| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 1.0, | |
| "eval_steps": 500, | |
| "global_step": 1418, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 5.953488372093023e-07, | |
| "loss": 0.3762, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.1906976744186047e-06, | |
| "loss": 0.3806, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.7860465116279068e-06, | |
| "loss": 0.3564, | |
| "step": 3 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 2.3813953488372093e-06, | |
| "loss": 0.3799, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 2.9767441860465116e-06, | |
| "loss": 0.3386, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 3.5720930232558135e-06, | |
| "loss": 0.3867, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 4.167441860465117e-06, | |
| "loss": 0.3584, | |
| "step": 7 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.762790697674419e-06, | |
| "loss": 0.4356, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 5.358139534883721e-06, | |
| "loss": 0.3867, | |
| "step": 9 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 5.953488372093023e-06, | |
| "loss": 0.3335, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6.548837209302325e-06, | |
| "loss": 0.3455, | |
| "step": 11 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 7.144186046511627e-06, | |
| "loss": 0.3419, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 7.73953488372093e-06, | |
| "loss": 0.3543, | |
| "step": 13 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.334883720930233e-06, | |
| "loss": 0.356, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.930232558139535e-06, | |
| "loss": 0.3398, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.525581395348837e-06, | |
| "loss": 0.3857, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.012093023255814e-05, | |
| "loss": 0.3408, | |
| "step": 17 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.0716279069767443e-05, | |
| "loss": 0.3078, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.1311627906976745e-05, | |
| "loss": 0.3459, | |
| "step": 19 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.1906976744186047e-05, | |
| "loss": 0.3024, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.2502325581395348e-05, | |
| "loss": 0.2673, | |
| "step": 21 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.309767441860465e-05, | |
| "loss": 0.2167, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.3693023255813952e-05, | |
| "loss": 0.2179, | |
| "step": 23 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.4288372093023254e-05, | |
| "loss": 0.3222, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.4883720930232558e-05, | |
| "loss": 0.9799, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.547906976744186e-05, | |
| "loss": 0.2555, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.607441860465116e-05, | |
| "loss": 0.3328, | |
| "step": 27 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.6669767441860467e-05, | |
| "loss": 0.3135, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.7265116279069765e-05, | |
| "loss": 0.3325, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.786046511627907e-05, | |
| "loss": 0.32, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.845581395348837e-05, | |
| "loss": 0.298, | |
| "step": 31 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.9051162790697675e-05, | |
| "loss": 0.3082, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 1.9646511627906976e-05, | |
| "loss": 0.295, | |
| "step": 33 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.024186046511628e-05, | |
| "loss": 0.2774, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.083720930232558e-05, | |
| "loss": 0.2568, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.1432558139534886e-05, | |
| "loss": 0.2376, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.2027906976744184e-05, | |
| "loss": 0.2348, | |
| "step": 37 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.262325581395349e-05, | |
| "loss": 0.2946, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.3218604651162788e-05, | |
| "loss": 0.3002, | |
| "step": 39 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.3813953488372093e-05, | |
| "loss": 0.3874, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.4409302325581395e-05, | |
| "loss": 0.1964, | |
| "step": 41 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.5004651162790697e-05, | |
| "loss": 0.4125, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.56e-05, | |
| "loss": 0.1628, | |
| "step": 43 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.5599966590180132e-05, | |
| "loss": 0.3108, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.5599866360894934e-05, | |
| "loss": 0.2632, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.559969931266764e-05, | |
| "loss": 0.1934, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.5599465446370275e-05, | |
| "loss": 0.199, | |
| "step": 47 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.5599164763223698e-05, | |
| "loss": 0.1933, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.5598797264797558e-05, | |
| "loss": 0.1568, | |
| "step": 49 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.5598362953010303e-05, | |
| "loss": 0.2147, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.559786183012916e-05, | |
| "loss": 0.1933, | |
| "step": 51 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.5597293898770134e-05, | |
| "loss": 0.1562, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.5596659161897994e-05, | |
| "loss": 0.1513, | |
| "step": 53 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.559595762282624e-05, | |
| "loss": 0.1082, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.5595189285217103e-05, | |
| "loss": 0.1502, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.5594354153081535e-05, | |
| "loss": 0.0769, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.5593452230779155e-05, | |
| "loss": 0.1964, | |
| "step": 57 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.5592483523018257e-05, | |
| "loss": 0.1502, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.5591448034855767e-05, | |
| "loss": 0.113, | |
| "step": 59 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.559034577169723e-05, | |
| "loss": 0.173, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.5589176739296777e-05, | |
| "loss": 0.2128, | |
| "step": 61 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.558794094375708e-05, | |
| "loss": 0.2959, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 2.5586638391529356e-05, | |
| "loss": 0.2, | |
| "step": 63 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.5585269089413288e-05, | |
| "loss": 0.2709, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.5583833044557024e-05, | |
| "loss": 0.1917, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.558233026445713e-05, | |
| "loss": 0.2401, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.558076075695854e-05, | |
| "loss": 0.1751, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.5579124530254534e-05, | |
| "loss": 0.2122, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.5577421592886686e-05, | |
| "loss": 0.2079, | |
| "step": 69 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.5575651953744796e-05, | |
| "loss": 0.2365, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.5573815622066902e-05, | |
| "loss": 0.1791, | |
| "step": 71 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.5571912607439166e-05, | |
| "loss": 0.1788, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.5569942919795875e-05, | |
| "loss": 0.1515, | |
| "step": 73 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.5567906569419345e-05, | |
| "loss": 0.195, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.5565803566939917e-05, | |
| "loss": 0.2371, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.5563633923335854e-05, | |
| "loss": 0.1201, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 2.5561397649933318e-05, | |
| "loss": 0.1287, | |
| "step": 77 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.555909475840629e-05, | |
| "loss": 0.1928, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.555672526077652e-05, | |
| "loss": 0.246, | |
| "step": 79 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.5554289169413454e-05, | |
| "loss": 0.0841, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.5551786497034187e-05, | |
| "loss": 0.1295, | |
| "step": 81 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.5549217256703376e-05, | |
| "loss": 0.1348, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.554658146183319e-05, | |
| "loss": 0.1311, | |
| "step": 83 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.5543879126183227e-05, | |
| "loss": 0.1291, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.5541110263860445e-05, | |
| "loss": 0.1207, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.5538274889319097e-05, | |
| "loss": 0.1859, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.5535373017360644e-05, | |
| "loss": 0.149, | |
| "step": 87 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.5532404663133676e-05, | |
| "loss": 0.1899, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.5529369842133853e-05, | |
| "loss": 0.1263, | |
| "step": 89 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.5526268570203802e-05, | |
| "loss": 0.1105, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.552310086353304e-05, | |
| "loss": 0.1231, | |
| "step": 91 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 2.5519866738657907e-05, | |
| "loss": 0.1514, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.551656621246144e-05, | |
| "loss": 0.0941, | |
| "step": 93 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.5513199302173338e-05, | |
| "loss": 0.1327, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.550976602536982e-05, | |
| "loss": 0.1082, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.5506266399973576e-05, | |
| "loss": 0.0906, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.5502700444253638e-05, | |
| "loss": 0.0972, | |
| "step": 97 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.5499068176825318e-05, | |
| "loss": 0.1775, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.549536961665007e-05, | |
| "loss": 0.1733, | |
| "step": 99 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.5491604783035443e-05, | |
| "loss": 0.1534, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.5487773695634935e-05, | |
| "loss": 0.211, | |
| "step": 101 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.548387637444791e-05, | |
| "loss": 0.2189, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.5479912839819495e-05, | |
| "loss": 0.1027, | |
| "step": 103 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.5475883112440467e-05, | |
| "loss": 0.1178, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.5471787213347145e-05, | |
| "loss": 0.2096, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 2.5467625163921297e-05, | |
| "loss": 0.147, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.5463396985889992e-05, | |
| "loss": 0.1291, | |
| "step": 107 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.5459102701325534e-05, | |
| "loss": 0.1225, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.54547423326453e-05, | |
| "loss": 0.1534, | |
| "step": 109 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.5450315902611658e-05, | |
| "loss": 0.1511, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.5445823434331826e-05, | |
| "loss": 0.1065, | |
| "step": 111 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.5441264951257778e-05, | |
| "loss": 0.1802, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.5436640477186083e-05, | |
| "loss": 0.0521, | |
| "step": 113 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.5431950036257814e-05, | |
| "loss": 0.0781, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.54271936529584e-05, | |
| "loss": 0.106, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.5422371352117522e-05, | |
| "loss": 0.1828, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.5417483158908956e-05, | |
| "loss": 0.1387, | |
| "step": 117 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.541252909885047e-05, | |
| "loss": 0.0849, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.5407509197803655e-05, | |
| "loss": 0.1304, | |
| "step": 119 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 2.5402423481973828e-05, | |
| "loss": 0.199, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.539727197790987e-05, | |
| "loss": 0.1687, | |
| "step": 121 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.5392054712504096e-05, | |
| "loss": 0.1528, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.5386771712992117e-05, | |
| "loss": 0.1409, | |
| "step": 123 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.5381423006952694e-05, | |
| "loss": 0.1412, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.537600862230759e-05, | |
| "loss": 0.1653, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.537052858732143e-05, | |
| "loss": 0.1309, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.5364982930601563e-05, | |
| "loss": 0.1091, | |
| "step": 127 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.5359371681097892e-05, | |
| "loss": 0.1168, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.5353694868102727e-05, | |
| "loss": 0.0669, | |
| "step": 129 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.5347952521250656e-05, | |
| "loss": 0.0654, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.534214467051836e-05, | |
| "loss": 0.1732, | |
| "step": 131 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.5336271346224467e-05, | |
| "loss": 0.0977, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.53303325790294e-05, | |
| "loss": 0.092, | |
| "step": 133 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 2.532432839993522e-05, | |
| "loss": 0.1091, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.5318258840285442e-05, | |
| "loss": 0.2038, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.5312123931764893e-05, | |
| "loss": 0.0583, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.5305923706399546e-05, | |
| "loss": 0.1289, | |
| "step": 137 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.5299658196556333e-05, | |
| "loss": 0.0615, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.5293327434943007e-05, | |
| "loss": 0.076, | |
| "step": 139 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.528693145460793e-05, | |
| "loss": 0.1594, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.528047028893995e-05, | |
| "loss": 0.1576, | |
| "step": 141 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.5273943971668187e-05, | |
| "loss": 0.0858, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.526735253686187e-05, | |
| "loss": 0.1616, | |
| "step": 143 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.526069601893016e-05, | |
| "loss": 0.1049, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.525397445262198e-05, | |
| "loss": 0.127, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.5247187873025812e-05, | |
| "loss": 0.1246, | |
| "step": 146 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.524033631556954e-05, | |
| "loss": 0.1713, | |
| "step": 147 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 2.523341981602023e-05, | |
| "loss": 0.1511, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.5226438410483984e-05, | |
| "loss": 0.1036, | |
| "step": 149 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.5219392135405724e-05, | |
| "loss": 0.1163, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.521228102756901e-05, | |
| "loss": 0.1119, | |
| "step": 151 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.520510512409584e-05, | |
| "loss": 0.0904, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.5197864462446477e-05, | |
| "loss": 0.0951, | |
| "step": 153 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.5190559080419233e-05, | |
| "loss": 0.1156, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.5183189016150282e-05, | |
| "loss": 0.1407, | |
| "step": 155 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.5175754308113447e-05, | |
| "loss": 0.1405, | |
| "step": 156 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.5168254995120028e-05, | |
| "loss": 0.0975, | |
| "step": 157 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.5160691116318566e-05, | |
| "loss": 0.0454, | |
| "step": 158 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.5153062711194654e-05, | |
| "loss": 0.2054, | |
| "step": 159 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.5145369819570747e-05, | |
| "loss": 0.1315, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.513761248160592e-05, | |
| "loss": 0.1083, | |
| "step": 161 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.5129790737795683e-05, | |
| "loss": 0.164, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 2.5121904628971767e-05, | |
| "loss": 0.0742, | |
| "step": 163 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.5113954196301896e-05, | |
| "loss": 0.1385, | |
| "step": 164 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.51059394812896e-05, | |
| "loss": 0.0774, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.5097860525773953e-05, | |
| "loss": 0.1382, | |
| "step": 166 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.5089717371929417e-05, | |
| "loss": 0.0979, | |
| "step": 167 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.5081510062265566e-05, | |
| "loss": 0.0985, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.5073238639626884e-05, | |
| "loss": 0.067, | |
| "step": 169 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.5064903147192557e-05, | |
| "loss": 0.1375, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.505650362847622e-05, | |
| "loss": 0.0705, | |
| "step": 171 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.504804012732575e-05, | |
| "loss": 0.1289, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.5039512687923035e-05, | |
| "loss": 0.1203, | |
| "step": 173 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.5030921354783726e-05, | |
| "loss": 0.1655, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.502226617275703e-05, | |
| "loss": 0.0626, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.5013547187025453e-05, | |
| "loss": 0.0906, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2.500476444310459e-05, | |
| "loss": 0.1442, | |
| "step": 177 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.4995917986842852e-05, | |
| "loss": 0.0884, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.498700786442126e-05, | |
| "loss": 0.0915, | |
| "step": 179 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.4978034122353188e-05, | |
| "loss": 0.1522, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.4968996807484115e-05, | |
| "loss": 0.1449, | |
| "step": 181 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.4959895966991398e-05, | |
| "loss": 0.0909, | |
| "step": 182 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.495073164838402e-05, | |
| "loss": 0.1254, | |
| "step": 183 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.494150389950232e-05, | |
| "loss": 0.1233, | |
| "step": 184 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.4932212768517775e-05, | |
| "loss": 0.1136, | |
| "step": 185 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.4922858303932728e-05, | |
| "loss": 0.0819, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.4913440554580145e-05, | |
| "loss": 0.1486, | |
| "step": 187 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.4903959569623357e-05, | |
| "loss": 0.0851, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.4894415398555803e-05, | |
| "loss": 0.2026, | |
| "step": 189 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.488480809120077e-05, | |
