch": 2.75, "learning_rate": 1.3544091181763647e-06, "loss": 0.0002, "step": 22910 }, { "epoch": 2.75, "learning_rate": 1.3543371325734853e-06, "loss": 0.0003, "step": 22911 }, { "epoch": 2.75, "learning_rate": 1.3542651469706058e-06, "loss": 0.0001, "step": 22912 }, { "epoch": 2.75, "learning_rate": 1.3541931613677264e-06, "loss": 0.0001, "step": 22913 }, { "epoch": 2.75, "learning_rate": 1.354121175764847e-06, "loss": 0.0001, "step": 22914 }, { "epoch": 2.75, "learning_rate": 1.3540491901619677e-06, "loss": 0.0002, "step": 22915 }, { "epoch": 2.75, "learning_rate": 1.3539772045590883e-06, "loss": 0.0001, "step": 22916 }, { "epoch": 2.75, "learning_rate": 1.3539052189562089e-06, "loss": 0.0001, "step": 22917 }, { "epoch": 2.75, "learning_rate": 1.3538332333533294e-06, "loss": 0.0002, "step": 22918 }, { "epoch": 2.75, "learning_rate": 1.35376124775045e-06, "loss": 0.0001, "step": 22919 }, { "epoch": 2.75, "learning_rate": 1.3536892621475704e-06, "loss": 0.0001, "step": 22920 }, { "epoch": 2.75, "learning_rate": 1.353617276544691e-06, "loss": 0.0001, "step": 22921 }, { "epoch": 2.75, "learning_rate": 1.3535452909418117e-06, "loss": 0.0002, "step": 22922 }, { "epoch": 2.75, "learning_rate": 1.3534733053389323e-06, "loss": 0.0003, "step": 22923 }, { "epoch": 2.75, "learning_rate": 1.3534013197360528e-06, "loss": 0.0005, "step": 22924 }, { "epoch": 2.75, "learning_rate": 1.3533293341331734e-06, "loss": 0.0002, "step": 22925 }, { "epoch": 2.75, "learning_rate": 1.353257348530294e-06, "loss": 0.0001, "step": 22926 }, { "epoch": 2.75, "learning_rate": 1.3531853629274145e-06, "loss": 0.0001, "step": 22927 }, { "epoch": 2.75, "learning_rate": 1.3531133773245351e-06, "loss": 0.0001, "step": 22928 }, { "epoch": 2.75, "learning_rate": 1.3530413917216557e-06, "loss": 0.0003, "step": 22929 }, { "epoch": 2.75, "learning_rate": 1.3529694061187762e-06, "loss": 0.0001, "step": 22930 }, { "epoch": 2.75, "learning_rate": 1.352897420515897e-06, "loss": 0.0002, "step": 22931 }, { "epoch": 2.75, "learning_rate": 1.3528254349130174e-06, "loss": 0.0003, "step": 22932 }, { "epoch": 2.75, "learning_rate": 1.352753449310138e-06, "loss": 0.0057, "step": 22933 }, { "epoch": 2.75, "learning_rate": 1.3526814637072585e-06, "loss": 0.0003, "step": 22934 }, { "epoch": 2.75, "learning_rate": 1.352609478104379e-06, "loss": 0.0001, "step": 22935 }, { "epoch": 2.75, "learning_rate": 1.3525374925014996e-06, "loss": 0.0002, "step": 22936 }, { "epoch": 2.75, "learning_rate": 1.3524655068986202e-06, "loss": 0.0001, "step": 22937 }, { "epoch": 2.75, "learning_rate": 1.352393521295741e-06, "loss": 0.9501, "step": 22938 }, { "epoch": 2.75, "learning_rate": 1.3523215356928616e-06, "loss": 0.0001, "step": 22939 }, { "epoch": 2.75, "learning_rate": 1.3522495500899821e-06, "loss": 0.0001, "step": 22940 }, { "epoch": 2.75, "learning_rate": 1.3521775644871027e-06, "loss": 0.0003, "step": 22941 }, { "epoch": 2.75, "learning_rate": 1.3521055788842233e-06, "loss": 0.0001, "step": 22942 }, { "epoch": 2.75, "learning_rate": 1.3520335932813436e-06, "loss": 0.0002, "step": 22943 }, { "epoch": 2.75, "learning_rate": 1.3519616076784642e-06, "loss": 0.0002, "step": 22944 }, { "epoch": 2.75, "learning_rate": 1.351889622075585e-06, "loss": 0.0003, "step": 22945 }, { "epoch": 2.75, "learning_rate": 1.3518176364727055e-06, "loss": 0.0002, "step": 22946 }, { "epoch": 2.75, "learning_rate": 1.351745650869826e-06, "loss": 0.0003, "step": 22947 }, { "epoch": 2.75, "learning_rate": 1.3516736652669467e-06, "loss": 0.0002, "step": 22948 }, { "epoch": 2.75, "learning_rate": 1.3516016796640672e-06, "loss": 0.0001, "step": 22949 }, { "epoch": 2.75, "learning_rate": 1.3515296940611878e-06, "loss": 0.0003, "step": 22950 }, { "epoch": 2.75, "learning_rate": 1.3514577084583084e-06, "loss": 0.0002, "step": 22951 }, { "epoch": 2.75, "learning_rate": 1.351385722855429e-06, "loss": 0.0003, "step": 22952 }, { "epoch": 2.75, "learning_rate": 1.3513137372525495e-06, "loss": 0.0001, "step": 22953 }, { "epoch": 2.75, "learning_rate": 1.3512417516496703e-06, "loss": 0.0002, "step": 22954 }, { "epoch": 2.75, "learning_rate": 1.3511697660467906e-06, "loss": 0.0001, "step": 22955 }, { "epoch": 2.75, "learning_rate": 1.3510977804439112e-06, "loss": 0.0004, "step": 22956 }, { "epoch": 2.75, "learning_rate": 1.3510257948410318e-06, "loss": 0.0001, "step": 22957 }, { "epoch": 2.75, "learning_rate": 1.3509538092381523e-06, "loss": 0.0001, "step": 22958 }, { "epoch": 2.75, "learning_rate": 1.3508818236352729e-06, "loss": 0.0002, "step": 22959 }, { "epoch": 2.75, "learning_rate": 1.3508098380323935e-06, "loss": 0.0002, "step": 22960 }, { "epoch": 2.75, "learning_rate": 1.3507378524295142e-06, "loss": 0.0001, "step": 22961 }, { "epoch": 2.75, "learning_rate": 1.3506658668266348e-06, "loss": 0.0001, "step": 22962 }, { "epoch": 2.75, "learning_rate": 1.3505938812237554e-06, "loss": 0.0002, "step": 22963 }, { "epoch": 2.75, "learning_rate": 1.350521895620876e-06, "loss": 0.0001, "step": 22964 }, { "epoch": 2.75, "learning_rate": 1.3504499100179965e-06, "loss": 0.0003, "step": 22965 }, { "epoch": 2.75, "learning_rate": 1.350377924415117e-06, "loss": 0.0003, "step": 22966 }, { "epoch": 2.75, "learning_rate": 1.3503059388122374e-06, "loss": 0.0001, "step": 22967 }, { "epoch": 2.75, "learning_rate": 1.3502339532093582e-06, "loss": 0.0002, "step": 22968 }, { "epoch": 2.75, "learning_rate": 1.3501619676064788e-06, "loss": 0.0005, "step": 22969 }, { "epoch": 2.75, "learning_rate": 1.3500899820035993e-06, "loss": 0.0001, "step": 22970 }, { "epoch": 2.75, "learning_rate": 1.35001799640072e-06, "loss": 0.0003, "step": 22971 }, { "epoch": 2.75, "learning_rate": 1.3499460107978405e-06, "loss": 0.0003, "step": 22972 }, { "epoch": 2.75, "learning_rate": 1.349874025194961e-06, "loss": 0.0001, "step": 22973 }, { "epoch": 2.75, "learning_rate": 1.3498020395920816e-06, "loss": 0.0002, "step": 22974 }, { "epoch": 2.75, "learning_rate": 1.3497300539892022e-06, "loss": 1.125, "step": 22975 }, { "epoch": 2.75, "learning_rate": 1.3496580683863227e-06, "loss": 0.0002, "step": 22976 }, { "epoch": 2.75, "learning_rate": 1.3495860827834435e-06, "loss": 0.0001, "step": 22977 }, { "epoch": 2.75, "learning_rate": 1.3495140971805639e-06, "loss": 0.0001, "step": 22978 }, { "epoch": 2.75, "learning_rate": 1.3494421115776844e-06, "loss": 0.0001, "step": 22979 }, { "epoch": 2.75, "learning_rate": 1.349370125974805e-06, "loss": 0.0002, "step": 22980 }, { "epoch": 2.75, "learning_rate": 1.3492981403719256e-06, "loss": 0.2969, "step": 22981 }, { "epoch": 2.75, "learning_rate": 1.3492261547690461e-06, "loss": 0.0002, "step": 22982 }, { "epoch": 2.75, "learning_rate": 1.3491541691661667e-06, "loss": 0.0002, "step": 22983 }, { "epoch": 2.75, "learning_rate": 1.3490821835632875e-06, "loss": 0.0002, "step": 22984 }, { "epoch": 2.75, "learning_rate": 1.349010197960408e-06, "loss": 0.0001, "step": 22985 }, { "epoch": 2.75, "learning_rate": 1.3489382123575286e-06, "loss": 0.0004, "step": 22986 }, { "epoch": 2.75, "learning_rate": 1.3488662267546492e-06, "loss": 0.0001, "step": 22987 }, { "epoch": 2.75, "learning_rate": 1.3487942411517698e-06, "loss": 0.0001, "step": 22988 }, { "epoch": 2.75, "learning_rate": 1.3487222555488903e-06, "loss": 0.0002, "step": 22989 }, { "epoch": 2.75, "learning_rate": 1.3486502699460107e-06, "loss": 0.0003, "step": 22990 }, { "epoch": 2.76, "learning_rate": 1.3485782843431315e-06, "loss": 0.0002, "step": 22991 }, { "epoch": 2.76, "learning_rate": 1.348506298740252e-06, "loss": 0.0002, "step": 22992 }, { "epoch": 2.76, "learning_rate": 1.3484343131373726e-06, "loss": 0.0013, "step": 22993 }, { "epoch": 2.76, "learning_rate": 1.3483623275344932e-06, "loss": 0.0004, "step": 22994 }, { "epoch": 2.76, "learning_rate": 1.3482903419316137e-06, "loss": 0.0003, "step": 22995 }, { "epoch": 2.76, "learning_rate": 1.3482183563287343e-06, "loss": 0.0002, "step": 22996 }, { "epoch": 2.76, "learning_rate": 1.3481463707258549e-06, "loss": 0.0004, "step": 22997 }, { "epoch": 2.76, "learning_rate": 1.3480743851229754e-06, "loss": 0.0002, "step": 22998 }, { "epoch": 2.76, "learning_rate": 1.348002399520096e-06, "loss": 0.0001, "step": 22999 }, { "epoch": 2.76, "learning_rate": 1.3479304139172168e-06, "loss": 0.0002, "step": 23000 }, { "epoch": 2.76, "learning_rate": 1.3478584283143371e-06, "loss": 0.0004, "step": 23001 }, { "epoch": 2.76, "learning_rate": 1.3477864427114577e-06, "loss": 0.0003, "step": 23002 }, { "epoch": 2.76, "learning_rate": 1.3477144571085783e-06, "loss": 0.0002, "step": 23003 }, { "epoch": 2.76, "learning_rate": 1.3476424715056988e-06, "loss": 0.0002, "step": 23004 }, { "epoch": 2.76, "learning_rate": 1.3475704859028194e-06, "loss": 0.0001, "step": 23005 }, { "epoch": 2.76, "learning_rate": 1.34749850029994e-06, "loss": 0.0001, "step": 23006 }, { "epoch": 2.76, "learning_rate": 1.3474265146970607e-06, "loss": 0.0001, "step": 23007 }, { "epoch": 2.76, "learning_rate": 1.3473545290941813e-06, "loss": 0.98, "step": 23008 }, { "epoch": 2.76, "learning_rate": 1.3472825434913019e-06, "loss": 0.0001, "step": 23009 }, { "epoch": 2.76, "learning_rate": 1.3472105578884224e-06, "loss": 0.0002, "step": 23010 }, { "epoch": 2.76, "learning_rate": 1.347138572285543e-06, "loss": 0.7461, "step": 23011 }, { "epoch": 2.76, "learning_rate": 1.3470665866826636e-06, "loss": 0.0002, "step": 23012 }, { "epoch": 2.76, "learning_rate": 1.346994601079784e-06, "loss": 0.0001, "step": 23013 }, { "epoch": 2.76, "learning_rate": 1.3469226154769047e-06, "loss": 0.0002, "step": 23014 }, { "epoch": 2.76, "learning_rate": 1.3468506298740253e-06, "loss": 0.0001, "step": 23015 }, { "epoch": 2.76, "learning_rate": 1.3467786442711458e-06, "loss": 0.0002, "step": 23016 }, { "epoch": 2.76, "learning_rate": 1.3467066586682664e-06, "loss": 0.0001, "step": 23017 }, { "epoch": 2.76, "learning_rate": 1.346634673065387e-06, "loss": 0.0001, "step": 23018 }, { "epoch": 2.76, "learning_rate": 1.3465626874625075e-06, "loss": 0.0002, "step": 23019 }, { "epoch": 2.76, "learning_rate": 1.346490701859628e-06, "loss": 0.0003, "step": 23020 }, { "epoch": 2.76, "learning_rate": 1.3464187162567487e-06, "loss": 0.0004, "step": 23021 }, { "epoch": 2.76, "learning_rate": 1.3463467306538692e-06, "loss": 0.0002, "step": 23022 }, { "epoch": 2.76, "learning_rate": 1.34627474505099e-06, "loss": 0.0001, "step": 23023 }, { "epoch": 2.76, "learning_rate": 1.3462027594481104e-06, "loss": 0.0001, "step": 23024 }, { "epoch": 2.76, "learning_rate": 1.346130773845231e-06, "loss": 0.0001, "step": 23025 }, { "epoch": 2.76, "learning_rate": 1.3460587882423515e-06, "loss": 0.0002, "step": 23026 }, { "epoch": 2.76, "learning_rate": 1.345986802639472e-06, "loss": 0.0001, "step": 23027 }, { "epoch": 2.76, "learning_rate": 1.3459148170365926e-06, "loss": 0.0005, "step": 23028 }, { "epoch": 2.76, "learning_rate": 1.3458428314337132e-06, "loss": 0.8643, "step": 23029 }, { "epoch": 2.76, "learning_rate": 1.345770845830834e-06, "loss": 0.0001, "step": 23030 }, { "epoch": 2.76, "learning_rate": 1.3456988602279545e-06, "loss": 0.0001, "step": 23031 }, { "epoch": 2.76, "learning_rate": 1.3456268746250751e-06, "loss": 0.0003, "step": 23032 }, { "epoch": 2.76, "learning_rate": 1.3455548890221957e-06, "loss": 0.0002, "step": 23033 }, { "epoch": 2.76, "learning_rate": 1.3454829034193162e-06, "loss": 0.0003, "step": 23034 }, { "epoch": 2.76, "learning_rate": 1.3454109178164368e-06, "loss": 0.0007, "step": 23035 }, { "epoch": 2.76, "learning_rate": 1.3453389322135572e-06, "loss": 0.0001, "step": 23036 }, { "epoch": 2.76, "learning_rate": 1.3452669466106777e-06, "loss": 0.0003, "step": 23037 }, { "epoch": 2.76, "learning_rate": 1.3451949610077985e-06, "loss": 0.0004, "step": 23038 }, { "epoch": 2.76, "learning_rate": 1.345122975404919e-06, "loss": 0.0003, "step": 23039 }, { "epoch": 2.76, "learning_rate": 1.3450509898020396e-06, "loss": 0.0003, "step": 23040 }, { "epoch": 2.76, "learning_rate": 1.3449790041991602e-06, "loss": 0.0002, "step": 23041 }, { "epoch": 2.76, "learning_rate": 1.3449070185962808e-06, "loss": 0.0001, "step": 23042 }, { "epoch": 2.76, "learning_rate": 1.3448350329934013e-06, "loss": 0.0004, "step": 23043 }, { "epoch": 2.76, "learning_rate": 1.344763047390522e-06, "loss": 0.0001, "step": 23044 }, { "epoch": 2.76, "learning_rate": 1.3446910617876425e-06, "loss": 0.0001, "step": 23045 }, { "epoch": 2.76, "learning_rate": 1.344619076184763e-06, "loss": 0.0001, "step": 23046 }, { "epoch": 2.76, "learning_rate": 1.3445470905818836e-06, "loss": 0.0002, "step": 23047 }, { "epoch": 2.76, "learning_rate": 1.3444751049790042e-06, "loss": 0.0006, "step": 23048 }, { "epoch": 2.76, "learning_rate": 1.3444031193761247e-06, "loss": 0.0003, "step": 23049 }, { "epoch": 2.76, "learning_rate": 1.3443311337732453e-06, "loss": 0.0002, "step": 23050 }, { "epoch": 2.76, "learning_rate": 1.3442591481703659e-06, "loss": 0.0005, "step": 23051 }, { "epoch": 2.76, "learning_rate": 1.3441871625674864e-06, "loss": 0.0002, "step": 23052 }, { "epoch": 2.76, "learning_rate": 1.344115176964607e-06, "loss": 0.0002, "step": 23053 }, { "epoch": 2.76, "learning_rate": 1.3440431913617278e-06, "loss": 0.0007, "step": 23054 }, { "epoch": 2.76, "learning_rate": 1.3439712057588484e-06, "loss": 0.0004, "step": 23055 }, { "epoch": 2.76, "learning_rate": 1.343899220155969e-06, "loss": 0.0001, "step": 23056 }, { "epoch": 2.76, "learning_rate": 1.3438272345530895e-06, "loss": 0.0002, "step": 23057 }, { "epoch": 2.76, "learning_rate": 1.34375524895021e-06, "loss": 0.0002, "step": 23058 }, { "epoch": 2.76, "learning_rate": 1.3436832633473304e-06, "loss": 0.0002, "step": 23059 }, { "epoch": 2.76, "learning_rate": 1.343611277744451e-06, "loss": 0.0001, "step": 23060 }, { "epoch": 2.76, "learning_rate": 1.3435392921415718e-06, "loss": 0.0003, "step": 23061 }, { "epoch": 2.76, "learning_rate": 1.3434673065386923e-06, "loss": 0.0003, "step": 23062 }, { "epoch": 2.76, "learning_rate": 1.343395320935813e-06, "loss": 0.0005, "step": 23063 }, { "epoch": 2.76, "learning_rate": 1.3433233353329335e-06, "loss": 0.0001, "step": 23064 }, { "epoch": 2.76, "learning_rate": 1.343251349730054e-06, "loss": 0.0003, "step": 23065 }, { "epoch": 2.76, "learning_rate": 1.3431793641271746e-06, "loss": 0.0001, "step": 23066 }, { "epoch": 2.76, "learning_rate": 1.3431073785242952e-06, "loss": 0.0008, "step": 23067 }, { "epoch": 2.76, "learning_rate": 1.3430353929214157e-06, "loss": 0.0005, "step": 23068 }, { "epoch": 2.76, "learning_rate": 1.3429634073185363e-06, "loss": 0.0002, "step": 23069 }, { "epoch": 2.76, "learning_rate": 1.3428914217156569e-06, "loss": 0.0003, "step": 23070 }, { "epoch": 2.76, "learning_rate": 1.3428194361127774e-06, "loss": 0.0001, "step": 23071 }, { "epoch": 2.76, "learning_rate": 1.342747450509898e-06, "loss": 0.0004, "step": 23072 }, { "epoch": 2.76, "learning_rate": 1.3426754649070186e-06, "loss": 0.0003, "step": 23073 }, { "epoch": 2.77, "learning_rate": 1.3426034793041391e-06, "loss": 0.0002, "step": 23074 }, { "epoch": 2.77, "learning_rate": 1.3425314937012597e-06, "loss": 0.0003, "step": 23075 }, { "epoch": 2.77, "learning_rate": 1.3424595080983803e-06, "loss": 0.0005, "step": 23076 }, { "epoch": 2.77, "learning_rate": 1.342387522495501e-06, "loss": 0.0001, "step": 23077 }, { "epoch": 2.77, "learning_rate": 1.3423155368926216e-06, "loss": 0.0002, "step": 23078 }, { "epoch": 2.77, "learning_rate": 1.3422435512897422e-06, "loss": 0.0009, "step": 23079 }, { "epoch": 2.77, "learning_rate": 1.3421715656868627e-06, "loss": 0.0002, "step": 23080 }, { "epoch": 2.77, "learning_rate": 1.3420995800839833e-06, "loss": 0.0002, "step": 23081 }, { "epoch": 2.77, "learning_rate": 1.3420275944811037e-06, "loss": 0.0004, "step": 23082 }, { "epoch": 2.77, "learning_rate": 1.3419556088782242e-06, "loss": 0.0071, "step": 23083 }, { "epoch": 2.77, "learning_rate": 1.341883623275345e-06, "loss": 0.0002, "step": 23084 }, { "epoch": 2.77, "learning_rate": 1.3418116376724656e-06, "loss": 0.0003, "step": 23085 }, { "epoch": 2.77, "learning_rate": 1.3417396520695861e-06, "loss": 0.0001, "step": 23086 }, { "epoch": 2.77, "learning_rate": 1.3416676664667067e-06, "loss": 0.0001, "step": 23087 }, { "epoch": 2.77, "learning_rate": 1.3415956808638273e-06, "loss": 0.0002, "step": 23088 }, { "epoch": 2.77, "learning_rate": 1.3415236952609478e-06, "loss": 0.0003, "step": 23089 }, { "epoch": 2.77, "learning_rate": 1.3414517096580684e-06, "loss": 0.0002, "step": 23090 }, { "epoch": 2.77, "learning_rate": 1.341379724055189e-06, "loss": 0.0008, "step": 23091 }, { "epoch": 2.77, "learning_rate": 1.3413077384523095e-06, "loss": 0.0002, "step": 23092 }, { "epoch": 2.77, "learning_rate": 1.3412357528494303e-06, "loss": 0.0003, "step": 23093 }, { "epoch": 2.77, "learning_rate": 1.3411637672465507e-06, "loss": 0.0002, "step": 23094 }, { "epoch": 2.77, "learning_rate": 1.3410917816436712e-06, "loss": 0.0002, "step": 23095 }, { "epoch": 2.77, "learning_rate": 1.3410197960407918e-06, "loss": 0.0002, "step": 23096 }, { "epoch": 2.77, "learning_rate": 1.3409478104379124e-06, "loss": 0.0001, "step": 23097 }, { "epoch": 2.77, "learning_rate": 1.340875824835033e-06, "loss": 0.0002, "step": 23098 }, { "epoch": 2.77, "learning_rate": 1.3408038392321535e-06, "loss": 0.0003, "step": 23099 }, { "epoch": 2.77, "learning_rate": 1.3407318536292743e-06, "loss": 0.729, "step": 23100 }, { "epoch": 2.77, "learning_rate": 1.3406598680263949e-06, "loss": 0.0001, "step": 23101 }, { "epoch": 2.77, "learning_rate": 1.3405878824235154e-06, "loss": 0.0001, "step": 23102 }, { "epoch": 2.77, "learning_rate": 1.340515896820636e-06, "loss": 0.0001, "step": 23103 }, { "epoch": 2.77, "learning_rate": 1.3404439112177566e-06, "loss": 0.0003, "step": 23104 }, { "epoch": 2.77, "learning_rate": 1.340371925614877e-06, "loss": 0.0002, "step": 23105 }, { "epoch": 2.77, "learning_rate": 1.3402999400119975e-06, "loss": 0.0005, "step": 23106 }, { "epoch": 2.77, "learning_rate": 1.3402279544091183e-06, "loss": 0.0006, "step": 23107 }, { "epoch": 2.77, "learning_rate": 1.3401559688062388e-06, "loss": 0.0003, "step": 23108 }, { "epoch": 2.77, "learning_rate": 1.3400839832033594e-06, "loss": 0.0001, "step": 23109 }, { "epoch": 2.77, "learning_rate": 1.34001199760048e-06, "loss": 0.0003, "step": 23110 }, { "epoch": 2.77, "learning_rate": 1.3399400119976005e-06, "loss": 0.0079, "step": 23111 }, { "epoch": 2.77, "learning_rate": 1.339868026394721e-06, "loss": 0.0001, "step": 23112 }, { "epoch": 2.77, "learning_rate": 1.3397960407918417e-06, "loss": 0.0002, "step": 23113 }, { "epoch": 2.77, "learning_rate": 1.3397240551889622e-06, "loss": 0.0001, "step": 23114 }, { "epoch": 2.77, "learning_rate": 1.3396520695860828e-06, "loss": 0.0001, "step": 23115 }, { "epoch": 2.77, "learning_rate": 1.3395800839832036e-06, "loss": 0.0003, "step": 23116 }, { "epoch": 2.77, "learning_rate": 1.339508098380324e-06, "loss": 0.0005, "step": 23117 }, { "epoch": 2.77, "learning_rate": 1.3394361127774445e-06, "loss": 0.0003, "step": 23118 }, { "epoch": 2.77, "learning_rate": 1.339364127174565e-06, "loss": 0.0004, "step": 23119 }, { "epoch": 2.77, "learning_rate": 1.3392921415716856e-06, "loss": 0.0001, "step": 23120 }, { "epoch": 2.77, "learning_rate": 1.3392201559688062e-06, "loss": 0.0004, "step": 23121 }, { "epoch": 2.77, "learning_rate": 1.3391481703659268e-06, "loss": 0.0001, "step": 23122 }, { "epoch": 2.77, "learning_rate": 1.3390761847630475e-06, "loss": 0.0004, "step": 23123 }, { "epoch": 2.77, "learning_rate": 1.339004199160168e-06, "loss": 0.0002, "step": 23124 }, { "epoch": 2.77, "learning_rate": 1.3389322135572887e-06, "loss": 0.0002, "step": 23125 }, { "epoch": 2.77, "learning_rate": 1.3388602279544092e-06, "loss": 0.0003, "step": 23126 }, { "epoch": 2.77, "learning_rate": 1.3387882423515298e-06, "loss": 0.0003, "step": 23127 }, { "epoch": 2.77, "learning_rate": 1.3387162567486502e-06, "loss": 0.0001, "step": 23128 }, { "epoch": 2.77, "learning_rate": 1.3386442711457707e-06, "loss": 0.0003, "step": 23129 }, { "epoch": 2.77, "learning_rate": 1.3385722855428915e-06, "loss": 0.0001, "step": 23130 }, { "epoch": 2.77, "learning_rate": 1.338500299940012e-06, "loss": 0.0162, "step": 23131 }, { "epoch": 2.77, "learning_rate": 1.3384283143371326e-06, "loss": 0.0002, "step": 23132 }, { "epoch": 2.77, "learning_rate": 1.3383563287342532e-06, "loss": 0.0002, "step": 23133 }, { "epoch": 2.77, "learning_rate": 1.3382843431313738e-06, "loss": 0.0006, "step": 23134 }, { "epoch": 2.77, "learning_rate": 1.3382123575284943e-06, "loss": 0.0004, "step": 23135 }, { "epoch": 2.77, "learning_rate": 1.338140371925615e-06, "loss": 0.0003, "step": 23136 }, { "epoch": 2.77, "learning_rate": 1.3380683863227355e-06, "loss": 0.0002, "step": 23137 }, { "epoch": 2.77, "learning_rate": 1.337996400719856e-06, "loss": 0.9757, "step": 23138 }, { "epoch": 2.77, "learning_rate": 1.3379244151169768e-06, "loss": 0.0001, "step": 23139 }, { "epoch": 2.77, "learning_rate": 1.3378524295140972e-06, "loss": 0.0003, "step": 23140 }, { "epoch": 2.77, "learning_rate": 1.3377804439112177e-06, "loss": 0.0003, "step": 23141 }, { "epoch": 2.77, "learning_rate": 1.3377084583083383e-06, "loss": 0.0002, "step": 23142 }, { "epoch": 2.77, "learning_rate": 1.3376364727054589e-06, "loss": 0.0001, "step": 23143 }, { "epoch": 2.77, "learning_rate": 1.3375644871025794e-06, "loss": 0.5349, "step": 23144 }, { "epoch": 2.77, "learning_rate": 1.3374925014997e-06, "loss": 0.0002, "step": 23145 }, { "epoch": 2.77, "learning_rate": 1.3374205158968208e-06, "loss": 0.0002, "step": 23146 }, { "epoch": 2.77, "learning_rate": 1.3373485302939414e-06, "loss": 0.0001, "step": 23147 }, { "epoch": 2.77, "learning_rate": 1.337276544691062e-06, "loss": 0.0542, "step": 23148 }, { "epoch": 2.77, "learning_rate": 1.3372045590881825e-06, "loss": 0.0769, "step": 23149 }, { "epoch": 2.77, "learning_rate": 1.337132573485303e-06, "loss": 0.0001, "step": 23150 }, { "epoch": 2.77, "learning_rate": 1.3370605878824234e-06, "loss": 0.0002, "step": 23151 }, { "epoch": 2.77, "learning_rate": 1.336988602279544e-06, "loss": 0.0415, "step": 23152 }, { "epoch": 2.77, "learning_rate": 1.3369166166766648e-06, "loss": 0.0003, "step": 23153 }, { "epoch": 2.77, "learning_rate": 1.3368446310737853e-06, "loss": 0.0003, "step": 23154 }, { "epoch": 2.77, "learning_rate": 1.3367726454709059e-06, "loss": 0.0002, "step": 23155 }, { "epoch": 2.77, "learning_rate": 1.3367006598680265e-06, "loss": 0.0023, "step": 23156 }, { "epoch": 2.77, "learning_rate": 1.336628674265147e-06, "loss": 0.0001, "step": 23157 }, { "epoch": 2.78, "learning_rate": 1.3365566886622676e-06, "loss": 0.0002, "step": 23158 }, { "epoch": 2.78, "learning_rate": 1.3364847030593882e-06, "loss": 0.0002, "step": 23159 }, { "epoch": 2.78, "learning_rate": 1.3364127174565087e-06, "loss": 0.0001, "step": 23160 }, { "epoch": 2.78, "learning_rate": 1.3363407318536293e-06, "loss": 0.8171, "step": 23161 }, { "epoch": 2.78, "learning_rate": 1.33626874625075e-06, "loss": 0.7956, "step": 23162 }, { "epoch": 2.78, "learning_rate": 1.3361967606478704e-06, "loss": 0.0003, "step": 23163 }, { "epoch": 2.78, "learning_rate": 1.336124775044991e-06, "loss": 0.0002, "step": 23164 }, { "epoch": 2.78, "learning_rate": 1.3360527894421116e-06, "loss": 0.8529, "step": 23165 }, { "epoch": 2.78, "learning_rate": 1.3359808038392321e-06, "loss": 0.0001, "step": 23166 }, { "epoch": 2.78, "learning_rate": 1.3359088182363527e-06, "loss": 0.0002, "step": 23167 }, { "epoch": 2.78, "learning_rate": 1.3358368326334733e-06, "loss": 0.0001, "step": 23168 }, { "epoch": 2.78, "learning_rate": 1.3357648470305938e-06, "loss": 0.0014, "step": 23169 }, { "epoch": 2.78, "learning_rate": 1.3356928614277146e-06, "loss": 0.0004, "step": 23170 }, { "epoch": 2.78, "learning_rate": 1.3356208758248352e-06, "loss": 0.0003, "step": 23171 }, { "epoch": 2.78, "learning_rate": 1.3355488902219557e-06, "loss": 0.0001, "step": 23172 }, { "epoch": 2.78, "learning_rate": 1.3354769046190763e-06, "loss": 0.0003, "step": 23173 }, { "epoch": 2.78, "learning_rate": 1.3354049190161967e-06, "loss": 0.0002, "step": 23174 }, { "epoch": 2.78, "learning_rate": 1.3353329334133172e-06, "loss": 0.0003, "step": 23175 }, { "epoch": 2.78, "learning_rate": 1.3352609478104378e-06, "loss": 0.0002, "step": 23176 }, { "epoch": 2.78, "learning_rate": 1.3351889622075586e-06, "loss": 0.0003, "step": 23177 }, { "epoch": 2.78, "learning_rate": 1.3351169766046791e-06, "loss": 0.0005, "step": 23178 }, { "epoch": 2.78, "learning_rate": 1.3350449910017997e-06, "loss": 0.0003, "step": 23179 }, { "epoch": 2.78, "learning_rate": 1.3349730053989203e-06, "loss": 0.0003, "step": 23180 }, { "epoch": 2.78, "learning_rate": 1.3349010197960408e-06, "loss": 0.0005, "step": 23181 }, { "epoch": 2.78, "learning_rate": 1.3348290341931614e-06, "loss": 0.9224, "step": 23182 }, { "epoch": 2.78, "learning_rate": 1.334757048590282e-06, "loss": 0.0002, "step": 23183 }, { "epoch": 2.78, "learning_rate": 1.3346850629874025e-06, "loss": 0.001, "step": 23184 }, { "epoch": 2.78, "learning_rate": 1.334613077384523e-06, "loss": 0.0003, "step": 23185 }, { "epoch": 2.78, "learning_rate": 1.3345410917816437e-06, "loss": 0.0017, "step": 23186 }, { "epoch": 2.78, "learning_rate": 1.3344691061787642e-06, "loss": 0.0006, "step": 23187 }, { "epoch": 2.78, "learning_rate": 1.3343971205758848e-06, "loss": 0.0006, "step": 23188 }, { "epoch": 2.78, "learning_rate": 1.3343251349730054e-06, "loss": 0.0004, "step": 23189 }, { "epoch": 2.78, "learning_rate": 1.334253149370126e-06, "loss": 0.0001, "step": 23190 }, { "epoch": 2.78, "learning_rate": 1.3341811637672465e-06, "loss": 0.0005, "step": 23191 }, { "epoch": 2.78, "learning_rate": 1.334109178164367e-06, "loss": 0.0616, "step": 23192 }, { "epoch": 2.78, "learning_rate": 1.3340371925614878e-06, "loss": 0.0004, "step": 23193 }, { "epoch": 2.78, "learning_rate": 1.3339652069586084e-06, "loss": 0.0009, "step": 23194 }, { "epoch": 2.78, "learning_rate": 1.333893221355729e-06, "loss": 0.0002, "step": 23195 }, { "epoch": 2.78, "learning_rate": 1.3338212357528495e-06, "loss": 0.0003, "step": 23196 }, { "epoch": 2.78, "learning_rate": 1.3337492501499701e-06, "loss": 0.0002, "step": 23197 }, { "epoch": 2.78, "learning_rate": 1.3336772645470905e-06, "loss": 0.0003, "step": 23198 }, { "epoch": 2.78, "learning_rate": 1.333605278944211e-06, "loss": 0.0004, "step": 23199 }, { "epoch": 2.78, "learning_rate": 1.3335332933413318e-06, "loss": 0.0005, "step": 23200 }, { "epoch": 2.78, "learning_rate": 1.3334613077384524e-06, "loss": 0.0004, "step": 23201 }, { "epoch": 2.78, "learning_rate": 1.333389322135573e-06, "loss": 0.001, "step": 23202 }, { "epoch": 2.78, "learning_rate": 1.3333173365326935e-06, "loss": 0.0001, "step": 23203 }, { "epoch": 2.78, "learning_rate": 1.333245350929814e-06, "loss": 0.0002, "step": 23204 }, { "epoch": 2.78, "learning_rate": 1.3331733653269347e-06, "loss": 0.0002, "step": 23205 }, { "epoch": 2.78, "learning_rate": 1.3331013797240552e-06, "loss": 0.0002, "step": 23206 }, { "epoch": 2.78, "learning_rate": 1.3330293941211758e-06, "loss": 0.0008, "step": 23207 }, { "epoch": 2.78, "learning_rate": 1.3329574085182964e-06, "loss": 0.0036, "step": 23208 }, { "epoch": 2.78, "learning_rate": 1.332885422915417e-06, "loss": 0.0002, "step": 23209 }, { "epoch": 2.78, "learning_rate": 1.3328134373125375e-06, "loss": 0.0003, "step": 23210 }, { "epoch": 2.78, "learning_rate": 1.332741451709658e-06, "loss": 0.0003, "step": 23211 }, { "epoch": 2.78, "learning_rate": 1.3326694661067786e-06, "loss": 0.0002, "step": 23212 }, { "epoch": 2.78, "learning_rate": 1.3325974805038992e-06, "loss": 0.0005, "step": 23213 }, { "epoch": 2.78, "learning_rate": 1.3325254949010198e-06, "loss": 0.0002, "step": 23214 }, { "epoch": 2.78, "learning_rate": 1.3324535092981403e-06, "loss": 0.0001, "step": 23215 }, { "epoch": 2.78, "learning_rate": 1.332381523695261e-06, "loss": 0.0003, "step": 23216 }, { "epoch": 2.78, "learning_rate": 1.3323095380923817e-06, "loss": 0.0004, "step": 23217 }, { "epoch": 2.78, "learning_rate": 1.3322375524895022e-06, "loss": 0.0006, "step": 23218 }, { "epoch": 2.78, "learning_rate": 1.3321655668866228e-06, "loss": 0.0002, "step": 23219 }, { "epoch": 2.78, "learning_rate": 1.3320935812837434e-06, "loss": 0.7245, "step": 23220 }, { "epoch": 2.78, "learning_rate": 1.3320215956808637e-06, "loss": 0.7247, "step": 23221 }, { "epoch": 2.78, "learning_rate": 1.3319496100779843e-06, "loss": 0.0001, "step": 23222 }, { "epoch": 2.78, "learning_rate": 1.331877624475105e-06, "loss": 0.0001, "step": 23223 }, { "epoch": 2.78, "learning_rate": 1.3318056388722256e-06, "loss": 0.0001, "step": 23224 }, { "epoch": 2.78, "learning_rate": 1.3317336532693462e-06, "loss": 0.0003, "step": 23225 }, { "epoch": 2.78, "learning_rate": 1.3316616676664668e-06, "loss": 0.0002, "step": 23226 }, { "epoch": 2.78, "learning_rate": 1.3315896820635873e-06, "loss": 0.0002, "step": 23227 }, { "epoch": 2.78, "learning_rate": 1.331517696460708e-06, "loss": 0.0014, "step": 23228 }, { "epoch": 2.78, "learning_rate": 1.3314457108578285e-06, "loss": 0.0001, "step": 23229 }, { "epoch": 2.78, "learning_rate": 1.331373725254949e-06, "loss": 0.0001, "step": 23230 }, { "epoch": 2.78, "learning_rate": 1.3313017396520696e-06, "loss": 0.0011, "step": 23231 }, { "epoch": 2.78, "learning_rate": 1.3312297540491902e-06, "loss": 0.0004, "step": 23232 }, { "epoch": 2.78, "learning_rate": 1.3311577684463107e-06, "loss": 0.0001, "step": 23233 }, { "epoch": 2.78, "learning_rate": 1.3310857828434313e-06, "loss": 0.0003, "step": 23234 }, { "epoch": 2.78, "learning_rate": 1.3310137972405519e-06, "loss": 0.0004, "step": 23235 }, { "epoch": 2.78, "learning_rate": 1.3309418116376724e-06, "loss": 0.0001, "step": 23236 }, { "epoch": 2.78, "learning_rate": 1.330869826034793e-06, "loss": 0.0004, "step": 23237 }, { "epoch": 2.78, "learning_rate": 1.3307978404319136e-06, "loss": 0.0004, "step": 23238 }, { "epoch": 2.78, "learning_rate": 1.3307258548290343e-06, "loss": 0.0002, "step": 23239 }, { "epoch": 2.78, "learning_rate": 1.330653869226155e-06, "loss": 0.0005, "step": 23240 }, { "epoch": 2.79, "learning_rate": 1.3305818836232755e-06, "loss": 0.0004, "step": 23241 }, { "epoch": 2.79, "learning_rate": 1.330509898020396e-06, "loss": 0.0003, "step": 23242 }, { "epoch": 2.79, "learning_rate": 1.3304379124175166e-06, "loss": 0.0005, "step": 23243 }, { "epoch": 2.79, "learning_rate": 1.330365926814637e-06, "loss": 0.0002, "step": 23244 }, { "epoch": 2.79, "learning_rate": 1.3302939412117575e-06, "loss": 0.0004, "step": 23245 }, { "epoch": 2.79, "learning_rate": 1.3302219556088783e-06, "loss": 0.0003, "step": 23246 }, { "epoch": 2.79, "learning_rate": 1.3301499700059989e-06, "loss": 0.0014, "step": 23247 }, { "epoch": 2.79, "learning_rate": 1.3300779844031194e-06, "loss": 0.0002, "step": 23248 }, { "epoch": 2.79, "learning_rate": 1.33000599880024e-06, "loss": 0.0003, "step": 23249 }, { "epoch": 2.79, "learning_rate": 1.3299340131973606e-06, "loss": 0.0003, "step": 23250 }, { "epoch": 2.79, "learning_rate": 1.3298620275944811e-06, "loss": 0.0002, "step": 23251 }, { "epoch": 2.79, "learning_rate": 1.3297900419916017e-06, "loss": 0.0001, "step": 23252 }, { "epoch": 2.79, "learning_rate": 1.3297180563887223e-06, "loss": 0.0002, "step": 23253 }, { "epoch": 2.79, "learning_rate": 1.3296460707858428e-06, "loss": 0.0003, "step": 23254 }, { "epoch": 2.79, "learning_rate": 1.3295740851829634e-06, "loss": 0.0001, "step": 23255 }, { "epoch": 2.79, "learning_rate": 1.329502099580084e-06, "loss": 0.0008, "step": 23256 }, { "epoch": 2.79, "learning_rate": 1.3294301139772045e-06, "loss": 0.0002, "step": 23257 }, { "epoch": 2.79, "learning_rate": 1.3293581283743251e-06, "loss": 0.0001, "step": 23258 }, { "epoch": 2.79, "learning_rate": 1.3292861427714457e-06, "loss": 0.0083, "step": 23259 }, { "epoch": 2.79, "learning_rate": 1.3292141571685662e-06, "loss": 0.0002, "step": 23260 }, { "epoch": 2.79, "learning_rate": 1.3291421715656868e-06, "loss": 0.0003, "step": 23261 }, { "epoch": 2.79, "learning_rate": 1.3290701859628076e-06, "loss": 0.0002, "step": 23262 }, { "epoch": 2.79, "learning_rate": 1.3289982003599282e-06, "loss": 0.0002, "step": 23263 }, { "epoch": 2.79, "learning_rate": 1.3289262147570487e-06, "loss": 0.0001, "step": 23264 }, { "epoch": 2.79, "learning_rate": 1.3288542291541693e-06, "loss": 0.0002, "step": 23265 }, { "epoch": 2.79, "learning_rate": 1.3287822435512899e-06, "loss": 0.0001, "step": 23266 }, { "epoch": 2.79, "learning_rate": 1.3287102579484102e-06, "loss": 0.0001, "step": 23267 }, { "epoch": 2.79, "learning_rate": 1.3286382723455308e-06, "loss": 0.0001, "step": 23268 }, { "epoch": 2.79, "learning_rate": 1.3285662867426516e-06, "loss": 0.0009, "step": 23269 }, { "epoch": 2.79, "learning_rate": 1.3284943011397721e-06, "loss": 0.0003, "step": 23270 }, { "epoch": 2.79, "learning_rate": 1.3284223155368927e-06, "loss": 0.0002, "step": 23271 }, { "epoch": 2.79, "learning_rate": 1.3283503299340133e-06, "loss": 0.0001, "step": 23272 }, { "epoch": 2.79, "learning_rate": 1.3282783443311338e-06, "loss": 0.0002, "step": 23273 }, { "epoch": 2.79, "learning_rate": 1.3282063587282544e-06, "loss": 0.0007, "step": 23274 }, { "epoch": 2.79, "learning_rate": 1.328134373125375e-06, "loss": 0.0001, "step": 23275 }, { "epoch": 2.79, "learning_rate": 1.3280623875224955e-06, "loss": 0.0006, "step": 23276 }, { "epoch": 2.79, "learning_rate": 1.327990401919616e-06, "loss": 0.0001, "step": 23277 }, { "epoch": 2.79, "learning_rate": 1.3279184163167367e-06, "loss": 0.0001, "step": 23278 }, { "epoch": 2.79, "learning_rate": 1.3278464307138572e-06, "loss": 0.0002, "step": 23279 }, { "epoch": 2.79, "learning_rate": 1.3277744451109778e-06, "loss": 0.0001, "step": 23280 }, { "epoch": 2.79, "learning_rate": 1.3277024595080984e-06, "loss": 0.0001, "step": 23281 }, { "epoch": 2.79, "learning_rate": 1.327630473905219e-06, "loss": 0.0273, "step": 23282 }, { "epoch": 2.79, "learning_rate": 1.3275584883023395e-06, "loss": 0.0001, "step": 23283 }, { "epoch": 2.79, "learning_rate": 1.32748650269946e-06, "loss": 0.0001, "step": 23284 }, { "epoch": 2.79, "learning_rate": 1.3274145170965808e-06, "loss": 0.0001, "step": 23285 }, { "epoch": 2.79, "learning_rate": 1.3273425314937014e-06, "loss": 0.9414, "step": 23286 }, { "epoch": 2.79, "learning_rate": 1.327270545890822e-06, "loss": 0.0007, "step": 23287 }, { "epoch": 2.79, "learning_rate": 1.3271985602879425e-06, "loss": 0.0002, "step": 23288 }, { "epoch": 2.79, "learning_rate": 1.3271265746850631e-06, "loss": 0.0001, "step": 23289 }, { "epoch": 2.79, "learning_rate": 1.3270545890821835e-06, "loss": 0.0001, "step": 23290 }, { "epoch": 2.79, "learning_rate": 1.326982603479304e-06, "loss": 0.0002, "step": 23291 }, { "epoch": 2.79, "learning_rate": 1.3269106178764248e-06, "loss": 0.0001, "step": 23292 }, { "epoch": 2.79, "learning_rate": 1.3268386322735454e-06, "loss": 0.0002, "step": 23293 }, { "epoch": 2.79, "learning_rate": 1.326766646670666e-06, "loss": 0.0008, "step": 23294 }, { "epoch": 2.79, "learning_rate": 1.3266946610677865e-06, "loss": 0.0002, "step": 23295 }, { "epoch": 2.79, "learning_rate": 1.326622675464907e-06, "loss": 0.0004, "step": 23296 }, { "epoch": 2.79, "learning_rate": 1.3265506898620276e-06, "loss": 0.0012, "step": 23297 }, { "epoch": 2.79, "learning_rate": 1.3264787042591482e-06, "loss": 0.0001, "step": 23298 }, { "epoch": 2.79, "learning_rate": 1.3264067186562688e-06, "loss": 0.0003, "step": 23299 }, { "epoch": 2.79, "learning_rate": 1.3263347330533893e-06, "loss": 0.0003, "step": 23300 }, { "epoch": 2.79, "learning_rate": 1.32626274745051e-06, "loss": 0.0004, "step": 23301 }, { "epoch": 2.79, "learning_rate": 1.3261907618476305e-06, "loss": 0.0001, "step": 23302 }, { "epoch": 2.79, "learning_rate": 1.326118776244751e-06, "loss": 0.0001, "step": 23303 }, { "epoch": 2.79, "learning_rate": 1.3260467906418716e-06, "loss": 0.0001, "step": 23304 }, { "epoch": 2.79, "learning_rate": 1.3259748050389922e-06, "loss": 0.0001, "step": 23305 }, { "epoch": 2.79, "learning_rate": 1.3259028194361127e-06, "loss": 0.0006, "step": 23306 }, { "epoch": 2.79, "learning_rate": 1.3258308338332333e-06, "loss": 0.0002, "step": 23307 }, { "epoch": 2.79, "learning_rate": 1.3257588482303539e-06, "loss": 0.0003, "step": 23308 }, { "epoch": 2.79, "learning_rate": 1.3256868626274747e-06, "loss": 0.0001, "step": 23309 }, { "epoch": 2.79, "learning_rate": 1.3256148770245952e-06, "loss": 0.0005, "step": 23310 }, { "epoch": 2.79, "learning_rate": 1.3255428914217158e-06, "loss": 0.0001, "step": 23311 }, { "epoch": 2.79, "learning_rate": 1.3254709058188364e-06, "loss": 0.0004, "step": 23312 }, { "epoch": 2.79, "learning_rate": 1.3253989202159567e-06, "loss": 0.0001, "step": 23313 }, { "epoch": 2.79, "learning_rate": 1.3253269346130773e-06, "loss": 0.0004, "step": 23314 }, { "epoch": 2.79, "learning_rate": 1.3252549490101978e-06, "loss": 0.0001, "step": 23315 }, { "epoch": 2.79, "learning_rate": 1.3251829634073186e-06, "loss": 0.0001, "step": 23316 }, { "epoch": 2.79, "learning_rate": 1.3251109778044392e-06, "loss": 0.0, "step": 23317 }, { "epoch": 2.79, "learning_rate": 1.3250389922015598e-06, "loss": 0.0003, "step": 23318 }, { "epoch": 2.79, "learning_rate": 1.3249670065986803e-06, "loss": 0.0001, "step": 23319 }, { "epoch": 2.79, "learning_rate": 1.3248950209958009e-06, "loss": 0.0001, "step": 23320 }, { "epoch": 2.79, "learning_rate": 1.3248230353929215e-06, "loss": 0.0003, "step": 23321 }, { "epoch": 2.79, "learning_rate": 1.324751049790042e-06, "loss": 0.0002, "step": 23322 }, { "epoch": 2.79, "learning_rate": 1.3246790641871626e-06, "loss": 0.0108, "step": 23323 }, { "epoch": 2.79, "learning_rate": 1.3246070785842832e-06, "loss": 0.0002, "step": 23324 }, { "epoch": 2.8, "learning_rate": 1.3245350929814037e-06, "loss": 0.0002, "step": 23325 }, { "epoch": 2.8, "learning_rate": 1.3244631073785243e-06, "loss": 0.0001, "step": 23326 }, { "epoch": 2.8, "learning_rate": 1.3243911217756449e-06, "loss": 0.0002, "step": 23327 }, { "epoch": 2.8, "learning_rate": 1.3243191361727654e-06, "loss": 0.0001, "step": 23328 }, { "epoch": 2.8, "learning_rate": 1.324247150569886e-06, "loss": 0.0005, "step": 23329 }, { "epoch": 2.8, "learning_rate": 1.3241751649670066e-06, "loss": 0.0001, "step": 23330 }, { "epoch": 2.8, "learning_rate": 1.3241031793641271e-06, "loss": 0.0002, "step": 23331 }, { "epoch": 2.8, "learning_rate": 1.324031193761248e-06, "loss": 0.0002, "step": 23332 }, { "epoch": 2.8, "learning_rate": 1.3239592081583685e-06, "loss": 0.0002, "step": 23333 }, { "epoch": 2.8, "learning_rate": 1.323887222555489e-06, "loss": 0.0001, "step": 23334 }, { "epoch": 2.8, "learning_rate": 1.3238152369526096e-06, "loss": 0.0002, "step": 23335 }, { "epoch": 2.8, "learning_rate": 1.32374325134973e-06, "loss": 0.0001, "step": 23336 }, { "epoch": 2.8, "learning_rate": 1.3236712657468505e-06, "loss": 0.0001, "step": 23337 }, { "epoch": 2.8, "learning_rate": 1.323599280143971e-06, "loss": 0.0002, "step": 23338 }, { "epoch": 2.8, "learning_rate": 1.3235272945410919e-06, "loss": 0.0002, "step": 23339 }, { "epoch": 2.8, "learning_rate": 1.3234553089382124e-06, "loss": 0.0003, "step": 23340 }, { "epoch": 2.8, "learning_rate": 1.323383323335333e-06, "loss": 0.0001, "step": 23341 }, { "epoch": 2.8, "learning_rate": 1.3233113377324536e-06, "loss": 0.0001, "step": 23342 }, { "epoch": 2.8, "learning_rate": 1.3232393521295741e-06, "loss": 0.0001, "step": 23343 }, { "epoch": 2.8, "learning_rate": 1.3231673665266947e-06, "loss": 0.0002, "step": 23344 }, { "epoch": 2.8, "learning_rate": 1.3230953809238153e-06, "loss": 0.0004, "step": 23345 }, { "epoch": 2.8, "learning_rate": 1.3230233953209358e-06, "loss": 0.0002, "step": 23346 }, { "epoch": 2.8, "learning_rate": 1.3229514097180564e-06, "loss": 0.0001, "step": 23347 }, { "epoch": 2.8, "learning_rate": 1.322879424115177e-06, "loss": 0.0002, "step": 23348 }, { "epoch": 2.8, "learning_rate": 1.3228074385122975e-06, "loss": 0.0007, "step": 23349 }, { "epoch": 2.8, "learning_rate": 1.322735452909418e-06, "loss": 0.0002, "step": 23350 }, { "epoch": 2.8, "learning_rate": 1.3226634673065387e-06, "loss": 0.0002, "step": 23351 }, { "epoch": 2.8, "learning_rate": 1.3225914817036592e-06, "loss": 0.0001, "step": 23352 }, { "epoch": 2.8, "learning_rate": 1.3225194961007798e-06, "loss": 0.0001, "step": 23353 }, { "epoch": 2.8, "learning_rate": 1.3224475104979004e-06, "loss": 0.0002, "step": 23354 }, { "epoch": 2.8, "learning_rate": 1.3223755248950212e-06, "loss": 0.0002, "step": 23355 }, { "epoch": 2.8, "learning_rate": 1.3223035392921417e-06, "loss": 0.0003, "step": 23356 }, { "epoch": 2.8, "learning_rate": 1.3222315536892623e-06, "loss": 0.0005, "step": 23357 }, { "epoch": 2.8, "learning_rate": 1.3221595680863829e-06, "loss": 0.0004, "step": 23358 }, { "epoch": 2.8, "learning_rate": 1.3220875824835032e-06, "loss": 0.0003, "step": 23359 }, { "epoch": 2.8, "learning_rate": 1.3220155968806238e-06, "loss": 0.0005, "step": 23360 }, { "epoch": 2.8, "learning_rate": 1.3219436112777443e-06, "loss": 0.0002, "step": 23361 }, { "epoch": 2.8, "learning_rate": 1.3218716256748651e-06, "loss": 0.0001, "step": 23362 }, { "epoch": 2.8, "learning_rate": 1.3217996400719857e-06, "loss": 0.0002, "step": 23363 }, { "epoch": 2.8, "learning_rate": 1.3217276544691063e-06, "loss": 0.0001, "step": 23364 }, { "epoch": 2.8, "learning_rate": 1.3216556688662268e-06, "loss": 0.0001, "step": 23365 }, { "epoch": 2.8, "learning_rate": 1.3215836832633474e-06, "loss": 0.0001, "step": 23366 }, { "epoch": 2.8, "learning_rate": 1.321511697660468e-06, "loss": 0.0001, "step": 23367 }, { "epoch": 2.8, "learning_rate": 1.3214397120575885e-06, "loss": 0.0001, "step": 23368 }, { "epoch": 2.8, "learning_rate": 1.321367726454709e-06, "loss": 0.0002, "step": 23369 }, { "epoch": 2.8, "learning_rate": 1.3212957408518297e-06, "loss": 0.0003, "step": 23370 }, { "epoch": 2.8, "learning_rate": 1.3212237552489502e-06, "loss": 0.0002, "step": 23371 }, { "epoch": 2.8, "learning_rate": 1.3211517696460708e-06, "loss": 0.2255, "step": 23372 }, { "epoch": 2.8, "learning_rate": 1.3210797840431914e-06, "loss": 0.0001, "step": 23373 }, { "epoch": 2.8, "learning_rate": 1.321007798440312e-06, "loss": 0.0002, "step": 23374 }, { "epoch": 2.8, "learning_rate": 1.3209358128374325e-06, "loss": 0.1041, "step": 23375 }, { "epoch": 2.8, "learning_rate": 1.320863827234553e-06, "loss": 0.0001, "step": 23376 }, { "epoch": 2.8, "learning_rate": 1.3207918416316736e-06, "loss": 0.0002, "step": 23377 }, { "epoch": 2.8, "learning_rate": 1.3207198560287944e-06, "loss": 0.0002, "step": 23378 }, { "epoch": 2.8, "learning_rate": 1.320647870425915e-06, "loss": 0.0002, "step": 23379 }, { "epoch": 2.8, "learning_rate": 1.3205758848230355e-06, "loss": 0.0003, "step": 23380 }, { "epoch": 2.8, "learning_rate": 1.320503899220156e-06, "loss": 0.0001, "step": 23381 }, { "epoch": 2.8, "learning_rate": 1.3204319136172765e-06, "loss": 0.0001, "step": 23382 }, { "epoch": 2.8, "learning_rate": 1.320359928014397e-06, "loss": 0.0001, "step": 23383 }, { "epoch": 2.8, "learning_rate": 1.3202879424115176e-06, "loss": 0.0002, "step": 23384 }, { "epoch": 2.8, "learning_rate": 1.3202159568086384e-06, "loss": 0.0001, "step": 23385 }, { "epoch": 2.8, "learning_rate": 1.320143971205759e-06, "loss": 0.0003, "step": 23386 }, { "epoch": 2.8, "learning_rate": 1.3200719856028795e-06, "loss": 0.0002, "step": 23387 }, { "epoch": 2.8, "learning_rate": 1.32e-06, "loss": 0.0017, "step": 23388 }, { "epoch": 2.8, "learning_rate": 1.3199280143971206e-06, "loss": 0.0004, "step": 23389 }, { "epoch": 2.8, "learning_rate": 1.3198560287942412e-06, "loss": 0.0002, "step": 23390 }, { "epoch": 2.8, "learning_rate": 1.3197840431913618e-06, "loss": 0.0001, "step": 23391 }, { "epoch": 2.8, "learning_rate": 1.3197120575884823e-06, "loss": 0.0001, "step": 23392 }, { "epoch": 2.8, "learning_rate": 1.319640071985603e-06, "loss": 0.0001, "step": 23393 }, { "epoch": 2.8, "learning_rate": 1.3195680863827235e-06, "loss": 0.0001, "step": 23394 }, { "epoch": 2.8, "learning_rate": 1.319496100779844e-06, "loss": 0.0002, "step": 23395 }, { "epoch": 2.8, "learning_rate": 1.3194241151769646e-06, "loss": 0.0001, "step": 23396 }, { "epoch": 2.8, "learning_rate": 1.3193521295740852e-06, "loss": 0.0004, "step": 23397 }, { "epoch": 2.8, "learning_rate": 1.3192801439712057e-06, "loss": 0.0001, "step": 23398 }, { "epoch": 2.8, "learning_rate": 1.3192081583683263e-06, "loss": 0.0037, "step": 23399 }, { "epoch": 2.8, "learning_rate": 1.3191361727654469e-06, "loss": 0.0007, "step": 23400 }, { "epoch": 2.8, "learning_rate": 1.3190641871625676e-06, "loss": 0.0002, "step": 23401 }, { "epoch": 2.8, "learning_rate": 1.3189922015596882e-06, "loss": 0.0003, "step": 23402 }, { "epoch": 2.8, "learning_rate": 1.3189202159568088e-06, "loss": 0.0001, "step": 23403 }, { "epoch": 2.8, "learning_rate": 1.3188482303539293e-06, "loss": 0.0002, "step": 23404 }, { "epoch": 2.8, "learning_rate": 1.3187762447510497e-06, "loss": 0.0001, "step": 23405 }, { "epoch": 2.8, "learning_rate": 1.3187042591481703e-06, "loss": 0.9499, "step": 23406 }, { "epoch": 2.8, "learning_rate": 1.3186322735452908e-06, "loss": 0.0016, "step": 23407 }, { "epoch": 2.81, "learning_rate": 1.3185602879424116e-06, "loss": 0.0003, "step": 23408 }, { "epoch": 2.81, "learning_rate": 1.3184883023395322e-06, "loss": 0.0002, "step": 23409 }, { "epoch": 2.81, "learning_rate": 1.3184163167366527e-06, "loss": 1.1694, "step": 23410 }, { "epoch": 2.81, "learning_rate": 1.3183443311337733e-06, "loss": 0.0015, "step": 23411 }, { "epoch": 2.81, "learning_rate": 1.3182723455308939e-06, "loss": 0.0002, "step": 23412 }, { "epoch": 2.81, "learning_rate": 1.3182003599280144e-06, "loss": 0.0001, "step": 23413 }, { "epoch": 2.81, "learning_rate": 1.318128374325135e-06, "loss": 1.0288, "step": 23414 }, { "epoch": 2.81, "learning_rate": 1.3180563887222556e-06, "loss": 0.0002, "step": 23415 }, { "epoch": 2.81, "learning_rate": 1.3179844031193761e-06, "loss": 0.0004, "step": 23416 }, { "epoch": 2.81, "learning_rate": 1.3179124175164967e-06, "loss": 0.0007, "step": 23417 }, { "epoch": 2.81, "learning_rate": 1.3178404319136173e-06, "loss": 0.0002, "step": 23418 }, { "epoch": 2.81, "learning_rate": 1.3177684463107378e-06, "loss": 0.0002, "step": 23419 }, { "epoch": 2.81, "learning_rate": 1.3176964607078584e-06, "loss": 0.0003, "step": 23420 }, { "epoch": 2.81, "learning_rate": 1.317624475104979e-06, "loss": 0.0002, "step": 23421 }, { "epoch": 2.81, "learning_rate": 1.3175524895020996e-06, "loss": 0.0003, "step": 23422 }, { "epoch": 2.81, "learning_rate": 1.3174805038992201e-06, "loss": 0.0002, "step": 23423 }, { "epoch": 2.81, "learning_rate": 1.317408518296341e-06, "loss": 0.0003, "step": 23424 }, { "epoch": 2.81, "learning_rate": 1.3173365326934615e-06, "loss": 0.0002, "step": 23425 }, { "epoch": 2.81, "learning_rate": 1.317264547090582e-06, "loss": 0.0002, "step": 23426 }, { "epoch": 2.81, "learning_rate": 1.3171925614877026e-06, "loss": 0.0004, "step": 23427 }, { "epoch": 2.81, "learning_rate": 1.317120575884823e-06, "loss": 0.0001, "step": 23428 }, { "epoch": 2.81, "learning_rate": 1.3170485902819435e-06, "loss": 0.0002, "step": 23429 }, { "epoch": 2.81, "learning_rate": 1.316976604679064e-06, "loss": 0.0002, "step": 23430 }, { "epoch": 2.81, "learning_rate": 1.3169046190761849e-06, "loss": 0.0001, "step": 23431 }, { "epoch": 2.81, "learning_rate": 1.3168326334733054e-06, "loss": 0.0002, "step": 23432 }, { "epoch": 2.81, "learning_rate": 1.316760647870426e-06, "loss": 0.0003, "step": 23433 }, { "epoch": 2.81, "learning_rate": 1.3166886622675466e-06, "loss": 0.0004, "step": 23434 }, { "epoch": 2.81, "learning_rate": 1.3166166766646671e-06, "loss": 0.0001, "step": 23435 }, { "epoch": 2.81, "learning_rate": 1.3165446910617877e-06, "loss": 0.0002, "step": 23436 }, { "epoch": 2.81, "learning_rate": 1.3164727054589083e-06, "loss": 0.0001, "step": 23437 }, { "epoch": 2.81, "learning_rate": 1.3164007198560288e-06, "loss": 0.0002, "step": 23438 }, { "epoch": 2.81, "learning_rate": 1.3163287342531494e-06, "loss": 0.0002, "step": 23439 }, { "epoch": 2.81, "learning_rate": 1.31625674865027e-06, "loss": 0.0003, "step": 23440 }, { "epoch": 2.81, "learning_rate": 1.3161847630473905e-06, "loss": 0.0003, "step": 23441 }, { "epoch": 2.81, "learning_rate": 1.316112777444511e-06, "loss": 0.0004, "step": 23442 }, { "epoch": 2.81, "learning_rate": 1.3160407918416317e-06, "loss": 0.0001, "step": 23443 }, { "epoch": 2.81, "learning_rate": 1.3159688062387522e-06, "loss": 0.0001, "step": 23444 }, { "epoch": 2.81, "learning_rate": 1.3158968206358728e-06, "loss": 0.0004, "step": 23445 }, { "epoch": 2.81, "learning_rate": 1.3158248350329934e-06, "loss": 0.0002, "step": 23446 }, { "epoch": 2.81, "learning_rate": 1.315752849430114e-06, "loss": 0.0003, "step": 23447 }, { "epoch": 2.81, "learning_rate": 1.3156808638272347e-06, "loss": 0.0002, "step": 23448 }, { "epoch": 2.81, "learning_rate": 1.3156088782243553e-06, "loss": 0.0003, "step": 23449 }, { "epoch": 2.81, "learning_rate": 1.3155368926214758e-06, "loss": 0.0002, "step": 23450 }, { "epoch": 2.81, "learning_rate": 1.3154649070185964e-06, "loss": 0.0002, "step": 23451 }, { "epoch": 2.81, "learning_rate": 1.3153929214157168e-06, "loss": 0.0001, "step": 23452 }, { "epoch": 2.81, "learning_rate": 1.3153209358128373e-06, "loss": 0.0002, "step": 23453 }, { "epoch": 2.81, "learning_rate": 1.315248950209958e-06, "loss": 0.0003, "step": 23454 }, { "epoch": 2.81, "learning_rate": 1.3151769646070787e-06, "loss": 0.0002, "step": 23455 }, { "epoch": 2.81, "learning_rate": 1.3151049790041992e-06, "loss": 0.0001, "step": 23456 }, { "epoch": 2.81, "learning_rate": 1.3150329934013198e-06, "loss": 0.0001, "step": 23457 }, { "epoch": 2.81, "learning_rate": 1.3149610077984404e-06, "loss": 0.0002, "step": 23458 }, { "epoch": 2.81, "learning_rate": 1.314889022195561e-06, "loss": 0.0003, "step": 23459 }, { "epoch": 2.81, "learning_rate": 1.3148170365926815e-06, "loss": 0.0003, "step": 23460 }, { "epoch": 2.81, "learning_rate": 1.314745050989802e-06, "loss": 0.0001, "step": 23461 }, { "epoch": 2.81, "learning_rate": 1.3146730653869226e-06, "loss": 0.0002, "step": 23462 }, { "epoch": 2.81, "learning_rate": 1.3146010797840432e-06, "loss": 0.0003, "step": 23463 }, { "epoch": 2.81, "learning_rate": 1.3145290941811638e-06, "loss": 0.0001, "step": 23464 }, { "epoch": 2.81, "learning_rate": 1.3144571085782843e-06, "loss": 0.0004, "step": 23465 }, { "epoch": 2.81, "learning_rate": 1.314385122975405e-06, "loss": 0.0002, "step": 23466 }, { "epoch": 2.81, "learning_rate": 1.3143131373725255e-06, "loss": 0.0001, "step": 23467 }, { "epoch": 2.81, "learning_rate": 1.314241151769646e-06, "loss": 0.0004, "step": 23468 }, { "epoch": 2.81, "learning_rate": 1.3141691661667666e-06, "loss": 0.0001, "step": 23469 }, { "epoch": 2.81, "learning_rate": 1.3140971805638872e-06, "loss": 0.0001, "step": 23470 }, { "epoch": 2.81, "learning_rate": 1.314025194961008e-06, "loss": 0.0002, "step": 23471 }, { "epoch": 2.81, "learning_rate": 1.3139532093581285e-06, "loss": 0.0001, "step": 23472 }, { "epoch": 2.81, "learning_rate": 1.313881223755249e-06, "loss": 0.0002, "step": 23473 }, { "epoch": 2.81, "learning_rate": 1.3138092381523697e-06, "loss": 0.0002, "step": 23474 }, { "epoch": 2.81, "learning_rate": 1.31373725254949e-06, "loss": 0.0002, "step": 23475 }, { "epoch": 2.81, "learning_rate": 1.3136652669466106e-06, "loss": 0.0002, "step": 23476 }, { "epoch": 2.81, "learning_rate": 1.3135932813437311e-06, "loss": 0.0001, "step": 23477 }, { "epoch": 2.81, "learning_rate": 1.313521295740852e-06, "loss": 0.0002, "step": 23478 }, { "epoch": 2.81, "learning_rate": 1.3134493101379725e-06, "loss": 0.0002, "step": 23479 }, { "epoch": 2.81, "learning_rate": 1.313377324535093e-06, "loss": 0.0003, "step": 23480 }, { "epoch": 2.81, "learning_rate": 1.3133053389322136e-06, "loss": 0.0002, "step": 23481 }, { "epoch": 2.81, "learning_rate": 1.3132333533293342e-06, "loss": 0.0002, "step": 23482 }, { "epoch": 2.81, "learning_rate": 1.3131613677264548e-06, "loss": 0.0003, "step": 23483 }, { "epoch": 2.81, "learning_rate": 1.3130893821235753e-06, "loss": 0.0001, "step": 23484 }, { "epoch": 2.81, "learning_rate": 1.3130173965206959e-06, "loss": 0.0003, "step": 23485 }, { "epoch": 2.81, "learning_rate": 1.3129454109178165e-06, "loss": 0.0004, "step": 23486 }, { "epoch": 2.81, "learning_rate": 1.312873425314937e-06, "loss": 0.0001, "step": 23487 }, { "epoch": 2.81, "learning_rate": 1.3128014397120576e-06, "loss": 0.5624, "step": 23488 }, { "epoch": 2.81, "learning_rate": 1.3127294541091782e-06, "loss": 0.0006, "step": 23489 }, { "epoch": 2.81, "learning_rate": 1.3126574685062987e-06, "loss": 0.1809, "step": 23490 }, { "epoch": 2.81, "learning_rate": 1.3125854829034193e-06, "loss": 0.0001, "step": 23491 }, { "epoch": 2.82, "learning_rate": 1.3125134973005399e-06, "loss": 0.0001, "step": 23492 }, { "epoch": 2.82, "learning_rate": 1.3124415116976604e-06, "loss": 0.0001, "step": 23493 }, { "epoch": 2.82, "learning_rate": 1.3123695260947812e-06, "loss": 0.0001, "step": 23494 }, { "epoch": 2.82, "learning_rate": 1.3122975404919018e-06, "loss": 0.9111, "step": 23495 }, { "epoch": 2.82, "learning_rate": 1.3122255548890223e-06, "loss": 0.0002, "step": 23496 }, { "epoch": 2.82, "learning_rate": 1.312153569286143e-06, "loss": 0.0001, "step": 23497 }, { "epoch": 2.82, "learning_rate": 1.3120815836832633e-06, "loss": 0.0002, "step": 23498 }, { "epoch": 2.82, "learning_rate": 1.3120095980803838e-06, "loss": 0.0002, "step": 23499 }, { "epoch": 2.82, "learning_rate": 1.3119376124775044e-06, "loss": 0.0002, "step": 23500 }, { "epoch": 2.82, "learning_rate": 1.3118656268746252e-06, "loss": 0.0002, "step": 23501 }, { "epoch": 2.82, "learning_rate": 1.3117936412717457e-06, "loss": 0.0002, "step": 23502 }, { "epoch": 2.82, "learning_rate": 1.3117216556688663e-06, "loss": 0.0002, "step": 23503 }, { "epoch": 2.82, "learning_rate": 1.3116496700659869e-06, "loss": 0.0002, "step": 23504 }, { "epoch": 2.82, "learning_rate": 1.3115776844631074e-06, "loss": 0.0004, "step": 23505 }, { "epoch": 2.82, "learning_rate": 1.311505698860228e-06, "loss": 0.0001, "step": 23506 }, { "epoch": 2.82, "learning_rate": 1.3114337132573486e-06, "loss": 0.0002, "step": 23507 }, { "epoch": 2.82, "learning_rate": 1.3113617276544691e-06, "loss": 0.0035, "step": 23508 }, { "epoch": 2.82, "learning_rate": 1.3112897420515897e-06, "loss": 0.0002, "step": 23509 }, { "epoch": 2.82, "learning_rate": 1.3112177564487103e-06, "loss": 0.0002, "step": 23510 }, { "epoch": 2.82, "learning_rate": 1.3111457708458308e-06, "loss": 0.0001, "step": 23511 }, { "epoch": 2.82, "learning_rate": 1.3110737852429514e-06, "loss": 0.0002, "step": 23512 }, { "epoch": 2.82, "learning_rate": 1.311001799640072e-06, "loss": 0.0001, "step": 23513 }, { "epoch": 2.82, "learning_rate": 1.3109298140371925e-06, "loss": 0.0001, "step": 23514 }, { "epoch": 2.82, "learning_rate": 1.3108578284343131e-06, "loss": 0.0001, "step": 23515 }, { "epoch": 2.82, "learning_rate": 1.3107858428314337e-06, "loss": 0.0003, "step": 23516 }, { "epoch": 2.82, "learning_rate": 1.3107138572285545e-06, "loss": 0.0005, "step": 23517 }, { "epoch": 2.82, "learning_rate": 1.310641871625675e-06, "loss": 0.0001, "step": 23518 }, { "epoch": 2.82, "learning_rate": 1.3105698860227956e-06, "loss": 0.0002, "step": 23519 }, { "epoch": 2.82, "learning_rate": 1.3104979004199162e-06, "loss": 0.0002, "step": 23520 }, { "epoch": 2.82, "learning_rate": 1.3104259148170365e-06, "loss": 0.0002, "step": 23521 }, { "epoch": 2.82, "learning_rate": 1.310353929214157e-06, "loss": 0.0002, "step": 23522 }, { "epoch": 2.82, "learning_rate": 1.3102819436112776e-06, "loss": 0.0002, "step": 23523 }, { "epoch": 2.82, "learning_rate": 1.3102099580083984e-06, "loss": 0.0001, "step": 23524 }, { "epoch": 2.82, "learning_rate": 1.310137972405519e-06, "loss": 0.0002, "step": 23525 }, { "epoch": 2.82, "learning_rate": 1.3100659868026396e-06, "loss": 0.0002, "step": 23526 }, { "epoch": 2.82, "learning_rate": 1.3099940011997601e-06, "loss": 0.0001, "step": 23527 }, { "epoch": 2.82, "learning_rate": 1.3099220155968807e-06, "loss": 0.0002, "step": 23528 }, { "epoch": 2.82, "learning_rate": 1.3098500299940013e-06, "loss": 0.0002, "step": 23529 }, { "epoch": 2.82, "learning_rate": 1.3097780443911218e-06, "loss": 0.001, "step": 23530 }, { "epoch": 2.82, "learning_rate": 1.3097060587882424e-06, "loss": 0.0002, "step": 23531 }, { "epoch": 2.82, "learning_rate": 1.309634073185363e-06, "loss": 0.0001, "step": 23532 }, { "epoch": 2.82, "learning_rate": 1.3095620875824835e-06, "loss": 0.0003, "step": 23533 }, { "epoch": 2.82, "learning_rate": 1.309490101979604e-06, "loss": 0.0011, "step": 23534 }, { "epoch": 2.82, "learning_rate": 1.3094181163767247e-06, "loss": 0.0002, "step": 23535 }, { "epoch": 2.82, "learning_rate": 1.3093461307738452e-06, "loss": 0.0001, "step": 23536 }, { "epoch": 2.82, "learning_rate": 1.3092741451709658e-06, "loss": 0.0002, "step": 23537 }, { "epoch": 2.82, "learning_rate": 1.3092021595680864e-06, "loss": 0.0002, "step": 23538 }, { "epoch": 2.82, "learning_rate": 1.309130173965207e-06, "loss": 0.0001, "step": 23539 }, { "epoch": 2.82, "learning_rate": 1.3090581883623277e-06, "loss": 0.0002, "step": 23540 }, { "epoch": 2.82, "learning_rate": 1.3089862027594483e-06, "loss": 0.0549, "step": 23541 }, { "epoch": 2.82, "learning_rate": 1.3089142171565688e-06, "loss": 0.0001, "step": 23542 }, { "epoch": 2.82, "learning_rate": 1.3088422315536894e-06, "loss": 0.1907, "step": 23543 }, { "epoch": 2.82, "learning_rate": 1.3087702459508098e-06, "loss": 0.0001, "step": 23544 }, { "epoch": 2.82, "learning_rate": 1.3086982603479303e-06, "loss": 0.0001, "step": 23545 }, { "epoch": 2.82, "learning_rate": 1.3086262747450509e-06, "loss": 0.0002, "step": 23546 }, { "epoch": 2.82, "learning_rate": 1.3085542891421717e-06, "loss": 0.0003, "step": 23547 }, { "epoch": 2.82, "learning_rate": 1.3084823035392922e-06, "loss": 0.0003, "step": 23548 }, { "epoch": 2.82, "learning_rate": 1.3084103179364128e-06, "loss": 0.0002, "step": 23549 }, { "epoch": 2.82, "learning_rate": 1.3083383323335334e-06, "loss": 0.0001, "step": 23550 }, { "epoch": 2.82, "learning_rate": 1.308266346730654e-06, "loss": 0.0002, "step": 23551 }, { "epoch": 2.82, "learning_rate": 1.3081943611277745e-06, "loss": 0.0002, "step": 23552 }, { "epoch": 2.82, "learning_rate": 1.308122375524895e-06, "loss": 0.0001, "step": 23553 }, { "epoch": 2.82, "learning_rate": 1.3080503899220156e-06, "loss": 0.0003, "step": 23554 }, { "epoch": 2.82, "learning_rate": 1.3079784043191362e-06, "loss": 0.0004, "step": 23555 }, { "epoch": 2.82, "learning_rate": 1.3079064187162568e-06, "loss": 0.0002, "step": 23556 }, { "epoch": 2.82, "learning_rate": 1.3078344331133773e-06, "loss": 0.0002, "step": 23557 }, { "epoch": 2.82, "learning_rate": 1.307762447510498e-06, "loss": 0.745, "step": 23558 }, { "epoch": 2.82, "learning_rate": 1.3076904619076185e-06, "loss": 0.0001, "step": 23559 }, { "epoch": 2.82, "learning_rate": 1.307618476304739e-06, "loss": 0.0002, "step": 23560 }, { "epoch": 2.82, "learning_rate": 1.3075464907018596e-06, "loss": 0.0001, "step": 23561 }, { "epoch": 2.82, "learning_rate": 1.3074745050989802e-06, "loss": 0.0004, "step": 23562 }, { "epoch": 2.82, "learning_rate": 1.307402519496101e-06, "loss": 0.0001, "step": 23563 }, { "epoch": 2.82, "learning_rate": 1.3073305338932215e-06, "loss": 0.0002, "step": 23564 }, { "epoch": 2.82, "learning_rate": 1.307258548290342e-06, "loss": 0.0001, "step": 23565 }, { "epoch": 2.82, "learning_rate": 1.3071865626874627e-06, "loss": 0.0002, "step": 23566 }, { "epoch": 2.82, "learning_rate": 1.307114577084583e-06, "loss": 0.0001, "step": 23567 }, { "epoch": 2.82, "learning_rate": 1.3070425914817036e-06, "loss": 0.0008, "step": 23568 }, { "epoch": 2.82, "learning_rate": 1.3069706058788241e-06, "loss": 0.0003, "step": 23569 }, { "epoch": 2.82, "learning_rate": 1.3068986202759447e-06, "loss": 0.0001, "step": 23570 }, { "epoch": 2.82, "learning_rate": 1.3068266346730655e-06, "loss": 0.0003, "step": 23571 }, { "epoch": 2.82, "learning_rate": 1.306754649070186e-06, "loss": 0.0004, "step": 23572 }, { "epoch": 2.82, "learning_rate": 1.3066826634673066e-06, "loss": 0.0001, "step": 23573 }, { "epoch": 2.82, "learning_rate": 1.3066106778644272e-06, "loss": 0.0002, "step": 23574 }, { "epoch": 2.83, "learning_rate": 1.3065386922615478e-06, "loss": 0.0002, "step": 23575 }, { "epoch": 2.83, "learning_rate": 1.3064667066586683e-06, "loss": 0.7034, "step": 23576 }, { "epoch": 2.83, "learning_rate": 1.3063947210557889e-06, "loss": 0.0002, "step": 23577 }, { "epoch": 2.83, "learning_rate": 1.3063227354529095e-06, "loss": 0.0004, "step": 23578 }, { "epoch": 2.83, "learning_rate": 1.30625074985003e-06, "loss": 0.0002, "step": 23579 }, { "epoch": 2.83, "learning_rate": 1.3061787642471506e-06, "loss": 0.0002, "step": 23580 }, { "epoch": 2.83, "learning_rate": 1.3061067786442712e-06, "loss": 0.0001, "step": 23581 }, { "epoch": 2.83, "learning_rate": 1.3060347930413917e-06, "loss": 0.9673, "step": 23582 }, { "epoch": 2.83, "learning_rate": 1.3059628074385123e-06, "loss": 1.0598, "step": 23583 }, { "epoch": 2.83, "learning_rate": 1.3058908218356329e-06, "loss": 0.0001, "step": 23584 }, { "epoch": 2.83, "learning_rate": 1.3058188362327534e-06, "loss": 0.0007, "step": 23585 }, { "epoch": 2.83, "learning_rate": 1.305746850629874e-06, "loss": 0.0002, "step": 23586 }, { "epoch": 2.83, "learning_rate": 1.3056748650269948e-06, "loss": 0.0002, "step": 23587 }, { "epoch": 2.83, "learning_rate": 1.3056028794241153e-06, "loss": 0.0, "step": 23588 }, { "epoch": 2.83, "learning_rate": 1.305530893821236e-06, "loss": 0.0003, "step": 23589 }, { "epoch": 2.83, "learning_rate": 1.3054589082183563e-06, "loss": 0.0001, "step": 23590 }, { "epoch": 2.83, "learning_rate": 1.3053869226154768e-06, "loss": 0.0004, "step": 23591 }, { "epoch": 2.83, "learning_rate": 1.3053149370125974e-06, "loss": 0.0003, "step": 23592 }, { "epoch": 2.83, "learning_rate": 1.305242951409718e-06, "loss": 0.0001, "step": 23593 }, { "epoch": 2.83, "learning_rate": 1.3051709658068387e-06, "loss": 0.0003, "step": 23594 }, { "epoch": 2.83, "learning_rate": 1.3050989802039593e-06, "loss": 0.0002, "step": 23595 }, { "epoch": 2.83, "learning_rate": 1.3050269946010799e-06, "loss": 0.0004, "step": 23596 }, { "epoch": 2.83, "learning_rate": 1.3049550089982004e-06, "loss": 0.0001, "step": 23597 }, { "epoch": 2.83, "learning_rate": 1.304883023395321e-06, "loss": 0.0001, "step": 23598 }, { "epoch": 2.83, "learning_rate": 1.3048110377924416e-06, "loss": 0.0002, "step": 23599 }, { "epoch": 2.83, "learning_rate": 1.3047390521895621e-06, "loss": 0.0003, "step": 23600 }, { "epoch": 2.83, "learning_rate": 1.3046670665866827e-06, "loss": 0.0, "step": 23601 }, { "epoch": 2.83, "learning_rate": 1.3045950809838033e-06, "loss": 0.0002, "step": 23602 }, { "epoch": 2.83, "learning_rate": 1.3045230953809238e-06, "loss": 0.0002, "step": 23603 }, { "epoch": 2.83, "learning_rate": 1.3044511097780444e-06, "loss": 0.0001, "step": 23604 }, { "epoch": 2.83, "learning_rate": 1.304379124175165e-06, "loss": 0.0002, "step": 23605 }, { "epoch": 2.83, "learning_rate": 1.3043071385722855e-06, "loss": 0.0003, "step": 23606 }, { "epoch": 2.83, "learning_rate": 1.304235152969406e-06, "loss": 0.0004, "step": 23607 }, { "epoch": 2.83, "learning_rate": 1.3041631673665267e-06, "loss": 0.0003, "step": 23608 }, { "epoch": 2.83, "learning_rate": 1.3040911817636472e-06, "loss": 0.0016, "step": 23609 }, { "epoch": 2.83, "learning_rate": 1.304019196160768e-06, "loss": 0.0004, "step": 23610 }, { "epoch": 2.83, "learning_rate": 1.3039472105578886e-06, "loss": 0.1215, "step": 23611 }, { "epoch": 2.83, "learning_rate": 1.3038752249550091e-06, "loss": 0.5898, "step": 23612 }, { "epoch": 2.83, "learning_rate": 1.3038032393521295e-06, "loss": 0.0003, "step": 23613 }, { "epoch": 2.83, "learning_rate": 1.30373125374925e-06, "loss": 0.0002, "step": 23614 }, { "epoch": 2.83, "learning_rate": 1.3036592681463706e-06, "loss": 0.0002, "step": 23615 }, { "epoch": 2.83, "learning_rate": 1.3035872825434912e-06, "loss": 0.0001, "step": 23616 }, { "epoch": 2.83, "learning_rate": 1.303515296940612e-06, "loss": 0.0002, "step": 23617 }, { "epoch": 2.83, "learning_rate": 1.3034433113377325e-06, "loss": 0.0001, "step": 23618 }, { "epoch": 2.83, "learning_rate": 1.3033713257348531e-06, "loss": 0.0015, "step": 23619 }, { "epoch": 2.83, "learning_rate": 1.3032993401319737e-06, "loss": 0.0005, "step": 23620 }, { "epoch": 2.83, "learning_rate": 1.3032273545290942e-06, "loss": 0.0002, "step": 23621 }, { "epoch": 2.83, "learning_rate": 1.3031553689262148e-06, "loss": 0.0008, "step": 23622 }, { "epoch": 2.83, "learning_rate": 1.3030833833233354e-06, "loss": 0.0001, "step": 23623 }, { "epoch": 2.83, "learning_rate": 1.303011397720456e-06, "loss": 0.0003, "step": 23624 }, { "epoch": 2.83, "learning_rate": 1.3029394121175765e-06, "loss": 0.0003, "step": 23625 }, { "epoch": 2.83, "learning_rate": 1.302867426514697e-06, "loss": 0.0711, "step": 23626 }, { "epoch": 2.83, "learning_rate": 1.3027954409118176e-06, "loss": 0.0001, "step": 23627 }, { "epoch": 2.83, "learning_rate": 1.3027234553089382e-06, "loss": 0.0001, "step": 23628 }, { "epoch": 2.83, "learning_rate": 1.3026514697060588e-06, "loss": 0.0008, "step": 23629 }, { "epoch": 2.83, "learning_rate": 1.3025794841031793e-06, "loss": 0.0289, "step": 23630 }, { "epoch": 2.83, "learning_rate": 1.3025074985003e-06, "loss": 0.0001, "step": 23631 }, { "epoch": 2.83, "learning_rate": 1.3024355128974205e-06, "loss": 0.0002, "step": 23632 }, { "epoch": 2.83, "learning_rate": 1.3023635272945413e-06, "loss": 0.0004, "step": 23633 }, { "epoch": 2.83, "learning_rate": 1.3022915416916618e-06, "loss": 0.0002, "step": 23634 }, { "epoch": 2.83, "learning_rate": 1.3022195560887824e-06, "loss": 0.0002, "step": 23635 }, { "epoch": 2.83, "learning_rate": 1.3021475704859027e-06, "loss": 0.0002, "step": 23636 }, { "epoch": 2.83, "learning_rate": 1.3020755848830233e-06, "loss": 0.0002, "step": 23637 }, { "epoch": 2.83, "learning_rate": 1.3020035992801439e-06, "loss": 0.0003, "step": 23638 }, { "epoch": 2.83, "learning_rate": 1.3019316136772644e-06, "loss": 0.0003, "step": 23639 }, { "epoch": 2.83, "learning_rate": 1.3018596280743852e-06, "loss": 0.0003, "step": 23640 }, { "epoch": 2.83, "learning_rate": 1.3017876424715058e-06, "loss": 0.0002, "step": 23641 }, { "epoch": 2.83, "learning_rate": 1.3017156568686264e-06, "loss": 0.0006, "step": 23642 }, { "epoch": 2.83, "learning_rate": 1.301643671265747e-06, "loss": 0.5775, "step": 23643 }, { "epoch": 2.83, "learning_rate": 1.3015716856628675e-06, "loss": 0.0003, "step": 23644 }, { "epoch": 2.83, "learning_rate": 1.301499700059988e-06, "loss": 0.0004, "step": 23645 }, { "epoch": 2.83, "learning_rate": 1.3014277144571086e-06, "loss": 0.0001, "step": 23646 }, { "epoch": 2.83, "learning_rate": 1.3013557288542292e-06, "loss": 0.0001, "step": 23647 }, { "epoch": 2.83, "learning_rate": 1.3012837432513498e-06, "loss": 0.0014, "step": 23648 }, { "epoch": 2.83, "learning_rate": 1.3012117576484703e-06, "loss": 0.0002, "step": 23649 }, { "epoch": 2.83, "learning_rate": 1.301139772045591e-06, "loss": 0.0004, "step": 23650 }, { "epoch": 2.83, "learning_rate": 1.3010677864427115e-06, "loss": 0.0004, "step": 23651 }, { "epoch": 2.83, "learning_rate": 1.300995800839832e-06, "loss": 0.0019, "step": 23652 }, { "epoch": 2.83, "learning_rate": 1.3009238152369526e-06, "loss": 0.2228, "step": 23653 }, { "epoch": 2.83, "learning_rate": 1.3008518296340732e-06, "loss": 0.0003, "step": 23654 }, { "epoch": 2.83, "learning_rate": 1.3007798440311937e-06, "loss": 0.0002, "step": 23655 }, { "epoch": 2.83, "learning_rate": 1.3007078584283145e-06, "loss": 0.0001, "step": 23656 }, { "epoch": 2.83, "learning_rate": 1.300635872825435e-06, "loss": 1.1547, "step": 23657 }, { "epoch": 2.83, "learning_rate": 1.3005638872225556e-06, "loss": 0.0001, "step": 23658 }, { "epoch": 2.84, "learning_rate": 1.300491901619676e-06, "loss": 0.0004, "step": 23659 }, { "epoch": 2.84, "learning_rate": 1.3004199160167966e-06, "loss": 0.0001, "step": 23660 }, { "epoch": 2.84, "learning_rate": 1.3003479304139171e-06, "loss": 0.0004, "step": 23661 }, { "epoch": 2.84, "learning_rate": 1.3002759448110377e-06, "loss": 0.0005, "step": 23662 }, { "epoch": 2.84, "learning_rate": 1.3002039592081585e-06, "loss": 0.0005, "step": 23663 }, { "epoch": 2.84, "learning_rate": 1.300131973605279e-06, "loss": 0.0002, "step": 23664 }, { "epoch": 2.84, "learning_rate": 1.3000599880023996e-06, "loss": 0.0004, "step": 23665 }, { "epoch": 2.84, "learning_rate": 1.2999880023995202e-06, "loss": 0.0004, "step": 23666 }, { "epoch": 2.84, "learning_rate": 1.2999160167966407e-06, "loss": 0.0002, "step": 23667 }, { "epoch": 2.84, "learning_rate": 1.2998440311937613e-06, "loss": 0.0005, "step": 23668 }, { "epoch": 2.84, "learning_rate": 1.2997720455908819e-06, "loss": 0.0001, "step": 23669 }, { "epoch": 2.84, "learning_rate": 1.2997000599880024e-06, "loss": 0.0005, "step": 23670 }, { "epoch": 2.84, "learning_rate": 1.299628074385123e-06, "loss": 0.0002, "step": 23671 }, { "epoch": 2.84, "learning_rate": 1.2995560887822436e-06, "loss": 0.0002, "step": 23672 }, { "epoch": 2.84, "learning_rate": 1.2994841031793641e-06, "loss": 0.0007, "step": 23673 }, { "epoch": 2.84, "learning_rate": 1.2994121175764847e-06, "loss": 0.0001, "step": 23674 }, { "epoch": 2.84, "learning_rate": 1.2993401319736053e-06, "loss": 0.0001, "step": 23675 }, { "epoch": 2.84, "learning_rate": 1.2992681463707258e-06, "loss": 0.0002, "step": 23676 }, { "epoch": 2.84, "learning_rate": 1.2991961607678464e-06, "loss": 0.4027, "step": 23677 }, { "epoch": 2.84, "learning_rate": 1.299124175164967e-06, "loss": 0.0001, "step": 23678 }, { "epoch": 2.84, "learning_rate": 1.2990521895620878e-06, "loss": 0.0002, "step": 23679 }, { "epoch": 2.84, "learning_rate": 1.2989802039592083e-06, "loss": 0.5444, "step": 23680 }, { "epoch": 2.84, "learning_rate": 1.2989082183563289e-06, "loss": 0.0006, "step": 23681 }, { "epoch": 2.84, "learning_rate": 1.2988362327534492e-06, "loss": 0.0002, "step": 23682 }, { "epoch": 2.84, "learning_rate": 1.2987642471505698e-06, "loss": 0.0001, "step": 23683 }, { "epoch": 2.84, "learning_rate": 1.2986922615476904e-06, "loss": 0.0004, "step": 23684 }, { "epoch": 2.84, "learning_rate": 1.298620275944811e-06, "loss": 0.0002, "step": 23685 }, { "epoch": 2.84, "learning_rate": 1.2985482903419317e-06, "loss": 0.0001, "step": 23686 }, { "epoch": 2.84, "learning_rate": 1.2984763047390523e-06, "loss": 0.0004, "step": 23687 }, { "epoch": 2.84, "learning_rate": 1.2984043191361729e-06, "loss": 0.0001, "step": 23688 }, { "epoch": 2.84, "learning_rate": 1.2983323335332934e-06, "loss": 0.0003, "step": 23689 }, { "epoch": 2.84, "learning_rate": 1.298260347930414e-06, "loss": 0.0001, "step": 23690 }, { "epoch": 2.84, "learning_rate": 1.2981883623275346e-06, "loss": 0.0001, "step": 23691 }, { "epoch": 2.84, "learning_rate": 1.2981163767246551e-06, "loss": 0.0002, "step": 23692 }, { "epoch": 2.84, "learning_rate": 1.2980443911217757e-06, "loss": 0.0002, "step": 23693 }, { "epoch": 2.84, "learning_rate": 1.2979724055188963e-06, "loss": 0.001, "step": 23694 }, { "epoch": 2.84, "learning_rate": 1.2979004199160168e-06, "loss": 0.0001, "step": 23695 }, { "epoch": 2.84, "learning_rate": 1.2978284343131374e-06, "loss": 0.0002, "step": 23696 }, { "epoch": 2.84, "learning_rate": 1.297756448710258e-06, "loss": 0.0003, "step": 23697 }, { "epoch": 2.84, "learning_rate": 1.2976844631073785e-06, "loss": 0.0006, "step": 23698 }, { "epoch": 2.84, "learning_rate": 1.297612477504499e-06, "loss": 0.0005, "step": 23699 }, { "epoch": 2.84, "learning_rate": 1.2975404919016197e-06, "loss": 0.0003, "step": 23700 }, { "epoch": 2.84, "learning_rate": 1.2974685062987402e-06, "loss": 0.0002, "step": 23701 }, { "epoch": 2.84, "learning_rate": 1.297396520695861e-06, "loss": 0.0001, "step": 23702 }, { "epoch": 2.84, "learning_rate": 1.2973245350929816e-06, "loss": 0.0002, "step": 23703 }, { "epoch": 2.84, "learning_rate": 1.2972525494901021e-06, "loss": 0.0002, "step": 23704 }, { "epoch": 2.84, "learning_rate": 1.2971805638872227e-06, "loss": 0.0001, "step": 23705 }, { "epoch": 2.84, "learning_rate": 1.297108578284343e-06, "loss": 0.0004, "step": 23706 }, { "epoch": 2.84, "learning_rate": 1.2970365926814636e-06, "loss": 0.0001, "step": 23707 }, { "epoch": 2.84, "learning_rate": 1.2969646070785842e-06, "loss": 0.0003, "step": 