| { |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 1.0471204188481675, |
| "eval_steps": 1000, |
| "global_step": 2000, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.0, |
| "learning_rate": 0.00019998952879581153, |
| "loss": 1.1519, |
| "step": 1 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 0.00019997905759162304, |
| "loss": 1.1227, |
| "step": 2 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 0.00019996858638743456, |
| "loss": 0.8782, |
| "step": 3 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 0.00019995811518324608, |
| "loss": 0.8829, |
| "step": 4 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 0.0001999476439790576, |
| "loss": 0.6515, |
| "step": 5 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 0.0001999371727748691, |
| "loss": 0.624, |
| "step": 6 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 0.00019992670157068063, |
| "loss": 0.6735, |
| "step": 7 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 0.00019991623036649217, |
| "loss": 0.5841, |
| "step": 8 |
| }, |
| { |
| "epoch": 0.0, |
| "learning_rate": 0.00019990575916230366, |
| "loss": 0.5819, |
| "step": 9 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.00019989528795811518, |
| "loss": 0.5001, |
| "step": 10 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.00019988481675392672, |
| "loss": 0.5033, |
| "step": 11 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.00019987434554973824, |
| "loss": 0.4789, |
| "step": 12 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.00019986387434554973, |
| "loss": 0.4262, |
| "step": 13 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.00019985340314136128, |
| "loss": 0.4104, |
| "step": 14 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.0001998429319371728, |
| "loss": 0.4281, |
| "step": 15 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.00019983246073298428, |
| "loss": 0.3888, |
| "step": 16 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.00019982198952879583, |
| "loss": 0.3937, |
| "step": 17 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.00019981151832460734, |
| "loss": 0.3207, |
| "step": 18 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.00019980104712041886, |
| "loss": 0.4089, |
| "step": 19 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.00019979057591623038, |
| "loss": 0.3359, |
| "step": 20 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.0001997801047120419, |
| "loss": 0.3908, |
| "step": 21 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.0001997696335078534, |
| "loss": 0.4128, |
| "step": 22 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.00019975916230366493, |
| "loss": 0.4216, |
| "step": 23 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.00019974869109947645, |
| "loss": 0.3463, |
| "step": 24 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.00019973821989528796, |
| "loss": 0.3821, |
| "step": 25 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.00019972774869109948, |
| "loss": 0.2799, |
| "step": 26 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.000199717277486911, |
| "loss": 0.342, |
| "step": 27 |
| }, |
| { |
| "epoch": 0.01, |
| "learning_rate": 0.00019970680628272251, |
| "loss": 0.3226, |
| "step": 28 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.00019969633507853406, |
| "loss": 0.3278, |
| "step": 29 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.00019968586387434555, |
| "loss": 0.3023, |
| "step": 30 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.00019967539267015707, |
| "loss": 0.3994, |
| "step": 31 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.0001996649214659686, |
| "loss": 0.2725, |
| "step": 32 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.0001996544502617801, |
| "loss": 0.2774, |
| "step": 33 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.00019964397905759162, |
| "loss": 0.2764, |
| "step": 34 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.00019963350785340316, |
| "loss": 0.2594, |
| "step": 35 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.00019962303664921468, |
| "loss": 0.3228, |
| "step": 36 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.00019961256544502617, |
| "loss": 0.2873, |
| "step": 37 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.0001996020942408377, |
| "loss": 0.2992, |
| "step": 38 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.00019959162303664923, |
| "loss": 0.1936, |
| "step": 39 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.00019958115183246072, |
| "loss": 0.2057, |
| "step": 40 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.00019957068062827226, |
| "loss": 0.2435, |
| "step": 41 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.00019956020942408378, |
| "loss": 0.2612, |
| "step": 42 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.0001995497382198953, |
| "loss": 0.256, |
| "step": 43 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.00019953926701570682, |
| "loss": 0.3077, |
| "step": 44 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.00019952879581151833, |
| "loss": 0.1841, |
| "step": 45 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.00019951832460732985, |
| "loss": 0.2836, |
| "step": 46 |
| }, |
| { |
| "epoch": 0.02, |
| "learning_rate": 0.00019950785340314137, |
| "loss": 0.2789, |
| "step": 47 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.00019949738219895288, |
| "loss": 0.1964, |
| "step": 48 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.0001994869109947644, |
| "loss": 0.1811, |
| "step": 49 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.00019947643979057592, |
| "loss": 0.3166, |
| "step": 50 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.00019946596858638743, |
| "loss": 0.2803, |
| "step": 51 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.00019945549738219895, |
| "loss": 0.2589, |
| "step": 52 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.0001994450261780105, |
| "loss": 0.3395, |
| "step": 53 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.00019943455497382199, |
| "loss": 0.2242, |
| "step": 54 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.0001994240837696335, |
| "loss": 0.2891, |
| "step": 55 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.00019941361256544505, |
| "loss": 0.2428, |
| "step": 56 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.00019940314136125656, |
| "loss": 0.2437, |
| "step": 57 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.00019939267015706805, |
| "loss": 0.2419, |
| "step": 58 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.0001993821989528796, |
| "loss": 0.2362, |
| "step": 59 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.00019937172774869112, |
| "loss": 0.2638, |
| "step": 60 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.0001993612565445026, |
| "loss": 0.1832, |
| "step": 61 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.00019935078534031415, |
| "loss": 0.1869, |
| "step": 62 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.00019934031413612567, |
| "loss": 0.2495, |
| "step": 63 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.00019932984293193718, |
| "loss": 0.2249, |
| "step": 64 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.0001993193717277487, |
| "loss": 0.1894, |
| "step": 65 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 0.00019930890052356022, |
| "loss": 0.2461, |
| "step": 66 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.00019929842931937174, |
| "loss": 0.2523, |
| "step": 67 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.00019928795811518325, |
| "loss": 0.1984, |
| "step": 68 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.00019927748691099477, |
| "loss": 0.1535, |
| "step": 69 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.00019926701570680629, |
| "loss": 0.1559, |
| "step": 70 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.0001992565445026178, |
| "loss": 0.2654, |
| "step": 71 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.00019924607329842932, |
| "loss": 0.2284, |
| "step": 72 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.00019923560209424084, |
| "loss": 0.2399, |
| "step": 73 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.00019922513089005238, |
| "loss": 0.1787, |
| "step": 74 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.00019921465968586387, |
| "loss": 0.21, |
| "step": 75 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.0001992041884816754, |
| "loss": 0.2099, |
| "step": 76 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.00019919371727748693, |
| "loss": 0.2181, |
| "step": 77 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.00019918324607329842, |
| "loss": 0.2419, |
| "step": 78 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.00019917277486910997, |
| "loss": 0.1828, |
| "step": 79 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.00019916230366492148, |
| "loss": 0.1793, |
| "step": 80 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.000199151832460733, |
| "loss": 0.256, |
| "step": 81 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.00019914136125654452, |
| "loss": 0.1811, |
| "step": 82 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.00019913089005235604, |
| "loss": 0.1897, |
| "step": 83 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.00019912041884816755, |
| "loss": 0.2366, |
| "step": 84 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 0.00019910994764397907, |
| "loss": 0.2209, |
| "step": 85 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.0001990994764397906, |
| "loss": 0.2023, |
| "step": 86 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.0001990890052356021, |
| "loss": 0.2068, |
| "step": 87 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.00019907853403141362, |
| "loss": 0.2392, |
| "step": 88 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.00019906806282722514, |
| "loss": 0.2348, |
| "step": 89 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.00019905759162303665, |
| "loss": 0.2099, |
| "step": 90 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.0001990471204188482, |
| "loss": 0.1773, |
| "step": 91 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.0001990366492146597, |
| "loss": 0.192, |
| "step": 92 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.0001990261780104712, |
| "loss": 0.2178, |
| "step": 93 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.00019901570680628275, |
| "loss": 0.1702, |
| "step": 94 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.00019900523560209424, |
| "loss": 0.2278, |
| "step": 95 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.00019899476439790576, |
| "loss": 0.1716, |
| "step": 96 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.0001989842931937173, |
| "loss": 0.2223, |
| "step": 97 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.00019897382198952882, |
| "loss": 0.2069, |
| "step": 98 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.0001989633507853403, |
| "loss": 0.1805, |
| "step": 99 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.00019895287958115185, |
| "loss": 0.1748, |
| "step": 100 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.00019894240837696337, |
| "loss": 0.2392, |
| "step": 101 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.0001989319371727749, |
| "loss": 0.164, |
| "step": 102 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.0001989214659685864, |
| "loss": 0.1483, |
| "step": 103 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.00019891099476439792, |
| "loss": 0.1653, |
| "step": 104 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 0.00019890052356020944, |
| "loss": 0.2162, |
| "step": 105 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.00019889005235602096, |
| "loss": 0.1487, |
| "step": 106 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.00019887958115183247, |
| "loss": 0.1585, |
| "step": 107 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.000198869109947644, |
| "loss": 0.196, |
| "step": 108 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.0001988586387434555, |
| "loss": 0.1954, |
| "step": 109 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.00019884816753926702, |
| "loss": 0.1944, |
| "step": 110 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.00019883769633507854, |
| "loss": 0.1297, |
| "step": 111 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.00019882722513089008, |
| "loss": 0.1807, |
| "step": 112 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.00019881675392670157, |
| "loss": 0.1871, |
| "step": 113 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.0001988062827225131, |
| "loss": 0.1721, |
| "step": 114 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.00019879581151832464, |
| "loss": 0.1688, |
| "step": 115 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.00019878534031413613, |
| "loss": 0.1751, |
| "step": 116 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.00019877486910994764, |
| "loss": 0.176, |
| "step": 117 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.0001987643979057592, |
| "loss": 0.1694, |
| "step": 118 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.0001987539267015707, |
| "loss": 0.1609, |
| "step": 119 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.0001987434554973822, |
| "loss": 0.1667, |
| "step": 120 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.00019873298429319374, |
| "loss": 0.1646, |
| "step": 121 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.00019872251308900526, |
| "loss": 0.1906, |
| "step": 122 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.00019871204188481675, |
| "loss": 0.1945, |
| "step": 123 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 0.0001987015706806283, |
| "loss": 0.1606, |
| "step": 124 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.0001986910994764398, |
| "loss": 0.1613, |
| "step": 125 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.00019868062827225132, |
| "loss": 0.164, |
| "step": 126 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.00019867015706806284, |
| "loss": 0.1585, |
| "step": 127 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.00019865968586387436, |
| "loss": 0.2032, |
| "step": 128 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.00019864921465968588, |
| "loss": 0.1667, |
| "step": 129 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.0001986387434554974, |
| "loss": 0.1539, |
| "step": 130 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.0001986282722513089, |
| "loss": 0.1746, |
| "step": 131 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.00019861780104712043, |
| "loss": 0.1678, |
| "step": 132 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.00019860732984293194, |
| "loss": 0.1924, |
| "step": 133 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.00019859685863874346, |
| "loss": 0.1273, |
| "step": 134 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.00019858638743455498, |
| "loss": 0.1732, |
| "step": 135 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.00019857591623036652, |
| "loss": 0.1503, |
| "step": 136 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.000198565445026178, |
| "loss": 0.1496, |
| "step": 137 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.00019855497382198953, |
| "loss": 0.1729, |
| "step": 138 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.00019854450261780107, |
| "loss": 0.1609, |
| "step": 139 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.00019853403141361256, |
| "loss": 0.1875, |
| "step": 140 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.00019852356020942408, |
| "loss": 0.1444, |
| "step": 141 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.00019851308900523562, |
| "loss": 0.1881, |
| "step": 142 |
| }, |
| { |
| "epoch": 0.07, |
| "learning_rate": 0.00019850261780104714, |
| "loss": 0.1946, |
| "step": 143 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.00019849214659685863, |
| "loss": 0.1391, |
| "step": 144 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.00019848167539267018, |
| "loss": 0.1387, |
| "step": 145 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.0001984712041884817, |
| "loss": 0.1709, |
| "step": 146 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.0001984607329842932, |
| "loss": 0.1557, |
| "step": 147 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.00019845026178010473, |
| "loss": 0.1937, |
| "step": 148 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.00019843979057591624, |
| "loss": 0.1931, |
| "step": 149 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.00019842931937172776, |
| "loss": 0.1904, |
| "step": 150 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.00019841884816753928, |
| "loss": 0.1934, |
| "step": 151 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.0001984083769633508, |
| "loss": 0.1969, |
| "step": 152 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.0001983979057591623, |
| "loss": 0.1735, |
| "step": 153 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.00019838743455497383, |
| "loss": 0.1638, |
| "step": 154 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.00019837696335078535, |
| "loss": 0.1631, |
| "step": 155 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.00019836649214659686, |
| "loss": 0.1526, |
| "step": 156 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.00019835602094240838, |
| "loss": 0.203, |
| "step": 157 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.0001983455497382199, |
| "loss": 0.142, |
| "step": 158 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.00019833507853403141, |
| "loss": 0.174, |
| "step": 159 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.00019832460732984296, |
| "loss": 0.1416, |
| "step": 160 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.00019831413612565445, |
| "loss": 0.1407, |
| "step": 161 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 0.00019830366492146597, |
| "loss": 0.1908, |
| "step": 162 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.0001982931937172775, |
| "loss": 0.1726, |
| "step": 163 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.00019828272251308903, |
| "loss": 0.182, |
| "step": 164 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.00019827225130890052, |
| "loss": 0.1267, |
| "step": 165 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.00019826178010471206, |
| "loss": 0.1385, |
| "step": 166 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.00019825130890052358, |
| "loss": 0.1471, |
| "step": 167 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.00019824083769633507, |
| "loss": 0.1652, |
| "step": 168 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.0001982303664921466, |
| "loss": 0.1164, |
| "step": 169 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.00019821989528795813, |
| "loss": 0.1326, |
| "step": 170 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.00019820942408376965, |
| "loss": 0.1566, |
| "step": 171 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.00019819895287958116, |
| "loss": 0.1364, |
| "step": 172 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.00019818848167539268, |
| "loss": 0.1707, |
| "step": 173 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.0001981780104712042, |
| "loss": 0.1316, |
| "step": 174 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.00019816753926701571, |
| "loss": 0.1385, |
| "step": 175 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.00019815706806282723, |
| "loss": 0.1817, |
| "step": 176 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.00019814659685863875, |
| "loss": 0.1872, |
| "step": 177 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.00019813612565445027, |
| "loss": 0.1673, |
| "step": 178 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.00019812565445026178, |
| "loss": 0.1684, |
| "step": 179 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.0001981151832460733, |
| "loss": 0.1618, |
| "step": 180 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 0.00019810471204188484, |
| "loss": 0.1534, |
| "step": 181 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.00019809424083769633, |
| "loss": 0.143, |
| "step": 182 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.00019808376963350785, |
| "loss": 0.1766, |
| "step": 183 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.0001980732984293194, |
| "loss": 0.1725, |
| "step": 184 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.00019806282722513089, |
| "loss": 0.1215, |
| "step": 185 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.0001980523560209424, |
| "loss": 0.1472, |
| "step": 186 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.00019804188481675395, |
| "loss": 0.1514, |
| "step": 187 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.00019803141361256546, |
| "loss": 0.1604, |
| "step": 188 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.00019802094240837695, |
| "loss": 0.1396, |
| "step": 189 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.0001980104712041885, |
| "loss": 0.1377, |
| "step": 190 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.00019800000000000002, |
| "loss": 0.2043, |
| "step": 191 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.0001979895287958115, |
| "loss": 0.171, |
| "step": 192 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.00019797905759162305, |
| "loss": 0.1071, |
| "step": 193 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.00019796858638743457, |
| "loss": 0.1506, |
| "step": 194 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.00019795811518324608, |
| "loss": 0.141, |
| "step": 195 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.0001979476439790576, |
| "loss": 0.1679, |
| "step": 196 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.00019793717277486912, |
| "loss": 0.166, |
| "step": 197 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.00019792670157068063, |
| "loss": 0.1489, |
| "step": 198 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.00019791623036649215, |
| "loss": 0.1344, |
| "step": 199 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 0.00019790575916230367, |
| "loss": 0.1666, |
| "step": 200 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.00019789528795811519, |
| "loss": 0.1716, |
| "step": 201 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.0001978848167539267, |
| "loss": 0.1584, |
| "step": 202 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.00019787434554973822, |
| "loss": 0.1696, |
| "step": 203 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.00019786387434554974, |
| "loss": 0.1488, |
| "step": 204 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.00019785340314136128, |
| "loss": 0.198, |
| "step": 205 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.00019784293193717277, |
| "loss": 0.1237, |
| "step": 206 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.0001978324607329843, |
| "loss": 0.2552, |
| "step": 207 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.00019782198952879583, |
| "loss": 0.1742, |
| "step": 208 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.00019781151832460735, |
| "loss": 0.1197, |
| "step": 209 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.00019780104712041884, |
| "loss": 0.1325, |
| "step": 210 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.00019779057591623038, |
| "loss": 0.1693, |
| "step": 211 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.0001977801047120419, |
| "loss": 0.1146, |
| "step": 212 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.0001977696335078534, |
| "loss": 0.2437, |
| "step": 213 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.00019775916230366494, |
| "loss": 0.1061, |
| "step": 214 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.00019774869109947645, |
| "loss": 0.1589, |
| "step": 215 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.00019773821989528797, |
| "loss": 0.1678, |
| "step": 216 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.00019772774869109949, |
| "loss": 0.1557, |
| "step": 217 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.000197717277486911, |
| "loss": 0.1725, |
| "step": 218 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 0.00019770680628272252, |
| "loss": 0.1609, |
| "step": 219 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.00019769633507853404, |
| "loss": 0.1542, |
| "step": 220 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.00019768586387434555, |
| "loss": 0.0978, |
| "step": 221 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.00019767539267015707, |
| "loss": 0.1271, |
| "step": 222 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.0001976649214659686, |
| "loss": 0.1559, |
| "step": 223 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.0001976544502617801, |
| "loss": 0.1656, |
| "step": 224 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.00019764397905759162, |
| "loss": 0.165, |
| "step": 225 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.00019763350785340317, |
| "loss": 0.2254, |
| "step": 226 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.00019762303664921466, |
| "loss": 0.1921, |
| "step": 227 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.00019761256544502617, |
| "loss": 0.1527, |
| "step": 228 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.00019760209424083772, |
| "loss": 0.1932, |
| "step": 229 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.0001975916230366492, |
| "loss": 0.1426, |
| "step": 230 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.00019758115183246073, |
| "loss": 0.1112, |
| "step": 231 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.00019757068062827227, |
| "loss": 0.1909, |
| "step": 232 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.0001975602094240838, |
| "loss": 0.1682, |
| "step": 233 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.0001975497382198953, |
| "loss": 0.1676, |
| "step": 234 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.00019753926701570682, |
| "loss": 0.16, |
| "step": 235 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.00019752879581151834, |
| "loss": 0.1849, |
| "step": 236 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.00019751832460732985, |
| "loss": 0.2124, |
| "step": 237 |
| }, |
| { |
| "epoch": 0.12, |
| "learning_rate": 0.00019750785340314137, |
| "loss": 0.1558, |
| "step": 238 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.0001974973821989529, |
| "loss": 0.1538, |
| "step": 239 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.0001974869109947644, |
| "loss": 0.1709, |
| "step": 240 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.00019747643979057592, |
| "loss": 0.1998, |
| "step": 241 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.00019746596858638744, |
| "loss": 0.1544, |
| "step": 242 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.00019745549738219898, |
| "loss": 0.1634, |
| "step": 243 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.00019744502617801047, |
| "loss": 0.1367, |
| "step": 244 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.000197434554973822, |
| "loss": 0.1651, |
| "step": 245 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.00019742408376963354, |
| "loss": 0.1257, |
| "step": 246 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.00019741361256544503, |
| "loss": 0.1465, |
| "step": 247 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.00019740314136125654, |
| "loss": 0.1452, |
| "step": 248 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.0001973926701570681, |
| "loss": 0.1399, |
| "step": 249 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.0001973821989528796, |
| "loss": 0.2168, |
| "step": 250 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.0001973717277486911, |