| "loss": 0.1033, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 2.4875137697711128e-05, | |
| "loss": 0.1236, | |
| "step": 191 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.486540426856909e-05, | |
| "loss": 0.0614, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.485560785458592e-05, | |
| "loss": 0.1039, | |
| "step": 193 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.4845748506901676e-05, | |
| "loss": 0.1269, | |
| "step": 194 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.483582627698497e-05, | |
| "loss": 0.0635, | |
| "step": 195 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.482584121663265e-05, | |
| "loss": 0.1355, | |
| "step": 196 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.481579337796959e-05, | |
| "loss": 0.0861, | |
| "step": 197 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.4805682813448344e-05, | |
| "loss": 0.0592, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.4795509575848947e-05, | |
| "loss": 0.1151, | |
| "step": 199 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.4785273718278583e-05, | |
| "loss": 0.1741, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.4774975294171352e-05, | |
| "loss": 0.1577, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.4764614357287945e-05, | |
| "loss": 0.0944, | |
| "step": 202 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.4754190961715398e-05, | |
| "loss": 0.117, | |
| "step": 203 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.4743705161866806e-05, | |
| "loss": 0.0992, | |
| "step": 204 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 2.4733157012481025e-05, | |
| "loss": 0.1416, | |
| "step": 205 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.4722546568622394e-05, | |
| "loss": 0.1639, | |
| "step": 206 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.4711873885680444e-05, | |
| "loss": 0.0868, | |
| "step": 207 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.4701139019369617e-05, | |
| "loss": 0.1122, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.4690342025728966e-05, | |
| "loss": 0.107, | |
| "step": 209 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.4679482961121872e-05, | |
| "loss": 0.0472, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.4668561882235738e-05, | |
| "loss": 0.094, | |
| "step": 211 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.4657578846081698e-05, | |
| "loss": 0.0853, | |
| "step": 212 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.4646533909994325e-05, | |
| "loss": 0.1425, | |
| "step": 213 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.463542713163133e-05, | |
| "loss": 0.0854, | |
| "step": 214 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.4624258568973253e-05, | |
| "loss": 0.1004, | |
| "step": 215 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.4613028280323165e-05, | |
| "loss": 0.1886, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.4601736324306368e-05, | |
| "loss": 0.1498, | |
| "step": 217 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.459038275987008e-05, | |
| "loss": 0.1738, | |
| "step": 218 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 2.457896764628314e-05, | |
| "loss": 0.1373, | |
| "step": 219 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.4567491043135685e-05, | |
| "loss": 0.1726, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.4555953010338845e-05, | |
| "loss": 0.0954, | |
| "step": 221 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.4544353608124434e-05, | |
| "loss": 0.1091, | |
| "step": 222 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.453269289704463e-05, | |
| "loss": 0.1237, | |
| "step": 223 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.4520970937971657e-05, | |
| "loss": 0.082, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.4509187792097484e-05, | |
| "loss": 0.1353, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.4497343520933473e-05, | |
| "loss": 0.1597, | |
| "step": 226 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.4485438186310092e-05, | |
| "loss": 0.0796, | |
| "step": 227 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.4473471850376575e-05, | |
| "loss": 0.0963, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.4461444575600592e-05, | |
| "loss": 0.0535, | |
| "step": 229 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.4449356424767944e-05, | |
| "loss": 0.0847, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.443720746098221e-05, | |
| "loss": 0.1054, | |
| "step": 231 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.442499774766444e-05, | |
| "loss": 0.1562, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 2.441272734855281e-05, | |
| "loss": 0.1007, | |
| "step": 233 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.4400396327702285e-05, | |
| "loss": 0.1864, | |
| "step": 234 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.438800474948431e-05, | |
| "loss": 0.0881, | |
| "step": 235 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.437555267858644e-05, | |
| "loss": 0.0552, | |
| "step": 236 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.4363040180012028e-05, | |
| "loss": 0.0312, | |
| "step": 237 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.4350467319079876e-05, | |
| "loss": 0.1568, | |
| "step": 238 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.4337834161423892e-05, | |
| "loss": 0.0426, | |
| "step": 239 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.432514077299275e-05, | |
| "loss": 0.1396, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.431238722004955e-05, | |
| "loss": 0.1679, | |
| "step": 241 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.4299573569171463e-05, | |
| "loss": 0.1548, | |
| "step": 242 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.4286699887249387e-05, | |
| "loss": 0.1308, | |
| "step": 243 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.4273766241487607e-05, | |
| "loss": 0.0792, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.4260772699403433e-05, | |
| "loss": 0.1189, | |
| "step": 245 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.4247719328826844e-05, | |
| "loss": 0.1032, | |
| "step": 246 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.4234606197900147e-05, | |
| "loss": 0.0697, | |
| "step": 247 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 2.4221433375077622e-05, | |
| "loss": 0.0549, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.4208200929125144e-05, | |
| "loss": 0.0668, | |
| "step": 249 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.4194908929119838e-05, | |
| "loss": 0.1114, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.418155744444973e-05, | |
| "loss": 0.0566, | |
| "step": 251 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.416814654481337e-05, | |
| "loss": 0.0881, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.415467630021946e-05, | |
| "loss": 0.0785, | |
| "step": 253 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.4141146780986513e-05, | |
| "loss": 0.2032, | |
| "step": 254 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.412755805774246e-05, | |
| "loss": 0.0383, | |
| "step": 255 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.411391020142431e-05, | |
| "loss": 0.0932, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.4100203283277743e-05, | |
| "loss": 0.1189, | |
| "step": 257 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.4086437374856777e-05, | |
| "loss": 0.0336, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.4072612548023364e-05, | |
| "loss": 0.1295, | |
| "step": 259 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.4058728874947034e-05, | |
| "loss": 0.1226, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.4044786428104508e-05, | |
| "loss": 0.0911, | |
| "step": 261 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 2.4030785280279326e-05, | |
| "loss": 0.0704, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.4016725504561452e-05, | |
| "loss": 0.1615, | |
| "step": 263 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.400260717434692e-05, | |
| "loss": 0.11, | |
| "step": 264 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.398843036333743e-05, | |
| "loss": 0.0995, | |
| "step": 265 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.397419514553996e-05, | |
| "loss": 0.1188, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.3959901595266404e-05, | |
| "loss": 0.0988, | |
| "step": 267 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.394554978713315e-05, | |
| "loss": 0.1354, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.3931139796060716e-05, | |
| "loss": 0.0843, | |
| "step": 269 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.3916671697273363e-05, | |
| "loss": 0.0929, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.3902145566298675e-05, | |
| "loss": 0.0913, | |
| "step": 271 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.388756147896719e-05, | |
| "loss": 0.043, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.387291951141198e-05, | |
| "loss": 0.0638, | |
| "step": 273 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.385821974006829e-05, | |
| "loss": 0.0699, | |
| "step": 274 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.3843462241673105e-05, | |
| "loss": 0.073, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 2.3828647093264758e-05, | |
| "loss": 0.0396, | |
| "step": 276 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.3813774372182544e-05, | |
| "loss": 0.1452, | |
| "step": 277 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.3798844156066292e-05, | |
| "loss": 0.121, | |
| "step": 278 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.3783856522855973e-05, | |
| "loss": 0.1391, | |
| "step": 279 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.3768811550791294e-05, | |
| "loss": 0.1278, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.3753709318411293e-05, | |
| "loss": 0.0287, | |
| "step": 281 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.373854990455391e-05, | |
| "loss": 0.1055, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.3723333388355605e-05, | |
| "loss": 0.0373, | |
| "step": 283 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.3708059849250915e-05, | |
| "loss": 0.1605, | |
| "step": 284 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.3692729366972054e-05, | |
| "loss": 0.066, | |
| "step": 285 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.36773420215485e-05, | |
| "loss": 0.1212, | |
| "step": 286 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.366189789330658e-05, | |
| "loss": 0.102, | |
| "step": 287 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.3646397062869024e-05, | |
| "loss": 0.0982, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.3630839611154577e-05, | |
| "loss": 0.1633, | |
| "step": 289 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 2.3615225619377568e-05, | |
| "loss": 0.1359, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.3599555169047465e-05, | |
| "loss": 0.1202, | |
| "step": 291 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.3583828341968478e-05, | |
| "loss": 0.139, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.356804522023912e-05, | |
| "loss": 0.0854, | |
| "step": 293 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.3552205886251767e-05, | |
| "loss": 0.0805, | |
| "step": 294 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.353631042269225e-05, | |
| "loss": 0.1007, | |
| "step": 295 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.3520358912539414e-05, | |
| "loss": 0.125, | |
| "step": 296 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.350435143906467e-05, | |
| "loss": 0.1219, | |
| "step": 297 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.3488288085831584e-05, | |
| "loss": 0.0497, | |
| "step": 298 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.347216893669543e-05, | |
| "loss": 0.2013, | |
| "step": 299 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.3455994075802744e-05, | |
| "loss": 0.0622, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.3439763587590907e-05, | |
| "loss": 0.1746, | |
| "step": 301 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.342347755678768e-05, | |
| "loss": 0.1532, | |
| "step": 302 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.3407136068410772e-05, | |
| "loss": 0.1318, | |
| "step": 303 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 2.3390739207767397e-05, | |
| "loss": 0.131, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.337428706045384e-05, | |
| "loss": 0.1095, | |
| "step": 305 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.335777971235498e-05, | |
| "loss": 0.123, | |
| "step": 306 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.334121724964387e-05, | |
| "loss": 0.094, | |
| "step": 307 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.3324599758781277e-05, | |
| "loss": 0.1023, | |
| "step": 308 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.3307927326515224e-05, | |
| "loss": 0.1007, | |
| "step": 309 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.329120003988055e-05, | |
| "loss": 0.0897, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.327441798619845e-05, | |
| "loss": 0.0856, | |
| "step": 311 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.3257581253076017e-05, | |
| "loss": 0.088, | |
| "step": 312 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.3240689928405782e-05, | |
| "loss": 0.0709, | |
| "step": 313 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.3223744100365262e-05, | |
| "loss": 0.0916, | |
| "step": 314 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.3206743857416504e-05, | |
| "loss": 0.0865, | |
| "step": 315 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.3189689288305606e-05, | |
| "loss": 0.0863, | |
| "step": 316 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.317258048206227e-05, | |
| "loss": 0.0785, | |
| "step": 317 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.3155417527999328e-05, | |
| "loss": 0.0718, | |
| "step": 318 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 2.3138200515712278e-05, | |
| "loss": 0.1762, | |
| "step": 319 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.3120929535078824e-05, | |
| "loss": 0.1684, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.3103604676258394e-05, | |
| "loss": 0.0441, | |
| "step": 321 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.3086226029691683e-05, | |
| "loss": 0.1375, | |
| "step": 322 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.306879368610016e-05, | |
| "loss": 0.1505, | |
| "step": 323 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.305130773648562e-05, | |
| "loss": 0.0875, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.303376827212969e-05, | |
| "loss": 0.0631, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.301617538459336e-05, | |
| "loss": 0.1098, | |
| "step": 326 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.2998529165716514e-05, | |
| "loss": 0.0745, | |
| "step": 327 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.2980829707617417e-05, | |
| "loss": 0.0903, | |
| "step": 328 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.2963077102692293e-05, | |
| "loss": 0.0896, | |
| "step": 329 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.294527144361477e-05, | |
| "loss": 0.1022, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.2927412823335462e-05, | |
| "loss": 0.0572, | |
| "step": 331 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.2909501335081437e-05, | |
| "loss": 0.1167, | |
| "step": 332 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 2.2891537072355764e-05, | |
| "loss": 0.0923, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.2873520128937e-05, | |
| "loss": 0.1604, | |
| "step": 334 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.2855450598878716e-05, | |
| "loss": 0.0825, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.283732857650899e-05, | |
| "loss": 0.0824, | |
| "step": 336 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.2819154156429937e-05, | |
| "loss": 0.0903, | |
| "step": 337 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.28009274335172e-05, | |
| "loss": 0.0427, | |
| "step": 338 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.2782648502919445e-05, | |
| "loss": 0.0325, | |
| "step": 339 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.27643174600579e-05, | |
| "loss": 0.1439, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.2745934400625813e-05, | |
| "loss": 0.0655, | |
| "step": 341 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.2727499420587985e-05, | |
| "loss": 0.1137, | |
| "step": 342 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.2709012616180255e-05, | |
| "loss": 0.0999, | |
| "step": 343 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.2690474083908996e-05, | |
| "loss": 0.0777, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.267188392055062e-05, | |
| "loss": 0.0576, | |
| "step": 345 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.265324222315106e-05, | |
| "loss": 0.0715, | |
| "step": 346 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 2.2634549089025286e-05, | |
| "loss": 0.123, | |
| "step": 347 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.261580461575677e-05, | |
| "loss": 0.0919, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.2597008901196993e-05, | |
| "loss": 0.1564, | |
| "step": 349 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.257816204346493e-05, | |
| "loss": 0.1736, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.255926414094653e-05, | |
| "loss": 0.0653, | |
| "step": 351 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.2540315292294227e-05, | |
| "loss": 0.0712, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.25213155964264e-05, | |
| "loss": 0.0829, | |
| "step": 353 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.250226515252686e-05, | |
| "loss": 0.1199, | |
| "step": 354 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.248316406004434e-05, | |
| "loss": 0.0543, | |
| "step": 355 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.2464012418691977e-05, | |
| "loss": 0.1094, | |
| "step": 356 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.2444810328446783e-05, | |
| "loss": 0.0739, | |
| "step": 357 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.2425557889549132e-05, | |
| "loss": 0.0738, | |
| "step": 358 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.2406255202502226e-05, | |
| "loss": 0.1215, | |
| "step": 359 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.238690236807158e-05, | |
| "loss": 0.1059, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 2.2367499487284498e-05, | |
| "loss": 0.0457, | |
| "step": 361 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.2348046661429532e-05, | |
| "loss": 0.0627, | |
| "step": 362 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.232854399205596e-05, | |
| "loss": 0.278, | |
| "step": 363 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.230899158097327e-05, | |
| "loss": 0.0801, | |
| "step": 364 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.2289389530250598e-05, | |
| "loss": 0.1078, | |
| "step": 365 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.2269737942216234e-05, | |
| "loss": 0.0694, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.2250036919457046e-05, | |
| "loss": 0.0766, | |
| "step": 367 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.223028656481798e-05, | |
| "loss": 0.075, | |
| "step": 368 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.2210486981401502e-05, | |
| "loss": 0.0641, | |
| "step": 369 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.219063827256707e-05, | |
