{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.9618834080717489, "global_step": 3500, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 6e-07, "loss": 8.6539, "step": 1 }, { "epoch": 0.0, "learning_rate": 1.2e-06, "loss": 8.3505, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.8e-06, "loss": 8.8541, "step": 3 }, { "epoch": 0.0, "learning_rate": 2.4e-06, "loss": 8.7158, "step": 4 }, { "epoch": 0.0, "learning_rate": 2.9999999999999997e-06, "loss": 8.5395, "step": 5 }, { "epoch": 0.0, "learning_rate": 3.6e-06, "loss": 9.6752, "step": 6 }, { "epoch": 0.0, "learning_rate": 4.2e-06, "loss": 8.4508, "step": 7 }, { "epoch": 0.0, "learning_rate": 4.8e-06, "loss": 8.9823, "step": 8 }, { "epoch": 0.01, "learning_rate": 5.399999999999999e-06, "loss": 9.1018, "step": 9 }, { "epoch": 0.01, "learning_rate": 5.999999999999999e-06, "loss": 8.4243, "step": 10 }, { "epoch": 0.01, "learning_rate": 6.599999999999999e-06, "loss": 9.3172, "step": 11 }, { "epoch": 0.01, "learning_rate": 7.2e-06, "loss": 8.6402, "step": 12 }, { "epoch": 0.01, "learning_rate": 7.799999999999998e-06, "loss": 8.8896, "step": 13 }, { "epoch": 0.01, "learning_rate": 8.4e-06, "loss": 8.7231, "step": 14 }, { "epoch": 0.01, "learning_rate": 8.999999999999999e-06, "loss": 8.1952, "step": 15 }, { "epoch": 0.01, "learning_rate": 9.6e-06, "loss": 7.8843, "step": 16 }, { "epoch": 0.01, "learning_rate": 1.02e-05, "loss": 8.063, "step": 17 }, { "epoch": 0.01, "learning_rate": 1.0799999999999998e-05, "loss": 8.6748, "step": 18 }, { "epoch": 0.01, "learning_rate": 1.14e-05, "loss": 8.4858, "step": 19 }, { "epoch": 0.01, "learning_rate": 1.1999999999999999e-05, "loss": 8.2403, "step": 20 }, { "epoch": 0.01, "learning_rate": 1.26e-05, "loss": 8.8869, "step": 21 }, { "epoch": 0.01, "learning_rate": 1.3199999999999997e-05, "loss": 8.3035, "step": 22 }, { "epoch": 0.01, "learning_rate": 1.3799999999999998e-05, "loss": 7.7211, "step": 23 }, { "epoch": 0.01, "learning_rate": 1.44e-05, "loss": 7.753, "step": 24 }, { "epoch": 0.01, "learning_rate": 1.4999999999999999e-05, "loss": 8.0471, "step": 25 }, { "epoch": 0.01, "learning_rate": 1.5599999999999996e-05, "loss": 7.7415, "step": 26 }, { "epoch": 0.02, "learning_rate": 1.6199999999999997e-05, "loss": 7.4489, "step": 27 }, { "epoch": 0.02, "learning_rate": 1.6199999999999997e-05, "loss": 7.3258, "step": 28 }, { "epoch": 0.02, "learning_rate": 1.68e-05, "loss": 7.0812, "step": 29 }, { "epoch": 0.02, "learning_rate": 1.74e-05, "loss": 6.9329, "step": 30 }, { "epoch": 0.02, "learning_rate": 1.7999999999999997e-05, "loss": 7.0746, "step": 31 }, { "epoch": 0.02, "learning_rate": 1.8599999999999998e-05, "loss": 6.785, "step": 32 }, { "epoch": 0.02, "learning_rate": 1.92e-05, "loss": 6.1227, "step": 33 }, { "epoch": 0.02, "learning_rate": 1.98e-05, "loss": 6.0981, "step": 34 }, { "epoch": 0.02, "learning_rate": 2.04e-05, "loss": 5.2057, "step": 35 }, { "epoch": 0.02, "learning_rate": 2.1e-05, "loss": 5.7122, "step": 36 }, { "epoch": 0.02, "learning_rate": 2.1599999999999996e-05, "loss": 4.6245, "step": 37 }, { "epoch": 0.02, "learning_rate": 2.2199999999999998e-05, "loss": 4.5981, "step": 38 }, { "epoch": 0.02, "learning_rate": 2.28e-05, "loss": 4.2869, "step": 39 }, { "epoch": 0.02, "learning_rate": 2.34e-05, "loss": 3.987, "step": 40 }, { "epoch": 0.02, "learning_rate": 2.3999999999999997e-05, "loss": 3.7459, "step": 41 }, { "epoch": 0.02, "learning_rate": 2.4599999999999998e-05, "loss": 3.525, "step": 42 }, { "epoch": 0.02, "learning_rate": 2.52e-05, "loss": 3.4569, "step": 43 }, { "epoch": 0.02, "learning_rate": 2.5799999999999997e-05, "loss": 3.3914, "step": 44 }, { "epoch": 0.03, "learning_rate": 2.6399999999999995e-05, "loss": 3.318, "step": 45 }, { "epoch": 0.03, "learning_rate": 2.6999999999999996e-05, "loss": 3.2911, "step": 46 }, { "epoch": 0.03, "learning_rate": 2.7599999999999997e-05, "loss": 3.2887, "step": 47 }, { "epoch": 0.03, "learning_rate": 2.8199999999999998e-05, "loss": 3.3229, "step": 48 }, { "epoch": 0.03, "learning_rate": 2.88e-05, "loss": 3.2328, "step": 49 }, { "epoch": 0.03, "learning_rate": 2.94e-05, "loss": 3.25, "step": 50 }, { "epoch": 0.03, "learning_rate": 2.9999999999999997e-05, "loss": 3.1102, "step": 51 }, { "epoch": 0.03, "learning_rate": 3.06e-05, "loss": 3.0985, "step": 52 }, { "epoch": 0.03, "learning_rate": 3.119999999999999e-05, "loss": 3.073, "step": 53 }, { "epoch": 0.03, "learning_rate": 3.1799999999999994e-05, "loss": 3.0488, "step": 54 }, { "epoch": 0.03, "learning_rate": 3.2399999999999995e-05, "loss": 3.0633, "step": 55 }, { "epoch": 0.03, "learning_rate": 3.2999999999999996e-05, "loss": 3.0112, "step": 56 }, { "epoch": 0.03, "learning_rate": 3.36e-05, "loss": 2.9985, "step": 57 }, { "epoch": 0.03, "learning_rate": 3.42e-05, "loss": 2.9938, "step": 58 }, { "epoch": 0.03, "learning_rate": 3.48e-05, "loss": 2.9952, "step": 59 }, { "epoch": 0.03, "learning_rate": 3.539999999999999e-05, "loss": 2.9772, "step": 60 }, { "epoch": 0.03, "learning_rate": 3.5999999999999994e-05, "loss": 2.9817, "step": 61 }, { "epoch": 0.03, "learning_rate": 3.6599999999999995e-05, "loss": 2.9694, "step": 62 }, { "epoch": 0.04, "learning_rate": 3.7199999999999996e-05, "loss": 2.9556, "step": 63 }, { "epoch": 0.04, "learning_rate": 3.78e-05, "loss": 2.9413, "step": 64 }, { "epoch": 0.04, "learning_rate": 3.84e-05, "loss": 2.9505, "step": 65 }, { "epoch": 0.04, "learning_rate": 3.9e-05, "loss": 2.9523, "step": 66 }, { "epoch": 0.04, "learning_rate": 3.96e-05, "loss": 2.9224, "step": 67 }, { "epoch": 0.04, "learning_rate": 4.02e-05, "loss": 2.9289, "step": 68 }, { "epoch": 0.04, "learning_rate": 4.08e-05, "loss": 2.9396, "step": 69 }, { "epoch": 0.04, "learning_rate": 4.14e-05, "loss": 2.903, "step": 70 }, { "epoch": 0.04, "learning_rate": 4.2e-05, "loss": 2.9346, "step": 71 }, { "epoch": 0.04, "learning_rate": 4.259999999999999e-05, "loss": 2.9075, "step": 72 }, { "epoch": 0.04, "learning_rate": 4.319999999999999e-05, "loss": 2.906, "step": 73 }, { "epoch": 0.04, "learning_rate": 4.3799999999999994e-05, "loss": 2.8996, "step": 74 }, { "epoch": 0.04, "learning_rate": 4.4399999999999995e-05, "loss": 2.927, "step": 75 }, { "epoch": 0.04, "learning_rate": 4.4999999999999996e-05, "loss": 2.9048, "step": 76 }, { "epoch": 0.04, "learning_rate": 4.56e-05, "loss": 2.9225, "step": 77 }, { "epoch": 0.04, "learning_rate": 4.62e-05, "loss": 2.8971, "step": 78 }, { "epoch": 0.04, "learning_rate": 4.68e-05, "loss": 2.9187, "step": 79 }, { "epoch": 0.04, "learning_rate": 4.7399999999999993e-05, "loss": 2.8989, "step": 80 }, { "epoch": 0.05, "learning_rate": 4.7999999999999994e-05, "loss": 2.9046, "step": 81 }, { "epoch": 0.05, "learning_rate": 4.8599999999999995e-05, "loss": 2.8927, "step": 82 }, { "epoch": 0.05, "learning_rate": 4.9199999999999997e-05, "loss": 2.8798, "step": 83 }, { "epoch": 0.05, "learning_rate": 4.98e-05, "loss": 2.9027, "step": 84 }, { "epoch": 0.05, "learning_rate": 5.04e-05, "loss": 2.875, "step": 85 }, { "epoch": 0.05, "learning_rate": 5.1e-05, "loss": 2.8983, "step": 86 }, { "epoch": 0.05, "learning_rate": 5.1599999999999994e-05, "loss": 2.9012, "step": 87 }, { "epoch": 0.05, "learning_rate": 5.2199999999999995e-05, "loss": 2.88, "step": 88 }, { "epoch": 0.05, "learning_rate": 5.279999999999999e-05, "loss": 2.8736, "step": 89 }, { "epoch": 0.05, "learning_rate": 5.339999999999999e-05, "loss": 2.8714, "step": 90 }, { "epoch": 0.05, "learning_rate": 5.399999999999999e-05, "loss": 2.878, "step": 91 }, { "epoch": 0.05, "learning_rate": 5.459999999999999e-05, "loss": 2.8687, "step": 92 }, { "epoch": 0.05, "learning_rate": 5.519999999999999e-05, "loss": 2.8857, "step": 93 }, { "epoch": 0.05, "learning_rate": 5.5799999999999994e-05, "loss": 2.8951, "step": 94 }, { "epoch": 0.05, "learning_rate": 5.6399999999999995e-05, "loss": 2.8974, "step": 95 }, { "epoch": 0.05, "learning_rate": 5.6999999999999996e-05, "loss": 2.8892, "step": 96 }, { "epoch": 0.05, "learning_rate": 5.76e-05, "loss": 2.8919, "step": 97 }, { "epoch": 0.05, "learning_rate": 5.82e-05, "loss": 2.9137, "step": 98 }, { "epoch": 0.06, "learning_rate": 5.88e-05, "loss": 2.9416, "step": 99 }, { "epoch": 0.06, "learning_rate": 5.94e-05, "loss": 2.9585, "step": 100 }, { "epoch": 0.06, "learning_rate": 5.9999999999999995e-05, "loss": 2.8703, "step": 101 }, { "epoch": 0.06, "learning_rate": 6.0599999999999996e-05, "loss": 2.8721, "step": 102 }, { "epoch": 0.06, "learning_rate": 6.12e-05, "loss": 2.8477, "step": 103 }, { "epoch": 0.06, "learning_rate": 6.18e-05, "loss": 2.8837, "step": 104 }, { "epoch": 0.06, "learning_rate": 6.239999999999999e-05, "loss": 2.8575, "step": 105 }, { "epoch": 0.06, "learning_rate": 6.299999999999999e-05, "loss": 2.8769, "step": 106 }, { "epoch": 0.06, "learning_rate": 6.359999999999999e-05, "loss": 2.8903, "step": 107 }, { "epoch": 0.06, "learning_rate": 6.419999999999999e-05, "loss": 2.8798, "step": 108 }, { "epoch": 0.06, "learning_rate": 6.479999999999999e-05, "loss": 2.8609, "step": 109 }, { "epoch": 0.06, "learning_rate": 6.539999999999999e-05, "loss": 2.8632, "step": 110 }, { "epoch": 0.06, "learning_rate": 6.599999999999999e-05, "loss": 2.8585, "step": 111 }, { "epoch": 0.06, "learning_rate": 6.659999999999999e-05, "loss": 2.8533, "step": 112 }, { "epoch": 0.06, "learning_rate": 6.72e-05, "loss": 2.8666, "step": 113 }, { "epoch": 0.06, "learning_rate": 6.78e-05, "loss": 2.8536, "step": 114 }, { "epoch": 0.06, "learning_rate": 6.84e-05, "loss": 2.8567, "step": 115 }, { "epoch": 0.07, "learning_rate": 6.9e-05, "loss": 2.8534, "step": 116 }, { "epoch": 0.07, "learning_rate": 6.96e-05, "loss": 2.8689, "step": 117 }, { "epoch": 0.07, "learning_rate": 7.02e-05, "loss": 2.8416, "step": 118 }, { "epoch": 0.07, "learning_rate": 7.079999999999999e-05, "loss": 2.8515, "step": 119 }, { "epoch": 0.07, "learning_rate": 7.139999999999999e-05, "loss": 2.8649, "step": 120 }, { "epoch": 0.07, "learning_rate": 7.199999999999999e-05, "loss": 2.8756, "step": 121 }, { "epoch": 0.07, "learning_rate": 7.259999999999999e-05, "loss": 2.8578, "step": 122 }, { "epoch": 0.07, "learning_rate": 7.319999999999999e-05, "loss": 2.8636, "step": 123 }, { "epoch": 0.07, "learning_rate": 7.379999999999999e-05, "loss": 2.85, "step": 124 }, { "epoch": 0.07, "learning_rate": 7.439999999999999e-05, "loss": 2.8351, "step": 125 }, { "epoch": 0.07, "learning_rate": 7.5e-05, "loss": 2.8771, "step": 126 }, { "epoch": 0.07, "learning_rate": 7.56e-05, "loss": 2.8482, "step": 127 }, { "epoch": 0.07, "learning_rate": 7.62e-05, "loss": 2.8511, "step": 128 }, { "epoch": 0.07, "learning_rate": 7.68e-05, "loss": 2.8642, "step": 129 }, { "epoch": 0.07, "learning_rate": 7.74e-05, "loss": 2.835, "step": 130 }, { "epoch": 0.07, "learning_rate": 7.8e-05, "loss": 2.8549, "step": 131 }, { "epoch": 0.07, "learning_rate": 7.86e-05, "loss": 2.8562, "step": 132 }, { "epoch": 0.07, "learning_rate": 7.92e-05, "loss": 2.8217, "step": 133 }, { "epoch": 0.08, "learning_rate": 7.98e-05, "loss": 2.8431, "step": 134 }, { "epoch": 0.08, "learning_rate": 8.04e-05, "loss": 2.8402, "step": 135 }, { "epoch": 0.08, "learning_rate": 8.1e-05, "loss": 2.8696, "step": 136 }, { "epoch": 0.08, "learning_rate": 8.16e-05, "loss": 2.8718, "step": 137 }, { "epoch": 0.08, "learning_rate": 8.22e-05, "loss": 2.8515, "step": 138 }, { "epoch": 0.08, "learning_rate": 8.28e-05, "loss": 2.8496, "step": 139 }, { "epoch": 0.08, "learning_rate": 8.34e-05, "loss": 2.8522, "step": 140 }, { "epoch": 0.08, "learning_rate": 8.4e-05, "loss": 2.8909, "step": 141 }, { "epoch": 0.08, "learning_rate": 8.459999999999998e-05, "loss": 2.8543, "step": 142 }, { "epoch": 0.08, "learning_rate": 8.519999999999998e-05, "loss": 2.845, "step": 143 }, { "epoch": 0.08, "learning_rate": 8.579999999999998e-05, "loss": 2.8662, "step": 144 }, { "epoch": 0.08, "learning_rate": 8.639999999999999e-05, "loss": 2.8556, "step": 145 }, { "epoch": 0.08, "learning_rate": 8.699999999999999e-05, "loss": 2.8757, "step": 146 }, { "epoch": 0.08, "learning_rate": 8.759999999999999e-05, "loss": 2.8927, "step": 147 }, { "epoch": 0.08, "learning_rate": 8.819999999999999e-05, "loss": 2.8969, "step": 148 }, { "epoch": 0.08, "learning_rate": 8.879999999999999e-05, "loss": 2.9139, "step": 149 }, { "epoch": 0.08, "learning_rate": 8.939999999999999e-05, "loss": 2.88, "step": 150 }, { "epoch": 0.08, "learning_rate": 8.999999999999999e-05, "loss": 2.8476, "step": 151 }, { "epoch": 0.09, "learning_rate": 9.059999999999999e-05, "loss": 2.8483, "step": 152 }, { "epoch": 0.09, "learning_rate": 9.12e-05, "loss": 2.8331, "step": 153 }, { "epoch": 0.09, "learning_rate": 9.18e-05, "loss": 2.8499, "step": 154 }, { "epoch": 0.09, "learning_rate": 9.24e-05, "loss": 2.8482, "step": 155 }, { "epoch": 0.09, "learning_rate": 9.3e-05, "loss": 2.8501, "step": 156 }, { "epoch": 0.09, "learning_rate": 9.36e-05, "loss": 2.8426, "step": 157 }, { "epoch": 0.09, "learning_rate": 9.419999999999999e-05, "loss": 2.8689, "step": 158 }, { "epoch": 0.09, "learning_rate": 9.479999999999999e-05, "loss": 2.8503, "step": 159 }, { "epoch": 0.09, "learning_rate": 9.539999999999999e-05, "loss": 2.8657, "step": 160 }, { "epoch": 0.09, "learning_rate": 9.599999999999999e-05, "loss": 2.8583, "step": 161 }, { "epoch": 0.09, "learning_rate": 9.659999999999999e-05, "loss": 2.8485, "step": 162 }, { "epoch": 0.09, "learning_rate": 9.719999999999999e-05, "loss": 2.8454, "step": 163 }, { "epoch": 0.09, "learning_rate": 9.779999999999999e-05, "loss": 2.8582, "step": 164 }, { "epoch": 0.09, "learning_rate": 9.839999999999999e-05, "loss": 2.8465, "step": 165 }, { "epoch": 0.09, "learning_rate": 9.9e-05, "loss": 2.8329, "step": 166 }, { "epoch": 0.09, "learning_rate": 9.96e-05, "loss": 2.8466, "step": 167 }, { "epoch": 0.09, "learning_rate": 0.0001002, "loss": 2.8338, "step": 168 }, { "epoch": 0.09, "learning_rate": 0.0001008, "loss": 2.8338, "step": 169 }, { "epoch": 0.1, "learning_rate": 0.0001014, "loss": 2.857, "step": 170 }, { "epoch": 0.1, "learning_rate": 0.000102, "loss": 2.8688, "step": 171 }, { "epoch": 0.1, "learning_rate": 0.0001026, "loss": 2.8199, "step": 172 }, { "epoch": 0.1, "learning_rate": 0.00010319999999999999, "loss": 2.8485, "step": 173 }, { "epoch": 0.1, "learning_rate": 0.00010379999999999999, "loss": 2.837, "step": 174 }, { "epoch": 0.1, "learning_rate": 0.00010439999999999999, "loss": 2.8405, "step": 175 }, { "epoch": 0.1, "learning_rate": 0.00010499999999999999, "loss": 2.833, "step": 176 }, { "epoch": 0.1, "learning_rate": 0.00010559999999999998, "loss": 2.8276, "step": 177 }, { "epoch": 0.1, "learning_rate": 0.00010619999999999998, "loss": 2.8735, "step": 178 }, { "epoch": 0.1, "learning_rate": 0.00010679999999999998, "loss": 2.839, "step": 179 }, { "epoch": 0.1, "learning_rate": 0.00010739999999999998, "loss": 2.8523, "step": 180 }, { "epoch": 0.1, "learning_rate": 0.00010799999999999998, "loss": 2.8358, "step": 181 }, { "epoch": 0.1, "learning_rate": 0.00010859999999999998, "loss": 2.8331, "step": 182 }, { "epoch": 0.1, "learning_rate": 0.00010919999999999998, "loss": 2.8428, "step": 183 }, { "epoch": 0.1, "learning_rate": 0.00010979999999999999, "loss": 2.8361, "step": 184 }, { "epoch": 0.1, "learning_rate": 0.00011039999999999999, "loss": 2.8447, "step": 185 }, { "epoch": 0.1, "learning_rate": 0.00011099999999999999, "loss": 2.8355, "step": 186 }, { "epoch": 0.1, "learning_rate": 0.00011159999999999999, "loss": 2.8265, "step": 187 }, { "epoch": 0.11, "learning_rate": 0.00011219999999999999, "loss": 2.8336, "step": 188 }, { "epoch": 0.11, "learning_rate": 0.00011279999999999999, "loss": 2.8416, "step": 189 }, { "epoch": 0.11, "learning_rate": 0.00011339999999999999, "loss": 2.8327, "step": 190 }, { "epoch": 0.11, "learning_rate": 0.00011399999999999999, "loss": 2.8459, "step": 191 }, { "epoch": 0.11, "learning_rate": 0.0001146, "loss": 2.8251, "step": 192 }, { "epoch": 0.11, "learning_rate": 0.0001152, "loss": 2.8309, "step": 193 }, { "epoch": 0.11, "learning_rate": 0.0001158, "loss": 2.8608, "step": 194 }, { "epoch": 0.11, "learning_rate": 0.0001164, "loss": 2.8405, "step": 195 }, { "epoch": 0.11, "learning_rate": 0.000117, "loss": 2.8846, "step": 196 }, { "epoch": 0.11, "learning_rate": 0.0001176, "loss": 2.8642, "step": 197 }, { "epoch": 0.11, "learning_rate": 0.0001182, "loss": 2.8596, "step": 198 }, { "epoch": 0.11, "learning_rate": 0.0001188, "loss": 2.9025, "step": 199 }, { "epoch": 0.11, "learning_rate": 0.0001194, "loss": 2.8799, "step": 200 }, { "epoch": 0.11, "learning_rate": 0.00011999999999999999, "loss": 2.8474, "step": 201 }, { "epoch": 0.11, "learning_rate": 0.00012059999999999999, "loss": 2.8313, "step": 202 }, { "epoch": 0.11, "learning_rate": 0.00012119999999999999, "loss": 2.845, "step": 203 }, { "epoch": 0.11, "learning_rate": 0.00012179999999999999, "loss": 2.8127, "step": 204 }, { "epoch": 0.11, "learning_rate": 0.0001224, "loss": 2.8425, "step": 205 }, { "epoch": 0.12, "learning_rate": 0.00012299999999999998, "loss": 2.8447, "step": 206 }, { "epoch": 0.12, "learning_rate": 0.0001236, "loss": 2.8423, "step": 207 }, { "epoch": 0.12, "learning_rate": 0.00012419999999999998, "loss": 2.8432, "step": 208 }, { "epoch": 0.12, "learning_rate": 0.00012479999999999997, "loss": 2.8422, "step": 209 }, { "epoch": 0.12, "learning_rate": 0.00012539999999999999, "loss": 2.8687, "step": 210 }, { "epoch": 0.12, "learning_rate": 0.00012599999999999997, "loss": 2.8257, "step": 211 }, { "epoch": 0.12, "learning_rate": 0.0001266, "loss": 2.8164, "step": 212 }, { "epoch": 0.12, "learning_rate": 0.00012719999999999997, "loss": 2.8235, "step": 213 }, { "epoch": 0.12, "learning_rate": 0.0001278, "loss": 2.817, "step": 214 }, { "epoch": 0.12, "learning_rate": 0.00012839999999999998, "loss": 2.8365, "step": 215 }, { "epoch": 0.12, "learning_rate": 0.000129, "loss": 2.8313, "step": 216 }, { "epoch": 0.12, "learning_rate": 0.00012959999999999998, "loss": 2.8317, "step": 217 }, { "epoch": 0.12, "learning_rate": 0.0001302, "loss": 2.8452, "step": 218 }, { "epoch": 0.12, "learning_rate": 0.00013079999999999998, "loss": 2.8352, "step": 219 }, { "epoch": 0.12, "learning_rate": 0.0001314, "loss": 2.8425, "step": 220 }, { "epoch": 0.12, "learning_rate": 0.00013199999999999998, "loss": 2.8473, "step": 221 }, { "epoch": 0.12, "learning_rate": 0.0001326, "loss": 2.8539, "step": 222 }, { "epoch": 0.12, "learning_rate": 0.00013319999999999999, "loss": 2.8363, "step": 223 }, { "epoch": 0.13, "learning_rate": 0.0001338, "loss": 2.8587, "step": 224 }, { "epoch": 0.13, "learning_rate": 0.0001344, "loss": 2.8581, "step": 225 }, { "epoch": 0.13, "learning_rate": 0.000135, "loss": 2.8461, "step": 226 }, { "epoch": 0.13, "learning_rate": 0.0001356, "loss": 2.8715, "step": 227 }, { "epoch": 0.13, "learning_rate": 0.0001362, "loss": 2.8478, "step": 228 }, { "epoch": 0.13, "learning_rate": 0.0001368, "loss": 2.8397, "step": 229 }, { "epoch": 0.13, "learning_rate": 0.0001374, "loss": 2.8272, "step": 230 }, { "epoch": 0.13, "learning_rate": 0.000138, "loss": 2.8311, "step": 231 }, { "epoch": 0.13, "learning_rate": 0.0001386, "loss": 2.8436, "step": 232 }, { "epoch": 0.13, "learning_rate": 0.0001392, "loss": 2.8538, "step": 233 }, { "epoch": 0.13, "learning_rate": 0.00013979999999999998, "loss": 2.832, "step": 234 }, { "epoch": 0.13, "learning_rate": 0.0001404, "loss": 2.8392, "step": 235 }, { "epoch": 0.13, "learning_rate": 0.00014099999999999998, "loss": 2.8644, "step": 236 }, { "epoch": 0.13, "learning_rate": 0.00014159999999999997, "loss": 2.8533, "step": 237 }, { "epoch": 0.13, "learning_rate": 0.0001422, "loss": 2.8347, "step": 238 }, { "epoch": 0.13, "learning_rate": 0.00014279999999999997, "loss": 2.8443, "step": 239 }, { "epoch": 0.13, "learning_rate": 0.0001434, "loss": 2.8389, "step": 240 }, { "epoch": 0.14, "learning_rate": 0.00014399999999999998, "loss": 2.8599, "step": 241 }, { "epoch": 0.14, "learning_rate": 0.0001446, "loss": 2.8439, "step": 242 }, { "epoch": 0.14, "learning_rate": 0.00014519999999999998, "loss": 2.8538, "step": 243 }, { "epoch": 0.14, "learning_rate": 0.0001458, "loss": 2.8571, "step": 244 }, { "epoch": 0.14, "learning_rate": 0.00014639999999999998, "loss": 2.8643, "step": 245 }, { "epoch": 0.14, "learning_rate": 0.000147, "loss": 2.8505, "step": 246 }, { "epoch": 0.14, "learning_rate": 0.00014759999999999998, "loss": 2.8529, "step": 247 }, { "epoch": 0.14, "learning_rate": 0.0001482, "loss": 2.8767, "step": 248 }, { "epoch": 0.14, "learning_rate": 0.00014879999999999998, "loss": 2.8678, "step": 249 }, { "epoch": 0.14, "learning_rate": 0.0001494, "loss": 2.917, "step": 250 }, { "epoch": 0.14, "learning_rate": 0.00015, "loss": 2.8354, "step": 251 }, { "epoch": 0.14, "learning_rate": 0.00015059999999999997, "loss": 2.8371, "step": 252 }, { "epoch": 0.14, "learning_rate": 0.0001512, "loss": 2.8482, "step": 253 }, { "epoch": 0.14, "learning_rate": 0.00015179999999999998, "loss": 2.8431, "step": 254 }, { "epoch": 0.14, "learning_rate": 0.0001524, "loss": 2.8402, "step": 255 }, { "epoch": 0.14, "learning_rate": 0.00015299999999999998, "loss": 2.8375, "step": 256 }, { "epoch": 0.14, "learning_rate": 0.0001536, "loss": 2.8247, "step": 257 }, { "epoch": 0.14, "learning_rate": 0.00015419999999999998, "loss": 2.8333, "step": 258 }, { "epoch": 0.15, "learning_rate": 0.0001548, "loss": 2.8496, "step": 259 }, { "epoch": 0.15, "learning_rate": 0.00015539999999999998, "loss": 2.8194, "step": 260 }, { "epoch": 0.15, "learning_rate": 0.000156, "loss": 2.8143, "step": 261 }, { "epoch": 0.15, "learning_rate": 0.00015659999999999998, "loss": 2.832, "step": 262 }, { "epoch": 0.15, "learning_rate": 0.0001572, "loss": 2.8484, "step": 263 }, { "epoch": 0.15, "learning_rate": 0.0001578, "loss": 2.8402, "step": 264 }, { "epoch": 0.15, "learning_rate": 0.0001584, "loss": 2.8321, "step": 265 }, { "epoch": 0.15, "learning_rate": 0.000159, "loss": 2.8285, "step": 266 }, { "epoch": 0.15, "learning_rate": 0.0001596, "loss": 2.8635, "step": 267 }, { "epoch": 0.15, "learning_rate": 0.0001602, "loss": 2.8371, "step": 268 }, { "epoch": 0.15, "learning_rate": 0.0001608, "loss": 2.8254, "step": 269 }, { "epoch": 0.15, "learning_rate": 0.0001614, "loss": 2.8175, "step": 270 }, { "epoch": 0.15, "learning_rate": 0.000162, "loss": 2.8204, "step": 271 }, { "epoch": 0.15, "learning_rate": 0.0001626, "loss": 2.8316, "step": 272 }, { "epoch": 0.15, "learning_rate": 0.0001632, "loss": 2.832, "step": 273 }, { "epoch": 0.15, "learning_rate": 0.0001638, "loss": 2.8301, "step": 274 }, { "epoch": 0.15, "learning_rate": 0.0001644, "loss": 2.8486, "step": 275 }, { "epoch": 0.15, "learning_rate": 0.000165, "loss": 2.8344, "step": 276 }, { "epoch": 0.16, "learning_rate": 0.0001656, "loss": 2.8457, "step": 277 }, { "epoch": 0.16, "learning_rate": 0.0001662, "loss": 2.8511, "step": 278 }, { "epoch": 0.16, "learning_rate": 0.0001668, "loss": 2.8364, "step": 279 }, { "epoch": 0.16, "learning_rate": 0.0001674, "loss": 2.8296, "step": 280 }, { "epoch": 0.16, "learning_rate": 0.000168, "loss": 2.8471, "step": 281 }, { "epoch": 0.16, "learning_rate": 0.0001686, "loss": 2.8431, "step": 282 }, { "epoch": 0.16, "learning_rate": 0.00016919999999999997, "loss": 2.8291, "step": 283 }, { "epoch": 0.16, "learning_rate": 0.00016979999999999998, "loss": 2.8335, "step": 284 }, { "epoch": 0.16, "learning_rate": 0.00017039999999999997, "loss": 2.8266, "step": 285 }, { "epoch": 0.16, "learning_rate": 0.00017099999999999998, "loss": 2.8349, "step": 286 }, { "epoch": 0.16, "learning_rate": 0.00017159999999999997, "loss": 2.8442, "step": 287 }, { "epoch": 0.16, "learning_rate": 0.00017219999999999998, "loss": 2.8355, "step": 288 }, { "epoch": 0.16, "learning_rate": 0.00017279999999999997, "loss": 2.8266, "step": 289 }, { "epoch": 0.16, "learning_rate": 0.00017339999999999996, "loss": 2.8386, "step": 290 }, { "epoch": 0.16, "learning_rate": 0.00017399999999999997, "loss": 2.8245, "step": 291 }, { "epoch": 0.16, "learning_rate": 0.00017459999999999996, "loss": 2.8394, "step": 292 }, { "epoch": 0.16, "learning_rate": 0.00017519999999999998, "loss": 2.8482, "step": 293 }, { "epoch": 0.16, "learning_rate": 0.00017579999999999996, "loss": 2.8274, "step": 294 }, { "epoch": 0.17, "learning_rate": 0.00017639999999999998, "loss": 2.836, "step": 295 }, { "epoch": 0.17, "learning_rate": 0.00017699999999999997, "loss": 2.8375, "step": 296 }, { "epoch": 0.17, "learning_rate": 0.00017759999999999998, "loss": 2.8221, "step": 297 }, { "epoch": 0.17, "learning_rate": 0.00017819999999999997, "loss": 2.8713, "step": 298 }, { "epoch": 0.17, "learning_rate": 0.00017879999999999998, "loss": 2.8454, "step": 299 }, { "epoch": 0.17, "learning_rate": 0.00017939999999999997, "loss": 2.8664, "step": 300 }, { "epoch": 0.17, "eval_loss": 2.8438608646392822, "eval_runtime": 270.317, "eval_samples_per_second": 9.774, "eval_steps_per_second": 0.614, "eval_wer": 1.0, "step": 300 }, { "epoch": 0.17, "learning_rate": 0.00017999999999999998, "loss": 2.8232, "step": 301 }, { "epoch": 0.17, "learning_rate": 0.00018059999999999997, "loss": 2.8037, "step": 302 }, { "epoch": 0.17, "learning_rate": 0.00018119999999999999, "loss": 2.8107, "step": 303 }, { "epoch": 0.17, "learning_rate": 0.00018179999999999997, "loss": 2.7888, "step": 304 }, { "epoch": 0.17, "learning_rate": 0.0001824, "loss": 2.8045, "step": 305 }, { "epoch": 0.17, "learning_rate": 0.00018299999999999998, "loss": 2.83, "step": 306 }, { "epoch": 0.17, "learning_rate": 0.0001836, "loss": 2.8206, "step": 307 }, { "epoch": 0.17, "learning_rate": 0.00018419999999999998, "loss": 2.8199, "step": 308 }, { "epoch": 0.17, "learning_rate": 0.0001848, "loss": 2.7871, "step": 309 }, { "epoch": 0.17, "learning_rate": 0.00018539999999999998, "loss": 2.7388, "step": 310 }, { "epoch": 0.17, "learning_rate": 0.000186, "loss": 2.7397, "step": 311 }, { "epoch": 0.17, "learning_rate": 0.00018659999999999998, "loss": 2.7277, "step": 312 }, { "epoch": 0.18, "learning_rate": 0.0001872, "loss": 2.6869, "step": 313 }, { "epoch": 0.18, "learning_rate": 0.00018779999999999998, "loss": 2.6928, "step": 314 }, { "epoch": 0.18, "learning_rate": 0.00018839999999999997, "loss": 2.6591, "step": 315 }, { "epoch": 0.18, "learning_rate": 0.00018899999999999999, "loss": 2.6356, "step": 316 }, { "epoch": 0.18, "learning_rate": 0.00018959999999999997, "loss": 2.5756, "step": 317 }, { "epoch": 0.18, "learning_rate": 0.0001902, "loss": 2.574, "step": 318 }, { "epoch": 0.18, "learning_rate": 0.00019079999999999998, "loss": 2.5523, "step": 319 }, { "epoch": 0.18, "learning_rate": 0.0001914, "loss": 2.4891, "step": 320 }, { "epoch": 0.18, "learning_rate": 0.00019199999999999998, "loss": 2.5411, "step": 321 }, { "epoch": 0.18, "learning_rate": 0.0001926, "loss": 2.431, "step": 322 }, { "epoch": 0.18, "learning_rate": 0.00019319999999999998, "loss": 2.3815, "step": 323 }, { "epoch": 0.18, "learning_rate": 0.0001938, "loss": 2.3422, "step": 324 }, { "epoch": 0.18, "learning_rate": 0.00019439999999999998, "loss": 2.2644, "step": 325 }, { "epoch": 0.18, "learning_rate": 0.000195, "loss": 2.2154, "step": 326 }, { "epoch": 0.18, "learning_rate": 0.00019559999999999998, "loss": 2.1733, "step": 327 }, { "epoch": 0.18, "learning_rate": 0.0001962, "loss": 2.154, "step": 328 }, { "epoch": 0.18, "learning_rate": 0.00019679999999999999, "loss": 2.0589, "step": 329 }, { "epoch": 0.18, "learning_rate": 0.0001974, "loss": 2.0108, "step": 330 }, { "epoch": 0.19, "learning_rate": 0.000198, "loss": 1.9365, "step": 331 }, { "epoch": 0.19, "learning_rate": 0.0001986, "loss": 1.8997, "step": 332 }, { "epoch": 0.19, "learning_rate": 0.0001992, "loss": 1.8173, "step": 333 }, { "epoch": 0.19, "learning_rate": 0.0001998, "loss": 1.7661, "step": 334 }, { "epoch": 0.19, "learning_rate": 0.0002004, "loss": 1.6816, "step": 335 }, { "epoch": 0.19, "learning_rate": 0.000201, "loss": 1.6392, "step": 336 }, { "epoch": 0.19, "learning_rate": 0.0002016, "loss": 1.6353, "step": 337 }, { "epoch": 0.19, "learning_rate": 0.0002022, "loss": 1.4882, "step": 338 }, { "epoch": 0.19, "learning_rate": 0.0002028, "loss": 1.4628, "step": 339 }, { "epoch": 0.19, "learning_rate": 0.00020339999999999998, "loss": 1.4176, "step": 340 }, { "epoch": 0.19, "learning_rate": 0.000204, "loss": 1.3511, "step": 341 }, { "epoch": 0.19, "learning_rate": 0.00020459999999999999, "loss": 1.3456, "step": 342 }, { "epoch": 0.19, "learning_rate": 0.0002052, "loss": 1.3598, "step": 343 }, { "epoch": 0.19, "learning_rate": 0.0002058, "loss": 1.2623, "step": 344 }, { "epoch": 0.19, "learning_rate": 0.00020639999999999998, "loss": 1.1237, "step": 345 }, { "epoch": 0.19, "learning_rate": 0.00020699999999999996, "loss": 1.2213, "step": 346 }, { "epoch": 0.19, "learning_rate": 0.00020759999999999998, "loss": 1.3576, "step": 347 }, { "epoch": 0.2, "learning_rate": 0.00020819999999999996, "loss": 1.2658, "step": 348 }, { "epoch": 0.2, "learning_rate": 0.00020879999999999998, "loss": 1.1761, "step": 349 }, { "epoch": 0.2, "learning_rate": 0.00020939999999999997, "loss": 1.2569, "step": 350 }, { "epoch": 0.2, "learning_rate": 0.00020999999999999998, "loss": 1.1303, "step": 351 }, { "epoch": 0.2, "learning_rate": 0.00021059999999999997, "loss": 1.0442, "step": 352 }, { "epoch": 0.2, "learning_rate": 0.00021119999999999996, "loss": 1.0645, "step": 353 }, { "epoch": 0.2, "learning_rate": 0.00021179999999999997, "loss": 0.9704, "step": 354 }, { "epoch": 0.2, "learning_rate": 0.00021239999999999996, "loss": 0.9132, "step": 355 }, { "epoch": 0.2, "learning_rate": 0.00021299999999999997, "loss": 0.9064, "step": 356 }, { "epoch": 0.2, "learning_rate": 0.00021359999999999996, "loss": 0.9122, "step": 357 }, { "epoch": 0.2, "learning_rate": 0.00021419999999999998, "loss": 0.8354, "step": 358 }, { "epoch": 0.2, "learning_rate": 0.00021479999999999996, "loss": 0.9233, "step": 359 }, { "epoch": 0.2, "learning_rate": 0.00021539999999999998, "loss": 0.8359, "step": 360 }, { "epoch": 0.2, "learning_rate": 0.00021599999999999996, "loss": 0.828, "step": 361 }, { "epoch": 0.2, "learning_rate": 0.00021659999999999998, "loss": 0.8157, "step": 362 }, { "epoch": 0.2, "learning_rate": 0.00021719999999999997, "loss": 0.8129, "step": 363 }, { "epoch": 0.2, "learning_rate": 0.00021779999999999998, "loss": 0.7811, "step": 364 }, { "epoch": 0.2, "learning_rate": 0.00021839999999999997, "loss": 0.6942, "step": 365 }, { "epoch": 0.21, "learning_rate": 0.00021899999999999998, "loss": 0.7684, "step": 366 }, { "epoch": 0.21, "learning_rate": 0.00021959999999999997, "loss": 0.6955, "step": 367 }, { "epoch": 0.21, "learning_rate": 0.00022019999999999999, "loss": 0.7254, "step": 368 }, { "epoch": 0.21, "learning_rate": 0.00022079999999999997, "loss": 0.6773, "step": 369 }, { "epoch": 0.21, "learning_rate": 0.0002214, "loss": 0.6325, "step": 370 }, { "epoch": 0.21, "learning_rate": 0.00022199999999999998, "loss": 0.6525, "step": 371 }, { "epoch": 0.21, "learning_rate": 0.0002226, "loss": 0.6926, "step": 372 }, { "epoch": 0.21, "learning_rate": 0.00022319999999999998, "loss": 0.6075, "step": 373 }, { "epoch": 0.21, "learning_rate": 0.0002238, "loss": 0.6703, "step": 374 }, { "epoch": 0.21, "learning_rate": 0.00022439999999999998, "loss": 0.6126, "step": 375 }, { "epoch": 0.21, "learning_rate": 0.000225, "loss": 0.6206, "step": 376 }, { "epoch": 0.21, "learning_rate": 0.00022559999999999998, "loss": 0.5673, "step": 377 }, { "epoch": 0.21, "learning_rate": 0.00022619999999999997, "loss": 0.5591, "step": 378 }, { "epoch": 0.21, "learning_rate": 0.00022679999999999998, "loss": 0.5877, "step": 379 }, { "epoch": 0.21, "learning_rate": 0.00022739999999999997, "loss": 0.593, "step": 380 }, { "epoch": 0.21, "learning_rate": 0.00022799999999999999, "loss": 0.5084, "step": 381 }, { "epoch": 0.21, "learning_rate": 0.00022859999999999997, "loss": 0.4967, "step": 382 }, { "epoch": 0.21, "learning_rate": 0.0002292, "loss": 0.5684, "step": 383 }, { "epoch": 0.22, "learning_rate": 0.00022979999999999997, "loss": 0.5225, "step": 384 }, { "epoch": 0.22, "learning_rate": 0.0002304, "loss": 0.5312, "step": 385 }, { "epoch": 0.22, "learning_rate": 0.00023099999999999998, "loss": 0.5945, "step": 386 }, { "epoch": 0.22, "learning_rate": 0.0002316, "loss": 0.4996, "step": 387 }, { "epoch": 0.22, "learning_rate": 0.00023219999999999998, "loss": 0.4608, "step": 388 }, { "epoch": 0.22, "learning_rate": 0.0002328, "loss": 0.447, "step": 389 }, { "epoch": 0.22, "learning_rate": 0.00023339999999999998, "loss": 0.5573, "step": 390 }, { "epoch": 0.22, "learning_rate": 0.000234, "loss": 0.4697, "step": 391 }, { "epoch": 0.22, "learning_rate": 0.00023459999999999998, "loss": 0.5904, "step": 392 }, { "epoch": 0.22, "learning_rate": 0.0002352, "loss": 0.5688, "step": 393 }, { "epoch": 0.22, "learning_rate": 0.00023579999999999999, "loss": 0.5212, "step": 394 }, { "epoch": 0.22, "learning_rate": 0.0002364, "loss": 0.6459, "step": 395 }, { "epoch": 0.22, "learning_rate": 0.000237, "loss": 0.5103, "step": 396 }, { "epoch": 0.22, "learning_rate": 0.0002376, "loss": 0.651, "step": 397 }, { "epoch": 0.22, "learning_rate": 0.0002382, "loss": 0.6448, "step": 398 }, { "epoch": 0.22, "learning_rate": 0.0002388, "loss": 0.9588, "step": 399 }, { "epoch": 0.22, "learning_rate": 0.0002394, "loss": 0.8209, "step": 400 }, { "epoch": 0.22, "learning_rate": 0.00023999999999999998, "loss": 0.5767, "step": 401 }, { "epoch": 0.23, "learning_rate": 0.0002406, "loss": 0.5286, "step": 402 }, { "epoch": 0.23, "learning_rate": 0.00024119999999999998, "loss": 0.4224, "step": 403 }, { "epoch": 0.23, "learning_rate": 0.0002418, "loss": 0.4969, "step": 404 }, { "epoch": 0.23, "learning_rate": 0.00024239999999999998, "loss": 0.3959, "step": 405 }, { "epoch": 0.23, "learning_rate": 0.000243, "loss": 0.4425, "step": 406 }, { "epoch": 0.23, "learning_rate": 0.00024359999999999999, "loss": 0.3984, "step": 407 }, { "epoch": 0.23, "learning_rate": 0.00024419999999999997, "loss": 0.4048, "step": 408 }, { "epoch": 0.23, "learning_rate": 0.0002448, "loss": 0.4521, "step": 409 }, { "epoch": 0.23, "learning_rate": 0.00024539999999999995, "loss": 0.4126, "step": 410 }, { "epoch": 0.23, "learning_rate": 0.00024599999999999996, "loss": 0.3925, "step": 411 }, { "epoch": 0.23, "learning_rate": 0.0002466, "loss": 0.4517, "step": 412 }, { "epoch": 0.23, "learning_rate": 0.0002472, "loss": 0.406, "step": 413 }, { "epoch": 0.23, "learning_rate": 0.00024779999999999995, "loss": 0.4112, "step": 414 }, { "epoch": 0.23, "learning_rate": 0.00024839999999999997, "loss": 0.392, "step": 415 }, { "epoch": 0.23, "learning_rate": 0.000249, "loss": 0.405, "step": 416 }, { "epoch": 0.23, "learning_rate": 0.00024959999999999994, "loss": 0.3648, "step": 417 }, { "epoch": 0.23, "learning_rate": 0.00025019999999999996, "loss": 0.366, "step": 418 }, { "epoch": 0.23, "learning_rate": 0.00025079999999999997, "loss": 0.355, "step": 419 }, { "epoch": 0.24, "learning_rate": 0.0002514, "loss": 0.3887, "step": 420 }, { "epoch": 0.24, "learning_rate": 0.00025199999999999995, "loss": 0.3729, "step": 421 }, { "epoch": 0.24, "learning_rate": 0.00025259999999999996, "loss": 0.3481, "step": 422 }, { "epoch": 0.24, "learning_rate": 0.0002532, "loss": 0.34, "step": 423 }, { "epoch": 0.24, "learning_rate": 0.0002538, "loss": 0.3587, "step": 424 }, { "epoch": 0.24, "learning_rate": 0.00025439999999999995, "loss": 0.4206, "step": 425 }, { "epoch": 0.24, "learning_rate": 0.00025499999999999996, "loss": 0.3684, "step": 426 }, { "epoch": 0.24, "learning_rate": 0.0002556, "loss": 0.3518, "step": 427 }, { "epoch": 0.24, "learning_rate": 0.0002562, "loss": 0.319, "step": 428 }, { "epoch": 0.24, "learning_rate": 0.00025679999999999995, "loss": 0.3958, "step": 429 }, { "epoch": 0.24, "learning_rate": 0.00025739999999999997, "loss": 0.3346, "step": 430 }, { "epoch": 0.24, "learning_rate": 0.000258, "loss": 0.3694, "step": 431 }, { "epoch": 0.24, "learning_rate": 0.0002586, "loss": 0.3041, "step": 432 }, { "epoch": 0.24, "learning_rate": 0.00025919999999999996, "loss": 0.2968, "step": 433 }, { "epoch": 0.24, "learning_rate": 0.00025979999999999997, "loss": 0.33, "step": 434 }, { "epoch": 0.24, "learning_rate": 0.0002604, "loss": 0.3416, "step": 435 }, { "epoch": 0.24, "learning_rate": 0.000261, "loss": 0.2966, "step": 436 }, { "epoch": 0.24, "learning_rate": 0.00026159999999999996, "loss": 0.4014, "step": 437 }, { "epoch": 0.25, "learning_rate": 0.0002622, "loss": 0.3433, "step": 438 }, { "epoch": 0.25, "learning_rate": 0.0002628, "loss": 0.3556, "step": 439 }, { "epoch": 0.25, "learning_rate": 0.00026339999999999995, "loss": 0.356, "step": 440 }, { "epoch": 0.25, "learning_rate": 0.00026399999999999997, "loss": 0.3989, "step": 441 }, { "epoch": 0.25, "learning_rate": 0.0002646, "loss": 0.3233, "step": 442 }, { "epoch": 0.25, "learning_rate": 0.0002652, "loss": 0.3714, "step": 443 }, { "epoch": 0.25, "learning_rate": 0.00026579999999999996, "loss": 0.2637, "step": 444 }, { "epoch": 0.25, "learning_rate": 0.00026639999999999997, "loss": 0.3756, "step": 445 }, { "epoch": 0.25, "learning_rate": 0.000267, "loss": 0.4774, "step": 446 }, { "epoch": 0.25, "learning_rate": 0.0002676, "loss": 0.5429, "step": 447 }, { "epoch": 0.25, "learning_rate": 0.00026819999999999996, "loss": 0.4754, "step": 448 }, { "epoch": 0.25, "learning_rate": 0.0002688, "loss": 0.5277, "step": 449 }, { "epoch": 0.25, "learning_rate": 0.0002694, "loss": 0.4841, "step": 450 }, { "epoch": 0.25, "learning_rate": 0.00027, "loss": 0.3522, "step": 451 }, { "epoch": 0.25, "learning_rate": 0.00027059999999999996, "loss": 0.35, "step": 452 }, { "epoch": 0.25, "learning_rate": 0.0002712, "loss": 0.3277, "step": 453 }, { "epoch": 0.25, "learning_rate": 0.0002718, "loss": 0.2986, "step": 454 }, { "epoch": 0.26, "learning_rate": 0.0002724, "loss": 0.3387, "step": 455 }, { "epoch": 0.26, "learning_rate": 0.00027299999999999997, "loss": 0.3342, "step": 456 }, { "epoch": 0.26, "learning_rate": 0.0002736, "loss": 0.3751, "step": 457 }, { "epoch": 0.26, "learning_rate": 0.0002742, "loss": 0.3029, "step": 458 }, { "epoch": 0.26, "learning_rate": 0.0002748, "loss": 0.2693, "step": 459 }, { "epoch": 0.26, "learning_rate": 0.00027539999999999997, "loss": 0.3257, "step": 460 }, { "epoch": 0.26, "learning_rate": 0.000276, "loss": 0.2682, "step": 461 }, { "epoch": 0.26, "learning_rate": 0.0002766, "loss": 0.3072, "step": 462 }, { "epoch": 0.26, "learning_rate": 0.0002772, "loss": 0.3104, "step": 463 }, { "epoch": 0.26, "learning_rate": 0.0002778, "loss": 0.336, "step": 464 }, { "epoch": 0.26, "learning_rate": 0.0002784, "loss": 0.2431, "step": 465 }, { "epoch": 0.26, "learning_rate": 0.000279, "loss": 0.3127, "step": 466 }, { "epoch": 0.26, "learning_rate": 0.00027959999999999997, "loss": 0.2735, "step": 467 }, { "epoch": 0.26, "learning_rate": 0.0002802, "loss": 0.3, "step": 468 }, { "epoch": 0.26, "learning_rate": 0.0002808, "loss": 0.2681, "step": 469 }, { "epoch": 0.26, "learning_rate": 0.00028139999999999996, "loss": 0.3238, "step": 470 }, { "epoch": 0.26, "learning_rate": 0.00028199999999999997, "loss": 