| "loss": 0.1125, |
| "step": 251 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.00019736125654450264, |
| "loss": 0.1583, |
| "step": 252 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.00019735078534031416, |
| "loss": 0.1561, |
| "step": 253 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.00019734031413612567, |
| "loss": 0.1661, |
| "step": 254 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.0001973298429319372, |
| "loss": 0.1498, |
| "step": 255 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.0001973193717277487, |
| "loss": 0.1538, |
| "step": 256 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 0.00019730890052356022, |
| "loss": 0.1726, |
| "step": 257 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.00019729842931937174, |
| "loss": 0.1527, |
| "step": 258 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.00019728795811518326, |
| "loss": 0.1103, |
| "step": 259 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.00019727748691099477, |
| "loss": 0.1695, |
| "step": 260 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.0001972670157068063, |
| "loss": 0.1415, |
| "step": 261 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.0001972565445026178, |
| "loss": 0.1714, |
| "step": 262 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.00019724607329842933, |
| "loss": 0.1367, |
| "step": 263 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.00019723560209424087, |
| "loss": 0.1394, |
| "step": 264 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.00019722513089005236, |
| "loss": 0.1384, |
| "step": 265 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.00019721465968586388, |
| "loss": 0.1455, |
| "step": 266 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.00019720418848167542, |
| "loss": 0.1535, |
| "step": 267 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.0001971937172774869, |
| "loss": 0.0964, |
| "step": 268 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.00019718324607329843, |
| "loss": 0.1877, |
| "step": 269 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.00019717277486910997, |
| "loss": 0.1508, |
| "step": 270 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.0001971623036649215, |
| "loss": 0.1698, |
| "step": 271 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.00019715183246073298, |
| "loss": 0.1473, |
| "step": 272 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.00019714136125654452, |
| "loss": 0.1403, |
| "step": 273 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.00019713089005235604, |
| "loss": 0.1342, |
| "step": 274 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.00019712041884816753, |
| "loss": 0.1209, |
| "step": 275 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 0.00019710994764397908, |
| "loss": 0.1566, |
| "step": 276 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.0001970994764397906, |
| "loss": 0.1843, |
| "step": 277 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.0001970890052356021, |
| "loss": 0.1659, |
| "step": 278 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.00019707853403141363, |
| "loss": 0.2108, |
| "step": 279 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.00019706806282722514, |
| "loss": 0.1637, |
| "step": 280 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.00019705759162303666, |
| "loss": 0.1498, |
| "step": 281 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.00019704712041884818, |
| "loss": 0.1766, |
| "step": 282 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.0001970366492146597, |
| "loss": 0.1253, |
| "step": 283 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.0001970261780104712, |
| "loss": 0.1062, |
| "step": 284 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.00019701570680628273, |
| "loss": 0.1692, |
| "step": 285 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.00019700523560209425, |
| "loss": 0.1443, |
| "step": 286 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.00019699476439790576, |
| "loss": 0.1504, |
| "step": 287 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.0001969842931937173, |
| "loss": 0.1067, |
| "step": 288 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.0001969738219895288, |
| "loss": 0.1141, |
| "step": 289 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.00019696335078534031, |
| "loss": 0.1528, |
| "step": 290 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.00019695287958115186, |
| "loss": 0.1302, |
| "step": 291 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.00019694240837696335, |
| "loss": 0.1246, |
| "step": 292 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.00019693193717277487, |
| "loss": 0.1629, |
| "step": 293 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.0001969214659685864, |
| "loss": 0.1655, |
| "step": 294 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.00019691099476439793, |
| "loss": 0.1417, |
| "step": 295 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 0.00019690052356020942, |
| "loss": 0.1307, |
| "step": 296 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.00019689005235602096, |
| "loss": 0.1313, |
| "step": 297 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.00019687958115183248, |
| "loss": 0.1644, |
| "step": 298 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.000196869109947644, |
| "loss": 0.1447, |
| "step": 299 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.0001968586387434555, |
| "loss": 0.1598, |
| "step": 300 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.00019684816753926703, |
| "loss": 0.154, |
| "step": 301 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.00019683769633507855, |
| "loss": 0.1458, |
| "step": 302 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.00019682722513089006, |
| "loss": 0.1189, |
| "step": 303 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.00019681675392670158, |
| "loss": 0.1681, |
| "step": 304 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.0001968062827225131, |
| "loss": 0.1466, |
| "step": 305 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.00019679581151832461, |
| "loss": 0.101, |
| "step": 306 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.00019678534031413613, |
| "loss": 0.1155, |
| "step": 307 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.00019677486910994765, |
| "loss": 0.133, |
| "step": 308 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.00019676439790575917, |
| "loss": 0.1563, |
| "step": 309 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.00019675392670157068, |
| "loss": 0.1737, |
| "step": 310 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.0001967434554973822, |
| "loss": 0.1427, |
| "step": 311 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.00019673298429319374, |
| "loss": 0.1533, |
| "step": 312 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.00019672251308900523, |
| "loss": 0.147, |
| "step": 313 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.00019671204188481675, |
| "loss": 0.1135, |
| "step": 314 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 0.0001967015706806283, |
| "loss": 0.1349, |
| "step": 315 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.0001966910994764398, |
| "loss": 0.1195, |
| "step": 316 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.0001966806282722513, |
| "loss": 0.1526, |
| "step": 317 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.00019667015706806285, |
| "loss": 0.1185, |
| "step": 318 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.00019665968586387436, |
| "loss": 0.179, |
| "step": 319 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.00019664921465968585, |
| "loss": 0.1214, |
| "step": 320 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.0001966387434554974, |
| "loss": 0.1762, |
| "step": 321 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.00019662827225130891, |
| "loss": 0.1353, |
| "step": 322 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.00019661780104712043, |
| "loss": 0.1528, |
| "step": 323 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.00019660732984293195, |
| "loss": 0.1553, |
| "step": 324 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.00019659685863874347, |
| "loss": 0.1623, |
| "step": 325 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.00019658638743455498, |
| "loss": 0.1716, |
| "step": 326 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.0001965759162303665, |
| "loss": 0.1524, |
| "step": 327 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.00019656544502617802, |
| "loss": 0.1452, |
| "step": 328 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.00019655497382198953, |
| "loss": 0.1315, |
| "step": 329 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.00019654450261780105, |
| "loss": 0.1714, |
| "step": 330 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.00019653403141361257, |
| "loss": 0.1629, |
| "step": 331 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.00019652356020942409, |
| "loss": 0.133, |
| "step": 332 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.00019651308900523563, |
| "loss": 0.0879, |
| "step": 333 |
| }, |
| { |
| "epoch": 0.17, |
| "learning_rate": 0.00019650261780104712, |
| "loss": 0.1551, |
| "step": 334 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.00019649214659685864, |
| "loss": 0.123, |
| "step": 335 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.00019648167539267018, |
| "loss": 0.1038, |
| "step": 336 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.00019647120418848167, |
| "loss": 0.125, |
| "step": 337 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.0001964607329842932, |
| "loss": 0.1686, |
| "step": 338 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.00019645026178010473, |
| "loss": 0.1429, |
| "step": 339 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.00019643979057591625, |
| "loss": 0.1463, |
| "step": 340 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.00019642931937172774, |
| "loss": 0.1758, |
| "step": 341 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.00019641884816753928, |
| "loss": 0.1482, |
| "step": 342 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.0001964083769633508, |
| "loss": 0.1377, |
| "step": 343 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.0001963979057591623, |
| "loss": 0.1141, |
| "step": 344 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.00019638743455497383, |
| "loss": 0.1169, |
| "step": 345 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.00019637696335078535, |
| "loss": 0.1179, |
| "step": 346 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.00019636649214659687, |
| "loss": 0.142, |
| "step": 347 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.00019635602094240839, |
| "loss": 0.113, |
| "step": 348 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.0001963455497382199, |
| "loss": 0.1322, |
| "step": 349 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.00019633507853403142, |
| "loss": 0.1371, |
| "step": 350 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.00019632460732984294, |
| "loss": 0.159, |
| "step": 351 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.00019631413612565445, |
| "loss": 0.1235, |
| "step": 352 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 0.00019630366492146597, |
| "loss": 0.1276, |
| "step": 353 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.0001962931937172775, |
| "loss": 0.1189, |
| "step": 354 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.000196282722513089, |
| "loss": 0.1245, |
| "step": 355 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.00019627225130890052, |
| "loss": 0.1526, |
| "step": 356 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.00019626178010471207, |
| "loss": 0.1547, |
| "step": 357 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.00019625130890052356, |
| "loss": 0.1131, |
| "step": 358 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.00019624083769633507, |
| "loss": 0.1762, |
| "step": 359 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.00019623036649214662, |
| "loss": 0.1065, |
| "step": 360 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.00019621989528795814, |
| "loss": 0.1312, |
| "step": 361 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.00019620942408376963, |
| "loss": 0.1321, |
| "step": 362 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.00019619895287958117, |
| "loss": 0.1548, |
| "step": 363 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.00019618848167539269, |
| "loss": 0.1693, |
| "step": 364 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.00019617801047120418, |
| "loss": 0.131, |
| "step": 365 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.00019616753926701572, |
| "loss": 0.1382, |
| "step": 366 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.00019615706806282724, |
| "loss": 0.1468, |
| "step": 367 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.00019614659685863875, |
| "loss": 0.1039, |
| "step": 368 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.00019613612565445027, |
| "loss": 0.1748, |
| "step": 369 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.0001961256544502618, |
| "loss": 0.1541, |
| "step": 370 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.0001961151832460733, |
| "loss": 0.1272, |
| "step": 371 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 0.00019610471204188482, |
| "loss": 0.1169, |
| "step": 372 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.00019609424083769634, |
| "loss": 0.1551, |
| "step": 373 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.00019608376963350786, |
| "loss": 0.1408, |
| "step": 374 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.00019607329842931937, |
| "loss": 0.1251, |
| "step": 375 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.0001960628272251309, |
| "loss": 0.1612, |
| "step": 376 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.0001960523560209424, |
| "loss": 0.1496, |
| "step": 377 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.00019604188481675395, |
| "loss": 0.1382, |
| "step": 378 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.00019603141361256544, |
| "loss": 0.1535, |
| "step": 379 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.00019602094240837696, |
| "loss": 0.117, |
| "step": 380 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.0001960104712041885, |
| "loss": 0.1572, |
| "step": 381 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.000196, |
| "loss": 0.1325, |
| "step": 382 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.0001959895287958115, |
| "loss": 0.132, |
| "step": 383 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.00019597905759162305, |
| "loss": 0.1435, |
| "step": 384 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.00019596858638743457, |
| "loss": 0.121, |
| "step": 385 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.0001959581151832461, |
| "loss": 0.124, |
| "step": 386 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.0001959476439790576, |
| "loss": 0.1731, |
| "step": 387 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.00019593717277486912, |
| "loss": 0.1406, |
| "step": 388 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.00019592670157068064, |
| "loss": 0.1184, |
| "step": 389 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.00019591623036649216, |
| "loss": 0.1412, |
| "step": 390 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 0.00019590575916230367, |
| "loss": 0.1347, |
| "step": 391 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.0001958952879581152, |
| "loss": 0.1258, |
| "step": 392 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.0001958848167539267, |
| "loss": 0.1431, |
| "step": 393 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.00019587434554973823, |
| "loss": 0.1317, |
| "step": 394 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.00019586387434554977, |
| "loss": 0.1482, |
| "step": 395 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.00019585340314136126, |
| "loss": 0.1536, |
| "step": 396 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.00019584293193717278, |
| "loss": 0.1724, |
| "step": 397 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.00019583246073298432, |
| "loss": 0.1783, |
| "step": 398 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.0001958219895287958, |
| "loss": 0.101, |
| "step": 399 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.00019581151832460733, |
| "loss": 0.1259, |
| "step": 400 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.00019580104712041887, |
| "loss": 0.1056, |
| "step": 401 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.0001957905759162304, |
| "loss": 0.139, |
| "step": 402 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.00019578010471204188, |
| "loss": 0.1672, |
| "step": 403 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.00019576963350785342, |
| "loss": 0.161, |
| "step": 404 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.00019575916230366494, |
| "loss": 0.1206, |
| "step": 405 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.00019574869109947646, |
| "loss": 0.163, |
| "step": 406 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.00019573821989528797, |
| "loss": 0.1235, |
| "step": 407 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.0001957277486910995, |
| "loss": 0.1271, |
| "step": 408 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.000195717277486911, |
| "loss": 0.1268, |
| "step": 409 |
| }, |
| { |
| "epoch": 0.21, |
| "learning_rate": 0.00019570680628272253, |
| "loss": 0.1321, |
| "step": 410 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.00019569633507853404, |
| "loss": 0.136, |
| "step": 411 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.00019568586387434556, |
| "loss": 0.1386, |
| "step": 412 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.00019567539267015708, |
| "loss": 0.1326, |
| "step": 413 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.0001956649214659686, |
| "loss": 0.1773, |
| "step": 414 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.0001956544502617801, |
| "loss": 0.1088, |
| "step": 415 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.00019564397905759163, |
| "loss": 0.1526, |
| "step": 416 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.00019563350785340315, |
| "loss": 0.1338, |
| "step": 417 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.00019562303664921466, |
| "loss": 0.1192, |
| "step": 418 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.0001956125654450262, |
| "loss": 0.1376, |
| "step": 419 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.0001956020942408377, |
| "loss": 0.166, |
| "step": 420 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.00019559162303664921, |
| "loss": 0.1217, |
| "step": 421 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.00019558115183246076, |
| "loss": 0.1602, |
| "step": 422 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.00019557068062827228, |
| "loss": 0.1408, |
| "step": 423 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.00019556020942408377, |
| "loss": 0.1316, |
| "step": 424 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.0001955497382198953, |
| "loss": 0.1407, |
| "step": 425 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.00019553926701570683, |
| "loss": 0.2199, |
| "step": 426 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.00019552879581151832, |
| "loss": 0.0953, |
| "step": 427 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.00019551832460732986, |
| "loss": 0.1232, |
| "step": 428 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 0.00019550785340314138, |
| "loss": 0.1054, |
| "step": 429 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.0001954973821989529, |
| "loss": 0.1441, |
| "step": 430 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.0001954869109947644, |
| "loss": 0.0949, |
| "step": 431 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.00019547643979057593, |
| "loss": 0.133, |
| "step": 432 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.00019546596858638745, |
| "loss": 0.1957, |
| "step": 433 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.00019545549738219896, |
| "loss": 0.1277, |
| "step": 434 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.00019544502617801048, |
| "loss": 0.1512, |
| "step": 435 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.000195434554973822, |
| "loss": 0.1316, |
| "step": 436 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.00019542408376963351, |
| "loss": 0.1142, |
| "step": 437 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.00019541361256544503, |
| "loss": 0.1233, |
| "step": 438 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.00019540314136125655, |
| "loss": 0.1651, |
| "step": 439 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.0001953926701570681, |
| "loss": 0.1288, |
| "step": 440 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.00019538219895287958, |
| "loss": 0.1289, |
| "step": 441 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.0001953717277486911, |
| "loss": 0.1408, |
| "step": 442 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.00019536125654450264, |
| "loss": 0.1002, |
| "step": 443 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.00019535078534031413, |
| "loss": 0.1535, |
| "step": 444 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.00019534031413612565, |
| "loss": 0.1113, |
| "step": 445 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.0001953298429319372, |
| "loss": 0.1057, |
| "step": 446 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.0001953193717277487, |
| "loss": 0.1263, |
| "step": 447 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 0.0001953089005235602, |
| "loss": 0.1393, |
| "step": 448 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019529842931937175, |
| "loss": 0.1484, |
| "step": 449 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019528795811518326, |
| "loss": 0.1279, |
| "step": 450 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019527748691099478, |
| "loss": 0.1451, |
| "step": 451 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.0001952670157068063, |
| "loss": 0.1597, |
| "step": 452 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019525654450261781, |
| "loss": 0.1348, |
| "step": 453 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019524607329842933, |
| "loss": 0.1762, |
| "step": 454 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019523560209424085, |
| "loss": 0.1071, |
| "step": 455 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019522513089005237, |
| "loss": 0.1425, |
| "step": 456 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019521465968586388, |
| "loss": 0.1679, |
| "step": 457 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.0001952041884816754, |
| "loss": 0.1359, |
| "step": 458 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019519371727748692, |
| "loss": 0.145, |
| "step": 459 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019518324607329843, |
| "loss": 0.1133, |
| "step": 460 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019517277486910995, |
| "loss": 0.13, |
| "step": 461 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019516230366492147, |
| "loss": 0.1419, |
| "step": 462 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019515183246073299, |
| "loss": 0.1229, |
| "step": 463 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019514136125654453, |
| "loss": 0.1032, |
| "step": 464 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019513089005235602, |
| "loss": 0.1608, |
| "step": 465 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019512041884816754, |
| "loss": 0.1794, |
| "step": 466 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 0.00019510994764397908, |
| "loss": 0.1275, |
| "step": 467 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.0001950994764397906, |
| "loss": 0.1371, |
| "step": 468 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.0001950890052356021, |
| "loss": 0.182, |
| "step": 469 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.00019507853403141363, |
| "loss": 0.1609, |
| "step": 470 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.00019506806282722515, |
| "loss": 0.1328, |
| "step": 471 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.00019505759162303664, |
| "loss": 0.0937, |
| "step": 472 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.00019504712041884818, |
| "loss": 0.1486, |
| "step": 473 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.0001950366492146597, |
| "loss": 0.12, |
| "step": 474 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.00019502617801047122, |
| "loss": 0.1577, |
| "step": 475 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.00019501570680628273, |
| "loss": 0.1106, |
| "step": 476 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.00019500523560209425, |
| "loss": 0.1257, |
| "step": 477 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.00019499476439790577, |
| "loss": 0.1039, |
| "step": 478 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.00019498429319371729, |
| "loss": 0.1877, |
| "step": 479 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.0001949738219895288, |
| "loss": 0.1447, |
| "step": 480 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.00019496335078534032, |
| "loss": 0.1302, |
| "step": 481 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.00019495287958115184, |
| "loss": 0.1109, |
| "step": 482 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.00019494240837696335, |
| "loss": 0.1162, |
| "step": 483 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.00019493193717277487, |
| "loss": 0.1436, |
| "step": 484 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.00019492146596858642, |
| "loss": 0.1417, |
| "step": 485 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.0001949109947643979, |
| "loss": 0.1233, |
| "step": 486 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 0.00019490052356020942, |
| "loss": 0.1224, |
| "step": 487 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019489005235602097, |
| "loss": 0.1283, |
| "step": 488 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019487958115183246, |
| "loss": 0.1297, |
| "step": 489 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019486910994764397, |
| "loss": 0.16, |
| "step": 490 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019485863874345552, |
| "loss": 0.1202, |
| "step": 491 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019484816753926703, |
| "loss": 0.1227, |
| "step": 492 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019483769633507852, |
| "loss": 0.1035, |
| "step": 493 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019482722513089007, |
| "loss": 0.1162, |
| "step": 494 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019481675392670159, |
| "loss": 0.1709, |
| "step": 495 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019480628272251308, |
| "loss": 0.106, |
| "step": 496 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019479581151832462, |
| "loss": 0.1173, |
| "step": 497 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019478534031413614, |
| "loss": 0.1649, |
| "step": 498 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019477486910994765, |
| "loss": 0.1443, |
| "step": 499 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019476439790575917, |
| "loss": 0.1576, |
| "step": 500 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.0001947539267015707, |
| "loss": 0.1404, |
| "step": 501 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.0001947434554973822, |
| "loss": 0.1507, |
| "step": 502 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019473298429319372, |
| "loss": 0.1293, |