| "loss": 0.1061, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.2170740541930582e-05, | |
| "loss": 0.1009, | |
| "step": 371 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.2150793893363856e-05, | |
| "loss": 0.0862, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.2130798430994066e-05, | |
| "loss": 0.0779, | |
| "step": 373 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.211075425920321e-05, | |
| "loss": 0.1166, | |
| "step": 374 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 2.209066148262757e-05, | |
| "loss": 0.0912, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.2070520206157153e-05, | |
| "loss": 0.0493, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.205033053493514e-05, | |
| "loss": 0.0558, | |
| "step": 377 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.2030092574357367e-05, | |
| "loss": 0.1363, | |
| "step": 378 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.200980643007174e-05, | |
| "loss": 0.0405, | |
| "step": 379 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.1989472207977692e-05, | |
| "loss": 0.1043, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.1969090014225656e-05, | |
| "loss": 0.0991, | |
| "step": 381 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.194865995521647e-05, | |
| "loss": 0.06, | |
| "step": 382 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.192818213760086e-05, | |
| "loss": 0.1145, | |
| "step": 383 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.1907656668278855e-05, | |
| "loss": 0.1527, | |
| "step": 384 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.1887083654399233e-05, | |
| "loss": 0.1988, | |
| "step": 385 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.1866463203358984e-05, | |
| "loss": 0.114, | |
| "step": 386 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.184579542280273e-05, | |
| "loss": 0.1654, | |
| "step": 387 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.182508042062215e-05, | |
| "loss": 0.1382, | |
| "step": 388 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 2.1804318304955456e-05, | |
| "loss": 0.0672, | |
| "step": 389 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.1783509184186793e-05, | |
| "loss": 0.0633, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.1762653166945687e-05, | |
| "loss": 0.0685, | |
| "step": 391 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.1741750362106476e-05, | |
| "loss": 0.0784, | |
| "step": 392 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.172080087878776e-05, | |
| "loss": 0.0944, | |
| "step": 393 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.1699804826351777e-05, | |
| "loss": 0.0839, | |
| "step": 394 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.16787623144039e-05, | |
| "loss": 0.1111, | |
| "step": 395 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.1657673452792033e-05, | |
| "loss": 0.0565, | |
| "step": 396 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.1636538351606025e-05, | |
| "loss": 0.078, | |
| "step": 397 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.1615357121177108e-05, | |
| "loss": 0.1572, | |
| "step": 398 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.159412987207733e-05, | |
| "loss": 0.0941, | |
| "step": 399 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.1572856715118978e-05, | |
| "loss": 0.1196, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.1551537761353972e-05, | |
| "loss": 0.1031, | |
| "step": 401 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.1530173122073306e-05, | |
| "loss": 0.1038, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.1508762908806483e-05, | |
| "loss": 0.1106, | |
| "step": 403 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 2.1487307233320903e-05, | |
| "loss": 0.0866, | |
| "step": 404 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.1465806207621287e-05, | |
| "loss": 0.1026, | |
| "step": 405 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.144425994394911e-05, | |
| "loss": 0.1461, | |
| "step": 406 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.1422668554781996e-05, | |
| "loss": 0.0712, | |
| "step": 407 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.1401032152833124e-05, | |
| "loss": 0.1592, | |
| "step": 408 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.1379350851050676e-05, | |
| "loss": 0.0891, | |
| "step": 409 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.135762476261721e-05, | |
| "loss": 0.1323, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.1335854000949084e-05, | |
| "loss": 0.0896, | |
| "step": 411 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.1314038679695866e-05, | |
| "loss": 0.1054, | |
| "step": 412 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.1292178912739735e-05, | |
| "loss": 0.0746, | |
| "step": 413 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.1270274814194887e-05, | |
| "loss": 0.0474, | |
| "step": 414 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.1248326498406955e-05, | |
| "loss": 0.1027, | |
| "step": 415 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.122633407995238e-05, | |
| "loss": 0.0869, | |
| "step": 416 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.120429767363784e-05, | |
| "loss": 0.0507, | |
| "step": 417 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 2.118221739449965e-05, | |
| "loss": 0.1043, | |
| "step": 418 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.1160093357803134e-05, | |
| "loss": 0.1305, | |
| "step": 419 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.1137925679042063e-05, | |
| "loss": 0.1193, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.111571447393802e-05, | |
| "loss": 0.0478, | |
| "step": 421 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.109345985843982e-05, | |
| "loss": 0.0759, | |
| "step": 422 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.1071161948722867e-05, | |
| "loss": 0.1078, | |
| "step": 423 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.1048820861188608e-05, | |
| "loss": 0.0976, | |
| "step": 424 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.102643671246386e-05, | |
| "loss": 0.0387, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.100400961940025e-05, | |
| "loss": 0.0547, | |
| "step": 426 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.0981539699073583e-05, | |
| "loss": 0.1642, | |
| "step": 427 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.0959027068783227e-05, | |
| "loss": 0.0777, | |
| "step": 428 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.0936471846051516e-05, | |
| "loss": 0.0616, | |
| "step": 429 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.091387414862313e-05, | |
| "loss": 0.1307, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.089123409446447e-05, | |
| "loss": 0.0764, | |
| "step": 431 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 2.086855180176306e-05, | |
| "loss": 0.0871, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.0845827388926918e-05, | |
| "loss": 0.0596, | |
| "step": 433 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.0823060974583937e-05, | |
| "loss": 0.0679, | |
| "step": 434 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.0800252677581275e-05, | |
| "loss": 0.1365, | |
| "step": 435 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.077740261698473e-05, | |
| "loss": 0.066, | |
| "step": 436 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.0754510912078115e-05, | |
| "loss": 0.0753, | |
| "step": 437 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.0731577682362638e-05, | |
| "loss": 0.0837, | |
| "step": 438 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.070860304755628e-05, | |
| "loss": 0.1214, | |
| "step": 439 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.068558712759317e-05, | |
| "loss": 0.0862, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.0662530042622945e-05, | |
| "loss": 0.1243, | |
| "step": 441 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.0639431913010155e-05, | |
| "loss": 0.0555, | |
| "step": 442 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.061629285933361e-05, | |
| "loss": 0.0806, | |
| "step": 443 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.0593113002385733e-05, | |
| "loss": 0.1031, | |
| "step": 444 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.056989246317198e-05, | |
| "loss": 0.0286, | |
| "step": 445 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 2.054663136291017e-05, | |
| "loss": 0.0717, | |
| "step": 446 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.052332982302985e-05, | |
| "loss": 0.0495, | |
| "step": 447 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.0499987965171693e-05, | |
| "loss": 0.0435, | |
| "step": 448 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.0476605911186833e-05, | |
| "loss": 0.0712, | |
| "step": 449 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.045318378313624e-05, | |
| "loss": 0.0621, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.0429721703290083e-05, | |
| "loss": 0.1567, | |
| "step": 451 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.0406219794127086e-05, | |
| "loss": 0.0385, | |
| "step": 452 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.038267817833391e-05, | |
| "loss": 0.1161, | |
| "step": 453 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.0359096978804466e-05, | |
| "loss": 0.0811, | |
| "step": 454 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.0335476318639337e-05, | |
| "loss": 0.0908, | |
| "step": 455 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.0311816321145072e-05, | |
| "loss": 0.0763, | |
| "step": 456 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.0288117109833596e-05, | |
| "loss": 0.0373, | |
| "step": 457 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.0264378808421528e-05, | |
| "loss": 0.0831, | |
| "step": 458 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.024060154082955e-05, | |
| "loss": 0.0521, | |
| "step": 459 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 2.021678543118176e-05, | |
| "loss": 0.0409, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.0192930603805024e-05, | |
| "loss": 0.0742, | |
| "step": 461 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.016903718322832e-05, | |
| "loss": 0.0593, | |
| "step": 462 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.014510529418211e-05, | |
| "loss": 0.0784, | |
| "step": 463 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.0121135061597647e-05, | |
| "loss": 0.0858, | |
| "step": 464 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.0097126610606362e-05, | |
| "loss": 0.0813, | |
| "step": 465 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.0073080066539203e-05, | |
| "loss": 0.1355, | |
| "step": 466 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.0048995554925965e-05, | |
| "loss": 0.0933, | |
| "step": 467 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.002487320149465e-05, | |
| "loss": 0.0593, | |
| "step": 468 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 2.000071313217079e-05, | |
| "loss": 0.0472, | |
| "step": 469 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9976515473076827e-05, | |
| "loss": 0.1254, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9952280350531423e-05, | |
| "loss": 0.1048, | |
| "step": 471 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9928007891048806e-05, | |
| "loss": 0.0534, | |
| "step": 472 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9903698221338117e-05, | |
| "loss": 0.0996, | |
| "step": 473 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9879351468302745e-05, | |
| "loss": 0.092, | |
| "step": 474 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9854967759039663e-05, | |
| "loss": 0.0289, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.983054722083877e-05, | |
| "loss": 0.0443, | |
| "step": 476 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9806089981182214e-05, | |
| "loss": 0.1331, | |
| "step": 477 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9781596167743745e-05, | |
| "loss": 0.0574, | |
| "step": 478 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9757065908388035e-05, | |
| "loss": 0.0558, | |
| "step": 479 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9732499331170003e-05, | |
| "loss": 0.0642, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9707896564334184e-05, | |
| "loss": 0.0491, | |
| "step": 481 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9683257736314005e-05, | |
| "loss": 0.1349, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9658582975731156e-05, | |
| "loss": 0.0804, | |
| "step": 483 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9633872411394906e-05, | |
| "loss": 0.0862, | |
| "step": 484 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.960912617230143e-05, | |
| "loss": 0.1643, | |
| "step": 485 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9584344387633124e-05, | |
| "loss": 0.0878, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.955952718675796e-05, | |
| "loss": 0.1199, | |
| "step": 487 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.953467469922878e-05, | |
| "loss": 0.0717, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9509787054782634e-05, | |
| "loss": 0.0662, | |
| "step": 489 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9484864383340107e-05, | |
| "loss": 0.1219, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9459906815004625e-05, | |
| "loss": 0.1284, | |
| "step": 491 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9434914480061795e-05, | |
| "loss": 0.1024, | |
| "step": 492 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9409887508978713e-05, | |
| "loss": 0.0765, | |
| "step": 493 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.938482603240328e-05, | |
| "loss": 0.0727, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.935973018116354e-05, | |
| "loss": 0.1085, | |
| "step": 495 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.933460008626696e-05, | |
| "loss": 0.0742, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.930943587889979e-05, | |
| "loss": 0.0995, | |
| "step": 497 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9284237690426346e-05, | |
| "loss": 0.0762, | |
| "step": 498 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9259005652388337e-05, | |
| "loss": 0.0855, | |
| "step": 499 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9233739896504173e-05, | |
| "loss": 0.0799, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9208440554668287e-05, | |
| "loss": 0.0836, | |
| "step": 501 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9183107758950437e-05, | |
| "loss": 0.0826, | |
| "step": 502 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9157741641595023e-05, | |
| "loss": 0.1164, | |
| "step": 503 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.9132342335020387e-05, | |
| "loss": 0.0695, | |
| "step": 504 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.9106909971818132e-05, | |
| "loss": 0.1127, | |
| "step": 505 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.9081444684752426e-05, | |
| "loss": 0.0659, | |
| "step": 506 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.905594660675931e-05, | |
| "loss": 0.1089, | |
| "step": 507 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.9030415870946e-05, | |
| "loss": 0.0391, | |
| "step": 508 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.90048526105902e-05, | |
| "loss": 0.1057, | |
| "step": 509 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.897925695913939e-05, | |
| "loss": 0.0298, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.8953629050210155e-05, | |
| "loss": 0.0857, | |
| "step": 511 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.8927969017587468e-05, | |
| "loss": 0.0873, | |
| "step": 512 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.8902276995224e-05, | |
| "loss": 0.0697, | |
| "step": 513 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.8876553117239402e-05, | |
| "loss": 0.057, | |
| "step": 514 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.8850797517919635e-05, | |
| "loss": 0.1297, | |
| "step": 515 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.882501033171626e-05, | |
| "loss": 0.044, | |
| "step": 516 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.8799191693245706e-05, | |
| "loss": 0.1268, | |
| "step": 517 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.8773341737288622e-05, | |
| "loss": 0.0677, | |
| "step": 518 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.8747460598789126e-05, | |
| "loss": 0.0507, | |
| "step": 519 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.8721548412854113e-05, | |
| "loss": 0.1031, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.8695605314752566e-05, | |
| "loss": 0.0757, | |
| "step": 521 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.8669631439914838e-05, | |
| "loss": 0.0592, | |
| "step": 522 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.8643626923931934e-05, | |
| "loss": 0.0538, | |
| "step": 523 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.861759190255483e-05, | |
| "loss": 0.1023, | |
| "step": 524 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.8591526511693733e-05, | |
| "loss": 0.0902, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.8565430887417404e-05, | |
| "loss": 0.1166, | |
| "step": 526 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.8539305165952414e-05, | |
| "loss": 0.0726, | |
| "step": 527 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.8513149483682455e-05, | |
| "loss": 0.0849, | |
| "step": 528 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.848696397714764e-05, | |
| "loss": 0.081, | |
| "step": 529 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.8460748783043747e-05, | |
| "loss": 0.1022, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.843450403822155e-05, | |
| "loss": 0.0956, | |
| "step": 531 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.8408229879686078e-05, | |
| "loss": 0.102, | |
| "step": 532 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.8381926444595907e-05, | |
| "loss": 0.0749, | |
| "step": 533 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.835559387026245e-05, | |
| "loss": 0.0554, | |
| "step": 534 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.8329232294149235e-05, | |
| "loss": 0.0984, | |
| "step": 535 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.830284185387118e-05, | |
| "loss": 0.0985, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.827642268719389e-05, | |
| "loss": 0.0465, | |
| "step": 537 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.824997493203293e-05, | |
| "loss": 0.0568, | |
| "step": 538 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.8223498726453098e-05, | |
| "loss": 0.0848, | |
| "step": 539 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.8196994208667718e-05, | |
| "loss": 0.0669, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.8170461517037915e-05, | |
| "loss": 0.122, | |
| "step": 541 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.8143900790071873e-05, | |
| "loss": 0.071, | |
| "step": 542 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.8117312166424145e-05, | |
| "loss": 0.1753, | |
| "step": 543 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.809069578489491e-05, | |
| "loss": 0.0825, | |
| "step": 544 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.8064051784429245e-05, | |
| "loss": 0.1365, | |
| "step": 545 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.803738030411641e-05, | |
| "loss": 0.0669, | |
| "step": 546 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.8010681483189106e-05, | |
| "loss": 0.076, | |
| "step": 547 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.7983955461022787e-05, | |
| "loss": 0.0562, | |