23708 }, { "epoch": 2.84, "learning_rate": 1.2968926214757048e-06, "loss": 0.0001, "step": 23709 }, { "epoch": 2.84, "learning_rate": 1.2968206358728255e-06, "loss": 0.0003, "step": 23710 }, { "epoch": 2.84, "learning_rate": 1.2967486502699461e-06, "loss": 0.0001, "step": 23711 }, { "epoch": 2.84, "learning_rate": 1.2966766646670667e-06, "loss": 0.0002, "step": 23712 }, { "epoch": 2.84, "learning_rate": 1.2966046790641872e-06, "loss": 0.0002, "step": 23713 }, { "epoch": 2.84, "learning_rate": 1.2965326934613078e-06, "loss": 0.0002, "step": 23714 }, { "epoch": 2.84, "learning_rate": 1.2964607078584284e-06, "loss": 0.0003, "step": 23715 }, { "epoch": 2.84, "learning_rate": 1.296388722255549e-06, "loss": 0.0003, "step": 23716 }, { "epoch": 2.84, "learning_rate": 1.2963167366526695e-06, "loss": 0.0001, "step": 23717 }, { "epoch": 2.84, "learning_rate": 1.29624475104979e-06, "loss": 0.0206, "step": 23718 }, { "epoch": 2.84, "learning_rate": 1.2961727654469106e-06, "loss": 0.0001, "step": 23719 }, { "epoch": 2.84, "learning_rate": 1.2961007798440312e-06, "loss": 0.0001, "step": 23720 }, { "epoch": 2.84, "learning_rate": 1.2960287942411518e-06, "loss": 0.0024, "step": 23721 }, { "epoch": 2.84, "learning_rate": 1.2959568086382723e-06, "loss": 0.0001, "step": 23722 }, { "epoch": 2.84, "learning_rate": 1.295884823035393e-06, "loss": 0.0002, "step": 23723 }, { "epoch": 2.84, "learning_rate": 1.2958128374325135e-06, "loss": 0.0002, "step": 23724 }, { "epoch": 2.84, "learning_rate": 1.295740851829634e-06, "loss": 0.0003, "step": 23725 }, { "epoch": 2.84, "learning_rate": 1.2956688662267548e-06, "loss": 0.0002, "step": 23726 }, { "epoch": 2.84, "learning_rate": 1.2955968806238754e-06, "loss": 0.0001, "step": 23727 }, { "epoch": 2.84, "learning_rate": 1.295524895020996e-06, "loss": 0.0002, "step": 23728 }, { "epoch": 2.84, "learning_rate": 1.2954529094181163e-06, "loss": 0.0002, "step": 23729 }, { "epoch": 2.84, "learning_rate": 1.2953809238152369e-06, "loss": 0.0003, "step": 23730 }, { "epoch": 2.84, "learning_rate": 1.2953089382123574e-06, "loss": 0.0001, "step": 23731 }, { "epoch": 2.84, "learning_rate": 1.295236952609478e-06, "loss": 0.0004, "step": 23732 }, { "epoch": 2.84, "learning_rate": 1.2951649670065988e-06, "loss": 0.0001, "step": 23733 }, { "epoch": 2.84, "learning_rate": 1.2950929814037194e-06, "loss": 0.0003, "step": 23734 }, { "epoch": 2.84, "learning_rate": 1.29502099580084e-06, "loss": 0.0001, "step": 23735 }, { "epoch": 2.84, "learning_rate": 1.2949490101979605e-06, "loss": 0.0002, "step": 23736 }, { "epoch": 2.84, "learning_rate": 1.294877024595081e-06, "loss": 0.0001, "step": 23737 }, { "epoch": 2.84, "learning_rate": 1.2948050389922016e-06, "loss": 0.0001, "step": 23738 }, { "epoch": 2.84, "learning_rate": 1.2947330533893222e-06, "loss": 0.0005, "step": 23739 }, { "epoch": 2.84, "learning_rate": 1.2946610677864428e-06, "loss": 0.0001, "step": 23740 }, { "epoch": 2.84, "learning_rate": 1.2945890821835633e-06, "loss": 0.0001, "step": 23741 }, { "epoch": 2.85, "learning_rate": 1.2945170965806839e-06, "loss": 0.0001, "step": 23742 }, { "epoch": 2.85, "learning_rate": 1.2944451109778045e-06, "loss": 0.0003, "step": 23743 }, { "epoch": 2.85, "learning_rate": 1.294373125374925e-06, "loss": 0.0003, "step": 23744 }, { "epoch": 2.85, "learning_rate": 1.2943011397720456e-06, "loss": 0.0001, "step": 23745 }, { "epoch": 2.85, "learning_rate": 1.2942291541691662e-06, "loss": 0.0001, "step": 23746 }, { "epoch": 2.85, "learning_rate": 1.2941571685662867e-06, "loss": 0.0003, "step": 23747 }, { "epoch": 2.85, "learning_rate": 1.2940851829634073e-06, "loss": 0.0002, "step": 23748 }, { "epoch": 2.85, "learning_rate": 1.294013197360528e-06, "loss": 0.0001, "step": 23749 }, { "epoch": 2.85, "learning_rate": 1.2939412117576486e-06, "loss": 0.4772, "step": 23750 }, { "epoch": 2.85, "learning_rate": 1.2938692261547692e-06, "loss": 0.0001, "step": 23751 }, { "epoch": 2.85, "learning_rate": 1.2937972405518896e-06, "loss": 0.0001, "step": 23752 }, { "epoch": 2.85, "learning_rate": 1.2937252549490101e-06, "loss": 0.0003, "step": 23753 }, { "epoch": 2.85, "learning_rate": 1.2936532693461307e-06, "loss": 1.0378, "step": 23754 }, { "epoch": 2.85, "learning_rate": 1.2935812837432513e-06, "loss": 0.0001, "step": 23755 }, { "epoch": 2.85, "learning_rate": 1.293509298140372e-06, "loss": 0.0001, "step": 23756 }, { "epoch": 2.85, "learning_rate": 1.2934373125374926e-06, "loss": 0.0223, "step": 23757 }, { "epoch": 2.85, "learning_rate": 1.2933653269346132e-06, "loss": 0.0001, "step": 23758 }, { "epoch": 2.85, "learning_rate": 1.2932933413317337e-06, "loss": 0.0002, "step": 23759 }, { "epoch": 2.85, "learning_rate": 1.2932213557288543e-06, "loss": 0.0001, "step": 23760 }, { "epoch": 2.85, "learning_rate": 1.2931493701259749e-06, "loss": 0.0001, "step": 23761 }, { "epoch": 2.85, "learning_rate": 1.2930773845230954e-06, "loss": 0.0004, "step": 23762 }, { "epoch": 2.85, "learning_rate": 1.293005398920216e-06, "loss": 0.0001, "step": 23763 }, { "epoch": 2.85, "learning_rate": 1.2929334133173366e-06, "loss": 0.0001, "step": 23764 }, { "epoch": 2.85, "learning_rate": 1.2928614277144571e-06, "loss": 0.0001, "step": 23765 }, { "epoch": 2.85, "learning_rate": 1.2927894421115777e-06, "loss": 0.0004, "step": 23766 }, { "epoch": 2.85, "learning_rate": 1.2927174565086983e-06, "loss": 0.0001, "step": 23767 }, { "epoch": 2.85, "learning_rate": 1.2926454709058188e-06, "loss": 0.0001, "step": 23768 }, { "epoch": 2.85, "learning_rate": 1.2925734853029394e-06, "loss": 0.0002, "step": 23769 }, { "epoch": 2.85, "learning_rate": 1.29250149970006e-06, "loss": 0.0551, "step": 23770 }, { "epoch": 2.85, "learning_rate": 1.2924295140971805e-06, "loss": 0.3229, "step": 23771 }, { "epoch": 2.85, "learning_rate": 1.2923575284943013e-06, "loss": 0.0001, "step": 23772 }, { "epoch": 2.85, "learning_rate": 1.2922855428914219e-06, "loss": 0.0001, "step": 23773 }, { "epoch": 2.85, "learning_rate": 1.2922135572885424e-06, "loss": 0.0002, "step": 23774 }, { "epoch": 2.85, "learning_rate": 1.2921415716856628e-06, "loss": 0.0002, "step": 23775 }, { "epoch": 2.85, "learning_rate": 1.2920695860827834e-06, "loss": 0.0002, "step": 23776 }, { "epoch": 2.85, "learning_rate": 1.291997600479904e-06, "loss": 0.0229, "step": 23777 }, { "epoch": 2.85, "learning_rate": 1.2919256148770245e-06, "loss": 0.0003, "step": 23778 }, { "epoch": 2.85, "learning_rate": 1.2918536292741453e-06, "loss": 0.0001, "step": 23779 }, { "epoch": 2.85, "learning_rate": 1.2917816436712658e-06, "loss": 0.0002, "step": 23780 }, { "epoch": 2.85, "learning_rate": 1.2917096580683864e-06, "loss": 0.0001, "step": 23781 }, { "epoch": 2.85, "learning_rate": 1.291637672465507e-06, "loss": 0.0225, "step": 23782 }, { "epoch": 2.85, "learning_rate": 1.2915656868626276e-06, "loss": 0.0002, "step": 23783 }, { "epoch": 2.85, "learning_rate": 1.2914937012597481e-06, "loss": 0.0001, "step": 23784 }, { "epoch": 2.85, "learning_rate": 1.2914217156568687e-06, "loss": 0.0003, "step": 23785 }, { "epoch": 2.85, "learning_rate": 1.2913497300539893e-06, "loss": 0.0002, "step": 23786 }, { "epoch": 2.85, "learning_rate": 1.2912777444511098e-06, "loss": 0.0002, "step": 23787 }, { "epoch": 2.85, "learning_rate": 1.2912057588482304e-06, "loss": 0.0001, "step": 23788 }, { "epoch": 2.85, "learning_rate": 1.291133773245351e-06, "loss": 0.0002, "step": 23789 }, { "epoch": 2.85, "learning_rate": 1.2910617876424715e-06, "loss": 0.0004, "step": 23790 }, { "epoch": 2.85, "learning_rate": 1.290989802039592e-06, "loss": 0.0001, "step": 23791 }, { "epoch": 2.85, "learning_rate": 1.2909178164367127e-06, "loss": 0.0001, "step": 23792 }, { "epoch": 2.85, "learning_rate": 1.2908458308338332e-06, "loss": 0.0001, "step": 23793 }, { "epoch": 2.85, "learning_rate": 1.2907738452309538e-06, "loss": 0.0003, "step": 23794 }, { "epoch": 2.85, "learning_rate": 1.2907018596280746e-06, "loss": 0.0006, "step": 23795 }, { "epoch": 2.85, "learning_rate": 1.2906298740251951e-06, "loss": 0.0004, "step": 23796 }, { "epoch": 2.85, "learning_rate": 1.2905578884223157e-06, "loss": 0.0001, "step": 23797 }, { "epoch": 2.85, "learning_rate": 1.290485902819436e-06, "loss": 0.0003, "step": 23798 }, { "epoch": 2.85, "learning_rate": 1.2904139172165566e-06, "loss": 0.0003, "step": 23799 }, { "epoch": 2.85, "learning_rate": 1.2903419316136772e-06, "loss": 0.0001, "step": 23800 }, { "epoch": 2.85, "learning_rate": 1.2902699460107978e-06, "loss": 0.0004, "step": 23801 }, { "epoch": 2.85, "learning_rate": 1.2901979604079185e-06, "loss": 0.0003, "step": 23802 }, { "epoch": 2.85, "learning_rate": 1.290125974805039e-06, "loss": 0.0001, "step": 23803 }, { "epoch": 2.85, "learning_rate": 1.2900539892021597e-06, "loss": 0.0002, "step": 23804 }, { "epoch": 2.85, "learning_rate": 1.2899820035992802e-06, "loss": 0.6979, "step": 23805 }, { "epoch": 2.85, "learning_rate": 1.2899100179964008e-06, "loss": 0.0001, "step": 23806 }, { "epoch": 2.85, "learning_rate": 1.2898380323935214e-06, "loss": 0.0002, "step": 23807 }, { "epoch": 2.85, "learning_rate": 1.289766046790642e-06, "loss": 0.0051, "step": 23808 }, { "epoch": 2.85, "learning_rate": 1.2896940611877625e-06, "loss": 0.0004, "step": 23809 }, { "epoch": 2.85, "learning_rate": 1.289622075584883e-06, "loss": 0.0004, "step": 23810 }, { "epoch": 2.85, "learning_rate": 1.2895500899820036e-06, "loss": 0.0002, "step": 23811 }, { "epoch": 2.85, "learning_rate": 1.2894781043791242e-06, "loss": 0.0001, "step": 23812 }, { "epoch": 2.85, "learning_rate": 1.2894061187762448e-06, "loss": 0.0002, "step": 23813 }, { "epoch": 2.85, "learning_rate": 1.2893341331733653e-06, "loss": 0.0009, "step": 23814 }, { "epoch": 2.85, "learning_rate": 1.289262147570486e-06, "loss": 0.0002, "step": 23815 }, { "epoch": 2.85, "learning_rate": 1.2891901619676065e-06, "loss": 0.0002, "step": 23816 }, { "epoch": 2.85, "learning_rate": 1.289118176364727e-06, "loss": 0.0002, "step": 23817 }, { "epoch": 2.85, "learning_rate": 1.2890461907618478e-06, "loss": 0.18, "step": 23818 }, { "epoch": 2.85, "learning_rate": 1.2889742051589684e-06, "loss": 0.0003, "step": 23819 }, { "epoch": 2.85, "learning_rate": 1.288902219556089e-06, "loss": 0.0001, "step": 23820 }, { "epoch": 2.85, "learning_rate": 1.2888302339532093e-06, "loss": 0.0005, "step": 23821 }, { "epoch": 2.85, "learning_rate": 1.2887582483503299e-06, "loss": 0.0002, "step": 23822 }, { "epoch": 2.85, "learning_rate": 1.2886862627474504e-06, "loss": 0.0046, "step": 23823 }, { "epoch": 2.85, "learning_rate": 1.288614277144571e-06, "loss": 0.0002, "step": 23824 }, { "epoch": 2.86, "learning_rate": 1.2885422915416918e-06, "loss": 0.0002, "step": 23825 }, { "epoch": 2.86, "learning_rate": 1.2884703059388123e-06, "loss": 0.0001, "step": 23826 }, { "epoch": 2.86, "learning_rate": 1.288398320335933e-06, "loss": 0.0002, "step": 23827 }, { "epoch": 2.86, "learning_rate": 1.2883263347330535e-06, "loss": 0.0003, "step": 23828 }, { "epoch": 2.86, "learning_rate": 1.288254349130174e-06, "loss": 0.0124, "step": 23829 }, { "epoch": 2.86, "learning_rate": 1.2881823635272946e-06, "loss": 0.0004, "step": 23830 }, { "epoch": 2.86, "learning_rate": 1.2881103779244152e-06, "loss": 0.0001, "step": 23831 }, { "epoch": 2.86, "learning_rate": 1.2880383923215357e-06, "loss": 0.0002, "step": 23832 }, { "epoch": 2.86, "learning_rate": 1.2879664067186563e-06, "loss": 0.0001, "step": 23833 }, { "epoch": 2.86, "learning_rate": 1.2878944211157769e-06, "loss": 0.0002, "step": 23834 }, { "epoch": 2.86, "learning_rate": 1.2878224355128974e-06, "loss": 0.0001, "step": 23835 }, { "epoch": 2.86, "learning_rate": 1.287750449910018e-06, "loss": 0.0001, "step": 23836 }, { "epoch": 2.86, "learning_rate": 1.2876784643071386e-06, "loss": 0.0001, "step": 23837 }, { "epoch": 2.86, "learning_rate": 1.2876064787042591e-06, "loss": 0.0006, "step": 23838 }, { "epoch": 2.86, "learning_rate": 1.2875344931013797e-06, "loss": 0.0002, "step": 23839 }, { "epoch": 2.86, "learning_rate": 1.2874625074985003e-06, "loss": 0.0001, "step": 23840 }, { "epoch": 2.86, "learning_rate": 1.2873905218956208e-06, "loss": 0.0004, "step": 23841 }, { "epoch": 2.86, "learning_rate": 1.2873185362927416e-06, "loss": 0.0001, "step": 23842 }, { "epoch": 2.86, "learning_rate": 1.2872465506898622e-06, "loss": 0.0005, "step": 23843 }, { "epoch": 2.86, "learning_rate": 1.2871745650869825e-06, "loss": 0.0002, "step": 23844 }, { "epoch": 2.86, "learning_rate": 1.2871025794841031e-06, "loss": 0.0002, "step": 23845 }, { "epoch": 2.86, "learning_rate": 1.2870305938812237e-06, "loss": 0.0002, "step": 23846 }, { "epoch": 2.86, "learning_rate": 1.2869586082783442e-06, "loss": 0.0001, "step": 23847 }, { "epoch": 2.86, "learning_rate": 1.2868866226754648e-06, "loss": 0.0002, "step": 23848 }, { "epoch": 2.86, "learning_rate": 1.2868146370725856e-06, "loss": 0.0002, "step": 23849 }, { "epoch": 2.86, "learning_rate": 1.2867426514697062e-06, "loss": 0.0001, "step": 23850 }, { "epoch": 2.86, "learning_rate": 1.2866706658668267e-06, "loss": 0.0004, "step": 23851 }, { "epoch": 2.86, "learning_rate": 1.2865986802639473e-06, "loss": 0.6461, "step": 23852 }, { "epoch": 2.86, "learning_rate": 1.2865266946610679e-06, "loss": 0.0003, "step": 23853 }, { "epoch": 2.86, "learning_rate": 1.2864547090581884e-06, "loss": 0.0002, "step": 23854 }, { "epoch": 2.86, "learning_rate": 1.286382723455309e-06, "loss": 0.0002, "step": 23855 }, { "epoch": 2.86, "learning_rate": 1.2863107378524296e-06, "loss": 0.0007, "step": 23856 }, { "epoch": 2.86, "learning_rate": 1.2862387522495501e-06, "loss": 0.0002, "step": 23857 }, { "epoch": 2.86, "learning_rate": 1.2861667666466707e-06, "loss": 0.0001, "step": 23858 }, { "epoch": 2.86, "learning_rate": 1.2860947810437913e-06, "loss": 0.0003, "step": 23859 }, { "epoch": 2.86, "learning_rate": 1.2860227954409118e-06, "loss": 0.0002, "step": 23860 }, { "epoch": 2.86, "learning_rate": 1.2859508098380324e-06, "loss": 0.0001, "step": 23861 }, { "epoch": 2.86, "learning_rate": 1.285878824235153e-06, "loss": 0.7756, "step": 23862 }, { "epoch": 2.86, "learning_rate": 1.2858068386322735e-06, "loss": 0.0001, "step": 23863 }, { "epoch": 2.86, "learning_rate": 1.285734853029394e-06, "loss": 0.0002, "step": 23864 }, { "epoch": 2.86, "learning_rate": 1.2856628674265149e-06, "loss": 0.0003, "step": 23865 }, { "epoch": 2.86, "learning_rate": 1.2855908818236354e-06, "loss": 0.0001, "step": 23866 }, { "epoch": 2.86, "learning_rate": 1.2855188962207558e-06, "loss": 0.0001, "step": 23867 }, { "epoch": 2.86, "learning_rate": 1.2854469106178764e-06, "loss": 0.0002, "step": 23868 }, { "epoch": 2.86, "learning_rate": 1.285374925014997e-06, "loss": 0.0002, "step": 23869 }, { "epoch": 2.86, "learning_rate": 1.2853029394121175e-06, "loss": 0.0001, "step": 23870 }, { "epoch": 2.86, "learning_rate": 1.285230953809238e-06, "loss": 0.0002, "step": 23871 }, { "epoch": 2.86, "learning_rate": 1.2851589682063588e-06, "loss": 0.0002, "step": 23872 }, { "epoch": 2.86, "learning_rate": 1.2850869826034794e-06, "loss": 0.0002, "step": 23873 }, { "epoch": 2.86, "learning_rate": 1.2850149970006e-06, "loss": 0.0001, "step": 23874 }, { "epoch": 2.86, "learning_rate": 1.2849430113977205e-06, "loss": 0.0007, "step": 23875 }, { "epoch": 2.86, "learning_rate": 1.2848710257948411e-06, "loss": 0.3297, "step": 23876 }, { "epoch": 2.86, "learning_rate": 1.2847990401919617e-06, "loss": 0.0005, "step": 23877 }, { "epoch": 2.86, "learning_rate": 1.2847270545890822e-06, "loss": 0.0003, "step": 23878 }, { "epoch": 2.86, "learning_rate": 1.2846550689862028e-06, "loss": 0.0002, "step": 23879 }, { "epoch": 2.86, "learning_rate": 1.2845830833833234e-06, "loss": 0.0001, "step": 23880 }, { "epoch": 2.86, "learning_rate": 1.284511097780444e-06, "loss": 0.0001, "step": 23881 }, { "epoch": 2.86, "learning_rate": 1.2844391121775645e-06, "loss": 0.0002, "step": 23882 }, { "epoch": 2.86, "learning_rate": 1.284367126574685e-06, "loss": 0.0003, "step": 23883 }, { "epoch": 2.86, "learning_rate": 1.2842951409718056e-06, "loss": 0.0002, "step": 23884 }, { "epoch": 2.86, "learning_rate": 1.2842231553689262e-06, "loss": 0.0001, "step": 23885 }, { "epoch": 2.86, "learning_rate": 1.2841511697660468e-06, "loss": 0.0001, "step": 23886 }, { "epoch": 2.86, "learning_rate": 1.2840791841631673e-06, "loss": 0.0002, "step": 23887 }, { "epoch": 2.86, "learning_rate": 1.2840071985602881e-06, "loss": 0.0002, "step": 23888 }, { "epoch": 2.86, "learning_rate": 1.2839352129574087e-06, "loss": 0.0004, "step": 23889 }, { "epoch": 2.86, "learning_rate": 1.283863227354529e-06, "loss": 0.0001, "step": 23890 }, { "epoch": 2.86, "learning_rate": 1.2837912417516496e-06, "loss": 0.0003, "step": 23891 }, { "epoch": 2.86, "learning_rate": 1.2837192561487702e-06, "loss": 0.0001, "step": 23892 }, { "epoch": 2.86, "learning_rate": 1.2836472705458907e-06, "loss": 0.0001, "step": 23893 }, { "epoch": 2.86, "learning_rate": 1.2835752849430113e-06, "loss": 0.0003, "step": 23894 }, { "epoch": 2.86, "learning_rate": 1.283503299340132e-06, "loss": 0.0002, "step": 23895 }, { "epoch": 2.86, "learning_rate": 1.2834313137372527e-06, "loss": 0.0002, "step": 23896 }, { "epoch": 2.86, "learning_rate": 1.2833593281343732e-06, "loss": 0.0002, "step": 23897 }, { "epoch": 2.86, "learning_rate": 1.2832873425314938e-06, "loss": 0.0004, "step": 23898 }, { "epoch": 2.86, "learning_rate": 1.2832153569286144e-06, "loss": 0.0003, "step": 23899 }, { "epoch": 2.86, "learning_rate": 1.283143371325735e-06, "loss": 0.0002, "step": 23900 }, { "epoch": 2.86, "learning_rate": 1.2830713857228555e-06, "loss": 0.0001, "step": 23901 }, { "epoch": 2.86, "learning_rate": 1.282999400119976e-06, "loss": 0.0003, "step": 23902 }, { "epoch": 2.86, "learning_rate": 1.2829274145170966e-06, "loss": 0.0001, "step": 23903 }, { "epoch": 2.86, "learning_rate": 1.2828554289142172e-06, "loss": 0.0009, "step": 23904 }, { "epoch": 2.86, "learning_rate": 1.2827834433113378e-06, "loss": 0.6657, "step": 23905 }, { "epoch": 2.86, "learning_rate": 1.2827114577084583e-06, "loss": 0.0002, "step": 23906 }, { "epoch": 2.86, "learning_rate": 1.2826394721055789e-06, "loss": 0.0001, "step": 23907 }, { "epoch": 2.86, "learning_rate": 1.2825674865026995e-06, "loss": 0.0001, "step": 23908 }, { "epoch": 2.87, "learning_rate": 1.28249550089982e-06, "loss": 0.0001, "step": 23909 }, { "epoch": 2.87, "learning_rate": 1.2824235152969406e-06, "loss": 0.0001, "step": 23910 }, { "epoch": 2.87, "learning_rate": 1.2823515296940614e-06, "loss": 0.0002, "step": 23911 }, { "epoch": 2.87, "learning_rate": 1.282279544091182e-06, "loss": 0.0001, "step": 23912 }, { "epoch": 2.87, "learning_rate": 1.2822075584883023e-06, "loss": 0.0001, "step": 23913 }, { "epoch": 2.87, "learning_rate": 1.2821355728854229e-06, "loss": 0.0002, "step": 23914 }, { "epoch": 2.87, "learning_rate": 1.2820635872825434e-06, "loss": 0.0003, "step": 23915 }, { "epoch": 2.87, "learning_rate": 1.281991601679664e-06, "loss": 0.0002, "step": 23916 }, { "epoch": 2.87, "learning_rate": 1.2819196160767846e-06, "loss": 0.0001, "step": 23917 }, { "epoch": 2.87, "learning_rate": 1.2818476304739053e-06, "loss": 0.6633, "step": 23918 }, { "epoch": 2.87, "learning_rate": 1.281775644871026e-06, "loss": 0.0001, "step": 23919 }, { "epoch": 2.87, "learning_rate": 1.2817036592681465e-06, "loss": 0.0001, "step": 23920 }, { "epoch": 2.87, "learning_rate": 1.281631673665267e-06, "loss": 0.0002, "step": 23921 }, { "epoch": 2.87, "learning_rate": 1.2815596880623876e-06, "loss": 0.0003, "step": 23922 }, { "epoch": 2.87, "learning_rate": 1.2814877024595082e-06, "loss": 0.0002, "step": 23923 }, { "epoch": 2.87, "learning_rate": 1.2814157168566287e-06, "loss": 0.0002, "step": 23924 }, { "epoch": 2.87, "learning_rate": 1.2813437312537493e-06, "loss": 0.0004, "step": 23925 }, { "epoch": 2.87, "learning_rate": 1.2812717456508699e-06, "loss": 0.0488, "step": 23926 }, { "epoch": 2.87, "learning_rate": 1.2811997600479904e-06, "loss": 0.0054, "step": 23927 }, { "epoch": 2.87, "learning_rate": 1.281127774445111e-06, "loss": 0.0003, "step": 23928 }, { "epoch": 2.87, "learning_rate": 1.2810557888422316e-06, "loss": 0.0002, "step": 23929 }, { "epoch": 2.87, "learning_rate": 1.2809838032393521e-06, "loss": 0.0003, "step": 23930 }, { "epoch": 2.87, "learning_rate": 1.2809118176364727e-06, "loss": 0.0001, "step": 23931 }, { "epoch": 2.87, "learning_rate": 1.2808398320335933e-06, "loss": 0.0004, "step": 23932 }, { "epoch": 2.87, "learning_rate": 1.2807678464307138e-06, "loss": 0.0009, "step": 23933 }, { "epoch": 2.87, "learning_rate": 1.2806958608278346e-06, "loss": 0.0002, "step": 23934 }, { "epoch": 2.87, "learning_rate": 1.2806238752249552e-06, "loss": 0.0003, "step": 23935 }, { "epoch": 2.87, "learning_rate": 1.2805518896220758e-06, "loss": 0.0001, "step": 23936 }, { "epoch": 2.87, "learning_rate": 1.2804799040191961e-06, "loss": 0.9013, "step": 23937 }, { "epoch": 2.87, "learning_rate": 1.2804079184163167e-06, "loss": 0.0002, "step": 23938 }, { "epoch": 2.87, "learning_rate": 1.2803359328134372e-06, "loss": 0.0003, "step": 23939 }, { "epoch": 2.87, "learning_rate": 1.2802639472105578e-06, "loss": 0.0001, "step": 23940 }, { "epoch": 2.87, "learning_rate": 1.2801919616076786e-06, "loss": 0.0002, "step": 23941 }, { "epoch": 2.87, "learning_rate": 1.2801199760047992e-06, "loss": 0.0002, "step": 23942 }, { "epoch": 2.87, "learning_rate": 1.2800479904019197e-06, "loss": 0.0002, "step": 23943 }, { "epoch": 2.87, "learning_rate": 1.2799760047990403e-06, "loss": 0.0002, "step": 23944 }, { "epoch": 2.87, "learning_rate": 1.2799040191961609e-06, "loss": 0.0012, "step": 23945 }, { "epoch": 2.87, "learning_rate": 1.2798320335932814e-06, "loss": 0.0001, "step": 23946 }, { "epoch": 2.87, "learning_rate": 1.279760047990402e-06, "loss": 0.0002, "step": 23947 }, { "epoch": 2.87, "learning_rate": 1.2796880623875226e-06, "loss": 0.0002, "step": 23948 }, { "epoch": 2.87, "learning_rate": 1.2796160767846431e-06, "loss": 0.0002, "step": 23949 }, { "epoch": 2.87, "learning_rate": 1.2795440911817637e-06, "loss": 0.0002, "step": 23950 }, { "epoch": 2.87, "learning_rate": 1.2794721055788843e-06, "loss": 0.0001, "step": 23951 }, { "epoch": 2.87, "learning_rate": 1.2794001199760048e-06, "loss": 0.0003, "step": 23952 }, { "epoch": 2.87, "learning_rate": 1.2793281343731254e-06, "loss": 0.0001, "step": 23953 }, { "epoch": 2.87, "learning_rate": 1.279256148770246e-06, "loss": 0.0004, "step": 23954 }, { "epoch": 2.87, "learning_rate": 1.2791841631673665e-06, "loss": 0.0012, "step": 23955 }, { "epoch": 2.87, "learning_rate": 1.279112177564487e-06, "loss": 0.0002, "step": 23956 }, { "epoch": 2.87, "learning_rate": 1.2790401919616079e-06, "loss": 0.0002, "step": 23957 }, { "epoch": 2.87, "learning_rate": 1.2789682063587284e-06, "loss": 0.0003, "step": 23958 }, { "epoch": 2.87, "learning_rate": 1.278896220755849e-06, "loss": 0.0002, "step": 23959 }, { "epoch": 2.87, "learning_rate": 1.2788242351529694e-06, "loss": 0.0002, "step": 23960 }, { "epoch": 2.87, "learning_rate": 1.27875224955009e-06, "loss": 0.0001, "step": 23961 }, { "epoch": 2.87, "learning_rate": 1.2786802639472105e-06, "loss": 0.0064, "step": 23962 }, { "epoch": 2.87, "learning_rate": 1.278608278344331e-06, "loss": 0.0028, "step": 23963 }, { "epoch": 2.87, "learning_rate": 1.2785362927414518e-06, "loss": 0.0005, "step": 23964 }, { "epoch": 2.87, "learning_rate": 1.2784643071385724e-06, "loss": 0.0001, "step": 23965 }, { "epoch": 2.87, "learning_rate": 1.278392321535693e-06, "loss": 0.0001, "step": 23966 }, { "epoch": 2.87, "learning_rate": 1.2783203359328135e-06, "loss": 0.0058, "step": 23967 }, { "epoch": 2.87, "learning_rate": 1.278248350329934e-06, "loss": 0.0001, "step": 23968 }, { "epoch": 2.87, "learning_rate": 1.2781763647270547e-06, "loss": 0.0001, "step": 23969 }, { "epoch": 2.87, "learning_rate": 1.2781043791241752e-06, "loss": 0.0001, "step": 23970 }, { "epoch": 2.87, "learning_rate": 1.2780323935212956e-06, "loss": 1.0077, "step": 23971 }, { "epoch": 2.87, "learning_rate": 1.2779604079184164e-06, "loss": 0.0001, "step": 23972 }, { "epoch": 2.87, "learning_rate": 1.277888422315537e-06, "loss": 0.0001, "step": 23973 }, { "epoch": 2.87, "learning_rate": 1.2778164367126575e-06, "loss": 