| "step": 503 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019472251308900524, |
| "loss": 0.1146, |
| "step": 504 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019471204188481676, |
| "loss": 0.1343, |
| "step": 505 |
| }, |
| { |
| "epoch": 0.26, |
| "learning_rate": 0.00019470157068062827, |
| "loss": 0.0991, |
| "step": 506 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.0001946910994764398, |
| "loss": 0.1379, |
| "step": 507 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.0001946806282722513, |
| "loss": 0.1462, |
| "step": 508 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.00019467015706806285, |
| "loss": 0.1292, |
| "step": 509 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.00019465968586387434, |
| "loss": 0.1603, |
| "step": 510 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.00019464921465968586, |
| "loss": 0.1445, |
| "step": 511 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.0001946387434554974, |
| "loss": 0.1131, |
| "step": 512 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.00019462827225130892, |
| "loss": 0.1546, |
| "step": 513 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.0001946178010471204, |
| "loss": 0.1047, |
| "step": 514 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.00019460732984293195, |
| "loss": 0.2546, |
| "step": 515 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.00019459685863874347, |
| "loss": 0.1485, |
| "step": 516 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.00019458638743455496, |
| "loss": 0.1452, |
| "step": 517 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.0001945759162303665, |
| "loss": 0.1425, |
| "step": 518 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.00019456544502617802, |
| "loss": 0.0973, |
| "step": 519 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.00019455497382198954, |
| "loss": 0.1234, |
| "step": 520 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.00019454450261780106, |
| "loss": 0.1484, |
| "step": 521 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.00019453403141361257, |
| "loss": 0.1207, |
| "step": 522 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.0001945235602094241, |
| "loss": 0.1729, |
| "step": 523 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.0001945130890052356, |
| "loss": 0.1295, |
| "step": 524 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 0.00019450261780104713, |
| "loss": 0.1594, |
| "step": 525 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.00019449214659685864, |
| "loss": 0.1048, |
| "step": 526 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.00019448167539267016, |
| "loss": 0.1356, |
| "step": 527 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.00019447120418848168, |
| "loss": 0.1457, |
| "step": 528 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.0001944607329842932, |
| "loss": 0.117, |
| "step": 529 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.00019445026178010474, |
| "loss": 0.1167, |
| "step": 530 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.00019443979057591623, |
| "loss": 0.1022, |
| "step": 531 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.00019442931937172774, |
| "loss": 0.1225, |
| "step": 532 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.0001944188481675393, |
| "loss": 0.1328, |
| "step": 533 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.00019440837696335078, |
| "loss": 0.1389, |
| "step": 534 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.0001943979057591623, |
| "loss": 0.1716, |
| "step": 535 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.00019438743455497384, |
| "loss": 0.1556, |
| "step": 536 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.00019437696335078536, |
| "loss": 0.1389, |
| "step": 537 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.00019436649214659685, |
| "loss": 0.1438, |
| "step": 538 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.0001943560209424084, |
| "loss": 0.1237, |
| "step": 539 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.0001943455497382199, |
| "loss": 0.1413, |
| "step": 540 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.00019433507853403143, |
| "loss": 0.1416, |
| "step": 541 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.00019432460732984294, |
| "loss": 0.0988, |
| "step": 542 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.00019431413612565446, |
| "loss": 0.1503, |
| "step": 543 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 0.00019430366492146598, |
| "loss": 0.1282, |
| "step": 544 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.0001942931937172775, |
| "loss": 0.0956, |
| "step": 545 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.000194282722513089, |
| "loss": 0.1049, |
| "step": 546 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.00019427225130890056, |
| "loss": 0.1907, |
| "step": 547 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.00019426178010471205, |
| "loss": 0.1665, |
| "step": 548 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.00019425130890052356, |
| "loss": 0.1021, |
| "step": 549 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.0001942408376963351, |
| "loss": 0.1236, |
| "step": 550 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.0001942303664921466, |
| "loss": 0.1173, |
| "step": 551 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.0001942198952879581, |
| "loss": 0.1265, |
| "step": 552 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.00019420942408376966, |
| "loss": 0.115, |
| "step": 553 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.00019419895287958117, |
| "loss": 0.1333, |
| "step": 554 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.00019418848167539266, |
| "loss": 0.1274, |
| "step": 555 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.0001941780104712042, |
| "loss": 0.1263, |
| "step": 556 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.00019416753926701573, |
| "loss": 0.1566, |
| "step": 557 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.00019415706806282724, |
| "loss": 0.104, |
| "step": 558 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.00019414659685863876, |
| "loss": 0.1398, |
| "step": 559 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.00019413612565445028, |
| "loss": 0.1335, |
| "step": 560 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.0001941256544502618, |
| "loss": 0.1428, |
| "step": 561 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.0001941151832460733, |
| "loss": 0.1026, |
| "step": 562 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 0.00019410471204188483, |
| "loss": 0.11, |
| "step": 563 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.00019409424083769635, |
| "loss": 0.1188, |
| "step": 564 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.00019408376963350786, |
| "loss": 0.1461, |
| "step": 565 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.00019407329842931938, |
| "loss": 0.1207, |
| "step": 566 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.0001940628272251309, |
| "loss": 0.1725, |
| "step": 567 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.00019405235602094241, |
| "loss": 0.1496, |
| "step": 568 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.00019404188481675393, |
| "loss": 0.1325, |
| "step": 569 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.00019403141361256545, |
| "loss": 0.1367, |
| "step": 570 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.000194020942408377, |
| "loss": 0.1085, |
| "step": 571 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.00019401047120418848, |
| "loss": 0.1113, |
| "step": 572 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.000194, |
| "loss": 0.1481, |
| "step": 573 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.00019398952879581154, |
| "loss": 0.1005, |
| "step": 574 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.00019397905759162306, |
| "loss": 0.1559, |
| "step": 575 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.00019396858638743455, |
| "loss": 0.1385, |
| "step": 576 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.0001939581151832461, |
| "loss": 0.1041, |
| "step": 577 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.0001939476439790576, |
| "loss": 0.1636, |
| "step": 578 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.0001939371727748691, |
| "loss": 0.1368, |
| "step": 579 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.00019392670157068065, |
| "loss": 0.1397, |
| "step": 580 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.00019391623036649216, |
| "loss": 0.1091, |
| "step": 581 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 0.00019390575916230368, |
| "loss": 0.1062, |
| "step": 582 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.0001938952879581152, |
| "loss": 0.1647, |
| "step": 583 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.00019388481675392671, |
| "loss": 0.1247, |
| "step": 584 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.00019387434554973823, |
| "loss": 0.1179, |
| "step": 585 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.00019386387434554975, |
| "loss": 0.1992, |
| "step": 586 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.00019385340314136127, |
| "loss": 0.1549, |
| "step": 587 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.00019384293193717278, |
| "loss": 0.0983, |
| "step": 588 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.0001938324607329843, |
| "loss": 0.0953, |
| "step": 589 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.00019382198952879582, |
| "loss": 0.1243, |
| "step": 590 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.00019381151832460733, |
| "loss": 0.133, |
| "step": 591 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.00019380104712041888, |
| "loss": 0.1311, |
| "step": 592 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.00019379057591623037, |
| "loss": 0.1142, |
| "step": 593 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.00019378010471204188, |
| "loss": 0.0874, |
| "step": 594 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.00019376963350785343, |
| "loss": 0.154, |
| "step": 595 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.00019375916230366492, |
| "loss": 0.1517, |
| "step": 596 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.00019374869109947644, |
| "loss": 0.1116, |
| "step": 597 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.00019373821989528798, |
| "loss": 0.1306, |
| "step": 598 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.0001937277486910995, |
| "loss": 0.1352, |
| "step": 599 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.000193717277486911, |
| "loss": 0.1412, |
| "step": 600 |
| }, |
| { |
| "epoch": 0.31, |
| "learning_rate": 0.00019370680628272253, |
| "loss": 0.1305, |
| "step": 601 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.00019369633507853405, |
| "loss": 0.1072, |
| "step": 602 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.00019368586387434557, |
| "loss": 0.1507, |
| "step": 603 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.00019367539267015708, |
| "loss": 0.1576, |
| "step": 604 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.0001936649214659686, |
| "loss": 0.1121, |
| "step": 605 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.00019365445026178012, |
| "loss": 0.1576, |
| "step": 606 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.00019364397905759163, |
| "loss": 0.1049, |
| "step": 607 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.00019363350785340315, |
| "loss": 0.105, |
| "step": 608 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.00019362303664921467, |
| "loss": 0.1155, |
| "step": 609 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.00019361256544502619, |
| "loss": 0.127, |
| "step": 610 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.0001936020942408377, |
| "loss": 0.0964, |
| "step": 611 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.00019359162303664922, |
| "loss": 0.1037, |
| "step": 612 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.00019358115183246074, |
| "loss": 0.1576, |
| "step": 613 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.00019357068062827225, |
| "loss": 0.1276, |
| "step": 614 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.00019356020942408377, |
| "loss": 0.1081, |
| "step": 615 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.00019354973821989531, |
| "loss": 0.106, |
| "step": 616 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.0001935392670157068, |
| "loss": 0.1865, |
| "step": 617 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.00019352879581151832, |
| "loss": 0.106, |
| "step": 618 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.00019351832460732987, |
| "loss": 0.1429, |
| "step": 619 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 0.00019350785340314138, |
| "loss": 0.1074, |
| "step": 620 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.00019349738219895287, |
| "loss": 0.1162, |
| "step": 621 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.00019348691099476442, |
| "loss": 0.1199, |
| "step": 622 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.00019347643979057593, |
| "loss": 0.1477, |
| "step": 623 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.00019346596858638742, |
| "loss": 0.121, |
| "step": 624 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.00019345549738219897, |
| "loss": 0.1161, |
| "step": 625 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.00019344502617801049, |
| "loss": 0.1534, |
| "step": 626 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.000193434554973822, |
| "loss": 0.1025, |
| "step": 627 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.00019342408376963352, |
| "loss": 0.1184, |
| "step": 628 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.00019341361256544504, |
| "loss": 0.1387, |
| "step": 629 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.00019340314136125655, |
| "loss": 0.1134, |
| "step": 630 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.00019339267015706807, |
| "loss": 0.1471, |
| "step": 631 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.0001933821989528796, |
| "loss": 0.1367, |
| "step": 632 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.0001933717277486911, |
| "loss": 0.1548, |
| "step": 633 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.00019336125654450262, |
| "loss": 0.143, |
| "step": 634 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.00019335078534031414, |
| "loss": 0.1526, |
| "step": 635 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.00019334031413612566, |
| "loss": 0.0817, |
| "step": 636 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.0001933298429319372, |
| "loss": 0.1512, |
| "step": 637 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.0001933193717277487, |
| "loss": 0.1363, |
| "step": 638 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 0.0001933089005235602, |
| "loss": 0.1166, |
| "step": 639 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.00019329842931937175, |
| "loss": 0.1072, |
| "step": 640 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.00019328795811518324, |
| "loss": 0.1369, |
| "step": 641 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.00019327748691099476, |
| "loss": 0.1248, |
| "step": 642 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.0001932670157068063, |
| "loss": 0.1238, |
| "step": 643 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.00019325654450261782, |
| "loss": 0.1212, |
| "step": 644 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.0001932460732984293, |
| "loss": 0.133, |
| "step": 645 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.00019323560209424085, |
| "loss": 0.127, |
| "step": 646 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.00019322513089005237, |
| "loss": 0.1587, |
| "step": 647 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.00019321465968586386, |
| "loss": 0.1602, |
| "step": 648 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.0001932041884816754, |
| "loss": 0.1199, |
| "step": 649 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.00019319371727748692, |
| "loss": 0.1286, |
| "step": 650 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.00019318324607329844, |
| "loss": 0.1229, |
| "step": 651 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.00019317277486910996, |
| "loss": 0.111, |
| "step": 652 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.00019316230366492147, |
| "loss": 0.1403, |
| "step": 653 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.000193151832460733, |
| "loss": 0.1586, |
| "step": 654 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.0001931413612565445, |
| "loss": 0.1016, |
| "step": 655 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.00019313089005235603, |
| "loss": 0.1157, |
| "step": 656 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.00019312041884816754, |
| "loss": 0.1277, |
| "step": 657 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 0.00019310994764397906, |
| "loss": 0.1143, |
| "step": 658 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.00019309947643979058, |
| "loss": 0.127, |
| "step": 659 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.0001930890052356021, |
| "loss": 0.1242, |
| "step": 660 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.00019307853403141364, |
| "loss": 0.1556, |
| "step": 661 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.00019306806282722513, |
| "loss": 0.1236, |
| "step": 662 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.00019305759162303664, |
| "loss": 0.1018, |
| "step": 663 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.0001930471204188482, |
| "loss": 0.1067, |
| "step": 664 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.0001930366492146597, |
| "loss": 0.1448, |
| "step": 665 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.0001930261780104712, |
| "loss": 0.1703, |
| "step": 666 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.00019301570680628274, |
| "loss": 0.1275, |
| "step": 667 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.00019300523560209426, |
| "loss": 0.1104, |
| "step": 668 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.00019299476439790575, |
| "loss": 0.1124, |
| "step": 669 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.0001929842931937173, |
| "loss": 0.1114, |
| "step": 670 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.0001929738219895288, |
| "loss": 0.1133, |
| "step": 671 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.00019296335078534033, |
| "loss": 0.1595, |
| "step": 672 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.00019295287958115184, |
| "loss": 0.1218, |
| "step": 673 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.00019294240837696336, |
| "loss": 0.148, |
| "step": 674 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.00019293193717277488, |
| "loss": 0.159, |
| "step": 675 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.0001929214659685864, |
| "loss": 0.1259, |
| "step": 676 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.0001929109947643979, |
| "loss": 0.1177, |
| "step": 677 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 0.00019290052356020943, |
| "loss": 0.1082, |
| "step": 678 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019289005235602094, |
| "loss": 0.1522, |
| "step": 679 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019287958115183246, |
| "loss": 0.1052, |
| "step": 680 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019286910994764398, |
| "loss": 0.1246, |
| "step": 681 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019285863874345552, |
| "loss": 0.1356, |
| "step": 682 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.000192848167539267, |
| "loss": 0.1462, |
| "step": 683 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019283769633507853, |
| "loss": 0.116, |
| "step": 684 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019282722513089007, |
| "loss": 0.1472, |
| "step": 685 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019281675392670156, |
| "loss": 0.154, |
| "step": 686 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019280628272251308, |
| "loss": 0.1473, |
| "step": 687 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019279581151832463, |
| "loss": 0.1436, |
| "step": 688 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019278534031413614, |
| "loss": 0.1147, |
| "step": 689 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019277486910994763, |
| "loss": 0.1587, |
| "step": 690 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019276439790575918, |
| "loss": 0.1194, |
| "step": 691 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.0001927539267015707, |
| "loss": 0.12, |
| "step": 692 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019274345549738218, |
| "loss": 0.134, |
| "step": 693 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019273298429319373, |
| "loss": 0.1634, |
| "step": 694 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019272251308900525, |
| "loss": 0.1353, |
| "step": 695 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019271204188481676, |
| "loss": 0.1441, |
| "step": 696 |
| }, |
| { |
| "epoch": 0.36, |
| "learning_rate": 0.00019270157068062828, |
| "loss": 0.1267, |
| "step": 697 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.0001926910994764398, |
| "loss": 0.1527, |
| "step": 698 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.00019268062827225134, |
| "loss": 0.1314, |
| "step": 699 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.00019267015706806283, |
| "loss": 0.1531, |
| "step": 700 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.00019265968586387435, |
| "loss": 0.1752, |
| "step": 701 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.0001926492146596859, |
| "loss": 0.1612, |
| "step": 702 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.00019263874345549738, |
| "loss": 0.1025, |
| "step": 703 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.0001926282722513089, |
| "loss": 0.1436, |
| "step": 704 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.00019261780104712044, |
| "loss": 0.1286, |
| "step": 705 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.00019260732984293196, |
| "loss": 0.0974, |
| "step": 706 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.00019259685863874345, |
| "loss": 0.1291, |
| "step": 707 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.000192586387434555, |
| "loss": 0.1181, |
| "step": 708 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.0001925759162303665, |
| "loss": 0.166, |
| "step": 709 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.00019256544502617803, |
| "loss": 0.1252, |
| "step": 710 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.00019255497382198955, |
| "loss": 0.0987, |
| "step": 711 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.00019254450261780106, |
| "loss": 0.1271, |
| "step": 712 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.00019253403141361258, |
| "loss": 0.1262, |
| "step": 713 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.0001925235602094241, |
| "loss": 0.1157, |
| "step": 714 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.00019251308900523561, |
| "loss": 0.13, |
| "step": 715 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 0.00019250261780104713, |
| "loss": 0.1341, |
| "step": 716 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.00019249214659685865, |
| "loss": 0.1529, |
| "step": 717 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.00019248167539267017, |
| "loss": 0.1359, |
| "step": 718 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.00019247120418848168, |
| "loss": 0.1535, |
| "step": 719 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.0001924607329842932, |
| "loss": 0.0994, |
| "step": 720 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.00019245026178010472, |
| "loss": 0.135, |
| "step": 721 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.00019243979057591623, |
| "loss": 0.1228, |
| "step": 722 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.00019242931937172778, |
| "loss": 0.1177, |
| "step": 723 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.00019241884816753927, |
| "loss": 0.1412, |
| "step": 724 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.00019240837696335078, |
| "loss": 0.1444, |
| "step": 725 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.00019239790575916233, |
| "loss": 0.1168, |
| "step": 726 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.00019238743455497385, |
| "loss": 0.1493, |
| "step": 727 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.00019237696335078534, |
| "loss": 0.1032, |
| "step": 728 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.00019236649214659688, |
| "loss": 0.1064, |
| "step": 729 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.0001923560209424084, |
| "loss": 0.1495, |
| "step": 730 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.0001923455497382199, |
| "loss": 0.0989, |
| "step": 731 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.00019233507853403143, |
| "loss": 0.1086, |
| "step": 732 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.00019232460732984295, |
| "loss": 0.1392, |
| "step": 733 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.00019231413612565447, |
| "loss": 0.0898, |
| "step": 734 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 0.00019230366492146598, |
| "loss": 0.1248, |
| "step": 735 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.0001922931937172775, |
| "loss": 0.1224, |
| "step": 736 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.00019228272251308902, |
| "loss": 0.1293, |
| "step": 737 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.00019227225130890053, |
| "loss": 0.0999, |
| "step": 738 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.00019226178010471205, |
| "loss": 0.1333, |
| "step": 739 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.00019225130890052357, |
| "loss": 0.1472, |
| "step": 740 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.00019224083769633508, |
| "loss": 0.15, |
| "step": 741 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.0001922303664921466, |
| "loss": 0.101, |
| "step": 742 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.00019221989528795812, |
| "loss": 0.0794, |
| "step": 743 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.00019220942408376966, |
| "loss": 0.1133, |
| "step": 744 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.00019219895287958115, |
| "loss": 0.1466, |
| "step": 745 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.00019218848167539267, |
| "loss": 0.1565, |
| "step": 746 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.00019217801047120421, |
| "loss": 0.0789, |
| "step": 747 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.0001921675392670157, |
| "loss": 0.115, |
| "step": 748 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.00019215706806282722, |
| "loss": 0.0834, |
| "step": 749 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.00019214659685863877, |
| "loss": 0.1294, |
| "step": 750 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.00019213612565445028, |
| "loss": 0.1448, |
| "step": 751 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.00019212565445026177, |
| "loss": 0.16, |
| "step": 752 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.00019211518324607332, |
| "loss": 0.1073, |
| "step": 753 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 0.00019210471204188483, |
| "loss": 0.1237, |
| "step": 754 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.00019209424083769632, |
| "loss": 0.1284, |