| "step": 548 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.795720237713488e-05, | |
| "loss": 0.0896, | |
| "step": 549 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.7930422371184092e-05, | |
| "loss": 0.1198, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.7903615582969667e-05, | |
| "loss": 0.0749, | |
| "step": 551 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.7876782152430666e-05, | |
| "loss": 0.0706, | |
| "step": 552 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.7849922219645218e-05, | |
| "loss": 0.0691, | |
| "step": 553 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.782303592482982e-05, | |
| "loss": 0.0465, | |
| "step": 554 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.7796123408338575e-05, | |
| "loss": 0.0763, | |
| "step": 555 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.7769184810662463e-05, | |
| "loss": 0.1329, | |
| "step": 556 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.774222027242863e-05, | |
| "loss": 0.1491, | |
| "step": 557 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.7715229934399634e-05, | |
| "loss": 0.0721, | |
| "step": 558 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.7688213937472715e-05, | |
| "loss": 0.1087, | |
| "step": 559 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.7661172422679056e-05, | |
| "loss": 0.0537, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.7634105531183053e-05, | |
| "loss": 0.0942, | |
| "step": 561 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.760701340428158e-05, | |
| "loss": 0.0694, | |
| "step": 562 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.757989618340324e-05, | |
| "loss": 0.1129, | |
| "step": 563 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.7552754010107637e-05, | |
| "loss": 0.091, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.7525587026084635e-05, | |
| "loss": 0.0709, | |
| "step": 565 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.7498395373153614e-05, | |
| "loss": 0.1242, | |
| "step": 566 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.747117919326274e-05, | |
| "loss": 0.0807, | |
| "step": 567 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.7443938628488203e-05, | |
| "loss": 0.0751, | |
| "step": 568 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.7416673821033503e-05, | |
| "loss": 0.0751, | |
| "step": 569 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.7389384913228687e-05, | |
| "loss": 0.049, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.7362072047529608e-05, | |
| "loss": 0.0518, | |
| "step": 571 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.73347353665172e-05, | |
| "loss": 0.0445, | |
| "step": 572 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.7307375012896705e-05, | |
| "loss": 0.1322, | |
| "step": 573 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.7279991129496948e-05, | |
| "loss": 0.0592, | |
| "step": 574 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.725258385926959e-05, | |
| "loss": 0.0838, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.7225153345288377e-05, | |
| "loss": 0.0631, | |
| "step": 576 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.719769973074839e-05, | |
| "loss": 0.1046, | |
| "step": 577 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.7170223158965302e-05, | |
| "loss": 0.0321, | |
| "step": 578 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.714272377337463e-05, | |
| "loss": 0.0675, | |
| "step": 579 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.7115201717530992e-05, | |
| "loss": 0.0802, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.708765713510734e-05, | |
| "loss": 0.1105, | |
| "step": 581 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.706009016989423e-05, | |
| "loss": 0.0671, | |
| "step": 582 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.7032500965799055e-05, | |
| "loss": 0.1109, | |
| "step": 583 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.7004889666845313e-05, | |
| "loss": 0.0901, | |
| "step": 584 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.6977256417171825e-05, | |
| "loss": 0.0942, | |
| "step": 585 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.6949601361032016e-05, | |
| "loss": 0.0828, | |
| "step": 586 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.6921924642793155e-05, | |
| "loss": 0.1247, | |
| "step": 587 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.6894226406935563e-05, | |
| "loss": 0.0651, | |
| "step": 588 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.6866506798051917e-05, | |
| "loss": 0.0507, | |
| "step": 589 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.683876596084646e-05, | |
| "loss": 0.0683, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.6811004040134244e-05, | |
| "loss": 0.0722, | |
| "step": 591 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.6783221180840392e-05, | |
| "loss": 0.047, | |
| "step": 592 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.6755417527999328e-05, | |
| "loss": 0.1056, | |
| "step": 593 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.6727593226754027e-05, | |
| "loss": 0.0722, | |
| "step": 594 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.6699748422355252e-05, | |
| "loss": 0.0639, | |
| "step": 595 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.667188326016081e-05, | |
| "loss": 0.0585, | |
| "step": 596 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.6643997885634756e-05, | |
| "loss": 0.0521, | |
| "step": 597 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.661609244434668e-05, | |
| "loss": 0.066, | |
| "step": 598 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.6588167081970924e-05, | |
| "loss": 0.0807, | |
| "step": 599 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.656022194428582e-05, | |
| "loss": 0.0595, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.653225717717293e-05, | |
| "loss": 0.0664, | |
| "step": 601 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.650427292661629e-05, | |
| "loss": 0.0679, | |
| "step": 602 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.647626933870165e-05, | |
| "loss": 0.0509, | |
| "step": 603 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.6448246559615693e-05, | |
| "loss": 0.1185, | |
| "step": 604 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.6420204735645306e-05, | |
| "loss": 0.0381, | |
| "step": 605 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.6392144013176772e-05, | |
| "loss": 0.0972, | |
| "step": 606 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.6364064538695048e-05, | |
| "loss": 0.1339, | |
| "step": 607 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.633596645878298e-05, | |
| "loss": 0.0664, | |
| "step": 608 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.6307849920120524e-05, | |
| "loss": 0.0805, | |
| "step": 609 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.6279715069484014e-05, | |
| "loss": 0.0457, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.625156205374537e-05, | |
| "loss": 0.1329, | |
| "step": 611 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.622339101987134e-05, | |
| "loss": 0.0646, | |
| "step": 612 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.6195202114922722e-05, | |
| "loss": 0.0718, | |
| "step": 613 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.616699548605363e-05, | |
| "loss": 0.0493, | |
| "step": 614 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.613877128051068e-05, | |
| "loss": 0.0397, | |
| "step": 615 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.611052964563225e-05, | |
| "loss": 0.0359, | |
| "step": 616 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.608227072884771e-05, | |
| "loss": 0.0537, | |
| "step": 617 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.6053994677676634e-05, | |
| "loss": 0.0577, | |
| "step": 618 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.602570163972805e-05, | |
| "loss": 0.1191, | |
| "step": 619 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.599739176269967e-05, | |
| "loss": 0.0854, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.596906519437709e-05, | |
| "loss": 0.1596, | |
| "step": 621 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.594072208263306e-05, | |
| "loss": 0.0295, | |
| "step": 622 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.5912362575426682e-05, | |
| "loss": 0.1256, | |
| "step": 623 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.5883986820802645e-05, | |
| "loss": 0.0198, | |
| "step": 624 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.585559496689046e-05, | |
| "loss": 0.0553, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.5827187161903678e-05, | |
| "loss": 0.0675, | |
| "step": 626 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.579876355413911e-05, | |
| "loss": 0.1082, | |
| "step": 627 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.5770324291976085e-05, | |
| "loss": 0.0471, | |
| "step": 628 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.5741869523875625e-05, | |
| "loss": 0.0634, | |
| "step": 629 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.571339939837972e-05, | |
| "loss": 0.0982, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.5684914064110516e-05, | |
| "loss": 0.0949, | |
| "step": 631 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.5656413669769558e-05, | |
| "loss": 0.0584, | |
| "step": 632 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.5627898364137012e-05, | |
| "loss": 0.0469, | |
| "step": 633 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.5599368296070877e-05, | |
| "loss": 0.0796, | |
| "step": 634 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.5570823614506228e-05, | |
| "loss": 0.0578, | |
| "step": 635 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.5542264468454418e-05, | |
| "loss": 0.046, | |
| "step": 636 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.5513691007002302e-05, | |
| "loss": 0.0515, | |
| "step": 637 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.5485103379311478e-05, | |
| "loss": 0.0469, | |
| "step": 638 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.5456501734617495e-05, | |
| "loss": 0.0949, | |
| "step": 639 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.5427886222229072e-05, | |
| "loss": 0.0588, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.5399256991527315e-05, | |
| "loss": 0.0547, | |
| "step": 641 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.537061419196495e-05, | |
| "loss": 0.0437, | |
| "step": 642 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.534195797306554e-05, | |
| "loss": 0.0429, | |
| "step": 643 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.5313288484422695e-05, | |
| "loss": 0.0972, | |
| "step": 644 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.5284605875699296e-05, | |
| "loss": 0.0818, | |
| "step": 645 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.5255910296626716e-05, | |
| "loss": 0.0052, | |
| "step": 646 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.5227201897004042e-05, | |
| "loss": 0.0823, | |
| "step": 647 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.5198480826697277e-05, | |
| "loss": 0.0885, | |
| "step": 648 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.5169747235638577e-05, | |
| "loss": 0.0501, | |
| "step": 649 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.514100127382546e-05, | |
| "loss": 0.097, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.5112243091320018e-05, | |
| "loss": 0.0961, | |
| "step": 651 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.5083472838248144e-05, | |
| "loss": 0.0217, | |
| "step": 652 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.5054690664798738e-05, | |
| "loss": 0.0624, | |
| "step": 653 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.5025896721222934e-05, | |
| "loss": 0.097, | |
| "step": 654 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.4997091157833302e-05, | |
| "loss": 0.0803, | |
| "step": 655 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.4968274125003068e-05, | |
| "loss": 0.1194, | |
| "step": 656 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.4939445773165348e-05, | |
| "loss": 0.0449, | |
| "step": 657 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.491060625281233e-05, | |
| "loss": 0.1214, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.4881755714494511e-05, | |
| "loss": 0.0506, | |
| "step": 659 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.4852894308819907e-05, | |
| "loss": 0.048, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.4824022186453259e-05, | |
| "loss": 0.0789, | |
| "step": 661 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.4795139498115258e-05, | |
| "loss": 0.105, | |
| "step": 662 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.4766246394581752e-05, | |
| "loss": 0.067, | |
| "step": 663 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.4737343026682959e-05, | |
| "loss": 0.0499, | |
| "step": 664 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.470842954530267e-05, | |
| "loss": 0.0479, | |
| "step": 665 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.4679506101377484e-05, | |
| "loss": 0.079, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.4650572845896004e-05, | |
| "loss": 0.0554, | |
| "step": 667 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.462162992989805e-05, | |
| "loss": 0.0586, | |
| "step": 668 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.4592677504473874e-05, | |
| "loss": 0.0699, | |
| "step": 669 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.4563715720763366e-05, | |
| "loss": 0.0772, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.4534744729955281e-05, | |
| "loss": 0.0369, | |
| "step": 671 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.4505764683286417e-05, | |
| "loss": 0.043, | |
| "step": 672 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.4476775732040869e-05, | |
| "loss": 0.1027, | |
| "step": 673 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.4447778027549197e-05, | |
| "loss": 0.0745, | |
| "step": 674 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.4418771721187672e-05, | |
| "loss": 0.0493, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.4389756964377457e-05, | |
| "loss": 0.0668, | |
| "step": 676 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.4360733908583833e-05, | |
| "loss": 0.0373, | |
| "step": 677 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.4331702705315402e-05, | |
| "loss": 0.0634, | |
| "step": 678 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.4302663506123306e-05, | |
| "loss": 0.0383, | |
| "step": 679 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.4273616462600417e-05, | |
| "loss": 0.0495, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.4244561726380564e-05, | |
| "loss": 0.0612, | |
| "step": 681 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.4215499449137735e-05, | |
| "loss": 0.0889, | |
| "step": 682 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.4186429782585276e-05, | |
| "loss": 0.1333, | |
| "step": 683 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.4157352878475115e-05, | |
| "loss": 0.1282, | |
| "step": 684 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.4128268888596956e-05, | |
| "loss": 0.0262, | |
| "step": 685 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.4099177964777504e-05, | |
| "loss": 0.0177, | |
| "step": 686 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.4070080258879645e-05, | |
| "loss": 0.1859, | |
| "step": 687 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.4040975922801682e-05, | |
| "loss": 0.0993, | |
| "step": 688 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.4011865108476528e-05, | |
| "loss": 0.0767, | |
| "step": 689 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.3982747967870906e-05, | |
| "loss": 0.0937, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.3953624652984572e-05, | |
| "loss": 0.0891, | |
| "step": 691 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.3924495315849514e-05, | |
| "loss": 0.0644, | |
| "step": 692 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.3895360108529149e-05, | |
| "loss": 0.0902, | |
| "step": 693 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.3866219183117547e-05, | |
| "loss": 0.1261, | |
| "step": 694 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.3837072691738626e-05, | |
| "loss": 0.0666, | |
| "step": 695 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.380792078654536e-05, | |
| "loss": 0.1429, | |
| "step": 696 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.377876361971898e-05, | |
| "loss": 0.0529, | |
| "step": 697 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.3749601343468188e-05, | |
| "loss": 0.0535, | |
| "step": 698 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.3720434110028361e-05, | |
| "loss": 0.1006, | |
| "step": 699 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.3691262071660747e-05, | |
| "loss": 0.0325, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.3662085380651688e-05, | |
| "loss": 0.0837, | |
| "step": 701 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.3632904189311803e-05, | |
| "loss": 0.0609, | |
| "step": 702 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.3603718649975212e-05, | |
| "loss": 0.0564, | |
| "step": 703 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.3574528914998725e-05, | |
| "loss": 0.1126, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.3545335136761062e-05, | |
| "loss": 0.0819, | |
| "step": 705 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.3516137467662047e-05, | |
| "loss": 0.1139, | |
| "step": 706 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.3486936060121809e-05, | |
| "loss": 0.0773, | |
| "step": 707 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.3457731066580006e-05, | |
| "loss": 0.0584, | |
| "step": 708 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.3428522639495009e-05, | |
| "loss": 0.0456, | |
| "step": 709 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.3399310931343112e-05, | |
| "loss": 0.0698, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.3370096094617732e-05, | |
| "loss": 0.1087, | |
| "step": 711 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.334087828182863e-05, | |
| "loss": 0.0646, | |
| "step": 712 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.3311657645501099e-05, | |
| "loss": 0.0372, | |
| "step": 713 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.3282434338175165e-05, | |
| "loss": 0.0776, | |
| "step": 714 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.32532085124048e-05, | |
| "loss": 0.0885, | |
| "step": 715 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.322398032075713e-05, | |
| "loss": 0.132, | |
| "step": 716 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.3194749915811626e-05, | |
| "loss": 0.1065, | |
| "step": 717 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.316551745015931e-05, | |
| "loss": 0.0433, | |
| "step": 718 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.3136283076401972e-05, | |
| "loss": 0.0445, | |
| "step": 719 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.3107046947151351e-05, | |
| "loss": 0.067, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.3077809215028352e-05, | |
| "loss": 0.0486, | |
| "step": 721 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.3048570032662257e-05, | |
| "loss": 0.0591, | |
| "step": 722 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.301932955268991e-05, | |
| "loss": 0.0937, | |
| "step": 723 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.299008792775493e-05, | |
| "loss": 0.045, | |
| "step": 724 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.2960845310506914e-05, | |