0.284, "step": 471 }, { "epoch": 0.26, "learning_rate": 0.0002826, "loss": 0.2585, "step": 472 }, { "epoch": 0.27, "learning_rate": 0.00028319999999999994, "loss": 0.2473, "step": 473 }, { "epoch": 0.27, "learning_rate": 0.00028379999999999996, "loss": 0.3014, "step": 474 }, { "epoch": 0.27, "learning_rate": 0.0002844, "loss": 0.2269, "step": 475 }, { "epoch": 0.27, "learning_rate": 0.000285, "loss": 0.2754, "step": 476 }, { "epoch": 0.27, "learning_rate": 0.00028559999999999995, "loss": 0.2439, "step": 477 }, { "epoch": 0.27, "learning_rate": 0.00028619999999999996, "loss": 0.2529, "step": 478 }, { "epoch": 0.27, "learning_rate": 0.0002868, "loss": 0.2533, "step": 479 }, { "epoch": 0.27, "learning_rate": 0.00028739999999999994, "loss": 0.3259, "step": 480 }, { "epoch": 0.27, "learning_rate": 0.00028799999999999995, "loss": 0.2961, "step": 481 }, { "epoch": 0.27, "learning_rate": 0.00028859999999999997, "loss": 0.2768, "step": 482 }, { "epoch": 0.27, "learning_rate": 0.0002892, "loss": 0.3291, "step": 483 }, { "epoch": 0.27, "learning_rate": 0.00028979999999999994, "loss": 0.2298, "step": 484 }, { "epoch": 0.27, "learning_rate": 0.00029039999999999996, "loss": 0.2599, "step": 485 }, { "epoch": 0.27, "learning_rate": 0.00029099999999999997, "loss": 0.2474, "step": 486 }, { "epoch": 0.27, "learning_rate": 0.0002916, "loss": 0.2072, "step": 487 }, { "epoch": 0.27, "learning_rate": 0.00029219999999999995, "loss": 0.271, "step": 488 }, { "epoch": 0.27, "learning_rate": 0.00029279999999999996, "loss": 0.3078, "step": 489 }, { "epoch": 0.27, "learning_rate": 0.0002934, "loss": 0.2259, "step": 490 }, { "epoch": 0.28, "learning_rate": 0.000294, "loss": 0.27, "step": 491 }, { "epoch": 0.28, "learning_rate": 0.00029459999999999995, "loss": 0.3388, "step": 492 }, { "epoch": 0.28, "learning_rate": 0.00029519999999999997, "loss": 0.249, "step": 493 }, { "epoch": 0.28, "learning_rate": 0.0002958, "loss": 0.3002, "step": 494 }, { "epoch": 0.28, "learning_rate": 0.0002964, "loss": 0.2501, "step": 495 }, { "epoch": 0.28, "learning_rate": 0.00029699999999999996, "loss": 0.2995, "step": 496 }, { "epoch": 0.28, "learning_rate": 0.00029759999999999997, "loss": 0.1916, "step": 497 }, { "epoch": 0.28, "learning_rate": 0.0002982, "loss": 0.4367, "step": 498 }, { "epoch": 0.28, "learning_rate": 0.0002988, "loss": 0.4111, "step": 499 }, { "epoch": 0.28, "learning_rate": 0.00029939999999999996, "loss": 0.8271, "step": 500 }, { "epoch": 0.28, "learning_rate": 0.0003, "loss": 0.2911, "step": 501 }, { "epoch": 0.28, "learning_rate": 0.0002999381698268755, "loss": 0.3028, "step": 502 }, { "epoch": 0.28, "learning_rate": 0.000299876339653751, "loss": 0.2886, "step": 503 }, { "epoch": 0.28, "learning_rate": 0.0002998145094806265, "loss": 0.2963, "step": 504 }, { "epoch": 0.28, "learning_rate": 0.00029975267930750204, "loss": 0.2973, "step": 505 }, { "epoch": 0.28, "learning_rate": 0.00029969084913437753, "loss": 0.2763, "step": 506 }, { "epoch": 0.28, "learning_rate": 0.0002996290189612531, "loss": 0.2558, "step": 507 }, { "epoch": 0.28, "learning_rate": 0.00029956718878812856, "loss": 0.225, "step": 508 }, { "epoch": 0.29, "learning_rate": 0.0002995053586150041, "loss": 0.2362, "step": 509 }, { "epoch": 0.29, "learning_rate": 0.0002994435284418796, "loss": 0.289, "step": 510 }, { "epoch": 0.29, "learning_rate": 0.00029938169826875514, "loss": 0.2464, "step": 511 }, { "epoch": 0.29, "learning_rate": 0.00029931986809563063, "loss": 0.2758, "step": 512 }, { "epoch": 0.29, "learning_rate": 0.0002992580379225061, "loss": 0.2434, "step": 513 }, { "epoch": 0.29, "learning_rate": 0.00029919620774938166, "loss": 0.2059, "step": 514 }, { "epoch": 0.29, "learning_rate": 0.0002991343775762572, "loss": 0.2723, "step": 515 }, { "epoch": 0.29, "learning_rate": 0.0002990725474031327, "loss": 0.2814, "step": 516 }, { "epoch": 0.29, "learning_rate": 0.00029901071723000824, "loss": 0.2401, "step": 517 }, { "epoch": 0.29, "learning_rate": 0.00029894888705688373, "loss": 0.3119, "step": 518 }, { "epoch": 0.29, "learning_rate": 0.00029888705688375927, "loss": 0.2685, "step": 519 }, { "epoch": 0.29, "learning_rate": 0.00029882522671063476, "loss": 0.2237, "step": 520 }, { "epoch": 0.29, "learning_rate": 0.0002987633965375103, "loss": 0.2165, "step": 521 }, { "epoch": 0.29, "learning_rate": 0.0002987015663643858, "loss": 0.2316, "step": 522 }, { "epoch": 0.29, "learning_rate": 0.0002986397361912613, "loss": 0.2629, "step": 523 }, { "epoch": 0.29, "learning_rate": 0.00029857790601813683, "loss": 0.2383, "step": 524 }, { "epoch": 0.29, "learning_rate": 0.00029851607584501237, "loss": 0.2605, "step": 525 }, { "epoch": 0.29, "learning_rate": 0.00029845424567188786, "loss": 0.2216, "step": 526 }, { "epoch": 0.3, "learning_rate": 0.00029839241549876335, "loss": 0.2801, "step": 527 }, { "epoch": 0.3, "learning_rate": 0.0002983305853256389, "loss": 0.1986, "step": 528 }, { "epoch": 0.3, "learning_rate": 0.0002982687551525144, "loss": 0.2625, "step": 529 }, { "epoch": 0.3, "learning_rate": 0.00029820692497938993, "loss": 0.1977, "step": 530 }, { "epoch": 0.3, "learning_rate": 0.0002981450948062654, "loss": 0.1902, "step": 531 }, { "epoch": 0.3, "learning_rate": 0.00029808326463314096, "loss": 0.2608, "step": 532 }, { "epoch": 0.3, "learning_rate": 0.00029802143446001645, "loss": 0.149, "step": 533 }, { "epoch": 0.3, "learning_rate": 0.000297959604286892, "loss": 0.2039, "step": 534 }, { "epoch": 0.3, "learning_rate": 0.0002978977741137675, "loss": 0.3103, "step": 535 }, { "epoch": 0.3, "learning_rate": 0.000297835943940643, "loss": 0.2779, "step": 536 }, { "epoch": 0.3, "learning_rate": 0.0002977741137675185, "loss": 0.2217, "step": 537 }, { "epoch": 0.3, "learning_rate": 0.00029771228359439406, "loss": 0.177, "step": 538 }, { "epoch": 0.3, "learning_rate": 0.00029765045342126955, "loss": 0.2878, "step": 539 }, { "epoch": 0.3, "learning_rate": 0.00029758862324814504, "loss": 0.2747, "step": 540 }, { "epoch": 0.3, "learning_rate": 0.0002975267930750206, "loss": 0.2096, "step": 541 }, { "epoch": 0.3, "learning_rate": 0.00029746496290189607, "loss": 0.3022, "step": 542 }, { "epoch": 0.3, "learning_rate": 0.0002974031327287716, "loss": 0.2189, "step": 543 }, { "epoch": 0.3, "learning_rate": 0.00029734130255564716, "loss": 0.277, "step": 544 }, { "epoch": 0.31, "learning_rate": 0.00029727947238252265, "loss": 0.3466, "step": 545 }, { "epoch": 0.31, "learning_rate": 0.00029721764220939814, "loss": 0.2607, "step": 546 }, { "epoch": 0.31, "learning_rate": 0.0002971558120362737, "loss": 0.4083, "step": 547 }, { "epoch": 0.31, "learning_rate": 0.0002970939818631492, "loss": 0.3234, "step": 548 }, { "epoch": 0.31, "learning_rate": 0.0002970321516900247, "loss": 0.4487, "step": 549 }, { "epoch": 0.31, "learning_rate": 0.0002969703215169002, "loss": 0.5202, "step": 550 }, { "epoch": 0.31, "learning_rate": 0.00029690849134377575, "loss": 0.3198, "step": 551 }, { "epoch": 0.31, "learning_rate": 0.00029684666117065124, "loss": 0.267, "step": 552 }, { "epoch": 0.31, "learning_rate": 0.0002967848309975268, "loss": 0.2284, "step": 553 }, { "epoch": 0.31, "learning_rate": 0.00029672300082440227, "loss": 0.329, "step": 554 }, { "epoch": 0.31, "learning_rate": 0.0002966611706512778, "loss": 0.2391, "step": 555 }, { "epoch": 0.31, "learning_rate": 0.0002965993404781533, "loss": 0.2573, "step": 556 }, { "epoch": 0.31, "learning_rate": 0.00029653751030502885, "loss": 0.2879, "step": 557 }, { "epoch": 0.31, "learning_rate": 0.00029647568013190434, "loss": 0.235, "step": 558 }, { "epoch": 0.31, "learning_rate": 0.00029641384995877983, "loss": 0.2025, "step": 559 }, { "epoch": 0.31, "learning_rate": 0.00029635201978565537, "loss": 0.2147, "step": 560 }, { "epoch": 0.31, "learning_rate": 0.0002962901896125309, "loss": 0.3112, "step": 561 }, { "epoch": 0.32, "learning_rate": 0.0002962283594394064, "loss": 0.244, "step": 562 }, { "epoch": 0.32, "learning_rate": 0.0002961665292662819, "loss": 0.1702, "step": 563 }, { "epoch": 0.32, "learning_rate": 0.00029610469909315744, "loss": 0.2282, "step": 564 }, { "epoch": 0.32, "learning_rate": 0.0002960428689200329, "loss": 0.2336, "step": 565 }, { "epoch": 0.32, "learning_rate": 0.00029598103874690847, "loss": 0.2141, "step": 566 }, { "epoch": 0.32, "learning_rate": 0.000295919208573784, "loss": 0.2626, "step": 567 }, { "epoch": 0.32, "learning_rate": 0.0002958573784006595, "loss": 0.1905, "step": 568 }, { "epoch": 0.32, "learning_rate": 0.000295795548227535, "loss": 0.1974, "step": 569 }, { "epoch": 0.32, "learning_rate": 0.00029573371805441054, "loss": 0.1913, "step": 570 }, { "epoch": 0.32, "learning_rate": 0.000295671887881286, "loss": 0.2098, "step": 571 }, { "epoch": 0.32, "learning_rate": 0.0002956100577081615, "loss": 0.1945, "step": 572 }, { "epoch": 0.32, "learning_rate": 0.00029554822753503706, "loss": 0.2043, "step": 573 }, { "epoch": 0.32, "learning_rate": 0.0002954863973619126, "loss": 0.1681, "step": 574 }, { "epoch": 0.32, "learning_rate": 0.0002954245671887881, "loss": 0.2182, "step": 575 }, { "epoch": 0.32, "learning_rate": 0.00029536273701566364, "loss": 0.1675, "step": 576 }, { "epoch": 0.32, "learning_rate": 0.0002953009068425391, "loss": 0.248, "step": 577 }, { "epoch": 0.32, "learning_rate": 0.00029523907666941467, "loss": 0.1669, "step": 578 }, { "epoch": 0.32, "learning_rate": 0.00029517724649629016, "loss": 0.2212, "step": 579 }, { "epoch": 0.33, "learning_rate": 0.0002951154163231657, "loss": 0.2468, "step": 580 }, { "epoch": 0.33, "learning_rate": 0.0002950535861500412, "loss": 0.1757, "step": 581 }, { "epoch": 0.33, "learning_rate": 0.0002949917559769167, "loss": 0.2105, "step": 582 }, { "epoch": 0.33, "learning_rate": 0.0002949299258037922, "loss": 0.2455, "step": 583 }, { "epoch": 0.33, "learning_rate": 0.00029486809563066777, "loss": 0.1815, "step": 584 }, { "epoch": 0.33, "learning_rate": 0.00029480626545754326, "loss": 0.2341, "step": 585 }, { "epoch": 0.33, "learning_rate": 0.00029474443528441875, "loss": 0.2344, "step": 586 }, { "epoch": 0.33, "learning_rate": 0.0002946826051112943, "loss": 0.214, "step": 587 }, { "epoch": 0.33, "learning_rate": 0.0002946207749381698, "loss": 0.2023, "step": 588 }, { "epoch": 0.33, "learning_rate": 0.0002945589447650453, "loss": 0.2087, "step": 589 }, { "epoch": 0.33, "learning_rate": 0.00029449711459192087, "loss": 0.2418, "step": 590 }, { "epoch": 0.33, "learning_rate": 0.00029443528441879636, "loss": 0.2382, "step": 591 }, { "epoch": 0.33, "learning_rate": 0.00029437345424567185, "loss": 0.184, "step": 592 }, { "epoch": 0.33, "learning_rate": 0.0002943116240725474, "loss": 0.2368, "step": 593 }, { "epoch": 0.33, "learning_rate": 0.0002942497938994229, "loss": 0.1775, "step": 594 }, { "epoch": 0.33, "learning_rate": 0.00029418796372629837, "loss": 0.2557, "step": 595 }, { "epoch": 0.33, "learning_rate": 0.0002941261335531739, "loss": 0.1742, "step": 596 }, { "epoch": 0.33, "learning_rate": 0.00029406430338004946, "loss": 0.3597, "step": 597 }, { "epoch": 0.34, "learning_rate": 0.00029400247320692495, "loss": 0.2881, "step": 598 }, { "epoch": 0.34, "learning_rate": 0.0002939406430338005, "loss": 0.4498, "step": 599 }, { "epoch": 0.34, "learning_rate": 0.000293878812860676, "loss": 0.5009, "step": 600 }, { "epoch": 0.34, "eval_loss": 0.27086159586906433, "eval_runtime": 273.1852, "eval_samples_per_second": 9.671, "eval_steps_per_second": 0.608, "eval_wer": 0.21616422054740184, "step": 600 }, { "epoch": 0.34, "learning_rate": 0.00029381698268755147, "loss": 0.2592, "step": 601 }, { "epoch": 0.34, "learning_rate": 0.000293755152514427, "loss": 0.2159, "step": 602 }, { "epoch": 0.34, "learning_rate": 0.00029369332234130256, "loss": 0.1903, "step": 603 }, { "epoch": 0.34, "learning_rate": 0.00029363149216817805, "loss": 0.211, "step": 604 }, { "epoch": 0.34, "learning_rate": 0.00029356966199505354, "loss": 0.2078, "step": 605 }, { "epoch": 0.34, "learning_rate": 0.0002935078318219291, "loss": 0.2204, "step": 606 }, { "epoch": 0.34, "learning_rate": 0.00029344600164880457, "loss": 0.2711, "step": 607 }, { "epoch": 0.34, "learning_rate": 0.0002933841714756801, "loss": 0.2279, "step": 608 }, { "epoch": 0.34, "learning_rate": 0.0002933223413025556, "loss": 0.2166, "step": 609 }, { "epoch": 0.34, "learning_rate": 0.00029326051112943115, "loss": 0.2198, "step": 610 }, { "epoch": 0.34, "learning_rate": 0.00029319868095630664, "loss": 0.2594, "step": 611 }, { "epoch": 0.34, "learning_rate": 0.0002931368507831822, "loss": 0.213, "step": 612 }, { "epoch": 0.34, "learning_rate": 0.0002930750206100577, "loss": 0.1991, "step": 613 }, { "epoch": 0.34, "learning_rate": 0.0002930131904369332, "loss": 0.1902, "step": 614 }, { "epoch": 0.34, "learning_rate": 0.0002929513602638087, "loss": 0.201, "step": 615 }, { "epoch": 0.35, "learning_rate": 0.00029288953009068425, "loss": 0.2235, "step": 616 }, { "epoch": 0.35, "learning_rate": 0.00029282769991755974, "loss": 0.1968, "step": 617 }, { "epoch": 0.35, "learning_rate": 0.0002927658697444352, "loss": 0.1838, "step": 618 }, { "epoch": 0.35, "learning_rate": 0.00029270403957131077, "loss": 0.1867, "step": 619 }, { "epoch": 0.35, "learning_rate": 0.0002926422093981863, "loss": 0.2616, "step": 620 }, { "epoch": 0.35, "learning_rate": 0.0002925803792250618, "loss": 0.2187, "step": 621 }, { "epoch": 0.35, "learning_rate": 0.00029251854905193735, "loss": 0.2081, "step": 622 }, { "epoch": 0.35, "learning_rate": 0.00029245671887881283, "loss": 0.1788, "step": 623 }, { "epoch": 0.35, "learning_rate": 0.0002923948887056883, "loss": 0.216, "step": 624 }, { "epoch": 0.35, "learning_rate": 0.00029233305853256387, "loss": 0.1669, "step": 625 }, { "epoch": 0.35, "learning_rate": 0.0002922712283594394, "loss": 0.1908, "step": 626 }, { "epoch": 0.35, "learning_rate": 0.0002922093981863149, "loss": 0.1747, "step": 627 }, { "epoch": 0.35, "learning_rate": 0.0002921475680131904, "loss": 0.2048, "step": 628 }, { "epoch": 0.35, "learning_rate": 0.00029208573784006593, "loss": 0.2613, "step": 629 }, { "epoch": 0.35, "learning_rate": 0.0002920239076669414, "loss": 0.196, "step": 630 }, { "epoch": 0.35, "learning_rate": 0.00029196207749381697, "loss": 0.18, "step": 631 }, { "epoch": 0.35, "learning_rate": 0.00029190024732069246, "loss": 0.1654, "step": 632 }, { "epoch": 0.35, "learning_rate": 0.000291838417147568, "loss": 0.1815, "step": 633 }, { "epoch": 0.36, "learning_rate": 0.0002917765869744435, "loss": 0.2123, "step": 634 }, { "epoch": 0.36, "learning_rate": 0.00029171475680131903, "loss": 0.1342, "step": 635 }, { "epoch": 0.36, "learning_rate": 0.0002916529266281945, "loss": 0.1996, "step": 636 }, { "epoch": 0.36, "learning_rate": 0.00029159109645507, "loss": 0.2196, "step": 637 }, { "epoch": 0.36, "learning_rate": 0.00029152926628194556, "loss": 0.1465, "step": 638 }, { "epoch": 0.36, "learning_rate": 0.0002914674361088211, "loss": 0.1814, "step": 639 }, { "epoch": 0.36, "learning_rate": 0.0002914056059356966, "loss": 0.2112, "step": 640 }, { "epoch": 0.36, "learning_rate": 0.0002913437757625721, "loss": 0.1922, "step": 641 }, { "epoch": 0.36, "learning_rate": 0.0002912819455894476, "loss": 0.1963, "step": 642 }, { "epoch": 0.36, "learning_rate": 0.00029122011541632317, "loss": 0.1911, "step": 643 }, { "epoch": 0.36, "learning_rate": 0.00029115828524319866, "loss": 0.2979, "step": 644 }, { "epoch": 0.36, "learning_rate": 0.0002910964550700742, "loss": 0.1595, "step": 645 }, { "epoch": 0.36, "learning_rate": 0.0002910346248969497, "loss": 0.3294, "step": 646 }, { "epoch": 0.36, "learning_rate": 0.0002909727947238252, "loss": 0.257, "step": 647 }, { "epoch": 0.36, "learning_rate": 0.0002909109645507007, "loss": 0.2642, "step": 648 }, { "epoch": 0.36, "learning_rate": 0.00029084913437757627, "loss": 0.2287, "step": 649 }, { "epoch": 0.36, "learning_rate": 0.00029078730420445176, "loss": 0.3652, "step": 650 }, { "epoch": 0.36, "learning_rate": 0.00029072547403132724, "loss": 0.2273, "step": 651 }, { "epoch": 0.37, "learning_rate": 0.0002906636438582028, "loss": 0.163, "step": 652 }, { "epoch": 0.37, "learning_rate": 0.0002906018136850783, "loss": 0.2131, "step": 653 }, { "epoch": 0.37, "learning_rate": 0.0002905399835119538, "loss": 0.2034, "step": 654 }, { "epoch": 0.37, "learning_rate": 0.0002904781533388293, "loss": 0.2016, "step": 655 }, { "epoch": 0.37, "learning_rate": 0.00029041632316570485, "loss": 0.1596, "step": 656 }, { "epoch": 0.37, "learning_rate": 0.00029035449299258034, "loss": 0.1713, "step": 657 }, { "epoch": 0.37, "learning_rate": 0.0002902926628194559, "loss": 0.2177, "step": 658 }, { "epoch": 0.37, "learning_rate": 0.0002902308326463314, "loss": 0.1585, "step": 659 }, { "epoch": 0.37, "learning_rate": 0.00029016900247320687, "loss": 0.1781, "step": 660 }, { "epoch": 0.37, "learning_rate": 0.0002901071723000824, "loss": 0.1734, "step": 661 }, { "epoch": 0.37, "learning_rate": 0.00029004534212695795, "loss": 0.1945, "step": 662 }, { "epoch": 0.37, "learning_rate": 0.00028998351195383344, "loss": 0.2126, "step": 663 }, { "epoch": 0.37, "learning_rate": 0.00028992168178070893, "loss": 0.2046, "step": 664 }, { "epoch": 0.37, "learning_rate": 0.0002898598516075845, "loss": 0.1244, "step": 665 }, { "epoch": 0.37, "learning_rate": 0.00028979802143445997, "loss": 0.1743, "step": 666 }, { "epoch": 0.37, "learning_rate": 0.0002897361912613355, "loss": 0.1811, "step": 667 }, { "epoch": 0.37, "learning_rate": 0.00028967436108821105, "loss": 0.2008, "step": 668 }, { "epoch": 0.38, "learning_rate": 0.00028961253091508654, "loss": 0.2328, "step": 669 }, { "epoch": 0.38, "learning_rate": 0.00028955070074196203, "loss": 0.1876, "step": 670 }, { "epoch": 0.38, "learning_rate": 0.0002894888705688376, "loss": 0.1925, "step": 671 }, { "epoch": 0.38, "learning_rate": 0.0002894270403957131, "loss": 0.134, "step": 672 }, { "epoch": 0.38, "learning_rate": 0.0002893652102225886, "loss": 0.1882, "step": 673 }, { "epoch": 0.38, "learning_rate": 0.0002893033800494641, "loss": 0.1698, "step": 674 }, { "epoch": 0.38, "learning_rate": 0.00028924154987633964, "loss": 0.2014, "step": 675 }, { "epoch": 0.38, "learning_rate": 0.00028917971970321513, "loss": 0.1823, "step": 676 }, { "epoch": 0.38, "learning_rate": 0.0002891178895300907, "loss": 0.1475, "step": 677 }, { "epoch": 0.38, "learning_rate": 0.00028905605935696617, "loss": 0.184, "step": 678 }, { "epoch": 0.38, "learning_rate": 0.0002889942291838417, "loss": 0.1606, "step": 679 }, { "epoch": 0.38, "learning_rate": 0.0002889323990107172, "loss": 0.2174, "step": 680 }, { "epoch": 0.38, "learning_rate": 0.00028887056883759274, "loss": 0.2283, "step": 681 }, { "epoch": 0.38, "learning_rate": 0.00028880873866446823, "loss": 0.1863, "step": 682 }, { "epoch": 0.38, "learning_rate": 0.0002887469084913437, "loss": 0.1935, "step": 683 }, { "epoch": 0.38, "learning_rate": 0.00028868507831821927, "loss": 0.1502, "step": 684 }, { "epoch": 0.38, "learning_rate": 0.0002886232481450948, "loss": 0.2315, "step": 685 }, { "epoch": 0.38, "learning_rate": 0.0002885614179719703, "loss": 0.198, "step": 686 }, { "epoch": 0.39, "learning_rate": 0.0002884995877988458, "loss": 0.1588, "step": 687 }, { "epoch": 0.39, "learning_rate": 0.00028843775762572133, "loss": 0.1513, "step": 688 }, { "epoch": 0.39, "learning_rate": 0.0002883759274525968, "loss": 0.2376, "step": 689 }, { "epoch": 0.39, "learning_rate": 0.00028831409727947236, "loss": 0.1911, "step": 690 }, { "epoch": 0.39, "learning_rate": 0.00028825226710634785, "loss": 0.1609, "step": 691 }, { "epoch": 0.39, "learning_rate": 0.0002881904369332234, "loss": 0.2073, "step": 692 }, { "epoch": 0.39, "learning_rate": 0.0002881286067600989, "loss": 0.1857, "step": 693 }, { "epoch": 0.39, "learning_rate": 0.00028806677658697443, "loss": 0.2565, "step": 694 }, { "epoch": 0.39, "learning_rate": 0.0002880049464138499, "loss": 0.2006, "step": 695 }, { "epoch": 0.39, "learning_rate": 0.0002879431162407254, "loss": 0.255, "step": 696 }, { "epoch": 0.39, "learning_rate": 0.00028788128606760095, "loss": 0.2625, "step": 697 }, { "epoch": 0.39, "learning_rate": 0.0002878194558944765, "loss": 0.3728, "step": 698 }, { "epoch": 0.39, "learning_rate": 0.000287757625721352, "loss": 0.4282, "step": 699 }, { "epoch": 0.39, "learning_rate": 0.00028769579554822753, "loss": 0.464, "step": 700 }, { "epoch": 0.39, "learning_rate": 0.000287633965375103, "loss": 0.1967, "step": 701 }, { "epoch": 0.39, "learning_rate": 0.00028757213520197856, "loss": 0.2464, "step": 702 }, { "epoch": 0.39, "learning_rate": 0.00028751030502885405, "loss": 0.2236, "step": 703 }, { "epoch": 0.39, "learning_rate": 0.0002874484748557296, "loss": 0.2505, "step": 704 }, { "epoch": 0.4, "learning_rate": 0.0002873866446826051, "loss": 0.1595, "step": 705 }, { "epoch": 0.4, "learning_rate": 0.0002873248145094806, "loss": 0.2188, "step": 706 }, { "epoch": 0.4, "learning_rate": 0.0002872629843363561, "loss": 0.2139, "step": 707 }, { "epoch": 0.4, "learning_rate": 0.00028720115416323166, "loss": 0.19, "step": 708 }, { "epoch": 0.4, "learning_rate": 0.00028713932399010715, "loss": 0.1697, "step": 709 }, { "epoch": 0.4, "learning_rate": 0.00028707749381698264, "loss": 0.1985, "step": 710 }, { "epoch": 0.4, "learning_rate": 0.0002870156636438582, "loss": 0.1732, "step": 711 }, { "epoch": 0.4, "learning_rate": 0.0002869538334707337, "loss": 0.1423, "step": 712 }, { "epoch": 0.4, "learning_rate": 0.0002868920032976092, "loss": 0.1464, "step": 713 }, { "epoch": 0.4, "learning_rate": 0.0002868301731244847, "loss": 0.1506, "step": 714 }, { "epoch": 0.4, "learning_rate": 0.00028676834295136025, "loss": 0.181, "step": 715 }, { "epoch": 0.4, "learning_rate": 0.00028670651277823574, "loss": 0.1121, "step": 716 }, { "epoch": 0.4, "learning_rate": 0.0002866446826051113, "loss": 0.2019, "step": 717 }, { "epoch": 0.4, "learning_rate": 0.0002865828524319868, "loss": 0.1563, "step": 718 }, { "epoch": 0.4, "learning_rate": 0.00028652102225886226, "loss": 0.1731, "step": 719 }, { "epoch": 0.4, "learning_rate": 0.0002864591920857378, "loss": 0.1477, "step": 720 }, { "epoch": 0.4, "learning_rate": 0.00028639736191261335, "loss": 0.1749, "step": 721 }, { "epoch": 0.4, "learning_rate": 0.00028633553173948884, "loss": 0.2094, "step": 722 }, { "epoch": 0.41, "learning_rate": 0.0002862737015663644, "loss": 0.1553, "step": 723 }, { "epoch": 0.41, "learning_rate": 0.0002862118713932399, "loss": 0.1699, "step": 724 }, { "epoch": 0.41, "learning_rate": 0.00028615004122011536, "loss": 0.1697, "step": 725 }, { "epoch": 0.41, "learning_rate": 0.0002860882110469909, "loss": 0.1659, "step": 726 }, { "epoch": 0.41, "learning_rate": 0.00028602638087386645, "loss": 0.1232, "step": 727 }, { "epoch": 0.41, "learning_rate": 0.00028596455070074194, "loss": 0.1561, "step": 728 }, { "epoch": 0.41, "learning_rate": 0.00028590272052761743, "loss": 0.1264, "step": 729 }, { "epoch": 0.41, "learning_rate": 0.000285840890354493, "loss": 0.1843, "step": 730 }, { "epoch": 0.41, "learning_rate": 0.0002857790601813685, "loss": 0.124, "step": 731 }, { "epoch": 0.41, "learning_rate": 0.000285717230008244, "loss": 0.1843, "step": 732 }, { "epoch": 0.41, "learning_rate": 0.0002856553998351195, "loss": 0.1447, "step": 733 }, { "epoch": 0.41, "learning_rate": 0.00028559356966199504, "loss": 0.1601, "step": 734 }, { "epoch": 0.41, "learning_rate": 0.00028553173948887053, "loss": 0.1542, "step": 735 }, { "epoch": 0.41, "learning_rate": 0.0002854699093157461, "loss": 0.1294, "step": 736 }, { "epoch": 0.41, "learning_rate": 0.00028540807914262156, "loss": 0.1541, "step": 737 }, { "epoch": 0.41, "learning_rate": 0.0002853462489694971, "loss": 0.1869, "step": 738 }, { "epoch": 0.41, "learning_rate": 0.0002852844187963726, "loss": 0.1548, "step": 739 }, { "epoch": 0.41, "learning_rate": 0.00028522258862324814, "loss": 0.1802, "step": 740 }, { "epoch": 0.42, "learning_rate": 0.00028516075845012363, "loss": 0.1732, "step": 741 }, { "epoch": 0.42, "learning_rate": 0.0002850989282769991, "loss": 0.1513, "step": 742 }, { "epoch": 0.42, "learning_rate": 0.00028503709810387466, "loss": 0.2004, "step": 743 }, { "epoch": 0.42, "learning_rate": 0.0002849752679307502, "loss": 0.2644, "step": 744 }, { "epoch": 0.42, "learning_rate": 0.0002849134377576257, "loss": 0.2793, "step": 745 }, { "epoch": 0.42, "learning_rate": 0.0002848516075845012, "loss": 0.1746, "step": 746 }, { "epoch": 0.42, "learning_rate": 0.00028478977741137673, "loss": 0.2939, "step": 747 }, { "epoch": 0.42, "learning_rate": 0.0002847279472382522, "loss": 0.2139, "step": 748 }, { "epoch": 0.42, "learning_rate": 0.00028466611706512776, "loss": 0.205, "step": 749 }, { "epoch": 0.42, "learning_rate": 0.0002846042868920033, "loss": 0.2406, "step": 750 }, { "epoch": 0.42, "learning_rate": 0.0002845424567188788, "loss": 0.2229, "step": 751 }, { "epoch": 0.42, "learning_rate": 0.0002844806265457543, "loss": 0.1866, "step": 752 }, { "epoch": 0.42, "learning_rate": 0.00028441879637262983, "loss": 0.137, "step": 753 }, { "epoch": 0.42, "learning_rate": 0.0002843569661995053, "loss": 0.1644, "step": 754 }, { "epoch": 0.42, "learning_rate": 0.00028429513602638086, "loss": 0.1622, "step": 755 }, { "epoch": 0.42, "learning_rate": 0.00028423330585325635, "loss": 0.1285, "step": 756 }, { "epoch": 0.42, "learning_rate": 0.0002841714756801319, "loss": 0.1829, "step": 757 }, { "epoch": 0.42, "learning_rate": 0.0002841096455070074, "loss": 0.2087, "step": 758 }, { "epoch": 0.43, "learning_rate": 0.00028404781533388293, "loss": 0.1591, "step": 759 }, { "epoch": 0.43, "learning_rate": 0.0002839859851607584, "loss": 0.1853, "step": 760 }, { "epoch": 0.43, "learning_rate": 0.00028392415498763396, "loss": 0.1572, "step": 761 }, { "epoch": 0.43, "learning_rate": 0.00028386232481450945, "loss": 0.1691, "step": 762 }, { "epoch": 0.43, "learning_rate": 0.000283800494641385, "loss": 0.1426, "step": 763 }, { "epoch": 0.43, "learning_rate": 0.0002837386644682605, "loss": 0.1324, "step": 764 }, { "epoch": 0.43, "learning_rate": 0.000283676834295136, "loss": 0.1354, "step": 765 }, { "epoch": 0.43, "learning_rate": 0.0002836150041220115, "loss": 0.1481, "step": 766 }, { "epoch": 0.43, "learning_rate": 0.00028355317394888706, "loss": 0.1952, "step": 767 }, { "epoch": 0.43, "learning_rate": 0.00028349134377576255, "loss": 0.1729, "step": 768 }, { "epoch": 0.43, "learning_rate": 0.00028342951360263804, "loss": 0.155, "step": 769 }, { "epoch": 0.43, "learning_rate": 0.0002833676834295136, "loss": 0.1204, "step": 770 }, { "epoch": 0.43, "learning_rate": 0.00028330585325638907, "loss": 0.1732, "step": 771 }, { "epoch": 0.43, "learning_rate": 0.0002832440230832646, "loss": 0.14, "step": 772 }, { "epoch": 0.43, "learning_rate": 0.00028318219291014016, "loss": 0.1761, "step": 773 }, { "epoch": 0.43, "learning_rate": 0.00028312036273701565, "loss": 0.1492, "step": 774 }, { "epoch": 0.43, "learning_rate": 0.00028305853256389114, "loss": 0.1644, "step": 775 }, { "epoch": 0.43, "learning_rate": 0.0002829967023907667, "loss": 0.1796, "step": 776 }, { "epoch": 0.44, "learning_rate": 0.00028293487221764217, "loss": 0.1583, "step": 777 }, { "epoch": 0.44, "learning_rate": 0.00028287304204451766, "loss": 0.1592, "step": 778 }, { "epoch": 0.44, "learning_rate": 0.0002828112118713932, "loss": 0.1478, "step": 779 }, { "epoch": 0.44, "learning_rate": 0.00028274938169826875, "loss": 0.1252, "step": 780 }, { "epoch": 0.44, "learning_rate": 0.00028268755152514424, "loss": 0.1431, "step": 781 }, { "epoch": 0.44, "learning_rate": 0.0002826257213520198, "loss": 0.118, "step": 782 }, { "epoch": 0.44, "learning_rate": 0.00028256389117889527, "loss": 0.1487, "step": 783 }, { "epoch": 0.44, "learning_rate": 0.00028250206100577076, "loss": 0.1726, "step": 784 }, { "epoch": 0.44, "learning_rate": 0.0002824402308326463, "loss": 0.1505, "step": 785 }, { "epoch": 0.44, "learning_rate": 0.00028237840065952185, "loss": 0.1335, "step": 786 }, { "epoch": 0.44, "learning_rate": 0.00028231657048639734, "loss": 0.1609, "step": 787 }, { "epoch": 0.44, "learning_rate": 0.00028225474031327283, "loss": 0.1929, "step": 788 }, { "epoch": 0.44, "learning_rate": 0.00028219291014014837, "loss": 0.1987, "step": 789 }, { "epoch": 0.44, "learning_rate": 0.00028213107996702386, "loss": 0.1678, "step": 790 }, { "epoch": 0.44, "learning_rate": 0.0002820692497938994, "loss": 0.1321, "step": 791 }, { "epoch": 0.44, "learning_rate": 0.0002820074196207749, "loss": 0.2386, "step": 792 }, { "epoch": 0.44, "learning_rate": 0.00028194558944765044, "loss": 0.1976, "step": 793 }, { "epoch": 0.45, "learning_rate": 0.0002818837592745259, "loss": 0.1798, "step": 794 }, { "epoch": 0.45, "learning_rate": 0.00028182192910140147, "loss": 0.2646, "step": 795 }, { "epoch": 0.45, "learning_rate": 0.000281760098928277, "loss": 0.2346, "step": 796 }, { "epoch": 0.45, "learning_rate": 0.0002816982687551525, "loss": 0.2303, "step": 797 }, { "epoch": 0.45, "learning_rate": 0.000281636438582028, "loss": 0.298, "step": 798 }, { "epoch": 0.45, "learning_rate": 0.00028157460840890354, "loss": 0.3555, "step": 799 }, { "epoch": 0.45, "learning_rate": 0.000281512778235779, "loss": 0.3332, "step": 800 }, { "epoch": 0.45, "learning_rate": 0.0002814509480626545, "loss": 0.2179, "step": 801 }, { "epoch": 0.45, "learning_rate": 0.00028138911788953006, "loss": 0.151, "step": 802 }, { "epoch": 0.45, "learning_rate": 0.0002813272877164056, "loss": 0.1935, "step": 803 }, { "epoch": 0.45, "learning_rate": 0.0002812654575432811, "loss": 0.2319, "step": 804 }, { "epoch": 0.45, "learning_rate": 0.00028120362737015664, "loss": 0.1608, "step": 805 }, { "epoch": 0.45, "learning_rate": 0.0002811417971970321, "loss": 0.1645, "step": 806 }, { "epoch": 0.45, "learning_rate": 0.0002810799670239076, "loss": 0.1821, "step": 807 }, { "epoch": 0.45, "learning_rate": 0.00028101813685078316, "loss": 0.1663, "step": 808 }, { "epoch": 0.45, "learning_rate": 0.0002809563066776587, "loss": 0.1972, "step": 809 }, { "epoch": 0.45, "learning_rate": 0.0002808944765045342, "loss": 0.1856, "step": 810 }, { "epoch": 0.45, "learning_rate": 0.0002808326463314097, "loss": 0.1734, "step": 811 }, { "epoch": 0.46, "learning_rate": 0.0002807708161582852, "loss": 0.1678, "step": 812 }, { "epoch": 0.46, "learning_rate": 0.0002807089859851607, "loss": 0.1826, "step": 813 }, { "epoch": 0.46, "learning_rate": 0.00028064715581203626, "loss": 0.1648, "step": 814 }, { "epoch": 0.46, "learning_rate": 0.00028058532563891175, "loss": 0.1234, "step": 815 }, { "epoch": 0.46, "learning_rate": 0.0002805234954657873, "loss": 0.1884, "step": 816 }, { "epoch": 0.46, "learning_rate": 0.0002804616652926628, "loss": 0.1482, "step": 817 }, { "epoch": 0.46, "learning_rate": 0.0002803998351195383, "loss": 0.1672, "step": 818 }, { "epoch": 0.46, "learning_rate": 0.0002803380049464138, "loss": 0.1914, "step": 819 }, { "epoch": 0.46, "learning_rate": 0.0002802761747732893, "loss": 0.1533, "step": 820 }, { "epoch": 0.46, "learning_rate": 0.00028021434460016485, "loss": 0.159, "step": 821 }, { "epoch": 0.46, "learning_rate": 0.0002801525144270404, "loss": 0.1422, "step": 822 }, { "epoch": 0.46, "learning_rate": 0.0002800906842539159, "loss": 0.164, "step": 823 }, { "epoch": 0.46, "learning_rate": 0.00028002885408079137, "loss": 0.2057, "step": 824 }, { "epoch": 0.46, "learning_rate": 0.0002799670239076669, "loss": 0.1465, "step": 825 }, { "epoch": 0.46, "learning_rate": 0.00027990519373454246, "loss": 0.1565, "step": 826 }, { "epoch": 0.46, "learning_rate": 0.00027984336356141795, "loss": 0.1866, "step": 827 }, { "epoch": 0.46, "learning_rate": 0.0002797815333882935, "loss": 0.1508, "step": 828 }, { "epoch": 0.46, "learning_rate": 0.000279719703215169, "loss": 0.1838, "step": 829 }, { "epoch": 0.47, "learning_rate": 0.00027965787304204447, "loss": 0.1617, "step": 830 }, { "epoch": 0.47, "learning_rate": 0.00027959604286892, "loss": 0.2064, "step": 831 }, { "epoch": 0.47, "learning_rate": 0.00027953421269579556, "loss": 0.1738, "step": 832 }, { "epoch": 0.47, "learning_rate": 0.00027947238252267105, "loss": 0.1429, "step": 833 }, { "epoch": 0.47, "learning_rate": 0.00027941055234954654, "loss": 0.1117, "step": 834 }, { "epoch": 0.47, "learning_rate": 0.0002793487221764221, "loss": 0.1924, "step": 835 }, { "epoch": 0.47, "learning_rate": 0.00027928689200329757, "loss": 0.1519, "step": 836 }, { "epoch": 0.47, "learning_rate": 0.0002792250618301731, "loss": 0.1273, "step": 837 }, { "epoch": 0.47, "learning_rate": 0.0002791632316570486, "loss": 0.1195, "step": 838 }, { "epoch": 0.47, "learning_rate": 0.00027910140148392415, "loss": 0.1337, "step": 839 }, { "epoch": 0.47, "learning_rate": 0.00027903957131079964, "loss": 0.1895, "step": 840 }, { "epoch": 0.47, "learning_rate": 0.0002789777411376752, "loss": 0.1841, "step": 841 }, { "epoch": 0.47, "learning_rate": 0.00027891591096455067, "loss": 0.1852, "step": 842 }, { "epoch": 0.47, "learning_rate": 0.00027885408079142616, "loss": 0.2053, "step": 843 }, { "epoch": 0.47, "learning_rate": 0.0002787922506183017, "loss": 0.1628, "step": 844 }, { "epoch": 0.47, "learning_rate": 0.00027873042044517725, "loss": 0.1652, "step": 845 }, { "epoch": 0.47, "learning_rate": 0.00027866859027205274, "loss": 0.1965, "step": 846 }, { "epoch": 0.47, "learning_rate": 0.0002786067600989282, "loss": 0.1481, "step": 847 }, { "epoch": 0.48, "learning_rate": 0.00027854492992580377, "loss": 0.3173, "step": 848 }, { "epoch": 0.48, "learning_rate": 0.00027848309975267926, "loss": 0.3504, "step": 849 }, { "epoch": 0.48, "learning_rate": 0.0002784212695795548, "loss": 0.639, "step": 850 }, { "epoch": 0.48, "learning_rate": 0.00027835943940643035, "loss": 0.1711, "step": 851 }, { "epoch": 0.48, "learning_rate": 0.00027829760923330583, "loss": 0.1938, "step": 852 }, { "epoch": 0.48, "learning_rate": 0.0002782357790601813, "loss": 0.1479, "step": 853 }, { "epoch": 0.48, "learning_rate": 0.00027817394888705687, "loss": 0.1527, "step": 854 }, { "epoch": 0.48, "learning_rate": 0.0002781121187139324, "loss": 0.1625, "step": 855 }, { "epoch": 0.48, "learning_rate": 0.0002780502885408079, "loss": 0.1241, "step": 856 }, { "epoch": 0.48, "learning_rate": 0.0002779884583676834, "loss": 0.145, "step": 857 }, { "epoch": 0.48, "learning_rate": 0.00027792662819455893, "loss": 0.1343, "step": 858 }, { "epoch": 0.48, "learning_rate": 0.0002778647980214344, "loss": 0.1804, "step": 859 }, { "epoch": 0.48, "learning_rate": 0.00027780296784830997, "loss": 0.1458, "step": 860 }, { "epoch": 0.48, "learning_rate": 0.00027774113767518546, "loss": 0.1593, "step": 861 }, { "epoch": 0.48, "learning_rate": 0.000277679307502061, "loss": 0.1662, "step": 862 }, { "epoch": 0.48, "learning_rate": 0.0002776174773289365, "loss": 0.1605, "step": 863 }, { "epoch": 0.48, "learning_rate": 0.00027755564715581203, "loss": 0.1004, "step": 864 }, { "epoch": 0.48, "learning_rate": 0.0002774938169826875, "loss": 0.1874, "step": 865 }, { "epoch": 0.49, "learning_rate": 0.000277431986809563, "loss": 0.1387, "step": 866 }, { "epoch": 0.49, "learning_rate": 0.00027737015663643856, "loss": 0.1134, "step": 867 }, { "epoch": 0.49, "learning_rate": 0.0002773083264633141, "loss": 0.196, "step": 868 }, { "epoch": 0.49, "learning_rate": 0.0002772464962901896, "loss": 0.1664, "step": 869 }, { "epoch": 0.49, "learning_rate": 0.0002771846661170651, "loss": 0.1929, "step": 870 }, { "epoch": 0.49, "learning_rate": 0.0002771228359439406, "loss": 0.175, "step": 871 }, { "epoch": 0.49, "learning_rate": 0.0002770610057708161, "loss": 0.122, "step": 872 }, { "epoch": 0.49, "learning_rate": 0.00027699917559769166, "loss": 0.1544, "step": 873 }, { "epoch": 0.49, "learning_rate": 0.0002769373454245672, "loss": 0.1249, "step": 874 }, { "epoch": 0.49, "learning_rate": 0.0002768755152514427, "loss": 0.1379, "step": 875 }, { "epoch": 0.49, "learning_rate": 0.0002768136850783182, "loss": 0.153, "step": 876 }, { "epoch": 0.49, "learning_rate": 0.0002767518549051937, "loss": 0.1445, "step": 877 }, { "epoch": 0.49, "learning_rate": 0.0002766900247320692, "loss": 0.1484, "step": 878 }, { "epoch": 0.49, "learning_rate": 0.0002766281945589447, "loss": 0.0882, "step": 879 }, { "epoch": 0.49, "learning_rate": 0.00027656636438582024, "loss": 0.138, "step": 880 }, { "epoch": 0.49, "learning_rate": 0.0002765045342126958, "loss": 0.1321, "step": 881 }, { "epoch": 0.49, "learning_rate": 0.0002764427040395713, "loss": 0.1253, "step": 882 }, { "epoch": 0.49, "learning_rate": 0.0002763808738664468, "loss": 0.2232, "step": 883 }, { "epoch": 0.5, "learning_rate": 0.0002763190436933223, "loss": 0.2084, "step": 884 }, { "epoch": 0.5, "learning_rate": 0.00027625721352019786, "loss": 0.1587, "step": 885 }, { "epoch": 0.5, "learning_rate": 0.00027619538334707334, "loss": 0.195, "step": 886 }, { "epoch": 0.5, "learning_rate": 0.0002761335531739489, "loss": 0.1956, "step": 887 }, { "epoch": 0.5, "learning_rate": 0.0002760717230008244, "loss": 0.1009, "step": 888 }, { "epoch": 0.5, "learning_rate": 0.00027600989282769987, "loss": 0.1401, "step": 889 }, { "epoch": 0.5, "learning_rate": 0.0002759480626545754, "loss": 0.1773, "step": 890 }, { "epoch": 0.5, "learning_rate": 0.00027588623248145095, "loss": 0.1992, "step": 891 }, { "epoch": 0.5, "learning_rate": 0.00027582440230832644, "loss": 0.221, "step": 892 }, { "epoch": 0.5, "learning_rate": 0.00027576257213520193, "loss": 0.236, "step": 893 }, { "epoch": 0.5, "learning_rate": 0.0002757007419620775, "loss": 0.2025, "step": 894 }, { "epoch": 0.5, "learning_rate": 0.00027563891178895297, "loss": 0.2056, "step": 895 }, { "epoch": 0.5, "learning_rate": 0.0002755770816158285, "loss": 0.1843, "step": 896 }, { "epoch": 0.5, "learning_rate": 0.000275515251442704, "loss": 0.3171, "step": 897 }, { "epoch": 0.5, "learning_rate": 0.00027545342126957954, "loss": 0.2157, "step": 898 }, { "epoch": 0.5, "learning_rate": 0.00027539159109645503, "loss": 0.2615, "step": 899 }, { "epoch": 0.5, "learning_rate": 0.0002753297609233306, "loss": 0.2056, "step": 900 }, { "epoch": 0.5, "eval_loss": 0.19344282150268555, "eval_runtime": 270.5572, "eval_samples_per_second": 9.765, "eval_steps_per_second": 0.614, "eval_wer": 0.16021420071400239, "step": 900 }, { "epoch": 0.51, "learning_rate": 0.00027526793075020607, "loss": 0.1603, "step": 901 }, { "epoch": 0.51, "learning_rate": 0.00027520610057708156, "loss": 0.197, "step": 902 }, { "epoch": 0.51, "learning_rate": 0.0002751442704039571, "loss": 0.1887, "step": 903 }, { "epoch": 0.51, "learning_rate": 0.00027508244023083264, "loss": 0.1289, "step": 904 }, { "epoch": 0.51, "learning_rate": 0.00027502061005770813, "loss": 0.1588, "step": 905 }, { "epoch": 0.51, "learning_rate": 0.0002749587798845837, "loss": 0.1782, "step": 906 }, { "epoch": 0.51, "learning_rate": 0.00027489694971145917, "loss": 0.1656, "step": 907 }, { "epoch": 0.51, "learning_rate": 0.00027483511953833466, "loss": 0.099, "step": 908 }, { "epoch": 0.51, "learning_rate": 0.0002747732893652102, "loss": 0.1638, "step": 909 }, { "epoch": 0.51, "learning_rate": 