| "step": 755 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.00019208376963350787, |
| "loss": 0.1263, |
| "step": 756 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.00019207329842931939, |
| "loss": 0.097, |
| "step": 757 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.0001920628272251309, |
| "loss": 0.1539, |
| "step": 758 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.00019205235602094242, |
| "loss": 0.1682, |
| "step": 759 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.00019204188481675394, |
| "loss": 0.1176, |
| "step": 760 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.00019203141361256545, |
| "loss": 0.1017, |
| "step": 761 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.00019202094240837697, |
| "loss": 0.1216, |
| "step": 762 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.0001920104712041885, |
| "loss": 0.1587, |
| "step": 763 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.000192, |
| "loss": 0.1428, |
| "step": 764 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.00019198952879581152, |
| "loss": 0.0945, |
| "step": 765 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.00019197905759162304, |
| "loss": 0.1229, |
| "step": 766 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.00019196858638743456, |
| "loss": 0.1002, |
| "step": 767 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.0001919581151832461, |
| "loss": 0.1405, |
| "step": 768 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.0001919476439790576, |
| "loss": 0.1119, |
| "step": 769 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.0001919371727748691, |
| "loss": 0.1757, |
| "step": 770 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.00019192670157068065, |
| "loss": 0.1293, |
| "step": 771 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.00019191623036649217, |
| "loss": 0.108, |
| "step": 772 |
| }, |
| { |
| "epoch": 0.4, |
| "learning_rate": 0.00019190575916230366, |
| "loss": 0.1441, |
| "step": 773 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.0001918952879581152, |
| "loss": 0.1343, |
| "step": 774 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.00019188481675392672, |
| "loss": 0.1062, |
| "step": 775 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.0001918743455497382, |
| "loss": 0.0866, |
| "step": 776 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.00019186387434554975, |
| "loss": 0.0932, |
| "step": 777 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.00019185340314136127, |
| "loss": 0.0931, |
| "step": 778 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.0001918429319371728, |
| "loss": 0.1203, |
| "step": 779 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.0001918324607329843, |
| "loss": 0.1278, |
| "step": 780 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.00019182198952879582, |
| "loss": 0.0876, |
| "step": 781 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.00019181151832460734, |
| "loss": 0.1254, |
| "step": 782 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.00019180104712041886, |
| "loss": 0.1267, |
| "step": 783 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.00019179057591623037, |
| "loss": 0.1642, |
| "step": 784 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.0001917801047120419, |
| "loss": 0.1232, |
| "step": 785 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.0001917696335078534, |
| "loss": 0.1186, |
| "step": 786 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.00019175916230366492, |
| "loss": 0.1138, |
| "step": 787 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.00019174869109947644, |
| "loss": 0.113, |
| "step": 788 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.00019173821989528799, |
| "loss": 0.1243, |
| "step": 789 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.00019172774869109948, |
| "loss": 0.1453, |
| "step": 790 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.000191717277486911, |
| "loss": 0.1565, |
| "step": 791 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 0.00019170680628272254, |
| "loss": 0.1044, |
| "step": 792 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.00019169633507853403, |
| "loss": 0.1549, |
| "step": 793 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.00019168586387434554, |
| "loss": 0.1509, |
| "step": 794 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.0001916753926701571, |
| "loss": 0.1174, |
| "step": 795 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.0001916649214659686, |
| "loss": 0.1154, |
| "step": 796 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.0001916544502617801, |
| "loss": 0.1234, |
| "step": 797 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.00019164397905759164, |
| "loss": 0.104, |
| "step": 798 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.00019163350785340316, |
| "loss": 0.1056, |
| "step": 799 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.00019162303664921465, |
| "loss": 0.0879, |
| "step": 800 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.0001916125654450262, |
| "loss": 0.1414, |
| "step": 801 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.0001916020942408377, |
| "loss": 0.1125, |
| "step": 802 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.00019159162303664923, |
| "loss": 0.1371, |
| "step": 803 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.00019158115183246074, |
| "loss": 0.1398, |
| "step": 804 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.00019157068062827226, |
| "loss": 0.1353, |
| "step": 805 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.00019156020942408378, |
| "loss": 0.0883, |
| "step": 806 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.0001915497382198953, |
| "loss": 0.1243, |
| "step": 807 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.0001915392670157068, |
| "loss": 0.1107, |
| "step": 808 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.00019152879581151833, |
| "loss": 0.1238, |
| "step": 809 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.00019151832460732984, |
| "loss": 0.1472, |
| "step": 810 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 0.00019150785340314136, |
| "loss": 0.1085, |
| "step": 811 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.00019149738219895288, |
| "loss": 0.1297, |
| "step": 812 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.00019148691099476442, |
| "loss": 0.1295, |
| "step": 813 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.0001914764397905759, |
| "loss": 0.1392, |
| "step": 814 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.00019146596858638743, |
| "loss": 0.1192, |
| "step": 815 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.00019145549738219897, |
| "loss": 0.0856, |
| "step": 816 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.0001914450261780105, |
| "loss": 0.1542, |
| "step": 817 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.00019143455497382198, |
| "loss": 0.0963, |
| "step": 818 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.00019142408376963353, |
| "loss": 0.1281, |
| "step": 819 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.00019141361256544504, |
| "loss": 0.1161, |
| "step": 820 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.00019140314136125653, |
| "loss": 0.1402, |
| "step": 821 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.00019139267015706808, |
| "loss": 0.1661, |
| "step": 822 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.0001913821989528796, |
| "loss": 0.1157, |
| "step": 823 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.0001913717277486911, |
| "loss": 0.0945, |
| "step": 824 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.00019136125654450263, |
| "loss": 0.1179, |
| "step": 825 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.00019135078534031414, |
| "loss": 0.1531, |
| "step": 826 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.00019134031413612566, |
| "loss": 0.1317, |
| "step": 827 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.00019132984293193718, |
| "loss": 0.1053, |
| "step": 828 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.0001913193717277487, |
| "loss": 0.0952, |
| "step": 829 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 0.0001913089005235602, |
| "loss": 0.1174, |
| "step": 830 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.00019129842931937173, |
| "loss": 0.1152, |
| "step": 831 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.00019128795811518325, |
| "loss": 0.1493, |
| "step": 832 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.00019127748691099476, |
| "loss": 0.0878, |
| "step": 833 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.0001912670157068063, |
| "loss": 0.1427, |
| "step": 834 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.0001912565445026178, |
| "loss": 0.0662, |
| "step": 835 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.00019124607329842932, |
| "loss": 0.1429, |
| "step": 836 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.00019123560209424086, |
| "loss": 0.2017, |
| "step": 837 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.00019122513089005235, |
| "loss": 0.1352, |
| "step": 838 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.00019121465968586387, |
| "loss": 0.1048, |
| "step": 839 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.0001912041884816754, |
| "loss": 0.1164, |
| "step": 840 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.00019119371727748693, |
| "loss": 0.1234, |
| "step": 841 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.00019118324607329842, |
| "loss": 0.124, |
| "step": 842 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.00019117277486910996, |
| "loss": 0.1439, |
| "step": 843 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.00019116230366492148, |
| "loss": 0.1163, |
| "step": 844 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.00019115183246073297, |
| "loss": 0.1298, |
| "step": 845 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.00019114136125654451, |
| "loss": 0.1351, |
| "step": 846 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.00019113089005235603, |
| "loss": 0.1116, |
| "step": 847 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.00019112041884816755, |
| "loss": 0.1011, |
| "step": 848 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 0.00019110994764397906, |
| "loss": 0.1031, |
| "step": 849 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019109947643979058, |
| "loss": 0.1048, |
| "step": 850 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019108900523560213, |
| "loss": 0.14, |
| "step": 851 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019107853403141362, |
| "loss": 0.0877, |
| "step": 852 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019106806282722513, |
| "loss": 0.1588, |
| "step": 853 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019105759162303668, |
| "loss": 0.1078, |
| "step": 854 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019104712041884817, |
| "loss": 0.1383, |
| "step": 855 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019103664921465968, |
| "loss": 0.1205, |
| "step": 856 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019102617801047123, |
| "loss": 0.1097, |
| "step": 857 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019101570680628275, |
| "loss": 0.104, |
| "step": 858 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019100523560209424, |
| "loss": 0.0952, |
| "step": 859 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019099476439790578, |
| "loss": 0.1205, |
| "step": 860 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.0001909842931937173, |
| "loss": 0.1029, |
| "step": 861 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019097382198952881, |
| "loss": 0.0897, |
| "step": 862 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019096335078534033, |
| "loss": 0.1094, |
| "step": 863 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019095287958115185, |
| "loss": 0.1179, |
| "step": 864 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019094240837696337, |
| "loss": 0.1071, |
| "step": 865 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019093193717277488, |
| "loss": 0.1503, |
| "step": 866 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.0001909214659685864, |
| "loss": 0.1035, |
| "step": 867 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019091099476439792, |
| "loss": 0.146, |
| "step": 868 |
| }, |
| { |
| "epoch": 0.45, |
| "learning_rate": 0.00019090052356020943, |
| "loss": 0.1571, |
| "step": 869 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019089005235602095, |
| "loss": 0.0951, |
| "step": 870 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019087958115183247, |
| "loss": 0.126, |
| "step": 871 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019086910994764398, |
| "loss": 0.1262, |
| "step": 872 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.0001908586387434555, |
| "loss": 0.1185, |
| "step": 873 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019084816753926702, |
| "loss": 0.1127, |
| "step": 874 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019083769633507856, |
| "loss": 0.1295, |
| "step": 875 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019082722513089005, |
| "loss": 0.136, |
| "step": 876 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019081675392670157, |
| "loss": 0.1368, |
| "step": 877 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019080628272251311, |
| "loss": 0.0959, |
| "step": 878 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019079581151832463, |
| "loss": 0.1368, |
| "step": 879 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019078534031413612, |
| "loss": 0.0974, |
| "step": 880 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019077486910994767, |
| "loss": 0.1203, |
| "step": 881 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019076439790575918, |
| "loss": 0.1165, |
| "step": 882 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019075392670157067, |
| "loss": 0.0886, |
| "step": 883 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019074345549738222, |
| "loss": 0.1212, |
| "step": 884 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019073298429319373, |
| "loss": 0.148, |
| "step": 885 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019072251308900525, |
| "loss": 0.0996, |
| "step": 886 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019071204188481677, |
| "loss": 0.118, |
| "step": 887 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 0.00019070157068062829, |
| "loss": 0.1023, |
| "step": 888 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.0001906910994764398, |
| "loss": 0.1533, |
| "step": 889 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.00019068062827225132, |
| "loss": 0.1172, |
| "step": 890 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.00019067015706806284, |
| "loss": 0.0968, |
| "step": 891 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.00019065968586387435, |
| "loss": 0.1236, |
| "step": 892 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.00019064921465968587, |
| "loss": 0.091, |
| "step": 893 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.0001906387434554974, |
| "loss": 0.1637, |
| "step": 894 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.0001906282722513089, |
| "loss": 0.1534, |
| "step": 895 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.00019061780104712045, |
| "loss": 0.1216, |
| "step": 896 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.00019060732984293194, |
| "loss": 0.1179, |
| "step": 897 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.00019059685863874346, |
| "loss": 0.0647, |
| "step": 898 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.000190586387434555, |
| "loss": 0.095, |
| "step": 899 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.0001905759162303665, |
| "loss": 0.1111, |
| "step": 900 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.000190565445026178, |
| "loss": 0.1286, |
| "step": 901 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.00019055497382198955, |
| "loss": 0.1137, |
| "step": 902 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.00019054450261780107, |
| "loss": 0.1354, |
| "step": 903 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.00019053403141361256, |
| "loss": 0.1361, |
| "step": 904 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.0001905235602094241, |
| "loss": 0.1241, |
| "step": 905 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.00019051308900523562, |
| "loss": 0.1118, |
| "step": 906 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 0.0001905026178010471, |
| "loss": 0.0932, |
| "step": 907 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.00019049214659685865, |
| "loss": 0.1509, |
| "step": 908 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.00019048167539267017, |
| "loss": 0.1374, |
| "step": 909 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.0001904712041884817, |
| "loss": 0.1068, |
| "step": 910 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.0001904607329842932, |
| "loss": 0.0821, |
| "step": 911 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.00019045026178010472, |
| "loss": 0.1186, |
| "step": 912 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.00019043979057591624, |
| "loss": 0.133, |
| "step": 913 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.00019042931937172776, |
| "loss": 0.1174, |
| "step": 914 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.00019041884816753927, |
| "loss": 0.1335, |
| "step": 915 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.0001904083769633508, |
| "loss": 0.1028, |
| "step": 916 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.0001903979057591623, |
| "loss": 0.1327, |
| "step": 917 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.00019038743455497382, |
| "loss": 0.1334, |
| "step": 918 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.00019037696335078534, |
| "loss": 0.1002, |
| "step": 919 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.00019036649214659689, |
| "loss": 0.0959, |
| "step": 920 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.00019035602094240838, |
| "loss": 0.1188, |
| "step": 921 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.0001903455497382199, |
| "loss": 0.1053, |
| "step": 922 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.00019033507853403144, |
| "loss": 0.1315, |
| "step": 923 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.00019032460732984295, |
| "loss": 0.1092, |
| "step": 924 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.00019031413612565444, |
| "loss": 0.1209, |
| "step": 925 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 0.000190303664921466, |
| "loss": 0.0814, |
| "step": 926 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.0001902931937172775, |
| "loss": 0.1257, |
| "step": 927 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.000190282722513089, |
| "loss": 0.1058, |
| "step": 928 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.00019027225130890054, |
| "loss": 0.1192, |
| "step": 929 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.00019026178010471206, |
| "loss": 0.1054, |
| "step": 930 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.00019025130890052357, |
| "loss": 0.0967, |
| "step": 931 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.0001902408376963351, |
| "loss": 0.1279, |
| "step": 932 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.0001902303664921466, |
| "loss": 0.1145, |
| "step": 933 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.00019021989528795812, |
| "loss": 0.0936, |
| "step": 934 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.00019020942408376964, |
| "loss": 0.1356, |
| "step": 935 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.00019019895287958116, |
| "loss": 0.158, |
| "step": 936 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.00019018848167539268, |
| "loss": 0.1244, |
| "step": 937 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.0001901780104712042, |
| "loss": 0.1292, |
| "step": 938 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.0001901675392670157, |
| "loss": 0.114, |
| "step": 939 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.00019015706806282723, |
| "loss": 0.0814, |
| "step": 940 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.00019014659685863877, |
| "loss": 0.1054, |
| "step": 941 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.00019013612565445026, |
| "loss": 0.1069, |
| "step": 942 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.00019012565445026178, |
| "loss": 0.136, |
| "step": 943 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.00019011518324607332, |
| "loss": 0.1451, |
| "step": 944 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 0.0001901047120418848, |
| "loss": 0.1203, |
| "step": 945 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.00019009424083769633, |
| "loss": 0.1302, |
| "step": 946 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.00019008376963350787, |
| "loss": 0.1257, |
| "step": 947 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.0001900732984293194, |
| "loss": 0.1097, |
| "step": 948 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.00019006282722513088, |
| "loss": 0.1286, |
| "step": 949 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.00019005235602094243, |
| "loss": 0.1216, |
| "step": 950 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.00019004188481675394, |
| "loss": 0.1039, |
| "step": 951 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.00019003141361256543, |
| "loss": 0.1247, |
| "step": 952 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.00019002094240837698, |
| "loss": 0.0874, |
| "step": 953 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.0001900104712041885, |
| "loss": 0.1241, |
| "step": 954 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.00019, |
| "loss": 0.123, |
| "step": 955 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.00018998952879581153, |
| "loss": 0.1241, |
| "step": 956 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.00018997905759162304, |
| "loss": 0.1249, |
| "step": 957 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.00018996858638743456, |
| "loss": 0.1413, |
| "step": 958 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.00018995811518324608, |
| "loss": 0.0843, |
| "step": 959 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.0001899476439790576, |
| "loss": 0.114, |
| "step": 960 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.0001899371727748691, |
| "loss": 0.119, |
| "step": 961 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.00018992670157068063, |
| "loss": 0.1176, |
| "step": 962 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.00018991623036649215, |
| "loss": 0.0932, |
| "step": 963 |
| }, |
| { |
| "epoch": 0.5, |
| "learning_rate": 0.00018990575916230366, |
| "loss": 0.1434, |
| "step": 964 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.0001898952879581152, |
| "loss": 0.1241, |
| "step": 965 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.0001898848167539267, |
| "loss": 0.1103, |
| "step": 966 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.00018987434554973822, |
| "loss": 0.0896, |
| "step": 967 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.00018986387434554976, |
| "loss": 0.1007, |
| "step": 968 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.00018985340314136128, |
| "loss": 0.1036, |
| "step": 969 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.00018984293193717277, |
| "loss": 0.0908, |
| "step": 970 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.0001898324607329843, |
| "loss": 0.1487, |
| "step": 971 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.00018982198952879583, |
| "loss": 0.1275, |
| "step": 972 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.00018981151832460732, |
| "loss": 0.1104, |
| "step": 973 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.00018980104712041886, |
| "loss": 0.1228, |
| "step": 974 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.00018979057591623038, |
| "loss": 0.1134, |
| "step": 975 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.0001897801047120419, |
| "loss": 0.1058, |
| "step": 976 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.0001897696335078534, |
| "loss": 0.1141, |
| "step": 977 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.00018975916230366493, |
| "loss": 0.1694, |
| "step": 978 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.00018974869109947645, |
| "loss": 0.11, |
| "step": 979 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.00018973821989528796, |
| "loss": 0.1067, |
| "step": 980 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.00018972774869109948, |
| "loss": 0.1671, |
| "step": 981 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.000189717277486911, |
| "loss": 0.1414, |
| "step": 982 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 0.00018970680628272252, |
| "loss": 0.1116, |
| "step": 983 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.00018969633507853403, |
| "loss": 0.1191, |
| "step": 984 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.00018968586387434555, |
| "loss": 0.1171, |
| "step": 985 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.0001896753926701571, |
| "loss": 0.1249, |
| "step": 986 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.00018966492146596858, |
| "loss": 0.0931, |
| "step": 987 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.0001896544502617801, |
| "loss": 0.0993, |
| "step": 988 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.00018964397905759165, |
| "loss": 0.1204, |
| "step": 989 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.00018963350785340314, |
| "loss": 0.087, |
| "step": 990 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.00018962303664921465, |
| "loss": 0.1457, |
| "step": 991 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.0001896125654450262, |
| "loss": 0.1073, |
| "step": 992 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.00018960209424083771, |
| "loss": 0.1171, |
| "step": 993 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.0001895916230366492, |
| "loss": 0.0736, |
| "step": 994 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.00018958115183246075, |
| "loss": 0.1, |
| "step": 995 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.00018957068062827226, |
| "loss": 0.1303, |
| "step": 996 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.00018956020942408375, |
| "loss": 0.1324, |
| "step": 997 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.0001895497382198953, |
| "loss": 0.1118, |
| "step": 998 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.00018953926701570682, |
| "loss": 0.1291, |
| "step": 999 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.00018952879581151833, |
| "loss": 0.1445, |
| "step": 1000 |
| }, |
| { |
| "epoch": 0.52, |
| "eval_loss": 0.11310110986232758, |
| "eval_runtime": 319.8032, |
| "eval_samples_per_second": 0.619, |
| "eval_steps_per_second": 0.078, |
| "step": 1000 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.00018951832460732985, |
| "loss": 0.0786, |
| "step": 1001 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 0.00018950785340314137, |
| "loss": 0.0932, |
| "step": 1002 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.0001894973821989529, |
| "loss": 0.1094, |
| "step": 1003 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.0001894869109947644, |
| "loss": 0.1331, |
| "step": 1004 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.00018947643979057592, |
| "loss": 0.1176, |
| "step": 1005 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.00018946596858638746, |