| "loss": 0.072, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.2931601853600642e-05, | |
| "loss": 0.0386, | |
| "step": 726 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.2902357709695268e-05, | |
| "loss": 0.0631, | |
| "step": 727 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.2873113031453546e-05, | |
| "loss": 0.0685, | |
| "step": 728 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.2843867971541013e-05, | |
| "loss": 0.0348, | |
| "step": 729 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.281462268262519e-05, | |
| "loss": 0.0622, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.278537731737481e-05, | |
| "loss": 0.0449, | |
| "step": 731 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.2756132028458986e-05, | |
| "loss": 0.1279, | |
| "step": 732 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.2726886968546453e-05, | |
| "loss": 0.0348, | |
| "step": 733 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.2697642290304734e-05, | |
| "loss": 0.1615, | |
| "step": 734 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.266839814639936e-05, | |
| "loss": 0.0664, | |
| "step": 735 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.2639154689493084e-05, | |
| "loss": 0.0578, | |
| "step": 736 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.2609912072245074e-05, | |
| "loss": 0.0382, | |
| "step": 737 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.258067044731009e-05, | |
| "loss": 0.1043, | |
| "step": 738 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.2551429967337742e-05, | |
| "loss": 0.0533, | |
| "step": 739 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.2522190784971649e-05, | |
| "loss": 0.0557, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.249295305284865e-05, | |
| "loss": 0.0275, | |
| "step": 741 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.2463716923598028e-05, | |
| "loss": 0.093, | |
| "step": 742 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.2434482549840689e-05, | |
| "loss": 0.0403, | |
| "step": 743 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.2405250084188374e-05, | |
| "loss": 0.0319, | |
| "step": 744 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.237601967924287e-05, | |
| "loss": 0.0599, | |
| "step": 745 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.23467914875952e-05, | |
| "loss": 0.036, | |
| "step": 746 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.2317565661824836e-05, | |
| "loss": 0.0397, | |
| "step": 747 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.2288342354498903e-05, | |
| "loss": 0.1234, | |
| "step": 748 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.2259121718171367e-05, | |
| "loss": 0.0308, | |
| "step": 749 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.222990390538227e-05, | |
| "loss": 0.1097, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.2200689068656892e-05, | |
| "loss": 0.0932, | |
| "step": 751 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.2171477360504989e-05, | |
| "loss": 0.0472, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.2142268933419994e-05, | |
| "loss": 0.0417, | |
| "step": 753 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.2113063939878192e-05, | |
| "loss": 0.0209, | |
| "step": 754 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.2083862532337955e-05, | |
| "loss": 0.0767, | |
| "step": 755 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.2054664863238939e-05, | |
| "loss": 0.056, | |
| "step": 756 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.2025471085001279e-05, | |
| "loss": 0.0615, | |
| "step": 757 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.1996281350024789e-05, | |
| "loss": 0.0139, | |
| "step": 758 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.1967095810688197e-05, | |
| "loss": 0.0362, | |
| "step": 759 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.193791461934831e-05, | |
| "loss": 0.0723, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.1908737928339251e-05, | |
| "loss": 0.0111, | |
| "step": 761 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.1879565889971644e-05, | |
| "loss": 0.0886, | |
| "step": 762 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.1850398656531812e-05, | |
| "loss": 0.0598, | |
| "step": 763 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.1821236380281022e-05, | |
| "loss": 0.0615, | |
| "step": 764 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.1792079213454643e-05, | |
| "loss": 0.0243, | |
| "step": 765 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.1762927308261372e-05, | |
| "loss": 0.0728, | |
| "step": 766 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.1733780816882452e-05, | |
| "loss": 0.0309, | |
| "step": 767 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.1704639891470853e-05, | |
| "loss": 0.0612, | |
| "step": 768 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.1675504684150488e-05, | |
| "loss": 0.0485, | |
| "step": 769 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.164637534701543e-05, | |
| "loss": 0.0994, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.1617252032129095e-05, | |
| "loss": 0.0993, | |
| "step": 771 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.1588134891523473e-05, | |
| "loss": 0.0262, | |
| "step": 772 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.1559024077198316e-05, | |
| "loss": 0.0812, | |
| "step": 773 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.1529919741120354e-05, | |
| "loss": 0.0769, | |
| "step": 774 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.1500822035222498e-05, | |
| "loss": 0.0286, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.1471731111403044e-05, | |
| "loss": 0.0378, | |
| "step": 776 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.1442647121524884e-05, | |
| "loss": 0.0746, | |
| "step": 777 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.1413570217414724e-05, | |
| "loss": 0.0352, | |
| "step": 778 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.1384500550862269e-05, | |
| "loss": 0.0538, | |
| "step": 779 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.1355438273619433e-05, | |
| "loss": 0.0369, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.1326383537399585e-05, | |
| "loss": 0.0651, | |
| "step": 781 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.1297336493876696e-05, | |
| "loss": 0.0509, | |
| "step": 782 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.1268297294684599e-05, | |
| "loss": 0.0426, | |
| "step": 783 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.123926609141617e-05, | |
| "loss": 0.0367, | |
| "step": 784 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.121024303562254e-05, | |
| "loss": 0.0894, | |
| "step": 785 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.1181228278812328e-05, | |
| "loss": 0.0619, | |
| "step": 786 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.1152221972450802e-05, | |
| "loss": 0.099, | |
| "step": 787 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.112322426795913e-05, | |
| "loss": 0.0498, | |
| "step": 788 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.1094235316713583e-05, | |
| "loss": 0.0465, | |
| "step": 789 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.1065255270044725e-05, | |
| "loss": 0.0611, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.1036284279236633e-05, | |
| "loss": 0.0659, | |
| "step": 791 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.1007322495526126e-05, | |
| "loss": 0.0583, | |
| "step": 792 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.097837007010195e-05, | |
| "loss": 0.0853, | |
| "step": 793 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.0949427154103995e-05, | |
| "loss": 0.0445, | |
| "step": 794 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.0920493898622516e-05, | |
| "loss": 0.0465, | |
| "step": 795 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.0891570454697334e-05, | |
| "loss": 0.0512, | |
| "step": 796 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.0862656973317042e-05, | |
| "loss": 0.1106, | |
| "step": 797 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.0833753605418248e-05, | |
| "loss": 0.0345, | |
| "step": 798 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.0804860501884737e-05, | |
| "loss": 0.0675, | |
| "step": 799 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.0775977813546742e-05, | |
| "loss": 0.0447, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.0747105691180096e-05, | |
| "loss": 0.0941, | |
| "step": 801 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.071824428550549e-05, | |
| "loss": 0.0434, | |
| "step": 802 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.068939374718767e-05, | |
| "loss": 0.0241, | |
| "step": 803 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.0660554226834654e-05, | |
| "loss": 0.0545, | |
| "step": 804 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.0631725874996929e-05, | |
| "loss": 0.0633, | |
| "step": 805 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.06029088421667e-05, | |
| "loss": 0.0137, | |
| "step": 806 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.0574103278777067e-05, | |
| "loss": 0.12, | |
| "step": 807 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.0545309335201258e-05, | |
| "loss": 0.0692, | |
| "step": 808 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.0516527161751855e-05, | |
| "loss": 0.1564, | |
| "step": 809 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.0487756908679983e-05, | |
| "loss": 0.0749, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.0458998726174541e-05, | |
| "loss": 0.0931, | |
| "step": 811 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.0430252764361423e-05, | |
| "loss": 0.0594, | |
| "step": 812 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.0401519173302721e-05, | |
| "loss": 0.0775, | |
| "step": 813 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.0372798102995958e-05, | |
| "loss": 0.0691, | |
| "step": 814 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.0344089703373284e-05, | |
| "loss": 0.0358, | |
| "step": 815 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.0315394124300702e-05, | |
| "loss": 0.0532, | |
| "step": 816 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.0286711515577304e-05, | |
| "loss": 0.0279, | |
| "step": 817 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.025804202693446e-05, | |
| "loss": 0.0492, | |
| "step": 818 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.0229385808035047e-05, | |
| "loss": 0.0707, | |
| "step": 819 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.0200743008472684e-05, | |
| "loss": 0.064, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.0172113777770932e-05, | |
| "loss": 0.0308, | |
| "step": 821 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.0143498265382503e-05, | |
| "loss": 0.0182, | |
| "step": 822 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.0114896620688523e-05, | |
| "loss": 0.0821, | |
| "step": 823 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.0086308992997704e-05, | |
| "loss": 0.0376, | |
| "step": 824 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.0057735531545586e-05, | |
| "loss": 0.0319, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.0029176385493773e-05, | |
| "loss": 0.062, | |
| "step": 826 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.0000631703929118e-05, | |
| "loss": 0.0169, | |
| "step": 827 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 9.972101635862988e-06, | |
| "loss": 0.0241, | |
| "step": 828 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 9.943586330230442e-06, | |
| "loss": 0.0972, | |
| "step": 829 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 9.915085935889485e-06, | |
| "loss": 0.0691, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 9.886600601620282e-06, | |
| "loss": 0.0686, | |
| "step": 831 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 9.858130476124377e-06, | |
| "loss": 0.0444, | |
| "step": 832 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 9.829675708023916e-06, | |
| "loss": 0.0628, | |
| "step": 833 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 9.80123644586089e-06, | |
| "loss": 0.036, | |
| "step": 834 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 9.772812838096326e-06, | |
| "loss": 0.0412, | |
| "step": 835 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 9.744405033109538e-06, | |
| "loss": 0.0544, | |
| "step": 836 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 9.716013179197356e-06, | |
| "loss": 0.0901, | |
| "step": 837 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 9.687637424573316e-06, | |
| "loss": 0.0484, | |
| "step": 838 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 9.659277917366937e-06, | |
| "loss": 0.0527, | |
| "step": 839 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 9.630934805622908e-06, | |
| "loss": 0.1124, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 9.60260823730033e-06, | |
| "loss": 0.0478, | |
| "step": 841 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 9.574298360271947e-06, | |
| "loss": 0.0593, | |
| "step": 842 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 9.54600532232337e-06, | |
| "loss": 0.0511, | |
| "step": 843 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 9.51772927115229e-06, | |
| "loss": 0.0526, | |
| "step": 844 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 9.489470354367749e-06, | |
| "loss": 0.0429, | |
| "step": 845 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 9.461228719489321e-06, | |
| "loss": 0.0274, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 9.433004513946369e-06, | |
| "loss": 0.0261, | |
| "step": 847 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 9.404797885077278e-06, | |
| "loss": 0.0704, | |
| "step": 848 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 9.376608980128665e-06, | |
| "loss": 0.0343, | |
| "step": 849 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 9.348437946254629e-06, | |
| "loss": 0.0319, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 9.320284930515987e-06, | |
| "loss": 0.0395, | |
| "step": 851 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 9.29215007987948e-06, | |
| "loss": 0.0894, | |
| "step": 852 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 9.26403354121702e-06, | |
| "loss": 0.0325, | |
| "step": 853 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 9.235935461304951e-06, | |
| "loss": 0.0531, | |
| "step": 854 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 9.207855986823225e-06, | |
| "loss": 0.0914, | |
| "step": 855 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 9.179795264354698e-06, | |
| "loss": 0.0078, | |
| "step": 856 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 9.151753440384305e-06, | |
| "loss": 0.0632, | |
| "step": 857 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 9.123730661298352e-06, | |
| "loss": 0.053, | |
| "step": 858 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 9.09572707338371e-06, | |
| "loss": 0.0686, | |
| "step": 859 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 9.067742822827071e-06, | |
| "loss": 0.0353, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 9.039778055714177e-06, | |
| "loss": 0.1007, | |
| "step": 861 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 9.011832918029074e-06, | |
| "loss": 0.0647, | |
| "step": 862 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 8.983907555653322e-06, | |
| "loss": 0.0526, | |
| "step": 863 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 8.956002114365244e-06, | |
| "loss": 0.0741, | |
| "step": 864 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 8.928116739839194e-06, | |
| "loss": 0.093, | |
| "step": 865 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 8.900251577644746e-06, | |
| "loss": 0.0231, | |
| "step": 866 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 8.872406773245974e-06, | |
| "loss": 0.0317, | |
| "step": 867 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 8.844582472000674e-06, | |
| "loss": 0.0477, | |
| "step": 868 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 8.816778819159607e-06, | |
| "loss": 0.0649, | |
| "step": 869 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 8.788995959865757e-06, | |
| "loss": 0.0817, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 8.761234039153542e-06, | |
| "loss": 0.0341, | |
| "step": 871 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 8.733493201948079e-06, | |
| "loss": 0.0691, | |
| "step": 872 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 8.705773593064436e-06, | |
| "loss": 0.0247, | |
| "step": 873 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 8.67807535720685e-06, | |
| "loss": 0.0588, | |
| "step": 874 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 8.65039863896798e-06, | |
| "loss": 0.0748, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 8.622743582828178e-06, | |
| "loss": 0.0294, | |
| "step": 876 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 8.595110333154691e-06, | |
| "loss": 0.102, | |
| "step": 877 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 8.567499034200944e-06, | |
| "loss": 0.0419, | |
| "step": 878 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 8.539909830105772e-06, | |
| "loss": 0.0725, | |
| "step": 879 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 8.512342864892662e-06, | |
| "loss": 0.0224, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 8.484798282469008e-06, | |
| "loss": 0.062, | |
| "step": 881 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 8.45727622662537e-06, | |
| "loss": 0.0275, | |
| "step": 882 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 8.429776841034696e-06, | |
| "loss": 0.0496, | |
| "step": 883 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 8.402300269251612e-06, | |
| "loss": 0.0342, | |
| "step": 884 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 8.374846654711624e-06, | |
| "loss": 0.0452, | |
| "step": 885 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 8.347416140730407e-06, | |
| "loss": 0.0333, | |
| "step": 886 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 8.320008870503051e-06, | |
| "loss": 0.0428, | |
| "step": 887 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 8.2926249871033e-06, | |
| "loss": 0.0363, | |
| "step": 888 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 8.2652646334828e-06, | |
| "loss": 0.1065, | |
| "step": 889 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 8.237927952470392e-06, | |
| "loss": 0.0499, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 8.210615086771317e-06, | |
| "loss": 0.0285, | |
| "step": 891 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 8.183326178966496e-06, | |
| "loss": 0.0579, | |
| "step": 892 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 8.156061371511796e-06, | |
| "loss": 0.0288, | |
| "step": 893 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 8.128820806737258e-06, | |
| "loss": 0.1202, | |
| "step": 894 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 8.101604626846383e-06, | |
| "loss": 0.0844, | |
| "step": 895 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 8.074412973915365e-06, | |
| "loss": 0.0406, | |
| "step": 896 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 8.047245989892363e-06, | |
| "loss": 0.0937, | |
| "step": 897 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 8.020103816596761e-06, | |
| "loss": 0.0535, | |
| "step": 898 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 7.992986595718424e-06, | |
| "loss": 0.0346, | |
| "step": 899 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 7.965894468816947e-06, | |