0.0001, "step": 23974 }, { "epoch": 2.87, "learning_rate": 1.277744451109778e-06, "loss": 0.0003, "step": 23975 }, { "epoch": 2.87, "learning_rate": 1.2776724655068986e-06, "loss": 0.0002, "step": 23976 }, { "epoch": 2.87, "learning_rate": 1.2776004799040192e-06, "loss": 0.0001, "step": 23977 }, { "epoch": 2.87, "learning_rate": 1.2775284943011398e-06, "loss": 0.0002, "step": 23978 }, { "epoch": 2.87, "learning_rate": 1.2774565086982603e-06, "loss": 0.0001, "step": 23979 }, { "epoch": 2.87, "learning_rate": 1.277384523095381e-06, "loss": 0.0001, "step": 23980 }, { "epoch": 2.87, "learning_rate": 1.2773125374925017e-06, "loss": 0.0002, "step": 23981 }, { "epoch": 2.87, "learning_rate": 1.2772405518896222e-06, "loss": 0.0007, "step": 23982 }, { "epoch": 2.87, "learning_rate": 1.2771685662867426e-06, "loss": 0.0002, "step": 23983 }, { "epoch": 2.87, "learning_rate": 1.2770965806838632e-06, "loss": 0.0002, "step": 23984 }, { "epoch": 2.87, "learning_rate": 1.2770245950809837e-06, "loss": 0.0001, "step": 23985 }, { "epoch": 2.87, "learning_rate": 1.2769526094781043e-06, "loss": 0.0004, "step": 23986 }, { "epoch": 2.87, "learning_rate": 1.2768806238752249e-06, "loss": 0.0002, "step": 23987 }, { "epoch": 2.87, "learning_rate": 1.2768086382723456e-06, "loss": 0.001, "step": 23988 }, { "epoch": 2.87, "learning_rate": 1.2767366526694662e-06, "loss": 0.0001, "step": 23989 }, { "epoch": 2.87, "learning_rate": 1.2766646670665868e-06, "loss": 0.0002, "step": 23990 }, { "epoch": 2.87, "learning_rate": 1.2765926814637073e-06, "loss": 0.0003, "step": 23991 }, { "epoch": 2.88, "learning_rate": 1.276520695860828e-06, "loss": 0.0002, "step": 23992 }, { "epoch": 2.88, "learning_rate": 1.2764487102579485e-06, "loss": 0.9459, "step": 23993 }, { "epoch": 2.88, "learning_rate": 1.2763767246550688e-06, "loss": 0.0003, "step": 23994 }, { "epoch": 2.88, "learning_rate": 1.2763047390521896e-06, "loss": 0.0002, "step": 23995 }, { "epoch": 2.88, "learning_rate": 1.2762327534493102e-06, "loss": 0.0002, "step": 23996 }, { "epoch": 2.88, "learning_rate": 1.2761607678464307e-06, "loss": 0.0002, "step": 23997 }, { "epoch": 2.88, "learning_rate": 1.2760887822435513e-06, "loss": 0.0003, "step": 23998 }, { "epoch": 2.88, "learning_rate": 1.2760167966406719e-06, "loss": 0.0002, "step": 23999 }, { "epoch": 2.88, "learning_rate": 1.2759448110377925e-06, "loss": 0.0001, "step": 24000 }, { "epoch": 2.88, "learning_rate": 1.275872825434913e-06, "loss": 0.0002, "step": 24001 }, { "epoch": 2.88, "learning_rate": 1.2758008398320336e-06, "loss": 0.0002, "step": 24002 }, { "epoch": 2.88, "learning_rate": 1.2757288542291542e-06, "loss": 0.0003, "step": 24003 }, { "epoch": 2.88, "learning_rate": 1.275656868626275e-06, "loss": 0.0002, "step": 24004 }, { "epoch": 2.88, "learning_rate": 1.2755848830233955e-06, "loss": 0.0002, "step": 24005 }, { "epoch": 2.88, "learning_rate": 1.2755128974205159e-06, "loss": 0.0001, "step": 24006 }, { "epoch": 2.88, "learning_rate": 1.2754409118176364e-06, "loss": 0.0003, "step": 24007 }, { "epoch": 2.88, "learning_rate": 1.275368926214757e-06, "loss": 0.0003, "step": 24008 }, { "epoch": 2.88, "learning_rate": 1.2752969406118776e-06, "loss": 0.0002, "step": 24009 }, { "epoch": 2.88, "learning_rate": 1.2752249550089981e-06, "loss": 0.0002, "step": 24010 }, { "epoch": 2.88, "learning_rate": 1.275152969406119e-06, "loss": 0.0001, "step": 24011 }, { "epoch": 2.88, "learning_rate": 1.2750809838032395e-06, "loss": 0.0002, "step": 24012 }, { "epoch": 2.88, "learning_rate": 1.27500899820036e-06, "loss": 0.0001, "step": 24013 }, { "epoch": 2.88, "learning_rate": 1.2749370125974806e-06, "loss": 0.0001, "step": 24014 }, { "epoch": 2.88, "learning_rate": 1.2748650269946012e-06, "loss": 0.0003, "step": 24015 }, { "epoch": 2.88, "learning_rate": 1.2747930413917217e-06, "loss": 0.0007, "step": 24016 }, { "epoch": 2.88, "learning_rate": 1.274721055788842e-06, "loss": 0.0001, "step": 24017 }, { "epoch": 2.88, "learning_rate": 1.2746490701859629e-06, "loss": 0.0001, "step": 24018 }, { "epoch": 2.88, "learning_rate": 1.2745770845830834e-06, "loss": 0.0021, "step": 24019 }, { "epoch": 2.88, "learning_rate": 1.274505098980204e-06, "loss": 0.0001, "step": 24020 }, { "epoch": 2.88, "learning_rate": 1.2744331133773246e-06, "loss": 0.0001, "step": 24021 }, { "epoch": 2.88, "learning_rate": 1.2743611277744451e-06, "loss": 0.0002, "step": 24022 }, { "epoch": 2.88, "learning_rate": 1.2742891421715657e-06, "loss": 0.0001, "step": 24023 }, { "epoch": 2.88, "learning_rate": 1.2742171565686863e-06, "loss": 0.0001, "step": 24024 }, { "epoch": 2.88, "learning_rate": 1.2741451709658068e-06, "loss": 0.0003, "step": 24025 }, { "epoch": 2.88, "learning_rate": 1.2740731853629274e-06, "loss": 0.0003, "step": 24026 }, { "epoch": 2.88, "learning_rate": 1.2740011997600482e-06, "loss": 0.0001, "step": 24027 }, { "epoch": 2.88, "learning_rate": 1.2739292141571687e-06, "loss": 0.0002, "step": 24028 }, { "epoch": 2.88, "learning_rate": 1.273857228554289e-06, "loss": 0.0002, "step": 24029 }, { "epoch": 2.88, "learning_rate": 1.2737852429514097e-06, "loss": 0.0004, "step": 24030 }, { "epoch": 2.88, "learning_rate": 1.2737132573485302e-06, "loss": 0.0003, "step": 24031 }, { "epoch": 2.88, "learning_rate": 1.2736412717456508e-06, "loss": 0.0002, "step": 24032 }, { "epoch": 2.88, "learning_rate": 1.2735692861427714e-06, "loss": 0.0001, "step": 24033 }, { "epoch": 2.88, "learning_rate": 1.2734973005398921e-06, "loss": 0.0861, "step": 24034 }, { "epoch": 2.88, "learning_rate": 1.2734253149370127e-06, "loss": 0.0003, "step": 24035 }, { "epoch": 2.88, "learning_rate": 1.2733533293341333e-06, "loss": 0.0002, "step": 24036 }, { "epoch": 2.88, "learning_rate": 1.2732813437312538e-06, "loss": 0.0001, "step": 24037 }, { "epoch": 2.88, "learning_rate": 1.2732093581283744e-06, "loss": 0.0001, "step": 24038 }, { "epoch": 2.88, "learning_rate": 1.273137372525495e-06, "loss": 0.0001, "step": 24039 }, { "epoch": 2.88, "learning_rate": 1.2730653869226153e-06, "loss": 0.7306, "step": 24040 }, { "epoch": 2.88, "learning_rate": 1.2729934013197361e-06, "loss": 0.0003, "step": 24041 }, { "epoch": 2.88, "learning_rate": 1.2729214157168567e-06, "loss": 0.0002, "step": 24042 }, { "epoch": 2.88, "learning_rate": 1.2728494301139772e-06, "loss": 0.0005, "step": 24043 }, { "epoch": 2.88, "learning_rate": 1.2727774445110978e-06, "loss": 0.0003, "step": 24044 }, { "epoch": 2.88, "learning_rate": 1.2727054589082184e-06, "loss": 0.0001, "step": 24045 }, { "epoch": 2.88, "learning_rate": 1.272633473305339e-06, "loss": 0.0001, "step": 24046 }, { "epoch": 2.88, "learning_rate": 1.2725614877024595e-06, "loss": 0.0006, "step": 24047 }, { "epoch": 2.88, "learning_rate": 1.27248950209958e-06, "loss": 0.0001, "step": 24048 }, { "epoch": 2.88, "learning_rate": 1.2724175164967006e-06, "loss": 0.0001, "step": 24049 }, { "epoch": 2.88, "learning_rate": 1.2723455308938214e-06, "loss": 0.0002, "step": 24050 }, { "epoch": 2.88, "learning_rate": 1.272273545290942e-06, "loss": 0.0002, "step": 24051 }, { "epoch": 2.88, "learning_rate": 1.2722015596880623e-06, "loss": 0.0001, "step": 24052 }, { "epoch": 2.88, "learning_rate": 1.272129574085183e-06, "loss": 0.0006, "step": 24053 }, { "epoch": 2.88, "learning_rate": 1.2720575884823035e-06, "loss": 0.0001, "step": 24054 }, { "epoch": 2.88, "learning_rate": 1.271985602879424e-06, "loss": 0.0001, "step": 24055 }, { "epoch": 2.88, "learning_rate": 1.2719136172765446e-06, "loss": 0.0002, "step": 24056 }, { "epoch": 2.88, "learning_rate": 1.2718416316736654e-06, "loss": 0.0001, "step": 24057 }, { "epoch": 2.88, "learning_rate": 1.271769646070786e-06, "loss": 0.0007, "step": 24058 }, { "epoch": 2.88, "learning_rate": 1.2716976604679065e-06, "loss": 0.0002, "step": 24059 }, { "epoch": 2.88, "learning_rate": 1.271625674865027e-06, "loss": 0.0004, "step": 24060 }, { "epoch": 2.88, "learning_rate": 1.2715536892621477e-06, "loss": 0.0002, "step": 24061 }, { "epoch": 2.88, "learning_rate": 1.2714817036592682e-06, "loss": 0.0004, "step": 24062 }, { "epoch": 2.88, "learning_rate": 1.2714097180563888e-06, "loss": 0.0002, "step": 24063 }, { "epoch": 2.88, "learning_rate": 1.2713377324535094e-06, "loss": 0.0001, "step": 24064 }, { "epoch": 2.88, "learning_rate": 1.27126574685063e-06, "loss": 0.0002, "step": 24065 }, { "epoch": 2.88, "learning_rate": 1.2711937612477505e-06, "loss": 0.0002, "step": 24066 }, { "epoch": 2.88, "learning_rate": 1.271121775644871e-06, "loss": 0.0003, "step": 24067 }, { "epoch": 2.88, "learning_rate": 1.2710497900419916e-06, "loss": 0.0002, "step": 24068 }, { "epoch": 2.88, "learning_rate": 1.2709778044391122e-06, "loss": 0.0001, "step": 24069 }, { "epoch": 2.88, "learning_rate": 1.2709058188362328e-06, "loss": 0.0002, "step": 24070 }, { "epoch": 2.88, "learning_rate": 1.2708338332333533e-06, "loss": 0.0001, "step": 24071 }, { "epoch": 2.88, "learning_rate": 1.270761847630474e-06, "loss": 0.0002, "step": 24072 }, { "epoch": 2.88, "learning_rate": 1.2706898620275947e-06, "loss": 0.0002, "step": 24073 }, { "epoch": 2.88, "learning_rate": 1.2706178764247152e-06, "loss": 0.8028, "step": 24074 }, { "epoch": 2.88, "learning_rate": 1.2705458908218356e-06, "loss": 0.0002, "step": 24075 }, { "epoch": 2.89, "learning_rate": 1.2704739052189562e-06, "loss": 0.0002, "step": 24076 }, { "epoch": 2.89, "learning_rate": 1.2704019196160767e-06, "loss": 0.0001, "step": 24077 }, { "epoch": 2.89, "learning_rate": 1.2703299340131973e-06, "loss": 0.0006, "step": 24078 }, { "epoch": 2.89, "learning_rate": 1.2702579484103179e-06, "loss": 0.0002, "step": 24079 }, { "epoch": 2.89, "learning_rate": 1.2701859628074386e-06, "loss": 0.0, "step": 24080 }, { "epoch": 2.89, "learning_rate": 1.2701139772045592e-06, "loss": 0.0003, "step": 24081 }, { "epoch": 2.89, "learning_rate": 1.2700419916016798e-06, "loss": 0.0002, "step": 24082 }, { "epoch": 2.89, "learning_rate": 1.2699700059988003e-06, "loss": 0.0024, "step": 24083 }, { "epoch": 2.89, "learning_rate": 1.269898020395921e-06, "loss": 0.0001, "step": 24084 }, { "epoch": 2.89, "learning_rate": 1.2698260347930415e-06, "loss": 0.0003, "step": 24085 }, { "epoch": 2.89, "learning_rate": 1.269754049190162e-06, "loss": 0.0007, "step": 24086 }, { "epoch": 2.89, "learning_rate": 1.2696820635872826e-06, "loss": 0.0001, "step": 24087 }, { "epoch": 2.89, "learning_rate": 1.2696100779844032e-06, "loss": 0.0002, "step": 24088 }, { "epoch": 2.89, "learning_rate": 1.2695380923815237e-06, "loss": 0.9546, "step": 24089 }, { "epoch": 2.89, "learning_rate": 1.2694661067786443e-06, "loss": 0.0002, "step": 24090 }, { "epoch": 2.89, "learning_rate": 1.2693941211757649e-06, "loss": 0.0005, "step": 24091 }, { "epoch": 2.89, "learning_rate": 1.2693221355728854e-06, "loss": 0.0001, "step": 24092 }, { "epoch": 2.89, "learning_rate": 1.269250149970006e-06, "loss": 0.0002, "step": 24093 }, { "epoch": 2.89, "learning_rate": 1.2691781643671266e-06, "loss": 0.0002, "step": 24094 }, { "epoch": 2.89, "learning_rate": 1.2691061787642471e-06, "loss": 0.0001, "step": 24095 }, { "epoch": 2.89, "learning_rate": 1.269034193161368e-06, "loss": 0.0002, "step": 24096 }, { "epoch": 2.89, "learning_rate": 1.2689622075584885e-06, "loss": 0.0002, "step": 24097 }, { "epoch": 2.89, "learning_rate": 1.2688902219556088e-06, "loss": 0.0002, "step": 24098 }, { "epoch": 2.89, "learning_rate": 1.2688182363527294e-06, "loss": 0.0005, "step": 24099 }, { "epoch": 2.89, "learning_rate": 1.26874625074985e-06, "loss": 0.0005, "step": 24100 }, { "epoch": 2.89, "learning_rate": 1.2686742651469705e-06, "loss": 0.0002, "step": 24101 }, { "epoch": 2.89, "learning_rate": 1.2686022795440911e-06, "loss": 0.0001, "step": 24102 }, { "epoch": 2.89, "learning_rate": 1.2685302939412119e-06, "loss": 0.0002, "step": 24103 }, { "epoch": 2.89, "learning_rate": 1.2684583083383325e-06, "loss": 0.0003, "step": 24104 }, { "epoch": 2.89, "learning_rate": 1.268386322735453e-06, "loss": 0.0, "step": 24105 }, { "epoch": 2.89, "learning_rate": 1.2683143371325736e-06, "loss": 0.0002, "step": 24106 }, { "epoch": 2.89, "learning_rate": 1.2682423515296942e-06, "loss": 0.0002, "step": 24107 }, { "epoch": 2.89, "learning_rate": 1.2681703659268147e-06, "loss": 0.0002, "step": 24108 }, { "epoch": 2.89, "learning_rate": 1.2680983803239353e-06, "loss": 0.0006, "step": 24109 }, { "epoch": 2.89, "learning_rate": 1.2680263947210556e-06, "loss": 0.0001, "step": 24110 }, { "epoch": 2.89, "learning_rate": 1.2679544091181764e-06, "loss": 0.0022, "step": 24111 }, { "epoch": 2.89, "learning_rate": 1.267882423515297e-06, "loss": 0.0003, "step": 24112 }, { "epoch": 2.89, "learning_rate": 1.2678104379124176e-06, "loss": 0.0001, "step": 24113 }, { "epoch": 2.89, "learning_rate": 1.2677384523095381e-06, "loss": 0.0001, "step": 24114 }, { "epoch": 2.89, "learning_rate": 1.2676664667066587e-06, "loss": 0.0001, "step": 24115 }, { "epoch": 2.89, "learning_rate": 1.2675944811037793e-06, "loss": 0.0001, "step": 24116 }, { "epoch": 2.89, "learning_rate": 1.2675224955008998e-06, "loss": 0.0001, "step": 24117 }, { "epoch": 2.89, "learning_rate": 1.2674505098980204e-06, "loss": 0.0001, "step": 24118 }, { "epoch": 2.89, "learning_rate": 1.267378524295141e-06, "loss": 0.0008, "step": 24119 }, { "epoch": 2.89, "learning_rate": 1.2673065386922617e-06, "loss": 0.0003, "step": 24120 }, { "epoch": 2.89, "learning_rate": 1.267234553089382e-06, "loss": 0.0002, "step": 24121 }, { "epoch": 2.89, "learning_rate": 1.2671625674865027e-06, "loss": 0.0002, "step": 24122 }, { "epoch": 2.89, "learning_rate": 1.2670905818836232e-06, "loss": 0.0003, "step": 24123 }, { "epoch": 2.89, "learning_rate": 1.2670185962807438e-06, "loss": 0.0001, "step": 24124 }, { "epoch": 2.89, "learning_rate": 1.2669466106778644e-06, "loss": 0.0002, "step": 24125 }, { "epoch": 2.89, "learning_rate": 1.266874625074985e-06, "loss": 0.0232, "step": 24126 }, { "epoch": 2.89, "learning_rate": 1.2668026394721057e-06, "loss": 0.0001, "step": 24127 }, { "epoch": 2.89, "learning_rate": 1.2667306538692263e-06, "loss": 0.0003, "step": 24128 }, { "epoch": 2.89, "learning_rate": 1.2666586682663468e-06, "loss": 0.0002, "step": 24129 }, { "epoch": 2.89, "learning_rate": 1.2665866826634674e-06, "loss": 0.0003, "step": 24130 }, { "epoch": 2.89, "learning_rate": 1.266514697060588e-06, "loss": 0.0002, "step": 24131 }, { "epoch": 2.89, "learning_rate": 1.2664427114577085e-06, "loss": 0.0001, "step": 24132 }, { "epoch": 2.89, "learning_rate": 1.2663707258548289e-06, "loss": 0.0001, "step": 24133 }, { "epoch": 2.89, "learning_rate": 1.2662987402519497e-06, "loss": 0.0003, "step": 24134 }, { "epoch": 2.89, "learning_rate": 1.2662267546490702e-06, "loss": 0.0001, "step": 24135 }, { "epoch": 2.89, "learning_rate": 1.2661547690461908e-06, "loss": 0.0002, "step": 24136 }, { "epoch": 2.89, "learning_rate": 1.2660827834433114e-06, "loss": 0.0002, "step": 24137 }, { "epoch": 2.89, "learning_rate": 1.266010797840432e-06, "loss": 0.0002, "step": 24138 }, { "epoch": 2.89, "learning_rate": 1.2659388122375525e-06, "loss": 0.0002, "step": 24139 }, { "epoch": 2.89, "learning_rate": 1.265866826634673e-06, "loss": 0.0001, "step": 24140 }, { "epoch": 2.89, "learning_rate": 1.2657948410317936e-06, "loss": 0.0002, "step": 24141 }, { "epoch": 2.89, "learning_rate": 1.2657228554289142e-06, "loss": 0.0001, "step": 24142 }, { "epoch": 2.89, "learning_rate": 1.265650869826035e-06, "loss": 0.0002, "step": 24143 }, { "epoch": 2.89, "learning_rate": 1.2655788842231553e-06, "loss": 0.0001, "step": 24144 }, { "epoch": 2.89, "learning_rate": 1.265506898620276e-06, "loss": 0.0004, "step": 24145 }, { "epoch": 2.89, "learning_rate": 1.2654349130173965e-06, "loss": 0.0003, "step": 24146 }, { "epoch": 2.89, "learning_rate": 1.265362927414517e-06, "loss": 0.0055, "step": 24147 }, { "epoch": 2.89, "learning_rate": 1.2652909418116376e-06, "loss": 0.0011, "step": 24148 }, { "epoch": 2.89, "learning_rate": 1.2652189562087582e-06, "loss": 0.0001, "step": 24149 }, { "epoch": 2.89, "learning_rate": 1.265146970605879e-06, "loss": 0.0002, "step": 24150 }, { "epoch": 2.89, "learning_rate": 1.2650749850029995e-06, "loss": 0.0002, "step": 24151 }, { "epoch": 2.89, "learning_rate": 1.26500299940012e-06, "loss": 0.0006, "step": 24152 }, { "epoch": 2.89, "learning_rate": 1.2649310137972407e-06, "loss": 0.0003, "step": 24153 }, { "epoch": 2.89, "learning_rate": 1.2648590281943612e-06, "loss": 0.0923, "step": 24154 }, { "epoch": 2.89, "learning_rate": 1.2647870425914818e-06, "loss": 0.0002, "step": 24155 }, { "epoch": 2.89, "learning_rate": 1.2647150569886021e-06, "loss": 0.5317, "step": 24156 }, { "epoch": 2.89, "learning_rate": 1.264643071385723e-06, "loss": 0.0003, "step": 24157 }, { "epoch": 2.89, "learning_rate": 1.2645710857828435e-06, "loss": 0.0001, "step": 24158 }, { "epoch": 2.9, "learning_rate": 1.264499100179964e-06, "loss": 0.0001, "step": 24159 }, { "epoch": 2.9, "learning_rate": 1.2644271145770846e-06, "loss": 0.0002, "step": 24160 }, { "epoch": 2.9, "learning_rate": 1.2643551289742052e-06, "loss": 0.0001, "step": 24161 }, { "epoch": 2.9, "learning_rate": 1.2642831433713258e-06, "loss": 0.0001, "step": 24162 }, { "epoch": 2.9, "learning_rate": 1.2642111577684463e-06, "loss": 0.0002, "step": 24163 }, { "epoch": 2.9, "learning_rate": 1.2641391721655669e-06, "loss": 0.0047, "step": 24164 }, { "epoch": 2.9, "learning_rate": 1.2640671865626875e-06, "loss": 0.001, "step": 24165 }, { "epoch": 2.9, "learning_rate": 1.2639952009598082e-06, "loss": 0.0002, "step": 24166 }, { "epoch": 2.9, "learning_rate": 1.2639232153569286e-06, "loss": 0.0001, "step": 24167 }, { "epoch": 2.9, "learning_rate": 1.2638512297540492e-06, "loss": 0.0002, "step": 24168 }, { "epoch": 2.9, "learning_rate": 1.2637792441511697e-06, "loss": 0.0003, "step": 24169 }, { "epoch": 2.9, "learning_rate": 1.2637072585482903e-06, "loss": 0.0001, "step": 24170 }, { "epoch": 2.9, "learning_rate": 1.2636352729454109e-06, "loss": 0.0002, "step": 24171 }, { "epoch": 2.9, "learning_rate": 1.2635632873425314e-06, "loss": 0.0001, "step": 24172 }, { "epoch": 2.9, "learning_rate": 1.2634913017396522e-06, "loss": 0.0001, "step": 24173 }, { "epoch": 2.9, "learning_rate": 1.2634193161367728e-06, "loss": 0.0001, "step": 24174 }, { "epoch": 2.9, "learning_rate": 1.2633473305338933e-06, "loss": 0.0002, "step": 24175 }, { "epoch": 2.9, "learning_rate": 1.263275344931014e-06, "loss": 0.0003, "step": 24176 }, { "epoch": 2.9, "learning_rate": 1.2632033593281345e-06, "loss": 0.0003, "step": 24177 }, { "epoch": 2.9, "learning_rate": 1.263131373725255e-06, "loss": 0.0002, "step": 24178 }, { "epoch": 2.9, "learning_rate": 1.2630593881223754e-06, "loss": 0.0001, "step": 24179 }, { "epoch": 2.9, "learning_rate": 1.2629874025194962e-06, "loss": 0.0001, "step": 24180 }, { "epoch": 2.9, "learning_rate": 1.2629154169166167e-06, "loss": 0.0002, "step": 24181 }, { "epoch": 2.9, "learning_rate": 1.2628434313137373e-06, "loss": 0.0005, "step": 24182 }, { "epoch": 2.9, "learning_rate": 1.2627714457108579e-06, "loss": 0.0001, "step": 24183 }, { "epoch": 2.9, "learning_rate": 1.2626994601079784e-06, "loss": 0.0002, "step": 24184 }, { "epoch": 2.9, "learning_rate": 1.262627474505099e-06, "loss": 0.0001, "step": 24185 }, { "epoch": 2.9, "learning_rate": 1.2625554889022196e-06, "loss": 0.0002, "step": 24186 }, { "epoch": 2.9, "learning_rate": 1.2624835032993401e-06, "loss": 0.0001, "step": 24187 }, { "epoch": 2.9, "learning_rate": 1.2624115176964607e-06, "loss": 0.0002, "step": 24188 }, { "epoch": 2.9, "learning_rate": 1.2623395320935815e-06, "loss": 0.0001, "step": 24189 }, { "epoch": 2.9, "learning_rate": 1.262267546490702e-06, "loss": 0.5549, "step": 24190 }, { "epoch": 2.9, "learning_rate": 1.2621955608878224e-06, "loss": 0.0001, "step": 24191 }, { "epoch": 2.9, "learning_rate": 1.262123575284943e-06, "loss": 0.0005, "step": 24192 }, { "epoch": 2.9, "learning_rate": 1.2620515896820635e-06, "loss": 0.0002, "step": 24193 }, { "epoch": 2.9, "learning_rate": 1.261979604079184e-06, "loss": 0.0008, "step": 24194 }, { "epoch": 2.9, "learning_rate": 1.2619076184763047e-06, "loss": 0.0001, "step": 24195 }, { "epoch": 2.9, "learning_rate": 1.2618356328734254e-06, "loss": 0.0001, "step": 24196 }, { "epoch": 2.9, "learning_rate": 1.261763647270546e-06, "loss": 0.0002, "step": 24197 }, { "epoch": 2.9, "learning_rate": 1.2616916616676666e-06, "loss": 0.0001, "step": 24198 }, { "epoch": 2.9, "learning_rate": 1.2616196760647871e-06, "loss": 0.0002, "step": 24199 }, { "epoch": 2.9, "learning_rate": 1.2615476904619077e-06, "loss": 0.0002, "step": 24200 }, { "epoch": 2.9, "learning_rate": 1.2614757048590283e-06, "loss": 0.5259, "step": 24201 }, { "epoch": 2.9, "learning_rate": 1.2614037192561486e-06, "loss": 0.0001, "step": 24202 }, { "epoch": 2.9, "learning_rate": 1.2613317336532694e-06, "loss": 0.0685, "step": 24203 }, { "epoch": 2.9, "learning_rate": 1.26125974805039e-06, "loss": 0.0001, "step": 24204 }, { "epoch": 2.9, "learning_rate": 1.2611877624475105e-06, "loss": 0.0001, "step": 24205 }, { "epoch": 2.9, "learning_rate": 1.2611157768446311e-06, "loss": 0.0001, "step": 24206 }, { "epoch": 2.9, "learning_rate": 1.2610437912417517e-06, "loss": 0.0002, "step": 24207 }, { "epoch": 2.9, "learning_rate": 1.2609718056388722e-06, "loss": 0.0001, "step": 24208 }, { "epoch": 2.9, "learning_rate": 1.2608998200359928e-06, "loss": 0.0001, "step": 24209 }, { "epoch": 2.9, "learning_rate": 1.2608278344331134e-06, "loss": 0.0002, "step": 24210 }, { "epoch": 2.9, "learning_rate": 1.260755848830234e-06, "loss": 0.0002, "step": 24211 }, { "epoch": 2.9, "learning_rate": 1.2606838632273547e-06, "loss": 0.0004, "step": 24212 }, { "epoch": 2.9, "learning_rate": 1.2606118776244753e-06, "loss": 0.0002, "step": 24213 }, { "epoch": 2.9, "learning_rate": 1.2605398920215956e-06, "loss": 0.0001, "step": 24214 }, { "epoch": 2.9, "learning_rate": 1.2604679064187162e-06, "loss": 0.0001, "step": 24215 }, { "epoch": 2.9, "learning_rate": 1.2603959208158368e-06, "loss": 0.0001, "step": 24216 }, { "epoch": 2.9, "learning_rate": 1.2603239352129574e-06, "loss": 0.0002, "step": 24217 }, { "epoch": 2.9, "learning_rate": 1.260251949610078e-06, "loss": 0.0001, "step": 24218 }, { "epoch": 2.9, "learning_rate": 1.2601799640071987e-06, "loss": 0.0001, "step": 24219 }, { "epoch": 2.9, "learning_rate": 1.2601079784043193e-06, "loss": 0.0001, "step": 24220 }, { "epoch": 2.9, "learning_rate": 1.2600359928014398e-06, "loss": 0.0001, "step": 24221 }, { "epoch": 2.9, "learning_rate": 1.2599640071985604e-06, "loss": 0.0002, "step": 24222 }, { "epoch": 2.9, "learning_rate": 1.259892021595681e-06, "loss": 0.0002, "step": 24223 }, { "epoch": 2.9, "learning_rate": 1.2598200359928015e-06, "loss": 0.0003, "step": 24224 }, { "epoch": 2.9, "learning_rate": 1.2597480503899219e-06, "loss": 0.0003, "step": 24225 }, { "epoch": 2.9, "learning_rate": 1.2596760647870427e-06, "loss": 0.0001, "step": 24226 }, { "epoch": 2.9, "learning_rate": 1.2596040791841632e-06, "loss": 0.0002, "step": 24227 }, { "epoch": 2.9, "learning_rate": 1.2595320935812838e-06, "loss": 0.0002, "step": 24228 }, { "epoch": 2.9, "learning_rate": 1.2594601079784044e-06, "loss": 0.0003, "step": 24229 }, { "epoch": 2.9, "learning_rate": 1.259388122375525e-06, "loss": 0.0002, "step": 24230 }, { "epoch": 2.9, "learning_rate": 1.2593161367726455e-06, "loss": 0.0001, "step": 24231 }, { "epoch": 2.9, "learning_rate": 1.259244151169766e-06, "loss": 0.0002, "step": 24232 }, { "epoch": 2.9, "learning_rate": 1.2591721655668866e-06, "loss": 0.0001, "step": 24233 }, { "epoch": 2.9, "learning_rate": 1.2591001799640072e-06, "loss": 0.0001, "step": 24234 }, { "epoch": 2.9, "learning_rate": 1.259028194361128e-06, "loss": 0.8717, "step": 24235 }, { "epoch": 2.9, "learning_rate": 1.2589562087582485e-06, "loss": 0.0001, "step": 24236 }, { "epoch": 2.9, "learning_rate": 1.258884223155369e-06, "loss": 0.0001, "step": 24237 }, { "epoch": 2.9, "learning_rate": 1.2588122375524895e-06, "loss": 0.0001, "step": 24238 }, { "epoch": 2.9, "learning_rate": 1.25874025194961e-06, "loss": 0.0001, "step": 24239 }, { "epoch": 2.9, "learning_rate": 1.2586682663467306e-06, "loss": 0.0001, "step": 24240 }, { "epoch": 2.9, "learning_rate": 1.2585962807438512e-06, "loss": 0.0003, "step": 24241 }, { "epoch": 2.9, "learning_rate": 1.2585242951409717e-06, "loss": 0.0012, "step": 24242 }, { "epoch": 2.91, "learning_rate": 1.2584523095380925e-06, "loss": 0.0001, "step": 24243 }, { "epoch": 2.91, "learning_rate": 1.258380323935213e-06, "loss": 0.0001, "step": 24244 }, { "epoch": 2.91, "learning_rate": 1.2583083383323336e-06, "loss": 0.0001, "step": 24245 }, { "epoch": 2.91, "learning_rate": 1.2582363527294542e-06, "loss": 0.0003, "step": 24246 }, { "epoch": 2.91, "learning_rate": 1.2581643671265748e-06, "loss": 0.0001, "step": 24247 }, { "epoch": 2.91, "learning_rate": 1.2580923815236951e-06, "loss": 0.0003, "step": 24248 }, { "epoch": 2.91, "learning_rate": 1.2580203959208157e-06, "loss": 0.0005, "step": 24249 }, { "epoch": 2.91, "learning_rate": 1.2579484103179365e-06, "loss": 0.0002, "step": 24250 }, { "epoch": 2.91, "learning_rate": 1.257876424715057e-06, "loss": 0.0002, "step": 24251 }, { "epoch": 2.91, "learning_rate": 1.2578044391121776e-06, "loss": 0.0002, "step": 24252 }, { "epoch": 2.91, "learning_rate": 1.2577324535092982e-06, "loss": 0.0001, "step": 24253 }, { "epoch": 2.91, "learning_rate": 1.2576604679064187e-06, "loss": 0.0001, "step": 24254 }, { "epoch": 2.91, "learning_rate": 1.2575884823035393e-06, "loss": 0.0001, "step": 24255 }, { "epoch": 2.91, "learning_rate": 1.2575164967006599e-06, "loss": 0.0001, "step": 24256 }, { "epoch": 2.91, "learning_rate": 1.2574445110977804e-06, "loss": 0.0001, "step": 24257 }, { "epoch": 2.91, "learning_rate": 1.257372525494901e-06, "loss": 0.0001, "step": 24258 }, { "epoch": 2.91, "learning_rate": 1.2573005398920218e-06, "loss": 0.0003, "step": 24259 }, { "epoch": 2.91, "learning_rate": 1.2572285542891421e-06, "loss": 0.0001, "step": 24260 }, { "epoch": 2.91, "learning_rate": 1.2571565686862627e-06, "loss": 0.0001, "step": 24261 }, { "epoch": 2.91, "learning_rate": 1.2570845830833833e-06, "loss": 0.2248, "step": 24262 }, { "epoch": 2.91, "learning_rate": 1.2570125974805038e-06, "loss": 0.0002, "step": 24263 }, { "epoch": 2.91, "learning_rate": 1.2569406118776244e-06, "loss": 0.5744, "step": 24264 }, { "epoch": 2.91, "learning_rate": 1.256868626274745e-06, "loss": 0.0001, "step": 24265 }, { "epoch": 2.91, "learning_rate": 1.2567966406718658e-06, "loss": 0.0001, "step": 24266 }, { "epoch": 2.91, "learning_rate": 1.2567246550689863e-06, "loss": 0.0021, "step": 24267 }, { "epoch": 2.91, "learning_rate": 1.2566526694661069e-06, "loss": 0.0002, "step": 24268 }, { "epoch": 2.91, "learning_rate": 1.2565806838632275e-06, "loss": 0.0042, "step": 24269 }, { "epoch": 2.91, "learning_rate": 1.256508698260348e-06, "loss": 0.001, "step": 24270 }, { "epoch": 2.91, "learning_rate": 1.2564367126574684e-06, "loss": 0.0001, "step": 24271 }, { "epoch": 2.91, "learning_rate": 1.256364727054589e-06, "loss": 0.0001, "step": 24272 }, { "epoch": 2.91, "learning_rate": 1.2562927414517097e-06, "loss": 0.0003, "step": 24273 }, { "epoch": 2.91, "learning_rate": 1.2562207558488303e-06, "loss": 0.0001, "step": 24274 }, { "epoch": 2.91, "learning_rate": 1.2561487702459509e-06, "loss": 0.0001, "step": 24275 }, { "epoch": 2.91, "learning_rate": 1.2560767846430714e-06, "loss": 0.0003, "step": 24276 }, { "epoch": 2.91, "learning_rate": 1.256004799040192e-06, "loss": 0.0001, "step": 24277 }, { "epoch": 2.91, "learning_rate": 1.2559328134373126e-06, "loss": 0.0003, "step": 24278 }, { "epoch": 2.91, "learning_rate": 1.2558608278344331e-06, "loss": 0.0001, "step": 24279 }, { "epoch": 2.91, "learning_rate": 1.2557888422315537e-06, "loss": 0.0001, "step": 24280 }, { "epoch": 2.91, "learning_rate": 1.2557168566286743e-06, "loss": 0.0001, "step": 24281 }, { "epoch": 2.91, "learning_rate": 1.255644871025795e-06, "loss": 0.0002, "step": 24282 }, { "epoch": 2.91, "learning_rate": 1.2555728854229154e-06, "loss": 0.0001, "step": 24283 }, { "epoch": 2.91, "learning_rate": 1.255500899820036e-06, "loss": 0.0001, "step": 24284 }, { "epoch": 2.91, "learning_rate": 1.2554289142171565e-06, "loss": 0.0001, "step": 24285 }, { "epoch": 2.91, "learning_rate": 1.255356928614277e-06, "loss": 0.0001, "step": 24286 }, { "epoch": 2.91, "learning_rate": 1.2552849430113977e-06, "loss": 0.0001, "step": 24287 }, { "epoch": 2.91, "learning_rate": 1.2552129574085182e-06, "loss": 0.0001, "step": 24288 }, { "epoch": 2.91, "learning_rate": 1.255140971805639e-06, "loss": 0.0001, "step": 24289 }, { "epoch": 2.91, "learning_rate": 1.2550689862027596e-06, "loss": 0.0001, "step": 24290 }, { "epoch": 2.91, "learning_rate": 1.2549970005998801e-06, "loss": 0.0002, "step": 24291 }, { "epoch": 2.91, "learning_rate": 1.2549250149970007e-06, "loss": 0.0001, "step": 24292 }, { "epoch": 2.91, "learning_rate": 1.2548530293941213e-06, "loss": 0.0002, "step": 24293 }, { "epoch": 2.91, "learning_rate": 1.2547810437912418e-06, "loss": 0.0003, "step": 24294 }, { "epoch": 2.91, "learning_rate": 1.2547090581883622e-06, "loss": 0.0006, "step": 24295 }, { "epoch": 2.91, "learning_rate": 1.254637072585483e-06, "loss": 0.0001, "step": 24296 }, { "epoch": 2.91, "learning_rate": 1.2545650869826035e-06, "loss": 0.0001, "step": 24297 }, { "epoch": 2.91, "learning_rate": 1.2544931013797241e-06, "loss": 0.0001, "step": 24298 }, { "epoch": 2.91, "learning_rate": 1.2544211157768447e-06, "loss": 0.0006, "step": 24299 }, { "epoch": 2.91, "learning_rate": 1.2543491301739652e-06, "loss": 0.0002, "step": 24300 }, { "epoch": 2.91, "learning_rate": 1.2542771445710858e-06, "loss": 0.0001, "step": 24301 }, { "epoch": 2.91, "learning_rate": 1.2542051589682064e-06, "loss": 0.0001, "step": 24302 }, { "epoch": 2.91, "learning_rate": 1.254133173365327e-06, "loss": 0.0002, "step": 24303 }, { "epoch": 2.91, "learning_rate": 1.2540611877624475e-06, "loss": 0.0001, "step": 24304 }, { "epoch": 2.91, "learning_rate": 1.2539892021595683e-06, "loss": 0.0001, "step": 24305 }, { "epoch": 2.91, "learning_rate": 1.2539172165566886e-06, "loss": 0.0001, "step": 24306 }, { "epoch": 2.91, "learning_rate": 1.2538452309538092e-06, "loss": 0.0004, "step": 24307 }, { "epoch": 2.91, "learning_rate": 1.2537732453509298e-06, "loss": 0.0001, "step": 24308 }, { "epoch": 2.91, "learning_rate": 1.2537012597480503e-06, "loss": 0.0001, "step": 24309 }, { "epoch": 2.91, "learning_rate": 1.253629274145171e-06, "loss": 0.0001, "step": 24310 }, { "epoch": 2.91, "learning_rate": 1.2535572885422915e-06, "loss": 0.0001, "step": 24311 }, { "epoch": 2.91, "learning_rate": 1.2534853029394123e-06, "loss": 0.0001, "step": 24312 }, { "epoch": 2.91, "learning_rate": 1.2534133173365328e-06, "loss": 0.0003, "step": 24313 }, { "epoch": 2.91, "learning_rate": 1.2533413317336534e-06, "loss": 0.0002, "step": 24314 }, { "epoch": 2.91, "learning_rate": 1.253269346130774e-06, "loss": 0.0001, "step": 24315 }, { "epoch": 2.91, "learning_rate": 1.2531973605278945e-06, "loss": 0.0001, "step": 24316 }, { "epoch": 2.91, "learning_rate": 1.253125374925015e-06, "loss": 0.0001, "step": 24317 }, { "epoch": 2.91, "learning_rate": 1.2530533893221354e-06, "loss": 0.0001, "step": 24318 }, { "epoch": 2.91, "learning_rate": 1.2529814037192562e-06, "loss": 0.0004, "step": 24319 }, { "epoch": 2.91, "learning_rate": 1.2529094181163768e-06, "loss": 0.0001, "step": 24320 }, { "epoch": 2.91, "learning_rate": 1.2528374325134974e-06, "loss": 0.0001, "step": 24321 }, { "epoch": 2.91, "learning_rate": 1.252765446910618e-06, "loss": 0.0001, "step": 24322 }, { "epoch": 2.91, "learning_rate": 1.2526934613077385e-06, "loss": 0.0001, "step": 24323 }, { "epoch": 2.91, "learning_rate": 1.252621475704859e-06, "loss": 0.0001, "step": 24324 }, { "epoch": 2.91, "learning_rate": 1.2525494901019796e-06, "loss": 0.0001, "step": 24325 }, { "epoch": 2.92, "learning_rate": 1.2524775044991002e-06, "loss": 0.0001, "step": 24326 }, { "epoch": 2.92, "learning_rate": 1.2524055188962208e-06, "loss": 0.0002, "step": 24327 }, { "epoch": 2.92, "learning_rate": 1.2523335332933415e-06, "loss": 0.0001, "step": 24328 }, { "epoch": 2.92, "learning_rate": 1.2522615476904619e-06, "loss": 0.0001, "step": 24329 }, { "epoch": 2.92, "learning_rate": 1.2521895620875825e-06, "loss": 0.0001, "step": 24330 }, { "epoch": 2.92, "learning_rate": 1.252117576484703e-06, "loss": 0.0001, "step": 24331 }, { "epoch": 2.92, "learning_rate": 1.2520455908818236e-06, "loss": 0.0001, "step": 24332 }, { "epoch": 2.92, "learning_rate": 1.2519736052789442e-06, "loss": 0.0001, "step": 24333 }, { "epoch": 2.92, "learning_rate": 1.2519016196760647e-06, "loss": 0.0001, "step": 24334 }, { "epoch": 2.92, "learning_rate": 1.2518296340731855e-06, "loss": 0.0001, "step": 24335 }, { "epoch": 2.92, "learning_rate": 1.251757648470306e-06, "loss": 0.0001, "step": 24336 }, { "epoch": 2.92, "learning_rate": 1.2516856628674266e-06, "loss": 0.0001, "step": 24337 }, { "epoch": 2.92, "learning_rate": 1.2516136772645472e-06, "loss": 0.0003, "step": 24338 }, { "epoch": 2.92, "learning_rate": 1.2515416916616678e-06, "loss": 0.0001, "step": 24339 }, { "epoch": 2.92, "learning_rate": 1.2514697060587883e-06, "loss": 0.0, "step": 24340 }, { "epoch": 2.92, "learning_rate": 1.2513977204559087e-06, "loss": 0.0002, "step": 24341 }, { "epoch": 2.92, "learning_rate": 1.2513257348530295e-06, "loss": 1.0398, "step": 24342 }, { "epoch": 2.92, "learning_rate": 1.25125374925015e-06, "loss": 0.0001, "step": 24343 }, { "epoch": 2.92, "learning_rate": 1.2511817636472706e-06, "loss": 0.0001, "step": 24344 }, { "epoch": 2.92, "learning_rate": 1.2511097780443912e-06, "loss": 0.0001, "step": 24345 }, { "epoch": 2.92, "learning_rate": 1.2510377924415117e-06, "loss": 0.0001, "step": 24346 }, { "epoch": 2.92, "learning_rate": 1.2509658068386323e-06, "loss": 0.0002, "step": 24347 }, { "epoch": 2.92, "learning_rate": 1.2508938212357529e-06, "loss": 0.0001, "step": 24348 }, { "epoch": 2.92, "learning_rate": 1.2508218356328734e-06, "loss": 0.0001, "step": 24349 }, { "epoch": 2.92, "learning_rate": 1.250749850029994e-06, "loss": 0.0001, "step": 24350 }, { "epoch": 2.92, "learning_rate": 1.2506778644271148e-06, "loss": 0.0008, "step": 24351 }, { "epoch": 2.92, "learning_rate": 1.2506058788242351e-06, "loss": 0.0001, "step": 24352 }, { "epoch": 2.92, "learning_rate": 1.2505338932213557e-06, "loss": 0.0002, "step": 24353 }, { "epoch": 2.92, "learning_rate": 1.2504619076184763e-06, "loss": 0.0003, "step": 24354 }, { "epoch": 2.92, "learning_rate": 1.2503899220155968e-06, "loss": 0.0002, "step": 24355 }, { "epoch": 2.92, "learning_rate": 1.2503179364127174e-06, "loss": 0.0001, "step": 24356 }, { "epoch": 2.92, "learning_rate": 1.250245950809838e-06, "loss": 0.0001, "step": 24357 }, { "epoch": 2.92, "learning_rate": 1.2501739652069588e-06, "loss": 0.0001, "step": 24358 }, { "epoch": 2.92, "learning_rate": 1.2501019796040793e-06, "loss": 0.0002, "step": 24359 }, { "epoch": 2.92, "learning_rate": 1.2500299940011999e-06, "loss": 0.0002, "step": 24360 }, { "epoch": 2.92, "learning_rate": 1.2499580083983205e-06, "loss": 0.2769, "step": 24361 }, { "epoch": 2.92, "learning_rate": 1.249886022795441e-06, "loss": 0.0003, "step": 24362 }, { "epoch": 2.92, "learning_rate": 1.2498140371925616e-06, "loss": 0.0001, "step": 24363 }, { "epoch": 2.92, "learning_rate": 1.249742051589682e-06, "loss": 0.2916, "step": 24364 }, { "epoch": 2.92, "learning_rate": 1.2496700659868027e-06, "loss": 0.0001, "step": 24365 }, { "epoch": 2.92, "learning_rate": 1.2495980803839233e-06, "loss": 0.0001, "step": 24366 }, { "epoch": 2.92, "learning_rate": 1.2495260947810439e-06, "loss": 0.0001, "step": 24367 }, { "epoch": 2.92, "learning_rate": 1.2494541091781644e-06, "loss": 0.0001, "step": 24368 }, { "epoch": 2.92, "learning_rate": 1.249382123575285e-06, "loss": 0.0001, "step": 24369 }, { "epoch": 2.92, "learning_rate": 1.2493101379724056e-06, "loss": 0.0002, "step": 24370 }, { "epoch": 2.92, "learning_rate": 1.2492381523695261e-06, "loss": 0.0001, "step": 24371 }, { "epoch": 2.92, "learning_rate": 1.2491661667666467e-06, "loss": 0.0001, "step": 24372 }, { "epoch": 2.92, "learning_rate": 1.2490941811637673e-06, "loss": 0.0001, "step": 24373 }, { "epoch": 2.92, "learning_rate": 1.249022195560888e-06, "loss": 0.0001, "step": 24374 }, { "epoch": 2.92, "learning_rate": 1.2489502099580084e-06, "loss": 0.0016, "step": 24375 }, { "epoch": 2.92, "learning_rate": 1.248878224355129e-06, "loss": 0.0001, "step": 24376 }, { "epoch": 2.92, "learning_rate": 1.2488062387522495e-06, "loss": 0.0008, "step": 24377 }, { "epoch": 2.92, "learning_rate": 1.24873425314937e-06, "loss": 0.0003, "step": 24378 }, { "epoch": 2.92, "learning_rate": 1.2486622675464907e-06, "loss": 0.0001, "step": 24379 }, { "epoch": 2.92, "learning_rate": 1.2485902819436112e-06, "loss": 0.0067, "step": 24380 }, { "epoch": 2.92, "learning_rate": 1.2485182963407318e-06, "loss": 0.0027, "step": 24381 }, { "epoch": 2.92, "learning_rate": 1.2484463107378526e-06, "loss": 0.0002, "step": 24382 }, { "epoch": 2.92, "learning_rate": 1.2483743251349731e-06, "loss": 0.0002, "step": 24383 }, { "epoch": 2.92, "learning_rate": 1.2483023395320937e-06, "loss": 0.0002, "step": 24384 }, { "epoch": 2.92, "learning_rate": 1.2482303539292143e-06, "loss": 0.0001, "step": 24385 }, { "epoch": 2.92, "learning_rate": 1.2481583683263348e-06, "loss": 0.0002, "step": 24386 }, { "epoch": 2.92, "learning_rate": 1.2480863827234552e-06, "loss": 0.0002, "step": 24387 }, { "epoch": 2.92, "learning_rate": 1.2480143971205758e-06, "loss": 0.0002, "step": 24388 }, { "epoch": 2.92, "learning_rate": 1.2479424115176965e-06, "loss": 0.0001, "step": 24389 }, { "epoch": 2.92, "learning_rate": 1.247870425914817e-06, "loss": 0.0002, "step": 24390 }, { "epoch": 2.92, "learning_rate": 1.2477984403119377e-06, "loss": 0.0001, "step": 24391 }, { "epoch": 2.92, "learning_rate": 1.2477264547090582e-06, "loss": 0.0002, "step": 24392 }, { "epoch": 2.92, "learning_rate": 1.2476544691061788e-06, "loss": 0.0001, "step": 24393 }, { "epoch": 2.92, "learning_rate": 1.2475824835032994e-06, "loss": 1.0642, "step": 24394 }, { "epoch": 2.92, "learning_rate": 1.24751049790042e-06, "loss": 0.0024, "step": 24395 }, { "epoch": 2.92, "learning_rate": 1.2474385122975405e-06, "loss": 0.0002, "step": 24396 }, { "epoch": 2.92, "learning_rate": 1.247366526694661e-06, "loss": 0.0014, "step": 24397 }, { "epoch": 2.92, "learning_rate": 1.2472945410917816e-06, "loss": 0.0003, "step": 24398 }, { "epoch": 2.92, "learning_rate": 1.2472225554889022e-06, "loss": 0.0001, "step": 24399 }, { "epoch": 2.92, "learning_rate": 1.2471505698860228e-06, "loss": 0.0002, "step": 24400 }, { "epoch": 2.92, "learning_rate": 1.2470785842831433e-06, "loss": 0.0002, "step": 24401 }, { "epoch": 2.92, "learning_rate": 1.247006598680264e-06, "loss": 0.0001, "step": 24402 }, { "epoch": 2.92, "learning_rate": 1.2469346130773845e-06, "loss": 0.0003, "step": 24403 }, { "epoch": 2.92, "learning_rate": 1.246862627474505e-06, "loss": 0.0005, "step": 24404 }, { "epoch": 2.92, "learning_rate": 1.2467906418716258e-06, "loss": 0.0001, "step": 24405 }, { "epoch": 2.92, "learning_rate": 1.2467186562687464e-06, "loss": 0.7895, "step": 24406 }, { "epoch": 2.92, "learning_rate": 1.246646670665867e-06, "loss": 0.0002, "step": 24407 }, { "epoch": 2.92, "learning_rate": 1.2465746850629875e-06, "loss": 0.0002, "step": 24408 }, { "epoch": 2.92, "learning_rate": 1.246502699460108e-06, "loss": 0.0002, "step": 24409 }, { "epoch": 2.93, "learning_rate": 1.2464307138572284e-06, "loss": 0.0003, "step": 24410 }, { "epoch": 2.93, "learning_rate": 1.246358728254349e-06, "loss": 0.0002, "step": 24411 }, { "epoch": 2.93, "learning_rate": 1.2462867426514698e-06, "loss": 0.0001, "step": 24412 }, { "epoch": 2.93, "learning_rate": 1.2462147570485903e-06, "loss": 0.1753, "step": 24413 }, { "epoch": 2.93, "learning_rate": 1.246142771445711e-06, "loss": 0.0001, "step": 24414 }, { "epoch": 2.93, "learning_rate": 1.2460707858428315e-06, "loss": 0.0001, "step": 24415 }, { "epoch": 2.93, "learning_rate": 1.245998800239952e-06, "loss": 0.0001, "step": 24416 }, { "epoch": 2.93, "learning_rate": 1.2459268146370726e-06, "loss": 0.0, "step": 24417 }, { "epoch": 2.93, "learning_rate": 1.2458548290341932e-06, "loss": 0.0003, "step": 24418 }, { "epoch": 2.93, "learning_rate": 1.2457828434313137e-06, "loss": 0.0001, "step": 24419 }, { "epoch": 2.93, "learning_rate": 1.2457108578284343e-06, "loss": 0.0001, "step": 24420 }, { "epoch": 2.93, "learning_rate": 1.245638872225555e-06, "loss": 0.0001, "step": 24421 }, { "epoch": 2.93, "learning_rate": 1.2455668866226754e-06, "loss": 0.0003, "step": 24422 }, { "epoch": 2.93, "learning_rate": 1.245494901019796e-06, "loss": 0.0004, "step": 24423 }, { "epoch": 2.93, "learning_rate": 1.2454229154169166e-06, "loss": 0.0002, "step": 24424 }, { "epoch": 2.93, "learning_rate": 1.2453509298140371e-06, "loss": 0.0002, "step": 24425 }, { "epoch": 2.93, "learning_rate": 1.2452789442111577e-06, "loss": 0.0023, "step": 24426 }, { "epoch": 2.93, "learning_rate": 1.2452069586082783e-06, "loss": 0.0001, "step": 24427 }, { "epoch": 2.93, "learning_rate": 1.245134973005399e-06, "loss": 0.0001, "step": 24428 }, { "epoch": 2.93, "learning_rate": 1.2450629874025196e-06, "loss": 0.0001, "step": 24429 }, { "epoch": 2.93, "learning_rate": 1.2449910017996402e-06, "loss": 0.0002, "step": 24430 }, { "epoch": 2.93, "learning_rate": 1.2449190161967608e-06, "loss": 0.0008, "step": 24431 }, { "epoch": 2.93, "learning_rate": 1.2448470305938813e-06, "loss": 0.0001, "step": 24432 }, { "epoch": 2.93, "learning_rate": 1.2447750449910017e-06, "loss": 0.0001, "step": 24433 }, { "epoch": 2.93, "learning_rate": 1.2447030593881222e-06, "loss": 0.0003, "step": 24434 }, { "epoch": 2.93, "learning_rate": 1.244631073785243e-06, "loss": 0.0002, "step": 24435 }, { "epoch": 2.93, "learning_rate": 1.2445590881823636e-06, "loss": 0.0001, "step": 24436 }, { "epoch": 2.93, "learning_rate": 1.2444871025794842e-06, "loss": 0.0001, "step": 24437 }, { "epoch": 2.93, "learning_rate": 1.2444151169766047e-06, "loss": 0.0002, "step": 24438 }, { "epoch": 2.93, "learning_rate": 1.2443431313737253e-06, "loss": 0.0002, "step": 24439 }, { "epoch": 2.93, "learning_rate": 1.2442711457708459e-06, "loss": 0.0005, "step": 24440 }, { "epoch": 2.93, "learning_rate": 1.2441991601679664e-06, "loss": 0.0001, "step": 24441 }, { "epoch": 2.93, "learning_rate": 1.244127174565087e-06, "loss": 0.0001, "step": 24442 }, { "epoch": 2.93, "learning_rate": 1.2440551889622076e-06, "loss": 0.0001, "step": 24443 }, { "epoch": 2.93, "learning_rate": 1.2439832033593283e-06, "loss": 0.0001, "step": 24444 }, { "epoch": 2.93, "learning_rate": 1.2439112177564487e-06, "loss": 0.0001, "step": 24445 }, { "epoch": 2.93, "learning_rate": 1.2438392321535693e-06, "loss": 0.0, "step": 24446 }, { "epoch": 2.93, "learning_rate": 1.2437672465506898e-06, "loss": 0.0003, "step": 24447 }, { "epoch": 2.93, "learning_rate": 1.2436952609478104e-06, "loss": 0.0001, "step": 24448 }, { "epoch": 2.93, "learning_rate": 1.243623275344931e-06, "loss": 0.0001, "step": 24449 }, { "epoch": 2.93, "learning_rate": 1.2435512897420515e-06, "loss": 0.0001, "step": 24450 }, { "epoch": 2.93, "learning_rate": 1.2434793041391723e-06, "loss": 0.0003, "step": 24451 }, { "epoch": 2.93, "learning_rate": 1.2434073185362929e-06, "loss": 0.0004, "step": 24452 }, { "epoch": 2.93, "learning_rate": 1.2433353329334134e-06, "loss": 0.0001, "step": 24453 }, { "epoch": 2.93, "learning_rate": 1.243263347330534e-06, "loss": 0.0004, "step": 24454 }, { "epoch": 2.93, "learning_rate": 1.2431913617276546e-06, "loss": 0.5579, "step": 24455 }, { "epoch": 2.93, "learning_rate": 1.243119376124775e-06, "loss": 0.0002, "step": 24456 }, { "epoch": 2.93, "learning_rate": 1.2430473905218955e-06, "loss": 0.0001, "step": 24457 }, { "epoch": 2.93, "learning_rate": 1.2429754049190163e-06, "loss": 0.0001, "step": 24458 }, { "epoch": 2.93, "learning_rate": 1.2429034193161368e-06, "loss": 0.0002, "step": 24459 }, { "epoch": 2.93, "learning_rate": 1.2428314337132574e-06, "loss": 0.0002, "step": 24460 }, { "epoch": 2.93, "learning_rate": 1.242759448110378e-06, "loss": 0.0003, "step": 24461 }, { "epoch": 2.93, "learning_rate": 1.2426874625074985e-06, "loss": 0.0004, "step": 24462 }, { "epoch": 2.93, "learning_rate": 1.2426154769046191e-06, "loss": 0.7301, "step": 24463 }, { "epoch": 2.93, "learning_rate": 1.2425434913017397e-06, "loss": 0.0001, "step": 24464 }, { "epoch": 2.93, "learning_rate": 1.2424715056988602e-06, "loss": 0.0002, "step": 24465 }, { "epoch": 2.93, "learning_rate": 1.2423995200959808e-06, "loss": 0.0001, "step": 24466 }, { "epoch": 2.93, "learning_rate": 1.2423275344931016e-06, "loss": 0.0012, "step": 24467 }, { "epoch": 2.93, "learning_rate": 1.242255548890222e-06, "loss": 0.0001, "step": 24468 }, { "epoch": 2.93, "learning_rate": 1.2421835632873425e-06, "loss": 0.0002, "step": 24469 }, { "epoch": 2.93, "learning_rate": 1.242111577684463e-06, "loss": 0.7634, "step": 24470 }, { "epoch": 2.93, "learning_rate": 1.2420395920815836e-06, "loss": 0.0001, "step": 24471 }, { "epoch": 2.93, "learning_rate": 1.2419676064787042e-06, "loss": 0.0003, "step": 24472 }, { "epoch": 2.93, "learning_rate": 1.2418956208758248e-06, "loss": 0.0001, "step": 24473 }, { "epoch": 2.93, "learning_rate": 1.2418236352729456e-06, "loss": 0.0001, "step": 24474 }, { "epoch": 2.93, "learning_rate": 1.2417516496700661e-06, "loss": 0.0001, "step": 24475 }, { "epoch": 2.93, "learning_rate": 1.2416796640671867e-06, "loss": 0.0121, "step": 24476 }, { "epoch": 2.93, "learning_rate": 1.2416076784643073e-06, "loss": 0.0001, "step": 24477 }, { "epoch": 2.93, "learning_rate": 1.2415356928614278e-06, "loss": 0.3448, "step": 24478 }, { "epoch": 2.93, "learning_rate": 1.2414637072585482e-06, "loss": 0.0002, "step": 24479 }, { "epoch": 2.93, "learning_rate": 1.2413917216556687e-06, "loss": 0.0001, "step": 24480 }, { "epoch": 2.93, "learning_rate": 1.2413197360527895e-06, "loss": 0.0005, "step": 24481 }, { "epoch": 2.93, "learning_rate": 1.24124775044991e-06, "loss": 0.0002, "step": 24482 }, { "epoch": 2.93, "learning_rate": 1.2411757648470307e-06, "loss": 0.0003, "step": 24483 }, { "epoch": 2.93, "learning_rate": 1.2411037792441512e-06, "loss": 0.0002, "step": 24484 }, { "epoch": 2.93, "learning_rate": 1.2410317936412718e-06, "loss": 0.0001, "step": 24485 }, { "epoch": 2.93, "learning_rate": 1.2409598080383924e-06, "loss": 0.0002, "step": 24486 }, { "epoch": 2.93, "learning_rate": 1.240887822435513e-06, "loss": 0.0001, "step": 24487 }, { "epoch": 2.93, "learning_rate": 1.2408158368326335e-06, "loss": 0.0002, "step": 24488 }, { "epoch": 2.93, "learning_rate": 1.240743851229754e-06, "loss": 0.0002, "step": 24489 }, { "epoch": 2.93, "learning_rate": 1.2406718656268748e-06, "loss": 0.0001, "step": 24490 }, { "epoch": 2.93, "learning_rate": 1.2405998800239952e-06, "loss": 0.0002, "step": 24491 }, { "epoch": 2.93, "learning_rate": 1.2405278944211158e-06, "loss": 0.0001, "step": 24492 }, { "epoch": 2.94, "learning_rate": 1.2404559088182363e-06, "loss": 0.0001, "step": 24493 }, { "epoch": 2.94, "learning_rate": 1.2403839232153569e-06, "loss": 0.0, "step": 24494 }, { "epoch": 2.94, "learning_rate": 1.2403119376124775e-06, "loss": 0.0001, "step": 24495 }, { "epoch": 2.94, "learning_rate": 1.240239952009598e-06, "loss": 0.0006, "step": 24496 }, { "epoch": 2.94, "learning_rate": 1.2401679664067188e-06, "loss": 0.0005, "step": 24497 }, { "epoch": 2.94, "learning_rate": 1.2400959808038394e-06, "loss": 0.0003, "step": 24498 }, { "epoch": 2.94, "learning_rate": 1.24002399520096e-06, "loss": 0.0001, "step": 24499 }, { "epoch": 2.94, "learning_rate": 1.2399520095980805e-06, "loss": 0.0002, "step": 24500 }, { "epoch": 2.94, "learning_rate": 1.239880023995201e-06, "loss": 0.0001, "step": 24501 }, { "epoch": 2.94, "learning_rate": 1.2398080383923214e-06, "loss": 0.0002, "step": 24502 }, { "epoch": 2.94, "learning_rate": 1.239736052789442e-06, "loss": 0.0001, "step": 24503 }, { "epoch": 2.94, "learning_rate": 1.2396640671865626e-06, "loss": 0.0002, "step": 24504 }, { "epoch": 2.94, "learning_rate": 1.2395920815836833e-06, "loss": 0.0002, "step": 24505 }, { "epoch": 2.94, "learning_rate": 1.239520095980804e-06, "loss": 0.0001, "step": 24506 }, { "epoch": 2.94, "learning_rate": 1.2394481103779245e-06, "loss": 0.0002, "step": 24507 }, { "epoch": 2.94, "learning_rate": 1.239376124775045e-06, "loss": 0.0001, "step": 24508 }, { "epoch": 2.94, "learning_rate": 1.2393041391721656e-06, "loss": 0.0001, "step": 24509 }, { "epoch": 2.94, "learning_rate": 1.2392321535692862e-06, "loss": 0.0002, "step": 24510 }, { "epoch": 2.94, "learning_rate": 1.2391601679664067e-06, "loss": 0.0001, "step": 24511 }, { "epoch": 2.94, "learning_rate": 1.2390881823635273e-06, "loss": 0.0001, "step": 24512 }, { "epoch": 2.94, "learning_rate": 1.2390161967606479e-06, "loss": 0.1443, "step": 24513 }, { "epoch": 2.94, "learning_rate": 1.2389442111577684e-06, "loss": 0.0001, "step": 24514 }, { "epoch": 2.94, "learning_rate": 1.238872225554889e-06, "loss": 0.0001, "step": 24515 }, { "epoch": 2.94, "learning_rate": 1.2388002399520096e-06, "loss": 0.0001, "step": 24516 }, { "epoch": 2.94, "learning_rate": 1.2387282543491301e-06, "loss": 0.0049, "step": 24517 }, { "epoch": 2.94, "learning_rate": 1.2386562687462507e-06, "loss": 0.0002, "step": 24518 }, { "epoch": 2.94, "learning_rate": 1.2385842831433713e-06, "loss": 0.0002, "step": 24519 }, { "epoch": 2.94, "learning_rate": 1.2385122975404918e-06, "loss": 0.0001, "step": 24520 }, { "epoch": 2.94, "learning_rate": 1.2384403119376126e-06, "loss": 0.0002, "step": 24521 }, { "epoch": 2.94, "learning_rate": 1.2383683263347332e-06, "loss": 0.0001, "step": 24522 }, { "epoch": 2.94, "learning_rate": 1.2382963407318538e-06, "loss": 0.2003, "step": 24523 }, { "epoch": 2.94, "learning_rate": 1.2382243551289743e-06, "loss": 0.0003, "step": 24524 }, { "epoch": 2.94, "learning_rate": 1.2381523695260947e-06, "loss": 0.0002, "step": 24525 }, { "epoch": 2.94, "learning_rate": 1.2380803839232152e-06, "loss": 0.0001, "step": 24526 }, { "epoch": 2.94, "learning_rate": 1.2380083983203358e-06, "loss": 0.0009, "step": 24527 }, { "epoch": 2.94, "learning_rate": 1.2379364127174566e-06, "loss": 0.0001, "step": 24528 }, { "epoch": 2.94, "learning_rate": 1.2378644271145772e-06, "loss": 0.0003, "step": 24529 }, { "epoch": 2.94, "learning_rate": 1.2377924415116977e-06, "loss": 0.0002, "step": 24530 }, { "epoch": 2.94, "learning_rate": 1.2377204559088183e-06, "loss": 0.0001, "step": 24531 }, { "epoch": 2.94, "learning_rate": 1.2376484703059389e-06, "loss": 0.0002, "step": 24532 }, { "epoch": 2.94, "learning_rate": 1.2375764847030594e-06, "loss": 0.0001, "step": 24533 }, { "epoch": 2.94, "learning_rate": 1.23750449910018e-06, "loss": 0.0002, "step": 24534 }, { "epoch": 2.94, "learning_rate": 1.2374325134973006e-06, "loss": 0.0009, "step": 24535 }, { "epoch": 2.94, "learning_rate": 1.2373605278944211e-06, "loss": 0.0002, "step": 24536 }, { "epoch": 2.94, "learning_rate": 1.2372885422915417e-06, "loss": 0.0001, "step": 24537 }, { "epoch": 2.94, "learning_rate": 1.2372165566886623e-06, "loss": 0.0002, "step": 24538 }, { "epoch": 2.94, "learning_rate": 1.2371445710857828e-06, "loss": 0.4708, "step": 24539 }, { "epoch": 2.94, "learning_rate": 1.2370725854829034e-06, "loss": 0.0001, "step": 24540 }, { "epoch": 2.94, "learning_rate": 1.237000599880024e-06, "loss": 0.0001, "step": 24541 }, { "epoch": 2.94, "learning_rate": 1.2369286142771445e-06, "loss": 0.0003, "step": 24542 }, { "epoch": 2.94, "learning_rate": 1.236856628674265e-06, "loss": 0.0001, "step": 24543 }, { "epoch": 2.94, "learning_rate": 1.2367846430713859e-06, "loss": 0.0001, "step": 24544 }, { "epoch": 2.94, "learning_rate": 1.2367126574685064e-06, "loss": 0.0002, "step": 24545 }, { "epoch": 2.94, "learning_rate": 1.236640671865627e-06, "loss": 0.0001, "step": 24546 }, { "epoch": 2.94, "learning_rate": 1.2365686862627476e-06, "loss": 0.0001, "step": 24547 }, { "epoch": 2.94, "learning_rate": 1.2364967006598681e-06, "loss": 0.0001, "step": 24548 }, { "epoch": 2.94, "learning_rate": 1.2364247150569885e-06, "loss": 1.0544, "step": 24549 }, { "epoch": 2.94, "learning_rate": 1.236352729454109e-06, "loss": 0.0001, "step": 24550 }, { "epoch": 2.94, "learning_rate": 1.2362807438512298e-06, "loss": 0.0001, "step": 24551 }, { "epoch": 2.94, "learning_rate": 1.2362087582483504e-06, "loss": 0.0001, "step": 24552 }, { "epoch": 2.94, "learning_rate": 1.236136772645471e-06, "loss": 0.0005, "step": 24553 }, { "epoch": 2.94, "learning_rate": 1.2360647870425915e-06, "loss": 0.0003, "step": 24554 }, { "epoch": 2.94, "learning_rate": 1.235992801439712e-06, "loss": 0.0002, "step": 24555 }, { "epoch": 2.94, "learning_rate": 1.2359208158368327e-06, "loss": 0.0001, "step": 24556 }, { "epoch": 2.94, "learning_rate": 1.2358488302339532e-06, "loss": 0.0005, "step": 24557 }, { "epoch": 2.94, "learning_rate": 1.2357768446310738e-06, "loss": 0.0005, "step": 24558 }, { "epoch": 2.94, "learning_rate": 1.2357048590281944e-06, "loss": 0.0001, "step": 24559 }, { "epoch": 2.94, "learning_rate": 1.235632873425315e-06, "loss": 0.0001, "step": 24560 }, { "epoch": 2.94, "learning_rate": 1.2355608878224355e-06, "loss": 0.0478, "step": 24561 }, { "epoch": 2.94, "learning_rate": 1.235488902219556e-06, "loss": 0.0002, "step": 24562 }, { "epoch": 2.94, "learning_rate": 1.2354169166166766e-06, "loss": 0.0001, "step": 24563 }, { "epoch": 2.94, "learning_rate": 1.2353449310137972e-06, "loss": 0.0002, "step": 24564 }, { "epoch": 2.94, "learning_rate": 1.2352729454109178e-06, "loss": 0.0001, "step": 24565 }, { "epoch": 2.94, "learning_rate": 1.2352009598080383e-06, "loss": 0.0004, "step": 24566 }, { "epoch": 2.94, "learning_rate": 1.2351289742051591e-06, "loss": 0.0001, "step": 24567 }, { "epoch": 2.94, "learning_rate": 1.2350569886022797e-06, "loss": 0.0001, "step": 24568 }, { "epoch": 2.94, "learning_rate": 1.2349850029994002e-06, "loss": 0.0002, "step": 24569 }, { "epoch": 2.94, "learning_rate": 1.2349130173965208e-06, "loss": 0.0001, "step": 24570 }, { "epoch": 2.94, "learning_rate": 1.2348410317936414e-06, "loss": 0.0002, "step": 24571 }, { "epoch": 2.94, "learning_rate": 1.2347690461907617e-06, "loss": 0.0001, "step": 24572 }, { "epoch": 2.94, "learning_rate": 1.2346970605878823e-06, "loss": 0.0001, "step": 24573 }, { "epoch": 2.94, "learning_rate": 1.234625074985003e-06, "loss": 0.0002, "step": 24574 }, { "epoch": 2.94, "learning_rate": 1.2345530893821237e-06, "loss": 0.0002, "step": 24575 }, { "epoch": 2.94, "learning_rate": 1.2344811037792442e-06, "loss": 0.0002, "step": 24576 }, { "epoch": 2.95, "learning_rate": 1.2344091181763648e-06, "loss": 0.0001, "step": 24577 }, { "epoch": 2.95, "learning_rate": 1.2343371325734854e-06, "loss": 0.0009, "step": 24578 }, { "epoch": 2.95, "learning_rate": 1.234265146970606e-06, "loss": 0.0003, "step": 24579 }, { "epoch": 2.95, "learning_rate": 1.2341931613677265e-06, "loss": 0.0002, "step": 24580 }, { "epoch": 2.95, "learning_rate": 1.234121175764847e-06, "loss": 0.0002, "step": 24581 }, { "epoch": 2.95, "learning_rate": 1.2340491901619676e-06, "loss": 0.0002, "step": 24582 }, { "epoch": 2.95, "learning_rate": 1.2339772045590882e-06, "loss": 0.6036, "step": 24583 }, { "epoch": 2.95, "learning_rate": 1.2339052189562088e-06, "loss": 0.0003, "step": 24584 }, { "epoch": 2.95, "learning_rate": 1.2338332333533293e-06, "loss": 0.0006, "step": 24585 }, { "epoch": 2.95, "learning_rate": 1.2337612477504499e-06, "loss": 0.0002, "step": 24586 }, { "epoch": 2.95, "learning_rate": 1.2336892621475705e-06, "loss": 0.0076, "step": 24587 }, { "epoch": 2.95, "learning_rate": 1.233617276544691e-06, "loss": 0.0001, "step": 24588 }, { "epoch": 2.95, "learning_rate": 1.2335452909418116e-06, "loss": 0.0001, "step": 24589 }, { "epoch": 2.95, "learning_rate": 1.2334733053389324e-06, "loss": 0.0002, "step": 24590 }, { "epoch": 2.95, "learning_rate": 1.233401319736053e-06, "loss": 0.0001, "step": 24591 }, { "epoch": 2.95, "learning_rate": 1.2333293341331735e-06, "loss": 0.0002, "step": 24592 }, { "epoch": 2.95, "learning_rate": 1.233257348530294e-06, "loss": 0.0001, "step": 24593 }, { "epoch": 2.95, "learning_rate": 1.2331853629274146e-06, "loss": 0.0002, "step": 24594 }, { "epoch": 2.95, "learning_rate": 1.233113377324535e-06, "loss": 0.0001, "step": 24595 }, { "epoch": 2.95, "learning_rate": 1.2330413917216556e-06, "loss": 0.0002, "step": 24596 }, { "epoch": 2.95, "learning_rate": 1.2329694061187763e-06, "loss": 0.0001, "step": 24597 }, { "epoch": 2.95, "learning_rate": 1.232897420515897e-06, "loss": 0.0001, "step": 24598 }, { "epoch": 2.95, "learning_rate": 1.2328254349130175e-06, "loss": 0.0004, "step": 24599 }, { "epoch": 2.95, "learning_rate": 1.232753449310138e-06, "loss": 0.0001, "step": 24600 }, { "epoch": 2.95, "learning_rate": 1.2326814637072586e-06, "loss": 0.0004, "step": 24601 }, { "epoch": 2.95, "learning_rate": 1.2326094781043792e-06, "loss": 0.0002, "step": 24602 }, { "epoch": 2.95, "learning_rate": 1.2325374925014997e-06, "loss": 0.0003, "step": 24603 }, { "epoch": 2.95, "learning_rate": 1.2324655068986203e-06, "loss": 0.0002, "step": 24604 }, { "epoch": 2.95, "learning_rate": 1.2323935212957409e-06, "loss": 0.0002, "step": 24605 }, { "epoch": 2.95, "learning_rate": 1.2323215356928614e-06, "loss": 0.0002, "step": 24606 }, { "epoch": 2.95, "learning_rate": 1.232249550089982e-06, "loss": 0.0001, "step": 24607 }, { "epoch": 2.95, "learning_rate": 1.2321775644871026e-06, "loss": 0.0002, "step": 24608 }, { "epoch": 2.95, "learning_rate": 1.2321055788842231e-06, "loss": 0.0002, "step": 24609 }, { "epoch": 2.95, "learning_rate": 1.2320335932813437e-06, "loss": 0.0003, "step": 24610 }, { "epoch": 2.95, "learning_rate": 1.2319616076784643e-06, "loss": 0.0001, "step": 24611 }, { "epoch": 2.95, "learning_rate": 1.2318896220755848e-06, "loss": 0.0001, "step": 24612 }, { "epoch": 2.95, "learning_rate": 1.2318176364727056e-06, "loss": 0.0002, "step": 24613 }, { "epoch": 2.95, "learning_rate": 1.2317456508698262e-06, "loss": 0.0002, "step": 24614 }, { "epoch": 2.95, "learning_rate": 1.2316736652669467e-06, "loss": 0.0001, "step": 24615 }, { "epoch": 2.95, "learning_rate": 1.2316016796640673e-06, "loss": 0.0002, "step": 24616 }, { "epoch": 2.95, "learning_rate": 1.2315296940611879e-06, "loss": 0.0005, "step": 24617 }, { "epoch": 2.95, "learning_rate": 1.2314577084583082e-06, "loss": 0.0001, "step": 24618 }, { "epoch": 2.95, "learning_rate": 1.2313857228554288e-06, "loss": 0.0001, "step": 24619 }, { "epoch": 2.95, "learning_rate": 1.2313137372525496e-06, "loss": 0.0001, "step": 24620 }, { "epoch": 2.95, "learning_rate": 1.2312417516496701e-06, "loss": 0.0001, "step": 24621 }, { "epoch": 2.95, "learning_rate": 1.2311697660467907e-06, "loss": 0.0005, "step": 24622 }, { "epoch": 2.95, "learning_rate": 1.2310977804439113e-06, "loss": 0.0002, "step": 24623 }, { "epoch": 2.95, "learning_rate": 1.2310257948410318e-06, "loss": 0.0002, "step": 24624 }, { "epoch": 2.95, "learning_rate": 1.2309538092381524e-06, "loss": 0.0001, "step": 24625 }, { "epoch": 2.95, "learning_rate": 1.230881823635273e-06, "loss": 0.0001, "step": 24626 }, { "epoch": 2.95, "learning_rate": 1.2308098380323935e-06, "loss": 0.0002, "step": 24627 }, { "epoch": 2.95, "learning_rate": 1.2307378524295141e-06, "loss": 0.0001, "step": 24628 }, { "epoch": 2.95, "learning_rate": 1.2306658668266347e-06, "loss": 0.0001, "step": 24629 }, { "epoch": 2.95, "learning_rate": 1.2305938812237552e-06, "loss": 0.0001, "step": 24630 }, { "epoch": 2.95, "learning_rate": 1.2305218956208758e-06, "loss": 0.0001, "step": 24631 }, { "epoch": 2.95, "learning_rate": 1.2304499100179964e-06, "loss": 0.0001, "step": 24632 }, { "epoch": 2.95, "learning_rate": 1.230377924415117e-06, "loss": 0.0002, "step": 24633 }, { "epoch": 2.95, "learning_rate": 1.2303059388122375e-06, "loss": 0.0001, "step": 24634 }, { "epoch": 2.95, "learning_rate": 1.230233953209358e-06, "loss": 0.0001, "step": 24635 }, { "epoch": 2.95, "learning_rate": 1.2301619676064789e-06, "loss": 0.0001, "step": 24636 }, { "epoch": 2.95, "learning_rate": 1.2300899820035994e-06, "loss": 0.0004, "step": 24637 }, { "epoch": 2.95, "learning_rate": 1.23001799640072e-06, "loss": 0.0001, "step": 24638 }, { "epoch": 2.95, "learning_rate": 1.2299460107978406e-06, "loss": 0.0002, "step": 24639 }, { "epoch": 2.95, "learning_rate": 1.2298740251949611e-06, "loss": 0.0002, "step": 24640 }, { "epoch": 2.95, "learning_rate": 1.2298020395920815e-06, "loss": 0.0083, "step": 24641 }, { "epoch": 2.95, "learning_rate": 1.229730053989202e-06, "loss": 0.0142, "step": 24642 }, { "epoch": 2.95, "learning_rate": 1.2296580683863226e-06, "loss": 0.0002, "step": 24643 }, { "epoch": 2.95, "learning_rate": 1.2295860827834434e-06, "loss": 0.0003, "step": 24644 }, { "epoch": 2.95, "learning_rate": 1.229514097180564e-06, "loss": 0.0001, "step": 24645 }, { "epoch": 2.95, "learning_rate": 1.2294421115776845e-06, "loss": 0.0001, "step": 24646 }, { "epoch": 2.95, "learning_rate": 1.229370125974805e-06, "loss": 0.0001, "step": 24647 }, { "epoch": 2.95, "learning_rate": 1.2292981403719257e-06, "loss": 0.0001, "step": 24648 }, { "epoch": 2.95, "learning_rate": 1.2292261547690462e-06, "loss": 0.0001, "step": 24649 }, { "epoch": 2.95, "learning_rate": 1.2291541691661668e-06, "loss": 0.0004, "step": 24650 }, { "epoch": 2.95, "learning_rate": 1.2290821835632874e-06, "loss": 0.066, "step": 24651 }, { "epoch": 2.95, "learning_rate": 1.229010197960408e-06, "loss": 0.0001, "step": 24652 }, { "epoch": 2.95, "learning_rate": 1.2289382123575285e-06, "loss": 0.0001, "step": 24653 }, { "epoch": 2.95, "learning_rate": 1.228866226754649e-06, "loss": 0.0001, "step": 24654 }, { "epoch": 2.95, "learning_rate": 1.2287942411517696e-06, "loss": 0.0001, "step": 24655 }, { "epoch": 2.95, "learning_rate": 1.2287222555488902e-06, "loss": 0.0003, "step": 24656 }, { "epoch": 2.95, "learning_rate": 1.2286502699460108e-06, "loss": 0.0911, "step": 24657 }, { "epoch": 2.95, "learning_rate": 1.2285782843431313e-06, "loss": 0.0001, "step": 24658 }, { "epoch": 2.95, "learning_rate": 1.228506298740252e-06, "loss": 0.0002, "step": 24659 }, { "epoch": 2.96, "learning_rate": 1.2284343131373727e-06, "loss": 0.0002, "step": 24660 }, { "epoch": 2.96, "learning_rate": 1.2283623275344932e-06, "loss": 0.0003, "step": 24661 }, { "epoch": 2.96, "learning_rate": 1.2282903419316138e-06, "loss": 0.0001, "step": 24662 }, { "epoch": 2.96, "learning_rate": 1.2282183563287344e-06, "loss": 0.0002, "step": 24663 }, { "epoch": 2.96, "learning_rate": 1.2281463707258547e-06, "loss": 0.0002, "step": 24664 }, { "epoch": 2.96, "learning_rate": 1.2280743851229753e-06, "loss": 0.0005, "step": 24665 }, { "epoch": 2.96, "learning_rate": 1.2280023995200959e-06, "loss": 0.0001, "step": 24666 }, { "epoch": 2.96, "learning_rate": 1.2279304139172166e-06, "loss": 0.0001, "step": 24667 }, { "epoch": 2.96, "learning_rate": 1.2278584283143372e-06, "loss": 0.0001, "step": 24668 }, { "epoch": 2.96, "learning_rate": 1.2277864427114578e-06, "loss": 0.0001, "step": 24669 }, { "epoch": 2.96, "learning_rate": 1.2277144571085783e-06, "loss": 0.0001, "step": 24670 }, { "epoch": 2.96, "learning_rate": 1.227642471505699e-06, "loss": 0.0001, "step": 24671 }, { "epoch": 2.96, "learning_rate": 1.2275704859028195e-06, "loss": 0.0001, "step": 24672 }, { "epoch": 2.96, "learning_rate": 1.22749850029994e-06, "loss": 0.0002, "step": 24673 }, { "epoch": 2.96, "learning_rate": 1.2274265146970606e-06, "loss": 0.0001, "step": 24674 }, { "epoch": 2.96, "learning_rate": 1.2273545290941812e-06, "loss": 0.0004, "step": 24675 }, { "epoch": 2.96, "learning_rate": 1.2272825434913017e-06, "loss": 0.0002, "step": 24676 }, { "epoch": 2.96, "learning_rate": 1.2272105578884223e-06, "loss": 0.0002, "step": 24677 }, { "epoch": 2.96, "learning_rate": 1.2271385722855429e-06, "loss": 0.0002, "step": 24678 }, { "epoch": 2.96, "learning_rate": 1.2270665866826634e-06, "loss": 0.0001, "step": 24679 }, { "epoch": 2.96, "learning_rate": 1.226994601079784e-06, "loss": 0.0001, "step": 24680 }, { "epoch": 2.96, "learning_rate": 1.2269226154769046e-06, "loss": 0.0001, "step": 24681 }, { "epoch": 2.96, "learning_rate": 1.2268506298740251e-06, "loss": 0.0005, "step": 24682 }, { "epoch": 2.96, "learning_rate": 1.226778644271146e-06, "loss": 0.0002, "step": 24683 }, { "epoch": 2.96, "learning_rate": 1.2267066586682665e-06, "loss": 0.0001, "step": 24684 }, { "epoch": 2.96, "learning_rate": 1.226634673065387e-06, "loss": 0.0001, "step": 24685 }, { "epoch": 2.96, "learning_rate": 1.2265626874625076e-06, "loss": 0.0002, "step": 24686 }, { "epoch": 2.96, "learning_rate": 1.226490701859628e-06, "loss": 0.0002, "step": 24687 }, { "epoch": 2.96, "learning_rate": 1.2264187162567485e-06, "loss": 0.0001, "step": 24688 }, { "epoch": 2.96, "learning_rate": 1.2263467306538691e-06, "loss": 0.0001, "step": 24689 }, { "epoch": 2.96, "learning_rate": 1.2262747450509899e-06, "loss": 0.0002, "step": 24690 }, { "epoch": 2.96, "learning_rate": 1.2262027594481105e-06, "loss": 0.0001, "step": 24691 }, { "epoch": 2.96, "learning_rate": 1.226130773845231e-06, "loss": 0.0001, "step": 24692 }, { "epoch": 2.96, "learning_rate": 1.2260587882423516e-06, "loss": 0.0001, "step": 24693 }, { "epoch": 2.96, "learning_rate": 1.2259868026394722e-06, "loss": 0.0001, "step": 24694 }, { "epoch": 2.96, "learning_rate": 1.2259148170365927e-06, "loss": 0.0001, "step": 24695 }, { "epoch": 2.96, "learning_rate": 1.2258428314337133e-06, "loss": 0.0002, "step": 24696 }, { "epoch": 2.96, "learning_rate": 1.2257708458308339e-06, "loss": 0.0001, "step": 24697 }, { "epoch": 2.96, "learning_rate": 1.2256988602279544e-06, "loss": 0.0001, "step": 24698 }, { "epoch": 2.96, "learning_rate": 1.225626874625075e-06, "loss": 0.0002, "step": 24699 }, { "epoch": 2.96, "learning_rate": 1.2255548890221956e-06, "loss": 0.0004, "step": 24700 }, { "epoch": 2.96, "learning_rate": 1.2254829034193161e-06, "loss": 0.0078, "step": 24701 }, { "epoch": 2.96, "learning_rate": 1.2254109178164367e-06, "loss": 0.0001, "step": 24702 }, { "epoch": 2.96, "learning_rate": 1.2253389322135573e-06, "loss": 0.0005, "step": 24703 }, { "epoch": 2.96, "learning_rate": 1.2252669466106778e-06, "loss": 0.0001, "step": 24704 }, { "epoch": 2.96, "learning_rate": 1.2251949610077984e-06, "loss": 0.0001, "step": 24705 }, { "epoch": 2.96, "learning_rate": 1.2251229754049192e-06, "loss": 0.0004, "step": 24706 }, { "epoch": 2.96, "learning_rate": 1.2250509898020397e-06, "loss": 0.0002, "step": 24707 }, { "epoch": 2.96, "learning_rate": 1.2249790041991603e-06, "loss": 0.0001, "step": 24708 }, { "epoch": 2.96, "learning_rate": 1.2249070185962809e-06, "loss": 0.0001, "step": 24709 }, { "epoch": 2.96, "learning_rate": 1.2248350329934012e-06, "loss": 0.0002, "step": 24710 }, { "epoch": 2.96, "learning_rate": 1.2247630473905218e-06, "loss": 0.0002, "step": 24711 }, { "epoch": 2.96, "learning_rate": 1.2246910617876424e-06, "loss": 0.0001, "step": 24712 }, { "epoch": 2.96, "learning_rate": 1.2246190761847631e-06, "loss": 0.0001, "step": 24713 }, { "epoch": 2.96, "learning_rate": 1.2245470905818837e-06, "loss": 0.0001, "step": 24714 }, { "epoch": 2.96, "learning_rate": 1.2244751049790043e-06, "loss": 0.0003, "step": 24715 }, { "epoch": 2.96, "learning_rate": 1.2244031193761248e-06, "loss": 0.0001, "step": 24716 }, { "epoch": 2.96, "learning_rate": 1.2243311337732454e-06, "loss": 0.0, "step": 24717 }, { "epoch": 2.96, "learning_rate": 1.224259148170366e-06, "loss": 0.0002, "step": 24718 }, { "epoch": 2.96, "learning_rate": 1.2241871625674865e-06, "loss": 0.0001, "step": 24719 }, { "epoch": 2.96, "learning_rate": 1.224115176964607e-06, "loss": 0.0001, "step": 24720 }, { "epoch": 2.96, "learning_rate": 1.2240431913617277e-06, "loss": 0.0002, "step": 24721 }, { "epoch": 2.96, "learning_rate": 1.2239712057588482e-06, "loss": 0.0001, "step": 24722 }, { "epoch": 2.96, "learning_rate": 1.2238992201559688e-06, "loss": 0.0001, "step": 24723 }, { "epoch": 2.96, "learning_rate": 1.2238272345530894e-06, "loss": 0.0001, "step": 24724 }, { "epoch": 2.96, "learning_rate": 1.22375524895021e-06, "loss": 0.0001, "step": 24725 }, { "epoch": 2.96, "learning_rate": 1.2236832633473305e-06, "loss": 0.0001, "step": 24726 }, { "epoch": 2.96, "learning_rate": 1.223611277744451e-06, "loss": 0.0002, "step": 24727 }, { "epoch": 2.96, "learning_rate": 1.2235392921415716e-06, "loss": 0.0001, "step": 24728 }, { "epoch": 2.96, "learning_rate": 1.2234673065386924e-06, "loss": 0.0001, "step": 24729 }, { "epoch": 2.96, "learning_rate": 1.223395320935813e-06, "loss": 0.8662, "step": 24730 }, { "epoch": 2.96, "learning_rate": 1.2233233353329336e-06, "loss": 0.0006, "step": 24731 }, { "epoch": 2.96, "learning_rate": 1.2232513497300541e-06, "loss": 0.0001, "step": 24732 }, { "epoch": 2.96, "learning_rate": 1.2231793641271745e-06, "loss": 0.0001, "step": 24733 }, { "epoch": 2.96, "learning_rate": 1.223107378524295e-06, "loss": 0.0001, "step": 24734 }, { "epoch": 2.96, "learning_rate": 1.2230353929214156e-06, "loss": 0.0001, "step": 24735 }, { "epoch": 2.96, "learning_rate": 1.2229634073185364e-06, "loss": 0.0001, "step": 24736 }, { "epoch": 2.96, "learning_rate": 1.222891421715657e-06, "loss": 0.0002, "step": 24737 }, { "epoch": 2.96, "learning_rate": 1.2228194361127775e-06, "loss": 0.0001, "step": 24738 }, { "epoch": 2.96, "learning_rate": 1.222747450509898e-06, "loss": 0.001, "step": 24739 }, { "epoch": 2.96, "learning_rate": 1.2226754649070187e-06, "loss": 0.0001, "step": 24740 }, { "epoch": 2.96, "learning_rate": 1.2226034793041392e-06, "loss": 0.0001, "step": 24741 }, { "epoch": 2.96, "learning_rate": 1.2225314937012598e-06, "loss": 0.0004, "step": 24742 }, { "epoch": 2.97, "learning_rate": 1.2224595080983804e-06, "loss": 0.0005, "step": 24743 }, { "epoch": 2.97, "learning_rate": 1.222387522495501e-06, "loss": 0.0001, "step": 24744 }, { "epoch": 2.97, "learning_rate": 1.2223155368926215e-06, "loss": 0.3453, "step": 24745 }, { "epoch": 2.97, "learning_rate": 1.222243551289742e-06, "loss": 0.0002, "step": 24746 }, { "epoch": 2.97, "learning_rate": 1.2221715656868626e-06, "loss": 0.0001, "step": 24747 }, { "epoch": 2.97, "learning_rate": 1.2220995800839832e-06, "loss": 0.0002, "step": 24748 }, { "epoch": 2.97, "learning_rate": 1.2220275944811038e-06, "loss": 0.3432, "step": 24749 }, { "epoch": 2.97, "learning_rate": 1.2219556088782243e-06, "loss": 0.0001, "step": 24750 }, { "epoch": 2.97, "learning_rate": 1.2218836232753449e-06, "loss": 0.1544, "step": 24751 }, { "epoch": 2.97, "learning_rate": 1.2218116376724657e-06, "loss": 0.0002, "step": 24752 }, { "epoch": 2.97, "learning_rate": 1.2217396520695862e-06, "loss": 0.0002, "step": 24753 }, { "epoch": 2.97, "learning_rate": 1.2216676664667068e-06, "loss": 0.0001, "step": 24754 }, { "epoch": 2.97, "learning_rate": 1.2215956808638274e-06, "loss": 0.0003, "step": 24755 }, { "epoch": 2.97, "learning_rate": 1.2215236952609477e-06, "loss": 0.0001, "step": 24756 }, { "epoch": 2.97, "learning_rate": 1.2214517096580683e-06, "loss": 0.0008, "step": 24757 }, { "epoch": 2.97, "learning_rate": 1.2213797240551889e-06, "loss": 0.0001, "step": 24758 }, { "epoch": 2.97, "learning_rate": 1.2213077384523096e-06, "loss": 0.0001, "step": 24759 }, { "epoch": 2.97, "learning_rate": 1.2212357528494302e-06, "loss": 0.0002, "step": 24760 }, { "epoch": 2.97, "learning_rate": 1.2211637672465508e-06, "loss": 0.0001, "step": 24761 }, { "epoch": 2.97, "learning_rate": 1.2210917816436713e-06, "loss": 0.0001, "step": 24762 }, { "epoch": 2.97, "learning_rate": 1.221019796040792e-06, "loss": 0.0052, "step": 24763 }, { "epoch": 2.97, "learning_rate": 1.2209478104379125e-06, "loss": 0.0002, "step": 24764 }, { "epoch": 2.97, "learning_rate": 1.220875824835033e-06, "loss": 0.0001, "step": 24765 }, { "epoch": 2.97, "learning_rate": 1.2208038392321536e-06, "loss": 0.0001, "step": 24766 }, { "epoch": 2.97, "learning_rate": 1.2207318536292742e-06, "loss": 0.0001, "step": 24767 }, { "epoch": 2.97, "learning_rate": 1.2206598680263947e-06, "loss": 0.0001, "step": 24768 }, { "epoch": 2.97, "learning_rate": 1.2205878824235153e-06, "loss": 0.0001, "step": 24769 }, { "epoch": 2.97, "learning_rate": 1.2205158968206359e-06, "loss": 0.0001, "step": 24770 }, { "epoch": 2.97, "learning_rate": 1.2204439112177564e-06, "loss": 0.0001, "step": 24771 }, { "epoch": 2.97, "learning_rate": 1.220371925614877e-06, "loss": 0.0001, "step": 24772 }, { "epoch": 2.97, "learning_rate": 1.2202999400119976e-06, "loss": 0.0001, "step": 24773 }, { "epoch": 2.97, "learning_rate": 1.2202279544091181e-06, "loss": 0.0003, "step": 24774 }, { "epoch": 2.97, "learning_rate": 1.220155968806239e-06, "loss": 0.0001, "step": 24775 }, { "epoch": 2.97, "learning_rate": 1.2200839832033595e-06, "loss": 0.0001, "step": 24776 }, { "epoch": 2.97, "learning_rate": 1.22001199760048e-06, "loss": 0.0001, "step": 24777 }, { "epoch": 2.97, "learning_rate": 1.2199400119976006e-06, "loss": 0.0001, "step": 24778 }, { "epoch": 2.97, "learning_rate": 1.2198680263947212e-06, "loss": 0.0001, "step": 24779 }, { "epoch": 2.97, "learning_rate": 1.2197960407918415e-06, "loss": 0.0002, "step": 24780 }, { "epoch": 2.97, "learning_rate": 1.219724055188962e-06, "loss": 0.0001, "step": 24781 }, { "epoch": 2.97, "learning_rate": 1.2196520695860827e-06, "loss": 0.0002, "step": 24782 }, { "epoch": 2.97, "learning_rate": 1.2195800839832034e-06, "loss": 0.0001, "step": 24783 }, { "epoch": 2.97, "learning_rate": 1.219508098380324e-06, "loss": 0.0003, "step": 24784 }, { "epoch": 2.97, "learning_rate": 1.2194361127774446e-06, "loss": 0.0003, "step": 24785 }, { "epoch": 2.97, "learning_rate": 1.2193641271745651e-06, "loss": 0.0001, "step": 24786 }, { "epoch": 2.97, "learning_rate": 1.2192921415716857e-06, "loss": 0.0001, "step": 24787 }, { "epoch": 2.97, "learning_rate": 1.2192201559688063e-06, "loss": 0.0001, "step": 24788 }, { "epoch": 2.97, "learning_rate": 1.2191481703659268e-06, "loss": 0.0003, "step": 24789 }, { "epoch": 2.97, "learning_rate": 1.2190761847630474e-06, "loss": 0.0001, "step": 24790 }, { "epoch": 2.97, "learning_rate": 1.219004199160168e-06, "loss": 0.0001, "step": 24791 }, { "epoch": 2.97, "learning_rate": 1.2189322135572885e-06, "loss": 0.5954, "step": 24792 }, { "epoch": 2.97, "learning_rate": 1.2188602279544091e-06, "loss": 0.0001, "step": 24793 }, { "epoch": 2.97, "learning_rate": 1.2187882423515297e-06, "loss": 0.0001, "step": 24794 }, { "epoch": 2.97, "learning_rate": 1.2187162567486503e-06, "loss": 0.0001, "step": 24795 }, { "epoch": 2.97, "learning_rate": 1.2186442711457708e-06, "loss": 0.0001, "step": 24796 }, { "epoch": 2.97, "learning_rate": 1.2185722855428914e-06, "loss": 0.0001, "step": 24797 }, { "epoch": 2.97, "learning_rate": 1.218500299940012e-06, "loss": 0.0001, "step": 24798 }, { "epoch": 2.97, "learning_rate": 1.2184283143371327e-06, "loss": 0.0003, "step": 24799 }, { "epoch": 2.97, "learning_rate": 1.2183563287342533e-06, "loss": 0.0001, "step": 24800 }, { "epoch": 2.97, "learning_rate": 1.2182843431313739e-06, "loss": 0.262, "step": 24801 }, { "epoch": 2.97, "learning_rate": 1.2182123575284944e-06, "loss": 0.0001, "step": 24802 }, { "epoch": 2.97, "learning_rate": 1.2181403719256148e-06, "loss": 0.0001, "step": 24803 }, { "epoch": 2.97, "learning_rate": 1.2180683863227354e-06, "loss": 0.0001, "step": 24804 }, { "epoch": 2.97, "learning_rate": 1.217996400719856e-06, "loss": 0.0001, "step": 24805 }, { "epoch": 2.97, "learning_rate": 1.2179244151169767e-06, "loss": 0.0001, "step": 24806 }, { "epoch": 2.97, "learning_rate": 1.2178524295140973e-06, "loss": 0.0003, "step": 24807 }, { "epoch": 2.97, "learning_rate": 1.2177804439112178e-06, "loss": 0.2799, "step": 24808 }, { "epoch": 2.97, "learning_rate": 1.2177084583083384e-06, "loss": 0.805, "step": 24809 }, { "epoch": 2.97, "learning_rate": 1.217636472705459e-06, "loss": 0.0001, "step": 24810 }, { "epoch": 2.97, "learning_rate": 1.2175644871025795e-06, "loss": 0.0006, "step": 24811 }, { "epoch": 2.97, "learning_rate": 1.2174925014997e-06, "loss": 0.0002, "step": 24812 }, { "epoch": 2.97, "learning_rate": 1.2174205158968207e-06, "loss": 0.0001, "step": 24813 }, { "epoch": 2.97, "learning_rate": 1.2173485302939412e-06, "loss": 0.0001, "step": 24814 }, { "epoch": 2.97, "learning_rate": 1.2172765446910618e-06, "loss": 0.0001, "step": 24815 }, { "epoch": 2.97, "learning_rate": 1.2172045590881824e-06, "loss": 0.0001, "step": 24816 }, { "epoch": 2.97, "learning_rate": 1.217132573485303e-06, "loss": 0.0001, "step": 24817 }, { "epoch": 2.97, "learning_rate": 1.2170605878824235e-06, "loss": 0.0001, "step": 24818 }, { "epoch": 2.97, "learning_rate": 1.216988602279544e-06, "loss": 0.0001, "step": 24819 }, { "epoch": 2.97, "learning_rate": 1.2169166166766646e-06, "loss": 0.0005, "step": 24820 }, { "epoch": 2.97, "learning_rate": 1.2168446310737852e-06, "loss": 0.0001, "step": 24821 }, { "epoch": 2.97, "learning_rate": 1.216772645470906e-06, "loss": 0.6119, "step": 24822 }, { "epoch": 2.97, "learning_rate": 1.2167006598680265e-06, "loss": 0.0001, "step": 24823 }, { "epoch": 2.97, "learning_rate": 1.2166286742651471e-06, "loss": 0.0003, "step": 24824 }, { "epoch": 2.97, "learning_rate": 1.2165566886622677e-06, "loss": 0.0001, "step": 24825 }, { "epoch": 2.97, "learning_rate": 1.216484703059388e-06, "loss": 0.0002, "step": 24826 }, { "epoch": 2.98, "learning_rate": 1.2164127174565086e-06, "loss": 0.6686, "step": 24827 }, { "epoch": 2.98, "learning_rate": 1.2163407318536292e-06, "loss": 0.0001, "step": 24828 }, { "epoch": 2.98, "learning_rate": 1.21626874625075e-06, "loss": 0.0001, "step": 24829 }, { "epoch": 2.98, "learning_rate": 1.2161967606478705e-06, "loss": 0.0028, "step": 24830 }, { "epoch": 2.98, "learning_rate": 1.216124775044991e-06, "loss": 0.0001, "step": 24831 }, { "epoch": 2.98, "learning_rate": 1.2160527894421116e-06, "loss": 0.0001, "step": 24832 }, { "epoch": 2.98, "learning_rate": 1.2159808038392322e-06, "loss": 0.0001, "step": 24833 }, { "epoch": 2.98, "learning_rate": 1.2159088182363528e-06, "loss": 0.0002, "step": 24834 }, { "epoch": 2.98, "learning_rate": 1.2158368326334733e-06, "loss": 0.0004, "step": 24835 }, { "epoch": 2.98, "learning_rate": 1.215764847030594e-06, "loss": 0.0001, "step": 24836 }, { "epoch": 2.98, "learning_rate": 1.2156928614277145e-06, "loss": 0.0001, "step": 24837 }, { "epoch": 2.98, "learning_rate": 1.215620875824835e-06, "loss": 0.0003, "step": 24838 }, { "epoch": 2.98, "learning_rate": 1.2155488902219556e-06, "loss": 0.0001, "step": 24839 }, { "epoch": 2.98, "learning_rate": 1.2154769046190762e-06, "loss": 0.0017, "step": 24840 }, { "epoch": 2.98, "learning_rate": 1.2154049190161967e-06, "loss": 0.0012, "step": 24841 }, { "epoch": 2.98, "learning_rate": 1.2153329334133173e-06, "loss": 0.0001, "step": 24842 }, { "epoch": 2.98, "learning_rate": 1.2152609478104379e-06, "loss": 0.0001, "step": 24843 }, { "epoch": 2.98, "learning_rate": 1.2151889622075584e-06, "loss": 0.0002, "step": 24844 }, { "epoch": 2.98, "learning_rate": 1.2151169766046792e-06, "loss": 0.0001, "step": 24845 }, { "epoch": 2.98, "learning_rate": 1.2150449910017998e-06, "loss": 0.0225, "step": 24846 }, { "epoch": 2.98, "learning_rate": 1.2149730053989204e-06, "loss": 0.0001, "step": 24847 }, { "epoch": 2.98, "learning_rate": 1.214901019796041e-06, "loss": 0.0001, "step": 24848 }, { "epoch": 2.98, "learning_rate": 1.2148290341931613e-06, "loss": 0.0001, "step": 24849 }, { "epoch": 2.98, "learning_rate": 1.2147570485902818e-06, "loss": 0.0002, "step": 24850 }, { "epoch": 2.98, "learning_rate": 1.2146850629874024e-06, "loss": 0.058, "step": 24851 }, { "epoch": 2.98, "learning_rate": 1.2146130773845232e-06, "loss": 0.0001, "step": 24852 }, { "epoch": 2.98, "learning_rate": 1.2145410917816438e-06, "loss": 0.0001, "step": 24853 }, { "epoch": 2.98, "learning_rate": 1.2144691061787643e-06, "loss": 0.0003, "step": 24854 }, { "epoch": 2.98, "learning_rate": 1.2143971205758849e-06, "loss": 0.0001, "step": 24855 }, { "epoch": 2.98, "learning_rate": 1.2143251349730055e-06, "loss": 0.0002, "step": 24856 }, { "epoch": 2.98, "learning_rate": 1.214253149370126e-06, "loss": 0.0002, "step": 24857 }, { "epoch": 2.98, "learning_rate": 1.2141811637672466e-06, "loss": 0.0002, "step": 24858 }, { "epoch": 2.98, "learning_rate": 1.2141091781643672e-06, "loss": 0.0003, "step": 24859 }, { "epoch": 2.98, "learning_rate": 1.2140371925614877e-06, "loss": 0.0001, "step": 24860 }, { "epoch": 2.98, "learning_rate": 1.2139652069586083e-06, "loss": 0.0003, "step": 24861 }, { "epoch": 2.98, "learning_rate": 1.2138932213557289e-06, "loss": 0.0002, "step": 24862 }, { "epoch": 2.98, "learning_rate": 1.2138212357528494e-06, "loss": 0.0002, "step": 24863 }, { "epoch": 2.98, "learning_rate": 1.21374925014997e-06, "loss": 0.0001, "step": 24864 }, { "epoch": 2.98, "learning_rate": 1.2136772645470906e-06, "loss": 0.0001, "step": 24865 }, { "epoch": 2.98, "learning_rate": 1.2136052789442111e-06, "loss": 0.0001, "step": 24866 }, { "epoch": 2.98, "learning_rate": 1.2135332933413317e-06, "loss": 0.0004, "step": 24867 }, { "epoch": 2.98, "learning_rate": 1.2134613077384525e-06, "loss": 0.0239, "step": 24868 }, { "epoch": 2.98, "learning_rate": 1.213389322135573e-06, "loss": 0.0002, "step": 24869 }, { "epoch": 2.98, "learning_rate": 1.2133173365326936e-06, "loss": 0.0002, "step": 24870 }, { "epoch": 2.98, "learning_rate": 1.2132453509298142e-06, "loss": 0.0001, "step": 24871 }, { "epoch": 2.98, "learning_rate": 1.2131733653269345e-06, "loss": 0.0001, "step": 24872 }, { "epoch": 2.98, "learning_rate": 1.213101379724055e-06, "loss": 0.0001, "step": 24873 }, { "epoch": 2.98, "learning_rate": 1.2130293941211757e-06, "loss": 0.0001, "step": 24874 }, { "epoch": 2.98, "learning_rate": 1.2129574085182964e-06, "loss": 0.0012, "step": 24875 }, { "epoch": 2.98, "learning_rate": 1.212885422915417e-06, "loss": 0.0008, "step": 24876 }, { "epoch": 2.98, "learning_rate": 1.2128134373125376e-06, "loss": 0.0001, "step": 24877 }, { "epoch": 2.98, "learning_rate": 1.2127414517096581e-06, "loss": 0.0002, "step": 24878 }, { "epoch": 2.98, "learning_rate": 1.2126694661067787e-06, "loss": 0.0001, "step": 24879 }, { "epoch": 2.98, "learning_rate": 1.2125974805038993e-06, "loss": 0.0001, "step": 24880 }, { "epoch": 2.98, "learning_rate": 1.2125254949010198e-06, "loss": 0.0002, "step": 24881 }, { "epoch": 2.98, "learning_rate": 1.2124535092981404e-06, "loss": 0.0002, "step": 24882 }, { "epoch": 2.98, "learning_rate": 1.212381523695261e-06, "loss": 0.0001, "step": 24883 }, { "epoch": 2.98, "learning_rate": 1.2123095380923815e-06, "loss": 0.0001, "step": 24884 }, { "epoch": 2.98, "learning_rate": 1.2122375524895021e-06, "loss": 0.0015, "step": 24885 }, { "epoch": 2.98, "learning_rate": 1.2121655668866227e-06, "loss": 0.0001, "step": 24886 }, { "epoch": 2.98, "learning_rate": 1.2120935812837432e-06, "loss": 0.0001, "step": 24887 }, { "epoch": 2.98, "learning_rate": 1.2120215956808638e-06, "loss": 0.0001, "step": 24888 }, { "epoch": 2.98, "learning_rate": 1.2119496100779844e-06, "loss": 0.0002, "step": 24889 }, { "epoch": 2.98, "learning_rate": 1.211877624475105e-06, "loss": 0.0001, "step": 24890 }, { "epoch": 2.98, "learning_rate": 1.2118056388722257e-06, "loss": 0.0002, "step": 24891 }, { "epoch": 2.98, "learning_rate": 1.2117336532693463e-06, "loss": 0.0001, "step": 24892 }, { "epoch": 2.98, "learning_rate": 1.2116616676664669e-06, "loss": 0.0001, "step": 24893 }, { "epoch": 2.98, "learning_rate": 1.2115896820635874e-06, "loss": 0.0001, "step": 24894 }, { "epoch": 2.98, "learning_rate": 1.2115176964607078e-06, "loss": 0.0002, "step": 24895 }, { "epoch": 2.98, "learning_rate": 1.2114457108578283e-06, "loss": 0.0001, "step": 24896 }, { "epoch": 2.98, "learning_rate": 1.211373725254949e-06, "loss": 0.0001, "step": 24897 }, { "epoch": 2.98, "learning_rate": 1.2113017396520697e-06, "loss": 0.0001, "step": 24898 }, { "epoch": 2.98, "learning_rate": 1.2112297540491903e-06, "loss": 0.0001, "step": 24899 }, { "epoch": 2.98, "learning_rate": 1.2111577684463108e-06, "loss": 0.0002, "step": 24900 }, { "epoch": 2.98, "learning_rate": 1.2110857828434314e-06, "loss": 0.0088, "step": 24901 }, { "epoch": 2.98, "learning_rate": 1.211013797240552e-06, "loss": 0.0001, "step": 24902 }, { "epoch": 2.98, "learning_rate": 1.2109418116376725e-06, "loss": 0.0001, "step": 24903 }, { "epoch": 2.98, "learning_rate": 1.210869826034793e-06, "loss": 0.0001, "step": 24904 }, { "epoch": 2.98, "learning_rate": 1.2107978404319137e-06, "loss": 0.0003, "step": 24905 }, { "epoch": 2.98, "learning_rate": 1.2107258548290342e-06, "loss": 0.0001, "step": 24906 }, { "epoch": 2.98, "learning_rate": 1.2106538692261548e-06, "loss": 0.0002, "step": 24907 }, { "epoch": 2.98, "learning_rate": 1.2105818836232754e-06, "loss": 0.0001, "step": 24908 }, { "epoch": 2.98, "learning_rate": 1.210509898020396e-06, "loss": 0.0001, "step": 24909 }, { "epoch": 2.99, "learning_rate": 1.2104379124175165e-06, "loss": 0.0001, "step": 24910 }, { "epoch": 2.99, "learning_rate": 1.210365926814637e-06, "loss": 0.0001, "step": 24911 }, { "epoch": 2.99, "learning_rate": 1.2102939412117576e-06, "loss": 0.0001, "step": 24912 }, { "epoch": 2.99, "learning_rate": 1.2102219556088782e-06, "loss": 0.0001, "step": 24913 }, { "epoch": 2.99, "learning_rate": 1.2101499700059988e-06, "loss": 0.0001, "step": 24914 }, { "epoch": 2.99, "learning_rate": 1.2100779844031195e-06, "loss": 0.0001, "step": 24915 }, { "epoch": 2.99, "learning_rate": 1.21000599880024e-06, "loss": 0.0001, "step": 24916 }, { "epoch": 2.99, "learning_rate": 1.2099340131973607e-06, "loss": 0.0001, "step": 24917 }, { "epoch": 2.99, "learning_rate": 1.209862027594481e-06, "loss": 0.0003, "step": 24918 }, { "epoch": 2.99, "learning_rate": 1.2097900419916016e-06, "loss": 0.0, "step": 24919 }, { "epoch": 2.99, "learning_rate": 1.2097180563887222e-06, "loss": 0.0003, "step": 24920 }, { "epoch": 2.99, "learning_rate": 1.2096460707858427e-06, "loss": 0.0001, "step": 24921 }, { "epoch": 2.99, "learning_rate": 1.2095740851829635e-06, "loss": 0.0, "step": 24922 }, { "epoch": 2.99, "learning_rate": 1.209502099580084e-06, "loss": 0.0003, "step": 24923 }, { "epoch": 2.99, "learning_rate": 1.2094301139772046e-06, "loss": 0.0001, "step": 24924 }, { "epoch": 2.99, "learning_rate": 1.2093581283743252e-06, "loss": 0.0001, "step": 24925 }, { "epoch": 2.99, "learning_rate": 1.2092861427714458e-06, "loss": 0.0001, "step": 24926 }, { "epoch": 2.99, "learning_rate": 1.2092141571685663e-06, "loss": 0.0001, "step": 24927 }, { "epoch": 2.99, "learning_rate": 1.209142171565687e-06, "loss": 0.0001, "step": 24928 }, { "epoch": 2.99, "learning_rate": 1.2090701859628075e-06, "loss": 0.0001, "step": 24929 }, { "epoch": 2.99, "learning_rate": 1.208998200359928e-06, "loss": 0.0005, "step": 24930 }, { "epoch": 2.99, "learning_rate": 1.2089262147570486e-06, "loss": 0.1462, "step": 24931 }, { "epoch": 2.99, "learning_rate": 1.2088542291541692e-06, "loss": 0.0001, "step": 24932 }, { "epoch": 2.99, "learning_rate": 1.2087822435512897e-06, "loss": 0.0004, "step": 24933 }, { "epoch": 2.99, "learning_rate": 1.2087102579484103e-06, "loss": 0.0001, "step": 24934 }, { "epoch": 2.99, "learning_rate": 1.2086382723455309e-06, "loss": 0.0001, "step": 24935 }, { "epoch": 2.99, "learning_rate": 1.2085662867426514e-06, "loss": 0.0001, "step": 24936 }, { "epoch": 2.99, "learning_rate": 1.208494301139772e-06, "loss": 0.0001, "step": 24937 }, { "epoch": 2.99, "learning_rate": 1.2084223155368928e-06, "loss": 0.0001, "step": 24938 }, { "epoch": 2.99, "learning_rate": 1.2083503299340134e-06, "loss": 0.0001, "step": 24939 }, { "epoch": 2.99, "learning_rate": 1.208278344331134e-06, "loss": 0.0001, "step": 24940 }, { "epoch": 2.99, "learning_rate": 1.2082063587282543e-06, "loss": 0.0, "step": 24941 }, { "epoch": 2.99, "learning_rate": 1.2081343731253748e-06, "loss": 0.0001, "step": 24942 }, { "epoch": 2.99, "learning_rate": 1.2080623875224954e-06, "loss": 0.0001, "step": 24943 }, { "epoch": 2.99, "learning_rate": 1.207990401919616e-06, "loss": 0.0006, "step": 24944 }, { "epoch": 2.99, "learning_rate": 1.2079184163167368e-06, "loss": 0.0001, "step": 24945 }, { "epoch": 2.99, "learning_rate": 1.2078464307138573e-06, "loss": 0.0004, "step": 24946 }, { "epoch": 2.99, "learning_rate": 1.2077744451109779e-06, "loss": 0.0001, "step": 24947 }, { "epoch": 2.99, "learning_rate": 1.2077024595080985e-06, "loss": 0.0001, "step": 24948 }, { "epoch": 2.99, "learning_rate": 1.207630473905219e-06, "loss": 0.0001, "step": 24949 }, { "epoch": 2.99, "learning_rate": 1.2075584883023396e-06, "loss": 0.0001, "step": 24950 }, { "epoch": 2.99, "learning_rate": 1.2074865026994602e-06, "loss": 0.0001, "step": 24951 }, { "epoch": 2.99, "learning_rate": 1.2074145170965807e-06, "loss": 0.0002, "step": 24952 }, { "epoch": 2.99, "learning_rate": 1.2073425314937013e-06, "loss": 0.0001, "step": 24953 }, { "epoch": 2.99, "learning_rate": 1.2072705458908219e-06, "loss": 0.0001, "step": 24954 }, { "epoch": 2.99, "learning_rate": 1.2071985602879424e-06, "loss": 0.0001, "step": 24955 }, { "epoch": 2.99, "learning_rate": 1.207126574685063e-06, "loss": 0.8375, "step": 24956 }, { "epoch": 2.99, "learning_rate": 1.2070545890821836e-06, "loss": 0.0001, "step": 24957 }, { "epoch": 2.99, "learning_rate": 1.2069826034793041e-06, "loss": 0.0001, "step": 24958 }, { "epoch": 2.99, "learning_rate": 1.2069106178764247e-06, "loss": 0.009, "step": 24959 }, { "epoch": 2.99, "learning_rate": 1.2068386322735453e-06, "loss": 0.6511, "step": 24960 }, { "epoch": 2.99, "learning_rate": 1.206766646670666e-06, "loss": 0.0001, "step": 24961 }, { "epoch": 2.99, "learning_rate": 1.2066946610677866e-06, "loss": 0.0001, "step": 24962 }, { "epoch": 2.99, "learning_rate": 1.2066226754649072e-06, "loss": 0.8015, "step": 24963 }, { "epoch": 2.99, "learning_rate": 1.2065506898620275e-06, "loss": 0.0001, "step": 24964 }, { "epoch": 2.99, "learning_rate": 1.206478704259148e-06, "loss": 0.0001, "step": 24965 }, { "epoch": 2.99, "learning_rate": 1.2064067186562687e-06, "loss": 0.0003, "step": 24966 }, { "epoch": 2.99, "learning_rate": 1.2063347330533892e-06, "loss": 0.0001, "step": 24967 }, { "epoch": 2.99, "learning_rate": 1.20626274745051e-06, "loss": 0.0001, "step": 24968 }, { "epoch": 2.99, "learning_rate": 1.2061907618476306e-06, "loss": 0.0001, "step": 24969 }, { "epoch": 2.99, "learning_rate": 1.2061187762447511e-06, "loss": 0.0001, "step": 24970 }, { "epoch": 2.99, "learning_rate": 1.2060467906418717e-06, "loss": 0.0001, "step": 24971 }, { "epoch": 2.99, "learning_rate": 1.2059748050389923e-06, "loss": 0.0, "step": 24972 }, { "epoch": 2.99, "learning_rate": 1.2059028194361128e-06, "loss": 0.0001, "step": 24973 }, { "epoch": 2.99, "learning_rate": 1.2058308338332334e-06, "loss": 0.0001, "step": 24974 }, { "epoch": 2.99, "learning_rate": 1.205758848230354e-06, "loss": 0.0001, "step": 24975 }, { "epoch": 2.99, "learning_rate": 1.2056868626274745e-06, "loss": 0.0001, "step": 24976 }, { "epoch": 2.99, "learning_rate": 1.205614877024595e-06, "loss": 0.0001, "step": 24977 }, { "epoch": 2.99, "learning_rate": 1.2055428914217157e-06, "loss": 0.0001, "step": 24978 }, { "epoch": 2.99, "learning_rate": 1.2054709058188362e-06, "loss": 0.0001, "step": 24979 }, { "epoch": 2.99, "learning_rate": 1.2053989202159568e-06, "loss": 0.0001, "step": 24980 }, { "epoch": 2.99, "learning_rate": 1.2053269346130774e-06, "loss": 0.0001, "step": 24981 }, { "epoch": 2.99, "learning_rate": 1.205254949010198e-06, "loss": 0.0001, "step": 24982 }, { "epoch": 2.99, "learning_rate": 1.2051829634073185e-06, "loss": 0.0001, "step": 24983 }, { "epoch": 2.99, "learning_rate": 1.2051109778044393e-06, "loss": 0.0001, "step": 24984 }, { "epoch": 2.99, "learning_rate": 1.2050389922015598e-06, "loss": 0.8612, "step": 24985 }, { "epoch": 2.99, "learning_rate": 1.2049670065986804e-06, "loss": 0.0001, "step": 24986 }, { "epoch": 2.99, "learning_rate": 1.2048950209958008e-06, "loss": 0.0001, "step": 24987 }, { "epoch": 2.99, "learning_rate": 1.2048230353929213e-06, "loss": 0.0001, "step": 24988 }, { "epoch": 2.99, "learning_rate": 1.204751049790042e-06, "loss": 0.0001, "step": 24989 }, { "epoch": 2.99, "learning_rate": 1.2046790641871625e-06, "loss": 0.0001, "step": 24990 }, { "epoch": 2.99, "learning_rate": 1.2046070785842832e-06, "loss": 0.0, "step": 24991 }, { "epoch": 2.99, "learning_rate": 1.2045350929814038e-06, "loss": 0.0001, "step": 24992 }, { "epoch": 2.99, "learning_rate": 1.2044631073785244e-06, "loss": 0.0001, "step": 24993 }, { "epoch": 3.0, "learning_rate": 1.204391121775645e-06, "loss": 0.0001, "step": 24994 }, { "epoch": 3.0, "learning_rate": 1.2043191361727655e-06, "loss": 0.0001, "step": 24995 }, { "epoch": 3.0, "learning_rate": 1.204247150569886e-06, "loss": 0.0001, "step": 24996 }, { "epoch": 3.0, "learning_rate": 1.2041751649670066e-06, "loss": 0.0001, "step": 24997 }, { "epoch": 3.0, "learning_rate": 1.2041031793641272e-06, "loss": 0.0001, "step": 24998 }, { "epoch": 3.0, "learning_rate": 1.2040311937612478e-06, "loss": 0.0075, "step": 24999 }, { "epoch": 3.0, "learning_rate": 1.2039592081583683e-06, "loss": 0.0001, "step": 25000 }, { "epoch": 3.0, "learning_rate": 1.203887222555489e-06, "loss": 0.0001, "step": 25001 }, { "epoch": 3.0, "learning_rate": 1.2038152369526095e-06, "loss": 0.0008, "step": 25002 }, { "epoch": 3.0, "learning_rate": 1.20374325134973e-06, "loss": 0.0001, "step": 25003 }, { "epoch": 3.0, "learning_rate": 1.2036712657468506e-06, "loss": 0.0003, "step": 25004 }, { "epoch": 3.0, "learning_rate": 1.2035992801439712e-06, "loss": 0.0001, "step": 25005 }, { "epoch": 3.0, "learning_rate": 1.2035272945410917e-06, "loss": 0.0002, "step": 25006 }, { "epoch": 3.0, "learning_rate": 1.2034553089382125e-06, "loss": 0.0001, "step": 25007 }, { "epoch": 3.0, "learning_rate": 1.203383323335333e-06, "loss": 0.0001, "step": 25008 }, { "epoch": 3.0, "learning_rate": 1.2033113377324537e-06, "loss": 0.0001, "step": 25009 }, { "epoch": 3.0, "learning_rate": 1.203239352129574e-06, "loss": 0.0002, "step": 25010 }, { "epoch": 3.0, "learning_rate": 1.2031673665266946e-06, "loss": 0.0005, "step": 25011 }, { "epoch": 3.0, "learning_rate": 1.2030953809238152e-06, "loss": 0.0002, "step": 25012 }, { "epoch": 3.0, "learning_rate": 1.2030233953209357e-06, "loss": 0.0001, "step": 25013 }, { "epoch": 3.0, "learning_rate": 1.2029514097180565e-06, "loss": 0.0001, "step": 25014 }, { "epoch": 3.0, "learning_rate": 1.202879424115177e-06, "loss": 0.0001, "step": 25015 }, { "epoch": 3.0, "learning_rate": 1.2028074385122976e-06, "loss": 0.0003, "step": 25016 }, { "epoch": 3.0, "learning_rate": 1.2027354529094182e-06, "loss": 0.0006, "step": 25017 }, { "epoch": 3.0, "learning_rate": 1.2026634673065388e-06, "loss": 0.0002, "step": 25018 }, { "epoch": 3.0, "learning_rate": 1.2025914817036593e-06, "loss": 0.0002, "step": 25019 }, { "epoch": 3.0, "learning_rate": 1.20251949610078e-06, "loss": 0.0001, "step": 25020 }, { "epoch": 3.0, "learning_rate": 1.2024475104979005e-06, "loss": 0.0001, "step": 25021 }, { "epoch": 3.0, "learning_rate": 1.202375524895021e-06, "loss": 0.0001, "step": 25022 }, { "epoch": 3.0, "learning_rate": 1.2023035392921416e-06, "loss": 0.0001, "step": 25023 }, { "epoch": 3.0, "learning_rate": 1.2022315536892622e-06, "loss": 0.0001, "step": 25024 }, { "epoch": 3.0, "learning_rate": 1.2021595680863827e-06, "loss": 0.8403, "step": 25025 }, { "epoch": 3.0, "learning_rate": 1.2020875824835033e-06, "loss": 0.3806, "step": 25026 }, { "epoch": 3.0, "learning_rate": 1.2020155968806239e-06, "loss": 0.0001, "step": 25027 }, { "epoch": 3.0, "learning_rate": 1.2019436112777444e-06, "loss": 0.0001, "step": 25028 }, { "epoch": 3.0, "learning_rate": 1.201871625674865e-06, "loss": 0.0, "step": 25029 }, { "epoch": 3.0, "learning_rate": 1.2017996400719858e-06, "loss": 0.0001, "step": 25030 }, { "epoch": 3.0, "learning_rate": 1.2017276544691063e-06, "loss": 0.0004, "step": 25031 }, { "epoch": 3.0, "learning_rate": 1.201655668866227e-06, "loss": 0.0001, "step": 25032 }, { "epoch": 3.0, "learning_rate": 1.2015836832633475e-06, "loss": 0.0002, "step": 25033 }, { "epoch": 3.0, "learning_rate": 1.2015116976604678e-06, "loss": 0.0001, "step": 25034 }, { "epoch": 3.0, "learning_rate": 1.2014397120575884e-06, "loss": 0.0001, "step": 25035 }, { "epoch": 3.0, "eval_accuracy": 0.9919712402636309, "eval_loss": 0.04723460599780083, "eval_runtime": 158.1022, "eval_samples_per_second": 105.565, "eval_steps_per_second": 1.651, "step": 25035 } ], "logging_steps": 1, "max_steps": 41725, "num_train_epochs": 5, "save_steps": 500, "total_flos": 2.5634588543205624e+16, "trial_name": null, "trial_params": null }