| "loss": 0.1038, |
| "step": 1006 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.00018945549738219895, |
| "loss": 0.1234, |
| "step": 1007 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.00018944502617801047, |
| "loss": 0.1382, |
| "step": 1008 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.00018943455497382201, |
| "loss": 0.1497, |
| "step": 1009 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.00018942408376963353, |
| "loss": 0.1239, |
| "step": 1010 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.00018941361256544502, |
| "loss": 0.1121, |
| "step": 1011 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.00018940314136125657, |
| "loss": 0.12, |
| "step": 1012 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.00018939267015706808, |
| "loss": 0.1464, |
| "step": 1013 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.0001893821989528796, |
| "loss": 0.1287, |
| "step": 1014 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.00018937172774869112, |
| "loss": 0.1221, |
| "step": 1015 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.00018936125654450263, |
| "loss": 0.114, |
| "step": 1016 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.00018935078534031415, |
| "loss": 0.1235, |
| "step": 1017 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.00018934031413612567, |
| "loss": 0.1267, |
| "step": 1018 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.00018932984293193718, |
| "loss": 0.1444, |
| "step": 1019 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.0001893193717277487, |
| "loss": 0.1065, |
| "step": 1020 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 0.00018930890052356022, |
| "loss": 0.109, |
| "step": 1021 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.00018929842931937174, |
| "loss": 0.144, |
| "step": 1022 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.00018928795811518325, |
| "loss": 0.1133, |
| "step": 1023 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.00018927748691099477, |
| "loss": 0.1175, |
| "step": 1024 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.0001892670157068063, |
| "loss": 0.1429, |
| "step": 1025 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.0001892565445026178, |
| "loss": 0.1179, |
| "step": 1026 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.00018924607329842935, |
| "loss": 0.1207, |
| "step": 1027 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.00018923560209424084, |
| "loss": 0.1429, |
| "step": 1028 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.00018922513089005236, |
| "loss": 0.1322, |
| "step": 1029 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.0001892146596858639, |
| "loss": 0.1214, |
| "step": 1030 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.00018920418848167542, |
| "loss": 0.1382, |
| "step": 1031 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.0001891937172774869, |
| "loss": 0.0919, |
| "step": 1032 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.00018918324607329845, |
| "loss": 0.1194, |
| "step": 1033 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.00018917277486910997, |
| "loss": 0.1016, |
| "step": 1034 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.00018916230366492146, |
| "loss": 0.1022, |
| "step": 1035 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.000189151832460733, |
| "loss": 0.1371, |
| "step": 1036 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.00018914136125654452, |
| "loss": 0.098, |
| "step": 1037 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.00018913089005235604, |
| "loss": 0.0865, |
| "step": 1038 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.00018912041884816755, |
| "loss": 0.098, |
| "step": 1039 |
| }, |
| { |
| "epoch": 0.54, |
| "learning_rate": 0.00018910994764397907, |
| "loss": 0.0773, |
| "step": 1040 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.0001890994764397906, |
| "loss": 0.1264, |
| "step": 1041 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.0001890890052356021, |
| "loss": 0.1255, |
| "step": 1042 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.00018907853403141362, |
| "loss": 0.1021, |
| "step": 1043 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.00018906806282722514, |
| "loss": 0.1258, |
| "step": 1044 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.00018905759162303666, |
| "loss": 0.0916, |
| "step": 1045 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.00018904712041884817, |
| "loss": 0.1202, |
| "step": 1046 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.0001890366492146597, |
| "loss": 0.1046, |
| "step": 1047 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.00018902617801047123, |
| "loss": 0.1167, |
| "step": 1048 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.00018901570680628272, |
| "loss": 0.113, |
| "step": 1049 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.00018900523560209424, |
| "loss": 0.1097, |
| "step": 1050 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.00018899476439790579, |
| "loss": 0.0906, |
| "step": 1051 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.00018898429319371728, |
| "loss": 0.09, |
| "step": 1052 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.0001889738219895288, |
| "loss": 0.1346, |
| "step": 1053 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.00018896335078534034, |
| "loss": 0.0935, |
| "step": 1054 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.00018895287958115185, |
| "loss": 0.1067, |
| "step": 1055 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.00018894240837696334, |
| "loss": 0.1263, |
| "step": 1056 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.0001889319371727749, |
| "loss": 0.1385, |
| "step": 1057 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.0001889214659685864, |
| "loss": 0.0972, |
| "step": 1058 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.0001889109947643979, |
| "loss": 0.0966, |
| "step": 1059 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 0.00018890052356020944, |
| "loss": 0.106, |
| "step": 1060 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.00018889005235602096, |
| "loss": 0.1392, |
| "step": 1061 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.00018887958115183247, |
| "loss": 0.0985, |
| "step": 1062 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.000188869109947644, |
| "loss": 0.0991, |
| "step": 1063 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.0001888586387434555, |
| "loss": 0.1261, |
| "step": 1064 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.00018884816753926702, |
| "loss": 0.1108, |
| "step": 1065 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.00018883769633507854, |
| "loss": 0.0955, |
| "step": 1066 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.00018882722513089006, |
| "loss": 0.1058, |
| "step": 1067 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.00018881675392670158, |
| "loss": 0.0917, |
| "step": 1068 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.0001888062827225131, |
| "loss": 0.0868, |
| "step": 1069 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.0001887958115183246, |
| "loss": 0.0868, |
| "step": 1070 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.00018878534031413613, |
| "loss": 0.091, |
| "step": 1071 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.00018877486910994767, |
| "loss": 0.1154, |
| "step": 1072 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.00018876439790575916, |
| "loss": 0.1139, |
| "step": 1073 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.00018875392670157068, |
| "loss": 0.1474, |
| "step": 1074 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.00018874345549738222, |
| "loss": 0.13, |
| "step": 1075 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.00018873298429319374, |
| "loss": 0.1115, |
| "step": 1076 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.00018872251308900523, |
| "loss": 0.0987, |
| "step": 1077 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.00018871204188481677, |
| "loss": 0.1081, |
| "step": 1078 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 0.0001887015706806283, |
| "loss": 0.0904, |
| "step": 1079 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.00018869109947643978, |
| "loss": 0.092, |
| "step": 1080 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.00018868062827225132, |
| "loss": 0.1071, |
| "step": 1081 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.00018867015706806284, |
| "loss": 0.1089, |
| "step": 1082 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.00018865968586387436, |
| "loss": 0.0928, |
| "step": 1083 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.00018864921465968588, |
| "loss": 0.0929, |
| "step": 1084 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.0001886387434554974, |
| "loss": 0.1294, |
| "step": 1085 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.0001886282722513089, |
| "loss": 0.1094, |
| "step": 1086 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.00018861780104712043, |
| "loss": 0.0909, |
| "step": 1087 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.00018860732984293194, |
| "loss": 0.1071, |
| "step": 1088 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.00018859685863874346, |
| "loss": 0.1431, |
| "step": 1089 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.00018858638743455498, |
| "loss": 0.1176, |
| "step": 1090 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.0001885759162303665, |
| "loss": 0.1106, |
| "step": 1091 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.000188565445026178, |
| "loss": 0.1356, |
| "step": 1092 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.00018855497382198956, |
| "loss": 0.1208, |
| "step": 1093 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.00018854450261780105, |
| "loss": 0.1203, |
| "step": 1094 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.00018853403141361256, |
| "loss": 0.1023, |
| "step": 1095 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.0001885235602094241, |
| "loss": 0.1079, |
| "step": 1096 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.0001885130890052356, |
| "loss": 0.1321, |
| "step": 1097 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 0.00018850261780104712, |
| "loss": 0.1144, |
| "step": 1098 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.00018849214659685866, |
| "loss": 0.0914, |
| "step": 1099 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.00018848167539267018, |
| "loss": 0.1057, |
| "step": 1100 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.00018847120418848167, |
| "loss": 0.1358, |
| "step": 1101 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.0001884607329842932, |
| "loss": 0.0937, |
| "step": 1102 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.00018845026178010473, |
| "loss": 0.1141, |
| "step": 1103 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.00018843979057591622, |
| "loss": 0.1147, |
| "step": 1104 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.00018842931937172776, |
| "loss": 0.1161, |
| "step": 1105 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.00018841884816753928, |
| "loss": 0.1485, |
| "step": 1106 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.0001884083769633508, |
| "loss": 0.1025, |
| "step": 1107 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.0001883979057591623, |
| "loss": 0.1188, |
| "step": 1108 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.00018838743455497383, |
| "loss": 0.1207, |
| "step": 1109 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.00018837696335078535, |
| "loss": 0.1194, |
| "step": 1110 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.00018836649214659686, |
| "loss": 0.0959, |
| "step": 1111 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.00018835602094240838, |
| "loss": 0.128, |
| "step": 1112 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.0001883455497382199, |
| "loss": 0.1383, |
| "step": 1113 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.00018833507853403142, |
| "loss": 0.116, |
| "step": 1114 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.00018832460732984293, |
| "loss": 0.1088, |
| "step": 1115 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.00018831413612565445, |
| "loss": 0.0791, |
| "step": 1116 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 0.000188303664921466, |
| "loss": 0.1041, |
| "step": 1117 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.00018829319371727748, |
| "loss": 0.1297, |
| "step": 1118 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.000188282722513089, |
| "loss": 0.1079, |
| "step": 1119 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.00018827225130890054, |
| "loss": 0.1326, |
| "step": 1120 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.00018826178010471206, |
| "loss": 0.0979, |
| "step": 1121 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.00018825130890052355, |
| "loss": 0.1242, |
| "step": 1122 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.0001882408376963351, |
| "loss": 0.0956, |
| "step": 1123 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.0001882303664921466, |
| "loss": 0.0918, |
| "step": 1124 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.0001882198952879581, |
| "loss": 0.1019, |
| "step": 1125 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.00018820942408376965, |
| "loss": 0.1085, |
| "step": 1126 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.00018819895287958116, |
| "loss": 0.0865, |
| "step": 1127 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.00018818848167539268, |
| "loss": 0.1284, |
| "step": 1128 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.0001881780104712042, |
| "loss": 0.1555, |
| "step": 1129 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.00018816753926701572, |
| "loss": 0.1422, |
| "step": 1130 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.00018815706806282723, |
| "loss": 0.1143, |
| "step": 1131 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.00018814659685863875, |
| "loss": 0.0986, |
| "step": 1132 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.00018813612565445027, |
| "loss": 0.1585, |
| "step": 1133 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.00018812565445026178, |
| "loss": 0.0947, |
| "step": 1134 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.0001881151832460733, |
| "loss": 0.1057, |
| "step": 1135 |
| }, |
| { |
| "epoch": 0.59, |
| "learning_rate": 0.00018810471204188482, |
| "loss": 0.1296, |
| "step": 1136 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.00018809424083769634, |
| "loss": 0.11, |
| "step": 1137 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.00018808376963350788, |
| "loss": 0.0855, |
| "step": 1138 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.00018807329842931937, |
| "loss": 0.1063, |
| "step": 1139 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.0001880628272251309, |
| "loss": 0.1533, |
| "step": 1140 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.00018805235602094243, |
| "loss": 0.1164, |
| "step": 1141 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.00018804188481675392, |
| "loss": 0.1011, |
| "step": 1142 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.00018803141361256544, |
| "loss": 0.1336, |
| "step": 1143 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.00018802094240837698, |
| "loss": 0.0961, |
| "step": 1144 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.0001880104712041885, |
| "loss": 0.0954, |
| "step": 1145 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.000188, |
| "loss": 0.1424, |
| "step": 1146 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.00018798952879581153, |
| "loss": 0.1119, |
| "step": 1147 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.00018797905759162305, |
| "loss": 0.1013, |
| "step": 1148 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.00018796858638743454, |
| "loss": 0.1342, |
| "step": 1149 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.00018795811518324608, |
| "loss": 0.1011, |
| "step": 1150 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.0001879476439790576, |
| "loss": 0.1228, |
| "step": 1151 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.00018793717277486912, |
| "loss": 0.1005, |
| "step": 1152 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.00018792670157068064, |
| "loss": 0.109, |
| "step": 1153 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.00018791623036649215, |
| "loss": 0.0921, |
| "step": 1154 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 0.00018790575916230367, |
| "loss": 0.1553, |
| "step": 1155 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.0001878952879581152, |
| "loss": 0.1262, |
| "step": 1156 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.0001878848167539267, |
| "loss": 0.123, |
| "step": 1157 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.00018787434554973825, |
| "loss": 0.1093, |
| "step": 1158 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.00018786387434554974, |
| "loss": 0.1126, |
| "step": 1159 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.00018785340314136126, |
| "loss": 0.092, |
| "step": 1160 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.0001878429319371728, |
| "loss": 0.0928, |
| "step": 1161 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.00018783246073298432, |
| "loss": 0.099, |
| "step": 1162 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.0001878219895287958, |
| "loss": 0.1035, |
| "step": 1163 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.00018781151832460735, |
| "loss": 0.0989, |
| "step": 1164 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.00018780104712041887, |
| "loss": 0.1106, |
| "step": 1165 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.00018779057591623038, |
| "loss": 0.0932, |
| "step": 1166 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.0001877801047120419, |
| "loss": 0.098, |
| "step": 1167 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.00018776963350785342, |
| "loss": 0.0765, |
| "step": 1168 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.00018775916230366494, |
| "loss": 0.1515, |
| "step": 1169 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.00018774869109947645, |
| "loss": 0.1255, |
| "step": 1170 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.00018773821989528797, |
| "loss": 0.1039, |
| "step": 1171 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.0001877277486910995, |
| "loss": 0.1148, |
| "step": 1172 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.000187717277486911, |
| "loss": 0.0952, |
| "step": 1173 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 0.00018770680628272252, |
| "loss": 0.1105, |
| "step": 1174 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.00018769633507853404, |
| "loss": 0.0927, |
| "step": 1175 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.00018768586387434556, |
| "loss": 0.108, |
| "step": 1176 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.00018767539267015707, |
| "loss": 0.0811, |
| "step": 1177 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.0001876649214659686, |
| "loss": 0.126, |
| "step": 1178 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.00018765445026178013, |
| "loss": 0.1081, |
| "step": 1179 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.00018764397905759162, |
| "loss": 0.0976, |
| "step": 1180 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.00018763350785340314, |
| "loss": 0.1377, |
| "step": 1181 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.00018762303664921469, |
| "loss": 0.1208, |
| "step": 1182 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.0001876125654450262, |
| "loss": 0.1006, |
| "step": 1183 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.0001876020942408377, |
| "loss": 0.1165, |
| "step": 1184 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.00018759162303664924, |
| "loss": 0.1217, |
| "step": 1185 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.00018758115183246075, |
| "loss": 0.1334, |
| "step": 1186 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.00018757068062827224, |
| "loss": 0.1088, |
| "step": 1187 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.0001875602094240838, |
| "loss": 0.1044, |
| "step": 1188 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.0001875497382198953, |
| "loss": 0.1095, |
| "step": 1189 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.00018753926701570682, |
| "loss": 0.1297, |
| "step": 1190 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.00018752879581151834, |
| "loss": 0.1178, |
| "step": 1191 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.00018751832460732986, |
| "loss": 0.1022, |
| "step": 1192 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 0.00018750785340314137, |
| "loss": 0.0994, |
| "step": 1193 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.0001874973821989529, |
| "loss": 0.1399, |
| "step": 1194 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.0001874869109947644, |
| "loss": 0.1169, |
| "step": 1195 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.00018747643979057592, |
| "loss": 0.0976, |
| "step": 1196 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.00018746596858638744, |
| "loss": 0.1124, |
| "step": 1197 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.00018745549738219896, |
| "loss": 0.0869, |
| "step": 1198 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.00018744502617801048, |
| "loss": 0.135, |
| "step": 1199 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.00018743455497382202, |
| "loss": 0.0846, |
| "step": 1200 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.0001874240837696335, |
| "loss": 0.1041, |
| "step": 1201 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.00018741361256544503, |
| "loss": 0.1319, |
| "step": 1202 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.00018740314136125657, |
| "loss": 0.1079, |
| "step": 1203 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.00018739267015706806, |
| "loss": 0.0713, |
| "step": 1204 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.00018738219895287958, |
| "loss": 0.1327, |
| "step": 1205 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.00018737172774869112, |
| "loss": 0.0626, |
| "step": 1206 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.00018736125654450264, |
| "loss": 0.1374, |
| "step": 1207 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.00018735078534031413, |
| "loss": 0.1029, |
| "step": 1208 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.00018734031413612567, |
| "loss": 0.128, |
| "step": 1209 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.0001873298429319372, |
| "loss": 0.1035, |
| "step": 1210 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.00018731937172774868, |
| "loss": 0.1122, |
| "step": 1211 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 0.00018730890052356022, |
| "loss": 0.1029, |
| "step": 1212 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.00018729842931937174, |
| "loss": 0.0885, |
| "step": 1213 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.00018728795811518326, |
| "loss": 0.1258, |
| "step": 1214 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.00018727748691099478, |
| "loss": 0.0897, |
| "step": 1215 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.0001872670157068063, |
| "loss": 0.1342, |
| "step": 1216 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.0001872565445026178, |
| "loss": 0.1036, |
| "step": 1217 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.00018724607329842933, |
| "loss": 0.0914, |
| "step": 1218 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.00018723560209424084, |
| "loss": 0.1114, |
| "step": 1219 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.00018722513089005236, |
| "loss": 0.0979, |
| "step": 1220 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.00018721465968586388, |
| "loss": 0.1239, |
| "step": 1221 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.00018721465968586388, |
| "loss": 0.1324, |
| "step": 1222 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.0001872041884816754, |
| "loss": 0.0963, |
| "step": 1223 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.0001871937172774869, |
| "loss": 0.07, |
| "step": 1224 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.00018718324607329846, |
| "loss": 0.1117, |
| "step": 1225 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.00018717277486910995, |
| "loss": 0.085, |
| "step": 1226 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.00018716230366492146, |
| "loss": 0.1158, |
| "step": 1227 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.000187151832460733, |
| "loss": 0.0789, |
| "step": 1228 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.00018714136125654452, |
| "loss": 0.1343, |
| "step": 1229 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.00018713089005235601, |
| "loss": 0.156, |
| "step": 1230 |
| }, |
| { |
| "epoch": 0.64, |
| "learning_rate": 0.00018712041884816756, |
| "loss": 0.1134, |
| "step": 1231 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.00018710994764397908, |
| "loss": 0.0989, |
| "step": 1232 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.00018709947643979057, |
| "loss": 0.0735, |
| "step": 1233 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.0001870890052356021, |
| "loss": 0.0984, |
| "step": 1234 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.00018707853403141363, |
| "loss": 0.1307, |
| "step": 1235 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.00018706806282722514, |
| "loss": 0.1375, |
| "step": 1236 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.00018705759162303666, |
| "loss": 0.0765, |
| "step": 1237 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.00018704712041884818, |
| "loss": 0.1267, |
| "step": 1238 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.0001870366492146597, |
| "loss": 0.1015, |
| "step": 1239 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.0001870261780104712, |
| "loss": 0.1004, |
| "step": 1240 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.00018701570680628273, |
| "loss": 0.1278, |
| "step": 1241 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.00018700523560209425, |
| "loss": 0.0986, |
| "step": 1242 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.00018699476439790576, |
| "loss": 0.103, |
| "step": 1243 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.00018698429319371728, |
| "loss": 0.0998, |
| "step": 1244 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.0001869738219895288, |
| "loss": 0.1019, |
| "step": 1245 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.00018696335078534034, |
| "loss": 0.1265, |
| "step": 1246 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.00018695287958115183, |
| "loss": 0.0975, |
| "step": 1247 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.00018694240837696335, |
| "loss": 0.0975, |
| "step": 1248 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.0001869319371727749, |
| "loss": 0.152, |
| "step": 1249 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.00018692146596858638, |
| "loss": 0.062, |
| "step": 1250 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 0.0001869109947643979, |
| "loss": 0.1286, |
| "step": 1251 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.00018690052356020944, |
| "loss": 0.1199, |
| "step": 1252 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.00018689005235602096, |
| "loss": 0.1334, |
| "step": 1253 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.00018687958115183245, |
| "loss": 0.0853, |
| "step": 1254 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.000186869109947644, |
| "loss": 0.1009, |