| "loss": 0.0497, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 7.938827577320946e-06, | |
| "loss": 0.0599, | |
| "step": 901 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 7.911786062527289e-06, | |
| "loss": 0.0383, | |
| "step": 902 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 7.884770065600366e-06, | |
| "loss": 0.0203, | |
| "step": 903 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 7.857779727571371e-06, | |
| "loss": 0.0175, | |
| "step": 904 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 7.83081518933754e-06, | |
| "loss": 0.0624, | |
| "step": 905 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 7.803876591661427e-06, | |
| "loss": 0.0426, | |
| "step": 906 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 7.77696407517018e-06, | |
| "loss": 0.0279, | |
| "step": 907 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 7.750077780354777e-06, | |
| "loss": 0.0422, | |
| "step": 908 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 7.723217847569335e-06, | |
| "loss": 0.0909, | |
| "step": 909 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 7.696384417030333e-06, | |
| "loss": 0.0303, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 7.669577628815907e-06, | |
| "loss": 0.0317, | |
| "step": 911 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 7.642797622865118e-06, | |
| "loss": 0.0677, | |
| "step": 912 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 7.616044538977212e-06, | |
| "loss": 0.067, | |
| "step": 913 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 7.589318516810891e-06, | |
| "loss": 0.0515, | |
| "step": 914 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 7.5626196958835945e-06, | |
| "loss": 0.0708, | |
| "step": 915 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 7.535948215570758e-06, | |
| "loss": 0.0403, | |
| "step": 916 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 7.509304215105089e-06, | |
| "loss": 0.043, | |
| "step": 917 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 7.482687833575855e-06, | |
| "loss": 0.0554, | |
| "step": 918 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 7.4560992099281275e-06, | |
| "loss": 0.0615, | |
| "step": 919 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 7.429538482962084e-06, | |
| "loss": 0.1071, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 7.403005791332279e-06, | |
| "loss": 0.0421, | |
| "step": 921 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 7.376501273546898e-06, | |
| "loss": 0.0349, | |
| "step": 922 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 7.35002506796707e-06, | |
| "loss": 0.0389, | |
| "step": 923 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 7.32357731280611e-06, | |
| "loss": 0.1106, | |
| "step": 924 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 7.297158146128817e-06, | |
| "loss": 0.0269, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 7.270767705850765e-06, | |
| "loss": 0.0527, | |
| "step": 926 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 7.24440612973755e-06, | |
| "loss": 0.0861, | |
| "step": 927 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 7.21807355540409e-06, | |
| "loss": 0.0698, | |
| "step": 928 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 7.191770120313921e-06, | |
| "loss": 0.0321, | |
| "step": 929 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 7.16549596177845e-06, | |
| "loss": 0.0647, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 7.139251216956251e-06, | |
| "loss": 0.0687, | |
| "step": 931 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 7.113036022852361e-06, | |
| "loss": 0.0354, | |
| "step": 932 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 7.086850516317545e-06, | |
| "loss": 0.0494, | |
| "step": 933 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 7.060694834047589e-06, | |
| "loss": 0.0387, | |
| "step": 934 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 7.034569112582599e-06, | |
| "loss": 0.0342, | |
| "step": 935 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 7.0084734883062645e-06, | |
| "loss": 0.042, | |
| "step": 936 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 6.982408097445173e-06, | |
| "loss": 0.0553, | |
| "step": 937 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 6.9563730760680635e-06, | |
| "loss": 0.0584, | |
| "step": 938 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 6.930368560085165e-06, | |
| "loss": 0.0378, | |
| "step": 939 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 6.904394685247431e-06, | |
| "loss": 0.0529, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 6.878451587145889e-06, | |
| "loss": 0.094, | |
| "step": 941 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 6.852539401210873e-06, | |
| "loss": 0.0299, | |
| "step": 942 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 6.826658262711376e-06, | |
| "loss": 0.0546, | |
| "step": 943 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 6.8008083067542935e-06, | |
| "loss": 0.0317, | |
| "step": 944 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 6.774989668283743e-06, | |
| "loss": 0.0873, | |
| "step": 945 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 6.749202482080365e-06, | |
| "loss": 0.0442, | |
| "step": 946 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 6.723446882760604e-06, | |
| "loss": 0.0228, | |
| "step": 947 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 6.697723004776001e-06, | |
| "loss": 0.0317, | |
| "step": 948 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 6.67203098241253e-06, | |
| "loss": 0.026, | |
| "step": 949 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 6.64637094978984e-06, | |
| "loss": 0.0415, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 6.620743040860611e-06, | |
| "loss": 0.1225, | |
| "step": 951 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 6.595147389409805e-06, | |
| "loss": 0.0888, | |
| "step": 952 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 6.569584129054e-06, | |
| "loss": 0.0312, | |
| "step": 953 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 6.544053393240692e-06, | |
| "loss": 0.0288, | |
| "step": 954 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 6.518555315247579e-06, | |
| "loss": 0.0758, | |
| "step": 955 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 6.493090028181869e-06, | |
| "loss": 0.0498, | |
| "step": 956 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 6.467657664979612e-06, | |
| "loss": 0.0196, | |
| "step": 957 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 6.4422583584049775e-06, | |
| "loss": 0.0182, | |
| "step": 958 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 6.416892241049558e-06, | |
| "loss": 0.0574, | |
| "step": 959 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 6.391559445331713e-06, | |
| "loss": 0.0484, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 6.366260103495829e-06, | |
| "loss": 0.0717, | |
| "step": 961 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 6.3409943476116635e-06, | |
| "loss": 0.1165, | |
| "step": 962 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 6.315762309573655e-06, | |
| "loss": 0.0595, | |
| "step": 963 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 6.290564121100206e-06, | |
| "loss": 0.0627, | |
| "step": 964 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 6.265399913733039e-06, | |
| "loss": 0.013, | |
| "step": 965 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 6.2402698188364634e-06, | |
| "loss": 0.0295, | |
| "step": 966 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 6.215173967596716e-06, | |
| "loss": 0.0821, | |
| "step": 967 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 6.190112491021288e-06, | |
| "loss": 0.1053, | |
| "step": 968 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 6.165085519938208e-06, | |
| "loss": 0.0566, | |
| "step": 969 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 6.140093184995376e-06, | |
| "loss": 0.1067, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 6.115135616659897e-06, | |
| "loss": 0.0594, | |
| "step": 971 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 6.090212945217369e-06, | |
| "loss": 0.056, | |
| "step": 972 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 6.065325300771221e-06, | |
| "loss": 0.0467, | |
| "step": 973 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 6.040472813242043e-06, | |
| "loss": 0.0769, | |
| "step": 974 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 6.015655612366875e-06, | |
| "loss": 0.0271, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 5.990873827698568e-06, | |
| "loss": 0.0858, | |
| "step": 976 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 5.9661275886050924e-06, | |
| "loss": 0.0856, | |
| "step": 977 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 5.941417024268839e-06, | |
| "loss": 0.1011, | |
| "step": 978 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 5.916742263685995e-06, | |
| "loss": 0.06, | |
| "step": 979 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 5.892103435665818e-06, | |
| "loss": 0.0498, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 5.867500668829993e-06, | |
| "loss": 0.0434, | |
| "step": 981 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 5.842934091611967e-06, | |
| "loss": 0.0455, | |
| "step": 982 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 5.818403832256257e-06, | |
| "loss": 0.0798, | |
| "step": 983 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 5.793910018817784e-06, | |
| "loss": 0.0488, | |
| "step": 984 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 5.769452779161231e-06, | |
| "loss": 0.0391, | |
| "step": 985 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 5.745032240960338e-06, | |
| "loss": 0.0459, | |
| "step": 986 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 5.7206485316972545e-06, | |
| "loss": 0.021, | |
| "step": 987 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 5.6963017786618846e-06, | |
| "loss": 0.1121, | |
| "step": 988 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 5.671992108951198e-06, | |
| "loss": 0.0656, | |
| "step": 989 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 5.647719649468578e-06, | |
| "loss": 0.035, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 5.623484526923175e-06, | |
| "loss": 0.0436, | |
| "step": 991 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 5.599286867829211e-06, | |
| "loss": 0.025, | |
| "step": 992 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 5.575126798505357e-06, | |
| "loss": 0.0359, | |
| "step": 993 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 5.551004445074034e-06, | |
| "loss": 0.0837, | |
| "step": 994 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 5.526919933460796e-06, | |
| "loss": 0.0313, | |
| "step": 995 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 5.502873389393634e-06, | |
| "loss": 0.0488, | |
| "step": 996 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 5.478864938402354e-06, | |
| "loss": 0.0844, | |
| "step": 997 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 5.454894705817888e-06, | |
| "loss": 0.0769, | |
| "step": 998 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 5.430962816771675e-06, | |
| "loss": 0.0362, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 5.407069396194977e-06, | |
| "loss": 0.0431, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 5.383214568818238e-06, | |
| "loss": 0.0304, | |
| "step": 1001 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 5.359398459170451e-06, | |
| "loss": 0.038, | |
| "step": 1002 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 5.335621191578471e-06, | |
| "loss": 0.0338, | |
| "step": 1003 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 5.3118828901664025e-06, | |
| "loss": 0.035, | |
| "step": 1004 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 5.288183678854929e-06, | |
| "loss": 0.0148, | |
| "step": 1005 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 5.264523681360664e-06, | |
| "loss": 0.0472, | |
| "step": 1006 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 5.240903021195534e-06, | |
| "loss": 0.0145, | |
| "step": 1007 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 5.217321821666095e-06, | |
| "loss": 0.0241, | |
| "step": 1008 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 5.19378020587291e-06, | |
| "loss": 0.0615, | |
| "step": 1009 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 5.170278296709918e-06, | |
| "loss": 0.0385, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 5.1468162168637635e-06, | |
| "loss": 0.0406, | |
| "step": 1011 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 5.1233940888131665e-06, | |
| "loss": 0.0476, | |
| "step": 1012 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 5.100012034828305e-06, | |
| "loss": 0.1815, | |
| "step": 1013 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 5.0766701769701495e-06, | |
| "loss": 0.017, | |
| "step": 1014 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 5.053368637089831e-06, | |
| "loss": 0.04, | |
| "step": 1015 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 5.030107536828018e-06, | |
| "loss": 0.0591, | |
| "step": 1016 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 5.006886997614264e-06, | |
| "loss": 0.0238, | |
| "step": 1017 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.983707140666391e-06, | |
| "loss": 0.0466, | |
| "step": 1018 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.960568086989842e-06, | |
| "loss": 0.0286, | |
| "step": 1019 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.937469957377051e-06, | |
| "loss": 0.0755, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.914412872406833e-06, | |
| "loss": 0.0581, | |
| "step": 1021 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.891396952443722e-06, | |
| "loss": 0.0259, | |
| "step": 1022 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.868422317637362e-06, | |
| "loss": 0.0197, | |
| "step": 1023 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.8454890879218865e-06, | |
| "loss": 0.0438, | |
| "step": 1024 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.8225973830152726e-06, | |
| "loss": 0.0394, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.799747322418724e-06, | |
| "loss": 0.0283, | |
| "step": 1026 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.7769390254160655e-06, | |
| "loss": 0.0649, | |
| "step": 1027 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 4.754172611073086e-06, | |
| "loss": 0.0984, | |
| "step": 1028 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.7314481982369395e-06, | |
| "loss": 0.0219, | |
| "step": 1029 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.7087659055355315e-06, | |
| "loss": 0.1168, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.6861258513768684e-06, | |
| "loss": 0.0631, | |
| "step": 1031 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.663528153948483e-06, | |
| "loss": 0.0876, | |
| "step": 1032 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.640972931216771e-06, | |
| "loss": 0.0386, | |
| "step": 1033 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.618460300926414e-06, | |
| "loss": 0.0505, | |
| "step": 1034 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.595990380599747e-06, | |
| "loss": 0.0154, | |
| "step": 1035 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.5735632875361406e-06, | |
| "loss": 0.0281, | |
| "step": 1036 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.551179138811391e-06, | |
| "loss": 0.0484, | |
| "step": 1037 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.528838051277131e-06, | |
| "loss": 0.0486, | |
| "step": 1038 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.506540141560184e-06, | |
| "loss": 0.1264, | |
| "step": 1039 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.4842855260619755e-06, | |
| "loss": 0.0591, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.462074320957936e-06, | |
| "loss": 0.0592, | |
| "step": 1041 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 4.439906642196866e-06, | |
| "loss": 0.0783, | |
| "step": 1042 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 4.417782605500352e-06, | |
| "loss": 0.0647, | |
| "step": 1043 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 4.395702326362159e-06, | |
| "loss": 0.087, | |
| "step": 1044 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 4.3736659200476195e-06, | |
| "loss": 0.0292, | |
| "step": 1045 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 4.351673501593046e-06, | |
| "loss": 0.0331, | |
| "step": 1046 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 4.329725185805112e-06, | |
| "loss": 0.0161, | |
| "step": 1047 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 4.307821087260265e-06, | |
| "loss": 0.0541, | |
| "step": 1048 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 4.285961320304136e-06, | |
| "loss": 0.0248, | |
| "step": 1049 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 4.264145999050914e-06, | |
| "loss": 0.0368, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 4.24237523738279e-06, | |
| "loss": 0.0762, | |
| "step": 1051 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 4.220649148949321e-06, | |
| "loss": 0.0437, | |
| "step": 1052 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 4.1989678471668755e-06, | |
| "loss": 0.0508, | |
| "step": 1053 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 4.1773314452180045e-06, | |
| "loss": 0.0392, | |
| "step": 1054 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 4.155740056050887e-06, | |
| "loss": 0.0748, | |
| "step": 1055 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 4.1341937923787115e-06, | |
| "loss": 0.0908, | |
| "step": 1056 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 4.112692766679095e-06, | |
| "loss": 0.0319, | |
| "step": 1057 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 4.0912370911935156e-06, | |
| "loss": 0.058, | |
| "step": 1058 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 4.069826877926691e-06, | |
| "loss": 0.1191, | |
| "step": 1059 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 4.04846223864603e-06, | |
| "loss": 0.0626, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 4.027143284881023e-06, | |
| "loss": 0.048, | |
| "step": 1061 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 4.005870127922665e-06, | |
| "loss": 0.0341, | |
| "step": 1062 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.984642878822893e-06, | |
| "loss": 0.0436, | |
| "step": 1063 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.96346164839398e-06, | |
| "loss": 0.0626, | |
| "step": 1064 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.942326547207966e-06, | |
| "loss": 0.0331, | |
| "step": 1065 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.921237685596098e-06, | |
| "loss": 0.0467, | |
| "step": 1066 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.900195173648228e-06, | |
| "loss": 0.057, | |
| "step": 1067 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.879199121212245e-06, | |
| "loss": 0.0829, | |
| "step": 1068 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.858249637893523e-06, | |
| "loss": 0.0514, | |
| "step": 1069 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 3.837346833054313e-06, | |
| "loss": 0.0498, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.816490815813205e-06, | |
| "loss": 0.0298, | |
| "step": 1071 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.7956816950445443e-06, | |
| "loss": 0.0269, | |
| "step": 1072 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.7749195793778466e-06, | |
| "loss": 0.0263, | |
| "step": 1073 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.7542045771972722e-06, | |
| "loss": 0.0152, | |