0.00027471145919208574, "loss": 0.1457, "step": 910 }, { "epoch": 0.51, "learning_rate": 0.00027464962901896123, "loss": 0.1728, "step": 911 }, { "epoch": 0.51, "learning_rate": 0.0002745877988458367, "loss": 0.1539, "step": 912 }, { "epoch": 0.51, "learning_rate": 0.00027452596867271227, "loss": 0.1253, "step": 913 }, { "epoch": 0.51, "learning_rate": 0.0002744641384995878, "loss": 0.1893, "step": 914 }, { "epoch": 0.51, "learning_rate": 0.0002744023083264633, "loss": 0.1743, "step": 915 }, { "epoch": 0.51, "learning_rate": 0.0002743404781533388, "loss": 0.1324, "step": 916 }, { "epoch": 0.51, "learning_rate": 0.00027427864798021433, "loss": 0.1254, "step": 917 }, { "epoch": 0.51, "learning_rate": 0.0002742168178070898, "loss": 0.195, "step": 918 }, { "epoch": 0.52, "learning_rate": 0.00027415498763396536, "loss": 0.1532, "step": 919 }, { "epoch": 0.52, "learning_rate": 0.00027409315746084085, "loss": 0.1385, "step": 920 }, { "epoch": 0.52, "learning_rate": 0.0002740313272877164, "loss": 0.1528, "step": 921 }, { "epoch": 0.52, "learning_rate": 0.0002739694971145919, "loss": 0.1596, "step": 922 }, { "epoch": 0.52, "learning_rate": 0.00027390766694146743, "loss": 0.1132, "step": 923 }, { "epoch": 0.52, "learning_rate": 0.0002738458367683429, "loss": 0.1369, "step": 924 }, { "epoch": 0.52, "learning_rate": 0.0002737840065952184, "loss": 0.1735, "step": 925 }, { "epoch": 0.52, "learning_rate": 0.00027372217642209395, "loss": 0.1754, "step": 926 }, { "epoch": 0.52, "learning_rate": 0.0002736603462489695, "loss": 0.1322, "step": 927 }, { "epoch": 0.52, "learning_rate": 0.000273598516075845, "loss": 0.1261, "step": 928 }, { "epoch": 0.52, "learning_rate": 0.0002735366859027205, "loss": 0.1374, "step": 929 }, { "epoch": 0.52, "learning_rate": 0.000273474855729596, "loss": 0.1141, "step": 930 }, { "epoch": 0.52, "learning_rate": 0.0002734130255564715, "loss": 0.1848, "step": 931 }, { "epoch": 0.52, "learning_rate": 0.00027335119538334705, "loss": 0.1868, "step": 932 }, { "epoch": 0.52, "learning_rate": 0.0002732893652102226, "loss": 0.1547, "step": 933 }, { "epoch": 0.52, "learning_rate": 0.0002732275350370981, "loss": 0.1687, "step": 934 }, { "epoch": 0.52, "learning_rate": 0.0002731657048639736, "loss": 0.1674, "step": 935 }, { "epoch": 0.52, "learning_rate": 0.0002731038746908491, "loss": 0.1415, "step": 936 }, { "epoch": 0.53, "learning_rate": 0.0002730420445177246, "loss": 0.1885, "step": 937 }, { "epoch": 0.53, "learning_rate": 0.00027298021434460015, "loss": 0.132, "step": 938 }, { "epoch": 0.53, "learning_rate": 0.00027291838417147564, "loss": 0.1394, "step": 939 }, { "epoch": 0.53, "learning_rate": 0.0002728565539983512, "loss": 0.1466, "step": 940 }, { "epoch": 0.53, "learning_rate": 0.0002727947238252267, "loss": 0.192, "step": 941 }, { "epoch": 0.53, "learning_rate": 0.0002727328936521022, "loss": 0.1616, "step": 942 }, { "epoch": 0.53, "learning_rate": 0.0002726710634789777, "loss": 0.1383, "step": 943 }, { "epoch": 0.53, "learning_rate": 0.00027260923330585325, "loss": 0.1686, "step": 944 }, { "epoch": 0.53, "learning_rate": 0.00027254740313272874, "loss": 0.266, "step": 945 }, { "epoch": 0.53, "learning_rate": 0.0002724855729596043, "loss": 0.19, "step": 946 }, { "epoch": 0.53, "learning_rate": 0.0002724237427864798, "loss": 0.2676, "step": 947 }, { "epoch": 0.53, "learning_rate": 0.00027236191261335526, "loss": 0.2308, "step": 948 }, { "epoch": 0.53, "learning_rate": 0.0002723000824402308, "loss": 0.3344, "step": 949 }, { "epoch": 0.53, "learning_rate": 0.00027223825226710635, "loss": 0.4768, "step": 950 }, { "epoch": 0.53, "learning_rate": 0.00027217642209398184, "loss": 0.1369, "step": 951 }, { "epoch": 0.53, "learning_rate": 0.00027211459192085733, "loss": 0.1325, "step": 952 }, { "epoch": 0.53, "learning_rate": 0.0002720527617477329, "loss": 0.164, "step": 953 }, { "epoch": 0.53, "learning_rate": 0.00027199093157460836, "loss": 0.1167, "step": 954 }, { "epoch": 0.54, "learning_rate": 0.0002719291014014839, "loss": 0.1906, "step": 955 }, { "epoch": 0.54, "learning_rate": 0.00027186727122835945, "loss": 0.1144, "step": 956 }, { "epoch": 0.54, "learning_rate": 0.00027180544105523494, "loss": 0.1974, "step": 957 }, { "epoch": 0.54, "learning_rate": 0.00027174361088211043, "loss": 0.1409, "step": 958 }, { "epoch": 0.54, "learning_rate": 0.000271681780708986, "loss": 0.181, "step": 959 }, { "epoch": 0.54, "learning_rate": 0.00027161995053586146, "loss": 0.1456, "step": 960 }, { "epoch": 0.54, "learning_rate": 0.000271558120362737, "loss": 0.1234, "step": 961 }, { "epoch": 0.54, "learning_rate": 0.0002714962901896125, "loss": 0.1878, "step": 962 }, { "epoch": 0.54, "learning_rate": 0.00027143446001648804, "loss": 0.1497, "step": 963 }, { "epoch": 0.54, "learning_rate": 0.00027137262984336353, "loss": 0.1231, "step": 964 }, { "epoch": 0.54, "learning_rate": 0.0002713107996702391, "loss": 0.149, "step": 965 }, { "epoch": 0.54, "learning_rate": 0.00027124896949711456, "loss": 0.1085, "step": 966 }, { "epoch": 0.54, "learning_rate": 0.00027118713932399005, "loss": 0.1421, "step": 967 }, { "epoch": 0.54, "learning_rate": 0.0002711253091508656, "loss": 0.179, "step": 968 }, { "epoch": 0.54, "learning_rate": 0.00027106347897774114, "loss": 0.1377, "step": 969 }, { "epoch": 0.54, "learning_rate": 0.00027100164880461663, "loss": 0.1091, "step": 970 }, { "epoch": 0.54, "learning_rate": 0.0002709398186314921, "loss": 0.1831, "step": 971 }, { "epoch": 0.54, "learning_rate": 0.00027087798845836766, "loss": 0.1478, "step": 972 }, { "epoch": 0.55, "learning_rate": 0.00027081615828524315, "loss": 0.1867, "step": 973 }, { "epoch": 0.55, "learning_rate": 0.0002707543281121187, "loss": 0.1807, "step": 974 }, { "epoch": 0.55, "learning_rate": 0.0002706924979389942, "loss": 0.1373, "step": 975 }, { "epoch": 0.55, "learning_rate": 0.00027063066776586973, "loss": 0.1513, "step": 976 }, { "epoch": 0.55, "learning_rate": 0.0002705688375927452, "loss": 0.2116, "step": 977 }, { "epoch": 0.55, "learning_rate": 0.00027050700741962076, "loss": 0.1462, "step": 978 }, { "epoch": 0.55, "learning_rate": 0.0002704451772464963, "loss": 0.153, "step": 979 }, { "epoch": 0.55, "learning_rate": 0.0002703833470733718, "loss": 0.1017, "step": 980 }, { "epoch": 0.55, "learning_rate": 0.0002703215169002473, "loss": 0.1375, "step": 981 }, { "epoch": 0.55, "learning_rate": 0.00027025968672712283, "loss": 0.1478, "step": 982 }, { "epoch": 0.55, "learning_rate": 0.0002701978565539983, "loss": 0.1604, "step": 983 }, { "epoch": 0.55, "learning_rate": 0.0002701360263808738, "loss": 0.1645, "step": 984 }, { "epoch": 0.55, "learning_rate": 0.00027007419620774935, "loss": 0.1004, "step": 985 }, { "epoch": 0.55, "learning_rate": 0.0002700123660346249, "loss": 0.2081, "step": 986 }, { "epoch": 0.55, "learning_rate": 0.0002699505358615004, "loss": 0.1759, "step": 987 }, { "epoch": 0.55, "learning_rate": 0.00026988870568837593, "loss": 0.1218, "step": 988 }, { "epoch": 0.55, "learning_rate": 0.0002698268755152514, "loss": 0.1745, "step": 989 }, { "epoch": 0.55, "learning_rate": 0.0002697650453421269, "loss": 0.1599, "step": 990 }, { "epoch": 0.56, "learning_rate": 0.00026970321516900245, "loss": 0.1461, "step": 991 }, { "epoch": 0.56, "learning_rate": 0.000269641384995878, "loss": 0.1631, "step": 992 }, { "epoch": 0.56, "learning_rate": 0.0002695795548227535, "loss": 0.1486, "step": 993 }, { "epoch": 0.56, "learning_rate": 0.000269517724649629, "loss": 0.1357, "step": 994 }, { "epoch": 0.56, "learning_rate": 0.0002694558944765045, "loss": 0.1515, "step": 995 }, { "epoch": 0.56, "learning_rate": 0.00026939406430338, "loss": 0.1451, "step": 996 }, { "epoch": 0.56, "learning_rate": 0.00026933223413025555, "loss": 0.1598, "step": 997 }, { "epoch": 0.56, "learning_rate": 0.00026927040395713104, "loss": 0.3699, "step": 998 }, { "epoch": 0.56, "learning_rate": 0.0002692085737840066, "loss": 0.3928, "step": 999 }, { "epoch": 0.56, "learning_rate": 0.00026914674361088207, "loss": 0.465, "step": 1000 }, { "epoch": 0.56, "learning_rate": 0.0002690849134377576, "loss": 0.2188, "step": 1001 }, { "epoch": 0.56, "learning_rate": 0.0002690230832646331, "loss": 0.1736, "step": 1002 }, { "epoch": 0.56, "learning_rate": 0.0002689612530915086, "loss": 0.1544, "step": 1003 }, { "epoch": 0.56, "learning_rate": 0.00026889942291838414, "loss": 0.1614, "step": 1004 }, { "epoch": 0.56, "learning_rate": 0.0002688375927452597, "loss": 0.1797, "step": 1005 }, { "epoch": 0.56, "learning_rate": 0.00026877576257213517, "loss": 0.1592, "step": 1006 }, { "epoch": 0.56, "learning_rate": 0.00026871393239901066, "loss": 0.1588, "step": 1007 }, { "epoch": 0.57, "learning_rate": 0.0002686521022258862, "loss": 0.1508, "step": 1008 }, { "epoch": 0.57, "learning_rate": 0.00026859027205276175, "loss": 0.1683, "step": 1009 }, { "epoch": 0.57, "learning_rate": 0.00026852844187963724, "loss": 0.1438, "step": 1010 }, { "epoch": 0.57, "learning_rate": 0.0002684666117065128, "loss": 0.1623, "step": 1011 }, { "epoch": 0.57, "learning_rate": 0.00026840478153338827, "loss": 0.1554, "step": 1012 }, { "epoch": 0.57, "learning_rate": 0.00026834295136026376, "loss": 0.1266, "step": 1013 }, { "epoch": 0.57, "learning_rate": 0.0002682811211871393, "loss": 0.1623, "step": 1014 }, { "epoch": 0.57, "learning_rate": 0.00026821929101401485, "loss": 0.1586, "step": 1015 }, { "epoch": 0.57, "learning_rate": 0.00026815746084089034, "loss": 0.1346, "step": 1016 }, { "epoch": 0.57, "learning_rate": 0.00026809563066776583, "loss": 0.144, "step": 1017 }, { "epoch": 0.57, "learning_rate": 0.00026803380049464137, "loss": 0.1977, "step": 1018 }, { "epoch": 0.57, "learning_rate": 0.00026797197032151686, "loss": 0.171, "step": 1019 }, { "epoch": 0.57, "learning_rate": 0.0002679101401483924, "loss": 0.1698, "step": 1020 }, { "epoch": 0.57, "learning_rate": 0.0002678483099752679, "loss": 0.1318, "step": 1021 }, { "epoch": 0.57, "learning_rate": 0.00026778647980214344, "loss": 0.1379, "step": 1022 }, { "epoch": 0.57, "learning_rate": 0.0002677246496290189, "loss": 0.1181, "step": 1023 }, { "epoch": 0.57, "learning_rate": 0.00026766281945589447, "loss": 0.1219, "step": 1024 }, { "epoch": 0.57, "learning_rate": 0.00026760098928276996, "loss": 0.1461, "step": 1025 }, { "epoch": 0.58, "learning_rate": 0.00026753915910964545, "loss": 0.1434, "step": 1026 }, { "epoch": 0.58, "learning_rate": 0.000267477328936521, "loss": 0.1069, "step": 1027 }, { "epoch": 0.58, "learning_rate": 0.00026741549876339654, "loss": 0.1363, "step": 1028 }, { "epoch": 0.58, "learning_rate": 0.000267353668590272, "loss": 0.1341, "step": 1029 }, { "epoch": 0.58, "learning_rate": 0.0002672918384171475, "loss": 0.178, "step": 1030 }, { "epoch": 0.58, "learning_rate": 0.00026723000824402306, "loss": 0.1752, "step": 1031 }, { "epoch": 0.58, "learning_rate": 0.00026716817807089855, "loss": 0.1378, "step": 1032 }, { "epoch": 0.58, "learning_rate": 0.0002671063478977741, "loss": 0.162, "step": 1033 }, { "epoch": 0.58, "learning_rate": 0.00026704451772464964, "loss": 0.1021, "step": 1034 }, { "epoch": 0.58, "learning_rate": 0.0002669826875515251, "loss": 0.1515, "step": 1035 }, { "epoch": 0.58, "learning_rate": 0.0002669208573784006, "loss": 0.1608, "step": 1036 }, { "epoch": 0.58, "learning_rate": 0.00026685902720527616, "loss": 0.136, "step": 1037 }, { "epoch": 0.58, "learning_rate": 0.0002667971970321517, "loss": 0.1267, "step": 1038 }, { "epoch": 0.58, "learning_rate": 0.0002667353668590272, "loss": 0.1353, "step": 1039 }, { "epoch": 0.58, "learning_rate": 0.0002666735366859027, "loss": 0.1803, "step": 1040 }, { "epoch": 0.58, "learning_rate": 0.0002666117065127782, "loss": 0.1244, "step": 1041 }, { "epoch": 0.58, "learning_rate": 0.0002665498763396537, "loss": 0.1441, "step": 1042 }, { "epoch": 0.58, "learning_rate": 0.00026648804616652926, "loss": 0.1978, "step": 1043 }, { "epoch": 0.59, "learning_rate": 0.00026642621599340475, "loss": 0.1371, "step": 1044 }, { "epoch": 0.59, "learning_rate": 0.0002663643858202803, "loss": 0.3337, "step": 1045 }, { "epoch": 0.59, "learning_rate": 0.0002663025556471558, "loss": 0.1758, "step": 1046 }, { "epoch": 0.59, "learning_rate": 0.0002662407254740313, "loss": 0.2584, "step": 1047 }, { "epoch": 0.59, "learning_rate": 0.0002661788953009068, "loss": 0.1847, "step": 1048 }, { "epoch": 0.59, "learning_rate": 0.0002661170651277823, "loss": 0.4219, "step": 1049 }, { "epoch": 0.59, "learning_rate": 0.00026605523495465785, "loss": 0.5732, "step": 1050 }, { "epoch": 0.59, "learning_rate": 0.0002659934047815334, "loss": 0.2307, "step": 1051 }, { "epoch": 0.59, "learning_rate": 0.0002659315746084089, "loss": 0.1866, "step": 1052 }, { "epoch": 0.59, "learning_rate": 0.00026586974443528437, "loss": 0.131, "step": 1053 }, { "epoch": 0.59, "learning_rate": 0.0002658079142621599, "loss": 0.18, "step": 1054 }, { "epoch": 0.59, "learning_rate": 0.0002657460840890354, "loss": 0.1296, "step": 1055 }, { "epoch": 0.59, "learning_rate": 0.00026568425391591095, "loss": 0.1688, "step": 1056 }, { "epoch": 0.59, "learning_rate": 0.0002656224237427865, "loss": 0.2268, "step": 1057 }, { "epoch": 0.59, "learning_rate": 0.000265560593569662, "loss": 0.1386, "step": 1058 }, { "epoch": 0.59, "learning_rate": 0.00026549876339653747, "loss": 0.1561, "step": 1059 }, { "epoch": 0.59, "learning_rate": 0.000265436933223413, "loss": 0.129, "step": 1060 }, { "epoch": 0.59, "learning_rate": 0.0002653751030502885, "loss": 0.1258, "step": 1061 }, { "epoch": 0.6, "learning_rate": 0.000265313272877164, "loss": 0.1279, "step": 1062 }, { "epoch": 0.6, "learning_rate": 0.00026525144270403954, "loss": 0.191, "step": 1063 }, { "epoch": 0.6, "learning_rate": 0.0002651896125309151, "loss": 0.0892, "step": 1064 }, { "epoch": 0.6, "learning_rate": 0.00026512778235779057, "loss": 0.1443, "step": 1065 }, { "epoch": 0.6, "learning_rate": 0.0002650659521846661, "loss": 0.1466, "step": 1066 }, { "epoch": 0.6, "learning_rate": 0.0002650041220115416, "loss": 0.1235, "step": 1067 }, { "epoch": 0.6, "learning_rate": 0.00026494229183841715, "loss": 0.1164, "step": 1068 }, { "epoch": 0.6, "learning_rate": 0.00026488046166529264, "loss": 0.1335, "step": 1069 }, { "epoch": 0.6, "learning_rate": 0.0002648186314921682, "loss": 0.1887, "step": 1070 }, { "epoch": 0.6, "learning_rate": 0.00026475680131904367, "loss": 0.1536, "step": 1071 }, { "epoch": 0.6, "learning_rate": 0.00026469497114591916, "loss": 0.0871, "step": 1072 }, { "epoch": 0.6, "learning_rate": 0.0002646331409727947, "loss": 0.1611, "step": 1073 }, { "epoch": 0.6, "learning_rate": 0.00026457131079967025, "loss": 0.1251, "step": 1074 }, { "epoch": 0.6, "learning_rate": 0.00026450948062654574, "loss": 0.1031, "step": 1075 }, { "epoch": 0.6, "learning_rate": 0.0002644476504534212, "loss": 0.1244, "step": 1076 }, { "epoch": 0.6, "learning_rate": 0.00026438582028029677, "loss": 0.1212, "step": 1077 }, { "epoch": 0.6, "learning_rate": 0.00026432399010717226, "loss": 0.1339, "step": 1078 }, { "epoch": 0.6, "learning_rate": 0.0002642621599340478, "loss": 0.1607, "step": 1079 }, { "epoch": 0.61, "learning_rate": 0.00026420032976092335, "loss": 0.1304, "step": 1080 }, { "epoch": 0.61, "learning_rate": 0.00026413849958779883, "loss": 0.1447, "step": 1081 }, { "epoch": 0.61, "learning_rate": 0.0002640766694146743, "loss": 0.1728, "step": 1082 }, { "epoch": 0.61, "learning_rate": 0.00026401483924154987, "loss": 0.1152, "step": 1083 }, { "epoch": 0.61, "learning_rate": 0.00026395300906842536, "loss": 0.1535, "step": 1084 }, { "epoch": 0.61, "learning_rate": 0.00026389117889530085, "loss": 0.1176, "step": 1085 }, { "epoch": 0.61, "learning_rate": 0.0002638293487221764, "loss": 0.1888, "step": 1086 }, { "epoch": 0.61, "learning_rate": 0.00026376751854905193, "loss": 0.1254, "step": 1087 }, { "epoch": 0.61, "learning_rate": 0.0002637056883759274, "loss": 0.1827, "step": 1088 }, { "epoch": 0.61, "learning_rate": 0.00026364385820280297, "loss": 0.1821, "step": 1089 }, { "epoch": 0.61, "learning_rate": 0.00026358202802967846, "loss": 0.1329, "step": 1090 }, { "epoch": 0.61, "learning_rate": 0.00026352019785655395, "loss": 0.1203, "step": 1091 }, { "epoch": 0.61, "learning_rate": 0.0002634583676834295, "loss": 0.1623, "step": 1092 }, { "epoch": 0.61, "learning_rate": 0.00026339653751030503, "loss": 0.1886, "step": 1093 }, { "epoch": 0.61, "learning_rate": 0.0002633347073371805, "loss": 0.1489, "step": 1094 }, { "epoch": 0.61, "learning_rate": 0.000263272877164056, "loss": 0.1599, "step": 1095 }, { "epoch": 0.61, "learning_rate": 0.00026321104699093156, "loss": 0.2067, "step": 1096 }, { "epoch": 0.61, "learning_rate": 0.0002631492168178071, "loss": 0.2604, "step": 1097 }, { "epoch": 0.62, "learning_rate": 0.0002630873866446826, "loss": 0.2436, "step": 1098 }, { "epoch": 0.62, "learning_rate": 0.0002630255564715581, "loss": 0.2807, "step": 1099 }, { "epoch": 0.62, "learning_rate": 0.0002629637262984336, "loss": 0.302, "step": 1100 }, { "epoch": 0.62, "learning_rate": 0.0002629018961253091, "loss": 0.1704, "step": 1101 }, { "epoch": 0.62, "learning_rate": 0.00026284006595218466, "loss": 0.1716, "step": 1102 }, { "epoch": 0.62, "learning_rate": 0.00026277823577906015, "loss": 0.173, "step": 1103 }, { "epoch": 0.62, "learning_rate": 0.0002627164056059357, "loss": 0.17, "step": 1104 }, { "epoch": 0.62, "learning_rate": 0.0002626545754328112, "loss": 0.1496, "step": 1105 }, { "epoch": 0.62, "learning_rate": 0.0002625927452596867, "loss": 0.1111, "step": 1106 }, { "epoch": 0.62, "learning_rate": 0.0002625309150865622, "loss": 0.1194, "step": 1107 }, { "epoch": 0.62, "learning_rate": 0.0002624690849134377, "loss": 0.1692, "step": 1108 }, { "epoch": 0.62, "learning_rate": 0.00026240725474031325, "loss": 0.1473, "step": 1109 }, { "epoch": 0.62, "learning_rate": 0.0002623454245671888, "loss": 0.1237, "step": 1110 }, { "epoch": 0.62, "learning_rate": 0.0002622835943940643, "loss": 0.136, "step": 1111 }, { "epoch": 0.62, "learning_rate": 0.0002622217642209398, "loss": 0.1533, "step": 1112 }, { "epoch": 0.62, "learning_rate": 0.0002621599340478153, "loss": 0.1425, "step": 1113 }, { "epoch": 0.62, "learning_rate": 0.0002620981038746908, "loss": 0.1641, "step": 1114 }, { "epoch": 0.62, "learning_rate": 0.00026203627370156634, "loss": 0.1519, "step": 1115 }, { "epoch": 0.63, "learning_rate": 0.0002619744435284419, "loss": 0.1188, "step": 1116 }, { "epoch": 0.63, "learning_rate": 0.0002619126133553174, "loss": 0.1157, "step": 1117 }, { "epoch": 0.63, "learning_rate": 0.00026185078318219287, "loss": 0.1615, "step": 1118 }, { "epoch": 0.63, "learning_rate": 0.0002617889530090684, "loss": 0.1464, "step": 1119 }, { "epoch": 0.63, "learning_rate": 0.0002617271228359439, "loss": 0.1389, "step": 1120 }, { "epoch": 0.63, "learning_rate": 0.00026166529266281944, "loss": 0.1226, "step": 1121 }, { "epoch": 0.63, "learning_rate": 0.00026160346248969493, "loss": 0.1254, "step": 1122 }, { "epoch": 0.63, "learning_rate": 0.0002615416323165705, "loss": 0.1128, "step": 1123 }, { "epoch": 0.63, "learning_rate": 0.00026147980214344597, "loss": 0.1166, "step": 1124 }, { "epoch": 0.63, "learning_rate": 0.0002614179719703215, "loss": 0.1408, "step": 1125 }, { "epoch": 0.63, "learning_rate": 0.000261356141797197, "loss": 0.1145, "step": 1126 }, { "epoch": 0.63, "learning_rate": 0.0002612943116240725, "loss": 0.1185, "step": 1127 }, { "epoch": 0.63, "learning_rate": 0.00026123248145094803, "loss": 0.1418, "step": 1128 }, { "epoch": 0.63, "learning_rate": 0.0002611706512778236, "loss": 0.1496, "step": 1129 }, { "epoch": 0.63, "learning_rate": 0.00026110882110469907, "loss": 0.1786, "step": 1130 }, { "epoch": 0.63, "learning_rate": 0.00026104699093157456, "loss": 0.1519, "step": 1131 }, { "epoch": 0.63, "learning_rate": 0.0002609851607584501, "loss": 0.1846, "step": 1132 }, { "epoch": 0.64, "learning_rate": 0.00026092333058532564, "loss": 0.1241, "step": 1133 }, { "epoch": 0.64, "learning_rate": 0.00026086150041220113, "loss": 0.1153, "step": 1134 }, { "epoch": 0.64, "learning_rate": 0.0002607996702390766, "loss": 0.1747, "step": 1135 }, { "epoch": 0.64, "learning_rate": 0.00026073784006595217, "loss": 0.124, "step": 1136 }, { "epoch": 0.64, "learning_rate": 0.00026067600989282766, "loss": 0.1917, "step": 1137 }, { "epoch": 0.64, "learning_rate": 0.0002606141797197032, "loss": 0.1378, "step": 1138 }, { "epoch": 0.64, "learning_rate": 0.00026055234954657874, "loss": 0.1458, "step": 1139 }, { "epoch": 0.64, "learning_rate": 0.00026049051937345423, "loss": 0.167, "step": 1140 }, { "epoch": 0.64, "learning_rate": 0.0002604286892003297, "loss": 0.1756, "step": 1141 }, { "epoch": 0.64, "learning_rate": 0.00026036685902720527, "loss": 0.1851, "step": 1142 }, { "epoch": 0.64, "learning_rate": 0.00026030502885408075, "loss": 0.1574, "step": 1143 }, { "epoch": 0.64, "learning_rate": 0.0002602431986809563, "loss": 0.208, "step": 1144 }, { "epoch": 0.64, "learning_rate": 0.0002601813685078318, "loss": 0.1467, "step": 1145 }, { "epoch": 0.64, "learning_rate": 0.00026011953833470733, "loss": 0.2386, "step": 1146 }, { "epoch": 0.64, "learning_rate": 0.0002600577081615828, "loss": 0.2821, "step": 1147 }, { "epoch": 0.64, "learning_rate": 0.00025999587798845836, "loss": 0.1893, "step": 1148 }, { "epoch": 0.64, "learning_rate": 0.00025993404781533385, "loss": 0.3428, "step": 1149 }, { "epoch": 0.64, "learning_rate": 0.00025987221764220934, "loss": 0.3464, "step": 1150 }, { "epoch": 0.65, "learning_rate": 0.0002598103874690849, "loss": 0.1533, "step": 1151 }, { "epoch": 0.65, "learning_rate": 0.00025974855729596043, "loss": 0.1426, "step": 1152 }, { "epoch": 0.65, "learning_rate": 0.0002596867271228359, "loss": 0.1388, "step": 1153 }, { "epoch": 0.65, "learning_rate": 0.0002596248969497114, "loss": 0.1706, "step": 1154 }, { "epoch": 0.65, "learning_rate": 0.00025956306677658695, "loss": 0.1424, "step": 1155 }, { "epoch": 0.65, "learning_rate": 0.00025950123660346244, "loss": 0.1256, "step": 1156 }, { "epoch": 0.65, "learning_rate": 0.000259439406430338, "loss": 0.1407, "step": 1157 }, { "epoch": 0.65, "learning_rate": 0.0002593775762572135, "loss": 0.135, "step": 1158 }, { "epoch": 0.65, "learning_rate": 0.000259315746084089, "loss": 0.0996, "step": 1159 }, { "epoch": 0.65, "learning_rate": 0.0002592539159109645, "loss": 0.1252, "step": 1160 }, { "epoch": 0.65, "learning_rate": 0.00025919208573784005, "loss": 0.1478, "step": 1161 }, { "epoch": 0.65, "learning_rate": 0.0002591302555647156, "loss": 0.1126, "step": 1162 }, { "epoch": 0.65, "learning_rate": 0.0002590684253915911, "loss": 0.1403, "step": 1163 }, { "epoch": 0.65, "learning_rate": 0.0002590065952184666, "loss": 0.1153, "step": 1164 }, { "epoch": 0.65, "learning_rate": 0.0002589447650453421, "loss": 0.1144, "step": 1165 }, { "epoch": 0.65, "learning_rate": 0.0002588829348722176, "loss": 0.1109, "step": 1166 }, { "epoch": 0.65, "learning_rate": 0.00025882110469909315, "loss": 0.1733, "step": 1167 }, { "epoch": 0.65, "learning_rate": 0.00025875927452596864, "loss": 0.1056, "step": 1168 }, { "epoch": 0.66, "learning_rate": 0.0002586974443528442, "loss": 0.1361, "step": 1169 }, { "epoch": 0.66, "learning_rate": 0.0002586356141797197, "loss": 0.1261, "step": 1170 }, { "epoch": 0.66, "learning_rate": 0.0002585737840065952, "loss": 0.1458, "step": 1171 }, { "epoch": 0.66, "learning_rate": 0.0002585119538334707, "loss": 0.1338, "step": 1172 }, { "epoch": 0.66, "learning_rate": 0.0002584501236603462, "loss": 0.1128, "step": 1173 }, { "epoch": 0.66, "learning_rate": 0.00025838829348722174, "loss": 0.1315, "step": 1174 }, { "epoch": 0.66, "learning_rate": 0.0002583264633140973, "loss": 0.158, "step": 1175 }, { "epoch": 0.66, "learning_rate": 0.0002582646331409728, "loss": 0.1095, "step": 1176 }, { "epoch": 0.66, "learning_rate": 0.00025820280296784826, "loss": 0.1196, "step": 1177 }, { "epoch": 0.66, "learning_rate": 0.0002581409727947238, "loss": 0.1441, "step": 1178 }, { "epoch": 0.66, "learning_rate": 0.0002580791426215993, "loss": 0.1078, "step": 1179 }, { "epoch": 0.66, "learning_rate": 0.00025801731244847484, "loss": 0.132, "step": 1180 }, { "epoch": 0.66, "learning_rate": 0.00025795548227535033, "loss": 0.133, "step": 1181 }, { "epoch": 0.66, "learning_rate": 0.0002578936521022259, "loss": 0.1084, "step": 1182 }, { "epoch": 0.66, "learning_rate": 0.00025783182192910136, "loss": 0.1194, "step": 1183 }, { "epoch": 0.66, "learning_rate": 0.0002577699917559769, "loss": 0.1053, "step": 1184 }, { "epoch": 0.66, "learning_rate": 0.0002577081615828524, "loss": 0.2066, "step": 1185 }, { "epoch": 0.66, "learning_rate": 0.0002576463314097279, "loss": 0.1253, "step": 1186 }, { "epoch": 0.67, "learning_rate": 0.00025758450123660343, "loss": 0.1289, "step": 1187 }, { "epoch": 0.67, "learning_rate": 0.000257522671063479, "loss": 0.1323, "step": 1188 }, { "epoch": 0.67, "learning_rate": 0.00025746084089035446, "loss": 0.1483, "step": 1189 }, { "epoch": 0.67, "learning_rate": 0.00025739901071722995, "loss": 0.1379, "step": 1190 }, { "epoch": 0.67, "learning_rate": 0.0002573371805441055, "loss": 0.129, "step": 1191 }, { "epoch": 0.67, "learning_rate": 0.00025727535037098104, "loss": 0.1165, "step": 1192 }, { "epoch": 0.67, "learning_rate": 0.00025721352019785653, "loss": 0.1729, "step": 1193 }, { "epoch": 0.67, "learning_rate": 0.0002571516900247321, "loss": 0.179, "step": 1194 }, { "epoch": 0.67, "learning_rate": 0.00025708985985160756, "loss": 0.0853, "step": 1195 }, { "epoch": 0.67, "learning_rate": 0.00025702802967848305, "loss": 0.1471, "step": 1196 }, { "epoch": 0.67, "learning_rate": 0.0002569661995053586, "loss": 0.1442, "step": 1197 }, { "epoch": 0.67, "learning_rate": 0.00025690436933223414, "loss": 0.3266, "step": 1198 }, { "epoch": 0.67, "learning_rate": 0.00025684253915910963, "loss": 0.3439, "step": 1199 }, { "epoch": 0.67, "learning_rate": 0.0002567807089859851, "loss": 0.1648, "step": 1200 }, { "epoch": 0.67, "eval_loss": 0.15760938823223114, "eval_runtime": 273.1625, "eval_samples_per_second": 9.672, "eval_steps_per_second": 0.608, "eval_wer": 0.13064260214200715, "step": 1200 }, { "epoch": 0.67, "learning_rate": 0.00025671887881286066, "loss": 0.1318, "step": 1201 }, { "epoch": 0.67, "learning_rate": 0.00025665704863973615, "loss": 0.1433, "step": 1202 }, { "epoch": 0.67, "learning_rate": 0.0002565952184666117, "loss": 0.1535, "step": 1203 }, { "epoch": 0.67, "learning_rate": 0.0002565333882934872, "loss": 0.1229, "step": 1204 }, { "epoch": 0.68, "learning_rate": 0.00025647155812036273, "loss": 0.1689, "step": 1205 }, { "epoch": 0.68, "learning_rate": 0.0002564097279472382, "loss": 0.1536, "step": 1206 }, { "epoch": 0.68, "learning_rate": 0.00025634789777411376, "loss": 0.1722, "step": 1207 }, { "epoch": 0.68, "learning_rate": 0.00025628606760098925, "loss": 0.1426, "step": 1208 }, { "epoch": 0.68, "learning_rate": 0.00025622423742786474, "loss": 0.1374, "step": 1209 }, { "epoch": 0.68, "learning_rate": 0.0002561624072547403, "loss": 0.1259, "step": 1210 }, { "epoch": 0.68, "learning_rate": 0.00025610057708161583, "loss": 0.1486, "step": 1211 }, { "epoch": 0.68, "learning_rate": 0.0002560387469084913, "loss": 0.0819, "step": 1212 }, { "epoch": 0.68, "learning_rate": 0.0002559769167353668, "loss": 0.139, "step": 1213 }, { "epoch": 0.68, "learning_rate": 0.00025591508656224235, "loss": 0.1421, "step": 1214 }, { "epoch": 0.68, "learning_rate": 0.00025585325638911784, "loss": 0.1257, "step": 1215 }, { "epoch": 0.68, "learning_rate": 0.0002557914262159934, "loss": 0.1213, "step": 1216 }, { "epoch": 0.68, "learning_rate": 0.00025572959604286893, "loss": 0.102, "step": 1217 }, { "epoch": 0.68, "learning_rate": 0.0002556677658697444, "loss": 0.089, "step": 1218 }, { "epoch": 0.68, "learning_rate": 0.0002556059356966199, "loss": 0.1249, "step": 1219 }, { "epoch": 0.68, "learning_rate": 0.00025554410552349545, "loss": 0.1045, "step": 1220 }, { "epoch": 0.68, "learning_rate": 0.000255482275350371, "loss": 0.1166, "step": 1221 }, { "epoch": 0.68, "learning_rate": 0.0002554204451772465, "loss": 0.1145, "step": 1222 }, { "epoch": 0.69, "learning_rate": 0.000255358615004122, "loss": 0.1231, "step": 1223 }, { "epoch": 0.69, "learning_rate": 0.0002552967848309975, "loss": 0.1018, "step": 1224 }, { "epoch": 0.69, "learning_rate": 0.000255234954657873, "loss": 0.1156, "step": 1225 }, { "epoch": 0.69, "learning_rate": 0.00025517312448474855, "loss": 0.1395, "step": 1226 }, { "epoch": 0.69, "learning_rate": 0.00025511129431162404, "loss": 0.1229, "step": 1227 }, { "epoch": 0.69, "learning_rate": 0.0002550494641384996, "loss": 0.113, "step": 1228 }, { "epoch": 0.69, "learning_rate": 0.00025498763396537507, "loss": 0.1237, "step": 1229 }, { "epoch": 0.69, "learning_rate": 0.0002549258037922506, "loss": 0.1636, "step": 1230 }, { "epoch": 0.69, "learning_rate": 0.0002548639736191261, "loss": 0.1239, "step": 1231 }, { "epoch": 0.69, "learning_rate": 0.0002548021434460016, "loss": 0.1558, "step": 1232 }, { "epoch": 0.69, "learning_rate": 0.00025474031327287714, "loss": 0.1969, "step": 1233 }, { "epoch": 0.69, "learning_rate": 0.0002546784830997527, "loss": 0.1241, "step": 1234 }, { "epoch": 0.69, "learning_rate": 0.00025461665292662817, "loss": 0.1263, "step": 1235 }, { "epoch": 0.69, "learning_rate": 0.00025455482275350366, "loss": 0.1275, "step": 1236 }, { "epoch": 0.69, "learning_rate": 0.0002544929925803792, "loss": 0.1466, "step": 1237 }, { "epoch": 0.69, "learning_rate": 0.0002544311624072547, "loss": 0.0893, "step": 1238 }, { "epoch": 0.69, "learning_rate": 0.00025436933223413024, "loss": 0.114, "step": 1239 }, { "epoch": 0.7, "learning_rate": 0.0002543075020610058, "loss": 0.0919, "step": 1240 }, { "epoch": 0.7, "learning_rate": 0.00025424567188788127, "loss": 0.1613, "step": 1241 }, { "epoch": 0.7, "learning_rate": 0.00025418384171475676, "loss": 0.2013, "step": 1242 }, { "epoch": 0.7, "learning_rate": 0.0002541220115416323, "loss": 0.177, "step": 1243 }, { "epoch": 0.7, "learning_rate": 0.0002540601813685078, "loss": 0.1502, "step": 1244 }, { "epoch": 0.7, "learning_rate": 0.0002539983511953833, "loss": 0.1079, "step": 1245 }, { "epoch": 0.7, "learning_rate": 0.00025393652102225883, "loss": 0.1157, "step": 1246 }, { "epoch": 0.7, "learning_rate": 0.00025387469084913437, "loss": 0.2448, "step": 1247 }, { "epoch": 0.7, "learning_rate": 0.00025381286067600986, "loss": 0.2866, "step": 1248 }, { "epoch": 0.7, "learning_rate": 0.0002537510305028854, "loss": 0.2697, "step": 1249 }, { "epoch": 0.7, "learning_rate": 0.0002536892003297609, "loss": 0.2761, "step": 1250 }, { "epoch": 0.7, "learning_rate": 0.00025362737015663644, "loss": 0.1119, "step": 1251 }, { "epoch": 0.7, "learning_rate": 0.00025356553998351193, "loss": 0.138, "step": 1252 }, { "epoch": 0.7, "learning_rate": 0.00025350370981038747, "loss": 0.1474, "step": 1253 }, { "epoch": 0.7, "learning_rate": 0.00025344187963726296, "loss": 0.136, "step": 1254 }, { "epoch": 0.7, "learning_rate": 0.00025338004946413845, "loss": 0.1787, "step": 1255 }, { "epoch": 0.7, "learning_rate": 0.000253318219291014, "loss": 0.1111, "step": 1256 }, { "epoch": 0.7, "learning_rate": 0.00025325638911788954, "loss": 0.1201, "step": 1257 }, { "epoch": 0.71, "learning_rate": 0.000253194558944765, "loss": 0.1411, "step": 1258 }, { "epoch": 0.71, "learning_rate": 0.0002531327287716405, "loss": 0.1402, "step": 1259 }, { "epoch": 0.71, "learning_rate": 0.00025307089859851606, "loss": 0.1572, "step": 1260 }, { "epoch": 0.71, "learning_rate": 0.00025300906842539155, "loss": 0.1007, "step": 1261 }, { "epoch": 0.71, "learning_rate": 0.0002529472382522671, "loss": 0.1487, "step": 1262 }, { "epoch": 0.71, "learning_rate": 0.00025288540807914264, "loss": 0.1222, "step": 1263 }, { "epoch": 0.71, "learning_rate": 0.0002528235779060181, "loss": 0.13, "step": 1264 }, { "epoch": 0.71, "learning_rate": 0.0002527617477328936, "loss": 0.1134, "step": 1265 }, { "epoch": 0.71, "learning_rate": 0.00025269991755976916, "loss": 0.1175, "step": 1266 }, { "epoch": 0.71, "learning_rate": 0.00025263808738664465, "loss": 0.1455, "step": 1267 }, { "epoch": 0.71, "learning_rate": 0.00025257625721352014, "loss": 0.1412, "step": 1268 }, { "epoch": 0.71, "learning_rate": 0.0002525144270403957, "loss": 0.115, "step": 1269 }, { "epoch": 0.71, "learning_rate": 0.0002524525968672712, "loss": 0.1306, "step": 1270 }, { "epoch": 0.71, "learning_rate": 0.0002523907666941467, "loss": 0.0876, "step": 1271 }, { "epoch": 0.71, "learning_rate": 0.00025232893652102226, "loss": 0.1209, "step": 1272 }, { "epoch": 0.71, "learning_rate": 0.00025226710634789775, "loss": 0.1274, "step": 1273 }, { "epoch": 0.71, "learning_rate": 0.00025220527617477324, "loss": 0.1041, "step": 1274 }, { "epoch": 0.71, "learning_rate": 0.0002521434460016488, "loss": 0.1386, "step": 1275 }, { "epoch": 0.72, "learning_rate": 0.0002520816158285243, "loss": 0.1139, "step": 1276 }, { "epoch": 0.72, "learning_rate": 0.0002520197856553998, "loss": 0.1136, "step": 1277 }, { "epoch": 0.72, "learning_rate": 0.0002519579554822753, "loss": 0.1099, "step": 1278 }, { "epoch": 0.72, "learning_rate": 0.00025189612530915085, "loss": 0.1251, "step": 1279 }, { "epoch": 0.72, "learning_rate": 0.0002518342951360264, "loss": 0.1099, "step": 1280 }, { "epoch": 0.72, "learning_rate": 0.0002517724649629019, "loss": 0.0563, "step": 1281 }, { "epoch": 0.72, "learning_rate": 0.00025171063478977737, "loss": 0.1199, "step": 1282 }, { "epoch": 0.72, "learning_rate": 0.0002516488046166529, "loss": 0.1591, "step": 1283 }, { "epoch": 0.72, "learning_rate": 0.0002515869744435284, "loss": 0.1077, "step": 1284 }, { "epoch": 0.72, "learning_rate": 0.00025152514427040395, "loss": 0.0946, "step": 1285 }, { "epoch": 0.72, "learning_rate": 0.0002514633140972795, "loss": 0.1277, "step": 1286 }, { "epoch": 0.72, "learning_rate": 0.000251401483924155, "loss": 0.1174, "step": 1287 }, { "epoch": 0.72, "learning_rate": 0.00025133965375103047, "loss": 0.0932, "step": 1288 }, { "epoch": 0.72, "learning_rate": 0.000251277823577906, "loss": 0.1028, "step": 1289 }, { "epoch": 0.72, "learning_rate": 0.0002512159934047815, "loss": 0.0698, "step": 1290 }, { "epoch": 0.72, "learning_rate": 0.000251154163231657, "loss": 0.1178, "step": 1291 }, { "epoch": 0.72, "learning_rate": 0.00025109233305853254, "loss": 0.1202, "step": 1292 }, { "epoch": 0.72, "learning_rate": 0.0002510305028854081, "loss": 0.2176, "step": 1293 }, { "epoch": 0.73, "learning_rate": 0.00025096867271228357, "loss": 0.1342, "step": 1294 }, { "epoch": 0.73, "learning_rate": 0.0002509068425391591, "loss": 0.1234, "step": 1295 }, { "epoch": 0.73, "learning_rate": 0.0002508450123660346, "loss": 0.0931, "step": 1296 }, { "epoch": 0.73, "learning_rate": 0.0002507831821929101, "loss": 0.0964, "step": 1297 }, { "epoch": 0.73, "learning_rate": 0.00025072135201978564, "loss": 0.1205, "step": 1298 }, { "epoch": 0.73, "learning_rate": 0.0002506595218466612, "loss": 0.419, "step": 1299 }, { "epoch": 0.73, "learning_rate": 0.00025059769167353667, "loss": 0.3994, "step": 1300 }, { "epoch": 0.73, "learning_rate": 0.00025053586150041216, "loss": 0.1459, "step": 1301 }, { "epoch": 0.73, "learning_rate": 0.0002504740313272877, "loss": 0.1509, "step": 1302 }, { "epoch": 0.73, "learning_rate": 0.0002504122011541632, "loss": 0.1403, "step": 1303 }, { "epoch": 0.73, "learning_rate": 0.00025035037098103874, "loss": 0.1498, "step": 1304 }, { "epoch": 0.73, "learning_rate": 0.0002502885408079142, "loss": 0.1234, "step": 1305 }, { "epoch": 0.73, "learning_rate": 0.00025022671063478977, "loss": 0.1305, "step": 1306 }, { "epoch": 0.73, "learning_rate": 0.00025016488046166526, "loss": 0.0836, "step": 1307 }, { "epoch": 0.73, "learning_rate": 0.0002501030502885408, "loss": 0.1359, "step": 1308 }, { "epoch": 0.73, "learning_rate": 0.0002500412201154163, "loss": 0.1332, "step": 1309 }, { "epoch": 0.73, "learning_rate": 0.0002499793899422918, "loss": 0.1247, "step": 1310 }, { "epoch": 0.73, "learning_rate": 0.0002499175597691673, "loss": 0.1141, "step": 1311 }, { "epoch": 0.74, "learning_rate": 0.00024985572959604287, "loss": 0.1535, "step": 1312 }, { "epoch": 0.74, "learning_rate": 0.00024979389942291836, "loss": 0.1196, "step": 1313 }, { "epoch": 0.74, "learning_rate": 0.00024973206924979385, "loss": 0.1027, "step": 1314 }, { "epoch": 0.74, "learning_rate": 0.0002496702390766694, "loss": 0.1381, "step": 1315 }, { "epoch": 0.74, "learning_rate": 0.00024960840890354493, "loss": 0.1245, "step": 1316 }, { "epoch": 0.74, "learning_rate": 0.0002495465787304204, "loss": 0.1, "step": 1317 }, { "epoch": 0.74, "learning_rate": 0.00024948474855729597, "loss": 0.1478, "step": 1318 }, { "epoch": 0.74, "learning_rate": 0.00024942291838417146, "loss": 0.1146, "step": 1319 }, { "epoch": 0.74, "learning_rate": 0.00024936108821104695, "loss": 0.1428, "step": 1320 }, { "epoch": 0.74, "learning_rate": 0.0002492992580379225, "loss": 0.0961, "step": 1321 }, { "epoch": 0.74, "learning_rate": 0.00024923742786479803, "loss": 0.1366, "step": 1322 }, { "epoch": 0.74, "learning_rate": 0.0002491755976916735, "loss": 0.1097, "step": 1323 }, { "epoch": 0.74, "learning_rate": 0.000249113767518549, "loss": 0.1336, "step": 1324 }, { "epoch": 0.74, "learning_rate": 0.00024905193734542456, "loss": 0.1177, "step": 1325 }, { "epoch": 0.74, "learning_rate": 0.00024899010717230005, "loss": 0.1145, "step": 1326 }, { "epoch": 0.74, "learning_rate": 0.0002489282769991756, "loss": 0.1094, "step": 1327 }, { "epoch": 0.74, "learning_rate": 0.0002488664468260511, "loss": 0.1171, "step": 1328 }, { "epoch": 0.74, "learning_rate": 0.0002488046166529266, "loss": 0.1406, "step": 1329 }, { "epoch": 0.75, "learning_rate": 0.0002487427864798021, "loss": 0.1338, "step": 1330 }, { "epoch": 0.75, "learning_rate": 0.00024868095630667766, "loss": 0.1219, "step": 1331 }, { "epoch": 0.75, "learning_rate": 0.00024861912613355315, "loss": 0.0683, "step": 1332 }, { "epoch": 0.75, "learning_rate": 0.00024855729596042864, "loss": 0.141, "step": 1333 }, { "epoch": 0.75, "learning_rate": 0.0002484954657873042, "loss": 0.149, "step": 1334 }, { "epoch": 0.75, "learning_rate": 0.0002484336356141797, "loss": 0.1034, "step": 1335 }, { "epoch": 0.75, "learning_rate": 0.0002483718054410552, "loss": 0.1596, "step": 1336 }, { "epoch": 0.75, "learning_rate": 0.0002483099752679307, "loss": 0.1215, "step": 1337 }, { "epoch": 0.75, "learning_rate": 0.00024824814509480625, "loss": 0.1203, "step": 1338 }, { "epoch": 0.75, "learning_rate": 0.00024818631492168173, "loss": 0.1992, "step": 1339 }, { "epoch": 0.75, "learning_rate": 0.0002481244847485573, "loss": 0.1137, "step": 1340 }, { "epoch": 0.75, "learning_rate": 0.00024806265457543277, "loss": 0.1748, "step": 1341 }, { "epoch": 0.75, "learning_rate": 0.0002480008244023083, "loss": 0.1484, "step": 1342 }, { "epoch": 0.75, "learning_rate": 0.0002479389942291838, "loss": 0.1504, "step": 1343 }, { "epoch": 0.75, "learning_rate": 0.00024787716405605934, "loss": 0.1332, "step": 1344 }, { "epoch": 0.75, "learning_rate": 0.0002478153338829349, "loss": 