| "step": 1255 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.0001868586387434555, |
| "loss": 0.115, |
| "step": 1256 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.000186848167539267, |
| "loss": 0.0803, |
| "step": 1257 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.00018683769633507855, |
| "loss": 0.1234, |
| "step": 1258 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.00018682722513089006, |
| "loss": 0.1043, |
| "step": 1259 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.00018681675392670158, |
| "loss": 0.135, |
| "step": 1260 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.0001868062827225131, |
| "loss": 0.0895, |
| "step": 1261 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.00018679581151832462, |
| "loss": 0.0818, |
| "step": 1262 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.00018678534031413613, |
| "loss": 0.0889, |
| "step": 1263 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.00018677486910994765, |
| "loss": 0.0868, |
| "step": 1264 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.00018676439790575917, |
| "loss": 0.1124, |
| "step": 1265 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.00018675392670157068, |
| "loss": 0.1316, |
| "step": 1266 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.0001867434554973822, |
| "loss": 0.0695, |
| "step": 1267 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.00018673298429319372, |
| "loss": 0.1221, |
| "step": 1268 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.00018672251308900523, |
| "loss": 0.1345, |
| "step": 1269 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 0.00018671204188481678, |
| "loss": 0.118, |
| "step": 1270 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.00018670157068062827, |
| "loss": 0.0895, |
| "step": 1271 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.00018669109947643979, |
| "loss": 0.0984, |
| "step": 1272 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.00018668062827225133, |
| "loss": 0.083, |
| "step": 1273 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.00018667015706806285, |
| "loss": 0.1065, |
| "step": 1274 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.00018665968586387434, |
| "loss": 0.111, |
| "step": 1275 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.00018664921465968588, |
| "loss": 0.1127, |
| "step": 1276 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.0001866387434554974, |
| "loss": 0.1088, |
| "step": 1277 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.0001866282722513089, |
| "loss": 0.0929, |
| "step": 1278 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.00018661780104712043, |
| "loss": 0.0832, |
| "step": 1279 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.00018660732984293195, |
| "loss": 0.075, |
| "step": 1280 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.00018659685863874347, |
| "loss": 0.0767, |
| "step": 1281 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.00018658638743455498, |
| "loss": 0.0902, |
| "step": 1282 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.0001865759162303665, |
| "loss": 0.0935, |
| "step": 1283 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.00018656544502617802, |
| "loss": 0.1193, |
| "step": 1284 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.00018655497382198954, |
| "loss": 0.1261, |
| "step": 1285 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.00018654450261780105, |
| "loss": 0.0867, |
| "step": 1286 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.00018653403141361257, |
| "loss": 0.1003, |
| "step": 1287 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.0001865235602094241, |
| "loss": 0.0927, |
| "step": 1288 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 0.0001865130890052356, |
| "loss": 0.085, |
| "step": 1289 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.00018650261780104712, |
| "loss": 0.0992, |
| "step": 1290 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.00018649214659685866, |
| "loss": 0.1055, |
| "step": 1291 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.00018648167539267015, |
| "loss": 0.0761, |
| "step": 1292 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.00018647120418848167, |
| "loss": 0.072, |
| "step": 1293 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.00018646073298429322, |
| "loss": 0.1158, |
| "step": 1294 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.0001864502617801047, |
| "loss": 0.1149, |
| "step": 1295 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.00018643979057591622, |
| "loss": 0.0968, |
| "step": 1296 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.00018642931937172777, |
| "loss": 0.1005, |
| "step": 1297 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.00018641884816753928, |
| "loss": 0.1134, |
| "step": 1298 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.00018640837696335077, |
| "loss": 0.0795, |
| "step": 1299 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.00018639790575916232, |
| "loss": 0.1031, |
| "step": 1300 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.00018638743455497384, |
| "loss": 0.1112, |
| "step": 1301 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.00018637696335078533, |
| "loss": 0.1065, |
| "step": 1302 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.00018636649214659687, |
| "loss": 0.076, |
| "step": 1303 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.0001863560209424084, |
| "loss": 0.0888, |
| "step": 1304 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.0001863455497382199, |
| "loss": 0.1196, |
| "step": 1305 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.00018633507853403142, |
| "loss": 0.0974, |
| "step": 1306 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.00018632460732984294, |
| "loss": 0.0959, |
| "step": 1307 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 0.00018631413612565446, |
| "loss": 0.1028, |
| "step": 1308 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.00018630366492146597, |
| "loss": 0.0868, |
| "step": 1309 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.0001862931937172775, |
| "loss": 0.0824, |
| "step": 1310 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.00018628272251308903, |
| "loss": 0.084, |
| "step": 1311 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.00018627225130890052, |
| "loss": 0.1233, |
| "step": 1312 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.00018626178010471204, |
| "loss": 0.0902, |
| "step": 1313 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.00018625130890052358, |
| "loss": 0.0985, |
| "step": 1314 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.0001862408376963351, |
| "loss": 0.0773, |
| "step": 1315 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.0001862303664921466, |
| "loss": 0.0933, |
| "step": 1316 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.00018621989528795814, |
| "loss": 0.0957, |
| "step": 1317 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.00018620942408376965, |
| "loss": 0.1089, |
| "step": 1318 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.00018619895287958117, |
| "loss": 0.0736, |
| "step": 1319 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.0001861884816753927, |
| "loss": 0.1263, |
| "step": 1320 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.0001861780104712042, |
| "loss": 0.1237, |
| "step": 1321 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.00018616753926701572, |
| "loss": 0.0962, |
| "step": 1322 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.00018615706806282724, |
| "loss": 0.127, |
| "step": 1323 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.00018614659685863876, |
| "loss": 0.1194, |
| "step": 1324 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.00018613612565445027, |
| "loss": 0.1321, |
| "step": 1325 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.0001861256544502618, |
| "loss": 0.0985, |
| "step": 1326 |
| }, |
| { |
| "epoch": 0.69, |
| "learning_rate": 0.0001861151832460733, |
| "loss": 0.086, |
| "step": 1327 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.00018610471204188482, |
| "loss": 0.0842, |
| "step": 1328 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.00018609424083769634, |
| "loss": 0.0866, |
| "step": 1329 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.00018608376963350786, |
| "loss": 0.1083, |
| "step": 1330 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.00018607329842931938, |
| "loss": 0.1274, |
| "step": 1331 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.00018606282722513092, |
| "loss": 0.1294, |
| "step": 1332 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.0001860523560209424, |
| "loss": 0.1, |
| "step": 1333 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.00018604188481675393, |
| "loss": 0.0974, |
| "step": 1334 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.00018603141361256547, |
| "loss": 0.1256, |
| "step": 1335 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.000186020942408377, |
| "loss": 0.0859, |
| "step": 1336 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.00018601047120418848, |
| "loss": 0.0969, |
| "step": 1337 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.00018600000000000002, |
| "loss": 0.0991, |
| "step": 1338 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.00018598952879581154, |
| "loss": 0.1271, |
| "step": 1339 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.00018597905759162303, |
| "loss": 0.1099, |
| "step": 1340 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.00018596858638743457, |
| "loss": 0.1065, |
| "step": 1341 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.0001859581151832461, |
| "loss": 0.098, |
| "step": 1342 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.0001859476439790576, |
| "loss": 0.0672, |
| "step": 1343 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.00018593717277486912, |
| "loss": 0.0875, |
| "step": 1344 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.00018592670157068064, |
| "loss": 0.1241, |
| "step": 1345 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 0.00018591623036649216, |
| "loss": 0.1061, |
| "step": 1346 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.00018590575916230368, |
| "loss": 0.1351, |
| "step": 1347 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.0001858952879581152, |
| "loss": 0.1073, |
| "step": 1348 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.0001858848167539267, |
| "loss": 0.1155, |
| "step": 1349 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.00018587434554973823, |
| "loss": 0.113, |
| "step": 1350 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.00018586387434554974, |
| "loss": 0.1113, |
| "step": 1351 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.00018585340314136126, |
| "loss": 0.1468, |
| "step": 1352 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.0001858429319371728, |
| "loss": 0.0923, |
| "step": 1353 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.0001858324607329843, |
| "loss": 0.0867, |
| "step": 1354 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.0001858219895287958, |
| "loss": 0.1254, |
| "step": 1355 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.00018581151832460736, |
| "loss": 0.1008, |
| "step": 1356 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.00018580104712041885, |
| "loss": 0.1046, |
| "step": 1357 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.00018579057591623036, |
| "loss": 0.0668, |
| "step": 1358 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.0001857801047120419, |
| "loss": 0.115, |
| "step": 1359 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.00018576963350785342, |
| "loss": 0.1103, |
| "step": 1360 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.00018575916230366491, |
| "loss": 0.0961, |
| "step": 1361 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.00018574869109947646, |
| "loss": 0.0879, |
| "step": 1362 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.00018573821989528798, |
| "loss": 0.0824, |
| "step": 1363 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.00018572774869109947, |
| "loss": 0.1167, |
| "step": 1364 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 0.000185717277486911, |
| "loss": 0.1049, |
| "step": 1365 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.00018570680628272253, |
| "loss": 0.0923, |
| "step": 1366 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.00018569633507853404, |
| "loss": 0.1516, |
| "step": 1367 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.00018568586387434556, |
| "loss": 0.14, |
| "step": 1368 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.00018567539267015708, |
| "loss": 0.1145, |
| "step": 1369 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.0001856649214659686, |
| "loss": 0.0681, |
| "step": 1370 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.0001856544502617801, |
| "loss": 0.1209, |
| "step": 1371 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.00018564397905759163, |
| "loss": 0.1075, |
| "step": 1372 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.00018563350785340315, |
| "loss": 0.1255, |
| "step": 1373 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.00018562303664921466, |
| "loss": 0.1123, |
| "step": 1374 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.00018561256544502618, |
| "loss": 0.1318, |
| "step": 1375 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.0001856020942408377, |
| "loss": 0.1033, |
| "step": 1376 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.00018559162303664924, |
| "loss": 0.0981, |
| "step": 1377 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.00018558115183246073, |
| "loss": 0.117, |
| "step": 1378 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.00018557068062827225, |
| "loss": 0.082, |
| "step": 1379 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.0001855602094240838, |
| "loss": 0.1087, |
| "step": 1380 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.0001855497382198953, |
| "loss": 0.1008, |
| "step": 1381 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.0001855392670157068, |
| "loss": 0.1083, |
| "step": 1382 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.00018552879581151834, |
| "loss": 0.1006, |
| "step": 1383 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 0.00018551832460732986, |
| "loss": 0.1019, |
| "step": 1384 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.00018550785340314135, |
| "loss": 0.0913, |
| "step": 1385 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.0001854973821989529, |
| "loss": 0.0727, |
| "step": 1386 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.0001854869109947644, |
| "loss": 0.1288, |
| "step": 1387 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.00018547643979057593, |
| "loss": 0.1186, |
| "step": 1388 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.00018546596858638745, |
| "loss": 0.1424, |
| "step": 1389 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.00018545549738219896, |
| "loss": 0.1072, |
| "step": 1390 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.00018544502617801048, |
| "loss": 0.1135, |
| "step": 1391 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.000185434554973822, |
| "loss": 0.1424, |
| "step": 1392 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.00018542408376963352, |
| "loss": 0.0836, |
| "step": 1393 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.00018541361256544503, |
| "loss": 0.1315, |
| "step": 1394 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.00018540314136125655, |
| "loss": 0.1171, |
| "step": 1395 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.00018539267015706807, |
| "loss": 0.1124, |
| "step": 1396 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.00018538219895287958, |
| "loss": 0.0931, |
| "step": 1397 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.00018537172774869113, |
| "loss": 0.0852, |
| "step": 1398 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.00018536125654450262, |
| "loss": 0.0732, |
| "step": 1399 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.00018535078534031413, |
| "loss": 0.1002, |
| "step": 1400 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.00018534031413612568, |
| "loss": 0.0724, |
| "step": 1401 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.00018532984293193717, |
| "loss": 0.0842, |
| "step": 1402 |
| }, |
| { |
| "epoch": 0.73, |
| "learning_rate": 0.00018531937172774869, |
| "loss": 0.0886, |
| "step": 1403 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.00018530890052356023, |
| "loss": 0.1045, |
| "step": 1404 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.00018529842931937175, |
| "loss": 0.128, |
| "step": 1405 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.00018528795811518324, |
| "loss": 0.0828, |
| "step": 1406 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.00018527748691099478, |
| "loss": 0.0829, |
| "step": 1407 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.0001852670157068063, |
| "loss": 0.151, |
| "step": 1408 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.0001852565445026178, |
| "loss": 0.0955, |
| "step": 1409 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.00018524607329842933, |
| "loss": 0.0887, |
| "step": 1410 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.00018523560209424085, |
| "loss": 0.1269, |
| "step": 1411 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.00018522513089005237, |
| "loss": 0.1185, |
| "step": 1412 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.00018521465968586388, |
| "loss": 0.1171, |
| "step": 1413 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.0001852041884816754, |
| "loss": 0.0933, |
| "step": 1414 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.00018519371727748692, |
| "loss": 0.1108, |
| "step": 1415 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.00018518324607329843, |
| "loss": 0.0887, |
| "step": 1416 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.00018517277486910995, |
| "loss": 0.1047, |
| "step": 1417 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.00018516230366492147, |
| "loss": 0.145, |
| "step": 1418 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.00018515183246073299, |
| "loss": 0.1135, |
| "step": 1419 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.0001851413612565445, |
| "loss": 0.0971, |
| "step": 1420 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.00018513089005235602, |
| "loss": 0.0979, |
| "step": 1421 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 0.00018512041884816756, |
| "loss": 0.1243, |
| "step": 1422 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.00018510994764397905, |
| "loss": 0.0909, |
| "step": 1423 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.00018509947643979057, |
| "loss": 0.1001, |
| "step": 1424 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.00018508900523560212, |
| "loss": 0.1299, |
| "step": 1425 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.00018507853403141363, |
| "loss": 0.1148, |
| "step": 1426 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.00018506806282722512, |
| "loss": 0.107, |
| "step": 1427 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.00018505759162303667, |
| "loss": 0.0916, |
| "step": 1428 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.00018504712041884818, |
| "loss": 0.0826, |
| "step": 1429 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.00018503664921465967, |
| "loss": 0.094, |
| "step": 1430 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.00018502617801047122, |
| "loss": 0.1244, |
| "step": 1431 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.00018501570680628274, |
| "loss": 0.1406, |
| "step": 1432 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.00018500523560209425, |
| "loss": 0.08, |
| "step": 1433 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.00018499476439790577, |
| "loss": 0.0779, |
| "step": 1434 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.0001849842931937173, |
| "loss": 0.1138, |
| "step": 1435 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.0001849738219895288, |
| "loss": 0.1092, |
| "step": 1436 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.00018496335078534032, |
| "loss": 0.094, |
| "step": 1437 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.00018495287958115184, |
| "loss": 0.104, |
| "step": 1438 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.00018494240837696335, |
| "loss": 0.0837, |
| "step": 1439 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.00018493193717277487, |
| "loss": 0.1157, |
| "step": 1440 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.0001849214659685864, |
| "loss": 0.0732, |
| "step": 1441 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 0.0001849109947643979, |
| "loss": 0.0744, |
| "step": 1442 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.00018490052356020945, |
| "loss": 0.1244, |
| "step": 1443 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.00018489005235602094, |
| "loss": 0.106, |
| "step": 1444 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.00018487958115183246, |
| "loss": 0.1019, |
| "step": 1445 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.000184869109947644, |
| "loss": 0.0976, |
| "step": 1446 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.0001848586387434555, |
| "loss": 0.1404, |
| "step": 1447 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.000184848167539267, |
| "loss": 0.0708, |
| "step": 1448 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.00018483769633507855, |
| "loss": 0.0816, |
| "step": 1449 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.00018482722513089007, |
| "loss": 0.0797, |
| "step": 1450 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.00018481675392670156, |
| "loss": 0.0801, |
| "step": 1451 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.0001848062827225131, |
| "loss": 0.0743, |
| "step": 1452 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.00018479581151832462, |
| "loss": 0.1347, |
| "step": 1453 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.0001847853403141361, |
| "loss": 0.102, |
| "step": 1454 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.00018477486910994766, |
| "loss": 0.0815, |
| "step": 1455 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.00018476439790575917, |
| "loss": 0.098, |
| "step": 1456 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.0001847539267015707, |
| "loss": 0.1122, |
| "step": 1457 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.0001847434554973822, |
| "loss": 0.1183, |
| "step": 1458 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.00018473298429319372, |
| "loss": 0.0703, |
| "step": 1459 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.00018472251308900524, |
| "loss": 0.0672, |
| "step": 1460 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 0.00018471204188481676, |
| "loss": 0.1077, |
| "step": 1461 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.00018470157068062827, |
| "loss": 0.1231, |
| "step": 1462 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.0001846910994764398, |
| "loss": 0.1165, |
| "step": 1463 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.0001846806282722513, |
| "loss": 0.107, |
| "step": 1464 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.00018467015706806283, |
| "loss": 0.1151, |
| "step": 1465 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.00018465968586387437, |
| "loss": 0.0872, |
| "step": 1466 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.0001846492146596859, |
| "loss": 0.1005, |
| "step": 1467 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.00018463874345549738, |
| "loss": 0.088, |
| "step": 1468 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.00018462827225130892, |
| "loss": 0.0772, |
| "step": 1469 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.00018461780104712044, |
| "loss": 0.09, |
| "step": 1470 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.00018460732984293196, |
| "loss": 0.078, |
| "step": 1471 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.00018459685863874347, |
| "loss": 0.1316, |
| "step": 1472 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.000184586387434555, |
| "loss": 0.1011, |
| "step": 1473 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.0001845759162303665, |
| "loss": 0.1152, |
| "step": 1474 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.00018456544502617802, |
| "loss": 0.0778, |
| "step": 1475 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.00018455497382198954, |
| "loss": 0.1362, |
| "step": 1476 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.00018454450261780106, |
| "loss": 0.1159, |
| "step": 1477 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.00018453403141361258, |
| "loss": 0.1178, |
| "step": 1478 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.0001845235602094241, |
| "loss": 0.0916, |
| "step": 1479 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 0.0001845130890052356, |
| "loss": 0.1032, |
| "step": 1480 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.00018450261780104713, |
| "loss": 0.1043, |
| "step": 1481 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.00018449214659685864, |
| "loss": 0.0889, |
| "step": 1482 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.00018448167539267016, |
| "loss": 0.0975, |
| "step": 1483 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.0001844712041884817, |
| "loss": 0.0721, |
| "step": 1484 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.0001844607329842932, |
| "loss": 0.0906, |
| "step": 1485 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.0001844502617801047, |
| "loss": 0.1057, |
| "step": 1486 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.00018443979057591626, |
| "loss": 0.1231, |
| "step": 1487 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.00018442931937172777, |
| "loss": 0.0894, |
| "step": 1488 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.00018441884816753926, |
| "loss": 0.1116, |
| "step": 1489 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.0001844083769633508, |
| "loss": 0.0899, |
| "step": 1490 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.00018439790575916232, |
| "loss": 0.0908, |
| "step": 1491 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.00018438743455497381, |
| "loss": 0.1646, |
| "step": 1492 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.00018437696335078536, |
| "loss": 0.1188, |
| "step": 1493 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.00018436649214659688, |
| "loss": 0.1022, |
| "step": 1494 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.0001843560209424084, |
| "loss": 0.1759, |
| "step": 1495 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.0001843455497382199, |
| "loss": 0.0977, |
| "step": 1496 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.00018433507853403143, |
| "loss": 0.0954, |
| "step": 1497 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.00018432460732984294, |
| "loss": 0.1113, |
| "step": 1498 |
| }, |
| { |
| "epoch": 0.78, |
| "learning_rate": 0.00018431413612565446, |
| "loss": 0.0883, |
| "step": 1499 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.00018430366492146598, |
| "loss": 0.0792, |
| "step": 1500 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.0001842931937172775, |
| "loss": 0.1174, |
| "step": 1501 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.000184282722513089, |
| "loss": 0.1327, |
| "step": 1502 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.00018427225130890053, |
| "loss": 0.1028, |
| "step": 1503 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.00018426178010471205, |
| "loss": 0.1135, |