| "step": 1074 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.733536796641016e-06, | |
| "loss": 0.0518, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.7129163456007657e-06, | |
| "loss": 0.076, | |
| "step": 1076 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.6923433317211473e-06, | |
| "loss": 0.0464, | |
| "step": 1077 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.6718178623991405e-06, | |
| "loss": 0.048, | |
| "step": 1078 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.651340044783528e-06, | |
| "loss": 0.0326, | |
| "step": 1079 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.630909985774346e-06, | |
| "loss": 0.023, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.61052779202231e-06, | |
| "loss": 0.05, | |
| "step": 1081 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.5901935699282632e-06, | |
| "loss": 0.072, | |
| "step": 1082 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.569907425642634e-06, | |
| "loss": 0.0651, | |
| "step": 1083 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 3.5496694650648604e-06, | |
| "loss": 0.0126, | |
| "step": 1084 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.5294797938428487e-06, | |
| "loss": 0.036, | |
| "step": 1085 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.5093385173724297e-06, | |
| "loss": 0.0426, | |
| "step": 1086 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.4892457407967864e-06, | |
| "loss": 0.0756, | |
| "step": 1087 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.469201569005935e-06, | |
| "loss": 0.0634, | |
| "step": 1088 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.4492061066361428e-06, | |
| "loss": 0.0119, | |
| "step": 1089 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.4292594580694127e-06, | |
| "loss": 0.0833, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.409361727432929e-06, | |
| "loss": 0.0547, | |
| "step": 1091 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.389513018598495e-06, | |
| "loss": 0.0175, | |
| "step": 1092 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.3697134351820153e-06, | |
| "loss": 0.0295, | |
| "step": 1093 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.3499630805429515e-06, | |
| "loss": 0.1331, | |
| "step": 1094 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.3302620577837672e-06, | |
| "loss": 0.0158, | |
| "step": 1095 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.3106104697493987e-06, | |
| "loss": 0.0495, | |
| "step": 1096 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.291008419026731e-06, | |
| "loss": 0.0531, | |
| "step": 1097 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 3.2714560079440416e-06, | |
| "loss": 0.0147, | |
| "step": 1098 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.251953338570469e-06, | |
| "loss": 0.0062, | |
| "step": 1099 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.232500512715502e-06, | |
| "loss": 0.0785, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.2130976319284173e-06, | |
| "loss": 0.0444, | |
| "step": 1101 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.1937447974977744e-06, | |
| "loss": 0.07, | |
| "step": 1102 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.1744421104508703e-06, | |
| "loss": 0.0577, | |
| "step": 1103 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.1551896715532152e-06, | |
| "loss": 0.1068, | |
| "step": 1104 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.1359875813080234e-06, | |
| "loss": 0.0343, | |
| "step": 1105 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.1168359399556577e-06, | |
| "loss": 0.0227, | |
| "step": 1106 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.0977348474731398e-06, | |
| "loss": 0.0305, | |
| "step": 1107 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.0786844035735964e-06, | |
| "loss": 0.0498, | |
| "step": 1108 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.059684707705769e-06, | |
| "loss": 0.1045, | |
| "step": 1109 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.0407358590534657e-06, | |
| "loss": 0.098, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.0218379565350716e-06, | |
| "loss": 0.0361, | |
| "step": 1111 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 3.0029910988030037e-06, | |
| "loss": 0.0532, | |
| "step": 1112 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 2.9841953842432276e-06, | |
| "loss": 0.0717, | |
| "step": 1113 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.9654509109747124e-06, | |
| "loss": 0.0731, | |
| "step": 1114 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.9467577768489363e-06, | |
| "loss": 0.0827, | |
| "step": 1115 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.9281160794493814e-06, | |
| "loss": 0.0359, | |
| "step": 1116 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.909525916091006e-06, | |
| "loss": 0.0325, | |
| "step": 1117 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.890987383819744e-06, | |
| "loss": 0.0635, | |
| "step": 1118 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.8725005794120147e-06, | |
| "loss": 0.0587, | |
| "step": 1119 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.8540655993741893e-06, | |
| "loss": 0.0551, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.835682539942101e-06, | |
| "loss": 0.0409, | |
| "step": 1121 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.8173514970805554e-06, | |
| "loss": 0.0086, | |
| "step": 1122 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.799072566482806e-06, | |
| "loss": 0.0711, | |
| "step": 1123 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.7808458435700643e-06, | |
| "loss": 0.0425, | |
| "step": 1124 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.7626714234910123e-06, | |
| "loss": 0.0856, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.744549401121286e-06, | |
| "loss": 0.0448, | |
| "step": 1126 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 2.726479871062996e-06, | |
| "loss": 0.0455, | |
| "step": 1127 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.708462927644234e-06, | |
| "loss": 0.0344, | |
| "step": 1128 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.69049866491856e-06, | |
| "loss": 0.0333, | |
| "step": 1129 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.672587176664538e-06, | |
| "loss": 0.0441, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.6547285563852297e-06, | |
| "loss": 0.0614, | |
| "step": 1131 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.636922897307707e-06, | |
| "loss": 0.0313, | |
| "step": 1132 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.6191702923825786e-06, | |
| "loss": 0.0401, | |
| "step": 1133 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.601470834283488e-06, | |
| "loss": 0.0267, | |
| "step": 1134 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.5838246154066366e-06, | |
| "loss": 0.0372, | |
| "step": 1135 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.56623172787031e-06, | |
| "loss": 0.0509, | |
| "step": 1136 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.5486922635143813e-06, | |
| "loss": 0.0571, | |
| "step": 1137 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.531206313899838e-06, | |
| "loss": 0.0394, | |
| "step": 1138 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.513773970308317e-06, | |
| "loss": 0.0192, | |
| "step": 1139 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.496395323741602e-06, | |
| "loss": 0.0204, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 2.4790704649211747e-06, | |
| "loss": 0.0517, | |
| "step": 1141 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.4617994842877237e-06, | |
| "loss": 0.0338, | |
| "step": 1142 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.444582472000674e-06, | |
| "loss": 0.0347, | |
| "step": 1143 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.4274195179377313e-06, | |
| "loss": 0.0737, | |
| "step": 1144 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.410310711694393e-06, | |
| "loss": 0.0678, | |
| "step": 1145 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.3932561425834947e-06, | |
| "loss": 0.0779, | |
| "step": 1146 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.3762558996347337e-06, | |
| "loss": 0.0457, | |
| "step": 1147 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.3593100715942184e-06, | |
| "loss": 0.0397, | |
| "step": 1148 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.342418746923981e-06, | |
| "loss": 0.0268, | |
| "step": 1149 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.325582013801548e-06, | |
| "loss": 0.0534, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.3087999601194495e-06, | |
| "loss": 0.1161, | |
| "step": 1151 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.292072673484776e-06, | |
| "loss": 0.086, | |
| "step": 1152 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.275400241218725e-06, | |
| "loss": 0.0423, | |
| "step": 1153 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.2587827503561284e-06, | |
| "loss": 0.0285, | |
| "step": 1154 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 2.2422202876450197e-06, | |
| "loss": 0.0157, | |
| "step": 1155 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.2257129395461602e-06, | |
| "loss": 0.0427, | |
| "step": 1156 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.2092607922325993e-06, | |
| "loss": 0.0449, | |
| "step": 1157 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.1928639315892295e-06, | |
| "loss": 0.0771, | |
| "step": 1158 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.176522443212322e-06, | |
| "loss": 0.0126, | |
| "step": 1159 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.160236412409091e-06, | |
| "loss": 0.0264, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.144005924197255e-06, | |
| "loss": 0.0897, | |
| "step": 1161 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.1278310633045735e-06, | |
| "loss": 0.0409, | |
| "step": 1162 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.1117119141684154e-06, | |
| "loss": 0.0656, | |
| "step": 1163 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.0956485609353292e-06, | |
| "loss": 0.0243, | |
| "step": 1164 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.079641087460587e-06, | |
| "loss": 0.0147, | |
| "step": 1165 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.063689577307747e-06, | |
| "loss": 0.0207, | |
| "step": 1166 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.047794113748233e-06, | |
| "loss": 0.1049, | |
| "step": 1167 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.031954779760879e-06, | |
| "loss": 0.0361, | |
| "step": 1168 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 2.01617165803152e-06, | |
| "loss": 0.0305, | |
| "step": 1169 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.0004448309525357e-06, | |
| "loss": 0.0218, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.9847743806224314e-06, | |
| "loss": 0.0428, | |
| "step": 1171 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.9691603888454194e-06, | |
| "loss": 0.0277, | |
| "step": 1172 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.9536029371309765e-06, | |
| "loss": 0.0431, | |
| "step": 1173 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.9381021066934196e-06, | |
| "loss": 0.0233, | |
| "step": 1174 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.9226579784514966e-06, | |
| "loss": 0.0528, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.9072706330279458e-06, | |
| "loss": 0.0046, | |
| "step": 1176 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.8919401507490846e-06, | |
| "loss": 0.0424, | |
| "step": 1177 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.8766666116443942e-06, | |
| "loss": 0.0348, | |
| "step": 1178 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.8614500954460893e-06, | |
| "loss": 0.0777, | |
| "step": 1179 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.846290681588708e-06, | |
| "loss": 0.0421, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.8311884492087073e-06, | |
| "loss": 0.0689, | |
| "step": 1181 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.8161434771440298e-06, | |
| "loss": 0.0336, | |
| "step": 1182 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.801155843933708e-06, | |
| "loss": 0.0484, | |
| "step": 1183 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.7862256278174555e-06, | |
| "loss": 0.0815, | |
| "step": 1184 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.7713529067352368e-06, | |
| "loss": 0.0246, | |
| "step": 1185 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.756537758326894e-06, | |
| "loss": 0.0493, | |
| "step": 1186 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.7417802599317084e-06, | |
| "loss": 0.027, | |
| "step": 1187 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.727080488588018e-06, | |
| "loss": 0.0794, | |
| "step": 1188 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.7124385210328128e-06, | |
| "loss": 0.0401, | |
| "step": 1189 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.697854433701326e-06, | |
| "loss": 0.0148, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.6833283027266361e-06, | |
| "loss": 0.0288, | |
| "step": 1191 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.6688602039392833e-06, | |
| "loss": 0.0292, | |
| "step": 1192 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.654450212866854e-06, | |
| "loss": 0.0698, | |
| "step": 1193 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.640098404733598e-06, | |
| "loss": 0.0214, | |
| "step": 1194 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.6258048544600399e-06, | |
| "loss": 0.0209, | |
| "step": 1195 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.6115696366625698e-06, | |
| "loss": 0.0693, | |
| "step": 1196 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.5973928256530781e-06, | |
| "loss": 0.0244, | |
| "step": 1197 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.5832744954385488e-06, | |
| "loss": 0.0746, | |
| "step": 1198 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.569214719720675e-06, | |
| "loss": 0.0404, | |
| "step": 1199 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.55521357189549e-06, | |
| "loss": 0.0265, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.5412711250529624e-06, | |
| "loss": 0.0899, | |
| "step": 1201 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.5273874519766352e-06, | |
| "loss": 0.046, | |
| "step": 1202 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.5135626251432213e-06, | |
| "loss": 0.0502, | |
| "step": 1203 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.499796716722257e-06, | |
| "loss": 0.0509, | |
| "step": 1204 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.4860897985756907e-06, | |
| "loss": 0.0244, | |
| "step": 1205 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.472441942257538e-06, | |
| "loss": 0.036, | |
| "step": 1206 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.4588532190134856e-06, | |
| "loss": 0.0714, | |
| "step": 1207 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.445323699780539e-06, | |
| "loss": 0.0426, | |
| "step": 1208 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.4318534551866306e-06, | |
| "loss": 0.0102, | |
| "step": 1209 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.418442555550267e-06, | |
| "loss": 0.0605, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.4050910708801622e-06, | |
| "loss": 0.0779, | |
| "step": 1211 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.3917990708748603e-06, | |
| "loss": 0.0608, | |
| "step": 1212 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.3785666249223779e-06, | |
| "loss": 0.0416, | |
| "step": 1213 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.3653938020998508e-06, | |
| "loss": 0.0329, | |
| "step": 1214 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.3522806711731575e-06, | |
| "loss": 0.0123, | |
| "step": 1215 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.3392273005965676e-06, | |
| "loss": 0.0368, | |
| "step": 1216 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.326233758512393e-06, | |
| "loss": 0.0237, | |
| "step": 1217 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.3133001127506147e-06, | |
| "loss": 0.0257, | |
| "step": 1218 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.30042643082854e-06, | |
| "loss": 0.0255, | |
| "step": 1219 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.2876127799504502e-06, | |
| "loss": 0.0562, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.2748592270072507e-06, | |
| "loss": 0.0501, | |
| "step": 1221 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.2621658385761079e-06, | |
| "loss": 0.064, | |
| "step": 1222 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.249532680920123e-06, | |
| "loss": 0.0293, | |
| "step": 1223 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.2369598199879689e-06, | |
| "loss": 0.0132, | |
| "step": 1224 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.2244473214135595e-06, | |
| "loss": 0.0353, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.2119952505156903e-06, | |
| "loss": 0.02, | |
| "step": 1226 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.1996036722977123e-06, | |
| "loss": 0.0319, | |
| "step": 1227 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.1872726514471907e-06, | |
| "loss": 0.0467, | |
| "step": 1228 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.1750022523355596e-06, | |
| "loss": 0.0506, | |
| "step": 1229 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.1627925390177877e-06, | |
| "loss": 0.0247, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.1506435752320555e-06, | |
| "loss": 0.0525, | |
| "step": 1231 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.1385554243994079e-06, | |
| "loss": 0.0147, | |
| "step": 1232 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.1265281496234252e-06, | |
| "loss": 0.0228, | |
| "step": 1233 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.1145618136899074e-06, | |
| "loss": 0.08, | |
| "step": 1234 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.1026564790665262e-06, | |
| "loss": 0.0125, | |
| "step": 1235 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.090812207902516e-06, | |
| "loss": 0.0231, | |
| "step": 1236 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.0790290620283414e-06, | |
| "loss": 0.021, | |
| "step": 1237 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.067307102955371e-06, | |
| "loss": 0.0171, | |
| "step": 1238 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.0556463918755682e-06, | |
| "loss": 0.0292, | |
| "step": 1239 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.044046989661156e-06, | |
| "loss": 0.0178, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.032508956864315e-06, | |
| "loss": 0.0325, | |
| "step": 1241 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.021032353716859e-06, | |
| "loss": 0.1254, | |
| "step": 1242 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.00961724012992e-06, | |
| "loss": 0.0522, | |
| "step": 1243 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 9.9826367569363e-07, | |
| "loss": 0.0924, | |
| "step": 1244 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 9.869717196768334e-07, | |
| "loss": 0.0869, | |
| "step": 1245 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 9.757414310267464e-07, | |
| "loss": 0.0519, | |
| "step": 1246 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 9.64572868368667e-07, | |
| "loss": 0.0577, | |
| "step": 1247 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 9.534660900056721e-07, | |
| "loss": 0.0242, | |
| "step": 1248 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 9.424211539183005e-07, | |