0.1411, "step": 1345 }, { "epoch": 0.75, "learning_rate": 0.0002477535037098104, "loss": 0.1674, "step": 1346 }, { "epoch": 0.76, "learning_rate": 0.00024769167353668587, "loss": 0.2306, "step": 1347 }, { "epoch": 0.76, "learning_rate": 0.0002476298433635614, "loss": 0.1895, "step": 1348 }, { "epoch": 0.76, "learning_rate": 0.0002475680131904369, "loss": 0.3214, "step": 1349 }, { "epoch": 0.76, "learning_rate": 0.00024750618301731244, "loss": 0.3671, "step": 1350 }, { "epoch": 0.76, "learning_rate": 0.00024744435284418793, "loss": 0.1253, "step": 1351 }, { "epoch": 0.76, "learning_rate": 0.0002473825226710635, "loss": 0.1777, "step": 1352 }, { "epoch": 0.76, "learning_rate": 0.00024732069249793897, "loss": 0.1333, "step": 1353 }, { "epoch": 0.76, "learning_rate": 0.0002472588623248145, "loss": 0.1456, "step": 1354 }, { "epoch": 0.76, "learning_rate": 0.00024719703215169, "loss": 0.1476, "step": 1355 }, { "epoch": 0.76, "learning_rate": 0.0002471352019785655, "loss": 0.1668, "step": 1356 }, { "epoch": 0.76, "learning_rate": 0.00024707337180544103, "loss": 0.0972, "step": 1357 }, { "epoch": 0.76, "learning_rate": 0.0002470115416323166, "loss": 0.1599, "step": 1358 }, { "epoch": 0.76, "learning_rate": 0.00024694971145919207, "loss": 0.1305, "step": 1359 }, { "epoch": 0.76, "learning_rate": 0.00024688788128606756, "loss": 0.1488, "step": 1360 }, { "epoch": 0.76, "learning_rate": 0.0002468260511129431, "loss": 0.1186, "step": 1361 }, { "epoch": 0.76, "learning_rate": 0.0002467642209398186, "loss": 0.1177, "step": 1362 }, { "epoch": 0.76, "learning_rate": 0.00024670239076669413, "loss": 0.1274, "step": 1363 }, { "epoch": 0.76, "learning_rate": 0.0002466405605935696, "loss": 0.1582, "step": 1364 }, { "epoch": 0.77, "learning_rate": 0.00024657873042044517, "loss": 0.0866, "step": 1365 }, { "epoch": 0.77, "learning_rate": 0.00024651690024732066, "loss": 0.1448, "step": 1366 }, { "epoch": 0.77, "learning_rate": 0.0002464550700741962, "loss": 0.1459, "step": 1367 }, { "epoch": 0.77, "learning_rate": 0.0002463932399010717, "loss": 0.1079, "step": 1368 }, { "epoch": 0.77, "learning_rate": 0.0002463314097279472, "loss": 0.1451, "step": 1369 }, { "epoch": 0.77, "learning_rate": 0.0002462695795548227, "loss": 0.1132, "step": 1370 }, { "epoch": 0.77, "learning_rate": 0.00024620774938169827, "loss": 0.1147, "step": 1371 }, { "epoch": 0.77, "learning_rate": 0.00024614591920857375, "loss": 0.0882, "step": 1372 }, { "epoch": 0.77, "learning_rate": 0.00024608408903544924, "loss": 0.0955, "step": 1373 }, { "epoch": 0.77, "learning_rate": 0.0002460222588623248, "loss": 0.1019, "step": 1374 }, { "epoch": 0.77, "learning_rate": 0.00024596042868920033, "loss": 0.1053, "step": 1375 }, { "epoch": 0.77, "learning_rate": 0.0002458985985160758, "loss": 0.0919, "step": 1376 }, { "epoch": 0.77, "learning_rate": 0.00024583676834295137, "loss": 0.1075, "step": 1377 }, { "epoch": 0.77, "learning_rate": 0.00024577493816982685, "loss": 0.1309, "step": 1378 }, { "epoch": 0.77, "learning_rate": 0.00024571310799670234, "loss": 0.1289, "step": 1379 }, { "epoch": 0.77, "learning_rate": 0.0002456512778235779, "loss": 0.1197, "step": 1380 }, { "epoch": 0.77, "learning_rate": 0.00024558944765045343, "loss": 0.1372, "step": 1381 }, { "epoch": 0.77, "learning_rate": 0.0002455276174773289, "loss": 0.1759, "step": 1382 }, { "epoch": 0.78, "learning_rate": 0.0002454657873042044, "loss": 0.1269, "step": 1383 }, { "epoch": 0.78, "learning_rate": 0.00024540395713107995, "loss": 0.1463, "step": 1384 }, { "epoch": 0.78, "learning_rate": 0.00024534212695795544, "loss": 0.1528, "step": 1385 }, { "epoch": 0.78, "learning_rate": 0.000245280296784831, "loss": 0.1006, "step": 1386 }, { "epoch": 0.78, "learning_rate": 0.0002452184666117065, "loss": 0.1652, "step": 1387 }, { "epoch": 0.78, "learning_rate": 0.000245156636438582, "loss": 0.1237, "step": 1388 }, { "epoch": 0.78, "learning_rate": 0.0002450948062654575, "loss": 0.1242, "step": 1389 }, { "epoch": 0.78, "learning_rate": 0.00024503297609233305, "loss": 0.0944, "step": 1390 }, { "epoch": 0.78, "learning_rate": 0.00024497114591920854, "loss": 0.0969, "step": 1391 }, { "epoch": 0.78, "learning_rate": 0.00024490931574608403, "loss": 0.1564, "step": 1392 }, { "epoch": 0.78, "learning_rate": 0.0002448474855729596, "loss": 0.1229, "step": 1393 }, { "epoch": 0.78, "learning_rate": 0.0002447856553998351, "loss": 0.2018, "step": 1394 }, { "epoch": 0.78, "learning_rate": 0.0002447238252267106, "loss": 0.1977, "step": 1395 }, { "epoch": 0.78, "learning_rate": 0.0002446619950535861, "loss": 0.1767, "step": 1396 }, { "epoch": 0.78, "learning_rate": 0.00024460016488046164, "loss": 0.1683, "step": 1397 }, { "epoch": 0.78, "learning_rate": 0.00024453833470733713, "loss": 0.1263, "step": 1398 }, { "epoch": 0.78, "learning_rate": 0.0002444765045342127, "loss": 0.2258, "step": 1399 }, { "epoch": 0.78, "learning_rate": 0.0002444146743610882, "loss": 0.2436, "step": 1400 }, { "epoch": 0.79, "learning_rate": 0.0002443528441879637, "loss": 0.1114, "step": 1401 }, { "epoch": 0.79, "learning_rate": 0.0002442910140148392, "loss": 0.121, "step": 1402 }, { "epoch": 0.79, "learning_rate": 0.00024422918384171474, "loss": 0.118, "step": 1403 }, { "epoch": 0.79, "learning_rate": 0.0002441673536685903, "loss": 0.082, "step": 1404 }, { "epoch": 0.79, "learning_rate": 0.00024410552349546578, "loss": 0.1205, "step": 1405 }, { "epoch": 0.79, "learning_rate": 0.00024404369332234126, "loss": 0.1839, "step": 1406 }, { "epoch": 0.79, "learning_rate": 0.00024398186314921678, "loss": 0.1257, "step": 1407 }, { "epoch": 0.79, "learning_rate": 0.00024392003297609232, "loss": 0.1102, "step": 1408 }, { "epoch": 0.79, "learning_rate": 0.00024385820280296784, "loss": 0.1236, "step": 1409 }, { "epoch": 0.79, "learning_rate": 0.00024379637262984333, "loss": 0.121, "step": 1410 }, { "epoch": 0.79, "learning_rate": 0.00024373454245671885, "loss": 0.1389, "step": 1411 }, { "epoch": 0.79, "learning_rate": 0.00024367271228359436, "loss": 0.1118, "step": 1412 }, { "epoch": 0.79, "learning_rate": 0.0002436108821104699, "loss": 0.1342, "step": 1413 }, { "epoch": 0.79, "learning_rate": 0.00024354905193734542, "loss": 0.1167, "step": 1414 }, { "epoch": 0.79, "learning_rate": 0.00024348722176422091, "loss": 0.0926, "step": 1415 }, { "epoch": 0.79, "learning_rate": 0.00024342539159109643, "loss": 0.107, "step": 1416 }, { "epoch": 0.79, "learning_rate": 0.00024336356141797195, "loss": 0.2233, "step": 1417 }, { "epoch": 0.79, "learning_rate": 0.00024330173124484746, "loss": 0.1258, "step": 1418 }, { "epoch": 0.8, "learning_rate": 0.00024323990107172295, "loss": 0.0849, "step": 1419 }, { "epoch": 0.8, "learning_rate": 0.0002431780708985985, "loss": 0.1233, "step": 1420 }, { "epoch": 0.8, "learning_rate": 0.000243116240725474, "loss": 0.1565, "step": 1421 }, { "epoch": 0.8, "learning_rate": 0.00024305441055234953, "loss": 0.125, "step": 1422 }, { "epoch": 0.8, "learning_rate": 0.00024299258037922505, "loss": 0.1084, "step": 1423 }, { "epoch": 0.8, "learning_rate": 0.00024293075020610054, "loss": 0.1162, "step": 1424 }, { "epoch": 0.8, "learning_rate": 0.00024286892003297608, "loss": 0.0847, "step": 1425 }, { "epoch": 0.8, "learning_rate": 0.0002428070898598516, "loss": 0.1265, "step": 1426 }, { "epoch": 0.8, "learning_rate": 0.0002427452596867271, "loss": 0.1123, "step": 1427 }, { "epoch": 0.8, "learning_rate": 0.0002426834295136026, "loss": 0.1073, "step": 1428 }, { "epoch": 0.8, "learning_rate": 0.00024262159934047812, "loss": 0.1509, "step": 1429 }, { "epoch": 0.8, "learning_rate": 0.00024255976916735364, "loss": 0.1224, "step": 1430 }, { "epoch": 0.8, "learning_rate": 0.00024249793899422918, "loss": 0.1323, "step": 1431 }, { "epoch": 0.8, "learning_rate": 0.0002424361088211047, "loss": 0.1303, "step": 1432 }, { "epoch": 0.8, "learning_rate": 0.00024237427864798019, "loss": 0.1189, "step": 1433 }, { "epoch": 0.8, "learning_rate": 0.0002423124484748557, "loss": 0.1111, "step": 1434 }, { "epoch": 0.8, "learning_rate": 0.00024225061830173122, "loss": 0.1237, "step": 1435 }, { "epoch": 0.8, "learning_rate": 0.00024218878812860674, "loss": 0.0778, "step": 1436 }, { "epoch": 0.81, "learning_rate": 0.00024212695795548228, "loss": 0.1282, "step": 1437 }, { "epoch": 0.81, "learning_rate": 0.00024206512778235777, "loss": 0.0991, "step": 1438 }, { "epoch": 0.81, "learning_rate": 0.00024200329760923328, "loss": 0.1139, "step": 1439 }, { "epoch": 0.81, "learning_rate": 0.0002419414674361088, "loss": 0.1572, "step": 1440 }, { "epoch": 0.81, "learning_rate": 0.00024187963726298432, "loss": 0.1076, "step": 1441 }, { "epoch": 0.81, "learning_rate": 0.0002418178070898598, "loss": 0.1027, "step": 1442 }, { "epoch": 0.81, "learning_rate": 0.00024175597691673535, "loss": 0.1462, "step": 1443 }, { "epoch": 0.81, "learning_rate": 0.00024169414674361087, "loss": 0.1058, "step": 1444 }, { "epoch": 0.81, "learning_rate": 0.00024163231657048638, "loss": 0.1865, "step": 1445 }, { "epoch": 0.81, "learning_rate": 0.0002415704863973619, "loss": 0.2194, "step": 1446 }, { "epoch": 0.81, "learning_rate": 0.0002415086562242374, "loss": 0.2434, "step": 1447 }, { "epoch": 0.81, "learning_rate": 0.0002414468260511129, "loss": 0.1365, "step": 1448 }, { "epoch": 0.81, "learning_rate": 0.00024138499587798845, "loss": 0.2527, "step": 1449 }, { "epoch": 0.81, "learning_rate": 0.00024132316570486397, "loss": 0.3855, "step": 1450 }, { "epoch": 0.81, "learning_rate": 0.00024126133553173946, "loss": 0.1261, "step": 1451 }, { "epoch": 0.81, "learning_rate": 0.00024119950535861497, "loss": 0.1177, "step": 1452 }, { "epoch": 0.81, "learning_rate": 0.0002411376751854905, "loss": 0.1127, "step": 1453 }, { "epoch": 0.82, "learning_rate": 0.000241075845012366, "loss": 0.0841, "step": 1454 }, { "epoch": 0.82, "learning_rate": 0.00024101401483924155, "loss": 0.1393, "step": 1455 }, { "epoch": 0.82, "learning_rate": 0.00024095218466611704, "loss": 0.139, "step": 1456 }, { "epoch": 0.82, "learning_rate": 0.00024089035449299256, "loss": 0.1143, "step": 1457 }, { "epoch": 0.82, "learning_rate": 0.00024082852431986807, "loss": 0.1322, "step": 1458 }, { "epoch": 0.82, "learning_rate": 0.0002407666941467436, "loss": 0.1039, "step": 1459 }, { "epoch": 0.82, "learning_rate": 0.00024070486397361908, "loss": 0.1502, "step": 1460 }, { "epoch": 0.82, "learning_rate": 0.00024064303380049462, "loss": 0.1258, "step": 1461 }, { "epoch": 0.82, "learning_rate": 0.00024058120362737014, "loss": 0.119, "step": 1462 }, { "epoch": 0.82, "learning_rate": 0.00024051937345424566, "loss": 0.1246, "step": 1463 }, { "epoch": 0.82, "learning_rate": 0.00024045754328112117, "loss": 0.1004, "step": 1464 }, { "epoch": 0.82, "learning_rate": 0.00024039571310799666, "loss": 0.1258, "step": 1465 }, { "epoch": 0.82, "learning_rate": 0.00024033388293487218, "loss": 0.1025, "step": 1466 }, { "epoch": 0.82, "learning_rate": 0.00024027205276174772, "loss": 0.1197, "step": 1467 }, { "epoch": 0.82, "learning_rate": 0.00024021022258862324, "loss": 0.0934, "step": 1468 }, { "epoch": 0.82, "learning_rate": 0.00024014839241549876, "loss": 0.0719, "step": 1469 }, { "epoch": 0.82, "learning_rate": 0.00024008656224237424, "loss": 0.1176, "step": 1470 }, { "epoch": 0.82, "learning_rate": 0.00024002473206924976, "loss": 0.1553, "step": 1471 }, { "epoch": 0.83, "learning_rate": 0.0002399629018961253, "loss": 0.1044, "step": 1472 }, { "epoch": 0.83, "learning_rate": 0.00023990107172300082, "loss": 0.114, "step": 1473 }, { "epoch": 0.83, "learning_rate": 0.0002398392415498763, "loss": 0.0909, "step": 1474 }, { "epoch": 0.83, "learning_rate": 0.00023977741137675183, "loss": 0.1132, "step": 1475 }, { "epoch": 0.83, "learning_rate": 0.00023971558120362734, "loss": 0.1125, "step": 1476 }, { "epoch": 0.83, "learning_rate": 0.00023965375103050286, "loss": 0.146, "step": 1477 }, { "epoch": 0.83, "learning_rate": 0.0002395919208573784, "loss": 0.0863, "step": 1478 }, { "epoch": 0.83, "learning_rate": 0.0002395300906842539, "loss": 0.1108, "step": 1479 }, { "epoch": 0.83, "learning_rate": 0.0002394682605111294, "loss": 0.0822, "step": 1480 }, { "epoch": 0.83, "learning_rate": 0.00023940643033800493, "loss": 0.1287, "step": 1481 }, { "epoch": 0.83, "learning_rate": 0.00023934460016488044, "loss": 0.1254, "step": 1482 }, { "epoch": 0.83, "learning_rate": 0.00023928276999175593, "loss": 0.128, "step": 1483 }, { "epoch": 0.83, "learning_rate": 0.00023922093981863145, "loss": 0.0991, "step": 1484 }, { "epoch": 0.83, "learning_rate": 0.000239159109645507, "loss": 0.1486, "step": 1485 }, { "epoch": 0.83, "learning_rate": 0.0002390972794723825, "loss": 0.0649, "step": 1486 }, { "epoch": 0.83, "learning_rate": 0.00023903544929925803, "loss": 0.123, "step": 1487 }, { "epoch": 0.83, "learning_rate": 0.00023897361912613352, "loss": 0.1623, "step": 1488 }, { "epoch": 0.83, "learning_rate": 0.00023891178895300903, "loss": 0.099, "step": 1489 }, { "epoch": 0.84, "learning_rate": 0.00023884995877988458, "loss": 0.1252, "step": 1490 }, { "epoch": 0.84, "learning_rate": 0.0002387881286067601, "loss": 0.1402, "step": 1491 }, { "epoch": 0.84, "learning_rate": 0.00023872629843363558, "loss": 0.1829, "step": 1492 }, { "epoch": 0.84, "learning_rate": 0.0002386644682605111, "loss": 0.1177, "step": 1493 }, { "epoch": 0.84, "learning_rate": 0.00023860263808738662, "loss": 0.1147, "step": 1494 }, { "epoch": 0.84, "learning_rate": 0.00023854080791426213, "loss": 0.1338, "step": 1495 }, { "epoch": 0.84, "learning_rate": 0.00023847897774113768, "loss": 0.1648, "step": 1496 }, { "epoch": 0.84, "learning_rate": 0.00023841714756801317, "loss": 0.1906, "step": 1497 }, { "epoch": 0.84, "learning_rate": 0.00023835531739488868, "loss": 0.2006, "step": 1498 }, { "epoch": 0.84, "learning_rate": 0.0002382934872217642, "loss": 0.1567, "step": 1499 }, { "epoch": 0.84, "learning_rate": 0.00023823165704863972, "loss": 0.1922, "step": 1500 }, { "epoch": 0.84, "eval_loss": 0.13580209016799927, "eval_runtime": 272.0203, "eval_samples_per_second": 9.713, "eval_steps_per_second": 0.61, "eval_wer": 0.11140420468068227, "step": 1500 }, { "epoch": 0.84, "learning_rate": 0.00023816982687551526, "loss": 0.1029, "step": 1501 }, { "epoch": 0.84, "learning_rate": 0.00023810799670239075, "loss": 0.1642, "step": 1502 }, { "epoch": 0.84, "learning_rate": 0.00023804616652926627, "loss": 0.095, "step": 1503 }, { "epoch": 0.84, "learning_rate": 0.00023798433635614178, "loss": 0.1525, "step": 1504 }, { "epoch": 0.84, "learning_rate": 0.0002379225061830173, "loss": 0.1158, "step": 1505 }, { "epoch": 0.84, "learning_rate": 0.0002378606760098928, "loss": 0.1437, "step": 1506 }, { "epoch": 0.84, "learning_rate": 0.0002377988458367683, "loss": 0.089, "step": 1507 }, { "epoch": 0.85, "learning_rate": 0.00023773701566364385, "loss": 0.0811, "step": 1508 }, { "epoch": 0.85, "learning_rate": 0.00023767518549051936, "loss": 0.108, "step": 1509 }, { "epoch": 0.85, "learning_rate": 0.00023761335531739488, "loss": 0.1048, "step": 1510 }, { "epoch": 0.85, "learning_rate": 0.00023755152514427037, "loss": 0.1038, "step": 1511 }, { "epoch": 0.85, "learning_rate": 0.0002374896949711459, "loss": 0.0991, "step": 1512 }, { "epoch": 0.85, "learning_rate": 0.0002374278647980214, "loss": 0.1276, "step": 1513 }, { "epoch": 0.85, "learning_rate": 0.00023736603462489695, "loss": 0.088, "step": 1514 }, { "epoch": 0.85, "learning_rate": 0.00023730420445177244, "loss": 0.1069, "step": 1515 }, { "epoch": 0.85, "learning_rate": 0.00023724237427864795, "loss": 0.1702, "step": 1516 }, { "epoch": 0.85, "learning_rate": 0.00023718054410552347, "loss": 0.1294, "step": 1517 }, { "epoch": 0.85, "learning_rate": 0.000237118713932399, "loss": 0.1492, "step": 1518 }, { "epoch": 0.85, "learning_rate": 0.00023705688375927453, "loss": 0.0975, "step": 1519 }, { "epoch": 0.85, "learning_rate": 0.00023699505358615002, "loss": 0.1554, "step": 1520 }, { "epoch": 0.85, "learning_rate": 0.00023693322341302554, "loss": 0.1016, "step": 1521 }, { "epoch": 0.85, "learning_rate": 0.00023687139323990105, "loss": 0.1215, "step": 1522 }, { "epoch": 0.85, "learning_rate": 0.00023680956306677657, "loss": 0.1195, "step": 1523 }, { "epoch": 0.85, "learning_rate": 0.00023674773289365209, "loss": 0.1097, "step": 1524 }, { "epoch": 0.85, "learning_rate": 0.00023668590272052758, "loss": 0.0887, "step": 1525 }, { "epoch": 0.86, "learning_rate": 0.00023662407254740312, "loss": 0.1192, "step": 1526 }, { "epoch": 0.86, "learning_rate": 0.00023656224237427864, "loss": 0.0956, "step": 1527 }, { "epoch": 0.86, "learning_rate": 0.00023650041220115415, "loss": 0.1101, "step": 1528 }, { "epoch": 0.86, "learning_rate": 0.00023643858202802964, "loss": 0.1398, "step": 1529 }, { "epoch": 0.86, "learning_rate": 0.00023637675185490516, "loss": 0.1117, "step": 1530 }, { "epoch": 0.86, "learning_rate": 0.0002363149216817807, "loss": 0.1022, "step": 1531 }, { "epoch": 0.86, "learning_rate": 0.00023625309150865622, "loss": 0.1162, "step": 1532 }, { "epoch": 0.86, "learning_rate": 0.00023619126133553174, "loss": 0.1359, "step": 1533 }, { "epoch": 0.86, "learning_rate": 0.00023612943116240723, "loss": 0.1334, "step": 1534 }, { "epoch": 0.86, "learning_rate": 0.00023606760098928274, "loss": 0.1081, "step": 1535 }, { "epoch": 0.86, "learning_rate": 0.00023600577081615826, "loss": 0.1392, "step": 1536 }, { "epoch": 0.86, "learning_rate": 0.0002359439406430338, "loss": 0.1394, "step": 1537 }, { "epoch": 0.86, "learning_rate": 0.0002358821104699093, "loss": 0.1575, "step": 1538 }, { "epoch": 0.86, "learning_rate": 0.0002358202802967848, "loss": 0.1132, "step": 1539 }, { "epoch": 0.86, "learning_rate": 0.00023575845012366032, "loss": 0.138, "step": 1540 }, { "epoch": 0.86, "learning_rate": 0.00023569661995053584, "loss": 0.2089, "step": 1541 }, { "epoch": 0.86, "learning_rate": 0.00023563478977741136, "loss": 0.0889, "step": 1542 }, { "epoch": 0.86, "learning_rate": 0.00023557295960428685, "loss": 0.1279, "step": 1543 }, { "epoch": 0.87, "learning_rate": 0.0002355111294311624, "loss": 0.1013, "step": 1544 }, { "epoch": 0.87, "learning_rate": 0.0002354492992580379, "loss": 0.1907, "step": 1545 }, { "epoch": 0.87, "learning_rate": 0.00023538746908491342, "loss": 0.1326, "step": 1546 }, { "epoch": 0.87, "learning_rate": 0.00023532563891178891, "loss": 0.1756, "step": 1547 }, { "epoch": 0.87, "learning_rate": 0.00023526380873866443, "loss": 0.1573, "step": 1548 }, { "epoch": 0.87, "learning_rate": 0.00023520197856553997, "loss": 0.1373, "step": 1549 }, { "epoch": 0.87, "learning_rate": 0.0002351401483924155, "loss": 0.2646, "step": 1550 }, { "epoch": 0.87, "learning_rate": 0.000235078318219291, "loss": 0.1164, "step": 1551 }, { "epoch": 0.87, "learning_rate": 0.0002350164880461665, "loss": 0.1133, "step": 1552 }, { "epoch": 0.87, "learning_rate": 0.000234954657873042, "loss": 0.1188, "step": 1553 }, { "epoch": 0.87, "learning_rate": 0.00023489282769991753, "loss": 0.1219, "step": 1554 }, { "epoch": 0.87, "learning_rate": 0.00023483099752679307, "loss": 0.1348, "step": 1555 }, { "epoch": 0.87, "learning_rate": 0.0002347691673536686, "loss": 0.1333, "step": 1556 }, { "epoch": 0.87, "learning_rate": 0.00023470733718054408, "loss": 0.1173, "step": 1557 }, { "epoch": 0.87, "learning_rate": 0.0002346455070074196, "loss": 0.0921, "step": 1558 }, { "epoch": 0.87, "learning_rate": 0.0002345836768342951, "loss": 0.1322, "step": 1559 }, { "epoch": 0.87, "learning_rate": 0.00023452184666117063, "loss": 0.1356, "step": 1560 }, { "epoch": 0.88, "learning_rate": 0.00023446001648804612, "loss": 0.1412, "step": 1561 }, { "epoch": 0.88, "learning_rate": 0.00023439818631492166, "loss": 0.133, "step": 1562 }, { "epoch": 0.88, "learning_rate": 0.00023433635614179718, "loss": 0.0812, "step": 1563 }, { "epoch": 0.88, "learning_rate": 0.0002342745259686727, "loss": 0.1028, "step": 1564 }, { "epoch": 0.88, "learning_rate": 0.0002342126957955482, "loss": 0.1121, "step": 1565 }, { "epoch": 0.88, "learning_rate": 0.0002341508656224237, "loss": 0.1186, "step": 1566 }, { "epoch": 0.88, "learning_rate": 0.00023408903544929925, "loss": 0.096, "step": 1567 }, { "epoch": 0.88, "learning_rate": 0.00023402720527617476, "loss": 0.1319, "step": 1568 }, { "epoch": 0.88, "learning_rate": 0.00023396537510305028, "loss": 0.0965, "step": 1569 }, { "epoch": 0.88, "learning_rate": 0.00023390354492992577, "loss": 0.1031, "step": 1570 }, { "epoch": 0.88, "learning_rate": 0.00023384171475680128, "loss": 0.1106, "step": 1571 }, { "epoch": 0.88, "learning_rate": 0.0002337798845836768, "loss": 0.1212, "step": 1572 }, { "epoch": 0.88, "learning_rate": 0.00023371805441055234, "loss": 0.0844, "step": 1573 }, { "epoch": 0.88, "learning_rate": 0.00023365622423742786, "loss": 0.0917, "step": 1574 }, { "epoch": 0.88, "learning_rate": 0.00023359439406430335, "loss": 0.0746, "step": 1575 }, { "epoch": 0.88, "learning_rate": 0.00023353256389117887, "loss": 0.1207, "step": 1576 }, { "epoch": 0.88, "learning_rate": 0.00023347073371805438, "loss": 0.1297, "step": 1577 }, { "epoch": 0.88, "learning_rate": 0.00023340890354492993, "loss": 0.1301, "step": 1578 }, { "epoch": 0.89, "learning_rate": 0.00023334707337180542, "loss": 0.1324, "step": 1579 }, { "epoch": 0.89, "learning_rate": 0.00023328524319868093, "loss": 0.0994, "step": 1580 }, { "epoch": 0.89, "learning_rate": 0.00023322341302555645, "loss": 0.1364, "step": 1581 }, { "epoch": 0.89, "learning_rate": 0.00023316158285243197, "loss": 0.1119, "step": 1582 }, { "epoch": 0.89, "learning_rate": 0.00023309975267930748, "loss": 0.0986, "step": 1583 }, { "epoch": 0.89, "learning_rate": 0.00023303792250618297, "loss": 0.1408, "step": 1584 }, { "epoch": 0.89, "learning_rate": 0.00023297609233305852, "loss": 0.1042, "step": 1585 }, { "epoch": 0.89, "learning_rate": 0.00023291426215993403, "loss": 0.1046, "step": 1586 }, { "epoch": 0.89, "learning_rate": 0.00023285243198680955, "loss": 0.1085, "step": 1587 }, { "epoch": 0.89, "learning_rate": 0.00023279060181368507, "loss": 0.0864, "step": 1588 }, { "epoch": 0.89, "learning_rate": 0.00023272877164056056, "loss": 0.0786, "step": 1589 }, { "epoch": 0.89, "learning_rate": 0.00023266694146743607, "loss": 0.1373, "step": 1590 }, { "epoch": 0.89, "learning_rate": 0.00023260511129431162, "loss": 0.1247, "step": 1591 }, { "epoch": 0.89, "learning_rate": 0.00023254328112118713, "loss": 0.1561, "step": 1592 }, { "epoch": 0.89, "learning_rate": 0.00023248145094806262, "loss": 0.1285, "step": 1593 }, { "epoch": 0.89, "learning_rate": 0.00023241962077493814, "loss": 0.1134, "step": 1594 }, { "epoch": 0.89, "learning_rate": 0.00023235779060181366, "loss": 0.1368, "step": 1595 }, { "epoch": 0.89, "learning_rate": 0.0002322959604286892, "loss": 0.1674, "step": 1596 }, { "epoch": 0.9, "learning_rate": 0.00023223413025556472, "loss": 0.1512, "step": 1597 }, { "epoch": 0.9, "learning_rate": 0.0002321723000824402, "loss": 0.1204, "step": 1598 }, { "epoch": 0.9, "learning_rate": 0.00023211046990931572, "loss": 0.3194, "step": 1599 }, { "epoch": 0.9, "learning_rate": 0.00023204863973619124, "loss": 0.2799, "step": 1600 }, { "epoch": 0.9, "learning_rate": 0.00023198680956306676, "loss": 0.1064, "step": 1601 }, { "epoch": 0.9, "learning_rate": 0.00023192497938994224, "loss": 0.0938, "step": 1602 }, { "epoch": 0.9, "learning_rate": 0.0002318631492168178, "loss": 0.1271, "step": 1603 }, { "epoch": 0.9, "learning_rate": 0.0002318013190436933, "loss": 0.1424, "step": 1604 }, { "epoch": 0.9, "learning_rate": 0.00023173948887056882, "loss": 0.0988, "step": 1605 }, { "epoch": 0.9, "learning_rate": 0.00023167765869744434, "loss": 0.1418, "step": 1606 }, { "epoch": 0.9, "learning_rate": 0.00023161582852431983, "loss": 0.1345, "step": 1607 }, { "epoch": 0.9, "learning_rate": 0.00023155399835119537, "loss": 0.142, "step": 1608 }, { "epoch": 0.9, "learning_rate": 0.0002314921681780709, "loss": 0.1282, "step": 1609 }, { "epoch": 0.9, "learning_rate": 0.0002314303380049464, "loss": 0.1246, "step": 1610 }, { "epoch": 0.9, "learning_rate": 0.00023136850783182192, "loss": 0.1135, "step": 1611 }, { "epoch": 0.9, "learning_rate": 0.0002313066776586974, "loss": 0.1054, "step": 1612 }, { "epoch": 0.9, "learning_rate": 0.00023124484748557293, "loss": 0.1026, "step": 1613 }, { "epoch": 0.9, "learning_rate": 0.00023118301731244847, "loss": 0.1123, "step": 1614 }, { "epoch": 0.91, "learning_rate": 0.000231121187139324, "loss": 0.0995, "step": 1615 }, { "epoch": 0.91, "learning_rate": 0.00023105935696619948, "loss": 0.1013, "step": 1616 }, { "epoch": 0.91, "learning_rate": 0.000230997526793075, "loss": 0.1279, "step": 1617 }, { "epoch": 0.91, "learning_rate": 0.0002309356966199505, "loss": 0.1018, "step": 1618 }, { "epoch": 0.91, "learning_rate": 0.00023087386644682603, "loss": 0.0894, "step": 1619 }, { "epoch": 0.91, "learning_rate": 0.00023081203627370157, "loss": 0.0976, "step": 1620 }, { "epoch": 0.91, "learning_rate": 0.00023075020610057706, "loss": 0.1237, "step": 1621 }, { "epoch": 0.91, "learning_rate": 0.00023068837592745258, "loss": 0.14, "step": 1622 }, { "epoch": 0.91, "learning_rate": 0.0002306265457543281, "loss": 0.0878, "step": 1623 }, { "epoch": 0.91, "learning_rate": 0.0002305647155812036, "loss": 0.0838, "step": 1624 }, { "epoch": 0.91, "learning_rate": 0.0002305028854080791, "loss": 0.1091, "step": 1625 }, { "epoch": 0.91, "learning_rate": 0.00023044105523495464, "loss": 0.104, "step": 1626 }, { "epoch": 0.91, "learning_rate": 0.00023037922506183016, "loss": 0.1139, "step": 1627 }, { "epoch": 0.91, "learning_rate": 0.00023031739488870568, "loss": 0.1133, "step": 1628 }, { "epoch": 0.91, "learning_rate": 0.0002302555647155812, "loss": 0.1165, "step": 1629 }, { "epoch": 0.91, "learning_rate": 0.00023019373454245668, "loss": 0.1052, "step": 1630 }, { "epoch": 0.91, "learning_rate": 0.0002301319043693322, "loss": 0.1047, "step": 1631 }, { "epoch": 0.91, "learning_rate": 0.00023007007419620774, "loss": 0.1524, "step": 1632 }, { "epoch": 0.92, "learning_rate": 0.00023000824402308326, "loss": 0.1189, "step": 1633 }, { "epoch": 0.92, "learning_rate": 0.00022994641384995875, "loss": 0.1767, "step": 1634 }, { "epoch": 0.92, "learning_rate": 0.00022988458367683426, "loss": 0.099, "step": 1635 }, { "epoch": 0.92, "learning_rate": 0.00022982275350370978, "loss": 0.0819, "step": 1636 }, { "epoch": 0.92, "learning_rate": 0.0002297609233305853, "loss": 0.1005, "step": 1637 }, { "epoch": 0.92, "learning_rate": 0.00022969909315746084, "loss": 0.1214, "step": 1638 }, { "epoch": 0.92, "learning_rate": 0.00022963726298433633, "loss": 0.0784, "step": 1639 }, { "epoch": 0.92, "learning_rate": 0.00022957543281121185, "loss": 0.1195, "step": 1640 }, { "epoch": 0.92, "learning_rate": 0.00022951360263808736, "loss": 0.1189, "step": 1641 }, { "epoch": 0.92, "learning_rate": 0.00022945177246496288, "loss": 0.123, "step": 1642 }, { "epoch": 0.92, "learning_rate": 0.00022938994229183842, "loss": 0.1297, "step": 1643 }, { "epoch": 0.92, "learning_rate": 0.00022932811211871391, "loss": 0.1157, "step": 1644 }, { "epoch": 0.92, "learning_rate": 0.00022926628194558943, "loss": 0.1945, "step": 1645 }, { "epoch": 0.92, "learning_rate": 0.00022920445177246495, "loss": 0.0955, "step": 1646 }, { "epoch": 0.92, "learning_rate": 0.00022914262159934046, "loss": 0.1715, "step": 1647 }, { "epoch": 0.92, "learning_rate": 0.00022908079142621595, "loss": 0.2128, "step": 1648 }, { "epoch": 0.92, "learning_rate": 0.00022901896125309147, "loss": 0.2373, "step": 1649 }, { "epoch": 0.92, "learning_rate": 0.00022895713107996701, "loss": 0.363, "step": 1650 }, { "epoch": 0.93, "learning_rate": 0.00022889530090684253, "loss": 0.1364, "step": 1651 }, { "epoch": 0.93, "learning_rate": 0.00022883347073371805, "loss": 0.1355, "step": 1652 }, { "epoch": 0.93, "learning_rate": 0.00022877164056059354, "loss": 0.1161, "step": 1653 }, { "epoch": 0.93, "learning_rate": 0.00022870981038746905, "loss": 0.1186, "step": 1654 }, { "epoch": 0.93, "learning_rate": 0.0002286479802143446, "loss": 0.1389, "step": 1655 }, { "epoch": 0.93, "learning_rate": 0.0002285861500412201, "loss": 0.1294, "step": 1656 }, { "epoch": 0.93, "learning_rate": 0.0002285243198680956, "loss": 0.106, "step": 1657 }, { "epoch": 0.93, "learning_rate": 0.00022846248969497112, "loss": 0.0865, "step": 1658 }, { "epoch": 0.93, "learning_rate": 0.00022840065952184664, "loss": 0.1083, "step": 1659 }, { "epoch": 0.93, "learning_rate": 0.00022833882934872215, "loss": 0.1139, "step": 1660 }, { "epoch": 0.93, "learning_rate": 0.0002282769991755977, "loss": 0.1197, "step": 1661 }, { "epoch": 0.93, "learning_rate": 0.00022821516900247319, "loss": 0.1063, "step": 1662 }, { "epoch": 0.93, "learning_rate": 0.0002281533388293487, "loss": 0.1081, "step": 1663 }, { "epoch": 0.93, "learning_rate": 0.00022809150865622422, "loss": 0.1647, "step": 1664 }, { "epoch": 0.93, "learning_rate": 0.00022802967848309974, "loss": 0.1377, "step": 1665 }, { "epoch": 0.93, "learning_rate": 0.00022796784830997522, "loss": 0.1128, "step": 1666 }, { "epoch": 0.93, "learning_rate": 0.00022790601813685074, "loss": 0.1149, "step": 1667 }, { "epoch": 0.93, "learning_rate": 0.00022784418796372629, "loss": 0.1179, "step": 1668 }, { "epoch": 0.94, "learning_rate": 0.0002277823577906018, "loss": 0.0654, "step": 1669 }, { "epoch": 0.94, "learning_rate": 0.00022772052761747732, "loss": 0.1155, "step": 1670 }, { "epoch": 0.94, "learning_rate": 0.0002276586974443528, "loss": 0.1191, "step": 1671 }, { "epoch": 0.94, "learning_rate": 0.00022759686727122832, "loss": 0.1403, "step": 1672 }, { "epoch": 0.94, "learning_rate": 0.00022753503709810387, "loss": 0.112, "step": 1673 }, { "epoch": 0.94, "learning_rate": 0.00022747320692497938, "loss": 0.1052, "step": 1674 }, { "epoch": 0.94, "learning_rate": 0.0002274113767518549, "loss": 0.104, "step": 1675 }, { "epoch": 0.94, "learning_rate": 0.0002273495465787304, "loss": 0.0613, "step": 1676 }, { "epoch": 0.94, "learning_rate": 0.0002272877164056059, "loss": 0.0749, "step": 1677 }, { "epoch": 0.94, "learning_rate": 0.00022722588623248142, "loss": 0.1096, "step": 1678 }, { "epoch": 0.94, "learning_rate": 0.00022716405605935697, "loss": 0.1167, "step": 1679 }, { "epoch": 0.94, "learning_rate": 0.00022710222588623246, "loss": 0.0981, "step": 1680 }, { "epoch": 0.94, "learning_rate": 0.00022704039571310797, "loss": 0.1165, "step": 1681 }, { "epoch": 0.94, "learning_rate": 0.0002269785655399835, "loss": 0.0924, "step": 1682 }, { "epoch": 0.94, "learning_rate": 0.000226916735366859, "loss": 0.1228, "step": 1683 }, { "epoch": 0.94, "learning_rate": 0.00022685490519373455, "loss": 0.0865, "step": 1684 }, { "epoch": 0.94, "learning_rate": 0.00022679307502061004, "loss": 0.1315, "step": 1685 }, { "epoch": 0.95, "learning_rate": 0.00022673124484748556, "loss": 0.0872, "step": 1686 }, { "epoch": 0.95, "learning_rate": 0.00022666941467436107, "loss": 0.0954, "step": 1687 }, { "epoch": 0.95, "learning_rate": 0.0002266075845012366, "loss": 0.1015, "step": 1688 }, { "epoch": 0.95, "learning_rate": 0.00022654575432811208, "loss": 0.1321, "step": 1689 }, { "epoch": 0.95, "learning_rate": 0.0002264839241549876, "loss": 0.1028, "step": 1690 }, { "epoch": 0.95, "learning_rate": 0.00022642209398186314, "loss": 0.1459, "step": 1691 }, { "epoch": 0.95, "learning_rate": 0.00022636026380873866, "loss": 0.1117, "step": 1692 }, { "epoch": 0.95, "learning_rate": 0.00022629843363561417, "loss": 0.1288, "step": 1693 }, { "epoch": 0.95, "learning_rate": 0.00022623660346248966, "loss": 0.1367, "step": 1694 }, { "epoch": 0.95, "learning_rate": 0.00022617477328936518, "loss": 0.2844, "step": 1695 }, { "epoch": 0.95, "learning_rate": 0.0002261129431162407, "loss": 0.1915, "step": 1696 }, { "epoch": 0.95, "learning_rate": 0.00022605111294311624, "loss": 0.1948, "step": 1697 }, { "epoch": 0.95, "learning_rate": 0.00022598928276999173, "loss": 0.3076, "step": 1698 }, { "epoch": 0.95, "learning_rate": 0.00022592745259686725, "loss": 0.2572, "step": 1699 }, { "epoch": 0.95, "learning_rate": 0.00022586562242374276, "loss": 0.2626, "step": 1700 }, { "epoch": 0.95, "learning_rate": 0.00022580379225061828, "loss": 0.1379, "step": 1701 }, { "epoch": 0.95, "learning_rate": 0.00022574196207749382, "loss": 0.1371, "step": 1702 }, { "epoch": 0.95, "learning_rate": 0.0002256801319043693, "loss": 0.1192, "step": 1703 }, { "epoch": 0.96, "learning_rate": 0.00022561830173124483, "loss": 0.1109, "step": 1704 }, { "epoch": 0.96, "learning_rate": 0.00022555647155812034, "loss": 0.0911, "step": 1705 }, { "epoch": 0.96, "learning_rate": 0.00022549464138499586, "loss": 0.0993, "step": 1706 }, { "epoch": 0.96, "learning_rate": 0.00022543281121187138, "loss": 0.0782, "step": 1707 }, { "epoch": 0.96, "learning_rate": 0.00022537098103874687, "loss": 0.1885, "step": 1708 }, { "epoch": 0.96, "learning_rate": 0.0002253091508656224, "loss": 0.1363, "step": 1709 }, { "epoch": 0.96, "learning_rate": 0.00022524732069249793, "loss": 0.1213, "step": 1710 }, { "epoch": 0.96, "learning_rate": 0.00022518549051937344, "loss": 0.0983, "step": 1711 }, { "epoch": 0.96, "learning_rate": 0.00022512366034624893, "loss": 0.1065, "step": 1712 }, { "epoch": 0.96, "learning_rate": 0.00022506183017312445, "loss": 0.1507, "step": 1713 }, { "epoch": 0.96, "learning_rate": 0.000225, "loss": 0.1294, "step": 1714 }, { "epoch": 0.96, "learning_rate": 0.0002249381698268755, "loss": 0.1145, "step": 1715 }, { "epoch": 0.96, "learning_rate": 0.00022487633965375103, "loss": 0.1135, "step": 1716 }, { "epoch": 0.96, "learning_rate": 0.00022481450948062652, "loss": 0.1121, "step": 1717 }, { "epoch": 0.96, "learning_rate": 0.00022475267930750203, "loss": 0.1473, "step": 1718 }, { "epoch": 0.96, "learning_rate": 0.00022469084913437755, "loss": 0.1264, "step": 1719 }, { "epoch": 0.96, "learning_rate": 0.0002246290189612531, "loss": 0.1334, "step": 1720 }, { "epoch": 0.96, "learning_rate": 0.00022456718878812858, "loss": 0.1471, "step": 1721 }, { "epoch": 0.97, "learning_rate": 0.0002245053586150041, "loss": 0.117, "step": 1722 }, { "epoch": 0.97, "learning_rate": 0.00022444352844187962, "loss": 0.113, "step": 1723 }, { "epoch": 0.97, "learning_rate": 0.00022438169826875513, "loss": 0.1064, "step": 1724 }, { "epoch": 0.97, "learning_rate": 0.00022431986809563065, "loss": 0.1355, "step": 1725 }, { "epoch": 0.97, "learning_rate": 0.00022425803792250614, "loss": 0.1157, "step": 1726 }, { "epoch": 0.97, "learning_rate": 0.00022419620774938168, "loss": 0.0896, "step": 1727 }, { "epoch": 0.97, "learning_rate": 0.0002241343775762572, "loss": 0.1229, "step": 1728 }, { "epoch": 0.97, "learning_rate": 0.00022407254740313272, "loss": 0.1121, "step": 1729 }, { "epoch": 0.97, "learning_rate": 0.00022401071723000823, "loss": 0.1235, "step": 1730 }, { "epoch": 0.97, "learning_rate": 0.00022394888705688372, "loss": 0.085, "step": 1731 }, { "epoch": 0.97, "learning_rate": 0.00022388705688375927, "loss": 0.0795, "step": 1732 }, { "epoch": 0.97, "learning_rate": 0.00022382522671063478, "loss": 0.1247, "step": 1733 }, { "epoch": 0.97, "learning_rate": 0.0002237633965375103, "loss": 0.1008, "step": 1734 }, { "epoch": 0.97, "learning_rate": 0.0002237015663643858, "loss": 0.1528, "step": 1735 }, { "epoch": 0.97, "learning_rate": 0.0002236397361912613, "loss": 0.1409, "step": 1736 }, { "epoch": 0.97, "learning_rate": 0.00022357790601813682, "loss": 0.122, "step": 1737 }, { "epoch": 0.97, "learning_rate": 0.00022351607584501236, "loss": 0.0985, "step": 1738 }, { "epoch": 0.97, "learning_rate": 0.00022345424567188788, "loss": 0.1065, "step": 1739 }, { "epoch": 0.98, "learning_rate": 0.00022339241549876337, "loss": 0.0853, "step": 1740 }, { "epoch": 0.98, "learning_rate": 0.0002233305853256389, "loss": 0.1022, "step": 1741 }, { "epoch": 0.98, "learning_rate": 0.0002232687551525144, "loss": 0.1789, "step": 1742 }, { "epoch": 0.98, "learning_rate": 0.00022320692497938992, "loss": 0.2318, "step": 1743 }, { "epoch": 0.98, "learning_rate": 0.0002231450948062654, "loss": 0.1359, "step": 1744 }, { "epoch": 0.98, "learning_rate": 0.00022308326463314095, "loss": 0.1876, "step": 1745 }, { "epoch": 0.98, "learning_rate": 0.00022302143446001647, "loss": 0.1396, "step": 1746 }, { "epoch": 0.98, "learning_rate": 0.000222959604286892, "loss": 0.073, "step": 1747 }, { "epoch": 0.98, "learning_rate": 0.0002228977741137675, "loss": 0.1561, "step": 1748 }, { "epoch": 0.98, "learning_rate": 0.000222835943940643, "loss": 0.2238, "step": 1749 }, { "epoch": 0.98, "learning_rate": 0.00022277411376751854, "loss": 0.264, "step": 1750 }, { "epoch": 0.98, "learning_rate": 0.00022271228359439405, "loss": 0.0973, "step": 1751 }, { "epoch": 0.98, "learning_rate": 0.00022265045342126957, "loss": 0.1136, "step": 1752 }, { "epoch": 0.98, "learning_rate": 0.00022258862324814506, "loss": 0.1163, "step": 1753 }, { "epoch": 0.98, "learning_rate": 0.00022252679307502058, "loss": 0.095, "step": 1754 }, { "epoch": 0.98, "learning_rate": 0.0002224649629018961, "loss": 0.082, "step": 1755 }, { "epoch": 0.98, "learning_rate": 0.00022240313272877164, "loss": 0.0939, "step": 1756 }, { "epoch": 0.98, "learning_rate": 0.00022234130255564715, "loss": 0.1055, "step": 1757 }, { "epoch": 0.99, "learning_rate": 0.00022227947238252264, "loss": 0.1169, "step": 1758 }, { "epoch": 0.99, "learning_rate": 0.00022221764220939816, "loss": 0.1005, "step": 1759 }, { "epoch": 0.99, "learning_rate": 0.00022215581203627368, "loss": 0.1394, "step": 1760 }, { "epoch": 0.99, "learning_rate": 0.00022209398186314922, "loss": 0.1111, "step": 1761 }, { "epoch": 0.99, "learning_rate": 0.00022203215169002474, "loss": 0.1284, "step": 1762 }, { "epoch": 0.99, "learning_rate": 0.00022197032151690023, "loss": 0.1367, "step": 1763 }, { "epoch": 0.99, "learning_rate": 0.00022190849134377574, "loss": 0.0889, "step": 1764 }, { "epoch": 0.99, "learning_rate": 0.00022184666117065126, "loss": 0.0963, "step": 1765 }, { "epoch": 0.99, "learning_rate": 0.00022178483099752678, "loss": 0.1283, "step": 1766 }, { "epoch": 0.99, "learning_rate": 0.00022172300082440226, "loss": 0.0993, "step": 1767 }, { "epoch": 0.99, "learning_rate": 0.0002216611706512778, "loss": 0.0907, "step": 1768 }, { "epoch": 0.99, "learning_rate": 0.00022159934047815332, "loss": 0.1137, "step": 1769 }, { "epoch": 0.99, "learning_rate": 0.00022153751030502884, "loss": 0.0584, "step": 1770 }, { "epoch": 0.99, "learning_rate": 0.00022147568013190436, "loss": 0.1238, "step": 1771 }, { "epoch": 0.99, "learning_rate": 0.00022141384995877985, "loss": 0.1043, "step": 1772 }, { "epoch": 0.99, "learning_rate": 0.00022135201978565536, "loss": 0.1268, "step": 1773 }, { "epoch": 0.99, "learning_rate": 0.0002212901896125309, "loss": 0.1161, "step": 1774 }, { "epoch": 0.99, "learning_rate": 0.00022122835943940642, "loss": 0.0977, "step": 1775 }, { "epoch": 1.0, "learning_rate": 0.00022116652926628191, "loss": 0.1215, "step": 1776 }, { "epoch": 1.0, "learning_rate": 0.00022110469909315743, "loss": 0.1102, "step": 1777 }, { "epoch": 1.0, "learning_rate": 0.00022104286892003295, "loss": 0.207, "step": 1778 }, { "epoch": 1.0, "learning_rate": 0.0002209810387469085, "loss": 0.1254, "step": 1779 }, { "epoch": 1.0, "learning_rate": 0.000220919208573784, "loss": 0.1235, "step": 1780 }, { "epoch": 1.0, "learning_rate": 0.0002208573784006595, "loss": 0.1555, "step": 1781 }, { "epoch": 1.0, "learning_rate": 0.000220795548227535, "loss": 0.1646, "step": 1782 }, { "epoch": 1.0, "learning_rate": 0.00022073371805441053, "loss": 0.1866, "step": 1783 }, { "epoch": 1.0, "learning_rate": 0.00022067188788128605, "loss": 0.1921, "step": 1784 }, { "epoch": 1.0, "learning_rate": 0.00022061005770816154, "loss": 0.1183, "step": 1785 }, { "epoch": 1.0, "learning_rate": 0.00022054822753503708, "loss": 0.1128, "step": 1786 }, { "epoch": 1.0, "learning_rate": 0.0002204863973619126, "loss": 0.0971, "step": 1787 }, { "epoch": 1.0, "learning_rate": 0.0002204245671887881, "loss": 0.0646, "step": 1788 }, { "epoch": 1.0, "learning_rate": 0.00022036273701566363, "loss": 0.0825, "step": 1789 }, { "epoch": 1.0, "learning_rate": 0.00022030090684253912, "loss": 0.0724, "step": 1790 }, { "epoch": 1.0, "learning_rate": 0.00022023907666941466, "loss": 0.1201, "step": 1791 }, { "epoch": 1.0, "learning_rate": 0.00022017724649629018, "loss": 0.1096, "step": 1792 }, { "epoch": 1.01, "learning_rate": 0.0002201154163231657, "loss": 0.1002, "step": 1793 }, { "epoch": 1.01, "learning_rate": 0.0002200535861500412, "loss": 0.1588, "step": 1794 }, { "epoch": 1.01, "learning_rate": 0.0002199917559769167, "loss": 0.0855, "step": 1795 }, { "epoch": 1.01, "learning_rate": 0.00021992992580379222, "loss": 0.0808, "step": 1796 }, { "epoch": 1.01, "learning_rate": 0.00021986809563066776, "loss": 0.103, "step": 1797 }, { "epoch": 1.01, "learning_rate": 0.00021980626545754328, "loss": 0.0964, "step": 1798 }, { "epoch": 1.01, "learning_rate": 0.00021974443528441877, "loss": 0.0846, "step": 1799 }, { "epoch": 1.01, "learning_rate": 0.00021968260511129428, "loss": 0.093, "step": 1800 }, { "epoch": 1.01, "eval_loss": 0.12774577736854553, "eval_runtime": 270.0693, "eval_samples_per_second": 9.783, "eval_steps_per_second": 0.615, "eval_wer": 0.10347084490281634, "step": 1800 }, { "epoch": 1.01, "learning_rate": 0.0002196207749381698, "loss": 0.1077, "step": 1801 }, { "epoch": 1.01, "learning_rate": 0.00021955894476504532, "loss": 0.0985, "step": 1802 }, { "epoch": 1.01, "learning_rate": 0.00021949711459192086, "loss": 0.0856, "step": 1803 }, { "epoch": 1.01, "learning_rate": 0.00021943528441879635, "loss": 0.1209, "step": 1804 }, { "epoch": 1.01, "learning_rate": 0.00021937345424567187, "loss": 0.0786, "step": 1805 }, { "epoch": 1.01, "learning_rate": 0.00021931162407254738, "loss": 0.0696, "step": 1806 }, { "epoch": 1.01, "learning_rate": 0.0002192497938994229, "loss": 0.0786, "step": 1807 }, { "epoch": 1.01, "learning_rate": 0.0002191879637262984, "loss": 0.0856, "step": 1808 }, { "epoch": 1.01, "learning_rate": 0.00021912613355317393, "loss": 0.0821, "step": 1809 }, { "epoch": 1.01, "learning_rate": 0.00021906430338004945, "loss": 0.0761, "step": 1810 }, { "epoch": 1.02, "learning_rate": 0.00021900247320692497, "loss": 0.0964, "step": 1811 }, { "epoch": 1.02, "learning_rate": 0.00021894064303380048, "loss": 0.1215, "step": 1812 }, { "epoch": 1.02, "learning_rate": 0.00021887881286067597, "loss": 0.0729, "step": 1813 }, { "epoch": 1.02, "learning_rate": 0.0002188169826875515, "loss": 0.0956, "step": 1814 }, { "epoch": 1.02, "learning_rate": 0.00021875515251442703, "loss": 0.1003, "step": 1815 }, { "epoch": 1.02, "learning_rate": 0.00021869332234130255, "loss": 0.1308, "step": 1816 }, { "epoch": 1.02, "learning_rate": 0.00021863149216817804, "loss": 0.0809, "step": 1817 }, { "epoch": 1.02, "learning_rate": 0.00021856966199505356, "loss": 0.1058, "step": 1818 }, { "epoch": 1.02, "learning_rate": 0.00021850783182192907, "loss": 0.0623, "step": 1819 }, { "epoch": 1.02, "learning_rate": 0.0002184460016488046, "loss": 0.0785, "step": 1820 }, { "epoch": 1.02, "learning_rate": 0.00021838417147568013, "loss": 0.0846, "step": 1821 }, { "epoch": 1.02, "learning_rate": 0.00021832234130255562, "loss": 0.0845, "step": 1822 }, { "epoch": 1.02, "learning_rate": 0.00021826051112943114, "loss": 0.0795, "step": 1823 }, { "epoch": 1.02, "learning_rate": 0.00021819868095630666, "loss": 0.1183, "step": 1824 }, { "epoch": 1.02, "learning_rate": 0.00021813685078318217, "loss": 0.0876, "step": 1825 }, { "epoch": 1.02, "learning_rate": 0.00021807502061005772, "loss": 0.1462, "step": 1826 }, { "epoch": 1.02, "learning_rate": 0.0002180131904369332, "loss": 0.0578, "step": 1827 }, { "epoch": 1.02, "learning_rate": 0.00021795136026380872, "loss": 0.0903, "step": 1828 }, { "epoch": 1.03, "learning_rate": 0.00021788953009068424, "loss": 0.1359, "step": 1829 }, { "epoch": 1.03, "learning_rate": 0.00021782769991755976, "loss": 0.1153, "step": 1830 }, { "epoch": 1.03, "learning_rate": 0.00021776586974443524, "loss": 0.0432, "step": 1831 }, { "epoch": 1.03, "learning_rate": 0.00021770403957131076, "loss": 0.1355, "step": 1832 }, { "epoch": 1.03, "learning_rate": 0.0002176422093981863, "loss": 0.1865, "step": 1833 }, { "epoch": 1.03, "learning_rate": 0.00021758037922506182, "loss": 0.1327, "step": 1834 }, { "epoch": 1.03, "learning_rate": 0.00021751854905193734, "loss": 0.1093, "step": 1835 }, { "epoch": 1.03, "learning_rate": 0.00021745671887881283, "loss": 0.1243, "step": 1836 }, { "epoch": 1.03, "learning_rate": 0.00021739488870568834, "loss": 0.0818, "step": 1837 }, { "epoch": 1.03, "learning_rate": 0.0002173330585325639, "loss": 0.0706, "step": 1838 }, { "epoch": 1.03, "learning_rate": 0.0002172712283594394, "loss": 0.0841, "step": 1839 }, { "epoch": 1.03, "learning_rate": 0.0002172093981863149, "loss": 0.0902, "step": 1840 }, { "epoch": 1.03, "learning_rate": 0.0002171475680131904, "loss": 0.098, "step": 1841 }, { "epoch": 1.03, "learning_rate": 0.00021708573784006593, "loss": 0.102, "step": 1842 }, { "epoch": 1.03, "learning_rate": 0.00021702390766694144, "loss": 0.1041, "step": 1843 }, { "epoch": 1.03, "learning_rate": 0.000216962077493817, "loss": 0.1174, "step": 1844 }, { "epoch": 1.03, "learning_rate": 0.00021690024732069248, "loss": 0.0638, "step": 1845 }, { "epoch": 1.03, "learning_rate": 0.000216838417147568, "loss": 0.0585, "step": 1846 }, { "epoch": 1.04, "learning_rate": 0.0002167765869744435, "loss": 0.1154, "step": 1847 }, { "epoch": 1.04, "learning_rate": 0.00021671475680131903, "loss": 0.0778, "step": 1848 }, { "epoch": 1.04, "learning_rate": 0.00021665292662819454, "loss": 0.0978, "step": 1849 }, { "epoch": 1.04, "learning_rate": 0.00021659109645507003, "loss": 0.0896, "step": 1850 }, { "epoch": 1.04, "learning_rate": 0.00021652926628194558, "loss": 0.0787, "step": 1851 }, { "epoch": 1.04, "learning_rate": 0.0002164674361088211, "loss": 0.089, "step": 1852 }, { "epoch": 1.04, "learning_rate": 0.0002164056059356966, "loss": 0.0554, "step": 1853 }, { "epoch": 1.04, "learning_rate": 0.0002163437757625721, "loss": 0.0807, "step": 1854 }, { "epoch": 1.04, "learning_rate": 0.00021628194558944762, "loss": 0.104, "step": 1855 }, { "epoch": 1.04, "learning_rate": 0.00021622011541632316, "loss": 0.0465, "step": 1856 }, { "epoch": 1.04, "learning_rate": 0.00021615828524319868, "loss": 0.0894, "step": 1857 }, { "epoch": 1.04, "learning_rate": 0.0002160964550700742, "loss": 0.0781, "step": 1858 }, { "epoch": 1.04, "learning_rate": 0.00021603462489694968, "loss": 0.0787, "step": 1859 }, { "epoch": 1.04, "learning_rate": 0.0002159727947238252, "loss": 0.0915, "step": 1860 }, { "epoch": 1.04, "learning_rate": 0.00021591096455070072, "loss": 0.0559, "step": 1861 }, { "epoch": 1.04, "learning_rate": 0.00021584913437757626, "loss": 0.0859, "step": 1862 }, { "epoch": 1.04, "learning_rate": 0.00021578730420445175, "loss": 0.1358, "step": 1863 }, { "epoch": 1.04, "learning_rate": 0.00021572547403132726, "loss": 0.0764, "step": 1864 }, { "epoch": 1.05, "learning_rate": 0.00021566364385820278, "loss": 0.1052, "step": 1865 }, { "epoch": 1.05, "learning_rate": 0.0002156018136850783, "loss": 0.0593, "step": 1866 }, { "epoch": 1.05, "learning_rate": 0.00021553998351195384, "loss": 0.0818, "step": 1867 }, { "epoch": 1.05, "learning_rate": 0.00021547815333882933, "loss": 0.0637, "step": 1868 }, { "epoch": 1.05, "learning_rate": 0.00021541632316570485, "loss": 0.0929, "step": 1869 }, { "epoch": 1.05, "learning_rate": 0.00021535449299258036, "loss": 0.0695, "step": 1870 }, { "epoch": 1.05, "learning_rate": 0.00021529266281945588, "loss": 0.1178, "step": 1871 }, { "epoch": 1.05, "learning_rate": 0.00021523083264633137, "loss": 0.0775, "step": 1872 }, { "epoch": 1.05, "learning_rate": 0.0002151690024732069, "loss": 0.071, "step": 1873 }, { "epoch": 1.05, "learning_rate": 0.00021510717230008243, "loss": 0.0817, "step": 1874 }, { "epoch": 1.05, "learning_rate": 0.00021504534212695795, "loss": 0.1031, "step": 1875 }, { "epoch": 1.05, "learning_rate": 0.00021498351195383346, "loss": 0.0598, "step": 1876 }, { "epoch": 1.05, "learning_rate": 0.00021492168178070895, "loss": 0.1371, "step": 1877 }, { "epoch": 1.05, "learning_rate": 0.00021485985160758447, "loss": 0.0561, "step": 1878 }, { "epoch": 1.05, "learning_rate": 0.00021479802143446, "loss": 0.1439, "step": 1879 }, { "epoch": 1.05, "learning_rate": 0.00021473619126133553, "loss": 0.0632, "step": 1880 }, { "epoch": 1.05, "learning_rate": 0.00021467436108821105, "loss": 0.1254, "step": 1881 }, { "epoch": 1.05, "learning_rate": 0.00021461253091508654, "loss": 0.2267, "step": 1882 }, { "epoch": 1.06, "learning_rate": 0.00021455070074196205, "loss": 0.2657, "step": 1883 }, { "epoch": 1.06, "learning_rate": 0.00021448887056883757, "loss": 0.2697, "step": 1884 }, { "epoch": 1.06, "learning_rate": 0.0002144270403957131, "loss": 0.0798, "step": 1885 }, { "epoch": 1.06, "learning_rate": 0.0002143652102225886, "loss": 0.0785, "step": 1886 }, { "epoch": 1.06, "learning_rate": 0.00021430338004946412, "loss": 0.0846, "step": 1887 }, { "epoch": 1.06, "learning_rate": 0.00021424154987633964, "loss": 0.129, "step": 1888 }, { "epoch": 1.06, "learning_rate": 0.00021417971970321515, "loss": 0.0863, "step": 1889 }, { "epoch": 1.06, "learning_rate": 0.00021411788953009067, "loss": 0.109, "step": 1890 }, { "epoch": 1.06, "learning_rate": 0.00021405605935696616, "loss": 0.1047, "step": 1891 }, { "epoch": 1.06, "learning_rate": 0.0002139942291838417, "loss": 0.1369, "step": 1892 }, { "epoch": 1.06, "learning_rate": 0.00021393239901071722, "loss": 0.1069, "step": 1893 }, { "epoch": 1.06, "learning_rate": 0.00021387056883759274, "loss": 0.0661, "step": 1894 }, { "epoch": 1.06, "learning_rate": 0.00021380873866446822, "loss": 0.067, "step": 1895 }, { "epoch": 1.06, "learning_rate": 0.00021374690849134374, "loss": 0.0583, "step": 1896 }, { "epoch": 1.06, "learning_rate": 0.00021368507831821926, "loss": 0.1064, "step": 1897 }, { "epoch": 1.06, "learning_rate": 0.0002136232481450948, "loss": 0.0858, "step": 1898 }, { "epoch": 1.06, "learning_rate": 0.00021356141797197032, "loss": 0.0716, "step": 1899 }, { "epoch": 1.07, "learning_rate": 0.0002134995877988458, "loss": 0.1326, "step": 1900 }, { "epoch": 1.07, "learning_rate": 0.00021343775762572132, "loss": 0.1038, "step": 1901 }, { "epoch": 1.07, "learning_rate": 0.00021337592745259684, "loss": 0.0499, "step": 1902 }, { "epoch": 1.07, "learning_rate": 0.00021331409727947238, "loss": 0.0713, "step": 1903 }, { "epoch": 1.07, "learning_rate": 0.00021325226710634787, "loss": 0.1082, "step": 1904 }, { "epoch": 1.07, "learning_rate": 0.0002131904369332234, "loss": 0.0877, "step": 1905 }, { "epoch": 1.07, "learning_rate": 0.0002131286067600989, "loss": 0.0831, "step": 1906 }, { "epoch": 1.07, "learning_rate": 0.00021306677658697442, "loss": 0.0912, "step": 1907 }, { "epoch": 1.07, "learning_rate": 0.00021300494641384994, "loss": 0.0582, "step": 1908 }, { "epoch": 1.07, "learning_rate": 0.00021294311624072543, "loss": 0.1076, "step": 1909 }, { "epoch": 1.07, "learning_rate": 0.00021288128606760097, "loss": 0.1143, "step": 1910 }, { "epoch": 1.07, "learning_rate": 0.0002128194558944765, "loss": 0.1127, "step": 1911 }, { "epoch": 1.07, "learning_rate": 0.000212757625721352, "loss": 0.0955, "step": 1912 }, { "epoch": 1.07, "learning_rate": 0.00021269579554822752, "loss": 0.0614, "step": 1913 }, { "epoch": 1.07, "learning_rate": 0.000212633965375103, "loss": 0.0854, "step": 1914 }, { "epoch": 1.07, "learning_rate": 0.00021257213520197856, "loss": 0.0732, "step": 1915 }, { "epoch": 1.07, "learning_rate": 0.00021251030502885407, "loss": 0.0664, "step": 1916 }, { "epoch": 1.07, "learning_rate": 0.0002124484748557296, "loss": 0.0971, "step": 1917 }, { "epoch": 1.08, "learning_rate": 0.00021238664468260508, "loss": 0.1093, "step": 1918 }, { "epoch": 1.08, "learning_rate": 0.0002123248145094806, "loss": 0.0612, "step": 1919 }, { "epoch": 1.08, "learning_rate": 0.0002122629843363561, "loss": 0.0833, "step": 1920 }, { "epoch": 1.08, "learning_rate": 0.00021220115416323166, "loss": 0.0858, "step": 1921 }, { "epoch": 1.08, "learning_rate": 0.00021213932399010717, "loss": 0.1039, "step": 1922 }, { "epoch": 1.08, "learning_rate": 0.00021207749381698266, "loss": 0.0946, "step": 1923 }, { "epoch": 1.08, "learning_rate": 0.00021201566364385818, "loss": 0.0969, "step": 1924 }, { "epoch": 1.08, "learning_rate": 0.0002119538334707337, "loss": 0.1349, "step": 1925 }, { "epoch": 1.08, "learning_rate": 0.0002118920032976092, "loss": 0.0968, "step": 1926 }, { "epoch": 1.08, "learning_rate": 0.0002118301731244847, "loss": 0.0875, "step": 1927 }, { "epoch": 1.08, "learning_rate": 0.00021176834295136025, "loss": 0.1203, "step": 1928 }, { "epoch": 1.08, "learning_rate": 0.00021170651277823576, "loss": 0.084, "step": 1929 }, { "epoch": 1.08, "learning_rate": 0.00021164468260511128, "loss": 0.1266, "step": 1930 }, { "epoch": 1.08, "learning_rate": 0.0002115828524319868, "loss": 0.107, "step": 1931 }, { "epoch": 1.08, "learning_rate": 0.00021152102225886228, "loss": 0.1283, "step": 1932 }, { "epoch": 1.08, "learning_rate": 0.00021145919208573783, "loss": 0.1764, "step": 1933 }, { "epoch": 1.08, "learning_rate": 0.00021139736191261334, "loss": 0.2608, "step": 1934 }, { "epoch": 1.08, "learning_rate": 0.00021133553173948886, "loss": 0.0695, "step": 1935 }, { "epoch": 1.09, "learning_rate": 0.00021127370156636438, "loss": 0.0743, "step": 1936 }, { "epoch": 1.09, "learning_rate": 0.00021121187139323987, "loss": 0.0766, "step": 1937 }, { "epoch": 1.09, "learning_rate": 0.00021115004122011538, "loss": 0.0774, "step": 1938 }, { "epoch": 1.09, "learning_rate": 0.00021108821104699093, "loss": 0.0998, "step": 1939 }, { "epoch": 1.09, "learning_rate": 0.00021102638087386644, "loss": 0.1181, "step": 1940 }, { "epoch": 1.09, "learning_rate": 0.00021096455070074193, "loss": 0.0969, "step": 1941 }, { "epoch": 1.09, "learning_rate": 0.00021090272052761745, "loss": 0.0607, "step": 1942 }, { "epoch": 1.09, "learning_rate": 0.00021084089035449297, "loss": 0.1248, "step": 1943 }, { "epoch": 1.09, "learning_rate": 0.0002107790601813685, "loss": 0.0737, "step": 1944 }, { "epoch": 1.09, "learning_rate": 0.00021071723000824403, "loss": 0.0835, "step": 1945 }, { "epoch": 1.09, "learning_rate": 0.00021065539983511952, "loss": 0.0895, "step": 1946 }, { "epoch": 1.09, "learning_rate": 0.00021059356966199503, "loss": 0.1079, "step": 1947 }, { "epoch": 1.09, "learning_rate": 0.00021053173948887055, "loss": 0.12, "step": 1948 }, { "epoch": 1.09, "learning_rate": 0.00021046990931574607, "loss": 0.0946, "step": 1949 }, { "epoch": 1.09, "learning_rate": 0.00021040807914262156, "loss": 0.048, "step": 1950 }, { "epoch": 1.09, "learning_rate": 0.0002103462489694971, "loss": 0.0827, "step": 1951 }, { "epoch": 1.09, "learning_rate": 0.00021028441879637262, "loss": 0.0795, "step": 1952 }, { "epoch": 1.09, "learning_rate": 0.00021022258862324813, "loss": 0.1059, "step": 1953 }, { "epoch": 1.1, "learning_rate": 0.00021016075845012365, "loss": 0.1049, "step": 1954 }, { "epoch": 1.1, "learning_rate": 0.00021009892827699914, "loss": 0.0766, "step": 1955 }, { "epoch": 1.1, "learning_rate": 0.00021003709810387466, "loss": 0.0741, "step": 1956 }, { "epoch": 1.1, "learning_rate": 0.0002099752679307502, "loss": 0.0827, "step": 1957 }, { "epoch": 1.1, "learning_rate": 0.00020991343775762572, "loss": 0.0875, "step": 1958 }, { "epoch": 1.1, "learning_rate": 0.0002098516075845012, "loss": 0.1027, "step": 1959 }, { "epoch": 1.1, "learning_rate": 0.00020978977741137672, "loss": 0.0692, "step": 1960 }, { "epoch": 1.1, "learning_rate": 0.00020972794723825224, "loss": 0.0912, "step": 1961 }, { "epoch": 1.1, "learning_rate": 0.00020966611706512778, "loss": 0.0792, "step": 1962 }, { "epoch": 1.1, "learning_rate": 0.0002096042868920033, "loss": 0.0615, "step": 1963 }, { "epoch": 1.1, "learning_rate": 0.0002095424567188788, "loss": 0.1258, "step": 1964 }, { "epoch": 1.1, "learning_rate": 0.0002094806265457543, "loss": 0.0958, "step": 1965 }, { "epoch": 1.1, "learning_rate": 0.00020941879637262982, "loss": 0.0578, "step": 1966 }, { "epoch": 1.1, "learning_rate": 0.00020935696619950534, "loss": 0.0784, "step": 1967 }, { "epoch": 1.1, "learning_rate": 0.00020929513602638088, "loss": 0.0279, "step": 1968 }, { "epoch": 1.1, "learning_rate": 0.00020923330585325637, "loss": 0.0867, "step": 1969 }, { "epoch": 1.1, "learning_rate": 0.0002091714756801319, "loss": 0.0655, "step": 1970 }, { "epoch": 1.1, "learning_rate": 0.0002091096455070074, "loss": 0.0621, "step": 1971 }, { "epoch": 1.11, "learning_rate": 0.00020904781533388292, "loss": 0.0938, "step": 1972 }, { "epoch": 1.11, "learning_rate": 0.0002089859851607584, "loss": 0.0927, "step": 1973 }, { "epoch": 1.11, "learning_rate": 0.00020892415498763395, "loss": 0.0959, "step": 1974 }, { "epoch": 1.11, "learning_rate": 0.00020886232481450947, "loss": 0.1179, "step": 1975 }, { "epoch": 1.11, "learning_rate": 0.000208800494641385, "loss": 0.1233, "step": 1976 }, { "epoch": 1.11, "learning_rate": 0.0002087386644682605, "loss": 0.0592, "step": 1977 }, { "epoch": 1.11, "learning_rate": 0.000208676834295136, "loss": 0.0644, "step": 1978 }, { "epoch": 1.11, "learning_rate": 0.0002086150041220115, "loss": 0.0831, "step": 1979 }, { "epoch": 1.11, "learning_rate": 0.00020855317394888705, "loss": 0.087, "step": 1980 }, { "epoch": 1.11, "learning_rate": 0.00020849134377576257, "loss": 0.2104, "step": 1981 }, { "epoch": 1.11, "learning_rate": 0.00020842951360263806, "loss": 0.1361, "step": 1982 }, { "epoch": 1.11, "learning_rate": 0.00020836768342951358, "loss": 0.2488, "step": 1983 }, { "epoch": 1.11, "learning_rate": 0.0002083058532563891, "loss": 0.2882, "step": 1984 }, { "epoch": 1.11, "learning_rate": 0.0002082440230832646, "loss": 0.0881, "step": 1985 }, { "epoch": 1.11, "learning_rate": 0.00020818219291014015, "loss": 0.0992, "step": 1986 }, { "epoch": 1.11, "learning_rate": 0.00020812036273701564, "loss": 0.077, "step": 1987 }, { "epoch": 1.11, "learning_rate": 0.00020805853256389116, "loss": 0.1063, "step": 1988 }, { "epoch": 1.11, "learning_rate": 0.00020799670239076668, "loss": 0.0827, "step": 1989 }, { "epoch": 1.12, "learning_rate": 0.0002079348722176422, "loss": 0.0907, "step": 1990 }, { "epoch": 1.12, "learning_rate": 0.00020787304204451768, "loss": 0.08, "step": 1991 }, { "epoch": 1.12, "learning_rate": 0.00020781121187139323, "loss": 0.0776, "step": 1992 }, { "epoch": 1.12, "learning_rate": 0.00020774938169826874, "loss": 0.0771, "step": 1993 }, { "epoch": 1.12, "learning_rate": 0.00020768755152514426, "loss": 0.0755, "step": 1994 }, { "epoch": 1.12, "learning_rate": 0.00020762572135201978, "loss": 0.0848, "step": 1995 }, { "epoch": 1.12, "learning_rate": 0.00020756389117889526, "loss": 0.0968, "step": 1996 }, { "epoch": 1.12, "learning_rate": 0.00020750206100577078, "loss": 0.0771, "step": 1997 }, { "epoch": 1.12, "learning_rate": 0.00020744023083264632, "loss": 0.093, "step": 1998 }, { "epoch": 1.12, "learning_rate": 0.00020737840065952184, "loss": 0.1155, "step": 1999 }, { "epoch": 1.12, "learning_rate": 0.00020731657048639736, "loss": 0.1031, "step": 2000 }, { "epoch": 1.12, "learning_rate": 0.00020725474031327285, "loss": 0.1007, "step": 2001 }, { "epoch": 1.12, "learning_rate": 0.00020719291014014836, "loss": 0.0806, "step": 2002 }, { "epoch": 1.12, "learning_rate": 0.00020713107996702388, "loss": 0.0682, "step": 2003 }, { "epoch": 1.12, "learning_rate": 0.00020706924979389942, "loss": 0.1091, "step": 2004 }, { "epoch": 1.12, "learning_rate": 0.00020700741962077491, "loss": 0.0813, "step": 2005 }, { "epoch": 1.12, "learning_rate": 0.00020694558944765043, "loss": 0.0691, "step": 2006 }, { "epoch": 1.12, "learning_rate": 0.00020688375927452595, "loss": 0.0586, "step": 2007 }, { "epoch": 1.13, "learning_rate": 0.00020682192910140146, "loss": 0.059, "step": 2008 }, { "epoch": 1.13, "learning_rate": 0.000206760098928277, "loss": 0.0921, "step": 2009 }, { "epoch": 1.13, "learning_rate": 0.0002066982687551525, "loss": 0.069, "step": 2010 }, { "epoch": 1.13, "learning_rate": 0.000206636438582028, "loss": 0.0506, "step": 2011 }, { "epoch": 1.13, "learning_rate": 0.00020657460840890353, "loss": 0.1181, "step": 2012 }, { "epoch": 1.13, "learning_rate": 0.00020651277823577905, "loss": 0.0652, "step": 2013 }, { "epoch": 1.13, "learning_rate": 0.00020645094806265454, "loss": 0.1039, "step": 2014 }, { "epoch": 1.13, "learning_rate": 0.00020638911788953005, "loss": 0.0575, "step": 2015 }, { "epoch": 1.13, "learning_rate": 0.0002063272877164056, "loss": 0.0786, "step": 2016 }, { "epoch": 1.13, "learning_rate": 0.0002062654575432811, "loss": 0.099, "step": 2017 }, { "epoch": 1.13, "learning_rate": 0.00020620362737015663, "loss": 0.0979, "step": 2018 }, { "epoch": 1.13, "learning_rate": 0.00020614179719703212, "loss": 0.078, "step": 2019 }, { "epoch": 1.13, "learning_rate": 0.00020607996702390764, "loss": 0.0837, "step": 2020 }, { "epoch": 1.13, "learning_rate": 0.00020601813685078318, "loss": 0.1042, "step": 2021 }, { "epoch": 1.13, "learning_rate": 0.0002059563066776587, "loss": 0.063, "step": 2022 }, { "epoch": 1.13, "learning_rate": 0.00020589447650453419, "loss": 0.0611, "step": 2023 }, { "epoch": 1.13, "learning_rate": 0.0002058326463314097, "loss": 0.1136, "step": 2024 }, { "epoch": 1.14, "learning_rate": 0.00020577081615828522, "loss": 0.1112, "step": 2025 }, { "epoch": 1.14, "learning_rate": 0.00020570898598516074, "loss": 0.1111, "step": 2026 }, { "epoch": 1.14, "learning_rate": 0.00020564715581203628, "loss": 0.0813, "step": 2027 }, { "epoch": 1.14, "learning_rate": 0.00020558532563891177, "loss": 0.0493, "step": 2028 }, { "epoch": 1.14, "learning_rate": 0.00020552349546578728, "loss": 0.1125, "step": 2029 }, { "epoch": 1.14, "learning_rate": 0.0002054616652926628, "loss": 0.0873, "step": 2030 }, { "epoch": 1.14, "learning_rate": 0.00020539983511953832, "loss": 0.2142, "step": 2031 }, { "epoch": 1.14, "learning_rate": 0.00020533800494641383, "loss": 0.0783, "step": 2032 }, { "epoch": 1.14, "learning_rate": 0.00020527617477328932, "loss": 0.1983, "step": 2033 }, { "epoch": 1.14, "learning_rate": 0.00020521434460016487, "loss": 0.1395, "step": 2034 }, { "epoch": 1.14, "learning_rate": 0.00020515251442704038, "loss": 0.0966, "step": 2035 }, { "epoch": 1.14, "learning_rate": 0.0002050906842539159, "loss": 0.0805, "step": 2036 }, { "epoch": 1.14, "learning_rate": 0.0002050288540807914, "loss": 0.0851, "step": 2037 }, { "epoch": 1.14, "learning_rate": 0.0002049670239076669, "loss": 0.1095, "step": 2038 }, { "epoch": 1.14, "learning_rate": 0.00020490519373454245, "loss": 0.0795, "step": 2039 }, { "epoch": 1.14, "learning_rate": 0.00020484336356141797, "loss": 0.0898, "step": 2040 }, { "epoch": 1.14, "learning_rate": 0.00020478153338829348, "loss": 0.1162, "step": 2041 }, { "epoch": 1.14, "learning_rate": 0.00020471970321516897, "loss": 0.1315, "step": 2042 }, { "epoch": 1.15, "learning_rate": 0.0002046578730420445, "loss": 0.0931, "step": 2043 }, { "epoch": 1.15, "learning_rate": 0.00020459604286892, "loss": 0.0742, "step": 2044 }, { "epoch": 1.15, "learning_rate": 0.00020453421269579555, "loss": 0.1063, "step": 2045 }, { "epoch": 1.15, "learning_rate": 0.00020447238252267104, "loss": 0.1003, "step": 2046 }, { "epoch": 1.15, "learning_rate": 0.00020441055234954656, "loss": 0.062, "step": 2047 }, { "epoch": 1.15, "learning_rate": 0.00020434872217642207, "loss": 0.0971, "step": 2048 }, { "epoch": 1.15, "learning_rate": 0.0002042868920032976, "loss": 0.1057, "step": 2049 }, { "epoch": 1.15, "learning_rate": 0.00020422506183017313, "loss": 0.0752, "step": 2050 }, { "epoch": 1.15, "learning_rate": 0.00020416323165704862, "loss": 0.0637, "step": 2051 }, { "epoch": 1.15, "learning_rate": 0.00020410140148392414, "loss": 0.0754, "step": 2052 }, { "epoch": 1.15, "learning_rate": 0.00020403957131079966, "loss": 0.0694, "step": 2053 }, { "epoch": 1.15, "learning_rate": 0.00020397774113767517, "loss": 0.0752, "step": 2054 }, { "epoch": 1.15, "learning_rate": 0.0002039159109645507, "loss": 0.0724, "step": 2055 }, { "epoch": 1.15, "learning_rate": 0.00020385408079142618, "loss": 0.0945, "step": 2056 }, { "epoch": 1.15, "learning_rate": 0.00020379225061830172, "loss": 0.0763, "step": 2057 }, { "epoch": 1.15, "learning_rate": 0.00020373042044517724, "loss": 0.0714, "step": 2058 }, { "epoch": 1.15, "learning_rate": 0.00020366859027205276, "loss": 0.0961, "step": 2059 }, { "epoch": 1.15, "learning_rate": 0.00020360676009892824, "loss": 0.0725, "step": 2060 }, { "epoch": 1.16, "learning_rate": 0.00020354492992580376, "loss": 0.0649, "step": 2061 }, { "epoch": 1.16, "learning_rate": 0.00020348309975267928, "loss": 0.0538, "step": 2062 }, { "epoch": 1.16, "learning_rate": 0.00020342126957955482, "loss": 0.0852, "step": 2063 }, { "epoch": 1.16, "learning_rate": 0.00020335943940643034, "loss": 0.086, "step": 2064 }, { "epoch": 1.16, "learning_rate": 0.00020329760923330583, "loss": 0.0883, "step": 2065 }, { "epoch": 1.16, "learning_rate": 0.00020323577906018134, "loss": 0.1046, "step": 2066 }, { "epoch": 1.16, "learning_rate": 0.00020317394888705686, "loss": 0.0734, "step": 2067 }, { "epoch": 1.16, "learning_rate": 0.0002031121187139324, "loss": 0.0705, "step": 2068 }, { "epoch": 1.16, "learning_rate": 0.0002030502885408079, "loss": 0.1194, "step": 2069 }, { "epoch": 1.16, "learning_rate": 0.0002029884583676834, "loss": 0.0858, "step": 2070 }, { "epoch": 1.16, "learning_rate": 0.00020292662819455893, "loss": 0.0704, "step": 2071 }, { "epoch": 1.16, "learning_rate": 0.00020286479802143444, "loss": 0.0638, "step": 2072 }, { "epoch": 1.16, "learning_rate": 0.00020280296784830996, "loss": 0.0535, "step": 2073 }, { "epoch": 1.16, "learning_rate": 0.00020274113767518545, "loss": 0.0993, "step": 2074 }, { "epoch": 1.16, "learning_rate": 0.000202679307502061, "loss": 0.0634, "step": 2075 }, { "epoch": 1.16, "learning_rate": 0.0002026174773289365, "loss": 0.0839, "step": 2076 }, { "epoch": 1.16, "learning_rate": 0.00020255564715581203, "loss": 0.0557, "step": 2077 }, { "epoch": 1.16, "learning_rate": 0.00020249381698268752, "loss": 0.0903, "step": 2078 }, { "epoch": 1.17, "learning_rate": 0.00020243198680956303, "loss": 0.0739, "step": 2079 }, { "epoch": 1.17, "learning_rate": 0.00020237015663643855, "loss": 0.1675, "step": 2080 }, { "epoch": 1.17, "learning_rate": 0.0002023083264633141, "loss": 0.2446, "step": 2081 }, { "epoch": 1.17, "learning_rate": 0.0002022464962901896, "loss": 0.1563, "step": 2082 }, { "epoch": 1.17, "learning_rate": 0.0002021846661170651, "loss": 0.2471, "step": 2083 }, { "epoch": 1.17, "learning_rate": 0.00020212283594394062, "loss": 0.2868, "step": 2084 }, { "epoch": 1.17, "learning_rate": 0.00020206100577081613, "loss": 0.0897, "step": 2085 }, { "epoch": 1.17, "learning_rate": 0.00020199917559769168, "loss": 0.0682, "step": 2086 }, { "epoch": 1.17, "learning_rate": 0.0002019373454245672, "loss": 0.1152, "step": 2087 }, { "epoch": 1.17, "learning_rate": 0.00020187551525144268, "loss": 0.0641, "step": 2088 }, { "epoch": 1.17, "learning_rate": 0.0002018136850783182, "loss": 0.1173, "step": 2089 }, { "epoch": 1.17, "learning_rate": 0.00020175185490519372, "loss": 0.1116, "step": 2090 }, { "epoch": 1.17, "learning_rate": 0.00020169002473206923, "loss": 0.0753, "step": 2091 }, { "epoch": 1.17, "learning_rate": 0.00020162819455894472, "loss": 0.0923, "step": 2092 }, { "epoch": 1.17, "learning_rate": 0.00020156636438582027, "loss": 0.0921, "step": 2093 }, { "epoch": 1.17, "learning_rate": 0.00020150453421269578, "loss": 0.0732, "step": 2094 }, { "epoch": 1.17, "learning_rate": 0.0002014427040395713, "loss": 0.1132, "step": 2095 }, { "epoch": 1.17, "learning_rate": 0.00020138087386644681, "loss": 0.1302, "step": 2096 }, { "epoch": 1.18, "learning_rate": 0.0002013190436933223, "loss": 0.1063, "step": 2097 }, { "epoch": 1.18, "learning_rate": 0.00020125721352019785, "loss": 0.0824, "step": 2098 }, { "epoch": 1.18, "learning_rate": 0.00020119538334707336, "loss": 0.0952, "step": 2099 }, { "epoch": 1.18, "learning_rate": 0.00020113355317394888, "loss": 0.0652, "step": 2100 }, { "epoch": 1.18, "eval_loss": 0.12512445449829102, "eval_runtime": 269.5086, "eval_samples_per_second": 9.803, "eval_steps_per_second": 0.616, "eval_wer": 0.10053550178500595, "step": 2100 }, { "epoch": 1.18, "learning_rate": 0.00020107172300082437, "loss": 0.0873, "step": 2101 }, { "epoch": 1.18, "learning_rate": 0.0002010098928276999, "loss": 0.0832, "step": 2102 }, { "epoch": 1.18, "learning_rate": 0.0002009480626545754, "loss": 0.0666, "step": 2103 }, { "epoch": 1.18, "learning_rate": 0.00020088623248145095, "loss": 0.0757, "step": 2104 }, { "epoch": 1.18, "learning_rate": 0.00020082440230832646, "loss": 0.0874, "step": 2105 }, { "epoch": 1.18, "learning_rate": 0.00020076257213520195, "loss": 0.0648, "step": 2106 }, { "epoch": 1.18, "learning_rate": 0.00020070074196207747, "loss": 0.1355, "step": 2107 }, { "epoch": 1.18, "learning_rate": 0.000200638911788953, "loss": 0.0736, "step": 2108 }, { "epoch": 1.18, "learning_rate": 0.0002005770816158285, "loss": 0.0914, "step": 2109 }, { "epoch": 1.18, "learning_rate": 0.000200515251442704, "loss": 0.0802, "step": 2110 }, { "epoch": 1.18, "learning_rate": 0.00020045342126957954, "loss": 0.0872, "step": 2111 }, { "epoch": 1.18, "learning_rate": 0.00020039159109645505, "loss": 0.0601, "step": 2112 }, { "epoch": 1.18, "learning_rate": 0.00020032976092333057, "loss": 0.0881, "step": 2113 }, { "epoch": 1.18, "learning_rate": 0.00020026793075020609, "loss": 0.0798, "step": 2114 }, { "epoch": 1.19, "learning_rate": 0.00020020610057708158, "loss": 0.0733, "step": 2115 }, { "epoch": 1.19, "learning_rate": 0.00020014427040395712, "loss": 0.0865, "step": 2116 }, { "epoch": 1.19, "learning_rate": 0.00020008244023083264, "loss": 0.0964, "step": 2117 }, { "epoch": 1.19, "learning_rate": 0.00020002061005770815, "loss": 0.0772, "step": 2118 }, { "epoch": 1.19, "learning_rate": 0.00019995877988458367, "loss": 0.0621, "step": 2119 }, { "epoch": 1.19, "learning_rate": 0.00019989694971145916, "loss": 0.0741, "step": 2120 }, { "epoch": 1.19, "learning_rate": 0.00019983511953833468, "loss": 0.0878, "step": 2121 }, { "epoch": 1.19, "learning_rate": 0.00019977328936521022, "loss": 0.1216, "step": 2122 }, { "epoch": 1.19, "learning_rate": 0.00019971145919208574, "loss": 0.0635, "step": 2123 }, { "epoch": 1.19, "learning_rate": 0.00019964962901896123, "loss": 0.0543, "step": 2124 }, { "epoch": 1.19, "learning_rate": 0.00019958779884583674, "loss": 0.0828, "step": 2125 }, { "epoch": 1.19, "learning_rate": 0.00019952596867271226, "loss": 0.1144, "step": 2126 }, { "epoch": 1.19, "learning_rate": 0.0001994641384995878, "loss": 0.0718, "step": 2127 }, { "epoch": 1.19, "learning_rate": 0.00019940230832646332, "loss": 0.086, "step": 2128 }, { "epoch": 1.19, "learning_rate": 0.0001993404781533388, "loss": 0.0919, "step": 2129 }, { "epoch": 1.19, "learning_rate": 0.00019927864798021432, "loss": 0.0997, "step": 2130 }, { "epoch": 1.19, "learning_rate": 0.00019921681780708984, "loss": 0.1087, "step": 2131 }, { "epoch": 1.2, "learning_rate": 0.00019915498763396536, "loss": 0.1622, "step": 2132 }, { "epoch": 1.2, "learning_rate": 0.00019909315746084085, "loss": 0.1715, "step": 2133 }, { "epoch": 1.2, "learning_rate": 0.0001990313272877164, "loss": 0.2882, "step": 2134 }, { "epoch": 1.2, "learning_rate": 0.0001989694971145919, "loss": 0.0627, "step": 2135 }, { "epoch": 1.2, "learning_rate": 0.00019890766694146742, "loss": 0.1077, "step": 2136 }, { "epoch": 1.2, "learning_rate": 0.00019884583676834294, "loss": 0.0838, "step": 2137 }, { "epoch": 1.2, "learning_rate": 0.00019878400659521843, "loss": 0.0769, "step": 2138 }, { "epoch": 1.2, "learning_rate": 0.00019872217642209395, "loss": 0.0621, "step": 2139 }, { "epoch": 1.2, "learning_rate": 0.0001986603462489695, "loss": 0.1242, "step": 2140 }, { "epoch": 1.2, "learning_rate": 0.000198598516075845, "loss": 0.072, "step": 2141 }, { "epoch": 1.2, "learning_rate": 0.0001985366859027205, "loss": 0.0837, "step": 2142 }, { "epoch": 1.2, "learning_rate": 0.000198474855729596, "loss": 0.086, "step": 2143 }, { "epoch": 1.2, "learning_rate": 0.00019841302555647153, "loss": 0.0836, "step": 2144 }, { "epoch": 1.2, "learning_rate": 0.00019835119538334707, "loss": 0.084, "step": 2145 }, { "epoch": 1.2, "learning_rate": 0.0001982893652102226, "loss": 0.0714, "step": 2146 }, { "epoch": 1.2, "learning_rate": 0.00019822753503709808, "loss": 0.0938, "step": 2147 }, { "epoch": 1.2, "learning_rate": 0.0001981657048639736, "loss": 0.0943, "step": 2148 }, { "epoch": 1.2, "learning_rate": 0.0001981038746908491, "loss": 0.0639, "step": 2149 }, { "epoch": 1.21, "learning_rate": 0.00019804204451772463, "loss": 0.1092, "step": 2150 }, { "epoch": 1.21, "learning_rate": 0.00019798021434460017, "loss": 0.1114, "step": 2151 }, { "epoch": 1.21, "learning_rate": 0.00019791838417147566, "loss": 0.115, "step": 2152 }, { "epoch": 1.21, "learning_rate": 0.00019785655399835118, "loss": 0.0788, "step": 2153 }, { "epoch": 1.21, "learning_rate": 0.0001977947238252267, "loss": 0.0957, "step": 2154 }, { "epoch": 1.21, "learning_rate": 0.0001977328936521022, "loss": 0.1138, "step": 2155 }, { "epoch": 1.21, "learning_rate": 0.0001976710634789777, "loss": 0.0607, "step": 2156 }, { "epoch": 1.21, "learning_rate": 0.00019760923330585325, "loss": 0.0947, "step": 2157 }, { "epoch": 1.21, "learning_rate": 0.00019754740313272876, "loss": 0.0756, "step": 2158 }, { "epoch": 1.21, "learning_rate": 0.00019748557295960428, "loss": 0.0709, "step": 2159 }, { "epoch": 1.21, "learning_rate": 0.0001974237427864798, "loss": 0.1069, "step": 2160 }, { "epoch": 1.21, "learning_rate": 0.00019736191261335528, "loss": 0.0662, "step": 2161 }, { "epoch": 1.21, "learning_rate": 0.0001973000824402308, "loss": 0.0774, "step": 2162 }, { "epoch": 1.21, "learning_rate": 0.00019723825226710634, "loss": 0.0969, "step": 2163 }, { "epoch": 1.21, "learning_rate": 0.00019717642209398186, "loss": 0.1068, "step": 2164 }, { "epoch": 1.21, "learning_rate": 0.00019711459192085735, "loss": 0.088, "step": 2165 }, { "epoch": 1.21, "learning_rate": 0.00019705276174773287, "loss": 0.0838, "step": 2166 }, { "epoch": 1.21, "learning_rate": 0.00019699093157460838, "loss": 0.0786, "step": 2167 }, { "epoch": 1.22, "learning_rate": 0.0001969291014014839, "loss": 0.069, "step": 2168 }, { "epoch": 1.22, "learning_rate": 0.00019686727122835944, "loss": 0.0688, "step": 2169 }, { "epoch": 1.22, "learning_rate": 0.00019680544105523493, "loss": 0.1414, "step": 2170 }, { "epoch": 1.22, "learning_rate": 0.00019674361088211045, "loss": 0.0737, "step": 2171 }, { "epoch": 1.22, "learning_rate": 0.00019668178070898597, "loss": 0.0736, "step": 2172 }, { "epoch": 1.22, "learning_rate": 0.00019661995053586148, "loss": 0.0827, "step": 2173 }, { "epoch": 1.22, "learning_rate": 0.00019655812036273703, "loss": 0.1322, "step": 2174 }, { "epoch": 1.22, "learning_rate": 0.00019649629018961252, "loss": 0.0415, "step": 2175 }, { "epoch": 1.22, "learning_rate": 0.00019643446001648803, "loss": 0.0686, "step": 2176 }, { "epoch": 1.22, "learning_rate": 0.00019637262984336355, "loss": 0.0673, "step": 2177 }, { "epoch": 1.22, "learning_rate": 0.00019631079967023907, "loss": 0.0614, "step": 2178 }, { "epoch": 1.22, "learning_rate": 0.00019624896949711456, "loss": 0.0859, "step": 2179 }, { "epoch": 1.22, "learning_rate": 0.00019618713932399007, "loss": 0.0718, "step": 2180 }, { "epoch": 1.22, "learning_rate": 0.00019612530915086562, "loss": 0.1286, "step": 2181 }, { "epoch": 1.22, "learning_rate": 0.00019606347897774113, "loss": 0.0503, "step": 2182 }, { "epoch": 1.22, "learning_rate": 0.00019600164880461665, "loss": 0.1262, "step": 2183 }, { "epoch": 1.22, "learning_rate": 0.00019593981863149214, "loss": 0.1536, "step": 2184 }, { "epoch": 1.22, "learning_rate": 0.00019587798845836766, "loss": 0.1152, "step": 2185 }, { "epoch": 1.23, "learning_rate": 0.00019581615828524317, "loss": 0.102, "step": 2186 }, { "epoch": 1.23, "learning_rate": 0.00019575432811211872, "loss": 0.0492, "step": 2187 }, { "epoch": 1.23, "learning_rate": 0.0001956924979389942, "loss": 0.1001, "step": 2188 }, { "epoch": 1.23, "learning_rate": 0.00019563066776586972, "loss": 0.1417, "step": 2189 }, { "epoch": 1.23, "learning_rate": 0.00019556883759274524, "loss": 0.0738, "step": 2190 }, { "epoch": 1.23, "learning_rate": 0.00019550700741962076, "loss": 0.0923, "step": 2191 }, { "epoch": 1.23, "learning_rate": 0.0001954451772464963, "loss": 0.1094, "step": 2192 }, { "epoch": 1.23, "learning_rate": 0.0001953833470733718, "loss": 0.0825, "step": 2193 }, { "epoch": 1.23, "learning_rate": 0.0001953215169002473, "loss": 0.0848, "step": 2194 }, { "epoch": 1.23, "learning_rate": 0.00019525968672712282, "loss": 0.0809, "step": 2195 }, { "epoch": 1.23, "learning_rate": 0.00019519785655399834, "loss": 0.0955, "step": 2196 }, { "epoch": 1.23, "learning_rate": 0.00019513602638087383, "loss": 0.0738, "step": 2197 }, { "epoch": 1.23, "learning_rate": 0.00019507419620774934, "loss": 0.0935, "step": 2198 }, { "epoch": 1.23, "learning_rate": 0.0001950123660346249, "loss": 0.0605, "step": 2199 }, { "epoch": 1.23, "learning_rate": 0.0001949505358615004, "loss": 0.0603, "step": 2200 }, { "epoch": 1.23, "learning_rate": 0.00019488870568837592, "loss": 0.1155, "step": 2201 }, { "epoch": 1.23, "learning_rate": 0.0001948268755152514, "loss": 0.1349, "step": 2202 }, { "epoch": 1.23, "learning_rate": 0.00019476504534212693, "loss": 0.0509, "step": 2203 }, { "epoch": 1.24, "learning_rate": 0.00019470321516900247, "loss": 0.0933, "step": 2204 }, { "epoch": 1.24, "learning_rate": 0.000194641384995878, "loss": 0.077, "step": 2205 }, { "epoch": 1.24, "learning_rate": 0.0001945795548227535, "loss": 0.0884, "step": 2206 }, { "epoch": 1.24, "learning_rate": 0.000194517724649629, "loss": 0.1018, "step": 2207 }, { "epoch": 1.24, "learning_rate": 0.0001944558944765045, "loss": 0.0699, "step": 2208 }, { "epoch": 1.24, "learning_rate": 0.00019439406430338003, "loss": 0.0611, "step": 2209 }, { "epoch": 1.24, "learning_rate": 0.00019433223413025557, "loss": 0.0899, "step": 2210 }, { "epoch": 1.24, "learning_rate": 0.00019427040395713106, "loss": 0.0656, "step": 2211 }, { "epoch": 1.24, "learning_rate": 0.00019420857378400658, "loss": 0.105, "step": 2212 }, { "epoch": 1.24, "learning_rate": 0.0001941467436108821, "loss": 0.091, "step": 2213 }, { "epoch": 1.24, "learning_rate": 0.0001940849134377576, "loss": 0.1047, "step": 2214 }, { "epoch": 1.24, "learning_rate": 0.00019402308326463313, "loss": 0.0828, "step": 2215 }, { "epoch": 1.24, "learning_rate": 0.00019396125309150862, "loss": 0.093, "step": 2216 }, { "epoch": 1.24, "learning_rate": 0.00019389942291838416, "loss": 0.0732, "step": 2217 }, { "epoch": 1.24, "learning_rate": 0.00019383759274525968, "loss": 0.0636, "step": 2218 }, { "epoch": 1.24, "learning_rate": 0.0001937757625721352, "loss": 0.0982, "step": 2219 }, { "epoch": 1.24, "learning_rate": 0.00019371393239901068, "loss": 0.0738, "step": 2220 }, { "epoch": 1.24, "learning_rate": 0.0001936521022258862, "loss": 0.0818, "step": 2221 }, { "epoch": 1.25, "learning_rate": 0.00019359027205276174, "loss": 0.0662, "step": 2222 }, { "epoch": 1.25, "learning_rate": 0.00019352844187963726, "loss": 0.1105, "step": 2223 }, { "epoch": 1.25, "learning_rate": 0.00019346661170651278, "loss": 0.0847, "step": 2224 }, { "epoch": 1.25, "learning_rate": 0.00019340478153338826, "loss": 0.126, "step": 2225 }, { "epoch": 1.25, "learning_rate": 0.00019334295136026378, "loss": 0.0886, "step": 2226 }, { "epoch": 1.25, "learning_rate": 0.0001932811211871393, "loss": 0.0944, "step": 2227 }, { "epoch": 1.25, "learning_rate": 0.00019321929101401484, "loss": 0.1107, "step": 2228 }, { "epoch": 1.25, "learning_rate": 0.00019315746084089033, "loss": 0.1181, "step": 2229 }, { "epoch": 1.25, "learning_rate": 0.00019309563066776585, "loss": 0.0955, "step": 2230 }, { "epoch": 1.25, "learning_rate": 0.00019303380049464136, "loss": 0.1028, "step": 2231 }, { "epoch": 1.25, "learning_rate": 0.00019297197032151688, "loss": 0.0857, "step": 2232 }, { "epoch": 1.25, "learning_rate": 0.00019291014014839242, "loss": 0.1636, "step": 2233 }, { "epoch": 1.25, "learning_rate": 0.00019284830997526791, "loss": 0.3423, "step": 2234 }, { "epoch": 1.25, "learning_rate": 0.00019278647980214343, "loss": 0.0892, "step": 2235 }, { "epoch": 1.25, "learning_rate": 0.00019272464962901895, "loss": 0.0534, "step": 2236 }, { "epoch": 1.25, "learning_rate": 0.00019266281945589446, "loss": 0.0699, "step": 2237 }, { "epoch": 1.25, "learning_rate": 0.00019260098928276998, "loss": 0.117, "step": 2238 }, { "epoch": 1.26, "learning_rate": 0.00019253915910964547, "loss": 0.0678, "step": 2239 }, { "epoch": 1.26, "learning_rate": 0.00019247732893652101, "loss": 0.0763, "step": 2240 }, { "epoch": 1.26, "learning_rate": 0.00019241549876339653, "loss": 0.0699, "step": 2241 }, { "epoch": 1.26, "learning_rate": 0.00019235366859027205, "loss": 0.1058, "step": 2242 }, { "epoch": 1.26, "learning_rate": 0.00019229183841714754, "loss": 0.0896, "step": 2243 }, { "epoch": 1.26, "learning_rate": 0.00019223000824402305, "loss": 0.0655, "step": 2244 }, { "epoch": 1.26, "learning_rate": 0.00019216817807089857, "loss": 0.093, "step": 2245 }, { "epoch": 1.26, "learning_rate": 0.0001921063478977741, "loss": 0.0993, "step": 2246 }, { "epoch": 1.26, "learning_rate": 0.00019204451772464963, "loss": 0.0731, "step": 2247 }, { "epoch": 1.26, "learning_rate": 0.00019198268755152512, "loss": 0.0946, "step": 2248 }, { "epoch": 1.26, "learning_rate": 0.00019192085737840064, "loss": 0.0625, "step": 2249 }, { "epoch": 1.26, "learning_rate": 0.00019185902720527615, "loss": 0.0678, "step": 2250 }, { "epoch": 1.26, "learning_rate": 0.0001917971970321517, "loss": 0.0787, "step": 2251 }, { "epoch": 1.26, "learning_rate": 0.00019173536685902719, "loss": 0.0751, "step": 2252 }, { "epoch": 1.26, "learning_rate": 0.0001916735366859027, "loss": 0.0831, "step": 2253 }, { "epoch": 1.26, "learning_rate": 0.00019161170651277822, "loss": 0.0979, "step": 2254 }, { "epoch": 1.26, "learning_rate": 0.00019154987633965374, "loss": 0.0832, "step": 2255 }, { "epoch": 1.26, "learning_rate": 0.00019148804616652925, "loss": 0.0821, "step": 2256 }, { "epoch": 1.27, "learning_rate": 0.00019142621599340474, "loss": 0.0728, "step": 2257 }, { "epoch": 1.27, "learning_rate": 0.00019136438582028029, "loss": 0.0888, "step": 2258 }, { "epoch": 1.27, "learning_rate": 0.0001913025556471558, "loss": 0.1258, "step": 2259 }, { "epoch": 1.27, "learning_rate": 0.00019124072547403132, "loss": 0.0807, "step": 2260 }, { "epoch": 1.27, "learning_rate": 0.0001911788953009068, "loss": 0.0825, "step": 2261 }, { "epoch": 1.27, "learning_rate": 0.00019111706512778232, "loss": 0.0812, "step": 2262 }, { "epoch": 1.27, "learning_rate": 0.00019105523495465784, "loss": 0.0515, "step": 2263 }, { "epoch": 1.27, "learning_rate": 0.00019099340478153338, "loss": 0.0603, "step": 2264 }, { "epoch": 1.27, "learning_rate": 0.0001909315746084089, "loss": 0.0631, "step": 2265 }, { "epoch": 1.27, "learning_rate": 0.0001908697444352844, "loss": 0.0673, "step": 2266 }, { "epoch": 1.27, "learning_rate": 0.0001908079142621599, "loss": 0.1144, "step": 2267 }, { "epoch": 1.27, "learning_rate": 0.00019074608408903542, "loss": 0.0696, "step": 2268 }, { "epoch": 1.27, "learning_rate": 0.00019068425391591097, "loss": 0.0714, "step": 2269 }, { "epoch": 1.27, "learning_rate": 0.00019062242374278648, "loss": 0.0712, "step": 2270 }, { "epoch": 1.27, "learning_rate": 0.00019056059356966197, "loss": 0.1035, "step": 2271 }, { "epoch": 1.27, "learning_rate": 0.0001904987633965375, "loss": 0.0841, "step": 2272 }, { "epoch": 1.27, "learning_rate": 0.000190436933223413, "loss": 0.0686, "step": 2273 }, { "epoch": 1.27, "learning_rate": 0.00019037510305028852, "loss": 0.0537, "step": 2274 }, { "epoch": 1.28, "learning_rate": 0.000190313272877164, "loss": 0.1384, "step": 2275 }, { "epoch": 1.28, "learning_rate": 0.00019025144270403956, "loss": 0.1137, "step": 2276 }, { "epoch": 1.28, "learning_rate": 0.00019018961253091507, "loss": 0.0859, "step": 2277 }, { "epoch": 1.28, "learning_rate": 0.0001901277823577906, "loss": 0.0781, "step": 2278 }, { "epoch": 1.28, "learning_rate": 0.0001900659521846661, "loss": 0.1526, "step": 2279 }, { "epoch": 1.28, "learning_rate": 0.0001900041220115416, "loss": 0.0766, "step": 2280 }, { "epoch": 1.28, "learning_rate": 0.00018994229183841714, "loss": 0.1861, "step": 2281 }, { "epoch": 1.28, "learning_rate": 0.00018988046166529266, "loss": 0.1914, "step": 2282 }, { "epoch": 1.28, "learning_rate": 0.00018981863149216817, "loss": 0.0995, "step": 2283 }, { "epoch": 1.28, "learning_rate": 0.00018975680131904366, "loss": 0.2009, "step": 2284 }, { "epoch": 1.28, "learning_rate": 0.00018969497114591918, "loss": 0.0954, "step": 2285 }, { "epoch": 1.28, "learning_rate": 0.0001896331409727947, "loss": 0.0987, "step": 2286 }, { "epoch": 1.28, "learning_rate": 0.00018957131079967024, "loss": 0.0755, "step": 2287 }, { "epoch": 1.28, "learning_rate": 0.00018950948062654576, "loss": 0.0738, "step": 2288 }, { "epoch": 1.28, "learning_rate": 0.00018944765045342124, "loss": 0.0944, "step": 2289 }, { "epoch": 1.28, "learning_rate": 0.00018938582028029676, "loss": 0.0583, "step": 2290 }, { "epoch": 1.28, "learning_rate": 0.00018932399010717228, "loss": 0.1035, "step": 2291 }, { "epoch": 1.28, "learning_rate": 0.0001892621599340478, "loss": 0.0778, "step": 2292 }, { "epoch": 1.29, "learning_rate": 0.00018920032976092334, "loss": 0.0548, "step": 2293 }, { "epoch": 1.29, "learning_rate": 0.00018913849958779883, "loss": 0.0794, "step": 2294 }, { "epoch": 1.29, "learning_rate": 0.00018907666941467434, "loss": 0.0637, "step": 2295 }, { "epoch": 1.29, "learning_rate": 0.00018901483924154986, "loss": 0.0911, "step": 2296 }, { "epoch": 1.29, "learning_rate": 0.00018895300906842538, "loss": 0.0791, "step": 2297 }, { "epoch": 1.29, "learning_rate": 0.00018889117889530087, "loss": 0.0764, "step": 2298 }, { "epoch": 1.29, "learning_rate": 0.0001888293487221764, "loss": 0.0937, "step": 2299 }, { "epoch": 1.29, "learning_rate": 0.00018876751854905193, "loss": 0.0727, "step": 2300 }, { "epoch": 1.29, "learning_rate": 0.00018870568837592744, "loss": 0.0823, "step": 2301 }, { "epoch": 1.29, "learning_rate": 0.00018864385820280296, "loss": 0.0981, "step": 2302 }, { "epoch": 1.29, "learning_rate": 0.00018858202802967845, "loss": 0.0752, "step": 2303 }, { "epoch": 1.29, "learning_rate": 0.00018852019785655397, "loss": 0.0698, "step": 2304 }, { "epoch": 1.29, "learning_rate": 0.0001884583676834295, "loss": 0.0809, "step": 2305 }, { "epoch": 1.29, "learning_rate": 0.00018839653751030503, "loss": 0.0765, "step": 2306 }, { "epoch": 1.29, "learning_rate": 0.00018833470733718052, "loss": 0.09, "step": 2307 }, { "epoch": 1.29, "learning_rate": 0.00018827287716405603, "loss": 0.0626, "step": 2308 }, { "epoch": 1.29, "learning_rate": 0.00018821104699093155, "loss": 0.0638, "step": 2309 }, { "epoch": 1.29, "learning_rate": 0.0001881492168178071, "loss": 0.0731, "step": 2310 }, { "epoch": 1.3, "learning_rate": 0.0001880873866446826, "loss": 0.0559, "step": 2311 }, { "epoch": 1.3, "learning_rate": 0.0001880255564715581, "loss": 0.0792, "step": 2312 }, { "epoch": 1.3, "learning_rate": 0.00018796372629843362, "loss": 0.0807, "step": 2313 }, { "epoch": 1.3, "learning_rate": 0.00018790189612530913, "loss": 0.0749, "step": 2314 }, { "epoch": 1.3, "learning_rate": 0.00018784006595218465, "loss": 0.0781, "step": 2315 }, { "epoch": 1.3, "learning_rate": 0.00018777823577906014, "loss": 0.055, "step": 2316 }, { "epoch": 1.3, "learning_rate": 0.00018771640560593568, "loss": 0.0958, "step": 2317 }, { "epoch": 1.3, "learning_rate": 0.0001876545754328112, "loss": 0.1208, "step": 2318 }, { "epoch": 1.3, "learning_rate": 0.00018759274525968672, "loss": 0.0543, "step": 2319 }, { "epoch": 1.3, "learning_rate": 0.00018753091508656223, "loss": 0.1199, "step": 2320 }, { "epoch": 1.3, "learning_rate": 0.00018746908491343772, "loss": 0.0672, "step": 2321 }, { "epoch": 1.3, "learning_rate": 0.00018740725474031324, "loss": 0.0655, "step": 2322 }, { "epoch": 1.3, "learning_rate": 0.00018734542456718878, "loss": 0.0986, "step": 2323 }, { "epoch": 1.3, "learning_rate": 0.0001872835943940643, "loss": 0.0574, "step": 2324 }, { "epoch": 1.3, "learning_rate": 0.00018722176422093981, "loss": 0.0839, "step": 2325 }, { "epoch": 1.3, "learning_rate": 0.0001871599340478153, "loss": 0.0628, "step": 2326 }, { "epoch": 1.3, "learning_rate": 0.00018709810387469082, "loss": 0.0856, "step": 2327 }, { "epoch": 1.3, "learning_rate": 0.00018703627370156636, "loss": 0.0836, "step": 2328 }, { "epoch": 1.31, "learning_rate": 0.00018697444352844188, "loss": 0.0926, "step": 2329 }, { "epoch": 1.31, "learning_rate": 0.00018691261335531737, "loss": 0.0547, "step": 2330 }, { "epoch": 1.31, "learning_rate": 0.0001868507831821929, "loss": 0.0971, "step": 2331 }, { "epoch": 1.31, "learning_rate": 0.0001867889530090684, "loss": 0.1074, "step": 2332 }, { "epoch": 1.31, "learning_rate": 0.00018672712283594392, "loss": 0.1077, "step": 2333 }, { "epoch": 1.31, "learning_rate": 0.00018666529266281946, "loss": 0.256, "step": 2334 }, { "epoch": 1.31, "learning_rate": 0.00018660346248969495, "loss": 0.1282, "step": 2335 }, { "epoch": 1.31, "learning_rate": 0.00018654163231657047, "loss": 0.0812, "step": 2336 }, { "epoch": 1.31, "learning_rate": 0.000186479802143446, "loss": 0.1052, "step": 2337 }, { "epoch": 1.31, "learning_rate": 0.0001864179719703215, "loss": 0.0692, "step": 2338 }, { "epoch": 1.31, "learning_rate": 0.000186356141797197, "loss": 0.0797, "step": 2339 }, { "epoch": 1.31, "learning_rate": 0.0001862943116240725, "loss": 0.097, "step": 2340 }, { "epoch": 1.31, "learning_rate": 0.00018623248145094805, "loss": 0.1095, "step": 2341 }, { "epoch": 1.31, "learning_rate": 0.00018617065127782357, "loss": 0.1052, "step": 2342 }, { "epoch": 1.31, "learning_rate": 0.00018610882110469909, "loss": 0.094, "step": 2343 }, { "epoch": 1.31, "learning_rate": 0.00018604699093157458, "loss": 0.0796, "step": 2344 }, { "epoch": 1.31, "learning_rate": 0.0001859851607584501, "loss": 0.0903, "step": 2345 }, { "epoch": 1.32, "learning_rate": 0.00018592333058532564, "loss": 0.0704, "step": 2346 }, { "epoch": 1.32, "learning_rate": 0.00018586150041220115, "loss": 0.0877, "step": 2347 }, { "epoch": 1.32, "learning_rate": 0.00018579967023907664, "loss": 0.0757, "step": 2348 }, { "epoch": 1.32, "learning_rate": 0.00018573784006595216, "loss": 0.0997, "step": 2349 }, { "epoch": 1.32, "learning_rate": 0.00018567600989282768, "loss": 0.1174, "step": 2350 }, { "epoch": 1.32, "learning_rate": 0.0001856141797197032, "loss": 0.064, "step": 2351 }, { "epoch": 1.32, "learning_rate": 0.00018555234954657874, "loss": 0.068, "step": 2352 }, { "epoch": 1.32, "learning_rate": 0.00018549051937345423, "loss": 0.0815, "step": 2353 }, { "epoch": 1.32, "learning_rate": 0.00018542868920032974, "loss": 0.1053, "step": 2354 }, { "epoch": 1.32, "learning_rate": 0.00018536685902720526, "loss": 0.1098, "step": 2355 }, { "epoch": 1.32, "learning_rate": 0.00018530502885408077, "loss": 0.0935, "step": 2356 }, { "epoch": 1.32, "learning_rate": 0.00018524319868095632, "loss": 0.115, "step": 2357 }, { "epoch": 1.32, "learning_rate": 0.0001851813685078318, "loss": 0.066, "step": 2358 }, { "epoch": 1.32, "learning_rate": 0.00018511953833470732, "loss": 0.0729, "step": 2359 }, { "epoch": 1.32, "learning_rate": 0.00018505770816158284, "loss": 0.0881, "step": 2360 }, { "epoch": 1.32, "learning_rate": 0.00018499587798845836, "loss": 0.0551, "step": 2361 }, { "epoch": 1.32, "learning_rate": 0.00018493404781533385, "loss": 0.0752, "step": 2362 }, { "epoch": 1.32, "learning_rate": 0.00018487221764220936, "loss": 0.0685, "step": 2363 }, { "epoch": 1.33, "learning_rate": 0.0001848103874690849, "loss": 0.1051, "step": 2364 }, { "epoch": 1.33, "learning_rate": 0.00018474855729596042, "loss": 0.0938, "step": 2365 }, { "epoch": 1.33, "learning_rate": 0.00018468672712283594, "loss": 0.0902, "step": 2366 }, { "epoch": 1.33, "learning_rate": 0.00018462489694971143, "loss": 0.0851, "step": 2367 }, { "epoch": 1.33, "learning_rate": 0.00018456306677658695, "loss": 0.1089, "step": 2368 }, { "epoch": 1.33, "learning_rate": 0.00018450123660346246, "loss": 0.0898, "step": 2369 }, { "epoch": 1.33, "learning_rate": 0.000184439406430338, "loss": 0.0868, "step": 2370 }, { "epoch": 1.33, "learning_rate": 0.0001843775762572135, "loss": 0.0743, "step": 2371 }, { "epoch": 1.33, "learning_rate": 0.000184315746084089, "loss": 0.1174, "step": 2372 }, { "epoch": 1.33, "learning_rate": 0.00018425391591096453, "loss": 0.0967, "step": 2373 }, { "epoch": 1.33, "learning_rate": 0.00018419208573784005, "loss": 0.1208, "step": 2374 }, { "epoch": 1.33, "learning_rate": 0.0001841302555647156, "loss": 0.0775, "step": 2375 }, { "epoch": 1.33, "learning_rate": 0.00018406842539159108, "loss": 0.0639, "step": 2376 }, { "epoch": 1.33, "learning_rate": 0.0001840065952184666, "loss": 0.0585, "step": 2377 }, { "epoch": 1.33, "learning_rate": 0.0001839447650453421, "loss": 0.1096, "step": 2378 }, { "epoch": 1.33, "learning_rate": 0.00018388293487221763, "loss": 0.0707, "step": 2379 }, { "epoch": 1.33, "learning_rate": 0.00018382110469909315, "loss": 0.1101, "step": 2380 }, { "epoch": 1.33, "learning_rate": 0.00018375927452596864, "loss": 0.0859, "step": 2381 }, { "epoch": 1.34, "learning_rate": 0.00018369744435284418, "loss": 0.0622, "step": 2382 }, { "epoch": 1.34, "learning_rate": 0.0001836356141797197, "loss": 0.1051, "step": 2383 }, { "epoch": 1.34, "learning_rate": 0.0001835737840065952, "loss": 0.272, "step": 2384 }, { "epoch": 1.34, "learning_rate": 0.0001835119538334707, "loss": 0.0643, "step": 2385 }, { "epoch": 1.34, "learning_rate": 0.00018345012366034622, "loss": 0.0809, "step": 2386 }, { "epoch": 1.34, "learning_rate": 0.00018338829348722176, "loss": 0.0963, "step": 2387 }, { "epoch": 1.34, "learning_rate": 0.00018332646331409728, "loss": 0.1109, "step": 2388 }, { "epoch": 1.34, "learning_rate": 0.0001832646331409728, "loss": 0.1211, "step": 2389 }, { "epoch": 1.34, "learning_rate": 0.00018320280296784828, "loss": 0.0995, "step": 2390 }, { "epoch": 1.34, "learning_rate": 0.0001831409727947238, "loss": 0.0876, "step": 2391 }, { "epoch": 1.34, "learning_rate": 0.00018307914262159932, "loss": 0.0918, "step": 2392 }, { "epoch": 1.34, "learning_rate": 0.00018301731244847486, "loss": 0.0869, "step": 2393 }, { "epoch": 1.34, "learning_rate": 0.00018295548227535035, "loss": 0.1072, "step": 2394 }, { "epoch": 1.34, "learning_rate": 0.00018289365210222587, "loss": 0.0865, "step": 2395 }, { "epoch": 1.34, "learning_rate": 0.00018283182192910138, "loss": 0.0631, "step": 2396 }, { "epoch": 1.34, "learning_rate": 0.0001827699917559769, "loss": 0.0667, "step": 2397 }, { "epoch": 1.34, "learning_rate": 0.00018270816158285242, "loss": 0.081, "step": 2398 }, { "epoch": 1.34, "learning_rate": 0.0001826463314097279, "loss": 0.1045, "step": 2399 }, { "epoch": 1.35, "learning_rate": 0.00018258450123660345, "loss": 0.0848, "step": 2400 }, { "epoch": 1.35, "eval_loss": 0.11876089870929718, "eval_runtime": 267.8535, "eval_samples_per_second": 9.864, "eval_steps_per_second": 0.62, "eval_wer": 0.096449821499405, "step": 2400 }, { "epoch": 1.35, "learning_rate": 0.00018252267106347897, "loss": 0.107, "step": 2401 }, { "epoch": 1.35, "learning_rate": 0.00018246084089035448, "loss": 0.0746, "step": 2402 }, { "epoch": 1.35, "learning_rate": 0.00018239901071722997, "loss": 0.108, "step": 2403 }, { "epoch": 1.35, "learning_rate": 0.0001823371805441055, "loss": 0.0708, "step": 2404 }, { "epoch": 1.35, "learning_rate": 0.00018227535037098103, "loss": 0.0958, "step": 2405 }, { "epoch": 1.35, "learning_rate": 0.00018221352019785655, "loss": 0.0942, "step": 2406 }, { "epoch": 1.35, "learning_rate": 0.00018215169002473207, "loss": 0.0856, "step": 2407 }, { "epoch": 1.35, "learning_rate": 0.00018208985985160756, "loss": 0.0753, "step": 2408 }, { "epoch": 1.35, "learning_rate": 0.00018202802967848307, "loss": 0.1033, "step": 2409 }, { "epoch": 1.35, "learning_rate": 0.0001819661995053586, "loss": 0.0725, "step": 2410 }, { "epoch": 1.35, "learning_rate": 0.00018190436933223413, "loss": 0.0848, "step": 2411 }, { "epoch": 1.35, "learning_rate": 0.00018184253915910965, "loss": 0.0623, "step": 2412 }, { "epoch": 1.35, "learning_rate": 0.00018178070898598514, "loss": 0.0982, "step": 2413 }, { "epoch": 1.35, "learning_rate": 0.00018171887881286066, "loss": 0.0795, "step": 2414 }, { "epoch": 1.35, "learning_rate": 0.00018165704863973617, "loss": 0.073, "step": 2415 }, { "epoch": 1.35, "learning_rate": 0.00018159521846661172, "loss": 0.0674, "step": 2416 }, { "epoch": 1.35, "learning_rate": 0.0001815333882934872, "loss": 0.0743, "step": 2417 }, { "epoch": 1.36, "learning_rate": 0.00018147155812036272, "loss": 0.0658, "step": 2418 }, { "epoch": 1.36, "learning_rate": 0.00018140972794723824, "loss": 0.0936, "step": 2419 }, { "epoch": 1.36, "learning_rate": 0.00018134789777411376, "loss": 0.1101, "step": 2420 }, { "epoch": 1.36, "learning_rate": 0.00018128606760098927, "loss": 0.0919, "step": 2421 }, { "epoch": 1.36, "learning_rate": 0.00018122423742786476, "loss": 0.0774, "step": 2422 }, { "epoch": 1.36, "learning_rate": 0.0001811624072547403, "loss": 0.0619, "step": 2423 }, { "epoch": 1.36, "learning_rate": 0.00018110057708161582, "loss": 0.1003, "step": 2424 }, { "epoch": 1.36, "learning_rate": 0.00018103874690849134, "loss": 0.0865, "step": 2425 }, { "epoch": 1.36, "learning_rate": 0.00018097691673536683, "loss": 0.1352, "step": 2426 }, { "epoch": 1.36, "learning_rate": 0.00018091508656224234, "loss": 0.0824, "step": 2427 }, { "epoch": 1.36, "learning_rate": 0.00018085325638911786, "loss": 0.0574, "step": 2428 }, { "epoch": 1.36, "learning_rate": 0.0001807914262159934, "loss": 0.0655, "step": 2429 }, { "epoch": 1.36, "learning_rate": 0.00018072959604286892, "loss": 0.0874, "step": 2430 }, { "epoch": 1.36, "learning_rate": 0.0001806677658697444, "loss": 0.1457, "step": 2431 }, { "epoch": 1.36, "learning_rate": 0.00018060593569661993, "loss": 0.1115, "step": 2432 }, { "epoch": 1.36, "learning_rate": 0.00018054410552349544, "loss": 0.1084, "step": 2433 }, { "epoch": 1.36, "learning_rate": 0.000180482275350371, "loss": 0.3598, "step": 2434 }, { "epoch": 1.36, "learning_rate": 0.00018042044517724648, "loss": 0.077, "step": 2435 }, { "epoch": 1.37, "learning_rate": 0.000180358615004122, "loss": 0.0578, "step": 2436 }, { "epoch": 1.37, "learning_rate": 0.0001802967848309975, "loss": 0.0796, "step": 2437 }, { "epoch": 1.37, "learning_rate": 0.00018023495465787303, "loss": 0.0855, "step": 2438 }, { "epoch": 1.37, "learning_rate": 0.00018017312448474854, "loss": 0.0808, "step": 2439 }, { "epoch": 1.37, "learning_rate": 0.00018011129431162403, "loss": 0.1005, "step": 2440 }, { "epoch": 1.37, "learning_rate": 0.00018004946413849958, "loss": 0.0871, "step": 2441 }, { "epoch": 1.37, "learning_rate": 0.0001799876339653751, "loss": 0.0959, "step": 2442 }, { "epoch": 1.37, "learning_rate": 0.0001799258037922506, "loss": 0.0866, "step": 2443 }, { "epoch": 1.37, "learning_rate": 0.00017986397361912613, "loss": 0.0913, "step": 2444 }, { "epoch": 1.37, "learning_rate": 0.00017980214344600162, "loss": 0.0548, "step": 2445 }, { "epoch": 1.37, "learning_rate": 0.00017974031327287713, "loss": 0.1291, "step": 2446 }, { "epoch": 1.37, "learning_rate": 0.00017967848309975268, "loss": 0.1139, "step": 2447 }, { "epoch": 1.37, "learning_rate": 0.0001796166529266282, "loss": 0.0636, "step": 2448 }, { "epoch": 1.37, "learning_rate": 0.00017955482275350368, "loss": 0.1008, "step": 2449 }, { "epoch": 1.37, "learning_rate": 0.0001794929925803792, "loss": 0.0699, "step": 2450 }, { "epoch": 1.37, "learning_rate": 0.00017943116240725472, "loss": 0.0673, "step": 2451 }, { "epoch": 1.37, "learning_rate": 0.00017936933223413026, "loss": 0.0902, "step": 2452 }, { "epoch": 1.38, "learning_rate": 0.00017930750206100578, "loss": 0.1222, "step": 2453 }, { "epoch": 1.38, "learning_rate": 0.00017924567188788126, "loss": 0.0755, "step": 2454 }, { "epoch": 1.38, "learning_rate": 0.00017918384171475678, "loss": 0.0872, "step": 2455 }, { "epoch": 1.38, "learning_rate": 0.0001791220115416323, "loss": 0.0657, "step": 2456 }, { "epoch": 1.38, "learning_rate": 0.00017906018136850781, "loss": 0.0888, "step": 2457 }, { "epoch": 1.38, "learning_rate": 0.0001789983511953833, "loss": 0.0638, "step": 2458 }, { "epoch": 1.38, "learning_rate": 0.00017893652102225885, "loss": 0.1135, "step": 2459 }, { "epoch": 1.38, "learning_rate": 0.00017887469084913436, "loss": 0.0641, "step": 2460 }, { "epoch": 1.38, "learning_rate": 0.00017881286067600988, "loss": 0.0788, "step": 2461 }, { "epoch": 1.38, "learning_rate": 0.0001787510305028854, "loss": 0.1065, "step": 2462 }, { "epoch": 1.38, "learning_rate": 0.0001786892003297609, "loss": 0.0831, "step": 2463 }, { "epoch": 1.38, "learning_rate": 0.00017862737015663643, "loss": 0.049, "step": 2464 }, { "epoch": 1.38, "learning_rate": 0.00017856553998351195, "loss": 0.1187, "step": 2465 }, { "epoch": 1.38, "learning_rate": 0.00017850370981038746, "loss": 0.071, "step": 2466 }, { "epoch": 1.38, "learning_rate": 0.00017844187963726295, "loss": 0.0897, "step": 2467 }, { "epoch": 1.38, "learning_rate": 0.00017838004946413847, "loss": 0.0712, "step": 2468 }, { "epoch": 1.38, "learning_rate": 0.000178318219291014, "loss": 0.0773, "step": 2469 }, { "epoch": 1.38, "learning_rate": 0.00017825638911788953, "loss": 0.0786, "step": 2470 }, { "epoch": 1.39, "learning_rate": 0.00017819455894476505, "loss": 0.0628, "step": 2471 }, { "epoch": 1.39, "learning_rate": 0.00017813272877164054, "loss": 0.0672, "step": 2472 }, { "epoch": 1.39, "learning_rate": 0.00017807089859851605, "loss": 0.0617, "step": 2473 }, { "epoch": 1.39, "learning_rate": 0.00017800906842539157, "loss": 0.0978, "step": 2474 }, { "epoch": 1.39, "learning_rate": 0.00017794723825226709, "loss": 0.0961, "step": 2475 }, { "epoch": 1.39, "learning_rate": 0.00017788540807914263, "loss": 0.0768, "step": 2476 }, { "epoch": 1.39, "learning_rate": 0.00017782357790601812, "loss": 0.1133, "step": 2477 }, { "epoch": 1.39, "learning_rate": 0.00017776174773289364, "loss": 0.0862, "step": 2478 }, { "epoch": 1.39, "learning_rate": 0.00017769991755976915, "loss": 0.0796, "step": 2479 }, { "epoch": 1.39, "learning_rate": 0.00017763808738664467, "loss": 0.1727, "step": 2480 }, { "epoch": 1.39, "learning_rate": 0.00017757625721352016, "loss": 0.069, "step": 2481 }, { "epoch": 1.39, "learning_rate": 0.0001775144270403957, "loss": 0.1398, "step": 2482 }, { "epoch": 1.39, "learning_rate": 0.00017745259686727122, "loss": 0.1146, "step": 2483 }, { "epoch": 1.39, "learning_rate": 0.00017739076669414674, "loss": 0.3214, "step": 2484 }, { "epoch": 1.39, "learning_rate": 0.00017732893652102225, "loss": 0.0966, "step": 2485 }, { "epoch": 1.39, "learning_rate": 0.00017726710634789774, "loss": 0.101, "step": 2486 }, { "epoch": 1.39, "learning_rate": 0.00017720527617477326, "loss": 0.0983, "step": 2487 }, { "epoch": 1.39, "learning_rate": 0.0001771434460016488, "loss": 0.1211, "step": 2488 }, { "epoch": 1.4, "learning_rate": 0.00017708161582852432, "loss": 0.0975, "step": 2489 }, { "epoch": 1.4, "learning_rate": 0.0001770197856553998, "loss": 0.0947, "step": 2490 }, { "epoch": 1.4, "learning_rate": 0.00017695795548227532, "loss": 0.0832, "step": 2491 }, { "epoch": 1.4, "learning_rate": 0.00017689612530915084, "loss": 0.0635, "step": 2492 }, { "epoch": 1.4, "learning_rate": 0.00017683429513602638, "loss": 0.0977, "step": 2493 }, { "epoch": 1.4, "learning_rate": 0.0001767724649629019, "loss": 0.0872, "step": 2494 }, { "epoch": 1.4, "learning_rate": 0.0001767106347897774, "loss": 0.1144, "step": 2495 }, { "epoch": 1.4, "learning_rate": 0.0001766488046166529, "loss": 0.0693, "step": 2496 }, { "epoch": 1.4, "learning_rate": 0.00017658697444352842, "loss": 0.0422, "step": 2497 }, { "epoch": 1.4, "learning_rate": 0.00017652514427040394, "loss": 0.0725, "step": 2498 }, { "epoch": 1.4, "learning_rate": 0.00017646331409727948, "loss": 0.0599, "step": 2499 }, { "epoch": 1.4, "learning_rate": 0.00017640148392415497, "loss": 0.0841, "step": 2500 }, { "epoch": 1.4, "learning_rate": 0.0001763396537510305, "loss": 0.0635, "step": 2501 }, { "epoch": 1.4, "learning_rate": 0.000176277823577906, "loss": 0.1011, "step": 2502 }, { "epoch": 1.4, "learning_rate": 0.00017621599340478152, "loss": 0.0749, "step": 2503 }, { "epoch": 1.4, "learning_rate": 0.000176154163231657, "loss": 0.0787, "step": 2504 }, { "epoch": 1.4, "learning_rate": 0.00017609233305853253, "loss": 0.0752, "step": 2505 }, { "epoch": 1.4, "learning_rate": 0.00017603050288540807, "loss": 0.0867, "step": 2506 }, { "epoch": 1.41, "learning_rate": 0.0001759686727122836, "loss": 0.0984, "step": 2507 }, { "epoch": 1.41, "learning_rate": 0.0001759068425391591, "loss": 0.0947, "step": 2508 }, { "epoch": 1.41, "learning_rate": 0.0001758450123660346, "loss": 0.0956, "step": 2509 }, { "epoch": 1.41, "learning_rate": 0.0001757831821929101, "loss": 0.0901, "step": 2510 }, { "epoch": 1.41, "learning_rate": 0.00017572135201978566, "loss": 0.0784, "step": 2511 }, { "epoch": 1.41, "learning_rate": 0.00017565952184666117, "loss": 0.0653, "step": 2512 }, { "epoch": 1.41, "learning_rate": 0.00017559769167353666, "loss": 0.0682, "step": 2513 }, { "epoch": 1.41, "learning_rate": 0.00017553586150041218, "loss": 0.0742, "step": 2514 }, { "epoch": 1.41, "learning_rate": 0.0001754740313272877, "loss": 0.0684, "step": 2515 }, { "epoch": 1.41, "learning_rate": 0.0001754122011541632, "loss": 0.0485, "step": 2516 }, { "epoch": 1.41, "learning_rate": 0.00017535037098103876, "loss": 0.0613, "step": 2517 }, { "epoch": 1.41, "learning_rate": 0.00017528854080791425, "loss": 0.0719, "step": 2518 }, { "epoch": 1.41, "learning_rate": 0.00017522671063478976, "loss": 0.0788, "step": 2519 }, { "epoch": 1.41, "learning_rate": 0.00017516488046166528, "loss": 0.0965, "step": 2520 }, { "epoch": 1.41, "learning_rate": 0.0001751030502885408, "loss": 0.0824, "step": 2521 }, { "epoch": 1.41, "learning_rate": 0.00017504122011541628, "loss": 0.0662, "step": 2522 }, { "epoch": 1.41, "learning_rate": 0.0001749793899422918, "loss": 0.1061, "step": 2523 }, { "epoch": 1.41, "learning_rate": 0.00017491755976916734, "loss": 0.0515, "step": 2524 }, { "epoch": 1.42, "learning_rate": 0.00017485572959604286, "loss": 0.0813, "step": 2525 }, { "epoch": 1.42, "learning_rate": 0.00017479389942291838, "loss": 0.0579, "step": 2526 }, { "epoch": 1.42, "learning_rate": 0.00017473206924979387, "loss": 0.0643, "step": 2527 }, { "epoch": 1.42, "learning_rate": 0.00017467023907666938, "loss": 0.072, "step": 2528 }, { "epoch": 1.42, "learning_rate": 0.00017460840890354493, "loss": 0.0627, "step": 2529 }, { "epoch": 1.42, "learning_rate": 0.00017454657873042044, "loss": 0.1254, "step": 2530 }, { "epoch": 1.42, "learning_rate": 0.00017448474855729596, "loss": 0.1125, "step": 2531 }, { "epoch": 1.42, "learning_rate": 0.00017442291838417145, "loss": 0.1643, "step": 2532 }, { "epoch": 1.42, "learning_rate": 0.00017436108821104697, "loss": 0.1264, "step": 2533 }, { "epoch": 1.42, "learning_rate": 0.00017429925803792248, "loss": 0.2839, "step": 2534 }, { "epoch": 1.42, "learning_rate": 0.00017423742786479803, "loss": 0.0718, "step": 2535 }, { "epoch": 1.42, "learning_rate": 0.00017417559769167352, "loss": 0.0786, "step": 2536 }, { "epoch": 1.42, "learning_rate": 0.00017411376751854903, "loss": 0.1073, "step": 2537 }, { "epoch": 1.42, "learning_rate": 0.00017405193734542455, "loss": 0.063, "step": 2538 }, { "epoch": 1.42, "learning_rate": 0.00017399010717230007, "loss": 0.0679, "step": 2539 }, { "epoch": 1.42, "learning_rate": 0.0001739282769991756, "loss": 0.0652, "step": 2540 }, { "epoch": 1.42, "learning_rate": 0.0001738664468260511, "loss": 0.0583, "step": 2541 }, { "epoch": 1.42, "learning_rate": 0.00017380461665292662, "loss": 0.0835, "step": 2542 }, { "epoch": 1.43, "learning_rate": 0.00017374278647980213, "loss": 0.0925, "step": 2543 }, { "epoch": 1.43, "learning_rate": 0.00017368095630667765, "loss": 0.074, "step": 2544 }, { "epoch": 1.43, "learning_rate": 0.00017361912613355314, "loss": 0.114, "step": 2545 }, { "epoch": 1.43, "learning_rate": 0.00017355729596042866, "loss": 0.069, "step": 2546 }, { "epoch": 1.43, "learning_rate": 0.0001734954657873042, "loss": 0.0812, "step": 2547 }, { "epoch": 1.43, "learning_rate": 0.00017343363561417972, "loss": 0.0765, "step": 2548 }, { "epoch": 1.43, "learning_rate": 0.00017337180544105523, "loss": 0.0906, "step": 2549 }, { "epoch": 1.43, "learning_rate": 0.00017330997526793072, "loss": 0.0528, "step": 2550 }, { "epoch": 1.43, "learning_rate": 0.00017324814509480624, "loss": 0.0668, "step": 2551 }, { "epoch": 1.43, "learning_rate": 0.00017318631492168175, "loss": 0.1026, "step": 2552 }, { "epoch": 1.43, "learning_rate": 0.0001731244847485573, "loss": 0.1057, "step": 2553 }, { "epoch": 1.43, "learning_rate": 0.0001730626545754328, "loss": 0.056, "step": 2554 }, { "epoch": 1.43, "learning_rate": 0.0001730008244023083, "loss": 0.0771, "step": 2555 }, { "epoch": 1.43, "learning_rate": 0.00017293899422918382, "loss": 0.0703, "step": 2556 }, { "epoch": 1.43, "learning_rate": 0.00017287716405605934, "loss": 0.0705, "step": 2557 }, { "epoch": 1.43, "learning_rate": 0.00017281533388293488, "loss": 0.0611, "step": 2558 }, { "epoch": 1.43, "learning_rate": 0.00017275350370981037, "loss": 0.0762, "step": 2559 }, { "epoch": 1.43, "learning_rate": 0.0001726916735366859, "loss": 0.0705, "step": 2560 }, { "epoch": 1.44, "learning_rate": 0.0001726298433635614, "loss": 0.0642, "step": 2561 }, { "epoch": 1.44, "learning_rate": 0.00017256801319043692, "loss": 0.0641, "step": 2562 }, { "epoch": 1.44, "learning_rate": 0.00017250618301731244, "loss": 0.0672, "step": 2563 }, { "epoch": 1.44, "learning_rate": 0.00017244435284418793, "loss": 0.0856, "step": 2564 }, { "epoch": 1.44, "learning_rate": 0.00017238252267106347, "loss": 0.0638, "step": 2565 }, { "epoch": 1.44, "learning_rate": 0.000172320692497939, "loss": 0.0763, "step": 2566 }, { "epoch": 1.44, "learning_rate": 0.0001722588623248145, "loss": 0.0753, "step": 2567 }, { "epoch": 1.44, "learning_rate": 0.00017219703215169, "loss": 0.1087, "step": 2568 }, { "epoch": 1.44, "learning_rate": 0.0001721352019785655, "loss": 0.0948, "step": 2569 }, { "epoch": 1.44, "learning_rate": 0.00017207337180544105, "loss": 0.0863, "step": 2570 }, { "epoch": 1.44, "learning_rate": 0.00017201154163231657, "loss": 0.0856, "step": 2571 }, { "epoch": 1.44, "learning_rate": 0.0001719497114591921, "loss": 0.0486, "step": 2572 }, { "epoch": 1.44, "learning_rate": 0.00017188788128606758, "loss": 0.0797, "step": 2573 }, { "epoch": 1.44, "learning_rate": 0.0001718260511129431, "loss": 0.0915, "step": 2574 }, { "epoch": 1.44, "learning_rate": 0.0001717642209398186, "loss": 0.0431, "step": 2575 }, { "epoch": 1.44, "learning_rate": 0.00017170239076669415, "loss": 0.0598, "step": 2576 }, { "epoch": 1.44, "learning_rate": 0.00017164056059356964, "loss": 0.115, "step": 2577 }, { "epoch": 1.45, "learning_rate": 0.00017157873042044516, "loss": 0.0607, "step": 2578 }, { "epoch": 1.45, "learning_rate": 0.00017151690024732068, "loss": 0.0654, "step": 2579 }, { "epoch": 1.45, "learning_rate": 0.0001714550700741962, "loss": 0.1019, "step": 2580 }, { "epoch": 1.45, "learning_rate": 0.0001713932399010717, "loss": 0.1492, "step": 2581 }, { "epoch": 1.45, "learning_rate": 0.0001713314097279472, "loss": 0.1934, "step": 2582 }, { "epoch": 1.45, "learning_rate": 0.00017126957955482274, "loss": 0.0947, "step": 2583 }, { "epoch": 1.45, "learning_rate": 0.00017120774938169826, "loss": 0.1454, "step": 2584 }, { "epoch": 1.45, "learning_rate": 0.00017114591920857378, "loss": 0.0693, "step": 2585 }, { "epoch": 1.45, "learning_rate": 0.00017108408903544926, "loss": 0.1006, "step": 2586 }, { "epoch": 1.45, "learning_rate": 0.00017102225886232478, "loss": 0.0744, "step": 2587 }, { "epoch": 1.45, "learning_rate": 0.00017096042868920032, "loss": 0.0891, "step": 2588 }, { "epoch": 1.45, "learning_rate": 0.00017089859851607584, "loss": 0.0764, "step": 2589 }, { "epoch": 1.45, "learning_rate": 0.00017083676834295136, "loss": 0.058, "step": 2590 }, { "epoch": 1.45, "learning_rate": 0.00017077493816982685, "loss": 0.074, "step": 2591 }, { "epoch": 1.45, "learning_rate": 0.00017071310799670236, "loss": 0.1166, "step": 2592 }, { "epoch": 1.45, "learning_rate": 0.00017065127782357788, "loss": 0.0782, "step": 2593 }, { "epoch": 1.45, "learning_rate": 0.00017058944765045342, "loss": 0.0776, "step": 2594 }, { "epoch": 1.45, "learning_rate": 0.00017052761747732894, "loss": 0.0915, "step": 2595 }, { "epoch": 1.46, "learning_rate": 0.00017046578730420443, "loss": 0.0619, "step": 2596 }, { "epoch": 1.46, "learning_rate": 0.00017040395713107995, "loss": 0.1016, "step": 2597 }, { "epoch": 1.46, "learning_rate": 0.00017034212695795546, "loss": 0.063, "step": 2598 }, { "epoch": 1.46, "learning_rate": 0.000170280296784831, "loss": 0.066, "step": 2599 }, { "epoch": 1.46, "learning_rate": 0.0001702184666117065, "loss": 0.085, "step": 2600 }, { "epoch": 1.46, "learning_rate": 0.000170156636438582, "loss": 0.0571, "step": 2601 }, { "epoch": 1.46, "learning_rate": 0.00017009480626545753, "loss": 0.0629, "step": 2602 }, { "epoch": 1.46, "learning_rate": 0.00017003297609233305, "loss": 0.1, "step": 2603 }, { "epoch": 1.46, "learning_rate": 0.00016997114591920856, "loss": 0.069, "step": 2604 }, { "epoch": 1.46, "learning_rate": 0.00016990931574608405, "loss": 0.0961, "step": 2605 }, { "epoch": 1.46, "learning_rate": 0.0001698474855729596, "loss": 0.0751, "step": 2606 }, { "epoch": 1.46, "learning_rate": 0.0001697856553998351, "loss": 0.0442, "step": 2607 }, { "epoch": 1.46, "learning_rate": 0.00016972382522671063, "loss": 0.0586, "step": 2608 }, { "epoch": 1.46, "learning_rate": 0.00016966199505358612, "loss": 0.0621, "step": 2609 }, { "epoch": 1.46, "learning_rate": 0.00016960016488046164, "loss": 0.0606, "step": 2610 }, { "epoch": 1.46, "learning_rate": 0.00016953833470733715, "loss": 0.094, "step": 2611 }, { "epoch": 1.46, "learning_rate": 0.0001694765045342127, "loss": 0.0772, "step": 2612 }, { "epoch": 1.46, "learning_rate": 0.0001694146743610882, "loss": 0.0967, "step": 2613 }, { "epoch": 1.47, "learning_rate": 0.0001693528441879637, "loss": 0.0716, "step": 2614 }, { "epoch": 1.47, "learning_rate": 0.00016929101401483922, "loss": 0.0648, "step": 2615 }, { "epoch": 1.47, "learning_rate": 0.00016922918384171474, "loss": 0.045, "step": 2616 }, { "epoch": 1.47, "learning_rate": 0.00016916735366859028, "loss": 0.0772, "step": 2617 }, { "epoch": 1.47, "learning_rate": 0.0001691055234954658, "loss": 0.0534, "step": 2618 }, { "epoch": 1.47, "learning_rate": 0.00016904369332234128, "loss": 0.0567, "step": 2619 }, { "epoch": 1.47, "learning_rate": 0.0001689818631492168, "loss": 0.0815, "step": 2620 }, { "epoch": 1.47, "learning_rate": 0.00016892003297609232, "loss": 0.1043, "step": 2621 }, { "epoch": 1.47, "learning_rate": 0.00016885820280296783, "loss": 0.0594, "step": 2622 }, { "epoch": 1.47, "learning_rate": 0.00016879637262984332, "loss": 0.0679, "step": 2623 }, { "epoch": 1.47, "learning_rate": 0.00016873454245671887, "loss": 0.0809, "step": 2624 }, { "epoch": 1.47, "learning_rate": 0.00016867271228359438, "loss": 0.1027, "step": 2625 }, { "epoch": 1.47, "learning_rate": 0.0001686108821104699, "loss": 0.1137, "step": 2626 }, { "epoch": 1.47, "learning_rate": 0.00016854905193734542, "loss": 0.0843, "step": 2627 }, { "epoch": 1.47, "learning_rate": 0.0001684872217642209, "loss": 0.0884, "step": 2628 }, { "epoch": 1.47, "learning_rate": 0.00016842539159109642, "loss": 0.1521, "step": 2629 }, { "epoch": 1.47, "learning_rate": 0.00016836356141797197, "loss": 0.1066, "step": 2630 }, { "epoch": 1.47, "learning_rate": 0.00016830173124484748, "loss": 0.0897, "step": 2631 }, { "epoch": 1.48, "learning_rate": 0.00016823990107172297, "loss": 0.2221, "step": 2632 }, { "epoch": 1.48, "learning_rate": 0.0001681780708985985, "loss": 0.1751, "step": 2633 }, { "epoch": 1.48, "learning_rate": 0.000168116240725474, "loss": 0.2122, "step": 2634 }, { "epoch": 1.48, "learning_rate": 0.00016805441055234955, "loss": 0.0894, "step": 2635 }, { "epoch": 1.48, "learning_rate": 0.00016799258037922507, "loss": 0.0927, "step": 2636 }, { "epoch": 1.48, "learning_rate": 0.00016793075020610056, "loss": 0.0908, "step": 2637 }, { "epoch": 1.48, "learning_rate": 0.00016786892003297607, "loss": 0.0912, "step": 2638 }, { "epoch": 1.48, "learning_rate": 0.0001678070898598516, "loss": 0.0553, "step": 2639 }, { "epoch": 1.48, "learning_rate": 0.0001677452596867271, "loss": 0.0899, "step": 2640 }, { "epoch": 1.48, "learning_rate": 0.0001676834295136026, "loss": 0.0807, "step": 2641 }, { "epoch": 1.48, "learning_rate": 0.00016762159934047814, "loss": 0.0626, "step": 2642 }, { "epoch": 1.48, "learning_rate": 0.00016755976916735366, "loss": 0.0742, "step": 2643 }, { "epoch": 1.48, "learning_rate": 0.00016749793899422917, "loss": 0.065, "step": 2644 }, { "epoch": 1.48, "learning_rate": 0.0001674361088211047, "loss": 0.1016, "step": 2645 }, { "epoch": 1.48, "learning_rate": 0.00016737427864798018, "loss": 0.0525, "step": 2646 }, { "epoch": 1.48, "learning_rate": 0.00016731244847485572, "loss": 0.0775, "step": 2647 }, { "epoch": 1.48, "learning_rate": 0.00016725061830173124, "loss": 0.0557, "step": 2648 }, { "epoch": 1.48, "learning_rate": 0.00016718878812860676, "loss": 0.0666, "step": 2649 }, { "epoch": 1.49, "learning_rate": 0.00016712695795548227, "loss": 0.0669, "step": 2650 }, { "epoch": 1.49, "learning_rate": 0.00016706512778235776, "loss": 0.0744, "step": 2651 }, { "epoch": 1.49, "learning_rate": 0.00016700329760923328, "loss": 0.0707, "step": 2652 }, { "epoch": 1.49, "learning_rate": 0.00016694146743610882, "loss": 0.0634, "step": 2653 }, { "epoch": 1.49, "learning_rate": 0.00016687963726298434, "loss": 0.0798, "step": 2654 }, { "epoch": 1.49, "learning_rate": 0.00016681780708985983, "loss": 0.0726, "step": 2655 }, { "epoch": 1.49, "learning_rate": 0.00016675597691673534, "loss": 0.0617, "step": 2656 }, { "epoch": 1.49, "learning_rate": 0.00016669414674361086, "loss": 0.0523, "step": 2657 }, { "epoch": 1.49, "learning_rate": 0.00016663231657048638, "loss": 0.0656, "step": 2658 }, { "epoch": 1.49, "learning_rate": 0.00016657048639736192, "loss": 0.0534, "step": 2659 }, { "epoch": 1.49, "learning_rate": 0.0001665086562242374, "loss": 0.0526, "step": 2660 }, { "epoch": 1.49, "learning_rate": 0.00016644682605111293, "loss": 0.087, "step": 2661 }, { "epoch": 1.49, "learning_rate": 0.00016638499587798844, "loss": 0.0588, "step": 2662 }, { "epoch": 1.49, "learning_rate": 0.00016632316570486396, "loss": 0.105, "step": 2663 }, { "epoch": 1.49, "learning_rate": 0.00016626133553173945, "loss": 0.0818, "step": 2664 }, { "epoch": 1.49, "learning_rate": 0.000166199505358615, "loss": 0.0671, "step": 2665 }, { "epoch": 1.49, "learning_rate": 0.0001661376751854905, "loss": 0.1151, "step": 2666 }, { "epoch": 1.49, "learning_rate": 0.00016607584501236603, "loss": 0.0801, "step": 2667 }, { "epoch": 1.5, "learning_rate": 0.00016601401483924154, "loss": 0.0743, "step": 2668 }, { "epoch": 1.5, "learning_rate": 0.00016595218466611703, "loss": 0.0646, "step": 2669 }, { "epoch": 1.5, "learning_rate": 0.00016589035449299255, "loss": 0.0885, "step": 2670 }, { "epoch": 1.5, "learning_rate": 0.0001658285243198681, "loss": 0.0709, "step": 2671 }, { "epoch": 1.5, "learning_rate": 0.0001657666941467436, "loss": 0.0426, "step": 2672 }, { "epoch": 1.5, "learning_rate": 0.0001657048639736191, "loss": 0.1207, "step": 2673 }, { "epoch": 1.5, "learning_rate": 0.00016564303380049462, "loss": 0.1103, "step": 2674 }, { "epoch": 1.5, "learning_rate": 0.00016558120362737013, "loss": 0.0999, "step": 2675 }, { "epoch": 1.5, "learning_rate": 0.00016551937345424568, "loss": 0.1024, "step": 2676 }, { "epoch": 1.5, "learning_rate": 0.0001654575432811212, "loss": 0.1148, "step": 2677 }, { "epoch": 1.5, "learning_rate": 0.00016539571310799668, "loss": 0.1036, "step": 2678 }, { "epoch": 1.5, "learning_rate": 0.0001653338829348722, "loss": 0.1133, "step": 2679 }, { "epoch": 1.5, "learning_rate": 0.00016527205276174772, "loss": 0.0971, "step": 2680 }, { "epoch": 1.5, "learning_rate": 0.00016521022258862323, "loss": 0.0983, "step": 2681 }, { "epoch": 1.5, "learning_rate": 0.00016514839241549878, "loss": 0.0986, "step": 2682 }, { "epoch": 1.5, "learning_rate": 0.00016508656224237427, "loss": 0.1137, "step": 2683 }, { "epoch": 1.5, "learning_rate": 0.00016502473206924978, "loss": 0.1183, "step": 2684 }, { "epoch": 1.51, "learning_rate": 0.0001649629018961253, "loss": 0.0985, "step": 2685 }, { "epoch": 1.51, "learning_rate": 0.00016490107172300081, "loss": 0.0909, "step": 2686 }, { "epoch": 1.51, "learning_rate": 0.0001648392415498763, "loss": 0.0864, "step": 2687 }, { "epoch": 1.51, "learning_rate": 0.00016477741137675182, "loss": 0.0948, "step": 2688 }, { "epoch": 1.51, "learning_rate": 0.00016471558120362736, "loss": 0.0961, "step": 2689 }, { "epoch": 1.51, "learning_rate": 0.00016465375103050288, "loss": 0.0758, "step": 2690 }, { "epoch": 1.51, "learning_rate": 0.0001645919208573784, "loss": 0.0769, "step": 2691 }, { "epoch": 1.51, "learning_rate": 0.0001645300906842539, "loss": 0.0876, "step": 2692 }, { "epoch": 1.51, "learning_rate": 0.0001644682605111294, "loss": 0.0482, "step": 2693 }, { "epoch": 1.51, "learning_rate": 0.00016440643033800495, "loss": 0.0718, "step": 2694 }, { "epoch": 1.51, "learning_rate": 0.00016434460016488046, "loss": 0.0638, "step": 2695 }, { "epoch": 1.51, "learning_rate": 0.00016428276999175595, "loss": 0.0974, "step": 2696 }, { "epoch": 1.51, "learning_rate": 0.00016422093981863147, "loss": 0.0738, "step": 2697 }, { "epoch": 1.51, "learning_rate": 0.000164159109645507, "loss": 0.1013, "step": 2698 }, { "epoch": 1.51, "learning_rate": 0.0001640972794723825, "loss": 0.0778, "step": 2699 }, { "epoch": 1.51, "learning_rate": 0.00016403544929925805, "loss": 0.0706, "step": 2700 }, { "epoch": 1.51, "eval_loss": 0.10914234071969986, "eval_runtime": 271.2482, "eval_samples_per_second": 9.74, "eval_steps_per_second": 0.612, "eval_wer": 0.09051963506545022, "step": 2700 }, { "epoch": 1.51, "learning_rate": 0.00016397361912613354, "loss": 0.0915, "step": 2701 }, { "epoch": 1.51, "learning_rate": 0.00016391178895300905, "loss": 0.0553, "step": 2702 }, { "epoch": 1.52, "learning_rate": 0.00016384995877988457, "loss": 0.0919, "step": 2703 }, { "epoch": 1.52, "learning_rate": 0.00016378812860676009, "loss": 0.0752, "step": 2704 }, { "epoch": 1.52, "learning_rate": 0.00016372629843363558, "loss": 0.0676, "step": 2705 }, { "epoch": 1.52, "learning_rate": 0.0001636644682605111, "loss": 0.08, "step": 2706 }, { "epoch": 1.52, "learning_rate": 0.00016360263808738664, "loss": 0.072, "step": 2707 }, { "epoch": 1.52, "learning_rate": 0.00016354080791426215, "loss": 0.0707, "step": 2708 }, { "epoch": 1.52, "learning_rate": 0.00016347897774113767, "loss": 0.0898, "step": 2709 }, { "epoch": 1.52, "learning_rate": 0.00016341714756801316, "loss": 0.0617, "step": 2710 }, { "epoch": 1.52, "learning_rate": 0.00016335531739488868, "loss": 0.0604, "step": 2711 }, { "epoch": 1.52, "learning_rate": 0.00016329348722176422, "loss": 0.0827, "step": 2712 }, { "epoch": 1.52, "learning_rate": 0.00016323165704863974, "loss": 0.075, "step": 2713 }, { "epoch": 1.52, "learning_rate": 0.00016316982687551525, "loss": 0.0984, "step": 2714 }, { "epoch": 1.52, "learning_rate": 0.00016310799670239074, "loss": 0.0544, "step": 2715 }, { "epoch": 1.52, "learning_rate": 0.00016304616652926626, "loss": 0.0629, "step": 2716 }, { "epoch": 1.52, "learning_rate": 0.00016298433635614177, "loss": 0.0747, "step": 2717 }, { "epoch": 1.52, "learning_rate": 0.00016292250618301732, "loss": 0.0658, "step": 2718 }, { "epoch": 1.52, "learning_rate": 0.0001628606760098928, "loss": 0.0861, "step": 2719 }, { "epoch": 1.52, "learning_rate": 0.00016279884583676832, "loss": 0.0736, "step": 2720 }, { "epoch": 1.53, "learning_rate": 0.00016273701566364384, "loss": 0.091, "step": 2721 }, { "epoch": 1.53, "learning_rate": 0.00016267518549051936, "loss": 0.104, "step": 2722 }, { "epoch": 1.53, "learning_rate": 0.0001626133553173949, "loss": 0.0935, "step": 2723 }, { "epoch": 1.53, "learning_rate": 0.0001625515251442704, "loss": 0.097, "step": 2724 }, { "epoch": 1.53, "learning_rate": 0.0001624896949711459, "loss": 0.0874, "step": 2725 }, { "epoch": 1.53, "learning_rate": 0.00016242786479802142, "loss": 0.0957, "step": 2726 }, { "epoch": 1.53, "learning_rate": 0.00016236603462489694, "loss": 0.0887, "step": 2727 }, { "epoch": 1.53, "learning_rate": 0.00016230420445177243, "loss": 0.0904, "step": 2728 }, { "epoch": 1.53, "learning_rate": 0.00016224237427864795, "loss": 0.0869, "step": 2729 }, { "epoch": 1.53, "learning_rate": 0.0001621805441055235, "loss": 0.1293, "step": 2730 }, { "epoch": 1.53, "learning_rate": 0.000162118713932399, "loss": 0.1331, "step": 2731 }, { "epoch": 1.53, "learning_rate": 0.00016205688375927452, "loss": 0.0771, "step": 2732 }, { "epoch": 1.53, "learning_rate": 0.00016199505358615, "loss": 0.1944, "step": 2733 }, { "epoch": 1.53, "learning_rate": 0.00016193322341302553, "loss": 0.1468, "step": 2734 }, { "epoch": 1.53, "learning_rate": 0.00016187139323990105, "loss": 0.1146, "step": 2735 }, { "epoch": 1.53, "learning_rate": 0.0001618095630667766, "loss": 0.0766, "step": 2736 }, { "epoch": 1.53, "learning_rate": 0.0001617477328936521, "loss": 0.0893, "step": 2737 }, { "epoch": 1.53, "learning_rate": 0.0001616859027205276, "loss": 0.0561, "step": 2738 }, { "epoch": 1.54, "learning_rate": 0.0001616240725474031, "loss": 0.0754, "step": 2739 }, { "epoch": 1.54, "learning_rate": 0.00016156224237427863, "loss": 0.0735, "step": 2740 }, { "epoch": 1.54, "learning_rate": 0.00016150041220115417, "loss": 0.0676, "step": 2741 }, { "epoch": 1.54, "learning_rate": 0.00016143858202802966, "loss": 0.0803, "step": 2742 }, { "epoch": 1.54, "learning_rate": 0.00016137675185490518, "loss": 0.0624, "step": 2743 }, { "epoch": 1.54, "learning_rate": 0.0001613149216817807, "loss": 0.1083, "step": 2744 }, { "epoch": 1.54, "learning_rate": 0.0001612530915086562, "loss": 0.0595, "step": 2745 }, { "epoch": 1.54, "learning_rate": 0.00016119126133553173, "loss": 0.0781, "step": 2746 }, { "epoch": 1.54, "learning_rate": 0.00016112943116240722, "loss": 0.0647, "step": 2747 }, { "epoch": 1.54, "learning_rate": 0.00016106760098928276, "loss": 0.0725, "step": 2748 }, { "epoch": 1.54, "learning_rate": 0.00016100577081615828, "loss": 0.1269, "step": 2749 }, { "epoch": 1.54, "learning_rate": 0.0001609439406430338, "loss": 0.1106, "step": 2750 }, { "epoch": 1.54, "learning_rate": 0.00016088211046990928, "loss": 0.081, "step": 2751 }, { "epoch": 1.54, "learning_rate": 0.0001608202802967848, "loss": 0.0702, "step": 2752 }, { "epoch": 1.54, "learning_rate": 0.00016075845012366034, "loss": 0.053, "step": 2753 }, { "epoch": 1.54, "learning_rate": 0.00016069661995053586, "loss": 0.0699, "step": 2754 }, { "epoch": 1.54, "learning_rate": 0.00016063478977741138, "loss": 0.082, "step": 2755 }, { "epoch": 1.54, "learning_rate": 0.00016057295960428687, "loss": 0.0622, "step": 2756 }, { "epoch": 1.55, "learning_rate": 0.00016051112943116238, "loss": 0.0771, "step": 2757 }, { "epoch": 1.55, "learning_rate": 0.0001604492992580379, "loss": 0.1261, "step": 2758 }, { "epoch": 1.55, "learning_rate": 0.00016038746908491344, "loss": 0.0724, "step": 2759 }, { "epoch": 1.55, "learning_rate": 0.00016032563891178893, "loss": 0.0604, "step": 2760 }, { "epoch": 1.55, "learning_rate": 0.00016026380873866445, "loss": 0.0763, "step": 2761 }, { "epoch": 1.55, "learning_rate": 0.00016020197856553997, "loss": 0.0576, "step": 2762 }, { "epoch": 1.55, "learning_rate": 0.00016014014839241548, "loss": 0.0799, "step": 2763 }, { "epoch": 1.55, "learning_rate": 0.000160078318219291, "loss": 0.0637, "step": 2764 }, { "epoch": 1.55, "learning_rate": 0.0001600164880461665, "loss": 0.0713, "step": 2765 }, { "epoch": 1.55, "learning_rate": 0.00015995465787304203, "loss": 0.052, "step": 2766 }, { "epoch": 1.55, "learning_rate": 0.00015989282769991755, "loss": 0.079, "step": 2767 }, { "epoch": 1.55, "learning_rate": 0.00015983099752679307, "loss": 0.0634, "step": 2768 }, { "epoch": 1.55, "learning_rate": 0.00015976916735366858, "loss": 0.0371, "step": 2769 }, { "epoch": 1.55, "learning_rate": 0.00015970733718054407, "loss": 0.0743, "step": 2770 }, { "epoch": 1.55, "learning_rate": 0.00015964550700741962, "loss": 0.0526, "step": 2771 }, { "epoch": 1.55, "learning_rate": 0.00015958367683429513, "loss": 0.0739, "step": 2772 }, { "epoch": 1.55, "learning_rate": 0.00015952184666117065, "loss": 0.0845, "step": 2773 }, { "epoch": 1.55, "learning_rate": 0.00015946001648804614, "loss": 0.0699, "step": 2774 }, { "epoch": 1.56, "learning_rate": 0.00015939818631492166, "loss": 0.0499, "step": 2775 }, { "epoch": 1.56, "learning_rate": 0.00015933635614179717, "loss": 0.1197, "step": 2776 }, { "epoch": 1.56, "learning_rate": 0.00015927452596867272, "loss": 0.0683, "step": 2777 }, { "epoch": 1.56, "learning_rate": 0.00015921269579554823, "loss": 0.1114, "step": 2778 }, { "epoch": 1.56, "learning_rate": 0.00015915086562242372, "loss": 0.0529, "step": 2779 }, { "epoch": 1.56, "learning_rate": 0.00015908903544929924, "loss": 0.0554, "step": 2780 }, { "epoch": 1.56, "learning_rate": 0.00015902720527617475, "loss": 0.0875, "step": 2781 }, { "epoch": 1.56, "learning_rate": 0.0001589653751030503, "loss": 0.1094, "step": 2782 }, { "epoch": 1.56, "learning_rate": 0.00015890354492992576, "loss": 0.1606, "step": 2783 }, { "epoch": 1.56, "learning_rate": 0.0001588417147568013, "loss": 0.1567, "step": 2784 }, { "epoch": 1.56, "learning_rate": 0.00015877988458367682, "loss": 0.1163, "step": 2785 }, { "epoch": 1.56, "learning_rate": 0.00015871805441055234, "loss": 0.0727, "step": 2786 }, { "epoch": 1.56, "learning_rate": 0.00015865622423742785, "loss": 0.0907, "step": 2787 }, { "epoch": 1.56, "learning_rate": 0.00015859439406430334, "loss": 0.0728, "step": 2788 }, { "epoch": 1.56, "learning_rate": 0.0001585325638911789, "loss": 0.0742, "step": 2789 }, { "epoch": 1.56, "learning_rate": 0.0001584707337180544, "loss": 0.0791, "step": 2790 }, { "epoch": 1.56, "learning_rate": 0.00015840890354492992, "loss": 0.082, "step": 2791 }, { "epoch": 1.57, "learning_rate": 0.0001583470733718054, "loss": 0.0868, "step": 2792 }, { "epoch": 1.57, "learning_rate": 0.00015828524319868093, "loss": 0.081, "step": 2793 }, { "epoch": 1.57, "learning_rate": 0.00015822341302555644, "loss": 0.0539, "step": 2794 }, { "epoch": 1.57, "learning_rate": 0.000158161582852432, "loss": 0.0698, "step": 2795 }, { "epoch": 1.57, "learning_rate": 0.0001580997526793075, "loss": 0.0676, "step": 2796 }, { "epoch": 1.57, "learning_rate": 0.000158037922506183, "loss": 0.0487, "step": 2797 }, { "epoch": 1.57, "learning_rate": 0.0001579760923330585, "loss": 0.0726, "step": 2798 }, { "epoch": 1.57, "learning_rate": 0.00015791426215993403, "loss": 0.1204, "step": 2799 }, { "epoch": 1.57, "learning_rate": 0.00015785243198680957, "loss": 0.0742, "step": 2800 }, { "epoch": 1.57, "learning_rate": 0.0001577906018136851, "loss": 0.121, "step": 2801 }, { "epoch": 1.57, "learning_rate": 0.00015772877164056058, "loss": 0.0873, "step": 2802 }, { "epoch": 1.57, "learning_rate": 0.0001576669414674361, "loss": 0.038, "step": 2803 }, { "epoch": 1.57, "learning_rate": 0.0001576051112943116, "loss": 0.0898, "step": 2804 }, { "epoch": 1.57, "learning_rate": 0.00015754328112118713, "loss": 0.0611, "step": 2805 }, { "epoch": 1.57, "learning_rate": 0.00015748145094806262, "loss": 0.076, "step": 2806 }, { "epoch": 1.57, "learning_rate": 0.00015741962077493816, "loss": 0.078, "step": 2807 }, { "epoch": 1.57, "learning_rate": 0.00015735779060181368, "loss": 0.0798, "step": 2808 }, { "epoch": 1.57, "learning_rate": 0.0001572959604286892, "loss": 0.0675, "step": 2809 }, { "epoch": 1.58, "learning_rate": 0.0001572341302555647, "loss": 0.0525, "step": 2810 }, { "epoch": 1.58, "learning_rate": 0.0001571723000824402, "loss": 0.0704, "step": 2811 }, { "epoch": 1.58, "learning_rate": 0.00015711046990931571, "loss": 0.0927, "step": 2812 }, { "epoch": 1.58, "learning_rate": 0.00015704863973619126, "loss": 0.0853, "step": 2813 }, { "epoch": 1.58, "learning_rate": 0.00015698680956306678, "loss": 0.0917, "step": 2814 }, { "epoch": 1.58, "learning_rate": 0.00015692497938994226, "loss": 0.0582, "step": 2815 }, { "epoch": 1.58, "learning_rate": 0.00015686314921681778, "loss": 0.0524, "step": 2816 }, { "epoch": 1.58, "learning_rate": 0.0001568013190436933, "loss": 0.0522, "step": 2817 }, { "epoch": 1.58, "learning_rate": 0.00015673948887056884, "loss": 0.0477, "step": 2818 }, { "epoch": 1.58, "learning_rate": 0.00015667765869744436, "loss": 0.0673, "step": 2819 }, { "epoch": 1.58, "learning_rate": 0.00015661582852431985, "loss": 0.0698, "step": 2820 }, { "epoch": 1.58, "learning_rate": 0.00015655399835119536, "loss": 0.0891, "step": 2821 }, { "epoch": 1.58, "learning_rate": 0.00015649216817807088, "loss": 0.0693, "step": 2822 }, { "epoch": 1.58, "learning_rate": 0.0001564303380049464, "loss": 0.0628, "step": 2823 }, { "epoch": 1.58, "learning_rate": 0.00015636850783182194, "loss": 0.0783, "step": 2824 }, { "epoch": 1.58, "learning_rate": 0.00015630667765869743, "loss": 0.1033, "step": 2825 }, { "epoch": 1.58, "learning_rate": 0.00015624484748557295, "loss": 0.0557, "step": 2826 }, { "epoch": 1.58, "learning_rate": 0.00015618301731244846, "loss": 0.1003, "step": 2827 }, { "epoch": 1.59, "learning_rate": 0.00015612118713932398, "loss": 0.1266, "step": 2828 }, { "epoch": 1.59, "learning_rate": 0.00015605935696619947, "loss": 0.0789, "step": 2829 }, { "epoch": 1.59, "learning_rate": 0.000155997526793075, "loss": 0.1073, "step": 2830 }, { "epoch": 1.59, "learning_rate": 0.00015593569661995053, "loss": 0.1032, "step": 2831 }, { "epoch": 1.59, "learning_rate": 0.00015587386644682605, "loss": 0.1153, "step": 2832 }, { "epoch": 1.59, "learning_rate": 0.00015581203627370156, "loss": 0.1515, "step": 2833 }, { "epoch": 1.59, "learning_rate": 0.00015575020610057705, "loss": 0.1632, "step": 2834 }, { "epoch": 1.59, "learning_rate": 0.00015568837592745257, "loss": 0.0971, "step": 2835 }, { "epoch": 1.59, "learning_rate": 0.0001556265457543281, "loss": 0.0937, "step": 2836 }, { "epoch": 1.59, "learning_rate": 0.00015556471558120363, "loss": 0.0618, "step": 2837 }, { "epoch": 1.59, "learning_rate": 0.00015550288540807912, "loss": 0.1074, "step": 2838 }, { "epoch": 1.59, "learning_rate": 0.00015544105523495464, "loss": 0.0856, "step": 2839 }, { "epoch": 1.59, "learning_rate": 0.00015537922506183015, "loss": 0.1062, "step": 2840 }, { "epoch": 1.59, "learning_rate": 0.00015531739488870567, "loss": 0.1461, "step": 2841 }, { "epoch": 1.59, "learning_rate": 0.0001552555647155812, "loss": 0.0955, "step": 2842 }, { "epoch": 1.59, "learning_rate": 0.0001551937345424567, "loss": 0.0938, "step": 2843 }, { "epoch": 1.59, "learning_rate": 0.00015513190436933222, "loss": 0.0921, "step": 2844 }, { "epoch": 1.59, "learning_rate": 0.00015507007419620774, "loss": 0.0669, "step": 2845 }, { "epoch": 1.6, "learning_rate": 0.00015500824402308325, "loss": 0.0995, "step": 2846 }, { "epoch": 1.6, "learning_rate": 0.00015494641384995874, "loss": 0.0747, "step": 2847 }, { "epoch": 1.6, "learning_rate": 0.00015488458367683428, "loss": 0.0798, "step": 2848 }, { "epoch": 1.6, "learning_rate": 0.0001548227535037098, "loss": 0.071, "step": 2849 }, { "epoch": 1.6, "learning_rate": 0.00015476092333058532, "loss": 0.1107, "step": 2850 }, { "epoch": 1.6, "learning_rate": 0.00015469909315746083, "loss": 0.0878, "step": 2851 }, { "epoch": 1.6, "learning_rate": 0.00015463726298433632, "loss": 0.0602, "step": 2852 }, { "epoch": 1.6, "learning_rate": 0.00015457543281121184, "loss": 0.0991, "step": 2853 }, { "epoch": 1.6, "learning_rate": 0.00015451360263808738, "loss": 0.0518, "step": 2854 }, { "epoch": 1.6, "learning_rate": 0.0001544517724649629, "loss": 0.0751, "step": 2855 }, { "epoch": 1.6, "learning_rate": 0.00015438994229183842, "loss": 0.0804, "step": 2856 }, { "epoch": 1.6, "learning_rate": 0.0001543281121187139, "loss": 0.0534, "step": 2857 }, { "epoch": 1.6, "learning_rate": 0.00015426628194558942, "loss": 0.0676, "step": 2858 }, { "epoch": 1.6, "learning_rate": 0.00015420445177246497, "loss": 0.0502, "step": 2859 }, { "epoch": 1.6, "learning_rate": 0.00015414262159934048, "loss": 0.0989, "step": 2860 }, { "epoch": 1.6, "learning_rate": 0.00015408079142621597, "loss": 0.0568, "step": 2861 }, { "epoch": 1.6, "learning_rate": 0.0001540189612530915, "loss": 0.0642, "step": 2862 }, { "epoch": 1.6, "learning_rate": 0.000153957131079967, "loss": 0.0879, "step": 2863 }, { "epoch": 1.61, "learning_rate": 0.00015389530090684252, "loss": 0.0601, "step": 2864 }, { "epoch": 1.61, "learning_rate": 0.00015383347073371807, "loss": 0.0477, "step": 2865 }, { "epoch": 1.61, "learning_rate": 0.00015377164056059356, "loss": 0.089, "step": 2866 }, { "epoch": 1.61, "learning_rate": 0.00015370981038746907, "loss": 0.0893, "step": 2867 }, { "epoch": 1.61, "learning_rate": 0.0001536479802143446, "loss": 0.0753, "step": 2868 }, { "epoch": 1.61, "learning_rate": 0.0001535861500412201, "loss": 0.0978, "step": 2869 }, { "epoch": 1.61, "learning_rate": 0.0001535243198680956, "loss": 0.0862, "step": 2870 }, { "epoch": 1.61, "learning_rate": 0.0001534624896949711, "loss": 0.076, "step": 2871 }, { "epoch": 1.61, "learning_rate": 0.00015340065952184666, "loss": 0.0508, "step": 2872 }, { "epoch": 1.61, "learning_rate": 0.00015333882934872217, "loss": 0.0519, "step": 2873 }, { "epoch": 1.61, "learning_rate": 0.0001532769991755977, "loss": 0.0597, "step": 2874 }, { "epoch": 1.61, "learning_rate": 0.00015321516900247318, "loss": 0.0925, "step": 2875 }, { "epoch": 1.61, "learning_rate": 0.0001531533388293487, "loss": 0.0866, "step": 2876 }, { "epoch": 1.61, "learning_rate": 0.00015309150865622424, "loss": 0.0715, "step": 2877 }, { "epoch": 1.61, "learning_rate": 0.00015302967848309976, "loss": 0.0979, "step": 2878 }, { "epoch": 1.61, "learning_rate": 0.00015296784830997524, "loss": 0.0587, "step": 2879 }, { "epoch": 1.61, "learning_rate": 0.00015290601813685076, "loss": 0.0835, "step": 2880 }, { "epoch": 1.61, "learning_rate": 0.00015284418796372628, "loss": 0.1398, "step": 2881 }, { "epoch": 1.62, "learning_rate": 0.0001527823577906018, "loss": 0.1971, "step": 2882 }, { "epoch": 1.62, "learning_rate": 0.00015272052761747734, "loss": 0.1953, "step": 2883 }, { "epoch": 1.62, "learning_rate": 0.00015265869744435283, "loss": 0.3312, "step": 2884 }, { "epoch": 1.62, "learning_rate": 0.00015259686727122834, "loss": 0.0801, "step": 2885 }, { "epoch": 1.62, "learning_rate": 0.00015253503709810386, "loss": 0.0619, "step": 2886 }, { "epoch": 1.62, "learning_rate": 0.00015247320692497938, "loss": 0.0732, "step": 2887 }, { "epoch": 1.62, "learning_rate": 0.00015241137675185492, "loss": 0.0615, "step": 2888 }, { "epoch": 1.62, "learning_rate": 0.00015234954657873038, "loss": 0.0925, "step": 2889 }, { "epoch": 1.62, "learning_rate": 0.00015228771640560593, "loss": 0.0619, "step": 2890 }, { "epoch": 1.62, "learning_rate": 0.00015222588623248144, "loss": 0.0576, "step": 2891 }, { "epoch": 1.62, "learning_rate": 0.00015216405605935696, "loss": 0.072, "step": 2892 }, { "epoch": 1.62, "learning_rate": 0.00015210222588623245, "loss": 0.0621, "step": 2893 }, { "epoch": 1.62, "learning_rate": 0.00015204039571310797, "loss": 0.0656, "step": 2894 }, { "epoch": 1.62, "learning_rate": 0.0001519785655399835, "loss": 0.0785, "step": 2895 }, { "epoch": 1.62, "learning_rate": 0.00015191673536685903, "loss": 0.0762, "step": 2896 }, { "epoch": 1.62, "learning_rate": 0.00015185490519373454, "loss": 0.127, "step": 2897 }, { "epoch": 1.62, "learning_rate": 0.00015179307502061003, "loss": 0.0922, "step": 2898 }, { "epoch": 1.62, "learning_rate": 0.00015173124484748555, "loss": 0.0604, "step": 2899 }, { "epoch": 1.63, "learning_rate": 0.00015166941467436107, "loss": 0.0853, "step": 2900 }, { "epoch": 1.63, "learning_rate": 0.0001516075845012366, "loss": 0.0793, "step": 2901 }, { "epoch": 1.63, "learning_rate": 0.0001515457543281121, "loss": 0.0646, "step": 2902 }, { "epoch": 1.63, "learning_rate": 0.00015148392415498762, "loss": 0.0749, "step": 2903 }, { "epoch": 1.63, "learning_rate": 0.00015142209398186313, "loss": 0.0375, "step": 2904 }, { "epoch": 1.63, "learning_rate": 0.00015136026380873865, "loss": 0.0497, "step": 2905 }, { "epoch": 1.63, "learning_rate": 0.0001512984336356142, "loss": 0.0674, "step": 2906 }, { "epoch": 1.63, "learning_rate": 0.00015123660346248968, "loss": 0.0881, "step": 2907 }, { "epoch": 1.63, "learning_rate": 0.0001511747732893652, "loss": 0.0817, "step": 2908 }, { "epoch": 1.63, "learning_rate": 0.00015111294311624072, "loss": 0.0709, "step": 2909 }, { "epoch": 1.63, "learning_rate": 0.00015105111294311623, "loss": 0.0833, "step": 2910 }, { "epoch": 1.63, "learning_rate": 0.00015098928276999172, "loss": 0.076, "step": 2911 }, { "epoch": 1.63, "learning_rate": 0.00015092745259686724, "loss": 0.0586, "step": 2912 }, { "epoch": 1.63, "learning_rate": 0.00015086562242374278, "loss": 0.1082, "step": 2913 }, { "epoch": 1.63, "learning_rate": 0.0001508037922506183, "loss": 0.062, "step": 2914 }, { "epoch": 1.63, "learning_rate": 0.00015074196207749381, "loss": 0.0687, "step": 2915 }, { "epoch": 1.63, "learning_rate": 0.0001506801319043693, "loss": 0.0667, "step": 2916 }, { "epoch": 1.64, "learning_rate": 0.00015061830173124482, "loss": 0.0441, "step": 2917 }, { "epoch": 1.64, "learning_rate": 0.00015055647155812034, "loss": 0.0874, "step": 2918 }, { "epoch": 1.64, "learning_rate": 0.00015049464138499588, "loss": 0.0688, "step": 2919 }, { "epoch": 1.64, "learning_rate": 0.0001504328112118714, "loss": 0.0645, "step": 2920 }, { "epoch": 1.64, "learning_rate": 0.0001503709810387469, "loss": 0.0591, "step": 2921 }, { "epoch": 1.64, "learning_rate": 0.0001503091508656224, "loss": 0.0678, "step": 2922 }, { "epoch": 1.64, "learning_rate": 0.00015024732069249792, "loss": 0.05, "step": 2923 }, { "epoch": 1.64, "learning_rate": 0.00015018549051937346, "loss": 0.0674, "step": 2924 }, { "epoch": 1.64, "learning_rate": 0.00015012366034624895, "loss": 0.0635, "step": 2925 }, { "epoch": 1.64, "learning_rate": 0.00015006183017312447, "loss": 0.1038, "step": 2926 }, { "epoch": 1.64, "learning_rate": 0.00015, "loss": 0.0736, "step": 2927 }, { "epoch": 1.64, "learning_rate": 0.0001499381698268755, "loss": 0.124, "step": 2928 }, { "epoch": 1.64, "learning_rate": 0.00014987633965375102, "loss": 0.0928, "step": 2929 }, { "epoch": 1.64, "learning_rate": 0.00014981450948062654, "loss": 0.1274, "step": 2930 }, { "epoch": 1.64, "learning_rate": 0.00014975267930750205, "loss": 0.0382, "step": 2931 }, { "epoch": 1.64, "learning_rate": 0.00014969084913437757, "loss": 0.2008, "step": 2932 }, { "epoch": 1.64, "learning_rate": 0.00014962901896125306, "loss": 0.2981, "step": 2933 }, { "epoch": 1.64, "learning_rate": 0.0001495671887881286, "loss": 0.1432, "step": 2934 }, { "epoch": 1.65, "learning_rate": 0.00014950535861500412, "loss": 0.0695, "step": 2935 }, { "epoch": 1.65, "learning_rate": 0.00014944352844187964, "loss": 0.0787, "step": 2936 }, { "epoch": 1.65, "learning_rate": 0.00014938169826875515, "loss": 0.0674, "step": 2937 }, { "epoch": 1.65, "learning_rate": 0.00014931986809563064, "loss": 0.0704, "step": 2938 }, { "epoch": 1.65, "learning_rate": 0.00014925803792250619, "loss": 0.0729, "step": 2939 }, { "epoch": 1.65, "learning_rate": 0.00014919620774938168, "loss": 0.0585, "step": 2940 }, { "epoch": 1.65, "learning_rate": 0.0001491343775762572, "loss": 0.0974, "step": 2941 }, { "epoch": 1.65, "learning_rate": 0.0001490725474031327, "loss": 0.0648, "step": 2942 }, { "epoch": 1.65, "learning_rate": 0.00014901071723000823, "loss": 0.0606, "step": 2943 }, { "epoch": 1.65, "learning_rate": 0.00014894888705688374, "loss": 0.0608, "step": 2944 }, { "epoch": 1.65, "learning_rate": 0.00014888705688375926, "loss": 0.1118, "step": 2945 }, { "epoch": 1.65, "learning_rate": 0.00014882522671063477, "loss": 0.0689, "step": 2946 }, { "epoch": 1.65, "learning_rate": 0.0001487633965375103, "loss": 0.0771, "step": 2947 }, { "epoch": 1.65, "learning_rate": 0.0001487015663643858, "loss": 0.0714, "step": 2948 }, { "epoch": 1.65, "learning_rate": 0.00014863973619126132, "loss": 0.0391, "step": 2949 }, { "epoch": 1.65, "learning_rate": 0.00014857790601813684, "loss": 0.0749, "step": 2950 }, { "epoch": 1.65, "learning_rate": 0.00014851607584501236, "loss": 0.0532, "step": 2951 }, { "epoch": 1.65, "learning_rate": 0.00014845424567188787, "loss": 0.0928, "step": 2952 }, { "epoch": 1.66, "learning_rate": 0.0001483924154987634, "loss": 0.0522, "step": 2953 }, { "epoch": 1.66, "learning_rate": 0.0001483305853256389, "loss": 0.0832, "step": 2954 }, { "epoch": 1.66, "learning_rate": 0.00014826875515251442, "loss": 0.0779, "step": 2955 }, { "epoch": 1.66, "learning_rate": 0.00014820692497938991, "loss": 0.0769, "step": 2956 }, { "epoch": 1.66, "learning_rate": 0.00014814509480626546, "loss": 0.0455, "step": 2957 }, { "epoch": 1.66, "learning_rate": 0.00014808326463314095, "loss": 0.1004, "step": 2958 }, { "epoch": 1.66, "learning_rate": 0.00014802143446001646, "loss": 0.0802, "step": 2959 }, { "epoch": 1.66, "learning_rate": 0.000147959604286892, "loss": 0.0875, "step": 2960 }, { "epoch": 1.66, "learning_rate": 0.0001478977741137675, "loss": 0.0801, "step": 2961 }, { "epoch": 1.66, "learning_rate": 0.000147835943940643, "loss": 0.0454, "step": 2962 }, { "epoch": 1.66, "learning_rate": 0.00014777411376751853, "loss": 0.1039, "step": 2963 }, { "epoch": 1.66, "learning_rate": 0.00014771228359439405, "loss": 0.0877, "step": 2964 }, { "epoch": 1.66, "learning_rate": 0.00014765045342126956, "loss": 0.073, "step": 2965 }, { "epoch": 1.66, "learning_rate": 0.00014758862324814508, "loss": 0.056, "step": 2966 }, { "epoch": 1.66, "learning_rate": 0.0001475267930750206, "loss": 0.0951, "step": 2967 }, { "epoch": 1.66, "learning_rate": 0.0001474649629018961, "loss": 0.0437, "step": 2968 }, { "epoch": 1.66, "learning_rate": 0.00014740313272877163, "loss": 0.0471, "step": 2969 }, { "epoch": 1.66, "learning_rate": 0.00014734130255564715, "loss": 0.0517, "step": 2970 }, { "epoch": 1.67, "learning_rate": 0.00014727947238252266, "loss": 0.0683, "step": 2971 }, { "epoch": 1.67, "learning_rate": 0.00014721764220939818, "loss": 0.0425, "step": 2972 }, { "epoch": 1.67, "learning_rate": 0.0001471558120362737, "loss": 0.0592, "step": 2973 }, { "epoch": 1.67, "learning_rate": 0.00014709398186314919, "loss": 0.0636, "step": 2974 }, { "epoch": 1.67, "learning_rate": 0.00014703215169002473, "loss": 0.0863, "step": 2975 }, { "epoch": 1.67, "learning_rate": 0.00014697032151690025, "loss": 0.0495, "step": 2976 }, { "epoch": 1.67, "learning_rate": 0.00014690849134377573, "loss": 0.1078, "step": 2977 }, { "epoch": 1.67, "learning_rate": 0.00014684666117065128, "loss": 0.0744, "step": 2978 }, { "epoch": 1.67, "learning_rate": 0.00014678483099752677, "loss": 0.105, "step": 2979 }, { "epoch": 1.67, "learning_rate": 0.00014672300082440228, "loss": 0.0914, "step": 2980 }, { "epoch": 1.67, "learning_rate": 0.0001466611706512778, "loss": 0.0363, "step": 2981 }, { "epoch": 1.67, "learning_rate": 0.00014659934047815332, "loss": 0.0964, "step": 2982 }, { "epoch": 1.67, "learning_rate": 0.00014653751030502886, "loss": 0.1903, "step": 2983 }, { "epoch": 1.67, "learning_rate": 0.00014647568013190435, "loss": 0.2177, "step": 2984 }, { "epoch": 1.67, "learning_rate": 0.00014641384995877987, "loss": 0.0957, "step": 2985 }, { "epoch": 1.67, "learning_rate": 0.00014635201978565538, "loss": 0.0635, "step": 2986 }, { "epoch": 1.67, "learning_rate": 0.0001462901896125309, "loss": 0.0492, "step": 2987 }, { "epoch": 1.67, "learning_rate": 0.00014622835943940642, "loss": 0.0607, "step": 2988 }, { "epoch": 1.68, "learning_rate": 0.00014616652926628193, "loss": 0.0851, "step": 2989 }, { "epoch": 1.68, "learning_rate": 0.00014610469909315745, "loss": 0.0814, "step": 2990 }, { "epoch": 1.68, "learning_rate": 0.00014604286892003297, "loss": 0.0649, "step": 2991 }, { "epoch": 1.68, "learning_rate": 0.00014598103874690848, "loss": 0.0545, "step": 2992 }, { "epoch": 1.68, "learning_rate": 0.000145919208573784, "loss": 0.0845, "step": 2993 }, { "epoch": 1.68, "learning_rate": 0.00014585737840065952, "loss": 0.0946, "step": 2994 }, { "epoch": 1.68, "learning_rate": 0.000145795548227535, "loss": 0.0764, "step": 2995 }, { "epoch": 1.68, "learning_rate": 0.00014573371805441055, "loss": 0.0536, "step": 2996 }, { "epoch": 1.68, "learning_rate": 0.00014567188788128604, "loss": 0.0795, "step": 2997 }, { "epoch": 1.68, "learning_rate": 0.00014561005770816158, "loss": 0.0554, "step": 2998 }, { "epoch": 1.68, "learning_rate": 0.0001455482275350371, "loss": 0.0635, "step": 2999 }, { "epoch": 1.68, "learning_rate": 0.0001454863973619126, "loss": 0.0846, "step": 3000 }, { "epoch": 1.68, "eval_loss": 0.10184536129236221, "eval_runtime": 270.6236, "eval_samples_per_second": 9.763, "eval_steps_per_second": 0.613, "eval_wer": 0.08403411344704483, "step": 3000 }, { "epoch": 1.68, "learning_rate": 0.00014542456718878813, "loss": 0.0773, "step": 3001 }, { "epoch": 1.68, "learning_rate": 0.00014536273701566362, "loss": 0.1022, "step": 3002 }, { "epoch": 1.68, "learning_rate": 0.00014530090684253914, "loss": 0.0542, "step": 3003 }, { "epoch": 1.68, "learning_rate": 0.00014523907666941466, "loss": 0.0538, "step": 3004 }, { "epoch": 1.68, "learning_rate": 0.00014517724649629017, "loss": 0.0777, "step": 3005 }, { "epoch": 1.68, "learning_rate": 0.0001451154163231657, "loss": 0.1058, "step": 3006 }, { "epoch": 1.69, "learning_rate": 0.0001450535861500412, "loss": 0.1067, "step": 3007 }, { "epoch": 1.69, "learning_rate": 0.00014499175597691672, "loss": 0.0551, "step": 3008 }, { "epoch": 1.69, "learning_rate": 0.00014492992580379224, "loss": 0.058, "step": 3009 }, { "epoch": 1.69, "learning_rate": 0.00014486809563066776, "loss": 0.0574, "step": 3010 }, { "epoch": 1.69, "learning_rate": 0.00014480626545754327, "loss": 0.072, "step": 3011 }, { "epoch": 1.69, "learning_rate": 0.0001447444352844188, "loss": 0.0653, "step": 3012 }, { "epoch": 1.69, "learning_rate": 0.0001446826051112943, "loss": 0.0741, "step": 3013 }, { "epoch": 1.69, "learning_rate": 0.00014462077493816982, "loss": 0.0884, "step": 3014 }, { "epoch": 1.69, "learning_rate": 0.00014455894476504534, "loss": 0.0601, "step": 3015 }, { "epoch": 1.69, "learning_rate": 0.00014449711459192085, "loss": 0.0742, "step": 3016 }, { "epoch": 1.69, "learning_rate": 0.00014443528441879637, "loss": 0.121, "step": 3017 }, { "epoch": 1.69, "learning_rate": 0.00014437345424567186, "loss": 0.0593, "step": 3018 }, { "epoch": 1.69, "learning_rate": 0.0001443116240725474, "loss": 0.1013, "step": 3019 }, { "epoch": 1.69, "learning_rate": 0.0001442497938994229, "loss": 0.0407, "step": 3020 }, { "epoch": 1.69, "learning_rate": 0.0001441879637262984, "loss": 0.0846, "step": 3021 }, { "epoch": 1.69, "learning_rate": 0.00014412613355317393, "loss": 0.0512, "step": 3022 }, { "epoch": 1.69, "learning_rate": 0.00014406430338004944, "loss": 0.0759, "step": 3023 }, { "epoch": 1.7, "learning_rate": 0.00014400247320692496, "loss": 0.0611, "step": 3024 }, { "epoch": 1.7, "learning_rate": 0.00014394064303380048, "loss": 0.096, "step": 3025 }, { "epoch": 1.7, "learning_rate": 0.000143878812860676, "loss": 0.0672, "step": 3026 }, { "epoch": 1.7, "learning_rate": 0.0001438169826875515, "loss": 0.0814, "step": 3027 }, { "epoch": 1.7, "learning_rate": 0.00014375515251442703, "loss": 0.0895, "step": 3028 }, { "epoch": 1.7, "learning_rate": 0.00014369332234130254, "loss": 0.112, "step": 3029 }, { "epoch": 1.7, "learning_rate": 0.00014363149216817806, "loss": 0.1675, "step": 3030 }, { "epoch": 1.7, "learning_rate": 0.00014356966199505358, "loss": 0.1278, "step": 3031 }, { "epoch": 1.7, "learning_rate": 0.0001435078318219291, "loss": 0.1045, "step": 3032 }, { "epoch": 1.7, "learning_rate": 0.0001434460016488046, "loss": 0.1403, "step": 3033 }, { "epoch": 1.7, "learning_rate": 0.00014338417147568013, "loss": 0.234, "step": 3034 }, { "epoch": 1.7, "learning_rate": 0.00014332234130255564, "loss": 0.0512, "step": 3035 }, { "epoch": 1.7, "learning_rate": 0.00014326051112943113, "loss": 0.0701, "step": 3036 }, { "epoch": 1.7, "learning_rate": 0.00014319868095630668, "loss": 0.0583, "step": 3037 }, { "epoch": 1.7, "learning_rate": 0.0001431368507831822, "loss": 0.0702, "step": 3038 }, { "epoch": 1.7, "learning_rate": 0.00014307502061005768, "loss": 0.081, "step": 3039 }, { "epoch": 1.7, "learning_rate": 0.00014301319043693323, "loss": 0.0576, "step": 3040 }, { "epoch": 1.7, "learning_rate": 0.00014295136026380872, "loss": 0.0641, "step": 3041 }, { "epoch": 1.71, "learning_rate": 0.00014288953009068426, "loss": 0.0327, "step": 3042 }, { "epoch": 1.71, "learning_rate": 0.00014282769991755975, "loss": 0.0801, "step": 3043 }, { "epoch": 1.71, "learning_rate": 0.00014276586974443526, "loss": 0.0739, "step": 3044 }, { "epoch": 1.71, "learning_rate": 0.00014270403957131078, "loss": 0.0779, "step": 3045 }, { "epoch": 1.71, "learning_rate": 0.0001426422093981863, "loss": 0.0711, "step": 3046 }, { "epoch": 1.71, "learning_rate": 0.00014258037922506181, "loss": 0.0464, "step": 3047 }, { "epoch": 1.71, "learning_rate": 0.00014251854905193733, "loss": 0.076, "step": 3048 }, { "epoch": 1.71, "learning_rate": 0.00014245671887881285, "loss": 0.0827, "step": 3049 }, { "epoch": 1.71, "learning_rate": 0.00014239488870568836, "loss": 0.0642, "step": 3050 }, { "epoch": 1.71, "learning_rate": 0.00014233305853256388, "loss": 0.0757, "step": 3051 }, { "epoch": 1.71, "learning_rate": 0.0001422712283594394, "loss": 0.0913, "step": 3052 }, { "epoch": 1.71, "learning_rate": 0.00014220939818631491, "loss": 0.0892, "step": 3053 }, { "epoch": 1.71, "learning_rate": 0.00014214756801319043, "loss": 0.081, "step": 3054 }, { "epoch": 1.71, "learning_rate": 0.00014208573784006595, "loss": 0.0911, "step": 3055 }, { "epoch": 1.71, "learning_rate": 0.00014202390766694146, "loss": 0.0759, "step": 3056 }, { "epoch": 1.71, "learning_rate": 0.00014196207749381698, "loss": 0.0542, "step": 3057 }, { "epoch": 1.71, "learning_rate": 0.0001419002473206925, "loss": 0.0789, "step": 3058 }, { "epoch": 1.71, "learning_rate": 0.000141838417147568, "loss": 0.082, "step": 3059 }, { "epoch": 1.72, "learning_rate": 0.00014177658697444353, "loss": 0.0733, "step": 3060 }, { "epoch": 1.72, "learning_rate": 0.00014171475680131902, "loss": 0.0828, "step": 3061 }, { "epoch": 1.72, "learning_rate": 0.00014165292662819454, "loss": 0.0916, "step": 3062 }, { "epoch": 1.72, "learning_rate": 0.00014159109645507008, "loss": 0.051, "step": 3063 }, { "epoch": 1.72, "learning_rate": 0.00014152926628194557, "loss": 0.0595, "step": 3064 }, { "epoch": 1.72, "learning_rate": 0.00014146743610882109, "loss": 0.0742, "step": 3065 }, { "epoch": 1.72, "learning_rate": 0.0001414056059356966, "loss": 0.0871, "step": 3066 }, { "epoch": 1.72, "learning_rate": 0.00014134377576257212, "loss": 0.0829, "step": 3067 }, { "epoch": 1.72, "learning_rate": 0.00014128194558944764, "loss": 0.0645, "step": 3068 }, { "epoch": 1.72, "learning_rate": 0.00014122011541632315, "loss": 0.1052, "step": 3069 }, { "epoch": 1.72, "learning_rate": 0.00014115828524319867, "loss": 0.0924, "step": 3070 }, { "epoch": 1.72, "learning_rate": 0.00014109645507007419, "loss": 0.0929, "step": 3071 }, { "epoch": 1.72, "learning_rate": 0.0001410346248969497, "loss": 0.0704, "step": 3072 }, { "epoch": 1.72, "learning_rate": 0.00014097279472382522, "loss": 0.0782, "step": 3073 }, { "epoch": 1.72, "learning_rate": 0.00014091096455070074, "loss": 0.0767, "step": 3074 }, { "epoch": 1.72, "learning_rate": 0.00014084913437757625, "loss": 0.0744, "step": 3075 }, { "epoch": 1.72, "learning_rate": 0.00014078730420445177, "loss": 0.0582, "step": 3076 }, { "epoch": 1.72, "learning_rate": 0.00014072547403132726, "loss": 0.0857, "step": 3077 }, { "epoch": 1.73, "learning_rate": 0.0001406636438582028, "loss": 0.0875, "step": 3078 }, { "epoch": 1.73, "learning_rate": 0.00014060181368507832, "loss": 0.0654, "step": 3079 }, { "epoch": 1.73, "learning_rate": 0.0001405399835119538, "loss": 0.0722, "step": 3080 }, { "epoch": 1.73, "learning_rate": 0.00014047815333882935, "loss": 0.0995, "step": 3081 }, { "epoch": 1.73, "learning_rate": 0.00014041632316570484, "loss": 0.1707, "step": 3082 }, { "epoch": 1.73, "learning_rate": 0.00014035449299258036, "loss": 0.1692, "step": 3083 }, { "epoch": 1.73, "learning_rate": 0.00014029266281945587, "loss": 0.1278, "step": 3084 }, { "epoch": 1.73, "learning_rate": 0.0001402308326463314, "loss": 0.0667, "step": 3085 }, { "epoch": 1.73, "learning_rate": 0.0001401690024732069, "loss": 0.084, "step": 3086 }, { "epoch": 1.73, "learning_rate": 0.00014010717230008242, "loss": 0.0665, "step": 3087 }, { "epoch": 1.73, "learning_rate": 0.00014004534212695794, "loss": 0.0811, "step": 3088 }, { "epoch": 1.73, "learning_rate": 0.00013998351195383346, "loss": 0.0773, "step": 3089 }, { "epoch": 1.73, "learning_rate": 0.00013992168178070897, "loss": 0.0702, "step": 3090 }, { "epoch": 1.73, "learning_rate": 0.0001398598516075845, "loss": 0.0695, "step": 3091 }, { "epoch": 1.73, "learning_rate": 0.00013979802143446, "loss": 0.0965, "step": 3092 }, { "epoch": 1.73, "learning_rate": 0.00013973619126133552, "loss": 0.0741, "step": 3093 }, { "epoch": 1.73, "learning_rate": 0.00013967436108821104, "loss": 0.0713, "step": 3094 }, { "epoch": 1.73, "learning_rate": 0.00013961253091508656, "loss": 0.0538, "step": 3095 }, { "epoch": 1.74, "learning_rate": 0.00013955070074196207, "loss": 0.0864, "step": 3096 }, { "epoch": 1.74, "learning_rate": 0.0001394888705688376, "loss": 0.0477, "step": 3097 }, { "epoch": 1.74, "learning_rate": 0.00013942704039571308, "loss": 0.0462, "step": 3098 }, { "epoch": 1.74, "learning_rate": 0.00013936521022258862, "loss": 0.0591, "step": 3099 }, { "epoch": 1.74, "learning_rate": 0.0001393033800494641, "loss": 0.0684, "step": 3100 }, { "epoch": 1.74, "learning_rate": 0.00013924154987633963, "loss": 0.0688, "step": 3101 }, { "epoch": 1.74, "learning_rate": 0.00013917971970321517, "loss": 0.0802, "step": 3102 }, { "epoch": 1.74, "learning_rate": 0.00013911788953009066, "loss": 0.0858, "step": 3103 }, { "epoch": 1.74, "learning_rate": 0.0001390560593569662, "loss": 0.0737, "step": 3104 }, { "epoch": 1.74, "learning_rate": 0.0001389942291838417, "loss": 0.0844, "step": 3105 }, { "epoch": 1.74, "learning_rate": 0.0001389323990107172, "loss": 0.0577, "step": 3106 }, { "epoch": 1.74, "learning_rate": 0.00013887056883759273, "loss": 0.0904, "step": 3107 }, { "epoch": 1.74, "learning_rate": 0.00013880873866446825, "loss": 0.0678, "step": 3108 }, { "epoch": 1.74, "learning_rate": 0.00013874690849134376, "loss": 0.0785, "step": 3109 }, { "epoch": 1.74, "learning_rate": 0.00013868507831821928, "loss": 0.1072, "step": 3110 }, { "epoch": 1.74, "learning_rate": 0.0001386232481450948, "loss": 0.0633, "step": 3111 }, { "epoch": 1.74, "learning_rate": 0.0001385614179719703, "loss": 0.1323, "step": 3112 }, { "epoch": 1.74, "learning_rate": 0.00013849958779884583, "loss": 0.0646, "step": 3113 }, { "epoch": 1.75, "learning_rate": 0.00013843775762572134, "loss": 0.0738, "step": 3114 }, { "epoch": 1.75, "learning_rate": 0.00013837592745259686, "loss": 0.0909, "step": 3115 }, { "epoch": 1.75, "learning_rate": 0.00013831409727947235, "loss": 0.0623, "step": 3116 }, { "epoch": 1.75, "learning_rate": 0.0001382522671063479, "loss": 0.0788, "step": 3117 }, { "epoch": 1.75, "learning_rate": 0.0001381904369332234, "loss": 0.0547, "step": 3118 }, { "epoch": 1.75, "learning_rate": 0.00013812860676009893, "loss": 0.0543, "step": 3119 }, { "epoch": 1.75, "learning_rate": 0.00013806677658697444, "loss": 0.0615, "step": 3120 }, { "epoch": 1.75, "learning_rate": 0.00013800494641384993, "loss": 0.1016, "step": 3121 }, { "epoch": 1.75, "learning_rate": 0.00013794311624072548, "loss": 0.056, "step": 3122 }, { "epoch": 1.75, "learning_rate": 0.00013788128606760097, "loss": 0.0382, "step": 3123 }, { "epoch": 1.75, "learning_rate": 0.00013781945589447648, "loss": 0.0661, "step": 3124 }, { "epoch": 1.75, "learning_rate": 0.000137757625721352, "loss": 0.101, "step": 3125 }, { "epoch": 1.75, "learning_rate": 0.00013769579554822752, "loss": 0.0862, "step": 3126 }, { "epoch": 1.75, "learning_rate": 0.00013763396537510303, "loss": 0.0624, "step": 3127 }, { "epoch": 1.75, "learning_rate": 0.00013757213520197855, "loss": 0.0814, "step": 3128 }, { "epoch": 1.75, "learning_rate": 0.00013751030502885407, "loss": 0.0761, "step": 3129 }, { "epoch": 1.75, "learning_rate": 0.00013744847485572958, "loss": 0.0993, "step": 3130 }, { "epoch": 1.76, "learning_rate": 0.0001373866446826051, "loss": 0.0717, "step": 3131 }, { "epoch": 1.76, "learning_rate": 0.00013732481450948062, "loss": 0.1937, "step": 3132 }, { "epoch": 1.76, "learning_rate": 0.00013726298433635613, "loss": 0.2313, "step": 3133 }, { "epoch": 1.76, "learning_rate": 0.00013720115416323165, "loss": 0.1096, "step": 3134 }, { "epoch": 1.76, "learning_rate": 0.00013713932399010717, "loss": 0.0666, "step": 3135 }, { "epoch": 1.76, "learning_rate": 0.00013707749381698268, "loss": 0.0634, "step": 3136 }, { "epoch": 1.76, "learning_rate": 0.0001370156636438582, "loss": 0.0696, "step": 3137 }, { "epoch": 1.76, "learning_rate": 0.00013695383347073372, "loss": 0.0819, "step": 3138 }, { "epoch": 1.76, "learning_rate": 0.0001368920032976092, "loss": 0.0667, "step": 3139 }, { "epoch": 1.76, "learning_rate": 0.00013683017312448475, "loss": 0.0612, "step": 3140 }, { "epoch": 1.76, "learning_rate": 0.00013676834295136024, "loss": 0.0888, "step": 3141 }, { "epoch": 1.76, "learning_rate": 0.00013670651277823575, "loss": 0.0727, "step": 3142 }, { "epoch": 1.76, "learning_rate": 0.0001366446826051113, "loss": 0.0619, "step": 3143 }, { "epoch": 1.76, "learning_rate": 0.0001365828524319868, "loss": 0.074, "step": 3144 }, { "epoch": 1.76, "learning_rate": 0.0001365210222588623, "loss": 0.0607, "step": 3145 }, { "epoch": 1.76, "learning_rate": 0.00013645919208573782, "loss": 0.0491, "step": 3146 }, { "epoch": 1.76, "learning_rate": 0.00013639736191261334, "loss": 0.0768, "step": 3147 }, { "epoch": 1.76, "learning_rate": 0.00013633553173948885, "loss": 0.0952, "step": 3148 }, { "epoch": 1.77, "learning_rate": 0.00013627370156636437, "loss": 0.0618, "step": 3149 }, { "epoch": 1.77, "learning_rate": 0.0001362118713932399, "loss": 0.0581, "step": 3150 }, { "epoch": 1.77, "learning_rate": 0.0001361500412201154, "loss": 0.0844, "step": 3151 }, { "epoch": 1.77, "learning_rate": 0.00013608821104699092, "loss": 0.0935, "step": 3152 }, { "epoch": 1.77, "learning_rate": 0.00013602638087386644, "loss": 0.0793, "step": 3153 }, { "epoch": 1.77, "learning_rate": 0.00013596455070074195, "loss": 0.0967, "step": 3154 }, { "epoch": 1.77, "learning_rate": 0.00013590272052761747, "loss": 0.0614, "step": 3155 }, { "epoch": 1.77, "learning_rate": 0.000135840890354493, "loss": 0.0893, "step": 3156 }, { "epoch": 1.77, "learning_rate": 0.0001357790601813685, "loss": 0.0719, "step": 3157 }, { "epoch": 1.77, "learning_rate": 0.00013571723000824402, "loss": 0.0535, "step": 3158 }, { "epoch": 1.77, "learning_rate": 0.00013565539983511954, "loss": 0.0541, "step": 3159 }, { "epoch": 1.77, "learning_rate": 0.00013559356966199503, "loss": 0.0434, "step": 3160 }, { "epoch": 1.77, "learning_rate": 0.00013553173948887057, "loss": 0.0677, "step": 3161 }, { "epoch": 1.77, "learning_rate": 0.00013546990931574606, "loss": 0.0927, "step": 3162 }, { "epoch": 1.77, "learning_rate": 0.00013540807914262158, "loss": 0.0552, "step": 3163 }, { "epoch": 1.77, "learning_rate": 0.0001353462489694971, "loss": 0.0904, "step": 3164 }, { "epoch": 1.77, "learning_rate": 0.0001352844187963726, "loss": 0.059, "step": 3165 }, { "epoch": 1.77, "learning_rate": 0.00013522258862324815, "loss": 0.0468, "step": 3166 }, { "epoch": 1.78, "learning_rate": 0.00013516075845012364, "loss": 0.1241, "step": 3167 }, { "epoch": 1.78, "learning_rate": 0.00013509892827699916, "loss": 0.085, "step": 3168 }, { "epoch": 1.78, "learning_rate": 0.00013503709810387468, "loss": 0.0686, "step": 3169 }, { "epoch": 1.78, "learning_rate": 0.0001349752679307502, "loss": 0.074, "step": 3170 }, { "epoch": 1.78, "learning_rate": 0.0001349134377576257, "loss": 0.0771, "step": 3171 }, { "epoch": 1.78, "learning_rate": 0.00013485160758450123, "loss": 0.0731, "step": 3172 }, { "epoch": 1.78, "learning_rate": 0.00013478977741137674, "loss": 0.0742, "step": 3173 }, { "epoch": 1.78, "learning_rate": 0.00013472794723825226, "loss": 0.0904, "step": 3174 }, { "epoch": 1.78, "learning_rate": 0.00013466611706512778, "loss": 0.0489, "step": 3175 }, { "epoch": 1.78, "learning_rate": 0.0001346042868920033, "loss": 0.1128, "step": 3176 }, { "epoch": 1.78, "learning_rate": 0.0001345424567188788, "loss": 0.1229, "step": 3177 }, { "epoch": 1.78, "learning_rate": 0.0001344806265457543, "loss": 0.0469, "step": 3178 }, { "epoch": 1.78, "learning_rate": 0.00013441879637262984, "loss": 0.0639, "step": 3179 }, { "epoch": 1.78, "learning_rate": 0.00013435696619950533, "loss": 0.0891, "step": 3180 }, { "epoch": 1.78, "learning_rate": 0.00013429513602638087, "loss": 0.1098, "step": 3181 }, { "epoch": 1.78, "learning_rate": 0.0001342333058532564, "loss": 0.1006, "step": 3182 }, { "epoch": 1.78, "learning_rate": 0.00013417147568013188, "loss": 0.1876, "step": 3183 }, { "epoch": 1.78, "learning_rate": 0.00013410964550700742, "loss": 0.2237, "step": 3184 }, { "epoch": 1.79, "learning_rate": 0.00013404781533388291, "loss": 0.0867, "step": 3185 }, { "epoch": 1.79, "learning_rate": 0.00013398598516075843, "loss": 0.0698, "step": 3186 }, { "epoch": 1.79, "learning_rate": 0.00013392415498763395, "loss": 0.0652, "step": 3187 }, { "epoch": 1.79, "learning_rate": 0.00013386232481450946, "loss": 0.0927, "step": 3188 }, { "epoch": 1.79, "learning_rate": 0.00013380049464138498, "loss": 0.0584, "step": 3189 }, { "epoch": 1.79, "learning_rate": 0.0001337386644682605, "loss": 0.0666, "step": 3190 }, { "epoch": 1.79, "learning_rate": 0.000133676834295136, "loss": 0.0879, "step": 3191 }, { "epoch": 1.79, "learning_rate": 0.00013361500412201153, "loss": 0.0744, "step": 3192 }, { "epoch": 1.79, "learning_rate": 0.00013355317394888705, "loss": 0.0742, "step": 3193 }, { "epoch": 1.79, "learning_rate": 0.00013349134377576256, "loss": 0.0934, "step": 3194 }, { "epoch": 1.79, "learning_rate": 0.00013342951360263808, "loss": 0.0572, "step": 3195 }, { "epoch": 1.79, "learning_rate": 0.0001333676834295136, "loss": 0.084, "step": 3196 }, { "epoch": 1.79, "learning_rate": 0.0001333058532563891, "loss": 0.0657, "step": 3197 }, { "epoch": 1.79, "learning_rate": 0.00013324402308326463, "loss": 0.0643, "step": 3198 }, { "epoch": 1.79, "learning_rate": 0.00013318219291014015, "loss": 0.0932, "step": 3199 }, { "epoch": 1.79, "learning_rate": 0.00013312036273701566, "loss": 0.0637, "step": 3200 }, { "epoch": 1.79, "learning_rate": 0.00013305853256389115, "loss": 0.0922, "step": 3201 }, { "epoch": 1.79, "learning_rate": 0.0001329967023907667, "loss": 0.0944, "step": 3202 }, { "epoch": 1.8, "learning_rate": 0.00013293487221764219, "loss": 0.0877, "step": 3203 }, { "epoch": 1.8, "learning_rate": 0.0001328730420445177, "loss": 0.0875, "step": 3204 }, { "epoch": 1.8, "learning_rate": 0.00013281121187139325, "loss": 0.0623, "step": 3205 }, { "epoch": 1.8, "learning_rate": 0.00013274938169826873, "loss": 0.0872, "step": 3206 }, { "epoch": 1.8, "learning_rate": 0.00013268755152514425, "loss": 0.0812, "step": 3207 }, { "epoch": 1.8, "learning_rate": 0.00013262572135201977, "loss": 0.0506, "step": 3208 }, { "epoch": 1.8, "learning_rate": 0.00013256389117889528, "loss": 0.0896, "step": 3209 }, { "epoch": 1.8, "learning_rate": 0.0001325020610057708, "loss": 0.0415, "step": 3210 }, { "epoch": 1.8, "learning_rate": 0.00013244023083264632, "loss": 0.0598, "step": 3211 }, { "epoch": 1.8, "learning_rate": 0.00013237840065952183, "loss": 0.079, "step": 3212 }, { "epoch": 1.8, "learning_rate": 0.00013231657048639735, "loss": 0.0427, "step": 3213 }, { "epoch": 1.8, "learning_rate": 0.00013225474031327287, "loss": 0.0669, "step": 3214 }, { "epoch": 1.8, "learning_rate": 0.00013219291014014838, "loss": 0.1134, "step": 3215 }, { "epoch": 1.8, "learning_rate": 0.0001321310799670239, "loss": 0.0551, "step": 3216 }, { "epoch": 1.8, "learning_rate": 0.00013206924979389942, "loss": 0.0689, "step": 3217 }, { "epoch": 1.8, "learning_rate": 0.00013200741962077493, "loss": 0.0622, "step": 3218 }, { "epoch": 1.8, "learning_rate": 0.00013194558944765042, "loss": 0.075, "step": 3219 }, { "epoch": 1.8, "learning_rate": 0.00013188375927452597, "loss": 0.0713, "step": 3220 }, { "epoch": 1.81, "learning_rate": 0.00013182192910140148, "loss": 0.0909, "step": 3221 }, { "epoch": 1.81, "learning_rate": 0.00013176009892827697, "loss": 0.0497, "step": 3222 }, { "epoch": 1.81, "learning_rate": 0.00013169826875515252, "loss": 0.082, "step": 3223 }, { "epoch": 1.81, "learning_rate": 0.000131636438582028, "loss": 0.0497, "step": 3224 }, { "epoch": 1.81, "learning_rate": 0.00013157460840890355, "loss": 0.0635, "step": 3225 }, { "epoch": 1.81, "learning_rate": 0.00013151277823577904, "loss": 0.0786, "step": 3226 }, { "epoch": 1.81, "learning_rate": 0.00013145094806265456, "loss": 0.0896, "step": 3227 }, { "epoch": 1.81, "learning_rate": 0.00013138911788953007, "loss": 0.073, "step": 3228 }, { "epoch": 1.81, "learning_rate": 0.0001313272877164056, "loss": 0.0706, "step": 3229 }, { "epoch": 1.81, "learning_rate": 0.0001312654575432811, "loss": 0.1021, "step": 3230 }, { "epoch": 1.81, "learning_rate": 0.00013120362737015662, "loss": 0.1333, "step": 3231 }, { "epoch": 1.81, "learning_rate": 0.00013114179719703214, "loss": 0.1307, "step": 3232 }, { "epoch": 1.81, "learning_rate": 0.00013107996702390766, "loss": 0.1707, "step": 3233 }, { "epoch": 1.81, "learning_rate": 0.00013101813685078317, "loss": 0.1365, "step": 3234 }, { "epoch": 1.81, "learning_rate": 0.0001309563066776587, "loss": 0.079, "step": 3235 }, { "epoch": 1.81, "learning_rate": 0.0001308944765045342, "loss": 0.0978, "step": 3236 }, { "epoch": 1.81, "learning_rate": 0.00013083264633140972, "loss": 0.0732, "step": 3237 }, { "epoch": 1.82, "learning_rate": 0.00013077081615828524, "loss": 0.0618, "step": 3238 }, { "epoch": 1.82, "learning_rate": 0.00013070898598516076, "loss": 0.0542, "step": 3239 }, { "epoch": 1.82, "learning_rate": 0.00013064715581203624, "loss": 0.0454, "step": 3240 }, { "epoch": 1.82, "learning_rate": 0.0001305853256389118, "loss": 0.0627, "step": 3241 }, { "epoch": 1.82, "learning_rate": 0.00013052349546578728, "loss": 0.0925, "step": 3242 }, { "epoch": 1.82, "learning_rate": 0.00013046166529266282, "loss": 0.0767, "step": 3243 }, { "epoch": 1.82, "learning_rate": 0.0001303998351195383, "loss": 0.0882, "step": 3244 }, { "epoch": 1.82, "learning_rate": 0.00013033800494641383, "loss": 0.0588, "step": 3245 }, { "epoch": 1.82, "learning_rate": 0.00013027617477328937, "loss": 0.0877, "step": 3246 }, { "epoch": 1.82, "learning_rate": 0.00013021434460016486, "loss": 0.0884, "step": 3247 }, { "epoch": 1.82, "learning_rate": 0.00013015251442704038, "loss": 0.0666, "step": 3248 }, { "epoch": 1.82, "learning_rate": 0.0001300906842539159, "loss": 0.0696, "step": 3249 }, { "epoch": 1.82, "learning_rate": 0.0001300288540807914, "loss": 0.0884, "step": 3250 }, { "epoch": 1.82, "learning_rate": 0.00012996702390766693, "loss": 0.0611, "step": 3251 }, { "epoch": 1.82, "learning_rate": 0.00012990519373454244, "loss": 0.0909, "step": 3252 }, { "epoch": 1.82, "learning_rate": 0.00012984336356141796, "loss": 0.0753, "step": 3253 }, { "epoch": 1.82, "learning_rate": 0.00012978153338829348, "loss": 0.0611, "step": 3254 }, { "epoch": 1.82, "learning_rate": 0.000129719703215169, "loss": 0.0431, "step": 3255 }, { "epoch": 1.83, "learning_rate": 0.0001296578730420445, "loss": 0.0429, "step": 3256 }, { "epoch": 1.83, "learning_rate": 0.00012959604286892003, "loss": 0.1077, "step": 3257 }, { "epoch": 1.83, "learning_rate": 0.00012953421269579554, "loss": 0.1027, "step": 3258 }, { "epoch": 1.83, "learning_rate": 0.00012947238252267106, "loss": 0.0947, "step": 3259 }, { "epoch": 1.83, "learning_rate": 0.00012941055234954658, "loss": 0.0587, "step": 3260 }, { "epoch": 1.83, "learning_rate": 0.0001293487221764221, "loss": 0.0539, "step": 3261 }, { "epoch": 1.83, "learning_rate": 0.0001292868920032976, "loss": 0.0756, "step": 3262 }, { "epoch": 1.83, "learning_rate": 0.0001292250618301731, "loss": 0.0739, "step": 3263 }, { "epoch": 1.83, "learning_rate": 0.00012916323165704864, "loss": 0.0724, "step": 3264 }, { "epoch": 1.83, "learning_rate": 0.00012910140148392413, "loss": 0.0793, "step": 3265 }, { "epoch": 1.83, "learning_rate": 0.00012903957131079965, "loss": 0.0747, "step": 3266 }, { "epoch": 1.83, "learning_rate": 0.00012897774113767517, "loss": 0.06, "step": 3267 }, { "epoch": 1.83, "learning_rate": 0.00012891591096455068, "loss": 0.0735, "step": 3268 }, { "epoch": 1.83, "learning_rate": 0.0001288540807914262, "loss": 0.0686, "step": 3269 }, { "epoch": 1.83, "learning_rate": 0.00012879225061830172, "loss": 0.1015, "step": 3270 }, { "epoch": 1.83, "learning_rate": 0.00012873042044517723, "loss": 0.0531, "step": 3271 }, { "epoch": 1.83, "learning_rate": 0.00012866859027205275, "loss": 0.0635, "step": 3272 }, { "epoch": 1.83, "learning_rate": 0.00012860676009892826, "loss": 0.0721, "step": 3273 }, { "epoch": 1.84, "learning_rate": 0.00012854492992580378, "loss": 0.0893, "step": 3274 }, { "epoch": 1.84, "learning_rate": 0.0001284830997526793, "loss": 0.1108, "step": 3275 }, { "epoch": 1.84, "learning_rate": 0.00012842126957955481, "loss": 0.0757, "step": 3276 }, { "epoch": 1.84, "learning_rate": 0.00012835943940643033, "loss": 0.0539, "step": 3277 }, { "epoch": 1.84, "learning_rate": 0.00012829760923330585, "loss": 0.0815, "step": 3278 }, { "epoch": 1.84, "learning_rate": 0.00012823577906018136, "loss": 0.1021, "step": 3279 }, { "epoch": 1.84, "learning_rate": 0.00012817394888705688, "loss": 0.0472, "step": 3280 }, { "epoch": 1.84, "learning_rate": 0.00012811211871393237, "loss": 0.116, "step": 3281 }, { "epoch": 1.84, "learning_rate": 0.00012805028854080791, "loss": 0.0516, "step": 3282 }, { "epoch": 1.84, "learning_rate": 0.0001279884583676834, "loss": 0.2085, "step": 3283 }, { "epoch": 1.84, "learning_rate": 0.00012792662819455892, "loss": 0.2577, "step": 3284 }, { "epoch": 1.84, "learning_rate": 0.00012786479802143446, "loss": 0.0452, "step": 3285 }, { "epoch": 1.84, "learning_rate": 0.00012780296784830995, "loss": 0.091, "step": 3286 }, { "epoch": 1.84, "learning_rate": 0.0001277411376751855, "loss": 0.0609, "step": 3287 }, { "epoch": 1.84, "learning_rate": 0.000127679307502061, "loss": 0.0705, "step": 3288 }, { "epoch": 1.84, "learning_rate": 0.0001276174773289365, "loss": 0.0813, "step": 3289 }, { "epoch": 1.84, "learning_rate": 0.00012755564715581202, "loss": 0.0874, "step": 3290 }, { "epoch": 1.84, "learning_rate": 0.00012749381698268754, "loss": 0.0831, "step": 3291 }, { "epoch": 1.85, "learning_rate": 0.00012743198680956305, "loss": 0.0886, "step": 3292 }, { "epoch": 1.85, "learning_rate": 0.00012737015663643857, "loss": 0.0806, "step": 3293 }, { "epoch": 1.85, "learning_rate": 0.00012730832646331409, "loss": 0.0459, "step": 3294 }, { "epoch": 1.85, "learning_rate": 0.0001272464962901896, "loss": 0.0483, "step": 3295 }, { "epoch": 1.85, "learning_rate": 0.00012718466611706512, "loss": 0.0691, "step": 3296 }, { "epoch": 1.85, "learning_rate": 0.00012712283594394064, "loss": 0.0655, "step": 3297 }, { "epoch": 1.85, "learning_rate": 0.00012706100577081615, "loss": 0.1171, "step": 3298 }, { "epoch": 1.85, "learning_rate": 0.00012699917559769164, "loss": 0.0919, "step": 3299 }, { "epoch": 1.85, "learning_rate": 0.00012693734542456719, "loss": 0.0684, "step": 3300 }, { "epoch": 1.85, "eval_loss": 0.09782690554857254, "eval_runtime": 272.1082, "eval_samples_per_second": 9.709, "eval_steps_per_second": 0.61, "eval_wer": 0.08092026973423244, "step": 3300 }, { "epoch": 1.85, "learning_rate": 0.0001268755152514427, "loss": 0.0734, "step": 3301 }, { "epoch": 1.85, "learning_rate": 0.00012681368507831822, "loss": 0.0431, "step": 3302 }, { "epoch": 1.85, "learning_rate": 0.00012675185490519374, "loss": 0.0956, "step": 3303 }, { "epoch": 1.85, "learning_rate": 0.00012669002473206922, "loss": 0.0714, "step": 3304 }, { "epoch": 1.85, "learning_rate": 0.00012662819455894477, "loss": 0.0748, "step": 3305 }, { "epoch": 1.85, "learning_rate": 0.00012656636438582026, "loss": 0.0797, "step": 3306 }, { "epoch": 1.85, "learning_rate": 0.00012650453421269577, "loss": 0.0671, "step": 3307 }, { "epoch": 1.85, "learning_rate": 0.00012644270403957132, "loss": 0.0733, "step": 3308 }, { "epoch": 1.85, "learning_rate": 0.0001263808738664468, "loss": 0.0628, "step": 3309 }, { "epoch": 1.86, "learning_rate": 0.00012631904369332232, "loss": 0.0735, "step": 3310 }, { "epoch": 1.86, "learning_rate": 0.00012625721352019784, "loss": 0.0672, "step": 3311 }, { "epoch": 1.86, "learning_rate": 0.00012619538334707336, "loss": 0.0704, "step": 3312 }, { "epoch": 1.86, "learning_rate": 0.00012613355317394887, "loss": 0.0761, "step": 3313 }, { "epoch": 1.86, "learning_rate": 0.0001260717230008244, "loss": 0.0499, "step": 3314 }, { "epoch": 1.86, "learning_rate": 0.0001260098928276999, "loss": 0.0524, "step": 3315 }, { "epoch": 1.86, "learning_rate": 0.00012594806265457542, "loss": 0.0819, "step": 3316 }, { "epoch": 1.86, "learning_rate": 0.00012588623248145094, "loss": 0.0424, "step": 3317 }, { "epoch": 1.86, "learning_rate": 0.00012582440230832646, "loss": 0.0786, "step": 3318 }, { "epoch": 1.86, "learning_rate": 0.00012576257213520197, "loss": 0.0508, "step": 3319 }, { "epoch": 1.86, "learning_rate": 0.0001257007419620775, "loss": 0.0594, "step": 3320 }, { "epoch": 1.86, "learning_rate": 0.000125638911788953, "loss": 0.0518, "step": 3321 }, { "epoch": 1.86, "learning_rate": 0.0001255770816158285, "loss": 0.0553, "step": 3322 }, { "epoch": 1.86, "learning_rate": 0.00012551525144270404, "loss": 0.0708, "step": 3323 }, { "epoch": 1.86, "learning_rate": 0.00012545342126957956, "loss": 0.068, "step": 3324 }, { "epoch": 1.86, "learning_rate": 0.00012539159109645505, "loss": 0.0254, "step": 3325 }, { "epoch": 1.86, "learning_rate": 0.0001253297609233306, "loss": 0.0802, "step": 3326 }, { "epoch": 1.86, "learning_rate": 0.00012526793075020608, "loss": 0.0954, "step": 3327 }, { "epoch": 1.87, "learning_rate": 0.0001252061005770816, "loss": 0.05, "step": 3328 }, { "epoch": 1.87, "learning_rate": 0.0001251442704039571, "loss": 0.0715, "step": 3329 }, { "epoch": 1.87, "learning_rate": 0.00012508244023083263, "loss": 0.0939, "step": 3330 }, { "epoch": 1.87, "learning_rate": 0.00012502061005770815, "loss": 0.1638, "step": 3331 }, { "epoch": 1.87, "learning_rate": 0.00012495877988458366, "loss": 0.1616, "step": 3332 }, { "epoch": 1.87, "learning_rate": 0.00012489694971145918, "loss": 0.1328, "step": 3333 }, { "epoch": 1.87, "learning_rate": 0.0001248351195383347, "loss": 0.2434, "step": 3334 }, { "epoch": 1.87, "learning_rate": 0.0001247732893652102, "loss": 0.0871, "step": 3335 }, { "epoch": 1.87, "learning_rate": 0.00012471145919208573, "loss": 0.0648, "step": 3336 }, { "epoch": 1.87, "learning_rate": 0.00012464962901896125, "loss": 0.0552, "step": 3337 }, { "epoch": 1.87, "learning_rate": 0.00012458779884583676, "loss": 0.052, "step": 3338 }, { "epoch": 1.87, "learning_rate": 0.00012452596867271228, "loss": 0.0539, "step": 3339 }, { "epoch": 1.87, "learning_rate": 0.0001244641384995878, "loss": 0.0616, "step": 3340 }, { "epoch": 1.87, "learning_rate": 0.0001244023083264633, "loss": 0.0652, "step": 3341 }, { "epoch": 1.87, "learning_rate": 0.00012434047815333883, "loss": 0.0748, "step": 3342 }, { "epoch": 1.87, "learning_rate": 0.00012427864798021432, "loss": 0.0598, "step": 3343 }, { "epoch": 1.87, "learning_rate": 0.00012421681780708986, "loss": 0.0619, "step": 3344 }, { "epoch": 1.88, "learning_rate": 0.00012415498763396535, "loss": 0.0689, "step": 3345 }, { "epoch": 1.88, "learning_rate": 0.00012409315746084087, "loss": 0.084, "step": 3346 }, { "epoch": 1.88, "learning_rate": 0.00012403132728771638, "loss": 0.0441, "step": 3347 }, { "epoch": 1.88, "learning_rate": 0.0001239694971145919, "loss": 0.0658, "step": 3348 }, { "epoch": 1.88, "learning_rate": 0.00012390766694146744, "loss": 0.0668, "step": 3349 }, { "epoch": 1.88, "learning_rate": 0.00012384583676834293, "loss": 0.0517, "step": 3350 }, { "epoch": 1.88, "learning_rate": 0.00012378400659521845, "loss": 0.0749, "step": 3351 }, { "epoch": 1.88, "learning_rate": 0.00012372217642209397, "loss": 0.0812, "step": 3352 }, { "epoch": 1.88, "learning_rate": 0.00012366034624896948, "loss": 0.0712, "step": 3353 }, { "epoch": 1.88, "learning_rate": 0.000123598516075845, "loss": 0.0518, "step": 3354 }, { "epoch": 1.88, "learning_rate": 0.00012353668590272052, "loss": 0.0599, "step": 3355 }, { "epoch": 1.88, "learning_rate": 0.00012347485572959603, "loss": 0.045, "step": 3356 }, { "epoch": 1.88, "learning_rate": 0.00012341302555647155, "loss": 0.0827, "step": 3357 }, { "epoch": 1.88, "learning_rate": 0.00012335119538334707, "loss": 0.0432, "step": 3358 }, { "epoch": 1.88, "learning_rate": 0.00012328936521022258, "loss": 0.0737, "step": 3359 }, { "epoch": 1.88, "learning_rate": 0.0001232275350370981, "loss": 0.0482, "step": 3360 }, { "epoch": 1.88, "learning_rate": 0.0001231657048639736, "loss": 0.0976, "step": 3361 }, { "epoch": 1.88, "learning_rate": 0.00012310387469084913, "loss": 0.0725, "step": 3362 }, { "epoch": 1.89, "learning_rate": 0.00012304204451772462, "loss": 0.0838, "step": 3363 }, { "epoch": 1.89, "learning_rate": 0.00012298021434460017, "loss": 0.0687, "step": 3364 }, { "epoch": 1.89, "learning_rate": 0.00012291838417147568, "loss": 0.0542, "step": 3365 }, { "epoch": 1.89, "learning_rate": 0.00012285655399835117, "loss": 0.059, "step": 3366 }, { "epoch": 1.89, "learning_rate": 0.00012279472382522672, "loss": 0.0725, "step": 3367 }, { "epoch": 1.89, "learning_rate": 0.0001227328936521022, "loss": 0.066, "step": 3368 }, { "epoch": 1.89, "learning_rate": 0.00012267106347897772, "loss": 0.0505, "step": 3369 }, { "epoch": 1.89, "learning_rate": 0.00012260923330585324, "loss": 0.0337, "step": 3370 }, { "epoch": 1.89, "learning_rate": 0.00012254740313272875, "loss": 0.0569, "step": 3371 }, { "epoch": 1.89, "learning_rate": 0.00012248557295960427, "loss": 0.0748, "step": 3372 }, { "epoch": 1.89, "learning_rate": 0.0001224237427864798, "loss": 0.0419, "step": 3373 }, { "epoch": 1.89, "learning_rate": 0.0001223619126133553, "loss": 0.0633, "step": 3374 }, { "epoch": 1.89, "learning_rate": 0.00012230008244023082, "loss": 0.0975, "step": 3375 }, { "epoch": 1.89, "learning_rate": 0.00012223825226710634, "loss": 0.0617, "step": 3376 }, { "epoch": 1.89, "learning_rate": 0.00012217642209398185, "loss": 0.0854, "step": 3377 }, { "epoch": 1.89, "learning_rate": 0.00012211459192085737, "loss": 0.0753, "step": 3378 }, { "epoch": 1.89, "learning_rate": 0.00012205276174773289, "loss": 0.0629, "step": 3379 }, { "epoch": 1.89, "learning_rate": 0.00012199093157460839, "loss": 0.1046, "step": 3380 }, { "epoch": 1.9, "learning_rate": 0.00012192910140148392, "loss": 0.067, "step": 3381 }, { "epoch": 1.9, "learning_rate": 0.00012186727122835942, "loss": 0.1327, "step": 3382 }, { "epoch": 1.9, "learning_rate": 0.00012180544105523495, "loss": 0.1564, "step": 3383 }, { "epoch": 1.9, "learning_rate": 0.00012174361088211046, "loss": 0.1622, "step": 3384 }, { "epoch": 1.9, "learning_rate": 0.00012168178070898597, "loss": 0.0787, "step": 3385 }, { "epoch": 1.9, "learning_rate": 0.00012161995053586148, "loss": 0.0602, "step": 3386 }, { "epoch": 1.9, "learning_rate": 0.000121558120362737, "loss": 0.0816, "step": 3387 }, { "epoch": 1.9, "learning_rate": 0.00012149629018961252, "loss": 0.0957, "step": 3388 }, { "epoch": 1.9, "learning_rate": 0.00012143446001648804, "loss": 0.0937, "step": 3389 }, { "epoch": 1.9, "learning_rate": 0.00012137262984336356, "loss": 0.0818, "step": 3390 }, { "epoch": 1.9, "learning_rate": 0.00012131079967023906, "loss": 0.0878, "step": 3391 }, { "epoch": 1.9, "learning_rate": 0.00012124896949711459, "loss": 0.0547, "step": 3392 }, { "epoch": 1.9, "learning_rate": 0.00012118713932399009, "loss": 0.066, "step": 3393 }, { "epoch": 1.9, "learning_rate": 0.00012112530915086561, "loss": 0.0687, "step": 3394 }, { "epoch": 1.9, "learning_rate": 0.00012106347897774114, "loss": 0.0511, "step": 3395 }, { "epoch": 1.9, "learning_rate": 0.00012100164880461664, "loss": 0.0927, "step": 3396 }, { "epoch": 1.9, "learning_rate": 0.00012093981863149216, "loss": 0.0544, "step": 3397 }, { "epoch": 1.9, "learning_rate": 0.00012087798845836768, "loss": 0.0746, "step": 3398 }, { "epoch": 1.91, "learning_rate": 0.00012081615828524319, "loss": 0.0636, "step": 3399 }, { "epoch": 1.91, "learning_rate": 0.0001207543281121187, "loss": 0.0849, "step": 3400 }, { "epoch": 1.91, "learning_rate": 0.00012069249793899423, "loss": 0.0599, "step": 3401 }, { "epoch": 1.91, "learning_rate": 0.00012063066776586973, "loss": 0.0872, "step": 3402 }, { "epoch": 1.91, "learning_rate": 0.00012056883759274525, "loss": 0.0794, "step": 3403 }, { "epoch": 1.91, "learning_rate": 0.00012050700741962078, "loss": 0.063, "step": 3404 }, { "epoch": 1.91, "learning_rate": 0.00012044517724649628, "loss": 0.0597, "step": 3405 }, { "epoch": 1.91, "learning_rate": 0.0001203833470733718, "loss": 0.0705, "step": 3406 }, { "epoch": 1.91, "learning_rate": 0.00012032151690024731, "loss": 0.0528, "step": 3407 }, { "epoch": 1.91, "learning_rate": 0.00012025968672712283, "loss": 0.0534, "step": 3408 }, { "epoch": 1.91, "learning_rate": 0.00012019785655399833, "loss": 0.0707, "step": 3409 }, { "epoch": 1.91, "learning_rate": 0.00012013602638087386, "loss": 0.0806, "step": 3410 }, { "epoch": 1.91, "learning_rate": 0.00012007419620774938, "loss": 0.0562, "step": 3411 }, { "epoch": 1.91, "learning_rate": 0.00012001236603462488, "loss": 0.0417, "step": 3412 }, { "epoch": 1.91, "learning_rate": 0.00011995053586150041, "loss": 0.0447, "step": 3413 }, { "epoch": 1.91, "learning_rate": 0.00011988870568837591, "loss": 0.0679, "step": 3414 }, { "epoch": 1.91, "learning_rate": 0.00011982687551525143, "loss": 0.0713, "step": 3415 }, { "epoch": 1.91, "learning_rate": 0.00011976504534212695, "loss": 0.066, "step": 3416 }, { "epoch": 1.92, "learning_rate": 0.00011970321516900246, "loss": 0.078, "step": 3417 }, { "epoch": 1.92, "learning_rate": 0.00011964138499587797, "loss": 0.0641, "step": 3418 }, { "epoch": 1.92, "learning_rate": 0.0001195795548227535, "loss": 0.0638, "step": 3419 }, { "epoch": 1.92, "learning_rate": 0.00011951772464962901, "loss": 0.0711, "step": 3420 }, { "epoch": 1.92, "learning_rate": 0.00011945589447650452, "loss": 0.0894, "step": 3421 }, { "epoch": 1.92, "learning_rate": 0.00011939406430338005, "loss": 0.0733, "step": 3422 }, { "epoch": 1.92, "learning_rate": 0.00011933223413025555, "loss": 0.0703, "step": 3423 }, { "epoch": 1.92, "learning_rate": 0.00011927040395713107, "loss": 0.0561, "step": 3424 }, { "epoch": 1.92, "learning_rate": 0.00011920857378400658, "loss": 0.065, "step": 3425 }, { "epoch": 1.92, "learning_rate": 0.0001191467436108821, "loss": 0.0377, "step": 3426 }, { "epoch": 1.92, "learning_rate": 0.00011908491343775763, "loss": 0.0941, "step": 3427 }, { "epoch": 1.92, "learning_rate": 0.00011902308326463313, "loss": 0.1043, "step": 3428 }, { "epoch": 1.92, "learning_rate": 0.00011896125309150865, "loss": 0.049, "step": 3429 }, { "epoch": 1.92, "learning_rate": 0.00011889942291838415, "loss": 0.0605, "step": 3430 }, { "epoch": 1.92, "learning_rate": 0.00011883759274525968, "loss": 0.0948, "step": 3431 }, { "epoch": 1.92, "learning_rate": 0.00011877576257213519, "loss": 0.0969, "step": 3432 }, { "epoch": 1.92, "learning_rate": 0.0001187139323990107, "loss": 0.16, "step": 3433 }, { "epoch": 1.92, "learning_rate": 0.00011865210222588622, "loss": 0.1396, "step": 3434 }, { "epoch": 1.93, "learning_rate": 0.00011859027205276174, "loss": 0.0729, "step": 3435 }, { "epoch": 1.93, "learning_rate": 0.00011852844187963727, "loss": 0.0852, "step": 3436 }, { "epoch": 1.93, "learning_rate": 0.00011846661170651277, "loss": 0.0526, "step": 3437 }, { "epoch": 1.93, "learning_rate": 0.00011840478153338828, "loss": 0.0657, "step": 3438 }, { "epoch": 1.93, "learning_rate": 0.00011834295136026379, "loss": 0.0687, "step": 3439 }, { "epoch": 1.93, "learning_rate": 0.00011828112118713932, "loss": 0.0728, "step": 3440 }, { "epoch": 1.93, "learning_rate": 0.00011821929101401482, "loss": 0.0496, "step": 3441 }, { "epoch": 1.93, "learning_rate": 0.00011815746084089035, "loss": 0.0827, "step": 3442 }, { "epoch": 1.93, "learning_rate": 0.00011809563066776587, "loss": 0.0584, "step": 3443 }, { "epoch": 1.93, "learning_rate": 0.00011803380049464137, "loss": 0.0911, "step": 3444 }, { "epoch": 1.93, "learning_rate": 0.0001179719703215169, "loss": 0.0496, "step": 3445 }, { "epoch": 1.93, "learning_rate": 0.0001179101401483924, "loss": 0.0798, "step": 3446 }, { "epoch": 1.93, "learning_rate": 0.00011784830997526792, "loss": 0.0802, "step": 3447 }, { "epoch": 1.93, "learning_rate": 0.00011778647980214342, "loss": 0.0713, "step": 3448 }, { "epoch": 1.93, "learning_rate": 0.00011772464962901895, "loss": 0.0613, "step": 3449 }, { "epoch": 1.93, "learning_rate": 0.00011766281945589446, "loss": 0.0744, "step": 3450 }, { "epoch": 1.93, "learning_rate": 0.00011760098928276999, "loss": 0.055, "step": 3451 }, { "epoch": 1.93, "learning_rate": 0.0001175391591096455, "loss": 0.0712, "step": 3452 }, { "epoch": 1.94, "learning_rate": 0.000117477328936521, "loss": 0.061, "step": 3453 }, { "epoch": 1.94, "learning_rate": 0.00011741549876339654, "loss": 0.0453, "step": 3454 }, { "epoch": 1.94, "learning_rate": 0.00011735366859027204, "loss": 0.0475, "step": 3455 }, { "epoch": 1.94, "learning_rate": 0.00011729183841714756, "loss": 0.074, "step": 3456 }, { "epoch": 1.94, "learning_rate": 0.00011723000824402306, "loss": 0.0838, "step": 3457 }, { "epoch": 1.94, "learning_rate": 0.00011716817807089859, "loss": 0.073, "step": 3458 }, { "epoch": 1.94, "learning_rate": 0.0001171063478977741, "loss": 0.082, "step": 3459 }, { "epoch": 1.94, "learning_rate": 0.00011704451772464962, "loss": 0.0682, "step": 3460 }, { "epoch": 1.94, "learning_rate": 0.00011698268755152514, "loss": 0.0761, "step": 3461 }, { "epoch": 1.94, "learning_rate": 0.00011692085737840064, "loss": 0.0378, "step": 3462 }, { "epoch": 1.94, "learning_rate": 0.00011685902720527617, "loss": 0.0706, "step": 3463 }, { "epoch": 1.94, "learning_rate": 0.00011679719703215168, "loss": 0.0804, "step": 3464 }, { "epoch": 1.94, "learning_rate": 0.00011673536685902719, "loss": 0.041, "step": 3465 }, { "epoch": 1.94, "learning_rate": 0.00011667353668590271, "loss": 0.0961, "step": 3466 }, { "epoch": 1.94, "learning_rate": 0.00011661170651277823, "loss": 0.0943, "step": 3467 }, { "epoch": 1.94, "learning_rate": 0.00011654987633965374, "loss": 0.0727, "step": 3468 }, { "epoch": 1.94, "learning_rate": 0.00011648804616652926, "loss": 0.0838, "step": 3469 }, { "epoch": 1.95, "learning_rate": 0.00011642621599340477, "loss": 0.097, "step": 3470 }, { "epoch": 1.95, "learning_rate": 0.00011636438582028028, "loss": 0.0632, "step": 3471 }, { "epoch": 1.95, "learning_rate": 0.00011630255564715581, "loss": 0.0746, "step": 3472 }, { "epoch": 1.95, "learning_rate": 0.00011624072547403131, "loss": 0.0703, "step": 3473 }, { "epoch": 1.95, "learning_rate": 0.00011617889530090683, "loss": 0.0761, "step": 3474 }, { "epoch": 1.95, "learning_rate": 0.00011611706512778236, "loss": 0.0597, "step": 3475 }, { "epoch": 1.95, "learning_rate": 0.00011605523495465786, "loss": 0.0587, "step": 3476 }, { "epoch": 1.95, "learning_rate": 0.00011599340478153338, "loss": 0.0897, "step": 3477 }, { "epoch": 1.95, "learning_rate": 0.0001159315746084089, "loss": 0.0892, "step": 3478 }, { "epoch": 1.95, "learning_rate": 0.00011586974443528441, "loss": 0.0954, "step": 3479 }, { "epoch": 1.95, "learning_rate": 0.00011580791426215991, "loss": 0.1112, "step": 3480 }, { "epoch": 1.95, "learning_rate": 0.00011574608408903544, "loss": 0.0545, "step": 3481 }, { "epoch": 1.95, "learning_rate": 0.00011568425391591096, "loss": 0.1409, "step": 3482 }, { "epoch": 1.95, "learning_rate": 0.00011562242374278646, "loss": 0.0992, "step": 3483 }, { "epoch": 1.95, "learning_rate": 0.000115560593569662, "loss": 0.2692, "step": 3484 }, { "epoch": 1.95, "learning_rate": 0.0001154987633965375, "loss": 0.1051, "step": 3485 }, { "epoch": 1.95, "learning_rate": 0.00011543693322341301, "loss": 0.0861, "step": 3486 }, { "epoch": 1.95, "learning_rate": 0.00011537510305028853, "loss": 0.0507, "step": 3487 }, { "epoch": 1.96, "learning_rate": 0.00011531327287716405, "loss": 0.0868, "step": 3488 }, { "epoch": 1.96, "learning_rate": 0.00011525144270403955, "loss": 0.0762, "step": 3489 }, { "epoch": 1.96, "learning_rate": 0.00011518961253091508, "loss": 0.0908, "step": 3490 }, { "epoch": 1.96, "learning_rate": 0.0001151277823577906, "loss": 0.0839, "step": 3491 }, { "epoch": 1.96, "learning_rate": 0.0001150659521846661, "loss": 0.0717, "step": 3492 }, { "epoch": 1.96, "learning_rate": 0.00011500412201154163, "loss": 0.0762, "step": 3493 }, { "epoch": 1.96, "learning_rate": 0.00011494229183841713, "loss": 0.076, "step": 3494 }, { "epoch": 1.96, "learning_rate": 0.00011488046166529265, "loss": 0.1009, "step": 3495 }, { "epoch": 1.96, "learning_rate": 0.00011481863149216817, "loss": 0.0609, "step": 3496 }, { "epoch": 1.96, "learning_rate": 0.00011475680131904368, "loss": 0.073, "step": 3497 }, { "epoch": 1.96, "learning_rate": 0.00011469497114591921, "loss": 0.0784, "step": 3498 }, { "epoch": 1.96, "learning_rate": 0.00011463314097279472, "loss": 0.0494, "step": 3499 }, { "epoch": 1.96, "learning_rate": 0.00011457131079967023, "loss": 0.0635, "step": 3500 }, { "epoch": 1.96, "step": 3500, "total_flos": 6.519395428891662e+18, "train_loss": 0.0, "train_runtime": 7.6753, "train_samples_per_second": 3718.173, "train_steps_per_second": 232.435 } ], "max_steps": 1784, "num_train_epochs": 1, "total_flos": 6.519395428891662e+18, "trial_name": null, "trial_params": null }