| "step": 1504 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.0001842513089005236, |
| "loss": 0.1379, |
| "step": 1505 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.00018424083769633508, |
| "loss": 0.1218, |
| "step": 1506 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.0001842303664921466, |
| "loss": 0.0939, |
| "step": 1507 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.00018421989528795814, |
| "loss": 0.0963, |
| "step": 1508 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.00018420942408376963, |
| "loss": 0.0721, |
| "step": 1509 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.00018419895287958115, |
| "loss": 0.0965, |
| "step": 1510 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.0001841884816753927, |
| "loss": 0.1051, |
| "step": 1511 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.0001841780104712042, |
| "loss": 0.095, |
| "step": 1512 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.0001841675392670157, |
| "loss": 0.0844, |
| "step": 1513 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.00018415706806282724, |
| "loss": 0.0791, |
| "step": 1514 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.00018414659685863876, |
| "loss": 0.0706, |
| "step": 1515 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.00018413612565445025, |
| "loss": 0.1009, |
| "step": 1516 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.0001841256544502618, |
| "loss": 0.1128, |
| "step": 1517 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 0.0001841151832460733, |
| "loss": 0.1624, |
| "step": 1518 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018410471204188483, |
| "loss": 0.1182, |
| "step": 1519 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018409424083769635, |
| "loss": 0.0566, |
| "step": 1520 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018408376963350786, |
| "loss": 0.0911, |
| "step": 1521 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018407329842931938, |
| "loss": 0.0985, |
| "step": 1522 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.0001840628272251309, |
| "loss": 0.099, |
| "step": 1523 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018405235602094241, |
| "loss": 0.1076, |
| "step": 1524 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018404188481675393, |
| "loss": 0.0936, |
| "step": 1525 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018403141361256545, |
| "loss": 0.1086, |
| "step": 1526 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018402094240837697, |
| "loss": 0.0981, |
| "step": 1527 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018401047120418848, |
| "loss": 0.1076, |
| "step": 1528 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018400000000000003, |
| "loss": 0.1159, |
| "step": 1529 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018398952879581152, |
| "loss": 0.1392, |
| "step": 1530 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018397905759162303, |
| "loss": 0.119, |
| "step": 1531 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018396858638743458, |
| "loss": 0.1047, |
| "step": 1532 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.0001839581151832461, |
| "loss": 0.121, |
| "step": 1533 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018394764397905759, |
| "loss": 0.1039, |
| "step": 1534 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018393717277486913, |
| "loss": 0.1305, |
| "step": 1535 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018392670157068065, |
| "loss": 0.0919, |
| "step": 1536 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 0.00018391623036649214, |
| "loss": 0.095, |
| "step": 1537 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.00018390575916230368, |
| "loss": 0.1354, |
| "step": 1538 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.0001838952879581152, |
| "loss": 0.0895, |
| "step": 1539 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.00018388481675392672, |
| "loss": 0.0824, |
| "step": 1540 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.00018387434554973823, |
| "loss": 0.1588, |
| "step": 1541 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.00018386387434554975, |
| "loss": 0.1235, |
| "step": 1542 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.00018385340314136127, |
| "loss": 0.0932, |
| "step": 1543 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.00018384293193717278, |
| "loss": 0.0794, |
| "step": 1544 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.0001838324607329843, |
| "loss": 0.0776, |
| "step": 1545 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.00018382198952879582, |
| "loss": 0.1276, |
| "step": 1546 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.00018381151832460733, |
| "loss": 0.0982, |
| "step": 1547 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.00018380104712041885, |
| "loss": 0.1076, |
| "step": 1548 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.00018379057591623037, |
| "loss": 0.1478, |
| "step": 1549 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.0001837801047120419, |
| "loss": 0.0903, |
| "step": 1550 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.0001837696335078534, |
| "loss": 0.0902, |
| "step": 1551 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.00018375916230366492, |
| "loss": 0.0893, |
| "step": 1552 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.00018374869109947646, |
| "loss": 0.1127, |
| "step": 1553 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.00018373821989528795, |
| "loss": 0.1048, |
| "step": 1554 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.00018372774869109947, |
| "loss": 0.1, |
| "step": 1555 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 0.00018371727748691102, |
| "loss": 0.1198, |
| "step": 1556 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.00018370680628272253, |
| "loss": 0.1077, |
| "step": 1557 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.00018369633507853402, |
| "loss": 0.0972, |
| "step": 1558 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.00018368586387434557, |
| "loss": 0.0781, |
| "step": 1559 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.00018367539267015708, |
| "loss": 0.1363, |
| "step": 1560 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.00018366492146596857, |
| "loss": 0.1419, |
| "step": 1561 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.00018365445026178012, |
| "loss": 0.0562, |
| "step": 1562 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.00018364397905759163, |
| "loss": 0.0959, |
| "step": 1563 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.00018363350785340315, |
| "loss": 0.1013, |
| "step": 1564 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.00018362303664921467, |
| "loss": 0.1102, |
| "step": 1565 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.00018361256544502619, |
| "loss": 0.1168, |
| "step": 1566 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.0001836020942408377, |
| "loss": 0.0922, |
| "step": 1567 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.00018359162303664922, |
| "loss": 0.1012, |
| "step": 1568 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.00018358115183246074, |
| "loss": 0.0887, |
| "step": 1569 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.00018357068062827225, |
| "loss": 0.1163, |
| "step": 1570 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.00018356020942408377, |
| "loss": 0.1063, |
| "step": 1571 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.0001835497382198953, |
| "loss": 0.1111, |
| "step": 1572 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.0001835392670157068, |
| "loss": 0.1229, |
| "step": 1573 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.00018352879581151835, |
| "loss": 0.0861, |
| "step": 1574 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 0.00018351832460732984, |
| "loss": 0.1138, |
| "step": 1575 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.00018350785340314136, |
| "loss": 0.0964, |
| "step": 1576 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.0001834973821989529, |
| "loss": 0.1121, |
| "step": 1577 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.00018348691099476442, |
| "loss": 0.0877, |
| "step": 1578 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.0001834764397905759, |
| "loss": 0.0723, |
| "step": 1579 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.00018346596858638745, |
| "loss": 0.0731, |
| "step": 1580 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.00018345549738219897, |
| "loss": 0.0965, |
| "step": 1581 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.00018344502617801046, |
| "loss": 0.1272, |
| "step": 1582 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.000183434554973822, |
| "loss": 0.0716, |
| "step": 1583 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.00018342408376963352, |
| "loss": 0.0995, |
| "step": 1584 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.00018341361256544504, |
| "loss": 0.0825, |
| "step": 1585 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.00018340314136125655, |
| "loss": 0.1164, |
| "step": 1586 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.00018339267015706807, |
| "loss": 0.1025, |
| "step": 1587 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.0001833821989528796, |
| "loss": 0.0761, |
| "step": 1588 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.0001833717277486911, |
| "loss": 0.0857, |
| "step": 1589 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.00018336125654450262, |
| "loss": 0.1263, |
| "step": 1590 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.00018335078534031414, |
| "loss": 0.1068, |
| "step": 1591 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.00018334031413612566, |
| "loss": 0.1256, |
| "step": 1592 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.00018332984293193717, |
| "loss": 0.0934, |
| "step": 1593 |
| }, |
| { |
| "epoch": 0.83, |
| "learning_rate": 0.0001833193717277487, |
| "loss": 0.0983, |
| "step": 1594 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.00018330890052356024, |
| "loss": 0.0725, |
| "step": 1595 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.00018329842931937173, |
| "loss": 0.0809, |
| "step": 1596 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.00018328795811518324, |
| "loss": 0.0829, |
| "step": 1597 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.0001832774869109948, |
| "loss": 0.0745, |
| "step": 1598 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.00018326701570680628, |
| "loss": 0.1004, |
| "step": 1599 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.0001832565445026178, |
| "loss": 0.0887, |
| "step": 1600 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.00018324607329842934, |
| "loss": 0.1212, |
| "step": 1601 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.00018323560209424086, |
| "loss": 0.1033, |
| "step": 1602 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.00018322513089005235, |
| "loss": 0.062, |
| "step": 1603 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.0001832146596858639, |
| "loss": 0.0611, |
| "step": 1604 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.0001832041884816754, |
| "loss": 0.1057, |
| "step": 1605 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.0001831937172774869, |
| "loss": 0.1134, |
| "step": 1606 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.00018318324607329844, |
| "loss": 0.1196, |
| "step": 1607 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.00018317277486910996, |
| "loss": 0.1098, |
| "step": 1608 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.00018316230366492147, |
| "loss": 0.087, |
| "step": 1609 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.000183151832460733, |
| "loss": 0.0724, |
| "step": 1610 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.0001831413612565445, |
| "loss": 0.0677, |
| "step": 1611 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.00018313089005235603, |
| "loss": 0.0726, |
| "step": 1612 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 0.00018312041884816754, |
| "loss": 0.1005, |
| "step": 1613 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.00018310994764397906, |
| "loss": 0.1392, |
| "step": 1614 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.00018309947643979058, |
| "loss": 0.0995, |
| "step": 1615 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.0001830890052356021, |
| "loss": 0.1207, |
| "step": 1616 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.0001830785340314136, |
| "loss": 0.1068, |
| "step": 1617 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.00018306806282722513, |
| "loss": 0.0728, |
| "step": 1618 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.00018305759162303667, |
| "loss": 0.0869, |
| "step": 1619 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.00018304712041884816, |
| "loss": 0.119, |
| "step": 1620 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.0001830366492146597, |
| "loss": 0.1233, |
| "step": 1621 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.00018302617801047122, |
| "loss": 0.0791, |
| "step": 1622 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.00018301570680628274, |
| "loss": 0.0712, |
| "step": 1623 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.00018300523560209426, |
| "loss": 0.0788, |
| "step": 1624 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.00018299476439790578, |
| "loss": 0.0881, |
| "step": 1625 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.0001829842931937173, |
| "loss": 0.1579, |
| "step": 1626 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.0001829738219895288, |
| "loss": 0.0723, |
| "step": 1627 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.00018296335078534033, |
| "loss": 0.0987, |
| "step": 1628 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.00018295287958115184, |
| "loss": 0.0937, |
| "step": 1629 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.00018294240837696336, |
| "loss": 0.0762, |
| "step": 1630 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.00018293193717277488, |
| "loss": 0.0902, |
| "step": 1631 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.0001829214659685864, |
| "loss": 0.0967, |
| "step": 1632 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 0.0001829109947643979, |
| "loss": 0.0789, |
| "step": 1633 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.00018290052356020943, |
| "loss": 0.1087, |
| "step": 1634 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.00018289005235602095, |
| "loss": 0.0766, |
| "step": 1635 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.0001828795811518325, |
| "loss": 0.1317, |
| "step": 1636 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.00018286910994764398, |
| "loss": 0.0891, |
| "step": 1637 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.0001828586387434555, |
| "loss": 0.0897, |
| "step": 1638 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.00018284816753926704, |
| "loss": 0.0828, |
| "step": 1639 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.00018283769633507856, |
| "loss": 0.0972, |
| "step": 1640 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.00018282722513089005, |
| "loss": 0.0621, |
| "step": 1641 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.0001828167539267016, |
| "loss": 0.0909, |
| "step": 1642 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.0001828062827225131, |
| "loss": 0.0709, |
| "step": 1643 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.0001827958115183246, |
| "loss": 0.0928, |
| "step": 1644 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.00018278534031413614, |
| "loss": 0.1047, |
| "step": 1645 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.00018277486910994766, |
| "loss": 0.0973, |
| "step": 1646 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.00018276439790575918, |
| "loss": 0.1009, |
| "step": 1647 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.0001827539267015707, |
| "loss": 0.0765, |
| "step": 1648 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.0001827434554973822, |
| "loss": 0.0912, |
| "step": 1649 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.00018273298429319373, |
| "loss": 0.122, |
| "step": 1650 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.00018272251308900525, |
| "loss": 0.1274, |
| "step": 1651 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 0.00018271204188481676, |
| "loss": 0.091, |
| "step": 1652 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.00018270157068062828, |
| "loss": 0.0862, |
| "step": 1653 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.0001826910994764398, |
| "loss": 0.0855, |
| "step": 1654 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.00018268062827225131, |
| "loss": 0.1361, |
| "step": 1655 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.00018267015706806283, |
| "loss": 0.0932, |
| "step": 1656 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.00018265968586387438, |
| "loss": 0.1249, |
| "step": 1657 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.00018264921465968587, |
| "loss": 0.1512, |
| "step": 1658 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.00018263874345549738, |
| "loss": 0.1064, |
| "step": 1659 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.00018262827225130893, |
| "loss": 0.1091, |
| "step": 1660 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.00018261780104712042, |
| "loss": 0.1152, |
| "step": 1661 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.00018260732984293193, |
| "loss": 0.0966, |
| "step": 1662 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.00018259685863874348, |
| "loss": 0.0874, |
| "step": 1663 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.000182586387434555, |
| "loss": 0.1, |
| "step": 1664 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.00018257591623036649, |
| "loss": 0.0984, |
| "step": 1665 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.00018256544502617803, |
| "loss": 0.1135, |
| "step": 1666 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.00018255497382198955, |
| "loss": 0.1183, |
| "step": 1667 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.00018254450261780104, |
| "loss": 0.1125, |
| "step": 1668 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.00018253403141361258, |
| "loss": 0.1041, |
| "step": 1669 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.0001825235602094241, |
| "loss": 0.0843, |
| "step": 1670 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 0.00018251308900523561, |
| "loss": 0.069, |
| "step": 1671 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.00018250261780104713, |
| "loss": 0.1333, |
| "step": 1672 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.00018249214659685865, |
| "loss": 0.1065, |
| "step": 1673 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.00018248167539267017, |
| "loss": 0.1251, |
| "step": 1674 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.00018247120418848168, |
| "loss": 0.0826, |
| "step": 1675 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.0001824607329842932, |
| "loss": 0.0954, |
| "step": 1676 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.00018245026178010472, |
| "loss": 0.1206, |
| "step": 1677 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.00018243979057591623, |
| "loss": 0.0837, |
| "step": 1678 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.00018242931937172775, |
| "loss": 0.1458, |
| "step": 1679 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.00018241884816753927, |
| "loss": 0.0779, |
| "step": 1680 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.0001824083769633508, |
| "loss": 0.091, |
| "step": 1681 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.0001823979057591623, |
| "loss": 0.1175, |
| "step": 1682 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.00018238743455497382, |
| "loss": 0.1097, |
| "step": 1683 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.00018237696335078536, |
| "loss": 0.1052, |
| "step": 1684 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.00018236649214659688, |
| "loss": 0.1042, |
| "step": 1685 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.00018235602094240837, |
| "loss": 0.0947, |
| "step": 1686 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.00018234554973821992, |
| "loss": 0.0928, |
| "step": 1687 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.00018233507853403143, |
| "loss": 0.0818, |
| "step": 1688 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.00018232460732984292, |
| "loss": 0.0728, |
| "step": 1689 |
| }, |
| { |
| "epoch": 0.88, |
| "learning_rate": 0.00018231413612565447, |
| "loss": 0.1066, |
| "step": 1690 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.00018230366492146598, |
| "loss": 0.1005, |
| "step": 1691 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.0001822931937172775, |
| "loss": 0.1144, |
| "step": 1692 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.00018228272251308902, |
| "loss": 0.0917, |
| "step": 1693 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.00018227225130890053, |
| "loss": 0.1358, |
| "step": 1694 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.00018226178010471205, |
| "loss": 0.1179, |
| "step": 1695 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.00018225130890052357, |
| "loss": 0.0775, |
| "step": 1696 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.00018224083769633509, |
| "loss": 0.1568, |
| "step": 1697 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.0001822303664921466, |
| "loss": 0.0759, |
| "step": 1698 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.00018221989528795812, |
| "loss": 0.0949, |
| "step": 1699 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.00018220942408376964, |
| "loss": 0.0658, |
| "step": 1700 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.00018219895287958115, |
| "loss": 0.1075, |
| "step": 1701 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.0001821884816753927, |
| "loss": 0.1172, |
| "step": 1702 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.0001821780104712042, |
| "loss": 0.0947, |
| "step": 1703 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.0001821675392670157, |
| "loss": 0.1198, |
| "step": 1704 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.00018215706806282725, |
| "loss": 0.1026, |
| "step": 1705 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.00018214659685863874, |
| "loss": 0.0519, |
| "step": 1706 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.00018213612565445026, |
| "loss": 0.096, |
| "step": 1707 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.0001821256544502618, |
| "loss": 0.1172, |
| "step": 1708 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 0.00018211518324607332, |
| "loss": 0.1015, |
| "step": 1709 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.0001821047120418848, |
| "loss": 0.121, |
| "step": 1710 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.00018209424083769635, |
| "loss": 0.1131, |
| "step": 1711 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.00018208376963350787, |
| "loss": 0.0771, |
| "step": 1712 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.00018207329842931936, |
| "loss": 0.1169, |
| "step": 1713 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.0001820628272251309, |
| "loss": 0.1037, |
| "step": 1714 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.00018205235602094242, |
| "loss": 0.1173, |
| "step": 1715 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.00018204188481675394, |
| "loss": 0.0766, |
| "step": 1716 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.00018203141361256545, |
| "loss": 0.0942, |
| "step": 1717 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.00018202094240837697, |
| "loss": 0.1011, |
| "step": 1718 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.0001820104712041885, |
| "loss": 0.0761, |
| "step": 1719 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.000182, |
| "loss": 0.1093, |
| "step": 1720 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.00018198952879581152, |
| "loss": 0.0903, |
| "step": 1721 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.00018197905759162304, |
| "loss": 0.0902, |
| "step": 1722 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.00018196858638743456, |
| "loss": 0.0921, |
| "step": 1723 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.00018195811518324607, |
| "loss": 0.133, |
| "step": 1724 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.0001819476439790576, |
| "loss": 0.1286, |
| "step": 1725 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.00018193717277486914, |
| "loss": 0.1042, |
| "step": 1726 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.00018192670157068063, |
| "loss": 0.1123, |
| "step": 1727 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 0.00018191623036649214, |
| "loss": 0.0826, |
| "step": 1728 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.0001819057591623037, |
| "loss": 0.0941, |
| "step": 1729 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.0001818952879581152, |
| "loss": 0.0846, |
| "step": 1730 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.0001818848167539267, |
| "loss": 0.0711, |
| "step": 1731 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.00018187434554973824, |
| "loss": 0.0835, |
| "step": 1732 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.00018186387434554975, |
| "loss": 0.0769, |
| "step": 1733 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.00018185340314136124, |
| "loss": 0.082, |
| "step": 1734 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.0001818429319371728, |
| "loss": 0.1134, |
| "step": 1735 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.0001818324607329843, |
| "loss": 0.0699, |
| "step": 1736 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.00018182198952879582, |
| "loss": 0.1061, |
| "step": 1737 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.00018181151832460734, |
| "loss": 0.1075, |
| "step": 1738 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.00018180104712041886, |
| "loss": 0.0902, |
| "step": 1739 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.00018179057591623037, |
| "loss": 0.0915, |
| "step": 1740 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.0001817801047120419, |
| "loss": 0.0827, |
| "step": 1741 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.0001817696335078534, |
| "loss": 0.0992, |
| "step": 1742 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.00018175916230366493, |
| "loss": 0.0882, |
| "step": 1743 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.00018174869109947644, |
| "loss": 0.1101, |
| "step": 1744 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.00018173821989528796, |
| "loss": 0.089, |
| "step": 1745 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.00018172774869109948, |
| "loss": 0.0847, |
| "step": 1746 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 0.00018171727748691102, |
| "loss": 0.0655, |
| "step": 1747 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.0001817068062827225, |
| "loss": 0.1227, |
| "step": 1748 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.00018169633507853403, |
| "loss": 0.0781, |
| "step": 1749 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.00018168586387434557, |
| "loss": 0.0863, |
| "step": 1750 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.00018167539267015706, |
| "loss": 0.0712, |
| "step": 1751 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.00018166492146596858, |
| "loss": 0.0895, |
| "step": 1752 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.00018165445026178012, |
| "loss": 0.1087, |
| "step": 1753 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.00018164397905759164, |
| "loss": 0.1357, |
| "step": 1754 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.00018163350785340313, |
| "loss": 0.1386, |
| "step": 1755 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.00018162303664921467, |
| "loss": 0.0849, |
| "step": 1756 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.0001816125654450262, |
| "loss": 0.0843, |
| "step": 1757 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.00018160209424083768, |
| "loss": 0.1019, |
| "step": 1758 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.00018159162303664923, |
| "loss": 0.1211, |
| "step": 1759 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.00018158115183246074, |
| "loss": 0.1031, |
| "step": 1760 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.00018157068062827226, |
| "loss": 0.1173, |
| "step": 1761 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.00018156020942408378, |
| "loss": 0.1098, |
| "step": 1762 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.0001815497382198953, |
| "loss": 0.0977, |
| "step": 1763 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.0001815392670157068, |
| "loss": 0.0853, |
| "step": 1764 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.00018152879581151833, |
| "loss": 0.0849, |
| "step": 1765 |
| }, |
| { |
| "epoch": 0.92, |
| "learning_rate": 0.00018151832460732985, |
| "loss": 0.0827, |
| "step": 1766 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.00018150785340314136, |
| "loss": 0.0909, |
| "step": 1767 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.00018149738219895288, |
| "loss": 0.0694, |
| "step": 1768 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.0001814869109947644, |
| "loss": 0.1282, |
| "step": 1769 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.00018147643979057591, |
| "loss": 0.092, |
| "step": 1770 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.00018146596858638746, |
| "loss": 0.0682, |
| "step": 1771 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.00018145549738219895, |
| "loss": 0.0788, |
| "step": 1772 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.0001814450261780105, |
| "loss": 0.088, |
| "step": 1773 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.000181434554973822, |
| "loss": 0.0831, |
| "step": 1774 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.0001814240837696335, |
| "loss": 0.1276, |
| "step": 1775 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.00018141361256544504, |
| "loss": 0.08, |
| "step": 1776 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.00018140314136125656, |
| "loss": 0.0636, |
| "step": 1777 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.00018139267015706808, |
| "loss": 0.0945, |
| "step": 1778 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.0001813821989528796, |
| "loss": 0.0888, |
| "step": 1779 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.0001813717277486911, |
| "loss": 0.1071, |
| "step": 1780 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.00018136125654450263, |
| "loss": 0.0902, |
| "step": 1781 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.00018135078534031415, |
| "loss": 0.0728, |
| "step": 1782 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.00018134031413612566, |
| "loss": 0.0821, |
| "step": 1783 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.00018132984293193718, |
| "loss": 0.1063, |
| "step": 1784 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 0.0001813193717277487, |
| "loss": 0.0846, |
| "step": 1785 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018130890052356021, |
| "loss": 0.1033, |
| "step": 1786 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018129842931937173, |
| "loss": 0.0784, |
| "step": 1787 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018128795811518328, |
| "loss": 0.0755, |
| "step": 1788 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018127748691099477, |
| "loss": 0.0939, |
| "step": 1789 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018126701570680628, |
| "loss": 0.1193, |
| "step": 1790 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018125654450261783, |
| "loss": 0.0868, |
| "step": 1791 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018124607329842934, |
| "loss": 0.0722, |
| "step": 1792 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018123560209424083, |
| "loss": 0.0798, |
| "step": 1793 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018122513089005238, |
| "loss": 0.0792, |
| "step": 1794 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.0001812146596858639, |
| "loss": 0.061, |
| "step": 1795 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018120418848167538, |
| "loss": 0.0676, |
| "step": 1796 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018119371727748693, |
| "loss": 0.0951, |
| "step": 1797 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018118324607329845, |
| "loss": 0.0874, |
| "step": 1798 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018117277486910996, |
| "loss": 0.0848, |
| "step": 1799 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018116230366492148, |
| "loss": 0.0917, |
| "step": 1800 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.000181151832460733, |
| "loss": 0.0803, |
| "step": 1801 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018114136125654451, |
| "loss": 0.0435, |
| "step": 1802 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018113089005235603, |
| "loss": 0.0775, |
| "step": 1803 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 0.00018112041884816755, |
| "loss": 0.1081, |
| "step": 1804 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.00018110994764397907, |
| "loss": 0.0729, |
| "step": 1805 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.00018109947643979058, |
| "loss": 0.0955, |
| "step": 1806 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.0001810890052356021, |
| "loss": 0.0832, |
| "step": 1807 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.00018107853403141362, |
| "loss": 0.0766, |
| "step": 1808 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.00018106806282722516, |
| "loss": 0.0736, |
| "step": 1809 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.00018105759162303665, |
| "loss": 0.0844, |
| "step": 1810 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.00018104712041884817, |
| "loss": 0.0747, |
| "step": 1811 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.0001810366492146597, |
| "loss": 0.0824, |
| "step": 1812 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.0001810261780104712, |
| "loss": 0.0851, |
| "step": 1813 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.00018101570680628272, |
| "loss": 0.1205, |
| "step": 1814 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.00018100523560209426, |
| "loss": 0.059, |
| "step": 1815 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.00018099476439790578, |
| "loss": 0.0954, |
| "step": 1816 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.00018098429319371727, |
| "loss": 0.1035, |
| "step": 1817 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.00018097382198952881, |
| "loss": 0.0803, |
| "step": 1818 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.00018096335078534033, |
| "loss": 0.076, |
| "step": 1819 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.00018095287958115182, |
| "loss": 0.1121, |
| "step": 1820 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.00018094240837696337, |
| "loss": 0.0772, |
| "step": 1821 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.00018093193717277488, |
| "loss": 0.1251, |
| "step": 1822 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.0001809214659685864, |
| "loss": 0.1156, |
| "step": 1823 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 0.00018091099476439792, |
| "loss": 0.1104, |
| "step": 1824 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.00018090052356020943, |
| "loss": 0.0727, |
| "step": 1825 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.00018089005235602095, |
| "loss": 0.1049, |
| "step": 1826 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.00018087958115183247, |
| "loss": 0.0857, |
| "step": 1827 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.00018086910994764399, |
| "loss": 0.0873, |
| "step": 1828 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.0001808586387434555, |
| "loss": 0.1134, |
| "step": 1829 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.00018084816753926702, |
| "loss": 0.0767, |
| "step": 1830 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.00018083769633507854, |
| "loss": 0.0978, |
| "step": 1831 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.00018082722513089005, |
| "loss": 0.0995, |
| "step": 1832 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.0001808167539267016, |
| "loss": 0.1067, |
| "step": 1833 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.0001808062827225131, |
| "loss": 0.1133, |
| "step": 1834 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.0001807958115183246, |
| "loss": 0.1016, |
| "step": 1835 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.00018078534031413615, |
| "loss": 0.0818, |
| "step": 1836 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.00018077486910994767, |
| "loss": 0.0665, |
| "step": 1837 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.00018076439790575916, |
| "loss": 0.0855, |
| "step": 1838 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.0001807539267015707, |
| "loss": 0.1225, |
| "step": 1839 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.00018074345549738222, |
| "loss": 0.069, |
| "step": 1840 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.0001807329842931937, |
| "loss": 0.1052, |
| "step": 1841 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.00018072251308900525, |
| "loss": 0.057, |
| "step": 1842 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 0.00018071204188481677, |
| "loss": 0.0735, |
| "step": 1843 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.00018070157068062829, |
| "loss": 0.0745, |
| "step": 1844 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.0001806910994764398, |
| "loss": 0.1267, |
| "step": 1845 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.00018068062827225132, |
| "loss": 0.094, |
| "step": 1846 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.00018067015706806284, |
| "loss": 0.0963, |
| "step": 1847 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.00018065968586387435, |
| "loss": 0.0897, |
| "step": 1848 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.00018064921465968587, |
| "loss": 0.0714, |
| "step": 1849 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.0001806387434554974, |
| "loss": 0.1001, |
| "step": 1850 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.0001806282722513089, |
| "loss": 0.0809, |
| "step": 1851 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.00018061780104712042, |
| "loss": 0.0712, |
| "step": 1852 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.00018060732984293194, |
| "loss": 0.0738, |
| "step": 1853 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.00018059685863874348, |
| "loss": 0.0673, |
| "step": 1854 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.00018058638743455497, |
| "loss": 0.1039, |
| "step": 1855 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.0001805759162303665, |
| "loss": 0.0923, |
| "step": 1856 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.00018056544502617803, |
| "loss": 0.09, |
| "step": 1857 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.00018055497382198952, |
| "loss": 0.114, |
| "step": 1858 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.00018054450261780104, |
| "loss": 0.0782, |
| "step": 1859 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.00018053403141361259, |
| "loss": 0.1094, |
| "step": 1860 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.0001805235602094241, |
| "loss": 0.0893, |
| "step": 1861 |
| }, |
| { |
| "epoch": 0.97, |
| "learning_rate": 0.0001805130890052356, |
| "loss": 0.0808, |
| "step": 1862 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.00018050261780104714, |
| "loss": 0.0887, |
| "step": 1863 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.00018049214659685865, |
| "loss": 0.0771, |
| "step": 1864 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.00018048167539267014, |
| "loss": 0.0739, |
| "step": 1865 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.0001804712041884817, |
| "loss": 0.0711, |
| "step": 1866 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.0001804607329842932, |
| "loss": 0.0833, |
| "step": 1867 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.00018045026178010472, |
| "loss": 0.1198, |
| "step": 1868 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.00018043979057591624, |
| "loss": 0.0826, |
| "step": 1869 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.00018042931937172776, |
| "loss": 0.0506, |
| "step": 1870 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.00018041884816753927, |
| "loss": 0.0693, |
| "step": 1871 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.0001804083769633508, |
| "loss": 0.0857, |
| "step": 1872 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.0001803979057591623, |
| "loss": 0.0876, |
| "step": 1873 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.00018038743455497383, |
| "loss": 0.0871, |
| "step": 1874 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.00018037696335078534, |
| "loss": 0.0926, |
| "step": 1875 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.00018036649214659686, |
| "loss": 0.0687, |
| "step": 1876 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.00018035602094240838, |
| "loss": 0.1153, |
| "step": 1877 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.00018034554973821992, |
| "loss": 0.088, |
| "step": 1878 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.0001803350785340314, |
| "loss": 0.1041, |
| "step": 1879 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.00018032460732984293, |
| "loss": 0.1169, |
| "step": 1880 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 0.00018031413612565447, |
| "loss": 0.0905, |
| "step": 1881 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.000180303664921466, |
| "loss": 0.0773, |
| "step": 1882 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.00018029319371727748, |
| "loss": 0.0763, |
| "step": 1883 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.00018028272251308902, |
| "loss": 0.0903, |
| "step": 1884 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.00018027225130890054, |
| "loss": 0.1226, |
| "step": 1885 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.00018026178010471203, |
| "loss": 0.0943, |
| "step": 1886 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.00018025130890052357, |
| "loss": 0.0904, |
| "step": 1887 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.0001802408376963351, |
| "loss": 0.0915, |
| "step": 1888 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.0001802303664921466, |
| "loss": 0.1119, |
| "step": 1889 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.00018021989528795813, |
| "loss": 0.0889, |
| "step": 1890 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.00018020942408376964, |
| "loss": 0.0945, |
| "step": 1891 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.00018019895287958116, |
| "loss": 0.0853, |
| "step": 1892 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.00018018848167539268, |
| "loss": 0.0976, |
| "step": 1893 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.0001801780104712042, |
| "loss": 0.0862, |
| "step": 1894 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.0001801675392670157, |
| "loss": 0.0868, |
| "step": 1895 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.00018015706806282723, |
| "loss": 0.083, |
| "step": 1896 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.00018014659685863875, |
| "loss": 0.1187, |
| "step": 1897 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.00018013612565445026, |
| "loss": 0.089, |
| "step": 1898 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.0001801256544502618, |
| "loss": 0.1268, |
| "step": 1899 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 0.0001801151832460733, |
| "loss": 0.0882, |
| "step": 1900 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.0001801047120418848, |
| "loss": 0.0848, |
| "step": 1901 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.00018009424083769636, |
| "loss": 0.0722, |
| "step": 1902 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.00018008376963350785, |
| "loss": 0.0617, |
| "step": 1903 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.00018007329842931936, |
| "loss": 0.109, |
| "step": 1904 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.0001800628272251309, |
| "loss": 0.1081, |
| "step": 1905 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.00018005235602094243, |
| "loss": 0.0884, |
| "step": 1906 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.00018004188481675392, |
| "loss": 0.1093, |
| "step": 1907 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.00018003141361256546, |
| "loss": 0.0862, |
| "step": 1908 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.00018002094240837698, |
| "loss": 0.0963, |
| "step": 1909 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.00018001047120418847, |
| "loss": 0.0625, |
| "step": 1910 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.00018, |
| "loss": 0.097, |
| "step": 1911 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.00017998952879581153, |
| "loss": 0.0895, |
| "step": 1912 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.00017997905759162305, |
| "loss": 0.0831, |
| "step": 1913 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.00017996858638743456, |
| "loss": 0.061, |
| "step": 1914 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.00017995811518324608, |
| "loss": 0.0768, |
| "step": 1915 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.0001799476439790576, |
| "loss": 0.0787, |
| "step": 1916 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.00017993717277486911, |
| "loss": 0.0901, |
| "step": 1917 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.00017992670157068063, |
| "loss": 0.0825, |
| "step": 1918 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 0.00017991623036649215, |
| "loss": 0.0766, |
| "step": 1919 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.00017990575916230367, |
| "loss": 0.0789, |
| "step": 1920 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.00017989528795811518, |
| "loss": 0.1093, |
| "step": 1921 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.0001798848167539267, |
| "loss": 0.0597, |
| "step": 1922 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.00017987434554973824, |
| "loss": 0.0676, |
| "step": 1923 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.00017986387434554973, |
| "loss": 0.0745, |
| "step": 1924 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.00017985340314136125, |
| "loss": 0.0747, |
| "step": 1925 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.0001798429319371728, |
| "loss": 0.0712, |
| "step": 1926 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.00017983246073298428, |
| "loss": 0.0753, |
| "step": 1927 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.00017982198952879583, |
| "loss": 0.0617, |
| "step": 1928 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.00017981151832460735, |
| "loss": 0.0685, |
| "step": 1929 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.00017980104712041886, |
| "loss": 0.0805, |
| "step": 1930 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.00017979057591623038, |
| "loss": 0.0912, |
| "step": 1931 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.0001797801047120419, |
| "loss": 0.0844, |
| "step": 1932 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.00017976963350785341, |
| "loss": 0.056, |
| "step": 1933 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.00017975916230366493, |
| "loss": 0.0926, |
| "step": 1934 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.00017974869109947645, |
| "loss": 0.0721, |
| "step": 1935 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.00017973821989528797, |
| "loss": 0.0543, |
| "step": 1936 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.00017972774869109948, |
| "loss": 0.0859, |
| "step": 1937 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 0.000179717277486911, |
| "loss": 0.0905, |
| "step": 1938 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017970680628272252, |
| "loss": 0.0669, |
| "step": 1939 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017969633507853406, |
| "loss": 0.0777, |
| "step": 1940 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017968586387434555, |
| "loss": 0.069, |
| "step": 1941 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017967539267015707, |
| "loss": 0.0921, |
| "step": 1942 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.0001796649214659686, |
| "loss": 0.0653, |
| "step": 1943 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017965445026178013, |
| "loss": 0.0727, |
| "step": 1944 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017964397905759162, |
| "loss": 0.0777, |
| "step": 1945 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017963350785340316, |
| "loss": 0.0793, |
| "step": 1946 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017962303664921468, |
| "loss": 0.0678, |
| "step": 1947 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017961256544502617, |
| "loss": 0.0535, |
| "step": 1948 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017960209424083771, |
| "loss": 0.0586, |
| "step": 1949 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017959162303664923, |
| "loss": 0.0732, |
| "step": 1950 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017958115183246075, |
| "loss": 0.0632, |
| "step": 1951 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017957068062827227, |
| "loss": 0.0605, |
| "step": 1952 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017956020942408378, |
| "loss": 0.0664, |
| "step": 1953 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.0001795497382198953, |
| "loss": 0.0961, |
| "step": 1954 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017953926701570682, |
| "loss": 0.0934, |
| "step": 1955 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017952879581151833, |
| "loss": 0.0706, |
| "step": 1956 |
| }, |
| { |
| "epoch": 1.02, |
| "learning_rate": 0.00017951832460732985, |
| "loss": 0.0943, |
| "step": 1957 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.00017950785340314137, |
| "loss": 0.0626, |
| "step": 1958 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.00017949738219895289, |
| "loss": 0.1044, |
| "step": 1959 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.0001794869109947644, |
| "loss": 0.0495, |
| "step": 1960 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.00017947643979057595, |
| "loss": 0.0954, |
| "step": 1961 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.00017946596858638744, |
| "loss": 0.0777, |
| "step": 1962 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.00017945549738219895, |
| "loss": 0.0715, |
| "step": 1963 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.0001794450261780105, |
| "loss": 0.0685, |
| "step": 1964 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.000179434554973822, |
| "loss": 0.0606, |
| "step": 1965 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.0001794240837696335, |
| "loss": 0.0825, |
| "step": 1966 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.00017941361256544505, |
| "loss": 0.0926, |
| "step": 1967 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.00017940314136125657, |
| "loss": 0.081, |
| "step": 1968 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.00017939267015706806, |
| "loss": 0.1007, |
| "step": 1969 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.0001793821989528796, |
| "loss": 0.0631, |
| "step": 1970 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.00017937172774869112, |
| "loss": 0.0983, |
| "step": 1971 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.0001793612565445026, |
| "loss": 0.1076, |
| "step": 1972 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.00017935078534031415, |
| "loss": 0.0877, |
| "step": 1973 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.00017934031413612567, |
| "loss": 0.0896, |
| "step": 1974 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.00017932984293193719, |
| "loss": 0.0901, |
| "step": 1975 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 0.0001793193717277487, |
| "loss": 0.087, |
| "step": 1976 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.00017930890052356022, |
| "loss": 0.0781, |
| "step": 1977 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.00017929842931937174, |
| "loss": 0.0519, |
| "step": 1978 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.00017928795811518325, |
| "loss": 0.0803, |
| "step": 1979 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.00017927748691099477, |
| "loss": 0.0814, |
| "step": 1980 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.0001792670157068063, |
| "loss": 0.0821, |
| "step": 1981 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.0001792565445026178, |
| "loss": 0.0801, |
| "step": 1982 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.00017924607329842932, |
| "loss": 0.0732, |
| "step": 1983 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.00017923560209424084, |
| "loss": 0.0829, |
| "step": 1984 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.00017922513089005238, |
| "loss": 0.0664, |
| "step": 1985 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.00017921465968586387, |
| "loss": 0.0853, |
| "step": 1986 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.0001792041884816754, |
| "loss": 0.0898, |
| "step": 1987 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.00017919371727748693, |
| "loss": 0.0791, |
| "step": 1988 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.00017918324607329845, |
| "loss": 0.0891, |
| "step": 1989 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.00017917277486910994, |
| "loss": 0.0765, |
| "step": 1990 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.00017916230366492149, |
| "loss": 0.0789, |
| "step": 1991 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.000179151832460733, |
| "loss": 0.0827, |
| "step": 1992 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.0001791413612565445, |
| "loss": 0.0858, |
| "step": 1993 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.00017913089005235604, |
| "loss": 0.0983, |
| "step": 1994 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 0.00017912041884816755, |
| "loss": 0.0869, |
| "step": 1995 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 0.00017910994764397907, |
| "loss": 0.0925, |
| "step": 1996 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 0.0001790994764397906, |
| "loss": 0.0762, |
| "step": 1997 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 0.0001790890052356021, |
| "loss": 0.0609, |
| "step": 1998 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 0.00017907853403141362, |
| "loss": 0.0677, |
| "step": 1999 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 0.00017906806282722514, |
| "loss": 0.0748, |
| "step": 2000 |
| }, |
| { |
| "epoch": 1.05, |
| "eval_loss": 0.08917588740587234, |
| "eval_runtime": 319.5812, |
| "eval_samples_per_second": 0.62, |
| "eval_steps_per_second": 0.078, |
| "step": 2000 |
| } |
| ], |
| "logging_steps": 1, |
| "max_steps": 19100, |
| "num_train_epochs": 10, |
| "save_steps": 1000, |
| "total_flos": 2.1773654843070874e+17, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