| "loss": 0.0369, | |
| "step": 1249 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 9.31438117764263e-07, | |
| "loss": 0.0733, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 9.205170388781283e-07, | |
| "loss": 0.0388, | |
| "step": 1251 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 9.096579742710318e-07, | |
| "loss": 0.041, | |
| "step": 1252 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.988609806303827e-07, | |
| "loss": 0.0489, | |
| "step": 1253 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 8.881261143195572e-07, | |
| "loss": 0.0421, | |
| "step": 1254 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 8.774534313776058e-07, | |
| "loss": 0.036, | |
| "step": 1255 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 8.668429875189744e-07, | |
| "loss": 0.0543, | |
| "step": 1256 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 8.562948381331921e-07, | |
| "loss": 0.0468, | |
| "step": 1257 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 8.458090382846009e-07, | |
| "loss": 0.0188, | |
| "step": 1258 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 8.353856427120561e-07, | |
| "loss": 0.014, | |
| "step": 1259 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 8.25024705828649e-07, | |
| "loss": 0.0334, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 8.147262817214127e-07, | |
| "loss": 0.0429, | |
| "step": 1261 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 8.044904241510536e-07, | |
| "loss": 0.0594, | |
| "step": 1262 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 7.943171865516532e-07, | |
| "loss": 0.0073, | |
| "step": 1263 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 7.842066220304119e-07, | |
| "loss": 0.0115, | |
| "step": 1264 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 7.741587833673464e-07, | |
| "loss": 0.0528, | |
| "step": 1265 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 7.641737230150311e-07, | |
| "loss": 0.0364, | |
| "step": 1266 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 7.542514930983245e-07, | |
| "loss": 0.0622, | |
| "step": 1267 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 7.443921454140849e-07, | |
| "loss": 0.0428, | |
| "step": 1268 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 7.345957314309103e-07, | |
| "loss": 0.0416, | |
| "step": 1269 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 7.248623022888708e-07, | |
| "loss": 0.0515, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 7.151919087992311e-07, | |
| "loss": 0.0396, | |
| "step": 1271 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 7.055846014441954e-07, | |
| "loss": 0.029, | |
| "step": 1272 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 6.960404303766424e-07, | |
| "loss": 0.0277, | |
| "step": 1273 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 6.865594454198558e-07, | |
| "loss": 0.0272, | |
| "step": 1274 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 6.771416960672738e-07, | |
| "loss": 0.0429, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 6.677872314822266e-07, | |
| "loss": 0.0199, | |
| "step": 1276 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 6.584961004976819e-07, | |
| "loss": 0.018, | |
| "step": 1277 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 6.492683516159801e-07, | |
| "loss": 0.0232, | |
| "step": 1278 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 6.40104033008599e-07, | |
| "loss": 0.0686, | |
| "step": 1279 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 6.310031925158838e-07, | |
| "loss": 0.0527, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 6.219658776468137e-07, | |
| "loss": 0.057, | |
| "step": 1281 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 6.129921355787402e-07, | |
| "loss": 0.0183, | |
| "step": 1282 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 6.040820131571478e-07, | |
| "loss": 0.0492, | |
| "step": 1283 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 5.952355568954103e-07, | |
| "loss": 0.1228, | |
| "step": 1284 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 5.864528129745451e-07, | |
| "loss": 0.1016, | |
| "step": 1285 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 5.777338272429702e-07, | |
| "loss": 0.028, | |
| "step": 1286 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 5.69078645216274e-07, | |
| "loss": 0.0569, | |
| "step": 1287 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 5.604873120769667e-07, | |
| "loss": 0.0411, | |
| "step": 1288 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 5.519598726742487e-07, | |
| "loss": 0.0112, | |
| "step": 1289 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 5.434963715237814e-07, | |
| "loss": 0.0616, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 5.350968528074431e-07, | |
| "loss": 0.0533, | |
| "step": 1291 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 5.267613603731149e-07, | |
| "loss": 0.0779, | |
| "step": 1292 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 5.184899377344351e-07, | |
| "loss": 0.0792, | |
| "step": 1293 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 5.102826280705799e-07, | |
| "loss": 0.0319, | |
| "step": 1294 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 5.02139474226044e-07, | |
| "loss": 0.0212, | |
| "step": 1295 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.940605187104011e-07, | |
| "loss": 0.0378, | |
| "step": 1296 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 4.860458036981001e-07, | |
| "loss": 0.0953, | |
| "step": 1297 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 4.780953710282318e-07, | |
| "loss": 0.0699, | |
| "step": 1298 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 4.702092622043153e-07, | |
| "loss": 0.082, | |
| "step": 1299 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 4.623875183940782e-07, | |
| "loss": 0.0314, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 4.5463018042925253e-07, | |
| "loss": 0.0215, | |
| "step": 1301 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 4.469372888053442e-07, | |
| "loss": 0.0309, | |
| "step": 1302 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 4.393088836814357e-07, | |
| "loss": 0.01, | |
| "step": 1303 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 4.3174500487997276e-07, | |
| "loss": 0.0229, | |
| "step": 1304 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 4.2424569188655e-07, | |
| "loss": 0.0567, | |
| "step": 1305 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 4.1681098384971873e-07, | |
| "loss": 0.0299, | |
| "step": 1306 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 4.094409195807657e-07, | |
| "loss": 0.0306, | |
| "step": 1307 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 4.021355375535222e-07, | |
| "loss": 0.0191, | |
| "step": 1308 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.948948759041599e-07, | |
| "loss": 0.0185, | |
| "step": 1309 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.8771897243099287e-07, | |
| "loss": 0.0392, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 3.8060786459427614e-07, | |
| "loss": 0.0803, | |
| "step": 1311 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.7356158951601513e-07, | |
| "loss": 0.0704, | |
| "step": 1312 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.6658018397976944e-07, | |
| "loss": 0.0555, | |
| "step": 1313 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.596636844304626e-07, | |
| "loss": 0.0537, | |
| "step": 1314 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.5281212697418453e-07, | |
| "loss": 0.0657, | |
| "step": 1315 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.4602554737801935e-07, | |
| "loss": 0.0503, | |
| "step": 1316 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.393039810698383e-07, | |
| "loss": 0.066, | |
| "step": 1317 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.326474631381316e-07, | |
| "loss": 0.0296, | |
| "step": 1318 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.260560283318114e-07, | |
| "loss": 0.0358, | |
| "step": 1319 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.195297110600478e-07, | |
| "loss": 0.0367, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.130685453920677e-07, | |
| "loss": 0.0396, | |
| "step": 1321 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.0667256505699353e-07, | |
| "loss": 0.0447, | |
| "step": 1322 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 3.003418034436649e-07, | |
| "loss": 0.0342, | |
| "step": 1323 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.940762936004546e-07, | |
| "loss": 0.0891, | |
| "step": 1324 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 2.8787606823510573e-07, | |
| "loss": 0.0227, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.817411597145593e-07, | |
| "loss": 0.0947, | |
| "step": 1326 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.756716000647813e-07, | |
| "loss": 0.0642, | |
| "step": 1327 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.696674209705975e-07, | |
| "loss": 0.0449, | |
| "step": 1328 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.637286537755344e-07, | |
| "loss": 0.0731, | |
| "step": 1329 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.578553294816402e-07, | |
| "loss": 0.0489, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.5204747874934124e-07, | |
| "loss": 0.0286, | |
| "step": 1331 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.4630513189727027e-07, | |
| "loss": 0.0393, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.4062831890210813e-07, | |
| "loss": 0.0357, | |
| "step": 1333 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.3501706939843387e-07, | |
| "loss": 0.0489, | |
| "step": 1334 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.2947141267856637e-07, | |
| "loss": 0.0463, | |
| "step": 1335 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.2399137769240978e-07, | |
| "loss": 0.0752, | |
| "step": 1336 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.1857699304730572e-07, | |
| "loss": 0.0208, | |
| "step": 1337 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.1322828700788108e-07, | |
| "loss": 0.0761, | |
| "step": 1338 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.079452874959017e-07, | |
| "loss": 0.0291, | |
| "step": 1339 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.0272802209013035e-07, | |
| "loss": 0.0548, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.9757651802617175e-07, | |
| "loss": 0.0503, | |
| "step": 1341 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.924908021963432e-07, | |
| "loss": 0.0232, | |
| "step": 1342 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.874709011495298e-07, | |
| "loss": 0.0094, | |
| "step": 1343 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.8251684109103934e-07, | |
| "loss": 0.0294, | |
| "step": 1344 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.7762864788247725e-07, | |
| "loss": 0.0377, | |
| "step": 1345 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.7280634704159893e-07, | |
| "loss": 0.0968, | |
| "step": 1346 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.6804996374218744e-07, | |
| "loss": 0.0431, | |
| "step": 1347 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.6335952281391712e-07, | |
| "loss": 0.0228, | |
| "step": 1348 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.5873504874222134e-07, | |
| "loss": 0.0212, | |
| "step": 1349 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.5417656566817043e-07, | |
| "loss": 0.0241, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.4968409738834366e-07, | |
| "loss": 0.0416, | |
| "step": 1351 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.4525766735470143e-07, | |
| "loss": 0.0318, | |
| "step": 1352 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.408972986744672e-07, | |
| "loss": 0.0393, | |
| "step": 1353 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.3660301411000546e-07, | |
| "loss": 0.079, | |
| "step": 1354 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.3237483607870358e-07, | |
| "loss": 0.024, | |
| "step": 1355 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.282127866528512e-07, | |
| "loss": 0.0244, | |
| "step": 1356 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2411688755953206e-07, | |
| "loss": 0.031, | |
| "step": 1357 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.2008716018050336e-07, | |
| "loss": 0.052, | |
| "step": 1358 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.1612362555208904e-07, | |
| "loss": 0.0791, | |
| "step": 1359 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.1222630436506477e-07, | |
| "loss": 0.0228, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.0839521696455562e-07, | |
| "loss": 0.0426, | |
| "step": 1361 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.0463038334992802e-07, | |
| "loss": 0.0347, | |
| "step": 1362 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.0093182317468461e-07, | |
| "loss": 0.0094, | |
| "step": 1363 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.729955574635909e-08, | |
| "loss": 0.0388, | |
| "step": 1364 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.373360002642243e-08, | |
| "loss": 0.0741, | |
| "step": 1365 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 9.02339746301763e-08, | |
| "loss": 0.0106, | |
| "step": 1366 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 8.680069782666066e-08, | |
| "loss": 0.0254, | |
| "step": 1367 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 8.343378753855717e-08, | |
| "loss": 0.0648, | |
| "step": 1368 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 8.013326134209535e-08, | |
| "loss": 0.0631, | |
| "step": 1369 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 7.689913646695884e-08, | |
| "loss": 0.038, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 7.373142979619871e-08, | |
| "loss": 0.0663, | |
| "step": 1371 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 7.063015786614812e-08, | |
| "loss": 0.0376, | |
| "step": 1372 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 6.759533686632437e-08, | |
| "loss": 0.071, | |
| "step": 1373 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 6.462698263935778e-08, | |
| "loss": 0.0628, | |
| "step": 1374 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 6.172511068090216e-08, | |
| "loss": 0.0365, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.888973613955386e-08, | |
| "loss": 0.0173, | |
| "step": 1376 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.612087381677355e-08, | |
| "loss": 0.0604, | |
| "step": 1377 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.341853816680952e-08, | |
| "loss": 0.0176, | |
| "step": 1378 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 5.078274329662378e-08, | |
| "loss": 0.0493, | |
| "step": 1379 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.821350296581386e-08, | |
| "loss": 0.0208, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.571083058654608e-08, | |
| "loss": 0.0297, | |
| "step": 1381 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 4.3274739223480195e-08, | |
| "loss": 0.0419, | |
| "step": 1382 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 4.09052415937083e-08, | |
| "loss": 0.0767, | |
| "step": 1383 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.8602350066678066e-08, | |
| "loss": 0.0464, | |
| "step": 1384 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.636607666414164e-08, | |
| "loss": 0.0254, | |
| "step": 1385 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.4196433060080265e-08, | |
| "loss": 0.0175, | |
| "step": 1386 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.2093430580651726e-08, | |
| "loss": 0.0403, | |
| "step": 1387 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 3.0057080204126405e-08, | |
| "loss": 0.0497, | |
| "step": 1388 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.8087392560831858e-08, | |
| "loss": 0.0293, | |
| "step": 1389 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.618437793309738e-08, | |
| "loss": 0.0447, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.4348046255201437e-08, | |
| "loss": 0.0121, | |
| "step": 1391 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.2578407113316246e-08, | |
| "loss": 0.0805, | |
| "step": 1392 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 2.0875469745463702e-08, | |
| "loss": 0.0513, | |
| "step": 1393 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.9239243041458563e-08, | |
| "loss": 0.0625, | |
| "step": 1394 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.7669735542871476e-08, | |
| "loss": 0.0601, | |
| "step": 1395 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.6166955442976414e-08, | |
| "loss": 0.0673, | |
| "step": 1396 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.4730910586713718e-08, | |
| "loss": 0.0099, | |
| "step": 1397 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.3361608470644626e-08, | |
| "loss": 0.0854, | |
| "step": 1398 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.2059056242917165e-08, | |
| "loss": 0.1167, | |
| "step": 1399 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.0823260703223525e-08, | |
| "loss": 0.0846, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 9.654228302768785e-09, | |
| "loss": 0.0629, | |
| "step": 1401 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 8.551965144232553e-09, | |
| "loss": 0.0679, | |
| "step": 1402 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 7.51647698174338e-09, | |
| "loss": 0.0445, | |
| "step": 1403 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 6.547769220843235e-09, | |
| "loss": 0.0699, | |
| "step": 1404 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 5.6458469184633485e-09, | |
| "loss": 0.0306, | |
| "step": 1405 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 4.810714782892944e-09, | |
| "loss": 0.0765, | |
| "step": 1406 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 4.042377173760769e-09, | |
| "loss": 0.0323, | |
| "step": 1407 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.34083810200525e-09, | |
| "loss": 0.0075, | |
| "step": 1408 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.7061012298631226e-09, | |
| "loss": 0.0244, | |
| "step": 1409 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.1381698708395903e-09, | |
| "loss": 0.0231, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.6370469896969552e-09, | |
| "loss": 0.0233, | |
| "step": 1411 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.202735202440408e-09, | |
| "loss": 0.0114, | |
| "step": 1412 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 8.352367763009738e-10, | |
| "loss": 0.0485, | |
| "step": 1413 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 5.345536297241438e-10, | |
| "loss": 0.0134, | |
| "step": 1414 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3.006873323627701e-10, | |
| "loss": 0.0224, | |
| "step": 1415 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.33639105064276e-10, | |
| "loss": 0.1117, | |
| "step": 1416 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3.340981986781344e-11, | |
| "loss": 0.0546, | |
| "step": 1417 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 0.0, | |
| "loss": 0.014, | |
| "step": 1418 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "step": 1418, | |
| "total_flos": 43933127933952.0, | |
| "train_loss": 0.08294331620851918, | |
| "train_runtime": 8926.9052, | |
| "train_samples_per_second": 10.164, | |
| "train_steps_per_second": 0.159 | |
| } | |
| ], | |
| "logging_steps": 1.0, | |
| "max_steps": 1418, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 1, | |
| "save_steps": 50000, | |
| "total_flos": 43933127933952.0, | |
| "train_batch_size": 16, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |